]> 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
super-intel: don't mark structs 'packed' unnecessarily
[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
cdddbdbc
DW		 2186static void detail_super_imsm(struct supertype *st, char *homehost)
2187{
3ebe00a1
DW		 2188 struct mdinfo info;
2189 char nbuf[64];
2190
a5d85af7 2191 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2192 fname_from_uuid(st, &info, nbuf, ':');
65884368 2193 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2194}
2195
2196static void brief_detail_super_imsm(struct supertype *st)
2197{
ff54de6e
N		 2198 struct mdinfo info;
2199 char nbuf[64];
a5d85af7 2200 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2201 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2202 printf(" UUID=%s", nbuf + 5);
cdddbdbc 2203}
d665cc31
DW		 2204
2205static int imsm_read_serial(int fd, char *devname, __u8 *serial);
2206static void fd2devname(int fd, char *name);
2207
120dc887 2208static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2209{
120dc887
LM		 2210 /* dump an unsorted list of devices attached to AHCI Intel storage
2211 * controller, as well as non-connected ports
d665cc31
DW		 2212 */
2213 int hba_len = strlen(hba_path) + 1;
2214 struct dirent *ent;
2215 DIR *dir;
2216 char *path = NULL;
2217 int err = 0;
2218 unsigned long port_mask = (1 << port_count) - 1;
2219
f21e18ca 2220 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2221 if (verbose > 0)
e7b84f9d 2222 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2223 return 2;
2224 }
2225
2226 /* scroll through /sys/dev/block looking for devices attached to
2227 * this hba
2228 */
2229 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2230 if (!dir)
2231 return 1;
2232
2233 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2234 int fd;
2235 char model[64];
2236 char vendor[64];
2237 char buf[1024];
2238 int major, minor;
2239 char *device;
2240 char *c;
2241 int port;
2242 int type;
2243
2244 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2245 continue;
2246 path = devt_to_devpath(makedev(major, minor));
2247 if (!path)
2248 continue;
2249 if (!path_attached_to_hba(path, hba_path)) {
2250 free(path);
2251 path = NULL;
2252 continue;
2253 }
2254
2255 /* retrieve the scsi device type */
2256 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2257 if (verbose > 0)
e7b84f9d 2258 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2259 err = 2;
2260 break;
2261 }
2262 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2263 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2264 if (verbose > 0)
e7b84f9d 2265 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2266 path);
2267 err = 2;
2268 free(device);
2269 break;
2270 }
2271 type = strtoul(buf, NULL, 10);
2272
2273 /* if it's not a disk print the vendor and model */
2274 if (!(type == 0 || type == 7 || type == 14)) {
2275 vendor[0] = '\0';
2276 model[0] = '\0';
2277 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2278 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2279 strncpy(vendor, buf, sizeof(vendor));
2280 vendor[sizeof(vendor) - 1] = '\0';
2281 c = (char *) &vendor[sizeof(vendor) - 1];
2282 while (isspace(*c) || *c == '\0')
2283 *c-- = '\0';
2284
2285 }
2286 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2287 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2288 strncpy(model, buf, sizeof(model));
2289 model[sizeof(model) - 1] = '\0';
2290 c = (char *) &model[sizeof(model) - 1];
2291 while (isspace(*c) || *c == '\0')
2292 *c-- = '\0';
2293 }
2294
2295 if (vendor[0] && model[0])
2296 sprintf(buf, "%.64s %.64s", vendor, model);
2297 else
2298 switch (type) { /* numbers from hald/linux/device.c */
2299 case 1: sprintf(buf, "tape"); break;
2300 case 2: sprintf(buf, "printer"); break;
2301 case 3: sprintf(buf, "processor"); break;
2302 case 4:
2303 case 5: sprintf(buf, "cdrom"); break;
2304 case 6: sprintf(buf, "scanner"); break;
2305 case 8: sprintf(buf, "media_changer"); break;
2306 case 9: sprintf(buf, "comm"); break;
2307 case 12: sprintf(buf, "raid"); break;
2308 default: sprintf(buf, "unknown");
2309 }
2310 } else
2311 buf[0] = '\0';
2312 free(device);
2313
2314 /* chop device path to 'host%d' and calculate the port number */
2315 c = strchr(&path[hba_len], '/');
4e5e717d 2316 if (!c) {
ba728be7 2317 if (verbose > 0)
e7b84f9d 2318 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2319 err = 2;
2320 break;
2321 }
d665cc31 2322 *c = '\0';
0858eccf
AP		 2323 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2324 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2325 port -= host_base;
2326 else {
ba728be7 2327 if (verbose > 0) {
d665cc31 2328 *c = '/'; /* repair the full string */
e7b84f9d 2329 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2330 path);
2331 }
2332 err = 2;
2333 break;
2334 }
2335
2336 /* mark this port as used */
2337 port_mask &= ~(1 << port);
2338
2339 /* print out the device information */
2340 if (buf[0]) {
2341 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2342 continue;
2343 }
2344
2345 fd = dev_open(ent->d_name, O_RDONLY);
2346 if (fd < 0)
2347 printf(" Port%d : - disk info unavailable -\n", port);
2348 else {
2349 fd2devname(fd, buf);
2350 printf(" Port%d : %s", port, buf);
2351 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2352 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2353 else
664d5325 2354 printf(" ()\n");
4dab422a 2355 close(fd);
d665cc31 2356 }
d665cc31
DW		 2357 free(path);
2358 path = NULL;
2359 }
2360 if (path)
2361 free(path);
2362 if (dir)
2363 closedir(dir);
2364 if (err == 0) {
2365 int i;
2366
2367 for (i = 0; i < port_count; i++)
2368 if (port_mask & (1 << i))
2369 printf(" Port%d : - no device attached -\n", i);
2370 }
2371
2372 return err;
2373}
2374
b5eece69 2375static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2376{
2377 struct dirent *ent;
2378 DIR *dir;
2379 char path[292];
2380 char link[256];
2381 char *c, *rp;
2382
2383 if (hba->type != SYS_DEV_VMD)
b5eece69 2384 return 1;
60f0f54d
PB		 2385
2386 /* scroll through /sys/dev/block looking for devices attached to
2387 * this hba
2388 */
2389 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2390 if (!dir)
b5eece69 2391 return 1;
b9135011
JS		 2392
2393 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2394 int n;
2395
2396 /* is 'ent' a device? check that the 'subsystem' link exists and
2397 * that its target matches 'bus'
2398 */
2399 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2400 ent->d_name);
2401 n = readlink(path, link, sizeof(link));
2402 if (n < 0 || n >= (int)sizeof(link))
2403 continue;
2404 link[n] = '\0';
2405 c = strrchr(link, '/');
2406 if (!c)
2407 continue;
2408 if (strncmp("pci", c+1, strlen("pci")) != 0)
2409 continue;
2410
2411 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2412
2413 rp = realpath(path, NULL);
2414 if (!rp)
2415 continue;
2416
2417 if (path_attached_to_hba(rp, hba->path)) {
2418 printf(" NVMe under VMD : %s\n", rp);
2419 }
2420 free(rp);
2421 }
2422
b9135011 2423 closedir(dir);
b5eece69 2424 return 0;
60f0f54d
PB		 2425}
2426
120dc887
LM		 2427static void print_found_intel_controllers(struct sys_dev *elem)
2428{
2429 for (; elem; elem = elem->next) {
e7b84f9d 2430 pr_err("found Intel(R) ");
120dc887
LM		 2431 if (elem->type == SYS_DEV_SATA)
2432 fprintf(stderr, "SATA ");
155cbb4c
LM		 2433 else if (elem->type == SYS_DEV_SAS)
2434 fprintf(stderr, "SAS ");
0858eccf
AP		 2435 else if (elem->type == SYS_DEV_NVME)
2436 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2437
2438 if (elem->type == SYS_DEV_VMD)
2439 fprintf(stderr, "VMD domain");
2440 else
2441 fprintf(stderr, "RAID controller");
2442
120dc887
LM		 2443 if (elem->pci_id)
2444 fprintf(stderr, " at %s", elem->pci_id);
2445 fprintf(stderr, ".\n");
2446 }
2447 fflush(stderr);
2448}
2449
120dc887
LM		 2450static int ahci_get_port_count(const char *hba_path, int *port_count)
2451{
2452 struct dirent *ent;
2453 DIR *dir;
2454 int host_base = -1;
2455
2456 *port_count = 0;
2457 if ((dir = opendir(hba_path)) == NULL)
2458 return -1;
2459
2460 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2461 int host;
2462
0858eccf
AP		 2463 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2464 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2465 continue;
2466 if (*port_count == 0)
2467 host_base = host;
2468 else if (host < host_base)
2469 host_base = host;
2470
2471 if (host + 1 > *port_count + host_base)
2472 *port_count = host + 1 - host_base;
2473 }
2474 closedir(dir);
2475 return host_base;
2476}
2477
a891a3c2
LM		 2478static void print_imsm_capability(const struct imsm_orom *orom)
2479{
0858eccf
AP		 2480 printf(" Platform : Intel(R) ");
2481 if (orom->capabilities == 0 && orom->driver_features == 0)
2482 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2483 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2484 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2485 else
2486 printf("Rapid Storage Technology%s\n",
2487 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2488 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2489 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2490 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2491 printf(" RAID Levels :%s%s%s%s%s\n",
2492 imsm_orom_has_raid0(orom) ? " raid0" : "",
2493 imsm_orom_has_raid1(orom) ? " raid1" : "",
2494 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2495 imsm_orom_has_raid10(orom) ? " raid10" : "",
2496 imsm_orom_has_raid5(orom) ? " raid5" : "");
2497 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2498 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2499 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2500 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2501 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2502 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2503 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2504 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2505 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2506 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2507 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2508 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2509 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2510 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2511 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2512 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2513 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2514 printf(" 2TB volumes :%s supported\n",
2515 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2516 printf(" 2TB disks :%s supported\n",
2517 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2518 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2519 printf(" Max Volumes : %d per array, %d per %s\n",
2520 orom->vpa, orom->vphba,
2521 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2522 return;
2523}
2524
e50cf220
MN		 2525static void print_imsm_capability_export(const struct imsm_orom *orom)
2526{
2527 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2528 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2529 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2530 orom->hotfix_ver, orom->build);
e50cf220
MN		 2531 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2532 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2533 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2534 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2535 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2536 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2537 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2538 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2539 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2540 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2541 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2542 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2543 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2544 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2545 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2546 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2547 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2548 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2549 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2550 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2551 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2552 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2553 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2554 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2555 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2556 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2557 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2558 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2559}
2560
9eafa1de 2561static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2562{
2563 /* There are two components to imsm platform support, the ahci SATA
2564 * controller and the option-rom. To find the SATA controller we
2565 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2566 * controller with the Intel vendor id is present. This approach
2567 * allows mdadm to leverage the kernel's ahci detection logic, with the
2568 * caveat that if ahci.ko is not loaded mdadm will not be able to
2569 * detect platform raid capabilities. The option-rom resides in a
2570 * platform "Adapter ROM". We scan for its signature to retrieve the
2571 * platform capabilities. If raid support is disabled in the BIOS the
2572 * option-rom capability structure will not be available.
2573 */
d665cc31 2574 struct sys_dev *list, *hba;
d665cc31
DW		 2575 int host_base = 0;
2576 int port_count = 0;
9eafa1de 2577 int result=1;
d665cc31 2578
5615172f 2579 if (enumerate_only) {
a891a3c2 2580 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2581 return 0;
a891a3c2
LM		 2582 list = find_intel_devices();
2583 if (!list)
2584 return 2;
2585 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2586 if (find_imsm_capability(hba)) {
2587 result = 0;
a891a3c2
LM		 2588 break;
2589 }
9eafa1de 2590 else
6b781d33 2591 result = 2;
a891a3c2 2592 }
a891a3c2 2593 return result;
5615172f
DW		 2594 }
2595
155cbb4c
LM		 2596 list = find_intel_devices();
2597 if (!list) {
ba728be7 2598 if (verbose > 0)
7a862a02 2599 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2600 return 2;
ba728be7 2601 } else if (verbose > 0)
155cbb4c 2602 print_found_intel_controllers(list);
d665cc31 2603
a891a3c2 2604 for (hba = list; hba; hba = hba->next) {
0858eccf 2605 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2606 continue;
0858eccf 2607 if (!find_imsm_capability(hba)) {
60f0f54d 2608 char buf[PATH_MAX];
e7b84f9d 2609 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2610 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2611 get_sys_dev_type(hba->type));
0858eccf
AP		 2612 continue;
2613 }
2614 result = 0;
2615 }
2616
2617 if (controller_path && result == 1) {
2618 pr_err("no active Intel(R) RAID controller found under %s\n",
2619 controller_path);
2620 return result;
2621 }
2622
5e1d6128 2623 const struct orom_entry *entry;
0858eccf 2624
5e1d6128 2625 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2626 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2627 print_imsm_capability(&entry->orom);
32716c51
PB		 2628 printf(" 3rd party NVMe :%s supported\n",
2629 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2630 for (hba = list; hba; hba = hba->next) {
2631 if (hba->type == SYS_DEV_VMD) {
2632 char buf[PATH_MAX];
60f0f54d
PB		 2633 printf(" I/O Controller : %s (%s)\n",
2634 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2635 if (print_vmd_attached_devs(hba)) {
2636 if (verbose > 0)
2637 pr_err("failed to get devices attached to VMD domain.\n");
2638 result |= 2;
2639 }
60f0f54d
PB		 2640 }
2641 }
07cb1e57 2642 printf("\n");
60f0f54d
PB		 2643 continue;
2644 }
0858eccf 2645
60f0f54d
PB		 2646 print_imsm_capability(&entry->orom);
2647 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2648 for (hba = list; hba; hba = hba->next) {
2649 if (hba->type == SYS_DEV_NVME)
2650 printf(" NVMe Device : %s\n", hba->path);
2651 }
60f0f54d 2652 printf("\n");
0858eccf
AP		 2653 continue;
2654 }
2655
2656 struct devid_list *devid;
5e1d6128 2657 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2658 hba = device_by_id(devid->devid);
2659 if (!hba)
2660 continue;
2661
9eafa1de
MN		 2662 printf(" I/O Controller : %s (%s)\n",
2663 hba->path, get_sys_dev_type(hba->type));
2664 if (hba->type == SYS_DEV_SATA) {
2665 host_base = ahci_get_port_count(hba->path, &port_count);
2666 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2667 if (verbose > 0)
7a862a02 2668 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2669 result |= 2;
2670 }
120dc887
LM		 2671 }
2672 }
0858eccf 2673 printf("\n");
d665cc31 2674 }
155cbb4c 2675
120dc887 2676 return result;
d665cc31 2677}
e50cf220 2678
9eafa1de 2679static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2680{
e50cf220
MN		 2681 struct sys_dev *list, *hba;
2682 int result=1;
2683
2684 list = find_intel_devices();
2685 if (!list) {
2686 if (verbose > 0)
2687 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2688 result = 2;
e50cf220
MN		 2689 return result;
2690 }
2691
2692 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2693 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2694 continue;
60f0f54d
PB		 2695 if (!find_imsm_capability(hba) && verbose > 0) {
2696 char buf[PATH_MAX];
2697 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2698 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2699 }
0858eccf 2700 else
e50cf220 2701 result = 0;
e50cf220
MN		 2702 }
2703
5e1d6128 2704 const struct orom_entry *entry;
0858eccf 2705
60f0f54d
PB		 2706 for (entry = orom_entries; entry; entry = entry->next) {
2707 if (entry->type == SYS_DEV_VMD) {
2708 for (hba = list; hba; hba = hba->next)
2709 print_imsm_capability_export(&entry->orom);
2710 continue;
2711 }
5e1d6128 2712 print_imsm_capability_export(&entry->orom);
60f0f54d 2713 }
0858eccf 2714
e50cf220
MN		 2715 return result;
2716}
2717
cdddbdbc
DW		 2718static int match_home_imsm(struct supertype *st, char *homehost)
2719{
5115ca67
DW		 2720 /* the imsm metadata format does not specify any host
2721 * identification information. We return -1 since we can never
2722 * confirm nor deny whether a given array is "meant" for this
148acb7b 2723 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2724 * exclude member disks that do not belong, and we rely on
2725 * mdadm.conf to specify the arrays that should be assembled.
2726 * Auto-assembly may still pick up "foreign" arrays.
2727 */
cdddbdbc 2728
9362c1c8 2729 return -1;
cdddbdbc
DW		 2730}
2731
2732static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2733{
51006d85
N		 2734 /* The uuid returned here is used for:
2735 * uuid to put into bitmap file (Create, Grow)
2736 * uuid for backup header when saving critical section (Grow)
2737 * comparing uuids when re-adding a device into an array
2738 * In these cases the uuid required is that of the data-array,
2739 * not the device-set.
2740 * uuid to recognise same set when adding a missing device back
2741 * to an array. This is a uuid for the device-set.
1011e834 2742 *
51006d85
N		 2743 * For each of these we can make do with a truncated
2744 * or hashed uuid rather than the original, as long as
2745 * everyone agrees.
2746 * In each case the uuid required is that of the data-array,
2747 * not the device-set.
43dad3d6 2748 */
51006d85
N		 2749 /* imsm does not track uuid's so we synthesis one using sha1 on
2750 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2751 * - the orig_family_num of the container
51006d85
N		 2752 * - the index number of the volume
2753 * - the 'serial' number of the volume.
2754 * Hopefully these are all constant.
2755 */
2756 struct intel_super *super = st->sb;
43dad3d6 2757
51006d85
N		 2758 char buf[20];
2759 struct sha1_ctx ctx;
2760 struct imsm_dev *dev = NULL;
148acb7b 2761 __u32 family_num;
51006d85 2762
148acb7b
DW		 2763 /* some mdadm versions failed to set ->orig_family_num, in which
2764 * case fall back to ->family_num. orig_family_num will be
2765 * fixed up with the first metadata update.
2766 */
2767 family_num = super->anchor->orig_family_num;
2768 if (family_num == 0)
2769 family_num = super->anchor->family_num;
51006d85 2770 sha1_init_ctx(&ctx);
92bd8f8d 2771 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2772 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2773 if (super->current_vol >= 0)
2774 dev = get_imsm_dev(super, super->current_vol);
2775 if (dev) {
2776 __u32 vol = super->current_vol;
2777 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2778 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2779 }
2780 sha1_finish_ctx(&ctx, buf);
2781 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2782}
2783
0d481d37 2784#if 0
4f5bc454
DW		 2785static void
2786get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2787{
cdddbdbc
DW		 2788 __u8 *v = get_imsm_version(mpb);
2789 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2790 char major[] = { 0, 0, 0 };
2791 char minor[] = { 0 ,0, 0 };
2792 char patch[] = { 0, 0, 0 };
2793 char *ver_parse[] = { major, minor, patch };
2794 int i, j;
2795
2796 i = j = 0;
2797 while (*v != '\0' && v < end) {
2798 if (*v != '.' && j < 2)
2799 ver_parse[i][j++] = *v;
2800 else {
2801 i++;
2802 j = 0;
2803 }
2804 v++;
2805 }
2806
4f5bc454
DW		 2807 *m = strtol(minor, NULL, 0);
2808 *p = strtol(patch, NULL, 0);
2809}
0d481d37 2810#endif
4f5bc454 2811
1e5c6983
DW		 2812static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2813{
2814 /* migr_strip_size when repairing or initializing parity */
238c0a71 2815 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2816 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2817
2818 switch (get_imsm_raid_level(map)) {
2819 case 5:
2820 case 10:
2821 return chunk;
2822 default:
2823 return 128*1024 >> 9;
2824 }
2825}
2826
2827static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2828{
2829 /* migr_strip_size when rebuilding a degraded disk, no idea why
2830 * this is different than migr_strip_size_resync(), but it's good
2831 * to be compatible
2832 */
238c0a71 2833 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2834 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2835
2836 switch (get_imsm_raid_level(map)) {
2837 case 1:
2838 case 10:
2839 if (map->num_members % map->num_domains == 0)
2840 return 128*1024 >> 9;
2841 else
2842 return chunk;
2843 case 5:
2844 return max((__u32) 64*1024 >> 9, chunk);
2845 default:
2846 return 128*1024 >> 9;
2847 }
2848}
2849
2850static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2851{
238c0a71
AK		 2852 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2853 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2854 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2855 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2856
2857 return max((__u32) 1, hi_chunk / lo_chunk);
2858}
2859
2860static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2861{
238c0a71 2862 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2863 int level = get_imsm_raid_level(lo);
2864
2865 if (level == 1 || level == 10) {
238c0a71 2866 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2867
2868 return hi->num_domains;
2869 } else
2870 return num_stripes_per_unit_resync(dev);
2871}
2872
9529d343 2873static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2874{
2875 /* named 'imsm_' because raid0, raid1 and raid10
2876 * counter-intuitively have the same number of data disks
2877 */
1e5c6983
DW		 2878 switch (get_imsm_raid_level(map)) {
2879 case 0:
36fd8ccc
AK		 2880 return map->num_members;
2881 break;
1e5c6983
DW		 2882 case 1:
2883 case 10:
36fd8ccc 2884 return map->num_members/2;
1e5c6983
DW		 2885 case 5:
2886 return map->num_members - 1;
2887 default:
1ade5cc1 2888 dprintf("unsupported raid level\n");
1e5c6983
DW		 2889 return 0;
2890 }
2891}
2892
44490938
MD		 2893static unsigned long long calc_component_size(struct imsm_map *map,
2894 struct imsm_dev *dev)
2895{
2896 unsigned long long component_size;
2897 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2898 long long calc_dev_size = 0;
44490938
MD		 2899 unsigned int member_disks = imsm_num_data_members(map);
2900
2901 if (member_disks == 0)
2902 return 0;
2903
2904 component_size = per_dev_array_size(map);
2905 calc_dev_size = component_size * member_disks;
2906
2907 /* Component size is rounded to 1MB so difference between size from
2908 * metadata and size calculated from num_data_stripes equals up to
2909 * 2048 blocks per each device. If the difference is higher it means
2910 * that array size was expanded and num_data_stripes was not updated.
2911 */
a4f7290c 2912 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2913 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2914 component_size = dev_size / member_disks;
2915 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2916 component_size / map->blocks_per_strip,
2917 num_data_stripes(map));
2918 }
2919
2920 return component_size;
2921}
2922
1e5c6983
DW		 2923static __u32 parity_segment_depth(struct imsm_dev *dev)
2924{
238c0a71 2925 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2926 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2927
2928 switch(get_imsm_raid_level(map)) {
2929 case 1:
2930 case 10:
2931 return chunk * map->num_domains;
2932 case 5:
2933 return chunk * map->num_members;
2934 default:
2935 return chunk;
2936 }
2937}
2938
2939static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2940{
238c0a71 2941 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2942 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2943 __u32 strip = block / chunk;
2944
2945 switch (get_imsm_raid_level(map)) {
2946 case 1:
2947 case 10: {
2948 __u32 vol_strip = (strip * map->num_domains) + 1;
2949 __u32 vol_stripe = vol_strip / map->num_members;
2950
2951 return vol_stripe * chunk + block % chunk;
2952 } case 5: {
2953 __u32 stripe = strip / (map->num_members - 1);
2954
2955 return stripe * chunk + block % chunk;
2956 }
2957 default:
2958 return 0;
2959 }
2960}
2961
c47b0ff6
AK		 2962static __u64 blocks_per_migr_unit(struct intel_super *super,
2963 struct imsm_dev *dev)
1e5c6983
DW		 2964{
2965 /* calculate the conversion factor between per member 'blocks'
2966 * (md/{resync,rebuild}_start) and imsm migration units, return
2967 * 0 for the 'not migrating' and 'unsupported migration' cases
2968 */
2969 if (!dev->vol.migr_state)
2970 return 0;
2971
2972 switch (migr_type(dev)) {
c47b0ff6
AK		 2973 case MIGR_GEN_MIGR: {
2974 struct migr_record *migr_rec = super->migr_rec;
2975 return __le32_to_cpu(migr_rec->blocks_per_unit);
2976 }
1e5c6983
DW		 2977 case MIGR_VERIFY:
2978 case MIGR_REPAIR:
2979 case MIGR_INIT: {
238c0a71 2980 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2981 __u32 stripes_per_unit;
2982 __u32 blocks_per_unit;
2983 __u32 parity_depth;
2984 __u32 migr_chunk;
2985 __u32 block_map;
2986 __u32 block_rel;
2987 __u32 segment;
2988 __u32 stripe;
2989 __u8 disks;
2990
2991 /* yes, this is really the translation of migr_units to
2992 * per-member blocks in the 'resync' case
2993 */
2994 stripes_per_unit = num_stripes_per_unit_resync(dev);
2995 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2996 disks = imsm_num_data_members(map);
1e5c6983 2997 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2998 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2999 segment = blocks_per_unit / stripe;
3000 block_rel = blocks_per_unit - segment * stripe;
3001 parity_depth = parity_segment_depth(dev);
3002 block_map = map_migr_block(dev, block_rel);
3003 return block_map + parity_depth * segment;
3004 }
3005 case MIGR_REBUILD: {
3006 __u32 stripes_per_unit;
3007 __u32 migr_chunk;
3008
3009 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
3010 migr_chunk = migr_strip_blocks_rebuild(dev);
3011 return migr_chunk * stripes_per_unit;
3012 }
1e5c6983
DW		 3013 case MIGR_STATE_CHANGE:
3014 default:
3015 return 0;
3016 }
3017}
3018
c2c087e6
DW		 3019static int imsm_level_to_layout(int level)
3020{
3021 switch (level) {
3022 case 0:
3023 case 1:
3024 return 0;
3025 case 5:
3026 case 6:
a380c027 3027 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 3028 case 10:
c92a2527 3029 return 0x102;
c2c087e6 3030 }
a18a888e 3031 return UnSet;
c2c087e6
DW		 3032}
3033
8e59f3d8
AK		 3034/*******************************************************************************
3035 * Function: read_imsm_migr_rec
3036 * Description: Function reads imsm migration record from last sector of disk
3037 * Parameters:
3038 * fd : disk descriptor
3039 * super : metadata info
3040 * Returns:
3041 * 0 : success,
3042 * -1 : fail
3043 ******************************************************************************/
3044static int read_imsm_migr_rec(int fd, struct intel_super *super)
3045{
3046 int ret_val = -1;
de44e46f 3047 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3048 unsigned long long dsize;
3049
3050 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3051 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3052 SEEK_SET) < 0) {
e7b84f9d
N		 3053 pr_err("Cannot seek to anchor block: %s\n",
3054 strerror(errno));
8e59f3d8
AK		 3055 goto out;
3056 }
466070ad 3057 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3058 MIGR_REC_BUF_SECTORS*sector_size) !=
3059 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3060 pr_err("Cannot read migr record block: %s\n",
3061 strerror(errno));
8e59f3d8
AK		 3062 goto out;
3063 }
3064 ret_val = 0;
de44e46f
PB		 3065 if (sector_size == 4096)
3066 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3067
3068out:
3069 return ret_val;
3070}
3071
3136abe5
AK		 3072static struct imsm_dev *imsm_get_device_during_migration(
3073 struct intel_super *super)
3074{
3075
3076 struct intel_dev *dv;
3077
3078 for (dv = super->devlist; dv; dv = dv->next) {
3079 if (is_gen_migration(dv->dev))
3080 return dv->dev;
3081 }
3082 return NULL;
3083}
3084
8e59f3d8
AK		 3085/*******************************************************************************
3086 * Function: load_imsm_migr_rec
3087 * Description: Function reads imsm migration record (it is stored at the last
3088 * sector of disk)
3089 * Parameters:
3090 * super : imsm internal array info
3091 * info : general array info
3092 * Returns:
3093 * 0 : success
3094 * -1 : fail
4c965cc9 3095 * -2 : no migration in progress
8e59f3d8
AK		 3096 ******************************************************************************/
3097static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3098{
3099 struct mdinfo *sd;
594dc1b8 3100 struct dl *dl;
8e59f3d8
AK		 3101 char nm[30];
3102 int retval = -1;
3103 int fd = -1;
3136abe5 3104 struct imsm_dev *dev;
594dc1b8 3105 struct imsm_map *map;
b4ab44d8 3106 int slot = -1;
3136abe5
AK		 3107
3108 /* find map under migration */
3109 dev = imsm_get_device_during_migration(super);
3110 /* nothing to load,no migration in progress?
3111 */
3112 if (dev == NULL)
4c965cc9 3113 return -2;
8e59f3d8
AK		 3114
3115 if (info) {
3116 for (sd = info->devs ; sd ; sd = sd->next) {
3117 /* read only from one of the first two slots */
12fe93e9
TM		 3118 if ((sd->disk.raid_disk < 0) ||
3119 (sd->disk.raid_disk > 1))
8e59f3d8 3120 continue;
3136abe5 3121
8e59f3d8
AK		 3122 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3123 fd = dev_open(nm, O_RDONLY);
3124 if (fd >= 0)
3125 break;
3126 }
3127 }
3128 if (fd < 0) {
12fe93e9 3129 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3130 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3131 /* skip spare and failed disks
3132 */
3133 if (dl->index < 0)
3134 continue;
8e59f3d8 3135 /* read only from one of the first two slots */
3136abe5
AK		 3136 if (map)
3137 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3138 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3139 continue;
3140 sprintf(nm, "%d:%d", dl->major, dl->minor);
3141 fd = dev_open(nm, O_RDONLY);
3142 if (fd >= 0)
3143 break;
3144 }
3145 }
3146 if (fd < 0)
3147 goto out;
3148 retval = read_imsm_migr_rec(fd, super);
3149
3150out:
3151 if (fd >= 0)
3152 close(fd);
3153 return retval;
3154}
3155
c17608ea
AK		 3156/*******************************************************************************
3157 * function: imsm_create_metadata_checkpoint_update
3158 * Description: It creates update for checkpoint change.
3159 * Parameters:
3160 * super : imsm internal array info
3161 * u : pointer to prepared update
3162 * Returns:
3163 * Uptate length.
3164 * If length is equal to 0, input pointer u contains no update
3165 ******************************************************************************/
3166static int imsm_create_metadata_checkpoint_update(
3167 struct intel_super *super,
3168 struct imsm_update_general_migration_checkpoint **u)
3169{
3170
3171 int update_memory_size = 0;
3172
1ade5cc1 3173 dprintf("(enter)\n");
c17608ea
AK		 3174
3175 if (u == NULL)
3176 return 0;
3177 *u = NULL;
3178
3179 /* size of all update data without anchor */
3180 update_memory_size =
3181 sizeof(struct imsm_update_general_migration_checkpoint);
3182
503975b9 3183 *u = xcalloc(1, update_memory_size);
c17608ea 3184 if (*u == NULL) {
1ade5cc1 3185 dprintf("error: cannot get memory\n");
c17608ea
AK		 3186 return 0;
3187 }
3188 (*u)->type = update_general_migration_checkpoint;
9f421827 3189 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3190 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3191
3192 return update_memory_size;
3193}
3194
c17608ea
AK		 3195static void imsm_update_metadata_locally(struct supertype *st,
3196 void *buf, int len);
3197
687629c2
AK		 3198/*******************************************************************************
3199 * Function: write_imsm_migr_rec
3200 * Description: Function writes imsm migration record
3201 * (at the last sector of disk)
3202 * Parameters:
3203 * super : imsm internal array info
3204 * Returns:
3205 * 0 : success
3206 * -1 : if fail
3207 ******************************************************************************/
3208static int write_imsm_migr_rec(struct supertype *st)
3209{
3210 struct intel_super *super = st->sb;
de44e46f 3211 unsigned int sector_size = super->sector_size;
687629c2
AK		 3212 unsigned long long dsize;
3213 char nm[30];
3214 int fd = -1;
3215 int retval = -1;
3216 struct dl *sd;
c17608ea
AK		 3217 int len;
3218 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3219 struct imsm_dev *dev;
594dc1b8 3220 struct imsm_map *map;
3136abe5
AK		 3221
3222 /* find map under migration */
3223 dev = imsm_get_device_during_migration(super);
3224 /* if no migration, write buffer anyway to clear migr_record
3225 * on disk based on first available device
3226 */
3227 if (dev == NULL)
3228 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3229 super->current_vol);
3230
44bfe6df 3231 map = get_imsm_map(dev, MAP_0);
687629c2 3232
de44e46f
PB		 3233 if (sector_size == 4096)
3234 convert_to_4k_imsm_migr_rec(super);
687629c2 3235 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3236 int slot = -1;
3136abe5
AK		 3237
3238 /* skip failed and spare devices */
3239 if (sd->index < 0)
3240 continue;
687629c2 3241 /* write to 2 first slots only */
3136abe5
AK		 3242 if (map)
3243 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3244 if (map == NULL || slot > 1 || slot < 0)
687629c2 3245 continue;
3136abe5 3246
687629c2
AK		 3247 sprintf(nm, "%d:%d", sd->major, sd->minor);
3248 fd = dev_open(nm, O_RDWR);
3249 if (fd < 0)
3250 continue;
3251 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3252 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3253 SEEK_SET) < 0) {
e7b84f9d
N		 3254 pr_err("Cannot seek to anchor block: %s\n",
3255 strerror(errno));
687629c2
AK		 3256 goto out;
3257 }
466070ad 3258 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3259 MIGR_REC_BUF_SECTORS*sector_size) !=
3260 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3261 pr_err("Cannot write migr record block: %s\n",
3262 strerror(errno));
687629c2
AK		 3263 goto out;
3264 }
3265 close(fd);
3266 fd = -1;
3267 }
de44e46f
PB		 3268 if (sector_size == 4096)
3269 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3270 /* update checkpoint information in metadata */
3271 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3272 if (len <= 0) {
3273 dprintf("imsm: Cannot prepare update\n");
3274 goto out;
3275 }
3276 /* update metadata locally */
3277 imsm_update_metadata_locally(st, u, len);
3278 /* and possibly remotely */
3279 if (st->update_tail) {
3280 append_metadata_update(st, u, len);
3281 /* during reshape we do all work inside metadata handler
3282 * manage_reshape(), so metadata update has to be triggered
3283 * insida it
3284 */
3285 flush_metadata_updates(st);
3286 st->update_tail = &st->updates;
3287 } else
3288 free(u);
687629c2
AK		 3289
3290 retval = 0;
3291 out:
3292 if (fd >= 0)
3293 close(fd);
3294 return retval;
3295}
3296
e2962bfc
AK		 3297/* spare/missing disks activations are not allowe when
3298 * array/container performs reshape operation, because
3299 * all arrays in container works on the same disks set
3300 */
3301int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3302{
3303 int rv = 0;
3304 struct intel_dev *i_dev;
3305 struct imsm_dev *dev;
3306
3307 /* check whole container
3308 */
3309 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3310 dev = i_dev->dev;
3ad25638 3311 if (is_gen_migration(dev)) {
e2962bfc
AK		 3312 /* No repair during any migration in container
3313 */
3314 rv = 1;
3315 break;
3316 }
3317 }
3318 return rv;
3319}
3e684231 3320static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3321 int chunk_size,
f36a9ecd 3322 unsigned int sector_size,
c41e00b2
AK		 3323 unsigned long long component_size)
3324{
3e684231 3325 unsigned int component_size_alignment;
c41e00b2 3326
3e684231 3327 /* check component size alignment
c41e00b2 3328 */
3e684231 3329 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3330
3e684231 3331 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3332 level, chunk_size, component_size,
3e684231 3333 component_size_alignment);
c41e00b2 3334
3e684231
MZ		 3335 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3336 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3337 component_size);
3e684231 3338 component_size -= component_size_alignment;
1ade5cc1 3339 dprintf_cont("to %llu (%i).\n",
3e684231 3340 component_size, component_size_alignment);
c41e00b2
AK		 3341 }
3342
3343 return component_size;
3344}
e2962bfc 3345
2432ce9b
AP		 3346static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3347{
3348 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3349 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3350
3351 return pba_of_lba0(map) +
3352 (num_data_stripes(map) * map->blocks_per_strip);
3353}
3354
a5d85af7 3355static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3356{
3357 struct intel_super *super = st->sb;
c47b0ff6 3358 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3359 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3360 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3361 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3362 struct imsm_map *map_to_analyse = map;
efb30e7f 3363 struct dl *dl;
a5d85af7 3364 int map_disks = info->array.raid_disks;
bf5a934a 3365
95eeceeb 3366 memset(info, 0, sizeof(*info));
b335e593
AK		 3367 if (prev_map)
3368 map_to_analyse = prev_map;
3369
ca0748fa 3370 dl = super->current_disk;
9894ec0d 3371
bf5a934a 3372 info->container_member = super->current_vol;
cd0430a1 3373 info->array.raid_disks = map->num_members;
b335e593 3374 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3375 info->array.layout = imsm_level_to_layout(info->array.level);
3376 info->array.md_minor = -1;
3377 info->array.ctime = 0;
3378 info->array.utime = 0;
b335e593
AK		 3379 info->array.chunk_size =
3380 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3381 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3382 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3383 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3384
3f510843 3385 if (is_gen_migration(dev)) {
3f83228a 3386 info->reshape_active = 1;
b335e593
AK		 3387 info->new_level = get_imsm_raid_level(map);
3388 info->new_layout = imsm_level_to_layout(info->new_level);
3389 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3390 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3391 if (info->delta_disks) {
3392 /* this needs to be applied to every array
3393 * in the container.
3394 */
81219e70 3395 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3396 }
3f83228a
N		 3397 /* We shape information that we give to md might have to be
3398 * modify to cope with md's requirement for reshaping arrays.
3399 * For example, when reshaping a RAID0, md requires it to be
3400 * presented as a degraded RAID4.
3401 * Also if a RAID0 is migrating to a RAID5 we need to specify
3402 * the array as already being RAID5, but the 'before' layout
3403 * is a RAID4-like layout.
