]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 98 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
105 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
106 | #define IMSM_DISK_FILLERS 3 | |
107 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
108 | }; |
109 | ||
3b451610 AK |
110 | /* map selector for map managment |
111 | */ | |
238c0a71 AK |
112 | #define MAP_0 0 |
113 | #define MAP_1 1 | |
114 | #define MAP_X -1 | |
3b451610 | 115 | |
cdddbdbc DW |
116 | /* RAID map configuration infos. */ |
117 | struct imsm_map { | |
5551b113 CA |
118 | __u32 pba_of_lba0_lo; /* start address of partition */ |
119 | __u32 blocks_per_member_lo;/* blocks per member */ | |
120 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
121 | __u16 blocks_per_strip; |
122 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
123 | #define IMSM_T_STATE_NORMAL 0 | |
124 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
125 | #define IMSM_T_STATE_DEGRADED 2 |
126 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
127 | __u8 raid_level; |
128 | #define IMSM_T_RAID0 0 | |
129 | #define IMSM_T_RAID1 1 | |
130 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
131 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
132 | __u8 num_domains; /* number of parity domains */ |
133 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 134 | __u8 ddf; |
5551b113 CA |
135 | __u32 pba_of_lba0_hi; |
136 | __u32 blocks_per_member_hi; | |
137 | __u32 num_data_stripes_hi; | |
138 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 139 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 140 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
141 | * top byte contains some flags |
142 | */ | |
cdddbdbc DW |
143 | } __attribute__ ((packed)); |
144 | ||
145 | struct imsm_vol { | |
f8f603f1 | 146 | __u32 curr_migr_unit; |
fe7ed8cb | 147 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 148 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
149 | #define MIGR_INIT 0 |
150 | #define MIGR_REBUILD 1 | |
151 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
152 | #define MIGR_GEN_MIGR 3 | |
153 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 154 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
155 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
156 | __u8 dirty; | |
fe7ed8cb DW |
157 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
158 | __u16 verify_errors; /* number of mismatches */ | |
159 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
160 | __u32 filler[4]; | |
cdddbdbc DW |
161 | struct imsm_map map[1]; |
162 | /* here comes another one if migr_state */ | |
163 | } __attribute__ ((packed)); | |
164 | ||
165 | struct imsm_dev { | |
fe7ed8cb | 166 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
167 | __u32 size_low; |
168 | __u32 size_high; | |
fe7ed8cb DW |
169 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
170 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
171 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
172 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
173 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
174 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
175 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
176 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
177 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
178 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
179 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
180 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
181 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
182 | __u32 status; /* Persistent RaidDev status */ |
183 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
184 | __u8 migr_priority; |
185 | __u8 num_sub_vols; | |
186 | __u8 tid; | |
187 | __u8 cng_master_disk; | |
188 | __u16 cache_policy; | |
189 | __u8 cng_state; | |
190 | __u8 cng_sub_state; | |
191 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
192 | __u32 filler[IMSM_DEV_FILLERS]; |
193 | struct imsm_vol vol; | |
194 | } __attribute__ ((packed)); | |
195 | ||
196 | struct imsm_super { | |
197 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
198 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
199 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
200 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
201 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
202 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
203 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
204 | __u8 num_disks; /* 0x38 Number of configured disks */ |
205 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
206 | __u8 error_log_pos; /* 0x3A */ |
207 | __u8 fill[1]; /* 0x3B */ | |
208 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
209 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
210 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
211 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
212 | #define IMSM_FILLERS 35 | |
213 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
214 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
215 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 216 | /* here comes BBM logs */ |
cdddbdbc DW |
217 | } __attribute__ ((packed)); |
218 | ||
604b746f JD |
219 | #define BBM_LOG_MAX_ENTRIES 254 |
220 | ||
221 | struct bbm_log_entry { | |
222 | __u64 defective_block_start; | |
223 | #define UNREADABLE 0xFFFFFFFF | |
224 | __u32 spare_block_offset; | |
225 | __u16 remapped_marked_count; | |
226 | __u16 disk_ordinal; | |
227 | } __attribute__ ((__packed__)); | |
228 | ||
229 | struct bbm_log { | |
230 | __u32 signature; /* 0xABADB10C */ | |
231 | __u32 entry_count; | |
232 | __u32 reserved_spare_block_count; /* 0 */ | |
233 | __u32 reserved; /* 0xFFFF */ | |
234 | __u64 first_spare_lba; | |
235 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
236 | } __attribute__ ((__packed__)); | |
237 | ||
cdddbdbc DW |
238 | #ifndef MDASSEMBLE |
239 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
240 | #endif | |
241 | ||
8e59f3d8 AK |
242 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
243 | ||
244 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
245 | ||
17a4eaf9 AK |
246 | #define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */ |
247 | #define MIGR_REC_POSITION 512 /* migr_record position offset on disk, | |
248 | * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION | |
249 | */ | |
250 | ||
8e59f3d8 AK |
251 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
252 | * be recovered using srcMap */ | |
253 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
254 | * already been migrated and must | |
255 | * be recovered from checkpoint area */ | |
256 | struct migr_record { | |
257 | __u32 rec_status; /* Status used to determine how to restart | |
258 | * migration in case it aborts | |
259 | * in some fashion */ | |
260 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
261 | __u32 family_num; /* Family number of MPB | |
262 | * containing the RaidDev | |
263 | * that is migrating */ | |
264 | __u32 ascending_migr; /* True if migrating in increasing | |
265 | * order of lbas */ | |
266 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
267 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
268 | * member disk | |
269 | * advances per unit-of-operation */ | |
270 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
271 | __u32 dest_1st_member_lba; /* First member lba on first | |
272 | * stripe of destination */ | |
273 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
274 | __u32 post_migr_vol_cap; /* Size of volume after | |
275 | * migration completes */ | |
276 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
277 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
278 | * migration ckpt record was read from | |
279 | * (for recovered migrations) */ | |
280 | } __attribute__ ((__packed__)); | |
281 | ||
ec50f7b6 LM |
282 | struct md_list { |
283 | /* usage marker: | |
284 | * 1: load metadata | |
285 | * 2: metadata does not match | |
286 | * 4: already checked | |
287 | */ | |
288 | int used; | |
289 | char *devname; | |
290 | int found; | |
291 | int container; | |
292 | dev_t st_rdev; | |
293 | struct md_list *next; | |
294 | }; | |
295 | ||
e7b84f9d | 296 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 297 | |
1484e727 DW |
298 | static __u8 migr_type(struct imsm_dev *dev) |
299 | { | |
300 | if (dev->vol.migr_type == MIGR_VERIFY && | |
301 | dev->status & DEV_VERIFY_AND_FIX) | |
302 | return MIGR_REPAIR; | |
303 | else | |
304 | return dev->vol.migr_type; | |
305 | } | |
306 | ||
307 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
308 | { | |
309 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
310 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
311 | */ | |
312 | if (migr_type == MIGR_REPAIR) { | |
313 | dev->vol.migr_type = MIGR_VERIFY; | |
314 | dev->status |= DEV_VERIFY_AND_FIX; | |
315 | } else { | |
316 | dev->vol.migr_type = migr_type; | |
317 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
318 | } | |
319 | } | |
320 | ||
87eb16df | 321 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 322 | { |
654a3817 | 323 | return ROUND_UP(bytes, 512) / 512; |
87eb16df | 324 | } |
cdddbdbc | 325 | |
87eb16df DW |
326 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
327 | { | |
328 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
329 | } |
330 | ||
ba2de7ba DW |
331 | struct intel_dev { |
332 | struct imsm_dev *dev; | |
333 | struct intel_dev *next; | |
f21e18ca | 334 | unsigned index; |
ba2de7ba DW |
335 | }; |
336 | ||
88654014 LM |
337 | struct intel_hba { |
338 | enum sys_dev_type type; | |
339 | char *path; | |
340 | char *pci_id; | |
341 | struct intel_hba *next; | |
342 | }; | |
343 | ||
1a64be56 LM |
344 | enum action { |
345 | DISK_REMOVE = 1, | |
346 | DISK_ADD | |
347 | }; | |
cdddbdbc DW |
348 | /* internal representation of IMSM metadata */ |
349 | struct intel_super { | |
350 | union { | |
949c47a0 DW |
351 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
352 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 353 | }; |
8e59f3d8 AK |
354 | union { |
355 | void *migr_rec_buf; /* buffer for I/O operations */ | |
356 | struct migr_record *migr_rec; /* migration record */ | |
357 | }; | |
51d83f5d AK |
358 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
359 | array, it indicates that mdmon is allowed to clean migration | |
360 | record */ | |
949c47a0 | 361 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
362 | void *next_buf; /* for realloc'ing buf from the manager */ |
363 | size_t next_len; | |
c2c087e6 | 364 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 365 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 366 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 367 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 368 | struct intel_dev *devlist; |
cdddbdbc DW |
369 | struct dl { |
370 | struct dl *next; | |
371 | int index; | |
372 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
373 | int major, minor; | |
374 | char *devname; | |
b9f594fe | 375 | struct imsm_disk disk; |
cdddbdbc | 376 | int fd; |
0dcecb2e DW |
377 | int extent_cnt; |
378 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 379 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 380 | enum action action; |
ca0748fa | 381 | } *disks, *current_disk; |
1a64be56 LM |
382 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
383 | active */ | |
47ee5a45 | 384 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 385 | struct bbm_log *bbm_log; |
88654014 | 386 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 387 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
388 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
389 | }; | |
390 | ||
391 | struct intel_disk { | |
392 | struct imsm_disk disk; | |
393 | #define IMSM_UNKNOWN_OWNER (-1) | |
394 | int owner; | |
395 | struct intel_disk *next; | |
cdddbdbc DW |
396 | }; |
397 | ||
c2c087e6 DW |
398 | struct extent { |
399 | unsigned long long start, size; | |
400 | }; | |
401 | ||
694575e7 KW |
402 | /* definitions of reshape process types */ |
403 | enum imsm_reshape_type { | |
404 | CH_TAKEOVER, | |
b5347799 | 405 | CH_MIGRATION, |
7abc9871 | 406 | CH_ARRAY_SIZE, |
694575e7 KW |
407 | }; |
408 | ||
88758e9d DW |
409 | /* definition of messages passed to imsm_process_update */ |
410 | enum imsm_update_type { | |
411 | update_activate_spare, | |
8273f55e | 412 | update_create_array, |
33414a01 | 413 | update_kill_array, |
aa534678 | 414 | update_rename_array, |
1a64be56 | 415 | update_add_remove_disk, |
78b10e66 | 416 | update_reshape_container_disks, |
48c5303a | 417 | update_reshape_migration, |
2d40f3a1 AK |
418 | update_takeover, |
419 | update_general_migration_checkpoint, | |
f3871fdc | 420 | update_size_change, |
88758e9d DW |
421 | }; |
422 | ||
423 | struct imsm_update_activate_spare { | |
424 | enum imsm_update_type type; | |
d23fe947 | 425 | struct dl *dl; |
88758e9d DW |
426 | int slot; |
427 | int array; | |
428 | struct imsm_update_activate_spare *next; | |
429 | }; | |
430 | ||
78b10e66 | 431 | struct geo_params { |
4dd2df09 | 432 | char devnm[32]; |
78b10e66 | 433 | char *dev_name; |
d04f65f4 | 434 | unsigned long long size; |
78b10e66 N |
435 | int level; |
436 | int layout; | |
437 | int chunksize; | |
438 | int raid_disks; | |
439 | }; | |
440 | ||
bb025c2f KW |
441 | enum takeover_direction { |
442 | R10_TO_R0, | |
443 | R0_TO_R10 | |
444 | }; | |
445 | struct imsm_update_takeover { | |
446 | enum imsm_update_type type; | |
447 | int subarray; | |
448 | enum takeover_direction direction; | |
449 | }; | |
78b10e66 N |
450 | |
451 | struct imsm_update_reshape { | |
452 | enum imsm_update_type type; | |
453 | int old_raid_disks; | |
454 | int new_raid_disks; | |
48c5303a PC |
455 | |
456 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
457 | }; | |
458 | ||
459 | struct imsm_update_reshape_migration { | |
460 | enum imsm_update_type type; | |
461 | int old_raid_disks; | |
462 | int new_raid_disks; | |
463 | /* fields for array migration changes | |
464 | */ | |
465 | int subdev; | |
466 | int new_level; | |
467 | int new_layout; | |
4bba0439 | 468 | int new_chunksize; |
48c5303a | 469 | |
d195167d | 470 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
471 | }; |
472 | ||
f3871fdc AK |
473 | struct imsm_update_size_change { |
474 | enum imsm_update_type type; | |
475 | int subdev; | |
476 | long long new_size; | |
477 | }; | |
478 | ||
2d40f3a1 AK |
479 | struct imsm_update_general_migration_checkpoint { |
480 | enum imsm_update_type type; | |
481 | __u32 curr_migr_unit; | |
482 | }; | |
483 | ||
54c2c1ea DW |
484 | struct disk_info { |
485 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
486 | }; | |
487 | ||
8273f55e DW |
488 | struct imsm_update_create_array { |
489 | enum imsm_update_type type; | |
8273f55e | 490 | int dev_idx; |
6a3e913e | 491 | struct imsm_dev dev; |
8273f55e DW |
492 | }; |
493 | ||
33414a01 DW |
494 | struct imsm_update_kill_array { |
495 | enum imsm_update_type type; | |
496 | int dev_idx; | |
497 | }; | |
498 | ||
aa534678 DW |
499 | struct imsm_update_rename_array { |
500 | enum imsm_update_type type; | |
501 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
502 | int dev_idx; | |
503 | }; | |
504 | ||
1a64be56 | 505 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
506 | enum imsm_update_type type; |
507 | }; | |
508 | ||
88654014 LM |
509 | static const char *_sys_dev_type[] = { |
510 | [SYS_DEV_UNKNOWN] = "Unknown", | |
511 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 512 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
513 | [SYS_DEV_NVME] = "NVMe", |
514 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
515 | }; |
516 | ||
517 | const char *get_sys_dev_type(enum sys_dev_type type) | |
518 | { | |
519 | if (type >= SYS_DEV_MAX) | |
520 | type = SYS_DEV_UNKNOWN; | |
521 | ||
522 | return _sys_dev_type[type]; | |
523 | } | |
524 | ||
525 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
526 | { | |
503975b9 N |
527 | struct intel_hba *result = xmalloc(sizeof(*result)); |
528 | ||
529 | result->type = device->type; | |
530 | result->path = xstrdup(device->path); | |
531 | result->next = NULL; | |
532 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
533 | result->pci_id++; | |
534 | ||
88654014 LM |
535 | return result; |
536 | } | |
537 | ||
538 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
539 | { | |
594dc1b8 JS |
540 | struct intel_hba *result; |
541 | ||
88654014 LM |
542 | for (result = hba; result; result = result->next) { |
543 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
544 | break; | |
545 | } | |
546 | return result; | |
547 | } | |
548 | ||
b4cf4cba | 549 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
550 | { |
551 | struct intel_hba *hba; | |
552 | ||
553 | /* check if disk attached to Intel HBA */ | |
554 | hba = find_intel_hba(super->hba, device); | |
555 | if (hba != NULL) | |
556 | return 1; | |
557 | /* Check if HBA is already attached to super */ | |
558 | if (super->hba == NULL) { | |
559 | super->hba = alloc_intel_hba(device); | |
560 | return 1; | |
6b781d33 AP |
561 | } |
562 | ||
563 | hba = super->hba; | |
564 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 565 | * Do not support HBA types mixing |
6b781d33 AP |
566 | */ |
567 | if (device->type != hba->type) | |
88654014 | 568 | return 2; |
6b781d33 | 569 | |
60f0f54d PB |
570 | /* Always forbid spanning between VMD domains (seen as different controllers by mdadm) */ |
571 | if (device->type == SYS_DEV_VMD && !path_attached_to_hba(device->path, hba->path)) | |
572 | return 2; | |
573 | ||
6b781d33 AP |
574 | /* Multiple same type HBAs can be used if they share the same OROM */ |
575 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
576 | ||
577 | if (device_orom != super->orom) | |
578 | return 2; | |
579 | ||
580 | while (hba->next) | |
581 | hba = hba->next; | |
582 | ||
583 | hba->next = alloc_intel_hba(device); | |
584 | return 1; | |
88654014 LM |
585 | } |
586 | ||
587 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
588 | { | |
9bc4ae77 | 589 | struct sys_dev *list, *elem; |
88654014 LM |
590 | char *disk_path; |
591 | ||
592 | if ((list = find_intel_devices()) == NULL) | |
593 | return 0; | |
594 | ||
595 | if (fd < 0) | |
596 | disk_path = (char *) devname; | |
597 | else | |
598 | disk_path = diskfd_to_devpath(fd); | |
599 | ||
9bc4ae77 | 600 | if (!disk_path) |
88654014 | 601 | return 0; |
88654014 | 602 | |
9bc4ae77 N |
603 | for (elem = list; elem; elem = elem->next) |
604 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 605 | return elem; |
9bc4ae77 | 606 | |
88654014 LM |
607 | if (disk_path != devname) |
608 | free(disk_path); | |
88654014 LM |
609 | |
610 | return NULL; | |
611 | } | |
612 | ||
d424212e N |
613 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
614 | char *devname); | |
f2f5c343 | 615 | |
cdddbdbc DW |
616 | static struct supertype *match_metadata_desc_imsm(char *arg) |
617 | { | |
618 | struct supertype *st; | |
619 | ||
620 | if (strcmp(arg, "imsm") != 0 && | |
621 | strcmp(arg, "default") != 0 | |
622 | ) | |
623 | return NULL; | |
624 | ||
503975b9 | 625 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
626 | st->ss = &super_imsm; |
627 | st->max_devs = IMSM_MAX_DEVICES; | |
628 | st->minor_version = 0; | |
629 | st->sb = NULL; | |
630 | return st; | |
631 | } | |
632 | ||
0e600426 | 633 | #ifndef MDASSEMBLE |
cdddbdbc DW |
634 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
635 | { | |
636 | return &mpb->sig[MPB_SIG_LEN]; | |
637 | } | |
9e2d750d | 638 | #endif |
cdddbdbc | 639 | |
949c47a0 DW |
640 | /* retrieve a disk directly from the anchor when the anchor is known to be |
641 | * up-to-date, currently only at load time | |
642 | */ | |
643 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 644 | { |
949c47a0 | 645 | if (index >= mpb->num_disks) |
cdddbdbc DW |
646 | return NULL; |
647 | return &mpb->disk[index]; | |
648 | } | |
649 | ||
95d07a2c LM |
650 | /* retrieve the disk description based on a index of the disk |
651 | * in the sub-array | |
652 | */ | |
653 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 654 | { |
b9f594fe DW |
655 | struct dl *d; |
656 | ||
657 | for (d = super->disks; d; d = d->next) | |
658 | if (d->index == index) | |
95d07a2c LM |
659 | return d; |
660 | ||
661 | return NULL; | |
662 | } | |
663 | /* retrieve a disk from the parsed metadata */ | |
664 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
665 | { | |
666 | struct dl *dl; | |
667 | ||
668 | dl = get_imsm_dl_disk(super, index); | |
669 | if (dl) | |
670 | return &dl->disk; | |
671 | ||
b9f594fe | 672 | return NULL; |
949c47a0 DW |
673 | } |
674 | ||
675 | /* generate a checksum directly from the anchor when the anchor is known to be | |
676 | * up-to-date, currently only at load or write_super after coalescing | |
677 | */ | |
678 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
679 | { |
680 | __u32 end = mpb->mpb_size / sizeof(end); | |
681 | __u32 *p = (__u32 *) mpb; | |
682 | __u32 sum = 0; | |
683 | ||
5d500228 N |
684 | while (end--) { |
685 | sum += __le32_to_cpu(*p); | |
97f734fd N |
686 | p++; |
687 | } | |
cdddbdbc | 688 | |
5d500228 | 689 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
690 | } |
691 | ||
a965f303 DW |
692 | static size_t sizeof_imsm_map(struct imsm_map *map) |
693 | { | |
694 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
695 | } | |
696 | ||
697 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 698 | { |
5e7b0330 AK |
699 | /* A device can have 2 maps if it is in the middle of a migration. |
700 | * If second_map is: | |
238c0a71 AK |
701 | * MAP_0 - we return the first map |
702 | * MAP_1 - we return the second map if it exists, else NULL | |
703 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 704 | */ |
a965f303 | 705 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 706 | struct imsm_map *map2 = NULL; |
a965f303 | 707 | |
9535fc47 AK |
708 | if (dev->vol.migr_state) |
709 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 710 | |
9535fc47 | 711 | switch (second_map) { |
3b451610 | 712 | case MAP_0: |
9535fc47 | 713 | break; |
3b451610 | 714 | case MAP_1: |
9535fc47 AK |
715 | map = map2; |
716 | break; | |
238c0a71 | 717 | case MAP_X: |
9535fc47 AK |
718 | if (map2) |
719 | map = map2; | |
720 | break; | |
9535fc47 AK |
721 | default: |
722 | map = NULL; | |
723 | } | |
724 | return map; | |
5e7b0330 | 725 | |
a965f303 | 726 | } |
cdddbdbc | 727 | |
3393c6af DW |
728 | /* return the size of the device. |
729 | * migr_state increases the returned size if map[0] were to be duplicated | |
730 | */ | |
731 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
732 | { |
733 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 734 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
735 | |
736 | /* migrating means an additional map */ | |
a965f303 | 737 | if (dev->vol.migr_state) |
238c0a71 | 738 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 739 | else if (migr_state) |
238c0a71 | 740 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
741 | |
742 | return size; | |
743 | } | |
744 | ||
54c2c1ea DW |
745 | #ifndef MDASSEMBLE |
746 | /* retrieve disk serial number list from a metadata update */ | |
747 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
748 | { | |
749 | void *u = update; | |
750 | struct disk_info *inf; | |
751 | ||
752 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
753 | sizeof_imsm_dev(&update->dev, 0); | |
754 | ||
755 | return inf; | |
756 | } | |
757 | #endif | |
758 | ||
949c47a0 | 759 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
760 | { |
761 | int offset; | |
762 | int i; | |
763 | void *_mpb = mpb; | |
764 | ||
949c47a0 | 765 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
766 | return NULL; |
767 | ||
768 | /* devices start after all disks */ | |
769 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
770 | ||
771 | for (i = 0; i <= index; i++) | |
772 | if (i == index) | |
773 | return _mpb + offset; | |
774 | else | |
3393c6af | 775 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
776 | |
777 | return NULL; | |
778 | } | |
779 | ||
949c47a0 DW |
780 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
781 | { | |
ba2de7ba DW |
782 | struct intel_dev *dv; |
783 | ||
949c47a0 DW |
784 | if (index >= super->anchor->num_raid_devs) |
785 | return NULL; | |
ba2de7ba DW |
786 | for (dv = super->devlist; dv; dv = dv->next) |
787 | if (dv->index == index) | |
788 | return dv->dev; | |
789 | return NULL; | |
949c47a0 DW |
790 | } |
791 | ||
98130f40 AK |
792 | /* |
793 | * for second_map: | |
238c0a71 AK |
794 | * == MAP_0 get first map |
795 | * == MAP_1 get second map | |
796 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
797 | */ |
798 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
799 | int slot, | |
800 | int second_map) | |
7eef0453 DW |
801 | { |
802 | struct imsm_map *map; | |
803 | ||
5e7b0330 | 804 | map = get_imsm_map(dev, second_map); |
7eef0453 | 805 | |
ff077194 DW |
806 | /* top byte identifies disk under rebuild */ |
807 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
808 | } | |
809 | ||
810 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 811 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 812 | { |
98130f40 | 813 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
814 | |
815 | return ord_to_idx(ord); | |
7eef0453 DW |
816 | } |
817 | ||
be73972f DW |
818 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
819 | { | |
820 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
821 | } | |
822 | ||
f21e18ca | 823 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
824 | { |
825 | int slot; | |
826 | __u32 ord; | |
827 | ||
828 | for (slot = 0; slot < map->num_members; slot++) { | |
829 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
830 | if (ord_to_idx(ord) == idx) | |
831 | return slot; | |
832 | } | |
833 | ||
834 | return -1; | |
835 | } | |
836 | ||
cdddbdbc DW |
837 | static int get_imsm_raid_level(struct imsm_map *map) |
838 | { | |
839 | if (map->raid_level == 1) { | |
840 | if (map->num_members == 2) | |
841 | return 1; | |
842 | else | |
843 | return 10; | |
844 | } | |
845 | ||
846 | return map->raid_level; | |
847 | } | |
848 | ||
c2c087e6 DW |
849 | static int cmp_extent(const void *av, const void *bv) |
850 | { | |
851 | const struct extent *a = av; | |
852 | const struct extent *b = bv; | |
853 | if (a->start < b->start) | |
854 | return -1; | |
855 | if (a->start > b->start) | |
856 | return 1; | |
857 | return 0; | |
858 | } | |
859 | ||
0dcecb2e | 860 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 861 | { |
c2c087e6 | 862 | int memberships = 0; |
620b1713 | 863 | int i; |
c2c087e6 | 864 | |
949c47a0 DW |
865 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
866 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 867 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 868 | |
620b1713 DW |
869 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
870 | memberships++; | |
c2c087e6 | 871 | } |
0dcecb2e DW |
872 | |
873 | return memberships; | |
874 | } | |
875 | ||
b81221b7 CA |
876 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
877 | ||
5551b113 CA |
878 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
879 | { | |
880 | if (lo == 0 || hi == 0) | |
881 | return 1; | |
882 | *lo = __le32_to_cpu((unsigned)n); | |
883 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
884 | return 0; | |
885 | } | |
886 | ||
887 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
888 | { | |
889 | return (unsigned long long)__le32_to_cpu(lo) | | |
890 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
891 | } | |
892 | ||
893 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
894 | { | |
895 | if (disk == NULL) | |
896 | return 0; | |
897 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
898 | } | |
899 | ||
900 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
901 | { | |
902 | if (map == NULL) | |
903 | return 0; | |
904 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
905 | } | |
906 | ||
907 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
908 | { | |
909 | if (map == NULL) | |
910 | return 0; | |
911 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
912 | } | |
913 | ||
ae2416e7 | 914 | #ifndef MDASSEMBLE |
5551b113 CA |
915 | static unsigned long long num_data_stripes(struct imsm_map *map) |
916 | { | |
917 | if (map == NULL) | |
918 | return 0; | |
919 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
920 | } | |
921 | ||
922 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
923 | { | |
924 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
925 | } | |
ae2416e7 | 926 | #endif |
5551b113 CA |
927 | |
928 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
929 | { | |
930 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
931 | } | |
932 | ||
933 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
934 | { | |
935 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
936 | } | |
937 | ||
938 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
939 | { | |
940 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
941 | } | |
942 | ||
0dcecb2e DW |
943 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
944 | { | |
945 | /* find a list of used extents on the given physical device */ | |
946 | struct extent *rv, *e; | |
620b1713 | 947 | int i; |
0dcecb2e | 948 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
949 | __u32 reservation; |
950 | ||
951 | /* trim the reserved area for spares, so they can join any array | |
952 | * regardless of whether the OROM has assigned sectors from the | |
953 | * IMSM_RESERVED_SECTORS region | |
954 | */ | |
955 | if (dl->index == -1) | |
b81221b7 | 956 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
957 | else |
958 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 959 | |
503975b9 | 960 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
961 | e = rv; |
962 | ||
949c47a0 DW |
963 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
964 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 965 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 966 | |
620b1713 | 967 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
968 | e->start = pba_of_lba0(map); |
969 | e->size = blocks_per_member(map); | |
620b1713 | 970 | e++; |
c2c087e6 DW |
971 | } |
972 | } | |
973 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
974 | ||
1011e834 | 975 | /* determine the start of the metadata |
14e8215b DW |
976 | * when no raid devices are defined use the default |
977 | * ...otherwise allow the metadata to truncate the value | |
978 | * as is the case with older versions of imsm | |
979 | */ | |
980 | if (memberships) { | |
981 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 982 | unsigned long long remainder; |
14e8215b | 983 | |
5551b113 | 984 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
985 | /* round down to 1k block to satisfy precision of the kernel |
986 | * 'size' interface | |
987 | */ | |
988 | remainder &= ~1UL; | |
989 | /* make sure remainder is still sane */ | |
f21e18ca | 990 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 991 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
992 | if (reservation > remainder) |
993 | reservation = remainder; | |
994 | } | |
5551b113 | 995 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
996 | e->size = 0; |
997 | return rv; | |
998 | } | |
999 | ||
14e8215b DW |
1000 | /* try to determine how much space is reserved for metadata from |
1001 | * the last get_extents() entry, otherwise fallback to the | |
1002 | * default | |
1003 | */ | |
1004 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1005 | { | |
1006 | struct extent *e; | |
1007 | int i; | |
1008 | __u32 rv; | |
1009 | ||
1010 | /* for spares just return a minimal reservation which will grow | |
1011 | * once the spare is picked up by an array | |
1012 | */ | |
1013 | if (dl->index == -1) | |
1014 | return MPB_SECTOR_CNT; | |
1015 | ||
1016 | e = get_extents(super, dl); | |
1017 | if (!e) | |
1018 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1019 | ||
1020 | /* scroll to last entry */ | |
1021 | for (i = 0; e[i].size; i++) | |
1022 | continue; | |
1023 | ||
5551b113 | 1024 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1025 | |
1026 | free(e); | |
1027 | ||
1028 | return rv; | |
1029 | } | |
1030 | ||
25ed7e59 DW |
1031 | static int is_spare(struct imsm_disk *disk) |
1032 | { | |
1033 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1034 | } | |
1035 | ||
1036 | static int is_configured(struct imsm_disk *disk) | |
1037 | { | |
1038 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1039 | } | |
1040 | ||
1041 | static int is_failed(struct imsm_disk *disk) | |
1042 | { | |
1043 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1044 | } | |
1045 | ||
b81221b7 CA |
1046 | /* try to determine how much space is reserved for metadata from |
1047 | * the last get_extents() entry on the smallest active disk, | |
1048 | * otherwise fallback to the default | |
1049 | */ | |
1050 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1051 | { | |
1052 | struct extent *e; | |
1053 | int i; | |
5551b113 CA |
1054 | unsigned long long min_active; |
1055 | __u32 remainder; | |
b81221b7 CA |
1056 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1057 | struct dl *dl, *dl_min = NULL; | |
1058 | ||
1059 | if (!super) | |
1060 | return rv; | |
1061 | ||
1062 | min_active = 0; | |
1063 | for (dl = super->disks; dl; dl = dl->next) { | |
1064 | if (dl->index < 0) | |
1065 | continue; | |
5551b113 CA |
1066 | unsigned long long blocks = total_blocks(&dl->disk); |
1067 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1068 | dl_min = dl; |
5551b113 | 1069 | min_active = blocks; |
b81221b7 CA |
1070 | } |
1071 | } | |
1072 | if (!dl_min) | |
1073 | return rv; | |
1074 | ||
1075 | /* find last lba used by subarrays on the smallest active disk */ | |
1076 | e = get_extents(super, dl_min); | |
1077 | if (!e) | |
1078 | return rv; | |
1079 | for (i = 0; e[i].size; i++) | |
1080 | continue; | |
1081 | ||
1082 | remainder = min_active - e[i].start; | |
1083 | free(e); | |
1084 | ||
1085 | /* to give priority to recovery we should not require full | |
1086 | IMSM_RESERVED_SECTORS from the spare */ | |
1087 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1088 | ||
1089 | /* if real reservation is smaller use that value */ | |
1090 | return (remainder < rv) ? remainder : rv; | |
1091 | } | |
1092 | ||
80e7f8c3 AC |
1093 | /* Return minimum size of a spare that can be used in this array*/ |
1094 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1095 | { | |
1096 | struct intel_super *super = st->sb; | |
1097 | struct dl *dl; | |
1098 | struct extent *e; | |
1099 | int i; | |
1100 | unsigned long long rv = 0; | |
1101 | ||
1102 | if (!super) | |
1103 | return rv; | |
1104 | /* find first active disk in array */ | |
1105 | dl = super->disks; | |
1106 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1107 | dl = dl->next; | |
1108 | if (!dl) | |
1109 | return rv; | |
1110 | /* find last lba used by subarrays */ | |
1111 | e = get_extents(super, dl); | |
1112 | if (!e) | |
1113 | return rv; | |
1114 | for (i = 0; e[i].size; i++) | |
1115 | continue; | |
1116 | if (i > 0) | |
1117 | rv = e[i-1].start + e[i-1].size; | |
1118 | free(e); | |
b81221b7 | 1119 | |
80e7f8c3 | 1120 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1121 | rv = rv + imsm_min_reserved_sectors(super); |
1122 | ||
80e7f8c3 AC |
1123 | return rv * 512; |
1124 | } | |
1125 | ||
d1e02575 AK |
1126 | static int is_gen_migration(struct imsm_dev *dev); |
1127 | ||
1799c9e8 | 1128 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1129 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1130 | struct imsm_dev *dev); | |
1e5c6983 | 1131 | |
c47b0ff6 AK |
1132 | static void print_imsm_dev(struct intel_super *super, |
1133 | struct imsm_dev *dev, | |
1134 | char *uuid, | |
1135 | int disk_idx) | |
cdddbdbc DW |
1136 | { |
1137 | __u64 sz; | |
0d80bb2f | 1138 | int slot, i; |
238c0a71 AK |
1139 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1140 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1141 | __u32 ord; |
cdddbdbc DW |
1142 | |
1143 | printf("\n"); | |
1e7bc0ed | 1144 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1145 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1146 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1147 | if (map2) | |
1148 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1149 | printf("\n"); | |
1150 | printf(" Members : %d", map->num_members); | |
1151 | if (map2) | |
1152 | printf(" <-- %d", map2->num_members); | |
1153 | printf("\n"); | |
0d80bb2f DW |
1154 | printf(" Slots : ["); |
1155 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1156 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1157 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1158 | } | |
dd8bcb3b AK |
1159 | printf("]"); |
1160 | if (map2) { | |
1161 | printf(" <-- ["); | |
1162 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1163 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1164 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1165 | } | |
1166 | printf("]"); | |
1167 | } | |
1168 | printf("\n"); | |
7095bccb AK |
1169 | printf(" Failed disk : "); |
1170 | if (map->failed_disk_num == 0xff) | |
1171 | printf("none"); | |
1172 | else | |
1173 | printf("%i", map->failed_disk_num); | |
1174 | printf("\n"); | |
620b1713 DW |
1175 | slot = get_imsm_disk_slot(map, disk_idx); |
1176 | if (slot >= 0) { | |
238c0a71 | 1177 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1178 | printf(" This Slot : %d%s\n", slot, |
1179 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1180 | } else | |
cdddbdbc DW |
1181 | printf(" This Slot : ?\n"); |
1182 | sz = __le32_to_cpu(dev->size_high); | |
1183 | sz <<= 32; | |
1184 | sz += __le32_to_cpu(dev->size_low); | |
1185 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1186 | human_size(sz * 512)); | |
5551b113 | 1187 | sz = blocks_per_member(map); |
cdddbdbc DW |
1188 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1189 | human_size(sz * 512)); | |
5551b113 CA |
1190 | printf(" Sector Offset : %llu\n", |
1191 | pba_of_lba0(map)); | |
1192 | printf(" Num Stripes : %llu\n", | |
1193 | num_data_stripes(map)); | |
dd8bcb3b | 1194 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1195 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1196 | if (map2) |
1197 | printf(" <-- %u KiB", | |
1198 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1199 | printf("\n"); | |
cdddbdbc | 1200 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1201 | printf(" Migrate State : "); |
1484e727 DW |
1202 | if (dev->vol.migr_state) { |
1203 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1204 | printf("initialize\n"); |
1484e727 | 1205 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1206 | printf("rebuild\n"); |
1484e727 | 1207 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1208 | printf("check\n"); |
1484e727 | 1209 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1210 | printf("general migration\n"); |
1484e727 | 1211 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1212 | printf("state change\n"); |
1484e727 | 1213 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1214 | printf("repair\n"); |
1484e727 | 1215 | else |
8655a7b1 DW |
1216 | printf("<unknown:%d>\n", migr_type(dev)); |
1217 | } else | |
1218 | printf("idle\n"); | |
3393c6af DW |
1219 | printf(" Map State : %s", map_state_str[map->map_state]); |
1220 | if (dev->vol.migr_state) { | |
238c0a71 | 1221 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1222 | |
b10b37b8 | 1223 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1224 | printf("\n Checkpoint : %u ", |
1225 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1226 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1227 | printf("(N/A)"); |
1228 | else | |
1229 | printf("(%llu)", (unsigned long long) | |
1230 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1231 | } |
1232 | printf("\n"); | |
cdddbdbc | 1233 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1234 | } |
1235 | ||
0ec1f4e8 | 1236 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1237 | { |
1f24f035 | 1238 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1239 | __u64 sz; |
1240 | ||
0ec1f4e8 | 1241 | if (index < -1 || !disk) |
e9d82038 DW |
1242 | return; |
1243 | ||
cdddbdbc | 1244 | printf("\n"); |
1f24f035 | 1245 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1246 | if (index >= 0) |
1247 | printf(" Disk%02d Serial : %s\n", index, str); | |
1248 | else | |
1249 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1250 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1251 | is_configured(disk) ? " active" : "", | |
1252 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1253 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1254 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1255 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1256 | human_size(sz * 512)); | |
1257 | } | |
1258 | ||
520e69e2 AK |
1259 | void examine_migr_rec_imsm(struct intel_super *super) |
1260 | { | |
1261 | struct migr_record *migr_rec = super->migr_rec; | |
1262 | struct imsm_super *mpb = super->anchor; | |
1263 | int i; | |
1264 | ||
1265 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1266 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1267 | struct imsm_map *map; |
b4ab44d8 | 1268 | int slot = -1; |
3136abe5 | 1269 | |
520e69e2 AK |
1270 | if (is_gen_migration(dev) == 0) |
1271 | continue; | |
1272 | ||
1273 | printf("\nMigration Record Information:"); | |
3136abe5 | 1274 | |
44bfe6df AK |
1275 | /* first map under migration */ |
1276 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1277 | if (map) |
1278 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1279 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1280 | printf(" Empty\n "); |
1281 | printf("Examine one of first two disks in array\n"); | |
1282 | break; | |
1283 | } | |
1284 | printf("\n Status : "); | |
1285 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1286 | printf("Normal\n"); | |
1287 | else | |
1288 | printf("Contains Data\n"); | |
1289 | printf(" Current Unit : %u\n", | |
1290 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1291 | printf(" Family : %u\n", | |
1292 | __le32_to_cpu(migr_rec->family_num)); | |
1293 | printf(" Ascending : %u\n", | |
1294 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1295 | printf(" Blocks Per Unit : %u\n", | |
1296 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1297 | printf(" Dest. Depth Per Unit : %u\n", | |
1298 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1299 | printf(" Checkpoint Area pba : %u\n", | |
1300 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1301 | printf(" First member lba : %u\n", | |
1302 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1303 | printf(" Total Number of Units : %u\n", | |
1304 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1305 | printf(" Size of volume : %u\n", | |
1306 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1307 | printf(" Expansion space for LBA64 : %u\n", | |
1308 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1309 | printf(" Record was read from : %u\n", | |
1310 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1311 | ||
1312 | break; | |
1313 | } | |
1314 | } | |
9e2d750d | 1315 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1316 | /******************************************************************************* |
1317 | * function: imsm_check_attributes | |
1318 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1319 | * are supported by mdadm. |
19482bcc AK |
1320 | * Parameters: |
1321 | * attributes - Attributes read from metadata | |
1322 | * Returns: | |
1011e834 N |
1323 | * 0 - passed attributes contains unsupported features flags |
1324 | * 1 - all features are supported | |
19482bcc AK |
1325 | ******************************************************************************/ |
1326 | static int imsm_check_attributes(__u32 attributes) | |
1327 | { | |
1328 | int ret_val = 1; | |
418f9b36 N |
1329 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1330 | ||
1331 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1332 | |
1333 | not_supported &= attributes; | |
1334 | if (not_supported) { | |
e7b84f9d | 1335 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1336 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1337 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1338 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1339 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1340 | } | |
1341 | if (not_supported & MPB_ATTRIB_2TB) { | |
1342 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1343 | not_supported ^= MPB_ATTRIB_2TB; | |
1344 | } | |
1345 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1346 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1347 | not_supported ^= MPB_ATTRIB_RAID0; | |
1348 | } | |
1349 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1350 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1351 | not_supported ^= MPB_ATTRIB_RAID1; | |
1352 | } | |
1353 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1354 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1355 | not_supported ^= MPB_ATTRIB_RAID10; | |
1356 | } | |
1357 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1358 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1359 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1360 | } | |
1361 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1362 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1363 | not_supported ^= MPB_ATTRIB_RAID5; | |
1364 | } | |
1365 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1366 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1367 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1368 | } | |
1369 | if (not_supported & MPB_ATTRIB_BBM) { | |
1370 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1371 | not_supported ^= MPB_ATTRIB_BBM; | |
1372 | } | |
1373 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1374 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1375 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1376 | } | |
1377 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1378 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1379 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1380 | } | |
1381 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1382 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1383 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1384 | } | |
1385 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1386 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1387 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1388 | } | |
1389 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1390 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1391 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1392 | } | |
1393 | ||
1394 | if (not_supported) | |
1ade5cc1 | 1395 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1396 | |
1397 | ret_val = 0; | |
1398 | } | |
1399 | ||
1400 | return ret_val; | |
1401 | } | |
1402 | ||
9e2d750d | 1403 | #ifndef MDASSEMBLE |
a5d85af7 | 1404 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1405 | |
cdddbdbc DW |
1406 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1407 | { | |
1408 | struct intel_super *super = st->sb; | |
949c47a0 | 1409 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1410 | char str[MAX_SIGNATURE_LENGTH]; |
1411 | int i; | |
27fd6274 DW |
1412 | struct mdinfo info; |
1413 | char nbuf[64]; | |
cdddbdbc | 1414 | __u32 sum; |
14e8215b | 1415 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1416 | struct dl *dl; |
27fd6274 | 1417 | |
cdddbdbc DW |
1418 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1419 | printf(" Magic : %s\n", str); | |
1420 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1421 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1422 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1423 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1424 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1425 | printf(" Attributes : "); |
1426 | if (imsm_check_attributes(mpb->attributes)) | |
1427 | printf("All supported\n"); | |
1428 | else | |
1429 | printf("not supported\n"); | |
a5d85af7 | 1430 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1431 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1432 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1433 | sum = __le32_to_cpu(mpb->check_sum); |
1434 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1435 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1436 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1437 | printf(" Disks : %d\n", mpb->num_disks); |
1438 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1439 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1440 | if (super->bbm_log) { |
1441 | struct bbm_log *log = super->bbm_log; | |
1442 | ||
