]> 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: fill working_disks according to metadata.
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
cdddbdbc
DW		 1/*
2 * mdadm - Intel(R) Matrix Storage Manager Support
3 *
a54d5262 4 * Copyright (C) 2002-2008 Intel Corporation
cdddbdbc
DW		 5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 */
19
51006d85 20#define HAVE_STDINT_H 1
cdddbdbc 21#include "mdadm.h"
c2a1e7da 22#include "mdmon.h"
51006d85 23#include "sha1.h"
88c32bb1 24#include "platform-intel.h"
cdddbdbc
DW		 25#include <values.h>
26#include <scsi/sg.h>
27#include <ctype.h>
d665cc31 28#include <dirent.h>
cdddbdbc
DW		 29
30/* MPB == Metadata Parameter Block */
31#define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
32#define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
33#define MPB_VERSION_RAID0 "1.0.00"
34#define MPB_VERSION_RAID1 "1.1.00"
fe7ed8cb
DW		 35#define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00"
36#define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01"
cdddbdbc 37#define MPB_VERSION_RAID5 "1.2.02"
fe7ed8cb
DW		 38#define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04"
39#define MPB_VERSION_CNG "1.2.06"
40#define MPB_VERSION_ATTRIBS "1.3.00"
cdddbdbc
DW		 41#define MAX_SIGNATURE_LENGTH 32
42#define MAX_RAID_SERIAL_LEN 16
fe7ed8cb 43
19482bcc
AK		 44/* supports RAID0 */
45#define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001)
46/* supports RAID1 */
47#define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002)
48/* supports RAID10 */
49#define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004)
50/* supports RAID1E */
51#define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008)
52/* supports RAID5 */
53#define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010)
54/* supports RAID CNG */
55#define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020)
56/* supports expanded stripe sizes of 256K, 512K and 1MB */
57#define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040)
58
59/* The OROM Support RST Caching of Volumes */
60#define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000)
61/* The OROM supports creating disks greater than 2TB */
62#define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000)
63/* The OROM supports Bad Block Management */
64#define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000)
65
66/* THe OROM Supports NVM Caching of Volumes */
67#define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000)
68/* The OROM supports creating volumes greater than 2TB */
69#define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000)
70/* originally for PMP, now it's wasted b/c. Never use this bit! */
71#define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000)
72/* Verify MPB contents against checksum after reading MPB */
73#define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000)
74
75/* Define all supported attributes that have to be accepted by mdadm
76 */
418f9b36 77#define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \
19482bcc
AK		 78 MPB_ATTRIB_2TB | \
79 MPB_ATTRIB_2TB_DISK | \
80 MPB_ATTRIB_RAID0 | \
81 MPB_ATTRIB_RAID1 | \
82 MPB_ATTRIB_RAID10 | \
83 MPB_ATTRIB_RAID5 | \
bbab0940
TM		 84 MPB_ATTRIB_EXP_STRIPE_SIZE | \
85 MPB_ATTRIB_BBM)
418f9b36
N		 86
87/* Define attributes that are unused but not harmful */
88#define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE)
fe7ed8cb 89
8e59f3d8 90#define MPB_SECTOR_CNT 2210
611d9529
MD		 91#define IMSM_RESERVED_SECTORS 8192
92#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048
979d38be 93#define SECT_PER_MB_SHIFT 11
f36a9ecd 94#define MAX_SECTOR_SIZE 4096
c2462068
PB		 95#define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole
96 * mutliple PPL area
97 */
cdddbdbc 98
761e3bd9
N		 99/*
100 * This macro let's us ensure that no-one accidentally
101 * changes the size of a struct
102 */
103#define ASSERT_SIZE(_struct, size) \
104static inline void __assert_size_##_struct(void) \
105{ \
106 switch (0) { \
107 case 0: break; \
108 case (sizeof(struct _struct) == size): break; \
109 } \
110}
111
cdddbdbc
DW		 112/* Disk configuration info. */
113#define IMSM_MAX_DEVICES 255
114struct imsm_disk {
115 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 116 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 117 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 118#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
119#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
120#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 121#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 122 __u32 status; /* 0xF0 - 0xF3 */
1011e834 123 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 124 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
125#define IMSM_DISK_FILLERS 3
126 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 127};
761e3bd9 128ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 129
3b451610
AK		 130/* map selector for map managment
131 */
238c0a71
AK		 132#define MAP_0 0
133#define MAP_1 1
134#define MAP_X -1
3b451610 135
cdddbdbc
DW		 136/* RAID map configuration infos. */
137struct imsm_map {
5551b113
CA		 138 __u32 pba_of_lba0_lo; /* start address of partition */
139 __u32 blocks_per_member_lo;/* blocks per member */
140 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 141 __u16 blocks_per_strip;
142 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
143#define IMSM_T_STATE_NORMAL 0
144#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 145#define IMSM_T_STATE_DEGRADED 2
146#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 147 __u8 raid_level;
148#define IMSM_T_RAID0 0
149#define IMSM_T_RAID1 1
150#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
151 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 152 __u8 num_domains; /* number of parity domains */
153 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 154 __u8 ddf;
5551b113
CA		 155 __u32 pba_of_lba0_hi;
156 __u32 blocks_per_member_hi;
157 __u32 num_data_stripes_hi;
158 __u32 filler[4]; /* expansion area */
7eef0453 159#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 160 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 161 * top byte contains some flags
162 */
761e3bd9
N		 163};
164ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 165
166struct imsm_vol {
f8f603f1 167 __u32 curr_migr_unit;
fe7ed8cb 168 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 169 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 170#define MIGR_INIT 0
171#define MIGR_REBUILD 1
172#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
173#define MIGR_GEN_MIGR 3
174#define MIGR_STATE_CHANGE 4
1484e727 175#define MIGR_REPAIR 5
cdddbdbc 176 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 177#define RAIDVOL_CLEAN 0
178#define RAIDVOL_DIRTY 1
179#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 180 __u8 dirty;
fe7ed8cb
DW		 181 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
182 __u16 verify_errors; /* number of mismatches */
183 __u16 bad_blocks; /* number of bad blocks during verify */
184 __u32 filler[4];
cdddbdbc
DW		 185 struct imsm_map map[1];
186 /* here comes another one if migr_state */
761e3bd9
N		 187};
188ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 189
190struct imsm_dev {
fe7ed8cb 191 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 192 __u32 size_low;
193 __u32 size_high;
fe7ed8cb
DW		 194#define DEV_BOOTABLE __cpu_to_le32(0x01)
195#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
196#define DEV_READ_COALESCING __cpu_to_le32(0x04)
197#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
198#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
199#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
200#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
201#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
202#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
203#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
204#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
205#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
206#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 207 __u32 status; /* Persistent RaidDev status */
208 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 209 __u8 migr_priority;
210 __u8 num_sub_vols;
211 __u8 tid;
212 __u8 cng_master_disk;
213 __u16 cache_policy;
214 __u8 cng_state;
215 __u8 cng_sub_state;
2432ce9b
AP		 216 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
217
218 /* NVM_EN */
219 __u8 nv_cache_mode;
220 __u8 nv_cache_flags;
221
222 /* Unique Volume Id of the NvCache Volume associated with this volume */
223 __u32 nvc_vol_orig_family_num;
224 __u16 nvc_vol_raid_dev_num;
225
226#define RWH_OFF 0
227#define RWH_DISTRIBUTED 1
228#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 229#define RWH_MULTIPLE_DISTRIBUTED 3
230#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
231#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 232 __u8 rwh_policy; /* Raid Write Hole Policy */
233 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
234 __u8 filler1;
235
236#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 237 __u32 filler[IMSM_DEV_FILLERS];
238 struct imsm_vol vol;
761e3bd9
N		 239};
240ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 241
242struct imsm_super {
243 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
244 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
245 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
246 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
247 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 248 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
249 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 250 __u8 num_disks; /* 0x38 Number of configured disks */
251 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 252 __u8 error_log_pos; /* 0x3A */
253 __u8 fill[1]; /* 0x3B */
254 __u32 cache_size; /* 0x3c - 0x40 in mb */
255 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
256 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
257 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 258 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
259 * volume IDs for raid_dev created in this array
260 * (starts at 1)
261 */
262 __u16 filler1; /* 0x4E - 0x4F */
263#define IMSM_FILLERS 34
264 __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 265 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
266 /* here comes imsm_dev[num_raid_devs] */
604b746f 267 /* here comes BBM logs */
761e3bd9
N		 268};
269ASSERT_SIZE(imsm_super, 264)
cdddbdbc 270
604b746f 271#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 272#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
273#define BBM_LOG_SIGNATURE 0xabadb10c
274
275struct bbm_log_block_addr {
276 __u16 w1;
277 __u32 dw1;
278} __attribute__ ((__packed__));
604b746f
JD		 279
280struct bbm_log_entry {
8d67477f
TM		 281 __u8 marked_count; /* Number of blocks marked - 1 */
282 __u8 disk_ordinal; /* Disk entry within the imsm_super */
283 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 284} __attribute__ ((__packed__));
285
286struct bbm_log {
287 __u32 signature; /* 0xABADB10C */
288 __u32 entry_count;
8d67477f 289 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 290};
291ASSERT_SIZE(bbm_log, 2040)
604b746f 292
cdddbdbc 293static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 294
b53bfba6
TM		 295#define BLOCKS_PER_KB (1024/512)
296
8e59f3d8
AK		 297#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
298
299#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
300
de44e46f
PB		 301#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
302#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
303 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 304 */
305
8e59f3d8
AK		 306#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
307 * be recovered using srcMap */
308#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
309 * already been migrated and must
310 * be recovered from checkpoint area */
2432ce9b 311
c2462068 312#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 313
8e59f3d8
AK		 314struct migr_record {
315 __u32 rec_status; /* Status used to determine how to restart
316 * migration in case it aborts
317 * in some fashion */
9f421827 318 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 319 __u32 family_num; /* Family number of MPB
320 * containing the RaidDev
321 * that is migrating */
322 __u32 ascending_migr; /* True if migrating in increasing
323 * order of lbas */
324 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
325 __u32 dest_depth_per_unit; /* Num member blocks each destMap
326 * member disk
327 * advances per unit-of-operation */
9f421827
PB		 328 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
329 __u32 dest_1st_member_lba_lo; /* First member lba on first
330 * stripe of destination */
331 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 332 __u32 post_migr_vol_cap; /* Size of volume after
333 * migration completes */
334 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
335 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
336 * migration ckpt record was read from
337 * (for recovered migrations) */
9f421827
PB		 338 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
339 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
340 * high order 32 bits */
341 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
342 * destination - high order 32 bits */
343 __u32 num_migr_units_hi; /* Total num migration units-of-op
344 * high order 32 bits */
761e3bd9
N		 345};
346ASSERT_SIZE(migr_record, 64)
8e59f3d8 347
ec50f7b6
LM		 348struct md_list {
349 /* usage marker:
350 * 1: load metadata
351 * 2: metadata does not match
352 * 4: already checked
353 */
354 int used;
355 char *devname;
356 int found;
357 int container;
358 dev_t st_rdev;
359 struct md_list *next;
360};
361
e7b84f9d 362#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 363
1484e727
DW		 364static __u8 migr_type(struct imsm_dev *dev)
365{
366 if (dev->vol.migr_type == MIGR_VERIFY &&
367 dev->status & DEV_VERIFY_AND_FIX)
368 return MIGR_REPAIR;
369 else
370 return dev->vol.migr_type;
371}
372
373static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
374{
375 /* for compatibility with older oroms convert MIGR_REPAIR, into
376 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
377 */
378 if (migr_type == MIGR_REPAIR) {
379 dev->vol.migr_type = MIGR_VERIFY;
380 dev->status |= DEV_VERIFY_AND_FIX;
381 } else {
382 dev->vol.migr_type = migr_type;
383 dev->status &= ~DEV_VERIFY_AND_FIX;
384 }
385}
386
f36a9ecd 387static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 388{
f36a9ecd 389 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 390}
cdddbdbc 391
f36a9ecd
PB		 392static unsigned int mpb_sectors(struct imsm_super *mpb,
393 unsigned int sector_size)
87eb16df 394{
f36a9ecd 395 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 396}
397
ba2de7ba
DW		 398struct intel_dev {
399 struct imsm_dev *dev;
400 struct intel_dev *next;
f21e18ca 401 unsigned index;
ba2de7ba
DW		 402};
403
88654014
LM		 404struct intel_hba {
405 enum sys_dev_type type;
406 char *path;
407 char *pci_id;
408 struct intel_hba *next;
409};
410
1a64be56
LM		 411enum action {
412 DISK_REMOVE = 1,
413 DISK_ADD
414};
cdddbdbc
DW		 415/* internal representation of IMSM metadata */
416struct intel_super {
417 union {
949c47a0
DW		 418 void *buf; /* O_DIRECT buffer for reading/writing metadata */
419 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 420 };
8e59f3d8
AK		 421 union {
422 void *migr_rec_buf; /* buffer for I/O operations */
423 struct migr_record *migr_rec; /* migration record */
424 };
51d83f5d
AK		 425 int clean_migration_record_by_mdmon; /* when reshape is switched to next
426 array, it indicates that mdmon is allowed to clean migration
427 record */
949c47a0 428 size_t len; /* size of the 'buf' allocation */
bbab0940 429 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 430 void *next_buf; /* for realloc'ing buf from the manager */
431 size_t next_len;
c2c087e6 432 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 433 int current_vol; /* index of raid device undergoing creation */
5551b113 434 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 435 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 436 struct intel_dev *devlist;
fa7bb6f8 437 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 438 struct dl {
439 struct dl *next;
440 int index;
441 __u8 serial[MAX_RAID_SERIAL_LEN];
442 int major, minor;
443 char *devname;
b9f594fe 444 struct imsm_disk disk;
cdddbdbc 445 int fd;
0dcecb2e
DW		 446 int extent_cnt;
447 struct extent *e; /* for determining freespace @ create */
efb30e7f 448 int raiddisk; /* slot to fill in autolayout */
1a64be56 449 enum action action;
ca0748fa 450 } *disks, *current_disk;
1a64be56
LM		 451 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
452 active */
47ee5a45 453 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 454 struct bbm_log *bbm_log;
88654014 455 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 456 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 457 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 458 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 459};
460
461struct intel_disk {
462 struct imsm_disk disk;
463 #define IMSM_UNKNOWN_OWNER (-1)
464 int owner;
465 struct intel_disk *next;
cdddbdbc
DW		 466};
467
c2c087e6
DW		 468struct extent {
469 unsigned long long start, size;
470};
471
694575e7
KW		 472/* definitions of reshape process types */
473enum imsm_reshape_type {
474 CH_TAKEOVER,
b5347799 475 CH_MIGRATION,
7abc9871 476 CH_ARRAY_SIZE,
694575e7
KW		 477};
478
88758e9d
DW		 479/* definition of messages passed to imsm_process_update */
480enum imsm_update_type {
481 update_activate_spare,
8273f55e 482 update_create_array,
33414a01 483 update_kill_array,
aa534678 484 update_rename_array,
1a64be56 485 update_add_remove_disk,
78b10e66 486 update_reshape_container_disks,
48c5303a 487 update_reshape_migration,
2d40f3a1
AK		 488 update_takeover,
489 update_general_migration_checkpoint,
f3871fdc 490 update_size_change,
bbab0940 491 update_prealloc_badblocks_mem,
e6e9dd3f 492 update_rwh_policy,
88758e9d
DW		 493};
494
495struct imsm_update_activate_spare {
496 enum imsm_update_type type;
d23fe947 497 struct dl *dl;
88758e9d
DW		 498 int slot;
499 int array;
500 struct imsm_update_activate_spare *next;
501};
502
78b10e66 503struct geo_params {
4dd2df09 504 char devnm[32];
78b10e66 505 char *dev_name;
d04f65f4 506 unsigned long long size;
78b10e66
N		 507 int level;
508 int layout;
509 int chunksize;
510 int raid_disks;
511};
512
bb025c2f
KW		 513enum takeover_direction {
514 R10_TO_R0,
515 R0_TO_R10
516};
517struct imsm_update_takeover {
518 enum imsm_update_type type;
519 int subarray;
520 enum takeover_direction direction;
521};
78b10e66
N		 522
523struct imsm_update_reshape {
524 enum imsm_update_type type;
525 int old_raid_disks;
526 int new_raid_disks;
48c5303a
PC		 527
528 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
529};
530
531struct imsm_update_reshape_migration {
532 enum imsm_update_type type;
533 int old_raid_disks;
534 int new_raid_disks;
535 /* fields for array migration changes
536 */
537 int subdev;
538 int new_level;
539 int new_layout;
4bba0439 540 int new_chunksize;
48c5303a 541
d195167d 542 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 543};
544
f3871fdc
AK		 545struct imsm_update_size_change {
546 enum imsm_update_type type;
547 int subdev;
548 long long new_size;
549};
550
2d40f3a1
AK		 551struct imsm_update_general_migration_checkpoint {
552 enum imsm_update_type type;
553 __u32 curr_migr_unit;
554};
555
54c2c1ea
DW		 556struct disk_info {
557 __u8 serial[MAX_RAID_SERIAL_LEN];
558};
559
8273f55e
DW		 560struct imsm_update_create_array {
561 enum imsm_update_type type;
8273f55e 562 int dev_idx;
6a3e913e 563 struct imsm_dev dev;
8273f55e
DW		 564};
565
33414a01
DW		 566struct imsm_update_kill_array {
567 enum imsm_update_type type;
568 int dev_idx;
569};
570
aa534678
DW		 571struct imsm_update_rename_array {
572 enum imsm_update_type type;
573 __u8 name[MAX_RAID_SERIAL_LEN];
574 int dev_idx;
575};
576
1a64be56 577struct imsm_update_add_remove_disk {
43dad3d6
DW		 578 enum imsm_update_type type;
579};
580
bbab0940
TM		 581struct imsm_update_prealloc_bb_mem {
582 enum imsm_update_type type;
583};
584
e6e9dd3f
AP		 585struct imsm_update_rwh_policy {
586 enum imsm_update_type type;
587 int new_policy;
588 int dev_idx;
589};
590
88654014
LM		 591static const char *_sys_dev_type[] = {
592 [SYS_DEV_UNKNOWN] = "Unknown",
593 [SYS_DEV_SAS] = "SAS",
614902f6 594 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 595 [SYS_DEV_NVME] = "NVMe",
596 [SYS_DEV_VMD] = "VMD"
88654014
LM		 597};
598
599const char *get_sys_dev_type(enum sys_dev_type type)
600{
601 if (type >= SYS_DEV_MAX)
602 type = SYS_DEV_UNKNOWN;
603
604 return _sys_dev_type[type];
605}
606
607static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
608{
503975b9
N		 609 struct intel_hba *result = xmalloc(sizeof(*result));
610
611 result->type = device->type;
612 result->path = xstrdup(device->path);
613 result->next = NULL;
614 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
615 result->pci_id++;
616
88654014
LM		 617 return result;
618}
619
620static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
621{
594dc1b8
JS		 622 struct intel_hba *result;
623
88654014
LM		 624 for (result = hba; result; result = result->next) {
625 if (result->type == device->type && strcmp(result->path, device->path) == 0)
626 break;
627 }
628 return result;
629}
630
b4cf4cba 631static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 632{
633 struct intel_hba *hba;
634
635 /* check if disk attached to Intel HBA */
636 hba = find_intel_hba(super->hba, device);
637 if (hba != NULL)
638 return 1;
639 /* Check if HBA is already attached to super */
640 if (super->hba == NULL) {
641 super->hba = alloc_intel_hba(device);
642 return 1;
6b781d33
AP		 643 }
644
645 hba = super->hba;
646 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 647 * Do not support HBA types mixing
6b781d33
AP		 648 */
649 if (device->type != hba->type)
88654014 650 return 2;
6b781d33
AP		 651
652 /* Multiple same type HBAs can be used if they share the same OROM */
653 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
654
655 if (device_orom != super->orom)
656 return 2;
657
658 while (hba->next)
659 hba = hba->next;
660
661 hba->next = alloc_intel_hba(device);
662 return 1;
88654014
LM		 663}
664
665static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
666{
9bc4ae77 667 struct sys_dev *list, *elem;
88654014
LM		 668 char *disk_path;
669
670 if ((list = find_intel_devices()) == NULL)
671 return 0;
672
673 if (fd < 0)
674 disk_path = (char *) devname;
675 else
676 disk_path = diskfd_to_devpath(fd);
677
9bc4ae77 678 if (!disk_path)
88654014 679 return 0;
88654014 680
9bc4ae77
N		 681 for (elem = list; elem; elem = elem->next)
682 if (path_attached_to_hba(disk_path, elem->path))
88654014 683 return elem;
9bc4ae77 684
88654014
LM		 685 if (disk_path != devname)
686 free(disk_path);
88654014
LM		 687
688 return NULL;
689}
690
d424212e
N		 691static int find_intel_hba_capability(int fd, struct intel_super *super,
692 char *devname);
f2f5c343 693
cdddbdbc
DW		 694static struct supertype *match_metadata_desc_imsm(char *arg)
695{
696 struct supertype *st;
697
698 if (strcmp(arg, "imsm") != 0 &&
699 strcmp(arg, "default") != 0
700 )
701 return NULL;
702
503975b9 703 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 704 st->ss = &super_imsm;
705 st->max_devs = IMSM_MAX_DEVICES;
706 st->minor_version = 0;
707 st->sb = NULL;
708 return st;
709}
710
cdddbdbc
DW		 711static __u8 *get_imsm_version(struct imsm_super *mpb)
712{
713 return &mpb->sig[MPB_SIG_LEN];
714}
715
949c47a0
DW		 716/* retrieve a disk directly from the anchor when the anchor is known to be
717 * up-to-date, currently only at load time
718 */
719static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 720{
949c47a0 721 if (index >= mpb->num_disks)
cdddbdbc
DW		 722 return NULL;
723 return &mpb->disk[index];
724}
725
95d07a2c
LM		 726/* retrieve the disk description based on a index of the disk
727 * in the sub-array
728 */
729static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 730{
b9f594fe
DW		 731 struct dl *d;
732
733 for (d = super->disks; d; d = d->next)
734 if (d->index == index)
95d07a2c
LM		 735 return d;
736
737 return NULL;
738}
739/* retrieve a disk from the parsed metadata */
740static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
741{
742 struct dl *dl;
743
744 dl = get_imsm_dl_disk(super, index);
745 if (dl)
746 return &dl->disk;
747
b9f594fe 748 return NULL;
949c47a0
DW		 749}
750
751/* generate a checksum directly from the anchor when the anchor is known to be
752 * up-to-date, currently only at load or write_super after coalescing
753 */
754static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 755{
756 __u32 end = mpb->mpb_size / sizeof(end);
757 __u32 *p = (__u32 *) mpb;
758 __u32 sum = 0;
759
5d500228
N		 760 while (end--) {
761 sum += __le32_to_cpu(*p);
97f734fd
N		 762 p++;
763 }
cdddbdbc 764
5d500228 765 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 766}
767
a965f303
DW		 768static size_t sizeof_imsm_map(struct imsm_map *map)
769{
770 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
771}
772
773struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 774{
5e7b0330
AK		 775 /* A device can have 2 maps if it is in the middle of a migration.
776 * If second_map is:
238c0a71
AK		 777 * MAP_0 - we return the first map
778 * MAP_1 - we return the second map if it exists, else NULL
779 * MAP_X - we return the second map if it exists, else the first
5e7b0330 780 */
a965f303 781 struct imsm_map *map = &dev->vol.map[0];
9535fc47 782 struct imsm_map *map2 = NULL;
a965f303 783
9535fc47
AK		 784 if (dev->vol.migr_state)
785 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 786
9535fc47 787 switch (second_map) {
3b451610 788 case MAP_0:
9535fc47 789 break;
3b451610 790 case MAP_1:
9535fc47
AK		 791 map = map2;
792 break;
238c0a71 793 case MAP_X:
9535fc47
AK		 794 if (map2)
795 map = map2;
796 break;
9535fc47
AK		 797 default:
798 map = NULL;
799 }
800 return map;
5e7b0330 801
a965f303 802}
cdddbdbc 803
3393c6af
DW		 804/* return the size of the device.
805 * migr_state increases the returned size if map[0] were to be duplicated
806 */
807static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 808{
809 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 810 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 811
812 /* migrating means an additional map */
a965f303 813 if (dev->vol.migr_state)
238c0a71 814 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 815 else if (migr_state)
238c0a71 816 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 817
818 return size;
819}
820
54c2c1ea
DW		 821/* retrieve disk serial number list from a metadata update */
822static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
823{
824 void *u = update;
825 struct disk_info *inf;
826
827 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
828 sizeof_imsm_dev(&update->dev, 0);
829
830 return inf;
831}
54c2c1ea 832
949c47a0 833static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 834{
835 int offset;
836 int i;
837 void *_mpb = mpb;
838
949c47a0 839 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 840 return NULL;
841
842 /* devices start after all disks */
843 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
844
845 for (i = 0; i <= index; i++)
846 if (i == index)
847 return _mpb + offset;
848 else
3393c6af 849 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 850
851 return NULL;
852}
853
949c47a0
DW		 854static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
855{
ba2de7ba
DW		 856 struct intel_dev *dv;
857
949c47a0
DW		 858 if (index >= super->anchor->num_raid_devs)
859 return NULL;
ba2de7ba
DW		 860 for (dv = super->devlist; dv; dv = dv->next)
861 if (dv->index == index)
862 return dv->dev;
863 return NULL;
949c47a0
DW		 864}
865
8d67477f
TM		 866static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
867 *addr)
868{
869 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
870 __le16_to_cpu(addr->w1));
871}
872
873static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
874{
875 struct bbm_log_block_addr addr;
876
877 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
878 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
879 return addr;
880}
881
8d67477f
TM		 882/* get size of the bbm log */
883static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
884{
885 if (!log || log->entry_count == 0)
886 return 0;
887
888 return sizeof(log->signature) +
889 sizeof(log->entry_count) +
890 log->entry_count * sizeof(struct bbm_log_entry);
891}
6f50473f
TM		 892
893/* check if bad block is not partially stored in bbm log */
894static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
895 long long sector, const int length, __u32 *pos)
896{
897 __u32 i;
898
899 for (i = *pos; i < log->entry_count; i++) {
900 struct bbm_log_entry *entry = &log->marked_block_entries[i];
901 unsigned long long bb_start;
902 unsigned long long bb_end;
903
904 bb_start = __le48_to_cpu(&entry->defective_block_start);
905 bb_end = bb_start + (entry->marked_count + 1);
906
907 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
908 (bb_end <= sector + length)) {
909 *pos = i;
910 return 1;
911 }
912 }
913 return 0;
914}
915
916/* record new bad block in bbm log */
917static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
918 long long sector, int length)
919{
920 int new_bb = 0;
921 __u32 pos = 0;
922 struct bbm_log_entry *entry = NULL;
923
924 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
925 struct bbm_log_entry *e = &log->marked_block_entries[pos];
926
927 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
928 (__le48_to_cpu(&e->defective_block_start) == sector)) {
929 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
930 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
931 pos = pos + 1;
932 continue;
933 }
934 entry = e;
935 break;
936 }
937
938 if (entry) {
939 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
940 BBM_LOG_MAX_LBA_ENTRY_VAL;
941 entry->defective_block_start = __cpu_to_le48(sector);
942 entry->marked_count = cnt - 1;
943 if (cnt == length)
944 return 1;
945 sector += cnt;
946 length -= cnt;
947 }
948
949 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
950 BBM_LOG_MAX_LBA_ENTRY_VAL;
951 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
952 return 0;
953
954 while (length > 0) {
955 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
956 BBM_LOG_MAX_LBA_ENTRY_VAL;
957 struct bbm_log_entry *entry =
958 &log->marked_block_entries[log->entry_count];
959
960 entry->defective_block_start = __cpu_to_le48(sector);
961 entry->marked_count = cnt - 1;
962 entry->disk_ordinal = idx;
963
964 sector += cnt;
965 length -= cnt;
966
967 log->entry_count++;
968 }
969
970 return new_bb;
971}
c07a5a4f 972
4c9e8c1e
TM		 973/* clear all bad blocks for given disk */
974static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
975{
976 __u32 i = 0;
977
978 while (i < log->entry_count) {
979 struct bbm_log_entry *entries = log->marked_block_entries;
980
981 if (entries[i].disk_ordinal == idx) {
982 if (i < log->entry_count - 1)
983 entries[i] = entries[log->entry_count - 1];
984 log->entry_count--;
985 } else {
986 i++;
987 }
988 }
989}
990
c07a5a4f
TM		 991/* clear given bad block */
992static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
993 long long sector, const int length) {
994 __u32 i = 0;
995
996 while (i < log->entry_count) {
997 struct bbm_log_entry *entries = log->marked_block_entries;
998
999 if ((entries[i].disk_ordinal == idx) &&
1000 (__le48_to_cpu(&entries[i].defective_block_start) ==
1001 sector) && (entries[i].marked_count + 1 == length)) {
1002 if (i < log->entry_count - 1)
1003 entries[i] = entries[log->entry_count - 1];
1004 log->entry_count--;
1005 break;
1006 }
1007 i++;
1008 }
1009
1010 return 1;
1011}
8d67477f
TM		 1012
1013/* allocate and load BBM log from metadata */
1014static int load_bbm_log(struct intel_super *super)
1015{
1016 struct imsm_super *mpb = super->anchor;
1017 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1018
1019 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1020 if (!super->bbm_log)
1021 return 1;
1022
1023 if (bbm_log_size) {
1024 struct bbm_log *log = (void *)mpb +
1025 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1026
1027 __u32 entry_count;
1028
1029 if (bbm_log_size < sizeof(log->signature) +
1030 sizeof(log->entry_count))
1031 return 2;
1032
1033 entry_count = __le32_to_cpu(log->entry_count);
1034 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1035 (entry_count > BBM_LOG_MAX_ENTRIES))
1036 return 3;
1037
1038 if (bbm_log_size !=
1039 sizeof(log->signature) + sizeof(log->entry_count) +
1040 entry_count * sizeof(struct bbm_log_entry))
1041 return 4;
1042
1043 memcpy(super->bbm_log, log, bbm_log_size);
1044 } else {
1045 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1046 super->bbm_log->entry_count = 0;
1047 }
1048
1049 return 0;
1050}
1051
b12796be
TM		 1052/* checks if bad block is within volume boundaries */
1053static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1054 const unsigned long long start_sector,
1055 const unsigned long long size)
1056{
1057 unsigned long long bb_start;
1058 unsigned long long bb_end;
1059
1060 bb_start = __le48_to_cpu(&entry->defective_block_start);
1061 bb_end = bb_start + (entry->marked_count + 1);
1062
1063 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1064 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1065 return 1;
1066
1067 return 0;
1068}
1069
1070/* get list of bad blocks on a drive for a volume */
1071static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1072 const unsigned long long start_sector,
1073 const unsigned long long size,
1074 struct md_bb *bbs)
1075{
1076 __u32 count = 0;
1077 __u32 i;
1078
1079 for (i = 0; i < log->entry_count; i++) {
1080 const struct bbm_log_entry *ent =
1081 &log->marked_block_entries[i];
1082 struct md_bb_entry *bb;
1083
1084 if ((ent->disk_ordinal == idx) &&
1085 is_bad_block_in_volume(ent, start_sector, size)) {
1086
1087 if (!bbs->entries) {
1088 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1089 sizeof(*bb));
1090 if (!bbs->entries)
1091 break;
1092 }
1093
1094 bb = &bbs->entries[count++];
1095 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1096 bb->length = ent->marked_count + 1;
1097 }
1098 }
1099 bbs->count = count;
1100}
1101
98130f40
AK		 1102/*
1103 * for second_map:
238c0a71
AK		 1104 * == MAP_0 get first map
1105 * == MAP_1 get second map
1106 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1107 */
1108static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1109 int slot,
1110 int second_map)
7eef0453
DW		 1111{
1112 struct imsm_map *map;
1113
5e7b0330 1114 map = get_imsm_map(dev, second_map);
7eef0453 1115
ff077194
DW		 1116 /* top byte identifies disk under rebuild */
1117 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1118}
1119
1120#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1121static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1122{
98130f40 1123 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1124
1125 return ord_to_idx(ord);
7eef0453
DW		 1126}
1127
be73972f
DW		 1128static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1129{
1130 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1131}
1132
f21e18ca 1133static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1134{
1135 int slot;
1136 __u32 ord;
1137
1138 for (slot = 0; slot < map->num_members; slot++) {
1139 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1140 if (ord_to_idx(ord) == idx)
1141 return slot;
1142 }
1143
1144 return -1;
1145}
1146
cdddbdbc
DW		 1147static int get_imsm_raid_level(struct imsm_map *map)
1148{
1149 if (map->raid_level == 1) {
1150 if (map->num_members == 2)
1151 return 1;
1152 else
1153 return 10;
1154 }
1155
1156 return map->raid_level;
1157}
1158
c2c087e6
DW		 1159static int cmp_extent(const void *av, const void *bv)
1160{
1161 const struct extent *a = av;
1162 const struct extent *b = bv;
1163 if (a->start < b->start)
1164 return -1;
1165 if (a->start > b->start)
1166 return 1;
1167 return 0;
1168}
1169
0dcecb2e 1170static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1171{
c2c087e6 1172 int memberships = 0;
620b1713 1173 int i;
c2c087e6 1174
949c47a0
DW		 1175 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1176 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1177 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1178
620b1713
DW		 1179 if (get_imsm_disk_slot(map, dl->index) >= 0)
1180 memberships++;
c2c087e6 1181 }
0dcecb2e
DW		 1182
1183 return memberships;
1184}
1185
b81221b7
CA		 1186static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1187
486720e0 1188static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1189{
1190 if (lo == 0 || hi == 0)
1191 return 1;
486720e0
JS		 1192 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1193 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1194 return 0;
1195}
1196
1197static unsigned long long join_u32(__u32 lo, __u32 hi)
1198{
1199 return (unsigned long long)__le32_to_cpu(lo) |
1200 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1201}
1202
1203static unsigned long long total_blocks(struct imsm_disk *disk)
1204{
1205 if (disk == NULL)
1206 return 0;
1207 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1208}
1209
1210static unsigned long long pba_of_lba0(struct imsm_map *map)
1211{
1212 if (map == NULL)
1213 return 0;
1214 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1215}
1216
1217static unsigned long long blocks_per_member(struct imsm_map *map)
1218{
1219 if (map == NULL)
1220 return 0;
1221 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1222}
1223
1224static unsigned long long num_data_stripes(struct imsm_map *map)
1225{
1226 if (map == NULL)
1227 return 0;
1228 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1229}
1230
fcc2c9da
MD		 1231static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1232{
1233 if (dev == NULL)
1234 return 0;
1235 return join_u32(dev->size_low, dev->size_high);
1236}
1237
9f421827
PB		 1238static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1239{
1240 if (migr_rec == NULL)
1241 return 0;
1242 return join_u32(migr_rec->ckpt_area_pba_lo,
1243 migr_rec->ckpt_area_pba_hi);
1244}
1245
1246static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1247{
1248 if (migr_rec == NULL)
1249 return 0;
1250 return join_u32(migr_rec->curr_migr_unit_lo,
1251 migr_rec->curr_migr_unit_hi);
1252}
1253
1254static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1255{
1256 if (migr_rec == NULL)
1257 return 0;
1258 return join_u32(migr_rec->dest_1st_member_lba_lo,
1259 migr_rec->dest_1st_member_lba_hi);
1260}
1261
1262static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1263{
1264 if (migr_rec == NULL)
1265 return 0;
1266 return join_u32(migr_rec->num_migr_units_lo,
1267 migr_rec->num_migr_units_hi);
1268}
1269
5551b113
CA		 1270static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1271{
1272 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1273}
1274
1275static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1276{
1277 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1278}
1279
1280static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1281{
1282 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1283}
1284
1285static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1286{
1287 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1288}
1289
fcc2c9da
MD		 1290static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1291{
1292 split_ull(n, &dev->size_low, &dev->size_high);
1293}
1294
9f421827
PB		 1295static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1296 unsigned long long n)
1297{
1298 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1299}
1300
1301static void set_current_migr_unit(struct migr_record *migr_rec,
1302 unsigned long long n)
1303{
1304 split_ull(n, &migr_rec->curr_migr_unit_lo,
1305 &migr_rec->curr_migr_unit_hi);
1306}
1307
1308static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1309 unsigned long long n)
1310{
1311 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1312 &migr_rec->dest_1st_member_lba_hi);
1313}
1314
1315static void set_num_migr_units(struct migr_record *migr_rec,
1316 unsigned long long n)
1317{
1318 split_ull(n, &migr_rec->num_migr_units_lo,
1319 &migr_rec->num_migr_units_hi);
1320}
1321
44490938
MD		 1322static unsigned long long per_dev_array_size(struct imsm_map *map)
1323{
1324 unsigned long long array_size = 0;
1325
1326 if (map == NULL)
1327 return array_size;
1328
1329 array_size = num_data_stripes(map) * map->blocks_per_strip;
1330 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1331 array_size *= 2;
1332
1333 return array_size;
1334}
1335
05501181
PB		 1336static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1337 int get_minimal_reservation)
0dcecb2e
DW		 1338{
1339 /* find a list of used extents on the given physical device */
1340 struct extent *rv, *e;
620b1713 1341 int i;
0dcecb2e 1342 int memberships = count_memberships(dl, super);
b276dd33
DW		 1343 __u32 reservation;
1344
1345 /* trim the reserved area for spares, so they can join any array
1346 * regardless of whether the OROM has assigned sectors from the
1347 * IMSM_RESERVED_SECTORS region
1348 */
05501181 1349 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1350 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1351 else
1352 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1353
503975b9 1354 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1355 e = rv;
1356
949c47a0
DW		 1357 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1358 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1359 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1360
620b1713 1361 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1362 e->start = pba_of_lba0(map);
44490938 1363 e->size = per_dev_array_size(map);
620b1713 1364 e++;
c2c087e6
DW		 1365 }
1366 }
1367 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1368
1011e834 1369 /* determine the start of the metadata
14e8215b
DW		 1370 * when no raid devices are defined use the default
1371 * ...otherwise allow the metadata to truncate the value
1372 * as is the case with older versions of imsm
1373 */
1374 if (memberships) {
1375 struct extent *last = &rv[memberships - 1];
5551b113 1376 unsigned long long remainder;
14e8215b 1377
5551b113 1378 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1379 /* round down to 1k block to satisfy precision of the kernel
1380 * 'size' interface
1381 */
1382 remainder &= ~1UL;
1383 /* make sure remainder is still sane */
f21e18ca 1384 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1385 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1386 if (reservation > remainder)
1387 reservation = remainder;
1388 }
5551b113 1389 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1390 e->size = 0;
1391 return rv;
1392}
1393
14e8215b
DW		 1394/* try to determine how much space is reserved for metadata from
1395 * the last get_extents() entry, otherwise fallback to the
1396 * default
1397 */
1398static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1399{
1400 struct extent *e;
1401 int i;
1402 __u32 rv;
1403
1404 /* for spares just return a minimal reservation which will grow
1405 * once the spare is picked up by an array
1406 */
1407 if (dl->index == -1)
1408 return MPB_SECTOR_CNT;
1409
05501181 1410 e = get_extents(super, dl, 0);
14e8215b
DW		 1411 if (!e)
1412 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1413
1414 /* scroll to last entry */
1415 for (i = 0; e[i].size; i++)
1416 continue;
1417
5551b113 1418 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1419
1420 free(e);
1421
1422 return rv;
1423}
1424
25ed7e59
DW		 1425static int is_spare(struct imsm_disk *disk)
1426{
1427 return (disk->status & SPARE_DISK) == SPARE_DISK;
1428}
1429
1430static int is_configured(struct imsm_disk *disk)
1431{
1432 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1433}
1434
1435static int is_failed(struct imsm_disk *disk)
1436{
1437 return (disk->status & FAILED_DISK) == FAILED_DISK;
1438}
1439
2432ce9b
AP		 1440static int is_journal(struct imsm_disk *disk)
1441{
1442 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1443}
1444
b53bfba6
TM		 1445/* round array size down to closest MB and ensure it splits evenly
1446 * between members
1447 */
1448static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1449 disk_count)
1450{
1451 size /= disk_count;
1452 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1453 size *= disk_count;
1454
1455 return size;
1456}
1457
8b9cd157
MK		 1458static int able_to_resync(int raid_level, int missing_disks)
1459{
1460 int max_missing_disks = 0;
1461
1462 switch (raid_level) {
1463 case 10:
1464 max_missing_disks = 1;
1465 break;
1466 default:
1467 max_missing_disks = 0;
1468 }
1469 return missing_disks <= max_missing_disks;
1470}
1471
b81221b7
CA		 1472/* try to determine how much space is reserved for metadata from
1473 * the last get_extents() entry on the smallest active disk,
1474 * otherwise fallback to the default
1475 */
1476static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1477{
1478 struct extent *e;
1479 int i;
5551b113
CA		 1480 unsigned long long min_active;
1481 __u32 remainder;
b81221b7
CA		 1482 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1483 struct dl *dl, *dl_min = NULL;
1484
1485 if (!super)
1486 return rv;
1487
1488 min_active = 0;
1489 for (dl = super->disks; dl; dl = dl->next) {
1490 if (dl->index < 0)
1491 continue;
5551b113
CA		 1492 unsigned long long blocks = total_blocks(&dl->disk);
1493 if (blocks < min_active || min_active == 0) {
b81221b7 1494 dl_min = dl;
5551b113 1495 min_active = blocks;
b81221b7
CA		 1496 }
1497 }
1498 if (!dl_min)
1499 return rv;
1500
1501 /* find last lba used by subarrays on the smallest active disk */
05501181 1502 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1503 if (!e)
1504 return rv;
1505 for (i = 0; e[i].size; i++)
1506 continue;
1507
1508 remainder = min_active - e[i].start;
1509 free(e);
1510
1511 /* to give priority to recovery we should not require full
1512 IMSM_RESERVED_SECTORS from the spare */
1513 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1514
1515 /* if real reservation is smaller use that value */
1516 return (remainder < rv) ? remainder : rv;
1517}
1518
fbfdcb06
AO		 1519/*
1520 * Return minimum size of a spare and sector size
1521 * that can be used in this array
1522 */
1523int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1524{
1525 struct intel_super *super = st->sb;
1526 struct dl *dl;
1527 struct extent *e;
1528 int i;
fbfdcb06
AO		 1529 unsigned long long size = 0;
1530
1531 c->min_size = 0;
4b57ecf6 1532 c->sector_size = 0;
80e7f8c3
AC		 1533
1534 if (!super)
fbfdcb06 1535 return -EINVAL;
80e7f8c3
AC		 1536 /* find first active disk in array */
1537 dl = super->disks;
1538 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1539 dl = dl->next;
1540 if (!dl)
fbfdcb06 1541 return -EINVAL;
80e7f8c3 1542 /* find last lba used by subarrays */
05501181 1543 e = get_extents(super, dl, 0);
80e7f8c3 1544 if (!e)
fbfdcb06 1545 return -EINVAL;
80e7f8c3
AC		 1546 for (i = 0; e[i].size; i++)
1547 continue;
1548 if (i > 0)
fbfdcb06 1549 size = e[i-1].start + e[i-1].size;
80e7f8c3 1550 free(e);
b81221b7 1551
80e7f8c3 1552 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1553 size += imsm_min_reserved_sectors(super);
1554
1555 c->min_size = size * 512;
4b57ecf6 1556 c->sector_size = super->sector_size;
b81221b7 1557
fbfdcb06 1558 return 0;
80e7f8c3
AC		 1559}
1560
d1e02575
AK		 1561static int is_gen_migration(struct imsm_dev *dev);
1562
f36a9ecd
PB		 1563#define IMSM_4K_DIV 8
1564
c47b0ff6
AK		 1565static __u64 blocks_per_migr_unit(struct intel_super *super,
1566 struct imsm_dev *dev);
1e5c6983 1567
c47b0ff6
AK		 1568static void print_imsm_dev(struct intel_super *super,
1569 struct imsm_dev *dev,
1570 char *uuid,
1571 int disk_idx)
cdddbdbc
DW		 1572{
1573 __u64 sz;
0d80bb2f 1574 int slot, i;
238c0a71
AK		 1575 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1576 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1577 __u32 ord;
cdddbdbc
DW		 1578
1579 printf("\n");
1e7bc0ed 1580 printf("[%.16s]:\n", dev->volume);
44470971 1581 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1582 printf(" RAID Level : %d", get_imsm_raid_level(map));
1583 if (map2)
1584 printf(" <-- %d", get_imsm_raid_level(map2));
1585 printf("\n");
1586 printf(" Members : %d", map->num_members);
1587 if (map2)
1588 printf(" <-- %d", map2->num_members);
1589 printf("\n");
0d80bb2f
DW		 1590 printf(" Slots : [");
1591 for (i = 0; i < map->num_members; i++) {
238c0a71 1592 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1593 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1594 }
dd8bcb3b
AK		 1595 printf("]");
1596 if (map2) {
1597 printf(" <-- [");
1598 for (i = 0; i < map2->num_members; i++) {
238c0a71 1599 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1600 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1601 }
1602 printf("]");
1603 }
1604 printf("\n");
7095bccb
AK		 1605 printf(" Failed disk : ");
1606 if (map->failed_disk_num == 0xff)
1607 printf("none");
1608 else
1609 printf("%i", map->failed_disk_num);
1610 printf("\n");
620b1713
DW		 1611 slot = get_imsm_disk_slot(map, disk_idx);
1612 if (slot >= 0) {
238c0a71 1613 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1614 printf(" This Slot : %d%s\n", slot,
1615 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1616 } else
cdddbdbc 1617 printf(" This Slot : ?\n");
84918897 1618 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1619 sz = imsm_dev_size(dev);
84918897
MK		 1620 printf(" Array Size : %llu%s\n",
1621 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1622 human_size(sz * 512));
5551b113 1623 sz = blocks_per_member(map);
84918897
MK		 1624 printf(" Per Dev Size : %llu%s\n",
1625 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1626 human_size(sz * 512));
5551b113
CA		 1627 printf(" Sector Offset : %llu\n",
1628 pba_of_lba0(map));
1629 printf(" Num Stripes : %llu\n",
1630 num_data_stripes(map));
dd8bcb3b 1631 printf(" Chunk Size : %u KiB",
cdddbdbc 1632 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1633 if (map2)
1634 printf(" <-- %u KiB",
1635 __le16_to_cpu(map2->blocks_per_strip) / 2);
1636 printf("\n");
cdddbdbc 1637 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1638 printf(" Migrate State : ");
1484e727
DW		 1639 if (dev->vol.migr_state) {
1640 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1641 printf("initialize\n");
1484e727 1642 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1643 printf("rebuild\n");
1484e727 1644 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1645 printf("check\n");
1484e727 1646 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1647 printf("general migration\n");
1484e727 1648 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1649 printf("state change\n");
1484e727 1650 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1651 printf("repair\n");
1484e727 1652 else
8655a7b1
DW		 1653 printf("<unknown:%d>\n", migr_type(dev));
1654 } else
1655 printf("idle\n");
3393c6af
DW		 1656 printf(" Map State : %s", map_state_str[map->map_state]);
1657 if (dev->vol.migr_state) {
238c0a71 1658 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1659
b10b37b8 1660 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1661 printf("\n Checkpoint : %u ",
1662 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1663 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1664 printf("(N/A)");
1665 else
1666 printf("(%llu)", (unsigned long long)
1667 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1668 }
1669 printf("\n");
2432ce9b
AP		 1670 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1671 "dirty" : "clean");
1672 printf(" RWH Policy : ");
c2462068 1673 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1674 printf("off\n");
1675 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1676 printf("PPL distributed\n");
1677 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1678 printf("PPL journaling drive\n");
c2462068
PB		 1679 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1680 printf("Multiple distributed PPLs\n");
1681 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1682 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1683 else
1684 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1685}
1686
ef5c214e
MK		 1687static void print_imsm_disk(struct imsm_disk *disk,
1688 int index,
1689 __u32 reserved,
1690 unsigned int sector_size) {
1f24f035 1691 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1692 __u64 sz;
1693
0ec1f4e8 1694 if (index < -1 || !disk)
e9d82038
DW		 1695 return;
1696
cdddbdbc 1697 printf("\n");
1f24f035 1698 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1699 if (index >= 0)
1700 printf(" Disk%02d Serial : %s\n", index, str);
1701 else
1702 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1703 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1704 is_configured(disk) ? " active" : "",
1705 is_failed(disk) ? " failed" : "",
1706 is_journal(disk) ? " journal" : "");
cdddbdbc 1707 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1708 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1709 printf(" Usable Size : %llu%s\n",
1710 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1711 human_size(sz * 512));
1712}
1713
de44e46f
PB		 1714void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1715{
1716 struct migr_record *migr_rec = super->migr_rec;
1717
1718 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1719 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1720 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1721 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1722 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1723 set_migr_chkp_area_pba(migr_rec,
1724 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1725 set_migr_dest_1st_member_lba(migr_rec,
1726 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1727}
1728
f36a9ecd
PB		 1729void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1730{
1731 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1732}
1733
1734void convert_to_4k(struct intel_super *super)
1735{
1736 struct imsm_super *mpb = super->anchor;
1737 struct imsm_disk *disk;
1738 int i;
e4467bc7 1739 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1740
1741 for (i = 0; i < mpb->num_disks ; i++) {
1742 disk = __get_imsm_disk(mpb, i);
1743 /* disk */
1744 convert_to_4k_imsm_disk(disk);
1745 }
1746 for (i = 0; i < mpb->num_raid_devs; i++) {
1747 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1748 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1749 /* dev */
fcc2c9da 1750 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1751 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1752
1753 /* map0 */
1754 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1755 map->blocks_per_strip /= IMSM_4K_DIV;
1756 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1757
1758 if (dev->vol.migr_state) {
1759 /* map1 */
1760 map = get_imsm_map(dev, MAP_1);
1761 set_blocks_per_member(map,
1762 blocks_per_member(map)/IMSM_4K_DIV);
1763 map->blocks_per_strip /= IMSM_4K_DIV;
1764 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1765 }
1766 }
e4467bc7
TM		 1767 if (bbm_log_size) {
1768 struct bbm_log *log = (void *)mpb +
1769 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1770 __u32 i;
1771
1772 for (i = 0; i < log->entry_count; i++) {
1773 struct bbm_log_entry *entry =
1774 &log->marked_block_entries[i];
1775
1776 __u8 count = entry->marked_count + 1;
1777 unsigned long long sector =
1778 __le48_to_cpu(&entry->defective_block_start);
1779
1780 entry->defective_block_start =
1781 __cpu_to_le48(sector/IMSM_4K_DIV);
1782 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1783 }
1784 }
f36a9ecd
PB		 1785
1786 mpb->check_sum = __gen_imsm_checksum(mpb);
1787}
1788
520e69e2
AK		 1789void examine_migr_rec_imsm(struct intel_super *super)
1790{
1791 struct migr_record *migr_rec = super->migr_rec;
1792 struct imsm_super *mpb = super->anchor;
1793 int i;
1794
1795 for (i = 0; i < mpb->num_raid_devs; i++) {
1796 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1797 struct imsm_map *map;
b4ab44d8 1798 int slot = -1;
3136abe5 1799
520e69e2
AK		 1800 if (is_gen_migration(dev) == 0)
1801 continue;
1802
1803 printf("\nMigration Record Information:");
3136abe5 1804
44bfe6df
AK		 1805 /* first map under migration */
1806 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1807 if (map)
1808 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1809 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1810 printf(" Empty\n ");
1811 printf("Examine one of first two disks in array\n");
1812 break;
1813 }
1814 printf("\n Status : ");
1815 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1816 printf("Normal\n");
1817 else
1818 printf("Contains Data\n");
9f421827
PB		 1819 printf(" Current Unit : %llu\n",
1820 current_migr_unit(migr_rec));
520e69e2
AK		 1821 printf(" Family : %u\n",
1822 __le32_to_cpu(migr_rec->family_num));
1823 printf(" Ascending : %u\n",
1824 __le32_to_cpu(migr_rec->ascending_migr));
1825 printf(" Blocks Per Unit : %u\n",
1826 __le32_to_cpu(migr_rec->blocks_per_unit));
1827 printf(" Dest. Depth Per Unit : %u\n",
1828 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1829 printf(" Checkpoint Area pba : %llu\n",
1830 migr_chkp_area_pba(migr_rec));
1831 printf(" First member lba : %llu\n",
1832 migr_dest_1st_member_lba(migr_rec));
1833 printf(" Total Number of Units : %llu\n",
1834 get_num_migr_units(migr_rec));
1835 printf(" Size of volume : %llu\n",
1836 join_u32(migr_rec->post_migr_vol_cap,
1837 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1838 printf(" Record was read from : %u\n",
1839 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1840
1841 break;
1842 }
1843}
f36a9ecd 1844
de44e46f
PB		 1845void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1846{
1847 struct migr_record *migr_rec = super->migr_rec;
1848
1849 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1850 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1851 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1852 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1853 &migr_rec->post_migr_vol_cap,
1854 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1855 set_migr_chkp_area_pba(migr_rec,
1856 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1857 set_migr_dest_1st_member_lba(migr_rec,
1858 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1859}
1860
f36a9ecd
PB		 1861void convert_from_4k(struct intel_super *super)
1862{
1863 struct imsm_super *mpb = super->anchor;
1864 struct imsm_disk *disk;
1865 int i;
e4467bc7 1866 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1867
1868 for (i = 0; i < mpb->num_disks ; i++) {
1869 disk = __get_imsm_disk(mpb, i);
1870 /* disk */
1871 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1872 }
1873
1874 for (i = 0; i < mpb->num_raid_devs; i++) {
1875 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1876 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1877 /* dev */
fcc2c9da 1878 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1879 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1880
1881 /* map0 */
1882 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1883 map->blocks_per_strip *= IMSM_4K_DIV;
1884 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1885
1886 if (dev->vol.migr_state) {
1887 /* map1 */
1888 map = get_imsm_map(dev, MAP_1);
1889 set_blocks_per_member(map,
1890 blocks_per_member(map)*IMSM_4K_DIV);
1891 map->blocks_per_strip *= IMSM_4K_DIV;
1892 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1893 }
1894 }
e4467bc7
TM		 1895 if (bbm_log_size) {
1896 struct bbm_log *log = (void *)mpb +
1897 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1898 __u32 i;
1899
1900 for (i = 0; i < log->entry_count; i++) {
1901 struct bbm_log_entry *entry =
1902 &log->marked_block_entries[i];
1903
1904 __u8 count = entry->marked_count + 1;
1905 unsigned long long sector =
1906 __le48_to_cpu(&entry->defective_block_start);
1907
1908 entry->defective_block_start =
1909 __cpu_to_le48(sector*IMSM_4K_DIV);
1910 entry->marked_count = count*IMSM_4K_DIV - 1;
1911 }
1912 }
f36a9ecd
PB		 1913
1914 mpb->check_sum = __gen_imsm_checksum(mpb);
1915}
1916
19482bcc
AK		 1917/*******************************************************************************
1918 * function: imsm_check_attributes
1919 * Description: Function checks if features represented by attributes flags
1011e834 1920 * are supported by mdadm.
