]> 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
Manage, imsm: Write metadata before add
[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);
ba1b3bc8 1582 printf(" Subarray : %d\n", super->current_vol);
44470971 1583 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1584 printf(" RAID Level : %d", get_imsm_raid_level(map));
1585 if (map2)
1586 printf(" <-- %d", get_imsm_raid_level(map2));
1587 printf("\n");
1588 printf(" Members : %d", map->num_members);
1589 if (map2)
1590 printf(" <-- %d", map2->num_members);
1591 printf("\n");
0d80bb2f
DW		 1592 printf(" Slots : [");
1593 for (i = 0; i < map->num_members; i++) {
238c0a71 1594 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1595 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1596 }
dd8bcb3b
AK		 1597 printf("]");
1598 if (map2) {
1599 printf(" <-- [");
1600 for (i = 0; i < map2->num_members; i++) {
238c0a71 1601 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1602 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1603 }
1604 printf("]");
1605 }
1606 printf("\n");
7095bccb
AK		 1607 printf(" Failed disk : ");
1608 if (map->failed_disk_num == 0xff)
1609 printf("none");
1610 else
1611 printf("%i", map->failed_disk_num);
1612 printf("\n");
620b1713
DW		 1613 slot = get_imsm_disk_slot(map, disk_idx);
1614 if (slot >= 0) {
238c0a71 1615 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1616 printf(" This Slot : %d%s\n", slot,
1617 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1618 } else
cdddbdbc 1619 printf(" This Slot : ?\n");
84918897 1620 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1621 sz = imsm_dev_size(dev);
84918897
MK		 1622 printf(" Array Size : %llu%s\n",
1623 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1624 human_size(sz * 512));
5551b113 1625 sz = blocks_per_member(map);
84918897
MK		 1626 printf(" Per Dev Size : %llu%s\n",
1627 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1628 human_size(sz * 512));
5551b113
CA		 1629 printf(" Sector Offset : %llu\n",
1630 pba_of_lba0(map));
1631 printf(" Num Stripes : %llu\n",
1632 num_data_stripes(map));
dd8bcb3b 1633 printf(" Chunk Size : %u KiB",
cdddbdbc 1634 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1635 if (map2)
1636 printf(" <-- %u KiB",
1637 __le16_to_cpu(map2->blocks_per_strip) / 2);
1638 printf("\n");
cdddbdbc 1639 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1640 printf(" Migrate State : ");
1484e727
DW		 1641 if (dev->vol.migr_state) {
1642 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1643 printf("initialize\n");
1484e727 1644 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1645 printf("rebuild\n");
1484e727 1646 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1647 printf("check\n");
1484e727 1648 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1649 printf("general migration\n");
1484e727 1650 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1651 printf("state change\n");
1484e727 1652 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1653 printf("repair\n");
1484e727 1654 else
8655a7b1
DW		 1655 printf("<unknown:%d>\n", migr_type(dev));
1656 } else
1657 printf("idle\n");
3393c6af
DW		 1658 printf(" Map State : %s", map_state_str[map->map_state]);
1659 if (dev->vol.migr_state) {
238c0a71 1660 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1661
b10b37b8 1662 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1663 printf("\n Checkpoint : %u ",
1664 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1665 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1666 printf("(N/A)");
1667 else
1668 printf("(%llu)", (unsigned long long)
1669 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1670 }
1671 printf("\n");
2432ce9b
AP		 1672 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1673 "dirty" : "clean");
1674 printf(" RWH Policy : ");
c2462068 1675 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1676 printf("off\n");
1677 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1678 printf("PPL distributed\n");
1679 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1680 printf("PPL journaling drive\n");
c2462068
PB		 1681 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1682 printf("Multiple distributed PPLs\n");
1683 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1684 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1685 else
1686 printf("<unknown:%d>\n", dev->rwh_policy);
ba1b3bc8
AP		 1687
1688 printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num);
cdddbdbc
DW		 1689}
1690
ef5c214e
MK		 1691static void print_imsm_disk(struct imsm_disk *disk,
1692 int index,
1693 __u32 reserved,
1694 unsigned int sector_size) {
1f24f035 1695 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1696 __u64 sz;
1697
0ec1f4e8 1698 if (index < -1 || !disk)
e9d82038
DW		 1699 return;
1700
cdddbdbc 1701 printf("\n");
