]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
imsm: support the Array Creation Time field in metadata
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
cdddbdbc
DW		 1/*
2 * mdadm - Intel(R) Matrix Storage Manager Support
3 *
a54d5262 4 * Copyright (C) 2002-2008 Intel Corporation
cdddbdbc
DW		 5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 */
19
51006d85 20#define HAVE_STDINT_H 1
cdddbdbc 21#include "mdadm.h"
c2a1e7da 22#include "mdmon.h"
51006d85 23#include "sha1.h"
88c32bb1 24#include "platform-intel.h"
cdddbdbc
DW		 25#include <values.h>
26#include <scsi/sg.h>
27#include <ctype.h>
d665cc31 28#include <dirent.h>
cdddbdbc
DW		 29
30/* MPB == Metadata Parameter Block */
31#define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
32#define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
33#define MPB_VERSION_RAID0 "1.0.00"
34#define MPB_VERSION_RAID1 "1.1.00"
fe7ed8cb
DW		 35#define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00"
36#define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01"
cdddbdbc 37#define MPB_VERSION_RAID5 "1.2.02"
fe7ed8cb
DW		 38#define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04"
39#define MPB_VERSION_CNG "1.2.06"
40#define MPB_VERSION_ATTRIBS "1.3.00"
cdddbdbc
DW		 41#define MAX_SIGNATURE_LENGTH 32
42#define MAX_RAID_SERIAL_LEN 16
fe7ed8cb 43
19482bcc
AK		 44/* supports RAID0 */
45#define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001)
46/* supports RAID1 */
47#define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002)
48/* supports RAID10 */
49#define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004)
50/* supports RAID1E */
51#define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008)
52/* supports RAID5 */
53#define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010)
54/* supports RAID CNG */
55#define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020)
56/* supports expanded stripe sizes of 256K, 512K and 1MB */
57#define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040)
58
59/* The OROM Support RST Caching of Volumes */
60#define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000)
61/* The OROM supports creating disks greater than 2TB */
62#define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000)
63/* The OROM supports Bad Block Management */
64#define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000)
65
66/* THe OROM Supports NVM Caching of Volumes */
67#define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000)
68/* The OROM supports creating volumes greater than 2TB */
69#define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000)
70/* originally for PMP, now it's wasted b/c. Never use this bit! */
71#define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000)
72/* Verify MPB contents against checksum after reading MPB */
73#define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000)
74
75/* Define all supported attributes that have to be accepted by mdadm
76 */
418f9b36 77#define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \
19482bcc
AK		 78 MPB_ATTRIB_2TB | \
79 MPB_ATTRIB_2TB_DISK | \
80 MPB_ATTRIB_RAID0 | \
81 MPB_ATTRIB_RAID1 | \
82 MPB_ATTRIB_RAID10 | \
83 MPB_ATTRIB_RAID5 | \
bbab0940
TM		 84 MPB_ATTRIB_EXP_STRIPE_SIZE | \
85 MPB_ATTRIB_BBM)
418f9b36
N		 86
87/* Define attributes that are unused but not harmful */
88#define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE)
fe7ed8cb 89
8e59f3d8 90#define MPB_SECTOR_CNT 2210
611d9529
MD		 91#define IMSM_RESERVED_SECTORS 8192
92#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048
979d38be 93#define SECT_PER_MB_SHIFT 11
f36a9ecd 94#define MAX_SECTOR_SIZE 4096
c2462068
PB		 95#define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole
96 * mutliple PPL area
97 */
cdddbdbc 98
761e3bd9
N		 99/*
100 * This macro let's us ensure that no-one accidentally
101 * changes the size of a struct
102 */
103#define ASSERT_SIZE(_struct, size) \
104static inline void __assert_size_##_struct(void) \
105{ \
106 switch (0) { \
107 case 0: break; \
108 case (sizeof(struct _struct) == size): break; \
109 } \
110}
111
cdddbdbc
DW		 112/* Disk configuration info. */
113#define IMSM_MAX_DEVICES 255
114struct imsm_disk {
115 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 116 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 117 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 118#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
119#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
120#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 121#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 122 __u32 status; /* 0xF0 - 0xF3 */
1011e834 123 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 124 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
125#define IMSM_DISK_FILLERS 3
126 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 127};
761e3bd9 128ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 129
3b451610
AK		 130/* map selector for map managment
131 */
238c0a71
AK		 132#define MAP_0 0
133#define MAP_1 1
134#define MAP_X -1
3b451610 135
cdddbdbc
DW		 136/* RAID map configuration infos. */
137struct imsm_map {
5551b113
CA		 138 __u32 pba_of_lba0_lo; /* start address of partition */
139 __u32 blocks_per_member_lo;/* blocks per member */
140 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 141 __u16 blocks_per_strip;
142 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
143#define IMSM_T_STATE_NORMAL 0
144#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 145#define IMSM_T_STATE_DEGRADED 2
146#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 147 __u8 raid_level;
148#define IMSM_T_RAID0 0
149#define IMSM_T_RAID1 1
150#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
151 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 152 __u8 num_domains; /* number of parity domains */
153 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 154 __u8 ddf;
5551b113
CA		 155 __u32 pba_of_lba0_hi;
156 __u32 blocks_per_member_hi;
157 __u32 num_data_stripes_hi;
158 __u32 filler[4]; /* expansion area */
7eef0453 159#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 160 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 161 * top byte contains some flags
162 */
761e3bd9
N		 163};
164ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 165
166struct imsm_vol {
f8f603f1 167 __u32 curr_migr_unit;
fe7ed8cb 168 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 169 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 170#define MIGR_INIT 0
171#define MIGR_REBUILD 1
172#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
173#define MIGR_GEN_MIGR 3
174#define MIGR_STATE_CHANGE 4
1484e727 175#define MIGR_REPAIR 5
cdddbdbc 176 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 177#define RAIDVOL_CLEAN 0
178#define RAIDVOL_DIRTY 1
179#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 180 __u8 dirty;
fe7ed8cb
DW		 181 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
182 __u16 verify_errors; /* number of mismatches */
183 __u16 bad_blocks; /* number of bad blocks during verify */
184 __u32 filler[4];
cdddbdbc
DW		 185 struct imsm_map map[1];
186 /* here comes another one if migr_state */
761e3bd9
N		 187};
188ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 189
190struct imsm_dev {
fe7ed8cb 191 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 192 __u32 size_low;
193 __u32 size_high;
fe7ed8cb
DW		 194#define DEV_BOOTABLE __cpu_to_le32(0x01)
195#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
196#define DEV_READ_COALESCING __cpu_to_le32(0x04)
197#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
198#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
199#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
200#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
201#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
202#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
203#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
204#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
205#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
206#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 207 __u32 status; /* Persistent RaidDev status */
208 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 209 __u8 migr_priority;
210 __u8 num_sub_vols;
211 __u8 tid;
212 __u8 cng_master_disk;
213 __u16 cache_policy;
214 __u8 cng_state;
215 __u8 cng_sub_state;
2432ce9b
AP		 216 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
217
218 /* NVM_EN */
219 __u8 nv_cache_mode;
220 __u8 nv_cache_flags;
221
222 /* Unique Volume Id of the NvCache Volume associated with this volume */
223 __u32 nvc_vol_orig_family_num;
224 __u16 nvc_vol_raid_dev_num;
225
226#define RWH_OFF 0
227#define RWH_DISTRIBUTED 1
228#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 229#define RWH_MULTIPLE_DISTRIBUTED 3
230#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
231#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 232 __u8 rwh_policy; /* Raid Write Hole Policy */
233 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
234 __u8 filler1;
235
236#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 237 __u32 filler[IMSM_DEV_FILLERS];
238 struct imsm_vol vol;
761e3bd9
N		 239};
240ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 241
242struct imsm_super {
243 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
244 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
245 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
246 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
247 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 248 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
249 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 250 __u8 num_disks; /* 0x38 Number of configured disks */
251 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 252 __u8 error_log_pos; /* 0x3A */
253 __u8 fill[1]; /* 0x3B */
254 __u32 cache_size; /* 0x3c - 0x40 in mb */
255 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
256 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
257 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 258 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
259 * volume IDs for raid_dev created in this array
260 * (starts at 1)
261 */
262 __u16 filler1; /* 0x4E - 0x4F */
e48aed3c
AP		 263 __u64 creation_time; /* 0x50 - 0x57 Array creation time */
264#define IMSM_FILLERS 32
265 __u32 filler[IMSM_FILLERS]; /* 0x58 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 266 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
267 /* here comes imsm_dev[num_raid_devs] */
604b746f 268 /* here comes BBM logs */
761e3bd9
N		 269};
270ASSERT_SIZE(imsm_super, 264)
cdddbdbc 271
604b746f 272#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 273#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
274#define BBM_LOG_SIGNATURE 0xabadb10c
275
276struct bbm_log_block_addr {
277 __u16 w1;
278 __u32 dw1;
279} __attribute__ ((__packed__));
604b746f
JD		 280
281struct bbm_log_entry {
8d67477f
TM		 282 __u8 marked_count; /* Number of blocks marked - 1 */
283 __u8 disk_ordinal; /* Disk entry within the imsm_super */
284 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 285} __attribute__ ((__packed__));
286
287struct bbm_log {
288 __u32 signature; /* 0xABADB10C */
289 __u32 entry_count;
8d67477f 290 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 291};
292ASSERT_SIZE(bbm_log, 2040)
604b746f 293
cdddbdbc 294static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 295
b53bfba6
TM		 296#define BLOCKS_PER_KB (1024/512)
297
8e59f3d8
AK		 298#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
299
300#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
301
de44e46f
PB		 302#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
303#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
304 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 305 */
306
8e59f3d8
AK		 307#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
308 * be recovered using srcMap */
309#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
310 * already been migrated and must
311 * be recovered from checkpoint area */
2432ce9b 312
c2462068 313#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 314
8e59f3d8
AK		 315struct migr_record {
316 __u32 rec_status; /* Status used to determine how to restart
317 * migration in case it aborts
318 * in some fashion */
9f421827 319 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 320 __u32 family_num; /* Family number of MPB
321 * containing the RaidDev
322 * that is migrating */
323 __u32 ascending_migr; /* True if migrating in increasing
324 * order of lbas */
325 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
326 __u32 dest_depth_per_unit; /* Num member blocks each destMap
327 * member disk
328 * advances per unit-of-operation */
9f421827
PB		 329 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
330 __u32 dest_1st_member_lba_lo; /* First member lba on first
331 * stripe of destination */
332 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 333 __u32 post_migr_vol_cap; /* Size of volume after
334 * migration completes */
335 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
336 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
337 * migration ckpt record was read from
338 * (for recovered migrations) */
9f421827
PB		 339 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
340 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
341 * high order 32 bits */
342 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
343 * destination - high order 32 bits */
344 __u32 num_migr_units_hi; /* Total num migration units-of-op
345 * high order 32 bits */
761e3bd9
N		 346};
347ASSERT_SIZE(migr_record, 64)
8e59f3d8 348
ec50f7b6
LM		 349struct md_list {
350 /* usage marker:
351 * 1: load metadata
352 * 2: metadata does not match
353 * 4: already checked
354 */
355 int used;
356 char *devname;
357 int found;
358 int container;
359 dev_t st_rdev;
360 struct md_list *next;
361};
362
e7b84f9d 363#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 364
1484e727
DW		 365static __u8 migr_type(struct imsm_dev *dev)
366{
367 if (dev->vol.migr_type == MIGR_VERIFY &&
368 dev->status & DEV_VERIFY_AND_FIX)
369 return MIGR_REPAIR;
370 else
371 return dev->vol.migr_type;
372}
373
374static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
375{
376 /* for compatibility with older oroms convert MIGR_REPAIR, into
377 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
378 */
379 if (migr_type == MIGR_REPAIR) {
380 dev->vol.migr_type = MIGR_VERIFY;
381 dev->status |= DEV_VERIFY_AND_FIX;
382 } else {
383 dev->vol.migr_type = migr_type;
384 dev->status &= ~DEV_VERIFY_AND_FIX;
385 }
386}
387
f36a9ecd 388static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 389{
f36a9ecd 390 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 391}
cdddbdbc 392
f36a9ecd
PB		 393static unsigned int mpb_sectors(struct imsm_super *mpb,
394 unsigned int sector_size)
87eb16df 395{
f36a9ecd 396 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 397}
398
ba2de7ba
DW		 399struct intel_dev {
400 struct imsm_dev *dev;
401 struct intel_dev *next;
f21e18ca 402 unsigned index;
ba2de7ba
DW		 403};
404
88654014
LM		 405struct intel_hba {
406 enum sys_dev_type type;
407 char *path;
408 char *pci_id;
409 struct intel_hba *next;
410};
411
1a64be56
LM		 412enum action {
413 DISK_REMOVE = 1,
414 DISK_ADD
415};
cdddbdbc
DW		 416/* internal representation of IMSM metadata */
417struct intel_super {
418 union {
949c47a0
DW		 419 void *buf; /* O_DIRECT buffer for reading/writing metadata */
420 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 421 };
8e59f3d8
AK		 422 union {
423 void *migr_rec_buf; /* buffer for I/O operations */
424 struct migr_record *migr_rec; /* migration record */
425 };
51d83f5d
AK		 426 int clean_migration_record_by_mdmon; /* when reshape is switched to next
427 array, it indicates that mdmon is allowed to clean migration
428 record */
949c47a0 429 size_t len; /* size of the 'buf' allocation */
bbab0940 430 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 431 void *next_buf; /* for realloc'ing buf from the manager */
432 size_t next_len;
c2c087e6 433 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 434 int current_vol; /* index of raid device undergoing creation */
5551b113 435 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 436 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 437 struct intel_dev *devlist;
fa7bb6f8 438 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 439 struct dl {
440 struct dl *next;
441 int index;
442 __u8 serial[MAX_RAID_SERIAL_LEN];
443 int major, minor;
444 char *devname;
b9f594fe 445 struct imsm_disk disk;
cdddbdbc 446 int fd;
0dcecb2e
DW		 447 int extent_cnt;
448 struct extent *e; /* for determining freespace @ create */
efb30e7f 449 int raiddisk; /* slot to fill in autolayout */
1a64be56 450 enum action action;
ca0748fa 451 } *disks, *current_disk;
1a64be56
LM		 452 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
453 active */
47ee5a45 454 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 455 struct bbm_log *bbm_log;
88654014 456 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 457 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 458 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 459 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 460};
461
462struct intel_disk {
463 struct imsm_disk disk;
464 #define IMSM_UNKNOWN_OWNER (-1)
465 int owner;
466 struct intel_disk *next;
cdddbdbc
DW		 467};
468
c2c087e6
DW		 469struct extent {
470 unsigned long long start, size;
471};
472
694575e7
KW		 473/* definitions of reshape process types */
474enum imsm_reshape_type {
475 CH_TAKEOVER,
b5347799 476 CH_MIGRATION,
7abc9871 477 CH_ARRAY_SIZE,
694575e7
KW		 478};
479
88758e9d
DW		 480/* definition of messages passed to imsm_process_update */
481enum imsm_update_type {
482 update_activate_spare,
8273f55e 483 update_create_array,
33414a01 484 update_kill_array,
aa534678 485 update_rename_array,
1a64be56 486 update_add_remove_disk,
78b10e66 487 update_reshape_container_disks,
48c5303a 488 update_reshape_migration,
2d40f3a1
AK		 489 update_takeover,
490 update_general_migration_checkpoint,
f3871fdc 491 update_size_change,
bbab0940 492 update_prealloc_badblocks_mem,
e6e9dd3f 493 update_rwh_policy,
88758e9d
DW		 494};
495
496struct imsm_update_activate_spare {
497 enum imsm_update_type type;
d23fe947 498 struct dl *dl;
88758e9d
DW		 499 int slot;
500 int array;
501 struct imsm_update_activate_spare *next;
502};
503
78b10e66 504struct geo_params {
4dd2df09 505 char devnm[32];
78b10e66 506 char *dev_name;
d04f65f4 507 unsigned long long size;
78b10e66
N		 508 int level;
509 int layout;
510 int chunksize;
511 int raid_disks;
512};
513
bb025c2f
KW		 514enum takeover_direction {
515 R10_TO_R0,
516 R0_TO_R10
517};
518struct imsm_update_takeover {
519 enum imsm_update_type type;
520 int subarray;
521 enum takeover_direction direction;
522};
78b10e66
N		 523
524struct imsm_update_reshape {
525 enum imsm_update_type type;
526 int old_raid_disks;
527 int new_raid_disks;
48c5303a
PC		 528
529 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
530};
531
532struct imsm_update_reshape_migration {
533 enum imsm_update_type type;
534 int old_raid_disks;
535 int new_raid_disks;
536 /* fields for array migration changes
537 */
538 int subdev;
539 int new_level;
540 int new_layout;
4bba0439 541 int new_chunksize;
48c5303a 542
d195167d 543 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 544};
545
f3871fdc
AK		 546struct imsm_update_size_change {
547 enum imsm_update_type type;
548 int subdev;
549 long long new_size;
550};
551
2d40f3a1
AK		 552struct imsm_update_general_migration_checkpoint {
553 enum imsm_update_type type;
554 __u32 curr_migr_unit;
555};
556
54c2c1ea
DW		 557struct disk_info {
558 __u8 serial[MAX_RAID_SERIAL_LEN];
559};
560
8273f55e
DW		 561struct imsm_update_create_array {
562 enum imsm_update_type type;
8273f55e 563 int dev_idx;
6a3e913e 564 struct imsm_dev dev;
8273f55e
DW		 565};
566
33414a01
DW		 567struct imsm_update_kill_array {
568 enum imsm_update_type type;
569 int dev_idx;
570};
571
aa534678
DW		 572struct imsm_update_rename_array {
573 enum imsm_update_type type;
574 __u8 name[MAX_RAID_SERIAL_LEN];
575 int dev_idx;
576};
577
1a64be56 578struct imsm_update_add_remove_disk {
43dad3d6
DW		 579 enum imsm_update_type type;
580};
581
bbab0940
TM		 582struct imsm_update_prealloc_bb_mem {
583 enum imsm_update_type type;
584};
585
e6e9dd3f
AP		 586struct imsm_update_rwh_policy {
587 enum imsm_update_type type;
588 int new_policy;
589 int dev_idx;
590};
591
88654014
LM		 592static const char *_sys_dev_type[] = {
593 [SYS_DEV_UNKNOWN] = "Unknown",
594 [SYS_DEV_SAS] = "SAS",
614902f6 595 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 596 [SYS_DEV_NVME] = "NVMe",
597 [SYS_DEV_VMD] = "VMD"
88654014
LM		 598};
599
600const char *get_sys_dev_type(enum sys_dev_type type)
601{
602 if (type >= SYS_DEV_MAX)
603 type = SYS_DEV_UNKNOWN;
604
605 return _sys_dev_type[type];
606}
607
608static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
609{
503975b9
N		 610 struct intel_hba *result = xmalloc(sizeof(*result));
611
612 result->type = device->type;
613 result->path = xstrdup(device->path);
614 result->next = NULL;
615 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
616 result->pci_id++;
617
88654014
LM		 618 return result;
619}
620
621static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
622{
594dc1b8
JS		 623 struct intel_hba *result;
624
88654014
LM		 625 for (result = hba; result; result = result->next) {
626 if (result->type == device->type && strcmp(result->path, device->path) == 0)
627 break;
628 }
629 return result;
630}
631
b4cf4cba 632static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 633{
634 struct intel_hba *hba;
635
636 /* check if disk attached to Intel HBA */
637 hba = find_intel_hba(super->hba, device);
638 if (hba != NULL)
639 return 1;
640 /* Check if HBA is already attached to super */
641 if (super->hba == NULL) {
642 super->hba = alloc_intel_hba(device);
643 return 1;
6b781d33
AP		 644 }
645
646 hba = super->hba;
647 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 648 * Do not support HBA types mixing
6b781d33
AP		 649 */
650 if (device->type != hba->type)
88654014 651 return 2;
6b781d33
AP		 652
653 /* Multiple same type HBAs can be used if they share the same OROM */
654 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
655
656 if (device_orom != super->orom)
657 return 2;
658
659 while (hba->next)
660 hba = hba->next;
661
662 hba->next = alloc_intel_hba(device);
663 return 1;
88654014
LM		 664}
665
666static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
667{
9bc4ae77 668 struct sys_dev *list, *elem;
88654014
LM		 669 char *disk_path;
670
671 if ((list = find_intel_devices()) == NULL)
672 return 0;
673
674 if (fd < 0)
675 disk_path = (char *) devname;
676 else
677 disk_path = diskfd_to_devpath(fd);
678
9bc4ae77 679 if (!disk_path)
88654014 680 return 0;
88654014 681
9bc4ae77
N		 682 for (elem = list; elem; elem = elem->next)
683 if (path_attached_to_hba(disk_path, elem->path))
88654014 684 return elem;
9bc4ae77 685
88654014
LM		 686 if (disk_path != devname)
687 free(disk_path);
88654014
LM		 688
689 return NULL;
690}
691
d424212e
N		 692static int find_intel_hba_capability(int fd, struct intel_super *super,
693 char *devname);
f2f5c343 694
cdddbdbc
DW		 695static struct supertype *match_metadata_desc_imsm(char *arg)
696{
697 struct supertype *st;
698
699 if (strcmp(arg, "imsm") != 0 &&
700 strcmp(arg, "default") != 0
701 )
702 return NULL;
703
503975b9 704 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 705 st->ss = &super_imsm;
706 st->max_devs = IMSM_MAX_DEVICES;
707 st->minor_version = 0;
708 st->sb = NULL;
709 return st;
710}
711
cdddbdbc
DW		 712static __u8 *get_imsm_version(struct imsm_super *mpb)
713{
714 return &mpb->sig[MPB_SIG_LEN];
715}
716
949c47a0
DW		 717/* retrieve a disk directly from the anchor when the anchor is known to be
718 * up-to-date, currently only at load time
719 */
720static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 721{
949c47a0 722 if (index >= mpb->num_disks)
cdddbdbc
DW		 723 return NULL;
724 return &mpb->disk[index];
725}
726
95d07a2c
LM		 727/* retrieve the disk description based on a index of the disk
728 * in the sub-array
729 */
730static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 731{
b9f594fe
DW		 732 struct dl *d;
733
734 for (d = super->disks; d; d = d->next)
735 if (d->index == index)
95d07a2c
LM		 736 return d;
737
738 return NULL;
739}
740/* retrieve a disk from the parsed metadata */
741static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
742{
743 struct dl *dl;
744
745 dl = get_imsm_dl_disk(super, index);
746 if (dl)
747 return &dl->disk;
748
b9f594fe 749 return NULL;
949c47a0
DW		 750}
751
752/* generate a checksum directly from the anchor when the anchor is known to be
753 * up-to-date, currently only at load or write_super after coalescing
754 */
755static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 756{
757 __u32 end = mpb->mpb_size / sizeof(end);
758 __u32 *p = (__u32 *) mpb;
759 __u32 sum = 0;
760
5d500228
N		 761 while (end--) {
762 sum += __le32_to_cpu(*p);
97f734fd
N		 763 p++;
764 }
cdddbdbc 765
5d500228 766 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 767}
768
a965f303
DW		 769static size_t sizeof_imsm_map(struct imsm_map *map)
770{
771 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
772}
773
774struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 775{
5e7b0330
AK		 776 /* A device can have 2 maps if it is in the middle of a migration.
777 * If second_map is:
238c0a71
AK		 778 * MAP_0 - we return the first map
779 * MAP_1 - we return the second map if it exists, else NULL
780 * MAP_X - we return the second map if it exists, else the first
5e7b0330 781 */
a965f303 782 struct imsm_map *map = &dev->vol.map[0];
9535fc47 783 struct imsm_map *map2 = NULL;
a965f303 784
9535fc47
AK		 785 if (dev->vol.migr_state)
786 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 787
9535fc47 788 switch (second_map) {
3b451610 789 case MAP_0:
9535fc47 790 break;
3b451610 791 case MAP_1:
9535fc47
AK		 792 map = map2;
793 break;
238c0a71 794 case MAP_X:
9535fc47
AK		 795 if (map2)
796 map = map2;
797 break;
9535fc47
AK		 798 default:
799 map = NULL;
800 }
801 return map;
5e7b0330 802
a965f303 803}
cdddbdbc 804
3393c6af
DW		 805/* return the size of the device.
806 * migr_state increases the returned size if map[0] were to be duplicated
807 */
808static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 809{
810 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 811 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 812
813 /* migrating means an additional map */
a965f303 814 if (dev->vol.migr_state)
238c0a71 815 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 816 else if (migr_state)
238c0a71 817 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 818
819 return size;
820}
821
54c2c1ea
DW		 822/* retrieve disk serial number list from a metadata update */
823static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
824{
825 void *u = update;
826 struct disk_info *inf;
827
828 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
829 sizeof_imsm_dev(&update->dev, 0);
830
831 return inf;
832}
54c2c1ea 833
949c47a0 834static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 835{
836 int offset;
837 int i;
838 void *_mpb = mpb;
839
949c47a0 840 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 841 return NULL;
842
843 /* devices start after all disks */
844 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
845
846 for (i = 0; i <= index; i++)
847 if (i == index)
848 return _mpb + offset;
849 else
3393c6af 850 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 851
852 return NULL;
853}
854
949c47a0
DW		 855static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
856{
ba2de7ba
DW		 857 struct intel_dev *dv;
858
949c47a0
DW		 859 if (index >= super->anchor->num_raid_devs)
860 return NULL;
ba2de7ba
DW		 861 for (dv = super->devlist; dv; dv = dv->next)
862 if (dv->index == index)
863 return dv->dev;
864 return NULL;
949c47a0
DW		 865}
866
8d67477f
TM		 867static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
868 *addr)
869{
870 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
871 __le16_to_cpu(addr->w1));
872}
873
874static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
875{
876 struct bbm_log_block_addr addr;
877
878 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
879 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
880 return addr;
881}
882
8d67477f
TM		 883/* get size of the bbm log */
884static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
885{
886 if (!log || log->entry_count == 0)
887 return 0;
888
889 return sizeof(log->signature) +
890 sizeof(log->entry_count) +
891 log->entry_count * sizeof(struct bbm_log_entry);
892}
6f50473f
TM		 893
894/* check if bad block is not partially stored in bbm log */
895static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
896 long long sector, const int length, __u32 *pos)
897{
898 __u32 i;
899
900 for (i = *pos; i < log->entry_count; i++) {
901 struct bbm_log_entry *entry = &log->marked_block_entries[i];
902 unsigned long long bb_start;
903 unsigned long long bb_end;
904
905 bb_start = __le48_to_cpu(&entry->defective_block_start);
906 bb_end = bb_start + (entry->marked_count + 1);
907
908 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
909 (bb_end <= sector + length)) {
910 *pos = i;
911 return 1;
912 }
913 }
914 return 0;
915}
916
917/* record new bad block in bbm log */
918static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
919 long long sector, int length)
920{
921 int new_bb = 0;
922 __u32 pos = 0;
923 struct bbm_log_entry *entry = NULL;
924
925 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
926 struct bbm_log_entry *e = &log->marked_block_entries[pos];
927
928 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
929 (__le48_to_cpu(&e->defective_block_start) == sector)) {
930 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
931 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
932 pos = pos + 1;
933 continue;
934 }
935 entry = e;
936 break;
937 }
938
939 if (entry) {
940 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
941 BBM_LOG_MAX_LBA_ENTRY_VAL;
942 entry->defective_block_start = __cpu_to_le48(sector);
943 entry->marked_count = cnt - 1;
944 if (cnt == length)
945 return 1;
946 sector += cnt;
947 length -= cnt;
948 }
949
950 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
951 BBM_LOG_MAX_LBA_ENTRY_VAL;
952 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
953 return 0;
954
955 while (length > 0) {
956 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
957 BBM_LOG_MAX_LBA_ENTRY_VAL;
958 struct bbm_log_entry *entry =
959 &log->marked_block_entries[log->entry_count];
960
961 entry->defective_block_start = __cpu_to_le48(sector);
962 entry->marked_count = cnt - 1;
963 entry->disk_ordinal = idx;
964
965 sector += cnt;
966 length -= cnt;
967
968 log->entry_count++;
969 }
970
971 return new_bb;
972}
c07a5a4f 973
4c9e8c1e
TM		 974/* clear all bad blocks for given disk */
975static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
976{
977 __u32 i = 0;
978
979 while (i < log->entry_count) {
980 struct bbm_log_entry *entries = log->marked_block_entries;
981
982 if (entries[i].disk_ordinal == idx) {
983 if (i < log->entry_count - 1)
984 entries[i] = entries[log->entry_count - 1];
985 log->entry_count--;
986 } else {
987 i++;
988 }
989 }
990}
991
c07a5a4f
TM		 992/* clear given bad block */
993static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
994 long long sector, const int length) {
995 __u32 i = 0;
996
997 while (i < log->entry_count) {
998 struct bbm_log_entry *entries = log->marked_block_entries;
999
1000 if ((entries[i].disk_ordinal == idx) &&
1001 (__le48_to_cpu(&entries[i].defective_block_start) ==
1002 sector) && (entries[i].marked_count + 1 == length)) {
1003 if (i < log->entry_count - 1)
1004 entries[i] = entries[log->entry_count - 1];
1005 log->entry_count--;
1006 break;
1007 }
1008 i++;
1009 }
1010
1011 return 1;
1012}
8d67477f
TM		 1013
1014/* allocate and load BBM log from metadata */
1015static int load_bbm_log(struct intel_super *super)
1016{
1017 struct imsm_super *mpb = super->anchor;
1018 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1019
1020 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1021 if (!super->bbm_log)
1022 return 1;
1023
1024 if (bbm_log_size) {
1025 struct bbm_log *log = (void *)mpb +
1026 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1027
1028 __u32 entry_count;
1029
1030 if (bbm_log_size < sizeof(log->signature) +
1031 sizeof(log->entry_count))
1032 return 2;
1033
1034 entry_count = __le32_to_cpu(log->entry_count);
1035 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1036 (entry_count > BBM_LOG_MAX_ENTRIES))
1037 return 3;
1038
1039 if (bbm_log_size !=
1040 sizeof(log->signature) + sizeof(log->entry_count) +
1041 entry_count * sizeof(struct bbm_log_entry))
1042 return 4;
1043
1044 memcpy(super->bbm_log, log, bbm_log_size);
1045 } else {
1046 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1047 super->bbm_log->entry_count = 0;
1048 }
1049
1050 return 0;
1051}
1052
b12796be
TM		 1053/* checks if bad block is within volume boundaries */
1054static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1055 const unsigned long long start_sector,
1056 const unsigned long long size)
1057{
1058 unsigned long long bb_start;
1059 unsigned long long bb_end;
1060
1061 bb_start = __le48_to_cpu(&entry->defective_block_start);
1062 bb_end = bb_start + (entry->marked_count + 1);
1063
1064 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1065 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1066 return 1;
1067
1068 return 0;
1069}
1070
1071/* get list of bad blocks on a drive for a volume */
1072static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1073 const unsigned long long start_sector,
1074 const unsigned long long size,
1075 struct md_bb *bbs)
1076{
1077 __u32 count = 0;
1078 __u32 i;
1079
1080 for (i = 0; i < log->entry_count; i++) {
1081 const struct bbm_log_entry *ent =
1082 &log->marked_block_entries[i];
1083 struct md_bb_entry *bb;
1084
1085 if ((ent->disk_ordinal == idx) &&
1086 is_bad_block_in_volume(ent, start_sector, size)) {
1087
1088 if (!bbs->entries) {
1089 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1090 sizeof(*bb));
1091 if (!bbs->entries)
1092 break;
1093 }
1094
1095 bb = &bbs->entries[count++];
1096 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1097 bb->length = ent->marked_count + 1;
1098 }
1099 }
1100 bbs->count = count;
1101}
1102
98130f40
AK		 1103/*
1104 * for second_map:
238c0a71
AK		 1105 * == MAP_0 get first map
1106 * == MAP_1 get second map
1107 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1108 */
1109static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1110 int slot,
1111 int second_map)
7eef0453
DW		 1112{
1113 struct imsm_map *map;
1114
5e7b0330 1115 map = get_imsm_map(dev, second_map);
7eef0453 1116
ff077194
DW		 1117 /* top byte identifies disk under rebuild */
1118 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1119}
1120
1121#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1122static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1123{
98130f40 1124 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1125
1126 return ord_to_idx(ord);
7eef0453
DW		 1127}
1128
be73972f
DW		 1129static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1130{
1131 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1132}
1133
f21e18ca 1134static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1135{
1136 int slot;
1137 __u32 ord;
1138
1139 for (slot = 0; slot < map->num_members; slot++) {
1140 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1141 if (ord_to_idx(ord) == idx)
1142 return slot;
1143 }
1144
1145 return -1;
1146}
1147
cdddbdbc
DW		 1148static int get_imsm_raid_level(struct imsm_map *map)
1149{
1150 if (map->raid_level == 1) {
1151 if (map->num_members == 2)
1152 return 1;
1153 else
1154 return 10;
1155 }
1156
1157 return map->raid_level;
1158}
1159
c2c087e6
DW		 1160static int cmp_extent(const void *av, const void *bv)
1161{
1162 const struct extent *a = av;
1163 const struct extent *b = bv;
1164 if (a->start < b->start)
1165 return -1;
1166 if (a->start > b->start)
1167 return 1;
1168 return 0;
1169}
1170
0dcecb2e 1171static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1172{
c2c087e6 1173 int memberships = 0;
620b1713 1174 int i;
c2c087e6 1175
949c47a0
DW		 1176 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1177 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1178 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1179
620b1713
DW		 1180 if (get_imsm_disk_slot(map, dl->index) >= 0)
1181 memberships++;
c2c087e6 1182 }
0dcecb2e
DW		 1183
1184 return memberships;
1185}
1186
b81221b7
CA		 1187static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1188
486720e0 1189static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1190{
1191 if (lo == 0 || hi == 0)
1192 return 1;
486720e0
JS		 1193 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1194 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1195 return 0;
1196}
1197
1198static unsigned long long join_u32(__u32 lo, __u32 hi)
1199{
1200 return (unsigned long long)__le32_to_cpu(lo) |
1201 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1202}
1203
1204static unsigned long long total_blocks(struct imsm_disk *disk)
1205{
1206 if (disk == NULL)
1207 return 0;
1208 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1209}
1210
1211static unsigned long long pba_of_lba0(struct imsm_map *map)
1212{
1213 if (map == NULL)
1214 return 0;
1215 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1216}
1217
1218static unsigned long long blocks_per_member(struct imsm_map *map)
1219{
1220 if (map == NULL)
1221 return 0;
1222 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1223}
1224
1225static unsigned long long num_data_stripes(struct imsm_map *map)
1226{
1227 if (map == NULL)
1228 return 0;
1229 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1230}
1231
fcc2c9da
MD		 1232static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1233{
1234 if (dev == NULL)
1235 return 0;
1236 return join_u32(dev->size_low, dev->size_high);
1237}
1238
9f421827
PB		 1239static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1240{
1241 if (migr_rec == NULL)
1242 return 0;
1243 return join_u32(migr_rec->ckpt_area_pba_lo,
1244 migr_rec->ckpt_area_pba_hi);
1245}
1246
1247static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1248{
1249 if (migr_rec == NULL)
1250 return 0;
1251 return join_u32(migr_rec->curr_migr_unit_lo,
1252 migr_rec->curr_migr_unit_hi);
1253}
1254
1255static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1256{
1257 if (migr_rec == NULL)
1258 return 0;
1259 return join_u32(migr_rec->dest_1st_member_lba_lo,
1260 migr_rec->dest_1st_member_lba_hi);
1261}
1262
1263static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1264{
1265 if (migr_rec == NULL)
1266 return 0;
1267 return join_u32(migr_rec->num_migr_units_lo,
1268 migr_rec->num_migr_units_hi);
1269}
1270
5551b113
CA		 1271static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1272{
1273 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1274}
1275
1276static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1277{
1278 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1279}
1280
1281static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1282{
1283 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1284}
1285
1286static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1287{
1288 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1289}
1290
fcc2c9da
MD		 1291static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1292{
1293 split_ull(n, &dev->size_low, &dev->size_high);
1294}
1295
9f421827
PB		 1296static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1297 unsigned long long n)
1298{
1299 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1300}
1301
1302static void set_current_migr_unit(struct migr_record *migr_rec,
1303 unsigned long long n)
1304{
1305 split_ull(n, &migr_rec->curr_migr_unit_lo,
1306 &migr_rec->curr_migr_unit_hi);
1307}
1308
1309static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1310 unsigned long long n)
1311{
1312 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1313 &migr_rec->dest_1st_member_lba_hi);
1314}
1315
1316static void set_num_migr_units(struct migr_record *migr_rec,
1317 unsigned long long n)
1318{
1319 split_ull(n, &migr_rec->num_migr_units_lo,
1320 &migr_rec->num_migr_units_hi);
1321}
1322
44490938
MD		 1323static unsigned long long per_dev_array_size(struct imsm_map *map)
1324{
1325 unsigned long long array_size = 0;
1326
1327 if (map == NULL)
1328 return array_size;
1329
1330 array_size = num_data_stripes(map) * map->blocks_per_strip;
1331 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1332 array_size *= 2;
1333
1334 return array_size;
1335}
1336
05501181
PB		 1337static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1338 int get_minimal_reservation)
0dcecb2e
DW		 1339{
1340 /* find a list of used extents on the given physical device */
1341 struct extent *rv, *e;
620b1713 1342 int i;
0dcecb2e 1343 int memberships = count_memberships(dl, super);
b276dd33
DW		 1344 __u32 reservation;
1345
1346 /* trim the reserved area for spares, so they can join any array
1347 * regardless of whether the OROM has assigned sectors from the
1348 * IMSM_RESERVED_SECTORS region
1349 */
05501181 1350 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1351 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1352 else
1353 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1354
503975b9 1355 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1356 e = rv;
1357
949c47a0
DW		 1358 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1359 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1360 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1361
620b1713 1362 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1363 e->start = pba_of_lba0(map);
44490938 1364 e->size = per_dev_array_size(map);
620b1713 1365 e++;
c2c087e6
DW		 1366 }
1367 }
1368 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1369
1011e834 1370 /* determine the start of the metadata
14e8215b
DW		 1371 * when no raid devices are defined use the default
1372 * ...otherwise allow the metadata to truncate the value
1373 * as is the case with older versions of imsm
1374 */
1375 if (memberships) {
1376 struct extent *last = &rv[memberships - 1];
5551b113 1377 unsigned long long remainder;
14e8215b 1378
5551b113 1379 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1380 /* round down to 1k block to satisfy precision of the kernel
1381 * 'size' interface
1382 */
1383 remainder &= ~1UL;
1384 /* make sure remainder is still sane */
f21e18ca 1385 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1386 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1387 if (reservation > remainder)
1388 reservation = remainder;
1389 }
5551b113 1390 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1391 e->size = 0;
1392 return rv;
1393}
1394
14e8215b
DW		 1395/* try to determine how much space is reserved for metadata from
1396 * the last get_extents() entry, otherwise fallback to the
1397 * default
1398 */
1399static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1400{
1401 struct extent *e;
1402 int i;
1403 __u32 rv;
1404
1405 /* for spares just return a minimal reservation which will grow
1406 * once the spare is picked up by an array
1407 */
1408 if (dl->index == -1)
1409 return MPB_SECTOR_CNT;
1410
05501181 1411 e = get_extents(super, dl, 0);
14e8215b
DW		 1412 if (!e)
1413 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1414
1415 /* scroll to last entry */
1416 for (i = 0; e[i].size; i++)
1417 continue;
1418
5551b113 1419 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1420
1421 free(e);
1422
1423 return rv;
1424}
1425
25ed7e59
DW		 1426static int is_spare(struct imsm_disk *disk)
1427{
1428 return (disk->status & SPARE_DISK) == SPARE_DISK;
1429}
1430
1431static int is_configured(struct imsm_disk *disk)
1432{
1433 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1434}
1435
1436static int is_failed(struct imsm_disk *disk)
1437{
1438 return (disk->status & FAILED_DISK) == FAILED_DISK;
1439}
1440
2432ce9b
AP		 1441static int is_journal(struct imsm_disk *disk)
1442{
1443 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1444}
1445
b53bfba6
TM		 1446/* round array size down to closest MB and ensure it splits evenly
1447 * between members
1448 */
1449static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1450 disk_count)
1451{
1452 size /= disk_count;
1453 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1454 size *= disk_count;
1455
1456 return size;
1457}
1458
8b9cd157
MK		 1459static int able_to_resync(int raid_level, int missing_disks)
1460{
1461 int max_missing_disks = 0;
1462
1463 switch (raid_level) {
1464 case 10:
1465 max_missing_disks = 1;
1466 break;
1467 default:
1468 max_missing_disks = 0;
1469 }
1470 return missing_disks <= max_missing_disks;
1471}
1472
b81221b7
CA		 1473/* try to determine how much space is reserved for metadata from
1474 * the last get_extents() entry on the smallest active disk,
1475 * otherwise fallback to the default
1476 */
1477static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1478{
1479 struct extent *e;
1480 int i;
5551b113
CA		 1481 unsigned long long min_active;
1482 __u32 remainder;
b81221b7
CA		 1483 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1484 struct dl *dl, *dl_min = NULL;
1485
1486 if (!super)
1487 return rv;
1488
1489 min_active = 0;
1490 for (dl = super->disks; dl; dl = dl->next) {
1491 if (dl->index < 0)
1492 continue;
5551b113
CA		 1493 unsigned long long blocks = total_blocks(&dl->disk);
1494 if (blocks < min_active || min_active == 0) {
b81221b7 1495 dl_min = dl;
5551b113 1496 min_active = blocks;
b81221b7
CA		 1497 }
1498 }
1499 if (!dl_min)
1500 return rv;
1501
1502 /* find last lba used by subarrays on the smallest active disk */
05501181 1503 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1504 if (!e)
1505 return rv;
1506 for (i = 0; e[i].size; i++)
1507 continue;
1508
1509 remainder = min_active - e[i].start;
1510 free(e);
1511
1512 /* to give priority to recovery we should not require full
1513 IMSM_RESERVED_SECTORS from the spare */
1514 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1515
1516 /* if real reservation is smaller use that value */
1517 return (remainder < rv) ? remainder : rv;
1518}
1519
fbfdcb06
AO		 1520/*
1521 * Return minimum size of a spare and sector size
1522 * that can be used in this array
1523 */
1524int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1525{
1526 struct intel_super *super = st->sb;
1527 struct dl *dl;
1528 struct extent *e;
1529 int i;
fbfdcb06
AO		 1530 unsigned long long size = 0;
1531
1532 c->min_size = 0;
4b57ecf6 1533 c->sector_size = 0;
80e7f8c3
AC		 1534
1535 if (!super)
fbfdcb06 1536 return -EINVAL;
80e7f8c3
AC		 1537 /* find first active disk in array */
1538 dl = super->disks;
1539 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1540 dl = dl->next;
1541 if (!dl)
fbfdcb06 1542 return -EINVAL;
80e7f8c3 1543 /* find last lba used by subarrays */
05501181 1544 e = get_extents(super, dl, 0);
80e7f8c3 1545 if (!e)
fbfdcb06 1546 return -EINVAL;
80e7f8c3
AC		 1547 for (i = 0; e[i].size; i++)
1548 continue;
1549 if (i > 0)
fbfdcb06 1550 size = e[i-1].start + e[i-1].size;
80e7f8c3 1551 free(e);
b81221b7 1552
80e7f8c3 1553 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1554 size += imsm_min_reserved_sectors(super);
1555
1556 c->min_size = size * 512;
4b57ecf6 1557 c->sector_size = super->sector_size;
b81221b7 1558
fbfdcb06 1559 return 0;
80e7f8c3
AC		 1560}
1561
d1e02575
AK		 1562static int is_gen_migration(struct imsm_dev *dev);
1563
f36a9ecd
PB		 1564#define IMSM_4K_DIV 8
1565
c47b0ff6
AK		 1566static __u64 blocks_per_migr_unit(struct intel_super *super,
1567 struct imsm_dev *dev);
1e5c6983 1568
c47b0ff6
AK		 1569static void print_imsm_dev(struct intel_super *super,
1570 struct imsm_dev *dev,
1571 char *uuid,
1572 int disk_idx)
cdddbdbc
DW		 1573{
1574 __u64 sz;
0d80bb2f 1575 int slot, i;
238c0a71
AK		 1576 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1577 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1578 __u32 ord;
cdddbdbc
DW		 1579
1580 printf("\n");
1e7bc0ed 1581 printf("[%.16s]:\n", dev->volume);
44470971 1582 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1583 printf(" RAID Level : %d", get_imsm_raid_level(map));
1584 if (map2)
1585 printf(" <-- %d", get_imsm_raid_level(map2));
1586 printf("\n");
1587 printf(" Members : %d", map->num_members);
1588 if (map2)
1589 printf(" <-- %d", map2->num_members);
1590 printf("\n");
0d80bb2f
DW		 1591 printf(" Slots : [");
1592 for (i = 0; i < map->num_members; i++) {
238c0a71 1593 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1594 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1595 }
dd8bcb3b
AK		 1596 printf("]");
1597 if (map2) {
1598 printf(" <-- [");
1599 for (i = 0; i < map2->num_members; i++) {
238c0a71 1600 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1601 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1602 }
1603 printf("]");
1604 }
1605 printf("\n");
7095bccb
AK		 1606 printf(" Failed disk : ");
1607 if (map->failed_disk_num == 0xff)
1608 printf("none");
1609 else
1610 printf("%i", map->failed_disk_num);
1611 printf("\n");
620b1713
DW		 1612 slot = get_imsm_disk_slot(map, disk_idx);
1613 if (slot >= 0) {
238c0a71 1614 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1615 printf(" This Slot : %d%s\n", slot,
1616 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1617 } else
cdddbdbc 1618 printf(" This Slot : ?\n");
84918897 1619 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1620 sz = imsm_dev_size(dev);
84918897
MK		 1621 printf(" Array Size : %llu%s\n",
1622 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1623 human_size(sz * 512));
5551b113 1624 sz = blocks_per_member(map);
84918897
MK		 1625 printf(" Per Dev Size : %llu%s\n",
1626 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1627 human_size(sz * 512));
5551b113
CA		 1628 printf(" Sector Offset : %llu\n",
1629 pba_of_lba0(map));
1630 printf(" Num Stripes : %llu\n",
1631 num_data_stripes(map));
dd8bcb3b 1632 printf(" Chunk Size : %u KiB",
cdddbdbc 1633 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1634 if (map2)
1635 printf(" <-- %u KiB",
1636 __le16_to_cpu(map2->blocks_per_strip) / 2);
1637 printf("\n");
cdddbdbc 1638 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1639 printf(" Migrate State : ");
1484e727
DW		 1640 if (dev->vol.migr_state) {
1641 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1642 printf("initialize\n");
1484e727 1643 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1644 printf("rebuild\n");
1484e727 1645 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1646 printf("check\n");
1484e727 1647 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1648 printf("general migration\n");
1484e727 1649 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1650 printf("state change\n");
1484e727 1651 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1652 printf("repair\n");
1484e727 1653 else
8655a7b1
DW		 1654 printf("<unknown:%d>\n", migr_type(dev));
1655 } else
1656 printf("idle\n");
3393c6af
DW		 1657 printf(" Map State : %s", map_state_str[map->map_state]);
1658 if (dev->vol.migr_state) {
238c0a71 1659 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1660
b10b37b8 1661 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1662 printf("\n Checkpoint : %u ",
1663 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1664 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1665 printf("(N/A)");
1666 else
1667 printf("(%llu)", (unsigned long long)
1668 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1669 }
1670 printf("\n");
2432ce9b
AP		 1671 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1672 "dirty" : "clean");
1673 printf(" RWH Policy : ");
c2462068 1674 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1675 printf("off\n");
1676 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1677 printf("PPL distributed\n");
1678 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1679 printf("PPL journaling drive\n");
c2462068
PB		 1680 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1681 printf("Multiple distributed PPLs\n");
1682 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1683 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1684 else
1685 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1686}
1687
ef5c214e
MK		 1688static void print_imsm_disk(struct imsm_disk *disk,
1689 int index,
1690 __u32 reserved,
1691 unsigned int sector_size) {
1f24f035 1692 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1693 __u64 sz;
1694
0ec1f4e8 1695 if (index < -1 || !disk)
e9d82038
DW		 1696 return;
1697
cdddbdbc 1698 printf("\n");
1f24f035 1699 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1700 if (index >= 0)
1701 printf(" Disk%02d Serial : %s\n", index, str);
1702 else
1703 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1704 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1705 is_configured(disk) ? " active" : "",
1706 is_failed(disk) ? " failed" : "",
1707 is_journal(disk) ? " journal" : "");
cdddbdbc 1708 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1709 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1710 printf(" Usable Size : %llu%s\n",
1711 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1712 human_size(sz * 512));
1713}
1714
de44e46f
PB		 1715void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1716{
1717 struct migr_record *migr_rec = super->migr_rec;
1718
1719 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1720 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1721 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1722 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1723 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1724 set_migr_chkp_area_pba(migr_rec,
1725 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1726 set_migr_dest_1st_member_lba(migr_rec,
1727 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1728}
1729
f36a9ecd
PB		 1730void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1731{
1732 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1733}
1734
1735void convert_to_4k(struct intel_super *super)
1736{
1737 struct imsm_super *mpb = super->anchor;
1738 struct imsm_disk *disk;
1739 int i;
e4467bc7 1740 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1741
1742 for (i = 0; i < mpb->num_disks ; i++) {
1743 disk = __get_imsm_disk(mpb, i);
1744 /* disk */
1745 convert_to_4k_imsm_disk(disk);
1746 }
1747 for (i = 0; i < mpb->num_raid_devs; i++) {
1748 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1749 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1750 /* dev */
fcc2c9da 1751 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1752 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1753
1754 /* map0 */
1755 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1756 map->blocks_per_strip /= IMSM_4K_DIV;
1757 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1758
1759 if (dev->vol.migr_state) {
1760 /* map1 */
1761 map = get_imsm_map(dev, MAP_1);
1762 set_blocks_per_member(map,
1763 blocks_per_member(map)/IMSM_4K_DIV);
1764 map->blocks_per_strip /= IMSM_4K_DIV;
1765 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1766 }
1767 }
e4467bc7
TM		 1768 if (bbm_log_size) {
1769 struct bbm_log *log = (void *)mpb +
1770 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1771 __u32 i;
1772
1773 for (i = 0; i < log->entry_count; i++) {
1774 struct bbm_log_entry *entry =
1775 &log->marked_block_entries[i];
1776
1777 __u8 count = entry->marked_count + 1;
1778 unsigned long long sector =
1779 __le48_to_cpu(&entry->defective_block_start);
1780
1781 entry->defective_block_start =
1782 __cpu_to_le48(sector/IMSM_4K_DIV);
1783 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1784 }
1785 }
f36a9ecd
PB		 1786
1787 mpb->check_sum = __gen_imsm_checksum(mpb);
1788}
1789
520e69e2
AK		 1790void examine_migr_rec_imsm(struct intel_super *super)
1791{
1792 struct migr_record *migr_rec = super->migr_rec;
1793 struct imsm_super *mpb = super->anchor;
1794 int i;
1795
1796 for (i = 0; i < mpb->num_raid_devs; i++) {
1797 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1798 struct imsm_map *map;
b4ab44d8 1799 int slot = -1;
3136abe5 1800
520e69e2
AK		 1801 if (is_gen_migration(dev) == 0)
1802 continue;
1803
1804 printf("\nMigration Record Information:");
3136abe5 1805
44bfe6df
AK		 1806 /* first map under migration */
1807 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1808 if (map)
1809 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1810 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1811 printf(" Empty\n ");
1812 printf("Examine one of first two disks in array\n");
1813 break;
1814 }
1815 printf("\n Status : ");
1816 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1817 printf("Normal\n");
1818 else
1819 printf("Contains Data\n");
9f421827
PB		 1820 printf(" Current Unit : %llu\n",
1821 current_migr_unit(migr_rec));
520e69e2
AK		 1822 printf(" Family : %u\n",
1823 __le32_to_cpu(migr_rec->family_num));
1824 printf(" Ascending : %u\n",
1825 __le32_to_cpu(migr_rec->ascending_migr));
1826 printf(" Blocks Per Unit : %u\n",
1827 __le32_to_cpu(migr_rec->blocks_per_unit));
1828 printf(" Dest. Depth Per Unit : %u\n",
1829 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1830 printf(" Checkpoint Area pba : %llu\n",
1831 migr_chkp_area_pba(migr_rec));
1832 printf(" First member lba : %llu\n",
1833 migr_dest_1st_member_lba(migr_rec));
1834 printf(" Total Number of Units : %llu\n",
1835 get_num_migr_units(migr_rec));
1836 printf(" Size of volume : %llu\n",
1837 join_u32(migr_rec->post_migr_vol_cap,
1838 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1839 printf(" Record was read from : %u\n",
1840 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1841
1842 break;
1843 }
1844}
f36a9ecd 1845
de44e46f
PB		 1846void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1847{
1848 struct migr_record *migr_rec = super->migr_rec;
1849
1850 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1851 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1852 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1853 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1854 &migr_rec->post_migr_vol_cap,
1855 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1856 set_migr_chkp_area_pba(migr_rec,
1857 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1858 set_migr_dest_1st_member_lba(migr_rec,
1859 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1860}
1861
f36a9ecd
PB		 1862void convert_from_4k(struct intel_super *super)
1863{
1864 struct imsm_super *mpb = super->anchor;
1865 struct imsm_disk *disk;
1866 int i;
e4467bc7 1867 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1868
1869 for (i = 0; i < mpb->num_disks ; i++) {
1870 disk = __get_imsm_disk(mpb, i);
1871 /* disk */
1872 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1873 }
1874
1875 for (i = 0; i < mpb->num_raid_devs; i++) {
1876 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1877 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1878 /* dev */
fcc2c9da 1879 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1880 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1881
1882 /* map0 */
1883 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1884 map->blocks_per_strip *= IMSM_4K_DIV;
1885 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1886
1887 if (dev->vol.migr_state) {
1888 /* map1 */
1889 map = get_imsm_map(dev, MAP_1);
1890 set_blocks_per_member(map,
1891 blocks_per_member(map)*IMSM_4K_DIV);
1892 map->blocks_per_strip *= IMSM_4K_DIV;
1893 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1894 }
1895 }
e4467bc7
TM		 1896 if (bbm_log_size) {
1897 struct bbm_log *log = (void *)mpb +
1898 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1899 __u32 i;
1900
1901 for (i = 0; i < log->entry_count; i++) {
1902 struct bbm_log_entry *entry =
1903 &log->marked_block_entries[i];
1904
1905 __u8 count = entry->marked_count + 1;
1906 unsigned long long sector =
1907 __le48_to_cpu(&entry->defective_block_start);
1908
1909 entry->defective_block_start =
1910 __cpu_to_le48(sector*IMSM_4K_DIV);
1911 entry->marked_count = count*IMSM_4K_DIV - 1;
1912 }
1913 }
f36a9ecd
PB		 1914
1915 mpb->check_sum = __gen_imsm_checksum(mpb);
1916}
1917
19482bcc
AK		 1918/*******************************************************************************
1919 * function: imsm_check_attributes
1920 * Description: Function checks if features represented by attributes flags
1011e834 1921 * are supported by mdadm.
