]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
imsm: return correct uuid for volume in detail
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
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 | \
bbab0940
TM		 84 MPB_ATTRIB_EXP_STRIPE_SIZE | \
85 MPB_ATTRIB_BBM)
418f9b36
N		 86
87/* Define attributes that are unused but not harmful */
88#define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE)
fe7ed8cb 89
8e59f3d8 90#define MPB_SECTOR_CNT 2210
611d9529
MD		 91#define IMSM_RESERVED_SECTORS 8192
92#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048
979d38be 93#define SECT_PER_MB_SHIFT 11
f36a9ecd 94#define MAX_SECTOR_SIZE 4096
c2462068
PB		 95#define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole
96 * mutliple PPL area
97 */
cdddbdbc 98
761e3bd9
N		 99/*
100 * This macro let's us ensure that no-one accidentally
101 * changes the size of a struct
102 */
103#define ASSERT_SIZE(_struct, size) \
104static inline void __assert_size_##_struct(void) \
105{ \
106 switch (0) { \
107 case 0: break; \
108 case (sizeof(struct _struct) == size): break; \
109 } \
110}
111
cdddbdbc
DW		 112/* Disk configuration info. */
113#define IMSM_MAX_DEVICES 255
114struct imsm_disk {
115 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 116 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 117 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 118#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
119#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
120#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 121#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 122 __u32 status; /* 0xF0 - 0xF3 */
1011e834 123 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 124 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
125#define IMSM_DISK_FILLERS 3
126 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 127};
761e3bd9 128ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 129
3b451610
AK		 130/* map selector for map managment
131 */
238c0a71
AK		 132#define MAP_0 0
133#define MAP_1 1
134#define MAP_X -1
3b451610 135
cdddbdbc
DW		 136/* RAID map configuration infos. */
137struct imsm_map {
5551b113
CA		 138 __u32 pba_of_lba0_lo; /* start address of partition */
139 __u32 blocks_per_member_lo;/* blocks per member */
140 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 141 __u16 blocks_per_strip;
142 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
143#define IMSM_T_STATE_NORMAL 0
144#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 145#define IMSM_T_STATE_DEGRADED 2
146#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 147 __u8 raid_level;
148#define IMSM_T_RAID0 0
149#define IMSM_T_RAID1 1
150#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
151 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 152 __u8 num_domains; /* number of parity domains */
153 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 154 __u8 ddf;
5551b113
CA		 155 __u32 pba_of_lba0_hi;
156 __u32 blocks_per_member_hi;
157 __u32 num_data_stripes_hi;
158 __u32 filler[4]; /* expansion area */
7eef0453 159#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 160 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 161 * top byte contains some flags
162 */
761e3bd9
N		 163};
164ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 165
166struct imsm_vol {
f8f603f1 167 __u32 curr_migr_unit;
fe7ed8cb 168 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 169 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 170#define MIGR_INIT 0
171#define MIGR_REBUILD 1
172#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
173#define MIGR_GEN_MIGR 3
174#define MIGR_STATE_CHANGE 4
1484e727 175#define MIGR_REPAIR 5
cdddbdbc 176 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 177#define RAIDVOL_CLEAN 0
178#define RAIDVOL_DIRTY 1
179#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 180 __u8 dirty;
fe7ed8cb
DW		 181 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
182 __u16 verify_errors; /* number of mismatches */
183 __u16 bad_blocks; /* number of bad blocks during verify */
184 __u32 filler[4];
cdddbdbc
DW		 185 struct imsm_map map[1];
186 /* here comes another one if migr_state */
761e3bd9
N		 187};
188ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 189
190struct imsm_dev {
fe7ed8cb 191 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 192 __u32 size_low;
193 __u32 size_high;
fe7ed8cb
DW		 194#define DEV_BOOTABLE __cpu_to_le32(0x01)
195#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
196#define DEV_READ_COALESCING __cpu_to_le32(0x04)
197#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
198#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
199#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
200#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
201#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
202#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
203#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
204#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
205#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
206#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 207 __u32 status; /* Persistent RaidDev status */
208 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 209 __u8 migr_priority;
210 __u8 num_sub_vols;
211 __u8 tid;
212 __u8 cng_master_disk;
213 __u16 cache_policy;
214 __u8 cng_state;
215 __u8 cng_sub_state;
2432ce9b
AP		 216 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
217
218 /* NVM_EN */
219 __u8 nv_cache_mode;
220 __u8 nv_cache_flags;
221
222 /* Unique Volume Id of the NvCache Volume associated with this volume */
223 __u32 nvc_vol_orig_family_num;
224 __u16 nvc_vol_raid_dev_num;
225
226#define RWH_OFF 0
227#define RWH_DISTRIBUTED 1
228#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 229#define RWH_MULTIPLE_DISTRIBUTED 3
230#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
231#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 232 __u8 rwh_policy; /* Raid Write Hole Policy */
233 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
234 __u8 filler1;
235
236#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 237 __u32 filler[IMSM_DEV_FILLERS];
238 struct imsm_vol vol;
761e3bd9
N		 239};
240ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 241
242struct imsm_super {
243 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
244 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
245 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
246 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
247 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 248 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
249 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 250 __u8 num_disks; /* 0x38 Number of configured disks */
251 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 252 __u8 error_log_pos; /* 0x3A */
253 __u8 fill[1]; /* 0x3B */
254 __u32 cache_size; /* 0x3c - 0x40 in mb */
255 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
256 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
257 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 258 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
259 * volume IDs for raid_dev created in this array
260 * (starts at 1)
261 */
262 __u16 filler1; /* 0x4E - 0x4F */
263#define IMSM_FILLERS 34
264 __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 265 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
266 /* here comes imsm_dev[num_raid_devs] */
604b746f 267 /* here comes BBM logs */
761e3bd9
N		 268};
269ASSERT_SIZE(imsm_super, 264)
cdddbdbc 270
604b746f 271#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 272#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
273#define BBM_LOG_SIGNATURE 0xabadb10c
274
275struct bbm_log_block_addr {
276 __u16 w1;
277 __u32 dw1;
278} __attribute__ ((__packed__));
604b746f
JD		 279
280struct bbm_log_entry {
8d67477f
TM		 281 __u8 marked_count; /* Number of blocks marked - 1 */
282 __u8 disk_ordinal; /* Disk entry within the imsm_super */
283 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 284} __attribute__ ((__packed__));
285
286struct bbm_log {
287 __u32 signature; /* 0xABADB10C */
288 __u32 entry_count;
8d67477f 289 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 290};
291ASSERT_SIZE(bbm_log, 2040)
604b746f 292
cdddbdbc 293static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 294
b53bfba6
TM		 295#define BLOCKS_PER_KB (1024/512)
296
8e59f3d8
AK		 297#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
298
299#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
300
de44e46f
PB		 301#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
302#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
303 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 304 */
305
8e59f3d8
AK		 306#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
307 * be recovered using srcMap */
308#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
309 * already been migrated and must
310 * be recovered from checkpoint area */
2432ce9b 311
c2462068 312#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 313
8e59f3d8
AK		 314struct migr_record {
315 __u32 rec_status; /* Status used to determine how to restart
316 * migration in case it aborts
317 * in some fashion */
9f421827 318 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 319 __u32 family_num; /* Family number of MPB
320 * containing the RaidDev
321 * that is migrating */
322 __u32 ascending_migr; /* True if migrating in increasing
323 * order of lbas */
324 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
325 __u32 dest_depth_per_unit; /* Num member blocks each destMap
326 * member disk
327 * advances per unit-of-operation */
9f421827
PB		 328 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
329 __u32 dest_1st_member_lba_lo; /* First member lba on first
330 * stripe of destination */
331 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 332 __u32 post_migr_vol_cap; /* Size of volume after
333 * migration completes */
334 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
335 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
336 * migration ckpt record was read from
337 * (for recovered migrations) */
9f421827
PB		 338 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
339 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
340 * high order 32 bits */
341 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
342 * destination - high order 32 bits */
343 __u32 num_migr_units_hi; /* Total num migration units-of-op
344 * high order 32 bits */
761e3bd9
N		 345};
346ASSERT_SIZE(migr_record, 64)
8e59f3d8 347
ec50f7b6
LM		 348struct md_list {
349 /* usage marker:
350 * 1: load metadata
351 * 2: metadata does not match
352 * 4: already checked
353 */
354 int used;
355 char *devname;
356 int found;
357 int container;
358 dev_t st_rdev;
359 struct md_list *next;
360};
361
e7b84f9d 362#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 363
1484e727
DW		 364static __u8 migr_type(struct imsm_dev *dev)
365{
366 if (dev->vol.migr_type == MIGR_VERIFY &&
367 dev->status & DEV_VERIFY_AND_FIX)
368 return MIGR_REPAIR;
369 else
370 return dev->vol.migr_type;
371}
372
373static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
374{
375 /* for compatibility with older oroms convert MIGR_REPAIR, into
376 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
377 */
378 if (migr_type == MIGR_REPAIR) {
379 dev->vol.migr_type = MIGR_VERIFY;
380 dev->status |= DEV_VERIFY_AND_FIX;
381 } else {
382 dev->vol.migr_type = migr_type;
383 dev->status &= ~DEV_VERIFY_AND_FIX;
384 }
385}
386
f36a9ecd 387static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 388{
f36a9ecd 389 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 390}
cdddbdbc 391
f36a9ecd
PB		 392static unsigned int mpb_sectors(struct imsm_super *mpb,
393 unsigned int sector_size)
87eb16df 394{
f36a9ecd 395 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 396}
397
ba2de7ba
DW		 398struct intel_dev {
399 struct imsm_dev *dev;
400 struct intel_dev *next;
f21e18ca 401 unsigned index;
ba2de7ba
DW		 402};
403
88654014
LM		 404struct intel_hba {
405 enum sys_dev_type type;
406 char *path;
407 char *pci_id;
408 struct intel_hba *next;
409};
410
1a64be56
LM		 411enum action {
412 DISK_REMOVE = 1,
413 DISK_ADD
414};
cdddbdbc
DW		 415/* internal representation of IMSM metadata */
416struct intel_super {
417 union {
949c47a0
DW		 418 void *buf; /* O_DIRECT buffer for reading/writing metadata */
419 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 420 };
8e59f3d8
AK		 421 union {
422 void *migr_rec_buf; /* buffer for I/O operations */
423 struct migr_record *migr_rec; /* migration record */
424 };
51d83f5d
AK		 425 int clean_migration_record_by_mdmon; /* when reshape is switched to next
426 array, it indicates that mdmon is allowed to clean migration
427 record */
949c47a0 428 size_t len; /* size of the 'buf' allocation */
bbab0940 429 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 430 void *next_buf; /* for realloc'ing buf from the manager */
431 size_t next_len;
c2c087e6 432 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 433 int current_vol; /* index of raid device undergoing creation */
5551b113 434 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 435 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 436 struct intel_dev *devlist;
fa7bb6f8 437 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 438 struct dl {
439 struct dl *next;
440 int index;
441 __u8 serial[MAX_RAID_SERIAL_LEN];
442 int major, minor;
443 char *devname;
b9f594fe 444 struct imsm_disk disk;
cdddbdbc 445 int fd;
0dcecb2e
DW		 446 int extent_cnt;
447 struct extent *e; /* for determining freespace @ create */
efb30e7f 448 int raiddisk; /* slot to fill in autolayout */
1a64be56 449 enum action action;
ca0748fa 450 } *disks, *current_disk;
1a64be56
LM		 451 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
452 active */
47ee5a45 453 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 454 struct bbm_log *bbm_log;
88654014 455 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 456 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 457 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 458 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 459};
460
461struct intel_disk {
462 struct imsm_disk disk;
463 #define IMSM_UNKNOWN_OWNER (-1)
464 int owner;
465 struct intel_disk *next;
cdddbdbc
DW		 466};
467
c2c087e6
DW		 468struct extent {
469 unsigned long long start, size;
470};
471
694575e7
KW		 472/* definitions of reshape process types */
473enum imsm_reshape_type {
474 CH_TAKEOVER,
b5347799 475 CH_MIGRATION,
7abc9871 476 CH_ARRAY_SIZE,
694575e7
KW		 477};
478
88758e9d
DW		 479/* definition of messages passed to imsm_process_update */
480enum imsm_update_type {
481 update_activate_spare,
8273f55e 482 update_create_array,
33414a01 483 update_kill_array,
aa534678 484 update_rename_array,
1a64be56 485 update_add_remove_disk,
78b10e66 486 update_reshape_container_disks,
48c5303a 487 update_reshape_migration,
2d40f3a1
AK		 488 update_takeover,
489 update_general_migration_checkpoint,
f3871fdc 490 update_size_change,
bbab0940 491 update_prealloc_badblocks_mem,
e6e9dd3f 492 update_rwh_policy,
88758e9d
DW		 493};
494
495struct imsm_update_activate_spare {
496 enum imsm_update_type type;
d23fe947 497 struct dl *dl;
88758e9d
DW		 498 int slot;
499 int array;
500 struct imsm_update_activate_spare *next;
501};
502
78b10e66 503struct geo_params {
4dd2df09 504 char devnm[32];
78b10e66 505 char *dev_name;
d04f65f4 506 unsigned long long size;
78b10e66
N		 507 int level;
508 int layout;
509 int chunksize;
510 int raid_disks;
511};
512
bb025c2f
KW		 513enum takeover_direction {
514 R10_TO_R0,
515 R0_TO_R10
516};
517struct imsm_update_takeover {
518 enum imsm_update_type type;
519 int subarray;
520 enum takeover_direction direction;
521};
78b10e66
N		 522
523struct imsm_update_reshape {
524 enum imsm_update_type type;
525 int old_raid_disks;
526 int new_raid_disks;
48c5303a
PC		 527
528 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
529};
530
531struct imsm_update_reshape_migration {
532 enum imsm_update_type type;
533 int old_raid_disks;
534 int new_raid_disks;
535 /* fields for array migration changes
536 */
537 int subdev;
538 int new_level;
539 int new_layout;
4bba0439 540 int new_chunksize;
48c5303a 541
d195167d 542 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 543};
544
f3871fdc
AK		 545struct imsm_update_size_change {
546 enum imsm_update_type type;
547 int subdev;
548 long long new_size;
549};
550
2d40f3a1
AK		 551struct imsm_update_general_migration_checkpoint {
552 enum imsm_update_type type;
553 __u32 curr_migr_unit;
554};
555
54c2c1ea
DW		 556struct disk_info {
557 __u8 serial[MAX_RAID_SERIAL_LEN];
558};
559
8273f55e
DW		 560struct imsm_update_create_array {
561 enum imsm_update_type type;
8273f55e 562 int dev_idx;
6a3e913e 563 struct imsm_dev dev;
8273f55e
DW		 564};
565
33414a01
DW		 566struct imsm_update_kill_array {
567 enum imsm_update_type type;
568 int dev_idx;
569};
570
aa534678
DW		 571struct imsm_update_rename_array {
572 enum imsm_update_type type;
573 __u8 name[MAX_RAID_SERIAL_LEN];
574 int dev_idx;
575};
576
1a64be56 577struct imsm_update_add_remove_disk {
43dad3d6
DW		 578 enum imsm_update_type type;
579};
580
bbab0940
TM		 581struct imsm_update_prealloc_bb_mem {
582 enum imsm_update_type type;
583};
584
e6e9dd3f
AP		 585struct imsm_update_rwh_policy {
586 enum imsm_update_type type;
587 int new_policy;
588 int dev_idx;
589};
590
88654014
LM		 591static const char *_sys_dev_type[] = {
592 [SYS_DEV_UNKNOWN] = "Unknown",
593 [SYS_DEV_SAS] = "SAS",
614902f6 594 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 595 [SYS_DEV_NVME] = "NVMe",
596 [SYS_DEV_VMD] = "VMD"
88654014
LM		 597};
598
599const char *get_sys_dev_type(enum sys_dev_type type)
600{
601 if (type >= SYS_DEV_MAX)
602 type = SYS_DEV_UNKNOWN;
603
604 return _sys_dev_type[type];
605}
606
607static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
608{
503975b9
N		 609 struct intel_hba *result = xmalloc(sizeof(*result));
610
611 result->type = device->type;
612 result->path = xstrdup(device->path);
613 result->next = NULL;
614 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
615 result->pci_id++;
616
88654014
LM		 617 return result;
618}
619
620static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
621{
594dc1b8
JS		 622 struct intel_hba *result;
623
88654014
LM		 624 for (result = hba; result; result = result->next) {
625 if (result->type == device->type && strcmp(result->path, device->path) == 0)
626 break;
627 }
628 return result;
629}
630
b4cf4cba 631static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 632{
633 struct intel_hba *hba;
634
635 /* check if disk attached to Intel HBA */
636 hba = find_intel_hba(super->hba, device);
637 if (hba != NULL)
638 return 1;
639 /* Check if HBA is already attached to super */
640 if (super->hba == NULL) {
641 super->hba = alloc_intel_hba(device);
642 return 1;
6b781d33
AP		 643 }
644
645 hba = super->hba;
646 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 647 * Do not support HBA types mixing
6b781d33
AP		 648 */
649 if (device->type != hba->type)
88654014 650 return 2;
6b781d33
AP		 651
652 /* Multiple same type HBAs can be used if they share the same OROM */
653 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
654
655 if (device_orom != super->orom)
656 return 2;
657
658 while (hba->next)
659 hba = hba->next;
660
661 hba->next = alloc_intel_hba(device);
662 return 1;
88654014
LM		 663}
664
665static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
666{
9bc4ae77 667 struct sys_dev *list, *elem;
88654014
LM		 668 char *disk_path;
669
670 if ((list = find_intel_devices()) == NULL)
671 return 0;
672
673 if (fd < 0)
674 disk_path = (char *) devname;
675 else
676 disk_path = diskfd_to_devpath(fd);
677
9bc4ae77 678 if (!disk_path)
88654014 679 return 0;
88654014 680
9bc4ae77
N		 681 for (elem = list; elem; elem = elem->next)
682 if (path_attached_to_hba(disk_path, elem->path))
88654014 683 return elem;
9bc4ae77 684
88654014
LM		 685 if (disk_path != devname)
686 free(disk_path);
88654014
LM		 687
688 return NULL;
689}
690
d424212e
N		 691static int find_intel_hba_capability(int fd, struct intel_super *super,
692 char *devname);
f2f5c343 693
cdddbdbc
DW		 694static struct supertype *match_metadata_desc_imsm(char *arg)
695{
696 struct supertype *st;
697
698 if (strcmp(arg, "imsm") != 0 &&
699 strcmp(arg, "default") != 0
700 )
701 return NULL;
702
503975b9 703 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 704 st->ss = &super_imsm;
705 st->max_devs = IMSM_MAX_DEVICES;
706 st->minor_version = 0;
707 st->sb = NULL;
708 return st;
709}
710
cdddbdbc
DW		 711static __u8 *get_imsm_version(struct imsm_super *mpb)
712{
713 return &mpb->sig[MPB_SIG_LEN];
714}
715
949c47a0
DW		 716/* retrieve a disk directly from the anchor when the anchor is known to be
717 * up-to-date, currently only at load time
718 */
719static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 720{
949c47a0 721 if (index >= mpb->num_disks)
cdddbdbc
DW		 722 return NULL;
723 return &mpb->disk[index];
724}
725
95d07a2c
LM		 726/* retrieve the disk description based on a index of the disk
727 * in the sub-array
728 */
729static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 730{
b9f594fe
DW		 731 struct dl *d;
732
733 for (d = super->disks; d; d = d->next)
734 if (d->index == index)
95d07a2c
LM		 735 return d;
736
737 return NULL;
738}
739/* retrieve a disk from the parsed metadata */
740static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
741{
742 struct dl *dl;
743
744 dl = get_imsm_dl_disk(super, index);
745 if (dl)
746 return &dl->disk;
747
b9f594fe 748 return NULL;
949c47a0
DW		 749}
750
751/* generate a checksum directly from the anchor when the anchor is known to be
752 * up-to-date, currently only at load or write_super after coalescing
753 */
754static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 755{
756 __u32 end = mpb->mpb_size / sizeof(end);
757 __u32 *p = (__u32 *) mpb;
758 __u32 sum = 0;
759
5d500228
N		 760 while (end--) {
761 sum += __le32_to_cpu(*p);
97f734fd
N		 762 p++;
763 }
cdddbdbc 764
5d500228 765 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 766}
767
a965f303
DW		 768static size_t sizeof_imsm_map(struct imsm_map *map)
769{
770 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
771}
772
773struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 774{
5e7b0330
AK		 775 /* A device can have 2 maps if it is in the middle of a migration.
776 * If second_map is:
238c0a71
AK		 777 * MAP_0 - we return the first map
778 * MAP_1 - we return the second map if it exists, else NULL
779 * MAP_X - we return the second map if it exists, else the first
5e7b0330 780 */
a965f303 781 struct imsm_map *map = &dev->vol.map[0];
9535fc47 782 struct imsm_map *map2 = NULL;
a965f303 783
9535fc47
AK		 784 if (dev->vol.migr_state)
785 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 786
9535fc47 787 switch (second_map) {
3b451610 788 case MAP_0:
9535fc47 789 break;
3b451610 790 case MAP_1:
9535fc47
AK		 791 map = map2;
792 break;
238c0a71 793 case MAP_X:
9535fc47
AK		 794 if (map2)
795 map = map2;
796 break;
9535fc47
AK		 797 default:
798 map = NULL;
799 }
800 return map;
5e7b0330 801
a965f303 802}
cdddbdbc 803
3393c6af
DW		 804/* return the size of the device.
805 * migr_state increases the returned size if map[0] were to be duplicated
806 */
807static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 808{
809 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 810 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 811
812 /* migrating means an additional map */
a965f303 813 if (dev->vol.migr_state)
238c0a71 814 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 815 else if (migr_state)
238c0a71 816 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 817
818 return size;
819}
820
54c2c1ea
DW		 821/* retrieve disk serial number list from a metadata update */
822static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
823{
824 void *u = update;
825 struct disk_info *inf;
826
827 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
828 sizeof_imsm_dev(&update->dev, 0);
829
830 return inf;
831}
54c2c1ea 832
949c47a0 833static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 834{
835 int offset;
836 int i;
837 void *_mpb = mpb;
838
949c47a0 839 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 840 return NULL;
841
842 /* devices start after all disks */
843 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
844
845 for (i = 0; i <= index; i++)
846 if (i == index)
847 return _mpb + offset;
848 else
3393c6af 849 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 850
851 return NULL;
852}
853
949c47a0
DW		 854static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
855{
ba2de7ba
DW		 856 struct intel_dev *dv;
857
949c47a0
DW		 858 if (index >= super->anchor->num_raid_devs)
859 return NULL;
ba2de7ba
DW		 860 for (dv = super->devlist; dv; dv = dv->next)
861 if (dv->index == index)
862 return dv->dev;
863 return NULL;
949c47a0
DW		 864}
865
8d67477f
TM		 866static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
867 *addr)
868{
869 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
870 __le16_to_cpu(addr->w1));
871}
872
873static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
874{
875 struct bbm_log_block_addr addr;
876
877 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
878 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
879 return addr;
880}
881
8d67477f
TM		 882/* get size of the bbm log */
883static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
884{
885 if (!log || log->entry_count == 0)
886 return 0;
887
888 return sizeof(log->signature) +
889 sizeof(log->entry_count) +
890 log->entry_count * sizeof(struct bbm_log_entry);
891}
6f50473f
TM		 892
893/* check if bad block is not partially stored in bbm log */
894static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
895 long long sector, const int length, __u32 *pos)
896{
897 __u32 i;
898
899 for (i = *pos; i < log->entry_count; i++) {
900 struct bbm_log_entry *entry = &log->marked_block_entries[i];
901 unsigned long long bb_start;
902 unsigned long long bb_end;
903
904 bb_start = __le48_to_cpu(&entry->defective_block_start);
905 bb_end = bb_start + (entry->marked_count + 1);
906
907 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
908 (bb_end <= sector + length)) {
909 *pos = i;
910 return 1;
911 }
912 }
913 return 0;
914}
915
916/* record new bad block in bbm log */
917static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
918 long long sector, int length)
919{
920 int new_bb = 0;
921 __u32 pos = 0;
922 struct bbm_log_entry *entry = NULL;
923
924 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
925 struct bbm_log_entry *e = &log->marked_block_entries[pos];
926
927 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
928 (__le48_to_cpu(&e->defective_block_start) == sector)) {
929 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
930 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
931 pos = pos + 1;
932 continue;
933 }
934 entry = e;
935 break;
936 }
937
938 if (entry) {
939 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
940 BBM_LOG_MAX_LBA_ENTRY_VAL;
941 entry->defective_block_start = __cpu_to_le48(sector);
942 entry->marked_count = cnt - 1;
943 if (cnt == length)
944 return 1;
945 sector += cnt;
946 length -= cnt;
947 }
948
949 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
950 BBM_LOG_MAX_LBA_ENTRY_VAL;
951 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
952 return 0;
953
954 while (length > 0) {
955 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
956 BBM_LOG_MAX_LBA_ENTRY_VAL;
957 struct bbm_log_entry *entry =
958 &log->marked_block_entries[log->entry_count];
959
960 entry->defective_block_start = __cpu_to_le48(sector);
961 entry->marked_count = cnt - 1;
962 entry->disk_ordinal = idx;
963
964 sector += cnt;
965 length -= cnt;
966
967 log->entry_count++;
968 }
969
970 return new_bb;
971}
c07a5a4f 972
4c9e8c1e
TM		 973/* clear all bad blocks for given disk */
974static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
975{
976 __u32 i = 0;
977
978 while (i < log->entry_count) {
979 struct bbm_log_entry *entries = log->marked_block_entries;
980
981 if (entries[i].disk_ordinal == idx) {
982 if (i < log->entry_count - 1)
983 entries[i] = entries[log->entry_count - 1];
984 log->entry_count--;
985 } else {
986 i++;
987 }
988 }
989}
990
c07a5a4f
TM		 991/* clear given bad block */
992static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
993 long long sector, const int length) {
994 __u32 i = 0;
995
996 while (i < log->entry_count) {
997 struct bbm_log_entry *entries = log->marked_block_entries;
998
999 if ((entries[i].disk_ordinal == idx) &&
1000 (__le48_to_cpu(&entries[i].defective_block_start) ==
1001 sector) && (entries[i].marked_count + 1 == length)) {
1002 if (i < log->entry_count - 1)
1003 entries[i] = entries[log->entry_count - 1];
1004 log->entry_count--;
1005 break;
1006 }
1007 i++;
1008 }
1009
1010 return 1;
1011}
8d67477f
TM		 1012
1013/* allocate and load BBM log from metadata */
1014static int load_bbm_log(struct intel_super *super)
1015{
1016 struct imsm_super *mpb = super->anchor;
1017 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1018
1019 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1020 if (!super->bbm_log)
1021 return 1;
1022
1023 if (bbm_log_size) {
1024 struct bbm_log *log = (void *)mpb +
1025 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1026
1027 __u32 entry_count;
1028
1029 if (bbm_log_size < sizeof(log->signature) +
1030 sizeof(log->entry_count))
1031 return 2;
1032
1033 entry_count = __le32_to_cpu(log->entry_count);
1034 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1035 (entry_count > BBM_LOG_MAX_ENTRIES))
1036 return 3;
1037
1038 if (bbm_log_size !=
1039 sizeof(log->signature) + sizeof(log->entry_count) +
1040 entry_count * sizeof(struct bbm_log_entry))
1041 return 4;
1042
1043 memcpy(super->bbm_log, log, bbm_log_size);
1044 } else {
1045 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1046 super->bbm_log->entry_count = 0;
1047 }
1048
1049 return 0;
1050}
1051
b12796be
TM		 1052/* checks if bad block is within volume boundaries */
1053static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1054 const unsigned long long start_sector,
1055 const unsigned long long size)
1056{
1057 unsigned long long bb_start;
1058 unsigned long long bb_end;
1059
1060 bb_start = __le48_to_cpu(&entry->defective_block_start);
1061 bb_end = bb_start + (entry->marked_count + 1);
1062
1063 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1064 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1065 return 1;
1066
1067 return 0;
1068}
1069
1070/* get list of bad blocks on a drive for a volume */
1071static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1072 const unsigned long long start_sector,
1073 const unsigned long long size,
1074 struct md_bb *bbs)
1075{
1076 __u32 count = 0;
1077 __u32 i;
1078
1079 for (i = 0; i < log->entry_count; i++) {
1080 const struct bbm_log_entry *ent =
1081 &log->marked_block_entries[i];
1082 struct md_bb_entry *bb;
1083
1084 if ((ent->disk_ordinal == idx) &&
1085 is_bad_block_in_volume(ent, start_sector, size)) {
1086
1087 if (!bbs->entries) {
1088 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1089 sizeof(*bb));
1090 if (!bbs->entries)
1091 break;
1092 }
1093
1094 bb = &bbs->entries[count++];
1095 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1096 bb->length = ent->marked_count + 1;
1097 }
1098 }
1099 bbs->count = count;
1100}
1101
98130f40
AK		 1102/*
1103 * for second_map:
238c0a71
AK		 1104 * == MAP_0 get first map
1105 * == MAP_1 get second map
1106 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1107 */
1108static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1109 int slot,
1110 int second_map)
7eef0453
DW		 1111{
1112 struct imsm_map *map;
1113
5e7b0330 1114 map = get_imsm_map(dev, second_map);
7eef0453 1115
ff077194
DW		 1116 /* top byte identifies disk under rebuild */
1117 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1118}
1119
1120#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1121static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1122{
98130f40 1123 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1124
1125 return ord_to_idx(ord);
7eef0453
DW		 1126}
1127
be73972f
DW		 1128static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1129{
1130 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1131}
1132
f21e18ca 1133static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1134{
1135 int slot;
1136 __u32 ord;
1137
1138 for (slot = 0; slot < map->num_members; slot++) {
1139 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1140 if (ord_to_idx(ord) == idx)
1141 return slot;
1142 }
1143
1144 return -1;
1145}
1146
cdddbdbc
DW		 1147static int get_imsm_raid_level(struct imsm_map *map)
1148{
1149 if (map->raid_level == 1) {
1150 if (map->num_members == 2)
1151 return 1;
1152 else
1153 return 10;
1154 }
1155
1156 return map->raid_level;
1157}
1158
c2c087e6
DW		 1159static int cmp_extent(const void *av, const void *bv)
1160{
1161 const struct extent *a = av;
1162 const struct extent *b = bv;
1163 if (a->start < b->start)
1164 return -1;
1165 if (a->start > b->start)
1166 return 1;
1167 return 0;
1168}
1169
0dcecb2e 1170static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1171{
c2c087e6 1172 int memberships = 0;
620b1713 1173 int i;
c2c087e6 1174
949c47a0
DW		 1175 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1176 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1177 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1178
620b1713
DW		 1179 if (get_imsm_disk_slot(map, dl->index) >= 0)
1180 memberships++;
c2c087e6 1181 }
0dcecb2e
DW		 1182
1183 return memberships;
1184}
1185
b81221b7
CA		 1186static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1187
486720e0 1188static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1189{
1190 if (lo == 0 || hi == 0)
1191 return 1;
486720e0
JS		 1192 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1193 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1194 return 0;
1195}
1196
1197static unsigned long long join_u32(__u32 lo, __u32 hi)
1198{
1199 return (unsigned long long)__le32_to_cpu(lo) |
1200 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1201}
1202
1203static unsigned long long total_blocks(struct imsm_disk *disk)
1204{
1205 if (disk == NULL)
1206 return 0;
1207 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1208}
1209
1210static unsigned long long pba_of_lba0(struct imsm_map *map)
1211{
1212 if (map == NULL)
1213 return 0;
1214 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1215}
1216
1217static unsigned long long blocks_per_member(struct imsm_map *map)
1218{
1219 if (map == NULL)
1220 return 0;
1221 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1222}
1223
1224static unsigned long long num_data_stripes(struct imsm_map *map)
1225{
1226 if (map == NULL)
1227 return 0;
1228 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1229}
1230
fcc2c9da
MD		 1231static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1232{
1233 if (dev == NULL)
1234 return 0;
1235 return join_u32(dev->size_low, dev->size_high);
1236}
1237
9f421827
PB		 1238static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1239{
1240 if (migr_rec == NULL)
1241 return 0;
1242 return join_u32(migr_rec->ckpt_area_pba_lo,
1243 migr_rec->ckpt_area_pba_hi);
1244}
1245
1246static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1247{
1248 if (migr_rec == NULL)
1249 return 0;
1250 return join_u32(migr_rec->curr_migr_unit_lo,
1251 migr_rec->curr_migr_unit_hi);
1252}
1253
1254static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1255{
1256 if (migr_rec == NULL)
1257 return 0;
1258 return join_u32(migr_rec->dest_1st_member_lba_lo,
1259 migr_rec->dest_1st_member_lba_hi);
1260}
1261
1262static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1263{
1264 if (migr_rec == NULL)
1265 return 0;
1266 return join_u32(migr_rec->num_migr_units_lo,
1267 migr_rec->num_migr_units_hi);
1268}
1269
5551b113
CA		 1270static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1271{
1272 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1273}
1274
1275static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1276{
1277 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1278}
1279
1280static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1281{
1282 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1283}
1284
1285static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1286{
1287 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1288}
1289
fcc2c9da
MD		 1290static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1291{
1292 split_ull(n, &dev->size_low, &dev->size_high);
1293}
1294
9f421827
PB		 1295static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1296 unsigned long long n)
1297{
1298 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1299}
1300
1301static void set_current_migr_unit(struct migr_record *migr_rec,
1302 unsigned long long n)
1303{
1304 split_ull(n, &migr_rec->curr_migr_unit_lo,
1305 &migr_rec->curr_migr_unit_hi);
1306}
1307
1308static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1309 unsigned long long n)
1310{
1311 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1312 &migr_rec->dest_1st_member_lba_hi);
1313}
1314
1315static void set_num_migr_units(struct migr_record *migr_rec,
1316 unsigned long long n)
1317{
1318 split_ull(n, &migr_rec->num_migr_units_lo,
1319 &migr_rec->num_migr_units_hi);
1320}
1321
44490938
MD		 1322static unsigned long long per_dev_array_size(struct imsm_map *map)
1323{
1324 unsigned long long array_size = 0;
1325
1326 if (map == NULL)
1327 return array_size;
1328
1329 array_size = num_data_stripes(map) * map->blocks_per_strip;
1330 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1331 array_size *= 2;
1332
1333 return array_size;
1334}
1335
05501181
PB		 1336static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1337 int get_minimal_reservation)
0dcecb2e
DW		 1338{
1339 /* find a list of used extents on the given physical device */
1340 struct extent *rv, *e;
620b1713 1341 int i;
0dcecb2e 1342 int memberships = count_memberships(dl, super);
b276dd33
DW		 1343 __u32 reservation;
1344
1345 /* trim the reserved area for spares, so they can join any array
1346 * regardless of whether the OROM has assigned sectors from the
1347 * IMSM_RESERVED_SECTORS region
1348 */
05501181 1349 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1350 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1351 else
1352 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1353
503975b9 1354 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1355 e = rv;
1356
949c47a0
DW		 1357 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1358 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1359 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1360
620b1713 1361 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1362 e->start = pba_of_lba0(map);
44490938 1363 e->size = per_dev_array_size(map);
620b1713 1364 e++;
c2c087e6
DW		 1365 }
1366 }
1367 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1368
1011e834 1369 /* determine the start of the metadata
14e8215b
DW		 1370 * when no raid devices are defined use the default
1371 * ...otherwise allow the metadata to truncate the value
1372 * as is the case with older versions of imsm
1373 */
1374 if (memberships) {
1375 struct extent *last = &rv[memberships - 1];
5551b113 1376 unsigned long long remainder;
14e8215b 1377
5551b113 1378 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1379 /* round down to 1k block to satisfy precision of the kernel
1380 * 'size' interface
1381 */
1382 remainder &= ~1UL;
1383 /* make sure remainder is still sane */
f21e18ca 1384 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1385 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1386 if (reservation > remainder)
1387 reservation = remainder;
1388 }
5551b113 1389 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1390 e->size = 0;
1391 return rv;
1392}
1393
14e8215b
DW		 1394/* try to determine how much space is reserved for metadata from
1395 * the last get_extents() entry, otherwise fallback to the
1396 * default
1397 */
1398static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1399{
1400 struct extent *e;
1401 int i;
1402 __u32 rv;
1403
1404 /* for spares just return a minimal reservation which will grow
1405 * once the spare is picked up by an array
1406 */
1407 if (dl->index == -1)
1408 return MPB_SECTOR_CNT;
1409
05501181 1410 e = get_extents(super, dl, 0);
14e8215b
DW		 1411 if (!e)
1412 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1413
1414 /* scroll to last entry */
1415 for (i = 0; e[i].size; i++)
1416 continue;
1417
5551b113 1418 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1419
1420 free(e);
1421
1422 return rv;
1423}
1424
25ed7e59
DW		 1425static int is_spare(struct imsm_disk *disk)
1426{
1427 return (disk->status & SPARE_DISK) == SPARE_DISK;
1428}
1429
1430static int is_configured(struct imsm_disk *disk)
1431{
1432 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1433}
1434
1435static int is_failed(struct imsm_disk *disk)
1436{
1437 return (disk->status & FAILED_DISK) == FAILED_DISK;
1438}
1439
2432ce9b
AP		 1440static int is_journal(struct imsm_disk *disk)
1441{
1442 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1443}
1444
b53bfba6
TM		 1445/* round array size down to closest MB and ensure it splits evenly
1446 * between members
1447 */
1448static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1449 disk_count)
1450{
1451 size /= disk_count;
1452 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1453 size *= disk_count;
1454
1455 return size;
1456}
1457
8b9cd157
MK		 1458static int able_to_resync(int raid_level, int missing_disks)
1459{
1460 int max_missing_disks = 0;
1461
1462 switch (raid_level) {
1463 case 10:
1464 max_missing_disks = 1;
1465 break;
1466 default:
1467 max_missing_disks = 0;
1468 }
1469 return missing_disks <= max_missing_disks;
1470}
1471
b81221b7
CA		 1472/* try to determine how much space is reserved for metadata from
1473 * the last get_extents() entry on the smallest active disk,
1474 * otherwise fallback to the default
1475 */
1476static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1477{
1478 struct extent *e;
1479 int i;
5551b113
CA		 1480 unsigned long long min_active;
1481 __u32 remainder;
b81221b7
CA		 1482 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1483 struct dl *dl, *dl_min = NULL;
1484
1485 if (!super)
1486 return rv;
1487
1488 min_active = 0;
1489 for (dl = super->disks; dl; dl = dl->next) {
1490 if (dl->index < 0)
1491 continue;
5551b113
CA		 1492 unsigned long long blocks = total_blocks(&dl->disk);
1493 if (blocks < min_active || min_active == 0) {
b81221b7 1494 dl_min = dl;
5551b113 1495 min_active = blocks;
b81221b7
CA		 1496 }
1497 }
1498 if (!dl_min)
1499 return rv;
1500
1501 /* find last lba used by subarrays on the smallest active disk */
05501181 1502 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1503 if (!e)
1504 return rv;
1505 for (i = 0; e[i].size; i++)
1506 continue;
1507
1508 remainder = min_active - e[i].start;
1509 free(e);
1510
1511 /* to give priority to recovery we should not require full
1512 IMSM_RESERVED_SECTORS from the spare */
1513 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1514
1515 /* if real reservation is smaller use that value */
1516 return (remainder < rv) ? remainder : rv;
1517}
1518
fbfdcb06
AO		 1519/*
1520 * Return minimum size of a spare and sector size
1521 * that can be used in this array
1522 */
1523int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1524{
1525 struct intel_super *super = st->sb;
1526 struct dl *dl;
1527 struct extent *e;
1528 int i;
fbfdcb06
AO		 1529 unsigned long long size = 0;
1530
1531 c->min_size = 0;
4b57ecf6 1532 c->sector_size = 0;
80e7f8c3
AC		 1533
1534 if (!super)
fbfdcb06 1535 return -EINVAL;
80e7f8c3
AC		 1536 /* find first active disk in array */
1537 dl = super->disks;
1538 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1539 dl = dl->next;
1540 if (!dl)
fbfdcb06 1541 return -EINVAL;
80e7f8c3 1542 /* find last lba used by subarrays */
05501181 1543 e = get_extents(super, dl, 0);
80e7f8c3 1544 if (!e)
fbfdcb06 1545 return -EINVAL;
80e7f8c3
AC		 1546 for (i = 0; e[i].size; i++)
1547 continue;
1548 if (i > 0)
fbfdcb06 1549 size = e[i-1].start + e[i-1].size;
80e7f8c3 1550 free(e);
b81221b7 1551
80e7f8c3 1552 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1553 size += imsm_min_reserved_sectors(super);
1554
1555 c->min_size = size * 512;
4b57ecf6 1556 c->sector_size = super->sector_size;
b81221b7 1557
fbfdcb06 1558 return 0;
80e7f8c3
AC		 1559}
1560
d1e02575
AK		 1561static int is_gen_migration(struct imsm_dev *dev);
1562
f36a9ecd
PB		 1563#define IMSM_4K_DIV 8
1564
c47b0ff6
AK		 1565static __u64 blocks_per_migr_unit(struct intel_super *super,
1566 struct imsm_dev *dev);
1e5c6983 1567
c47b0ff6
AK		 1568static void print_imsm_dev(struct intel_super *super,
1569 struct imsm_dev *dev,
1570 char *uuid,
1571 int disk_idx)
cdddbdbc
DW		 1572{
1573 __u64 sz;
0d80bb2f 1574 int slot, i;
238c0a71
AK		 1575 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1576 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1577 __u32 ord;
cdddbdbc
DW		 1578
1579 printf("\n");
1e7bc0ed 1580 printf("[%.16s]:\n", dev->volume);
44470971 1581 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1582 printf(" RAID Level : %d", get_imsm_raid_level(map));
1583 if (map2)
1584 printf(" <-- %d", get_imsm_raid_level(map2));
1585 printf("\n");
1586 printf(" Members : %d", map->num_members);
1587 if (map2)
1588 printf(" <-- %d", map2->num_members);
1589 printf("\n");
0d80bb2f
DW		 1590 printf(" Slots : [");
1591 for (i = 0; i < map->num_members; i++) {
238c0a71 1592 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1593 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1594 }
dd8bcb3b
AK		 1595 printf("]");
1596 if (map2) {
1597 printf(" <-- [");
1598 for (i = 0; i < map2->num_members; i++) {
238c0a71 1599 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1600 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1601 }
1602 printf("]");
1603 }
1604 printf("\n");
7095bccb
AK		 1605 printf(" Failed disk : ");
1606 if (map->failed_disk_num == 0xff)
1607 printf("none");
1608 else
1609 printf("%i", map->failed_disk_num);
1610 printf("\n");
620b1713
DW		 1611 slot = get_imsm_disk_slot(map, disk_idx);
1612 if (slot >= 0) {
238c0a71 1613 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1614 printf(" This Slot : %d%s\n", slot,
1615 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1616 } else
cdddbdbc 1617 printf(" This Slot : ?\n");
84918897 1618 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1619 sz = imsm_dev_size(dev);
84918897
MK		 1620 printf(" Array Size : %llu%s\n",
1621 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1622 human_size(sz * 512));
5551b113 1623 sz = blocks_per_member(map);
84918897
MK		 1624 printf(" Per Dev Size : %llu%s\n",
1625 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1626 human_size(sz * 512));
5551b113
CA		 1627 printf(" Sector Offset : %llu\n",
1628 pba_of_lba0(map));
1629 printf(" Num Stripes : %llu\n",
1630 num_data_stripes(map));
dd8bcb3b 1631 printf(" Chunk Size : %u KiB",
cdddbdbc 1632 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1633 if (map2)
1634 printf(" <-- %u KiB",
1635 __le16_to_cpu(map2->blocks_per_strip) / 2);
1636 printf("\n");
cdddbdbc 1637 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1638 printf(" Migrate State : ");
1484e727
DW		 1639 if (dev->vol.migr_state) {
1640 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1641 printf("initialize\n");
1484e727 1642 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1643 printf("rebuild\n");
1484e727 1644 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1645 printf("check\n");
1484e727 1646 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1647 printf("general migration\n");
1484e727 1648 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1649 printf("state change\n");
1484e727 1650 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1651 printf("repair\n");
1484e727 1652 else
8655a7b1
DW		 1653 printf("<unknown:%d>\n", migr_type(dev));
1654 } else
1655 printf("idle\n");
3393c6af
DW		 1656 printf(" Map State : %s", map_state_str[map->map_state]);
1657 if (dev->vol.migr_state) {
238c0a71 1658 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1659
b10b37b8 1660 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1661 printf("\n Checkpoint : %u ",
1662 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1663 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1664 printf("(N/A)");
1665 else
1666 printf("(%llu)", (unsigned long long)
1667 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1668 }
1669 printf("\n");
2432ce9b
AP		 1670 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1671 "dirty" : "clean");
1672 printf(" RWH Policy : ");
c2462068 1673 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1674 printf("off\n");
1675 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1676 printf("PPL distributed\n");
1677 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1678 printf("PPL journaling drive\n");
c2462068
PB		 1679 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1680 printf("Multiple distributed PPLs\n");
1681 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1682 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1683 else
1684 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1685}
1686
ef5c214e
MK		 1687static void print_imsm_disk(struct imsm_disk *disk,
1688 int index,
1689 __u32 reserved,
1690 unsigned int sector_size) {
1f24f035 1691 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1692 __u64 sz;
1693
0ec1f4e8 1694 if (index < -1 || !disk)
e9d82038
DW		 1695 return;
1696
cdddbdbc 1697 printf("\n");
1f24f035 1698 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1699 if (index >= 0)
1700 printf(" Disk%02d Serial : %s\n", index, str);
1701 else
1702 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1703 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1704 is_configured(disk) ? " active" : "",
1705 is_failed(disk) ? " failed" : "",
1706 is_journal(disk) ? " journal" : "");
cdddbdbc 1707 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1708 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1709 printf(" Usable Size : %llu%s\n",
1710 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1711 human_size(sz * 512));
1712}
1713
de44e46f
PB		 1714void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1715{
1716 struct migr_record *migr_rec = super->migr_rec;
1717
1718 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1719 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1720 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1721 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1722 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1723 set_migr_chkp_area_pba(migr_rec,
1724 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1725 set_migr_dest_1st_member_lba(migr_rec,
1726 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1727}
1728
f36a9ecd
PB		 1729void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1730{
1731 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1732}
1733
1734void convert_to_4k(struct intel_super *super)
1735{
1736 struct imsm_super *mpb = super->anchor;
1737 struct imsm_disk *disk;
1738 int i;
e4467bc7 1739 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1740
1741 for (i = 0; i < mpb->num_disks ; i++) {
1742 disk = __get_imsm_disk(mpb, i);
1743 /* disk */
1744 convert_to_4k_imsm_disk(disk);
1745 }
1746 for (i = 0; i < mpb->num_raid_devs; i++) {
1747 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1748 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1749 /* dev */
fcc2c9da 1750 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1751 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1752
1753 /* map0 */
1754 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1755 map->blocks_per_strip /= IMSM_4K_DIV;
1756 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1757
1758 if (dev->vol.migr_state) {
1759 /* map1 */
1760 map = get_imsm_map(dev, MAP_1);
1761 set_blocks_per_member(map,
1762 blocks_per_member(map)/IMSM_4K_DIV);
1763 map->blocks_per_strip /= IMSM_4K_DIV;
1764 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1765 }
1766 }
e4467bc7
TM		 1767 if (bbm_log_size) {
1768 struct bbm_log *log = (void *)mpb +
1769 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1770 __u32 i;
1771
1772 for (i = 0; i < log->entry_count; i++) {
1773 struct bbm_log_entry *entry =
1774 &log->marked_block_entries[i];
1775
1776 __u8 count = entry->marked_count + 1;
1777 unsigned long long sector =
1778 __le48_to_cpu(&entry->defective_block_start);
1779
1780 entry->defective_block_start =
1781 __cpu_to_le48(sector/IMSM_4K_DIV);
1782 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1783 }
1784 }
f36a9ecd
PB		 1785
1786 mpb->check_sum = __gen_imsm_checksum(mpb);
1787}
1788
520e69e2
AK		 1789void examine_migr_rec_imsm(struct intel_super *super)
1790{
1791 struct migr_record *migr_rec = super->migr_rec;
1792 struct imsm_super *mpb = super->anchor;
1793 int i;
1794
1795 for (i = 0; i < mpb->num_raid_devs; i++) {
1796 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1797 struct imsm_map *map;
b4ab44d8 1798 int slot = -1;
3136abe5 1799
520e69e2
AK		 1800 if (is_gen_migration(dev) == 0)
1801 continue;
1802
1803 printf("\nMigration Record Information:");
3136abe5 1804
44bfe6df
AK		 1805 /* first map under migration */
1806 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1807 if (map)
1808 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1809 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1810 printf(" Empty\n ");
1811 printf("Examine one of first two disks in array\n");
1812 break;
1813 }
1814 printf("\n Status : ");
1815 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1816 printf("Normal\n");
1817 else
1818 printf("Contains Data\n");
9f421827
PB		 1819 printf(" Current Unit : %llu\n",
1820 current_migr_unit(migr_rec));
520e69e2
AK		 1821 printf(" Family : %u\n",
1822 __le32_to_cpu(migr_rec->family_num));
1823 printf(" Ascending : %u\n",
1824 __le32_to_cpu(migr_rec->ascending_migr));
1825 printf(" Blocks Per Unit : %u\n",
1826 __le32_to_cpu(migr_rec->blocks_per_unit));
1827 printf(" Dest. Depth Per Unit : %u\n",
1828 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1829 printf(" Checkpoint Area pba : %llu\n",
1830 migr_chkp_area_pba(migr_rec));
1831 printf(" First member lba : %llu\n",
1832 migr_dest_1st_member_lba(migr_rec));
1833 printf(" Total Number of Units : %llu\n",
1834 get_num_migr_units(migr_rec));
1835 printf(" Size of volume : %llu\n",
1836 join_u32(migr_rec->post_migr_vol_cap,
1837 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1838 printf(" Record was read from : %u\n",
1839 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1840
1841 break;
1842 }
1843}
f36a9ecd 1844
de44e46f
PB		 1845void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1846{
1847 struct migr_record *migr_rec = super->migr_rec;
1848
1849 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1850 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1851 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1852 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1853 &migr_rec->post_migr_vol_cap,
1854 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1855 set_migr_chkp_area_pba(migr_rec,
1856 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1857 set_migr_dest_1st_member_lba(migr_rec,
1858 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1859}
1860
f36a9ecd
PB		 1861void convert_from_4k(struct intel_super *super)
1862{
1863 struct imsm_super *mpb = super->anchor;
1864 struct imsm_disk *disk;
1865 int i;
e4467bc7 1866 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1867
1868 for (i = 0; i < mpb->num_disks ; i++) {
1869 disk = __get_imsm_disk(mpb, i);
1870 /* disk */
1871 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1872 }
1873
1874 for (i = 0; i < mpb->num_raid_devs; i++) {
1875 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1876 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1877 /* dev */
fcc2c9da 1878 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1879 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1880
1881 /* map0 */
1882 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1883 map->blocks_per_strip *= IMSM_4K_DIV;
1884 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1885
1886 if (dev->vol.migr_state) {
1887 /* map1 */
1888 map = get_imsm_map(dev, MAP_1);
1889 set_blocks_per_member(map,
1890 blocks_per_member(map)*IMSM_4K_DIV);
1891 map->blocks_per_strip *= IMSM_4K_DIV;
1892 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1893 }
1894 }
e4467bc7
TM		 1895 if (bbm_log_size) {
1896 struct bbm_log *log = (void *)mpb +
1897 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1898 __u32 i;
1899
1900 for (i = 0; i < log->entry_count; i++) {
1901 struct bbm_log_entry *entry =
1902 &log->marked_block_entries[i];
1903
1904 __u8 count = entry->marked_count + 1;
1905 unsigned long long sector =
1906 __le48_to_cpu(&entry->defective_block_start);
1907
1908 entry->defective_block_start =
1909 __cpu_to_le48(sector*IMSM_4K_DIV);
1910 entry->marked_count = count*IMSM_4K_DIV - 1;
1911 }
1912 }
f36a9ecd
PB		 1913
1914 mpb->check_sum = __gen_imsm_checksum(mpb);
1915}
1916
19482bcc
AK		 1917/*******************************************************************************
1918 * function: imsm_check_attributes
1919 * Description: Function checks if features represented by attributes flags
1011e834 1920 * are supported by mdadm.