1f24f035 1702 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1703 if (index >= 0)
1704 printf(" Disk%02d Serial : %s\n", index, str);
1705 else
1706 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1707 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1708 is_configured(disk) ? " active" : "",
1709 is_failed(disk) ? " failed" : "",
1710 is_journal(disk) ? " journal" : "");
cdddbdbc 1711 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1712 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1713 printf(" Usable Size : %llu%s\n",
1714 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1715 human_size(sz * 512));
1716}
1717
de44e46f
PB		 1718void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1719{
1720 struct migr_record *migr_rec = super->migr_rec;
1721
1722 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1723 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1724 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1725 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1726 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1727 set_migr_chkp_area_pba(migr_rec,
1728 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1729 set_migr_dest_1st_member_lba(migr_rec,
1730 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1731}
1732
f36a9ecd
PB		 1733void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1734{
1735 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1736}
1737
1738void convert_to_4k(struct intel_super *super)
1739{
1740 struct imsm_super *mpb = super->anchor;
1741 struct imsm_disk *disk;
1742 int i;
e4467bc7 1743 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1744
1745 for (i = 0; i < mpb->num_disks ; i++) {
1746 disk = __get_imsm_disk(mpb, i);
1747 /* disk */
1748 convert_to_4k_imsm_disk(disk);
1749 }
1750 for (i = 0; i < mpb->num_raid_devs; i++) {
1751 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1752 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1753 /* dev */
fcc2c9da 1754 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1755 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1756
1757 /* map0 */
1758 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1759 map->blocks_per_strip /= IMSM_4K_DIV;
1760 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1761
1762 if (dev->vol.migr_state) {
1763 /* map1 */
1764 map = get_imsm_map(dev, MAP_1);
1765 set_blocks_per_member(map,
1766 blocks_per_member(map)/IMSM_4K_DIV);
1767 map->blocks_per_strip /= IMSM_4K_DIV;
1768 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1769 }
1770 }
e4467bc7
TM		 1771 if (bbm_log_size) {
1772 struct bbm_log *log = (void *)mpb +
1773 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1774 __u32 i;
1775
1776 for (i = 0; i < log->entry_count; i++) {
1777 struct bbm_log_entry *entry =
1778 &log->marked_block_entries[i];
1779
1780 __u8 count = entry->marked_count + 1;
1781 unsigned long long sector =
1782 __le48_to_cpu(&entry->defective_block_start);
1783
1784 entry->defective_block_start =
1785 __cpu_to_le48(sector/IMSM_4K_DIV);
1786 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1787 }
1788 }
f36a9ecd
PB		 1789
1790 mpb->check_sum = __gen_imsm_checksum(mpb);
1791}
1792
520e69e2
AK		 1793void examine_migr_rec_imsm(struct intel_super *super)
1794{
1795 struct migr_record *migr_rec = super->migr_rec;
1796 struct imsm_super *mpb = super->anchor;
1797 int i;
1798
1799 for (i = 0; i < mpb->num_raid_devs; i++) {
1800 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1801 struct imsm_map *map;
b4ab44d8 1802 int slot = -1;
3136abe5 1803
520e69e2
AK		 1804 if (is_gen_migration(dev) == 0)
1805 continue;
1806
1807 printf("\nMigration Record Information:");
3136abe5 1808
44bfe6df
AK		 1809 /* first map under migration */
1810 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1811 if (map)
1812 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1813 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1814 printf(" Empty\n ");
1815 printf("Examine one of first two disks in array\n");
1816 break;
1817 }
1818 printf("\n Status : ");
1819 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1820 printf("Normal\n");
1821 else
1822 printf("Contains Data\n");
9f421827
PB		 1823 printf(" Current Unit : %llu\n",
1824 current_migr_unit(migr_rec));
520e69e2
AK		 1825 printf(" Family : %u\n",
1826 __le32_to_cpu(migr_rec->family_num));
1827 printf(" Ascending : %u\n",
1828 __le32_to_cpu(migr_rec->ascending_migr));
1829 printf(" Blocks Per Unit : %u\n",
1830 __le32_to_cpu(migr_rec->blocks_per_unit));
1831 printf(" Dest. Depth Per Unit : %u\n",
1832 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1833 printf(" Checkpoint Area pba : %llu\n",
1834 migr_chkp_area_pba(migr_rec));
1835 printf(" First member lba : %llu\n",
1836 migr_dest_1st_member_lba(migr_rec));
1837 printf(" Total Number of Units : %llu\n",
1838 get_num_migr_units(migr_rec));
1839 printf(" Size of volume : %llu\n",