19482bcc
AK		 1922 * Parameters:
1923 * attributes - Attributes read from metadata
1924 * Returns:
1011e834
N		 1925 * 0 - passed attributes contains unsupported features flags
1926 * 1 - all features are supported
19482bcc
AK		 1927 ******************************************************************************/
1928static int imsm_check_attributes(__u32 attributes)
1929{
1930 int ret_val = 1;
418f9b36
N		 1931 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1932
1933 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1934
1935 not_supported &= attributes;
1936 if (not_supported) {
e7b84f9d 1937 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1938 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1939 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1940 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1941 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1942 }
1943 if (not_supported & MPB_ATTRIB_2TB) {
1944 dprintf("\t\tMPB_ATTRIB_2TB\n");
1945 not_supported ^= MPB_ATTRIB_2TB;
1946 }
1947 if (not_supported & MPB_ATTRIB_RAID0) {
1948 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1949 not_supported ^= MPB_ATTRIB_RAID0;
1950 }
1951 if (not_supported & MPB_ATTRIB_RAID1) {
1952 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1953 not_supported ^= MPB_ATTRIB_RAID1;
1954 }
1955 if (not_supported & MPB_ATTRIB_RAID10) {
1956 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1957 not_supported ^= MPB_ATTRIB_RAID10;
1958 }
1959 if (not_supported & MPB_ATTRIB_RAID1E) {
1960 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1961 not_supported ^= MPB_ATTRIB_RAID1E;
1962 }
1963 if (not_supported & MPB_ATTRIB_RAID5) {
1964 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1965 not_supported ^= MPB_ATTRIB_RAID5;
1966 }
1967 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1968 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1969 not_supported ^= MPB_ATTRIB_RAIDCNG;
1970 }
1971 if (not_supported & MPB_ATTRIB_BBM) {
1972 dprintf("\t\tMPB_ATTRIB_BBM\n");
1973 not_supported ^= MPB_ATTRIB_BBM;
1974 }
1975 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1976 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1977 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1978 }
1979 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1980 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1981 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1982 }
1983 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1984 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1985 not_supported ^= MPB_ATTRIB_2TB_DISK;
1986 }
1987 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1988 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1989 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1990 }
1991 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1992 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1993 not_supported ^= MPB_ATTRIB_NEVER_USE;
1994 }
1995
1996 if (not_supported)
1ade5cc1 1997 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1998
1999 ret_val = 0;
2000 }
2001
2002 return ret_val;
2003}
2004
a5d85af7 2005static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2006
cdddbdbc
DW		 2007static void examine_super_imsm(struct supertype *st, char *homehost)
2008{
2009 struct intel_super *super = st->sb;
949c47a0 2010 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2011 char str[MAX_SIGNATURE_LENGTH];
2012 int i;
27fd6274
DW		 2013 struct mdinfo info;
2014 char nbuf[64];
cdddbdbc 2015 __u32 sum;
14e8215b 2016 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2017 struct dl *dl;
e48aed3c 2018 time_t creation_time;
27fd6274 2019
618f4e6d
XN		 2020 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2021 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2022 printf(" Magic : %s\n", str);
cdddbdbc 2023 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2024 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2025 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2026 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
e48aed3c
AP		 2027 creation_time = __le64_to_cpu(mpb->creation_time);
2028 printf(" Creation Time : %.24s\n",
2029 creation_time ? ctime(&creation_time) : "Unknown");
19482bcc
AK		 2030 printf(" Attributes : ");
2031 if (imsm_check_attributes(mpb->attributes))
2032 printf("All supported\n");
2033 else
2034 printf("not supported\n");
a5d85af7 2035 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2036 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2037 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2038 sum = __le32_to_cpu(mpb->check_sum);
2039 printf(" Checksum : %08x %s\n", sum,
949c47a0 2040 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2041 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2042 printf(" Disks : %d\n", mpb->num_disks);
2043 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2044 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2045 super->disks->index, reserved, super->sector_size);
8d67477f 2046 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2047 struct bbm_log *log = super->bbm_log;
2048
2049 printf("\n");
2050 printf("Bad Block Management Log:\n");
2051 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2052 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2053 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2054 }
44470971
DW		 2055 for (i = 0; i < mpb->num_raid_devs; i++) {
2056 struct mdinfo info;
2057 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2058
2059 super->current_vol = i;
a5d85af7 2060 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2061 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2062 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2063 }
cdddbdbc
DW		 2064 for (i = 0; i < mpb->num_disks; i++) {
2065 if (i == super->disks->index)
2066 continue;
ef5c214e
MK		 2067 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2068 super->sector_size);
cdddbdbc 2069 }
94827db3 2070
0ec1f4e8
DW		 2071 for (dl = super->disks; dl; dl = dl->next)
2072 if (dl->index == -1)
ef5c214e
MK		 2073 print_imsm_disk(&dl->disk, -1, reserved,
2074 super->sector_size);
520e69e2
AK		 2075
2076 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2077}
2078
061f2c6a 2079static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2080{
27fd6274 2081 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2082 struct mdinfo info;
2083 char nbuf[64];
1e7bc0ed 2084 struct intel_super *super = st->sb;
1e7bc0ed 2085
0d5a423f
DW		 2086 if (!super->anchor->num_raid_devs) {
2087 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2088 return;
0d5a423f 2089 }
ff54de6e 2090
a5d85af7 2091 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2092 fname_from_uuid(st, &info, nbuf, ':');
2093 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2094}
2095
2096static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2097{
2098 /* We just write a generic IMSM ARRAY entry */
2099 struct mdinfo info;
2100 char nbuf[64];
2101 char nbuf1[64];
2102 struct intel_super *super = st->sb;
2103 int i;
2104
2105 if (!super->anchor->num_raid_devs)
2106 return;
2107
a5d85af7 2108 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2109 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2110 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2111 struct imsm_dev *dev = get_imsm_dev(super, i);
2112
2113 super->current_vol = i;
a5d85af7 2114 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2115 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2116 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2117 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2118 }
cdddbdbc
DW		 2119}
2120
9d84c8ea
DW		 2121static void export_examine_super_imsm(struct supertype *st)
2122{
2123 struct intel_super *super = st->sb;
2124 struct imsm_super *mpb = super->anchor;
2125 struct mdinfo info;
2126 char nbuf[64];
2127
a5d85af7 2128 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2129 fname_from_uuid(st, &info, nbuf, ':');
2130 printf("MD_METADATA=imsm\n");
2131 printf("MD_LEVEL=container\n");
2132 printf("MD_UUID=%s\n", nbuf+5);
2133 printf("MD_DEVICES=%u\n", mpb->num_disks);
e48aed3c 2134 printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time));
9d84c8ea
DW		 2135}
2136
b771faef
BK		 2137static void detail_super_imsm(struct supertype *st, char *homehost,
2138 char *subarray)
cdddbdbc 2139{
3ebe00a1
DW		 2140 struct mdinfo info;
2141 char nbuf[64];
b771faef
BK		 2142 struct intel_super *super = st->sb;
2143 int temp_vol = super->current_vol;
2144
2145 if (subarray)
2146 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2147
a5d85af7 2148 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2149 fname_from_uuid(st, &info, nbuf, ':');
65884368 2150 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2151
2152 super->current_vol = temp_vol;
cdddbdbc
DW		 2153}
2154
b771faef 2155static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2156{
ff54de6e
N		 2157 struct mdinfo info;
2158 char nbuf[64];
b771faef
BK		 2159 struct intel_super *super = st->sb;
2160 int temp_vol = super->current_vol;
2161
2162 if (subarray)
2163 super->current_vol = strtoul(subarray, NULL, 10);
2164
a5d85af7 2165 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2166 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2167 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2168
2169 super->current_vol = temp_vol;
cdddbdbc 2170}
d665cc31 2171
6da53c0e
BK		 2172static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2173 size_t serial_buf_len);
d665cc31
DW		 2174static void fd2devname(int fd, char *name);
2175
120dc887 2176static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2177{
120dc887
LM		 2178 /* dump an unsorted list of devices attached to AHCI Intel storage
2179 * controller, as well as non-connected ports
d665cc31
DW		 2180 */
2181 int hba_len = strlen(hba_path) + 1;
2182 struct dirent *ent;
2183 DIR *dir;
2184 char *path = NULL;
2185 int err = 0;
2186 unsigned long port_mask = (1 << port_count) - 1;
2187
f21e18ca 2188 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2189 if (verbose > 0)
e7b84f9d 2190 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2191 return 2;
2192 }
2193
2194 /* scroll through /sys/dev/block looking for devices attached to
2195 * this hba
2196 */
2197 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2198 if (!dir)
2199 return 1;
2200
2201 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2202 int fd;
2203 char model[64];
2204 char vendor[64];
2205 char buf[1024];
2206 int major, minor;
2207 char *device;
2208 char *c;
2209 int port;
2210 int type;
2211
2212 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2213 continue;
2214 path = devt_to_devpath(makedev(major, minor));
2215 if (!path)
2216 continue;
2217 if (!path_attached_to_hba(path, hba_path)) {
2218 free(path);
2219 path = NULL;
2220 continue;
2221 }
2222
2223 /* retrieve the scsi device type */
2224 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2225 if (verbose > 0)
e7b84f9d 2226 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2227 err = 2;
2228 break;
2229 }
2230 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2231 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2232 if (verbose > 0)
e7b84f9d 2233 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2234 path);
2235 err = 2;
2236 free(device);
2237 break;
2238 }
2239 type = strtoul(buf, NULL, 10);
2240
2241 /* if it's not a disk print the vendor and model */
2242 if (!(type == 0 || type == 7 || type == 14)) {
2243 vendor[0] = '\0';
2244 model[0] = '\0';
2245 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2246 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2247 strncpy(vendor, buf, sizeof(vendor));
2248 vendor[sizeof(vendor) - 1] = '\0';
2249 c = (char *) &vendor[sizeof(vendor) - 1];
2250 while (isspace(*c) || *c == '\0')
2251 *c-- = '\0';
2252
2253 }
2254 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2255 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2256 strncpy(model, buf, sizeof(model));
2257 model[sizeof(model) - 1] = '\0';
2258 c = (char *) &model[sizeof(model) - 1];
2259 while (isspace(*c) || *c == '\0')
2260 *c-- = '\0';
2261 }
2262
2263 if (vendor[0] && model[0])
2264 sprintf(buf, "%.64s %.64s", vendor, model);
2265 else
2266 switch (type) { /* numbers from hald/linux/device.c */
2267 case 1: sprintf(buf, "tape"); break;
2268 case 2: sprintf(buf, "printer"); break;
2269 case 3: sprintf(buf, "processor"); break;
2270 case 4:
2271 case 5: sprintf(buf, "cdrom"); break;
2272 case 6: sprintf(buf, "scanner"); break;
2273 case 8: sprintf(buf, "media_changer"); break;
2274 case 9: sprintf(buf, "comm"); break;
2275 case 12: sprintf(buf, "raid"); break;
2276 default: sprintf(buf, "unknown");
2277 }
2278 } else
2279 buf[0] = '\0';
2280 free(device);
2281
2282 /* chop device path to 'host%d' and calculate the port number */
2283 c = strchr(&path[hba_len], '/');
4e5e717d 2284 if (!c) {
ba728be7 2285 if (verbose > 0)
e7b84f9d 2286 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2287 err = 2;
2288 break;
2289 }
d665cc31 2290 *c = '\0';
0858eccf
AP		 2291 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2292 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2293 port -= host_base;
2294 else {
ba728be7 2295 if (verbose > 0) {
d665cc31 2296 *c = '/'; /* repair the full string */
e7b84f9d 2297 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2298 path);
2299 }
2300 err = 2;
2301 break;
2302 }
2303
2304 /* mark this port as used */
2305 port_mask &= ~(1 << port);
2306
2307 /* print out the device information */
2308 if (buf[0]) {
2309 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2310 continue;
2311 }
2312
2313 fd = dev_open(ent->d_name, O_RDONLY);
2314 if (fd < 0)
2315 printf(" Port%d : - disk info unavailable -\n", port);
2316 else {
2317 fd2devname(fd, buf);
2318 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2319 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2320 sizeof(buf)) == 0)
2321 printf(" (%s)\n", buf);
d665cc31 2322 else
664d5325 2323 printf(" ()\n");
4dab422a 2324 close(fd);
d665cc31 2325 }
d665cc31
DW		 2326 free(path);
2327 path = NULL;
2328 }
2329 if (path)
2330 free(path);
2331 if (dir)
2332 closedir(dir);
2333 if (err == 0) {
2334 int i;
2335
2336 for (i = 0; i < port_count; i++)
2337 if (port_mask & (1 << i))
2338 printf(" Port%d : - no device attached -\n", i);
2339 }
2340
2341 return err;
2342}
2343
6da53c0e 2344static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2345{
6da53c0e 2346 char buf[1024];
60f0f54d
PB		 2347 struct dirent *ent;
2348 DIR *dir;
6da53c0e
BK		 2349 char *rp;
2350 int fd;
60f0f54d 2351
6da53c0e 2352 dir = opendir("/sys/block/");
b9135011 2353 if (!dir)
b5eece69 2354 return 1;
b9135011
JS		 2355
2356 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e
BK		 2357 if (strstr(ent->d_name, "nvme")) {
2358 sprintf(buf, "/sys/block/%s", ent->d_name);
2359 rp = realpath(buf, NULL);
2360 if (!rp)
2361 continue;
2362 if (path_attached_to_hba(rp, hba->path)) {
2363 fd = open_dev(ent->d_name);
2364 if (fd < 0) {
2365 free(rp);
2366 continue;
2367 }
60f0f54d 2368
6da53c0e
BK		 2369 fd2devname(fd, buf);
2370 if (hba->type == SYS_DEV_VMD)
2371 printf(" NVMe under VMD : %s", buf);
2372 else if (hba->type == SYS_DEV_NVME)
2373 printf(" NVMe Device : %s", buf);
2374 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2375 sizeof(buf)))
2376 printf(" (%s)\n", buf);
2377 else
2378 printf("()\n");
2379 close(fd);
2380 }
2381 free(rp);
60f0f54d 2382 }
60f0f54d
PB		 2383 }
2384
b9135011 2385 closedir(dir);
b5eece69 2386 return 0;
60f0f54d
PB		 2387}
2388
120dc887
LM		 2389static void print_found_intel_controllers(struct sys_dev *elem)
2390{
2391 for (; elem; elem = elem->next) {
e7b84f9d 2392 pr_err("found Intel(R) ");
120dc887
LM		 2393 if (elem->type == SYS_DEV_SATA)
2394 fprintf(stderr, "SATA ");
155cbb4c
LM		 2395 else if (elem->type == SYS_DEV_SAS)
2396 fprintf(stderr, "SAS ");
0858eccf
AP		 2397 else if (elem->type == SYS_DEV_NVME)
2398 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2399
2400 if (elem->type == SYS_DEV_VMD)
2401 fprintf(stderr, "VMD domain");
2402 else
2403 fprintf(stderr, "RAID controller");
2404
120dc887
LM		 2405 if (elem->pci_id)
2406 fprintf(stderr, " at %s", elem->pci_id);
2407 fprintf(stderr, ".\n");
2408 }
2409 fflush(stderr);
2410}
2411
120dc887
LM		 2412static int ahci_get_port_count(const char *hba_path, int *port_count)
2413{
2414 struct dirent *ent;
2415 DIR *dir;
2416 int host_base = -1;
2417
2418 *port_count = 0;
2419 if ((dir = opendir(hba_path)) == NULL)
2420 return -1;
2421
2422 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2423 int host;
2424
0858eccf
AP		 2425 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2426 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2427 continue;
2428 if (*port_count == 0)
2429 host_base = host;
2430 else if (host < host_base)
2431 host_base = host;
2432
2433 if (host + 1 > *port_count + host_base)
2434 *port_count = host + 1 - host_base;
2435 }
2436 closedir(dir);
2437 return host_base;
2438}
2439
a891a3c2
LM		 2440static void print_imsm_capability(const struct imsm_orom *orom)
2441{
0858eccf
AP		 2442 printf(" Platform : Intel(R) ");
2443 if (orom->capabilities == 0 && orom->driver_features == 0)
2444 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2445 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2446 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2447 else
2448 printf("Rapid Storage Technology%s\n",
2449 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2450 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2451 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2452 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2453 printf(" RAID Levels :%s%s%s%s%s\n",
2454 imsm_orom_has_raid0(orom) ? " raid0" : "",
2455 imsm_orom_has_raid1(orom) ? " raid1" : "",
2456 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2457 imsm_orom_has_raid10(orom) ? " raid10" : "",
2458 imsm_orom_has_raid5(orom) ? " raid5" : "");
2459 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2460 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2461 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2462 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2463 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2464 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2465 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2466 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2467 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2468 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2469 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2470 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2471 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2472 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2473 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2474 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2475 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2476 printf(" 2TB volumes :%s supported\n",
2477 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2478 printf(" 2TB disks :%s supported\n",
2479 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2480 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2481 printf(" Max Volumes : %d per array, %d per %s\n",
2482 orom->vpa, orom->vphba,
2483 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2484 return;
2485}
2486
e50cf220
MN		 2487static void print_imsm_capability_export(const struct imsm_orom *orom)
2488{
2489 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2490 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2491 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2492 orom->hotfix_ver, orom->build);
e50cf220
MN		 2493 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2494 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2495 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2496 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2497 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2498 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2499 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2500 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2501 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2502 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2503 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2504 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2505 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2506 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2507 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2508 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2509 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2510 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2511 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2512 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2513 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2514 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2515 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2516 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2517 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2518 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2519 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2520 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2521}
2522
9eafa1de 2523static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2524{
2525 /* There are two components to imsm platform support, the ahci SATA
2526 * controller and the option-rom. To find the SATA controller we
2527 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2528 * controller with the Intel vendor id is present. This approach
2529 * allows mdadm to leverage the kernel's ahci detection logic, with the
2530 * caveat that if ahci.ko is not loaded mdadm will not be able to
2531 * detect platform raid capabilities. The option-rom resides in a
2532 * platform "Adapter ROM". We scan for its signature to retrieve the
2533 * platform capabilities. If raid support is disabled in the BIOS the
2534 * option-rom capability structure will not be available.
2535 */
d665cc31 2536 struct sys_dev *list, *hba;
d665cc31
DW		 2537 int host_base = 0;
2538 int port_count = 0;
9eafa1de 2539 int result=1;
d665cc31 2540
5615172f 2541 if (enumerate_only) {
a891a3c2 2542 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2543 return 0;
a891a3c2
LM		 2544 list = find_intel_devices();
2545 if (!list)
2546 return 2;
2547 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2548 if (find_imsm_capability(hba)) {
2549 result = 0;
a891a3c2
LM		 2550 break;
2551 }
9eafa1de 2552 else
6b781d33 2553 result = 2;
a891a3c2 2554 }
a891a3c2 2555 return result;
5615172f
DW		 2556 }
2557
155cbb4c
LM		 2558 list = find_intel_devices();
2559 if (!list) {
ba728be7 2560 if (verbose > 0)
7a862a02 2561 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2562 return 2;
ba728be7 2563 } else if (verbose > 0)
155cbb4c 2564 print_found_intel_controllers(list);
d665cc31 2565
a891a3c2 2566 for (hba = list; hba; hba = hba->next) {
0858eccf 2567 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2568 continue;
0858eccf 2569 if (!find_imsm_capability(hba)) {
60f0f54d 2570 char buf[PATH_MAX];
e7b84f9d 2571 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2572 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2573 get_sys_dev_type(hba->type));
0858eccf
AP		 2574 continue;
2575 }
2576 result = 0;
2577 }
2578
2579 if (controller_path && result == 1) {
2580 pr_err("no active Intel(R) RAID controller found under %s\n",
2581 controller_path);
2582 return result;
2583 }
2584
5e1d6128 2585 const struct orom_entry *entry;
0858eccf 2586
5e1d6128 2587 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2588 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2589 print_imsm_capability(&entry->orom);
32716c51
PB		 2590 printf(" 3rd party NVMe :%s supported\n",
2591 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2592 for (hba = list; hba; hba = hba->next) {
2593 if (hba->type == SYS_DEV_VMD) {
2594 char buf[PATH_MAX];
60f0f54d
PB		 2595 printf(" I/O Controller : %s (%s)\n",
2596 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2597 if (print_nvme_info(hba)) {
b5eece69
PB		 2598 if (verbose > 0)
2599 pr_err("failed to get devices attached to VMD domain.\n");
2600 result |= 2;
2601 }
60f0f54d
PB		 2602 }
2603 }
07cb1e57 2604 printf("\n");
60f0f54d
PB		 2605 continue;
2606 }
0858eccf 2607
60f0f54d
PB		 2608 print_imsm_capability(&entry->orom);
2609 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2610 for (hba = list; hba; hba = hba->next) {
2611 if (hba->type == SYS_DEV_NVME)
6da53c0e 2612 print_nvme_info(hba);
0858eccf 2613 }
60f0f54d 2614 printf("\n");
0858eccf
AP		 2615 continue;
2616 }
2617
2618 struct devid_list *devid;
5e1d6128 2619 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2620 hba = device_by_id(devid->devid);
2621 if (!hba)
2622 continue;
2623
9eafa1de
MN		 2624 printf(" I/O Controller : %s (%s)\n",
2625 hba->path, get_sys_dev_type(hba->type));
2626 if (hba->type == SYS_DEV_SATA) {
2627 host_base = ahci_get_port_count(hba->path, &port_count);
2628 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2629 if (verbose > 0)
7a862a02 2630 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2631 result |= 2;
2632 }
120dc887
LM		 2633 }
2634 }
0858eccf 2635 printf("\n");
d665cc31 2636 }
155cbb4c 2637
120dc887 2638 return result;
d665cc31 2639}
e50cf220 2640
9eafa1de 2641static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2642{
e50cf220
MN		 2643 struct sys_dev *list, *hba;
2644 int result=1;
2645
2646 list = find_intel_devices();
2647 if (!list) {
2648 if (verbose > 0)
2649 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2650 result = 2;
e50cf220
MN		 2651 return result;
2652 }
2653
2654 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2655 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2656 continue;
60f0f54d
PB		 2657 if (!find_imsm_capability(hba) && verbose > 0) {
2658 char buf[PATH_MAX];
2659 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2660 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2661 }
0858eccf 2662 else
e50cf220 2663 result = 0;
e50cf220
MN		 2664 }
2665
5e1d6128 2666 const struct orom_entry *entry;
0858eccf 2667
60f0f54d
PB		 2668 for (entry = orom_entries; entry; entry = entry->next) {
2669 if (entry->type == SYS_DEV_VMD) {
2670 for (hba = list; hba; hba = hba->next)
2671 print_imsm_capability_export(&entry->orom);
2672 continue;
2673 }
5e1d6128 2674 print_imsm_capability_export(&entry->orom);
60f0f54d 2675 }
0858eccf 2676
e50cf220
MN		 2677 return result;
2678}
2679
cdddbdbc
DW		 2680static int match_home_imsm(struct supertype *st, char *homehost)
2681{
5115ca67
DW		 2682 /* the imsm metadata format does not specify any host
2683 * identification information. We return -1 since we can never
2684 * confirm nor deny whether a given array is "meant" for this
148acb7b 2685 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2686 * exclude member disks that do not belong, and we rely on
2687 * mdadm.conf to specify the arrays that should be assembled.
2688 * Auto-assembly may still pick up "foreign" arrays.
2689 */
cdddbdbc 2690
9362c1c8 2691 return -1;
cdddbdbc
DW		 2692}
2693
2694static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2695{
51006d85
N		 2696 /* The uuid returned here is used for:
2697 * uuid to put into bitmap file (Create, Grow)
2698 * uuid for backup header when saving critical section (Grow)
2699 * comparing uuids when re-adding a device into an array
2700 * In these cases the uuid required is that of the data-array,
2701 * not the device-set.
2702 * uuid to recognise same set when adding a missing device back
2703 * to an array. This is a uuid for the device-set.
1011e834 2704 *
51006d85
N		 2705 * For each of these we can make do with a truncated
2706 * or hashed uuid rather than the original, as long as
2707 * everyone agrees.
2708 * In each case the uuid required is that of the data-array,
2709 * not the device-set.
43dad3d6 2710 */
51006d85
N		 2711 /* imsm does not track uuid's so we synthesis one using sha1 on
2712 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2713 * - the orig_family_num of the container
51006d85
N		 2714 * - the index number of the volume
2715 * - the 'serial' number of the volume.
2716 * Hopefully these are all constant.
2717 */
2718 struct intel_super *super = st->sb;
43dad3d6 2719
51006d85
N		 2720 char buf[20];
2721 struct sha1_ctx ctx;
2722 struct imsm_dev *dev = NULL;
148acb7b 2723 __u32 family_num;
51006d85 2724
148acb7b
DW		 2725 /* some mdadm versions failed to set ->orig_family_num, in which
2726 * case fall back to ->family_num. orig_family_num will be
2727 * fixed up with the first metadata update.
2728 */
2729 family_num = super->anchor->orig_family_num;
2730 if (family_num == 0)
2731 family_num = super->anchor->family_num;
51006d85 2732 sha1_init_ctx(&ctx);
92bd8f8d 2733 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2734 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2735 if (super->current_vol >= 0)
2736 dev = get_imsm_dev(super, super->current_vol);
2737 if (dev) {
2738 __u32 vol = super->current_vol;
2739 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2740 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2741 }
2742 sha1_finish_ctx(&ctx, buf);
2743 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2744}
2745
0d481d37 2746#if 0
4f5bc454
DW		 2747static void
2748get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2749{
cdddbdbc
DW		 2750 __u8 *v = get_imsm_version(mpb);
2751 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2752 char major[] = { 0, 0, 0 };
2753 char minor[] = { 0 ,0, 0 };
2754 char patch[] = { 0, 0, 0 };
2755 char *ver_parse[] = { major, minor, patch };
2756 int i, j;
2757
2758 i = j = 0;
2759 while (*v != '\0' && v < end) {
2760 if (*v != '.' && j < 2)
2761 ver_parse[i][j++] = *v;
2762 else {
2763 i++;
2764 j = 0;
2765 }
2766 v++;
2767 }
2768
4f5bc454
DW		 2769 *m = strtol(minor, NULL, 0);
2770 *p = strtol(patch, NULL, 0);
2771}
0d481d37 2772#endif
4f5bc454 2773
1e5c6983
DW		 2774static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2775{
2776 /* migr_strip_size when repairing or initializing parity */
238c0a71 2777 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2778 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2779
2780 switch (get_imsm_raid_level(map)) {
2781 case 5:
2782 case 10:
2783 return chunk;
2784 default:
2785 return 128*1024 >> 9;
2786 }
2787}
2788
2789static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2790{
2791 /* migr_strip_size when rebuilding a degraded disk, no idea why
2792 * this is different than migr_strip_size_resync(), but it's good
2793 * to be compatible
2794 */
238c0a71 2795 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2796 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2797
2798 switch (get_imsm_raid_level(map)) {
2799 case 1:
2800 case 10:
2801 if (map->num_members % map->num_domains == 0)
2802 return 128*1024 >> 9;
2803 else
2804 return chunk;
2805 case 5:
2806 return max((__u32) 64*1024 >> 9, chunk);
2807 default:
2808 return 128*1024 >> 9;
2809 }
2810}
2811
2812static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2813{
238c0a71
AK		 2814 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2815 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2816 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2817 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2818
2819 return max((__u32) 1, hi_chunk / lo_chunk);
2820}
2821
2822static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2823{
238c0a71 2824 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2825 int level = get_imsm_raid_level(lo);
2826
2827 if (level == 1 || level == 10) {
238c0a71 2828 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2829
2830 return hi->num_domains;
2831 } else
2832 return num_stripes_per_unit_resync(dev);
2833}
2834
9529d343 2835static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2836{
2837 /* named 'imsm_' because raid0, raid1 and raid10
2838 * counter-intuitively have the same number of data disks
2839 */
1e5c6983
DW		 2840 switch (get_imsm_raid_level(map)) {
2841 case 0:
36fd8ccc
AK		 2842 return map->num_members;
2843 break;
1e5c6983
DW		 2844 case 1:
2845 case 10:
36fd8ccc 2846 return map->num_members/2;
1e5c6983
DW		 2847 case 5:
2848 return map->num_members - 1;
2849 default:
1ade5cc1 2850 dprintf("unsupported raid level\n");
1e5c6983
DW		 2851 return 0;
2852 }
2853}
2854
44490938
MD		 2855static unsigned long long calc_component_size(struct imsm_map *map,
2856 struct imsm_dev *dev)
2857{
2858 unsigned long long component_size;
2859 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2860 long long calc_dev_size = 0;
44490938
MD		 2861 unsigned int member_disks = imsm_num_data_members(map);
2862
2863 if (member_disks == 0)
2864 return 0;
2865
2866 component_size = per_dev_array_size(map);
2867 calc_dev_size = component_size * member_disks;
2868
2869 /* Component size is rounded to 1MB so difference between size from
2870 * metadata and size calculated from num_data_stripes equals up to
2871 * 2048 blocks per each device. If the difference is higher it means
2872 * that array size was expanded and num_data_stripes was not updated.
2873 */
a4f7290c 2874 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2875 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2876 component_size = dev_size / member_disks;
2877 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2878 component_size / map->blocks_per_strip,
2879 num_data_stripes(map));
2880 }
2881
2882 return component_size;
2883}
2884
1e5c6983
DW		 2885static __u32 parity_segment_depth(struct imsm_dev *dev)
2886{
238c0a71 2887 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2888 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2889
2890 switch(get_imsm_raid_level(map)) {
2891 case 1:
2892 case 10:
2893 return chunk * map->num_domains;
2894 case 5:
2895 return chunk * map->num_members;
2896 default:
2897 return chunk;
2898 }
2899}
2900
2901static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2902{
238c0a71 2903 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2904 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2905 __u32 strip = block / chunk;
2906
2907 switch (get_imsm_raid_level(map)) {
2908 case 1:
2909 case 10: {
2910 __u32 vol_strip = (strip * map->num_domains) + 1;
2911 __u32 vol_stripe = vol_strip / map->num_members;
2912
2913 return vol_stripe * chunk + block % chunk;
2914 } case 5: {
2915 __u32 stripe = strip / (map->num_members - 1);
2916
2917 return stripe * chunk + block % chunk;
2918 }
2919 default:
2920 return 0;
2921 }
2922}
2923
c47b0ff6
AK		 2924static __u64 blocks_per_migr_unit(struct intel_super *super,
2925 struct imsm_dev *dev)
1e5c6983
DW		 2926{
2927 /* calculate the conversion factor between per member 'blocks'
2928 * (md/{resync,rebuild}_start) and imsm migration units, return
2929 * 0 for the 'not migrating' and 'unsupported migration' cases
2930 */
2931 if (!dev->vol.migr_state)
2932 return 0;
2933
2934 switch (migr_type(dev)) {
c47b0ff6
AK		 2935 case MIGR_GEN_MIGR: {
2936 struct migr_record *migr_rec = super->migr_rec;
2937 return __le32_to_cpu(migr_rec->blocks_per_unit);
2938 }
1e5c6983
DW		 2939 case MIGR_VERIFY:
2940 case MIGR_REPAIR:
2941 case MIGR_INIT: {
238c0a71 2942 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2943 __u32 stripes_per_unit;
2944 __u32 blocks_per_unit;
2945 __u32 parity_depth;
2946 __u32 migr_chunk;
2947 __u32 block_map;
2948 __u32 block_rel;
2949 __u32 segment;
2950 __u32 stripe;
2951 __u8 disks;
2952
2953 /* yes, this is really the translation of migr_units to
2954 * per-member blocks in the 'resync' case
2955 */
2956 stripes_per_unit = num_stripes_per_unit_resync(dev);
2957 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2958 disks = imsm_num_data_members(map);
1e5c6983 2959 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2960 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2961 segment = blocks_per_unit / stripe;
2962 block_rel = blocks_per_unit - segment * stripe;
2963 parity_depth = parity_segment_depth(dev);
2964 block_map = map_migr_block(dev, block_rel);
2965 return block_map + parity_depth * segment;
2966 }
2967 case MIGR_REBUILD: {
2968 __u32 stripes_per_unit;
2969 __u32 migr_chunk;
2970
2971 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2972 migr_chunk = migr_strip_blocks_rebuild(dev);
2973 return migr_chunk * stripes_per_unit;
2974 }
1e5c6983
DW		 2975 case MIGR_STATE_CHANGE:
2976 default:
2977 return 0;
2978 }
2979}
2980
c2c087e6
DW		 2981static int imsm_level_to_layout(int level)
2982{
2983 switch (level) {
2984 case 0:
2985 case 1:
2986 return 0;
2987 case 5:
2988 case 6:
a380c027 2989 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2990 case 10:
c92a2527 2991 return 0x102;
c2c087e6 2992 }
a18a888e 2993 return UnSet;
c2c087e6
DW		 2994}
2995
8e59f3d8
AK		 2996/*******************************************************************************
2997 * Function: read_imsm_migr_rec
2998 * Description: Function reads imsm migration record from last sector of disk
2999 * Parameters:
3000 * fd : disk descriptor
3001 * super : metadata info
3002 * Returns:
3003 * 0 : success,
3004 * -1 : fail
3005 ******************************************************************************/
3006static int read_imsm_migr_rec(int fd, struct intel_super *super)
3007{
3008 int ret_val = -1;
de44e46f 3009 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3010 unsigned long long dsize;
3011
3012 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3013 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3014 SEEK_SET) < 0) {
e7b84f9d
N		 3015 pr_err("Cannot seek to anchor block: %s\n",
3016 strerror(errno));
8e59f3d8
AK		 3017 goto out;
3018 }
466070ad 3019 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3020 MIGR_REC_BUF_SECTORS*sector_size) !=
3021 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3022 pr_err("Cannot read migr record block: %s\n",
3023 strerror(errno));
8e59f3d8
AK		 3024 goto out;
3025 }
3026 ret_val = 0;
de44e46f
PB		 3027 if (sector_size == 4096)
3028 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3029
3030out:
3031 return ret_val;
3032}
3033
3136abe5
AK		 3034static struct imsm_dev *imsm_get_device_during_migration(
3035 struct intel_super *super)
3036{
3037
3038 struct intel_dev *dv;
3039
3040 for (dv = super->devlist; dv; dv = dv->next) {
3041 if (is_gen_migration(dv->dev))
3042 return dv->dev;
3043 }
3044 return NULL;
3045}
3046
8e59f3d8
AK		 3047/*******************************************************************************
3048 * Function: load_imsm_migr_rec
3049 * Description: Function reads imsm migration record (it is stored at the last
3050 * sector of disk)
3051 * Parameters:
3052 * super : imsm internal array info
3053 * info : general array info
3054 * Returns:
3055 * 0 : success
3056 * -1 : fail
4c965cc9 3057 * -2 : no migration in progress
8e59f3d8
AK		 3058 ******************************************************************************/
3059static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3060{
3061 struct mdinfo *sd;
594dc1b8 3062 struct dl *dl;
8e59f3d8
AK		 3063 char nm[30];
3064 int retval = -1;
3065 int fd = -1;
3136abe5 3066 struct imsm_dev *dev;
594dc1b8 3067 struct imsm_map *map;
b4ab44d8 3068 int slot = -1;
3136abe5
AK		 3069
3070 /* find map under migration */
3071 dev = imsm_get_device_during_migration(super);
3072 /* nothing to load,no migration in progress?
3073 */
3074 if (dev == NULL)
4c965cc9 3075 return -2;
8e59f3d8
AK		 3076
3077 if (info) {
3078 for (sd = info->devs ; sd ; sd = sd->next) {
3079 /* read only from one of the first two slots */
12fe93e9
TM		 3080 if ((sd->disk.raid_disk < 0) ||
3081 (sd->disk.raid_disk > 1))
8e59f3d8 3082 continue;
3136abe5 3083
8e59f3d8
AK		 3084 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3085 fd = dev_open(nm, O_RDONLY);
3086 if (fd >= 0)
3087 break;
3088 }
3089 }
3090 if (fd < 0) {
12fe93e9 3091 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3092 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3093 /* skip spare and failed disks
3094 */
3095 if (dl->index < 0)
3096 continue;
8e59f3d8 3097 /* read only from one of the first two slots */
3136abe5
AK		 3098 if (map)
3099 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3100 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3101 continue;
3102 sprintf(nm, "%d:%d", dl->major, dl->minor);
3103 fd = dev_open(nm, O_RDONLY);
3104 if (fd >= 0)
3105 break;
3106 }
3107 }
3108 if (fd < 0)
3109 goto out;
3110 retval = read_imsm_migr_rec(fd, super);
3111
3112out:
3113 if (fd >= 0)
3114 close(fd);
3115 return retval;
3116}
3117
c17608ea
AK		 3118/*******************************************************************************
3119 * function: imsm_create_metadata_checkpoint_update
3120 * Description: It creates update for checkpoint change.
3121 * Parameters:
3122 * super : imsm internal array info
3123 * u : pointer to prepared update
3124 * Returns:
3125 * Uptate length.