3404 */
3405 switch (info->array.level) {
3406 case 0:
3407 switch(info->new_level) {
3408 case 0:
3409 /* conversion is happening as RAID4 */
3410 info->array.level = 4;
3411 info->array.raid_disks += 1;
3412 break;
3413 case 5:
3414 /* conversion is happening as RAID5 */
3415 info->array.level = 5;
3416 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3417 info->delta_disks -= 1;
3418 break;
3419 default:
3420 /* FIXME error message */
3421 info->array.level = UnSet;
3422 break;
3423 }
3424 break;
3425 }
b335e593
AK		 3426 } else {
3427 info->new_level = UnSet;
3428 info->new_layout = UnSet;
3429 info->new_chunk = info->array.chunk_size;
3f83228a 3430 info->delta_disks = 0;
b335e593 3431 }
ca0748fa 3432
efb30e7f
DW		 3433 if (dl) {
3434 info->disk.major = dl->major;
3435 info->disk.minor = dl->minor;
ca0748fa 3436 info->disk.number = dl->index;
656b6b5a
N		 3437 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3438 dl->index);
efb30e7f 3439 }
bf5a934a 3440
5551b113 3441 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3442 info->component_size = calc_component_size(map, dev);
3e684231 3443 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3444 info->array.level,
3445 info->array.chunk_size,
f36a9ecd 3446 super->sector_size,
c41e00b2 3447 info->component_size);
5e46202e 3448 info->bb.supported = 1;
139dae11 3449
301406c9 3450 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3451 info->recovery_start = MaxSector;
bf5a934a 3452
c2462068
PB		 3453 if (info->array.level == 5 &&
3454 (dev->rwh_policy == RWH_DISTRIBUTED ||
3455 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3456 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3457 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3458 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3459 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3460 else
3461 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3462 >> 9;
2432ce9b
AP		 3463 } else if (info->array.level <= 0) {
3464 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3465 } else {
3466 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3467 }
3468
d2e6d5d6 3469 info->reshape_progress = 0;
b6796ce1 3470 info->resync_start = MaxSector;
b9172665 3471 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3472 !(info->array.state & 1)) &&
b9172665 3473 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3474 info->resync_start = 0;
b6796ce1
AK		 3475 }
3476 if (dev->vol.migr_state) {
1e5c6983
DW		 3477 switch (migr_type(dev)) {
3478 case MIGR_REPAIR:
3479 case MIGR_INIT: {
c47b0ff6
AK		 3480 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3481 dev);
1e5c6983
DW		 3482 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3483
3484 info->resync_start = blocks_per_unit * units;
3485 break;
3486 }
d2e6d5d6 3487 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3488 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3489 dev);
9f421827 3490 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3491 unsigned long long array_blocks;
3492 int used_disks;
d2e6d5d6 3493
befb629b
AK		 3494 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3495 (units <
9f421827 3496 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3497 (super->migr_rec->rec_status ==
3498 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3499 units++;
3500
d2e6d5d6 3501 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3502
7a862a02 3503 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3504 (unsigned long long)units,
3505 (unsigned long long)blocks_per_unit,
3506 info->reshape_progress);
75156c46 3507
9529d343 3508 used_disks = imsm_num_data_members(prev_map);
75156c46 3509 if (used_disks > 0) {
44490938 3510 array_blocks = per_dev_array_size(map) *
75156c46 3511 used_disks;
b53bfba6
TM		 3512 info->custom_array_size =
3513 round_size_to_mb(array_blocks,
3514 used_disks);
3515
75156c46 3516 }
d2e6d5d6 3517 }
1e5c6983
DW		 3518 case MIGR_VERIFY:
3519 /* we could emulate the checkpointing of
3520 * 'sync_action=check' migrations, but for now
3521 * we just immediately complete them
3522 */
3523 case MIGR_REBUILD:
3524 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3525 case MIGR_STATE_CHANGE:
3526 /* FIXME handle other migrations */
3527 default:
3528 /* we are not dirty, so... */
3529 info->resync_start = MaxSector;
3530 }
b6796ce1 3531 }
301406c9
DW		 3532
3533 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3534 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3535
f35f2525
N		 3536 info->array.major_version = -1;
3537 info->array.minor_version = -2;
4dd2df09 3538 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3539 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3540 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3541
3542 if (dmap) {
3543 int i, j;
3544 for (i=0; i<map_disks; i++) {
3545 dmap[i] = 0;
3546 if (i < info->array.raid_disks) {
3547 struct imsm_disk *dsk;
238c0a71 3548 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3549 dsk = get_imsm_disk(super, j);
3550 if (dsk && (dsk->status & CONFIGURED_DISK))
3551 dmap[i] = 1;
3552 }
3553 }
3554 }
81ac8b4d 3555}
bf5a934a 3556
3b451610
AK		 3557static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3558 int failed, int look_in_map);
3559
3560static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3561 int look_in_map);
3562
3563static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3564{
3565 if (is_gen_migration(dev)) {
3566 int failed;
3567 __u8 map_state;
3568 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3569
3570 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3571 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3572 if (map2->map_state != map_state) {
3573 map2->map_state = map_state;
3574 super->updates_pending++;
3575 }
3576 }
3577}
97b4d0e9
DW		 3578
3579static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3580{
3581 struct dl *d;
3582
3583 for (d = super->missing; d; d = d->next)
3584 if (d->index == index)
3585 return &d->disk;
3586 return NULL;
3587}
3588
a5d85af7 3589static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3590{
3591 struct intel_super *super = st->sb;
4f5bc454 3592 struct imsm_disk *disk;
a5d85af7 3593 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3594 int max_enough = -1;
3595 int i;
3596 struct imsm_super *mpb;
4f5bc454 3597
bf5a934a 3598 if (super->current_vol >= 0) {
a5d85af7 3599 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3600 return;
3601 }
95eeceeb 3602 memset(info, 0, sizeof(*info));
d23fe947
DW		 3603
3604 /* Set raid_disks to zero so that Assemble will always pull in valid
3605 * spares
3606 */
3607 info->array.raid_disks = 0;
cdddbdbc
DW		 3608 info->array.level = LEVEL_CONTAINER;
3609 info->array.layout = 0;
3610 info->array.md_minor = -1;
1011e834 3611 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3612 info->array.utime = 0;
3613 info->array.chunk_size = 0;
3614
3615 info->disk.major = 0;
3616 info->disk.minor = 0;
cdddbdbc 3617 info->disk.raid_disk = -1;
c2c087e6 3618 info->reshape_active = 0;
f35f2525
N		 3619 info->array.major_version = -1;
3620 info->array.minor_version = -2;
c2c087e6 3621 strcpy(info->text_version, "imsm");
a67dd8cc 3622 info->safe_mode_delay = 0;
c2c087e6
DW		 3623 info->disk.number = -1;
3624 info->disk.state = 0;
c5afc314 3625 info->name[0] = 0;
921d9e16 3626 info->recovery_start = MaxSector;
3ad25638 3627 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3628 info->bb.supported = 1;
c2c087e6 3629
97b4d0e9 3630 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3631 mpb = super->anchor;
b7d81a38 3632 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3633
ab3cb6b3
N		 3634 for (i = 0; i < mpb->num_raid_devs; i++) {
3635 struct imsm_dev *dev = get_imsm_dev(super, i);
3636 int failed, enough, j, missing = 0;
3637 struct imsm_map *map;
3638 __u8 state;
97b4d0e9 3639
3b451610
AK		 3640 failed = imsm_count_failed(super, dev, MAP_0);
3641 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3642 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3643
3644 /* any newly missing disks?
3645 * (catches single-degraded vs double-degraded)
3646 */
3647 for (j = 0; j < map->num_members; j++) {
238c0a71 3648 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3649 __u32 idx = ord_to_idx(ord);
3650
20dc76d1
MT		 3651 if (super->disks && super->disks->index == (int)idx)
3652 info->disk.raid_disk = j;
3653
ab3cb6b3
N		 3654 if (!(ord & IMSM_ORD_REBUILD) &&
3655 get_imsm_missing(super, idx)) {
3656 missing = 1;
3657 break;
3658 }
97b4d0e9 3659 }
ab3cb6b3
N		 3660
3661 if (state == IMSM_T_STATE_FAILED)
3662 enough = -1;
3663 else if (state == IMSM_T_STATE_DEGRADED &&
3664 (state != map->map_state || missing))
3665 enough = 0;
3666 else /* we're normal, or already degraded */
3667 enough = 1;
d2bde6d3
AK		 3668 if (is_gen_migration(dev) && missing) {
3669 /* during general migration we need all disks
3670 * that process is running on.
3671 * No new missing disk is allowed.
3672 */
3673 max_enough = -1;
3674 enough = -1;
3675 /* no more checks necessary
3676 */
3677 break;
3678 }
ab3cb6b3
N		 3679 /* in the missing/failed disk case check to see
3680 * if at least one array is runnable
3681 */
3682 max_enough = max(max_enough, enough);
3683 }
1ade5cc1 3684 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3685 info->container_enough = max_enough;
97b4d0e9 3686
4a04ec6c 3687 if (super->disks) {
14e8215b
DW		 3688 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3689
b9f594fe 3690 disk = &super->disks->disk;
5551b113 3691 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3692 info->component_size = reserved;
25ed7e59 3693 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3694 /* we don't change info->disk.raid_disk here because
3695 * this state will be finalized in mdmon after we have
3696 * found the 'most fresh' version of the metadata
3697 */
25ed7e59 3698 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3699 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3700 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3701 }
a575e2a7
DW		 3702
3703 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3704 * ->compare_super may have updated the 'num_raid_devs' field for spares
3705 */
3706 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3707 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3708 else
3709 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3710
3711 /* I don't know how to compute 'map' on imsm, so use safe default */
3712 if (map) {
3713 int i;
3714 for (i = 0; i < map_disks; i++)
3715 map[i] = 1;
3716 }
3717
cdddbdbc
DW		 3718}
3719
5c4cd5da
AC		 3720/* allocates memory and fills disk in mdinfo structure
3721 * for each disk in array */
3722struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3723{
594dc1b8 3724 struct mdinfo *mddev;
5c4cd5da
AC		 3725 struct intel_super *super = st->sb;
3726 struct imsm_disk *disk;
3727 int count = 0;
3728 struct dl *dl;
3729 if (!super || !super->disks)
3730 return NULL;
3731 dl = super->disks;
503975b9 3732 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3733 while (dl) {
3734 struct mdinfo *tmp;
3735 disk = &dl->disk;
503975b9 3736 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3737 if (mddev->devs)
3738 tmp->next = mddev->devs;
3739 mddev->devs = tmp;
3740 tmp->disk.number = count++;
3741 tmp->disk.major = dl->major;
3742 tmp->disk.minor = dl->minor;
3743 tmp->disk.state = is_configured(disk) ?
3744 (1 << MD_DISK_ACTIVE) : 0;
3745 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3746 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3747 tmp->disk.raid_disk = -1;
3748 dl = dl->next;
3749 }
3750 return mddev;
3751}
3752
cdddbdbc
DW		 3753static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3754 char *update, char *devname, int verbose,
3755 int uuid_set, char *homehost)
3756{
f352c545
DW		 3757 /* For 'assemble' and 'force' we need to return non-zero if any
3758 * change was made. For others, the return value is ignored.
3759 * Update options are:
3760 * force-one : This device looks a bit old but needs to be included,
3761 * update age info appropriately.
3762 * assemble: clear any 'faulty' flag to allow this device to
3763 * be assembled.
3764 * force-array: Array is degraded but being forced, mark it clean
3765 * if that will be needed to assemble it.
3766 *
3767 * newdev: not used ????
3768 * grow: Array has gained a new device - this is currently for
3769 * linear only
3770 * resync: mark as dirty so a resync will happen.
3771 * name: update the name - preserving the homehost
6e46bf34 3772 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3773 *
3774 * Following are not relevant for this imsm:
3775 * sparc2.2 : update from old dodgey metadata
3776 * super-minor: change the preferred_minor number
3777 * summaries: update redundant counters.
f352c545
DW		 3778 * homehost: update the recorded homehost
3779 * _reshape_progress: record new reshape_progress position.
3780 */
6e46bf34
DW		 3781 int rv = 1;
3782 struct intel_super *super = st->sb;
3783 struct imsm_super *mpb;
f352c545 3784
6e46bf34
DW		 3785 /* we can only update container info */
3786 if (!super || super->current_vol >= 0 || !super->anchor)
3787 return 1;
3788
3789 mpb = super->anchor;
3790
81a5b4f5
N		 3791 if (strcmp(update, "uuid") == 0) {
3792 /* We take this to mean that the family_num should be updated.
3793 * However that is much smaller than the uuid so we cannot really
3794 * allow an explicit uuid to be given. And it is hard to reliably
3795 * know if one was.
3796 * So if !uuid_set we know the current uuid is random and just used
3797 * the first 'int' and copy it to the other 3 positions.
3798 * Otherwise we require the 4 'int's to be the same as would be the
3799 * case if we are using a random uuid. So an explicit uuid will be
3800 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3801 */
81a5b4f5
N		 3802 if (!uuid_set) {
3803 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3804 rv = 0;
81a5b4f5
N		 3805 } else {
3806 if (info->uuid[0] != info->uuid[1] ||
3807 info->uuid[1] != info->uuid[2] ||
3808 info->uuid[2] != info->uuid[3])
3809 rv = -1;
3810 else
3811 rv = 0;
6e46bf34 3812 }
81a5b4f5
N		 3813 if (rv == 0)
3814 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3815 } else if (strcmp(update, "assemble") == 0)
3816 rv = 0;
3817 else
1e2b2765 3818 rv = -1;
f352c545 3819
6e46bf34
DW		 3820 /* successful update? recompute checksum */
3821 if (rv == 0)
3822 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3823
3824 return rv;
cdddbdbc
DW		 3825}
3826
c2c087e6 3827static size_t disks_to_mpb_size(int disks)
cdddbdbc 3828{
c2c087e6 3829 size_t size;
cdddbdbc 3830
c2c087e6
DW		 3831 size = sizeof(struct imsm_super);
3832 size += (disks - 1) * sizeof(struct imsm_disk);
3833 size += 2 * sizeof(struct imsm_dev);
3834 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3835 size += (4 - 2) * sizeof(struct imsm_map);
3836 /* 4 possible disk_ord_tbl's */
3837 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3838 /* maximum bbm log */
3839 size += sizeof(struct bbm_log);
c2c087e6
DW		 3840
3841 return size;
3842}
3843
387fcd59
N		 3844static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3845 unsigned long long data_offset)
c2c087e6
DW		 3846{
3847 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3848 return 0;
3849
3850 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3851}
3852
ba2de7ba
DW		 3853static void free_devlist(struct intel_super *super)
3854{
3855 struct intel_dev *dv;
3856
3857 while (super->devlist) {
3858 dv = super->devlist->next;
3859 free(super->devlist->dev);
3860 free(super->devlist);
3861 super->devlist = dv;
3862 }
3863}
3864
3865static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3866{
3867 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3868}
3869
cdddbdbc
DW		 3870static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3871{
3872 /*
3873 * return:
3874 * 0 same, or first was empty, and second was copied
3875 * 1 second had wrong number
3876 * 2 wrong uuid
3877 * 3 wrong other info
3878 */
3879 struct intel_super *first = st->sb;
3880 struct intel_super *sec = tst->sb;
3881
5d500228
N		 3882 if (!first) {
3883 st->sb = tst->sb;
3884 tst->sb = NULL;
3885 return 0;
3886 }
8603ea6f
LM		 3887 /* in platform dependent environment test if the disks
3888 * use the same Intel hba
cb8f6859 3889 * If not on Intel hba at all, allow anything.
8603ea6f 3890 */
6b781d33
AP		 3891 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3892 if (first->hba->type != sec->hba->type) {
8603ea6f 3893 fprintf(stderr,
6b781d33
AP		 3894 "HBAs of devices do not match %s != %s\n",
3895 get_sys_dev_type(first->hba->type),
3896 get_sys_dev_type(sec->hba->type));
3897 return 3;
3898 }
3899 if (first->orom != sec->orom) {
3900 fprintf(stderr,
3901 "HBAs of devices do not match %s != %s\n",
3902 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3903 return 3;
3904 }
3905 }
cdddbdbc 3906
d23fe947
DW		 3907 /* if an anchor does not have num_raid_devs set then it is a free
3908 * floating spare
3909 */
3910 if (first->anchor->num_raid_devs > 0 &&
3911 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3912 /* Determine if these disks might ever have been
3913 * related. Further disambiguation can only take place
3914 * in load_super_imsm_all
3915 */
3916 __u32 first_family = first->anchor->orig_family_num;
3917 __u32 sec_family = sec->anchor->orig_family_num;
3918
f796af5d
DW		 3919 if (memcmp(first->anchor->sig, sec->anchor->sig,
3920 MAX_SIGNATURE_LENGTH) != 0)
3921 return 3;
3922
a2b97981
DW		 3923 if (first_family == 0)
3924 first_family = first->anchor->family_num;
3925 if (sec_family == 0)
3926 sec_family = sec->anchor->family_num;
3927
3928 if (first_family != sec_family)
d23fe947 3929 return 3;
f796af5d 3930
d23fe947 3931 }
cdddbdbc 3932
3e372e5a
DW		 3933 /* if 'first' is a spare promote it to a populated mpb with sec's
3934 * family number
3935 */
3936 if (first->anchor->num_raid_devs == 0 &&
3937 sec->anchor->num_raid_devs > 0) {
78d30f94 3938 int i;
ba2de7ba
DW		 3939 struct intel_dev *dv;
3940 struct imsm_dev *dev;
78d30f94
DW		 3941
3942 /* we need to copy raid device info from sec if an allocation
3943 * fails here we don't associate the spare
3944 */
3945 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3946 dv = xmalloc(sizeof(*dv));
3947 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3948 dv->dev = dev;
3949 dv->index = i;
3950 dv->next = first->devlist;
3951 first->devlist = dv;
78d30f94 3952 }
709743c5 3953 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3954 /* allocation failure */
3955 free_devlist(first);
e12b3daa 3956 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3957 return 3;
78d30f94 3958 }
3e372e5a 3959 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3960 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3961 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3962 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3963 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3964 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3965 }
3966
cdddbdbc
DW		 3967 return 0;
3968}
3969
0030e8d6
DW		 3970static void fd2devname(int fd, char *name)
3971{
3972 struct stat st;
3973 char path[256];
33a6535d 3974 char dname[PATH_MAX];
0030e8d6
DW		 3975 char *nm;
3976 int rv;
3977
3978 name[0] = '\0';
3979 if (fstat(fd, &st) != 0)
3980 return;
3981 sprintf(path, "/sys/dev/block/%d:%d",
3982 major(st.st_rdev), minor(st.st_rdev));
3983
9cf014ec 3984 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3985 if (rv <= 0)
3986 return;
9587c373 3987
0030e8d6
DW		 3988 dname[rv] = '\0';
3989 nm = strrchr(dname, '/');
7897de29
JS		 3990 if (nm) {
3991 nm++;
3992 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3993 }
0030e8d6
DW		 3994}
3995
21e9380b
AP		 3996static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3997{
3998 char path[60];
3999 char *name = fd2kname(fd);
4000
4001 if (!name)
4002 return 1;
4003
4004 if (strncmp(name, "nvme", 4) != 0)
4005 return 1;
4006
4007 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
4008
4009 return load_sys(path, buf, buf_len);
4010}
4011
cdddbdbc
DW		 4012extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
4013
4014static int imsm_read_serial(int fd, char *devname,
4015 __u8 serial[MAX_RAID_SERIAL_LEN])
4016{
21e9380b 4017 char buf[50];
cdddbdbc 4018 int rv;
1f24f035 4019 int len;
316e2bf4
DW		 4020 char *dest;
4021 char *src;
21e9380b
AP		 4022 unsigned int i;
4023
4024 memset(buf, 0, sizeof(buf));
cdddbdbc 4025
21e9380b 4026 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 4027
21e9380b
AP		 4028 if (rv)
4029 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 4030
40ebbb9c 4031 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 4032 memset(serial, 0, MAX_RAID_SERIAL_LEN);
4033 fd2devname(fd, (char *) serial);
0030e8d6
DW		 4034 return 0;
4035 }
4036
cdddbdbc
DW		 4037 if (rv != 0) {
4038 if (devname)
e7b84f9d
N		 4039 pr_err("Failed to retrieve serial for %s\n",
4040 devname);
cdddbdbc
DW		 4041 return rv;
4042 }
4043
316e2bf4
DW		 4044 /* trim all whitespace and non-printable characters and convert
4045 * ':' to ';'
4046 */
21e9380b
AP		 4047 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4048 src = &buf[i];
316e2bf4
DW		 4049 if (*src > 0x20) {
4050 /* ':' is reserved for use in placeholder serial
4051 * numbers for missing disks
4052 */
4053 if (*src == ':')
4054 *dest++ = ';';
4055 else
4056 *dest++ = *src;
4057 }
4058 }
21e9380b
AP		 4059 len = dest - buf;
4060 dest = buf;
316e2bf4
DW		 4061
4062 /* truncate leading characters */
4063 if (len > MAX_RAID_SERIAL_LEN) {
4064 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 4065 len = MAX_RAID_SERIAL_LEN;
316e2bf4 4066 }
5c3db629 4067
5c3db629 4068 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 4069 memcpy(serial, dest, len);
cdddbdbc
DW		 4070
4071 return 0;
4072}
4073
1f24f035
DW		 4074static int serialcmp(__u8 *s1, __u8 *s2)
4075{
4076 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4077}
4078
4079static void serialcpy(__u8 *dest, __u8 *src)
4080{
4081 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4082}
4083
54c2c1ea
DW		 4084static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4085{
4086 struct dl *dl;
4087
4088 for (dl = super->disks; dl; dl = dl->next)
4089 if (serialcmp(dl->serial, serial) == 0)
4090 break;
4091
4092 return dl;
4093}
4094
a2b97981
DW		 4095static struct imsm_disk *
4096__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4097{
4098 int i;
4099
4100 for (i = 0; i < mpb->num_disks; i++) {
4101 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4102
4103 if (serialcmp(disk->serial, serial) == 0) {
4104 if (idx)
4105 *idx = i;
4106 return disk;
4107 }
4108 }
4109
4110 return NULL;
4111}
4112
cdddbdbc
DW		 4113static int
4114load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4115{
a2b97981 4116 struct imsm_disk *disk;
cdddbdbc
DW		 4117 struct dl *dl;
4118 struct stat stb;
cdddbdbc 4119 int rv;
a2b97981 4120 char name[40];
d23fe947
DW		 4121 __u8 serial[MAX_RAID_SERIAL_LEN];
4122
4123 rv = imsm_read_serial(fd, devname, serial);
4124
4125 if (rv != 0)
4126 return 2;
4127
503975b9 4128 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4129
a2b97981
DW		 4130 fstat(fd, &stb);
4131 dl->major = major(stb.st_rdev);
4132 dl->minor = minor(stb.st_rdev);
4133 dl->next = super->disks;
4134 dl->fd = keep_fd ? fd : -1;
4135 assert(super->disks == NULL);
4136 super->disks = dl;
4137 serialcpy(dl->serial, serial);
4138 dl->index = -2;
4139 dl->e = NULL;
4140 fd2devname(fd, name);
4141 if (devname)
503975b9 4142 dl->devname = xstrdup(devname);
a2b97981 4143 else
503975b9 4144 dl->devname = xstrdup(name);
cdddbdbc 4145
d23fe947 4146 /* look up this disk's index in the current anchor */
a2b97981
DW		 4147 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4148 if (disk) {
4149 dl->disk = *disk;
4150 /* only set index on disks that are a member of a
4151 * populated contianer, i.e. one with raid_devs
4152 */
4153 if (is_failed(&dl->disk))
3f6efecc 4154 dl->index = -2;
2432ce9b 4155 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4156 dl->index = -1;
3f6efecc
DW		 4157 }
4158
949c47a0
DW		 4159 return 0;
4160}
4161
0c046afd
DW		 4162/* When migrating map0 contains the 'destination' state while map1
4163 * contains the current state. When not migrating map0 contains the
4164 * current state. This routine assumes that map[0].map_state is set to
4165 * the current array state before being called.
4166 *
4167 * Migration is indicated by one of the following states
4168 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4169 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4170 * map1state=unitialized)
1484e727 4171 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4172 * map1state=normal)
e3bba0e0 4173 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4174 * map1state=degraded)
8e59f3d8
AK		 4175 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4176 * map1state=normal)
0c046afd 4177 */
8e59f3d8
AK		 4178static void migrate(struct imsm_dev *dev, struct intel_super *super,
4179 __u8 to_state, int migr_type)
3393c6af 4180{
0c046afd 4181 struct imsm_map *dest;
238c0a71 4182 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4183
0c046afd 4184 dev->vol.migr_state = 1;
1484e727 4185 set_migr_type(dev, migr_type);
f8f603f1 4186 dev->vol.curr_migr_unit = 0;
238c0a71 4187 dest = get_imsm_map(dev, MAP_1);
0c046afd 4188
0556e1a2 4189 /* duplicate and then set the target end state in map[0] */
3393c6af 4190 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4191 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4192 __u32 ord;
4193 int i;
4194
4195 for (i = 0; i < src->num_members; i++) {
4196 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4197 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4198 }
4199 }
4200
8e59f3d8
AK		 4201 if (migr_type == MIGR_GEN_MIGR)
4202 /* Clear migration record */
4203 memset(super->migr_rec, 0, sizeof(struct migr_record));
4204
0c046afd 4205 src->map_state = to_state;
949c47a0 4206}
f8f603f1 4207
809da78e
AK		 4208static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4209 __u8 map_state)
f8f603f1 4210{
238c0a71
AK		 4211 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4212 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4213 MAP_0 : MAP_1);
28bce06f 4214 int i, j;
0556e1a2
DW		 4215
4216 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4217 * completed in the last migration.
4218 *
28bce06f 4219 * FIXME add support for raid-level-migration
0556e1a2 4220 */
089f9d79
JS		 4221 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4222 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4223 /* when final map state is other than expected
4224 * merge maps (not for migration)
4225 */
4226 int failed;
4227
4228 for (i = 0; i < prev->num_members; i++)
4229 for (j = 0; j < map->num_members; j++)
4230 /* during online capacity expansion
4231 * disks position can be changed
4232 * if takeover is used
4233 */
4234 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4235 ord_to_idx(prev->disk_ord_tbl[i])) {
4236 map->disk_ord_tbl[j] |=
4237 prev->disk_ord_tbl[i];
4238 break;
4239 }
4240 failed = imsm_count_failed(super, dev, MAP_0);
4241 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4242 }
f8f603f1
DW		 4243
4244 dev->vol.migr_state = 0;
ea672ee1 4245 set_migr_type(dev, 0);
f8f603f1
DW		 4246 dev->vol.curr_migr_unit = 0;
4247 map->map_state = map_state;
4248}
949c47a0
DW		 4249
4250static int parse_raid_devices(struct intel_super *super)
4251{
4252 int i;
4253 struct imsm_dev *dev_new;
4d7b1503 4254 size_t len, len_migr;
401d313b 4255 size_t max_len = 0;
4d7b1503
DW		 4256 size_t space_needed = 0;
4257 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4258
4259 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4260 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4261 struct intel_dev *dv;
949c47a0 4262
4d7b1503
DW		 4263 len = sizeof_imsm_dev(dev_iter, 0);
4264 len_migr = sizeof_imsm_dev(dev_iter, 1);
4265 if (len_migr > len)
4266 space_needed += len_migr - len;
ca9de185 4267
503975b9 4268 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4269 if (max_len < len_migr)
4270 max_len = len_migr;
4271 if (max_len > len_migr)
4272 space_needed += max_len - len_migr;
503975b9 4273 dev_new = xmalloc(max_len);
949c47a0 4274 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4275 dv->dev = dev_new;
4276 dv->index = i;
4277 dv->next = super->devlist;
4278 super->devlist = dv;
949c47a0 4279 }
cdddbdbc 4280
4d7b1503
DW		 4281 /* ensure that super->buf is large enough when all raid devices
4282 * are migrating
4283 */
4284 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4285 void *buf;
4286
f36a9ecd
PB		 4287 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4288 super->sector_size);
4289 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4290 return 1;
4291
1f45a8ad
DW		 4292 memcpy(buf, super->buf, super->len);
4293 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4294 free(super->buf);
4295 super->buf = buf;
4296 super->len = len;
4297 }
ca9de185 4298
bbab0940
TM		 4299 super->extra_space += space_needed;
4300
cdddbdbc
DW		 4301 return 0;
4302}
4303
e2f41b2c
AK		 4304/*******************************************************************************
4305 * Function: check_mpb_migr_compatibility
4306 * Description: Function checks for unsupported migration features:
4307 * - migration optimization area (pba_of_lba0)
4308 * - descending reshape (ascending_migr)
4309 * Parameters:
4310 * super : imsm metadata information
4311 * Returns:
4312 * 0 : migration is compatible
4313 * -1 : migration is not compatible
4314 ******************************************************************************/
4315int check_mpb_migr_compatibility(struct intel_super *super)
4316{
4317 struct imsm_map *map0, *map1;
4318 struct migr_record *migr_rec = super->migr_rec;
4319 int i;
4320
4321 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4322 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4323
4324 if (dev_iter &&
4325 dev_iter->vol.migr_state == 1 &&
4326 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4327 /* This device is migrating */
238c0a71
AK		 4328 map0 = get_imsm_map(dev_iter, MAP_0);
4329 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4330 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4331 /* migration optimization area was used */
4332 return -1;
fc54fe7a
JS		 4333 if (migr_rec->ascending_migr == 0 &&
4334 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4335 /* descending reshape not supported yet */
4336 return -1;
4337 }
4338 }
4339 return 0;
4340}
4341
d23fe947 4342static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4343
cdddbdbc 4344/* load_imsm_mpb - read matrix metadata
f2f5c343 4345 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4346 */
4347static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4348{
4349 unsigned long long dsize;
cdddbdbc 4350 unsigned long long sectors;
f36a9ecd 4351 unsigned int sector_size = super->sector_size;
cdddbdbc 4352 struct stat;
6416d527 4353 struct imsm_super *anchor;
cdddbdbc
DW		 4354 __u32 check_sum;
4355
cdddbdbc 4356 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4357 if (dsize < 2*sector_size) {
64436f06 4358 if (devname)
e7b84f9d
N		 4359 pr_err("%s: device to small for imsm\n",
4360 devname);
64436f06
N		 4361 return 1;
4362 }
cdddbdbc 4363
f36a9ecd 4364 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4365 if (devname)
e7b84f9d
N		 4366 pr_err("Cannot seek to anchor block on %s: %s\n",
4367 devname, strerror(errno));
cdddbdbc
DW		 4368 return 1;
4369 }
4370
f36a9ecd 4371 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4372 if (devname)
7a862a02 4373 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4374 return 1;
4375 }
466070ad 4376 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4377 if (devname)
e7b84f9d
N		 4378 pr_err("Cannot read anchor block on %s: %s\n",
4379 devname, strerror(errno));
6416d527 4380 free(anchor);
cdddbdbc
DW		 4381 return 1;
4382 }
4383
6416d527 4384 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4385 if (devname)
e7b84f9d 4386 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4387 free(anchor);
cdddbdbc
DW		 4388 return 2;
4389 }
4390
d23fe947 4391 __free_imsm(super, 0);
f2f5c343
LM		 4392 /* reload capability and hba */
4393
4394 /* capability and hba must be updated with new super allocation */
d424212e 4395 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4396 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4397 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4398 if (devname)
e7b84f9d
N		 4399 pr_err("unable to allocate %zu byte mpb buffer\n",
4400 super->len);
6416d527 4401 free(anchor);
cdddbdbc
DW		 4402 return 2;
4403 }
f36a9ecd 4404 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4405
f36a9ecd 4406 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4407 free(anchor);
8e59f3d8 4408
85337573
AO		 4409 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4410 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4411 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4412 free(super->buf);
4413 return 2;
4414 }
51d83f5d 4415 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4416
949c47a0 4417 if (!sectors) {
ecf45690
DW		 4418 check_sum = __gen_imsm_checksum(super->anchor);
4419 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4420 if (devname)
e7b84f9d
N		 4421 pr_err("IMSM checksum %x != %x on %s\n",
4422 check_sum,
4423 __le32_to_cpu(super->anchor->check_sum),
4424 devname);
ecf45690
DW		 4425 return 2;
4426 }
4427
a2b97981 4428 return 0;
949c47a0 4429 }
cdddbdbc
DW		 4430
4431 /* read the extended mpb */
f36a9ecd 4432 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4433 if (devname)
e7b84f9d
N		 4434 pr_err("Cannot seek to extended mpb on %s: %s\n",
4435 devname, strerror(errno));
cdddbdbc
DW		 4436 return 1;
4437 }
4438
f36a9ecd
PB		 4439 if ((unsigned int)read(fd, super->buf + sector_size,
4440 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4441 if (devname)
e7b84f9d
N		 4442 pr_err("Cannot read extended mpb on %s: %s\n",
4443 devname, strerror(errno));
cdddbdbc
DW		 4444 return 2;
4445 }
4446
949c47a0
DW		 4447 check_sum = __gen_imsm_checksum(super->anchor);
4448 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4449 if (devname)
e7b84f9d
N		 4450 pr_err("IMSM checksum %x != %x on %s\n",
4451 check_sum, __le32_to_cpu(super->anchor->check_sum),
4452 devname);
db575f3b 4453 return 3;
cdddbdbc
DW		 4454 }
4455
a2b97981
DW		 4456 return 0;
4457}
4458
8e59f3d8
AK		 4459static int read_imsm_migr_rec(int fd, struct intel_super *super);
4460
97f81ee2
CA		 4461/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4462static void clear_hi(struct intel_super *super)
4463{
4464 struct imsm_super *mpb = super->anchor;
4465 int i, n;
4466 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4467 return;
4468 for (i = 0; i < mpb->num_disks; ++i) {
4469 struct imsm_disk *disk = &mpb->disk[i];
4470 disk->total_blocks_hi = 0;
4471 }
4472 for (i = 0; i < mpb->num_raid_devs; ++i) {
4473 struct imsm_dev *dev = get_imsm_dev(super, i);
4474 if (!dev)
4475 return;
4476 for (n = 0; n < 2; ++n) {
4477 struct imsm_map *map = get_imsm_map(dev, n);
4478 if (!map)
4479 continue;
4480 map->pba_of_lba0_hi = 0;
4481 map->blocks_per_member_hi = 0;
4482 map->num_data_stripes_hi = 0;
4483 }
4484 }
4485}
4486
a2b97981
DW		 4487static int
4488load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4489{
4490 int err;
4491
4492 err = load_imsm_mpb(fd, super, devname);
4493 if (err)
4494 return err;
f36a9ecd
PB		 4495 if (super->sector_size == 4096)
4496 convert_from_4k(super);
a2b97981
DW		 4497 err = load_imsm_disk(fd, super, devname, keep_fd);
4498 if (err)
4499 return err;
4500 err = parse_raid_devices(super);
8d67477f
TM		 4501 if (err)
4502 return err;
4503 err = load_bbm_log(super);
97f81ee2 4504 clear_hi(super);
a2b97981 4505 return err;
cdddbdbc
DW		 4506}
4507
ae6aad82
DW		 4508static void __free_imsm_disk(struct dl *d)
4509{
4510 if (d->fd >= 0)
4511 close(d->fd);
4512 if (d->devname)
4513 free(d->devname);
0dcecb2e
DW		 4514 if (d->e)
4515 free(d->e);
ae6aad82
DW		 4516 free(d);
4517
4518}
1a64be56 4519
cdddbdbc
DW		 4520static void free_imsm_disks(struct intel_super *super)
4521{
47ee5a45 4522 struct dl *d;
cdddbdbc 4523
47ee5a45
DW		 4524 while (super->disks) {
4525 d = super->disks;
cdddbdbc 4526 super->disks = d->next;
ae6aad82 4527 __free_imsm_disk(d);
cdddbdbc 4528 }
cb82edca
AK		 4529 while (super->disk_mgmt_list) {
4530 d = super->disk_mgmt_list;
4531 super->disk_mgmt_list = d->next;
4532 __free_imsm_disk(d);
4533 }
47ee5a45
DW		 4534 while (super->missing) {
4535 d = super->missing;
4536 super->missing = d->next;
4537 __free_imsm_disk(d);
4538 }
4539
cdddbdbc
DW		 4540}
4541
9ca2c81c 4542/* free all the pieces hanging off of a super pointer */
d23fe947 4543static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4544{
88654014
LM		 4545 struct intel_hba *elem, *next;
4546
9ca2c81c 4547 if (super->buf) {
949c47a0 4548 free(super->buf);
9ca2c81c
DW		 4549 super->buf = NULL;
4550 }
f2f5c343
LM		 4551 /* unlink capability description */
4552 super->orom = NULL;
8e59f3d8
AK		 4553 if (super->migr_rec_buf) {
4554 free(super->migr_rec_buf);
4555 super->migr_rec_buf = NULL;
4556 }
d23fe947
DW		 4557 if (free_disks)
4558 free_imsm_disks(super);
ba2de7ba 4559 free_devlist(super);
88654014
LM		 4560 elem = super->hba;
4561 while (elem) {
4562 if (elem->path)
4563 free((void *)elem->path);
4564 next = elem->next;
4565 free(elem);
4566 elem = next;
88c32bb1 4567 }
8d67477f
TM		 4568 if (super->bbm_log)
4569 free(super->bbm_log);
88654014 4570 super->hba = NULL;
cdddbdbc
DW		 4571}
4572
9ca2c81c
DW		 4573static void free_imsm(struct intel_super *super)
4574{
d23fe947 4575 __free_imsm(super, 1);
928f1424 4576 free(super->bb.entries);
9ca2c81c
DW		 4577 free(super);
4578}
cdddbdbc
DW		 4579
4580static void free_super_imsm(struct supertype *st)
4581{
4582 struct intel_super *super = st->sb;
4583
4584 if (!super)
4585 return;
4586
4587 free_imsm(super);
4588 st->sb = NULL;
4589}
4590
49133e57 4591static struct intel_super *alloc_super(void)
c2c087e6 4592{
503975b9 4593 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4594
503975b9
N		 4595 super->current_vol = -1;
4596 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4597
4598 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4599 sizeof(struct md_bb_entry));
4600 if (!super->bb.entries) {
4601 free(super);
4602 return NULL;
4603 }
4604
c2c087e6
DW		 4605 return super;
4606}
4607
f0f5a016
LM		 4608/*
4609 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4610 */
d424212e 4611static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4612{
4613 struct sys_dev *hba_name;
4614 int rv = 0;
4615
3a30e28e
MT		 4616 if (fd >= 0 && test_partition(fd)) {
4617 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4618 devname);
4619 return 1;
4620 }
089f9d79 4621 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4622 super->orom = NULL;
f0f5a016
LM		 4623 super->hba = NULL;
4624 return 0;
4625 }
4626 hba_name = find_disk_attached_hba(fd, NULL);
4627 if (!hba_name) {
d424212e 4628 if (devname)
e7b84f9d
N		 4629 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4630 devname);
f0f5a016
LM		 4631 return 1;
4632 }
4633 rv = attach_hba_to_super(super, hba_name);
4634 if (rv == 2) {
d424212e
N		 4635 if (devname) {
4636 struct intel_hba *hba = super->hba;
f0f5a016 4637
60f0f54d
PB		 4638 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4639 " but the container is assigned to Intel(R) %s %s (",
d424212e 4640 devname,
614902f6 4641 get_sys_dev_type(hba_name->type),
60f0f54d 4642 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4643 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4644 get_sys_dev_type(super->hba->type),
4645 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4646
f0f5a016
LM		 4647 while (hba) {
4648 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4649 if (hba->next)
4650 fprintf(stderr, ", ");
4651 hba = hba->next;
4652 }
6b781d33 4653 fprintf(stderr, ").\n"
cca67208 4654 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4655 }
f0f5a016
LM		 4656 return 2;
4657 }
6b781d33 4658 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4659 if (!super->orom)
4660 return 3;
614902f6 4661
f0f5a016
LM		 4662 return 0;
4663}
4664
47ee5a45
DW		 4665/* find_missing - helper routine for load_super_imsm_all that identifies
4666 * disks that have disappeared from the system. This routine relies on
4667 * the mpb being uptodate, which it is at load time.