1840 join_u32(migr_rec->post_migr_vol_cap,
1841 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1842 printf(" Record was read from : %u\n",
1843 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1844
1845 break;
1846 }
1847}
f36a9ecd 1848
de44e46f
PB		 1849void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1850{
1851 struct migr_record *migr_rec = super->migr_rec;
1852
1853 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1854 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1855 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1856 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1857 &migr_rec->post_migr_vol_cap,
1858 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1859 set_migr_chkp_area_pba(migr_rec,
1860 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1861 set_migr_dest_1st_member_lba(migr_rec,
1862 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1863}
1864
f36a9ecd
PB		 1865void convert_from_4k(struct intel_super *super)
1866{
1867 struct imsm_super *mpb = super->anchor;
1868 struct imsm_disk *disk;
1869 int i;
e4467bc7 1870 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1871
1872 for (i = 0; i < mpb->num_disks ; i++) {
1873 disk = __get_imsm_disk(mpb, i);
1874 /* disk */
1875 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1876 }
1877
1878 for (i = 0; i < mpb->num_raid_devs; i++) {
1879 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1880 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1881 /* dev */
fcc2c9da 1882 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1883 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1884
1885 /* map0 */
1886 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1887 map->blocks_per_strip *= IMSM_4K_DIV;
1888 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1889
1890 if (dev->vol.migr_state) {
1891 /* map1 */
1892 map = get_imsm_map(dev, MAP_1);
1893 set_blocks_per_member(map,
1894 blocks_per_member(map)*IMSM_4K_DIV);
1895 map->blocks_per_strip *= IMSM_4K_DIV;
1896 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1897 }
1898 }
e4467bc7
TM		 1899 if (bbm_log_size) {
1900 struct bbm_log *log = (void *)mpb +
1901 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1902 __u32 i;
1903
1904 for (i = 0; i < log->entry_count; i++) {
1905 struct bbm_log_entry *entry =
1906 &log->marked_block_entries[i];
1907
1908 __u8 count = entry->marked_count + 1;
1909 unsigned long long sector =
1910 __le48_to_cpu(&entry->defective_block_start);
1911
1912 entry->defective_block_start =
1913 __cpu_to_le48(sector*IMSM_4K_DIV);
1914 entry->marked_count = count*IMSM_4K_DIV - 1;
1915 }
1916 }
f36a9ecd
PB		 1917
1918 mpb->check_sum = __gen_imsm_checksum(mpb);
1919}
1920
19482bcc
AK		 1921/*******************************************************************************
1922 * function: imsm_check_attributes
1923 * Description: Function checks if features represented by attributes flags
1011e834 1924 * are supported by mdadm.
19482bcc
AK		 1925 * Parameters:
1926 * attributes - Attributes read from metadata
1927 * Returns:
1011e834
N		 1928 * 0 - passed attributes contains unsupported features flags
1929 * 1 - all features are supported
19482bcc
AK		 1930 ******************************************************************************/
1931static int imsm_check_attributes(__u32 attributes)
1932{
1933 int ret_val = 1;
418f9b36
N		 1934 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1935
1936 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1937
1938 not_supported &= attributes;
1939 if (not_supported) {
e7b84f9d 1940 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1941 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1942 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1943 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1944 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1945 }
1946 if (not_supported & MPB_ATTRIB_2TB) {
1947 dprintf("\t\tMPB_ATTRIB_2TB\n");
1948 not_supported ^= MPB_ATTRIB_2TB;
1949 }
1950 if (not_supported & MPB_ATTRIB_RAID0) {
1951 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1952 not_supported ^= MPB_ATTRIB_RAID0;
1953 }
1954 if (not_supported & MPB_ATTRIB_RAID1) {
1955 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1956 not_supported ^= MPB_ATTRIB_RAID1;
1957 }
1958 if (not_supported & MPB_ATTRIB_RAID10) {
1959 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1960 not_supported ^= MPB_ATTRIB_RAID10;
1961 }
1962 if (not_supported & MPB_ATTRIB_RAID1E) {
1963 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1964 not_supported ^= MPB_ATTRIB_RAID1E;
1965 }
1966 if (not_supported & MPB_ATTRIB_RAID5) {
1967 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1968 not_supported ^= MPB_ATTRIB_RAID5;
1969 }
1970 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1971 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1972 not_supported ^= MPB_ATTRIB_RAIDCNG;