3126 * If length is equal to 0, input pointer u contains no update
3127 ******************************************************************************/
3128static int imsm_create_metadata_checkpoint_update(
3129 struct intel_super *super,
3130 struct imsm_update_general_migration_checkpoint **u)
3131{
3132
3133 int update_memory_size = 0;
3134
1ade5cc1 3135 dprintf("(enter)\n");
c17608ea
AK		 3136
3137 if (u == NULL)
3138 return 0;
3139 *u = NULL;
3140
3141 /* size of all update data without anchor */
3142 update_memory_size =
3143 sizeof(struct imsm_update_general_migration_checkpoint);
3144
503975b9 3145 *u = xcalloc(1, update_memory_size);
c17608ea 3146 if (*u == NULL) {
1ade5cc1 3147 dprintf("error: cannot get memory\n");
c17608ea
AK		 3148 return 0;
3149 }
3150 (*u)->type = update_general_migration_checkpoint;
9f421827 3151 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3152 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3153
3154 return update_memory_size;
3155}
3156
c17608ea
AK		 3157static void imsm_update_metadata_locally(struct supertype *st,
3158 void *buf, int len);
3159
687629c2
AK		 3160/*******************************************************************************
3161 * Function: write_imsm_migr_rec
3162 * Description: Function writes imsm migration record
3163 * (at the last sector of disk)
3164 * Parameters:
3165 * super : imsm internal array info
3166 * Returns:
3167 * 0 : success
3168 * -1 : if fail
3169 ******************************************************************************/
3170static int write_imsm_migr_rec(struct supertype *st)
3171{
3172 struct intel_super *super = st->sb;
de44e46f 3173 unsigned int sector_size = super->sector_size;
687629c2
AK		 3174 unsigned long long dsize;
3175 char nm[30];
3176 int fd = -1;
3177 int retval = -1;
3178 struct dl *sd;
c17608ea
AK		 3179 int len;
3180 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3181 struct imsm_dev *dev;
594dc1b8 3182 struct imsm_map *map;
3136abe5
AK		 3183
3184 /* find map under migration */
3185 dev = imsm_get_device_during_migration(super);
3186 /* if no migration, write buffer anyway to clear migr_record
3187 * on disk based on first available device
3188 */
3189 if (dev == NULL)
3190 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3191 super->current_vol);
3192
44bfe6df 3193 map = get_imsm_map(dev, MAP_0);
687629c2 3194
de44e46f
PB		 3195 if (sector_size == 4096)
3196 convert_to_4k_imsm_migr_rec(super);
687629c2 3197 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3198 int slot = -1;
3136abe5
AK		 3199
3200 /* skip failed and spare devices */
3201 if (sd->index < 0)
3202 continue;
687629c2 3203 /* write to 2 first slots only */
3136abe5
AK		 3204 if (map)
3205 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3206 if (map == NULL || slot > 1 || slot < 0)
687629c2 3207 continue;
3136abe5 3208
687629c2
AK		 3209 sprintf(nm, "%d:%d", sd->major, sd->minor);
3210 fd = dev_open(nm, O_RDWR);
3211 if (fd < 0)
3212 continue;
3213 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3214 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3215 SEEK_SET) < 0) {
e7b84f9d
N		 3216 pr_err("Cannot seek to anchor block: %s\n",
3217 strerror(errno));
687629c2
AK		 3218 goto out;
3219 }
466070ad 3220 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3221 MIGR_REC_BUF_SECTORS*sector_size) !=
3222 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3223 pr_err("Cannot write migr record block: %s\n",
3224 strerror(errno));
687629c2
AK		 3225 goto out;
3226 }
3227 close(fd);
3228 fd = -1;
3229 }
de44e46f
PB		 3230 if (sector_size == 4096)
3231 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3232 /* update checkpoint information in metadata */
3233 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3234 if (len <= 0) {
3235 dprintf("imsm: Cannot prepare update\n");
3236 goto out;
3237 }
3238 /* update metadata locally */
3239 imsm_update_metadata_locally(st, u, len);
3240 /* and possibly remotely */
3241 if (st->update_tail) {
3242 append_metadata_update(st, u, len);
3243 /* during reshape we do all work inside metadata handler
3244 * manage_reshape(), so metadata update has to be triggered
3245 * insida it
3246 */
3247 flush_metadata_updates(st);
3248 st->update_tail = &st->updates;
3249 } else
3250 free(u);
687629c2
AK		 3251
3252 retval = 0;
3253 out:
3254 if (fd >= 0)
3255 close(fd);
3256 return retval;
3257}
3258
e2962bfc
AK		 3259/* spare/missing disks activations are not allowe when
3260 * array/container performs reshape operation, because
3261 * all arrays in container works on the same disks set
3262 */
3263int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3264{
3265 int rv = 0;
3266 struct intel_dev *i_dev;
3267 struct imsm_dev *dev;
3268
3269 /* check whole container
3270 */
3271 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3272 dev = i_dev->dev;
3ad25638 3273 if (is_gen_migration(dev)) {
e2962bfc
AK		 3274 /* No repair during any migration in container
3275 */
3276 rv = 1;
3277 break;
3278 }
3279 }
3280 return rv;
3281}
3e684231 3282static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3283 int chunk_size,
f36a9ecd 3284 unsigned int sector_size,
c41e00b2
AK		 3285 unsigned long long component_size)
3286{
3e684231 3287 unsigned int component_size_alignment;
c41e00b2 3288
3e684231 3289 /* check component size alignment
c41e00b2 3290 */
3e684231 3291 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3292
3e684231 3293 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3294 level, chunk_size, component_size,
3e684231 3295 component_size_alignment);
c41e00b2 3296
3e684231
MZ		 3297 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3298 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3299 component_size);
3e684231 3300 component_size -= component_size_alignment;
1ade5cc1 3301 dprintf_cont("to %llu (%i).\n",
3e684231 3302 component_size, component_size_alignment);
c41e00b2
AK		 3303 }
3304
3305 return component_size;
3306}
e2962bfc 3307
2432ce9b
AP		 3308static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3309{
3310 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3311 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3312
3313 return pba_of_lba0(map) +
3314 (num_data_stripes(map) * map->blocks_per_strip);
3315}
3316
a5d85af7 3317static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3318{
3319 struct intel_super *super = st->sb;
c47b0ff6 3320 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3321 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3322 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3323 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3324 struct imsm_map *map_to_analyse = map;
efb30e7f 3325 struct dl *dl;
a5d85af7 3326 int map_disks = info->array.raid_disks;
bf5a934a 3327
95eeceeb 3328 memset(info, 0, sizeof(*info));
b335e593
AK		 3329 if (prev_map)
3330 map_to_analyse = prev_map;
3331
ca0748fa 3332 dl = super->current_disk;
9894ec0d 3333
bf5a934a 3334 info->container_member = super->current_vol;
cd0430a1 3335 info->array.raid_disks = map->num_members;
b335e593 3336 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3337 info->array.layout = imsm_level_to_layout(info->array.level);
3338 info->array.md_minor = -1;
3339 info->array.ctime = 0;
3340 info->array.utime = 0;
b335e593
AK		 3341 info->array.chunk_size =
3342 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3343 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3344 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3345 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3346
3f510843 3347 if (is_gen_migration(dev)) {
3f83228a 3348 info->reshape_active = 1;
b335e593
AK		 3349 info->new_level = get_imsm_raid_level(map);
3350 info->new_layout = imsm_level_to_layout(info->new_level);
3351 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3352 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3353 if (info->delta_disks) {
3354 /* this needs to be applied to every array
3355 * in the container.
3356 */
81219e70 3357 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3358 }
3f83228a
N		 3359 /* We shape information that we give to md might have to be
3360 * modify to cope with md's requirement for reshaping arrays.
3361 * For example, when reshaping a RAID0, md requires it to be
3362 * presented as a degraded RAID4.
3363 * Also if a RAID0 is migrating to a RAID5 we need to specify
3364 * the array as already being RAID5, but the 'before' layout
3365 * is a RAID4-like layout.
3366 */
3367 switch (info->array.level) {
3368 case 0:
3369 switch(info->new_level) {
3370 case 0:
3371 /* conversion is happening as RAID4 */
3372 info->array.level = 4;
3373 info->array.raid_disks += 1;
3374 break;
3375 case 5:
3376 /* conversion is happening as RAID5 */
3377 info->array.level = 5;
3378 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3379 info->delta_disks -= 1;
3380 break;
3381 default:
3382 /* FIXME error message */
3383 info->array.level = UnSet;
3384 break;
3385 }
3386 break;
3387 }
b335e593
AK		 3388 } else {
3389 info->new_level = UnSet;
3390 info->new_layout = UnSet;
3391 info->new_chunk = info->array.chunk_size;
3f83228a 3392 info->delta_disks = 0;
b335e593 3393 }
ca0748fa 3394
efb30e7f
DW		 3395 if (dl) {
3396 info->disk.major = dl->major;
3397 info->disk.minor = dl->minor;
ca0748fa 3398 info->disk.number = dl->index;
656b6b5a
N		 3399 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3400 dl->index);
efb30e7f 3401 }
bf5a934a 3402
5551b113 3403 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3404 info->component_size = calc_component_size(map, dev);
3e684231 3405 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3406 info->array.level,
3407 info->array.chunk_size,
f36a9ecd 3408 super->sector_size,
c41e00b2 3409 info->component_size);
5e46202e 3410 info->bb.supported = 1;
139dae11 3411
301406c9 3412 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3413 info->recovery_start = MaxSector;
bf5a934a 3414
c2462068
PB		 3415 if (info->array.level == 5 &&
3416 (dev->rwh_policy == RWH_DISTRIBUTED ||
3417 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3418 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3419 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3420 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3421 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3422 else
3423 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3424 >> 9;
2432ce9b
AP		 3425 } else if (info->array.level <= 0) {
3426 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3427 } else {
3428 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3429 }
3430
d2e6d5d6 3431 info->reshape_progress = 0;
b6796ce1 3432 info->resync_start = MaxSector;
b9172665 3433 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3434 !(info->array.state & 1)) &&
b9172665 3435 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3436 info->resync_start = 0;
b6796ce1
AK		 3437 }
3438 if (dev->vol.migr_state) {
1e5c6983
DW		 3439 switch (migr_type(dev)) {
3440 case MIGR_REPAIR:
3441 case MIGR_INIT: {
c47b0ff6
AK		 3442 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3443 dev);
1e5c6983
DW		 3444 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3445
3446 info->resync_start = blocks_per_unit * units;
3447 break;
3448 }
d2e6d5d6 3449 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3450 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3451 dev);
9f421827 3452 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3453 unsigned long long array_blocks;
3454 int used_disks;
d2e6d5d6 3455
befb629b
AK		 3456 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3457 (units <
9f421827 3458 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3459 (super->migr_rec->rec_status ==
3460 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3461 units++;
3462
d2e6d5d6 3463 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3464
7a862a02 3465 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3466 (unsigned long long)units,
3467 (unsigned long long)blocks_per_unit,
3468 info->reshape_progress);
75156c46 3469
9529d343 3470 used_disks = imsm_num_data_members(prev_map);
75156c46 3471 if (used_disks > 0) {
44490938 3472 array_blocks = per_dev_array_size(map) *
75156c46 3473 used_disks;
b53bfba6
TM		 3474 info->custom_array_size =
3475 round_size_to_mb(array_blocks,
3476 used_disks);
3477
75156c46 3478 }
d2e6d5d6 3479 }
1e5c6983
DW		 3480 case MIGR_VERIFY:
3481 /* we could emulate the checkpointing of
3482 * 'sync_action=check' migrations, but for now
3483 * we just immediately complete them
3484 */
3485 case MIGR_REBUILD:
3486 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3487 case MIGR_STATE_CHANGE:
3488 /* FIXME handle other migrations */
3489 default:
3490 /* we are not dirty, so... */
3491 info->resync_start = MaxSector;
3492 }
b6796ce1 3493 }
301406c9
DW		 3494
3495 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3496 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3497
f35f2525
N		 3498 info->array.major_version = -1;
3499 info->array.minor_version = -2;
4dd2df09 3500 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3501 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3502 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3503
3504 if (dmap) {
3505 int i, j;
3506 for (i=0; i<map_disks; i++) {
3507 dmap[i] = 0;
3508 if (i < info->array.raid_disks) {
3509 struct imsm_disk *dsk;
238c0a71 3510 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3511 dsk = get_imsm_disk(super, j);
3512 if (dsk && (dsk->status & CONFIGURED_DISK))
3513 dmap[i] = 1;
3514 }
3515 }
3516 }
81ac8b4d 3517}
bf5a934a 3518
3b451610
AK		 3519static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3520 int failed, int look_in_map);
3521
3522static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3523 int look_in_map);
3524
3525static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3526{
3527 if (is_gen_migration(dev)) {
3528 int failed;
3529 __u8 map_state;
3530 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3531
3532 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3533 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3534 if (map2->map_state != map_state) {
3535 map2->map_state = map_state;
3536 super->updates_pending++;
3537 }
3538 }
3539}
97b4d0e9
DW		 3540
3541static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3542{
3543 struct dl *d;
3544
3545 for (d = super->missing; d; d = d->next)
3546 if (d->index == index)
3547 return &d->disk;
3548 return NULL;
3549}
3550
a5d85af7 3551static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3552{
3553 struct intel_super *super = st->sb;
4f5bc454 3554 struct imsm_disk *disk;
a5d85af7 3555 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3556 int max_enough = -1;
3557 int i;
3558 struct imsm_super *mpb;
4f5bc454 3559
bf5a934a 3560 if (super->current_vol >= 0) {
a5d85af7 3561 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3562 return;
3563 }
95eeceeb 3564 memset(info, 0, sizeof(*info));
d23fe947
DW		 3565
3566 /* Set raid_disks to zero so that Assemble will always pull in valid
3567 * spares
3568 */
3569 info->array.raid_disks = 0;
cdddbdbc
DW		 3570 info->array.level = LEVEL_CONTAINER;
3571 info->array.layout = 0;
3572 info->array.md_minor = -1;
1011e834 3573 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3574 info->array.utime = 0;
3575 info->array.chunk_size = 0;
3576
3577 info->disk.major = 0;
3578 info->disk.minor = 0;
cdddbdbc 3579 info->disk.raid_disk = -1;
c2c087e6 3580 info->reshape_active = 0;
f35f2525
N		 3581 info->array.major_version = -1;
3582 info->array.minor_version = -2;
c2c087e6 3583 strcpy(info->text_version, "imsm");
a67dd8cc 3584 info->safe_mode_delay = 0;
c2c087e6
DW		 3585 info->disk.number = -1;
3586 info->disk.state = 0;
c5afc314 3587 info->name[0] = 0;
921d9e16 3588 info->recovery_start = MaxSector;
3ad25638 3589 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3590 info->bb.supported = 1;
c2c087e6 3591
97b4d0e9 3592 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3593 mpb = super->anchor;
b7d81a38 3594 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3595
ab3cb6b3
N		 3596 for (i = 0; i < mpb->num_raid_devs; i++) {
3597 struct imsm_dev *dev = get_imsm_dev(super, i);
3598 int failed, enough, j, missing = 0;
3599 struct imsm_map *map;
3600 __u8 state;
97b4d0e9 3601
3b451610
AK		 3602 failed = imsm_count_failed(super, dev, MAP_0);
3603 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3604 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3605
3606 /* any newly missing disks?
3607 * (catches single-degraded vs double-degraded)
3608 */
3609 for (j = 0; j < map->num_members; j++) {
238c0a71 3610 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3611 __u32 idx = ord_to_idx(ord);
3612
20dc76d1
MT		 3613 if (super->disks && super->disks->index == (int)idx)
3614 info->disk.raid_disk = j;
3615
ab3cb6b3
N		 3616 if (!(ord & IMSM_ORD_REBUILD) &&
3617 get_imsm_missing(super, idx)) {
3618 missing = 1;
3619 break;
3620 }
97b4d0e9 3621 }
ab3cb6b3
N		 3622
3623 if (state == IMSM_T_STATE_FAILED)
3624 enough = -1;
3625 else if (state == IMSM_T_STATE_DEGRADED &&
3626 (state != map->map_state || missing))
3627 enough = 0;
3628 else /* we're normal, or already degraded */
3629 enough = 1;
d2bde6d3
AK		 3630 if (is_gen_migration(dev) && missing) {
3631 /* during general migration we need all disks
3632 * that process is running on.
3633 * No new missing disk is allowed.
3634 */
3635 max_enough = -1;
3636 enough = -1;
3637 /* no more checks necessary
3638 */
3639 break;
3640 }
ab3cb6b3
N		 3641 /* in the missing/failed disk case check to see
3642 * if at least one array is runnable
3643 */
3644 max_enough = max(max_enough, enough);
3645 }
1ade5cc1 3646 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3647 info->container_enough = max_enough;
97b4d0e9 3648
4a04ec6c 3649 if (super->disks) {
14e8215b
DW		 3650 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3651
b9f594fe 3652 disk = &super->disks->disk;
5551b113 3653 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3654 info->component_size = reserved;
25ed7e59 3655 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3656 /* we don't change info->disk.raid_disk here because
3657 * this state will be finalized in mdmon after we have
3658 * found the 'most fresh' version of the metadata
3659 */
25ed7e59 3660 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3661 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3662 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3663 }
a575e2a7
DW		 3664
3665 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3666 * ->compare_super may have updated the 'num_raid_devs' field for spares
3667 */
3668 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3669 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3670 else
3671 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3672
3673 /* I don't know how to compute 'map' on imsm, so use safe default */
3674 if (map) {
3675 int i;
3676 for (i = 0; i < map_disks; i++)
3677 map[i] = 1;
3678 }
3679
cdddbdbc
DW		 3680}
3681
5c4cd5da
AC		 3682/* allocates memory and fills disk in mdinfo structure
3683 * for each disk in array */
3684struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3685{
594dc1b8 3686 struct mdinfo *mddev;
5c4cd5da
AC		 3687 struct intel_super *super = st->sb;
3688 struct imsm_disk *disk;
3689 int count = 0;
3690 struct dl *dl;
3691 if (!super || !super->disks)
3692 return NULL;
3693 dl = super->disks;
503975b9 3694 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3695 while (dl) {
3696 struct mdinfo *tmp;
3697 disk = &dl->disk;
503975b9 3698 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3699 if (mddev->devs)
3700 tmp->next = mddev->devs;
3701 mddev->devs = tmp;
3702 tmp->disk.number = count++;
3703 tmp->disk.major = dl->major;
3704 tmp->disk.minor = dl->minor;
3705 tmp->disk.state = is_configured(disk) ?
3706 (1 << MD_DISK_ACTIVE) : 0;
3707 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3708 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3709 tmp->disk.raid_disk = -1;
3710 dl = dl->next;
3711 }
3712 return mddev;
3713}
3714
cdddbdbc
DW		 3715static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3716 char *update, char *devname, int verbose,
3717 int uuid_set, char *homehost)
3718{
f352c545
DW		 3719 /* For 'assemble' and 'force' we need to return non-zero if any
3720 * change was made. For others, the return value is ignored.
3721 * Update options are:
3722 * force-one : This device looks a bit old but needs to be included,
3723 * update age info appropriately.
3724 * assemble: clear any 'faulty' flag to allow this device to
3725 * be assembled.
3726 * force-array: Array is degraded but being forced, mark it clean
3727 * if that will be needed to assemble it.
3728 *
3729 * newdev: not used ????
3730 * grow: Array has gained a new device - this is currently for
3731 * linear only
3732 * resync: mark as dirty so a resync will happen.
3733 * name: update the name - preserving the homehost
6e46bf34 3734 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3735 *
3736 * Following are not relevant for this imsm:
3737 * sparc2.2 : update from old dodgey metadata
3738 * super-minor: change the preferred_minor number
3739 * summaries: update redundant counters.
f352c545
DW		 3740 * homehost: update the recorded homehost
3741 * _reshape_progress: record new reshape_progress position.
3742 */
6e46bf34
DW		 3743 int rv = 1;
3744 struct intel_super *super = st->sb;
3745 struct imsm_super *mpb;
f352c545 3746
6e46bf34
DW		 3747 /* we can only update container info */
3748 if (!super || super->current_vol >= 0 || !super->anchor)
3749 return 1;
3750
3751 mpb = super->anchor;
3752
81a5b4f5
N		 3753 if (strcmp(update, "uuid") == 0) {
3754 /* We take this to mean that the family_num should be updated.
3755 * However that is much smaller than the uuid so we cannot really
3756 * allow an explicit uuid to be given. And it is hard to reliably
3757 * know if one was.
3758 * So if !uuid_set we know the current uuid is random and just used
3759 * the first 'int' and copy it to the other 3 positions.
3760 * Otherwise we require the 4 'int's to be the same as would be the
3761 * case if we are using a random uuid. So an explicit uuid will be
3762 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3763 */
81a5b4f5
N		 3764 if (!uuid_set) {
3765 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3766 rv = 0;
81a5b4f5
N		 3767 } else {
3768 if (info->uuid[0] != info->uuid[1] ||
3769 info->uuid[1] != info->uuid[2] ||
3770 info->uuid[2] != info->uuid[3])
3771 rv = -1;
3772 else
3773 rv = 0;
6e46bf34 3774 }
81a5b4f5
N		 3775 if (rv == 0)
3776 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3777 } else if (strcmp(update, "assemble") == 0)
3778 rv = 0;
3779 else
1e2b2765 3780 rv = -1;
f352c545 3781
6e46bf34
DW		 3782 /* successful update? recompute checksum */
3783 if (rv == 0)
3784 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3785
3786 return rv;
cdddbdbc
DW		 3787}
3788
c2c087e6 3789static size_t disks_to_mpb_size(int disks)
cdddbdbc 3790{
c2c087e6 3791 size_t size;
cdddbdbc 3792
c2c087e6
DW		 3793 size = sizeof(struct imsm_super);
3794 size += (disks - 1) * sizeof(struct imsm_disk);
3795 size += 2 * sizeof(struct imsm_dev);
3796 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3797 size += (4 - 2) * sizeof(struct imsm_map);
3798 /* 4 possible disk_ord_tbl's */
3799 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3800 /* maximum bbm log */
3801 size += sizeof(struct bbm_log);
c2c087e6
DW		 3802
3803 return size;
3804}
3805
387fcd59
N		 3806static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3807 unsigned long long data_offset)
c2c087e6
DW		 3808{
3809 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3810 return 0;
3811
3812 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3813}
3814
ba2de7ba
DW		 3815static void free_devlist(struct intel_super *super)
3816{
3817 struct intel_dev *dv;
3818
3819 while (super->devlist) {
3820 dv = super->devlist->next;
3821 free(super->devlist->dev);
3822 free(super->devlist);
3823 super->devlist = dv;
3824 }
3825}
3826
3827static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3828{
3829 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3830}
3831
cdddbdbc
DW		 3832static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3833{
3834 /*
3835 * return:
3836 * 0 same, or first was empty, and second was copied
3837 * 1 second had wrong number
3838 * 2 wrong uuid
3839 * 3 wrong other info
3840 */
3841 struct intel_super *first = st->sb;
3842 struct intel_super *sec = tst->sb;
3843
5d500228
N		 3844 if (!first) {
3845 st->sb = tst->sb;
3846 tst->sb = NULL;
3847 return 0;
3848 }
8603ea6f
LM		 3849 /* in platform dependent environment test if the disks
3850 * use the same Intel hba
cb8f6859 3851 * If not on Intel hba at all, allow anything.
8603ea6f 3852 */
6b781d33
AP		 3853 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3854 if (first->hba->type != sec->hba->type) {
8603ea6f 3855 fprintf(stderr,
6b781d33
AP		 3856 "HBAs of devices do not match %s != %s\n",
3857 get_sys_dev_type(first->hba->type),
3858 get_sys_dev_type(sec->hba->type));
3859 return 3;
3860 }
3861 if (first->orom != sec->orom) {
3862 fprintf(stderr,
3863 "HBAs of devices do not match %s != %s\n",
3864 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3865 return 3;
3866 }
3867 }
cdddbdbc 3868
d23fe947
DW		 3869 /* if an anchor does not have num_raid_devs set then it is a free
3870 * floating spare
3871 */
3872 if (first->anchor->num_raid_devs > 0 &&
3873 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3874 /* Determine if these disks might ever have been
3875 * related. Further disambiguation can only take place
3876 * in load_super_imsm_all
3877 */
3878 __u32 first_family = first->anchor->orig_family_num;
3879 __u32 sec_family = sec->anchor->orig_family_num;
3880
f796af5d
DW		 3881 if (memcmp(first->anchor->sig, sec->anchor->sig,
3882 MAX_SIGNATURE_LENGTH) != 0)
3883 return 3;
3884
a2b97981
DW		 3885 if (first_family == 0)
3886 first_family = first->anchor->family_num;
3887 if (sec_family == 0)
3888 sec_family = sec->anchor->family_num;
3889
3890 if (first_family != sec_family)
d23fe947 3891 return 3;
f796af5d 3892
d23fe947 3893 }
cdddbdbc 3894
3e372e5a
DW		 3895 /* if 'first' is a spare promote it to a populated mpb with sec's
3896 * family number
3897 */
3898 if (first->anchor->num_raid_devs == 0 &&
3899 sec->anchor->num_raid_devs > 0) {
78d30f94 3900 int i;
ba2de7ba
DW		 3901 struct intel_dev *dv;
3902 struct imsm_dev *dev;
78d30f94
DW		 3903
3904 /* we need to copy raid device info from sec if an allocation
3905 * fails here we don't associate the spare
3906 */
3907 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3908 dv = xmalloc(sizeof(*dv));
3909 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3910 dv->dev = dev;
3911 dv->index = i;
3912 dv->next = first->devlist;
3913 first->devlist = dv;
78d30f94 3914 }
709743c5 3915 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3916 /* allocation failure */
3917 free_devlist(first);
e12b3daa 3918 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3919 return 3;
78d30f94 3920 }
3e372e5a 3921 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3922 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3923 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3924 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3925 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3926 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3927 }
3928
cdddbdbc
DW		 3929 return 0;
3930}
3931
0030e8d6
DW		 3932static void fd2devname(int fd, char *name)
3933{
3934 struct stat st;
3935 char path[256];
33a6535d 3936 char dname[PATH_MAX];
0030e8d6
DW		 3937 char *nm;
3938 int rv;
3939
3940 name[0] = '\0';
3941 if (fstat(fd, &st) != 0)
3942 return;
3943 sprintf(path, "/sys/dev/block/%d:%d",
3944 major(st.st_rdev), minor(st.st_rdev));
3945
9cf014ec 3946 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3947 if (rv <= 0)
3948 return;
9587c373 3949
0030e8d6
DW		 3950 dname[rv] = '\0';
3951 nm = strrchr(dname, '/');
7897de29
JS		 3952 if (nm) {
3953 nm++;
3954 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3955 }
0030e8d6
DW		 3956}
3957
21e9380b
AP		 3958static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3959{
3960 char path[60];
3961 char *name = fd2kname(fd);
3962
3963 if (!name)
3964 return 1;
3965
3966 if (strncmp(name, "nvme", 4) != 0)
3967 return 1;
3968
3969 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3970
3971 return load_sys(path, buf, buf_len);
3972}
3973
cdddbdbc
DW		 3974extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3975
3976static int imsm_read_serial(int fd, char *devname,
6da53c0e 3977 __u8 *serial, size_t serial_buf_len)
cdddbdbc 3978{
21e9380b 3979 char buf[50];
cdddbdbc 3980 int rv;
6da53c0e 3981 size_t len;
316e2bf4
DW		 3982 char *dest;
3983 char *src;
21e9380b
AP		 3984 unsigned int i;
3985
3986 memset(buf, 0, sizeof(buf));
cdddbdbc 3987
21e9380b 3988 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3989
21e9380b
AP		 3990 if (rv)
3991 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3992
40ebbb9c 3993 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3994 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3995 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3996 return 0;
3997 }
3998
cdddbdbc
DW		 3999 if (rv != 0) {
4000 if (devname)
e7b84f9d
N		 4001 pr_err("Failed to retrieve serial for %s\n",
4002 devname);
cdddbdbc
DW		 4003 return rv;
4004 }
4005
316e2bf4
DW		 4006 /* trim all whitespace and non-printable characters and convert
4007 * ':' to ';'
4008 */
21e9380b
AP		 4009 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4010 src = &buf[i];
316e2bf4
DW		 4011 if (*src > 0x20) {
4012 /* ':' is reserved for use in placeholder serial
4013 * numbers for missing disks
4014 */
4015 if (*src == ':')
4016 *dest++ = ';';
4017 else
4018 *dest++ = *src;
4019 }
4020 }
21e9380b
AP		 4021 len = dest - buf;
4022 dest = buf;
316e2bf4 4023
6da53c0e
BK		 4024 if (len > serial_buf_len) {
4025 /* truncate leading characters */
4026 dest += len - serial_buf_len;
4027 len = serial_buf_len;
316e2bf4 4028 }
5c3db629 4029
6da53c0e 4030 memset(serial, 0, serial_buf_len);
316e2bf4 4031 memcpy(serial, dest, len);
cdddbdbc
DW		 4032
4033 return 0;
4034}
4035
1f24f035
DW		 4036static int serialcmp(__u8 *s1, __u8 *s2)
4037{
4038 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4039}
4040
4041static void serialcpy(__u8 *dest, __u8 *src)
4042{
4043 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4044}
4045
54c2c1ea
DW		 4046static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4047{
4048 struct dl *dl;
4049
4050 for (dl = super->disks; dl; dl = dl->next)
4051 if (serialcmp(dl->serial, serial) == 0)
4052 break;
4053
4054 return dl;
4055}
4056
a2b97981
DW		 4057static struct imsm_disk *
4058__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4059{
4060 int i;
4061
4062 for (i = 0; i < mpb->num_disks; i++) {
4063 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4064
4065 if (serialcmp(disk->serial, serial) == 0) {
4066 if (idx)
4067 *idx = i;
4068 return disk;
4069 }
4070 }
4071
4072 return NULL;
4073}
4074
cdddbdbc
DW		 4075static int
4076load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4077{
a2b97981 4078 struct imsm_disk *disk;
cdddbdbc
DW		 4079 struct dl *dl;
4080 struct stat stb;
cdddbdbc 4081 int rv;
a2b97981 4082 char name[40];
d23fe947
DW		 4083 __u8 serial[MAX_RAID_SERIAL_LEN];
4084
6da53c0e 4085 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4086
4087 if (rv != 0)
4088 return 2;
4089
503975b9 4090 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4091
a2b97981
DW		 4092 fstat(fd, &stb);
4093 dl->major = major(stb.st_rdev);
4094 dl->minor = minor(stb.st_rdev);
4095 dl->next = super->disks;
4096 dl->fd = keep_fd ? fd : -1;
4097 assert(super->disks == NULL);
4098 super->disks = dl;
4099 serialcpy(dl->serial, serial);
4100 dl->index = -2;
4101 dl->e = NULL;
4102 fd2devname(fd, name);
4103 if (devname)
503975b9 4104 dl->devname = xstrdup(devname);
a2b97981 4105 else
503975b9 4106 dl->devname = xstrdup(name);
cdddbdbc 4107
d23fe947 4108 /* look up this disk's index in the current anchor */
a2b97981
DW		 4109 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4110 if (disk) {
4111 dl->disk = *disk;
4112 /* only set index on disks that are a member of a
4113 * populated contianer, i.e. one with raid_devs
4114 */
4115 if (is_failed(&dl->disk))
3f6efecc 4116 dl->index = -2;
2432ce9b 4117 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4118 dl->index = -1;
3f6efecc
DW		 4119 }
4120
949c47a0
DW		 4121 return 0;
4122}
4123
0c046afd
DW		 4124/* When migrating map0 contains the 'destination' state while map1
4125 * contains the current state. When not migrating map0 contains the
4126 * current state. This routine assumes that map[0].map_state is set to
4127 * the current array state before being called.
4128 *
4129 * Migration is indicated by one of the following states
4130 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4131 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4132 * map1state=unitialized)
1484e727 4133 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4134 * map1state=normal)
e3bba0e0 4135 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4136 * map1state=degraded)
8e59f3d8
AK		 4137 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4138 * map1state=normal)
0c046afd 4139 */
8e59f3d8
AK		 4140static void migrate(struct imsm_dev *dev, struct intel_super *super,
4141 __u8 to_state, int migr_type)
3393c6af 4142{
0c046afd 4143 struct imsm_map *dest;
238c0a71 4144 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4145
0c046afd 4146 dev->vol.migr_state = 1;
1484e727 4147 set_migr_type(dev, migr_type);
f8f603f1 4148 dev->vol.curr_migr_unit = 0;
238c0a71 4149 dest = get_imsm_map(dev, MAP_1);
0c046afd 4150
0556e1a2 4151 /* duplicate and then set the target end state in map[0] */
3393c6af 4152 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4153 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4154 __u32 ord;
4155 int i;
4156
4157 for (i = 0; i < src->num_members; i++) {
4158 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4159 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4160 }
4161 }
4162
8e59f3d8
AK		 4163 if (migr_type == MIGR_GEN_MIGR)
4164 /* Clear migration record */
4165 memset(super->migr_rec, 0, sizeof(struct migr_record));
4166
0c046afd 4167 src->map_state = to_state;
949c47a0 4168}
f8f603f1 4169
809da78e
AK		 4170static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4171 __u8 map_state)
f8f603f1 4172{
238c0a71
AK		 4173 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4174 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4175 MAP_0 : MAP_1);
28bce06f 4176 int i, j;
0556e1a2
DW		 4177
4178 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4179 * completed in the last migration.
4180 *
28bce06f 4181 * FIXME add support for raid-level-migration
0556e1a2 4182 */
089f9d79
JS		 4183 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4184 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4185 /* when final map state is other than expected
4186 * merge maps (not for migration)
4187 */
4188 int failed;
4189
4190 for (i = 0; i < prev->num_members; i++)
4191 for (j = 0; j < map->num_members; j++)
4192 /* during online capacity expansion
4193 * disks position can be changed
4194 * if takeover is used
4195 */
4196 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4197 ord_to_idx(prev->disk_ord_tbl[i])) {
4198 map->disk_ord_tbl[j] |=
4199 prev->disk_ord_tbl[i];
4200 break;
4201 }
4202 failed = imsm_count_failed(super, dev, MAP_0);
4203 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4204 }
f8f603f1
DW		 4205
4206 dev->vol.migr_state = 0;
ea672ee1 4207 set_migr_type(dev, 0);
f8f603f1
DW		 4208 dev->vol.curr_migr_unit = 0;
4209 map->map_state = map_state;
4210}
949c47a0
DW		 4211
4212static int parse_raid_devices(struct intel_super *super)
4213{
4214 int i;
4215 struct imsm_dev *dev_new;
4d7b1503 4216 size_t len, len_migr;
401d313b 4217 size_t max_len = 0;
4d7b1503
DW		 4218 size_t space_needed = 0;
4219 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4220
4221 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4222 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4223 struct intel_dev *dv;
949c47a0 4224
4d7b1503
DW		 4225 len = sizeof_imsm_dev(dev_iter, 0);
4226 len_migr = sizeof_imsm_dev(dev_iter, 1);
4227 if (len_migr > len)
4228 space_needed += len_migr - len;
ca9de185 4229
503975b9 4230 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4231 if (max_len < len_migr)
4232 max_len = len_migr;
4233 if (max_len > len_migr)
4234 space_needed += max_len - len_migr;
503975b9 4235 dev_new = xmalloc(max_len);
949c47a0 4236 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4237 dv->dev = dev_new;
4238 dv->index = i;
4239 dv->next = super->devlist;
4240 super->devlist = dv;
949c47a0 4241 }
cdddbdbc 4242
4d7b1503
DW		 4243 /* ensure that super->buf is large enough when all raid devices
4244 * are migrating
4245 */
4246 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4247 void *buf;
4248
f36a9ecd
PB		 4249 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4250 super->sector_size);
4251 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4252 return 1;
4253
1f45a8ad
DW		 4254 memcpy(buf, super->buf, super->len);
4255 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4256 free(super->buf);
4257 super->buf = buf;
4258 super->len = len;
4259 }
ca9de185 4260
bbab0940
TM		 4261 super->extra_space += space_needed;
4262
cdddbdbc
DW		 4263 return 0;
4264}
4265
e2f41b2c
AK		 4266/*******************************************************************************
4267 * Function: check_mpb_migr_compatibility
4268 * Description: Function checks for unsupported migration features:
4269 * - migration optimization area (pba_of_lba0)
4270 * - descending reshape (ascending_migr)
4271 * Parameters:
4272 * super : imsm metadata information
4273 * Returns:
4274 * 0 : migration is compatible
4275 * -1 : migration is not compatible
4276 ******************************************************************************/
4277int check_mpb_migr_compatibility(struct intel_super *super)
4278{
4279 struct imsm_map *map0, *map1;
4280 struct migr_record *migr_rec = super->migr_rec;
4281 int i;
4282
4283 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4284 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4285
4286 if (dev_iter &&
4287 dev_iter->vol.migr_state == 1 &&
4288 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4289 /* This device is migrating */
238c0a71
AK		 4290 map0 = get_imsm_map(dev_iter, MAP_0);
4291 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4292 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4293 /* migration optimization area was used */
4294 return -1;
fc54fe7a
JS		 4295 if (migr_rec->ascending_migr == 0 &&
4296 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4297 /* descending reshape not supported yet */
4298 return -1;
4299 }
4300 }
4301 return 0;
4302}
4303
d23fe947 4304static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4305
cdddbdbc 4306/* load_imsm_mpb - read matrix metadata
f2f5c343 4307 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4308 */
4309static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4310{
4311 unsigned long long dsize;
cdddbdbc 4312 unsigned long long sectors;
f36a9ecd 4313 unsigned int sector_size = super->sector_size;
cdddbdbc 4314 struct stat;
6416d527 4315 struct imsm_super *anchor;
cdddbdbc
DW		 4316 __u32 check_sum;
4317
cdddbdbc 4318 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4319 if (dsize < 2*sector_size) {
64436f06 4320 if (devname)
e7b84f9d
N		 4321 pr_err("%s: device to small for imsm\n",
4322 devname);
64436f06
N		 4323 return 1;
4324 }
cdddbdbc 4325
f36a9ecd 4326 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4327 if (devname)
e7b84f9d
N		 4328 pr_err("Cannot seek to anchor block on %s: %s\n",
4329 devname, strerror(errno));
cdddbdbc
DW		 4330 return 1;
4331 }
4332
f36a9ecd 4333 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4334 if (devname)
7a862a02 4335 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4336 return 1;
4337 }
466070ad 4338 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4339 if (devname)
e7b84f9d
N		 4340 pr_err("Cannot read anchor block on %s: %s\n",
4341 devname, strerror(errno));
6416d527 4342 free(anchor);
cdddbdbc
DW		 4343 return 1;
4344 }
4345
6416d527 4346 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4347 if (devname)
e7b84f9d 4348 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4349 free(anchor);
cdddbdbc
DW		 4350 return 2;
4351 }
4352
d23fe947 4353 __free_imsm(super, 0);
f2f5c343
LM		 4354 /* reload capability and hba */
4355
4356 /* capability and hba must be updated with new super allocation */
d424212e 4357 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4358 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4359 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4360 if (devname)
e7b84f9d
N		 4361 pr_err("unable to allocate %zu byte mpb buffer\n",
4362 super->len);
6416d527 4363 free(anchor);
cdddbdbc
DW		 4364 return 2;
4365 }
f36a9ecd 4366 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4367
f36a9ecd 4368 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4369 free(anchor);
8e59f3d8 4370
85337573
AO		 4371 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4372 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4373 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4374 free(super->buf);
4375 return 2;
4376 }
51d83f5d 4377 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4378
949c47a0 4379 if (!sectors) {
ecf45690
DW		 4380 check_sum = __gen_imsm_checksum(super->anchor);
4381 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4382 if (devname)
e7b84f9d
N		 4383 pr_err("IMSM checksum %x != %x on %s\n",
4384 check_sum,
4385 __le32_to_cpu(super->anchor->check_sum),
4386 devname);
ecf45690
DW		 4387 return 2;
4388 }
4389
a2b97981 4390 return 0;
949c47a0 4391 }
cdddbdbc
DW		 4392
4393 /* read the extended mpb */
f36a9ecd 4394 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4395 if (devname)
e7b84f9d
N		 4396 pr_err("Cannot seek to extended mpb on %s: %s\n",
4397 devname, strerror(errno));
cdddbdbc
DW		 4398 return 1;
4399 }
4400
f36a9ecd
PB		 4401 if ((unsigned int)read(fd, super->buf + sector_size,
4402 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4403 if (devname)
e7b84f9d
N		 4404 pr_err("Cannot read extended mpb on %s: %s\n",
4405 devname, strerror(errno));
cdddbdbc
DW		 4406 return 2;
4407 }
4408
949c47a0
DW		 4409 check_sum = __gen_imsm_checksum(super->anchor);
4410 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4411 if (devname)
e7b84f9d
N		 4412 pr_err("IMSM checksum %x != %x on %s\n",
4413 check_sum, __le32_to_cpu(super->anchor->check_sum),
4414 devname);
db575f3b 4415 return 3;
cdddbdbc
DW		 4416 }
4417
a2b97981
DW		 4418 return 0;
4419}
4420
8e59f3d8
AK		 4421static int read_imsm_migr_rec(int fd, struct intel_super *super);
4422
97f81ee2
CA		 4423/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4424static void clear_hi(struct intel_super *super)
4425{
4426 struct imsm_super *mpb = super->anchor;
4427 int i, n;
4428 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4429 return;
4430 for (i = 0; i < mpb->num_disks; ++i) {
4431 struct imsm_disk *disk = &mpb->disk[i];
4432 disk->total_blocks_hi = 0;
4433 }
4434 for (i = 0; i < mpb->num_raid_devs; ++i) {
4435 struct imsm_dev *dev = get_imsm_dev(super, i);
4436 if (!dev)
4437 return;
4438 for (n = 0; n < 2; ++n) {
4439 struct imsm_map *map = get_imsm_map(dev, n);
4440 if (!map)
4441 continue;
4442 map->pba_of_lba0_hi = 0;
4443 map->blocks_per_member_hi = 0;
4444 map->num_data_stripes_hi = 0;
4445 }
4446 }
4447}
4448
a2b97981
DW		 4449static int
4450load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4451{
4452 int err;
4453
4454 err = load_imsm_mpb(fd, super, devname);
4455 if (err)
4456 return err;
f36a9ecd
PB		 4457 if (super->sector_size == 4096)
4458 convert_from_4k(super);
a2b97981
DW		 4459 err = load_imsm_disk(fd, super, devname, keep_fd);
4460 if (err)
4461 return err;
4462 err = parse_raid_devices(super);
8d67477f
TM		 4463 if (err)
4464 return err;
4465 err = load_bbm_log(super);
97f81ee2 4466 clear_hi(super);
a2b97981 4467 return err;
cdddbdbc
DW		 4468}
4469
ae6aad82
DW		 4470static void __free_imsm_disk(struct dl *d)
4471{
4472 if (d->fd >= 0)
4473 close(d->fd);
4474 if (d->devname)
4475 free(d->devname);
0dcecb2e
DW		 4476 if (d->e)
4477 free(d->e);
ae6aad82
DW		 4478 free(d);
4479
4480}
1a64be56 4481
cdddbdbc
DW		 4482static void free_imsm_disks(struct intel_super *super)
4483{
47ee5a45 4484 struct dl *d;
cdddbdbc 4485
47ee5a45
DW		 4486 while (super->disks) {
4487 d = super->disks;
cdddbdbc 4488 super->disks = d->next;
ae6aad82 4489 __free_imsm_disk(d);
cdddbdbc 4490 }
cb82edca
AK		 4491 while (super->disk_mgmt_list) {
4492 d = super->disk_mgmt_list;
4493 super->disk_mgmt_list = d->next;
4494 __free_imsm_disk(d);
4495 }
47ee5a45
DW		 4496 while (super->missing) {
4497 d = super->missing;
4498 super->missing = d->next;
4499 __free_imsm_disk(d);
4500 }
4501
cdddbdbc
DW		 4502}
4503
9ca2c81c 4504/* free all the pieces hanging off of a super pointer */
d23fe947 4505static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4506{
88654014
LM		 4507 struct intel_hba *elem, *next;
4508
9ca2c81c 4509 if (super->buf) {
949c47a0 4510 free(super->buf);
9ca2c81c
DW		 4511 super->buf = NULL;
4512 }
f2f5c343
LM		 4513 /* unlink capability description */
4514 super->orom = NULL;
8e59f3d8
AK		 4515 if (super->migr_rec_buf) {
4516 free(super->migr_rec_buf);
4517 super->migr_rec_buf = NULL;
4518 }
d23fe947
DW		 4519 if (free_disks)
4520 free_imsm_disks(super);
ba2de7ba 4521 free_devlist(super);
88654014
LM		 4522 elem = super->hba;
4523 while (elem) {
4524 if (elem->path)
4525 free((void *)elem->path);
4526 next = elem->next;
4527 free(elem);
4528 elem = next;
88c32bb1 4529 }
8d67477f
TM		 4530 if (super->bbm_log)
4531 free(super->bbm_log);
88654014 4532 super->hba = NULL;
cdddbdbc
DW		 4533}
4534
9ca2c81c
DW		 4535static void free_imsm(struct intel_super *super)
4536{
d23fe947 4537 __free_imsm(super, 1);
928f1424 4538 free(super->bb.entries);
9ca2c81c
DW		 4539 free(super);
4540}
cdddbdbc
DW		 4541
4542static void free_super_imsm(struct supertype *st)
4543{
4544 struct intel_super *super = st->sb;
4545
4546 if (!super)
4547 return;
4548
4549 free_imsm(super);
4550 st->sb = NULL;
4551}
4552
49133e57 4553static struct intel_super *alloc_super(void)
c2c087e6 4554{
503975b9 4555 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4556
503975b9
N		 4557 super->current_vol = -1;
4558 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4559
4560 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4561 sizeof(struct md_bb_entry));
4562 if (!super->bb.entries) {
4563 free(super);
4564 return NULL;
4565 }
4566
c2c087e6
DW		 4567 return super;
4568}
4569
f0f5a016
LM		 4570/*
4571 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4572 */
d424212e 4573static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4574{
4575 struct sys_dev *hba_name;
4576 int rv = 0;
4577
3a30e28e
MT		 4578 if (fd >= 0 && test_partition(fd)) {
4579 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4580 devname);
4581 return 1;
4582 }
089f9d79 4583 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4584 super->orom = NULL;
f0f5a016
LM		 4585 super->hba = NULL;
4586 return 0;
4587 }
4588 hba_name = find_disk_attached_hba(fd, NULL);
4589 if (!hba_name) {
d424212e 4590 if (devname)
e7b84f9d
N		 4591 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4592 devname);
f0f5a016
LM		 4593 return 1;
4594 }
4595 rv = attach_hba_to_super(super, hba_name);
4596 if (rv == 2) {
d424212e
N		 4597 if (devname) {
4598 struct intel_hba *hba = super->hba;
f0f5a016 4599
60f0f54d
PB		 4600 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4601 " but the container is assigned to Intel(R) %s %s (",
d424212e 4602 devname,
614902f6 4603 get_sys_dev_type(hba_name->type),
60f0f54d 4604 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4605 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4606 get_sys_dev_type(super->hba->type),
4607 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4608
f0f5a016
LM		 4609 while (hba) {
4610 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4611 if (hba->next)
4612 fprintf(stderr, ", ");
4613 hba = hba->next;
4614 }
6b781d33 4615 fprintf(stderr, ").\n"
cca67208 4616 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4617 }
f0f5a016
LM		 4618 return 2;
4619 }
6b781d33 4620 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4621 if (!super->orom)
4622 return 3;
614902f6 4623
f0f5a016
LM		 4624 return 0;
4625}
4626
47ee5a45
DW		 4627/* find_missing - helper routine for load_super_imsm_all that identifies
4628 * disks that have disappeared from the system. This routine relies on
4629 * the mpb being uptodate, which it is at load time.
