]> 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: Correct minimal device size.
[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
b771faef
BK		 2131static void detail_super_imsm(struct supertype *st, char *homehost,
2132 char *subarray)
cdddbdbc 2133{
3ebe00a1
DW		 2134 struct mdinfo info;
2135 char nbuf[64];
b771faef
BK		 2136 struct intel_super *super = st->sb;
2137 int temp_vol = super->current_vol;
2138
2139 if (subarray)
2140 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2141
a5d85af7 2142 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2143 fname_from_uuid(st, &info, nbuf, ':');
65884368 2144 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2145
2146 super->current_vol = temp_vol;
cdddbdbc
DW		 2147}
2148
b771faef 2149static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2150{
ff54de6e
N		 2151 struct mdinfo info;
2152 char nbuf[64];
b771faef
BK		 2153 struct intel_super *super = st->sb;
2154 int temp_vol = super->current_vol;
2155
2156 if (subarray)
2157 super->current_vol = strtoul(subarray, NULL, 10);
2158
a5d85af7 2159 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2160 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2161 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2162
2163 super->current_vol = temp_vol;
cdddbdbc 2164}
d665cc31 2165
6da53c0e
BK		 2166static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2167 size_t serial_buf_len);
d665cc31
DW		 2168static void fd2devname(int fd, char *name);
2169
120dc887 2170static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2171{
120dc887
LM		 2172 /* dump an unsorted list of devices attached to AHCI Intel storage
2173 * controller, as well as non-connected ports
d665cc31
DW		 2174 */
2175 int hba_len = strlen(hba_path) + 1;
2176 struct dirent *ent;
2177 DIR *dir;
2178 char *path = NULL;
2179 int err = 0;
2180 unsigned long port_mask = (1 << port_count) - 1;
2181
f21e18ca 2182 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2183 if (verbose > 0)
e7b84f9d 2184 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2185 return 2;
2186 }
2187
2188 /* scroll through /sys/dev/block looking for devices attached to
2189 * this hba
2190 */
2191 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2192 if (!dir)
2193 return 1;
2194
2195 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2196 int fd;
2197 char model[64];
2198 char vendor[64];
2199 char buf[1024];
2200 int major, minor;
2201 char *device;
2202 char *c;
2203 int port;
2204 int type;
2205
2206 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2207 continue;
2208 path = devt_to_devpath(makedev(major, minor));
2209 if (!path)
2210 continue;
2211 if (!path_attached_to_hba(path, hba_path)) {
2212 free(path);
2213 path = NULL;
2214 continue;
2215 }
2216
2217 /* retrieve the scsi device type */
2218 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2219 if (verbose > 0)
e7b84f9d 2220 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2221 err = 2;
2222 break;
2223 }
2224 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2225 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2226 if (verbose > 0)
e7b84f9d 2227 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2228 path);
2229 err = 2;
2230 free(device);
2231 break;
2232 }
2233 type = strtoul(buf, NULL, 10);
2234
2235 /* if it's not a disk print the vendor and model */
2236 if (!(type == 0 || type == 7 || type == 14)) {
2237 vendor[0] = '\0';
2238 model[0] = '\0';
2239 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2240 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2241 strncpy(vendor, buf, sizeof(vendor));
2242 vendor[sizeof(vendor) - 1] = '\0';
2243 c = (char *) &vendor[sizeof(vendor) - 1];
2244 while (isspace(*c) || *c == '\0')
2245 *c-- = '\0';
2246
2247 }
2248 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2249 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2250 strncpy(model, buf, sizeof(model));
2251 model[sizeof(model) - 1] = '\0';
2252 c = (char *) &model[sizeof(model) - 1];
2253 while (isspace(*c) || *c == '\0')
2254 *c-- = '\0';
2255 }
2256
2257 if (vendor[0] && model[0])
2258 sprintf(buf, "%.64s %.64s", vendor, model);
2259 else
2260 switch (type) { /* numbers from hald/linux/device.c */
2261 case 1: sprintf(buf, "tape"); break;
2262 case 2: sprintf(buf, "printer"); break;
2263 case 3: sprintf(buf, "processor"); break;
2264 case 4:
2265 case 5: sprintf(buf, "cdrom"); break;
2266 case 6: sprintf(buf, "scanner"); break;
2267 case 8: sprintf(buf, "media_changer"); break;
2268 case 9: sprintf(buf, "comm"); break;
2269 case 12: sprintf(buf, "raid"); break;
2270 default: sprintf(buf, "unknown");
2271 }
2272 } else
2273 buf[0] = '\0';
2274 free(device);
2275
2276 /* chop device path to 'host%d' and calculate the port number */
2277 c = strchr(&path[hba_len], '/');
4e5e717d 2278 if (!c) {
ba728be7 2279 if (verbose > 0)
e7b84f9d 2280 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2281 err = 2;
2282 break;
2283 }
d665cc31 2284 *c = '\0';
0858eccf
AP		 2285 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2286 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2287 port -= host_base;
2288 else {
ba728be7 2289 if (verbose > 0) {
d665cc31 2290 *c = '/'; /* repair the full string */
e7b84f9d 2291 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2292 path);
2293 }
2294 err = 2;
2295 break;
2296 }
2297
2298 /* mark this port as used */
2299 port_mask &= ~(1 << port);
2300
2301 /* print out the device information */
2302 if (buf[0]) {
2303 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2304 continue;
2305 }
2306
2307 fd = dev_open(ent->d_name, O_RDONLY);
2308 if (fd < 0)
2309 printf(" Port%d : - disk info unavailable -\n", port);
2310 else {
2311 fd2devname(fd, buf);
2312 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2313 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2314 sizeof(buf)) == 0)
2315 printf(" (%s)\n", buf);
d665cc31 2316 else
664d5325 2317 printf(" ()\n");
4dab422a 2318 close(fd);
d665cc31 2319 }
d665cc31
DW		 2320 free(path);
2321 path = NULL;
2322 }
2323 if (path)
2324 free(path);
2325 if (dir)
2326 closedir(dir);
2327 if (err == 0) {
2328 int i;
2329
2330 for (i = 0; i < port_count; i++)
2331 if (port_mask & (1 << i))
2332 printf(" Port%d : - no device attached -\n", i);
2333 }
2334
2335 return err;
2336}
2337
6da53c0e 2338static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2339{
6da53c0e 2340 char buf[1024];
60f0f54d
PB		 2341 struct dirent *ent;
2342 DIR *dir;
6da53c0e
BK		 2343 char *rp;
2344 int fd;
60f0f54d 2345
6da53c0e 2346 dir = opendir("/sys/block/");
b9135011 2347 if (!dir)
b5eece69 2348 return 1;
b9135011
JS		 2349
2350 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e
BK		 2351 if (strstr(ent->d_name, "nvme")) {
2352 sprintf(buf, "/sys/block/%s", ent->d_name);
2353 rp = realpath(buf, NULL);
2354 if (!rp)
2355 continue;
2356 if (path_attached_to_hba(rp, hba->path)) {
2357 fd = open_dev(ent->d_name);
2358 if (fd < 0) {
2359 free(rp);
2360 continue;
2361 }
60f0f54d 2362
6da53c0e
BK		 2363 fd2devname(fd, buf);
2364 if (hba->type == SYS_DEV_VMD)
2365 printf(" NVMe under VMD : %s", buf);
2366 else if (hba->type == SYS_DEV_NVME)
2367 printf(" NVMe Device : %s", buf);
2368 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2369 sizeof(buf)))
2370 printf(" (%s)\n", buf);
2371 else
2372 printf("()\n");
2373 close(fd);
2374 }
2375 free(rp);
60f0f54d 2376 }
60f0f54d
PB		 2377 }
2378
b9135011 2379 closedir(dir);
b5eece69 2380 return 0;
60f0f54d
PB		 2381}
2382
120dc887
LM		 2383static void print_found_intel_controllers(struct sys_dev *elem)
2384{
2385 for (; elem; elem = elem->next) {
e7b84f9d 2386 pr_err("found Intel(R) ");
120dc887
LM		 2387 if (elem->type == SYS_DEV_SATA)
2388 fprintf(stderr, "SATA ");
155cbb4c
LM		 2389 else if (elem->type == SYS_DEV_SAS)
2390 fprintf(stderr, "SAS ");
0858eccf
AP		 2391 else if (elem->type == SYS_DEV_NVME)
2392 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2393
2394 if (elem->type == SYS_DEV_VMD)
2395 fprintf(stderr, "VMD domain");
2396 else
2397 fprintf(stderr, "RAID controller");
2398
120dc887
LM		 2399 if (elem->pci_id)
2400 fprintf(stderr, " at %s", elem->pci_id);
2401 fprintf(stderr, ".\n");
2402 }
2403 fflush(stderr);
2404}
2405
120dc887
LM		 2406static int ahci_get_port_count(const char *hba_path, int *port_count)
2407{
2408 struct dirent *ent;
2409 DIR *dir;
2410 int host_base = -1;
2411
2412 *port_count = 0;
2413 if ((dir = opendir(hba_path)) == NULL)
2414 return -1;
2415
2416 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2417 int host;
2418
0858eccf
AP		 2419 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2420 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2421 continue;
2422 if (*port_count == 0)
2423 host_base = host;
2424 else if (host < host_base)
2425 host_base = host;
2426
2427 if (host + 1 > *port_count + host_base)
2428 *port_count = host + 1 - host_base;
2429 }
2430 closedir(dir);
2431 return host_base;
2432}
2433
a891a3c2
LM		 2434static void print_imsm_capability(const struct imsm_orom *orom)
2435{
0858eccf
AP		 2436 printf(" Platform : Intel(R) ");
2437 if (orom->capabilities == 0 && orom->driver_features == 0)
2438 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2439 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2440 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2441 else
2442 printf("Rapid Storage Technology%s\n",
2443 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2444 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2445 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2446 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2447 printf(" RAID Levels :%s%s%s%s%s\n",
2448 imsm_orom_has_raid0(orom) ? " raid0" : "",
2449 imsm_orom_has_raid1(orom) ? " raid1" : "",
2450 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2451 imsm_orom_has_raid10(orom) ? " raid10" : "",
2452 imsm_orom_has_raid5(orom) ? " raid5" : "");
2453 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2454 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2455 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2456 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2457 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2458 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2459 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2460 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2461 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2462 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2463 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2464 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2465 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2466 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2467 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2468 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2469 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2470 printf(" 2TB volumes :%s supported\n",
2471 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2472 printf(" 2TB disks :%s supported\n",
2473 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2474 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2475 printf(" Max Volumes : %d per array, %d per %s\n",
2476 orom->vpa, orom->vphba,
2477 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2478 return;
2479}
2480
e50cf220
MN		 2481static void print_imsm_capability_export(const struct imsm_orom *orom)
2482{
2483 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2484 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2485 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2486 orom->hotfix_ver, orom->build);
e50cf220
MN		 2487 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2488 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2489 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2490 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2491 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2492 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2493 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2494 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2495 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2496 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2497 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2498 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2499 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2500 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2501 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2502 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2503 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2504 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2505 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2506 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2507 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2508 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2509 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2510 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2511 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2512 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2513 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2514 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2515}
2516
9eafa1de 2517static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2518{
2519 /* There are two components to imsm platform support, the ahci SATA
2520 * controller and the option-rom. To find the SATA controller we
2521 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2522 * controller with the Intel vendor id is present. This approach
2523 * allows mdadm to leverage the kernel's ahci detection logic, with the
2524 * caveat that if ahci.ko is not loaded mdadm will not be able to
2525 * detect platform raid capabilities. The option-rom resides in a
2526 * platform "Adapter ROM". We scan for its signature to retrieve the
2527 * platform capabilities. If raid support is disabled in the BIOS the
2528 * option-rom capability structure will not be available.
2529 */
d665cc31 2530 struct sys_dev *list, *hba;
d665cc31
DW		 2531 int host_base = 0;
2532 int port_count = 0;
9eafa1de 2533 int result=1;
d665cc31 2534
5615172f 2535 if (enumerate_only) {
a891a3c2 2536 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2537 return 0;
a891a3c2
LM		 2538 list = find_intel_devices();
2539 if (!list)
2540 return 2;
2541 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2542 if (find_imsm_capability(hba)) {
2543 result = 0;
a891a3c2
LM		 2544 break;
2545 }
9eafa1de 2546 else
6b781d33 2547 result = 2;
a891a3c2 2548 }
a891a3c2 2549 return result;
5615172f
DW		 2550 }
2551
155cbb4c
LM		 2552 list = find_intel_devices();
2553 if (!list) {
ba728be7 2554 if (verbose > 0)
7a862a02 2555 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2556 return 2;
ba728be7 2557 } else if (verbose > 0)
155cbb4c 2558 print_found_intel_controllers(list);
d665cc31 2559
a891a3c2 2560 for (hba = list; hba; hba = hba->next) {
0858eccf 2561 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2562 continue;
0858eccf 2563 if (!find_imsm_capability(hba)) {
60f0f54d 2564 char buf[PATH_MAX];
e7b84f9d 2565 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2566 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2567 get_sys_dev_type(hba->type));
0858eccf
AP		 2568 continue;
2569 }
2570 result = 0;
2571 }
2572
2573 if (controller_path && result == 1) {
2574 pr_err("no active Intel(R) RAID controller found under %s\n",
2575 controller_path);
2576 return result;
2577 }
2578
5e1d6128 2579 const struct orom_entry *entry;
0858eccf 2580
5e1d6128 2581 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2582 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2583 print_imsm_capability(&entry->orom);
32716c51
PB		 2584 printf(" 3rd party NVMe :%s supported\n",
2585 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2586 for (hba = list; hba; hba = hba->next) {
2587 if (hba->type == SYS_DEV_VMD) {
2588 char buf[PATH_MAX];
60f0f54d
PB		 2589 printf(" I/O Controller : %s (%s)\n",
2590 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2591 if (print_nvme_info(hba)) {
b5eece69
PB		 2592 if (verbose > 0)
2593 pr_err("failed to get devices attached to VMD domain.\n");
2594 result |= 2;
2595 }
60f0f54d
PB		 2596 }
2597 }
07cb1e57 2598 printf("\n");
60f0f54d
PB		 2599 continue;
2600 }
0858eccf 2601
60f0f54d
PB		 2602 print_imsm_capability(&entry->orom);
2603 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2604 for (hba = list; hba; hba = hba->next) {
2605 if (hba->type == SYS_DEV_NVME)
6da53c0e 2606 print_nvme_info(hba);
0858eccf 2607 }
60f0f54d 2608 printf("\n");
0858eccf
AP		 2609 continue;
2610 }
2611
2612 struct devid_list *devid;
5e1d6128 2613 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2614 hba = device_by_id(devid->devid);
2615 if (!hba)
2616 continue;
2617
9eafa1de
MN		 2618 printf(" I/O Controller : %s (%s)\n",
2619 hba->path, get_sys_dev_type(hba->type));
2620 if (hba->type == SYS_DEV_SATA) {
2621 host_base = ahci_get_port_count(hba->path, &port_count);
2622 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2623 if (verbose > 0)
7a862a02 2624 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2625 result |= 2;
2626 }
120dc887
LM		 2627 }
2628 }
0858eccf 2629 printf("\n");
d665cc31 2630 }
155cbb4c 2631
120dc887 2632 return result;
d665cc31 2633}
e50cf220 2634
9eafa1de 2635static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2636{
e50cf220
MN		 2637 struct sys_dev *list, *hba;
2638 int result=1;
2639
2640 list = find_intel_devices();
2641 if (!list) {
2642 if (verbose > 0)
2643 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2644 result = 2;
e50cf220
MN		 2645 return result;
2646 }
2647
2648 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2649 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2650 continue;
60f0f54d
PB		 2651 if (!find_imsm_capability(hba) && verbose > 0) {
2652 char buf[PATH_MAX];
2653 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2654 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2655 }
0858eccf 2656 else
e50cf220 2657 result = 0;
e50cf220
MN		 2658 }
2659
5e1d6128 2660 const struct orom_entry *entry;
0858eccf 2661
60f0f54d
PB		 2662 for (entry = orom_entries; entry; entry = entry->next) {
2663 if (entry->type == SYS_DEV_VMD) {
2664 for (hba = list; hba; hba = hba->next)
2665 print_imsm_capability_export(&entry->orom);
2666 continue;
2667 }
5e1d6128 2668 print_imsm_capability_export(&entry->orom);
60f0f54d 2669 }
0858eccf 2670
e50cf220
MN		 2671 return result;
2672}
2673
cdddbdbc
DW		 2674static int match_home_imsm(struct supertype *st, char *homehost)
2675{
5115ca67
DW		 2676 /* the imsm metadata format does not specify any host
2677 * identification information. We return -1 since we can never
2678 * confirm nor deny whether a given array is "meant" for this
148acb7b 2679 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2680 * exclude member disks that do not belong, and we rely on
2681 * mdadm.conf to specify the arrays that should be assembled.
2682 * Auto-assembly may still pick up "foreign" arrays.
2683 */
cdddbdbc 2684
9362c1c8 2685 return -1;
cdddbdbc
DW		 2686}
2687
2688static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2689{
51006d85
N		 2690 /* The uuid returned here is used for:
2691 * uuid to put into bitmap file (Create, Grow)
2692 * uuid for backup header when saving critical section (Grow)
2693 * comparing uuids when re-adding a device into an array
2694 * In these cases the uuid required is that of the data-array,
2695 * not the device-set.
2696 * uuid to recognise same set when adding a missing device back
2697 * to an array. This is a uuid for the device-set.
1011e834 2698 *
51006d85
N		 2699 * For each of these we can make do with a truncated
2700 * or hashed uuid rather than the original, as long as
2701 * everyone agrees.
2702 * In each case the uuid required is that of the data-array,
2703 * not the device-set.
43dad3d6 2704 */
51006d85
N		 2705 /* imsm does not track uuid's so we synthesis one using sha1 on
2706 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2707 * - the orig_family_num of the container
51006d85
N		 2708 * - the index number of the volume
2709 * - the 'serial' number of the volume.
2710 * Hopefully these are all constant.
2711 */
2712 struct intel_super *super = st->sb;
43dad3d6 2713
51006d85
N		 2714 char buf[20];
2715 struct sha1_ctx ctx;
2716 struct imsm_dev *dev = NULL;
148acb7b 2717 __u32 family_num;
51006d85 2718
148acb7b
DW		 2719 /* some mdadm versions failed to set ->orig_family_num, in which
2720 * case fall back to ->family_num. orig_family_num will be
2721 * fixed up with the first metadata update.
2722 */
2723 family_num = super->anchor->orig_family_num;
2724 if (family_num == 0)
2725 family_num = super->anchor->family_num;
51006d85 2726 sha1_init_ctx(&ctx);
92bd8f8d 2727 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2728 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2729 if (super->current_vol >= 0)
2730 dev = get_imsm_dev(super, super->current_vol);
2731 if (dev) {
2732 __u32 vol = super->current_vol;
2733 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2734 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2735 }
2736 sha1_finish_ctx(&ctx, buf);
2737 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2738}
2739
0d481d37 2740#if 0
4f5bc454
DW		 2741static void
2742get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2743{
cdddbdbc
DW		 2744 __u8 *v = get_imsm_version(mpb);
2745 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2746 char major[] = { 0, 0, 0 };
2747 char minor[] = { 0 ,0, 0 };
2748 char patch[] = { 0, 0, 0 };
2749 char *ver_parse[] = { major, minor, patch };
2750 int i, j;
2751
2752 i = j = 0;
2753 while (*v != '\0' && v < end) {
2754 if (*v != '.' && j < 2)
2755 ver_parse[i][j++] = *v;
2756 else {
2757 i++;
2758 j = 0;
2759 }
2760 v++;
2761 }
2762
4f5bc454
DW		 2763 *m = strtol(minor, NULL, 0);
2764 *p = strtol(patch, NULL, 0);
2765}
0d481d37 2766#endif
4f5bc454 2767
1e5c6983
DW		 2768static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2769{
2770 /* migr_strip_size when repairing or initializing parity */
238c0a71 2771 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2772 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2773
2774 switch (get_imsm_raid_level(map)) {
2775 case 5:
2776 case 10:
2777 return chunk;
2778 default:
2779 return 128*1024 >> 9;
2780 }
2781}
2782
2783static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2784{
2785 /* migr_strip_size when rebuilding a degraded disk, no idea why
2786 * this is different than migr_strip_size_resync(), but it's good
2787 * to be compatible
2788 */
238c0a71 2789 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2790 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2791
2792 switch (get_imsm_raid_level(map)) {
2793 case 1:
2794 case 10:
2795 if (map->num_members % map->num_domains == 0)
2796 return 128*1024 >> 9;
2797 else
2798 return chunk;
2799 case 5:
2800 return max((__u32) 64*1024 >> 9, chunk);
2801 default:
2802 return 128*1024 >> 9;
2803 }
2804}
2805
2806static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2807{
238c0a71
AK		 2808 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2809 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2810 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2811 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2812
2813 return max((__u32) 1, hi_chunk / lo_chunk);
2814}
2815
2816static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2817{
238c0a71 2818 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2819 int level = get_imsm_raid_level(lo);
2820
2821 if (level == 1 || level == 10) {
238c0a71 2822 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2823
2824 return hi->num_domains;
2825 } else
2826 return num_stripes_per_unit_resync(dev);
2827}
2828
9529d343 2829static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2830{
2831 /* named 'imsm_' because raid0, raid1 and raid10
2832 * counter-intuitively have the same number of data disks
2833 */
1e5c6983
DW		 2834 switch (get_imsm_raid_level(map)) {
2835 case 0:
36fd8ccc
AK		 2836 return map->num_members;
2837 break;
1e5c6983
DW		 2838 case 1:
2839 case 10:
36fd8ccc 2840 return map->num_members/2;
1e5c6983
DW		 2841 case 5:
2842 return map->num_members - 1;
2843 default:
1ade5cc1 2844 dprintf("unsupported raid level\n");
1e5c6983
DW		 2845 return 0;
2846 }
2847}
2848
44490938
MD		 2849static unsigned long long calc_component_size(struct imsm_map *map,
2850 struct imsm_dev *dev)
2851{
2852 unsigned long long component_size;
2853 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2854 long long calc_dev_size = 0;
44490938
MD		 2855 unsigned int member_disks = imsm_num_data_members(map);
2856
2857 if (member_disks == 0)
2858 return 0;
2859
2860 component_size = per_dev_array_size(map);
2861 calc_dev_size = component_size * member_disks;
2862
2863 /* Component size is rounded to 1MB so difference between size from
2864 * metadata and size calculated from num_data_stripes equals up to
2865 * 2048 blocks per each device. If the difference is higher it means
2866 * that array size was expanded and num_data_stripes was not updated.
2867 */
a4f7290c 2868 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2869 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2870 component_size = dev_size / member_disks;
2871 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2872 component_size / map->blocks_per_strip,
2873 num_data_stripes(map));
2874 }
2875
2876 return component_size;
2877}
2878
1e5c6983
DW		 2879static __u32 parity_segment_depth(struct imsm_dev *dev)
2880{
238c0a71 2881 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2882 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2883
2884 switch(get_imsm_raid_level(map)) {
2885 case 1:
2886 case 10:
2887 return chunk * map->num_domains;
2888 case 5:
2889 return chunk * map->num_members;
2890 default:
2891 return chunk;
2892 }
2893}
2894
2895static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2896{
238c0a71 2897 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2898 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2899 __u32 strip = block / chunk;
2900
2901 switch (get_imsm_raid_level(map)) {
2902 case 1:
2903 case 10: {
2904 __u32 vol_strip = (strip * map->num_domains) + 1;
2905 __u32 vol_stripe = vol_strip / map->num_members;
2906
2907 return vol_stripe * chunk + block % chunk;
2908 } case 5: {
2909 __u32 stripe = strip / (map->num_members - 1);
2910
2911 return stripe * chunk + block % chunk;
2912 }
2913 default:
2914 return 0;
2915 }
2916}
2917
c47b0ff6
AK		 2918static __u64 blocks_per_migr_unit(struct intel_super *super,
2919 struct imsm_dev *dev)
1e5c6983
DW		 2920{
2921 /* calculate the conversion factor between per member 'blocks'
2922 * (md/{resync,rebuild}_start) and imsm migration units, return
2923 * 0 for the 'not migrating' and 'unsupported migration' cases
2924 */
2925 if (!dev->vol.migr_state)
2926 return 0;
2927
2928 switch (migr_type(dev)) {
c47b0ff6
AK		 2929 case MIGR_GEN_MIGR: {
2930 struct migr_record *migr_rec = super->migr_rec;
2931 return __le32_to_cpu(migr_rec->blocks_per_unit);
2932 }
1e5c6983
DW		 2933 case MIGR_VERIFY:
2934 case MIGR_REPAIR:
2935 case MIGR_INIT: {
238c0a71 2936 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2937 __u32 stripes_per_unit;
2938 __u32 blocks_per_unit;
2939 __u32 parity_depth;
2940 __u32 migr_chunk;
2941 __u32 block_map;
2942 __u32 block_rel;
2943 __u32 segment;
2944 __u32 stripe;
2945 __u8 disks;
2946
2947 /* yes, this is really the translation of migr_units to
2948 * per-member blocks in the 'resync' case
2949 */
2950 stripes_per_unit = num_stripes_per_unit_resync(dev);
2951 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2952 disks = imsm_num_data_members(map);
1e5c6983 2953 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2954 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2955 segment = blocks_per_unit / stripe;
2956 block_rel = blocks_per_unit - segment * stripe;
2957 parity_depth = parity_segment_depth(dev);
2958 block_map = map_migr_block(dev, block_rel);
2959 return block_map + parity_depth * segment;
2960 }
2961 case MIGR_REBUILD: {
2962 __u32 stripes_per_unit;
2963 __u32 migr_chunk;
2964
2965 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2966 migr_chunk = migr_strip_blocks_rebuild(dev);
2967 return migr_chunk * stripes_per_unit;
2968 }
1e5c6983
DW		 2969 case MIGR_STATE_CHANGE:
2970 default:
2971 return 0;
2972 }
2973}
2974
c2c087e6
DW		 2975static int imsm_level_to_layout(int level)
2976{
2977 switch (level) {
2978 case 0:
2979 case 1:
2980 return 0;
2981 case 5:
2982 case 6:
a380c027 2983 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2984 case 10:
c92a2527 2985 return 0x102;
c2c087e6 2986 }
a18a888e 2987 return UnSet;
c2c087e6
DW		 2988}
2989
8e59f3d8
AK		 2990/*******************************************************************************
2991 * Function: read_imsm_migr_rec
2992 * Description: Function reads imsm migration record from last sector of disk
2993 * Parameters:
2994 * fd : disk descriptor
2995 * super : metadata info
2996 * Returns:
2997 * 0 : success,
2998 * -1 : fail
2999 ******************************************************************************/
3000static int read_imsm_migr_rec(int fd, struct intel_super *super)
3001{
3002 int ret_val = -1;
de44e46f 3003 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3004 unsigned long long dsize;
3005
3006 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3007 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3008 SEEK_SET) < 0) {
e7b84f9d
N		 3009 pr_err("Cannot seek to anchor block: %s\n",
3010 strerror(errno));
8e59f3d8
AK		 3011 goto out;
3012 }
466070ad 3013 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3014 MIGR_REC_BUF_SECTORS*sector_size) !=
3015 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3016 pr_err("Cannot read migr record block: %s\n",
3017 strerror(errno));
8e59f3d8
AK		 3018 goto out;
3019 }
3020 ret_val = 0;
de44e46f
PB		 3021 if (sector_size == 4096)
3022 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3023
3024out:
3025 return ret_val;
3026}
3027
3136abe5
AK		 3028static struct imsm_dev *imsm_get_device_during_migration(
3029 struct intel_super *super)
3030{
3031
3032 struct intel_dev *dv;
3033
3034 for (dv = super->devlist; dv; dv = dv->next) {
3035 if (is_gen_migration(dv->dev))
3036 return dv->dev;
3037 }
3038 return NULL;
3039}
3040
8e59f3d8
AK		 3041/*******************************************************************************
3042 * Function: load_imsm_migr_rec
3043 * Description: Function reads imsm migration record (it is stored at the last
3044 * sector of disk)
3045 * Parameters:
3046 * super : imsm internal array info
3047 * info : general array info
3048 * Returns:
3049 * 0 : success
3050 * -1 : fail
4c965cc9 3051 * -2 : no migration in progress
8e59f3d8
AK		 3052 ******************************************************************************/
3053static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3054{
3055 struct mdinfo *sd;
594dc1b8 3056 struct dl *dl;
8e59f3d8
AK		 3057 char nm[30];
3058 int retval = -1;
3059 int fd = -1;
3136abe5 3060 struct imsm_dev *dev;
594dc1b8 3061 struct imsm_map *map;
b4ab44d8 3062 int slot = -1;
3136abe5
AK		 3063
3064 /* find map under migration */
3065 dev = imsm_get_device_during_migration(super);
3066 /* nothing to load,no migration in progress?
3067 */
3068 if (dev == NULL)
4c965cc9 3069 return -2;
8e59f3d8
AK		 3070
3071 if (info) {
3072 for (sd = info->devs ; sd ; sd = sd->next) {
3073 /* read only from one of the first two slots */
12fe93e9
TM		 3074 if ((sd->disk.raid_disk < 0) ||
3075 (sd->disk.raid_disk > 1))
8e59f3d8 3076 continue;
3136abe5 3077
8e59f3d8
AK		 3078 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3079 fd = dev_open(nm, O_RDONLY);
3080 if (fd >= 0)
3081 break;
3082 }
3083 }
3084 if (fd < 0) {
12fe93e9 3085 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3086 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3087 /* skip spare and failed disks
3088 */
3089 if (dl->index < 0)
3090 continue;
8e59f3d8 3091 /* read only from one of the first two slots */
3136abe5
AK		 3092 if (map)
3093 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3094 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3095 continue;
3096 sprintf(nm, "%d:%d", dl->major, dl->minor);
3097 fd = dev_open(nm, O_RDONLY);
3098 if (fd >= 0)
3099 break;
3100 }
3101 }
3102 if (fd < 0)
3103 goto out;
3104 retval = read_imsm_migr_rec(fd, super);
3105
3106out:
3107 if (fd >= 0)
3108 close(fd);
3109 return retval;
3110}
3111
c17608ea
AK		 3112/*******************************************************************************
3113 * function: imsm_create_metadata_checkpoint_update
3114 * Description: It creates update for checkpoint change.
3115 * Parameters:
3116 * super : imsm internal array info
3117 * u : pointer to prepared update
3118 * Returns:
3119 * Uptate length.
3120 * If length is equal to 0, input pointer u contains no update
3121 ******************************************************************************/
3122static int imsm_create_metadata_checkpoint_update(
3123 struct intel_super *super,
3124 struct imsm_update_general_migration_checkpoint **u)
3125{
3126
3127 int update_memory_size = 0;
3128
1ade5cc1 3129 dprintf("(enter)\n");
c17608ea
AK		 3130
3131 if (u == NULL)
3132 return 0;
3133 *u = NULL;
3134
3135 /* size of all update data without anchor */
3136 update_memory_size =
3137 sizeof(struct imsm_update_general_migration_checkpoint);
3138
503975b9 3139 *u = xcalloc(1, update_memory_size);
c17608ea 3140 if (*u == NULL) {
1ade5cc1 3141 dprintf("error: cannot get memory\n");
c17608ea
AK		 3142 return 0;
3143 }
3144 (*u)->type = update_general_migration_checkpoint;
9f421827 3145 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3146 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3147
3148 return update_memory_size;
3149}
3150
c17608ea
AK		 3151static void imsm_update_metadata_locally(struct supertype *st,
3152 void *buf, int len);
3153
687629c2
AK		 3154/*******************************************************************************
3155 * Function: write_imsm_migr_rec
3156 * Description: Function writes imsm migration record
3157 * (at the last sector of disk)
3158 * Parameters:
3159 * super : imsm internal array info
3160 * Returns:
3161 * 0 : success
3162 * -1 : if fail
3163 ******************************************************************************/
3164static int write_imsm_migr_rec(struct supertype *st)
3165{
3166 struct intel_super *super = st->sb;
de44e46f 3167 unsigned int sector_size = super->sector_size;
687629c2
AK		 3168 unsigned long long dsize;
3169 char nm[30];
3170 int fd = -1;
3171 int retval = -1;
3172 struct dl *sd;
c17608ea
AK		 3173 int len;
3174 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3175 struct imsm_dev *dev;
594dc1b8 3176 struct imsm_map *map;
3136abe5
AK		 3177
3178 /* find map under migration */
3179 dev = imsm_get_device_during_migration(super);
3180 /* if no migration, write buffer anyway to clear migr_record
3181 * on disk based on first available device
3182 */
3183 if (dev == NULL)
3184 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3185 super->current_vol);
3186
44bfe6df 3187 map = get_imsm_map(dev, MAP_0);
687629c2 3188
de44e46f
PB		 3189 if (sector_size == 4096)
3190 convert_to_4k_imsm_migr_rec(super);
687629c2 3191 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3192 int slot = -1;
3136abe5
AK		 3193
3194 /* skip failed and spare devices */
3195 if (sd->index < 0)
3196 continue;
687629c2 3197 /* write to 2 first slots only */
3136abe5
AK		 3198 if (map)
3199 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3200 if (map == NULL || slot > 1 || slot < 0)
687629c2 3201 continue;
3136abe5 3202
687629c2
AK		 3203 sprintf(nm, "%d:%d", sd->major, sd->minor);
3204 fd = dev_open(nm, O_RDWR);
3205 if (fd < 0)
3206 continue;
3207 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3208 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3209 SEEK_SET) < 0) {
e7b84f9d
N		 3210 pr_err("Cannot seek to anchor block: %s\n",
3211 strerror(errno));
687629c2
AK		 3212 goto out;
3213 }
466070ad 3214 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3215 MIGR_REC_BUF_SECTORS*sector_size) !=
3216 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3217 pr_err("Cannot write migr record block: %s\n",
3218 strerror(errno));
687629c2
AK		 3219 goto out;
3220 }
3221 close(fd);
3222 fd = -1;
3223 }
de44e46f
PB		 3224 if (sector_size == 4096)
3225 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3226 /* update checkpoint information in metadata */
3227 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3228 if (len <= 0) {
3229 dprintf("imsm: Cannot prepare update\n");
3230 goto out;
3231 }
3232 /* update metadata locally */
3233 imsm_update_metadata_locally(st, u, len);
3234 /* and possibly remotely */
3235 if (st->update_tail) {
3236 append_metadata_update(st, u, len);
3237 /* during reshape we do all work inside metadata handler
3238 * manage_reshape(), so metadata update has to be triggered
3239 * insida it
3240 */
3241 flush_metadata_updates(st);
3242 st->update_tail = &st->updates;
3243 } else
3244 free(u);
687629c2
AK		 3245
3246 retval = 0;
3247 out:
3248 if (fd >= 0)
3249 close(fd);
3250 return retval;
3251}
3252
e2962bfc
AK		 3253/* spare/missing disks activations are not allowe when
3254 * array/container performs reshape operation, because
3255 * all arrays in container works on the same disks set
3256 */
3257int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3258{
3259 int rv = 0;
3260 struct intel_dev *i_dev;
3261 struct imsm_dev *dev;
3262
3263 /* check whole container
3264 */
3265 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3266 dev = i_dev->dev;
3ad25638 3267 if (is_gen_migration(dev)) {
e2962bfc
AK		 3268 /* No repair during any migration in container
3269 */
3270 rv = 1;
3271 break;
3272 }
3273 }
3274 return rv;
3275}
3e684231 3276static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3277 int chunk_size,
f36a9ecd 3278 unsigned int sector_size,
c41e00b2
AK		 3279 unsigned long long component_size)
3280{
3e684231 3281 unsigned int component_size_alignment;
c41e00b2 3282
3e684231 3283 /* check component size alignment
c41e00b2 3284 */
3e684231 3285 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3286
3e684231 3287 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3288 level, chunk_size, component_size,
3e684231 3289 component_size_alignment);
c41e00b2 3290
3e684231
MZ		 3291 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3292 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3293 component_size);
3e684231 3294 component_size -= component_size_alignment;
1ade5cc1 3295 dprintf_cont("to %llu (%i).\n",
3e684231 3296 component_size, component_size_alignment);
c41e00b2
AK		 3297 }
3298
3299 return component_size;
3300}
e2962bfc 3301
2432ce9b
AP		 3302static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3303{
3304 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3305 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3306
3307 return pba_of_lba0(map) +
3308 (num_data_stripes(map) * map->blocks_per_strip);
3309}
3310
a5d85af7 3311static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3312{
3313 struct intel_super *super = st->sb;
c47b0ff6 3314 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3315 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3316 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3317 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3318 struct imsm_map *map_to_analyse = map;
efb30e7f 3319 struct dl *dl;
a5d85af7 3320 int map_disks = info->array.raid_disks;
bf5a934a 3321
95eeceeb 3322 memset(info, 0, sizeof(*info));
b335e593
AK		 3323 if (prev_map)
3324 map_to_analyse = prev_map;
3325
ca0748fa 3326 dl = super->current_disk;
9894ec0d 3327
bf5a934a 3328 info->container_member = super->current_vol;
cd0430a1 3329 info->array.raid_disks = map->num_members;
b335e593 3330 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3331 info->array.layout = imsm_level_to_layout(info->array.level);
3332 info->array.md_minor = -1;
3333 info->array.ctime = 0;
3334 info->array.utime = 0;
b335e593
AK		 3335 info->array.chunk_size =
3336 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3337 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3338 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3339 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3340
3f510843 3341 if (is_gen_migration(dev)) {
3f83228a 3342 info->reshape_active = 1;
b335e593
AK		 3343 info->new_level = get_imsm_raid_level(map);
3344 info->new_layout = imsm_level_to_layout(info->new_level);
3345 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3346 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3347 if (info->delta_disks) {
3348 /* this needs to be applied to every array
3349 * in the container.
3350 */
81219e70 3351 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3352 }
3f83228a
N		 3353 /* We shape information that we give to md might have to be
3354 * modify to cope with md's requirement for reshaping arrays.
3355 * For example, when reshaping a RAID0, md requires it to be
3356 * presented as a degraded RAID4.
3357 * Also if a RAID0 is migrating to a RAID5 we need to specify
3358 * the array as already being RAID5, but the 'before' layout
3359 * is a RAID4-like layout.