1443 | printf("\n"); | |
1444 | printf("Bad Block Management Log:\n"); | |
1445 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1446 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1447 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1448 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1449 | printf(" First Spare : %llx\n", |
1450 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1451 | } |
44470971 DW |
1452 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1453 | struct mdinfo info; | |
1454 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1455 | ||
1456 | super->current_vol = i; | |
a5d85af7 | 1457 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1458 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1459 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1460 | } |
cdddbdbc DW |
1461 | for (i = 0; i < mpb->num_disks; i++) { |
1462 | if (i == super->disks->index) | |
1463 | continue; | |
0ec1f4e8 | 1464 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1465 | } |
94827db3 | 1466 | |
0ec1f4e8 DW |
1467 | for (dl = super->disks; dl; dl = dl->next) |
1468 | if (dl->index == -1) | |
1469 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1470 | |
1471 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1472 | } |
1473 | ||
061f2c6a | 1474 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1475 | { |
27fd6274 | 1476 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1477 | struct mdinfo info; |
1478 | char nbuf[64]; | |
1e7bc0ed | 1479 | struct intel_super *super = st->sb; |
1e7bc0ed | 1480 | |
0d5a423f DW |
1481 | if (!super->anchor->num_raid_devs) { |
1482 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1483 | return; |
0d5a423f | 1484 | } |
ff54de6e | 1485 | |
a5d85af7 | 1486 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1487 | fname_from_uuid(st, &info, nbuf, ':'); |
1488 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1489 | } | |
1490 | ||
1491 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1492 | { | |
1493 | /* We just write a generic IMSM ARRAY entry */ | |
1494 | struct mdinfo info; | |
1495 | char nbuf[64]; | |
1496 | char nbuf1[64]; | |
1497 | struct intel_super *super = st->sb; | |
1498 | int i; | |
1499 | ||
1500 | if (!super->anchor->num_raid_devs) | |
1501 | return; | |
1502 | ||
a5d85af7 | 1503 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1504 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1505 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1506 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1507 | ||
1508 | super->current_vol = i; | |
a5d85af7 | 1509 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1510 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1511 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1512 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1513 | } |
cdddbdbc DW |
1514 | } |
1515 | ||
9d84c8ea DW |
1516 | static void export_examine_super_imsm(struct supertype *st) |
1517 | { | |
1518 | struct intel_super *super = st->sb; | |
1519 | struct imsm_super *mpb = super->anchor; | |
1520 | struct mdinfo info; | |
1521 | char nbuf[64]; | |
1522 | ||
a5d85af7 | 1523 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1524 | fname_from_uuid(st, &info, nbuf, ':'); |
1525 | printf("MD_METADATA=imsm\n"); | |
1526 | printf("MD_LEVEL=container\n"); | |
1527 | printf("MD_UUID=%s\n", nbuf+5); | |
1528 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1529 | } | |
1530 | ||
74db60b0 N |
1531 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
1532 | { | |
1533 | /* The second last 512byte sector of the device contains | |
1534 | * the "struct imsm_super" metadata. | |
1535 | * This contains mpb_size which is the size in bytes of the | |
1536 | * extended metadata. This is located immediately before | |
1537 | * the imsm_super. | |
1538 | * We want to read all that, plus the last sector which | |
1539 | * may contain a migration record, and write it all | |
1540 | * to the target. | |
1541 | */ | |
1542 | void *buf; | |
1543 | unsigned long long dsize, offset; | |
1544 | int sectors; | |
1545 | struct imsm_super *sb; | |
1546 | int written = 0; | |
1547 | ||
1548 | if (posix_memalign(&buf, 4096, 4096) != 0) | |
1549 | return 1; | |
1550 | ||
1551 | if (!get_dev_size(from, NULL, &dsize)) | |
1552 | goto err; | |
1553 | ||
1554 | if (lseek64(from, dsize-1024, 0) < 0) | |
1555 | goto err; | |
1556 | if (read(from, buf, 512) != 512) | |
1557 | goto err; | |
1558 | sb = buf; | |
1559 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
1560 | goto err; | |
1561 | ||
1562 | sectors = mpb_sectors(sb) + 2; | |
1563 | offset = dsize - sectors * 512; | |
1564 | if (lseek64(from, offset, 0) < 0 || | |
1565 | lseek64(to, offset, 0) < 0) | |
1566 | goto err; | |
1567 | while (written < sectors * 512) { | |
1568 | int n = sectors*512 - written; | |
1569 | if (n > 4096) | |
1570 | n = 4096; | |
1571 | if (read(from, buf, n) != n) | |
1572 | goto err; | |
1573 | if (write(to, buf, n) != n) | |
1574 | goto err; | |
1575 | written += n; | |
1576 | } | |
1577 | free(buf); | |
1578 | return 0; | |
1579 | err: | |
1580 | free(buf); | |
1581 | return 1; | |
1582 | } | |
1583 | ||
cdddbdbc DW |
1584 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1585 | { | |
3ebe00a1 DW |
1586 | struct mdinfo info; |
1587 | char nbuf[64]; | |
1588 | ||
a5d85af7 | 1589 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1590 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1591 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1592 | } |
1593 | ||
1594 | static void brief_detail_super_imsm(struct supertype *st) | |
1595 | { | |
ff54de6e N |
1596 | struct mdinfo info; |
1597 | char nbuf[64]; | |
a5d85af7 | 1598 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1599 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1600 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1601 | } |
d665cc31 DW |
1602 | |
1603 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1604 | static void fd2devname(int fd, char *name); | |
1605 | ||
120dc887 | 1606 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1607 | { |
120dc887 LM |
1608 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1609 | * controller, as well as non-connected ports | |
d665cc31 DW |
1610 | */ |
1611 | int hba_len = strlen(hba_path) + 1; | |
1612 | struct dirent *ent; | |
1613 | DIR *dir; | |
1614 | char *path = NULL; | |
1615 | int err = 0; | |
1616 | unsigned long port_mask = (1 << port_count) - 1; | |
1617 | ||
f21e18ca | 1618 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 1619 | if (verbose > 0) |
e7b84f9d | 1620 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
1621 | return 2; |
1622 | } | |
1623 | ||
1624 | /* scroll through /sys/dev/block looking for devices attached to | |
1625 | * this hba | |
1626 | */ | |
1627 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
1628 | if (!dir) |
1629 | return 1; | |
1630 | ||
1631 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
1632 | int fd; |
1633 | char model[64]; | |
1634 | char vendor[64]; | |
1635 | char buf[1024]; | |
1636 | int major, minor; | |
1637 | char *device; | |
1638 | char *c; | |
1639 | int port; | |
1640 | int type; | |
1641 | ||
1642 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1643 | continue; | |
1644 | path = devt_to_devpath(makedev(major, minor)); | |
1645 | if (!path) | |
1646 | continue; | |
1647 | if (!path_attached_to_hba(path, hba_path)) { | |
1648 | free(path); | |
1649 | path = NULL; | |
1650 | continue; | |
1651 | } | |
1652 | ||
1653 | /* retrieve the scsi device type */ | |
1654 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 1655 | if (verbose > 0) |
e7b84f9d | 1656 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
1657 | err = 2; |
1658 | break; | |
1659 | } | |
1660 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 1661 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 1662 | if (verbose > 0) |
e7b84f9d | 1663 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
1664 | path); |
1665 | err = 2; | |
1666 | free(device); | |
1667 | break; | |
1668 | } | |
1669 | type = strtoul(buf, NULL, 10); | |
1670 | ||
1671 | /* if it's not a disk print the vendor and model */ | |
1672 | if (!(type == 0 || type == 7 || type == 14)) { | |
1673 | vendor[0] = '\0'; | |
1674 | model[0] = '\0'; | |
1675 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 1676 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
1677 | strncpy(vendor, buf, sizeof(vendor)); |
1678 | vendor[sizeof(vendor) - 1] = '\0'; | |
1679 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1680 | while (isspace(*c) || *c == '\0') | |
1681 | *c-- = '\0'; | |
1682 | ||
1683 | } | |
1684 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 1685 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
1686 | strncpy(model, buf, sizeof(model)); |
1687 | model[sizeof(model) - 1] = '\0'; | |
1688 | c = (char *) &model[sizeof(model) - 1]; | |
1689 | while (isspace(*c) || *c == '\0') | |
1690 | *c-- = '\0'; | |
1691 | } | |
1692 | ||
1693 | if (vendor[0] && model[0]) | |
1694 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1695 | else | |
1696 | switch (type) { /* numbers from hald/linux/device.c */ | |
1697 | case 1: sprintf(buf, "tape"); break; | |
1698 | case 2: sprintf(buf, "printer"); break; | |
1699 | case 3: sprintf(buf, "processor"); break; | |
1700 | case 4: | |
1701 | case 5: sprintf(buf, "cdrom"); break; | |
1702 | case 6: sprintf(buf, "scanner"); break; | |
1703 | case 8: sprintf(buf, "media_changer"); break; | |
1704 | case 9: sprintf(buf, "comm"); break; | |
1705 | case 12: sprintf(buf, "raid"); break; | |
1706 | default: sprintf(buf, "unknown"); | |
1707 | } | |
1708 | } else | |
1709 | buf[0] = '\0'; | |
1710 | free(device); | |
1711 | ||
1712 | /* chop device path to 'host%d' and calculate the port number */ | |
1713 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 1714 | if (!c) { |
ba728be7 | 1715 | if (verbose > 0) |
e7b84f9d | 1716 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
1717 | err = 2; |
1718 | break; | |
1719 | } | |
d665cc31 | 1720 | *c = '\0'; |
0858eccf AP |
1721 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
1722 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
1723 | port -= host_base; |
1724 | else { | |
ba728be7 | 1725 | if (verbose > 0) { |
d665cc31 | 1726 | *c = '/'; /* repair the full string */ |
e7b84f9d | 1727 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
1728 | path); |
1729 | } | |
1730 | err = 2; | |
1731 | break; | |
1732 | } | |
1733 | ||
1734 | /* mark this port as used */ | |
1735 | port_mask &= ~(1 << port); | |
1736 | ||
1737 | /* print out the device information */ | |
1738 | if (buf[0]) { | |
1739 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1740 | continue; | |
1741 | } | |
1742 | ||
1743 | fd = dev_open(ent->d_name, O_RDONLY); | |
1744 | if (fd < 0) | |
1745 | printf(" Port%d : - disk info unavailable -\n", port); | |
1746 | else { | |
1747 | fd2devname(fd, buf); | |
1748 | printf(" Port%d : %s", port, buf); | |
1749 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1750 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1751 | else |
664d5325 | 1752 | printf(" ()\n"); |
4dab422a | 1753 | close(fd); |
d665cc31 | 1754 | } |
d665cc31 DW |
1755 | free(path); |
1756 | path = NULL; | |
1757 | } | |
1758 | if (path) | |
1759 | free(path); | |
1760 | if (dir) | |
1761 | closedir(dir); | |
1762 | if (err == 0) { | |
1763 | int i; | |
1764 | ||
1765 | for (i = 0; i < port_count; i++) | |
1766 | if (port_mask & (1 << i)) | |
1767 | printf(" Port%d : - no device attached -\n", i); | |
1768 | } | |
1769 | ||
1770 | return err; | |
1771 | } | |
1772 | ||
b5eece69 | 1773 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
1774 | { |
1775 | struct dirent *ent; | |
1776 | DIR *dir; | |
1777 | char path[292]; | |
1778 | char link[256]; | |
1779 | char *c, *rp; | |
1780 | ||
1781 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 1782 | return 1; |
60f0f54d PB |
1783 | |
1784 | /* scroll through /sys/dev/block looking for devices attached to | |
1785 | * this hba | |
1786 | */ | |
1787 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 1788 | if (!dir) |
b5eece69 | 1789 | return 1; |
b9135011 JS |
1790 | |
1791 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
1792 | int n; |
1793 | ||
1794 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
1795 | * that its target matches 'bus' | |
1796 | */ | |
1797 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
1798 | ent->d_name); | |
1799 | n = readlink(path, link, sizeof(link)); | |
1800 | if (n < 0 || n >= (int)sizeof(link)) | |
1801 | continue; | |
1802 | link[n] = '\0'; | |
1803 | c = strrchr(link, '/'); | |
1804 | if (!c) | |
1805 | continue; | |
1806 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
1807 | continue; | |
1808 | ||
1809 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
1810 | /* if not a intel NVMe - skip it*/ | |
1811 | if (devpath_to_vendor(path) != 0x8086) | |
1812 | continue; | |
1813 | ||
1814 | rp = realpath(path, NULL); | |
1815 | if (!rp) | |
1816 | continue; | |
1817 | ||
1818 | if (path_attached_to_hba(rp, hba->path)) { | |
1819 | printf(" NVMe under VMD : %s\n", rp); | |
1820 | } | |
1821 | free(rp); | |
1822 | } | |
1823 | ||
b9135011 | 1824 | closedir(dir); |
b5eece69 | 1825 | return 0; |
60f0f54d PB |
1826 | } |
1827 | ||
120dc887 LM |
1828 | static void print_found_intel_controllers(struct sys_dev *elem) |
1829 | { | |
1830 | for (; elem; elem = elem->next) { | |
e7b84f9d | 1831 | pr_err("found Intel(R) "); |
120dc887 LM |
1832 | if (elem->type == SYS_DEV_SATA) |
1833 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1834 | else if (elem->type == SYS_DEV_SAS) |
1835 | fprintf(stderr, "SAS "); | |
0858eccf AP |
1836 | else if (elem->type == SYS_DEV_NVME) |
1837 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
1838 | |
1839 | if (elem->type == SYS_DEV_VMD) | |
1840 | fprintf(stderr, "VMD domain"); | |
1841 | else | |
1842 | fprintf(stderr, "RAID controller"); | |
1843 | ||
120dc887 LM |
1844 | if (elem->pci_id) |
1845 | fprintf(stderr, " at %s", elem->pci_id); | |
1846 | fprintf(stderr, ".\n"); | |
1847 | } | |
1848 | fflush(stderr); | |
1849 | } | |
1850 | ||
120dc887 LM |
1851 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1852 | { | |
1853 | struct dirent *ent; | |
1854 | DIR *dir; | |
1855 | int host_base = -1; | |
1856 | ||
1857 | *port_count = 0; | |
1858 | if ((dir = opendir(hba_path)) == NULL) | |
1859 | return -1; | |
1860 | ||
1861 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1862 | int host; | |
1863 | ||
0858eccf AP |
1864 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
1865 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
1866 | continue; |
1867 | if (*port_count == 0) | |
1868 | host_base = host; | |
1869 | else if (host < host_base) | |
1870 | host_base = host; | |
1871 | ||
1872 | if (host + 1 > *port_count + host_base) | |
1873 | *port_count = host + 1 - host_base; | |
1874 | } | |
1875 | closedir(dir); | |
1876 | return host_base; | |
1877 | } | |
1878 | ||
a891a3c2 LM |
1879 | static void print_imsm_capability(const struct imsm_orom *orom) |
1880 | { | |
0858eccf AP |
1881 | printf(" Platform : Intel(R) "); |
1882 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
1883 | printf("Matrix Storage Manager\n"); | |
1884 | else | |
1885 | printf("Rapid Storage Technology%s\n", | |
1886 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
1887 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
1888 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
1889 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
1890 | printf(" RAID Levels :%s%s%s%s%s\n", |
1891 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1892 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1893 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1894 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1895 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1896 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1897 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1898 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1899 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1900 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1901 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1902 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1903 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1904 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1905 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1906 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1907 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1908 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1909 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1910 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1911 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1912 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1913 | printf(" 2TB volumes :%s supported\n", |
1914 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1915 | printf(" 2TB disks :%s supported\n", | |
1916 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 1917 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
1918 | printf(" Max Volumes : %d per array, %d per %s\n", |
1919 | orom->vpa, orom->vphba, | |
1920 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
1921 | return; |
1922 | } | |
1923 | ||
e50cf220 MN |
1924 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
1925 | { | |
1926 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
1927 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
1928 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1929 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
1930 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
1931 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
1932 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
1933 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
1934 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
1935 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
1936 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1937 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
1938 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
1939 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
1940 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
1941 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
1942 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
1943 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
1944 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
1945 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
1946 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
1947 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
1948 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
1949 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
1950 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
1951 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
1952 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
1953 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
1954 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
1955 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
1956 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
1957 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
1958 | } | |
1959 | ||
9eafa1de | 1960 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
1961 | { |
1962 | /* There are two components to imsm platform support, the ahci SATA | |
1963 | * controller and the option-rom. To find the SATA controller we | |
1964 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1965 | * controller with the Intel vendor id is present. This approach | |
1966 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1967 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1968 | * detect platform raid capabilities. The option-rom resides in a | |
1969 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1970 | * platform capabilities. If raid support is disabled in the BIOS the | |
1971 | * option-rom capability structure will not be available. | |
1972 | */ | |
d665cc31 | 1973 | struct sys_dev *list, *hba; |
d665cc31 DW |
1974 | int host_base = 0; |
1975 | int port_count = 0; | |
9eafa1de | 1976 | int result=1; |
d665cc31 | 1977 | |
5615172f | 1978 | if (enumerate_only) { |
a891a3c2 | 1979 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1980 | return 0; |
a891a3c2 LM |
1981 | list = find_intel_devices(); |
1982 | if (!list) | |
1983 | return 2; | |
1984 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
1985 | if (find_imsm_capability(hba)) { |
1986 | result = 0; | |
a891a3c2 LM |
1987 | break; |
1988 | } | |
9eafa1de | 1989 | else |
6b781d33 | 1990 | result = 2; |
a891a3c2 | 1991 | } |
a891a3c2 | 1992 | return result; |
5615172f DW |
1993 | } |
1994 | ||
155cbb4c LM |
1995 | list = find_intel_devices(); |
1996 | if (!list) { | |
ba728be7 | 1997 | if (verbose > 0) |
7a862a02 | 1998 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 1999 | return 2; |
ba728be7 | 2000 | } else if (verbose > 0) |
155cbb4c | 2001 | print_found_intel_controllers(list); |
d665cc31 | 2002 | |
a891a3c2 | 2003 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2004 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2005 | continue; |
0858eccf | 2006 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2007 | char buf[PATH_MAX]; |
e7b84f9d | 2008 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2009 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2010 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2011 | continue; |
2012 | } | |
2013 | result = 0; | |
2014 | } | |
2015 | ||
2016 | if (controller_path && result == 1) { | |
2017 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2018 | controller_path); | |
2019 | return result; | |
2020 | } | |
2021 | ||
5e1d6128 | 2022 | const struct orom_entry *entry; |
0858eccf | 2023 | |
5e1d6128 | 2024 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d PB |
2025 | if (entry->type == SYS_DEV_VMD) { |
2026 | for (hba = list; hba; hba = hba->next) { | |
2027 | if (hba->type == SYS_DEV_VMD) { | |
2028 | char buf[PATH_MAX]; | |
2029 | print_imsm_capability(&entry->orom); | |
2030 | printf(" I/O Controller : %s (%s)\n", | |
2031 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2032 | if (print_vmd_attached_devs(hba)) { |
2033 | if (verbose > 0) | |
2034 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2035 | result |= 2; | |
2036 | } | |
60f0f54d PB |
2037 | printf("\n"); |
2038 | } | |
2039 | } | |
2040 | continue; | |
2041 | } | |
0858eccf | 2042 | |
60f0f54d PB |
2043 | print_imsm_capability(&entry->orom); |
2044 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2045 | for (hba = list; hba; hba = hba->next) { |
2046 | if (hba->type == SYS_DEV_NVME) | |
2047 | printf(" NVMe Device : %s\n", hba->path); | |
2048 | } | |
60f0f54d | 2049 | printf("\n"); |
0858eccf AP |
2050 | continue; |
2051 | } | |
2052 | ||
2053 | struct devid_list *devid; | |
5e1d6128 | 2054 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2055 | hba = device_by_id(devid->devid); |
2056 | if (!hba) | |
2057 | continue; | |
2058 | ||
9eafa1de MN |
2059 | printf(" I/O Controller : %s (%s)\n", |
2060 | hba->path, get_sys_dev_type(hba->type)); | |
2061 | if (hba->type == SYS_DEV_SATA) { | |
2062 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2063 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2064 | if (verbose > 0) | |
7a862a02 | 2065 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2066 | result |= 2; |
2067 | } | |
120dc887 LM |
2068 | } |
2069 | } | |
0858eccf | 2070 | printf("\n"); |
d665cc31 | 2071 | } |
155cbb4c | 2072 | |
120dc887 | 2073 | return result; |
d665cc31 | 2074 | } |
e50cf220 | 2075 | |
9eafa1de | 2076 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2077 | { |
e50cf220 MN |
2078 | struct sys_dev *list, *hba; |
2079 | int result=1; | |
2080 | ||
2081 | list = find_intel_devices(); | |
2082 | if (!list) { | |
2083 | if (verbose > 0) | |
2084 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2085 | result = 2; | |
e50cf220 MN |
2086 | return result; |
2087 | } | |
2088 | ||
2089 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2090 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2091 | continue; | |
60f0f54d PB |
2092 | if (!find_imsm_capability(hba) && verbose > 0) { |
2093 | char buf[PATH_MAX]; | |
2094 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2095 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2096 | } | |
0858eccf | 2097 | else |
e50cf220 | 2098 | result = 0; |
e50cf220 MN |
2099 | } |
2100 | ||
5e1d6128 | 2101 | const struct orom_entry *entry; |
0858eccf | 2102 | |
60f0f54d PB |
2103 | for (entry = orom_entries; entry; entry = entry->next) { |
2104 | if (entry->type == SYS_DEV_VMD) { | |
2105 | for (hba = list; hba; hba = hba->next) | |
2106 | print_imsm_capability_export(&entry->orom); | |
2107 | continue; | |
2108 | } | |
5e1d6128 | 2109 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2110 | } |
0858eccf | 2111 | |
e50cf220 MN |
2112 | return result; |
2113 | } | |
2114 | ||
cdddbdbc DW |
2115 | #endif |
2116 | ||
2117 | static int match_home_imsm(struct supertype *st, char *homehost) | |
2118 | { | |
5115ca67 DW |
2119 | /* the imsm metadata format does not specify any host |
2120 | * identification information. We return -1 since we can never | |
2121 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2122 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2123 | * exclude member disks that do not belong, and we rely on |
2124 | * mdadm.conf to specify the arrays that should be assembled. | |
2125 | * Auto-assembly may still pick up "foreign" arrays. | |
2126 | */ | |
cdddbdbc | 2127 | |
9362c1c8 | 2128 | return -1; |
cdddbdbc DW |
2129 | } |
2130 | ||
2131 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2132 | { | |
51006d85 N |
2133 | /* The uuid returned here is used for: |
2134 | * uuid to put into bitmap file (Create, Grow) | |
2135 | * uuid for backup header when saving critical section (Grow) | |
2136 | * comparing uuids when re-adding a device into an array | |
2137 | * In these cases the uuid required is that of the data-array, | |
2138 | * not the device-set. | |
2139 | * uuid to recognise same set when adding a missing device back | |
2140 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2141 | * |
51006d85 N |
2142 | * For each of these we can make do with a truncated |
2143 | * or hashed uuid rather than the original, as long as | |
2144 | * everyone agrees. | |
2145 | * In each case the uuid required is that of the data-array, | |
2146 | * not the device-set. | |
43dad3d6 | 2147 | */ |
51006d85 N |
2148 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2149 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2150 | * - the orig_family_num of the container |
51006d85 N |
2151 | * - the index number of the volume |
2152 | * - the 'serial' number of the volume. | |
2153 | * Hopefully these are all constant. | |
2154 | */ | |
2155 | struct intel_super *super = st->sb; | |
43dad3d6 | 2156 | |
51006d85 N |
2157 | char buf[20]; |
2158 | struct sha1_ctx ctx; | |
2159 | struct imsm_dev *dev = NULL; | |
148acb7b | 2160 | __u32 family_num; |
51006d85 | 2161 | |
148acb7b DW |
2162 | /* some mdadm versions failed to set ->orig_family_num, in which |
2163 | * case fall back to ->family_num. orig_family_num will be | |
2164 | * fixed up with the first metadata update. | |
2165 | */ | |
2166 | family_num = super->anchor->orig_family_num; | |
2167 | if (family_num == 0) | |
2168 | family_num = super->anchor->family_num; | |
51006d85 | 2169 | sha1_init_ctx(&ctx); |
92bd8f8d | 2170 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2171 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2172 | if (super->current_vol >= 0) |
2173 | dev = get_imsm_dev(super, super->current_vol); | |
2174 | if (dev) { | |
2175 | __u32 vol = super->current_vol; | |
2176 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2177 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2178 | } | |
2179 | sha1_finish_ctx(&ctx, buf); | |
2180 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2181 | } |
2182 | ||
0d481d37 | 2183 | #if 0 |
4f5bc454 DW |
2184 | static void |
2185 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2186 | { |
cdddbdbc DW |
2187 | __u8 *v = get_imsm_version(mpb); |
2188 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2189 | char major[] = { 0, 0, 0 }; | |
2190 | char minor[] = { 0 ,0, 0 }; | |
2191 | char patch[] = { 0, 0, 0 }; | |
2192 | char *ver_parse[] = { major, minor, patch }; | |
2193 | int i, j; | |
2194 | ||
2195 | i = j = 0; | |
2196 | while (*v != '\0' && v < end) { | |
2197 | if (*v != '.' && j < 2) | |
2198 | ver_parse[i][j++] = *v; | |
2199 | else { | |
2200 | i++; | |
2201 | j = 0; | |
2202 | } | |
2203 | v++; | |
2204 | } | |
2205 | ||
4f5bc454 DW |
2206 | *m = strtol(minor, NULL, 0); |
2207 | *p = strtol(patch, NULL, 0); | |
2208 | } | |
0d481d37 | 2209 | #endif |
4f5bc454 | 2210 | |
1e5c6983 DW |
2211 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2212 | { | |
2213 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2214 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2215 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2216 | ||
2217 | switch (get_imsm_raid_level(map)) { | |
2218 | case 5: | |
2219 | case 10: | |
2220 | return chunk; | |
2221 | default: | |
2222 | return 128*1024 >> 9; | |
2223 | } | |
2224 | } | |
2225 | ||
2226 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2227 | { | |
2228 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2229 | * this is different than migr_strip_size_resync(), but it's good | |
2230 | * to be compatible | |
2231 | */ | |
238c0a71 | 2232 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2233 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2234 | ||
2235 | switch (get_imsm_raid_level(map)) { | |
2236 | case 1: | |
2237 | case 10: | |
2238 | if (map->num_members % map->num_domains == 0) | |
2239 | return 128*1024 >> 9; | |
2240 | else | |
2241 | return chunk; | |
2242 | case 5: | |
2243 | return max((__u32) 64*1024 >> 9, chunk); | |
2244 | default: | |
2245 | return 128*1024 >> 9; | |
2246 | } | |
2247 | } | |
2248 | ||
2249 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2250 | { | |
238c0a71 AK |
2251 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2252 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2253 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2254 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2255 | ||
2256 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2257 | } | |
2258 | ||
2259 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2260 | { | |
238c0a71 | 2261 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2262 | int level = get_imsm_raid_level(lo); |
2263 | ||
2264 | if (level == 1 || level == 10) { | |
238c0a71 | 2265 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2266 | |
2267 | return hi->num_domains; | |
2268 | } else | |
2269 | return num_stripes_per_unit_resync(dev); | |
2270 | } | |
2271 | ||
98130f40 | 2272 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2273 | { |
2274 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2275 | * counter-intuitively have the same number of data disks | |
2276 | */ | |
98130f40 | 2277 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2278 | |
2279 | switch (get_imsm_raid_level(map)) { | |
2280 | case 0: | |
36fd8ccc AK |
2281 | return map->num_members; |
2282 | break; | |
1e5c6983 DW |
2283 | case 1: |
2284 | case 10: | |
36fd8ccc | 2285 | return map->num_members/2; |
1e5c6983 DW |
2286 | case 5: |
2287 | return map->num_members - 1; | |
2288 | default: | |
1ade5cc1 | 2289 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2290 | return 0; |
2291 | } | |
2292 | } | |
2293 | ||
2294 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2295 | { | |
238c0a71 | 2296 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2297 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2298 | ||
2299 | switch(get_imsm_raid_level(map)) { | |
2300 | case 1: | |
2301 | case 10: | |
2302 | return chunk * map->num_domains; | |
2303 | case 5: | |
2304 | return chunk * map->num_members; | |
2305 | default: | |
2306 | return chunk; | |
2307 | } | |
2308 | } | |
2309 | ||
2310 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2311 | { | |
238c0a71 | 2312 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2313 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2314 | __u32 strip = block / chunk; | |
2315 | ||
2316 | switch (get_imsm_raid_level(map)) { | |
2317 | case 1: | |
2318 | case 10: { | |
2319 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2320 | __u32 vol_stripe = vol_strip / map->num_members; | |
2321 | ||
2322 | return vol_stripe * chunk + block % chunk; | |
2323 | } case 5: { | |
2324 | __u32 stripe = strip / (map->num_members - 1); | |
2325 | ||
2326 | return stripe * chunk + block % chunk; | |
2327 | } | |
2328 | default: | |
2329 | return 0; | |
2330 | } | |
2331 | } | |
2332 | ||
c47b0ff6 AK |
2333 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2334 | struct imsm_dev *dev) | |
1e5c6983 DW |
2335 | { |
2336 | /* calculate the conversion factor between per member 'blocks' | |
2337 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2338 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2339 | */ | |
2340 | if (!dev->vol.migr_state) | |
2341 | return 0; | |
2342 | ||
2343 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2344 | case MIGR_GEN_MIGR: { |
2345 | struct migr_record *migr_rec = super->migr_rec; | |
2346 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2347 | } | |
1e5c6983 DW |
2348 | case MIGR_VERIFY: |
2349 | case MIGR_REPAIR: | |
2350 | case MIGR_INIT: { | |
238c0a71 | 2351 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2352 | __u32 stripes_per_unit; |
2353 | __u32 blocks_per_unit; | |
2354 | __u32 parity_depth; | |
2355 | __u32 migr_chunk; | |
2356 | __u32 block_map; | |
2357 | __u32 block_rel; | |
2358 | __u32 segment; | |
2359 | __u32 stripe; | |
2360 | __u8 disks; | |
2361 | ||
2362 | /* yes, this is really the translation of migr_units to | |
2363 | * per-member blocks in the 'resync' case | |
2364 | */ | |
2365 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2366 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2367 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2368 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2369 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2370 | segment = blocks_per_unit / stripe; |
2371 | block_rel = blocks_per_unit - segment * stripe; | |
2372 | parity_depth = parity_segment_depth(dev); | |
2373 | block_map = map_migr_block(dev, block_rel); | |
2374 | return block_map + parity_depth * segment; | |
2375 | } | |
2376 | case MIGR_REBUILD: { | |
2377 | __u32 stripes_per_unit; | |
2378 | __u32 migr_chunk; | |
2379 | ||
2380 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2381 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2382 | return migr_chunk * stripes_per_unit; | |
2383 | } | |
1e5c6983 DW |
2384 | case MIGR_STATE_CHANGE: |
2385 | default: | |
2386 | return 0; | |
2387 | } | |
2388 | } | |
2389 | ||
c2c087e6 DW |
2390 | static int imsm_level_to_layout(int level) |
2391 | { | |
2392 | switch (level) { | |
2393 | case 0: | |
2394 | case 1: | |
2395 | return 0; | |
2396 | case 5: | |
2397 | case 6: | |
a380c027 | 2398 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2399 | case 10: |
c92a2527 | 2400 | return 0x102; |
c2c087e6 | 2401 | } |
a18a888e | 2402 | return UnSet; |
c2c087e6 DW |
2403 | } |
2404 | ||
8e59f3d8 AK |
2405 | /******************************************************************************* |
2406 | * Function: read_imsm_migr_rec | |
2407 | * Description: Function reads imsm migration record from last sector of disk | |
2408 | * Parameters: | |
2409 | * fd : disk descriptor | |
2410 | * super : metadata info | |
2411 | * Returns: | |
2412 | * 0 : success, | |
2413 | * -1 : fail | |
2414 | ******************************************************************************/ | |
2415 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2416 | { | |
2417 | int ret_val = -1; | |
2418 | unsigned long long dsize; | |
2419 | ||
2420 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2421 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2422 | pr_err("Cannot seek to anchor block: %s\n", |
2423 | strerror(errno)); | |
8e59f3d8 AK |
2424 | goto out; |
2425 | } | |
17a4eaf9 AK |
2426 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2427 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2428 | pr_err("Cannot read migr record block: %s\n", |
2429 | strerror(errno)); | |
8e59f3d8 AK |
2430 | goto out; |
2431 | } | |
2432 | ret_val = 0; | |
2433 | ||
2434 | out: | |
2435 | return ret_val; | |
2436 | } | |
2437 | ||
3136abe5 AK |
2438 | static struct imsm_dev *imsm_get_device_during_migration( |
2439 | struct intel_super *super) | |
2440 | { | |
2441 | ||
2442 | struct intel_dev *dv; | |
2443 | ||
2444 | for (dv = super->devlist; dv; dv = dv->next) { | |
2445 | if (is_gen_migration(dv->dev)) | |
2446 | return dv->dev; | |
2447 | } | |
2448 | return NULL; | |
2449 | } | |
2450 | ||
8e59f3d8 AK |
2451 | /******************************************************************************* |
2452 | * Function: load_imsm_migr_rec | |
2453 | * Description: Function reads imsm migration record (it is stored at the last | |
2454 | * sector of disk) | |
2455 | * Parameters: | |
2456 | * super : imsm internal array info | |
2457 | * info : general array info | |
2458 | * Returns: | |
2459 | * 0 : success | |
2460 | * -1 : fail | |
4c965cc9 | 2461 | * -2 : no migration in progress |
8e59f3d8 AK |
2462 | ******************************************************************************/ |
2463 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2464 | { | |
2465 | struct mdinfo *sd; | |
594dc1b8 | 2466 | struct dl *dl; |
8e59f3d8 AK |
2467 | char nm[30]; |
2468 | int retval = -1; | |
2469 | int fd = -1; | |
3136abe5 | 2470 | struct imsm_dev *dev; |
594dc1b8 | 2471 | struct imsm_map *map; |
b4ab44d8 | 2472 | int slot = -1; |
3136abe5 AK |
2473 | |
2474 | /* find map under migration */ | |
2475 | dev = imsm_get_device_during_migration(super); | |
2476 | /* nothing to load,no migration in progress? | |
2477 | */ | |
2478 | if (dev == NULL) | |
4c965cc9 | 2479 | return -2; |
44bfe6df | 2480 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2481 | |
2482 | if (info) { | |
2483 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2484 | /* skip spare and failed disks |
2485 | */ | |
2486 | if (sd->disk.raid_disk < 0) | |
2487 | continue; | |
8e59f3d8 | 2488 | /* read only from one of the first two slots */ |
3136abe5 AK |
2489 | if (map) |
2490 | slot = get_imsm_disk_slot(map, | |
2491 | sd->disk.raid_disk); | |
2492 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2493 | continue; |
3136abe5 | 2494 | |
8e59f3d8 AK |
2495 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2496 | fd = dev_open(nm, O_RDONLY); | |
2497 | if (fd >= 0) | |
2498 | break; | |
2499 | } | |
2500 | } | |
2501 | if (fd < 0) { | |
2502 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2503 | /* skip spare and failed disks |
2504 | */ | |
2505 | if (dl->index < 0) | |
2506 | continue; | |
8e59f3d8 | 2507 | /* read only from one of the first two slots */ |
3136abe5 AK |
2508 | if (map) |
2509 | slot = get_imsm_disk_slot(map, dl->index); | |
2510 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2511 | continue; |
2512 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2513 | fd = dev_open(nm, O_RDONLY); | |
2514 | if (fd >= 0) | |
2515 | break; | |
2516 | } | |
2517 | } | |
2518 | if (fd < 0) | |
2519 | goto out; | |
2520 | retval = read_imsm_migr_rec(fd, super); | |
2521 | ||
2522 | out: | |
2523 | if (fd >= 0) | |
2524 | close(fd); | |
2525 | return retval; | |
2526 | } | |
2527 | ||
9e2d750d | 2528 | #ifndef MDASSEMBLE |
c17608ea AK |
2529 | /******************************************************************************* |
2530 | * function: imsm_create_metadata_checkpoint_update | |
2531 | * Description: It creates update for checkpoint change. | |
2532 | * Parameters: | |
2533 | * super : imsm internal array info | |
2534 | * u : pointer to prepared update | |
2535 | * Returns: | |
2536 | * Uptate length. | |
2537 | * If length is equal to 0, input pointer u contains no update | |
2538 | ******************************************************************************/ | |
2539 | static int imsm_create_metadata_checkpoint_update( | |
2540 | struct intel_super *super, | |
2541 | struct imsm_update_general_migration_checkpoint **u) | |
2542 | { | |
2543 | ||
2544 | int update_memory_size = 0; | |
2545 | ||
1ade5cc1 | 2546 | dprintf("(enter)\n"); |
c17608ea AK |
2547 | |
2548 | if (u == NULL) | |
2549 | return 0; | |
2550 | *u = NULL; | |
2551 | ||
2552 | /* size of all update data without anchor */ | |
2553 | update_memory_size = | |
2554 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2555 | ||
503975b9 | 2556 | *u = xcalloc(1, update_memory_size); |
c17608ea | 2557 | if (*u == NULL) { |
1ade5cc1 | 2558 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
2559 | return 0; |
2560 | } | |
2561 | (*u)->type = update_general_migration_checkpoint; | |
2562 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 2563 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
2564 | |
2565 | return update_memory_size; | |
2566 | } | |
2567 | ||
c17608ea AK |
2568 | static void imsm_update_metadata_locally(struct supertype *st, |
2569 | void *buf, int len); | |
2570 | ||
687629c2 AK |
2571 | /******************************************************************************* |
2572 | * Function: write_imsm_migr_rec | |
2573 | * Description: Function writes imsm migration record | |
2574 | * (at the last sector of disk) | |
2575 | * Parameters: | |
2576 | * super : imsm internal array info | |
2577 | * Returns: | |
2578 | * 0 : success | |
2579 | * -1 : if fail | |
2580 | ******************************************************************************/ | |
2581 | static int write_imsm_migr_rec(struct supertype *st) | |
2582 | { | |
2583 | struct intel_super *super = st->sb; | |
2584 | unsigned long long dsize; | |
2585 | char nm[30]; | |
2586 | int fd = -1; | |
2587 | int retval = -1; | |
2588 | struct dl *sd; | |
c17608ea AK |
2589 | int len; |
2590 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 2591 | struct imsm_dev *dev; |
594dc1b8 | 2592 | struct imsm_map *map; |
3136abe5 AK |
2593 | |
2594 | /* find map under migration */ | |
2595 | dev = imsm_get_device_during_migration(super); | |
2596 | /* if no migration, write buffer anyway to clear migr_record | |
2597 | * on disk based on first available device | |
2598 | */ | |
2599 | if (dev == NULL) | |
2600 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2601 | super->current_vol); | |
2602 | ||
44bfe6df | 2603 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2604 | |
2605 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2606 | int slot = -1; |
3136abe5 AK |
2607 | |
2608 | /* skip failed and spare devices */ | |
2609 | if (sd->index < 0) | |
2610 | continue; | |
687629c2 | 2611 | /* write to 2 first slots only */ |
3136abe5 AK |
2612 | if (map) |
2613 | slot = get_imsm_disk_slot(map, sd->index); | |
2614 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2615 | continue; |
3136abe5 | 2616 | |
687629c2 AK |
2617 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2618 | fd = dev_open(nm, O_RDWR); | |
2619 | if (fd < 0) | |
2620 | continue; | |
2621 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2622 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2623 | pr_err("Cannot seek to anchor block: %s\n", |
2624 | strerror(errno)); | |
687629c2 AK |
2625 | goto out; |
2626 | } | |
17a4eaf9 AK |
2627 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2628 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2629 | pr_err("Cannot write migr record block: %s\n", |
2630 | strerror(errno)); | |
687629c2 AK |
2631 | goto out; |
2632 | } | |
2633 | close(fd); | |
2634 | fd = -1; | |
2635 | } | |
c17608ea AK |
2636 | /* update checkpoint information in metadata */ |
2637 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2638 | ||
2639 | if (len <= 0) { | |
2640 | dprintf("imsm: Cannot prepare update\n"); | |
2641 | goto out; | |
2642 | } | |
2643 | /* update metadata locally */ | |
2644 | imsm_update_metadata_locally(st, u, len); | |
2645 | /* and possibly remotely */ | |
2646 | if (st->update_tail) { | |
2647 | append_metadata_update(st, u, len); | |
2648 | /* during reshape we do all work inside metadata handler | |
2649 | * manage_reshape(), so metadata update has to be triggered | |
2650 | * insida it | |
2651 | */ | |
2652 | flush_metadata_updates(st); | |
2653 | st->update_tail = &st->updates; | |
2654 | } else | |
2655 | free(u); | |
687629c2 AK |
2656 | |
2657 | retval = 0; | |
2658 | out: | |
2659 | if (fd >= 0) | |
2660 | close(fd); | |
2661 | return retval; | |
2662 | } | |
9e2d750d | 2663 | #endif /* MDASSEMBLE */ |
687629c2 | 2664 | |
e2962bfc AK |
2665 | /* spare/missing disks activations are not allowe when |
2666 | * array/container performs reshape operation, because | |
2667 | * all arrays in container works on the same disks set | |
2668 | */ | |
2669 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2670 | { | |
2671 | int rv = 0; | |
2672 | struct intel_dev *i_dev; | |
2673 | struct imsm_dev *dev; | |
2674 | ||
2675 | /* check whole container | |
2676 | */ | |
2677 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2678 | dev = i_dev->dev; | |
3ad25638 | 2679 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2680 | /* No repair during any migration in container |
2681 | */ | |
2682 | rv = 1; | |
2683 | break; | |
2684 | } | |
2685 | } | |
2686 | return rv; | |
2687 | } | |
c41e00b2 AK |
2688 | static unsigned long long imsm_component_size_aligment_check(int level, |
2689 | int chunk_size, | |
2690 | unsigned long long component_size) | |
2691 | { | |
2692 | unsigned int component_size_alligment; | |
2693 | ||
2694 | /* check component size aligment | |
2695 | */ | |
2696 | component_size_alligment = component_size % (chunk_size/512); | |
2697 | ||
1ade5cc1 | 2698 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
2699 | level, chunk_size, component_size, |
2700 | component_size_alligment); | |
2701 | ||
2702 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
2703 | dprintf("imsm: reported component size alligned from %llu ", | |
2704 | component_size); | |
2705 | component_size -= component_size_alligment; | |
1ade5cc1 | 2706 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
2707 | component_size, component_size_alligment); |
2708 | } | |
2709 | ||
2710 | return component_size; | |
2711 | } | |
e2962bfc | 2712 | |
a5d85af7 | 2713 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2714 | { |
2715 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2716 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2717 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2718 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2719 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2720 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2721 | struct dl *dl; |
a5d85af7 | 2722 | int map_disks = info->array.raid_disks; |
bf5a934a | 2723 | |
95eeceeb | 2724 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2725 | if (prev_map) |
2726 | map_to_analyse = prev_map; | |
2727 | ||
ca0748fa | 2728 | dl = super->current_disk; |
9894ec0d | 2729 | |
bf5a934a | 2730 | info->container_member = super->current_vol; |
cd0430a1 | 2731 | info->array.raid_disks = map->num_members; |
b335e593 | 2732 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2733 | info->array.layout = imsm_level_to_layout(info->array.level); |
2734 | info->array.md_minor = -1; | |
2735 | info->array.ctime = 0; | |
2736 | info->array.utime = 0; | |
b335e593 AK |
2737 | info->array.chunk_size = |
2738 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2739 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2740 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2741 | info->custom_array_size <<= 32; | |
2742 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2743 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2744 | ||
3f510843 | 2745 | if (is_gen_migration(dev)) { |
3f83228a | 2746 | info->reshape_active = 1; |
b335e593 AK |
2747 | info->new_level = get_imsm_raid_level(map); |
2748 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2749 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2750 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2751 | if (info->delta_disks) { |
2752 | /* this needs to be applied to every array | |
2753 | * in the container. | |
2754 | */ | |
81219e70 | 2755 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2756 | } |
3f83228a N |
2757 | /* We shape information that we give to md might have to be |