4630 */
4631static int find_missing(struct intel_super *super)
4632{
4633 int i;
4634 struct imsm_super *mpb = super->anchor;
4635 struct dl *dl;
4636 struct imsm_disk *disk;
47ee5a45
DW		 4637
4638 for (i = 0; i < mpb->num_disks; i++) {
4639 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4640 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4641 if (dl)
4642 continue;
47ee5a45 4643
503975b9 4644 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4645 dl->major = 0;
4646 dl->minor = 0;
4647 dl->fd = -1;
503975b9 4648 dl->devname = xstrdup("missing");
47ee5a45
DW		 4649 dl->index = i;
4650 serialcpy(dl->serial, disk->serial);
4651 dl->disk = *disk;
689c9bf3 4652 dl->e = NULL;
47ee5a45
DW		 4653 dl->next = super->missing;
4654 super->missing = dl;
4655 }
4656
4657 return 0;
4658}
4659
a2b97981
DW		 4660static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4661{
4662 struct intel_disk *idisk = disk_list;
4663
4664 while (idisk) {
4665 if (serialcmp(idisk->disk.serial, serial) == 0)
4666 break;
4667 idisk = idisk->next;
4668 }
4669
4670 return idisk;
4671}
4672
4673static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4674 struct intel_super *super,
4675 struct intel_disk **disk_list)
4676{
4677 struct imsm_disk *d = &super->disks->disk;
4678 struct imsm_super *mpb = super->anchor;
4679 int i, j;
4680
4681 for (i = 0; i < tbl_size; i++) {
4682 struct imsm_super *tbl_mpb = table[i]->anchor;
4683 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4684
4685 if (tbl_mpb->family_num == mpb->family_num) {
4686 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4687 dprintf("mpb from %d:%d matches %d:%d\n",
4688 super->disks->major,
a2b97981
DW		 4689 super->disks->minor,
4690 table[i]->disks->major,
4691 table[i]->disks->minor);
4692 break;
4693 }
4694
4695 if (((is_configured(d) && !is_configured(tbl_d)) ||
4696 is_configured(d) == is_configured(tbl_d)) &&
4697 tbl_mpb->generation_num < mpb->generation_num) {
4698 /* current version of the mpb is a
4699 * better candidate than the one in
4700 * super_table, but copy over "cross
4701 * generational" status
4702 */
4703 struct intel_disk *idisk;
4704
1ade5cc1
N		 4705 dprintf("mpb from %d:%d replaces %d:%d\n",
4706 super->disks->major,
a2b97981
DW		 4707 super->disks->minor,
4708 table[i]->disks->major,
4709 table[i]->disks->minor);
4710
4711 idisk = disk_list_get(tbl_d->serial, *disk_list);
4712 if (idisk && is_failed(&idisk->disk))
4713 tbl_d->status |= FAILED_DISK;
4714 break;
4715 } else {
4716 struct intel_disk *idisk;
4717 struct imsm_disk *disk;
4718
4719 /* tbl_mpb is more up to date, but copy
4720 * over cross generational status before
4721 * returning
4722 */
4723 disk = __serial_to_disk(d->serial, mpb, NULL);
4724 if (disk && is_failed(disk))
4725 d->status |= FAILED_DISK;
4726
4727 idisk = disk_list_get(d->serial, *disk_list);
4728 if (idisk) {
4729 idisk->owner = i;
4730 if (disk && is_configured(disk))
4731 idisk->disk.status |= CONFIGURED_DISK;
4732 }
4733
1ade5cc1
N		 4734 dprintf("mpb from %d:%d prefer %d:%d\n",
4735 super->disks->major,
a2b97981
DW		 4736 super->disks->minor,
4737 table[i]->disks->major,
4738 table[i]->disks->minor);
4739
4740 return tbl_size;
4741 }
4742 }
4743 }
4744
4745 if (i >= tbl_size)
4746 table[tbl_size++] = super;
4747 else
4748 table[i] = super;
4749
4750 /* update/extend the merged list of imsm_disk records */
4751 for (j = 0; j < mpb->num_disks; j++) {
4752 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4753 struct intel_disk *idisk;
4754
4755 idisk = disk_list_get(disk->serial, *disk_list);
4756 if (idisk) {
4757 idisk->disk.status |= disk->status;
4758 if (is_configured(&idisk->disk) ||
4759 is_failed(&idisk->disk))
4760 idisk->disk.status &= ~(SPARE_DISK);
4761 } else {
503975b9 4762 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4763 idisk->owner = IMSM_UNKNOWN_OWNER;
4764 idisk->disk = *disk;
4765 idisk->next = *disk_list;
4766 *disk_list = idisk;
4767 }
4768
4769 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4770 idisk->owner = i;
4771 }
4772
4773 return tbl_size;
4774}
4775
4776static struct intel_super *
4777validate_members(struct intel_super *super, struct intel_disk *disk_list,
4778 const int owner)
4779{
4780 struct imsm_super *mpb = super->anchor;
4781 int ok_count = 0;
4782 int i;
4783
4784 for (i = 0; i < mpb->num_disks; i++) {
4785 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4786 struct intel_disk *idisk;
4787
4788 idisk = disk_list_get(disk->serial, disk_list);
4789 if (idisk) {
4790 if (idisk->owner == owner ||
4791 idisk->owner == IMSM_UNKNOWN_OWNER)
4792 ok_count++;
4793 else
1ade5cc1
N		 4794 dprintf("'%.16s' owner %d != %d\n",
4795 disk->serial, idisk->owner,
a2b97981
DW		 4796 owner);
4797 } else {
1ade5cc1
N		 4798 dprintf("unknown disk %x [%d]: %.16s\n",
4799 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4800 disk->serial);
4801 break;
4802 }
4803 }
4804
4805 if (ok_count == mpb->num_disks)
4806 return super;
4807 return NULL;
4808}
4809
4810static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4811{
4812 struct intel_super *s;
4813
4814 for (s = super_list; s; s = s->next) {
4815 if (family_num != s->anchor->family_num)
4816 continue;
e12b3daa 4817 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4818 __le32_to_cpu(family_num), s->disks->devname);
4819 }
4820}
4821
4822static struct intel_super *
4823imsm_thunderdome(struct intel_super **super_list, int len)
4824{
4825 struct intel_super *super_table[len];
4826 struct intel_disk *disk_list = NULL;
4827 struct intel_super *champion, *spare;
4828 struct intel_super *s, **del;
4829 int tbl_size = 0;
4830 int conflict;
4831 int i;
4832
4833 memset(super_table, 0, sizeof(super_table));
4834 for (s = *super_list; s; s = s->next)
4835 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4836
4837 for (i = 0; i < tbl_size; i++) {
4838 struct imsm_disk *d;
4839 struct intel_disk *idisk;
4840 struct imsm_super *mpb = super_table[i]->anchor;
4841
4842 s = super_table[i];
4843 d = &s->disks->disk;
4844
4845 /* 'd' must appear in merged disk list for its
4846 * configuration to be valid
4847 */
4848 idisk = disk_list_get(d->serial, disk_list);
4849 if (idisk && idisk->owner == i)
4850 s = validate_members(s, disk_list, i);
4851 else
4852 s = NULL;
4853
4854 if (!s)
1ade5cc1
N		 4855 dprintf("marking family: %#x from %d:%d offline\n",
4856 mpb->family_num,
a2b97981
DW		 4857 super_table[i]->disks->major,
4858 super_table[i]->disks->minor);
4859 super_table[i] = s;
4860 }
4861
4862 /* This is where the mdadm implementation differs from the Windows
4863 * driver which has no strict concept of a container. We can only
4864 * assemble one family from a container, so when returning a prodigal
4865 * array member to this system the code will not be able to disambiguate
4866 * the container contents that should be assembled ("foreign" versus
4867 * "local"). It requires user intervention to set the orig_family_num
4868 * to a new value to establish a new container. The Windows driver in
4869 * this situation fixes up the volume name in place and manages the
4870 * foreign array as an independent entity.
4871 */
4872 s = NULL;
4873 spare = NULL;
4874 conflict = 0;
4875 for (i = 0; i < tbl_size; i++) {
4876 struct intel_super *tbl_ent = super_table[i];
4877 int is_spare = 0;
4878
4879 if (!tbl_ent)
4880 continue;
4881
4882 if (tbl_ent->anchor->num_raid_devs == 0) {
4883 spare = tbl_ent;
4884 is_spare = 1;
4885 }
4886
4887 if (s && !is_spare) {
4888 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4889 conflict++;
4890 } else if (!s && !is_spare)
4891 s = tbl_ent;
4892 }
4893
4894 if (!s)
4895 s = spare;
4896 if (!s) {
4897 champion = NULL;
4898 goto out;
4899 }
4900 champion = s;
4901
4902 if (conflict)
7a862a02 4903 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4904 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4905
4906 /* collect all dl's onto 'champion', and update them to
4907 * champion's version of the status
4908 */
4909 for (s = *super_list; s; s = s->next) {
4910 struct imsm_super *mpb = champion->anchor;
4911 struct dl *dl = s->disks;
4912
4913 if (s == champion)
4914 continue;
4915
5d7b407a
CA		 4916 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4917
a2b97981
DW		 4918 for (i = 0; i < mpb->num_disks; i++) {
4919 struct imsm_disk *disk;
4920
4921 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4922 if (disk) {
4923 dl->disk = *disk;
4924 /* only set index on disks that are a member of
4925 * a populated contianer, i.e. one with
4926 * raid_devs
4927 */
4928 if (is_failed(&dl->disk))
4929 dl->index = -2;
4930 else if (is_spare(&dl->disk))
4931 dl->index = -1;
4932 break;
4933 }
4934 }
4935
4936 if (i >= mpb->num_disks) {
4937 struct intel_disk *idisk;
4938
4939 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4940 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4941 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4942 dl->index = -1;
4943 else {
4944 dl->index = -2;
4945 continue;
4946 }
4947 }
4948
4949 dl->next = champion->disks;
4950 champion->disks = dl;
4951 s->disks = NULL;
4952 }
4953
4954 /* delete 'champion' from super_list */
4955 for (del = super_list; *del; ) {
4956 if (*del == champion) {
4957 *del = (*del)->next;
4958 break;
4959 } else
4960 del = &(*del)->next;
4961 }
4962 champion->next = NULL;
4963
4964 out:
4965 while (disk_list) {
4966 struct intel_disk *idisk = disk_list;
4967
4968 disk_list = disk_list->next;
4969 free(idisk);
4970 }
4971
4972 return champion;
4973}
4974
9587c373
LM		 4975static int
4976get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4977static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4978 int major, int minor, int keep_fd);
ec50f7b6
LM		 4979static int
4980get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4981 int *max, int keep_fd);
4982
cdddbdbc 4983static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4984 char *devname, struct md_list *devlist,
4985 int keep_fd)
cdddbdbc 4986{
a2b97981
DW		 4987 struct intel_super *super_list = NULL;
4988 struct intel_super *super = NULL;
a2b97981 4989 int err = 0;
9587c373 4990 int i = 0;
dab4a513 4991
9587c373
LM		 4992 if (fd >= 0)
4993 /* 'fd' is an opened container */
4994 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4995 else
ec50f7b6
LM		 4996 /* get super block from devlist devices */
4997 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4998 if (err)
1602d52c 4999 goto error;
a2b97981
DW		 5000 /* all mpbs enter, maybe one leaves */
5001 super = imsm_thunderdome(&super_list, i);
5002 if (!super) {
5003 err = 1;
5004 goto error;
cdddbdbc
DW		 5005 }
5006
47ee5a45
DW		 5007 if (find_missing(super) != 0) {
5008 free_imsm(super);
a2b97981
DW		 5009 err = 2;
5010 goto error;
47ee5a45 5011 }
8e59f3d8
AK		 5012
5013 /* load migration record */
5014 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5015 if (err == -1) {
5016 /* migration is in progress,
5017 * but migr_rec cannot be loaded,
5018 */
8e59f3d8
AK		 5019 err = 4;
5020 goto error;
5021 }
e2f41b2c
AK		 5022
5023 /* Check migration compatibility */
089f9d79 5024 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5025 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5026 if (devname)
5027 fprintf(stderr, " on %s\n", devname);
5028 else
5029 fprintf(stderr, " (IMSM).\n");
5030
5031 err = 5;
5032 goto error;
5033 }
5034
a2b97981
DW		 5035 err = 0;
5036
5037 error:
5038 while (super_list) {
5039 struct intel_super *s = super_list;
5040
5041 super_list = super_list->next;
5042 free_imsm(s);
5043 }
9587c373 5044
a2b97981
DW		 5045 if (err)
5046 return err;
f7e7067b 5047
cdddbdbc 5048 *sbp = super;
9587c373 5049 if (fd >= 0)
4dd2df09 5050 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5051 else
4dd2df09 5052 st->container_devnm[0] = 0;
a2b97981 5053 if (err == 0 && st->ss == NULL) {
bf5a934a 5054 st->ss = &super_imsm;
cdddbdbc
DW		 5055 st->minor_version = 0;
5056 st->max_devs = IMSM_MAX_DEVICES;
5057 }
cdddbdbc
DW		 5058 return 0;
5059}
2b959fbf 5060
ec50f7b6
LM		 5061static int
5062get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5063 int *max, int keep_fd)
5064{
5065 struct md_list *tmpdev;
5066 int err = 0;
5067 int i = 0;
9587c373 5068
ec50f7b6
LM		 5069 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5070 if (tmpdev->used != 1)
5071 continue;
5072 if (tmpdev->container == 1) {
ca9de185 5073 int lmax = 0;
ec50f7b6
LM		 5074 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5075 if (fd < 0) {
e7b84f9d 5076 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5077 tmpdev->devname, strerror(errno));
5078 err = 8;
5079 goto error;
5080 }
5081 err = get_sra_super_block(fd, super_list,
5082 tmpdev->devname, &lmax,
5083 keep_fd);
5084 i += lmax;
5085 close(fd);
5086 if (err) {
5087 err = 7;
5088 goto error;
5089 }
5090 } else {
5091 int major = major(tmpdev->st_rdev);
5092 int minor = minor(tmpdev->st_rdev);
5093 err = get_super_block(super_list,
4dd2df09 5094 NULL,
ec50f7b6
LM		 5095 tmpdev->devname,
5096 major, minor,
5097 keep_fd);
5098 i++;
5099 if (err) {
5100 err = 6;
5101 goto error;
5102 }
5103 }
5104 }
5105 error:
5106 *max = i;
5107 return err;
5108}
9587c373 5109
4dd2df09 5110static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5111 int major, int minor, int keep_fd)
5112{
594dc1b8 5113 struct intel_super *s;
9587c373
LM		 5114 char nm[32];
5115 int dfd = -1;
9587c373
LM		 5116 int err = 0;
5117 int retry;
5118
5119 s = alloc_super();
5120 if (!s) {
5121 err = 1;
5122 goto error;
5123 }
5124
5125 sprintf(nm, "%d:%d", major, minor);
5126 dfd = dev_open(nm, O_RDWR);
5127 if (dfd < 0) {
5128 err = 2;
5129 goto error;
5130 }
5131
fa7bb6f8 5132 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5133 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5134 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5135
5136 /* retry the load if we might have raced against mdmon */
4dd2df09 5137 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5138 for (retry = 0; retry < 3; retry++) {
5139 usleep(3000);
5140 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5141 if (err != 3)
5142 break;
5143 }
5144 error:
5145 if (!err) {
5146 s->next = *super_list;
5147 *super_list = s;
5148 } else {
5149 if (s)
8d67477f 5150 free_imsm(s);
36614e95 5151 if (dfd >= 0)
9587c373
LM		 5152 close(dfd);
5153 }
089f9d79 5154 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5155 close(dfd);
5156 return err;
5157
5158}
5159
5160static int
5161get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5162{
5163 struct mdinfo *sra;
4dd2df09 5164 char *devnm;
9587c373
LM		 5165 struct mdinfo *sd;
5166 int err = 0;
5167 int i = 0;
4dd2df09 5168 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5169 if (!sra)
5170 return 1;
5171
5172 if (sra->array.major_version != -1 ||
5173 sra->array.minor_version != -2 ||
5174 strcmp(sra->text_version, "imsm") != 0) {
5175 err = 1;
5176 goto error;
5177 }
5178 /* load all mpbs */
4dd2df09 5179 devnm = fd2devnm(fd);
9587c373 5180 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5181 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5182 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5183 err = 7;
5184 goto error;
5185 }
5186 }
5187 error:
5188 sysfs_free(sra);
5189 *max = i;
5190 return err;
5191}
5192
2b959fbf
N		 5193static int load_container_imsm(struct supertype *st, int fd, char *devname)
5194{
ec50f7b6 5195 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5196}
cdddbdbc
DW		 5197
5198static int load_super_imsm(struct supertype *st, int fd, char *devname)
5199{
5200 struct intel_super *super;
5201 int rv;
8a3544f8 5202 int retry;
cdddbdbc 5203
357ac106 5204 if (test_partition(fd))
691c6ee1
N		 5205 /* IMSM not allowed on partitions */
5206 return 1;
5207
37424f13
DW		 5208 free_super_imsm(st);
5209
49133e57 5210 super = alloc_super();
fa7bb6f8 5211 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5212 if (!super)
5213 return 1;
ea2bc72b
LM		 5214 /* Load hba and capabilities if they exist.
5215 * But do not preclude loading metadata in case capabilities or hba are
5216 * non-compliant and ignore_hw_compat is set.
5217 */
d424212e 5218 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5219 /* no orom/efi or non-intel hba of the disk */
089f9d79 5220 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5221 if (devname)
e7b84f9d 5222 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5223 free_imsm(super);
5224 return 2;
5225 }
a2b97981 5226 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5227
8a3544f8
AP		 5228 /* retry the load if we might have raced against mdmon */
5229 if (rv == 3) {
f96b1302
AP		 5230 struct mdstat_ent *mdstat = NULL;
5231 char *name = fd2kname(fd);
5232
5233 if (name)
5234 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5235
5236 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5237 for (retry = 0; retry < 3; retry++) {
5238 usleep(3000);
5239 rv = load_and_parse_mpb(fd, super, devname, 0);
5240 if (rv != 3)
5241 break;
5242 }
5243 }
5244
5245 free_mdstat(mdstat);
5246 }
5247
cdddbdbc
DW		 5248 if (rv) {
5249 if (devname)
7a862a02 5250 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5251 free_imsm(super);
5252 return rv;
5253 }
5254
5255 st->sb = super;
5256 if (st->ss == NULL) {
5257 st->ss = &super_imsm;
5258 st->minor_version = 0;
5259 st->max_devs = IMSM_MAX_DEVICES;
5260 }
8e59f3d8
AK		 5261
5262 /* load migration record */
2e062e82
AK		 5263 if (load_imsm_migr_rec(super, NULL) == 0) {
5264 /* Check for unsupported migration features */
5265 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5266 pr_err("Unsupported migration detected");
2e062e82
AK		 5267 if (devname)
5268 fprintf(stderr, " on %s\n", devname);
5269 else
5270 fprintf(stderr, " (IMSM).\n");
5271 return 3;
5272 }
e2f41b2c
AK		 5273 }
5274
cdddbdbc
DW		 5275 return 0;
5276}
5277
ef6ffade
DW		 5278static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5279{
5280 if (info->level == 1)
5281 return 128;
5282 return info->chunk_size >> 9;
5283}
5284
5551b113
CA		 5285static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5286 unsigned long long size)
fcfd9599 5287{
4025c288 5288 if (info->level == 1)
5551b113 5289 return size * 2;
4025c288 5290 else
5551b113 5291 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5292}
5293
4d1313e9
DW		 5294static void imsm_update_version_info(struct intel_super *super)
5295{
5296 /* update the version and attributes */
5297 struct imsm_super *mpb = super->anchor;
5298 char *version;
5299 struct imsm_dev *dev;
5300 struct imsm_map *map;
5301 int i;
5302
5303 for (i = 0; i < mpb->num_raid_devs; i++) {
5304 dev = get_imsm_dev(super, i);
238c0a71 5305 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5306 if (__le32_to_cpu(dev->size_high) > 0)
5307 mpb->attributes |= MPB_ATTRIB_2TB;
5308
5309 /* FIXME detect when an array spans a port multiplier */
5310 #if 0
5311 mpb->attributes |= MPB_ATTRIB_PM;
5312 #endif
5313
5314 if (mpb->num_raid_devs > 1 ||
5315 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5316 version = MPB_VERSION_ATTRIBS;
5317 switch (get_imsm_raid_level(map)) {
5318 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5319 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5320 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5321 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5322 }
5323 } else {
5324 if (map->num_members >= 5)
5325 version = MPB_VERSION_5OR6_DISK_ARRAY;
5326 else if (dev->status == DEV_CLONE_N_GO)
5327 version = MPB_VERSION_CNG;
5328 else if (get_imsm_raid_level(map) == 5)
5329 version = MPB_VERSION_RAID5;
5330 else if (map->num_members >= 3)
5331 version = MPB_VERSION_3OR4_DISK_ARRAY;
5332 else if (get_imsm_raid_level(map) == 1)
5333 version = MPB_VERSION_RAID1;
5334 else
5335 version = MPB_VERSION_RAID0;
5336 }
5337 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5338 }
5339}
5340
aa534678
DW		 5341static int check_name(struct intel_super *super, char *name, int quiet)
5342{
5343 struct imsm_super *mpb = super->anchor;
5344 char *reason = NULL;
9bd99a90
RS		 5345 char *start = name;
5346 size_t len = strlen(name);
aa534678
DW		 5347 int i;
5348
9bd99a90
RS		 5349 if (len > 0) {
5350 while (isspace(start[len - 1]))
5351 start[--len] = 0;
5352 while (*start && isspace(*start))
5353 ++start, --len;
5354 memmove(name, start, len + 1);
5355 }
5356
5357 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5358 reason = "must be 16 characters or less";
9bd99a90
RS		 5359 else if (len == 0)
5360 reason = "must be a non-empty string";
aa534678
DW		 5361
5362 for (i = 0; i < mpb->num_raid_devs; i++) {
5363 struct imsm_dev *dev = get_imsm_dev(super, i);
5364
5365 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5366 reason = "already exists";
5367 break;
5368 }
5369 }
5370
5371 if (reason && !quiet)
e7b84f9d 5372 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5373
5374 return !reason;
5375}
5376
8b353278 5377static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5378 struct shape *s, char *name,
83cd1e97
N		 5379 char *homehost, int *uuid,
5380 long long data_offset)
cdddbdbc 5381{
c2c087e6
DW		 5382 /* We are creating a volume inside a pre-existing container.
5383 * so st->sb is already set.
5384 */
5385 struct intel_super *super = st->sb;
f36a9ecd 5386 unsigned int sector_size = super->sector_size;
949c47a0 5387 struct imsm_super *mpb = super->anchor;
ba2de7ba 5388 struct intel_dev *dv;
c2c087e6
DW		 5389 struct imsm_dev *dev;
5390 struct imsm_vol *vol;
5391 struct imsm_map *map;
5392 int idx = mpb->num_raid_devs;
5393 int i;
760365f9 5394 int namelen;
c2c087e6 5395 unsigned long long array_blocks;
2c092cad 5396 size_t size_old, size_new;
5551b113 5397 unsigned long long num_data_stripes;
b53bfba6
TM		 5398 unsigned int data_disks;
5399 unsigned long long size_per_member;
cdddbdbc 5400
88c32bb1 5401 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5402 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5403 return 0;
5404 }
5405
2c092cad
DW		 5406 /* ensure the mpb is large enough for the new data */
5407 size_old = __le32_to_cpu(mpb->mpb_size);
5408 size_new = disks_to_mpb_size(info->nr_disks);
5409 if (size_new > size_old) {
5410 void *mpb_new;
f36a9ecd 5411 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5412
f36a9ecd 5413 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5414 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5415 return 0;
5416 }
85337573
AO		 5417 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5418 MIGR_REC_BUF_SECTORS*
5419 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5420 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5421 free(super->buf);
5422 free(super);
ea944c8f 5423 free(mpb_new);
8e59f3d8
AK		 5424 return 0;
5425 }
2c092cad
DW		 5426 memcpy(mpb_new, mpb, size_old);
5427 free(mpb);
5428 mpb = mpb_new;
949c47a0 5429 super->anchor = mpb_new;
2c092cad
DW		 5430 mpb->mpb_size = __cpu_to_le32(size_new);
5431 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5432 super->len = size_round;
2c092cad 5433 }
bf5a934a 5434 super->current_vol = idx;
3960e579
DW		 5435
5436 /* handle 'failed_disks' by either:
5437 * a) create dummy disk entries in the table if this the first
5438 * volume in the array. We add them here as this is the only
5439 * opportunity to add them. add_to_super_imsm_volume()
5440 * handles the non-failed disks and continues incrementing
5441 * mpb->num_disks.
5442 * b) validate that 'failed_disks' matches the current number
5443 * of missing disks if the container is populated
d23fe947 5444 */
3960e579 5445 if (super->current_vol == 0) {
d23fe947 5446 mpb->num_disks = 0;
3960e579
DW		 5447 for (i = 0; i < info->failed_disks; i++) {
5448 struct imsm_disk *disk;
5449
5450 mpb->num_disks++;
5451 disk = __get_imsm_disk(mpb, i);
5452 disk->status = CONFIGURED_DISK | FAILED_DISK;
5453 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5454 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5455 "missing:%d", (__u8)i);
3960e579
DW		 5456 }
5457 find_missing(super);
5458 } else {
5459 int missing = 0;
5460 struct dl *d;
5461
5462 for (d = super->missing; d; d = d->next)
5463 missing++;
5464 if (info->failed_disks > missing) {
e7b84f9d 5465 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5466 return 0;
5467 }
5468 }
5a038140 5469
aa534678
DW		 5470 if (!check_name(super, name, 0))
5471 return 0;
503975b9
N		 5472 dv = xmalloc(sizeof(*dv));
5473 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5474 /*
5475 * Explicitly allow truncating to not confuse gcc's
5476 * -Werror=stringop-truncation
5477 */
5478 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5479 memcpy(dev->volume, name, namelen);
e03640bd 5480 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5481 info->layout, info->chunk_size,
b53bfba6
TM		 5482 s->size * BLOCKS_PER_KB);
5483 data_disks = get_data_disks(info->level, info->layout,
5484 info->raid_disks);
5485 array_blocks = round_size_to_mb(array_blocks, data_disks);
5486 size_per_member = array_blocks / data_disks;
979d38be 5487
fcc2c9da 5488 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5489 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5490 vol = &dev->vol;
5491 vol->migr_state = 0;
1484e727 5492 set_migr_type(dev, MIGR_INIT);
3960e579 5493 vol->dirty = !info->state;
f8f603f1 5494 vol->curr_migr_unit = 0;
238c0a71 5495 map = get_imsm_map(dev, MAP_0);
5551b113 5496 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5497 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5498 map->failed_disk_num = ~0;
bf4442ab 5499 if (info->level > 0)
fffaf1ff
N		 5500 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5501 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5502 else
5503 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5504 IMSM_T_STATE_NORMAL;
252d23c0 5505 map->ddf = 1;
ef6ffade
DW		 5506
5507 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5508 free(dev);
5509 free(dv);
7a862a02 5510 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5511 return 0;
5512 }
81062a36
DW		 5513
5514 map->raid_level = info->level;
4d1313e9 5515 if (info->level == 10) {
c2c087e6 5516 map->raid_level = 1;
4d1313e9 5517 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5518 } else if (info->level == 1)
5519 map->num_domains = info->raid_disks;
5520 else
ff596308 5521 map->num_domains = 1;
81062a36 5522
5551b113 5523 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5524 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5525 num_data_stripes /= map->num_domains;
5526 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5527
44490938
MD		 5528 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5529 set_blocks_per_member(map, info_to_blocks_per_member(info,
5530 size_per_member /
5531 BLOCKS_PER_KB));
5532
c2c087e6
DW		 5533 map->num_members = info->raid_disks;
5534 for (i = 0; i < map->num_members; i++) {
5535 /* initialized in add_to_super */
4eb26970 5536 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5537 }
949c47a0 5538 mpb->num_raid_devs++;
2a24dc1b
PB		 5539 mpb->num_raid_devs_created++;
5540 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5541
b7580566 5542 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5543 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5544 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5545 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5546 } else {
5547 free(dev);
5548 free(dv);
5549 pr_err("imsm does not support consistency policy %s\n",
5550 map_num(consistency_policies, s->consistency_policy));
5551 return 0;
5552 }
5553
ba2de7ba
DW		 5554 dv->dev = dev;
5555 dv->index = super->current_vol;
5556 dv->next = super->devlist;
5557 super->devlist = dv;
c2c087e6 5558
4d1313e9
DW		 5559 imsm_update_version_info(super);
5560
c2c087e6 5561 return 1;
cdddbdbc
DW		 5562}
5563
bf5a934a 5564static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5565 struct shape *s, char *name,
83cd1e97
N		 5566 char *homehost, int *uuid,
5567 unsigned long long data_offset)
bf5a934a
DW		 5568{
5569 /* This is primarily called by Create when creating a new array.
5570 * We will then get add_to_super called for each component, and then
5571 * write_init_super called to write it out to each device.
5572 * For IMSM, Create can create on fresh devices or on a pre-existing
5573 * array.
5574 * To create on a pre-existing array a different method will be called.
5575 * This one is just for fresh drives.
5576 */
5577 struct intel_super *super;
5578 struct imsm_super *mpb;
5579 size_t mpb_size;
4d1313e9 5580 char *version;
bf5a934a 5581
83cd1e97 5582 if (data_offset != INVALID_SECTORS) {
ed503f89 5583 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5584 return 0;
5585 }
5586
bf5a934a 5587 if (st->sb)
5308f117 5588 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5589 data_offset);
e683ca88
DW		 5590
5591 if (info)
5592 mpb_size = disks_to_mpb_size(info->nr_disks);
5593 else
f36a9ecd 5594 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5595
49133e57 5596 super = alloc_super();
f36a9ecd
PB		 5597 if (super &&
5598 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5599 free_imsm(super);
e683ca88
DW		 5600 super = NULL;
5601 }
5602 if (!super) {
1ade5cc1 5603 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5604 return 0;
5605 }
de44e46f
PB		 5606 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5607 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5608 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5609 free(super->buf);
8d67477f 5610 free_imsm(super);
8e59f3d8
AK		 5611 return 0;
5612 }
e683ca88 5613 memset(super->buf, 0, mpb_size);
ef649044 5614 mpb = super->buf;
e683ca88
DW		 5615 mpb->mpb_size = __cpu_to_le32(mpb_size);
5616 st->sb = super;
5617
5618 if (info == NULL) {
5619 /* zeroing superblock */
5620 return 0;
5621 }
bf5a934a 5622
4d1313e9
DW		 5623 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5624
5625 version = (char *) mpb->sig;
5626 strcpy(version, MPB_SIGNATURE);
5627 version += strlen(MPB_SIGNATURE);
5628 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5629
bf5a934a
DW		 5630 return 1;
5631}
5632
f2cc4f7d
AO		 5633static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5634{
5635 unsigned int member_sector_size;
5636
5637 if (dl->fd < 0) {
5638 pr_err("Invalid file descriptor for %s\n", dl->devname);
5639 return 0;
5640 }
5641
5642 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5643 return 0;
5644 if (member_sector_size != super->sector_size)
5645 return 0;
5646 return 1;
5647}
5648
f20c3968 5649static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5650 int fd, char *devname)
5651{
5652 struct intel_super *super = st->sb;
d23fe947 5653 struct imsm_super *mpb = super->anchor;
3960e579 5654 struct imsm_disk *_disk;
bf5a934a
DW		 5655 struct imsm_dev *dev;
5656 struct imsm_map *map;
3960e579 5657 struct dl *dl, *df;
4eb26970 5658 int slot;
bf5a934a 5659
949c47a0 5660 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5661 map = get_imsm_map(dev, MAP_0);
bf5a934a 5662
208933a7 5663 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5664 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5665 devname);
5666 return 1;
5667 }
5668
efb30e7f
DW		 5669 if (fd == -1) {
5670 /* we're doing autolayout so grab the pre-marked (in
5671 * validate_geometry) raid_disk
5672 */
5673 for (dl = super->disks; dl; dl = dl->next)
5674 if (dl->raiddisk == dk->raid_disk)
5675 break;
5676 } else {
5677 for (dl = super->disks; dl ; dl = dl->next)
5678 if (dl->major == dk->major &&
5679 dl->minor == dk->minor)
5680 break;
5681 }
d23fe947 5682
208933a7 5683 if (!dl) {
e7b84f9d 5684 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5685 return 1;
208933a7 5686 }
bf5a934a 5687
59632db9
MZ		 5688 if (mpb->num_disks == 0)
5689 if (!get_dev_sector_size(dl->fd, dl->devname,
5690 &super->sector_size))
5691 return 1;
5692
f2cc4f7d
AO		 5693 if (!drive_validate_sector_size(super, dl)) {
5694 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5695 return 1;
5696 }
5697
d23fe947
DW		 5698 /* add a pristine spare to the metadata */
5699 if (dl->index < 0) {
5700 dl->index = super->anchor->num_disks;
5701 super->anchor->num_disks++;
5702 }
4eb26970
DW		 5703 /* Check the device has not already been added */
5704 slot = get_imsm_disk_slot(map, dl->index);
5705 if (slot >= 0 &&
238c0a71 5706 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5707 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5708 devname);
5709 return 1;
5710 }
656b6b5a 5711 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5712 dl->disk.status = CONFIGURED_DISK;
d23fe947 5713
3960e579
DW		 5714 /* update size of 'missing' disks to be at least as large as the
5715 * largest acitve member (we only have dummy missing disks when
5716 * creating the first volume)
5717 */
5718 if (super->current_vol == 0) {
5719 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5720 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5721 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5722 _disk = __get_imsm_disk(mpb, df->index);
5723 *_disk = df->disk;
5724 }
5725 }
5726
5727 /* refresh unset/failed slots to point to valid 'missing' entries */
5728 for (df = super->missing; df; df = df->next)
5729 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5730 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5731
5732 if ((ord & IMSM_ORD_REBUILD) == 0)
5733 continue;
5734 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5735 if (is_gen_migration(dev)) {
238c0a71
AK		 5736 struct imsm_map *map2 = get_imsm_map(dev,
5737 MAP_1);
0a108d63 5738 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5739 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5740 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5741 slot2,
5742 MAP_1);
1ace8403
AK		 5743 if ((unsigned)df->index ==
5744 ord_to_idx(ord2))
5745 set_imsm_ord_tbl_ent(map2,
0a108d63 5746 slot2,
1ace8403
AK		 5747 df->index |
5748 IMSM_ORD_REBUILD);
5749 }
5750 }
3960e579
DW		 5751 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5752 break;
5753 }
5754
d23fe947
DW		 5755 /* if we are creating the first raid device update the family number */
5756 if (super->current_vol == 0) {
5757 __u32 sum;
5758 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5759
3960e579 5760 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5761 if (!_dev || !_disk) {
e7b84f9d 5762 pr_err("BUG mpb setup error\n");
791b666a
AW		 5763 return 1;
5764 }
d23fe947
DW		 5765 *_dev = *dev;
5766 *_disk = dl->disk;
148acb7b
DW		 5767 sum = random32();
5768 sum += __gen_imsm_checksum(mpb);
d23fe947 5769 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5770 mpb->orig_family_num = mpb->family_num;
e48aed3c 5771 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5772 }
ca0748fa 5773 super->current_disk = dl;
f20c3968 5774 return 0;
bf5a934a
DW		 5775}
5776
a8619d23
AK		 5777/* mark_spare()
5778 * Function marks disk as spare and restores disk serial
5779 * in case it was previously marked as failed by takeover operation
5780 * reruns:
5781 * -1 : critical error
5782 * 0 : disk is marked as spare but serial is not set
5783 * 1 : success
5784 */
5785int mark_spare(struct dl *disk)
5786{
5787 __u8 serial[MAX_RAID_SERIAL_LEN];
5788 int ret_val = -1;
5789
5790 if (!disk)
5791 return ret_val;
5792
5793 ret_val = 0;
6da53c0e 5794 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5795 /* Restore disk serial number, because takeover marks disk
5796 * as failed and adds to serial ':0' before it becomes
5797 * a spare disk.
5798 */
5799 serialcpy(disk->serial, serial);
5800 serialcpy(disk->disk.serial, serial);
5801 ret_val = 1;
5802 }
5803 disk->disk.status = SPARE_DISK;
5804 disk->index = -1;
5805
5806 return ret_val;
5807}
88654014 5808
f20c3968 5809static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5810 int fd, char *devname,
5811 unsigned long long data_offset)
cdddbdbc 5812{
c2c087e6 5813 struct intel_super *super = st->sb;
c2c087e6
DW		 5814 struct dl *dd;
5815 unsigned long long size;
fa7bb6f8 5816 unsigned int member_sector_size;
f2f27e63 5817 __u32 id;
c2c087e6
DW		 5818 int rv;
5819 struct stat stb;
5820
88654014
LM		 5821 /* If we are on an RAID enabled platform check that the disk is
5822 * attached to the raid controller.
5823 * We do not need to test disks attachment for container based additions,
5824 * they shall be already tested when container was created/assembled.
88c32bb1 5825 */
d424212e 5826 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5827 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5828 if (rv != 0) {
5829 dprintf("capability: %p fd: %d ret: %d\n",
5830 super->orom, fd, rv);
5831 return 1;
88c32bb1
DW		 5832 }
5833
f20c3968
DW		 5834 if (super->current_vol >= 0)
5835 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5836
c2c087e6 5837 fstat(fd, &stb);
503975b9 5838 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5839 dd->major = major(stb.st_rdev);
5840 dd->minor = minor(stb.st_rdev);
503975b9 5841 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5842 dd->fd = fd;
689c9bf3 5843 dd->e = NULL;
1a64be56 5844 dd->action = DISK_ADD;
6da53c0e 5845 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5846 if (rv) {
e7b84f9d 5847 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5848 if (dd->devname)
5849 free(dd->devname);
949c47a0 5850 free(dd);
0030e8d6 5851 abort();
c2c087e6 5852 }
20bee0f8
PB		 5853 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5854 (super->hba->type == SYS_DEV_VMD))) {
5855 int i;
5856 char *devpath = diskfd_to_devpath(fd);
5857 char controller_path[PATH_MAX];
5858
5859 if (!devpath) {
5860 pr_err("failed to get devpath, aborting\n");
5861 if (dd->devname)
5862 free(dd->devname);
5863 free(dd);
5864 return 1;
5865 }
5866
5867 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5868 free(devpath);
5869
5870 if (devpath_to_vendor(controller_path) == 0x8086) {
5871 /*
5872 * If Intel's NVMe drive has serial ended with
5873 * "-A","-B","-1" or "-2" it means that this is "x8"
5874 * device (double drive on single PCIe card).
5875 * User should be warned about potential data loss.
5876 */
5877 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5878 /* Skip empty character at the end */
5879 if (dd->serial[i] == 0)
5880 continue;
5881
5882 if (((dd->serial[i] == 'A') ||
5883 (dd->serial[i] == 'B') ||
5884 (dd->serial[i] == '1') ||
5885 (dd->serial[i] == '2')) &&
5886 (dd->serial[i-1] == '-'))
5887 pr_err("\tThe action you are about to take may put your data at risk.\n"
5888 "\tPlease note that x8 devices may consist of two separate x4 devices "
5889 "located on a single PCIe port.\n"
5890 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5891 break;
5892 }
32716c51
PB		 5893 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5894 !imsm_orom_has_tpv_support(super->orom)) {
5895 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5896 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5897 free(dd->devname);
5898 free(dd);
5899 return 1;
20bee0f8
PB		 5900 }
5901 }
c2c087e6 5902
c2c087e6 5903 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5904 get_dev_sector_size(fd, NULL, &member_sector_size);
5905
5906 if (super->sector_size == 0) {
5907 /* this a first device, so sector_size is not set yet */
5908 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5909 }
5910
71e5411e 5911 /* clear migr_rec when adding disk to container */
85337573
AO		 5912 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5913 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5914 SEEK_SET) >= 0) {
466070ad 5915 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5916 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5917 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5918 perror("Write migr_rec failed");
5919 }
5920
c2c087e6 5921 size /= 512;
1f24f035 5922 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5923 set_total_blocks(&dd->disk, size);
5924 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5925 struct imsm_super *mpb = super->anchor;
5926 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5927 }
a8619d23 5928 mark_spare(dd);
c2c087e6 5929 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5930 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5931 else
b9f594fe 5932 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5933
5934 if (st->update_tail) {
1a64be56
LM		 5935 dd->next = super->disk_mgmt_list;
5936 super->disk_mgmt_list = dd;
43dad3d6
DW		 5937 } else {
5938 dd->next = super->disks;
5939 super->disks = dd;
ceaf0ee1 5940 super->updates_pending++;
43dad3d6 5941 }
f20c3968
DW		 5942
5943 return 0;
cdddbdbc
DW		 5944}
5945
1a64be56
LM		 5946static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5947{
5948 struct intel_super *super = st->sb;
5949 struct dl *dd;
5950
5951 /* remove from super works only in mdmon - for communication
5952 * manager - monitor. Check if communication memory buffer
5953 * is prepared.
5954 */
5955 if (!st->update_tail) {
1ade5cc1 5956 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5957 return 1;
5958 }
503975b9 5959 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5960 dd->major = dk->major;
5961 dd->minor = dk->minor;
1a64be56 5962 dd->fd = -1;
a8619d23 5963 mark_spare(dd);
1a64be56
LM		 5964 dd->action = DISK_REMOVE;
5965
5966 dd->next = super->disk_mgmt_list;
5967 super->disk_mgmt_list = dd;
5968
1a64be56
LM		 5969 return 0;
5970}
5971
f796af5d
DW		 5972static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5973
5974static union {
f36a9ecd 5975 char buf[MAX_SECTOR_SIZE];
f796af5d 5976 struct imsm_super anchor;
f36a9ecd 5977} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5978
d23fe947
DW		 5979/* spare records have their own family number and do not have any defined raid
5980 * devices
5981 */
5982static int write_super_imsm_spares(struct intel_super *super, int doclose)
5983{
d23fe947 5984 struct imsm_super *mpb = super->anchor;
f796af5d 5985 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5986 __u32 sum;
5987 struct dl *d;
5988
68641cdb
JS		 5989 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5990 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5991 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5992 spare->num_disks = 1;
5993 spare->num_raid_devs = 0;
5994 spare->cache_size = mpb->cache_size;
5995 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5996
5997 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5998 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5999
6000 for (d = super->disks; d; d = d->next) {
8796fdc4 6001 if (d->index != -1)
d23fe947
DW		 6002 continue;
6003
f796af5d 6004 spare->disk[0] = d->disk;
027c374f
CA		 6005 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6006 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6007
f36a9ecd
PB		 6008 if (super->sector_size == 4096)
6009 convert_to_4k_imsm_disk(&spare->disk[0]);
6010
f796af5d
DW		 6011 sum = __gen_imsm_checksum(spare);
6012 spare->family_num = __cpu_to_le32(sum);
6013 spare->orig_family_num = 0;
6014 sum = __gen_imsm_checksum(spare);
6015 spare->check_sum = __cpu_to_le32(sum);
d23fe947 6016
f796af5d 6017 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 6018 pr_err("failed for device %d:%d %s\n",
6019 d->major, d->minor, strerror(errno));
e74255d9 6020 return 1;
d23fe947
DW		 6021 }
6022 if (doclose) {
6023 close(d->fd);
6024 d->fd = -1;
6025 }
6026 }
6027
e74255d9 6028 return 0;
d23fe947
DW		 6029}
6030
36988a3d 6031static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6032{
36988a3d 6033 struct intel_super *super = st->sb;
f36a9ecd 6034 unsigned int sector_size = super->sector_size;
949c47a0 6035 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6036 struct dl *d;
6037 __u32 generation;
6038 __u32 sum;
d23fe947 6039 int spares = 0;
949c47a0 6040 int i;
a48ac0a8 6041 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6042 int num_disks = 0;
146c6260 6043 int clear_migration_record = 1;
bbab0940 6044 __u32 bbm_log_size;
cdddbdbc 6045
c2c087e6
DW		 6046 /* 'generation' is incremented everytime the metadata is written */
6047 generation = __le32_to_cpu(mpb->generation_num);
6048 generation++;
6049 mpb->generation_num = __cpu_to_le32(generation);
6050
148acb7b
DW		 6051 /* fix up cases where previous mdadm releases failed to set
6052 * orig_family_num
6053 */
6054 if (mpb->orig_family_num == 0)
6055 mpb->orig_family_num = mpb->family_num;
6056
d23fe947 6057 for (d = super->disks; d; d = d->next) {
8796fdc4 6058 if (d->index == -1)
d23fe947 6059 spares++;
36988a3d 6060 else {
d23fe947 6061 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6062 num_disks++;
6063 }
d23fe947 6064 }
36988a3d 6065 for (d = super->missing; d; d = d->next) {
47ee5a45 6066 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6067 num_disks++;
6068 }
6069 mpb->num_disks = num_disks;
6070 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6071
949c47a0
DW		 6072 for (i = 0; i < mpb->num_raid_devs; i++) {
6073 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6074 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6075 if (dev && dev2) {
6076 imsm_copy_dev(dev, dev2);
6077 mpb_size += sizeof_imsm_dev(dev, 0);
6078 }
146c6260
AK		 6079 if (is_gen_migration(dev2))
6080 clear_migration_record = 0;
949c47a0 6081 }
bbab0940
TM		 6082
6083 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6084
6085 if (bbm_log_size) {
6086 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6087 mpb->attributes |= MPB_ATTRIB_BBM;
6088 } else
6089 mpb->attributes &= ~MPB_ATTRIB_BBM;
6090
6091 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6092 mpb_size += bbm_log_size;
a48ac0a8 6093 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6094
bbab0940
TM		 6095#ifdef DEBUG
6096 assert(super->len == 0 || mpb_size <= super->len);
6097#endif
6098
c2c087e6 6099 /* recalculate checksum */
949c47a0 6100 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6101 mpb->check_sum = __cpu_to_le32(sum);
6102
51d83f5d
AK		 6103 if (super->clean_migration_record_by_mdmon) {
6104 clear_migration_record = 1;
6105 super->clean_migration_record_by_mdmon = 0;
6106 }
146c6260 6107 if (clear_migration_record)
de44e46f 6108 memset(super->migr_rec_buf, 0,
85337573 6109 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6110
f36a9ecd
PB		 6111 if (sector_size == 4096)
6112 convert_to_4k(super);
6113
d23fe947 6114 /* write the mpb for disks that compose raid devices */
c2c087e6 6115 for (d = super->disks; d ; d = d->next) {
86c54047 6116 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6117 continue;
30602f53 6118
146c6260
AK		 6119 if (clear_migration_record) {
6120 unsigned long long dsize;
6121
6122 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6123 if (lseek64(d->fd, dsize - sector_size,
6124 SEEK_SET) >= 0) {
466070ad
PB		 6125 if ((unsigned int)write(d->fd,
6126 super->migr_rec_buf,
de44e46f
PB		 6127 MIGR_REC_BUF_SECTORS*sector_size) !=
6128 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6129 perror("Write migr_rec failed");
146c6260
AK		 6130 }
6131 }
51d83f5d
AK		 6132
6133 if (store_imsm_mpb(d->fd, mpb))
6134 fprintf(stderr,
1ade5cc1
N		 6135 "failed for device %d:%d (fd: %d)%s\n",
6136 d->major, d->minor,
51d83f5d
AK		 6137 d->fd, strerror(errno));
6138
c2c087e6
DW		 6139 if (doclose) {
6140 close(d->fd);
6141 d->fd = -1;
6142 }
6143 }
6144
d23fe947
DW		 6145 if (spares)
6146 return write_super_imsm_spares(super, doclose);
6147
e74255d9 6148 return 0;
c2c087e6
DW		 6149}
6150
9b1fb677 6151static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6152{
6153 size_t len;
6154 struct imsm_update_create_array *u;
6155 struct intel_super *super = st->sb;
9b1fb677 6156 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6157 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6158 struct disk_info *inf;
6159 struct imsm_disk *disk;
6160 int i;
43dad3d6 6161
54c2c1ea
DW		 6162 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6163 sizeof(*inf) * map->num_members;
503975b9 6164 u = xmalloc(len);
43dad3d6 6165 u->type = update_create_array;
9b1fb677 6166 u->dev_idx = dev_idx;
43dad3d6 6167 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6168 inf = get_disk_info(u);
6169 for (i = 0; i < map->num_members; i++) {
238c0a71 6170 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6171
54c2c1ea 6172 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6173 if (!disk)
6174 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6175 serialcpy(inf[i].serial, disk->serial);
6176 }
43dad3d6
DW		 6177 append_metadata_update(st, u, len);
6178
6179 return 0;
6180}
6181
1a64be56 6182static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6183{
6184 struct intel_super *super = st->sb;
6185 size_t len;
1a64be56 6186 struct imsm_update_add_remove_disk *u;
43dad3d6 6187
1a64be56 6188 if (!super->disk_mgmt_list)
43dad3d6
DW		 6189 return 0;
6190
6191 len = sizeof(*u);
503975b9 6192 u = xmalloc(len);
1a64be56 6193 u->type = update_add_remove_disk;
43dad3d6
DW		 6194 append_metadata_update(st, u, len);
6195
6196 return 0;
6197}
2432ce9b
AP		 6198
6199__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6200
e397cefe
AP		 6201static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6202{
6203 struct ppl_header *ppl_hdr = buf;
6204 int ret;
6205
6206 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6207
6208 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6209 ret = -errno;
6210 perror("Failed to seek to PPL header location");
6211 return ret;
6212 }
6213
6214 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6215 ret = -errno;
6216 perror("Write PPL header failed");
6217 return ret;
6218 }
6219
6220 fsync(fd);
6221
6222 return 0;
6223}
6224
2432ce9b
AP		 6225static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6226{
6227 struct intel_super *super = st->sb;
6228 void *buf;
6229 struct ppl_header *ppl_hdr;
6230 int ret;
6231
b2514242
PB		 6232 /* first clear entire ppl space */
6233 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6234 if (ret)
6235 return ret;
6236
6237 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6238 if (ret) {
6239 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6240 return -ret;
2432ce9b
AP		 6241 }
6242
6243 memset(buf, 0, PPL_HEADER_SIZE);
6244 ppl_hdr = buf;
6245 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6246 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6247
6248 if (info->mismatch_cnt) {
6249 /*
6250 * We are overwriting an invalid ppl. Make one entry with wrong
6251 * checksum to prevent the kernel from skipping resync.