3360 */
3361 switch (info->array.level) {
3362 case 0:
3363 switch(info->new_level) {
3364 case 0:
3365 /* conversion is happening as RAID4 */
3366 info->array.level = 4;
3367 info->array.raid_disks += 1;
3368 break;
3369 case 5:
3370 /* conversion is happening as RAID5 */
3371 info->array.level = 5;
3372 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3373 info->delta_disks -= 1;
3374 break;
3375 default:
3376 /* FIXME error message */
3377 info->array.level = UnSet;
3378 break;
3379 }
3380 break;
3381 }
b335e593
AK		 3382 } else {
3383 info->new_level = UnSet;
3384 info->new_layout = UnSet;
3385 info->new_chunk = info->array.chunk_size;
3f83228a 3386 info->delta_disks = 0;
b335e593 3387 }
ca0748fa 3388
efb30e7f
DW		 3389 if (dl) {
3390 info->disk.major = dl->major;
3391 info->disk.minor = dl->minor;
ca0748fa 3392 info->disk.number = dl->index;
656b6b5a
N		 3393 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3394 dl->index);
efb30e7f 3395 }
bf5a934a 3396
5551b113 3397 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3398 info->component_size = calc_component_size(map, dev);
3e684231 3399 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3400 info->array.level,
3401 info->array.chunk_size,
f36a9ecd 3402 super->sector_size,
c41e00b2 3403 info->component_size);
5e46202e 3404 info->bb.supported = 1;
139dae11 3405
301406c9 3406 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3407 info->recovery_start = MaxSector;
bf5a934a 3408
c2462068
PB		 3409 if (info->array.level == 5 &&
3410 (dev->rwh_policy == RWH_DISTRIBUTED ||
3411 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3412 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3413 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3414 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3415 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3416 else
3417 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3418 >> 9;
2432ce9b
AP		 3419 } else if (info->array.level <= 0) {
3420 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3421 } else {
3422 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3423 }
3424
d2e6d5d6 3425 info->reshape_progress = 0;
b6796ce1 3426 info->resync_start = MaxSector;
b9172665 3427 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3428 !(info->array.state & 1)) &&
b9172665 3429 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3430 info->resync_start = 0;
b6796ce1
AK		 3431 }
3432 if (dev->vol.migr_state) {
1e5c6983
DW		 3433 switch (migr_type(dev)) {
3434 case MIGR_REPAIR:
3435 case MIGR_INIT: {
c47b0ff6
AK		 3436 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3437 dev);
1e5c6983
DW		 3438 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3439
3440 info->resync_start = blocks_per_unit * units;
3441 break;
3442 }
d2e6d5d6 3443 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3444 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3445 dev);
9f421827 3446 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3447 unsigned long long array_blocks;
3448 int used_disks;
d2e6d5d6 3449
befb629b
AK		 3450 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3451 (units <
9f421827 3452 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3453 (super->migr_rec->rec_status ==
3454 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3455 units++;
3456
d2e6d5d6 3457 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3458
7a862a02 3459 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3460 (unsigned long long)units,
3461 (unsigned long long)blocks_per_unit,
3462 info->reshape_progress);
75156c46 3463
9529d343 3464 used_disks = imsm_num_data_members(prev_map);
75156c46 3465 if (used_disks > 0) {
44490938 3466 array_blocks = per_dev_array_size(map) *
75156c46 3467 used_disks;
b53bfba6
TM		 3468 info->custom_array_size =
3469 round_size_to_mb(array_blocks,
3470 used_disks);
3471
75156c46 3472 }
d2e6d5d6 3473 }
1e5c6983
DW		 3474 case MIGR_VERIFY:
3475 /* we could emulate the checkpointing of
3476 * 'sync_action=check' migrations, but for now
3477 * we just immediately complete them
3478 */
3479 case MIGR_REBUILD:
3480 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3481 case MIGR_STATE_CHANGE:
3482 /* FIXME handle other migrations */
3483 default:
3484 /* we are not dirty, so... */
3485 info->resync_start = MaxSector;
3486 }
b6796ce1 3487 }
301406c9
DW		 3488
3489 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3490 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3491
f35f2525
N		 3492 info->array.major_version = -1;
3493 info->array.minor_version = -2;
4dd2df09 3494 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3495 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3496 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3497
3498 if (dmap) {
3499 int i, j;
3500 for (i=0; i<map_disks; i++) {
3501 dmap[i] = 0;
3502 if (i < info->array.raid_disks) {
3503 struct imsm_disk *dsk;
238c0a71 3504 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3505 dsk = get_imsm_disk(super, j);
3506 if (dsk && (dsk->status & CONFIGURED_DISK))
3507 dmap[i] = 1;
3508 }
3509 }
3510 }
81ac8b4d 3511}
bf5a934a 3512
3b451610
AK		 3513static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3514 int failed, int look_in_map);
3515
3516static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3517 int look_in_map);
3518
3519static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3520{
3521 if (is_gen_migration(dev)) {
3522 int failed;
3523 __u8 map_state;
3524 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3525
3526 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3527 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3528 if (map2->map_state != map_state) {
3529 map2->map_state = map_state;
3530 super->updates_pending++;
3531 }
3532 }
3533}
97b4d0e9
DW		 3534
3535static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3536{
3537 struct dl *d;
3538
3539 for (d = super->missing; d; d = d->next)
3540 if (d->index == index)
3541 return &d->disk;
3542 return NULL;
3543}
3544
a5d85af7 3545static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3546{
3547 struct intel_super *super = st->sb;
4f5bc454 3548 struct imsm_disk *disk;
a5d85af7 3549 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3550 int max_enough = -1;
3551 int i;
3552 struct imsm_super *mpb;
4f5bc454 3553
bf5a934a 3554 if (super->current_vol >= 0) {
a5d85af7 3555 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3556 return;
3557 }
95eeceeb 3558 memset(info, 0, sizeof(*info));
d23fe947
DW		 3559
3560 /* Set raid_disks to zero so that Assemble will always pull in valid
3561 * spares
3562 */
3563 info->array.raid_disks = 0;
cdddbdbc
DW		 3564 info->array.level = LEVEL_CONTAINER;
3565 info->array.layout = 0;
3566 info->array.md_minor = -1;
1011e834 3567 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3568 info->array.utime = 0;
3569 info->array.chunk_size = 0;
3570
3571 info->disk.major = 0;
3572 info->disk.minor = 0;
cdddbdbc 3573 info->disk.raid_disk = -1;
c2c087e6 3574 info->reshape_active = 0;
f35f2525
N		 3575 info->array.major_version = -1;
3576 info->array.minor_version = -2;
c2c087e6 3577 strcpy(info->text_version, "imsm");
a67dd8cc 3578 info->safe_mode_delay = 0;
c2c087e6
DW		 3579 info->disk.number = -1;
3580 info->disk.state = 0;
c5afc314 3581 info->name[0] = 0;
921d9e16 3582 info->recovery_start = MaxSector;
3ad25638 3583 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3584 info->bb.supported = 1;
c2c087e6 3585
97b4d0e9 3586 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3587 mpb = super->anchor;
b7d81a38 3588 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3589
ab3cb6b3
N		 3590 for (i = 0; i < mpb->num_raid_devs; i++) {
3591 struct imsm_dev *dev = get_imsm_dev(super, i);
3592 int failed, enough, j, missing = 0;
3593 struct imsm_map *map;
3594 __u8 state;
97b4d0e9 3595
3b451610
AK		 3596 failed = imsm_count_failed(super, dev, MAP_0);
3597 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3598 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3599
3600 /* any newly missing disks?
3601 * (catches single-degraded vs double-degraded)
3602 */
3603 for (j = 0; j < map->num_members; j++) {
238c0a71 3604 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3605 __u32 idx = ord_to_idx(ord);
3606
20dc76d1
MT		 3607 if (super->disks && super->disks->index == (int)idx)
3608 info->disk.raid_disk = j;
3609
ab3cb6b3
N		 3610 if (!(ord & IMSM_ORD_REBUILD) &&
3611 get_imsm_missing(super, idx)) {
3612 missing = 1;
3613 break;
3614 }
97b4d0e9 3615 }
ab3cb6b3
N		 3616
3617 if (state == IMSM_T_STATE_FAILED)
3618 enough = -1;
3619 else if (state == IMSM_T_STATE_DEGRADED &&
3620 (state != map->map_state || missing))
3621 enough = 0;
3622 else /* we're normal, or already degraded */
3623 enough = 1;
d2bde6d3
AK		 3624 if (is_gen_migration(dev) && missing) {
3625 /* during general migration we need all disks
3626 * that process is running on.
3627 * No new missing disk is allowed.
3628 */
3629 max_enough = -1;
3630 enough = -1;
3631 /* no more checks necessary
3632 */
3633 break;
3634 }
ab3cb6b3
N		 3635 /* in the missing/failed disk case check to see
3636 * if at least one array is runnable
3637 */
3638 max_enough = max(max_enough, enough);
3639 }
1ade5cc1 3640 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3641 info->container_enough = max_enough;
97b4d0e9 3642
4a04ec6c 3643 if (super->disks) {
14e8215b
DW		 3644 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3645
b9f594fe 3646 disk = &super->disks->disk;
5551b113 3647 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3648 info->component_size = reserved;
25ed7e59 3649 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3650 /* we don't change info->disk.raid_disk here because
3651 * this state will be finalized in mdmon after we have
3652 * found the 'most fresh' version of the metadata
3653 */
25ed7e59 3654 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3655 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3656 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3657 }
a575e2a7
DW		 3658
3659 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3660 * ->compare_super may have updated the 'num_raid_devs' field for spares
3661 */
3662 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3663 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3664 else
3665 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3666
3667 /* I don't know how to compute 'map' on imsm, so use safe default */
3668 if (map) {
3669 int i;
3670 for (i = 0; i < map_disks; i++)
3671 map[i] = 1;
3672 }
3673
cdddbdbc
DW		 3674}
3675
5c4cd5da
AC		 3676/* allocates memory and fills disk in mdinfo structure
3677 * for each disk in array */
3678struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3679{
594dc1b8 3680 struct mdinfo *mddev;
5c4cd5da
AC		 3681 struct intel_super *super = st->sb;
3682 struct imsm_disk *disk;
3683 int count = 0;
3684 struct dl *dl;
3685 if (!super || !super->disks)
3686 return NULL;
3687 dl = super->disks;
503975b9 3688 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3689 while (dl) {
3690 struct mdinfo *tmp;
3691 disk = &dl->disk;
503975b9 3692 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3693 if (mddev->devs)
3694 tmp->next = mddev->devs;
3695 mddev->devs = tmp;
3696 tmp->disk.number = count++;
3697 tmp->disk.major = dl->major;
3698 tmp->disk.minor = dl->minor;
3699 tmp->disk.state = is_configured(disk) ?
3700 (1 << MD_DISK_ACTIVE) : 0;
3701 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3702 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3703 tmp->disk.raid_disk = -1;
3704 dl = dl->next;
3705 }
3706 return mddev;
3707}
3708
cdddbdbc
DW		 3709static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3710 char *update, char *devname, int verbose,
3711 int uuid_set, char *homehost)
3712{
f352c545
DW		 3713 /* For 'assemble' and 'force' we need to return non-zero if any
3714 * change was made. For others, the return value is ignored.
3715 * Update options are:
3716 * force-one : This device looks a bit old but needs to be included,
3717 * update age info appropriately.
3718 * assemble: clear any 'faulty' flag to allow this device to
3719 * be assembled.
3720 * force-array: Array is degraded but being forced, mark it clean
3721 * if that will be needed to assemble it.
3722 *
3723 * newdev: not used ????
3724 * grow: Array has gained a new device - this is currently for
3725 * linear only
3726 * resync: mark as dirty so a resync will happen.
3727 * name: update the name - preserving the homehost
6e46bf34 3728 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3729 *
3730 * Following are not relevant for this imsm:
3731 * sparc2.2 : update from old dodgey metadata
3732 * super-minor: change the preferred_minor number
3733 * summaries: update redundant counters.
f352c545
DW		 3734 * homehost: update the recorded homehost
3735 * _reshape_progress: record new reshape_progress position.
3736 */
6e46bf34
DW		 3737 int rv = 1;
3738 struct intel_super *super = st->sb;
3739 struct imsm_super *mpb;
f352c545 3740
6e46bf34
DW		 3741 /* we can only update container info */
3742 if (!super || super->current_vol >= 0 || !super->anchor)
3743 return 1;
3744
3745 mpb = super->anchor;
3746
81a5b4f5
N		 3747 if (strcmp(update, "uuid") == 0) {
3748 /* We take this to mean that the family_num should be updated.
3749 * However that is much smaller than the uuid so we cannot really
3750 * allow an explicit uuid to be given. And it is hard to reliably
3751 * know if one was.
3752 * So if !uuid_set we know the current uuid is random and just used
3753 * the first 'int' and copy it to the other 3 positions.
3754 * Otherwise we require the 4 'int's to be the same as would be the
3755 * case if we are using a random uuid. So an explicit uuid will be
3756 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3757 */
81a5b4f5
N		 3758 if (!uuid_set) {
3759 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3760 rv = 0;
81a5b4f5
N		 3761 } else {
3762 if (info->uuid[0] != info->uuid[1] ||
3763 info->uuid[1] != info->uuid[2] ||
3764 info->uuid[2] != info->uuid[3])
3765 rv = -1;
3766 else
3767 rv = 0;
6e46bf34 3768 }
81a5b4f5
N		 3769 if (rv == 0)
3770 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3771 } else if (strcmp(update, "assemble") == 0)
3772 rv = 0;
3773 else
1e2b2765 3774 rv = -1;
f352c545 3775
6e46bf34
DW		 3776 /* successful update? recompute checksum */
3777 if (rv == 0)
3778 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3779
3780 return rv;
cdddbdbc
DW		 3781}
3782
c2c087e6 3783static size_t disks_to_mpb_size(int disks)
cdddbdbc 3784{
c2c087e6 3785 size_t size;
cdddbdbc 3786
c2c087e6
DW		 3787 size = sizeof(struct imsm_super);
3788 size += (disks - 1) * sizeof(struct imsm_disk);
3789 size += 2 * sizeof(struct imsm_dev);
3790 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3791 size += (4 - 2) * sizeof(struct imsm_map);
3792 /* 4 possible disk_ord_tbl's */
3793 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3794 /* maximum bbm log */
3795 size += sizeof(struct bbm_log);
c2c087e6
DW		 3796
3797 return size;
3798}
3799
387fcd59
N		 3800static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3801 unsigned long long data_offset)
c2c087e6
DW		 3802{
3803 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3804 return 0;
3805
3806 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3807}
3808
ba2de7ba
DW		 3809static void free_devlist(struct intel_super *super)
3810{
3811 struct intel_dev *dv;
3812
3813 while (super->devlist) {
3814 dv = super->devlist->next;
3815 free(super->devlist->dev);
3816 free(super->devlist);
3817 super->devlist = dv;
3818 }
3819}
3820
3821static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3822{
3823 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3824}
3825
cdddbdbc
DW		 3826static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3827{
3828 /*
3829 * return:
3830 * 0 same, or first was empty, and second was copied
3831 * 1 second had wrong number
3832 * 2 wrong uuid
3833 * 3 wrong other info
3834 */
3835 struct intel_super *first = st->sb;
3836 struct intel_super *sec = tst->sb;
3837
5d500228
N		 3838 if (!first) {
3839 st->sb = tst->sb;
3840 tst->sb = NULL;
3841 return 0;
3842 }
8603ea6f
LM		 3843 /* in platform dependent environment test if the disks
3844 * use the same Intel hba
cb8f6859 3845 * If not on Intel hba at all, allow anything.
8603ea6f 3846 */
6b781d33
AP		 3847 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3848 if (first->hba->type != sec->hba->type) {
8603ea6f 3849 fprintf(stderr,
6b781d33
AP		 3850 "HBAs of devices do not match %s != %s\n",
3851 get_sys_dev_type(first->hba->type),
3852 get_sys_dev_type(sec->hba->type));
3853 return 3;
3854 }
3855 if (first->orom != sec->orom) {
3856 fprintf(stderr,
3857 "HBAs of devices do not match %s != %s\n",
3858 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3859 return 3;
3860 }
3861 }
cdddbdbc 3862
d23fe947
DW		 3863 /* if an anchor does not have num_raid_devs set then it is a free
3864 * floating spare
3865 */
3866 if (first->anchor->num_raid_devs > 0 &&
3867 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3868 /* Determine if these disks might ever have been
3869 * related. Further disambiguation can only take place
3870 * in load_super_imsm_all
3871 */
3872 __u32 first_family = first->anchor->orig_family_num;
3873 __u32 sec_family = sec->anchor->orig_family_num;
3874
f796af5d
DW		 3875 if (memcmp(first->anchor->sig, sec->anchor->sig,
3876 MAX_SIGNATURE_LENGTH) != 0)
3877 return 3;
3878
a2b97981
DW		 3879 if (first_family == 0)
3880 first_family = first->anchor->family_num;
3881 if (sec_family == 0)
3882 sec_family = sec->anchor->family_num;
3883
3884 if (first_family != sec_family)
d23fe947 3885 return 3;
f796af5d 3886
d23fe947 3887 }
cdddbdbc 3888
3e372e5a
DW		 3889 /* if 'first' is a spare promote it to a populated mpb with sec's
3890 * family number
3891 */
3892 if (first->anchor->num_raid_devs == 0 &&
3893 sec->anchor->num_raid_devs > 0) {
78d30f94 3894 int i;
ba2de7ba
DW		 3895 struct intel_dev *dv;
3896 struct imsm_dev *dev;
78d30f94
DW		 3897
3898 /* we need to copy raid device info from sec if an allocation
3899 * fails here we don't associate the spare
3900 */
3901 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3902 dv = xmalloc(sizeof(*dv));
3903 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3904 dv->dev = dev;
3905 dv->index = i;
3906 dv->next = first->devlist;
3907 first->devlist = dv;
78d30f94 3908 }
709743c5 3909 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3910 /* allocation failure */
3911 free_devlist(first);
e12b3daa 3912 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3913 return 3;
78d30f94 3914 }
3e372e5a 3915 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3916 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3917 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3918 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3919 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3920 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3921 }
3922
cdddbdbc
DW		 3923 return 0;
3924}
3925
0030e8d6
DW		 3926static void fd2devname(int fd, char *name)
3927{
3928 struct stat st;
3929 char path[256];
33a6535d 3930 char dname[PATH_MAX];
0030e8d6
DW		 3931 char *nm;
3932 int rv;
3933
3934 name[0] = '\0';
3935 if (fstat(fd, &st) != 0)
3936 return;
3937 sprintf(path, "/sys/dev/block/%d:%d",
3938 major(st.st_rdev), minor(st.st_rdev));
3939
9cf014ec 3940 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3941 if (rv <= 0)
3942 return;
9587c373 3943
0030e8d6
DW		 3944 dname[rv] = '\0';
3945 nm = strrchr(dname, '/');
7897de29
JS		 3946 if (nm) {
3947 nm++;
3948 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3949 }
0030e8d6
DW		 3950}
3951
21e9380b
AP		 3952static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3953{
3954 char path[60];
3955 char *name = fd2kname(fd);
3956
3957 if (!name)
3958 return 1;
3959
3960 if (strncmp(name, "nvme", 4) != 0)
3961 return 1;
3962
3963 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3964
3965 return load_sys(path, buf, buf_len);
3966}
3967
cdddbdbc
DW		 3968extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3969
3970static int imsm_read_serial(int fd, char *devname,
6da53c0e 3971 __u8 *serial, size_t serial_buf_len)
cdddbdbc 3972{
21e9380b 3973 char buf[50];
cdddbdbc 3974 int rv;
6da53c0e 3975 size_t len;
316e2bf4
DW		 3976 char *dest;
3977 char *src;
21e9380b
AP		 3978 unsigned int i;
3979
3980 memset(buf, 0, sizeof(buf));
cdddbdbc 3981
21e9380b 3982 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3983
21e9380b
AP		 3984 if (rv)
3985 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3986
40ebbb9c 3987 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3988 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3989 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3990 return 0;
3991 }
3992
cdddbdbc
DW		 3993 if (rv != 0) {
3994 if (devname)
e7b84f9d
N		 3995 pr_err("Failed to retrieve serial for %s\n",
3996 devname);
cdddbdbc
DW		 3997 return rv;
3998 }
3999
316e2bf4
DW		 4000 /* trim all whitespace and non-printable characters and convert
4001 * ':' to ';'
4002 */
21e9380b
AP		 4003 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4004 src = &buf[i];
316e2bf4
DW		 4005 if (*src > 0x20) {
4006 /* ':' is reserved for use in placeholder serial
4007 * numbers for missing disks
4008 */
4009 if (*src == ':')
4010 *dest++ = ';';
4011 else
4012 *dest++ = *src;
4013 }
4014 }
21e9380b
AP		 4015 len = dest - buf;
4016 dest = buf;
316e2bf4 4017
6da53c0e
BK		 4018 if (len > serial_buf_len) {
4019 /* truncate leading characters */
4020 dest += len - serial_buf_len;
4021 len = serial_buf_len;
316e2bf4 4022 }
5c3db629 4023
6da53c0e 4024 memset(serial, 0, serial_buf_len);
316e2bf4 4025 memcpy(serial, dest, len);
cdddbdbc
DW		 4026
4027 return 0;
4028}
4029
1f24f035
DW		 4030static int serialcmp(__u8 *s1, __u8 *s2)
4031{
4032 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4033}
4034
4035static void serialcpy(__u8 *dest, __u8 *src)
4036{
4037 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4038}
4039
54c2c1ea
DW		 4040static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4041{
4042 struct dl *dl;
4043
4044 for (dl = super->disks; dl; dl = dl->next)
4045 if (serialcmp(dl->serial, serial) == 0)
4046 break;
4047
4048 return dl;
4049}
4050
a2b97981
DW		 4051static struct imsm_disk *
4052__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4053{
4054 int i;
4055
4056 for (i = 0; i < mpb->num_disks; i++) {
4057 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4058
4059 if (serialcmp(disk->serial, serial) == 0) {
4060 if (idx)
4061 *idx = i;
4062 return disk;
4063 }
4064 }
4065
4066 return NULL;
4067}
4068
cdddbdbc
DW		 4069static int
4070load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4071{
a2b97981 4072 struct imsm_disk *disk;
cdddbdbc
DW		 4073 struct dl *dl;
4074 struct stat stb;
cdddbdbc 4075 int rv;
a2b97981 4076 char name[40];
d23fe947
DW		 4077 __u8 serial[MAX_RAID_SERIAL_LEN];
4078
6da53c0e 4079 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4080
4081 if (rv != 0)
4082 return 2;
4083
503975b9 4084 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4085
a2b97981
DW		 4086 fstat(fd, &stb);
4087 dl->major = major(stb.st_rdev);
4088 dl->minor = minor(stb.st_rdev);
4089 dl->next = super->disks;
4090 dl->fd = keep_fd ? fd : -1;
4091 assert(super->disks == NULL);
4092 super->disks = dl;
4093 serialcpy(dl->serial, serial);
4094 dl->index = -2;
4095 dl->e = NULL;
4096 fd2devname(fd, name);
4097 if (devname)
503975b9 4098 dl->devname = xstrdup(devname);
a2b97981 4099 else
503975b9 4100 dl->devname = xstrdup(name);
cdddbdbc 4101
d23fe947 4102 /* look up this disk's index in the current anchor */
a2b97981
DW		 4103 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4104 if (disk) {
4105 dl->disk = *disk;
4106 /* only set index on disks that are a member of a
4107 * populated contianer, i.e. one with raid_devs
4108 */
4109 if (is_failed(&dl->disk))
3f6efecc 4110 dl->index = -2;
2432ce9b 4111 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4112 dl->index = -1;
3f6efecc
DW		 4113 }
4114
949c47a0
DW		 4115 return 0;
4116}
4117
0c046afd
DW		 4118/* When migrating map0 contains the 'destination' state while map1
4119 * contains the current state. When not migrating map0 contains the
4120 * current state. This routine assumes that map[0].map_state is set to
4121 * the current array state before being called.
4122 *
4123 * Migration is indicated by one of the following states
4124 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4125 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4126 * map1state=unitialized)
1484e727 4127 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4128 * map1state=normal)
e3bba0e0 4129 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4130 * map1state=degraded)
8e59f3d8
AK		 4131 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4132 * map1state=normal)
0c046afd 4133 */
8e59f3d8
AK		 4134static void migrate(struct imsm_dev *dev, struct intel_super *super,
4135 __u8 to_state, int migr_type)
3393c6af 4136{
0c046afd 4137 struct imsm_map *dest;
238c0a71 4138 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4139
0c046afd 4140 dev->vol.migr_state = 1;
1484e727 4141 set_migr_type(dev, migr_type);
f8f603f1 4142 dev->vol.curr_migr_unit = 0;
238c0a71 4143 dest = get_imsm_map(dev, MAP_1);
0c046afd 4144
0556e1a2 4145 /* duplicate and then set the target end state in map[0] */
3393c6af 4146 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4147 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4148 __u32 ord;
4149 int i;
4150
4151 for (i = 0; i < src->num_members; i++) {
4152 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4153 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4154 }
4155 }
4156
8e59f3d8
AK		 4157 if (migr_type == MIGR_GEN_MIGR)
4158 /* Clear migration record */
4159 memset(super->migr_rec, 0, sizeof(struct migr_record));
4160
0c046afd 4161 src->map_state = to_state;
949c47a0 4162}
f8f603f1 4163
809da78e
AK		 4164static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4165 __u8 map_state)
f8f603f1 4166{
238c0a71
AK		 4167 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4168 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4169 MAP_0 : MAP_1);
28bce06f 4170 int i, j;
0556e1a2
DW		 4171
4172 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4173 * completed in the last migration.
4174 *
28bce06f 4175 * FIXME add support for raid-level-migration
0556e1a2 4176 */
089f9d79
JS		 4177 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4178 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4179 /* when final map state is other than expected
4180 * merge maps (not for migration)
4181 */
4182 int failed;
4183
4184 for (i = 0; i < prev->num_members; i++)
4185 for (j = 0; j < map->num_members; j++)
4186 /* during online capacity expansion
4187 * disks position can be changed
4188 * if takeover is used
4189 */
4190 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4191 ord_to_idx(prev->disk_ord_tbl[i])) {
4192 map->disk_ord_tbl[j] |=
4193 prev->disk_ord_tbl[i];
4194 break;
4195 }
4196 failed = imsm_count_failed(super, dev, MAP_0);
4197 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4198 }
f8f603f1
DW		 4199
4200 dev->vol.migr_state = 0;
ea672ee1 4201 set_migr_type(dev, 0);
f8f603f1
DW		 4202 dev->vol.curr_migr_unit = 0;
4203 map->map_state = map_state;
4204}
949c47a0
DW		 4205
4206static int parse_raid_devices(struct intel_super *super)
4207{
4208 int i;
4209 struct imsm_dev *dev_new;
4d7b1503 4210 size_t len, len_migr;
401d313b 4211 size_t max_len = 0;
4d7b1503
DW		 4212 size_t space_needed = 0;
4213 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4214
4215 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4216 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4217 struct intel_dev *dv;
949c47a0 4218
4d7b1503
DW		 4219 len = sizeof_imsm_dev(dev_iter, 0);
4220 len_migr = sizeof_imsm_dev(dev_iter, 1);
4221 if (len_migr > len)
4222 space_needed += len_migr - len;
ca9de185 4223
503975b9 4224 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4225 if (max_len < len_migr)
4226 max_len = len_migr;
4227 if (max_len > len_migr)
4228 space_needed += max_len - len_migr;
503975b9 4229 dev_new = xmalloc(max_len);
949c47a0 4230 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4231 dv->dev = dev_new;
4232 dv->index = i;
4233 dv->next = super->devlist;
4234 super->devlist = dv;
949c47a0 4235 }
cdddbdbc 4236
4d7b1503
DW		 4237 /* ensure that super->buf is large enough when all raid devices
4238 * are migrating
4239 */
4240 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4241 void *buf;
4242
f36a9ecd
PB		 4243 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4244 super->sector_size);
4245 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4246 return 1;
4247
1f45a8ad
DW		 4248 memcpy(buf, super->buf, super->len);
4249 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4250 free(super->buf);
4251 super->buf = buf;
4252 super->len = len;
4253 }
ca9de185 4254
bbab0940
TM		 4255 super->extra_space += space_needed;
4256
cdddbdbc
DW		 4257 return 0;
4258}
4259
e2f41b2c
AK		 4260/*******************************************************************************
4261 * Function: check_mpb_migr_compatibility
4262 * Description: Function checks for unsupported migration features:
4263 * - migration optimization area (pba_of_lba0)
4264 * - descending reshape (ascending_migr)
4265 * Parameters:
4266 * super : imsm metadata information
4267 * Returns:
4268 * 0 : migration is compatible
4269 * -1 : migration is not compatible
4270 ******************************************************************************/
4271int check_mpb_migr_compatibility(struct intel_super *super)
4272{
4273 struct imsm_map *map0, *map1;
4274 struct migr_record *migr_rec = super->migr_rec;
4275 int i;
4276
4277 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4278 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4279
4280 if (dev_iter &&
4281 dev_iter->vol.migr_state == 1 &&
4282 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4283 /* This device is migrating */
238c0a71
AK		 4284 map0 = get_imsm_map(dev_iter, MAP_0);
4285 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4286 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4287 /* migration optimization area was used */
4288 return -1;
fc54fe7a
JS		 4289 if (migr_rec->ascending_migr == 0 &&
4290 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4291 /* descending reshape not supported yet */
4292 return -1;
4293 }
4294 }
4295 return 0;
4296}
4297
d23fe947 4298static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4299
cdddbdbc 4300/* load_imsm_mpb - read matrix metadata
f2f5c343 4301 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4302 */
4303static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4304{
4305 unsigned long long dsize;
cdddbdbc 4306 unsigned long long sectors;
f36a9ecd 4307 unsigned int sector_size = super->sector_size;
cdddbdbc 4308 struct stat;
6416d527 4309 struct imsm_super *anchor;
cdddbdbc
DW		 4310 __u32 check_sum;
4311
cdddbdbc 4312 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4313 if (dsize < 2*sector_size) {
64436f06 4314 if (devname)
e7b84f9d
N		 4315 pr_err("%s: device to small for imsm\n",
4316 devname);
64436f06
N		 4317 return 1;
4318 }
cdddbdbc 4319
f36a9ecd 4320 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4321 if (devname)
e7b84f9d
N		 4322 pr_err("Cannot seek to anchor block on %s: %s\n",
4323 devname, strerror(errno));
cdddbdbc
DW		 4324 return 1;
4325 }
4326
f36a9ecd 4327 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4328 if (devname)
7a862a02 4329 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4330 return 1;
4331 }
466070ad 4332 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4333 if (devname)
e7b84f9d
N		 4334 pr_err("Cannot read anchor block on %s: %s\n",
4335 devname, strerror(errno));
6416d527 4336 free(anchor);
cdddbdbc
DW		 4337 return 1;
4338 }
4339
6416d527 4340 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4341 if (devname)
e7b84f9d 4342 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4343 free(anchor);
cdddbdbc
DW		 4344 return 2;
4345 }
4346
d23fe947 4347 __free_imsm(super, 0);
f2f5c343
LM		 4348 /* reload capability and hba */
4349
4350 /* capability and hba must be updated with new super allocation */
d424212e 4351 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4352 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4353 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4354 if (devname)
e7b84f9d
N		 4355 pr_err("unable to allocate %zu byte mpb buffer\n",
4356 super->len);
6416d527 4357 free(anchor);
cdddbdbc
DW		 4358 return 2;
4359 }
f36a9ecd 4360 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4361
f36a9ecd 4362 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4363 free(anchor);
8e59f3d8 4364
85337573
AO		 4365 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4366 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4367 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4368 free(super->buf);
4369 return 2;
4370 }
51d83f5d 4371 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4372
949c47a0 4373 if (!sectors) {
ecf45690
DW		 4374 check_sum = __gen_imsm_checksum(super->anchor);
4375 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4376 if (devname)
e7b84f9d
N		 4377 pr_err("IMSM checksum %x != %x on %s\n",
4378 check_sum,
4379 __le32_to_cpu(super->anchor->check_sum),
4380 devname);
ecf45690
DW		 4381 return 2;
4382 }
4383
a2b97981 4384 return 0;
949c47a0 4385 }
cdddbdbc
DW		 4386
4387 /* read the extended mpb */
f36a9ecd 4388 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4389 if (devname)
e7b84f9d
N		 4390 pr_err("Cannot seek to extended mpb on %s: %s\n",
4391 devname, strerror(errno));
cdddbdbc
DW		 4392 return 1;
4393 }
4394
f36a9ecd
PB		 4395 if ((unsigned int)read(fd, super->buf + sector_size,
4396 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4397 if (devname)
e7b84f9d
N		 4398 pr_err("Cannot read extended mpb on %s: %s\n",
4399 devname, strerror(errno));
cdddbdbc
DW		 4400 return 2;
4401 }
4402
949c47a0
DW		 4403 check_sum = __gen_imsm_checksum(super->anchor);
4404 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4405 if (devname)
e7b84f9d
N		 4406 pr_err("IMSM checksum %x != %x on %s\n",
4407 check_sum, __le32_to_cpu(super->anchor->check_sum),
4408 devname);
db575f3b 4409 return 3;
cdddbdbc
DW		 4410 }
4411
a2b97981
DW		 4412 return 0;
4413}
4414
8e59f3d8
AK		 4415static int read_imsm_migr_rec(int fd, struct intel_super *super);
4416
97f81ee2
CA		 4417/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4418static void clear_hi(struct intel_super *super)
4419{
4420 struct imsm_super *mpb = super->anchor;
4421 int i, n;
4422 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4423 return;
4424 for (i = 0; i < mpb->num_disks; ++i) {
4425 struct imsm_disk *disk = &mpb->disk[i];
4426 disk->total_blocks_hi = 0;
4427 }
4428 for (i = 0; i < mpb->num_raid_devs; ++i) {
4429 struct imsm_dev *dev = get_imsm_dev(super, i);
4430 if (!dev)
4431 return;
4432 for (n = 0; n < 2; ++n) {
4433 struct imsm_map *map = get_imsm_map(dev, n);
4434 if (!map)
4435 continue;
4436 map->pba_of_lba0_hi = 0;
4437 map->blocks_per_member_hi = 0;
4438 map->num_data_stripes_hi = 0;
4439 }
4440 }
4441}
4442
a2b97981
DW		 4443static int
4444load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4445{
4446 int err;
4447
4448 err = load_imsm_mpb(fd, super, devname);
4449 if (err)
4450 return err;
f36a9ecd
PB		 4451 if (super->sector_size == 4096)
4452 convert_from_4k(super);
a2b97981
DW		 4453 err = load_imsm_disk(fd, super, devname, keep_fd);
4454 if (err)
4455 return err;
4456 err = parse_raid_devices(super);
8d67477f
TM		 4457 if (err)
4458 return err;
4459 err = load_bbm_log(super);
97f81ee2 4460 clear_hi(super);
a2b97981 4461 return err;
cdddbdbc
DW		 4462}
4463
ae6aad82
DW		 4464static void __free_imsm_disk(struct dl *d)
4465{
4466 if (d->fd >= 0)
4467 close(d->fd);
4468 if (d->devname)
4469 free(d->devname);
0dcecb2e
DW		 4470 if (d->e)
4471 free(d->e);
ae6aad82
DW		 4472 free(d);
4473
4474}
1a64be56 4475
cdddbdbc
DW		 4476static void free_imsm_disks(struct intel_super *super)
4477{
47ee5a45 4478 struct dl *d;
cdddbdbc 4479
47ee5a45
DW		 4480 while (super->disks) {
4481 d = super->disks;
cdddbdbc 4482 super->disks = d->next;
ae6aad82 4483 __free_imsm_disk(d);
cdddbdbc 4484 }
cb82edca
AK		 4485 while (super->disk_mgmt_list) {
4486 d = super->disk_mgmt_list;
4487 super->disk_mgmt_list = d->next;
4488 __free_imsm_disk(d);
4489 }
47ee5a45
DW		 4490 while (super->missing) {
4491 d = super->missing;
4492 super->missing = d->next;
4493 __free_imsm_disk(d);
4494 }
4495
cdddbdbc
DW		 4496}
4497
9ca2c81c 4498/* free all the pieces hanging off of a super pointer */
d23fe947 4499static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4500{
88654014
LM		 4501 struct intel_hba *elem, *next;
4502
9ca2c81c 4503 if (super->buf) {
949c47a0 4504 free(super->buf);
9ca2c81c
DW		 4505 super->buf = NULL;
4506 }
f2f5c343
LM		 4507 /* unlink capability description */
4508 super->orom = NULL;
8e59f3d8
AK		 4509 if (super->migr_rec_buf) {
4510 free(super->migr_rec_buf);
4511 super->migr_rec_buf = NULL;
4512 }
d23fe947
DW		 4513 if (free_disks)
4514 free_imsm_disks(super);
ba2de7ba 4515 free_devlist(super);
88654014
LM		 4516 elem = super->hba;
4517 while (elem) {
4518 if (elem->path)
4519 free((void *)elem->path);
4520 next = elem->next;
4521 free(elem);
4522 elem = next;
88c32bb1 4523 }
8d67477f
TM		 4524 if (super->bbm_log)
4525 free(super->bbm_log);
88654014 4526 super->hba = NULL;
cdddbdbc
DW		 4527}
4528
9ca2c81c
DW		 4529static void free_imsm(struct intel_super *super)
4530{
d23fe947 4531 __free_imsm(super, 1);
928f1424 4532 free(super->bb.entries);
9ca2c81c
DW		 4533 free(super);
4534}
cdddbdbc
DW		 4535
4536static void free_super_imsm(struct supertype *st)
4537{
4538 struct intel_super *super = st->sb;
4539
4540 if (!super)
4541 return;
4542
4543 free_imsm(super);
4544 st->sb = NULL;
4545}
4546
49133e57 4547static struct intel_super *alloc_super(void)
c2c087e6 4548{
503975b9 4549 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4550
503975b9
N		 4551 super->current_vol = -1;
4552 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4553
4554 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4555 sizeof(struct md_bb_entry));
4556 if (!super->bb.entries) {
4557 free(super);
4558 return NULL;
4559 }
4560
c2c087e6
DW		 4561 return super;
4562}
4563
f0f5a016
LM		 4564/*
4565 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4566 */
d424212e 4567static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4568{
4569 struct sys_dev *hba_name;
4570 int rv = 0;
4571
3a30e28e
MT		 4572 if (fd >= 0 && test_partition(fd)) {
4573 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4574 devname);
4575 return 1;
4576 }
089f9d79 4577 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4578 super->orom = NULL;
f0f5a016
LM		 4579 super->hba = NULL;
4580 return 0;
4581 }
4582 hba_name = find_disk_attached_hba(fd, NULL);
4583 if (!hba_name) {
d424212e 4584 if (devname)
e7b84f9d
N		 4585 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4586 devname);
f0f5a016
LM		 4587 return 1;
4588 }
4589 rv = attach_hba_to_super(super, hba_name);
4590 if (rv == 2) {
d424212e
N		 4591 if (devname) {
4592 struct intel_hba *hba = super->hba;
f0f5a016 4593
60f0f54d
PB		 4594 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4595 " but the container is assigned to Intel(R) %s %s (",
d424212e 4596 devname,
614902f6 4597 get_sys_dev_type(hba_name->type),
60f0f54d 4598 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4599 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4600 get_sys_dev_type(super->hba->type),
4601 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4602
f0f5a016
LM		 4603 while (hba) {
4604 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4605 if (hba->next)
4606 fprintf(stderr, ", ");
4607 hba = hba->next;
4608 }
6b781d33 4609 fprintf(stderr, ").\n"
cca67208 4610 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4611 }
f0f5a016
LM		 4612 return 2;
4613 }
6b781d33 4614 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4615 if (!super->orom)
4616 return 3;
614902f6 4617
f0f5a016
LM		 4618 return 0;
4619}
4620
47ee5a45
DW		 4621/* find_missing - helper routine for load_super_imsm_all that identifies
4622 * disks that have disappeared from the system. This routine relies on
4623 * the mpb being uptodate, which it is at load time.