19482bcc
AK		 1921 * Parameters:
1922 * attributes - Attributes read from metadata
1923 * Returns:
1011e834
N		 1924 * 0 - passed attributes contains unsupported features flags
1925 * 1 - all features are supported
19482bcc
AK		 1926 ******************************************************************************/
1927static int imsm_check_attributes(__u32 attributes)
1928{
1929 int ret_val = 1;
418f9b36
N		 1930 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1931
1932 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1933
1934 not_supported &= attributes;
1935 if (not_supported) {
e7b84f9d 1936 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1937 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1938 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1939 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1940 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1941 }
1942 if (not_supported & MPB_ATTRIB_2TB) {
1943 dprintf("\t\tMPB_ATTRIB_2TB\n");
1944 not_supported ^= MPB_ATTRIB_2TB;
1945 }
1946 if (not_supported & MPB_ATTRIB_RAID0) {
1947 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1948 not_supported ^= MPB_ATTRIB_RAID0;
1949 }
1950 if (not_supported & MPB_ATTRIB_RAID1) {
1951 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1952 not_supported ^= MPB_ATTRIB_RAID1;
1953 }
1954 if (not_supported & MPB_ATTRIB_RAID10) {
1955 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1956 not_supported ^= MPB_ATTRIB_RAID10;
1957 }
1958 if (not_supported & MPB_ATTRIB_RAID1E) {
1959 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1960 not_supported ^= MPB_ATTRIB_RAID1E;
1961 }
1962 if (not_supported & MPB_ATTRIB_RAID5) {
1963 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1964 not_supported ^= MPB_ATTRIB_RAID5;
1965 }
1966 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1967 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1968 not_supported ^= MPB_ATTRIB_RAIDCNG;
1969 }
1970 if (not_supported & MPB_ATTRIB_BBM) {
1971 dprintf("\t\tMPB_ATTRIB_BBM\n");
1972 not_supported ^= MPB_ATTRIB_BBM;
1973 }
1974 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1975 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1976 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1977 }
1978 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1979 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1980 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1981 }
1982 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1983 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1984 not_supported ^= MPB_ATTRIB_2TB_DISK;
1985 }
1986 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1987 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1988 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1989 }
1990 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1991 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1992 not_supported ^= MPB_ATTRIB_NEVER_USE;
1993 }
1994
1995 if (not_supported)
1ade5cc1 1996 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1997
1998 ret_val = 0;
1999 }
2000
2001 return ret_val;
2002}
2003
a5d85af7 2004static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2005
cdddbdbc
DW		 2006static void examine_super_imsm(struct supertype *st, char *homehost)
2007{
2008 struct intel_super *super = st->sb;
949c47a0 2009 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2010 char str[MAX_SIGNATURE_LENGTH];
2011 int i;
27fd6274
DW		 2012 struct mdinfo info;
2013 char nbuf[64];
cdddbdbc 2014 __u32 sum;
14e8215b 2015 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2016 struct dl *dl;
27fd6274 2017
618f4e6d
XN		 2018 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2019 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2020 printf(" Magic : %s\n", str);
cdddbdbc 2021 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2022 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2023 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2024 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 2025 printf(" Attributes : ");
2026 if (imsm_check_attributes(mpb->attributes))
2027 printf("All supported\n");
2028 else
2029 printf("not supported\n");
a5d85af7 2030 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2031 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2032 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2033 sum = __le32_to_cpu(mpb->check_sum);
2034 printf(" Checksum : %08x %s\n", sum,
949c47a0 2035 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2036 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2037 printf(" Disks : %d\n", mpb->num_disks);
2038 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2039 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2040 super->disks->index, reserved, super->sector_size);
8d67477f 2041 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2042 struct bbm_log *log = super->bbm_log;
2043
2044 printf("\n");
2045 printf("Bad Block Management Log:\n");
2046 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2047 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2048 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2049 }
44470971
DW		 2050 for (i = 0; i < mpb->num_raid_devs; i++) {
2051 struct mdinfo info;
2052 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2053
2054 super->current_vol = i;
a5d85af7 2055 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2056 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2057 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2058 }
cdddbdbc
DW		 2059 for (i = 0; i < mpb->num_disks; i++) {
2060 if (i == super->disks->index)
2061 continue;
ef5c214e
MK		 2062 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2063 super->sector_size);
cdddbdbc 2064 }
94827db3 2065
0ec1f4e8
DW		 2066 for (dl = super->disks; dl; dl = dl->next)
2067 if (dl->index == -1)
ef5c214e
MK		 2068 print_imsm_disk(&dl->disk, -1, reserved,
2069 super->sector_size);
520e69e2
AK		 2070
2071 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2072}
2073
061f2c6a 2074static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2075{
27fd6274 2076 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2077 struct mdinfo info;
2078 char nbuf[64];
1e7bc0ed 2079 struct intel_super *super = st->sb;
1e7bc0ed 2080
0d5a423f
DW		 2081 if (!super->anchor->num_raid_devs) {
2082 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2083 return;
0d5a423f 2084 }
ff54de6e 2085
a5d85af7 2086 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2087 fname_from_uuid(st, &info, nbuf, ':');
2088 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2089}
2090
2091static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2092{
2093 /* We just write a generic IMSM ARRAY entry */
2094 struct mdinfo info;
2095 char nbuf[64];
2096 char nbuf1[64];
2097 struct intel_super *super = st->sb;
2098 int i;
2099
2100 if (!super->anchor->num_raid_devs)
2101 return;
2102
a5d85af7 2103 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2104 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2105 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2106 struct imsm_dev *dev = get_imsm_dev(super, i);
2107
2108 super->current_vol = i;
a5d85af7 2109 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2110 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2111 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2112 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2113 }
cdddbdbc
DW		 2114}
2115
9d84c8ea
DW		 2116static void export_examine_super_imsm(struct supertype *st)
2117{
2118 struct intel_super *super = st->sb;
2119 struct imsm_super *mpb = super->anchor;
2120 struct mdinfo info;
2121 char nbuf[64];
2122
a5d85af7 2123 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2124 fname_from_uuid(st, &info, nbuf, ':');
2125 printf("MD_METADATA=imsm\n");
2126 printf("MD_LEVEL=container\n");
2127 printf("MD_UUID=%s\n", nbuf+5);
2128 printf("MD_DEVICES=%u\n", mpb->num_disks);
2129}
2130
74db60b0
N		 2131static int copy_metadata_imsm(struct supertype *st, int from, int to)
2132{
f36a9ecd 2133 /* The second last sector of the device contains
74db60b0
N		 2134 * the "struct imsm_super" metadata.
2135 * This contains mpb_size which is the size in bytes of the
2136 * extended metadata. This is located immediately before
2137 * the imsm_super.
2138 * We want to read all that, plus the last sector which
2139 * may contain a migration record, and write it all
2140 * to the target.
2141 */
2142 void *buf;
2143 unsigned long long dsize, offset;
2144 int sectors;
2145 struct imsm_super *sb;
f36a9ecd
PB		 2146 struct intel_super *super = st->sb;
2147 unsigned int sector_size = super->sector_size;
2148 unsigned int written = 0;
74db60b0 2149
de44e46f 2150 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 2151 return 1;
2152
2153 if (!get_dev_size(from, NULL, &dsize))
2154 goto err;
2155
f36a9ecd 2156 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 2157 goto err;
466070ad 2158 if ((unsigned int)read(from, buf, sector_size) != sector_size)
74db60b0
N		 2159 goto err;
2160 sb = buf;
2161 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
2162 goto err;
2163
f36a9ecd
PB		 2164 sectors = mpb_sectors(sb, sector_size) + 2;
2165 offset = dsize - sectors * sector_size;
74db60b0
N		 2166 if (lseek64(from, offset, 0) < 0 ||
2167 lseek64(to, offset, 0) < 0)
2168 goto err;
f36a9ecd
PB		 2169 while (written < sectors * sector_size) {
2170 int n = sectors*sector_size - written;
74db60b0
N		 2171 if (n > 4096)
2172 n = 4096;
2173 if (read(from, buf, n) != n)
2174 goto err;
2175 if (write(to, buf, n) != n)
2176 goto err;
2177 written += n;
2178 }
2179 free(buf);
2180 return 0;
2181err:
2182 free(buf);
2183 return 1;
2184}
2185
b771faef
BK		 2186static void detail_super_imsm(struct supertype *st, char *homehost,
2187 char *subarray)
cdddbdbc 2188{
3ebe00a1
DW		 2189 struct mdinfo info;
2190 char nbuf[64];
b771faef
BK		 2191 struct intel_super *super = st->sb;
2192 int temp_vol = super->current_vol;
2193
2194 if (subarray)
2195 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2196
a5d85af7 2197 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2198 fname_from_uuid(st, &info, nbuf, ':');
65884368 2199 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2200
2201 super->current_vol = temp_vol;
cdddbdbc
DW		 2202}
2203
b771faef 2204static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2205{
ff54de6e
N		 2206 struct mdinfo info;
2207 char nbuf[64];
b771faef
BK		 2208 struct intel_super *super = st->sb;
2209 int temp_vol = super->current_vol;
2210
2211 if (subarray)
2212 super->current_vol = strtoul(subarray, NULL, 10);
2213
a5d85af7 2214 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2215 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2216 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2217
2218 super->current_vol = temp_vol;
cdddbdbc 2219}
d665cc31 2220
6da53c0e
BK		 2221static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2222 size_t serial_buf_len);
d665cc31
DW		 2223static void fd2devname(int fd, char *name);
2224
120dc887 2225static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2226{
120dc887
LM		 2227 /* dump an unsorted list of devices attached to AHCI Intel storage
2228 * controller, as well as non-connected ports
d665cc31
DW		 2229 */
2230 int hba_len = strlen(hba_path) + 1;
2231 struct dirent *ent;
2232 DIR *dir;
2233 char *path = NULL;
2234 int err = 0;
2235 unsigned long port_mask = (1 << port_count) - 1;
2236
f21e18ca 2237 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2238 if (verbose > 0)
e7b84f9d 2239 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2240 return 2;
2241 }
2242
2243 /* scroll through /sys/dev/block looking for devices attached to
2244 * this hba
2245 */
2246 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2247 if (!dir)
2248 return 1;
2249
2250 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2251 int fd;
2252 char model[64];
2253 char vendor[64];
2254 char buf[1024];
2255 int major, minor;
2256 char *device;
2257 char *c;
2258 int port;
2259 int type;
2260
2261 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2262 continue;
2263 path = devt_to_devpath(makedev(major, minor));
2264 if (!path)
2265 continue;
2266 if (!path_attached_to_hba(path, hba_path)) {
2267 free(path);
2268 path = NULL;
2269 continue;
2270 }
2271
2272 /* retrieve the scsi device type */
2273 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2274 if (verbose > 0)
e7b84f9d 2275 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2276 err = 2;
2277 break;
2278 }
2279 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2280 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2281 if (verbose > 0)
e7b84f9d 2282 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2283 path);
2284 err = 2;
2285 free(device);
2286 break;
2287 }
2288 type = strtoul(buf, NULL, 10);
2289
2290 /* if it's not a disk print the vendor and model */
2291 if (!(type == 0 || type == 7 || type == 14)) {
2292 vendor[0] = '\0';
2293 model[0] = '\0';
2294 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2295 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2296 strncpy(vendor, buf, sizeof(vendor));
2297 vendor[sizeof(vendor) - 1] = '\0';
2298 c = (char *) &vendor[sizeof(vendor) - 1];
2299 while (isspace(*c) || *c == '\0')
2300 *c-- = '\0';
2301
2302 }
2303 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2304 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2305 strncpy(model, buf, sizeof(model));
2306 model[sizeof(model) - 1] = '\0';
2307 c = (char *) &model[sizeof(model) - 1];
2308 while (isspace(*c) || *c == '\0')
2309 *c-- = '\0';
2310 }
2311
2312 if (vendor[0] && model[0])
2313 sprintf(buf, "%.64s %.64s", vendor, model);
2314 else
2315 switch (type) { /* numbers from hald/linux/device.c */
2316 case 1: sprintf(buf, "tape"); break;
2317 case 2: sprintf(buf, "printer"); break;
2318 case 3: sprintf(buf, "processor"); break;
2319 case 4:
2320 case 5: sprintf(buf, "cdrom"); break;
2321 case 6: sprintf(buf, "scanner"); break;
2322 case 8: sprintf(buf, "media_changer"); break;
2323 case 9: sprintf(buf, "comm"); break;
2324 case 12: sprintf(buf, "raid"); break;
2325 default: sprintf(buf, "unknown");
2326 }
2327 } else
2328 buf[0] = '\0';
2329 free(device);
2330
2331 /* chop device path to 'host%d' and calculate the port number */
2332 c = strchr(&path[hba_len], '/');
4e5e717d 2333 if (!c) {
ba728be7 2334 if (verbose > 0)
e7b84f9d 2335 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2336 err = 2;
2337 break;
2338 }
d665cc31 2339 *c = '\0';
0858eccf
AP		 2340 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2341 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2342 port -= host_base;
2343 else {
ba728be7 2344 if (verbose > 0) {
d665cc31 2345 *c = '/'; /* repair the full string */
e7b84f9d 2346 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2347 path);
2348 }
2349 err = 2;
2350 break;
2351 }
2352
2353 /* mark this port as used */
2354 port_mask &= ~(1 << port);
2355
2356 /* print out the device information */
2357 if (buf[0]) {
2358 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2359 continue;
2360 }
2361
2362 fd = dev_open(ent->d_name, O_RDONLY);
2363 if (fd < 0)
2364 printf(" Port%d : - disk info unavailable -\n", port);
2365 else {
2366 fd2devname(fd, buf);
2367 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2368 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2369 sizeof(buf)) == 0)
2370 printf(" (%s)\n", buf);
d665cc31 2371 else
664d5325 2372 printf(" ()\n");
4dab422a 2373 close(fd);
d665cc31 2374 }
d665cc31
DW		 2375 free(path);
2376 path = NULL;
2377 }
2378 if (path)
2379 free(path);
2380 if (dir)
2381 closedir(dir);
2382 if (err == 0) {
2383 int i;
2384
2385 for (i = 0; i < port_count; i++)
2386 if (port_mask & (1 << i))
2387 printf(" Port%d : - no device attached -\n", i);
2388 }
2389
2390 return err;
2391}
2392
6da53c0e 2393static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2394{
6da53c0e 2395 char buf[1024];
60f0f54d
PB		 2396 struct dirent *ent;
2397 DIR *dir;
6da53c0e
BK		 2398 char *rp;
2399 int fd;
60f0f54d 2400
6da53c0e 2401 dir = opendir("/sys/block/");
b9135011 2402 if (!dir)
b5eece69 2403 return 1;
b9135011
JS		 2404
2405 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e
BK		 2406 if (strstr(ent->d_name, "nvme")) {
2407 sprintf(buf, "/sys/block/%s", ent->d_name);
2408 rp = realpath(buf, NULL);
2409 if (!rp)
2410 continue;
2411 if (path_attached_to_hba(rp, hba->path)) {
2412 fd = open_dev(ent->d_name);
2413 if (fd < 0) {
2414 free(rp);
2415 continue;
2416 }
60f0f54d 2417
6da53c0e
BK		 2418 fd2devname(fd, buf);
2419 if (hba->type == SYS_DEV_VMD)
2420 printf(" NVMe under VMD : %s", buf);
2421 else if (hba->type == SYS_DEV_NVME)
2422 printf(" NVMe Device : %s", buf);
2423 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2424 sizeof(buf)))
2425 printf(" (%s)\n", buf);
2426 else
2427 printf("()\n");
2428 close(fd);
2429 }
2430 free(rp);
60f0f54d 2431 }
60f0f54d
PB		 2432 }
2433
b9135011 2434 closedir(dir);
b5eece69 2435 return 0;
60f0f54d
PB		 2436}
2437
120dc887
LM		 2438static void print_found_intel_controllers(struct sys_dev *elem)
2439{
2440 for (; elem; elem = elem->next) {
e7b84f9d 2441 pr_err("found Intel(R) ");
120dc887
LM		 2442 if (elem->type == SYS_DEV_SATA)
2443 fprintf(stderr, "SATA ");
155cbb4c
LM		 2444 else if (elem->type == SYS_DEV_SAS)
2445 fprintf(stderr, "SAS ");
0858eccf
AP		 2446 else if (elem->type == SYS_DEV_NVME)
2447 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2448
2449 if (elem->type == SYS_DEV_VMD)
2450 fprintf(stderr, "VMD domain");
2451 else
2452 fprintf(stderr, "RAID controller");
2453
120dc887
LM		 2454 if (elem->pci_id)
2455 fprintf(stderr, " at %s", elem->pci_id);
2456 fprintf(stderr, ".\n");
2457 }
2458 fflush(stderr);
2459}
2460
120dc887
LM		 2461static int ahci_get_port_count(const char *hba_path, int *port_count)
2462{
2463 struct dirent *ent;
2464 DIR *dir;
2465 int host_base = -1;
2466
2467 *port_count = 0;
2468 if ((dir = opendir(hba_path)) == NULL)
2469 return -1;
2470
2471 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2472 int host;
2473
0858eccf
AP		 2474 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2475 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2476 continue;
2477 if (*port_count == 0)
2478 host_base = host;
2479 else if (host < host_base)
2480 host_base = host;
2481
2482 if (host + 1 > *port_count + host_base)
2483 *port_count = host + 1 - host_base;
2484 }
2485 closedir(dir);
2486 return host_base;
2487}
2488
a891a3c2
LM		 2489static void print_imsm_capability(const struct imsm_orom *orom)
2490{
0858eccf
AP		 2491 printf(" Platform : Intel(R) ");
2492 if (orom->capabilities == 0 && orom->driver_features == 0)
2493 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2494 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2495 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2496 else
2497 printf("Rapid Storage Technology%s\n",
2498 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2499 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2500 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2501 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2502 printf(" RAID Levels :%s%s%s%s%s\n",
2503 imsm_orom_has_raid0(orom) ? " raid0" : "",
2504 imsm_orom_has_raid1(orom) ? " raid1" : "",
2505 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2506 imsm_orom_has_raid10(orom) ? " raid10" : "",
2507 imsm_orom_has_raid5(orom) ? " raid5" : "");
2508 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2509 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2510 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2511 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2512 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2513 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2514 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2515 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2516 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2517 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2518 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2519 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2520 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2521 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2522 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2523 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2524 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2525 printf(" 2TB volumes :%s supported\n",
2526 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2527 printf(" 2TB disks :%s supported\n",
2528 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2529 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2530 printf(" Max Volumes : %d per array, %d per %s\n",
2531 orom->vpa, orom->vphba,
2532 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2533 return;
2534}
2535
e50cf220
MN		 2536static void print_imsm_capability_export(const struct imsm_orom *orom)
2537{
2538 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2539 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2540 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2541 orom->hotfix_ver, orom->build);
e50cf220
MN		 2542 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2543 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2544 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2545 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2546 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2547 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2548 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2549 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2550 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2551 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2552 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2553 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2554 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2555 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2556 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2557 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2558 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2559 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2560 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2561 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2562 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2563 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2564 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2565 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2566 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2567 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2568 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2569 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2570}
2571
9eafa1de 2572static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2573{
2574 /* There are two components to imsm platform support, the ahci SATA
2575 * controller and the option-rom. To find the SATA controller we
2576 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2577 * controller with the Intel vendor id is present. This approach
2578 * allows mdadm to leverage the kernel's ahci detection logic, with the
2579 * caveat that if ahci.ko is not loaded mdadm will not be able to
2580 * detect platform raid capabilities. The option-rom resides in a
2581 * platform "Adapter ROM". We scan for its signature to retrieve the
2582 * platform capabilities. If raid support is disabled in the BIOS the
2583 * option-rom capability structure will not be available.
2584 */
d665cc31 2585 struct sys_dev *list, *hba;
d665cc31
DW		 2586 int host_base = 0;
2587 int port_count = 0;
9eafa1de 2588 int result=1;
d665cc31 2589
5615172f 2590 if (enumerate_only) {
a891a3c2 2591 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2592 return 0;
a891a3c2
LM		 2593 list = find_intel_devices();
2594 if (!list)
2595 return 2;
2596 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2597 if (find_imsm_capability(hba)) {
2598 result = 0;
a891a3c2
LM		 2599 break;
2600 }
9eafa1de 2601 else
6b781d33 2602 result = 2;
a891a3c2 2603 }
a891a3c2 2604 return result;
5615172f
DW		 2605 }
2606
155cbb4c
LM		 2607 list = find_intel_devices();
2608 if (!list) {
ba728be7 2609 if (verbose > 0)
7a862a02 2610 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2611 return 2;
ba728be7 2612 } else if (verbose > 0)
155cbb4c 2613 print_found_intel_controllers(list);
d665cc31 2614
a891a3c2 2615 for (hba = list; hba; hba = hba->next) {
0858eccf 2616 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2617 continue;
0858eccf 2618 if (!find_imsm_capability(hba)) {
60f0f54d 2619 char buf[PATH_MAX];
e7b84f9d 2620 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2621 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2622 get_sys_dev_type(hba->type));
0858eccf
AP		 2623 continue;
2624 }
2625 result = 0;
2626 }
2627
2628 if (controller_path && result == 1) {
2629 pr_err("no active Intel(R) RAID controller found under %s\n",
2630 controller_path);
2631 return result;
2632 }
2633
5e1d6128 2634 const struct orom_entry *entry;
0858eccf 2635
5e1d6128 2636 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2637 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2638 print_imsm_capability(&entry->orom);
32716c51
PB		 2639 printf(" 3rd party NVMe :%s supported\n",
2640 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2641 for (hba = list; hba; hba = hba->next) {
2642 if (hba->type == SYS_DEV_VMD) {
2643 char buf[PATH_MAX];
60f0f54d
PB		 2644 printf(" I/O Controller : %s (%s)\n",
2645 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2646 if (print_nvme_info(hba)) {
b5eece69
PB		 2647 if (verbose > 0)
2648 pr_err("failed to get devices attached to VMD domain.\n");
2649 result |= 2;
2650 }
60f0f54d
PB		 2651 }
2652 }
07cb1e57 2653 printf("\n");
60f0f54d
PB		 2654 continue;
2655 }
0858eccf 2656
60f0f54d
PB		 2657 print_imsm_capability(&entry->orom);
2658 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2659 for (hba = list; hba; hba = hba->next) {
2660 if (hba->type == SYS_DEV_NVME)
6da53c0e 2661 print_nvme_info(hba);
0858eccf 2662 }
60f0f54d 2663 printf("\n");
0858eccf
AP		 2664 continue;
2665 }
2666
2667 struct devid_list *devid;
5e1d6128 2668 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2669 hba = device_by_id(devid->devid);
2670 if (!hba)
2671 continue;
2672
9eafa1de
MN		 2673 printf(" I/O Controller : %s (%s)\n",
2674 hba->path, get_sys_dev_type(hba->type));
2675 if (hba->type == SYS_DEV_SATA) {
2676 host_base = ahci_get_port_count(hba->path, &port_count);
2677 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2678 if (verbose > 0)
7a862a02 2679 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2680 result |= 2;
2681 }
120dc887
LM		 2682 }
2683 }
0858eccf 2684 printf("\n");
d665cc31 2685 }
155cbb4c 2686
120dc887 2687 return result;
d665cc31 2688}
e50cf220 2689
9eafa1de 2690static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2691{
e50cf220
MN		 2692 struct sys_dev *list, *hba;
2693 int result=1;
2694
2695 list = find_intel_devices();
2696 if (!list) {
2697 if (verbose > 0)
2698 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2699 result = 2;
e50cf220
MN		 2700 return result;
2701 }
2702
2703 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2704 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2705 continue;
60f0f54d
PB		 2706 if (!find_imsm_capability(hba) && verbose > 0) {
2707 char buf[PATH_MAX];
2708 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2709 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2710 }
0858eccf 2711 else
e50cf220 2712 result = 0;
e50cf220
MN		 2713 }
2714
5e1d6128 2715 const struct orom_entry *entry;
0858eccf 2716
60f0f54d
PB		 2717 for (entry = orom_entries; entry; entry = entry->next) {
2718 if (entry->type == SYS_DEV_VMD) {
2719 for (hba = list; hba; hba = hba->next)
2720 print_imsm_capability_export(&entry->orom);
2721 continue;
2722 }
5e1d6128 2723 print_imsm_capability_export(&entry->orom);
60f0f54d 2724 }
0858eccf 2725
e50cf220
MN		 2726 return result;
2727}
2728
cdddbdbc
DW		 2729static int match_home_imsm(struct supertype *st, char *homehost)
2730{
5115ca67
DW		 2731 /* the imsm metadata format does not specify any host
2732 * identification information. We return -1 since we can never
2733 * confirm nor deny whether a given array is "meant" for this
148acb7b 2734 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2735 * exclude member disks that do not belong, and we rely on
2736 * mdadm.conf to specify the arrays that should be assembled.
2737 * Auto-assembly may still pick up "foreign" arrays.
2738 */
cdddbdbc 2739
9362c1c8 2740 return -1;
cdddbdbc
DW		 2741}
2742
2743static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2744{
51006d85
N		 2745 /* The uuid returned here is used for:
2746 * uuid to put into bitmap file (Create, Grow)
2747 * uuid for backup header when saving critical section (Grow)
2748 * comparing uuids when re-adding a device into an array
2749 * In these cases the uuid required is that of the data-array,
2750 * not the device-set.
2751 * uuid to recognise same set when adding a missing device back
2752 * to an array. This is a uuid for the device-set.
1011e834 2753 *
51006d85
N		 2754 * For each of these we can make do with a truncated
2755 * or hashed uuid rather than the original, as long as
2756 * everyone agrees.
2757 * In each case the uuid required is that of the data-array,
2758 * not the device-set.
43dad3d6 2759 */
51006d85
N		 2760 /* imsm does not track uuid's so we synthesis one using sha1 on
2761 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2762 * - the orig_family_num of the container
51006d85
N		 2763 * - the index number of the volume
2764 * - the 'serial' number of the volume.
2765 * Hopefully these are all constant.
2766 */
2767 struct intel_super *super = st->sb;
43dad3d6 2768
51006d85
N		 2769 char buf[20];
2770 struct sha1_ctx ctx;
2771 struct imsm_dev *dev = NULL;
148acb7b 2772 __u32 family_num;
51006d85 2773
148acb7b
DW		 2774 /* some mdadm versions failed to set ->orig_family_num, in which
2775 * case fall back to ->family_num. orig_family_num will be
2776 * fixed up with the first metadata update.
2777 */
2778 family_num = super->anchor->orig_family_num;
2779 if (family_num == 0)
2780 family_num = super->anchor->family_num;
51006d85 2781 sha1_init_ctx(&ctx);
92bd8f8d 2782 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2783 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2784 if (super->current_vol >= 0)
2785 dev = get_imsm_dev(super, super->current_vol);
2786 if (dev) {
2787 __u32 vol = super->current_vol;
2788 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2789 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2790 }
2791 sha1_finish_ctx(&ctx, buf);
2792 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2793}
2794
0d481d37 2795#if 0
4f5bc454
DW		 2796static void
2797get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2798{
cdddbdbc
DW		 2799 __u8 *v = get_imsm_version(mpb);
2800 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2801 char major[] = { 0, 0, 0 };
2802 char minor[] = { 0 ,0, 0 };
2803 char patch[] = { 0, 0, 0 };
2804 char *ver_parse[] = { major, minor, patch };
2805 int i, j;
2806
2807 i = j = 0;
2808 while (*v != '\0' && v < end) {
2809 if (*v != '.' && j < 2)
2810 ver_parse[i][j++] = *v;
2811 else {
2812 i++;
2813 j = 0;
2814 }
2815 v++;
2816 }
2817
4f5bc454
DW		 2818 *m = strtol(minor, NULL, 0);
2819 *p = strtol(patch, NULL, 0);
2820}
0d481d37 2821#endif
4f5bc454 2822
1e5c6983
DW		 2823static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2824{
2825 /* migr_strip_size when repairing or initializing parity */
238c0a71 2826 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2827 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2828
2829 switch (get_imsm_raid_level(map)) {
2830 case 5:
2831 case 10:
2832 return chunk;
2833 default:
2834 return 128*1024 >> 9;
2835 }
2836}
2837
2838static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2839{
2840 /* migr_strip_size when rebuilding a degraded disk, no idea why
2841 * this is different than migr_strip_size_resync(), but it's good
2842 * to be compatible
2843 */
238c0a71 2844 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2845 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2846
2847 switch (get_imsm_raid_level(map)) {
2848 case 1:
2849 case 10:
2850 if (map->num_members % map->num_domains == 0)
2851 return 128*1024 >> 9;
2852 else
2853 return chunk;
2854 case 5:
2855 return max((__u32) 64*1024 >> 9, chunk);
2856 default:
2857 return 128*1024 >> 9;
2858 }
2859}
2860
2861static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2862{
238c0a71
AK		 2863 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2864 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2865 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2866 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2867
2868 return max((__u32) 1, hi_chunk / lo_chunk);
2869}
2870
2871static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2872{
238c0a71 2873 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2874 int level = get_imsm_raid_level(lo);
2875
2876 if (level == 1 || level == 10) {
238c0a71 2877 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2878
2879 return hi->num_domains;
2880 } else
2881 return num_stripes_per_unit_resync(dev);
2882}
2883
9529d343 2884static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2885{
2886 /* named 'imsm_' because raid0, raid1 and raid10
2887 * counter-intuitively have the same number of data disks
2888 */
1e5c6983
DW		 2889 switch (get_imsm_raid_level(map)) {
2890 case 0:
36fd8ccc
AK		 2891 return map->num_members;
2892 break;
1e5c6983
DW		 2893 case 1:
2894 case 10:
36fd8ccc 2895 return map->num_members/2;
1e5c6983
DW		 2896 case 5:
2897 return map->num_members - 1;
2898 default:
1ade5cc1 2899 dprintf("unsupported raid level\n");
1e5c6983
DW		 2900 return 0;
2901 }
2902}
2903
44490938
MD		 2904static unsigned long long calc_component_size(struct imsm_map *map,
2905 struct imsm_dev *dev)
2906{
2907 unsigned long long component_size;
2908 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2909 long long calc_dev_size = 0;
44490938
MD		 2910 unsigned int member_disks = imsm_num_data_members(map);
2911
2912 if (member_disks == 0)
2913 return 0;
2914
2915 component_size = per_dev_array_size(map);
2916 calc_dev_size = component_size * member_disks;
2917
2918 /* Component size is rounded to 1MB so difference between size from
2919 * metadata and size calculated from num_data_stripes equals up to
2920 * 2048 blocks per each device. If the difference is higher it means
2921 * that array size was expanded and num_data_stripes was not updated.
2922 */
a4f7290c 2923 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2924 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2925 component_size = dev_size / member_disks;
2926 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2927 component_size / map->blocks_per_strip,
2928 num_data_stripes(map));
2929 }
2930
2931 return component_size;
2932}
2933
1e5c6983
DW		 2934static __u32 parity_segment_depth(struct imsm_dev *dev)
2935{
238c0a71 2936 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2937 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2938
2939 switch(get_imsm_raid_level(map)) {
2940 case 1:
2941 case 10:
2942 return chunk * map->num_domains;
2943 case 5:
2944 return chunk * map->num_members;
2945 default:
2946 return chunk;
2947 }
2948}
2949
2950static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2951{
238c0a71 2952 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2953 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2954 __u32 strip = block / chunk;
2955
2956 switch (get_imsm_raid_level(map)) {
2957 case 1:
2958 case 10: {
2959 __u32 vol_strip = (strip * map->num_domains) + 1;
2960 __u32 vol_stripe = vol_strip / map->num_members;
2961
2962 return vol_stripe * chunk + block % chunk;
2963 } case 5: {
2964 __u32 stripe = strip / (map->num_members - 1);
2965
2966 return stripe * chunk + block % chunk;
2967 }
2968 default:
2969 return 0;
2970 }
2971}
2972
c47b0ff6
AK		 2973static __u64 blocks_per_migr_unit(struct intel_super *super,
2974 struct imsm_dev *dev)
1e5c6983
DW		 2975{
2976 /* calculate the conversion factor between per member 'blocks'
2977 * (md/{resync,rebuild}_start) and imsm migration units, return
2978 * 0 for the 'not migrating' and 'unsupported migration' cases
2979 */
2980 if (!dev->vol.migr_state)
2981 return 0;
2982
2983 switch (migr_type(dev)) {
c47b0ff6
AK		 2984 case MIGR_GEN_MIGR: {
2985 struct migr_record *migr_rec = super->migr_rec;
2986 return __le32_to_cpu(migr_rec->blocks_per_unit);
2987 }
1e5c6983
DW		 2988 case MIGR_VERIFY:
2989 case MIGR_REPAIR:
2990 case MIGR_INIT: {
238c0a71 2991 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2992 __u32 stripes_per_unit;
2993 __u32 blocks_per_unit;
2994 __u32 parity_depth;
2995 __u32 migr_chunk;
2996 __u32 block_map;
2997 __u32 block_rel;
2998 __u32 segment;
2999 __u32 stripe;
3000 __u8 disks;
3001
3002 /* yes, this is really the translation of migr_units to
3003 * per-member blocks in the 'resync' case
3004 */
3005 stripes_per_unit = num_stripes_per_unit_resync(dev);
3006 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 3007 disks = imsm_num_data_members(map);
1e5c6983 3008 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 3009 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 3010 segment = blocks_per_unit / stripe;
3011 block_rel = blocks_per_unit - segment * stripe;
3012 parity_depth = parity_segment_depth(dev);
3013 block_map = map_migr_block(dev, block_rel);
3014 return block_map + parity_depth * segment;
3015 }
3016 case MIGR_REBUILD: {
3017 __u32 stripes_per_unit;
3018 __u32 migr_chunk;
3019
3020 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
3021 migr_chunk = migr_strip_blocks_rebuild(dev);
3022 return migr_chunk * stripes_per_unit;
3023 }
1e5c6983
DW		 3024 case MIGR_STATE_CHANGE:
3025 default:
3026 return 0;
3027 }
3028}
3029
c2c087e6
DW		 3030static int imsm_level_to_layout(int level)
3031{
3032 switch (level) {
3033 case 0:
3034 case 1:
3035 return 0;
3036 case 5:
3037 case 6:
a380c027 3038 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 3039 case 10:
c92a2527 3040 return 0x102;
c2c087e6 3041 }
a18a888e 3042 return UnSet;
c2c087e6
DW		 3043}
3044
8e59f3d8
AK		 3045/*******************************************************************************
3046 * Function: read_imsm_migr_rec
3047 * Description: Function reads imsm migration record from last sector of disk
3048 * Parameters:
3049 * fd : disk descriptor
3050 * super : metadata info
3051 * Returns:
3052 * 0 : success,
3053 * -1 : fail
3054 ******************************************************************************/
3055static int read_imsm_migr_rec(int fd, struct intel_super *super)
3056{
3057 int ret_val = -1;
de44e46f 3058 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3059 unsigned long long dsize;
3060
3061 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3062 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3063 SEEK_SET) < 0) {
e7b84f9d
N		 3064 pr_err("Cannot seek to anchor block: %s\n",
3065 strerror(errno));
8e59f3d8
AK		 3066 goto out;
3067 }
466070ad 3068 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3069 MIGR_REC_BUF_SECTORS*sector_size) !=
3070 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3071 pr_err("Cannot read migr record block: %s\n",
3072 strerror(errno));
8e59f3d8
AK		 3073 goto out;
3074 }
3075 ret_val = 0;
de44e46f
PB		 3076 if (sector_size == 4096)
3077 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3078
3079out:
3080 return ret_val;
3081}
3082
3136abe5
AK		 3083static struct imsm_dev *imsm_get_device_during_migration(
3084 struct intel_super *super)
3085{
3086
3087 struct intel_dev *dv;
3088
3089 for (dv = super->devlist; dv; dv = dv->next) {
3090 if (is_gen_migration(dv->dev))
3091 return dv->dev;
3092 }
3093 return NULL;
3094}
3095
8e59f3d8
AK		 3096/*******************************************************************************
3097 * Function: load_imsm_migr_rec
3098 * Description: Function reads imsm migration record (it is stored at the last
3099 * sector of disk)
3100 * Parameters:
3101 * super : imsm internal array info
3102 * info : general array info
3103 * Returns:
3104 * 0 : success
3105 * -1 : fail
4c965cc9 3106 * -2 : no migration in progress
8e59f3d8
AK		 3107 ******************************************************************************/
3108static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3109{
3110 struct mdinfo *sd;
594dc1b8 3111 struct dl *dl;
8e59f3d8
AK		 3112 char nm[30];
3113 int retval = -1;
3114 int fd = -1;
3136abe5 3115 struct imsm_dev *dev;
594dc1b8 3116 struct imsm_map *map;
b4ab44d8 3117 int slot = -1;
3136abe5
AK		 3118
3119 /* find map under migration */
3120 dev = imsm_get_device_during_migration(super);
3121 /* nothing to load,no migration in progress?
3122 */
3123 if (dev == NULL)
4c965cc9 3124 return -2;
8e59f3d8
AK		 3125
3126 if (info) {
3127 for (sd = info->devs ; sd ; sd = sd->next) {
3128 /* read only from one of the first two slots */
12fe93e9
TM		 3129 if ((sd->disk.raid_disk < 0) ||
3130 (sd->disk.raid_disk > 1))
8e59f3d8 3131 continue;
3136abe5 3132
8e59f3d8
AK		 3133 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3134 fd = dev_open(nm, O_RDONLY);
3135 if (fd >= 0)
3136 break;
3137 }
3138 }
3139 if (fd < 0) {
12fe93e9 3140 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3141 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3142 /* skip spare and failed disks
3143 */
3144 if (dl->index < 0)
3145 continue;
8e59f3d8 3146 /* read only from one of the first two slots */
3136abe5
AK		 3147 if (map)
3148 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3149 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3150 continue;
3151 sprintf(nm, "%d:%d", dl->major, dl->minor);
3152 fd = dev_open(nm, O_RDONLY);
3153 if (fd >= 0)
3154 break;
3155 }
3156 }
3157 if (fd < 0)
3158 goto out;
3159 retval = read_imsm_migr_rec(fd, super);
3160
3161out:
3162 if (fd >= 0)
3163 close(fd);
3164 return retval;
3165}
3166
c17608ea
AK		 3167/*******************************************************************************
3168 * function: imsm_create_metadata_checkpoint_update
3169 * Description: It creates update for checkpoint change.
3170 * Parameters:
3171 * super : imsm internal array info
3172 * u : pointer to prepared update
3173 * Returns:
3174 * Uptate length.