6252 */
6253 ppl_hdr->entries_count = __cpu_to_le32(1);
6254 ppl_hdr->entries[0].checksum = ~0;
6255 }
6256
e397cefe 6257 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6258
6259 free(buf);
6260 return ret;
6261}
6262
e397cefe
AP		 6263static int is_rebuilding(struct imsm_dev *dev);
6264
2432ce9b
AP		 6265static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6266 struct mdinfo *disk)
6267{
6268 struct intel_super *super = st->sb;
6269 struct dl *d;
e397cefe 6270 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6271 int ret = 0;
e397cefe 6272 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6273 __u32 crc;
6274 struct imsm_dev *dev;
2432ce9b 6275 __u32 idx;
44b6b876
PB		 6276 unsigned int i;
6277 unsigned long long ppl_offset = 0;
6278 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6279
6280 if (disk->disk.raid_disk < 0)
6281 return 0;
6282
2432ce9b 6283 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6284 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6285 d = get_imsm_dl_disk(super, idx);
6286
6287 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6288 return 0;
6289
6290 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6291 pr_err("Failed to allocate PPL header buffer\n");
6292 return -1;
6293 }
6294 buf = buf_orig;
2432ce9b 6295
44b6b876
PB		 6296 ret = 1;
6297 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6298 void *tmp;
6299
44b6b876 6300 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6301
44b6b876
PB		 6302 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6303 SEEK_SET) < 0) {
6304 perror("Failed to seek to PPL header location");
6305 ret = -1;
e397cefe 6306 break;
44b6b876 6307 }
2432ce9b 6308
44b6b876
PB		 6309 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6310 perror("Read PPL header failed");
6311 ret = -1;
e397cefe 6312 break;
44b6b876 6313 }
2432ce9b 6314
44b6b876 6315 ppl_hdr = buf;
2432ce9b 6316
44b6b876
PB		 6317 crc = __le32_to_cpu(ppl_hdr->checksum);
6318 ppl_hdr->checksum = 0;
6319
6320 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6321 dprintf("Wrong PPL header checksum on %s\n",
6322 d->devname);
e397cefe 6323 break;
44b6b876
PB		 6324 }
6325
6326 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6327 /* previous was newest, it was already checked */
e397cefe 6328 break;
44b6b876
PB		 6329 }
6330
6331 if ((__le32_to_cpu(ppl_hdr->signature) !=
6332 super->anchor->orig_family_num)) {
6333 dprintf("Wrong PPL header signature on %s\n",
6334 d->devname);
6335 ret = 1;
e397cefe 6336 break;
44b6b876
PB		 6337 }
6338
6339 ret = 0;
6340 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6341
44b6b876
PB		 6342 ppl_offset += PPL_HEADER_SIZE;
6343 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6344 ppl_offset +=
6345 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6346
6347 if (!buf_prev)
6348 buf_prev = buf + PPL_HEADER_SIZE;
6349 tmp = buf_prev;
6350 buf_prev = buf;
6351 buf = tmp;
2432ce9b
AP		 6352 }
6353
e397cefe
AP		 6354 if (buf_prev) {
6355 buf = buf_prev;
6356 ppl_hdr = buf_prev;
6357 }
2432ce9b 6358
54148aba
PB		 6359 /*
6360 * Update metadata to use mutliple PPLs area (1MB).
6361 * This is done once for all RAID members
6362 */
6363 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6364 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6365 char subarray[20];
6366 struct mdinfo *member_dev;
6367
6368 sprintf(subarray, "%d", info->container_member);
6369
6370 if (mdmon_running(st->container_devnm))
6371 st->update_tail = &st->updates;
6372
6373 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6374 pr_err("Failed to update subarray %s\n",
6375 subarray);
6376 } else {
6377 if (st->update_tail)
6378 flush_metadata_updates(st);
6379 else
6380 st->ss->sync_metadata(st);
6381 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6382 for (member_dev = info->devs; member_dev;
6383 member_dev = member_dev->next)
6384 member_dev->ppl_size =
6385 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6386 }
6387 }
6388
b23d0750 6389 if (ret == 1) {
2fc0fc63
AP		 6390 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6391
50b9c10d
PB		 6392 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6393 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6394 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6395 (is_rebuilding(dev) &&
2ec9d182
AP		 6396 dev->vol.curr_migr_unit == 0 &&
6397 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6398 ret = st->ss->write_init_ppl(st, info, d->fd);
6399 else
6400 info->mismatch_cnt++;
e397cefe
AP		 6401 } else if (ret == 0 &&
6402 ppl_hdr->entries_count == 0 &&
6403 is_rebuilding(dev) &&
6404 info->resync_start == 0) {
6405 /*
6406 * The header has no entries - add a single empty entry and
6407 * rewrite the header to prevent the kernel from going into
6408 * resync after an interrupted rebuild.
6409 */
6410 ppl_hdr->entries_count = __cpu_to_le32(1);
6411 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6412 }
2432ce9b 6413
e397cefe
AP		 6414 free(buf_orig);
6415
2432ce9b
AP		 6416 return ret;
6417}
6418
2432ce9b
AP		 6419static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6420{
6421 struct intel_super *super = st->sb;
6422 struct dl *d;
6423 int ret = 0;
6424
6425 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6426 info->array.level != 5)
6427 return 0;
6428
6429 for (d = super->disks; d ; d = d->next) {
6430 if (d->index < 0 || is_failed(&d->disk))
6431 continue;
6432
6433 ret = st->ss->write_init_ppl(st, info, d->fd);
6434 if (ret)
6435 break;
6436 }
6437
6438 return ret;
6439}
43dad3d6 6440
c2c087e6
DW		 6441static int write_init_super_imsm(struct supertype *st)
6442{
9b1fb677
DW		 6443 struct intel_super *super = st->sb;
6444 int current_vol = super->current_vol;
2432ce9b
AP		 6445 int rv = 0;
6446 struct mdinfo info;
6447
6448 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6449
6450 /* we are done with current_vol reset it to point st at the container */
6451 super->current_vol = -1;
6452
8273f55e 6453 if (st->update_tail) {
43dad3d6
DW		 6454 /* queue the recently created array / added disk
6455 * as a metadata update */
8273f55e 6456
43dad3d6 6457 /* determine if we are creating a volume or adding a disk */
9b1fb677 6458 if (current_vol < 0) {
1a64be56
LM		 6459 /* in the mgmt (add/remove) disk case we are running
6460 * in mdmon context, so don't close fd's
43dad3d6 6461 */
2432ce9b
AP		 6462 rv = mgmt_disk(st);
6463 } else {
6464 rv = write_init_ppl_imsm_all(st, &info);
6465 if (!rv)
6466 rv = create_array(st, current_vol);
6467 }
d682f344
N		 6468 } else {
6469 struct dl *d;
6470 for (d = super->disks; d; d = d->next)
ba728be7 6471 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6472 if (current_vol >= 0)
6473 rv = write_init_ppl_imsm_all(st, &info);
6474 if (!rv)
6475 rv = write_super_imsm(st, 1);
d682f344 6476 }
2432ce9b
AP		 6477
6478 return rv;
cdddbdbc
DW		 6479}
6480
e683ca88 6481static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6482{
e683ca88
DW		 6483 struct intel_super *super = st->sb;
6484 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6485
e683ca88 6486 if (!mpb)
ad97895e
DW		 6487 return 1;
6488
f36a9ecd
PB		 6489 if (super->sector_size == 4096)
6490 convert_to_4k(super);
e683ca88 6491 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6492}
6493
cdddbdbc
DW		 6494static int validate_geometry_imsm_container(struct supertype *st, int level,
6495 int layout, int raiddisks, int chunk,
af4348dd
N		 6496 unsigned long long size,
6497 unsigned long long data_offset,
6498 char *dev,
2c514b71
NB		 6499 unsigned long long *freesize,
6500 int verbose)
cdddbdbc 6501{
c2c087e6
DW		 6502 int fd;
6503 unsigned long long ldsize;
594dc1b8 6504 struct intel_super *super;
f2f5c343 6505 int rv = 0;
cdddbdbc 6506
c2c087e6
DW		 6507 if (level != LEVEL_CONTAINER)
6508 return 0;
6509 if (!dev)
6510 return 1;
6511
6512 fd = open(dev, O_RDONLY|O_EXCL, 0);
6513 if (fd < 0) {
ba728be7 6514 if (verbose > 0)
e7b84f9d 6515 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6516 dev, strerror(errno));
c2c087e6
DW		 6517 return 0;
6518 }
6519 if (!get_dev_size(fd, dev, &ldsize)) {
6520 close(fd);
6521 return 0;
6522 }
f2f5c343
LM		 6523
6524 /* capabilities retrieve could be possible
6525 * note that there is no fd for the disks in array.
6526 */
6527 super = alloc_super();
8d67477f
TM		 6528 if (!super) {
6529 close(fd);
6530 return 0;
6531 }
fa7bb6f8
PB		 6532 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6533 close(fd);
6534 free_imsm(super);
6535 return 0;
6536 }
6537
ba728be7 6538 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6539 if (rv != 0) {
6540#if DEBUG
6541 char str[256];
6542 fd2devname(fd, str);
1ade5cc1 6543 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6544 fd, str, super->orom, rv, raiddisks);
6545#endif
6546 /* no orom/efi or non-intel hba of the disk */
6547 close(fd);
6548 free_imsm(super);
6549 return 0;
6550 }
c2c087e6 6551 close(fd);
9126b9a8
CA		 6552 if (super->orom) {
6553 if (raiddisks > super->orom->tds) {
6554 if (verbose)
7a862a02 6555 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6556 raiddisks, super->orom->tds);
6557 free_imsm(super);
6558 return 0;
6559 }
6560 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6561 (ldsize >> 9) >> 32 > 0) {
6562 if (verbose)
e7b84f9d 6563 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6564 free_imsm(super);
6565 return 0;
6566 }
f2f5c343 6567 }
c2c087e6 6568
af4348dd 6569 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6570 free_imsm(super);
c2c087e6
DW		 6571
6572 return 1;
cdddbdbc
DW		 6573}
6574
0dcecb2e
DW		 6575static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6576{
6577 const unsigned long long base_start = e[*idx].start;
6578 unsigned long long end = base_start + e[*idx].size;
6579 int i;
6580
6581 if (base_start == end)
6582 return 0;
6583
6584 *idx = *idx + 1;
6585 for (i = *idx; i < num_extents; i++) {
6586 /* extend overlapping extents */
6587 if (e[i].start >= base_start &&
6588 e[i].start <= end) {
6589 if (e[i].size == 0)
6590 return 0;
6591 if (e[i].start + e[i].size > end)
6592 end = e[i].start + e[i].size;
6593 } else if (e[i].start > end) {
6594 *idx = i;
6595 break;
6596 }
6597 }
6598
6599 return end - base_start;
6600}
6601
6602static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6603{
6604 /* build a composite disk with all known extents and generate a new
6605 * 'maxsize' given the "all disks in an array must share a common start
6606 * offset" constraint
6607 */
503975b9 6608 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6609 struct dl *dl;
6610 int i, j;
6611 int start_extent;
6612 unsigned long long pos;
b9d77223 6613 unsigned long long start = 0;
0dcecb2e
DW		 6614 unsigned long long maxsize;
6615 unsigned long reserve;
6616
0dcecb2e
DW		 6617 /* coalesce and sort all extents. also, check to see if we need to
6618 * reserve space between member arrays
6619 */
6620 j = 0;
6621 for (dl = super->disks; dl; dl = dl->next) {
6622 if (!dl->e)
6623 continue;
6624 for (i = 0; i < dl->extent_cnt; i++)
6625 e[j++] = dl->e[i];
6626 }
6627 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6628
6629 /* merge extents */
6630 i = 0;
6631 j = 0;
6632 while (i < sum_extents) {
6633 e[j].start = e[i].start;
6634 e[j].size = find_size(e, &i, sum_extents);
6635 j++;
6636 if (e[j-1].size == 0)
6637 break;
6638 }
6639
6640 pos = 0;
6641 maxsize = 0;
6642 start_extent = 0;
6643 i = 0;
6644 do {
6645 unsigned long long esize;
6646
6647 esize = e[i].start - pos;
6648 if (esize >= maxsize) {
6649 maxsize = esize;
6650 start = pos;
6651 start_extent = i;
6652 }
6653 pos = e[i].start + e[i].size;
6654 i++;
6655 } while (e[i-1].size);
6656 free(e);
6657
a7dd165b
DW		 6658 if (maxsize == 0)
6659 return 0;
6660
6661 /* FIXME assumes volume at offset 0 is the first volume in a
6662 * container
6663 */
0dcecb2e
DW		 6664 if (start_extent > 0)
6665 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6666 else
6667 reserve = 0;
6668
6669 if (maxsize < reserve)
a7dd165b 6670 return 0;
0dcecb2e 6671
5551b113 6672 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6673 if (start + reserve > super->create_offset)
a7dd165b 6674 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6675 super->create_offset = start + reserve;
6676
6677 return maxsize - reserve;
6678}
6679
88c32bb1
DW		 6680static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6681{
6682 if (level < 0 || level == 6 || level == 4)
6683 return 0;
6684
6685 /* if we have an orom prevent invalid raid levels */
6686 if (orom)
6687 switch (level) {
6688 case 0: return imsm_orom_has_raid0(orom);
6689 case 1:
6690 if (raiddisks > 2)
6691 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6692 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6693 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6694 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6695 }
6696 else
6697 return 1; /* not on an Intel RAID platform so anything goes */
6698
6699 return 0;
6700}
6701
ca9de185
LM		 6702static int
6703active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6704 int dpa, int verbose)
6705{
6706 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6707 struct mdstat_ent *memb;
ca9de185
LM		 6708 int count = 0;
6709 int num = 0;
594dc1b8 6710 struct md_list *dv;
ca9de185
LM		 6711 int found;
6712
6713 for (memb = mdstat ; memb ; memb = memb->next) {
6714 if (memb->metadata_version &&
fc54fe7a 6715 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6716 (strcmp(&memb->metadata_version[9], name) == 0) &&
6717 !is_subarray(memb->metadata_version+9) &&
6718 memb->members) {
6719 struct dev_member *dev = memb->members;
6720 int fd = -1;
6721 while(dev && (fd < 0)) {
503975b9
N		 6722 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6723 num = sprintf(path, "%s%s", "/dev/", dev->name);
6724 if (num > 0)
6725 fd = open(path, O_RDONLY, 0);
089f9d79 6726 if (num <= 0 || fd < 0) {
676e87a8 6727 pr_vrb("Cannot open %s: %s\n",
503975b9 6728 dev->name, strerror(errno));
ca9de185 6729 }
503975b9 6730 free(path);
ca9de185
LM		 6731 dev = dev->next;
6732 }
6733 found = 0;
089f9d79 6734 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6735 struct mdstat_ent *vol;
6736 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6737 if (vol->active > 0 &&
ca9de185 6738 vol->metadata_version &&
9581efb1 6739 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6740 found++;
6741 count++;
6742 }
6743 }
6744 if (*devlist && (found < dpa)) {
503975b9 6745 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6746 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6747 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6748 dv->found = found;
6749 dv->used = 0;
6750 dv->next = *devlist;
6751 *devlist = dv;
ca9de185
LM		 6752 }
6753 }
6754 if (fd >= 0)
6755 close(fd);
6756 }
6757 }
6758 free_mdstat(mdstat);
6759 return count;
6760}
6761
6762#ifdef DEBUG_LOOP
6763static struct md_list*
6764get_loop_devices(void)
6765{
6766 int i;
6767 struct md_list *devlist = NULL;
594dc1b8 6768 struct md_list *dv;
ca9de185
LM		 6769
6770 for(i = 0; i < 12; i++) {
503975b9
N		 6771 dv = xcalloc(1, sizeof(*dv));
6772 dv->devname = xmalloc(40);
ca9de185
LM		 6773 sprintf(dv->devname, "/dev/loop%d", i);
6774 dv->next = devlist;
6775 devlist = dv;
6776 }
6777 return devlist;
6778}
6779#endif
6780
6781static struct md_list*
6782get_devices(const char *hba_path)
6783{
6784 struct md_list *devlist = NULL;
594dc1b8 6785 struct md_list *dv;
ca9de185
LM		 6786 struct dirent *ent;
6787 DIR *dir;
6788 int err = 0;
6789
6790#if DEBUG_LOOP
6791 devlist = get_loop_devices();
6792 return devlist;
6793#endif
6794 /* scroll through /sys/dev/block looking for devices attached to
6795 * this hba
6796 */
6797 dir = opendir("/sys/dev/block");
6798 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6799 int fd;
6800 char buf[1024];
6801 int major, minor;
6802 char *path = NULL;
6803 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6804 continue;
6805 path = devt_to_devpath(makedev(major, minor));
6806 if (!path)
6807 continue;
6808 if (!path_attached_to_hba(path, hba_path)) {
6809 free(path);
6810 path = NULL;
6811 continue;
6812 }
6813 free(path);
6814 path = NULL;
6815 fd = dev_open(ent->d_name, O_RDONLY);
6816 if (fd >= 0) {
6817 fd2devname(fd, buf);
6818 close(fd);
6819 } else {
e7b84f9d 6820 pr_err("cannot open device: %s\n",
ca9de185
LM		 6821 ent->d_name);
6822 continue;
6823 }
6824
503975b9
N		 6825 dv = xcalloc(1, sizeof(*dv));
6826 dv->devname = xstrdup(buf);
ca9de185
LM		 6827 dv->next = devlist;
6828 devlist = dv;
6829 }
6830 if (err) {
6831 while(devlist) {
6832 dv = devlist;
6833 devlist = devlist->next;
6834 free(dv->devname);
6835 free(dv);
6836 }
6837 }
562aa102 6838 closedir(dir);
ca9de185
LM		 6839 return devlist;
6840}
6841
6842static int
6843count_volumes_list(struct md_list *devlist, char *homehost,
6844 int verbose, int *found)
6845{
6846 struct md_list *tmpdev;
6847 int count = 0;
594dc1b8 6848 struct supertype *st;
ca9de185
LM		 6849
6850 /* first walk the list of devices to find a consistent set
6851 * that match the criterea, if that is possible.
6852 * We flag the ones we like with 'used'.
6853 */
6854 *found = 0;
6855 st = match_metadata_desc_imsm("imsm");
6856 if (st == NULL) {
676e87a8 6857 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6858 return 0;
6859 }
6860
6861 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6862 char *devname = tmpdev->devname;
0a6bff09 6863 dev_t rdev;
ca9de185
LM		 6864 struct supertype *tst;
6865 int dfd;
6866 if (tmpdev->used > 1)
6867 continue;
6868 tst = dup_super(st);
6869 if (tst == NULL) {
676e87a8 6870 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6871 goto err_1;
6872 }
6873 tmpdev->container = 0;
6874 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6875 if (dfd < 0) {
1ade5cc1 6876 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6877 devname, strerror(errno));
6878 tmpdev->used = 2;
0a6bff09 6879 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6880 tmpdev->used = 2;
6881 } else if (must_be_container(dfd)) {
6882 struct supertype *cst;
6883 cst = super_by_fd(dfd, NULL);
6884 if (cst == NULL) {
1ade5cc1 6885 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6886 devname);
6887 tmpdev->used = 2;
6888 } else if (tst->ss != st->ss) {
1ade5cc1 6889 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6890 devname);
6891 tmpdev->used = 2;
6892 } else if (!tst->ss->load_container ||
6893 tst->ss->load_container(tst, dfd, NULL))
6894 tmpdev->used = 2;
6895 else {
6896 tmpdev->container = 1;
6897 }
6898 if (cst)
6899 cst->ss->free_super(cst);
6900 } else {
0a6bff09 6901 tmpdev->st_rdev = rdev;
ca9de185 6902 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6903 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6904 devname);
6905 tmpdev->used = 2;
6906 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6907 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6908 tst->ss->name, devname);
6909 tmpdev->used = 2;
6910 }
6911 }
6912 if (dfd >= 0)
6913 close(dfd);
6914 if (tmpdev->used == 2 || tmpdev->used == 4) {
6915 /* Ignore unrecognised devices during auto-assembly */
6916 goto loop;
6917 }
6918 else {
6919 struct mdinfo info;
6920 tst->ss->getinfo_super(tst, &info, NULL);
6921
6922 if (st->minor_version == -1)
6923 st->minor_version = tst->minor_version;
6924
6925 if (memcmp(info.uuid, uuid_zero,
6926 sizeof(int[4])) == 0) {
6927 /* this is a floating spare. It cannot define
6928 * an array unless there are no more arrays of
6929 * this type to be found. It can be included
6930 * in an array of this type though.
6931 */
6932 tmpdev->used = 3;
6933 goto loop;
6934 }
6935
6936 if (st->ss != tst->ss ||
6937 st->minor_version != tst->minor_version ||
6938 st->ss->compare_super(st, tst) != 0) {
6939 /* Some mismatch. If exactly one array matches this host,
6940 * we can resolve on that one.
6941 * Or, if we are auto assembling, we just ignore the second
6942 * for now.
6943 */
1ade5cc1 6944 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6945 devname);
6946 goto loop;
6947 }
6948 tmpdev->used = 1;
6949 *found = 1;
6950 dprintf("found: devname: %s\n", devname);
6951 }
6952 loop:
6953 if (tst)
6954 tst->ss->free_super(tst);
6955 }
6956 if (*found != 0) {
6957 int err;
6958 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6959 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6960 for (iter = head; iter; iter = iter->next) {
6961 dprintf("content->text_version: %s vol\n",
6962 iter->text_version);
6963 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6964 /* do not assemble arrays with unsupported
6965 configurations */
1ade5cc1 6966 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6967 iter->text_version);
6968 } else
6969 count++;
6970 }
6971 sysfs_free(head);
6972
6973 } else {
1ade5cc1 6974 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6975 err, st->sb);
6976 }
6977 } else {
1ade5cc1 6978 dprintf("no more devices to examine\n");
ca9de185
LM		 6979 }
6980
6981 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6982 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6983 if (count) {
6984 if (count < tmpdev->found)
6985 count = 0;
6986 else
6987 count -= tmpdev->found;
6988 }
6989 }
6990 if (tmpdev->used == 1)
6991 tmpdev->used = 4;
6992 }
6993 err_1:
6994 if (st)
6995 st->ss->free_super(st);
6996 return count;
6997}
6998
d3c11416
AO		 6999static int __count_volumes(char *hba_path, int dpa, int verbose,
7000 int cmp_hba_path)
ca9de185 7001{
72a45777 7002 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7003 int count = 0;
72a45777
PB		 7004 const struct orom_entry *entry;
7005 struct devid_list *dv, *devid_list;
ca9de185 7006
d3c11416 7007 if (!hba_path)
ca9de185
LM		 7008 return 0;
7009
72a45777 7010 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7011 if (strstr(idev->path, hba_path))
7012 break;
72a45777
PB		 7013 }
7014
7015 if (!idev || !idev->dev_id)
ca9de185 7016 return 0;
72a45777
PB		 7017
7018 entry = get_orom_entry_by_device_id(idev->dev_id);
7019
7020 if (!entry || !entry->devid_list)
7021 return 0;
7022
7023 devid_list = entry->devid_list;
7024 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7025 struct md_list *devlist;
d3c11416
AO		 7026 struct sys_dev *device = NULL;
7027 char *hpath;
72a45777
PB		 7028 int found = 0;
7029
d3c11416
AO		 7030 if (cmp_hba_path)
7031 device = device_by_id_and_path(dv->devid, hba_path);
7032 else
7033 device = device_by_id(dv->devid);
7034
72a45777 7035 if (device)
d3c11416 7036 hpath = device->path;
72a45777
PB		 7037 else
7038 return 0;
7039
d3c11416 7040 devlist = get_devices(hpath);
72a45777
PB		 7041 /* if no intel devices return zero volumes */
7042 if (devlist == NULL)
7043 return 0;
7044
d3c11416
AO		 7045 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7046 verbose);
7047 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7048 if (devlist == NULL)
7049 return 0;
7050 do {
7051 found = 0;
7052 count += count_volumes_list(devlist,
7053 NULL,
7054 verbose,
7055 &found);
7056 dprintf("found %d count: %d\n", found, count);
7057 } while (found);
7058
d3c11416 7059 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7060
7061 while (devlist) {
7062 struct md_list *dv = devlist;
7063 devlist = devlist->next;
7064 free(dv->devname);
7065 free(dv);
7066 }
ca9de185
LM		 7067 }
7068 return count;
7069}
7070
d3c11416
AO		 7071static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7072{
7073 if (!hba)
7074 return 0;
7075 if (hba->type == SYS_DEV_VMD) {
7076 struct sys_dev *dev;
7077 int count = 0;
7078
7079 for (dev = find_intel_devices(); dev; dev = dev->next) {
7080 if (dev->type == SYS_DEV_VMD)
7081 count += __count_volumes(dev->path, dpa,
7082 verbose, 1);
7083 }
7084 return count;
7085 }
7086 return __count_volumes(hba->path, dpa, verbose, 0);
7087}
7088
cd9d1ac7
DW		 7089static int imsm_default_chunk(const struct imsm_orom *orom)
7090{
7091 /* up to 512 if the plaform supports it, otherwise the platform max.
7092 * 128 if no platform detected
7093 */
7094 int fs = max(7, orom ? fls(orom->sss) : 0);
7095
7096 return min(512, (1 << fs));
7097}
73408129 7098
6592ce37
DW		 7099static int
7100validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7101 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7102{
660260d0
DW		 7103 /* check/set platform and metadata limits/defaults */
7104 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7105 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7106 super->orom->dpa);
73408129
LM		 7107 return 0;
7108 }
7109
5d500228 7110 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7111 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7112 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7113 level, raiddisks, raiddisks > 1 ? "s" : "");
7114 return 0;
7115 }
cd9d1ac7 7116
7ccc4cc4 7117 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7118 *chunk = imsm_default_chunk(super->orom);
7119
7ccc4cc4 7120 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7121 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7122 return 0;
6592ce37 7123 }
cd9d1ac7 7124
6592ce37
DW		 7125 if (layout != imsm_level_to_layout(level)) {
7126 if (level == 5)
676e87a8 7127 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7128 else if (level == 10)
676e87a8 7129 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7130 else
676e87a8 7131 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7132 layout, level);
7133 return 0;
7134 }
2cc699af 7135
7ccc4cc4 7136 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7137 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7138 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7139 return 0;
7140 }
614902f6 7141
6592ce37
DW		 7142 return 1;
7143}
7144
1011e834 7145/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7146 * FIX ME add ahci details
7147 */
8b353278 7148static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7149 int layout, int raiddisks, int *chunk,
af4348dd
N		 7150 unsigned long long size,
7151 unsigned long long data_offset,
7152 char *dev,
2c514b71
NB		 7153 unsigned long long *freesize,
7154 int verbose)
cdddbdbc 7155{
9e04ac1c 7156 dev_t rdev;
c2c087e6 7157 struct intel_super *super = st->sb;
b2916f25 7158 struct imsm_super *mpb;
c2c087e6
DW		 7159 struct dl *dl;
7160 unsigned long long pos = 0;
7161 unsigned long long maxsize;
7162 struct extent *e;
7163 int i;
cdddbdbc 7164
88c32bb1
DW		 7165 /* We must have the container info already read in. */
7166 if (!super)
c2c087e6
DW		 7167 return 0;
7168
b2916f25
JS		 7169 mpb = super->anchor;
7170
2cc699af 7171 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7172 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7173 return 0;
d54559f0 7174 }
c2c087e6
DW		 7175 if (!dev) {
7176 /* General test: make sure there is space for
2da8544a
DW		 7177 * 'raiddisks' device extents of size 'size' at a given
7178 * offset
c2c087e6 7179 */
e46273eb 7180 unsigned long long minsize = size;
b7528a20 7181 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7182 int dcnt = 0;
7183 if (minsize == 0)
7184 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7185 for (dl = super->disks; dl ; dl = dl->next) {
7186 int found = 0;
7187
bf5a934a 7188 pos = 0;
c2c087e6 7189 i = 0;
05501181 7190 e = get_extents(super, dl, 0);
c2c087e6
DW		 7191 if (!e) continue;
7192 do {
7193 unsigned long long esize;
7194 esize = e[i].start - pos;
7195 if (esize >= minsize)
7196 found = 1;
b7528a20 7197 if (found && start_offset == MaxSector) {
2da8544a
DW		 7198 start_offset = pos;
7199 break;
7200 } else if (found && pos != start_offset) {
7201 found = 0;
7202 break;
7203 }
c2c087e6
DW		 7204 pos = e[i].start + e[i].size;
7205 i++;
7206 } while (e[i-1].size);
7207 if (found)
7208 dcnt++;
7209 free(e);
7210 }
7211 if (dcnt < raiddisks) {
2c514b71 7212 if (verbose)
7a862a02 7213 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7214 dcnt, raiddisks);
c2c087e6
DW		 7215 return 0;
7216 }
7217 return 1;
7218 }
0dcecb2e 7219
c2c087e6 7220 /* This device must be a member of the set */
9e04ac1c 7221 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7222 return 0;
7223 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7224 if (dl->major == (int)major(rdev) &&
7225 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7226 break;
7227 }
7228 if (!dl) {
2c514b71 7229 if (verbose)
7a862a02 7230 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7231 return 0;
a20d2ba5
DW		 7232 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7233 /* If a volume is present then the current creation attempt
7234 * cannot incorporate new spares because the orom may not
7235 * understand this configuration (all member disks must be
7236 * members of each array in the container).
7237 */
7a862a02
N		 7238 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7239 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7240 return 0;
5fe62b94
WD		 7241 } else if (super->orom && mpb->num_raid_devs > 0 &&
7242 mpb->num_disks != raiddisks) {
7a862a02 7243 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7244 return 0;
c2c087e6 7245 }
0dcecb2e
DW		 7246
7247 /* retrieve the largest free space block */
05501181 7248 e = get_extents(super, dl, 0);
c2c087e6
DW		 7249 maxsize = 0;
7250 i = 0;
0dcecb2e
DW		 7251 if (e) {
7252 do {
7253 unsigned long long esize;
7254
7255 esize = e[i].start - pos;
7256 if (esize >= maxsize)
7257 maxsize = esize;
7258 pos = e[i].start + e[i].size;
7259 i++;
7260 } while (e[i-1].size);
7261 dl->e = e;
7262 dl->extent_cnt = i;
7263 } else {
7264 if (verbose)
e7b84f9d 7265 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7266 dev);
7267 return 0;
7268 }
7269 if (maxsize < size) {
7270 if (verbose)
e7b84f9d 7271 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7272 dev, maxsize, size);
7273 return 0;
7274 }
7275
7276 /* count total number of extents for merge */
7277 i = 0;
7278 for (dl = super->disks; dl; dl = dl->next)
7279 if (dl->e)
7280 i += dl->extent_cnt;
7281
7282 maxsize = merge_extents(super, i);
3baa56ab 7283
1a1ced1e
KS		 7284 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7285 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7286
a7dd165b 7287 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7288 if (verbose) {
7289 if (maxsize == 0)
7a862a02 7290 pr_err("no free space left on device. Aborting...\n");
b3071342 7291 else
7a862a02 7292 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7293 maxsize, size);
7294 }
0dcecb2e 7295 return 0;
0dcecb2e
DW		 7296 }
7297
c2c087e6
DW		 7298 *freesize = maxsize;
7299
ca9de185 7300 if (super->orom) {
72a45777 7301 int count = count_volumes(super->hba,
ca9de185
LM		 7302 super->orom->dpa, verbose);
7303 if (super->orom->vphba <= count) {
676e87a8 7304 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7305 super->orom->vphba);
7306 return 0;
7307 }
7308 }
c2c087e6 7309 return 1;
cdddbdbc
DW		 7310}
7311
13bcac90 7312static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7313 unsigned long long size, int chunk,
7314 unsigned long long *freesize)
7315{
7316 struct intel_super *super = st->sb;
7317 struct imsm_super *mpb = super->anchor;
7318 struct dl *dl;
7319 int i;
7320 int extent_cnt;
7321 struct extent *e;
7322 unsigned long long maxsize;
7323 unsigned long long minsize;
7324 int cnt;
7325 int used;
7326
7327 /* find the largest common start free region of the possible disks */
7328 used = 0;
7329 extent_cnt = 0;
7330 cnt = 0;
7331 for (dl = super->disks; dl; dl = dl->next) {
7332 dl->raiddisk = -1;
7333
7334 if (dl->index >= 0)
7335 used++;
7336
7337 /* don't activate new spares if we are orom constrained
7338 * and there is already a volume active in the container
7339 */
7340 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7341 continue;
7342
05501181 7343 e = get_extents(super, dl, 0);
efb30e7f
DW		 7344 if (!e)
7345 continue;
7346 for (i = 1; e[i-1].size; i++)
7347 ;
7348 dl->e = e;
7349 dl->extent_cnt = i;
7350 extent_cnt += i;
7351 cnt++;
7352 }
7353
7354 maxsize = merge_extents(super, extent_cnt);
7355 minsize = size;
7356 if (size == 0)
612e59d8
CA		 7357 /* chunk is in K */
7358 minsize = chunk * 2;
efb30e7f
DW		 7359
7360 if (cnt < raiddisks ||
7361 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7362 maxsize < minsize ||
7363 maxsize == 0) {
e7b84f9d 7364 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7365 return 0; /* No enough free spaces large enough */
7366 }
7367
7368 if (size == 0) {
7369 size = maxsize;
7370 if (chunk) {
612e59d8
CA		 7371 size /= 2 * chunk;
7372 size *= 2 * chunk;
efb30e7f 7373 }
f878b242
LM		 7374 maxsize = size;
7375 }
1a1ced1e
KS		 7376 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7377 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7378 cnt = 0;
7379 for (dl = super->disks; dl; dl = dl->next)
7380 if (dl->e)
7381 dl->raiddisk = cnt++;
7382
7383 *freesize = size;
7384
13bcac90
AK		 7385 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7386
efb30e7f
DW		 7387 return 1;
7388}
7389
13bcac90
AK		 7390static int reserve_space(struct supertype *st, int raiddisks,
7391 unsigned long long size, int chunk,
7392 unsigned long long *freesize)
7393{
7394 struct intel_super *super = st->sb;
7395 struct dl *dl;
7396 int cnt;
7397 int rv = 0;
7398
7399 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7400 if (rv) {
7401 cnt = 0;
7402 for (dl = super->disks; dl; dl = dl->next)
7403 if (dl->e)
7404 dl->raiddisk = cnt++;
7405 rv = 1;
7406 }
7407
7408 return rv;
7409}
7410
bf5a934a 7411static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7412 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7413 unsigned long long data_offset,
bf5a934a 7414 char *dev, unsigned long long *freesize,
5308f117 7415 int consistency_policy, int verbose)
bf5a934a
DW		 7416{
7417 int fd, cfd;
7418 struct mdinfo *sra;
20cbe8d2 7419 int is_member = 0;
bf5a934a 7420
d54559f0
LM		 7421 /* load capability
7422 * if given unused devices create a container
bf5a934a
DW		 7423 * if given given devices in a container create a member volume
7424 */
7425 if (level == LEVEL_CONTAINER) {
7426 /* Must be a fresh device to add to a container */
7427 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7428 raiddisks,
7ccc4cc4 7429 *chunk,
af4348dd 7430 size, data_offset,
bf5a934a
DW		 7431 dev, freesize,
7432 verbose);
7433 }
9587c373 7434
06a6101c
BK		 7435 /*
7436 * Size is given in sectors.
7437 */
7438 if (size && (size < 2048)) {
22dc741f 7439 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7440 /* Depends on algorithm in Create.c :
7441 * if container was given (dev == NULL) return -1,
7442 * if block device was given ( dev != NULL) return 0.
7443 */
7444 return dev ? -1 : 0;
7445 }
7446
8592f29d 7447 if (!dev) {
e91a3bad 7448 if (st->sb) {
ca9de185 7449 struct intel_super *super = st->sb;
e91a3bad 7450 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7451 raiddisks, chunk, size,
e91a3bad
LM		 7452 verbose))
7453 return 0;
efb30e7f
DW		 7454 /* we are being asked to automatically layout a
7455 * new volume based on the current contents of
7456 * the container. If the the parameters can be
7457 * satisfied reserve_space will record the disks,
7458 * start offset, and size of the volume to be
7459 * created. add_to_super and getinfo_super
7460 * detect when autolayout is in progress.
7461 */
ca9de185
LM		 7462 /* assuming that freesize is always given when array is
7463 created */
7464 if (super->orom && freesize) {
7465 int count;
72a45777 7466 count = count_volumes(super->hba,
ca9de185
LM		 7467 super->orom->dpa, verbose);
7468 if (super->orom->vphba <= count) {
676e87a8 7469 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7470 super->orom->vphba);
7471 return 0;
7472 }
7473 }
e91a3bad
LM		 7474 if (freesize)
7475 return reserve_space(st, raiddisks, size,
7ccc4cc4 7476 *chunk, freesize);
8592f29d
N		 7477 }
7478 return 1;
7479 }
bf5a934a
DW		 7480 if (st->sb) {
7481 /* creating in a given container */
7482 return validate_geometry_imsm_volume(st, level, layout,
7483 raiddisks, chunk, size,
af4348dd 7484 data_offset,
bf5a934a
DW		 7485 dev, freesize, verbose);
7486 }
7487
bf5a934a
DW		 7488 /* This device needs to be a device in an 'imsm' container */
7489 fd = open(dev, O_RDONLY|O_EXCL, 0);
7490 if (fd >= 0) {
7491 if (verbose)
e7b84f9d
N		 7492 pr_err("Cannot create this array on device %s\n",
7493 dev);
bf5a934a
DW		 7494 close(fd);
7495 return 0;
7496 }
7497 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7498 if (verbose)
e7b84f9d 7499 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7500 dev, strerror(errno));
7501 return 0;
7502 }
7503 /* Well, it is in use by someone, maybe an 'imsm' container. */
7504 cfd = open_container(fd);
20cbe8d2 7505 close(fd);
bf5a934a 7506 if (cfd < 0) {
bf5a934a 7507 if (verbose)
e7b84f9d 7508 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7509 dev);
7510 return 0;
7511 }
4dd2df09 7512 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7513 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7514 strcmp(sra->text_version, "imsm") == 0)
7515 is_member = 1;
7516 sysfs_free(sra);
7517 if (is_member) {
bf5a934a
DW		 7518 /* This is a member of a imsm container. Load the container
7519 * and try to create a volume
7520 */
7521 struct intel_super *super;
7522
ec50f7b6 7523 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7524 st->sb = super;
4dd2df09 7525 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7526 close(cfd);
7527 return validate_geometry_imsm_volume(st, level, layout,
7528 raiddisks, chunk,
af4348dd 7529 size, data_offset, dev,
ecbd9e81
N		 7530 freesize, 1)
7531 ? 1 : -1;
bf5a934a 7532 }
20cbe8d2 7533 }
bf5a934a 7534
20cbe8d2 7535 if (verbose)
e7b84f9d 7536 pr_err("failed container membership check\n");
20cbe8d2
AW		 7537
7538 close(cfd);
7539 return 0;
bf5a934a 7540}
0bd16cf2 7541
30f58b22 7542static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7543{
7544 struct intel_super *super = st->sb;
7545
30f58b22
DW		 7546 if (level && *level == UnSet)
7547 *level = LEVEL_CONTAINER;
7548
7549 if (level && layout && *layout == UnSet)
7550 *layout = imsm_level_to_layout(*level);
0bd16cf2 7551
cd9d1ac7
DW		 7552 if (chunk && (*chunk == UnSet || *chunk == 0))
7553 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7554}
7555
33414a01
DW		 7556static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7557
3364781b 7558static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7559{
3364781b 7560 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7561 __u8 i;
7562 struct intel_dev **dp;
7563 struct intel_super *super = st->sb;
3364781b 7564 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7565 struct imsm_super *mpb = super->anchor;
7566
3364781b 7567 if (mpb->num_raid_devs == 0)
33414a01 7568 return 2;
33414a01
DW		 7569
7570 /* block deletions that would change the uuid of active subarrays
7571 *
7572 * FIXME when immutable ids are available, but note that we'll
7573 * also need to fixup the invalidated/active subarray indexes in
7574 * mdstat
7575 */
7576 for (i = 0; i < mpb->num_raid_devs; i++) {
7577 char subarray[4];
7578
7579 if (i < current_vol)
7580 continue;
7581 sprintf(subarray, "%u", i);
4dd2df09 7582 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7583 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7584 current_vol, i);
33414a01
DW		 7585
7586 return 2;
7587 }
7588 }
7589
7590 if (st->update_tail) {
503975b9 7591 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7592
33414a01
DW		 7593 u->type = update_kill_array;
7594 u->dev_idx = current_vol;
7595 append_metadata_update(st, u, sizeof(*u));
7596
7597 return 0;
7598 }
7599
7600 for (dp = &super->devlist; *dp;)
7601 if ((*dp)->index == current_vol) {
7602 *dp = (*dp)->next;
7603 } else {
7604 handle_missing(super, (*dp)->dev);
7605 if ((*dp)->index > current_vol)
7606 (*dp)->index--;
7607 dp = &(*dp)->next;
7608 }
7609
7610 /* no more raid devices, all active components are now spares,
7611 * but of course failed are still failed
7612 */
7613 if (--mpb->num_raid_devs == 0) {
7614 struct dl *d;
7615
7616 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7617 if (d->index > -2)
7618 mark_spare(d);
33414a01
DW		 7619 }
7620
7621 super->updates_pending++;
7622
7623 return 0;
7624}
aa534678 7625
a951a4f7 7626static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7627 char *update, struct mddev_ident *ident)
aa534678
DW		 7628{
7629 /* update the subarray currently referenced by ->current_vol */
7630 struct intel_super *super = st->sb;
7631 struct imsm_super *mpb = super->anchor;
7632
aa534678
DW		 7633 if (strcmp(update, "name") == 0) {
7634 char *name = ident->name;
a951a4f7
N		 7635 char *ep;
7636 int vol;
aa534678 7637
4dd2df09 7638 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7639 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7640 return 2;
7641 }
7642
7643 if (!check_name(super, name, 0))
7644 return 2;
7645
a951a4f7
N		 7646 vol = strtoul(subarray, &ep, 10);
7647 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7648 return 2;
7649
aa534678 7650 if (st->update_tail) {
503975b9 7651 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7652
aa534678 7653 u->type = update_rename_array;
a951a4f7 7654 u->dev_idx = vol;
618f4e6d
XN		 7655 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7656 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7657 append_metadata_update(st, u, sizeof(*u));
7658 } else {
7659 struct imsm_dev *dev;
ebad3af2 7660 int i, namelen;
aa534678 7661
a951a4f7 7662 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7663 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7664 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7665 memcpy(dev->volume, name, namelen);
aa534678
DW		 7666 for (i = 0; i < mpb->num_raid_devs; i++) {
7667 dev = get_imsm_dev(super, i);
7668 handle_missing(super, dev);
7669 }
7670 super->updates_pending++;
7671 }
e6e9dd3f
AP		 7672 } else if (strcmp(update, "ppl") == 0 ||
7673 strcmp(update, "no-ppl") == 0) {
7674 int new_policy;
7675 char *ep;
7676 int vol = strtoul(subarray, &ep, 10);
7677
7678 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7679 return 2;
7680
7681 if (strcmp(update, "ppl") == 0)
c2462068 7682 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7683 else
c2462068 7684 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7685
7686 if (st->update_tail) {
7687 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7688
7689 u->type = update_rwh_policy;
7690 u->dev_idx = vol;
7691 u->new_policy = new_policy;
7692 append_metadata_update(st, u, sizeof(*u));
7693 } else {
7694 struct imsm_dev *dev;
7695
7696 dev = get_imsm_dev(super, vol);
7697 dev->rwh_policy = new_policy;
7698 super->updates_pending++;
7699 }
aa534678
DW		 7700 } else
7701 return 2;
7702
7703 return 0;
7704}
bf5a934a 7705
28bce06f
AK		 7706static int is_gen_migration(struct imsm_dev *dev)
7707{
7534230b
AK		 7708 if (dev == NULL)
7709 return 0;
7710
28bce06f
AK		 7711 if (!dev->vol.migr_state)
7712 return 0;
7713
7714 if (migr_type(dev) == MIGR_GEN_MIGR)
7715 return 1;
7716
7717 return 0;
7718}
7719
1e5c6983
DW		 7720static int is_rebuilding(struct imsm_dev *dev)
7721{
7722 struct imsm_map *migr_map;
7723
7724 if (!dev->vol.migr_state)
7725 return 0;
7726
7727 if (migr_type(dev) != MIGR_REBUILD)
7728 return 0;
7729
238c0a71 7730 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7731
7732 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7733 return 1;
7734 else
7735 return 0;
7736}
7737
6ce1fbf1
AK		 7738static int is_initializing(struct imsm_dev *dev)
7739{
7740 struct imsm_map *migr_map;
7741
7742 if (!dev->vol.migr_state)
7743 return 0;
7744
7745 if (migr_type(dev) != MIGR_INIT)
7746 return 0;
7747
238c0a71 7748 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7749
7750 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7751 return 1;
7752
7753 return 0;
6ce1fbf1
AK		 7754}
7755
c47b0ff6
AK		 7756static void update_recovery_start(struct intel_super *super,
7757 struct imsm_dev *dev,
7758 struct mdinfo *array)
1e5c6983
DW		 7759{
7760 struct mdinfo *rebuild = NULL;
7761 struct mdinfo *d;
7762 __u32 units;
7763
7764 if (!is_rebuilding(dev))
7765 return;
7766
7767 /* Find the rebuild target, but punt on the dual rebuild case */
7768 for (d = array->devs; d; d = d->next)
7769 if (d->recovery_start == 0) {
7770 if (rebuild)
7771 return;
7772 rebuild = d;
7773 }
7774
4363fd80
DW		 7775 if (!rebuild) {
7776 /* (?) none of the disks are marked with
7777 * IMSM_ORD_REBUILD, so assume they are missing and the
7778 * disk_ord_tbl was not correctly updated
7779 */
1ade5cc1 7780 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7781 return;
7782 }
7783
1e5c6983 7784 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7785 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7786}
7787
276d77db 7788static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7789
00bbdbda 7790static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7791{
4f5bc454
DW		 7792 /* Given a container loaded by load_super_imsm_all,
7793 * extract information about all the arrays into
7794 * an mdinfo tree.
00bbdbda 7795 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7796 *
7797 * For each imsm_dev create an mdinfo, fill it in,
7798 * then look for matching devices in super->disks
7799 * and create appropriate device mdinfo.