4624 */
4625static int find_missing(struct intel_super *super)
4626{
4627 int i;
4628 struct imsm_super *mpb = super->anchor;
4629 struct dl *dl;
4630 struct imsm_disk *disk;
47ee5a45
DW		 4631
4632 for (i = 0; i < mpb->num_disks; i++) {
4633 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4634 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4635 if (dl)
4636 continue;
47ee5a45 4637
503975b9 4638 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4639 dl->major = 0;
4640 dl->minor = 0;
4641 dl->fd = -1;
503975b9 4642 dl->devname = xstrdup("missing");
47ee5a45
DW		 4643 dl->index = i;
4644 serialcpy(dl->serial, disk->serial);
4645 dl->disk = *disk;
689c9bf3 4646 dl->e = NULL;
47ee5a45
DW		 4647 dl->next = super->missing;
4648 super->missing = dl;
4649 }
4650
4651 return 0;
4652}
4653
a2b97981
DW		 4654static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4655{
4656 struct intel_disk *idisk = disk_list;
4657
4658 while (idisk) {
4659 if (serialcmp(idisk->disk.serial, serial) == 0)
4660 break;
4661 idisk = idisk->next;
4662 }
4663
4664 return idisk;
4665}
4666
4667static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4668 struct intel_super *super,
4669 struct intel_disk **disk_list)
4670{
4671 struct imsm_disk *d = &super->disks->disk;
4672 struct imsm_super *mpb = super->anchor;
4673 int i, j;
4674
4675 for (i = 0; i < tbl_size; i++) {
4676 struct imsm_super *tbl_mpb = table[i]->anchor;
4677 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4678
4679 if (tbl_mpb->family_num == mpb->family_num) {
4680 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4681 dprintf("mpb from %d:%d matches %d:%d\n",
4682 super->disks->major,
a2b97981
DW		 4683 super->disks->minor,
4684 table[i]->disks->major,
4685 table[i]->disks->minor);
4686 break;
4687 }
4688
4689 if (((is_configured(d) && !is_configured(tbl_d)) ||
4690 is_configured(d) == is_configured(tbl_d)) &&
4691 tbl_mpb->generation_num < mpb->generation_num) {
4692 /* current version of the mpb is a
4693 * better candidate than the one in
4694 * super_table, but copy over "cross
4695 * generational" status
4696 */
4697 struct intel_disk *idisk;
4698
1ade5cc1
N		 4699 dprintf("mpb from %d:%d replaces %d:%d\n",
4700 super->disks->major,
a2b97981
DW		 4701 super->disks->minor,
4702 table[i]->disks->major,
4703 table[i]->disks->minor);
4704
4705 idisk = disk_list_get(tbl_d->serial, *disk_list);
4706 if (idisk && is_failed(&idisk->disk))
4707 tbl_d->status |= FAILED_DISK;
4708 break;
4709 } else {
4710 struct intel_disk *idisk;
4711 struct imsm_disk *disk;
4712
4713 /* tbl_mpb is more up to date, but copy
4714 * over cross generational status before
4715 * returning
4716 */
4717 disk = __serial_to_disk(d->serial, mpb, NULL);
4718 if (disk && is_failed(disk))
4719 d->status |= FAILED_DISK;
4720
4721 idisk = disk_list_get(d->serial, *disk_list);
4722 if (idisk) {
4723 idisk->owner = i;
4724 if (disk && is_configured(disk))
4725 idisk->disk.status |= CONFIGURED_DISK;
4726 }
4727
1ade5cc1
N		 4728 dprintf("mpb from %d:%d prefer %d:%d\n",
4729 super->disks->major,
a2b97981
DW		 4730 super->disks->minor,
4731 table[i]->disks->major,
4732 table[i]->disks->minor);
4733
4734 return tbl_size;
4735 }
4736 }
4737 }
4738
4739 if (i >= tbl_size)
4740 table[tbl_size++] = super;
4741 else
4742 table[i] = super;
4743
4744 /* update/extend the merged list of imsm_disk records */
4745 for (j = 0; j < mpb->num_disks; j++) {
4746 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4747 struct intel_disk *idisk;
4748
4749 idisk = disk_list_get(disk->serial, *disk_list);
4750 if (idisk) {
4751 idisk->disk.status |= disk->status;
4752 if (is_configured(&idisk->disk) ||
4753 is_failed(&idisk->disk))
4754 idisk->disk.status &= ~(SPARE_DISK);
4755 } else {
503975b9 4756 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4757 idisk->owner = IMSM_UNKNOWN_OWNER;
4758 idisk->disk = *disk;
4759 idisk->next = *disk_list;
4760 *disk_list = idisk;
4761 }
4762
4763 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4764 idisk->owner = i;
4765 }
4766
4767 return tbl_size;
4768}
4769
4770static struct intel_super *
4771validate_members(struct intel_super *super, struct intel_disk *disk_list,
4772 const int owner)
4773{
4774 struct imsm_super *mpb = super->anchor;
4775 int ok_count = 0;
4776 int i;
4777
4778 for (i = 0; i < mpb->num_disks; i++) {
4779 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4780 struct intel_disk *idisk;
4781
4782 idisk = disk_list_get(disk->serial, disk_list);
4783 if (idisk) {
4784 if (idisk->owner == owner ||
4785 idisk->owner == IMSM_UNKNOWN_OWNER)
4786 ok_count++;
4787 else
1ade5cc1
N		 4788 dprintf("'%.16s' owner %d != %d\n",
4789 disk->serial, idisk->owner,
a2b97981
DW		 4790 owner);
4791 } else {
1ade5cc1
N		 4792 dprintf("unknown disk %x [%d]: %.16s\n",
4793 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4794 disk->serial);
4795 break;
4796 }
4797 }
4798
4799 if (ok_count == mpb->num_disks)
4800 return super;
4801 return NULL;
4802}
4803
4804static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4805{
4806 struct intel_super *s;
4807
4808 for (s = super_list; s; s = s->next) {
4809 if (family_num != s->anchor->family_num)
4810 continue;
e12b3daa 4811 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4812 __le32_to_cpu(family_num), s->disks->devname);
4813 }
4814}
4815
4816static struct intel_super *
4817imsm_thunderdome(struct intel_super **super_list, int len)
4818{
4819 struct intel_super *super_table[len];
4820 struct intel_disk *disk_list = NULL;
4821 struct intel_super *champion, *spare;
4822 struct intel_super *s, **del;
4823 int tbl_size = 0;
4824 int conflict;
4825 int i;
4826
4827 memset(super_table, 0, sizeof(super_table));
4828 for (s = *super_list; s; s = s->next)
4829 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4830
4831 for (i = 0; i < tbl_size; i++) {
4832 struct imsm_disk *d;
4833 struct intel_disk *idisk;
4834 struct imsm_super *mpb = super_table[i]->anchor;
4835
4836 s = super_table[i];
4837 d = &s->disks->disk;
4838
4839 /* 'd' must appear in merged disk list for its
4840 * configuration to be valid
4841 */
4842 idisk = disk_list_get(d->serial, disk_list);
4843 if (idisk && idisk->owner == i)
4844 s = validate_members(s, disk_list, i);
4845 else
4846 s = NULL;
4847
4848 if (!s)
1ade5cc1
N		 4849 dprintf("marking family: %#x from %d:%d offline\n",
4850 mpb->family_num,
a2b97981
DW		 4851 super_table[i]->disks->major,
4852 super_table[i]->disks->minor);
4853 super_table[i] = s;
4854 }
4855
4856 /* This is where the mdadm implementation differs from the Windows
4857 * driver which has no strict concept of a container. We can only
4858 * assemble one family from a container, so when returning a prodigal
4859 * array member to this system the code will not be able to disambiguate
4860 * the container contents that should be assembled ("foreign" versus
4861 * "local"). It requires user intervention to set the orig_family_num
4862 * to a new value to establish a new container. The Windows driver in
4863 * this situation fixes up the volume name in place and manages the
4864 * foreign array as an independent entity.
4865 */
4866 s = NULL;
4867 spare = NULL;
4868 conflict = 0;
4869 for (i = 0; i < tbl_size; i++) {
4870 struct intel_super *tbl_ent = super_table[i];
4871 int is_spare = 0;
4872
4873 if (!tbl_ent)
4874 continue;
4875
4876 if (tbl_ent->anchor->num_raid_devs == 0) {
4877 spare = tbl_ent;
4878 is_spare = 1;
4879 }
4880
4881 if (s && !is_spare) {
4882 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4883 conflict++;
4884 } else if (!s && !is_spare)
4885 s = tbl_ent;
4886 }
4887
4888 if (!s)
4889 s = spare;
4890 if (!s) {
4891 champion = NULL;
4892 goto out;
4893 }
4894 champion = s;
4895
4896 if (conflict)
7a862a02 4897 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4898 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4899
4900 /* collect all dl's onto 'champion', and update them to
4901 * champion's version of the status
4902 */
4903 for (s = *super_list; s; s = s->next) {
4904 struct imsm_super *mpb = champion->anchor;
4905 struct dl *dl = s->disks;
4906
4907 if (s == champion)
4908 continue;
4909
5d7b407a
CA		 4910 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4911
a2b97981
DW		 4912 for (i = 0; i < mpb->num_disks; i++) {
4913 struct imsm_disk *disk;
4914
4915 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4916 if (disk) {
4917 dl->disk = *disk;
4918 /* only set index on disks that are a member of
4919 * a populated contianer, i.e. one with
4920 * raid_devs
4921 */
4922 if (is_failed(&dl->disk))
4923 dl->index = -2;
4924 else if (is_spare(&dl->disk))
4925 dl->index = -1;
4926 break;
4927 }
4928 }
4929
4930 if (i >= mpb->num_disks) {
4931 struct intel_disk *idisk;
4932
4933 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4934 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4935 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4936 dl->index = -1;
4937 else {
4938 dl->index = -2;
4939 continue;
4940 }
4941 }
4942
4943 dl->next = champion->disks;
4944 champion->disks = dl;
4945 s->disks = NULL;
4946 }
4947
4948 /* delete 'champion' from super_list */
4949 for (del = super_list; *del; ) {
4950 if (*del == champion) {
4951 *del = (*del)->next;
4952 break;
4953 } else
4954 del = &(*del)->next;
4955 }
4956 champion->next = NULL;
4957
4958 out:
4959 while (disk_list) {
4960 struct intel_disk *idisk = disk_list;
4961
4962 disk_list = disk_list->next;
4963 free(idisk);
4964 }
4965
4966 return champion;
4967}
4968
9587c373
LM		 4969static int
4970get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4971static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4972 int major, int minor, int keep_fd);
ec50f7b6
LM		 4973static int
4974get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4975 int *max, int keep_fd);
4976
cdddbdbc 4977static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4978 char *devname, struct md_list *devlist,
4979 int keep_fd)
cdddbdbc 4980{
a2b97981
DW		 4981 struct intel_super *super_list = NULL;
4982 struct intel_super *super = NULL;
a2b97981 4983 int err = 0;
9587c373 4984 int i = 0;
dab4a513 4985
9587c373
LM		 4986 if (fd >= 0)
4987 /* 'fd' is an opened container */
4988 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4989 else
ec50f7b6
LM		 4990 /* get super block from devlist devices */
4991 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4992 if (err)
1602d52c 4993 goto error;
a2b97981
DW		 4994 /* all mpbs enter, maybe one leaves */
4995 super = imsm_thunderdome(&super_list, i);
4996 if (!super) {
4997 err = 1;
4998 goto error;
cdddbdbc
DW		 4999 }
5000
47ee5a45
DW		 5001 if (find_missing(super) != 0) {
5002 free_imsm(super);
a2b97981
DW		 5003 err = 2;
5004 goto error;
47ee5a45 5005 }
8e59f3d8
AK		 5006
5007 /* load migration record */
5008 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5009 if (err == -1) {
5010 /* migration is in progress,
5011 * but migr_rec cannot be loaded,
5012 */
8e59f3d8
AK		 5013 err = 4;
5014 goto error;
5015 }
e2f41b2c
AK		 5016
5017 /* Check migration compatibility */
089f9d79 5018 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5019 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5020 if (devname)
5021 fprintf(stderr, " on %s\n", devname);
5022 else
5023 fprintf(stderr, " (IMSM).\n");
5024
5025 err = 5;
5026 goto error;
5027 }
5028
a2b97981
DW		 5029 err = 0;
5030
5031 error:
5032 while (super_list) {
5033 struct intel_super *s = super_list;
5034
5035 super_list = super_list->next;
5036 free_imsm(s);
5037 }
9587c373 5038
a2b97981
DW		 5039 if (err)
5040 return err;
f7e7067b 5041
cdddbdbc 5042 *sbp = super;
9587c373 5043 if (fd >= 0)
4dd2df09 5044 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5045 else
4dd2df09 5046 st->container_devnm[0] = 0;
a2b97981 5047 if (err == 0 && st->ss == NULL) {
bf5a934a 5048 st->ss = &super_imsm;
cdddbdbc
DW		 5049 st->minor_version = 0;
5050 st->max_devs = IMSM_MAX_DEVICES;
5051 }
cdddbdbc
DW		 5052 return 0;
5053}
2b959fbf 5054
ec50f7b6
LM		 5055static int
5056get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5057 int *max, int keep_fd)
5058{
5059 struct md_list *tmpdev;
5060 int err = 0;
5061 int i = 0;
9587c373 5062
ec50f7b6
LM		 5063 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5064 if (tmpdev->used != 1)
5065 continue;
5066 if (tmpdev->container == 1) {
ca9de185 5067 int lmax = 0;
ec50f7b6
LM		 5068 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5069 if (fd < 0) {
e7b84f9d 5070 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5071 tmpdev->devname, strerror(errno));
5072 err = 8;
5073 goto error;
5074 }
5075 err = get_sra_super_block(fd, super_list,
5076 tmpdev->devname, &lmax,
5077 keep_fd);
5078 i += lmax;
5079 close(fd);
5080 if (err) {
5081 err = 7;
5082 goto error;
5083 }
5084 } else {
5085 int major = major(tmpdev->st_rdev);
5086 int minor = minor(tmpdev->st_rdev);
5087 err = get_super_block(super_list,
4dd2df09 5088 NULL,
ec50f7b6
LM		 5089 tmpdev->devname,
5090 major, minor,
5091 keep_fd);
5092 i++;
5093 if (err) {
5094 err = 6;
5095 goto error;
5096 }
5097 }
5098 }
5099 error:
5100 *max = i;
5101 return err;
5102}
9587c373 5103
4dd2df09 5104static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5105 int major, int minor, int keep_fd)
5106{
594dc1b8 5107 struct intel_super *s;
9587c373
LM		 5108 char nm[32];
5109 int dfd = -1;
9587c373
LM		 5110 int err = 0;
5111 int retry;
5112
5113 s = alloc_super();
5114 if (!s) {
5115 err = 1;
5116 goto error;
5117 }
5118
5119 sprintf(nm, "%d:%d", major, minor);
5120 dfd = dev_open(nm, O_RDWR);
5121 if (dfd < 0) {
5122 err = 2;
5123 goto error;
5124 }
5125
fa7bb6f8 5126 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5127 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5128 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5129
5130 /* retry the load if we might have raced against mdmon */
4dd2df09 5131 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5132 for (retry = 0; retry < 3; retry++) {
5133 usleep(3000);
5134 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5135 if (err != 3)
5136 break;
5137 }
5138 error:
5139 if (!err) {
5140 s->next = *super_list;
5141 *super_list = s;
5142 } else {
5143 if (s)
8d67477f 5144 free_imsm(s);
36614e95 5145 if (dfd >= 0)
9587c373
LM		 5146 close(dfd);
5147 }
089f9d79 5148 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5149 close(dfd);
5150 return err;
5151
5152}
5153
5154static int
5155get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5156{
5157 struct mdinfo *sra;
4dd2df09 5158 char *devnm;
9587c373
LM		 5159 struct mdinfo *sd;
5160 int err = 0;
5161 int i = 0;
4dd2df09 5162 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5163 if (!sra)
5164 return 1;
5165
5166 if (sra->array.major_version != -1 ||
5167 sra->array.minor_version != -2 ||
5168 strcmp(sra->text_version, "imsm") != 0) {
5169 err = 1;
5170 goto error;
5171 }
5172 /* load all mpbs */
4dd2df09 5173 devnm = fd2devnm(fd);
9587c373 5174 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5175 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5176 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5177 err = 7;
5178 goto error;
5179 }
5180 }
5181 error:
5182 sysfs_free(sra);
5183 *max = i;
5184 return err;
5185}
5186
2b959fbf
N		 5187static int load_container_imsm(struct supertype *st, int fd, char *devname)
5188{
ec50f7b6 5189 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5190}
cdddbdbc
DW		 5191
5192static int load_super_imsm(struct supertype *st, int fd, char *devname)
5193{
5194 struct intel_super *super;
5195 int rv;
8a3544f8 5196 int retry;
cdddbdbc 5197
357ac106 5198 if (test_partition(fd))
691c6ee1
N		 5199 /* IMSM not allowed on partitions */
5200 return 1;
5201
37424f13
DW		 5202 free_super_imsm(st);
5203
49133e57 5204 super = alloc_super();
fa7bb6f8 5205 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5206 if (!super)
5207 return 1;
ea2bc72b
LM		 5208 /* Load hba and capabilities if they exist.
5209 * But do not preclude loading metadata in case capabilities or hba are
5210 * non-compliant and ignore_hw_compat is set.
5211 */
d424212e 5212 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5213 /* no orom/efi or non-intel hba of the disk */
089f9d79 5214 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5215 if (devname)
e7b84f9d 5216 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5217 free_imsm(super);
5218 return 2;
5219 }
a2b97981 5220 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5221
8a3544f8
AP		 5222 /* retry the load if we might have raced against mdmon */
5223 if (rv == 3) {
f96b1302
AP		 5224 struct mdstat_ent *mdstat = NULL;
5225 char *name = fd2kname(fd);
5226
5227 if (name)
5228 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5229
5230 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5231 for (retry = 0; retry < 3; retry++) {
5232 usleep(3000);
5233 rv = load_and_parse_mpb(fd, super, devname, 0);
5234 if (rv != 3)
5235 break;
5236 }
5237 }
5238
5239 free_mdstat(mdstat);
5240 }
5241
cdddbdbc
DW		 5242 if (rv) {
5243 if (devname)
7a862a02 5244 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5245 free_imsm(super);
5246 return rv;
5247 }
5248
5249 st->sb = super;
5250 if (st->ss == NULL) {
5251 st->ss = &super_imsm;
5252 st->minor_version = 0;
5253 st->max_devs = IMSM_MAX_DEVICES;
5254 }
8e59f3d8
AK		 5255
5256 /* load migration record */
2e062e82
AK		 5257 if (load_imsm_migr_rec(super, NULL) == 0) {
5258 /* Check for unsupported migration features */
5259 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5260 pr_err("Unsupported migration detected");
2e062e82
AK		 5261 if (devname)
5262 fprintf(stderr, " on %s\n", devname);
5263 else
5264 fprintf(stderr, " (IMSM).\n");
5265 return 3;
5266 }
e2f41b2c
AK		 5267 }
5268
cdddbdbc
DW		 5269 return 0;
5270}
5271
ef6ffade
DW		 5272static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5273{
5274 if (info->level == 1)
5275 return 128;
5276 return info->chunk_size >> 9;
5277}
5278
5551b113
CA		 5279static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5280 unsigned long long size)
fcfd9599 5281{
4025c288 5282 if (info->level == 1)
5551b113 5283 return size * 2;
4025c288 5284 else
5551b113 5285 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5286}
5287
4d1313e9
DW		 5288static void imsm_update_version_info(struct intel_super *super)
5289{
5290 /* update the version and attributes */
5291 struct imsm_super *mpb = super->anchor;
5292 char *version;
5293 struct imsm_dev *dev;
5294 struct imsm_map *map;
5295 int i;
5296
5297 for (i = 0; i < mpb->num_raid_devs; i++) {
5298 dev = get_imsm_dev(super, i);
238c0a71 5299 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5300 if (__le32_to_cpu(dev->size_high) > 0)
5301 mpb->attributes |= MPB_ATTRIB_2TB;
5302
5303 /* FIXME detect when an array spans a port multiplier */
5304 #if 0
5305 mpb->attributes |= MPB_ATTRIB_PM;
5306 #endif
5307
5308 if (mpb->num_raid_devs > 1 ||
5309 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5310 version = MPB_VERSION_ATTRIBS;
5311 switch (get_imsm_raid_level(map)) {
5312 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5313 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5314 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5315 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5316 }
5317 } else {
5318 if (map->num_members >= 5)
5319 version = MPB_VERSION_5OR6_DISK_ARRAY;
5320 else if (dev->status == DEV_CLONE_N_GO)
5321 version = MPB_VERSION_CNG;
5322 else if (get_imsm_raid_level(map) == 5)
5323 version = MPB_VERSION_RAID5;
5324 else if (map->num_members >= 3)
5325 version = MPB_VERSION_3OR4_DISK_ARRAY;
5326 else if (get_imsm_raid_level(map) == 1)
5327 version = MPB_VERSION_RAID1;
5328 else
5329 version = MPB_VERSION_RAID0;
5330 }
5331 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5332 }
5333}
5334
aa534678
DW		 5335static int check_name(struct intel_super *super, char *name, int quiet)
5336{
5337 struct imsm_super *mpb = super->anchor;
5338 char *reason = NULL;
9bd99a90
RS		 5339 char *start = name;
5340 size_t len = strlen(name);
aa534678
DW		 5341 int i;
5342
9bd99a90
RS		 5343 if (len > 0) {
5344 while (isspace(start[len - 1]))
5345 start[--len] = 0;
5346 while (*start && isspace(*start))
5347 ++start, --len;
5348 memmove(name, start, len + 1);
5349 }
5350
5351 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5352 reason = "must be 16 characters or less";
9bd99a90
RS		 5353 else if (len == 0)
5354 reason = "must be a non-empty string";
aa534678
DW		 5355
5356 for (i = 0; i < mpb->num_raid_devs; i++) {
5357 struct imsm_dev *dev = get_imsm_dev(super, i);
5358
5359 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5360 reason = "already exists";
5361 break;
5362 }
5363 }
5364
5365 if (reason && !quiet)
e7b84f9d 5366 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5367
5368 return !reason;
5369}
5370
8b353278 5371static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5372 struct shape *s, char *name,
83cd1e97
N		 5373 char *homehost, int *uuid,
5374 long long data_offset)
cdddbdbc 5375{
c2c087e6
DW		 5376 /* We are creating a volume inside a pre-existing container.
5377 * so st->sb is already set.
5378 */
5379 struct intel_super *super = st->sb;
f36a9ecd 5380 unsigned int sector_size = super->sector_size;
949c47a0 5381 struct imsm_super *mpb = super->anchor;
ba2de7ba 5382 struct intel_dev *dv;
c2c087e6
DW		 5383 struct imsm_dev *dev;
5384 struct imsm_vol *vol;
5385 struct imsm_map *map;
5386 int idx = mpb->num_raid_devs;
5387 int i;
760365f9 5388 int namelen;
c2c087e6 5389 unsigned long long array_blocks;
2c092cad 5390 size_t size_old, size_new;
5551b113 5391 unsigned long long num_data_stripes;
b53bfba6
TM		 5392 unsigned int data_disks;
5393 unsigned long long size_per_member;
cdddbdbc 5394
88c32bb1 5395 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5396 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5397 return 0;
5398 }
5399
2c092cad
DW		 5400 /* ensure the mpb is large enough for the new data */
5401 size_old = __le32_to_cpu(mpb->mpb_size);
5402 size_new = disks_to_mpb_size(info->nr_disks);
5403 if (size_new > size_old) {
5404 void *mpb_new;
f36a9ecd 5405 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5406
f36a9ecd 5407 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5408 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5409 return 0;
5410 }
85337573
AO		 5411 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5412 MIGR_REC_BUF_SECTORS*
5413 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5414 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5415 free(super->buf);
5416 free(super);
ea944c8f 5417 free(mpb_new);
8e59f3d8
AK		 5418 return 0;
5419 }
2c092cad
DW		 5420 memcpy(mpb_new, mpb, size_old);
5421 free(mpb);
5422 mpb = mpb_new;
949c47a0 5423 super->anchor = mpb_new;
2c092cad
DW		 5424 mpb->mpb_size = __cpu_to_le32(size_new);
5425 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5426 super->len = size_round;
2c092cad 5427 }
bf5a934a 5428 super->current_vol = idx;
3960e579
DW		 5429
5430 /* handle 'failed_disks' by either:
5431 * a) create dummy disk entries in the table if this the first
5432 * volume in the array. We add them here as this is the only
5433 * opportunity to add them. add_to_super_imsm_volume()
5434 * handles the non-failed disks and continues incrementing
5435 * mpb->num_disks.
5436 * b) validate that 'failed_disks' matches the current number
5437 * of missing disks if the container is populated
d23fe947 5438 */
3960e579 5439 if (super->current_vol == 0) {
d23fe947 5440 mpb->num_disks = 0;
3960e579
DW		 5441 for (i = 0; i < info->failed_disks; i++) {
5442 struct imsm_disk *disk;
5443
5444 mpb->num_disks++;
5445 disk = __get_imsm_disk(mpb, i);
5446 disk->status = CONFIGURED_DISK | FAILED_DISK;
5447 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5448 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5449 "missing:%d", (__u8)i);
3960e579
DW		 5450 }
5451 find_missing(super);
5452 } else {
5453 int missing = 0;
5454 struct dl *d;
5455
5456 for (d = super->missing; d; d = d->next)
5457 missing++;
5458 if (info->failed_disks > missing) {
e7b84f9d 5459 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5460 return 0;
5461 }
5462 }
5a038140 5463
aa534678
DW		 5464 if (!check_name(super, name, 0))
5465 return 0;
503975b9
N		 5466 dv = xmalloc(sizeof(*dv));
5467 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5468 /*
5469 * Explicitly allow truncating to not confuse gcc's
5470 * -Werror=stringop-truncation
5471 */
5472 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5473 memcpy(dev->volume, name, namelen);
e03640bd 5474 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5475 info->layout, info->chunk_size,
b53bfba6
TM		 5476 s->size * BLOCKS_PER_KB);
5477 data_disks = get_data_disks(info->level, info->layout,
5478 info->raid_disks);
5479 array_blocks = round_size_to_mb(array_blocks, data_disks);
5480 size_per_member = array_blocks / data_disks;
979d38be 5481
fcc2c9da 5482 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5483 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5484 vol = &dev->vol;
5485 vol->migr_state = 0;
1484e727 5486 set_migr_type(dev, MIGR_INIT);
3960e579 5487 vol->dirty = !info->state;
f8f603f1 5488 vol->curr_migr_unit = 0;
238c0a71 5489 map = get_imsm_map(dev, MAP_0);
5551b113 5490 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5491 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5492 map->failed_disk_num = ~0;
bf4442ab 5493 if (info->level > 0)
fffaf1ff
N		 5494 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5495 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5496 else
5497 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5498 IMSM_T_STATE_NORMAL;
252d23c0 5499 map->ddf = 1;
ef6ffade
DW		 5500
5501 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5502 free(dev);
5503 free(dv);
7a862a02 5504 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5505 return 0;
5506 }
81062a36
DW		 5507
5508 map->raid_level = info->level;
4d1313e9 5509 if (info->level == 10) {
c2c087e6 5510 map->raid_level = 1;
4d1313e9 5511 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5512 } else if (info->level == 1)
5513 map->num_domains = info->raid_disks;
5514 else
ff596308 5515 map->num_domains = 1;
81062a36 5516
5551b113 5517 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5518 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5519 num_data_stripes /= map->num_domains;
5520 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5521
44490938
MD		 5522 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5523 set_blocks_per_member(map, info_to_blocks_per_member(info,
5524 size_per_member /
5525 BLOCKS_PER_KB));
5526
c2c087e6
DW		 5527 map->num_members = info->raid_disks;
5528 for (i = 0; i < map->num_members; i++) {
5529 /* initialized in add_to_super */
4eb26970 5530 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5531 }
949c47a0 5532 mpb->num_raid_devs++;
2a24dc1b
PB		 5533 mpb->num_raid_devs_created++;
5534 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5535
b7580566 5536 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5537 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5538 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5539 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5540 } else {
5541 free(dev);
5542 free(dv);
5543 pr_err("imsm does not support consistency policy %s\n",
5544 map_num(consistency_policies, s->consistency_policy));
5545 return 0;
5546 }
5547
ba2de7ba
DW		 5548 dv->dev = dev;
5549 dv->index = super->current_vol;
5550 dv->next = super->devlist;
5551 super->devlist = dv;
c2c087e6 5552
4d1313e9
DW		 5553 imsm_update_version_info(super);
5554
c2c087e6 5555 return 1;
cdddbdbc
DW		 5556}
5557
bf5a934a 5558static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5559 struct shape *s, char *name,
83cd1e97
N		 5560 char *homehost, int *uuid,
5561 unsigned long long data_offset)
bf5a934a
DW		 5562{
5563 /* This is primarily called by Create when creating a new array.
5564 * We will then get add_to_super called for each component, and then
5565 * write_init_super called to write it out to each device.
5566 * For IMSM, Create can create on fresh devices or on a pre-existing
5567 * array.
5568 * To create on a pre-existing array a different method will be called.
5569 * This one is just for fresh drives.
5570 */
5571 struct intel_super *super;
5572 struct imsm_super *mpb;
5573 size_t mpb_size;
4d1313e9 5574 char *version;
bf5a934a 5575
83cd1e97 5576 if (data_offset != INVALID_SECTORS) {
ed503f89 5577 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5578 return 0;
5579 }
5580
bf5a934a 5581 if (st->sb)
5308f117 5582 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5583 data_offset);
e683ca88
DW		 5584
5585 if (info)
5586 mpb_size = disks_to_mpb_size(info->nr_disks);
5587 else
f36a9ecd 5588 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5589
49133e57 5590 super = alloc_super();
f36a9ecd
PB		 5591 if (super &&
5592 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5593 free_imsm(super);
e683ca88
DW		 5594 super = NULL;
5595 }
5596 if (!super) {
1ade5cc1 5597 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5598 return 0;
5599 }
de44e46f
PB		 5600 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5601 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5602 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5603 free(super->buf);
8d67477f 5604 free_imsm(super);
8e59f3d8
AK		 5605 return 0;
5606 }
e683ca88 5607 memset(super->buf, 0, mpb_size);
ef649044 5608 mpb = super->buf;
e683ca88
DW		 5609 mpb->mpb_size = __cpu_to_le32(mpb_size);
5610 st->sb = super;
5611
5612 if (info == NULL) {
5613 /* zeroing superblock */
5614 return 0;
5615 }
bf5a934a 5616
4d1313e9
DW		 5617 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5618
5619 version = (char *) mpb->sig;
5620 strcpy(version, MPB_SIGNATURE);
5621 version += strlen(MPB_SIGNATURE);
5622 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5623
bf5a934a
DW		 5624 return 1;
5625}
5626
f2cc4f7d
AO		 5627static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5628{
5629 unsigned int member_sector_size;
5630
5631 if (dl->fd < 0) {
5632 pr_err("Invalid file descriptor for %s\n", dl->devname);
5633 return 0;
5634 }
5635
5636 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5637 return 0;
5638 if (member_sector_size != super->sector_size)
5639 return 0;
5640 return 1;
5641}
5642
f20c3968 5643static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5644 int fd, char *devname)
5645{
5646 struct intel_super *super = st->sb;
d23fe947 5647 struct imsm_super *mpb = super->anchor;
3960e579 5648 struct imsm_disk *_disk;
bf5a934a
DW		 5649 struct imsm_dev *dev;
5650 struct imsm_map *map;
3960e579 5651 struct dl *dl, *df;
4eb26970 5652 int slot;
bf5a934a 5653
949c47a0 5654 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5655 map = get_imsm_map(dev, MAP_0);
bf5a934a 5656
208933a7 5657 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5658 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5659 devname);
5660 return 1;
5661 }
5662
efb30e7f
DW		 5663 if (fd == -1) {
5664 /* we're doing autolayout so grab the pre-marked (in
5665 * validate_geometry) raid_disk
5666 */
5667 for (dl = super->disks; dl; dl = dl->next)
5668 if (dl->raiddisk == dk->raid_disk)
5669 break;
5670 } else {
5671 for (dl = super->disks; dl ; dl = dl->next)
5672 if (dl->major == dk->major &&
5673 dl->minor == dk->minor)
5674 break;
5675 }
d23fe947 5676
208933a7 5677 if (!dl) {
e7b84f9d 5678 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5679 return 1;
208933a7 5680 }
bf5a934a 5681
59632db9
MZ		 5682 if (mpb->num_disks == 0)
5683 if (!get_dev_sector_size(dl->fd, dl->devname,
5684 &super->sector_size))
5685 return 1;
5686
f2cc4f7d
AO		 5687 if (!drive_validate_sector_size(super, dl)) {
5688 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5689 return 1;
5690 }
5691
d23fe947
DW		 5692 /* add a pristine spare to the metadata */
5693 if (dl->index < 0) {
5694 dl->index = super->anchor->num_disks;
5695 super->anchor->num_disks++;
5696 }
4eb26970
DW		 5697 /* Check the device has not already been added */
5698 slot = get_imsm_disk_slot(map, dl->index);
5699 if (slot >= 0 &&
238c0a71 5700 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5701 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5702 devname);
5703 return 1;
5704 }
656b6b5a 5705 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5706 dl->disk.status = CONFIGURED_DISK;
d23fe947 5707
3960e579
DW		 5708 /* update size of 'missing' disks to be at least as large as the
5709 * largest acitve member (we only have dummy missing disks when
5710 * creating the first volume)
5711 */
5712 if (super->current_vol == 0) {
5713 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5714 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5715 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5716 _disk = __get_imsm_disk(mpb, df->index);
5717 *_disk = df->disk;
5718 }
5719 }
5720
5721 /* refresh unset/failed slots to point to valid 'missing' entries */
5722 for (df = super->missing; df; df = df->next)
5723 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5724 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5725
5726 if ((ord & IMSM_ORD_REBUILD) == 0)
5727 continue;
5728 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5729 if (is_gen_migration(dev)) {
238c0a71
AK		 5730 struct imsm_map *map2 = get_imsm_map(dev,
5731 MAP_1);
0a108d63 5732 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5733 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5734 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5735 slot2,
5736 MAP_1);
1ace8403
AK		 5737 if ((unsigned)df->index ==
5738 ord_to_idx(ord2))
5739 set_imsm_ord_tbl_ent(map2,
0a108d63 5740 slot2,
1ace8403
AK		 5741 df->index |
5742 IMSM_ORD_REBUILD);
5743 }
5744 }
3960e579
DW		 5745 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5746 break;
5747 }
5748
d23fe947
DW		 5749 /* if we are creating the first raid device update the family number */
5750 if (super->current_vol == 0) {
5751 __u32 sum;
5752 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5753
3960e579 5754 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5755 if (!_dev || !_disk) {
e7b84f9d 5756 pr_err("BUG mpb setup error\n");
791b666a
AW		 5757 return 1;
5758 }
d23fe947
DW		 5759 *_dev = *dev;
5760 *_disk = dl->disk;
148acb7b
DW		 5761 sum = random32();
5762 sum += __gen_imsm_checksum(mpb);
d23fe947 5763 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5764 mpb->orig_family_num = mpb->family_num;
d23fe947 5765 }
ca0748fa 5766 super->current_disk = dl;
f20c3968 5767 return 0;
bf5a934a
DW		 5768}
5769
a8619d23
AK		 5770/* mark_spare()
5771 * Function marks disk as spare and restores disk serial
5772 * in case it was previously marked as failed by takeover operation
5773 * reruns:
5774 * -1 : critical error
5775 * 0 : disk is marked as spare but serial is not set
5776 * 1 : success
5777 */
5778int mark_spare(struct dl *disk)
5779{
5780 __u8 serial[MAX_RAID_SERIAL_LEN];
5781 int ret_val = -1;
5782
5783 if (!disk)
5784 return ret_val;
5785
5786 ret_val = 0;
6da53c0e 5787 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5788 /* Restore disk serial number, because takeover marks disk
5789 * as failed and adds to serial ':0' before it becomes
5790 * a spare disk.
5791 */
5792 serialcpy(disk->serial, serial);
5793 serialcpy(disk->disk.serial, serial);
5794 ret_val = 1;
5795 }
5796 disk->disk.status = SPARE_DISK;
5797 disk->index = -1;
5798
5799 return ret_val;
5800}
88654014 5801
f20c3968 5802static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5803 int fd, char *devname,
5804 unsigned long long data_offset)
cdddbdbc 5805{
c2c087e6 5806 struct intel_super *super = st->sb;
c2c087e6
DW		 5807 struct dl *dd;
5808 unsigned long long size;
fa7bb6f8 5809 unsigned int member_sector_size;
f2f27e63 5810 __u32 id;
c2c087e6
DW		 5811 int rv;
5812 struct stat stb;
5813
88654014
LM		 5814 /* If we are on an RAID enabled platform check that the disk is
5815 * attached to the raid controller.
5816 * We do not need to test disks attachment for container based additions,
5817 * they shall be already tested when container was created/assembled.
88c32bb1 5818 */
d424212e 5819 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5820 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5821 if (rv != 0) {
5822 dprintf("capability: %p fd: %d ret: %d\n",
5823 super->orom, fd, rv);
5824 return 1;
88c32bb1
DW		 5825 }
5826
f20c3968
DW		 5827 if (super->current_vol >= 0)
5828 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5829
c2c087e6 5830 fstat(fd, &stb);
503975b9 5831 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5832 dd->major = major(stb.st_rdev);
5833 dd->minor = minor(stb.st_rdev);
503975b9 5834 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5835 dd->fd = fd;
689c9bf3 5836 dd->e = NULL;
1a64be56 5837 dd->action = DISK_ADD;
6da53c0e 5838 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5839 if (rv) {
e7b84f9d 5840 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5841 if (dd->devname)
5842 free(dd->devname);
949c47a0 5843 free(dd);
0030e8d6 5844 abort();
c2c087e6 5845 }
20bee0f8
PB		 5846 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5847 (super->hba->type == SYS_DEV_VMD))) {
5848 int i;
5849 char *devpath = diskfd_to_devpath(fd);
5850 char controller_path[PATH_MAX];
5851
5852 if (!devpath) {
5853 pr_err("failed to get devpath, aborting\n");
5854 if (dd->devname)
5855 free(dd->devname);
5856 free(dd);
5857 return 1;
5858 }
5859
5860 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5861 free(devpath);
5862
5863 if (devpath_to_vendor(controller_path) == 0x8086) {
5864 /*
5865 * If Intel's NVMe drive has serial ended with
5866 * "-A","-B","-1" or "-2" it means that this is "x8"
5867 * device (double drive on single PCIe card).
5868 * User should be warned about potential data loss.
5869 */
5870 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5871 /* Skip empty character at the end */
5872 if (dd->serial[i] == 0)
5873 continue;
5874
5875 if (((dd->serial[i] == 'A') ||
5876 (dd->serial[i] == 'B') ||
5877 (dd->serial[i] == '1') ||
5878 (dd->serial[i] == '2')) &&
5879 (dd->serial[i-1] == '-'))
5880 pr_err("\tThe action you are about to take may put your data at risk.\n"
5881 "\tPlease note that x8 devices may consist of two separate x4 devices "
5882 "located on a single PCIe port.\n"
5883 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5884 break;
5885 }
32716c51
PB		 5886 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5887 !imsm_orom_has_tpv_support(super->orom)) {
5888 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5889 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5890 free(dd->devname);
5891 free(dd);
5892 return 1;
20bee0f8
PB		 5893 }
5894 }
c2c087e6 5895
c2c087e6 5896 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5897 get_dev_sector_size(fd, NULL, &member_sector_size);
5898
5899 if (super->sector_size == 0) {
5900 /* this a first device, so sector_size is not set yet */
5901 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5902 }
5903
71e5411e 5904 /* clear migr_rec when adding disk to container */
85337573
AO		 5905 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5906 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5907 SEEK_SET) >= 0) {
466070ad 5908 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5909 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5910 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5911 perror("Write migr_rec failed");
5912 }
5913
c2c087e6 5914 size /= 512;
1f24f035 5915 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5916 set_total_blocks(&dd->disk, size);
5917 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5918 struct imsm_super *mpb = super->anchor;
5919 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5920 }
a8619d23 5921 mark_spare(dd);
c2c087e6 5922 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5923 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5924 else
b9f594fe 5925 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5926
5927 if (st->update_tail) {
1a64be56
LM		 5928 dd->next = super->disk_mgmt_list;
5929 super->disk_mgmt_list = dd;
43dad3d6
DW		 5930 } else {
5931 dd->next = super->disks;
5932 super->disks = dd;
ceaf0ee1 5933 super->updates_pending++;
43dad3d6 5934 }
f20c3968
DW		 5935
5936 return 0;
cdddbdbc
DW		 5937}
5938
1a64be56
LM		 5939static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5940{
5941 struct intel_super *super = st->sb;
5942 struct dl *dd;
5943
5944 /* remove from super works only in mdmon - for communication
5945 * manager - monitor. Check if communication memory buffer
5946 * is prepared.