19482bcc
AK		 1921 * Parameters:
1922 * attributes - Attributes read from metadata
1923 * Returns:
1011e834
N		 1924 * 0 - passed attributes contains unsupported features flags
1925 * 1 - all features are supported
19482bcc
AK		 1926 ******************************************************************************/
1927static int imsm_check_attributes(__u32 attributes)
1928{
1929 int ret_val = 1;
418f9b36
N		 1930 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1931
1932 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1933
1934 not_supported &= attributes;
1935 if (not_supported) {
e7b84f9d 1936 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1937 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1938 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1939 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1940 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1941 }
1942 if (not_supported & MPB_ATTRIB_2TB) {
1943 dprintf("\t\tMPB_ATTRIB_2TB\n");
1944 not_supported ^= MPB_ATTRIB_2TB;
1945 }
1946 if (not_supported & MPB_ATTRIB_RAID0) {
1947 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1948 not_supported ^= MPB_ATTRIB_RAID0;
1949 }
1950 if (not_supported & MPB_ATTRIB_RAID1) {
1951 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1952 not_supported ^= MPB_ATTRIB_RAID1;
1953 }
1954 if (not_supported & MPB_ATTRIB_RAID10) {
1955 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1956 not_supported ^= MPB_ATTRIB_RAID10;
1957 }
1958 if (not_supported & MPB_ATTRIB_RAID1E) {
1959 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1960 not_supported ^= MPB_ATTRIB_RAID1E;
1961 }
1962 if (not_supported & MPB_ATTRIB_RAID5) {
1963 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1964 not_supported ^= MPB_ATTRIB_RAID5;
1965 }
1966 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1967 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1968 not_supported ^= MPB_ATTRIB_RAIDCNG;
1969 }
1970 if (not_supported & MPB_ATTRIB_BBM) {
1971 dprintf("\t\tMPB_ATTRIB_BBM\n");
1972 not_supported ^= MPB_ATTRIB_BBM;
1973 }
1974 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1975 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1976 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1977 }
1978 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1979 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1980 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1981 }
1982 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1983 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1984 not_supported ^= MPB_ATTRIB_2TB_DISK;
1985 }
1986 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1987 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1988 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1989 }
1990 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1991 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1992 not_supported ^= MPB_ATTRIB_NEVER_USE;
1993 }
1994
1995 if (not_supported)
1ade5cc1 1996 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1997
1998 ret_val = 0;
1999 }
2000
2001 return ret_val;
2002}
2003
a5d85af7 2004static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2005
cdddbdbc
DW		 2006static void examine_super_imsm(struct supertype *st, char *homehost)
2007{
2008 struct intel_super *super = st->sb;
949c47a0 2009 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2010 char str[MAX_SIGNATURE_LENGTH];
2011 int i;
27fd6274
DW		 2012 struct mdinfo info;
2013 char nbuf[64];
cdddbdbc 2014 __u32 sum;
14e8215b 2015 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2016 struct dl *dl;
27fd6274 2017
618f4e6d
XN		 2018 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2019 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2020 printf(" Magic : %s\n", str);
cdddbdbc 2021 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2022 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2023 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2024 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 2025 printf(" Attributes : ");
2026 if (imsm_check_attributes(mpb->attributes))
2027 printf("All supported\n");
2028 else
2029 printf("not supported\n");
a5d85af7 2030 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2031 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2032 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2033 sum = __le32_to_cpu(mpb->check_sum);
2034 printf(" Checksum : %08x %s\n", sum,
949c47a0 2035 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2036 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2037 printf(" Disks : %d\n", mpb->num_disks);
2038 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2039 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2040 super->disks->index, reserved, super->sector_size);
8d67477f 2041 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2042 struct bbm_log *log = super->bbm_log;
2043
2044 printf("\n");
2045 printf("Bad Block Management Log:\n");
2046 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2047 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2048 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2049 }
44470971
DW		 2050 for (i = 0; i < mpb->num_raid_devs; i++) {
2051 struct mdinfo info;
2052 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2053
2054 super->current_vol = i;
a5d85af7 2055 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2056 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2057 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2058 }
cdddbdbc
DW		 2059 for (i = 0; i < mpb->num_disks; i++) {
2060 if (i == super->disks->index)
2061 continue;
ef5c214e
MK		 2062 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2063 super->sector_size);
cdddbdbc 2064 }
94827db3 2065
0ec1f4e8
DW		 2066 for (dl = super->disks; dl; dl = dl->next)
2067 if (dl->index == -1)
ef5c214e
MK		 2068 print_imsm_disk(&dl->disk, -1, reserved,
2069 super->sector_size);
520e69e2
AK		 2070
2071 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2072}
2073
061f2c6a 2074static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2075{
27fd6274 2076 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2077 struct mdinfo info;
2078 char nbuf[64];
1e7bc0ed 2079 struct intel_super *super = st->sb;
1e7bc0ed 2080
0d5a423f
DW		 2081 if (!super->anchor->num_raid_devs) {
2082 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2083 return;
0d5a423f 2084 }
ff54de6e 2085
a5d85af7 2086 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2087 fname_from_uuid(st, &info, nbuf, ':');
2088 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2089}
2090
2091static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2092{
2093 /* We just write a generic IMSM ARRAY entry */
2094 struct mdinfo info;
2095 char nbuf[64];
2096 char nbuf1[64];
2097 struct intel_super *super = st->sb;
2098 int i;
2099
2100 if (!super->anchor->num_raid_devs)
2101 return;
2102
a5d85af7 2103 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2104 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2105 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2106 struct imsm_dev *dev = get_imsm_dev(super, i);
2107
2108 super->current_vol = i;
a5d85af7 2109 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2110 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2111 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2112 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2113 }
cdddbdbc
DW		 2114}
2115
9d84c8ea
DW		 2116static void export_examine_super_imsm(struct supertype *st)
2117{
2118 struct intel_super *super = st->sb;
2119 struct imsm_super *mpb = super->anchor;
2120 struct mdinfo info;
2121 char nbuf[64];
2122
a5d85af7 2123 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2124 fname_from_uuid(st, &info, nbuf, ':');
2125 printf("MD_METADATA=imsm\n");
2126 printf("MD_LEVEL=container\n");
2127 printf("MD_UUID=%s\n", nbuf+5);
2128 printf("MD_DEVICES=%u\n", mpb->num_disks);
2129}
2130
74db60b0
N		 2131static int copy_metadata_imsm(struct supertype *st, int from, int to)
2132{
f36a9ecd 2133 /* The second last sector of the device contains
74db60b0
N		 2134 * the "struct imsm_super" metadata.
2135 * This contains mpb_size which is the size in bytes of the
2136 * extended metadata. This is located immediately before
2137 * the imsm_super.
2138 * We want to read all that, plus the last sector which
2139 * may contain a migration record, and write it all
2140 * to the target.
2141 */
2142 void *buf;
2143 unsigned long long dsize, offset;
2144 int sectors;
2145 struct imsm_super *sb;
f36a9ecd
PB		 2146 struct intel_super *super = st->sb;
2147 unsigned int sector_size = super->sector_size;
2148 unsigned int written = 0;
74db60b0 2149
de44e46f 2150 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 2151 return 1;
2152
2153 if (!get_dev_size(from, NULL, &dsize))
2154 goto err;
2155
f36a9ecd 2156 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 2157 goto err;
466070ad 2158 if ((unsigned int)read(from, buf, sector_size) != sector_size)
74db60b0
N		 2159 goto err;
2160 sb = buf;
2161 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
2162 goto err;
2163
f36a9ecd
PB		 2164 sectors = mpb_sectors(sb, sector_size) + 2;
2165 offset = dsize - sectors * sector_size;
74db60b0
N		 2166 if (lseek64(from, offset, 0) < 0 ||
2167 lseek64(to, offset, 0) < 0)
2168 goto err;
f36a9ecd
PB		 2169 while (written < sectors * sector_size) {
2170 int n = sectors*sector_size - written;
74db60b0
N		 2171 if (n > 4096)
2172 n = 4096;
2173 if (read(from, buf, n) != n)
2174 goto err;
2175 if (write(to, buf, n) != n)
2176 goto err;
2177 written += n;
2178 }
2179 free(buf);
2180 return 0;
2181err:
2182 free(buf);
2183 return 1;
2184}
2185
b771faef
BK		 2186static void detail_super_imsm(struct supertype *st, char *homehost,
2187 char *subarray)
cdddbdbc 2188{
3ebe00a1
DW		 2189 struct mdinfo info;
2190 char nbuf[64];
b771faef
BK		 2191 struct intel_super *super = st->sb;
2192 int temp_vol = super->current_vol;
2193
2194 if (subarray)
2195 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2196
a5d85af7 2197 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2198 fname_from_uuid(st, &info, nbuf, ':');
65884368 2199 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2200
2201 super->current_vol = temp_vol;
cdddbdbc
DW		 2202}
2203
b771faef 2204static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2205{
ff54de6e
N		 2206 struct mdinfo info;
2207 char nbuf[64];
b771faef
BK		 2208 struct intel_super *super = st->sb;
2209 int temp_vol = super->current_vol;
2210
2211 if (subarray)
2212 super->current_vol = strtoul(subarray, NULL, 10);
2213
a5d85af7 2214 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2215 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2216 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2217
2218 super->current_vol = temp_vol;
cdddbdbc 2219}
d665cc31
DW		 2220
2221static int imsm_read_serial(int fd, char *devname, __u8 *serial);
2222static void fd2devname(int fd, char *name);
2223
120dc887 2224static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2225{
120dc887
LM		 2226 /* dump an unsorted list of devices attached to AHCI Intel storage
2227 * controller, as well as non-connected ports
d665cc31
DW		 2228 */
2229 int hba_len = strlen(hba_path) + 1;
2230 struct dirent *ent;
2231 DIR *dir;
2232 char *path = NULL;
2233 int err = 0;
2234 unsigned long port_mask = (1 << port_count) - 1;
2235
f21e18ca 2236 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2237 if (verbose > 0)
e7b84f9d 2238 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2239 return 2;
2240 }
2241
2242 /* scroll through /sys/dev/block looking for devices attached to
2243 * this hba
2244 */
2245 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2246 if (!dir)
2247 return 1;
2248
2249 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2250 int fd;
2251 char model[64];
2252 char vendor[64];
2253 char buf[1024];
2254 int major, minor;
2255 char *device;
2256 char *c;
2257 int port;
2258 int type;
2259
2260 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2261 continue;
2262 path = devt_to_devpath(makedev(major, minor));
2263 if (!path)
2264 continue;
2265 if (!path_attached_to_hba(path, hba_path)) {
2266 free(path);
2267 path = NULL;
2268 continue;
2269 }
2270
2271 /* retrieve the scsi device type */
2272 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2273 if (verbose > 0)
e7b84f9d 2274 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2275 err = 2;
2276 break;
2277 }
2278 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2279 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2280 if (verbose > 0)
e7b84f9d 2281 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2282 path);
2283 err = 2;
2284 free(device);
2285 break;
2286 }
2287 type = strtoul(buf, NULL, 10);
2288
2289 /* if it's not a disk print the vendor and model */
2290 if (!(type == 0 || type == 7 || type == 14)) {
2291 vendor[0] = '\0';
2292 model[0] = '\0';
2293 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2294 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2295 strncpy(vendor, buf, sizeof(vendor));
2296 vendor[sizeof(vendor) - 1] = '\0';
2297 c = (char *) &vendor[sizeof(vendor) - 1];
2298 while (isspace(*c) || *c == '\0')
2299 *c-- = '\0';
2300
2301 }
2302 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2303 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2304 strncpy(model, buf, sizeof(model));
2305 model[sizeof(model) - 1] = '\0';
2306 c = (char *) &model[sizeof(model) - 1];
2307 while (isspace(*c) || *c == '\0')
2308 *c-- = '\0';
2309 }
2310
2311 if (vendor[0] && model[0])
2312 sprintf(buf, "%.64s %.64s", vendor, model);
2313 else
2314 switch (type) { /* numbers from hald/linux/device.c */
2315 case 1: sprintf(buf, "tape"); break;
2316 case 2: sprintf(buf, "printer"); break;
2317 case 3: sprintf(buf, "processor"); break;
2318 case 4:
2319 case 5: sprintf(buf, "cdrom"); break;
2320 case 6: sprintf(buf, "scanner"); break;
2321 case 8: sprintf(buf, "media_changer"); break;
2322 case 9: sprintf(buf, "comm"); break;
2323 case 12: sprintf(buf, "raid"); break;
2324 default: sprintf(buf, "unknown");
2325 }
2326 } else
2327 buf[0] = '\0';
2328 free(device);
2329
2330 /* chop device path to 'host%d' and calculate the port number */
2331 c = strchr(&path[hba_len], '/');
4e5e717d 2332 if (!c) {
ba728be7 2333 if (verbose > 0)
e7b84f9d 2334 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2335 err = 2;
2336 break;
2337 }
d665cc31 2338 *c = '\0';
0858eccf
AP		 2339 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2340 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2341 port -= host_base;
2342 else {
ba728be7 2343 if (verbose > 0) {
d665cc31 2344 *c = '/'; /* repair the full string */
e7b84f9d 2345 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2346 path);
2347 }
2348 err = 2;
2349 break;
2350 }
2351
2352 /* mark this port as used */
2353 port_mask &= ~(1 << port);
2354
2355 /* print out the device information */
2356 if (buf[0]) {
2357 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2358 continue;
2359 }
2360
2361 fd = dev_open(ent->d_name, O_RDONLY);
2362 if (fd < 0)
2363 printf(" Port%d : - disk info unavailable -\n", port);
2364 else {
2365 fd2devname(fd, buf);
2366 printf(" Port%d : %s", port, buf);
2367 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2368 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2369 else
664d5325 2370 printf(" ()\n");
4dab422a 2371 close(fd);
d665cc31 2372 }
d665cc31
DW		 2373 free(path);
2374 path = NULL;
2375 }
2376 if (path)
2377 free(path);
2378 if (dir)
2379 closedir(dir);
2380 if (err == 0) {
2381 int i;
2382
2383 for (i = 0; i < port_count; i++)
2384 if (port_mask & (1 << i))
2385 printf(" Port%d : - no device attached -\n", i);
2386 }
2387
2388 return err;
2389}
2390
b5eece69 2391static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2392{
2393 struct dirent *ent;
2394 DIR *dir;
2395 char path[292];
2396 char link[256];
2397 char *c, *rp;
2398
2399 if (hba->type != SYS_DEV_VMD)
b5eece69 2400 return 1;
60f0f54d
PB		 2401
2402 /* scroll through /sys/dev/block looking for devices attached to
2403 * this hba
2404 */
2405 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2406 if (!dir)
b5eece69 2407 return 1;
b9135011
JS		 2408
2409 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2410 int n;
2411
2412 /* is 'ent' a device? check that the 'subsystem' link exists and
2413 * that its target matches 'bus'
2414 */
2415 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2416 ent->d_name);
2417 n = readlink(path, link, sizeof(link));
2418 if (n < 0 || n >= (int)sizeof(link))
2419 continue;
2420 link[n] = '\0';
2421 c = strrchr(link, '/');
2422 if (!c)
2423 continue;
2424 if (strncmp("pci", c+1, strlen("pci")) != 0)
2425 continue;
2426
2427 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2428
2429 rp = realpath(path, NULL);
2430 if (!rp)
2431 continue;
2432
2433 if (path_attached_to_hba(rp, hba->path)) {
2434 printf(" NVMe under VMD : %s\n", rp);
2435 }
2436 free(rp);
2437 }
2438
b9135011 2439 closedir(dir);
b5eece69 2440 return 0;
60f0f54d
PB		 2441}
2442
120dc887
LM		 2443static void print_found_intel_controllers(struct sys_dev *elem)
2444{
2445 for (; elem; elem = elem->next) {
e7b84f9d 2446 pr_err("found Intel(R) ");
120dc887
LM		 2447 if (elem->type == SYS_DEV_SATA)
2448 fprintf(stderr, "SATA ");
155cbb4c
LM		 2449 else if (elem->type == SYS_DEV_SAS)
2450 fprintf(stderr, "SAS ");
0858eccf
AP		 2451 else if (elem->type == SYS_DEV_NVME)
2452 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2453
2454 if (elem->type == SYS_DEV_VMD)
2455 fprintf(stderr, "VMD domain");
2456 else
2457 fprintf(stderr, "RAID controller");
2458
120dc887
LM		 2459 if (elem->pci_id)
2460 fprintf(stderr, " at %s", elem->pci_id);
2461 fprintf(stderr, ".\n");
2462 }
2463 fflush(stderr);
2464}
2465
120dc887
LM		 2466static int ahci_get_port_count(const char *hba_path, int *port_count)
2467{
2468 struct dirent *ent;
2469 DIR *dir;
2470 int host_base = -1;
2471
2472 *port_count = 0;
2473 if ((dir = opendir(hba_path)) == NULL)
2474 return -1;
2475
2476 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2477 int host;
2478
0858eccf
AP		 2479 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2480 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2481 continue;
2482 if (*port_count == 0)
2483 host_base = host;
2484 else if (host < host_base)
2485 host_base = host;
2486
2487 if (host + 1 > *port_count + host_base)
2488 *port_count = host + 1 - host_base;
2489 }
2490 closedir(dir);
2491 return host_base;
2492}
2493
a891a3c2
LM		 2494static void print_imsm_capability(const struct imsm_orom *orom)
2495{
0858eccf
AP		 2496 printf(" Platform : Intel(R) ");
2497 if (orom->capabilities == 0 && orom->driver_features == 0)
2498 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2499 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2500 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2501 else
2502 printf("Rapid Storage Technology%s\n",
2503 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2504 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2505 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2506 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2507 printf(" RAID Levels :%s%s%s%s%s\n",
2508 imsm_orom_has_raid0(orom) ? " raid0" : "",
2509 imsm_orom_has_raid1(orom) ? " raid1" : "",
2510 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2511 imsm_orom_has_raid10(orom) ? " raid10" : "",
2512 imsm_orom_has_raid5(orom) ? " raid5" : "");
2513 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2514 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2515 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2516 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2517 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2518 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2519 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2520 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2521 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2522 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2523 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2524 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2525 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2526 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2527 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2528 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2529 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2530 printf(" 2TB volumes :%s supported\n",
2531 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2532 printf(" 2TB disks :%s supported\n",
2533 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2534 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2535 printf(" Max Volumes : %d per array, %d per %s\n",
2536 orom->vpa, orom->vphba,
2537 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2538 return;
2539}
2540
e50cf220
MN		 2541static void print_imsm_capability_export(const struct imsm_orom *orom)
2542{
2543 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2544 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2545 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2546 orom->hotfix_ver, orom->build);
e50cf220
MN		 2547 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2548 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2549 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2550 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2551 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2552 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2553 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2554 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2555 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2556 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2557 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2558 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2559 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2560 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2561 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2562 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2563 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2564 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2565 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2566 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2567 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2568 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2569 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2570 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2571 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2572 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2573 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2574 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2575}
2576
9eafa1de 2577static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2578{
2579 /* There are two components to imsm platform support, the ahci SATA
2580 * controller and the option-rom. To find the SATA controller we
2581 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2582 * controller with the Intel vendor id is present. This approach
2583 * allows mdadm to leverage the kernel's ahci detection logic, with the
2584 * caveat that if ahci.ko is not loaded mdadm will not be able to
2585 * detect platform raid capabilities. The option-rom resides in a
2586 * platform "Adapter ROM". We scan for its signature to retrieve the
2587 * platform capabilities. If raid support is disabled in the BIOS the
2588 * option-rom capability structure will not be available.
2589 */
d665cc31 2590 struct sys_dev *list, *hba;
d665cc31
DW		 2591 int host_base = 0;
2592 int port_count = 0;
9eafa1de 2593 int result=1;
d665cc31 2594
5615172f 2595 if (enumerate_only) {
a891a3c2 2596 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2597 return 0;
a891a3c2
LM		 2598 list = find_intel_devices();
2599 if (!list)
2600 return 2;
2601 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2602 if (find_imsm_capability(hba)) {
2603 result = 0;
a891a3c2
LM		 2604 break;
2605 }
9eafa1de 2606 else
6b781d33 2607 result = 2;
a891a3c2 2608 }
a891a3c2 2609 return result;
5615172f
DW		 2610 }
2611
155cbb4c
LM		 2612 list = find_intel_devices();
2613 if (!list) {
ba728be7 2614 if (verbose > 0)
7a862a02 2615 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2616 return 2;
ba728be7 2617 } else if (verbose > 0)
155cbb4c 2618 print_found_intel_controllers(list);
d665cc31 2619
a891a3c2 2620 for (hba = list; hba; hba = hba->next) {
0858eccf 2621 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2622 continue;
0858eccf 2623 if (!find_imsm_capability(hba)) {
60f0f54d 2624 char buf[PATH_MAX];
e7b84f9d 2625 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2626 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2627 get_sys_dev_type(hba->type));
0858eccf
AP		 2628 continue;
2629 }
2630 result = 0;
2631 }
2632
2633 if (controller_path && result == 1) {
2634 pr_err("no active Intel(R) RAID controller found under %s\n",
2635 controller_path);
2636 return result;
2637 }
2638
5e1d6128 2639 const struct orom_entry *entry;
0858eccf 2640
5e1d6128 2641 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2642 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2643 print_imsm_capability(&entry->orom);
32716c51
PB		 2644 printf(" 3rd party NVMe :%s supported\n",
2645 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2646 for (hba = list; hba; hba = hba->next) {
2647 if (hba->type == SYS_DEV_VMD) {
2648 char buf[PATH_MAX];
60f0f54d
PB		 2649 printf(" I/O Controller : %s (%s)\n",
2650 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2651 if (print_vmd_attached_devs(hba)) {
2652 if (verbose > 0)
2653 pr_err("failed to get devices attached to VMD domain.\n");
2654 result |= 2;
2655 }
60f0f54d
PB		 2656 }
2657 }
07cb1e57 2658 printf("\n");
60f0f54d
PB		 2659 continue;
2660 }
0858eccf 2661
60f0f54d
PB		 2662 print_imsm_capability(&entry->orom);
2663 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2664 for (hba = list; hba; hba = hba->next) {
2665 if (hba->type == SYS_DEV_NVME)
2666 printf(" NVMe Device : %s\n", hba->path);
2667 }
60f0f54d 2668 printf("\n");
0858eccf
AP		 2669 continue;
2670 }
2671
2672 struct devid_list *devid;
5e1d6128 2673 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2674 hba = device_by_id(devid->devid);
2675 if (!hba)
2676 continue;
2677
9eafa1de
MN		 2678 printf(" I/O Controller : %s (%s)\n",
2679 hba->path, get_sys_dev_type(hba->type));
2680 if (hba->type == SYS_DEV_SATA) {
2681 host_base = ahci_get_port_count(hba->path, &port_count);
2682 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2683 if (verbose > 0)
7a862a02 2684 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2685 result |= 2;
2686 }
120dc887
LM		 2687 }
2688 }
0858eccf 2689 printf("\n");
d665cc31 2690 }
155cbb4c 2691
120dc887 2692 return result;
d665cc31 2693}
e50cf220 2694
9eafa1de 2695static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2696{
e50cf220
MN		 2697 struct sys_dev *list, *hba;
2698 int result=1;
2699
2700 list = find_intel_devices();
2701 if (!list) {
2702 if (verbose > 0)
2703 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2704 result = 2;
e50cf220
MN		 2705 return result;
2706 }
2707
2708 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2709 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2710 continue;
60f0f54d
PB		 2711 if (!find_imsm_capability(hba) && verbose > 0) {
2712 char buf[PATH_MAX];
2713 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2714 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2715 }
0858eccf 2716 else
e50cf220 2717 result = 0;
e50cf220
MN		 2718 }
2719
5e1d6128 2720 const struct orom_entry *entry;
0858eccf 2721
60f0f54d
PB		 2722 for (entry = orom_entries; entry; entry = entry->next) {
2723 if (entry->type == SYS_DEV_VMD) {
2724 for (hba = list; hba; hba = hba->next)
2725 print_imsm_capability_export(&entry->orom);
2726 continue;
2727 }
5e1d6128 2728 print_imsm_capability_export(&entry->orom);
60f0f54d 2729 }
0858eccf 2730
e50cf220
MN		 2731 return result;
2732}
2733
cdddbdbc
DW		 2734static int match_home_imsm(struct supertype *st, char *homehost)
2735{
5115ca67
DW		 2736 /* the imsm metadata format does not specify any host
2737 * identification information. We return -1 since we can never
2738 * confirm nor deny whether a given array is "meant" for this
148acb7b 2739 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2740 * exclude member disks that do not belong, and we rely on
2741 * mdadm.conf to specify the arrays that should be assembled.
2742 * Auto-assembly may still pick up "foreign" arrays.
2743 */
cdddbdbc 2744
9362c1c8 2745 return -1;
cdddbdbc
DW		 2746}
2747
2748static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2749{
51006d85
N		 2750 /* The uuid returned here is used for:
2751 * uuid to put into bitmap file (Create, Grow)
2752 * uuid for backup header when saving critical section (Grow)
2753 * comparing uuids when re-adding a device into an array
2754 * In these cases the uuid required is that of the data-array,
2755 * not the device-set.
2756 * uuid to recognise same set when adding a missing device back
2757 * to an array. This is a uuid for the device-set.
1011e834 2758 *
51006d85
N		 2759 * For each of these we can make do with a truncated
2760 * or hashed uuid rather than the original, as long as
2761 * everyone agrees.
2762 * In each case the uuid required is that of the data-array,
2763 * not the device-set.
43dad3d6 2764 */
51006d85
N		 2765 /* imsm does not track uuid's so we synthesis one using sha1 on
2766 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2767 * - the orig_family_num of the container
51006d85
N		 2768 * - the index number of the volume
2769 * - the 'serial' number of the volume.
2770 * Hopefully these are all constant.
2771 */
2772 struct intel_super *super = st->sb;
43dad3d6 2773
51006d85
N		 2774 char buf[20];
2775 struct sha1_ctx ctx;
2776 struct imsm_dev *dev = NULL;
148acb7b 2777 __u32 family_num;
51006d85 2778
148acb7b
DW		 2779 /* some mdadm versions failed to set ->orig_family_num, in which
2780 * case fall back to ->family_num. orig_family_num will be
2781 * fixed up with the first metadata update.
2782 */
2783 family_num = super->anchor->orig_family_num;
2784 if (family_num == 0)
2785 family_num = super->anchor->family_num;
51006d85 2786 sha1_init_ctx(&ctx);
92bd8f8d 2787 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2788 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2789 if (super->current_vol >= 0)
2790 dev = get_imsm_dev(super, super->current_vol);
2791 if (dev) {
2792 __u32 vol = super->current_vol;
2793 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2794 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2795 }
2796 sha1_finish_ctx(&ctx, buf);
2797 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2798}
2799
0d481d37 2800#if 0
4f5bc454
DW		 2801static void
2802get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2803{
cdddbdbc
DW		 2804 __u8 *v = get_imsm_version(mpb);
2805 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2806 char major[] = { 0, 0, 0 };
2807 char minor[] = { 0 ,0, 0 };
2808 char patch[] = { 0, 0, 0 };
2809 char *ver_parse[] = { major, minor, patch };
2810 int i, j;
2811
2812 i = j = 0;
2813 while (*v != '\0' && v < end) {
2814 if (*v != '.' && j < 2)
2815 ver_parse[i][j++] = *v;
2816 else {
2817 i++;
2818 j = 0;
2819 }
2820 v++;
2821 }
2822
4f5bc454
DW		 2823 *m = strtol(minor, NULL, 0);
2824 *p = strtol(patch, NULL, 0);
2825}
0d481d37 2826#endif
4f5bc454 2827
1e5c6983
DW		 2828static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2829{
2830 /* migr_strip_size when repairing or initializing parity */
238c0a71 2831 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2832 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2833
2834 switch (get_imsm_raid_level(map)) {
2835 case 5:
2836 case 10:
2837 return chunk;
2838 default:
2839 return 128*1024 >> 9;
2840 }
2841}
2842
2843static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2844{
2845 /* migr_strip_size when rebuilding a degraded disk, no idea why
2846 * this is different than migr_strip_size_resync(), but it's good
2847 * to be compatible
2848 */
238c0a71 2849 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2850 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2851
2852 switch (get_imsm_raid_level(map)) {
2853 case 1:
2854 case 10:
2855 if (map->num_members % map->num_domains == 0)
2856 return 128*1024 >> 9;
2857 else
2858 return chunk;
2859 case 5:
2860 return max((__u32) 64*1024 >> 9, chunk);
2861 default:
2862 return 128*1024 >> 9;
2863 }
2864}
2865
2866static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2867{
238c0a71
AK		 2868 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2869 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2870 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2871 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2872
2873 return max((__u32) 1, hi_chunk / lo_chunk);
2874}
2875
2876static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2877{
238c0a71 2878 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2879 int level = get_imsm_raid_level(lo);
2880
2881 if (level == 1 || level == 10) {
238c0a71 2882 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2883
2884 return hi->num_domains;
2885 } else
2886 return num_stripes_per_unit_resync(dev);
2887}
2888
9529d343 2889static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2890{
2891 /* named 'imsm_' because raid0, raid1 and raid10
2892 * counter-intuitively have the same number of data disks
2893 */
1e5c6983
DW		 2894 switch (get_imsm_raid_level(map)) {
2895 case 0:
36fd8ccc
AK		 2896 return map->num_members;
2897 break;
1e5c6983
DW		 2898 case 1:
2899 case 10:
36fd8ccc 2900 return map->num_members/2;
1e5c6983
DW		 2901 case 5:
2902 return map->num_members - 1;
2903 default:
1ade5cc1 2904 dprintf("unsupported raid level\n");
1e5c6983
DW		 2905 return 0;
2906 }
2907}
2908
44490938
MD		 2909static unsigned long long calc_component_size(struct imsm_map *map,
2910 struct imsm_dev *dev)
2911{
2912 unsigned long long component_size;
2913 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2914 long long calc_dev_size = 0;
44490938
MD		 2915 unsigned int member_disks = imsm_num_data_members(map);
2916
2917 if (member_disks == 0)
2918 return 0;
2919
2920 component_size = per_dev_array_size(map);
2921 calc_dev_size = component_size * member_disks;
2922
2923 /* Component size is rounded to 1MB so difference between size from
2924 * metadata and size calculated from num_data_stripes equals up to
2925 * 2048 blocks per each device. If the difference is higher it means
2926 * that array size was expanded and num_data_stripes was not updated.
2927 */
a4f7290c 2928 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2929 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2930 component_size = dev_size / member_disks;
2931 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2932 component_size / map->blocks_per_strip,
2933 num_data_stripes(map));
2934 }
2935
2936 return component_size;
2937}
2938
1e5c6983
DW		 2939static __u32 parity_segment_depth(struct imsm_dev *dev)
2940{
238c0a71 2941 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2942 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2943
2944 switch(get_imsm_raid_level(map)) {
2945 case 1:
2946 case 10:
2947 return chunk * map->num_domains;
2948 case 5:
2949 return chunk * map->num_members;
2950 default:
2951 return chunk;
2952 }
2953}
2954
2955static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2956{
238c0a71 2957 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2958 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2959 __u32 strip = block / chunk;
2960
2961 switch (get_imsm_raid_level(map)) {
2962 case 1:
2963 case 10: {
2964 __u32 vol_strip = (strip * map->num_domains) + 1;
2965 __u32 vol_stripe = vol_strip / map->num_members;
2966
2967 return vol_stripe * chunk + block % chunk;
2968 } case 5: {
2969 __u32 stripe = strip / (map->num_members - 1);
2970
2971 return stripe * chunk + block % chunk;
2972 }
2973 default:
2974 return 0;
2975 }
2976}
2977
c47b0ff6
AK		 2978static __u64 blocks_per_migr_unit(struct intel_super *super,
2979 struct imsm_dev *dev)
1e5c6983
DW		 2980{
2981 /* calculate the conversion factor between per member 'blocks'
2982 * (md/{resync,rebuild}_start) and imsm migration units, return
2983 * 0 for the 'not migrating' and 'unsupported migration' cases
2984 */
2985 if (!dev->vol.migr_state)
2986 return 0;
2987
2988 switch (migr_type(dev)) {
c47b0ff6
AK		 2989 case MIGR_GEN_MIGR: {
2990 struct migr_record *migr_rec = super->migr_rec;
2991 return __le32_to_cpu(migr_rec->blocks_per_unit);
2992 }
1e5c6983
DW		 2993 case MIGR_VERIFY:
2994 case MIGR_REPAIR:
2995 case MIGR_INIT: {
238c0a71 2996 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2997 __u32 stripes_per_unit;
2998 __u32 blocks_per_unit;
2999 __u32 parity_depth;
3000 __u32 migr_chunk;
3001 __u32 block_map;
3002 __u32 block_rel;
3003 __u32 segment;
3004 __u32 stripe;
3005 __u8 disks;
3006
3007 /* yes, this is really the translation of migr_units to
3008 * per-member blocks in the 'resync' case
3009 */
3010 stripes_per_unit = num_stripes_per_unit_resync(dev);
3011 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 3012 disks = imsm_num_data_members(map);
1e5c6983 3013 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 3014 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 3015 segment = blocks_per_unit / stripe;
3016 block_rel = blocks_per_unit - segment * stripe;
3017 parity_depth = parity_segment_depth(dev);
3018 block_map = map_migr_block(dev, block_rel);
3019 return block_map + parity_depth * segment;
3020 }
3021 case MIGR_REBUILD: {
3022 __u32 stripes_per_unit;
3023 __u32 migr_chunk;
3024
3025 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
3026 migr_chunk = migr_strip_blocks_rebuild(dev);
3027 return migr_chunk * stripes_per_unit;
3028 }
1e5c6983
DW		 3029 case MIGR_STATE_CHANGE:
3030 default:
3031 return 0;
3032 }
3033}
3034
c2c087e6
DW		 3035static int imsm_level_to_layout(int level)
3036{
3037 switch (level) {
3038 case 0:
3039 case 1:
3040 return 0;
3041 case 5:
3042 case 6:
a380c027 3043 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 3044 case 10:
c92a2527 3045 return 0x102;
c2c087e6 3046 }
a18a888e 3047 return UnSet;
c2c087e6
DW		 3048}
3049
8e59f3d8
AK		 3050/*******************************************************************************
3051 * Function: read_imsm_migr_rec
3052 * Description: Function reads imsm migration record from last sector of disk
3053 * Parameters:
3054 * fd : disk descriptor
3055 * super : metadata info
3056 * Returns:
3057 * 0 : success,
3058 * -1 : fail
3059 ******************************************************************************/
3060static int read_imsm_migr_rec(int fd, struct intel_super *super)
3061{
3062 int ret_val = -1;
de44e46f 3063 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3064 unsigned long long dsize;
3065
3066 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3067 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3068 SEEK_SET) < 0) {
e7b84f9d
N		 3069 pr_err("Cannot seek to anchor block: %s\n",
3070 strerror(errno));
8e59f3d8
AK		 3071 goto out;
3072 }
466070ad 3073 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3074 MIGR_REC_BUF_SECTORS*sector_size) !=
3075 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3076 pr_err("Cannot read migr record block: %s\n",
3077 strerror(errno));
8e59f3d8
AK		 3078 goto out;
3079 }
3080 ret_val = 0;
de44e46f
PB		 3081 if (sector_size == 4096)
3082 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3083
3084out:
3085 return ret_val;
3086}
3087
3136abe5
AK		 3088static struct imsm_dev *imsm_get_device_during_migration(
3089 struct intel_super *super)
3090{
3091
3092 struct intel_dev *dv;
3093
3094 for (dv = super->devlist; dv; dv = dv->next) {
3095 if (is_gen_migration(dv->dev))
3096 return dv->dev;
3097 }
3098 return NULL;
3099}
3100
8e59f3d8
AK		 3101/*******************************************************************************
3102 * Function: load_imsm_migr_rec
3103 * Description: Function reads imsm migration record (it is stored at the last
3104 * sector of disk)
3105 * Parameters:
3106 * super : imsm internal array info
3107 * info : general array info
3108 * Returns:
3109 * 0 : success
3110 * -1 : fail
4c965cc9 3111 * -2 : no migration in progress
8e59f3d8
AK		 3112 ******************************************************************************/
3113static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3114{
3115 struct mdinfo *sd;
594dc1b8 3116 struct dl *dl;
8e59f3d8
AK		 3117 char nm[30];
3118 int retval = -1;
3119 int fd = -1;
3136abe5 3120 struct imsm_dev *dev;
594dc1b8 3121 struct imsm_map *map;
b4ab44d8 3122 int slot = -1;
3136abe5
AK		 3123
3124 /* find map under migration */
3125 dev = imsm_get_device_during_migration(super);
3126 /* nothing to load,no migration in progress?
3127 */
3128 if (dev == NULL)
4c965cc9 3129 return -2;
8e59f3d8
AK		 3130
3131 if (info) {
3132 for (sd = info->devs ; sd ; sd = sd->next) {
3133 /* read only from one of the first two slots */
12fe93e9
TM		 3134 if ((sd->disk.raid_disk < 0) ||
3135 (sd->disk.raid_disk > 1))
8e59f3d8 3136 continue;
3136abe5 3137
8e59f3d8
AK		 3138 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3139 fd = dev_open(nm, O_RDONLY);
3140 if (fd >= 0)
3141 break;
3142 }
3143 }
3144 if (fd < 0) {
12fe93e9 3145 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3146 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3147 /* skip spare and failed disks
3148 */
3149 if (dl->index < 0)
3150 continue;
8e59f3d8 3151 /* read only from one of the first two slots */
3136abe5
AK		 3152 if (map)
3153 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3154 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3155 continue;
3156 sprintf(nm, "%d:%d", dl->major, dl->minor);
3157 fd = dev_open(nm, O_RDONLY);
3158 if (fd >= 0)
3159 break;
3160 }
3161 }
3162 if (fd < 0)
3163 goto out;
3164 retval = read_imsm_migr_rec(fd, super);
3165
3166out:
3167 if (fd >= 0)
3168 close(fd);
3169 return retval;
3170}
3171
c17608ea
AK		 3172/*******************************************************************************
3173 * function: imsm_create_metadata_checkpoint_update
3174 * Description: It creates update for checkpoint change.
3175 * Parameters:
3176 * super : imsm internal array info
3177 * u : pointer to prepared update
3178 * Returns:
3179 * Uptate length.
3180 * If length is equal to 0, input pointer u contains no update
3181 ******************************************************************************/
3182static int imsm_create_metadata_checkpoint_update(
3183 struct intel_super *super,
3184 struct imsm_update_general_migration_checkpoint **u)
3185{
3186
3187 int update_memory_size = 0;
3188
1ade5cc1 3189 dprintf("(enter)\n");
c17608ea
AK		 3190
3191 if (u == NULL)
3192 return 0;
3193 *u = NULL;
3194
3195 /* size of all update data without anchor */
3196 update_memory_size =
3197 sizeof(struct imsm_update_general_migration_checkpoint);
3198
503975b9 3199 *u = xcalloc(1, update_memory_size);
c17608ea 3200 if (*u == NULL) {
1ade5cc1 3201 dprintf("error: cannot get memory\n");
c17608ea
AK		 3202 return 0;
3203 }
3204 (*u)->type = update_general_migration_checkpoint;
9f421827 3205 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3206 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3207
3208 return update_memory_size;
3209}
3210
c17608ea
AK		 3211static void imsm_update_metadata_locally(struct supertype *st,
3212 void *buf, int len);
3213
687629c2
AK		 3214/*******************************************************************************
3215 * Function: write_imsm_migr_rec
3216 * Description: Function writes imsm migration record
3217 * (at the last sector of disk)
3218 * Parameters:
3219 * super : imsm internal array info
3220 * Returns:
3221 * 0 : success
3222 * -1 : if fail
3223 ******************************************************************************/
3224static int write_imsm_migr_rec(struct supertype *st)
3225{
3226 struct intel_super *super = st->sb;
de44e46f 3227 unsigned int sector_size = super->sector_size;
687629c2
AK		 3228 unsigned long long dsize;
3229 char nm[30];
3230 int fd = -1;
3231 int retval = -1;
3232 struct dl *sd;
c17608ea
AK		 3233 int len;
3234 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3235 struct imsm_dev *dev;
594dc1b8 3236 struct imsm_map *map;
3136abe5
AK		 3237
3238 /* find map under migration */
3239 dev = imsm_get_device_during_migration(super);
3240 /* if no migration, write buffer anyway to clear migr_record
3241 * on disk based on first available device
3242 */
3243 if (dev == NULL)
3244 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3245 super->current_vol);
3246
44bfe6df 3247 map = get_imsm_map(dev, MAP_0);
687629c2 3248
de44e46f
PB		 3249 if (sector_size == 4096)
3250 convert_to_4k_imsm_migr_rec(super);
687629c2 3251 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3252 int slot = -1;
3136abe5
AK		 3253
3254 /* skip failed and spare devices */
3255 if (sd->index < 0)
3256 continue;
687629c2 3257 /* write to 2 first slots only */
3136abe5
AK		 3258 if (map)
3259 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3260 if (map == NULL || slot > 1 || slot < 0)
687629c2 3261 continue;
3136abe5 3262
687629c2
AK		 3263 sprintf(nm, "%d:%d", sd->major, sd->minor);
3264 fd = dev_open(nm, O_RDWR);
3265 if (fd < 0)
3266 continue;
3267 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3268 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3269 SEEK_SET) < 0) {
e7b84f9d
N		 3270 pr_err("Cannot seek to anchor block: %s\n",
3271 strerror(errno));
687629c2
AK		 3272 goto out;
3273 }
466070ad 3274 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3275 MIGR_REC_BUF_SECTORS*sector_size) !=
3276 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3277 pr_err("Cannot write migr record block: %s\n",
3278 strerror(errno));
687629c2
AK		 3279 goto out;
3280 }
3281 close(fd);
3282 fd = -1;
3283 }
de44e46f
PB		 3284 if (sector_size == 4096)
3285 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3286 /* update checkpoint information in metadata */
3287 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3288 if (len <= 0) {
3289 dprintf("imsm: Cannot prepare update\n");
3290 goto out;
3291 }
3292 /* update metadata locally */
3293 imsm_update_metadata_locally(st, u, len);
3294 /* and possibly remotely */
3295 if (st->update_tail) {
3296 append_metadata_update(st, u, len);
3297 /* during reshape we do all work inside metadata handler
3298 * manage_reshape(), so metadata update has to be triggered
3299 * insida it
3300 */
3301 flush_metadata_updates(st);
3302 st->update_tail = &st->updates;
3303 } else
3304 free(u);
687629c2
AK		 3305
3306 retval = 0;
3307 out:
3308 if (fd >= 0)
3309 close(fd);
3310 return retval;
3311}
3312
e2962bfc
AK		 3313/* spare/missing disks activations are not allowe when
3314 * array/container performs reshape operation, because
3315 * all arrays in container works on the same disks set
3316 */
3317int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3318{
3319 int rv = 0;
3320 struct intel_dev *i_dev;
3321 struct imsm_dev *dev;
3322
3323 /* check whole container
3324 */
3325 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3326 dev = i_dev->dev;
3ad25638 3327 if (is_gen_migration(dev)) {
e2962bfc
AK		 3328 /* No repair during any migration in container
3329 */
3330 rv = 1;
3331 break;
3332 }
3333 }
3334 return rv;
3335}
3e684231 3336static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3337 int chunk_size,
f36a9ecd 3338 unsigned int sector_size,
c41e00b2
AK		 3339 unsigned long long component_size)
3340{
3e684231 3341 unsigned int component_size_alignment;
c41e00b2 3342
3e684231 3343 /* check component size alignment
c41e00b2 3344 */
3e684231 3345 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3346
3e684231 3347 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3348 level, chunk_size, component_size,
3e684231 3349 component_size_alignment);
c41e00b2 3350
3e684231
MZ		 3351 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3352 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3353 component_size);
3e684231 3354 component_size -= component_size_alignment;
1ade5cc1 3355 dprintf_cont("to %llu (%i).\n",
3e684231 3356 component_size, component_size_alignment);
c41e00b2
AK		 3357 }
3358
3359 return component_size;
3360}
e2962bfc 3361
2432ce9b
AP		 3362static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3363{
3364 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3365 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3366
3367 return pba_of_lba0(map) +
3368 (num_data_stripes(map) * map->blocks_per_strip);
3369}
3370
a5d85af7 3371static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3372{
3373 struct intel_super *super = st->sb;
c47b0ff6 3374 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3375 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3376 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3377 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3378 struct imsm_map *map_to_analyse = map;
efb30e7f 3379 struct dl *dl;
a5d85af7 3380 int map_disks = info->array.raid_disks;
bf5a934a 3381
95eeceeb 3382 memset(info, 0, sizeof(*info));
b335e593
AK		 3383 if (prev_map)
3384 map_to_analyse = prev_map;
3385
ca0748fa 3386 dl = super->current_disk;
9894ec0d 3387
bf5a934a 3388 info->container_member = super->current_vol;
cd0430a1 3389 info->array.raid_disks = map->num_members;
b335e593 3390 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3391 info->array.layout = imsm_level_to_layout(info->array.level);
3392 info->array.md_minor = -1;
3393 info->array.ctime = 0;
3394 info->array.utime = 0;
b335e593
AK		 3395 info->array.chunk_size =
3396 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3397 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3398 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3399 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3400
3f510843 3401 if (is_gen_migration(dev)) {
3f83228a 3402 info->reshape_active = 1;
b335e593
AK		 3403 info->new_level = get_imsm_raid_level(map);
3404 info->new_layout = imsm_level_to_layout(info->new_level);
3405 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3406 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3407 if (info->delta_disks) {
3408 /* this needs to be applied to every array
3409 * in the container.