7800 */
7801 struct intel_super *super = st->sb;
949c47a0 7802 struct imsm_super *mpb = super->anchor;
4f5bc454 7803 struct mdinfo *rest = NULL;
00bbdbda 7804 unsigned int i;
81219e70 7805 int sb_errors = 0;
abef11a3
AK		 7806 struct dl *d;
7807 int spare_disks = 0;
b6180160 7808 int current_vol = super->current_vol;
cdddbdbc 7809
19482bcc
AK		 7810 /* do not assemble arrays when not all attributes are supported */
7811 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7812 sb_errors = 1;
7a862a02 7813 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7814 }
7815
abef11a3
AK		 7816 /* count spare devices, not used in maps
7817 */
7818 for (d = super->disks; d; d = d->next)
7819 if (d->index == -1)
7820 spare_disks++;
7821
4f5bc454 7822 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7823 struct imsm_dev *dev;
7824 struct imsm_map *map;
86e3692b 7825 struct imsm_map *map2;
4f5bc454 7826 struct mdinfo *this;
a6482415 7827 int slot;
a6482415 7828 int chunk;
00bbdbda 7829 char *ep;
8b9cd157 7830 int level;
00bbdbda
N		 7831
7832 if (subarray &&
7833 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7834 continue;
7835
7836 dev = get_imsm_dev(super, i);
238c0a71
AK		 7837 map = get_imsm_map(dev, MAP_0);
7838 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7839 level = get_imsm_raid_level(map);
4f5bc454 7840
1ce0101c
DW		 7841 /* do not publish arrays that are in the middle of an
7842 * unsupported migration
7843 */
7844 if (dev->vol.migr_state &&
28bce06f 7845 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7846 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7847 dev->volume);
7848 continue;
7849 }
2db86302
LM		 7850 /* do not publish arrays that are not support by controller's
7851 * OROM/EFI
7852 */
1ce0101c 7853
503975b9 7854 this = xmalloc(sizeof(*this));
4f5bc454 7855
301406c9 7856 super->current_vol = i;
a5d85af7 7857 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7858 this->next = rest;
a6482415 7859 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7860 /* mdadm does not support all metadata features- set the bit in all arrays state */
7861 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7862 level, /* RAID level */
7863 imsm_level_to_layout(level),
81219e70 7864 map->num_members, /* raid disks */
fcc2c9da 7865 &chunk, imsm_dev_size(dev),
81219e70 7866 1 /* verbose */)) {
7a862a02 7867 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7868 dev->volume);
7869 this->array.state |=
7870 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7871 (1<<MD_SB_BLOCK_VOLUME);
7872 }
81219e70
LM		 7873
7874 /* if array has bad blocks, set suitable bit in all arrays state */
7875 if (sb_errors)
7876 this->array.state |=
7877 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7878 (1<<MD_SB_BLOCK_VOLUME);
7879
4f5bc454 7880 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7881 unsigned long long recovery_start;
4f5bc454
DW		 7882 struct mdinfo *info_d;
7883 struct dl *d;
7884 int idx;
9a1608e5 7885 int skip;
7eef0453 7886 __u32 ord;
8b9cd157 7887 int missing = 0;
4f5bc454 7888
9a1608e5 7889 skip = 0;
238c0a71
AK		 7890 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7891 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7892 for (d = super->disks; d ; d = d->next)
7893 if (d->index == idx)
0fbd635c 7894 break;
4f5bc454 7895
1e5c6983 7896 recovery_start = MaxSector;
4f5bc454 7897 if (d == NULL)
9a1608e5 7898 skip = 1;
25ed7e59 7899 if (d && is_failed(&d->disk))
9a1608e5 7900 skip = 1;
8b9cd157 7901 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7902 recovery_start = 0;
1e93d0d1
BK		 7903 if (!(ord & IMSM_ORD_REBUILD))
7904 this->array.working_disks++;
1011e834 7905 /*
9a1608e5 7906 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7907 * reset resync start to avoid a dirty-degraded
7908 * situation when performing the intial sync
9a1608e5 7909 */
8b9cd157
MK		 7910 if (skip)
7911 missing++;
7912
7913 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7914 if ((!able_to_resync(level, missing) ||
7915 recovery_start == 0))
7916 this->resync_start = MaxSector;
7917 } else {
7918 /*
7919 * FIXME handle dirty degraded
7920 */
7921 }
7922
9a1608e5
DW		 7923 if (skip)
7924 continue;
4f5bc454 7925
503975b9 7926 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7927 info_d->next = this->devs;
7928 this->devs = info_d;
7929
4f5bc454
DW		 7930 info_d->disk.number = d->index;
7931 info_d->disk.major = d->major;
7932 info_d->disk.minor = d->minor;
7933 info_d->disk.raid_disk = slot;
1e5c6983 7934 info_d->recovery_start = recovery_start;
86e3692b
AK		 7935 if (map2) {
7936 if (slot < map2->num_members)
7937 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7938 else
7939 this->array.spare_disks++;
86e3692b
AK		 7940 } else {
7941 if (slot < map->num_members)
7942 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7943 else
7944 this->array.spare_disks++;
86e3692b 7945 }
4f5bc454
DW		 7946
7947 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7948 info_d->data_offset = pba_of_lba0(map);
44490938 7949 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7950
7951 if (map->raid_level == 5) {
2432ce9b
AP		 7952 info_d->ppl_sector = this->ppl_sector;
7953 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7954 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7955 recovery_start == 0)
7956 this->resync_start = 0;
06fb291a 7957 }
b12796be 7958
5e46202e 7959 info_d->bb.supported = 1;
b12796be
TM		 7960 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7961 info_d->data_offset,
7962 info_d->component_size,
7963 &info_d->bb);
4f5bc454 7964 }
1e5c6983 7965 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7966 update_recovery_start(super, dev, this);
abef11a3 7967 this->array.spare_disks += spare_disks;
276d77db
AK		 7968
7969 /* check for reshape */
7970 if (this->reshape_active == 1)
7971 recover_backup_imsm(st, this);
9a1608e5 7972 rest = this;
4f5bc454
DW		 7973 }
7974
b6180160 7975 super->current_vol = current_vol;
4f5bc454 7976 return rest;
cdddbdbc
DW		 7977}
7978
3b451610
AK		 7979static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7980 int failed, int look_in_map)
c2a1e7da 7981{
3b451610
AK		 7982 struct imsm_map *map;
7983
7984 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7985
7986 if (!failed)
1011e834 7987 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7988 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7989
7990 switch (get_imsm_raid_level(map)) {
7991 case 0:
7992 return IMSM_T_STATE_FAILED;
7993 break;
7994 case 1:
7995 if (failed < map->num_members)
7996 return IMSM_T_STATE_DEGRADED;
7997 else
7998 return IMSM_T_STATE_FAILED;
7999 break;
8000 case 10:
8001 {
8002 /**
c92a2527
DW		 8003 * check to see if any mirrors have failed, otherwise we
8004 * are degraded. Even numbered slots are mirrored on
8005 * slot+1
c2a1e7da 8006 */
c2a1e7da 8007 int i;
d9b420a5
N		 8008 /* gcc -Os complains that this is unused */
8009 int insync = insync;
c2a1e7da
DW		 8010
8011 for (i = 0; i < map->num_members; i++) {
238c0a71 8012 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8013 int idx = ord_to_idx(ord);
8014 struct imsm_disk *disk;
c2a1e7da 8015
c92a2527 8016 /* reset the potential in-sync count on even-numbered
1011e834 8017 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8018 */
8019 if ((i & 1) == 0)
8020 insync = 2;
c2a1e7da 8021
c92a2527 8022 disk = get_imsm_disk(super, idx);
25ed7e59 8023 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8024 insync--;
c2a1e7da 8025
c92a2527
DW		 8026 /* no in-sync disks left in this mirror the
8027 * array has failed
8028 */
8029 if (insync == 0)
8030 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8031 }
8032
8033 return IMSM_T_STATE_DEGRADED;
8034 }
8035 case 5:
8036 if (failed < 2)
8037 return IMSM_T_STATE_DEGRADED;
8038 else
8039 return IMSM_T_STATE_FAILED;
8040 break;
8041 default:
8042 break;
8043 }
8044
8045 return map->map_state;
8046}
8047
3b451610
AK		 8048static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8049 int look_in_map)
c2a1e7da
DW		 8050{
8051 int i;
8052 int failed = 0;
8053 struct imsm_disk *disk;
d5985138
AK		 8054 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8055 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8056 struct imsm_map *map_for_loop;
0556e1a2
DW		 8057 __u32 ord;
8058 int idx;
d5985138 8059 int idx_1;
c2a1e7da 8060
0556e1a2
DW		 8061 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8062 * disks that are being rebuilt. New failures are recorded to
8063 * map[0]. So we look through all the disks we started with and
8064 * see if any failures are still present, or if any new ones
8065 * have arrived
0556e1a2 8066 */
d5985138
AK		 8067 map_for_loop = map;
8068 if (prev && (map->num_members < prev->num_members))
8069 map_for_loop = prev;
68fe4598
LD		 8070
8071 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8072 idx_1 = -255;
238c0a71
AK		 8073 /* when MAP_X is passed both maps failures are counted
8074 */
d5985138 8075 if (prev &&
089f9d79
JS		 8076 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8077 i < prev->num_members) {
d5985138
AK		 8078 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8079 idx_1 = ord_to_idx(ord);
c2a1e7da 8080
d5985138
AK		 8081 disk = get_imsm_disk(super, idx_1);
8082 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8083 failed++;
8084 }
089f9d79
JS		 8085 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8086 i < map->num_members) {
d5985138
AK		 8087 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8088 idx = ord_to_idx(ord);
8089
8090 if (idx != idx_1) {
8091 disk = get_imsm_disk(super, idx);
8092 if (!disk || is_failed(disk) ||
8093 ord & IMSM_ORD_REBUILD)
8094 failed++;
8095 }
8096 }
c2a1e7da
DW		 8097 }
8098
8099 return failed;
845dea95
NB		 8100}
8101
97b4d0e9
DW		 8102static int imsm_open_new(struct supertype *c, struct active_array *a,
8103 char *inst)
8104{
8105 struct intel_super *super = c->sb;
8106 struct imsm_super *mpb = super->anchor;
bbab0940 8107 struct imsm_update_prealloc_bb_mem u;
9587c373 8108
97b4d0e9 8109 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8110 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8111 return -ENODEV;
8112 }
8113
8114 dprintf("imsm: open_new %s\n", inst);
8115 a->info.container_member = atoi(inst);
bbab0940
TM		 8116
8117 u.type = update_prealloc_badblocks_mem;
8118 imsm_update_metadata_locally(c, &u, sizeof(u));
8119
97b4d0e9
DW		 8120 return 0;
8121}
8122
0c046afd
DW		 8123static int is_resyncing(struct imsm_dev *dev)
8124{
8125 struct imsm_map *migr_map;
8126
8127 if (!dev->vol.migr_state)
8128 return 0;
8129
1484e727
DW		 8130 if (migr_type(dev) == MIGR_INIT ||
8131 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8132 return 1;
8133
4c9bc37b
AK		 8134 if (migr_type(dev) == MIGR_GEN_MIGR)
8135 return 0;
8136
238c0a71 8137 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8138
089f9d79
JS		 8139 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8140 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8141 return 1;
8142 else
8143 return 0;
8144}
8145
0556e1a2 8146/* return true if we recorded new information */
4c9e8c1e
TM		 8147static int mark_failure(struct intel_super *super,
8148 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8149{
0556e1a2
DW		 8150 __u32 ord;
8151 int slot;
8152 struct imsm_map *map;
86c54047
DW		 8153 char buf[MAX_RAID_SERIAL_LEN+3];
8154 unsigned int len, shift = 0;
0556e1a2
DW		 8155
8156 /* new failures are always set in map[0] */
238c0a71 8157 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8158
8159 slot = get_imsm_disk_slot(map, idx);
8160 if (slot < 0)
8161 return 0;
8162
8163 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8164 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8165 return 0;
8166
7d0c5e24
LD		 8167 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8168 buf[MAX_RAID_SERIAL_LEN] = '\000';
8169 strcat(buf, ":0");
86c54047
DW		 8170 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8171 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8172 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8173
f2f27e63 8174 disk->status |= FAILED_DISK;
0556e1a2 8175 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8176 /* mark failures in second map if second map exists and this disk
8177 * in this slot.
8178 * This is valid for migration, initialization and rebuild
8179 */
8180 if (dev->vol.migr_state) {
238c0a71 8181 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8182 int slot2 = get_imsm_disk_slot(map2, idx);
8183
089f9d79 8184 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8185 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8186 idx | IMSM_ORD_REBUILD);
8187 }
d7a1fda2
MT		 8188 if (map->failed_disk_num == 0xff ||
8189 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8190 map->failed_disk_num = slot;
4c9e8c1e
TM		 8191
8192 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8193
0556e1a2
DW		 8194 return 1;
8195}
8196
4c9e8c1e
TM		 8197static void mark_missing(struct intel_super *super,
8198 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8199{
4c9e8c1e 8200 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8201
8202 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8203 return;
8204
47ee5a45
DW		 8205 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8206 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8207}
8208
33414a01
DW		 8209static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8210{
33414a01 8211 struct dl *dl;
33414a01
DW		 8212
8213 if (!super->missing)
8214 return;
33414a01 8215
79b68f1b
PC		 8216 /* When orom adds replacement for missing disk it does
8217 * not remove entry of missing disk, but just updates map with
8218 * new added disk. So it is not enough just to test if there is
8219 * any missing disk, we have to look if there are any failed disks
8220 * in map to stop migration */
8221
33414a01 8222 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8223 /* end process for initialization and rebuild only
8224 */
8225 if (is_gen_migration(dev) == 0) {
fb12a745 8226 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8227
fb12a745
TM		 8228 if (failed) {
8229 __u8 map_state;
8230 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8231 struct imsm_map *map1;
8232 int i, ord, ord_map1;
8233 int rebuilt = 1;
3d59f0c0 8234
fb12a745
TM		 8235 for (i = 0; i < map->num_members; i++) {
8236 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8237 if (!(ord & IMSM_ORD_REBUILD))
8238 continue;
8239
8240 map1 = get_imsm_map(dev, MAP_1);
8241 if (!map1)
8242 continue;
8243
8244 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8245 if (ord_map1 & IMSM_ORD_REBUILD)
8246 rebuilt = 0;
8247 }
8248
8249 if (rebuilt) {
8250 map_state = imsm_check_degraded(super, dev,
8251 failed, MAP_0);
8252 end_migration(dev, super, map_state);
8253 }
8254 }
3d59f0c0 8255 }
33414a01 8256 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8257 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8258 super->updates_pending++;
8259}
8260
f3871fdc
AK		 8261static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8262 long long new_size)
70bdf0dc 8263{
70bdf0dc 8264 unsigned long long array_blocks;
9529d343
MD		 8265 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8266 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8267
8268 if (used_disks == 0) {
8269 /* when problems occures
8270 * return current array_blocks value
8271 */
fcc2c9da 8272 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8273
8274 return array_blocks;
8275 }
8276
8277 /* set array size in metadata
8278 */
9529d343 8279 if (new_size <= 0)
f3871fdc
AK		 8280 /* OLCE size change is caused by added disks
8281 */
44490938 8282 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8283 else
f3871fdc
AK		 8284 /* Online Volume Size Change
8285 * Using available free space
8286 */
8287 array_blocks = new_size;
70bdf0dc 8288
b53bfba6 8289 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8290 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8291
8292 return array_blocks;
8293}
8294
28bce06f
AK		 8295static void imsm_set_disk(struct active_array *a, int n, int state);
8296
0e2d1a4e
AK		 8297static void imsm_progress_container_reshape(struct intel_super *super)
8298{
8299 /* if no device has a migr_state, but some device has a
8300 * different number of members than the previous device, start
8301 * changing the number of devices in this device to match
8302 * previous.
8303 */
8304 struct imsm_super *mpb = super->anchor;
8305 int prev_disks = -1;
8306 int i;
1dfaa380 8307 int copy_map_size;
0e2d1a4e
AK		 8308
8309 for (i = 0; i < mpb->num_raid_devs; i++) {
8310 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8311 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8312 struct imsm_map *map2;
8313 int prev_num_members;
0e2d1a4e
AK		 8314
8315 if (dev->vol.migr_state)
8316 return;
8317
8318 if (prev_disks == -1)
8319 prev_disks = map->num_members;
8320 if (prev_disks == map->num_members)
8321 continue;
8322
8323 /* OK, this array needs to enter reshape mode.
8324 * i.e it needs a migr_state
8325 */
8326
1dfaa380 8327 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8328 prev_num_members = map->num_members;
8329 map->num_members = prev_disks;
8330 dev->vol.migr_state = 1;
8331 dev->vol.curr_migr_unit = 0;
ea672ee1 8332 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8333 for (i = prev_num_members;
8334 i < map->num_members; i++)
8335 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8336 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8337 /* Copy the current map */
1dfaa380 8338 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8339 map2->num_members = prev_num_members;
8340
f3871fdc 8341 imsm_set_array_size(dev, -1);
51d83f5d 8342 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8343 super->updates_pending++;
8344 }
8345}
8346
aad6f216 8347/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8348 * states are handled in imsm_set_disk() with one exception, when a
8349 * resync is stopped due to a new failure this routine will set the
8350 * 'degraded' state for the array.
8351 */
01f157d7 8352static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8353{
8354 int inst = a->info.container_member;
8355 struct intel_super *super = a->container->sb;
949c47a0 8356 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8357 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8358 int failed = imsm_count_failed(super, dev, MAP_0);
8359 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8360 __u32 blocks_per_unit;
a862209d 8361
1af97990
AK		 8362 if (dev->vol.migr_state &&
8363 dev->vol.migr_type == MIGR_GEN_MIGR) {
8364 /* array state change is blocked due to reshape action
aad6f216
N		 8365 * We might need to
8366 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8367 * - finish the reshape (if last_checkpoint is big and action != reshape)
8368 * - update curr_migr_unit
1af97990 8369 */
aad6f216
N		 8370 if (a->curr_action == reshape) {
8371 /* still reshaping, maybe update curr_migr_unit */
633b5610 8372 goto mark_checkpoint;
aad6f216
N		 8373 } else {
8374 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8375 /* for some reason we aborted the reshape.
b66e591b
AK		 8376 *
8377 * disable automatic metadata rollback
8378 * user action is required to recover process
aad6f216 8379 */
b66e591b 8380 if (0) {
238c0a71
AK		 8381 struct imsm_map *map2 =
8382 get_imsm_map(dev, MAP_1);
8383 dev->vol.migr_state = 0;
8384 set_migr_type(dev, 0);
8385 dev->vol.curr_migr_unit = 0;
8386 memcpy(map, map2,
8387 sizeof_imsm_map(map2));
8388 super->updates_pending++;
b66e591b 8389 }
aad6f216
N		 8390 }
8391 if (a->last_checkpoint >= a->info.component_size) {
8392 unsigned long long array_blocks;
8393 int used_disks;
e154ced3 8394 struct mdinfo *mdi;
aad6f216 8395
9529d343 8396 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8397 if (used_disks > 0) {
8398 array_blocks =
44490938 8399 per_dev_array_size(map) *
d55adef9 8400 used_disks;
b53bfba6
TM		 8401 array_blocks =
8402 round_size_to_mb(array_blocks,
8403 used_disks);
d55adef9
AK		 8404 a->info.custom_array_size = array_blocks;
8405 /* encourage manager to update array
8406 * size
8407 */
e154ced3 8408
d55adef9 8409 a->check_reshape = 1;
633b5610 8410 }
e154ced3
AK		 8411 /* finalize online capacity expansion/reshape */
8412 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8413 imsm_set_disk(a,
8414 mdi->disk.raid_disk,
8415 mdi->curr_state);
8416
0e2d1a4e 8417 imsm_progress_container_reshape(super);
e154ced3 8418 }
aad6f216 8419 }
1af97990
AK		 8420 }
8421
47ee5a45 8422 /* before we activate this array handle any missing disks */
33414a01
DW		 8423 if (consistent == 2)
8424 handle_missing(super, dev);
1e5c6983 8425
0c046afd 8426 if (consistent == 2 &&
b7941fd6 8427 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8428 map_state != IMSM_T_STATE_NORMAL ||
8429 dev->vol.migr_state))
01f157d7 8430 consistent = 0;
272906ef 8431
b7941fd6 8432 if (is_resync_complete(&a->info)) {
0c046afd 8433 /* complete intialization / resync,
0556e1a2
DW		 8434 * recovery and interrupted recovery is completed in
8435 * ->set_disk
0c046afd
DW		 8436 */
8437 if (is_resyncing(dev)) {
8438 dprintf("imsm: mark resync done\n");
809da78e 8439 end_migration(dev, super, map_state);
115c3803 8440 super->updates_pending++;
484240d8 8441 a->last_checkpoint = 0;
115c3803 8442 }
b9172665
AK		 8443 } else if ((!is_resyncing(dev) && !failed) &&
8444 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8445 /* mark the start of the init process if nothing is failed */
b7941fd6 8446 dprintf("imsm: mark resync start\n");
1484e727 8447 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8448 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8449 else
8e59f3d8 8450 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8451 super->updates_pending++;
115c3803 8452 }
a862209d 8453
633b5610 8454mark_checkpoint:
5b83bacf
AK		 8455 /* skip checkpointing for general migration,
8456 * it is controlled in mdadm
8457 */
8458 if (is_gen_migration(dev))
8459 goto skip_mark_checkpoint;
8460
1e5c6983 8461 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8462 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8463 if (blocks_per_unit) {
1e5c6983
DW		 8464 __u32 units32;
8465 __u64 units;
8466
4f0a7acc 8467 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8468 units32 = units;
8469
8470 /* check that we did not overflow 32-bits, and that
8471 * curr_migr_unit needs updating
8472 */
8473 if (units32 == units &&
bfd80a56 8474 units32 != 0 &&
1e5c6983
DW		 8475 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8476 dprintf("imsm: mark checkpoint (%u)\n", units32);
8477 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8478 super->updates_pending++;
8479 }
8480 }
f8f603f1 8481
5b83bacf 8482skip_mark_checkpoint:
3393c6af 8483 /* mark dirty / clean */
2432ce9b
AP		 8484 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8485 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8486 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8487 if (consistent) {
8488 dev->vol.dirty = RAIDVOL_CLEAN;
8489 } else {
8490 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8491 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8492 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8493 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8494 }
a862209d
DW		 8495 super->updates_pending++;
8496 }
28bce06f 8497
01f157d7 8498 return consistent;
a862209d
DW		 8499}
8500
6f50473f
TM		 8501static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8502{
8503 int inst = a->info.container_member;
8504 struct intel_super *super = a->container->sb;
8505 struct imsm_dev *dev = get_imsm_dev(super, inst);
8506 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8507
8508 if (slot > map->num_members) {
8509 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8510 slot, map->num_members - 1);
8511 return -1;
8512 }
8513
8514 if (slot < 0)
8515 return -1;
8516
8517 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8518}
8519
8d45d196 8520static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8521{
8d45d196
DW		 8522 int inst = a->info.container_member;
8523 struct intel_super *super = a->container->sb;
949c47a0 8524 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8525 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8526 struct imsm_disk *disk;
7ce05701
LD		 8527 struct mdinfo *mdi;
8528 int recovery_not_finished = 0;
0c046afd 8529 int failed;
6f50473f 8530 int ord;
0c046afd 8531 __u8 map_state;
fb12a745
TM		 8532 int rebuild_done = 0;
8533 int i;
8d45d196 8534
fb12a745 8535 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8536 if (ord < 0)
8d45d196
DW		 8537 return;
8538
4e6e574a 8539 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8540 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8541
5802a811 8542 /* check for new failures */
ae7d61e3 8543 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8544 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8545 super->updates_pending++;
8d45d196 8546 }
47ee5a45 8547
19859edc 8548 /* check if in_sync */
0556e1a2 8549 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8550 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8551
8552 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8553 rebuild_done = 1;
19859edc
DW		 8554 super->updates_pending++;
8555 }
8d45d196 8556
3b451610
AK		 8557 failed = imsm_count_failed(super, dev, MAP_0);
8558 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8559
0c046afd 8560 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8561 dprintf("imsm: Detected transition to state ");
8562 switch (map_state) {
8563 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8564 dprintf("normal: ");
8565 if (is_rebuilding(dev)) {
1ade5cc1 8566 dprintf_cont("while rebuilding");
7ce05701
LD		 8567 /* check if recovery is really finished */
8568 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8569 if (mdi->recovery_start != MaxSector) {
8570 recovery_not_finished = 1;
8571 break;
8572 }
8573 if (recovery_not_finished) {
1ade5cc1
N		 8574 dprintf_cont("\n");
8575 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8576 if (a->last_checkpoint < mdi->recovery_start) {
8577 a->last_checkpoint = mdi->recovery_start;
8578 super->updates_pending++;
8579 }
8580 break;
8581 }
94002678 8582 end_migration(dev, super, map_state);
238c0a71 8583 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8584 map->failed_disk_num = ~0;
8585 super->updates_pending++;
8586 a->last_checkpoint = 0;
8587 break;
8588 }
8589 if (is_gen_migration(dev)) {
1ade5cc1 8590 dprintf_cont("while general migration");
bf2f0071 8591 if (a->last_checkpoint >= a->info.component_size)
809da78e 8592 end_migration(dev, super, map_state);
94002678
AK		 8593 else
8594 map->map_state = map_state;
238c0a71 8595 map = get_imsm_map(dev, MAP_0);
28bce06f 8596 map->failed_disk_num = ~0;
94002678 8597 super->updates_pending++;
bf2f0071 8598 break;
94002678
AK		 8599 }
8600 break;
8601 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8602 dprintf_cont("degraded: ");
089f9d79 8603 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8604 dprintf_cont("mark degraded");
94002678
AK		 8605 map->map_state = map_state;
8606 super->updates_pending++;
8607 a->last_checkpoint = 0;
8608 break;
8609 }
8610 if (is_rebuilding(dev)) {
d7a1fda2 8611 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8612 if (state & DS_FAULTY) {
8613 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8614 if (n == map->failed_disk_num) {
8615 dprintf_cont("end migration");
8616 end_migration(dev, super, map_state);
a4e96fd8 8617 a->last_checkpoint = 0;
d7a1fda2 8618 } else {
a4e96fd8 8619 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8620 map->map_state = map_state;
8621 }
94002678 8622 super->updates_pending++;
fb12a745
TM		 8623 }
8624
a4e96fd8
MT		 8625 if (!rebuild_done)
8626 break;
8627
fb12a745
TM		 8628 /* check if recovery is really finished */
8629 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8630 if (mdi->recovery_start != MaxSector) {
8631 recovery_not_finished = 1;
8632 break;
8633 }
8634 if (recovery_not_finished) {
8635 dprintf_cont("\n");
a4e96fd8 8636 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8637 if (a->last_checkpoint < mdi->recovery_start) {
8638 a->last_checkpoint =
8639 mdi->recovery_start;
8640 super->updates_pending++;
8641 }
8642 break;
94002678 8643 }
fb12a745
TM		 8644
8645 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8646 end_migration(dev, super, map_state);
8647 a->last_checkpoint = 0;
8648 super->updates_pending++;
fb12a745
TM		 8649
8650 for (i = 0; i < map->num_members; i++) {
8651 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8652
8653 if (idx & IMSM_ORD_REBUILD)
8654 map->failed_disk_num = i;
8655 }
8656 super->updates_pending++;
94002678
AK		 8657 break;
8658 }
8659 if (is_gen_migration(dev)) {
1ade5cc1 8660 dprintf_cont("while general migration");
bf2f0071 8661 if (a->last_checkpoint >= a->info.component_size)
809da78e 8662 end_migration(dev, super, map_state);
94002678
AK		 8663 else {
8664 map->map_state = map_state;
3b451610 8665 manage_second_map(super, dev);
94002678
AK		 8666 }
8667 super->updates_pending++;
bf2f0071 8668 break;
28bce06f 8669 }
6ce1fbf1 8670 if (is_initializing(dev)) {
1ade5cc1 8671 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8672 map->map_state = map_state;
8673 super->updates_pending++;
8674 break;
8675 }
94002678
AK		 8676 break;
8677 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8678 dprintf_cont("failed: ");
94002678 8679 if (is_gen_migration(dev)) {
1ade5cc1 8680 dprintf_cont("while general migration");
94002678
AK		 8681 map->map_state = map_state;
8682 super->updates_pending++;
8683 break;
8684 }
8685 if (map->map_state != map_state) {
1ade5cc1 8686 dprintf_cont("mark failed");
94002678
AK		 8687 end_migration(dev, super, map_state);
8688 super->updates_pending++;
8689 a->last_checkpoint = 0;
8690 break;
8691 }
8692 break;
8693 default:
1ade5cc1 8694 dprintf_cont("state %i\n", map_state);
5802a811 8695 }
1ade5cc1 8696 dprintf_cont("\n");
845dea95
NB		 8697}
8698
f796af5d 8699static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8700{
f796af5d 8701 void *buf = mpb;
c2a1e7da
DW		 8702 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8703 unsigned long long dsize;
8704 unsigned long long sectors;
f36a9ecd 8705 unsigned int sector_size;
c2a1e7da 8706
f36a9ecd 8707 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8708 get_dev_size(fd, NULL, &dsize);
8709
f36a9ecd 8710 if (mpb_size > sector_size) {
272f648f 8711 /* -1 to account for anchor */
f36a9ecd 8712 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8713
272f648f 8714 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8715 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8716 SEEK_SET) < 0)
272f648f 8717 return 1;
c2a1e7da 8718
f36a9ecd
PB		 8719 if ((unsigned long long)write(fd, buf + sector_size,
8720 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8721 return 1;
8722 }
c2a1e7da 8723
272f648f 8724 /* first block is stored on second to last sector of the disk */
f36a9ecd 8725 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8726 return 1;
8727
466070ad 8728 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8729 return 1;
8730
c2a1e7da
DW		 8731 return 0;
8732}
8733
2e735d19 8734static void imsm_sync_metadata(struct supertype *container)
845dea95 8735{
2e735d19 8736 struct intel_super *super = container->sb;
c2a1e7da 8737
1a64be56 8738 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8739 if (!super->updates_pending)
8740 return;
8741
36988a3d 8742 write_super_imsm(container, 0);
c2a1e7da
DW		 8743
8744 super->updates_pending = 0;
845dea95
NB		 8745}
8746
272906ef
DW		 8747static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8748{
8749 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8750 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8751 struct dl *dl;
8752
8753 for (dl = super->disks; dl; dl = dl->next)
8754 if (dl->index == i)
8755 break;
8756
25ed7e59 8757 if (dl && is_failed(&dl->disk))
272906ef
DW		 8758 dl = NULL;
8759
8760 if (dl)
1ade5cc1 8761 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8762
8763 return dl;
8764}
8765
a20d2ba5 8766static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8767 struct active_array *a, int activate_new,
8768 struct mdinfo *additional_test_list)
272906ef
DW		 8769{
8770 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8771 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8772 struct imsm_super *mpb = super->anchor;
8773 struct imsm_map *map;
272906ef
DW		 8774 unsigned long long pos;
8775 struct mdinfo *d;
8776 struct extent *ex;
a20d2ba5 8777 int i, j;
272906ef 8778 int found;
569cc43f
DW		 8779 __u32 array_start = 0;
8780 __u32 array_end = 0;
272906ef 8781 struct dl *dl;
6c932028 8782 struct mdinfo *test_list;
272906ef
DW		 8783
8784 for (dl = super->disks; dl; dl = dl->next) {
8785 /* If in this array, skip */
8786 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8787 if (d->state_fd >= 0 &&
8788 d->disk.major == dl->major &&
272906ef 8789 d->disk.minor == dl->minor) {
8ba77d32
AK		 8790 dprintf("%x:%x already in array\n",
8791 dl->major, dl->minor);
272906ef
DW		 8792 break;
8793 }
8794 if (d)
8795 continue;
6c932028
AK		 8796 test_list = additional_test_list;
8797 while (test_list) {
8798 if (test_list->disk.major == dl->major &&
8799 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8800 dprintf("%x:%x already in additional test list\n",
8801 dl->major, dl->minor);
8802 break;
8803 }
6c932028 8804 test_list = test_list->next;
8ba77d32 8805 }
6c932028 8806 if (test_list)
8ba77d32 8807 continue;
272906ef 8808
e553d2a4 8809 /* skip in use or failed drives */
25ed7e59 8810 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8811 dl->index == -2) {
8812 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8813 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8814 continue;
8815 }
8816
a20d2ba5
DW		 8817 /* skip pure spares when we are looking for partially
8818 * assimilated drives
8819 */
8820 if (dl->index == -1 && !activate_new)
8821 continue;
8822
f2cc4f7d
AO		 8823 if (!drive_validate_sector_size(super, dl))
8824 continue;
8825
272906ef 8826 /* Does this unused device have the requisite free space?
a20d2ba5 8827 * It needs to be able to cover all member volumes
272906ef 8828 */
05501181 8829 ex = get_extents(super, dl, 1);
272906ef
DW		 8830 if (!ex) {
8831 dprintf("cannot get extents\n");
8832 continue;
8833 }
a20d2ba5
DW		 8834 for (i = 0; i < mpb->num_raid_devs; i++) {
8835 dev = get_imsm_dev(super, i);
238c0a71 8836 map = get_imsm_map(dev, MAP_0);
272906ef 8837
a20d2ba5
DW		 8838 /* check if this disk is already a member of
8839 * this array
272906ef 8840 */
620b1713 8841 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8842 continue;
8843
8844 found = 0;
8845 j = 0;
8846 pos = 0;
5551b113 8847 array_start = pba_of_lba0(map);
329c8278 8848 array_end = array_start +
44490938 8849 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8850
8851 do {
8852 /* check that we can start at pba_of_lba0 with
44490938 8853 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8854 */
329c8278 8855 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8856 found = 1;
8857 break;
8858 }
8859 pos = ex[j].start + ex[j].size;
8860 j++;
8861 } while (ex[j-1].size);
8862
8863 if (!found)
272906ef 8864 break;
a20d2ba5 8865 }
272906ef
DW		 8866
8867 free(ex);
a20d2ba5 8868 if (i < mpb->num_raid_devs) {
329c8278
DW		 8869 dprintf("%x:%x does not have %u to %u available\n",
8870 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8871 /* No room */
8872 continue;
a20d2ba5
DW		 8873 }
8874 return dl;
272906ef
DW		 8875 }
8876
8877 return dl;
8878}
8879
95d07a2c
LM		 8880static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8881{
8882 struct imsm_dev *dev2;
8883 struct imsm_map *map;
8884 struct dl *idisk;
8885 int slot;
8886 int idx;
8887 __u8 state;
8888
8889 dev2 = get_imsm_dev(cont->sb, dev_idx);
8890 if (dev2) {
238c0a71 8891 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8892 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8893 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8894 if (!map)
8895 return 1;
8896 for (slot = 0; slot < map->num_members; slot++) {
8897 /*
8898 * Check if failed disks are deleted from intel
8899 * disk list or are marked to be deleted
8900 */
238c0a71 8901 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8902 idisk = get_imsm_dl_disk(cont->sb, idx);
8903 /*
8904 * Do not rebuild the array if failed disks
8905 * from failed sub-array are not removed from
8906 * container.
8907 */
8908 if (idisk &&
8909 is_failed(&idisk->disk) &&
8910 (idisk->action != DISK_REMOVE))
8911 return 0;
8912 }
8913 }
8914 }
8915 return 1;
8916}
8917
88758e9d
DW		 8918static struct mdinfo *imsm_activate_spare(struct active_array *a,
8919 struct metadata_update **updates)
8920{
8921 /**
d23fe947
DW		 8922 * Find a device with unused free space and use it to replace a
8923 * failed/vacant region in an array. We replace failed regions one a
8924 * array at a time. The result is that a new spare disk will be added
8925 * to the first failed array and after the monitor has finished
8926 * propagating failures the remainder will be consumed.
88758e9d 8927 *
d23fe947
DW		 8928 * FIXME add a capability for mdmon to request spares from another
8929 * container.
88758e9d
DW		 8930 */
8931
8932 struct intel_super *super = a->container->sb;
88758e9d 8933 int inst = a->info.container_member;
949c47a0 8934 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8935 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8936 int failed = a->info.array.raid_disks;
8937 struct mdinfo *rv = NULL;
8938 struct mdinfo *d;
8939 struct mdinfo *di;
8940 struct metadata_update *mu;
8941 struct dl *dl;
8942 struct imsm_update_activate_spare *u;
8943 int num_spares = 0;
8944 int i;
95d07a2c 8945 int allowed;
88758e9d
DW		 8946
8947 for (d = a->info.devs ; d ; d = d->next) {
8948 if ((d->curr_state & DS_FAULTY) &&
8949 d->state_fd >= 0)
8950 /* wait for Removal to happen */
8951 return NULL;
8952 if (d->state_fd >= 0)
8953 failed--;
8954 }
8955
8956 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8957 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8958
e2962bfc
AK		 8959 if (imsm_reshape_blocks_arrays_changes(super))
8960 return NULL;
1af97990 8961
fc8ca064
AK		 8962 /* Cannot activate another spare if rebuild is in progress already
8963 */
8964 if (is_rebuilding(dev)) {
7a862a02 8965 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8966 return NULL;
8967 }
8968
89c67882
AK		 8969 if (a->info.array.level == 4)
8970 /* No repair for takeovered array
8971 * imsm doesn't support raid4
8972 */
8973 return NULL;
8974
3b451610
AK		 8975 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8976 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8977 return NULL;
8978
83ca7d45
AP		 8979 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8980 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8981 return NULL;
8982 }
8983
95d07a2c
LM		 8984 /*
8985 * If there are any failed disks check state of the other volume.
8986 * Block rebuild if the another one is failed until failed disks
8987 * are removed from container.
8988 */
8989 if (failed) {
7a862a02 8990 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8991 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8992 /* check if states of the other volumes allow for rebuild */
8993 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8994 if (i != inst) {
8995 allowed = imsm_rebuild_allowed(a->container,
8996 i, failed);
8997 if (!allowed)
8998 return NULL;
8999 }
9000 }
9001 }
9002
88758e9d 9003 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9004 for (i = 0; i < a->info.array.raid_disks; i++) {
9005 for (d = a->info.devs ; d ; d = d->next)
9006 if (d->disk.raid_disk == i)
9007 break;
9008 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9009 if (d && (d->state_fd >= 0))
9010 continue;
9011
272906ef 9012 /*
a20d2ba5
DW		 9013 * OK, this device needs recovery. Try to re-add the
9014 * previous occupant of this slot, if this fails see if
9015 * we can continue the assimilation of a spare that was
9016 * partially assimilated, finally try to activate a new
9017 * spare.
272906ef
DW		 9018 */
9019 dl = imsm_readd(super, i, a);
9020 if (!dl)
b303fe21 9021 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9022 if (!dl)
b303fe21 9023 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9024 if (!dl)
9025 continue;
1011e834 9026
272906ef 9027 /* found a usable disk with enough space */
503975b9 9028 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9029
9030 /* dl->index will be -1 in the case we are activating a
9031 * pristine spare. imsm_process_update() will create a
9032 * new index in this case. Once a disk is found to be
9033 * failed in all member arrays it is kicked from the
9034 * metadata
9035 */
9036 di->disk.number = dl->index;
d23fe947 9037
272906ef
DW		 9038 /* (ab)use di->devs to store a pointer to the device
9039 * we chose
9040 */
9041 di->devs = (struct mdinfo *) dl;
9042
9043 di->disk.raid_disk = i;
9044 di->disk.major = dl->major;
9045 di->disk.minor = dl->minor;
9046 di->disk.state = 0;
d23534e4 9047 di->recovery_start = 0;
5551b113 9048 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9049 di->component_size = a->info.component_size;
9050 di->container_member = inst;
5e46202e 9051 di->bb.supported = 1;
2c8890e9 9052 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9053 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9054 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9055 }
148acb7b 9056 super->random = random32();
272906ef
DW		 9057 di->next = rv;
9058 rv = di;
9059 num_spares++;
9060 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9061 i, di->data_offset);
88758e9d
DW		 9062 }
9063
9064 if (!rv)
9065 /* No spares found */
9066 return rv;
9067 /* Now 'rv' has a list of devices to return.
9068 * Create a metadata_update record to update the
9069 * disk_ord_tbl for the array
9070 */
503975b9 9071 mu = xmalloc(sizeof(*mu));
1011e834 9072 mu->buf = xcalloc(num_spares,
503975b9 9073 sizeof(struct imsm_update_activate_spare));
88758e9d 9074 mu->space = NULL;
cb23f1f4 9075 mu->space_list = NULL;
88758e9d
DW		 9076 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9077 mu->next = *updates;
9078 u = (struct imsm_update_activate_spare *) mu->buf;
9079
9080 for (di = rv ; di ; di = di->next) {
9081 u->type = update_activate_spare;
d23fe947
DW		 9082 u->dl = (struct dl *) di->devs;
9083 di->devs = NULL;
88758e9d
DW		 9084 u->slot = di->disk.raid_disk;
9085 u->array = inst;
9086 u->next = u + 1;
9087 u++;
9088 }
9089 (u-1)->next = NULL;
9090 *updates = mu;
9091
9092 return rv;
9093}
9094
54c2c1ea 9095static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9096{
54c2c1ea 9097 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9098 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9099 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9100 struct disk_info *inf = get_disk_info(u);
9101 struct imsm_disk *disk;
8273f55e
DW		 9102 int i;
9103 int j;
8273f55e 9104
54c2c1ea 9105 for (i = 0; i < map->num_members; i++) {
238c0a71 9106 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9107 for (j = 0; j < new_map->num_members; j++)
9108 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9109 return 1;
9110 }
9111
9112 return 0;
9113}
9114
1a64be56
LM		 9115static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9116{
594dc1b8
JS		 9117 struct dl *dl;
9118
1a64be56 9119 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9120 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9121 return dl;
9122 return NULL;
9123}
9124
9125static int remove_disk_super(struct intel_super *super, int major, int minor)
9126{
594dc1b8 9127 struct dl *prev;
1a64be56
LM		 9128 struct dl *dl;
9129
9130 prev = NULL;
9131 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9132 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9133 /* remove */
9134 if (prev)
9135 prev->next = dl->next;
9136 else
9137 super->disks = dl->next;
9138 dl->next = NULL;
9139 __free_imsm_disk(dl);
1ade5cc1 9140 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9141 break;
9142 }
9143 prev = dl;
9144 }
9145 return 0;
9146}
9147
f21e18ca 9148static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9149
1a64be56
LM		 9150static int add_remove_disk_update(struct intel_super *super)
9151{
9152 int check_degraded = 0;
594dc1b8
JS		 9153 struct dl *disk;
9154
1a64be56
LM		 9155 /* add/remove some spares to/from the metadata/contrainer */
9156 while (super->disk_mgmt_list) {
9157 struct dl *disk_cfg;
9158
9159 disk_cfg = super->disk_mgmt_list;
9160 super->disk_mgmt_list = disk_cfg->next;
9161 disk_cfg->next = NULL;
9162
9163 if (disk_cfg->action == DISK_ADD) {
9164 disk_cfg->next = super->disks;
9165 super->disks = disk_cfg;
9166 check_degraded = 1;
1ade5cc1
N		 9167 dprintf("added %x:%x\n",
9168 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9169 } else if (disk_cfg->action == DISK_REMOVE) {
9170 dprintf("Disk remove action processed: %x.%x\n",
9171 disk_cfg->major, disk_cfg->minor);
9172 disk = get_disk_super(super,
9173 disk_cfg->major,
9174 disk_cfg->minor);
9175 if (disk) {
9176 /* store action status */
9177 disk->action = DISK_REMOVE;
9178 /* remove spare disks only */
9179 if (disk->index == -1) {
9180 remove_disk_super(super,
9181 disk_cfg->major,
9182 disk_cfg->minor);
91c97c54
MT		 9183 } else {
9184 disk_cfg->fd = disk->fd;
9185 disk->fd = -1;
1a64be56
LM		 9186 }
9187 }
9188 /* release allocate disk structure */
9189 __free_imsm_disk(disk_cfg);
9190 }
9191 }
9192 return check_degraded;
9193}
9194
a29911da
PC		 9195static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9196 struct intel_super *super,
9197 void ***space_list)
9198{
9199 struct intel_dev *id;
9200 void **tofree = NULL;
9201 int ret_val = 0;
9202
1ade5cc1 9203 dprintf("(enter)\n");
089f9d79 9204 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9205 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9206 return ret_val;
9207 }
089f9d79 9208 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9209 dprintf("imsm: Error: Memory is not allocated\n");
9210 return ret_val;
9211 }
9212
9213 for (id = super->devlist ; id; id = id->next) {
9214 if (id->index == (unsigned)u->subdev) {
9215 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9216 struct imsm_map *map;
9217 struct imsm_dev *new_dev =
9218 (struct imsm_dev *)*space_list;
238c0a71 9219 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9220 int to_state;
9221 struct dl *new_disk;
9222
9223 if (new_dev == NULL)
9224 return ret_val;
9225 *space_list = **space_list;
9226 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9227 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9228 if (migr_map) {
9229 dprintf("imsm: Error: migration in progress");
9230 return ret_val;
9231 }
9232
9233 to_state = map->map_state;
9234 if ((u->new_level == 5) && (map->raid_level == 0)) {
9235 map->num_members++;
9236 /* this should not happen */
9237 if (u->new_disks[0] < 0) {
9238 map->failed_disk_num =
9239 map->num_members - 1;
9240 to_state = IMSM_T_STATE_DEGRADED;
9241 } else
9242 to_state = IMSM_T_STATE_NORMAL;
9243 }
8e59f3d8 9244 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9245 if (u->new_level > -1)
9246 map->raid_level = u->new_level;
238c0a71 9247 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9248 if ((u->new_level == 5) &&
9249 (migr_map->raid_level == 0)) {
9250 int ord = map->num_members - 1;
9251 migr_map->num_members--;
9252 if (u->new_disks[0] < 0)
9253 ord |= IMSM_ORD_REBUILD;
9254 set_imsm_ord_tbl_ent(map,
9255 map->num_members - 1,
9256 ord);
9257 }
9258 id->dev = new_dev;
9259 tofree = (void **)dev;
9260
4bba0439
PC		 9261 /* update chunk size
9262 */
06fb291a
PB		 9263 if (u->new_chunksize > 0) {
9264 unsigned long long num_data_stripes;
9529d343
MD		 9265 struct imsm_map *dest_map =
9266 get_imsm_map(dev, MAP_0);
06fb291a 9267 int used_disks =
9529d343 9268 imsm_num_data_members(dest_map);
06fb291a
PB		 9269
9270 if (used_disks == 0)
9271 return ret_val;
9272
4bba0439
PC		 9273 map->blocks_per_strip =
9274 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9275 num_data_stripes =
fcc2c9da 9276 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9277 num_data_stripes /= map->blocks_per_strip;
9278 num_data_stripes /= map->num_domains;
9279 set_num_data_stripes(map, num_data_stripes);
9280 }
4bba0439 9281
44490938
MD		 9282 /* ensure blocks_per_member has valid value
9283 */
9284 set_blocks_per_member(map,
9285 per_dev_array_size(map) +
9286 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9287
a29911da
PC		 9288 /* add disk
9289 */
089f9d79
JS		 9290 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9291 migr_map->raid_level == map->raid_level)
a29911da
PC		 9292 goto skip_disk_add;
9293
9294 if (u->new_disks[0] >= 0) {
9295 /* use passes spare
9296 */
9297 new_disk = get_disk_super(super,
9298 major(u->new_disks[0]),
9299 minor(u->new_disks[0]));
7a862a02 9300 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9301 major(u->new_disks[0]),
9302 minor(u->new_disks[0]),
9303 new_disk, new_disk->index);
9304 if (new_disk == NULL)
9305 goto error_disk_add;
9306
9307 new_disk->index = map->num_members - 1;
9308 /* slot to fill in autolayout
9309 */
9310 new_disk->raiddisk = new_disk->index;
9311 new_disk->disk.status |= CONFIGURED_DISK;
9312 new_disk->disk.status &= ~SPARE_DISK;
9313 } else
9314 goto error_disk_add;
9315
9316skip_disk_add:
9317 *tofree = *space_list;
9318 /* calculate new size
9319 */
f3871fdc 9320 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9321
9322 ret_val = 1;
9323 }
9324 }
9325
9326 if (tofree)
9327 *space_list = tofree;
9328 return ret_val;
9329
9330error_disk_add:
9331 dprintf("Error: imsm: Cannot find disk.\n");
9332 return ret_val;
9333}
9334
f3871fdc
AK		 9335static int apply_size_change_update(struct imsm_update_size_change *u,
9336 struct intel_super *super)
9337{
9338 struct intel_dev *id;
9339 int ret_val = 0;
9340
1ade5cc1 9341 dprintf("(enter)\n");
089f9d79 9342 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9343 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9344 return ret_val;
9345 }
9346
9347 for (id = super->devlist ; id; id = id->next) {
9348 if (id->index == (unsigned)u->subdev) {
9349 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9350 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9351 int used_disks = imsm_num_data_members(map);
f3871fdc 9352 unsigned long long blocks_per_member;
06fb291a 9353 unsigned long long num_data_stripes;
44490938
MD		 9354 unsigned long long new_size_per_disk;
9355
9356 if (used_disks == 0)
9357 return 0;
f3871fdc
AK		 9358
9359 /* calculate new size
9360 */
44490938
MD		 9361 new_size_per_disk = u->new_size / used_disks;
9362 blocks_per_member = new_size_per_disk +
9363 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9364 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9365 map->blocks_per_strip;
9366 num_data_stripes /= map->num_domains;
9367 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9368 u->new_size, new_size_per_disk,
06fb291a 9369 num_data_stripes);
f3871fdc 9370 set_blocks_per_member(map, blocks_per_member);
06fb291a 9371 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9372 imsm_set_array_size(dev, u->new_size);
9373
9374 ret_val = 1;
9375 break;
9376 }
9377 }
9378
9379 return ret_val;
9380}
9381
061d7da3 9382static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9383 struct intel_super *super,
061d7da3
LO		 9384 struct active_array *active_array)
9385{
9386 struct imsm_super *mpb = super->anchor;
9387 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9388 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9389 struct imsm_map *migr_map;
9390 struct active_array *a;
9391 struct imsm_disk *disk;
9392 __u8 to_state;
9393 struct dl *dl;
9394 unsigned int found;
9395 int failed;
5961eeec 9396 int victim;
061d7da3 9397 int i;
5961eeec 9398 int second_map_created = 0;
061d7da3 9399
5961eeec 9400 for (; u; u = u->next) {
238c0a71 9401 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9402
5961eeec 9403 if (victim < 0)
9404 return 0;
061d7da3 9405
5961eeec 9406 for (dl = super->disks; dl; dl = dl->next)
9407 if (dl == u->dl)
9408 break;
061d7da3 9409
5961eeec 9410 if (!dl) {
7a862a02 9411 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9412 u->dl->index);
9413 return 0;
9414 }
061d7da3 9415
5961eeec 9416 /* count failures (excluding rebuilds and the victim)
9417 * to determine map[0] state
9418 */
9419 failed = 0;
9420 for (i = 0; i < map->num_members; i++) {
9421 if (i == u->slot)
9422 continue;
9423 disk = get_imsm_disk(super,
238c0a71 9424 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9425 if (!disk || is_failed(disk))
9426 failed++;
9427 }
061d7da3 9428
5961eeec 9429 /* adding a pristine spare, assign a new index */
9430 if (dl->index < 0) {
9431 dl->index = super->anchor->num_disks;
9432 super->anchor->num_disks++;
9433 }
9434 disk = &dl->disk;
9435 disk->status |= CONFIGURED_DISK;
9436 disk->status &= ~SPARE_DISK;
9437
9438 /* mark rebuild */
238c0a71 9439 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9440 if (!second_map_created) {
9441 second_map_created = 1;
9442 map->map_state = IMSM_T_STATE_DEGRADED;
9443 migrate(dev, super, to_state, MIGR_REBUILD);
9444 } else
9445 map->map_state = to_state;
238c0a71 9446 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9447 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9448 set_imsm_ord_tbl_ent(migr_map, u->slot,
9449 dl->index | IMSM_ORD_REBUILD);
9450
9451 /* update the family_num to mark a new container
9452 * generation, being careful to record the existing
9453 * family_num in orig_family_num to clean up after
9454 * earlier mdadm versions that neglected to set it.