4668 */
4669static int find_missing(struct intel_super *super)
4670{
4671 int i;
4672 struct imsm_super *mpb = super->anchor;
4673 struct dl *dl;
4674 struct imsm_disk *disk;
47ee5a45
DW		 4675
4676 for (i = 0; i < mpb->num_disks; i++) {
4677 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4678 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4679 if (dl)
4680 continue;
47ee5a45 4681
503975b9 4682 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4683 dl->major = 0;
4684 dl->minor = 0;
4685 dl->fd = -1;
503975b9 4686 dl->devname = xstrdup("missing");
47ee5a45
DW		 4687 dl->index = i;
4688 serialcpy(dl->serial, disk->serial);
4689 dl->disk = *disk;
689c9bf3 4690 dl->e = NULL;
47ee5a45
DW		 4691 dl->next = super->missing;
4692 super->missing = dl;
4693 }
4694
4695 return 0;
4696}
4697
a2b97981
DW		 4698static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4699{
4700 struct intel_disk *idisk = disk_list;
4701
4702 while (idisk) {
4703 if (serialcmp(idisk->disk.serial, serial) == 0)
4704 break;
4705 idisk = idisk->next;
4706 }
4707
4708 return idisk;
4709}
4710
4711static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4712 struct intel_super *super,
4713 struct intel_disk **disk_list)
4714{
4715 struct imsm_disk *d = &super->disks->disk;
4716 struct imsm_super *mpb = super->anchor;
4717 int i, j;
4718
4719 for (i = 0; i < tbl_size; i++) {
4720 struct imsm_super *tbl_mpb = table[i]->anchor;
4721 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4722
4723 if (tbl_mpb->family_num == mpb->family_num) {
4724 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4725 dprintf("mpb from %d:%d matches %d:%d\n",
4726 super->disks->major,
a2b97981
DW		 4727 super->disks->minor,
4728 table[i]->disks->major,
4729 table[i]->disks->minor);
4730 break;
4731 }
4732
4733 if (((is_configured(d) && !is_configured(tbl_d)) ||
4734 is_configured(d) == is_configured(tbl_d)) &&
4735 tbl_mpb->generation_num < mpb->generation_num) {
4736 /* current version of the mpb is a
4737 * better candidate than the one in
4738 * super_table, but copy over "cross
4739 * generational" status
4740 */
4741 struct intel_disk *idisk;
4742
1ade5cc1
N		 4743 dprintf("mpb from %d:%d replaces %d:%d\n",
4744 super->disks->major,
a2b97981
DW		 4745 super->disks->minor,
4746 table[i]->disks->major,
4747 table[i]->disks->minor);
4748
4749 idisk = disk_list_get(tbl_d->serial, *disk_list);
4750 if (idisk && is_failed(&idisk->disk))
4751 tbl_d->status |= FAILED_DISK;
4752 break;
4753 } else {
4754 struct intel_disk *idisk;
4755 struct imsm_disk *disk;
4756
4757 /* tbl_mpb is more up to date, but copy
4758 * over cross generational status before
4759 * returning
4760 */
4761 disk = __serial_to_disk(d->serial, mpb, NULL);
4762 if (disk && is_failed(disk))
4763 d->status |= FAILED_DISK;
4764
4765 idisk = disk_list_get(d->serial, *disk_list);
4766 if (idisk) {
4767 idisk->owner = i;
4768 if (disk && is_configured(disk))
4769 idisk->disk.status |= CONFIGURED_DISK;
4770 }
4771
1ade5cc1
N		 4772 dprintf("mpb from %d:%d prefer %d:%d\n",
4773 super->disks->major,
a2b97981
DW		 4774 super->disks->minor,
4775 table[i]->disks->major,
4776 table[i]->disks->minor);
4777
4778 return tbl_size;
4779 }
4780 }
4781 }
4782
4783 if (i >= tbl_size)
4784 table[tbl_size++] = super;
4785 else
4786 table[i] = super;
4787
4788 /* update/extend the merged list of imsm_disk records */
4789 for (j = 0; j < mpb->num_disks; j++) {
4790 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4791 struct intel_disk *idisk;
4792
4793 idisk = disk_list_get(disk->serial, *disk_list);
4794 if (idisk) {
4795 idisk->disk.status |= disk->status;
4796 if (is_configured(&idisk->disk) ||
4797 is_failed(&idisk->disk))
4798 idisk->disk.status &= ~(SPARE_DISK);
4799 } else {
503975b9 4800 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4801 idisk->owner = IMSM_UNKNOWN_OWNER;
4802 idisk->disk = *disk;
4803 idisk->next = *disk_list;
4804 *disk_list = idisk;
4805 }
4806
4807 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4808 idisk->owner = i;
4809 }
4810
4811 return tbl_size;
4812}
4813
4814static struct intel_super *
4815validate_members(struct intel_super *super, struct intel_disk *disk_list,
4816 const int owner)
4817{
4818 struct imsm_super *mpb = super->anchor;
4819 int ok_count = 0;
4820 int i;
4821
4822 for (i = 0; i < mpb->num_disks; i++) {
4823 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4824 struct intel_disk *idisk;
4825
4826 idisk = disk_list_get(disk->serial, disk_list);
4827 if (idisk) {
4828 if (idisk->owner == owner ||
4829 idisk->owner == IMSM_UNKNOWN_OWNER)
4830 ok_count++;
4831 else
1ade5cc1
N		 4832 dprintf("'%.16s' owner %d != %d\n",
4833 disk->serial, idisk->owner,
a2b97981
DW		 4834 owner);
4835 } else {
1ade5cc1
N		 4836 dprintf("unknown disk %x [%d]: %.16s\n",
4837 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4838 disk->serial);
4839 break;
4840 }
4841 }
4842
4843 if (ok_count == mpb->num_disks)
4844 return super;
4845 return NULL;
4846}
4847
4848static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4849{
4850 struct intel_super *s;
4851
4852 for (s = super_list; s; s = s->next) {
4853 if (family_num != s->anchor->family_num)
4854 continue;
e12b3daa 4855 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4856 __le32_to_cpu(family_num), s->disks->devname);
4857 }
4858}
4859
4860static struct intel_super *
4861imsm_thunderdome(struct intel_super **super_list, int len)
4862{
4863 struct intel_super *super_table[len];
4864 struct intel_disk *disk_list = NULL;
4865 struct intel_super *champion, *spare;
4866 struct intel_super *s, **del;
4867 int tbl_size = 0;
4868 int conflict;
4869 int i;
4870
4871 memset(super_table, 0, sizeof(super_table));
4872 for (s = *super_list; s; s = s->next)
4873 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4874
4875 for (i = 0; i < tbl_size; i++) {
4876 struct imsm_disk *d;
4877 struct intel_disk *idisk;
4878 struct imsm_super *mpb = super_table[i]->anchor;
4879
4880 s = super_table[i];
4881 d = &s->disks->disk;
4882
4883 /* 'd' must appear in merged disk list for its
4884 * configuration to be valid
4885 */
4886 idisk = disk_list_get(d->serial, disk_list);
4887 if (idisk && idisk->owner == i)
4888 s = validate_members(s, disk_list, i);
4889 else
4890 s = NULL;
4891
4892 if (!s)
1ade5cc1
N		 4893 dprintf("marking family: %#x from %d:%d offline\n",
4894 mpb->family_num,
a2b97981
DW		 4895 super_table[i]->disks->major,
4896 super_table[i]->disks->minor);
4897 super_table[i] = s;
4898 }
4899
4900 /* This is where the mdadm implementation differs from the Windows
4901 * driver which has no strict concept of a container. We can only
4902 * assemble one family from a container, so when returning a prodigal
4903 * array member to this system the code will not be able to disambiguate
4904 * the container contents that should be assembled ("foreign" versus
4905 * "local"). It requires user intervention to set the orig_family_num
4906 * to a new value to establish a new container. The Windows driver in
4907 * this situation fixes up the volume name in place and manages the
4908 * foreign array as an independent entity.
4909 */
4910 s = NULL;
4911 spare = NULL;
4912 conflict = 0;
4913 for (i = 0; i < tbl_size; i++) {
4914 struct intel_super *tbl_ent = super_table[i];
4915 int is_spare = 0;
4916
4917 if (!tbl_ent)
4918 continue;
4919
4920 if (tbl_ent->anchor->num_raid_devs == 0) {
4921 spare = tbl_ent;
4922 is_spare = 1;
4923 }
4924
4925 if (s && !is_spare) {
4926 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4927 conflict++;
4928 } else if (!s && !is_spare)
4929 s = tbl_ent;
4930 }
4931
4932 if (!s)
4933 s = spare;
4934 if (!s) {
4935 champion = NULL;
4936 goto out;
4937 }
4938 champion = s;
4939
4940 if (conflict)
7a862a02 4941 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4942 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4943
4944 /* collect all dl's onto 'champion', and update them to
4945 * champion's version of the status
4946 */
4947 for (s = *super_list; s; s = s->next) {
4948 struct imsm_super *mpb = champion->anchor;
4949 struct dl *dl = s->disks;
4950
4951 if (s == champion)
4952 continue;
4953
5d7b407a
CA		 4954 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4955
a2b97981
DW		 4956 for (i = 0; i < mpb->num_disks; i++) {
4957 struct imsm_disk *disk;
4958
4959 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4960 if (disk) {
4961 dl->disk = *disk;
4962 /* only set index on disks that are a member of
4963 * a populated contianer, i.e. one with
4964 * raid_devs
4965 */
4966 if (is_failed(&dl->disk))
4967 dl->index = -2;
4968 else if (is_spare(&dl->disk))
4969 dl->index = -1;
4970 break;
4971 }
4972 }
4973
4974 if (i >= mpb->num_disks) {
4975 struct intel_disk *idisk;
4976
4977 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4978 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4979 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4980 dl->index = -1;
4981 else {
4982 dl->index = -2;
4983 continue;
4984 }
4985 }
4986
4987 dl->next = champion->disks;
4988 champion->disks = dl;
4989 s->disks = NULL;
4990 }
4991
4992 /* delete 'champion' from super_list */
4993 for (del = super_list; *del; ) {
4994 if (*del == champion) {
4995 *del = (*del)->next;
4996 break;
4997 } else
4998 del = &(*del)->next;
4999 }
5000 champion->next = NULL;
5001
5002 out:
5003 while (disk_list) {
5004 struct intel_disk *idisk = disk_list;
5005
5006 disk_list = disk_list->next;
5007 free(idisk);
5008 }
5009
5010 return champion;
5011}
5012
9587c373
LM		 5013static int
5014get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 5015static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 5016 int major, int minor, int keep_fd);
ec50f7b6
LM		 5017static int
5018get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5019 int *max, int keep_fd);
5020
cdddbdbc 5021static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 5022 char *devname, struct md_list *devlist,
5023 int keep_fd)
cdddbdbc 5024{
a2b97981
DW		 5025 struct intel_super *super_list = NULL;
5026 struct intel_super *super = NULL;
a2b97981 5027 int err = 0;
9587c373 5028 int i = 0;
dab4a513 5029
9587c373
LM		 5030 if (fd >= 0)
5031 /* 'fd' is an opened container */
5032 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
5033 else
ec50f7b6
LM		 5034 /* get super block from devlist devices */
5035 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5036 if (err)
1602d52c 5037 goto error;
a2b97981
DW		 5038 /* all mpbs enter, maybe one leaves */
5039 super = imsm_thunderdome(&super_list, i);
5040 if (!super) {
5041 err = 1;
5042 goto error;
cdddbdbc
DW		 5043 }
5044
47ee5a45
DW		 5045 if (find_missing(super) != 0) {
5046 free_imsm(super);
a2b97981
DW		 5047 err = 2;
5048 goto error;
47ee5a45 5049 }
8e59f3d8
AK		 5050
5051 /* load migration record */
5052 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5053 if (err == -1) {
5054 /* migration is in progress,
5055 * but migr_rec cannot be loaded,
5056 */
8e59f3d8
AK		 5057 err = 4;
5058 goto error;
5059 }
e2f41b2c
AK		 5060
5061 /* Check migration compatibility */
089f9d79 5062 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5063 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5064 if (devname)
5065 fprintf(stderr, " on %s\n", devname);
5066 else
5067 fprintf(stderr, " (IMSM).\n");
5068
5069 err = 5;
5070 goto error;
5071 }
5072
a2b97981
DW		 5073 err = 0;
5074
5075 error:
5076 while (super_list) {
5077 struct intel_super *s = super_list;
5078
5079 super_list = super_list->next;
5080 free_imsm(s);
5081 }
9587c373 5082
a2b97981
DW		 5083 if (err)
5084 return err;
f7e7067b 5085
cdddbdbc 5086 *sbp = super;
9587c373 5087 if (fd >= 0)
4dd2df09 5088 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5089 else
4dd2df09 5090 st->container_devnm[0] = 0;
a2b97981 5091 if (err == 0 && st->ss == NULL) {
bf5a934a 5092 st->ss = &super_imsm;
cdddbdbc
DW		 5093 st->minor_version = 0;
5094 st->max_devs = IMSM_MAX_DEVICES;
5095 }
cdddbdbc
DW		 5096 return 0;
5097}
2b959fbf 5098
ec50f7b6
LM		 5099static int
5100get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5101 int *max, int keep_fd)
5102{
5103 struct md_list *tmpdev;
5104 int err = 0;
5105 int i = 0;
9587c373 5106
ec50f7b6
LM		 5107 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5108 if (tmpdev->used != 1)
5109 continue;
5110 if (tmpdev->container == 1) {
ca9de185 5111 int lmax = 0;
ec50f7b6
LM		 5112 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5113 if (fd < 0) {
e7b84f9d 5114 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5115 tmpdev->devname, strerror(errno));
5116 err = 8;
5117 goto error;
5118 }
5119 err = get_sra_super_block(fd, super_list,
5120 tmpdev->devname, &lmax,
5121 keep_fd);
5122 i += lmax;
5123 close(fd);
5124 if (err) {
5125 err = 7;
5126 goto error;
5127 }
5128 } else {
5129 int major = major(tmpdev->st_rdev);
5130 int minor = minor(tmpdev->st_rdev);
5131 err = get_super_block(super_list,
4dd2df09 5132 NULL,
ec50f7b6
LM		 5133 tmpdev->devname,
5134 major, minor,
5135 keep_fd);
5136 i++;
5137 if (err) {
5138 err = 6;
5139 goto error;
5140 }
5141 }
5142 }
5143 error:
5144 *max = i;
5145 return err;
5146}
9587c373 5147
4dd2df09 5148static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5149 int major, int minor, int keep_fd)
5150{
594dc1b8 5151 struct intel_super *s;
9587c373
LM		 5152 char nm[32];
5153 int dfd = -1;
9587c373
LM		 5154 int err = 0;
5155 int retry;
5156
5157 s = alloc_super();
5158 if (!s) {
5159 err = 1;
5160 goto error;
5161 }
5162
5163 sprintf(nm, "%d:%d", major, minor);
5164 dfd = dev_open(nm, O_RDWR);
5165 if (dfd < 0) {
5166 err = 2;
5167 goto error;
5168 }
5169
fa7bb6f8 5170 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5171 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5172 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5173
5174 /* retry the load if we might have raced against mdmon */
4dd2df09 5175 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5176 for (retry = 0; retry < 3; retry++) {
5177 usleep(3000);
5178 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5179 if (err != 3)
5180 break;
5181 }
5182 error:
5183 if (!err) {
5184 s->next = *super_list;
5185 *super_list = s;
5186 } else {
5187 if (s)
8d67477f 5188 free_imsm(s);
36614e95 5189 if (dfd >= 0)
9587c373
LM		 5190 close(dfd);
5191 }
089f9d79 5192 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5193 close(dfd);
5194 return err;
5195
5196}
5197
5198static int
5199get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5200{
5201 struct mdinfo *sra;
4dd2df09 5202 char *devnm;
9587c373
LM		 5203 struct mdinfo *sd;
5204 int err = 0;
5205 int i = 0;
4dd2df09 5206 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5207 if (!sra)
5208 return 1;
5209
5210 if (sra->array.major_version != -1 ||
5211 sra->array.minor_version != -2 ||
5212 strcmp(sra->text_version, "imsm") != 0) {
5213 err = 1;
5214 goto error;
5215 }
5216 /* load all mpbs */
4dd2df09 5217 devnm = fd2devnm(fd);
9587c373 5218 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5219 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5220 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5221 err = 7;
5222 goto error;
5223 }
5224 }
5225 error:
5226 sysfs_free(sra);
5227 *max = i;
5228 return err;
5229}
5230
2b959fbf
N		 5231static int load_container_imsm(struct supertype *st, int fd, char *devname)
5232{
ec50f7b6 5233 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5234}
cdddbdbc
DW		 5235
5236static int load_super_imsm(struct supertype *st, int fd, char *devname)
5237{
5238 struct intel_super *super;
5239 int rv;
8a3544f8 5240 int retry;
cdddbdbc 5241
357ac106 5242 if (test_partition(fd))
691c6ee1
N		 5243 /* IMSM not allowed on partitions */
5244 return 1;
5245
37424f13
DW		 5246 free_super_imsm(st);
5247
49133e57 5248 super = alloc_super();
fa7bb6f8 5249 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5250 if (!super)
5251 return 1;
ea2bc72b
LM		 5252 /* Load hba and capabilities if they exist.
5253 * But do not preclude loading metadata in case capabilities or hba are
5254 * non-compliant and ignore_hw_compat is set.
5255 */
d424212e 5256 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5257 /* no orom/efi or non-intel hba of the disk */
089f9d79 5258 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5259 if (devname)
e7b84f9d 5260 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5261 free_imsm(super);
5262 return 2;
5263 }
a2b97981 5264 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5265
8a3544f8
AP		 5266 /* retry the load if we might have raced against mdmon */
5267 if (rv == 3) {
f96b1302
AP		 5268 struct mdstat_ent *mdstat = NULL;
5269 char *name = fd2kname(fd);
5270
5271 if (name)
5272 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5273
5274 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5275 for (retry = 0; retry < 3; retry++) {
5276 usleep(3000);
5277 rv = load_and_parse_mpb(fd, super, devname, 0);
5278 if (rv != 3)
5279 break;
5280 }
5281 }
5282
5283 free_mdstat(mdstat);
5284 }
5285
cdddbdbc
DW		 5286 if (rv) {
5287 if (devname)
7a862a02 5288 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5289 free_imsm(super);
5290 return rv;
5291 }
5292
5293 st->sb = super;
5294 if (st->ss == NULL) {
5295 st->ss = &super_imsm;
5296 st->minor_version = 0;
5297 st->max_devs = IMSM_MAX_DEVICES;
5298 }
8e59f3d8
AK		 5299
5300 /* load migration record */
2e062e82
AK		 5301 if (load_imsm_migr_rec(super, NULL) == 0) {
5302 /* Check for unsupported migration features */
5303 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5304 pr_err("Unsupported migration detected");
2e062e82
AK		 5305 if (devname)
5306 fprintf(stderr, " on %s\n", devname);
5307 else
5308 fprintf(stderr, " (IMSM).\n");
5309 return 3;
5310 }
e2f41b2c
AK		 5311 }
5312
cdddbdbc
DW		 5313 return 0;
5314}
5315
ef6ffade
DW		 5316static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5317{
5318 if (info->level == 1)
5319 return 128;
5320 return info->chunk_size >> 9;
5321}
5322
5551b113
CA		 5323static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5324 unsigned long long size)
fcfd9599 5325{
4025c288 5326 if (info->level == 1)
5551b113 5327 return size * 2;
4025c288 5328 else
5551b113 5329 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5330}
5331
4d1313e9
DW		 5332static void imsm_update_version_info(struct intel_super *super)
5333{
5334 /* update the version and attributes */
5335 struct imsm_super *mpb = super->anchor;
5336 char *version;
5337 struct imsm_dev *dev;
5338 struct imsm_map *map;
5339 int i;
5340
5341 for (i = 0; i < mpb->num_raid_devs; i++) {
5342 dev = get_imsm_dev(super, i);
238c0a71 5343 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5344 if (__le32_to_cpu(dev->size_high) > 0)
5345 mpb->attributes |= MPB_ATTRIB_2TB;
5346
5347 /* FIXME detect when an array spans a port multiplier */
5348 #if 0
5349 mpb->attributes |= MPB_ATTRIB_PM;
5350 #endif
5351
5352 if (mpb->num_raid_devs > 1 ||
5353 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5354 version = MPB_VERSION_ATTRIBS;
5355 switch (get_imsm_raid_level(map)) {
5356 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5357 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5358 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5359 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5360 }
5361 } else {
5362 if (map->num_members >= 5)
5363 version = MPB_VERSION_5OR6_DISK_ARRAY;
5364 else if (dev->status == DEV_CLONE_N_GO)
5365 version = MPB_VERSION_CNG;
5366 else if (get_imsm_raid_level(map) == 5)
5367 version = MPB_VERSION_RAID5;
5368 else if (map->num_members >= 3)
5369 version = MPB_VERSION_3OR4_DISK_ARRAY;
5370 else if (get_imsm_raid_level(map) == 1)
5371 version = MPB_VERSION_RAID1;
5372 else
5373 version = MPB_VERSION_RAID0;
5374 }
5375 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5376 }
5377}
5378
aa534678
DW		 5379static int check_name(struct intel_super *super, char *name, int quiet)
5380{
5381 struct imsm_super *mpb = super->anchor;
5382 char *reason = NULL;
9bd99a90
RS		 5383 char *start = name;
5384 size_t len = strlen(name);
aa534678
DW		 5385 int i;
5386
9bd99a90
RS		 5387 if (len > 0) {
5388 while (isspace(start[len - 1]))
5389 start[--len] = 0;
5390 while (*start && isspace(*start))
5391 ++start, --len;
5392 memmove(name, start, len + 1);
5393 }
5394
5395 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5396 reason = "must be 16 characters or less";
9bd99a90
RS		 5397 else if (len == 0)
5398 reason = "must be a non-empty string";
aa534678
DW		 5399
5400 for (i = 0; i < mpb->num_raid_devs; i++) {
5401 struct imsm_dev *dev = get_imsm_dev(super, i);
5402
5403 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5404 reason = "already exists";
5405 break;
5406 }
5407 }
5408
5409 if (reason && !quiet)
e7b84f9d 5410 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5411
5412 return !reason;
5413}
5414
8b353278 5415static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5416 struct shape *s, char *name,
83cd1e97
N		 5417 char *homehost, int *uuid,
5418 long long data_offset)
cdddbdbc 5419{
c2c087e6
DW		 5420 /* We are creating a volume inside a pre-existing container.
5421 * so st->sb is already set.
5422 */
5423 struct intel_super *super = st->sb;
f36a9ecd 5424 unsigned int sector_size = super->sector_size;
949c47a0 5425 struct imsm_super *mpb = super->anchor;
ba2de7ba 5426 struct intel_dev *dv;
c2c087e6
DW		 5427 struct imsm_dev *dev;
5428 struct imsm_vol *vol;
5429 struct imsm_map *map;
5430 int idx = mpb->num_raid_devs;
5431 int i;
760365f9 5432 int namelen;
c2c087e6 5433 unsigned long long array_blocks;
2c092cad 5434 size_t size_old, size_new;
5551b113 5435 unsigned long long num_data_stripes;
b53bfba6
TM		 5436 unsigned int data_disks;
5437 unsigned long long size_per_member;
cdddbdbc 5438
88c32bb1 5439 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5440 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5441 return 0;
5442 }
5443
2c092cad
DW		 5444 /* ensure the mpb is large enough for the new data */
5445 size_old = __le32_to_cpu(mpb->mpb_size);
5446 size_new = disks_to_mpb_size(info->nr_disks);
5447 if (size_new > size_old) {
5448 void *mpb_new;
f36a9ecd 5449 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5450
f36a9ecd 5451 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5452 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5453 return 0;
5454 }
85337573
AO		 5455 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5456 MIGR_REC_BUF_SECTORS*
5457 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5458 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5459 free(super->buf);
5460 free(super);
ea944c8f 5461 free(mpb_new);
8e59f3d8
AK		 5462 return 0;
5463 }
2c092cad
DW		 5464 memcpy(mpb_new, mpb, size_old);
5465 free(mpb);
5466 mpb = mpb_new;
949c47a0 5467 super->anchor = mpb_new;
2c092cad
DW		 5468 mpb->mpb_size = __cpu_to_le32(size_new);
5469 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5470 super->len = size_round;
2c092cad 5471 }
bf5a934a 5472 super->current_vol = idx;
3960e579
DW		 5473
5474 /* handle 'failed_disks' by either:
5475 * a) create dummy disk entries in the table if this the first
5476 * volume in the array. We add them here as this is the only
5477 * opportunity to add them. add_to_super_imsm_volume()
5478 * handles the non-failed disks and continues incrementing
5479 * mpb->num_disks.
5480 * b) validate that 'failed_disks' matches the current number
5481 * of missing disks if the container is populated
d23fe947 5482 */
3960e579 5483 if (super->current_vol == 0) {
d23fe947 5484 mpb->num_disks = 0;
3960e579
DW		 5485 for (i = 0; i < info->failed_disks; i++) {
5486 struct imsm_disk *disk;
5487
5488 mpb->num_disks++;
5489 disk = __get_imsm_disk(mpb, i);
5490 disk->status = CONFIGURED_DISK | FAILED_DISK;
5491 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5492 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5493 "missing:%d", (__u8)i);
3960e579
DW		 5494 }
5495 find_missing(super);
5496 } else {
5497 int missing = 0;
5498 struct dl *d;
5499
5500 for (d = super->missing; d; d = d->next)
5501 missing++;
5502 if (info->failed_disks > missing) {
e7b84f9d 5503 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5504 return 0;
5505 }
5506 }
5a038140 5507
aa534678
DW		 5508 if (!check_name(super, name, 0))
5509 return 0;
503975b9
N		 5510 dv = xmalloc(sizeof(*dv));
5511 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5512 /*
5513 * Explicitly allow truncating to not confuse gcc's
5514 * -Werror=stringop-truncation
5515 */
5516 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5517 memcpy(dev->volume, name, namelen);
e03640bd 5518 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5519 info->layout, info->chunk_size,
b53bfba6
TM		 5520 s->size * BLOCKS_PER_KB);
5521 data_disks = get_data_disks(info->level, info->layout,
5522 info->raid_disks);
5523 array_blocks = round_size_to_mb(array_blocks, data_disks);
5524 size_per_member = array_blocks / data_disks;
979d38be 5525
fcc2c9da 5526 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5527 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5528 vol = &dev->vol;
5529 vol->migr_state = 0;
1484e727 5530 set_migr_type(dev, MIGR_INIT);
3960e579 5531 vol->dirty = !info->state;
f8f603f1 5532 vol->curr_migr_unit = 0;
238c0a71 5533 map = get_imsm_map(dev, MAP_0);
5551b113 5534 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5535 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5536 map->failed_disk_num = ~0;
bf4442ab 5537 if (info->level > 0)
fffaf1ff
N		 5538 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5539 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5540 else
5541 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5542 IMSM_T_STATE_NORMAL;
252d23c0 5543 map->ddf = 1;
ef6ffade
DW		 5544
5545 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5546 free(dev);
5547 free(dv);
7a862a02 5548 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5549 return 0;
5550 }
81062a36
DW		 5551
5552 map->raid_level = info->level;
4d1313e9 5553 if (info->level == 10) {
c2c087e6 5554 map->raid_level = 1;
4d1313e9 5555 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5556 } else if (info->level == 1)
5557 map->num_domains = info->raid_disks;
5558 else
ff596308 5559 map->num_domains = 1;
81062a36 5560
5551b113 5561 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5562 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5563 num_data_stripes /= map->num_domains;
5564 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5565
44490938
MD		 5566 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5567 set_blocks_per_member(map, info_to_blocks_per_member(info,
5568 size_per_member /
5569 BLOCKS_PER_KB));
5570
c2c087e6
DW		 5571 map->num_members = info->raid_disks;
5572 for (i = 0; i < map->num_members; i++) {
5573 /* initialized in add_to_super */
4eb26970 5574 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5575 }
949c47a0 5576 mpb->num_raid_devs++;
2a24dc1b
PB		 5577 mpb->num_raid_devs_created++;
5578 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5579
b7580566 5580 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5581 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5582 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5583 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5584 } else {
5585 free(dev);
5586 free(dv);
5587 pr_err("imsm does not support consistency policy %s\n",
5588 map_num(consistency_policies, s->consistency_policy));
5589 return 0;
5590 }
5591
ba2de7ba
DW		 5592 dv->dev = dev;
5593 dv->index = super->current_vol;
5594 dv->next = super->devlist;
5595 super->devlist = dv;
c2c087e6 5596
4d1313e9
DW		 5597 imsm_update_version_info(super);
5598
c2c087e6 5599 return 1;
cdddbdbc
DW		 5600}
5601
bf5a934a 5602static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5603 struct shape *s, char *name,
83cd1e97
N		 5604 char *homehost, int *uuid,
5605 unsigned long long data_offset)
bf5a934a
DW		 5606{
5607 /* This is primarily called by Create when creating a new array.
5608 * We will then get add_to_super called for each component, and then
5609 * write_init_super called to write it out to each device.
5610 * For IMSM, Create can create on fresh devices or on a pre-existing
5611 * array.
5612 * To create on a pre-existing array a different method will be called.
5613 * This one is just for fresh drives.
5614 */
5615 struct intel_super *super;
5616 struct imsm_super *mpb;
5617 size_t mpb_size;
4d1313e9 5618 char *version;
bf5a934a 5619
83cd1e97 5620 if (data_offset != INVALID_SECTORS) {
ed503f89 5621 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5622 return 0;
5623 }
5624
bf5a934a 5625 if (st->sb)
5308f117 5626 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5627 data_offset);
e683ca88
DW		 5628
5629 if (info)
5630 mpb_size = disks_to_mpb_size(info->nr_disks);
5631 else
f36a9ecd 5632 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5633
49133e57 5634 super = alloc_super();
f36a9ecd
PB		 5635 if (super &&
5636 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5637 free_imsm(super);
e683ca88
DW		 5638 super = NULL;
5639 }
5640 if (!super) {
1ade5cc1 5641 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5642 return 0;
5643 }
de44e46f
PB		 5644 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5645 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5646 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5647 free(super->buf);
8d67477f 5648 free_imsm(super);
8e59f3d8
AK		 5649 return 0;
5650 }
e683ca88 5651 memset(super->buf, 0, mpb_size);
ef649044 5652 mpb = super->buf;
e683ca88
DW		 5653 mpb->mpb_size = __cpu_to_le32(mpb_size);
5654 st->sb = super;
5655
5656 if (info == NULL) {
5657 /* zeroing superblock */
5658 return 0;
5659 }
bf5a934a 5660
4d1313e9
DW		 5661 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5662
5663 version = (char *) mpb->sig;
5664 strcpy(version, MPB_SIGNATURE);
5665 version += strlen(MPB_SIGNATURE);
5666 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5667
bf5a934a
DW		 5668 return 1;
5669}
5670
f2cc4f7d
AO		 5671static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5672{
5673 unsigned int member_sector_size;
5674
5675 if (dl->fd < 0) {
5676 pr_err("Invalid file descriptor for %s\n", dl->devname);
5677 return 0;
5678 }
5679
5680 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5681 return 0;
5682 if (member_sector_size != super->sector_size)
5683 return 0;
5684 return 1;
5685}
5686
f20c3968 5687static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5688 int fd, char *devname)
5689{
5690 struct intel_super *super = st->sb;
d23fe947 5691 struct imsm_super *mpb = super->anchor;
3960e579 5692 struct imsm_disk *_disk;
bf5a934a
DW		 5693 struct imsm_dev *dev;
5694 struct imsm_map *map;
3960e579 5695 struct dl *dl, *df;
4eb26970 5696 int slot;
bf5a934a 5697
949c47a0 5698 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5699 map = get_imsm_map(dev, MAP_0);
bf5a934a 5700
208933a7 5701 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5702 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5703 devname);
5704 return 1;
5705 }
5706
efb30e7f
DW		 5707 if (fd == -1) {
5708 /* we're doing autolayout so grab the pre-marked (in
5709 * validate_geometry) raid_disk
5710 */
5711 for (dl = super->disks; dl; dl = dl->next)
5712 if (dl->raiddisk == dk->raid_disk)
5713 break;
5714 } else {
5715 for (dl = super->disks; dl ; dl = dl->next)
5716 if (dl->major == dk->major &&
5717 dl->minor == dk->minor)
5718 break;
5719 }
d23fe947 5720
208933a7 5721 if (!dl) {
e7b84f9d 5722 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5723 return 1;
208933a7 5724 }
bf5a934a 5725
59632db9
MZ		 5726 if (mpb->num_disks == 0)
5727 if (!get_dev_sector_size(dl->fd, dl->devname,
5728 &super->sector_size))
5729 return 1;
5730
f2cc4f7d
AO		 5731 if (!drive_validate_sector_size(super, dl)) {
5732 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5733 return 1;
5734 }
5735
d23fe947
DW		 5736 /* add a pristine spare to the metadata */
5737 if (dl->index < 0) {
5738 dl->index = super->anchor->num_disks;
5739 super->anchor->num_disks++;
5740 }
4eb26970
DW		 5741 /* Check the device has not already been added */
5742 slot = get_imsm_disk_slot(map, dl->index);
5743 if (slot >= 0 &&
238c0a71 5744 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5745 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5746 devname);
5747 return 1;
5748 }
656b6b5a 5749 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5750 dl->disk.status = CONFIGURED_DISK;
d23fe947 5751
3960e579
DW		 5752 /* update size of 'missing' disks to be at least as large as the
5753 * largest acitve member (we only have dummy missing disks when
5754 * creating the first volume)
5755 */
5756 if (super->current_vol == 0) {
5757 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5758 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5759 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5760 _disk = __get_imsm_disk(mpb, df->index);
5761 *_disk = df->disk;
5762 }
5763 }
5764
5765 /* refresh unset/failed slots to point to valid 'missing' entries */
5766 for (df = super->missing; df; df = df->next)
5767 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5768 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5769
5770 if ((ord & IMSM_ORD_REBUILD) == 0)
5771 continue;
5772 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5773 if (is_gen_migration(dev)) {
238c0a71
AK		 5774 struct imsm_map *map2 = get_imsm_map(dev,
5775 MAP_1);
0a108d63 5776 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5777 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5778 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5779 slot2,
5780 MAP_1);
1ace8403
AK		 5781 if ((unsigned)df->index ==
5782 ord_to_idx(ord2))
5783 set_imsm_ord_tbl_ent(map2,
0a108d63 5784 slot2,
1ace8403
AK		 5785 df->index |
5786 IMSM_ORD_REBUILD);
5787 }
5788 }
3960e579
DW		 5789 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5790 break;
5791 }
5792
d23fe947
DW		 5793 /* if we are creating the first raid device update the family number */
5794 if (super->current_vol == 0) {
5795 __u32 sum;
5796 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5797
3960e579 5798 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5799 if (!_dev || !_disk) {
e7b84f9d 5800 pr_err("BUG mpb setup error\n");
791b666a
AW		 5801 return 1;
5802 }
d23fe947
DW		 5803 *_dev = *dev;
5804 *_disk = dl->disk;
148acb7b
DW		 5805 sum = random32();
5806 sum += __gen_imsm_checksum(mpb);
d23fe947 5807 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5808 mpb->orig_family_num = mpb->family_num;
d23fe947 5809 }
ca0748fa 5810 super->current_disk = dl;
f20c3968 5811 return 0;
bf5a934a
DW		 5812}
5813
a8619d23
AK		 5814/* mark_spare()
5815 * Function marks disk as spare and restores disk serial
5816 * in case it was previously marked as failed by takeover operation
5817 * reruns:
5818 * -1 : critical error
5819 * 0 : disk is marked as spare but serial is not set
5820 * 1 : success
5821 */
5822int mark_spare(struct dl *disk)
5823{
5824 __u8 serial[MAX_RAID_SERIAL_LEN];
5825 int ret_val = -1;
5826
5827 if (!disk)
5828 return ret_val;
5829
5830 ret_val = 0;
5831 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5832 /* Restore disk serial number, because takeover marks disk
5833 * as failed and adds to serial ':0' before it becomes
5834 * a spare disk.
5835 */
5836 serialcpy(disk->serial, serial);
5837 serialcpy(disk->disk.serial, serial);
5838 ret_val = 1;
5839 }
5840 disk->disk.status = SPARE_DISK;
5841 disk->index = -1;
5842
5843 return ret_val;
5844}
88654014 5845
f20c3968 5846static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5847 int fd, char *devname,
5848 unsigned long long data_offset)
cdddbdbc 5849{
c2c087e6 5850 struct intel_super *super = st->sb;
c2c087e6
DW		 5851 struct dl *dd;
5852 unsigned long long size;
fa7bb6f8 5853 unsigned int member_sector_size;
f2f27e63 5854 __u32 id;
c2c087e6
DW		 5855 int rv;
5856 struct stat stb;
5857
88654014
LM		 5858 /* If we are on an RAID enabled platform check that the disk is
5859 * attached to the raid controller.
5860 * We do not need to test disks attachment for container based additions,
5861 * they shall be already tested when container was created/assembled.
88c32bb1 5862 */
d424212e 5863 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5864 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5865 if (rv != 0) {
5866 dprintf("capability: %p fd: %d ret: %d\n",
5867 super->orom, fd, rv);
5868 return 1;
88c32bb1
DW		 5869 }
5870
f20c3968
DW		 5871 if (super->current_vol >= 0)
5872 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5873
c2c087e6 5874 fstat(fd, &stb);
503975b9 5875 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5876 dd->major = major(stb.st_rdev);
5877 dd->minor = minor(stb.st_rdev);
503975b9 5878 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5879 dd->fd = fd;
689c9bf3 5880 dd->e = NULL;
1a64be56 5881 dd->action = DISK_ADD;
c2c087e6 5882 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5883 if (rv) {
e7b84f9d 5884 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5885 if (dd->devname)
5886 free(dd->devname);
949c47a0 5887 free(dd);
0030e8d6 5888 abort();
c2c087e6 5889 }
20bee0f8
PB		 5890 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5891 (super->hba->type == SYS_DEV_VMD))) {
5892 int i;
5893 char *devpath = diskfd_to_devpath(fd);
5894 char controller_path[PATH_MAX];
5895
5896 if (!devpath) {
5897 pr_err("failed to get devpath, aborting\n");
5898 if (dd->devname)
5899 free(dd->devname);
5900 free(dd);
5901 return 1;
5902 }
5903
5904 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5905 free(devpath);
5906
5907 if (devpath_to_vendor(controller_path) == 0x8086) {
5908 /*
5909 * If Intel's NVMe drive has serial ended with
5910 * "-A","-B","-1" or "-2" it means that this is "x8"
5911 * device (double drive on single PCIe card).
5912 * User should be warned about potential data loss.
5913 */
5914 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5915 /* Skip empty character at the end */
5916 if (dd->serial[i] == 0)
5917 continue;
5918
5919 if (((dd->serial[i] == 'A') ||
5920 (dd->serial[i] == 'B') ||
5921 (dd->serial[i] == '1') ||
5922 (dd->serial[i] == '2')) &&
5923 (dd->serial[i-1] == '-'))
5924 pr_err("\tThe action you are about to take may put your data at risk.\n"
5925 "\tPlease note that x8 devices may consist of two separate x4 devices "
5926 "located on a single PCIe port.\n"
5927 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5928 break;
5929 }
32716c51
PB		 5930 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5931 !imsm_orom_has_tpv_support(super->orom)) {
5932 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5933 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5934 free(dd->devname);
5935 free(dd);
5936 return 1;
20bee0f8
PB		 5937 }
5938 }
c2c087e6 5939
c2c087e6 5940 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5941 get_dev_sector_size(fd, NULL, &member_sector_size);
5942
5943 if (super->sector_size == 0) {
5944 /* this a first device, so sector_size is not set yet */
5945 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5946 }
5947
71e5411e 5948 /* clear migr_rec when adding disk to container */
85337573
AO		 5949 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5950 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5951 SEEK_SET) >= 0) {
466070ad 5952 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5953 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5954 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5955 perror("Write migr_rec failed");
5956 }
5957
c2c087e6 5958 size /= 512;
1f24f035 5959 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5960 set_total_blocks(&dd->disk, size);
5961 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5962 struct imsm_super *mpb = super->anchor;
5963 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5964 }
a8619d23 5965 mark_spare(dd);
c2c087e6 5966 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5967 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5968 else
b9f594fe 5969 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5970
5971 if (st->update_tail) {
1a64be56
LM		 5972 dd->next = super->disk_mgmt_list;
5973 super->disk_mgmt_list = dd;
43dad3d6
DW		 5974 } else {
5975 dd->next = super->disks;
5976 super->disks = dd;
ceaf0ee1 5977 super->updates_pending++;
43dad3d6 5978 }
f20c3968
DW		 5979
5980 return 0;
cdddbdbc
DW		 5981}
5982
1a64be56
LM		 5983static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5984{
5985 struct intel_super *super = st->sb;
5986 struct dl *dd;
5987
5988 /* remove from super works only in mdmon - for communication
5989 * manager - monitor. Check if communication memory buffer
5990 * is prepared.