2758 | * modify to cope with md's requirement for reshaping arrays. | |
2759 | * For example, when reshaping a RAID0, md requires it to be | |
2760 | * presented as a degraded RAID4. | |
2761 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2762 | * the array as already being RAID5, but the 'before' layout | |
2763 | * is a RAID4-like layout. | |
2764 | */ | |
2765 | switch (info->array.level) { | |
2766 | case 0: | |
2767 | switch(info->new_level) { | |
2768 | case 0: | |
2769 | /* conversion is happening as RAID4 */ | |
2770 | info->array.level = 4; | |
2771 | info->array.raid_disks += 1; | |
2772 | break; | |
2773 | case 5: | |
2774 | /* conversion is happening as RAID5 */ | |
2775 | info->array.level = 5; | |
2776 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2777 | info->delta_disks -= 1; |
2778 | break; | |
2779 | default: | |
2780 | /* FIXME error message */ | |
2781 | info->array.level = UnSet; | |
2782 | break; | |
2783 | } | |
2784 | break; | |
2785 | } | |
b335e593 AK |
2786 | } else { |
2787 | info->new_level = UnSet; | |
2788 | info->new_layout = UnSet; | |
2789 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2790 | info->delta_disks = 0; |
b335e593 | 2791 | } |
ca0748fa | 2792 | |
efb30e7f DW |
2793 | if (dl) { |
2794 | info->disk.major = dl->major; | |
2795 | info->disk.minor = dl->minor; | |
ca0748fa | 2796 | info->disk.number = dl->index; |
656b6b5a N |
2797 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2798 | dl->index); | |
efb30e7f | 2799 | } |
bf5a934a | 2800 | |
5551b113 CA |
2801 | info->data_offset = pba_of_lba0(map_to_analyse); |
2802 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 | 2803 | |
c41e00b2 AK |
2804 | info->component_size = imsm_component_size_aligment_check( |
2805 | info->array.level, | |
2806 | info->array.chunk_size, | |
2807 | info->component_size); | |
139dae11 | 2808 | |
301406c9 | 2809 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2810 | info->recovery_start = MaxSector; |
bf5a934a | 2811 | |
d2e6d5d6 | 2812 | info->reshape_progress = 0; |
b6796ce1 | 2813 | info->resync_start = MaxSector; |
b9172665 AK |
2814 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2815 | dev->vol.dirty) && | |
2816 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2817 | info->resync_start = 0; |
b6796ce1 AK |
2818 | } |
2819 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2820 | switch (migr_type(dev)) { |
2821 | case MIGR_REPAIR: | |
2822 | case MIGR_INIT: { | |
c47b0ff6 AK |
2823 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2824 | dev); | |
1e5c6983 DW |
2825 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2826 | ||
2827 | info->resync_start = blocks_per_unit * units; | |
2828 | break; | |
2829 | } | |
d2e6d5d6 | 2830 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2831 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2832 | dev); | |
2833 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2834 | unsigned long long array_blocks; |
2835 | int used_disks; | |
d2e6d5d6 | 2836 | |
befb629b AK |
2837 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2838 | (units < | |
2839 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2840 | (super->migr_rec->rec_status == | |
2841 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2842 | units++; | |
2843 | ||
d2e6d5d6 | 2844 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2845 | |
7a862a02 | 2846 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
2847 | (unsigned long long)units, |
2848 | (unsigned long long)blocks_per_unit, | |
2849 | info->reshape_progress); | |
75156c46 | 2850 | |
238c0a71 | 2851 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2852 | if (used_disks > 0) { |
5551b113 | 2853 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2854 | used_disks; |
2855 | /* round array size down to closest MB | |
2856 | */ | |
2857 | info->custom_array_size = (array_blocks | |
2858 | >> SECT_PER_MB_SHIFT) | |
2859 | << SECT_PER_MB_SHIFT; | |
2860 | } | |
d2e6d5d6 | 2861 | } |
1e5c6983 DW |
2862 | case MIGR_VERIFY: |
2863 | /* we could emulate the checkpointing of | |
2864 | * 'sync_action=check' migrations, but for now | |
2865 | * we just immediately complete them | |
2866 | */ | |
2867 | case MIGR_REBUILD: | |
2868 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2869 | case MIGR_STATE_CHANGE: |
2870 | /* FIXME handle other migrations */ | |
2871 | default: | |
2872 | /* we are not dirty, so... */ | |
2873 | info->resync_start = MaxSector; | |
2874 | } | |
b6796ce1 | 2875 | } |
301406c9 DW |
2876 | |
2877 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2878 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2879 | |
f35f2525 N |
2880 | info->array.major_version = -1; |
2881 | info->array.minor_version = -2; | |
4dd2df09 | 2882 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 2883 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2884 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2885 | |
2886 | if (dmap) { | |
2887 | int i, j; | |
2888 | for (i=0; i<map_disks; i++) { | |
2889 | dmap[i] = 0; | |
2890 | if (i < info->array.raid_disks) { | |
2891 | struct imsm_disk *dsk; | |
238c0a71 | 2892 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2893 | dsk = get_imsm_disk(super, j); |
2894 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2895 | dmap[i] = 1; | |
2896 | } | |
2897 | } | |
2898 | } | |
81ac8b4d | 2899 | } |
bf5a934a | 2900 | |
3b451610 AK |
2901 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2902 | int failed, int look_in_map); | |
2903 | ||
2904 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2905 | int look_in_map); | |
2906 | ||
b4ab44d8 | 2907 | #ifndef MDASSEMBLE |
3b451610 AK |
2908 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2909 | { | |
2910 | if (is_gen_migration(dev)) { | |
2911 | int failed; | |
2912 | __u8 map_state; | |
2913 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2914 | ||
2915 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2916 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2917 | if (map2->map_state != map_state) { |
2918 | map2->map_state = map_state; | |
2919 | super->updates_pending++; | |
2920 | } | |
2921 | } | |
2922 | } | |
b4ab44d8 | 2923 | #endif |
97b4d0e9 DW |
2924 | |
2925 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2926 | { | |
2927 | struct dl *d; | |
2928 | ||
2929 | for (d = super->missing; d; d = d->next) | |
2930 | if (d->index == index) | |
2931 | return &d->disk; | |
2932 | return NULL; | |
2933 | } | |
2934 | ||
a5d85af7 | 2935 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2936 | { |
2937 | struct intel_super *super = st->sb; | |
4f5bc454 | 2938 | struct imsm_disk *disk; |
a5d85af7 | 2939 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2940 | int max_enough = -1; |
2941 | int i; | |
2942 | struct imsm_super *mpb; | |
4f5bc454 | 2943 | |
bf5a934a | 2944 | if (super->current_vol >= 0) { |
a5d85af7 | 2945 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2946 | return; |
2947 | } | |
95eeceeb | 2948 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2949 | |
2950 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2951 | * spares | |
2952 | */ | |
2953 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2954 | info->array.level = LEVEL_CONTAINER; |
2955 | info->array.layout = 0; | |
2956 | info->array.md_minor = -1; | |
1011e834 | 2957 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2958 | info->array.utime = 0; |
2959 | info->array.chunk_size = 0; | |
2960 | ||
2961 | info->disk.major = 0; | |
2962 | info->disk.minor = 0; | |
cdddbdbc | 2963 | info->disk.raid_disk = -1; |
c2c087e6 | 2964 | info->reshape_active = 0; |
f35f2525 N |
2965 | info->array.major_version = -1; |
2966 | info->array.minor_version = -2; | |
c2c087e6 | 2967 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2968 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2969 | info->disk.number = -1; |
2970 | info->disk.state = 0; | |
c5afc314 | 2971 | info->name[0] = 0; |
921d9e16 | 2972 | info->recovery_start = MaxSector; |
3ad25638 | 2973 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2974 | |
97b4d0e9 | 2975 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2976 | mpb = super->anchor; |
97b4d0e9 | 2977 | |
ab3cb6b3 N |
2978 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2979 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2980 | int failed, enough, j, missing = 0; | |
2981 | struct imsm_map *map; | |
2982 | __u8 state; | |
97b4d0e9 | 2983 | |
3b451610 AK |
2984 | failed = imsm_count_failed(super, dev, MAP_0); |
2985 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2986 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2987 | |
2988 | /* any newly missing disks? | |
2989 | * (catches single-degraded vs double-degraded) | |
2990 | */ | |
2991 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2992 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2993 | __u32 idx = ord_to_idx(ord); |
2994 | ||
2995 | if (!(ord & IMSM_ORD_REBUILD) && | |
2996 | get_imsm_missing(super, idx)) { | |
2997 | missing = 1; | |
2998 | break; | |
2999 | } | |
97b4d0e9 | 3000 | } |
ab3cb6b3 N |
3001 | |
3002 | if (state == IMSM_T_STATE_FAILED) | |
3003 | enough = -1; | |
3004 | else if (state == IMSM_T_STATE_DEGRADED && | |
3005 | (state != map->map_state || missing)) | |
3006 | enough = 0; | |
3007 | else /* we're normal, or already degraded */ | |
3008 | enough = 1; | |
d2bde6d3 AK |
3009 | if (is_gen_migration(dev) && missing) { |
3010 | /* during general migration we need all disks | |
3011 | * that process is running on. | |
3012 | * No new missing disk is allowed. | |
3013 | */ | |
3014 | max_enough = -1; | |
3015 | enough = -1; | |
3016 | /* no more checks necessary | |
3017 | */ | |
3018 | break; | |
3019 | } | |
ab3cb6b3 N |
3020 | /* in the missing/failed disk case check to see |
3021 | * if at least one array is runnable | |
3022 | */ | |
3023 | max_enough = max(max_enough, enough); | |
3024 | } | |
1ade5cc1 | 3025 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3026 | info->container_enough = max_enough; |
97b4d0e9 | 3027 | |
4a04ec6c | 3028 | if (super->disks) { |
14e8215b DW |
3029 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3030 | ||
b9f594fe | 3031 | disk = &super->disks->disk; |
5551b113 | 3032 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3033 | info->component_size = reserved; |
25ed7e59 | 3034 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3035 | /* we don't change info->disk.raid_disk here because |
3036 | * this state will be finalized in mdmon after we have | |
3037 | * found the 'most fresh' version of the metadata | |
3038 | */ | |
25ed7e59 DW |
3039 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
3040 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3041 | } |
a575e2a7 DW |
3042 | |
3043 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3044 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3045 | */ | |
3046 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3047 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3048 | else |
3049 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3050 | |
3051 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3052 | if (map) { | |
3053 | int i; | |
3054 | for (i = 0; i < map_disks; i++) | |
3055 | map[i] = 1; | |
3056 | } | |
3057 | ||
cdddbdbc DW |
3058 | } |
3059 | ||
5c4cd5da AC |
3060 | /* allocates memory and fills disk in mdinfo structure |
3061 | * for each disk in array */ | |
3062 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3063 | { | |
594dc1b8 | 3064 | struct mdinfo *mddev; |
5c4cd5da AC |
3065 | struct intel_super *super = st->sb; |
3066 | struct imsm_disk *disk; | |
3067 | int count = 0; | |
3068 | struct dl *dl; | |
3069 | if (!super || !super->disks) | |
3070 | return NULL; | |
3071 | dl = super->disks; | |
503975b9 | 3072 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3073 | while (dl) { |
3074 | struct mdinfo *tmp; | |
3075 | disk = &dl->disk; | |
503975b9 | 3076 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3077 | if (mddev->devs) |
3078 | tmp->next = mddev->devs; | |
3079 | mddev->devs = tmp; | |
3080 | tmp->disk.number = count++; | |
3081 | tmp->disk.major = dl->major; | |
3082 | tmp->disk.minor = dl->minor; | |
3083 | tmp->disk.state = is_configured(disk) ? | |
3084 | (1 << MD_DISK_ACTIVE) : 0; | |
3085 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3086 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3087 | tmp->disk.raid_disk = -1; | |
3088 | dl = dl->next; | |
3089 | } | |
3090 | return mddev; | |
3091 | } | |
3092 | ||
cdddbdbc DW |
3093 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3094 | char *update, char *devname, int verbose, | |
3095 | int uuid_set, char *homehost) | |
3096 | { | |
f352c545 DW |
3097 | /* For 'assemble' and 'force' we need to return non-zero if any |
3098 | * change was made. For others, the return value is ignored. | |
3099 | * Update options are: | |
3100 | * force-one : This device looks a bit old but needs to be included, | |
3101 | * update age info appropriately. | |
3102 | * assemble: clear any 'faulty' flag to allow this device to | |
3103 | * be assembled. | |
3104 | * force-array: Array is degraded but being forced, mark it clean | |
3105 | * if that will be needed to assemble it. | |
3106 | * | |
3107 | * newdev: not used ???? | |
3108 | * grow: Array has gained a new device - this is currently for | |
3109 | * linear only | |
3110 | * resync: mark as dirty so a resync will happen. | |
3111 | * name: update the name - preserving the homehost | |
6e46bf34 | 3112 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3113 | * |
3114 | * Following are not relevant for this imsm: | |
3115 | * sparc2.2 : update from old dodgey metadata | |
3116 | * super-minor: change the preferred_minor number | |
3117 | * summaries: update redundant counters. | |
f352c545 DW |
3118 | * homehost: update the recorded homehost |
3119 | * _reshape_progress: record new reshape_progress position. | |
3120 | */ | |
6e46bf34 DW |
3121 | int rv = 1; |
3122 | struct intel_super *super = st->sb; | |
3123 | struct imsm_super *mpb; | |
f352c545 | 3124 | |
6e46bf34 DW |
3125 | /* we can only update container info */ |
3126 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3127 | return 1; | |
3128 | ||
3129 | mpb = super->anchor; | |
3130 | ||
81a5b4f5 N |
3131 | if (strcmp(update, "uuid") == 0) { |
3132 | /* We take this to mean that the family_num should be updated. | |
3133 | * However that is much smaller than the uuid so we cannot really | |
3134 | * allow an explicit uuid to be given. And it is hard to reliably | |
3135 | * know if one was. | |
3136 | * So if !uuid_set we know the current uuid is random and just used | |
3137 | * the first 'int' and copy it to the other 3 positions. | |
3138 | * Otherwise we require the 4 'int's to be the same as would be the | |
3139 | * case if we are using a random uuid. So an explicit uuid will be | |
3140 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3141 | */ |
81a5b4f5 N |
3142 | if (!uuid_set) { |
3143 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3144 | rv = 0; |
81a5b4f5 N |
3145 | } else { |
3146 | if (info->uuid[0] != info->uuid[1] || | |
3147 | info->uuid[1] != info->uuid[2] || | |
3148 | info->uuid[2] != info->uuid[3]) | |
3149 | rv = -1; | |
3150 | else | |
3151 | rv = 0; | |
6e46bf34 | 3152 | } |
81a5b4f5 N |
3153 | if (rv == 0) |
3154 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3155 | } else if (strcmp(update, "assemble") == 0) |
3156 | rv = 0; | |
3157 | else | |
1e2b2765 | 3158 | rv = -1; |
f352c545 | 3159 | |
6e46bf34 DW |
3160 | /* successful update? recompute checksum */ |
3161 | if (rv == 0) | |
3162 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3163 | |
3164 | return rv; | |
cdddbdbc DW |
3165 | } |
3166 | ||
c2c087e6 | 3167 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3168 | { |
c2c087e6 | 3169 | size_t size; |
cdddbdbc | 3170 | |
c2c087e6 DW |
3171 | size = sizeof(struct imsm_super); |
3172 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3173 | size += 2 * sizeof(struct imsm_dev); | |
3174 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3175 | size += (4 - 2) * sizeof(struct imsm_map); | |
3176 | /* 4 possible disk_ord_tbl's */ | |
3177 | size += 4 * (disks - 1) * sizeof(__u32); | |
3178 | ||
3179 | return size; | |
3180 | } | |
3181 | ||
387fcd59 N |
3182 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3183 | unsigned long long data_offset) | |
c2c087e6 DW |
3184 | { |
3185 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3186 | return 0; | |
3187 | ||
3188 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3189 | } |
3190 | ||
ba2de7ba DW |
3191 | static void free_devlist(struct intel_super *super) |
3192 | { | |
3193 | struct intel_dev *dv; | |
3194 | ||
3195 | while (super->devlist) { | |
3196 | dv = super->devlist->next; | |
3197 | free(super->devlist->dev); | |
3198 | free(super->devlist); | |
3199 | super->devlist = dv; | |
3200 | } | |
3201 | } | |
3202 | ||
3203 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3204 | { | |
3205 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3206 | } | |
3207 | ||
cdddbdbc DW |
3208 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3209 | { | |
3210 | /* | |
3211 | * return: | |
3212 | * 0 same, or first was empty, and second was copied | |
3213 | * 1 second had wrong number | |
3214 | * 2 wrong uuid | |
3215 | * 3 wrong other info | |
3216 | */ | |
3217 | struct intel_super *first = st->sb; | |
3218 | struct intel_super *sec = tst->sb; | |
3219 | ||
5d500228 N |
3220 | if (!first) { |
3221 | st->sb = tst->sb; | |
3222 | tst->sb = NULL; | |
3223 | return 0; | |
3224 | } | |
8603ea6f LM |
3225 | /* in platform dependent environment test if the disks |
3226 | * use the same Intel hba | |
cb8f6859 | 3227 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3228 | */ |
6b781d33 AP |
3229 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3230 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3231 | fprintf(stderr, |
6b781d33 AP |
3232 | "HBAs of devices do not match %s != %s\n", |
3233 | get_sys_dev_type(first->hba->type), | |
3234 | get_sys_dev_type(sec->hba->type)); | |
3235 | return 3; | |
3236 | } | |
3237 | if (first->orom != sec->orom) { | |
3238 | fprintf(stderr, | |
3239 | "HBAs of devices do not match %s != %s\n", | |
3240 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3241 | return 3; |
3242 | } | |
3243 | } | |
cdddbdbc | 3244 | |
d23fe947 DW |
3245 | /* if an anchor does not have num_raid_devs set then it is a free |
3246 | * floating spare | |
3247 | */ | |
3248 | if (first->anchor->num_raid_devs > 0 && | |
3249 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3250 | /* Determine if these disks might ever have been |
3251 | * related. Further disambiguation can only take place | |
3252 | * in load_super_imsm_all | |
3253 | */ | |
3254 | __u32 first_family = first->anchor->orig_family_num; | |
3255 | __u32 sec_family = sec->anchor->orig_family_num; | |
3256 | ||
f796af5d DW |
3257 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3258 | MAX_SIGNATURE_LENGTH) != 0) | |
3259 | return 3; | |
3260 | ||
a2b97981 DW |
3261 | if (first_family == 0) |
3262 | first_family = first->anchor->family_num; | |
3263 | if (sec_family == 0) | |
3264 | sec_family = sec->anchor->family_num; | |
3265 | ||
3266 | if (first_family != sec_family) | |
d23fe947 | 3267 | return 3; |
f796af5d | 3268 | |
d23fe947 | 3269 | } |
cdddbdbc | 3270 | |
3e372e5a DW |
3271 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3272 | * family number | |
3273 | */ | |
3274 | if (first->anchor->num_raid_devs == 0 && | |
3275 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3276 | int i; |
ba2de7ba DW |
3277 | struct intel_dev *dv; |
3278 | struct imsm_dev *dev; | |
78d30f94 DW |
3279 | |
3280 | /* we need to copy raid device info from sec if an allocation | |
3281 | * fails here we don't associate the spare | |
3282 | */ | |
3283 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3284 | dv = xmalloc(sizeof(*dv)); |
3285 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3286 | dv->dev = dev; |
3287 | dv->index = i; | |
3288 | dv->next = first->devlist; | |
3289 | first->devlist = dv; | |
78d30f94 | 3290 | } |
709743c5 | 3291 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3292 | /* allocation failure */ |
3293 | free_devlist(first); | |
e12b3daa | 3294 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3295 | return 3; |
78d30f94 | 3296 | } |
3e372e5a | 3297 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3298 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3299 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3300 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3301 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3302 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3303 | } |
3304 | ||
cdddbdbc DW |
3305 | return 0; |
3306 | } | |
3307 | ||
0030e8d6 DW |
3308 | static void fd2devname(int fd, char *name) |
3309 | { | |
3310 | struct stat st; | |
3311 | char path[256]; | |
33a6535d | 3312 | char dname[PATH_MAX]; |
0030e8d6 DW |
3313 | char *nm; |
3314 | int rv; | |
3315 | ||
3316 | name[0] = '\0'; | |
3317 | if (fstat(fd, &st) != 0) | |
3318 | return; | |
3319 | sprintf(path, "/sys/dev/block/%d:%d", | |
3320 | major(st.st_rdev), minor(st.st_rdev)); | |
3321 | ||
9cf014ec | 3322 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3323 | if (rv <= 0) |
3324 | return; | |
9587c373 | 3325 | |
0030e8d6 DW |
3326 | dname[rv] = '\0'; |
3327 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3328 | if (nm) { |
3329 | nm++; | |
3330 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3331 | } | |
0030e8d6 DW |
3332 | } |
3333 | ||
cdddbdbc DW |
3334 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3335 | ||
3336 | static int imsm_read_serial(int fd, char *devname, | |
3337 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3338 | { | |
3339 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3340 | int rv; |
3341 | int rsp_len; | |
1f24f035 | 3342 | int len; |
316e2bf4 DW |
3343 | char *dest; |
3344 | char *src; | |
3345 | char *rsp_buf; | |
3346 | int i; | |
cdddbdbc DW |
3347 | |
3348 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3349 | |
f9ba0ff1 DW |
3350 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3351 | ||
40ebbb9c | 3352 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3353 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3354 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3355 | return 0; |
3356 | } | |
3357 | ||
cdddbdbc DW |
3358 | if (rv != 0) { |
3359 | if (devname) | |
e7b84f9d N |
3360 | pr_err("Failed to retrieve serial for %s\n", |
3361 | devname); | |
cdddbdbc DW |
3362 | return rv; |
3363 | } | |
3364 | ||
3365 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3366 | if (!rsp_len) { |
3367 | if (devname) | |
e7b84f9d N |
3368 | pr_err("Failed to retrieve serial for %s\n", |
3369 | devname); | |
03cd4cc8 DW |
3370 | return 2; |
3371 | } | |
1f24f035 | 3372 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3373 | |
316e2bf4 DW |
3374 | /* trim all whitespace and non-printable characters and convert |
3375 | * ':' to ';' | |
3376 | */ | |
3377 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3378 | src = &rsp_buf[i]; | |
3379 | if (*src > 0x20) { | |
3380 | /* ':' is reserved for use in placeholder serial | |
3381 | * numbers for missing disks | |
3382 | */ | |
3383 | if (*src == ':') | |
3384 | *dest++ = ';'; | |
3385 | else | |
3386 | *dest++ = *src; | |
3387 | } | |
3388 | } | |
3389 | len = dest - rsp_buf; | |
3390 | dest = rsp_buf; | |
3391 | ||
3392 | /* truncate leading characters */ | |
3393 | if (len > MAX_RAID_SERIAL_LEN) { | |
3394 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3395 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3396 | } |
5c3db629 | 3397 | |
5c3db629 | 3398 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3399 | memcpy(serial, dest, len); |
cdddbdbc DW |
3400 | |
3401 | return 0; | |
3402 | } | |
3403 | ||
1f24f035 DW |
3404 | static int serialcmp(__u8 *s1, __u8 *s2) |
3405 | { | |
3406 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3407 | } | |
3408 | ||
3409 | static void serialcpy(__u8 *dest, __u8 *src) | |
3410 | { | |
3411 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3412 | } | |
3413 | ||
54c2c1ea DW |
3414 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3415 | { | |
3416 | struct dl *dl; | |
3417 | ||
3418 | for (dl = super->disks; dl; dl = dl->next) | |
3419 | if (serialcmp(dl->serial, serial) == 0) | |
3420 | break; | |
3421 | ||
3422 | return dl; | |
3423 | } | |
3424 | ||
a2b97981 DW |
3425 | static struct imsm_disk * |
3426 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3427 | { | |
3428 | int i; | |
3429 | ||
3430 | for (i = 0; i < mpb->num_disks; i++) { | |
3431 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3432 | ||
3433 | if (serialcmp(disk->serial, serial) == 0) { | |
3434 | if (idx) | |
3435 | *idx = i; | |
3436 | return disk; | |
3437 | } | |
3438 | } | |
3439 | ||
3440 | return NULL; | |
3441 | } | |
3442 | ||
cdddbdbc DW |
3443 | static int |
3444 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3445 | { | |
a2b97981 | 3446 | struct imsm_disk *disk; |
cdddbdbc DW |
3447 | struct dl *dl; |
3448 | struct stat stb; | |
cdddbdbc | 3449 | int rv; |
a2b97981 | 3450 | char name[40]; |
d23fe947 DW |
3451 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3452 | ||
3453 | rv = imsm_read_serial(fd, devname, serial); | |
3454 | ||
3455 | if (rv != 0) | |
3456 | return 2; | |
3457 | ||
503975b9 | 3458 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3459 | |
a2b97981 DW |
3460 | fstat(fd, &stb); |
3461 | dl->major = major(stb.st_rdev); | |
3462 | dl->minor = minor(stb.st_rdev); | |
3463 | dl->next = super->disks; | |
3464 | dl->fd = keep_fd ? fd : -1; | |
3465 | assert(super->disks == NULL); | |
3466 | super->disks = dl; | |
3467 | serialcpy(dl->serial, serial); | |
3468 | dl->index = -2; | |
3469 | dl->e = NULL; | |
3470 | fd2devname(fd, name); | |
3471 | if (devname) | |
503975b9 | 3472 | dl->devname = xstrdup(devname); |
a2b97981 | 3473 | else |
503975b9 | 3474 | dl->devname = xstrdup(name); |
cdddbdbc | 3475 | |
d23fe947 | 3476 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3477 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3478 | if (disk) { | |
3479 | dl->disk = *disk; | |
3480 | /* only set index on disks that are a member of a | |
3481 | * populated contianer, i.e. one with raid_devs | |
3482 | */ | |
3483 | if (is_failed(&dl->disk)) | |
3f6efecc | 3484 | dl->index = -2; |
a2b97981 DW |
3485 | else if (is_spare(&dl->disk)) |
3486 | dl->index = -1; | |
3f6efecc DW |
3487 | } |
3488 | ||
949c47a0 DW |
3489 | return 0; |
3490 | } | |
3491 | ||
0e600426 | 3492 | #ifndef MDASSEMBLE |
0c046afd DW |
3493 | /* When migrating map0 contains the 'destination' state while map1 |
3494 | * contains the current state. When not migrating map0 contains the | |
3495 | * current state. This routine assumes that map[0].map_state is set to | |
3496 | * the current array state before being called. | |
3497 | * | |
3498 | * Migration is indicated by one of the following states | |
3499 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3500 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3501 | * map1state=unitialized) |
1484e727 | 3502 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3503 | * map1state=normal) |
e3bba0e0 | 3504 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3505 | * map1state=degraded) |
8e59f3d8 AK |
3506 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3507 | * map1state=normal) | |
0c046afd | 3508 | */ |
8e59f3d8 AK |
3509 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3510 | __u8 to_state, int migr_type) | |
3393c6af | 3511 | { |
0c046afd | 3512 | struct imsm_map *dest; |
238c0a71 | 3513 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3514 | |
0c046afd | 3515 | dev->vol.migr_state = 1; |
1484e727 | 3516 | set_migr_type(dev, migr_type); |
f8f603f1 | 3517 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3518 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3519 | |
0556e1a2 | 3520 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3521 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3522 | if ((migr_type == MIGR_REBUILD) || |
3523 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3524 | __u32 ord; |
3525 | int i; | |
3526 | ||
3527 | for (i = 0; i < src->num_members; i++) { | |
3528 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3529 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3530 | } | |
3531 | } | |
3532 | ||
8e59f3d8 AK |
3533 | if (migr_type == MIGR_GEN_MIGR) |
3534 | /* Clear migration record */ | |
3535 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3536 | ||
0c046afd | 3537 | src->map_state = to_state; |
949c47a0 | 3538 | } |
f8f603f1 | 3539 | |
809da78e AK |
3540 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3541 | __u8 map_state) | |
f8f603f1 | 3542 | { |
238c0a71 AK |
3543 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3544 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3545 | MAP_0 : MAP_1); | |
28bce06f | 3546 | int i, j; |
0556e1a2 DW |
3547 | |
3548 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3549 | * completed in the last migration. | |
3550 | * | |
28bce06f | 3551 | * FIXME add support for raid-level-migration |
0556e1a2 | 3552 | */ |
809da78e AK |
3553 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3554 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3555 | /* when final map state is other than expected | |
3556 | * merge maps (not for migration) | |
3557 | */ | |
3558 | int failed; | |
3559 | ||
3560 | for (i = 0; i < prev->num_members; i++) | |
3561 | for (j = 0; j < map->num_members; j++) | |
3562 | /* during online capacity expansion | |
3563 | * disks position can be changed | |
3564 | * if takeover is used | |
3565 | */ | |
3566 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3567 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3568 | map->disk_ord_tbl[j] |= | |
3569 | prev->disk_ord_tbl[i]; | |
3570 | break; | |
3571 | } | |
3572 | failed = imsm_count_failed(super, dev, MAP_0); | |
3573 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3574 | } | |
f8f603f1 DW |
3575 | |
3576 | dev->vol.migr_state = 0; | |
ea672ee1 | 3577 | set_migr_type(dev, 0); |
f8f603f1 DW |
3578 | dev->vol.curr_migr_unit = 0; |
3579 | map->map_state = map_state; | |
3580 | } | |
0e600426 | 3581 | #endif |
949c47a0 DW |
3582 | |
3583 | static int parse_raid_devices(struct intel_super *super) | |
3584 | { | |
3585 | int i; | |
3586 | struct imsm_dev *dev_new; | |
4d7b1503 | 3587 | size_t len, len_migr; |
401d313b | 3588 | size_t max_len = 0; |
4d7b1503 DW |
3589 | size_t space_needed = 0; |
3590 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3591 | |
3592 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3593 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3594 | struct intel_dev *dv; |
949c47a0 | 3595 | |
4d7b1503 DW |
3596 | len = sizeof_imsm_dev(dev_iter, 0); |
3597 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3598 | if (len_migr > len) | |
3599 | space_needed += len_migr - len; | |
ca9de185 | 3600 | |
503975b9 | 3601 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
3602 | if (max_len < len_migr) |
3603 | max_len = len_migr; | |
3604 | if (max_len > len_migr) | |
3605 | space_needed += max_len - len_migr; | |
503975b9 | 3606 | dev_new = xmalloc(max_len); |
949c47a0 | 3607 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3608 | dv->dev = dev_new; |
3609 | dv->index = i; | |
3610 | dv->next = super->devlist; | |
3611 | super->devlist = dv; | |
949c47a0 | 3612 | } |
cdddbdbc | 3613 | |
4d7b1503 DW |
3614 | /* ensure that super->buf is large enough when all raid devices |
3615 | * are migrating | |
3616 | */ | |
3617 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3618 | void *buf; | |
3619 | ||
3620 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3621 | if (posix_memalign(&buf, 512, len) != 0) | |
3622 | return 1; | |
3623 | ||
1f45a8ad DW |
3624 | memcpy(buf, super->buf, super->len); |
3625 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3626 | free(super->buf); |
3627 | super->buf = buf; | |
3628 | super->len = len; | |
3629 | } | |
ca9de185 | 3630 | |
cdddbdbc DW |
3631 | return 0; |
3632 | } | |
3633 | ||
604b746f JD |
3634 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3635 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3636 | { | |
3637 | void *ptr = NULL; | |
3638 | ||
3639 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3640 | ptr = mpb; | |
3641 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
1011e834 | 3642 | } |
604b746f JD |
3643 | |
3644 | return ptr; | |
3645 | } | |
3646 | ||
e2f41b2c AK |
3647 | /******************************************************************************* |
3648 | * Function: check_mpb_migr_compatibility | |
3649 | * Description: Function checks for unsupported migration features: | |
3650 | * - migration optimization area (pba_of_lba0) | |
3651 | * - descending reshape (ascending_migr) | |
3652 | * Parameters: | |
3653 | * super : imsm metadata information | |
3654 | * Returns: | |
3655 | * 0 : migration is compatible | |
3656 | * -1 : migration is not compatible | |
3657 | ******************************************************************************/ | |
3658 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3659 | { | |
3660 | struct imsm_map *map0, *map1; | |
3661 | struct migr_record *migr_rec = super->migr_rec; | |
3662 | int i; | |
3663 | ||
3664 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3665 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3666 | ||
3667 | if (dev_iter && | |
3668 | dev_iter->vol.migr_state == 1 && | |
3669 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3670 | /* This device is migrating */ | |
238c0a71 AK |
3671 | map0 = get_imsm_map(dev_iter, MAP_0); |
3672 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3673 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3674 | /* migration optimization area was used */ |
3675 | return -1; | |
3676 | if (migr_rec->ascending_migr == 0 | |
3677 | && migr_rec->dest_depth_per_unit > 0) | |
3678 | /* descending reshape not supported yet */ | |
3679 | return -1; | |
3680 | } | |
3681 | } | |
3682 | return 0; | |
3683 | } | |
3684 | ||
d23fe947 | 3685 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3686 | |
cdddbdbc | 3687 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3688 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3689 | */ |
3690 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3691 | { | |
3692 | unsigned long long dsize; | |
cdddbdbc DW |
3693 | unsigned long long sectors; |
3694 | struct stat; | |
6416d527 | 3695 | struct imsm_super *anchor; |
cdddbdbc DW |
3696 | __u32 check_sum; |
3697 | ||
cdddbdbc | 3698 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3699 | if (dsize < 1024) { |
3700 | if (devname) | |
e7b84f9d N |
3701 | pr_err("%s: device to small for imsm\n", |
3702 | devname); | |
64436f06 N |
3703 | return 1; |
3704 | } | |
cdddbdbc DW |
3705 | |
3706 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3707 | if (devname) | |
e7b84f9d N |
3708 | pr_err("Cannot seek to anchor block on %s: %s\n", |
3709 | devname, strerror(errno)); | |
cdddbdbc DW |
3710 | return 1; |
3711 | } | |
3712 | ||
949c47a0 | 3713 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e | 3714 | if (devname) |
7a862a02 | 3715 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
3716 | return 1; |
3717 | } | |
949c47a0 | 3718 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc | 3719 | if (devname) |
e7b84f9d N |
3720 | pr_err("Cannot read anchor block on %s: %s\n", |
3721 | devname, strerror(errno)); | |
6416d527 | 3722 | free(anchor); |
cdddbdbc DW |
3723 | return 1; |
3724 | } | |
3725 | ||
6416d527 | 3726 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 3727 | if (devname) |
e7b84f9d | 3728 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 3729 | free(anchor); |
cdddbdbc DW |
3730 | return 2; |
3731 | } | |
3732 | ||
d23fe947 | 3733 | __free_imsm(super, 0); |
f2f5c343 LM |
3734 | /* reload capability and hba */ |
3735 | ||
3736 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3737 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3738 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3739 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc | 3740 | if (devname) |
e7b84f9d N |
3741 | pr_err("unable to allocate %zu byte mpb buffer\n", |
3742 | super->len); | |
6416d527 | 3743 | free(anchor); |
cdddbdbc DW |
3744 | return 2; |
3745 | } | |
949c47a0 | 3746 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3747 | |
6416d527 NB |
3748 | sectors = mpb_sectors(anchor) - 1; |
3749 | free(anchor); | |
8e59f3d8 | 3750 | |
17a4eaf9 | 3751 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
1ade5cc1 | 3752 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
3753 | free(super->buf); |
3754 | return 2; | |
3755 | } | |
51d83f5d | 3756 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3757 | |
949c47a0 | 3758 | if (!sectors) { |
ecf45690 DW |
3759 | check_sum = __gen_imsm_checksum(super->anchor); |
3760 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3761 | if (devname) | |
e7b84f9d N |
3762 | pr_err("IMSM checksum %x != %x on %s\n", |
3763 | check_sum, | |
3764 | __le32_to_cpu(super->anchor->check_sum), | |
3765 | devname); | |
ecf45690 DW |
3766 | return 2; |
3767 | } | |
3768 | ||
a2b97981 | 3769 | return 0; |
949c47a0 | 3770 | } |
cdddbdbc DW |
3771 | |
3772 | /* read the extended mpb */ | |
3773 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3774 | if (devname) | |
e7b84f9d N |
3775 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
3776 | devname, strerror(errno)); | |
cdddbdbc DW |
3777 | return 1; |
3778 | } | |
3779 | ||
f21e18ca | 3780 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc | 3781 | if (devname) |
e7b84f9d N |
3782 | pr_err("Cannot read extended mpb on %s: %s\n", |
3783 | devname, strerror(errno)); | |
cdddbdbc DW |
3784 | return 2; |
3785 | } | |
3786 | ||
949c47a0 DW |
3787 | check_sum = __gen_imsm_checksum(super->anchor); |
3788 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 3789 | if (devname) |
e7b84f9d N |
3790 | pr_err("IMSM checksum %x != %x on %s\n", |
3791 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
3792 | devname); | |
db575f3b | 3793 | return 3; |
cdddbdbc DW |
3794 | } |
3795 | ||
604b746f JD |
3796 | /* FIXME the BBM log is disk specific so we cannot use this global |
3797 | * buffer for all disks. Ok for now since we only look at the global | |
3798 | * bbm_log_size parameter to gate assembly | |
3799 | */ | |
3800 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3801 | ||
a2b97981 DW |
3802 | return 0; |
3803 | } | |
3804 | ||
8e59f3d8 AK |
3805 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3806 | ||
97f81ee2 CA |
3807 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3808 | static void clear_hi(struct intel_super *super) | |
3809 | { | |
3810 | struct imsm_super *mpb = super->anchor; | |
3811 | int i, n; | |
3812 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3813 | return; | |
3814 | for (i = 0; i < mpb->num_disks; ++i) { | |
3815 | struct imsm_disk *disk = &mpb->disk[i]; | |
3816 | disk->total_blocks_hi = 0; | |
3817 | } | |
3818 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3819 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3820 | if (!dev) | |
3821 | return; | |
3822 | for (n = 0; n < 2; ++n) { | |
3823 | struct imsm_map *map = get_imsm_map(dev, n); | |
3824 | if (!map) | |
3825 | continue; | |
3826 | map->pba_of_lba0_hi = 0; | |
3827 | map->blocks_per_member_hi = 0; | |
3828 | map->num_data_stripes_hi = 0; | |
3829 | } | |
3830 | } | |
3831 | } | |
3832 | ||
a2b97981 DW |
3833 | static int |
3834 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3835 | { | |
3836 | int err; | |
3837 | ||
3838 | err = load_imsm_mpb(fd, super, devname); | |
3839 | if (err) | |
3840 | return err; | |
3841 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3842 | if (err) | |
3843 | return err; | |
3844 | err = parse_raid_devices(super); | |
97f81ee2 | 3845 | clear_hi(super); |
a2b97981 | 3846 | return err; |
cdddbdbc DW |
3847 | } |
3848 | ||
ae6aad82 DW |
3849 | static void __free_imsm_disk(struct dl *d) |
3850 | { | |
3851 | if (d->fd >= 0) | |
3852 | close(d->fd); | |
3853 | if (d->devname) | |
3854 | free(d->devname); | |
0dcecb2e DW |
3855 | if (d->e) |
3856 | free(d->e); | |
ae6aad82 DW |
3857 | free(d); |
3858 | ||
3859 | } | |
1a64be56 | 3860 | |
cdddbdbc DW |
3861 | static void free_imsm_disks(struct intel_super *super) |
3862 | { | |
47ee5a45 | 3863 | struct dl *d; |
cdddbdbc | 3864 | |
47ee5a45 DW |
3865 | while (super->disks) { |
3866 | d = super->disks; | |
cdddbdbc | 3867 | super->disks = d->next; |
ae6aad82 | 3868 | __free_imsm_disk(d); |
cdddbdbc | 3869 | } |
cb82edca AK |
3870 | while (super->disk_mgmt_list) { |
3871 | d = super->disk_mgmt_list; | |
3872 | super->disk_mgmt_list = d->next; | |
3873 | __free_imsm_disk(d); | |
3874 | } | |
47ee5a45 DW |
3875 | while (super->missing) { |
3876 | d = super->missing; | |
3877 | super->missing = d->next; | |
3878 | __free_imsm_disk(d); | |
3879 | } | |
3880 | ||
cdddbdbc DW |
3881 | } |
3882 | ||
9ca2c81c | 3883 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3884 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3885 | { |
88654014 LM |
3886 | struct intel_hba *elem, *next; |
3887 | ||
9ca2c81c | 3888 | if (super->buf) { |
949c47a0 | 3889 | free(super->buf); |
9ca2c81c DW |
3890 | super->buf = NULL; |
3891 | } | |
f2f5c343 LM |
3892 | /* unlink capability description */ |
3893 | super->orom = NULL; | |
8e59f3d8 AK |
3894 | if (super->migr_rec_buf) { |
3895 | free(super->migr_rec_buf); | |
3896 | super->migr_rec_buf = NULL; | |
3897 | } | |
d23fe947 DW |
3898 | if (free_disks) |
3899 | free_imsm_disks(super); | |
ba2de7ba | 3900 | free_devlist(super); |
88654014 LM |
3901 | elem = super->hba; |
3902 | while (elem) { | |
3903 | if (elem->path) | |
3904 | free((void *)elem->path); | |
3905 | next = elem->next; | |
3906 | free(elem); | |
3907 | elem = next; | |
88c32bb1 | 3908 | } |
88654014 | 3909 | super->hba = NULL; |
cdddbdbc DW |
3910 | } |
3911 | ||
9ca2c81c DW |
3912 | static void free_imsm(struct intel_super *super) |
3913 | { | |
d23fe947 | 3914 | __free_imsm(super, 1); |
9ca2c81c DW |
3915 | free(super); |
3916 | } | |
cdddbdbc DW |
3917 | |
3918 | static void free_super_imsm(struct supertype *st) | |
3919 | { | |
3920 | struct intel_super *super = st->sb; | |
3921 | ||
3922 | if (!super) | |
3923 | return; | |
3924 | ||
3925 | free_imsm(super); | |
3926 | st->sb = NULL; | |
3927 | } | |
3928 | ||
49133e57 | 3929 | static struct intel_super *alloc_super(void) |
c2c087e6 | 3930 | { |
503975b9 | 3931 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 3932 | |
503975b9 N |
3933 | super->current_vol = -1; |
3934 | super->create_offset = ~((unsigned long long) 0); | |
c2c087e6 DW |
3935 | return super; |
3936 | } | |
3937 | ||
f0f5a016 LM |
3938 | /* |
3939 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3940 | */ | |
d424212e | 3941 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3942 | { |
3943 | struct sys_dev *hba_name; | |
3944 | int rv = 0; | |
3945 | ||
3946 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3947 | super->orom = NULL; |
f0f5a016 LM |
3948 | super->hba = NULL; |
3949 | return 0; | |
3950 | } | |
3951 | hba_name = find_disk_attached_hba(fd, NULL); | |
3952 | if (!hba_name) { | |
d424212e | 3953 | if (devname) |
e7b84f9d N |
3954 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
3955 | devname); | |
f0f5a016 LM |
3956 | return 1; |
3957 | } | |
3958 | rv = attach_hba_to_super(super, hba_name); | |
3959 | if (rv == 2) { | |
d424212e N |
3960 | if (devname) { |
3961 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3962 | |
60f0f54d PB |
3963 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
3964 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 3965 | devname, |
614902f6 | 3966 | get_sys_dev_type(hba_name->type), |
60f0f54d | 3967 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 3968 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
3969 | get_sys_dev_type(super->hba->type), |
3970 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 3971 | |
f0f5a016 LM |
3972 | while (hba) { |
3973 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3974 | if (hba->next) | |
3975 | fprintf(stderr, ", "); | |
3976 | hba = hba->next; | |
3977 | } | |
6b781d33 | 3978 | fprintf(stderr, ").\n" |
60f0f54d PB |
3979 | " Mixing devices attached to different %s is not allowed.\n", |
3980 | hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers"); | |
f0f5a016 | 3981 | } |
f0f5a016 LM |
3982 | return 2; |
3983 | } | |
6b781d33 | 3984 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
3985 | if (!super->orom) |
3986 | return 3; | |
614902f6 | 3987 | |
f0f5a016 LM |
3988 | return 0; |
3989 | } | |
3990 | ||
47ee5a45 DW |
3991 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3992 | * disks that have disappeared from the system. This routine relies on | |
3993 | * the mpb being uptodate, which it is at load time. | |
3994 | */ | |
3995 | static int find_missing(struct intel_super *super) | |
3996 | { | |
3997 | int i; | |
3998 | struct imsm_super *mpb = super->anchor; | |
3999 | struct dl *dl; | |
4000 | struct imsm_disk *disk; | |
47ee5a45 DW |
4001 | |
4002 | for (i = 0; i < mpb->num_disks; i++) { | |
4003 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4004 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4005 | if (dl) |
4006 | continue; | |
47ee5a45 | 4007 | |
503975b9 | 4008 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4009 | dl->major = 0; |
4010 | dl->minor = 0; | |
4011 | dl->fd = -1; | |
503975b9 | 4012 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4013 | dl->index = i; |
4014 | serialcpy(dl->serial, disk->serial); | |
4015 | dl->disk = *disk; | |
689c9bf3 | 4016 | dl->e = NULL; |
47ee5a45 DW |
4017 | dl->next = super->missing; |
4018 | super->missing = dl; | |
4019 | } | |
4020 | ||
4021 | return 0; | |
4022 | } | |
4023 | ||
3960e579 | 4024 | #ifndef MDASSEMBLE |
a2b97981 DW |
4025 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4026 | { | |
4027 | struct intel_disk *idisk = disk_list; | |
4028 | ||
4029 | while (idisk) { | |
4030 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4031 | break; | |
4032 | idisk = idisk->next; | |
4033 | } | |
4034 | ||
4035 | return idisk; | |
4036 | } | |
4037 | ||
4038 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4039 | struct intel_super *super, | |
4040 | struct intel_disk **disk_list) | |
4041 | { | |
4042 | struct imsm_disk *d = &super->disks->disk; | |
4043 | struct imsm_super *mpb = super->anchor; | |
4044 | int i, j; | |
4045 | ||
4046 | for (i = 0; i < tbl_size; i++) { | |
4047 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4048 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4049 | ||
4050 | if (tbl_mpb->family_num == mpb->family_num) { | |
4051 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4052 | dprintf("mpb from %d:%d matches %d:%d\n", |
4053 | super->disks->major, | |
a2b97981 DW |
4054 | super->disks->minor, |
4055 | table[i]->disks->major, | |
4056 | table[i]->disks->minor); | |
4057 | break; | |
4058 | } | |
4059 | ||
4060 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4061 | is_configured(d) == is_configured(tbl_d)) && | |
4062 | tbl_mpb->generation_num < mpb->generation_num) { | |
4063 | /* current version of the mpb is a | |
4064 | * better candidate than the one in | |
4065 | * super_table, but copy over "cross | |
4066 | * generational" status | |
4067 | */ | |
4068 | struct intel_disk *idisk; | |
4069 | ||
1ade5cc1 N |
4070 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4071 | super->disks->major, | |
a2b97981 DW |
4072 | super->disks->minor, |
4073 | table[i]->disks->major, | |
4074 | table[i]->disks->minor); | |
4075 | ||
4076 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4077 | if (idisk && is_failed(&idisk->disk)) | |
4078 | tbl_d->status |= FAILED_DISK; | |
4079 | break; | |
4080 | } else { | |
4081 | struct intel_disk *idisk; | |
4082 | struct imsm_disk *disk; | |
4083 | ||
4084 | /* tbl_mpb is more up to date, but copy | |
4085 | * over cross generational status before | |
4086 | * returning | |
4087 | */ | |
4088 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4089 | if (disk && is_failed(disk)) | |
4090 | d->status |= FAILED_DISK; | |
4091 | ||
4092 | idisk = disk_list_get(d->serial, *disk_list); | |
4093 | if (idisk) { | |
4094 | idisk->owner = i; | |
4095 | if (disk && is_configured(disk)) | |
4096 | idisk->disk.status |= CONFIGURED_DISK; | |
4097 | } | |
4098 | ||
1ade5cc1 N |
4099 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4100 | super->disks->major, | |
a2b97981 DW |
4101 | super->disks->minor, |
4102 | table[i]->disks->major, | |
4103 | table[i]->disks->minor); | |
4104 | ||
4105 | return tbl_size; | |