3175 * If length is equal to 0, input pointer u contains no update
3176 ******************************************************************************/
3177static int imsm_create_metadata_checkpoint_update(
3178 struct intel_super *super,
3179 struct imsm_update_general_migration_checkpoint **u)
3180{
3181
3182 int update_memory_size = 0;
3183
1ade5cc1 3184 dprintf("(enter)\n");
c17608ea
AK		 3185
3186 if (u == NULL)
3187 return 0;
3188 *u = NULL;
3189
3190 /* size of all update data without anchor */
3191 update_memory_size =
3192 sizeof(struct imsm_update_general_migration_checkpoint);
3193
503975b9 3194 *u = xcalloc(1, update_memory_size);
c17608ea 3195 if (*u == NULL) {
1ade5cc1 3196 dprintf("error: cannot get memory\n");
c17608ea
AK		 3197 return 0;
3198 }
3199 (*u)->type = update_general_migration_checkpoint;
9f421827 3200 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3201 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3202
3203 return update_memory_size;
3204}
3205
c17608ea
AK		 3206static void imsm_update_metadata_locally(struct supertype *st,
3207 void *buf, int len);
3208
687629c2
AK		 3209/*******************************************************************************
3210 * Function: write_imsm_migr_rec
3211 * Description: Function writes imsm migration record
3212 * (at the last sector of disk)
3213 * Parameters:
3214 * super : imsm internal array info
3215 * Returns:
3216 * 0 : success
3217 * -1 : if fail
3218 ******************************************************************************/
3219static int write_imsm_migr_rec(struct supertype *st)
3220{
3221 struct intel_super *super = st->sb;
de44e46f 3222 unsigned int sector_size = super->sector_size;
687629c2
AK		 3223 unsigned long long dsize;
3224 char nm[30];
3225 int fd = -1;
3226 int retval = -1;
3227 struct dl *sd;
c17608ea
AK		 3228 int len;
3229 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3230 struct imsm_dev *dev;
594dc1b8 3231 struct imsm_map *map;
3136abe5
AK		 3232
3233 /* find map under migration */
3234 dev = imsm_get_device_during_migration(super);
3235 /* if no migration, write buffer anyway to clear migr_record
3236 * on disk based on first available device
3237 */
3238 if (dev == NULL)
3239 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3240 super->current_vol);
3241
44bfe6df 3242 map = get_imsm_map(dev, MAP_0);
687629c2 3243
de44e46f
PB		 3244 if (sector_size == 4096)
3245 convert_to_4k_imsm_migr_rec(super);
687629c2 3246 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3247 int slot = -1;
3136abe5
AK		 3248
3249 /* skip failed and spare devices */
3250 if (sd->index < 0)
3251 continue;
687629c2 3252 /* write to 2 first slots only */
3136abe5
AK		 3253 if (map)
3254 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3255 if (map == NULL || slot > 1 || slot < 0)
687629c2 3256 continue;
3136abe5 3257
687629c2
AK		 3258 sprintf(nm, "%d:%d", sd->major, sd->minor);
3259 fd = dev_open(nm, O_RDWR);
3260 if (fd < 0)
3261 continue;
3262 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3263 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3264 SEEK_SET) < 0) {
e7b84f9d
N		 3265 pr_err("Cannot seek to anchor block: %s\n",
3266 strerror(errno));
687629c2
AK		 3267 goto out;
3268 }
466070ad 3269 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3270 MIGR_REC_BUF_SECTORS*sector_size) !=
3271 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3272 pr_err("Cannot write migr record block: %s\n",
3273 strerror(errno));
687629c2
AK		 3274 goto out;
3275 }
3276 close(fd);
3277 fd = -1;
3278 }
de44e46f
PB		 3279 if (sector_size == 4096)
3280 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3281 /* update checkpoint information in metadata */
3282 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3283 if (len <= 0) {
3284 dprintf("imsm: Cannot prepare update\n");
3285 goto out;
3286 }
3287 /* update metadata locally */
3288 imsm_update_metadata_locally(st, u, len);
3289 /* and possibly remotely */
3290 if (st->update_tail) {
3291 append_metadata_update(st, u, len);
3292 /* during reshape we do all work inside metadata handler
3293 * manage_reshape(), so metadata update has to be triggered
3294 * insida it
3295 */
3296 flush_metadata_updates(st);
3297 st->update_tail = &st->updates;
3298 } else
3299 free(u);
687629c2
AK		 3300
3301 retval = 0;
3302 out:
3303 if (fd >= 0)
3304 close(fd);
3305 return retval;
3306}
3307
e2962bfc
AK		 3308/* spare/missing disks activations are not allowe when
3309 * array/container performs reshape operation, because
3310 * all arrays in container works on the same disks set
3311 */
3312int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3313{
3314 int rv = 0;
3315 struct intel_dev *i_dev;
3316 struct imsm_dev *dev;
3317
3318 /* check whole container
3319 */
3320 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3321 dev = i_dev->dev;
3ad25638 3322 if (is_gen_migration(dev)) {
e2962bfc
AK		 3323 /* No repair during any migration in container
3324 */
3325 rv = 1;
3326 break;
3327 }
3328 }
3329 return rv;
3330}
3e684231 3331static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3332 int chunk_size,
f36a9ecd 3333 unsigned int sector_size,
c41e00b2
AK		 3334 unsigned long long component_size)
3335{
3e684231 3336 unsigned int component_size_alignment;
c41e00b2 3337
3e684231 3338 /* check component size alignment
c41e00b2 3339 */
3e684231 3340 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3341
3e684231 3342 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3343 level, chunk_size, component_size,
3e684231 3344 component_size_alignment);
c41e00b2 3345
3e684231
MZ		 3346 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3347 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3348 component_size);
3e684231 3349 component_size -= component_size_alignment;
1ade5cc1 3350 dprintf_cont("to %llu (%i).\n",
3e684231 3351 component_size, component_size_alignment);
c41e00b2
AK		 3352 }
3353
3354 return component_size;
3355}
e2962bfc 3356
2432ce9b
AP		 3357static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3358{
3359 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3360 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3361
3362 return pba_of_lba0(map) +
3363 (num_data_stripes(map) * map->blocks_per_strip);
3364}
3365
a5d85af7 3366static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3367{
3368 struct intel_super *super = st->sb;
c47b0ff6 3369 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3370 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3371 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3372 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3373 struct imsm_map *map_to_analyse = map;
efb30e7f 3374 struct dl *dl;
a5d85af7 3375 int map_disks = info->array.raid_disks;
bf5a934a 3376
95eeceeb 3377 memset(info, 0, sizeof(*info));
b335e593
AK		 3378 if (prev_map)
3379 map_to_analyse = prev_map;
3380
ca0748fa 3381 dl = super->current_disk;
9894ec0d 3382
bf5a934a 3383 info->container_member = super->current_vol;
cd0430a1 3384 info->array.raid_disks = map->num_members;
b335e593 3385 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3386 info->array.layout = imsm_level_to_layout(info->array.level);
3387 info->array.md_minor = -1;
3388 info->array.ctime = 0;
3389 info->array.utime = 0;
b335e593
AK		 3390 info->array.chunk_size =
3391 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3392 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3393 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3394 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3395
3f510843 3396 if (is_gen_migration(dev)) {
3f83228a 3397 info->reshape_active = 1;
b335e593
AK		 3398 info->new_level = get_imsm_raid_level(map);
3399 info->new_layout = imsm_level_to_layout(info->new_level);
3400 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3401 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3402 if (info->delta_disks) {
3403 /* this needs to be applied to every array
3404 * in the container.
3405 */
81219e70 3406 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3407 }
3f83228a
N		 3408 /* We shape information that we give to md might have to be
3409 * modify to cope with md's requirement for reshaping arrays.
3410 * For example, when reshaping a RAID0, md requires it to be
3411 * presented as a degraded RAID4.
3412 * Also if a RAID0 is migrating to a RAID5 we need to specify
3413 * the array as already being RAID5, but the 'before' layout
3414 * is a RAID4-like layout.
3415 */
3416 switch (info->array.level) {
3417 case 0:
3418 switch(info->new_level) {
3419 case 0:
3420 /* conversion is happening as RAID4 */
3421 info->array.level = 4;
3422 info->array.raid_disks += 1;
3423 break;
3424 case 5:
3425 /* conversion is happening as RAID5 */
3426 info->array.level = 5;
3427 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3428 info->delta_disks -= 1;
3429 break;
3430 default:
3431 /* FIXME error message */
3432 info->array.level = UnSet;
3433 break;
3434 }
3435 break;
3436 }
b335e593
AK		 3437 } else {
3438 info->new_level = UnSet;
3439 info->new_layout = UnSet;
3440 info->new_chunk = info->array.chunk_size;
3f83228a 3441 info->delta_disks = 0;
b335e593 3442 }
ca0748fa 3443
efb30e7f
DW		 3444 if (dl) {
3445 info->disk.major = dl->major;
3446 info->disk.minor = dl->minor;
ca0748fa 3447 info->disk.number = dl->index;
656b6b5a
N		 3448 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3449 dl->index);
efb30e7f 3450 }
bf5a934a 3451
5551b113 3452 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3453 info->component_size = calc_component_size(map, dev);
3e684231 3454 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3455 info->array.level,
3456 info->array.chunk_size,
f36a9ecd 3457 super->sector_size,
c41e00b2 3458 info->component_size);
5e46202e 3459 info->bb.supported = 1;
139dae11 3460
301406c9 3461 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3462 info->recovery_start = MaxSector;
bf5a934a 3463
c2462068
PB		 3464 if (info->array.level == 5 &&
3465 (dev->rwh_policy == RWH_DISTRIBUTED ||
3466 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3467 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3468 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3469 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3470 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3471 else
3472 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3473 >> 9;
2432ce9b
AP		 3474 } else if (info->array.level <= 0) {
3475 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3476 } else {
3477 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3478 }
3479
d2e6d5d6 3480 info->reshape_progress = 0;
b6796ce1 3481 info->resync_start = MaxSector;
b9172665 3482 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3483 !(info->array.state & 1)) &&
b9172665 3484 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3485 info->resync_start = 0;
b6796ce1
AK		 3486 }
3487 if (dev->vol.migr_state) {
1e5c6983
DW		 3488 switch (migr_type(dev)) {
3489 case MIGR_REPAIR:
3490 case MIGR_INIT: {
c47b0ff6
AK		 3491 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3492 dev);
1e5c6983
DW		 3493 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3494
3495 info->resync_start = blocks_per_unit * units;
3496 break;
3497 }
d2e6d5d6 3498 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3499 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3500 dev);
9f421827 3501 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3502 unsigned long long array_blocks;
3503 int used_disks;
d2e6d5d6 3504
befb629b
AK		 3505 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3506 (units <
9f421827 3507 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3508 (super->migr_rec->rec_status ==
3509 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3510 units++;
3511
d2e6d5d6 3512 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3513
7a862a02 3514 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3515 (unsigned long long)units,
3516 (unsigned long long)blocks_per_unit,
3517 info->reshape_progress);
75156c46 3518
9529d343 3519 used_disks = imsm_num_data_members(prev_map);
75156c46 3520 if (used_disks > 0) {
44490938 3521 array_blocks = per_dev_array_size(map) *
75156c46 3522 used_disks;
b53bfba6
TM		 3523 info->custom_array_size =
3524 round_size_to_mb(array_blocks,
3525 used_disks);
3526
75156c46 3527 }
d2e6d5d6 3528 }
1e5c6983
DW		 3529 case MIGR_VERIFY:
3530 /* we could emulate the checkpointing of
3531 * 'sync_action=check' migrations, but for now
3532 * we just immediately complete them
3533 */
3534 case MIGR_REBUILD:
3535 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3536 case MIGR_STATE_CHANGE:
3537 /* FIXME handle other migrations */
3538 default:
3539 /* we are not dirty, so... */
3540 info->resync_start = MaxSector;
3541 }
b6796ce1 3542 }
301406c9
DW		 3543
3544 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3545 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3546
f35f2525
N		 3547 info->array.major_version = -1;
3548 info->array.minor_version = -2;
4dd2df09 3549 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3550 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3551 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3552
3553 if (dmap) {
3554 int i, j;
3555 for (i=0; i<map_disks; i++) {
3556 dmap[i] = 0;
3557 if (i < info->array.raid_disks) {
3558 struct imsm_disk *dsk;
238c0a71 3559 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3560 dsk = get_imsm_disk(super, j);
3561 if (dsk && (dsk->status & CONFIGURED_DISK))
3562 dmap[i] = 1;
3563 }
3564 }
3565 }
81ac8b4d 3566}
bf5a934a 3567
3b451610
AK		 3568static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3569 int failed, int look_in_map);
3570
3571static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3572 int look_in_map);
3573
3574static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3575{
3576 if (is_gen_migration(dev)) {
3577 int failed;
3578 __u8 map_state;
3579 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3580
3581 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3582 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3583 if (map2->map_state != map_state) {
3584 map2->map_state = map_state;
3585 super->updates_pending++;
3586 }
3587 }
3588}
97b4d0e9
DW		 3589
3590static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3591{
3592 struct dl *d;
3593
3594 for (d = super->missing; d; d = d->next)
3595 if (d->index == index)
3596 return &d->disk;
3597 return NULL;
3598}
3599
a5d85af7 3600static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3601{
3602 struct intel_super *super = st->sb;
4f5bc454 3603 struct imsm_disk *disk;
a5d85af7 3604 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3605 int max_enough = -1;
3606 int i;
3607 struct imsm_super *mpb;
4f5bc454 3608
bf5a934a 3609 if (super->current_vol >= 0) {
a5d85af7 3610 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3611 return;
3612 }
95eeceeb 3613 memset(info, 0, sizeof(*info));
d23fe947
DW		 3614
3615 /* Set raid_disks to zero so that Assemble will always pull in valid
3616 * spares
3617 */
3618 info->array.raid_disks = 0;
cdddbdbc
DW		 3619 info->array.level = LEVEL_CONTAINER;
3620 info->array.layout = 0;
3621 info->array.md_minor = -1;
1011e834 3622 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3623 info->array.utime = 0;
3624 info->array.chunk_size = 0;
3625
3626 info->disk.major = 0;
3627 info->disk.minor = 0;
cdddbdbc 3628 info->disk.raid_disk = -1;
c2c087e6 3629 info->reshape_active = 0;
f35f2525
N		 3630 info->array.major_version = -1;
3631 info->array.minor_version = -2;
c2c087e6 3632 strcpy(info->text_version, "imsm");
a67dd8cc 3633 info->safe_mode_delay = 0;
c2c087e6
DW		 3634 info->disk.number = -1;
3635 info->disk.state = 0;
c5afc314 3636 info->name[0] = 0;
921d9e16 3637 info->recovery_start = MaxSector;
3ad25638 3638 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3639 info->bb.supported = 1;
c2c087e6 3640
97b4d0e9 3641 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3642 mpb = super->anchor;
b7d81a38 3643 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3644
ab3cb6b3
N		 3645 for (i = 0; i < mpb->num_raid_devs; i++) {
3646 struct imsm_dev *dev = get_imsm_dev(super, i);
3647 int failed, enough, j, missing = 0;
3648 struct imsm_map *map;
3649 __u8 state;
97b4d0e9 3650
3b451610
AK		 3651 failed = imsm_count_failed(super, dev, MAP_0);
3652 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3653 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3654
3655 /* any newly missing disks?
3656 * (catches single-degraded vs double-degraded)
3657 */
3658 for (j = 0; j < map->num_members; j++) {
238c0a71 3659 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3660 __u32 idx = ord_to_idx(ord);
3661
20dc76d1
MT		 3662 if (super->disks && super->disks->index == (int)idx)
3663 info->disk.raid_disk = j;
3664
ab3cb6b3
N		 3665 if (!(ord & IMSM_ORD_REBUILD) &&
3666 get_imsm_missing(super, idx)) {
3667 missing = 1;
3668 break;
3669 }
97b4d0e9 3670 }
ab3cb6b3
N		 3671
3672 if (state == IMSM_T_STATE_FAILED)
3673 enough = -1;
3674 else if (state == IMSM_T_STATE_DEGRADED &&
3675 (state != map->map_state || missing))
3676 enough = 0;
3677 else /* we're normal, or already degraded */
3678 enough = 1;
d2bde6d3
AK		 3679 if (is_gen_migration(dev) && missing) {
3680 /* during general migration we need all disks
3681 * that process is running on.
3682 * No new missing disk is allowed.
3683 */
3684 max_enough = -1;
3685 enough = -1;
3686 /* no more checks necessary
3687 */
3688 break;
3689 }
ab3cb6b3
N		 3690 /* in the missing/failed disk case check to see
3691 * if at least one array is runnable
3692 */
3693 max_enough = max(max_enough, enough);
3694 }
1ade5cc1 3695 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3696 info->container_enough = max_enough;
97b4d0e9 3697
4a04ec6c 3698 if (super->disks) {
14e8215b
DW		 3699 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3700
b9f594fe 3701 disk = &super->disks->disk;
5551b113 3702 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3703 info->component_size = reserved;
25ed7e59 3704 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3705 /* we don't change info->disk.raid_disk here because
3706 * this state will be finalized in mdmon after we have
3707 * found the 'most fresh' version of the metadata
3708 */
25ed7e59 3709 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3710 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3711 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3712 }
a575e2a7
DW		 3713
3714 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3715 * ->compare_super may have updated the 'num_raid_devs' field for spares
3716 */
3717 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3718 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3719 else
3720 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3721
3722 /* I don't know how to compute 'map' on imsm, so use safe default */
3723 if (map) {
3724 int i;
3725 for (i = 0; i < map_disks; i++)
3726 map[i] = 1;
3727 }
3728
cdddbdbc
DW		 3729}
3730
5c4cd5da
AC		 3731/* allocates memory and fills disk in mdinfo structure
3732 * for each disk in array */
3733struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3734{
594dc1b8 3735 struct mdinfo *mddev;
5c4cd5da
AC		 3736 struct intel_super *super = st->sb;
3737 struct imsm_disk *disk;
3738 int count = 0;
3739 struct dl *dl;
3740 if (!super || !super->disks)
3741 return NULL;
3742 dl = super->disks;
503975b9 3743 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3744 while (dl) {
3745 struct mdinfo *tmp;
3746 disk = &dl->disk;
503975b9 3747 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3748 if (mddev->devs)
3749 tmp->next = mddev->devs;
3750 mddev->devs = tmp;
3751 tmp->disk.number = count++;
3752 tmp->disk.major = dl->major;
3753 tmp->disk.minor = dl->minor;
3754 tmp->disk.state = is_configured(disk) ?
3755 (1 << MD_DISK_ACTIVE) : 0;
3756 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3757 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3758 tmp->disk.raid_disk = -1;
3759 dl = dl->next;
3760 }
3761 return mddev;
3762}
3763
cdddbdbc
DW		 3764static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3765 char *update, char *devname, int verbose,
3766 int uuid_set, char *homehost)
3767{
f352c545
DW		 3768 /* For 'assemble' and 'force' we need to return non-zero if any
3769 * change was made. For others, the return value is ignored.
3770 * Update options are:
3771 * force-one : This device looks a bit old but needs to be included,
3772 * update age info appropriately.
3773 * assemble: clear any 'faulty' flag to allow this device to
3774 * be assembled.
3775 * force-array: Array is degraded but being forced, mark it clean
3776 * if that will be needed to assemble it.
3777 *
3778 * newdev: not used ????
3779 * grow: Array has gained a new device - this is currently for
3780 * linear only
3781 * resync: mark as dirty so a resync will happen.
3782 * name: update the name - preserving the homehost
6e46bf34 3783 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3784 *
3785 * Following are not relevant for this imsm:
3786 * sparc2.2 : update from old dodgey metadata
3787 * super-minor: change the preferred_minor number
3788 * summaries: update redundant counters.
f352c545
DW		 3789 * homehost: update the recorded homehost
3790 * _reshape_progress: record new reshape_progress position.
3791 */
6e46bf34
DW		 3792 int rv = 1;
3793 struct intel_super *super = st->sb;
3794 struct imsm_super *mpb;
f352c545 3795
6e46bf34
DW		 3796 /* we can only update container info */
3797 if (!super || super->current_vol >= 0 || !super->anchor)
3798 return 1;
3799
3800 mpb = super->anchor;
3801
81a5b4f5
N		 3802 if (strcmp(update, "uuid") == 0) {
3803 /* We take this to mean that the family_num should be updated.
3804 * However that is much smaller than the uuid so we cannot really
3805 * allow an explicit uuid to be given. And it is hard to reliably
3806 * know if one was.
3807 * So if !uuid_set we know the current uuid is random and just used
3808 * the first 'int' and copy it to the other 3 positions.
3809 * Otherwise we require the 4 'int's to be the same as would be the
3810 * case if we are using a random uuid. So an explicit uuid will be
3811 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3812 */
81a5b4f5
N		 3813 if (!uuid_set) {
3814 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3815 rv = 0;
81a5b4f5
N		 3816 } else {
3817 if (info->uuid[0] != info->uuid[1] ||
3818 info->uuid[1] != info->uuid[2] ||
3819 info->uuid[2] != info->uuid[3])
3820 rv = -1;
3821 else
3822 rv = 0;
6e46bf34 3823 }
81a5b4f5
N		 3824 if (rv == 0)
3825 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3826 } else if (strcmp(update, "assemble") == 0)
3827 rv = 0;
3828 else
1e2b2765 3829 rv = -1;
f352c545 3830
6e46bf34
DW		 3831 /* successful update? recompute checksum */
3832 if (rv == 0)
3833 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3834
3835 return rv;
cdddbdbc
DW		 3836}
3837
c2c087e6 3838static size_t disks_to_mpb_size(int disks)
cdddbdbc 3839{
c2c087e6 3840 size_t size;
cdddbdbc 3841
c2c087e6
DW		 3842 size = sizeof(struct imsm_super);
3843 size += (disks - 1) * sizeof(struct imsm_disk);
3844 size += 2 * sizeof(struct imsm_dev);
3845 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3846 size += (4 - 2) * sizeof(struct imsm_map);
3847 /* 4 possible disk_ord_tbl's */
3848 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3849 /* maximum bbm log */
3850 size += sizeof(struct bbm_log);
c2c087e6
DW		 3851
3852 return size;
3853}
3854
387fcd59
N		 3855static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3856 unsigned long long data_offset)
c2c087e6
DW		 3857{
3858 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3859 return 0;
3860
3861 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3862}
3863
ba2de7ba
DW		 3864static void free_devlist(struct intel_super *super)
3865{
3866 struct intel_dev *dv;
3867
3868 while (super->devlist) {
3869 dv = super->devlist->next;
3870 free(super->devlist->dev);
3871 free(super->devlist);
3872 super->devlist = dv;
3873 }
3874}
3875
3876static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3877{
3878 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3879}
3880
cdddbdbc
DW		 3881static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3882{
3883 /*
3884 * return:
3885 * 0 same, or first was empty, and second was copied
3886 * 1 second had wrong number
3887 * 2 wrong uuid
3888 * 3 wrong other info
3889 */
3890 struct intel_super *first = st->sb;
3891 struct intel_super *sec = tst->sb;
3892
5d500228
N		 3893 if (!first) {
3894 st->sb = tst->sb;
3895 tst->sb = NULL;
3896 return 0;
3897 }
8603ea6f
LM		 3898 /* in platform dependent environment test if the disks
3899 * use the same Intel hba
cb8f6859 3900 * If not on Intel hba at all, allow anything.
8603ea6f 3901 */
6b781d33
AP		 3902 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3903 if (first->hba->type != sec->hba->type) {
8603ea6f 3904 fprintf(stderr,
6b781d33
AP		 3905 "HBAs of devices do not match %s != %s\n",
3906 get_sys_dev_type(first->hba->type),
3907 get_sys_dev_type(sec->hba->type));
3908 return 3;
3909 }
3910 if (first->orom != sec->orom) {
3911 fprintf(stderr,
3912 "HBAs of devices do not match %s != %s\n",
3913 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3914 return 3;
3915 }
3916 }
cdddbdbc 3917
d23fe947
DW		 3918 /* if an anchor does not have num_raid_devs set then it is a free
3919 * floating spare
3920 */
3921 if (first->anchor->num_raid_devs > 0 &&
3922 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3923 /* Determine if these disks might ever have been
3924 * related. Further disambiguation can only take place
3925 * in load_super_imsm_all
3926 */
3927 __u32 first_family = first->anchor->orig_family_num;
3928 __u32 sec_family = sec->anchor->orig_family_num;
3929
f796af5d
DW		 3930 if (memcmp(first->anchor->sig, sec->anchor->sig,
3931 MAX_SIGNATURE_LENGTH) != 0)
3932 return 3;
3933
a2b97981
DW		 3934 if (first_family == 0)
3935 first_family = first->anchor->family_num;
3936 if (sec_family == 0)
3937 sec_family = sec->anchor->family_num;
3938
3939 if (first_family != sec_family)
d23fe947 3940 return 3;
f796af5d 3941
d23fe947 3942 }
cdddbdbc 3943
3e372e5a
DW		 3944 /* if 'first' is a spare promote it to a populated mpb with sec's
3945 * family number
3946 */
3947 if (first->anchor->num_raid_devs == 0 &&
3948 sec->anchor->num_raid_devs > 0) {
78d30f94 3949 int i;
ba2de7ba
DW		 3950 struct intel_dev *dv;
3951 struct imsm_dev *dev;
78d30f94
DW		 3952
3953 /* we need to copy raid device info from sec if an allocation
3954 * fails here we don't associate the spare
3955 */
3956 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3957 dv = xmalloc(sizeof(*dv));
3958 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3959 dv->dev = dev;
3960 dv->index = i;
3961 dv->next = first->devlist;
3962 first->devlist = dv;
78d30f94 3963 }
709743c5 3964 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3965 /* allocation failure */
3966 free_devlist(first);
e12b3daa 3967 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3968 return 3;
78d30f94 3969 }
3e372e5a 3970 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3971 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3972 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3973 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3974 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3975 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3976 }
3977
cdddbdbc
DW		 3978 return 0;
3979}
3980
0030e8d6
DW		 3981static void fd2devname(int fd, char *name)
3982{
3983 struct stat st;
3984 char path[256];
33a6535d 3985 char dname[PATH_MAX];
0030e8d6
DW		 3986 char *nm;
3987 int rv;
3988
3989 name[0] = '\0';
3990 if (fstat(fd, &st) != 0)
3991 return;
3992 sprintf(path, "/sys/dev/block/%d:%d",
3993 major(st.st_rdev), minor(st.st_rdev));
3994
9cf014ec 3995 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3996 if (rv <= 0)
3997 return;
9587c373 3998
0030e8d6
DW		 3999 dname[rv] = '\0';
4000 nm = strrchr(dname, '/');
7897de29
JS		 4001 if (nm) {
4002 nm++;
4003 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
4004 }
0030e8d6
DW		 4005}
4006
21e9380b
AP		 4007static int nvme_get_serial(int fd, void *buf, size_t buf_len)
4008{
4009 char path[60];
4010 char *name = fd2kname(fd);
4011
4012 if (!name)
4013 return 1;
4014
4015 if (strncmp(name, "nvme", 4) != 0)
4016 return 1;
4017
4018 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
4019
4020 return load_sys(path, buf, buf_len);
4021}
4022
cdddbdbc
DW		 4023extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
4024
4025static int imsm_read_serial(int fd, char *devname,
6da53c0e 4026 __u8 *serial, size_t serial_buf_len)
cdddbdbc 4027{
21e9380b 4028 char buf[50];
cdddbdbc 4029 int rv;
6da53c0e 4030 size_t len;
316e2bf4
DW		 4031 char *dest;
4032 char *src;
21e9380b
AP		 4033 unsigned int i;
4034
4035 memset(buf, 0, sizeof(buf));
cdddbdbc 4036
21e9380b 4037 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 4038
21e9380b
AP		 4039 if (rv)
4040 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 4041
40ebbb9c 4042 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 4043 memset(serial, 0, MAX_RAID_SERIAL_LEN);
4044 fd2devname(fd, (char *) serial);
0030e8d6
DW		 4045 return 0;
4046 }
4047
cdddbdbc
DW		 4048 if (rv != 0) {
4049 if (devname)
e7b84f9d
N		 4050 pr_err("Failed to retrieve serial for %s\n",
4051 devname);
cdddbdbc
DW		 4052 return rv;
4053 }
4054
316e2bf4
DW		 4055 /* trim all whitespace and non-printable characters and convert
4056 * ':' to ';'
4057 */
21e9380b
AP		 4058 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4059 src = &buf[i];
316e2bf4
DW		 4060 if (*src > 0x20) {
4061 /* ':' is reserved for use in placeholder serial
4062 * numbers for missing disks
4063 */
4064 if (*src == ':')
4065 *dest++ = ';';
4066 else
4067 *dest++ = *src;
4068 }
4069 }
21e9380b
AP		 4070 len = dest - buf;
4071 dest = buf;
316e2bf4 4072
6da53c0e
BK		 4073 if (len > serial_buf_len) {
4074 /* truncate leading characters */
4075 dest += len - serial_buf_len;
4076 len = serial_buf_len;
316e2bf4 4077 }
5c3db629 4078
6da53c0e 4079 memset(serial, 0, serial_buf_len);
316e2bf4 4080 memcpy(serial, dest, len);
cdddbdbc
DW		 4081
4082 return 0;
4083}
4084
1f24f035
DW		 4085static int serialcmp(__u8 *s1, __u8 *s2)
4086{
4087 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4088}
4089
4090static void serialcpy(__u8 *dest, __u8 *src)
4091{
4092 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4093}
4094
54c2c1ea
DW		 4095static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4096{
4097 struct dl *dl;
4098
4099 for (dl = super->disks; dl; dl = dl->next)
4100 if (serialcmp(dl->serial, serial) == 0)
4101 break;
4102
4103 return dl;
4104}
4105
a2b97981
DW		 4106static struct imsm_disk *
4107__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4108{
4109 int i;
4110
4111 for (i = 0; i < mpb->num_disks; i++) {
4112 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4113
4114 if (serialcmp(disk->serial, serial) == 0) {
4115 if (idx)
4116 *idx = i;
4117 return disk;
4118 }
4119 }
4120
4121 return NULL;
4122}
4123
cdddbdbc
DW		 4124static int
4125load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4126{
a2b97981 4127 struct imsm_disk *disk;
cdddbdbc
DW		 4128 struct dl *dl;
4129 struct stat stb;
cdddbdbc 4130 int rv;
a2b97981 4131 char name[40];
d23fe947
DW		 4132 __u8 serial[MAX_RAID_SERIAL_LEN];
4133
6da53c0e 4134 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4135
4136 if (rv != 0)
4137 return 2;
4138
503975b9 4139 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4140
a2b97981
DW		 4141 fstat(fd, &stb);
4142 dl->major = major(stb.st_rdev);
4143 dl->minor = minor(stb.st_rdev);
4144 dl->next = super->disks;
4145 dl->fd = keep_fd ? fd : -1;
4146 assert(super->disks == NULL);
4147 super->disks = dl;
4148 serialcpy(dl->serial, serial);
4149 dl->index = -2;
4150 dl->e = NULL;
4151 fd2devname(fd, name);
4152 if (devname)
503975b9 4153 dl->devname = xstrdup(devname);
a2b97981 4154 else
503975b9 4155 dl->devname = xstrdup(name);
cdddbdbc 4156
d23fe947 4157 /* look up this disk's index in the current anchor */
a2b97981
DW		 4158 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4159 if (disk) {
4160 dl->disk = *disk;
4161 /* only set index on disks that are a member of a
4162 * populated contianer, i.e. one with raid_devs
4163 */
4164 if (is_failed(&dl->disk))
3f6efecc 4165 dl->index = -2;
2432ce9b 4166 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4167 dl->index = -1;
3f6efecc
DW		 4168 }
4169
949c47a0
DW		 4170 return 0;
4171}
4172
0c046afd
DW		 4173/* When migrating map0 contains the 'destination' state while map1
4174 * contains the current state. When not migrating map0 contains the
4175 * current state. This routine assumes that map[0].map_state is set to
4176 * the current array state before being called.
4177 *
4178 * Migration is indicated by one of the following states
4179 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4180 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4181 * map1state=unitialized)
1484e727 4182 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4183 * map1state=normal)
e3bba0e0 4184 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4185 * map1state=degraded)
8e59f3d8
AK		 4186 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4187 * map1state=normal)
0c046afd 4188 */
8e59f3d8
AK		 4189static void migrate(struct imsm_dev *dev, struct intel_super *super,
4190 __u8 to_state, int migr_type)
3393c6af 4191{
0c046afd 4192 struct imsm_map *dest;
238c0a71 4193 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4194
0c046afd 4195 dev->vol.migr_state = 1;
1484e727 4196 set_migr_type(dev, migr_type);
f8f603f1 4197 dev->vol.curr_migr_unit = 0;
238c0a71 4198 dest = get_imsm_map(dev, MAP_1);
0c046afd 4199
0556e1a2 4200 /* duplicate and then set the target end state in map[0] */
3393c6af 4201 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4202 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4203 __u32 ord;
4204 int i;
4205
4206 for (i = 0; i < src->num_members; i++) {
4207 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4208 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4209 }
4210 }
4211
8e59f3d8
AK		 4212 if (migr_type == MIGR_GEN_MIGR)
4213 /* Clear migration record */
4214 memset(super->migr_rec, 0, sizeof(struct migr_record));
4215
0c046afd 4216 src->map_state = to_state;
949c47a0 4217}
f8f603f1 4218
809da78e
AK		 4219static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4220 __u8 map_state)
f8f603f1 4221{
238c0a71
AK		 4222 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4223 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4224 MAP_0 : MAP_1);
28bce06f 4225 int i, j;
0556e1a2
DW		 4226
4227 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4228 * completed in the last migration.
4229 *
28bce06f 4230 * FIXME add support for raid-level-migration
0556e1a2 4231 */
089f9d79
JS		 4232 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4233 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4234 /* when final map state is other than expected
4235 * merge maps (not for migration)
4236 */
4237 int failed;
4238
4239 for (i = 0; i < prev->num_members; i++)
4240 for (j = 0; j < map->num_members; j++)
4241 /* during online capacity expansion
4242 * disks position can be changed
4243 * if takeover is used
4244 */
4245 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4246 ord_to_idx(prev->disk_ord_tbl[i])) {
4247 map->disk_ord_tbl[j] |=
4248 prev->disk_ord_tbl[i];
4249 break;
4250 }
4251 failed = imsm_count_failed(super, dev, MAP_0);
4252 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4253 }
f8f603f1
DW		 4254
4255 dev->vol.migr_state = 0;
ea672ee1 4256 set_migr_type(dev, 0);
f8f603f1
DW		 4257 dev->vol.curr_migr_unit = 0;
4258 map->map_state = map_state;
4259}
949c47a0
DW		 4260
4261static int parse_raid_devices(struct intel_super *super)
4262{
4263 int i;
4264 struct imsm_dev *dev_new;
4d7b1503 4265 size_t len, len_migr;
401d313b 4266 size_t max_len = 0;
4d7b1503
DW		 4267 size_t space_needed = 0;
4268 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4269
4270 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4271 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4272 struct intel_dev *dv;
949c47a0 4273
4d7b1503
DW		 4274 len = sizeof_imsm_dev(dev_iter, 0);
4275 len_migr = sizeof_imsm_dev(dev_iter, 1);
4276 if (len_migr > len)
4277 space_needed += len_migr - len;
ca9de185 4278
503975b9 4279 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4280 if (max_len < len_migr)
4281 max_len = len_migr;
4282 if (max_len > len_migr)
4283 space_needed += max_len - len_migr;
503975b9 4284 dev_new = xmalloc(max_len);
949c47a0 4285 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4286 dv->dev = dev_new;
4287 dv->index = i;
4288 dv->next = super->devlist;
4289 super->devlist = dv;
949c47a0 4290 }
cdddbdbc 4291
4d7b1503
DW		 4292 /* ensure that super->buf is large enough when all raid devices
4293 * are migrating
4294 */
4295 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4296 void *buf;
4297
f36a9ecd
PB		 4298 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4299 super->sector_size);
4300 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4301 return 1;
4302
1f45a8ad
DW		 4303 memcpy(buf, super->buf, super->len);
4304 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4305 free(super->buf);
4306 super->buf = buf;
4307 super->len = len;
4308 }
ca9de185 4309
bbab0940
TM		 4310 super->extra_space += space_needed;
4311
cdddbdbc
DW		 4312 return 0;
4313}
4314
e2f41b2c
AK		 4315/*******************************************************************************
4316 * Function: check_mpb_migr_compatibility
4317 * Description: Function checks for unsupported migration features:
4318 * - migration optimization area (pba_of_lba0)
4319 * - descending reshape (ascending_migr)
4320 * Parameters:
4321 * super : imsm metadata information
4322 * Returns:
4323 * 0 : migration is compatible
4324 * -1 : migration is not compatible
4325 ******************************************************************************/
4326int check_mpb_migr_compatibility(struct intel_super *super)
4327{
4328 struct imsm_map *map0, *map1;
4329 struct migr_record *migr_rec = super->migr_rec;
4330 int i;
4331
4332 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4333 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4334
4335 if (dev_iter &&
4336 dev_iter->vol.migr_state == 1 &&
4337 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4338 /* This device is migrating */
238c0a71
AK		 4339 map0 = get_imsm_map(dev_iter, MAP_0);
4340 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4341 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4342 /* migration optimization area was used */
4343 return -1;
fc54fe7a
JS		 4344 if (migr_rec->ascending_migr == 0 &&
4345 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4346 /* descending reshape not supported yet */
4347 return -1;
4348 }
4349 }
4350 return 0;
4351}
4352
d23fe947 4353static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4354
cdddbdbc 4355/* load_imsm_mpb - read matrix metadata
f2f5c343 4356 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4357 */
4358static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4359{
4360 unsigned long long dsize;
cdddbdbc 4361 unsigned long long sectors;
f36a9ecd 4362 unsigned int sector_size = super->sector_size;
cdddbdbc 4363 struct stat;
6416d527 4364 struct imsm_super *anchor;
cdddbdbc
DW		 4365 __u32 check_sum;
4366
cdddbdbc 4367 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4368 if (dsize < 2*sector_size) {
64436f06 4369 if (devname)
e7b84f9d
N		 4370 pr_err("%s: device to small for imsm\n",
4371 devname);
64436f06
N		 4372 return 1;
4373 }
cdddbdbc 4374
f36a9ecd 4375 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4376 if (devname)
e7b84f9d
N		 4377 pr_err("Cannot seek to anchor block on %s: %s\n",
4378 devname, strerror(errno));
cdddbdbc
DW		 4379 return 1;
4380 }
4381
f36a9ecd 4382 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4383 if (devname)
7a862a02 4384 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4385 return 1;
4386 }
466070ad 4387 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4388 if (devname)
e7b84f9d
N		 4389 pr_err("Cannot read anchor block on %s: %s\n",
4390 devname, strerror(errno));
6416d527 4391 free(anchor);
cdddbdbc
DW		 4392 return 1;
4393 }
4394
6416d527 4395 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4396 if (devname)
e7b84f9d 4397 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4398 free(anchor);
cdddbdbc
DW		 4399 return 2;
4400 }
4401
d23fe947 4402 __free_imsm(super, 0);
f2f5c343
LM		 4403 /* reload capability and hba */
4404
4405 /* capability and hba must be updated with new super allocation */
d424212e 4406 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4407 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4408 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4409 if (devname)
e7b84f9d
N		 4410 pr_err("unable to allocate %zu byte mpb buffer\n",
4411 super->len);
6416d527 4412 free(anchor);
cdddbdbc
DW		 4413 return 2;
4414 }
f36a9ecd 4415 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4416
f36a9ecd 4417 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4418 free(anchor);
8e59f3d8 4419
85337573
AO		 4420 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4421 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4422 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4423 free(super->buf);
4424 return 2;
4425 }
51d83f5d 4426 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4427
949c47a0 4428 if (!sectors) {
ecf45690
DW		 4429 check_sum = __gen_imsm_checksum(super->anchor);
4430 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4431 if (devname)
e7b84f9d
N		 4432 pr_err("IMSM checksum %x != %x on %s\n",
4433 check_sum,
4434 __le32_to_cpu(super->anchor->check_sum),
4435 devname);
ecf45690
DW		 4436 return 2;
4437 }
4438
a2b97981 4439 return 0;
949c47a0 4440 }
cdddbdbc
DW		 4441
4442 /* read the extended mpb */
f36a9ecd 4443 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4444 if (devname)
e7b84f9d
N		 4445 pr_err("Cannot seek to extended mpb on %s: %s\n",
4446 devname, strerror(errno));
cdddbdbc
DW		 4447 return 1;
4448 }
4449
f36a9ecd
PB		 4450 if ((unsigned int)read(fd, super->buf + sector_size,
4451 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4452 if (devname)
e7b84f9d
N		 4453 pr_err("Cannot read extended mpb on %s: %s\n",
4454 devname, strerror(errno));
cdddbdbc
DW		 4455 return 2;
4456 }
4457
949c47a0
DW		 4458 check_sum = __gen_imsm_checksum(super->anchor);
4459 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4460 if (devname)
e7b84f9d
N		 4461 pr_err("IMSM checksum %x != %x on %s\n",
4462 check_sum, __le32_to_cpu(super->anchor->check_sum),
4463 devname);
db575f3b 4464 return 3;
cdddbdbc
DW		 4465 }
4466
a2b97981
DW		 4467 return 0;
4468}
4469
8e59f3d8
AK		 4470static int read_imsm_migr_rec(int fd, struct intel_super *super);
4471
97f81ee2
CA		 4472/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4473static void clear_hi(struct intel_super *super)
4474{
4475 struct imsm_super *mpb = super->anchor;
4476 int i, n;
4477 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4478 return;
4479 for (i = 0; i < mpb->num_disks; ++i) {
4480 struct imsm_disk *disk = &mpb->disk[i];
4481 disk->total_blocks_hi = 0;
4482 }
4483 for (i = 0; i < mpb->num_raid_devs; ++i) {
4484 struct imsm_dev *dev = get_imsm_dev(super, i);
4485 if (!dev)
4486 return;
4487 for (n = 0; n < 2; ++n) {
4488 struct imsm_map *map = get_imsm_map(dev, n);
4489 if (!map)
4490 continue;
4491 map->pba_of_lba0_hi = 0;
4492 map->blocks_per_member_hi = 0;
4493 map->num_data_stripes_hi = 0;
4494 }
4495 }
4496}
4497
a2b97981
DW		 4498static int
4499load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4500{
4501 int err;
4502
4503 err = load_imsm_mpb(fd, super, devname);
4504 if (err)
4505 return err;
f36a9ecd
PB		 4506 if (super->sector_size == 4096)
4507 convert_from_4k(super);
a2b97981
DW		 4508 err = load_imsm_disk(fd, super, devname, keep_fd);
4509 if (err)
4510 return err;
4511 err = parse_raid_devices(super);
8d67477f
TM		 4512 if (err)
4513 return err;
4514 err = load_bbm_log(super);
97f81ee2 4515 clear_hi(super);
a2b97981 4516 return err;
cdddbdbc
DW		 4517}
4518
ae6aad82
DW		 4519static void __free_imsm_disk(struct dl *d)
4520{
4521 if (d->fd >= 0)
4522 close(d->fd);
4523 if (d->devname)
4524 free(d->devname);
0dcecb2e
DW		 4525 if (d->e)
4526 free(d->e);
ae6aad82
DW		 4527 free(d);
4528
4529}
1a64be56 4530
cdddbdbc
DW		 4531static void free_imsm_disks(struct intel_super *super)
4532{
47ee5a45 4533 struct dl *d;
cdddbdbc 4534
47ee5a45
DW		 4535 while (super->disks) {
4536 d = super->disks;
cdddbdbc 4537 super->disks = d->next;
ae6aad82 4538 __free_imsm_disk(d);
cdddbdbc 4539 }
cb82edca
AK		 4540 while (super->disk_mgmt_list) {
4541 d = super->disk_mgmt_list;
4542 super->disk_mgmt_list = d->next;
4543 __free_imsm_disk(d);
4544 }
47ee5a45
DW		 4545 while (super->missing) {
4546 d = super->missing;
4547 super->missing = d->next;
4548 __free_imsm_disk(d);
4549 }
4550
cdddbdbc
DW		 4551}
4552
9ca2c81c 4553/* free all the pieces hanging off of a super pointer */
d23fe947 4554static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4555{
88654014
LM		 4556 struct intel_hba *elem, *next;
4557
9ca2c81c 4558 if (super->buf) {
949c47a0 4559 free(super->buf);
9ca2c81c
DW		 4560 super->buf = NULL;
4561 }
f2f5c343
LM		 4562 /* unlink capability description */
4563 super->orom = NULL;
8e59f3d8
AK		 4564 if (super->migr_rec_buf) {
4565 free(super->migr_rec_buf);
4566 super->migr_rec_buf = NULL;
4567 }
d23fe947
DW		 4568 if (free_disks)
4569 free_imsm_disks(super);
ba2de7ba 4570 free_devlist(super);
88654014
LM		 4571 elem = super->hba;
4572 while (elem) {
4573 if (elem->path)
4574 free((void *)elem->path);
4575 next = elem->next;
4576 free(elem);
4577 elem = next;
88c32bb1 4578 }
8d67477f
TM		 4579 if (super->bbm_log)
4580 free(super->bbm_log);
88654014 4581 super->hba = NULL;
cdddbdbc
DW		 4582}
4583
9ca2c81c
DW		 4584static void free_imsm(struct intel_super *super)
4585{
d23fe947 4586 __free_imsm(super, 1);
928f1424 4587 free(super->bb.entries);
9ca2c81c
DW		 4588 free(super);
4589}
cdddbdbc
DW		 4590
4591static void free_super_imsm(struct supertype *st)
4592{
4593 struct intel_super *super = st->sb;
4594
4595 if (!super)
4596 return;
4597
4598 free_imsm(super);
4599 st->sb = NULL;
4600}
4601
49133e57 4602static struct intel_super *alloc_super(void)
c2c087e6 4603{
503975b9 4604 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4605
503975b9
N		 4606 super->current_vol = -1;
4607 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4608
4609 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4610 sizeof(struct md_bb_entry));
4611 if (!super->bb.entries) {
4612 free(super);
4613 return NULL;
4614 }
4615
c2c087e6
DW		 4616 return super;
4617}
4618
f0f5a016
LM		 4619/*
4620 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4621 */
d424212e 4622static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4623{
4624 struct sys_dev *hba_name;
4625 int rv = 0;
4626
3a30e28e
MT		 4627 if (fd >= 0 && test_partition(fd)) {
4628 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4629 devname);
4630 return 1;
4631 }
089f9d79 4632 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4633 super->orom = NULL;
f0f5a016
LM		 4634 super->hba = NULL;
4635 return 0;
4636 }
4637 hba_name = find_disk_attached_hba(fd, NULL);
4638 if (!hba_name) {
d424212e 4639 if (devname)
e7b84f9d
N		 4640 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4641 devname);
f0f5a016
LM		 4642 return 1;
4643 }
4644 rv = attach_hba_to_super(super, hba_name);
4645 if (rv == 2) {
d424212e
N		 4646 if (devname) {
4647 struct intel_hba *hba = super->hba;
f0f5a016 4648
60f0f54d
PB		 4649 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4650 " but the container is assigned to Intel(R) %s %s (",
d424212e 4651 devname,
614902f6 4652 get_sys_dev_type(hba_name->type),
60f0f54d 4653 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4654 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4655 get_sys_dev_type(super->hba->type),
4656 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4657
f0f5a016
LM		 4658 while (hba) {
4659 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4660 if (hba->next)
4661 fprintf(stderr, ", ");
4662 hba = hba->next;
4663 }
6b781d33 4664 fprintf(stderr, ").\n"
cca67208 4665 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4666 }
f0f5a016
LM		 4667 return 2;
4668 }
6b781d33 4669 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4670 if (!super->orom)
4671 return 3;
614902f6 4672
f0f5a016
LM		 4673 return 0;
4674}
4675
47ee5a45
DW		 4676/* find_missing - helper routine for load_super_imsm_all that identifies
4677 * disks that have disappeared from the system. This routine relies on
4678 * the mpb being uptodate, which it is at load time.