5947 */
5948 if (!st->update_tail) {
1ade5cc1 5949 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5950 return 1;
5951 }
503975b9 5952 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5953 dd->major = dk->major;
5954 dd->minor = dk->minor;
1a64be56 5955 dd->fd = -1;
a8619d23 5956 mark_spare(dd);
1a64be56
LM		 5957 dd->action = DISK_REMOVE;
5958
5959 dd->next = super->disk_mgmt_list;
5960 super->disk_mgmt_list = dd;
5961
1a64be56
LM		 5962 return 0;
5963}
5964
f796af5d
DW		 5965static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5966
5967static union {
f36a9ecd 5968 char buf[MAX_SECTOR_SIZE];
f796af5d 5969 struct imsm_super anchor;
f36a9ecd 5970} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5971
d23fe947
DW		 5972/* spare records have their own family number and do not have any defined raid
5973 * devices
5974 */
5975static int write_super_imsm_spares(struct intel_super *super, int doclose)
5976{
d23fe947 5977 struct imsm_super *mpb = super->anchor;
f796af5d 5978 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5979 __u32 sum;
5980 struct dl *d;
5981
68641cdb
JS		 5982 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5983 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5984 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5985 spare->num_disks = 1;
5986 spare->num_raid_devs = 0;
5987 spare->cache_size = mpb->cache_size;
5988 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5989
5990 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5991 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5992
5993 for (d = super->disks; d; d = d->next) {
8796fdc4 5994 if (d->index != -1)
d23fe947
DW		 5995 continue;
5996
f796af5d 5997 spare->disk[0] = d->disk;
027c374f
CA		 5998 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5999 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6000
f36a9ecd
PB		 6001 if (super->sector_size == 4096)
6002 convert_to_4k_imsm_disk(&spare->disk[0]);
6003
f796af5d
DW		 6004 sum = __gen_imsm_checksum(spare);
6005 spare->family_num = __cpu_to_le32(sum);
6006 spare->orig_family_num = 0;
6007 sum = __gen_imsm_checksum(spare);
6008 spare->check_sum = __cpu_to_le32(sum);
d23fe947 6009
f796af5d 6010 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 6011 pr_err("failed for device %d:%d %s\n",
6012 d->major, d->minor, strerror(errno));
e74255d9 6013 return 1;
d23fe947
DW		 6014 }
6015 if (doclose) {
6016 close(d->fd);
6017 d->fd = -1;
6018 }
6019 }
6020
e74255d9 6021 return 0;
d23fe947
DW		 6022}
6023
36988a3d 6024static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6025{
36988a3d 6026 struct intel_super *super = st->sb;
f36a9ecd 6027 unsigned int sector_size = super->sector_size;
949c47a0 6028 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6029 struct dl *d;
6030 __u32 generation;
6031 __u32 sum;
d23fe947 6032 int spares = 0;
949c47a0 6033 int i;
a48ac0a8 6034 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6035 int num_disks = 0;
146c6260 6036 int clear_migration_record = 1;
bbab0940 6037 __u32 bbm_log_size;
cdddbdbc 6038
c2c087e6
DW		 6039 /* 'generation' is incremented everytime the metadata is written */
6040 generation = __le32_to_cpu(mpb->generation_num);
6041 generation++;
6042 mpb->generation_num = __cpu_to_le32(generation);
6043
148acb7b
DW		 6044 /* fix up cases where previous mdadm releases failed to set
6045 * orig_family_num
6046 */
6047 if (mpb->orig_family_num == 0)
6048 mpb->orig_family_num = mpb->family_num;
6049
d23fe947 6050 for (d = super->disks; d; d = d->next) {
8796fdc4 6051 if (d->index == -1)
d23fe947 6052 spares++;
36988a3d 6053 else {
d23fe947 6054 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6055 num_disks++;
6056 }
d23fe947 6057 }
36988a3d 6058 for (d = super->missing; d; d = d->next) {
47ee5a45 6059 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6060 num_disks++;
6061 }
6062 mpb->num_disks = num_disks;
6063 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6064
949c47a0
DW		 6065 for (i = 0; i < mpb->num_raid_devs; i++) {
6066 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6067 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6068 if (dev && dev2) {
6069 imsm_copy_dev(dev, dev2);
6070 mpb_size += sizeof_imsm_dev(dev, 0);
6071 }
146c6260
AK		 6072 if (is_gen_migration(dev2))
6073 clear_migration_record = 0;
949c47a0 6074 }
bbab0940
TM		 6075
6076 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6077
6078 if (bbm_log_size) {
6079 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6080 mpb->attributes |= MPB_ATTRIB_BBM;
6081 } else
6082 mpb->attributes &= ~MPB_ATTRIB_BBM;
6083
6084 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6085 mpb_size += bbm_log_size;
a48ac0a8 6086 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6087
bbab0940
TM		 6088#ifdef DEBUG
6089 assert(super->len == 0 || mpb_size <= super->len);
6090#endif
6091
c2c087e6 6092 /* recalculate checksum */
949c47a0 6093 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6094 mpb->check_sum = __cpu_to_le32(sum);
6095
51d83f5d
AK		 6096 if (super->clean_migration_record_by_mdmon) {
6097 clear_migration_record = 1;
6098 super->clean_migration_record_by_mdmon = 0;
6099 }
146c6260 6100 if (clear_migration_record)
de44e46f 6101 memset(super->migr_rec_buf, 0,
85337573 6102 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6103
f36a9ecd
PB		 6104 if (sector_size == 4096)
6105 convert_to_4k(super);
6106
d23fe947 6107 /* write the mpb for disks that compose raid devices */
c2c087e6 6108 for (d = super->disks; d ; d = d->next) {
86c54047 6109 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6110 continue;
30602f53 6111
146c6260
AK		 6112 if (clear_migration_record) {
6113 unsigned long long dsize;
6114
6115 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6116 if (lseek64(d->fd, dsize - sector_size,
6117 SEEK_SET) >= 0) {
466070ad
PB		 6118 if ((unsigned int)write(d->fd,
6119 super->migr_rec_buf,
de44e46f
PB		 6120 MIGR_REC_BUF_SECTORS*sector_size) !=
6121 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6122 perror("Write migr_rec failed");
146c6260
AK		 6123 }
6124 }
51d83f5d
AK		 6125
6126 if (store_imsm_mpb(d->fd, mpb))
6127 fprintf(stderr,
1ade5cc1
N		 6128 "failed for device %d:%d (fd: %d)%s\n",
6129 d->major, d->minor,
51d83f5d
AK		 6130 d->fd, strerror(errno));
6131
c2c087e6
DW		 6132 if (doclose) {
6133 close(d->fd);
6134 d->fd = -1;
6135 }
6136 }
6137
d23fe947
DW		 6138 if (spares)
6139 return write_super_imsm_spares(super, doclose);
6140
e74255d9 6141 return 0;
c2c087e6
DW		 6142}
6143
9b1fb677 6144static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6145{
6146 size_t len;
6147 struct imsm_update_create_array *u;
6148 struct intel_super *super = st->sb;
9b1fb677 6149 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6150 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6151 struct disk_info *inf;
6152 struct imsm_disk *disk;
6153 int i;
43dad3d6 6154
54c2c1ea
DW		 6155 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6156 sizeof(*inf) * map->num_members;
503975b9 6157 u = xmalloc(len);
43dad3d6 6158 u->type = update_create_array;
9b1fb677 6159 u->dev_idx = dev_idx;
43dad3d6 6160 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6161 inf = get_disk_info(u);
6162 for (i = 0; i < map->num_members; i++) {
238c0a71 6163 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6164
54c2c1ea 6165 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6166 if (!disk)
6167 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6168 serialcpy(inf[i].serial, disk->serial);
6169 }
43dad3d6
DW		 6170 append_metadata_update(st, u, len);
6171
6172 return 0;
6173}
6174
1a64be56 6175static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6176{
6177 struct intel_super *super = st->sb;
6178 size_t len;
1a64be56 6179 struct imsm_update_add_remove_disk *u;
43dad3d6 6180
1a64be56 6181 if (!super->disk_mgmt_list)
43dad3d6
DW		 6182 return 0;
6183
6184 len = sizeof(*u);
503975b9 6185 u = xmalloc(len);
1a64be56 6186 u->type = update_add_remove_disk;
43dad3d6
DW		 6187 append_metadata_update(st, u, len);
6188
6189 return 0;
6190}
2432ce9b
AP		 6191
6192__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6193
e397cefe
AP		 6194static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6195{
6196 struct ppl_header *ppl_hdr = buf;
6197 int ret;
6198
6199 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6200
6201 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6202 ret = -errno;
6203 perror("Failed to seek to PPL header location");
6204 return ret;
6205 }
6206
6207 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6208 ret = -errno;
6209 perror("Write PPL header failed");
6210 return ret;
6211 }
6212
6213 fsync(fd);
6214
6215 return 0;
6216}
6217
2432ce9b
AP		 6218static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6219{
6220 struct intel_super *super = st->sb;
6221 void *buf;
6222 struct ppl_header *ppl_hdr;
6223 int ret;
6224
b2514242
PB		 6225 /* first clear entire ppl space */
6226 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6227 if (ret)
6228 return ret;
6229
6230 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6231 if (ret) {
6232 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6233 return -ret;
2432ce9b
AP		 6234 }
6235
6236 memset(buf, 0, PPL_HEADER_SIZE);
6237 ppl_hdr = buf;
6238 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6239 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6240
6241 if (info->mismatch_cnt) {
6242 /*
6243 * We are overwriting an invalid ppl. Make one entry with wrong
6244 * checksum to prevent the kernel from skipping resync.
6245 */
6246 ppl_hdr->entries_count = __cpu_to_le32(1);
6247 ppl_hdr->entries[0].checksum = ~0;
6248 }
6249
e397cefe 6250 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6251
6252 free(buf);
6253 return ret;
6254}
6255
e397cefe
AP		 6256static int is_rebuilding(struct imsm_dev *dev);
6257
2432ce9b
AP		 6258static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6259 struct mdinfo *disk)
6260{
6261 struct intel_super *super = st->sb;
6262 struct dl *d;
e397cefe 6263 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6264 int ret = 0;
e397cefe 6265 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6266 __u32 crc;
6267 struct imsm_dev *dev;
2432ce9b 6268 __u32 idx;
44b6b876
PB		 6269 unsigned int i;
6270 unsigned long long ppl_offset = 0;
6271 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6272
6273 if (disk->disk.raid_disk < 0)
6274 return 0;
6275
2432ce9b 6276 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6277 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6278 d = get_imsm_dl_disk(super, idx);
6279
6280 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6281 return 0;
6282
6283 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6284 pr_err("Failed to allocate PPL header buffer\n");
6285 return -1;
6286 }
6287 buf = buf_orig;
2432ce9b 6288
44b6b876
PB		 6289 ret = 1;
6290 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6291 void *tmp;
6292
44b6b876 6293 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6294
44b6b876
PB		 6295 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6296 SEEK_SET) < 0) {
6297 perror("Failed to seek to PPL header location");
6298 ret = -1;
e397cefe 6299 break;
44b6b876 6300 }
2432ce9b 6301
44b6b876
PB		 6302 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6303 perror("Read PPL header failed");
6304 ret = -1;
e397cefe 6305 break;
44b6b876 6306 }
2432ce9b 6307
44b6b876 6308 ppl_hdr = buf;
2432ce9b 6309
44b6b876
PB		 6310 crc = __le32_to_cpu(ppl_hdr->checksum);
6311 ppl_hdr->checksum = 0;
6312
6313 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6314 dprintf("Wrong PPL header checksum on %s\n",
6315 d->devname);
e397cefe 6316 break;
44b6b876
PB		 6317 }
6318
6319 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6320 /* previous was newest, it was already checked */
e397cefe 6321 break;
44b6b876
PB		 6322 }
6323
6324 if ((__le32_to_cpu(ppl_hdr->signature) !=
6325 super->anchor->orig_family_num)) {
6326 dprintf("Wrong PPL header signature on %s\n",
6327 d->devname);
6328 ret = 1;
e397cefe 6329 break;
44b6b876
PB		 6330 }
6331
6332 ret = 0;
6333 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6334
44b6b876
PB		 6335 ppl_offset += PPL_HEADER_SIZE;
6336 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6337 ppl_offset +=
6338 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6339
6340 if (!buf_prev)
6341 buf_prev = buf + PPL_HEADER_SIZE;
6342 tmp = buf_prev;
6343 buf_prev = buf;
6344 buf = tmp;
2432ce9b
AP		 6345 }
6346
e397cefe
AP		 6347 if (buf_prev) {
6348 buf = buf_prev;
6349 ppl_hdr = buf_prev;
6350 }
2432ce9b 6351
54148aba
PB		 6352 /*
6353 * Update metadata to use mutliple PPLs area (1MB).
6354 * This is done once for all RAID members
6355 */
6356 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6357 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6358 char subarray[20];
6359 struct mdinfo *member_dev;
6360
6361 sprintf(subarray, "%d", info->container_member);
6362
6363 if (mdmon_running(st->container_devnm))
6364 st->update_tail = &st->updates;
6365
6366 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6367 pr_err("Failed to update subarray %s\n",
6368 subarray);
6369 } else {
6370 if (st->update_tail)
6371 flush_metadata_updates(st);
6372 else
6373 st->ss->sync_metadata(st);
6374 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6375 for (member_dev = info->devs; member_dev;
6376 member_dev = member_dev->next)
6377 member_dev->ppl_size =
6378 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6379 }
6380 }
6381
b23d0750 6382 if (ret == 1) {
2fc0fc63
AP		 6383 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6384
50b9c10d
PB		 6385 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6386 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6387 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6388 (is_rebuilding(dev) &&
2ec9d182
AP		 6389 dev->vol.curr_migr_unit == 0 &&
6390 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6391 ret = st->ss->write_init_ppl(st, info, d->fd);
6392 else
6393 info->mismatch_cnt++;
e397cefe
AP		 6394 } else if (ret == 0 &&
6395 ppl_hdr->entries_count == 0 &&
6396 is_rebuilding(dev) &&
6397 info->resync_start == 0) {
6398 /*
6399 * The header has no entries - add a single empty entry and
6400 * rewrite the header to prevent the kernel from going into
6401 * resync after an interrupted rebuild.
6402 */
6403 ppl_hdr->entries_count = __cpu_to_le32(1);
6404 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6405 }
2432ce9b 6406
e397cefe
AP		 6407 free(buf_orig);
6408
2432ce9b
AP		 6409 return ret;
6410}
6411
2432ce9b
AP		 6412static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6413{
6414 struct intel_super *super = st->sb;
6415 struct dl *d;
6416 int ret = 0;
6417
6418 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6419 info->array.level != 5)
6420 return 0;
6421
6422 for (d = super->disks; d ; d = d->next) {
6423 if (d->index < 0 || is_failed(&d->disk))
6424 continue;
6425
6426 ret = st->ss->write_init_ppl(st, info, d->fd);
6427 if (ret)
6428 break;
6429 }
6430
6431 return ret;
6432}
43dad3d6 6433
c2c087e6
DW		 6434static int write_init_super_imsm(struct supertype *st)
6435{
9b1fb677
DW		 6436 struct intel_super *super = st->sb;
6437 int current_vol = super->current_vol;
2432ce9b
AP		 6438 int rv = 0;
6439 struct mdinfo info;
6440
6441 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6442
6443 /* we are done with current_vol reset it to point st at the container */
6444 super->current_vol = -1;
6445
8273f55e 6446 if (st->update_tail) {
43dad3d6
DW		 6447 /* queue the recently created array / added disk
6448 * as a metadata update */
8273f55e 6449
43dad3d6 6450 /* determine if we are creating a volume or adding a disk */
9b1fb677 6451 if (current_vol < 0) {
1a64be56
LM		 6452 /* in the mgmt (add/remove) disk case we are running
6453 * in mdmon context, so don't close fd's
43dad3d6 6454 */
2432ce9b
AP		 6455 rv = mgmt_disk(st);
6456 } else {
6457 rv = write_init_ppl_imsm_all(st, &info);
6458 if (!rv)
6459 rv = create_array(st, current_vol);
6460 }
d682f344
N		 6461 } else {
6462 struct dl *d;
6463 for (d = super->disks; d; d = d->next)
ba728be7 6464 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6465 if (current_vol >= 0)
6466 rv = write_init_ppl_imsm_all(st, &info);
6467 if (!rv)
6468 rv = write_super_imsm(st, 1);
d682f344 6469 }
2432ce9b
AP		 6470
6471 return rv;
cdddbdbc
DW		 6472}
6473
e683ca88 6474static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6475{
e683ca88
DW		 6476 struct intel_super *super = st->sb;
6477 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6478
e683ca88 6479 if (!mpb)
ad97895e
DW		 6480 return 1;
6481
f36a9ecd
PB		 6482 if (super->sector_size == 4096)
6483 convert_to_4k(super);
e683ca88 6484 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6485}
6486
cdddbdbc
DW		 6487static int validate_geometry_imsm_container(struct supertype *st, int level,
6488 int layout, int raiddisks, int chunk,
af4348dd
N		 6489 unsigned long long size,
6490 unsigned long long data_offset,
6491 char *dev,
2c514b71
NB		 6492 unsigned long long *freesize,
6493 int verbose)
cdddbdbc 6494{
c2c087e6
DW		 6495 int fd;
6496 unsigned long long ldsize;
594dc1b8 6497 struct intel_super *super;
f2f5c343 6498 int rv = 0;
cdddbdbc 6499
c2c087e6
DW		 6500 if (level != LEVEL_CONTAINER)
6501 return 0;
6502 if (!dev)
6503 return 1;
6504
6505 fd = open(dev, O_RDONLY|O_EXCL, 0);
6506 if (fd < 0) {
ba728be7 6507 if (verbose > 0)
e7b84f9d 6508 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6509 dev, strerror(errno));
c2c087e6
DW		 6510 return 0;
6511 }
6512 if (!get_dev_size(fd, dev, &ldsize)) {
6513 close(fd);
6514 return 0;
6515 }
f2f5c343
LM		 6516
6517 /* capabilities retrieve could be possible
6518 * note that there is no fd for the disks in array.
6519 */
6520 super = alloc_super();
8d67477f
TM		 6521 if (!super) {
6522 close(fd);
6523 return 0;
6524 }
fa7bb6f8
PB		 6525 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6526 close(fd);
6527 free_imsm(super);
6528 return 0;
6529 }
6530
ba728be7 6531 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6532 if (rv != 0) {
6533#if DEBUG
6534 char str[256];
6535 fd2devname(fd, str);
1ade5cc1 6536 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6537 fd, str, super->orom, rv, raiddisks);
6538#endif
6539 /* no orom/efi or non-intel hba of the disk */
6540 close(fd);
6541 free_imsm(super);
6542 return 0;
6543 }
c2c087e6 6544 close(fd);
9126b9a8
CA		 6545 if (super->orom) {
6546 if (raiddisks > super->orom->tds) {
6547 if (verbose)
7a862a02 6548 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6549 raiddisks, super->orom->tds);
6550 free_imsm(super);
6551 return 0;
6552 }
6553 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6554 (ldsize >> 9) >> 32 > 0) {
6555 if (verbose)
e7b84f9d 6556 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6557 free_imsm(super);
6558 return 0;
6559 }
f2f5c343 6560 }
c2c087e6 6561
af4348dd 6562 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6563 free_imsm(super);
c2c087e6
DW		 6564
6565 return 1;
cdddbdbc
DW		 6566}
6567
0dcecb2e
DW		 6568static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6569{
6570 const unsigned long long base_start = e[*idx].start;
6571 unsigned long long end = base_start + e[*idx].size;
6572 int i;
6573
6574 if (base_start == end)
6575 return 0;
6576
6577 *idx = *idx + 1;
6578 for (i = *idx; i < num_extents; i++) {
6579 /* extend overlapping extents */
6580 if (e[i].start >= base_start &&
6581 e[i].start <= end) {
6582 if (e[i].size == 0)
6583 return 0;
6584 if (e[i].start + e[i].size > end)
6585 end = e[i].start + e[i].size;
6586 } else if (e[i].start > end) {
6587 *idx = i;
6588 break;
6589 }
6590 }
6591
6592 return end - base_start;
6593}
6594
6595static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6596{
6597 /* build a composite disk with all known extents and generate a new
6598 * 'maxsize' given the "all disks in an array must share a common start
6599 * offset" constraint
6600 */
503975b9 6601 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6602 struct dl *dl;
6603 int i, j;
6604 int start_extent;
6605 unsigned long long pos;
b9d77223 6606 unsigned long long start = 0;
0dcecb2e
DW		 6607 unsigned long long maxsize;
6608 unsigned long reserve;
6609
0dcecb2e
DW		 6610 /* coalesce and sort all extents. also, check to see if we need to
6611 * reserve space between member arrays
6612 */
6613 j = 0;
6614 for (dl = super->disks; dl; dl = dl->next) {
6615 if (!dl->e)
6616 continue;
6617 for (i = 0; i < dl->extent_cnt; i++)
6618 e[j++] = dl->e[i];
6619 }
6620 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6621
6622 /* merge extents */
6623 i = 0;
6624 j = 0;
6625 while (i < sum_extents) {
6626 e[j].start = e[i].start;
6627 e[j].size = find_size(e, &i, sum_extents);
6628 j++;
6629 if (e[j-1].size == 0)
6630 break;
6631 }
6632
6633 pos = 0;
6634 maxsize = 0;
6635 start_extent = 0;
6636 i = 0;
6637 do {
6638 unsigned long long esize;
6639
6640 esize = e[i].start - pos;
6641 if (esize >= maxsize) {
6642 maxsize = esize;
6643 start = pos;
6644 start_extent = i;
6645 }
6646 pos = e[i].start + e[i].size;
6647 i++;
6648 } while (e[i-1].size);
6649 free(e);
6650
a7dd165b
DW		 6651 if (maxsize == 0)
6652 return 0;
6653
6654 /* FIXME assumes volume at offset 0 is the first volume in a
6655 * container
6656 */
0dcecb2e
DW		 6657 if (start_extent > 0)
6658 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6659 else
6660 reserve = 0;
6661
6662 if (maxsize < reserve)
a7dd165b 6663 return 0;
0dcecb2e 6664
5551b113 6665 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6666 if (start + reserve > super->create_offset)
a7dd165b 6667 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6668 super->create_offset = start + reserve;
6669
6670 return maxsize - reserve;
6671}
6672
88c32bb1
DW		 6673static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6674{
6675 if (level < 0 || level == 6 || level == 4)
6676 return 0;
6677
6678 /* if we have an orom prevent invalid raid levels */
6679 if (orom)
6680 switch (level) {
6681 case 0: return imsm_orom_has_raid0(orom);
6682 case 1:
6683 if (raiddisks > 2)
6684 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6685 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6686 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6687 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6688 }
6689 else
6690 return 1; /* not on an Intel RAID platform so anything goes */
6691
6692 return 0;
6693}
6694
ca9de185
LM		 6695static int
6696active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6697 int dpa, int verbose)
6698{
6699 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6700 struct mdstat_ent *memb;
ca9de185
LM		 6701 int count = 0;
6702 int num = 0;
594dc1b8 6703 struct md_list *dv;
ca9de185
LM		 6704 int found;
6705
6706 for (memb = mdstat ; memb ; memb = memb->next) {
6707 if (memb->metadata_version &&
fc54fe7a 6708 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6709 (strcmp(&memb->metadata_version[9], name) == 0) &&
6710 !is_subarray(memb->metadata_version+9) &&
6711 memb->members) {
6712 struct dev_member *dev = memb->members;
6713 int fd = -1;
6714 while(dev && (fd < 0)) {
503975b9
N		 6715 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6716 num = sprintf(path, "%s%s", "/dev/", dev->name);
6717 if (num > 0)
6718 fd = open(path, O_RDONLY, 0);
089f9d79 6719 if (num <= 0 || fd < 0) {
676e87a8 6720 pr_vrb("Cannot open %s: %s\n",
503975b9 6721 dev->name, strerror(errno));
ca9de185 6722 }
503975b9 6723 free(path);
ca9de185
LM		 6724 dev = dev->next;
6725 }
6726 found = 0;
089f9d79 6727 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6728 struct mdstat_ent *vol;
6729 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6730 if (vol->active > 0 &&
ca9de185 6731 vol->metadata_version &&
9581efb1 6732 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6733 found++;
6734 count++;
6735 }
6736 }
6737 if (*devlist && (found < dpa)) {
503975b9 6738 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6739 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6740 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6741 dv->found = found;
6742 dv->used = 0;
6743 dv->next = *devlist;
6744 *devlist = dv;
ca9de185
LM		 6745 }
6746 }
6747 if (fd >= 0)
6748 close(fd);
6749 }
6750 }
6751 free_mdstat(mdstat);
6752 return count;
6753}
6754
6755#ifdef DEBUG_LOOP
6756static struct md_list*
6757get_loop_devices(void)
6758{
6759 int i;
6760 struct md_list *devlist = NULL;
594dc1b8 6761 struct md_list *dv;
ca9de185
LM		 6762
6763 for(i = 0; i < 12; i++) {
503975b9
N		 6764 dv = xcalloc(1, sizeof(*dv));
6765 dv->devname = xmalloc(40);
ca9de185
LM		 6766 sprintf(dv->devname, "/dev/loop%d", i);
6767 dv->next = devlist;
6768 devlist = dv;
6769 }
6770 return devlist;
6771}
6772#endif
6773
6774static struct md_list*
6775get_devices(const char *hba_path)
6776{
6777 struct md_list *devlist = NULL;
594dc1b8 6778 struct md_list *dv;
ca9de185
LM		 6779 struct dirent *ent;
6780 DIR *dir;
6781 int err = 0;
6782
6783#if DEBUG_LOOP
6784 devlist = get_loop_devices();
6785 return devlist;
6786#endif
6787 /* scroll through /sys/dev/block looking for devices attached to
6788 * this hba
6789 */
6790 dir = opendir("/sys/dev/block");
6791 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6792 int fd;
6793 char buf[1024];
6794 int major, minor;
6795 char *path = NULL;
6796 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6797 continue;
6798 path = devt_to_devpath(makedev(major, minor));
6799 if (!path)
6800 continue;
6801 if (!path_attached_to_hba(path, hba_path)) {
6802 free(path);
6803 path = NULL;
6804 continue;
6805 }
6806 free(path);
6807 path = NULL;
6808 fd = dev_open(ent->d_name, O_RDONLY);
6809 if (fd >= 0) {
6810 fd2devname(fd, buf);
6811 close(fd);
6812 } else {
e7b84f9d 6813 pr_err("cannot open device: %s\n",
ca9de185
LM		 6814 ent->d_name);
6815 continue;
6816 }
6817
503975b9
N		 6818 dv = xcalloc(1, sizeof(*dv));
6819 dv->devname = xstrdup(buf);
ca9de185
LM		 6820 dv->next = devlist;
6821 devlist = dv;
6822 }
6823 if (err) {
6824 while(devlist) {
6825 dv = devlist;
6826 devlist = devlist->next;
6827 free(dv->devname);
6828 free(dv);
6829 }
6830 }
562aa102 6831 closedir(dir);
ca9de185
LM		 6832 return devlist;
6833}
6834
6835static int
6836count_volumes_list(struct md_list *devlist, char *homehost,
6837 int verbose, int *found)
6838{
6839 struct md_list *tmpdev;
6840 int count = 0;
594dc1b8 6841 struct supertype *st;
ca9de185
LM		 6842
6843 /* first walk the list of devices to find a consistent set
6844 * that match the criterea, if that is possible.
6845 * We flag the ones we like with 'used'.
6846 */
6847 *found = 0;
6848 st = match_metadata_desc_imsm("imsm");
6849 if (st == NULL) {
676e87a8 6850 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6851 return 0;
6852 }
6853
6854 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6855 char *devname = tmpdev->devname;
0a6bff09 6856 dev_t rdev;
ca9de185
LM		 6857 struct supertype *tst;
6858 int dfd;
6859 if (tmpdev->used > 1)
6860 continue;
6861 tst = dup_super(st);
6862 if (tst == NULL) {
676e87a8 6863 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6864 goto err_1;
6865 }
6866 tmpdev->container = 0;
6867 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6868 if (dfd < 0) {
1ade5cc1 6869 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6870 devname, strerror(errno));
6871 tmpdev->used = 2;
0a6bff09 6872 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6873 tmpdev->used = 2;
6874 } else if (must_be_container(dfd)) {
6875 struct supertype *cst;
6876 cst = super_by_fd(dfd, NULL);
6877 if (cst == NULL) {
1ade5cc1 6878 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6879 devname);
6880 tmpdev->used = 2;
6881 } else if (tst->ss != st->ss) {
1ade5cc1 6882 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6883 devname);
6884 tmpdev->used = 2;
6885 } else if (!tst->ss->load_container ||
6886 tst->ss->load_container(tst, dfd, NULL))
6887 tmpdev->used = 2;
6888 else {
6889 tmpdev->container = 1;
6890 }
6891 if (cst)
6892 cst->ss->free_super(cst);
6893 } else {
0a6bff09 6894 tmpdev->st_rdev = rdev;
ca9de185 6895 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6896 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6897 devname);
6898 tmpdev->used = 2;
6899 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6900 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6901 tst->ss->name, devname);
6902 tmpdev->used = 2;
6903 }
6904 }
6905 if (dfd >= 0)
6906 close(dfd);
6907 if (tmpdev->used == 2 || tmpdev->used == 4) {
6908 /* Ignore unrecognised devices during auto-assembly */
6909 goto loop;
6910 }
6911 else {
6912 struct mdinfo info;
6913 tst->ss->getinfo_super(tst, &info, NULL);
6914
6915 if (st->minor_version == -1)
6916 st->minor_version = tst->minor_version;
6917
6918 if (memcmp(info.uuid, uuid_zero,
6919 sizeof(int[4])) == 0) {
6920 /* this is a floating spare. It cannot define
6921 * an array unless there are no more arrays of
6922 * this type to be found. It can be included
6923 * in an array of this type though.
6924 */
6925 tmpdev->used = 3;
6926 goto loop;
6927 }
6928
6929 if (st->ss != tst->ss ||
6930 st->minor_version != tst->minor_version ||
6931 st->ss->compare_super(st, tst) != 0) {
6932 /* Some mismatch. If exactly one array matches this host,
6933 * we can resolve on that one.
6934 * Or, if we are auto assembling, we just ignore the second
6935 * for now.
6936 */
1ade5cc1 6937 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6938 devname);
6939 goto loop;
6940 }
6941 tmpdev->used = 1;
6942 *found = 1;
6943 dprintf("found: devname: %s\n", devname);
6944 }
6945 loop:
6946 if (tst)
6947 tst->ss->free_super(tst);
6948 }
6949 if (*found != 0) {
6950 int err;
6951 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6952 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6953 for (iter = head; iter; iter = iter->next) {
6954 dprintf("content->text_version: %s vol\n",
6955 iter->text_version);
6956 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6957 /* do not assemble arrays with unsupported
6958 configurations */
1ade5cc1 6959 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6960 iter->text_version);
6961 } else
6962 count++;
6963 }
6964 sysfs_free(head);
6965
6966 } else {
1ade5cc1 6967 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6968 err, st->sb);
6969 }
6970 } else {
1ade5cc1 6971 dprintf("no more devices to examine\n");
ca9de185
LM		 6972 }
6973
6974 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6975 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6976 if (count) {
6977 if (count < tmpdev->found)
6978 count = 0;
6979 else
6980 count -= tmpdev->found;
6981 }
6982 }
6983 if (tmpdev->used == 1)
6984 tmpdev->used = 4;
6985 }
6986 err_1:
6987 if (st)
6988 st->ss->free_super(st);
6989 return count;
6990}
6991
d3c11416
AO		 6992static int __count_volumes(char *hba_path, int dpa, int verbose,
6993 int cmp_hba_path)
ca9de185 6994{
72a45777 6995 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6996 int count = 0;
72a45777
PB		 6997 const struct orom_entry *entry;
6998 struct devid_list *dv, *devid_list;
ca9de185 6999
d3c11416 7000 if (!hba_path)
ca9de185
LM		 7001 return 0;
7002
72a45777 7003 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7004 if (strstr(idev->path, hba_path))
7005 break;
72a45777
PB		 7006 }
7007
7008 if (!idev || !idev->dev_id)
ca9de185 7009 return 0;
72a45777
PB		 7010
7011 entry = get_orom_entry_by_device_id(idev->dev_id);
7012
7013 if (!entry || !entry->devid_list)
7014 return 0;
7015
7016 devid_list = entry->devid_list;
7017 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7018 struct md_list *devlist;
d3c11416
AO		 7019 struct sys_dev *device = NULL;
7020 char *hpath;
72a45777
PB		 7021 int found = 0;
7022
d3c11416
AO		 7023 if (cmp_hba_path)
7024 device = device_by_id_and_path(dv->devid, hba_path);
7025 else
7026 device = device_by_id(dv->devid);
7027
72a45777 7028 if (device)
d3c11416 7029 hpath = device->path;
72a45777
PB		 7030 else
7031 return 0;
7032
d3c11416 7033 devlist = get_devices(hpath);
72a45777
PB		 7034 /* if no intel devices return zero volumes */
7035 if (devlist == NULL)
7036 return 0;
7037
d3c11416
AO		 7038 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7039 verbose);
7040 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7041 if (devlist == NULL)
7042 return 0;
7043 do {
7044 found = 0;
7045 count += count_volumes_list(devlist,
7046 NULL,
7047 verbose,
7048 &found);
7049 dprintf("found %d count: %d\n", found, count);
7050 } while (found);
7051
d3c11416 7052 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7053
7054 while (devlist) {
7055 struct md_list *dv = devlist;
7056 devlist = devlist->next;
7057 free(dv->devname);
7058 free(dv);
7059 }
ca9de185
LM		 7060 }
7061 return count;
7062}
7063
d3c11416
AO		 7064static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7065{
7066 if (!hba)
7067 return 0;
7068 if (hba->type == SYS_DEV_VMD) {
7069 struct sys_dev *dev;
7070 int count = 0;
7071
7072 for (dev = find_intel_devices(); dev; dev = dev->next) {
7073 if (dev->type == SYS_DEV_VMD)
7074 count += __count_volumes(dev->path, dpa,
7075 verbose, 1);
7076 }
7077 return count;
7078 }
7079 return __count_volumes(hba->path, dpa, verbose, 0);
7080}
7081
cd9d1ac7
DW		 7082static int imsm_default_chunk(const struct imsm_orom *orom)
7083{
7084 /* up to 512 if the plaform supports it, otherwise the platform max.
7085 * 128 if no platform detected
7086 */
7087 int fs = max(7, orom ? fls(orom->sss) : 0);
7088
7089 return min(512, (1 << fs));
7090}
73408129 7091
6592ce37
DW		 7092static int
7093validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7094 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7095{
660260d0
DW		 7096 /* check/set platform and metadata limits/defaults */
7097 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7098 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7099 super->orom->dpa);
73408129
LM		 7100 return 0;
7101 }
7102
5d500228 7103 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7104 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7105 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7106 level, raiddisks, raiddisks > 1 ? "s" : "");
7107 return 0;
7108 }
cd9d1ac7 7109
7ccc4cc4 7110 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7111 *chunk = imsm_default_chunk(super->orom);
7112
7ccc4cc4 7113 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7114 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7115 return 0;
6592ce37 7116 }
cd9d1ac7 7117
6592ce37
DW		 7118 if (layout != imsm_level_to_layout(level)) {
7119 if (level == 5)
676e87a8 7120 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7121 else if (level == 10)
676e87a8 7122 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7123 else
676e87a8 7124 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7125 layout, level);
7126 return 0;
7127 }
2cc699af 7128
7ccc4cc4 7129 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7130 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7131 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7132 return 0;
7133 }
614902f6 7134
6592ce37
DW		 7135 return 1;
7136}
7137
1011e834 7138/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7139 * FIX ME add ahci details
7140 */
8b353278 7141static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7142 int layout, int raiddisks, int *chunk,
af4348dd
N		 7143 unsigned long long size,
7144 unsigned long long data_offset,
7145 char *dev,
2c514b71
NB		 7146 unsigned long long *freesize,
7147 int verbose)
cdddbdbc 7148{
9e04ac1c 7149 dev_t rdev;
c2c087e6 7150 struct intel_super *super = st->sb;
b2916f25 7151 struct imsm_super *mpb;
c2c087e6
DW		 7152 struct dl *dl;
7153 unsigned long long pos = 0;
7154 unsigned long long maxsize;
7155 struct extent *e;
7156 int i;
cdddbdbc 7157
88c32bb1
DW		 7158 /* We must have the container info already read in. */
7159 if (!super)
c2c087e6
DW		 7160 return 0;
7161
b2916f25
JS		 7162 mpb = super->anchor;
7163
2cc699af 7164 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7165 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7166 return 0;
d54559f0 7167 }
c2c087e6
DW		 7168 if (!dev) {
7169 /* General test: make sure there is space for
2da8544a
DW		 7170 * 'raiddisks' device extents of size 'size' at a given
7171 * offset
c2c087e6 7172 */
e46273eb 7173 unsigned long long minsize = size;
b7528a20 7174 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7175 int dcnt = 0;
7176 if (minsize == 0)
7177 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7178 for (dl = super->disks; dl ; dl = dl->next) {
7179 int found = 0;
7180
bf5a934a 7181 pos = 0;
c2c087e6 7182 i = 0;
05501181 7183 e = get_extents(super, dl, 0);
c2c087e6
DW		 7184 if (!e) continue;
7185 do {
7186 unsigned long long esize;
7187 esize = e[i].start - pos;
7188 if (esize >= minsize)
7189 found = 1;
b7528a20 7190 if (found && start_offset == MaxSector) {
2da8544a
DW		 7191 start_offset = pos;
7192 break;
7193 } else if (found && pos != start_offset) {
7194 found = 0;
7195 break;
7196 }
c2c087e6
DW		 7197 pos = e[i].start + e[i].size;
7198 i++;
7199 } while (e[i-1].size);
7200 if (found)
7201 dcnt++;
7202 free(e);
7203 }
7204 if (dcnt < raiddisks) {
2c514b71 7205 if (verbose)
7a862a02 7206 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7207 dcnt, raiddisks);
c2c087e6
DW		 7208 return 0;
7209 }
7210 return 1;
7211 }
0dcecb2e 7212
c2c087e6 7213 /* This device must be a member of the set */
9e04ac1c 7214 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7215 return 0;
7216 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7217 if (dl->major == (int)major(rdev) &&
7218 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7219 break;
7220 }
7221 if (!dl) {
2c514b71 7222 if (verbose)
7a862a02 7223 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7224 return 0;
a20d2ba5
DW		 7225 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7226 /* If a volume is present then the current creation attempt
7227 * cannot incorporate new spares because the orom may not
7228 * understand this configuration (all member disks must be
7229 * members of each array in the container).