4106 | } | |
4107 | } | |
4108 | } | |
4109 | ||
4110 | if (i >= tbl_size) | |
4111 | table[tbl_size++] = super; | |
4112 | else | |
4113 | table[i] = super; | |
4114 | ||
4115 | /* update/extend the merged list of imsm_disk records */ | |
4116 | for (j = 0; j < mpb->num_disks; j++) { | |
4117 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4118 | struct intel_disk *idisk; | |
4119 | ||
4120 | idisk = disk_list_get(disk->serial, *disk_list); | |
4121 | if (idisk) { | |
4122 | idisk->disk.status |= disk->status; | |
4123 | if (is_configured(&idisk->disk) || | |
4124 | is_failed(&idisk->disk)) | |
4125 | idisk->disk.status &= ~(SPARE_DISK); | |
4126 | } else { | |
503975b9 | 4127 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4128 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4129 | idisk->disk = *disk; | |
4130 | idisk->next = *disk_list; | |
4131 | *disk_list = idisk; | |
4132 | } | |
4133 | ||
4134 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4135 | idisk->owner = i; | |
4136 | } | |
4137 | ||
4138 | return tbl_size; | |
4139 | } | |
4140 | ||
4141 | static struct intel_super * | |
4142 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4143 | const int owner) | |
4144 | { | |
4145 | struct imsm_super *mpb = super->anchor; | |
4146 | int ok_count = 0; | |
4147 | int i; | |
4148 | ||
4149 | for (i = 0; i < mpb->num_disks; i++) { | |
4150 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4151 | struct intel_disk *idisk; | |
4152 | ||
4153 | idisk = disk_list_get(disk->serial, disk_list); | |
4154 | if (idisk) { | |
4155 | if (idisk->owner == owner || | |
4156 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4157 | ok_count++; | |
4158 | else | |
1ade5cc1 N |
4159 | dprintf("'%.16s' owner %d != %d\n", |
4160 | disk->serial, idisk->owner, | |
a2b97981 DW |
4161 | owner); |
4162 | } else { | |
1ade5cc1 N |
4163 | dprintf("unknown disk %x [%d]: %.16s\n", |
4164 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4165 | disk->serial); |
4166 | break; | |
4167 | } | |
4168 | } | |
4169 | ||
4170 | if (ok_count == mpb->num_disks) | |
4171 | return super; | |
4172 | return NULL; | |
4173 | } | |
4174 | ||
4175 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4176 | { | |
4177 | struct intel_super *s; | |
4178 | ||
4179 | for (s = super_list; s; s = s->next) { | |
4180 | if (family_num != s->anchor->family_num) | |
4181 | continue; | |
e12b3daa | 4182 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4183 | __le32_to_cpu(family_num), s->disks->devname); |
4184 | } | |
4185 | } | |
4186 | ||
4187 | static struct intel_super * | |
4188 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4189 | { | |
4190 | struct intel_super *super_table[len]; | |
4191 | struct intel_disk *disk_list = NULL; | |
4192 | struct intel_super *champion, *spare; | |
4193 | struct intel_super *s, **del; | |
4194 | int tbl_size = 0; | |
4195 | int conflict; | |
4196 | int i; | |
4197 | ||
4198 | memset(super_table, 0, sizeof(super_table)); | |
4199 | for (s = *super_list; s; s = s->next) | |
4200 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4201 | ||
4202 | for (i = 0; i < tbl_size; i++) { | |
4203 | struct imsm_disk *d; | |
4204 | struct intel_disk *idisk; | |
4205 | struct imsm_super *mpb = super_table[i]->anchor; | |
4206 | ||
4207 | s = super_table[i]; | |
4208 | d = &s->disks->disk; | |
4209 | ||
4210 | /* 'd' must appear in merged disk list for its | |
4211 | * configuration to be valid | |
4212 | */ | |
4213 | idisk = disk_list_get(d->serial, disk_list); | |
4214 | if (idisk && idisk->owner == i) | |
4215 | s = validate_members(s, disk_list, i); | |
4216 | else | |
4217 | s = NULL; | |
4218 | ||
4219 | if (!s) | |
1ade5cc1 N |
4220 | dprintf("marking family: %#x from %d:%d offline\n", |
4221 | mpb->family_num, | |
a2b97981 DW |
4222 | super_table[i]->disks->major, |
4223 | super_table[i]->disks->minor); | |
4224 | super_table[i] = s; | |
4225 | } | |
4226 | ||
4227 | /* This is where the mdadm implementation differs from the Windows | |
4228 | * driver which has no strict concept of a container. We can only | |
4229 | * assemble one family from a container, so when returning a prodigal | |
4230 | * array member to this system the code will not be able to disambiguate | |
4231 | * the container contents that should be assembled ("foreign" versus | |
4232 | * "local"). It requires user intervention to set the orig_family_num | |
4233 | * to a new value to establish a new container. The Windows driver in | |
4234 | * this situation fixes up the volume name in place and manages the | |
4235 | * foreign array as an independent entity. | |
4236 | */ | |
4237 | s = NULL; | |
4238 | spare = NULL; | |
4239 | conflict = 0; | |
4240 | for (i = 0; i < tbl_size; i++) { | |
4241 | struct intel_super *tbl_ent = super_table[i]; | |
4242 | int is_spare = 0; | |
4243 | ||
4244 | if (!tbl_ent) | |
4245 | continue; | |
4246 | ||
4247 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4248 | spare = tbl_ent; | |
4249 | is_spare = 1; | |
4250 | } | |
4251 | ||
4252 | if (s && !is_spare) { | |
4253 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4254 | conflict++; | |
4255 | } else if (!s && !is_spare) | |
4256 | s = tbl_ent; | |
4257 | } | |
4258 | ||
4259 | if (!s) | |
4260 | s = spare; | |
4261 | if (!s) { | |
4262 | champion = NULL; | |
4263 | goto out; | |
4264 | } | |
4265 | champion = s; | |
4266 | ||
4267 | if (conflict) | |
7a862a02 | 4268 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4269 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4270 | ||
4271 | /* collect all dl's onto 'champion', and update them to | |
4272 | * champion's version of the status | |
4273 | */ | |
4274 | for (s = *super_list; s; s = s->next) { | |
4275 | struct imsm_super *mpb = champion->anchor; | |
4276 | struct dl *dl = s->disks; | |
4277 | ||
4278 | if (s == champion) | |
4279 | continue; | |
4280 | ||
5d7b407a CA |
4281 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4282 | ||
a2b97981 DW |
4283 | for (i = 0; i < mpb->num_disks; i++) { |
4284 | struct imsm_disk *disk; | |
4285 | ||
4286 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4287 | if (disk) { | |
4288 | dl->disk = *disk; | |
4289 | /* only set index on disks that are a member of | |
4290 | * a populated contianer, i.e. one with | |
4291 | * raid_devs | |
4292 | */ | |
4293 | if (is_failed(&dl->disk)) | |
4294 | dl->index = -2; | |
4295 | else if (is_spare(&dl->disk)) | |
4296 | dl->index = -1; | |
4297 | break; | |
4298 | } | |
4299 | } | |
4300 | ||
4301 | if (i >= mpb->num_disks) { | |
4302 | struct intel_disk *idisk; | |
4303 | ||
4304 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4305 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4306 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4307 | dl->index = -1; | |
4308 | else { | |
4309 | dl->index = -2; | |
4310 | continue; | |
4311 | } | |
4312 | } | |
4313 | ||
4314 | dl->next = champion->disks; | |
4315 | champion->disks = dl; | |
4316 | s->disks = NULL; | |
4317 | } | |
4318 | ||
4319 | /* delete 'champion' from super_list */ | |
4320 | for (del = super_list; *del; ) { | |
4321 | if (*del == champion) { | |
4322 | *del = (*del)->next; | |
4323 | break; | |
4324 | } else | |
4325 | del = &(*del)->next; | |
4326 | } | |
4327 | champion->next = NULL; | |
4328 | ||
4329 | out: | |
4330 | while (disk_list) { | |
4331 | struct intel_disk *idisk = disk_list; | |
4332 | ||
4333 | disk_list = disk_list->next; | |
4334 | free(idisk); | |
4335 | } | |
4336 | ||
4337 | return champion; | |
4338 | } | |
4339 | ||
9587c373 LM |
4340 | static int |
4341 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4342 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4343 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4344 | static int |
4345 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4346 | int *max, int keep_fd); | |
4347 | ||
cdddbdbc | 4348 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4349 | char *devname, struct md_list *devlist, |
4350 | int keep_fd) | |
cdddbdbc | 4351 | { |
a2b97981 DW |
4352 | struct intel_super *super_list = NULL; |
4353 | struct intel_super *super = NULL; | |
a2b97981 | 4354 | int err = 0; |
9587c373 | 4355 | int i = 0; |
dab4a513 | 4356 | |
9587c373 LM |
4357 | if (fd >= 0) |
4358 | /* 'fd' is an opened container */ | |
4359 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4360 | else | |
ec50f7b6 LM |
4361 | /* get super block from devlist devices */ |
4362 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4363 | if (err) |
1602d52c | 4364 | goto error; |
a2b97981 DW |
4365 | /* all mpbs enter, maybe one leaves */ |
4366 | super = imsm_thunderdome(&super_list, i); | |
4367 | if (!super) { | |
4368 | err = 1; | |
4369 | goto error; | |
cdddbdbc DW |
4370 | } |
4371 | ||
47ee5a45 DW |
4372 | if (find_missing(super) != 0) { |
4373 | free_imsm(super); | |
a2b97981 DW |
4374 | err = 2; |
4375 | goto error; | |
47ee5a45 | 4376 | } |
8e59f3d8 AK |
4377 | |
4378 | /* load migration record */ | |
4379 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4380 | if (err == -1) { |
4381 | /* migration is in progress, | |
4382 | * but migr_rec cannot be loaded, | |
4383 | */ | |
8e59f3d8 AK |
4384 | err = 4; |
4385 | goto error; | |
4386 | } | |
e2f41b2c AK |
4387 | |
4388 | /* Check migration compatibility */ | |
4c965cc9 | 4389 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e7b84f9d | 4390 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4391 | if (devname) |
4392 | fprintf(stderr, " on %s\n", devname); | |
4393 | else | |
4394 | fprintf(stderr, " (IMSM).\n"); | |
4395 | ||
4396 | err = 5; | |
4397 | goto error; | |
4398 | } | |
4399 | ||
a2b97981 DW |
4400 | err = 0; |
4401 | ||
4402 | error: | |
4403 | while (super_list) { | |
4404 | struct intel_super *s = super_list; | |
4405 | ||
4406 | super_list = super_list->next; | |
4407 | free_imsm(s); | |
4408 | } | |
9587c373 | 4409 | |
a2b97981 DW |
4410 | if (err) |
4411 | return err; | |
f7e7067b | 4412 | |
cdddbdbc | 4413 | *sbp = super; |
9587c373 | 4414 | if (fd >= 0) |
4dd2df09 | 4415 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4416 | else |
4dd2df09 | 4417 | st->container_devnm[0] = 0; |
a2b97981 | 4418 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4419 | st->ss = &super_imsm; |
cdddbdbc DW |
4420 | st->minor_version = 0; |
4421 | st->max_devs = IMSM_MAX_DEVICES; | |
4422 | } | |
cdddbdbc DW |
4423 | return 0; |
4424 | } | |
2b959fbf | 4425 | |
ec50f7b6 LM |
4426 | static int |
4427 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4428 | int *max, int keep_fd) | |
4429 | { | |
4430 | struct md_list *tmpdev; | |
4431 | int err = 0; | |
4432 | int i = 0; | |
9587c373 | 4433 | |
ec50f7b6 LM |
4434 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4435 | if (tmpdev->used != 1) | |
4436 | continue; | |
4437 | if (tmpdev->container == 1) { | |
ca9de185 | 4438 | int lmax = 0; |
ec50f7b6 LM |
4439 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4440 | if (fd < 0) { | |
e7b84f9d | 4441 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4442 | tmpdev->devname, strerror(errno)); |
4443 | err = 8; | |
4444 | goto error; | |
4445 | } | |
4446 | err = get_sra_super_block(fd, super_list, | |
4447 | tmpdev->devname, &lmax, | |
4448 | keep_fd); | |
4449 | i += lmax; | |
4450 | close(fd); | |
4451 | if (err) { | |
4452 | err = 7; | |
4453 | goto error; | |
4454 | } | |
4455 | } else { | |
4456 | int major = major(tmpdev->st_rdev); | |
4457 | int minor = minor(tmpdev->st_rdev); | |
4458 | err = get_super_block(super_list, | |
4dd2df09 | 4459 | NULL, |
ec50f7b6 LM |
4460 | tmpdev->devname, |
4461 | major, minor, | |
4462 | keep_fd); | |
4463 | i++; | |
4464 | if (err) { | |
4465 | err = 6; | |
4466 | goto error; | |
4467 | } | |
4468 | } | |
4469 | } | |
4470 | error: | |
4471 | *max = i; | |
4472 | return err; | |
4473 | } | |
9587c373 | 4474 | |
4dd2df09 | 4475 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
4476 | int major, int minor, int keep_fd) |
4477 | { | |
594dc1b8 | 4478 | struct intel_super *s; |
9587c373 LM |
4479 | char nm[32]; |
4480 | int dfd = -1; | |
9587c373 LM |
4481 | int err = 0; |
4482 | int retry; | |
4483 | ||
4484 | s = alloc_super(); | |
4485 | if (!s) { | |
4486 | err = 1; | |
4487 | goto error; | |
4488 | } | |
4489 | ||
4490 | sprintf(nm, "%d:%d", major, minor); | |
4491 | dfd = dev_open(nm, O_RDWR); | |
4492 | if (dfd < 0) { | |
4493 | err = 2; | |
4494 | goto error; | |
4495 | } | |
4496 | ||
cb8f6859 | 4497 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
4498 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
4499 | ||
4500 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 4501 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
4502 | for (retry = 0; retry < 3; retry++) { |
4503 | usleep(3000); | |
4504 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4505 | if (err != 3) | |
4506 | break; | |
4507 | } | |
4508 | error: | |
4509 | if (!err) { | |
4510 | s->next = *super_list; | |
4511 | *super_list = s; | |
4512 | } else { | |
4513 | if (s) | |
4514 | free(s); | |
36614e95 | 4515 | if (dfd >= 0) |
9587c373 LM |
4516 | close(dfd); |
4517 | } | |
4518 | if ((dfd >= 0) && (!keep_fd)) | |
4519 | close(dfd); | |
4520 | return err; | |
4521 | ||
4522 | } | |
4523 | ||
4524 | static int | |
4525 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4526 | { | |
4527 | struct mdinfo *sra; | |
4dd2df09 | 4528 | char *devnm; |
9587c373 LM |
4529 | struct mdinfo *sd; |
4530 | int err = 0; | |
4531 | int i = 0; | |
4dd2df09 | 4532 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
4533 | if (!sra) |
4534 | return 1; | |
4535 | ||
4536 | if (sra->array.major_version != -1 || | |
4537 | sra->array.minor_version != -2 || | |
4538 | strcmp(sra->text_version, "imsm") != 0) { | |
4539 | err = 1; | |
4540 | goto error; | |
4541 | } | |
4542 | /* load all mpbs */ | |
4dd2df09 | 4543 | devnm = fd2devnm(fd); |
9587c373 | 4544 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 4545 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
4546 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
4547 | err = 7; | |
4548 | goto error; | |
4549 | } | |
4550 | } | |
4551 | error: | |
4552 | sysfs_free(sra); | |
4553 | *max = i; | |
4554 | return err; | |
4555 | } | |
4556 | ||
2b959fbf N |
4557 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4558 | { | |
ec50f7b6 | 4559 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4560 | } |
cdddbdbc DW |
4561 | #endif |
4562 | ||
4563 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4564 | { | |
4565 | struct intel_super *super; | |
4566 | int rv; | |
8a3544f8 | 4567 | int retry; |
cdddbdbc | 4568 | |
357ac106 | 4569 | if (test_partition(fd)) |
691c6ee1 N |
4570 | /* IMSM not allowed on partitions */ |
4571 | return 1; | |
4572 | ||
37424f13 DW |
4573 | free_super_imsm(st); |
4574 | ||
49133e57 | 4575 | super = alloc_super(); |
ea2bc72b LM |
4576 | /* Load hba and capabilities if they exist. |
4577 | * But do not preclude loading metadata in case capabilities or hba are | |
4578 | * non-compliant and ignore_hw_compat is set. | |
4579 | */ | |
d424212e | 4580 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4581 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4582 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 | 4583 | if (devname) |
e7b84f9d | 4584 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
4585 | free_imsm(super); |
4586 | return 2; | |
4587 | } | |
a2b97981 | 4588 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 4589 | |
8a3544f8 AP |
4590 | /* retry the load if we might have raced against mdmon */ |
4591 | if (rv == 3) { | |
4592 | struct mdstat_ent *mdstat = mdstat_by_component(fd2devnm(fd)); | |
4593 | ||
4594 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
4595 | for (retry = 0; retry < 3; retry++) { | |
4596 | usleep(3000); | |
4597 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
4598 | if (rv != 3) | |
4599 | break; | |
4600 | } | |
4601 | } | |
4602 | ||
4603 | free_mdstat(mdstat); | |
4604 | } | |
4605 | ||
cdddbdbc DW |
4606 | if (rv) { |
4607 | if (devname) | |
7a862a02 | 4608 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
4609 | free_imsm(super); |
4610 | return rv; | |
4611 | } | |
4612 | ||
4613 | st->sb = super; | |
4614 | if (st->ss == NULL) { | |
4615 | st->ss = &super_imsm; | |
4616 | st->minor_version = 0; | |
4617 | st->max_devs = IMSM_MAX_DEVICES; | |
4618 | } | |
8e59f3d8 AK |
4619 | |
4620 | /* load migration record */ | |
2e062e82 AK |
4621 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4622 | /* Check for unsupported migration features */ | |
4623 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 4624 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
4625 | if (devname) |
4626 | fprintf(stderr, " on %s\n", devname); | |
4627 | else | |
4628 | fprintf(stderr, " (IMSM).\n"); | |
4629 | return 3; | |
4630 | } | |
e2f41b2c AK |
4631 | } |
4632 | ||
cdddbdbc DW |
4633 | return 0; |
4634 | } | |
4635 | ||
ef6ffade DW |
4636 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4637 | { | |
4638 | if (info->level == 1) | |
4639 | return 128; | |
4640 | return info->chunk_size >> 9; | |
4641 | } | |
4642 | ||
5551b113 CA |
4643 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4644 | unsigned long long size) | |
fcfd9599 | 4645 | { |
4025c288 | 4646 | if (info->level == 1) |
5551b113 | 4647 | return size * 2; |
4025c288 | 4648 | else |
5551b113 | 4649 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4650 | } |
4651 | ||
4d1313e9 DW |
4652 | static void imsm_update_version_info(struct intel_super *super) |
4653 | { | |
4654 | /* update the version and attributes */ | |
4655 | struct imsm_super *mpb = super->anchor; | |
4656 | char *version; | |
4657 | struct imsm_dev *dev; | |
4658 | struct imsm_map *map; | |
4659 | int i; | |
4660 | ||
4661 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4662 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4663 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4664 | if (__le32_to_cpu(dev->size_high) > 0) |
4665 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4666 | ||
4667 | /* FIXME detect when an array spans a port multiplier */ | |
4668 | #if 0 | |
4669 | mpb->attributes |= MPB_ATTRIB_PM; | |
4670 | #endif | |
4671 | ||
4672 | if (mpb->num_raid_devs > 1 || | |
4673 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4674 | version = MPB_VERSION_ATTRIBS; | |
4675 | switch (get_imsm_raid_level(map)) { | |
4676 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4677 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4678 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4679 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4680 | } | |
4681 | } else { | |
4682 | if (map->num_members >= 5) | |
4683 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4684 | else if (dev->status == DEV_CLONE_N_GO) | |
4685 | version = MPB_VERSION_CNG; | |
4686 | else if (get_imsm_raid_level(map) == 5) | |
4687 | version = MPB_VERSION_RAID5; | |
4688 | else if (map->num_members >= 3) | |
4689 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4690 | else if (get_imsm_raid_level(map) == 1) | |
4691 | version = MPB_VERSION_RAID1; | |
4692 | else | |
4693 | version = MPB_VERSION_RAID0; | |
4694 | } | |
4695 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4696 | } | |
4697 | } | |
4698 | ||
aa534678 DW |
4699 | static int check_name(struct intel_super *super, char *name, int quiet) |
4700 | { | |
4701 | struct imsm_super *mpb = super->anchor; | |
4702 | char *reason = NULL; | |
4703 | int i; | |
4704 | ||
4705 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4706 | reason = "must be 16 characters or less"; | |
4707 | ||
4708 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4709 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4710 | ||
4711 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4712 | reason = "already exists"; | |
4713 | break; | |
4714 | } | |
4715 | } | |
4716 | ||
4717 | if (reason && !quiet) | |
e7b84f9d | 4718 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
4719 | |
4720 | return !reason; | |
4721 | } | |
4722 | ||
8b353278 DW |
4723 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4724 | unsigned long long size, char *name, | |
83cd1e97 N |
4725 | char *homehost, int *uuid, |
4726 | long long data_offset) | |
cdddbdbc | 4727 | { |
c2c087e6 DW |
4728 | /* We are creating a volume inside a pre-existing container. |
4729 | * so st->sb is already set. | |
4730 | */ | |
4731 | struct intel_super *super = st->sb; | |
949c47a0 | 4732 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4733 | struct intel_dev *dv; |
c2c087e6 DW |
4734 | struct imsm_dev *dev; |
4735 | struct imsm_vol *vol; | |
4736 | struct imsm_map *map; | |
4737 | int idx = mpb->num_raid_devs; | |
4738 | int i; | |
4739 | unsigned long long array_blocks; | |
2c092cad | 4740 | size_t size_old, size_new; |
5551b113 | 4741 | unsigned long long num_data_stripes; |
cdddbdbc | 4742 | |
88c32bb1 | 4743 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 4744 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4745 | return 0; |
4746 | } | |
4747 | ||
2c092cad DW |
4748 | /* ensure the mpb is large enough for the new data */ |
4749 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4750 | size_new = disks_to_mpb_size(info->nr_disks); | |
4751 | if (size_new > size_old) { | |
4752 | void *mpb_new; | |
4753 | size_t size_round = ROUND_UP(size_new, 512); | |
4754 | ||
4755 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
e7b84f9d | 4756 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
4757 | return 0; |
4758 | } | |
17a4eaf9 AK |
4759 | if (posix_memalign(&super->migr_rec_buf, 512, |
4760 | MIGR_REC_BUF_SIZE) != 0) { | |
1ade5cc1 | 4761 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4762 | free(super->buf); |
4763 | free(super); | |
ea944c8f | 4764 | free(mpb_new); |
8e59f3d8 AK |
4765 | return 0; |
4766 | } | |
2c092cad DW |
4767 | memcpy(mpb_new, mpb, size_old); |
4768 | free(mpb); | |
4769 | mpb = mpb_new; | |
949c47a0 | 4770 | super->anchor = mpb_new; |
2c092cad DW |
4771 | mpb->mpb_size = __cpu_to_le32(size_new); |
4772 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4773 | } | |
bf5a934a | 4774 | super->current_vol = idx; |
3960e579 DW |
4775 | |
4776 | /* handle 'failed_disks' by either: | |
4777 | * a) create dummy disk entries in the table if this the first | |
4778 | * volume in the array. We add them here as this is the only | |
4779 | * opportunity to add them. add_to_super_imsm_volume() | |
4780 | * handles the non-failed disks and continues incrementing | |
4781 | * mpb->num_disks. | |
4782 | * b) validate that 'failed_disks' matches the current number | |
4783 | * of missing disks if the container is populated | |
d23fe947 | 4784 | */ |
3960e579 | 4785 | if (super->current_vol == 0) { |
d23fe947 | 4786 | mpb->num_disks = 0; |
3960e579 DW |
4787 | for (i = 0; i < info->failed_disks; i++) { |
4788 | struct imsm_disk *disk; | |
4789 | ||
4790 | mpb->num_disks++; | |
4791 | disk = __get_imsm_disk(mpb, i); | |
4792 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4793 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4794 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4795 | "missing:%d", i); | |
4796 | } | |
4797 | find_missing(super); | |
4798 | } else { | |
4799 | int missing = 0; | |
4800 | struct dl *d; | |
4801 | ||
4802 | for (d = super->missing; d; d = d->next) | |
4803 | missing++; | |
4804 | if (info->failed_disks > missing) { | |
e7b84f9d | 4805 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
4806 | return 0; |
4807 | } | |
4808 | } | |
5a038140 | 4809 | |
aa534678 DW |
4810 | if (!check_name(super, name, 0)) |
4811 | return 0; | |
503975b9 N |
4812 | dv = xmalloc(sizeof(*dv)); |
4813 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 4814 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4815 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4816 | info->layout, info->chunk_size, |
5551b113 | 4817 | size * 2); |
979d38be DW |
4818 | /* round array size down to closest MB */ |
4819 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4820 | ||
c2c087e6 DW |
4821 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4822 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4823 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4824 | vol = &dev->vol; |
4825 | vol->migr_state = 0; | |
1484e727 | 4826 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4827 | vol->dirty = !info->state; |
f8f603f1 | 4828 | vol->curr_migr_unit = 0; |
238c0a71 | 4829 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4830 | set_pba_of_lba0(map, super->create_offset); |
4831 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4832 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4833 | map->failed_disk_num = ~0; |
bf4442ab | 4834 | if (info->level > 0) |
fffaf1ff N |
4835 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
4836 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
4837 | else |
4838 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4839 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4840 | map->ddf = 1; |
ef6ffade DW |
4841 | |
4842 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4843 | free(dev); |
4844 | free(dv); | |
7a862a02 | 4845 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
4846 | return 0; |
4847 | } | |
81062a36 DW |
4848 | |
4849 | map->raid_level = info->level; | |
4d1313e9 | 4850 | if (info->level == 10) { |
c2c087e6 | 4851 | map->raid_level = 1; |
4d1313e9 | 4852 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4853 | } else if (info->level == 1) |
4854 | map->num_domains = info->raid_disks; | |
4855 | else | |
ff596308 | 4856 | map->num_domains = 1; |
81062a36 | 4857 | |
5551b113 CA |
4858 | /* info->size is only int so use the 'size' parameter instead */ |
4859 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4860 | num_data_stripes /= map->num_domains; | |
4861 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4862 | |
c2c087e6 DW |
4863 | map->num_members = info->raid_disks; |
4864 | for (i = 0; i < map->num_members; i++) { | |
4865 | /* initialized in add_to_super */ | |
4eb26970 | 4866 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4867 | } |
949c47a0 | 4868 | mpb->num_raid_devs++; |
ba2de7ba DW |
4869 | |
4870 | dv->dev = dev; | |
4871 | dv->index = super->current_vol; | |
4872 | dv->next = super->devlist; | |
4873 | super->devlist = dv; | |
c2c087e6 | 4874 | |
4d1313e9 DW |
4875 | imsm_update_version_info(super); |
4876 | ||
c2c087e6 | 4877 | return 1; |
cdddbdbc DW |
4878 | } |
4879 | ||
bf5a934a DW |
4880 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4881 | unsigned long long size, char *name, | |
83cd1e97 N |
4882 | char *homehost, int *uuid, |
4883 | unsigned long long data_offset) | |
bf5a934a DW |
4884 | { |
4885 | /* This is primarily called by Create when creating a new array. | |
4886 | * We will then get add_to_super called for each component, and then | |
4887 | * write_init_super called to write it out to each device. | |
4888 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4889 | * array. | |
4890 | * To create on a pre-existing array a different method will be called. | |
4891 | * This one is just for fresh drives. | |
4892 | */ | |
4893 | struct intel_super *super; | |
4894 | struct imsm_super *mpb; | |
4895 | size_t mpb_size; | |
4d1313e9 | 4896 | char *version; |
bf5a934a | 4897 | |
83cd1e97 | 4898 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 4899 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
4900 | return 0; |
4901 | } | |
4902 | ||
bf5a934a | 4903 | if (st->sb) |
83cd1e97 N |
4904 | return init_super_imsm_volume(st, info, size, name, homehost, uuid, |
4905 | data_offset); | |
e683ca88 DW |
4906 | |
4907 | if (info) | |
4908 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4909 | else | |
4910 | mpb_size = 512; | |
bf5a934a | 4911 | |
49133e57 | 4912 | super = alloc_super(); |
e683ca88 | 4913 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4914 | free(super); |
e683ca88 DW |
4915 | super = NULL; |
4916 | } | |
4917 | if (!super) { | |
1ade5cc1 | 4918 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
4919 | return 0; |
4920 | } | |
17a4eaf9 | 4921 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
1ade5cc1 | 4922 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4923 | free(super->buf); |
4924 | free(super); | |
4925 | return 0; | |
4926 | } | |
e683ca88 | 4927 | memset(super->buf, 0, mpb_size); |
ef649044 | 4928 | mpb = super->buf; |
e683ca88 DW |
4929 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4930 | st->sb = super; | |
4931 | ||
4932 | if (info == NULL) { | |
4933 | /* zeroing superblock */ | |
4934 | return 0; | |
4935 | } | |
bf5a934a | 4936 | |
4d1313e9 DW |
4937 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4938 | ||
4939 | version = (char *) mpb->sig; | |
4940 | strcpy(version, MPB_SIGNATURE); | |
4941 | version += strlen(MPB_SIGNATURE); | |
4942 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4943 | |
bf5a934a DW |
4944 | return 1; |
4945 | } | |
4946 | ||
0e600426 | 4947 | #ifndef MDASSEMBLE |
f20c3968 | 4948 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4949 | int fd, char *devname) |
4950 | { | |
4951 | struct intel_super *super = st->sb; | |
d23fe947 | 4952 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4953 | struct imsm_disk *_disk; |
bf5a934a DW |
4954 | struct imsm_dev *dev; |
4955 | struct imsm_map *map; | |
3960e579 | 4956 | struct dl *dl, *df; |
4eb26970 | 4957 | int slot; |
bf5a934a | 4958 | |
949c47a0 | 4959 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4960 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4961 | |
208933a7 | 4962 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 4963 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
4964 | devname); |
4965 | return 1; | |
4966 | } | |
4967 | ||
efb30e7f DW |
4968 | if (fd == -1) { |
4969 | /* we're doing autolayout so grab the pre-marked (in | |
4970 | * validate_geometry) raid_disk | |
4971 | */ | |
4972 | for (dl = super->disks; dl; dl = dl->next) | |
4973 | if (dl->raiddisk == dk->raid_disk) | |
4974 | break; | |
4975 | } else { | |
4976 | for (dl = super->disks; dl ; dl = dl->next) | |
4977 | if (dl->major == dk->major && | |
4978 | dl->minor == dk->minor) | |
4979 | break; | |
4980 | } | |
d23fe947 | 4981 | |
208933a7 | 4982 | if (!dl) { |
e7b84f9d | 4983 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 4984 | return 1; |
208933a7 | 4985 | } |
bf5a934a | 4986 | |
d23fe947 DW |
4987 | /* add a pristine spare to the metadata */ |
4988 | if (dl->index < 0) { | |
4989 | dl->index = super->anchor->num_disks; | |
4990 | super->anchor->num_disks++; | |
4991 | } | |
4eb26970 DW |
4992 | /* Check the device has not already been added */ |
4993 | slot = get_imsm_disk_slot(map, dl->index); | |
4994 | if (slot >= 0 && | |
238c0a71 | 4995 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 4996 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
4997 | devname); |
4998 | return 1; | |
4999 | } | |
656b6b5a | 5000 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5001 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5002 | |
3960e579 DW |
5003 | /* update size of 'missing' disks to be at least as large as the |
5004 | * largest acitve member (we only have dummy missing disks when | |
5005 | * creating the first volume) | |
5006 | */ | |
5007 | if (super->current_vol == 0) { | |
5008 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5009 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5010 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5011 | _disk = __get_imsm_disk(mpb, df->index); |
5012 | *_disk = df->disk; | |
5013 | } | |
5014 | } | |
5015 | ||
5016 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5017 | for (df = super->missing; df; df = df->next) | |
5018 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5019 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5020 | |
5021 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5022 | continue; | |
5023 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5024 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5025 | struct imsm_map *map2 = get_imsm_map(dev, |
5026 | MAP_1); | |
0a108d63 AK |
5027 | int slot2 = get_imsm_disk_slot(map2, df->index); |
5028 | if ((slot2 < map2->num_members) && | |
5029 | (slot2 >= 0)) { | |
1ace8403 | 5030 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5031 | slot2, |
5032 | MAP_1); | |
1ace8403 AK |
5033 | if ((unsigned)df->index == |
5034 | ord_to_idx(ord2)) | |
5035 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5036 | slot2, |
1ace8403 AK |
5037 | df->index | |
5038 | IMSM_ORD_REBUILD); | |
5039 | } | |
5040 | } | |
3960e579 DW |
5041 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5042 | break; | |
5043 | } | |
5044 | ||
d23fe947 DW |
5045 | /* if we are creating the first raid device update the family number */ |
5046 | if (super->current_vol == 0) { | |
5047 | __u32 sum; | |
5048 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5049 | |
3960e579 | 5050 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5051 | if (!_dev || !_disk) { |
e7b84f9d | 5052 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5053 | return 1; |
5054 | } | |
d23fe947 DW |
5055 | *_dev = *dev; |
5056 | *_disk = dl->disk; | |
148acb7b DW |
5057 | sum = random32(); |
5058 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5059 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5060 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5061 | } |
ca0748fa | 5062 | super->current_disk = dl; |
f20c3968 | 5063 | return 0; |
bf5a934a DW |
5064 | } |
5065 | ||
a8619d23 AK |
5066 | /* mark_spare() |
5067 | * Function marks disk as spare and restores disk serial | |
5068 | * in case it was previously marked as failed by takeover operation | |
5069 | * reruns: | |
5070 | * -1 : critical error | |
5071 | * 0 : disk is marked as spare but serial is not set | |
5072 | * 1 : success | |
5073 | */ | |
5074 | int mark_spare(struct dl *disk) | |
5075 | { | |
5076 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5077 | int ret_val = -1; | |
5078 | ||
5079 | if (!disk) | |
5080 | return ret_val; | |
5081 | ||
5082 | ret_val = 0; | |
5083 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5084 | /* Restore disk serial number, because takeover marks disk | |
5085 | * as failed and adds to serial ':0' before it becomes | |
5086 | * a spare disk. | |
5087 | */ | |
5088 | serialcpy(disk->serial, serial); | |
5089 | serialcpy(disk->disk.serial, serial); | |
5090 | ret_val = 1; | |
5091 | } | |
5092 | disk->disk.status = SPARE_DISK; | |
5093 | disk->index = -1; | |
5094 | ||
5095 | return ret_val; | |
5096 | } | |
88654014 | 5097 | |
f20c3968 | 5098 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5099 | int fd, char *devname, |
5100 | unsigned long long data_offset) | |
cdddbdbc | 5101 | { |
c2c087e6 | 5102 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5103 | struct dl *dd; |
5104 | unsigned long long size; | |
f2f27e63 | 5105 | __u32 id; |
c2c087e6 DW |
5106 | int rv; |
5107 | struct stat stb; | |
5108 | ||
88654014 LM |
5109 | /* If we are on an RAID enabled platform check that the disk is |
5110 | * attached to the raid controller. | |
5111 | * We do not need to test disks attachment for container based additions, | |
5112 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5113 | */ |
d424212e | 5114 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5115 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5116 | if (rv != 0) { |
5117 | dprintf("capability: %p fd: %d ret: %d\n", | |
5118 | super->orom, fd, rv); | |
5119 | return 1; | |
88c32bb1 DW |
5120 | } |
5121 | ||
f20c3968 DW |
5122 | if (super->current_vol >= 0) |
5123 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5124 | |
c2c087e6 | 5125 | fstat(fd, &stb); |
503975b9 | 5126 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5127 | dd->major = major(stb.st_rdev); |
5128 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5129 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5130 | dd->fd = fd; |
689c9bf3 | 5131 | dd->e = NULL; |
1a64be56 | 5132 | dd->action = DISK_ADD; |
c2c087e6 | 5133 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5134 | if (rv) { |
e7b84f9d | 5135 | pr_err("failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 5136 | free(dd); |
0030e8d6 | 5137 | abort(); |
c2c087e6 DW |
5138 | } |
5139 | ||
c2c087e6 | 5140 | get_dev_size(fd, NULL, &size); |
71e5411e PB |
5141 | /* clear migr_rec when adding disk to container */ |
5142 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); | |
5143 | if (lseek64(fd, size - MIGR_REC_POSITION, SEEK_SET) >= 0) { | |
5144 | if (write(fd, super->migr_rec_buf, | |
5145 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
5146 | perror("Write migr_rec failed"); | |
5147 | } | |
5148 | ||
c2c087e6 | 5149 | size /= 512; |
1f24f035 | 5150 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5151 | set_total_blocks(&dd->disk, size); |
5152 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5153 | struct imsm_super *mpb = super->anchor; | |
5154 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5155 | } | |
a8619d23 | 5156 | mark_spare(dd); |
c2c087e6 | 5157 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5158 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5159 | else |
b9f594fe | 5160 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5161 | |
5162 | if (st->update_tail) { | |
1a64be56 LM |
5163 | dd->next = super->disk_mgmt_list; |
5164 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5165 | } else { |
5166 | dd->next = super->disks; | |
5167 | super->disks = dd; | |
ceaf0ee1 | 5168 | super->updates_pending++; |
43dad3d6 | 5169 | } |
f20c3968 DW |
5170 | |
5171 | return 0; | |
cdddbdbc DW |
5172 | } |
5173 | ||
1a64be56 LM |
5174 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5175 | { | |
5176 | struct intel_super *super = st->sb; | |
5177 | struct dl *dd; | |
5178 | ||
5179 | /* remove from super works only in mdmon - for communication | |
5180 | * manager - monitor. Check if communication memory buffer | |
5181 | * is prepared. | |
5182 | */ | |
5183 | if (!st->update_tail) { | |
1ade5cc1 | 5184 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5185 | return 1; |
5186 | } | |
503975b9 | 5187 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5188 | dd->major = dk->major; |
5189 | dd->minor = dk->minor; | |
1a64be56 | 5190 | dd->fd = -1; |
a8619d23 | 5191 | mark_spare(dd); |
1a64be56 LM |
5192 | dd->action = DISK_REMOVE; |
5193 | ||
5194 | dd->next = super->disk_mgmt_list; | |
5195 | super->disk_mgmt_list = dd; | |
5196 | ||
1a64be56 LM |
5197 | return 0; |
5198 | } | |
5199 | ||
f796af5d DW |
5200 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5201 | ||
5202 | static union { | |
5203 | char buf[512]; | |
5204 | struct imsm_super anchor; | |
5205 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5206 | |
d23fe947 DW |
5207 | /* spare records have their own family number and do not have any defined raid |
5208 | * devices | |
5209 | */ | |
5210 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5211 | { | |
d23fe947 | 5212 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5213 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5214 | __u32 sum; |
5215 | struct dl *d; | |
5216 | ||
68641cdb JS |
5217 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5218 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5219 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5220 | spare->num_disks = 1; |
5221 | spare->num_raid_devs = 0; | |
5222 | spare->cache_size = mpb->cache_size; | |
5223 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5224 | |
5225 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5226 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5227 | |
5228 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5229 | if (d->index != -1) |
d23fe947 DW |
5230 | continue; |
5231 | ||
f796af5d | 5232 | spare->disk[0] = d->disk; |
027c374f CA |
5233 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5234 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5235 | ||
f796af5d DW |
5236 | sum = __gen_imsm_checksum(spare); |
5237 | spare->family_num = __cpu_to_le32(sum); | |
5238 | spare->orig_family_num = 0; | |
5239 | sum = __gen_imsm_checksum(spare); | |
5240 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5241 | |
f796af5d | 5242 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5243 | pr_err("failed for device %d:%d %s\n", |
5244 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5245 | return 1; |
d23fe947 DW |
5246 | } |
5247 | if (doclose) { | |
5248 | close(d->fd); | |
5249 | d->fd = -1; | |
5250 | } | |
5251 | } | |
5252 | ||
e74255d9 | 5253 | return 0; |
d23fe947 DW |
5254 | } |
5255 | ||
36988a3d | 5256 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5257 | { |
36988a3d | 5258 | struct intel_super *super = st->sb; |
949c47a0 | 5259 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5260 | struct dl *d; |
5261 | __u32 generation; | |
5262 | __u32 sum; | |
d23fe947 | 5263 | int spares = 0; |
949c47a0 | 5264 | int i; |
a48ac0a8 | 5265 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5266 | int num_disks = 0; |
146c6260 | 5267 | int clear_migration_record = 1; |
cdddbdbc | 5268 | |
c2c087e6 DW |
5269 | /* 'generation' is incremented everytime the metadata is written */ |
5270 | generation = __le32_to_cpu(mpb->generation_num); | |
5271 | generation++; | |
5272 | mpb->generation_num = __cpu_to_le32(generation); | |
5273 | ||
148acb7b DW |
5274 | /* fix up cases where previous mdadm releases failed to set |
5275 | * orig_family_num | |
5276 | */ | |
5277 | if (mpb->orig_family_num == 0) | |
5278 | mpb->orig_family_num = mpb->family_num; | |
5279 | ||
d23fe947 | 5280 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5281 | if (d->index == -1) |
d23fe947 | 5282 | spares++; |
36988a3d | 5283 | else { |
d23fe947 | 5284 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5285 | num_disks++; |
5286 | } | |
d23fe947 | 5287 | } |
36988a3d | 5288 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5289 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5290 | num_disks++; |
5291 | } | |
5292 | mpb->num_disks = num_disks; | |
5293 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5294 | |
949c47a0 DW |
5295 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5296 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5297 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5298 | if (dev && dev2) { | |
5299 | imsm_copy_dev(dev, dev2); | |
5300 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5301 | } | |
146c6260 AK |
5302 | if (is_gen_migration(dev2)) |
5303 | clear_migration_record = 0; | |
949c47a0 | 5304 | } |
a48ac0a8 DW |
5305 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5306 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5307 | |
c2c087e6 | 5308 | /* recalculate checksum */ |
949c47a0 | 5309 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5310 | mpb->check_sum = __cpu_to_le32(sum); |
5311 | ||
51d83f5d AK |
5312 | if (super->clean_migration_record_by_mdmon) { |
5313 | clear_migration_record = 1; | |
5314 | super->clean_migration_record_by_mdmon = 0; | |
5315 | } | |
146c6260 | 5316 | if (clear_migration_record) |
17a4eaf9 | 5317 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5318 | |
d23fe947 | 5319 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5320 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5321 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5322 | continue; |
30602f53 | 5323 | |
146c6260 AK |
5324 | if (clear_migration_record) { |
5325 | unsigned long long dsize; | |
5326 | ||
5327 | get_dev_size(d->fd, NULL, &dsize); | |
5328 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5329 | if (write(d->fd, super->migr_rec_buf, |
5330 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5331 | perror("Write migr_rec failed"); |
146c6260 AK |
5332 | } |
5333 | } | |
51d83f5d AK |
5334 | |
5335 | if (store_imsm_mpb(d->fd, mpb)) | |
5336 | fprintf(stderr, | |
1ade5cc1 N |
5337 | "failed for device %d:%d (fd: %d)%s\n", |
5338 | d->major, d->minor, | |
51d83f5d AK |
5339 | d->fd, strerror(errno)); |
5340 | ||
c2c087e6 DW |
5341 | if (doclose) { |
5342 | close(d->fd); | |
5343 | d->fd = -1; | |
5344 | } | |
5345 | } | |
5346 | ||
d23fe947 DW |
5347 | if (spares) |
5348 | return write_super_imsm_spares(super, doclose); | |
5349 | ||
e74255d9 | 5350 | return 0; |
c2c087e6 DW |
5351 | } |
5352 | ||
9b1fb677 | 5353 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5354 | { |
5355 | size_t len; | |
5356 | struct imsm_update_create_array *u; | |
5357 | struct intel_super *super = st->sb; | |
9b1fb677 | 5358 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5359 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5360 | struct disk_info *inf; |
5361 | struct imsm_disk *disk; | |
5362 | int i; | |
43dad3d6 | 5363 | |
54c2c1ea DW |
5364 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5365 | sizeof(*inf) * map->num_members; | |
503975b9 | 5366 | u = xmalloc(len); |
43dad3d6 | 5367 | u->type = update_create_array; |
9b1fb677 | 5368 | u->dev_idx = dev_idx; |
43dad3d6 | 5369 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5370 | inf = get_disk_info(u); |
5371 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5372 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5373 | |
54c2c1ea | 5374 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
5375 | if (!disk) |
5376 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
5377 | serialcpy(inf[i].serial, disk->serial); |
5378 | } | |
43dad3d6 DW |
5379 | append_metadata_update(st, u, len); |
5380 | ||
5381 | return 0; | |
5382 | } | |
5383 | ||
1a64be56 | 5384 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5385 | { |
5386 | struct intel_super *super = st->sb; | |
5387 | size_t len; | |
1a64be56 | 5388 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5389 | |
1a64be56 | 5390 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5391 | return 0; |
5392 | ||
5393 | len = sizeof(*u); | |
503975b9 | 5394 | u = xmalloc(len); |
1a64be56 | 5395 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5396 | append_metadata_update(st, u, len); |
5397 | ||
5398 | return 0; | |
5399 | } | |
5400 | ||
c2c087e6 DW |
5401 | static int write_init_super_imsm(struct supertype *st) |
5402 | { | |
9b1fb677 DW |
5403 | struct intel_super *super = st->sb; |
5404 | int current_vol = super->current_vol; | |
5405 | ||
5406 | /* we are done with current_vol reset it to point st at the container */ | |
5407 | super->current_vol = -1; | |
5408 | ||
8273f55e | 5409 | if (st->update_tail) { |
43dad3d6 DW |
5410 | /* queue the recently created array / added disk |
5411 | * as a metadata update */ | |
43dad3d6 | 5412 | int rv; |
8273f55e | 5413 | |
43dad3d6 | 5414 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5415 | if (current_vol < 0) { |
1a64be56 LM |
5416 | /* in the mgmt (add/remove) disk case we are running |
5417 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5418 | */ |
1a64be56 | 5419 | return mgmt_disk(st); |
43dad3d6 | 5420 | } else |
9b1fb677 | 5421 | rv = create_array(st, current_vol); |
8273f55e | 5422 | |
43dad3d6 | 5423 | return rv; |
d682f344 N |
5424 | } else { |
5425 | struct dl *d; | |
5426 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5427 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5428 | return write_super_imsm(st, 1); |
d682f344 | 5429 | } |
cdddbdbc | 5430 | } |
0e600426 | 5431 | #endif |
cdddbdbc | 5432 | |
e683ca88 | 5433 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5434 | { |
e683ca88 DW |
5435 | struct intel_super *super = st->sb; |
5436 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5437 | |
e683ca88 | 5438 | if (!mpb) |
ad97895e DW |
5439 | return 1; |
5440 | ||
1799c9e8 | 5441 | #ifndef MDASSEMBLE |
e683ca88 | 5442 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5443 | #else |
5444 | return 1; | |
5445 | #endif | |
cdddbdbc DW |
5446 | } |
5447 | ||
0e600426 N |
5448 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5449 | { | |
5450 | return __le32_to_cpu(mpb->bbm_log_size); | |
5451 | } | |
5452 | ||
5453 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5454 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5455 | int layout, int raiddisks, int chunk, | |
af4348dd N |
5456 | unsigned long long size, |
5457 | unsigned long long data_offset, | |
5458 | char *dev, | |
2c514b71 NB |
5459 | unsigned long long *freesize, |
5460 | int verbose) | |
cdddbdbc | 5461 | { |
c2c087e6 DW |
5462 | int fd; |
5463 | unsigned long long ldsize; | |
594dc1b8 | 5464 | struct intel_super *super; |
f2f5c343 | 5465 | int rv = 0; |
cdddbdbc | 5466 | |
c2c087e6 DW |
5467 | if (level != LEVEL_CONTAINER) |
5468 | return 0; | |
5469 | if (!dev) | |
5470 | return 1; | |
5471 | ||
5472 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5473 | if (fd < 0) { | |
ba728be7 | 5474 | if (verbose > 0) |
e7b84f9d | 5475 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 5476 | dev, strerror(errno)); |