4679 */
4680static int find_missing(struct intel_super *super)
4681{
4682 int i;
4683 struct imsm_super *mpb = super->anchor;
4684 struct dl *dl;
4685 struct imsm_disk *disk;
47ee5a45
DW		 4686
4687 for (i = 0; i < mpb->num_disks; i++) {
4688 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4689 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4690 if (dl)
4691 continue;
47ee5a45 4692
503975b9 4693 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4694 dl->major = 0;
4695 dl->minor = 0;
4696 dl->fd = -1;
503975b9 4697 dl->devname = xstrdup("missing");
47ee5a45
DW		 4698 dl->index = i;
4699 serialcpy(dl->serial, disk->serial);
4700 dl->disk = *disk;
689c9bf3 4701 dl->e = NULL;
47ee5a45
DW		 4702 dl->next = super->missing;
4703 super->missing = dl;
4704 }
4705
4706 return 0;
4707}
4708
a2b97981
DW		 4709static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4710{
4711 struct intel_disk *idisk = disk_list;
4712
4713 while (idisk) {
4714 if (serialcmp(idisk->disk.serial, serial) == 0)
4715 break;
4716 idisk = idisk->next;
4717 }
4718
4719 return idisk;
4720}
4721
4722static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4723 struct intel_super *super,
4724 struct intel_disk **disk_list)
4725{
4726 struct imsm_disk *d = &super->disks->disk;
4727 struct imsm_super *mpb = super->anchor;
4728 int i, j;
4729
4730 for (i = 0; i < tbl_size; i++) {
4731 struct imsm_super *tbl_mpb = table[i]->anchor;
4732 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4733
4734 if (tbl_mpb->family_num == mpb->family_num) {
4735 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4736 dprintf("mpb from %d:%d matches %d:%d\n",
4737 super->disks->major,
a2b97981
DW		 4738 super->disks->minor,
4739 table[i]->disks->major,
4740 table[i]->disks->minor);
4741 break;
4742 }
4743
4744 if (((is_configured(d) && !is_configured(tbl_d)) ||
4745 is_configured(d) == is_configured(tbl_d)) &&
4746 tbl_mpb->generation_num < mpb->generation_num) {
4747 /* current version of the mpb is a
4748 * better candidate than the one in
4749 * super_table, but copy over "cross
4750 * generational" status
4751 */
4752 struct intel_disk *idisk;
4753
1ade5cc1
N		 4754 dprintf("mpb from %d:%d replaces %d:%d\n",
4755 super->disks->major,
a2b97981
DW		 4756 super->disks->minor,
4757 table[i]->disks->major,
4758 table[i]->disks->minor);
4759
4760 idisk = disk_list_get(tbl_d->serial, *disk_list);
4761 if (idisk && is_failed(&idisk->disk))
4762 tbl_d->status |= FAILED_DISK;
4763 break;
4764 } else {
4765 struct intel_disk *idisk;
4766 struct imsm_disk *disk;
4767
4768 /* tbl_mpb is more up to date, but copy
4769 * over cross generational status before
4770 * returning
4771 */
4772 disk = __serial_to_disk(d->serial, mpb, NULL);
4773 if (disk && is_failed(disk))
4774 d->status |= FAILED_DISK;
4775
4776 idisk = disk_list_get(d->serial, *disk_list);
4777 if (idisk) {
4778 idisk->owner = i;
4779 if (disk && is_configured(disk))
4780 idisk->disk.status |= CONFIGURED_DISK;
4781 }
4782
1ade5cc1
N		 4783 dprintf("mpb from %d:%d prefer %d:%d\n",
4784 super->disks->major,
a2b97981
DW		 4785 super->disks->minor,
4786 table[i]->disks->major,
4787 table[i]->disks->minor);
4788
4789 return tbl_size;
4790 }
4791 }
4792 }
4793
4794 if (i >= tbl_size)
4795 table[tbl_size++] = super;
4796 else
4797 table[i] = super;
4798
4799 /* update/extend the merged list of imsm_disk records */
4800 for (j = 0; j < mpb->num_disks; j++) {
4801 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4802 struct intel_disk *idisk;
4803
4804 idisk = disk_list_get(disk->serial, *disk_list);
4805 if (idisk) {
4806 idisk->disk.status |= disk->status;
4807 if (is_configured(&idisk->disk) ||
4808 is_failed(&idisk->disk))
4809 idisk->disk.status &= ~(SPARE_DISK);
4810 } else {
503975b9 4811 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4812 idisk->owner = IMSM_UNKNOWN_OWNER;
4813 idisk->disk = *disk;
4814 idisk->next = *disk_list;
4815 *disk_list = idisk;
4816 }
4817
4818 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4819 idisk->owner = i;
4820 }
4821
4822 return tbl_size;
4823}
4824
4825static struct intel_super *
4826validate_members(struct intel_super *super, struct intel_disk *disk_list,
4827 const int owner)
4828{
4829 struct imsm_super *mpb = super->anchor;
4830 int ok_count = 0;
4831 int i;
4832
4833 for (i = 0; i < mpb->num_disks; i++) {
4834 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4835 struct intel_disk *idisk;
4836
4837 idisk = disk_list_get(disk->serial, disk_list);
4838 if (idisk) {
4839 if (idisk->owner == owner ||
4840 idisk->owner == IMSM_UNKNOWN_OWNER)
4841 ok_count++;
4842 else
1ade5cc1
N		 4843 dprintf("'%.16s' owner %d != %d\n",
4844 disk->serial, idisk->owner,
a2b97981
DW		 4845 owner);
4846 } else {
1ade5cc1
N		 4847 dprintf("unknown disk %x [%d]: %.16s\n",
4848 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4849 disk->serial);
4850 break;
4851 }
4852 }
4853
4854 if (ok_count == mpb->num_disks)
4855 return super;
4856 return NULL;
4857}
4858
4859static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4860{
4861 struct intel_super *s;
4862
4863 for (s = super_list; s; s = s->next) {
4864 if (family_num != s->anchor->family_num)
4865 continue;
e12b3daa 4866 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4867 __le32_to_cpu(family_num), s->disks->devname);
4868 }
4869}
4870
4871static struct intel_super *
4872imsm_thunderdome(struct intel_super **super_list, int len)
4873{
4874 struct intel_super *super_table[len];
4875 struct intel_disk *disk_list = NULL;
4876 struct intel_super *champion, *spare;
4877 struct intel_super *s, **del;
4878 int tbl_size = 0;
4879 int conflict;
4880 int i;
4881
4882 memset(super_table, 0, sizeof(super_table));
4883 for (s = *super_list; s; s = s->next)
4884 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4885
4886 for (i = 0; i < tbl_size; i++) {
4887 struct imsm_disk *d;
4888 struct intel_disk *idisk;
4889 struct imsm_super *mpb = super_table[i]->anchor;
4890
4891 s = super_table[i];
4892 d = &s->disks->disk;
4893
4894 /* 'd' must appear in merged disk list for its
4895 * configuration to be valid
4896 */
4897 idisk = disk_list_get(d->serial, disk_list);
4898 if (idisk && idisk->owner == i)
4899 s = validate_members(s, disk_list, i);
4900 else
4901 s = NULL;
4902
4903 if (!s)
1ade5cc1
N		 4904 dprintf("marking family: %#x from %d:%d offline\n",
4905 mpb->family_num,
a2b97981
DW		 4906 super_table[i]->disks->major,
4907 super_table[i]->disks->minor);
4908 super_table[i] = s;
4909 }
4910
4911 /* This is where the mdadm implementation differs from the Windows
4912 * driver which has no strict concept of a container. We can only
4913 * assemble one family from a container, so when returning a prodigal
4914 * array member to this system the code will not be able to disambiguate
4915 * the container contents that should be assembled ("foreign" versus
4916 * "local"). It requires user intervention to set the orig_family_num
4917 * to a new value to establish a new container. The Windows driver in
4918 * this situation fixes up the volume name in place and manages the
4919 * foreign array as an independent entity.
4920 */
4921 s = NULL;
4922 spare = NULL;
4923 conflict = 0;
4924 for (i = 0; i < tbl_size; i++) {
4925 struct intel_super *tbl_ent = super_table[i];
4926 int is_spare = 0;
4927
4928 if (!tbl_ent)
4929 continue;
4930
4931 if (tbl_ent->anchor->num_raid_devs == 0) {
4932 spare = tbl_ent;
4933 is_spare = 1;
4934 }
4935
4936 if (s && !is_spare) {
4937 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4938 conflict++;
4939 } else if (!s && !is_spare)
4940 s = tbl_ent;
4941 }
4942
4943 if (!s)
4944 s = spare;
4945 if (!s) {
4946 champion = NULL;
4947 goto out;
4948 }
4949 champion = s;
4950
4951 if (conflict)
7a862a02 4952 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4953 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4954
4955 /* collect all dl's onto 'champion', and update them to
4956 * champion's version of the status
4957 */
4958 for (s = *super_list; s; s = s->next) {
4959 struct imsm_super *mpb = champion->anchor;
4960 struct dl *dl = s->disks;
4961
4962 if (s == champion)
4963 continue;
4964
5d7b407a
CA		 4965 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4966
a2b97981
DW		 4967 for (i = 0; i < mpb->num_disks; i++) {
4968 struct imsm_disk *disk;
4969
4970 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4971 if (disk) {
4972 dl->disk = *disk;
4973 /* only set index on disks that are a member of
4974 * a populated contianer, i.e. one with
4975 * raid_devs
4976 */
4977 if (is_failed(&dl->disk))
4978 dl->index = -2;
4979 else if (is_spare(&dl->disk))
4980 dl->index = -1;
4981 break;
4982 }
4983 }
4984
4985 if (i >= mpb->num_disks) {
4986 struct intel_disk *idisk;
4987
4988 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4989 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4990 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4991 dl->index = -1;
4992 else {
4993 dl->index = -2;
4994 continue;
4995 }
4996 }
4997
4998 dl->next = champion->disks;
4999 champion->disks = dl;
5000 s->disks = NULL;
5001 }
5002
5003 /* delete 'champion' from super_list */
5004 for (del = super_list; *del; ) {
5005 if (*del == champion) {
5006 *del = (*del)->next;
5007 break;
5008 } else
5009 del = &(*del)->next;
5010 }
5011 champion->next = NULL;
5012
5013 out:
5014 while (disk_list) {
5015 struct intel_disk *idisk = disk_list;
5016
5017 disk_list = disk_list->next;
5018 free(idisk);
5019 }
5020
5021 return champion;
5022}
5023
9587c373
LM		 5024static int
5025get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 5026static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 5027 int major, int minor, int keep_fd);
ec50f7b6
LM		 5028static int
5029get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5030 int *max, int keep_fd);
5031
cdddbdbc 5032static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 5033 char *devname, struct md_list *devlist,
5034 int keep_fd)
cdddbdbc 5035{
a2b97981
DW		 5036 struct intel_super *super_list = NULL;
5037 struct intel_super *super = NULL;
a2b97981 5038 int err = 0;
9587c373 5039 int i = 0;
dab4a513 5040
9587c373
LM		 5041 if (fd >= 0)
5042 /* 'fd' is an opened container */
5043 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
5044 else
ec50f7b6
LM		 5045 /* get super block from devlist devices */
5046 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5047 if (err)
1602d52c 5048 goto error;
a2b97981
DW		 5049 /* all mpbs enter, maybe one leaves */
5050 super = imsm_thunderdome(&super_list, i);
5051 if (!super) {
5052 err = 1;
5053 goto error;
cdddbdbc
DW		 5054 }
5055
47ee5a45
DW		 5056 if (find_missing(super) != 0) {
5057 free_imsm(super);
a2b97981
DW		 5058 err = 2;
5059 goto error;
47ee5a45 5060 }
8e59f3d8
AK		 5061
5062 /* load migration record */
5063 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5064 if (err == -1) {
5065 /* migration is in progress,
5066 * but migr_rec cannot be loaded,
5067 */
8e59f3d8
AK		 5068 err = 4;
5069 goto error;
5070 }
e2f41b2c
AK		 5071
5072 /* Check migration compatibility */
089f9d79 5073 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5074 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5075 if (devname)
5076 fprintf(stderr, " on %s\n", devname);
5077 else
5078 fprintf(stderr, " (IMSM).\n");
5079
5080 err = 5;
5081 goto error;
5082 }
5083
a2b97981
DW		 5084 err = 0;
5085
5086 error:
5087 while (super_list) {
5088 struct intel_super *s = super_list;
5089
5090 super_list = super_list->next;
5091 free_imsm(s);
5092 }
9587c373 5093
a2b97981
DW		 5094 if (err)
5095 return err;
f7e7067b 5096
cdddbdbc 5097 *sbp = super;
9587c373 5098 if (fd >= 0)
4dd2df09 5099 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5100 else
4dd2df09 5101 st->container_devnm[0] = 0;
a2b97981 5102 if (err == 0 && st->ss == NULL) {
bf5a934a 5103 st->ss = &super_imsm;
cdddbdbc
DW		 5104 st->minor_version = 0;
5105 st->max_devs = IMSM_MAX_DEVICES;
5106 }
cdddbdbc
DW		 5107 return 0;
5108}
2b959fbf 5109
ec50f7b6
LM		 5110static int
5111get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5112 int *max, int keep_fd)
5113{
5114 struct md_list *tmpdev;
5115 int err = 0;
5116 int i = 0;
9587c373 5117
ec50f7b6
LM		 5118 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5119 if (tmpdev->used != 1)
5120 continue;
5121 if (tmpdev->container == 1) {
ca9de185 5122 int lmax = 0;
ec50f7b6
LM		 5123 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5124 if (fd < 0) {
e7b84f9d 5125 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5126 tmpdev->devname, strerror(errno));
5127 err = 8;
5128 goto error;
5129 }
5130 err = get_sra_super_block(fd, super_list,
5131 tmpdev->devname, &lmax,
5132 keep_fd);
5133 i += lmax;
5134 close(fd);
5135 if (err) {
5136 err = 7;
5137 goto error;
5138 }
5139 } else {
5140 int major = major(tmpdev->st_rdev);
5141 int minor = minor(tmpdev->st_rdev);
5142 err = get_super_block(super_list,
4dd2df09 5143 NULL,
ec50f7b6
LM		 5144 tmpdev->devname,
5145 major, minor,
5146 keep_fd);
5147 i++;
5148 if (err) {
5149 err = 6;
5150 goto error;
5151 }
5152 }
5153 }
5154 error:
5155 *max = i;
5156 return err;
5157}
9587c373 5158
4dd2df09 5159static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5160 int major, int minor, int keep_fd)
5161{
594dc1b8 5162 struct intel_super *s;
9587c373
LM		 5163 char nm[32];
5164 int dfd = -1;
9587c373
LM		 5165 int err = 0;
5166 int retry;
5167
5168 s = alloc_super();
5169 if (!s) {
5170 err = 1;
5171 goto error;
5172 }
5173
5174 sprintf(nm, "%d:%d", major, minor);
5175 dfd = dev_open(nm, O_RDWR);
5176 if (dfd < 0) {
5177 err = 2;
5178 goto error;
5179 }
5180
fa7bb6f8 5181 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5182 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5183 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5184
5185 /* retry the load if we might have raced against mdmon */
4dd2df09 5186 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5187 for (retry = 0; retry < 3; retry++) {
5188 usleep(3000);
5189 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5190 if (err != 3)
5191 break;
5192 }
5193 error:
5194 if (!err) {
5195 s->next = *super_list;
5196 *super_list = s;
5197 } else {
5198 if (s)
8d67477f 5199 free_imsm(s);
36614e95 5200 if (dfd >= 0)
9587c373
LM		 5201 close(dfd);
5202 }
089f9d79 5203 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5204 close(dfd);
5205 return err;
5206
5207}
5208
5209static int
5210get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5211{
5212 struct mdinfo *sra;
4dd2df09 5213 char *devnm;
9587c373
LM		 5214 struct mdinfo *sd;
5215 int err = 0;
5216 int i = 0;
4dd2df09 5217 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5218 if (!sra)
5219 return 1;
5220
5221 if (sra->array.major_version != -1 ||
5222 sra->array.minor_version != -2 ||
5223 strcmp(sra->text_version, "imsm") != 0) {
5224 err = 1;
5225 goto error;
5226 }
5227 /* load all mpbs */
4dd2df09 5228 devnm = fd2devnm(fd);
9587c373 5229 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5230 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5231 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5232 err = 7;
5233 goto error;
5234 }
5235 }
5236 error:
5237 sysfs_free(sra);
5238 *max = i;
5239 return err;
5240}
5241
2b959fbf
N		 5242static int load_container_imsm(struct supertype *st, int fd, char *devname)
5243{
ec50f7b6 5244 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5245}
cdddbdbc
DW		 5246
5247static int load_super_imsm(struct supertype *st, int fd, char *devname)
5248{
5249 struct intel_super *super;
5250 int rv;
8a3544f8 5251 int retry;
cdddbdbc 5252
357ac106 5253 if (test_partition(fd))
691c6ee1
N		 5254 /* IMSM not allowed on partitions */
5255 return 1;
5256
37424f13
DW		 5257 free_super_imsm(st);
5258
49133e57 5259 super = alloc_super();
fa7bb6f8 5260 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5261 if (!super)
5262 return 1;
ea2bc72b
LM		 5263 /* Load hba and capabilities if they exist.
5264 * But do not preclude loading metadata in case capabilities or hba are
5265 * non-compliant and ignore_hw_compat is set.
5266 */
d424212e 5267 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5268 /* no orom/efi or non-intel hba of the disk */
089f9d79 5269 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5270 if (devname)
e7b84f9d 5271 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5272 free_imsm(super);
5273 return 2;
5274 }
a2b97981 5275 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5276
8a3544f8
AP		 5277 /* retry the load if we might have raced against mdmon */
5278 if (rv == 3) {
f96b1302
AP		 5279 struct mdstat_ent *mdstat = NULL;
5280 char *name = fd2kname(fd);
5281
5282 if (name)
5283 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5284
5285 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5286 for (retry = 0; retry < 3; retry++) {
5287 usleep(3000);
5288 rv = load_and_parse_mpb(fd, super, devname, 0);
5289 if (rv != 3)
5290 break;
5291 }
5292 }
5293
5294 free_mdstat(mdstat);
5295 }
5296
cdddbdbc
DW		 5297 if (rv) {
5298 if (devname)
7a862a02 5299 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5300 free_imsm(super);
5301 return rv;
5302 }
5303
5304 st->sb = super;
5305 if (st->ss == NULL) {
5306 st->ss = &super_imsm;
5307 st->minor_version = 0;
5308 st->max_devs = IMSM_MAX_DEVICES;
5309 }
8e59f3d8
AK		 5310
5311 /* load migration record */
2e062e82
AK		 5312 if (load_imsm_migr_rec(super, NULL) == 0) {
5313 /* Check for unsupported migration features */
5314 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5315 pr_err("Unsupported migration detected");
2e062e82
AK		 5316 if (devname)
5317 fprintf(stderr, " on %s\n", devname);
5318 else
5319 fprintf(stderr, " (IMSM).\n");
5320 return 3;
5321 }
e2f41b2c
AK		 5322 }
5323
cdddbdbc
DW		 5324 return 0;
5325}
5326
ef6ffade
DW		 5327static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5328{
5329 if (info->level == 1)
5330 return 128;
5331 return info->chunk_size >> 9;
5332}
5333
5551b113
CA		 5334static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5335 unsigned long long size)
fcfd9599 5336{
4025c288 5337 if (info->level == 1)
5551b113 5338 return size * 2;
4025c288 5339 else
5551b113 5340 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5341}
5342
4d1313e9
DW		 5343static void imsm_update_version_info(struct intel_super *super)
5344{
5345 /* update the version and attributes */
5346 struct imsm_super *mpb = super->anchor;
5347 char *version;
5348 struct imsm_dev *dev;
5349 struct imsm_map *map;
5350 int i;
5351
5352 for (i = 0; i < mpb->num_raid_devs; i++) {
5353 dev = get_imsm_dev(super, i);
238c0a71 5354 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5355 if (__le32_to_cpu(dev->size_high) > 0)
5356 mpb->attributes |= MPB_ATTRIB_2TB;
5357
5358 /* FIXME detect when an array spans a port multiplier */
5359 #if 0
5360 mpb->attributes |= MPB_ATTRIB_PM;
5361 #endif
5362
5363 if (mpb->num_raid_devs > 1 ||
5364 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5365 version = MPB_VERSION_ATTRIBS;
5366 switch (get_imsm_raid_level(map)) {
5367 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5368 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5369 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5370 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5371 }
5372 } else {
5373 if (map->num_members >= 5)
5374 version = MPB_VERSION_5OR6_DISK_ARRAY;
5375 else if (dev->status == DEV_CLONE_N_GO)
5376 version = MPB_VERSION_CNG;
5377 else if (get_imsm_raid_level(map) == 5)
5378 version = MPB_VERSION_RAID5;
5379 else if (map->num_members >= 3)
5380 version = MPB_VERSION_3OR4_DISK_ARRAY;
5381 else if (get_imsm_raid_level(map) == 1)
5382 version = MPB_VERSION_RAID1;
5383 else
5384 version = MPB_VERSION_RAID0;
5385 }
5386 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5387 }
5388}
5389
aa534678
DW		 5390static int check_name(struct intel_super *super, char *name, int quiet)
5391{
5392 struct imsm_super *mpb = super->anchor;
5393 char *reason = NULL;
9bd99a90
RS		 5394 char *start = name;
5395 size_t len = strlen(name);
aa534678
DW		 5396 int i;
5397
9bd99a90
RS		 5398 if (len > 0) {
5399 while (isspace(start[len - 1]))
5400 start[--len] = 0;
5401 while (*start && isspace(*start))
5402 ++start, --len;
5403 memmove(name, start, len + 1);
5404 }
5405
5406 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5407 reason = "must be 16 characters or less";
9bd99a90
RS		 5408 else if (len == 0)
5409 reason = "must be a non-empty string";
aa534678
DW		 5410
5411 for (i = 0; i < mpb->num_raid_devs; i++) {
5412 struct imsm_dev *dev = get_imsm_dev(super, i);
5413
5414 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5415 reason = "already exists";
5416 break;
5417 }
5418 }
5419
5420 if (reason && !quiet)
e7b84f9d 5421 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5422
5423 return !reason;
5424}
5425
8b353278 5426static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5427 struct shape *s, char *name,
83cd1e97
N		 5428 char *homehost, int *uuid,
5429 long long data_offset)
cdddbdbc 5430{
c2c087e6
DW		 5431 /* We are creating a volume inside a pre-existing container.
5432 * so st->sb is already set.
5433 */
5434 struct intel_super *super = st->sb;
f36a9ecd 5435 unsigned int sector_size = super->sector_size;
949c47a0 5436 struct imsm_super *mpb = super->anchor;
ba2de7ba 5437 struct intel_dev *dv;
c2c087e6
DW		 5438 struct imsm_dev *dev;
5439 struct imsm_vol *vol;
5440 struct imsm_map *map;
5441 int idx = mpb->num_raid_devs;
5442 int i;
760365f9 5443 int namelen;
c2c087e6 5444 unsigned long long array_blocks;
2c092cad 5445 size_t size_old, size_new;
5551b113 5446 unsigned long long num_data_stripes;
b53bfba6
TM		 5447 unsigned int data_disks;
5448 unsigned long long size_per_member;
cdddbdbc 5449
88c32bb1 5450 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5451 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5452 return 0;
5453 }
5454
2c092cad
DW		 5455 /* ensure the mpb is large enough for the new data */
5456 size_old = __le32_to_cpu(mpb->mpb_size);
5457 size_new = disks_to_mpb_size(info->nr_disks);
5458 if (size_new > size_old) {
5459 void *mpb_new;
f36a9ecd 5460 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5461
f36a9ecd 5462 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5463 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5464 return 0;
5465 }
85337573
AO		 5466 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5467 MIGR_REC_BUF_SECTORS*
5468 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5469 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5470 free(super->buf);
5471 free(super);
ea944c8f 5472 free(mpb_new);
8e59f3d8
AK		 5473 return 0;
5474 }
2c092cad
DW		 5475 memcpy(mpb_new, mpb, size_old);
5476 free(mpb);
5477 mpb = mpb_new;
949c47a0 5478 super->anchor = mpb_new;
2c092cad
DW		 5479 mpb->mpb_size = __cpu_to_le32(size_new);
5480 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5481 super->len = size_round;
2c092cad 5482 }
bf5a934a 5483 super->current_vol = idx;
3960e579
DW		 5484
5485 /* handle 'failed_disks' by either:
5486 * a) create dummy disk entries in the table if this the first
5487 * volume in the array. We add them here as this is the only
5488 * opportunity to add them. add_to_super_imsm_volume()
5489 * handles the non-failed disks and continues incrementing
5490 * mpb->num_disks.
5491 * b) validate that 'failed_disks' matches the current number
5492 * of missing disks if the container is populated
d23fe947 5493 */
3960e579 5494 if (super->current_vol == 0) {
d23fe947 5495 mpb->num_disks = 0;
3960e579
DW		 5496 for (i = 0; i < info->failed_disks; i++) {
5497 struct imsm_disk *disk;
5498
5499 mpb->num_disks++;
5500 disk = __get_imsm_disk(mpb, i);
5501 disk->status = CONFIGURED_DISK | FAILED_DISK;
5502 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5503 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5504 "missing:%d", (__u8)i);
3960e579
DW		 5505 }
5506 find_missing(super);
5507 } else {
5508 int missing = 0;
5509 struct dl *d;
5510
5511 for (d = super->missing; d; d = d->next)
5512 missing++;
5513 if (info->failed_disks > missing) {
e7b84f9d 5514 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5515 return 0;
5516 }
5517 }
5a038140 5518
aa534678
DW		 5519 if (!check_name(super, name, 0))
5520 return 0;
503975b9
N		 5521 dv = xmalloc(sizeof(*dv));
5522 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5523 /*
5524 * Explicitly allow truncating to not confuse gcc's
5525 * -Werror=stringop-truncation
5526 */
5527 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5528 memcpy(dev->volume, name, namelen);
e03640bd 5529 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5530 info->layout, info->chunk_size,
b53bfba6
TM		 5531 s->size * BLOCKS_PER_KB);
5532 data_disks = get_data_disks(info->level, info->layout,
5533 info->raid_disks);
5534 array_blocks = round_size_to_mb(array_blocks, data_disks);
5535 size_per_member = array_blocks / data_disks;
979d38be 5536
fcc2c9da 5537 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5538 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5539 vol = &dev->vol;
5540 vol->migr_state = 0;
1484e727 5541 set_migr_type(dev, MIGR_INIT);
3960e579 5542 vol->dirty = !info->state;
f8f603f1 5543 vol->curr_migr_unit = 0;
238c0a71 5544 map = get_imsm_map(dev, MAP_0);
5551b113 5545 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5546 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5547 map->failed_disk_num = ~0;
bf4442ab 5548 if (info->level > 0)
fffaf1ff
N		 5549 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5550 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5551 else
5552 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5553 IMSM_T_STATE_NORMAL;
252d23c0 5554 map->ddf = 1;
ef6ffade
DW		 5555
5556 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5557 free(dev);
5558 free(dv);
7a862a02 5559 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5560 return 0;
5561 }
81062a36
DW		 5562
5563 map->raid_level = info->level;
4d1313e9 5564 if (info->level == 10) {
c2c087e6 5565 map->raid_level = 1;
4d1313e9 5566 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5567 } else if (info->level == 1)
5568 map->num_domains = info->raid_disks;
5569 else
ff596308 5570 map->num_domains = 1;
81062a36 5571
5551b113 5572 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5573 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5574 num_data_stripes /= map->num_domains;
5575 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5576
44490938
MD		 5577 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5578 set_blocks_per_member(map, info_to_blocks_per_member(info,
5579 size_per_member /
5580 BLOCKS_PER_KB));
5581
c2c087e6
DW		 5582 map->num_members = info->raid_disks;
5583 for (i = 0; i < map->num_members; i++) {
5584 /* initialized in add_to_super */
4eb26970 5585 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5586 }
949c47a0 5587 mpb->num_raid_devs++;
2a24dc1b
PB		 5588 mpb->num_raid_devs_created++;
5589 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5590
b7580566 5591 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5592 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5593 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5594 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5595 } else {
5596 free(dev);
5597 free(dv);
5598 pr_err("imsm does not support consistency policy %s\n",
5599 map_num(consistency_policies, s->consistency_policy));
5600 return 0;
5601 }
5602
ba2de7ba
DW		 5603 dv->dev = dev;
5604 dv->index = super->current_vol;
5605 dv->next = super->devlist;
5606 super->devlist = dv;
c2c087e6 5607
4d1313e9
DW		 5608 imsm_update_version_info(super);
5609
c2c087e6 5610 return 1;
cdddbdbc
DW		 5611}
5612
bf5a934a 5613static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5614 struct shape *s, char *name,
83cd1e97
N		 5615 char *homehost, int *uuid,
5616 unsigned long long data_offset)
bf5a934a
DW		 5617{
5618 /* This is primarily called by Create when creating a new array.
5619 * We will then get add_to_super called for each component, and then
5620 * write_init_super called to write it out to each device.
5621 * For IMSM, Create can create on fresh devices or on a pre-existing
5622 * array.
5623 * To create on a pre-existing array a different method will be called.
5624 * This one is just for fresh drives.
5625 */
5626 struct intel_super *super;
5627 struct imsm_super *mpb;
5628 size_t mpb_size;
4d1313e9 5629 char *version;
bf5a934a 5630
83cd1e97 5631 if (data_offset != INVALID_SECTORS) {
ed503f89 5632 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5633 return 0;
5634 }
5635
bf5a934a 5636 if (st->sb)
5308f117 5637 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5638 data_offset);
e683ca88
DW		 5639
5640 if (info)
5641 mpb_size = disks_to_mpb_size(info->nr_disks);
5642 else
f36a9ecd 5643 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5644
49133e57 5645 super = alloc_super();
f36a9ecd
PB		 5646 if (super &&
5647 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5648 free_imsm(super);
e683ca88
DW		 5649 super = NULL;
5650 }
5651 if (!super) {
1ade5cc1 5652 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5653 return 0;
5654 }
de44e46f
PB		 5655 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5656 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5657 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5658 free(super->buf);
8d67477f 5659 free_imsm(super);
8e59f3d8
AK		 5660 return 0;
5661 }
e683ca88 5662 memset(super->buf, 0, mpb_size);
ef649044 5663 mpb = super->buf;
e683ca88
DW		 5664 mpb->mpb_size = __cpu_to_le32(mpb_size);
5665 st->sb = super;
5666
5667 if (info == NULL) {
5668 /* zeroing superblock */
5669 return 0;
5670 }
bf5a934a 5671
4d1313e9
DW		 5672 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5673
5674 version = (char *) mpb->sig;
5675 strcpy(version, MPB_SIGNATURE);
5676 version += strlen(MPB_SIGNATURE);
5677 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5678
bf5a934a
DW		 5679 return 1;
5680}
5681
f2cc4f7d
AO		 5682static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5683{
5684 unsigned int member_sector_size;
5685
5686 if (dl->fd < 0) {
5687 pr_err("Invalid file descriptor for %s\n", dl->devname);
5688 return 0;
5689 }
5690
5691 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5692 return 0;
5693 if (member_sector_size != super->sector_size)
5694 return 0;
5695 return 1;
5696}
5697
f20c3968 5698static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5699 int fd, char *devname)
5700{
5701 struct intel_super *super = st->sb;
d23fe947 5702 struct imsm_super *mpb = super->anchor;
3960e579 5703 struct imsm_disk *_disk;
bf5a934a
DW		 5704 struct imsm_dev *dev;
5705 struct imsm_map *map;
3960e579 5706 struct dl *dl, *df;
4eb26970 5707 int slot;
bf5a934a 5708
949c47a0 5709 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5710 map = get_imsm_map(dev, MAP_0);
bf5a934a 5711
208933a7 5712 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5713 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5714 devname);
5715 return 1;
5716 }
5717
efb30e7f
DW		 5718 if (fd == -1) {
5719 /* we're doing autolayout so grab the pre-marked (in
5720 * validate_geometry) raid_disk
5721 */
5722 for (dl = super->disks; dl; dl = dl->next)
5723 if (dl->raiddisk == dk->raid_disk)
5724 break;
5725 } else {
5726 for (dl = super->disks; dl ; dl = dl->next)
5727 if (dl->major == dk->major &&
5728 dl->minor == dk->minor)
5729 break;
5730 }
d23fe947 5731
208933a7 5732 if (!dl) {
e7b84f9d 5733 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5734 return 1;
208933a7 5735 }
bf5a934a 5736
59632db9
MZ		 5737 if (mpb->num_disks == 0)
5738 if (!get_dev_sector_size(dl->fd, dl->devname,
5739 &super->sector_size))
5740 return 1;
5741
f2cc4f7d
AO		 5742 if (!drive_validate_sector_size(super, dl)) {
5743 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5744 return 1;
5745 }
5746
d23fe947
DW		 5747 /* add a pristine spare to the metadata */
5748 if (dl->index < 0) {
5749 dl->index = super->anchor->num_disks;
5750 super->anchor->num_disks++;
5751 }
4eb26970
DW		 5752 /* Check the device has not already been added */
5753 slot = get_imsm_disk_slot(map, dl->index);
5754 if (slot >= 0 &&
238c0a71 5755 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5756 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5757 devname);
5758 return 1;
5759 }
656b6b5a 5760 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5761 dl->disk.status = CONFIGURED_DISK;
d23fe947 5762
3960e579
DW		 5763 /* update size of 'missing' disks to be at least as large as the
5764 * largest acitve member (we only have dummy missing disks when
5765 * creating the first volume)
5766 */
5767 if (super->current_vol == 0) {
5768 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5769 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5770 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5771 _disk = __get_imsm_disk(mpb, df->index);
5772 *_disk = df->disk;
5773 }
5774 }
5775
5776 /* refresh unset/failed slots to point to valid 'missing' entries */
5777 for (df = super->missing; df; df = df->next)
5778 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5779 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5780
5781 if ((ord & IMSM_ORD_REBUILD) == 0)
5782 continue;
5783 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5784 if (is_gen_migration(dev)) {
238c0a71
AK		 5785 struct imsm_map *map2 = get_imsm_map(dev,
5786 MAP_1);
0a108d63 5787 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5788 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5789 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5790 slot2,
5791 MAP_1);
1ace8403
AK		 5792 if ((unsigned)df->index ==
5793 ord_to_idx(ord2))
5794 set_imsm_ord_tbl_ent(map2,
0a108d63 5795 slot2,
1ace8403
AK		 5796 df->index |
5797 IMSM_ORD_REBUILD);
5798 }
5799 }
3960e579
DW		 5800 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5801 break;
5802 }
5803
d23fe947
DW		 5804 /* if we are creating the first raid device update the family number */
5805 if (super->current_vol == 0) {
5806 __u32 sum;
5807 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5808
3960e579 5809 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5810 if (!_dev || !_disk) {
e7b84f9d 5811 pr_err("BUG mpb setup error\n");
791b666a
AW		 5812 return 1;
5813 }
d23fe947
DW		 5814 *_dev = *dev;
5815 *_disk = dl->disk;
148acb7b
DW		 5816 sum = random32();
5817 sum += __gen_imsm_checksum(mpb);
d23fe947 5818 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5819 mpb->orig_family_num = mpb->family_num;
d23fe947 5820 }
ca0748fa 5821 super->current_disk = dl;
f20c3968 5822 return 0;
bf5a934a
DW		 5823}
5824
a8619d23
AK		 5825/* mark_spare()
5826 * Function marks disk as spare and restores disk serial
5827 * in case it was previously marked as failed by takeover operation
5828 * reruns:
5829 * -1 : critical error
5830 * 0 : disk is marked as spare but serial is not set
5831 * 1 : success
5832 */
5833int mark_spare(struct dl *disk)
5834{
5835 __u8 serial[MAX_RAID_SERIAL_LEN];
5836 int ret_val = -1;
5837
5838 if (!disk)
5839 return ret_val;
5840
5841 ret_val = 0;
6da53c0e 5842 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5843 /* Restore disk serial number, because takeover marks disk
5844 * as failed and adds to serial ':0' before it becomes
5845 * a spare disk.
5846 */
5847 serialcpy(disk->serial, serial);
5848 serialcpy(disk->disk.serial, serial);
5849 ret_val = 1;
5850 }
5851 disk->disk.status = SPARE_DISK;
5852 disk->index = -1;
5853
5854 return ret_val;
5855}
88654014 5856
f20c3968 5857static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5858 int fd, char *devname,
5859 unsigned long long data_offset)
cdddbdbc 5860{
c2c087e6 5861 struct intel_super *super = st->sb;
c2c087e6
DW		 5862 struct dl *dd;
5863 unsigned long long size;
fa7bb6f8 5864 unsigned int member_sector_size;
f2f27e63 5865 __u32 id;
c2c087e6
DW		 5866 int rv;
5867 struct stat stb;
5868
88654014
LM		 5869 /* If we are on an RAID enabled platform check that the disk is
5870 * attached to the raid controller.
5871 * We do not need to test disks attachment for container based additions,
5872 * they shall be already tested when container was created/assembled.
88c32bb1 5873 */
d424212e 5874 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5875 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5876 if (rv != 0) {
5877 dprintf("capability: %p fd: %d ret: %d\n",
5878 super->orom, fd, rv);
5879 return 1;
88c32bb1
DW		 5880 }
5881
f20c3968
DW		 5882 if (super->current_vol >= 0)
5883 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5884
c2c087e6 5885 fstat(fd, &stb);
503975b9 5886 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5887 dd->major = major(stb.st_rdev);
5888 dd->minor = minor(stb.st_rdev);
503975b9 5889 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5890 dd->fd = fd;
689c9bf3 5891 dd->e = NULL;
1a64be56 5892 dd->action = DISK_ADD;
6da53c0e 5893 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5894 if (rv) {
e7b84f9d 5895 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5896 if (dd->devname)
5897 free(dd->devname);
949c47a0 5898 free(dd);
0030e8d6 5899 abort();
c2c087e6 5900 }
20bee0f8
PB		 5901 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5902 (super->hba->type == SYS_DEV_VMD))) {
5903 int i;
5904 char *devpath = diskfd_to_devpath(fd);
5905 char controller_path[PATH_MAX];
5906
5907 if (!devpath) {
5908 pr_err("failed to get devpath, aborting\n");
5909 if (dd->devname)
5910 free(dd->devname);
5911 free(dd);
5912 return 1;
5913 }
5914
5915 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5916 free(devpath);
5917
5918 if (devpath_to_vendor(controller_path) == 0x8086) {
5919 /*
5920 * If Intel's NVMe drive has serial ended with
5921 * "-A","-B","-1" or "-2" it means that this is "x8"
5922 * device (double drive on single PCIe card).
5923 * User should be warned about potential data loss.
5924 */
5925 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5926 /* Skip empty character at the end */
5927 if (dd->serial[i] == 0)
5928 continue;
5929
5930 if (((dd->serial[i] == 'A') ||
5931 (dd->serial[i] == 'B') ||
5932 (dd->serial[i] == '1') ||
5933 (dd->serial[i] == '2')) &&
5934 (dd->serial[i-1] == '-'))
5935 pr_err("\tThe action you are about to take may put your data at risk.\n"
5936 "\tPlease note that x8 devices may consist of two separate x4 devices "
5937 "located on a single PCIe port.\n"
5938 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5939 break;
5940 }
32716c51
PB		 5941 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5942 !imsm_orom_has_tpv_support(super->orom)) {
5943 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5944 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5945 free(dd->devname);
5946 free(dd);
5947 return 1;
20bee0f8
PB		 5948 }
5949 }
c2c087e6 5950
c2c087e6 5951 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5952 get_dev_sector_size(fd, NULL, &member_sector_size);
5953
5954 if (super->sector_size == 0) {
5955 /* this a first device, so sector_size is not set yet */
5956 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5957 }
5958
71e5411e 5959 /* clear migr_rec when adding disk to container */
85337573
AO		 5960 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5961 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5962 SEEK_SET) >= 0) {
466070ad 5963 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5964 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5965 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5966 perror("Write migr_rec failed");
5967 }
5968
c2c087e6 5969 size /= 512;
1f24f035 5970 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5971 set_total_blocks(&dd->disk, size);
5972 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5973 struct imsm_super *mpb = super->anchor;
5974 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5975 }
a8619d23 5976 mark_spare(dd);
c2c087e6 5977 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5978 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5979 else
b9f594fe 5980 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5981
5982 if (st->update_tail) {
1a64be56
LM		 5983 dd->next = super->disk_mgmt_list;
5984 super->disk_mgmt_list = dd;
43dad3d6
DW		 5985 } else {
5986 dd->next = super->disks;
5987 super->disks = dd;
ceaf0ee1 5988 super->updates_pending++;
43dad3d6 5989 }
f20c3968
DW		 5990
5991 return 0;
cdddbdbc
DW		 5992}
5993
1a64be56
LM		 5994static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5995{
5996 struct intel_super *super = st->sb;
5997 struct dl *dd;
5998
5999 /* remove from super works only in mdmon - for communication
6000 * manager - monitor. Check if communication memory buffer
6001 * is prepared.
6002 */
6003 if (!st->update_tail) {
1ade5cc1 6004 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 6005 return 1;
6006 }
503975b9 6007 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 6008 dd->major = dk->major;
6009 dd->minor = dk->minor;
1a64be56 6010 dd->fd = -1;
a8619d23 6011 mark_spare(dd);
1a64be56
LM		 6012 dd->action = DISK_REMOVE;
6013
6014 dd->next = super->disk_mgmt_list;
6015 super->disk_mgmt_list = dd;
6016
1a64be56
LM		 6017 return 0;
6018}
6019
f796af5d
DW		 6020static int store_imsm_mpb(int fd, struct imsm_super *mpb);
6021
6022static union {
f36a9ecd 6023 char buf[MAX_SECTOR_SIZE];
f796af5d 6024 struct imsm_super anchor;
f36a9ecd 6025} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 6026
d23fe947
DW		 6027/* spare records have their own family number and do not have any defined raid
6028 * devices
6029 */
6030static int write_super_imsm_spares(struct intel_super *super, int doclose)
6031{
d23fe947 6032 struct imsm_super *mpb = super->anchor;
f796af5d 6033 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 6034 __u32 sum;
6035 struct dl *d;
6036
68641cdb
JS		 6037 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6038 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6039 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6040 spare->num_disks = 1;
6041 spare->num_raid_devs = 0;
6042 spare->cache_size = mpb->cache_size;
6043 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6044
6045 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6046 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 6047
6048 for (d = super->disks; d; d = d->next) {
8796fdc4 6049 if (d->index != -1)
d23fe947
DW		 6050 continue;
6051
f796af5d 6052 spare->disk[0] = d->disk;
027c374f
CA		 6053 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6054 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6055
f36a9ecd
PB		 6056 if (super->sector_size == 4096)
6057 convert_to_4k_imsm_disk(&spare->disk[0]);
6058
f796af5d
DW		 6059 sum = __gen_imsm_checksum(spare);
6060 spare->family_num = __cpu_to_le32(sum);
6061 spare->orig_family_num = 0;
6062 sum = __gen_imsm_checksum(spare);
6063 spare->check_sum = __cpu_to_le32(sum);
d23fe947 6064
f796af5d 6065 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 6066 pr_err("failed for device %d:%d %s\n",
6067 d->major, d->minor, strerror(errno));
e74255d9 6068 return 1;
d23fe947
DW		 6069 }
6070 if (doclose) {
6071 close(d->fd);
6072 d->fd = -1;
6073 }
6074 }
6075
e74255d9 6076 return 0;
d23fe947
DW		 6077}
6078
36988a3d 6079static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6080{
36988a3d 6081 struct intel_super *super = st->sb;
f36a9ecd 6082 unsigned int sector_size = super->sector_size;
949c47a0 6083 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6084 struct dl *d;
6085 __u32 generation;
6086 __u32 sum;
d23fe947 6087 int spares = 0;
949c47a0 6088 int i;
a48ac0a8 6089 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6090 int num_disks = 0;
146c6260 6091 int clear_migration_record = 1;
bbab0940 6092 __u32 bbm_log_size;
cdddbdbc 6093
c2c087e6
DW		 6094 /* 'generation' is incremented everytime the metadata is written */
6095 generation = __le32_to_cpu(mpb->generation_num);
6096 generation++;
6097 mpb->generation_num = __cpu_to_le32(generation);
6098
148acb7b
DW		 6099 /* fix up cases where previous mdadm releases failed to set
6100 * orig_family_num
6101 */
6102 if (mpb->orig_family_num == 0)
6103 mpb->orig_family_num = mpb->family_num;
6104
d23fe947 6105 for (d = super->disks; d; d = d->next) {
8796fdc4 6106 if (d->index == -1)
d23fe947 6107 spares++;
36988a3d 6108 else {
d23fe947 6109 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6110 num_disks++;
6111 }
d23fe947 6112 }
36988a3d 6113 for (d = super->missing; d; d = d->next) {
47ee5a45 6114 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6115 num_disks++;
6116 }
6117 mpb->num_disks = num_disks;
6118 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6119
949c47a0
DW		 6120 for (i = 0; i < mpb->num_raid_devs; i++) {
6121 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6122 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6123 if (dev && dev2) {
6124 imsm_copy_dev(dev, dev2);
6125 mpb_size += sizeof_imsm_dev(dev, 0);
6126 }
146c6260
AK		 6127 if (is_gen_migration(dev2))
6128 clear_migration_record = 0;
949c47a0 6129 }
bbab0940
TM		 6130
6131 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6132
6133 if (bbm_log_size) {
6134 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6135 mpb->attributes |= MPB_ATTRIB_BBM;
6136 } else
6137 mpb->attributes &= ~MPB_ATTRIB_BBM;
6138
6139 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6140 mpb_size += bbm_log_size;
a48ac0a8 6141 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6142
bbab0940
TM		 6143#ifdef DEBUG
6144 assert(super->len == 0 || mpb_size <= super->len);
6145#endif
6146
c2c087e6 6147 /* recalculate checksum */
949c47a0 6148 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6149 mpb->check_sum = __cpu_to_le32(sum);
6150
51d83f5d
AK		 6151 if (super->clean_migration_record_by_mdmon) {
6152 clear_migration_record = 1;
6153 super->clean_migration_record_by_mdmon = 0;
6154 }
146c6260 6155 if (clear_migration_record)
de44e46f 6156 memset(super->migr_rec_buf, 0,
85337573 6157 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6158
f36a9ecd
PB		 6159 if (sector_size == 4096)
6160 convert_to_4k(super);
6161
d23fe947 6162 /* write the mpb for disks that compose raid devices */
c2c087e6 6163 for (d = super->disks; d ; d = d->next) {
86c54047 6164 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6165 continue;
30602f53 6166
146c6260
AK		 6167 if (clear_migration_record) {
6168 unsigned long long dsize;
6169
6170 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6171 if (lseek64(d->fd, dsize - sector_size,
6172 SEEK_SET) >= 0) {
466070ad
PB		 6173 if ((unsigned int)write(d->fd,
6174 super->migr_rec_buf,
de44e46f
PB		 6175 MIGR_REC_BUF_SECTORS*sector_size) !=
6176 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6177 perror("Write migr_rec failed");
146c6260
AK		 6178 }
6179 }
51d83f5d
AK		 6180
6181 if (store_imsm_mpb(d->fd, mpb))
6182 fprintf(stderr,
1ade5cc1
N		 6183 "failed for device %d:%d (fd: %d)%s\n",
6184 d->major, d->minor,
51d83f5d
AK		 6185 d->fd, strerror(errno));
6186
c2c087e6
DW		 6187 if (doclose) {
6188 close(d->fd);
6189 d->fd = -1;
6190 }
6191 }
6192
d23fe947
DW		 6193 if (spares)
6194 return write_super_imsm_spares(super, doclose);
6195
e74255d9 6196 return 0;
c2c087e6
DW		 6197}
6198
9b1fb677 6199static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6200{
6201 size_t len;
6202 struct imsm_update_create_array *u;
6203 struct intel_super *super = st->sb;
9b1fb677 6204 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6205 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6206 struct disk_info *inf;
6207 struct imsm_disk *disk;
6208 int i;
43dad3d6 6209
54c2c1ea
DW		 6210 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6211 sizeof(*inf) * map->num_members;
503975b9 6212 u = xmalloc(len);
43dad3d6 6213 u->type = update_create_array;
9b1fb677 6214 u->dev_idx = dev_idx;
43dad3d6 6215 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6216 inf = get_disk_info(u);
6217 for (i = 0; i < map->num_members; i++) {
238c0a71 6218 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6219
54c2c1ea 6220 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6221 if (!disk)
6222 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6223 serialcpy(inf[i].serial, disk->serial);
6224 }
43dad3d6
DW		 6225 append_metadata_update(st, u, len);
6226
6227 return 0;
6228}
6229
1a64be56 6230static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6231{
6232 struct intel_super *super = st->sb;
6233 size_t len;
1a64be56 6234 struct imsm_update_add_remove_disk *u;
43dad3d6 6235
1a64be56 6236 if (!super->disk_mgmt_list)
43dad3d6
DW		 6237 return 0;
6238
6239 len = sizeof(*u);
503975b9 6240 u = xmalloc(len);
1a64be56 6241 u->type = update_add_remove_disk;
43dad3d6
DW		 6242 append_metadata_update(st, u, len);
6243
6244 return 0;
6245}
2432ce9b
AP		 6246
6247__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6248
e397cefe
AP		 6249static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6250{
6251 struct ppl_header *ppl_hdr = buf;
6252 int ret;
6253
6254 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6255
6256 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6257 ret = -errno;
6258 perror("Failed to seek to PPL header location");
6259 return ret;
6260 }
6261
6262 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6263 ret = -errno;
6264 perror("Write PPL header failed");
6265 return ret;
6266 }
6267
6268 fsync(fd);
6269
6270 return 0;
6271}
6272
2432ce9b
AP		 6273static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6274{
6275 struct intel_super *super = st->sb;
6276 void *buf;
6277 struct ppl_header *ppl_hdr;
6278 int ret;
6279
b2514242
PB		 6280 /* first clear entire ppl space */
6281 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6282 if (ret)
6283 return ret;
6284
6285 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6286 if (ret) {
6287 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6288 return -ret;
2432ce9b
AP		 6289 }
6290
6291 memset(buf, 0, PPL_HEADER_SIZE);
6292 ppl_hdr = buf;
6293 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6294 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6295
6296 if (info->mismatch_cnt) {
6297 /*
6298 * We are overwriting an invalid ppl. Make one entry with wrong
6299 * checksum to prevent the kernel from skipping resync.