5991 */
5992 if (!st->update_tail) {
1ade5cc1 5993 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5994 return 1;
5995 }
503975b9 5996 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5997 dd->major = dk->major;
5998 dd->minor = dk->minor;
1a64be56 5999 dd->fd = -1;
a8619d23 6000 mark_spare(dd);
1a64be56
LM		 6001 dd->action = DISK_REMOVE;
6002
6003 dd->next = super->disk_mgmt_list;
6004 super->disk_mgmt_list = dd;
6005
1a64be56
LM		 6006 return 0;
6007}
6008
f796af5d
DW		 6009static int store_imsm_mpb(int fd, struct imsm_super *mpb);
6010
6011static union {
f36a9ecd 6012 char buf[MAX_SECTOR_SIZE];
f796af5d 6013 struct imsm_super anchor;
f36a9ecd 6014} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 6015
d23fe947
DW		 6016/* spare records have their own family number and do not have any defined raid
6017 * devices
6018 */
6019static int write_super_imsm_spares(struct intel_super *super, int doclose)
6020{
d23fe947 6021 struct imsm_super *mpb = super->anchor;
f796af5d 6022 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 6023 __u32 sum;
6024 struct dl *d;
6025
68641cdb
JS		 6026 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6027 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6028 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6029 spare->num_disks = 1;
6030 spare->num_raid_devs = 0;
6031 spare->cache_size = mpb->cache_size;
6032 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6033
6034 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6035 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 6036
6037 for (d = super->disks; d; d = d->next) {
8796fdc4 6038 if (d->index != -1)
d23fe947
DW		 6039 continue;
6040
f796af5d 6041 spare->disk[0] = d->disk;
027c374f
CA		 6042 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6043 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6044
f36a9ecd
PB		 6045 if (super->sector_size == 4096)
6046 convert_to_4k_imsm_disk(&spare->disk[0]);
6047
f796af5d
DW		 6048 sum = __gen_imsm_checksum(spare);
6049 spare->family_num = __cpu_to_le32(sum);
6050 spare->orig_family_num = 0;
6051 sum = __gen_imsm_checksum(spare);
6052 spare->check_sum = __cpu_to_le32(sum);
d23fe947 6053
f796af5d 6054 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 6055 pr_err("failed for device %d:%d %s\n",
6056 d->major, d->minor, strerror(errno));
e74255d9 6057 return 1;
d23fe947
DW		 6058 }
6059 if (doclose) {
6060 close(d->fd);
6061 d->fd = -1;
6062 }
6063 }
6064
e74255d9 6065 return 0;
d23fe947
DW		 6066}
6067
36988a3d 6068static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6069{
36988a3d 6070 struct intel_super *super = st->sb;
f36a9ecd 6071 unsigned int sector_size = super->sector_size;
949c47a0 6072 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6073 struct dl *d;
6074 __u32 generation;
6075 __u32 sum;
d23fe947 6076 int spares = 0;
949c47a0 6077 int i;
a48ac0a8 6078 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6079 int num_disks = 0;
146c6260 6080 int clear_migration_record = 1;
bbab0940 6081 __u32 bbm_log_size;
cdddbdbc 6082
c2c087e6
DW		 6083 /* 'generation' is incremented everytime the metadata is written */
6084 generation = __le32_to_cpu(mpb->generation_num);
6085 generation++;
6086 mpb->generation_num = __cpu_to_le32(generation);
6087
148acb7b
DW		 6088 /* fix up cases where previous mdadm releases failed to set
6089 * orig_family_num
6090 */
6091 if (mpb->orig_family_num == 0)
6092 mpb->orig_family_num = mpb->family_num;
6093
d23fe947 6094 for (d = super->disks; d; d = d->next) {
8796fdc4 6095 if (d->index == -1)
d23fe947 6096 spares++;
36988a3d 6097 else {
d23fe947 6098 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6099 num_disks++;
6100 }
d23fe947 6101 }
36988a3d 6102 for (d = super->missing; d; d = d->next) {
47ee5a45 6103 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6104 num_disks++;
6105 }
6106 mpb->num_disks = num_disks;
6107 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6108
949c47a0
DW		 6109 for (i = 0; i < mpb->num_raid_devs; i++) {
6110 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6111 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6112 if (dev && dev2) {
6113 imsm_copy_dev(dev, dev2);
6114 mpb_size += sizeof_imsm_dev(dev, 0);
6115 }
146c6260
AK		 6116 if (is_gen_migration(dev2))
6117 clear_migration_record = 0;
949c47a0 6118 }
bbab0940
TM		 6119
6120 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6121
6122 if (bbm_log_size) {
6123 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6124 mpb->attributes |= MPB_ATTRIB_BBM;
6125 } else
6126 mpb->attributes &= ~MPB_ATTRIB_BBM;
6127
6128 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6129 mpb_size += bbm_log_size;
a48ac0a8 6130 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6131
bbab0940
TM		 6132#ifdef DEBUG
6133 assert(super->len == 0 || mpb_size <= super->len);
6134#endif
6135
c2c087e6 6136 /* recalculate checksum */
949c47a0 6137 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6138 mpb->check_sum = __cpu_to_le32(sum);
6139
51d83f5d
AK		 6140 if (super->clean_migration_record_by_mdmon) {
6141 clear_migration_record = 1;
6142 super->clean_migration_record_by_mdmon = 0;
6143 }
146c6260 6144 if (clear_migration_record)
de44e46f 6145 memset(super->migr_rec_buf, 0,
85337573 6146 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6147
f36a9ecd
PB		 6148 if (sector_size == 4096)
6149 convert_to_4k(super);
6150
d23fe947 6151 /* write the mpb for disks that compose raid devices */
c2c087e6 6152 for (d = super->disks; d ; d = d->next) {
86c54047 6153 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6154 continue;
30602f53 6155
146c6260
AK		 6156 if (clear_migration_record) {
6157 unsigned long long dsize;
6158
6159 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6160 if (lseek64(d->fd, dsize - sector_size,
6161 SEEK_SET) >= 0) {
466070ad
PB		 6162 if ((unsigned int)write(d->fd,
6163 super->migr_rec_buf,
de44e46f
PB		 6164 MIGR_REC_BUF_SECTORS*sector_size) !=
6165 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6166 perror("Write migr_rec failed");
146c6260
AK		 6167 }
6168 }
51d83f5d
AK		 6169
6170 if (store_imsm_mpb(d->fd, mpb))
6171 fprintf(stderr,
1ade5cc1
N		 6172 "failed for device %d:%d (fd: %d)%s\n",
6173 d->major, d->minor,
51d83f5d
AK		 6174 d->fd, strerror(errno));
6175
c2c087e6
DW		 6176 if (doclose) {
6177 close(d->fd);
6178 d->fd = -1;
6179 }
6180 }
6181
d23fe947
DW		 6182 if (spares)
6183 return write_super_imsm_spares(super, doclose);
6184
e74255d9 6185 return 0;
c2c087e6
DW		 6186}
6187
9b1fb677 6188static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6189{
6190 size_t len;
6191 struct imsm_update_create_array *u;
6192 struct intel_super *super = st->sb;
9b1fb677 6193 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6194 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6195 struct disk_info *inf;
6196 struct imsm_disk *disk;
6197 int i;
43dad3d6 6198
54c2c1ea
DW		 6199 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6200 sizeof(*inf) * map->num_members;
503975b9 6201 u = xmalloc(len);
43dad3d6 6202 u->type = update_create_array;
9b1fb677 6203 u->dev_idx = dev_idx;
43dad3d6 6204 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6205 inf = get_disk_info(u);
6206 for (i = 0; i < map->num_members; i++) {
238c0a71 6207 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6208
54c2c1ea 6209 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6210 if (!disk)
6211 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6212 serialcpy(inf[i].serial, disk->serial);
6213 }
43dad3d6
DW		 6214 append_metadata_update(st, u, len);
6215
6216 return 0;
6217}
6218
1a64be56 6219static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6220{
6221 struct intel_super *super = st->sb;
6222 size_t len;
1a64be56 6223 struct imsm_update_add_remove_disk *u;
43dad3d6 6224
1a64be56 6225 if (!super->disk_mgmt_list)
43dad3d6
DW		 6226 return 0;
6227
6228 len = sizeof(*u);
503975b9 6229 u = xmalloc(len);
1a64be56 6230 u->type = update_add_remove_disk;
43dad3d6
DW		 6231 append_metadata_update(st, u, len);
6232
6233 return 0;
6234}
2432ce9b
AP		 6235
6236__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6237
e397cefe
AP		 6238static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6239{
6240 struct ppl_header *ppl_hdr = buf;
6241 int ret;
6242
6243 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6244
6245 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6246 ret = -errno;
6247 perror("Failed to seek to PPL header location");
6248 return ret;
6249 }
6250
6251 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6252 ret = -errno;
6253 perror("Write PPL header failed");
6254 return ret;
6255 }
6256
6257 fsync(fd);
6258
6259 return 0;
6260}
6261
2432ce9b
AP		 6262static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6263{
6264 struct intel_super *super = st->sb;
6265 void *buf;
6266 struct ppl_header *ppl_hdr;
6267 int ret;
6268
b2514242
PB		 6269 /* first clear entire ppl space */
6270 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6271 if (ret)
6272 return ret;
6273
6274 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6275 if (ret) {
6276 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6277 return -ret;
2432ce9b
AP		 6278 }
6279
6280 memset(buf, 0, PPL_HEADER_SIZE);
6281 ppl_hdr = buf;
6282 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6283 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6284
6285 if (info->mismatch_cnt) {
6286 /*
6287 * We are overwriting an invalid ppl. Make one entry with wrong
6288 * checksum to prevent the kernel from skipping resync.
6289 */
6290 ppl_hdr->entries_count = __cpu_to_le32(1);
6291 ppl_hdr->entries[0].checksum = ~0;
6292 }
6293
e397cefe 6294 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6295
6296 free(buf);
6297 return ret;
6298}
6299
e397cefe
AP		 6300static int is_rebuilding(struct imsm_dev *dev);
6301
2432ce9b
AP		 6302static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6303 struct mdinfo *disk)
6304{
6305 struct intel_super *super = st->sb;
6306 struct dl *d;
e397cefe 6307 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6308 int ret = 0;
e397cefe 6309 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6310 __u32 crc;
6311 struct imsm_dev *dev;
2432ce9b 6312 __u32 idx;
44b6b876
PB		 6313 unsigned int i;
6314 unsigned long long ppl_offset = 0;
6315 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6316
6317 if (disk->disk.raid_disk < 0)
6318 return 0;
6319
2432ce9b 6320 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6321 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6322 d = get_imsm_dl_disk(super, idx);
6323
6324 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6325 return 0;
6326
6327 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6328 pr_err("Failed to allocate PPL header buffer\n");
6329 return -1;
6330 }
6331 buf = buf_orig;
2432ce9b 6332
44b6b876
PB		 6333 ret = 1;
6334 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6335 void *tmp;
6336
44b6b876 6337 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6338
44b6b876
PB		 6339 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6340 SEEK_SET) < 0) {
6341 perror("Failed to seek to PPL header location");
6342 ret = -1;
e397cefe 6343 break;
44b6b876 6344 }
2432ce9b 6345
44b6b876
PB		 6346 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6347 perror("Read PPL header failed");
6348 ret = -1;
e397cefe 6349 break;
44b6b876 6350 }
2432ce9b 6351
44b6b876 6352 ppl_hdr = buf;
2432ce9b 6353
44b6b876
PB		 6354 crc = __le32_to_cpu(ppl_hdr->checksum);
6355 ppl_hdr->checksum = 0;
6356
6357 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6358 dprintf("Wrong PPL header checksum on %s\n",
6359 d->devname);
e397cefe 6360 break;
44b6b876
PB		 6361 }
6362
6363 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6364 /* previous was newest, it was already checked */
e397cefe 6365 break;
44b6b876
PB		 6366 }
6367
6368 if ((__le32_to_cpu(ppl_hdr->signature) !=
6369 super->anchor->orig_family_num)) {
6370 dprintf("Wrong PPL header signature on %s\n",
6371 d->devname);
6372 ret = 1;
e397cefe 6373 break;
44b6b876
PB		 6374 }
6375
6376 ret = 0;
6377 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6378
44b6b876
PB		 6379 ppl_offset += PPL_HEADER_SIZE;
6380 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6381 ppl_offset +=
6382 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6383
6384 if (!buf_prev)
6385 buf_prev = buf + PPL_HEADER_SIZE;
6386 tmp = buf_prev;
6387 buf_prev = buf;
6388 buf = tmp;
2432ce9b
AP		 6389 }
6390
e397cefe
AP		 6391 if (buf_prev) {
6392 buf = buf_prev;
6393 ppl_hdr = buf_prev;
6394 }
2432ce9b 6395
54148aba
PB		 6396 /*
6397 * Update metadata to use mutliple PPLs area (1MB).
6398 * This is done once for all RAID members
6399 */
6400 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6401 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6402 char subarray[20];
6403 struct mdinfo *member_dev;
6404
6405 sprintf(subarray, "%d", info->container_member);
6406
6407 if (mdmon_running(st->container_devnm))
6408 st->update_tail = &st->updates;
6409
6410 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6411 pr_err("Failed to update subarray %s\n",
6412 subarray);
6413 } else {
6414 if (st->update_tail)
6415 flush_metadata_updates(st);
6416 else
6417 st->ss->sync_metadata(st);
6418 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6419 for (member_dev = info->devs; member_dev;
6420 member_dev = member_dev->next)
6421 member_dev->ppl_size =
6422 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6423 }
6424 }
6425
b23d0750 6426 if (ret == 1) {
2fc0fc63
AP		 6427 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6428
50b9c10d
PB		 6429 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6430 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6431 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6432 (is_rebuilding(dev) &&
2ec9d182
AP		 6433 dev->vol.curr_migr_unit == 0 &&
6434 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6435 ret = st->ss->write_init_ppl(st, info, d->fd);
6436 else
6437 info->mismatch_cnt++;
e397cefe
AP		 6438 } else if (ret == 0 &&
6439 ppl_hdr->entries_count == 0 &&
6440 is_rebuilding(dev) &&
6441 info->resync_start == 0) {
6442 /*
6443 * The header has no entries - add a single empty entry and
6444 * rewrite the header to prevent the kernel from going into
6445 * resync after an interrupted rebuild.
6446 */
6447 ppl_hdr->entries_count = __cpu_to_le32(1);
6448 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6449 }
2432ce9b 6450
e397cefe
AP		 6451 free(buf_orig);
6452
2432ce9b
AP		 6453 return ret;
6454}
6455
2432ce9b
AP		 6456static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6457{
6458 struct intel_super *super = st->sb;
6459 struct dl *d;
6460 int ret = 0;
6461
6462 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6463 info->array.level != 5)
6464 return 0;
6465
6466 for (d = super->disks; d ; d = d->next) {
6467 if (d->index < 0 || is_failed(&d->disk))
6468 continue;
6469
6470 ret = st->ss->write_init_ppl(st, info, d->fd);
6471 if (ret)
6472 break;
6473 }
6474
6475 return ret;
6476}
43dad3d6 6477
c2c087e6
DW		 6478static int write_init_super_imsm(struct supertype *st)
6479{
9b1fb677
DW		 6480 struct intel_super *super = st->sb;
6481 int current_vol = super->current_vol;
2432ce9b
AP		 6482 int rv = 0;
6483 struct mdinfo info;
6484
6485 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6486
6487 /* we are done with current_vol reset it to point st at the container */
6488 super->current_vol = -1;
6489
8273f55e 6490 if (st->update_tail) {
43dad3d6
DW		 6491 /* queue the recently created array / added disk
6492 * as a metadata update */
8273f55e 6493
43dad3d6 6494 /* determine if we are creating a volume or adding a disk */
9b1fb677 6495 if (current_vol < 0) {
1a64be56
LM		 6496 /* in the mgmt (add/remove) disk case we are running
6497 * in mdmon context, so don't close fd's
43dad3d6 6498 */
2432ce9b
AP		 6499 rv = mgmt_disk(st);
6500 } else {
6501 rv = write_init_ppl_imsm_all(st, &info);
6502 if (!rv)
6503 rv = create_array(st, current_vol);
6504 }
d682f344
N		 6505 } else {
6506 struct dl *d;
6507 for (d = super->disks; d; d = d->next)
ba728be7 6508 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6509 if (current_vol >= 0)
6510 rv = write_init_ppl_imsm_all(st, &info);
6511 if (!rv)
6512 rv = write_super_imsm(st, 1);
d682f344 6513 }
2432ce9b
AP		 6514
6515 return rv;
cdddbdbc
DW		 6516}
6517
e683ca88 6518static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6519{
e683ca88
DW		 6520 struct intel_super *super = st->sb;
6521 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6522
e683ca88 6523 if (!mpb)
ad97895e
DW		 6524 return 1;
6525
f36a9ecd
PB		 6526 if (super->sector_size == 4096)
6527 convert_to_4k(super);
e683ca88 6528 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6529}
6530
cdddbdbc
DW		 6531static int validate_geometry_imsm_container(struct supertype *st, int level,
6532 int layout, int raiddisks, int chunk,
af4348dd
N		 6533 unsigned long long size,
6534 unsigned long long data_offset,
6535 char *dev,
2c514b71
NB		 6536 unsigned long long *freesize,
6537 int verbose)
cdddbdbc 6538{
c2c087e6
DW		 6539 int fd;
6540 unsigned long long ldsize;
594dc1b8 6541 struct intel_super *super;
f2f5c343 6542 int rv = 0;
cdddbdbc 6543
c2c087e6
DW		 6544 if (level != LEVEL_CONTAINER)
6545 return 0;
6546 if (!dev)
6547 return 1;
6548
6549 fd = open(dev, O_RDONLY|O_EXCL, 0);
6550 if (fd < 0) {
ba728be7 6551 if (verbose > 0)
e7b84f9d 6552 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6553 dev, strerror(errno));
c2c087e6
DW		 6554 return 0;
6555 }
6556 if (!get_dev_size(fd, dev, &ldsize)) {
6557 close(fd);
6558 return 0;
6559 }
f2f5c343
LM		 6560
6561 /* capabilities retrieve could be possible
6562 * note that there is no fd for the disks in array.
6563 */
6564 super = alloc_super();
8d67477f
TM		 6565 if (!super) {
6566 close(fd);
6567 return 0;
6568 }
fa7bb6f8
PB		 6569 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6570 close(fd);
6571 free_imsm(super);
6572 return 0;
6573 }
6574
ba728be7 6575 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6576 if (rv != 0) {
6577#if DEBUG
6578 char str[256];
6579 fd2devname(fd, str);
1ade5cc1 6580 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6581 fd, str, super->orom, rv, raiddisks);
6582#endif
6583 /* no orom/efi or non-intel hba of the disk */
6584 close(fd);
6585 free_imsm(super);
6586 return 0;
6587 }
c2c087e6 6588 close(fd);
9126b9a8
CA		 6589 if (super->orom) {
6590 if (raiddisks > super->orom->tds) {
6591 if (verbose)
7a862a02 6592 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6593 raiddisks, super->orom->tds);
6594 free_imsm(super);
6595 return 0;
6596 }
6597 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6598 (ldsize >> 9) >> 32 > 0) {
6599 if (verbose)
e7b84f9d 6600 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6601 free_imsm(super);
6602 return 0;
6603 }
f2f5c343 6604 }
c2c087e6 6605
af4348dd 6606 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6607 free_imsm(super);
c2c087e6
DW		 6608
6609 return 1;
cdddbdbc
DW		 6610}
6611
0dcecb2e
DW		 6612static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6613{
6614 const unsigned long long base_start = e[*idx].start;
6615 unsigned long long end = base_start + e[*idx].size;
6616 int i;
6617
6618 if (base_start == end)
6619 return 0;
6620
6621 *idx = *idx + 1;
6622 for (i = *idx; i < num_extents; i++) {
6623 /* extend overlapping extents */
6624 if (e[i].start >= base_start &&
6625 e[i].start <= end) {
6626 if (e[i].size == 0)
6627 return 0;
6628 if (e[i].start + e[i].size > end)
6629 end = e[i].start + e[i].size;
6630 } else if (e[i].start > end) {
6631 *idx = i;
6632 break;
6633 }
6634 }
6635
6636 return end - base_start;
6637}
6638
6639static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6640{
6641 /* build a composite disk with all known extents and generate a new
6642 * 'maxsize' given the "all disks in an array must share a common start
6643 * offset" constraint
6644 */
503975b9 6645 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6646 struct dl *dl;
6647 int i, j;
6648 int start_extent;
6649 unsigned long long pos;
b9d77223 6650 unsigned long long start = 0;
0dcecb2e
DW		 6651 unsigned long long maxsize;
6652 unsigned long reserve;
6653
0dcecb2e
DW		 6654 /* coalesce and sort all extents. also, check to see if we need to
6655 * reserve space between member arrays
6656 */
6657 j = 0;
6658 for (dl = super->disks; dl; dl = dl->next) {
6659 if (!dl->e)
6660 continue;
6661 for (i = 0; i < dl->extent_cnt; i++)
6662 e[j++] = dl->e[i];
6663 }
6664 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6665
6666 /* merge extents */
6667 i = 0;
6668 j = 0;
6669 while (i < sum_extents) {
6670 e[j].start = e[i].start;
6671 e[j].size = find_size(e, &i, sum_extents);
6672 j++;
6673 if (e[j-1].size == 0)
6674 break;
6675 }
6676
6677 pos = 0;
6678 maxsize = 0;
6679 start_extent = 0;
6680 i = 0;
6681 do {
6682 unsigned long long esize;
6683
6684 esize = e[i].start - pos;
6685 if (esize >= maxsize) {
6686 maxsize = esize;
6687 start = pos;
6688 start_extent = i;
6689 }
6690 pos = e[i].start + e[i].size;
6691 i++;
6692 } while (e[i-1].size);
6693 free(e);
6694
a7dd165b
DW		 6695 if (maxsize == 0)
6696 return 0;
6697
6698 /* FIXME assumes volume at offset 0 is the first volume in a
6699 * container
6700 */
0dcecb2e
DW		 6701 if (start_extent > 0)
6702 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6703 else
6704 reserve = 0;
6705
6706 if (maxsize < reserve)
a7dd165b 6707 return 0;
0dcecb2e 6708
5551b113 6709 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6710 if (start + reserve > super->create_offset)
a7dd165b 6711 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6712 super->create_offset = start + reserve;
6713
6714 return maxsize - reserve;
6715}
6716
88c32bb1
DW		 6717static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6718{
6719 if (level < 0 || level == 6 || level == 4)
6720 return 0;
6721
6722 /* if we have an orom prevent invalid raid levels */
6723 if (orom)
6724 switch (level) {
6725 case 0: return imsm_orom_has_raid0(orom);
6726 case 1:
6727 if (raiddisks > 2)
6728 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6729 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6730 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6731 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6732 }
6733 else
6734 return 1; /* not on an Intel RAID platform so anything goes */
6735
6736 return 0;
6737}
6738
ca9de185
LM		 6739static int
6740active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6741 int dpa, int verbose)
6742{
6743 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6744 struct mdstat_ent *memb;
ca9de185
LM		 6745 int count = 0;
6746 int num = 0;
594dc1b8 6747 struct md_list *dv;
ca9de185
LM		 6748 int found;
6749
6750 for (memb = mdstat ; memb ; memb = memb->next) {
6751 if (memb->metadata_version &&
fc54fe7a 6752 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6753 (strcmp(&memb->metadata_version[9], name) == 0) &&
6754 !is_subarray(memb->metadata_version+9) &&
6755 memb->members) {
6756 struct dev_member *dev = memb->members;
6757 int fd = -1;
6758 while(dev && (fd < 0)) {
503975b9
N		 6759 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6760 num = sprintf(path, "%s%s", "/dev/", dev->name);
6761 if (num > 0)
6762 fd = open(path, O_RDONLY, 0);
089f9d79 6763 if (num <= 0 || fd < 0) {
676e87a8 6764 pr_vrb("Cannot open %s: %s\n",
503975b9 6765 dev->name, strerror(errno));
ca9de185 6766 }
503975b9 6767 free(path);
ca9de185
LM		 6768 dev = dev->next;
6769 }
6770 found = 0;
089f9d79 6771 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6772 struct mdstat_ent *vol;
6773 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6774 if (vol->active > 0 &&
ca9de185 6775 vol->metadata_version &&
9581efb1 6776 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6777 found++;
6778 count++;
6779 }
6780 }
6781 if (*devlist && (found < dpa)) {
503975b9 6782 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6783 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6784 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6785 dv->found = found;
6786 dv->used = 0;
6787 dv->next = *devlist;
6788 *devlist = dv;
ca9de185
LM		 6789 }
6790 }
6791 if (fd >= 0)
6792 close(fd);
6793 }
6794 }
6795 free_mdstat(mdstat);
6796 return count;
6797}
6798
6799#ifdef DEBUG_LOOP
6800static struct md_list*
6801get_loop_devices(void)
6802{
6803 int i;
6804 struct md_list *devlist = NULL;
594dc1b8 6805 struct md_list *dv;
ca9de185
LM		 6806
6807 for(i = 0; i < 12; i++) {
503975b9
N		 6808 dv = xcalloc(1, sizeof(*dv));
6809 dv->devname = xmalloc(40);
ca9de185
LM		 6810 sprintf(dv->devname, "/dev/loop%d", i);
6811 dv->next = devlist;
6812 devlist = dv;
6813 }
6814 return devlist;
6815}
6816#endif
6817
6818static struct md_list*
6819get_devices(const char *hba_path)
6820{
6821 struct md_list *devlist = NULL;
594dc1b8 6822 struct md_list *dv;
ca9de185
LM		 6823 struct dirent *ent;
6824 DIR *dir;
6825 int err = 0;
6826
6827#if DEBUG_LOOP
6828 devlist = get_loop_devices();
6829 return devlist;
6830#endif
6831 /* scroll through /sys/dev/block looking for devices attached to
6832 * this hba
6833 */
6834 dir = opendir("/sys/dev/block");
6835 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6836 int fd;
6837 char buf[1024];
6838 int major, minor;
6839 char *path = NULL;
6840 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6841 continue;
6842 path = devt_to_devpath(makedev(major, minor));
6843 if (!path)
6844 continue;
6845 if (!path_attached_to_hba(path, hba_path)) {
6846 free(path);
6847 path = NULL;
6848 continue;
6849 }
6850 free(path);
6851 path = NULL;
6852 fd = dev_open(ent->d_name, O_RDONLY);
6853 if (fd >= 0) {
6854 fd2devname(fd, buf);
6855 close(fd);
6856 } else {
e7b84f9d 6857 pr_err("cannot open device: %s\n",
ca9de185
LM		 6858 ent->d_name);
6859 continue;
6860 }
6861
503975b9
N		 6862 dv = xcalloc(1, sizeof(*dv));
6863 dv->devname = xstrdup(buf);
ca9de185
LM		 6864 dv->next = devlist;
6865 devlist = dv;
6866 }
6867 if (err) {
6868 while(devlist) {
6869 dv = devlist;
6870 devlist = devlist->next;
6871 free(dv->devname);
6872 free(dv);
6873 }
6874 }
562aa102 6875 closedir(dir);
ca9de185
LM		 6876 return devlist;
6877}
6878
6879static int
6880count_volumes_list(struct md_list *devlist, char *homehost,
6881 int verbose, int *found)
6882{
6883 struct md_list *tmpdev;
6884 int count = 0;
594dc1b8 6885 struct supertype *st;
ca9de185
LM		 6886
6887 /* first walk the list of devices to find a consistent set
6888 * that match the criterea, if that is possible.
6889 * We flag the ones we like with 'used'.
6890 */
6891 *found = 0;
6892 st = match_metadata_desc_imsm("imsm");
6893 if (st == NULL) {
676e87a8 6894 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6895 return 0;
6896 }
6897
6898 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6899 char *devname = tmpdev->devname;
0a6bff09 6900 dev_t rdev;
ca9de185
LM		 6901 struct supertype *tst;
6902 int dfd;
6903 if (tmpdev->used > 1)
6904 continue;
6905 tst = dup_super(st);
6906 if (tst == NULL) {
676e87a8 6907 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6908 goto err_1;
6909 }
6910 tmpdev->container = 0;
6911 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6912 if (dfd < 0) {
1ade5cc1 6913 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6914 devname, strerror(errno));
6915 tmpdev->used = 2;
0a6bff09 6916 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6917 tmpdev->used = 2;
6918 } else if (must_be_container(dfd)) {
6919 struct supertype *cst;
6920 cst = super_by_fd(dfd, NULL);
6921 if (cst == NULL) {
1ade5cc1 6922 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6923 devname);
6924 tmpdev->used = 2;
6925 } else if (tst->ss != st->ss) {
1ade5cc1 6926 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6927 devname);
6928 tmpdev->used = 2;
6929 } else if (!tst->ss->load_container ||
6930 tst->ss->load_container(tst, dfd, NULL))
6931 tmpdev->used = 2;
6932 else {
6933 tmpdev->container = 1;
6934 }
6935 if (cst)
6936 cst->ss->free_super(cst);
6937 } else {
0a6bff09 6938 tmpdev->st_rdev = rdev;
ca9de185 6939 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6940 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6941 devname);
6942 tmpdev->used = 2;
6943 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6944 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6945 tst->ss->name, devname);
6946 tmpdev->used = 2;
6947 }
6948 }
6949 if (dfd >= 0)
6950 close(dfd);
6951 if (tmpdev->used == 2 || tmpdev->used == 4) {
6952 /* Ignore unrecognised devices during auto-assembly */
6953 goto loop;
6954 }
6955 else {
6956 struct mdinfo info;
6957 tst->ss->getinfo_super(tst, &info, NULL);
6958
6959 if (st->minor_version == -1)
6960 st->minor_version = tst->minor_version;
6961
6962 if (memcmp(info.uuid, uuid_zero,
6963 sizeof(int[4])) == 0) {
6964 /* this is a floating spare. It cannot define
6965 * an array unless there are no more arrays of
6966 * this type to be found. It can be included
6967 * in an array of this type though.
6968 */
6969 tmpdev->used = 3;
6970 goto loop;
6971 }
6972
6973 if (st->ss != tst->ss ||
6974 st->minor_version != tst->minor_version ||
6975 st->ss->compare_super(st, tst) != 0) {
6976 /* Some mismatch. If exactly one array matches this host,
6977 * we can resolve on that one.
6978 * Or, if we are auto assembling, we just ignore the second
6979 * for now.
6980 */
1ade5cc1 6981 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6982 devname);
6983 goto loop;
6984 }
6985 tmpdev->used = 1;
6986 *found = 1;
6987 dprintf("found: devname: %s\n", devname);
6988 }
6989 loop:
6990 if (tst)
6991 tst->ss->free_super(tst);
6992 }
6993 if (*found != 0) {
6994 int err;
6995 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6996 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6997 for (iter = head; iter; iter = iter->next) {
6998 dprintf("content->text_version: %s vol\n",
6999 iter->text_version);
7000 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
7001 /* do not assemble arrays with unsupported
7002 configurations */
1ade5cc1 7003 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 7004 iter->text_version);
7005 } else
7006 count++;
7007 }
7008 sysfs_free(head);
7009
7010 } else {
1ade5cc1 7011 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7012 err, st->sb);
7013 }
7014 } else {
1ade5cc1 7015 dprintf("no more devices to examine\n");
ca9de185
LM		 7016 }
7017
7018 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7019 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7020 if (count) {
7021 if (count < tmpdev->found)
7022 count = 0;
7023 else
7024 count -= tmpdev->found;
7025 }
7026 }
7027 if (tmpdev->used == 1)
7028 tmpdev->used = 4;
7029 }
7030 err_1:
7031 if (st)
7032 st->ss->free_super(st);
7033 return count;
7034}
7035
d3c11416
AO		 7036static int __count_volumes(char *hba_path, int dpa, int verbose,
7037 int cmp_hba_path)
ca9de185 7038{
72a45777 7039 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7040 int count = 0;
72a45777
PB		 7041 const struct orom_entry *entry;
7042 struct devid_list *dv, *devid_list;
ca9de185 7043
d3c11416 7044 if (!hba_path)
ca9de185
LM		 7045 return 0;
7046
72a45777 7047 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7048 if (strstr(idev->path, hba_path))
7049 break;
72a45777
PB		 7050 }
7051
7052 if (!idev || !idev->dev_id)
ca9de185 7053 return 0;
72a45777
PB		 7054
7055 entry = get_orom_entry_by_device_id(idev->dev_id);
7056
7057 if (!entry || !entry->devid_list)
7058 return 0;
7059
7060 devid_list = entry->devid_list;
7061 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7062 struct md_list *devlist;
d3c11416
AO		 7063 struct sys_dev *device = NULL;
7064 char *hpath;
72a45777
PB		 7065 int found = 0;
7066
d3c11416
AO		 7067 if (cmp_hba_path)
7068 device = device_by_id_and_path(dv->devid, hba_path);
7069 else
7070 device = device_by_id(dv->devid);
7071
72a45777 7072 if (device)
d3c11416 7073 hpath = device->path;
72a45777
PB		 7074 else
7075 return 0;
7076
d3c11416 7077 devlist = get_devices(hpath);
72a45777
PB		 7078 /* if no intel devices return zero volumes */
7079 if (devlist == NULL)
7080 return 0;
7081
d3c11416
AO		 7082 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7083 verbose);
7084 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7085 if (devlist == NULL)
7086 return 0;
7087 do {
7088 found = 0;
7089 count += count_volumes_list(devlist,
7090 NULL,
7091 verbose,
7092 &found);
7093 dprintf("found %d count: %d\n", found, count);
7094 } while (found);
7095
d3c11416 7096 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7097
7098 while (devlist) {
7099 struct md_list *dv = devlist;
7100 devlist = devlist->next;
7101 free(dv->devname);
7102 free(dv);
7103 }
ca9de185
LM		 7104 }
7105 return count;
7106}
7107
d3c11416
AO		 7108static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7109{
7110 if (!hba)
7111 return 0;
7112 if (hba->type == SYS_DEV_VMD) {
7113 struct sys_dev *dev;
7114 int count = 0;
7115
7116 for (dev = find_intel_devices(); dev; dev = dev->next) {
7117 if (dev->type == SYS_DEV_VMD)
7118 count += __count_volumes(dev->path, dpa,
7119 verbose, 1);
7120 }
7121 return count;
7122 }
7123 return __count_volumes(hba->path, dpa, verbose, 0);
7124}
7125
cd9d1ac7
DW		 7126static int imsm_default_chunk(const struct imsm_orom *orom)
7127{
7128 /* up to 512 if the plaform supports it, otherwise the platform max.
7129 * 128 if no platform detected
7130 */
7131 int fs = max(7, orom ? fls(orom->sss) : 0);
7132
7133 return min(512, (1 << fs));
7134}
73408129 7135
6592ce37
DW		 7136static int
7137validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7138 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7139{
660260d0
DW		 7140 /* check/set platform and metadata limits/defaults */
7141 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7142 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7143 super->orom->dpa);
73408129
LM		 7144 return 0;
7145 }
7146
5d500228 7147 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7148 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7149 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7150 level, raiddisks, raiddisks > 1 ? "s" : "");
7151 return 0;
7152 }
cd9d1ac7 7153
7ccc4cc4 7154 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7155 *chunk = imsm_default_chunk(super->orom);
7156
7ccc4cc4 7157 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7158 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7159 return 0;
6592ce37 7160 }
cd9d1ac7 7161
6592ce37
DW		 7162 if (layout != imsm_level_to_layout(level)) {
7163 if (level == 5)
676e87a8 7164 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7165 else if (level == 10)
676e87a8 7166 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7167 else
676e87a8 7168 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7169 layout, level);
7170 return 0;
7171 }
2cc699af 7172
7ccc4cc4 7173 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7174 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7175 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7176 return 0;
7177 }
614902f6 7178
6592ce37
DW		 7179 return 1;
7180}
7181
1011e834 7182/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7183 * FIX ME add ahci details
7184 */
8b353278 7185static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7186 int layout, int raiddisks, int *chunk,
af4348dd
N		 7187 unsigned long long size,
7188 unsigned long long data_offset,
7189 char *dev,
2c514b71
NB		 7190 unsigned long long *freesize,
7191 int verbose)
cdddbdbc 7192{
9e04ac1c 7193 dev_t rdev;
c2c087e6 7194 struct intel_super *super = st->sb;
b2916f25 7195 struct imsm_super *mpb;
c2c087e6
DW		 7196 struct dl *dl;
7197 unsigned long long pos = 0;
7198 unsigned long long maxsize;
7199 struct extent *e;
7200 int i;
cdddbdbc 7201
88c32bb1
DW		 7202 /* We must have the container info already read in. */
7203 if (!super)
c2c087e6
DW		 7204 return 0;
7205
b2916f25
JS		 7206 mpb = super->anchor;
7207
2cc699af 7208 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7209 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7210 return 0;
d54559f0 7211 }
c2c087e6
DW		 7212 if (!dev) {
7213 /* General test: make sure there is space for
2da8544a
DW		 7214 * 'raiddisks' device extents of size 'size' at a given
7215 * offset
c2c087e6 7216 */
e46273eb 7217 unsigned long long minsize = size;
b7528a20 7218 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7219 int dcnt = 0;
7220 if (minsize == 0)
7221 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7222 for (dl = super->disks; dl ; dl = dl->next) {
7223 int found = 0;
7224
bf5a934a 7225 pos = 0;
c2c087e6 7226 i = 0;
05501181 7227 e = get_extents(super, dl, 0);
c2c087e6
DW		 7228 if (!e) continue;
7229 do {
7230 unsigned long long esize;
7231 esize = e[i].start - pos;
7232 if (esize >= minsize)
7233 found = 1;
b7528a20 7234 if (found && start_offset == MaxSector) {
2da8544a
DW		 7235 start_offset = pos;
7236 break;
7237 } else if (found && pos != start_offset) {
7238 found = 0;
7239 break;
7240 }
c2c087e6
DW		 7241 pos = e[i].start + e[i].size;
7242 i++;
7243 } while (e[i-1].size);
7244 if (found)
7245 dcnt++;
7246 free(e);
7247 }
7248 if (dcnt < raiddisks) {
2c514b71 7249 if (verbose)
7a862a02 7250 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7251 dcnt, raiddisks);
c2c087e6
DW		 7252 return 0;
7253 }
7254 return 1;
7255 }
0dcecb2e 7256
c2c087e6 7257 /* This device must be a member of the set */
9e04ac1c 7258 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7259 return 0;
7260 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7261 if (dl->major == (int)major(rdev) &&
7262 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7263 break;
7264 }
7265 if (!dl) {
2c514b71 7266 if (verbose)
7a862a02 7267 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7268 return 0;
a20d2ba5
DW		 7269 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7270 /* If a volume is present then the current creation attempt
7271 * cannot incorporate new spares because the orom may not
7272 * understand this configuration (all member disks must be
7273 * members of each array in the container).