3410 */
81219e70 3411 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3412 }
3f83228a
N		 3413 /* We shape information that we give to md might have to be
3414 * modify to cope with md's requirement for reshaping arrays.
3415 * For example, when reshaping a RAID0, md requires it to be
3416 * presented as a degraded RAID4.
3417 * Also if a RAID0 is migrating to a RAID5 we need to specify
3418 * the array as already being RAID5, but the 'before' layout
3419 * is a RAID4-like layout.
3420 */
3421 switch (info->array.level) {
3422 case 0:
3423 switch(info->new_level) {
3424 case 0:
3425 /* conversion is happening as RAID4 */
3426 info->array.level = 4;
3427 info->array.raid_disks += 1;
3428 break;
3429 case 5:
3430 /* conversion is happening as RAID5 */
3431 info->array.level = 5;
3432 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3433 info->delta_disks -= 1;
3434 break;
3435 default:
3436 /* FIXME error message */
3437 info->array.level = UnSet;
3438 break;
3439 }
3440 break;
3441 }
b335e593
AK		 3442 } else {
3443 info->new_level = UnSet;
3444 info->new_layout = UnSet;
3445 info->new_chunk = info->array.chunk_size;
3f83228a 3446 info->delta_disks = 0;
b335e593 3447 }
ca0748fa 3448
efb30e7f
DW		 3449 if (dl) {
3450 info->disk.major = dl->major;
3451 info->disk.minor = dl->minor;
ca0748fa 3452 info->disk.number = dl->index;
656b6b5a
N		 3453 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3454 dl->index);
efb30e7f 3455 }
bf5a934a 3456
5551b113 3457 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3458 info->component_size = calc_component_size(map, dev);
3e684231 3459 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3460 info->array.level,
3461 info->array.chunk_size,
f36a9ecd 3462 super->sector_size,
c41e00b2 3463 info->component_size);
5e46202e 3464 info->bb.supported = 1;
139dae11 3465
301406c9 3466 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3467 info->recovery_start = MaxSector;
bf5a934a 3468
c2462068
PB		 3469 if (info->array.level == 5 &&
3470 (dev->rwh_policy == RWH_DISTRIBUTED ||
3471 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3472 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3473 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3474 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3475 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3476 else
3477 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3478 >> 9;
2432ce9b
AP		 3479 } else if (info->array.level <= 0) {
3480 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3481 } else {
3482 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3483 }
3484
d2e6d5d6 3485 info->reshape_progress = 0;
b6796ce1 3486 info->resync_start = MaxSector;
b9172665 3487 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3488 !(info->array.state & 1)) &&
b9172665 3489 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3490 info->resync_start = 0;
b6796ce1
AK		 3491 }
3492 if (dev->vol.migr_state) {
1e5c6983
DW		 3493 switch (migr_type(dev)) {
3494 case MIGR_REPAIR:
3495 case MIGR_INIT: {
c47b0ff6
AK		 3496 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3497 dev);
1e5c6983
DW		 3498 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3499
3500 info->resync_start = blocks_per_unit * units;
3501 break;
3502 }
d2e6d5d6 3503 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3504 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3505 dev);
9f421827 3506 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3507 unsigned long long array_blocks;
3508 int used_disks;
d2e6d5d6 3509
befb629b
AK		 3510 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3511 (units <
9f421827 3512 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3513 (super->migr_rec->rec_status ==
3514 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3515 units++;
3516
d2e6d5d6 3517 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3518
7a862a02 3519 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3520 (unsigned long long)units,
3521 (unsigned long long)blocks_per_unit,
3522 info->reshape_progress);
75156c46 3523
9529d343 3524 used_disks = imsm_num_data_members(prev_map);
75156c46 3525 if (used_disks > 0) {
44490938 3526 array_blocks = per_dev_array_size(map) *
75156c46 3527 used_disks;
b53bfba6
TM		 3528 info->custom_array_size =
3529 round_size_to_mb(array_blocks,
3530 used_disks);
3531
75156c46 3532 }
d2e6d5d6 3533 }
1e5c6983
DW		 3534 case MIGR_VERIFY:
3535 /* we could emulate the checkpointing of
3536 * 'sync_action=check' migrations, but for now
3537 * we just immediately complete them
3538 */
3539 case MIGR_REBUILD:
3540 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3541 case MIGR_STATE_CHANGE:
3542 /* FIXME handle other migrations */
3543 default:
3544 /* we are not dirty, so... */
3545 info->resync_start = MaxSector;
3546 }
b6796ce1 3547 }
301406c9
DW		 3548
3549 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3550 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3551
f35f2525
N		 3552 info->array.major_version = -1;
3553 info->array.minor_version = -2;
4dd2df09 3554 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3555 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3556 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3557
3558 if (dmap) {
3559 int i, j;
3560 for (i=0; i<map_disks; i++) {
3561 dmap[i] = 0;
3562 if (i < info->array.raid_disks) {
3563 struct imsm_disk *dsk;
238c0a71 3564 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3565 dsk = get_imsm_disk(super, j);
3566 if (dsk && (dsk->status & CONFIGURED_DISK))
3567 dmap[i] = 1;
3568 }
3569 }
3570 }
81ac8b4d 3571}
bf5a934a 3572
3b451610
AK		 3573static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3574 int failed, int look_in_map);
3575
3576static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3577 int look_in_map);
3578
3579static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3580{
3581 if (is_gen_migration(dev)) {
3582 int failed;
3583 __u8 map_state;
3584 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3585
3586 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3587 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3588 if (map2->map_state != map_state) {
3589 map2->map_state = map_state;
3590 super->updates_pending++;
3591 }
3592 }
3593}
97b4d0e9
DW		 3594
3595static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3596{
3597 struct dl *d;
3598
3599 for (d = super->missing; d; d = d->next)
3600 if (d->index == index)
3601 return &d->disk;
3602 return NULL;
3603}
3604
a5d85af7 3605static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3606{
3607 struct intel_super *super = st->sb;
4f5bc454 3608 struct imsm_disk *disk;
a5d85af7 3609 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3610 int max_enough = -1;
3611 int i;
3612 struct imsm_super *mpb;
4f5bc454 3613
bf5a934a 3614 if (super->current_vol >= 0) {
a5d85af7 3615 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3616 return;
3617 }
95eeceeb 3618 memset(info, 0, sizeof(*info));
d23fe947
DW		 3619
3620 /* Set raid_disks to zero so that Assemble will always pull in valid
3621 * spares
3622 */
3623 info->array.raid_disks = 0;
cdddbdbc
DW		 3624 info->array.level = LEVEL_CONTAINER;
3625 info->array.layout = 0;
3626 info->array.md_minor = -1;
1011e834 3627 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3628 info->array.utime = 0;
3629 info->array.chunk_size = 0;
3630
3631 info->disk.major = 0;
3632 info->disk.minor = 0;
cdddbdbc 3633 info->disk.raid_disk = -1;
c2c087e6 3634 info->reshape_active = 0;
f35f2525
N		 3635 info->array.major_version = -1;
3636 info->array.minor_version = -2;
c2c087e6 3637 strcpy(info->text_version, "imsm");
a67dd8cc 3638 info->safe_mode_delay = 0;
c2c087e6
DW		 3639 info->disk.number = -1;
3640 info->disk.state = 0;
c5afc314 3641 info->name[0] = 0;
921d9e16 3642 info->recovery_start = MaxSector;
3ad25638 3643 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3644 info->bb.supported = 1;
c2c087e6 3645
97b4d0e9 3646 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3647 mpb = super->anchor;
b7d81a38 3648 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3649
ab3cb6b3
N		 3650 for (i = 0; i < mpb->num_raid_devs; i++) {
3651 struct imsm_dev *dev = get_imsm_dev(super, i);
3652 int failed, enough, j, missing = 0;
3653 struct imsm_map *map;
3654 __u8 state;
97b4d0e9 3655
3b451610
AK		 3656 failed = imsm_count_failed(super, dev, MAP_0);
3657 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3658 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3659
3660 /* any newly missing disks?
3661 * (catches single-degraded vs double-degraded)
3662 */
3663 for (j = 0; j < map->num_members; j++) {
238c0a71 3664 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3665 __u32 idx = ord_to_idx(ord);
3666
20dc76d1
MT		 3667 if (super->disks && super->disks->index == (int)idx)
3668 info->disk.raid_disk = j;
3669
ab3cb6b3
N		 3670 if (!(ord & IMSM_ORD_REBUILD) &&
3671 get_imsm_missing(super, idx)) {
3672 missing = 1;
3673 break;
3674 }
97b4d0e9 3675 }
ab3cb6b3
N		 3676
3677 if (state == IMSM_T_STATE_FAILED)
3678 enough = -1;
3679 else if (state == IMSM_T_STATE_DEGRADED &&
3680 (state != map->map_state || missing))
3681 enough = 0;
3682 else /* we're normal, or already degraded */
3683 enough = 1;
d2bde6d3
AK		 3684 if (is_gen_migration(dev) && missing) {
3685 /* during general migration we need all disks
3686 * that process is running on.
3687 * No new missing disk is allowed.
3688 */
3689 max_enough = -1;
3690 enough = -1;
3691 /* no more checks necessary
3692 */
3693 break;
3694 }
ab3cb6b3
N		 3695 /* in the missing/failed disk case check to see
3696 * if at least one array is runnable
3697 */
3698 max_enough = max(max_enough, enough);
3699 }
1ade5cc1 3700 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3701 info->container_enough = max_enough;
97b4d0e9 3702
4a04ec6c 3703 if (super->disks) {
14e8215b
DW		 3704 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3705
b9f594fe 3706 disk = &super->disks->disk;
5551b113 3707 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3708 info->component_size = reserved;
25ed7e59 3709 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3710 /* we don't change info->disk.raid_disk here because
3711 * this state will be finalized in mdmon after we have
3712 * found the 'most fresh' version of the metadata
3713 */
25ed7e59 3714 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3715 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3716 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3717 }
a575e2a7
DW		 3718
3719 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3720 * ->compare_super may have updated the 'num_raid_devs' field for spares
3721 */
3722 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3723 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3724 else
3725 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3726
3727 /* I don't know how to compute 'map' on imsm, so use safe default */
3728 if (map) {
3729 int i;
3730 for (i = 0; i < map_disks; i++)
3731 map[i] = 1;
3732 }
3733
cdddbdbc
DW		 3734}
3735
5c4cd5da
AC		 3736/* allocates memory and fills disk in mdinfo structure
3737 * for each disk in array */
3738struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3739{
594dc1b8 3740 struct mdinfo *mddev;
5c4cd5da
AC		 3741 struct intel_super *super = st->sb;
3742 struct imsm_disk *disk;
3743 int count = 0;
3744 struct dl *dl;
3745 if (!super || !super->disks)
3746 return NULL;
3747 dl = super->disks;
503975b9 3748 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3749 while (dl) {
3750 struct mdinfo *tmp;
3751 disk = &dl->disk;
503975b9 3752 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3753 if (mddev->devs)
3754 tmp->next = mddev->devs;
3755 mddev->devs = tmp;
3756 tmp->disk.number = count++;
3757 tmp->disk.major = dl->major;
3758 tmp->disk.minor = dl->minor;
3759 tmp->disk.state = is_configured(disk) ?
3760 (1 << MD_DISK_ACTIVE) : 0;
3761 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3762 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3763 tmp->disk.raid_disk = -1;
3764 dl = dl->next;
3765 }
3766 return mddev;
3767}
3768
cdddbdbc
DW		 3769static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3770 char *update, char *devname, int verbose,
3771 int uuid_set, char *homehost)
3772{
f352c545
DW		 3773 /* For 'assemble' and 'force' we need to return non-zero if any
3774 * change was made. For others, the return value is ignored.
3775 * Update options are:
3776 * force-one : This device looks a bit old but needs to be included,
3777 * update age info appropriately.
3778 * assemble: clear any 'faulty' flag to allow this device to
3779 * be assembled.
3780 * force-array: Array is degraded but being forced, mark it clean
3781 * if that will be needed to assemble it.
3782 *
3783 * newdev: not used ????
3784 * grow: Array has gained a new device - this is currently for
3785 * linear only
3786 * resync: mark as dirty so a resync will happen.
3787 * name: update the name - preserving the homehost
6e46bf34 3788 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3789 *
3790 * Following are not relevant for this imsm:
3791 * sparc2.2 : update from old dodgey metadata
3792 * super-minor: change the preferred_minor number
3793 * summaries: update redundant counters.
f352c545
DW		 3794 * homehost: update the recorded homehost
3795 * _reshape_progress: record new reshape_progress position.
3796 */
6e46bf34
DW		 3797 int rv = 1;
3798 struct intel_super *super = st->sb;
3799 struct imsm_super *mpb;
f352c545 3800
6e46bf34
DW		 3801 /* we can only update container info */
3802 if (!super || super->current_vol >= 0 || !super->anchor)
3803 return 1;
3804
3805 mpb = super->anchor;
3806
81a5b4f5
N		 3807 if (strcmp(update, "uuid") == 0) {
3808 /* We take this to mean that the family_num should be updated.
3809 * However that is much smaller than the uuid so we cannot really
3810 * allow an explicit uuid to be given. And it is hard to reliably
3811 * know if one was.
3812 * So if !uuid_set we know the current uuid is random and just used
3813 * the first 'int' and copy it to the other 3 positions.
3814 * Otherwise we require the 4 'int's to be the same as would be the
3815 * case if we are using a random uuid. So an explicit uuid will be
3816 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3817 */
81a5b4f5
N		 3818 if (!uuid_set) {
3819 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3820 rv = 0;
81a5b4f5
N		 3821 } else {
3822 if (info->uuid[0] != info->uuid[1] ||
3823 info->uuid[1] != info->uuid[2] ||
3824 info->uuid[2] != info->uuid[3])
3825 rv = -1;
3826 else
3827 rv = 0;
6e46bf34 3828 }
81a5b4f5
N		 3829 if (rv == 0)
3830 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3831 } else if (strcmp(update, "assemble") == 0)
3832 rv = 0;
3833 else
1e2b2765 3834 rv = -1;
f352c545 3835
6e46bf34
DW		 3836 /* successful update? recompute checksum */
3837 if (rv == 0)
3838 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3839
3840 return rv;
cdddbdbc
DW		 3841}
3842
c2c087e6 3843static size_t disks_to_mpb_size(int disks)
cdddbdbc 3844{
c2c087e6 3845 size_t size;
cdddbdbc 3846
c2c087e6
DW		 3847 size = sizeof(struct imsm_super);
3848 size += (disks - 1) * sizeof(struct imsm_disk);
3849 size += 2 * sizeof(struct imsm_dev);
3850 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3851 size += (4 - 2) * sizeof(struct imsm_map);
3852 /* 4 possible disk_ord_tbl's */
3853 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3854 /* maximum bbm log */
3855 size += sizeof(struct bbm_log);
c2c087e6
DW		 3856
3857 return size;
3858}
3859
387fcd59
N		 3860static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3861 unsigned long long data_offset)
c2c087e6
DW		 3862{
3863 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3864 return 0;
3865
3866 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3867}
3868
ba2de7ba
DW		 3869static void free_devlist(struct intel_super *super)
3870{
3871 struct intel_dev *dv;
3872
3873 while (super->devlist) {
3874 dv = super->devlist->next;
3875 free(super->devlist->dev);
3876 free(super->devlist);
3877 super->devlist = dv;
3878 }
3879}
3880
3881static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3882{
3883 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3884}
3885
cdddbdbc
DW		 3886static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3887{
3888 /*
3889 * return:
3890 * 0 same, or first was empty, and second was copied
3891 * 1 second had wrong number
3892 * 2 wrong uuid
3893 * 3 wrong other info
3894 */
3895 struct intel_super *first = st->sb;
3896 struct intel_super *sec = tst->sb;
3897
5d500228
N		 3898 if (!first) {
3899 st->sb = tst->sb;
3900 tst->sb = NULL;
3901 return 0;
3902 }
8603ea6f
LM		 3903 /* in platform dependent environment test if the disks
3904 * use the same Intel hba
cb8f6859 3905 * If not on Intel hba at all, allow anything.
8603ea6f 3906 */
6b781d33
AP		 3907 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3908 if (first->hba->type != sec->hba->type) {
8603ea6f 3909 fprintf(stderr,
6b781d33
AP		 3910 "HBAs of devices do not match %s != %s\n",
3911 get_sys_dev_type(first->hba->type),
3912 get_sys_dev_type(sec->hba->type));
3913 return 3;
3914 }
3915 if (first->orom != sec->orom) {
3916 fprintf(stderr,
3917 "HBAs of devices do not match %s != %s\n",
3918 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3919 return 3;
3920 }
3921 }
cdddbdbc 3922
d23fe947
DW		 3923 /* if an anchor does not have num_raid_devs set then it is a free
3924 * floating spare
3925 */
3926 if (first->anchor->num_raid_devs > 0 &&
3927 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3928 /* Determine if these disks might ever have been
3929 * related. Further disambiguation can only take place
3930 * in load_super_imsm_all
3931 */
3932 __u32 first_family = first->anchor->orig_family_num;
3933 __u32 sec_family = sec->anchor->orig_family_num;
3934
f796af5d
DW		 3935 if (memcmp(first->anchor->sig, sec->anchor->sig,
3936 MAX_SIGNATURE_LENGTH) != 0)
3937 return 3;
3938
a2b97981
DW		 3939 if (first_family == 0)
3940 first_family = first->anchor->family_num;
3941 if (sec_family == 0)
3942 sec_family = sec->anchor->family_num;
3943
3944 if (first_family != sec_family)
d23fe947 3945 return 3;
f796af5d 3946
d23fe947 3947 }
cdddbdbc 3948
3e372e5a
DW		 3949 /* if 'first' is a spare promote it to a populated mpb with sec's
3950 * family number
3951 */
3952 if (first->anchor->num_raid_devs == 0 &&
3953 sec->anchor->num_raid_devs > 0) {
78d30f94 3954 int i;
ba2de7ba
DW		 3955 struct intel_dev *dv;
3956 struct imsm_dev *dev;
78d30f94
DW		 3957
3958 /* we need to copy raid device info from sec if an allocation
3959 * fails here we don't associate the spare
3960 */
3961 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3962 dv = xmalloc(sizeof(*dv));
3963 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3964 dv->dev = dev;
3965 dv->index = i;
3966 dv->next = first->devlist;
3967 first->devlist = dv;
78d30f94 3968 }
709743c5 3969 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3970 /* allocation failure */
3971 free_devlist(first);
e12b3daa 3972 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3973 return 3;
78d30f94 3974 }
3e372e5a 3975 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3976 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3977 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3978 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3979 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3980 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3981 }
3982
cdddbdbc
DW		 3983 return 0;
3984}
3985
0030e8d6
DW		 3986static void fd2devname(int fd, char *name)
3987{
3988 struct stat st;
3989 char path[256];
33a6535d 3990 char dname[PATH_MAX];
0030e8d6
DW		 3991 char *nm;
3992 int rv;
3993
3994 name[0] = '\0';
3995 if (fstat(fd, &st) != 0)
3996 return;
3997 sprintf(path, "/sys/dev/block/%d:%d",
3998 major(st.st_rdev), minor(st.st_rdev));
3999
9cf014ec 4000 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 4001 if (rv <= 0)
4002 return;
9587c373 4003
0030e8d6
DW		 4004 dname[rv] = '\0';
4005 nm = strrchr(dname, '/');
7897de29
JS		 4006 if (nm) {
4007 nm++;
4008 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
4009 }
0030e8d6
DW		 4010}
4011
21e9380b
AP		 4012static int nvme_get_serial(int fd, void *buf, size_t buf_len)
4013{
4014 char path[60];
4015 char *name = fd2kname(fd);
4016
4017 if (!name)
4018 return 1;
4019
4020 if (strncmp(name, "nvme", 4) != 0)
4021 return 1;
4022
4023 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
4024
4025 return load_sys(path, buf, buf_len);
4026}
4027
cdddbdbc
DW		 4028extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
4029
4030static int imsm_read_serial(int fd, char *devname,
4031 __u8 serial[MAX_RAID_SERIAL_LEN])
4032{
21e9380b 4033 char buf[50];
cdddbdbc 4034 int rv;
1f24f035 4035 int len;
316e2bf4
DW		 4036 char *dest;
4037 char *src;
21e9380b
AP		 4038 unsigned int i;
4039
4040 memset(buf, 0, sizeof(buf));
cdddbdbc 4041
21e9380b 4042 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 4043
21e9380b
AP		 4044 if (rv)
4045 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 4046
40ebbb9c 4047 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 4048 memset(serial, 0, MAX_RAID_SERIAL_LEN);
4049 fd2devname(fd, (char *) serial);
0030e8d6
DW		 4050 return 0;
4051 }
4052
cdddbdbc
DW		 4053 if (rv != 0) {
4054 if (devname)
e7b84f9d
N		 4055 pr_err("Failed to retrieve serial for %s\n",
4056 devname);
cdddbdbc
DW		 4057 return rv;
4058 }
4059
316e2bf4
DW		 4060 /* trim all whitespace and non-printable characters and convert
4061 * ':' to ';'
4062 */
21e9380b
AP		 4063 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4064 src = &buf[i];
316e2bf4
DW		 4065 if (*src > 0x20) {
4066 /* ':' is reserved for use in placeholder serial
4067 * numbers for missing disks
4068 */
4069 if (*src == ':')
4070 *dest++ = ';';
4071 else
4072 *dest++ = *src;
4073 }
4074 }
21e9380b
AP		 4075 len = dest - buf;
4076 dest = buf;
316e2bf4
DW		 4077
4078 /* truncate leading characters */
4079 if (len > MAX_RAID_SERIAL_LEN) {
4080 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 4081 len = MAX_RAID_SERIAL_LEN;
316e2bf4 4082 }
5c3db629 4083
5c3db629 4084 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 4085 memcpy(serial, dest, len);
cdddbdbc
DW		 4086
4087 return 0;
4088}
4089
1f24f035
DW		 4090static int serialcmp(__u8 *s1, __u8 *s2)
4091{
4092 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4093}
4094
4095static void serialcpy(__u8 *dest, __u8 *src)
4096{
4097 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4098}
4099
54c2c1ea
DW		 4100static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4101{
4102 struct dl *dl;
4103
4104 for (dl = super->disks; dl; dl = dl->next)
4105 if (serialcmp(dl->serial, serial) == 0)
4106 break;
4107
4108 return dl;
4109}
4110
a2b97981
DW		 4111static struct imsm_disk *
4112__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4113{
4114 int i;
4115
4116 for (i = 0; i < mpb->num_disks; i++) {
4117 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4118
4119 if (serialcmp(disk->serial, serial) == 0) {
4120 if (idx)
4121 *idx = i;
4122 return disk;
4123 }
4124 }
4125
4126 return NULL;
4127}
4128
cdddbdbc
DW		 4129static int
4130load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4131{
a2b97981 4132 struct imsm_disk *disk;
cdddbdbc
DW		 4133 struct dl *dl;
4134 struct stat stb;
cdddbdbc 4135 int rv;
a2b97981 4136 char name[40];
d23fe947
DW		 4137 __u8 serial[MAX_RAID_SERIAL_LEN];
4138
4139 rv = imsm_read_serial(fd, devname, serial);
4140
4141 if (rv != 0)
4142 return 2;
4143
503975b9 4144 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4145
a2b97981
DW		 4146 fstat(fd, &stb);
4147 dl->major = major(stb.st_rdev);
4148 dl->minor = minor(stb.st_rdev);
4149 dl->next = super->disks;
4150 dl->fd = keep_fd ? fd : -1;
4151 assert(super->disks == NULL);
4152 super->disks = dl;
4153 serialcpy(dl->serial, serial);
4154 dl->index = -2;
4155 dl->e = NULL;
4156 fd2devname(fd, name);
4157 if (devname)
503975b9 4158 dl->devname = xstrdup(devname);
a2b97981 4159 else
503975b9 4160 dl->devname = xstrdup(name);
cdddbdbc 4161
d23fe947 4162 /* look up this disk's index in the current anchor */
a2b97981
DW		 4163 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4164 if (disk) {
4165 dl->disk = *disk;
4166 /* only set index on disks that are a member of a
4167 * populated contianer, i.e. one with raid_devs
4168 */
4169 if (is_failed(&dl->disk))
3f6efecc 4170 dl->index = -2;
2432ce9b 4171 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4172 dl->index = -1;
3f6efecc
DW		 4173 }
4174
949c47a0
DW		 4175 return 0;
4176}
4177
0c046afd
DW		 4178/* When migrating map0 contains the 'destination' state while map1
4179 * contains the current state. When not migrating map0 contains the
4180 * current state. This routine assumes that map[0].map_state is set to
4181 * the current array state before being called.
4182 *
4183 * Migration is indicated by one of the following states
4184 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4185 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4186 * map1state=unitialized)
1484e727 4187 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4188 * map1state=normal)
e3bba0e0 4189 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4190 * map1state=degraded)
8e59f3d8
AK		 4191 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4192 * map1state=normal)
0c046afd 4193 */
8e59f3d8
AK		 4194static void migrate(struct imsm_dev *dev, struct intel_super *super,
4195 __u8 to_state, int migr_type)
3393c6af 4196{
0c046afd 4197 struct imsm_map *dest;
238c0a71 4198 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4199
0c046afd 4200 dev->vol.migr_state = 1;
1484e727 4201 set_migr_type(dev, migr_type);
f8f603f1 4202 dev->vol.curr_migr_unit = 0;
238c0a71 4203 dest = get_imsm_map(dev, MAP_1);
0c046afd 4204
0556e1a2 4205 /* duplicate and then set the target end state in map[0] */
3393c6af 4206 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4207 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4208 __u32 ord;
4209 int i;
4210
4211 for (i = 0; i < src->num_members; i++) {
4212 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4213 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4214 }
4215 }
4216
8e59f3d8
AK		 4217 if (migr_type == MIGR_GEN_MIGR)
4218 /* Clear migration record */
4219 memset(super->migr_rec, 0, sizeof(struct migr_record));
4220
0c046afd 4221 src->map_state = to_state;
949c47a0 4222}
f8f603f1 4223
809da78e
AK		 4224static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4225 __u8 map_state)
f8f603f1 4226{
238c0a71
AK		 4227 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4228 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4229 MAP_0 : MAP_1);
28bce06f 4230 int i, j;
0556e1a2
DW		 4231
4232 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4233 * completed in the last migration.
4234 *
28bce06f 4235 * FIXME add support for raid-level-migration
0556e1a2 4236 */
089f9d79
JS		 4237 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4238 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4239 /* when final map state is other than expected
4240 * merge maps (not for migration)
4241 */
4242 int failed;
4243
4244 for (i = 0; i < prev->num_members; i++)
4245 for (j = 0; j < map->num_members; j++)
4246 /* during online capacity expansion
4247 * disks position can be changed
4248 * if takeover is used
4249 */
4250 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4251 ord_to_idx(prev->disk_ord_tbl[i])) {
4252 map->disk_ord_tbl[j] |=
4253 prev->disk_ord_tbl[i];
4254 break;
4255 }
4256 failed = imsm_count_failed(super, dev, MAP_0);
4257 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4258 }
f8f603f1
DW		 4259
4260 dev->vol.migr_state = 0;
ea672ee1 4261 set_migr_type(dev, 0);
f8f603f1
DW		 4262 dev->vol.curr_migr_unit = 0;
4263 map->map_state = map_state;
4264}
949c47a0
DW		 4265
4266static int parse_raid_devices(struct intel_super *super)
4267{
4268 int i;
4269 struct imsm_dev *dev_new;
4d7b1503 4270 size_t len, len_migr;
401d313b 4271 size_t max_len = 0;
4d7b1503
DW		 4272 size_t space_needed = 0;
4273 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4274
4275 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4276 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4277 struct intel_dev *dv;
949c47a0 4278
4d7b1503
DW		 4279 len = sizeof_imsm_dev(dev_iter, 0);
4280 len_migr = sizeof_imsm_dev(dev_iter, 1);
4281 if (len_migr > len)
4282 space_needed += len_migr - len;
ca9de185 4283
503975b9 4284 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4285 if (max_len < len_migr)
4286 max_len = len_migr;
4287 if (max_len > len_migr)
4288 space_needed += max_len - len_migr;
503975b9 4289 dev_new = xmalloc(max_len);
949c47a0 4290 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4291 dv->dev = dev_new;
4292 dv->index = i;
4293 dv->next = super->devlist;
4294 super->devlist = dv;
949c47a0 4295 }
cdddbdbc 4296
4d7b1503
DW		 4297 /* ensure that super->buf is large enough when all raid devices
4298 * are migrating
4299 */
4300 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4301 void *buf;
4302
f36a9ecd
PB		 4303 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4304 super->sector_size);
4305 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4306 return 1;
4307
1f45a8ad
DW		 4308 memcpy(buf, super->buf, super->len);
4309 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4310 free(super->buf);
4311 super->buf = buf;
4312 super->len = len;
4313 }
ca9de185 4314
bbab0940
TM		 4315 super->extra_space += space_needed;
4316
cdddbdbc
DW		 4317 return 0;
4318}
4319
e2f41b2c
AK		 4320/*******************************************************************************
4321 * Function: check_mpb_migr_compatibility
4322 * Description: Function checks for unsupported migration features:
4323 * - migration optimization area (pba_of_lba0)
4324 * - descending reshape (ascending_migr)
4325 * Parameters:
4326 * super : imsm metadata information
4327 * Returns:
4328 * 0 : migration is compatible
4329 * -1 : migration is not compatible
4330 ******************************************************************************/
4331int check_mpb_migr_compatibility(struct intel_super *super)
4332{
4333 struct imsm_map *map0, *map1;
4334 struct migr_record *migr_rec = super->migr_rec;
4335 int i;
4336
4337 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4338 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4339
4340 if (dev_iter &&
4341 dev_iter->vol.migr_state == 1 &&
4342 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4343 /* This device is migrating */
238c0a71
AK		 4344 map0 = get_imsm_map(dev_iter, MAP_0);
4345 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4346 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4347 /* migration optimization area was used */
4348 return -1;
fc54fe7a
JS		 4349 if (migr_rec->ascending_migr == 0 &&
4350 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4351 /* descending reshape not supported yet */
4352 return -1;
4353 }
4354 }
4355 return 0;
4356}
4357
d23fe947 4358static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4359
cdddbdbc 4360/* load_imsm_mpb - read matrix metadata
f2f5c343 4361 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4362 */
4363static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4364{
4365 unsigned long long dsize;
cdddbdbc 4366 unsigned long long sectors;
f36a9ecd 4367 unsigned int sector_size = super->sector_size;
cdddbdbc 4368 struct stat;
6416d527 4369 struct imsm_super *anchor;
cdddbdbc
DW		 4370 __u32 check_sum;
4371
cdddbdbc 4372 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4373 if (dsize < 2*sector_size) {
64436f06 4374 if (devname)
e7b84f9d
N		 4375 pr_err("%s: device to small for imsm\n",
4376 devname);
64436f06
N		 4377 return 1;
4378 }
cdddbdbc 4379
f36a9ecd 4380 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4381 if (devname)
e7b84f9d
N		 4382 pr_err("Cannot seek to anchor block on %s: %s\n",
4383 devname, strerror(errno));
cdddbdbc
DW		 4384 return 1;
4385 }
4386
f36a9ecd 4387 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4388 if (devname)
7a862a02 4389 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4390 return 1;
4391 }
466070ad 4392 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4393 if (devname)
e7b84f9d
N		 4394 pr_err("Cannot read anchor block on %s: %s\n",
4395 devname, strerror(errno));
6416d527 4396 free(anchor);
cdddbdbc
DW		 4397 return 1;
4398 }
4399
6416d527 4400 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4401 if (devname)
e7b84f9d 4402 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4403 free(anchor);
cdddbdbc
DW		 4404 return 2;
4405 }
4406
d23fe947 4407 __free_imsm(super, 0);
f2f5c343
LM		 4408 /* reload capability and hba */
4409
4410 /* capability and hba must be updated with new super allocation */
d424212e 4411 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4412 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4413 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4414 if (devname)
e7b84f9d
N		 4415 pr_err("unable to allocate %zu byte mpb buffer\n",
4416 super->len);
6416d527 4417 free(anchor);
cdddbdbc
DW		 4418 return 2;
4419 }
f36a9ecd 4420 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4421
f36a9ecd 4422 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4423 free(anchor);
8e59f3d8 4424
85337573
AO		 4425 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4426 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4427 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4428 free(super->buf);
4429 return 2;
4430 }
51d83f5d 4431 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4432
949c47a0 4433 if (!sectors) {
ecf45690
DW		 4434 check_sum = __gen_imsm_checksum(super->anchor);
4435 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4436 if (devname)
e7b84f9d
N		 4437 pr_err("IMSM checksum %x != %x on %s\n",
4438 check_sum,
4439 __le32_to_cpu(super->anchor->check_sum),
4440 devname);
ecf45690
DW		 4441 return 2;
4442 }
4443
a2b97981 4444 return 0;
949c47a0 4445 }
cdddbdbc
DW		 4446
4447 /* read the extended mpb */
f36a9ecd 4448 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4449 if (devname)
e7b84f9d
N		 4450 pr_err("Cannot seek to extended mpb on %s: %s\n",
4451 devname, strerror(errno));
cdddbdbc
DW		 4452 return 1;
4453 }
4454
f36a9ecd
PB		 4455 if ((unsigned int)read(fd, super->buf + sector_size,
4456 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4457 if (devname)
e7b84f9d
N		 4458 pr_err("Cannot read extended mpb on %s: %s\n",
4459 devname, strerror(errno));
cdddbdbc
DW		 4460 return 2;
4461 }
4462
949c47a0
DW		 4463 check_sum = __gen_imsm_checksum(super->anchor);
4464 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4465 if (devname)
e7b84f9d
N		 4466 pr_err("IMSM checksum %x != %x on %s\n",
4467 check_sum, __le32_to_cpu(super->anchor->check_sum),
4468 devname);
db575f3b 4469 return 3;
cdddbdbc
DW		 4470 }
4471
a2b97981
DW		 4472 return 0;
4473}
4474
8e59f3d8
AK		 4475static int read_imsm_migr_rec(int fd, struct intel_super *super);
4476
97f81ee2
CA		 4477/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4478static void clear_hi(struct intel_super *super)
4479{
4480 struct imsm_super *mpb = super->anchor;
4481 int i, n;
4482 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4483 return;
4484 for (i = 0; i < mpb->num_disks; ++i) {
4485 struct imsm_disk *disk = &mpb->disk[i];
4486 disk->total_blocks_hi = 0;
4487 }
4488 for (i = 0; i < mpb->num_raid_devs; ++i) {
4489 struct imsm_dev *dev = get_imsm_dev(super, i);
4490 if (!dev)
4491 return;
4492 for (n = 0; n < 2; ++n) {
4493 struct imsm_map *map = get_imsm_map(dev, n);
4494 if (!map)
4495 continue;
4496 map->pba_of_lba0_hi = 0;
4497 map->blocks_per_member_hi = 0;
4498 map->num_data_stripes_hi = 0;
4499 }
4500 }
4501}
4502
a2b97981
DW		 4503static int
4504load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4505{
4506 int err;
4507
4508 err = load_imsm_mpb(fd, super, devname);
4509 if (err)
4510 return err;
f36a9ecd
PB		 4511 if (super->sector_size == 4096)
4512 convert_from_4k(super);
a2b97981
DW		 4513 err = load_imsm_disk(fd, super, devname, keep_fd);
4514 if (err)
4515 return err;
4516 err = parse_raid_devices(super);
8d67477f
TM		 4517 if (err)
4518 return err;
4519 err = load_bbm_log(super);
97f81ee2 4520 clear_hi(super);
a2b97981 4521 return err;
cdddbdbc
DW		 4522}
4523
ae6aad82
DW		 4524static void __free_imsm_disk(struct dl *d)
4525{
4526 if (d->fd >= 0)
4527 close(d->fd);
4528 if (d->devname)
4529 free(d->devname);
0dcecb2e
DW		 4530 if (d->e)
4531 free(d->e);
ae6aad82
DW		 4532 free(d);
4533
4534}
1a64be56 4535
cdddbdbc
DW		 4536static void free_imsm_disks(struct intel_super *super)
4537{
47ee5a45 4538 struct dl *d;
cdddbdbc 4539
47ee5a45
DW		 4540 while (super->disks) {
4541 d = super->disks;
cdddbdbc 4542 super->disks = d->next;
ae6aad82 4543 __free_imsm_disk(d);
cdddbdbc 4544 }
cb82edca
AK		 4545 while (super->disk_mgmt_list) {
4546 d = super->disk_mgmt_list;
4547 super->disk_mgmt_list = d->next;
4548 __free_imsm_disk(d);
4549 }
47ee5a45
DW		 4550 while (super->missing) {
4551 d = super->missing;
4552 super->missing = d->next;
4553 __free_imsm_disk(d);
4554 }
4555
cdddbdbc
DW		 4556}
4557
9ca2c81c 4558/* free all the pieces hanging off of a super pointer */
d23fe947 4559static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4560{
88654014
LM		 4561 struct intel_hba *elem, *next;
4562
9ca2c81c 4563 if (super->buf) {
949c47a0 4564 free(super->buf);
9ca2c81c
DW		 4565 super->buf = NULL;
4566 }
f2f5c343
LM		 4567 /* unlink capability description */
4568 super->orom = NULL;
8e59f3d8
AK		 4569 if (super->migr_rec_buf) {
4570 free(super->migr_rec_buf);
4571 super->migr_rec_buf = NULL;
4572 }
d23fe947
DW		 4573 if (free_disks)
4574 free_imsm_disks(super);
ba2de7ba 4575 free_devlist(super);
88654014
LM		 4576 elem = super->hba;
4577 while (elem) {
4578 if (elem->path)
4579 free((void *)elem->path);
4580 next = elem->next;
4581 free(elem);
4582 elem = next;
88c32bb1 4583 }
8d67477f
TM		 4584 if (super->bbm_log)
4585 free(super->bbm_log);
88654014 4586 super->hba = NULL;
cdddbdbc
DW		 4587}
4588
9ca2c81c
DW		 4589static void free_imsm(struct intel_super *super)
4590{
d23fe947 4591 __free_imsm(super, 1);
928f1424 4592 free(super->bb.entries);
9ca2c81c
DW		 4593 free(super);
4594}
cdddbdbc
DW		 4595
4596static void free_super_imsm(struct supertype *st)
4597{
4598 struct intel_super *super = st->sb;
4599
4600 if (!super)
4601 return;
4602
4603 free_imsm(super);
4604 st->sb = NULL;
4605}
4606
49133e57 4607static struct intel_super *alloc_super(void)
c2c087e6 4608{
503975b9 4609 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4610
503975b9
N		 4611 super->current_vol = -1;
4612 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4613
4614 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4615 sizeof(struct md_bb_entry));
4616 if (!super->bb.entries) {
4617 free(super);
4618 return NULL;
4619 }
4620
c2c087e6
DW		 4621 return super;
4622}
4623
f0f5a016
LM		 4624/*
4625 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4626 */
d424212e 4627static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4628{
4629 struct sys_dev *hba_name;
4630 int rv = 0;
4631
3a30e28e
MT		 4632 if (fd >= 0 && test_partition(fd)) {
4633 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4634 devname);
4635 return 1;
4636 }
089f9d79 4637 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4638 super->orom = NULL;
f0f5a016
LM		 4639 super->hba = NULL;
4640 return 0;
4641 }
4642 hba_name = find_disk_attached_hba(fd, NULL);
4643 if (!hba_name) {
d424212e 4644 if (devname)
e7b84f9d
N		 4645 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4646 devname);
f0f5a016
LM		 4647 return 1;
4648 }
4649 rv = attach_hba_to_super(super, hba_name);
4650 if (rv == 2) {
d424212e
N		 4651 if (devname) {
4652 struct intel_hba *hba = super->hba;
f0f5a016 4653
60f0f54d
PB		 4654 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4655 " but the container is assigned to Intel(R) %s %s (",
d424212e 4656 devname,
614902f6 4657 get_sys_dev_type(hba_name->type),
60f0f54d 4658 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4659 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4660 get_sys_dev_type(super->hba->type),
4661 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4662
f0f5a016
LM		 4663 while (hba) {
4664 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4665 if (hba->next)
4666 fprintf(stderr, ", ");
4667 hba = hba->next;
4668 }
6b781d33 4669 fprintf(stderr, ").\n"
cca67208 4670 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4671 }
f0f5a016
LM		 4672 return 2;
4673 }
6b781d33 4674 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4675 if (!super->orom)
4676 return 3;
614902f6 4677
f0f5a016
LM		 4678 return 0;
4679}
4680
47ee5a45
DW		 4681/* find_missing - helper routine for load_super_imsm_all that identifies
4682 * disks that have disappeared from the system. This routine relies on
4683 * the mpb being uptodate, which it is at load time.