7230 */
7a862a02
N		 7231 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7232 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7233 return 0;
5fe62b94
WD		 7234 } else if (super->orom && mpb->num_raid_devs > 0 &&
7235 mpb->num_disks != raiddisks) {
7a862a02 7236 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7237 return 0;
c2c087e6 7238 }
0dcecb2e
DW		 7239
7240 /* retrieve the largest free space block */
05501181 7241 e = get_extents(super, dl, 0);
c2c087e6
DW		 7242 maxsize = 0;
7243 i = 0;
0dcecb2e
DW		 7244 if (e) {
7245 do {
7246 unsigned long long esize;
7247
7248 esize = e[i].start - pos;
7249 if (esize >= maxsize)
7250 maxsize = esize;
7251 pos = e[i].start + e[i].size;
7252 i++;
7253 } while (e[i-1].size);
7254 dl->e = e;
7255 dl->extent_cnt = i;
7256 } else {
7257 if (verbose)
e7b84f9d 7258 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7259 dev);
7260 return 0;
7261 }
7262 if (maxsize < size) {
7263 if (verbose)
e7b84f9d 7264 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7265 dev, maxsize, size);
7266 return 0;
7267 }
7268
7269 /* count total number of extents for merge */
7270 i = 0;
7271 for (dl = super->disks; dl; dl = dl->next)
7272 if (dl->e)
7273 i += dl->extent_cnt;
7274
7275 maxsize = merge_extents(super, i);
3baa56ab 7276
1a1ced1e
KS		 7277 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7278 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7279
a7dd165b 7280 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7281 if (verbose) {
7282 if (maxsize == 0)
7a862a02 7283 pr_err("no free space left on device. Aborting...\n");
b3071342 7284 else
7a862a02 7285 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7286 maxsize, size);
7287 }
0dcecb2e 7288 return 0;
0dcecb2e
DW		 7289 }
7290
c2c087e6
DW		 7291 *freesize = maxsize;
7292
ca9de185 7293 if (super->orom) {
72a45777 7294 int count = count_volumes(super->hba,
ca9de185
LM		 7295 super->orom->dpa, verbose);
7296 if (super->orom->vphba <= count) {
676e87a8 7297 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7298 super->orom->vphba);
7299 return 0;
7300 }
7301 }
c2c087e6 7302 return 1;
cdddbdbc
DW		 7303}
7304
13bcac90 7305static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7306 unsigned long long size, int chunk,
7307 unsigned long long *freesize)
7308{
7309 struct intel_super *super = st->sb;
7310 struct imsm_super *mpb = super->anchor;
7311 struct dl *dl;
7312 int i;
7313 int extent_cnt;
7314 struct extent *e;
7315 unsigned long long maxsize;
7316 unsigned long long minsize;
7317 int cnt;
7318 int used;
7319
7320 /* find the largest common start free region of the possible disks */
7321 used = 0;
7322 extent_cnt = 0;
7323 cnt = 0;
7324 for (dl = super->disks; dl; dl = dl->next) {
7325 dl->raiddisk = -1;
7326
7327 if (dl->index >= 0)
7328 used++;
7329
7330 /* don't activate new spares if we are orom constrained
7331 * and there is already a volume active in the container
7332 */
7333 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7334 continue;
7335
05501181 7336 e = get_extents(super, dl, 0);
efb30e7f
DW		 7337 if (!e)
7338 continue;
7339 for (i = 1; e[i-1].size; i++)
7340 ;
7341 dl->e = e;
7342 dl->extent_cnt = i;
7343 extent_cnt += i;
7344 cnt++;
7345 }
7346
7347 maxsize = merge_extents(super, extent_cnt);
7348 minsize = size;
7349 if (size == 0)
612e59d8
CA		 7350 /* chunk is in K */
7351 minsize = chunk * 2;
efb30e7f
DW		 7352
7353 if (cnt < raiddisks ||
7354 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7355 maxsize < minsize ||
7356 maxsize == 0) {
e7b84f9d 7357 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7358 return 0; /* No enough free spaces large enough */
7359 }
7360
7361 if (size == 0) {
7362 size = maxsize;
7363 if (chunk) {
612e59d8
CA		 7364 size /= 2 * chunk;
7365 size *= 2 * chunk;
efb30e7f 7366 }
f878b242
LM		 7367 maxsize = size;
7368 }
1a1ced1e
KS		 7369 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7370 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7371 cnt = 0;
7372 for (dl = super->disks; dl; dl = dl->next)
7373 if (dl->e)
7374 dl->raiddisk = cnt++;
7375
7376 *freesize = size;
7377
13bcac90
AK		 7378 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7379
efb30e7f
DW		 7380 return 1;
7381}
7382
13bcac90
AK		 7383static int reserve_space(struct supertype *st, int raiddisks,
7384 unsigned long long size, int chunk,
7385 unsigned long long *freesize)
7386{
7387 struct intel_super *super = st->sb;
7388 struct dl *dl;
7389 int cnt;
7390 int rv = 0;
7391
7392 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7393 if (rv) {
7394 cnt = 0;
7395 for (dl = super->disks; dl; dl = dl->next)
7396 if (dl->e)
7397 dl->raiddisk = cnt++;
7398 rv = 1;
7399 }
7400
7401 return rv;
7402}
7403
bf5a934a 7404static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7405 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7406 unsigned long long data_offset,
bf5a934a 7407 char *dev, unsigned long long *freesize,
5308f117 7408 int consistency_policy, int verbose)
bf5a934a
DW		 7409{
7410 int fd, cfd;
7411 struct mdinfo *sra;
20cbe8d2 7412 int is_member = 0;
bf5a934a 7413
d54559f0
LM		 7414 /* load capability
7415 * if given unused devices create a container
bf5a934a
DW		 7416 * if given given devices in a container create a member volume
7417 */
7418 if (level == LEVEL_CONTAINER) {
7419 /* Must be a fresh device to add to a container */
7420 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7421 raiddisks,
7ccc4cc4 7422 *chunk,
af4348dd 7423 size, data_offset,
bf5a934a
DW		 7424 dev, freesize,
7425 verbose);
7426 }
9587c373 7427
06a6101c
BK		 7428 /*
7429 * Size is given in sectors.
7430 */
7431 if (size && (size < 2048)) {
22dc741f 7432 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7433 /* Depends on algorithm in Create.c :
7434 * if container was given (dev == NULL) return -1,
7435 * if block device was given ( dev != NULL) return 0.
7436 */
7437 return dev ? -1 : 0;
7438 }
7439
8592f29d 7440 if (!dev) {
e91a3bad 7441 if (st->sb) {
ca9de185 7442 struct intel_super *super = st->sb;
e91a3bad 7443 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7444 raiddisks, chunk, size,
e91a3bad
LM		 7445 verbose))
7446 return 0;
efb30e7f
DW		 7447 /* we are being asked to automatically layout a
7448 * new volume based on the current contents of
7449 * the container. If the the parameters can be
7450 * satisfied reserve_space will record the disks,
7451 * start offset, and size of the volume to be
7452 * created. add_to_super and getinfo_super
7453 * detect when autolayout is in progress.
7454 */
ca9de185
LM		 7455 /* assuming that freesize is always given when array is
7456 created */
7457 if (super->orom && freesize) {
7458 int count;
72a45777 7459 count = count_volumes(super->hba,
ca9de185
LM		 7460 super->orom->dpa, verbose);
7461 if (super->orom->vphba <= count) {
676e87a8 7462 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7463 super->orom->vphba);
7464 return 0;
7465 }
7466 }
e91a3bad
LM		 7467 if (freesize)
7468 return reserve_space(st, raiddisks, size,
7ccc4cc4 7469 *chunk, freesize);
8592f29d
N		 7470 }
7471 return 1;
7472 }
bf5a934a
DW		 7473 if (st->sb) {
7474 /* creating in a given container */
7475 return validate_geometry_imsm_volume(st, level, layout,
7476 raiddisks, chunk, size,
af4348dd 7477 data_offset,
bf5a934a
DW		 7478 dev, freesize, verbose);
7479 }
7480
bf5a934a
DW		 7481 /* This device needs to be a device in an 'imsm' container */
7482 fd = open(dev, O_RDONLY|O_EXCL, 0);
7483 if (fd >= 0) {
7484 if (verbose)
e7b84f9d
N		 7485 pr_err("Cannot create this array on device %s\n",
7486 dev);
bf5a934a
DW		 7487 close(fd);
7488 return 0;
7489 }
7490 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7491 if (verbose)
e7b84f9d 7492 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7493 dev, strerror(errno));
7494 return 0;
7495 }
7496 /* Well, it is in use by someone, maybe an 'imsm' container. */
7497 cfd = open_container(fd);
20cbe8d2 7498 close(fd);
bf5a934a 7499 if (cfd < 0) {
bf5a934a 7500 if (verbose)
e7b84f9d 7501 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7502 dev);
7503 return 0;
7504 }
4dd2df09 7505 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7506 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7507 strcmp(sra->text_version, "imsm") == 0)
7508 is_member = 1;
7509 sysfs_free(sra);
7510 if (is_member) {
bf5a934a
DW		 7511 /* This is a member of a imsm container. Load the container
7512 * and try to create a volume
7513 */
7514 struct intel_super *super;
7515
ec50f7b6 7516 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7517 st->sb = super;
4dd2df09 7518 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7519 close(cfd);
7520 return validate_geometry_imsm_volume(st, level, layout,
7521 raiddisks, chunk,
af4348dd 7522 size, data_offset, dev,
ecbd9e81
N		 7523 freesize, 1)
7524 ? 1 : -1;
bf5a934a 7525 }
20cbe8d2 7526 }
bf5a934a 7527
20cbe8d2 7528 if (verbose)
e7b84f9d 7529 pr_err("failed container membership check\n");
20cbe8d2
AW		 7530
7531 close(cfd);
7532 return 0;
bf5a934a 7533}
0bd16cf2 7534
30f58b22 7535static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7536{
7537 struct intel_super *super = st->sb;
7538
30f58b22
DW		 7539 if (level && *level == UnSet)
7540 *level = LEVEL_CONTAINER;
7541
7542 if (level && layout && *layout == UnSet)
7543 *layout = imsm_level_to_layout(*level);
0bd16cf2 7544
cd9d1ac7
DW		 7545 if (chunk && (*chunk == UnSet || *chunk == 0))
7546 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7547}
7548
33414a01
DW		 7549static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7550
3364781b 7551static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7552{
3364781b 7553 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7554 __u8 i;
7555 struct intel_dev **dp;
7556 struct intel_super *super = st->sb;
3364781b 7557 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7558 struct imsm_super *mpb = super->anchor;
7559
3364781b 7560 if (mpb->num_raid_devs == 0)
33414a01 7561 return 2;
33414a01
DW		 7562
7563 /* block deletions that would change the uuid of active subarrays
7564 *
7565 * FIXME when immutable ids are available, but note that we'll
7566 * also need to fixup the invalidated/active subarray indexes in
7567 * mdstat
7568 */
7569 for (i = 0; i < mpb->num_raid_devs; i++) {
7570 char subarray[4];
7571
7572 if (i < current_vol)
7573 continue;
7574 sprintf(subarray, "%u", i);
4dd2df09 7575 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7576 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7577 current_vol, i);
33414a01
DW		 7578
7579 return 2;
7580 }
7581 }
7582
7583 if (st->update_tail) {
503975b9 7584 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7585
33414a01
DW		 7586 u->type = update_kill_array;
7587 u->dev_idx = current_vol;
7588 append_metadata_update(st, u, sizeof(*u));
7589
7590 return 0;
7591 }
7592
7593 for (dp = &super->devlist; *dp;)
7594 if ((*dp)->index == current_vol) {
7595 *dp = (*dp)->next;
7596 } else {
7597 handle_missing(super, (*dp)->dev);
7598 if ((*dp)->index > current_vol)
7599 (*dp)->index--;
7600 dp = &(*dp)->next;
7601 }
7602
7603 /* no more raid devices, all active components are now spares,
7604 * but of course failed are still failed
7605 */
7606 if (--mpb->num_raid_devs == 0) {
7607 struct dl *d;
7608
7609 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7610 if (d->index > -2)
7611 mark_spare(d);
33414a01
DW		 7612 }
7613
7614 super->updates_pending++;
7615
7616 return 0;
7617}
aa534678 7618
a951a4f7 7619static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7620 char *update, struct mddev_ident *ident)
aa534678
DW		 7621{
7622 /* update the subarray currently referenced by ->current_vol */
7623 struct intel_super *super = st->sb;
7624 struct imsm_super *mpb = super->anchor;
7625
aa534678
DW		 7626 if (strcmp(update, "name") == 0) {
7627 char *name = ident->name;
a951a4f7
N		 7628 char *ep;
7629 int vol;
aa534678 7630
4dd2df09 7631 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7632 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7633 return 2;
7634 }
7635
7636 if (!check_name(super, name, 0))
7637 return 2;
7638
a951a4f7
N		 7639 vol = strtoul(subarray, &ep, 10);
7640 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7641 return 2;
7642
aa534678 7643 if (st->update_tail) {
503975b9 7644 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7645
aa534678 7646 u->type = update_rename_array;
a951a4f7 7647 u->dev_idx = vol;
618f4e6d
XN		 7648 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7649 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7650 append_metadata_update(st, u, sizeof(*u));
7651 } else {
7652 struct imsm_dev *dev;
ebad3af2 7653 int i, namelen;
aa534678 7654
a951a4f7 7655 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7656 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7657 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7658 memcpy(dev->volume, name, namelen);
aa534678
DW		 7659 for (i = 0; i < mpb->num_raid_devs; i++) {
7660 dev = get_imsm_dev(super, i);
7661 handle_missing(super, dev);
7662 }
7663 super->updates_pending++;
7664 }
e6e9dd3f
AP		 7665 } else if (strcmp(update, "ppl") == 0 ||
7666 strcmp(update, "no-ppl") == 0) {
7667 int new_policy;
7668 char *ep;
7669 int vol = strtoul(subarray, &ep, 10);
7670
7671 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7672 return 2;
7673
7674 if (strcmp(update, "ppl") == 0)
c2462068 7675 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7676 else
c2462068 7677 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7678
7679 if (st->update_tail) {
7680 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7681
7682 u->type = update_rwh_policy;
7683 u->dev_idx = vol;
7684 u->new_policy = new_policy;
7685 append_metadata_update(st, u, sizeof(*u));
7686 } else {
7687 struct imsm_dev *dev;
7688
7689 dev = get_imsm_dev(super, vol);
7690 dev->rwh_policy = new_policy;
7691 super->updates_pending++;
7692 }
aa534678
DW		 7693 } else
7694 return 2;
7695
7696 return 0;
7697}
bf5a934a 7698
28bce06f
AK		 7699static int is_gen_migration(struct imsm_dev *dev)
7700{
7534230b
AK		 7701 if (dev == NULL)
7702 return 0;
7703
28bce06f
AK		 7704 if (!dev->vol.migr_state)
7705 return 0;
7706
7707 if (migr_type(dev) == MIGR_GEN_MIGR)
7708 return 1;
7709
7710 return 0;
7711}
7712
1e5c6983
DW		 7713static int is_rebuilding(struct imsm_dev *dev)
7714{
7715 struct imsm_map *migr_map;
7716
7717 if (!dev->vol.migr_state)
7718 return 0;
7719
7720 if (migr_type(dev) != MIGR_REBUILD)
7721 return 0;
7722
238c0a71 7723 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7724
7725 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7726 return 1;
7727 else
7728 return 0;
7729}
7730
6ce1fbf1
AK		 7731static int is_initializing(struct imsm_dev *dev)
7732{
7733 struct imsm_map *migr_map;
7734
7735 if (!dev->vol.migr_state)
7736 return 0;
7737
7738 if (migr_type(dev) != MIGR_INIT)
7739 return 0;
7740
238c0a71 7741 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7742
7743 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7744 return 1;
7745
7746 return 0;
6ce1fbf1
AK		 7747}
7748
c47b0ff6
AK		 7749static void update_recovery_start(struct intel_super *super,
7750 struct imsm_dev *dev,
7751 struct mdinfo *array)
1e5c6983
DW		 7752{
7753 struct mdinfo *rebuild = NULL;
7754 struct mdinfo *d;
7755 __u32 units;
7756
7757 if (!is_rebuilding(dev))
7758 return;
7759
7760 /* Find the rebuild target, but punt on the dual rebuild case */
7761 for (d = array->devs; d; d = d->next)
7762 if (d->recovery_start == 0) {
7763 if (rebuild)
7764 return;
7765 rebuild = d;
7766 }
7767
4363fd80
DW		 7768 if (!rebuild) {
7769 /* (?) none of the disks are marked with
7770 * IMSM_ORD_REBUILD, so assume they are missing and the
7771 * disk_ord_tbl was not correctly updated
7772 */
1ade5cc1 7773 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7774 return;
7775 }
7776
1e5c6983 7777 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7778 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7779}
7780
276d77db 7781static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7782
00bbdbda 7783static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7784{
4f5bc454
DW		 7785 /* Given a container loaded by load_super_imsm_all,
7786 * extract information about all the arrays into
7787 * an mdinfo tree.
00bbdbda 7788 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7789 *
7790 * For each imsm_dev create an mdinfo, fill it in,
7791 * then look for matching devices in super->disks
7792 * and create appropriate device mdinfo.
7793 */
7794 struct intel_super *super = st->sb;
949c47a0 7795 struct imsm_super *mpb = super->anchor;
4f5bc454 7796 struct mdinfo *rest = NULL;
00bbdbda 7797 unsigned int i;
81219e70 7798 int sb_errors = 0;
abef11a3
AK		 7799 struct dl *d;
7800 int spare_disks = 0;
b6180160 7801 int current_vol = super->current_vol;
cdddbdbc 7802
19482bcc
AK		 7803 /* do not assemble arrays when not all attributes are supported */
7804 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7805 sb_errors = 1;
7a862a02 7806 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7807 }
7808
abef11a3
AK		 7809 /* count spare devices, not used in maps
7810 */
7811 for (d = super->disks; d; d = d->next)
7812 if (d->index == -1)
7813 spare_disks++;
7814
4f5bc454 7815 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7816 struct imsm_dev *dev;
7817 struct imsm_map *map;
86e3692b 7818 struct imsm_map *map2;
4f5bc454 7819 struct mdinfo *this;
a6482415 7820 int slot;
a6482415 7821 int chunk;
00bbdbda 7822 char *ep;
8b9cd157 7823 int level;
00bbdbda
N		 7824
7825 if (subarray &&
7826 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7827 continue;
7828
7829 dev = get_imsm_dev(super, i);
238c0a71
AK		 7830 map = get_imsm_map(dev, MAP_0);
7831 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7832 level = get_imsm_raid_level(map);
4f5bc454 7833
1ce0101c
DW		 7834 /* do not publish arrays that are in the middle of an
7835 * unsupported migration
7836 */
7837 if (dev->vol.migr_state &&
28bce06f 7838 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7839 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7840 dev->volume);
7841 continue;
7842 }
2db86302
LM		 7843 /* do not publish arrays that are not support by controller's
7844 * OROM/EFI
7845 */
1ce0101c 7846
503975b9 7847 this = xmalloc(sizeof(*this));
4f5bc454 7848
301406c9 7849 super->current_vol = i;
a5d85af7 7850 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7851 this->next = rest;
a6482415 7852 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7853 /* mdadm does not support all metadata features- set the bit in all arrays state */
7854 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7855 level, /* RAID level */
7856 imsm_level_to_layout(level),
81219e70 7857 map->num_members, /* raid disks */
fcc2c9da 7858 &chunk, imsm_dev_size(dev),
81219e70 7859 1 /* verbose */)) {
7a862a02 7860 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7861 dev->volume);
7862 this->array.state |=
7863 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7864 (1<<MD_SB_BLOCK_VOLUME);
7865 }
81219e70
LM		 7866
7867 /* if array has bad blocks, set suitable bit in all arrays state */
7868 if (sb_errors)
7869 this->array.state |=
7870 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7871 (1<<MD_SB_BLOCK_VOLUME);
7872
4f5bc454 7873 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7874 unsigned long long recovery_start;
4f5bc454
DW		 7875 struct mdinfo *info_d;
7876 struct dl *d;
7877 int idx;
9a1608e5 7878 int skip;
7eef0453 7879 __u32 ord;
8b9cd157 7880 int missing = 0;
4f5bc454 7881
9a1608e5 7882 skip = 0;
238c0a71
AK		 7883 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7884 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7885 for (d = super->disks; d ; d = d->next)
7886 if (d->index == idx)
0fbd635c 7887 break;
4f5bc454 7888
1e5c6983 7889 recovery_start = MaxSector;
4f5bc454 7890 if (d == NULL)
9a1608e5 7891 skip = 1;
25ed7e59 7892 if (d && is_failed(&d->disk))
9a1608e5 7893 skip = 1;
8b9cd157 7894 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7895 recovery_start = 0;
1e93d0d1
BK		 7896 if (!(ord & IMSM_ORD_REBUILD))
7897 this->array.working_disks++;
1011e834 7898 /*
9a1608e5 7899 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7900 * reset resync start to avoid a dirty-degraded
7901 * situation when performing the intial sync
9a1608e5 7902 */
8b9cd157
MK		 7903 if (skip)
7904 missing++;
7905
7906 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7907 if ((!able_to_resync(level, missing) ||
7908 recovery_start == 0))
7909 this->resync_start = MaxSector;
7910 } else {
7911 /*
7912 * FIXME handle dirty degraded
7913 */
7914 }
7915
9a1608e5
DW		 7916 if (skip)
7917 continue;
4f5bc454 7918
503975b9 7919 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7920 info_d->next = this->devs;
7921 this->devs = info_d;
7922
4f5bc454
DW		 7923 info_d->disk.number = d->index;
7924 info_d->disk.major = d->major;
7925 info_d->disk.minor = d->minor;
7926 info_d->disk.raid_disk = slot;
1e5c6983 7927 info_d->recovery_start = recovery_start;
86e3692b
AK		 7928 if (map2) {
7929 if (slot < map2->num_members)
7930 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7931 else
7932 this->array.spare_disks++;
86e3692b
AK		 7933 } else {
7934 if (slot < map->num_members)
7935 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7936 else
7937 this->array.spare_disks++;
86e3692b 7938 }
4f5bc454
DW		 7939
7940 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7941 info_d->data_offset = pba_of_lba0(map);
44490938 7942 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7943
7944 if (map->raid_level == 5) {
2432ce9b
AP		 7945 info_d->ppl_sector = this->ppl_sector;
7946 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7947 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7948 recovery_start == 0)
7949 this->resync_start = 0;
06fb291a 7950 }
b12796be 7951
5e46202e 7952 info_d->bb.supported = 1;
b12796be
TM		 7953 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7954 info_d->data_offset,
7955 info_d->component_size,
7956 &info_d->bb);
4f5bc454 7957 }
1e5c6983 7958 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7959 update_recovery_start(super, dev, this);
abef11a3 7960 this->array.spare_disks += spare_disks;
276d77db
AK		 7961
7962 /* check for reshape */
7963 if (this->reshape_active == 1)
7964 recover_backup_imsm(st, this);
9a1608e5 7965 rest = this;
4f5bc454
DW		 7966 }
7967
b6180160 7968 super->current_vol = current_vol;
4f5bc454 7969 return rest;
cdddbdbc
DW		 7970}
7971
3b451610
AK		 7972static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7973 int failed, int look_in_map)
c2a1e7da 7974{
3b451610
AK		 7975 struct imsm_map *map;
7976
7977 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7978
7979 if (!failed)
1011e834 7980 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7981 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7982
7983 switch (get_imsm_raid_level(map)) {
7984 case 0:
7985 return IMSM_T_STATE_FAILED;
7986 break;
7987 case 1:
7988 if (failed < map->num_members)
7989 return IMSM_T_STATE_DEGRADED;
7990 else
7991 return IMSM_T_STATE_FAILED;
7992 break;
7993 case 10:
7994 {
7995 /**
c92a2527
DW		 7996 * check to see if any mirrors have failed, otherwise we
7997 * are degraded. Even numbered slots are mirrored on
7998 * slot+1
c2a1e7da 7999 */
c2a1e7da 8000 int i;
d9b420a5
N		 8001 /* gcc -Os complains that this is unused */
8002 int insync = insync;
c2a1e7da
DW		 8003
8004 for (i = 0; i < map->num_members; i++) {
238c0a71 8005 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8006 int idx = ord_to_idx(ord);
8007 struct imsm_disk *disk;
c2a1e7da 8008
c92a2527 8009 /* reset the potential in-sync count on even-numbered
1011e834 8010 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8011 */
8012 if ((i & 1) == 0)
8013 insync = 2;
c2a1e7da 8014
c92a2527 8015 disk = get_imsm_disk(super, idx);
25ed7e59 8016 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8017 insync--;
c2a1e7da 8018
c92a2527
DW		 8019 /* no in-sync disks left in this mirror the
8020 * array has failed
8021 */
8022 if (insync == 0)
8023 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8024 }
8025
8026 return IMSM_T_STATE_DEGRADED;
8027 }
8028 case 5:
8029 if (failed < 2)
8030 return IMSM_T_STATE_DEGRADED;
8031 else
8032 return IMSM_T_STATE_FAILED;
8033 break;
8034 default:
8035 break;
8036 }
8037
8038 return map->map_state;
8039}
8040
3b451610
AK		 8041static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8042 int look_in_map)
c2a1e7da
DW		 8043{
8044 int i;
8045 int failed = 0;
8046 struct imsm_disk *disk;
d5985138
AK		 8047 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8048 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8049 struct imsm_map *map_for_loop;
0556e1a2
DW		 8050 __u32 ord;
8051 int idx;
d5985138 8052 int idx_1;
c2a1e7da 8053
0556e1a2
DW		 8054 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8055 * disks that are being rebuilt. New failures are recorded to
8056 * map[0]. So we look through all the disks we started with and
8057 * see if any failures are still present, or if any new ones
8058 * have arrived
0556e1a2 8059 */
d5985138
AK		 8060 map_for_loop = map;
8061 if (prev && (map->num_members < prev->num_members))
8062 map_for_loop = prev;
68fe4598
LD		 8063
8064 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8065 idx_1 = -255;
238c0a71
AK		 8066 /* when MAP_X is passed both maps failures are counted
8067 */
d5985138 8068 if (prev &&
089f9d79
JS		 8069 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8070 i < prev->num_members) {
d5985138
AK		 8071 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8072 idx_1 = ord_to_idx(ord);
c2a1e7da 8073
d5985138
AK		 8074 disk = get_imsm_disk(super, idx_1);
8075 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8076 failed++;
8077 }
089f9d79
JS		 8078 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8079 i < map->num_members) {
d5985138
AK		 8080 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8081 idx = ord_to_idx(ord);
8082
8083 if (idx != idx_1) {
8084 disk = get_imsm_disk(super, idx);
8085 if (!disk || is_failed(disk) ||
8086 ord & IMSM_ORD_REBUILD)
8087 failed++;
8088 }
8089 }
c2a1e7da
DW		 8090 }
8091
8092 return failed;
845dea95
NB		 8093}
8094
97b4d0e9
DW		 8095static int imsm_open_new(struct supertype *c, struct active_array *a,
8096 char *inst)
8097{
8098 struct intel_super *super = c->sb;
8099 struct imsm_super *mpb = super->anchor;
bbab0940 8100 struct imsm_update_prealloc_bb_mem u;
9587c373 8101
97b4d0e9 8102 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8103 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8104 return -ENODEV;
8105 }
8106
8107 dprintf("imsm: open_new %s\n", inst);
8108 a->info.container_member = atoi(inst);
bbab0940
TM		 8109
8110 u.type = update_prealloc_badblocks_mem;
8111 imsm_update_metadata_locally(c, &u, sizeof(u));
8112
97b4d0e9
DW		 8113 return 0;
8114}
8115
0c046afd
DW		 8116static int is_resyncing(struct imsm_dev *dev)
8117{
8118 struct imsm_map *migr_map;
8119
8120 if (!dev->vol.migr_state)
8121 return 0;
8122
1484e727
DW		 8123 if (migr_type(dev) == MIGR_INIT ||
8124 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8125 return 1;
8126
4c9bc37b
AK		 8127 if (migr_type(dev) == MIGR_GEN_MIGR)
8128 return 0;
8129
238c0a71 8130 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8131
089f9d79
JS		 8132 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8133 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8134 return 1;
8135 else
8136 return 0;
8137}
8138
0556e1a2 8139/* return true if we recorded new information */
4c9e8c1e
TM		 8140static int mark_failure(struct intel_super *super,
8141 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8142{
0556e1a2
DW		 8143 __u32 ord;
8144 int slot;
8145 struct imsm_map *map;
86c54047
DW		 8146 char buf[MAX_RAID_SERIAL_LEN+3];
8147 unsigned int len, shift = 0;
0556e1a2
DW		 8148
8149 /* new failures are always set in map[0] */
238c0a71 8150 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8151
8152 slot = get_imsm_disk_slot(map, idx);
8153 if (slot < 0)
8154 return 0;
8155
8156 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8157 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8158 return 0;
8159
7d0c5e24
LD		 8160 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8161 buf[MAX_RAID_SERIAL_LEN] = '\000';
8162 strcat(buf, ":0");
86c54047
DW		 8163 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8164 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8165 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8166
f2f27e63 8167 disk->status |= FAILED_DISK;
0556e1a2 8168 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8169 /* mark failures in second map if second map exists and this disk
8170 * in this slot.
8171 * This is valid for migration, initialization and rebuild
8172 */
8173 if (dev->vol.migr_state) {
238c0a71 8174 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8175 int slot2 = get_imsm_disk_slot(map2, idx);
8176
089f9d79 8177 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8178 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8179 idx | IMSM_ORD_REBUILD);
8180 }
d7a1fda2
MT		 8181 if (map->failed_disk_num == 0xff ||
8182 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8183 map->failed_disk_num = slot;
4c9e8c1e
TM		 8184
8185 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8186
0556e1a2
DW		 8187 return 1;
8188}
8189
4c9e8c1e
TM		 8190static void mark_missing(struct intel_super *super,
8191 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8192{
4c9e8c1e 8193 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8194
8195 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8196 return;
8197
47ee5a45
DW		 8198 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8199 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8200}
8201
33414a01
DW		 8202static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8203{
33414a01 8204 struct dl *dl;
33414a01
DW		 8205
8206 if (!super->missing)
8207 return;
33414a01 8208
79b68f1b
PC		 8209 /* When orom adds replacement for missing disk it does
8210 * not remove entry of missing disk, but just updates map with
8211 * new added disk. So it is not enough just to test if there is
8212 * any missing disk, we have to look if there are any failed disks
8213 * in map to stop migration */
8214
33414a01 8215 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8216 /* end process for initialization and rebuild only
8217 */
8218 if (is_gen_migration(dev) == 0) {
fb12a745 8219 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8220
fb12a745
TM		 8221 if (failed) {
8222 __u8 map_state;
8223 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8224 struct imsm_map *map1;
8225 int i, ord, ord_map1;
8226 int rebuilt = 1;
3d59f0c0 8227
fb12a745
TM		 8228 for (i = 0; i < map->num_members; i++) {
8229 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8230 if (!(ord & IMSM_ORD_REBUILD))
8231 continue;
8232
8233 map1 = get_imsm_map(dev, MAP_1);
8234 if (!map1)
8235 continue;
8236
8237 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8238 if (ord_map1 & IMSM_ORD_REBUILD)
8239 rebuilt = 0;
8240 }
8241
8242 if (rebuilt) {
8243 map_state = imsm_check_degraded(super, dev,
8244 failed, MAP_0);
8245 end_migration(dev, super, map_state);
8246 }
8247 }
3d59f0c0 8248 }
33414a01 8249 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8250 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8251 super->updates_pending++;
8252}
8253
f3871fdc
AK		 8254static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8255 long long new_size)
70bdf0dc 8256{
70bdf0dc 8257 unsigned long long array_blocks;
9529d343
MD		 8258 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8259 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8260
8261 if (used_disks == 0) {
8262 /* when problems occures
8263 * return current array_blocks value
8264 */
fcc2c9da 8265 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8266
8267 return array_blocks;
8268 }
8269
8270 /* set array size in metadata
8271 */
9529d343 8272 if (new_size <= 0)
f3871fdc
AK		 8273 /* OLCE size change is caused by added disks
8274 */
44490938 8275 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8276 else
f3871fdc
AK		 8277 /* Online Volume Size Change
8278 * Using available free space
8279 */
8280 array_blocks = new_size;
70bdf0dc 8281
b53bfba6 8282 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8283 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8284
8285 return array_blocks;
8286}
8287
28bce06f
AK		 8288static void imsm_set_disk(struct active_array *a, int n, int state);
8289
0e2d1a4e
AK		 8290static void imsm_progress_container_reshape(struct intel_super *super)
8291{
8292 /* if no device has a migr_state, but some device has a
8293 * different number of members than the previous device, start
8294 * changing the number of devices in this device to match
8295 * previous.
8296 */
8297 struct imsm_super *mpb = super->anchor;
8298 int prev_disks = -1;
8299 int i;
1dfaa380 8300 int copy_map_size;
0e2d1a4e
AK		 8301
8302 for (i = 0; i < mpb->num_raid_devs; i++) {
8303 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8304 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8305 struct imsm_map *map2;
8306 int prev_num_members;
0e2d1a4e
AK		 8307
8308 if (dev->vol.migr_state)
8309 return;
8310
8311 if (prev_disks == -1)
8312 prev_disks = map->num_members;
8313 if (prev_disks == map->num_members)
8314 continue;
8315
8316 /* OK, this array needs to enter reshape mode.
8317 * i.e it needs a migr_state
8318 */
8319
1dfaa380 8320 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8321 prev_num_members = map->num_members;
8322 map->num_members = prev_disks;
8323 dev->vol.migr_state = 1;
8324 dev->vol.curr_migr_unit = 0;
ea672ee1 8325 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8326 for (i = prev_num_members;
8327 i < map->num_members; i++)
8328 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8329 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8330 /* Copy the current map */
1dfaa380 8331 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8332 map2->num_members = prev_num_members;
8333
f3871fdc 8334 imsm_set_array_size(dev, -1);
51d83f5d 8335 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8336 super->updates_pending++;
8337 }
8338}
8339
aad6f216 8340/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8341 * states are handled in imsm_set_disk() with one exception, when a
8342 * resync is stopped due to a new failure this routine will set the
8343 * 'degraded' state for the array.
8344 */
01f157d7 8345static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8346{
8347 int inst = a->info.container_member;
8348 struct intel_super *super = a->container->sb;
949c47a0 8349 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8350 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8351 int failed = imsm_count_failed(super, dev, MAP_0);
8352 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8353 __u32 blocks_per_unit;
a862209d 8354
1af97990
AK		 8355 if (dev->vol.migr_state &&
8356 dev->vol.migr_type == MIGR_GEN_MIGR) {
8357 /* array state change is blocked due to reshape action
aad6f216
N		 8358 * We might need to
8359 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8360 * - finish the reshape (if last_checkpoint is big and action != reshape)
8361 * - update curr_migr_unit
1af97990 8362 */
aad6f216
N		 8363 if (a->curr_action == reshape) {
8364 /* still reshaping, maybe update curr_migr_unit */
633b5610 8365 goto mark_checkpoint;
aad6f216
N		 8366 } else {
8367 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8368 /* for some reason we aborted the reshape.