7274 */
7a862a02
N		 7275 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7276 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7277 return 0;
5fe62b94
WD		 7278 } else if (super->orom && mpb->num_raid_devs > 0 &&
7279 mpb->num_disks != raiddisks) {
7a862a02 7280 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7281 return 0;
c2c087e6 7282 }
0dcecb2e
DW		 7283
7284 /* retrieve the largest free space block */
05501181 7285 e = get_extents(super, dl, 0);
c2c087e6
DW		 7286 maxsize = 0;
7287 i = 0;
0dcecb2e
DW		 7288 if (e) {
7289 do {
7290 unsigned long long esize;
7291
7292 esize = e[i].start - pos;
7293 if (esize >= maxsize)
7294 maxsize = esize;
7295 pos = e[i].start + e[i].size;
7296 i++;
7297 } while (e[i-1].size);
7298 dl->e = e;
7299 dl->extent_cnt = i;
7300 } else {
7301 if (verbose)
e7b84f9d 7302 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7303 dev);
7304 return 0;
7305 }
7306 if (maxsize < size) {
7307 if (verbose)
e7b84f9d 7308 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7309 dev, maxsize, size);
7310 return 0;
7311 }
7312
7313 /* count total number of extents for merge */
7314 i = 0;
7315 for (dl = super->disks; dl; dl = dl->next)
7316 if (dl->e)
7317 i += dl->extent_cnt;
7318
7319 maxsize = merge_extents(super, i);
3baa56ab 7320
1a1ced1e
KS		 7321 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7322 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7323
a7dd165b 7324 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7325 if (verbose) {
7326 if (maxsize == 0)
7a862a02 7327 pr_err("no free space left on device. Aborting...\n");
b3071342 7328 else
7a862a02 7329 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7330 maxsize, size);
7331 }
0dcecb2e 7332 return 0;
0dcecb2e
DW		 7333 }
7334
c2c087e6
DW		 7335 *freesize = maxsize;
7336
ca9de185 7337 if (super->orom) {
72a45777 7338 int count = count_volumes(super->hba,
ca9de185
LM		 7339 super->orom->dpa, verbose);
7340 if (super->orom->vphba <= count) {
676e87a8 7341 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7342 super->orom->vphba);
7343 return 0;
7344 }
7345 }
c2c087e6 7346 return 1;
cdddbdbc
DW		 7347}
7348
13bcac90 7349static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7350 unsigned long long size, int chunk,
7351 unsigned long long *freesize)
7352{
7353 struct intel_super *super = st->sb;
7354 struct imsm_super *mpb = super->anchor;
7355 struct dl *dl;
7356 int i;
7357 int extent_cnt;
7358 struct extent *e;
7359 unsigned long long maxsize;
7360 unsigned long long minsize;
7361 int cnt;
7362 int used;
7363
7364 /* find the largest common start free region of the possible disks */
7365 used = 0;
7366 extent_cnt = 0;
7367 cnt = 0;
7368 for (dl = super->disks; dl; dl = dl->next) {
7369 dl->raiddisk = -1;
7370
7371 if (dl->index >= 0)
7372 used++;
7373
7374 /* don't activate new spares if we are orom constrained
7375 * and there is already a volume active in the container
7376 */
7377 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7378 continue;
7379
05501181 7380 e = get_extents(super, dl, 0);
efb30e7f
DW		 7381 if (!e)
7382 continue;
7383 for (i = 1; e[i-1].size; i++)
7384 ;
7385 dl->e = e;
7386 dl->extent_cnt = i;
7387 extent_cnt += i;
7388 cnt++;
7389 }
7390
7391 maxsize = merge_extents(super, extent_cnt);
7392 minsize = size;
7393 if (size == 0)
612e59d8
CA		 7394 /* chunk is in K */
7395 minsize = chunk * 2;
efb30e7f
DW		 7396
7397 if (cnt < raiddisks ||
7398 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7399 maxsize < minsize ||
7400 maxsize == 0) {
e7b84f9d 7401 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7402 return 0; /* No enough free spaces large enough */
7403 }
7404
7405 if (size == 0) {
7406 size = maxsize;
7407 if (chunk) {
612e59d8
CA		 7408 size /= 2 * chunk;
7409 size *= 2 * chunk;
efb30e7f 7410 }
f878b242
LM		 7411 maxsize = size;
7412 }
1a1ced1e
KS		 7413 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7414 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7415 cnt = 0;
7416 for (dl = super->disks; dl; dl = dl->next)
7417 if (dl->e)
7418 dl->raiddisk = cnt++;
7419
7420 *freesize = size;
7421
13bcac90
AK		 7422 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7423
efb30e7f
DW		 7424 return 1;
7425}
7426
13bcac90
AK		 7427static int reserve_space(struct supertype *st, int raiddisks,
7428 unsigned long long size, int chunk,
7429 unsigned long long *freesize)
7430{
7431 struct intel_super *super = st->sb;
7432 struct dl *dl;
7433 int cnt;
7434 int rv = 0;
7435
7436 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7437 if (rv) {
7438 cnt = 0;
7439 for (dl = super->disks; dl; dl = dl->next)
7440 if (dl->e)
7441 dl->raiddisk = cnt++;
7442 rv = 1;
7443 }
7444
7445 return rv;
7446}
7447
bf5a934a 7448static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7449 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7450 unsigned long long data_offset,
bf5a934a 7451 char *dev, unsigned long long *freesize,
5308f117 7452 int consistency_policy, int verbose)
bf5a934a
DW		 7453{
7454 int fd, cfd;
7455 struct mdinfo *sra;
20cbe8d2 7456 int is_member = 0;
bf5a934a 7457
d54559f0
LM		 7458 /* load capability
7459 * if given unused devices create a container
bf5a934a
DW		 7460 * if given given devices in a container create a member volume
7461 */
7462 if (level == LEVEL_CONTAINER) {
7463 /* Must be a fresh device to add to a container */
7464 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7465 raiddisks,
7ccc4cc4 7466 *chunk,
af4348dd 7467 size, data_offset,
bf5a934a
DW		 7468 dev, freesize,
7469 verbose);
7470 }
9587c373 7471
22dc741f
MT		 7472 if (size && (size < 1024)) {
7473 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7474 /* Depends on algorithm in Create.c :
7475 * if container was given (dev == NULL) return -1,
7476 * if block device was given ( dev != NULL) return 0.
7477 */
7478 return dev ? -1 : 0;
7479 }
7480
8592f29d 7481 if (!dev) {
e91a3bad 7482 if (st->sb) {
ca9de185 7483 struct intel_super *super = st->sb;
e91a3bad 7484 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7485 raiddisks, chunk, size,
e91a3bad
LM		 7486 verbose))
7487 return 0;
efb30e7f
DW		 7488 /* we are being asked to automatically layout a
7489 * new volume based on the current contents of
7490 * the container. If the the parameters can be
7491 * satisfied reserve_space will record the disks,
7492 * start offset, and size of the volume to be
7493 * created. add_to_super and getinfo_super
7494 * detect when autolayout is in progress.
7495 */
ca9de185
LM		 7496 /* assuming that freesize is always given when array is
7497 created */
7498 if (super->orom && freesize) {
7499 int count;
72a45777 7500 count = count_volumes(super->hba,
ca9de185
LM		 7501 super->orom->dpa, verbose);
7502 if (super->orom->vphba <= count) {
676e87a8 7503 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7504 super->orom->vphba);
7505 return 0;
7506 }
7507 }
e91a3bad
LM		 7508 if (freesize)
7509 return reserve_space(st, raiddisks, size,
7ccc4cc4 7510 *chunk, freesize);
8592f29d
N		 7511 }
7512 return 1;
7513 }
bf5a934a
DW		 7514 if (st->sb) {
7515 /* creating in a given container */
7516 return validate_geometry_imsm_volume(st, level, layout,
7517 raiddisks, chunk, size,
af4348dd 7518 data_offset,
bf5a934a
DW		 7519 dev, freesize, verbose);
7520 }
7521
bf5a934a
DW		 7522 /* This device needs to be a device in an 'imsm' container */
7523 fd = open(dev, O_RDONLY|O_EXCL, 0);
7524 if (fd >= 0) {
7525 if (verbose)
e7b84f9d
N		 7526 pr_err("Cannot create this array on device %s\n",
7527 dev);
bf5a934a
DW		 7528 close(fd);
7529 return 0;
7530 }
7531 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7532 if (verbose)
e7b84f9d 7533 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7534 dev, strerror(errno));
7535 return 0;
7536 }
7537 /* Well, it is in use by someone, maybe an 'imsm' container. */
7538 cfd = open_container(fd);
20cbe8d2 7539 close(fd);
bf5a934a 7540 if (cfd < 0) {
bf5a934a 7541 if (verbose)
e7b84f9d 7542 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7543 dev);
7544 return 0;
7545 }
4dd2df09 7546 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7547 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7548 strcmp(sra->text_version, "imsm") == 0)
7549 is_member = 1;
7550 sysfs_free(sra);
7551 if (is_member) {
bf5a934a
DW		 7552 /* This is a member of a imsm container. Load the container
7553 * and try to create a volume
7554 */
7555 struct intel_super *super;
7556
ec50f7b6 7557 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7558 st->sb = super;
4dd2df09 7559 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7560 close(cfd);
7561 return validate_geometry_imsm_volume(st, level, layout,
7562 raiddisks, chunk,
af4348dd 7563 size, data_offset, dev,
ecbd9e81
N		 7564 freesize, 1)
7565 ? 1 : -1;
bf5a934a 7566 }
20cbe8d2 7567 }
bf5a934a 7568
20cbe8d2 7569 if (verbose)
e7b84f9d 7570 pr_err("failed container membership check\n");
20cbe8d2
AW		 7571
7572 close(cfd);
7573 return 0;
bf5a934a 7574}
0bd16cf2 7575
30f58b22 7576static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7577{
7578 struct intel_super *super = st->sb;
7579
30f58b22
DW		 7580 if (level && *level == UnSet)
7581 *level = LEVEL_CONTAINER;
7582
7583 if (level && layout && *layout == UnSet)
7584 *layout = imsm_level_to_layout(*level);
0bd16cf2 7585
cd9d1ac7
DW		 7586 if (chunk && (*chunk == UnSet || *chunk == 0))
7587 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7588}
7589
33414a01
DW		 7590static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7591
7592static int kill_subarray_imsm(struct supertype *st)
7593{
7594 /* remove the subarray currently referenced by ->current_vol */
7595 __u8 i;
7596 struct intel_dev **dp;
7597 struct intel_super *super = st->sb;
7598 __u8 current_vol = super->current_vol;
7599 struct imsm_super *mpb = super->anchor;
7600
7601 if (super->current_vol < 0)
7602 return 2;
7603 super->current_vol = -1; /* invalidate subarray cursor */
7604
7605 /* block deletions that would change the uuid of active subarrays
7606 *
7607 * FIXME when immutable ids are available, but note that we'll
7608 * also need to fixup the invalidated/active subarray indexes in
7609 * mdstat
7610 */
7611 for (i = 0; i < mpb->num_raid_devs; i++) {
7612 char subarray[4];
7613
7614 if (i < current_vol)
7615 continue;
7616 sprintf(subarray, "%u", i);
4dd2df09 7617 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7618 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7619 current_vol, i);
33414a01
DW		 7620
7621 return 2;
7622 }
7623 }
7624
7625 if (st->update_tail) {
503975b9 7626 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7627
33414a01
DW		 7628 u->type = update_kill_array;
7629 u->dev_idx = current_vol;
7630 append_metadata_update(st, u, sizeof(*u));
7631
7632 return 0;
7633 }
7634
7635 for (dp = &super->devlist; *dp;)
7636 if ((*dp)->index == current_vol) {
7637 *dp = (*dp)->next;
7638 } else {
7639 handle_missing(super, (*dp)->dev);
7640 if ((*dp)->index > current_vol)
7641 (*dp)->index--;
7642 dp = &(*dp)->next;
7643 }
7644
7645 /* no more raid devices, all active components are now spares,
7646 * but of course failed are still failed
7647 */
7648 if (--mpb->num_raid_devs == 0) {
7649 struct dl *d;
7650
7651 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7652 if (d->index > -2)
7653 mark_spare(d);
33414a01
DW		 7654 }
7655
7656 super->updates_pending++;
7657
7658 return 0;
7659}
aa534678 7660
a951a4f7 7661static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7662 char *update, struct mddev_ident *ident)
aa534678
DW		 7663{
7664 /* update the subarray currently referenced by ->current_vol */
7665 struct intel_super *super = st->sb;
7666 struct imsm_super *mpb = super->anchor;
7667
aa534678
DW		 7668 if (strcmp(update, "name") == 0) {
7669 char *name = ident->name;
a951a4f7
N		 7670 char *ep;
7671 int vol;
aa534678 7672
4dd2df09 7673 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7674 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7675 return 2;
7676 }
7677
7678 if (!check_name(super, name, 0))
7679 return 2;
7680
a951a4f7
N		 7681 vol = strtoul(subarray, &ep, 10);
7682 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7683 return 2;
7684
aa534678 7685 if (st->update_tail) {
503975b9 7686 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7687
aa534678 7688 u->type = update_rename_array;
a951a4f7 7689 u->dev_idx = vol;
618f4e6d
XN		 7690 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7691 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7692 append_metadata_update(st, u, sizeof(*u));
7693 } else {
7694 struct imsm_dev *dev;
ebad3af2 7695 int i, namelen;
aa534678 7696
a951a4f7 7697 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7698 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7699 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7700 memcpy(dev->volume, name, namelen);
aa534678
DW		 7701 for (i = 0; i < mpb->num_raid_devs; i++) {
7702 dev = get_imsm_dev(super, i);
7703 handle_missing(super, dev);
7704 }
7705 super->updates_pending++;
7706 }
e6e9dd3f
AP		 7707 } else if (strcmp(update, "ppl") == 0 ||
7708 strcmp(update, "no-ppl") == 0) {
7709 int new_policy;
7710 char *ep;
7711 int vol = strtoul(subarray, &ep, 10);
7712
7713 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7714 return 2;
7715
7716 if (strcmp(update, "ppl") == 0)
c2462068 7717 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7718 else
c2462068 7719 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7720
7721 if (st->update_tail) {
7722 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7723
7724 u->type = update_rwh_policy;
7725 u->dev_idx = vol;
7726 u->new_policy = new_policy;
7727 append_metadata_update(st, u, sizeof(*u));
7728 } else {
7729 struct imsm_dev *dev;
7730
7731 dev = get_imsm_dev(super, vol);
7732 dev->rwh_policy = new_policy;
7733 super->updates_pending++;
7734 }
aa534678
DW		 7735 } else
7736 return 2;
7737
7738 return 0;
7739}
bf5a934a 7740
28bce06f
AK		 7741static int is_gen_migration(struct imsm_dev *dev)
7742{
7534230b
AK		 7743 if (dev == NULL)
7744 return 0;
7745
28bce06f
AK		 7746 if (!dev->vol.migr_state)
7747 return 0;
7748
7749 if (migr_type(dev) == MIGR_GEN_MIGR)
7750 return 1;
7751
7752 return 0;
7753}
7754
1e5c6983
DW		 7755static int is_rebuilding(struct imsm_dev *dev)
7756{
7757 struct imsm_map *migr_map;
7758
7759 if (!dev->vol.migr_state)
7760 return 0;
7761
7762 if (migr_type(dev) != MIGR_REBUILD)
7763 return 0;
7764
238c0a71 7765 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7766
7767 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7768 return 1;
7769 else
7770 return 0;
7771}
7772
6ce1fbf1
AK		 7773static int is_initializing(struct imsm_dev *dev)
7774{
7775 struct imsm_map *migr_map;
7776
7777 if (!dev->vol.migr_state)
7778 return 0;
7779
7780 if (migr_type(dev) != MIGR_INIT)
7781 return 0;
7782
238c0a71 7783 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7784
7785 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7786 return 1;
7787
7788 return 0;
6ce1fbf1
AK		 7789}
7790
c47b0ff6
AK		 7791static void update_recovery_start(struct intel_super *super,
7792 struct imsm_dev *dev,
7793 struct mdinfo *array)
1e5c6983
DW		 7794{
7795 struct mdinfo *rebuild = NULL;
7796 struct mdinfo *d;
7797 __u32 units;
7798
7799 if (!is_rebuilding(dev))
7800 return;
7801
7802 /* Find the rebuild target, but punt on the dual rebuild case */
7803 for (d = array->devs; d; d = d->next)
7804 if (d->recovery_start == 0) {
7805 if (rebuild)
7806 return;
7807 rebuild = d;
7808 }
7809
4363fd80
DW		 7810 if (!rebuild) {
7811 /* (?) none of the disks are marked with
7812 * IMSM_ORD_REBUILD, so assume they are missing and the
7813 * disk_ord_tbl was not correctly updated
7814 */
1ade5cc1 7815 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7816 return;
7817 }
7818
1e5c6983 7819 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7820 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7821}
7822
276d77db 7823static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7824
00bbdbda 7825static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7826{
4f5bc454
DW		 7827 /* Given a container loaded by load_super_imsm_all,
7828 * extract information about all the arrays into
7829 * an mdinfo tree.
00bbdbda 7830 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7831 *
7832 * For each imsm_dev create an mdinfo, fill it in,
7833 * then look for matching devices in super->disks
7834 * and create appropriate device mdinfo.
7835 */
7836 struct intel_super *super = st->sb;
949c47a0 7837 struct imsm_super *mpb = super->anchor;
4f5bc454 7838 struct mdinfo *rest = NULL;
00bbdbda 7839 unsigned int i;
81219e70 7840 int sb_errors = 0;
abef11a3
AK		 7841 struct dl *d;
7842 int spare_disks = 0;
b6180160 7843 int current_vol = super->current_vol;
cdddbdbc 7844
19482bcc
AK		 7845 /* do not assemble arrays when not all attributes are supported */
7846 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7847 sb_errors = 1;
7a862a02 7848 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7849 }
7850
abef11a3
AK		 7851 /* count spare devices, not used in maps
7852 */
7853 for (d = super->disks; d; d = d->next)
7854 if (d->index == -1)
7855 spare_disks++;
7856
4f5bc454 7857 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7858 struct imsm_dev *dev;
7859 struct imsm_map *map;
86e3692b 7860 struct imsm_map *map2;
4f5bc454 7861 struct mdinfo *this;
a6482415 7862 int slot;
a6482415 7863 int chunk;
00bbdbda 7864 char *ep;
8b9cd157 7865 int level;
00bbdbda
N		 7866
7867 if (subarray &&
7868 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7869 continue;
7870
7871 dev = get_imsm_dev(super, i);
238c0a71
AK		 7872 map = get_imsm_map(dev, MAP_0);
7873 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7874 level = get_imsm_raid_level(map);
4f5bc454 7875
1ce0101c
DW		 7876 /* do not publish arrays that are in the middle of an
7877 * unsupported migration
7878 */
7879 if (dev->vol.migr_state &&
28bce06f 7880 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7881 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7882 dev->volume);
7883 continue;
7884 }
2db86302
LM		 7885 /* do not publish arrays that are not support by controller's
7886 * OROM/EFI
7887 */
1ce0101c 7888
503975b9 7889 this = xmalloc(sizeof(*this));
4f5bc454 7890
301406c9 7891 super->current_vol = i;
a5d85af7 7892 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7893 this->next = rest;
a6482415 7894 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7895 /* mdadm does not support all metadata features- set the bit in all arrays state */
7896 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7897 level, /* RAID level */
7898 imsm_level_to_layout(level),
81219e70 7899 map->num_members, /* raid disks */
fcc2c9da 7900 &chunk, imsm_dev_size(dev),
81219e70 7901 1 /* verbose */)) {
7a862a02 7902 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7903 dev->volume);
7904 this->array.state |=
7905 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7906 (1<<MD_SB_BLOCK_VOLUME);
7907 }
81219e70
LM		 7908
7909 /* if array has bad blocks, set suitable bit in all arrays state */
7910 if (sb_errors)
7911 this->array.state |=
7912 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7913 (1<<MD_SB_BLOCK_VOLUME);
7914
4f5bc454 7915 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7916 unsigned long long recovery_start;
4f5bc454
DW		 7917 struct mdinfo *info_d;
7918 struct dl *d;
7919 int idx;
9a1608e5 7920 int skip;
7eef0453 7921 __u32 ord;
8b9cd157 7922 int missing = 0;
4f5bc454 7923
9a1608e5 7924 skip = 0;
238c0a71
AK		 7925 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7926 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7927 for (d = super->disks; d ; d = d->next)
7928 if (d->index == idx)
0fbd635c 7929 break;
4f5bc454 7930
1e5c6983 7931 recovery_start = MaxSector;
4f5bc454 7932 if (d == NULL)
9a1608e5 7933 skip = 1;
25ed7e59 7934 if (d && is_failed(&d->disk))
9a1608e5 7935 skip = 1;
8b9cd157 7936 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7937 recovery_start = 0;
9a1608e5 7938
1011e834 7939 /*
9a1608e5 7940 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7941 * reset resync start to avoid a dirty-degraded
7942 * situation when performing the intial sync
9a1608e5 7943 */
8b9cd157
MK		 7944 if (skip)
7945 missing++;
7946
7947 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7948 if ((!able_to_resync(level, missing) ||
7949 recovery_start == 0))
7950 this->resync_start = MaxSector;
7951 } else {
7952 /*
7953 * FIXME handle dirty degraded
7954 */
7955 }
7956
9a1608e5
DW		 7957 if (skip)
7958 continue;
4f5bc454 7959
503975b9 7960 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7961 info_d->next = this->devs;
7962 this->devs = info_d;
7963
4f5bc454
DW		 7964 info_d->disk.number = d->index;
7965 info_d->disk.major = d->major;
7966 info_d->disk.minor = d->minor;
7967 info_d->disk.raid_disk = slot;
1e5c6983 7968 info_d->recovery_start = recovery_start;
86e3692b
AK		 7969 if (map2) {
7970 if (slot < map2->num_members)
7971 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7972 else
7973 this->array.spare_disks++;
86e3692b
AK		 7974 } else {
7975 if (slot < map->num_members)
7976 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7977 else
7978 this->array.spare_disks++;
86e3692b 7979 }
1e5c6983
DW		 7980 if (info_d->recovery_start == MaxSector)
7981 this->array.working_disks++;
4f5bc454
DW		 7982
7983 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7984 info_d->data_offset = pba_of_lba0(map);
44490938 7985 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7986
7987 if (map->raid_level == 5) {
2432ce9b
AP		 7988 info_d->ppl_sector = this->ppl_sector;
7989 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7990 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7991 recovery_start == 0)
7992 this->resync_start = 0;
06fb291a 7993 }
b12796be 7994
5e46202e 7995 info_d->bb.supported = 1;
b12796be
TM		 7996 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7997 info_d->data_offset,
7998 info_d->component_size,
7999 &info_d->bb);
4f5bc454 8000 }
1e5c6983 8001 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 8002 update_recovery_start(super, dev, this);
abef11a3 8003 this->array.spare_disks += spare_disks;
276d77db
AK		 8004
8005 /* check for reshape */
8006 if (this->reshape_active == 1)
8007 recover_backup_imsm(st, this);
9a1608e5 8008 rest = this;
4f5bc454
DW		 8009 }
8010
b6180160 8011 super->current_vol = current_vol;
4f5bc454 8012 return rest;
cdddbdbc
DW		 8013}
8014
3b451610
AK		 8015static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8016 int failed, int look_in_map)
c2a1e7da 8017{
3b451610
AK		 8018 struct imsm_map *map;
8019
8020 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8021
8022 if (!failed)
1011e834 8023 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8024 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8025
8026 switch (get_imsm_raid_level(map)) {
8027 case 0:
8028 return IMSM_T_STATE_FAILED;
8029 break;
8030 case 1:
8031 if (failed < map->num_members)
8032 return IMSM_T_STATE_DEGRADED;
8033 else
8034 return IMSM_T_STATE_FAILED;
8035 break;
8036 case 10:
8037 {
8038 /**
c92a2527
DW		 8039 * check to see if any mirrors have failed, otherwise we
8040 * are degraded. Even numbered slots are mirrored on
8041 * slot+1
c2a1e7da 8042 */
c2a1e7da 8043 int i;
d9b420a5
N		 8044 /* gcc -Os complains that this is unused */
8045 int insync = insync;
c2a1e7da
DW		 8046
8047 for (i = 0; i < map->num_members; i++) {
238c0a71 8048 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8049 int idx = ord_to_idx(ord);
8050 struct imsm_disk *disk;
c2a1e7da 8051
c92a2527 8052 /* reset the potential in-sync count on even-numbered
1011e834 8053 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8054 */
8055 if ((i & 1) == 0)
8056 insync = 2;
c2a1e7da 8057
c92a2527 8058 disk = get_imsm_disk(super, idx);
25ed7e59 8059 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8060 insync--;
c2a1e7da 8061
c92a2527
DW		 8062 /* no in-sync disks left in this mirror the
8063 * array has failed
8064 */
8065 if (insync == 0)
8066 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8067 }
8068
8069 return IMSM_T_STATE_DEGRADED;
8070 }
8071 case 5:
8072 if (failed < 2)
8073 return IMSM_T_STATE_DEGRADED;
8074 else
8075 return IMSM_T_STATE_FAILED;
8076 break;
8077 default:
8078 break;
8079 }
8080
8081 return map->map_state;
8082}
8083
3b451610
AK		 8084static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8085 int look_in_map)
c2a1e7da
DW		 8086{
8087 int i;
8088 int failed = 0;
8089 struct imsm_disk *disk;
d5985138
AK		 8090 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8091 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8092 struct imsm_map *map_for_loop;
0556e1a2
DW		 8093 __u32 ord;
8094 int idx;
d5985138 8095 int idx_1;
c2a1e7da 8096
0556e1a2
DW		 8097 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8098 * disks that are being rebuilt. New failures are recorded to
8099 * map[0]. So we look through all the disks we started with and
8100 * see if any failures are still present, or if any new ones
8101 * have arrived
0556e1a2 8102 */
d5985138
AK		 8103 map_for_loop = map;
8104 if (prev && (map->num_members < prev->num_members))
8105 map_for_loop = prev;
68fe4598
LD		 8106
8107 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8108 idx_1 = -255;
238c0a71
AK		 8109 /* when MAP_X is passed both maps failures are counted
8110 */
d5985138 8111 if (prev &&
089f9d79
JS		 8112 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8113 i < prev->num_members) {
d5985138
AK		 8114 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8115 idx_1 = ord_to_idx(ord);
c2a1e7da 8116
d5985138
AK		 8117 disk = get_imsm_disk(super, idx_1);
8118 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8119 failed++;
8120 }
089f9d79
JS		 8121 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8122 i < map->num_members) {
d5985138
AK		 8123 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8124 idx = ord_to_idx(ord);
8125
8126 if (idx != idx_1) {
8127 disk = get_imsm_disk(super, idx);
8128 if (!disk || is_failed(disk) ||
8129 ord & IMSM_ORD_REBUILD)
8130 failed++;
8131 }
8132 }
c2a1e7da
DW		 8133 }
8134
8135 return failed;
845dea95
NB		 8136}
8137
97b4d0e9
DW		 8138static int imsm_open_new(struct supertype *c, struct active_array *a,
8139 char *inst)
8140{
8141 struct intel_super *super = c->sb;
8142 struct imsm_super *mpb = super->anchor;
bbab0940 8143 struct imsm_update_prealloc_bb_mem u;
9587c373 8144
97b4d0e9 8145 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8146 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8147 return -ENODEV;
8148 }
8149
8150 dprintf("imsm: open_new %s\n", inst);
8151 a->info.container_member = atoi(inst);
bbab0940
TM		 8152
8153 u.type = update_prealloc_badblocks_mem;
8154 imsm_update_metadata_locally(c, &u, sizeof(u));
8155
97b4d0e9
DW		 8156 return 0;
8157}
8158
0c046afd
DW		 8159static int is_resyncing(struct imsm_dev *dev)
8160{
8161 struct imsm_map *migr_map;
8162
8163 if (!dev->vol.migr_state)
8164 return 0;
8165
1484e727
DW		 8166 if (migr_type(dev) == MIGR_INIT ||
8167 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8168 return 1;
8169
4c9bc37b
AK		 8170 if (migr_type(dev) == MIGR_GEN_MIGR)
8171 return 0;
8172
238c0a71 8173 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8174
089f9d79
JS		 8175 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8176 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8177 return 1;
8178 else
8179 return 0;
8180}
8181
0556e1a2 8182/* return true if we recorded new information */
4c9e8c1e
TM		 8183static int mark_failure(struct intel_super *super,
8184 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8185{
0556e1a2
DW		 8186 __u32 ord;
8187 int slot;
8188 struct imsm_map *map;
86c54047
DW		 8189 char buf[MAX_RAID_SERIAL_LEN+3];
8190 unsigned int len, shift = 0;
0556e1a2
DW		 8191
8192 /* new failures are always set in map[0] */
238c0a71 8193 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8194
8195 slot = get_imsm_disk_slot(map, idx);
8196 if (slot < 0)
8197 return 0;
8198
8199 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8200 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8201 return 0;
8202
7d0c5e24
LD		 8203 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8204 buf[MAX_RAID_SERIAL_LEN] = '\000';
8205 strcat(buf, ":0");
86c54047
DW		 8206 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8207 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8208 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8209
f2f27e63 8210 disk->status |= FAILED_DISK;
0556e1a2 8211 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8212 /* mark failures in second map if second map exists and this disk
8213 * in this slot.
8214 * This is valid for migration, initialization and rebuild
8215 */
8216 if (dev->vol.migr_state) {
238c0a71 8217 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8218 int slot2 = get_imsm_disk_slot(map2, idx);
8219
089f9d79 8220 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8221 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8222 idx | IMSM_ORD_REBUILD);
8223 }
d7a1fda2
MT		 8224 if (map->failed_disk_num == 0xff ||
8225 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8226 map->failed_disk_num = slot;
4c9e8c1e
TM		 8227
8228 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8229
0556e1a2
DW		 8230 return 1;
8231}
8232
4c9e8c1e
TM		 8233static void mark_missing(struct intel_super *super,
8234 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8235{
4c9e8c1e 8236 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8237
8238 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8239 return;
8240
47ee5a45
DW		 8241 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8242 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8243}
8244
33414a01
DW		 8245static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8246{
33414a01 8247 struct dl *dl;
33414a01
DW		 8248
8249 if (!super->missing)
8250 return;
33414a01 8251
79b68f1b
PC		 8252 /* When orom adds replacement for missing disk it does
8253 * not remove entry of missing disk, but just updates map with
8254 * new added disk. So it is not enough just to test if there is
8255 * any missing disk, we have to look if there are any failed disks
8256 * in map to stop migration */
8257
33414a01 8258 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8259 /* end process for initialization and rebuild only
8260 */
8261 if (is_gen_migration(dev) == 0) {
fb12a745 8262 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8263
fb12a745
TM		 8264 if (failed) {
8265 __u8 map_state;
8266 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8267 struct imsm_map *map1;
8268 int i, ord, ord_map1;
8269 int rebuilt = 1;
3d59f0c0 8270
fb12a745
TM		 8271 for (i = 0; i < map->num_members; i++) {
8272 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8273 if (!(ord & IMSM_ORD_REBUILD))
8274 continue;
8275
8276 map1 = get_imsm_map(dev, MAP_1);
8277 if (!map1)
8278 continue;
8279
8280 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8281 if (ord_map1 & IMSM_ORD_REBUILD)
8282 rebuilt = 0;
8283 }
8284
8285 if (rebuilt) {
8286 map_state = imsm_check_degraded(super, dev,
8287 failed, MAP_0);
8288 end_migration(dev, super, map_state);
8289 }
8290 }
3d59f0c0 8291 }
33414a01 8292 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8293 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8294 super->updates_pending++;
8295}
8296
f3871fdc
AK		 8297static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8298 long long new_size)
70bdf0dc 8299{
70bdf0dc 8300 unsigned long long array_blocks;
9529d343
MD		 8301 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8302 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8303
8304 if (used_disks == 0) {
8305 /* when problems occures
8306 * return current array_blocks value
8307 */
fcc2c9da 8308 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8309
8310 return array_blocks;
8311 }
8312
8313 /* set array size in metadata
8314 */
9529d343 8315 if (new_size <= 0)
f3871fdc
AK		 8316 /* OLCE size change is caused by added disks
8317 */
44490938 8318 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8319 else
f3871fdc
AK		 8320 /* Online Volume Size Change
8321 * Using available free space
8322 */
8323 array_blocks = new_size;
70bdf0dc 8324
b53bfba6 8325 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8326 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8327
8328 return array_blocks;
8329}
8330
28bce06f
AK		 8331static void imsm_set_disk(struct active_array *a, int n, int state);
8332
0e2d1a4e
AK		 8333static void imsm_progress_container_reshape(struct intel_super *super)
8334{
8335 /* if no device has a migr_state, but some device has a
8336 * different number of members than the previous device, start
8337 * changing the number of devices in this device to match
8338 * previous.
8339 */
8340 struct imsm_super *mpb = super->anchor;
8341 int prev_disks = -1;
8342 int i;
1dfaa380 8343 int copy_map_size;
0e2d1a4e
AK		 8344
8345 for (i = 0; i < mpb->num_raid_devs; i++) {
8346 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8347 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8348 struct imsm_map *map2;
8349 int prev_num_members;
0e2d1a4e
AK		 8350
8351 if (dev->vol.migr_state)
8352 return;
8353
8354 if (prev_disks == -1)
8355 prev_disks = map->num_members;
8356 if (prev_disks == map->num_members)
8357 continue;
8358
8359 /* OK, this array needs to enter reshape mode.
8360 * i.e it needs a migr_state
8361 */
8362
1dfaa380 8363 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8364 prev_num_members = map->num_members;
8365 map->num_members = prev_disks;
8366 dev->vol.migr_state = 1;
8367 dev->vol.curr_migr_unit = 0;
ea672ee1 8368 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8369 for (i = prev_num_members;
8370 i < map->num_members; i++)
8371 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8372 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8373 /* Copy the current map */
1dfaa380 8374 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8375 map2->num_members = prev_num_members;
8376
f3871fdc 8377 imsm_set_array_size(dev, -1);
51d83f5d 8378 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8379 super->updates_pending++;
8380 }
8381}
8382
aad6f216 8383/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8384 * states are handled in imsm_set_disk() with one exception, when a
8385 * resync is stopped due to a new failure this routine will set the
8386 * 'degraded' state for the array.
8387 */
01f157d7 8388static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8389{
8390 int inst = a->info.container_member;
8391 struct intel_super *super = a->container->sb;
949c47a0 8392 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8393 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8394 int failed = imsm_count_failed(super, dev, MAP_0);
8395 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8396 __u32 blocks_per_unit;
a862209d 8397
1af97990
AK		 8398 if (dev->vol.migr_state &&
8399 dev->vol.migr_type == MIGR_GEN_MIGR) {
8400 /* array state change is blocked due to reshape action
aad6f216
N		 8401 * We might need to
8402 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8403 * - finish the reshape (if last_checkpoint is big and action != reshape)
8404 * - update curr_migr_unit
1af97990 8405 */
aad6f216
N		 8406 if (a->curr_action == reshape) {
8407 /* still reshaping, maybe update curr_migr_unit */
633b5610 8408 goto mark_checkpoint;
aad6f216
N		 8409 } else {
8410 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8411 /* for some reason we aborted the reshape.