4684 */
4685static int find_missing(struct intel_super *super)
4686{
4687 int i;
4688 struct imsm_super *mpb = super->anchor;
4689 struct dl *dl;
4690 struct imsm_disk *disk;
47ee5a45
DW		 4691
4692 for (i = 0; i < mpb->num_disks; i++) {
4693 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4694 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4695 if (dl)
4696 continue;
47ee5a45 4697
503975b9 4698 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4699 dl->major = 0;
4700 dl->minor = 0;
4701 dl->fd = -1;
503975b9 4702 dl->devname = xstrdup("missing");
47ee5a45
DW		 4703 dl->index = i;
4704 serialcpy(dl->serial, disk->serial);
4705 dl->disk = *disk;
689c9bf3 4706 dl->e = NULL;
47ee5a45
DW		 4707 dl->next = super->missing;
4708 super->missing = dl;
4709 }
4710
4711 return 0;
4712}
4713
a2b97981
DW		 4714static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4715{
4716 struct intel_disk *idisk = disk_list;
4717
4718 while (idisk) {
4719 if (serialcmp(idisk->disk.serial, serial) == 0)
4720 break;
4721 idisk = idisk->next;
4722 }
4723
4724 return idisk;
4725}
4726
4727static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4728 struct intel_super *super,
4729 struct intel_disk **disk_list)
4730{
4731 struct imsm_disk *d = &super->disks->disk;
4732 struct imsm_super *mpb = super->anchor;
4733 int i, j;
4734
4735 for (i = 0; i < tbl_size; i++) {
4736 struct imsm_super *tbl_mpb = table[i]->anchor;
4737 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4738
4739 if (tbl_mpb->family_num == mpb->family_num) {
4740 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4741 dprintf("mpb from %d:%d matches %d:%d\n",
4742 super->disks->major,
a2b97981
DW		 4743 super->disks->minor,
4744 table[i]->disks->major,
4745 table[i]->disks->minor);
4746 break;
4747 }
4748
4749 if (((is_configured(d) && !is_configured(tbl_d)) ||
4750 is_configured(d) == is_configured(tbl_d)) &&
4751 tbl_mpb->generation_num < mpb->generation_num) {
4752 /* current version of the mpb is a
4753 * better candidate than the one in
4754 * super_table, but copy over "cross
4755 * generational" status
4756 */
4757 struct intel_disk *idisk;
4758
1ade5cc1
N		 4759 dprintf("mpb from %d:%d replaces %d:%d\n",
4760 super->disks->major,
a2b97981
DW		 4761 super->disks->minor,
4762 table[i]->disks->major,
4763 table[i]->disks->minor);
4764
4765 idisk = disk_list_get(tbl_d->serial, *disk_list);
4766 if (idisk && is_failed(&idisk->disk))
4767 tbl_d->status |= FAILED_DISK;
4768 break;
4769 } else {
4770 struct intel_disk *idisk;
4771 struct imsm_disk *disk;
4772
4773 /* tbl_mpb is more up to date, but copy
4774 * over cross generational status before
4775 * returning
4776 */
4777 disk = __serial_to_disk(d->serial, mpb, NULL);
4778 if (disk && is_failed(disk))
4779 d->status |= FAILED_DISK;
4780
4781 idisk = disk_list_get(d->serial, *disk_list);
4782 if (idisk) {
4783 idisk->owner = i;
4784 if (disk && is_configured(disk))
4785 idisk->disk.status |= CONFIGURED_DISK;
4786 }
4787
1ade5cc1
N		 4788 dprintf("mpb from %d:%d prefer %d:%d\n",
4789 super->disks->major,
a2b97981
DW		 4790 super->disks->minor,
4791 table[i]->disks->major,
4792 table[i]->disks->minor);
4793
4794 return tbl_size;
4795 }
4796 }
4797 }
4798
4799 if (i >= tbl_size)
4800 table[tbl_size++] = super;
4801 else
4802 table[i] = super;
4803
4804 /* update/extend the merged list of imsm_disk records */
4805 for (j = 0; j < mpb->num_disks; j++) {
4806 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4807 struct intel_disk *idisk;
4808
4809 idisk = disk_list_get(disk->serial, *disk_list);
4810 if (idisk) {
4811 idisk->disk.status |= disk->status;
4812 if (is_configured(&idisk->disk) ||
4813 is_failed(&idisk->disk))
4814 idisk->disk.status &= ~(SPARE_DISK);
4815 } else {
503975b9 4816 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4817 idisk->owner = IMSM_UNKNOWN_OWNER;
4818 idisk->disk = *disk;
4819 idisk->next = *disk_list;
4820 *disk_list = idisk;
4821 }
4822
4823 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4824 idisk->owner = i;
4825 }
4826
4827 return tbl_size;
4828}
4829
4830static struct intel_super *
4831validate_members(struct intel_super *super, struct intel_disk *disk_list,
4832 const int owner)
4833{
4834 struct imsm_super *mpb = super->anchor;
4835 int ok_count = 0;
4836 int i;
4837
4838 for (i = 0; i < mpb->num_disks; i++) {
4839 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4840 struct intel_disk *idisk;
4841
4842 idisk = disk_list_get(disk->serial, disk_list);
4843 if (idisk) {
4844 if (idisk->owner == owner ||
4845 idisk->owner == IMSM_UNKNOWN_OWNER)
4846 ok_count++;
4847 else
1ade5cc1
N		 4848 dprintf("'%.16s' owner %d != %d\n",
4849 disk->serial, idisk->owner,
a2b97981
DW		 4850 owner);
4851 } else {
1ade5cc1
N		 4852 dprintf("unknown disk %x [%d]: %.16s\n",
4853 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4854 disk->serial);
4855 break;
4856 }
4857 }
4858
4859 if (ok_count == mpb->num_disks)
4860 return super;
4861 return NULL;
4862}
4863
4864static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4865{
4866 struct intel_super *s;
4867
4868 for (s = super_list; s; s = s->next) {
4869 if (family_num != s->anchor->family_num)
4870 continue;
e12b3daa 4871 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4872 __le32_to_cpu(family_num), s->disks->devname);
4873 }
4874}
4875
4876static struct intel_super *
4877imsm_thunderdome(struct intel_super **super_list, int len)
4878{
4879 struct intel_super *super_table[len];
4880 struct intel_disk *disk_list = NULL;
4881 struct intel_super *champion, *spare;
4882 struct intel_super *s, **del;
4883 int tbl_size = 0;
4884 int conflict;
4885 int i;
4886
4887 memset(super_table, 0, sizeof(super_table));
4888 for (s = *super_list; s; s = s->next)
4889 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4890
4891 for (i = 0; i < tbl_size; i++) {
4892 struct imsm_disk *d;
4893 struct intel_disk *idisk;
4894 struct imsm_super *mpb = super_table[i]->anchor;
4895
4896 s = super_table[i];
4897 d = &s->disks->disk;
4898
4899 /* 'd' must appear in merged disk list for its
4900 * configuration to be valid
4901 */
4902 idisk = disk_list_get(d->serial, disk_list);
4903 if (idisk && idisk->owner == i)
4904 s = validate_members(s, disk_list, i);
4905 else
4906 s = NULL;
4907
4908 if (!s)
1ade5cc1
N		 4909 dprintf("marking family: %#x from %d:%d offline\n",
4910 mpb->family_num,
a2b97981
DW		 4911 super_table[i]->disks->major,
4912 super_table[i]->disks->minor);
4913 super_table[i] = s;
4914 }
4915
4916 /* This is where the mdadm implementation differs from the Windows
4917 * driver which has no strict concept of a container. We can only
4918 * assemble one family from a container, so when returning a prodigal
4919 * array member to this system the code will not be able to disambiguate
4920 * the container contents that should be assembled ("foreign" versus
4921 * "local"). It requires user intervention to set the orig_family_num
4922 * to a new value to establish a new container. The Windows driver in
4923 * this situation fixes up the volume name in place and manages the
4924 * foreign array as an independent entity.
4925 */
4926 s = NULL;
4927 spare = NULL;
4928 conflict = 0;
4929 for (i = 0; i < tbl_size; i++) {
4930 struct intel_super *tbl_ent = super_table[i];
4931 int is_spare = 0;
4932
4933 if (!tbl_ent)
4934 continue;
4935
4936 if (tbl_ent->anchor->num_raid_devs == 0) {
4937 spare = tbl_ent;
4938 is_spare = 1;
4939 }
4940
4941 if (s && !is_spare) {
4942 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4943 conflict++;
4944 } else if (!s && !is_spare)
4945 s = tbl_ent;
4946 }
4947
4948 if (!s)
4949 s = spare;
4950 if (!s) {
4951 champion = NULL;
4952 goto out;
4953 }
4954 champion = s;
4955
4956 if (conflict)
7a862a02 4957 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4958 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4959
4960 /* collect all dl's onto 'champion', and update them to
4961 * champion's version of the status
4962 */
4963 for (s = *super_list; s; s = s->next) {
4964 struct imsm_super *mpb = champion->anchor;
4965 struct dl *dl = s->disks;
4966
4967 if (s == champion)
4968 continue;
4969
5d7b407a
CA		 4970 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4971
a2b97981
DW		 4972 for (i = 0; i < mpb->num_disks; i++) {
4973 struct imsm_disk *disk;
4974
4975 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4976 if (disk) {
4977 dl->disk = *disk;
4978 /* only set index on disks that are a member of
4979 * a populated contianer, i.e. one with
4980 * raid_devs
4981 */
4982 if (is_failed(&dl->disk))
4983 dl->index = -2;
4984 else if (is_spare(&dl->disk))
4985 dl->index = -1;
4986 break;
4987 }
4988 }
4989
4990 if (i >= mpb->num_disks) {
4991 struct intel_disk *idisk;
4992
4993 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4994 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4995 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4996 dl->index = -1;
4997 else {
4998 dl->index = -2;
4999 continue;
5000 }
5001 }
5002
5003 dl->next = champion->disks;
5004 champion->disks = dl;
5005 s->disks = NULL;
5006 }
5007
5008 /* delete 'champion' from super_list */
5009 for (del = super_list; *del; ) {
5010 if (*del == champion) {
5011 *del = (*del)->next;
5012 break;
5013 } else
5014 del = &(*del)->next;
5015 }
5016 champion->next = NULL;
5017
5018 out:
5019 while (disk_list) {
5020 struct intel_disk *idisk = disk_list;
5021
5022 disk_list = disk_list->next;
5023 free(idisk);
5024 }
5025
5026 return champion;
5027}
5028
9587c373
LM		 5029static int
5030get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 5031static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 5032 int major, int minor, int keep_fd);
ec50f7b6
LM		 5033static int
5034get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5035 int *max, int keep_fd);
5036
cdddbdbc 5037static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 5038 char *devname, struct md_list *devlist,
5039 int keep_fd)
cdddbdbc 5040{
a2b97981
DW		 5041 struct intel_super *super_list = NULL;
5042 struct intel_super *super = NULL;
a2b97981 5043 int err = 0;
9587c373 5044 int i = 0;
dab4a513 5045
9587c373
LM		 5046 if (fd >= 0)
5047 /* 'fd' is an opened container */
5048 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
5049 else
ec50f7b6
LM		 5050 /* get super block from devlist devices */
5051 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5052 if (err)
1602d52c 5053 goto error;
a2b97981
DW		 5054 /* all mpbs enter, maybe one leaves */
5055 super = imsm_thunderdome(&super_list, i);
5056 if (!super) {
5057 err = 1;
5058 goto error;
cdddbdbc
DW		 5059 }
5060
47ee5a45
DW		 5061 if (find_missing(super) != 0) {
5062 free_imsm(super);
a2b97981
DW		 5063 err = 2;
5064 goto error;
47ee5a45 5065 }
8e59f3d8
AK		 5066
5067 /* load migration record */
5068 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5069 if (err == -1) {
5070 /* migration is in progress,
5071 * but migr_rec cannot be loaded,
5072 */
8e59f3d8
AK		 5073 err = 4;
5074 goto error;
5075 }
e2f41b2c
AK		 5076
5077 /* Check migration compatibility */
089f9d79 5078 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5079 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5080 if (devname)
5081 fprintf(stderr, " on %s\n", devname);
5082 else
5083 fprintf(stderr, " (IMSM).\n");
5084
5085 err = 5;
5086 goto error;
5087 }
5088
a2b97981
DW		 5089 err = 0;
5090
5091 error:
5092 while (super_list) {
5093 struct intel_super *s = super_list;
5094
5095 super_list = super_list->next;
5096 free_imsm(s);
5097 }
9587c373 5098
a2b97981
DW		 5099 if (err)
5100 return err;
f7e7067b 5101
cdddbdbc 5102 *sbp = super;
9587c373 5103 if (fd >= 0)
4dd2df09 5104 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5105 else
4dd2df09 5106 st->container_devnm[0] = 0;
a2b97981 5107 if (err == 0 && st->ss == NULL) {
bf5a934a 5108 st->ss = &super_imsm;
cdddbdbc
DW		 5109 st->minor_version = 0;
5110 st->max_devs = IMSM_MAX_DEVICES;
5111 }
cdddbdbc
DW		 5112 return 0;
5113}
2b959fbf 5114
ec50f7b6
LM		 5115static int
5116get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5117 int *max, int keep_fd)
5118{
5119 struct md_list *tmpdev;
5120 int err = 0;
5121 int i = 0;
9587c373 5122
ec50f7b6
LM		 5123 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5124 if (tmpdev->used != 1)
5125 continue;
5126 if (tmpdev->container == 1) {
ca9de185 5127 int lmax = 0;
ec50f7b6
LM		 5128 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5129 if (fd < 0) {
e7b84f9d 5130 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5131 tmpdev->devname, strerror(errno));
5132 err = 8;
5133 goto error;
5134 }
5135 err = get_sra_super_block(fd, super_list,
5136 tmpdev->devname, &lmax,
5137 keep_fd);
5138 i += lmax;
5139 close(fd);
5140 if (err) {
5141 err = 7;
5142 goto error;
5143 }
5144 } else {
5145 int major = major(tmpdev->st_rdev);
5146 int minor = minor(tmpdev->st_rdev);
5147 err = get_super_block(super_list,
4dd2df09 5148 NULL,
ec50f7b6
LM		 5149 tmpdev->devname,
5150 major, minor,
5151 keep_fd);
5152 i++;
5153 if (err) {
5154 err = 6;
5155 goto error;
5156 }
5157 }
5158 }
5159 error:
5160 *max = i;
5161 return err;
5162}
9587c373 5163
4dd2df09 5164static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5165 int major, int minor, int keep_fd)
5166{
594dc1b8 5167 struct intel_super *s;
9587c373
LM		 5168 char nm[32];
5169 int dfd = -1;
9587c373
LM		 5170 int err = 0;
5171 int retry;
5172
5173 s = alloc_super();
5174 if (!s) {
5175 err = 1;
5176 goto error;
5177 }
5178
5179 sprintf(nm, "%d:%d", major, minor);
5180 dfd = dev_open(nm, O_RDWR);
5181 if (dfd < 0) {
5182 err = 2;
5183 goto error;
5184 }
5185
fa7bb6f8 5186 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5187 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5188 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5189
5190 /* retry the load if we might have raced against mdmon */
4dd2df09 5191 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5192 for (retry = 0; retry < 3; retry++) {
5193 usleep(3000);
5194 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5195 if (err != 3)
5196 break;
5197 }
5198 error:
5199 if (!err) {
5200 s->next = *super_list;
5201 *super_list = s;
5202 } else {
5203 if (s)
8d67477f 5204 free_imsm(s);
36614e95 5205 if (dfd >= 0)
9587c373
LM		 5206 close(dfd);
5207 }
089f9d79 5208 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5209 close(dfd);
5210 return err;
5211
5212}
5213
5214static int
5215get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5216{
5217 struct mdinfo *sra;
4dd2df09 5218 char *devnm;
9587c373
LM		 5219 struct mdinfo *sd;
5220 int err = 0;
5221 int i = 0;
4dd2df09 5222 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5223 if (!sra)
5224 return 1;
5225
5226 if (sra->array.major_version != -1 ||
5227 sra->array.minor_version != -2 ||
5228 strcmp(sra->text_version, "imsm") != 0) {
5229 err = 1;
5230 goto error;
5231 }
5232 /* load all mpbs */
4dd2df09 5233 devnm = fd2devnm(fd);
9587c373 5234 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5235 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5236 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5237 err = 7;
5238 goto error;
5239 }
5240 }
5241 error:
5242 sysfs_free(sra);
5243 *max = i;
5244 return err;
5245}
5246
2b959fbf
N		 5247static int load_container_imsm(struct supertype *st, int fd, char *devname)
5248{
ec50f7b6 5249 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5250}
cdddbdbc
DW		 5251
5252static int load_super_imsm(struct supertype *st, int fd, char *devname)
5253{
5254 struct intel_super *super;
5255 int rv;
8a3544f8 5256 int retry;
cdddbdbc 5257
357ac106 5258 if (test_partition(fd))
691c6ee1
N		 5259 /* IMSM not allowed on partitions */
5260 return 1;
5261
37424f13
DW		 5262 free_super_imsm(st);
5263
49133e57 5264 super = alloc_super();
fa7bb6f8 5265 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5266 if (!super)
5267 return 1;
ea2bc72b
LM		 5268 /* Load hba and capabilities if they exist.
5269 * But do not preclude loading metadata in case capabilities or hba are
5270 * non-compliant and ignore_hw_compat is set.
5271 */
d424212e 5272 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5273 /* no orom/efi or non-intel hba of the disk */
089f9d79 5274 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5275 if (devname)
e7b84f9d 5276 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5277 free_imsm(super);
5278 return 2;
5279 }
a2b97981 5280 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5281
8a3544f8
AP		 5282 /* retry the load if we might have raced against mdmon */
5283 if (rv == 3) {
f96b1302
AP		 5284 struct mdstat_ent *mdstat = NULL;
5285 char *name = fd2kname(fd);
5286
5287 if (name)
5288 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5289
5290 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5291 for (retry = 0; retry < 3; retry++) {
5292 usleep(3000);
5293 rv = load_and_parse_mpb(fd, super, devname, 0);
5294 if (rv != 3)
5295 break;
5296 }
5297 }
5298
5299 free_mdstat(mdstat);
5300 }
5301
cdddbdbc
DW		 5302 if (rv) {
5303 if (devname)
7a862a02 5304 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5305 free_imsm(super);
5306 return rv;
5307 }
5308
5309 st->sb = super;
5310 if (st->ss == NULL) {
5311 st->ss = &super_imsm;
5312 st->minor_version = 0;
5313 st->max_devs = IMSM_MAX_DEVICES;
5314 }
8e59f3d8
AK		 5315
5316 /* load migration record */
2e062e82
AK		 5317 if (load_imsm_migr_rec(super, NULL) == 0) {
5318 /* Check for unsupported migration features */
5319 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5320 pr_err("Unsupported migration detected");
2e062e82
AK		 5321 if (devname)
5322 fprintf(stderr, " on %s\n", devname);
5323 else
5324 fprintf(stderr, " (IMSM).\n");
5325 return 3;
5326 }
e2f41b2c
AK		 5327 }
5328
cdddbdbc
DW		 5329 return 0;
5330}
5331
ef6ffade
DW		 5332static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5333{
5334 if (info->level == 1)
5335 return 128;
5336 return info->chunk_size >> 9;
5337}
5338
5551b113
CA		 5339static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5340 unsigned long long size)
fcfd9599 5341{
4025c288 5342 if (info->level == 1)
5551b113 5343 return size * 2;
4025c288 5344 else
5551b113 5345 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5346}
5347
4d1313e9
DW		 5348static void imsm_update_version_info(struct intel_super *super)
5349{
5350 /* update the version and attributes */
5351 struct imsm_super *mpb = super->anchor;
5352 char *version;
5353 struct imsm_dev *dev;
5354 struct imsm_map *map;
5355 int i;
5356
5357 for (i = 0; i < mpb->num_raid_devs; i++) {
5358 dev = get_imsm_dev(super, i);
238c0a71 5359 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5360 if (__le32_to_cpu(dev->size_high) > 0)
5361 mpb->attributes |= MPB_ATTRIB_2TB;
5362
5363 /* FIXME detect when an array spans a port multiplier */
5364 #if 0
5365 mpb->attributes |= MPB_ATTRIB_PM;
5366 #endif
5367
5368 if (mpb->num_raid_devs > 1 ||
5369 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5370 version = MPB_VERSION_ATTRIBS;
5371 switch (get_imsm_raid_level(map)) {
5372 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5373 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5374 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5375 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5376 }
5377 } else {
5378 if (map->num_members >= 5)
5379 version = MPB_VERSION_5OR6_DISK_ARRAY;
5380 else if (dev->status == DEV_CLONE_N_GO)
5381 version = MPB_VERSION_CNG;
5382 else if (get_imsm_raid_level(map) == 5)
5383 version = MPB_VERSION_RAID5;
5384 else if (map->num_members >= 3)
5385 version = MPB_VERSION_3OR4_DISK_ARRAY;
5386 else if (get_imsm_raid_level(map) == 1)
5387 version = MPB_VERSION_RAID1;
5388 else
5389 version = MPB_VERSION_RAID0;
5390 }
5391 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5392 }
5393}
5394
aa534678
DW		 5395static int check_name(struct intel_super *super, char *name, int quiet)
5396{
5397 struct imsm_super *mpb = super->anchor;
5398 char *reason = NULL;
9bd99a90
RS		 5399 char *start = name;
5400 size_t len = strlen(name);
aa534678
DW		 5401 int i;
5402
9bd99a90
RS		 5403 if (len > 0) {
5404 while (isspace(start[len - 1]))
5405 start[--len] = 0;
5406 while (*start && isspace(*start))
5407 ++start, --len;
5408 memmove(name, start, len + 1);
5409 }
5410
5411 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5412 reason = "must be 16 characters or less";
9bd99a90
RS		 5413 else if (len == 0)
5414 reason = "must be a non-empty string";
aa534678
DW		 5415
5416 for (i = 0; i < mpb->num_raid_devs; i++) {
5417 struct imsm_dev *dev = get_imsm_dev(super, i);
5418
5419 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5420 reason = "already exists";
5421 break;
5422 }
5423 }
5424
5425 if (reason && !quiet)
e7b84f9d 5426 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5427
5428 return !reason;
5429}
5430
8b353278 5431static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5432 struct shape *s, char *name,
83cd1e97
N		 5433 char *homehost, int *uuid,
5434 long long data_offset)
cdddbdbc 5435{
c2c087e6
DW		 5436 /* We are creating a volume inside a pre-existing container.
5437 * so st->sb is already set.
5438 */
5439 struct intel_super *super = st->sb;
f36a9ecd 5440 unsigned int sector_size = super->sector_size;
949c47a0 5441 struct imsm_super *mpb = super->anchor;
ba2de7ba 5442 struct intel_dev *dv;
c2c087e6
DW		 5443 struct imsm_dev *dev;
5444 struct imsm_vol *vol;
5445 struct imsm_map *map;
5446 int idx = mpb->num_raid_devs;
5447 int i;
760365f9 5448 int namelen;
c2c087e6 5449 unsigned long long array_blocks;
2c092cad 5450 size_t size_old, size_new;
5551b113 5451 unsigned long long num_data_stripes;
b53bfba6
TM		 5452 unsigned int data_disks;
5453 unsigned long long size_per_member;
cdddbdbc 5454
88c32bb1 5455 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5456 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5457 return 0;
5458 }
5459
2c092cad
DW		 5460 /* ensure the mpb is large enough for the new data */
5461 size_old = __le32_to_cpu(mpb->mpb_size);
5462 size_new = disks_to_mpb_size(info->nr_disks);
5463 if (size_new > size_old) {
5464 void *mpb_new;
f36a9ecd 5465 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5466
f36a9ecd 5467 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5468 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5469 return 0;
5470 }
85337573
AO		 5471 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5472 MIGR_REC_BUF_SECTORS*
5473 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5474 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5475 free(super->buf);
5476 free(super);
ea944c8f 5477 free(mpb_new);
8e59f3d8
AK		 5478 return 0;
5479 }
2c092cad
DW		 5480 memcpy(mpb_new, mpb, size_old);
5481 free(mpb);
5482 mpb = mpb_new;
949c47a0 5483 super->anchor = mpb_new;
2c092cad
DW		 5484 mpb->mpb_size = __cpu_to_le32(size_new);
5485 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5486 super->len = size_round;
2c092cad 5487 }
bf5a934a 5488 super->current_vol = idx;
3960e579
DW		 5489
5490 /* handle 'failed_disks' by either:
5491 * a) create dummy disk entries in the table if this the first
5492 * volume in the array. We add them here as this is the only
5493 * opportunity to add them. add_to_super_imsm_volume()
5494 * handles the non-failed disks and continues incrementing
5495 * mpb->num_disks.
5496 * b) validate that 'failed_disks' matches the current number
5497 * of missing disks if the container is populated
d23fe947 5498 */
3960e579 5499 if (super->current_vol == 0) {
d23fe947 5500 mpb->num_disks = 0;
3960e579
DW		 5501 for (i = 0; i < info->failed_disks; i++) {
5502 struct imsm_disk *disk;
5503
5504 mpb->num_disks++;
5505 disk = __get_imsm_disk(mpb, i);
5506 disk->status = CONFIGURED_DISK | FAILED_DISK;
5507 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5508 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5509 "missing:%d", (__u8)i);
3960e579
DW		 5510 }
5511 find_missing(super);
5512 } else {
5513 int missing = 0;
5514 struct dl *d;
5515
5516 for (d = super->missing; d; d = d->next)
5517 missing++;
5518 if (info->failed_disks > missing) {
e7b84f9d 5519 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5520 return 0;
5521 }
5522 }
5a038140 5523
aa534678
DW		 5524 if (!check_name(super, name, 0))
5525 return 0;
503975b9
N		 5526 dv = xmalloc(sizeof(*dv));
5527 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5528 /*
5529 * Explicitly allow truncating to not confuse gcc's
5530 * -Werror=stringop-truncation
5531 */
5532 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5533 memcpy(dev->volume, name, namelen);
e03640bd 5534 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5535 info->layout, info->chunk_size,
b53bfba6
TM		 5536 s->size * BLOCKS_PER_KB);
5537 data_disks = get_data_disks(info->level, info->layout,
5538 info->raid_disks);
5539 array_blocks = round_size_to_mb(array_blocks, data_disks);
5540 size_per_member = array_blocks / data_disks;
979d38be 5541
fcc2c9da 5542 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5543 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5544 vol = &dev->vol;
5545 vol->migr_state = 0;
1484e727 5546 set_migr_type(dev, MIGR_INIT);
3960e579 5547 vol->dirty = !info->state;
f8f603f1 5548 vol->curr_migr_unit = 0;
238c0a71 5549 map = get_imsm_map(dev, MAP_0);
5551b113 5550 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5551 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5552 map->failed_disk_num = ~0;
bf4442ab 5553 if (info->level > 0)
fffaf1ff
N		 5554 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5555 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5556 else
5557 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5558 IMSM_T_STATE_NORMAL;
252d23c0 5559 map->ddf = 1;
ef6ffade
DW		 5560
5561 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5562 free(dev);
5563 free(dv);
7a862a02 5564 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5565 return 0;
5566 }
81062a36
DW		 5567
5568 map->raid_level = info->level;
4d1313e9 5569 if (info->level == 10) {
c2c087e6 5570 map->raid_level = 1;
4d1313e9 5571 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5572 } else if (info->level == 1)
5573 map->num_domains = info->raid_disks;
5574 else
ff596308 5575 map->num_domains = 1;
81062a36 5576
5551b113 5577 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5578 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5579 num_data_stripes /= map->num_domains;
5580 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5581
44490938
MD		 5582 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5583 set_blocks_per_member(map, info_to_blocks_per_member(info,
5584 size_per_member /
5585 BLOCKS_PER_KB));
5586
c2c087e6
DW		 5587 map->num_members = info->raid_disks;
5588 for (i = 0; i < map->num_members; i++) {
5589 /* initialized in add_to_super */
4eb26970 5590 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5591 }
949c47a0 5592 mpb->num_raid_devs++;
2a24dc1b
PB		 5593 mpb->num_raid_devs_created++;
5594 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5595
b7580566 5596 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5597 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5598 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5599 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5600 } else {
5601 free(dev);
5602 free(dv);
5603 pr_err("imsm does not support consistency policy %s\n",
5604 map_num(consistency_policies, s->consistency_policy));
5605 return 0;
5606 }
5607
ba2de7ba
DW		 5608 dv->dev = dev;
5609 dv->index = super->current_vol;
5610 dv->next = super->devlist;
5611 super->devlist = dv;
c2c087e6 5612
4d1313e9
DW		 5613 imsm_update_version_info(super);
5614
c2c087e6 5615 return 1;
cdddbdbc
DW		 5616}
5617
bf5a934a 5618static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5619 struct shape *s, char *name,
83cd1e97
N		 5620 char *homehost, int *uuid,
5621 unsigned long long data_offset)
bf5a934a
DW		 5622{
5623 /* This is primarily called by Create when creating a new array.
5624 * We will then get add_to_super called for each component, and then
5625 * write_init_super called to write it out to each device.
5626 * For IMSM, Create can create on fresh devices or on a pre-existing
5627 * array.
5628 * To create on a pre-existing array a different method will be called.
5629 * This one is just for fresh drives.
5630 */
5631 struct intel_super *super;
5632 struct imsm_super *mpb;
5633 size_t mpb_size;
4d1313e9 5634 char *version;
bf5a934a 5635
83cd1e97 5636 if (data_offset != INVALID_SECTORS) {
ed503f89 5637 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5638 return 0;
5639 }
5640
bf5a934a 5641 if (st->sb)
5308f117 5642 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5643 data_offset);
e683ca88
DW		 5644
5645 if (info)
5646 mpb_size = disks_to_mpb_size(info->nr_disks);
5647 else
f36a9ecd 5648 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5649
49133e57 5650 super = alloc_super();
f36a9ecd
PB		 5651 if (super &&
5652 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5653 free_imsm(super);
e683ca88
DW		 5654 super = NULL;
5655 }
5656 if (!super) {
1ade5cc1 5657 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5658 return 0;
5659 }
de44e46f
PB		 5660 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5661 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5662 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5663 free(super->buf);
8d67477f 5664 free_imsm(super);
8e59f3d8
AK		 5665 return 0;
5666 }
e683ca88 5667 memset(super->buf, 0, mpb_size);
ef649044 5668 mpb = super->buf;
e683ca88
DW		 5669 mpb->mpb_size = __cpu_to_le32(mpb_size);
5670 st->sb = super;
5671
5672 if (info == NULL) {
5673 /* zeroing superblock */
5674 return 0;
5675 }
bf5a934a 5676
4d1313e9
DW		 5677 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5678
5679 version = (char *) mpb->sig;
5680 strcpy(version, MPB_SIGNATURE);
5681 version += strlen(MPB_SIGNATURE);
5682 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5683
bf5a934a
DW		 5684 return 1;
5685}
5686
f2cc4f7d
AO		 5687static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5688{
5689 unsigned int member_sector_size;
5690
5691 if (dl->fd < 0) {
5692 pr_err("Invalid file descriptor for %s\n", dl->devname);
5693 return 0;
5694 }
5695
5696 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5697 return 0;
5698 if (member_sector_size != super->sector_size)
5699 return 0;
5700 return 1;
5701}
5702
f20c3968 5703static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5704 int fd, char *devname)
5705{
5706 struct intel_super *super = st->sb;
d23fe947 5707 struct imsm_super *mpb = super->anchor;
3960e579 5708 struct imsm_disk *_disk;
bf5a934a
DW		 5709 struct imsm_dev *dev;
5710 struct imsm_map *map;
3960e579 5711 struct dl *dl, *df;
4eb26970 5712 int slot;
bf5a934a 5713
949c47a0 5714 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5715 map = get_imsm_map(dev, MAP_0);
bf5a934a 5716
208933a7 5717 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5718 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5719 devname);
5720 return 1;
5721 }
5722
efb30e7f
DW		 5723 if (fd == -1) {
5724 /* we're doing autolayout so grab the pre-marked (in
5725 * validate_geometry) raid_disk
5726 */
5727 for (dl = super->disks; dl; dl = dl->next)
5728 if (dl->raiddisk == dk->raid_disk)
5729 break;
5730 } else {
5731 for (dl = super->disks; dl ; dl = dl->next)
5732 if (dl->major == dk->major &&
5733 dl->minor == dk->minor)
5734 break;
5735 }
d23fe947 5736
208933a7 5737 if (!dl) {
e7b84f9d 5738 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5739 return 1;
208933a7 5740 }
bf5a934a 5741
59632db9
MZ		 5742 if (mpb->num_disks == 0)
5743 if (!get_dev_sector_size(dl->fd, dl->devname,
5744 &super->sector_size))
5745 return 1;
5746
f2cc4f7d
AO		 5747 if (!drive_validate_sector_size(super, dl)) {
5748 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5749 return 1;
5750 }
5751
d23fe947
DW		 5752 /* add a pristine spare to the metadata */
5753 if (dl->index < 0) {
5754 dl->index = super->anchor->num_disks;
5755 super->anchor->num_disks++;
5756 }
4eb26970
DW		 5757 /* Check the device has not already been added */
5758 slot = get_imsm_disk_slot(map, dl->index);
5759 if (slot >= 0 &&
238c0a71 5760 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5761 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5762 devname);
5763 return 1;
5764 }
656b6b5a 5765 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5766 dl->disk.status = CONFIGURED_DISK;
d23fe947 5767
3960e579
DW		 5768 /* update size of 'missing' disks to be at least as large as the
5769 * largest acitve member (we only have dummy missing disks when
5770 * creating the first volume)
5771 */
5772 if (super->current_vol == 0) {
5773 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5774 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5775 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5776 _disk = __get_imsm_disk(mpb, df->index);
5777 *_disk = df->disk;
5778 }
5779 }
5780
5781 /* refresh unset/failed slots to point to valid 'missing' entries */
5782 for (df = super->missing; df; df = df->next)
5783 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5784 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5785
5786 if ((ord & IMSM_ORD_REBUILD) == 0)
5787 continue;
5788 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5789 if (is_gen_migration(dev)) {
238c0a71
AK		 5790 struct imsm_map *map2 = get_imsm_map(dev,
5791 MAP_1);
0a108d63 5792 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5793 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5794 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5795 slot2,
5796 MAP_1);
1ace8403
AK		 5797 if ((unsigned)df->index ==
5798 ord_to_idx(ord2))
5799 set_imsm_ord_tbl_ent(map2,
0a108d63 5800 slot2,
1ace8403
AK		 5801 df->index |
5802 IMSM_ORD_REBUILD);
5803 }
5804 }
3960e579
DW		 5805 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5806 break;
5807 }
5808
d23fe947
DW		 5809 /* if we are creating the first raid device update the family number */
5810 if (super->current_vol == 0) {
5811 __u32 sum;
5812 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5813
3960e579 5814 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5815 if (!_dev || !_disk) {
e7b84f9d 5816 pr_err("BUG mpb setup error\n");
791b666a
AW		 5817 return 1;
5818 }
d23fe947
DW		 5819 *_dev = *dev;
5820 *_disk = dl->disk;
148acb7b
DW		 5821 sum = random32();
5822 sum += __gen_imsm_checksum(mpb);
d23fe947 5823 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5824 mpb->orig_family_num = mpb->family_num;
d23fe947 5825 }
ca0748fa 5826 super->current_disk = dl;
f20c3968 5827 return 0;
bf5a934a
DW		 5828}
5829
a8619d23
AK		 5830/* mark_spare()
5831 * Function marks disk as spare and restores disk serial
5832 * in case it was previously marked as failed by takeover operation
5833 * reruns:
5834 * -1 : critical error
5835 * 0 : disk is marked as spare but serial is not set
5836 * 1 : success
5837 */
5838int mark_spare(struct dl *disk)
5839{
5840 __u8 serial[MAX_RAID_SERIAL_LEN];
5841 int ret_val = -1;
5842
5843 if (!disk)
5844 return ret_val;
5845
5846 ret_val = 0;
5847 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5848 /* Restore disk serial number, because takeover marks disk
5849 * as failed and adds to serial ':0' before it becomes
5850 * a spare disk.
5851 */
5852 serialcpy(disk->serial, serial);
5853 serialcpy(disk->disk.serial, serial);
5854 ret_val = 1;
5855 }
5856 disk->disk.status = SPARE_DISK;
5857 disk->index = -1;
5858
5859 return ret_val;
5860}
88654014 5861
f20c3968 5862static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5863 int fd, char *devname,
5864 unsigned long long data_offset)
cdddbdbc 5865{
c2c087e6 5866 struct intel_super *super = st->sb;
c2c087e6
DW		 5867 struct dl *dd;
5868 unsigned long long size;
fa7bb6f8 5869 unsigned int member_sector_size;
f2f27e63 5870 __u32 id;
c2c087e6
DW		 5871 int rv;
5872 struct stat stb;
5873
88654014
LM		 5874 /* If we are on an RAID enabled platform check that the disk is
5875 * attached to the raid controller.
5876 * We do not need to test disks attachment for container based additions,
5877 * they shall be already tested when container was created/assembled.
88c32bb1 5878 */
d424212e 5879 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5880 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5881 if (rv != 0) {
5882 dprintf("capability: %p fd: %d ret: %d\n",
5883 super->orom, fd, rv);
5884 return 1;
88c32bb1
DW		 5885 }
5886
f20c3968
DW		 5887 if (super->current_vol >= 0)
5888 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5889
c2c087e6 5890 fstat(fd, &stb);
503975b9 5891 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5892 dd->major = major(stb.st_rdev);
5893 dd->minor = minor(stb.st_rdev);
503975b9 5894 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5895 dd->fd = fd;
689c9bf3 5896 dd->e = NULL;
1a64be56 5897 dd->action = DISK_ADD;
c2c087e6 5898 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5899 if (rv) {
e7b84f9d 5900 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5901 if (dd->devname)
5902 free(dd->devname);
949c47a0 5903 free(dd);
0030e8d6 5904 abort();
c2c087e6 5905 }
20bee0f8
PB		 5906 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5907 (super->hba->type == SYS_DEV_VMD))) {
5908 int i;
5909 char *devpath = diskfd_to_devpath(fd);
5910 char controller_path[PATH_MAX];
5911
5912 if (!devpath) {
5913 pr_err("failed to get devpath, aborting\n");
5914 if (dd->devname)
5915 free(dd->devname);
5916 free(dd);
5917 return 1;
5918 }
5919
5920 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5921 free(devpath);
5922
5923 if (devpath_to_vendor(controller_path) == 0x8086) {
5924 /*
5925 * If Intel's NVMe drive has serial ended with
5926 * "-A","-B","-1" or "-2" it means that this is "x8"
5927 * device (double drive on single PCIe card).
5928 * User should be warned about potential data loss.
5929 */
5930 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5931 /* Skip empty character at the end */
5932 if (dd->serial[i] == 0)
5933 continue;
5934
5935 if (((dd->serial[i] == 'A') ||
5936 (dd->serial[i] == 'B') ||
5937 (dd->serial[i] == '1') ||
5938 (dd->serial[i] == '2')) &&
5939 (dd->serial[i-1] == '-'))
5940 pr_err("\tThe action you are about to take may put your data at risk.\n"
5941 "\tPlease note that x8 devices may consist of two separate x4 devices "
5942 "located on a single PCIe port.\n"
5943 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5944 break;
5945 }
32716c51
PB		 5946 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5947 !imsm_orom_has_tpv_support(super->orom)) {
5948 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5949 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5950 free(dd->devname);
5951 free(dd);
5952 return 1;
20bee0f8
PB		 5953 }
5954 }
c2c087e6 5955
c2c087e6 5956 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5957 get_dev_sector_size(fd, NULL, &member_sector_size);
5958
5959 if (super->sector_size == 0) {
5960 /* this a first device, so sector_size is not set yet */
5961 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5962 }
5963
71e5411e 5964 /* clear migr_rec when adding disk to container */
85337573
AO		 5965 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5966 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5967 SEEK_SET) >= 0) {
466070ad 5968 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5969 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5970 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5971 perror("Write migr_rec failed");
5972 }
5973
c2c087e6 5974 size /= 512;
1f24f035 5975 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5976 set_total_blocks(&dd->disk, size);
5977 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5978 struct imsm_super *mpb = super->anchor;
5979 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5980 }
a8619d23 5981 mark_spare(dd);
c2c087e6 5982 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5983 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5984 else
b9f594fe 5985 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5986
5987 if (st->update_tail) {
1a64be56
LM		 5988 dd->next = super->disk_mgmt_list;
5989 super->disk_mgmt_list = dd;
43dad3d6
DW		 5990 } else {
5991 dd->next = super->disks;
5992 super->disks = dd;
ceaf0ee1 5993 super->updates_pending++;
43dad3d6 5994 }
f20c3968
DW		 5995
5996 return 0;
cdddbdbc
DW		 5997}
5998
1a64be56
LM		 5999static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
6000{
6001 struct intel_super *super = st->sb;
6002 struct dl *dd;
6003
6004 /* remove from super works only in mdmon - for communication
6005 * manager - monitor. Check if communication memory buffer
6006 * is prepared.
6007 */
6008 if (!st->update_tail) {
1ade5cc1 6009 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 6010 return 1;
6011 }
503975b9 6012 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 6013 dd->major = dk->major;
6014 dd->minor = dk->minor;
1a64be56 6015 dd->fd = -1;
a8619d23 6016 mark_spare(dd);
1a64be56
LM		 6017 dd->action = DISK_REMOVE;
6018
6019 dd->next = super->disk_mgmt_list;
6020 super->disk_mgmt_list = dd;
6021
1a64be56
LM		 6022 return 0;
6023}
6024
f796af5d
DW		 6025static int store_imsm_mpb(int fd, struct imsm_super *mpb);
6026
6027static union {
f36a9ecd 6028 char buf[MAX_SECTOR_SIZE];
f796af5d 6029 struct imsm_super anchor;
f36a9ecd 6030} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 6031
d23fe947
DW		 6032/* spare records have their own family number and do not have any defined raid
6033 * devices
6034 */
6035static int write_super_imsm_spares(struct intel_super *super, int doclose)
6036{
d23fe947 6037 struct imsm_super *mpb = super->anchor;
f796af5d 6038 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 6039 __u32 sum;
6040 struct dl *d;
6041
68641cdb
JS		 6042 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6043 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6044 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6045 spare->num_disks = 1;
6046 spare->num_raid_devs = 0;
6047 spare->cache_size = mpb->cache_size;
6048 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6049
6050 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6051 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 6052
6053 for (d = super->disks; d; d = d->next) {
8796fdc4 6054 if (d->index != -1)
d23fe947
DW		 6055 continue;
6056
f796af5d 6057 spare->disk[0] = d->disk;
027c374f
CA		 6058 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6059 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6060
f36a9ecd
PB		 6061 if (super->sector_size == 4096)
6062 convert_to_4k_imsm_disk(&spare->disk[0]);
6063
f796af5d
DW		 6064 sum = __gen_imsm_checksum(spare);
6065 spare->family_num = __cpu_to_le32(sum);
6066 spare->orig_family_num = 0;
6067 sum = __gen_imsm_checksum(spare);
6068 spare->check_sum = __cpu_to_le32(sum);
d23fe947 6069
f796af5d 6070 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 6071 pr_err("failed for device %d:%d %s\n",
6072 d->major, d->minor, strerror(errno));
e74255d9 6073 return 1;
d23fe947
DW		 6074 }
6075 if (doclose) {
6076 close(d->fd);
6077 d->fd = -1;
6078 }
6079 }
6080
e74255d9 6081 return 0;
d23fe947
DW		 6082}
6083
36988a3d 6084static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6085{
36988a3d 6086 struct intel_super *super = st->sb;
f36a9ecd 6087 unsigned int sector_size = super->sector_size;
949c47a0 6088 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6089 struct dl *d;
6090 __u32 generation;
6091 __u32 sum;
d23fe947 6092 int spares = 0;
949c47a0 6093 int i;
a48ac0a8 6094 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6095 int num_disks = 0;
146c6260 6096 int clear_migration_record = 1;
bbab0940 6097 __u32 bbm_log_size;
cdddbdbc 6098
c2c087e6
DW		 6099 /* 'generation' is incremented everytime the metadata is written */
6100 generation = __le32_to_cpu(mpb->generation_num);
6101 generation++;
6102 mpb->generation_num = __cpu_to_le32(generation);
6103
148acb7b
DW		 6104 /* fix up cases where previous mdadm releases failed to set
6105 * orig_family_num
6106 */
6107 if (mpb->orig_family_num == 0)
6108 mpb->orig_family_num = mpb->family_num;
6109
d23fe947 6110 for (d = super->disks; d; d = d->next) {
8796fdc4 6111 if (d->index == -1)
d23fe947 6112 spares++;
36988a3d 6113 else {
d23fe947 6114 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6115 num_disks++;
6116 }
d23fe947 6117 }
36988a3d 6118 for (d = super->missing; d; d = d->next) {
47ee5a45 6119 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6120 num_disks++;
6121 }
6122 mpb->num_disks = num_disks;
6123 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6124
949c47a0
DW		 6125 for (i = 0; i < mpb->num_raid_devs; i++) {
6126 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6127 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6128 if (dev && dev2) {
6129 imsm_copy_dev(dev, dev2);
6130 mpb_size += sizeof_imsm_dev(dev, 0);
6131 }
146c6260
AK		 6132 if (is_gen_migration(dev2))
6133 clear_migration_record = 0;
949c47a0 6134 }
bbab0940
TM		 6135
6136 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6137
6138 if (bbm_log_size) {
6139 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6140 mpb->attributes |= MPB_ATTRIB_BBM;
6141 } else
6142 mpb->attributes &= ~MPB_ATTRIB_BBM;
6143
6144 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6145 mpb_size += bbm_log_size;
a48ac0a8 6146 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6147
bbab0940
TM		 6148#ifdef DEBUG
6149 assert(super->len == 0 || mpb_size <= super->len);
6150#endif
6151
c2c087e6 6152 /* recalculate checksum */
949c47a0 6153 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6154 mpb->check_sum = __cpu_to_le32(sum);
6155
51d83f5d
AK		 6156 if (super->clean_migration_record_by_mdmon) {
6157 clear_migration_record = 1;
6158 super->clean_migration_record_by_mdmon = 0;
6159 }
146c6260 6160 if (clear_migration_record)
de44e46f 6161 memset(super->migr_rec_buf, 0,
85337573 6162 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6163
f36a9ecd
PB		 6164 if (sector_size == 4096)
6165 convert_to_4k(super);
6166
d23fe947 6167 /* write the mpb for disks that compose raid devices */
c2c087e6 6168 for (d = super->disks; d ; d = d->next) {
86c54047 6169 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6170 continue;
30602f53 6171
146c6260
AK		 6172 if (clear_migration_record) {
6173 unsigned long long dsize;
6174
6175 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6176 if (lseek64(d->fd, dsize - sector_size,
6177 SEEK_SET) >= 0) {
466070ad
PB		 6178 if ((unsigned int)write(d->fd,
6179 super->migr_rec_buf,
de44e46f
PB		 6180 MIGR_REC_BUF_SECTORS*sector_size) !=
6181 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6182 perror("Write migr_rec failed");
146c6260
AK		 6183 }
6184 }
51d83f5d
AK		 6185
6186 if (store_imsm_mpb(d->fd, mpb))
6187 fprintf(stderr,
1ade5cc1
N		 6188 "failed for device %d:%d (fd: %d)%s\n",
6189 d->major, d->minor,
51d83f5d
AK		 6190 d->fd, strerror(errno));
6191
c2c087e6
DW		 6192 if (doclose) {
6193 close(d->fd);
6194 d->fd = -1;
6195 }
6196 }
6197
d23fe947
DW		 6198 if (spares)
6199 return write_super_imsm_spares(super, doclose);
6200
e74255d9 6201 return 0;
c2c087e6
DW		 6202}
6203
9b1fb677 6204static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6205{
6206 size_t len;
6207 struct imsm_update_create_array *u;
6208 struct intel_super *super = st->sb;
9b1fb677 6209 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6210 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6211 struct disk_info *inf;
6212 struct imsm_disk *disk;
6213 int i;
43dad3d6 6214
54c2c1ea
DW		 6215 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6216 sizeof(*inf) * map->num_members;
503975b9 6217 u = xmalloc(len);
43dad3d6 6218 u->type = update_create_array;
9b1fb677 6219 u->dev_idx = dev_idx;
43dad3d6 6220 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6221 inf = get_disk_info(u);
6222 for (i = 0; i < map->num_members; i++) {
238c0a71 6223 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6224
54c2c1ea 6225 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6226 if (!disk)
6227 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6228 serialcpy(inf[i].serial, disk->serial);
6229 }
43dad3d6
DW		 6230 append_metadata_update(st, u, len);
6231
6232 return 0;
6233}
6234
1a64be56 6235static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6236{
6237 struct intel_super *super = st->sb;
6238 size_t len;
1a64be56 6239 struct imsm_update_add_remove_disk *u;
43dad3d6 6240
1a64be56 6241 if (!super->disk_mgmt_list)
43dad3d6
DW		 6242 return 0;
6243
6244 len = sizeof(*u);
503975b9 6245 u = xmalloc(len);
1a64be56 6246 u->type = update_add_remove_disk;
43dad3d6
DW		 6247 append_metadata_update(st, u, len);
6248
6249 return 0;
6250}
2432ce9b
AP		 6251
6252__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6253
e397cefe
AP		 6254static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6255{
6256 struct ppl_header *ppl_hdr = buf;
6257 int ret;
6258
6259 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6260
6261 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6262 ret = -errno;
6263 perror("Failed to seek to PPL header location");
6264 return ret;
6265 }
6266
6267 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6268 ret = -errno;
6269 perror("Write PPL header failed");
6270 return ret;
6271 }
6272
6273 fsync(fd);
6274
6275 return 0;
6276}
6277
2432ce9b
AP		 6278static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6279{
6280 struct intel_super *super = st->sb;
6281 void *buf;
6282 struct ppl_header *ppl_hdr;
6283 int ret;
6284
b2514242
PB		 6285 /* first clear entire ppl space */
6286 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6287 if (ret)
6288 return ret;
6289
6290 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6291 if (ret) {
6292 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6293 return -ret;
2432ce9b
AP		 6294 }
6295
6296 memset(buf, 0, PPL_HEADER_SIZE);
6297 ppl_hdr = buf;
6298 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6299 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6300
6301 if (info->mismatch_cnt) {
6302 /*
6303 * We are overwriting an invalid ppl. Make one entry with wrong
6304 * checksum to prevent the kernel from skipping resync.