1973 }
1974 if (not_supported & MPB_ATTRIB_BBM) {
1975 dprintf("\t\tMPB_ATTRIB_BBM\n");
1976 not_supported ^= MPB_ATTRIB_BBM;
1977 }
1978 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1979 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1980 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1981 }
1982 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1983 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1984 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1985 }
1986 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1987 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1988 not_supported ^= MPB_ATTRIB_2TB_DISK;
1989 }
1990 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1991 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1992 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1993 }
1994 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1995 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1996 not_supported ^= MPB_ATTRIB_NEVER_USE;
1997 }
1998
1999 if (not_supported)
1ade5cc1 2000 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 2001
2002 ret_val = 0;
2003 }
2004
2005 return ret_val;
2006}
2007
a5d85af7 2008static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2009
cdddbdbc
DW		 2010static void examine_super_imsm(struct supertype *st, char *homehost)
2011{
2012 struct intel_super *super = st->sb;
949c47a0 2013 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2014 char str[MAX_SIGNATURE_LENGTH];
2015 int i;
27fd6274
DW		 2016 struct mdinfo info;
2017 char nbuf[64];
cdddbdbc 2018 __u32 sum;
14e8215b 2019 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2020 struct dl *dl;
e48aed3c 2021 time_t creation_time;
27fd6274 2022
618f4e6d
XN		 2023 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2024 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2025 printf(" Magic : %s\n", str);
cdddbdbc 2026 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2027 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2028 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2029 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
e48aed3c
AP		 2030 creation_time = __le64_to_cpu(mpb->creation_time);
2031 printf(" Creation Time : %.24s\n",
2032 creation_time ? ctime(&creation_time) : "Unknown");
19482bcc
AK		 2033 printf(" Attributes : ");
2034 if (imsm_check_attributes(mpb->attributes))
2035 printf("All supported\n");
2036 else
2037 printf("not supported\n");
a5d85af7 2038 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2039 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2040 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2041 sum = __le32_to_cpu(mpb->check_sum);
2042 printf(" Checksum : %08x %s\n", sum,
949c47a0 2043 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2044 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2045 printf(" Disks : %d\n", mpb->num_disks);
2046 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2047 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2048 super->disks->index, reserved, super->sector_size);
8d67477f 2049 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2050 struct bbm_log *log = super->bbm_log;
2051
2052 printf("\n");
2053 printf("Bad Block Management Log:\n");
2054 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2055 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2056 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2057 }
44470971
DW		 2058 for (i = 0; i < mpb->num_raid_devs; i++) {
2059 struct mdinfo info;
2060 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2061
2062 super->current_vol = i;
a5d85af7 2063 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2064 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2065 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2066 }
cdddbdbc
DW		 2067 for (i = 0; i < mpb->num_disks; i++) {
2068 if (i == super->disks->index)
2069 continue;
ef5c214e
MK		 2070 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2071 super->sector_size);
cdddbdbc 2072 }
94827db3 2073
0ec1f4e8
DW		 2074 for (dl = super->disks; dl; dl = dl->next)
2075 if (dl->index == -1)
ef5c214e
MK		 2076 print_imsm_disk(&dl->disk, -1, reserved,
2077 super->sector_size);
520e69e2
AK		 2078
2079 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2080}
2081
061f2c6a 2082static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2083{
27fd6274 2084 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2085 struct mdinfo info;
2086 char nbuf[64];
1e7bc0ed 2087 struct intel_super *super = st->sb;
1e7bc0ed 2088
0d5a423f
DW		 2089 if (!super->anchor->num_raid_devs) {
2090 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2091 return;
0d5a423f 2092 }
ff54de6e 2093
a5d85af7 2094 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2095 fname_from_uuid(st, &info, nbuf, ':');