c2c087e6 DW |
5477 | return 0; |
5478 | } | |
5479 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5480 | close(fd); | |
5481 | return 0; | |
5482 | } | |
f2f5c343 LM |
5483 | |
5484 | /* capabilities retrieve could be possible | |
5485 | * note that there is no fd for the disks in array. | |
5486 | */ | |
5487 | super = alloc_super(); | |
ba728be7 | 5488 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
5489 | if (rv != 0) { |
5490 | #if DEBUG | |
5491 | char str[256]; | |
5492 | fd2devname(fd, str); | |
1ade5cc1 | 5493 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
5494 | fd, str, super->orom, rv, raiddisks); |
5495 | #endif | |
5496 | /* no orom/efi or non-intel hba of the disk */ | |
5497 | close(fd); | |
5498 | free_imsm(super); | |
5499 | return 0; | |
5500 | } | |
c2c087e6 | 5501 | close(fd); |
9126b9a8 CA |
5502 | if (super->orom) { |
5503 | if (raiddisks > super->orom->tds) { | |
5504 | if (verbose) | |
7a862a02 | 5505 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
5506 | raiddisks, super->orom->tds); |
5507 | free_imsm(super); | |
5508 | return 0; | |
5509 | } | |
5510 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5511 | (ldsize >> 9) >> 32 > 0) { | |
5512 | if (verbose) | |
e7b84f9d | 5513 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
5514 | free_imsm(super); |
5515 | return 0; | |
5516 | } | |
f2f5c343 | 5517 | } |
c2c087e6 | 5518 | |
af4348dd | 5519 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 5520 | free_imsm(super); |
c2c087e6 DW |
5521 | |
5522 | return 1; | |
cdddbdbc DW |
5523 | } |
5524 | ||
0dcecb2e DW |
5525 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5526 | { | |
5527 | const unsigned long long base_start = e[*idx].start; | |
5528 | unsigned long long end = base_start + e[*idx].size; | |
5529 | int i; | |
5530 | ||
5531 | if (base_start == end) | |
5532 | return 0; | |
5533 | ||
5534 | *idx = *idx + 1; | |
5535 | for (i = *idx; i < num_extents; i++) { | |
5536 | /* extend overlapping extents */ | |
5537 | if (e[i].start >= base_start && | |
5538 | e[i].start <= end) { | |
5539 | if (e[i].size == 0) | |
5540 | return 0; | |
5541 | if (e[i].start + e[i].size > end) | |
5542 | end = e[i].start + e[i].size; | |
5543 | } else if (e[i].start > end) { | |
5544 | *idx = i; | |
5545 | break; | |
5546 | } | |
5547 | } | |
5548 | ||
5549 | return end - base_start; | |
5550 | } | |
5551 | ||
5552 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5553 | { | |
5554 | /* build a composite disk with all known extents and generate a new | |
5555 | * 'maxsize' given the "all disks in an array must share a common start | |
5556 | * offset" constraint | |
5557 | */ | |
503975b9 | 5558 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
5559 | struct dl *dl; |
5560 | int i, j; | |
5561 | int start_extent; | |
5562 | unsigned long long pos; | |
b9d77223 | 5563 | unsigned long long start = 0; |
0dcecb2e DW |
5564 | unsigned long long maxsize; |
5565 | unsigned long reserve; | |
5566 | ||
0dcecb2e DW |
5567 | /* coalesce and sort all extents. also, check to see if we need to |
5568 | * reserve space between member arrays | |
5569 | */ | |
5570 | j = 0; | |
5571 | for (dl = super->disks; dl; dl = dl->next) { | |
5572 | if (!dl->e) | |
5573 | continue; | |
5574 | for (i = 0; i < dl->extent_cnt; i++) | |
5575 | e[j++] = dl->e[i]; | |
5576 | } | |
5577 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5578 | ||
5579 | /* merge extents */ | |
5580 | i = 0; | |
5581 | j = 0; | |
5582 | while (i < sum_extents) { | |
5583 | e[j].start = e[i].start; | |
5584 | e[j].size = find_size(e, &i, sum_extents); | |
5585 | j++; | |
5586 | if (e[j-1].size == 0) | |
5587 | break; | |
5588 | } | |
5589 | ||
5590 | pos = 0; | |
5591 | maxsize = 0; | |
5592 | start_extent = 0; | |
5593 | i = 0; | |
5594 | do { | |
5595 | unsigned long long esize; | |
5596 | ||
5597 | esize = e[i].start - pos; | |
5598 | if (esize >= maxsize) { | |
5599 | maxsize = esize; | |
5600 | start = pos; | |
5601 | start_extent = i; | |
5602 | } | |
5603 | pos = e[i].start + e[i].size; | |
5604 | i++; | |
5605 | } while (e[i-1].size); | |
5606 | free(e); | |
5607 | ||
a7dd165b DW |
5608 | if (maxsize == 0) |
5609 | return 0; | |
5610 | ||
5611 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5612 | * container | |
5613 | */ | |
0dcecb2e DW |
5614 | if (start_extent > 0) |
5615 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5616 | else | |
5617 | reserve = 0; | |
5618 | ||
5619 | if (maxsize < reserve) | |
a7dd165b | 5620 | return 0; |
0dcecb2e | 5621 | |
5551b113 | 5622 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5623 | if (start + reserve > super->create_offset) |
a7dd165b | 5624 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5625 | super->create_offset = start + reserve; |
5626 | ||
5627 | return maxsize - reserve; | |
5628 | } | |
5629 | ||
88c32bb1 DW |
5630 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5631 | { | |
5632 | if (level < 0 || level == 6 || level == 4) | |
5633 | return 0; | |
5634 | ||
5635 | /* if we have an orom prevent invalid raid levels */ | |
5636 | if (orom) | |
5637 | switch (level) { | |
5638 | case 0: return imsm_orom_has_raid0(orom); | |
5639 | case 1: | |
5640 | if (raiddisks > 2) | |
5641 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5642 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5643 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5644 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5645 | } |
5646 | else | |
5647 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5648 | ||
5649 | return 0; | |
5650 | } | |
5651 | ||
ca9de185 LM |
5652 | static int |
5653 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5654 | int dpa, int verbose) | |
5655 | { | |
5656 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 5657 | struct mdstat_ent *memb; |
ca9de185 LM |
5658 | int count = 0; |
5659 | int num = 0; | |
594dc1b8 | 5660 | struct md_list *dv; |
ca9de185 LM |
5661 | int found; |
5662 | ||
5663 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5664 | if (memb->metadata_version && | |
5665 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5666 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5667 | !is_subarray(memb->metadata_version+9) && | |
5668 | memb->members) { | |
5669 | struct dev_member *dev = memb->members; | |
5670 | int fd = -1; | |
5671 | while(dev && (fd < 0)) { | |
503975b9 N |
5672 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
5673 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5674 | if (num > 0) | |
5675 | fd = open(path, O_RDONLY, 0); | |
5676 | if ((num <= 0) || (fd < 0)) { | |
5677 | pr_vrb(": Cannot open %s: %s\n", | |
5678 | dev->name, strerror(errno)); | |
ca9de185 | 5679 | } |
503975b9 | 5680 | free(path); |
ca9de185 LM |
5681 | dev = dev->next; |
5682 | } | |
5683 | found = 0; | |
5684 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5685 | struct mdstat_ent *vol; | |
5686 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5687 | if ((vol->active > 0) && | |
5688 | vol->metadata_version && | |
9581efb1 | 5689 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
5690 | found++; |
5691 | count++; | |
5692 | } | |
5693 | } | |
5694 | if (*devlist && (found < dpa)) { | |
503975b9 | 5695 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
5696 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
5697 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
5698 | dv->found = found; |
5699 | dv->used = 0; | |
5700 | dv->next = *devlist; | |
5701 | *devlist = dv; | |
ca9de185 LM |
5702 | } |
5703 | } | |
5704 | if (fd >= 0) | |
5705 | close(fd); | |
5706 | } | |
5707 | } | |
5708 | free_mdstat(mdstat); | |
5709 | return count; | |
5710 | } | |
5711 | ||
5712 | #ifdef DEBUG_LOOP | |
5713 | static struct md_list* | |
5714 | get_loop_devices(void) | |
5715 | { | |
5716 | int i; | |
5717 | struct md_list *devlist = NULL; | |
594dc1b8 | 5718 | struct md_list *dv; |
ca9de185 LM |
5719 | |
5720 | for(i = 0; i < 12; i++) { | |
503975b9 N |
5721 | dv = xcalloc(1, sizeof(*dv)); |
5722 | dv->devname = xmalloc(40); | |
ca9de185 LM |
5723 | sprintf(dv->devname, "/dev/loop%d", i); |
5724 | dv->next = devlist; | |
5725 | devlist = dv; | |
5726 | } | |
5727 | return devlist; | |
5728 | } | |
5729 | #endif | |
5730 | ||
5731 | static struct md_list* | |
5732 | get_devices(const char *hba_path) | |
5733 | { | |
5734 | struct md_list *devlist = NULL; | |
594dc1b8 | 5735 | struct md_list *dv; |
ca9de185 LM |
5736 | struct dirent *ent; |
5737 | DIR *dir; | |
5738 | int err = 0; | |
5739 | ||
5740 | #if DEBUG_LOOP | |
5741 | devlist = get_loop_devices(); | |
5742 | return devlist; | |
5743 | #endif | |
5744 | /* scroll through /sys/dev/block looking for devices attached to | |
5745 | * this hba | |
5746 | */ | |
5747 | dir = opendir("/sys/dev/block"); | |
5748 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5749 | int fd; | |
5750 | char buf[1024]; | |
5751 | int major, minor; | |
5752 | char *path = NULL; | |
5753 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5754 | continue; | |
5755 | path = devt_to_devpath(makedev(major, minor)); | |
5756 | if (!path) | |
5757 | continue; | |
5758 | if (!path_attached_to_hba(path, hba_path)) { | |
5759 | free(path); | |
5760 | path = NULL; | |
5761 | continue; | |
5762 | } | |
5763 | free(path); | |
5764 | path = NULL; | |
5765 | fd = dev_open(ent->d_name, O_RDONLY); | |
5766 | if (fd >= 0) { | |
5767 | fd2devname(fd, buf); | |
5768 | close(fd); | |
5769 | } else { | |
e7b84f9d | 5770 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
5771 | ent->d_name); |
5772 | continue; | |
5773 | } | |
5774 | ||
503975b9 N |
5775 | dv = xcalloc(1, sizeof(*dv)); |
5776 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
5777 | dv->next = devlist; |
5778 | devlist = dv; | |
5779 | } | |
5780 | if (err) { | |
5781 | while(devlist) { | |
5782 | dv = devlist; | |
5783 | devlist = devlist->next; | |
5784 | free(dv->devname); | |
5785 | free(dv); | |
5786 | } | |
5787 | } | |
562aa102 | 5788 | closedir(dir); |
ca9de185 LM |
5789 | return devlist; |
5790 | } | |
5791 | ||
5792 | static int | |
5793 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5794 | int verbose, int *found) | |
5795 | { | |
5796 | struct md_list *tmpdev; | |
5797 | int count = 0; | |
594dc1b8 | 5798 | struct supertype *st; |
ca9de185 LM |
5799 | |
5800 | /* first walk the list of devices to find a consistent set | |
5801 | * that match the criterea, if that is possible. | |
5802 | * We flag the ones we like with 'used'. | |
5803 | */ | |
5804 | *found = 0; | |
5805 | st = match_metadata_desc_imsm("imsm"); | |
5806 | if (st == NULL) { | |
5807 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5808 | return 0; | |
5809 | } | |
5810 | ||
5811 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5812 | char *devname = tmpdev->devname; | |
5813 | struct stat stb; | |
5814 | struct supertype *tst; | |
5815 | int dfd; | |
5816 | if (tmpdev->used > 1) | |
5817 | continue; | |
5818 | tst = dup_super(st); | |
5819 | if (tst == NULL) { | |
5820 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5821 | goto err_1; | |
5822 | } | |
5823 | tmpdev->container = 0; | |
5824 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5825 | if (dfd < 0) { | |
1ade5cc1 | 5826 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
5827 | devname, strerror(errno)); |
5828 | tmpdev->used = 2; | |
5829 | } else if (fstat(dfd, &stb)< 0) { | |
5830 | /* Impossible! */ | |
1ade5cc1 | 5831 | dprintf("fstat failed for %s: %s\n", |
ca9de185 LM |
5832 | devname, strerror(errno)); |
5833 | tmpdev->used = 2; | |
5834 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
1ade5cc1 | 5835 | dprintf("%s is not a block device.\n", |
ca9de185 LM |
5836 | devname); |
5837 | tmpdev->used = 2; | |
5838 | } else if (must_be_container(dfd)) { | |
5839 | struct supertype *cst; | |
5840 | cst = super_by_fd(dfd, NULL); | |
5841 | if (cst == NULL) { | |
1ade5cc1 | 5842 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
5843 | devname); |
5844 | tmpdev->used = 2; | |
5845 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 5846 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
5847 | devname); |
5848 | tmpdev->used = 2; | |
5849 | } else if (!tst->ss->load_container || | |
5850 | tst->ss->load_container(tst, dfd, NULL)) | |
5851 | tmpdev->used = 2; | |
5852 | else { | |
5853 | tmpdev->container = 1; | |
5854 | } | |
5855 | if (cst) | |
5856 | cst->ss->free_super(cst); | |
5857 | } else { | |
5858 | tmpdev->st_rdev = stb.st_rdev; | |
5859 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
1ade5cc1 | 5860 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
5861 | devname); |
5862 | tmpdev->used = 2; | |
5863 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 5864 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
5865 | tst->ss->name, devname); |
5866 | tmpdev->used = 2; | |
5867 | } | |
5868 | } | |
5869 | if (dfd >= 0) | |
5870 | close(dfd); | |
5871 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5872 | /* Ignore unrecognised devices during auto-assembly */ | |
5873 | goto loop; | |
5874 | } | |
5875 | else { | |
5876 | struct mdinfo info; | |
5877 | tst->ss->getinfo_super(tst, &info, NULL); | |
5878 | ||
5879 | if (st->minor_version == -1) | |
5880 | st->minor_version = tst->minor_version; | |
5881 | ||
5882 | if (memcmp(info.uuid, uuid_zero, | |
5883 | sizeof(int[4])) == 0) { | |
5884 | /* this is a floating spare. It cannot define | |
5885 | * an array unless there are no more arrays of | |
5886 | * this type to be found. It can be included | |
5887 | * in an array of this type though. | |
5888 | */ | |
5889 | tmpdev->used = 3; | |
5890 | goto loop; | |
5891 | } | |
5892 | ||
5893 | if (st->ss != tst->ss || | |
5894 | st->minor_version != tst->minor_version || | |
5895 | st->ss->compare_super(st, tst) != 0) { | |
5896 | /* Some mismatch. If exactly one array matches this host, | |
5897 | * we can resolve on that one. | |
5898 | * Or, if we are auto assembling, we just ignore the second | |
5899 | * for now. | |
5900 | */ | |
1ade5cc1 | 5901 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
5902 | devname); |
5903 | goto loop; | |
5904 | } | |
5905 | tmpdev->used = 1; | |
5906 | *found = 1; | |
5907 | dprintf("found: devname: %s\n", devname); | |
5908 | } | |
5909 | loop: | |
5910 | if (tst) | |
5911 | tst->ss->free_super(tst); | |
5912 | } | |
5913 | if (*found != 0) { | |
5914 | int err; | |
5915 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5916 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5917 | for (iter = head; iter; iter = iter->next) { | |
5918 | dprintf("content->text_version: %s vol\n", | |
5919 | iter->text_version); | |
5920 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5921 | /* do not assemble arrays with unsupported | |
5922 | configurations */ | |
1ade5cc1 | 5923 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
5924 | iter->text_version); |
5925 | } else | |
5926 | count++; | |
5927 | } | |
5928 | sysfs_free(head); | |
5929 | ||
5930 | } else { | |
1ade5cc1 | 5931 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
5932 | err, st->sb); |
5933 | } | |
5934 | } else { | |
1ade5cc1 | 5935 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
5936 | } |
5937 | ||
5938 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5939 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5940 | if (count) { | |
5941 | if (count < tmpdev->found) | |
5942 | count = 0; | |
5943 | else | |
5944 | count -= tmpdev->found; | |
5945 | } | |
5946 | } | |
5947 | if (tmpdev->used == 1) | |
5948 | tmpdev->used = 4; | |
5949 | } | |
5950 | err_1: | |
5951 | if (st) | |
5952 | st->ss->free_super(st); | |
5953 | return count; | |
5954 | } | |
5955 | ||
ca9de185 | 5956 | static int |
72a45777 | 5957 | count_volumes(struct intel_hba *hba, int dpa, int verbose) |
ca9de185 | 5958 | { |
72a45777 | 5959 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 5960 | int count = 0; |
72a45777 PB |
5961 | const struct orom_entry *entry; |
5962 | struct devid_list *dv, *devid_list; | |
ca9de185 | 5963 | |
72a45777 | 5964 | if (!hba || !hba->path) |
ca9de185 LM |
5965 | return 0; |
5966 | ||
72a45777 PB |
5967 | for (idev = intel_devices; idev; idev = idev->next) { |
5968 | if (strstr(idev->path, hba->path)) | |
5969 | break; | |
5970 | } | |
5971 | ||
5972 | if (!idev || !idev->dev_id) | |
ca9de185 | 5973 | return 0; |
72a45777 PB |
5974 | |
5975 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
5976 | ||
5977 | if (!entry || !entry->devid_list) | |
5978 | return 0; | |
5979 | ||
5980 | devid_list = entry->devid_list; | |
5981 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 5982 | struct md_list *devlist; |
72a45777 PB |
5983 | struct sys_dev *device = device_by_id(dv->devid); |
5984 | char *hba_path; | |
5985 | int found = 0; | |
5986 | ||
5987 | if (device) | |
5988 | hba_path = device->path; | |
5989 | else | |
5990 | return 0; | |
5991 | ||
60f0f54d PB |
5992 | /* VMD has one orom entry for all domain, but spanning is not allowed. |
5993 | * VMD arrays should be counted per domain (controller), so skip | |
5994 | * domains that are not the given one. | |
5995 | */ | |
5996 | if ((hba->type == SYS_DEV_VMD) && | |
5997 | (strncmp(device->path, hba->path, strlen(device->path)) != 0)) | |
5998 | continue; | |
5999 | ||
72a45777 PB |
6000 | devlist = get_devices(hba_path); |
6001 | /* if no intel devices return zero volumes */ | |
6002 | if (devlist == NULL) | |
6003 | return 0; | |
6004 | ||
6005 | count += active_arrays_by_format("imsm", hba_path, &devlist, dpa, verbose); | |
6006 | dprintf("path: %s active arrays: %d\n", hba_path, count); | |
6007 | if (devlist == NULL) | |
6008 | return 0; | |
6009 | do { | |
6010 | found = 0; | |
6011 | count += count_volumes_list(devlist, | |
6012 | NULL, | |
6013 | verbose, | |
6014 | &found); | |
6015 | dprintf("found %d count: %d\n", found, count); | |
6016 | } while (found); | |
6017 | ||
6018 | dprintf("path: %s total number of volumes: %d\n", hba_path, count); | |
6019 | ||
6020 | while (devlist) { | |
6021 | struct md_list *dv = devlist; | |
6022 | devlist = devlist->next; | |
6023 | free(dv->devname); | |
6024 | free(dv); | |
6025 | } | |
ca9de185 LM |
6026 | } |
6027 | return count; | |
6028 | } | |
6029 | ||
cd9d1ac7 DW |
6030 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6031 | { | |
6032 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6033 | * 128 if no platform detected | |
6034 | */ | |
6035 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6036 | ||
6037 | return min(512, (1 << fs)); | |
6038 | } | |
73408129 | 6039 | |
6592ce37 DW |
6040 | static int |
6041 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6042 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6043 | { |
660260d0 DW |
6044 | /* check/set platform and metadata limits/defaults */ |
6045 | if (super->orom && raiddisks > super->orom->dpa) { | |
6046 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
6047 | super->orom->dpa); | |
73408129 LM |
6048 | return 0; |
6049 | } | |
6050 | ||
5d500228 | 6051 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6052 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
6053 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
6054 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
6055 | return 0; | |
6056 | } | |
cd9d1ac7 | 6057 | |
7ccc4cc4 | 6058 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6059 | *chunk = imsm_default_chunk(super->orom); |
6060 | ||
7ccc4cc4 | 6061 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
7a862a02 | 6062 | pr_vrb(": platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6063 | return 0; |
6592ce37 | 6064 | } |
cd9d1ac7 | 6065 | |
6592ce37 DW |
6066 | if (layout != imsm_level_to_layout(level)) { |
6067 | if (level == 5) | |
6068 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
6069 | else if (level == 10) | |
6070 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
6071 | else | |
6072 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
6073 | layout, level); | |
6074 | return 0; | |
6075 | } | |
2cc699af | 6076 | |
7ccc4cc4 | 6077 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af CA |
6078 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
6079 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
6080 | return 0; | |
6081 | } | |
614902f6 | 6082 | |
6592ce37 DW |
6083 | return 1; |
6084 | } | |
6085 | ||
1011e834 | 6086 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6087 | * FIX ME add ahci details |
6088 | */ | |
8b353278 | 6089 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6090 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6091 | unsigned long long size, |
6092 | unsigned long long data_offset, | |
6093 | char *dev, | |
2c514b71 NB |
6094 | unsigned long long *freesize, |
6095 | int verbose) | |
cdddbdbc | 6096 | { |
c2c087e6 DW |
6097 | struct stat stb; |
6098 | struct intel_super *super = st->sb; | |
b2916f25 | 6099 | struct imsm_super *mpb; |
c2c087e6 DW |
6100 | struct dl *dl; |
6101 | unsigned long long pos = 0; | |
6102 | unsigned long long maxsize; | |
6103 | struct extent *e; | |
6104 | int i; | |
cdddbdbc | 6105 | |
88c32bb1 DW |
6106 | /* We must have the container info already read in. */ |
6107 | if (!super) | |
c2c087e6 DW |
6108 | return 0; |
6109 | ||
b2916f25 JS |
6110 | mpb = super->anchor; |
6111 | ||
2cc699af | 6112 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 6113 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 6114 | return 0; |
d54559f0 | 6115 | } |
c2c087e6 DW |
6116 | if (!dev) { |
6117 | /* General test: make sure there is space for | |
2da8544a DW |
6118 | * 'raiddisks' device extents of size 'size' at a given |
6119 | * offset | |
c2c087e6 | 6120 | */ |
e46273eb | 6121 | unsigned long long minsize = size; |
b7528a20 | 6122 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
6123 | int dcnt = 0; |
6124 | if (minsize == 0) | |
6125 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
6126 | for (dl = super->disks; dl ; dl = dl->next) { | |
6127 | int found = 0; | |
6128 | ||
bf5a934a | 6129 | pos = 0; |
c2c087e6 DW |
6130 | i = 0; |
6131 | e = get_extents(super, dl); | |
6132 | if (!e) continue; | |
6133 | do { | |
6134 | unsigned long long esize; | |
6135 | esize = e[i].start - pos; | |
6136 | if (esize >= minsize) | |
6137 | found = 1; | |
b7528a20 | 6138 | if (found && start_offset == MaxSector) { |
2da8544a DW |
6139 | start_offset = pos; |
6140 | break; | |
6141 | } else if (found && pos != start_offset) { | |
6142 | found = 0; | |
6143 | break; | |
6144 | } | |
c2c087e6 DW |
6145 | pos = e[i].start + e[i].size; |
6146 | i++; | |
6147 | } while (e[i-1].size); | |
6148 | if (found) | |
6149 | dcnt++; | |
6150 | free(e); | |
6151 | } | |
6152 | if (dcnt < raiddisks) { | |
2c514b71 | 6153 | if (verbose) |
7a862a02 | 6154 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 6155 | dcnt, raiddisks); |
c2c087e6 DW |
6156 | return 0; |
6157 | } | |
6158 | return 1; | |
6159 | } | |
0dcecb2e | 6160 | |
c2c087e6 DW |
6161 | /* This device must be a member of the set */ |
6162 | if (stat(dev, &stb) < 0) | |
6163 | return 0; | |
6164 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
6165 | return 0; | |
6166 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
6167 | if (dl->major == (int)major(stb.st_rdev) && |
6168 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
6169 | break; |
6170 | } | |
6171 | if (!dl) { | |
2c514b71 | 6172 | if (verbose) |
7a862a02 | 6173 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 6174 | return 0; |
a20d2ba5 DW |
6175 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
6176 | /* If a volume is present then the current creation attempt | |
6177 | * cannot incorporate new spares because the orom may not | |
6178 | * understand this configuration (all member disks must be | |
6179 | * members of each array in the container). | |
6180 | */ | |
7a862a02 N |
6181 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
6182 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 6183 | return 0; |
5fe62b94 WD |
6184 | } else if (super->orom && mpb->num_raid_devs > 0 && |
6185 | mpb->num_disks != raiddisks) { | |
7a862a02 | 6186 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 6187 | return 0; |
c2c087e6 | 6188 | } |
0dcecb2e DW |
6189 | |
6190 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
6191 | e = get_extents(super, dl); |
6192 | maxsize = 0; | |
6193 | i = 0; | |
0dcecb2e DW |
6194 | if (e) { |
6195 | do { | |
6196 | unsigned long long esize; | |
6197 | ||
6198 | esize = e[i].start - pos; | |
6199 | if (esize >= maxsize) | |
6200 | maxsize = esize; | |
6201 | pos = e[i].start + e[i].size; | |
6202 | i++; | |
6203 | } while (e[i-1].size); | |
6204 | dl->e = e; | |
6205 | dl->extent_cnt = i; | |
6206 | } else { | |
6207 | if (verbose) | |
e7b84f9d | 6208 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
6209 | dev); |
6210 | return 0; | |
6211 | } | |
6212 | if (maxsize < size) { | |
6213 | if (verbose) | |
e7b84f9d | 6214 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
6215 | dev, maxsize, size); |
6216 | return 0; | |
6217 | } | |
6218 | ||
6219 | /* count total number of extents for merge */ | |
6220 | i = 0; | |
6221 | for (dl = super->disks; dl; dl = dl->next) | |
6222 | if (dl->e) | |
6223 | i += dl->extent_cnt; | |
6224 | ||
6225 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6226 | |
6227 | if (!check_env("IMSM_NO_PLATFORM") && | |
6228 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6229 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
6230 | return 0; |
6231 | } | |
6232 | ||
a7dd165b | 6233 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6234 | if (verbose) { |
6235 | if (maxsize == 0) | |
7a862a02 | 6236 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 6237 | else |
7a862a02 | 6238 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
6239 | maxsize, size); |
6240 | } | |
0dcecb2e | 6241 | return 0; |
0dcecb2e DW |
6242 | } |
6243 | ||
c2c087e6 DW |
6244 | *freesize = maxsize; |
6245 | ||
ca9de185 | 6246 | if (super->orom) { |
72a45777 | 6247 | int count = count_volumes(super->hba, |
ca9de185 LM |
6248 | super->orom->dpa, verbose); |
6249 | if (super->orom->vphba <= count) { | |
40110b90 | 6250 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6251 | super->orom->vphba); |
6252 | return 0; | |
6253 | } | |
6254 | } | |
c2c087e6 | 6255 | return 1; |
cdddbdbc DW |
6256 | } |
6257 | ||
13bcac90 | 6258 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6259 | unsigned long long size, int chunk, |
6260 | unsigned long long *freesize) | |
6261 | { | |
6262 | struct intel_super *super = st->sb; | |
6263 | struct imsm_super *mpb = super->anchor; | |
6264 | struct dl *dl; | |
6265 | int i; | |
6266 | int extent_cnt; | |
6267 | struct extent *e; | |
6268 | unsigned long long maxsize; | |
6269 | unsigned long long minsize; | |
6270 | int cnt; | |
6271 | int used; | |
6272 | ||
6273 | /* find the largest common start free region of the possible disks */ | |
6274 | used = 0; | |
6275 | extent_cnt = 0; | |
6276 | cnt = 0; | |
6277 | for (dl = super->disks; dl; dl = dl->next) { | |
6278 | dl->raiddisk = -1; | |
6279 | ||
6280 | if (dl->index >= 0) | |
6281 | used++; | |
6282 | ||
6283 | /* don't activate new spares if we are orom constrained | |
6284 | * and there is already a volume active in the container | |
6285 | */ | |
6286 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6287 | continue; | |
6288 | ||
6289 | e = get_extents(super, dl); | |
6290 | if (!e) | |
6291 | continue; | |
6292 | for (i = 1; e[i-1].size; i++) | |
6293 | ; | |
6294 | dl->e = e; | |
6295 | dl->extent_cnt = i; | |
6296 | extent_cnt += i; | |
6297 | cnt++; | |
6298 | } | |
6299 | ||
6300 | maxsize = merge_extents(super, extent_cnt); | |
6301 | minsize = size; | |
6302 | if (size == 0) | |
612e59d8 CA |
6303 | /* chunk is in K */ |
6304 | minsize = chunk * 2; | |
efb30e7f DW |
6305 | |
6306 | if (cnt < raiddisks || | |
6307 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6308 | maxsize < minsize || |
6309 | maxsize == 0) { | |
e7b84f9d | 6310 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6311 | return 0; /* No enough free spaces large enough */ |
6312 | } | |
6313 | ||
6314 | if (size == 0) { | |
6315 | size = maxsize; | |
6316 | if (chunk) { | |
612e59d8 CA |
6317 | size /= 2 * chunk; |
6318 | size *= 2 * chunk; | |
efb30e7f | 6319 | } |
f878b242 LM |
6320 | maxsize = size; |
6321 | } | |
6322 | if (!check_env("IMSM_NO_PLATFORM") && | |
6323 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6324 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 6325 | return 0; |
efb30e7f | 6326 | } |
efb30e7f DW |
6327 | cnt = 0; |
6328 | for (dl = super->disks; dl; dl = dl->next) | |
6329 | if (dl->e) | |
6330 | dl->raiddisk = cnt++; | |
6331 | ||
6332 | *freesize = size; | |
6333 | ||
13bcac90 AK |
6334 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6335 | ||
efb30e7f DW |
6336 | return 1; |
6337 | } | |
6338 | ||
13bcac90 AK |
6339 | static int reserve_space(struct supertype *st, int raiddisks, |
6340 | unsigned long long size, int chunk, | |
6341 | unsigned long long *freesize) | |
6342 | { | |
6343 | struct intel_super *super = st->sb; | |
6344 | struct dl *dl; | |
6345 | int cnt; | |
6346 | int rv = 0; | |
6347 | ||
6348 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6349 | if (rv) { | |
6350 | cnt = 0; | |
6351 | for (dl = super->disks; dl; dl = dl->next) | |
6352 | if (dl->e) | |
6353 | dl->raiddisk = cnt++; | |
6354 | rv = 1; | |
6355 | } | |
6356 | ||
6357 | return rv; | |
6358 | } | |
6359 | ||
bf5a934a | 6360 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6361 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 6362 | unsigned long long data_offset, |
bf5a934a DW |
6363 | char *dev, unsigned long long *freesize, |
6364 | int verbose) | |
6365 | { | |
6366 | int fd, cfd; | |
6367 | struct mdinfo *sra; | |
20cbe8d2 | 6368 | int is_member = 0; |
bf5a934a | 6369 | |
d54559f0 LM |
6370 | /* load capability |
6371 | * if given unused devices create a container | |
bf5a934a DW |
6372 | * if given given devices in a container create a member volume |
6373 | */ | |
6374 | if (level == LEVEL_CONTAINER) { | |
6375 | /* Must be a fresh device to add to a container */ | |
6376 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 6377 | raiddisks, |
7ccc4cc4 | 6378 | *chunk, |
af4348dd | 6379 | size, data_offset, |
bf5a934a DW |
6380 | dev, freesize, |
6381 | verbose); | |
6382 | } | |
9587c373 | 6383 | |
8592f29d | 6384 | if (!dev) { |
e91a3bad | 6385 | if (st->sb) { |
ca9de185 | 6386 | struct intel_super *super = st->sb; |
e91a3bad | 6387 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6388 | raiddisks, chunk, size, |
e91a3bad LM |
6389 | verbose)) |
6390 | return 0; | |
efb30e7f DW |
6391 | /* we are being asked to automatically layout a |
6392 | * new volume based on the current contents of | |
6393 | * the container. If the the parameters can be | |
6394 | * satisfied reserve_space will record the disks, | |
6395 | * start offset, and size of the volume to be | |
6396 | * created. add_to_super and getinfo_super | |
6397 | * detect when autolayout is in progress. | |
6398 | */ | |
ca9de185 LM |
6399 | /* assuming that freesize is always given when array is |
6400 | created */ | |
6401 | if (super->orom && freesize) { | |
6402 | int count; | |
72a45777 | 6403 | count = count_volumes(super->hba, |
ca9de185 LM |
6404 | super->orom->dpa, verbose); |
6405 | if (super->orom->vphba <= count) { | |
7a862a02 | 6406 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6407 | super->orom->vphba); |
6408 | return 0; | |
6409 | } | |
6410 | } | |
e91a3bad LM |
6411 | if (freesize) |
6412 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 6413 | *chunk, freesize); |
8592f29d N |
6414 | } |
6415 | return 1; | |
6416 | } | |
bf5a934a DW |
6417 | if (st->sb) { |
6418 | /* creating in a given container */ | |
6419 | return validate_geometry_imsm_volume(st, level, layout, | |
6420 | raiddisks, chunk, size, | |
af4348dd | 6421 | data_offset, |
bf5a934a DW |
6422 | dev, freesize, verbose); |
6423 | } | |
6424 | ||
bf5a934a DW |
6425 | /* This device needs to be a device in an 'imsm' container */ |
6426 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6427 | if (fd >= 0) { | |
6428 | if (verbose) | |
e7b84f9d N |
6429 | pr_err("Cannot create this array on device %s\n", |
6430 | dev); | |
bf5a934a DW |
6431 | close(fd); |
6432 | return 0; | |
6433 | } | |
6434 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6435 | if (verbose) | |
e7b84f9d | 6436 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6437 | dev, strerror(errno)); |
6438 | return 0; | |
6439 | } | |
6440 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6441 | cfd = open_container(fd); | |
20cbe8d2 | 6442 | close(fd); |
bf5a934a | 6443 | if (cfd < 0) { |
bf5a934a | 6444 | if (verbose) |
e7b84f9d | 6445 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6446 | dev); |
6447 | return 0; | |
6448 | } | |
4dd2df09 | 6449 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 6450 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6451 | strcmp(sra->text_version, "imsm") == 0) |
6452 | is_member = 1; | |
6453 | sysfs_free(sra); | |
6454 | if (is_member) { | |
bf5a934a DW |
6455 | /* This is a member of a imsm container. Load the container |
6456 | * and try to create a volume | |
6457 | */ | |
6458 | struct intel_super *super; | |
6459 | ||
ec50f7b6 | 6460 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 6461 | st->sb = super; |
4dd2df09 | 6462 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
6463 | close(cfd); |
6464 | return validate_geometry_imsm_volume(st, level, layout, | |
6465 | raiddisks, chunk, | |
af4348dd | 6466 | size, data_offset, dev, |
ecbd9e81 N |
6467 | freesize, 1) |
6468 | ? 1 : -1; | |
bf5a934a | 6469 | } |
20cbe8d2 | 6470 | } |
bf5a934a | 6471 | |
20cbe8d2 | 6472 | if (verbose) |
e7b84f9d | 6473 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
6474 | |
6475 | close(cfd); | |
6476 | return 0; | |
bf5a934a | 6477 | } |
0bd16cf2 | 6478 | |
30f58b22 | 6479 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6480 | { |
6481 | struct intel_super *super = st->sb; | |
6482 | ||
30f58b22 DW |
6483 | if (level && *level == UnSet) |
6484 | *level = LEVEL_CONTAINER; | |
6485 | ||
6486 | if (level && layout && *layout == UnSet) | |
6487 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6488 | |
cd9d1ac7 DW |
6489 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6490 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6491 | } |
6492 | ||
33414a01 DW |
6493 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6494 | ||
6495 | static int kill_subarray_imsm(struct supertype *st) | |
6496 | { | |
6497 | /* remove the subarray currently referenced by ->current_vol */ | |
6498 | __u8 i; | |
6499 | struct intel_dev **dp; | |
6500 | struct intel_super *super = st->sb; | |
6501 | __u8 current_vol = super->current_vol; | |
6502 | struct imsm_super *mpb = super->anchor; | |
6503 | ||
6504 | if (super->current_vol < 0) | |
6505 | return 2; | |
6506 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6507 | ||
6508 | /* block deletions that would change the uuid of active subarrays | |
6509 | * | |
6510 | * FIXME when immutable ids are available, but note that we'll | |
6511 | * also need to fixup the invalidated/active subarray indexes in | |
6512 | * mdstat | |
6513 | */ | |
6514 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6515 | char subarray[4]; | |
6516 | ||
6517 | if (i < current_vol) | |
6518 | continue; | |
6519 | sprintf(subarray, "%u", i); | |
4dd2df09 | 6520 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
6521 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
6522 | current_vol, i); | |
33414a01 DW |
6523 | |
6524 | return 2; | |
6525 | } | |
6526 | } | |
6527 | ||
6528 | if (st->update_tail) { | |
503975b9 | 6529 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 6530 | |
33414a01 DW |
6531 | u->type = update_kill_array; |
6532 | u->dev_idx = current_vol; | |
6533 | append_metadata_update(st, u, sizeof(*u)); | |
6534 | ||
6535 | return 0; | |
6536 | } | |
6537 | ||
6538 | for (dp = &super->devlist; *dp;) | |
6539 | if ((*dp)->index == current_vol) { | |
6540 | *dp = (*dp)->next; | |
6541 | } else { | |
6542 | handle_missing(super, (*dp)->dev); | |
6543 | if ((*dp)->index > current_vol) | |
6544 | (*dp)->index--; | |
6545 | dp = &(*dp)->next; | |
6546 | } | |
6547 | ||
6548 | /* no more raid devices, all active components are now spares, | |
6549 | * but of course failed are still failed | |
6550 | */ | |
6551 | if (--mpb->num_raid_devs == 0) { | |
6552 | struct dl *d; | |
6553 | ||
6554 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6555 | if (d->index > -2) |
6556 | mark_spare(d); | |
33414a01 DW |
6557 | } |
6558 | ||
6559 | super->updates_pending++; | |
6560 | ||
6561 | return 0; | |
6562 | } | |
aa534678 | 6563 | |
a951a4f7 | 6564 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6565 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6566 | { |
6567 | /* update the subarray currently referenced by ->current_vol */ | |
6568 | struct intel_super *super = st->sb; | |
6569 | struct imsm_super *mpb = super->anchor; | |
6570 | ||
aa534678 DW |
6571 | if (strcmp(update, "name") == 0) { |
6572 | char *name = ident->name; | |
a951a4f7 N |
6573 | char *ep; |
6574 | int vol; | |
aa534678 | 6575 | |
4dd2df09 | 6576 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 6577 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
6578 | return 2; |
6579 | } | |
6580 | ||
6581 | if (!check_name(super, name, 0)) | |
6582 | return 2; | |
6583 | ||
a951a4f7 N |
6584 | vol = strtoul(subarray, &ep, 10); |
6585 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6586 | return 2; | |
6587 | ||
aa534678 | 6588 | if (st->update_tail) { |
503975b9 | 6589 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 6590 | |
aa534678 | 6591 | u->type = update_rename_array; |
a951a4f7 | 6592 | u->dev_idx = vol; |
aa534678 DW |
6593 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6594 | append_metadata_update(st, u, sizeof(*u)); | |
6595 | } else { | |
6596 | struct imsm_dev *dev; | |
6597 | int i; | |
6598 | ||
a951a4f7 | 6599 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6600 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6601 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6602 | dev = get_imsm_dev(super, i); | |
6603 | handle_missing(super, dev); | |
6604 | } | |
6605 | super->updates_pending++; | |
6606 | } | |
6607 | } else | |
6608 | return 2; | |
6609 | ||
6610 | return 0; | |
6611 | } | |
d1e02575 | 6612 | #endif /* MDASSEMBLE */ |
bf5a934a | 6613 | |
28bce06f AK |
6614 | static int is_gen_migration(struct imsm_dev *dev) |
6615 | { | |
7534230b AK |
6616 | if (dev == NULL) |
6617 | return 0; | |
6618 | ||
28bce06f AK |
6619 | if (!dev->vol.migr_state) |
6620 | return 0; | |
6621 | ||
6622 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6623 | return 1; | |
6624 | ||
6625 | return 0; | |
6626 | } | |
6627 | ||
1e5c6983 DW |
6628 | static int is_rebuilding(struct imsm_dev *dev) |
6629 | { | |
6630 | struct imsm_map *migr_map; | |
6631 | ||
6632 | if (!dev->vol.migr_state) | |
6633 | return 0; | |
6634 | ||
6635 | if (migr_type(dev) != MIGR_REBUILD) | |
6636 | return 0; | |
6637 | ||
238c0a71 | 6638 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6639 | |
6640 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6641 | return 1; | |
6642 | else | |
6643 | return 0; | |
6644 | } | |
6645 | ||
b4ab44d8 | 6646 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6647 | static int is_initializing(struct imsm_dev *dev) |
6648 | { | |
6649 | struct imsm_map *migr_map; | |
6650 | ||
6651 | if (!dev->vol.migr_state) | |
6652 | return 0; | |
6653 | ||
6654 | if (migr_type(dev) != MIGR_INIT) | |
6655 | return 0; | |
6656 | ||
238c0a71 | 6657 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6658 | |
6659 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6660 | return 1; | |
6661 | ||
6662 | return 0; | |
6ce1fbf1 | 6663 | } |
b4ab44d8 | 6664 | #endif |
6ce1fbf1 | 6665 | |
c47b0ff6 AK |
6666 | static void update_recovery_start(struct intel_super *super, |
6667 | struct imsm_dev *dev, | |
6668 | struct mdinfo *array) | |
1e5c6983 DW |
6669 | { |
6670 | struct mdinfo *rebuild = NULL; | |
6671 | struct mdinfo *d; | |
6672 | __u32 units; | |
6673 | ||
6674 | if (!is_rebuilding(dev)) | |
6675 | return; | |
6676 | ||
6677 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6678 | for (d = array->devs; d; d = d->next) | |
6679 | if (d->recovery_start == 0) { | |
6680 | if (rebuild) | |
6681 | return; | |
6682 | rebuild = d; | |
6683 | } | |
6684 | ||
4363fd80 DW |
6685 | if (!rebuild) { |
6686 | /* (?) none of the disks are marked with | |
6687 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6688 | * disk_ord_tbl was not correctly updated | |
6689 | */ | |
1ade5cc1 | 6690 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
6691 | return; |
6692 | } | |
6693 | ||
1e5c6983 | 6694 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6695 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6696 | } |
6697 | ||
9e2d750d | 6698 | #ifndef MDASSEMBLE |
276d77db | 6699 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6700 | #endif |
1e5c6983 | 6701 | |
00bbdbda | 6702 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6703 | { |
4f5bc454 DW |
6704 | /* Given a container loaded by load_super_imsm_all, |
6705 | * extract information about all the arrays into | |
6706 | * an mdinfo tree. | |
00bbdbda | 6707 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6708 | * |
6709 | * For each imsm_dev create an mdinfo, fill it in, | |
6710 | * then look for matching devices in super->disks | |
6711 | * and create appropriate device mdinfo. | |
6712 | */ | |
6713 | struct intel_super *super = st->sb; | |
949c47a0 | 6714 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6715 | struct mdinfo *rest = NULL; |
00bbdbda | 6716 | unsigned int i; |
81219e70 | 6717 | int sb_errors = 0; |
abef11a3 AK |
6718 | struct dl *d; |
6719 | int spare_disks = 0; | |
cdddbdbc | 6720 | |
19482bcc AK |
6721 | /* do not assemble arrays when not all attributes are supported */ |
6722 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 6723 | sb_errors = 1; |
7a862a02 | 6724 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
6725 | } |
6726 | ||
a06d022d | 6727 | /* check for bad blocks */ |
81219e70 | 6728 | if (imsm_bbm_log_size(super->anchor)) { |
7a862a02 | 6729 | pr_err("BBM log found in IMSM metadata.Arrays activation is blocked.\n"); |
81219e70 LM |
6730 | sb_errors = 1; |
6731 | } | |
6732 | ||
abef11a3 AK |
6733 | /* count spare devices, not used in maps |
6734 | */ | |
6735 | for (d = super->disks; d; d = d->next) | |
6736 | if (d->index == -1) | |
6737 | spare_disks++; | |
6738 | ||
4f5bc454 | 6739 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6740 | struct imsm_dev *dev; |
6741 | struct imsm_map *map; | |
86e3692b | 6742 | struct imsm_map *map2; |
4f5bc454 | 6743 | struct mdinfo *this; |
a6482415 N |
6744 | int slot; |
6745 | #ifndef MDASSEMBLE | |
6746 | int chunk; | |
6747 | #endif | |
00bbdbda N |
6748 | char *ep; |
6749 | ||
6750 | if (subarray && | |
6751 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6752 | continue; | |
6753 | ||
6754 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6755 | map = get_imsm_map(dev, MAP_0); |
6756 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6757 | |
1ce0101c DW |
6758 | /* do not publish arrays that are in the middle of an |
6759 | * unsupported migration | |
6760 | */ | |
6761 | if (dev->vol.migr_state && | |
28bce06f | 6762 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 6763 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
6764 | dev->volume); |
6765 | continue; | |
6766 | } | |
2db86302 LM |
6767 | /* do not publish arrays that are not support by controller's |
6768 | * OROM/EFI | |
6769 | */ | |
1ce0101c | 6770 | |
503975b9 | 6771 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 6772 | |
301406c9 | 6773 | super->current_vol = i; |
a5d85af7 | 6774 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6775 | this->next = rest; |
81219e70 | 6776 | #ifndef MDASSEMBLE |
a6482415 | 6777 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6778 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6779 | if (!validate_geometry_imsm_orom(super, | |
6780 | get_imsm_raid_level(map), /* RAID level */ | |
6781 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6782 | map->num_members, /* raid disks */ | |
2cc699af | 6783 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6784 | 1 /* verbose */)) { |
7a862a02 | 6785 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
6786 | dev->volume); |
6787 | this->array.state |= | |
6788 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6789 | (1<<MD_SB_BLOCK_VOLUME); | |
6790 | } | |
6791 | #endif | |
6792 | ||
6793 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6794 | if (sb_errors) | |
6795 | this->array.state |= | |
6796 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6797 | (1<<MD_SB_BLOCK_VOLUME); | |
6798 | ||
4f5bc454 | 6799 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6800 | unsigned long long recovery_start; |
4f5bc454 DW |
6801 | struct mdinfo *info_d; |
6802 | struct dl *d; | |
6803 | int idx; | |
9a1608e5 | 6804 | int skip; |
7eef0453 | 6805 | __u32 ord; |
4f5bc454 | 6806 | |
9a1608e5 | 6807 | skip = 0; |
238c0a71 AK |
6808 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6809 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6810 | for (d = super->disks; d ; d = d->next) |
6811 | if (d->index == idx) | |
0fbd635c | 6812 | break; |
4f5bc454 | 6813 | |
1e5c6983 | 6814 | recovery_start = MaxSector; |
4f5bc454 | 6815 | if (d == NULL) |
9a1608e5 | 6816 | skip = 1; |
25ed7e59 | 6817 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6818 | skip = 1; |
7eef0453 | 6819 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6820 | recovery_start = 0; |
9a1608e5 | 6821 | |
1011e834 | 6822 | /* |
9a1608e5 | 6823 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
6824 | * reset resync start to avoid a dirty-degraded |
6825 | * situation when performing the intial sync | |
9a1608e5 DW |
6826 | * |
6827 | * FIXME handle dirty degraded | |
6828 | */ | |
1e5c6983 | 6829 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6830 | this->resync_start = MaxSector; |
9a1608e5 DW |
6831 | if (skip) |
6832 | continue; | |
4f5bc454 | 6833 | |
503975b9 | 6834 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
6835 | info_d->next = this->devs; |
6836 | this->devs = info_d; | |
6837 | ||
4f5bc454 DW |
6838 | info_d->disk.number = d->index; |
6839 | info_d->disk.major = d->major; | |
6840 | info_d->disk.minor = d->minor; | |
6841 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6842 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6843 | if (map2) { |
6844 | if (slot < map2->num_members) | |
6845 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6846 | else |
6847 | this->array.spare_disks++; | |
86e3692b AK |
6848 | } else { |
6849 | if (slot < map->num_members) | |
6850 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6851 | else |
6852 | this->array.spare_disks++; | |
86e3692b | 6853 | } |
1e5c6983 DW |
6854 | if (info_d->recovery_start == MaxSector) |
6855 | this->array.working_disks++; | |
4f5bc454 DW |
6856 | |
6857 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6858 | info_d->data_offset = pba_of_lba0(map); |
6859 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6860 | } |
1e5c6983 | 6861 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6862 | update_recovery_start(super, dev, this); |
abef11a3 | 6863 | this->array.spare_disks += spare_disks; |
276d77db | 6864 | |
9e2d750d | 6865 | #ifndef MDASSEMBLE |
276d77db AK |
6866 | /* check for reshape */ |
6867 | if (this->reshape_active == 1) | |
6868 | recover_backup_imsm(st, this); | |
9e2d750d | 6869 | #endif |
9a1608e5 | 6870 | rest = this; |
4f5bc454 DW |
6871 | } |
6872 | ||
6873 | return rest; | |
cdddbdbc DW |
6874 | } |
6875 | ||
3b451610 AK |
6876 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6877 | int failed, int look_in_map) | |
c2a1e7da | 6878 | { |
3b451610 AK |
6879 | struct imsm_map *map; |
6880 | ||
6881 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6882 | |
6883 | if (!failed) | |
1011e834 | 6884 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 6885 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
6886 | |
6887 | switch (get_imsm_raid_level(map)) { | |
6888 | case 0: | |
6889 | return IMSM_T_STATE_FAILED; | |
6890 | break; | |
6891 | case 1: | |
6892 | if (failed < map->num_members) | |
6893 | return IMSM_T_STATE_DEGRADED; | |
6894 | else | |
6895 | return IMSM_T_STATE_FAILED; | |
6896 | break; | |
6897 | case 10: | |
6898 | { | |
6899 | /** | |
c92a2527 DW |
6900 | * check to see if any mirrors have failed, otherwise we |
6901 | * are degraded. Even numbered slots are mirrored on | |