6300 */
6301 ppl_hdr->entries_count = __cpu_to_le32(1);
6302 ppl_hdr->entries[0].checksum = ~0;
6303 }
6304
e397cefe 6305 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6306
6307 free(buf);
6308 return ret;
6309}
6310
e397cefe
AP		 6311static int is_rebuilding(struct imsm_dev *dev);
6312
2432ce9b
AP		 6313static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6314 struct mdinfo *disk)
6315{
6316 struct intel_super *super = st->sb;
6317 struct dl *d;
e397cefe 6318 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6319 int ret = 0;
e397cefe 6320 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6321 __u32 crc;
6322 struct imsm_dev *dev;
2432ce9b 6323 __u32 idx;
44b6b876
PB		 6324 unsigned int i;
6325 unsigned long long ppl_offset = 0;
6326 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6327
6328 if (disk->disk.raid_disk < 0)
6329 return 0;
6330
2432ce9b 6331 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6332 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6333 d = get_imsm_dl_disk(super, idx);
6334
6335 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6336 return 0;
6337
6338 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6339 pr_err("Failed to allocate PPL header buffer\n");
6340 return -1;
6341 }
6342 buf = buf_orig;
2432ce9b 6343
44b6b876
PB		 6344 ret = 1;
6345 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6346 void *tmp;
6347
44b6b876 6348 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6349
44b6b876
PB		 6350 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6351 SEEK_SET) < 0) {
6352 perror("Failed to seek to PPL header location");
6353 ret = -1;
e397cefe 6354 break;
44b6b876 6355 }
2432ce9b 6356
44b6b876
PB		 6357 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6358 perror("Read PPL header failed");
6359 ret = -1;
e397cefe 6360 break;
44b6b876 6361 }
2432ce9b 6362
44b6b876 6363 ppl_hdr = buf;
2432ce9b 6364
44b6b876
PB		 6365 crc = __le32_to_cpu(ppl_hdr->checksum);
6366 ppl_hdr->checksum = 0;
6367
6368 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6369 dprintf("Wrong PPL header checksum on %s\n",
6370 d->devname);
e397cefe 6371 break;
44b6b876
PB		 6372 }
6373
6374 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6375 /* previous was newest, it was already checked */
e397cefe 6376 break;
44b6b876
PB		 6377 }
6378
6379 if ((__le32_to_cpu(ppl_hdr->signature) !=
6380 super->anchor->orig_family_num)) {
6381 dprintf("Wrong PPL header signature on %s\n",
6382 d->devname);
6383 ret = 1;
e397cefe 6384 break;
44b6b876
PB		 6385 }
6386
6387 ret = 0;
6388 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6389
44b6b876
PB		 6390 ppl_offset += PPL_HEADER_SIZE;
6391 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6392 ppl_offset +=
6393 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6394
6395 if (!buf_prev)
6396 buf_prev = buf + PPL_HEADER_SIZE;
6397 tmp = buf_prev;
6398 buf_prev = buf;
6399 buf = tmp;
2432ce9b
AP		 6400 }
6401
e397cefe
AP		 6402 if (buf_prev) {
6403 buf = buf_prev;
6404 ppl_hdr = buf_prev;
6405 }
2432ce9b 6406
54148aba
PB		 6407 /*
6408 * Update metadata to use mutliple PPLs area (1MB).
6409 * This is done once for all RAID members
6410 */
6411 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6412 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6413 char subarray[20];
6414 struct mdinfo *member_dev;
6415
6416 sprintf(subarray, "%d", info->container_member);
6417
6418 if (mdmon_running(st->container_devnm))
6419 st->update_tail = &st->updates;
6420
6421 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6422 pr_err("Failed to update subarray %s\n",
6423 subarray);
6424 } else {
6425 if (st->update_tail)
6426 flush_metadata_updates(st);
6427 else
6428 st->ss->sync_metadata(st);
6429 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6430 for (member_dev = info->devs; member_dev;
6431 member_dev = member_dev->next)
6432 member_dev->ppl_size =
6433 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6434 }
6435 }
6436
b23d0750 6437 if (ret == 1) {
2fc0fc63
AP		 6438 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6439
50b9c10d
PB		 6440 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6441 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6442 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6443 (is_rebuilding(dev) &&
2ec9d182
AP		 6444 dev->vol.curr_migr_unit == 0 &&
6445 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6446 ret = st->ss->write_init_ppl(st, info, d->fd);
6447 else
6448 info->mismatch_cnt++;
e397cefe
AP		 6449 } else if (ret == 0 &&
6450 ppl_hdr->entries_count == 0 &&
6451 is_rebuilding(dev) &&
6452 info->resync_start == 0) {
6453 /*
6454 * The header has no entries - add a single empty entry and
6455 * rewrite the header to prevent the kernel from going into
6456 * resync after an interrupted rebuild.
6457 */
6458 ppl_hdr->entries_count = __cpu_to_le32(1);
6459 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6460 }
2432ce9b 6461
e397cefe
AP		 6462 free(buf_orig);
6463
2432ce9b
AP		 6464 return ret;
6465}
6466
2432ce9b
AP		 6467static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6468{
6469 struct intel_super *super = st->sb;
6470 struct dl *d;
6471 int ret = 0;
6472
6473 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6474 info->array.level != 5)
6475 return 0;
6476
6477 for (d = super->disks; d ; d = d->next) {
6478 if (d->index < 0 || is_failed(&d->disk))
6479 continue;
6480
6481 ret = st->ss->write_init_ppl(st, info, d->fd);
6482 if (ret)
6483 break;
6484 }
6485
6486 return ret;
6487}
43dad3d6 6488
c2c087e6
DW		 6489static int write_init_super_imsm(struct supertype *st)
6490{
9b1fb677
DW		 6491 struct intel_super *super = st->sb;
6492 int current_vol = super->current_vol;
2432ce9b
AP		 6493 int rv = 0;
6494 struct mdinfo info;
6495
6496 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6497
6498 /* we are done with current_vol reset it to point st at the container */
6499 super->current_vol = -1;
6500
8273f55e 6501 if (st->update_tail) {
43dad3d6
DW		 6502 /* queue the recently created array / added disk
6503 * as a metadata update */
8273f55e 6504
43dad3d6 6505 /* determine if we are creating a volume or adding a disk */
9b1fb677 6506 if (current_vol < 0) {
1a64be56
LM		 6507 /* in the mgmt (add/remove) disk case we are running
6508 * in mdmon context, so don't close fd's
43dad3d6 6509 */
2432ce9b
AP		 6510 rv = mgmt_disk(st);
6511 } else {
6512 rv = write_init_ppl_imsm_all(st, &info);
6513 if (!rv)
6514 rv = create_array(st, current_vol);
6515 }
d682f344
N		 6516 } else {
6517 struct dl *d;
6518 for (d = super->disks; d; d = d->next)
ba728be7 6519 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6520 if (current_vol >= 0)
6521 rv = write_init_ppl_imsm_all(st, &info);
6522 if (!rv)
6523 rv = write_super_imsm(st, 1);
d682f344 6524 }
2432ce9b
AP		 6525
6526 return rv;
cdddbdbc
DW		 6527}
6528
e683ca88 6529static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6530{
e683ca88
DW		 6531 struct intel_super *super = st->sb;
6532 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6533
e683ca88 6534 if (!mpb)
ad97895e
DW		 6535 return 1;
6536
f36a9ecd
PB		 6537 if (super->sector_size == 4096)
6538 convert_to_4k(super);
e683ca88 6539 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6540}
6541
cdddbdbc
DW		 6542static int validate_geometry_imsm_container(struct supertype *st, int level,
6543 int layout, int raiddisks, int chunk,
af4348dd
N		 6544 unsigned long long size,
6545 unsigned long long data_offset,
6546 char *dev,
2c514b71
NB		 6547 unsigned long long *freesize,
6548 int verbose)
cdddbdbc 6549{
c2c087e6
DW		 6550 int fd;
6551 unsigned long long ldsize;
594dc1b8 6552 struct intel_super *super;
f2f5c343 6553 int rv = 0;
cdddbdbc 6554
c2c087e6
DW		 6555 if (level != LEVEL_CONTAINER)
6556 return 0;
6557 if (!dev)
6558 return 1;
6559
6560 fd = open(dev, O_RDONLY|O_EXCL, 0);
6561 if (fd < 0) {
ba728be7 6562 if (verbose > 0)
e7b84f9d 6563 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6564 dev, strerror(errno));
c2c087e6
DW		 6565 return 0;
6566 }
6567 if (!get_dev_size(fd, dev, &ldsize)) {
6568 close(fd);
6569 return 0;
6570 }
f2f5c343
LM		 6571
6572 /* capabilities retrieve could be possible
6573 * note that there is no fd for the disks in array.
6574 */
6575 super = alloc_super();
8d67477f
TM		 6576 if (!super) {
6577 close(fd);
6578 return 0;
6579 }
fa7bb6f8
PB		 6580 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6581 close(fd);
6582 free_imsm(super);
6583 return 0;
6584 }
6585
ba728be7 6586 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6587 if (rv != 0) {
6588#if DEBUG
6589 char str[256];
6590 fd2devname(fd, str);
1ade5cc1 6591 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6592 fd, str, super->orom, rv, raiddisks);
6593#endif
6594 /* no orom/efi or non-intel hba of the disk */
6595 close(fd);
6596 free_imsm(super);
6597 return 0;
6598 }
c2c087e6 6599 close(fd);
9126b9a8
CA		 6600 if (super->orom) {
6601 if (raiddisks > super->orom->tds) {
6602 if (verbose)
7a862a02 6603 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6604 raiddisks, super->orom->tds);
6605 free_imsm(super);
6606 return 0;
6607 }
6608 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6609 (ldsize >> 9) >> 32 > 0) {
6610 if (verbose)
e7b84f9d 6611 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6612 free_imsm(super);
6613 return 0;
6614 }
f2f5c343 6615 }
c2c087e6 6616
af4348dd 6617 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6618 free_imsm(super);
c2c087e6
DW		 6619
6620 return 1;
cdddbdbc
DW		 6621}
6622
0dcecb2e
DW		 6623static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6624{
6625 const unsigned long long base_start = e[*idx].start;
6626 unsigned long long end = base_start + e[*idx].size;
6627 int i;
6628
6629 if (base_start == end)
6630 return 0;
6631
6632 *idx = *idx + 1;
6633 for (i = *idx; i < num_extents; i++) {
6634 /* extend overlapping extents */
6635 if (e[i].start >= base_start &&
6636 e[i].start <= end) {
6637 if (e[i].size == 0)
6638 return 0;
6639 if (e[i].start + e[i].size > end)
6640 end = e[i].start + e[i].size;
6641 } else if (e[i].start > end) {
6642 *idx = i;
6643 break;
6644 }
6645 }
6646
6647 return end - base_start;
6648}
6649
6650static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6651{
6652 /* build a composite disk with all known extents and generate a new
6653 * 'maxsize' given the "all disks in an array must share a common start
6654 * offset" constraint
6655 */
503975b9 6656 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6657 struct dl *dl;
6658 int i, j;
6659 int start_extent;
6660 unsigned long long pos;
b9d77223 6661 unsigned long long start = 0;
0dcecb2e
DW		 6662 unsigned long long maxsize;
6663 unsigned long reserve;
6664
0dcecb2e
DW		 6665 /* coalesce and sort all extents. also, check to see if we need to
6666 * reserve space between member arrays
6667 */
6668 j = 0;
6669 for (dl = super->disks; dl; dl = dl->next) {
6670 if (!dl->e)
6671 continue;
6672 for (i = 0; i < dl->extent_cnt; i++)
6673 e[j++] = dl->e[i];
6674 }
6675 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6676
6677 /* merge extents */
6678 i = 0;
6679 j = 0;
6680 while (i < sum_extents) {
6681 e[j].start = e[i].start;
6682 e[j].size = find_size(e, &i, sum_extents);
6683 j++;
6684 if (e[j-1].size == 0)
6685 break;
6686 }
6687
6688 pos = 0;
6689 maxsize = 0;
6690 start_extent = 0;
6691 i = 0;
6692 do {
6693 unsigned long long esize;
6694
6695 esize = e[i].start - pos;
6696 if (esize >= maxsize) {
6697 maxsize = esize;
6698 start = pos;
6699 start_extent = i;
6700 }
6701 pos = e[i].start + e[i].size;
6702 i++;
6703 } while (e[i-1].size);
6704 free(e);
6705
a7dd165b
DW		 6706 if (maxsize == 0)
6707 return 0;
6708
6709 /* FIXME assumes volume at offset 0 is the first volume in a
6710 * container
6711 */
0dcecb2e
DW		 6712 if (start_extent > 0)
6713 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6714 else
6715 reserve = 0;
6716
6717 if (maxsize < reserve)
a7dd165b 6718 return 0;
0dcecb2e 6719
5551b113 6720 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6721 if (start + reserve > super->create_offset)
a7dd165b 6722 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6723 super->create_offset = start + reserve;
6724
6725 return maxsize - reserve;
6726}
6727
88c32bb1
DW		 6728static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6729{
6730 if (level < 0 || level == 6 || level == 4)
6731 return 0;
6732
6733 /* if we have an orom prevent invalid raid levels */
6734 if (orom)
6735 switch (level) {
6736 case 0: return imsm_orom_has_raid0(orom);
6737 case 1:
6738 if (raiddisks > 2)
6739 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6740 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6741 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6742 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6743 }
6744 else
6745 return 1; /* not on an Intel RAID platform so anything goes */
6746
6747 return 0;
6748}
6749
ca9de185
LM		 6750static int
6751active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6752 int dpa, int verbose)
6753{
6754 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6755 struct mdstat_ent *memb;
ca9de185
LM		 6756 int count = 0;
6757 int num = 0;
594dc1b8 6758 struct md_list *dv;
ca9de185
LM		 6759 int found;
6760
6761 for (memb = mdstat ; memb ; memb = memb->next) {
6762 if (memb->metadata_version &&
fc54fe7a 6763 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6764 (strcmp(&memb->metadata_version[9], name) == 0) &&
6765 !is_subarray(memb->metadata_version+9) &&
6766 memb->members) {
6767 struct dev_member *dev = memb->members;
6768 int fd = -1;
6769 while(dev && (fd < 0)) {
503975b9
N		 6770 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6771 num = sprintf(path, "%s%s", "/dev/", dev->name);
6772 if (num > 0)
6773 fd = open(path, O_RDONLY, 0);
089f9d79 6774 if (num <= 0 || fd < 0) {
676e87a8 6775 pr_vrb("Cannot open %s: %s\n",
503975b9 6776 dev->name, strerror(errno));
ca9de185 6777 }
503975b9 6778 free(path);
ca9de185
LM		 6779 dev = dev->next;
6780 }
6781 found = 0;
089f9d79 6782 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6783 struct mdstat_ent *vol;
6784 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6785 if (vol->active > 0 &&
ca9de185 6786 vol->metadata_version &&
9581efb1 6787 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6788 found++;
6789 count++;
6790 }
6791 }
6792 if (*devlist && (found < dpa)) {
503975b9 6793 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6794 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6795 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6796 dv->found = found;
6797 dv->used = 0;
6798 dv->next = *devlist;
6799 *devlist = dv;
ca9de185
LM		 6800 }
6801 }
6802 if (fd >= 0)
6803 close(fd);
6804 }
6805 }
6806 free_mdstat(mdstat);
6807 return count;
6808}
6809
6810#ifdef DEBUG_LOOP
6811static struct md_list*
6812get_loop_devices(void)
6813{
6814 int i;
6815 struct md_list *devlist = NULL;
594dc1b8 6816 struct md_list *dv;
ca9de185
LM		 6817
6818 for(i = 0; i < 12; i++) {
503975b9
N		 6819 dv = xcalloc(1, sizeof(*dv));
6820 dv->devname = xmalloc(40);
ca9de185
LM		 6821 sprintf(dv->devname, "/dev/loop%d", i);
6822 dv->next = devlist;
6823 devlist = dv;
6824 }
6825 return devlist;
6826}
6827#endif
6828
6829static struct md_list*
6830get_devices(const char *hba_path)
6831{
6832 struct md_list *devlist = NULL;
594dc1b8 6833 struct md_list *dv;
ca9de185
LM		 6834 struct dirent *ent;
6835 DIR *dir;
6836 int err = 0;
6837
6838#if DEBUG_LOOP
6839 devlist = get_loop_devices();
6840 return devlist;
6841#endif
6842 /* scroll through /sys/dev/block looking for devices attached to
6843 * this hba
6844 */
6845 dir = opendir("/sys/dev/block");
6846 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6847 int fd;
6848 char buf[1024];
6849 int major, minor;
6850 char *path = NULL;
6851 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6852 continue;
6853 path = devt_to_devpath(makedev(major, minor));
6854 if (!path)
6855 continue;
6856 if (!path_attached_to_hba(path, hba_path)) {
6857 free(path);
6858 path = NULL;
6859 continue;
6860 }
6861 free(path);
6862 path = NULL;
6863 fd = dev_open(ent->d_name, O_RDONLY);
6864 if (fd >= 0) {
6865 fd2devname(fd, buf);
6866 close(fd);
6867 } else {
e7b84f9d 6868 pr_err("cannot open device: %s\n",
ca9de185
LM		 6869 ent->d_name);
6870 continue;
6871 }
6872
503975b9
N		 6873 dv = xcalloc(1, sizeof(*dv));
6874 dv->devname = xstrdup(buf);
ca9de185
LM		 6875 dv->next = devlist;
6876 devlist = dv;
6877 }
6878 if (err) {
6879 while(devlist) {
6880 dv = devlist;
6881 devlist = devlist->next;
6882 free(dv->devname);
6883 free(dv);
6884 }
6885 }
562aa102 6886 closedir(dir);
ca9de185
LM		 6887 return devlist;
6888}
6889
6890static int
6891count_volumes_list(struct md_list *devlist, char *homehost,
6892 int verbose, int *found)
6893{
6894 struct md_list *tmpdev;
6895 int count = 0;
594dc1b8 6896 struct supertype *st;
ca9de185
LM		 6897
6898 /* first walk the list of devices to find a consistent set
6899 * that match the criterea, if that is possible.
6900 * We flag the ones we like with 'used'.
6901 */
6902 *found = 0;
6903 st = match_metadata_desc_imsm("imsm");
6904 if (st == NULL) {
676e87a8 6905 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6906 return 0;
6907 }
6908
6909 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6910 char *devname = tmpdev->devname;
0a6bff09 6911 dev_t rdev;
ca9de185
LM		 6912 struct supertype *tst;
6913 int dfd;
6914 if (tmpdev->used > 1)
6915 continue;
6916 tst = dup_super(st);
6917 if (tst == NULL) {
676e87a8 6918 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6919 goto err_1;
6920 }
6921 tmpdev->container = 0;
6922 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6923 if (dfd < 0) {
1ade5cc1 6924 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6925 devname, strerror(errno));
6926 tmpdev->used = 2;
0a6bff09 6927 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6928 tmpdev->used = 2;
6929 } else if (must_be_container(dfd)) {
6930 struct supertype *cst;
6931 cst = super_by_fd(dfd, NULL);
6932 if (cst == NULL) {
1ade5cc1 6933 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6934 devname);
6935 tmpdev->used = 2;
6936 } else if (tst->ss != st->ss) {
1ade5cc1 6937 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6938 devname);
6939 tmpdev->used = 2;
6940 } else if (!tst->ss->load_container ||
6941 tst->ss->load_container(tst, dfd, NULL))
6942 tmpdev->used = 2;
6943 else {
6944 tmpdev->container = 1;
6945 }
6946 if (cst)
6947 cst->ss->free_super(cst);
6948 } else {
0a6bff09 6949 tmpdev->st_rdev = rdev;
ca9de185 6950 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6951 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6952 devname);
6953 tmpdev->used = 2;
6954 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6955 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6956 tst->ss->name, devname);
6957 tmpdev->used = 2;
6958 }
6959 }
6960 if (dfd >= 0)
6961 close(dfd);
6962 if (tmpdev->used == 2 || tmpdev->used == 4) {
6963 /* Ignore unrecognised devices during auto-assembly */
6964 goto loop;
6965 }
6966 else {
6967 struct mdinfo info;
6968 tst->ss->getinfo_super(tst, &info, NULL);
6969
6970 if (st->minor_version == -1)
6971 st->minor_version = tst->minor_version;
6972
6973 if (memcmp(info.uuid, uuid_zero,
6974 sizeof(int[4])) == 0) {
6975 /* this is a floating spare. It cannot define
6976 * an array unless there are no more arrays of
6977 * this type to be found. It can be included
6978 * in an array of this type though.
6979 */
6980 tmpdev->used = 3;
6981 goto loop;
6982 }
6983
6984 if (st->ss != tst->ss ||
6985 st->minor_version != tst->minor_version ||
6986 st->ss->compare_super(st, tst) != 0) {
6987 /* Some mismatch. If exactly one array matches this host,
6988 * we can resolve on that one.
6989 * Or, if we are auto assembling, we just ignore the second
6990 * for now.
6991 */
1ade5cc1 6992 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6993 devname);
6994 goto loop;
6995 }
6996 tmpdev->used = 1;
6997 *found = 1;
6998 dprintf("found: devname: %s\n", devname);
6999 }
7000 loop:
7001 if (tst)
7002 tst->ss->free_super(tst);
7003 }
7004 if (*found != 0) {
7005 int err;
7006 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
7007 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
7008 for (iter = head; iter; iter = iter->next) {
7009 dprintf("content->text_version: %s vol\n",
7010 iter->text_version);
7011 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
7012 /* do not assemble arrays with unsupported
7013 configurations */
1ade5cc1 7014 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 7015 iter->text_version);
7016 } else
7017 count++;
7018 }
7019 sysfs_free(head);
7020
7021 } else {
1ade5cc1 7022 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7023 err, st->sb);
7024 }
7025 } else {
1ade5cc1 7026 dprintf("no more devices to examine\n");
ca9de185
LM		 7027 }
7028
7029 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7030 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7031 if (count) {
7032 if (count < tmpdev->found)
7033 count = 0;
7034 else
7035 count -= tmpdev->found;
7036 }
7037 }
7038 if (tmpdev->used == 1)
7039 tmpdev->used = 4;
7040 }
7041 err_1:
7042 if (st)
7043 st->ss->free_super(st);
7044 return count;
7045}
7046
d3c11416
AO		 7047static int __count_volumes(char *hba_path, int dpa, int verbose,
7048 int cmp_hba_path)
ca9de185 7049{
72a45777 7050 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7051 int count = 0;
72a45777
PB		 7052 const struct orom_entry *entry;
7053 struct devid_list *dv, *devid_list;
ca9de185 7054
d3c11416 7055 if (!hba_path)
ca9de185
LM		 7056 return 0;
7057
72a45777 7058 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7059 if (strstr(idev->path, hba_path))
7060 break;
72a45777
PB		 7061 }
7062
7063 if (!idev || !idev->dev_id)
ca9de185 7064 return 0;
72a45777
PB		 7065
7066 entry = get_orom_entry_by_device_id(idev->dev_id);
7067
7068 if (!entry || !entry->devid_list)
7069 return 0;
7070
7071 devid_list = entry->devid_list;
7072 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7073 struct md_list *devlist;
d3c11416
AO		 7074 struct sys_dev *device = NULL;
7075 char *hpath;
72a45777
PB		 7076 int found = 0;
7077
d3c11416
AO		 7078 if (cmp_hba_path)
7079 device = device_by_id_and_path(dv->devid, hba_path);
7080 else
7081 device = device_by_id(dv->devid);
7082
72a45777 7083 if (device)
d3c11416 7084 hpath = device->path;
72a45777
PB		 7085 else
7086 return 0;
7087
d3c11416 7088 devlist = get_devices(hpath);
72a45777
PB		 7089 /* if no intel devices return zero volumes */
7090 if (devlist == NULL)
7091 return 0;
7092
d3c11416
AO		 7093 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7094 verbose);
7095 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7096 if (devlist == NULL)
7097 return 0;
7098 do {
7099 found = 0;
7100 count += count_volumes_list(devlist,
7101 NULL,
7102 verbose,
7103 &found);
7104 dprintf("found %d count: %d\n", found, count);
7105 } while (found);
7106
d3c11416 7107 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7108
7109 while (devlist) {
7110 struct md_list *dv = devlist;
7111 devlist = devlist->next;
7112 free(dv->devname);
7113 free(dv);
7114 }
ca9de185
LM		 7115 }
7116 return count;
7117}
7118
d3c11416
AO		 7119static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7120{
7121 if (!hba)
7122 return 0;
7123 if (hba->type == SYS_DEV_VMD) {
7124 struct sys_dev *dev;
7125 int count = 0;
7126
7127 for (dev = find_intel_devices(); dev; dev = dev->next) {
7128 if (dev->type == SYS_DEV_VMD)
7129 count += __count_volumes(dev->path, dpa,
7130 verbose, 1);
7131 }
7132 return count;
7133 }
7134 return __count_volumes(hba->path, dpa, verbose, 0);
7135}
7136
cd9d1ac7
DW		 7137static int imsm_default_chunk(const struct imsm_orom *orom)
7138{
7139 /* up to 512 if the plaform supports it, otherwise the platform max.
7140 * 128 if no platform detected
7141 */
7142 int fs = max(7, orom ? fls(orom->sss) : 0);
7143
7144 return min(512, (1 << fs));
7145}
73408129 7146
6592ce37
DW		 7147static int
7148validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7149 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7150{
660260d0
DW		 7151 /* check/set platform and metadata limits/defaults */
7152 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7153 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7154 super->orom->dpa);
73408129
LM		 7155 return 0;
7156 }
7157
5d500228 7158 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7159 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7160 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7161 level, raiddisks, raiddisks > 1 ? "s" : "");
7162 return 0;
7163 }
cd9d1ac7 7164
7ccc4cc4 7165 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7166 *chunk = imsm_default_chunk(super->orom);
7167
7ccc4cc4 7168 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7169 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7170 return 0;
6592ce37 7171 }
cd9d1ac7 7172
6592ce37
DW		 7173 if (layout != imsm_level_to_layout(level)) {
7174 if (level == 5)
676e87a8 7175 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7176 else if (level == 10)
676e87a8 7177 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7178 else
676e87a8 7179 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7180 layout, level);
7181 return 0;
7182 }
2cc699af 7183
7ccc4cc4 7184 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7185 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7186 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7187 return 0;
7188 }
614902f6 7189
6592ce37
DW		 7190 return 1;
7191}
7192
1011e834 7193/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7194 * FIX ME add ahci details
7195 */
8b353278 7196static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7197 int layout, int raiddisks, int *chunk,
af4348dd
N		 7198 unsigned long long size,
7199 unsigned long long data_offset,
7200 char *dev,
2c514b71
NB		 7201 unsigned long long *freesize,
7202 int verbose)
cdddbdbc 7203{
9e04ac1c 7204 dev_t rdev;
c2c087e6 7205 struct intel_super *super = st->sb;
b2916f25 7206 struct imsm_super *mpb;
c2c087e6
DW		 7207 struct dl *dl;
7208 unsigned long long pos = 0;
7209 unsigned long long maxsize;
7210 struct extent *e;
7211 int i;
cdddbdbc 7212
88c32bb1
DW		 7213 /* We must have the container info already read in. */
7214 if (!super)
c2c087e6
DW		 7215 return 0;
7216
b2916f25
JS		 7217 mpb = super->anchor;
7218
2cc699af 7219 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7220 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7221 return 0;
d54559f0 7222 }
c2c087e6
DW		 7223 if (!dev) {
7224 /* General test: make sure there is space for
2da8544a
DW		 7225 * 'raiddisks' device extents of size 'size' at a given
7226 * offset
c2c087e6 7227 */
e46273eb 7228 unsigned long long minsize = size;
b7528a20 7229 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7230 int dcnt = 0;
7231 if (minsize == 0)
7232 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7233 for (dl = super->disks; dl ; dl = dl->next) {
7234 int found = 0;
7235
bf5a934a 7236 pos = 0;
c2c087e6 7237 i = 0;
05501181 7238 e = get_extents(super, dl, 0);
c2c087e6
DW		 7239 if (!e) continue;
7240 do {
7241 unsigned long long esize;
7242 esize = e[i].start - pos;
7243 if (esize >= minsize)
7244 found = 1;
b7528a20 7245 if (found && start_offset == MaxSector) {
2da8544a
DW		 7246 start_offset = pos;
7247 break;
7248 } else if (found && pos != start_offset) {
7249 found = 0;
7250 break;
7251 }
c2c087e6
DW		 7252 pos = e[i].start + e[i].size;
7253 i++;
7254 } while (e[i-1].size);
7255 if (found)
7256 dcnt++;
7257 free(e);
7258 }
7259 if (dcnt < raiddisks) {
2c514b71 7260 if (verbose)
7a862a02 7261 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7262 dcnt, raiddisks);
c2c087e6
DW		 7263 return 0;
7264 }
7265 return 1;
7266 }
0dcecb2e 7267
c2c087e6 7268 /* This device must be a member of the set */
9e04ac1c 7269 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7270 return 0;
7271 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7272 if (dl->major == (int)major(rdev) &&
7273 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7274 break;
7275 }
7276 if (!dl) {
2c514b71 7277 if (verbose)
7a862a02 7278 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7279 return 0;
a20d2ba5
DW		 7280 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7281 /* If a volume is present then the current creation attempt
7282 * cannot incorporate new spares because the orom may not
7283 * understand this configuration (all member disks must be
7284 * members of each array in the container).
7285 */
7a862a02
N		 7286 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7287 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7288 return 0;
5fe62b94
WD		 7289 } else if (super->orom && mpb->num_raid_devs > 0 &&
7290 mpb->num_disks != raiddisks) {
7a862a02 7291 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7292 return 0;
c2c087e6 7293 }
0dcecb2e
DW		 7294
7295 /* retrieve the largest free space block */
05501181 7296 e = get_extents(super, dl, 0);
c2c087e6
DW		 7297 maxsize = 0;
7298 i = 0;
0dcecb2e
DW		 7299 if (e) {
7300 do {
7301 unsigned long long esize;
7302
7303 esize = e[i].start - pos;
7304 if (esize >= maxsize)
7305 maxsize = esize;
7306 pos = e[i].start + e[i].size;
7307 i++;
7308 } while (e[i-1].size);
7309 dl->e = e;
7310 dl->extent_cnt = i;
7311 } else {
7312 if (verbose)
e7b84f9d 7313 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7314 dev);
7315 return 0;
7316 }
7317 if (maxsize < size) {
7318 if (verbose)
e7b84f9d 7319 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7320 dev, maxsize, size);
7321 return 0;
7322 }
7323
7324 /* count total number of extents for merge */
7325 i = 0;
7326 for (dl = super->disks; dl; dl = dl->next)
7327 if (dl->e)
7328 i += dl->extent_cnt;
7329
7330 maxsize = merge_extents(super, i);
3baa56ab 7331
1a1ced1e
KS		 7332 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7333 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7334
a7dd165b 7335 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7336 if (verbose) {
7337 if (maxsize == 0)
7a862a02 7338 pr_err("no free space left on device. Aborting...\n");
b3071342 7339 else
7a862a02 7340 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7341 maxsize, size);
7342 }
0dcecb2e 7343 return 0;
0dcecb2e
DW		 7344 }
7345
c2c087e6
DW		 7346 *freesize = maxsize;
7347
ca9de185 7348 if (super->orom) {
72a45777 7349 int count = count_volumes(super->hba,
ca9de185
LM		 7350 super->orom->dpa, verbose);
7351 if (super->orom->vphba <= count) {
676e87a8 7352 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7353 super->orom->vphba);
7354 return 0;
7355 }
7356 }
c2c087e6 7357 return 1;
cdddbdbc
DW		 7358}
7359
13bcac90 7360static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7361 unsigned long long size, int chunk,
7362 unsigned long long *freesize)
7363{
7364 struct intel_super *super = st->sb;
7365 struct imsm_super *mpb = super->anchor;
7366 struct dl *dl;
7367 int i;
7368 int extent_cnt;
7369 struct extent *e;
7370 unsigned long long maxsize;
7371 unsigned long long minsize;
7372 int cnt;
7373 int used;
7374
7375 /* find the largest common start free region of the possible disks */
7376 used = 0;
7377 extent_cnt = 0;
7378 cnt = 0;
7379 for (dl = super->disks; dl; dl = dl->next) {
7380 dl->raiddisk = -1;
7381
7382 if (dl->index >= 0)
7383 used++;
7384
7385 /* don't activate new spares if we are orom constrained
7386 * and there is already a volume active in the container
7387 */
7388 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7389 continue;
7390
05501181 7391 e = get_extents(super, dl, 0);
efb30e7f
DW		 7392 if (!e)
7393 continue;
7394 for (i = 1; e[i-1].size; i++)
7395 ;
7396 dl->e = e;
7397 dl->extent_cnt = i;
7398 extent_cnt += i;
7399 cnt++;
7400 }
7401
7402 maxsize = merge_extents(super, extent_cnt);
7403 minsize = size;
7404 if (size == 0)
612e59d8
CA		 7405 /* chunk is in K */
7406 minsize = chunk * 2;
efb30e7f
DW		 7407
7408 if (cnt < raiddisks ||
7409 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7410 maxsize < minsize ||
7411 maxsize == 0) {
e7b84f9d 7412 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7413 return 0; /* No enough free spaces large enough */
7414 }
7415
7416 if (size == 0) {
7417 size = maxsize;
7418 if (chunk) {
612e59d8
CA		 7419 size /= 2 * chunk;
7420 size *= 2 * chunk;
efb30e7f 7421 }
f878b242
LM		 7422 maxsize = size;
7423 }
1a1ced1e
KS		 7424 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7425 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7426 cnt = 0;
7427 for (dl = super->disks; dl; dl = dl->next)
7428 if (dl->e)
7429 dl->raiddisk = cnt++;
7430
7431 *freesize = size;
7432
13bcac90
AK		 7433 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7434
efb30e7f
DW		 7435 return 1;
7436}
7437
13bcac90
AK		 7438static int reserve_space(struct supertype *st, int raiddisks,
7439 unsigned long long size, int chunk,
7440 unsigned long long *freesize)
7441{
7442 struct intel_super *super = st->sb;
7443 struct dl *dl;
7444 int cnt;
7445 int rv = 0;
7446
7447 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7448 if (rv) {
7449 cnt = 0;
7450 for (dl = super->disks; dl; dl = dl->next)
7451 if (dl->e)
7452 dl->raiddisk = cnt++;
7453 rv = 1;
7454 }
7455
7456 return rv;
7457}
7458
bf5a934a 7459static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7460 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7461 unsigned long long data_offset,
bf5a934a 7462 char *dev, unsigned long long *freesize,
5308f117 7463 int consistency_policy, int verbose)
bf5a934a
DW		 7464{
7465 int fd, cfd;
7466 struct mdinfo *sra;
20cbe8d2 7467 int is_member = 0;
bf5a934a 7468
d54559f0
LM		 7469 /* load capability
7470 * if given unused devices create a container
bf5a934a
DW		 7471 * if given given devices in a container create a member volume
7472 */
7473 if (level == LEVEL_CONTAINER) {
7474 /* Must be a fresh device to add to a container */
7475 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7476 raiddisks,
7ccc4cc4 7477 *chunk,
af4348dd 7478 size, data_offset,
bf5a934a
DW		 7479 dev, freesize,
7480 verbose);
7481 }
9587c373 7482
22dc741f
MT		 7483 if (size && (size < 1024)) {
7484 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7485 /* Depends on algorithm in Create.c :
7486 * if container was given (dev == NULL) return -1,
7487 * if block device was given ( dev != NULL) return 0.
7488 */
7489 return dev ? -1 : 0;
7490 }
7491
8592f29d 7492 if (!dev) {
e91a3bad 7493 if (st->sb) {
ca9de185 7494 struct intel_super *super = st->sb;
e91a3bad 7495 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7496 raiddisks, chunk, size,
e91a3bad
LM		 7497 verbose))
7498 return 0;
efb30e7f
DW		 7499 /* we are being asked to automatically layout a
7500 * new volume based on the current contents of
7501 * the container. If the the parameters can be
7502 * satisfied reserve_space will record the disks,
7503 * start offset, and size of the volume to be
7504 * created. add_to_super and getinfo_super
7505 * detect when autolayout is in progress.
7506 */
ca9de185
LM		 7507 /* assuming that freesize is always given when array is
7508 created */
7509 if (super->orom && freesize) {
7510 int count;
72a45777 7511 count = count_volumes(super->hba,
ca9de185
LM		 7512 super->orom->dpa, verbose);
7513 if (super->orom->vphba <= count) {
676e87a8 7514 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7515 super->orom->vphba);
7516 return 0;
7517 }
7518 }
e91a3bad
LM		 7519 if (freesize)
7520 return reserve_space(st, raiddisks, size,
7ccc4cc4 7521 *chunk, freesize);
8592f29d
N		 7522 }
7523 return 1;
7524 }
bf5a934a
DW		 7525 if (st->sb) {
7526 /* creating in a given container */
7527 return validate_geometry_imsm_volume(st, level, layout,
7528 raiddisks, chunk, size,
af4348dd 7529 data_offset,
bf5a934a
DW		 7530 dev, freesize, verbose);
7531 }
7532
bf5a934a
DW		 7533 /* This device needs to be a device in an 'imsm' container */
7534 fd = open(dev, O_RDONLY|O_EXCL, 0);
7535 if (fd >= 0) {
7536 if (verbose)
e7b84f9d
N		 7537 pr_err("Cannot create this array on device %s\n",
7538 dev);
bf5a934a
DW		 7539 close(fd);
7540 return 0;
7541 }
7542 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7543 if (verbose)
e7b84f9d 7544 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7545 dev, strerror(errno));
7546 return 0;
7547 }
7548 /* Well, it is in use by someone, maybe an 'imsm' container. */
7549 cfd = open_container(fd);
20cbe8d2 7550 close(fd);
bf5a934a 7551 if (cfd < 0) {
bf5a934a 7552 if (verbose)
e7b84f9d 7553 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7554 dev);
7555 return 0;
7556 }
4dd2df09 7557 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7558 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7559 strcmp(sra->text_version, "imsm") == 0)
7560 is_member = 1;
7561 sysfs_free(sra);
7562 if (is_member) {
bf5a934a
DW		 7563 /* This is a member of a imsm container. Load the container
7564 * and try to create a volume
7565 */
7566 struct intel_super *super;
7567
ec50f7b6 7568 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7569 st->sb = super;
4dd2df09 7570 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7571 close(cfd);
7572 return validate_geometry_imsm_volume(st, level, layout,
7573 raiddisks, chunk,
af4348dd 7574 size, data_offset, dev,
ecbd9e81
N		 7575 freesize, 1)
7576 ? 1 : -1;
bf5a934a 7577 }
20cbe8d2 7578 }
bf5a934a 7579
20cbe8d2 7580 if (verbose)
e7b84f9d 7581 pr_err("failed container membership check\n");
20cbe8d2
AW		 7582
7583 close(cfd);
7584 return 0;
bf5a934a 7585}
0bd16cf2 7586
30f58b22 7587static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7588{
7589 struct intel_super *super = st->sb;
7590
30f58b22
DW		 7591 if (level && *level == UnSet)
7592 *level = LEVEL_CONTAINER;
7593
7594 if (level && layout && *layout == UnSet)
7595 *layout = imsm_level_to_layout(*level);
0bd16cf2 7596
cd9d1ac7
DW		 7597 if (chunk && (*chunk == UnSet || *chunk == 0))
7598 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7599}
7600
33414a01
DW		 7601static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7602
7603static int kill_subarray_imsm(struct supertype *st)
7604{
7605 /* remove the subarray currently referenced by ->current_vol */
7606 __u8 i;
7607 struct intel_dev **dp;
7608 struct intel_super *super = st->sb;
7609 __u8 current_vol = super->current_vol;
7610 struct imsm_super *mpb = super->anchor;
7611
7612 if (super->current_vol < 0)
7613 return 2;
7614 super->current_vol = -1; /* invalidate subarray cursor */
7615
7616 /* block deletions that would change the uuid of active subarrays
7617 *
7618 * FIXME when immutable ids are available, but note that we'll
7619 * also need to fixup the invalidated/active subarray indexes in
7620 * mdstat
7621 */
7622 for (i = 0; i < mpb->num_raid_devs; i++) {
7623 char subarray[4];
7624
7625 if (i < current_vol)
7626 continue;
7627 sprintf(subarray, "%u", i);
4dd2df09 7628 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7629 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7630 current_vol, i);
33414a01
DW		 7631
7632 return 2;
7633 }
7634 }
7635
7636 if (st->update_tail) {
503975b9 7637 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7638
33414a01
DW		 7639 u->type = update_kill_array;
7640 u->dev_idx = current_vol;
7641 append_metadata_update(st, u, sizeof(*u));
7642
7643 return 0;
7644 }
7645
7646 for (dp = &super->devlist; *dp;)
7647 if ((*dp)->index == current_vol) {
7648 *dp = (*dp)->next;
7649 } else {
7650 handle_missing(super, (*dp)->dev);
7651 if ((*dp)->index > current_vol)
7652 (*dp)->index--;
7653 dp = &(*dp)->next;
7654 }
7655
7656 /* no more raid devices, all active components are now spares,
7657 * but of course failed are still failed
7658 */
7659 if (--mpb->num_raid_devs == 0) {
7660 struct dl *d;
7661
7662 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7663 if (d->index > -2)
7664 mark_spare(d);
33414a01
DW		 7665 }
7666
7667 super->updates_pending++;
7668
7669 return 0;
7670}
aa534678 7671
a951a4f7 7672static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7673 char *update, struct mddev_ident *ident)
aa534678
DW		 7674{
7675 /* update the subarray currently referenced by ->current_vol */
7676 struct intel_super *super = st->sb;
7677 struct imsm_super *mpb = super->anchor;
7678
aa534678
DW		 7679 if (strcmp(update, "name") == 0) {
7680 char *name = ident->name;
a951a4f7
N		 7681 char *ep;
7682 int vol;
aa534678 7683
4dd2df09 7684 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7685 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7686 return 2;
7687 }
7688
7689 if (!check_name(super, name, 0))
7690 return 2;
7691
a951a4f7
N		 7692 vol = strtoul(subarray, &ep, 10);
7693 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7694 return 2;
7695
aa534678 7696 if (st->update_tail) {
503975b9 7697 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7698
aa534678 7699 u->type = update_rename_array;
a951a4f7 7700 u->dev_idx = vol;
618f4e6d
XN		 7701 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7702 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7703 append_metadata_update(st, u, sizeof(*u));
7704 } else {
7705 struct imsm_dev *dev;
ebad3af2 7706 int i, namelen;
aa534678 7707
a951a4f7 7708 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7709 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7710 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7711 memcpy(dev->volume, name, namelen);
aa534678
DW		 7712 for (i = 0; i < mpb->num_raid_devs; i++) {
7713 dev = get_imsm_dev(super, i);
7714 handle_missing(super, dev);
7715 }
7716 super->updates_pending++;
7717 }
e6e9dd3f
AP		 7718 } else if (strcmp(update, "ppl") == 0 ||
7719 strcmp(update, "no-ppl") == 0) {
7720 int new_policy;
7721 char *ep;
7722 int vol = strtoul(subarray, &ep, 10);
7723
7724 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7725 return 2;
7726
7727 if (strcmp(update, "ppl") == 0)
c2462068 7728 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7729 else
c2462068 7730 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7731
7732 if (st->update_tail) {
7733 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7734
7735 u->type = update_rwh_policy;
7736 u->dev_idx = vol;
7737 u->new_policy = new_policy;
7738 append_metadata_update(st, u, sizeof(*u));
7739 } else {
7740 struct imsm_dev *dev;
7741
7742 dev = get_imsm_dev(super, vol);
7743 dev->rwh_policy = new_policy;
7744 super->updates_pending++;
7745 }
aa534678
DW		 7746 } else
7747 return 2;
7748
7749 return 0;
7750}
bf5a934a 7751
28bce06f
AK		 7752static int is_gen_migration(struct imsm_dev *dev)
7753{
7534230b
AK		 7754 if (dev == NULL)
7755 return 0;
7756
28bce06f
AK		 7757 if (!dev->vol.migr_state)
7758 return 0;
7759
7760 if (migr_type(dev) == MIGR_GEN_MIGR)
7761 return 1;
7762
7763 return 0;
7764}
7765
1e5c6983
DW		 7766static int is_rebuilding(struct imsm_dev *dev)
7767{
7768 struct imsm_map *migr_map;
7769
7770 if (!dev->vol.migr_state)
7771 return 0;
7772
7773 if (migr_type(dev) != MIGR_REBUILD)
7774 return 0;
7775
238c0a71 7776 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7777
7778 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7779 return 1;
7780 else
7781 return 0;
7782}
7783
6ce1fbf1
AK		 7784static int is_initializing(struct imsm_dev *dev)
7785{
7786 struct imsm_map *migr_map;
7787
7788 if (!dev->vol.migr_state)
7789 return 0;
7790
7791 if (migr_type(dev) != MIGR_INIT)
7792 return 0;
7793
238c0a71 7794 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7795
7796 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7797 return 1;
7798
7799 return 0;
6ce1fbf1
AK		 7800}
7801
c47b0ff6
AK		 7802static void update_recovery_start(struct intel_super *super,
7803 struct imsm_dev *dev,
7804 struct mdinfo *array)
1e5c6983
DW		 7805{
7806 struct mdinfo *rebuild = NULL;
7807 struct mdinfo *d;
7808 __u32 units;
7809
7810 if (!is_rebuilding(dev))
7811 return;
7812
7813 /* Find the rebuild target, but punt on the dual rebuild case */
7814 for (d = array->devs; d; d = d->next)
7815 if (d->recovery_start == 0) {
7816 if (rebuild)
7817 return;
7818 rebuild = d;
7819 }
7820
4363fd80
DW		 7821 if (!rebuild) {
7822 /* (?) none of the disks are marked with
7823 * IMSM_ORD_REBUILD, so assume they are missing and the
7824 * disk_ord_tbl was not correctly updated
7825 */
1ade5cc1 7826 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7827 return;
7828 }
7829
1e5c6983 7830 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7831 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7832}
7833
276d77db 7834static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7835
00bbdbda 7836static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7837{
4f5bc454
DW		 7838 /* Given a container loaded by load_super_imsm_all,
7839 * extract information about all the arrays into
7840 * an mdinfo tree.
00bbdbda 7841 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7842 *
7843 * For each imsm_dev create an mdinfo, fill it in,
7844 * then look for matching devices in super->disks
7845 * and create appropriate device mdinfo.