9455 */
9456 if (mpb->orig_family_num == 0)
9457 mpb->orig_family_num = mpb->family_num;
9458 mpb->family_num += super->random;
9459
9460 /* count arrays using the victim in the metadata */
9461 found = 0;
9462 for (a = active_array; a ; a = a->next) {
9463 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9464 map = get_imsm_map(dev, MAP_0);
061d7da3 9465
5961eeec 9466 if (get_imsm_disk_slot(map, victim) >= 0)
9467 found++;
9468 }
061d7da3 9469
5961eeec 9470 /* delete the victim if it is no longer being
9471 * utilized anywhere
061d7da3 9472 */
5961eeec 9473 if (!found) {
9474 struct dl **dlp;
061d7da3 9475
5961eeec 9476 /* We know that 'manager' isn't touching anything,
9477 * so it is safe to delete
9478 */
9479 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9480 if ((*dlp)->index == victim)
9481 break;
5961eeec 9482
9483 /* victim may be on the missing list */
9484 if (!*dlp)
9485 for (dlp = &super->missing; *dlp;
9486 dlp = &(*dlp)->next)
9487 if ((*dlp)->index == victim)
9488 break;
9489 imsm_delete(super, dlp, victim);
9490 }
061d7da3
LO		 9491 }
9492
9493 return 1;
9494}
a29911da 9495
2e5dc010
N		 9496static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9497 struct intel_super *super,
9498 void ***space_list)
9499{
9500 struct dl *new_disk;
9501 struct intel_dev *id;
9502 int i;
9503 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9504 int disk_count = u->old_raid_disks;
2e5dc010
N		 9505 void **tofree = NULL;
9506 int devices_to_reshape = 1;
9507 struct imsm_super *mpb = super->anchor;
9508 int ret_val = 0;
d098291a 9509 unsigned int dev_id;
2e5dc010 9510
1ade5cc1 9511 dprintf("(enter)\n");
2e5dc010
N		 9512
9513 /* enable spares to use in array */
9514 for (i = 0; i < delta_disks; i++) {
9515 new_disk = get_disk_super(super,
9516 major(u->new_disks[i]),
9517 minor(u->new_disks[i]));
7a862a02 9518 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9519 major(u->new_disks[i]), minor(u->new_disks[i]),
9520 new_disk, new_disk->index);
089f9d79
JS		 9521 if (new_disk == NULL ||
9522 (new_disk->index >= 0 &&
9523 new_disk->index < u->old_raid_disks))
2e5dc010 9524 goto update_reshape_exit;
ee4beede 9525 new_disk->index = disk_count++;
2e5dc010
N		 9526 /* slot to fill in autolayout
9527 */
9528 new_disk->raiddisk = new_disk->index;
9529 new_disk->disk.status |=
9530 CONFIGURED_DISK;
9531 new_disk->disk.status &= ~SPARE_DISK;
9532 }
9533
ed7333bd
AK		 9534 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9535 mpb->num_raid_devs);
2e5dc010
N		 9536 /* manage changes in volume
9537 */
d098291a 9538 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9539 void **sp = *space_list;
9540 struct imsm_dev *newdev;
9541 struct imsm_map *newmap, *oldmap;
9542
d098291a
AK		 9543 for (id = super->devlist ; id; id = id->next) {
9544 if (id->index == dev_id)
9545 break;
9546 }
9547 if (id == NULL)
9548 break;
2e5dc010
N		 9549 if (!sp)
9550 continue;
9551 *space_list = *sp;
9552 newdev = (void*)sp;
9553 /* Copy the dev, but not (all of) the map */
9554 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9555 oldmap = get_imsm_map(id->dev, MAP_0);
9556 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9557 /* Copy the current map */
9558 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9559 /* update one device only
9560 */
9561 if (devices_to_reshape) {
ed7333bd
AK		 9562 dprintf("imsm: modifying subdev: %i\n",
9563 id->index);
2e5dc010
N		 9564 devices_to_reshape--;
9565 newdev->vol.migr_state = 1;
9566 newdev->vol.curr_migr_unit = 0;
ea672ee1 9567 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9568 newmap->num_members = u->new_raid_disks;
9569 for (i = 0; i < delta_disks; i++) {
9570 set_imsm_ord_tbl_ent(newmap,
9571 u->old_raid_disks + i,
9572 u->old_raid_disks + i);
9573 }
9574 /* New map is correct, now need to save old map
9575 */
238c0a71 9576 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9577 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9578
f3871fdc 9579 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9580 }
9581
9582 sp = (void **)id->dev;
9583 id->dev = newdev;
9584 *sp = tofree;
9585 tofree = sp;
8e59f3d8
AK		 9586
9587 /* Clear migration record */
9588 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9589 }
819bc634
AK		 9590 if (tofree)
9591 *space_list = tofree;
2e5dc010
N		 9592 ret_val = 1;
9593
9594update_reshape_exit:
9595
9596 return ret_val;
9597}
9598
bb025c2f 9599static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9600 struct intel_super *super,
9601 void ***space_list)
bb025c2f
KW		 9602{
9603 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9604 struct intel_dev *dv;
9605 struct imsm_dev *dev_new;
bb025c2f
KW		 9606 struct imsm_map *map;
9607 struct dl *dm, *du;
8ca6df95 9608 int i;
bb025c2f
KW		 9609
9610 for (dv = super->devlist; dv; dv = dv->next)
9611 if (dv->index == (unsigned int)u->subarray) {
9612 dev = dv->dev;
9613 break;
9614 }
9615
9616 if (dev == NULL)
9617 return 0;
9618
238c0a71 9619 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9620
9621 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9622 unsigned long long num_data_stripes;
9623
43d5ec18 9624 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9625 if (imsm_count_failed(super, dev, MAP_0) !=
9626 (map->num_members / 2))
43d5ec18
KW		 9627 return 0;
9628
bb025c2f
KW		 9629 /* iterate through devices to mark removed disks as spare */
9630 for (dm = super->disks; dm; dm = dm->next) {
9631 if (dm->disk.status & FAILED_DISK) {
9632 int idx = dm->index;
9633 /* update indexes on the disk list */
9634/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9635 the index values will end up being correct.... NB */
9636 for (du = super->disks; du; du = du->next)
9637 if (du->index > idx)
9638 du->index--;
9639 /* mark as spare disk */
a8619d23 9640 mark_spare(dm);
bb025c2f
KW		 9641 }
9642 }
bb025c2f
KW		 9643 /* update map */
9644 map->num_members = map->num_members / 2;
9645 map->map_state = IMSM_T_STATE_NORMAL;
9646 map->num_domains = 1;
9647 map->raid_level = 0;
9648 map->failed_disk_num = -1;
4a353e6e
RS		 9649 num_data_stripes = imsm_dev_size(dev) / 2;
9650 num_data_stripes /= map->blocks_per_strip;
9651 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9652 }
9653
8ca6df95
KW		 9654 if (u->direction == R0_TO_R10) {
9655 void **space;
4a353e6e
RS		 9656 unsigned long long num_data_stripes;
9657
8ca6df95
KW		 9658 /* update slots in current disk list */
9659 for (dm = super->disks; dm; dm = dm->next) {
9660 if (dm->index >= 0)
9661 dm->index *= 2;
9662 }
9663 /* create new *missing* disks */
9664 for (i = 0; i < map->num_members; i++) {
9665 space = *space_list;
9666 if (!space)
9667 continue;
9668 *space_list = *space;
9669 du = (void *)space;
9670 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9671 du->fd = -1;
9672 du->minor = 0;
9673 du->major = 0;
9674 du->index = (i * 2) + 1;
9675 sprintf((char *)du->disk.serial,
9676 " MISSING_%d", du->index);
9677 sprintf((char *)du->serial,
9678 "MISSING_%d", du->index);
9679 du->next = super->missing;
9680 super->missing = du;
9681 }
9682 /* create new dev and map */
9683 space = *space_list;
9684 if (!space)
9685 return 0;
9686 *space_list = *space;
9687 dev_new = (void *)space;
9688 memcpy(dev_new, dev, sizeof(*dev));
9689 /* update new map */
238c0a71 9690 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9691 map->num_members = map->num_members * 2;
1a2487c2 9692 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9693 map->num_domains = 2;
9694 map->raid_level = 1;
4a353e6e
RS		 9695 num_data_stripes = imsm_dev_size(dev) / 2;
9696 num_data_stripes /= map->blocks_per_strip;
9697 num_data_stripes /= map->num_domains;
9698 set_num_data_stripes(map, num_data_stripes);
9699
8ca6df95
KW		 9700 /* replace dev<->dev_new */
9701 dv->dev = dev_new;
9702 }
bb025c2f
KW		 9703 /* update disk order table */
9704 for (du = super->disks; du; du = du->next)
9705 if (du->index >= 0)
9706 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9707 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9708 if (du->index >= 0) {
9709 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9710 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9711 }
bb025c2f
KW		 9712
9713 return 1;
9714}
9715
e8319a19
DW		 9716static void imsm_process_update(struct supertype *st,
9717 struct metadata_update *update)
9718{
9719 /**
9720 * crack open the metadata_update envelope to find the update record
9721 * update can be one of:
d195167d
AK		 9722 * update_reshape_container_disks - all the arrays in the container
9723 * are being reshaped to have more devices. We need to mark
9724 * the arrays for general migration and convert selected spares
9725 * into active devices.
9726 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9727 * device in an array, update the disk_ord_tbl. If this disk is
9728 * present in all member arrays then also clear the SPARE_DISK
9729 * flag
d195167d
AK		 9730 * update_create_array
9731 * update_kill_array
9732 * update_rename_array
9733 * update_add_remove_disk
e8319a19
DW		 9734 */
9735 struct intel_super *super = st->sb;
4d7b1503 9736 struct imsm_super *mpb;
e8319a19
DW		 9737 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9738
4d7b1503
DW		 9739 /* update requires a larger buf but the allocation failed */
9740 if (super->next_len && !super->next_buf) {
9741 super->next_len = 0;
9742 return;
9743 }
9744
9745 if (super->next_buf) {
9746 memcpy(super->next_buf, super->buf, super->len);
9747 free(super->buf);
9748 super->len = super->next_len;
9749 super->buf = super->next_buf;
9750
9751 super->next_len = 0;
9752 super->next_buf = NULL;
9753 }
9754
9755 mpb = super->anchor;
9756
e8319a19 9757 switch (type) {
0ec5d470
AK		 9758 case update_general_migration_checkpoint: {
9759 struct intel_dev *id;
9760 struct imsm_update_general_migration_checkpoint *u =
9761 (void *)update->buf;
9762
1ade5cc1 9763 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9764
9765 /* find device under general migration */
9766 for (id = super->devlist ; id; id = id->next) {
9767 if (is_gen_migration(id->dev)) {
9768 id->dev->vol.curr_migr_unit =
9769 __cpu_to_le32(u->curr_migr_unit);
9770 super->updates_pending++;
9771 }
9772 }
9773 break;
9774 }
bb025c2f
KW		 9775 case update_takeover: {
9776 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9777 if (apply_takeover_update(u, super, &update->space_list)) {
9778 imsm_update_version_info(super);
bb025c2f 9779 super->updates_pending++;
1a2487c2 9780 }
bb025c2f
KW		 9781 break;
9782 }
9783
78b10e66 9784 case update_reshape_container_disks: {
d195167d 9785 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9786 if (apply_reshape_container_disks_update(
9787 u, super, &update->space_list))
9788 super->updates_pending++;
78b10e66
N		 9789 break;
9790 }
48c5303a 9791 case update_reshape_migration: {
a29911da
PC		 9792 struct imsm_update_reshape_migration *u = (void *)update->buf;
9793 if (apply_reshape_migration_update(
9794 u, super, &update->space_list))
9795 super->updates_pending++;
48c5303a
PC		 9796 break;
9797 }
f3871fdc
AK		 9798 case update_size_change: {
9799 struct imsm_update_size_change *u = (void *)update->buf;
9800 if (apply_size_change_update(u, super))
9801 super->updates_pending++;
9802 break;
9803 }
e8319a19 9804 case update_activate_spare: {
1011e834 9805 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9806 if (apply_update_activate_spare(u, super, st->arrays))
9807 super->updates_pending++;
8273f55e
DW		 9808 break;
9809 }
9810 case update_create_array: {
9811 /* someone wants to create a new array, we need to be aware of
9812 * a few races/collisions:
9813 * 1/ 'Create' called by two separate instances of mdadm
9814 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9815 * devices that have since been assimilated via
9816 * activate_spare.
9817 * In the event this update can not be carried out mdadm will
9818 * (FIX ME) notice that its update did not take hold.
9819 */
9820 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9821 struct intel_dev *dv;
8273f55e
DW		 9822 struct imsm_dev *dev;
9823 struct imsm_map *map, *new_map;
9824 unsigned long long start, end;
9825 unsigned long long new_start, new_end;
9826 int i;
54c2c1ea
DW		 9827 struct disk_info *inf;
9828 struct dl *dl;
8273f55e
DW		 9829
9830 /* handle racing creates: first come first serve */
9831 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9832 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9833 goto create_error;
8273f55e
DW		 9834 }
9835
9836 /* check update is next in sequence */
9837 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9838 dprintf("can not create array %d expected index %d\n",
9839 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9840 goto create_error;
8273f55e
DW		 9841 }
9842
238c0a71 9843 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9844 new_start = pba_of_lba0(new_map);
44490938 9845 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9846 inf = get_disk_info(u);
8273f55e
DW		 9847
9848 /* handle activate_spare versus create race:
9849 * check to make sure that overlapping arrays do not include
9850 * overalpping disks
9851 */
9852 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9853 dev = get_imsm_dev(super, i);
238c0a71 9854 map = get_imsm_map(dev, MAP_0);
5551b113 9855 start = pba_of_lba0(map);
44490938 9856 end = start + per_dev_array_size(map);
8273f55e
DW		 9857 if ((new_start >= start && new_start <= end) ||
9858 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9859 /* overlap */;
9860 else
9861 continue;
9862
9863 if (disks_overlap(super, i, u)) {
1ade5cc1 9864 dprintf("arrays overlap\n");
ba2de7ba 9865 goto create_error;
8273f55e
DW		 9866 }
9867 }
8273f55e 9868
949c47a0
DW		 9869 /* check that prepare update was successful */
9870 if (!update->space) {
1ade5cc1 9871 dprintf("prepare update failed\n");
ba2de7ba 9872 goto create_error;
949c47a0
DW		 9873 }
9874
54c2c1ea
DW		 9875 /* check that all disks are still active before committing
9876 * changes. FIXME: could we instead handle this by creating a
9877 * degraded array? That's probably not what the user expects,
9878 * so better to drop this update on the floor.
9879 */
9880 for (i = 0; i < new_map->num_members; i++) {
9881 dl = serial_to_dl(inf[i].serial, super);
9882 if (!dl) {
1ade5cc1 9883 dprintf("disk disappeared\n");
ba2de7ba 9884 goto create_error;
54c2c1ea 9885 }
949c47a0
DW		 9886 }
9887
8273f55e 9888 super->updates_pending++;
54c2c1ea
DW		 9889
9890 /* convert spares to members and fixup ord_tbl */
9891 for (i = 0; i < new_map->num_members; i++) {
9892 dl = serial_to_dl(inf[i].serial, super);
9893 if (dl->index == -1) {
9894 dl->index = mpb->num_disks;
9895 mpb->num_disks++;
9896 dl->disk.status |= CONFIGURED_DISK;
9897 dl->disk.status &= ~SPARE_DISK;
9898 }
9899 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9900 }
9901
ba2de7ba
DW		 9902 dv = update->space;
9903 dev = dv->dev;
949c47a0
DW		 9904 update->space = NULL;
9905 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9906 dv->index = u->dev_idx;
9907 dv->next = super->devlist;
9908 super->devlist = dv;
8273f55e 9909 mpb->num_raid_devs++;
8273f55e 9910
4d1313e9 9911 imsm_update_version_info(super);
8273f55e 9912 break;
ba2de7ba
DW		 9913 create_error:
9914 /* mdmon knows how to release update->space, but not
9915 * ((struct intel_dev *) update->space)->dev
9916 */
9917 if (update->space) {
9918 dv = update->space;
9919 free(dv->dev);
9920 }
8273f55e 9921 break;
e8319a19 9922 }
33414a01
DW		 9923 case update_kill_array: {
9924 struct imsm_update_kill_array *u = (void *) update->buf;
9925 int victim = u->dev_idx;
9926 struct active_array *a;
9927 struct intel_dev **dp;
9928 struct imsm_dev *dev;
9929
9930 /* sanity check that we are not affecting the uuid of
9931 * active arrays, or deleting an active array
9932 *
9933 * FIXME when immutable ids are available, but note that
9934 * we'll also need to fixup the invalidated/active
9935 * subarray indexes in mdstat
9936 */
9937 for (a = st->arrays; a; a = a->next)
9938 if (a->info.container_member >= victim)
9939 break;
9940 /* by definition if mdmon is running at least one array
9941 * is active in the container, so checking
9942 * mpb->num_raid_devs is just extra paranoia
9943 */
9944 dev = get_imsm_dev(super, victim);
9945 if (a || !dev || mpb->num_raid_devs == 1) {
9946 dprintf("failed to delete subarray-%d\n", victim);
9947 break;
9948 }
9949
9950 for (dp = &super->devlist; *dp;)
f21e18ca 9951 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9952 *dp = (*dp)->next;
9953 } else {
f21e18ca 9954 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9955 (*dp)->index--;
9956 dp = &(*dp)->next;
9957 }
9958 mpb->num_raid_devs--;
9959 super->updates_pending++;
9960 break;
9961 }
aa534678
DW		 9962 case update_rename_array: {
9963 struct imsm_update_rename_array *u = (void *) update->buf;
9964 char name[MAX_RAID_SERIAL_LEN+1];
9965 int target = u->dev_idx;
9966 struct active_array *a;
9967 struct imsm_dev *dev;
9968
9969 /* sanity check that we are not affecting the uuid of
9970 * an active array
9971 */
40659392 9972 memset(name, 0, sizeof(name));
aa534678
DW		 9973 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9974 name[MAX_RAID_SERIAL_LEN] = '\0';
9975 for (a = st->arrays; a; a = a->next)
9976 if (a->info.container_member == target)
9977 break;
9978 dev = get_imsm_dev(super, u->dev_idx);
9979 if (a || !dev || !check_name(super, name, 1)) {
9980 dprintf("failed to rename subarray-%d\n", target);
9981 break;
9982 }
9983
40659392 9984 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 9985 super->updates_pending++;
9986 break;
9987 }
1a64be56 9988 case update_add_remove_disk: {
43dad3d6 9989 /* we may be able to repair some arrays if disks are
095b8088 9990 * being added, check the status of add_remove_disk
1a64be56
LM		 9991 * if discs has been added.
9992 */
9993 if (add_remove_disk_update(super)) {
43dad3d6 9994 struct active_array *a;
072b727f
DW		 9995
9996 super->updates_pending++;
1a64be56 9997 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9998 a->check_degraded = 1;
9999 }
43dad3d6 10000 break;
e8319a19 10001 }
bbab0940
TM		 10002 case update_prealloc_badblocks_mem:
10003 break;
e6e9dd3f
AP		 10004 case update_rwh_policy: {
10005 struct imsm_update_rwh_policy *u = (void *)update->buf;
10006 int target = u->dev_idx;
10007 struct imsm_dev *dev = get_imsm_dev(super, target);
10008 if (!dev) {
10009 dprintf("could not find subarray-%d\n", target);
10010 break;
10011 }
10012
10013 if (dev->rwh_policy != u->new_policy) {
10014 dev->rwh_policy = u->new_policy;
10015 super->updates_pending++;
10016 }
10017 break;
10018 }
1a64be56 10019 default:
ebf3be99 10020 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10021 }
e8319a19 10022}
88758e9d 10023
bc0b9d34
PC		 10024static struct mdinfo *get_spares_for_grow(struct supertype *st);
10025
5fe6f031
N		 10026static int imsm_prepare_update(struct supertype *st,
10027 struct metadata_update *update)
8273f55e 10028{
949c47a0 10029 /**
4d7b1503
DW		 10030 * Allocate space to hold new disk entries, raid-device entries or a new
10031 * mpb if necessary. The manager synchronously waits for updates to
10032 * complete in the monitor, so new mpb buffers allocated here can be
10033 * integrated by the monitor thread without worrying about live pointers
10034 * in the manager thread.
8273f55e 10035 */
095b8088 10036 enum imsm_update_type type;
4d7b1503 10037 struct intel_super *super = st->sb;
f36a9ecd 10038 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10039 struct imsm_super *mpb = super->anchor;
10040 size_t buf_len;
10041 size_t len = 0;
949c47a0 10042
095b8088
N		 10043 if (update->len < (int)sizeof(type))
10044 return 0;
10045
10046 type = *(enum imsm_update_type *) update->buf;
10047
949c47a0 10048 switch (type) {
0ec5d470 10049 case update_general_migration_checkpoint:
095b8088
N		 10050 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10051 return 0;
1ade5cc1 10052 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10053 break;
abedf5fc
KW		 10054 case update_takeover: {
10055 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10056 if (update->len < (int)sizeof(*u))
10057 return 0;
abedf5fc
KW		 10058 if (u->direction == R0_TO_R10) {
10059 void **tail = (void **)&update->space_list;
10060 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10061 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10062 int num_members = map->num_members;
10063 void *space;
10064 int size, i;
abedf5fc
KW		 10065 /* allocate memory for added disks */
10066 for (i = 0; i < num_members; i++) {
10067 size = sizeof(struct dl);
503975b9 10068 space = xmalloc(size);
abedf5fc
KW		 10069 *tail = space;
10070 tail = space;
10071 *tail = NULL;
10072 }
10073 /* allocate memory for new device */
10074 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10075 (num_members * sizeof(__u32));
503975b9
N		 10076 space = xmalloc(size);
10077 *tail = space;
10078 tail = space;
10079 *tail = NULL;
10080 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10081 }
10082
10083 break;
10084 }
78b10e66 10085 case update_reshape_container_disks: {
d195167d
AK		 10086 /* Every raid device in the container is about to
10087 * gain some more devices, and we will enter a
10088 * reconfiguration.
10089 * So each 'imsm_map' will be bigger, and the imsm_vol
10090 * will now hold 2 of them.
10091 * Thus we need new 'struct imsm_dev' allocations sized
10092 * as sizeof_imsm_dev but with more devices in both maps.
10093 */
10094 struct imsm_update_reshape *u = (void *)update->buf;
10095 struct intel_dev *dl;
10096 void **space_tail = (void**)&update->space_list;
10097
095b8088
N		 10098 if (update->len < (int)sizeof(*u))
10099 return 0;
10100
1ade5cc1 10101 dprintf("for update_reshape\n");
d195167d
AK		 10102
10103 for (dl = super->devlist; dl; dl = dl->next) {
10104 int size = sizeof_imsm_dev(dl->dev, 1);
10105 void *s;
d677e0b8
AK		 10106 if (u->new_raid_disks > u->old_raid_disks)
10107 size += sizeof(__u32)*2*
10108 (u->new_raid_disks - u->old_raid_disks);
503975b9 10109 s = xmalloc(size);
d195167d
AK		 10110 *space_tail = s;
10111 space_tail = s;
10112 *space_tail = NULL;
10113 }
10114
10115 len = disks_to_mpb_size(u->new_raid_disks);
10116 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10117 break;
10118 }
48c5303a 10119 case update_reshape_migration: {
bc0b9d34
PC		 10120 /* for migration level 0->5 we need to add disks
10121 * so the same as for container operation we will copy
10122 * device to the bigger location.
10123 * in memory prepared device and new disk area are prepared
10124 * for usage in process update
10125 */
10126 struct imsm_update_reshape_migration *u = (void *)update->buf;
10127 struct intel_dev *id;
10128 void **space_tail = (void **)&update->space_list;
10129 int size;
10130 void *s;
10131 int current_level = -1;
10132
095b8088
N		 10133 if (update->len < (int)sizeof(*u))
10134 return 0;
10135
1ade5cc1 10136 dprintf("for update_reshape\n");
bc0b9d34
PC		 10137
10138 /* add space for bigger array in update
10139 */
10140 for (id = super->devlist; id; id = id->next) {
10141 if (id->index == (unsigned)u->subdev) {
10142 size = sizeof_imsm_dev(id->dev, 1);
10143 if (u->new_raid_disks > u->old_raid_disks)
10144 size += sizeof(__u32)*2*
10145 (u->new_raid_disks - u->old_raid_disks);
503975b9 10146 s = xmalloc(size);
bc0b9d34
PC		 10147 *space_tail = s;
10148 space_tail = s;
10149 *space_tail = NULL;
10150 break;
10151 }
10152 }
10153 if (update->space_list == NULL)
10154 break;
10155
10156 /* add space for disk in update
10157 */
10158 size = sizeof(struct dl);
503975b9 10159 s = xmalloc(size);
bc0b9d34
PC		 10160 *space_tail = s;
10161 space_tail = s;
10162 *space_tail = NULL;
10163
10164 /* add spare device to update
10165 */
10166 for (id = super->devlist ; id; id = id->next)
10167 if (id->index == (unsigned)u->subdev) {
10168 struct imsm_dev *dev;
10169 struct imsm_map *map;
10170
10171 dev = get_imsm_dev(super, u->subdev);
238c0a71 10172 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10173 current_level = map->raid_level;
10174 break;
10175 }
089f9d79 10176 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10177 struct mdinfo *spares;
10178
10179 spares = get_spares_for_grow(st);
10180 if (spares) {
10181 struct dl *dl;
10182 struct mdinfo *dev;
10183
10184 dev = spares->devs;
10185 if (dev) {
10186 u->new_disks[0] =
10187 makedev(dev->disk.major,
10188 dev->disk.minor);
10189 dl = get_disk_super(super,
10190 dev->disk.major,
10191 dev->disk.minor);
10192 dl->index = u->old_raid_disks;
10193 dev = dev->next;
10194 }
10195 sysfs_free(spares);
10196 }
10197 }
10198 len = disks_to_mpb_size(u->new_raid_disks);
10199 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10200 break;
10201 }
f3871fdc 10202 case update_size_change: {
095b8088
N		 10203 if (update->len < (int)sizeof(struct imsm_update_size_change))
10204 return 0;
10205 break;
10206 }
10207 case update_activate_spare: {
10208 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10209 return 0;
f3871fdc
AK		 10210 break;
10211 }
949c47a0
DW		 10212 case update_create_array: {
10213 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10214 struct intel_dev *dv;
54c2c1ea 10215 struct imsm_dev *dev = &u->dev;
238c0a71 10216 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10217 struct dl *dl;
10218 struct disk_info *inf;
10219 int i;
10220 int activate = 0;
949c47a0 10221
095b8088
N		 10222 if (update->len < (int)sizeof(*u))
10223 return 0;
10224
54c2c1ea
DW		 10225 inf = get_disk_info(u);
10226 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10227 /* allocate a new super->devlist entry */
503975b9
N		 10228 dv = xmalloc(sizeof(*dv));
10229 dv->dev = xmalloc(len);
10230 update->space = dv;
949c47a0 10231
54c2c1ea
DW		 10232 /* count how many spares will be converted to members */
10233 for (i = 0; i < map->num_members; i++) {
10234 dl = serial_to_dl(inf[i].serial, super);
10235 if (!dl) {
10236 /* hmm maybe it failed?, nothing we can do about
10237 * it here
10238 */
10239 continue;
10240 }
10241 if (count_memberships(dl, super) == 0)
10242 activate++;
10243 }
10244 len += activate * sizeof(struct imsm_disk);
949c47a0 10245 break;
095b8088
N		 10246 }
10247 case update_kill_array: {
10248 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10249 return 0;
949c47a0
DW		 10250 break;
10251 }
095b8088
N		 10252 case update_rename_array: {
10253 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10254 return 0;
10255 break;
10256 }
10257 case update_add_remove_disk:
10258 /* no update->len needed */
10259 break;
bbab0940
TM		 10260 case update_prealloc_badblocks_mem:
10261 super->extra_space += sizeof(struct bbm_log) -
10262 get_imsm_bbm_log_size(super->bbm_log);
10263 break;
e6e9dd3f
AP		 10264 case update_rwh_policy: {
10265 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10266 return 0;
10267 break;
10268 }
095b8088
N		 10269 default:
10270 return 0;
949c47a0 10271 }
8273f55e 10272
4d7b1503
DW		 10273 /* check if we need a larger metadata buffer */
10274 if (super->next_buf)
10275 buf_len = super->next_len;
10276 else
10277 buf_len = super->len;
10278
bbab0940 10279 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10280 /* ok we need a larger buf than what is currently allocated
10281 * if this allocation fails process_update will notice that
10282 * ->next_len is set and ->next_buf is NULL
10283 */
bbab0940
TM		 10284 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10285 super->extra_space + len, sector_size);
4d7b1503
DW		 10286 if (super->next_buf)
10287 free(super->next_buf);
10288
10289 super->next_len = buf_len;
f36a9ecd 10290 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10291 memset(super->next_buf, 0, buf_len);
10292 else
4d7b1503
DW		 10293 super->next_buf = NULL;
10294 }
5fe6f031 10295 return 1;
8273f55e
DW		 10296}
10297
ae6aad82 10298/* must be called while manager is quiesced */
f21e18ca 10299static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10300{
10301 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10302 struct dl *iter;
10303 struct imsm_dev *dev;
10304 struct imsm_map *map;
4c9e8c1e 10305 unsigned int i, j, num_members;
fb12a745 10306 __u32 ord, ord_map0;
4c9e8c1e 10307 struct bbm_log *log = super->bbm_log;
ae6aad82 10308
1ade5cc1 10309 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10310
10311 /* shift all indexes down one */
10312 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10313 if (iter->index > (int)index)
ae6aad82 10314 iter->index--;
47ee5a45 10315 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10316 if (iter->index > (int)index)
47ee5a45 10317 iter->index--;
ae6aad82
DW		 10318
10319 for (i = 0; i < mpb->num_raid_devs; i++) {
10320 dev = get_imsm_dev(super, i);
238c0a71 10321 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10322 num_members = map->num_members;
10323 for (j = 0; j < num_members; j++) {
10324 /* update ord entries being careful not to propagate
10325 * ord-flags to the first map
10326 */
238c0a71 10327 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10328 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10329
24565c9a
DW		 10330 if (ord_to_idx(ord) <= index)
10331 continue;
ae6aad82 10332
238c0a71 10333 map = get_imsm_map(dev, MAP_0);
fb12a745 10334 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10335 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10336 if (map)
10337 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10338 }
10339 }
10340
4c9e8c1e
TM		 10341 for (i = 0; i < log->entry_count; i++) {
10342 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10343
10344 if (entry->disk_ordinal <= index)
10345 continue;
10346 entry->disk_ordinal--;
10347 }
10348
ae6aad82
DW		 10349 mpb->num_disks--;
10350 super->updates_pending++;
24565c9a
DW		 10351 if (*dlp) {
10352 struct dl *dl = *dlp;
10353
10354 *dlp = (*dlp)->next;
10355 __free_imsm_disk(dl);
10356 }
ae6aad82 10357}
9a717282
AK		 10358
10359static void close_targets(int *targets, int new_disks)
10360{
10361 int i;
10362
10363 if (!targets)
10364 return;
10365
10366 for (i = 0; i < new_disks; i++) {
10367 if (targets[i] >= 0) {
10368 close(targets[i]);
10369 targets[i] = -1;
10370 }
10371 }
10372}
10373
10374static int imsm_get_allowed_degradation(int level, int raid_disks,
10375 struct intel_super *super,
10376 struct imsm_dev *dev)
10377{
10378 switch (level) {
bf5cf7c7 10379 case 1:
9a717282
AK		 10380 case 10:{
10381 int ret_val = 0;
10382 struct imsm_map *map;
10383 int i;
10384
10385 ret_val = raid_disks/2;
10386 /* check map if all disks pairs not failed
10387 * in both maps
10388 */
238c0a71 10389 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10390 for (i = 0; i < ret_val; i++) {
10391 int degradation = 0;
10392 if (get_imsm_disk(super, i) == NULL)
10393 degradation++;
10394 if (get_imsm_disk(super, i + 1) == NULL)
10395 degradation++;
10396 if (degradation == 2)
10397 return 0;
10398 }
238c0a71 10399 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10400 /* if there is no second map
10401 * result can be returned
10402 */
10403 if (map == NULL)
10404 return ret_val;
10405 /* check degradation in second map
10406 */
10407 for (i = 0; i < ret_val; i++) {
10408 int degradation = 0;
10409 if (get_imsm_disk(super, i) == NULL)
10410 degradation++;
10411 if (get_imsm_disk(super, i + 1) == NULL)
10412 degradation++;
10413 if (degradation == 2)
10414 return 0;
10415 }
10416 return ret_val;
10417 }
10418 case 5:
10419 return 1;
10420 case 6:
10421 return 2;
10422 default:
10423 return 0;
10424 }
10425}
10426
687629c2
AK		 10427/*******************************************************************************
10428 * Function: open_backup_targets
10429 * Description: Function opens file descriptors for all devices given in
10430 * info->devs
10431 * Parameters:
10432 * info : general array info
10433 * raid_disks : number of disks
10434 * raid_fds : table of device's file descriptors
9a717282
AK		 10435 * super : intel super for raid10 degradation check
10436 * dev : intel device for raid10 degradation check
687629c2
AK		 10437 * Returns:
10438 * 0 : success
10439 * -1 : fail
10440 ******************************************************************************/
9a717282
AK		 10441int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10442 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10443{
10444 struct mdinfo *sd;
f627f5ad 10445 int i;
9a717282 10446 int opened = 0;
f627f5ad
AK		 10447
10448 for (i = 0; i < raid_disks; i++)
10449 raid_fds[i] = -1;
687629c2
AK		 10450
10451 for (sd = info->devs ; sd ; sd = sd->next) {
10452 char *dn;
10453
10454 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10455 dprintf("disk is faulty!!\n");
10456 continue;
10457 }
10458
089f9d79 10459 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10460 continue;
10461
10462 dn = map_dev(sd->disk.major,
10463 sd->disk.minor, 1);
10464 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10465 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10466 pr_err("cannot open component\n");
9a717282 10467 continue;
687629c2 10468 }
9a717282
AK		 10469 opened++;
10470 }
10471 /* check if maximum array degradation level is not exceeded
10472 */
10473 if ((raid_disks - opened) >
089f9d79
JS		 10474 imsm_get_allowed_degradation(info->new_level, raid_disks,
10475 super, dev)) {
e12b3daa 10476 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10477 close_targets(raid_fds, raid_disks);
10478 return -2;
687629c2
AK		 10479 }
10480 return 0;
10481}
10482
d31ad643
PB		 10483/*******************************************************************************
10484 * Function: validate_container_imsm
10485 * Description: This routine validates container after assemble,
10486 * eg. if devices in container are under the same controller.
10487 *
10488 * Parameters:
10489 * info : linked list with info about devices used in array
10490 * Returns:
10491 * 1 : HBA mismatch
10492 * 0 : Success
10493 ******************************************************************************/
10494int validate_container_imsm(struct mdinfo *info)
10495{
6b781d33
AP		 10496 if (check_env("IMSM_NO_PLATFORM"))
10497 return 0;
d31ad643 10498
6b781d33
AP		 10499 struct sys_dev *idev;
10500 struct sys_dev *hba = NULL;
10501 struct sys_dev *intel_devices = find_intel_devices();
10502 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10503 info->disk.minor));
10504
10505 for (idev = intel_devices; idev; idev = idev->next) {
10506 if (dev_path && strstr(dev_path, idev->path)) {
10507 hba = idev;
10508 break;
d31ad643 10509 }
6b781d33
AP		 10510 }
10511 if (dev_path)
d31ad643
PB		 10512 free(dev_path);
10513
6b781d33
AP		 10514 if (!hba) {
10515 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10516 devid2kname(makedev(info->disk.major, info->disk.minor)));
10517 return 1;
10518 }
10519
10520 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10521 struct mdinfo *dev;
10522
10523 for (dev = info->next; dev; dev = dev->next) {
10524 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10525
10526 struct sys_dev *hba2 = NULL;
10527 for (idev = intel_devices; idev; idev = idev->next) {
10528 if (dev_path && strstr(dev_path, idev->path)) {
10529 hba2 = idev;
10530 break;
d31ad643
PB		 10531 }
10532 }
6b781d33
AP		 10533 if (dev_path)
10534 free(dev_path);
10535
10536 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10537 get_orom_by_device_id(hba2->dev_id);
10538
10539 if (hba2 && hba->type != hba2->type) {
10540 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10541 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10542 return 1;
10543 }
10544
07cb1e57 10545 if (orom != orom2) {
6b781d33
AP		 10546 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10547 " This operation is not supported and can lead to data loss.\n");
10548 return 1;
10549 }
10550
10551 if (!orom) {
10552 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10553 " This operation is not supported and can lead to data loss.\n");
10554 return 1;
10555 }
d31ad643 10556 }
6b781d33 10557
d31ad643
PB		 10558 return 0;
10559}
32141c17 10560
6f50473f
TM		 10561/*******************************************************************************
10562* Function: imsm_record_badblock
10563* Description: This routine stores new bad block record in BBM log
10564*
10565* Parameters:
10566* a : array containing a bad block
10567* slot : disk number containing a bad block
10568* sector : bad block sector
10569* length : bad block sectors range
10570* Returns:
10571* 1 : Success
10572* 0 : Error
10573******************************************************************************/
10574static int imsm_record_badblock(struct active_array *a, int slot,
10575 unsigned long long sector, int length)
10576{
10577 struct intel_super *super = a->container->sb;
10578 int ord;
10579 int ret;
10580
10581 ord = imsm_disk_slot_to_ord(a, slot);
10582 if (ord < 0)
10583 return 0;
10584
10585 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10586 length);
10587 if (ret)
10588 super->updates_pending++;
10589
10590 return ret;
10591}
c07a5a4f
TM		 10592/*******************************************************************************
10593* Function: imsm_clear_badblock
10594* Description: This routine clears bad block record from BBM log
10595*
10596* Parameters:
10597* a : array containing a bad block
10598* slot : disk number containing a bad block
10599* sector : bad block sector
10600* length : bad block sectors range
10601* Returns:
10602* 1 : Success
10603* 0 : Error
10604******************************************************************************/
10605static int imsm_clear_badblock(struct active_array *a, int slot,
10606 unsigned long long sector, int length)
10607{
10608 struct intel_super *super = a->container->sb;
10609 int ord;
10610 int ret;
10611
10612 ord = imsm_disk_slot_to_ord(a, slot);
10613 if (ord < 0)
10614 return 0;
10615
10616 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10617 if (ret)
10618 super->updates_pending++;
10619
10620 return ret;
10621}
928f1424
TM		 10622/*******************************************************************************
10623* Function: imsm_get_badblocks
10624* Description: This routine get list of bad blocks for an array
10625*
10626* Parameters:
10627* a : array
10628* slot : disk number
10629* Returns:
10630* bb : structure containing bad blocks
10631* NULL : error
10632******************************************************************************/
10633static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10634{
10635 int inst = a->info.container_member;
10636 struct intel_super *super = a->container->sb;
10637 struct imsm_dev *dev = get_imsm_dev(super, inst);
10638 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10639 int ord;
10640
10641 ord = imsm_disk_slot_to_ord(a, slot);
10642 if (ord < 0)
10643 return NULL;
10644
10645 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10646 per_dev_array_size(map), &super->bb);
928f1424
TM		 10647
10648 return &super->bb;
10649}
27156a57
TM		 10650/*******************************************************************************
10651* Function: examine_badblocks_imsm
10652* Description: Prints list of bad blocks on a disk to the standard output
10653*
10654* Parameters:
10655* st : metadata handler
10656* fd : open file descriptor for device
10657* devname : device name
10658* Returns:
10659* 0 : Success
10660* 1 : Error
10661******************************************************************************/
10662static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10663{
10664 struct intel_super *super = st->sb;
10665 struct bbm_log *log = super->bbm_log;
10666 struct dl *d = NULL;
10667 int any = 0;
10668
10669 for (d = super->disks; d ; d = d->next) {
10670 if (strcmp(d->devname, devname) == 0)
10671 break;
10672 }
10673
10674 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10675 pr_err("%s doesn't appear to be part of a raid array\n",
10676 devname);
10677 return 1;
10678 }
10679
10680 if (log != NULL) {
10681 unsigned int i;
10682 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10683
10684 for (i = 0; i < log->entry_count; i++) {
10685 if (entry[i].disk_ordinal == d->index) {
10686 unsigned long long sector = __le48_to_cpu(
10687 &entry[i].defective_block_start);
10688 int cnt = entry[i].marked_count + 1;
10689
10690 if (!any) {
10691 printf("Bad-blocks on %s:\n", devname);
10692 any = 1;
10693 }
10694
10695 printf("%20llu for %d sectors\n", sector, cnt);
10696 }
10697 }
10698 }
10699
10700 if (!any)
10701 printf("No bad-blocks list configured on %s\n", devname);
10702
10703 return 0;
10704}
687629c2
AK		 10705/*******************************************************************************
10706 * Function: init_migr_record_imsm
10707 * Description: Function inits imsm migration record
10708 * Parameters:
10709 * super : imsm internal array info
10710 * dev : device under migration
10711 * info : general array info to find the smallest device
10712 * Returns:
10713 * none
10714 ******************************************************************************/
10715void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10716 struct mdinfo *info)
10717{
10718 struct intel_super *super = st->sb;
10719 struct migr_record *migr_rec = super->migr_rec;
10720 int new_data_disks;
10721 unsigned long long dsize, dev_sectors;
10722 long long unsigned min_dev_sectors = -1LLU;
10723 struct mdinfo *sd;
10724 char nm[30];
10725 int fd;
238c0a71
AK		 10726 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10727 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10728 unsigned long long num_migr_units;
3ef4403c 10729 unsigned long long array_blocks;
687629c2
AK		 10730
10731 memset(migr_rec, 0, sizeof(struct migr_record));
10732 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10733
10734 /* only ascending reshape supported now */
10735 migr_rec->ascending_migr = __cpu_to_le32(1);
10736
10737 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10738 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10739 migr_rec->dest_depth_per_unit *=
10740 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10741 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10742 migr_rec->blocks_per_unit =
10743 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10744 migr_rec->dest_depth_per_unit =
10745 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10746 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10747 num_migr_units =
10748 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10749
10750 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10751 num_migr_units++;
9f421827 10752 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10753
10754 migr_rec->post_migr_vol_cap = dev->size_low;
10755 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10756
687629c2
AK		 10757 /* Find the smallest dev */
10758 for (sd = info->devs ; sd ; sd = sd->next) {
10759 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10760 fd = dev_open(nm, O_RDONLY);
10761 if (fd < 0)
10762 continue;
10763 get_dev_size(fd, NULL, &dsize);
10764 dev_sectors = dsize / 512;
10765 if (dev_sectors < min_dev_sectors)
10766 min_dev_sectors = dev_sectors;
10767 close(fd);
10768 }
9f421827 10769 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10770 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10771
10772 write_imsm_migr_rec(st);
10773
10774 return;
10775}
10776
10777/*******************************************************************************
10778 * Function: save_backup_imsm
10779 * Description: Function saves critical data stripes to Migration Copy Area
10780 * and updates the current migration unit status.
10781 * Use restore_stripes() to form a destination stripe,
10782 * and to write it to the Copy Area.