b66e591b
AK		 8412 *
8413 * disable automatic metadata rollback
8414 * user action is required to recover process
aad6f216 8415 */
b66e591b 8416 if (0) {
238c0a71
AK		 8417 struct imsm_map *map2 =
8418 get_imsm_map(dev, MAP_1);
8419 dev->vol.migr_state = 0;
8420 set_migr_type(dev, 0);
8421 dev->vol.curr_migr_unit = 0;
8422 memcpy(map, map2,
8423 sizeof_imsm_map(map2));
8424 super->updates_pending++;
b66e591b 8425 }
aad6f216
N		 8426 }
8427 if (a->last_checkpoint >= a->info.component_size) {
8428 unsigned long long array_blocks;
8429 int used_disks;
e154ced3 8430 struct mdinfo *mdi;
aad6f216 8431
9529d343 8432 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8433 if (used_disks > 0) {
8434 array_blocks =
44490938 8435 per_dev_array_size(map) *
d55adef9 8436 used_disks;
b53bfba6
TM		 8437 array_blocks =
8438 round_size_to_mb(array_blocks,
8439 used_disks);
d55adef9
AK		 8440 a->info.custom_array_size = array_blocks;
8441 /* encourage manager to update array
8442 * size
8443 */
e154ced3 8444
d55adef9 8445 a->check_reshape = 1;
633b5610 8446 }
e154ced3
AK		 8447 /* finalize online capacity expansion/reshape */
8448 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8449 imsm_set_disk(a,
8450 mdi->disk.raid_disk,
8451 mdi->curr_state);
8452
0e2d1a4e 8453 imsm_progress_container_reshape(super);
e154ced3 8454 }
aad6f216 8455 }
1af97990
AK		 8456 }
8457
47ee5a45 8458 /* before we activate this array handle any missing disks */
33414a01
DW		 8459 if (consistent == 2)
8460 handle_missing(super, dev);
1e5c6983 8461
0c046afd 8462 if (consistent == 2 &&
b7941fd6 8463 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8464 map_state != IMSM_T_STATE_NORMAL ||
8465 dev->vol.migr_state))
01f157d7 8466 consistent = 0;
272906ef 8467
b7941fd6 8468 if (is_resync_complete(&a->info)) {
0c046afd 8469 /* complete intialization / resync,
0556e1a2
DW		 8470 * recovery and interrupted recovery is completed in
8471 * ->set_disk
0c046afd
DW		 8472 */
8473 if (is_resyncing(dev)) {
8474 dprintf("imsm: mark resync done\n");
809da78e 8475 end_migration(dev, super, map_state);
115c3803 8476 super->updates_pending++;
484240d8 8477 a->last_checkpoint = 0;
115c3803 8478 }
b9172665
AK		 8479 } else if ((!is_resyncing(dev) && !failed) &&
8480 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8481 /* mark the start of the init process if nothing is failed */
b7941fd6 8482 dprintf("imsm: mark resync start\n");
1484e727 8483 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8484 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8485 else
8e59f3d8 8486 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8487 super->updates_pending++;
115c3803 8488 }
a862209d 8489
633b5610 8490mark_checkpoint:
5b83bacf
AK		 8491 /* skip checkpointing for general migration,
8492 * it is controlled in mdadm
8493 */
8494 if (is_gen_migration(dev))
8495 goto skip_mark_checkpoint;
8496
1e5c6983 8497 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8498 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8499 if (blocks_per_unit) {
1e5c6983
DW		 8500 __u32 units32;
8501 __u64 units;
8502
4f0a7acc 8503 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8504 units32 = units;
8505
8506 /* check that we did not overflow 32-bits, and that
8507 * curr_migr_unit needs updating
8508 */
8509 if (units32 == units &&
bfd80a56 8510 units32 != 0 &&
1e5c6983
DW		 8511 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8512 dprintf("imsm: mark checkpoint (%u)\n", units32);
8513 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8514 super->updates_pending++;
8515 }
8516 }
f8f603f1 8517
5b83bacf 8518skip_mark_checkpoint:
3393c6af 8519 /* mark dirty / clean */
2432ce9b
AP		 8520 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8521 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8522 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8523 if (consistent) {
8524 dev->vol.dirty = RAIDVOL_CLEAN;
8525 } else {
8526 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8527 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8528 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8529 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8530 }
a862209d
DW		 8531 super->updates_pending++;
8532 }
28bce06f 8533
01f157d7 8534 return consistent;
a862209d
DW		 8535}
8536
6f50473f
TM		 8537static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8538{
8539 int inst = a->info.container_member;
8540 struct intel_super *super = a->container->sb;
8541 struct imsm_dev *dev = get_imsm_dev(super, inst);
8542 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8543
8544 if (slot > map->num_members) {
8545 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8546 slot, map->num_members - 1);
8547 return -1;
8548 }
8549
8550 if (slot < 0)
8551 return -1;
8552
8553 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8554}
8555
8d45d196 8556static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8557{
8d45d196
DW		 8558 int inst = a->info.container_member;
8559 struct intel_super *super = a->container->sb;
949c47a0 8560 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8561 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8562 struct imsm_disk *disk;
7ce05701
LD		 8563 struct mdinfo *mdi;
8564 int recovery_not_finished = 0;
0c046afd 8565 int failed;
6f50473f 8566 int ord;
0c046afd 8567 __u8 map_state;
fb12a745
TM		 8568 int rebuild_done = 0;
8569 int i;
8d45d196 8570
fb12a745 8571 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8572 if (ord < 0)
8d45d196
DW		 8573 return;
8574
4e6e574a 8575 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8576 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8577
5802a811 8578 /* check for new failures */
ae7d61e3 8579 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8580 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8581 super->updates_pending++;
8d45d196 8582 }
47ee5a45 8583
19859edc 8584 /* check if in_sync */
0556e1a2 8585 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8586 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8587
8588 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8589 rebuild_done = 1;
19859edc
DW		 8590 super->updates_pending++;
8591 }
8d45d196 8592
3b451610
AK		 8593 failed = imsm_count_failed(super, dev, MAP_0);
8594 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8595
0c046afd 8596 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8597 dprintf("imsm: Detected transition to state ");
8598 switch (map_state) {
8599 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8600 dprintf("normal: ");
8601 if (is_rebuilding(dev)) {
1ade5cc1 8602 dprintf_cont("while rebuilding");
7ce05701
LD		 8603 /* check if recovery is really finished */
8604 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8605 if (mdi->recovery_start != MaxSector) {
8606 recovery_not_finished = 1;
8607 break;
8608 }
8609 if (recovery_not_finished) {
1ade5cc1
N		 8610 dprintf_cont("\n");
8611 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8612 if (a->last_checkpoint < mdi->recovery_start) {
8613 a->last_checkpoint = mdi->recovery_start;
8614 super->updates_pending++;
8615 }
8616 break;
8617 }
94002678 8618 end_migration(dev, super, map_state);
238c0a71 8619 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8620 map->failed_disk_num = ~0;
8621 super->updates_pending++;
8622 a->last_checkpoint = 0;
8623 break;
8624 }
8625 if (is_gen_migration(dev)) {
1ade5cc1 8626 dprintf_cont("while general migration");
bf2f0071 8627 if (a->last_checkpoint >= a->info.component_size)
809da78e 8628 end_migration(dev, super, map_state);
94002678
AK		 8629 else
8630 map->map_state = map_state;
238c0a71 8631 map = get_imsm_map(dev, MAP_0);
28bce06f 8632 map->failed_disk_num = ~0;
94002678 8633 super->updates_pending++;
bf2f0071 8634 break;
94002678
AK		 8635 }
8636 break;
8637 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8638 dprintf_cont("degraded: ");
089f9d79 8639 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8640 dprintf_cont("mark degraded");
94002678
AK		 8641 map->map_state = map_state;
8642 super->updates_pending++;
8643 a->last_checkpoint = 0;
8644 break;
8645 }
8646 if (is_rebuilding(dev)) {
d7a1fda2 8647 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8648 if (state & DS_FAULTY) {
8649 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8650 if (n == map->failed_disk_num) {
8651 dprintf_cont("end migration");
8652 end_migration(dev, super, map_state);
a4e96fd8 8653 a->last_checkpoint = 0;
d7a1fda2 8654 } else {
a4e96fd8 8655 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8656 map->map_state = map_state;
8657 }
94002678 8658 super->updates_pending++;
fb12a745
TM		 8659 }
8660
a4e96fd8
MT		 8661 if (!rebuild_done)
8662 break;
8663
fb12a745
TM		 8664 /* check if recovery is really finished */
8665 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8666 if (mdi->recovery_start != MaxSector) {
8667 recovery_not_finished = 1;
8668 break;
8669 }
8670 if (recovery_not_finished) {
8671 dprintf_cont("\n");
a4e96fd8 8672 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8673 if (a->last_checkpoint < mdi->recovery_start) {
8674 a->last_checkpoint =
8675 mdi->recovery_start;
8676 super->updates_pending++;
8677 }
8678 break;
94002678 8679 }
fb12a745
TM		 8680
8681 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8682 end_migration(dev, super, map_state);
8683 a->last_checkpoint = 0;
8684 super->updates_pending++;
fb12a745
TM		 8685
8686 for (i = 0; i < map->num_members; i++) {
8687 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8688
8689 if (idx & IMSM_ORD_REBUILD)
8690 map->failed_disk_num = i;
8691 }
8692 super->updates_pending++;
94002678
AK		 8693 break;
8694 }
8695 if (is_gen_migration(dev)) {
1ade5cc1 8696 dprintf_cont("while general migration");
bf2f0071 8697 if (a->last_checkpoint >= a->info.component_size)
809da78e 8698 end_migration(dev, super, map_state);
94002678
AK		 8699 else {
8700 map->map_state = map_state;
3b451610 8701 manage_second_map(super, dev);
94002678
AK		 8702 }
8703 super->updates_pending++;
bf2f0071 8704 break;
28bce06f 8705 }
6ce1fbf1 8706 if (is_initializing(dev)) {
1ade5cc1 8707 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8708 map->map_state = map_state;
8709 super->updates_pending++;
8710 break;
8711 }
94002678
AK		 8712 break;
8713 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8714 dprintf_cont("failed: ");
94002678 8715 if (is_gen_migration(dev)) {
1ade5cc1 8716 dprintf_cont("while general migration");
94002678
AK		 8717 map->map_state = map_state;
8718 super->updates_pending++;
8719 break;
8720 }
8721 if (map->map_state != map_state) {
1ade5cc1 8722 dprintf_cont("mark failed");
94002678
AK		 8723 end_migration(dev, super, map_state);
8724 super->updates_pending++;
8725 a->last_checkpoint = 0;
8726 break;
8727 }
8728 break;
8729 default:
1ade5cc1 8730 dprintf_cont("state %i\n", map_state);
5802a811 8731 }
1ade5cc1 8732 dprintf_cont("\n");
845dea95
NB		 8733}
8734
f796af5d 8735static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8736{
f796af5d 8737 void *buf = mpb;
c2a1e7da
DW		 8738 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8739 unsigned long long dsize;
8740 unsigned long long sectors;
f36a9ecd 8741 unsigned int sector_size;
c2a1e7da 8742
f36a9ecd 8743 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8744 get_dev_size(fd, NULL, &dsize);
8745
f36a9ecd 8746 if (mpb_size > sector_size) {
272f648f 8747 /* -1 to account for anchor */
f36a9ecd 8748 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8749
272f648f 8750 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8751 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8752 SEEK_SET) < 0)
272f648f 8753 return 1;
c2a1e7da 8754
f36a9ecd
PB		 8755 if ((unsigned long long)write(fd, buf + sector_size,
8756 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8757 return 1;
8758 }
c2a1e7da 8759
272f648f 8760 /* first block is stored on second to last sector of the disk */
f36a9ecd 8761 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8762 return 1;
8763
466070ad 8764 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8765 return 1;
8766
c2a1e7da
DW		 8767 return 0;
8768}
8769
2e735d19 8770static void imsm_sync_metadata(struct supertype *container)
845dea95 8771{
2e735d19 8772 struct intel_super *super = container->sb;
c2a1e7da 8773
1a64be56 8774 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8775 if (!super->updates_pending)
8776 return;
8777
36988a3d 8778 write_super_imsm(container, 0);
c2a1e7da
DW		 8779
8780 super->updates_pending = 0;
845dea95
NB		 8781}
8782
272906ef
DW		 8783static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8784{
8785 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8786 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8787 struct dl *dl;
8788
8789 for (dl = super->disks; dl; dl = dl->next)
8790 if (dl->index == i)
8791 break;
8792
25ed7e59 8793 if (dl && is_failed(&dl->disk))
272906ef
DW		 8794 dl = NULL;
8795
8796 if (dl)
1ade5cc1 8797 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8798
8799 return dl;
8800}
8801
a20d2ba5 8802static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8803 struct active_array *a, int activate_new,
8804 struct mdinfo *additional_test_list)
272906ef
DW		 8805{
8806 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8807 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8808 struct imsm_super *mpb = super->anchor;
8809 struct imsm_map *map;
272906ef
DW		 8810 unsigned long long pos;
8811 struct mdinfo *d;
8812 struct extent *ex;
a20d2ba5 8813 int i, j;
272906ef 8814 int found;
569cc43f
DW		 8815 __u32 array_start = 0;
8816 __u32 array_end = 0;
272906ef 8817 struct dl *dl;
6c932028 8818 struct mdinfo *test_list;
272906ef
DW		 8819
8820 for (dl = super->disks; dl; dl = dl->next) {
8821 /* If in this array, skip */
8822 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8823 if (d->state_fd >= 0 &&
8824 d->disk.major == dl->major &&
272906ef 8825 d->disk.minor == dl->minor) {
8ba77d32
AK		 8826 dprintf("%x:%x already in array\n",
8827 dl->major, dl->minor);
272906ef
DW		 8828 break;
8829 }
8830 if (d)
8831 continue;
6c932028
AK		 8832 test_list = additional_test_list;
8833 while (test_list) {
8834 if (test_list->disk.major == dl->major &&
8835 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8836 dprintf("%x:%x already in additional test list\n",
8837 dl->major, dl->minor);
8838 break;
8839 }
6c932028 8840 test_list = test_list->next;
8ba77d32 8841 }
6c932028 8842 if (test_list)
8ba77d32 8843 continue;
272906ef 8844
e553d2a4 8845 /* skip in use or failed drives */
25ed7e59 8846 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8847 dl->index == -2) {
8848 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8849 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8850 continue;
8851 }
8852
a20d2ba5
DW		 8853 /* skip pure spares when we are looking for partially
8854 * assimilated drives
8855 */
8856 if (dl->index == -1 && !activate_new)
8857 continue;
8858
f2cc4f7d
AO		 8859 if (!drive_validate_sector_size(super, dl))
8860 continue;
8861
272906ef 8862 /* Does this unused device have the requisite free space?
a20d2ba5 8863 * It needs to be able to cover all member volumes
272906ef 8864 */
05501181 8865 ex = get_extents(super, dl, 1);
272906ef
DW		 8866 if (!ex) {
8867 dprintf("cannot get extents\n");
8868 continue;
8869 }
a20d2ba5
DW		 8870 for (i = 0; i < mpb->num_raid_devs; i++) {
8871 dev = get_imsm_dev(super, i);
238c0a71 8872 map = get_imsm_map(dev, MAP_0);
272906ef 8873
a20d2ba5
DW		 8874 /* check if this disk is already a member of
8875 * this array
272906ef 8876 */
620b1713 8877 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8878 continue;
8879
8880 found = 0;
8881 j = 0;
8882 pos = 0;
5551b113 8883 array_start = pba_of_lba0(map);
329c8278 8884 array_end = array_start +
44490938 8885 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8886
8887 do {
8888 /* check that we can start at pba_of_lba0 with
44490938 8889 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8890 */
329c8278 8891 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8892 found = 1;
8893 break;
8894 }
8895 pos = ex[j].start + ex[j].size;
8896 j++;
8897 } while (ex[j-1].size);
8898
8899 if (!found)
272906ef 8900 break;
a20d2ba5 8901 }
272906ef
DW		 8902
8903 free(ex);
a20d2ba5 8904 if (i < mpb->num_raid_devs) {
329c8278
DW		 8905 dprintf("%x:%x does not have %u to %u available\n",
8906 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8907 /* No room */
8908 continue;
a20d2ba5
DW		 8909 }
8910 return dl;
272906ef
DW		 8911 }
8912
8913 return dl;
8914}
8915
95d07a2c
LM		 8916static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8917{
8918 struct imsm_dev *dev2;
8919 struct imsm_map *map;
8920 struct dl *idisk;
8921 int slot;
8922 int idx;
8923 __u8 state;
8924
8925 dev2 = get_imsm_dev(cont->sb, dev_idx);
8926 if (dev2) {
238c0a71 8927 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8928 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8929 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8930 if (!map)
8931 return 1;
8932 for (slot = 0; slot < map->num_members; slot++) {
8933 /*
8934 * Check if failed disks are deleted from intel
8935 * disk list or are marked to be deleted
8936 */
238c0a71 8937 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8938 idisk = get_imsm_dl_disk(cont->sb, idx);
8939 /*
8940 * Do not rebuild the array if failed disks
8941 * from failed sub-array are not removed from
8942 * container.
8943 */
8944 if (idisk &&
8945 is_failed(&idisk->disk) &&
8946 (idisk->action != DISK_REMOVE))
8947 return 0;
8948 }
8949 }
8950 }
8951 return 1;
8952}
8953
88758e9d
DW		 8954static struct mdinfo *imsm_activate_spare(struct active_array *a,
8955 struct metadata_update **updates)
8956{
8957 /**
d23fe947
DW		 8958 * Find a device with unused free space and use it to replace a
8959 * failed/vacant region in an array. We replace failed regions one a
8960 * array at a time. The result is that a new spare disk will be added
8961 * to the first failed array and after the monitor has finished
8962 * propagating failures the remainder will be consumed.
88758e9d 8963 *
d23fe947
DW		 8964 * FIXME add a capability for mdmon to request spares from another
8965 * container.
88758e9d
DW		 8966 */
8967
8968 struct intel_super *super = a->container->sb;
88758e9d 8969 int inst = a->info.container_member;
949c47a0 8970 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8971 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8972 int failed = a->info.array.raid_disks;
8973 struct mdinfo *rv = NULL;
8974 struct mdinfo *d;
8975 struct mdinfo *di;
8976 struct metadata_update *mu;
8977 struct dl *dl;
8978 struct imsm_update_activate_spare *u;
8979 int num_spares = 0;
8980 int i;
95d07a2c 8981 int allowed;
88758e9d
DW		 8982
8983 for (d = a->info.devs ; d ; d = d->next) {
8984 if ((d->curr_state & DS_FAULTY) &&
8985 d->state_fd >= 0)
8986 /* wait for Removal to happen */
8987 return NULL;
8988 if (d->state_fd >= 0)
8989 failed--;
8990 }
8991
8992 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8993 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8994
e2962bfc
AK		 8995 if (imsm_reshape_blocks_arrays_changes(super))
8996 return NULL;
1af97990 8997
fc8ca064
AK		 8998 /* Cannot activate another spare if rebuild is in progress already
8999 */
9000 if (is_rebuilding(dev)) {
7a862a02 9001 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 9002 return NULL;
9003 }
9004
89c67882
AK		 9005 if (a->info.array.level == 4)
9006 /* No repair for takeovered array
9007 * imsm doesn't support raid4
9008 */
9009 return NULL;
9010
3b451610
AK		 9011 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9012 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9013 return NULL;
9014
83ca7d45
AP		 9015 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9016 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9017 return NULL;
9018 }
9019
95d07a2c
LM		 9020 /*
9021 * If there are any failed disks check state of the other volume.
9022 * Block rebuild if the another one is failed until failed disks
9023 * are removed from container.
9024 */
9025 if (failed) {
7a862a02 9026 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9027 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9028 /* check if states of the other volumes allow for rebuild */
9029 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9030 if (i != inst) {
9031 allowed = imsm_rebuild_allowed(a->container,
9032 i, failed);
9033 if (!allowed)
9034 return NULL;
9035 }
9036 }
9037 }
9038
88758e9d 9039 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9040 for (i = 0; i < a->info.array.raid_disks; i++) {
9041 for (d = a->info.devs ; d ; d = d->next)
9042 if (d->disk.raid_disk == i)
9043 break;
9044 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9045 if (d && (d->state_fd >= 0))
9046 continue;
9047
272906ef 9048 /*
a20d2ba5
DW		 9049 * OK, this device needs recovery. Try to re-add the
9050 * previous occupant of this slot, if this fails see if
9051 * we can continue the assimilation of a spare that was
9052 * partially assimilated, finally try to activate a new
9053 * spare.
272906ef
DW		 9054 */
9055 dl = imsm_readd(super, i, a);
9056 if (!dl)
b303fe21 9057 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9058 if (!dl)
b303fe21 9059 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9060 if (!dl)
9061 continue;
1011e834 9062
272906ef 9063 /* found a usable disk with enough space */
503975b9 9064 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9065
9066 /* dl->index will be -1 in the case we are activating a
9067 * pristine spare. imsm_process_update() will create a
9068 * new index in this case. Once a disk is found to be
9069 * failed in all member arrays it is kicked from the
9070 * metadata
9071 */
9072 di->disk.number = dl->index;
d23fe947 9073
272906ef
DW		 9074 /* (ab)use di->devs to store a pointer to the device
9075 * we chose
9076 */
9077 di->devs = (struct mdinfo *) dl;
9078
9079 di->disk.raid_disk = i;
9080 di->disk.major = dl->major;
9081 di->disk.minor = dl->minor;
9082 di->disk.state = 0;
d23534e4 9083 di->recovery_start = 0;
5551b113 9084 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9085 di->component_size = a->info.component_size;
9086 di->container_member = inst;
5e46202e 9087 di->bb.supported = 1;
2c8890e9 9088 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9089 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9090 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9091 }
148acb7b 9092 super->random = random32();
272906ef
DW		 9093 di->next = rv;
9094 rv = di;
9095 num_spares++;
9096 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9097 i, di->data_offset);
88758e9d
DW		 9098 }
9099
9100 if (!rv)
9101 /* No spares found */
9102 return rv;
9103 /* Now 'rv' has a list of devices to return.
9104 * Create a metadata_update record to update the
9105 * disk_ord_tbl for the array
9106 */
503975b9 9107 mu = xmalloc(sizeof(*mu));
1011e834 9108 mu->buf = xcalloc(num_spares,
503975b9 9109 sizeof(struct imsm_update_activate_spare));
88758e9d 9110 mu->space = NULL;
cb23f1f4 9111 mu->space_list = NULL;
88758e9d
DW		 9112 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9113 mu->next = *updates;
9114 u = (struct imsm_update_activate_spare *) mu->buf;
9115
9116 for (di = rv ; di ; di = di->next) {
9117 u->type = update_activate_spare;
d23fe947
DW		 9118 u->dl = (struct dl *) di->devs;
9119 di->devs = NULL;
88758e9d
DW		 9120 u->slot = di->disk.raid_disk;
9121 u->array = inst;
9122 u->next = u + 1;
9123 u++;
9124 }
9125 (u-1)->next = NULL;
9126 *updates = mu;
9127
9128 return rv;
9129}
9130
54c2c1ea 9131static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9132{
54c2c1ea 9133 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9134 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9135 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9136 struct disk_info *inf = get_disk_info(u);
9137 struct imsm_disk *disk;
8273f55e
DW		 9138 int i;
9139 int j;
8273f55e 9140
54c2c1ea 9141 for (i = 0; i < map->num_members; i++) {
238c0a71 9142 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9143 for (j = 0; j < new_map->num_members; j++)
9144 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9145 return 1;
9146 }
9147
9148 return 0;
9149}
9150
1a64be56
LM		 9151static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9152{
594dc1b8
JS		 9153 struct dl *dl;
9154
1a64be56 9155 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9156 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9157 return dl;
9158 return NULL;
9159}
9160
9161static int remove_disk_super(struct intel_super *super, int major, int minor)
9162{
594dc1b8 9163 struct dl *prev;
1a64be56
LM		 9164 struct dl *dl;
9165
9166 prev = NULL;
9167 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9168 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9169 /* remove */
9170 if (prev)
9171 prev->next = dl->next;
9172 else
9173 super->disks = dl->next;
9174 dl->next = NULL;
9175 __free_imsm_disk(dl);
1ade5cc1 9176 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9177 break;
9178 }
9179 prev = dl;
9180 }
9181 return 0;
9182}
9183
f21e18ca 9184static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9185
1a64be56
LM		 9186static int add_remove_disk_update(struct intel_super *super)
9187{
9188 int check_degraded = 0;
594dc1b8
JS		 9189 struct dl *disk;
9190
1a64be56
LM		 9191 /* add/remove some spares to/from the metadata/contrainer */
9192 while (super->disk_mgmt_list) {
9193 struct dl *disk_cfg;
9194
9195 disk_cfg = super->disk_mgmt_list;
9196 super->disk_mgmt_list = disk_cfg->next;
9197 disk_cfg->next = NULL;
9198
9199 if (disk_cfg->action == DISK_ADD) {
9200 disk_cfg->next = super->disks;
9201 super->disks = disk_cfg;
9202 check_degraded = 1;
1ade5cc1
N		 9203 dprintf("added %x:%x\n",
9204 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9205 } else if (disk_cfg->action == DISK_REMOVE) {
9206 dprintf("Disk remove action processed: %x.%x\n",
9207 disk_cfg->major, disk_cfg->minor);
9208 disk = get_disk_super(super,
9209 disk_cfg->major,
9210 disk_cfg->minor);
9211 if (disk) {
9212 /* store action status */
9213 disk->action = DISK_REMOVE;
9214 /* remove spare disks only */
9215 if (disk->index == -1) {
9216 remove_disk_super(super,
9217 disk_cfg->major,
9218 disk_cfg->minor);
91c97c54
MT		 9219 } else {
9220 disk_cfg->fd = disk->fd;
9221 disk->fd = -1;
1a64be56
LM		 9222 }
9223 }
9224 /* release allocate disk structure */
9225 __free_imsm_disk(disk_cfg);
9226 }
9227 }
9228 return check_degraded;
9229}
9230
a29911da
PC		 9231static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9232 struct intel_super *super,
9233 void ***space_list)
9234{
9235 struct intel_dev *id;
9236 void **tofree = NULL;
9237 int ret_val = 0;
9238
1ade5cc1 9239 dprintf("(enter)\n");
089f9d79 9240 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9241 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9242 return ret_val;
9243 }
089f9d79 9244 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9245 dprintf("imsm: Error: Memory is not allocated\n");
9246 return ret_val;
9247 }
9248
9249 for (id = super->devlist ; id; id = id->next) {
9250 if (id->index == (unsigned)u->subdev) {
9251 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9252 struct imsm_map *map;
9253 struct imsm_dev *new_dev =
9254 (struct imsm_dev *)*space_list;
238c0a71 9255 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9256 int to_state;
9257 struct dl *new_disk;
9258
9259 if (new_dev == NULL)
9260 return ret_val;
9261 *space_list = **space_list;
9262 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9263 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9264 if (migr_map) {
9265 dprintf("imsm: Error: migration in progress");
9266 return ret_val;
9267 }
9268
9269 to_state = map->map_state;
9270 if ((u->new_level == 5) && (map->raid_level == 0)) {
9271 map->num_members++;
9272 /* this should not happen */
9273 if (u->new_disks[0] < 0) {
9274 map->failed_disk_num =
9275 map->num_members - 1;
9276 to_state = IMSM_T_STATE_DEGRADED;
9277 } else
9278 to_state = IMSM_T_STATE_NORMAL;
9279 }
8e59f3d8 9280 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9281 if (u->new_level > -1)
9282 map->raid_level = u->new_level;
238c0a71 9283 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9284 if ((u->new_level == 5) &&
9285 (migr_map->raid_level == 0)) {
9286 int ord = map->num_members - 1;
9287 migr_map->num_members--;
9288 if (u->new_disks[0] < 0)
9289 ord |= IMSM_ORD_REBUILD;
9290 set_imsm_ord_tbl_ent(map,
9291 map->num_members - 1,
9292 ord);
9293 }
9294 id->dev = new_dev;
9295 tofree = (void **)dev;
9296
4bba0439
PC		 9297 /* update chunk size
9298 */
06fb291a
PB		 9299 if (u->new_chunksize > 0) {
9300 unsigned long long num_data_stripes;
9529d343
MD		 9301 struct imsm_map *dest_map =
9302 get_imsm_map(dev, MAP_0);
06fb291a 9303 int used_disks =
9529d343 9304 imsm_num_data_members(dest_map);
06fb291a
PB		 9305
9306 if (used_disks == 0)
9307 return ret_val;
9308
4bba0439
PC		 9309 map->blocks_per_strip =
9310 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9311 num_data_stripes =
fcc2c9da 9312 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9313 num_data_stripes /= map->blocks_per_strip;
9314 num_data_stripes /= map->num_domains;
9315 set_num_data_stripes(map, num_data_stripes);
9316 }
4bba0439 9317
44490938
MD		 9318 /* ensure blocks_per_member has valid value
9319 */
9320 set_blocks_per_member(map,
9321 per_dev_array_size(map) +
9322 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9323
a29911da
PC		 9324 /* add disk
9325 */
089f9d79
JS		 9326 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9327 migr_map->raid_level == map->raid_level)
a29911da
PC		 9328 goto skip_disk_add;
9329
9330 if (u->new_disks[0] >= 0) {
9331 /* use passes spare
9332 */
9333 new_disk = get_disk_super(super,
9334 major(u->new_disks[0]),
9335 minor(u->new_disks[0]));
7a862a02 9336 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9337 major(u->new_disks[0]),
9338 minor(u->new_disks[0]),
9339 new_disk, new_disk->index);
9340 if (new_disk == NULL)
9341 goto error_disk_add;
9342
9343 new_disk->index = map->num_members - 1;
9344 /* slot to fill in autolayout
9345 */
9346 new_disk->raiddisk = new_disk->index;
9347 new_disk->disk.status |= CONFIGURED_DISK;
9348 new_disk->disk.status &= ~SPARE_DISK;
9349 } else
9350 goto error_disk_add;
9351
9352skip_disk_add:
9353 *tofree = *space_list;
9354 /* calculate new size
9355 */
f3871fdc 9356 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9357
9358 ret_val = 1;
9359 }
9360 }
9361
9362 if (tofree)
9363 *space_list = tofree;
9364 return ret_val;
9365
9366error_disk_add:
9367 dprintf("Error: imsm: Cannot find disk.\n");
9368 return ret_val;
9369}
9370
f3871fdc
AK		 9371static int apply_size_change_update(struct imsm_update_size_change *u,
9372 struct intel_super *super)
9373{
9374 struct intel_dev *id;
9375 int ret_val = 0;
9376
1ade5cc1 9377 dprintf("(enter)\n");
089f9d79 9378 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9379 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9380 return ret_val;
9381 }
9382
9383 for (id = super->devlist ; id; id = id->next) {
9384 if (id->index == (unsigned)u->subdev) {
9385 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9386 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9387 int used_disks = imsm_num_data_members(map);
f3871fdc 9388 unsigned long long blocks_per_member;
06fb291a 9389 unsigned long long num_data_stripes;
44490938
MD		 9390 unsigned long long new_size_per_disk;
9391
9392 if (used_disks == 0)
9393 return 0;
f3871fdc
AK		 9394
9395 /* calculate new size
9396 */
44490938
MD		 9397 new_size_per_disk = u->new_size / used_disks;
9398 blocks_per_member = new_size_per_disk +
9399 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9400 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9401 map->blocks_per_strip;
9402 num_data_stripes /= map->num_domains;
9403 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9404 u->new_size, new_size_per_disk,
06fb291a 9405 num_data_stripes);
f3871fdc 9406 set_blocks_per_member(map, blocks_per_member);
06fb291a 9407 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9408 imsm_set_array_size(dev, u->new_size);
9409
9410 ret_val = 1;
9411 break;
9412 }
9413 }
9414
9415 return ret_val;
9416}
9417
061d7da3 9418static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9419 struct intel_super *super,
061d7da3
LO		 9420 struct active_array *active_array)
9421{
9422 struct imsm_super *mpb = super->anchor;
9423 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9424 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9425 struct imsm_map *migr_map;
9426 struct active_array *a;
9427 struct imsm_disk *disk;
9428 __u8 to_state;
9429 struct dl *dl;
9430 unsigned int found;
9431 int failed;
5961eeec 9432 int victim;
061d7da3 9433 int i;
5961eeec 9434 int second_map_created = 0;
061d7da3 9435
5961eeec 9436 for (; u; u = u->next) {
238c0a71 9437 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9438
5961eeec 9439 if (victim < 0)
9440 return 0;
061d7da3 9441
5961eeec 9442 for (dl = super->disks; dl; dl = dl->next)
9443 if (dl == u->dl)
9444 break;
061d7da3 9445
5961eeec 9446 if (!dl) {
7a862a02 9447 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9448 u->dl->index);
9449 return 0;
9450 }
061d7da3 9451
5961eeec 9452 /* count failures (excluding rebuilds and the victim)
9453 * to determine map[0] state
9454 */
9455 failed = 0;
9456 for (i = 0; i < map->num_members; i++) {
9457 if (i == u->slot)
9458 continue;
9459 disk = get_imsm_disk(super,
238c0a71 9460 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9461 if (!disk || is_failed(disk))
9462 failed++;
9463 }
061d7da3 9464
5961eeec 9465 /* adding a pristine spare, assign a new index */
9466 if (dl->index < 0) {
9467 dl->index = super->anchor->num_disks;
9468 super->anchor->num_disks++;
9469 }
9470 disk = &dl->disk;
9471 disk->status |= CONFIGURED_DISK;
9472 disk->status &= ~SPARE_DISK;
9473
9474 /* mark rebuild */
238c0a71 9475 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9476 if (!second_map_created) {
9477 second_map_created = 1;
9478 map->map_state = IMSM_T_STATE_DEGRADED;
9479 migrate(dev, super, to_state, MIGR_REBUILD);
9480 } else
9481 map->map_state = to_state;
238c0a71 9482 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9483 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9484 set_imsm_ord_tbl_ent(migr_map, u->slot,
9485 dl->index | IMSM_ORD_REBUILD);
9486
9487 /* update the family_num to mark a new container
9488 * generation, being careful to record the existing
9489 * family_num in orig_family_num to clean up after
9490 * earlier mdadm versions that neglected to set it.
9491 */
9492 if (mpb->orig_family_num == 0)
9493 mpb->orig_family_num = mpb->family_num;
9494 mpb->family_num += super->random;
9495
9496 /* count arrays using the victim in the metadata */
9497 found = 0;
9498 for (a = active_array; a ; a = a->next) {
9499 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9500 map = get_imsm_map(dev, MAP_0);
061d7da3 9501
5961eeec 9502 if (get_imsm_disk_slot(map, victim) >= 0)
9503 found++;
9504 }
061d7da3 9505
5961eeec 9506 /* delete the victim if it is no longer being
9507 * utilized anywhere
061d7da3 9508 */
5961eeec 9509 if (!found) {
9510 struct dl **dlp;
061d7da3 9511
5961eeec 9512 /* We know that 'manager' isn't touching anything,
9513 * so it is safe to delete
9514 */
9515 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9516 if ((*dlp)->index == victim)
9517 break;
5961eeec 9518
9519 /* victim may be on the missing list */
9520 if (!*dlp)
9521 for (dlp = &super->missing; *dlp;
9522 dlp = &(*dlp)->next)
9523 if ((*dlp)->index == victim)
9524 break;
9525 imsm_delete(super, dlp, victim);
9526 }
061d7da3
LO		 9527 }
9528
9529 return 1;
9530}
a29911da 9531
2e5dc010
N		 9532static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9533 struct intel_super *super,
9534 void ***space_list)
9535{
9536 struct dl *new_disk;
9537 struct intel_dev *id;
9538 int i;
9539 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9540 int disk_count = u->old_raid_disks;
2e5dc010
N		 9541 void **tofree = NULL;
9542 int devices_to_reshape = 1;
9543 struct imsm_super *mpb = super->anchor;
9544 int ret_val = 0;
d098291a 9545 unsigned int dev_id;
2e5dc010 9546
1ade5cc1 9547 dprintf("(enter)\n");
2e5dc010
N		 9548
9549 /* enable spares to use in array */
9550 for (i = 0; i < delta_disks; i++) {
9551 new_disk = get_disk_super(super,
9552 major(u->new_disks[i]),
9553 minor(u->new_disks[i]));
7a862a02 9554 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9555 major(u->new_disks[i]), minor(u->new_disks[i]),
9556 new_disk, new_disk->index);
089f9d79
JS		 9557 if (new_disk == NULL ||
9558 (new_disk->index >= 0 &&
9559 new_disk->index < u->old_raid_disks))
2e5dc010 9560 goto update_reshape_exit;
ee4beede 9561 new_disk->index = disk_count++;
2e5dc010
N		 9562 /* slot to fill in autolayout
9563 */
9564 new_disk->raiddisk = new_disk->index;
9565 new_disk->disk.status |=
9566 CONFIGURED_DISK;
9567 new_disk->disk.status &= ~SPARE_DISK;
9568 }
9569
ed7333bd
AK		 9570 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9571 mpb->num_raid_devs);
2e5dc010
N		 9572 /* manage changes in volume
9573 */
d098291a 9574 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9575 void **sp = *space_list;
9576 struct imsm_dev *newdev;
9577 struct imsm_map *newmap, *oldmap;
9578
d098291a
AK		 9579 for (id = super->devlist ; id; id = id->next) {
9580 if (id->index == dev_id)
9581 break;
9582 }
9583 if (id == NULL)
9584 break;
2e5dc010
N		 9585 if (!sp)
9586 continue;
9587 *space_list = *sp;
9588 newdev = (void*)sp;
9589 /* Copy the dev, but not (all of) the map */
9590 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9591 oldmap = get_imsm_map(id->dev, MAP_0);
9592 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9593 /* Copy the current map */
9594 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9595 /* update one device only
9596 */
9597 if (devices_to_reshape) {
ed7333bd
AK		 9598 dprintf("imsm: modifying subdev: %i\n",
9599 id->index);
2e5dc010
N		 9600 devices_to_reshape--;
9601 newdev->vol.migr_state = 1;
9602 newdev->vol.curr_migr_unit = 0;
ea672ee1 9603 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9604 newmap->num_members = u->new_raid_disks;
9605 for (i = 0; i < delta_disks; i++) {
9606 set_imsm_ord_tbl_ent(newmap,
9607 u->old_raid_disks + i,
9608 u->old_raid_disks + i);
9609 }
9610 /* New map is correct, now need to save old map
9611 */
238c0a71 9612 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9613 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9614
f3871fdc 9615 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9616 }
9617
9618 sp = (void **)id->dev;
9619 id->dev = newdev;
9620 *sp = tofree;
9621 tofree = sp;
8e59f3d8
AK		 9622
9623 /* Clear migration record */
9624 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9625 }
819bc634
AK		 9626 if (tofree)
9627 *space_list = tofree;
2e5dc010
N		 9628 ret_val = 1;
9629
9630update_reshape_exit:
9631
9632 return ret_val;
9633}
9634
bb025c2f 9635static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9636 struct intel_super *super,
9637 void ***space_list)
bb025c2f
KW		 9638{
9639 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9640 struct intel_dev *dv;
9641 struct imsm_dev *dev_new;
bb025c2f
KW		 9642 struct imsm_map *map;
9643 struct dl *dm, *du;
8ca6df95 9644 int i;
bb025c2f
KW		 9645
9646 for (dv = super->devlist; dv; dv = dv->next)
9647 if (dv->index == (unsigned int)u->subarray) {
9648 dev = dv->dev;
9649 break;
9650 }
9651
9652 if (dev == NULL)
9653 return 0;
9654
238c0a71 9655 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9656
9657 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9658 unsigned long long num_data_stripes;
9659
43d5ec18 9660 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9661 if (imsm_count_failed(super, dev, MAP_0) !=
9662 (map->num_members / 2))
43d5ec18
KW		 9663 return 0;
9664
bb025c2f
KW		 9665 /* iterate through devices to mark removed disks as spare */
9666 for (dm = super->disks; dm; dm = dm->next) {
9667 if (dm->disk.status & FAILED_DISK) {
9668 int idx = dm->index;
9669 /* update indexes on the disk list */
9670/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9671 the index values will end up being correct.... NB */
9672 for (du = super->disks; du; du = du->next)
9673 if (du->index > idx)
9674 du->index--;
9675 /* mark as spare disk */
a8619d23 9676 mark_spare(dm);
bb025c2f
KW		 9677 }
9678 }
bb025c2f
KW		 9679 /* update map */
9680 map->num_members = map->num_members / 2;
9681 map->map_state = IMSM_T_STATE_NORMAL;
9682 map->num_domains = 1;
9683 map->raid_level = 0;
9684 map->failed_disk_num = -1;
4a353e6e
RS		 9685 num_data_stripes = imsm_dev_size(dev) / 2;
9686 num_data_stripes /= map->blocks_per_strip;
9687 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9688 }
9689
8ca6df95
KW		 9690 if (u->direction == R0_TO_R10) {
9691 void **space;
4a353e6e
RS		 9692 unsigned long long num_data_stripes;
9693
8ca6df95
KW		 9694 /* update slots in current disk list */
9695 for (dm = super->disks; dm; dm = dm->next) {
9696 if (dm->index >= 0)
9697 dm->index *= 2;
9698 }
9699 /* create new *missing* disks */
9700 for (i = 0; i < map->num_members; i++) {
9701 space = *space_list;
9702 if (!space)
9703 continue;
9704 *space_list = *space;
9705 du = (void *)space;
9706 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9707 du->fd = -1;
9708 du->minor = 0;
9709 du->major = 0;
9710 du->index = (i * 2) + 1;
9711 sprintf((char *)du->disk.serial,
9712 " MISSING_%d", du->index);
9713 sprintf((char *)du->serial,
9714 "MISSING_%d", du->index);
9715 du->next = super->missing;
9716 super->missing = du;
9717 }
9718 /* create new dev and map */
9719 space = *space_list;
9720 if (!space)
9721 return 0;
9722 *space_list = *space;
9723 dev_new = (void *)space;
9724 memcpy(dev_new, dev, sizeof(*dev));
9725 /* update new map */
238c0a71 9726 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9727 map->num_members = map->num_members * 2;
1a2487c2 9728 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9729 map->num_domains = 2;
9730 map->raid_level = 1;
4a353e6e
RS		 9731 num_data_stripes = imsm_dev_size(dev) / 2;
9732 num_data_stripes /= map->blocks_per_strip;
9733 num_data_stripes /= map->num_domains;
9734 set_num_data_stripes(map, num_data_stripes);
9735
8ca6df95
KW		 9736 /* replace dev<->dev_new */
9737 dv->dev = dev_new;
9738 }
bb025c2f
KW		 9739 /* update disk order table */
9740 for (du = super->disks; du; du = du->next)
9741 if (du->index >= 0)
9742 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9743 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9744 if (du->index >= 0) {
9745 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9746 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9747 }
bb025c2f
KW		 9748
9749 return 1;
9750}
9751
e8319a19
DW		 9752static void imsm_process_update(struct supertype *st,
9753 struct metadata_update *update)
9754{
9755 /**
9756 * crack open the metadata_update envelope to find the update record
9757 * update can be one of:
d195167d
AK		 9758 * update_reshape_container_disks - all the arrays in the container
9759 * are being reshaped to have more devices. We need to mark
9760 * the arrays for general migration and convert selected spares
9761 * into active devices.
9762 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9763 * device in an array, update the disk_ord_tbl. If this disk is
9764 * present in all member arrays then also clear the SPARE_DISK
9765 * flag
d195167d
AK		 9766 * update_create_array
9767 * update_kill_array
9768 * update_rename_array
9769 * update_add_remove_disk
e8319a19
DW		 9770 */
9771 struct intel_super *super = st->sb;
4d7b1503 9772 struct imsm_super *mpb;
e8319a19
DW		 9773 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9774
4d7b1503
DW		 9775 /* update requires a larger buf but the allocation failed */
9776 if (super->next_len && !super->next_buf) {
9777 super->next_len = 0;
9778 return;
9779 }
9780
9781 if (super->next_buf) {
9782 memcpy(super->next_buf, super->buf, super->len);
9783 free(super->buf);
9784 super->len = super->next_len;
9785 super->buf = super->next_buf;
9786
9787 super->next_len = 0;
9788 super->next_buf = NULL;
9789 }
9790
9791 mpb = super->anchor;
9792
e8319a19 9793 switch (type) {
0ec5d470
AK		 9794 case update_general_migration_checkpoint: {
9795 struct intel_dev *id;
9796 struct imsm_update_general_migration_checkpoint *u =
9797 (void *)update->buf;
9798
1ade5cc1 9799 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9800
9801 /* find device under general migration */
9802 for (id = super->devlist ; id; id = id->next) {
9803 if (is_gen_migration(id->dev)) {
9804 id->dev->vol.curr_migr_unit =
9805 __cpu_to_le32(u->curr_migr_unit);
9806 super->updates_pending++;
9807 }
9808 }
9809 break;
9810 }
bb025c2f
KW		 9811 case update_takeover: {
9812 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9813 if (apply_takeover_update(u, super, &update->space_list)) {
9814 imsm_update_version_info(super);
bb025c2f 9815 super->updates_pending++;
1a2487c2 9816 }
bb025c2f
KW		 9817 break;
9818 }
9819
78b10e66 9820 case update_reshape_container_disks: {
d195167d 9821 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9822 if (apply_reshape_container_disks_update(
9823 u, super, &update->space_list))
9824 super->updates_pending++;
78b10e66
N		 9825 break;
9826 }
48c5303a 9827 case update_reshape_migration: {
a29911da
PC		 9828 struct imsm_update_reshape_migration *u = (void *)update->buf;
9829 if (apply_reshape_migration_update(
9830 u, super, &update->space_list))
9831 super->updates_pending++;
48c5303a
PC		 9832 break;
9833 }
f3871fdc
AK		 9834 case update_size_change: {
9835 struct imsm_update_size_change *u = (void *)update->buf;
9836 if (apply_size_change_update(u, super))
9837 super->updates_pending++;
9838 break;
9839 }
e8319a19 9840 case update_activate_spare: {
1011e834 9841 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9842 if (apply_update_activate_spare(u, super, st->arrays))
9843 super->updates_pending++;
8273f55e
DW		 9844 break;
9845 }
9846 case update_create_array: {
9847 /* someone wants to create a new array, we need to be aware of
9848 * a few races/collisions:
9849 * 1/ 'Create' called by two separate instances of mdadm
9850 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9851 * devices that have since been assimilated via
9852 * activate_spare.