7846 */
7847 struct intel_super *super = st->sb;
949c47a0 7848 struct imsm_super *mpb = super->anchor;
4f5bc454 7849 struct mdinfo *rest = NULL;
00bbdbda 7850 unsigned int i;
81219e70 7851 int sb_errors = 0;
abef11a3
AK		 7852 struct dl *d;
7853 int spare_disks = 0;
b6180160 7854 int current_vol = super->current_vol;
cdddbdbc 7855
19482bcc
AK		 7856 /* do not assemble arrays when not all attributes are supported */
7857 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7858 sb_errors = 1;
7a862a02 7859 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7860 }
7861
abef11a3
AK		 7862 /* count spare devices, not used in maps
7863 */
7864 for (d = super->disks; d; d = d->next)
7865 if (d->index == -1)
7866 spare_disks++;
7867
4f5bc454 7868 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7869 struct imsm_dev *dev;
7870 struct imsm_map *map;
86e3692b 7871 struct imsm_map *map2;
4f5bc454 7872 struct mdinfo *this;
a6482415 7873 int slot;
a6482415 7874 int chunk;
00bbdbda 7875 char *ep;
8b9cd157 7876 int level;
00bbdbda
N		 7877
7878 if (subarray &&
7879 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7880 continue;
7881
7882 dev = get_imsm_dev(super, i);
238c0a71
AK		 7883 map = get_imsm_map(dev, MAP_0);
7884 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7885 level = get_imsm_raid_level(map);
4f5bc454 7886
1ce0101c
DW		 7887 /* do not publish arrays that are in the middle of an
7888 * unsupported migration
7889 */
7890 if (dev->vol.migr_state &&
28bce06f 7891 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7892 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7893 dev->volume);
7894 continue;
7895 }
2db86302
LM		 7896 /* do not publish arrays that are not support by controller's
7897 * OROM/EFI
7898 */
1ce0101c 7899
503975b9 7900 this = xmalloc(sizeof(*this));
4f5bc454 7901
301406c9 7902 super->current_vol = i;
a5d85af7 7903 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7904 this->next = rest;
a6482415 7905 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7906 /* mdadm does not support all metadata features- set the bit in all arrays state */
7907 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7908 level, /* RAID level */
7909 imsm_level_to_layout(level),
81219e70 7910 map->num_members, /* raid disks */
fcc2c9da 7911 &chunk, imsm_dev_size(dev),
81219e70 7912 1 /* verbose */)) {
7a862a02 7913 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7914 dev->volume);
7915 this->array.state |=
7916 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7917 (1<<MD_SB_BLOCK_VOLUME);
7918 }
81219e70
LM		 7919
7920 /* if array has bad blocks, set suitable bit in all arrays state */
7921 if (sb_errors)
7922 this->array.state |=
7923 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7924 (1<<MD_SB_BLOCK_VOLUME);
7925
4f5bc454 7926 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7927 unsigned long long recovery_start;
4f5bc454
DW		 7928 struct mdinfo *info_d;
7929 struct dl *d;
7930 int idx;
9a1608e5 7931 int skip;
7eef0453 7932 __u32 ord;
8b9cd157 7933 int missing = 0;
4f5bc454 7934
9a1608e5 7935 skip = 0;
238c0a71
AK		 7936 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7937 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7938 for (d = super->disks; d ; d = d->next)
7939 if (d->index == idx)
0fbd635c 7940 break;
4f5bc454 7941
1e5c6983 7942 recovery_start = MaxSector;
4f5bc454 7943 if (d == NULL)
9a1608e5 7944 skip = 1;
25ed7e59 7945 if (d && is_failed(&d->disk))
9a1608e5 7946 skip = 1;
8b9cd157 7947 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7948 recovery_start = 0;
1e93d0d1
BK		 7949 if (!(ord & IMSM_ORD_REBUILD))
7950 this->array.working_disks++;
1011e834 7951 /*
9a1608e5 7952 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7953 * reset resync start to avoid a dirty-degraded
7954 * situation when performing the intial sync
9a1608e5 7955 */
8b9cd157
MK		 7956 if (skip)
7957 missing++;
7958
7959 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7960 if ((!able_to_resync(level, missing) ||
7961 recovery_start == 0))
7962 this->resync_start = MaxSector;
7963 } else {
7964 /*
7965 * FIXME handle dirty degraded
7966 */
7967 }
7968
9a1608e5
DW		 7969 if (skip)
7970 continue;
4f5bc454 7971
503975b9 7972 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7973 info_d->next = this->devs;
7974 this->devs = info_d;
7975
4f5bc454
DW		 7976 info_d->disk.number = d->index;
7977 info_d->disk.major = d->major;
7978 info_d->disk.minor = d->minor;
7979 info_d->disk.raid_disk = slot;
1e5c6983 7980 info_d->recovery_start = recovery_start;
86e3692b
AK		 7981 if (map2) {
7982 if (slot < map2->num_members)
7983 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7984 else
7985 this->array.spare_disks++;
86e3692b
AK		 7986 } else {
7987 if (slot < map->num_members)
7988 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7989 else
7990 this->array.spare_disks++;
86e3692b 7991 }
4f5bc454
DW		 7992
7993 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7994 info_d->data_offset = pba_of_lba0(map);
44490938 7995 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7996
7997 if (map->raid_level == 5) {
2432ce9b
AP		 7998 info_d->ppl_sector = this->ppl_sector;
7999 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 8000 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
8001 recovery_start == 0)
8002 this->resync_start = 0;
06fb291a 8003 }
b12796be 8004
5e46202e 8005 info_d->bb.supported = 1;
b12796be
TM		 8006 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
8007 info_d->data_offset,
8008 info_d->component_size,
8009 &info_d->bb);
4f5bc454 8010 }
1e5c6983 8011 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 8012 update_recovery_start(super, dev, this);
abef11a3 8013 this->array.spare_disks += spare_disks;
276d77db
AK		 8014
8015 /* check for reshape */
8016 if (this->reshape_active == 1)
8017 recover_backup_imsm(st, this);
9a1608e5 8018 rest = this;
4f5bc454
DW		 8019 }
8020
b6180160 8021 super->current_vol = current_vol;
4f5bc454 8022 return rest;
cdddbdbc
DW		 8023}
8024
3b451610
AK		 8025static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8026 int failed, int look_in_map)
c2a1e7da 8027{
3b451610
AK		 8028 struct imsm_map *map;
8029
8030 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8031
8032 if (!failed)
1011e834 8033 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8034 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8035
8036 switch (get_imsm_raid_level(map)) {
8037 case 0:
8038 return IMSM_T_STATE_FAILED;
8039 break;
8040 case 1:
8041 if (failed < map->num_members)
8042 return IMSM_T_STATE_DEGRADED;
8043 else
8044 return IMSM_T_STATE_FAILED;
8045 break;
8046 case 10:
8047 {
8048 /**
c92a2527
DW		 8049 * check to see if any mirrors have failed, otherwise we
8050 * are degraded. Even numbered slots are mirrored on
8051 * slot+1
c2a1e7da 8052 */
c2a1e7da 8053 int i;
d9b420a5
N		 8054 /* gcc -Os complains that this is unused */
8055 int insync = insync;
c2a1e7da
DW		 8056
8057 for (i = 0; i < map->num_members; i++) {
238c0a71 8058 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8059 int idx = ord_to_idx(ord);
8060 struct imsm_disk *disk;
c2a1e7da 8061
c92a2527 8062 /* reset the potential in-sync count on even-numbered
1011e834 8063 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8064 */
8065 if ((i & 1) == 0)
8066 insync = 2;
c2a1e7da 8067
c92a2527 8068 disk = get_imsm_disk(super, idx);
25ed7e59 8069 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8070 insync--;
c2a1e7da 8071
c92a2527
DW		 8072 /* no in-sync disks left in this mirror the
8073 * array has failed
8074 */
8075 if (insync == 0)
8076 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8077 }
8078
8079 return IMSM_T_STATE_DEGRADED;
8080 }
8081 case 5:
8082 if (failed < 2)
8083 return IMSM_T_STATE_DEGRADED;
8084 else
8085 return IMSM_T_STATE_FAILED;
8086 break;
8087 default:
8088 break;
8089 }
8090
8091 return map->map_state;
8092}
8093
3b451610
AK		 8094static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8095 int look_in_map)
c2a1e7da
DW		 8096{
8097 int i;
8098 int failed = 0;
8099 struct imsm_disk *disk;
d5985138
AK		 8100 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8101 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8102 struct imsm_map *map_for_loop;
0556e1a2
DW		 8103 __u32 ord;
8104 int idx;
d5985138 8105 int idx_1;
c2a1e7da 8106
0556e1a2
DW		 8107 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8108 * disks that are being rebuilt. New failures are recorded to
8109 * map[0]. So we look through all the disks we started with and
8110 * see if any failures are still present, or if any new ones
8111 * have arrived
0556e1a2 8112 */
d5985138
AK		 8113 map_for_loop = map;
8114 if (prev && (map->num_members < prev->num_members))
8115 map_for_loop = prev;
68fe4598
LD		 8116
8117 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8118 idx_1 = -255;
238c0a71
AK		 8119 /* when MAP_X is passed both maps failures are counted
8120 */
d5985138 8121 if (prev &&
089f9d79
JS		 8122 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8123 i < prev->num_members) {
d5985138
AK		 8124 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8125 idx_1 = ord_to_idx(ord);
c2a1e7da 8126
d5985138
AK		 8127 disk = get_imsm_disk(super, idx_1);
8128 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8129 failed++;
8130 }
089f9d79
JS		 8131 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8132 i < map->num_members) {
d5985138
AK		 8133 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8134 idx = ord_to_idx(ord);
8135
8136 if (idx != idx_1) {
8137 disk = get_imsm_disk(super, idx);
8138 if (!disk || is_failed(disk) ||
8139 ord & IMSM_ORD_REBUILD)
8140 failed++;
8141 }
8142 }
c2a1e7da
DW		 8143 }
8144
8145 return failed;
845dea95
NB		 8146}
8147
97b4d0e9
DW		 8148static int imsm_open_new(struct supertype *c, struct active_array *a,
8149 char *inst)
8150{
8151 struct intel_super *super = c->sb;
8152 struct imsm_super *mpb = super->anchor;
bbab0940 8153 struct imsm_update_prealloc_bb_mem u;
9587c373 8154
97b4d0e9 8155 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8156 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8157 return -ENODEV;
8158 }
8159
8160 dprintf("imsm: open_new %s\n", inst);
8161 a->info.container_member = atoi(inst);
bbab0940
TM		 8162
8163 u.type = update_prealloc_badblocks_mem;
8164 imsm_update_metadata_locally(c, &u, sizeof(u));
8165
97b4d0e9
DW		 8166 return 0;
8167}
8168
0c046afd
DW		 8169static int is_resyncing(struct imsm_dev *dev)
8170{
8171 struct imsm_map *migr_map;
8172
8173 if (!dev->vol.migr_state)
8174 return 0;
8175
1484e727
DW		 8176 if (migr_type(dev) == MIGR_INIT ||
8177 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8178 return 1;
8179
4c9bc37b
AK		 8180 if (migr_type(dev) == MIGR_GEN_MIGR)
8181 return 0;
8182
238c0a71 8183 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8184
089f9d79
JS		 8185 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8186 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8187 return 1;
8188 else
8189 return 0;
8190}
8191
0556e1a2 8192/* return true if we recorded new information */
4c9e8c1e
TM		 8193static int mark_failure(struct intel_super *super,
8194 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8195{
0556e1a2
DW		 8196 __u32 ord;
8197 int slot;
8198 struct imsm_map *map;
86c54047
DW		 8199 char buf[MAX_RAID_SERIAL_LEN+3];
8200 unsigned int len, shift = 0;
0556e1a2
DW		 8201
8202 /* new failures are always set in map[0] */
238c0a71 8203 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8204
8205 slot = get_imsm_disk_slot(map, idx);
8206 if (slot < 0)
8207 return 0;
8208
8209 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8210 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8211 return 0;
8212
7d0c5e24
LD		 8213 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8214 buf[MAX_RAID_SERIAL_LEN] = '\000';
8215 strcat(buf, ":0");
86c54047
DW		 8216 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8217 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8218 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8219
f2f27e63 8220 disk->status |= FAILED_DISK;
0556e1a2 8221 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8222 /* mark failures in second map if second map exists and this disk
8223 * in this slot.
8224 * This is valid for migration, initialization and rebuild
8225 */
8226 if (dev->vol.migr_state) {
238c0a71 8227 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8228 int slot2 = get_imsm_disk_slot(map2, idx);
8229
089f9d79 8230 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8231 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8232 idx | IMSM_ORD_REBUILD);
8233 }
d7a1fda2
MT		 8234 if (map->failed_disk_num == 0xff ||
8235 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8236 map->failed_disk_num = slot;
4c9e8c1e
TM		 8237
8238 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8239
0556e1a2
DW		 8240 return 1;
8241}
8242
4c9e8c1e
TM		 8243static void mark_missing(struct intel_super *super,
8244 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8245{
4c9e8c1e 8246 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8247
8248 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8249 return;
8250
47ee5a45
DW		 8251 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8252 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8253}
8254
33414a01
DW		 8255static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8256{
33414a01 8257 struct dl *dl;
33414a01
DW		 8258
8259 if (!super->missing)
8260 return;
33414a01 8261
79b68f1b
PC		 8262 /* When orom adds replacement for missing disk it does
8263 * not remove entry of missing disk, but just updates map with
8264 * new added disk. So it is not enough just to test if there is
8265 * any missing disk, we have to look if there are any failed disks
8266 * in map to stop migration */
8267
33414a01 8268 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8269 /* end process for initialization and rebuild only
8270 */
8271 if (is_gen_migration(dev) == 0) {
fb12a745 8272 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8273
fb12a745
TM		 8274 if (failed) {
8275 __u8 map_state;
8276 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8277 struct imsm_map *map1;
8278 int i, ord, ord_map1;
8279 int rebuilt = 1;
3d59f0c0 8280
fb12a745
TM		 8281 for (i = 0; i < map->num_members; i++) {
8282 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8283 if (!(ord & IMSM_ORD_REBUILD))
8284 continue;
8285
8286 map1 = get_imsm_map(dev, MAP_1);
8287 if (!map1)
8288 continue;
8289
8290 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8291 if (ord_map1 & IMSM_ORD_REBUILD)
8292 rebuilt = 0;
8293 }
8294
8295 if (rebuilt) {
8296 map_state = imsm_check_degraded(super, dev,
8297 failed, MAP_0);
8298 end_migration(dev, super, map_state);
8299 }
8300 }
3d59f0c0 8301 }
33414a01 8302 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8303 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8304 super->updates_pending++;
8305}
8306
f3871fdc
AK		 8307static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8308 long long new_size)
70bdf0dc 8309{
70bdf0dc 8310 unsigned long long array_blocks;
9529d343
MD		 8311 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8312 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8313
8314 if (used_disks == 0) {
8315 /* when problems occures
8316 * return current array_blocks value
8317 */
fcc2c9da 8318 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8319
8320 return array_blocks;
8321 }
8322
8323 /* set array size in metadata
8324 */
9529d343 8325 if (new_size <= 0)
f3871fdc
AK		 8326 /* OLCE size change is caused by added disks
8327 */
44490938 8328 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8329 else
f3871fdc
AK		 8330 /* Online Volume Size Change
8331 * Using available free space
8332 */
8333 array_blocks = new_size;
70bdf0dc 8334
b53bfba6 8335 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8336 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8337
8338 return array_blocks;
8339}
8340
28bce06f
AK		 8341static void imsm_set_disk(struct active_array *a, int n, int state);
8342
0e2d1a4e
AK		 8343static void imsm_progress_container_reshape(struct intel_super *super)
8344{
8345 /* if no device has a migr_state, but some device has a
8346 * different number of members than the previous device, start
8347 * changing the number of devices in this device to match
8348 * previous.
8349 */
8350 struct imsm_super *mpb = super->anchor;
8351 int prev_disks = -1;
8352 int i;
1dfaa380 8353 int copy_map_size;
0e2d1a4e
AK		 8354
8355 for (i = 0; i < mpb->num_raid_devs; i++) {
8356 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8357 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8358 struct imsm_map *map2;
8359 int prev_num_members;
0e2d1a4e
AK		 8360
8361 if (dev->vol.migr_state)
8362 return;
8363
8364 if (prev_disks == -1)
8365 prev_disks = map->num_members;
8366 if (prev_disks == map->num_members)
8367 continue;
8368
8369 /* OK, this array needs to enter reshape mode.
8370 * i.e it needs a migr_state
8371 */
8372
1dfaa380 8373 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8374 prev_num_members = map->num_members;
8375 map->num_members = prev_disks;
8376 dev->vol.migr_state = 1;
8377 dev->vol.curr_migr_unit = 0;
ea672ee1 8378 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8379 for (i = prev_num_members;
8380 i < map->num_members; i++)
8381 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8382 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8383 /* Copy the current map */
1dfaa380 8384 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8385 map2->num_members = prev_num_members;
8386
f3871fdc 8387 imsm_set_array_size(dev, -1);
51d83f5d 8388 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8389 super->updates_pending++;
8390 }
8391}
8392
aad6f216 8393/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8394 * states are handled in imsm_set_disk() with one exception, when a
8395 * resync is stopped due to a new failure this routine will set the
8396 * 'degraded' state for the array.
8397 */
01f157d7 8398static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8399{
8400 int inst = a->info.container_member;
8401 struct intel_super *super = a->container->sb;
949c47a0 8402 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8403 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8404 int failed = imsm_count_failed(super, dev, MAP_0);
8405 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8406 __u32 blocks_per_unit;
a862209d 8407
1af97990
AK		 8408 if (dev->vol.migr_state &&
8409 dev->vol.migr_type == MIGR_GEN_MIGR) {
8410 /* array state change is blocked due to reshape action
aad6f216
N		 8411 * We might need to
8412 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8413 * - finish the reshape (if last_checkpoint is big and action != reshape)
8414 * - update curr_migr_unit
1af97990 8415 */
aad6f216
N		 8416 if (a->curr_action == reshape) {
8417 /* still reshaping, maybe update curr_migr_unit */
633b5610 8418 goto mark_checkpoint;
aad6f216
N		 8419 } else {
8420 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8421 /* for some reason we aborted the reshape.
b66e591b
AK		 8422 *
8423 * disable automatic metadata rollback
8424 * user action is required to recover process
aad6f216 8425 */
b66e591b 8426 if (0) {
238c0a71
AK		 8427 struct imsm_map *map2 =
8428 get_imsm_map(dev, MAP_1);
8429 dev->vol.migr_state = 0;
8430 set_migr_type(dev, 0);
8431 dev->vol.curr_migr_unit = 0;
8432 memcpy(map, map2,
8433 sizeof_imsm_map(map2));
8434 super->updates_pending++;
b66e591b 8435 }
aad6f216
N		 8436 }
8437 if (a->last_checkpoint >= a->info.component_size) {
8438 unsigned long long array_blocks;
8439 int used_disks;
e154ced3 8440 struct mdinfo *mdi;
aad6f216 8441
9529d343 8442 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8443 if (used_disks > 0) {
8444 array_blocks =
44490938 8445 per_dev_array_size(map) *
d55adef9 8446 used_disks;
b53bfba6
TM		 8447 array_blocks =
8448 round_size_to_mb(array_blocks,
8449 used_disks);
d55adef9
AK		 8450 a->info.custom_array_size = array_blocks;
8451 /* encourage manager to update array
8452 * size
8453 */
e154ced3 8454
d55adef9 8455 a->check_reshape = 1;
633b5610 8456 }
e154ced3
AK		 8457 /* finalize online capacity expansion/reshape */
8458 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8459 imsm_set_disk(a,
8460 mdi->disk.raid_disk,
8461 mdi->curr_state);
8462
0e2d1a4e 8463 imsm_progress_container_reshape(super);
e154ced3 8464 }
aad6f216 8465 }
1af97990
AK		 8466 }
8467
47ee5a45 8468 /* before we activate this array handle any missing disks */
33414a01
DW		 8469 if (consistent == 2)
8470 handle_missing(super, dev);
1e5c6983 8471
0c046afd 8472 if (consistent == 2 &&
b7941fd6 8473 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8474 map_state != IMSM_T_STATE_NORMAL ||
8475 dev->vol.migr_state))
01f157d7 8476 consistent = 0;
272906ef 8477
b7941fd6 8478 if (is_resync_complete(&a->info)) {
0c046afd 8479 /* complete intialization / resync,
0556e1a2
DW		 8480 * recovery and interrupted recovery is completed in
8481 * ->set_disk
0c046afd
DW		 8482 */
8483 if (is_resyncing(dev)) {
8484 dprintf("imsm: mark resync done\n");
809da78e 8485 end_migration(dev, super, map_state);
115c3803 8486 super->updates_pending++;
484240d8 8487 a->last_checkpoint = 0;
115c3803 8488 }
b9172665
AK		 8489 } else if ((!is_resyncing(dev) && !failed) &&
8490 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8491 /* mark the start of the init process if nothing is failed */
b7941fd6 8492 dprintf("imsm: mark resync start\n");
1484e727 8493 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8494 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8495 else
8e59f3d8 8496 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8497 super->updates_pending++;
115c3803 8498 }
a862209d 8499
633b5610 8500mark_checkpoint:
5b83bacf
AK		 8501 /* skip checkpointing for general migration,
8502 * it is controlled in mdadm
8503 */
8504 if (is_gen_migration(dev))
8505 goto skip_mark_checkpoint;
8506
1e5c6983 8507 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8508 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8509 if (blocks_per_unit) {
1e5c6983
DW		 8510 __u32 units32;
8511 __u64 units;
8512
4f0a7acc 8513 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8514 units32 = units;
8515
8516 /* check that we did not overflow 32-bits, and that
8517 * curr_migr_unit needs updating
8518 */
8519 if (units32 == units &&
bfd80a56 8520 units32 != 0 &&
1e5c6983
DW		 8521 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8522 dprintf("imsm: mark checkpoint (%u)\n", units32);
8523 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8524 super->updates_pending++;
8525 }
8526 }
f8f603f1 8527
5b83bacf 8528skip_mark_checkpoint:
3393c6af 8529 /* mark dirty / clean */
2432ce9b
AP		 8530 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8531 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8532 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8533 if (consistent) {
8534 dev->vol.dirty = RAIDVOL_CLEAN;
8535 } else {
8536 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8537 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8538 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8539 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8540 }
a862209d
DW		 8541 super->updates_pending++;
8542 }
28bce06f 8543
01f157d7 8544 return consistent;
a862209d
DW		 8545}
8546
6f50473f
TM		 8547static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8548{
8549 int inst = a->info.container_member;
8550 struct intel_super *super = a->container->sb;
8551 struct imsm_dev *dev = get_imsm_dev(super, inst);
8552 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8553
8554 if (slot > map->num_members) {
8555 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8556 slot, map->num_members - 1);
8557 return -1;
8558 }
8559
8560 if (slot < 0)
8561 return -1;
8562
8563 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8564}
8565
8d45d196 8566static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8567{
8d45d196
DW		 8568 int inst = a->info.container_member;
8569 struct intel_super *super = a->container->sb;
949c47a0 8570 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8571 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8572 struct imsm_disk *disk;
7ce05701
LD		 8573 struct mdinfo *mdi;
8574 int recovery_not_finished = 0;
0c046afd 8575 int failed;
6f50473f 8576 int ord;
0c046afd 8577 __u8 map_state;
fb12a745
TM		 8578 int rebuild_done = 0;
8579 int i;
8d45d196 8580
fb12a745 8581 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8582 if (ord < 0)
8d45d196
DW		 8583 return;
8584
4e6e574a 8585 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8586 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8587
5802a811 8588 /* check for new failures */
ae7d61e3 8589 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8590 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8591 super->updates_pending++;
8d45d196 8592 }
47ee5a45 8593
19859edc 8594 /* check if in_sync */
0556e1a2 8595 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8596 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8597
8598 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8599 rebuild_done = 1;
19859edc
DW		 8600 super->updates_pending++;
8601 }
8d45d196 8602
3b451610
AK		 8603 failed = imsm_count_failed(super, dev, MAP_0);
8604 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8605
0c046afd 8606 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8607 dprintf("imsm: Detected transition to state ");
8608 switch (map_state) {
8609 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8610 dprintf("normal: ");
8611 if (is_rebuilding(dev)) {
1ade5cc1 8612 dprintf_cont("while rebuilding");
7ce05701
LD		 8613 /* check if recovery is really finished */
8614 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8615 if (mdi->recovery_start != MaxSector) {
8616 recovery_not_finished = 1;
8617 break;
8618 }
8619 if (recovery_not_finished) {
1ade5cc1
N		 8620 dprintf_cont("\n");
8621 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8622 if (a->last_checkpoint < mdi->recovery_start) {
8623 a->last_checkpoint = mdi->recovery_start;
8624 super->updates_pending++;
8625 }
8626 break;
8627 }
94002678 8628 end_migration(dev, super, map_state);
238c0a71 8629 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8630 map->failed_disk_num = ~0;
8631 super->updates_pending++;
8632 a->last_checkpoint = 0;
8633 break;
8634 }
8635 if (is_gen_migration(dev)) {
1ade5cc1 8636 dprintf_cont("while general migration");
bf2f0071 8637 if (a->last_checkpoint >= a->info.component_size)
809da78e 8638 end_migration(dev, super, map_state);
94002678
AK		 8639 else
8640 map->map_state = map_state;
238c0a71 8641 map = get_imsm_map(dev, MAP_0);
28bce06f 8642 map->failed_disk_num = ~0;
94002678 8643 super->updates_pending++;
bf2f0071 8644 break;
94002678
AK		 8645 }
8646 break;
8647 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8648 dprintf_cont("degraded: ");
089f9d79 8649 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8650 dprintf_cont("mark degraded");
94002678
AK		 8651 map->map_state = map_state;
8652 super->updates_pending++;
8653 a->last_checkpoint = 0;
8654 break;
8655 }
8656 if (is_rebuilding(dev)) {
d7a1fda2 8657 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8658 if (state & DS_FAULTY) {
8659 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8660 if (n == map->failed_disk_num) {
8661 dprintf_cont("end migration");
8662 end_migration(dev, super, map_state);
a4e96fd8 8663 a->last_checkpoint = 0;
d7a1fda2 8664 } else {
a4e96fd8 8665 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8666 map->map_state = map_state;
8667 }
94002678 8668 super->updates_pending++;
fb12a745
TM		 8669 }
8670
a4e96fd8
MT		 8671 if (!rebuild_done)
8672 break;
8673
fb12a745
TM		 8674 /* check if recovery is really finished */
8675 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8676 if (mdi->recovery_start != MaxSector) {
8677 recovery_not_finished = 1;
8678 break;
8679 }
8680 if (recovery_not_finished) {
8681 dprintf_cont("\n");
a4e96fd8 8682 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8683 if (a->last_checkpoint < mdi->recovery_start) {
8684 a->last_checkpoint =
8685 mdi->recovery_start;
8686 super->updates_pending++;
8687 }
8688 break;
94002678 8689 }
fb12a745
TM		 8690
8691 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8692 end_migration(dev, super, map_state);
8693 a->last_checkpoint = 0;
8694 super->updates_pending++;
fb12a745
TM		 8695
8696 for (i = 0; i < map->num_members; i++) {
8697 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8698
8699 if (idx & IMSM_ORD_REBUILD)
8700 map->failed_disk_num = i;
8701 }
8702 super->updates_pending++;
94002678
AK		 8703 break;
8704 }
8705 if (is_gen_migration(dev)) {
1ade5cc1 8706 dprintf_cont("while general migration");
bf2f0071 8707 if (a->last_checkpoint >= a->info.component_size)
809da78e 8708 end_migration(dev, super, map_state);
94002678
AK		 8709 else {
8710 map->map_state = map_state;
3b451610 8711 manage_second_map(super, dev);
94002678
AK		 8712 }
8713 super->updates_pending++;
bf2f0071 8714 break;
28bce06f 8715 }
6ce1fbf1 8716 if (is_initializing(dev)) {
1ade5cc1 8717 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8718 map->map_state = map_state;
8719 super->updates_pending++;
8720 break;
8721 }
94002678
AK		 8722 break;
8723 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8724 dprintf_cont("failed: ");
94002678 8725 if (is_gen_migration(dev)) {
1ade5cc1 8726 dprintf_cont("while general migration");
94002678
AK		 8727 map->map_state = map_state;
8728 super->updates_pending++;
8729 break;
8730 }
8731 if (map->map_state != map_state) {
1ade5cc1 8732 dprintf_cont("mark failed");
94002678
AK		 8733 end_migration(dev, super, map_state);
8734 super->updates_pending++;
8735 a->last_checkpoint = 0;
8736 break;
8737 }
8738 break;
8739 default:
1ade5cc1 8740 dprintf_cont("state %i\n", map_state);
5802a811 8741 }
1ade5cc1 8742 dprintf_cont("\n");
845dea95
NB		 8743}
8744
f796af5d 8745static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8746{
f796af5d 8747 void *buf = mpb;
c2a1e7da
DW		 8748 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8749 unsigned long long dsize;
8750 unsigned long long sectors;
f36a9ecd 8751 unsigned int sector_size;
c2a1e7da 8752
f36a9ecd 8753 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8754 get_dev_size(fd, NULL, &dsize);
8755
f36a9ecd 8756 if (mpb_size > sector_size) {
272f648f 8757 /* -1 to account for anchor */
f36a9ecd 8758 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8759
272f648f 8760 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8761 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8762 SEEK_SET) < 0)
272f648f 8763 return 1;
c2a1e7da 8764
f36a9ecd
PB		 8765 if ((unsigned long long)write(fd, buf + sector_size,
8766 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8767 return 1;
8768 }
c2a1e7da 8769
272f648f 8770 /* first block is stored on second to last sector of the disk */
f36a9ecd 8771 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8772 return 1;
8773
466070ad 8774 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8775 return 1;
8776
c2a1e7da
DW		 8777 return 0;
8778}
8779
2e735d19 8780static void imsm_sync_metadata(struct supertype *container)
845dea95 8781{
2e735d19 8782 struct intel_super *super = container->sb;
c2a1e7da 8783
1a64be56 8784 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8785 if (!super->updates_pending)
8786 return;
8787
36988a3d 8788 write_super_imsm(container, 0);
c2a1e7da
DW		 8789
8790 super->updates_pending = 0;
845dea95
NB		 8791}
8792
272906ef
DW		 8793static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8794{
8795 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8796 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8797 struct dl *dl;
8798
8799 for (dl = super->disks; dl; dl = dl->next)
8800 if (dl->index == i)
8801 break;
8802
25ed7e59 8803 if (dl && is_failed(&dl->disk))
272906ef
DW		 8804 dl = NULL;
8805
8806 if (dl)
1ade5cc1 8807 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8808
8809 return dl;
8810}
8811
a20d2ba5 8812static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8813 struct active_array *a, int activate_new,
8814 struct mdinfo *additional_test_list)
272906ef
DW		 8815{
8816 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8817 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8818 struct imsm_super *mpb = super->anchor;
8819 struct imsm_map *map;
272906ef
DW		 8820 unsigned long long pos;
8821 struct mdinfo *d;
8822 struct extent *ex;
a20d2ba5 8823 int i, j;
272906ef 8824 int found;
569cc43f
DW		 8825 __u32 array_start = 0;
8826 __u32 array_end = 0;
272906ef 8827 struct dl *dl;
6c932028 8828 struct mdinfo *test_list;
272906ef
DW		 8829
8830 for (dl = super->disks; dl; dl = dl->next) {
8831 /* If in this array, skip */
8832 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8833 if (d->state_fd >= 0 &&
8834 d->disk.major == dl->major &&
272906ef 8835 d->disk.minor == dl->minor) {
8ba77d32
AK		 8836 dprintf("%x:%x already in array\n",
8837 dl->major, dl->minor);
272906ef
DW		 8838 break;
8839 }
8840 if (d)
8841 continue;
6c932028
AK		 8842 test_list = additional_test_list;
8843 while (test_list) {
8844 if (test_list->disk.major == dl->major &&
8845 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8846 dprintf("%x:%x already in additional test list\n",
8847 dl->major, dl->minor);
8848 break;
8849 }
6c932028 8850 test_list = test_list->next;
8ba77d32 8851 }
6c932028 8852 if (test_list)
8ba77d32 8853 continue;
272906ef 8854
e553d2a4 8855 /* skip in use or failed drives */
25ed7e59 8856 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8857 dl->index == -2) {
8858 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8859 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8860 continue;
8861 }
8862
a20d2ba5
DW		 8863 /* skip pure spares when we are looking for partially
8864 * assimilated drives
8865 */
8866 if (dl->index == -1 && !activate_new)
8867 continue;
8868
f2cc4f7d
AO		 8869 if (!drive_validate_sector_size(super, dl))
8870 continue;
8871
272906ef 8872 /* Does this unused device have the requisite free space?
a20d2ba5 8873 * It needs to be able to cover all member volumes
272906ef 8874 */
05501181 8875 ex = get_extents(super, dl, 1);
272906ef
DW		 8876 if (!ex) {
8877 dprintf("cannot get extents\n");
8878 continue;
8879 }
a20d2ba5
DW		 8880 for (i = 0; i < mpb->num_raid_devs; i++) {
8881 dev = get_imsm_dev(super, i);
238c0a71 8882 map = get_imsm_map(dev, MAP_0);
272906ef 8883
a20d2ba5
DW		 8884 /* check if this disk is already a member of
8885 * this array
272906ef 8886 */
620b1713 8887 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8888 continue;
8889
8890 found = 0;
8891 j = 0;
8892 pos = 0;
5551b113 8893 array_start = pba_of_lba0(map);
329c8278 8894 array_end = array_start +
44490938 8895 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8896
8897 do {
8898 /* check that we can start at pba_of_lba0 with
44490938 8899 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8900 */
329c8278 8901 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8902 found = 1;
8903 break;
8904 }
8905 pos = ex[j].start + ex[j].size;
8906 j++;
8907 } while (ex[j-1].size);
8908
8909 if (!found)
272906ef 8910 break;
a20d2ba5 8911 }
272906ef
DW		 8912
8913 free(ex);
a20d2ba5 8914 if (i < mpb->num_raid_devs) {
329c8278
DW		 8915 dprintf("%x:%x does not have %u to %u available\n",
8916 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8917 /* No room */
8918 continue;
a20d2ba5
DW		 8919 }
8920 return dl;
272906ef
DW		 8921 }
8922
8923 return dl;
8924}
8925
95d07a2c
LM		 8926static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8927{
8928 struct imsm_dev *dev2;
8929 struct imsm_map *map;
8930 struct dl *idisk;
8931 int slot;
8932 int idx;
8933 __u8 state;
8934
8935 dev2 = get_imsm_dev(cont->sb, dev_idx);
8936 if (dev2) {
238c0a71 8937 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8938 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8939 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8940 if (!map)
8941 return 1;
8942 for (slot = 0; slot < map->num_members; slot++) {
8943 /*
8944 * Check if failed disks are deleted from intel
8945 * disk list or are marked to be deleted
8946 */
238c0a71 8947 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8948 idisk = get_imsm_dl_disk(cont->sb, idx);
8949 /*
8950 * Do not rebuild the array if failed disks
8951 * from failed sub-array are not removed from
8952 * container.
8953 */
8954 if (idisk &&
8955 is_failed(&idisk->disk) &&
8956 (idisk->action != DISK_REMOVE))
8957 return 0;
8958 }
8959 }
8960 }
8961 return 1;
8962}
8963
88758e9d
DW		 8964static struct mdinfo *imsm_activate_spare(struct active_array *a,
8965 struct metadata_update **updates)
8966{
8967 /**
d23fe947
DW		 8968 * Find a device with unused free space and use it to replace a
8969 * failed/vacant region in an array. We replace failed regions one a
8970 * array at a time. The result is that a new spare disk will be added
8971 * to the first failed array and after the monitor has finished
8972 * propagating failures the remainder will be consumed.
88758e9d 8973 *
d23fe947
DW		 8974 * FIXME add a capability for mdmon to request spares from another
8975 * container.
88758e9d
DW		 8976 */
8977
8978 struct intel_super *super = a->container->sb;
88758e9d 8979 int inst = a->info.container_member;
949c47a0 8980 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8981 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8982 int failed = a->info.array.raid_disks;
8983 struct mdinfo *rv = NULL;
8984 struct mdinfo *d;
8985 struct mdinfo *di;
8986 struct metadata_update *mu;
8987 struct dl *dl;
8988 struct imsm_update_activate_spare *u;
8989 int num_spares = 0;
8990 int i;
95d07a2c 8991 int allowed;
88758e9d
DW		 8992
8993 for (d = a->info.devs ; d ; d = d->next) {
8994 if ((d->curr_state & DS_FAULTY) &&
8995 d->state_fd >= 0)
8996 /* wait for Removal to happen */
8997 return NULL;
8998 if (d->state_fd >= 0)
8999 failed--;
9000 }
9001
9002 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
9003 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 9004
e2962bfc
AK		 9005 if (imsm_reshape_blocks_arrays_changes(super))
9006 return NULL;
1af97990 9007
fc8ca064
AK		 9008 /* Cannot activate another spare if rebuild is in progress already
9009 */
9010 if (is_rebuilding(dev)) {
7a862a02 9011 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 9012 return NULL;
9013 }
9014
89c67882
AK		 9015 if (a->info.array.level == 4)
9016 /* No repair for takeovered array
9017 * imsm doesn't support raid4
9018 */
9019 return NULL;
9020
3b451610
AK		 9021 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9022 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9023 return NULL;
9024
83ca7d45
AP		 9025 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9026 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9027 return NULL;
9028 }
9029
95d07a2c
LM		 9030 /*
9031 * If there are any failed disks check state of the other volume.
9032 * Block rebuild if the another one is failed until failed disks
9033 * are removed from container.
9034 */
9035 if (failed) {
7a862a02 9036 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9037 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9038 /* check if states of the other volumes allow for rebuild */
9039 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9040 if (i != inst) {
9041 allowed = imsm_rebuild_allowed(a->container,
9042 i, failed);
9043 if (!allowed)
9044 return NULL;
9045 }
9046 }
9047 }
9048
88758e9d 9049 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9050 for (i = 0; i < a->info.array.raid_disks; i++) {
9051 for (d = a->info.devs ; d ; d = d->next)
9052 if (d->disk.raid_disk == i)
9053 break;
9054 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9055 if (d && (d->state_fd >= 0))
9056 continue;
9057
272906ef 9058 /*
a20d2ba5
DW		 9059 * OK, this device needs recovery. Try to re-add the
9060 * previous occupant of this slot, if this fails see if
9061 * we can continue the assimilation of a spare that was
9062 * partially assimilated, finally try to activate a new
9063 * spare.
272906ef
DW		 9064 */
9065 dl = imsm_readd(super, i, a);
9066 if (!dl)
b303fe21 9067 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9068 if (!dl)
b303fe21 9069 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9070 if (!dl)
9071 continue;
1011e834 9072
272906ef 9073 /* found a usable disk with enough space */
503975b9 9074 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9075
9076 /* dl->index will be -1 in the case we are activating a
9077 * pristine spare. imsm_process_update() will create a
9078 * new index in this case. Once a disk is found to be
9079 * failed in all member arrays it is kicked from the
9080 * metadata
9081 */
9082 di->disk.number = dl->index;
d23fe947 9083
272906ef
DW		 9084 /* (ab)use di->devs to store a pointer to the device
9085 * we chose
9086 */
9087 di->devs = (struct mdinfo *) dl;
9088
9089 di->disk.raid_disk = i;
9090 di->disk.major = dl->major;
9091 di->disk.minor = dl->minor;
9092 di->disk.state = 0;
d23534e4 9093 di->recovery_start = 0;
5551b113 9094 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9095 di->component_size = a->info.component_size;
9096 di->container_member = inst;
5e46202e 9097 di->bb.supported = 1;
2c8890e9 9098 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9099 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9100 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9101 }
148acb7b 9102 super->random = random32();
272906ef
DW		 9103 di->next = rv;
9104 rv = di;
9105 num_spares++;
9106 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9107 i, di->data_offset);
88758e9d
DW		 9108 }
9109
9110 if (!rv)
9111 /* No spares found */
9112 return rv;
9113 /* Now 'rv' has a list of devices to return.
9114 * Create a metadata_update record to update the
9115 * disk_ord_tbl for the array
9116 */
503975b9 9117 mu = xmalloc(sizeof(*mu));
1011e834 9118 mu->buf = xcalloc(num_spares,
503975b9 9119 sizeof(struct imsm_update_activate_spare));
88758e9d 9120 mu->space = NULL;
cb23f1f4 9121 mu->space_list = NULL;
88758e9d
DW		 9122 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9123 mu->next = *updates;
9124 u = (struct imsm_update_activate_spare *) mu->buf;
9125
9126 for (di = rv ; di ; di = di->next) {
9127 u->type = update_activate_spare;
d23fe947
DW		 9128 u->dl = (struct dl *) di->devs;
9129 di->devs = NULL;
88758e9d
DW		 9130 u->slot = di->disk.raid_disk;
9131 u->array = inst;
9132 u->next = u + 1;
9133 u++;
9134 }
9135 (u-1)->next = NULL;
9136 *updates = mu;
9137
9138 return rv;
9139}
9140
54c2c1ea 9141static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9142{
54c2c1ea 9143 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9144 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9145 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9146 struct disk_info *inf = get_disk_info(u);
9147 struct imsm_disk *disk;
8273f55e
DW		 9148 int i;
9149 int j;
8273f55e 9150
54c2c1ea 9151 for (i = 0; i < map->num_members; i++) {
238c0a71 9152 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9153 for (j = 0; j < new_map->num_members; j++)
9154 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9155 return 1;
9156 }
9157
9158 return 0;
9159}
9160
1a64be56
LM		 9161static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9162{
594dc1b8
JS		 9163 struct dl *dl;
9164
1a64be56 9165 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9166 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9167 return dl;
9168 return NULL;
9169}
9170
9171static int remove_disk_super(struct intel_super *super, int major, int minor)
9172{
594dc1b8 9173 struct dl *prev;
1a64be56
LM		 9174 struct dl *dl;
9175
9176 prev = NULL;
9177 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9178 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9179 /* remove */
9180 if (prev)
9181 prev->next = dl->next;
9182 else
9183 super->disks = dl->next;
9184 dl->next = NULL;
9185 __free_imsm_disk(dl);
1ade5cc1 9186 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9187 break;
9188 }
9189 prev = dl;
9190 }
9191 return 0;
9192}
9193
f21e18ca 9194static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9195
1a64be56
LM		 9196static int add_remove_disk_update(struct intel_super *super)
9197{
9198 int check_degraded = 0;
594dc1b8
JS		 9199 struct dl *disk;
9200
1a64be56
LM		 9201 /* add/remove some spares to/from the metadata/contrainer */
9202 while (super->disk_mgmt_list) {
9203 struct dl *disk_cfg;
9204
9205 disk_cfg = super->disk_mgmt_list;
9206 super->disk_mgmt_list = disk_cfg->next;
9207 disk_cfg->next = NULL;
9208
9209 if (disk_cfg->action == DISK_ADD) {
9210 disk_cfg->next = super->disks;
9211 super->disks = disk_cfg;
9212 check_degraded = 1;
1ade5cc1
N		 9213 dprintf("added %x:%x\n",
9214 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9215 } else if (disk_cfg->action == DISK_REMOVE) {
9216 dprintf("Disk remove action processed: %x.%x\n",
9217 disk_cfg->major, disk_cfg->minor);
9218 disk = get_disk_super(super,
9219 disk_cfg->major,
9220 disk_cfg->minor);
9221 if (disk) {
9222 /* store action status */
9223 disk->action = DISK_REMOVE;
9224 /* remove spare disks only */
9225 if (disk->index == -1) {
9226 remove_disk_super(super,
9227 disk_cfg->major,
9228 disk_cfg->minor);
91c97c54
MT		 9229 } else {
9230 disk_cfg->fd = disk->fd;
9231 disk->fd = -1;
1a64be56
LM		 9232 }
9233 }
9234 /* release allocate disk structure */
9235 __free_imsm_disk(disk_cfg);
9236 }
9237 }
9238 return check_degraded;
9239}
9240
a29911da
PC		 9241static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9242 struct intel_super *super,
9243 void ***space_list)
9244{
9245 struct intel_dev *id;
9246 void **tofree = NULL;
9247 int ret_val = 0;
9248
1ade5cc1 9249 dprintf("(enter)\n");
089f9d79 9250 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9251 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9252 return ret_val;
9253 }
089f9d79 9254 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9255 dprintf("imsm: Error: Memory is not allocated\n");
9256 return ret_val;
9257 }
9258
9259 for (id = super->devlist ; id; id = id->next) {
9260 if (id->index == (unsigned)u->subdev) {
9261 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9262 struct imsm_map *map;
9263 struct imsm_dev *new_dev =
9264 (struct imsm_dev *)*space_list;
238c0a71 9265 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9266 int to_state;
9267 struct dl *new_disk;
9268
9269 if (new_dev == NULL)
9270 return ret_val;
9271 *space_list = **space_list;
9272 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9273 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9274 if (migr_map) {
9275 dprintf("imsm: Error: migration in progress");
9276 return ret_val;
9277 }
9278
9279 to_state = map->map_state;
9280 if ((u->new_level == 5) && (map->raid_level == 0)) {
9281 map->num_members++;
9282 /* this should not happen */
9283 if (u->new_disks[0] < 0) {
9284 map->failed_disk_num =
9285 map->num_members - 1;
9286 to_state = IMSM_T_STATE_DEGRADED;
9287 } else
9288 to_state = IMSM_T_STATE_NORMAL;
9289 }
8e59f3d8 9290 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9291 if (u->new_level > -1)
9292 map->raid_level = u->new_level;
238c0a71 9293 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9294 if ((u->new_level == 5) &&
9295 (migr_map->raid_level == 0)) {
9296 int ord = map->num_members - 1;
9297 migr_map->num_members--;
9298 if (u->new_disks[0] < 0)
9299 ord |= IMSM_ORD_REBUILD;
9300 set_imsm_ord_tbl_ent(map,
9301 map->num_members - 1,
9302 ord);
9303 }
9304 id->dev = new_dev;
9305 tofree = (void **)dev;
9306
4bba0439
PC		 9307 /* update chunk size
9308 */
06fb291a
PB		 9309 if (u->new_chunksize > 0) {
9310 unsigned long long num_data_stripes;
9529d343
MD		 9311 struct imsm_map *dest_map =
9312 get_imsm_map(dev, MAP_0);
06fb291a 9313 int used_disks =
9529d343 9314 imsm_num_data_members(dest_map);
06fb291a
PB		 9315
9316 if (used_disks == 0)
9317 return ret_val;
9318
4bba0439
PC		 9319 map->blocks_per_strip =
9320 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9321 num_data_stripes =
fcc2c9da 9322 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9323 num_data_stripes /= map->blocks_per_strip;
9324 num_data_stripes /= map->num_domains;
9325 set_num_data_stripes(map, num_data_stripes);
9326 }
4bba0439 9327
44490938
MD		 9328 /* ensure blocks_per_member has valid value
9329 */
9330 set_blocks_per_member(map,
9331 per_dev_array_size(map) +
9332 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9333
a29911da
PC		 9334 /* add disk
9335 */
089f9d79
JS		 9336 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9337 migr_map->raid_level == map->raid_level)
a29911da
PC		 9338 goto skip_disk_add;
9339
9340 if (u->new_disks[0] >= 0) {
9341 /* use passes spare
9342 */
9343 new_disk = get_disk_super(super,
9344 major(u->new_disks[0]),
9345 minor(u->new_disks[0]));
7a862a02 9346 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9347 major(u->new_disks[0]),
9348 minor(u->new_disks[0]),
9349 new_disk, new_disk->index);
9350 if (new_disk == NULL)
9351 goto error_disk_add;
9352
9353 new_disk->index = map->num_members - 1;
9354 /* slot to fill in autolayout
9355 */
9356 new_disk->raiddisk = new_disk->index;
9357 new_disk->disk.status |= CONFIGURED_DISK;
9358 new_disk->disk.status &= ~SPARE_DISK;
9359 } else
9360 goto error_disk_add;
9361
9362skip_disk_add:
9363 *tofree = *space_list;
9364 /* calculate new size
9365 */
f3871fdc 9366 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9367
9368 ret_val = 1;
9369 }
9370 }
9371
9372 if (tofree)
9373 *space_list = tofree;
9374 return ret_val;
9375
9376error_disk_add:
9377 dprintf("Error: imsm: Cannot find disk.\n");
9378 return ret_val;
9379}
9380
f3871fdc
AK		 9381static int apply_size_change_update(struct imsm_update_size_change *u,
9382 struct intel_super *super)
9383{
9384 struct intel_dev *id;
9385 int ret_val = 0;
9386
1ade5cc1 9387 dprintf("(enter)\n");
089f9d79 9388 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9389 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9390 return ret_val;
9391 }
9392
9393 for (id = super->devlist ; id; id = id->next) {
9394 if (id->index == (unsigned)u->subdev) {
9395 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9396 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9397 int used_disks = imsm_num_data_members(map);
f3871fdc 9398 unsigned long long blocks_per_member;
06fb291a 9399 unsigned long long num_data_stripes;
44490938
MD		 9400 unsigned long long new_size_per_disk;
9401
9402 if (used_disks == 0)
9403 return 0;
f3871fdc
AK		 9404
9405 /* calculate new size
9406 */
44490938
MD		 9407 new_size_per_disk = u->new_size / used_disks;
9408 blocks_per_member = new_size_per_disk +
9409 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9410 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9411 map->blocks_per_strip;
9412 num_data_stripes /= map->num_domains;
9413 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9414 u->new_size, new_size_per_disk,
06fb291a 9415 num_data_stripes);
f3871fdc 9416 set_blocks_per_member(map, blocks_per_member);
06fb291a 9417 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9418 imsm_set_array_size(dev, u->new_size);
9419
9420 ret_val = 1;
9421 break;
9422 }
9423 }
9424
9425 return ret_val;
9426}
9427
061d7da3 9428static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9429 struct intel_super *super,
061d7da3
LO		 9430 struct active_array *active_array)
9431{
9432 struct imsm_super *mpb = super->anchor;
9433 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9434 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9435 struct imsm_map *migr_map;
9436 struct active_array *a;
9437 struct imsm_disk *disk;
9438 __u8 to_state;
9439 struct dl *dl;
9440 unsigned int found;
9441 int failed;
5961eeec 9442 int victim;
061d7da3 9443 int i;
5961eeec 9444 int second_map_created = 0;
061d7da3 9445
5961eeec 9446 for (; u; u = u->next) {
238c0a71 9447 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9448
5961eeec 9449 if (victim < 0)
9450 return 0;
061d7da3 9451
5961eeec 9452 for (dl = super->disks; dl; dl = dl->next)
9453 if (dl == u->dl)
9454 break;
061d7da3 9455
5961eeec 9456 if (!dl) {
7a862a02 9457 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9458 u->dl->index);
9459 return 0;
9460 }
061d7da3 9461
5961eeec 9462 /* count failures (excluding rebuilds and the victim)
9463 * to determine map[0] state
9464 */
9465 failed = 0;
9466 for (i = 0; i < map->num_members; i++) {
9467 if (i == u->slot)
9468 continue;
9469 disk = get_imsm_disk(super,
238c0a71 9470 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9471 if (!disk || is_failed(disk))
9472 failed++;
9473 }
061d7da3 9474
5961eeec 9475 /* adding a pristine spare, assign a new index */
9476 if (dl->index < 0) {
9477 dl->index = super->anchor->num_disks;
9478 super->anchor->num_disks++;
9479 }
9480 disk = &dl->disk;
9481 disk->status |= CONFIGURED_DISK;
9482 disk->status &= ~SPARE_DISK;
9483
9484 /* mark rebuild */
238c0a71 9485 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9486 if (!second_map_created) {
9487 second_map_created = 1;
9488 map->map_state = IMSM_T_STATE_DEGRADED;
9489 migrate(dev, super, to_state, MIGR_REBUILD);
9490 } else
9491 map->map_state = to_state;
238c0a71 9492 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9493 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9494 set_imsm_ord_tbl_ent(migr_map, u->slot,
9495 dl->index | IMSM_ORD_REBUILD);
9496
9497 /* update the family_num to mark a new container
9498 * generation, being careful to record the existing
9499 * family_num in orig_family_num to clean up after
9500 * earlier mdadm versions that neglected to set it.