10783 * Parameters:
10784 * st : supertype information
aea93171 10785 * dev : imsm device that backup is saved for
687629c2
AK		 10786 * info : general array info
10787 * buf : input buffer
687629c2
AK		 10788 * length : length of data to backup (blocks_per_unit)
10789 * Returns:
10790 * 0 : success
10791 *, -1 : fail
10792 ******************************************************************************/
10793int save_backup_imsm(struct supertype *st,
10794 struct imsm_dev *dev,
10795 struct mdinfo *info,
10796 void *buf,
687629c2
AK		 10797 int length)
10798{
10799 int rv = -1;
10800 struct intel_super *super = st->sb;
594dc1b8
JS		 10801 unsigned long long *target_offsets;
10802 int *targets;
687629c2 10803 int i;
238c0a71 10804 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10805 int new_disks = map_dest->num_members;
ab724b98
AK		 10806 int dest_layout = 0;
10807 int dest_chunk;
d1877f69 10808 unsigned long long start;
9529d343 10809 int data_disks = imsm_num_data_members(map_dest);
687629c2 10810
503975b9 10811 targets = xmalloc(new_disks * sizeof(int));
687629c2 10812
7e45b550
AK		 10813 for (i = 0; i < new_disks; i++)
10814 targets[i] = -1;
10815
503975b9 10816 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10817
d1877f69 10818 start = info->reshape_progress * 512;
687629c2 10819 for (i = 0; i < new_disks; i++) {
9f421827 10820 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10821 /* move back copy area adderss, it will be moved forward
10822 * in restore_stripes() using start input variable
10823 */
10824 target_offsets[i] -= start/data_disks;
687629c2
AK		 10825 }
10826
9a717282
AK		 10827 if (open_backup_targets(info, new_disks, targets,
10828 super, dev))
687629c2
AK		 10829 goto abort;
10830
68eb8bc6 10831 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10832 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10833
687629c2
AK		 10834 if (restore_stripes(targets, /* list of dest devices */
10835 target_offsets, /* migration record offsets */
10836 new_disks,
ab724b98
AK		 10837 dest_chunk,
10838 map_dest->raid_level,
10839 dest_layout,
10840 -1, /* source backup file descriptor */
10841 0, /* input buf offset
10842 * always 0 buf is already offseted */
d1877f69 10843 start,
687629c2
AK		 10844 length,
10845 buf) != 0) {
e7b84f9d 10846 pr_err("Error restoring stripes\n");
687629c2
AK		 10847 goto abort;
10848 }
10849
10850 rv = 0;
10851
10852abort:
10853 if (targets) {
9a717282 10854 close_targets(targets, new_disks);
687629c2
AK		 10855 free(targets);
10856 }
10857 free(target_offsets);
10858
10859 return rv;
10860}
10861
10862/*******************************************************************************
10863 * Function: save_checkpoint_imsm
10864 * Description: Function called for current unit status update
10865 * in the migration record. It writes it to disk.
10866 * Parameters:
10867 * super : imsm internal array info
10868 * info : general array info
10869 * Returns:
10870 * 0: success
10871 * 1: failure
0228d92c
AK		 10872 * 2: failure, means no valid migration record
10873 * / no general migration in progress /
687629c2
AK		 10874 ******************************************************************************/
10875int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10876{
10877 struct intel_super *super = st->sb;
f8b72ef5
AK		 10878 unsigned long long blocks_per_unit;
10879 unsigned long long curr_migr_unit;
10880
2e062e82 10881 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10882 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10883 return 1;
10884 }
10885
f8b72ef5
AK		 10886 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10887 if (blocks_per_unit == 0) {
0228d92c
AK		 10888 dprintf("imsm: no migration in progress.\n");
10889 return 2;
687629c2 10890 }
f8b72ef5
AK		 10891 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10892 /* check if array is alligned to copy area
10893 * if it is not alligned, add one to current migration unit value
10894 * this can happend on array reshape finish only
10895 */
10896 if (info->reshape_progress % blocks_per_unit)
10897 curr_migr_unit++;
687629c2 10898
9f421827 10899 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10900 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10901 set_migr_dest_1st_member_lba(super->migr_rec,
10902 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10903
687629c2 10904 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10905 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10906 return 1;
10907 }
10908
10909 return 0;
10910}
10911
276d77db
AK		 10912/*******************************************************************************
10913 * Function: recover_backup_imsm
10914 * Description: Function recovers critical data from the Migration Copy Area
10915 * while assembling an array.
10916 * Parameters:
10917 * super : imsm internal array info
10918 * info : general array info
10919 * Returns:
10920 * 0 : success (or there is no data to recover)
10921 * 1 : fail
10922 ******************************************************************************/
10923int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10924{
10925 struct intel_super *super = st->sb;
10926 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10927 struct imsm_map *map_dest;
276d77db
AK		 10928 struct intel_dev *id = NULL;
10929 unsigned long long read_offset;
10930 unsigned long long write_offset;
10931 unsigned unit_len;
10932 int *targets = NULL;
10933 int new_disks, i, err;
10934 char *buf = NULL;
10935 int retval = 1;
f36a9ecd 10936 unsigned int sector_size = super->sector_size;
9f421827
PB		 10937 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10938 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10939 char buffer[20];
6c3560c0 10940 int skipped_disks = 0;
276d77db
AK		 10941
10942 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10943 if (err < 1)
10944 return 1;
10945
10946 /* recover data only during assemblation */
10947 if (strncmp(buffer, "inactive", 8) != 0)
10948 return 0;
10949 /* no data to recover */
10950 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10951 return 0;
10952 if (curr_migr_unit >= num_migr_units)
10953 return 1;
10954
10955 /* find device during reshape */
10956 for (id = super->devlist; id; id = id->next)
10957 if (is_gen_migration(id->dev))
10958 break;
10959 if (id == NULL)
10960 return 1;
10961
238c0a71 10962 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10963 new_disks = map_dest->num_members;
10964
9f421827 10965 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 10966
9f421827 10967 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 10968 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10969
10970 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10971 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10972 goto abort;
503975b9 10973 targets = xcalloc(new_disks, sizeof(int));
276d77db 10974
9a717282 10975 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10976 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10977 goto abort;
10978 }
276d77db
AK		 10979
10980 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10981 if (targets[i] < 0) {
10982 skipped_disks++;
10983 continue;
10984 }
276d77db 10985 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10986 pr_err("Cannot seek to block: %s\n",
10987 strerror(errno));
137debce
AK		 10988 skipped_disks++;
10989 continue;
276d77db 10990 }
9ec11d1a 10991 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10992 pr_err("Cannot read copy area block: %s\n",
10993 strerror(errno));
137debce
AK		 10994 skipped_disks++;
10995 continue;
276d77db
AK		 10996 }
10997 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10998 pr_err("Cannot seek to block: %s\n",
10999 strerror(errno));
137debce
AK		 11000 skipped_disks++;
11001 continue;
276d77db 11002 }
9ec11d1a 11003 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11004 pr_err("Cannot restore block: %s\n",
11005 strerror(errno));
137debce
AK		 11006 skipped_disks++;
11007 continue;
276d77db
AK		 11008 }
11009 }
11010
137debce
AK		 11011 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11012 new_disks,
11013 super,
11014 id->dev)) {
7a862a02 11015 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11016 goto abort;
11017 }
11018
befb629b
AK		 11019 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11020 /* ignore error == 2, this can mean end of reshape here
11021 */
7a862a02 11022 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11023 } else
276d77db 11024 retval = 0;
276d77db
AK		 11025
11026abort:
11027 if (targets) {
11028 for (i = 0; i < new_disks; i++)
11029 if (targets[i])
11030 close(targets[i]);
11031 free(targets);
11032 }
11033 free(buf);
11034 return retval;
11035}
11036
2cda7640
ML		 11037static char disk_by_path[] = "/dev/disk/by-path/";
11038
11039static const char *imsm_get_disk_controller_domain(const char *path)
11040{
2cda7640 11041 char disk_path[PATH_MAX];
96234762
LM		 11042 char *drv=NULL;
11043 struct stat st;
2cda7640 11044
6d8d290a 11045 strcpy(disk_path, disk_by_path);
96234762
LM		 11046 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11047 if (stat(disk_path, &st) == 0) {
11048 struct sys_dev* hba;
594dc1b8 11049 char *path;
96234762
LM		 11050
11051 path = devt_to_devpath(st.st_rdev);
11052 if (path == NULL)
11053 return "unknown";
11054 hba = find_disk_attached_hba(-1, path);
11055 if (hba && hba->type == SYS_DEV_SAS)
11056 drv = "isci";
11057 else if (hba && hba->type == SYS_DEV_SATA)
11058 drv = "ahci";
c6839718
MT		 11059 else if (hba && hba->type == SYS_DEV_VMD)
11060 drv = "vmd";
11061 else if (hba && hba->type == SYS_DEV_NVME)
11062 drv = "nvme";
1011e834 11063 else
96234762
LM		 11064 drv = "unknown";
11065 dprintf("path: %s hba: %s attached: %s\n",
11066 path, (hba) ? hba->path : "NULL", drv);
11067 free(path);
2cda7640 11068 }
96234762 11069 return drv;
2cda7640
ML		 11070}
11071
4dd2df09 11072static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11073{
4dd2df09 11074 static char devnm[32];
78b10e66
N		 11075 char subdev_name[20];
11076 struct mdstat_ent *mdstat;
11077
11078 sprintf(subdev_name, "%d", subdev);
11079 mdstat = mdstat_by_subdev(subdev_name, container);
11080 if (!mdstat)
4dd2df09 11081 return NULL;
78b10e66 11082
4dd2df09 11083 strcpy(devnm, mdstat->devnm);
78b10e66 11084 free_mdstat(mdstat);
4dd2df09 11085 return devnm;
78b10e66
N		 11086}
11087
11088static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11089 struct geo_params *geo,
fbf3d202
AK		 11090 int *old_raid_disks,
11091 int direction)
78b10e66 11092{
694575e7
KW		 11093 /* currently we only support increasing the number of devices
11094 * for a container. This increases the number of device for each
11095 * member array. They must all be RAID0 or RAID5.
11096 */
78b10e66
N		 11097 int ret_val = 0;
11098 struct mdinfo *info, *member;
11099 int devices_that_can_grow = 0;
11100
7a862a02 11101 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11102
d04f65f4 11103 if (geo->size > 0 ||
78b10e66
N		 11104 geo->level != UnSet ||
11105 geo->layout != UnSet ||
11106 geo->chunksize != 0 ||
11107 geo->raid_disks == UnSet) {
7a862a02 11108 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11109 return ret_val;
11110 }
11111
fbf3d202 11112 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11113 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11114 return ret_val;
11115 }
11116
78b10e66
N		 11117 info = container_content_imsm(st, NULL);
11118 for (member = info; member; member = member->next) {
4dd2df09 11119 char *result;
78b10e66
N		 11120
11121 dprintf("imsm: checking device_num: %i\n",
11122 member->container_member);
11123
d7d205bd 11124 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11125 /* we work on container for Online Capacity Expansion
11126 * only so raid_disks has to grow
11127 */
7a862a02 11128 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11129 break;
11130 }
11131
089f9d79 11132 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11133 /* we cannot use this container with other raid level
11134 */
7a862a02 11135 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11136 info->array.level);
11137 break;
11138 } else {
11139 /* check for platform support
11140 * for this raid level configuration
11141 */
11142 struct intel_super *super = st->sb;
11143 if (!is_raid_level_supported(super->orom,
11144 member->array.level,
11145 geo->raid_disks)) {
7a862a02 11146 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11147 info->array.level,
11148 geo->raid_disks,
11149 geo->raid_disks > 1 ? "s" : "");
11150 break;
11151 }
2a4a08e7
AK		 11152 /* check if component size is aligned to chunk size
11153 */
11154 if (info->component_size %
11155 (info->array.chunk_size/512)) {
7a862a02 11156 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11157 break;
11158 }
78b10e66
N		 11159 }
11160
11161 if (*old_raid_disks &&
11162 info->array.raid_disks != *old_raid_disks)
11163 break;
11164 *old_raid_disks = info->array.raid_disks;
11165
11166 /* All raid5 and raid0 volumes in container
11167 * have to be ready for Online Capacity Expansion
11168 * so they need to be assembled. We have already
11169 * checked that no recovery etc is happening.
11170 */
4dd2df09
N		 11171 result = imsm_find_array_devnm_by_subdev(member->container_member,
11172 st->container_devnm);
11173 if (result == NULL) {
78b10e66
N		 11174 dprintf("imsm: cannot find array\n");
11175 break;
11176 }
11177 devices_that_can_grow++;
11178 }
11179 sysfs_free(info);
11180 if (!member && devices_that_can_grow)
11181 ret_val = 1;
11182
11183 if (ret_val)
1ade5cc1 11184 dprintf("Container operation allowed\n");
78b10e66 11185 else
1ade5cc1 11186 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11187
11188 return ret_val;
11189}
11190
11191/* Function: get_spares_for_grow
11192 * Description: Allocates memory and creates list of spare devices
1011e834 11193 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11194 * Parameters: Pointer to the supertype structure
11195 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11196 * NULL if fail
78b10e66
N		 11197 */
11198static struct mdinfo *get_spares_for_grow(struct supertype *st)
11199{
fbfdcb06
AO		 11200 struct spare_criteria sc;
11201
11202 get_spare_criteria_imsm(st, &sc);
11203 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11204}
11205
11206/******************************************************************************
11207 * function: imsm_create_metadata_update_for_reshape
11208 * Function creates update for whole IMSM container.
11209 *
11210 ******************************************************************************/
11211static int imsm_create_metadata_update_for_reshape(
11212 struct supertype *st,
11213 struct geo_params *geo,
11214 int old_raid_disks,
11215 struct imsm_update_reshape **updatep)
11216{
11217 struct intel_super *super = st->sb;
11218 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11219 int update_memory_size;
11220 struct imsm_update_reshape *u;
11221 struct mdinfo *spares;
78b10e66 11222 int i;
594dc1b8 11223 int delta_disks;
bbd24d86 11224 struct mdinfo *dev;
78b10e66 11225
1ade5cc1 11226 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11227
11228 delta_disks = geo->raid_disks - old_raid_disks;
11229
11230 /* size of all update data without anchor */
11231 update_memory_size = sizeof(struct imsm_update_reshape);
11232
11233 /* now add space for spare disks that we need to add. */
11234 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11235
503975b9 11236 u = xcalloc(1, update_memory_size);
78b10e66
N		 11237 u->type = update_reshape_container_disks;
11238 u->old_raid_disks = old_raid_disks;
11239 u->new_raid_disks = geo->raid_disks;
11240
11241 /* now get spare disks list
11242 */
11243 spares = get_spares_for_grow(st);
11244
d7be7d87 11245 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11246 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11247 i = -1;
78b10e66
N		 11248 goto abort;
11249 }
11250
11251 /* we have got spares
11252 * update disk list in imsm_disk list table in anchor
11253 */
11254 dprintf("imsm: %i spares are available.\n\n",
11255 spares->array.spare_disks);
11256
bbd24d86 11257 dev = spares->devs;
78b10e66 11258 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11259 struct dl *dl;
11260
bbd24d86
AK		 11261 if (dev == NULL)
11262 break;
78b10e66
N		 11263 u->new_disks[i] = makedev(dev->disk.major,
11264 dev->disk.minor);
11265 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11266 dl->index = mpb->num_disks;
11267 mpb->num_disks++;
bbd24d86 11268 dev = dev->next;
78b10e66 11269 }
78b10e66
N		 11270
11271abort:
11272 /* free spares
11273 */
11274 sysfs_free(spares);
11275
d677e0b8 11276 dprintf("imsm: reshape update preparation :");
78b10e66 11277 if (i == delta_disks) {
1ade5cc1 11278 dprintf_cont(" OK\n");
78b10e66
N		 11279 *updatep = u;
11280 return update_memory_size;
11281 }
11282 free(u);
1ade5cc1 11283 dprintf_cont(" Error\n");
78b10e66
N		 11284
11285 return 0;
11286}
11287
f3871fdc
AK		 11288/******************************************************************************
11289 * function: imsm_create_metadata_update_for_size_change()
11290 * Creates update for IMSM array for array size change.
11291 *
11292 ******************************************************************************/
11293static int imsm_create_metadata_update_for_size_change(
11294 struct supertype *st,
11295 struct geo_params *geo,
11296 struct imsm_update_size_change **updatep)
11297{
11298 struct intel_super *super = st->sb;
594dc1b8
JS		 11299 int update_memory_size;
11300 struct imsm_update_size_change *u;
f3871fdc 11301
1ade5cc1 11302 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11303
11304 /* size of all update data without anchor */
11305 update_memory_size = sizeof(struct imsm_update_size_change);
11306
503975b9 11307 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11308 u->type = update_size_change;
11309 u->subdev = super->current_vol;
11310 u->new_size = geo->size;
11311
11312 dprintf("imsm: reshape update preparation : OK\n");
11313 *updatep = u;
11314
11315 return update_memory_size;
11316}
11317
48c5303a
PC		 11318/******************************************************************************
11319 * function: imsm_create_metadata_update_for_migration()
11320 * Creates update for IMSM array.
11321 *
11322 ******************************************************************************/
11323static int imsm_create_metadata_update_for_migration(
11324 struct supertype *st,
11325 struct geo_params *geo,
11326 struct imsm_update_reshape_migration **updatep)
11327{
11328 struct intel_super *super = st->sb;
594dc1b8
JS		 11329 int update_memory_size;
11330 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11331 struct imsm_dev *dev;
11332 int previous_level = -1;
11333
1ade5cc1 11334 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11335
11336 /* size of all update data without anchor */
11337 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11338
503975b9 11339 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11340 u->type = update_reshape_migration;
11341 u->subdev = super->current_vol;
11342 u->new_level = geo->level;
11343 u->new_layout = geo->layout;
11344 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11345 u->new_disks[0] = -1;
4bba0439 11346 u->new_chunksize = -1;
48c5303a
PC		 11347
11348 dev = get_imsm_dev(super, u->subdev);
11349 if (dev) {
11350 struct imsm_map *map;
11351
238c0a71 11352 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11353 if (map) {
11354 int current_chunk_size =
11355 __le16_to_cpu(map->blocks_per_strip) / 2;
11356
11357 if (geo->chunksize != current_chunk_size) {
11358 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11359 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11360 current_chunk_size, u->new_chunksize);
11361 }
48c5303a 11362 previous_level = map->raid_level;
4bba0439 11363 }
48c5303a 11364 }
089f9d79 11365 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11366 struct mdinfo *spares = NULL;
11367
11368 u->new_raid_disks++;
11369 spares = get_spares_for_grow(st);
089f9d79 11370 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11371 free(u);
11372 sysfs_free(spares);
11373 update_memory_size = 0;
565cc99e 11374 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11375 return 0;
11376 }
11377 sysfs_free(spares);
11378 }
11379 dprintf("imsm: reshape update preparation : OK\n");
11380 *updatep = u;
11381
11382 return update_memory_size;
11383}
11384
8dd70bce
AK		 11385static void imsm_update_metadata_locally(struct supertype *st,
11386 void *buf, int len)
11387{
11388 struct metadata_update mu;
11389
11390 mu.buf = buf;
11391 mu.len = len;
11392 mu.space = NULL;
11393 mu.space_list = NULL;
11394 mu.next = NULL;
5fe6f031
N		 11395 if (imsm_prepare_update(st, &mu))
11396 imsm_process_update(st, &mu);
8dd70bce
AK		 11397
11398 while (mu.space_list) {
11399 void **space = mu.space_list;
11400 mu.space_list = *space;
11401 free(space);
11402 }
11403}
78b10e66 11404
471bceb6 11405/***************************************************************************
694575e7 11406* Function: imsm_analyze_change
471bceb6 11407* Description: Function analyze change for single volume
1011e834 11408* and validate if transition is supported
fbf3d202
AK		 11409* Parameters: Geometry parameters, supertype structure,
11410* metadata change direction (apply/rollback)
694575e7 11411* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11412****************************************************************************/
11413enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11414 struct geo_params *geo,
11415 int direction)
694575e7 11416{
471bceb6
KW		 11417 struct mdinfo info;
11418 int change = -1;
11419 int check_devs = 0;
c21e737b 11420 int chunk;
67a2db32
AK		 11421 /* number of added/removed disks in operation result */
11422 int devNumChange = 0;
11423 /* imsm compatible layout value for array geometry verification */
11424 int imsm_layout = -1;
7abc9871
AK		 11425 int data_disks;
11426 struct imsm_dev *dev;
9529d343 11427 struct imsm_map *map;
7abc9871 11428 struct intel_super *super;
d04f65f4 11429 unsigned long long current_size;
65d38cca 11430 unsigned long long free_size;
d04f65f4 11431 unsigned long long max_size;
65d38cca 11432 int rv;
471bceb6
KW		 11433
11434 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11435 if (geo->level != info.array.level && geo->level >= 0 &&
11436 geo->level != UnSet) {
471bceb6
KW		 11437 switch (info.array.level) {
11438 case 0:
11439 if (geo->level == 5) {
b5347799 11440 change = CH_MIGRATION;
e13ce846 11441 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11442 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11443 change = -1;
11444 goto analyse_change_exit;
11445 }
67a2db32 11446 imsm_layout = geo->layout;
471bceb6 11447 check_devs = 1;
e91a3bad
LM		 11448 devNumChange = 1; /* parity disk added */
11449 } else if (geo->level == 10) {
471bceb6
KW		 11450 change = CH_TAKEOVER;
11451 check_devs = 1;
e91a3bad 11452 devNumChange = 2; /* two mirrors added */
67a2db32 11453 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11454 }
dfe77a9e
KW		 11455 break;
11456 case 1:
471bceb6
KW		 11457 case 10:
11458 if (geo->level == 0) {
11459 change = CH_TAKEOVER;
11460 check_devs = 1;
e91a3bad 11461 devNumChange = -(geo->raid_disks/2);
67a2db32 11462 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11463 }
11464 break;
11465 }
11466 if (change == -1) {
7a862a02 11467 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11468 info.array.level, geo->level);
471bceb6
KW		 11469 goto analyse_change_exit;
11470 }
11471 } else
11472 geo->level = info.array.level;
11473
089f9d79
JS		 11474 if (geo->layout != info.array.layout &&
11475 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11476 change = CH_MIGRATION;
089f9d79
JS		 11477 if (info.array.layout == 0 && info.array.level == 5 &&
11478 geo->layout == 5) {
471bceb6 11479 /* reshape 5 -> 4 */
089f9d79
JS		 11480 } else if (info.array.layout == 5 && info.array.level == 5 &&
11481 geo->layout == 0) {
471bceb6
KW		 11482 /* reshape 4 -> 5 */
11483 geo->layout = 0;
11484 geo->level = 5;
11485 } else {
7a862a02 11486 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11487 info.array.layout, geo->layout);
471bceb6
KW		 11488 change = -1;
11489 goto analyse_change_exit;
11490 }
67a2db32 11491 } else {
471bceb6 11492 geo->layout = info.array.layout;
67a2db32
AK		 11493 if (imsm_layout == -1)
11494 imsm_layout = info.array.layout;
11495 }
471bceb6 11496
089f9d79
JS		 11497 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11498 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11499 if (info.array.level == 10) {
11500 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11501 change = -1;
11502 goto analyse_change_exit;
1e9b2c3f
PB		 11503 } else if (info.component_size % (geo->chunksize/512)) {
11504 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11505 geo->chunksize/1024, info.component_size/2);
11506 change = -1;
11507 goto analyse_change_exit;
2d2b0eb7 11508 }
b5347799 11509 change = CH_MIGRATION;
2d2b0eb7 11510 } else {
471bceb6 11511 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11512 }
471bceb6 11513
c21e737b 11514 chunk = geo->chunksize / 1024;
7abc9871
AK		 11515
11516 super = st->sb;
11517 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11518 map = get_imsm_map(dev, MAP_0);
11519 data_disks = imsm_num_data_members(map);
c41e00b2 11520 /* compute current size per disk member
7abc9871 11521 */
c41e00b2
AK		 11522 current_size = info.custom_array_size / data_disks;
11523
089f9d79 11524 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11525 /* align component size
11526 */
3e684231 11527 geo->size = imsm_component_size_alignment_check(
c41e00b2 11528 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11529 chunk * 1024, super->sector_size,
c41e00b2 11530 geo->size * 2);
65d0b4ce 11531 if (geo->size == 0) {
7a862a02 11532 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11533 current_size);
11534 goto analyse_change_exit;
11535 }
c41e00b2 11536 }
7abc9871 11537
089f9d79 11538 if (current_size != geo->size && geo->size > 0) {
7abc9871 11539 if (change != -1) {
7a862a02 11540 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11541 change = -1;
11542 goto analyse_change_exit;
11543 }
11544 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11545 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11546 super->current_vol, st->devnm);
7abc9871
AK		 11547 goto analyse_change_exit;
11548 }
65d38cca
LD		 11549 /* check the maximum available size
11550 */
11551 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11552 0, chunk, &free_size);
11553 if (rv == 0)
11554 /* Cannot find maximum available space
11555 */
11556 max_size = 0;
11557 else {
11558 max_size = free_size + current_size;
11559 /* align component size
11560 */
3e684231 11561 max_size = imsm_component_size_alignment_check(
65d38cca 11562 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11563 chunk * 1024, super->sector_size,
65d38cca
LD		 11564 max_size);
11565 }
d04f65f4 11566 if (geo->size == MAX_SIZE) {
b130333f
AK		 11567 /* requested size change to the maximum available size
11568 */
65d38cca 11569 if (max_size == 0) {
7a862a02 11570 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11571 change = -1;
11572 goto analyse_change_exit;
65d38cca
LD		 11573 } else
11574 geo->size = max_size;
c41e00b2 11575 }
b130333f 11576
681b7ae2 11577 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11578 /* accept size for rollback only
11579 */
11580 } else {
11581 /* round size due to metadata compatibility
11582 */
11583 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11584 << SECT_PER_MB_SHIFT;
11585 dprintf("Prepare update for size change to %llu\n",
11586 geo->size );
11587 if (current_size >= geo->size) {
7a862a02 11588 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11589 current_size, geo->size);
fbf3d202
AK		 11590 goto analyse_change_exit;
11591 }
65d38cca 11592 if (max_size && geo->size > max_size) {
7a862a02 11593 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11594 max_size, geo->size);
65d38cca
LD		 11595 goto analyse_change_exit;
11596 }
7abc9871
AK		 11597 }
11598 geo->size *= data_disks;
11599 geo->raid_disks = dev->vol.map->num_members;
11600 change = CH_ARRAY_SIZE;
11601 }
471bceb6
KW		 11602 if (!validate_geometry_imsm(st,
11603 geo->level,
67a2db32 11604 imsm_layout,
e91a3bad 11605 geo->raid_disks + devNumChange,
c21e737b 11606 &chunk,
af4348dd 11607 geo->size, INVALID_SECTORS,
5308f117 11608 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11609 change = -1;
11610
11611 if (check_devs) {
11612 struct intel_super *super = st->sb;
11613 struct imsm_super *mpb = super->anchor;
11614
11615 if (mpb->num_raid_devs > 1) {
7a862a02 11616 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11617 geo->dev_name);
471bceb6
KW		 11618 change = -1;
11619 }
11620 }
11621
11622analyse_change_exit:
089f9d79
JS		 11623 if (direction == ROLLBACK_METADATA_CHANGES &&
11624 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11625 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11626 change = -1;
11627 }
471bceb6 11628 return change;
694575e7
KW		 11629}
11630
bb025c2f
KW		 11631int imsm_takeover(struct supertype *st, struct geo_params *geo)
11632{
11633 struct intel_super *super = st->sb;
11634 struct imsm_update_takeover *u;
11635
503975b9 11636 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11637
11638 u->type = update_takeover;
11639 u->subarray = super->current_vol;
11640
11641 /* 10->0 transition */
11642 if (geo->level == 0)
11643 u->direction = R10_TO_R0;
11644
0529c688
KW		 11645 /* 0->10 transition */
11646 if (geo->level == 10)
11647 u->direction = R0_TO_R10;
11648
bb025c2f
KW		 11649 /* update metadata locally */
11650 imsm_update_metadata_locally(st, u,
11651 sizeof(struct imsm_update_takeover));
11652 /* and possibly remotely */
11653 if (st->update_tail)
11654 append_metadata_update(st, u,
11655 sizeof(struct imsm_update_takeover));
11656 else
11657 free(u);
11658
11659 return 0;
11660}
11661
d04f65f4
N		 11662static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11663 int level,
78b10e66 11664 int layout, int chunksize, int raid_disks,
41784c88 11665 int delta_disks, char *backup, char *dev,
016e00f5 11666 int direction, int verbose)
78b10e66 11667{
78b10e66
N		 11668 int ret_val = 1;
11669 struct geo_params geo;
11670
1ade5cc1 11671 dprintf("(enter)\n");
78b10e66 11672
71204a50 11673 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11674
11675 geo.dev_name = dev;
4dd2df09 11676 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11677 geo.size = size;
11678 geo.level = level;
11679 geo.layout = layout;
11680 geo.chunksize = chunksize;
11681 geo.raid_disks = raid_disks;
41784c88
AK		 11682 if (delta_disks != UnSet)
11683 geo.raid_disks += delta_disks;
78b10e66 11684
1ade5cc1
N		 11685 dprintf("for level : %i\n", geo.level);
11686 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11687
4dd2df09 11688 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11689 /* On container level we can only increase number of devices. */
11690 dprintf("imsm: info: Container operation\n");
78b10e66 11691 int old_raid_disks = 0;
6dc0be30 11692
78b10e66 11693 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11694 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11695 struct imsm_update_reshape *u = NULL;
11696 int len;
11697
11698 len = imsm_create_metadata_update_for_reshape(
11699 st, &geo, old_raid_disks, &u);
11700
ed08d51c
AK		 11701 if (len <= 0) {
11702 dprintf("imsm: Cannot prepare update\n");
11703 goto exit_imsm_reshape_super;
11704 }
11705
8dd70bce
AK		 11706 ret_val = 0;
11707 /* update metadata locally */
11708 imsm_update_metadata_locally(st, u, len);
11709 /* and possibly remotely */
11710 if (st->update_tail)
11711 append_metadata_update(st, u, len);
11712 else
ed08d51c 11713 free(u);
8dd70bce 11714
694575e7 11715 } else {
7a862a02 11716 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11717 }
11718 } else {
11719 /* On volume level we support following operations
471bceb6
KW		 11720 * - takeover: raid10 -> raid0; raid0 -> raid10
11721 * - chunk size migration
11722 * - migration: raid5 -> raid0; raid0 -> raid5
11723 */
11724 struct intel_super *super = st->sb;
11725 struct intel_dev *dev = super->devlist;
4dd2df09 11726 int change;
694575e7 11727 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11728 /* find requested device */
11729 while (dev) {
1011e834 11730 char *devnm =
4dd2df09
N		 11731 imsm_find_array_devnm_by_subdev(
11732 dev->index, st->container_devnm);
11733 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11734 break;
11735 dev = dev->next;
11736 }
11737 if (dev == NULL) {
4dd2df09
N		 11738 pr_err("Cannot find %s (%s) subarray\n",
11739 geo.dev_name, geo.devnm);
471bceb6
KW		 11740 goto exit_imsm_reshape_super;
11741 }
11742 super->current_vol = dev->index;
fbf3d202 11743 change = imsm_analyze_change(st, &geo, direction);
694575e7 11744 switch (change) {
471bceb6 11745 case CH_TAKEOVER:
bb025c2f 11746 ret_val = imsm_takeover(st, &geo);
694575e7 11747 break;
48c5303a
PC		 11748 case CH_MIGRATION: {
11749 struct imsm_update_reshape_migration *u = NULL;
11750 int len =
11751 imsm_create_metadata_update_for_migration(
11752 st, &geo, &u);
11753 if (len < 1) {
7a862a02 11754 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11755 break;
11756 }
471bceb6 11757 ret_val = 0;
48c5303a
PC		 11758 /* update metadata locally */
11759 imsm_update_metadata_locally(st, u, len);
11760 /* and possibly remotely */
11761 if (st->update_tail)
11762 append_metadata_update(st, u, len);
11763 else
11764 free(u);
11765 }
11766 break;
7abc9871 11767 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11768 struct imsm_update_size_change *u = NULL;
11769 int len =
11770 imsm_create_metadata_update_for_size_change(
11771 st, &geo, &u);
11772 if (len < 1) {
7a862a02 11773 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11774 break;
11775 }
11776 ret_val = 0;
11777 /* update metadata locally */
11778 imsm_update_metadata_locally(st, u, len);
11779 /* and possibly remotely */
11780 if (st->update_tail)
11781 append_metadata_update(st, u, len);
11782 else
11783 free(u);
7abc9871
AK		 11784 }
11785 break;
471bceb6
KW		 11786 default:
11787 ret_val = 1;
694575e7 11788 }
694575e7 11789 }
78b10e66 11790
ed08d51c 11791exit_imsm_reshape_super:
78b10e66
N		 11792 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11793 return ret_val;
11794}
2cda7640 11795
0febb20c
AO		 11796#define COMPLETED_OK 0
11797#define COMPLETED_NONE 1
11798#define COMPLETED_DELAYED 2
11799
11800static int read_completed(int fd, unsigned long long *val)
11801{
11802 int ret;
11803 char buf[50];
11804
11805 ret = sysfs_fd_get_str(fd, buf, 50);
11806 if (ret < 0)
11807 return ret;
11808
11809 ret = COMPLETED_OK;
11810 if (strncmp(buf, "none", 4) == 0) {
11811 ret = COMPLETED_NONE;
11812 } else if (strncmp(buf, "delayed", 7) == 0) {
11813 ret = COMPLETED_DELAYED;
11814 } else {
11815 char *ep;
11816 *val = strtoull(buf, &ep, 0);
11817 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11818 ret = -1;
11819 }
11820 return ret;
11821}
11822
eee67a47
AK		 11823/*******************************************************************************
11824 * Function: wait_for_reshape_imsm
11825 * Description: Function writes new sync_max value and waits until
11826 * reshape process reach new position
11827 * Parameters:
11828 * sra : general array info
eee67a47
AK		 11829 * ndata : number of disks in new array's layout
11830 * Returns:
11831 * 0 : success,
11832 * 1 : there is no reshape in progress,
11833 * -1 : fail
11834 ******************************************************************************/
ae9f01f8 11835int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11836{
85ca499c 11837 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11838 int retry = 3;
eee67a47 11839 unsigned long long completed;
ae9f01f8
AK		 11840 /* to_complete : new sync_max position */
11841 unsigned long long to_complete = sra->reshape_progress;
11842 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11843
ae9f01f8 11844 if (fd < 0) {
1ade5cc1 11845 dprintf("cannot open reshape_position\n");
eee67a47 11846 return 1;
ae9f01f8 11847 }
eee67a47 11848
df2647fa
PB		 11849 do {
11850 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11851 if (!retry) {
11852 dprintf("cannot read reshape_position (no reshape in progres)\n");
11853 close(fd);
11854 return 1;
11855 }
11856 usleep(30000);
11857 } else
11858 break;
11859 } while (retry--);
eee67a47 11860
85ca499c 11861 if (completed > position_to_set) {
1ade5cc1 11862 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11863 to_complete, position_to_set);
ae9f01f8
AK		 11864 close(fd);
11865 return -1;
11866 }
11867 dprintf("Position set: %llu\n", position_to_set);
11868 if (sysfs_set_num(sra, NULL, "sync_max",
11869 position_to_set) != 0) {
1ade5cc1 11870 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11871 position_to_set);
11872 close(fd);
11873 return -1;
eee67a47
AK		 11874 }
11875
eee67a47 11876 do {
0febb20c 11877 int rc;
eee67a47 11878 char action[20];
5ff3a780 11879 int timeout = 3000;
0febb20c 11880
5ff3a780 11881 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11882 if (sysfs_get_str(sra, NULL, "sync_action",
11883 action, 20) > 0 &&
d7d3809a 11884 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11885 if (strncmp(action, "idle", 4) == 0)
11886 break;
d7d3809a
AP		 11887 close(fd);
11888 return -1;
11889 }
0febb20c
AO		 11890
11891 rc = read_completed(fd, &completed);
11892 if (rc < 0) {
1ade5cc1 11893 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11894 close(fd);
11895 return 1;
0febb20c
AO		 11896 } else if (rc == COMPLETED_NONE)
11897 break;
85ca499c 11898 } while (completed < position_to_set);
b2be2b62 11899
eee67a47
AK		 11900 close(fd);
11901 return 0;
eee67a47
AK		 11902}
11903
b915c95f
AK		 11904/*******************************************************************************
11905 * Function: check_degradation_change
11906 * Description: Check that array hasn't become failed.
11907 * Parameters:
11908 * info : for sysfs access
11909 * sources : source disks descriptors
11910 * degraded: previous degradation level
11911 * Returns:
11912 * degradation level
11913 ******************************************************************************/
11914int check_degradation_change(struct mdinfo *info,
11915 int *sources,
11916 int degraded)
11917{
11918 unsigned long long new_degraded;
e1993023
LD		 11919 int rv;
11920
11921 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11922 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11923 /* check each device to ensure it is still working */
11924 struct mdinfo *sd;
11925 new_degraded = 0;
11926 for (sd = info->devs ; sd ; sd = sd->next) {
11927 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11928 continue;
11929 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11930 char sbuf[100];
11931
b915c95f 11932 if (sysfs_get_str(info,
cf52eff5 11933 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11934 strstr(sbuf, "faulty") ||
11935 strstr(sbuf, "in_sync") == NULL) {
11936 /* this device is dead */
11937 sd->disk.state = (1<<MD_DISK_FAULTY);
11938 if (sd->disk.raid_disk >= 0 &&
11939 sources[sd->disk.raid_disk] >= 0) {
11940 close(sources[
11941 sd->disk.raid_disk]);
11942 sources[sd->disk.raid_disk] =
11943 -1;
11944 }
11945 new_degraded++;
11946 }
11947 }
11948 }
11949 }
11950
11951 return new_degraded;
11952}
11953
10f22854
AK		 11954/*******************************************************************************
11955 * Function: imsm_manage_reshape
11956 * Description: Function finds array under reshape and it manages reshape
11957 * process. It creates stripes backups (if required) and sets
942e1cdb 11958 * checkpoints.
10f22854
AK		 11959 * Parameters:
11960 * afd : Backup handle (nattive) - not used
11961 * sra : general array info
11962 * reshape : reshape parameters - not used
11963 * st : supertype structure
11964 * blocks : size of critical section [blocks]
11965 * fds : table of source device descriptor
11966 * offsets : start of array (offest per devices)
11967 * dests : not used
11968 * destfd : table of destination device descriptor
11969 * destoffsets : table of destination offsets (per device)
11970 * Returns:
11971 * 1 : success, reshape is done
11972 * 0 : fail
11973 ******************************************************************************/
999b4972
N		 11974static int imsm_manage_reshape(
11975 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11976 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11977 int *fds, unsigned long long *offsets,
11978 int dests, int *destfd, unsigned long long *destoffsets)
11979{
10f22854
AK		 11980 int ret_val = 0;
11981 struct intel_super *super = st->sb;
594dc1b8 11982 struct intel_dev *dv;
de44e46f 11983 unsigned int sector_size = super->sector_size;
10f22854 11984 struct imsm_dev *dev = NULL;
9529d343 11985 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11986 int migr_vol_qan = 0;
11987 int ndata, odata; /* [bytes] */
11988 int chunk; /* [bytes] */
11989 struct migr_record *migr_rec;
11990 char *buf = NULL;
11991 unsigned int buf_size; /* [bytes] */
11992 unsigned long long max_position; /* array size [bytes] */
11993 unsigned long long next_step; /* [blocks]/[bytes] */
11994 unsigned long long old_data_stripe_length;
10f22854
AK		 11995 unsigned long long start_src; /* [bytes] */
11996 unsigned long long start; /* [bytes] */
11997 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11998 int degraded = 0;
ab724b98 11999 int source_layout = 0;
10f22854 12000
79a16a9b
JS		 12001 if (!sra)
12002 return ret_val;
12003
12004 if (!fds || !offsets)
10f22854
AK		 12005 goto abort;
12006
12007 /* Find volume during the reshape */
12008 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12009 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12010 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12011 dev = dv->dev;
12012 migr_vol_qan++;
12013 }
12014 }
12015 /* Only one volume can migrate at the same time */
12016 if (migr_vol_qan != 1) {
676e87a8 12017 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12018 "Number of migrating volumes greater than 1\n" :
12019 "There is no volume during migrationg\n");
12020 goto abort;
12021 }
12022
9529d343 12023 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12024 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12025 if (map_src == NULL)
12026 goto abort;
10f22854 12027
9529d343
MD		 12028 ndata = imsm_num_data_members(map_dest);
12029 odata = imsm_num_data_members(map_src);
10f22854 12030
7b1ab482 12031 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12032 old_data_stripe_length = odata * chunk;
12033
12034 migr_rec = super->migr_rec;
12035
10f22854
AK		 12036 /* initialize migration record for start condition */
12037 if (sra->reshape_progress == 0)
12038 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12039 else {
12040 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12041 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12042 goto abort;
12043 }
6a75c8ca
AK		 12044 /* Save checkpoint to update migration record for current
12045 * reshape position (in md). It can be farther than current
12046 * reshape position in metadata.
12047 */
12048 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12049 /* ignore error == 2, this can mean end of reshape here
12050 */
7a862a02 12051 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12052 goto abort;
12053 }
b2c59438 12054 }
10f22854
AK		 12055
12056 /* size for data */
12057 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12058 /* extend buffer size for parity disk */
12059 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12060 /* add space for stripe alignment */
10f22854 12061 buf_size += old_data_stripe_length;
de44e46f
PB		 12062 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12063 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12064 goto abort;
12065 }
12066
3ef4403c 12067 max_position = sra->component_size * ndata;
68eb8bc6 12068 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12069
9f421827
PB		 12070 while (current_migr_unit(migr_rec) <
12071 get_num_migr_units(migr_rec)) {
10f22854
AK		 12072 /* current reshape position [blocks] */
12073 unsigned long long current_position =
12074 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12075 * current_migr_unit(migr_rec);
10f22854
AK		 12076 unsigned long long border;
12077
b915c95f
AK		 12078 /* Check that array hasn't become failed.
12079 */
12080 degraded = check_degradation_change(sra, fds, degraded);
12081 if (degraded > 1) {
7a862a02 12082 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12083 goto abort;
12084 }
12085
10f22854
AK		 12086 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12087
12088 if ((current_position + next_step) > max_position)
12089 next_step = max_position - current_position;
12090
92144abf 12091 start = current_position * 512;
10f22854 12092
942e1cdb 12093 /* align reading start to old geometry */
10f22854
AK		 12094 start_buf_shift = start % old_data_stripe_length;
12095 start_src = start - start_buf_shift;
12096
12097 border = (start_src / odata) - (start / ndata);
12098 border /= 512;
12099 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12100 /* save critical stripes to buf
12101 * start - start address of current unit
12102 * to backup [bytes]
12103 * start_src - start address of current unit
12104 * to backup alligned to source array
12105 * [bytes]
12106 */
594dc1b8 12107 unsigned long long next_step_filler;
10f22854
AK		 12108 unsigned long long copy_length = next_step * 512;
12109
12110 /* allign copy area length to stripe in old geometry */
12111 next_step_filler = ((copy_length + start_buf_shift)
12112 % old_data_stripe_length);
12113 if (next_step_filler)
12114 next_step_filler = (old_data_stripe_length
12115 - next_step_filler);
7a862a02 12116 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		 12117 start, start_src, copy_length,
12118 start_buf_shift, next_step_filler);
12119
12120 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12121 chunk, map_src->raid_level,
12122 source_layout, 0, NULL, start_src,
10f22854
AK		 12123 copy_length +
12124 next_step_filler + start_buf_shift,
12125 buf)) {
7a862a02 12126 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12127 goto abort;
12128 }
12129 /* Convert data to destination format and store it
12130 * in backup general migration area
12131 */
12132 if (save_backup_imsm(st, dev, sra,
aea93171 12133 buf + start_buf_shift, copy_length)) {
7a862a02 12134 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12135 goto abort;
12136 }
12137 if (save_checkpoint_imsm(st, sra,
12138 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12139 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12140 goto abort;
12141 }
8016a6d4
AK		 12142 } else {
12143 /* set next step to use whole border area */
12144 border /= next_step;
12145 if (border > 1)
12146 next_step *= border;
10f22854
AK		 12147 }
12148 /* When data backed up, checkpoint stored,
12149 * kick the kernel to reshape unit of data
12150 */
12151 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12152 /* limit next step to array max position */
12153 if (next_step > max_position)
12154 next_step = max_position;
10f22854
AK		 12155 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12156 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12157 sra->reshape_progress = next_step;
10f22854
AK		 12158
12159 /* wait until reshape finish */
c85338c6 12160 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12161 dprintf("wait_for_reshape_imsm returned error!\n");
12162 goto abort;
12163 }
84d11e6c
N		 12164 if (sigterm)
12165 goto abort;
10f22854 12166
0228d92c
AK		 12167 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12168 /* ignore error == 2, this can mean end of reshape here
12169 */
7a862a02 12170 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12171 goto abort;
12172 }
12173
12174 }
12175
71e5411e
PB		 12176 /* clear migr_rec on disks after successful migration */
12177 struct dl *d;
12178
85337573 12179 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12180 for (d = super->disks; d; d = d->next) {
12181 if (d->index < 0 || is_failed(&d->disk))
12182 continue;
12183 unsigned long long dsize;
12184
12185 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12186 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12187 SEEK_SET) >= 0) {
466070ad 12188 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12189 MIGR_REC_BUF_SECTORS*sector_size) !=
12190 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12191 perror("Write migr_rec failed");
12192 }
12193 }
12194
10f22854
AK		 12195 /* return '1' if done */
12196 ret_val = 1;
12197abort:
12198 free(buf);
942e1cdb
N		 12199 /* See Grow.c: abort_reshape() for further explanation */
12200 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12201 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12202 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12203
12204 return ret_val;
999b4972 12205}
0c21b485 12206
cdddbdbc 12207struct superswitch super_imsm = {
cdddbdbc
DW		 12208 .examine_super = examine_super_imsm,
12209 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12210 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12211 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12212 .detail_super = detail_super_imsm,
12213 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12214 .write_init_super = write_init_super_imsm,
0e600426
N		 12215 .validate_geometry = validate_geometry_imsm,
12216 .add_to_super = add_to_super_imsm,
1a64be56 12217 .remove_from_super = remove_from_super_imsm,
d665cc31 12218 .detail_platform = detail_platform_imsm,
e50cf220 12219 .export_detail_platform = export_detail_platform_imsm,
33414a01 12220 .kill_subarray = kill_subarray_imsm,
aa534678 12221 .update_subarray = update_subarray_imsm,
2b959fbf 12222 .load_container = load_container_imsm,
71204a50
N		 12223 .default_geometry = default_geometry_imsm,
12224 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12225 .reshape_super = imsm_reshape_super,
12226 .manage_reshape = imsm_manage_reshape,
9e2d750d 12227 .recover_backup = recover_backup_imsm,
27156a57 12228 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12229 .match_home = match_home_imsm,
12230 .uuid_from_super= uuid_from_super_imsm,
12231 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12232 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12233 .update_super = update_super_imsm,
12234
12235 .avail_size = avail_size_imsm,
fbfdcb06 12236 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12237
12238 .compare_super = compare_super_imsm,
12239
12240 .load_super = load_super_imsm,
bf5a934a 12241 .init_super = init_super_imsm,
e683ca88 12242 .store_super = store_super_imsm,
cdddbdbc
DW		 12243 .free_super = free_super_imsm,
12244 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12245 .container_content = container_content_imsm,
0c21b485 12246 .validate_container = validate_container_imsm,
cdddbdbc 12247
2432ce9b
AP		 12248 .write_init_ppl = write_init_ppl_imsm,
12249 .validate_ppl = validate_ppl_imsm,
12250
cdddbdbc 12251 .external = 1,
4cce4069 12252 .name = "imsm",
845dea95
NB		 12253
12254/* for mdmon */
12255 .open_new = imsm_open_new,
ed9d66aa 12256 .set_array_state= imsm_set_array_state,
845dea95
NB		 12257 .set_disk = imsm_set_disk,
12258 .sync_metadata = imsm_sync_metadata,
88758e9d 12259 .activate_spare = imsm_activate_spare,
e8319a19 12260 .process_update = imsm_process_update,
8273f55e 12261 .prepare_update = imsm_prepare_update,
6f50473f 12262 .record_bad_block = imsm_record_badblock,
c07a5a4f 12263 .clear_bad_block = imsm_clear_badblock,
928f1424 12264 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12265};
Unnamed repository; edit this file 'description' to name the repository.