6902 | * slot+1 | |
c2a1e7da | 6903 | */ |
c2a1e7da | 6904 | int i; |
d9b420a5 N |
6905 | /* gcc -Os complains that this is unused */ |
6906 | int insync = insync; | |
c2a1e7da DW |
6907 | |
6908 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6909 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6910 | int idx = ord_to_idx(ord); |
6911 | struct imsm_disk *disk; | |
c2a1e7da | 6912 | |
c92a2527 | 6913 | /* reset the potential in-sync count on even-numbered |
1011e834 | 6914 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
6915 | */ |
6916 | if ((i & 1) == 0) | |
6917 | insync = 2; | |
c2a1e7da | 6918 | |
c92a2527 | 6919 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6920 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6921 | insync--; |
c2a1e7da | 6922 | |
c92a2527 DW |
6923 | /* no in-sync disks left in this mirror the |
6924 | * array has failed | |
6925 | */ | |
6926 | if (insync == 0) | |
6927 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6928 | } |
6929 | ||
6930 | return IMSM_T_STATE_DEGRADED; | |
6931 | } | |
6932 | case 5: | |
6933 | if (failed < 2) | |
6934 | return IMSM_T_STATE_DEGRADED; | |
6935 | else | |
6936 | return IMSM_T_STATE_FAILED; | |
6937 | break; | |
6938 | default: | |
6939 | break; | |
6940 | } | |
6941 | ||
6942 | return map->map_state; | |
6943 | } | |
6944 | ||
3b451610 AK |
6945 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6946 | int look_in_map) | |
c2a1e7da DW |
6947 | { |
6948 | int i; | |
6949 | int failed = 0; | |
6950 | struct imsm_disk *disk; | |
d5985138 AK |
6951 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6952 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6953 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6954 | __u32 ord; |
6955 | int idx; | |
d5985138 | 6956 | int idx_1; |
c2a1e7da | 6957 | |
0556e1a2 DW |
6958 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6959 | * disks that are being rebuilt. New failures are recorded to | |
6960 | * map[0]. So we look through all the disks we started with and | |
6961 | * see if any failures are still present, or if any new ones | |
6962 | * have arrived | |
0556e1a2 | 6963 | */ |
d5985138 AK |
6964 | map_for_loop = map; |
6965 | if (prev && (map->num_members < prev->num_members)) | |
6966 | map_for_loop = prev; | |
68fe4598 LD |
6967 | |
6968 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6969 | idx_1 = -255; |
238c0a71 AK |
6970 | /* when MAP_X is passed both maps failures are counted |
6971 | */ | |
d5985138 | 6972 | if (prev && |
238c0a71 AK |
6973 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6974 | (i < prev->num_members)) { | |
d5985138 AK |
6975 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6976 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6977 | |
d5985138 AK |
6978 | disk = get_imsm_disk(super, idx_1); |
6979 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6980 | failed++; | |
6981 | } | |
238c0a71 AK |
6982 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6983 | (i < map->num_members)) { | |
d5985138 AK |
6984 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6985 | idx = ord_to_idx(ord); | |
6986 | ||
6987 | if (idx != idx_1) { | |
6988 | disk = get_imsm_disk(super, idx); | |
6989 | if (!disk || is_failed(disk) || | |
6990 | ord & IMSM_ORD_REBUILD) | |
6991 | failed++; | |
6992 | } | |
6993 | } | |
c2a1e7da DW |
6994 | } |
6995 | ||
6996 | return failed; | |
845dea95 NB |
6997 | } |
6998 | ||
97b4d0e9 DW |
6999 | #ifndef MDASSEMBLE |
7000 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
7001 | char *inst) | |
7002 | { | |
7003 | struct intel_super *super = c->sb; | |
7004 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 7005 | |
97b4d0e9 | 7006 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 7007 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
7008 | return -ENODEV; |
7009 | } | |
7010 | ||
7011 | dprintf("imsm: open_new %s\n", inst); | |
7012 | a->info.container_member = atoi(inst); | |
7013 | return 0; | |
7014 | } | |
7015 | ||
0c046afd DW |
7016 | static int is_resyncing(struct imsm_dev *dev) |
7017 | { | |
7018 | struct imsm_map *migr_map; | |
7019 | ||
7020 | if (!dev->vol.migr_state) | |
7021 | return 0; | |
7022 | ||
1484e727 DW |
7023 | if (migr_type(dev) == MIGR_INIT || |
7024 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7025 | return 1; |
7026 | ||
4c9bc37b AK |
7027 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7028 | return 0; | |
7029 | ||
238c0a71 | 7030 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7031 | |
4c9bc37b AK |
7032 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
7033 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
7034 | return 1; |
7035 | else | |
7036 | return 0; | |
7037 | } | |
7038 | ||
0556e1a2 DW |
7039 | /* return true if we recorded new information */ |
7040 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7041 | { |
0556e1a2 DW |
7042 | __u32 ord; |
7043 | int slot; | |
7044 | struct imsm_map *map; | |
86c54047 DW |
7045 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7046 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7047 | |
7048 | /* new failures are always set in map[0] */ | |
238c0a71 | 7049 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7050 | |
7051 | slot = get_imsm_disk_slot(map, idx); | |
7052 | if (slot < 0) | |
7053 | return 0; | |
7054 | ||
7055 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 7056 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
7057 | return 0; |
7058 | ||
7d0c5e24 LD |
7059 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
7060 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
7061 | strcat(buf, ":0"); | |
86c54047 DW |
7062 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
7063 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
7064 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
7065 | ||
f2f27e63 | 7066 | disk->status |= FAILED_DISK; |
0556e1a2 | 7067 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
7068 | /* mark failures in second map if second map exists and this disk |
7069 | * in this slot. | |
7070 | * This is valid for migration, initialization and rebuild | |
7071 | */ | |
7072 | if (dev->vol.migr_state) { | |
238c0a71 | 7073 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
7074 | int slot2 = get_imsm_disk_slot(map2, idx); |
7075 | ||
7076 | if ((slot2 < map2->num_members) && | |
7077 | (slot2 >= 0)) | |
7078 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
7079 | idx | IMSM_ORD_REBUILD); |
7080 | } | |
f21e18ca | 7081 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
7082 | map->failed_disk_num = slot; |
7083 | return 1; | |
7084 | } | |
7085 | ||
7086 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
7087 | { | |
7088 | mark_failure(dev, disk, idx); | |
7089 | ||
7090 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
7091 | return; | |
7092 | ||
47ee5a45 DW |
7093 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
7094 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
7095 | } | |
7096 | ||
33414a01 DW |
7097 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
7098 | { | |
33414a01 | 7099 | struct dl *dl; |
33414a01 DW |
7100 | |
7101 | if (!super->missing) | |
7102 | return; | |
33414a01 | 7103 | |
79b68f1b PC |
7104 | /* When orom adds replacement for missing disk it does |
7105 | * not remove entry of missing disk, but just updates map with | |
7106 | * new added disk. So it is not enough just to test if there is | |
7107 | * any missing disk, we have to look if there are any failed disks | |
7108 | * in map to stop migration */ | |
7109 | ||
33414a01 | 7110 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
7111 | /* end process for initialization and rebuild only |
7112 | */ | |
7113 | if (is_gen_migration(dev) == 0) { | |
7114 | __u8 map_state; | |
7115 | int failed; | |
7116 | ||
7117 | failed = imsm_count_failed(super, dev, MAP_0); | |
7118 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
7119 | ||
79b68f1b PC |
7120 | if (failed) |
7121 | end_migration(dev, super, map_state); | |
3d59f0c0 | 7122 | } |
33414a01 DW |
7123 | for (dl = super->missing; dl; dl = dl->next) |
7124 | mark_missing(dev, &dl->disk, dl->index); | |
7125 | super->updates_pending++; | |
7126 | } | |
7127 | ||
f3871fdc AK |
7128 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
7129 | long long new_size) | |
70bdf0dc | 7130 | { |
238c0a71 | 7131 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
7132 | unsigned long long array_blocks; |
7133 | struct imsm_map *map; | |
7134 | ||
7135 | if (used_disks == 0) { | |
7136 | /* when problems occures | |
7137 | * return current array_blocks value | |
7138 | */ | |
7139 | array_blocks = __le32_to_cpu(dev->size_high); | |
7140 | array_blocks = array_blocks << 32; | |
7141 | array_blocks += __le32_to_cpu(dev->size_low); | |
7142 | ||
7143 | return array_blocks; | |
7144 | } | |
7145 | ||
7146 | /* set array size in metadata | |
7147 | */ | |
f3871fdc AK |
7148 | if (new_size <= 0) { |
7149 | /* OLCE size change is caused by added disks | |
7150 | */ | |
7151 | map = get_imsm_map(dev, MAP_0); | |
7152 | array_blocks = blocks_per_member(map) * used_disks; | |
7153 | } else { | |
7154 | /* Online Volume Size Change | |
7155 | * Using available free space | |
7156 | */ | |
7157 | array_blocks = new_size; | |
7158 | } | |
70bdf0dc AK |
7159 | |
7160 | /* round array size down to closest MB | |
7161 | */ | |
7162 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
7163 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
7164 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
7165 | ||
7166 | return array_blocks; | |
7167 | } | |
7168 | ||
28bce06f AK |
7169 | static void imsm_set_disk(struct active_array *a, int n, int state); |
7170 | ||
0e2d1a4e AK |
7171 | static void imsm_progress_container_reshape(struct intel_super *super) |
7172 | { | |
7173 | /* if no device has a migr_state, but some device has a | |
7174 | * different number of members than the previous device, start | |
7175 | * changing the number of devices in this device to match | |
7176 | * previous. | |
7177 | */ | |
7178 | struct imsm_super *mpb = super->anchor; | |
7179 | int prev_disks = -1; | |
7180 | int i; | |
1dfaa380 | 7181 | int copy_map_size; |
0e2d1a4e AK |
7182 | |
7183 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7184 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 7185 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
7186 | struct imsm_map *map2; |
7187 | int prev_num_members; | |
0e2d1a4e AK |
7188 | |
7189 | if (dev->vol.migr_state) | |
7190 | return; | |
7191 | ||
7192 | if (prev_disks == -1) | |
7193 | prev_disks = map->num_members; | |
7194 | if (prev_disks == map->num_members) | |
7195 | continue; | |
7196 | ||
7197 | /* OK, this array needs to enter reshape mode. | |
7198 | * i.e it needs a migr_state | |
7199 | */ | |
7200 | ||
1dfaa380 | 7201 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
7202 | prev_num_members = map->num_members; |
7203 | map->num_members = prev_disks; | |
7204 | dev->vol.migr_state = 1; | |
7205 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7206 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7207 | for (i = prev_num_members; |
7208 | i < map->num_members; i++) | |
7209 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7210 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7211 | /* Copy the current map */ |
1dfaa380 | 7212 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7213 | map2->num_members = prev_num_members; |
7214 | ||
f3871fdc | 7215 | imsm_set_array_size(dev, -1); |
51d83f5d | 7216 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7217 | super->updates_pending++; |
7218 | } | |
7219 | } | |
7220 | ||
aad6f216 | 7221 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7222 | * states are handled in imsm_set_disk() with one exception, when a |
7223 | * resync is stopped due to a new failure this routine will set the | |
7224 | * 'degraded' state for the array. | |
7225 | */ | |
01f157d7 | 7226 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7227 | { |
7228 | int inst = a->info.container_member; | |
7229 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7230 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7231 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7232 | int failed = imsm_count_failed(super, dev, MAP_0); |
7233 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7234 | __u32 blocks_per_unit; |
a862209d | 7235 | |
1af97990 AK |
7236 | if (dev->vol.migr_state && |
7237 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7238 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7239 | * We might need to |
7240 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7241 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7242 | * - update curr_migr_unit | |
1af97990 | 7243 | */ |
aad6f216 N |
7244 | if (a->curr_action == reshape) { |
7245 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7246 | goto mark_checkpoint; |
aad6f216 N |
7247 | } else { |
7248 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7249 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7250 | * |
7251 | * disable automatic metadata rollback | |
7252 | * user action is required to recover process | |
aad6f216 | 7253 | */ |
b66e591b | 7254 | if (0) { |
238c0a71 AK |
7255 | struct imsm_map *map2 = |
7256 | get_imsm_map(dev, MAP_1); | |
7257 | dev->vol.migr_state = 0; | |
7258 | set_migr_type(dev, 0); | |
7259 | dev->vol.curr_migr_unit = 0; | |
7260 | memcpy(map, map2, | |
7261 | sizeof_imsm_map(map2)); | |
7262 | super->updates_pending++; | |
b66e591b | 7263 | } |
aad6f216 N |
7264 | } |
7265 | if (a->last_checkpoint >= a->info.component_size) { | |
7266 | unsigned long long array_blocks; | |
7267 | int used_disks; | |
e154ced3 | 7268 | struct mdinfo *mdi; |
aad6f216 | 7269 | |
238c0a71 | 7270 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7271 | if (used_disks > 0) { |
7272 | array_blocks = | |
5551b113 | 7273 | blocks_per_member(map) * |
d55adef9 AK |
7274 | used_disks; |
7275 | /* round array size down to closest MB | |
7276 | */ | |
7277 | array_blocks = (array_blocks | |
7278 | >> SECT_PER_MB_SHIFT) | |
7279 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7280 | a->info.custom_array_size = array_blocks; |
7281 | /* encourage manager to update array | |
7282 | * size | |
7283 | */ | |
e154ced3 | 7284 | |
d55adef9 | 7285 | a->check_reshape = 1; |
633b5610 | 7286 | } |
e154ced3 AK |
7287 | /* finalize online capacity expansion/reshape */ |
7288 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7289 | imsm_set_disk(a, | |
7290 | mdi->disk.raid_disk, | |
7291 | mdi->curr_state); | |
7292 | ||
0e2d1a4e | 7293 | imsm_progress_container_reshape(super); |
e154ced3 | 7294 | } |
aad6f216 | 7295 | } |
1af97990 AK |
7296 | } |
7297 | ||
47ee5a45 | 7298 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7299 | if (consistent == 2) |
7300 | handle_missing(super, dev); | |
1e5c6983 | 7301 | |
0c046afd | 7302 | if (consistent == 2 && |
b7941fd6 | 7303 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7304 | map_state != IMSM_T_STATE_NORMAL || |
7305 | dev->vol.migr_state)) | |
01f157d7 | 7306 | consistent = 0; |
272906ef | 7307 | |
b7941fd6 | 7308 | if (is_resync_complete(&a->info)) { |
0c046afd | 7309 | /* complete intialization / resync, |
0556e1a2 DW |
7310 | * recovery and interrupted recovery is completed in |
7311 | * ->set_disk | |
0c046afd DW |
7312 | */ |
7313 | if (is_resyncing(dev)) { | |
7314 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7315 | end_migration(dev, super, map_state); |
115c3803 | 7316 | super->updates_pending++; |
484240d8 | 7317 | a->last_checkpoint = 0; |
115c3803 | 7318 | } |
b9172665 AK |
7319 | } else if ((!is_resyncing(dev) && !failed) && |
7320 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7321 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7322 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7323 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7324 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7325 | else |
8e59f3d8 | 7326 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7327 | super->updates_pending++; |
115c3803 | 7328 | } |
a862209d | 7329 | |
633b5610 | 7330 | mark_checkpoint: |
5b83bacf AK |
7331 | /* skip checkpointing for general migration, |
7332 | * it is controlled in mdadm | |
7333 | */ | |
7334 | if (is_gen_migration(dev)) | |
7335 | goto skip_mark_checkpoint; | |
7336 | ||
1e5c6983 | 7337 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7338 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7339 | if (blocks_per_unit) { |
1e5c6983 DW |
7340 | __u32 units32; |
7341 | __u64 units; | |
7342 | ||
4f0a7acc | 7343 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7344 | units32 = units; |
7345 | ||
7346 | /* check that we did not overflow 32-bits, and that | |
7347 | * curr_migr_unit needs updating | |
7348 | */ | |
7349 | if (units32 == units && | |
bfd80a56 | 7350 | units32 != 0 && |
1e5c6983 DW |
7351 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7352 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7353 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7354 | super->updates_pending++; | |
7355 | } | |
7356 | } | |
f8f603f1 | 7357 | |
5b83bacf | 7358 | skip_mark_checkpoint: |
3393c6af | 7359 | /* mark dirty / clean */ |
0c046afd | 7360 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7361 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7362 | if (consistent) |
7363 | dev->vol.dirty = 0; | |
7364 | else | |
7365 | dev->vol.dirty = 1; | |
a862209d DW |
7366 | super->updates_pending++; |
7367 | } | |
28bce06f | 7368 | |
01f157d7 | 7369 | return consistent; |
a862209d DW |
7370 | } |
7371 | ||
8d45d196 | 7372 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7373 | { |
8d45d196 DW |
7374 | int inst = a->info.container_member; |
7375 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7376 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7377 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7378 | struct imsm_disk *disk; |
7ce05701 LD |
7379 | struct mdinfo *mdi; |
7380 | int recovery_not_finished = 0; | |
0c046afd | 7381 | int failed; |
b10b37b8 | 7382 | __u32 ord; |
0c046afd | 7383 | __u8 map_state; |
8d45d196 DW |
7384 | |
7385 | if (n > map->num_members) | |
e12b3daa | 7386 | pr_err("imsm: set_disk %d out of range 0..%d\n", |
8d45d196 DW |
7387 | n, map->num_members - 1); |
7388 | ||
7389 | if (n < 0) | |
7390 | return; | |
7391 | ||
4e6e574a | 7392 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7393 | |
238c0a71 | 7394 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7395 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7396 | |
5802a811 | 7397 | /* check for new failures */ |
0556e1a2 DW |
7398 | if (state & DS_FAULTY) { |
7399 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7400 | super->updates_pending++; | |
8d45d196 | 7401 | } |
47ee5a45 | 7402 | |
19859edc | 7403 | /* check if in_sync */ |
0556e1a2 | 7404 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7405 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7406 | |
7407 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7408 | super->updates_pending++; |
7409 | } | |
8d45d196 | 7410 | |
3b451610 AK |
7411 | failed = imsm_count_failed(super, dev, MAP_0); |
7412 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7413 | |
0c046afd | 7414 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7415 | dprintf("imsm: Detected transition to state "); |
7416 | switch (map_state) { | |
7417 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7418 | dprintf("normal: "); | |
7419 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 7420 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
7421 | /* check if recovery is really finished */ |
7422 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
7423 | if (mdi->recovery_start != MaxSector) { | |
7424 | recovery_not_finished = 1; | |
7425 | break; | |
7426 | } | |
7427 | if (recovery_not_finished) { | |
1ade5cc1 N |
7428 | dprintf_cont("\n"); |
7429 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
7430 | if (a->last_checkpoint < mdi->recovery_start) { |
7431 | a->last_checkpoint = mdi->recovery_start; | |
7432 | super->updates_pending++; | |
7433 | } | |
7434 | break; | |
7435 | } | |
94002678 | 7436 | end_migration(dev, super, map_state); |
238c0a71 | 7437 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7438 | map->failed_disk_num = ~0; |
7439 | super->updates_pending++; | |
7440 | a->last_checkpoint = 0; | |
7441 | break; | |
7442 | } | |
7443 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 7444 | dprintf_cont("while general migration"); |
bf2f0071 | 7445 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7446 | end_migration(dev, super, map_state); |
94002678 AK |
7447 | else |
7448 | map->map_state = map_state; | |
238c0a71 | 7449 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7450 | map->failed_disk_num = ~0; |
94002678 | 7451 | super->updates_pending++; |
bf2f0071 | 7452 | break; |
94002678 AK |
7453 | } |
7454 | break; | |
7455 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 7456 | dprintf_cont("degraded: "); |
94002678 AK |
7457 | if ((map->map_state != map_state) && |
7458 | !dev->vol.migr_state) { | |
1ade5cc1 | 7459 | dprintf_cont("mark degraded"); |
94002678 AK |
7460 | map->map_state = map_state; |
7461 | super->updates_pending++; | |
7462 | a->last_checkpoint = 0; | |
7463 | break; | |
7464 | } | |
7465 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 7466 | dprintf_cont("while rebuilding."); |
94002678 | 7467 | if (map->map_state != map_state) { |
1ade5cc1 | 7468 | dprintf_cont(" Map state change"); |
94002678 AK |
7469 | end_migration(dev, super, map_state); |
7470 | super->updates_pending++; | |
7471 | } | |
7472 | break; | |
7473 | } | |
7474 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 7475 | dprintf_cont("while general migration"); |
bf2f0071 | 7476 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7477 | end_migration(dev, super, map_state); |
94002678 AK |
7478 | else { |
7479 | map->map_state = map_state; | |
3b451610 | 7480 | manage_second_map(super, dev); |
94002678 AK |
7481 | } |
7482 | super->updates_pending++; | |
bf2f0071 | 7483 | break; |
28bce06f | 7484 | } |
6ce1fbf1 | 7485 | if (is_initializing(dev)) { |
1ade5cc1 | 7486 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
7487 | map->map_state = map_state; |
7488 | super->updates_pending++; | |
7489 | break; | |
7490 | } | |
94002678 AK |
7491 | break; |
7492 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 7493 | dprintf_cont("failed: "); |
94002678 | 7494 | if (is_gen_migration(dev)) { |
1ade5cc1 | 7495 | dprintf_cont("while general migration"); |
94002678 AK |
7496 | map->map_state = map_state; |
7497 | super->updates_pending++; | |
7498 | break; | |
7499 | } | |
7500 | if (map->map_state != map_state) { | |
1ade5cc1 | 7501 | dprintf_cont("mark failed"); |
94002678 AK |
7502 | end_migration(dev, super, map_state); |
7503 | super->updates_pending++; | |
7504 | a->last_checkpoint = 0; | |
7505 | break; | |
7506 | } | |
7507 | break; | |
7508 | default: | |
1ade5cc1 | 7509 | dprintf_cont("state %i\n", map_state); |
5802a811 | 7510 | } |
1ade5cc1 | 7511 | dprintf_cont("\n"); |
845dea95 NB |
7512 | } |
7513 | ||
f796af5d | 7514 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7515 | { |
f796af5d | 7516 | void *buf = mpb; |
c2a1e7da DW |
7517 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7518 | unsigned long long dsize; | |
7519 | unsigned long long sectors; | |
7520 | ||
7521 | get_dev_size(fd, NULL, &dsize); | |
7522 | ||
272f648f DW |
7523 | if (mpb_size > 512) { |
7524 | /* -1 to account for anchor */ | |
7525 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7526 | |
272f648f DW |
7527 | /* write the extended mpb to the sectors preceeding the anchor */ |
7528 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7529 | return 1; | |
c2a1e7da | 7530 | |
f21e18ca N |
7531 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7532 | != 512 * sectors) | |
272f648f DW |
7533 | return 1; |
7534 | } | |
c2a1e7da | 7535 | |
272f648f DW |
7536 | /* first block is stored on second to last sector of the disk */ |
7537 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7538 | return 1; |
7539 | ||
f796af5d | 7540 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7541 | return 1; |
7542 | ||
c2a1e7da DW |
7543 | return 0; |
7544 | } | |
7545 | ||
2e735d19 | 7546 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7547 | { |
2e735d19 | 7548 | struct intel_super *super = container->sb; |
c2a1e7da | 7549 | |
1a64be56 | 7550 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7551 | if (!super->updates_pending) |
7552 | return; | |
7553 | ||
36988a3d | 7554 | write_super_imsm(container, 0); |
c2a1e7da DW |
7555 | |
7556 | super->updates_pending = 0; | |
845dea95 NB |
7557 | } |
7558 | ||
272906ef DW |
7559 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7560 | { | |
7561 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7562 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7563 | struct dl *dl; |
7564 | ||
7565 | for (dl = super->disks; dl; dl = dl->next) | |
7566 | if (dl->index == i) | |
7567 | break; | |
7568 | ||
25ed7e59 | 7569 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7570 | dl = NULL; |
7571 | ||
7572 | if (dl) | |
1ade5cc1 | 7573 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
7574 | |
7575 | return dl; | |
7576 | } | |
7577 | ||
a20d2ba5 | 7578 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7579 | struct active_array *a, int activate_new, |
7580 | struct mdinfo *additional_test_list) | |
272906ef DW |
7581 | { |
7582 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7583 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7584 | struct imsm_super *mpb = super->anchor; |
7585 | struct imsm_map *map; | |
272906ef DW |
7586 | unsigned long long pos; |
7587 | struct mdinfo *d; | |
7588 | struct extent *ex; | |
a20d2ba5 | 7589 | int i, j; |
272906ef | 7590 | int found; |
569cc43f DW |
7591 | __u32 array_start = 0; |
7592 | __u32 array_end = 0; | |
272906ef | 7593 | struct dl *dl; |
6c932028 | 7594 | struct mdinfo *test_list; |
272906ef DW |
7595 | |
7596 | for (dl = super->disks; dl; dl = dl->next) { | |
7597 | /* If in this array, skip */ | |
7598 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7599 | if (d->state_fd >= 0 && |
7600 | d->disk.major == dl->major && | |
272906ef | 7601 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7602 | dprintf("%x:%x already in array\n", |
7603 | dl->major, dl->minor); | |
272906ef DW |
7604 | break; |
7605 | } | |
7606 | if (d) | |
7607 | continue; | |
6c932028 AK |
7608 | test_list = additional_test_list; |
7609 | while (test_list) { | |
7610 | if (test_list->disk.major == dl->major && | |
7611 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7612 | dprintf("%x:%x already in additional test list\n", |
7613 | dl->major, dl->minor); | |
7614 | break; | |
7615 | } | |
6c932028 | 7616 | test_list = test_list->next; |
8ba77d32 | 7617 | } |
6c932028 | 7618 | if (test_list) |
8ba77d32 | 7619 | continue; |
272906ef | 7620 | |
e553d2a4 | 7621 | /* skip in use or failed drives */ |
25ed7e59 | 7622 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7623 | dl->index == -2) { |
7624 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7625 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7626 | continue; |
7627 | } | |
7628 | ||
a20d2ba5 DW |
7629 | /* skip pure spares when we are looking for partially |
7630 | * assimilated drives | |
7631 | */ | |
7632 | if (dl->index == -1 && !activate_new) | |
7633 | continue; | |
7634 | ||
272906ef | 7635 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7636 | * It needs to be able to cover all member volumes |
272906ef DW |
7637 | */ |
7638 | ex = get_extents(super, dl); | |
7639 | if (!ex) { | |
7640 | dprintf("cannot get extents\n"); | |
7641 | continue; | |
7642 | } | |
a20d2ba5 DW |
7643 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7644 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7645 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7646 | |
a20d2ba5 DW |
7647 | /* check if this disk is already a member of |
7648 | * this array | |
272906ef | 7649 | */ |
620b1713 | 7650 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7651 | continue; |
7652 | ||
7653 | found = 0; | |
7654 | j = 0; | |
7655 | pos = 0; | |
5551b113 | 7656 | array_start = pba_of_lba0(map); |
329c8278 | 7657 | array_end = array_start + |
5551b113 | 7658 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7659 | |
7660 | do { | |
7661 | /* check that we can start at pba_of_lba0 with | |
7662 | * blocks_per_member of space | |
7663 | */ | |
329c8278 | 7664 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7665 | found = 1; |
7666 | break; | |
7667 | } | |
7668 | pos = ex[j].start + ex[j].size; | |
7669 | j++; | |
7670 | } while (ex[j-1].size); | |
7671 | ||
7672 | if (!found) | |
272906ef | 7673 | break; |
a20d2ba5 | 7674 | } |
272906ef DW |
7675 | |
7676 | free(ex); | |
a20d2ba5 | 7677 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7678 | dprintf("%x:%x does not have %u to %u available\n", |
7679 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7680 | /* No room */ |
7681 | continue; | |
a20d2ba5 DW |
7682 | } |
7683 | return dl; | |
272906ef DW |
7684 | } |
7685 | ||
7686 | return dl; | |
7687 | } | |
7688 | ||
95d07a2c LM |
7689 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
7690 | { | |
7691 | struct imsm_dev *dev2; | |
7692 | struct imsm_map *map; | |
7693 | struct dl *idisk; | |
7694 | int slot; | |
7695 | int idx; | |
7696 | __u8 state; | |
7697 | ||
7698 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7699 | if (dev2) { | |
238c0a71 | 7700 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7701 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7702 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7703 | if (!map) |
7704 | return 1; | |
7705 | for (slot = 0; slot < map->num_members; slot++) { | |
7706 | /* | |
7707 | * Check if failed disks are deleted from intel | |
7708 | * disk list or are marked to be deleted | |
7709 | */ | |
238c0a71 | 7710 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7711 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7712 | /* | |
7713 | * Do not rebuild the array if failed disks | |
7714 | * from failed sub-array are not removed from | |
7715 | * container. | |
7716 | */ | |
7717 | if (idisk && | |
7718 | is_failed(&idisk->disk) && | |
7719 | (idisk->action != DISK_REMOVE)) | |
7720 | return 0; | |
7721 | } | |
7722 | } | |
7723 | } | |
7724 | return 1; | |
7725 | } | |
7726 | ||
88758e9d DW |
7727 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7728 | struct metadata_update **updates) | |
7729 | { | |
7730 | /** | |
d23fe947 DW |
7731 | * Find a device with unused free space and use it to replace a |
7732 | * failed/vacant region in an array. We replace failed regions one a | |
7733 | * array at a time. The result is that a new spare disk will be added | |
7734 | * to the first failed array and after the monitor has finished | |
7735 | * propagating failures the remainder will be consumed. | |
88758e9d | 7736 | * |
d23fe947 DW |
7737 | * FIXME add a capability for mdmon to request spares from another |
7738 | * container. | |
88758e9d DW |
7739 | */ |
7740 | ||
7741 | struct intel_super *super = a->container->sb; | |
88758e9d | 7742 | int inst = a->info.container_member; |
949c47a0 | 7743 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7744 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7745 | int failed = a->info.array.raid_disks; |
7746 | struct mdinfo *rv = NULL; | |
7747 | struct mdinfo *d; | |
7748 | struct mdinfo *di; | |
7749 | struct metadata_update *mu; | |
7750 | struct dl *dl; | |
7751 | struct imsm_update_activate_spare *u; | |
7752 | int num_spares = 0; | |
7753 | int i; | |
95d07a2c | 7754 | int allowed; |
88758e9d DW |
7755 | |
7756 | for (d = a->info.devs ; d ; d = d->next) { | |
7757 | if ((d->curr_state & DS_FAULTY) && | |
7758 | d->state_fd >= 0) | |
7759 | /* wait for Removal to happen */ | |
7760 | return NULL; | |
7761 | if (d->state_fd >= 0) | |
7762 | failed--; | |
7763 | } | |
7764 | ||
7765 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7766 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7767 | |
e2962bfc AK |
7768 | if (imsm_reshape_blocks_arrays_changes(super)) |
7769 | return NULL; | |
1af97990 | 7770 | |
fc8ca064 AK |
7771 | /* Cannot activate another spare if rebuild is in progress already |
7772 | */ | |
7773 | if (is_rebuilding(dev)) { | |
7a862a02 | 7774 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
7775 | return NULL; |
7776 | } | |
7777 | ||
89c67882 AK |
7778 | if (a->info.array.level == 4) |
7779 | /* No repair for takeovered array | |
7780 | * imsm doesn't support raid4 | |
7781 | */ | |
7782 | return NULL; | |
7783 | ||
3b451610 AK |
7784 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7785 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7786 | return NULL; |
7787 | ||
95d07a2c LM |
7788 | /* |
7789 | * If there are any failed disks check state of the other volume. | |
7790 | * Block rebuild if the another one is failed until failed disks | |
7791 | * are removed from container. | |
7792 | */ | |
7793 | if (failed) { | |
7a862a02 | 7794 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 7795 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7796 | /* check if states of the other volumes allow for rebuild */ |
7797 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7798 | if (i != inst) { | |
7799 | allowed = imsm_rebuild_allowed(a->container, | |
7800 | i, failed); | |
7801 | if (!allowed) | |
7802 | return NULL; | |
7803 | } | |
7804 | } | |
7805 | } | |
7806 | ||
88758e9d | 7807 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7808 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7809 | for (d = a->info.devs ; d ; d = d->next) | |
7810 | if (d->disk.raid_disk == i) | |
7811 | break; | |
7812 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7813 | if (d && (d->state_fd >= 0)) | |
7814 | continue; | |
7815 | ||
272906ef | 7816 | /* |
a20d2ba5 DW |
7817 | * OK, this device needs recovery. Try to re-add the |
7818 | * previous occupant of this slot, if this fails see if | |
7819 | * we can continue the assimilation of a spare that was | |
7820 | * partially assimilated, finally try to activate a new | |
7821 | * spare. | |
272906ef DW |
7822 | */ |
7823 | dl = imsm_readd(super, i, a); | |
7824 | if (!dl) | |
b303fe21 | 7825 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7826 | if (!dl) |
b303fe21 | 7827 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7828 | if (!dl) |
7829 | continue; | |
1011e834 | 7830 | |
272906ef | 7831 | /* found a usable disk with enough space */ |
503975b9 | 7832 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
7833 | |
7834 | /* dl->index will be -1 in the case we are activating a | |
7835 | * pristine spare. imsm_process_update() will create a | |
7836 | * new index in this case. Once a disk is found to be | |
7837 | * failed in all member arrays it is kicked from the | |
7838 | * metadata | |
7839 | */ | |
7840 | di->disk.number = dl->index; | |
d23fe947 | 7841 | |
272906ef DW |
7842 | /* (ab)use di->devs to store a pointer to the device |
7843 | * we chose | |
7844 | */ | |
7845 | di->devs = (struct mdinfo *) dl; | |
7846 | ||
7847 | di->disk.raid_disk = i; | |
7848 | di->disk.major = dl->major; | |
7849 | di->disk.minor = dl->minor; | |
7850 | di->disk.state = 0; | |
d23534e4 | 7851 | di->recovery_start = 0; |
5551b113 | 7852 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7853 | di->component_size = a->info.component_size; |
7854 | di->container_member = inst; | |
148acb7b | 7855 | super->random = random32(); |
272906ef DW |
7856 | di->next = rv; |
7857 | rv = di; | |
7858 | num_spares++; | |
7859 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7860 | i, di->data_offset); | |
88758e9d DW |
7861 | } |
7862 | ||
7863 | if (!rv) | |
7864 | /* No spares found */ | |
7865 | return rv; | |
7866 | /* Now 'rv' has a list of devices to return. | |
7867 | * Create a metadata_update record to update the | |
7868 | * disk_ord_tbl for the array | |
7869 | */ | |
503975b9 | 7870 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 7871 | mu->buf = xcalloc(num_spares, |
503975b9 | 7872 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 7873 | mu->space = NULL; |
cb23f1f4 | 7874 | mu->space_list = NULL; |
88758e9d DW |
7875 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7876 | mu->next = *updates; | |
7877 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7878 | ||
7879 | for (di = rv ; di ; di = di->next) { | |
7880 | u->type = update_activate_spare; | |
d23fe947 DW |
7881 | u->dl = (struct dl *) di->devs; |
7882 | di->devs = NULL; | |
88758e9d DW |
7883 | u->slot = di->disk.raid_disk; |
7884 | u->array = inst; | |
7885 | u->next = u + 1; | |
7886 | u++; | |
7887 | } | |
7888 | (u-1)->next = NULL; | |
7889 | *updates = mu; | |
7890 | ||
7891 | return rv; | |
7892 | } | |
7893 | ||
54c2c1ea | 7894 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7895 | { |
54c2c1ea | 7896 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7897 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7898 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7899 | struct disk_info *inf = get_disk_info(u); |
7900 | struct imsm_disk *disk; | |
8273f55e DW |
7901 | int i; |
7902 | int j; | |
8273f55e | 7903 | |
54c2c1ea | 7904 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7905 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7906 | for (j = 0; j < new_map->num_members; j++) |
7907 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7908 | return 1; |
7909 | } | |
7910 | ||
7911 | return 0; | |
7912 | } | |
7913 | ||
1a64be56 LM |
7914 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
7915 | { | |
594dc1b8 JS |
7916 | struct dl *dl; |
7917 | ||
1a64be56 LM |
7918 | for (dl = super->disks; dl; dl = dl->next) |
7919 | if ((dl->major == major) && (dl->minor == minor)) | |
7920 | return dl; | |
7921 | return NULL; | |
7922 | } | |
7923 | ||
7924 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7925 | { | |
594dc1b8 | 7926 | struct dl *prev; |
1a64be56 LM |
7927 | struct dl *dl; |
7928 | ||
7929 | prev = NULL; | |
7930 | for (dl = super->disks; dl; dl = dl->next) { | |
7931 | if ((dl->major == major) && (dl->minor == minor)) { | |
7932 | /* remove */ | |
7933 | if (prev) | |
7934 | prev->next = dl->next; | |
7935 | else | |
7936 | super->disks = dl->next; | |
7937 | dl->next = NULL; | |
7938 | __free_imsm_disk(dl); | |
1ade5cc1 | 7939 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
7940 | break; |
7941 | } | |
7942 | prev = dl; | |
7943 | } | |
7944 | return 0; | |
7945 | } | |
7946 | ||
f21e18ca | 7947 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7948 | |
1a64be56 LM |
7949 | static int add_remove_disk_update(struct intel_super *super) |
7950 | { | |
7951 | int check_degraded = 0; | |
594dc1b8 JS |
7952 | struct dl *disk; |
7953 | ||
1a64be56 LM |
7954 | /* add/remove some spares to/from the metadata/contrainer */ |
7955 | while (super->disk_mgmt_list) { | |
7956 | struct dl *disk_cfg; | |
7957 | ||
7958 | disk_cfg = super->disk_mgmt_list; | |
7959 | super->disk_mgmt_list = disk_cfg->next; | |
7960 | disk_cfg->next = NULL; | |
7961 | ||
7962 | if (disk_cfg->action == DISK_ADD) { | |
7963 | disk_cfg->next = super->disks; | |
7964 | super->disks = disk_cfg; | |
7965 | check_degraded = 1; | |
1ade5cc1 N |
7966 | dprintf("added %x:%x\n", |
7967 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
7968 | } else if (disk_cfg->action == DISK_REMOVE) { |
7969 | dprintf("Disk remove action processed: %x.%x\n", | |
7970 | disk_cfg->major, disk_cfg->minor); | |
7971 | disk = get_disk_super(super, | |
7972 | disk_cfg->major, | |
7973 | disk_cfg->minor); | |
7974 | if (disk) { | |
7975 | /* store action status */ | |
7976 | disk->action = DISK_REMOVE; | |
7977 | /* remove spare disks only */ | |
7978 | if (disk->index == -1) { | |
7979 | remove_disk_super(super, | |
7980 | disk_cfg->major, | |
7981 | disk_cfg->minor); | |
7982 | } | |
7983 | } | |
7984 | /* release allocate disk structure */ | |
7985 | __free_imsm_disk(disk_cfg); | |
7986 | } | |
7987 | } | |
7988 | return check_degraded; | |
7989 | } | |
7990 | ||
a29911da PC |
7991 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
7992 | struct intel_super *super, | |
7993 | void ***space_list) | |
7994 | { | |
7995 | struct intel_dev *id; | |
7996 | void **tofree = NULL; | |
7997 | int ret_val = 0; | |
7998 | ||
1ade5cc1 | 7999 | dprintf("(enter)\n"); |
a29911da PC |
8000 | if ((u->subdev < 0) || |
8001 | (u->subdev > 1)) { | |
8002 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
8003 | return ret_val; | |
8004 | } | |
8005 | if ((space_list == NULL) || (*space_list == NULL)) { | |
8006 | dprintf("imsm: Error: Memory is not allocated\n"); | |
8007 | return ret_val; | |
8008 | } | |
8009 | ||
8010 | for (id = super->devlist ; id; id = id->next) { | |
8011 | if (id->index == (unsigned)u->subdev) { | |
8012 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8013 | struct imsm_map *map; | |
8014 | struct imsm_dev *new_dev = | |
8015 | (struct imsm_dev *)*space_list; | |
238c0a71 | 8016 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
8017 | int to_state; |
8018 | struct dl *new_disk; | |
8019 | ||
8020 | if (new_dev == NULL) | |
8021 | return ret_val; | |
8022 | *space_list = **space_list; | |
8023 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 8024 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
8025 | if (migr_map) { |
8026 | dprintf("imsm: Error: migration in progress"); | |
8027 | return ret_val; | |
8028 | } | |
8029 | ||
8030 | to_state = map->map_state; | |
8031 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
8032 | map->num_members++; | |
8033 | /* this should not happen */ | |
8034 | if (u->new_disks[0] < 0) { | |
8035 | map->failed_disk_num = | |
8036 | map->num_members - 1; | |
8037 | to_state = IMSM_T_STATE_DEGRADED; | |
8038 | } else | |
8039 | to_state = IMSM_T_STATE_NORMAL; | |
8040 | } | |
8e59f3d8 | 8041 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
8042 | if (u->new_level > -1) |
8043 | map->raid_level = u->new_level; | |
238c0a71 | 8044 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
8045 | if ((u->new_level == 5) && |
8046 | (migr_map->raid_level == 0)) { | |
8047 | int ord = map->num_members - 1; | |
8048 | migr_map->num_members--; | |
8049 | if (u->new_disks[0] < 0) | |
8050 | ord |= IMSM_ORD_REBUILD; | |
8051 | set_imsm_ord_tbl_ent(map, | |
8052 | map->num_members - 1, | |
8053 | ord); | |
8054 | } | |
8055 | id->dev = new_dev; | |
8056 | tofree = (void **)dev; | |
8057 | ||
4bba0439 PC |
8058 | /* update chunk size |
8059 | */ | |
8060 | if (u->new_chunksize > 0) | |
8061 | map->blocks_per_strip = | |
8062 | __cpu_to_le16(u->new_chunksize * 2); | |
8063 | ||
a29911da PC |
8064 | /* add disk |
8065 | */ | |
8066 | if ((u->new_level != 5) || | |
8067 | (migr_map->raid_level != 0) || | |
8068 | (migr_map->raid_level == map->raid_level)) | |
8069 | goto skip_disk_add; | |
8070 | ||
8071 | if (u->new_disks[0] >= 0) { | |
8072 | /* use passes spare | |
8073 | */ | |
8074 | new_disk = get_disk_super(super, | |
8075 | major(u->new_disks[0]), | |
8076 | minor(u->new_disks[0])); | |
7a862a02 | 8077 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
8078 | major(u->new_disks[0]), |
8079 | minor(u->new_disks[0]), | |
8080 | new_disk, new_disk->index); | |
8081 | if (new_disk == NULL) | |
8082 | goto error_disk_add; | |
8083 | ||
8084 | new_disk->index = map->num_members - 1; | |
8085 | /* slot to fill in autolayout | |
8086 | */ | |
8087 | new_disk->raiddisk = new_disk->index; | |
8088 | new_disk->disk.status |= CONFIGURED_DISK; | |
8089 | new_disk->disk.status &= ~SPARE_DISK; | |
8090 | } else | |
8091 | goto error_disk_add; | |
8092 | ||
8093 | skip_disk_add: | |
8094 | *tofree = *space_list; | |
8095 | /* calculate new size | |
8096 | */ | |
f3871fdc | 8097 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
8098 | |
8099 | ret_val = 1; | |
8100 | } | |
8101 | } | |
8102 | ||
8103 | if (tofree) | |
8104 | *space_list = tofree; | |
8105 | return ret_val; | |
8106 | ||
8107 | error_disk_add: | |
8108 | dprintf("Error: imsm: Cannot find disk.\n"); | |
8109 | return ret_val; | |
8110 | } | |
8111 | ||
f3871fdc AK |
8112 | static int apply_size_change_update(struct imsm_update_size_change *u, |
8113 | struct intel_super *super) | |
8114 | { | |
8115 | struct intel_dev *id; | |
8116 | int ret_val = 0; | |
8117 | ||
1ade5cc1 | 8118 | dprintf("(enter)\n"); |
f3871fdc AK |
8119 | if ((u->subdev < 0) || |
8120 | (u->subdev > 1)) { | |
8121 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
8122 | return ret_val; | |
8123 | } | |
8124 | ||
8125 | for (id = super->devlist ; id; id = id->next) { | |
8126 | if (id->index == (unsigned)u->subdev) { | |
8127 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8128 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8129 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
8130 | unsigned long long blocks_per_member; | |
8131 | ||
8132 | /* calculate new size | |
8133 | */ | |
8134 | blocks_per_member = u->new_size / used_disks; | |
1ade5cc1 | 8135 | dprintf("(size: %llu, blocks per member: %llu)\n", |
f3871fdc AK |
8136 | u->new_size, blocks_per_member); |
8137 | set_blocks_per_member(map, blocks_per_member); | |
8138 | imsm_set_array_size(dev, u->new_size); | |
8139 | ||
8140 | ret_val = 1; | |
8141 | break; | |
8142 | } | |
8143 | } | |
8144 | ||
8145 | return ret_val; | |
8146 | } | |
8147 | ||
061d7da3 | 8148 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 8149 | struct intel_super *super, |
061d7da3 LO |
8150 | struct active_array *active_array) |
8151 | { | |
8152 | struct imsm_super *mpb = super->anchor; | |
8153 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 8154 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
8155 | struct imsm_map *migr_map; |
8156 | struct active_array *a; | |
8157 | struct imsm_disk *disk; | |
8158 | __u8 to_state; | |
8159 | struct dl *dl; | |
8160 | unsigned int found; | |
8161 | int failed; | |
5961eeec | 8162 | int victim; |
061d7da3 | 8163 | int i; |
5961eeec | 8164 | int second_map_created = 0; |
061d7da3 | 8165 | |
5961eeec | 8166 | for (; u; u = u->next) { |
238c0a71 | 8167 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 8168 | |
5961eeec | 8169 | if (victim < 0) |
8170 | return 0; | |
061d7da3 | 8171 | |
5961eeec | 8172 | for (dl = super->disks; dl; dl = dl->next) |
8173 | if (dl == u->dl) | |
8174 | break; | |
061d7da3 | 8175 | |
5961eeec | 8176 | if (!dl) { |
7a862a02 | 8177 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 8178 | u->dl->index); |
8179 | return 0; | |
8180 | } | |
061d7da3 | 8181 | |
5961eeec | 8182 | /* count failures (excluding rebuilds and the victim) |
8183 | * to determine map[0] state | |
8184 | */ | |
8185 | failed = 0; | |
8186 | for (i = 0; i < map->num_members; i++) { | |
8187 | if (i == u->slot) | |
8188 | continue; | |
8189 | disk = get_imsm_disk(super, | |
238c0a71 | 8190 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 8191 | if (!disk || is_failed(disk)) |
8192 | failed++; | |
8193 | } | |
061d7da3 | 8194 | |
5961eeec | 8195 | /* adding a pristine spare, assign a new index */ |
8196 | if (dl->index < 0) { | |
8197 | dl->index = super->anchor->num_disks; | |
8198 | super->anchor->num_disks++; | |
8199 | } | |
8200 | disk = &dl->disk; | |
8201 | disk->status |= CONFIGURED_DISK; | |
8202 | disk->status &= ~SPARE_DISK; | |
8203 | ||
8204 | /* mark rebuild */ | |
238c0a71 | 8205 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8206 | if (!second_map_created) { |
8207 | second_map_created = 1; | |
8208 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8209 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8210 | } else | |
8211 | map->map_state = to_state; | |
238c0a71 | 8212 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8213 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8214 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8215 | dl->index | IMSM_ORD_REBUILD); | |
8216 | ||
8217 | /* update the family_num to mark a new container | |
8218 | * generation, being careful to record the existing | |
8219 | * family_num in orig_family_num to clean up after | |
8220 | * earlier mdadm versions that neglected to set it. | |
8221 | */ | |
8222 | if (mpb->orig_family_num == 0) | |
8223 | mpb->orig_family_num = mpb->family_num; | |
8224 | mpb->family_num += super->random; | |
8225 | ||
8226 | /* count arrays using the victim in the metadata */ | |
8227 | found = 0; | |
8228 | for (a = active_array; a ; a = a->next) { | |
8229 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8230 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8231 | |
5961eeec | 8232 | if (get_imsm_disk_slot(map, victim) >= 0) |
8233 | found++; | |
8234 | } | |
061d7da3 | 8235 | |