9853 * In the event this update can not be carried out mdadm will
9854 * (FIX ME) notice that its update did not take hold.
9855 */
9856 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9857 struct intel_dev *dv;
8273f55e
DW		 9858 struct imsm_dev *dev;
9859 struct imsm_map *map, *new_map;
9860 unsigned long long start, end;
9861 unsigned long long new_start, new_end;
9862 int i;
54c2c1ea
DW		 9863 struct disk_info *inf;
9864 struct dl *dl;
8273f55e
DW		 9865
9866 /* handle racing creates: first come first serve */
9867 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9868 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9869 goto create_error;
8273f55e
DW		 9870 }
9871
9872 /* check update is next in sequence */
9873 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9874 dprintf("can not create array %d expected index %d\n",
9875 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9876 goto create_error;
8273f55e
DW		 9877 }
9878
238c0a71 9879 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9880 new_start = pba_of_lba0(new_map);
44490938 9881 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9882 inf = get_disk_info(u);
8273f55e
DW		 9883
9884 /* handle activate_spare versus create race:
9885 * check to make sure that overlapping arrays do not include
9886 * overalpping disks
9887 */
9888 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9889 dev = get_imsm_dev(super, i);
238c0a71 9890 map = get_imsm_map(dev, MAP_0);
5551b113 9891 start = pba_of_lba0(map);
44490938 9892 end = start + per_dev_array_size(map);
8273f55e
DW		 9893 if ((new_start >= start && new_start <= end) ||
9894 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9895 /* overlap */;
9896 else
9897 continue;
9898
9899 if (disks_overlap(super, i, u)) {
1ade5cc1 9900 dprintf("arrays overlap\n");
ba2de7ba 9901 goto create_error;
8273f55e
DW		 9902 }
9903 }
8273f55e 9904
949c47a0
DW		 9905 /* check that prepare update was successful */
9906 if (!update->space) {
1ade5cc1 9907 dprintf("prepare update failed\n");
ba2de7ba 9908 goto create_error;
949c47a0
DW		 9909 }
9910
54c2c1ea
DW		 9911 /* check that all disks are still active before committing
9912 * changes. FIXME: could we instead handle this by creating a
9913 * degraded array? That's probably not what the user expects,
9914 * so better to drop this update on the floor.
9915 */
9916 for (i = 0; i < new_map->num_members; i++) {
9917 dl = serial_to_dl(inf[i].serial, super);
9918 if (!dl) {
1ade5cc1 9919 dprintf("disk disappeared\n");
ba2de7ba 9920 goto create_error;
54c2c1ea 9921 }
949c47a0
DW		 9922 }
9923
8273f55e 9924 super->updates_pending++;
54c2c1ea
DW		 9925
9926 /* convert spares to members and fixup ord_tbl */
9927 for (i = 0; i < new_map->num_members; i++) {
9928 dl = serial_to_dl(inf[i].serial, super);
9929 if (dl->index == -1) {
9930 dl->index = mpb->num_disks;
9931 mpb->num_disks++;
9932 dl->disk.status |= CONFIGURED_DISK;
9933 dl->disk.status &= ~SPARE_DISK;
9934 }
9935 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9936 }
9937
ba2de7ba
DW		 9938 dv = update->space;
9939 dev = dv->dev;
949c47a0
DW		 9940 update->space = NULL;
9941 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9942 dv->index = u->dev_idx;
9943 dv->next = super->devlist;
9944 super->devlist = dv;
8273f55e 9945 mpb->num_raid_devs++;
8273f55e 9946
4d1313e9 9947 imsm_update_version_info(super);
8273f55e 9948 break;
ba2de7ba
DW		 9949 create_error:
9950 /* mdmon knows how to release update->space, but not
9951 * ((struct intel_dev *) update->space)->dev
9952 */
9953 if (update->space) {
9954 dv = update->space;
9955 free(dv->dev);
9956 }
8273f55e 9957 break;
e8319a19 9958 }
33414a01
DW		 9959 case update_kill_array: {
9960 struct imsm_update_kill_array *u = (void *) update->buf;
9961 int victim = u->dev_idx;
9962 struct active_array *a;
9963 struct intel_dev **dp;
9964 struct imsm_dev *dev;
9965
9966 /* sanity check that we are not affecting the uuid of
9967 * active arrays, or deleting an active array
9968 *
9969 * FIXME when immutable ids are available, but note that
9970 * we'll also need to fixup the invalidated/active
9971 * subarray indexes in mdstat
9972 */
9973 for (a = st->arrays; a; a = a->next)
9974 if (a->info.container_member >= victim)
9975 break;
9976 /* by definition if mdmon is running at least one array
9977 * is active in the container, so checking
9978 * mpb->num_raid_devs is just extra paranoia
9979 */
9980 dev = get_imsm_dev(super, victim);
9981 if (a || !dev || mpb->num_raid_devs == 1) {
9982 dprintf("failed to delete subarray-%d\n", victim);
9983 break;
9984 }
9985
9986 for (dp = &super->devlist; *dp;)
f21e18ca 9987 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9988 *dp = (*dp)->next;
9989 } else {
f21e18ca 9990 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9991 (*dp)->index--;
9992 dp = &(*dp)->next;
9993 }
9994 mpb->num_raid_devs--;
9995 super->updates_pending++;
9996 break;
9997 }
aa534678
DW		 9998 case update_rename_array: {
9999 struct imsm_update_rename_array *u = (void *) update->buf;
10000 char name[MAX_RAID_SERIAL_LEN+1];
10001 int target = u->dev_idx;
10002 struct active_array *a;
10003 struct imsm_dev *dev;
10004
10005 /* sanity check that we are not affecting the uuid of
10006 * an active array
10007 */
40659392 10008 memset(name, 0, sizeof(name));
aa534678
DW		 10009 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
10010 name[MAX_RAID_SERIAL_LEN] = '\0';
10011 for (a = st->arrays; a; a = a->next)
10012 if (a->info.container_member == target)
10013 break;
10014 dev = get_imsm_dev(super, u->dev_idx);
10015 if (a || !dev || !check_name(super, name, 1)) {
10016 dprintf("failed to rename subarray-%d\n", target);
10017 break;
10018 }
10019
40659392 10020 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10021 super->updates_pending++;
10022 break;
10023 }
1a64be56 10024 case update_add_remove_disk: {
43dad3d6 10025 /* we may be able to repair some arrays if disks are
095b8088 10026 * being added, check the status of add_remove_disk
1a64be56
LM		 10027 * if discs has been added.
10028 */
10029 if (add_remove_disk_update(super)) {
43dad3d6 10030 struct active_array *a;
072b727f
DW		 10031
10032 super->updates_pending++;
1a64be56 10033 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10034 a->check_degraded = 1;
10035 }
43dad3d6 10036 break;
e8319a19 10037 }
bbab0940
TM		 10038 case update_prealloc_badblocks_mem:
10039 break;
e6e9dd3f
AP		 10040 case update_rwh_policy: {
10041 struct imsm_update_rwh_policy *u = (void *)update->buf;
10042 int target = u->dev_idx;
10043 struct imsm_dev *dev = get_imsm_dev(super, target);
10044 if (!dev) {
10045 dprintf("could not find subarray-%d\n", target);
10046 break;
10047 }
10048
10049 if (dev->rwh_policy != u->new_policy) {
10050 dev->rwh_policy = u->new_policy;
10051 super->updates_pending++;
10052 }
10053 break;
10054 }
1a64be56 10055 default:
ebf3be99 10056 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10057 }
e8319a19 10058}
88758e9d 10059
bc0b9d34
PC		 10060static struct mdinfo *get_spares_for_grow(struct supertype *st);
10061
5fe6f031
N		 10062static int imsm_prepare_update(struct supertype *st,
10063 struct metadata_update *update)
8273f55e 10064{
949c47a0 10065 /**
4d7b1503
DW		 10066 * Allocate space to hold new disk entries, raid-device entries or a new
10067 * mpb if necessary. The manager synchronously waits for updates to
10068 * complete in the monitor, so new mpb buffers allocated here can be
10069 * integrated by the monitor thread without worrying about live pointers
10070 * in the manager thread.
8273f55e 10071 */
095b8088 10072 enum imsm_update_type type;
4d7b1503 10073 struct intel_super *super = st->sb;
f36a9ecd 10074 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10075 struct imsm_super *mpb = super->anchor;
10076 size_t buf_len;
10077 size_t len = 0;
949c47a0 10078
095b8088
N		 10079 if (update->len < (int)sizeof(type))
10080 return 0;
10081
10082 type = *(enum imsm_update_type *) update->buf;
10083
949c47a0 10084 switch (type) {
0ec5d470 10085 case update_general_migration_checkpoint:
095b8088
N		 10086 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10087 return 0;
1ade5cc1 10088 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10089 break;
abedf5fc
KW		 10090 case update_takeover: {
10091 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10092 if (update->len < (int)sizeof(*u))
10093 return 0;
abedf5fc
KW		 10094 if (u->direction == R0_TO_R10) {
10095 void **tail = (void **)&update->space_list;
10096 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10097 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10098 int num_members = map->num_members;
10099 void *space;
10100 int size, i;
abedf5fc
KW		 10101 /* allocate memory for added disks */
10102 for (i = 0; i < num_members; i++) {
10103 size = sizeof(struct dl);
503975b9 10104 space = xmalloc(size);
abedf5fc
KW		 10105 *tail = space;
10106 tail = space;
10107 *tail = NULL;
10108 }
10109 /* allocate memory for new device */
10110 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10111 (num_members * sizeof(__u32));
503975b9
N		 10112 space = xmalloc(size);
10113 *tail = space;
10114 tail = space;
10115 *tail = NULL;
10116 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10117 }
10118
10119 break;
10120 }
78b10e66 10121 case update_reshape_container_disks: {
d195167d
AK		 10122 /* Every raid device in the container is about to
10123 * gain some more devices, and we will enter a
10124 * reconfiguration.
10125 * So each 'imsm_map' will be bigger, and the imsm_vol
10126 * will now hold 2 of them.
10127 * Thus we need new 'struct imsm_dev' allocations sized
10128 * as sizeof_imsm_dev but with more devices in both maps.
10129 */
10130 struct imsm_update_reshape *u = (void *)update->buf;
10131 struct intel_dev *dl;
10132 void **space_tail = (void**)&update->space_list;
10133
095b8088
N		 10134 if (update->len < (int)sizeof(*u))
10135 return 0;
10136
1ade5cc1 10137 dprintf("for update_reshape\n");
d195167d
AK		 10138
10139 for (dl = super->devlist; dl; dl = dl->next) {
10140 int size = sizeof_imsm_dev(dl->dev, 1);
10141 void *s;
d677e0b8
AK		 10142 if (u->new_raid_disks > u->old_raid_disks)
10143 size += sizeof(__u32)*2*
10144 (u->new_raid_disks - u->old_raid_disks);
503975b9 10145 s = xmalloc(size);
d195167d
AK		 10146 *space_tail = s;
10147 space_tail = s;
10148 *space_tail = NULL;
10149 }
10150
10151 len = disks_to_mpb_size(u->new_raid_disks);
10152 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10153 break;
10154 }
48c5303a 10155 case update_reshape_migration: {
bc0b9d34
PC		 10156 /* for migration level 0->5 we need to add disks
10157 * so the same as for container operation we will copy
10158 * device to the bigger location.
10159 * in memory prepared device and new disk area are prepared
10160 * for usage in process update
10161 */
10162 struct imsm_update_reshape_migration *u = (void *)update->buf;
10163 struct intel_dev *id;
10164 void **space_tail = (void **)&update->space_list;
10165 int size;
10166 void *s;
10167 int current_level = -1;
10168
095b8088
N		 10169 if (update->len < (int)sizeof(*u))
10170 return 0;
10171
1ade5cc1 10172 dprintf("for update_reshape\n");
bc0b9d34
PC		 10173
10174 /* add space for bigger array in update
10175 */
10176 for (id = super->devlist; id; id = id->next) {
10177 if (id->index == (unsigned)u->subdev) {
10178 size = sizeof_imsm_dev(id->dev, 1);
10179 if (u->new_raid_disks > u->old_raid_disks)
10180 size += sizeof(__u32)*2*
10181 (u->new_raid_disks - u->old_raid_disks);
503975b9 10182 s = xmalloc(size);
bc0b9d34
PC		 10183 *space_tail = s;
10184 space_tail = s;
10185 *space_tail = NULL;
10186 break;
10187 }
10188 }
10189 if (update->space_list == NULL)
10190 break;
10191
10192 /* add space for disk in update
10193 */
10194 size = sizeof(struct dl);
503975b9 10195 s = xmalloc(size);
bc0b9d34
PC		 10196 *space_tail = s;
10197 space_tail = s;
10198 *space_tail = NULL;
10199
10200 /* add spare device to update
10201 */
10202 for (id = super->devlist ; id; id = id->next)
10203 if (id->index == (unsigned)u->subdev) {
10204 struct imsm_dev *dev;
10205 struct imsm_map *map;
10206
10207 dev = get_imsm_dev(super, u->subdev);
238c0a71 10208 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10209 current_level = map->raid_level;
10210 break;
10211 }
089f9d79 10212 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10213 struct mdinfo *spares;
10214
10215 spares = get_spares_for_grow(st);
10216 if (spares) {
10217 struct dl *dl;
10218 struct mdinfo *dev;
10219
10220 dev = spares->devs;
10221 if (dev) {
10222 u->new_disks[0] =
10223 makedev(dev->disk.major,
10224 dev->disk.minor);
10225 dl = get_disk_super(super,
10226 dev->disk.major,
10227 dev->disk.minor);
10228 dl->index = u->old_raid_disks;
10229 dev = dev->next;
10230 }
10231 sysfs_free(spares);
10232 }
10233 }
10234 len = disks_to_mpb_size(u->new_raid_disks);
10235 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10236 break;
10237 }
f3871fdc 10238 case update_size_change: {
095b8088
N		 10239 if (update->len < (int)sizeof(struct imsm_update_size_change))
10240 return 0;
10241 break;
10242 }
10243 case update_activate_spare: {
10244 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10245 return 0;
f3871fdc
AK		 10246 break;
10247 }
949c47a0
DW		 10248 case update_create_array: {
10249 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10250 struct intel_dev *dv;
54c2c1ea 10251 struct imsm_dev *dev = &u->dev;
238c0a71 10252 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10253 struct dl *dl;
10254 struct disk_info *inf;
10255 int i;
10256 int activate = 0;
949c47a0 10257
095b8088
N		 10258 if (update->len < (int)sizeof(*u))
10259 return 0;
10260
54c2c1ea
DW		 10261 inf = get_disk_info(u);
10262 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10263 /* allocate a new super->devlist entry */
503975b9
N		 10264 dv = xmalloc(sizeof(*dv));
10265 dv->dev = xmalloc(len);
10266 update->space = dv;
949c47a0 10267
54c2c1ea
DW		 10268 /* count how many spares will be converted to members */
10269 for (i = 0; i < map->num_members; i++) {
10270 dl = serial_to_dl(inf[i].serial, super);
10271 if (!dl) {
10272 /* hmm maybe it failed?, nothing we can do about
10273 * it here
10274 */
10275 continue;
10276 }
10277 if (count_memberships(dl, super) == 0)
10278 activate++;
10279 }
10280 len += activate * sizeof(struct imsm_disk);
949c47a0 10281 break;
095b8088
N		 10282 }
10283 case update_kill_array: {
10284 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10285 return 0;
949c47a0
DW		 10286 break;
10287 }
095b8088
N		 10288 case update_rename_array: {
10289 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10290 return 0;
10291 break;
10292 }
10293 case update_add_remove_disk:
10294 /* no update->len needed */
10295 break;
bbab0940
TM		 10296 case update_prealloc_badblocks_mem:
10297 super->extra_space += sizeof(struct bbm_log) -
10298 get_imsm_bbm_log_size(super->bbm_log);
10299 break;
e6e9dd3f
AP		 10300 case update_rwh_policy: {
10301 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10302 return 0;
10303 break;
10304 }
095b8088
N		 10305 default:
10306 return 0;
949c47a0 10307 }
8273f55e 10308
4d7b1503
DW		 10309 /* check if we need a larger metadata buffer */
10310 if (super->next_buf)
10311 buf_len = super->next_len;
10312 else
10313 buf_len = super->len;
10314
bbab0940 10315 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10316 /* ok we need a larger buf than what is currently allocated
10317 * if this allocation fails process_update will notice that
10318 * ->next_len is set and ->next_buf is NULL
10319 */
bbab0940
TM		 10320 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10321 super->extra_space + len, sector_size);
4d7b1503
DW		 10322 if (super->next_buf)
10323 free(super->next_buf);
10324
10325 super->next_len = buf_len;
f36a9ecd 10326 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10327 memset(super->next_buf, 0, buf_len);
10328 else
4d7b1503
DW		 10329 super->next_buf = NULL;
10330 }
5fe6f031 10331 return 1;
8273f55e
DW		 10332}
10333
ae6aad82 10334/* must be called while manager is quiesced */
f21e18ca 10335static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10336{
10337 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10338 struct dl *iter;
10339 struct imsm_dev *dev;
10340 struct imsm_map *map;
4c9e8c1e 10341 unsigned int i, j, num_members;
fb12a745 10342 __u32 ord, ord_map0;
4c9e8c1e 10343 struct bbm_log *log = super->bbm_log;
ae6aad82 10344
1ade5cc1 10345 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10346
10347 /* shift all indexes down one */
10348 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10349 if (iter->index > (int)index)
ae6aad82 10350 iter->index--;
47ee5a45 10351 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10352 if (iter->index > (int)index)
47ee5a45 10353 iter->index--;
ae6aad82
DW		 10354
10355 for (i = 0; i < mpb->num_raid_devs; i++) {
10356 dev = get_imsm_dev(super, i);
238c0a71 10357 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10358 num_members = map->num_members;
10359 for (j = 0; j < num_members; j++) {
10360 /* update ord entries being careful not to propagate
10361 * ord-flags to the first map
10362 */
238c0a71 10363 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10364 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10365
24565c9a
DW		 10366 if (ord_to_idx(ord) <= index)
10367 continue;
ae6aad82 10368
238c0a71 10369 map = get_imsm_map(dev, MAP_0);
fb12a745 10370 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10371 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10372 if (map)
10373 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10374 }
10375 }
10376
4c9e8c1e
TM		 10377 for (i = 0; i < log->entry_count; i++) {
10378 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10379
10380 if (entry->disk_ordinal <= index)
10381 continue;
10382 entry->disk_ordinal--;
10383 }
10384
ae6aad82
DW		 10385 mpb->num_disks--;
10386 super->updates_pending++;
24565c9a
DW		 10387 if (*dlp) {
10388 struct dl *dl = *dlp;
10389
10390 *dlp = (*dlp)->next;
10391 __free_imsm_disk(dl);
10392 }
ae6aad82 10393}
9a717282
AK		 10394
10395static void close_targets(int *targets, int new_disks)
10396{
10397 int i;
10398
10399 if (!targets)
10400 return;
10401
10402 for (i = 0; i < new_disks; i++) {
10403 if (targets[i] >= 0) {
10404 close(targets[i]);
10405 targets[i] = -1;
10406 }
10407 }
10408}
10409
10410static int imsm_get_allowed_degradation(int level, int raid_disks,
10411 struct intel_super *super,
10412 struct imsm_dev *dev)
10413{
10414 switch (level) {
bf5cf7c7 10415 case 1:
9a717282
AK		 10416 case 10:{
10417 int ret_val = 0;
10418 struct imsm_map *map;
10419 int i;
10420
10421 ret_val = raid_disks/2;
10422 /* check map if all disks pairs not failed
10423 * in both maps
10424 */
238c0a71 10425 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10426 for (i = 0; i < ret_val; i++) {
10427 int degradation = 0;
10428 if (get_imsm_disk(super, i) == NULL)
10429 degradation++;
10430 if (get_imsm_disk(super, i + 1) == NULL)
10431 degradation++;
10432 if (degradation == 2)
10433 return 0;
10434 }
238c0a71 10435 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10436 /* if there is no second map
10437 * result can be returned
10438 */
10439 if (map == NULL)
10440 return ret_val;
10441 /* check degradation in second map
10442 */
10443 for (i = 0; i < ret_val; i++) {
10444 int degradation = 0;
10445 if (get_imsm_disk(super, i) == NULL)
10446 degradation++;
10447 if (get_imsm_disk(super, i + 1) == NULL)
10448 degradation++;
10449 if (degradation == 2)
10450 return 0;
10451 }
10452 return ret_val;
10453 }
10454 case 5:
10455 return 1;
10456 case 6:
10457 return 2;
10458 default:
10459 return 0;
10460 }
10461}
10462
687629c2
AK		 10463/*******************************************************************************
10464 * Function: open_backup_targets
10465 * Description: Function opens file descriptors for all devices given in
10466 * info->devs
10467 * Parameters:
10468 * info : general array info
10469 * raid_disks : number of disks
10470 * raid_fds : table of device's file descriptors
9a717282
AK		 10471 * super : intel super for raid10 degradation check
10472 * dev : intel device for raid10 degradation check
687629c2
AK		 10473 * Returns:
10474 * 0 : success
10475 * -1 : fail
10476 ******************************************************************************/
9a717282
AK		 10477int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10478 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10479{
10480 struct mdinfo *sd;
f627f5ad 10481 int i;
9a717282 10482 int opened = 0;
f627f5ad
AK		 10483
10484 for (i = 0; i < raid_disks; i++)
10485 raid_fds[i] = -1;
687629c2
AK		 10486
10487 for (sd = info->devs ; sd ; sd = sd->next) {
10488 char *dn;
10489
10490 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10491 dprintf("disk is faulty!!\n");
10492 continue;
10493 }
10494
089f9d79 10495 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10496 continue;
10497
10498 dn = map_dev(sd->disk.major,
10499 sd->disk.minor, 1);
10500 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10501 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10502 pr_err("cannot open component\n");
9a717282 10503 continue;
687629c2 10504 }
9a717282
AK		 10505 opened++;
10506 }
10507 /* check if maximum array degradation level is not exceeded
10508 */
10509 if ((raid_disks - opened) >
089f9d79
JS		 10510 imsm_get_allowed_degradation(info->new_level, raid_disks,
10511 super, dev)) {
e12b3daa 10512 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10513 close_targets(raid_fds, raid_disks);
10514 return -2;
687629c2
AK		 10515 }
10516 return 0;
10517}
10518
d31ad643
PB		 10519/*******************************************************************************
10520 * Function: validate_container_imsm
10521 * Description: This routine validates container after assemble,
10522 * eg. if devices in container are under the same controller.
10523 *
10524 * Parameters:
10525 * info : linked list with info about devices used in array
10526 * Returns:
10527 * 1 : HBA mismatch
10528 * 0 : Success
10529 ******************************************************************************/
10530int validate_container_imsm(struct mdinfo *info)
10531{
6b781d33
AP		 10532 if (check_env("IMSM_NO_PLATFORM"))
10533 return 0;
d31ad643 10534
6b781d33
AP		 10535 struct sys_dev *idev;
10536 struct sys_dev *hba = NULL;
10537 struct sys_dev *intel_devices = find_intel_devices();
10538 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10539 info->disk.minor));
10540
10541 for (idev = intel_devices; idev; idev = idev->next) {
10542 if (dev_path && strstr(dev_path, idev->path)) {
10543 hba = idev;
10544 break;
d31ad643 10545 }
6b781d33
AP		 10546 }
10547 if (dev_path)
d31ad643
PB		 10548 free(dev_path);
10549
6b781d33
AP		 10550 if (!hba) {
10551 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10552 devid2kname(makedev(info->disk.major, info->disk.minor)));
10553 return 1;
10554 }
10555
10556 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10557 struct mdinfo *dev;
10558
10559 for (dev = info->next; dev; dev = dev->next) {
10560 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10561
10562 struct sys_dev *hba2 = NULL;
10563 for (idev = intel_devices; idev; idev = idev->next) {
10564 if (dev_path && strstr(dev_path, idev->path)) {
10565 hba2 = idev;
10566 break;
d31ad643
PB		 10567 }
10568 }
6b781d33
AP		 10569 if (dev_path)
10570 free(dev_path);
10571
10572 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10573 get_orom_by_device_id(hba2->dev_id);
10574
10575 if (hba2 && hba->type != hba2->type) {
10576 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10577 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10578 return 1;
10579 }
10580
07cb1e57 10581 if (orom != orom2) {
6b781d33
AP		 10582 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10583 " This operation is not supported and can lead to data loss.\n");
10584 return 1;
10585 }
10586
10587 if (!orom) {
10588 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10589 " This operation is not supported and can lead to data loss.\n");
10590 return 1;
10591 }
d31ad643 10592 }
6b781d33 10593
d31ad643
PB		 10594 return 0;
10595}
32141c17 10596
6f50473f
TM		 10597/*******************************************************************************
10598* Function: imsm_record_badblock
10599* Description: This routine stores new bad block record in BBM log
10600*
10601* Parameters:
10602* a : array containing a bad block
10603* slot : disk number containing a bad block
10604* sector : bad block sector
10605* length : bad block sectors range
10606* Returns:
10607* 1 : Success
10608* 0 : Error
10609******************************************************************************/
10610static int imsm_record_badblock(struct active_array *a, int slot,
10611 unsigned long long sector, int length)
10612{
10613 struct intel_super *super = a->container->sb;
10614 int ord;
10615 int ret;
10616
10617 ord = imsm_disk_slot_to_ord(a, slot);
10618 if (ord < 0)
10619 return 0;
10620
10621 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10622 length);
10623 if (ret)
10624 super->updates_pending++;
10625
10626 return ret;
10627}
c07a5a4f
TM		 10628/*******************************************************************************
10629* Function: imsm_clear_badblock
10630* Description: This routine clears bad block record from BBM log
10631*
10632* Parameters:
10633* a : array containing a bad block
10634* slot : disk number containing a bad block
10635* sector : bad block sector
10636* length : bad block sectors range
10637* Returns:
10638* 1 : Success
10639* 0 : Error
10640******************************************************************************/
10641static int imsm_clear_badblock(struct active_array *a, int slot,
10642 unsigned long long sector, int length)
10643{
10644 struct intel_super *super = a->container->sb;
10645 int ord;
10646 int ret;
10647
10648 ord = imsm_disk_slot_to_ord(a, slot);
10649 if (ord < 0)
10650 return 0;
10651
10652 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10653 if (ret)
10654 super->updates_pending++;
10655
10656 return ret;
10657}
928f1424
TM		 10658/*******************************************************************************
10659* Function: imsm_get_badblocks
10660* Description: This routine get list of bad blocks for an array
10661*
10662* Parameters:
10663* a : array
10664* slot : disk number
10665* Returns:
10666* bb : structure containing bad blocks
10667* NULL : error
10668******************************************************************************/
10669static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10670{
10671 int inst = a->info.container_member;
10672 struct intel_super *super = a->container->sb;
10673 struct imsm_dev *dev = get_imsm_dev(super, inst);
10674 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10675 int ord;
10676
10677 ord = imsm_disk_slot_to_ord(a, slot);
10678 if (ord < 0)
10679 return NULL;
10680
10681 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10682 per_dev_array_size(map), &super->bb);
928f1424
TM		 10683
10684 return &super->bb;
10685}
27156a57
TM		 10686/*******************************************************************************
10687* Function: examine_badblocks_imsm
10688* Description: Prints list of bad blocks on a disk to the standard output
10689*
10690* Parameters:
10691* st : metadata handler
10692* fd : open file descriptor for device
10693* devname : device name
10694* Returns:
10695* 0 : Success
10696* 1 : Error
10697******************************************************************************/
10698static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10699{
10700 struct intel_super *super = st->sb;
10701 struct bbm_log *log = super->bbm_log;
10702 struct dl *d = NULL;
10703 int any = 0;
10704
10705 for (d = super->disks; d ; d = d->next) {
10706 if (strcmp(d->devname, devname) == 0)
10707 break;
10708 }
10709
10710 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10711 pr_err("%s doesn't appear to be part of a raid array\n",
10712 devname);
10713 return 1;
10714 }
10715
10716 if (log != NULL) {
10717 unsigned int i;
10718 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10719
10720 for (i = 0; i < log->entry_count; i++) {
10721 if (entry[i].disk_ordinal == d->index) {
10722 unsigned long long sector = __le48_to_cpu(
10723 &entry[i].defective_block_start);
10724 int cnt = entry[i].marked_count + 1;
10725
10726 if (!any) {
10727 printf("Bad-blocks on %s:\n", devname);
10728 any = 1;
10729 }
10730
10731 printf("%20llu for %d sectors\n", sector, cnt);
10732 }
10733 }
10734 }
10735
10736 if (!any)
10737 printf("No bad-blocks list configured on %s\n", devname);
10738
10739 return 0;
10740}
687629c2
AK		 10741/*******************************************************************************
10742 * Function: init_migr_record_imsm
10743 * Description: Function inits imsm migration record
10744 * Parameters:
10745 * super : imsm internal array info
10746 * dev : device under migration
10747 * info : general array info to find the smallest device
10748 * Returns:
10749 * none
10750 ******************************************************************************/
10751void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10752 struct mdinfo *info)
10753{
10754 struct intel_super *super = st->sb;
10755 struct migr_record *migr_rec = super->migr_rec;
10756 int new_data_disks;
10757 unsigned long long dsize, dev_sectors;
10758 long long unsigned min_dev_sectors = -1LLU;
10759 struct mdinfo *sd;
10760 char nm[30];
10761 int fd;
238c0a71
AK		 10762 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10763 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10764 unsigned long long num_migr_units;
3ef4403c 10765 unsigned long long array_blocks;
687629c2
AK		 10766
10767 memset(migr_rec, 0, sizeof(struct migr_record));
10768 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10769
10770 /* only ascending reshape supported now */
10771 migr_rec->ascending_migr = __cpu_to_le32(1);
10772
10773 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10774 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10775 migr_rec->dest_depth_per_unit *=
10776 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10777 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10778 migr_rec->blocks_per_unit =
10779 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10780 migr_rec->dest_depth_per_unit =
10781 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10782 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10783 num_migr_units =
10784 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10785
10786 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10787 num_migr_units++;
9f421827 10788 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10789
10790 migr_rec->post_migr_vol_cap = dev->size_low;
10791 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10792
687629c2
AK		 10793 /* Find the smallest dev */
10794 for (sd = info->devs ; sd ; sd = sd->next) {
10795 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10796 fd = dev_open(nm, O_RDONLY);
10797 if (fd < 0)
10798 continue;
10799 get_dev_size(fd, NULL, &dsize);
10800 dev_sectors = dsize / 512;
10801 if (dev_sectors < min_dev_sectors)
10802 min_dev_sectors = dev_sectors;
10803 close(fd);
10804 }
9f421827 10805 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10806 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10807
10808 write_imsm_migr_rec(st);
10809
10810 return;
10811}
10812
10813/*******************************************************************************
10814 * Function: save_backup_imsm
10815 * Description: Function saves critical data stripes to Migration Copy Area
10816 * and updates the current migration unit status.
10817 * Use restore_stripes() to form a destination stripe,
10818 * and to write it to the Copy Area.
10819 * Parameters:
10820 * st : supertype information
aea93171 10821 * dev : imsm device that backup is saved for
687629c2
AK		 10822 * info : general array info
10823 * buf : input buffer
687629c2
AK		 10824 * length : length of data to backup (blocks_per_unit)
10825 * Returns:
10826 * 0 : success
10827 *, -1 : fail
10828 ******************************************************************************/
10829int save_backup_imsm(struct supertype *st,
10830 struct imsm_dev *dev,
10831 struct mdinfo *info,
10832 void *buf,
687629c2
AK		 10833 int length)
10834{
10835 int rv = -1;
10836 struct intel_super *super = st->sb;
594dc1b8
JS		 10837 unsigned long long *target_offsets;
10838 int *targets;
687629c2 10839 int i;
238c0a71 10840 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10841 int new_disks = map_dest->num_members;
ab724b98
AK		 10842 int dest_layout = 0;
10843 int dest_chunk;
d1877f69 10844 unsigned long long start;
9529d343 10845 int data_disks = imsm_num_data_members(map_dest);
687629c2 10846
503975b9 10847 targets = xmalloc(new_disks * sizeof(int));
687629c2 10848
7e45b550
AK		 10849 for (i = 0; i < new_disks; i++)
10850 targets[i] = -1;
10851
503975b9 10852 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10853
d1877f69 10854 start = info->reshape_progress * 512;
687629c2 10855 for (i = 0; i < new_disks; i++) {
9f421827 10856 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10857 /* move back copy area adderss, it will be moved forward
10858 * in restore_stripes() using start input variable
10859 */
10860 target_offsets[i] -= start/data_disks;
687629c2
AK		 10861 }
10862
9a717282
AK		 10863 if (open_backup_targets(info, new_disks, targets,
10864 super, dev))
687629c2
AK		 10865 goto abort;
10866
68eb8bc6 10867 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10868 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10869
687629c2
AK		 10870 if (restore_stripes(targets, /* list of dest devices */
10871 target_offsets, /* migration record offsets */
10872 new_disks,
ab724b98
AK		 10873 dest_chunk,
10874 map_dest->raid_level,
10875 dest_layout,
10876 -1, /* source backup file descriptor */
10877 0, /* input buf offset
10878 * always 0 buf is already offseted */
d1877f69 10879 start,
687629c2
AK		 10880 length,
10881 buf) != 0) {
e7b84f9d 10882 pr_err("Error restoring stripes\n");
687629c2
AK		 10883 goto abort;
10884 }
10885
10886 rv = 0;
10887
10888abort:
10889 if (targets) {
9a717282 10890 close_targets(targets, new_disks);
687629c2
AK		 10891 free(targets);
10892 }
10893 free(target_offsets);
10894
10895 return rv;
10896}
10897
10898/*******************************************************************************
10899 * Function: save_checkpoint_imsm
10900 * Description: Function called for current unit status update
10901 * in the migration record. It writes it to disk.
10902 * Parameters:
10903 * super : imsm internal array info
10904 * info : general array info
10905 * Returns:
10906 * 0: success
10907 * 1: failure
0228d92c
AK		 10908 * 2: failure, means no valid migration record
10909 * / no general migration in progress /
687629c2
AK		 10910 ******************************************************************************/
10911int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10912{
10913 struct intel_super *super = st->sb;
f8b72ef5
AK		 10914 unsigned long long blocks_per_unit;
10915 unsigned long long curr_migr_unit;
10916
2e062e82 10917 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10918 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10919 return 1;
10920 }
10921
f8b72ef5
AK		 10922 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10923 if (blocks_per_unit == 0) {
0228d92c
AK		 10924 dprintf("imsm: no migration in progress.\n");
10925 return 2;
687629c2 10926 }
f8b72ef5
AK		 10927 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10928 /* check if array is alligned to copy area
10929 * if it is not alligned, add one to current migration unit value
10930 * this can happend on array reshape finish only
10931 */
10932 if (info->reshape_progress % blocks_per_unit)
10933 curr_migr_unit++;
687629c2 10934
9f421827 10935 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10936 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10937 set_migr_dest_1st_member_lba(super->migr_rec,
10938 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10939
687629c2 10940 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10941 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10942 return 1;
10943 }
10944
10945 return 0;
10946}
10947
276d77db
AK		 10948/*******************************************************************************
10949 * Function: recover_backup_imsm
10950 * Description: Function recovers critical data from the Migration Copy Area
10951 * while assembling an array.
10952 * Parameters:
10953 * super : imsm internal array info
10954 * info : general array info
10955 * Returns:
10956 * 0 : success (or there is no data to recover)
10957 * 1 : fail
10958 ******************************************************************************/
10959int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10960{
10961 struct intel_super *super = st->sb;
10962 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10963 struct imsm_map *map_dest;
276d77db
AK		 10964 struct intel_dev *id = NULL;
10965 unsigned long long read_offset;
10966 unsigned long long write_offset;
10967 unsigned unit_len;
10968 int *targets = NULL;
10969 int new_disks, i, err;
10970 char *buf = NULL;
10971 int retval = 1;
f36a9ecd 10972 unsigned int sector_size = super->sector_size;
9f421827
PB		 10973 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10974 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10975 char buffer[20];
6c3560c0 10976 int skipped_disks = 0;
276d77db
AK		 10977
10978 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10979 if (err < 1)
10980 return 1;
10981
10982 /* recover data only during assemblation */
10983 if (strncmp(buffer, "inactive", 8) != 0)
10984 return 0;
10985 /* no data to recover */
10986 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10987 return 0;
10988 if (curr_migr_unit >= num_migr_units)
10989 return 1;
10990
10991 /* find device during reshape */
10992 for (id = super->devlist; id; id = id->next)
10993 if (is_gen_migration(id->dev))
10994 break;
10995 if (id == NULL)
10996 return 1;
10997
238c0a71 10998 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10999 new_disks = map_dest->num_members;
11000
9f421827 11001 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 11002
9f421827 11003 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 11004 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 11005
11006 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 11007 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 11008 goto abort;
503975b9 11009 targets = xcalloc(new_disks, sizeof(int));
276d77db 11010
9a717282 11011 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 11012 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 11013 goto abort;
11014 }
276d77db
AK		 11015
11016 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 11017 if (targets[i] < 0) {
11018 skipped_disks++;
11019 continue;
11020 }
276d77db 11021 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11022 pr_err("Cannot seek to block: %s\n",
11023 strerror(errno));
137debce
AK		 11024 skipped_disks++;
11025 continue;
276d77db 11026 }
9ec11d1a 11027 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11028 pr_err("Cannot read copy area block: %s\n",
11029 strerror(errno));
137debce
AK		 11030 skipped_disks++;
11031 continue;
276d77db
AK		 11032 }
11033 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11034 pr_err("Cannot seek to block: %s\n",
11035 strerror(errno));
137debce
AK		 11036 skipped_disks++;
11037 continue;
276d77db 11038 }
9ec11d1a 11039 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11040 pr_err("Cannot restore block: %s\n",
11041 strerror(errno));
137debce
AK		 11042 skipped_disks++;
11043 continue;
276d77db
AK		 11044 }
11045 }
11046
137debce
AK		 11047 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11048 new_disks,
11049 super,
11050 id->dev)) {
7a862a02 11051 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11052 goto abort;
11053 }
11054
befb629b
AK		 11055 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11056 /* ignore error == 2, this can mean end of reshape here
11057 */
7a862a02 11058 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11059 } else
276d77db 11060 retval = 0;
276d77db
AK		 11061
11062abort:
11063 if (targets) {
11064 for (i = 0; i < new_disks; i++)
11065 if (targets[i])
11066 close(targets[i]);
11067 free(targets);
11068 }
11069 free(buf);
11070 return retval;
11071}
11072
2cda7640
ML		 11073static char disk_by_path[] = "/dev/disk/by-path/";
11074
11075static const char *imsm_get_disk_controller_domain(const char *path)
11076{
2cda7640 11077 char disk_path[PATH_MAX];
96234762
LM		 11078 char *drv=NULL;
11079 struct stat st;
2cda7640 11080
6d8d290a 11081 strcpy(disk_path, disk_by_path);
96234762
LM		 11082 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11083 if (stat(disk_path, &st) == 0) {
11084 struct sys_dev* hba;
594dc1b8 11085 char *path;
96234762
LM		 11086
11087 path = devt_to_devpath(st.st_rdev);
11088 if (path == NULL)
11089 return "unknown";
11090 hba = find_disk_attached_hba(-1, path);
11091 if (hba && hba->type == SYS_DEV_SAS)
11092 drv = "isci";
11093 else if (hba && hba->type == SYS_DEV_SATA)
11094 drv = "ahci";
c6839718
MT		 11095 else if (hba && hba->type == SYS_DEV_VMD)
11096 drv = "vmd";
11097 else if (hba && hba->type == SYS_DEV_NVME)
11098 drv = "nvme";
1011e834 11099 else
96234762
LM		 11100 drv = "unknown";
11101 dprintf("path: %s hba: %s attached: %s\n",
11102 path, (hba) ? hba->path : "NULL", drv);
11103 free(path);
2cda7640 11104 }
96234762 11105 return drv;
2cda7640
ML		 11106}
11107
4dd2df09 11108static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11109{
4dd2df09 11110 static char devnm[32];
78b10e66
N		 11111 char subdev_name[20];
11112 struct mdstat_ent *mdstat;
11113
11114 sprintf(subdev_name, "%d", subdev);
11115 mdstat = mdstat_by_subdev(subdev_name, container);
11116 if (!mdstat)
4dd2df09 11117 return NULL;
78b10e66 11118
4dd2df09 11119 strcpy(devnm, mdstat->devnm);
78b10e66 11120 free_mdstat(mdstat);
4dd2df09 11121 return devnm;
78b10e66
N		 11122}
11123
11124static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11125 struct geo_params *geo,
fbf3d202
AK		 11126 int *old_raid_disks,
11127 int direction)
78b10e66 11128{
694575e7
KW		 11129 /* currently we only support increasing the number of devices
11130 * for a container. This increases the number of device for each
11131 * member array. They must all be RAID0 or RAID5.