6305 */
6306 ppl_hdr->entries_count = __cpu_to_le32(1);
6307 ppl_hdr->entries[0].checksum = ~0;
6308 }
6309
e397cefe 6310 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6311
6312 free(buf);
6313 return ret;
6314}
6315
e397cefe
AP		 6316static int is_rebuilding(struct imsm_dev *dev);
6317
2432ce9b
AP		 6318static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6319 struct mdinfo *disk)
6320{
6321 struct intel_super *super = st->sb;
6322 struct dl *d;
e397cefe 6323 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6324 int ret = 0;
e397cefe 6325 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6326 __u32 crc;
6327 struct imsm_dev *dev;
2432ce9b 6328 __u32 idx;
44b6b876
PB		 6329 unsigned int i;
6330 unsigned long long ppl_offset = 0;
6331 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6332
6333 if (disk->disk.raid_disk < 0)
6334 return 0;
6335
2432ce9b 6336 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6337 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6338 d = get_imsm_dl_disk(super, idx);
6339
6340 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6341 return 0;
6342
6343 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6344 pr_err("Failed to allocate PPL header buffer\n");
6345 return -1;
6346 }
6347 buf = buf_orig;
2432ce9b 6348
44b6b876
PB		 6349 ret = 1;
6350 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6351 void *tmp;
6352
44b6b876 6353 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6354
44b6b876
PB		 6355 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6356 SEEK_SET) < 0) {
6357 perror("Failed to seek to PPL header location");
6358 ret = -1;
e397cefe 6359 break;
44b6b876 6360 }
2432ce9b 6361
44b6b876
PB		 6362 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6363 perror("Read PPL header failed");
6364 ret = -1;
e397cefe 6365 break;
44b6b876 6366 }
2432ce9b 6367
44b6b876 6368 ppl_hdr = buf;
2432ce9b 6369
44b6b876
PB		 6370 crc = __le32_to_cpu(ppl_hdr->checksum);
6371 ppl_hdr->checksum = 0;
6372
6373 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6374 dprintf("Wrong PPL header checksum on %s\n",
6375 d->devname);
e397cefe 6376 break;
44b6b876
PB		 6377 }
6378
6379 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6380 /* previous was newest, it was already checked */
e397cefe 6381 break;
44b6b876
PB		 6382 }
6383
6384 if ((__le32_to_cpu(ppl_hdr->signature) !=
6385 super->anchor->orig_family_num)) {
6386 dprintf("Wrong PPL header signature on %s\n",
6387 d->devname);
6388 ret = 1;
e397cefe 6389 break;
44b6b876
PB		 6390 }
6391
6392 ret = 0;
6393 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6394
44b6b876
PB		 6395 ppl_offset += PPL_HEADER_SIZE;
6396 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6397 ppl_offset +=
6398 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6399
6400 if (!buf_prev)
6401 buf_prev = buf + PPL_HEADER_SIZE;
6402 tmp = buf_prev;
6403 buf_prev = buf;
6404 buf = tmp;
2432ce9b
AP		 6405 }
6406
e397cefe
AP		 6407 if (buf_prev) {
6408 buf = buf_prev;
6409 ppl_hdr = buf_prev;
6410 }
2432ce9b 6411
54148aba
PB		 6412 /*
6413 * Update metadata to use mutliple PPLs area (1MB).
6414 * This is done once for all RAID members
6415 */
6416 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6417 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6418 char subarray[20];
6419 struct mdinfo *member_dev;
6420
6421 sprintf(subarray, "%d", info->container_member);
6422
6423 if (mdmon_running(st->container_devnm))
6424 st->update_tail = &st->updates;
6425
6426 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6427 pr_err("Failed to update subarray %s\n",
6428 subarray);
6429 } else {
6430 if (st->update_tail)
6431 flush_metadata_updates(st);
6432 else
6433 st->ss->sync_metadata(st);
6434 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6435 for (member_dev = info->devs; member_dev;
6436 member_dev = member_dev->next)
6437 member_dev->ppl_size =
6438 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6439 }
6440 }
6441
b23d0750 6442 if (ret == 1) {
2fc0fc63
AP		 6443 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6444
50b9c10d
PB		 6445 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6446 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6447 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6448 (is_rebuilding(dev) &&
2ec9d182
AP		 6449 dev->vol.curr_migr_unit == 0 &&
6450 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6451 ret = st->ss->write_init_ppl(st, info, d->fd);
6452 else
6453 info->mismatch_cnt++;
e397cefe
AP		 6454 } else if (ret == 0 &&
6455 ppl_hdr->entries_count == 0 &&
6456 is_rebuilding(dev) &&
6457 info->resync_start == 0) {
6458 /*
6459 * The header has no entries - add a single empty entry and
6460 * rewrite the header to prevent the kernel from going into
6461 * resync after an interrupted rebuild.
6462 */
6463 ppl_hdr->entries_count = __cpu_to_le32(1);
6464 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6465 }
2432ce9b 6466
e397cefe
AP		 6467 free(buf_orig);
6468
2432ce9b
AP		 6469 return ret;
6470}
6471
2432ce9b
AP		 6472static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6473{
6474 struct intel_super *super = st->sb;
6475 struct dl *d;
6476 int ret = 0;
6477
6478 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6479 info->array.level != 5)
6480 return 0;
6481
6482 for (d = super->disks; d ; d = d->next) {
6483 if (d->index < 0 || is_failed(&d->disk))
6484 continue;
6485
6486 ret = st->ss->write_init_ppl(st, info, d->fd);
6487 if (ret)
6488 break;
6489 }
6490
6491 return ret;
6492}
43dad3d6 6493
c2c087e6
DW		 6494static int write_init_super_imsm(struct supertype *st)
6495{
9b1fb677
DW		 6496 struct intel_super *super = st->sb;
6497 int current_vol = super->current_vol;
2432ce9b
AP		 6498 int rv = 0;
6499 struct mdinfo info;
6500
6501 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6502
6503 /* we are done with current_vol reset it to point st at the container */
6504 super->current_vol = -1;
6505
8273f55e 6506 if (st->update_tail) {
43dad3d6
DW		 6507 /* queue the recently created array / added disk
6508 * as a metadata update */
8273f55e 6509
43dad3d6 6510 /* determine if we are creating a volume or adding a disk */
9b1fb677 6511 if (current_vol < 0) {
1a64be56
LM		 6512 /* in the mgmt (add/remove) disk case we are running
6513 * in mdmon context, so don't close fd's
43dad3d6 6514 */
2432ce9b
AP		 6515 rv = mgmt_disk(st);
6516 } else {
6517 rv = write_init_ppl_imsm_all(st, &info);
6518 if (!rv)
6519 rv = create_array(st, current_vol);
6520 }
d682f344
N		 6521 } else {
6522 struct dl *d;
6523 for (d = super->disks; d; d = d->next)
ba728be7 6524 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6525 if (current_vol >= 0)
6526 rv = write_init_ppl_imsm_all(st, &info);
6527 if (!rv)
6528 rv = write_super_imsm(st, 1);
d682f344 6529 }
2432ce9b
AP		 6530
6531 return rv;
cdddbdbc
DW		 6532}
6533
e683ca88 6534static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6535{
e683ca88
DW		 6536 struct intel_super *super = st->sb;
6537 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6538
e683ca88 6539 if (!mpb)
ad97895e
DW		 6540 return 1;
6541
f36a9ecd
PB		 6542 if (super->sector_size == 4096)
6543 convert_to_4k(super);
e683ca88 6544 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6545}
6546
cdddbdbc
DW		 6547static int validate_geometry_imsm_container(struct supertype *st, int level,
6548 int layout, int raiddisks, int chunk,
af4348dd
N		 6549 unsigned long long size,
6550 unsigned long long data_offset,
6551 char *dev,
2c514b71
NB		 6552 unsigned long long *freesize,
6553 int verbose)
cdddbdbc 6554{
c2c087e6
DW		 6555 int fd;
6556 unsigned long long ldsize;
594dc1b8 6557 struct intel_super *super;
f2f5c343 6558 int rv = 0;
cdddbdbc 6559
c2c087e6
DW		 6560 if (level != LEVEL_CONTAINER)
6561 return 0;
6562 if (!dev)
6563 return 1;
6564
6565 fd = open(dev, O_RDONLY|O_EXCL, 0);
6566 if (fd < 0) {
ba728be7 6567 if (verbose > 0)
e7b84f9d 6568 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6569 dev, strerror(errno));
c2c087e6
DW		 6570 return 0;
6571 }
6572 if (!get_dev_size(fd, dev, &ldsize)) {
6573 close(fd);
6574 return 0;
6575 }
f2f5c343
LM		 6576
6577 /* capabilities retrieve could be possible
6578 * note that there is no fd for the disks in array.
6579 */
6580 super = alloc_super();
8d67477f
TM		 6581 if (!super) {
6582 close(fd);
6583 return 0;
6584 }
fa7bb6f8
PB		 6585 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6586 close(fd);
6587 free_imsm(super);
6588 return 0;
6589 }
6590
ba728be7 6591 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6592 if (rv != 0) {
6593#if DEBUG
6594 char str[256];
6595 fd2devname(fd, str);
1ade5cc1 6596 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6597 fd, str, super->orom, rv, raiddisks);
6598#endif
6599 /* no orom/efi or non-intel hba of the disk */
6600 close(fd);
6601 free_imsm(super);
6602 return 0;
6603 }
c2c087e6 6604 close(fd);
9126b9a8
CA		 6605 if (super->orom) {
6606 if (raiddisks > super->orom->tds) {
6607 if (verbose)
7a862a02 6608 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6609 raiddisks, super->orom->tds);
6610 free_imsm(super);
6611 return 0;
6612 }
6613 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6614 (ldsize >> 9) >> 32 > 0) {
6615 if (verbose)
e7b84f9d 6616 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6617 free_imsm(super);
6618 return 0;
6619 }
f2f5c343 6620 }
c2c087e6 6621
af4348dd 6622 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6623 free_imsm(super);
c2c087e6
DW		 6624
6625 return 1;
cdddbdbc
DW		 6626}
6627
0dcecb2e
DW		 6628static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6629{
6630 const unsigned long long base_start = e[*idx].start;
6631 unsigned long long end = base_start + e[*idx].size;
6632 int i;
6633
6634 if (base_start == end)
6635 return 0;
6636
6637 *idx = *idx + 1;
6638 for (i = *idx; i < num_extents; i++) {
6639 /* extend overlapping extents */
6640 if (e[i].start >= base_start &&
6641 e[i].start <= end) {
6642 if (e[i].size == 0)
6643 return 0;
6644 if (e[i].start + e[i].size > end)
6645 end = e[i].start + e[i].size;
6646 } else if (e[i].start > end) {
6647 *idx = i;
6648 break;
6649 }
6650 }
6651
6652 return end - base_start;
6653}
6654
6655static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6656{
6657 /* build a composite disk with all known extents and generate a new
6658 * 'maxsize' given the "all disks in an array must share a common start
6659 * offset" constraint
6660 */
503975b9 6661 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6662 struct dl *dl;
6663 int i, j;
6664 int start_extent;
6665 unsigned long long pos;
b9d77223 6666 unsigned long long start = 0;
0dcecb2e
DW		 6667 unsigned long long maxsize;
6668 unsigned long reserve;
6669
0dcecb2e
DW		 6670 /* coalesce and sort all extents. also, check to see if we need to
6671 * reserve space between member arrays
6672 */
6673 j = 0;
6674 for (dl = super->disks; dl; dl = dl->next) {
6675 if (!dl->e)
6676 continue;
6677 for (i = 0; i < dl->extent_cnt; i++)
6678 e[j++] = dl->e[i];
6679 }
6680 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6681
6682 /* merge extents */
6683 i = 0;
6684 j = 0;
6685 while (i < sum_extents) {
6686 e[j].start = e[i].start;
6687 e[j].size = find_size(e, &i, sum_extents);
6688 j++;
6689 if (e[j-1].size == 0)
6690 break;
6691 }
6692
6693 pos = 0;
6694 maxsize = 0;
6695 start_extent = 0;
6696 i = 0;
6697 do {
6698 unsigned long long esize;
6699
6700 esize = e[i].start - pos;
6701 if (esize >= maxsize) {
6702 maxsize = esize;
6703 start = pos;
6704 start_extent = i;
6705 }
6706 pos = e[i].start + e[i].size;
6707 i++;
6708 } while (e[i-1].size);
6709 free(e);
6710
a7dd165b
DW		 6711 if (maxsize == 0)
6712 return 0;
6713
6714 /* FIXME assumes volume at offset 0 is the first volume in a
6715 * container
6716 */
0dcecb2e
DW		 6717 if (start_extent > 0)
6718 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6719 else
6720 reserve = 0;
6721
6722 if (maxsize < reserve)
a7dd165b 6723 return 0;
0dcecb2e 6724
5551b113 6725 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6726 if (start + reserve > super->create_offset)
a7dd165b 6727 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6728 super->create_offset = start + reserve;
6729
6730 return maxsize - reserve;
6731}
6732
88c32bb1
DW		 6733static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6734{
6735 if (level < 0 || level == 6 || level == 4)
6736 return 0;
6737
6738 /* if we have an orom prevent invalid raid levels */
6739 if (orom)
6740 switch (level) {
6741 case 0: return imsm_orom_has_raid0(orom);
6742 case 1:
6743 if (raiddisks > 2)
6744 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6745 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6746 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6747 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6748 }
6749 else
6750 return 1; /* not on an Intel RAID platform so anything goes */
6751
6752 return 0;
6753}
6754
ca9de185
LM		 6755static int
6756active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6757 int dpa, int verbose)
6758{
6759 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6760 struct mdstat_ent *memb;
ca9de185
LM		 6761 int count = 0;
6762 int num = 0;
594dc1b8 6763 struct md_list *dv;
ca9de185
LM		 6764 int found;
6765
6766 for (memb = mdstat ; memb ; memb = memb->next) {
6767 if (memb->metadata_version &&
fc54fe7a 6768 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6769 (strcmp(&memb->metadata_version[9], name) == 0) &&
6770 !is_subarray(memb->metadata_version+9) &&
6771 memb->members) {
6772 struct dev_member *dev = memb->members;
6773 int fd = -1;
6774 while(dev && (fd < 0)) {
503975b9
N		 6775 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6776 num = sprintf(path, "%s%s", "/dev/", dev->name);
6777 if (num > 0)
6778 fd = open(path, O_RDONLY, 0);
089f9d79 6779 if (num <= 0 || fd < 0) {
676e87a8 6780 pr_vrb("Cannot open %s: %s\n",
503975b9 6781 dev->name, strerror(errno));
ca9de185 6782 }
503975b9 6783 free(path);
ca9de185
LM		 6784 dev = dev->next;
6785 }
6786 found = 0;
089f9d79 6787 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6788 struct mdstat_ent *vol;
6789 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6790 if (vol->active > 0 &&
ca9de185 6791 vol->metadata_version &&
9581efb1 6792 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6793 found++;
6794 count++;
6795 }
6796 }
6797 if (*devlist && (found < dpa)) {
503975b9 6798 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6799 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6800 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6801 dv->found = found;
6802 dv->used = 0;
6803 dv->next = *devlist;
6804 *devlist = dv;
ca9de185
LM		 6805 }
6806 }
6807 if (fd >= 0)
6808 close(fd);
6809 }
6810 }
6811 free_mdstat(mdstat);
6812 return count;
6813}
6814
6815#ifdef DEBUG_LOOP
6816static struct md_list*
6817get_loop_devices(void)
6818{
6819 int i;
6820 struct md_list *devlist = NULL;
594dc1b8 6821 struct md_list *dv;
ca9de185
LM		 6822
6823 for(i = 0; i < 12; i++) {
503975b9
N		 6824 dv = xcalloc(1, sizeof(*dv));
6825 dv->devname = xmalloc(40);
ca9de185
LM		 6826 sprintf(dv->devname, "/dev/loop%d", i);
6827 dv->next = devlist;
6828 devlist = dv;
6829 }
6830 return devlist;
6831}
6832#endif
6833
6834static struct md_list*
6835get_devices(const char *hba_path)
6836{
6837 struct md_list *devlist = NULL;
594dc1b8 6838 struct md_list *dv;
ca9de185
LM		 6839 struct dirent *ent;
6840 DIR *dir;
6841 int err = 0;
6842
6843#if DEBUG_LOOP
6844 devlist = get_loop_devices();
6845 return devlist;
6846#endif
6847 /* scroll through /sys/dev/block looking for devices attached to
6848 * this hba
6849 */
6850 dir = opendir("/sys/dev/block");
6851 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6852 int fd;
6853 char buf[1024];
6854 int major, minor;
6855 char *path = NULL;
6856 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6857 continue;
6858 path = devt_to_devpath(makedev(major, minor));
6859 if (!path)
6860 continue;
6861 if (!path_attached_to_hba(path, hba_path)) {
6862 free(path);
6863 path = NULL;
6864 continue;
6865 }
6866 free(path);
6867 path = NULL;
6868 fd = dev_open(ent->d_name, O_RDONLY);
6869 if (fd >= 0) {
6870 fd2devname(fd, buf);
6871 close(fd);
6872 } else {
e7b84f9d 6873 pr_err("cannot open device: %s\n",
ca9de185
LM		 6874 ent->d_name);
6875 continue;
6876 }
6877
503975b9
N		 6878 dv = xcalloc(1, sizeof(*dv));
6879 dv->devname = xstrdup(buf);
ca9de185
LM		 6880 dv->next = devlist;
6881 devlist = dv;
6882 }
6883 if (err) {
6884 while(devlist) {
6885 dv = devlist;
6886 devlist = devlist->next;
6887 free(dv->devname);
6888 free(dv);
6889 }
6890 }
562aa102 6891 closedir(dir);
ca9de185
LM		 6892 return devlist;
6893}
6894
6895static int
6896count_volumes_list(struct md_list *devlist, char *homehost,
6897 int verbose, int *found)
6898{
6899 struct md_list *tmpdev;
6900 int count = 0;
594dc1b8 6901 struct supertype *st;
ca9de185
LM		 6902
6903 /* first walk the list of devices to find a consistent set
6904 * that match the criterea, if that is possible.
6905 * We flag the ones we like with 'used'.
6906 */
6907 *found = 0;
6908 st = match_metadata_desc_imsm("imsm");
6909 if (st == NULL) {
676e87a8 6910 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6911 return 0;
6912 }
6913
6914 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6915 char *devname = tmpdev->devname;
0a6bff09 6916 dev_t rdev;
ca9de185
LM		 6917 struct supertype *tst;
6918 int dfd;
6919 if (tmpdev->used > 1)
6920 continue;
6921 tst = dup_super(st);
6922 if (tst == NULL) {
676e87a8 6923 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6924 goto err_1;
6925 }
6926 tmpdev->container = 0;
6927 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6928 if (dfd < 0) {
1ade5cc1 6929 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6930 devname, strerror(errno));
6931 tmpdev->used = 2;
0a6bff09 6932 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6933 tmpdev->used = 2;
6934 } else if (must_be_container(dfd)) {
6935 struct supertype *cst;
6936 cst = super_by_fd(dfd, NULL);
6937 if (cst == NULL) {
1ade5cc1 6938 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6939 devname);
6940 tmpdev->used = 2;
6941 } else if (tst->ss != st->ss) {
1ade5cc1 6942 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6943 devname);
6944 tmpdev->used = 2;
6945 } else if (!tst->ss->load_container ||
6946 tst->ss->load_container(tst, dfd, NULL))
6947 tmpdev->used = 2;
6948 else {
6949 tmpdev->container = 1;
6950 }
6951 if (cst)
6952 cst->ss->free_super(cst);
6953 } else {
0a6bff09 6954 tmpdev->st_rdev = rdev;
ca9de185 6955 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6956 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6957 devname);
6958 tmpdev->used = 2;
6959 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6960 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6961 tst->ss->name, devname);
6962 tmpdev->used = 2;
6963 }
6964 }
6965 if (dfd >= 0)
6966 close(dfd);
6967 if (tmpdev->used == 2 || tmpdev->used == 4) {
6968 /* Ignore unrecognised devices during auto-assembly */
6969 goto loop;
6970 }
6971 else {
6972 struct mdinfo info;
6973 tst->ss->getinfo_super(tst, &info, NULL);
6974
6975 if (st->minor_version == -1)
6976 st->minor_version = tst->minor_version;
6977
6978 if (memcmp(info.uuid, uuid_zero,
6979 sizeof(int[4])) == 0) {
6980 /* this is a floating spare. It cannot define
6981 * an array unless there are no more arrays of
6982 * this type to be found. It can be included
6983 * in an array of this type though.
6984 */
6985 tmpdev->used = 3;
6986 goto loop;
6987 }
6988
6989 if (st->ss != tst->ss ||
6990 st->minor_version != tst->minor_version ||
6991 st->ss->compare_super(st, tst) != 0) {
6992 /* Some mismatch. If exactly one array matches this host,
6993 * we can resolve on that one.
6994 * Or, if we are auto assembling, we just ignore the second
6995 * for now.
6996 */
1ade5cc1 6997 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6998 devname);
6999 goto loop;
7000 }
7001 tmpdev->used = 1;
7002 *found = 1;
7003 dprintf("found: devname: %s\n", devname);
7004 }
7005 loop:
7006 if (tst)
7007 tst->ss->free_super(tst);
7008 }
7009 if (*found != 0) {
7010 int err;
7011 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
7012 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
7013 for (iter = head; iter; iter = iter->next) {
7014 dprintf("content->text_version: %s vol\n",
7015 iter->text_version);
7016 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
7017 /* do not assemble arrays with unsupported
7018 configurations */
1ade5cc1 7019 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 7020 iter->text_version);
7021 } else
7022 count++;
7023 }
7024 sysfs_free(head);
7025
7026 } else {
1ade5cc1 7027 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7028 err, st->sb);
7029 }
7030 } else {
1ade5cc1 7031 dprintf("no more devices to examine\n");
ca9de185
LM		 7032 }
7033
7034 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7035 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7036 if (count) {
7037 if (count < tmpdev->found)
7038 count = 0;
7039 else
7040 count -= tmpdev->found;
7041 }
7042 }
7043 if (tmpdev->used == 1)
7044 tmpdev->used = 4;
7045 }
7046 err_1:
7047 if (st)
7048 st->ss->free_super(st);
7049 return count;
7050}
7051
d3c11416
AO		 7052static int __count_volumes(char *hba_path, int dpa, int verbose,
7053 int cmp_hba_path)
ca9de185 7054{
72a45777 7055 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7056 int count = 0;
72a45777
PB		 7057 const struct orom_entry *entry;
7058 struct devid_list *dv, *devid_list;
ca9de185 7059
d3c11416 7060 if (!hba_path)
ca9de185
LM		 7061 return 0;
7062
72a45777 7063 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7064 if (strstr(idev->path, hba_path))
7065 break;
72a45777
PB		 7066 }
7067
7068 if (!idev || !idev->dev_id)
ca9de185 7069 return 0;
72a45777
PB		 7070
7071 entry = get_orom_entry_by_device_id(idev->dev_id);
7072
7073 if (!entry || !entry->devid_list)
7074 return 0;
7075
7076 devid_list = entry->devid_list;
7077 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7078 struct md_list *devlist;
d3c11416
AO		 7079 struct sys_dev *device = NULL;
7080 char *hpath;
72a45777
PB		 7081 int found = 0;
7082
d3c11416
AO		 7083 if (cmp_hba_path)
7084 device = device_by_id_and_path(dv->devid, hba_path);
7085 else
7086 device = device_by_id(dv->devid);
7087
72a45777 7088 if (device)
d3c11416 7089 hpath = device->path;
72a45777
PB		 7090 else
7091 return 0;
7092
d3c11416 7093 devlist = get_devices(hpath);
72a45777
PB		 7094 /* if no intel devices return zero volumes */
7095 if (devlist == NULL)
7096 return 0;
7097
d3c11416
AO		 7098 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7099 verbose);
7100 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7101 if (devlist == NULL)
7102 return 0;
7103 do {
7104 found = 0;
7105 count += count_volumes_list(devlist,
7106 NULL,
7107 verbose,
7108 &found);
7109 dprintf("found %d count: %d\n", found, count);
7110 } while (found);
7111
d3c11416 7112 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7113
7114 while (devlist) {
7115 struct md_list *dv = devlist;
7116 devlist = devlist->next;
7117 free(dv->devname);
7118 free(dv);
7119 }
ca9de185
LM		 7120 }
7121 return count;
7122}
7123
d3c11416
AO		 7124static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7125{
7126 if (!hba)
7127 return 0;
7128 if (hba->type == SYS_DEV_VMD) {
7129 struct sys_dev *dev;
7130 int count = 0;
7131
7132 for (dev = find_intel_devices(); dev; dev = dev->next) {
7133 if (dev->type == SYS_DEV_VMD)
7134 count += __count_volumes(dev->path, dpa,
7135 verbose, 1);
7136 }
7137 return count;
7138 }
7139 return __count_volumes(hba->path, dpa, verbose, 0);
7140}
7141
cd9d1ac7
DW		 7142static int imsm_default_chunk(const struct imsm_orom *orom)
7143{
7144 /* up to 512 if the plaform supports it, otherwise the platform max.
7145 * 128 if no platform detected
7146 */
7147 int fs = max(7, orom ? fls(orom->sss) : 0);
7148
7149 return min(512, (1 << fs));
7150}
73408129 7151
6592ce37
DW		 7152static int
7153validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7154 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7155{
660260d0
DW		 7156 /* check/set platform and metadata limits/defaults */
7157 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7158 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7159 super->orom->dpa);
73408129
LM		 7160 return 0;
7161 }
7162
5d500228 7163 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7164 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7165 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7166 level, raiddisks, raiddisks > 1 ? "s" : "");
7167 return 0;
7168 }
cd9d1ac7 7169
7ccc4cc4 7170 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7171 *chunk = imsm_default_chunk(super->orom);
7172
7ccc4cc4 7173 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7174 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7175 return 0;
6592ce37 7176 }
cd9d1ac7 7177
6592ce37
DW		 7178 if (layout != imsm_level_to_layout(level)) {
7179 if (level == 5)
676e87a8 7180 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7181 else if (level == 10)
676e87a8 7182 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7183 else
676e87a8 7184 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7185 layout, level);
7186 return 0;
7187 }
2cc699af 7188
7ccc4cc4 7189 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7190 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7191 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7192 return 0;
7193 }
614902f6 7194
6592ce37
DW		 7195 return 1;
7196}
7197
1011e834 7198/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7199 * FIX ME add ahci details
7200 */
8b353278 7201static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7202 int layout, int raiddisks, int *chunk,
af4348dd
N		 7203 unsigned long long size,
7204 unsigned long long data_offset,
7205 char *dev,
2c514b71
NB		 7206 unsigned long long *freesize,
7207 int verbose)
cdddbdbc 7208{
9e04ac1c 7209 dev_t rdev;
c2c087e6 7210 struct intel_super *super = st->sb;
b2916f25 7211 struct imsm_super *mpb;
c2c087e6
DW		 7212 struct dl *dl;
7213 unsigned long long pos = 0;
7214 unsigned long long maxsize;
7215 struct extent *e;
7216 int i;
cdddbdbc 7217
88c32bb1
DW		 7218 /* We must have the container info already read in. */
7219 if (!super)
c2c087e6
DW		 7220 return 0;
7221
b2916f25
JS		 7222 mpb = super->anchor;
7223
2cc699af 7224 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7225 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7226 return 0;
d54559f0 7227 }
c2c087e6
DW		 7228 if (!dev) {
7229 /* General test: make sure there is space for
2da8544a
DW		 7230 * 'raiddisks' device extents of size 'size' at a given
7231 * offset
c2c087e6 7232 */
e46273eb 7233 unsigned long long minsize = size;
b7528a20 7234 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7235 int dcnt = 0;
7236 if (minsize == 0)
7237 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7238 for (dl = super->disks; dl ; dl = dl->next) {
7239 int found = 0;
7240
bf5a934a 7241 pos = 0;
c2c087e6 7242 i = 0;
05501181 7243 e = get_extents(super, dl, 0);
c2c087e6
DW		 7244 if (!e) continue;
7245 do {
7246 unsigned long long esize;
7247 esize = e[i].start - pos;
7248 if (esize >= minsize)
7249 found = 1;
b7528a20 7250 if (found && start_offset == MaxSector) {
2da8544a
DW		 7251 start_offset = pos;
7252 break;
7253 } else if (found && pos != start_offset) {
7254 found = 0;
7255 break;
7256 }
c2c087e6
DW		 7257 pos = e[i].start + e[i].size;
7258 i++;
7259 } while (e[i-1].size);
7260 if (found)
7261 dcnt++;
7262 free(e);
7263 }
7264 if (dcnt < raiddisks) {
2c514b71 7265 if (verbose)
7a862a02 7266 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7267 dcnt, raiddisks);
c2c087e6
DW		 7268 return 0;
7269 }
7270 return 1;
7271 }
0dcecb2e 7272
c2c087e6 7273 /* This device must be a member of the set */
9e04ac1c 7274 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7275 return 0;
7276 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7277 if (dl->major == (int)major(rdev) &&
7278 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7279 break;
7280 }
7281 if (!dl) {
2c514b71 7282 if (verbose)
7a862a02 7283 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7284 return 0;
a20d2ba5
DW		 7285 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7286 /* If a volume is present then the current creation attempt
7287 * cannot incorporate new spares because the orom may not
7288 * understand this configuration (all member disks must be
7289 * members of each array in the container).
7290 */
7a862a02
N		 7291 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7292 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7293 return 0;
5fe62b94
WD		 7294 } else if (super->orom && mpb->num_raid_devs > 0 &&
7295 mpb->num_disks != raiddisks) {
7a862a02 7296 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7297 return 0;
c2c087e6 7298 }
0dcecb2e
DW		 7299
7300 /* retrieve the largest free space block */
05501181 7301 e = get_extents(super, dl, 0);
c2c087e6
DW		 7302 maxsize = 0;
7303 i = 0;
0dcecb2e
DW		 7304 if (e) {
7305 do {
7306 unsigned long long esize;
7307
7308 esize = e[i].start - pos;
7309 if (esize >= maxsize)
7310 maxsize = esize;
7311 pos = e[i].start + e[i].size;
7312 i++;
7313 } while (e[i-1].size);
7314 dl->e = e;
7315 dl->extent_cnt = i;
7316 } else {
7317 if (verbose)
e7b84f9d 7318 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7319 dev);
7320 return 0;
7321 }
7322 if (maxsize < size) {
7323 if (verbose)
e7b84f9d 7324 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7325 dev, maxsize, size);
7326 return 0;
7327 }
7328
7329 /* count total number of extents for merge */
7330 i = 0;
7331 for (dl = super->disks; dl; dl = dl->next)
7332 if (dl->e)
7333 i += dl->extent_cnt;
7334
7335 maxsize = merge_extents(super, i);
3baa56ab 7336
1a1ced1e
KS		 7337 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7338 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7339
a7dd165b 7340 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7341 if (verbose) {
7342 if (maxsize == 0)
7a862a02 7343 pr_err("no free space left on device. Aborting...\n");
b3071342 7344 else
7a862a02 7345 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7346 maxsize, size);
7347 }
0dcecb2e 7348 return 0;
0dcecb2e
DW		 7349 }
7350
c2c087e6
DW		 7351 *freesize = maxsize;
7352
ca9de185 7353 if (super->orom) {
72a45777 7354 int count = count_volumes(super->hba,
ca9de185
LM		 7355 super->orom->dpa, verbose);
7356 if (super->orom->vphba <= count) {
676e87a8 7357 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7358 super->orom->vphba);
7359 return 0;
7360 }
7361 }
c2c087e6 7362 return 1;
cdddbdbc
DW		 7363}
7364
13bcac90 7365static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7366 unsigned long long size, int chunk,
7367 unsigned long long *freesize)
7368{
7369 struct intel_super *super = st->sb;
7370 struct imsm_super *mpb = super->anchor;
7371 struct dl *dl;
7372 int i;
7373 int extent_cnt;
7374 struct extent *e;
7375 unsigned long long maxsize;
7376 unsigned long long minsize;
7377 int cnt;
7378 int used;
7379
7380 /* find the largest common start free region of the possible disks */
7381 used = 0;
7382 extent_cnt = 0;
7383 cnt = 0;
7384 for (dl = super->disks; dl; dl = dl->next) {
7385 dl->raiddisk = -1;
7386
7387 if (dl->index >= 0)
7388 used++;
7389
7390 /* don't activate new spares if we are orom constrained
7391 * and there is already a volume active in the container
7392 */
7393 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7394 continue;
7395
05501181 7396 e = get_extents(super, dl, 0);
efb30e7f
DW		 7397 if (!e)
7398 continue;
7399 for (i = 1; e[i-1].size; i++)
7400 ;
7401 dl->e = e;
7402 dl->extent_cnt = i;
7403 extent_cnt += i;
7404 cnt++;
7405 }
7406
7407 maxsize = merge_extents(super, extent_cnt);
7408 minsize = size;
7409 if (size == 0)
612e59d8
CA		 7410 /* chunk is in K */
7411 minsize = chunk * 2;
efb30e7f
DW		 7412
7413 if (cnt < raiddisks ||
7414 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7415 maxsize < minsize ||
7416 maxsize == 0) {
e7b84f9d 7417 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7418 return 0; /* No enough free spaces large enough */
7419 }
7420
7421 if (size == 0) {
7422 size = maxsize;
7423 if (chunk) {
612e59d8
CA		 7424 size /= 2 * chunk;
7425 size *= 2 * chunk;
efb30e7f 7426 }
f878b242
LM		 7427 maxsize = size;
7428 }
1a1ced1e
KS		 7429 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7430 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7431 cnt = 0;
7432 for (dl = super->disks; dl; dl = dl->next)
7433 if (dl->e)
7434 dl->raiddisk = cnt++;
7435
7436 *freesize = size;
7437
13bcac90
AK		 7438 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7439
efb30e7f
DW		 7440 return 1;
7441}
7442
13bcac90
AK		 7443static int reserve_space(struct supertype *st, int raiddisks,
7444 unsigned long long size, int chunk,
7445 unsigned long long *freesize)
7446{
7447 struct intel_super *super = st->sb;
7448 struct dl *dl;
7449 int cnt;
7450 int rv = 0;
7451
7452 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7453 if (rv) {
7454 cnt = 0;
7455 for (dl = super->disks; dl; dl = dl->next)
7456 if (dl->e)
7457 dl->raiddisk = cnt++;
7458 rv = 1;
7459 }
7460
7461 return rv;
7462}
7463
bf5a934a 7464static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7465 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7466 unsigned long long data_offset,
bf5a934a 7467 char *dev, unsigned long long *freesize,
5308f117 7468 int consistency_policy, int verbose)
bf5a934a
DW		 7469{
7470 int fd, cfd;
7471 struct mdinfo *sra;
20cbe8d2 7472 int is_member = 0;
bf5a934a 7473
d54559f0
LM		 7474 /* load capability
7475 * if given unused devices create a container
bf5a934a
DW		 7476 * if given given devices in a container create a member volume
7477 */
7478 if (level == LEVEL_CONTAINER) {
7479 /* Must be a fresh device to add to a container */
7480 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7481 raiddisks,
7ccc4cc4 7482 *chunk,
af4348dd 7483 size, data_offset,
bf5a934a
DW		 7484 dev, freesize,
7485 verbose);
7486 }
9587c373 7487
22dc741f
MT		 7488 if (size && (size < 1024)) {
7489 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7490 /* Depends on algorithm in Create.c :
7491 * if container was given (dev == NULL) return -1,
7492 * if block device was given ( dev != NULL) return 0.
7493 */
7494 return dev ? -1 : 0;
7495 }
7496
8592f29d 7497 if (!dev) {
e91a3bad 7498 if (st->sb) {
ca9de185 7499 struct intel_super *super = st->sb;
e91a3bad 7500 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7501 raiddisks, chunk, size,
e91a3bad
LM		 7502 verbose))
7503 return 0;
efb30e7f
DW		 7504 /* we are being asked to automatically layout a
7505 * new volume based on the current contents of
7506 * the container. If the the parameters can be
7507 * satisfied reserve_space will record the disks,
7508 * start offset, and size of the volume to be
7509 * created. add_to_super and getinfo_super
7510 * detect when autolayout is in progress.
7511 */
ca9de185
LM		 7512 /* assuming that freesize is always given when array is
7513 created */
7514 if (super->orom && freesize) {
7515 int count;
72a45777 7516 count = count_volumes(super->hba,
ca9de185
LM		 7517 super->orom->dpa, verbose);
7518 if (super->orom->vphba <= count) {
676e87a8 7519 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7520 super->orom->vphba);
7521 return 0;
7522 }
7523 }
e91a3bad
LM		 7524 if (freesize)
7525 return reserve_space(st, raiddisks, size,
7ccc4cc4 7526 *chunk, freesize);
8592f29d
N		 7527 }
7528 return 1;
7529 }
bf5a934a
DW		 7530 if (st->sb) {
7531 /* creating in a given container */
7532 return validate_geometry_imsm_volume(st, level, layout,
7533 raiddisks, chunk, size,
af4348dd 7534 data_offset,
bf5a934a
DW		 7535 dev, freesize, verbose);
7536 }
7537
bf5a934a
DW		 7538 /* This device needs to be a device in an 'imsm' container */
7539 fd = open(dev, O_RDONLY|O_EXCL, 0);
7540 if (fd >= 0) {
7541 if (verbose)
e7b84f9d
N		 7542 pr_err("Cannot create this array on device %s\n",
7543 dev);
bf5a934a
DW		 7544 close(fd);
7545 return 0;
7546 }
7547 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7548 if (verbose)
e7b84f9d 7549 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7550 dev, strerror(errno));
7551 return 0;
7552 }
7553 /* Well, it is in use by someone, maybe an 'imsm' container. */
7554 cfd = open_container(fd);
20cbe8d2 7555 close(fd);
bf5a934a 7556 if (cfd < 0) {
bf5a934a 7557 if (verbose)
e7b84f9d 7558 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7559 dev);
7560 return 0;
7561 }
4dd2df09 7562 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7563 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7564 strcmp(sra->text_version, "imsm") == 0)
7565 is_member = 1;
7566 sysfs_free(sra);
7567 if (is_member) {
bf5a934a
DW		 7568 /* This is a member of a imsm container. Load the container
7569 * and try to create a volume
7570 */
7571 struct intel_super *super;
7572
ec50f7b6 7573 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7574 st->sb = super;
4dd2df09 7575 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7576 close(cfd);
7577 return validate_geometry_imsm_volume(st, level, layout,
7578 raiddisks, chunk,
af4348dd 7579 size, data_offset, dev,
ecbd9e81
N		 7580 freesize, 1)
7581 ? 1 : -1;
bf5a934a 7582 }
20cbe8d2 7583 }
bf5a934a 7584
20cbe8d2 7585 if (verbose)
e7b84f9d 7586 pr_err("failed container membership check\n");
20cbe8d2
AW		 7587
7588 close(cfd);
7589 return 0;
bf5a934a 7590}
0bd16cf2 7591
30f58b22 7592static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7593{
7594 struct intel_super *super = st->sb;
7595
30f58b22
DW		 7596 if (level && *level == UnSet)
7597 *level = LEVEL_CONTAINER;
7598
7599 if (level && layout && *layout == UnSet)
7600 *layout = imsm_level_to_layout(*level);
0bd16cf2 7601
cd9d1ac7
DW		 7602 if (chunk && (*chunk == UnSet || *chunk == 0))
7603 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7604}
7605
33414a01
DW		 7606static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7607
7608static int kill_subarray_imsm(struct supertype *st)
7609{
7610 /* remove the subarray currently referenced by ->current_vol */
7611 __u8 i;
7612 struct intel_dev **dp;
7613 struct intel_super *super = st->sb;
7614 __u8 current_vol = super->current_vol;
7615 struct imsm_super *mpb = super->anchor;
7616
7617 if (super->current_vol < 0)
7618 return 2;
7619 super->current_vol = -1; /* invalidate subarray cursor */
7620
7621 /* block deletions that would change the uuid of active subarrays
7622 *
7623 * FIXME when immutable ids are available, but note that we'll
7624 * also need to fixup the invalidated/active subarray indexes in
7625 * mdstat
7626 */
7627 for (i = 0; i < mpb->num_raid_devs; i++) {
7628 char subarray[4];
7629
7630 if (i < current_vol)
7631 continue;
7632 sprintf(subarray, "%u", i);
4dd2df09 7633 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7634 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7635 current_vol, i);
33414a01
DW		 7636
7637 return 2;
7638 }
7639 }
7640
7641 if (st->update_tail) {
503975b9 7642 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7643
33414a01
DW		 7644 u->type = update_kill_array;
7645 u->dev_idx = current_vol;
7646 append_metadata_update(st, u, sizeof(*u));
7647
7648 return 0;
7649 }
7650
7651 for (dp = &super->devlist; *dp;)
7652 if ((*dp)->index == current_vol) {
7653 *dp = (*dp)->next;
7654 } else {
7655 handle_missing(super, (*dp)->dev);
7656 if ((*dp)->index > current_vol)
7657 (*dp)->index--;
7658 dp = &(*dp)->next;
7659 }
7660
7661 /* no more raid devices, all active components are now spares,
7662 * but of course failed are still failed
7663 */
7664 if (--mpb->num_raid_devs == 0) {
7665 struct dl *d;
7666
7667 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7668 if (d->index > -2)
7669 mark_spare(d);
33414a01
DW		 7670 }
7671
7672 super->updates_pending++;
7673
7674 return 0;
7675}
aa534678 7676
a951a4f7 7677static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7678 char *update, struct mddev_ident *ident)
aa534678
DW		 7679{
7680 /* update the subarray currently referenced by ->current_vol */
7681 struct intel_super *super = st->sb;
7682 struct imsm_super *mpb = super->anchor;
7683
aa534678
DW		 7684 if (strcmp(update, "name") == 0) {
7685 char *name = ident->name;
a951a4f7
N		 7686 char *ep;
7687 int vol;
aa534678 7688
4dd2df09 7689 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7690 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7691 return 2;
7692 }
7693
7694 if (!check_name(super, name, 0))
7695 return 2;
7696
a951a4f7
N		 7697 vol = strtoul(subarray, &ep, 10);
7698 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7699 return 2;
7700
aa534678 7701 if (st->update_tail) {
503975b9 7702 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7703
aa534678 7704 u->type = update_rename_array;
a951a4f7 7705 u->dev_idx = vol;
618f4e6d
XN		 7706 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7707 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7708 append_metadata_update(st, u, sizeof(*u));
7709 } else {
7710 struct imsm_dev *dev;
ebad3af2 7711 int i, namelen;
aa534678 7712
a951a4f7 7713 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7714 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7715 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7716 memcpy(dev->volume, name, namelen);
aa534678
DW		 7717 for (i = 0; i < mpb->num_raid_devs; i++) {
7718 dev = get_imsm_dev(super, i);
7719 handle_missing(super, dev);
7720 }
7721 super->updates_pending++;
7722 }
e6e9dd3f
AP		 7723 } else if (strcmp(update, "ppl") == 0 ||
7724 strcmp(update, "no-ppl") == 0) {
7725 int new_policy;
7726 char *ep;
7727 int vol = strtoul(subarray, &ep, 10);
7728
7729 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7730 return 2;
7731
7732 if (strcmp(update, "ppl") == 0)
c2462068 7733 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7734 else
c2462068 7735 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7736
7737 if (st->update_tail) {
7738 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7739
7740 u->type = update_rwh_policy;
7741 u->dev_idx = vol;
7742 u->new_policy = new_policy;
7743 append_metadata_update(st, u, sizeof(*u));
7744 } else {
7745 struct imsm_dev *dev;
7746
7747 dev = get_imsm_dev(super, vol);
7748 dev->rwh_policy = new_policy;
7749 super->updates_pending++;
7750 }
aa534678
DW		 7751 } else
7752 return 2;
7753
7754 return 0;
7755}
bf5a934a 7756
28bce06f
AK		 7757static int is_gen_migration(struct imsm_dev *dev)
7758{
7534230b
AK		 7759 if (dev == NULL)
7760 return 0;
7761
28bce06f
AK		 7762 if (!dev->vol.migr_state)
7763 return 0;
7764
7765 if (migr_type(dev) == MIGR_GEN_MIGR)
7766 return 1;
7767
7768 return 0;
7769}
7770
1e5c6983
DW		 7771static int is_rebuilding(struct imsm_dev *dev)
7772{
7773 struct imsm_map *migr_map;
7774
7775 if (!dev->vol.migr_state)
7776 return 0;
7777
7778 if (migr_type(dev) != MIGR_REBUILD)
7779 return 0;
7780
238c0a71 7781 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7782
7783 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7784 return 1;
7785 else
7786 return 0;
7787}
7788
6ce1fbf1
AK		 7789static int is_initializing(struct imsm_dev *dev)
7790{
7791 struct imsm_map *migr_map;
7792
7793 if (!dev->vol.migr_state)
7794 return 0;
7795
7796 if (migr_type(dev) != MIGR_INIT)
7797 return 0;
7798
238c0a71 7799 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7800
7801 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7802 return 1;
7803
7804 return 0;
6ce1fbf1
AK		 7805}
7806
c47b0ff6
AK		 7807static void update_recovery_start(struct intel_super *super,
7808 struct imsm_dev *dev,
7809 struct mdinfo *array)
1e5c6983
DW		 7810{
7811 struct mdinfo *rebuild = NULL;
7812 struct mdinfo *d;
7813 __u32 units;
7814
7815 if (!is_rebuilding(dev))
7816 return;
7817
7818 /* Find the rebuild target, but punt on the dual rebuild case */
7819 for (d = array->devs; d; d = d->next)
7820 if (d->recovery_start == 0) {
7821 if (rebuild)
7822 return;
7823 rebuild = d;
7824 }
7825
4363fd80
DW		 7826 if (!rebuild) {
7827 /* (?) none of the disks are marked with
7828 * IMSM_ORD_REBUILD, so assume they are missing and the
7829 * disk_ord_tbl was not correctly updated
7830 */
1ade5cc1 7831 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7832 return;
7833 }
7834
1e5c6983 7835 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7836 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7837}
7838
276d77db 7839static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7840
00bbdbda 7841static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7842{
4f5bc454
DW		 7843 /* Given a container loaded by load_super_imsm_all,
7844 * extract information about all the arrays into
7845 * an mdinfo tree.
00bbdbda 7846 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7847 *
7848 * For each imsm_dev create an mdinfo, fill it in,
7849 * then look for matching devices in super->disks
7850 * and create appropriate device mdinfo.