5961eeec | 8236 | /* delete the victim if it is no longer being |
8237 | * utilized anywhere | |
061d7da3 | 8238 | */ |
5961eeec | 8239 | if (!found) { |
8240 | struct dl **dlp; | |
061d7da3 | 8241 | |
5961eeec | 8242 | /* We know that 'manager' isn't touching anything, |
8243 | * so it is safe to delete | |
8244 | */ | |
8245 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8246 | if ((*dlp)->index == victim) |
8247 | break; | |
5961eeec | 8248 | |
8249 | /* victim may be on the missing list */ | |
8250 | if (!*dlp) | |
8251 | for (dlp = &super->missing; *dlp; | |
8252 | dlp = &(*dlp)->next) | |
8253 | if ((*dlp)->index == victim) | |
8254 | break; | |
8255 | imsm_delete(super, dlp, victim); | |
8256 | } | |
061d7da3 LO |
8257 | } |
8258 | ||
8259 | return 1; | |
8260 | } | |
a29911da | 8261 | |
2e5dc010 N |
8262 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8263 | struct intel_super *super, | |
8264 | void ***space_list) | |
8265 | { | |
8266 | struct dl *new_disk; | |
8267 | struct intel_dev *id; | |
8268 | int i; | |
8269 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8270 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8271 | void **tofree = NULL; |
8272 | int devices_to_reshape = 1; | |
8273 | struct imsm_super *mpb = super->anchor; | |
8274 | int ret_val = 0; | |
d098291a | 8275 | unsigned int dev_id; |
2e5dc010 | 8276 | |
1ade5cc1 | 8277 | dprintf("(enter)\n"); |
2e5dc010 N |
8278 | |
8279 | /* enable spares to use in array */ | |
8280 | for (i = 0; i < delta_disks; i++) { | |
8281 | new_disk = get_disk_super(super, | |
8282 | major(u->new_disks[i]), | |
8283 | minor(u->new_disks[i])); | |
7a862a02 | 8284 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
8285 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8286 | new_disk, new_disk->index); | |
8287 | if ((new_disk == NULL) || | |
8288 | ((new_disk->index >= 0) && | |
8289 | (new_disk->index < u->old_raid_disks))) | |
8290 | goto update_reshape_exit; | |
ee4beede | 8291 | new_disk->index = disk_count++; |
2e5dc010 N |
8292 | /* slot to fill in autolayout |
8293 | */ | |
8294 | new_disk->raiddisk = new_disk->index; | |
8295 | new_disk->disk.status |= | |
8296 | CONFIGURED_DISK; | |
8297 | new_disk->disk.status &= ~SPARE_DISK; | |
8298 | } | |
8299 | ||
ed7333bd AK |
8300 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8301 | mpb->num_raid_devs); | |
2e5dc010 N |
8302 | /* manage changes in volume |
8303 | */ | |
d098291a | 8304 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8305 | void **sp = *space_list; |
8306 | struct imsm_dev *newdev; | |
8307 | struct imsm_map *newmap, *oldmap; | |
8308 | ||
d098291a AK |
8309 | for (id = super->devlist ; id; id = id->next) { |
8310 | if (id->index == dev_id) | |
8311 | break; | |
8312 | } | |
8313 | if (id == NULL) | |
8314 | break; | |
2e5dc010 N |
8315 | if (!sp) |
8316 | continue; | |
8317 | *space_list = *sp; | |
8318 | newdev = (void*)sp; | |
8319 | /* Copy the dev, but not (all of) the map */ | |
8320 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8321 | oldmap = get_imsm_map(id->dev, MAP_0); |
8322 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8323 | /* Copy the current map */ |
8324 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8325 | /* update one device only | |
8326 | */ | |
8327 | if (devices_to_reshape) { | |
ed7333bd AK |
8328 | dprintf("imsm: modifying subdev: %i\n", |
8329 | id->index); | |
2e5dc010 N |
8330 | devices_to_reshape--; |
8331 | newdev->vol.migr_state = 1; | |
8332 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8333 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8334 | newmap->num_members = u->new_raid_disks; |
8335 | for (i = 0; i < delta_disks; i++) { | |
8336 | set_imsm_ord_tbl_ent(newmap, | |
8337 | u->old_raid_disks + i, | |
8338 | u->old_raid_disks + i); | |
8339 | } | |
8340 | /* New map is correct, now need to save old map | |
8341 | */ | |
238c0a71 | 8342 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8343 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8344 | ||
f3871fdc | 8345 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8346 | } |
8347 | ||
8348 | sp = (void **)id->dev; | |
8349 | id->dev = newdev; | |
8350 | *sp = tofree; | |
8351 | tofree = sp; | |
8e59f3d8 AK |
8352 | |
8353 | /* Clear migration record */ | |
8354 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8355 | } |
819bc634 AK |
8356 | if (tofree) |
8357 | *space_list = tofree; | |
2e5dc010 N |
8358 | ret_val = 1; |
8359 | ||
8360 | update_reshape_exit: | |
8361 | ||
8362 | return ret_val; | |
8363 | } | |
8364 | ||
bb025c2f | 8365 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8366 | struct intel_super *super, |
8367 | void ***space_list) | |
bb025c2f KW |
8368 | { |
8369 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8370 | struct intel_dev *dv; |
8371 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8372 | struct imsm_map *map; |
8373 | struct dl *dm, *du; | |
8ca6df95 | 8374 | int i; |
bb025c2f KW |
8375 | |
8376 | for (dv = super->devlist; dv; dv = dv->next) | |
8377 | if (dv->index == (unsigned int)u->subarray) { | |
8378 | dev = dv->dev; | |
8379 | break; | |
8380 | } | |
8381 | ||
8382 | if (dev == NULL) | |
8383 | return 0; | |
8384 | ||
238c0a71 | 8385 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8386 | |
8387 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8388 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8389 | if (imsm_count_failed(super, dev, MAP_0) != |
8390 | (map->num_members / 2)) | |
43d5ec18 KW |
8391 | return 0; |
8392 | ||
bb025c2f KW |
8393 | /* iterate through devices to mark removed disks as spare */ |
8394 | for (dm = super->disks; dm; dm = dm->next) { | |
8395 | if (dm->disk.status & FAILED_DISK) { | |
8396 | int idx = dm->index; | |
8397 | /* update indexes on the disk list */ | |
8398 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8399 | the index values will end up being correct.... NB */ | |
8400 | for (du = super->disks; du; du = du->next) | |
8401 | if (du->index > idx) | |
8402 | du->index--; | |
8403 | /* mark as spare disk */ | |
a8619d23 | 8404 | mark_spare(dm); |
bb025c2f KW |
8405 | } |
8406 | } | |
bb025c2f KW |
8407 | /* update map */ |
8408 | map->num_members = map->num_members / 2; | |
8409 | map->map_state = IMSM_T_STATE_NORMAL; | |
8410 | map->num_domains = 1; | |
8411 | map->raid_level = 0; | |
8412 | map->failed_disk_num = -1; | |
8413 | } | |
8414 | ||
8ca6df95 KW |
8415 | if (u->direction == R0_TO_R10) { |
8416 | void **space; | |
8417 | /* update slots in current disk list */ | |
8418 | for (dm = super->disks; dm; dm = dm->next) { | |
8419 | if (dm->index >= 0) | |
8420 | dm->index *= 2; | |
8421 | } | |
8422 | /* create new *missing* disks */ | |
8423 | for (i = 0; i < map->num_members; i++) { | |
8424 | space = *space_list; | |
8425 | if (!space) | |
8426 | continue; | |
8427 | *space_list = *space; | |
8428 | du = (void *)space; | |
8429 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8430 | du->fd = -1; |
8431 | du->minor = 0; | |
8432 | du->major = 0; | |
8433 | du->index = (i * 2) + 1; | |
8434 | sprintf((char *)du->disk.serial, | |
8435 | " MISSING_%d", du->index); | |
8436 | sprintf((char *)du->serial, | |
8437 | "MISSING_%d", du->index); | |
8438 | du->next = super->missing; | |
8439 | super->missing = du; | |
8440 | } | |
8441 | /* create new dev and map */ | |
8442 | space = *space_list; | |
8443 | if (!space) | |
8444 | return 0; | |
8445 | *space_list = *space; | |
8446 | dev_new = (void *)space; | |
8447 | memcpy(dev_new, dev, sizeof(*dev)); | |
8448 | /* update new map */ | |
238c0a71 | 8449 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8450 | map->num_members = map->num_members * 2; |
1a2487c2 | 8451 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8452 | map->num_domains = 2; |
8453 | map->raid_level = 1; | |
8454 | /* replace dev<->dev_new */ | |
8455 | dv->dev = dev_new; | |
8456 | } | |
bb025c2f KW |
8457 | /* update disk order table */ |
8458 | for (du = super->disks; du; du = du->next) | |
8459 | if (du->index >= 0) | |
8460 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8461 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8462 | if (du->index >= 0) { |
8463 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8464 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8465 | } |
bb025c2f KW |
8466 | |
8467 | return 1; | |
8468 | } | |
8469 | ||
e8319a19 DW |
8470 | static void imsm_process_update(struct supertype *st, |
8471 | struct metadata_update *update) | |
8472 | { | |
8473 | /** | |
8474 | * crack open the metadata_update envelope to find the update record | |
8475 | * update can be one of: | |
d195167d AK |
8476 | * update_reshape_container_disks - all the arrays in the container |
8477 | * are being reshaped to have more devices. We need to mark | |
8478 | * the arrays for general migration and convert selected spares | |
8479 | * into active devices. | |
8480 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
8481 | * device in an array, update the disk_ord_tbl. If this disk is |
8482 | * present in all member arrays then also clear the SPARE_DISK | |
8483 | * flag | |
d195167d AK |
8484 | * update_create_array |
8485 | * update_kill_array | |
8486 | * update_rename_array | |
8487 | * update_add_remove_disk | |
e8319a19 DW |
8488 | */ |
8489 | struct intel_super *super = st->sb; | |
4d7b1503 | 8490 | struct imsm_super *mpb; |
e8319a19 DW |
8491 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8492 | ||
4d7b1503 DW |
8493 | /* update requires a larger buf but the allocation failed */ |
8494 | if (super->next_len && !super->next_buf) { | |
8495 | super->next_len = 0; | |
8496 | return; | |
8497 | } | |
8498 | ||
8499 | if (super->next_buf) { | |
8500 | memcpy(super->next_buf, super->buf, super->len); | |
8501 | free(super->buf); | |
8502 | super->len = super->next_len; | |
8503 | super->buf = super->next_buf; | |
8504 | ||
8505 | super->next_len = 0; | |
8506 | super->next_buf = NULL; | |
8507 | } | |
8508 | ||
8509 | mpb = super->anchor; | |
8510 | ||
e8319a19 | 8511 | switch (type) { |
0ec5d470 AK |
8512 | case update_general_migration_checkpoint: { |
8513 | struct intel_dev *id; | |
8514 | struct imsm_update_general_migration_checkpoint *u = | |
8515 | (void *)update->buf; | |
8516 | ||
1ade5cc1 | 8517 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
8518 | |
8519 | /* find device under general migration */ | |
8520 | for (id = super->devlist ; id; id = id->next) { | |
8521 | if (is_gen_migration(id->dev)) { | |
8522 | id->dev->vol.curr_migr_unit = | |
8523 | __cpu_to_le32(u->curr_migr_unit); | |
8524 | super->updates_pending++; | |
8525 | } | |
8526 | } | |
8527 | break; | |
8528 | } | |
bb025c2f KW |
8529 | case update_takeover: { |
8530 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8531 | if (apply_takeover_update(u, super, &update->space_list)) { |
8532 | imsm_update_version_info(super); | |
bb025c2f | 8533 | super->updates_pending++; |
1a2487c2 | 8534 | } |
bb025c2f KW |
8535 | break; |
8536 | } | |
8537 | ||
78b10e66 | 8538 | case update_reshape_container_disks: { |
d195167d | 8539 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8540 | if (apply_reshape_container_disks_update( |
8541 | u, super, &update->space_list)) | |
8542 | super->updates_pending++; | |
78b10e66 N |
8543 | break; |
8544 | } | |
48c5303a | 8545 | case update_reshape_migration: { |
a29911da PC |
8546 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8547 | if (apply_reshape_migration_update( | |
8548 | u, super, &update->space_list)) | |
8549 | super->updates_pending++; | |
48c5303a PC |
8550 | break; |
8551 | } | |
f3871fdc AK |
8552 | case update_size_change: { |
8553 | struct imsm_update_size_change *u = (void *)update->buf; | |
8554 | if (apply_size_change_update(u, super)) | |
8555 | super->updates_pending++; | |
8556 | break; | |
8557 | } | |
e8319a19 | 8558 | case update_activate_spare: { |
1011e834 | 8559 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
8560 | if (apply_update_activate_spare(u, super, st->arrays)) |
8561 | super->updates_pending++; | |
8273f55e DW |
8562 | break; |
8563 | } | |
8564 | case update_create_array: { | |
8565 | /* someone wants to create a new array, we need to be aware of | |
8566 | * a few races/collisions: | |
8567 | * 1/ 'Create' called by two separate instances of mdadm | |
8568 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8569 | * devices that have since been assimilated via | |
8570 | * activate_spare. | |
8571 | * In the event this update can not be carried out mdadm will | |
8572 | * (FIX ME) notice that its update did not take hold. | |
8573 | */ | |
8574 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8575 | struct intel_dev *dv; |
8273f55e DW |
8576 | struct imsm_dev *dev; |
8577 | struct imsm_map *map, *new_map; | |
8578 | unsigned long long start, end; | |
8579 | unsigned long long new_start, new_end; | |
8580 | int i; | |
54c2c1ea DW |
8581 | struct disk_info *inf; |
8582 | struct dl *dl; | |
8273f55e DW |
8583 | |
8584 | /* handle racing creates: first come first serve */ | |
8585 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 8586 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 8587 | goto create_error; |
8273f55e DW |
8588 | } |
8589 | ||
8590 | /* check update is next in sequence */ | |
8591 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
8592 | dprintf("can not create array %d expected index %d\n", |
8593 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8594 | goto create_error; |
8273f55e DW |
8595 | } |
8596 | ||
238c0a71 | 8597 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8598 | new_start = pba_of_lba0(new_map); |
8599 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8600 | inf = get_disk_info(u); |
8273f55e DW |
8601 | |
8602 | /* handle activate_spare versus create race: | |
8603 | * check to make sure that overlapping arrays do not include | |
8604 | * overalpping disks | |
8605 | */ | |
8606 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8607 | dev = get_imsm_dev(super, i); |
238c0a71 | 8608 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8609 | start = pba_of_lba0(map); |
8610 | end = start + blocks_per_member(map); | |
8273f55e DW |
8611 | if ((new_start >= start && new_start <= end) || |
8612 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8613 | /* overlap */; |
8614 | else | |
8615 | continue; | |
8616 | ||
8617 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 8618 | dprintf("arrays overlap\n"); |
ba2de7ba | 8619 | goto create_error; |
8273f55e DW |
8620 | } |
8621 | } | |
8273f55e | 8622 | |
949c47a0 DW |
8623 | /* check that prepare update was successful */ |
8624 | if (!update->space) { | |
1ade5cc1 | 8625 | dprintf("prepare update failed\n"); |
ba2de7ba | 8626 | goto create_error; |
949c47a0 DW |
8627 | } |
8628 | ||
54c2c1ea DW |
8629 | /* check that all disks are still active before committing |
8630 | * changes. FIXME: could we instead handle this by creating a | |
8631 | * degraded array? That's probably not what the user expects, | |
8632 | * so better to drop this update on the floor. | |
8633 | */ | |
8634 | for (i = 0; i < new_map->num_members; i++) { | |
8635 | dl = serial_to_dl(inf[i].serial, super); | |
8636 | if (!dl) { | |
1ade5cc1 | 8637 | dprintf("disk disappeared\n"); |
ba2de7ba | 8638 | goto create_error; |
54c2c1ea | 8639 | } |
949c47a0 DW |
8640 | } |
8641 | ||
8273f55e | 8642 | super->updates_pending++; |
54c2c1ea DW |
8643 | |
8644 | /* convert spares to members and fixup ord_tbl */ | |
8645 | for (i = 0; i < new_map->num_members; i++) { | |
8646 | dl = serial_to_dl(inf[i].serial, super); | |
8647 | if (dl->index == -1) { | |
8648 | dl->index = mpb->num_disks; | |
8649 | mpb->num_disks++; | |
8650 | dl->disk.status |= CONFIGURED_DISK; | |
8651 | dl->disk.status &= ~SPARE_DISK; | |
8652 | } | |
8653 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8654 | } | |
8655 | ||
ba2de7ba DW |
8656 | dv = update->space; |
8657 | dev = dv->dev; | |
949c47a0 DW |
8658 | update->space = NULL; |
8659 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8660 | dv->index = u->dev_idx; |
8661 | dv->next = super->devlist; | |
8662 | super->devlist = dv; | |
8273f55e | 8663 | mpb->num_raid_devs++; |
8273f55e | 8664 | |
4d1313e9 | 8665 | imsm_update_version_info(super); |
8273f55e | 8666 | break; |
ba2de7ba DW |
8667 | create_error: |
8668 | /* mdmon knows how to release update->space, but not | |
8669 | * ((struct intel_dev *) update->space)->dev | |
8670 | */ | |
8671 | if (update->space) { | |
8672 | dv = update->space; | |
8673 | free(dv->dev); | |
8674 | } | |
8273f55e | 8675 | break; |
e8319a19 | 8676 | } |
33414a01 DW |
8677 | case update_kill_array: { |
8678 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8679 | int victim = u->dev_idx; | |
8680 | struct active_array *a; | |
8681 | struct intel_dev **dp; | |
8682 | struct imsm_dev *dev; | |
8683 | ||
8684 | /* sanity check that we are not affecting the uuid of | |
8685 | * active arrays, or deleting an active array | |
8686 | * | |
8687 | * FIXME when immutable ids are available, but note that | |
8688 | * we'll also need to fixup the invalidated/active | |
8689 | * subarray indexes in mdstat | |
8690 | */ | |
8691 | for (a = st->arrays; a; a = a->next) | |
8692 | if (a->info.container_member >= victim) | |
8693 | break; | |
8694 | /* by definition if mdmon is running at least one array | |
8695 | * is active in the container, so checking | |
8696 | * mpb->num_raid_devs is just extra paranoia | |
8697 | */ | |
8698 | dev = get_imsm_dev(super, victim); | |
8699 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8700 | dprintf("failed to delete subarray-%d\n", victim); | |
8701 | break; | |
8702 | } | |
8703 | ||
8704 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8705 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8706 | *dp = (*dp)->next; |
8707 | } else { | |
f21e18ca | 8708 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8709 | (*dp)->index--; |
8710 | dp = &(*dp)->next; | |
8711 | } | |
8712 | mpb->num_raid_devs--; | |
8713 | super->updates_pending++; | |
8714 | break; | |
8715 | } | |
aa534678 DW |
8716 | case update_rename_array: { |
8717 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8718 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8719 | int target = u->dev_idx; | |
8720 | struct active_array *a; | |
8721 | struct imsm_dev *dev; | |
8722 | ||
8723 | /* sanity check that we are not affecting the uuid of | |
8724 | * an active array | |
8725 | */ | |
8726 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8727 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8728 | for (a = st->arrays; a; a = a->next) | |
8729 | if (a->info.container_member == target) | |
8730 | break; | |
8731 | dev = get_imsm_dev(super, u->dev_idx); | |
8732 | if (a || !dev || !check_name(super, name, 1)) { | |
8733 | dprintf("failed to rename subarray-%d\n", target); | |
8734 | break; | |
8735 | } | |
8736 | ||
cdbe98cd | 8737 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8738 | super->updates_pending++; |
8739 | break; | |
8740 | } | |
1a64be56 | 8741 | case update_add_remove_disk: { |
43dad3d6 | 8742 | /* we may be able to repair some arrays if disks are |
095b8088 | 8743 | * being added, check the status of add_remove_disk |
1a64be56 LM |
8744 | * if discs has been added. |
8745 | */ | |
8746 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8747 | struct active_array *a; |
072b727f DW |
8748 | |
8749 | super->updates_pending++; | |
1a64be56 | 8750 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8751 | a->check_degraded = 1; |
8752 | } | |
43dad3d6 | 8753 | break; |
e8319a19 | 8754 | } |
1a64be56 | 8755 | default: |
7a862a02 | 8756 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 8757 | } |
e8319a19 | 8758 | } |
88758e9d | 8759 | |
bc0b9d34 PC |
8760 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8761 | ||
5fe6f031 N |
8762 | static int imsm_prepare_update(struct supertype *st, |
8763 | struct metadata_update *update) | |
8273f55e | 8764 | { |
949c47a0 | 8765 | /** |
4d7b1503 DW |
8766 | * Allocate space to hold new disk entries, raid-device entries or a new |
8767 | * mpb if necessary. The manager synchronously waits for updates to | |
8768 | * complete in the monitor, so new mpb buffers allocated here can be | |
8769 | * integrated by the monitor thread without worrying about live pointers | |
8770 | * in the manager thread. | |
8273f55e | 8771 | */ |
095b8088 | 8772 | enum imsm_update_type type; |
4d7b1503 DW |
8773 | struct intel_super *super = st->sb; |
8774 | struct imsm_super *mpb = super->anchor; | |
8775 | size_t buf_len; | |
8776 | size_t len = 0; | |
949c47a0 | 8777 | |
095b8088 N |
8778 | if (update->len < (int)sizeof(type)) |
8779 | return 0; | |
8780 | ||
8781 | type = *(enum imsm_update_type *) update->buf; | |
8782 | ||
949c47a0 | 8783 | switch (type) { |
0ec5d470 | 8784 | case update_general_migration_checkpoint: |
095b8088 N |
8785 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
8786 | return 0; | |
1ade5cc1 | 8787 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 8788 | break; |
abedf5fc KW |
8789 | case update_takeover: { |
8790 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
8791 | if (update->len < (int)sizeof(*u)) |
8792 | return 0; | |
abedf5fc KW |
8793 | if (u->direction == R0_TO_R10) { |
8794 | void **tail = (void **)&update->space_list; | |
8795 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8796 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8797 | int num_members = map->num_members; |
8798 | void *space; | |
8799 | int size, i; | |
abedf5fc KW |
8800 | /* allocate memory for added disks */ |
8801 | for (i = 0; i < num_members; i++) { | |
8802 | size = sizeof(struct dl); | |
503975b9 | 8803 | space = xmalloc(size); |
abedf5fc KW |
8804 | *tail = space; |
8805 | tail = space; | |
8806 | *tail = NULL; | |
8807 | } | |
8808 | /* allocate memory for new device */ | |
8809 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8810 | (num_members * sizeof(__u32)); | |
503975b9 N |
8811 | space = xmalloc(size); |
8812 | *tail = space; | |
8813 | tail = space; | |
8814 | *tail = NULL; | |
8815 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
8816 | } |
8817 | ||
8818 | break; | |
8819 | } | |
78b10e66 | 8820 | case update_reshape_container_disks: { |
d195167d AK |
8821 | /* Every raid device in the container is about to |
8822 | * gain some more devices, and we will enter a | |
8823 | * reconfiguration. | |
8824 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8825 | * will now hold 2 of them. | |
8826 | * Thus we need new 'struct imsm_dev' allocations sized | |
8827 | * as sizeof_imsm_dev but with more devices in both maps. | |
8828 | */ | |
8829 | struct imsm_update_reshape *u = (void *)update->buf; | |
8830 | struct intel_dev *dl; | |
8831 | void **space_tail = (void**)&update->space_list; | |
8832 | ||
095b8088 N |
8833 | if (update->len < (int)sizeof(*u)) |
8834 | return 0; | |
8835 | ||
1ade5cc1 | 8836 | dprintf("for update_reshape\n"); |
d195167d AK |
8837 | |
8838 | for (dl = super->devlist; dl; dl = dl->next) { | |
8839 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8840 | void *s; | |
d677e0b8 AK |
8841 | if (u->new_raid_disks > u->old_raid_disks) |
8842 | size += sizeof(__u32)*2* | |
8843 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8844 | s = xmalloc(size); |
d195167d AK |
8845 | *space_tail = s; |
8846 | space_tail = s; | |
8847 | *space_tail = NULL; | |
8848 | } | |
8849 | ||
8850 | len = disks_to_mpb_size(u->new_raid_disks); | |
8851 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8852 | break; |
8853 | } | |
48c5303a | 8854 | case update_reshape_migration: { |
bc0b9d34 PC |
8855 | /* for migration level 0->5 we need to add disks |
8856 | * so the same as for container operation we will copy | |
8857 | * device to the bigger location. | |
8858 | * in memory prepared device and new disk area are prepared | |
8859 | * for usage in process update | |
8860 | */ | |
8861 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8862 | struct intel_dev *id; | |
8863 | void **space_tail = (void **)&update->space_list; | |
8864 | int size; | |
8865 | void *s; | |
8866 | int current_level = -1; | |
8867 | ||
095b8088 N |
8868 | if (update->len < (int)sizeof(*u)) |
8869 | return 0; | |
8870 | ||
1ade5cc1 | 8871 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
8872 | |
8873 | /* add space for bigger array in update | |
8874 | */ | |
8875 | for (id = super->devlist; id; id = id->next) { | |
8876 | if (id->index == (unsigned)u->subdev) { | |
8877 | size = sizeof_imsm_dev(id->dev, 1); | |
8878 | if (u->new_raid_disks > u->old_raid_disks) | |
8879 | size += sizeof(__u32)*2* | |
8880 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8881 | s = xmalloc(size); |
bc0b9d34 PC |
8882 | *space_tail = s; |
8883 | space_tail = s; | |
8884 | *space_tail = NULL; | |
8885 | break; | |
8886 | } | |
8887 | } | |
8888 | if (update->space_list == NULL) | |
8889 | break; | |
8890 | ||
8891 | /* add space for disk in update | |
8892 | */ | |
8893 | size = sizeof(struct dl); | |
503975b9 | 8894 | s = xmalloc(size); |
bc0b9d34 PC |
8895 | *space_tail = s; |
8896 | space_tail = s; | |
8897 | *space_tail = NULL; | |
8898 | ||
8899 | /* add spare device to update | |
8900 | */ | |
8901 | for (id = super->devlist ; id; id = id->next) | |
8902 | if (id->index == (unsigned)u->subdev) { | |
8903 | struct imsm_dev *dev; | |
8904 | struct imsm_map *map; | |
8905 | ||
8906 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8907 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8908 | current_level = map->raid_level; |
8909 | break; | |
8910 | } | |
8911 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8912 | struct mdinfo *spares; | |
8913 | ||
8914 | spares = get_spares_for_grow(st); | |
8915 | if (spares) { | |
8916 | struct dl *dl; | |
8917 | struct mdinfo *dev; | |
8918 | ||
8919 | dev = spares->devs; | |
8920 | if (dev) { | |
8921 | u->new_disks[0] = | |
8922 | makedev(dev->disk.major, | |
8923 | dev->disk.minor); | |
8924 | dl = get_disk_super(super, | |
8925 | dev->disk.major, | |
8926 | dev->disk.minor); | |
8927 | dl->index = u->old_raid_disks; | |
8928 | dev = dev->next; | |
8929 | } | |
8930 | sysfs_free(spares); | |
8931 | } | |
8932 | } | |
8933 | len = disks_to_mpb_size(u->new_raid_disks); | |
8934 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8935 | break; |
8936 | } | |
f3871fdc | 8937 | case update_size_change: { |
095b8088 N |
8938 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
8939 | return 0; | |
8940 | break; | |
8941 | } | |
8942 | case update_activate_spare: { | |
8943 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
8944 | return 0; | |
f3871fdc AK |
8945 | break; |
8946 | } | |
949c47a0 DW |
8947 | case update_create_array: { |
8948 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8949 | struct intel_dev *dv; |
54c2c1ea | 8950 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8951 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8952 | struct dl *dl; |
8953 | struct disk_info *inf; | |
8954 | int i; | |
8955 | int activate = 0; | |
949c47a0 | 8956 | |
095b8088 N |
8957 | if (update->len < (int)sizeof(*u)) |
8958 | return 0; | |
8959 | ||
54c2c1ea DW |
8960 | inf = get_disk_info(u); |
8961 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 8962 | /* allocate a new super->devlist entry */ |
503975b9 N |
8963 | dv = xmalloc(sizeof(*dv)); |
8964 | dv->dev = xmalloc(len); | |
8965 | update->space = dv; | |
949c47a0 | 8966 | |
54c2c1ea DW |
8967 | /* count how many spares will be converted to members */ |
8968 | for (i = 0; i < map->num_members; i++) { | |
8969 | dl = serial_to_dl(inf[i].serial, super); | |
8970 | if (!dl) { | |
8971 | /* hmm maybe it failed?, nothing we can do about | |
8972 | * it here | |
8973 | */ | |
8974 | continue; | |
8975 | } | |
8976 | if (count_memberships(dl, super) == 0) | |
8977 | activate++; | |
8978 | } | |
8979 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 8980 | break; |
095b8088 N |
8981 | } |
8982 | case update_kill_array: { | |
8983 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
8984 | return 0; | |
949c47a0 DW |
8985 | break; |
8986 | } | |
095b8088 N |
8987 | case update_rename_array: { |
8988 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
8989 | return 0; | |
8990 | break; | |
8991 | } | |
8992 | case update_add_remove_disk: | |
8993 | /* no update->len needed */ | |
8994 | break; | |
8995 | default: | |
8996 | return 0; | |
949c47a0 | 8997 | } |
8273f55e | 8998 | |
4d7b1503 DW |
8999 | /* check if we need a larger metadata buffer */ |
9000 | if (super->next_buf) | |
9001 | buf_len = super->next_len; | |
9002 | else | |
9003 | buf_len = super->len; | |
9004 | ||
9005 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
9006 | /* ok we need a larger buf than what is currently allocated | |
9007 | * if this allocation fails process_update will notice that | |
9008 | * ->next_len is set and ->next_buf is NULL | |
9009 | */ | |
9010 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
9011 | if (super->next_buf) | |
9012 | free(super->next_buf); | |
9013 | ||
9014 | super->next_len = buf_len; | |
1f45a8ad DW |
9015 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
9016 | memset(super->next_buf, 0, buf_len); | |
9017 | else | |
4d7b1503 DW |
9018 | super->next_buf = NULL; |
9019 | } | |
5fe6f031 | 9020 | return 1; |
8273f55e DW |
9021 | } |
9022 | ||
ae6aad82 | 9023 | /* must be called while manager is quiesced */ |
f21e18ca | 9024 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
9025 | { |
9026 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
9027 | struct dl *iter; |
9028 | struct imsm_dev *dev; | |
9029 | struct imsm_map *map; | |
24565c9a DW |
9030 | int i, j, num_members; |
9031 | __u32 ord; | |
ae6aad82 | 9032 | |
1ade5cc1 | 9033 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
9034 | |
9035 | /* shift all indexes down one */ | |
9036 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 9037 | if (iter->index > (int)index) |
ae6aad82 | 9038 | iter->index--; |
47ee5a45 | 9039 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 9040 | if (iter->index > (int)index) |
47ee5a45 | 9041 | iter->index--; |
ae6aad82 DW |
9042 | |
9043 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
9044 | dev = get_imsm_dev(super, i); | |
238c0a71 | 9045 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
9046 | num_members = map->num_members; |
9047 | for (j = 0; j < num_members; j++) { | |
9048 | /* update ord entries being careful not to propagate | |
9049 | * ord-flags to the first map | |
9050 | */ | |
238c0a71 | 9051 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 9052 | |
24565c9a DW |
9053 | if (ord_to_idx(ord) <= index) |
9054 | continue; | |
ae6aad82 | 9055 | |
238c0a71 | 9056 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 9057 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 9058 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
9059 | if (map) |
9060 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
9061 | } |
9062 | } | |
9063 | ||
9064 | mpb->num_disks--; | |
9065 | super->updates_pending++; | |
24565c9a DW |
9066 | if (*dlp) { |
9067 | struct dl *dl = *dlp; | |
9068 | ||
9069 | *dlp = (*dlp)->next; | |
9070 | __free_imsm_disk(dl); | |
9071 | } | |
ae6aad82 | 9072 | } |
9e2d750d | 9073 | #endif /* MDASSEMBLE */ |
9a717282 AK |
9074 | |
9075 | static void close_targets(int *targets, int new_disks) | |
9076 | { | |
9077 | int i; | |
9078 | ||
9079 | if (!targets) | |
9080 | return; | |
9081 | ||
9082 | for (i = 0; i < new_disks; i++) { | |
9083 | if (targets[i] >= 0) { | |
9084 | close(targets[i]); | |
9085 | targets[i] = -1; | |
9086 | } | |
9087 | } | |
9088 | } | |
9089 | ||
9090 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
9091 | struct intel_super *super, | |
9092 | struct imsm_dev *dev) | |
9093 | { | |
9094 | switch (level) { | |
bf5cf7c7 | 9095 | case 1: |
9a717282 AK |
9096 | case 10:{ |
9097 | int ret_val = 0; | |
9098 | struct imsm_map *map; | |
9099 | int i; | |
9100 | ||
9101 | ret_val = raid_disks/2; | |
9102 | /* check map if all disks pairs not failed | |
9103 | * in both maps | |
9104 | */ | |
238c0a71 | 9105 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
9106 | for (i = 0; i < ret_val; i++) { |
9107 | int degradation = 0; | |
9108 | if (get_imsm_disk(super, i) == NULL) | |
9109 | degradation++; | |
9110 | if (get_imsm_disk(super, i + 1) == NULL) | |
9111 | degradation++; | |
9112 | if (degradation == 2) | |
9113 | return 0; | |
9114 | } | |
238c0a71 | 9115 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
9116 | /* if there is no second map |
9117 | * result can be returned | |
9118 | */ | |
9119 | if (map == NULL) | |
9120 | return ret_val; | |
9121 | /* check degradation in second map | |
9122 | */ | |
9123 | for (i = 0; i < ret_val; i++) { | |
9124 | int degradation = 0; | |
9125 | if (get_imsm_disk(super, i) == NULL) | |
9126 | degradation++; | |
9127 | if (get_imsm_disk(super, i + 1) == NULL) | |
9128 | degradation++; | |
9129 | if (degradation == 2) | |
9130 | return 0; | |
9131 | } | |
9132 | return ret_val; | |
9133 | } | |
9134 | case 5: | |
9135 | return 1; | |
9136 | case 6: | |
9137 | return 2; | |
9138 | default: | |
9139 | return 0; | |
9140 | } | |
9141 | } | |
9142 | ||
687629c2 AK |
9143 | /******************************************************************************* |
9144 | * Function: open_backup_targets | |
9145 | * Description: Function opens file descriptors for all devices given in | |
9146 | * info->devs | |
9147 | * Parameters: | |
9148 | * info : general array info | |
9149 | * raid_disks : number of disks | |
9150 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
9151 | * super : intel super for raid10 degradation check |
9152 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
9153 | * Returns: |
9154 | * 0 : success | |
9155 | * -1 : fail | |
9156 | ******************************************************************************/ | |
9a717282 AK |
9157 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
9158 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
9159 | { |
9160 | struct mdinfo *sd; | |
f627f5ad | 9161 | int i; |
9a717282 | 9162 | int opened = 0; |
f627f5ad AK |
9163 | |
9164 | for (i = 0; i < raid_disks; i++) | |
9165 | raid_fds[i] = -1; | |
687629c2 AK |
9166 | |
9167 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9168 | char *dn; | |
9169 | ||
9170 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
9171 | dprintf("disk is faulty!!\n"); | |
9172 | continue; | |
9173 | } | |
9174 | ||
9175 | if ((sd->disk.raid_disk >= raid_disks) || | |
9176 | (sd->disk.raid_disk < 0)) | |
9177 | continue; | |
9178 | ||
9179 | dn = map_dev(sd->disk.major, | |
9180 | sd->disk.minor, 1); | |
9181 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
9182 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 9183 | pr_err("cannot open component\n"); |
9a717282 | 9184 | continue; |
687629c2 | 9185 | } |
9a717282 AK |
9186 | opened++; |
9187 | } | |
9188 | /* check if maximum array degradation level is not exceeded | |
9189 | */ | |
9190 | if ((raid_disks - opened) > | |
9191 | imsm_get_allowed_degradation(info->new_level, | |
9192 | raid_disks, | |
9193 | super, dev)) { | |
e12b3daa | 9194 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
9195 | close_targets(raid_fds, raid_disks); |
9196 | return -2; | |
687629c2 AK |
9197 | } |
9198 | return 0; | |
9199 | } | |
9200 | ||
d31ad643 PB |
9201 | /******************************************************************************* |
9202 | * Function: validate_container_imsm | |
9203 | * Description: This routine validates container after assemble, | |
9204 | * eg. if devices in container are under the same controller. | |
9205 | * | |
9206 | * Parameters: | |
9207 | * info : linked list with info about devices used in array | |
9208 | * Returns: | |
9209 | * 1 : HBA mismatch | |
9210 | * 0 : Success | |
9211 | ******************************************************************************/ | |
9212 | int validate_container_imsm(struct mdinfo *info) | |
9213 | { | |
6b781d33 AP |
9214 | if (check_env("IMSM_NO_PLATFORM")) |
9215 | return 0; | |
d31ad643 | 9216 | |
6b781d33 AP |
9217 | struct sys_dev *idev; |
9218 | struct sys_dev *hba = NULL; | |
9219 | struct sys_dev *intel_devices = find_intel_devices(); | |
9220 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
9221 | info->disk.minor)); | |
9222 | ||
9223 | for (idev = intel_devices; idev; idev = idev->next) { | |
9224 | if (dev_path && strstr(dev_path, idev->path)) { | |
9225 | hba = idev; | |
9226 | break; | |
d31ad643 | 9227 | } |
6b781d33 AP |
9228 | } |
9229 | if (dev_path) | |
d31ad643 PB |
9230 | free(dev_path); |
9231 | ||
6b781d33 AP |
9232 | if (!hba) { |
9233 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
9234 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
9235 | return 1; | |
9236 | } | |
9237 | ||
9238 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
9239 | struct mdinfo *dev; | |
9240 | ||
9241 | for (dev = info->next; dev; dev = dev->next) { | |
9242 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
9243 | ||
9244 | struct sys_dev *hba2 = NULL; | |
9245 | for (idev = intel_devices; idev; idev = idev->next) { | |
9246 | if (dev_path && strstr(dev_path, idev->path)) { | |
9247 | hba2 = idev; | |
9248 | break; | |
d31ad643 PB |
9249 | } |
9250 | } | |
6b781d33 AP |
9251 | if (dev_path) |
9252 | free(dev_path); | |
9253 | ||
9254 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
9255 | get_orom_by_device_id(hba2->dev_id); | |
9256 | ||
9257 | if (hba2 && hba->type != hba2->type) { | |
9258 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
9259 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
9260 | return 1; | |
9261 | } | |
9262 | ||
60f0f54d | 9263 | if ((orom != orom2) || ((hba->type == SYS_DEV_VMD) && (hba != hba2))) { |
6b781d33 AP |
9264 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
9265 | " This operation is not supported and can lead to data loss.\n"); | |
9266 | return 1; | |
9267 | } | |
9268 | ||
9269 | if (!orom) { | |
9270 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
9271 | " This operation is not supported and can lead to data loss.\n"); | |
9272 | return 1; | |
9273 | } | |
d31ad643 | 9274 | } |
6b781d33 | 9275 | |
d31ad643 PB |
9276 | return 0; |
9277 | } | |
9e2d750d | 9278 | #ifndef MDASSEMBLE |
687629c2 AK |
9279 | /******************************************************************************* |
9280 | * Function: init_migr_record_imsm | |
9281 | * Description: Function inits imsm migration record | |
9282 | * Parameters: | |
9283 | * super : imsm internal array info | |
9284 | * dev : device under migration | |
9285 | * info : general array info to find the smallest device | |
9286 | * Returns: | |
9287 | * none | |
9288 | ******************************************************************************/ | |
9289 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
9290 | struct mdinfo *info) | |
9291 | { | |
9292 | struct intel_super *super = st->sb; | |
9293 | struct migr_record *migr_rec = super->migr_rec; | |
9294 | int new_data_disks; | |
9295 | unsigned long long dsize, dev_sectors; | |
9296 | long long unsigned min_dev_sectors = -1LLU; | |
9297 | struct mdinfo *sd; | |
9298 | char nm[30]; | |
9299 | int fd; | |
238c0a71 AK |
9300 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
9301 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 9302 | unsigned long long num_migr_units; |
3ef4403c | 9303 | unsigned long long array_blocks; |
687629c2 AK |
9304 | |
9305 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
9306 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
9307 | ||
9308 | /* only ascending reshape supported now */ | |
9309 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
9310 | ||
9311 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
9312 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
9313 | migr_rec->dest_depth_per_unit *= |
9314 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9315 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9316 | migr_rec->blocks_per_unit = |
9317 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9318 | migr_rec->dest_depth_per_unit = | |
9319 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9320 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9321 | num_migr_units = |
9322 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9323 | ||
9324 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9325 | num_migr_units++; | |
9326 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9327 | ||
9328 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9329 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9330 | ||
687629c2 AK |
9331 | /* Find the smallest dev */ |
9332 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9333 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9334 | fd = dev_open(nm, O_RDONLY); | |
9335 | if (fd < 0) | |
9336 | continue; | |
9337 | get_dev_size(fd, NULL, &dsize); | |
9338 | dev_sectors = dsize / 512; | |
9339 | if (dev_sectors < min_dev_sectors) | |
9340 | min_dev_sectors = dev_sectors; | |
9341 | close(fd); | |
9342 | } | |
9343 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9344 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9345 | ||
9346 | write_imsm_migr_rec(st); | |
9347 | ||
9348 | return; | |
9349 | } | |
9350 | ||
9351 | /******************************************************************************* | |
9352 | * Function: save_backup_imsm | |
9353 | * Description: Function saves critical data stripes to Migration Copy Area | |
9354 | * and updates the current migration unit status. | |
9355 | * Use restore_stripes() to form a destination stripe, | |
9356 | * and to write it to the Copy Area. | |
9357 | * Parameters: | |
9358 | * st : supertype information | |
aea93171 | 9359 | * dev : imsm device that backup is saved for |
687629c2 AK |
9360 | * info : general array info |
9361 | * buf : input buffer | |
687629c2 AK |
9362 | * length : length of data to backup (blocks_per_unit) |
9363 | * Returns: | |
9364 | * 0 : success | |
9365 | *, -1 : fail | |
9366 | ******************************************************************************/ | |
9367 | int save_backup_imsm(struct supertype *st, | |
9368 | struct imsm_dev *dev, | |
9369 | struct mdinfo *info, | |
9370 | void *buf, | |
687629c2 AK |
9371 | int length) |
9372 | { | |
9373 | int rv = -1; | |
9374 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
9375 | unsigned long long *target_offsets; |
9376 | int *targets; | |
687629c2 | 9377 | int i; |
238c0a71 | 9378 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9379 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9380 | int dest_layout = 0; |
9381 | int dest_chunk; | |
d1877f69 | 9382 | unsigned long long start; |
238c0a71 | 9383 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 9384 | |
503975b9 | 9385 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 9386 | |
7e45b550 AK |
9387 | for (i = 0; i < new_disks; i++) |
9388 | targets[i] = -1; | |
9389 | ||
503975b9 | 9390 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 9391 | |
d1877f69 | 9392 | start = info->reshape_progress * 512; |
687629c2 | 9393 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9394 | target_offsets[i] = (unsigned long long) |
9395 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9396 | /* move back copy area adderss, it will be moved forward |
9397 | * in restore_stripes() using start input variable | |
9398 | */ | |
9399 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9400 | } |
9401 | ||
9a717282 AK |
9402 | if (open_backup_targets(info, new_disks, targets, |
9403 | super, dev)) | |
687629c2 AK |
9404 | goto abort; |
9405 | ||
68eb8bc6 | 9406 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9407 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9408 | ||
687629c2 AK |
9409 | if (restore_stripes(targets, /* list of dest devices */ |
9410 | target_offsets, /* migration record offsets */ | |
9411 | new_disks, | |
ab724b98 AK |
9412 | dest_chunk, |
9413 | map_dest->raid_level, | |
9414 | dest_layout, | |
9415 | -1, /* source backup file descriptor */ | |
9416 | 0, /* input buf offset | |
9417 | * always 0 buf is already offseted */ | |
d1877f69 | 9418 | start, |
687629c2 AK |
9419 | length, |
9420 | buf) != 0) { | |
e7b84f9d | 9421 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
9422 | goto abort; |
9423 | } | |
9424 | ||
9425 | rv = 0; | |
9426 | ||
9427 | abort: | |
9428 | if (targets) { | |
9a717282 | 9429 | close_targets(targets, new_disks); |
687629c2 AK |
9430 | free(targets); |
9431 | } | |
9432 | free(target_offsets); | |
9433 | ||
9434 | return rv; | |
9435 | } | |
9436 | ||
9437 | /******************************************************************************* | |
9438 | * Function: save_checkpoint_imsm | |
9439 | * Description: Function called for current unit status update | |
9440 | * in the migration record. It writes it to disk. | |
9441 | * Parameters: | |
9442 | * super : imsm internal array info | |
9443 | * info : general array info | |
9444 | * Returns: | |
9445 | * 0: success | |
9446 | * 1: failure | |
0228d92c AK |
9447 | * 2: failure, means no valid migration record |
9448 | * / no general migration in progress / | |
687629c2 AK |
9449 | ******************************************************************************/ |
9450 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9451 | { | |
9452 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9453 | unsigned long long blocks_per_unit; |
9454 | unsigned long long curr_migr_unit; | |
9455 | ||
2e062e82 | 9456 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 9457 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
9458 | return 1; |
9459 | } | |
9460 | ||
f8b72ef5 AK |
9461 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9462 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9463 | dprintf("imsm: no migration in progress.\n"); |
9464 | return 2; | |
687629c2 | 9465 | } |
f8b72ef5 AK |
9466 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9467 | /* check if array is alligned to copy area | |
9468 | * if it is not alligned, add one to current migration unit value | |
9469 | * this can happend on array reshape finish only | |
9470 | */ | |
9471 | if (info->reshape_progress % blocks_per_unit) | |
9472 | curr_migr_unit++; | |
687629c2 AK |
9473 | |
9474 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9475 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9476 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9477 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9478 | __cpu_to_le32(curr_migr_unit * |
9479 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 9480 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 9481 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
9482 | return 1; |
9483 | } | |
9484 | ||
9485 | return 0; | |
9486 | } | |
9487 | ||
276d77db AK |
9488 | /******************************************************************************* |
9489 | * Function: recover_backup_imsm | |
9490 | * Description: Function recovers critical data from the Migration Copy Area | |
9491 | * while assembling an array. | |
9492 | * Parameters: | |
9493 | * super : imsm internal array info | |
9494 | * info : general array info | |
9495 | * Returns: | |
9496 | * 0 : success (or there is no data to recover) | |
9497 | * 1 : fail | |
9498 | ******************************************************************************/ | |
9499 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9500 | { | |
9501 | struct intel_super *super = st->sb; | |
9502 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 9503 | struct imsm_map *map_dest; |
276d77db AK |
9504 | struct intel_dev *id = NULL; |
9505 | unsigned long long read_offset; | |
9506 | unsigned long long write_offset; | |
9507 | unsigned unit_len; | |
9508 | int *targets = NULL; | |
9509 | int new_disks, i, err; | |
9510 | char *buf = NULL; | |
9511 | int retval = 1; | |
9512 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9513 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9514 | char buffer[20]; |
6c3560c0 | 9515 | int skipped_disks = 0; |
276d77db AK |
9516 | |
9517 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9518 | if (err < 1) | |
9519 | return 1; | |
9520 | ||
9521 | /* recover data only during assemblation */ | |
9522 | if (strncmp(buffer, "inactive", 8) != 0) | |
9523 | return 0; | |
9524 | /* no data to recover */ | |
9525 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9526 | return 0; | |
9527 | if (curr_migr_unit >= num_migr_units) | |
9528 | return 1; | |
9529 | ||
9530 | /* find device during reshape */ | |
9531 | for (id = super->devlist; id; id = id->next) | |
9532 | if (is_gen_migration(id->dev)) | |
9533 | break; | |
9534 | if (id == NULL) | |
9535 | return 1; | |
9536 | ||
238c0a71 | 9537 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9538 | new_disks = map_dest->num_members; |
9539 | ||