b66e591b
AK		 8369 *
8370 * disable automatic metadata rollback
8371 * user action is required to recover process
aad6f216 8372 */
b66e591b 8373 if (0) {
238c0a71
AK		 8374 struct imsm_map *map2 =
8375 get_imsm_map(dev, MAP_1);
8376 dev->vol.migr_state = 0;
8377 set_migr_type(dev, 0);
8378 dev->vol.curr_migr_unit = 0;
8379 memcpy(map, map2,
8380 sizeof_imsm_map(map2));
8381 super->updates_pending++;
b66e591b 8382 }
aad6f216
N		 8383 }
8384 if (a->last_checkpoint >= a->info.component_size) {
8385 unsigned long long array_blocks;
8386 int used_disks;
e154ced3 8387 struct mdinfo *mdi;
aad6f216 8388
9529d343 8389 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8390 if (used_disks > 0) {
8391 array_blocks =
44490938 8392 per_dev_array_size(map) *
d55adef9 8393 used_disks;
b53bfba6
TM		 8394 array_blocks =
8395 round_size_to_mb(array_blocks,
8396 used_disks);
d55adef9
AK		 8397 a->info.custom_array_size = array_blocks;
8398 /* encourage manager to update array
8399 * size
8400 */
e154ced3 8401
d55adef9 8402 a->check_reshape = 1;
633b5610 8403 }
e154ced3
AK		 8404 /* finalize online capacity expansion/reshape */
8405 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8406 imsm_set_disk(a,
8407 mdi->disk.raid_disk,
8408 mdi->curr_state);
8409
0e2d1a4e 8410 imsm_progress_container_reshape(super);
e154ced3 8411 }
aad6f216 8412 }
1af97990
AK		 8413 }
8414
47ee5a45 8415 /* before we activate this array handle any missing disks */
33414a01
DW		 8416 if (consistent == 2)
8417 handle_missing(super, dev);
1e5c6983 8418
0c046afd 8419 if (consistent == 2 &&
b7941fd6 8420 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8421 map_state != IMSM_T_STATE_NORMAL ||
8422 dev->vol.migr_state))
01f157d7 8423 consistent = 0;
272906ef 8424
b7941fd6 8425 if (is_resync_complete(&a->info)) {
0c046afd 8426 /* complete intialization / resync,
0556e1a2
DW		 8427 * recovery and interrupted recovery is completed in
8428 * ->set_disk
0c046afd
DW		 8429 */
8430 if (is_resyncing(dev)) {
8431 dprintf("imsm: mark resync done\n");
809da78e 8432 end_migration(dev, super, map_state);
115c3803 8433 super->updates_pending++;
484240d8 8434 a->last_checkpoint = 0;
115c3803 8435 }
b9172665
AK		 8436 } else if ((!is_resyncing(dev) && !failed) &&
8437 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8438 /* mark the start of the init process if nothing is failed */
b7941fd6 8439 dprintf("imsm: mark resync start\n");
1484e727 8440 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8441 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8442 else
8e59f3d8 8443 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8444 super->updates_pending++;
115c3803 8445 }
a862209d 8446
633b5610 8447mark_checkpoint:
5b83bacf
AK		 8448 /* skip checkpointing for general migration,
8449 * it is controlled in mdadm
8450 */
8451 if (is_gen_migration(dev))
8452 goto skip_mark_checkpoint;
8453
1e5c6983 8454 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8455 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8456 if (blocks_per_unit) {
1e5c6983
DW		 8457 __u32 units32;
8458 __u64 units;
8459
4f0a7acc 8460 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8461 units32 = units;
8462
8463 /* check that we did not overflow 32-bits, and that
8464 * curr_migr_unit needs updating
8465 */
8466 if (units32 == units &&
bfd80a56 8467 units32 != 0 &&
1e5c6983
DW		 8468 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8469 dprintf("imsm: mark checkpoint (%u)\n", units32);
8470 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8471 super->updates_pending++;
8472 }
8473 }
f8f603f1 8474
5b83bacf 8475skip_mark_checkpoint:
3393c6af 8476 /* mark dirty / clean */
2432ce9b
AP		 8477 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8478 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8479 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8480 if (consistent) {
8481 dev->vol.dirty = RAIDVOL_CLEAN;
8482 } else {
8483 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8484 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8485 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8486 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8487 }
a862209d
DW		 8488 super->updates_pending++;
8489 }
28bce06f 8490
01f157d7 8491 return consistent;
a862209d
DW		 8492}
8493
6f50473f
TM		 8494static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8495{
8496 int inst = a->info.container_member;
8497 struct intel_super *super = a->container->sb;
8498 struct imsm_dev *dev = get_imsm_dev(super, inst);
8499 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8500
8501 if (slot > map->num_members) {
8502 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8503 slot, map->num_members - 1);
8504 return -1;
8505 }
8506
8507 if (slot < 0)
8508 return -1;
8509
8510 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8511}
8512
8d45d196 8513static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8514{
8d45d196
DW		 8515 int inst = a->info.container_member;
8516 struct intel_super *super = a->container->sb;
949c47a0 8517 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8518 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8519 struct imsm_disk *disk;
7ce05701
LD		 8520 struct mdinfo *mdi;
8521 int recovery_not_finished = 0;
0c046afd 8522 int failed;
6f50473f 8523 int ord;
0c046afd 8524 __u8 map_state;
fb12a745
TM		 8525 int rebuild_done = 0;
8526 int i;
8d45d196 8527
fb12a745 8528 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8529 if (ord < 0)
8d45d196
DW		 8530 return;
8531
4e6e574a 8532 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8533 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8534
5802a811 8535 /* check for new failures */
ae7d61e3 8536 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8537 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8538 super->updates_pending++;
8d45d196 8539 }
47ee5a45 8540
19859edc 8541 /* check if in_sync */
0556e1a2 8542 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8543 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8544
8545 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8546 rebuild_done = 1;
19859edc
DW		 8547 super->updates_pending++;
8548 }
8d45d196 8549
3b451610
AK		 8550 failed = imsm_count_failed(super, dev, MAP_0);
8551 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8552
0c046afd 8553 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8554 dprintf("imsm: Detected transition to state ");
8555 switch (map_state) {
8556 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8557 dprintf("normal: ");
8558 if (is_rebuilding(dev)) {
1ade5cc1 8559 dprintf_cont("while rebuilding");
7ce05701
LD		 8560 /* check if recovery is really finished */
8561 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8562 if (mdi->recovery_start != MaxSector) {
8563 recovery_not_finished = 1;
8564 break;
8565 }
8566 if (recovery_not_finished) {
1ade5cc1
N		 8567 dprintf_cont("\n");
8568 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8569 if (a->last_checkpoint < mdi->recovery_start) {
8570 a->last_checkpoint = mdi->recovery_start;
8571 super->updates_pending++;
8572 }
8573 break;
8574 }
94002678 8575 end_migration(dev, super, map_state);
238c0a71 8576 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8577 map->failed_disk_num = ~0;
8578 super->updates_pending++;
8579 a->last_checkpoint = 0;
8580 break;
8581 }
8582 if (is_gen_migration(dev)) {
1ade5cc1 8583 dprintf_cont("while general migration");
bf2f0071 8584 if (a->last_checkpoint >= a->info.component_size)
809da78e 8585 end_migration(dev, super, map_state);
94002678
AK		 8586 else
8587 map->map_state = map_state;
238c0a71 8588 map = get_imsm_map(dev, MAP_0);
28bce06f 8589 map->failed_disk_num = ~0;
94002678 8590 super->updates_pending++;
bf2f0071 8591 break;
94002678
AK		 8592 }
8593 break;
8594 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8595 dprintf_cont("degraded: ");
089f9d79 8596 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8597 dprintf_cont("mark degraded");
94002678
AK		 8598 map->map_state = map_state;
8599 super->updates_pending++;
8600 a->last_checkpoint = 0;
8601 break;
8602 }
8603 if (is_rebuilding(dev)) {
d7a1fda2 8604 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8605 if (state & DS_FAULTY) {
8606 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8607 if (n == map->failed_disk_num) {
8608 dprintf_cont("end migration");
8609 end_migration(dev, super, map_state);
a4e96fd8 8610 a->last_checkpoint = 0;
d7a1fda2 8611 } else {
a4e96fd8 8612 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8613 map->map_state = map_state;
8614 }
94002678 8615 super->updates_pending++;
fb12a745
TM		 8616 }
8617
a4e96fd8
MT		 8618 if (!rebuild_done)
8619 break;
8620
fb12a745
TM		 8621 /* check if recovery is really finished */
8622 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8623 if (mdi->recovery_start != MaxSector) {
8624 recovery_not_finished = 1;
8625 break;
8626 }
8627 if (recovery_not_finished) {
8628 dprintf_cont("\n");
a4e96fd8 8629 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8630 if (a->last_checkpoint < mdi->recovery_start) {
8631 a->last_checkpoint =
8632 mdi->recovery_start;
8633 super->updates_pending++;
8634 }
8635 break;
94002678 8636 }
fb12a745
TM		 8637
8638 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8639 end_migration(dev, super, map_state);
8640 a->last_checkpoint = 0;
8641 super->updates_pending++;
fb12a745
TM		 8642
8643 for (i = 0; i < map->num_members; i++) {
8644 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8645
8646 if (idx & IMSM_ORD_REBUILD)
8647 map->failed_disk_num = i;
8648 }
8649 super->updates_pending++;
94002678
AK		 8650 break;
8651 }
8652 if (is_gen_migration(dev)) {
1ade5cc1 8653 dprintf_cont("while general migration");
bf2f0071 8654 if (a->last_checkpoint >= a->info.component_size)
809da78e 8655 end_migration(dev, super, map_state);
94002678
AK		 8656 else {
8657 map->map_state = map_state;
3b451610 8658 manage_second_map(super, dev);
94002678
AK		 8659 }
8660 super->updates_pending++;
bf2f0071 8661 break;
28bce06f 8662 }
6ce1fbf1 8663 if (is_initializing(dev)) {
1ade5cc1 8664 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8665 map->map_state = map_state;
8666 super->updates_pending++;
8667 break;
8668 }
94002678
AK		 8669 break;
8670 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8671 dprintf_cont("failed: ");
94002678 8672 if (is_gen_migration(dev)) {
1ade5cc1 8673 dprintf_cont("while general migration");
94002678
AK		 8674 map->map_state = map_state;
8675 super->updates_pending++;
8676 break;
8677 }
8678 if (map->map_state != map_state) {
1ade5cc1 8679 dprintf_cont("mark failed");
94002678
AK		 8680 end_migration(dev, super, map_state);
8681 super->updates_pending++;
8682 a->last_checkpoint = 0;
8683 break;
8684 }
8685 break;
8686 default:
1ade5cc1 8687 dprintf_cont("state %i\n", map_state);
5802a811 8688 }
1ade5cc1 8689 dprintf_cont("\n");
845dea95
NB		 8690}
8691
f796af5d 8692static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8693{
f796af5d 8694 void *buf = mpb;
c2a1e7da
DW		 8695 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8696 unsigned long long dsize;
8697 unsigned long long sectors;
f36a9ecd 8698 unsigned int sector_size;
c2a1e7da 8699
f36a9ecd 8700 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8701 get_dev_size(fd, NULL, &dsize);
8702
f36a9ecd 8703 if (mpb_size > sector_size) {
272f648f 8704 /* -1 to account for anchor */
f36a9ecd 8705 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8706
272f648f 8707 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8708 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8709 SEEK_SET) < 0)
272f648f 8710 return 1;
c2a1e7da 8711
f36a9ecd
PB		 8712 if ((unsigned long long)write(fd, buf + sector_size,
8713 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8714 return 1;
8715 }
c2a1e7da 8716
272f648f 8717 /* first block is stored on second to last sector of the disk */
f36a9ecd 8718 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8719 return 1;
8720
466070ad 8721 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8722 return 1;
8723
c2a1e7da
DW		 8724 return 0;
8725}
8726
2e735d19 8727static void imsm_sync_metadata(struct supertype *container)
845dea95 8728{
2e735d19 8729 struct intel_super *super = container->sb;
c2a1e7da 8730
1a64be56 8731 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8732 if (!super->updates_pending)
8733 return;
8734
36988a3d 8735 write_super_imsm(container, 0);
c2a1e7da
DW		 8736
8737 super->updates_pending = 0;
845dea95
NB		 8738}
8739
272906ef
DW		 8740static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8741{
8742 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8743 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8744 struct dl *dl;
8745
8746 for (dl = super->disks; dl; dl = dl->next)
8747 if (dl->index == i)
8748 break;
8749
25ed7e59 8750 if (dl && is_failed(&dl->disk))
272906ef
DW		 8751 dl = NULL;
8752
8753 if (dl)
1ade5cc1 8754 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8755
8756 return dl;
8757}
8758
a20d2ba5 8759static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8760 struct active_array *a, int activate_new,
8761 struct mdinfo *additional_test_list)
272906ef
DW		 8762{
8763 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8764 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8765 struct imsm_super *mpb = super->anchor;
8766 struct imsm_map *map;
272906ef
DW		 8767 unsigned long long pos;
8768 struct mdinfo *d;
8769 struct extent *ex;
a20d2ba5 8770 int i, j;
272906ef 8771 int found;
569cc43f
DW		 8772 __u32 array_start = 0;
8773 __u32 array_end = 0;
272906ef 8774 struct dl *dl;
6c932028 8775 struct mdinfo *test_list;
272906ef
DW		 8776
8777 for (dl = super->disks; dl; dl = dl->next) {
8778 /* If in this array, skip */
8779 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8780 if (d->state_fd >= 0 &&
8781 d->disk.major == dl->major &&
272906ef 8782 d->disk.minor == dl->minor) {
8ba77d32
AK		 8783 dprintf("%x:%x already in array\n",
8784 dl->major, dl->minor);
272906ef
DW		 8785 break;
8786 }
8787 if (d)
8788 continue;
6c932028
AK		 8789 test_list = additional_test_list;
8790 while (test_list) {
8791 if (test_list->disk.major == dl->major &&
8792 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8793 dprintf("%x:%x already in additional test list\n",
8794 dl->major, dl->minor);
8795 break;
8796 }
6c932028 8797 test_list = test_list->next;
8ba77d32 8798 }
6c932028 8799 if (test_list)
8ba77d32 8800 continue;
272906ef 8801
e553d2a4 8802 /* skip in use or failed drives */
25ed7e59 8803 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8804 dl->index == -2) {
8805 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8806 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8807 continue;
8808 }
8809
a20d2ba5
DW		 8810 /* skip pure spares when we are looking for partially
8811 * assimilated drives
8812 */
8813 if (dl->index == -1 && !activate_new)
8814 continue;
8815
f2cc4f7d
AO		 8816 if (!drive_validate_sector_size(super, dl))
8817 continue;
8818
272906ef 8819 /* Does this unused device have the requisite free space?
a20d2ba5 8820 * It needs to be able to cover all member volumes
272906ef 8821 */
05501181 8822 ex = get_extents(super, dl, 1);
272906ef
DW		 8823 if (!ex) {
8824 dprintf("cannot get extents\n");
8825 continue;
8826 }
a20d2ba5
DW		 8827 for (i = 0; i < mpb->num_raid_devs; i++) {
8828 dev = get_imsm_dev(super, i);
238c0a71 8829 map = get_imsm_map(dev, MAP_0);
272906ef 8830
a20d2ba5
DW		 8831 /* check if this disk is already a member of
8832 * this array
272906ef 8833 */
620b1713 8834 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8835 continue;
8836
8837 found = 0;
8838 j = 0;
8839 pos = 0;
5551b113 8840 array_start = pba_of_lba0(map);
329c8278 8841 array_end = array_start +
44490938 8842 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8843
8844 do {
8845 /* check that we can start at pba_of_lba0 with
44490938 8846 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8847 */
329c8278 8848 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8849 found = 1;
8850 break;
8851 }
8852 pos = ex[j].start + ex[j].size;
8853 j++;
8854 } while (ex[j-1].size);
8855
8856 if (!found)
272906ef 8857 break;
a20d2ba5 8858 }
272906ef
DW		 8859
8860 free(ex);
a20d2ba5 8861 if (i < mpb->num_raid_devs) {
329c8278
DW		 8862 dprintf("%x:%x does not have %u to %u available\n",
8863 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8864 /* No room */
8865 continue;
a20d2ba5
DW		 8866 }
8867 return dl;
272906ef
DW		 8868 }
8869
8870 return dl;
8871}
8872
95d07a2c
LM		 8873static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8874{
8875 struct imsm_dev *dev2;
8876 struct imsm_map *map;
8877 struct dl *idisk;
8878 int slot;
8879 int idx;
8880 __u8 state;
8881
8882 dev2 = get_imsm_dev(cont->sb, dev_idx);
8883 if (dev2) {
238c0a71 8884 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8885 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8886 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8887 if (!map)
8888 return 1;
8889 for (slot = 0; slot < map->num_members; slot++) {
8890 /*
8891 * Check if failed disks are deleted from intel
8892 * disk list or are marked to be deleted
8893 */
238c0a71 8894 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8895 idisk = get_imsm_dl_disk(cont->sb, idx);
8896 /*
8897 * Do not rebuild the array if failed disks
8898 * from failed sub-array are not removed from
8899 * container.
8900 */
8901 if (idisk &&
8902 is_failed(&idisk->disk) &&
8903 (idisk->action != DISK_REMOVE))
8904 return 0;
8905 }
8906 }
8907 }
8908 return 1;
8909}
8910
88758e9d
DW		 8911static struct mdinfo *imsm_activate_spare(struct active_array *a,
8912 struct metadata_update **updates)
8913{
8914 /**
d23fe947
DW		 8915 * Find a device with unused free space and use it to replace a
8916 * failed/vacant region in an array. We replace failed regions one a
8917 * array at a time. The result is that a new spare disk will be added
8918 * to the first failed array and after the monitor has finished
8919 * propagating failures the remainder will be consumed.
88758e9d 8920 *
d23fe947
DW		 8921 * FIXME add a capability for mdmon to request spares from another
8922 * container.
88758e9d
DW		 8923 */
8924
8925 struct intel_super *super = a->container->sb;
88758e9d 8926 int inst = a->info.container_member;
949c47a0 8927 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8928 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8929 int failed = a->info.array.raid_disks;
8930 struct mdinfo *rv = NULL;
8931 struct mdinfo *d;
8932 struct mdinfo *di;
8933 struct metadata_update *mu;
8934 struct dl *dl;
8935 struct imsm_update_activate_spare *u;
8936 int num_spares = 0;
8937 int i;
95d07a2c 8938 int allowed;
88758e9d
DW		 8939
8940 for (d = a->info.devs ; d ; d = d->next) {
8941 if ((d->curr_state & DS_FAULTY) &&
8942 d->state_fd >= 0)
8943 /* wait for Removal to happen */
8944 return NULL;
8945 if (d->state_fd >= 0)
8946 failed--;
8947 }
8948
8949 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8950 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8951
e2962bfc
AK		 8952 if (imsm_reshape_blocks_arrays_changes(super))
8953 return NULL;
1af97990 8954
fc8ca064
AK		 8955 /* Cannot activate another spare if rebuild is in progress already
8956 */
8957 if (is_rebuilding(dev)) {
7a862a02 8958 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8959 return NULL;
8960 }
8961
89c67882
AK		 8962 if (a->info.array.level == 4)
8963 /* No repair for takeovered array
8964 * imsm doesn't support raid4
8965 */
8966 return NULL;
8967
3b451610
AK		 8968 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8969 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8970 return NULL;
8971
83ca7d45
AP		 8972 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8973 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8974 return NULL;
8975 }
8976
95d07a2c
LM		 8977 /*
8978 * If there are any failed disks check state of the other volume.
8979 * Block rebuild if the another one is failed until failed disks
8980 * are removed from container.
8981 */
8982 if (failed) {
7a862a02 8983 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8984 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8985 /* check if states of the other volumes allow for rebuild */
8986 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8987 if (i != inst) {
8988 allowed = imsm_rebuild_allowed(a->container,
8989 i, failed);
8990 if (!allowed)
8991 return NULL;
8992 }
8993 }
8994 }
8995
88758e9d 8996 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8997 for (i = 0; i < a->info.array.raid_disks; i++) {
8998 for (d = a->info.devs ; d ; d = d->next)
8999 if (d->disk.raid_disk == i)
9000 break;
9001 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9002 if (d && (d->state_fd >= 0))
9003 continue;
9004
272906ef 9005 /*
a20d2ba5
DW		 9006 * OK, this device needs recovery. Try to re-add the
9007 * previous occupant of this slot, if this fails see if
9008 * we can continue the assimilation of a spare that was
9009 * partially assimilated, finally try to activate a new
9010 * spare.
272906ef
DW		 9011 */
9012 dl = imsm_readd(super, i, a);
9013 if (!dl)
b303fe21 9014 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9015 if (!dl)
b303fe21 9016 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9017 if (!dl)
9018 continue;
1011e834 9019
272906ef 9020 /* found a usable disk with enough space */
503975b9 9021 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9022
9023 /* dl->index will be -1 in the case we are activating a
9024 * pristine spare. imsm_process_update() will create a
9025 * new index in this case. Once a disk is found to be
9026 * failed in all member arrays it is kicked from the
9027 * metadata
9028 */
9029 di->disk.number = dl->index;
d23fe947 9030
272906ef
DW		 9031 /* (ab)use di->devs to store a pointer to the device
9032 * we chose
9033 */
9034 di->devs = (struct mdinfo *) dl;
9035
9036 di->disk.raid_disk = i;
9037 di->disk.major = dl->major;
9038 di->disk.minor = dl->minor;
9039 di->disk.state = 0;
d23534e4 9040 di->recovery_start = 0;
5551b113 9041 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9042 di->component_size = a->info.component_size;
9043 di->container_member = inst;
5e46202e 9044 di->bb.supported = 1;
2c8890e9 9045 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9046 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9047 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9048 }
148acb7b 9049 super->random = random32();
272906ef
DW		 9050 di->next = rv;
9051 rv = di;
9052 num_spares++;
9053 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9054 i, di->data_offset);
88758e9d
DW		 9055 }
9056
9057 if (!rv)
9058 /* No spares found */
9059 return rv;
9060 /* Now 'rv' has a list of devices to return.
9061 * Create a metadata_update record to update the
9062 * disk_ord_tbl for the array
9063 */
503975b9 9064 mu = xmalloc(sizeof(*mu));
1011e834 9065 mu->buf = xcalloc(num_spares,
503975b9 9066 sizeof(struct imsm_update_activate_spare));
88758e9d 9067 mu->space = NULL;
cb23f1f4 9068 mu->space_list = NULL;
88758e9d
DW		 9069 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9070 mu->next = *updates;
9071 u = (struct imsm_update_activate_spare *) mu->buf;
9072
9073 for (di = rv ; di ; di = di->next) {
9074 u->type = update_activate_spare;
d23fe947
DW		 9075 u->dl = (struct dl *) di->devs;
9076 di->devs = NULL;
88758e9d
DW		 9077 u->slot = di->disk.raid_disk;
9078 u->array = inst;
9079 u->next = u + 1;
9080 u++;
9081 }
9082 (u-1)->next = NULL;
9083 *updates = mu;
9084
9085 return rv;
9086}
9087
54c2c1ea 9088static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9089{
54c2c1ea 9090 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9091 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9092 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9093 struct disk_info *inf = get_disk_info(u);
9094 struct imsm_disk *disk;
8273f55e
DW		 9095 int i;
9096 int j;
8273f55e 9097
54c2c1ea 9098 for (i = 0; i < map->num_members; i++) {
238c0a71 9099 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9100 for (j = 0; j < new_map->num_members; j++)
9101 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9102 return 1;
9103 }
9104
9105 return 0;
9106}
9107
1a64be56
LM		 9108static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9109{
594dc1b8
JS		 9110 struct dl *dl;
9111
1a64be56 9112 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9113 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9114 return dl;
9115 return NULL;
9116}
9117
9118static int remove_disk_super(struct intel_super *super, int major, int minor)
9119{
594dc1b8 9120 struct dl *prev;
1a64be56
LM		 9121 struct dl *dl;
9122
9123 prev = NULL;
9124 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9125 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9126 /* remove */
9127 if (prev)
9128 prev->next = dl->next;
9129 else
9130 super->disks = dl->next;
9131 dl->next = NULL;
9132 __free_imsm_disk(dl);
1ade5cc1 9133 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9134 break;
9135 }
9136 prev = dl;
9137 }
9138 return 0;
9139}
9140
f21e18ca 9141static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9142
1a64be56
LM		 9143static int add_remove_disk_update(struct intel_super *super)
9144{
9145 int check_degraded = 0;
594dc1b8
JS		 9146 struct dl *disk;
9147
1a64be56
LM		 9148 /* add/remove some spares to/from the metadata/contrainer */
9149 while (super->disk_mgmt_list) {
9150 struct dl *disk_cfg;
9151
9152 disk_cfg = super->disk_mgmt_list;
9153 super->disk_mgmt_list = disk_cfg->next;
9154 disk_cfg->next = NULL;
9155
9156 if (disk_cfg->action == DISK_ADD) {
9157 disk_cfg->next = super->disks;
9158 super->disks = disk_cfg;
9159 check_degraded = 1;
1ade5cc1
N		 9160 dprintf("added %x:%x\n",
9161 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9162 } else if (disk_cfg->action == DISK_REMOVE) {
9163 dprintf("Disk remove action processed: %x.%x\n",
9164 disk_cfg->major, disk_cfg->minor);
9165 disk = get_disk_super(super,
9166 disk_cfg->major,
9167 disk_cfg->minor);
9168 if (disk) {
9169 /* store action status */
9170 disk->action = DISK_REMOVE;
9171 /* remove spare disks only */
9172 if (disk->index == -1) {
9173 remove_disk_super(super,
9174 disk_cfg->major,
9175 disk_cfg->minor);
91c97c54
MT		 9176 } else {
9177 disk_cfg->fd = disk->fd;
9178 disk->fd = -1;
1a64be56
LM		 9179 }
9180 }
9181 /* release allocate disk structure */
9182 __free_imsm_disk(disk_cfg);
9183 }
9184 }
9185 return check_degraded;
9186}
9187
a29911da
PC		 9188static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9189 struct intel_super *super,
9190 void ***space_list)
9191{
9192 struct intel_dev *id;
9193 void **tofree = NULL;
9194 int ret_val = 0;
9195
1ade5cc1 9196 dprintf("(enter)\n");
089f9d79 9197 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9198 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9199 return ret_val;
9200 }
089f9d79 9201 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9202 dprintf("imsm: Error: Memory is not allocated\n");
9203 return ret_val;
9204 }
9205
9206 for (id = super->devlist ; id; id = id->next) {
9207 if (id->index == (unsigned)u->subdev) {
9208 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9209 struct imsm_map *map;
9210 struct imsm_dev *new_dev =
9211 (struct imsm_dev *)*space_list;
238c0a71 9212 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9213 int to_state;
9214 struct dl *new_disk;
9215
9216 if (new_dev == NULL)
9217 return ret_val;
9218 *space_list = **space_list;
9219 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9220 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9221 if (migr_map) {
9222 dprintf("imsm: Error: migration in progress");
9223 return ret_val;
9224 }
9225
9226 to_state = map->map_state;
9227 if ((u->new_level == 5) && (map->raid_level == 0)) {
9228 map->num_members++;
9229 /* this should not happen */
9230 if (u->new_disks[0] < 0) {
9231 map->failed_disk_num =
9232 map->num_members - 1;
9233 to_state = IMSM_T_STATE_DEGRADED;
9234 } else
9235 to_state = IMSM_T_STATE_NORMAL;
9236 }
8e59f3d8 9237 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9238 if (u->new_level > -1)
9239 map->raid_level = u->new_level;
238c0a71 9240 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9241 if ((u->new_level == 5) &&
9242 (migr_map->raid_level == 0)) {
9243 int ord = map->num_members - 1;
9244 migr_map->num_members--;
9245 if (u->new_disks[0] < 0)
9246 ord |= IMSM_ORD_REBUILD;
9247 set_imsm_ord_tbl_ent(map,
9248 map->num_members - 1,
9249 ord);
9250 }
9251 id->dev = new_dev;
9252 tofree = (void **)dev;
9253
4bba0439
PC		 9254 /* update chunk size
9255 */
06fb291a
PB		 9256 if (u->new_chunksize > 0) {
9257 unsigned long long num_data_stripes;
9529d343
MD		 9258 struct imsm_map *dest_map =
9259 get_imsm_map(dev, MAP_0);
06fb291a 9260 int used_disks =
9529d343 9261 imsm_num_data_members(dest_map);
06fb291a
PB		 9262
9263 if (used_disks == 0)
9264 return ret_val;
9265
4bba0439
PC		 9266 map->blocks_per_strip =
9267 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9268 num_data_stripes =
fcc2c9da 9269 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9270 num_data_stripes /= map->blocks_per_strip;
9271 num_data_stripes /= map->num_domains;
9272 set_num_data_stripes(map, num_data_stripes);
9273 }
4bba0439 9274
44490938
MD		 9275 /* ensure blocks_per_member has valid value
9276 */
9277 set_blocks_per_member(map,
9278 per_dev_array_size(map) +
9279 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9280
a29911da
PC		 9281 /* add disk
9282 */
089f9d79
JS		 9283 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9284 migr_map->raid_level == map->raid_level)
a29911da
PC		 9285 goto skip_disk_add;
9286
9287 if (u->new_disks[0] >= 0) {
9288 /* use passes spare
9289 */
9290 new_disk = get_disk_super(super,
9291 major(u->new_disks[0]),
9292 minor(u->new_disks[0]));
7a862a02 9293 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9294 major(u->new_disks[0]),
9295 minor(u->new_disks[0]),
9296 new_disk, new_disk->index);
9297 if (new_disk == NULL)
9298 goto error_disk_add;
9299
9300 new_disk->index = map->num_members - 1;
9301 /* slot to fill in autolayout
9302 */
9303 new_disk->raiddisk = new_disk->index;
9304 new_disk->disk.status |= CONFIGURED_DISK;
9305 new_disk->disk.status &= ~SPARE_DISK;
9306 } else
9307 goto error_disk_add;
9308
9309skip_disk_add:
9310 *tofree = *space_list;
9311 /* calculate new size
9312 */
f3871fdc 9313 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9314
9315 ret_val = 1;
9316 }
9317 }
9318
9319 if (tofree)
9320 *space_list = tofree;
9321 return ret_val;
9322
9323error_disk_add:
9324 dprintf("Error: imsm: Cannot find disk.\n");
9325 return ret_val;
9326}
9327
f3871fdc
AK		 9328static int apply_size_change_update(struct imsm_update_size_change *u,
9329 struct intel_super *super)
9330{
9331 struct intel_dev *id;
9332 int ret_val = 0;
9333
1ade5cc1 9334 dprintf("(enter)\n");
089f9d79 9335 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9336 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9337 return ret_val;
9338 }
9339
9340 for (id = super->devlist ; id; id = id->next) {
9341 if (id->index == (unsigned)u->subdev) {
9342 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9343 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9344 int used_disks = imsm_num_data_members(map);
f3871fdc 9345 unsigned long long blocks_per_member;
06fb291a 9346 unsigned long long num_data_stripes;
44490938
MD		 9347 unsigned long long new_size_per_disk;
9348
9349 if (used_disks == 0)
9350 return 0;
f3871fdc
AK		 9351
9352 /* calculate new size
9353 */
44490938
MD		 9354 new_size_per_disk = u->new_size / used_disks;
9355 blocks_per_member = new_size_per_disk +
9356 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9357 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9358 map->blocks_per_strip;
9359 num_data_stripes /= map->num_domains;
9360 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9361 u->new_size, new_size_per_disk,
06fb291a 9362 num_data_stripes);
f3871fdc 9363 set_blocks_per_member(map, blocks_per_member);
06fb291a 9364 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9365 imsm_set_array_size(dev, u->new_size);
9366
9367 ret_val = 1;
9368 break;
9369 }
9370 }
9371
9372 return ret_val;
9373}
9374
061d7da3 9375static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9376 struct intel_super *super,
061d7da3
LO		 9377 struct active_array *active_array)
9378{
9379 struct imsm_super *mpb = super->anchor;
9380 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9381 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9382 struct imsm_map *migr_map;
9383 struct active_array *a;
9384 struct imsm_disk *disk;
9385 __u8 to_state;
9386 struct dl *dl;
9387 unsigned int found;
9388 int failed;
5961eeec 9389 int victim;
061d7da3 9390 int i;
5961eeec 9391 int second_map_created = 0;
061d7da3 9392
5961eeec 9393 for (; u; u = u->next) {
238c0a71 9394 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9395
5961eeec 9396 if (victim < 0)
9397 return 0;
061d7da3 9398
5961eeec 9399 for (dl = super->disks; dl; dl = dl->next)
9400 if (dl == u->dl)
9401 break;
061d7da3 9402
5961eeec 9403 if (!dl) {
7a862a02 9404 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9405 u->dl->index);
9406 return 0;
9407 }
061d7da3 9408
5961eeec 9409 /* count failures (excluding rebuilds and the victim)
9410 * to determine map[0] state
9411 */
9412 failed = 0;
9413 for (i = 0; i < map->num_members; i++) {
9414 if (i == u->slot)
9415 continue;
9416 disk = get_imsm_disk(super,
238c0a71 9417 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9418 if (!disk || is_failed(disk))
9419 failed++;
9420 }
061d7da3 9421
5961eeec 9422 /* adding a pristine spare, assign a new index */
9423 if (dl->index < 0) {
9424 dl->index = super->anchor->num_disks;
9425 super->anchor->num_disks++;
9426 }
9427 disk = &dl->disk;
9428 disk->status |= CONFIGURED_DISK;
9429 disk->status &= ~SPARE_DISK;
9430
9431 /* mark rebuild */
238c0a71 9432 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9433 if (!second_map_created) {
9434 second_map_created = 1;
9435 map->map_state = IMSM_T_STATE_DEGRADED;
9436 migrate(dev, super, to_state, MIGR_REBUILD);
9437 } else
9438 map->map_state = to_state;
238c0a71 9439 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9440 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9441 set_imsm_ord_tbl_ent(migr_map, u->slot,
9442 dl->index | IMSM_ORD_REBUILD);
9443
9444 /* update the family_num to mark a new container
9445 * generation, being careful to record the existing
9446 * family_num in orig_family_num to clean up after
9447 * earlier mdadm versions that neglected to set it.
9448 */
9449 if (mpb->orig_family_num == 0)
9450 mpb->orig_family_num = mpb->family_num;
9451 mpb->family_num += super->random;
9452
9453 /* count arrays using the victim in the metadata */
9454 found = 0;
9455 for (a = active_array; a ; a = a->next) {
9456 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9457 map = get_imsm_map(dev, MAP_0);
061d7da3 9458
5961eeec 9459 if (get_imsm_disk_slot(map, victim) >= 0)
9460 found++;
9461 }
061d7da3 9462
5961eeec 9463 /* delete the victim if it is no longer being
9464 * utilized anywhere
061d7da3 9465 */
5961eeec 9466 if (!found) {
9467 struct dl **dlp;
061d7da3 9468
5961eeec 9469 /* We know that 'manager' isn't touching anything,
9470 * so it is safe to delete
9471 */
9472 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9473 if ((*dlp)->index == victim)
9474 break;
5961eeec 9475
9476 /* victim may be on the missing list */
9477 if (!*dlp)
9478 for (dlp = &super->missing; *dlp;
9479 dlp = &(*dlp)->next)
9480 if ((*dlp)->index == victim)
9481 break;
9482 imsm_delete(super, dlp, victim);
9483 }
061d7da3
LO		 9484 }
9485
9486 return 1;
9487}
a29911da 9488
2e5dc010
N		 9489static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9490 struct intel_super *super,
9491 void ***space_list)
9492{
9493 struct dl *new_disk;
9494 struct intel_dev *id;
9495 int i;
9496 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9497 int disk_count = u->old_raid_disks;
2e5dc010
N		 9498 void **tofree = NULL;
9499 int devices_to_reshape = 1;
9500 struct imsm_super *mpb = super->anchor;
9501 int ret_val = 0;
d098291a 9502 unsigned int dev_id;
2e5dc010 9503
1ade5cc1 9504 dprintf("(enter)\n");
2e5dc010
N		 9505
9506 /* enable spares to use in array */
9507 for (i = 0; i < delta_disks; i++) {
9508 new_disk = get_disk_super(super,
9509 major(u->new_disks[i]),
9510 minor(u->new_disks[i]));
7a862a02 9511 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9512 major(u->new_disks[i]), minor(u->new_disks[i]),
9513 new_disk, new_disk->index);
089f9d79
JS		 9514 if (new_disk == NULL ||
9515 (new_disk->index >= 0 &&
9516 new_disk->index < u->old_raid_disks))
2e5dc010 9517 goto update_reshape_exit;
ee4beede 9518 new_disk->index = disk_count++;
2e5dc010
N		 9519 /* slot to fill in autolayout
9520 */
9521 new_disk->raiddisk = new_disk->index;
9522 new_disk->disk.status |=
9523 CONFIGURED_DISK;
9524 new_disk->disk.status &= ~SPARE_DISK;
9525 }
9526
ed7333bd
AK		 9527 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9528 mpb->num_raid_devs);
2e5dc010
N		 9529 /* manage changes in volume
9530 */
d098291a 9531 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9532 void **sp = *space_list;
9533 struct imsm_dev *newdev;
9534 struct imsm_map *newmap, *oldmap;
9535
d098291a
AK		 9536 for (id = super->devlist ; id; id = id->next) {
9537 if (id->index == dev_id)
9538 break;
9539 }
9540 if (id == NULL)
9541 break;
2e5dc010
N		 9542 if (!sp)
9543 continue;
9544 *space_list = *sp;
9545 newdev = (void*)sp;
9546 /* Copy the dev, but not (all of) the map */
9547 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9548 oldmap = get_imsm_map(id->dev, MAP_0);
9549 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9550 /* Copy the current map */
9551 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9552 /* update one device only
9553 */
9554 if (devices_to_reshape) {
ed7333bd
AK		 9555 dprintf("imsm: modifying subdev: %i\n",
9556 id->index);
2e5dc010
N		 9557 devices_to_reshape--;
9558 newdev->vol.migr_state = 1;
9559 newdev->vol.curr_migr_unit = 0;
ea672ee1 9560 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9561 newmap->num_members = u->new_raid_disks;
9562 for (i = 0; i < delta_disks; i++) {
9563 set_imsm_ord_tbl_ent(newmap,
9564 u->old_raid_disks + i,
9565 u->old_raid_disks + i);
9566 }
9567 /* New map is correct, now need to save old map
9568 */
238c0a71 9569 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9570 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9571
f3871fdc 9572 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9573 }
9574
9575 sp = (void **)id->dev;
9576 id->dev = newdev;
9577 *sp = tofree;
9578 tofree = sp;
8e59f3d8
AK		 9579
9580 /* Clear migration record */
9581 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9582 }
819bc634
AK		 9583 if (tofree)
9584 *space_list = tofree;
2e5dc010
N		 9585 ret_val = 1;
9586
9587update_reshape_exit:
9588
9589 return ret_val;
9590}
9591
bb025c2f 9592static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9593 struct intel_super *super,
9594 void ***space_list)
bb025c2f
KW		 9595{
9596 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9597 struct intel_dev *dv;
9598 struct imsm_dev *dev_new;
bb025c2f
KW		 9599 struct imsm_map *map;
9600 struct dl *dm, *du;
8ca6df95 9601 int i;
bb025c2f
KW		 9602
9603 for (dv = super->devlist; dv; dv = dv->next)
9604 if (dv->index == (unsigned int)u->subarray) {
9605 dev = dv->dev;
9606 break;
9607 }
9608
9609 if (dev == NULL)
9610 return 0;
9611
238c0a71 9612 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9613
9614 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9615 unsigned long long num_data_stripes;
9616
43d5ec18 9617 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9618 if (imsm_count_failed(super, dev, MAP_0) !=
9619 (map->num_members / 2))
43d5ec18
KW		 9620 return 0;
9621
bb025c2f
KW		 9622 /* iterate through devices to mark removed disks as spare */
9623 for (dm = super->disks; dm; dm = dm->next) {
9624 if (dm->disk.status & FAILED_DISK) {
9625 int idx = dm->index;
9626 /* update indexes on the disk list */
9627/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9628 the index values will end up being correct.... NB */
9629 for (du = super->disks; du; du = du->next)
9630 if (du->index > idx)
9631 du->index--;
9632 /* mark as spare disk */
a8619d23 9633 mark_spare(dm);
bb025c2f
KW		 9634 }
9635 }
bb025c2f
KW		 9636 /* update map */
9637 map->num_members = map->num_members / 2;
9638 map->map_state = IMSM_T_STATE_NORMAL;
9639 map->num_domains = 1;
9640 map->raid_level = 0;
9641 map->failed_disk_num = -1;
4a353e6e
RS		 9642 num_data_stripes = imsm_dev_size(dev) / 2;
9643 num_data_stripes /= map->blocks_per_strip;
9644 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9645 }
9646
8ca6df95
KW		 9647 if (u->direction == R0_TO_R10) {
9648 void **space;
4a353e6e
RS		 9649 unsigned long long num_data_stripes;
9650
8ca6df95
KW		 9651 /* update slots in current disk list */
9652 for (dm = super->disks; dm; dm = dm->next) {
9653 if (dm->index >= 0)
9654 dm->index *= 2;
9655 }
9656 /* create new *missing* disks */
9657 for (i = 0; i < map->num_members; i++) {
9658 space = *space_list;
9659 if (!space)
9660 continue;
9661 *space_list = *space;
9662 du = (void *)space;
9663 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9664 du->fd = -1;
9665 du->minor = 0;
9666 du->major = 0;
9667 du->index = (i * 2) + 1;
9668 sprintf((char *)du->disk.serial,
9669 " MISSING_%d", du->index);
9670 sprintf((char *)du->serial,
9671 "MISSING_%d", du->index);
9672 du->next = super->missing;
9673 super->missing = du;
9674 }
9675 /* create new dev and map */
9676 space = *space_list;
9677 if (!space)
9678 return 0;
9679 *space_list = *space;
9680 dev_new = (void *)space;
9681 memcpy(dev_new, dev, sizeof(*dev));
9682 /* update new map */
238c0a71 9683 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9684 map->num_members = map->num_members * 2;
1a2487c2 9685 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9686 map->num_domains = 2;
9687 map->raid_level = 1;
4a353e6e
RS		 9688 num_data_stripes = imsm_dev_size(dev) / 2;
9689 num_data_stripes /= map->blocks_per_strip;
9690 num_data_stripes /= map->num_domains;
9691 set_num_data_stripes(map, num_data_stripes);
9692
8ca6df95
KW		 9693 /* replace dev<->dev_new */
9694 dv->dev = dev_new;
9695 }
bb025c2f
KW		 9696 /* update disk order table */
9697 for (du = super->disks; du; du = du->next)
9698 if (du->index >= 0)
9699 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9700 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9701 if (du->index >= 0) {
9702 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9703 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9704 }
bb025c2f
KW		 9705
9706 return 1;
9707}
9708
e8319a19
DW		 9709static void imsm_process_update(struct supertype *st,
9710 struct metadata_update *update)
9711{
9712 /**
9713 * crack open the metadata_update envelope to find the update record
9714 * update can be one of:
d195167d
AK		 9715 * update_reshape_container_disks - all the arrays in the container
9716 * are being reshaped to have more devices. We need to mark
9717 * the arrays for general migration and convert selected spares
9718 * into active devices.
9719 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9720 * device in an array, update the disk_ord_tbl. If this disk is
9721 * present in all member arrays then also clear the SPARE_DISK
9722 * flag
d195167d
AK		 9723 * update_create_array
9724 * update_kill_array
9725 * update_rename_array
9726 * update_add_remove_disk
e8319a19
DW		 9727 */
9728 struct intel_super *super = st->sb;
4d7b1503 9729 struct imsm_super *mpb;
e8319a19
DW		 9730 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9731
4d7b1503
DW		 9732 /* update requires a larger buf but the allocation failed */
9733 if (super->next_len && !super->next_buf) {
9734 super->next_len = 0;
9735 return;
9736 }
9737
9738 if (super->next_buf) {
9739 memcpy(super->next_buf, super->buf, super->len);
9740 free(super->buf);
9741 super->len = super->next_len;
9742 super->buf = super->next_buf;
9743
9744 super->next_len = 0;
9745 super->next_buf = NULL;
9746 }
9747
9748 mpb = super->anchor;
9749
e8319a19 9750 switch (type) {
0ec5d470
AK		 9751 case update_general_migration_checkpoint: {
9752 struct intel_dev *id;
9753 struct imsm_update_general_migration_checkpoint *u =
9754 (void *)update->buf;
9755
1ade5cc1 9756 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9757
9758 /* find device under general migration */
9759 for (id = super->devlist ; id; id = id->next) {
9760 if (is_gen_migration(id->dev)) {
9761 id->dev->vol.curr_migr_unit =
9762 __cpu_to_le32(u->curr_migr_unit);
9763 super->updates_pending++;
9764 }
9765 }
9766 break;
9767 }
bb025c2f
KW		 9768 case update_takeover: {
9769 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9770 if (apply_takeover_update(u, super, &update->space_list)) {
9771 imsm_update_version_info(super);
bb025c2f 9772 super->updates_pending++;
1a2487c2 9773 }
bb025c2f
KW		 9774 break;
9775 }
9776
78b10e66 9777 case update_reshape_container_disks: {
d195167d 9778 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9779 if (apply_reshape_container_disks_update(
9780 u, super, &update->space_list))
9781 super->updates_pending++;
78b10e66
N		 9782 break;
9783 }
48c5303a 9784 case update_reshape_migration: {
a29911da
PC		 9785 struct imsm_update_reshape_migration *u = (void *)update->buf;
9786 if (apply_reshape_migration_update(
9787 u, super, &update->space_list))
9788 super->updates_pending++;
48c5303a
PC		 9789 break;
9790 }
f3871fdc
AK		 9791 case update_size_change: {
9792 struct imsm_update_size_change *u = (void *)update->buf;
9793 if (apply_size_change_update(u, super))
9794 super->updates_pending++;
9795 break;
9796 }
e8319a19 9797 case update_activate_spare: {
1011e834 9798 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9799 if (apply_update_activate_spare(u, super, st->arrays))
9800 super->updates_pending++;
8273f55e
DW		 9801 break;
9802 }
9803 case update_create_array: {
9804 /* someone wants to create a new array, we need to be aware of
9805 * a few races/collisions:
9806 * 1/ 'Create' called by two separate instances of mdadm
9807 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9808 * devices that have since been assimilated via
9809 * activate_spare.