7851 */
7852 struct intel_super *super = st->sb;
949c47a0 7853 struct imsm_super *mpb = super->anchor;
4f5bc454 7854 struct mdinfo *rest = NULL;
00bbdbda 7855 unsigned int i;
81219e70 7856 int sb_errors = 0;
abef11a3
AK		 7857 struct dl *d;
7858 int spare_disks = 0;
b6180160 7859 int current_vol = super->current_vol;
cdddbdbc 7860
19482bcc
AK		 7861 /* do not assemble arrays when not all attributes are supported */
7862 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7863 sb_errors = 1;
7a862a02 7864 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7865 }
7866
abef11a3
AK		 7867 /* count spare devices, not used in maps
7868 */
7869 for (d = super->disks; d; d = d->next)
7870 if (d->index == -1)
7871 spare_disks++;
7872
4f5bc454 7873 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7874 struct imsm_dev *dev;
7875 struct imsm_map *map;
86e3692b 7876 struct imsm_map *map2;
4f5bc454 7877 struct mdinfo *this;
a6482415 7878 int slot;
a6482415 7879 int chunk;
00bbdbda 7880 char *ep;
8b9cd157 7881 int level;
00bbdbda
N		 7882
7883 if (subarray &&
7884 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7885 continue;
7886
7887 dev = get_imsm_dev(super, i);
238c0a71
AK		 7888 map = get_imsm_map(dev, MAP_0);
7889 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7890 level = get_imsm_raid_level(map);
4f5bc454 7891
1ce0101c
DW		 7892 /* do not publish arrays that are in the middle of an
7893 * unsupported migration
7894 */
7895 if (dev->vol.migr_state &&
28bce06f 7896 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7897 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7898 dev->volume);
7899 continue;
7900 }
2db86302
LM		 7901 /* do not publish arrays that are not support by controller's
7902 * OROM/EFI
7903 */
1ce0101c 7904
503975b9 7905 this = xmalloc(sizeof(*this));
4f5bc454 7906
301406c9 7907 super->current_vol = i;
a5d85af7 7908 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7909 this->next = rest;
a6482415 7910 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7911 /* mdadm does not support all metadata features- set the bit in all arrays state */
7912 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7913 level, /* RAID level */
7914 imsm_level_to_layout(level),
81219e70 7915 map->num_members, /* raid disks */
fcc2c9da 7916 &chunk, imsm_dev_size(dev),
81219e70 7917 1 /* verbose */)) {
7a862a02 7918 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7919 dev->volume);
7920 this->array.state |=
7921 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7922 (1<<MD_SB_BLOCK_VOLUME);
7923 }
81219e70
LM		 7924
7925 /* if array has bad blocks, set suitable bit in all arrays state */
7926 if (sb_errors)
7927 this->array.state |=
7928 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7929 (1<<MD_SB_BLOCK_VOLUME);
7930
4f5bc454 7931 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7932 unsigned long long recovery_start;
4f5bc454
DW		 7933 struct mdinfo *info_d;
7934 struct dl *d;
7935 int idx;
9a1608e5 7936 int skip;
7eef0453 7937 __u32 ord;
8b9cd157 7938 int missing = 0;
4f5bc454 7939
9a1608e5 7940 skip = 0;
238c0a71
AK		 7941 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7942 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7943 for (d = super->disks; d ; d = d->next)
7944 if (d->index == idx)
0fbd635c 7945 break;
4f5bc454 7946
1e5c6983 7947 recovery_start = MaxSector;
4f5bc454 7948 if (d == NULL)
9a1608e5 7949 skip = 1;
25ed7e59 7950 if (d && is_failed(&d->disk))
9a1608e5 7951 skip = 1;
8b9cd157 7952 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7953 recovery_start = 0;
9a1608e5 7954
1011e834 7955 /*
9a1608e5 7956 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7957 * reset resync start to avoid a dirty-degraded
7958 * situation when performing the intial sync
9a1608e5 7959 */
8b9cd157
MK		 7960 if (skip)
7961 missing++;
7962
7963 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7964 if ((!able_to_resync(level, missing) ||
7965 recovery_start == 0))
7966 this->resync_start = MaxSector;
7967 } else {
7968 /*
7969 * FIXME handle dirty degraded
7970 */
7971 }
7972
9a1608e5
DW		 7973 if (skip)
7974 continue;
4f5bc454 7975
503975b9 7976 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7977 info_d->next = this->devs;
7978 this->devs = info_d;
7979
4f5bc454
DW		 7980 info_d->disk.number = d->index;
7981 info_d->disk.major = d->major;
7982 info_d->disk.minor = d->minor;
7983 info_d->disk.raid_disk = slot;
1e5c6983 7984 info_d->recovery_start = recovery_start;
86e3692b
AK		 7985 if (map2) {
7986 if (slot < map2->num_members)
7987 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7988 else
7989 this->array.spare_disks++;
86e3692b
AK		 7990 } else {
7991 if (slot < map->num_members)
7992 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7993 else
7994 this->array.spare_disks++;
86e3692b 7995 }
1e5c6983
DW		 7996 if (info_d->recovery_start == MaxSector)
7997 this->array.working_disks++;
4f5bc454
DW		 7998
7999 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 8000 info_d->data_offset = pba_of_lba0(map);
44490938 8001 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 8002
8003 if (map->raid_level == 5) {
2432ce9b
AP		 8004 info_d->ppl_sector = this->ppl_sector;
8005 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 8006 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
8007 recovery_start == 0)
8008 this->resync_start = 0;
06fb291a 8009 }
b12796be 8010
5e46202e 8011 info_d->bb.supported = 1;
b12796be
TM		 8012 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
8013 info_d->data_offset,
8014 info_d->component_size,
8015 &info_d->bb);
4f5bc454 8016 }
1e5c6983 8017 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 8018 update_recovery_start(super, dev, this);
abef11a3 8019 this->array.spare_disks += spare_disks;
276d77db
AK		 8020
8021 /* check for reshape */
8022 if (this->reshape_active == 1)
8023 recover_backup_imsm(st, this);
9a1608e5 8024 rest = this;
4f5bc454
DW		 8025 }
8026
b6180160 8027 super->current_vol = current_vol;
4f5bc454 8028 return rest;
cdddbdbc
DW		 8029}
8030
3b451610
AK		 8031static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8032 int failed, int look_in_map)
c2a1e7da 8033{
3b451610
AK		 8034 struct imsm_map *map;
8035
8036 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8037
8038 if (!failed)
1011e834 8039 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8040 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8041
8042 switch (get_imsm_raid_level(map)) {
8043 case 0:
8044 return IMSM_T_STATE_FAILED;
8045 break;
8046 case 1:
8047 if (failed < map->num_members)
8048 return IMSM_T_STATE_DEGRADED;
8049 else
8050 return IMSM_T_STATE_FAILED;
8051 break;
8052 case 10:
8053 {
8054 /**
c92a2527
DW		 8055 * check to see if any mirrors have failed, otherwise we
8056 * are degraded. Even numbered slots are mirrored on
8057 * slot+1
c2a1e7da 8058 */
c2a1e7da 8059 int i;
d9b420a5
N		 8060 /* gcc -Os complains that this is unused */
8061 int insync = insync;
c2a1e7da
DW		 8062
8063 for (i = 0; i < map->num_members; i++) {
238c0a71 8064 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8065 int idx = ord_to_idx(ord);
8066 struct imsm_disk *disk;
c2a1e7da 8067
c92a2527 8068 /* reset the potential in-sync count on even-numbered
1011e834 8069 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8070 */
8071 if ((i & 1) == 0)
8072 insync = 2;
c2a1e7da 8073
c92a2527 8074 disk = get_imsm_disk(super, idx);
25ed7e59 8075 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8076 insync--;
c2a1e7da 8077
c92a2527
DW		 8078 /* no in-sync disks left in this mirror the
8079 * array has failed
8080 */
8081 if (insync == 0)
8082 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8083 }
8084
8085 return IMSM_T_STATE_DEGRADED;
8086 }
8087 case 5:
8088 if (failed < 2)
8089 return IMSM_T_STATE_DEGRADED;
8090 else
8091 return IMSM_T_STATE_FAILED;
8092 break;
8093 default:
8094 break;
8095 }
8096
8097 return map->map_state;
8098}
8099
3b451610
AK		 8100static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8101 int look_in_map)
c2a1e7da
DW		 8102{
8103 int i;
8104 int failed = 0;
8105 struct imsm_disk *disk;
d5985138
AK		 8106 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8107 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8108 struct imsm_map *map_for_loop;
0556e1a2
DW		 8109 __u32 ord;
8110 int idx;
d5985138 8111 int idx_1;
c2a1e7da 8112
0556e1a2
DW		 8113 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8114 * disks that are being rebuilt. New failures are recorded to
8115 * map[0]. So we look through all the disks we started with and
8116 * see if any failures are still present, or if any new ones
8117 * have arrived
0556e1a2 8118 */
d5985138
AK		 8119 map_for_loop = map;
8120 if (prev && (map->num_members < prev->num_members))
8121 map_for_loop = prev;
68fe4598
LD		 8122
8123 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8124 idx_1 = -255;
238c0a71
AK		 8125 /* when MAP_X is passed both maps failures are counted
8126 */
d5985138 8127 if (prev &&
089f9d79
JS		 8128 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8129 i < prev->num_members) {
d5985138
AK		 8130 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8131 idx_1 = ord_to_idx(ord);
c2a1e7da 8132
d5985138
AK		 8133 disk = get_imsm_disk(super, idx_1);
8134 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8135 failed++;
8136 }
089f9d79
JS		 8137 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8138 i < map->num_members) {
d5985138
AK		 8139 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8140 idx = ord_to_idx(ord);
8141
8142 if (idx != idx_1) {
8143 disk = get_imsm_disk(super, idx);
8144 if (!disk || is_failed(disk) ||
8145 ord & IMSM_ORD_REBUILD)
8146 failed++;
8147 }
8148 }
c2a1e7da
DW		 8149 }
8150
8151 return failed;
845dea95
NB		 8152}
8153
97b4d0e9
DW		 8154static int imsm_open_new(struct supertype *c, struct active_array *a,
8155 char *inst)
8156{
8157 struct intel_super *super = c->sb;
8158 struct imsm_super *mpb = super->anchor;
bbab0940 8159 struct imsm_update_prealloc_bb_mem u;
9587c373 8160
97b4d0e9 8161 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8162 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8163 return -ENODEV;
8164 }
8165
8166 dprintf("imsm: open_new %s\n", inst);
8167 a->info.container_member = atoi(inst);
bbab0940
TM		 8168
8169 u.type = update_prealloc_badblocks_mem;
8170 imsm_update_metadata_locally(c, &u, sizeof(u));
8171
97b4d0e9
DW		 8172 return 0;
8173}
8174
0c046afd
DW		 8175static int is_resyncing(struct imsm_dev *dev)
8176{
8177 struct imsm_map *migr_map;
8178
8179 if (!dev->vol.migr_state)
8180 return 0;
8181
1484e727
DW		 8182 if (migr_type(dev) == MIGR_INIT ||
8183 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8184 return 1;
8185
4c9bc37b
AK		 8186 if (migr_type(dev) == MIGR_GEN_MIGR)
8187 return 0;
8188
238c0a71 8189 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8190
089f9d79
JS		 8191 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8192 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8193 return 1;
8194 else
8195 return 0;
8196}
8197
0556e1a2 8198/* return true if we recorded new information */
4c9e8c1e
TM		 8199static int mark_failure(struct intel_super *super,
8200 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8201{
0556e1a2
DW		 8202 __u32 ord;
8203 int slot;
8204 struct imsm_map *map;
86c54047
DW		 8205 char buf[MAX_RAID_SERIAL_LEN+3];
8206 unsigned int len, shift = 0;
0556e1a2
DW		 8207
8208 /* new failures are always set in map[0] */
238c0a71 8209 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8210
8211 slot = get_imsm_disk_slot(map, idx);
8212 if (slot < 0)
8213 return 0;
8214
8215 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8216 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8217 return 0;
8218
7d0c5e24
LD		 8219 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8220 buf[MAX_RAID_SERIAL_LEN] = '\000';
8221 strcat(buf, ":0");
86c54047
DW		 8222 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8223 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8224 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8225
f2f27e63 8226 disk->status |= FAILED_DISK;
0556e1a2 8227 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8228 /* mark failures in second map if second map exists and this disk
8229 * in this slot.
8230 * This is valid for migration, initialization and rebuild
8231 */
8232 if (dev->vol.migr_state) {
238c0a71 8233 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8234 int slot2 = get_imsm_disk_slot(map2, idx);
8235
089f9d79 8236 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8237 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8238 idx | IMSM_ORD_REBUILD);
8239 }
d7a1fda2
MT		 8240 if (map->failed_disk_num == 0xff ||
8241 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8242 map->failed_disk_num = slot;
4c9e8c1e
TM		 8243
8244 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8245
0556e1a2
DW		 8246 return 1;
8247}
8248
4c9e8c1e
TM		 8249static void mark_missing(struct intel_super *super,
8250 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8251{
4c9e8c1e 8252 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8253
8254 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8255 return;
8256
47ee5a45
DW		 8257 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8258 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8259}
8260
33414a01
DW		 8261static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8262{
33414a01 8263 struct dl *dl;
33414a01
DW		 8264
8265 if (!super->missing)
8266 return;
33414a01 8267
79b68f1b
PC		 8268 /* When orom adds replacement for missing disk it does
8269 * not remove entry of missing disk, but just updates map with
8270 * new added disk. So it is not enough just to test if there is
8271 * any missing disk, we have to look if there are any failed disks
8272 * in map to stop migration */
8273
33414a01 8274 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8275 /* end process for initialization and rebuild only
8276 */
8277 if (is_gen_migration(dev) == 0) {
fb12a745 8278 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8279
fb12a745
TM		 8280 if (failed) {
8281 __u8 map_state;
8282 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8283 struct imsm_map *map1;
8284 int i, ord, ord_map1;
8285 int rebuilt = 1;
3d59f0c0 8286
fb12a745
TM		 8287 for (i = 0; i < map->num_members; i++) {
8288 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8289 if (!(ord & IMSM_ORD_REBUILD))
8290 continue;
8291
8292 map1 = get_imsm_map(dev, MAP_1);
8293 if (!map1)
8294 continue;
8295
8296 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8297 if (ord_map1 & IMSM_ORD_REBUILD)
8298 rebuilt = 0;
8299 }
8300
8301 if (rebuilt) {
8302 map_state = imsm_check_degraded(super, dev,
8303 failed, MAP_0);
8304 end_migration(dev, super, map_state);
8305 }
8306 }
3d59f0c0 8307 }
33414a01 8308 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8309 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8310 super->updates_pending++;
8311}
8312
f3871fdc
AK		 8313static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8314 long long new_size)
70bdf0dc 8315{
70bdf0dc 8316 unsigned long long array_blocks;
9529d343
MD		 8317 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8318 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8319
8320 if (used_disks == 0) {
8321 /* when problems occures
8322 * return current array_blocks value
8323 */
fcc2c9da 8324 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8325
8326 return array_blocks;
8327 }
8328
8329 /* set array size in metadata
8330 */
9529d343 8331 if (new_size <= 0)
f3871fdc
AK		 8332 /* OLCE size change is caused by added disks
8333 */
44490938 8334 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8335 else
f3871fdc
AK		 8336 /* Online Volume Size Change
8337 * Using available free space
8338 */
8339 array_blocks = new_size;
70bdf0dc 8340
b53bfba6 8341 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8342 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8343
8344 return array_blocks;
8345}
8346
28bce06f
AK		 8347static void imsm_set_disk(struct active_array *a, int n, int state);
8348
0e2d1a4e
AK		 8349static void imsm_progress_container_reshape(struct intel_super *super)
8350{
8351 /* if no device has a migr_state, but some device has a
8352 * different number of members than the previous device, start
8353 * changing the number of devices in this device to match
8354 * previous.
8355 */
8356 struct imsm_super *mpb = super->anchor;
8357 int prev_disks = -1;
8358 int i;
1dfaa380 8359 int copy_map_size;
0e2d1a4e
AK		 8360
8361 for (i = 0; i < mpb->num_raid_devs; i++) {
8362 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8363 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8364 struct imsm_map *map2;
8365 int prev_num_members;
0e2d1a4e
AK		 8366
8367 if (dev->vol.migr_state)
8368 return;
8369
8370 if (prev_disks == -1)
8371 prev_disks = map->num_members;
8372 if (prev_disks == map->num_members)
8373 continue;
8374
8375 /* OK, this array needs to enter reshape mode.
8376 * i.e it needs a migr_state
8377 */
8378
1dfaa380 8379 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8380 prev_num_members = map->num_members;
8381 map->num_members = prev_disks;
8382 dev->vol.migr_state = 1;
8383 dev->vol.curr_migr_unit = 0;
ea672ee1 8384 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8385 for (i = prev_num_members;
8386 i < map->num_members; i++)
8387 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8388 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8389 /* Copy the current map */
1dfaa380 8390 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8391 map2->num_members = prev_num_members;
8392
f3871fdc 8393 imsm_set_array_size(dev, -1);
51d83f5d 8394 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8395 super->updates_pending++;
8396 }
8397}
8398
aad6f216 8399/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8400 * states are handled in imsm_set_disk() with one exception, when a
8401 * resync is stopped due to a new failure this routine will set the
8402 * 'degraded' state for the array.
8403 */
01f157d7 8404static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8405{
8406 int inst = a->info.container_member;
8407 struct intel_super *super = a->container->sb;
949c47a0 8408 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8409 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8410 int failed = imsm_count_failed(super, dev, MAP_0);
8411 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8412 __u32 blocks_per_unit;
a862209d 8413
1af97990
AK		 8414 if (dev->vol.migr_state &&
8415 dev->vol.migr_type == MIGR_GEN_MIGR) {
8416 /* array state change is blocked due to reshape action
aad6f216
N		 8417 * We might need to
8418 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8419 * - finish the reshape (if last_checkpoint is big and action != reshape)
8420 * - update curr_migr_unit
1af97990 8421 */
aad6f216
N		 8422 if (a->curr_action == reshape) {
8423 /* still reshaping, maybe update curr_migr_unit */
633b5610 8424 goto mark_checkpoint;
aad6f216
N		 8425 } else {
8426 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8427 /* for some reason we aborted the reshape.
b66e591b
AK		 8428 *
8429 * disable automatic metadata rollback
8430 * user action is required to recover process
aad6f216 8431 */
b66e591b 8432 if (0) {
238c0a71
AK		 8433 struct imsm_map *map2 =
8434 get_imsm_map(dev, MAP_1);
8435 dev->vol.migr_state = 0;
8436 set_migr_type(dev, 0);
8437 dev->vol.curr_migr_unit = 0;
8438 memcpy(map, map2,
8439 sizeof_imsm_map(map2));
8440 super->updates_pending++;
b66e591b 8441 }
aad6f216
N		 8442 }
8443 if (a->last_checkpoint >= a->info.component_size) {
8444 unsigned long long array_blocks;
8445 int used_disks;
e154ced3 8446 struct mdinfo *mdi;
aad6f216 8447
9529d343 8448 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8449 if (used_disks > 0) {
8450 array_blocks =
44490938 8451 per_dev_array_size(map) *
d55adef9 8452 used_disks;
b53bfba6
TM		 8453 array_blocks =
8454 round_size_to_mb(array_blocks,
8455 used_disks);
d55adef9
AK		 8456 a->info.custom_array_size = array_blocks;
8457 /* encourage manager to update array
8458 * size
8459 */
e154ced3 8460
d55adef9 8461 a->check_reshape = 1;
633b5610 8462 }
e154ced3
AK		 8463 /* finalize online capacity expansion/reshape */
8464 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8465 imsm_set_disk(a,
8466 mdi->disk.raid_disk,
8467 mdi->curr_state);
8468
0e2d1a4e 8469 imsm_progress_container_reshape(super);
e154ced3 8470 }
aad6f216 8471 }
1af97990
AK		 8472 }
8473
47ee5a45 8474 /* before we activate this array handle any missing disks */
33414a01
DW		 8475 if (consistent == 2)
8476 handle_missing(super, dev);
1e5c6983 8477
0c046afd 8478 if (consistent == 2 &&
b7941fd6 8479 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8480 map_state != IMSM_T_STATE_NORMAL ||
8481 dev->vol.migr_state))
01f157d7 8482 consistent = 0;
272906ef 8483
b7941fd6 8484 if (is_resync_complete(&a->info)) {
0c046afd 8485 /* complete intialization / resync,
0556e1a2
DW		 8486 * recovery and interrupted recovery is completed in
8487 * ->set_disk
0c046afd
DW		 8488 */
8489 if (is_resyncing(dev)) {
8490 dprintf("imsm: mark resync done\n");
809da78e 8491 end_migration(dev, super, map_state);
115c3803 8492 super->updates_pending++;
484240d8 8493 a->last_checkpoint = 0;
115c3803 8494 }
b9172665
AK		 8495 } else if ((!is_resyncing(dev) && !failed) &&
8496 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8497 /* mark the start of the init process if nothing is failed */
b7941fd6 8498 dprintf("imsm: mark resync start\n");
1484e727 8499 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8500 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8501 else
8e59f3d8 8502 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8503 super->updates_pending++;
115c3803 8504 }
a862209d 8505
633b5610 8506mark_checkpoint:
5b83bacf
AK		 8507 /* skip checkpointing for general migration,
8508 * it is controlled in mdadm
8509 */
8510 if (is_gen_migration(dev))
8511 goto skip_mark_checkpoint;
8512
1e5c6983 8513 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8514 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8515 if (blocks_per_unit) {
1e5c6983
DW		 8516 __u32 units32;
8517 __u64 units;
8518
4f0a7acc 8519 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8520 units32 = units;
8521
8522 /* check that we did not overflow 32-bits, and that
8523 * curr_migr_unit needs updating
8524 */
8525 if (units32 == units &&
bfd80a56 8526 units32 != 0 &&
1e5c6983
DW		 8527 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8528 dprintf("imsm: mark checkpoint (%u)\n", units32);
8529 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8530 super->updates_pending++;
8531 }
8532 }
f8f603f1 8533
5b83bacf 8534skip_mark_checkpoint:
3393c6af 8535 /* mark dirty / clean */
2432ce9b
AP		 8536 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8537 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8538 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8539 if (consistent) {
8540 dev->vol.dirty = RAIDVOL_CLEAN;
8541 } else {
8542 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8543 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8544 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8545 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8546 }
a862209d
DW		 8547 super->updates_pending++;
8548 }
28bce06f 8549
01f157d7 8550 return consistent;
a862209d
DW		 8551}
8552
6f50473f
TM		 8553static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8554{
8555 int inst = a->info.container_member;
8556 struct intel_super *super = a->container->sb;
8557 struct imsm_dev *dev = get_imsm_dev(super, inst);
8558 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8559
8560 if (slot > map->num_members) {
8561 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8562 slot, map->num_members - 1);
8563 return -1;
8564 }
8565
8566 if (slot < 0)
8567 return -1;
8568
8569 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8570}
8571
8d45d196 8572static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8573{
8d45d196
DW		 8574 int inst = a->info.container_member;
8575 struct intel_super *super = a->container->sb;
949c47a0 8576 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8577 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8578 struct imsm_disk *disk;
7ce05701
LD		 8579 struct mdinfo *mdi;
8580 int recovery_not_finished = 0;
0c046afd 8581 int failed;
6f50473f 8582 int ord;
0c046afd 8583 __u8 map_state;
fb12a745
TM		 8584 int rebuild_done = 0;
8585 int i;
8d45d196 8586
fb12a745 8587 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8588 if (ord < 0)
8d45d196
DW		 8589 return;
8590
4e6e574a 8591 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8592 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8593
5802a811 8594 /* check for new failures */
ae7d61e3 8595 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8596 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8597 super->updates_pending++;
8d45d196 8598 }
47ee5a45 8599
19859edc 8600 /* check if in_sync */
0556e1a2 8601 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8602 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8603
8604 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8605 rebuild_done = 1;
19859edc
DW		 8606 super->updates_pending++;
8607 }
8d45d196 8608
3b451610
AK		 8609 failed = imsm_count_failed(super, dev, MAP_0);
8610 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8611
0c046afd 8612 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8613 dprintf("imsm: Detected transition to state ");
8614 switch (map_state) {
8615 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8616 dprintf("normal: ");
8617 if (is_rebuilding(dev)) {
1ade5cc1 8618 dprintf_cont("while rebuilding");
7ce05701
LD		 8619 /* check if recovery is really finished */
8620 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8621 if (mdi->recovery_start != MaxSector) {
8622 recovery_not_finished = 1;
8623 break;
8624 }
8625 if (recovery_not_finished) {
1ade5cc1
N		 8626 dprintf_cont("\n");
8627 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8628 if (a->last_checkpoint < mdi->recovery_start) {
8629 a->last_checkpoint = mdi->recovery_start;
8630 super->updates_pending++;
8631 }
8632 break;
8633 }
94002678 8634 end_migration(dev, super, map_state);
238c0a71 8635 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8636 map->failed_disk_num = ~0;
8637 super->updates_pending++;
8638 a->last_checkpoint = 0;
8639 break;
8640 }
8641 if (is_gen_migration(dev)) {
1ade5cc1 8642 dprintf_cont("while general migration");
bf2f0071 8643 if (a->last_checkpoint >= a->info.component_size)
809da78e 8644 end_migration(dev, super, map_state);
94002678
AK		 8645 else
8646 map->map_state = map_state;
238c0a71 8647 map = get_imsm_map(dev, MAP_0);
28bce06f 8648 map->failed_disk_num = ~0;
94002678 8649 super->updates_pending++;
bf2f0071 8650 break;
94002678
AK		 8651 }
8652 break;
8653 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8654 dprintf_cont("degraded: ");
089f9d79 8655 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8656 dprintf_cont("mark degraded");
94002678
AK		 8657 map->map_state = map_state;
8658 super->updates_pending++;
8659 a->last_checkpoint = 0;
8660 break;
8661 }
8662 if (is_rebuilding(dev)) {
d7a1fda2 8663 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8664 if (state & DS_FAULTY) {
8665 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8666 if (n == map->failed_disk_num) {
8667 dprintf_cont("end migration");
8668 end_migration(dev, super, map_state);
a4e96fd8 8669 a->last_checkpoint = 0;
d7a1fda2 8670 } else {
a4e96fd8 8671 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8672 map->map_state = map_state;
8673 }
94002678 8674 super->updates_pending++;
fb12a745
TM		 8675 }
8676
a4e96fd8
MT		 8677 if (!rebuild_done)
8678 break;
8679
fb12a745
TM		 8680 /* check if recovery is really finished */
8681 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8682 if (mdi->recovery_start != MaxSector) {
8683 recovery_not_finished = 1;
8684 break;
8685 }
8686 if (recovery_not_finished) {
8687 dprintf_cont("\n");
a4e96fd8 8688 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8689 if (a->last_checkpoint < mdi->recovery_start) {
8690 a->last_checkpoint =
8691 mdi->recovery_start;
8692 super->updates_pending++;
8693 }
8694 break;
94002678 8695 }
fb12a745
TM		 8696
8697 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8698 end_migration(dev, super, map_state);
8699 a->last_checkpoint = 0;
8700 super->updates_pending++;
fb12a745
TM		 8701
8702 for (i = 0; i < map->num_members; i++) {
8703 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8704
8705 if (idx & IMSM_ORD_REBUILD)
8706 map->failed_disk_num = i;
8707 }
8708 super->updates_pending++;
94002678
AK		 8709 break;
8710 }
8711 if (is_gen_migration(dev)) {
1ade5cc1 8712 dprintf_cont("while general migration");
bf2f0071 8713 if (a->last_checkpoint >= a->info.component_size)
809da78e 8714 end_migration(dev, super, map_state);
94002678
AK		 8715 else {
8716 map->map_state = map_state;
3b451610 8717 manage_second_map(super, dev);
94002678
AK		 8718 }
8719 super->updates_pending++;
bf2f0071 8720 break;
28bce06f 8721 }
6ce1fbf1 8722 if (is_initializing(dev)) {
1ade5cc1 8723 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8724 map->map_state = map_state;
8725 super->updates_pending++;
8726 break;
8727 }
94002678
AK		 8728 break;
8729 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8730 dprintf_cont("failed: ");
94002678 8731 if (is_gen_migration(dev)) {
1ade5cc1 8732 dprintf_cont("while general migration");
94002678
AK		 8733 map->map_state = map_state;
8734 super->updates_pending++;
8735 break;
8736 }
8737 if (map->map_state != map_state) {
1ade5cc1 8738 dprintf_cont("mark failed");
94002678
AK		 8739 end_migration(dev, super, map_state);
8740 super->updates_pending++;
8741 a->last_checkpoint = 0;
8742 break;
8743 }
8744 break;
8745 default:
1ade5cc1 8746 dprintf_cont("state %i\n", map_state);
5802a811 8747 }
1ade5cc1 8748 dprintf_cont("\n");
845dea95
NB		 8749}
8750
f796af5d 8751static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8752{
f796af5d 8753 void *buf = mpb;
c2a1e7da
DW		 8754 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8755 unsigned long long dsize;
8756 unsigned long long sectors;
f36a9ecd 8757 unsigned int sector_size;
c2a1e7da 8758
f36a9ecd 8759 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8760 get_dev_size(fd, NULL, &dsize);
8761
f36a9ecd 8762 if (mpb_size > sector_size) {
272f648f 8763 /* -1 to account for anchor */
f36a9ecd 8764 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8765
272f648f 8766 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8767 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8768 SEEK_SET) < 0)
272f648f 8769 return 1;
c2a1e7da 8770
f36a9ecd
PB		 8771 if ((unsigned long long)write(fd, buf + sector_size,
8772 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8773 return 1;
8774 }
c2a1e7da 8775
272f648f 8776 /* first block is stored on second to last sector of the disk */
f36a9ecd 8777 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8778 return 1;
8779
466070ad 8780 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8781 return 1;
8782
c2a1e7da
DW		 8783 return 0;
8784}
8785
2e735d19 8786static void imsm_sync_metadata(struct supertype *container)
845dea95 8787{
2e735d19 8788 struct intel_super *super = container->sb;
c2a1e7da 8789
1a64be56 8790 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8791 if (!super->updates_pending)
8792 return;
8793
36988a3d 8794 write_super_imsm(container, 0);
c2a1e7da
DW		 8795
8796 super->updates_pending = 0;
845dea95
NB		 8797}
8798
272906ef
DW		 8799static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8800{
8801 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8802 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8803 struct dl *dl;
8804
8805 for (dl = super->disks; dl; dl = dl->next)
8806 if (dl->index == i)
8807 break;
8808
25ed7e59 8809 if (dl && is_failed(&dl->disk))
272906ef
DW		 8810 dl = NULL;
8811
8812 if (dl)
1ade5cc1 8813 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8814
8815 return dl;
8816}
8817
a20d2ba5 8818static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8819 struct active_array *a, int activate_new,
8820 struct mdinfo *additional_test_list)
272906ef
DW		 8821{
8822 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8823 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8824 struct imsm_super *mpb = super->anchor;
8825 struct imsm_map *map;
272906ef
DW		 8826 unsigned long long pos;
8827 struct mdinfo *d;
8828 struct extent *ex;
a20d2ba5 8829 int i, j;
272906ef 8830 int found;
569cc43f
DW		 8831 __u32 array_start = 0;
8832 __u32 array_end = 0;
272906ef 8833 struct dl *dl;
6c932028 8834 struct mdinfo *test_list;
272906ef
DW		 8835
8836 for (dl = super->disks; dl; dl = dl->next) {
8837 /* If in this array, skip */
8838 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8839 if (d->state_fd >= 0 &&
8840 d->disk.major == dl->major &&
272906ef 8841 d->disk.minor == dl->minor) {
8ba77d32
AK		 8842 dprintf("%x:%x already in array\n",
8843 dl->major, dl->minor);
272906ef
DW		 8844 break;
8845 }
8846 if (d)
8847 continue;
6c932028
AK		 8848 test_list = additional_test_list;
8849 while (test_list) {
8850 if (test_list->disk.major == dl->major &&
8851 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8852 dprintf("%x:%x already in additional test list\n",
8853 dl->major, dl->minor);
8854 break;
8855 }
6c932028 8856 test_list = test_list->next;
8ba77d32 8857 }
6c932028 8858 if (test_list)
8ba77d32 8859 continue;
272906ef 8860
e553d2a4 8861 /* skip in use or failed drives */
25ed7e59 8862 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8863 dl->index == -2) {
8864 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8865 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8866 continue;
8867 }
8868
a20d2ba5
DW		 8869 /* skip pure spares when we are looking for partially
8870 * assimilated drives
8871 */
8872 if (dl->index == -1 && !activate_new)
8873 continue;
8874
f2cc4f7d
AO		 8875 if (!drive_validate_sector_size(super, dl))
8876 continue;
8877
272906ef 8878 /* Does this unused device have the requisite free space?
a20d2ba5 8879 * It needs to be able to cover all member volumes
272906ef 8880 */
05501181 8881 ex = get_extents(super, dl, 1);
272906ef
DW		 8882 if (!ex) {
8883 dprintf("cannot get extents\n");
8884 continue;
8885 }
a20d2ba5
DW		 8886 for (i = 0; i < mpb->num_raid_devs; i++) {
8887 dev = get_imsm_dev(super, i);
238c0a71 8888 map = get_imsm_map(dev, MAP_0);
272906ef 8889
a20d2ba5
DW		 8890 /* check if this disk is already a member of
8891 * this array
272906ef 8892 */
620b1713 8893 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8894 continue;
8895
8896 found = 0;
8897 j = 0;
8898 pos = 0;
5551b113 8899 array_start = pba_of_lba0(map);
329c8278 8900 array_end = array_start +
44490938 8901 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8902
8903 do {
8904 /* check that we can start at pba_of_lba0 with
44490938 8905 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8906 */
329c8278 8907 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8908 found = 1;
8909 break;
8910 }
8911 pos = ex[j].start + ex[j].size;
8912 j++;
8913 } while (ex[j-1].size);
8914
8915 if (!found)
272906ef 8916 break;
a20d2ba5 8917 }
272906ef
DW		 8918
8919 free(ex);
a20d2ba5 8920 if (i < mpb->num_raid_devs) {
329c8278
DW		 8921 dprintf("%x:%x does not have %u to %u available\n",
8922 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8923 /* No room */
8924 continue;
a20d2ba5
DW		 8925 }
8926 return dl;
272906ef
DW		 8927 }
8928
8929 return dl;
8930}
8931
95d07a2c
LM		 8932static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8933{
8934 struct imsm_dev *dev2;
8935 struct imsm_map *map;
8936 struct dl *idisk;
8937 int slot;
8938 int idx;
8939 __u8 state;
8940
8941 dev2 = get_imsm_dev(cont->sb, dev_idx);
8942 if (dev2) {
238c0a71 8943 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8944 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8945 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8946 if (!map)
8947 return 1;
8948 for (slot = 0; slot < map->num_members; slot++) {
8949 /*
8950 * Check if failed disks are deleted from intel
8951 * disk list or are marked to be deleted
8952 */
238c0a71 8953 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8954 idisk = get_imsm_dl_disk(cont->sb, idx);
8955 /*
8956 * Do not rebuild the array if failed disks
8957 * from failed sub-array are not removed from
8958 * container.
8959 */
8960 if (idisk &&
8961 is_failed(&idisk->disk) &&
8962 (idisk->action != DISK_REMOVE))
8963 return 0;
8964 }
8965 }
8966 }
8967 return 1;
8968}
8969
88758e9d
DW		 8970static struct mdinfo *imsm_activate_spare(struct active_array *a,
8971 struct metadata_update **updates)
8972{
8973 /**
d23fe947
DW		 8974 * Find a device with unused free space and use it to replace a
8975 * failed/vacant region in an array. We replace failed regions one a
8976 * array at a time. The result is that a new spare disk will be added
8977 * to the first failed array and after the monitor has finished
8978 * propagating failures the remainder will be consumed.
88758e9d 8979 *
d23fe947
DW		 8980 * FIXME add a capability for mdmon to request spares from another
8981 * container.
88758e9d
DW		 8982 */
8983
8984 struct intel_super *super = a->container->sb;
88758e9d 8985 int inst = a->info.container_member;
949c47a0 8986 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8987 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8988 int failed = a->info.array.raid_disks;
8989 struct mdinfo *rv = NULL;
8990 struct mdinfo *d;
8991 struct mdinfo *di;
8992 struct metadata_update *mu;
8993 struct dl *dl;
8994 struct imsm_update_activate_spare *u;
8995 int num_spares = 0;
8996 int i;
95d07a2c 8997 int allowed;
88758e9d
DW		 8998
8999 for (d = a->info.devs ; d ; d = d->next) {
9000 if ((d->curr_state & DS_FAULTY) &&
9001 d->state_fd >= 0)
9002 /* wait for Removal to happen */
9003 return NULL;
9004 if (d->state_fd >= 0)
9005 failed--;
9006 }
9007
9008 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
9009 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 9010
e2962bfc
AK		 9011 if (imsm_reshape_blocks_arrays_changes(super))
9012 return NULL;
1af97990 9013
fc8ca064
AK		 9014 /* Cannot activate another spare if rebuild is in progress already
9015 */
9016 if (is_rebuilding(dev)) {
7a862a02 9017 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 9018 return NULL;
9019 }
9020
89c67882
AK		 9021 if (a->info.array.level == 4)
9022 /* No repair for takeovered array
9023 * imsm doesn't support raid4
9024 */
9025 return NULL;
9026
3b451610
AK		 9027 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9028 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9029 return NULL;
9030
83ca7d45
AP		 9031 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9032 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9033 return NULL;
9034 }
9035
95d07a2c
LM		 9036 /*
9037 * If there are any failed disks check state of the other volume.
9038 * Block rebuild if the another one is failed until failed disks
9039 * are removed from container.
9040 */
9041 if (failed) {
7a862a02 9042 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9043 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9044 /* check if states of the other volumes allow for rebuild */
9045 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9046 if (i != inst) {
9047 allowed = imsm_rebuild_allowed(a->container,
9048 i, failed);
9049 if (!allowed)
9050 return NULL;
9051 }
9052 }
9053 }
9054
88758e9d 9055 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9056 for (i = 0; i < a->info.array.raid_disks; i++) {
9057 for (d = a->info.devs ; d ; d = d->next)
9058 if (d->disk.raid_disk == i)
9059 break;
9060 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9061 if (d && (d->state_fd >= 0))
9062 continue;
9063
272906ef 9064 /*
a20d2ba5
DW		 9065 * OK, this device needs recovery. Try to re-add the
9066 * previous occupant of this slot, if this fails see if
9067 * we can continue the assimilation of a spare that was
9068 * partially assimilated, finally try to activate a new
9069 * spare.
272906ef
DW		 9070 */
9071 dl = imsm_readd(super, i, a);
9072 if (!dl)
b303fe21 9073 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9074 if (!dl)
b303fe21 9075 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9076 if (!dl)
9077 continue;
1011e834 9078
272906ef 9079 /* found a usable disk with enough space */
503975b9 9080 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9081
9082 /* dl->index will be -1 in the case we are activating a
9083 * pristine spare. imsm_process_update() will create a
9084 * new index in this case. Once a disk is found to be
9085 * failed in all member arrays it is kicked from the
9086 * metadata
9087 */
9088 di->disk.number = dl->index;
d23fe947 9089
272906ef
DW		 9090 /* (ab)use di->devs to store a pointer to the device
9091 * we chose
9092 */
9093 di->devs = (struct mdinfo *) dl;
9094
9095 di->disk.raid_disk = i;
9096 di->disk.major = dl->major;
9097 di->disk.minor = dl->minor;
9098 di->disk.state = 0;
d23534e4 9099 di->recovery_start = 0;
5551b113 9100 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9101 di->component_size = a->info.component_size;
9102 di->container_member = inst;
5e46202e 9103 di->bb.supported = 1;
2c8890e9 9104 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9105 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9106 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9107 }
148acb7b 9108 super->random = random32();
272906ef
DW		 9109 di->next = rv;
9110 rv = di;
9111 num_spares++;
9112 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9113 i, di->data_offset);
88758e9d
DW		 9114 }
9115
9116 if (!rv)
9117 /* No spares found */
9118 return rv;
9119 /* Now 'rv' has a list of devices to return.
9120 * Create a metadata_update record to update the
9121 * disk_ord_tbl for the array
9122 */
503975b9 9123 mu = xmalloc(sizeof(*mu));
1011e834 9124 mu->buf = xcalloc(num_spares,
503975b9 9125 sizeof(struct imsm_update_activate_spare));
88758e9d 9126 mu->space = NULL;
cb23f1f4 9127 mu->space_list = NULL;
88758e9d
DW		 9128 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9129 mu->next = *updates;
9130 u = (struct imsm_update_activate_spare *) mu->buf;
9131
9132 for (di = rv ; di ; di = di->next) {
9133 u->type = update_activate_spare;
d23fe947
DW		 9134 u->dl = (struct dl *) di->devs;
9135 di->devs = NULL;
88758e9d
DW		 9136 u->slot = di->disk.raid_disk;
9137 u->array = inst;
9138 u->next = u + 1;
9139 u++;
9140 }
9141 (u-1)->next = NULL;
9142 *updates = mu;
9143
9144 return rv;
9145}
9146
54c2c1ea 9147static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9148{
54c2c1ea 9149 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9150 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9151 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9152 struct disk_info *inf = get_disk_info(u);
9153 struct imsm_disk *disk;
8273f55e
DW		 9154 int i;
9155 int j;
8273f55e 9156
54c2c1ea 9157 for (i = 0; i < map->num_members; i++) {
238c0a71 9158 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9159 for (j = 0; j < new_map->num_members; j++)
9160 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9161 return 1;
9162 }
9163
9164 return 0;
9165}
9166
1a64be56
LM		 9167static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9168{
594dc1b8
JS		 9169 struct dl *dl;
9170
1a64be56 9171 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9172 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9173 return dl;
9174 return NULL;
9175}
9176
9177static int remove_disk_super(struct intel_super *super, int major, int minor)
9178{
594dc1b8 9179 struct dl *prev;
1a64be56
LM		 9180 struct dl *dl;
9181
9182 prev = NULL;
9183 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9184 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9185 /* remove */
9186 if (prev)
9187 prev->next = dl->next;
9188 else
9189 super->disks = dl->next;
9190 dl->next = NULL;
9191 __free_imsm_disk(dl);
1ade5cc1 9192 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9193 break;
9194 }
9195 prev = dl;
9196 }
9197 return 0;
9198}
9199
f21e18ca 9200static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9201
1a64be56
LM		 9202static int add_remove_disk_update(struct intel_super *super)
9203{
9204 int check_degraded = 0;
594dc1b8
JS		 9205 struct dl *disk;
9206
1a64be56
LM		 9207 /* add/remove some spares to/from the metadata/contrainer */
9208 while (super->disk_mgmt_list) {
9209 struct dl *disk_cfg;
9210
9211 disk_cfg = super->disk_mgmt_list;
9212 super->disk_mgmt_list = disk_cfg->next;
9213 disk_cfg->next = NULL;
9214
9215 if (disk_cfg->action == DISK_ADD) {
9216 disk_cfg->next = super->disks;
9217 super->disks = disk_cfg;
9218 check_degraded = 1;
1ade5cc1
N		 9219 dprintf("added %x:%x\n",
9220 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9221 } else if (disk_cfg->action == DISK_REMOVE) {
9222 dprintf("Disk remove action processed: %x.%x\n",
9223 disk_cfg->major, disk_cfg->minor);
9224 disk = get_disk_super(super,
9225 disk_cfg->major,
9226 disk_cfg->minor);
9227 if (disk) {
9228 /* store action status */
9229 disk->action = DISK_REMOVE;
9230 /* remove spare disks only */
9231 if (disk->index == -1) {
9232 remove_disk_super(super,
9233 disk_cfg->major,
9234 disk_cfg->minor);
91c97c54
MT		 9235 } else {
9236 disk_cfg->fd = disk->fd;
9237 disk->fd = -1;
1a64be56
LM		 9238 }
9239 }
9240 /* release allocate disk structure */
9241 __free_imsm_disk(disk_cfg);
9242 }
9243 }
9244 return check_degraded;
9245}
9246
a29911da
PC		 9247static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9248 struct intel_super *super,
9249 void ***space_list)
9250{
9251 struct intel_dev *id;
9252 void **tofree = NULL;
9253 int ret_val = 0;
9254
1ade5cc1 9255 dprintf("(enter)\n");
089f9d79 9256 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9257 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9258 return ret_val;
9259 }
089f9d79 9260 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9261 dprintf("imsm: Error: Memory is not allocated\n");
9262 return ret_val;
9263 }
9264
9265 for (id = super->devlist ; id; id = id->next) {
9266 if (id->index == (unsigned)u->subdev) {
9267 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9268 struct imsm_map *map;
9269 struct imsm_dev *new_dev =
9270 (struct imsm_dev *)*space_list;
238c0a71 9271 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9272 int to_state;
9273 struct dl *new_disk;
9274
9275 if (new_dev == NULL)
9276 return ret_val;
9277 *space_list = **space_list;
9278 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9279 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9280 if (migr_map) {
9281 dprintf("imsm: Error: migration in progress");
9282 return ret_val;
9283 }
9284
9285 to_state = map->map_state;
9286 if ((u->new_level == 5) && (map->raid_level == 0)) {
9287 map->num_members++;
9288 /* this should not happen */
9289 if (u->new_disks[0] < 0) {
9290 map->failed_disk_num =
9291 map->num_members - 1;
9292 to_state = IMSM_T_STATE_DEGRADED;
9293 } else
9294 to_state = IMSM_T_STATE_NORMAL;
9295 }
8e59f3d8 9296 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9297 if (u->new_level > -1)
9298 map->raid_level = u->new_level;
238c0a71 9299 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9300 if ((u->new_level == 5) &&
9301 (migr_map->raid_level == 0)) {
9302 int ord = map->num_members - 1;
9303 migr_map->num_members--;
9304 if (u->new_disks[0] < 0)
9305 ord |= IMSM_ORD_REBUILD;
9306 set_imsm_ord_tbl_ent(map,
9307 map->num_members - 1,
9308 ord);
9309 }
9310 id->dev = new_dev;
9311 tofree = (void **)dev;
9312
4bba0439
PC		 9313 /* update chunk size
9314 */
06fb291a
PB		 9315 if (u->new_chunksize > 0) {
9316 unsigned long long num_data_stripes;
9529d343
MD		 9317 struct imsm_map *dest_map =
9318 get_imsm_map(dev, MAP_0);
06fb291a 9319 int used_disks =
9529d343 9320 imsm_num_data_members(dest_map);
06fb291a
PB		 9321
9322 if (used_disks == 0)
9323 return ret_val;
9324
4bba0439
PC		 9325 map->blocks_per_strip =
9326 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9327 num_data_stripes =
fcc2c9da 9328 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9329 num_data_stripes /= map->blocks_per_strip;
9330 num_data_stripes /= map->num_domains;
9331 set_num_data_stripes(map, num_data_stripes);
9332 }
4bba0439 9333
44490938
MD		 9334 /* ensure blocks_per_member has valid value
9335 */
9336 set_blocks_per_member(map,
9337 per_dev_array_size(map) +
9338 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9339
a29911da
PC		 9340 /* add disk
9341 */
089f9d79
JS		 9342 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9343 migr_map->raid_level == map->raid_level)
a29911da
PC		 9344 goto skip_disk_add;
9345
9346 if (u->new_disks[0] >= 0) {
9347 /* use passes spare
9348 */
9349 new_disk = get_disk_super(super,
9350 major(u->new_disks[0]),
9351 minor(u->new_disks[0]));
7a862a02 9352 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9353 major(u->new_disks[0]),
9354 minor(u->new_disks[0]),
9355 new_disk, new_disk->index);
9356 if (new_disk == NULL)
9357 goto error_disk_add;
9358
9359 new_disk->index = map->num_members - 1;
9360 /* slot to fill in autolayout
9361 */
9362 new_disk->raiddisk = new_disk->index;
9363 new_disk->disk.status |= CONFIGURED_DISK;
9364 new_disk->disk.status &= ~SPARE_DISK;
9365 } else
9366 goto error_disk_add;
9367
9368skip_disk_add:
9369 *tofree = *space_list;
9370 /* calculate new size
9371 */
f3871fdc 9372 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9373
9374 ret_val = 1;
9375 }
9376 }
9377
9378 if (tofree)
9379 *space_list = tofree;
9380 return ret_val;
9381
9382error_disk_add:
9383 dprintf("Error: imsm: Cannot find disk.\n");
9384 return ret_val;
9385}
9386
f3871fdc
AK		 9387static int apply_size_change_update(struct imsm_update_size_change *u,
9388 struct intel_super *super)
9389{
9390 struct intel_dev *id;
9391 int ret_val = 0;
9392
1ade5cc1 9393 dprintf("(enter)\n");
089f9d79 9394 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9395 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9396 return ret_val;
9397 }
9398
9399 for (id = super->devlist ; id; id = id->next) {
9400 if (id->index == (unsigned)u->subdev) {
9401 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9402 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9403 int used_disks = imsm_num_data_members(map);
f3871fdc 9404 unsigned long long blocks_per_member;
06fb291a 9405 unsigned long long num_data_stripes;
44490938
MD		 9406 unsigned long long new_size_per_disk;
9407
9408 if (used_disks == 0)
9409 return 0;
f3871fdc
AK		 9410
9411 /* calculate new size
9412 */
44490938
MD		 9413 new_size_per_disk = u->new_size / used_disks;
9414 blocks_per_member = new_size_per_disk +
9415 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9416 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9417 map->blocks_per_strip;
9418 num_data_stripes /= map->num_domains;
9419 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9420 u->new_size, new_size_per_disk,
06fb291a 9421 num_data_stripes);
f3871fdc 9422 set_blocks_per_member(map, blocks_per_member);
06fb291a 9423 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9424 imsm_set_array_size(dev, u->new_size);
9425
9426 ret_val = 1;
9427 break;
9428 }
9429 }
9430
9431 return ret_val;
9432}
9433
061d7da3 9434static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9435 struct intel_super *super,
061d7da3
LO		 9436 struct active_array *active_array)
9437{
9438 struct imsm_super *mpb = super->anchor;
9439 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9440 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9441 struct imsm_map *migr_map;
9442 struct active_array *a;
9443 struct imsm_disk *disk;
9444 __u8 to_state;
9445 struct dl *dl;
9446 unsigned int found;
9447 int failed;
5961eeec 9448 int victim;
061d7da3 9449 int i;
5961eeec 9450 int second_map_created = 0;
061d7da3 9451
5961eeec 9452 for (; u; u = u->next) {
238c0a71 9453 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9454
5961eeec 9455 if (victim < 0)
9456 return 0;
061d7da3 9457
5961eeec 9458 for (dl = super->disks; dl; dl = dl->next)
9459 if (dl == u->dl)
9460 break;
061d7da3 9461
5961eeec 9462 if (!dl) {
7a862a02 9463 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9464 u->dl->index);
9465 return 0;
9466 }
061d7da3 9467
5961eeec 9468 /* count failures (excluding rebuilds and the victim)
9469 * to determine map[0] state
9470 */
9471 failed = 0;
9472 for (i = 0; i < map->num_members; i++) {
9473 if (i == u->slot)
9474 continue;
9475 disk = get_imsm_disk(super,
238c0a71 9476 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9477 if (!disk || is_failed(disk))
9478 failed++;
9479 }
061d7da3 9480
5961eeec 9481 /* adding a pristine spare, assign a new index */
9482 if (dl->index < 0) {
9483 dl->index = super->anchor->num_disks;
9484 super->anchor->num_disks++;
9485 }
9486 disk = &dl->disk;
9487 disk->status |= CONFIGURED_DISK;
9488 disk->status &= ~SPARE_DISK;
9489
9490 /* mark rebuild */
238c0a71 9491 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9492 if (!second_map_created) {
9493 second_map_created = 1;
9494 map->map_state = IMSM_T_STATE_DEGRADED;
9495 migrate(dev, super, to_state, MIGR_REBUILD);
9496 } else
9497 map->map_state = to_state;
238c0a71 9498 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9499 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9500 set_imsm_ord_tbl_ent(migr_map, u->slot,
9501 dl->index | IMSM_ORD_REBUILD);
9502
9503 /* update the family_num to mark a new container
9504 * generation, being careful to record the existing
9505 * family_num in orig_family_num to clean up after
9506 * earlier mdadm versions that neglected to set it.