9501 */
9502 if (mpb->orig_family_num == 0)
9503 mpb->orig_family_num = mpb->family_num;
9504 mpb->family_num += super->random;
9505
9506 /* count arrays using the victim in the metadata */
9507 found = 0;
9508 for (a = active_array; a ; a = a->next) {
9509 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9510 map = get_imsm_map(dev, MAP_0);
061d7da3 9511
5961eeec 9512 if (get_imsm_disk_slot(map, victim) >= 0)
9513 found++;
9514 }
061d7da3 9515
5961eeec 9516 /* delete the victim if it is no longer being
9517 * utilized anywhere
061d7da3 9518 */
5961eeec 9519 if (!found) {
9520 struct dl **dlp;
061d7da3 9521
5961eeec 9522 /* We know that 'manager' isn't touching anything,
9523 * so it is safe to delete
9524 */
9525 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9526 if ((*dlp)->index == victim)
9527 break;
5961eeec 9528
9529 /* victim may be on the missing list */
9530 if (!*dlp)
9531 for (dlp = &super->missing; *dlp;
9532 dlp = &(*dlp)->next)
9533 if ((*dlp)->index == victim)
9534 break;
9535 imsm_delete(super, dlp, victim);
9536 }
061d7da3
LO		 9537 }
9538
9539 return 1;
9540}
a29911da 9541
2e5dc010
N		 9542static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9543 struct intel_super *super,
9544 void ***space_list)
9545{
9546 struct dl *new_disk;
9547 struct intel_dev *id;
9548 int i;
9549 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9550 int disk_count = u->old_raid_disks;
2e5dc010
N		 9551 void **tofree = NULL;
9552 int devices_to_reshape = 1;
9553 struct imsm_super *mpb = super->anchor;
9554 int ret_val = 0;
d098291a 9555 unsigned int dev_id;
2e5dc010 9556
1ade5cc1 9557 dprintf("(enter)\n");
2e5dc010
N		 9558
9559 /* enable spares to use in array */
9560 for (i = 0; i < delta_disks; i++) {
9561 new_disk = get_disk_super(super,
9562 major(u->new_disks[i]),
9563 minor(u->new_disks[i]));
7a862a02 9564 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9565 major(u->new_disks[i]), minor(u->new_disks[i]),
9566 new_disk, new_disk->index);
089f9d79
JS		 9567 if (new_disk == NULL ||
9568 (new_disk->index >= 0 &&
9569 new_disk->index < u->old_raid_disks))
2e5dc010 9570 goto update_reshape_exit;
ee4beede 9571 new_disk->index = disk_count++;
2e5dc010
N		 9572 /* slot to fill in autolayout
9573 */
9574 new_disk->raiddisk = new_disk->index;
9575 new_disk->disk.status |=
9576 CONFIGURED_DISK;
9577 new_disk->disk.status &= ~SPARE_DISK;
9578 }
9579
ed7333bd
AK		 9580 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9581 mpb->num_raid_devs);
2e5dc010
N		 9582 /* manage changes in volume
9583 */
d098291a 9584 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9585 void **sp = *space_list;
9586 struct imsm_dev *newdev;
9587 struct imsm_map *newmap, *oldmap;
9588
d098291a
AK		 9589 for (id = super->devlist ; id; id = id->next) {
9590 if (id->index == dev_id)
9591 break;
9592 }
9593 if (id == NULL)
9594 break;
2e5dc010
N		 9595 if (!sp)
9596 continue;
9597 *space_list = *sp;
9598 newdev = (void*)sp;
9599 /* Copy the dev, but not (all of) the map */
9600 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9601 oldmap = get_imsm_map(id->dev, MAP_0);
9602 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9603 /* Copy the current map */
9604 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9605 /* update one device only
9606 */
9607 if (devices_to_reshape) {
ed7333bd
AK		 9608 dprintf("imsm: modifying subdev: %i\n",
9609 id->index);
2e5dc010
N		 9610 devices_to_reshape--;
9611 newdev->vol.migr_state = 1;
9612 newdev->vol.curr_migr_unit = 0;
ea672ee1 9613 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9614 newmap->num_members = u->new_raid_disks;
9615 for (i = 0; i < delta_disks; i++) {
9616 set_imsm_ord_tbl_ent(newmap,
9617 u->old_raid_disks + i,
9618 u->old_raid_disks + i);
9619 }
9620 /* New map is correct, now need to save old map
9621 */
238c0a71 9622 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9623 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9624
f3871fdc 9625 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9626 }
9627
9628 sp = (void **)id->dev;
9629 id->dev = newdev;
9630 *sp = tofree;
9631 tofree = sp;
8e59f3d8
AK		 9632
9633 /* Clear migration record */
9634 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9635 }
819bc634
AK		 9636 if (tofree)
9637 *space_list = tofree;
2e5dc010
N		 9638 ret_val = 1;
9639
9640update_reshape_exit:
9641
9642 return ret_val;
9643}
9644
bb025c2f 9645static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9646 struct intel_super *super,
9647 void ***space_list)
bb025c2f
KW		 9648{
9649 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9650 struct intel_dev *dv;
9651 struct imsm_dev *dev_new;
bb025c2f
KW		 9652 struct imsm_map *map;
9653 struct dl *dm, *du;
8ca6df95 9654 int i;
bb025c2f
KW		 9655
9656 for (dv = super->devlist; dv; dv = dv->next)
9657 if (dv->index == (unsigned int)u->subarray) {
9658 dev = dv->dev;
9659 break;
9660 }
9661
9662 if (dev == NULL)
9663 return 0;
9664
238c0a71 9665 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9666
9667 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9668 unsigned long long num_data_stripes;
9669
43d5ec18 9670 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9671 if (imsm_count_failed(super, dev, MAP_0) !=
9672 (map->num_members / 2))
43d5ec18
KW		 9673 return 0;
9674
bb025c2f
KW		 9675 /* iterate through devices to mark removed disks as spare */
9676 for (dm = super->disks; dm; dm = dm->next) {
9677 if (dm->disk.status & FAILED_DISK) {
9678 int idx = dm->index;
9679 /* update indexes on the disk list */
9680/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9681 the index values will end up being correct.... NB */
9682 for (du = super->disks; du; du = du->next)
9683 if (du->index > idx)
9684 du->index--;
9685 /* mark as spare disk */
a8619d23 9686 mark_spare(dm);
bb025c2f
KW		 9687 }
9688 }
bb025c2f
KW		 9689 /* update map */
9690 map->num_members = map->num_members / 2;
9691 map->map_state = IMSM_T_STATE_NORMAL;
9692 map->num_domains = 1;
9693 map->raid_level = 0;
9694 map->failed_disk_num = -1;
4a353e6e
RS		 9695 num_data_stripes = imsm_dev_size(dev) / 2;
9696 num_data_stripes /= map->blocks_per_strip;
9697 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9698 }
9699
8ca6df95
KW		 9700 if (u->direction == R0_TO_R10) {
9701 void **space;
4a353e6e
RS		 9702 unsigned long long num_data_stripes;
9703
8ca6df95
KW		 9704 /* update slots in current disk list */
9705 for (dm = super->disks; dm; dm = dm->next) {
9706 if (dm->index >= 0)
9707 dm->index *= 2;
9708 }
9709 /* create new *missing* disks */
9710 for (i = 0; i < map->num_members; i++) {
9711 space = *space_list;
9712 if (!space)
9713 continue;
9714 *space_list = *space;
9715 du = (void *)space;
9716 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9717 du->fd = -1;
9718 du->minor = 0;
9719 du->major = 0;
9720 du->index = (i * 2) + 1;
9721 sprintf((char *)du->disk.serial,
9722 " MISSING_%d", du->index);
9723 sprintf((char *)du->serial,
9724 "MISSING_%d", du->index);
9725 du->next = super->missing;
9726 super->missing = du;
9727 }
9728 /* create new dev and map */
9729 space = *space_list;
9730 if (!space)
9731 return 0;
9732 *space_list = *space;
9733 dev_new = (void *)space;
9734 memcpy(dev_new, dev, sizeof(*dev));
9735 /* update new map */
238c0a71 9736 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9737 map->num_members = map->num_members * 2;
1a2487c2 9738 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9739 map->num_domains = 2;
9740 map->raid_level = 1;
4a353e6e
RS		 9741 num_data_stripes = imsm_dev_size(dev) / 2;
9742 num_data_stripes /= map->blocks_per_strip;
9743 num_data_stripes /= map->num_domains;
9744 set_num_data_stripes(map, num_data_stripes);
9745
8ca6df95
KW		 9746 /* replace dev<->dev_new */
9747 dv->dev = dev_new;
9748 }
bb025c2f
KW		 9749 /* update disk order table */
9750 for (du = super->disks; du; du = du->next)
9751 if (du->index >= 0)
9752 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9753 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9754 if (du->index >= 0) {
9755 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9756 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9757 }
bb025c2f
KW		 9758
9759 return 1;
9760}
9761
e8319a19
DW		 9762static void imsm_process_update(struct supertype *st,
9763 struct metadata_update *update)
9764{
9765 /**
9766 * crack open the metadata_update envelope to find the update record
9767 * update can be one of:
d195167d
AK		 9768 * update_reshape_container_disks - all the arrays in the container
9769 * are being reshaped to have more devices. We need to mark
9770 * the arrays for general migration and convert selected spares
9771 * into active devices.
9772 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9773 * device in an array, update the disk_ord_tbl. If this disk is
9774 * present in all member arrays then also clear the SPARE_DISK
9775 * flag
d195167d
AK		 9776 * update_create_array
9777 * update_kill_array
9778 * update_rename_array
9779 * update_add_remove_disk
e8319a19
DW		 9780 */
9781 struct intel_super *super = st->sb;
4d7b1503 9782 struct imsm_super *mpb;
e8319a19
DW		 9783 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9784
4d7b1503
DW		 9785 /* update requires a larger buf but the allocation failed */
9786 if (super->next_len && !super->next_buf) {
9787 super->next_len = 0;
9788 return;
9789 }
9790
9791 if (super->next_buf) {
9792 memcpy(super->next_buf, super->buf, super->len);
9793 free(super->buf);
9794 super->len = super->next_len;
9795 super->buf = super->next_buf;
9796
9797 super->next_len = 0;
9798 super->next_buf = NULL;
9799 }
9800
9801 mpb = super->anchor;
9802
e8319a19 9803 switch (type) {
0ec5d470
AK		 9804 case update_general_migration_checkpoint: {
9805 struct intel_dev *id;
9806 struct imsm_update_general_migration_checkpoint *u =
9807 (void *)update->buf;
9808
1ade5cc1 9809 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9810
9811 /* find device under general migration */
9812 for (id = super->devlist ; id; id = id->next) {
9813 if (is_gen_migration(id->dev)) {
9814 id->dev->vol.curr_migr_unit =
9815 __cpu_to_le32(u->curr_migr_unit);
9816 super->updates_pending++;
9817 }
9818 }
9819 break;
9820 }
bb025c2f
KW		 9821 case update_takeover: {
9822 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9823 if (apply_takeover_update(u, super, &update->space_list)) {
9824 imsm_update_version_info(super);
bb025c2f 9825 super->updates_pending++;
1a2487c2 9826 }
bb025c2f
KW		 9827 break;
9828 }
9829
78b10e66 9830 case update_reshape_container_disks: {
d195167d 9831 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9832 if (apply_reshape_container_disks_update(
9833 u, super, &update->space_list))
9834 super->updates_pending++;
78b10e66
N		 9835 break;
9836 }
48c5303a 9837 case update_reshape_migration: {
a29911da
PC		 9838 struct imsm_update_reshape_migration *u = (void *)update->buf;
9839 if (apply_reshape_migration_update(
9840 u, super, &update->space_list))
9841 super->updates_pending++;
48c5303a
PC		 9842 break;
9843 }
f3871fdc
AK		 9844 case update_size_change: {
9845 struct imsm_update_size_change *u = (void *)update->buf;
9846 if (apply_size_change_update(u, super))
9847 super->updates_pending++;
9848 break;
9849 }
e8319a19 9850 case update_activate_spare: {
1011e834 9851 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9852 if (apply_update_activate_spare(u, super, st->arrays))
9853 super->updates_pending++;
8273f55e
DW		 9854 break;
9855 }
9856 case update_create_array: {
9857 /* someone wants to create a new array, we need to be aware of
9858 * a few races/collisions:
9859 * 1/ 'Create' called by two separate instances of mdadm
9860 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9861 * devices that have since been assimilated via
9862 * activate_spare.
9863 * In the event this update can not be carried out mdadm will
9864 * (FIX ME) notice that its update did not take hold.
9865 */
9866 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9867 struct intel_dev *dv;
8273f55e
DW		 9868 struct imsm_dev *dev;
9869 struct imsm_map *map, *new_map;
9870 unsigned long long start, end;
9871 unsigned long long new_start, new_end;
9872 int i;
54c2c1ea
DW		 9873 struct disk_info *inf;
9874 struct dl *dl;
8273f55e
DW		 9875
9876 /* handle racing creates: first come first serve */
9877 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9878 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9879 goto create_error;
8273f55e
DW		 9880 }
9881
9882 /* check update is next in sequence */
9883 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9884 dprintf("can not create array %d expected index %d\n",
9885 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9886 goto create_error;
8273f55e
DW		 9887 }
9888
238c0a71 9889 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9890 new_start = pba_of_lba0(new_map);
44490938 9891 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9892 inf = get_disk_info(u);
8273f55e
DW		 9893
9894 /* handle activate_spare versus create race:
9895 * check to make sure that overlapping arrays do not include
9896 * overalpping disks
9897 */
9898 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9899 dev = get_imsm_dev(super, i);
238c0a71 9900 map = get_imsm_map(dev, MAP_0);
5551b113 9901 start = pba_of_lba0(map);
44490938 9902 end = start + per_dev_array_size(map);
8273f55e
DW		 9903 if ((new_start >= start && new_start <= end) ||
9904 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9905 /* overlap */;
9906 else
9907 continue;
9908
9909 if (disks_overlap(super, i, u)) {
1ade5cc1 9910 dprintf("arrays overlap\n");
ba2de7ba 9911 goto create_error;
8273f55e
DW		 9912 }
9913 }
8273f55e 9914
949c47a0
DW		 9915 /* check that prepare update was successful */
9916 if (!update->space) {
1ade5cc1 9917 dprintf("prepare update failed\n");
ba2de7ba 9918 goto create_error;
949c47a0
DW		 9919 }
9920
54c2c1ea
DW		 9921 /* check that all disks are still active before committing
9922 * changes. FIXME: could we instead handle this by creating a
9923 * degraded array? That's probably not what the user expects,
9924 * so better to drop this update on the floor.
9925 */
9926 for (i = 0; i < new_map->num_members; i++) {
9927 dl = serial_to_dl(inf[i].serial, super);
9928 if (!dl) {
1ade5cc1 9929 dprintf("disk disappeared\n");
ba2de7ba 9930 goto create_error;
54c2c1ea 9931 }
949c47a0
DW		 9932 }
9933
8273f55e 9934 super->updates_pending++;
54c2c1ea
DW		 9935
9936 /* convert spares to members and fixup ord_tbl */
9937 for (i = 0; i < new_map->num_members; i++) {
9938 dl = serial_to_dl(inf[i].serial, super);
9939 if (dl->index == -1) {
9940 dl->index = mpb->num_disks;
9941 mpb->num_disks++;
9942 dl->disk.status |= CONFIGURED_DISK;
9943 dl->disk.status &= ~SPARE_DISK;
9944 }
9945 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9946 }
9947
ba2de7ba
DW		 9948 dv = update->space;
9949 dev = dv->dev;
949c47a0
DW		 9950 update->space = NULL;
9951 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9952 dv->index = u->dev_idx;
9953 dv->next = super->devlist;
9954 super->devlist = dv;
8273f55e 9955 mpb->num_raid_devs++;
8273f55e 9956
4d1313e9 9957 imsm_update_version_info(super);
8273f55e 9958 break;
ba2de7ba
DW		 9959 create_error:
9960 /* mdmon knows how to release update->space, but not
9961 * ((struct intel_dev *) update->space)->dev
9962 */
9963 if (update->space) {
9964 dv = update->space;
9965 free(dv->dev);
9966 }
8273f55e 9967 break;
e8319a19 9968 }
33414a01
DW		 9969 case update_kill_array: {
9970 struct imsm_update_kill_array *u = (void *) update->buf;
9971 int victim = u->dev_idx;
9972 struct active_array *a;
9973 struct intel_dev **dp;
9974 struct imsm_dev *dev;
9975
9976 /* sanity check that we are not affecting the uuid of
9977 * active arrays, or deleting an active array
9978 *
9979 * FIXME when immutable ids are available, but note that
9980 * we'll also need to fixup the invalidated/active
9981 * subarray indexes in mdstat
9982 */
9983 for (a = st->arrays; a; a = a->next)
9984 if (a->info.container_member >= victim)
9985 break;
9986 /* by definition if mdmon is running at least one array
9987 * is active in the container, so checking
9988 * mpb->num_raid_devs is just extra paranoia
9989 */
9990 dev = get_imsm_dev(super, victim);
9991 if (a || !dev || mpb->num_raid_devs == 1) {
9992 dprintf("failed to delete subarray-%d\n", victim);
9993 break;
9994 }
9995
9996 for (dp = &super->devlist; *dp;)
f21e18ca 9997 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9998 *dp = (*dp)->next;
9999 } else {
f21e18ca 10000 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 10001 (*dp)->index--;
10002 dp = &(*dp)->next;
10003 }
10004 mpb->num_raid_devs--;
10005 super->updates_pending++;
10006 break;
10007 }
aa534678
DW		 10008 case update_rename_array: {
10009 struct imsm_update_rename_array *u = (void *) update->buf;
10010 char name[MAX_RAID_SERIAL_LEN+1];
10011 int target = u->dev_idx;
10012 struct active_array *a;
10013 struct imsm_dev *dev;
10014
10015 /* sanity check that we are not affecting the uuid of
10016 * an active array
10017 */
40659392 10018 memset(name, 0, sizeof(name));
aa534678
DW		 10019 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
10020 name[MAX_RAID_SERIAL_LEN] = '\0';
10021 for (a = st->arrays; a; a = a->next)
10022 if (a->info.container_member == target)
10023 break;
10024 dev = get_imsm_dev(super, u->dev_idx);
10025 if (a || !dev || !check_name(super, name, 1)) {
10026 dprintf("failed to rename subarray-%d\n", target);
10027 break;
10028 }
10029
40659392 10030 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10031 super->updates_pending++;
10032 break;
10033 }
1a64be56 10034 case update_add_remove_disk: {
43dad3d6 10035 /* we may be able to repair some arrays if disks are
095b8088 10036 * being added, check the status of add_remove_disk
1a64be56
LM		 10037 * if discs has been added.
10038 */
10039 if (add_remove_disk_update(super)) {
43dad3d6 10040 struct active_array *a;
072b727f
DW		 10041
10042 super->updates_pending++;
1a64be56 10043 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10044 a->check_degraded = 1;
10045 }
43dad3d6 10046 break;
e8319a19 10047 }
bbab0940
TM		 10048 case update_prealloc_badblocks_mem:
10049 break;
e6e9dd3f
AP		 10050 case update_rwh_policy: {
10051 struct imsm_update_rwh_policy *u = (void *)update->buf;
10052 int target = u->dev_idx;
10053 struct imsm_dev *dev = get_imsm_dev(super, target);
10054 if (!dev) {
10055 dprintf("could not find subarray-%d\n", target);
10056 break;
10057 }
10058
10059 if (dev->rwh_policy != u->new_policy) {
10060 dev->rwh_policy = u->new_policy;
10061 super->updates_pending++;
10062 }
10063 break;
10064 }
1a64be56 10065 default:
ebf3be99 10066 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10067 }
e8319a19 10068}
88758e9d 10069
bc0b9d34
PC		 10070static struct mdinfo *get_spares_for_grow(struct supertype *st);
10071
5fe6f031
N		 10072static int imsm_prepare_update(struct supertype *st,
10073 struct metadata_update *update)
8273f55e 10074{
949c47a0 10075 /**
4d7b1503
DW		 10076 * Allocate space to hold new disk entries, raid-device entries or a new
10077 * mpb if necessary. The manager synchronously waits for updates to
10078 * complete in the monitor, so new mpb buffers allocated here can be
10079 * integrated by the monitor thread without worrying about live pointers
10080 * in the manager thread.
8273f55e 10081 */
095b8088 10082 enum imsm_update_type type;
4d7b1503 10083 struct intel_super *super = st->sb;
f36a9ecd 10084 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10085 struct imsm_super *mpb = super->anchor;
10086 size_t buf_len;
10087 size_t len = 0;
949c47a0 10088
095b8088
N		 10089 if (update->len < (int)sizeof(type))
10090 return 0;
10091
10092 type = *(enum imsm_update_type *) update->buf;
10093
949c47a0 10094 switch (type) {
0ec5d470 10095 case update_general_migration_checkpoint:
095b8088
N		 10096 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10097 return 0;
1ade5cc1 10098 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10099 break;
abedf5fc
KW		 10100 case update_takeover: {
10101 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10102 if (update->len < (int)sizeof(*u))
10103 return 0;
abedf5fc
KW		 10104 if (u->direction == R0_TO_R10) {
10105 void **tail = (void **)&update->space_list;
10106 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10107 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10108 int num_members = map->num_members;
10109 void *space;
10110 int size, i;
abedf5fc
KW		 10111 /* allocate memory for added disks */
10112 for (i = 0; i < num_members; i++) {
10113 size = sizeof(struct dl);
503975b9 10114 space = xmalloc(size);
abedf5fc
KW		 10115 *tail = space;
10116 tail = space;
10117 *tail = NULL;
10118 }
10119 /* allocate memory for new device */
10120 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10121 (num_members * sizeof(__u32));
503975b9
N		 10122 space = xmalloc(size);
10123 *tail = space;
10124 tail = space;
10125 *tail = NULL;
10126 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10127 }
10128
10129 break;
10130 }
78b10e66 10131 case update_reshape_container_disks: {
d195167d
AK		 10132 /* Every raid device in the container is about to
10133 * gain some more devices, and we will enter a
10134 * reconfiguration.
10135 * So each 'imsm_map' will be bigger, and the imsm_vol
10136 * will now hold 2 of them.
10137 * Thus we need new 'struct imsm_dev' allocations sized
10138 * as sizeof_imsm_dev but with more devices in both maps.
10139 */
10140 struct imsm_update_reshape *u = (void *)update->buf;
10141 struct intel_dev *dl;
10142 void **space_tail = (void**)&update->space_list;
10143
095b8088
N		 10144 if (update->len < (int)sizeof(*u))
10145 return 0;
10146
1ade5cc1 10147 dprintf("for update_reshape\n");
d195167d
AK		 10148
10149 for (dl = super->devlist; dl; dl = dl->next) {
10150 int size = sizeof_imsm_dev(dl->dev, 1);
10151 void *s;
d677e0b8
AK		 10152 if (u->new_raid_disks > u->old_raid_disks)
10153 size += sizeof(__u32)*2*
10154 (u->new_raid_disks - u->old_raid_disks);
503975b9 10155 s = xmalloc(size);
d195167d
AK		 10156 *space_tail = s;
10157 space_tail = s;
10158 *space_tail = NULL;
10159 }
10160
10161 len = disks_to_mpb_size(u->new_raid_disks);
10162 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10163 break;
10164 }
48c5303a 10165 case update_reshape_migration: {
bc0b9d34
PC		 10166 /* for migration level 0->5 we need to add disks
10167 * so the same as for container operation we will copy
10168 * device to the bigger location.
10169 * in memory prepared device and new disk area are prepared
10170 * for usage in process update
10171 */
10172 struct imsm_update_reshape_migration *u = (void *)update->buf;
10173 struct intel_dev *id;
10174 void **space_tail = (void **)&update->space_list;
10175 int size;
10176 void *s;
10177 int current_level = -1;
10178
095b8088
N		 10179 if (update->len < (int)sizeof(*u))
10180 return 0;
10181
1ade5cc1 10182 dprintf("for update_reshape\n");
bc0b9d34
PC		 10183
10184 /* add space for bigger array in update
10185 */
10186 for (id = super->devlist; id; id = id->next) {
10187 if (id->index == (unsigned)u->subdev) {
10188 size = sizeof_imsm_dev(id->dev, 1);
10189 if (u->new_raid_disks > u->old_raid_disks)
10190 size += sizeof(__u32)*2*
10191 (u->new_raid_disks - u->old_raid_disks);
503975b9 10192 s = xmalloc(size);
bc0b9d34
PC		 10193 *space_tail = s;
10194 space_tail = s;
10195 *space_tail = NULL;
10196 break;
10197 }
10198 }
10199 if (update->space_list == NULL)
10200 break;
10201
10202 /* add space for disk in update
10203 */
10204 size = sizeof(struct dl);
503975b9 10205 s = xmalloc(size);
bc0b9d34
PC		 10206 *space_tail = s;
10207 space_tail = s;
10208 *space_tail = NULL;
10209
10210 /* add spare device to update
10211 */
10212 for (id = super->devlist ; id; id = id->next)
10213 if (id->index == (unsigned)u->subdev) {
10214 struct imsm_dev *dev;
10215 struct imsm_map *map;
10216
10217 dev = get_imsm_dev(super, u->subdev);
238c0a71 10218 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10219 current_level = map->raid_level;
10220 break;
10221 }
089f9d79 10222 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10223 struct mdinfo *spares;
10224
10225 spares = get_spares_for_grow(st);
10226 if (spares) {
10227 struct dl *dl;
10228 struct mdinfo *dev;
10229
10230 dev = spares->devs;
10231 if (dev) {
10232 u->new_disks[0] =
10233 makedev(dev->disk.major,
10234 dev->disk.minor);
10235 dl = get_disk_super(super,
10236 dev->disk.major,
10237 dev->disk.minor);
10238 dl->index = u->old_raid_disks;
10239 dev = dev->next;
10240 }
10241 sysfs_free(spares);
10242 }
10243 }
10244 len = disks_to_mpb_size(u->new_raid_disks);
10245 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10246 break;
10247 }
f3871fdc 10248 case update_size_change: {
095b8088
N		 10249 if (update->len < (int)sizeof(struct imsm_update_size_change))
10250 return 0;
10251 break;
10252 }
10253 case update_activate_spare: {
10254 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10255 return 0;
f3871fdc
AK		 10256 break;
10257 }
949c47a0
DW		 10258 case update_create_array: {
10259 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10260 struct intel_dev *dv;
54c2c1ea 10261 struct imsm_dev *dev = &u->dev;
238c0a71 10262 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10263 struct dl *dl;
10264 struct disk_info *inf;
10265 int i;
10266 int activate = 0;
949c47a0 10267
095b8088
N		 10268 if (update->len < (int)sizeof(*u))
10269 return 0;
10270
54c2c1ea
DW		 10271 inf = get_disk_info(u);
10272 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10273 /* allocate a new super->devlist entry */
503975b9
N		 10274 dv = xmalloc(sizeof(*dv));
10275 dv->dev = xmalloc(len);
10276 update->space = dv;
949c47a0 10277
54c2c1ea
DW		 10278 /* count how many spares will be converted to members */
10279 for (i = 0; i < map->num_members; i++) {
10280 dl = serial_to_dl(inf[i].serial, super);
10281 if (!dl) {
10282 /* hmm maybe it failed?, nothing we can do about
10283 * it here
10284 */
10285 continue;
10286 }
10287 if (count_memberships(dl, super) == 0)
10288 activate++;
10289 }
10290 len += activate * sizeof(struct imsm_disk);
949c47a0 10291 break;
095b8088
N		 10292 }
10293 case update_kill_array: {
10294 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10295 return 0;
949c47a0
DW		 10296 break;
10297 }
095b8088
N		 10298 case update_rename_array: {
10299 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10300 return 0;
10301 break;
10302 }
10303 case update_add_remove_disk:
10304 /* no update->len needed */
10305 break;
bbab0940
TM		 10306 case update_prealloc_badblocks_mem:
10307 super->extra_space += sizeof(struct bbm_log) -
10308 get_imsm_bbm_log_size(super->bbm_log);
10309 break;
e6e9dd3f
AP		 10310 case update_rwh_policy: {
10311 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10312 return 0;
10313 break;
10314 }
095b8088
N		 10315 default:
10316 return 0;
949c47a0 10317 }
8273f55e 10318
4d7b1503
DW		 10319 /* check if we need a larger metadata buffer */
10320 if (super->next_buf)
10321 buf_len = super->next_len;
10322 else
10323 buf_len = super->len;
10324
bbab0940 10325 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10326 /* ok we need a larger buf than what is currently allocated
10327 * if this allocation fails process_update will notice that
10328 * ->next_len is set and ->next_buf is NULL
10329 */
bbab0940
TM		 10330 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10331 super->extra_space + len, sector_size);
4d7b1503
DW		 10332 if (super->next_buf)
10333 free(super->next_buf);
10334
10335 super->next_len = buf_len;
f36a9ecd 10336 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10337 memset(super->next_buf, 0, buf_len);
10338 else
4d7b1503
DW		 10339 super->next_buf = NULL;
10340 }
5fe6f031 10341 return 1;
8273f55e
DW		 10342}
10343
ae6aad82 10344/* must be called while manager is quiesced */
f21e18ca 10345static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10346{
10347 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10348 struct dl *iter;
10349 struct imsm_dev *dev;
10350 struct imsm_map *map;
4c9e8c1e 10351 unsigned int i, j, num_members;
fb12a745 10352 __u32 ord, ord_map0;
4c9e8c1e 10353 struct bbm_log *log = super->bbm_log;
ae6aad82 10354
1ade5cc1 10355 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10356
10357 /* shift all indexes down one */
10358 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10359 if (iter->index > (int)index)
ae6aad82 10360 iter->index--;
47ee5a45 10361 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10362 if (iter->index > (int)index)
47ee5a45 10363 iter->index--;
ae6aad82
DW		 10364
10365 for (i = 0; i < mpb->num_raid_devs; i++) {
10366 dev = get_imsm_dev(super, i);
238c0a71 10367 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10368 num_members = map->num_members;
10369 for (j = 0; j < num_members; j++) {
10370 /* update ord entries being careful not to propagate
10371 * ord-flags to the first map
10372 */
238c0a71 10373 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10374 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10375
24565c9a
DW		 10376 if (ord_to_idx(ord) <= index)
10377 continue;
ae6aad82 10378
238c0a71 10379 map = get_imsm_map(dev, MAP_0);
fb12a745 10380 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10381 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10382 if (map)
10383 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10384 }
10385 }
10386
4c9e8c1e
TM		 10387 for (i = 0; i < log->entry_count; i++) {
10388 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10389
10390 if (entry->disk_ordinal <= index)
10391 continue;
10392 entry->disk_ordinal--;
10393 }
10394
ae6aad82
DW		 10395 mpb->num_disks--;
10396 super->updates_pending++;
24565c9a
DW		 10397 if (*dlp) {
10398 struct dl *dl = *dlp;
10399
10400 *dlp = (*dlp)->next;
10401 __free_imsm_disk(dl);
10402 }
ae6aad82 10403}
9a717282
AK		 10404
10405static void close_targets(int *targets, int new_disks)
10406{
10407 int i;
10408
10409 if (!targets)
10410 return;
10411
10412 for (i = 0; i < new_disks; i++) {
10413 if (targets[i] >= 0) {
10414 close(targets[i]);
10415 targets[i] = -1;
10416 }
10417 }
10418}
10419
10420static int imsm_get_allowed_degradation(int level, int raid_disks,
10421 struct intel_super *super,
10422 struct imsm_dev *dev)
10423{
10424 switch (level) {
bf5cf7c7 10425 case 1:
9a717282
AK		 10426 case 10:{
10427 int ret_val = 0;
10428 struct imsm_map *map;
10429 int i;
10430
10431 ret_val = raid_disks/2;
10432 /* check map if all disks pairs not failed
10433 * in both maps
10434 */
238c0a71 10435 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10436 for (i = 0; i < ret_val; i++) {
10437 int degradation = 0;
10438 if (get_imsm_disk(super, i) == NULL)
10439 degradation++;
10440 if (get_imsm_disk(super, i + 1) == NULL)
10441 degradation++;
10442 if (degradation == 2)
10443 return 0;
10444 }
238c0a71 10445 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10446 /* if there is no second map
10447 * result can be returned
10448 */
10449 if (map == NULL)
10450 return ret_val;
10451 /* check degradation in second map
10452 */
10453 for (i = 0; i < ret_val; i++) {
10454 int degradation = 0;
10455 if (get_imsm_disk(super, i) == NULL)
10456 degradation++;
10457 if (get_imsm_disk(super, i + 1) == NULL)
10458 degradation++;
10459 if (degradation == 2)
10460 return 0;
10461 }
10462 return ret_val;
10463 }
10464 case 5:
10465 return 1;
10466 case 6:
10467 return 2;
10468 default:
10469 return 0;
10470 }
10471}
10472
687629c2
AK		 10473/*******************************************************************************
10474 * Function: open_backup_targets
10475 * Description: Function opens file descriptors for all devices given in
10476 * info->devs
10477 * Parameters:
10478 * info : general array info
10479 * raid_disks : number of disks
10480 * raid_fds : table of device's file descriptors
9a717282
AK		 10481 * super : intel super for raid10 degradation check
10482 * dev : intel device for raid10 degradation check
687629c2
AK		 10483 * Returns:
10484 * 0 : success
10485 * -1 : fail
10486 ******************************************************************************/
9a717282
AK		 10487int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10488 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10489{
10490 struct mdinfo *sd;
f627f5ad 10491 int i;
9a717282 10492 int opened = 0;
f627f5ad
AK		 10493
10494 for (i = 0; i < raid_disks; i++)
10495 raid_fds[i] = -1;
687629c2
AK		 10496
10497 for (sd = info->devs ; sd ; sd = sd->next) {
10498 char *dn;
10499
10500 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10501 dprintf("disk is faulty!!\n");
10502 continue;
10503 }
10504
089f9d79 10505 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10506 continue;
10507
10508 dn = map_dev(sd->disk.major,
10509 sd->disk.minor, 1);
10510 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10511 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10512 pr_err("cannot open component\n");
9a717282 10513 continue;
687629c2 10514 }
9a717282
AK		 10515 opened++;
10516 }
10517 /* check if maximum array degradation level is not exceeded
10518 */
10519 if ((raid_disks - opened) >
089f9d79
JS		 10520 imsm_get_allowed_degradation(info->new_level, raid_disks,
10521 super, dev)) {
e12b3daa 10522 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10523 close_targets(raid_fds, raid_disks);
10524 return -2;
687629c2
AK		 10525 }
10526 return 0;
10527}
10528
d31ad643
PB		 10529/*******************************************************************************
10530 * Function: validate_container_imsm
10531 * Description: This routine validates container after assemble,
10532 * eg. if devices in container are under the same controller.
10533 *
10534 * Parameters:
10535 * info : linked list with info about devices used in array
10536 * Returns:
10537 * 1 : HBA mismatch
10538 * 0 : Success
10539 ******************************************************************************/
10540int validate_container_imsm(struct mdinfo *info)
10541{
6b781d33
AP		 10542 if (check_env("IMSM_NO_PLATFORM"))
10543 return 0;
d31ad643 10544
6b781d33
AP		 10545 struct sys_dev *idev;
10546 struct sys_dev *hba = NULL;
10547 struct sys_dev *intel_devices = find_intel_devices();
10548 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10549 info->disk.minor));
10550
10551 for (idev = intel_devices; idev; idev = idev->next) {
10552 if (dev_path && strstr(dev_path, idev->path)) {
10553 hba = idev;
10554 break;
d31ad643 10555 }
6b781d33
AP		 10556 }
10557 if (dev_path)
d31ad643
PB		 10558 free(dev_path);
10559
6b781d33
AP		 10560 if (!hba) {
10561 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10562 devid2kname(makedev(info->disk.major, info->disk.minor)));
10563 return 1;
10564 }
10565
10566 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10567 struct mdinfo *dev;
10568
10569 for (dev = info->next; dev; dev = dev->next) {
10570 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10571
10572 struct sys_dev *hba2 = NULL;
10573 for (idev = intel_devices; idev; idev = idev->next) {
10574 if (dev_path && strstr(dev_path, idev->path)) {
10575 hba2 = idev;
10576 break;
d31ad643
PB		 10577 }
10578 }
6b781d33
AP		 10579 if (dev_path)
10580 free(dev_path);
10581
10582 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10583 get_orom_by_device_id(hba2->dev_id);
10584
10585 if (hba2 && hba->type != hba2->type) {
10586 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10587 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10588 return 1;
10589 }
10590
07cb1e57 10591 if (orom != orom2) {
6b781d33
AP		 10592 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10593 " This operation is not supported and can lead to data loss.\n");
10594 return 1;
10595 }
10596
10597 if (!orom) {
10598 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10599 " This operation is not supported and can lead to data loss.\n");
10600 return 1;
10601 }
d31ad643 10602 }
6b781d33 10603
d31ad643
PB		 10604 return 0;
10605}
32141c17 10606
6f50473f
TM		 10607/*******************************************************************************
10608* Function: imsm_record_badblock
10609* Description: This routine stores new bad block record in BBM log
10610*
10611* Parameters:
10612* a : array containing a bad block
10613* slot : disk number containing a bad block
10614* sector : bad block sector
10615* length : bad block sectors range
10616* Returns:
10617* 1 : Success
10618* 0 : Error
10619******************************************************************************/
10620static int imsm_record_badblock(struct active_array *a, int slot,
10621 unsigned long long sector, int length)
10622{
10623 struct intel_super *super = a->container->sb;
10624 int ord;
10625 int ret;
10626
10627 ord = imsm_disk_slot_to_ord(a, slot);
10628 if (ord < 0)
10629 return 0;
10630
10631 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10632 length);
10633 if (ret)
10634 super->updates_pending++;
10635
10636 return ret;
10637}
c07a5a4f
TM		 10638/*******************************************************************************
10639* Function: imsm_clear_badblock
10640* Description: This routine clears bad block record from BBM log
10641*
10642* Parameters:
10643* a : array containing a bad block
10644* slot : disk number containing a bad block
10645* sector : bad block sector
10646* length : bad block sectors range
10647* Returns:
10648* 1 : Success
10649* 0 : Error
10650******************************************************************************/
10651static int imsm_clear_badblock(struct active_array *a, int slot,
10652 unsigned long long sector, int length)
10653{
10654 struct intel_super *super = a->container->sb;
10655 int ord;
10656 int ret;
10657
10658 ord = imsm_disk_slot_to_ord(a, slot);
10659 if (ord < 0)
10660 return 0;
10661
10662 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10663 if (ret)
10664 super->updates_pending++;
10665
10666 return ret;
10667}
928f1424
TM		 10668/*******************************************************************************
10669* Function: imsm_get_badblocks
10670* Description: This routine get list of bad blocks for an array
10671*
10672* Parameters:
10673* a : array
10674* slot : disk number
10675* Returns:
10676* bb : structure containing bad blocks
10677* NULL : error
10678******************************************************************************/
10679static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10680{
10681 int inst = a->info.container_member;
10682 struct intel_super *super = a->container->sb;
10683 struct imsm_dev *dev = get_imsm_dev(super, inst);
10684 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10685 int ord;
10686
10687 ord = imsm_disk_slot_to_ord(a, slot);
10688 if (ord < 0)
10689 return NULL;
10690
10691 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10692 per_dev_array_size(map), &super->bb);
928f1424
TM		 10693
10694 return &super->bb;
10695}
27156a57
TM		 10696/*******************************************************************************
10697* Function: examine_badblocks_imsm
10698* Description: Prints list of bad blocks on a disk to the standard output
10699*
10700* Parameters:
10701* st : metadata handler
10702* fd : open file descriptor for device
10703* devname : device name
10704* Returns:
10705* 0 : Success
10706* 1 : Error
10707******************************************************************************/
10708static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10709{
10710 struct intel_super *super = st->sb;
10711 struct bbm_log *log = super->bbm_log;
10712 struct dl *d = NULL;
10713 int any = 0;
10714
10715 for (d = super->disks; d ; d = d->next) {
10716 if (strcmp(d->devname, devname) == 0)
10717 break;
10718 }
10719
10720 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10721 pr_err("%s doesn't appear to be part of a raid array\n",
10722 devname);
10723 return 1;
10724 }
10725
10726 if (log != NULL) {
10727 unsigned int i;
10728 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10729
10730 for (i = 0; i < log->entry_count; i++) {
10731 if (entry[i].disk_ordinal == d->index) {
10732 unsigned long long sector = __le48_to_cpu(
10733 &entry[i].defective_block_start);
10734 int cnt = entry[i].marked_count + 1;
10735
10736 if (!any) {
10737 printf("Bad-blocks on %s:\n", devname);
10738 any = 1;
10739 }
10740
10741 printf("%20llu for %d sectors\n", sector, cnt);
10742 }
10743 }
10744 }
10745
10746 if (!any)
10747 printf("No bad-blocks list configured on %s\n", devname);
10748
10749 return 0;
10750}
687629c2
AK		 10751/*******************************************************************************
10752 * Function: init_migr_record_imsm
10753 * Description: Function inits imsm migration record
10754 * Parameters:
10755 * super : imsm internal array info
10756 * dev : device under migration
10757 * info : general array info to find the smallest device
10758 * Returns:
10759 * none
10760 ******************************************************************************/
10761void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10762 struct mdinfo *info)
10763{
10764 struct intel_super *super = st->sb;
10765 struct migr_record *migr_rec = super->migr_rec;
10766 int new_data_disks;
10767 unsigned long long dsize, dev_sectors;
10768 long long unsigned min_dev_sectors = -1LLU;
10769 struct mdinfo *sd;
10770 char nm[30];
10771 int fd;
238c0a71
AK		 10772 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10773 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10774 unsigned long long num_migr_units;
3ef4403c 10775 unsigned long long array_blocks;
687629c2
AK		 10776
10777 memset(migr_rec, 0, sizeof(struct migr_record));
10778 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10779
10780 /* only ascending reshape supported now */
10781 migr_rec->ascending_migr = __cpu_to_le32(1);
10782
10783 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10784 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10785 migr_rec->dest_depth_per_unit *=
10786 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10787 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10788 migr_rec->blocks_per_unit =
10789 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10790 migr_rec->dest_depth_per_unit =
10791 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10792 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10793 num_migr_units =
10794 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10795
10796 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10797 num_migr_units++;
9f421827 10798 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10799
10800 migr_rec->post_migr_vol_cap = dev->size_low;
10801 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10802
687629c2
AK		 10803 /* Find the smallest dev */
10804 for (sd = info->devs ; sd ; sd = sd->next) {
10805 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10806 fd = dev_open(nm, O_RDONLY);
10807 if (fd < 0)
10808 continue;
10809 get_dev_size(fd, NULL, &dsize);
10810 dev_sectors = dsize / 512;
10811 if (dev_sectors < min_dev_sectors)
10812 min_dev_sectors = dev_sectors;
10813 close(fd);
10814 }
9f421827 10815 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10816 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10817
10818 write_imsm_migr_rec(st);
10819
10820 return;
10821}
10822
10823/*******************************************************************************
10824 * Function: save_backup_imsm
10825 * Description: Function saves critical data stripes to Migration Copy Area
10826 * and updates the current migration unit status.
10827 * Use restore_stripes() to form a destination stripe,
10828 * and to write it to the Copy Area.