9810 * In the event this update can not be carried out mdadm will
9811 * (FIX ME) notice that its update did not take hold.
9812 */
9813 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9814 struct intel_dev *dv;
8273f55e
DW		 9815 struct imsm_dev *dev;
9816 struct imsm_map *map, *new_map;
9817 unsigned long long start, end;
9818 unsigned long long new_start, new_end;
9819 int i;
54c2c1ea
DW		 9820 struct disk_info *inf;
9821 struct dl *dl;
8273f55e
DW		 9822
9823 /* handle racing creates: first come first serve */
9824 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9825 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9826 goto create_error;
8273f55e
DW		 9827 }
9828
9829 /* check update is next in sequence */
9830 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9831 dprintf("can not create array %d expected index %d\n",
9832 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9833 goto create_error;
8273f55e
DW		 9834 }
9835
238c0a71 9836 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9837 new_start = pba_of_lba0(new_map);
44490938 9838 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9839 inf = get_disk_info(u);
8273f55e
DW		 9840
9841 /* handle activate_spare versus create race:
9842 * check to make sure that overlapping arrays do not include
9843 * overalpping disks
9844 */
9845 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9846 dev = get_imsm_dev(super, i);
238c0a71 9847 map = get_imsm_map(dev, MAP_0);
5551b113 9848 start = pba_of_lba0(map);
44490938 9849 end = start + per_dev_array_size(map);
8273f55e
DW		 9850 if ((new_start >= start && new_start <= end) ||
9851 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9852 /* overlap */;
9853 else
9854 continue;
9855
9856 if (disks_overlap(super, i, u)) {
1ade5cc1 9857 dprintf("arrays overlap\n");
ba2de7ba 9858 goto create_error;
8273f55e
DW		 9859 }
9860 }
8273f55e 9861
949c47a0
DW		 9862 /* check that prepare update was successful */
9863 if (!update->space) {
1ade5cc1 9864 dprintf("prepare update failed\n");
ba2de7ba 9865 goto create_error;
949c47a0
DW		 9866 }
9867
54c2c1ea
DW		 9868 /* check that all disks are still active before committing
9869 * changes. FIXME: could we instead handle this by creating a
9870 * degraded array? That's probably not what the user expects,
9871 * so better to drop this update on the floor.
9872 */
9873 for (i = 0; i < new_map->num_members; i++) {
9874 dl = serial_to_dl(inf[i].serial, super);
9875 if (!dl) {
1ade5cc1 9876 dprintf("disk disappeared\n");
ba2de7ba 9877 goto create_error;
54c2c1ea 9878 }
949c47a0
DW		 9879 }
9880
8273f55e 9881 super->updates_pending++;
54c2c1ea
DW		 9882
9883 /* convert spares to members and fixup ord_tbl */
9884 for (i = 0; i < new_map->num_members; i++) {
9885 dl = serial_to_dl(inf[i].serial, super);
9886 if (dl->index == -1) {
9887 dl->index = mpb->num_disks;
9888 mpb->num_disks++;
9889 dl->disk.status |= CONFIGURED_DISK;
9890 dl->disk.status &= ~SPARE_DISK;
9891 }
9892 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9893 }
9894
ba2de7ba
DW		 9895 dv = update->space;
9896 dev = dv->dev;
949c47a0
DW		 9897 update->space = NULL;
9898 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9899 dv->index = u->dev_idx;
9900 dv->next = super->devlist;
9901 super->devlist = dv;
8273f55e 9902 mpb->num_raid_devs++;
8273f55e 9903
4d1313e9 9904 imsm_update_version_info(super);
8273f55e 9905 break;
ba2de7ba
DW		 9906 create_error:
9907 /* mdmon knows how to release update->space, but not
9908 * ((struct intel_dev *) update->space)->dev
9909 */
9910 if (update->space) {
9911 dv = update->space;
9912 free(dv->dev);
9913 }
8273f55e 9914 break;
e8319a19 9915 }
33414a01
DW		 9916 case update_kill_array: {
9917 struct imsm_update_kill_array *u = (void *) update->buf;
9918 int victim = u->dev_idx;
9919 struct active_array *a;
9920 struct intel_dev **dp;
9921 struct imsm_dev *dev;
9922
9923 /* sanity check that we are not affecting the uuid of
9924 * active arrays, or deleting an active array
9925 *
9926 * FIXME when immutable ids are available, but note that
9927 * we'll also need to fixup the invalidated/active
9928 * subarray indexes in mdstat
9929 */
9930 for (a = st->arrays; a; a = a->next)
9931 if (a->info.container_member >= victim)
9932 break;
9933 /* by definition if mdmon is running at least one array
9934 * is active in the container, so checking
9935 * mpb->num_raid_devs is just extra paranoia
9936 */
9937 dev = get_imsm_dev(super, victim);
9938 if (a || !dev || mpb->num_raid_devs == 1) {
9939 dprintf("failed to delete subarray-%d\n", victim);
9940 break;
9941 }
9942
9943 for (dp = &super->devlist; *dp;)
f21e18ca 9944 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9945 *dp = (*dp)->next;
9946 } else {
f21e18ca 9947 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9948 (*dp)->index--;
9949 dp = &(*dp)->next;
9950 }
9951 mpb->num_raid_devs--;
9952 super->updates_pending++;
9953 break;
9954 }
aa534678
DW		 9955 case update_rename_array: {
9956 struct imsm_update_rename_array *u = (void *) update->buf;
9957 char name[MAX_RAID_SERIAL_LEN+1];
9958 int target = u->dev_idx;
9959 struct active_array *a;
9960 struct imsm_dev *dev;
9961
9962 /* sanity check that we are not affecting the uuid of
9963 * an active array
9964 */
40659392 9965 memset(name, 0, sizeof(name));
aa534678
DW		 9966 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9967 name[MAX_RAID_SERIAL_LEN] = '\0';
9968 for (a = st->arrays; a; a = a->next)
9969 if (a->info.container_member == target)
9970 break;
9971 dev = get_imsm_dev(super, u->dev_idx);
9972 if (a || !dev || !check_name(super, name, 1)) {
9973 dprintf("failed to rename subarray-%d\n", target);
9974 break;
9975 }
9976
40659392 9977 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 9978 super->updates_pending++;
9979 break;
9980 }
1a64be56 9981 case update_add_remove_disk: {
43dad3d6 9982 /* we may be able to repair some arrays if disks are
095b8088 9983 * being added, check the status of add_remove_disk
1a64be56
LM		 9984 * if discs has been added.
9985 */
9986 if (add_remove_disk_update(super)) {
43dad3d6 9987 struct active_array *a;
072b727f
DW		 9988
9989 super->updates_pending++;
1a64be56 9990 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9991 a->check_degraded = 1;
9992 }
43dad3d6 9993 break;
e8319a19 9994 }
bbab0940
TM		 9995 case update_prealloc_badblocks_mem:
9996 break;
e6e9dd3f
AP		 9997 case update_rwh_policy: {
9998 struct imsm_update_rwh_policy *u = (void *)update->buf;
9999 int target = u->dev_idx;
10000 struct imsm_dev *dev = get_imsm_dev(super, target);
10001 if (!dev) {
10002 dprintf("could not find subarray-%d\n", target);
10003 break;
10004 }
10005
10006 if (dev->rwh_policy != u->new_policy) {
10007 dev->rwh_policy = u->new_policy;
10008 super->updates_pending++;
10009 }
10010 break;
10011 }
1a64be56 10012 default:
ebf3be99 10013 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10014 }
e8319a19 10015}
88758e9d 10016
bc0b9d34
PC		 10017static struct mdinfo *get_spares_for_grow(struct supertype *st);
10018
5fe6f031
N		 10019static int imsm_prepare_update(struct supertype *st,
10020 struct metadata_update *update)
8273f55e 10021{
949c47a0 10022 /**
4d7b1503
DW		 10023 * Allocate space to hold new disk entries, raid-device entries or a new
10024 * mpb if necessary. The manager synchronously waits for updates to
10025 * complete in the monitor, so new mpb buffers allocated here can be
10026 * integrated by the monitor thread without worrying about live pointers
10027 * in the manager thread.
8273f55e 10028 */
095b8088 10029 enum imsm_update_type type;
4d7b1503 10030 struct intel_super *super = st->sb;
f36a9ecd 10031 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10032 struct imsm_super *mpb = super->anchor;
10033 size_t buf_len;
10034 size_t len = 0;
949c47a0 10035
095b8088
N		 10036 if (update->len < (int)sizeof(type))
10037 return 0;
10038
10039 type = *(enum imsm_update_type *) update->buf;
10040
949c47a0 10041 switch (type) {
0ec5d470 10042 case update_general_migration_checkpoint:
095b8088
N		 10043 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10044 return 0;
1ade5cc1 10045 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10046 break;
abedf5fc
KW		 10047 case update_takeover: {
10048 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10049 if (update->len < (int)sizeof(*u))
10050 return 0;
abedf5fc
KW		 10051 if (u->direction == R0_TO_R10) {
10052 void **tail = (void **)&update->space_list;
10053 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10054 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10055 int num_members = map->num_members;
10056 void *space;
10057 int size, i;
abedf5fc
KW		 10058 /* allocate memory for added disks */
10059 for (i = 0; i < num_members; i++) {
10060 size = sizeof(struct dl);
503975b9 10061 space = xmalloc(size);
abedf5fc
KW		 10062 *tail = space;
10063 tail = space;
10064 *tail = NULL;
10065 }
10066 /* allocate memory for new device */
10067 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10068 (num_members * sizeof(__u32));
503975b9
N		 10069 space = xmalloc(size);
10070 *tail = space;
10071 tail = space;
10072 *tail = NULL;
10073 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10074 }
10075
10076 break;
10077 }
78b10e66 10078 case update_reshape_container_disks: {
d195167d
AK		 10079 /* Every raid device in the container is about to
10080 * gain some more devices, and we will enter a
10081 * reconfiguration.
10082 * So each 'imsm_map' will be bigger, and the imsm_vol
10083 * will now hold 2 of them.
10084 * Thus we need new 'struct imsm_dev' allocations sized
10085 * as sizeof_imsm_dev but with more devices in both maps.
10086 */
10087 struct imsm_update_reshape *u = (void *)update->buf;
10088 struct intel_dev *dl;
10089 void **space_tail = (void**)&update->space_list;
10090
095b8088
N		 10091 if (update->len < (int)sizeof(*u))
10092 return 0;
10093
1ade5cc1 10094 dprintf("for update_reshape\n");
d195167d
AK		 10095
10096 for (dl = super->devlist; dl; dl = dl->next) {
10097 int size = sizeof_imsm_dev(dl->dev, 1);
10098 void *s;
d677e0b8
AK		 10099 if (u->new_raid_disks > u->old_raid_disks)
10100 size += sizeof(__u32)*2*
10101 (u->new_raid_disks - u->old_raid_disks);
503975b9 10102 s = xmalloc(size);
d195167d
AK		 10103 *space_tail = s;
10104 space_tail = s;
10105 *space_tail = NULL;
10106 }
10107
10108 len = disks_to_mpb_size(u->new_raid_disks);
10109 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10110 break;
10111 }
48c5303a 10112 case update_reshape_migration: {
bc0b9d34
PC		 10113 /* for migration level 0->5 we need to add disks
10114 * so the same as for container operation we will copy
10115 * device to the bigger location.
10116 * in memory prepared device and new disk area are prepared
10117 * for usage in process update
10118 */
10119 struct imsm_update_reshape_migration *u = (void *)update->buf;
10120 struct intel_dev *id;
10121 void **space_tail = (void **)&update->space_list;
10122 int size;
10123 void *s;
10124 int current_level = -1;
10125
095b8088
N		 10126 if (update->len < (int)sizeof(*u))
10127 return 0;
10128
1ade5cc1 10129 dprintf("for update_reshape\n");
bc0b9d34
PC		 10130
10131 /* add space for bigger array in update
10132 */
10133 for (id = super->devlist; id; id = id->next) {
10134 if (id->index == (unsigned)u->subdev) {
10135 size = sizeof_imsm_dev(id->dev, 1);
10136 if (u->new_raid_disks > u->old_raid_disks)
10137 size += sizeof(__u32)*2*
10138 (u->new_raid_disks - u->old_raid_disks);
503975b9 10139 s = xmalloc(size);
bc0b9d34
PC		 10140 *space_tail = s;
10141 space_tail = s;
10142 *space_tail = NULL;
10143 break;
10144 }
10145 }
10146 if (update->space_list == NULL)
10147 break;
10148
10149 /* add space for disk in update
10150 */
10151 size = sizeof(struct dl);
503975b9 10152 s = xmalloc(size);
bc0b9d34
PC		 10153 *space_tail = s;
10154 space_tail = s;
10155 *space_tail = NULL;
10156
10157 /* add spare device to update
10158 */
10159 for (id = super->devlist ; id; id = id->next)
10160 if (id->index == (unsigned)u->subdev) {
10161 struct imsm_dev *dev;
10162 struct imsm_map *map;
10163
10164 dev = get_imsm_dev(super, u->subdev);
238c0a71 10165 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10166 current_level = map->raid_level;
10167 break;
10168 }
089f9d79 10169 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10170 struct mdinfo *spares;
10171
10172 spares = get_spares_for_grow(st);
10173 if (spares) {
10174 struct dl *dl;
10175 struct mdinfo *dev;
10176
10177 dev = spares->devs;
10178 if (dev) {
10179 u->new_disks[0] =
10180 makedev(dev->disk.major,
10181 dev->disk.minor);
10182 dl = get_disk_super(super,
10183 dev->disk.major,
10184 dev->disk.minor);
10185 dl->index = u->old_raid_disks;
10186 dev = dev->next;
10187 }
10188 sysfs_free(spares);
10189 }
10190 }
10191 len = disks_to_mpb_size(u->new_raid_disks);
10192 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10193 break;
10194 }
f3871fdc 10195 case update_size_change: {
095b8088
N		 10196 if (update->len < (int)sizeof(struct imsm_update_size_change))
10197 return 0;
10198 break;
10199 }
10200 case update_activate_spare: {
10201 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10202 return 0;
f3871fdc
AK		 10203 break;
10204 }
949c47a0
DW		 10205 case update_create_array: {
10206 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10207 struct intel_dev *dv;
54c2c1ea 10208 struct imsm_dev *dev = &u->dev;
238c0a71 10209 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10210 struct dl *dl;
10211 struct disk_info *inf;
10212 int i;
10213 int activate = 0;
949c47a0 10214
095b8088
N		 10215 if (update->len < (int)sizeof(*u))
10216 return 0;
10217
54c2c1ea
DW		 10218 inf = get_disk_info(u);
10219 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10220 /* allocate a new super->devlist entry */
503975b9
N		 10221 dv = xmalloc(sizeof(*dv));
10222 dv->dev = xmalloc(len);
10223 update->space = dv;
949c47a0 10224
54c2c1ea
DW		 10225 /* count how many spares will be converted to members */
10226 for (i = 0; i < map->num_members; i++) {
10227 dl = serial_to_dl(inf[i].serial, super);
10228 if (!dl) {
10229 /* hmm maybe it failed?, nothing we can do about
10230 * it here
10231 */
10232 continue;
10233 }
10234 if (count_memberships(dl, super) == 0)
10235 activate++;
10236 }
10237 len += activate * sizeof(struct imsm_disk);
949c47a0 10238 break;
095b8088
N		 10239 }
10240 case update_kill_array: {
10241 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10242 return 0;
949c47a0
DW		 10243 break;
10244 }
095b8088
N		 10245 case update_rename_array: {
10246 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10247 return 0;
10248 break;
10249 }
10250 case update_add_remove_disk:
10251 /* no update->len needed */
10252 break;
bbab0940
TM		 10253 case update_prealloc_badblocks_mem:
10254 super->extra_space += sizeof(struct bbm_log) -
10255 get_imsm_bbm_log_size(super->bbm_log);
10256 break;
e6e9dd3f
AP		 10257 case update_rwh_policy: {
10258 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10259 return 0;
10260 break;
10261 }
095b8088
N		 10262 default:
10263 return 0;
949c47a0 10264 }
8273f55e 10265
4d7b1503
DW		 10266 /* check if we need a larger metadata buffer */
10267 if (super->next_buf)
10268 buf_len = super->next_len;
10269 else
10270 buf_len = super->len;
10271
bbab0940 10272 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10273 /* ok we need a larger buf than what is currently allocated
10274 * if this allocation fails process_update will notice that
10275 * ->next_len is set and ->next_buf is NULL
10276 */
bbab0940
TM		 10277 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10278 super->extra_space + len, sector_size);
4d7b1503
DW		 10279 if (super->next_buf)
10280 free(super->next_buf);
10281
10282 super->next_len = buf_len;
f36a9ecd 10283 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10284 memset(super->next_buf, 0, buf_len);
10285 else
4d7b1503
DW		 10286 super->next_buf = NULL;
10287 }
5fe6f031 10288 return 1;
8273f55e
DW		 10289}
10290
ae6aad82 10291/* must be called while manager is quiesced */
f21e18ca 10292static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10293{
10294 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10295 struct dl *iter;
10296 struct imsm_dev *dev;
10297 struct imsm_map *map;
4c9e8c1e 10298 unsigned int i, j, num_members;
fb12a745 10299 __u32 ord, ord_map0;
4c9e8c1e 10300 struct bbm_log *log = super->bbm_log;
ae6aad82 10301
1ade5cc1 10302 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10303
10304 /* shift all indexes down one */
10305 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10306 if (iter->index > (int)index)
ae6aad82 10307 iter->index--;
47ee5a45 10308 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10309 if (iter->index > (int)index)
47ee5a45 10310 iter->index--;
ae6aad82
DW		 10311
10312 for (i = 0; i < mpb->num_raid_devs; i++) {
10313 dev = get_imsm_dev(super, i);
238c0a71 10314 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10315 num_members = map->num_members;
10316 for (j = 0; j < num_members; j++) {
10317 /* update ord entries being careful not to propagate
10318 * ord-flags to the first map
10319 */
238c0a71 10320 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10321 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10322
24565c9a
DW		 10323 if (ord_to_idx(ord) <= index)
10324 continue;
ae6aad82 10325
238c0a71 10326 map = get_imsm_map(dev, MAP_0);
fb12a745 10327 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10328 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10329 if (map)
10330 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10331 }
10332 }
10333
4c9e8c1e
TM		 10334 for (i = 0; i < log->entry_count; i++) {
10335 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10336
10337 if (entry->disk_ordinal <= index)
10338 continue;
10339 entry->disk_ordinal--;
10340 }
10341
ae6aad82
DW		 10342 mpb->num_disks--;
10343 super->updates_pending++;
24565c9a
DW		 10344 if (*dlp) {
10345 struct dl *dl = *dlp;
10346
10347 *dlp = (*dlp)->next;
10348 __free_imsm_disk(dl);
10349 }
ae6aad82 10350}
9a717282
AK		 10351
10352static void close_targets(int *targets, int new_disks)
10353{
10354 int i;
10355
10356 if (!targets)
10357 return;
10358
10359 for (i = 0; i < new_disks; i++) {
10360 if (targets[i] >= 0) {
10361 close(targets[i]);
10362 targets[i] = -1;
10363 }
10364 }
10365}
10366
10367static int imsm_get_allowed_degradation(int level, int raid_disks,
10368 struct intel_super *super,
10369 struct imsm_dev *dev)
10370{
10371 switch (level) {
bf5cf7c7 10372 case 1:
9a717282
AK		 10373 case 10:{
10374 int ret_val = 0;
10375 struct imsm_map *map;
10376 int i;
10377
10378 ret_val = raid_disks/2;
10379 /* check map if all disks pairs not failed
10380 * in both maps
10381 */
238c0a71 10382 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10383 for (i = 0; i < ret_val; i++) {
10384 int degradation = 0;
10385 if (get_imsm_disk(super, i) == NULL)
10386 degradation++;
10387 if (get_imsm_disk(super, i + 1) == NULL)
10388 degradation++;
10389 if (degradation == 2)
10390 return 0;
10391 }
238c0a71 10392 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10393 /* if there is no second map
10394 * result can be returned
10395 */
10396 if (map == NULL)
10397 return ret_val;
10398 /* check degradation in second map
10399 */
10400 for (i = 0; i < ret_val; i++) {
10401 int degradation = 0;
10402 if (get_imsm_disk(super, i) == NULL)
10403 degradation++;
10404 if (get_imsm_disk(super, i + 1) == NULL)
10405 degradation++;
10406 if (degradation == 2)
10407 return 0;
10408 }
10409 return ret_val;
10410 }
10411 case 5:
10412 return 1;
10413 case 6:
10414 return 2;
10415 default:
10416 return 0;
10417 }
10418}
10419
687629c2
AK		 10420/*******************************************************************************
10421 * Function: open_backup_targets
10422 * Description: Function opens file descriptors for all devices given in
10423 * info->devs
10424 * Parameters:
10425 * info : general array info
10426 * raid_disks : number of disks
10427 * raid_fds : table of device's file descriptors
9a717282
AK		 10428 * super : intel super for raid10 degradation check
10429 * dev : intel device for raid10 degradation check
687629c2
AK		 10430 * Returns:
10431 * 0 : success
10432 * -1 : fail
10433 ******************************************************************************/
9a717282
AK		 10434int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10435 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10436{
10437 struct mdinfo *sd;
f627f5ad 10438 int i;
9a717282 10439 int opened = 0;
f627f5ad
AK		 10440
10441 for (i = 0; i < raid_disks; i++)
10442 raid_fds[i] = -1;
687629c2
AK		 10443
10444 for (sd = info->devs ; sd ; sd = sd->next) {
10445 char *dn;
10446
10447 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10448 dprintf("disk is faulty!!\n");
10449 continue;
10450 }
10451
089f9d79 10452 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10453 continue;
10454
10455 dn = map_dev(sd->disk.major,
10456 sd->disk.minor, 1);
10457 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10458 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10459 pr_err("cannot open component\n");
9a717282 10460 continue;
687629c2 10461 }
9a717282
AK		 10462 opened++;
10463 }
10464 /* check if maximum array degradation level is not exceeded
10465 */
10466 if ((raid_disks - opened) >
089f9d79
JS		 10467 imsm_get_allowed_degradation(info->new_level, raid_disks,
10468 super, dev)) {
e12b3daa 10469 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10470 close_targets(raid_fds, raid_disks);
10471 return -2;
687629c2
AK		 10472 }
10473 return 0;
10474}
10475
d31ad643
PB		 10476/*******************************************************************************
10477 * Function: validate_container_imsm
10478 * Description: This routine validates container after assemble,
10479 * eg. if devices in container are under the same controller.
10480 *
10481 * Parameters:
10482 * info : linked list with info about devices used in array
10483 * Returns:
10484 * 1 : HBA mismatch
10485 * 0 : Success
10486 ******************************************************************************/
10487int validate_container_imsm(struct mdinfo *info)
10488{
6b781d33
AP		 10489 if (check_env("IMSM_NO_PLATFORM"))
10490 return 0;
d31ad643 10491
6b781d33
AP		 10492 struct sys_dev *idev;
10493 struct sys_dev *hba = NULL;
10494 struct sys_dev *intel_devices = find_intel_devices();
10495 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10496 info->disk.minor));
10497
10498 for (idev = intel_devices; idev; idev = idev->next) {
10499 if (dev_path && strstr(dev_path, idev->path)) {
10500 hba = idev;
10501 break;
d31ad643 10502 }
6b781d33
AP		 10503 }
10504 if (dev_path)
d31ad643
PB		 10505 free(dev_path);
10506
6b781d33
AP		 10507 if (!hba) {
10508 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10509 devid2kname(makedev(info->disk.major, info->disk.minor)));
10510 return 1;
10511 }
10512
10513 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10514 struct mdinfo *dev;
10515
10516 for (dev = info->next; dev; dev = dev->next) {
10517 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10518
10519 struct sys_dev *hba2 = NULL;
10520 for (idev = intel_devices; idev; idev = idev->next) {
10521 if (dev_path && strstr(dev_path, idev->path)) {
10522 hba2 = idev;
10523 break;
d31ad643
PB		 10524 }
10525 }
6b781d33
AP		 10526 if (dev_path)
10527 free(dev_path);
10528
10529 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10530 get_orom_by_device_id(hba2->dev_id);
10531
10532 if (hba2 && hba->type != hba2->type) {
10533 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10534 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10535 return 1;
10536 }
10537
07cb1e57 10538 if (orom != orom2) {
6b781d33
AP		 10539 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10540 " This operation is not supported and can lead to data loss.\n");
10541 return 1;
10542 }
10543
10544 if (!orom) {
10545 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10546 " This operation is not supported and can lead to data loss.\n");
10547 return 1;
10548 }
d31ad643 10549 }
6b781d33 10550
d31ad643
PB		 10551 return 0;
10552}
32141c17 10553
6f50473f
TM		 10554/*******************************************************************************
10555* Function: imsm_record_badblock
10556* Description: This routine stores new bad block record in BBM log
10557*
10558* Parameters:
10559* a : array containing a bad block
10560* slot : disk number containing a bad block
10561* sector : bad block sector
10562* length : bad block sectors range
10563* Returns:
10564* 1 : Success
10565* 0 : Error
10566******************************************************************************/
10567static int imsm_record_badblock(struct active_array *a, int slot,
10568 unsigned long long sector, int length)
10569{
10570 struct intel_super *super = a->container->sb;
10571 int ord;
10572 int ret;
10573
10574 ord = imsm_disk_slot_to_ord(a, slot);
10575 if (ord < 0)
10576 return 0;
10577
10578 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10579 length);
10580 if (ret)
10581 super->updates_pending++;
10582
10583 return ret;
10584}
c07a5a4f
TM		 10585/*******************************************************************************
10586* Function: imsm_clear_badblock
10587* Description: This routine clears bad block record from BBM log
10588*
10589* Parameters:
10590* a : array containing a bad block
10591* slot : disk number containing a bad block
10592* sector : bad block sector
10593* length : bad block sectors range
10594* Returns:
10595* 1 : Success
10596* 0 : Error
10597******************************************************************************/
10598static int imsm_clear_badblock(struct active_array *a, int slot,
10599 unsigned long long sector, int length)
10600{
10601 struct intel_super *super = a->container->sb;
10602 int ord;
10603 int ret;
10604
10605 ord = imsm_disk_slot_to_ord(a, slot);
10606 if (ord < 0)
10607 return 0;
10608
10609 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10610 if (ret)
10611 super->updates_pending++;
10612
10613 return ret;
10614}
928f1424
TM		 10615/*******************************************************************************
10616* Function: imsm_get_badblocks
10617* Description: This routine get list of bad blocks for an array
10618*
10619* Parameters:
10620* a : array
10621* slot : disk number
10622* Returns:
10623* bb : structure containing bad blocks
10624* NULL : error
10625******************************************************************************/
10626static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10627{
10628 int inst = a->info.container_member;
10629 struct intel_super *super = a->container->sb;
10630 struct imsm_dev *dev = get_imsm_dev(super, inst);
10631 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10632 int ord;
10633
10634 ord = imsm_disk_slot_to_ord(a, slot);
10635 if (ord < 0)
10636 return NULL;
10637
10638 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10639 per_dev_array_size(map), &super->bb);
928f1424
TM		 10640
10641 return &super->bb;
10642}
27156a57
TM		 10643/*******************************************************************************
10644* Function: examine_badblocks_imsm
10645* Description: Prints list of bad blocks on a disk to the standard output
10646*
10647* Parameters:
10648* st : metadata handler
10649* fd : open file descriptor for device
10650* devname : device name
10651* Returns:
10652* 0 : Success
10653* 1 : Error
10654******************************************************************************/
10655static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10656{
10657 struct intel_super *super = st->sb;
10658 struct bbm_log *log = super->bbm_log;
10659 struct dl *d = NULL;
10660 int any = 0;
10661
10662 for (d = super->disks; d ; d = d->next) {
10663 if (strcmp(d->devname, devname) == 0)
10664 break;
10665 }
10666
10667 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10668 pr_err("%s doesn't appear to be part of a raid array\n",
10669 devname);
10670 return 1;
10671 }
10672
10673 if (log != NULL) {
10674 unsigned int i;
10675 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10676
10677 for (i = 0; i < log->entry_count; i++) {
10678 if (entry[i].disk_ordinal == d->index) {
10679 unsigned long long sector = __le48_to_cpu(
10680 &entry[i].defective_block_start);
10681 int cnt = entry[i].marked_count + 1;
10682
10683 if (!any) {
10684 printf("Bad-blocks on %s:\n", devname);
10685 any = 1;
10686 }
10687
10688 printf("%20llu for %d sectors\n", sector, cnt);
10689 }
10690 }
10691 }
10692
10693 if (!any)
10694 printf("No bad-blocks list configured on %s\n", devname);
10695
10696 return 0;
10697}
687629c2
AK		 10698/*******************************************************************************
10699 * Function: init_migr_record_imsm
10700 * Description: Function inits imsm migration record
10701 * Parameters:
10702 * super : imsm internal array info
10703 * dev : device under migration
10704 * info : general array info to find the smallest device
10705 * Returns:
10706 * none
10707 ******************************************************************************/
10708void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10709 struct mdinfo *info)
10710{
10711 struct intel_super *super = st->sb;
10712 struct migr_record *migr_rec = super->migr_rec;
10713 int new_data_disks;
10714 unsigned long long dsize, dev_sectors;
10715 long long unsigned min_dev_sectors = -1LLU;
10716 struct mdinfo *sd;
10717 char nm[30];
10718 int fd;
238c0a71
AK		 10719 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10720 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10721 unsigned long long num_migr_units;
3ef4403c 10722 unsigned long long array_blocks;
687629c2
AK		 10723
10724 memset(migr_rec, 0, sizeof(struct migr_record));
10725 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10726
10727 /* only ascending reshape supported now */
10728 migr_rec->ascending_migr = __cpu_to_le32(1);
10729
10730 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10731 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10732 migr_rec->dest_depth_per_unit *=
10733 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10734 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10735 migr_rec->blocks_per_unit =
10736 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10737 migr_rec->dest_depth_per_unit =
10738 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10739 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10740 num_migr_units =
10741 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10742
10743 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10744 num_migr_units++;
9f421827 10745 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10746
10747 migr_rec->post_migr_vol_cap = dev->size_low;
10748 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10749
687629c2
AK		 10750 /* Find the smallest dev */
10751 for (sd = info->devs ; sd ; sd = sd->next) {
10752 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10753 fd = dev_open(nm, O_RDONLY);
10754 if (fd < 0)
10755 continue;
10756 get_dev_size(fd, NULL, &dsize);
10757 dev_sectors = dsize / 512;
10758 if (dev_sectors < min_dev_sectors)
10759 min_dev_sectors = dev_sectors;
10760 close(fd);
10761 }
9f421827 10762 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10763 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10764
10765 write_imsm_migr_rec(st);
10766
10767 return;
10768}
10769
10770/*******************************************************************************
10771 * Function: save_backup_imsm
10772 * Description: Function saves critical data stripes to Migration Copy Area
10773 * and updates the current migration unit status.
10774 * Use restore_stripes() to form a destination stripe,
10775 * and to write it to the Copy Area.
10776 * Parameters:
10777 * st : supertype information
aea93171 10778 * dev : imsm device that backup is saved for
687629c2
AK		 10779 * info : general array info
10780 * buf : input buffer
687629c2
AK		 10781 * length : length of data to backup (blocks_per_unit)
10782 * Returns:
10783 * 0 : success
10784 *, -1 : fail
10785 ******************************************************************************/
10786int save_backup_imsm(struct supertype *st,
10787 struct imsm_dev *dev,
10788 struct mdinfo *info,
10789 void *buf,
687629c2
AK		 10790 int length)
10791{
10792 int rv = -1;
10793 struct intel_super *super = st->sb;
594dc1b8
JS		 10794 unsigned long long *target_offsets;
10795 int *targets;
687629c2 10796 int i;
238c0a71 10797 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10798 int new_disks = map_dest->num_members;
ab724b98
AK		 10799 int dest_layout = 0;
10800 int dest_chunk;
d1877f69 10801 unsigned long long start;
9529d343 10802 int data_disks = imsm_num_data_members(map_dest);
687629c2 10803
503975b9 10804 targets = xmalloc(new_disks * sizeof(int));
687629c2 10805
7e45b550
AK		 10806 for (i = 0; i < new_disks; i++)
10807 targets[i] = -1;
10808
503975b9 10809 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10810
d1877f69 10811 start = info->reshape_progress * 512;
687629c2 10812 for (i = 0; i < new_disks; i++) {
9f421827 10813 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10814 /* move back copy area adderss, it will be moved forward
10815 * in restore_stripes() using start input variable
10816 */
10817 target_offsets[i] -= start/data_disks;
687629c2
AK		 10818 }
10819
9a717282
AK		 10820 if (open_backup_targets(info, new_disks, targets,
10821 super, dev))
687629c2
AK		 10822 goto abort;
10823
68eb8bc6 10824 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10825 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10826
687629c2
AK		 10827 if (restore_stripes(targets, /* list of dest devices */
10828 target_offsets, /* migration record offsets */
10829 new_disks,
ab724b98
AK		 10830 dest_chunk,
10831 map_dest->raid_level,
10832 dest_layout,
10833 -1, /* source backup file descriptor */
10834 0, /* input buf offset
10835 * always 0 buf is already offseted */
d1877f69 10836 start,
687629c2
AK		 10837 length,
10838 buf) != 0) {
e7b84f9d 10839 pr_err("Error restoring stripes\n");
687629c2
AK		 10840 goto abort;
10841 }
10842
10843 rv = 0;
10844
10845abort:
10846 if (targets) {
9a717282 10847 close_targets(targets, new_disks);
687629c2
AK		 10848 free(targets);
10849 }
10850 free(target_offsets);
10851
10852 return rv;
10853}
10854
10855/*******************************************************************************
10856 * Function: save_checkpoint_imsm
10857 * Description: Function called for current unit status update
10858 * in the migration record. It writes it to disk.
10859 * Parameters:
10860 * super : imsm internal array info
10861 * info : general array info
10862 * Returns:
10863 * 0: success
10864 * 1: failure
0228d92c
AK		 10865 * 2: failure, means no valid migration record
10866 * / no general migration in progress /
687629c2
AK		 10867 ******************************************************************************/
10868int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10869{
10870 struct intel_super *super = st->sb;
f8b72ef5
AK		 10871 unsigned long long blocks_per_unit;
10872 unsigned long long curr_migr_unit;
10873
2e062e82 10874 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10875 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10876 return 1;
10877 }
10878
f8b72ef5
AK		 10879 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10880 if (blocks_per_unit == 0) {
0228d92c
AK		 10881 dprintf("imsm: no migration in progress.\n");
10882 return 2;
687629c2 10883 }
f8b72ef5
AK		 10884 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10885 /* check if array is alligned to copy area
10886 * if it is not alligned, add one to current migration unit value
10887 * this can happend on array reshape finish only
10888 */
10889 if (info->reshape_progress % blocks_per_unit)
10890 curr_migr_unit++;
687629c2 10891
9f421827 10892 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10893 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10894 set_migr_dest_1st_member_lba(super->migr_rec,
10895 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10896
687629c2 10897 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10898 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10899 return 1;
10900 }
10901
10902 return 0;
10903}
10904
276d77db
AK		 10905/*******************************************************************************
10906 * Function: recover_backup_imsm
10907 * Description: Function recovers critical data from the Migration Copy Area
10908 * while assembling an array.
10909 * Parameters:
10910 * super : imsm internal array info
10911 * info : general array info
10912 * Returns:
10913 * 0 : success (or there is no data to recover)
10914 * 1 : fail
10915 ******************************************************************************/
10916int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10917{
10918 struct intel_super *super = st->sb;
10919 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10920 struct imsm_map *map_dest;
276d77db
AK		 10921 struct intel_dev *id = NULL;
10922 unsigned long long read_offset;
10923 unsigned long long write_offset;
10924 unsigned unit_len;
10925 int *targets = NULL;
10926 int new_disks, i, err;
10927 char *buf = NULL;
10928 int retval = 1;
f36a9ecd 10929 unsigned int sector_size = super->sector_size;
9f421827
PB		 10930 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10931 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10932 char buffer[20];
6c3560c0 10933 int skipped_disks = 0;
276d77db
AK		 10934
10935 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10936 if (err < 1)
10937 return 1;
10938
10939 /* recover data only during assemblation */
10940 if (strncmp(buffer, "inactive", 8) != 0)
10941 return 0;
10942 /* no data to recover */
10943 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10944 return 0;
10945 if (curr_migr_unit >= num_migr_units)
10946 return 1;
10947
10948 /* find device during reshape */
10949 for (id = super->devlist; id; id = id->next)
10950 if (is_gen_migration(id->dev))
10951 break;
10952 if (id == NULL)
10953 return 1;
10954
238c0a71 10955 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10956 new_disks = map_dest->num_members;
10957
9f421827 10958 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 10959
9f421827 10960 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 10961 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10962
10963 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10964 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10965 goto abort;
503975b9 10966 targets = xcalloc(new_disks, sizeof(int));
276d77db 10967
9a717282 10968 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10969 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10970 goto abort;
10971 }
276d77db
AK		 10972
10973 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10974 if (targets[i] < 0) {
10975 skipped_disks++;
10976 continue;
10977 }
276d77db 10978 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10979 pr_err("Cannot seek to block: %s\n",
10980 strerror(errno));
137debce
AK		 10981 skipped_disks++;
10982 continue;
276d77db 10983 }
9ec11d1a 10984 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10985 pr_err("Cannot read copy area block: %s\n",
10986 strerror(errno));
137debce
AK		 10987 skipped_disks++;
10988 continue;
276d77db
AK		 10989 }
10990 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10991 pr_err("Cannot seek to block: %s\n",
10992 strerror(errno));
137debce
AK		 10993 skipped_disks++;
10994 continue;
276d77db 10995 }
9ec11d1a 10996 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10997 pr_err("Cannot restore block: %s\n",
10998 strerror(errno));
137debce
AK		 10999 skipped_disks++;
11000 continue;
276d77db
AK		 11001 }
11002 }
11003
137debce
AK		 11004 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11005 new_disks,
11006 super,
11007 id->dev)) {
7a862a02 11008 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11009 goto abort;
11010 }
11011
befb629b
AK		 11012 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11013 /* ignore error == 2, this can mean end of reshape here
11014 */
7a862a02 11015 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11016 } else
276d77db 11017 retval = 0;
276d77db
AK		 11018
11019abort:
11020 if (targets) {
11021 for (i = 0; i < new_disks; i++)
11022 if (targets[i])
11023 close(targets[i]);
11024 free(targets);
11025 }
11026 free(buf);
11027 return retval;
11028}
11029
2cda7640
ML		 11030static char disk_by_path[] = "/dev/disk/by-path/";
11031
11032static const char *imsm_get_disk_controller_domain(const char *path)
11033{
2cda7640 11034 char disk_path[PATH_MAX];
96234762
LM		 11035 char *drv=NULL;
11036 struct stat st;
2cda7640 11037
6d8d290a 11038 strcpy(disk_path, disk_by_path);
96234762
LM		 11039 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11040 if (stat(disk_path, &st) == 0) {
11041 struct sys_dev* hba;
594dc1b8 11042 char *path;
96234762
LM		 11043
11044 path = devt_to_devpath(st.st_rdev);
11045 if (path == NULL)
11046 return "unknown";
11047 hba = find_disk_attached_hba(-1, path);
11048 if (hba && hba->type == SYS_DEV_SAS)
11049 drv = "isci";
11050 else if (hba && hba->type == SYS_DEV_SATA)
11051 drv = "ahci";
c6839718
MT		 11052 else if (hba && hba->type == SYS_DEV_VMD)
11053 drv = "vmd";
11054 else if (hba && hba->type == SYS_DEV_NVME)
11055 drv = "nvme";
1011e834 11056 else
96234762
LM		 11057 drv = "unknown";
11058 dprintf("path: %s hba: %s attached: %s\n",
11059 path, (hba) ? hba->path : "NULL", drv);
11060 free(path);
2cda7640 11061 }
96234762 11062 return drv;
2cda7640
ML		 11063}
11064
4dd2df09 11065static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11066{
4dd2df09 11067 static char devnm[32];
78b10e66
N		 11068 char subdev_name[20];
11069 struct mdstat_ent *mdstat;
11070
11071 sprintf(subdev_name, "%d", subdev);
11072 mdstat = mdstat_by_subdev(subdev_name, container);
11073 if (!mdstat)
4dd2df09 11074 return NULL;
78b10e66 11075
4dd2df09 11076 strcpy(devnm, mdstat->devnm);
78b10e66 11077 free_mdstat(mdstat);
4dd2df09 11078 return devnm;
78b10e66
N		 11079}
11080
11081static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11082 struct geo_params *geo,
fbf3d202
AK		 11083 int *old_raid_disks,
11084 int direction)
78b10e66 11085{
694575e7
KW		 11086 /* currently we only support increasing the number of devices
11087 * for a container. This increases the number of device for each
11088 * member array. They must all be RAID0 or RAID5.