9540 | read_offset = (unsigned long long) | |
9541 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9542 | ||
9543 | write_offset = ((unsigned long long) | |
9544 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9545 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9546 | |
9547 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9548 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9549 | goto abort; | |
503975b9 | 9550 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 9551 | |
9a717282 | 9552 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 9553 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
9554 | goto abort; |
9555 | } | |
276d77db AK |
9556 | |
9557 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9558 | if (targets[i] < 0) { |
9559 | skipped_disks++; | |
9560 | continue; | |
9561 | } | |
276d77db | 9562 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
9563 | pr_err("Cannot seek to block: %s\n", |
9564 | strerror(errno)); | |
137debce AK |
9565 | skipped_disks++; |
9566 | continue; | |
276d77db | 9567 | } |
9ec11d1a | 9568 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9569 | pr_err("Cannot read copy area block: %s\n", |
9570 | strerror(errno)); | |
137debce AK |
9571 | skipped_disks++; |
9572 | continue; | |
276d77db AK |
9573 | } |
9574 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
9575 | pr_err("Cannot seek to block: %s\n", |
9576 | strerror(errno)); | |
137debce AK |
9577 | skipped_disks++; |
9578 | continue; | |
276d77db | 9579 | } |
9ec11d1a | 9580 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9581 | pr_err("Cannot restore block: %s\n", |
9582 | strerror(errno)); | |
137debce AK |
9583 | skipped_disks++; |
9584 | continue; | |
276d77db AK |
9585 | } |
9586 | } | |
9587 | ||
137debce AK |
9588 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9589 | new_disks, | |
9590 | super, | |
9591 | id->dev)) { | |
7a862a02 | 9592 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
9593 | goto abort; |
9594 | } | |
9595 | ||
befb629b AK |
9596 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9597 | /* ignore error == 2, this can mean end of reshape here | |
9598 | */ | |
7a862a02 | 9599 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 9600 | } else |
276d77db | 9601 | retval = 0; |
276d77db AK |
9602 | |
9603 | abort: | |
9604 | if (targets) { | |
9605 | for (i = 0; i < new_disks; i++) | |
9606 | if (targets[i]) | |
9607 | close(targets[i]); | |
9608 | free(targets); | |
9609 | } | |
9610 | free(buf); | |
9611 | return retval; | |
9612 | } | |
9613 | ||
2cda7640 ML |
9614 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9615 | ||
9616 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9617 | { | |
2cda7640 | 9618 | char disk_path[PATH_MAX]; |
96234762 LM |
9619 | char *drv=NULL; |
9620 | struct stat st; | |
2cda7640 | 9621 | |
6d8d290a | 9622 | strcpy(disk_path, disk_by_path); |
96234762 LM |
9623 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
9624 | if (stat(disk_path, &st) == 0) { | |
9625 | struct sys_dev* hba; | |
594dc1b8 | 9626 | char *path; |
96234762 LM |
9627 | |
9628 | path = devt_to_devpath(st.st_rdev); | |
9629 | if (path == NULL) | |
9630 | return "unknown"; | |
9631 | hba = find_disk_attached_hba(-1, path); | |
9632 | if (hba && hba->type == SYS_DEV_SAS) | |
9633 | drv = "isci"; | |
9634 | else if (hba && hba->type == SYS_DEV_SATA) | |
9635 | drv = "ahci"; | |
1011e834 | 9636 | else |
96234762 LM |
9637 | drv = "unknown"; |
9638 | dprintf("path: %s hba: %s attached: %s\n", | |
9639 | path, (hba) ? hba->path : "NULL", drv); | |
9640 | free(path); | |
2cda7640 | 9641 | } |
96234762 | 9642 | return drv; |
2cda7640 ML |
9643 | } |
9644 | ||
4dd2df09 | 9645 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 9646 | { |
4dd2df09 | 9647 | static char devnm[32]; |
78b10e66 N |
9648 | char subdev_name[20]; |
9649 | struct mdstat_ent *mdstat; | |
9650 | ||
9651 | sprintf(subdev_name, "%d", subdev); | |
9652 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9653 | if (!mdstat) | |
4dd2df09 | 9654 | return NULL; |
78b10e66 | 9655 | |
4dd2df09 | 9656 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 9657 | free_mdstat(mdstat); |
4dd2df09 | 9658 | return devnm; |
78b10e66 N |
9659 | } |
9660 | ||
9661 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9662 | struct geo_params *geo, | |
fbf3d202 AK |
9663 | int *old_raid_disks, |
9664 | int direction) | |
78b10e66 | 9665 | { |
694575e7 KW |
9666 | /* currently we only support increasing the number of devices |
9667 | * for a container. This increases the number of device for each | |
9668 | * member array. They must all be RAID0 or RAID5. | |
9669 | */ | |
78b10e66 N |
9670 | int ret_val = 0; |
9671 | struct mdinfo *info, *member; | |
9672 | int devices_that_can_grow = 0; | |
9673 | ||
7a862a02 | 9674 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 9675 | |
d04f65f4 | 9676 | if (geo->size > 0 || |
78b10e66 N |
9677 | geo->level != UnSet || |
9678 | geo->layout != UnSet || | |
9679 | geo->chunksize != 0 || | |
9680 | geo->raid_disks == UnSet) { | |
7a862a02 | 9681 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
9682 | return ret_val; |
9683 | } | |
9684 | ||
fbf3d202 | 9685 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 9686 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
9687 | return ret_val; |
9688 | } | |
9689 | ||
78b10e66 N |
9690 | info = container_content_imsm(st, NULL); |
9691 | for (member = info; member; member = member->next) { | |
4dd2df09 | 9692 | char *result; |
78b10e66 N |
9693 | |
9694 | dprintf("imsm: checking device_num: %i\n", | |
9695 | member->container_member); | |
9696 | ||
d7d205bd | 9697 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9698 | /* we work on container for Online Capacity Expansion |
9699 | * only so raid_disks has to grow | |
9700 | */ | |
7a862a02 | 9701 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
9702 | break; |
9703 | } | |
9704 | ||
9705 | if ((info->array.level != 0) && | |
9706 | (info->array.level != 5)) { | |
9707 | /* we cannot use this container with other raid level | |
9708 | */ | |
7a862a02 | 9709 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
9710 | info->array.level); |
9711 | break; | |
9712 | } else { | |
9713 | /* check for platform support | |
9714 | * for this raid level configuration | |
9715 | */ | |
9716 | struct intel_super *super = st->sb; | |
9717 | if (!is_raid_level_supported(super->orom, | |
9718 | member->array.level, | |
9719 | geo->raid_disks)) { | |
7a862a02 | 9720 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
9721 | info->array.level, |
9722 | geo->raid_disks, | |
9723 | geo->raid_disks > 1 ? "s" : ""); | |
9724 | break; | |
9725 | } | |
2a4a08e7 AK |
9726 | /* check if component size is aligned to chunk size |
9727 | */ | |
9728 | if (info->component_size % | |
9729 | (info->array.chunk_size/512)) { | |
7a862a02 | 9730 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
9731 | break; |
9732 | } | |
78b10e66 N |
9733 | } |
9734 | ||
9735 | if (*old_raid_disks && | |
9736 | info->array.raid_disks != *old_raid_disks) | |
9737 | break; | |
9738 | *old_raid_disks = info->array.raid_disks; | |
9739 | ||
9740 | /* All raid5 and raid0 volumes in container | |
9741 | * have to be ready for Online Capacity Expansion | |
9742 | * so they need to be assembled. We have already | |
9743 | * checked that no recovery etc is happening. | |
9744 | */ | |
4dd2df09 N |
9745 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
9746 | st->container_devnm); | |
9747 | if (result == NULL) { | |
78b10e66 N |
9748 | dprintf("imsm: cannot find array\n"); |
9749 | break; | |
9750 | } | |
9751 | devices_that_can_grow++; | |
9752 | } | |
9753 | sysfs_free(info); | |
9754 | if (!member && devices_that_can_grow) | |
9755 | ret_val = 1; | |
9756 | ||
9757 | if (ret_val) | |
1ade5cc1 | 9758 | dprintf("Container operation allowed\n"); |
78b10e66 | 9759 | else |
1ade5cc1 | 9760 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
9761 | |
9762 | return ret_val; | |
9763 | } | |
9764 | ||
9765 | /* Function: get_spares_for_grow | |
9766 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 9767 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
9768 | * Parameters: Pointer to the supertype structure |
9769 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 9770 | * NULL if fail |
78b10e66 N |
9771 | */ |
9772 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9773 | { | |
78b10e66 | 9774 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9775 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9776 | } |
9777 | ||
9778 | /****************************************************************************** | |
9779 | * function: imsm_create_metadata_update_for_reshape | |
9780 | * Function creates update for whole IMSM container. | |
9781 | * | |
9782 | ******************************************************************************/ | |
9783 | static int imsm_create_metadata_update_for_reshape( | |
9784 | struct supertype *st, | |
9785 | struct geo_params *geo, | |
9786 | int old_raid_disks, | |
9787 | struct imsm_update_reshape **updatep) | |
9788 | { | |
9789 | struct intel_super *super = st->sb; | |
9790 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
9791 | int update_memory_size; |
9792 | struct imsm_update_reshape *u; | |
9793 | struct mdinfo *spares; | |
78b10e66 | 9794 | int i; |
594dc1b8 | 9795 | int delta_disks; |
bbd24d86 | 9796 | struct mdinfo *dev; |
78b10e66 | 9797 | |
1ade5cc1 | 9798 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
9799 | |
9800 | delta_disks = geo->raid_disks - old_raid_disks; | |
9801 | ||
9802 | /* size of all update data without anchor */ | |
9803 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9804 | ||
9805 | /* now add space for spare disks that we need to add. */ | |
9806 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9807 | ||
503975b9 | 9808 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
9809 | u->type = update_reshape_container_disks; |
9810 | u->old_raid_disks = old_raid_disks; | |
9811 | u->new_raid_disks = geo->raid_disks; | |
9812 | ||
9813 | /* now get spare disks list | |
9814 | */ | |
9815 | spares = get_spares_for_grow(st); | |
9816 | ||
9817 | if (spares == NULL | |
9818 | || delta_disks > spares->array.spare_disks) { | |
7a862a02 | 9819 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 9820 | i = -1; |
78b10e66 N |
9821 | goto abort; |
9822 | } | |
9823 | ||
9824 | /* we have got spares | |
9825 | * update disk list in imsm_disk list table in anchor | |
9826 | */ | |
9827 | dprintf("imsm: %i spares are available.\n\n", | |
9828 | spares->array.spare_disks); | |
9829 | ||
bbd24d86 | 9830 | dev = spares->devs; |
78b10e66 | 9831 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9832 | struct dl *dl; |
9833 | ||
bbd24d86 AK |
9834 | if (dev == NULL) |
9835 | break; | |
78b10e66 N |
9836 | u->new_disks[i] = makedev(dev->disk.major, |
9837 | dev->disk.minor); | |
9838 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9839 | dl->index = mpb->num_disks; |
9840 | mpb->num_disks++; | |
bbd24d86 | 9841 | dev = dev->next; |
78b10e66 | 9842 | } |
78b10e66 N |
9843 | |
9844 | abort: | |
9845 | /* free spares | |
9846 | */ | |
9847 | sysfs_free(spares); | |
9848 | ||
d677e0b8 | 9849 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9850 | if (i == delta_disks) { |
1ade5cc1 | 9851 | dprintf_cont(" OK\n"); |
78b10e66 N |
9852 | *updatep = u; |
9853 | return update_memory_size; | |
9854 | } | |
9855 | free(u); | |
1ade5cc1 | 9856 | dprintf_cont(" Error\n"); |
78b10e66 N |
9857 | |
9858 | return 0; | |
9859 | } | |
9860 | ||
f3871fdc AK |
9861 | /****************************************************************************** |
9862 | * function: imsm_create_metadata_update_for_size_change() | |
9863 | * Creates update for IMSM array for array size change. | |
9864 | * | |
9865 | ******************************************************************************/ | |
9866 | static int imsm_create_metadata_update_for_size_change( | |
9867 | struct supertype *st, | |
9868 | struct geo_params *geo, | |
9869 | struct imsm_update_size_change **updatep) | |
9870 | { | |
9871 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
9872 | int update_memory_size; |
9873 | struct imsm_update_size_change *u; | |
f3871fdc | 9874 | |
1ade5cc1 | 9875 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
9876 | |
9877 | /* size of all update data without anchor */ | |
9878 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9879 | ||
503975b9 | 9880 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
9881 | u->type = update_size_change; |
9882 | u->subdev = super->current_vol; | |
9883 | u->new_size = geo->size; | |
9884 | ||
9885 | dprintf("imsm: reshape update preparation : OK\n"); | |
9886 | *updatep = u; | |
9887 | ||
9888 | return update_memory_size; | |
9889 | } | |
9890 | ||
48c5303a PC |
9891 | /****************************************************************************** |
9892 | * function: imsm_create_metadata_update_for_migration() | |
9893 | * Creates update for IMSM array. | |
9894 | * | |
9895 | ******************************************************************************/ | |
9896 | static int imsm_create_metadata_update_for_migration( | |
9897 | struct supertype *st, | |
9898 | struct geo_params *geo, | |
9899 | struct imsm_update_reshape_migration **updatep) | |
9900 | { | |
9901 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
9902 | int update_memory_size; |
9903 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
9904 | struct imsm_dev *dev; |
9905 | int previous_level = -1; | |
9906 | ||
1ade5cc1 | 9907 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
9908 | |
9909 | /* size of all update data without anchor */ | |
9910 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9911 | ||
503975b9 | 9912 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
9913 | u->type = update_reshape_migration; |
9914 | u->subdev = super->current_vol; | |
9915 | u->new_level = geo->level; | |
9916 | u->new_layout = geo->layout; | |
9917 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9918 | u->new_disks[0] = -1; | |
4bba0439 | 9919 | u->new_chunksize = -1; |
48c5303a PC |
9920 | |
9921 | dev = get_imsm_dev(super, u->subdev); | |
9922 | if (dev) { | |
9923 | struct imsm_map *map; | |
9924 | ||
238c0a71 | 9925 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9926 | if (map) { |
9927 | int current_chunk_size = | |
9928 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9929 | ||
9930 | if (geo->chunksize != current_chunk_size) { | |
9931 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 9932 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
9933 | current_chunk_size, u->new_chunksize); |
9934 | } | |
48c5303a | 9935 | previous_level = map->raid_level; |
4bba0439 | 9936 | } |
48c5303a PC |
9937 | } |
9938 | if ((geo->level == 5) && (previous_level == 0)) { | |
9939 | struct mdinfo *spares = NULL; | |
9940 | ||
9941 | u->new_raid_disks++; | |
9942 | spares = get_spares_for_grow(st); | |
9943 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9944 | free(u); | |
9945 | sysfs_free(spares); | |
9946 | update_memory_size = 0; | |
7a862a02 | 9947 | dprintf("error: cannot get spare device for requested migration"); |
48c5303a PC |
9948 | return 0; |
9949 | } | |
9950 | sysfs_free(spares); | |
9951 | } | |
9952 | dprintf("imsm: reshape update preparation : OK\n"); | |
9953 | *updatep = u; | |
9954 | ||
9955 | return update_memory_size; | |
9956 | } | |
9957 | ||
8dd70bce AK |
9958 | static void imsm_update_metadata_locally(struct supertype *st, |
9959 | void *buf, int len) | |
9960 | { | |
9961 | struct metadata_update mu; | |
9962 | ||
9963 | mu.buf = buf; | |
9964 | mu.len = len; | |
9965 | mu.space = NULL; | |
9966 | mu.space_list = NULL; | |
9967 | mu.next = NULL; | |
5fe6f031 N |
9968 | if (imsm_prepare_update(st, &mu)) |
9969 | imsm_process_update(st, &mu); | |
8dd70bce AK |
9970 | |
9971 | while (mu.space_list) { | |
9972 | void **space = mu.space_list; | |
9973 | mu.space_list = *space; | |
9974 | free(space); | |
9975 | } | |
9976 | } | |
78b10e66 | 9977 | |
471bceb6 | 9978 | /*************************************************************************** |
694575e7 | 9979 | * Function: imsm_analyze_change |
471bceb6 | 9980 | * Description: Function analyze change for single volume |
1011e834 | 9981 | * and validate if transition is supported |
fbf3d202 AK |
9982 | * Parameters: Geometry parameters, supertype structure, |
9983 | * metadata change direction (apply/rollback) | |
694575e7 | 9984 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9985 | ****************************************************************************/ |
9986 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9987 | struct geo_params *geo, |
9988 | int direction) | |
694575e7 | 9989 | { |
471bceb6 KW |
9990 | struct mdinfo info; |
9991 | int change = -1; | |
9992 | int check_devs = 0; | |
c21e737b | 9993 | int chunk; |
67a2db32 AK |
9994 | /* number of added/removed disks in operation result */ |
9995 | int devNumChange = 0; | |
9996 | /* imsm compatible layout value for array geometry verification */ | |
9997 | int imsm_layout = -1; | |
7abc9871 AK |
9998 | int data_disks; |
9999 | struct imsm_dev *dev; | |
10000 | struct intel_super *super; | |
d04f65f4 | 10001 | unsigned long long current_size; |
65d38cca | 10002 | unsigned long long free_size; |
d04f65f4 | 10003 | unsigned long long max_size; |
65d38cca | 10004 | int rv; |
471bceb6 KW |
10005 | |
10006 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
10007 | if ((geo->level != info.array.level) && |
10008 | (geo->level >= 0) && | |
10009 | (geo->level != UnSet)) { | |
10010 | switch (info.array.level) { | |
10011 | case 0: | |
10012 | if (geo->level == 5) { | |
b5347799 | 10013 | change = CH_MIGRATION; |
e13ce846 | 10014 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 10015 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
10016 | change = -1; |
10017 | goto analyse_change_exit; | |
10018 | } | |
67a2db32 | 10019 | imsm_layout = geo->layout; |
471bceb6 | 10020 | check_devs = 1; |
e91a3bad LM |
10021 | devNumChange = 1; /* parity disk added */ |
10022 | } else if (geo->level == 10) { | |
471bceb6 KW |
10023 | change = CH_TAKEOVER; |
10024 | check_devs = 1; | |
e91a3bad | 10025 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 10026 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 10027 | } |
dfe77a9e KW |
10028 | break; |
10029 | case 1: | |
471bceb6 KW |
10030 | case 10: |
10031 | if (geo->level == 0) { | |
10032 | change = CH_TAKEOVER; | |
10033 | check_devs = 1; | |
e91a3bad | 10034 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 10035 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
10036 | } |
10037 | break; | |
10038 | } | |
10039 | if (change == -1) { | |
7a862a02 | 10040 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 10041 | info.array.level, geo->level); |
471bceb6 KW |
10042 | goto analyse_change_exit; |
10043 | } | |
10044 | } else | |
10045 | geo->level = info.array.level; | |
10046 | ||
10047 | if ((geo->layout != info.array.layout) | |
10048 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 10049 | change = CH_MIGRATION; |
471bceb6 KW |
10050 | if ((info.array.layout == 0) |
10051 | && (info.array.level == 5) | |
10052 | && (geo->layout == 5)) { | |
10053 | /* reshape 5 -> 4 */ | |
10054 | } else if ((info.array.layout == 5) | |
10055 | && (info.array.level == 5) | |
10056 | && (geo->layout == 0)) { | |
10057 | /* reshape 4 -> 5 */ | |
10058 | geo->layout = 0; | |
10059 | geo->level = 5; | |
10060 | } else { | |
7a862a02 | 10061 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 10062 | info.array.layout, geo->layout); |
471bceb6 KW |
10063 | change = -1; |
10064 | goto analyse_change_exit; | |
10065 | } | |
67a2db32 | 10066 | } else { |
471bceb6 | 10067 | geo->layout = info.array.layout; |
67a2db32 AK |
10068 | if (imsm_layout == -1) |
10069 | imsm_layout = info.array.layout; | |
10070 | } | |
471bceb6 KW |
10071 | |
10072 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
10073 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 10074 | change = CH_MIGRATION; |
471bceb6 KW |
10075 | else |
10076 | geo->chunksize = info.array.chunk_size; | |
10077 | ||
c21e737b | 10078 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
10079 | |
10080 | super = st->sb; | |
10081 | dev = get_imsm_dev(super, super->current_vol); | |
10082 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 10083 | /* compute current size per disk member |
7abc9871 | 10084 | */ |
c41e00b2 AK |
10085 | current_size = info.custom_array_size / data_disks; |
10086 | ||
d5d2c614 | 10087 | if ((geo->size > 0) && (geo->size != MAX_SIZE)) { |
c41e00b2 AK |
10088 | /* align component size |
10089 | */ | |
10090 | geo->size = imsm_component_size_aligment_check( | |
10091 | get_imsm_raid_level(dev->vol.map), | |
10092 | chunk * 1024, | |
10093 | geo->size * 2); | |
65d0b4ce | 10094 | if (geo->size == 0) { |
7a862a02 | 10095 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
10096 | current_size); |
10097 | goto analyse_change_exit; | |
10098 | } | |
c41e00b2 | 10099 | } |
7abc9871 | 10100 | |
d04f65f4 | 10101 | if ((current_size != geo->size) && (geo->size > 0)) { |
7abc9871 | 10102 | if (change != -1) { |
7a862a02 | 10103 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
10104 | change = -1; |
10105 | goto analyse_change_exit; | |
10106 | } | |
10107 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 10108 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 10109 | super->current_vol, st->devnm); |
7abc9871 AK |
10110 | goto analyse_change_exit; |
10111 | } | |
65d38cca LD |
10112 | /* check the maximum available size |
10113 | */ | |
10114 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
10115 | 0, chunk, &free_size); | |
10116 | if (rv == 0) | |
10117 | /* Cannot find maximum available space | |
10118 | */ | |
10119 | max_size = 0; | |
10120 | else { | |
10121 | max_size = free_size + current_size; | |
10122 | /* align component size | |
10123 | */ | |
10124 | max_size = imsm_component_size_aligment_check( | |
10125 | get_imsm_raid_level(dev->vol.map), | |
10126 | chunk * 1024, | |
10127 | max_size); | |
10128 | } | |
d04f65f4 | 10129 | if (geo->size == MAX_SIZE) { |
b130333f AK |
10130 | /* requested size change to the maximum available size |
10131 | */ | |
65d38cca | 10132 | if (max_size == 0) { |
7a862a02 | 10133 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
10134 | change = -1; |
10135 | goto analyse_change_exit; | |
65d38cca LD |
10136 | } else |
10137 | geo->size = max_size; | |
c41e00b2 | 10138 | } |
b130333f | 10139 | |
fbf3d202 AK |
10140 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
10141 | /* accept size for rollback only | |
10142 | */ | |
10143 | } else { | |
10144 | /* round size due to metadata compatibility | |
10145 | */ | |
10146 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
10147 | << SECT_PER_MB_SHIFT; | |
10148 | dprintf("Prepare update for size change to %llu\n", | |
10149 | geo->size ); | |
10150 | if (current_size >= geo->size) { | |
7a862a02 | 10151 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10152 | current_size, geo->size); |
fbf3d202 AK |
10153 | goto analyse_change_exit; |
10154 | } | |
65d38cca | 10155 | if (max_size && geo->size > max_size) { |
7a862a02 | 10156 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10157 | max_size, geo->size); |
65d38cca LD |
10158 | goto analyse_change_exit; |
10159 | } | |
7abc9871 AK |
10160 | } |
10161 | geo->size *= data_disks; | |
10162 | geo->raid_disks = dev->vol.map->num_members; | |
10163 | change = CH_ARRAY_SIZE; | |
10164 | } | |
471bceb6 KW |
10165 | if (!validate_geometry_imsm(st, |
10166 | geo->level, | |
67a2db32 | 10167 | imsm_layout, |
e91a3bad | 10168 | geo->raid_disks + devNumChange, |
c21e737b | 10169 | &chunk, |
af4348dd | 10170 | geo->size, INVALID_SECTORS, |
471bceb6 KW |
10171 | 0, 0, 1)) |
10172 | change = -1; | |
10173 | ||
10174 | if (check_devs) { | |
10175 | struct intel_super *super = st->sb; | |
10176 | struct imsm_super *mpb = super->anchor; | |
10177 | ||
10178 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 10179 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 10180 | geo->dev_name); |
471bceb6 KW |
10181 | change = -1; |
10182 | } | |
10183 | } | |
10184 | ||
10185 | analyse_change_exit: | |
fbf3d202 AK |
10186 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
10187 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
7a862a02 | 10188 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
10189 | change = -1; |
10190 | } | |
471bceb6 | 10191 | return change; |
694575e7 KW |
10192 | } |
10193 | ||
bb025c2f KW |
10194 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
10195 | { | |
10196 | struct intel_super *super = st->sb; | |
10197 | struct imsm_update_takeover *u; | |
10198 | ||
503975b9 | 10199 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
10200 | |
10201 | u->type = update_takeover; | |
10202 | u->subarray = super->current_vol; | |
10203 | ||
10204 | /* 10->0 transition */ | |
10205 | if (geo->level == 0) | |
10206 | u->direction = R10_TO_R0; | |
10207 | ||
0529c688 KW |
10208 | /* 0->10 transition */ |
10209 | if (geo->level == 10) | |
10210 | u->direction = R0_TO_R10; | |
10211 | ||
bb025c2f KW |
10212 | /* update metadata locally */ |
10213 | imsm_update_metadata_locally(st, u, | |
10214 | sizeof(struct imsm_update_takeover)); | |
10215 | /* and possibly remotely */ | |
10216 | if (st->update_tail) | |
10217 | append_metadata_update(st, u, | |
10218 | sizeof(struct imsm_update_takeover)); | |
10219 | else | |
10220 | free(u); | |
10221 | ||
10222 | return 0; | |
10223 | } | |
10224 | ||
d04f65f4 N |
10225 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
10226 | int level, | |
78b10e66 | 10227 | int layout, int chunksize, int raid_disks, |
41784c88 | 10228 | int delta_disks, char *backup, char *dev, |
016e00f5 | 10229 | int direction, int verbose) |
78b10e66 | 10230 | { |
78b10e66 N |
10231 | int ret_val = 1; |
10232 | struct geo_params geo; | |
10233 | ||
1ade5cc1 | 10234 | dprintf("(enter)\n"); |
78b10e66 | 10235 | |
71204a50 | 10236 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
10237 | |
10238 | geo.dev_name = dev; | |
4dd2df09 | 10239 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
10240 | geo.size = size; |
10241 | geo.level = level; | |
10242 | geo.layout = layout; | |
10243 | geo.chunksize = chunksize; | |
10244 | geo.raid_disks = raid_disks; | |
41784c88 AK |
10245 | if (delta_disks != UnSet) |
10246 | geo.raid_disks += delta_disks; | |
78b10e66 | 10247 | |
1ade5cc1 N |
10248 | dprintf("for level : %i\n", geo.level); |
10249 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
10250 | |
10251 | if (experimental() == 0) | |
10252 | return ret_val; | |
10253 | ||
4dd2df09 | 10254 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
10255 | /* On container level we can only increase number of devices. */ |
10256 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 10257 | int old_raid_disks = 0; |
6dc0be30 | 10258 | |
78b10e66 | 10259 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 10260 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
10261 | struct imsm_update_reshape *u = NULL; |
10262 | int len; | |
10263 | ||
10264 | len = imsm_create_metadata_update_for_reshape( | |
10265 | st, &geo, old_raid_disks, &u); | |
10266 | ||
ed08d51c AK |
10267 | if (len <= 0) { |
10268 | dprintf("imsm: Cannot prepare update\n"); | |
10269 | goto exit_imsm_reshape_super; | |
10270 | } | |
10271 | ||
8dd70bce AK |
10272 | ret_val = 0; |
10273 | /* update metadata locally */ | |
10274 | imsm_update_metadata_locally(st, u, len); | |
10275 | /* and possibly remotely */ | |
10276 | if (st->update_tail) | |
10277 | append_metadata_update(st, u, len); | |
10278 | else | |
ed08d51c | 10279 | free(u); |
8dd70bce | 10280 | |
694575e7 | 10281 | } else { |
7a862a02 | 10282 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
10283 | } |
10284 | } else { | |
10285 | /* On volume level we support following operations | |
471bceb6 KW |
10286 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10287 | * - chunk size migration | |
10288 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10289 | */ | |
10290 | struct intel_super *super = st->sb; | |
10291 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 10292 | int change; |
694575e7 | 10293 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10294 | /* find requested device */ |
10295 | while (dev) { | |
1011e834 | 10296 | char *devnm = |
4dd2df09 N |
10297 | imsm_find_array_devnm_by_subdev( |
10298 | dev->index, st->container_devnm); | |
10299 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
10300 | break; |
10301 | dev = dev->next; | |
10302 | } | |
10303 | if (dev == NULL) { | |
4dd2df09 N |
10304 | pr_err("Cannot find %s (%s) subarray\n", |
10305 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
10306 | goto exit_imsm_reshape_super; |
10307 | } | |
10308 | super->current_vol = dev->index; | |
fbf3d202 | 10309 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10310 | switch (change) { |
471bceb6 | 10311 | case CH_TAKEOVER: |
bb025c2f | 10312 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10313 | break; |
48c5303a PC |
10314 | case CH_MIGRATION: { |
10315 | struct imsm_update_reshape_migration *u = NULL; | |
10316 | int len = | |
10317 | imsm_create_metadata_update_for_migration( | |
10318 | st, &geo, &u); | |
10319 | if (len < 1) { | |
7a862a02 | 10320 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
10321 | break; |
10322 | } | |
471bceb6 | 10323 | ret_val = 0; |
48c5303a PC |
10324 | /* update metadata locally */ |
10325 | imsm_update_metadata_locally(st, u, len); | |
10326 | /* and possibly remotely */ | |
10327 | if (st->update_tail) | |
10328 | append_metadata_update(st, u, len); | |
10329 | else | |
10330 | free(u); | |
10331 | } | |
10332 | break; | |
7abc9871 | 10333 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10334 | struct imsm_update_size_change *u = NULL; |
10335 | int len = | |
10336 | imsm_create_metadata_update_for_size_change( | |
10337 | st, &geo, &u); | |
10338 | if (len < 1) { | |
7a862a02 | 10339 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
10340 | break; |
10341 | } | |
10342 | ret_val = 0; | |
10343 | /* update metadata locally */ | |
10344 | imsm_update_metadata_locally(st, u, len); | |
10345 | /* and possibly remotely */ | |
10346 | if (st->update_tail) | |
10347 | append_metadata_update(st, u, len); | |
10348 | else | |
10349 | free(u); | |
7abc9871 AK |
10350 | } |
10351 | break; | |
471bceb6 KW |
10352 | default: |
10353 | ret_val = 1; | |
694575e7 | 10354 | } |
694575e7 | 10355 | } |
78b10e66 | 10356 | |
ed08d51c | 10357 | exit_imsm_reshape_super: |
78b10e66 N |
10358 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10359 | return ret_val; | |
10360 | } | |
2cda7640 | 10361 | |
eee67a47 AK |
10362 | /******************************************************************************* |
10363 | * Function: wait_for_reshape_imsm | |
10364 | * Description: Function writes new sync_max value and waits until | |
10365 | * reshape process reach new position | |
10366 | * Parameters: | |
10367 | * sra : general array info | |
eee67a47 AK |
10368 | * ndata : number of disks in new array's layout |
10369 | * Returns: | |
10370 | * 0 : success, | |
10371 | * 1 : there is no reshape in progress, | |
10372 | * -1 : fail | |
10373 | ******************************************************************************/ | |
ae9f01f8 | 10374 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 10375 | { |
85ca499c | 10376 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
eee67a47 | 10377 | unsigned long long completed; |
ae9f01f8 AK |
10378 | /* to_complete : new sync_max position */ |
10379 | unsigned long long to_complete = sra->reshape_progress; | |
10380 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10381 | |
ae9f01f8 | 10382 | if (fd < 0) { |
1ade5cc1 | 10383 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 10384 | return 1; |
ae9f01f8 | 10385 | } |
eee67a47 | 10386 | |
ae9f01f8 | 10387 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
1ade5cc1 | 10388 | dprintf("cannot read reshape_position (no reshape in progres)\n"); |
ae9f01f8 | 10389 | close(fd); |
c85338c6 | 10390 | return 1; |
ae9f01f8 | 10391 | } |
eee67a47 | 10392 | |
85ca499c | 10393 | if (completed > position_to_set) { |
1ade5cc1 | 10394 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 10395 | to_complete, position_to_set); |
ae9f01f8 AK |
10396 | close(fd); |
10397 | return -1; | |
10398 | } | |
10399 | dprintf("Position set: %llu\n", position_to_set); | |
10400 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10401 | position_to_set) != 0) { | |
1ade5cc1 | 10402 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
10403 | position_to_set); |
10404 | close(fd); | |
10405 | return -1; | |
eee67a47 AK |
10406 | } |
10407 | ||
eee67a47 AK |
10408 | do { |
10409 | char action[20]; | |
5ff3a780 AP |
10410 | int timeout = 3000; |
10411 | sysfs_wait(fd, &timeout); | |
a47e44fb AK |
10412 | if (sysfs_get_str(sra, NULL, "sync_action", |
10413 | action, 20) > 0 && | |
d7d3809a AP |
10414 | strncmp(action, "reshape", 7) != 0) { |
10415 | close(fd); | |
10416 | return -1; | |
10417 | } | |
eee67a47 | 10418 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
1ade5cc1 | 10419 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
10420 | close(fd); |
10421 | return 1; | |
10422 | } | |
85ca499c | 10423 | } while (completed < position_to_set); |
eee67a47 AK |
10424 | close(fd); |
10425 | return 0; | |
10426 | ||
10427 | } | |
10428 | ||
b915c95f AK |
10429 | /******************************************************************************* |
10430 | * Function: check_degradation_change | |
10431 | * Description: Check that array hasn't become failed. | |
10432 | * Parameters: | |
10433 | * info : for sysfs access | |
10434 | * sources : source disks descriptors | |
10435 | * degraded: previous degradation level | |
10436 | * Returns: | |
10437 | * degradation level | |
10438 | ******************************************************************************/ | |
10439 | int check_degradation_change(struct mdinfo *info, | |
10440 | int *sources, | |
10441 | int degraded) | |
10442 | { | |
10443 | unsigned long long new_degraded; | |
e1993023 LD |
10444 | int rv; |
10445 | ||
10446 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10447 | if ((rv == -1) || (new_degraded != (unsigned long long)degraded)) { | |
b915c95f AK |
10448 | /* check each device to ensure it is still working */ |
10449 | struct mdinfo *sd; | |
10450 | new_degraded = 0; | |
10451 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10452 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10453 | continue; | |
10454 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10455 | char sbuf[20]; | |
10456 | if (sysfs_get_str(info, | |
10457 | sd, "state", sbuf, 20) < 0 || | |
10458 | strstr(sbuf, "faulty") || | |
10459 | strstr(sbuf, "in_sync") == NULL) { | |
10460 | /* this device is dead */ | |
10461 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10462 | if (sd->disk.raid_disk >= 0 && | |
10463 | sources[sd->disk.raid_disk] >= 0) { | |
10464 | close(sources[ | |
10465 | sd->disk.raid_disk]); | |
10466 | sources[sd->disk.raid_disk] = | |
10467 | -1; | |
10468 | } | |
10469 | new_degraded++; | |
10470 | } | |
10471 | } | |
10472 | } | |
10473 | } | |
10474 | ||
10475 | return new_degraded; | |
10476 | } | |
10477 | ||
10f22854 AK |
10478 | /******************************************************************************* |
10479 | * Function: imsm_manage_reshape | |
10480 | * Description: Function finds array under reshape and it manages reshape | |
10481 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 10482 | * checkpoints. |
10f22854 AK |
10483 | * Parameters: |
10484 | * afd : Backup handle (nattive) - not used | |
10485 | * sra : general array info | |
10486 | * reshape : reshape parameters - not used | |
10487 | * st : supertype structure | |
10488 | * blocks : size of critical section [blocks] | |
10489 | * fds : table of source device descriptor | |
10490 | * offsets : start of array (offest per devices) | |
10491 | * dests : not used | |
10492 | * destfd : table of destination device descriptor | |
10493 | * destoffsets : table of destination offsets (per device) | |
10494 | * Returns: | |
10495 | * 1 : success, reshape is done | |
10496 | * 0 : fail | |
10497 | ******************************************************************************/ | |
999b4972 N |
10498 | static int imsm_manage_reshape( |
10499 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10500 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10501 | int *fds, unsigned long long *offsets, |
10502 | int dests, int *destfd, unsigned long long *destoffsets) | |
10503 | { | |
10f22854 AK |
10504 | int ret_val = 0; |
10505 | struct intel_super *super = st->sb; | |
594dc1b8 | 10506 | struct intel_dev *dv; |
10f22854 | 10507 | struct imsm_dev *dev = NULL; |
a6b6d984 | 10508 | struct imsm_map *map_src; |
10f22854 AK |
10509 | int migr_vol_qan = 0; |
10510 | int ndata, odata; /* [bytes] */ | |
10511 | int chunk; /* [bytes] */ | |
10512 | struct migr_record *migr_rec; | |
10513 | char *buf = NULL; | |
10514 | unsigned int buf_size; /* [bytes] */ | |
10515 | unsigned long long max_position; /* array size [bytes] */ | |
10516 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10517 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10518 | unsigned long long start_src; /* [bytes] */ |
10519 | unsigned long long start; /* [bytes] */ | |
10520 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10521 | int degraded = 0; |
ab724b98 | 10522 | int source_layout = 0; |
10f22854 | 10523 | |
79a16a9b JS |
10524 | if (!sra) |
10525 | return ret_val; | |
10526 | ||
10527 | if (!fds || !offsets) | |
10f22854 AK |
10528 | goto abort; |
10529 | ||
10530 | /* Find volume during the reshape */ | |
10531 | for (dv = super->devlist; dv; dv = dv->next) { | |
10532 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10533 | && dv->dev->vol.migr_state == 1) { | |
10534 | dev = dv->dev; | |
10535 | migr_vol_qan++; | |
10536 | } | |
10537 | } | |
10538 | /* Only one volume can migrate at the same time */ | |
10539 | if (migr_vol_qan != 1) { | |
e7b84f9d | 10540 | pr_err(": %s", migr_vol_qan ? |
10f22854 AK |
10541 | "Number of migrating volumes greater than 1\n" : |
10542 | "There is no volume during migrationg\n"); | |
10543 | goto abort; | |
10544 | } | |
10545 | ||
238c0a71 | 10546 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10547 | if (map_src == NULL) |
10548 | goto abort; | |
10f22854 | 10549 | |
238c0a71 AK |
10550 | ndata = imsm_num_data_members(dev, MAP_0); |
10551 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10552 | |
7b1ab482 | 10553 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10554 | old_data_stripe_length = odata * chunk; |
10555 | ||
10556 | migr_rec = super->migr_rec; | |
10557 | ||
10f22854 AK |
10558 | /* initialize migration record for start condition */ |
10559 | if (sra->reshape_progress == 0) | |
10560 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10561 | else { |
10562 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 10563 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
10564 | goto abort; |
10565 | } | |
6a75c8ca AK |
10566 | /* Save checkpoint to update migration record for current |
10567 | * reshape position (in md). It can be farther than current | |
10568 | * reshape position in metadata. | |
10569 | */ | |
10570 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10571 | /* ignore error == 2, this can mean end of reshape here | |
10572 | */ | |
7a862a02 | 10573 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
10574 | goto abort; |
10575 | } | |
b2c59438 | 10576 | } |
10f22854 AK |
10577 | |
10578 | /* size for data */ | |
10579 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10580 | /* extend buffer size for parity disk */ | |
10581 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10582 | /* add space for stripe aligment */ | |
10583 | buf_size += old_data_stripe_length; | |
10584 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10585 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10586 | goto abort; | |
10587 | } | |
10588 | ||
3ef4403c | 10589 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10590 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10591 | |
10592 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10593 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10594 | /* current reshape position [blocks] */ | |
10595 | unsigned long long current_position = | |
10596 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10597 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10598 | unsigned long long border; | |
10599 | ||
b915c95f AK |
10600 | /* Check that array hasn't become failed. |
10601 | */ | |
10602 | degraded = check_degradation_change(sra, fds, degraded); | |
10603 | if (degraded > 1) { | |
7a862a02 | 10604 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
10605 | goto abort; |
10606 | } | |
10607 | ||
10f22854 AK |
10608 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10609 | ||
10610 | if ((current_position + next_step) > max_position) | |
10611 | next_step = max_position - current_position; | |
10612 | ||
92144abf | 10613 | start = current_position * 512; |
10f22854 | 10614 | |
942e1cdb | 10615 | /* align reading start to old geometry */ |
10f22854 AK |
10616 | start_buf_shift = start % old_data_stripe_length; |
10617 | start_src = start - start_buf_shift; | |
10618 | ||
10619 | border = (start_src / odata) - (start / ndata); | |
10620 | border /= 512; | |
10621 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10622 | /* save critical stripes to buf | |
10623 | * start - start address of current unit | |
10624 | * to backup [bytes] | |
10625 | * start_src - start address of current unit | |
10626 | * to backup alligned to source array | |
10627 | * [bytes] | |
10628 | */ | |
594dc1b8 | 10629 | unsigned long long next_step_filler; |
10f22854 AK |
10630 | unsigned long long copy_length = next_step * 512; |
10631 | ||
10632 | /* allign copy area length to stripe in old geometry */ | |
10633 | next_step_filler = ((copy_length + start_buf_shift) | |
10634 | % old_data_stripe_length); | |
10635 | if (next_step_filler) | |
10636 | next_step_filler = (old_data_stripe_length | |
10637 | - next_step_filler); | |
7a862a02 | 10638 | dprintf("save_stripes() parameters: start = %llu,\tstart_src = %llu,\tnext_step*512 = %llu,\tstart_in_buf_shift = %llu,\tnext_step_filler = %llu\n", |
10f22854 AK |
10639 | start, start_src, copy_length, |
10640 | start_buf_shift, next_step_filler); | |
10641 | ||
10642 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10643 | chunk, map_src->raid_level, |
10644 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10645 | copy_length + |
10646 | next_step_filler + start_buf_shift, | |
10647 | buf)) { | |
7a862a02 | 10648 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
10649 | goto abort; |
10650 | } | |
10651 | /* Convert data to destination format and store it | |
10652 | * in backup general migration area | |
10653 | */ | |
10654 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10655 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 10656 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
10657 | goto abort; |
10658 | } | |
10659 | if (save_checkpoint_imsm(st, sra, | |
10660 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 10661 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
10662 | goto abort; |
10663 | } | |
8016a6d4 AK |
10664 | } else { |
10665 | /* set next step to use whole border area */ | |
10666 | border /= next_step; | |
10667 | if (border > 1) | |
10668 | next_step *= border; | |
10f22854 AK |
10669 | } |
10670 | /* When data backed up, checkpoint stored, | |
10671 | * kick the kernel to reshape unit of data | |
10672 | */ | |
10673 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10674 | /* limit next step to array max position */ |
10675 | if (next_step > max_position) | |
10676 | next_step = max_position; | |
10f22854 AK |
10677 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10678 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10679 | sra->reshape_progress = next_step; |
10f22854 AK |
10680 | |
10681 | /* wait until reshape finish */ | |
c85338c6 | 10682 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
10683 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10684 | goto abort; | |
10685 | } | |
84d11e6c N |
10686 | if (sigterm) |
10687 | goto abort; | |
10f22854 | 10688 | |
0228d92c AK |
10689 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10690 | /* ignore error == 2, this can mean end of reshape here | |
10691 | */ | |
7a862a02 | 10692 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
10693 | goto abort; |
10694 | } | |
10695 | ||
10696 | } | |
10697 | ||
71e5411e PB |
10698 | /* clear migr_rec on disks after successful migration */ |
10699 | struct dl *d; | |
10700 | ||
10701 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); | |
10702 | for (d = super->disks; d; d = d->next) { | |
10703 | if (d->index < 0 || is_failed(&d->disk)) | |
10704 | continue; | |
10705 | unsigned long long dsize; | |
10706 | ||
10707 | get_dev_size(d->fd, NULL, &dsize); | |
10708 | if (lseek64(d->fd, dsize - MIGR_REC_POSITION, | |
10709 | SEEK_SET) >= 0) { | |
10710 | if (write(d->fd, super->migr_rec_buf, | |
10711 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
10712 | perror("Write migr_rec failed"); | |
10713 | } | |
10714 | } | |
10715 | ||
10f22854 AK |
10716 | /* return '1' if done */ |
10717 | ret_val = 1; | |
10718 | abort: | |
10719 | free(buf); | |
942e1cdb N |
10720 | /* See Grow.c: abort_reshape() for further explanation */ |
10721 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
10722 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
10723 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
10724 | |
10725 | return ret_val; | |
999b4972 | 10726 | } |
0c21b485 | 10727 | |
71204a50 | 10728 | #endif /* MDASSEMBLE */ |
999b4972 | 10729 | |
cdddbdbc DW |
10730 | struct superswitch super_imsm = { |
10731 | #ifndef MDASSEMBLE | |
10732 | .examine_super = examine_super_imsm, | |
10733 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10734 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10735 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10736 | .detail_super = detail_super_imsm, |
10737 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10738 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10739 | .validate_geometry = validate_geometry_imsm, |
10740 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10741 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10742 | .detail_platform = detail_platform_imsm, |
e50cf220 | 10743 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 10744 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10745 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10746 | .load_container = load_container_imsm, |
71204a50 N |
10747 | .default_geometry = default_geometry_imsm, |
10748 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10749 | .reshape_super = imsm_reshape_super, | |
10750 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10751 | .recover_backup = recover_backup_imsm, |
74db60b0 | 10752 | .copy_metadata = copy_metadata_imsm, |
cdddbdbc DW |
10753 | #endif |
10754 | .match_home = match_home_imsm, | |
10755 | .uuid_from_super= uuid_from_super_imsm, | |
10756 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10757 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10758 | .update_super = update_super_imsm, |
10759 | ||
10760 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10761 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10762 | |
10763 | .compare_super = compare_super_imsm, | |
10764 | ||
10765 | .load_super = load_super_imsm, | |
bf5a934a | 10766 | .init_super = init_super_imsm, |
e683ca88 | 10767 | .store_super = store_super_imsm, |
cdddbdbc DW |
10768 | .free_super = free_super_imsm, |
10769 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10770 | .container_content = container_content_imsm, |
0c21b485 | 10771 | .validate_container = validate_container_imsm, |
cdddbdbc | 10772 | |
cdddbdbc | 10773 | .external = 1, |
4cce4069 | 10774 | .name = "imsm", |
845dea95 | 10775 | |
0e600426 | 10776 | #ifndef MDASSEMBLE |
845dea95 NB |
10777 | /* for mdmon */ |
10778 | .open_new = imsm_open_new, | |
ed9d66aa | 10779 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10780 | .set_disk = imsm_set_disk, |
10781 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10782 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10783 | .process_update = imsm_process_update, |
8273f55e | 10784 | .prepare_update = imsm_prepare_update, |
0e600426 | 10785 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10786 | }; |