10829 * Parameters:
10830 * st : supertype information
aea93171 10831 * dev : imsm device that backup is saved for
687629c2
AK		 10832 * info : general array info
10833 * buf : input buffer
687629c2
AK		 10834 * length : length of data to backup (blocks_per_unit)
10835 * Returns:
10836 * 0 : success
10837 *, -1 : fail
10838 ******************************************************************************/
10839int save_backup_imsm(struct supertype *st,
10840 struct imsm_dev *dev,
10841 struct mdinfo *info,
10842 void *buf,
687629c2
AK		 10843 int length)
10844{
10845 int rv = -1;
10846 struct intel_super *super = st->sb;
594dc1b8
JS		 10847 unsigned long long *target_offsets;
10848 int *targets;
687629c2 10849 int i;
238c0a71 10850 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10851 int new_disks = map_dest->num_members;
ab724b98
AK		 10852 int dest_layout = 0;
10853 int dest_chunk;
d1877f69 10854 unsigned long long start;
9529d343 10855 int data_disks = imsm_num_data_members(map_dest);
687629c2 10856
503975b9 10857 targets = xmalloc(new_disks * sizeof(int));
687629c2 10858
7e45b550
AK		 10859 for (i = 0; i < new_disks; i++)
10860 targets[i] = -1;
10861
503975b9 10862 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10863
d1877f69 10864 start = info->reshape_progress * 512;
687629c2 10865 for (i = 0; i < new_disks; i++) {
9f421827 10866 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10867 /* move back copy area adderss, it will be moved forward
10868 * in restore_stripes() using start input variable
10869 */
10870 target_offsets[i] -= start/data_disks;
687629c2
AK		 10871 }
10872
9a717282
AK		 10873 if (open_backup_targets(info, new_disks, targets,
10874 super, dev))
687629c2
AK		 10875 goto abort;
10876
68eb8bc6 10877 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10878 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10879
687629c2
AK		 10880 if (restore_stripes(targets, /* list of dest devices */
10881 target_offsets, /* migration record offsets */
10882 new_disks,
ab724b98
AK		 10883 dest_chunk,
10884 map_dest->raid_level,
10885 dest_layout,
10886 -1, /* source backup file descriptor */
10887 0, /* input buf offset
10888 * always 0 buf is already offseted */
d1877f69 10889 start,
687629c2
AK		 10890 length,
10891 buf) != 0) {
e7b84f9d 10892 pr_err("Error restoring stripes\n");
687629c2
AK		 10893 goto abort;
10894 }
10895
10896 rv = 0;
10897
10898abort:
10899 if (targets) {
9a717282 10900 close_targets(targets, new_disks);
687629c2
AK		 10901 free(targets);
10902 }
10903 free(target_offsets);
10904
10905 return rv;
10906}
10907
10908/*******************************************************************************
10909 * Function: save_checkpoint_imsm
10910 * Description: Function called for current unit status update
10911 * in the migration record. It writes it to disk.
10912 * Parameters:
10913 * super : imsm internal array info
10914 * info : general array info
10915 * Returns:
10916 * 0: success
10917 * 1: failure
0228d92c
AK		 10918 * 2: failure, means no valid migration record
10919 * / no general migration in progress /
687629c2
AK		 10920 ******************************************************************************/
10921int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10922{
10923 struct intel_super *super = st->sb;
f8b72ef5
AK		 10924 unsigned long long blocks_per_unit;
10925 unsigned long long curr_migr_unit;
10926
2e062e82 10927 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10928 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10929 return 1;
10930 }
10931
f8b72ef5
AK		 10932 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10933 if (blocks_per_unit == 0) {
0228d92c
AK		 10934 dprintf("imsm: no migration in progress.\n");
10935 return 2;
687629c2 10936 }
f8b72ef5
AK		 10937 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10938 /* check if array is alligned to copy area
10939 * if it is not alligned, add one to current migration unit value
10940 * this can happend on array reshape finish only
10941 */
10942 if (info->reshape_progress % blocks_per_unit)
10943 curr_migr_unit++;
687629c2 10944
9f421827 10945 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10946 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10947 set_migr_dest_1st_member_lba(super->migr_rec,
10948 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10949
687629c2 10950 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10951 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10952 return 1;
10953 }
10954
10955 return 0;
10956}
10957
276d77db
AK		 10958/*******************************************************************************
10959 * Function: recover_backup_imsm
10960 * Description: Function recovers critical data from the Migration Copy Area
10961 * while assembling an array.
10962 * Parameters:
10963 * super : imsm internal array info
10964 * info : general array info
10965 * Returns:
10966 * 0 : success (or there is no data to recover)
10967 * 1 : fail
10968 ******************************************************************************/
10969int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10970{
10971 struct intel_super *super = st->sb;
10972 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10973 struct imsm_map *map_dest;
276d77db
AK		 10974 struct intel_dev *id = NULL;
10975 unsigned long long read_offset;
10976 unsigned long long write_offset;
10977 unsigned unit_len;
10978 int *targets = NULL;
10979 int new_disks, i, err;
10980 char *buf = NULL;
10981 int retval = 1;
f36a9ecd 10982 unsigned int sector_size = super->sector_size;
9f421827
PB		 10983 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10984 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10985 char buffer[20];
6c3560c0 10986 int skipped_disks = 0;
276d77db
AK		 10987
10988 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10989 if (err < 1)
10990 return 1;
10991
10992 /* recover data only during assemblation */
10993 if (strncmp(buffer, "inactive", 8) != 0)
10994 return 0;
10995 /* no data to recover */
10996 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10997 return 0;
10998 if (curr_migr_unit >= num_migr_units)
10999 return 1;
11000
11001 /* find device during reshape */
11002 for (id = super->devlist; id; id = id->next)
11003 if (is_gen_migration(id->dev))
11004 break;
11005 if (id == NULL)
11006 return 1;
11007
238c0a71 11008 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 11009 new_disks = map_dest->num_members;
11010
9f421827 11011 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 11012
9f421827 11013 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 11014 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 11015
11016 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 11017 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 11018 goto abort;
503975b9 11019 targets = xcalloc(new_disks, sizeof(int));
276d77db 11020
9a717282 11021 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 11022 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 11023 goto abort;
11024 }
276d77db
AK		 11025
11026 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 11027 if (targets[i] < 0) {
11028 skipped_disks++;
11029 continue;
11030 }
276d77db 11031 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11032 pr_err("Cannot seek to block: %s\n",
11033 strerror(errno));
137debce
AK		 11034 skipped_disks++;
11035 continue;
276d77db 11036 }
9ec11d1a 11037 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11038 pr_err("Cannot read copy area block: %s\n",
11039 strerror(errno));
137debce
AK		 11040 skipped_disks++;
11041 continue;
276d77db
AK		 11042 }
11043 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11044 pr_err("Cannot seek to block: %s\n",
11045 strerror(errno));
137debce
AK		 11046 skipped_disks++;
11047 continue;
276d77db 11048 }
9ec11d1a 11049 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11050 pr_err("Cannot restore block: %s\n",
11051 strerror(errno));
137debce
AK		 11052 skipped_disks++;
11053 continue;
276d77db
AK		 11054 }
11055 }
11056
137debce
AK		 11057 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11058 new_disks,
11059 super,
11060 id->dev)) {
7a862a02 11061 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11062 goto abort;
11063 }
11064
befb629b
AK		 11065 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11066 /* ignore error == 2, this can mean end of reshape here
11067 */
7a862a02 11068 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11069 } else
276d77db 11070 retval = 0;
276d77db
AK		 11071
11072abort:
11073 if (targets) {
11074 for (i = 0; i < new_disks; i++)
11075 if (targets[i])
11076 close(targets[i]);
11077 free(targets);
11078 }
11079 free(buf);
11080 return retval;
11081}
11082
2cda7640
ML		 11083static char disk_by_path[] = "/dev/disk/by-path/";
11084
11085static const char *imsm_get_disk_controller_domain(const char *path)
11086{
2cda7640 11087 char disk_path[PATH_MAX];
96234762
LM		 11088 char *drv=NULL;
11089 struct stat st;
2cda7640 11090
6d8d290a 11091 strcpy(disk_path, disk_by_path);
96234762
LM		 11092 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11093 if (stat(disk_path, &st) == 0) {
11094 struct sys_dev* hba;
594dc1b8 11095 char *path;
96234762
LM		 11096
11097 path = devt_to_devpath(st.st_rdev);
11098 if (path == NULL)
11099 return "unknown";
11100 hba = find_disk_attached_hba(-1, path);
11101 if (hba && hba->type == SYS_DEV_SAS)
11102 drv = "isci";
11103 else if (hba && hba->type == SYS_DEV_SATA)
11104 drv = "ahci";
c6839718
MT		 11105 else if (hba && hba->type == SYS_DEV_VMD)
11106 drv = "vmd";
11107 else if (hba && hba->type == SYS_DEV_NVME)
11108 drv = "nvme";
1011e834 11109 else
96234762
LM		 11110 drv = "unknown";
11111 dprintf("path: %s hba: %s attached: %s\n",
11112 path, (hba) ? hba->path : "NULL", drv);
11113 free(path);
2cda7640 11114 }
96234762 11115 return drv;
2cda7640
ML		 11116}
11117
4dd2df09 11118static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11119{
4dd2df09 11120 static char devnm[32];
78b10e66
N		 11121 char subdev_name[20];
11122 struct mdstat_ent *mdstat;
11123
11124 sprintf(subdev_name, "%d", subdev);
11125 mdstat = mdstat_by_subdev(subdev_name, container);
11126 if (!mdstat)
4dd2df09 11127 return NULL;
78b10e66 11128
4dd2df09 11129 strcpy(devnm, mdstat->devnm);
78b10e66 11130 free_mdstat(mdstat);
4dd2df09 11131 return devnm;
78b10e66
N		 11132}
11133
11134static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11135 struct geo_params *geo,
fbf3d202
AK		 11136 int *old_raid_disks,
11137 int direction)
78b10e66 11138{
694575e7
KW		 11139 /* currently we only support increasing the number of devices
11140 * for a container. This increases the number of device for each
11141 * member array. They must all be RAID0 or RAID5.
11142 */
78b10e66
N		 11143 int ret_val = 0;
11144 struct mdinfo *info, *member;
11145 int devices_that_can_grow = 0;
11146
7a862a02 11147 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11148
d04f65f4 11149 if (geo->size > 0 ||
78b10e66
N		 11150 geo->level != UnSet ||
11151 geo->layout != UnSet ||
11152 geo->chunksize != 0 ||
11153 geo->raid_disks == UnSet) {
7a862a02 11154 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11155 return ret_val;
11156 }
11157
fbf3d202 11158 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11159 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11160 return ret_val;
11161 }
11162
78b10e66
N		 11163 info = container_content_imsm(st, NULL);
11164 for (member = info; member; member = member->next) {
4dd2df09 11165 char *result;
78b10e66
N		 11166
11167 dprintf("imsm: checking device_num: %i\n",
11168 member->container_member);
11169
d7d205bd 11170 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11171 /* we work on container for Online Capacity Expansion
11172 * only so raid_disks has to grow
11173 */
7a862a02 11174 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11175 break;
11176 }
11177
089f9d79 11178 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11179 /* we cannot use this container with other raid level
11180 */
7a862a02 11181 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11182 info->array.level);
11183 break;
11184 } else {
11185 /* check for platform support
11186 * for this raid level configuration
11187 */
11188 struct intel_super *super = st->sb;
11189 if (!is_raid_level_supported(super->orom,
11190 member->array.level,
11191 geo->raid_disks)) {
7a862a02 11192 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11193 info->array.level,
11194 geo->raid_disks,
11195 geo->raid_disks > 1 ? "s" : "");
11196 break;
11197 }
2a4a08e7
AK		 11198 /* check if component size is aligned to chunk size
11199 */
11200 if (info->component_size %
11201 (info->array.chunk_size/512)) {
7a862a02 11202 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11203 break;
11204 }
78b10e66
N		 11205 }
11206
11207 if (*old_raid_disks &&
11208 info->array.raid_disks != *old_raid_disks)
11209 break;
11210 *old_raid_disks = info->array.raid_disks;
11211
11212 /* All raid5 and raid0 volumes in container
11213 * have to be ready for Online Capacity Expansion
11214 * so they need to be assembled. We have already
11215 * checked that no recovery etc is happening.
11216 */
4dd2df09
N		 11217 result = imsm_find_array_devnm_by_subdev(member->container_member,
11218 st->container_devnm);
11219 if (result == NULL) {
78b10e66
N		 11220 dprintf("imsm: cannot find array\n");
11221 break;
11222 }
11223 devices_that_can_grow++;
11224 }
11225 sysfs_free(info);
11226 if (!member && devices_that_can_grow)
11227 ret_val = 1;
11228
11229 if (ret_val)
1ade5cc1 11230 dprintf("Container operation allowed\n");
78b10e66 11231 else
1ade5cc1 11232 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11233
11234 return ret_val;
11235}
11236
11237/* Function: get_spares_for_grow
11238 * Description: Allocates memory and creates list of spare devices
1011e834 11239 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11240 * Parameters: Pointer to the supertype structure
11241 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11242 * NULL if fail
78b10e66
N		 11243 */
11244static struct mdinfo *get_spares_for_grow(struct supertype *st)
11245{
fbfdcb06
AO		 11246 struct spare_criteria sc;
11247
11248 get_spare_criteria_imsm(st, &sc);
11249 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11250}
11251
11252/******************************************************************************
11253 * function: imsm_create_metadata_update_for_reshape
11254 * Function creates update for whole IMSM container.
11255 *
11256 ******************************************************************************/
11257static int imsm_create_metadata_update_for_reshape(
11258 struct supertype *st,
11259 struct geo_params *geo,
11260 int old_raid_disks,
11261 struct imsm_update_reshape **updatep)
11262{
11263 struct intel_super *super = st->sb;
11264 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11265 int update_memory_size;
11266 struct imsm_update_reshape *u;
11267 struct mdinfo *spares;
78b10e66 11268 int i;
594dc1b8 11269 int delta_disks;
bbd24d86 11270 struct mdinfo *dev;
78b10e66 11271
1ade5cc1 11272 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11273
11274 delta_disks = geo->raid_disks - old_raid_disks;
11275
11276 /* size of all update data without anchor */
11277 update_memory_size = sizeof(struct imsm_update_reshape);
11278
11279 /* now add space for spare disks that we need to add. */
11280 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11281
503975b9 11282 u = xcalloc(1, update_memory_size);
78b10e66
N		 11283 u->type = update_reshape_container_disks;
11284 u->old_raid_disks = old_raid_disks;
11285 u->new_raid_disks = geo->raid_disks;
11286
11287 /* now get spare disks list
11288 */
11289 spares = get_spares_for_grow(st);
11290
d7be7d87 11291 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11292 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11293 i = -1;
78b10e66
N		 11294 goto abort;
11295 }
11296
11297 /* we have got spares
11298 * update disk list in imsm_disk list table in anchor
11299 */
11300 dprintf("imsm: %i spares are available.\n\n",
11301 spares->array.spare_disks);
11302
bbd24d86 11303 dev = spares->devs;
78b10e66 11304 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11305 struct dl *dl;
11306
bbd24d86
AK		 11307 if (dev == NULL)
11308 break;
78b10e66
N		 11309 u->new_disks[i] = makedev(dev->disk.major,
11310 dev->disk.minor);
11311 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11312 dl->index = mpb->num_disks;
11313 mpb->num_disks++;
bbd24d86 11314 dev = dev->next;
78b10e66 11315 }
78b10e66
N		 11316
11317abort:
11318 /* free spares
11319 */
11320 sysfs_free(spares);
11321
d677e0b8 11322 dprintf("imsm: reshape update preparation :");
78b10e66 11323 if (i == delta_disks) {
1ade5cc1 11324 dprintf_cont(" OK\n");
78b10e66
N		 11325 *updatep = u;
11326 return update_memory_size;
11327 }
11328 free(u);
1ade5cc1 11329 dprintf_cont(" Error\n");
78b10e66
N		 11330
11331 return 0;
11332}
11333
f3871fdc
AK		 11334/******************************************************************************
11335 * function: imsm_create_metadata_update_for_size_change()
11336 * Creates update for IMSM array for array size change.
11337 *
11338 ******************************************************************************/
11339static int imsm_create_metadata_update_for_size_change(
11340 struct supertype *st,
11341 struct geo_params *geo,
11342 struct imsm_update_size_change **updatep)
11343{
11344 struct intel_super *super = st->sb;
594dc1b8
JS		 11345 int update_memory_size;
11346 struct imsm_update_size_change *u;
f3871fdc 11347
1ade5cc1 11348 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11349
11350 /* size of all update data without anchor */
11351 update_memory_size = sizeof(struct imsm_update_size_change);
11352
503975b9 11353 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11354 u->type = update_size_change;
11355 u->subdev = super->current_vol;
11356 u->new_size = geo->size;
11357
11358 dprintf("imsm: reshape update preparation : OK\n");
11359 *updatep = u;
11360
11361 return update_memory_size;
11362}
11363
48c5303a
PC		 11364/******************************************************************************
11365 * function: imsm_create_metadata_update_for_migration()
11366 * Creates update for IMSM array.
11367 *
11368 ******************************************************************************/
11369static int imsm_create_metadata_update_for_migration(
11370 struct supertype *st,
11371 struct geo_params *geo,
11372 struct imsm_update_reshape_migration **updatep)
11373{
11374 struct intel_super *super = st->sb;
594dc1b8
JS		 11375 int update_memory_size;
11376 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11377 struct imsm_dev *dev;
11378 int previous_level = -1;
11379
1ade5cc1 11380 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11381
11382 /* size of all update data without anchor */
11383 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11384
503975b9 11385 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11386 u->type = update_reshape_migration;
11387 u->subdev = super->current_vol;
11388 u->new_level = geo->level;
11389 u->new_layout = geo->layout;
11390 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11391 u->new_disks[0] = -1;
4bba0439 11392 u->new_chunksize = -1;
48c5303a
PC		 11393
11394 dev = get_imsm_dev(super, u->subdev);
11395 if (dev) {
11396 struct imsm_map *map;
11397
238c0a71 11398 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11399 if (map) {
11400 int current_chunk_size =
11401 __le16_to_cpu(map->blocks_per_strip) / 2;
11402
11403 if (geo->chunksize != current_chunk_size) {
11404 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11405 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11406 current_chunk_size, u->new_chunksize);
11407 }
48c5303a 11408 previous_level = map->raid_level;
4bba0439 11409 }
48c5303a 11410 }
089f9d79 11411 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11412 struct mdinfo *spares = NULL;
11413
11414 u->new_raid_disks++;
11415 spares = get_spares_for_grow(st);
089f9d79 11416 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11417 free(u);
11418 sysfs_free(spares);
11419 update_memory_size = 0;
565cc99e 11420 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11421 return 0;
11422 }
11423 sysfs_free(spares);
11424 }
11425 dprintf("imsm: reshape update preparation : OK\n");
11426 *updatep = u;
11427
11428 return update_memory_size;
11429}
11430
8dd70bce
AK		 11431static void imsm_update_metadata_locally(struct supertype *st,
11432 void *buf, int len)
11433{
11434 struct metadata_update mu;
11435
11436 mu.buf = buf;
11437 mu.len = len;
11438 mu.space = NULL;
11439 mu.space_list = NULL;
11440 mu.next = NULL;
5fe6f031
N		 11441 if (imsm_prepare_update(st, &mu))
11442 imsm_process_update(st, &mu);
8dd70bce
AK		 11443
11444 while (mu.space_list) {
11445 void **space = mu.space_list;
11446 mu.space_list = *space;
11447 free(space);
11448 }
11449}
78b10e66 11450
471bceb6 11451/***************************************************************************
694575e7 11452* Function: imsm_analyze_change
471bceb6 11453* Description: Function analyze change for single volume
1011e834 11454* and validate if transition is supported
fbf3d202
AK		 11455* Parameters: Geometry parameters, supertype structure,
11456* metadata change direction (apply/rollback)
694575e7 11457* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11458****************************************************************************/
11459enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11460 struct geo_params *geo,
11461 int direction)
694575e7 11462{
471bceb6
KW		 11463 struct mdinfo info;
11464 int change = -1;
11465 int check_devs = 0;
c21e737b 11466 int chunk;
67a2db32
AK		 11467 /* number of added/removed disks in operation result */
11468 int devNumChange = 0;
11469 /* imsm compatible layout value for array geometry verification */
11470 int imsm_layout = -1;
7abc9871
AK		 11471 int data_disks;
11472 struct imsm_dev *dev;
9529d343 11473 struct imsm_map *map;
7abc9871 11474 struct intel_super *super;
d04f65f4 11475 unsigned long long current_size;
65d38cca 11476 unsigned long long free_size;
d04f65f4 11477 unsigned long long max_size;
65d38cca 11478 int rv;
471bceb6
KW		 11479
11480 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11481 if (geo->level != info.array.level && geo->level >= 0 &&
11482 geo->level != UnSet) {
471bceb6
KW		 11483 switch (info.array.level) {
11484 case 0:
11485 if (geo->level == 5) {
b5347799 11486 change = CH_MIGRATION;
e13ce846 11487 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11488 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11489 change = -1;
11490 goto analyse_change_exit;
11491 }
67a2db32 11492 imsm_layout = geo->layout;
471bceb6 11493 check_devs = 1;
e91a3bad
LM		 11494 devNumChange = 1; /* parity disk added */
11495 } else if (geo->level == 10) {
471bceb6
KW		 11496 change = CH_TAKEOVER;
11497 check_devs = 1;
e91a3bad 11498 devNumChange = 2; /* two mirrors added */
67a2db32 11499 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11500 }
dfe77a9e
KW		 11501 break;
11502 case 1:
471bceb6
KW		 11503 case 10:
11504 if (geo->level == 0) {
11505 change = CH_TAKEOVER;
11506 check_devs = 1;
e91a3bad 11507 devNumChange = -(geo->raid_disks/2);
67a2db32 11508 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11509 }
11510 break;
11511 }
11512 if (change == -1) {
7a862a02 11513 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11514 info.array.level, geo->level);
471bceb6
KW		 11515 goto analyse_change_exit;
11516 }
11517 } else
11518 geo->level = info.array.level;
11519
089f9d79
JS		 11520 if (geo->layout != info.array.layout &&
11521 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11522 change = CH_MIGRATION;
089f9d79
JS		 11523 if (info.array.layout == 0 && info.array.level == 5 &&
11524 geo->layout == 5) {
471bceb6 11525 /* reshape 5 -> 4 */
089f9d79
JS		 11526 } else if (info.array.layout == 5 && info.array.level == 5 &&
11527 geo->layout == 0) {
471bceb6
KW		 11528 /* reshape 4 -> 5 */
11529 geo->layout = 0;
11530 geo->level = 5;
11531 } else {
7a862a02 11532 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11533 info.array.layout, geo->layout);
471bceb6
KW		 11534 change = -1;
11535 goto analyse_change_exit;
11536 }
67a2db32 11537 } else {
471bceb6 11538 geo->layout = info.array.layout;
67a2db32
AK		 11539 if (imsm_layout == -1)
11540 imsm_layout = info.array.layout;
11541 }
471bceb6 11542
089f9d79
JS		 11543 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11544 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11545 if (info.array.level == 10) {
11546 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11547 change = -1;
11548 goto analyse_change_exit;
1e9b2c3f
PB		 11549 } else if (info.component_size % (geo->chunksize/512)) {
11550 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11551 geo->chunksize/1024, info.component_size/2);
11552 change = -1;
11553 goto analyse_change_exit;
2d2b0eb7 11554 }
b5347799 11555 change = CH_MIGRATION;
2d2b0eb7 11556 } else {
471bceb6 11557 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11558 }
471bceb6 11559
c21e737b 11560 chunk = geo->chunksize / 1024;
7abc9871
AK		 11561
11562 super = st->sb;
11563 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11564 map = get_imsm_map(dev, MAP_0);
11565 data_disks = imsm_num_data_members(map);
c41e00b2 11566 /* compute current size per disk member
7abc9871 11567 */
c41e00b2
AK		 11568 current_size = info.custom_array_size / data_disks;
11569
089f9d79 11570 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11571 /* align component size
11572 */
3e684231 11573 geo->size = imsm_component_size_alignment_check(
c41e00b2 11574 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11575 chunk * 1024, super->sector_size,
c41e00b2 11576 geo->size * 2);
65d0b4ce 11577 if (geo->size == 0) {
7a862a02 11578 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11579 current_size);
11580 goto analyse_change_exit;
11581 }
c41e00b2 11582 }
7abc9871 11583
089f9d79 11584 if (current_size != geo->size && geo->size > 0) {
7abc9871 11585 if (change != -1) {
7a862a02 11586 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11587 change = -1;
11588 goto analyse_change_exit;
11589 }
11590 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11591 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11592 super->current_vol, st->devnm);
7abc9871
AK		 11593 goto analyse_change_exit;
11594 }
65d38cca
LD		 11595 /* check the maximum available size
11596 */
11597 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11598 0, chunk, &free_size);
11599 if (rv == 0)
11600 /* Cannot find maximum available space
11601 */
11602 max_size = 0;
11603 else {
11604 max_size = free_size + current_size;
11605 /* align component size
11606 */
3e684231 11607 max_size = imsm_component_size_alignment_check(
65d38cca 11608 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11609 chunk * 1024, super->sector_size,
65d38cca
LD		 11610 max_size);
11611 }
d04f65f4 11612 if (geo->size == MAX_SIZE) {
b130333f
AK		 11613 /* requested size change to the maximum available size
11614 */
65d38cca 11615 if (max_size == 0) {
7a862a02 11616 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11617 change = -1;
11618 goto analyse_change_exit;
65d38cca
LD		 11619 } else
11620 geo->size = max_size;
c41e00b2 11621 }
b130333f 11622
681b7ae2 11623 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11624 /* accept size for rollback only
11625 */
11626 } else {
11627 /* round size due to metadata compatibility
11628 */
11629 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11630 << SECT_PER_MB_SHIFT;
11631 dprintf("Prepare update for size change to %llu\n",
11632 geo->size );
11633 if (current_size >= geo->size) {
7a862a02 11634 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11635 current_size, geo->size);
fbf3d202
AK		 11636 goto analyse_change_exit;
11637 }
65d38cca 11638 if (max_size && geo->size > max_size) {
7a862a02 11639 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11640 max_size, geo->size);
65d38cca
LD		 11641 goto analyse_change_exit;
11642 }
7abc9871
AK		 11643 }
11644 geo->size *= data_disks;
11645 geo->raid_disks = dev->vol.map->num_members;
11646 change = CH_ARRAY_SIZE;
11647 }
471bceb6
KW		 11648 if (!validate_geometry_imsm(st,
11649 geo->level,
67a2db32 11650 imsm_layout,
e91a3bad 11651 geo->raid_disks + devNumChange,
c21e737b 11652 &chunk,
af4348dd 11653 geo->size, INVALID_SECTORS,
5308f117 11654 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11655 change = -1;
11656
11657 if (check_devs) {
11658 struct intel_super *super = st->sb;
11659 struct imsm_super *mpb = super->anchor;
11660
11661 if (mpb->num_raid_devs > 1) {
7a862a02 11662 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11663 geo->dev_name);
471bceb6
KW		 11664 change = -1;
11665 }
11666 }
11667
11668analyse_change_exit:
089f9d79
JS		 11669 if (direction == ROLLBACK_METADATA_CHANGES &&
11670 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11671 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11672 change = -1;
11673 }
471bceb6 11674 return change;
694575e7
KW		 11675}
11676
bb025c2f
KW		 11677int imsm_takeover(struct supertype *st, struct geo_params *geo)
11678{
11679 struct intel_super *super = st->sb;
11680 struct imsm_update_takeover *u;
11681
503975b9 11682 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11683
11684 u->type = update_takeover;
11685 u->subarray = super->current_vol;
11686
11687 /* 10->0 transition */
11688 if (geo->level == 0)
11689 u->direction = R10_TO_R0;
11690
0529c688
KW		 11691 /* 0->10 transition */
11692 if (geo->level == 10)
11693 u->direction = R0_TO_R10;
11694
bb025c2f
KW		 11695 /* update metadata locally */
11696 imsm_update_metadata_locally(st, u,
11697 sizeof(struct imsm_update_takeover));
11698 /* and possibly remotely */
11699 if (st->update_tail)
11700 append_metadata_update(st, u,
11701 sizeof(struct imsm_update_takeover));
11702 else
11703 free(u);
11704
11705 return 0;
11706}
11707
d04f65f4
N		 11708static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11709 int level,
78b10e66 11710 int layout, int chunksize, int raid_disks,
41784c88 11711 int delta_disks, char *backup, char *dev,
016e00f5 11712 int direction, int verbose)
78b10e66 11713{
78b10e66
N		 11714 int ret_val = 1;
11715 struct geo_params geo;
11716
1ade5cc1 11717 dprintf("(enter)\n");
78b10e66 11718
71204a50 11719 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11720
11721 geo.dev_name = dev;
4dd2df09 11722 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11723 geo.size = size;
11724 geo.level = level;
11725 geo.layout = layout;
11726 geo.chunksize = chunksize;
11727 geo.raid_disks = raid_disks;
41784c88
AK		 11728 if (delta_disks != UnSet)
11729 geo.raid_disks += delta_disks;
78b10e66 11730
1ade5cc1
N		 11731 dprintf("for level : %i\n", geo.level);
11732 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11733
4dd2df09 11734 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11735 /* On container level we can only increase number of devices. */
11736 dprintf("imsm: info: Container operation\n");
78b10e66 11737 int old_raid_disks = 0;
6dc0be30 11738
78b10e66 11739 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11740 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11741 struct imsm_update_reshape *u = NULL;
11742 int len;
11743
11744 len = imsm_create_metadata_update_for_reshape(
11745 st, &geo, old_raid_disks, &u);
11746
ed08d51c
AK		 11747 if (len <= 0) {
11748 dprintf("imsm: Cannot prepare update\n");
11749 goto exit_imsm_reshape_super;
11750 }
11751
8dd70bce
AK		 11752 ret_val = 0;
11753 /* update metadata locally */
11754 imsm_update_metadata_locally(st, u, len);
11755 /* and possibly remotely */
11756 if (st->update_tail)
11757 append_metadata_update(st, u, len);
11758 else
ed08d51c 11759 free(u);
8dd70bce 11760
694575e7 11761 } else {
7a862a02 11762 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11763 }
11764 } else {
11765 /* On volume level we support following operations
471bceb6
KW		 11766 * - takeover: raid10 -> raid0; raid0 -> raid10
11767 * - chunk size migration
11768 * - migration: raid5 -> raid0; raid0 -> raid5
11769 */
11770 struct intel_super *super = st->sb;
11771 struct intel_dev *dev = super->devlist;
4dd2df09 11772 int change;
694575e7 11773 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11774 /* find requested device */
11775 while (dev) {
1011e834 11776 char *devnm =
4dd2df09
N		 11777 imsm_find_array_devnm_by_subdev(
11778 dev->index, st->container_devnm);
11779 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11780 break;
11781 dev = dev->next;
11782 }
11783 if (dev == NULL) {
4dd2df09
N		 11784 pr_err("Cannot find %s (%s) subarray\n",
11785 geo.dev_name, geo.devnm);
471bceb6
KW		 11786 goto exit_imsm_reshape_super;
11787 }
11788 super->current_vol = dev->index;
fbf3d202 11789 change = imsm_analyze_change(st, &geo, direction);
694575e7 11790 switch (change) {
471bceb6 11791 case CH_TAKEOVER:
bb025c2f 11792 ret_val = imsm_takeover(st, &geo);
694575e7 11793 break;
48c5303a
PC		 11794 case CH_MIGRATION: {
11795 struct imsm_update_reshape_migration *u = NULL;
11796 int len =
11797 imsm_create_metadata_update_for_migration(
11798 st, &geo, &u);
11799 if (len < 1) {
7a862a02 11800 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11801 break;
11802 }
471bceb6 11803 ret_val = 0;
48c5303a
PC		 11804 /* update metadata locally */
11805 imsm_update_metadata_locally(st, u, len);
11806 /* and possibly remotely */
11807 if (st->update_tail)
11808 append_metadata_update(st, u, len);
11809 else
11810 free(u);
11811 }
11812 break;
7abc9871 11813 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11814 struct imsm_update_size_change *u = NULL;
11815 int len =
11816 imsm_create_metadata_update_for_size_change(
11817 st, &geo, &u);
11818 if (len < 1) {
7a862a02 11819 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11820 break;
11821 }
11822 ret_val = 0;
11823 /* update metadata locally */
11824 imsm_update_metadata_locally(st, u, len);
11825 /* and possibly remotely */
11826 if (st->update_tail)
11827 append_metadata_update(st, u, len);
11828 else
11829 free(u);
7abc9871
AK		 11830 }
11831 break;
471bceb6
KW		 11832 default:
11833 ret_val = 1;
694575e7 11834 }
694575e7 11835 }
78b10e66 11836
ed08d51c 11837exit_imsm_reshape_super:
78b10e66
N		 11838 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11839 return ret_val;
11840}
2cda7640 11841
0febb20c
AO		 11842#define COMPLETED_OK 0
11843#define COMPLETED_NONE 1
11844#define COMPLETED_DELAYED 2
11845
11846static int read_completed(int fd, unsigned long long *val)
11847{
11848 int ret;
11849 char buf[50];
11850
11851 ret = sysfs_fd_get_str(fd, buf, 50);
11852 if (ret < 0)
11853 return ret;
11854
11855 ret = COMPLETED_OK;
11856 if (strncmp(buf, "none", 4) == 0) {
11857 ret = COMPLETED_NONE;
11858 } else if (strncmp(buf, "delayed", 7) == 0) {
11859 ret = COMPLETED_DELAYED;
11860 } else {
11861 char *ep;
11862 *val = strtoull(buf, &ep, 0);
11863 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11864 ret = -1;
11865 }
11866 return ret;
11867}
11868
eee67a47
AK		 11869/*******************************************************************************
11870 * Function: wait_for_reshape_imsm
11871 * Description: Function writes new sync_max value and waits until
11872 * reshape process reach new position
11873 * Parameters:
11874 * sra : general array info
eee67a47
AK		 11875 * ndata : number of disks in new array's layout
11876 * Returns:
11877 * 0 : success,
11878 * 1 : there is no reshape in progress,
11879 * -1 : fail
11880 ******************************************************************************/
ae9f01f8 11881int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11882{
85ca499c 11883 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11884 int retry = 3;
eee67a47 11885 unsigned long long completed;
ae9f01f8
AK		 11886 /* to_complete : new sync_max position */
11887 unsigned long long to_complete = sra->reshape_progress;
11888 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11889
ae9f01f8 11890 if (fd < 0) {
1ade5cc1 11891 dprintf("cannot open reshape_position\n");
eee67a47 11892 return 1;
ae9f01f8 11893 }
eee67a47 11894
df2647fa
PB		 11895 do {
11896 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11897 if (!retry) {
11898 dprintf("cannot read reshape_position (no reshape in progres)\n");
11899 close(fd);
11900 return 1;
11901 }
11902 usleep(30000);
11903 } else
11904 break;
11905 } while (retry--);
eee67a47 11906
85ca499c 11907 if (completed > position_to_set) {
1ade5cc1 11908 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11909 to_complete, position_to_set);
ae9f01f8
AK		 11910 close(fd);
11911 return -1;
11912 }
11913 dprintf("Position set: %llu\n", position_to_set);
11914 if (sysfs_set_num(sra, NULL, "sync_max",
11915 position_to_set) != 0) {
1ade5cc1 11916 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11917 position_to_set);
11918 close(fd);
11919 return -1;
eee67a47
AK		 11920 }
11921
eee67a47 11922 do {
0febb20c 11923 int rc;
eee67a47 11924 char action[20];
5ff3a780 11925 int timeout = 3000;
0febb20c 11926
5ff3a780 11927 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11928 if (sysfs_get_str(sra, NULL, "sync_action",
11929 action, 20) > 0 &&
d7d3809a 11930 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11931 if (strncmp(action, "idle", 4) == 0)
11932 break;
d7d3809a
AP		 11933 close(fd);
11934 return -1;
11935 }
0febb20c
AO		 11936
11937 rc = read_completed(fd, &completed);
11938 if (rc < 0) {
1ade5cc1 11939 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11940 close(fd);
11941 return 1;
0febb20c
AO		 11942 } else if (rc == COMPLETED_NONE)
11943 break;
85ca499c 11944 } while (completed < position_to_set);
b2be2b62 11945
eee67a47
AK		 11946 close(fd);
11947 return 0;
eee67a47
AK		 11948}
11949
b915c95f
AK		 11950/*******************************************************************************
11951 * Function: check_degradation_change
11952 * Description: Check that array hasn't become failed.
11953 * Parameters:
11954 * info : for sysfs access
11955 * sources : source disks descriptors
11956 * degraded: previous degradation level
11957 * Returns:
11958 * degradation level
11959 ******************************************************************************/
11960int check_degradation_change(struct mdinfo *info,
11961 int *sources,
11962 int degraded)
11963{
11964 unsigned long long new_degraded;
e1993023
LD		 11965 int rv;
11966
11967 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11968 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11969 /* check each device to ensure it is still working */
11970 struct mdinfo *sd;
11971 new_degraded = 0;
11972 for (sd = info->devs ; sd ; sd = sd->next) {
11973 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11974 continue;
11975 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11976 char sbuf[100];
11977
b915c95f 11978 if (sysfs_get_str(info,
cf52eff5 11979 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11980 strstr(sbuf, "faulty") ||
11981 strstr(sbuf, "in_sync") == NULL) {
11982 /* this device is dead */
11983 sd->disk.state = (1<<MD_DISK_FAULTY);
11984 if (sd->disk.raid_disk >= 0 &&
11985 sources[sd->disk.raid_disk] >= 0) {
11986 close(sources[
11987 sd->disk.raid_disk]);
11988 sources[sd->disk.raid_disk] =
11989 -1;
11990 }
11991 new_degraded++;
11992 }
11993 }
11994 }
11995 }
11996
11997 return new_degraded;
11998}
11999
10f22854
AK		 12000/*******************************************************************************
12001 * Function: imsm_manage_reshape
12002 * Description: Function finds array under reshape and it manages reshape
12003 * process. It creates stripes backups (if required) and sets
942e1cdb 12004 * checkpoints.
10f22854
AK		 12005 * Parameters:
12006 * afd : Backup handle (nattive) - not used
12007 * sra : general array info
12008 * reshape : reshape parameters - not used
12009 * st : supertype structure
12010 * blocks : size of critical section [blocks]
12011 * fds : table of source device descriptor
12012 * offsets : start of array (offest per devices)
12013 * dests : not used
12014 * destfd : table of destination device descriptor
12015 * destoffsets : table of destination offsets (per device)
12016 * Returns:
12017 * 1 : success, reshape is done
12018 * 0 : fail
12019 ******************************************************************************/
999b4972
N		 12020static int imsm_manage_reshape(
12021 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12022 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12023 int *fds, unsigned long long *offsets,
12024 int dests, int *destfd, unsigned long long *destoffsets)
12025{
10f22854
AK		 12026 int ret_val = 0;
12027 struct intel_super *super = st->sb;
594dc1b8 12028 struct intel_dev *dv;
de44e46f 12029 unsigned int sector_size = super->sector_size;
10f22854 12030 struct imsm_dev *dev = NULL;
9529d343 12031 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12032 int migr_vol_qan = 0;
12033 int ndata, odata; /* [bytes] */
12034 int chunk; /* [bytes] */
12035 struct migr_record *migr_rec;
12036 char *buf = NULL;
12037 unsigned int buf_size; /* [bytes] */
12038 unsigned long long max_position; /* array size [bytes] */
12039 unsigned long long next_step; /* [blocks]/[bytes] */
12040 unsigned long long old_data_stripe_length;
10f22854
AK		 12041 unsigned long long start_src; /* [bytes] */
12042 unsigned long long start; /* [bytes] */
12043 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12044 int degraded = 0;
ab724b98 12045 int source_layout = 0;
10f22854 12046
79a16a9b
JS		 12047 if (!sra)
12048 return ret_val;
12049
12050 if (!fds || !offsets)
10f22854
AK		 12051 goto abort;
12052
12053 /* Find volume during the reshape */
12054 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12055 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12056 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12057 dev = dv->dev;
12058 migr_vol_qan++;
12059 }
12060 }
12061 /* Only one volume can migrate at the same time */
12062 if (migr_vol_qan != 1) {
676e87a8 12063 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12064 "Number of migrating volumes greater than 1\n" :
12065 "There is no volume during migrationg\n");
12066 goto abort;
12067 }
12068
9529d343 12069 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12070 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12071 if (map_src == NULL)
12072 goto abort;
10f22854 12073
9529d343
MD		 12074 ndata = imsm_num_data_members(map_dest);
12075 odata = imsm_num_data_members(map_src);
10f22854 12076
7b1ab482 12077 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12078 old_data_stripe_length = odata * chunk;
12079
12080 migr_rec = super->migr_rec;
12081
10f22854
AK		 12082 /* initialize migration record for start condition */
12083 if (sra->reshape_progress == 0)
12084 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12085 else {
12086 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12087 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12088 goto abort;
12089 }
6a75c8ca
AK		 12090 /* Save checkpoint to update migration record for current
12091 * reshape position (in md). It can be farther than current
12092 * reshape position in metadata.
12093 */
12094 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12095 /* ignore error == 2, this can mean end of reshape here
12096 */
7a862a02 12097 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12098 goto abort;
12099 }
b2c59438 12100 }
10f22854
AK		 12101
12102 /* size for data */
12103 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12104 /* extend buffer size for parity disk */
12105 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12106 /* add space for stripe alignment */
10f22854 12107 buf_size += old_data_stripe_length;
de44e46f
PB		 12108 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12109 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12110 goto abort;
12111 }
12112
3ef4403c 12113 max_position = sra->component_size * ndata;
68eb8bc6 12114 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12115
9f421827
PB		 12116 while (current_migr_unit(migr_rec) <
12117 get_num_migr_units(migr_rec)) {
10f22854
AK		 12118 /* current reshape position [blocks] */
12119 unsigned long long current_position =
12120 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12121 * current_migr_unit(migr_rec);
10f22854
AK		 12122 unsigned long long border;
12123
b915c95f
AK		 12124 /* Check that array hasn't become failed.
12125 */
12126 degraded = check_degradation_change(sra, fds, degraded);
12127 if (degraded > 1) {
7a862a02 12128 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12129 goto abort;
12130 }
12131
10f22854
AK		 12132 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12133
12134 if ((current_position + next_step) > max_position)
12135 next_step = max_position - current_position;
12136
92144abf 12137 start = current_position * 512;
10f22854 12138
942e1cdb 12139 /* align reading start to old geometry */
10f22854
AK		 12140 start_buf_shift = start % old_data_stripe_length;
12141 start_src = start - start_buf_shift;
12142
12143 border = (start_src / odata) - (start / ndata);
12144 border /= 512;
12145 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12146 /* save critical stripes to buf
12147 * start - start address of current unit
12148 * to backup [bytes]
12149 * start_src - start address of current unit
12150 * to backup alligned to source array
12151 * [bytes]
12152 */
594dc1b8 12153 unsigned long long next_step_filler;
10f22854
AK		 12154 unsigned long long copy_length = next_step * 512;
12155
12156 /* allign copy area length to stripe in old geometry */
12157 next_step_filler = ((copy_length + start_buf_shift)
12158 % old_data_stripe_length);
12159 if (next_step_filler)
12160 next_step_filler = (old_data_stripe_length
12161 - next_step_filler);
7a862a02 12162 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		 12163 start, start_src, copy_length,
12164 start_buf_shift, next_step_filler);
12165
12166 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12167 chunk, map_src->raid_level,
12168 source_layout, 0, NULL, start_src,
10f22854
AK		 12169 copy_length +
12170 next_step_filler + start_buf_shift,
12171 buf)) {
7a862a02 12172 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12173 goto abort;
12174 }
12175 /* Convert data to destination format and store it
12176 * in backup general migration area
12177 */
12178 if (save_backup_imsm(st, dev, sra,
aea93171 12179 buf + start_buf_shift, copy_length)) {
7a862a02 12180 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12181 goto abort;
12182 }
12183 if (save_checkpoint_imsm(st, sra,
12184 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12185 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12186 goto abort;
12187 }
8016a6d4
AK		 12188 } else {
12189 /* set next step to use whole border area */
12190 border /= next_step;
12191 if (border > 1)
12192 next_step *= border;
10f22854
AK		 12193 }
12194 /* When data backed up, checkpoint stored,
12195 * kick the kernel to reshape unit of data
12196 */
12197 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12198 /* limit next step to array max position */
12199 if (next_step > max_position)
12200 next_step = max_position;
10f22854
AK		 12201 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12202 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12203 sra->reshape_progress = next_step;
10f22854
AK		 12204
12205 /* wait until reshape finish */
c85338c6 12206 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12207 dprintf("wait_for_reshape_imsm returned error!\n");
12208 goto abort;
12209 }
84d11e6c
N		 12210 if (sigterm)
12211 goto abort;
10f22854 12212
0228d92c
AK		 12213 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12214 /* ignore error == 2, this can mean end of reshape here
12215 */
7a862a02 12216 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12217 goto abort;
12218 }
12219
12220 }
12221
71e5411e
PB		 12222 /* clear migr_rec on disks after successful migration */
12223 struct dl *d;
12224
85337573 12225 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12226 for (d = super->disks; d; d = d->next) {
12227 if (d->index < 0 || is_failed(&d->disk))
12228 continue;
12229 unsigned long long dsize;
12230
12231 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12232 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12233 SEEK_SET) >= 0) {
466070ad 12234 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12235 MIGR_REC_BUF_SECTORS*sector_size) !=
12236 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12237 perror("Write migr_rec failed");
12238 }
12239 }
12240
10f22854
AK		 12241 /* return '1' if done */
12242 ret_val = 1;
12243abort:
12244 free(buf);
942e1cdb
N		 12245 /* See Grow.c: abort_reshape() for further explanation */
12246 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12247 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12248 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12249
12250 return ret_val;
999b4972 12251}
0c21b485 12252
cdddbdbc 12253struct superswitch super_imsm = {
cdddbdbc
DW		 12254 .examine_super = examine_super_imsm,
12255 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12256 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12257 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12258 .detail_super = detail_super_imsm,
12259 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12260 .write_init_super = write_init_super_imsm,
0e600426
N		 12261 .validate_geometry = validate_geometry_imsm,
12262 .add_to_super = add_to_super_imsm,
1a64be56 12263 .remove_from_super = remove_from_super_imsm,
d665cc31 12264 .detail_platform = detail_platform_imsm,
e50cf220 12265 .export_detail_platform = export_detail_platform_imsm,
33414a01 12266 .kill_subarray = kill_subarray_imsm,
aa534678 12267 .update_subarray = update_subarray_imsm,
2b959fbf 12268 .load_container = load_container_imsm,
71204a50
N		 12269 .default_geometry = default_geometry_imsm,
12270 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12271 .reshape_super = imsm_reshape_super,
12272 .manage_reshape = imsm_manage_reshape,
9e2d750d 12273 .recover_backup = recover_backup_imsm,
74db60b0 12274 .copy_metadata = copy_metadata_imsm,
27156a57 12275 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12276 .match_home = match_home_imsm,
12277 .uuid_from_super= uuid_from_super_imsm,
12278 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12279 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12280 .update_super = update_super_imsm,
12281
12282 .avail_size = avail_size_imsm,
fbfdcb06 12283 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12284
12285 .compare_super = compare_super_imsm,
12286
12287 .load_super = load_super_imsm,
bf5a934a 12288 .init_super = init_super_imsm,
e683ca88 12289 .store_super = store_super_imsm,
cdddbdbc
DW		 12290 .free_super = free_super_imsm,
12291 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12292 .container_content = container_content_imsm,
0c21b485 12293 .validate_container = validate_container_imsm,
cdddbdbc 12294
2432ce9b
AP		 12295 .write_init_ppl = write_init_ppl_imsm,
12296 .validate_ppl = validate_ppl_imsm,
12297
cdddbdbc 12298 .external = 1,
4cce4069 12299 .name = "imsm",
845dea95
NB		 12300
12301/* for mdmon */
12302 .open_new = imsm_open_new,
ed9d66aa 12303 .set_array_state= imsm_set_array_state,
845dea95
NB		 12304 .set_disk = imsm_set_disk,
12305 .sync_metadata = imsm_sync_metadata,
88758e9d 12306 .activate_spare = imsm_activate_spare,
e8319a19 12307 .process_update = imsm_process_update,
8273f55e 12308 .prepare_update = imsm_prepare_update,
6f50473f 12309 .record_bad_block = imsm_record_badblock,
c07a5a4f 12310 .clear_bad_block = imsm_clear_badblock,
928f1424 12311 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12312};
Unnamed repository; edit this file 'description' to name the repository.