11089 */
78b10e66
N		 11090 int ret_val = 0;
11091 struct mdinfo *info, *member;
11092 int devices_that_can_grow = 0;
11093
7a862a02 11094 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11095
d04f65f4 11096 if (geo->size > 0 ||
78b10e66
N		 11097 geo->level != UnSet ||
11098 geo->layout != UnSet ||
11099 geo->chunksize != 0 ||
11100 geo->raid_disks == UnSet) {
7a862a02 11101 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11102 return ret_val;
11103 }
11104
fbf3d202 11105 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11106 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11107 return ret_val;
11108 }
11109
78b10e66
N		 11110 info = container_content_imsm(st, NULL);
11111 for (member = info; member; member = member->next) {
4dd2df09 11112 char *result;
78b10e66
N		 11113
11114 dprintf("imsm: checking device_num: %i\n",
11115 member->container_member);
11116
d7d205bd 11117 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11118 /* we work on container for Online Capacity Expansion
11119 * only so raid_disks has to grow
11120 */
7a862a02 11121 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11122 break;
11123 }
11124
089f9d79 11125 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11126 /* we cannot use this container with other raid level
11127 */
7a862a02 11128 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11129 info->array.level);
11130 break;
11131 } else {
11132 /* check for platform support
11133 * for this raid level configuration
11134 */
11135 struct intel_super *super = st->sb;
11136 if (!is_raid_level_supported(super->orom,
11137 member->array.level,
11138 geo->raid_disks)) {
7a862a02 11139 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11140 info->array.level,
11141 geo->raid_disks,
11142 geo->raid_disks > 1 ? "s" : "");
11143 break;
11144 }
2a4a08e7
AK		 11145 /* check if component size is aligned to chunk size
11146 */
11147 if (info->component_size %
11148 (info->array.chunk_size/512)) {
7a862a02 11149 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11150 break;
11151 }
78b10e66
N		 11152 }
11153
11154 if (*old_raid_disks &&
11155 info->array.raid_disks != *old_raid_disks)
11156 break;
11157 *old_raid_disks = info->array.raid_disks;
11158
11159 /* All raid5 and raid0 volumes in container
11160 * have to be ready for Online Capacity Expansion
11161 * so they need to be assembled. We have already
11162 * checked that no recovery etc is happening.
11163 */
4dd2df09
N		 11164 result = imsm_find_array_devnm_by_subdev(member->container_member,
11165 st->container_devnm);
11166 if (result == NULL) {
78b10e66
N		 11167 dprintf("imsm: cannot find array\n");
11168 break;
11169 }
11170 devices_that_can_grow++;
11171 }
11172 sysfs_free(info);
11173 if (!member && devices_that_can_grow)
11174 ret_val = 1;
11175
11176 if (ret_val)
1ade5cc1 11177 dprintf("Container operation allowed\n");
78b10e66 11178 else
1ade5cc1 11179 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11180
11181 return ret_val;
11182}
11183
11184/* Function: get_spares_for_grow
11185 * Description: Allocates memory and creates list of spare devices
1011e834 11186 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11187 * Parameters: Pointer to the supertype structure
11188 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11189 * NULL if fail
78b10e66
N		 11190 */
11191static struct mdinfo *get_spares_for_grow(struct supertype *st)
11192{
fbfdcb06
AO		 11193 struct spare_criteria sc;
11194
11195 get_spare_criteria_imsm(st, &sc);
11196 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11197}
11198
11199/******************************************************************************
11200 * function: imsm_create_metadata_update_for_reshape
11201 * Function creates update for whole IMSM container.
11202 *
11203 ******************************************************************************/
11204static int imsm_create_metadata_update_for_reshape(
11205 struct supertype *st,
11206 struct geo_params *geo,
11207 int old_raid_disks,
11208 struct imsm_update_reshape **updatep)
11209{
11210 struct intel_super *super = st->sb;
11211 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11212 int update_memory_size;
11213 struct imsm_update_reshape *u;
11214 struct mdinfo *spares;
78b10e66 11215 int i;
594dc1b8 11216 int delta_disks;
bbd24d86 11217 struct mdinfo *dev;
78b10e66 11218
1ade5cc1 11219 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11220
11221 delta_disks = geo->raid_disks - old_raid_disks;
11222
11223 /* size of all update data without anchor */
11224 update_memory_size = sizeof(struct imsm_update_reshape);
11225
11226 /* now add space for spare disks that we need to add. */
11227 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11228
503975b9 11229 u = xcalloc(1, update_memory_size);
78b10e66
N		 11230 u->type = update_reshape_container_disks;
11231 u->old_raid_disks = old_raid_disks;
11232 u->new_raid_disks = geo->raid_disks;
11233
11234 /* now get spare disks list
11235 */
11236 spares = get_spares_for_grow(st);
11237
d7be7d87 11238 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11239 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11240 i = -1;
78b10e66
N		 11241 goto abort;
11242 }
11243
11244 /* we have got spares
11245 * update disk list in imsm_disk list table in anchor
11246 */
11247 dprintf("imsm: %i spares are available.\n\n",
11248 spares->array.spare_disks);
11249
bbd24d86 11250 dev = spares->devs;
78b10e66 11251 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11252 struct dl *dl;
11253
bbd24d86
AK		 11254 if (dev == NULL)
11255 break;
78b10e66
N		 11256 u->new_disks[i] = makedev(dev->disk.major,
11257 dev->disk.minor);
11258 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11259 dl->index = mpb->num_disks;
11260 mpb->num_disks++;
bbd24d86 11261 dev = dev->next;
78b10e66 11262 }
78b10e66
N		 11263
11264abort:
11265 /* free spares
11266 */
11267 sysfs_free(spares);
11268
d677e0b8 11269 dprintf("imsm: reshape update preparation :");
78b10e66 11270 if (i == delta_disks) {
1ade5cc1 11271 dprintf_cont(" OK\n");
78b10e66
N		 11272 *updatep = u;
11273 return update_memory_size;
11274 }
11275 free(u);
1ade5cc1 11276 dprintf_cont(" Error\n");
78b10e66
N		 11277
11278 return 0;
11279}
11280
f3871fdc
AK		 11281/******************************************************************************
11282 * function: imsm_create_metadata_update_for_size_change()
11283 * Creates update for IMSM array for array size change.
11284 *
11285 ******************************************************************************/
11286static int imsm_create_metadata_update_for_size_change(
11287 struct supertype *st,
11288 struct geo_params *geo,
11289 struct imsm_update_size_change **updatep)
11290{
11291 struct intel_super *super = st->sb;
594dc1b8
JS		 11292 int update_memory_size;
11293 struct imsm_update_size_change *u;
f3871fdc 11294
1ade5cc1 11295 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11296
11297 /* size of all update data without anchor */
11298 update_memory_size = sizeof(struct imsm_update_size_change);
11299
503975b9 11300 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11301 u->type = update_size_change;
11302 u->subdev = super->current_vol;
11303 u->new_size = geo->size;
11304
11305 dprintf("imsm: reshape update preparation : OK\n");
11306 *updatep = u;
11307
11308 return update_memory_size;
11309}
11310
48c5303a
PC		 11311/******************************************************************************
11312 * function: imsm_create_metadata_update_for_migration()
11313 * Creates update for IMSM array.
11314 *
11315 ******************************************************************************/
11316static int imsm_create_metadata_update_for_migration(
11317 struct supertype *st,
11318 struct geo_params *geo,
11319 struct imsm_update_reshape_migration **updatep)
11320{
11321 struct intel_super *super = st->sb;
594dc1b8
JS		 11322 int update_memory_size;
11323 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11324 struct imsm_dev *dev;
11325 int previous_level = -1;
11326
1ade5cc1 11327 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11328
11329 /* size of all update data without anchor */
11330 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11331
503975b9 11332 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11333 u->type = update_reshape_migration;
11334 u->subdev = super->current_vol;
11335 u->new_level = geo->level;
11336 u->new_layout = geo->layout;
11337 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11338 u->new_disks[0] = -1;
4bba0439 11339 u->new_chunksize = -1;
48c5303a
PC		 11340
11341 dev = get_imsm_dev(super, u->subdev);
11342 if (dev) {
11343 struct imsm_map *map;
11344
238c0a71 11345 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11346 if (map) {
11347 int current_chunk_size =
11348 __le16_to_cpu(map->blocks_per_strip) / 2;
11349
11350 if (geo->chunksize != current_chunk_size) {
11351 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11352 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11353 current_chunk_size, u->new_chunksize);
11354 }
48c5303a 11355 previous_level = map->raid_level;
4bba0439 11356 }
48c5303a 11357 }
089f9d79 11358 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11359 struct mdinfo *spares = NULL;
11360
11361 u->new_raid_disks++;
11362 spares = get_spares_for_grow(st);
089f9d79 11363 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11364 free(u);
11365 sysfs_free(spares);
11366 update_memory_size = 0;
565cc99e 11367 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11368 return 0;
11369 }
11370 sysfs_free(spares);
11371 }
11372 dprintf("imsm: reshape update preparation : OK\n");
11373 *updatep = u;
11374
11375 return update_memory_size;
11376}
11377
8dd70bce
AK		 11378static void imsm_update_metadata_locally(struct supertype *st,
11379 void *buf, int len)
11380{
11381 struct metadata_update mu;
11382
11383 mu.buf = buf;
11384 mu.len = len;
11385 mu.space = NULL;
11386 mu.space_list = NULL;
11387 mu.next = NULL;
5fe6f031
N		 11388 if (imsm_prepare_update(st, &mu))
11389 imsm_process_update(st, &mu);
8dd70bce
AK		 11390
11391 while (mu.space_list) {
11392 void **space = mu.space_list;
11393 mu.space_list = *space;
11394 free(space);
11395 }
11396}
78b10e66 11397
471bceb6 11398/***************************************************************************
694575e7 11399* Function: imsm_analyze_change
471bceb6 11400* Description: Function analyze change for single volume
1011e834 11401* and validate if transition is supported
fbf3d202
AK		 11402* Parameters: Geometry parameters, supertype structure,
11403* metadata change direction (apply/rollback)
694575e7 11404* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11405****************************************************************************/
11406enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11407 struct geo_params *geo,
11408 int direction)
694575e7 11409{
471bceb6
KW		 11410 struct mdinfo info;
11411 int change = -1;
11412 int check_devs = 0;
c21e737b 11413 int chunk;
67a2db32
AK		 11414 /* number of added/removed disks in operation result */
11415 int devNumChange = 0;
11416 /* imsm compatible layout value for array geometry verification */
11417 int imsm_layout = -1;
7abc9871
AK		 11418 int data_disks;
11419 struct imsm_dev *dev;
9529d343 11420 struct imsm_map *map;
7abc9871 11421 struct intel_super *super;
d04f65f4 11422 unsigned long long current_size;
65d38cca 11423 unsigned long long free_size;
d04f65f4 11424 unsigned long long max_size;
65d38cca 11425 int rv;
471bceb6
KW		 11426
11427 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11428 if (geo->level != info.array.level && geo->level >= 0 &&
11429 geo->level != UnSet) {
471bceb6
KW		 11430 switch (info.array.level) {
11431 case 0:
11432 if (geo->level == 5) {
b5347799 11433 change = CH_MIGRATION;
e13ce846 11434 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11435 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11436 change = -1;
11437 goto analyse_change_exit;
11438 }
67a2db32 11439 imsm_layout = geo->layout;
471bceb6 11440 check_devs = 1;
e91a3bad
LM		 11441 devNumChange = 1; /* parity disk added */
11442 } else if (geo->level == 10) {
471bceb6
KW		 11443 change = CH_TAKEOVER;
11444 check_devs = 1;
e91a3bad 11445 devNumChange = 2; /* two mirrors added */
67a2db32 11446 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11447 }
dfe77a9e
KW		 11448 break;
11449 case 1:
471bceb6
KW		 11450 case 10:
11451 if (geo->level == 0) {
11452 change = CH_TAKEOVER;
11453 check_devs = 1;
e91a3bad 11454 devNumChange = -(geo->raid_disks/2);
67a2db32 11455 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11456 }
11457 break;
11458 }
11459 if (change == -1) {
7a862a02 11460 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11461 info.array.level, geo->level);
471bceb6
KW		 11462 goto analyse_change_exit;
11463 }
11464 } else
11465 geo->level = info.array.level;
11466
089f9d79
JS		 11467 if (geo->layout != info.array.layout &&
11468 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11469 change = CH_MIGRATION;
089f9d79
JS		 11470 if (info.array.layout == 0 && info.array.level == 5 &&
11471 geo->layout == 5) {
471bceb6 11472 /* reshape 5 -> 4 */
089f9d79
JS		 11473 } else if (info.array.layout == 5 && info.array.level == 5 &&
11474 geo->layout == 0) {
471bceb6
KW		 11475 /* reshape 4 -> 5 */
11476 geo->layout = 0;
11477 geo->level = 5;
11478 } else {
7a862a02 11479 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11480 info.array.layout, geo->layout);
471bceb6
KW		 11481 change = -1;
11482 goto analyse_change_exit;
11483 }
67a2db32 11484 } else {
471bceb6 11485 geo->layout = info.array.layout;
67a2db32
AK		 11486 if (imsm_layout == -1)
11487 imsm_layout = info.array.layout;
11488 }
471bceb6 11489
089f9d79
JS		 11490 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11491 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11492 if (info.array.level == 10) {
11493 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11494 change = -1;
11495 goto analyse_change_exit;
1e9b2c3f
PB		 11496 } else if (info.component_size % (geo->chunksize/512)) {
11497 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11498 geo->chunksize/1024, info.component_size/2);
11499 change = -1;
11500 goto analyse_change_exit;
2d2b0eb7 11501 }
b5347799 11502 change = CH_MIGRATION;
2d2b0eb7 11503 } else {
471bceb6 11504 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11505 }
471bceb6 11506
c21e737b 11507 chunk = geo->chunksize / 1024;
7abc9871
AK		 11508
11509 super = st->sb;
11510 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11511 map = get_imsm_map(dev, MAP_0);
11512 data_disks = imsm_num_data_members(map);
c41e00b2 11513 /* compute current size per disk member
7abc9871 11514 */
c41e00b2
AK		 11515 current_size = info.custom_array_size / data_disks;
11516
089f9d79 11517 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11518 /* align component size
11519 */
3e684231 11520 geo->size = imsm_component_size_alignment_check(
c41e00b2 11521 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11522 chunk * 1024, super->sector_size,
c41e00b2 11523 geo->size * 2);
65d0b4ce 11524 if (geo->size == 0) {
7a862a02 11525 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11526 current_size);
11527 goto analyse_change_exit;
11528 }
c41e00b2 11529 }
7abc9871 11530
089f9d79 11531 if (current_size != geo->size && geo->size > 0) {
7abc9871 11532 if (change != -1) {
7a862a02 11533 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11534 change = -1;
11535 goto analyse_change_exit;
11536 }
11537 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11538 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11539 super->current_vol, st->devnm);
7abc9871
AK		 11540 goto analyse_change_exit;
11541 }
65d38cca
LD		 11542 /* check the maximum available size
11543 */
11544 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11545 0, chunk, &free_size);
11546 if (rv == 0)
11547 /* Cannot find maximum available space
11548 */
11549 max_size = 0;
11550 else {
11551 max_size = free_size + current_size;
11552 /* align component size
11553 */
3e684231 11554 max_size = imsm_component_size_alignment_check(
65d38cca 11555 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11556 chunk * 1024, super->sector_size,
65d38cca
LD		 11557 max_size);
11558 }
d04f65f4 11559 if (geo->size == MAX_SIZE) {
b130333f
AK		 11560 /* requested size change to the maximum available size
11561 */
65d38cca 11562 if (max_size == 0) {
7a862a02 11563 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11564 change = -1;
11565 goto analyse_change_exit;
65d38cca
LD		 11566 } else
11567 geo->size = max_size;
c41e00b2 11568 }
b130333f 11569
681b7ae2 11570 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11571 /* accept size for rollback only
11572 */
11573 } else {
11574 /* round size due to metadata compatibility
11575 */
11576 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11577 << SECT_PER_MB_SHIFT;
11578 dprintf("Prepare update for size change to %llu\n",
11579 geo->size );
11580 if (current_size >= geo->size) {
7a862a02 11581 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11582 current_size, geo->size);
fbf3d202
AK		 11583 goto analyse_change_exit;
11584 }
65d38cca 11585 if (max_size && geo->size > max_size) {
7a862a02 11586 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11587 max_size, geo->size);
65d38cca
LD		 11588 goto analyse_change_exit;
11589 }
7abc9871
AK		 11590 }
11591 geo->size *= data_disks;
11592 geo->raid_disks = dev->vol.map->num_members;
11593 change = CH_ARRAY_SIZE;
11594 }
471bceb6
KW		 11595 if (!validate_geometry_imsm(st,
11596 geo->level,
67a2db32 11597 imsm_layout,
e91a3bad 11598 geo->raid_disks + devNumChange,
c21e737b 11599 &chunk,
af4348dd 11600 geo->size, INVALID_SECTORS,
5308f117 11601 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11602 change = -1;
11603
11604 if (check_devs) {
11605 struct intel_super *super = st->sb;
11606 struct imsm_super *mpb = super->anchor;
11607
11608 if (mpb->num_raid_devs > 1) {
7a862a02 11609 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11610 geo->dev_name);
471bceb6
KW		 11611 change = -1;
11612 }
11613 }
11614
11615analyse_change_exit:
089f9d79
JS		 11616 if (direction == ROLLBACK_METADATA_CHANGES &&
11617 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11618 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11619 change = -1;
11620 }
471bceb6 11621 return change;
694575e7
KW		 11622}
11623
bb025c2f
KW		 11624int imsm_takeover(struct supertype *st, struct geo_params *geo)
11625{
11626 struct intel_super *super = st->sb;
11627 struct imsm_update_takeover *u;
11628
503975b9 11629 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11630
11631 u->type = update_takeover;
11632 u->subarray = super->current_vol;
11633
11634 /* 10->0 transition */
11635 if (geo->level == 0)
11636 u->direction = R10_TO_R0;
11637
0529c688
KW		 11638 /* 0->10 transition */
11639 if (geo->level == 10)
11640 u->direction = R0_TO_R10;
11641
bb025c2f
KW		 11642 /* update metadata locally */
11643 imsm_update_metadata_locally(st, u,
11644 sizeof(struct imsm_update_takeover));
11645 /* and possibly remotely */
11646 if (st->update_tail)
11647 append_metadata_update(st, u,
11648 sizeof(struct imsm_update_takeover));
11649 else
11650 free(u);
11651
11652 return 0;
11653}
11654
d04f65f4
N		 11655static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11656 int level,
78b10e66 11657 int layout, int chunksize, int raid_disks,
41784c88 11658 int delta_disks, char *backup, char *dev,
016e00f5 11659 int direction, int verbose)
78b10e66 11660{
78b10e66
N		 11661 int ret_val = 1;
11662 struct geo_params geo;
11663
1ade5cc1 11664 dprintf("(enter)\n");
78b10e66 11665
71204a50 11666 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11667
11668 geo.dev_name = dev;
4dd2df09 11669 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11670 geo.size = size;
11671 geo.level = level;
11672 geo.layout = layout;
11673 geo.chunksize = chunksize;
11674 geo.raid_disks = raid_disks;
41784c88
AK		 11675 if (delta_disks != UnSet)
11676 geo.raid_disks += delta_disks;
78b10e66 11677
1ade5cc1
N		 11678 dprintf("for level : %i\n", geo.level);
11679 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11680
4dd2df09 11681 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11682 /* On container level we can only increase number of devices. */
11683 dprintf("imsm: info: Container operation\n");
78b10e66 11684 int old_raid_disks = 0;
6dc0be30 11685
78b10e66 11686 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11687 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11688 struct imsm_update_reshape *u = NULL;
11689 int len;
11690
11691 len = imsm_create_metadata_update_for_reshape(
11692 st, &geo, old_raid_disks, &u);
11693
ed08d51c
AK		 11694 if (len <= 0) {
11695 dprintf("imsm: Cannot prepare update\n");
11696 goto exit_imsm_reshape_super;
11697 }
11698
8dd70bce
AK		 11699 ret_val = 0;
11700 /* update metadata locally */
11701 imsm_update_metadata_locally(st, u, len);
11702 /* and possibly remotely */
11703 if (st->update_tail)
11704 append_metadata_update(st, u, len);
11705 else
ed08d51c 11706 free(u);
8dd70bce 11707
694575e7 11708 } else {
7a862a02 11709 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11710 }
11711 } else {
11712 /* On volume level we support following operations
471bceb6
KW		 11713 * - takeover: raid10 -> raid0; raid0 -> raid10
11714 * - chunk size migration
11715 * - migration: raid5 -> raid0; raid0 -> raid5
11716 */
11717 struct intel_super *super = st->sb;
11718 struct intel_dev *dev = super->devlist;
4dd2df09 11719 int change;
694575e7 11720 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11721 /* find requested device */
11722 while (dev) {
1011e834 11723 char *devnm =
4dd2df09
N		 11724 imsm_find_array_devnm_by_subdev(
11725 dev->index, st->container_devnm);
11726 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11727 break;
11728 dev = dev->next;
11729 }
11730 if (dev == NULL) {
4dd2df09
N		 11731 pr_err("Cannot find %s (%s) subarray\n",
11732 geo.dev_name, geo.devnm);
471bceb6
KW		 11733 goto exit_imsm_reshape_super;
11734 }
11735 super->current_vol = dev->index;
fbf3d202 11736 change = imsm_analyze_change(st, &geo, direction);
694575e7 11737 switch (change) {
471bceb6 11738 case CH_TAKEOVER:
bb025c2f 11739 ret_val = imsm_takeover(st, &geo);
694575e7 11740 break;
48c5303a
PC		 11741 case CH_MIGRATION: {
11742 struct imsm_update_reshape_migration *u = NULL;
11743 int len =
11744 imsm_create_metadata_update_for_migration(
11745 st, &geo, &u);
11746 if (len < 1) {
7a862a02 11747 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11748 break;
11749 }
471bceb6 11750 ret_val = 0;
48c5303a
PC		 11751 /* update metadata locally */
11752 imsm_update_metadata_locally(st, u, len);
11753 /* and possibly remotely */
11754 if (st->update_tail)
11755 append_metadata_update(st, u, len);
11756 else
11757 free(u);
11758 }
11759 break;
7abc9871 11760 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11761 struct imsm_update_size_change *u = NULL;
11762 int len =
11763 imsm_create_metadata_update_for_size_change(
11764 st, &geo, &u);
11765 if (len < 1) {
7a862a02 11766 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11767 break;
11768 }
11769 ret_val = 0;
11770 /* update metadata locally */
11771 imsm_update_metadata_locally(st, u, len);
11772 /* and possibly remotely */
11773 if (st->update_tail)
11774 append_metadata_update(st, u, len);
11775 else
11776 free(u);
7abc9871
AK		 11777 }
11778 break;
471bceb6
KW		 11779 default:
11780 ret_val = 1;
694575e7 11781 }
694575e7 11782 }
78b10e66 11783
ed08d51c 11784exit_imsm_reshape_super:
78b10e66
N		 11785 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11786 return ret_val;
11787}
2cda7640 11788
0febb20c
AO		 11789#define COMPLETED_OK 0
11790#define COMPLETED_NONE 1
11791#define COMPLETED_DELAYED 2
11792
11793static int read_completed(int fd, unsigned long long *val)
11794{
11795 int ret;
11796 char buf[50];
11797
11798 ret = sysfs_fd_get_str(fd, buf, 50);
11799 if (ret < 0)
11800 return ret;
11801
11802 ret = COMPLETED_OK;
11803 if (strncmp(buf, "none", 4) == 0) {
11804 ret = COMPLETED_NONE;
11805 } else if (strncmp(buf, "delayed", 7) == 0) {
11806 ret = COMPLETED_DELAYED;
11807 } else {
11808 char *ep;
11809 *val = strtoull(buf, &ep, 0);
11810 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11811 ret = -1;
11812 }
11813 return ret;
11814}
11815
eee67a47
AK		 11816/*******************************************************************************
11817 * Function: wait_for_reshape_imsm
11818 * Description: Function writes new sync_max value and waits until
11819 * reshape process reach new position
11820 * Parameters:
11821 * sra : general array info
eee67a47
AK		 11822 * ndata : number of disks in new array's layout
11823 * Returns:
11824 * 0 : success,
11825 * 1 : there is no reshape in progress,
11826 * -1 : fail
11827 ******************************************************************************/
ae9f01f8 11828int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11829{
85ca499c 11830 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11831 int retry = 3;
eee67a47 11832 unsigned long long completed;
ae9f01f8
AK		 11833 /* to_complete : new sync_max position */
11834 unsigned long long to_complete = sra->reshape_progress;
11835 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11836
ae9f01f8 11837 if (fd < 0) {
1ade5cc1 11838 dprintf("cannot open reshape_position\n");
eee67a47 11839 return 1;
ae9f01f8 11840 }
eee67a47 11841
df2647fa
PB		 11842 do {
11843 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11844 if (!retry) {
11845 dprintf("cannot read reshape_position (no reshape in progres)\n");
11846 close(fd);
11847 return 1;
11848 }
11849 usleep(30000);
11850 } else
11851 break;
11852 } while (retry--);
eee67a47 11853
85ca499c 11854 if (completed > position_to_set) {
1ade5cc1 11855 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11856 to_complete, position_to_set);
ae9f01f8
AK		 11857 close(fd);
11858 return -1;
11859 }
11860 dprintf("Position set: %llu\n", position_to_set);
11861 if (sysfs_set_num(sra, NULL, "sync_max",
11862 position_to_set) != 0) {
1ade5cc1 11863 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11864 position_to_set);
11865 close(fd);
11866 return -1;
eee67a47
AK		 11867 }
11868
eee67a47 11869 do {
0febb20c 11870 int rc;
eee67a47 11871 char action[20];
5ff3a780 11872 int timeout = 3000;
0febb20c 11873
5ff3a780 11874 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11875 if (sysfs_get_str(sra, NULL, "sync_action",
11876 action, 20) > 0 &&
d7d3809a 11877 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11878 if (strncmp(action, "idle", 4) == 0)
11879 break;
d7d3809a
AP		 11880 close(fd);
11881 return -1;
11882 }
0febb20c
AO		 11883
11884 rc = read_completed(fd, &completed);
11885 if (rc < 0) {
1ade5cc1 11886 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11887 close(fd);
11888 return 1;
0febb20c
AO		 11889 } else if (rc == COMPLETED_NONE)
11890 break;
85ca499c 11891 } while (completed < position_to_set);
b2be2b62 11892
eee67a47
AK		 11893 close(fd);
11894 return 0;
eee67a47
AK		 11895}
11896
b915c95f
AK		 11897/*******************************************************************************
11898 * Function: check_degradation_change
11899 * Description: Check that array hasn't become failed.
11900 * Parameters:
11901 * info : for sysfs access
11902 * sources : source disks descriptors
11903 * degraded: previous degradation level
11904 * Returns:
11905 * degradation level
11906 ******************************************************************************/
11907int check_degradation_change(struct mdinfo *info,
11908 int *sources,
11909 int degraded)
11910{
11911 unsigned long long new_degraded;
e1993023
LD		 11912 int rv;
11913
11914 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11915 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11916 /* check each device to ensure it is still working */
11917 struct mdinfo *sd;
11918 new_degraded = 0;
11919 for (sd = info->devs ; sd ; sd = sd->next) {
11920 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11921 continue;
11922 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11923 char sbuf[100];
11924
b915c95f 11925 if (sysfs_get_str(info,
cf52eff5 11926 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11927 strstr(sbuf, "faulty") ||
11928 strstr(sbuf, "in_sync") == NULL) {
11929 /* this device is dead */
11930 sd->disk.state = (1<<MD_DISK_FAULTY);
11931 if (sd->disk.raid_disk >= 0 &&
11932 sources[sd->disk.raid_disk] >= 0) {
11933 close(sources[
11934 sd->disk.raid_disk]);
11935 sources[sd->disk.raid_disk] =
11936 -1;
11937 }
11938 new_degraded++;
11939 }
11940 }
11941 }
11942 }
11943
11944 return new_degraded;
11945}
11946
10f22854
AK		 11947/*******************************************************************************
11948 * Function: imsm_manage_reshape
11949 * Description: Function finds array under reshape and it manages reshape
11950 * process. It creates stripes backups (if required) and sets
942e1cdb 11951 * checkpoints.
10f22854
AK		 11952 * Parameters:
11953 * afd : Backup handle (nattive) - not used
11954 * sra : general array info
11955 * reshape : reshape parameters - not used
11956 * st : supertype structure
11957 * blocks : size of critical section [blocks]
11958 * fds : table of source device descriptor
11959 * offsets : start of array (offest per devices)
11960 * dests : not used
11961 * destfd : table of destination device descriptor
11962 * destoffsets : table of destination offsets (per device)
11963 * Returns:
11964 * 1 : success, reshape is done
11965 * 0 : fail
11966 ******************************************************************************/
999b4972
N		 11967static int imsm_manage_reshape(
11968 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11969 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11970 int *fds, unsigned long long *offsets,
11971 int dests, int *destfd, unsigned long long *destoffsets)
11972{
10f22854
AK		 11973 int ret_val = 0;
11974 struct intel_super *super = st->sb;
594dc1b8 11975 struct intel_dev *dv;
de44e46f 11976 unsigned int sector_size = super->sector_size;
10f22854 11977 struct imsm_dev *dev = NULL;
9529d343 11978 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11979 int migr_vol_qan = 0;
11980 int ndata, odata; /* [bytes] */
11981 int chunk; /* [bytes] */
11982 struct migr_record *migr_rec;
11983 char *buf = NULL;
11984 unsigned int buf_size; /* [bytes] */
11985 unsigned long long max_position; /* array size [bytes] */
11986 unsigned long long next_step; /* [blocks]/[bytes] */
11987 unsigned long long old_data_stripe_length;
10f22854
AK		 11988 unsigned long long start_src; /* [bytes] */
11989 unsigned long long start; /* [bytes] */
11990 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11991 int degraded = 0;
ab724b98 11992 int source_layout = 0;
10f22854 11993
79a16a9b
JS		 11994 if (!sra)
11995 return ret_val;
11996
11997 if (!fds || !offsets)
10f22854
AK		 11998 goto abort;
11999
12000 /* Find volume during the reshape */
12001 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12002 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12003 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12004 dev = dv->dev;
12005 migr_vol_qan++;
12006 }
12007 }
12008 /* Only one volume can migrate at the same time */
12009 if (migr_vol_qan != 1) {
676e87a8 12010 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12011 "Number of migrating volumes greater than 1\n" :
12012 "There is no volume during migrationg\n");
12013 goto abort;
12014 }
12015
9529d343 12016 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12017 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12018 if (map_src == NULL)
12019 goto abort;
10f22854 12020
9529d343
MD		 12021 ndata = imsm_num_data_members(map_dest);
12022 odata = imsm_num_data_members(map_src);
10f22854 12023
7b1ab482 12024 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12025 old_data_stripe_length = odata * chunk;
12026
12027 migr_rec = super->migr_rec;
12028
10f22854
AK		 12029 /* initialize migration record for start condition */
12030 if (sra->reshape_progress == 0)
12031 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12032 else {
12033 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12034 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12035 goto abort;
12036 }
6a75c8ca
AK		 12037 /* Save checkpoint to update migration record for current
12038 * reshape position (in md). It can be farther than current
12039 * reshape position in metadata.
12040 */
12041 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12042 /* ignore error == 2, this can mean end of reshape here
12043 */
7a862a02 12044 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12045 goto abort;
12046 }
b2c59438 12047 }
10f22854
AK		 12048
12049 /* size for data */
12050 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12051 /* extend buffer size for parity disk */
12052 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12053 /* add space for stripe alignment */
10f22854 12054 buf_size += old_data_stripe_length;
de44e46f
PB		 12055 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12056 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12057 goto abort;
12058 }
12059
3ef4403c 12060 max_position = sra->component_size * ndata;
68eb8bc6 12061 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12062
9f421827
PB		 12063 while (current_migr_unit(migr_rec) <
12064 get_num_migr_units(migr_rec)) {
10f22854
AK		 12065 /* current reshape position [blocks] */
12066 unsigned long long current_position =
12067 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12068 * current_migr_unit(migr_rec);
10f22854
AK		 12069 unsigned long long border;
12070
b915c95f
AK		 12071 /* Check that array hasn't become failed.
12072 */
12073 degraded = check_degradation_change(sra, fds, degraded);
12074 if (degraded > 1) {
7a862a02 12075 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12076 goto abort;
12077 }
12078
10f22854
AK		 12079 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12080
12081 if ((current_position + next_step) > max_position)
12082 next_step = max_position - current_position;
12083
92144abf 12084 start = current_position * 512;
10f22854 12085
942e1cdb 12086 /* align reading start to old geometry */
10f22854
AK		 12087 start_buf_shift = start % old_data_stripe_length;
12088 start_src = start - start_buf_shift;
12089
12090 border = (start_src / odata) - (start / ndata);
12091 border /= 512;
12092 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12093 /* save critical stripes to buf
12094 * start - start address of current unit
12095 * to backup [bytes]
12096 * start_src - start address of current unit
12097 * to backup alligned to source array
12098 * [bytes]
12099 */
594dc1b8 12100 unsigned long long next_step_filler;
10f22854
AK		 12101 unsigned long long copy_length = next_step * 512;
12102
12103 /* allign copy area length to stripe in old geometry */
12104 next_step_filler = ((copy_length + start_buf_shift)
12105 % old_data_stripe_length);
12106 if (next_step_filler)
12107 next_step_filler = (old_data_stripe_length
12108 - next_step_filler);
7a862a02 12109 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		 12110 start, start_src, copy_length,
12111 start_buf_shift, next_step_filler);
12112
12113 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12114 chunk, map_src->raid_level,
12115 source_layout, 0, NULL, start_src,
10f22854
AK		 12116 copy_length +
12117 next_step_filler + start_buf_shift,
12118 buf)) {
7a862a02 12119 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12120 goto abort;
12121 }
12122 /* Convert data to destination format and store it
12123 * in backup general migration area
12124 */
12125 if (save_backup_imsm(st, dev, sra,
aea93171 12126 buf + start_buf_shift, copy_length)) {
7a862a02 12127 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12128 goto abort;
12129 }
12130 if (save_checkpoint_imsm(st, sra,
12131 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12132 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12133 goto abort;
12134 }
8016a6d4
AK		 12135 } else {
12136 /* set next step to use whole border area */
12137 border /= next_step;
12138 if (border > 1)
12139 next_step *= border;
10f22854
AK		 12140 }
12141 /* When data backed up, checkpoint stored,
12142 * kick the kernel to reshape unit of data
12143 */
12144 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12145 /* limit next step to array max position */
12146 if (next_step > max_position)
12147 next_step = max_position;
10f22854
AK		 12148 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12149 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12150 sra->reshape_progress = next_step;
10f22854
AK		 12151
12152 /* wait until reshape finish */
c85338c6 12153 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12154 dprintf("wait_for_reshape_imsm returned error!\n");
12155 goto abort;
12156 }
84d11e6c
N		 12157 if (sigterm)
12158 goto abort;
10f22854 12159
0228d92c
AK		 12160 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12161 /* ignore error == 2, this can mean end of reshape here
12162 */
7a862a02 12163 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12164 goto abort;
12165 }
12166
12167 }
12168
71e5411e
PB		 12169 /* clear migr_rec on disks after successful migration */
12170 struct dl *d;
12171
85337573 12172 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12173 for (d = super->disks; d; d = d->next) {
12174 if (d->index < 0 || is_failed(&d->disk))
12175 continue;
12176 unsigned long long dsize;
12177
12178 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12179 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12180 SEEK_SET) >= 0) {
466070ad 12181 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12182 MIGR_REC_BUF_SECTORS*sector_size) !=
12183 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12184 perror("Write migr_rec failed");
12185 }
12186 }
12187
10f22854
AK		 12188 /* return '1' if done */
12189 ret_val = 1;
12190abort:
12191 free(buf);
942e1cdb
N		 12192 /* See Grow.c: abort_reshape() for further explanation */
12193 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12194 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12195 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12196
12197 return ret_val;
999b4972 12198}
0c21b485 12199
cdddbdbc 12200struct superswitch super_imsm = {
cdddbdbc
DW		 12201 .examine_super = examine_super_imsm,
12202 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12203 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12204 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12205 .detail_super = detail_super_imsm,
12206 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12207 .write_init_super = write_init_super_imsm,
0e600426
N		 12208 .validate_geometry = validate_geometry_imsm,
12209 .add_to_super = add_to_super_imsm,
1a64be56 12210 .remove_from_super = remove_from_super_imsm,
d665cc31 12211 .detail_platform = detail_platform_imsm,
e50cf220 12212 .export_detail_platform = export_detail_platform_imsm,
33414a01 12213 .kill_subarray = kill_subarray_imsm,
aa534678 12214 .update_subarray = update_subarray_imsm,
2b959fbf 12215 .load_container = load_container_imsm,
71204a50
N		 12216 .default_geometry = default_geometry_imsm,
12217 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12218 .reshape_super = imsm_reshape_super,
12219 .manage_reshape = imsm_manage_reshape,
9e2d750d 12220 .recover_backup = recover_backup_imsm,
27156a57 12221 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12222 .match_home = match_home_imsm,
12223 .uuid_from_super= uuid_from_super_imsm,
12224 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12225 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12226 .update_super = update_super_imsm,
12227
12228 .avail_size = avail_size_imsm,
fbfdcb06 12229 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12230
12231 .compare_super = compare_super_imsm,
12232
12233 .load_super = load_super_imsm,
bf5a934a 12234 .init_super = init_super_imsm,
e683ca88 12235 .store_super = store_super_imsm,
cdddbdbc
DW		 12236 .free_super = free_super_imsm,
12237 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12238 .container_content = container_content_imsm,
0c21b485 12239 .validate_container = validate_container_imsm,
cdddbdbc 12240
2432ce9b
AP		 12241 .write_init_ppl = write_init_ppl_imsm,
12242 .validate_ppl = validate_ppl_imsm,
12243
cdddbdbc 12244 .external = 1,
4cce4069 12245 .name = "imsm",
845dea95
NB		 12246
12247/* for mdmon */
12248 .open_new = imsm_open_new,
ed9d66aa 12249 .set_array_state= imsm_set_array_state,
845dea95
NB		 12250 .set_disk = imsm_set_disk,
12251 .sync_metadata = imsm_sync_metadata,
88758e9d 12252 .activate_spare = imsm_activate_spare,
e8319a19 12253 .process_update = imsm_process_update,
8273f55e 12254 .prepare_update = imsm_prepare_update,
6f50473f 12255 .record_bad_block = imsm_record_badblock,
c07a5a4f 12256 .clear_bad_block = imsm_clear_badblock,
928f1424 12257 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12258};
Unnamed repository; edit this file 'description' to name the repository.