9507 */
9508 if (mpb->orig_family_num == 0)
9509 mpb->orig_family_num = mpb->family_num;
9510 mpb->family_num += super->random;
9511
9512 /* count arrays using the victim in the metadata */
9513 found = 0;
9514 for (a = active_array; a ; a = a->next) {
9515 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9516 map = get_imsm_map(dev, MAP_0);
061d7da3 9517
5961eeec 9518 if (get_imsm_disk_slot(map, victim) >= 0)
9519 found++;
9520 }
061d7da3 9521
5961eeec 9522 /* delete the victim if it is no longer being
9523 * utilized anywhere
061d7da3 9524 */
5961eeec 9525 if (!found) {
9526 struct dl **dlp;
061d7da3 9527
5961eeec 9528 /* We know that 'manager' isn't touching anything,
9529 * so it is safe to delete
9530 */
9531 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9532 if ((*dlp)->index == victim)
9533 break;
5961eeec 9534
9535 /* victim may be on the missing list */
9536 if (!*dlp)
9537 for (dlp = &super->missing; *dlp;
9538 dlp = &(*dlp)->next)
9539 if ((*dlp)->index == victim)
9540 break;
9541 imsm_delete(super, dlp, victim);
9542 }
061d7da3
LO		 9543 }
9544
9545 return 1;
9546}
a29911da 9547
2e5dc010
N		 9548static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9549 struct intel_super *super,
9550 void ***space_list)
9551{
9552 struct dl *new_disk;
9553 struct intel_dev *id;
9554 int i;
9555 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9556 int disk_count = u->old_raid_disks;
2e5dc010
N		 9557 void **tofree = NULL;
9558 int devices_to_reshape = 1;
9559 struct imsm_super *mpb = super->anchor;
9560 int ret_val = 0;
d098291a 9561 unsigned int dev_id;
2e5dc010 9562
1ade5cc1 9563 dprintf("(enter)\n");
2e5dc010
N		 9564
9565 /* enable spares to use in array */
9566 for (i = 0; i < delta_disks; i++) {
9567 new_disk = get_disk_super(super,
9568 major(u->new_disks[i]),
9569 minor(u->new_disks[i]));
7a862a02 9570 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9571 major(u->new_disks[i]), minor(u->new_disks[i]),
9572 new_disk, new_disk->index);
089f9d79
JS		 9573 if (new_disk == NULL ||
9574 (new_disk->index >= 0 &&
9575 new_disk->index < u->old_raid_disks))
2e5dc010 9576 goto update_reshape_exit;
ee4beede 9577 new_disk->index = disk_count++;
2e5dc010
N		 9578 /* slot to fill in autolayout
9579 */
9580 new_disk->raiddisk = new_disk->index;
9581 new_disk->disk.status |=
9582 CONFIGURED_DISK;
9583 new_disk->disk.status &= ~SPARE_DISK;
9584 }
9585
ed7333bd
AK		 9586 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9587 mpb->num_raid_devs);
2e5dc010
N		 9588 /* manage changes in volume
9589 */
d098291a 9590 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9591 void **sp = *space_list;
9592 struct imsm_dev *newdev;
9593 struct imsm_map *newmap, *oldmap;
9594
d098291a
AK		 9595 for (id = super->devlist ; id; id = id->next) {
9596 if (id->index == dev_id)
9597 break;
9598 }
9599 if (id == NULL)
9600 break;
2e5dc010
N		 9601 if (!sp)
9602 continue;
9603 *space_list = *sp;
9604 newdev = (void*)sp;
9605 /* Copy the dev, but not (all of) the map */
9606 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9607 oldmap = get_imsm_map(id->dev, MAP_0);
9608 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9609 /* Copy the current map */
9610 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9611 /* update one device only
9612 */
9613 if (devices_to_reshape) {
ed7333bd
AK		 9614 dprintf("imsm: modifying subdev: %i\n",
9615 id->index);
2e5dc010
N		 9616 devices_to_reshape--;
9617 newdev->vol.migr_state = 1;
9618 newdev->vol.curr_migr_unit = 0;
ea672ee1 9619 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9620 newmap->num_members = u->new_raid_disks;
9621 for (i = 0; i < delta_disks; i++) {
9622 set_imsm_ord_tbl_ent(newmap,
9623 u->old_raid_disks + i,
9624 u->old_raid_disks + i);
9625 }
9626 /* New map is correct, now need to save old map
9627 */
238c0a71 9628 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9629 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9630
f3871fdc 9631 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9632 }
9633
9634 sp = (void **)id->dev;
9635 id->dev = newdev;
9636 *sp = tofree;
9637 tofree = sp;
8e59f3d8
AK		 9638
9639 /* Clear migration record */
9640 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9641 }
819bc634
AK		 9642 if (tofree)
9643 *space_list = tofree;
2e5dc010
N		 9644 ret_val = 1;
9645
9646update_reshape_exit:
9647
9648 return ret_val;
9649}
9650
bb025c2f 9651static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9652 struct intel_super *super,
9653 void ***space_list)
bb025c2f
KW		 9654{
9655 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9656 struct intel_dev *dv;
9657 struct imsm_dev *dev_new;
bb025c2f
KW		 9658 struct imsm_map *map;
9659 struct dl *dm, *du;
8ca6df95 9660 int i;
bb025c2f
KW		 9661
9662 for (dv = super->devlist; dv; dv = dv->next)
9663 if (dv->index == (unsigned int)u->subarray) {
9664 dev = dv->dev;
9665 break;
9666 }
9667
9668 if (dev == NULL)
9669 return 0;
9670
238c0a71 9671 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9672
9673 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9674 unsigned long long num_data_stripes;
9675
43d5ec18 9676 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9677 if (imsm_count_failed(super, dev, MAP_0) !=
9678 (map->num_members / 2))
43d5ec18
KW		 9679 return 0;
9680
bb025c2f
KW		 9681 /* iterate through devices to mark removed disks as spare */
9682 for (dm = super->disks; dm; dm = dm->next) {
9683 if (dm->disk.status & FAILED_DISK) {
9684 int idx = dm->index;
9685 /* update indexes on the disk list */
9686/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9687 the index values will end up being correct.... NB */
9688 for (du = super->disks; du; du = du->next)
9689 if (du->index > idx)
9690 du->index--;
9691 /* mark as spare disk */
a8619d23 9692 mark_spare(dm);
bb025c2f
KW		 9693 }
9694 }
bb025c2f
KW		 9695 /* update map */
9696 map->num_members = map->num_members / 2;
9697 map->map_state = IMSM_T_STATE_NORMAL;
9698 map->num_domains = 1;
9699 map->raid_level = 0;
9700 map->failed_disk_num = -1;
4a353e6e
RS		 9701 num_data_stripes = imsm_dev_size(dev) / 2;
9702 num_data_stripes /= map->blocks_per_strip;
9703 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9704 }
9705
8ca6df95
KW		 9706 if (u->direction == R0_TO_R10) {
9707 void **space;
4a353e6e
RS		 9708 unsigned long long num_data_stripes;
9709
8ca6df95
KW		 9710 /* update slots in current disk list */
9711 for (dm = super->disks; dm; dm = dm->next) {
9712 if (dm->index >= 0)
9713 dm->index *= 2;
9714 }
9715 /* create new *missing* disks */
9716 for (i = 0; i < map->num_members; i++) {
9717 space = *space_list;
9718 if (!space)
9719 continue;
9720 *space_list = *space;
9721 du = (void *)space;
9722 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9723 du->fd = -1;
9724 du->minor = 0;
9725 du->major = 0;
9726 du->index = (i * 2) + 1;
9727 sprintf((char *)du->disk.serial,
9728 " MISSING_%d", du->index);
9729 sprintf((char *)du->serial,
9730 "MISSING_%d", du->index);
9731 du->next = super->missing;
9732 super->missing = du;
9733 }
9734 /* create new dev and map */
9735 space = *space_list;
9736 if (!space)
9737 return 0;
9738 *space_list = *space;
9739 dev_new = (void *)space;
9740 memcpy(dev_new, dev, sizeof(*dev));
9741 /* update new map */
238c0a71 9742 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9743 map->num_members = map->num_members * 2;
1a2487c2 9744 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9745 map->num_domains = 2;
9746 map->raid_level = 1;
4a353e6e
RS		 9747 num_data_stripes = imsm_dev_size(dev) / 2;
9748 num_data_stripes /= map->blocks_per_strip;
9749 num_data_stripes /= map->num_domains;
9750 set_num_data_stripes(map, num_data_stripes);
9751
8ca6df95
KW		 9752 /* replace dev<->dev_new */
9753 dv->dev = dev_new;
9754 }
bb025c2f
KW		 9755 /* update disk order table */
9756 for (du = super->disks; du; du = du->next)
9757 if (du->index >= 0)
9758 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9759 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9760 if (du->index >= 0) {
9761 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9762 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9763 }
bb025c2f
KW		 9764
9765 return 1;
9766}
9767
e8319a19
DW		 9768static void imsm_process_update(struct supertype *st,
9769 struct metadata_update *update)
9770{
9771 /**
9772 * crack open the metadata_update envelope to find the update record
9773 * update can be one of:
d195167d
AK		 9774 * update_reshape_container_disks - all the arrays in the container
9775 * are being reshaped to have more devices. We need to mark
9776 * the arrays for general migration and convert selected spares
9777 * into active devices.
9778 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9779 * device in an array, update the disk_ord_tbl. If this disk is
9780 * present in all member arrays then also clear the SPARE_DISK
9781 * flag
d195167d
AK		 9782 * update_create_array
9783 * update_kill_array
9784 * update_rename_array
9785 * update_add_remove_disk
e8319a19
DW		 9786 */
9787 struct intel_super *super = st->sb;
4d7b1503 9788 struct imsm_super *mpb;
e8319a19
DW		 9789 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9790
4d7b1503
DW		 9791 /* update requires a larger buf but the allocation failed */
9792 if (super->next_len && !super->next_buf) {
9793 super->next_len = 0;
9794 return;
9795 }
9796
9797 if (super->next_buf) {
9798 memcpy(super->next_buf, super->buf, super->len);
9799 free(super->buf);
9800 super->len = super->next_len;
9801 super->buf = super->next_buf;
9802
9803 super->next_len = 0;
9804 super->next_buf = NULL;
9805 }
9806
9807 mpb = super->anchor;
9808
e8319a19 9809 switch (type) {
0ec5d470
AK		 9810 case update_general_migration_checkpoint: {
9811 struct intel_dev *id;
9812 struct imsm_update_general_migration_checkpoint *u =
9813 (void *)update->buf;
9814
1ade5cc1 9815 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9816
9817 /* find device under general migration */
9818 for (id = super->devlist ; id; id = id->next) {
9819 if (is_gen_migration(id->dev)) {
9820 id->dev->vol.curr_migr_unit =
9821 __cpu_to_le32(u->curr_migr_unit);
9822 super->updates_pending++;
9823 }
9824 }
9825 break;
9826 }
bb025c2f
KW		 9827 case update_takeover: {
9828 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9829 if (apply_takeover_update(u, super, &update->space_list)) {
9830 imsm_update_version_info(super);
bb025c2f 9831 super->updates_pending++;
1a2487c2 9832 }
bb025c2f
KW		 9833 break;
9834 }
9835
78b10e66 9836 case update_reshape_container_disks: {
d195167d 9837 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9838 if (apply_reshape_container_disks_update(
9839 u, super, &update->space_list))
9840 super->updates_pending++;
78b10e66
N		 9841 break;
9842 }
48c5303a 9843 case update_reshape_migration: {
a29911da
PC		 9844 struct imsm_update_reshape_migration *u = (void *)update->buf;
9845 if (apply_reshape_migration_update(
9846 u, super, &update->space_list))
9847 super->updates_pending++;
48c5303a
PC		 9848 break;
9849 }
f3871fdc
AK		 9850 case update_size_change: {
9851 struct imsm_update_size_change *u = (void *)update->buf;
9852 if (apply_size_change_update(u, super))
9853 super->updates_pending++;
9854 break;
9855 }
e8319a19 9856 case update_activate_spare: {
1011e834 9857 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9858 if (apply_update_activate_spare(u, super, st->arrays))
9859 super->updates_pending++;
8273f55e
DW		 9860 break;
9861 }
9862 case update_create_array: {
9863 /* someone wants to create a new array, we need to be aware of
9864 * a few races/collisions:
9865 * 1/ 'Create' called by two separate instances of mdadm
9866 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9867 * devices that have since been assimilated via
9868 * activate_spare.
9869 * In the event this update can not be carried out mdadm will
9870 * (FIX ME) notice that its update did not take hold.
9871 */
9872 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9873 struct intel_dev *dv;
8273f55e
DW		 9874 struct imsm_dev *dev;
9875 struct imsm_map *map, *new_map;
9876 unsigned long long start, end;
9877 unsigned long long new_start, new_end;
9878 int i;
54c2c1ea
DW		 9879 struct disk_info *inf;
9880 struct dl *dl;
8273f55e
DW		 9881
9882 /* handle racing creates: first come first serve */
9883 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9884 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9885 goto create_error;
8273f55e
DW		 9886 }
9887
9888 /* check update is next in sequence */
9889 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9890 dprintf("can not create array %d expected index %d\n",
9891 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9892 goto create_error;
8273f55e
DW		 9893 }
9894
238c0a71 9895 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9896 new_start = pba_of_lba0(new_map);
44490938 9897 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9898 inf = get_disk_info(u);
8273f55e
DW		 9899
9900 /* handle activate_spare versus create race:
9901 * check to make sure that overlapping arrays do not include
9902 * overalpping disks
9903 */
9904 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9905 dev = get_imsm_dev(super, i);
238c0a71 9906 map = get_imsm_map(dev, MAP_0);
5551b113 9907 start = pba_of_lba0(map);
44490938 9908 end = start + per_dev_array_size(map);
8273f55e
DW		 9909 if ((new_start >= start && new_start <= end) ||
9910 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9911 /* overlap */;
9912 else
9913 continue;
9914
9915 if (disks_overlap(super, i, u)) {
1ade5cc1 9916 dprintf("arrays overlap\n");
ba2de7ba 9917 goto create_error;
8273f55e
DW		 9918 }
9919 }
8273f55e 9920
949c47a0
DW		 9921 /* check that prepare update was successful */
9922 if (!update->space) {
1ade5cc1 9923 dprintf("prepare update failed\n");
ba2de7ba 9924 goto create_error;
949c47a0
DW		 9925 }
9926
54c2c1ea
DW		 9927 /* check that all disks are still active before committing
9928 * changes. FIXME: could we instead handle this by creating a
9929 * degraded array? That's probably not what the user expects,
9930 * so better to drop this update on the floor.
9931 */
9932 for (i = 0; i < new_map->num_members; i++) {
9933 dl = serial_to_dl(inf[i].serial, super);
9934 if (!dl) {
1ade5cc1 9935 dprintf("disk disappeared\n");
ba2de7ba 9936 goto create_error;
54c2c1ea 9937 }
949c47a0
DW		 9938 }
9939
8273f55e 9940 super->updates_pending++;
54c2c1ea
DW		 9941
9942 /* convert spares to members and fixup ord_tbl */
9943 for (i = 0; i < new_map->num_members; i++) {
9944 dl = serial_to_dl(inf[i].serial, super);
9945 if (dl->index == -1) {
9946 dl->index = mpb->num_disks;
9947 mpb->num_disks++;
9948 dl->disk.status |= CONFIGURED_DISK;
9949 dl->disk.status &= ~SPARE_DISK;
9950 }
9951 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9952 }
9953
ba2de7ba
DW		 9954 dv = update->space;
9955 dev = dv->dev;
949c47a0
DW		 9956 update->space = NULL;
9957 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9958 dv->index = u->dev_idx;
9959 dv->next = super->devlist;
9960 super->devlist = dv;
8273f55e 9961 mpb->num_raid_devs++;
8273f55e 9962
4d1313e9 9963 imsm_update_version_info(super);
8273f55e 9964 break;
ba2de7ba
DW		 9965 create_error:
9966 /* mdmon knows how to release update->space, but not
9967 * ((struct intel_dev *) update->space)->dev
9968 */
9969 if (update->space) {
9970 dv = update->space;
9971 free(dv->dev);
9972 }
8273f55e 9973 break;
e8319a19 9974 }
33414a01
DW		 9975 case update_kill_array: {
9976 struct imsm_update_kill_array *u = (void *) update->buf;
9977 int victim = u->dev_idx;
9978 struct active_array *a;
9979 struct intel_dev **dp;
9980 struct imsm_dev *dev;
9981
9982 /* sanity check that we are not affecting the uuid of
9983 * active arrays, or deleting an active array
9984 *
9985 * FIXME when immutable ids are available, but note that
9986 * we'll also need to fixup the invalidated/active
9987 * subarray indexes in mdstat
9988 */
9989 for (a = st->arrays; a; a = a->next)
9990 if (a->info.container_member >= victim)
9991 break;
9992 /* by definition if mdmon is running at least one array
9993 * is active in the container, so checking
9994 * mpb->num_raid_devs is just extra paranoia
9995 */
9996 dev = get_imsm_dev(super, victim);
9997 if (a || !dev || mpb->num_raid_devs == 1) {
9998 dprintf("failed to delete subarray-%d\n", victim);
9999 break;
10000 }
10001
10002 for (dp = &super->devlist; *dp;)
f21e18ca 10003 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 10004 *dp = (*dp)->next;
10005 } else {
f21e18ca 10006 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 10007 (*dp)->index--;
10008 dp = &(*dp)->next;
10009 }
10010 mpb->num_raid_devs--;
10011 super->updates_pending++;
10012 break;
10013 }
aa534678
DW		 10014 case update_rename_array: {
10015 struct imsm_update_rename_array *u = (void *) update->buf;
10016 char name[MAX_RAID_SERIAL_LEN+1];
10017 int target = u->dev_idx;
10018 struct active_array *a;
10019 struct imsm_dev *dev;
10020
10021 /* sanity check that we are not affecting the uuid of
10022 * an active array
10023 */
40659392 10024 memset(name, 0, sizeof(name));
aa534678
DW		 10025 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
10026 name[MAX_RAID_SERIAL_LEN] = '\0';
10027 for (a = st->arrays; a; a = a->next)
10028 if (a->info.container_member == target)
10029 break;
10030 dev = get_imsm_dev(super, u->dev_idx);
10031 if (a || !dev || !check_name(super, name, 1)) {
10032 dprintf("failed to rename subarray-%d\n", target);
10033 break;
10034 }
10035
40659392 10036 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10037 super->updates_pending++;
10038 break;
10039 }
1a64be56 10040 case update_add_remove_disk: {
43dad3d6 10041 /* we may be able to repair some arrays if disks are
095b8088 10042 * being added, check the status of add_remove_disk
1a64be56
LM		 10043 * if discs has been added.
10044 */
10045 if (add_remove_disk_update(super)) {
43dad3d6 10046 struct active_array *a;
072b727f
DW		 10047
10048 super->updates_pending++;
1a64be56 10049 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10050 a->check_degraded = 1;
10051 }
43dad3d6 10052 break;
e8319a19 10053 }
bbab0940
TM		 10054 case update_prealloc_badblocks_mem:
10055 break;
e6e9dd3f
AP		 10056 case update_rwh_policy: {
10057 struct imsm_update_rwh_policy *u = (void *)update->buf;
10058 int target = u->dev_idx;
10059 struct imsm_dev *dev = get_imsm_dev(super, target);
10060 if (!dev) {
10061 dprintf("could not find subarray-%d\n", target);
10062 break;
10063 }
10064
10065 if (dev->rwh_policy != u->new_policy) {
10066 dev->rwh_policy = u->new_policy;
10067 super->updates_pending++;
10068 }
10069 break;
10070 }
1a64be56 10071 default:
ebf3be99 10072 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10073 }
e8319a19 10074}
88758e9d 10075
bc0b9d34
PC		 10076static struct mdinfo *get_spares_for_grow(struct supertype *st);
10077
5fe6f031
N		 10078static int imsm_prepare_update(struct supertype *st,
10079 struct metadata_update *update)
8273f55e 10080{
949c47a0 10081 /**
4d7b1503
DW		 10082 * Allocate space to hold new disk entries, raid-device entries or a new
10083 * mpb if necessary. The manager synchronously waits for updates to
10084 * complete in the monitor, so new mpb buffers allocated here can be
10085 * integrated by the monitor thread without worrying about live pointers
10086 * in the manager thread.
8273f55e 10087 */
095b8088 10088 enum imsm_update_type type;
4d7b1503 10089 struct intel_super *super = st->sb;
f36a9ecd 10090 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10091 struct imsm_super *mpb = super->anchor;
10092 size_t buf_len;
10093 size_t len = 0;
949c47a0 10094
095b8088
N		 10095 if (update->len < (int)sizeof(type))
10096 return 0;
10097
10098 type = *(enum imsm_update_type *) update->buf;
10099
949c47a0 10100 switch (type) {
0ec5d470 10101 case update_general_migration_checkpoint:
095b8088
N		 10102 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10103 return 0;
1ade5cc1 10104 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10105 break;
abedf5fc
KW		 10106 case update_takeover: {
10107 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10108 if (update->len < (int)sizeof(*u))
10109 return 0;
abedf5fc
KW		 10110 if (u->direction == R0_TO_R10) {
10111 void **tail = (void **)&update->space_list;
10112 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10113 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10114 int num_members = map->num_members;
10115 void *space;
10116 int size, i;
abedf5fc
KW		 10117 /* allocate memory for added disks */
10118 for (i = 0; i < num_members; i++) {
10119 size = sizeof(struct dl);
503975b9 10120 space = xmalloc(size);
abedf5fc
KW		 10121 *tail = space;
10122 tail = space;
10123 *tail = NULL;
10124 }
10125 /* allocate memory for new device */
10126 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10127 (num_members * sizeof(__u32));
503975b9
N		 10128 space = xmalloc(size);
10129 *tail = space;
10130 tail = space;
10131 *tail = NULL;
10132 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10133 }
10134
10135 break;
10136 }
78b10e66 10137 case update_reshape_container_disks: {
d195167d
AK		 10138 /* Every raid device in the container is about to
10139 * gain some more devices, and we will enter a
10140 * reconfiguration.
10141 * So each 'imsm_map' will be bigger, and the imsm_vol
10142 * will now hold 2 of them.
10143 * Thus we need new 'struct imsm_dev' allocations sized
10144 * as sizeof_imsm_dev but with more devices in both maps.
10145 */
10146 struct imsm_update_reshape *u = (void *)update->buf;
10147 struct intel_dev *dl;
10148 void **space_tail = (void**)&update->space_list;
10149
095b8088
N		 10150 if (update->len < (int)sizeof(*u))
10151 return 0;
10152
1ade5cc1 10153 dprintf("for update_reshape\n");
d195167d
AK		 10154
10155 for (dl = super->devlist; dl; dl = dl->next) {
10156 int size = sizeof_imsm_dev(dl->dev, 1);
10157 void *s;
d677e0b8
AK		 10158 if (u->new_raid_disks > u->old_raid_disks)
10159 size += sizeof(__u32)*2*
10160 (u->new_raid_disks - u->old_raid_disks);
503975b9 10161 s = xmalloc(size);
d195167d
AK		 10162 *space_tail = s;
10163 space_tail = s;
10164 *space_tail = NULL;
10165 }
10166
10167 len = disks_to_mpb_size(u->new_raid_disks);
10168 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10169 break;
10170 }
48c5303a 10171 case update_reshape_migration: {
bc0b9d34
PC		 10172 /* for migration level 0->5 we need to add disks
10173 * so the same as for container operation we will copy
10174 * device to the bigger location.
10175 * in memory prepared device and new disk area are prepared
10176 * for usage in process update
10177 */
10178 struct imsm_update_reshape_migration *u = (void *)update->buf;
10179 struct intel_dev *id;
10180 void **space_tail = (void **)&update->space_list;
10181 int size;
10182 void *s;
10183 int current_level = -1;
10184
095b8088
N		 10185 if (update->len < (int)sizeof(*u))
10186 return 0;
10187
1ade5cc1 10188 dprintf("for update_reshape\n");
bc0b9d34
PC		 10189
10190 /* add space for bigger array in update
10191 */
10192 for (id = super->devlist; id; id = id->next) {
10193 if (id->index == (unsigned)u->subdev) {
10194 size = sizeof_imsm_dev(id->dev, 1);
10195 if (u->new_raid_disks > u->old_raid_disks)
10196 size += sizeof(__u32)*2*
10197 (u->new_raid_disks - u->old_raid_disks);
503975b9 10198 s = xmalloc(size);
bc0b9d34
PC		 10199 *space_tail = s;
10200 space_tail = s;
10201 *space_tail = NULL;
10202 break;
10203 }
10204 }
10205 if (update->space_list == NULL)
10206 break;
10207
10208 /* add space for disk in update
10209 */
10210 size = sizeof(struct dl);
503975b9 10211 s = xmalloc(size);
bc0b9d34
PC		 10212 *space_tail = s;
10213 space_tail = s;
10214 *space_tail = NULL;
10215
10216 /* add spare device to update
10217 */
10218 for (id = super->devlist ; id; id = id->next)
10219 if (id->index == (unsigned)u->subdev) {
10220 struct imsm_dev *dev;
10221 struct imsm_map *map;
10222
10223 dev = get_imsm_dev(super, u->subdev);
238c0a71 10224 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10225 current_level = map->raid_level;
10226 break;
10227 }
089f9d79 10228 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10229 struct mdinfo *spares;
10230
10231 spares = get_spares_for_grow(st);
10232 if (spares) {
10233 struct dl *dl;
10234 struct mdinfo *dev;
10235
10236 dev = spares->devs;
10237 if (dev) {
10238 u->new_disks[0] =
10239 makedev(dev->disk.major,
10240 dev->disk.minor);
10241 dl = get_disk_super(super,
10242 dev->disk.major,
10243 dev->disk.minor);
10244 dl->index = u->old_raid_disks;
10245 dev = dev->next;
10246 }
10247 sysfs_free(spares);
10248 }
10249 }
10250 len = disks_to_mpb_size(u->new_raid_disks);
10251 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10252 break;
10253 }
f3871fdc 10254 case update_size_change: {
095b8088
N		 10255 if (update->len < (int)sizeof(struct imsm_update_size_change))
10256 return 0;
10257 break;
10258 }
10259 case update_activate_spare: {
10260 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10261 return 0;
f3871fdc
AK		 10262 break;
10263 }
949c47a0
DW		 10264 case update_create_array: {
10265 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10266 struct intel_dev *dv;
54c2c1ea 10267 struct imsm_dev *dev = &u->dev;
238c0a71 10268 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10269 struct dl *dl;
10270 struct disk_info *inf;
10271 int i;
10272 int activate = 0;
949c47a0 10273
095b8088
N		 10274 if (update->len < (int)sizeof(*u))
10275 return 0;
10276
54c2c1ea
DW		 10277 inf = get_disk_info(u);
10278 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10279 /* allocate a new super->devlist entry */
503975b9
N		 10280 dv = xmalloc(sizeof(*dv));
10281 dv->dev = xmalloc(len);
10282 update->space = dv;
949c47a0 10283
54c2c1ea
DW		 10284 /* count how many spares will be converted to members */
10285 for (i = 0; i < map->num_members; i++) {
10286 dl = serial_to_dl(inf[i].serial, super);
10287 if (!dl) {
10288 /* hmm maybe it failed?, nothing we can do about
10289 * it here
10290 */
10291 continue;
10292 }
10293 if (count_memberships(dl, super) == 0)
10294 activate++;
10295 }
10296 len += activate * sizeof(struct imsm_disk);
949c47a0 10297 break;
095b8088
N		 10298 }
10299 case update_kill_array: {
10300 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10301 return 0;
949c47a0
DW		 10302 break;
10303 }
095b8088
N		 10304 case update_rename_array: {
10305 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10306 return 0;
10307 break;
10308 }
10309 case update_add_remove_disk:
10310 /* no update->len needed */
10311 break;
bbab0940
TM		 10312 case update_prealloc_badblocks_mem:
10313 super->extra_space += sizeof(struct bbm_log) -
10314 get_imsm_bbm_log_size(super->bbm_log);
10315 break;
e6e9dd3f
AP		 10316 case update_rwh_policy: {
10317 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10318 return 0;
10319 break;
10320 }
095b8088
N		 10321 default:
10322 return 0;
949c47a0 10323 }
8273f55e 10324
4d7b1503
DW		 10325 /* check if we need a larger metadata buffer */
10326 if (super->next_buf)
10327 buf_len = super->next_len;
10328 else
10329 buf_len = super->len;
10330
bbab0940 10331 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10332 /* ok we need a larger buf than what is currently allocated
10333 * if this allocation fails process_update will notice that
10334 * ->next_len is set and ->next_buf is NULL
10335 */
bbab0940
TM		 10336 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10337 super->extra_space + len, sector_size);
4d7b1503
DW		 10338 if (super->next_buf)
10339 free(super->next_buf);
10340
10341 super->next_len = buf_len;
f36a9ecd 10342 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10343 memset(super->next_buf, 0, buf_len);
10344 else
4d7b1503
DW		 10345 super->next_buf = NULL;
10346 }
5fe6f031 10347 return 1;
8273f55e
DW		 10348}
10349
ae6aad82 10350/* must be called while manager is quiesced */
f21e18ca 10351static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10352{
10353 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10354 struct dl *iter;
10355 struct imsm_dev *dev;
10356 struct imsm_map *map;
4c9e8c1e 10357 unsigned int i, j, num_members;
fb12a745 10358 __u32 ord, ord_map0;
4c9e8c1e 10359 struct bbm_log *log = super->bbm_log;
ae6aad82 10360
1ade5cc1 10361 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10362
10363 /* shift all indexes down one */
10364 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10365 if (iter->index > (int)index)
ae6aad82 10366 iter->index--;
47ee5a45 10367 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10368 if (iter->index > (int)index)
47ee5a45 10369 iter->index--;
ae6aad82
DW		 10370
10371 for (i = 0; i < mpb->num_raid_devs; i++) {
10372 dev = get_imsm_dev(super, i);
238c0a71 10373 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10374 num_members = map->num_members;
10375 for (j = 0; j < num_members; j++) {
10376 /* update ord entries being careful not to propagate
10377 * ord-flags to the first map
10378 */
238c0a71 10379 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10380 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10381
24565c9a
DW		 10382 if (ord_to_idx(ord) <= index)
10383 continue;
ae6aad82 10384
238c0a71 10385 map = get_imsm_map(dev, MAP_0);
fb12a745 10386 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10387 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10388 if (map)
10389 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10390 }
10391 }
10392
4c9e8c1e
TM		 10393 for (i = 0; i < log->entry_count; i++) {
10394 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10395
10396 if (entry->disk_ordinal <= index)
10397 continue;
10398 entry->disk_ordinal--;
10399 }
10400
ae6aad82
DW		 10401 mpb->num_disks--;
10402 super->updates_pending++;
24565c9a
DW		 10403 if (*dlp) {
10404 struct dl *dl = *dlp;
10405
10406 *dlp = (*dlp)->next;
10407 __free_imsm_disk(dl);
10408 }
ae6aad82 10409}
9a717282
AK		 10410
10411static void close_targets(int *targets, int new_disks)
10412{
10413 int i;
10414
10415 if (!targets)
10416 return;
10417
10418 for (i = 0; i < new_disks; i++) {
10419 if (targets[i] >= 0) {
10420 close(targets[i]);
10421 targets[i] = -1;
10422 }
10423 }
10424}
10425
10426static int imsm_get_allowed_degradation(int level, int raid_disks,
10427 struct intel_super *super,
10428 struct imsm_dev *dev)
10429{
10430 switch (level) {
bf5cf7c7 10431 case 1:
9a717282
AK		 10432 case 10:{
10433 int ret_val = 0;
10434 struct imsm_map *map;
10435 int i;
10436
10437 ret_val = raid_disks/2;
10438 /* check map if all disks pairs not failed
10439 * in both maps
10440 */
238c0a71 10441 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10442 for (i = 0; i < ret_val; i++) {
10443 int degradation = 0;
10444 if (get_imsm_disk(super, i) == NULL)
10445 degradation++;
10446 if (get_imsm_disk(super, i + 1) == NULL)
10447 degradation++;
10448 if (degradation == 2)
10449 return 0;
10450 }
238c0a71 10451 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10452 /* if there is no second map
10453 * result can be returned
10454 */
10455 if (map == NULL)
10456 return ret_val;
10457 /* check degradation in second map
10458 */
10459 for (i = 0; i < ret_val; i++) {
10460 int degradation = 0;
10461 if (get_imsm_disk(super, i) == NULL)
10462 degradation++;
10463 if (get_imsm_disk(super, i + 1) == NULL)
10464 degradation++;
10465 if (degradation == 2)
10466 return 0;
10467 }
10468 return ret_val;
10469 }
10470 case 5:
10471 return 1;
10472 case 6:
10473 return 2;
10474 default:
10475 return 0;
10476 }
10477}
10478
687629c2
AK		 10479/*******************************************************************************
10480 * Function: open_backup_targets
10481 * Description: Function opens file descriptors for all devices given in
10482 * info->devs
10483 * Parameters:
10484 * info : general array info
10485 * raid_disks : number of disks
10486 * raid_fds : table of device's file descriptors
9a717282
AK		 10487 * super : intel super for raid10 degradation check
10488 * dev : intel device for raid10 degradation check
687629c2
AK		 10489 * Returns:
10490 * 0 : success
10491 * -1 : fail
10492 ******************************************************************************/
9a717282
AK		 10493int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10494 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10495{
10496 struct mdinfo *sd;
f627f5ad 10497 int i;
9a717282 10498 int opened = 0;
f627f5ad
AK		 10499
10500 for (i = 0; i < raid_disks; i++)
10501 raid_fds[i] = -1;
687629c2
AK		 10502
10503 for (sd = info->devs ; sd ; sd = sd->next) {
10504 char *dn;
10505
10506 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10507 dprintf("disk is faulty!!\n");
10508 continue;
10509 }
10510
089f9d79 10511 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10512 continue;
10513
10514 dn = map_dev(sd->disk.major,
10515 sd->disk.minor, 1);
10516 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10517 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10518 pr_err("cannot open component\n");
9a717282 10519 continue;
687629c2 10520 }
9a717282
AK		 10521 opened++;
10522 }
10523 /* check if maximum array degradation level is not exceeded
10524 */
10525 if ((raid_disks - opened) >
089f9d79
JS		 10526 imsm_get_allowed_degradation(info->new_level, raid_disks,
10527 super, dev)) {
e12b3daa 10528 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10529 close_targets(raid_fds, raid_disks);
10530 return -2;
687629c2
AK		 10531 }
10532 return 0;
10533}
10534
d31ad643
PB		 10535/*******************************************************************************
10536 * Function: validate_container_imsm
10537 * Description: This routine validates container after assemble,
10538 * eg. if devices in container are under the same controller.
10539 *
10540 * Parameters:
10541 * info : linked list with info about devices used in array
10542 * Returns:
10543 * 1 : HBA mismatch
10544 * 0 : Success
10545 ******************************************************************************/
10546int validate_container_imsm(struct mdinfo *info)
10547{
6b781d33
AP		 10548 if (check_env("IMSM_NO_PLATFORM"))
10549 return 0;
d31ad643 10550
6b781d33
AP		 10551 struct sys_dev *idev;
10552 struct sys_dev *hba = NULL;
10553 struct sys_dev *intel_devices = find_intel_devices();
10554 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10555 info->disk.minor));
10556
10557 for (idev = intel_devices; idev; idev = idev->next) {
10558 if (dev_path && strstr(dev_path, idev->path)) {
10559 hba = idev;
10560 break;
d31ad643 10561 }
6b781d33
AP		 10562 }
10563 if (dev_path)
d31ad643
PB		 10564 free(dev_path);
10565
6b781d33
AP		 10566 if (!hba) {
10567 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10568 devid2kname(makedev(info->disk.major, info->disk.minor)));
10569 return 1;
10570 }
10571
10572 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10573 struct mdinfo *dev;
10574
10575 for (dev = info->next; dev; dev = dev->next) {
10576 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10577
10578 struct sys_dev *hba2 = NULL;
10579 for (idev = intel_devices; idev; idev = idev->next) {
10580 if (dev_path && strstr(dev_path, idev->path)) {
10581 hba2 = idev;
10582 break;
d31ad643
PB		 10583 }
10584 }
6b781d33
AP		 10585 if (dev_path)
10586 free(dev_path);
10587
10588 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10589 get_orom_by_device_id(hba2->dev_id);
10590
10591 if (hba2 && hba->type != hba2->type) {
10592 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10593 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10594 return 1;
10595 }
10596
07cb1e57 10597 if (orom != orom2) {
6b781d33
AP		 10598 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10599 " This operation is not supported and can lead to data loss.\n");
10600 return 1;
10601 }
10602
10603 if (!orom) {
10604 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10605 " This operation is not supported and can lead to data loss.\n");
10606 return 1;
10607 }
d31ad643 10608 }
6b781d33 10609
d31ad643
PB		 10610 return 0;
10611}
32141c17 10612
6f50473f
TM		 10613/*******************************************************************************
10614* Function: imsm_record_badblock
10615* Description: This routine stores new bad block record in BBM log
10616*
10617* Parameters:
10618* a : array containing a bad block
10619* slot : disk number containing a bad block
10620* sector : bad block sector
10621* length : bad block sectors range
10622* Returns:
10623* 1 : Success
10624* 0 : Error
10625******************************************************************************/
10626static int imsm_record_badblock(struct active_array *a, int slot,
10627 unsigned long long sector, int length)
10628{
10629 struct intel_super *super = a->container->sb;
10630 int ord;
10631 int ret;
10632
10633 ord = imsm_disk_slot_to_ord(a, slot);
10634 if (ord < 0)
10635 return 0;
10636
10637 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10638 length);
10639 if (ret)
10640 super->updates_pending++;
10641
10642 return ret;
10643}
c07a5a4f
TM		 10644/*******************************************************************************
10645* Function: imsm_clear_badblock
10646* Description: This routine clears bad block record from BBM log
10647*
10648* Parameters:
10649* a : array containing a bad block
10650* slot : disk number containing a bad block
10651* sector : bad block sector
10652* length : bad block sectors range
10653* Returns:
10654* 1 : Success
10655* 0 : Error
10656******************************************************************************/
10657static int imsm_clear_badblock(struct active_array *a, int slot,
10658 unsigned long long sector, int length)
10659{
10660 struct intel_super *super = a->container->sb;
10661 int ord;
10662 int ret;
10663
10664 ord = imsm_disk_slot_to_ord(a, slot);
10665 if (ord < 0)
10666 return 0;
10667
10668 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10669 if (ret)
10670 super->updates_pending++;
10671
10672 return ret;
10673}
928f1424
TM		 10674/*******************************************************************************
10675* Function: imsm_get_badblocks
10676* Description: This routine get list of bad blocks for an array
10677*
10678* Parameters:
10679* a : array
10680* slot : disk number
10681* Returns:
10682* bb : structure containing bad blocks
10683* NULL : error
10684******************************************************************************/
10685static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10686{
10687 int inst = a->info.container_member;
10688 struct intel_super *super = a->container->sb;
10689 struct imsm_dev *dev = get_imsm_dev(super, inst);
10690 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10691 int ord;
10692
10693 ord = imsm_disk_slot_to_ord(a, slot);
10694 if (ord < 0)
10695 return NULL;
10696
10697 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10698 per_dev_array_size(map), &super->bb);
928f1424
TM		 10699
10700 return &super->bb;
10701}
27156a57
TM		 10702/*******************************************************************************
10703* Function: examine_badblocks_imsm
10704* Description: Prints list of bad blocks on a disk to the standard output
10705*
10706* Parameters:
10707* st : metadata handler
10708* fd : open file descriptor for device
10709* devname : device name
10710* Returns:
10711* 0 : Success
10712* 1 : Error
10713******************************************************************************/
10714static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10715{
10716 struct intel_super *super = st->sb;
10717 struct bbm_log *log = super->bbm_log;
10718 struct dl *d = NULL;
10719 int any = 0;
10720
10721 for (d = super->disks; d ; d = d->next) {
10722 if (strcmp(d->devname, devname) == 0)
10723 break;
10724 }
10725
10726 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10727 pr_err("%s doesn't appear to be part of a raid array\n",
10728 devname);
10729 return 1;
10730 }
10731
10732 if (log != NULL) {
10733 unsigned int i;
10734 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10735
10736 for (i = 0; i < log->entry_count; i++) {
10737 if (entry[i].disk_ordinal == d->index) {
10738 unsigned long long sector = __le48_to_cpu(
10739 &entry[i].defective_block_start);
10740 int cnt = entry[i].marked_count + 1;
10741
10742 if (!any) {
10743 printf("Bad-blocks on %s:\n", devname);
10744 any = 1;
10745 }
10746
10747 printf("%20llu for %d sectors\n", sector, cnt);
10748 }
10749 }
10750 }
10751
10752 if (!any)
10753 printf("No bad-blocks list configured on %s\n", devname);
10754
10755 return 0;
10756}
687629c2
AK		 10757/*******************************************************************************
10758 * Function: init_migr_record_imsm
10759 * Description: Function inits imsm migration record
10760 * Parameters:
10761 * super : imsm internal array info
10762 * dev : device under migration
10763 * info : general array info to find the smallest device
10764 * Returns:
10765 * none
10766 ******************************************************************************/
10767void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10768 struct mdinfo *info)
10769{
10770 struct intel_super *super = st->sb;
10771 struct migr_record *migr_rec = super->migr_rec;
10772 int new_data_disks;
10773 unsigned long long dsize, dev_sectors;
10774 long long unsigned min_dev_sectors = -1LLU;
10775 struct mdinfo *sd;
10776 char nm[30];
10777 int fd;
238c0a71
AK		 10778 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10779 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10780 unsigned long long num_migr_units;
3ef4403c 10781 unsigned long long array_blocks;
687629c2
AK		 10782
10783 memset(migr_rec, 0, sizeof(struct migr_record));
10784 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10785
10786 /* only ascending reshape supported now */
10787 migr_rec->ascending_migr = __cpu_to_le32(1);
10788
10789 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10790 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10791 migr_rec->dest_depth_per_unit *=
10792 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10793 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10794 migr_rec->blocks_per_unit =
10795 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10796 migr_rec->dest_depth_per_unit =
10797 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10798 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10799 num_migr_units =
10800 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10801
10802 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10803 num_migr_units++;
9f421827 10804 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10805
10806 migr_rec->post_migr_vol_cap = dev->size_low;
10807 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10808
687629c2
AK		 10809 /* Find the smallest dev */
10810 for (sd = info->devs ; sd ; sd = sd->next) {
10811 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10812 fd = dev_open(nm, O_RDONLY);
10813 if (fd < 0)
10814 continue;
10815 get_dev_size(fd, NULL, &dsize);
10816 dev_sectors = dsize / 512;
10817 if (dev_sectors < min_dev_sectors)
10818 min_dev_sectors = dev_sectors;
10819 close(fd);
10820 }
9f421827 10821 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10822 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10823
10824 write_imsm_migr_rec(st);
10825
10826 return;
10827}
10828
10829/*******************************************************************************
10830 * Function: save_backup_imsm
10831 * Description: Function saves critical data stripes to Migration Copy Area
10832 * and updates the current migration unit status.
10833 * Use restore_stripes() to form a destination stripe,
10834 * and to write it to the Copy Area.