11132 */
78b10e66
N		 11133 int ret_val = 0;
11134 struct mdinfo *info, *member;
11135 int devices_that_can_grow = 0;
11136
7a862a02 11137 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11138
d04f65f4 11139 if (geo->size > 0 ||
78b10e66
N		 11140 geo->level != UnSet ||
11141 geo->layout != UnSet ||
11142 geo->chunksize != 0 ||
11143 geo->raid_disks == UnSet) {
7a862a02 11144 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11145 return ret_val;
11146 }
11147
fbf3d202 11148 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11149 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11150 return ret_val;
11151 }
11152
78b10e66
N		 11153 info = container_content_imsm(st, NULL);
11154 for (member = info; member; member = member->next) {
4dd2df09 11155 char *result;
78b10e66
N		 11156
11157 dprintf("imsm: checking device_num: %i\n",
11158 member->container_member);
11159
d7d205bd 11160 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11161 /* we work on container for Online Capacity Expansion
11162 * only so raid_disks has to grow
11163 */
7a862a02 11164 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11165 break;
11166 }
11167
089f9d79 11168 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11169 /* we cannot use this container with other raid level
11170 */
7a862a02 11171 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11172 info->array.level);
11173 break;
11174 } else {
11175 /* check for platform support
11176 * for this raid level configuration
11177 */
11178 struct intel_super *super = st->sb;
11179 if (!is_raid_level_supported(super->orom,
11180 member->array.level,
11181 geo->raid_disks)) {
7a862a02 11182 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11183 info->array.level,
11184 geo->raid_disks,
11185 geo->raid_disks > 1 ? "s" : "");
11186 break;
11187 }
2a4a08e7
AK		 11188 /* check if component size is aligned to chunk size
11189 */
11190 if (info->component_size %
11191 (info->array.chunk_size/512)) {
7a862a02 11192 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11193 break;
11194 }
78b10e66
N		 11195 }
11196
11197 if (*old_raid_disks &&
11198 info->array.raid_disks != *old_raid_disks)
11199 break;
11200 *old_raid_disks = info->array.raid_disks;
11201
11202 /* All raid5 and raid0 volumes in container
11203 * have to be ready for Online Capacity Expansion
11204 * so they need to be assembled. We have already
11205 * checked that no recovery etc is happening.
11206 */
4dd2df09
N		 11207 result = imsm_find_array_devnm_by_subdev(member->container_member,
11208 st->container_devnm);
11209 if (result == NULL) {
78b10e66
N		 11210 dprintf("imsm: cannot find array\n");
11211 break;
11212 }
11213 devices_that_can_grow++;
11214 }
11215 sysfs_free(info);
11216 if (!member && devices_that_can_grow)
11217 ret_val = 1;
11218
11219 if (ret_val)
1ade5cc1 11220 dprintf("Container operation allowed\n");
78b10e66 11221 else
1ade5cc1 11222 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11223
11224 return ret_val;
11225}
11226
11227/* Function: get_spares_for_grow
11228 * Description: Allocates memory and creates list of spare devices
1011e834 11229 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11230 * Parameters: Pointer to the supertype structure
11231 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11232 * NULL if fail
78b10e66
N		 11233 */
11234static struct mdinfo *get_spares_for_grow(struct supertype *st)
11235{
fbfdcb06
AO		 11236 struct spare_criteria sc;
11237
11238 get_spare_criteria_imsm(st, &sc);
11239 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11240}
11241
11242/******************************************************************************
11243 * function: imsm_create_metadata_update_for_reshape
11244 * Function creates update for whole IMSM container.
11245 *
11246 ******************************************************************************/
11247static int imsm_create_metadata_update_for_reshape(
11248 struct supertype *st,
11249 struct geo_params *geo,
11250 int old_raid_disks,
11251 struct imsm_update_reshape **updatep)
11252{
11253 struct intel_super *super = st->sb;
11254 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11255 int update_memory_size;
11256 struct imsm_update_reshape *u;
11257 struct mdinfo *spares;
78b10e66 11258 int i;
594dc1b8 11259 int delta_disks;
bbd24d86 11260 struct mdinfo *dev;
78b10e66 11261
1ade5cc1 11262 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11263
11264 delta_disks = geo->raid_disks - old_raid_disks;
11265
11266 /* size of all update data without anchor */
11267 update_memory_size = sizeof(struct imsm_update_reshape);
11268
11269 /* now add space for spare disks that we need to add. */
11270 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11271
503975b9 11272 u = xcalloc(1, update_memory_size);
78b10e66
N		 11273 u->type = update_reshape_container_disks;
11274 u->old_raid_disks = old_raid_disks;
11275 u->new_raid_disks = geo->raid_disks;
11276
11277 /* now get spare disks list
11278 */
11279 spares = get_spares_for_grow(st);
11280
d7be7d87 11281 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11282 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11283 i = -1;
78b10e66
N		 11284 goto abort;
11285 }
11286
11287 /* we have got spares
11288 * update disk list in imsm_disk list table in anchor
11289 */
11290 dprintf("imsm: %i spares are available.\n\n",
11291 spares->array.spare_disks);
11292
bbd24d86 11293 dev = spares->devs;
78b10e66 11294 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11295 struct dl *dl;
11296
bbd24d86
AK		 11297 if (dev == NULL)
11298 break;
78b10e66
N		 11299 u->new_disks[i] = makedev(dev->disk.major,
11300 dev->disk.minor);
11301 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11302 dl->index = mpb->num_disks;
11303 mpb->num_disks++;
bbd24d86 11304 dev = dev->next;
78b10e66 11305 }
78b10e66
N		 11306
11307abort:
11308 /* free spares
11309 */
11310 sysfs_free(spares);
11311
d677e0b8 11312 dprintf("imsm: reshape update preparation :");
78b10e66 11313 if (i == delta_disks) {
1ade5cc1 11314 dprintf_cont(" OK\n");
78b10e66
N		 11315 *updatep = u;
11316 return update_memory_size;
11317 }
11318 free(u);
1ade5cc1 11319 dprintf_cont(" Error\n");
78b10e66
N		 11320
11321 return 0;
11322}
11323
f3871fdc
AK		 11324/******************************************************************************
11325 * function: imsm_create_metadata_update_for_size_change()
11326 * Creates update for IMSM array for array size change.
11327 *
11328 ******************************************************************************/
11329static int imsm_create_metadata_update_for_size_change(
11330 struct supertype *st,
11331 struct geo_params *geo,
11332 struct imsm_update_size_change **updatep)
11333{
11334 struct intel_super *super = st->sb;
594dc1b8
JS		 11335 int update_memory_size;
11336 struct imsm_update_size_change *u;
f3871fdc 11337
1ade5cc1 11338 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11339
11340 /* size of all update data without anchor */
11341 update_memory_size = sizeof(struct imsm_update_size_change);
11342
503975b9 11343 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11344 u->type = update_size_change;
11345 u->subdev = super->current_vol;
11346 u->new_size = geo->size;
11347
11348 dprintf("imsm: reshape update preparation : OK\n");
11349 *updatep = u;
11350
11351 return update_memory_size;
11352}
11353
48c5303a
PC		 11354/******************************************************************************
11355 * function: imsm_create_metadata_update_for_migration()
11356 * Creates update for IMSM array.
11357 *
11358 ******************************************************************************/
11359static int imsm_create_metadata_update_for_migration(
11360 struct supertype *st,
11361 struct geo_params *geo,
11362 struct imsm_update_reshape_migration **updatep)
11363{
11364 struct intel_super *super = st->sb;
594dc1b8
JS		 11365 int update_memory_size;
11366 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11367 struct imsm_dev *dev;
11368 int previous_level = -1;
11369
1ade5cc1 11370 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11371
11372 /* size of all update data without anchor */
11373 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11374
503975b9 11375 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11376 u->type = update_reshape_migration;
11377 u->subdev = super->current_vol;
11378 u->new_level = geo->level;
11379 u->new_layout = geo->layout;
11380 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11381 u->new_disks[0] = -1;
4bba0439 11382 u->new_chunksize = -1;
48c5303a
PC		 11383
11384 dev = get_imsm_dev(super, u->subdev);
11385 if (dev) {
11386 struct imsm_map *map;
11387
238c0a71 11388 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11389 if (map) {
11390 int current_chunk_size =
11391 __le16_to_cpu(map->blocks_per_strip) / 2;
11392
11393 if (geo->chunksize != current_chunk_size) {
11394 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11395 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11396 current_chunk_size, u->new_chunksize);
11397 }
48c5303a 11398 previous_level = map->raid_level;
4bba0439 11399 }
48c5303a 11400 }
089f9d79 11401 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11402 struct mdinfo *spares = NULL;
11403
11404 u->new_raid_disks++;
11405 spares = get_spares_for_grow(st);
089f9d79 11406 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11407 free(u);
11408 sysfs_free(spares);
11409 update_memory_size = 0;
565cc99e 11410 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11411 return 0;
11412 }
11413 sysfs_free(spares);
11414 }
11415 dprintf("imsm: reshape update preparation : OK\n");
11416 *updatep = u;
11417
11418 return update_memory_size;
11419}
11420
8dd70bce
AK		 11421static void imsm_update_metadata_locally(struct supertype *st,
11422 void *buf, int len)
11423{
11424 struct metadata_update mu;
11425
11426 mu.buf = buf;
11427 mu.len = len;
11428 mu.space = NULL;
11429 mu.space_list = NULL;
11430 mu.next = NULL;
5fe6f031
N		 11431 if (imsm_prepare_update(st, &mu))
11432 imsm_process_update(st, &mu);
8dd70bce
AK		 11433
11434 while (mu.space_list) {
11435 void **space = mu.space_list;
11436 mu.space_list = *space;
11437 free(space);
11438 }
11439}
78b10e66 11440
471bceb6 11441/***************************************************************************
694575e7 11442* Function: imsm_analyze_change
471bceb6 11443* Description: Function analyze change for single volume
1011e834 11444* and validate if transition is supported
fbf3d202
AK		 11445* Parameters: Geometry parameters, supertype structure,
11446* metadata change direction (apply/rollback)
694575e7 11447* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11448****************************************************************************/
11449enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11450 struct geo_params *geo,
11451 int direction)
694575e7 11452{
471bceb6
KW		 11453 struct mdinfo info;
11454 int change = -1;
11455 int check_devs = 0;
c21e737b 11456 int chunk;
67a2db32
AK		 11457 /* number of added/removed disks in operation result */
11458 int devNumChange = 0;
11459 /* imsm compatible layout value for array geometry verification */
11460 int imsm_layout = -1;
7abc9871
AK		 11461 int data_disks;
11462 struct imsm_dev *dev;
9529d343 11463 struct imsm_map *map;
7abc9871 11464 struct intel_super *super;
d04f65f4 11465 unsigned long long current_size;
65d38cca 11466 unsigned long long free_size;
d04f65f4 11467 unsigned long long max_size;
65d38cca 11468 int rv;
471bceb6
KW		 11469
11470 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11471 if (geo->level != info.array.level && geo->level >= 0 &&
11472 geo->level != UnSet) {
471bceb6
KW		 11473 switch (info.array.level) {
11474 case 0:
11475 if (geo->level == 5) {
b5347799 11476 change = CH_MIGRATION;
e13ce846 11477 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11478 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11479 change = -1;
11480 goto analyse_change_exit;
11481 }
67a2db32 11482 imsm_layout = geo->layout;
471bceb6 11483 check_devs = 1;
e91a3bad
LM		 11484 devNumChange = 1; /* parity disk added */
11485 } else if (geo->level == 10) {
471bceb6
KW		 11486 change = CH_TAKEOVER;
11487 check_devs = 1;
e91a3bad 11488 devNumChange = 2; /* two mirrors added */
67a2db32 11489 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11490 }
dfe77a9e
KW		 11491 break;
11492 case 1:
471bceb6
KW		 11493 case 10:
11494 if (geo->level == 0) {
11495 change = CH_TAKEOVER;
11496 check_devs = 1;
e91a3bad 11497 devNumChange = -(geo->raid_disks/2);
67a2db32 11498 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11499 }
11500 break;
11501 }
11502 if (change == -1) {
7a862a02 11503 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11504 info.array.level, geo->level);
471bceb6
KW		 11505 goto analyse_change_exit;
11506 }
11507 } else
11508 geo->level = info.array.level;
11509
089f9d79
JS		 11510 if (geo->layout != info.array.layout &&
11511 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11512 change = CH_MIGRATION;
089f9d79
JS		 11513 if (info.array.layout == 0 && info.array.level == 5 &&
11514 geo->layout == 5) {
471bceb6 11515 /* reshape 5 -> 4 */
089f9d79
JS		 11516 } else if (info.array.layout == 5 && info.array.level == 5 &&
11517 geo->layout == 0) {
471bceb6
KW		 11518 /* reshape 4 -> 5 */
11519 geo->layout = 0;
11520 geo->level = 5;
11521 } else {
7a862a02 11522 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11523 info.array.layout, geo->layout);
471bceb6
KW		 11524 change = -1;
11525 goto analyse_change_exit;
11526 }
67a2db32 11527 } else {
471bceb6 11528 geo->layout = info.array.layout;
67a2db32
AK		 11529 if (imsm_layout == -1)
11530 imsm_layout = info.array.layout;
11531 }
471bceb6 11532
089f9d79
JS		 11533 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11534 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11535 if (info.array.level == 10) {
11536 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11537 change = -1;
11538 goto analyse_change_exit;
1e9b2c3f
PB		 11539 } else if (info.component_size % (geo->chunksize/512)) {
11540 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11541 geo->chunksize/1024, info.component_size/2);
11542 change = -1;
11543 goto analyse_change_exit;
2d2b0eb7 11544 }
b5347799 11545 change = CH_MIGRATION;
2d2b0eb7 11546 } else {
471bceb6 11547 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11548 }
471bceb6 11549
c21e737b 11550 chunk = geo->chunksize / 1024;
7abc9871
AK		 11551
11552 super = st->sb;
11553 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11554 map = get_imsm_map(dev, MAP_0);
11555 data_disks = imsm_num_data_members(map);
c41e00b2 11556 /* compute current size per disk member
7abc9871 11557 */
c41e00b2
AK		 11558 current_size = info.custom_array_size / data_disks;
11559
089f9d79 11560 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11561 /* align component size
11562 */
3e684231 11563 geo->size = imsm_component_size_alignment_check(
c41e00b2 11564 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11565 chunk * 1024, super->sector_size,
c41e00b2 11566 geo->size * 2);
65d0b4ce 11567 if (geo->size == 0) {
7a862a02 11568 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11569 current_size);
11570 goto analyse_change_exit;
11571 }
c41e00b2 11572 }
7abc9871 11573
089f9d79 11574 if (current_size != geo->size && geo->size > 0) {
7abc9871 11575 if (change != -1) {
7a862a02 11576 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11577 change = -1;
11578 goto analyse_change_exit;
11579 }
11580 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11581 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11582 super->current_vol, st->devnm);
7abc9871
AK		 11583 goto analyse_change_exit;
11584 }
65d38cca
LD		 11585 /* check the maximum available size
11586 */
11587 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11588 0, chunk, &free_size);
11589 if (rv == 0)
11590 /* Cannot find maximum available space
11591 */
11592 max_size = 0;
11593 else {
11594 max_size = free_size + current_size;
11595 /* align component size
11596 */
3e684231 11597 max_size = imsm_component_size_alignment_check(
65d38cca 11598 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11599 chunk * 1024, super->sector_size,
65d38cca
LD		 11600 max_size);
11601 }
d04f65f4 11602 if (geo->size == MAX_SIZE) {
b130333f
AK		 11603 /* requested size change to the maximum available size
11604 */
65d38cca 11605 if (max_size == 0) {
7a862a02 11606 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11607 change = -1;
11608 goto analyse_change_exit;
65d38cca
LD		 11609 } else
11610 geo->size = max_size;
c41e00b2 11611 }
b130333f 11612
681b7ae2 11613 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11614 /* accept size for rollback only
11615 */
11616 } else {
11617 /* round size due to metadata compatibility
11618 */
11619 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11620 << SECT_PER_MB_SHIFT;
11621 dprintf("Prepare update for size change to %llu\n",
11622 geo->size );
11623 if (current_size >= geo->size) {
7a862a02 11624 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11625 current_size, geo->size);
fbf3d202
AK		 11626 goto analyse_change_exit;
11627 }
65d38cca 11628 if (max_size && geo->size > max_size) {
7a862a02 11629 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11630 max_size, geo->size);
65d38cca
LD		 11631 goto analyse_change_exit;
11632 }
7abc9871
AK		 11633 }
11634 geo->size *= data_disks;
11635 geo->raid_disks = dev->vol.map->num_members;
11636 change = CH_ARRAY_SIZE;
11637 }
471bceb6
KW		 11638 if (!validate_geometry_imsm(st,
11639 geo->level,
67a2db32 11640 imsm_layout,
e91a3bad 11641 geo->raid_disks + devNumChange,
c21e737b 11642 &chunk,
af4348dd 11643 geo->size, INVALID_SECTORS,
5308f117 11644 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11645 change = -1;
11646
11647 if (check_devs) {
11648 struct intel_super *super = st->sb;
11649 struct imsm_super *mpb = super->anchor;
11650
11651 if (mpb->num_raid_devs > 1) {
7a862a02 11652 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11653 geo->dev_name);
471bceb6
KW		 11654 change = -1;
11655 }
11656 }
11657
11658analyse_change_exit:
089f9d79
JS		 11659 if (direction == ROLLBACK_METADATA_CHANGES &&
11660 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11661 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11662 change = -1;
11663 }
471bceb6 11664 return change;
694575e7
KW		 11665}
11666
bb025c2f
KW		 11667int imsm_takeover(struct supertype *st, struct geo_params *geo)
11668{
11669 struct intel_super *super = st->sb;
11670 struct imsm_update_takeover *u;
11671
503975b9 11672 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11673
11674 u->type = update_takeover;
11675 u->subarray = super->current_vol;
11676
11677 /* 10->0 transition */
11678 if (geo->level == 0)
11679 u->direction = R10_TO_R0;
11680
0529c688
KW		 11681 /* 0->10 transition */
11682 if (geo->level == 10)
11683 u->direction = R0_TO_R10;
11684
bb025c2f
KW		 11685 /* update metadata locally */
11686 imsm_update_metadata_locally(st, u,
11687 sizeof(struct imsm_update_takeover));
11688 /* and possibly remotely */
11689 if (st->update_tail)
11690 append_metadata_update(st, u,
11691 sizeof(struct imsm_update_takeover));
11692 else
11693 free(u);
11694
11695 return 0;
11696}
11697
d04f65f4
N		 11698static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11699 int level,
78b10e66 11700 int layout, int chunksize, int raid_disks,
41784c88 11701 int delta_disks, char *backup, char *dev,
016e00f5 11702 int direction, int verbose)
78b10e66 11703{
78b10e66
N		 11704 int ret_val = 1;
11705 struct geo_params geo;
11706
1ade5cc1 11707 dprintf("(enter)\n");
78b10e66 11708
71204a50 11709 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11710
11711 geo.dev_name = dev;
4dd2df09 11712 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11713 geo.size = size;
11714 geo.level = level;
11715 geo.layout = layout;
11716 geo.chunksize = chunksize;
11717 geo.raid_disks = raid_disks;
41784c88
AK		 11718 if (delta_disks != UnSet)
11719 geo.raid_disks += delta_disks;
78b10e66 11720
1ade5cc1
N		 11721 dprintf("for level : %i\n", geo.level);
11722 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11723
4dd2df09 11724 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11725 /* On container level we can only increase number of devices. */
11726 dprintf("imsm: info: Container operation\n");
78b10e66 11727 int old_raid_disks = 0;
6dc0be30 11728
78b10e66 11729 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11730 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11731 struct imsm_update_reshape *u = NULL;
11732 int len;
11733
11734 len = imsm_create_metadata_update_for_reshape(
11735 st, &geo, old_raid_disks, &u);
11736
ed08d51c
AK		 11737 if (len <= 0) {
11738 dprintf("imsm: Cannot prepare update\n");
11739 goto exit_imsm_reshape_super;
11740 }
11741
8dd70bce
AK		 11742 ret_val = 0;
11743 /* update metadata locally */
11744 imsm_update_metadata_locally(st, u, len);
11745 /* and possibly remotely */
11746 if (st->update_tail)
11747 append_metadata_update(st, u, len);
11748 else
ed08d51c 11749 free(u);
8dd70bce 11750
694575e7 11751 } else {
7a862a02 11752 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11753 }
11754 } else {
11755 /* On volume level we support following operations
471bceb6
KW		 11756 * - takeover: raid10 -> raid0; raid0 -> raid10
11757 * - chunk size migration
11758 * - migration: raid5 -> raid0; raid0 -> raid5
11759 */
11760 struct intel_super *super = st->sb;
11761 struct intel_dev *dev = super->devlist;
4dd2df09 11762 int change;
694575e7 11763 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11764 /* find requested device */
11765 while (dev) {
1011e834 11766 char *devnm =
4dd2df09
N		 11767 imsm_find_array_devnm_by_subdev(
11768 dev->index, st->container_devnm);
11769 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11770 break;
11771 dev = dev->next;
11772 }
11773 if (dev == NULL) {
4dd2df09
N		 11774 pr_err("Cannot find %s (%s) subarray\n",
11775 geo.dev_name, geo.devnm);
471bceb6
KW		 11776 goto exit_imsm_reshape_super;
11777 }
11778 super->current_vol = dev->index;
fbf3d202 11779 change = imsm_analyze_change(st, &geo, direction);
694575e7 11780 switch (change) {
471bceb6 11781 case CH_TAKEOVER:
bb025c2f 11782 ret_val = imsm_takeover(st, &geo);
694575e7 11783 break;
48c5303a
PC		 11784 case CH_MIGRATION: {
11785 struct imsm_update_reshape_migration *u = NULL;
11786 int len =
11787 imsm_create_metadata_update_for_migration(
11788 st, &geo, &u);
11789 if (len < 1) {
7a862a02 11790 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11791 break;
11792 }
471bceb6 11793 ret_val = 0;
48c5303a
PC		 11794 /* update metadata locally */
11795 imsm_update_metadata_locally(st, u, len);
11796 /* and possibly remotely */
11797 if (st->update_tail)
11798 append_metadata_update(st, u, len);
11799 else
11800 free(u);
11801 }
11802 break;
7abc9871 11803 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11804 struct imsm_update_size_change *u = NULL;
11805 int len =
11806 imsm_create_metadata_update_for_size_change(
11807 st, &geo, &u);
11808 if (len < 1) {
7a862a02 11809 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11810 break;
11811 }
11812 ret_val = 0;
11813 /* update metadata locally */
11814 imsm_update_metadata_locally(st, u, len);
11815 /* and possibly remotely */
11816 if (st->update_tail)
11817 append_metadata_update(st, u, len);
11818 else
11819 free(u);
7abc9871
AK		 11820 }
11821 break;
471bceb6
KW		 11822 default:
11823 ret_val = 1;
694575e7 11824 }
694575e7 11825 }
78b10e66 11826
ed08d51c 11827exit_imsm_reshape_super:
78b10e66
N		 11828 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11829 return ret_val;
11830}
2cda7640 11831
0febb20c
AO		 11832#define COMPLETED_OK 0
11833#define COMPLETED_NONE 1
11834#define COMPLETED_DELAYED 2
11835
11836static int read_completed(int fd, unsigned long long *val)
11837{
11838 int ret;
11839 char buf[50];
11840
11841 ret = sysfs_fd_get_str(fd, buf, 50);
11842 if (ret < 0)
11843 return ret;
11844
11845 ret = COMPLETED_OK;
11846 if (strncmp(buf, "none", 4) == 0) {
11847 ret = COMPLETED_NONE;
11848 } else if (strncmp(buf, "delayed", 7) == 0) {
11849 ret = COMPLETED_DELAYED;
11850 } else {
11851 char *ep;
11852 *val = strtoull(buf, &ep, 0);
11853 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11854 ret = -1;
11855 }
11856 return ret;
11857}
11858
eee67a47
AK		 11859/*******************************************************************************
11860 * Function: wait_for_reshape_imsm
11861 * Description: Function writes new sync_max value and waits until
11862 * reshape process reach new position
11863 * Parameters:
11864 * sra : general array info
eee67a47
AK		 11865 * ndata : number of disks in new array's layout
11866 * Returns:
11867 * 0 : success,
11868 * 1 : there is no reshape in progress,
11869 * -1 : fail
11870 ******************************************************************************/
ae9f01f8 11871int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11872{
85ca499c 11873 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11874 int retry = 3;
eee67a47 11875 unsigned long long completed;
ae9f01f8
AK		 11876 /* to_complete : new sync_max position */
11877 unsigned long long to_complete = sra->reshape_progress;
11878 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11879
ae9f01f8 11880 if (fd < 0) {
1ade5cc1 11881 dprintf("cannot open reshape_position\n");
eee67a47 11882 return 1;
ae9f01f8 11883 }
eee67a47 11884
df2647fa
PB		 11885 do {
11886 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11887 if (!retry) {
11888 dprintf("cannot read reshape_position (no reshape in progres)\n");
11889 close(fd);
11890 return 1;
11891 }
11892 usleep(30000);
11893 } else
11894 break;
11895 } while (retry--);
eee67a47 11896
85ca499c 11897 if (completed > position_to_set) {
1ade5cc1 11898 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11899 to_complete, position_to_set);
ae9f01f8
AK		 11900 close(fd);
11901 return -1;
11902 }
11903 dprintf("Position set: %llu\n", position_to_set);
11904 if (sysfs_set_num(sra, NULL, "sync_max",
11905 position_to_set) != 0) {
1ade5cc1 11906 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11907 position_to_set);
11908 close(fd);
11909 return -1;
eee67a47
AK		 11910 }
11911
eee67a47 11912 do {
0febb20c 11913 int rc;
eee67a47 11914 char action[20];
5ff3a780 11915 int timeout = 3000;
0febb20c 11916
5ff3a780 11917 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11918 if (sysfs_get_str(sra, NULL, "sync_action",
11919 action, 20) > 0 &&
d7d3809a 11920 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11921 if (strncmp(action, "idle", 4) == 0)
11922 break;
d7d3809a
AP		 11923 close(fd);
11924 return -1;
11925 }
0febb20c
AO		 11926
11927 rc = read_completed(fd, &completed);
11928 if (rc < 0) {
1ade5cc1 11929 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11930 close(fd);
11931 return 1;
0febb20c
AO		 11932 } else if (rc == COMPLETED_NONE)
11933 break;
85ca499c 11934 } while (completed < position_to_set);
b2be2b62 11935
eee67a47
AK		 11936 close(fd);
11937 return 0;
eee67a47
AK		 11938}
11939
b915c95f
AK		 11940/*******************************************************************************
11941 * Function: check_degradation_change
11942 * Description: Check that array hasn't become failed.
11943 * Parameters:
11944 * info : for sysfs access
11945 * sources : source disks descriptors
11946 * degraded: previous degradation level
11947 * Returns:
11948 * degradation level
11949 ******************************************************************************/
11950int check_degradation_change(struct mdinfo *info,
11951 int *sources,
11952 int degraded)
11953{
11954 unsigned long long new_degraded;
e1993023
LD		 11955 int rv;
11956
11957 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11958 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11959 /* check each device to ensure it is still working */
11960 struct mdinfo *sd;
11961 new_degraded = 0;
11962 for (sd = info->devs ; sd ; sd = sd->next) {
11963 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11964 continue;
11965 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11966 char sbuf[100];
11967
b915c95f 11968 if (sysfs_get_str(info,
cf52eff5 11969 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11970 strstr(sbuf, "faulty") ||
11971 strstr(sbuf, "in_sync") == NULL) {
11972 /* this device is dead */
11973 sd->disk.state = (1<<MD_DISK_FAULTY);
11974 if (sd->disk.raid_disk >= 0 &&
11975 sources[sd->disk.raid_disk] >= 0) {
11976 close(sources[
11977 sd->disk.raid_disk]);
11978 sources[sd->disk.raid_disk] =
11979 -1;
11980 }
11981 new_degraded++;
11982 }
11983 }
11984 }
11985 }
11986
11987 return new_degraded;
11988}
11989
10f22854
AK		 11990/*******************************************************************************
11991 * Function: imsm_manage_reshape
11992 * Description: Function finds array under reshape and it manages reshape
11993 * process. It creates stripes backups (if required) and sets
942e1cdb 11994 * checkpoints.
10f22854
AK		 11995 * Parameters:
11996 * afd : Backup handle (nattive) - not used
11997 * sra : general array info
11998 * reshape : reshape parameters - not used
11999 * st : supertype structure
12000 * blocks : size of critical section [blocks]
12001 * fds : table of source device descriptor
12002 * offsets : start of array (offest per devices)
12003 * dests : not used
12004 * destfd : table of destination device descriptor
12005 * destoffsets : table of destination offsets (per device)
12006 * Returns:
12007 * 1 : success, reshape is done
12008 * 0 : fail
12009 ******************************************************************************/
999b4972
N		 12010static int imsm_manage_reshape(
12011 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12012 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12013 int *fds, unsigned long long *offsets,
12014 int dests, int *destfd, unsigned long long *destoffsets)
12015{
10f22854
AK		 12016 int ret_val = 0;
12017 struct intel_super *super = st->sb;
594dc1b8 12018 struct intel_dev *dv;
de44e46f 12019 unsigned int sector_size = super->sector_size;
10f22854 12020 struct imsm_dev *dev = NULL;
9529d343 12021 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12022 int migr_vol_qan = 0;
12023 int ndata, odata; /* [bytes] */
12024 int chunk; /* [bytes] */
12025 struct migr_record *migr_rec;
12026 char *buf = NULL;
12027 unsigned int buf_size; /* [bytes] */
12028 unsigned long long max_position; /* array size [bytes] */
12029 unsigned long long next_step; /* [blocks]/[bytes] */
12030 unsigned long long old_data_stripe_length;
10f22854
AK		 12031 unsigned long long start_src; /* [bytes] */
12032 unsigned long long start; /* [bytes] */
12033 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12034 int degraded = 0;
ab724b98 12035 int source_layout = 0;
10f22854 12036
79a16a9b
JS		 12037 if (!sra)
12038 return ret_val;
12039
12040 if (!fds || !offsets)
10f22854
AK		 12041 goto abort;
12042
12043 /* Find volume during the reshape */
12044 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12045 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12046 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12047 dev = dv->dev;
12048 migr_vol_qan++;
12049 }
12050 }
12051 /* Only one volume can migrate at the same time */
12052 if (migr_vol_qan != 1) {
676e87a8 12053 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12054 "Number of migrating volumes greater than 1\n" :
12055 "There is no volume during migrationg\n");
12056 goto abort;
12057 }
12058
9529d343 12059 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12060 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12061 if (map_src == NULL)
12062 goto abort;
10f22854 12063
9529d343
MD		 12064 ndata = imsm_num_data_members(map_dest);
12065 odata = imsm_num_data_members(map_src);
10f22854 12066
7b1ab482 12067 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12068 old_data_stripe_length = odata * chunk;
12069
12070 migr_rec = super->migr_rec;
12071
10f22854
AK		 12072 /* initialize migration record for start condition */
12073 if (sra->reshape_progress == 0)
12074 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12075 else {
12076 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12077 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12078 goto abort;
12079 }
6a75c8ca
AK		 12080 /* Save checkpoint to update migration record for current
12081 * reshape position (in md). It can be farther than current
12082 * reshape position in metadata.
12083 */
12084 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12085 /* ignore error == 2, this can mean end of reshape here
12086 */
7a862a02 12087 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12088 goto abort;
12089 }
b2c59438 12090 }
10f22854
AK		 12091
12092 /* size for data */
12093 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12094 /* extend buffer size for parity disk */
12095 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12096 /* add space for stripe alignment */
10f22854 12097 buf_size += old_data_stripe_length;
de44e46f
PB		 12098 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12099 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12100 goto abort;
12101 }
12102
3ef4403c 12103 max_position = sra->component_size * ndata;
68eb8bc6 12104 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12105
9f421827
PB		 12106 while (current_migr_unit(migr_rec) <
12107 get_num_migr_units(migr_rec)) {
10f22854
AK		 12108 /* current reshape position [blocks] */
12109 unsigned long long current_position =
12110 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12111 * current_migr_unit(migr_rec);
10f22854
AK		 12112 unsigned long long border;
12113
b915c95f
AK		 12114 /* Check that array hasn't become failed.
12115 */
12116 degraded = check_degradation_change(sra, fds, degraded);
12117 if (degraded > 1) {
7a862a02 12118 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12119 goto abort;
12120 }
12121
10f22854
AK		 12122 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12123
12124 if ((current_position + next_step) > max_position)
12125 next_step = max_position - current_position;
12126
92144abf 12127 start = current_position * 512;
10f22854 12128
942e1cdb 12129 /* align reading start to old geometry */
10f22854
AK		 12130 start_buf_shift = start % old_data_stripe_length;
12131 start_src = start - start_buf_shift;
12132
12133 border = (start_src / odata) - (start / ndata);
12134 border /= 512;
12135 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12136 /* save critical stripes to buf
12137 * start - start address of current unit
12138 * to backup [bytes]
12139 * start_src - start address of current unit
12140 * to backup alligned to source array
12141 * [bytes]
12142 */
594dc1b8 12143 unsigned long long next_step_filler;
10f22854
AK		 12144 unsigned long long copy_length = next_step * 512;
12145
12146 /* allign copy area length to stripe in old geometry */
12147 next_step_filler = ((copy_length + start_buf_shift)
12148 % old_data_stripe_length);
12149 if (next_step_filler)
12150 next_step_filler = (old_data_stripe_length
12151 - next_step_filler);
7a862a02 12152 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		 12153 start, start_src, copy_length,
12154 start_buf_shift, next_step_filler);
12155
12156 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12157 chunk, map_src->raid_level,
12158 source_layout, 0, NULL, start_src,
10f22854
AK		 12159 copy_length +
12160 next_step_filler + start_buf_shift,
12161 buf)) {
7a862a02 12162 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12163 goto abort;
12164 }
12165 /* Convert data to destination format and store it
12166 * in backup general migration area
12167 */
12168 if (save_backup_imsm(st, dev, sra,
aea93171 12169 buf + start_buf_shift, copy_length)) {
7a862a02 12170 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12171 goto abort;
12172 }
12173 if (save_checkpoint_imsm(st, sra,
12174 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12175 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12176 goto abort;
12177 }
8016a6d4
AK		 12178 } else {
12179 /* set next step to use whole border area */
12180 border /= next_step;
12181 if (border > 1)
12182 next_step *= border;
10f22854
AK		 12183 }
12184 /* When data backed up, checkpoint stored,
12185 * kick the kernel to reshape unit of data
12186 */
12187 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12188 /* limit next step to array max position */
12189 if (next_step > max_position)
12190 next_step = max_position;
10f22854
AK		 12191 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12192 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12193 sra->reshape_progress = next_step;
10f22854
AK		 12194
12195 /* wait until reshape finish */
c85338c6 12196 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12197 dprintf("wait_for_reshape_imsm returned error!\n");
12198 goto abort;
12199 }
84d11e6c
N		 12200 if (sigterm)
12201 goto abort;
10f22854 12202
0228d92c
AK		 12203 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12204 /* ignore error == 2, this can mean end of reshape here
12205 */
7a862a02 12206 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12207 goto abort;
12208 }
12209
12210 }
12211
71e5411e
PB		 12212 /* clear migr_rec on disks after successful migration */
12213 struct dl *d;
12214
85337573 12215 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12216 for (d = super->disks; d; d = d->next) {
12217 if (d->index < 0 || is_failed(&d->disk))
12218 continue;
12219 unsigned long long dsize;
12220
12221 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12222 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12223 SEEK_SET) >= 0) {
466070ad 12224 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12225 MIGR_REC_BUF_SECTORS*sector_size) !=
12226 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12227 perror("Write migr_rec failed");
12228 }
12229 }
12230
10f22854
AK		 12231 /* return '1' if done */
12232 ret_val = 1;
12233abort:
12234 free(buf);
942e1cdb
N		 12235 /* See Grow.c: abort_reshape() for further explanation */
12236 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12237 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12238 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12239
12240 return ret_val;
999b4972 12241}
0c21b485 12242
cdddbdbc 12243struct superswitch super_imsm = {
cdddbdbc
DW		 12244 .examine_super = examine_super_imsm,
12245 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12246 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12247 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12248 .detail_super = detail_super_imsm,
12249 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12250 .write_init_super = write_init_super_imsm,
0e600426
N		 12251 .validate_geometry = validate_geometry_imsm,
12252 .add_to_super = add_to_super_imsm,
1a64be56 12253 .remove_from_super = remove_from_super_imsm,
d665cc31 12254 .detail_platform = detail_platform_imsm,
e50cf220 12255 .export_detail_platform = export_detail_platform_imsm,
33414a01 12256 .kill_subarray = kill_subarray_imsm,
aa534678 12257 .update_subarray = update_subarray_imsm,
2b959fbf 12258 .load_container = load_container_imsm,
71204a50
N		 12259 .default_geometry = default_geometry_imsm,
12260 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12261 .reshape_super = imsm_reshape_super,
12262 .manage_reshape = imsm_manage_reshape,
9e2d750d 12263 .recover_backup = recover_backup_imsm,
74db60b0 12264 .copy_metadata = copy_metadata_imsm,
27156a57 12265 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12266 .match_home = match_home_imsm,
12267 .uuid_from_super= uuid_from_super_imsm,
12268 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12269 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12270 .update_super = update_super_imsm,
12271
12272 .avail_size = avail_size_imsm,
fbfdcb06 12273 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12274
12275 .compare_super = compare_super_imsm,
12276
12277 .load_super = load_super_imsm,
bf5a934a 12278 .init_super = init_super_imsm,
e683ca88 12279 .store_super = store_super_imsm,
cdddbdbc
DW		 12280 .free_super = free_super_imsm,
12281 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12282 .container_content = container_content_imsm,
0c21b485 12283 .validate_container = validate_container_imsm,
cdddbdbc 12284
2432ce9b
AP		 12285 .write_init_ppl = write_init_ppl_imsm,
12286 .validate_ppl = validate_ppl_imsm,
12287
cdddbdbc 12288 .external = 1,
4cce4069 12289 .name = "imsm",
845dea95
NB		 12290
12291/* for mdmon */
12292 .open_new = imsm_open_new,
ed9d66aa 12293 .set_array_state= imsm_set_array_state,
845dea95
NB		 12294 .set_disk = imsm_set_disk,
12295 .sync_metadata = imsm_sync_metadata,
88758e9d 12296 .activate_spare = imsm_activate_spare,
e8319a19 12297 .process_update = imsm_process_update,
8273f55e 12298 .prepare_update = imsm_prepare_update,
6f50473f 12299 .record_bad_block = imsm_record_badblock,
c07a5a4f 12300 .clear_bad_block = imsm_clear_badblock,
928f1424 12301 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12302};
Unnamed repository; edit this file 'description' to name the repository.