10835 * Parameters:
10836 * st : supertype information
aea93171 10837 * dev : imsm device that backup is saved for
687629c2
AK		 10838 * info : general array info
10839 * buf : input buffer
687629c2
AK		 10840 * length : length of data to backup (blocks_per_unit)
10841 * Returns:
10842 * 0 : success
10843 *, -1 : fail
10844 ******************************************************************************/
10845int save_backup_imsm(struct supertype *st,
10846 struct imsm_dev *dev,
10847 struct mdinfo *info,
10848 void *buf,
687629c2
AK		 10849 int length)
10850{
10851 int rv = -1;
10852 struct intel_super *super = st->sb;
594dc1b8
JS		 10853 unsigned long long *target_offsets;
10854 int *targets;
687629c2 10855 int i;
238c0a71 10856 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10857 int new_disks = map_dest->num_members;
ab724b98
AK		 10858 int dest_layout = 0;
10859 int dest_chunk;
d1877f69 10860 unsigned long long start;
9529d343 10861 int data_disks = imsm_num_data_members(map_dest);
687629c2 10862
503975b9 10863 targets = xmalloc(new_disks * sizeof(int));
687629c2 10864
7e45b550
AK		 10865 for (i = 0; i < new_disks; i++)
10866 targets[i] = -1;
10867
503975b9 10868 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10869
d1877f69 10870 start = info->reshape_progress * 512;
687629c2 10871 for (i = 0; i < new_disks; i++) {
9f421827 10872 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10873 /* move back copy area adderss, it will be moved forward
10874 * in restore_stripes() using start input variable
10875 */
10876 target_offsets[i] -= start/data_disks;
687629c2
AK		 10877 }
10878
9a717282
AK		 10879 if (open_backup_targets(info, new_disks, targets,
10880 super, dev))
687629c2
AK		 10881 goto abort;
10882
68eb8bc6 10883 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10884 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10885
687629c2
AK		 10886 if (restore_stripes(targets, /* list of dest devices */
10887 target_offsets, /* migration record offsets */
10888 new_disks,
ab724b98
AK		 10889 dest_chunk,
10890 map_dest->raid_level,
10891 dest_layout,
10892 -1, /* source backup file descriptor */
10893 0, /* input buf offset
10894 * always 0 buf is already offseted */
d1877f69 10895 start,
687629c2
AK		 10896 length,
10897 buf) != 0) {
e7b84f9d 10898 pr_err("Error restoring stripes\n");
687629c2
AK		 10899 goto abort;
10900 }
10901
10902 rv = 0;
10903
10904abort:
10905 if (targets) {
9a717282 10906 close_targets(targets, new_disks);
687629c2
AK		 10907 free(targets);
10908 }
10909 free(target_offsets);
10910
10911 return rv;
10912}
10913
10914/*******************************************************************************
10915 * Function: save_checkpoint_imsm
10916 * Description: Function called for current unit status update
10917 * in the migration record. It writes it to disk.
10918 * Parameters:
10919 * super : imsm internal array info
10920 * info : general array info
10921 * Returns:
10922 * 0: success
10923 * 1: failure
0228d92c
AK		 10924 * 2: failure, means no valid migration record
10925 * / no general migration in progress /
687629c2
AK		 10926 ******************************************************************************/
10927int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10928{
10929 struct intel_super *super = st->sb;
f8b72ef5
AK		 10930 unsigned long long blocks_per_unit;
10931 unsigned long long curr_migr_unit;
10932
2e062e82 10933 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10934 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10935 return 1;
10936 }
10937
f8b72ef5
AK		 10938 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10939 if (blocks_per_unit == 0) {
0228d92c
AK		 10940 dprintf("imsm: no migration in progress.\n");
10941 return 2;
687629c2 10942 }
f8b72ef5
AK		 10943 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10944 /* check if array is alligned to copy area
10945 * if it is not alligned, add one to current migration unit value
10946 * this can happend on array reshape finish only
10947 */
10948 if (info->reshape_progress % blocks_per_unit)
10949 curr_migr_unit++;
687629c2 10950
9f421827 10951 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10952 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10953 set_migr_dest_1st_member_lba(super->migr_rec,
10954 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10955
687629c2 10956 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10957 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10958 return 1;
10959 }
10960
10961 return 0;
10962}
10963
276d77db
AK		 10964/*******************************************************************************
10965 * Function: recover_backup_imsm
10966 * Description: Function recovers critical data from the Migration Copy Area
10967 * while assembling an array.
10968 * Parameters:
10969 * super : imsm internal array info
10970 * info : general array info
10971 * Returns:
10972 * 0 : success (or there is no data to recover)
10973 * 1 : fail
10974 ******************************************************************************/
10975int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10976{
10977 struct intel_super *super = st->sb;
10978 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10979 struct imsm_map *map_dest;
276d77db
AK		 10980 struct intel_dev *id = NULL;
10981 unsigned long long read_offset;
10982 unsigned long long write_offset;
10983 unsigned unit_len;
10984 int *targets = NULL;
10985 int new_disks, i, err;
10986 char *buf = NULL;
10987 int retval = 1;
f36a9ecd 10988 unsigned int sector_size = super->sector_size;
9f421827
PB		 10989 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10990 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10991 char buffer[20];
6c3560c0 10992 int skipped_disks = 0;
276d77db
AK		 10993
10994 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10995 if (err < 1)
10996 return 1;
10997
10998 /* recover data only during assemblation */
10999 if (strncmp(buffer, "inactive", 8) != 0)
11000 return 0;
11001 /* no data to recover */
11002 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
11003 return 0;
11004 if (curr_migr_unit >= num_migr_units)
11005 return 1;
11006
11007 /* find device during reshape */
11008 for (id = super->devlist; id; id = id->next)
11009 if (is_gen_migration(id->dev))
11010 break;
11011 if (id == NULL)
11012 return 1;
11013
238c0a71 11014 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 11015 new_disks = map_dest->num_members;
11016
9f421827 11017 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 11018
9f421827 11019 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 11020 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 11021
11022 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 11023 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 11024 goto abort;
503975b9 11025 targets = xcalloc(new_disks, sizeof(int));
276d77db 11026
9a717282 11027 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 11028 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 11029 goto abort;
11030 }
276d77db
AK		 11031
11032 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 11033 if (targets[i] < 0) {
11034 skipped_disks++;
11035 continue;
11036 }
276d77db 11037 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11038 pr_err("Cannot seek to block: %s\n",
11039 strerror(errno));
137debce
AK		 11040 skipped_disks++;
11041 continue;
276d77db 11042 }
9ec11d1a 11043 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11044 pr_err("Cannot read copy area block: %s\n",
11045 strerror(errno));
137debce
AK		 11046 skipped_disks++;
11047 continue;
276d77db
AK		 11048 }
11049 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11050 pr_err("Cannot seek to block: %s\n",
11051 strerror(errno));
137debce
AK		 11052 skipped_disks++;
11053 continue;
276d77db 11054 }
9ec11d1a 11055 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11056 pr_err("Cannot restore block: %s\n",
11057 strerror(errno));
137debce
AK		 11058 skipped_disks++;
11059 continue;
276d77db
AK		 11060 }
11061 }
11062
137debce
AK		 11063 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11064 new_disks,
11065 super,
11066 id->dev)) {
7a862a02 11067 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11068 goto abort;
11069 }
11070
befb629b
AK		 11071 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11072 /* ignore error == 2, this can mean end of reshape here
11073 */
7a862a02 11074 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11075 } else
276d77db 11076 retval = 0;
276d77db
AK		 11077
11078abort:
11079 if (targets) {
11080 for (i = 0; i < new_disks; i++)
11081 if (targets[i])
11082 close(targets[i]);
11083 free(targets);
11084 }
11085 free(buf);
11086 return retval;
11087}
11088
2cda7640
ML		 11089static char disk_by_path[] = "/dev/disk/by-path/";
11090
11091static const char *imsm_get_disk_controller_domain(const char *path)
11092{
2cda7640 11093 char disk_path[PATH_MAX];
96234762
LM		 11094 char *drv=NULL;
11095 struct stat st;
2cda7640 11096
6d8d290a 11097 strcpy(disk_path, disk_by_path);
96234762
LM		 11098 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11099 if (stat(disk_path, &st) == 0) {
11100 struct sys_dev* hba;
594dc1b8 11101 char *path;
96234762
LM		 11102
11103 path = devt_to_devpath(st.st_rdev);
11104 if (path == NULL)
11105 return "unknown";
11106 hba = find_disk_attached_hba(-1, path);
11107 if (hba && hba->type == SYS_DEV_SAS)
11108 drv = "isci";
11109 else if (hba && hba->type == SYS_DEV_SATA)
11110 drv = "ahci";
c6839718
MT		 11111 else if (hba && hba->type == SYS_DEV_VMD)
11112 drv = "vmd";
11113 else if (hba && hba->type == SYS_DEV_NVME)
11114 drv = "nvme";
1011e834 11115 else
96234762
LM		 11116 drv = "unknown";
11117 dprintf("path: %s hba: %s attached: %s\n",
11118 path, (hba) ? hba->path : "NULL", drv);
11119 free(path);
2cda7640 11120 }
96234762 11121 return drv;
2cda7640
ML		 11122}
11123
4dd2df09 11124static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11125{
4dd2df09 11126 static char devnm[32];
78b10e66
N		 11127 char subdev_name[20];
11128 struct mdstat_ent *mdstat;
11129
11130 sprintf(subdev_name, "%d", subdev);
11131 mdstat = mdstat_by_subdev(subdev_name, container);
11132 if (!mdstat)
4dd2df09 11133 return NULL;
78b10e66 11134
4dd2df09 11135 strcpy(devnm, mdstat->devnm);
78b10e66 11136 free_mdstat(mdstat);
4dd2df09 11137 return devnm;
78b10e66
N		 11138}
11139
11140static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11141 struct geo_params *geo,
fbf3d202
AK		 11142 int *old_raid_disks,
11143 int direction)
78b10e66 11144{
694575e7
KW		 11145 /* currently we only support increasing the number of devices
11146 * for a container. This increases the number of device for each
11147 * member array. They must all be RAID0 or RAID5.
11148 */
78b10e66
N		 11149 int ret_val = 0;
11150 struct mdinfo *info, *member;
11151 int devices_that_can_grow = 0;
11152
7a862a02 11153 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11154
d04f65f4 11155 if (geo->size > 0 ||
78b10e66
N		 11156 geo->level != UnSet ||
11157 geo->layout != UnSet ||
11158 geo->chunksize != 0 ||
11159 geo->raid_disks == UnSet) {
7a862a02 11160 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11161 return ret_val;
11162 }
11163
fbf3d202 11164 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11165 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11166 return ret_val;
11167 }
11168
78b10e66
N		 11169 info = container_content_imsm(st, NULL);
11170 for (member = info; member; member = member->next) {
4dd2df09 11171 char *result;
78b10e66
N		 11172
11173 dprintf("imsm: checking device_num: %i\n",
11174 member->container_member);
11175
d7d205bd 11176 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11177 /* we work on container for Online Capacity Expansion
11178 * only so raid_disks has to grow
11179 */
7a862a02 11180 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11181 break;
11182 }
11183
089f9d79 11184 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11185 /* we cannot use this container with other raid level
11186 */
7a862a02 11187 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11188 info->array.level);
11189 break;
11190 } else {
11191 /* check for platform support
11192 * for this raid level configuration
11193 */
11194 struct intel_super *super = st->sb;
11195 if (!is_raid_level_supported(super->orom,
11196 member->array.level,
11197 geo->raid_disks)) {
7a862a02 11198 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11199 info->array.level,
11200 geo->raid_disks,
11201 geo->raid_disks > 1 ? "s" : "");
11202 break;
11203 }
2a4a08e7
AK		 11204 /* check if component size is aligned to chunk size
11205 */
11206 if (info->component_size %
11207 (info->array.chunk_size/512)) {
7a862a02 11208 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11209 break;
11210 }
78b10e66
N		 11211 }
11212
11213 if (*old_raid_disks &&
11214 info->array.raid_disks != *old_raid_disks)
11215 break;
11216 *old_raid_disks = info->array.raid_disks;
11217
11218 /* All raid5 and raid0 volumes in container
11219 * have to be ready for Online Capacity Expansion
11220 * so they need to be assembled. We have already
11221 * checked that no recovery etc is happening.
11222 */
4dd2df09
N		 11223 result = imsm_find_array_devnm_by_subdev(member->container_member,
11224 st->container_devnm);
11225 if (result == NULL) {
78b10e66
N		 11226 dprintf("imsm: cannot find array\n");
11227 break;
11228 }
11229 devices_that_can_grow++;
11230 }
11231 sysfs_free(info);
11232 if (!member && devices_that_can_grow)
11233 ret_val = 1;
11234
11235 if (ret_val)
1ade5cc1 11236 dprintf("Container operation allowed\n");
78b10e66 11237 else
1ade5cc1 11238 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11239
11240 return ret_val;
11241}
11242
11243/* Function: get_spares_for_grow
11244 * Description: Allocates memory and creates list of spare devices
1011e834 11245 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11246 * Parameters: Pointer to the supertype structure
11247 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11248 * NULL if fail
78b10e66
N		 11249 */
11250static struct mdinfo *get_spares_for_grow(struct supertype *st)
11251{
fbfdcb06
AO		 11252 struct spare_criteria sc;
11253
11254 get_spare_criteria_imsm(st, &sc);
11255 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11256}
11257
11258/******************************************************************************
11259 * function: imsm_create_metadata_update_for_reshape
11260 * Function creates update for whole IMSM container.
11261 *
11262 ******************************************************************************/
11263static int imsm_create_metadata_update_for_reshape(
11264 struct supertype *st,
11265 struct geo_params *geo,
11266 int old_raid_disks,
11267 struct imsm_update_reshape **updatep)
11268{
11269 struct intel_super *super = st->sb;
11270 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11271 int update_memory_size;
11272 struct imsm_update_reshape *u;
11273 struct mdinfo *spares;
78b10e66 11274 int i;
594dc1b8 11275 int delta_disks;
bbd24d86 11276 struct mdinfo *dev;
78b10e66 11277
1ade5cc1 11278 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11279
11280 delta_disks = geo->raid_disks - old_raid_disks;
11281
11282 /* size of all update data without anchor */
11283 update_memory_size = sizeof(struct imsm_update_reshape);
11284
11285 /* now add space for spare disks that we need to add. */
11286 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11287
503975b9 11288 u = xcalloc(1, update_memory_size);
78b10e66
N		 11289 u->type = update_reshape_container_disks;
11290 u->old_raid_disks = old_raid_disks;
11291 u->new_raid_disks = geo->raid_disks;
11292
11293 /* now get spare disks list
11294 */
11295 spares = get_spares_for_grow(st);
11296
d7be7d87 11297 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11298 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11299 i = -1;
78b10e66
N		 11300 goto abort;
11301 }
11302
11303 /* we have got spares
11304 * update disk list in imsm_disk list table in anchor
11305 */
11306 dprintf("imsm: %i spares are available.\n\n",
11307 spares->array.spare_disks);
11308
bbd24d86 11309 dev = spares->devs;
78b10e66 11310 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11311 struct dl *dl;
11312
bbd24d86
AK		 11313 if (dev == NULL)
11314 break;
78b10e66
N		 11315 u->new_disks[i] = makedev(dev->disk.major,
11316 dev->disk.minor);
11317 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11318 dl->index = mpb->num_disks;
11319 mpb->num_disks++;
bbd24d86 11320 dev = dev->next;
78b10e66 11321 }
78b10e66
N		 11322
11323abort:
11324 /* free spares
11325 */
11326 sysfs_free(spares);
11327
d677e0b8 11328 dprintf("imsm: reshape update preparation :");
78b10e66 11329 if (i == delta_disks) {
1ade5cc1 11330 dprintf_cont(" OK\n");
78b10e66
N		 11331 *updatep = u;
11332 return update_memory_size;
11333 }
11334 free(u);
1ade5cc1 11335 dprintf_cont(" Error\n");
78b10e66
N		 11336
11337 return 0;
11338}
11339
f3871fdc
AK		 11340/******************************************************************************
11341 * function: imsm_create_metadata_update_for_size_change()
11342 * Creates update for IMSM array for array size change.
11343 *
11344 ******************************************************************************/
11345static int imsm_create_metadata_update_for_size_change(
11346 struct supertype *st,
11347 struct geo_params *geo,
11348 struct imsm_update_size_change **updatep)
11349{
11350 struct intel_super *super = st->sb;
594dc1b8
JS		 11351 int update_memory_size;
11352 struct imsm_update_size_change *u;
f3871fdc 11353
1ade5cc1 11354 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11355
11356 /* size of all update data without anchor */
11357 update_memory_size = sizeof(struct imsm_update_size_change);
11358
503975b9 11359 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11360 u->type = update_size_change;
11361 u->subdev = super->current_vol;
11362 u->new_size = geo->size;
11363
11364 dprintf("imsm: reshape update preparation : OK\n");
11365 *updatep = u;
11366
11367 return update_memory_size;
11368}
11369
48c5303a
PC		 11370/******************************************************************************
11371 * function: imsm_create_metadata_update_for_migration()
11372 * Creates update for IMSM array.
11373 *
11374 ******************************************************************************/
11375static int imsm_create_metadata_update_for_migration(
11376 struct supertype *st,
11377 struct geo_params *geo,
11378 struct imsm_update_reshape_migration **updatep)
11379{
11380 struct intel_super *super = st->sb;
594dc1b8
JS		 11381 int update_memory_size;
11382 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11383 struct imsm_dev *dev;
11384 int previous_level = -1;
11385
1ade5cc1 11386 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11387
11388 /* size of all update data without anchor */
11389 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11390
503975b9 11391 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11392 u->type = update_reshape_migration;
11393 u->subdev = super->current_vol;
11394 u->new_level = geo->level;
11395 u->new_layout = geo->layout;
11396 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11397 u->new_disks[0] = -1;
4bba0439 11398 u->new_chunksize = -1;
48c5303a
PC		 11399
11400 dev = get_imsm_dev(super, u->subdev);
11401 if (dev) {
11402 struct imsm_map *map;
11403
238c0a71 11404 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11405 if (map) {
11406 int current_chunk_size =
11407 __le16_to_cpu(map->blocks_per_strip) / 2;
11408
11409 if (geo->chunksize != current_chunk_size) {
11410 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11411 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11412 current_chunk_size, u->new_chunksize);
11413 }
48c5303a 11414 previous_level = map->raid_level;
4bba0439 11415 }
48c5303a 11416 }
089f9d79 11417 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11418 struct mdinfo *spares = NULL;
11419
11420 u->new_raid_disks++;
11421 spares = get_spares_for_grow(st);
089f9d79 11422 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11423 free(u);
11424 sysfs_free(spares);
11425 update_memory_size = 0;
565cc99e 11426 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11427 return 0;
11428 }
11429 sysfs_free(spares);
11430 }
11431 dprintf("imsm: reshape update preparation : OK\n");
11432 *updatep = u;
11433
11434 return update_memory_size;
11435}
11436
8dd70bce
AK		 11437static void imsm_update_metadata_locally(struct supertype *st,
11438 void *buf, int len)
11439{
11440 struct metadata_update mu;
11441
11442 mu.buf = buf;
11443 mu.len = len;
11444 mu.space = NULL;
11445 mu.space_list = NULL;
11446 mu.next = NULL;
5fe6f031
N		 11447 if (imsm_prepare_update(st, &mu))
11448 imsm_process_update(st, &mu);
8dd70bce
AK		 11449
11450 while (mu.space_list) {
11451 void **space = mu.space_list;
11452 mu.space_list = *space;
11453 free(space);
11454 }
11455}
78b10e66 11456
471bceb6 11457/***************************************************************************
694575e7 11458* Function: imsm_analyze_change
471bceb6 11459* Description: Function analyze change for single volume
1011e834 11460* and validate if transition is supported
fbf3d202
AK		 11461* Parameters: Geometry parameters, supertype structure,
11462* metadata change direction (apply/rollback)
694575e7 11463* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11464****************************************************************************/
11465enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11466 struct geo_params *geo,
11467 int direction)
694575e7 11468{
471bceb6
KW		 11469 struct mdinfo info;
11470 int change = -1;
11471 int check_devs = 0;
c21e737b 11472 int chunk;
67a2db32
AK		 11473 /* number of added/removed disks in operation result */
11474 int devNumChange = 0;
11475 /* imsm compatible layout value for array geometry verification */
11476 int imsm_layout = -1;
7abc9871
AK		 11477 int data_disks;
11478 struct imsm_dev *dev;
9529d343 11479 struct imsm_map *map;
7abc9871 11480 struct intel_super *super;
d04f65f4 11481 unsigned long long current_size;
65d38cca 11482 unsigned long long free_size;
d04f65f4 11483 unsigned long long max_size;
65d38cca 11484 int rv;
471bceb6
KW		 11485
11486 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11487 if (geo->level != info.array.level && geo->level >= 0 &&
11488 geo->level != UnSet) {
471bceb6
KW		 11489 switch (info.array.level) {
11490 case 0:
11491 if (geo->level == 5) {
b5347799 11492 change = CH_MIGRATION;
e13ce846 11493 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11494 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11495 change = -1;
11496 goto analyse_change_exit;
11497 }
67a2db32 11498 imsm_layout = geo->layout;
471bceb6 11499 check_devs = 1;
e91a3bad
LM		 11500 devNumChange = 1; /* parity disk added */
11501 } else if (geo->level == 10) {
471bceb6
KW		 11502 change = CH_TAKEOVER;
11503 check_devs = 1;
e91a3bad 11504 devNumChange = 2; /* two mirrors added */
67a2db32 11505 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11506 }
dfe77a9e
KW		 11507 break;
11508 case 1:
471bceb6
KW		 11509 case 10:
11510 if (geo->level == 0) {
11511 change = CH_TAKEOVER;
11512 check_devs = 1;
e91a3bad 11513 devNumChange = -(geo->raid_disks/2);
67a2db32 11514 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11515 }
11516 break;
11517 }
11518 if (change == -1) {
7a862a02 11519 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11520 info.array.level, geo->level);
471bceb6
KW		 11521 goto analyse_change_exit;
11522 }
11523 } else
11524 geo->level = info.array.level;
11525
089f9d79
JS		 11526 if (geo->layout != info.array.layout &&
11527 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11528 change = CH_MIGRATION;
089f9d79
JS		 11529 if (info.array.layout == 0 && info.array.level == 5 &&
11530 geo->layout == 5) {
471bceb6 11531 /* reshape 5 -> 4 */
089f9d79
JS		 11532 } else if (info.array.layout == 5 && info.array.level == 5 &&
11533 geo->layout == 0) {
471bceb6
KW		 11534 /* reshape 4 -> 5 */
11535 geo->layout = 0;
11536 geo->level = 5;
11537 } else {
7a862a02 11538 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11539 info.array.layout, geo->layout);
471bceb6
KW		 11540 change = -1;
11541 goto analyse_change_exit;
11542 }
67a2db32 11543 } else {
471bceb6 11544 geo->layout = info.array.layout;
67a2db32
AK		 11545 if (imsm_layout == -1)
11546 imsm_layout = info.array.layout;
11547 }
471bceb6 11548
089f9d79
JS		 11549 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11550 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11551 if (info.array.level == 10) {
11552 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11553 change = -1;
11554 goto analyse_change_exit;
1e9b2c3f
PB		 11555 } else if (info.component_size % (geo->chunksize/512)) {
11556 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11557 geo->chunksize/1024, info.component_size/2);
11558 change = -1;
11559 goto analyse_change_exit;
2d2b0eb7 11560 }
b5347799 11561 change = CH_MIGRATION;
2d2b0eb7 11562 } else {
471bceb6 11563 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11564 }
471bceb6 11565
c21e737b 11566 chunk = geo->chunksize / 1024;
7abc9871
AK		 11567
11568 super = st->sb;
11569 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11570 map = get_imsm_map(dev, MAP_0);
11571 data_disks = imsm_num_data_members(map);
c41e00b2 11572 /* compute current size per disk member
7abc9871 11573 */
c41e00b2
AK		 11574 current_size = info.custom_array_size / data_disks;
11575
089f9d79 11576 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11577 /* align component size
11578 */
3e684231 11579 geo->size = imsm_component_size_alignment_check(
c41e00b2 11580 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11581 chunk * 1024, super->sector_size,
c41e00b2 11582 geo->size * 2);
65d0b4ce 11583 if (geo->size == 0) {
7a862a02 11584 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11585 current_size);
11586 goto analyse_change_exit;
11587 }
c41e00b2 11588 }
7abc9871 11589
089f9d79 11590 if (current_size != geo->size && geo->size > 0) {
7abc9871 11591 if (change != -1) {
7a862a02 11592 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11593 change = -1;
11594 goto analyse_change_exit;
11595 }
11596 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11597 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11598 super->current_vol, st->devnm);
7abc9871
AK		 11599 goto analyse_change_exit;
11600 }
65d38cca
LD		 11601 /* check the maximum available size
11602 */
11603 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11604 0, chunk, &free_size);
11605 if (rv == 0)
11606 /* Cannot find maximum available space
11607 */
11608 max_size = 0;
11609 else {
11610 max_size = free_size + current_size;
11611 /* align component size
11612 */
3e684231 11613 max_size = imsm_component_size_alignment_check(
65d38cca 11614 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11615 chunk * 1024, super->sector_size,
65d38cca
LD		 11616 max_size);
11617 }
d04f65f4 11618 if (geo->size == MAX_SIZE) {
b130333f
AK		 11619 /* requested size change to the maximum available size
11620 */
65d38cca 11621 if (max_size == 0) {
7a862a02 11622 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11623 change = -1;
11624 goto analyse_change_exit;
65d38cca
LD		 11625 } else
11626 geo->size = max_size;
c41e00b2 11627 }
b130333f 11628
681b7ae2 11629 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11630 /* accept size for rollback only
11631 */
11632 } else {
11633 /* round size due to metadata compatibility
11634 */
11635 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11636 << SECT_PER_MB_SHIFT;
11637 dprintf("Prepare update for size change to %llu\n",
11638 geo->size );
11639 if (current_size >= geo->size) {
7a862a02 11640 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11641 current_size, geo->size);
fbf3d202
AK		 11642 goto analyse_change_exit;
11643 }
65d38cca 11644 if (max_size && geo->size > max_size) {
7a862a02 11645 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11646 max_size, geo->size);
65d38cca
LD		 11647 goto analyse_change_exit;
11648 }
7abc9871
AK		 11649 }
11650 geo->size *= data_disks;
11651 geo->raid_disks = dev->vol.map->num_members;
11652 change = CH_ARRAY_SIZE;
11653 }
471bceb6
KW		 11654 if (!validate_geometry_imsm(st,
11655 geo->level,
67a2db32 11656 imsm_layout,
e91a3bad 11657 geo->raid_disks + devNumChange,
c21e737b 11658 &chunk,
af4348dd 11659 geo->size, INVALID_SECTORS,
5308f117 11660 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11661 change = -1;
11662
11663 if (check_devs) {
11664 struct intel_super *super = st->sb;
11665 struct imsm_super *mpb = super->anchor;
11666
11667 if (mpb->num_raid_devs > 1) {
7a862a02 11668 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11669 geo->dev_name);
471bceb6
KW		 11670 change = -1;
11671 }
11672 }
11673
11674analyse_change_exit:
089f9d79
JS		 11675 if (direction == ROLLBACK_METADATA_CHANGES &&
11676 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11677 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11678 change = -1;
11679 }
471bceb6 11680 return change;
694575e7
KW		 11681}
11682
bb025c2f
KW		 11683int imsm_takeover(struct supertype *st, struct geo_params *geo)
11684{
11685 struct intel_super *super = st->sb;
11686 struct imsm_update_takeover *u;
11687
503975b9 11688 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11689
11690 u->type = update_takeover;
11691 u->subarray = super->current_vol;
11692
11693 /* 10->0 transition */
11694 if (geo->level == 0)
11695 u->direction = R10_TO_R0;
11696
0529c688
KW		 11697 /* 0->10 transition */
11698 if (geo->level == 10)
11699 u->direction = R0_TO_R10;
11700
bb025c2f
KW		 11701 /* update metadata locally */
11702 imsm_update_metadata_locally(st, u,
11703 sizeof(struct imsm_update_takeover));
11704 /* and possibly remotely */
11705 if (st->update_tail)
11706 append_metadata_update(st, u,
11707 sizeof(struct imsm_update_takeover));
11708 else
11709 free(u);
11710
11711 return 0;
11712}
11713
d04f65f4
N		 11714static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11715 int level,
78b10e66 11716 int layout, int chunksize, int raid_disks,
41784c88 11717 int delta_disks, char *backup, char *dev,
016e00f5 11718 int direction, int verbose)
78b10e66 11719{
78b10e66
N		 11720 int ret_val = 1;
11721 struct geo_params geo;
11722
1ade5cc1 11723 dprintf("(enter)\n");
78b10e66 11724
71204a50 11725 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11726
11727 geo.dev_name = dev;
4dd2df09 11728 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11729 geo.size = size;
11730 geo.level = level;
11731 geo.layout = layout;
11732 geo.chunksize = chunksize;
11733 geo.raid_disks = raid_disks;
41784c88
AK		 11734 if (delta_disks != UnSet)
11735 geo.raid_disks += delta_disks;
78b10e66 11736
1ade5cc1
N		 11737 dprintf("for level : %i\n", geo.level);
11738 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11739
4dd2df09 11740 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11741 /* On container level we can only increase number of devices. */
11742 dprintf("imsm: info: Container operation\n");
78b10e66 11743 int old_raid_disks = 0;
6dc0be30 11744
78b10e66 11745 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11746 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11747 struct imsm_update_reshape *u = NULL;
11748 int len;
11749
11750 len = imsm_create_metadata_update_for_reshape(
11751 st, &geo, old_raid_disks, &u);
11752
ed08d51c
AK		 11753 if (len <= 0) {
11754 dprintf("imsm: Cannot prepare update\n");
11755 goto exit_imsm_reshape_super;
11756 }
11757
8dd70bce
AK		 11758 ret_val = 0;
11759 /* update metadata locally */
11760 imsm_update_metadata_locally(st, u, len);
11761 /* and possibly remotely */
11762 if (st->update_tail)
11763 append_metadata_update(st, u, len);
11764 else
ed08d51c 11765 free(u);
8dd70bce 11766
694575e7 11767 } else {
7a862a02 11768 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11769 }
11770 } else {
11771 /* On volume level we support following operations
471bceb6
KW		 11772 * - takeover: raid10 -> raid0; raid0 -> raid10
11773 * - chunk size migration
11774 * - migration: raid5 -> raid0; raid0 -> raid5
11775 */
11776 struct intel_super *super = st->sb;
11777 struct intel_dev *dev = super->devlist;
4dd2df09 11778 int change;
694575e7 11779 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11780 /* find requested device */
11781 while (dev) {
1011e834 11782 char *devnm =
4dd2df09
N		 11783 imsm_find_array_devnm_by_subdev(
11784 dev->index, st->container_devnm);
11785 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11786 break;
11787 dev = dev->next;
11788 }
11789 if (dev == NULL) {
4dd2df09
N		 11790 pr_err("Cannot find %s (%s) subarray\n",
11791 geo.dev_name, geo.devnm);
471bceb6
KW		 11792 goto exit_imsm_reshape_super;
11793 }
11794 super->current_vol = dev->index;
fbf3d202 11795 change = imsm_analyze_change(st, &geo, direction);
694575e7 11796 switch (change) {
471bceb6 11797 case CH_TAKEOVER:
bb025c2f 11798 ret_val = imsm_takeover(st, &geo);
694575e7 11799 break;
48c5303a
PC		 11800 case CH_MIGRATION: {
11801 struct imsm_update_reshape_migration *u = NULL;
11802 int len =
11803 imsm_create_metadata_update_for_migration(
11804 st, &geo, &u);
11805 if (len < 1) {
7a862a02 11806 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11807 break;
11808 }
471bceb6 11809 ret_val = 0;
48c5303a
PC		 11810 /* update metadata locally */
11811 imsm_update_metadata_locally(st, u, len);
11812 /* and possibly remotely */
11813 if (st->update_tail)
11814 append_metadata_update(st, u, len);
11815 else
11816 free(u);
11817 }
11818 break;
7abc9871 11819 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11820 struct imsm_update_size_change *u = NULL;
11821 int len =
11822 imsm_create_metadata_update_for_size_change(
11823 st, &geo, &u);
11824 if (len < 1) {
7a862a02 11825 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11826 break;
11827 }
11828 ret_val = 0;
11829 /* update metadata locally */
11830 imsm_update_metadata_locally(st, u, len);
11831 /* and possibly remotely */
11832 if (st->update_tail)
11833 append_metadata_update(st, u, len);
11834 else
11835 free(u);
7abc9871
AK		 11836 }
11837 break;
471bceb6
KW		 11838 default:
11839 ret_val = 1;
694575e7 11840 }
694575e7 11841 }
78b10e66 11842
ed08d51c 11843exit_imsm_reshape_super:
78b10e66
N		 11844 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11845 return ret_val;
11846}
2cda7640 11847
0febb20c
AO		 11848#define COMPLETED_OK 0
11849#define COMPLETED_NONE 1
11850#define COMPLETED_DELAYED 2
11851
11852static int read_completed(int fd, unsigned long long *val)
11853{
11854 int ret;
11855 char buf[50];
11856
11857 ret = sysfs_fd_get_str(fd, buf, 50);
11858 if (ret < 0)
11859 return ret;
11860
11861 ret = COMPLETED_OK;
11862 if (strncmp(buf, "none", 4) == 0) {
11863 ret = COMPLETED_NONE;
11864 } else if (strncmp(buf, "delayed", 7) == 0) {
11865 ret = COMPLETED_DELAYED;
11866 } else {
11867 char *ep;
11868 *val = strtoull(buf, &ep, 0);
11869 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11870 ret = -1;
11871 }
11872 return ret;
11873}
11874
eee67a47
AK		 11875/*******************************************************************************
11876 * Function: wait_for_reshape_imsm
11877 * Description: Function writes new sync_max value and waits until
11878 * reshape process reach new position
11879 * Parameters:
11880 * sra : general array info
eee67a47
AK		 11881 * ndata : number of disks in new array's layout
11882 * Returns:
11883 * 0 : success,
11884 * 1 : there is no reshape in progress,
11885 * -1 : fail
11886 ******************************************************************************/
ae9f01f8 11887int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11888{
85ca499c 11889 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11890 int retry = 3;
eee67a47 11891 unsigned long long completed;
ae9f01f8
AK		 11892 /* to_complete : new sync_max position */
11893 unsigned long long to_complete = sra->reshape_progress;
11894 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11895
ae9f01f8 11896 if (fd < 0) {
1ade5cc1 11897 dprintf("cannot open reshape_position\n");
eee67a47 11898 return 1;
ae9f01f8 11899 }
eee67a47 11900
df2647fa
PB		 11901 do {
11902 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11903 if (!retry) {
11904 dprintf("cannot read reshape_position (no reshape in progres)\n");
11905 close(fd);
11906 return 1;
11907 }
11908 usleep(30000);
11909 } else
11910 break;
11911 } while (retry--);
eee67a47 11912
85ca499c 11913 if (completed > position_to_set) {
1ade5cc1 11914 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11915 to_complete, position_to_set);
ae9f01f8
AK		 11916 close(fd);
11917 return -1;
11918 }
11919 dprintf("Position set: %llu\n", position_to_set);
11920 if (sysfs_set_num(sra, NULL, "sync_max",
11921 position_to_set) != 0) {
1ade5cc1 11922 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11923 position_to_set);
11924 close(fd);
11925 return -1;
eee67a47
AK		 11926 }
11927
eee67a47 11928 do {
0febb20c 11929 int rc;
eee67a47 11930 char action[20];
5ff3a780 11931 int timeout = 3000;
0febb20c 11932
5ff3a780 11933 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11934 if (sysfs_get_str(sra, NULL, "sync_action",
11935 action, 20) > 0 &&
d7d3809a 11936 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11937 if (strncmp(action, "idle", 4) == 0)
11938 break;
d7d3809a
AP		 11939 close(fd);
11940 return -1;
11941 }
0febb20c
AO		 11942
11943 rc = read_completed(fd, &completed);
11944 if (rc < 0) {
1ade5cc1 11945 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11946 close(fd);
11947 return 1;
0febb20c
AO		 11948 } else if (rc == COMPLETED_NONE)
11949 break;
85ca499c 11950 } while (completed < position_to_set);
b2be2b62 11951
eee67a47
AK		 11952 close(fd);
11953 return 0;
eee67a47
AK		 11954}
11955
b915c95f
AK		 11956/*******************************************************************************
11957 * Function: check_degradation_change
11958 * Description: Check that array hasn't become failed.
11959 * Parameters:
11960 * info : for sysfs access
11961 * sources : source disks descriptors
11962 * degraded: previous degradation level
11963 * Returns:
11964 * degradation level
11965 ******************************************************************************/
11966int check_degradation_change(struct mdinfo *info,
11967 int *sources,
11968 int degraded)
11969{
11970 unsigned long long new_degraded;
e1993023
LD		 11971 int rv;
11972
11973 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11974 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11975 /* check each device to ensure it is still working */
11976 struct mdinfo *sd;
11977 new_degraded = 0;
11978 for (sd = info->devs ; sd ; sd = sd->next) {
11979 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11980 continue;
11981 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11982 char sbuf[100];
11983
b915c95f 11984 if (sysfs_get_str(info,
cf52eff5 11985 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11986 strstr(sbuf, "faulty") ||
11987 strstr(sbuf, "in_sync") == NULL) {
11988 /* this device is dead */
11989 sd->disk.state = (1<<MD_DISK_FAULTY);
11990 if (sd->disk.raid_disk >= 0 &&
11991 sources[sd->disk.raid_disk] >= 0) {
11992 close(sources[
11993 sd->disk.raid_disk]);
11994 sources[sd->disk.raid_disk] =
11995 -1;
11996 }
11997 new_degraded++;
11998 }
11999 }
12000 }
12001 }
12002
12003 return new_degraded;
12004}
12005
10f22854
AK		 12006/*******************************************************************************
12007 * Function: imsm_manage_reshape
12008 * Description: Function finds array under reshape and it manages reshape
12009 * process. It creates stripes backups (if required) and sets
942e1cdb 12010 * checkpoints.
10f22854
AK		 12011 * Parameters:
12012 * afd : Backup handle (nattive) - not used
12013 * sra : general array info
12014 * reshape : reshape parameters - not used
12015 * st : supertype structure
12016 * blocks : size of critical section [blocks]
12017 * fds : table of source device descriptor
12018 * offsets : start of array (offest per devices)
12019 * dests : not used
12020 * destfd : table of destination device descriptor
12021 * destoffsets : table of destination offsets (per device)
12022 * Returns:
12023 * 1 : success, reshape is done
12024 * 0 : fail
12025 ******************************************************************************/
999b4972
N		 12026static int imsm_manage_reshape(
12027 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12028 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12029 int *fds, unsigned long long *offsets,
12030 int dests, int *destfd, unsigned long long *destoffsets)
12031{
10f22854
AK		 12032 int ret_val = 0;
12033 struct intel_super *super = st->sb;
594dc1b8 12034 struct intel_dev *dv;
de44e46f 12035 unsigned int sector_size = super->sector_size;
10f22854 12036 struct imsm_dev *dev = NULL;
9529d343 12037 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12038 int migr_vol_qan = 0;
12039 int ndata, odata; /* [bytes] */
12040 int chunk; /* [bytes] */
12041 struct migr_record *migr_rec;
12042 char *buf = NULL;
12043 unsigned int buf_size; /* [bytes] */
12044 unsigned long long max_position; /* array size [bytes] */
12045 unsigned long long next_step; /* [blocks]/[bytes] */
12046 unsigned long long old_data_stripe_length;
10f22854
AK		 12047 unsigned long long start_src; /* [bytes] */
12048 unsigned long long start; /* [bytes] */
12049 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12050 int degraded = 0;
ab724b98 12051 int source_layout = 0;
10f22854 12052
79a16a9b
JS		 12053 if (!sra)
12054 return ret_val;
12055
12056 if (!fds || !offsets)
10f22854
AK		 12057 goto abort;
12058
12059 /* Find volume during the reshape */
12060 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12061 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12062 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12063 dev = dv->dev;
12064 migr_vol_qan++;
12065 }
12066 }
12067 /* Only one volume can migrate at the same time */
12068 if (migr_vol_qan != 1) {
676e87a8 12069 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12070 "Number of migrating volumes greater than 1\n" :
12071 "There is no volume during migrationg\n");
12072 goto abort;
12073 }
12074
9529d343 12075 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12076 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12077 if (map_src == NULL)
12078 goto abort;
10f22854 12079
9529d343
MD		 12080 ndata = imsm_num_data_members(map_dest);
12081 odata = imsm_num_data_members(map_src);
10f22854 12082
7b1ab482 12083 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12084 old_data_stripe_length = odata * chunk;
12085
12086 migr_rec = super->migr_rec;
12087
10f22854
AK		 12088 /* initialize migration record for start condition */
12089 if (sra->reshape_progress == 0)
12090 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12091 else {
12092 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12093 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12094 goto abort;
12095 }
6a75c8ca
AK		 12096 /* Save checkpoint to update migration record for current
12097 * reshape position (in md). It can be farther than current
12098 * reshape position in metadata.
12099 */
12100 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12101 /* ignore error == 2, this can mean end of reshape here
12102 */
7a862a02 12103 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12104 goto abort;
12105 }
b2c59438 12106 }
10f22854
AK		 12107
12108 /* size for data */
12109 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12110 /* extend buffer size for parity disk */
12111 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12112 /* add space for stripe alignment */
10f22854 12113 buf_size += old_data_stripe_length;
de44e46f
PB		 12114 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12115 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12116 goto abort;
12117 }
12118
3ef4403c 12119 max_position = sra->component_size * ndata;
68eb8bc6 12120 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12121
9f421827
PB		 12122 while (current_migr_unit(migr_rec) <
12123 get_num_migr_units(migr_rec)) {
10f22854
AK		 12124 /* current reshape position [blocks] */
12125 unsigned long long current_position =
12126 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12127 * current_migr_unit(migr_rec);
10f22854
AK		 12128 unsigned long long border;
12129
b915c95f
AK		 12130 /* Check that array hasn't become failed.
12131 */
12132 degraded = check_degradation_change(sra, fds, degraded);
12133 if (degraded > 1) {
7a862a02 12134 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12135 goto abort;
12136 }
12137
10f22854
AK		 12138 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12139
12140 if ((current_position + next_step) > max_position)
12141 next_step = max_position - current_position;
12142
92144abf 12143 start = current_position * 512;
10f22854 12144
942e1cdb 12145 /* align reading start to old geometry */
10f22854
AK		 12146 start_buf_shift = start % old_data_stripe_length;
12147 start_src = start - start_buf_shift;
12148
12149 border = (start_src / odata) - (start / ndata);
12150 border /= 512;
12151 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12152 /* save critical stripes to buf
12153 * start - start address of current unit
12154 * to backup [bytes]
12155 * start_src - start address of current unit
12156 * to backup alligned to source array
12157 * [bytes]
12158 */
594dc1b8 12159 unsigned long long next_step_filler;
10f22854
AK		 12160 unsigned long long copy_length = next_step * 512;
12161
12162 /* allign copy area length to stripe in old geometry */
12163 next_step_filler = ((copy_length + start_buf_shift)
12164 % old_data_stripe_length);
12165 if (next_step_filler)
12166 next_step_filler = (old_data_stripe_length
12167 - next_step_filler);
7a862a02 12168 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		 12169 start, start_src, copy_length,
12170 start_buf_shift, next_step_filler);
12171
12172 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12173 chunk, map_src->raid_level,
12174 source_layout, 0, NULL, start_src,
10f22854
AK		 12175 copy_length +
12176 next_step_filler + start_buf_shift,
12177 buf)) {
7a862a02 12178 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12179 goto abort;
12180 }
12181 /* Convert data to destination format and store it
12182 * in backup general migration area
12183 */
12184 if (save_backup_imsm(st, dev, sra,
aea93171 12185 buf + start_buf_shift, copy_length)) {
7a862a02 12186 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12187 goto abort;
12188 }
12189 if (save_checkpoint_imsm(st, sra,
12190 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12191 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12192 goto abort;
12193 }
8016a6d4
AK		 12194 } else {
12195 /* set next step to use whole border area */
12196 border /= next_step;
12197 if (border > 1)
12198 next_step *= border;
10f22854
AK		 12199 }
12200 /* When data backed up, checkpoint stored,
12201 * kick the kernel to reshape unit of data
12202 */
12203 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12204 /* limit next step to array max position */
12205 if (next_step > max_position)
12206 next_step = max_position;
10f22854
AK		 12207 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12208 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12209 sra->reshape_progress = next_step;
10f22854
AK		 12210
12211 /* wait until reshape finish */
c85338c6 12212 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12213 dprintf("wait_for_reshape_imsm returned error!\n");
12214 goto abort;
12215 }
84d11e6c
N		 12216 if (sigterm)
12217 goto abort;
10f22854 12218
0228d92c
AK		 12219 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12220 /* ignore error == 2, this can mean end of reshape here
12221 */
7a862a02 12222 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12223 goto abort;
12224 }
12225
12226 }
12227
71e5411e
PB		 12228 /* clear migr_rec on disks after successful migration */
12229 struct dl *d;
12230
85337573 12231 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12232 for (d = super->disks; d; d = d->next) {
12233 if (d->index < 0 || is_failed(&d->disk))
12234 continue;
12235 unsigned long long dsize;
12236
12237 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12238 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12239 SEEK_SET) >= 0) {
466070ad 12240 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12241 MIGR_REC_BUF_SECTORS*sector_size) !=
12242 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12243 perror("Write migr_rec failed");
12244 }
12245 }
12246
10f22854
AK		 12247 /* return '1' if done */
12248 ret_val = 1;
12249abort:
12250 free(buf);
942e1cdb
N		 12251 /* See Grow.c: abort_reshape() for further explanation */
12252 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12253 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12254 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12255
12256 return ret_val;
999b4972 12257}
0c21b485 12258
cdddbdbc 12259struct superswitch super_imsm = {
cdddbdbc
DW		 12260 .examine_super = examine_super_imsm,
12261 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12262 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12263 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12264 .detail_super = detail_super_imsm,
12265 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12266 .write_init_super = write_init_super_imsm,
0e600426
N		 12267 .validate_geometry = validate_geometry_imsm,
12268 .add_to_super = add_to_super_imsm,
1a64be56 12269 .remove_from_super = remove_from_super_imsm,
d665cc31 12270 .detail_platform = detail_platform_imsm,
e50cf220 12271 .export_detail_platform = export_detail_platform_imsm,
33414a01 12272 .kill_subarray = kill_subarray_imsm,
aa534678 12273 .update_subarray = update_subarray_imsm,
2b959fbf 12274 .load_container = load_container_imsm,
71204a50
N		 12275 .default_geometry = default_geometry_imsm,
12276 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12277 .reshape_super = imsm_reshape_super,
12278 .manage_reshape = imsm_manage_reshape,
9e2d750d 12279 .recover_backup = recover_backup_imsm,
74db60b0 12280 .copy_metadata = copy_metadata_imsm,
27156a57 12281 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12282 .match_home = match_home_imsm,
12283 .uuid_from_super= uuid_from_super_imsm,
12284 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12285 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12286 .update_super = update_super_imsm,
12287
12288 .avail_size = avail_size_imsm,
fbfdcb06 12289 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12290
12291 .compare_super = compare_super_imsm,
12292
12293 .load_super = load_super_imsm,
bf5a934a 12294 .init_super = init_super_imsm,
e683ca88 12295 .store_super = store_super_imsm,
cdddbdbc
DW		 12296 .free_super = free_super_imsm,
12297 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12298 .container_content = container_content_imsm,
0c21b485 12299 .validate_container = validate_container_imsm,
cdddbdbc 12300
2432ce9b
AP		 12301 .write_init_ppl = write_init_ppl_imsm,
12302 .validate_ppl = validate_ppl_imsm,
12303
cdddbdbc 12304 .external = 1,
4cce4069 12305 .name = "imsm",
845dea95
NB		 12306
12307/* for mdmon */
12308 .open_new = imsm_open_new,
ed9d66aa 12309 .set_array_state= imsm_set_array_state,
845dea95
NB		 12310 .set_disk = imsm_set_disk,
12311 .sync_metadata = imsm_sync_metadata,
88758e9d 12312 .activate_spare = imsm_activate_spare,
e8319a19 12313 .process_update = imsm_process_update,
8273f55e 12314 .prepare_update = imsm_prepare_update,
6f50473f 12315 .record_bad_block = imsm_record_badblock,
c07a5a4f 12316 .clear_bad_block = imsm_clear_badblock,
928f1424 12317 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12318};
Unnamed repository; edit this file 'description' to name the repository.