2096 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2097}
2098
2099static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2100{
2101 /* We just write a generic IMSM ARRAY entry */
2102 struct mdinfo info;
2103 char nbuf[64];
2104 char nbuf1[64];
2105 struct intel_super *super = st->sb;
2106 int i;
2107
2108 if (!super->anchor->num_raid_devs)
2109 return;
2110
a5d85af7 2111 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2112 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2113 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2114 struct imsm_dev *dev = get_imsm_dev(super, i);
2115
2116 super->current_vol = i;
a5d85af7 2117 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2118 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2119 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2120 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2121 }
cdddbdbc
DW		 2122}
2123
9d84c8ea
DW		 2124static void export_examine_super_imsm(struct supertype *st)
2125{
2126 struct intel_super *super = st->sb;
2127 struct imsm_super *mpb = super->anchor;
2128 struct mdinfo info;
2129 char nbuf[64];
2130
a5d85af7 2131 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2132 fname_from_uuid(st, &info, nbuf, ':');
2133 printf("MD_METADATA=imsm\n");
2134 printf("MD_LEVEL=container\n");
2135 printf("MD_UUID=%s\n", nbuf+5);
2136 printf("MD_DEVICES=%u\n", mpb->num_disks);
e48aed3c 2137 printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time));
9d84c8ea
DW		 2138}
2139
b771faef
BK		 2140static void detail_super_imsm(struct supertype *st, char *homehost,
2141 char *subarray)
cdddbdbc 2142{
3ebe00a1
DW		 2143 struct mdinfo info;
2144 char nbuf[64];
b771faef
BK		 2145 struct intel_super *super = st->sb;
2146 int temp_vol = super->current_vol;
2147
2148 if (subarray)
2149 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2150
a5d85af7 2151 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2152 fname_from_uuid(st, &info, nbuf, ':');
65884368 2153 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2154
2155 super->current_vol = temp_vol;
cdddbdbc
DW		 2156}
2157
b771faef 2158static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2159{
ff54de6e
N		 2160 struct mdinfo info;
2161 char nbuf[64];
b771faef
BK		 2162 struct intel_super *super = st->sb;
2163 int temp_vol = super->current_vol;
2164
2165 if (subarray)
2166 super->current_vol = strtoul(subarray, NULL, 10);
2167
a5d85af7 2168 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2169 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2170 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2171
2172 super->current_vol = temp_vol;
cdddbdbc 2173}
d665cc31 2174
6da53c0e
BK		 2175static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2176 size_t serial_buf_len);
d665cc31
DW		 2177static void fd2devname(int fd, char *name);
2178
120dc887 2179static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2180{
120dc887
LM		 2181 /* dump an unsorted list of devices attached to AHCI Intel storage
2182 * controller, as well as non-connected ports
d665cc31
DW		 2183 */
2184 int hba_len = strlen(hba_path) + 1;
2185 struct dirent *ent;
2186 DIR *dir;
2187 char *path = NULL;
2188 int err = 0;
2189 unsigned long port_mask = (1 << port_count) - 1;
2190
f21e18ca 2191 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2192 if (verbose > 0)
e7b84f9d 2193 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2194 return 2;
2195 }
2196
2197 /* scroll through /sys/dev/block looking for devices attached to
2198 * this hba
2199 */
2200 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2201 if (!dir)
2202 return 1;
2203
2204 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2205 int fd;
2206 char model[64];
2207 char vendor[64];
2208 char buf[1024];
2209 int major, minor;
2210 char *device;
2211 char *c;
2212 int port;
2213 int type;
2214
2215 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2216 continue;
2217 path = devt_to_devpath(makedev(major, minor));
2218 if (!path)
2219 continue;
2220 if (!path_attached_to_hba(path, hba_path)) {
2221 free(path);
2222 path = NULL;
2223 continue;
2224 }
2225
2226 /* retrieve the scsi device type */
2227 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2228 if (verbose > 0)
e7b84f9d 2229 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2230 err = 2;
2231 break;
2232 }
2233 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2234 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2235 if (verbose > 0)
e7b84f9d 2236 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2237 path);
2238 err = 2;
2239 free(device);
2240 break;
2241 }
2242 type = strtoul(buf, NULL, 10);
2243
2244 /* if it's not a disk print the vendor and model */
2245 if (!(type == 0 || type == 7 || type == 14)) {
2246 vendor[0] = '\0';
2247 model[0] = '\0';
2248 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2249 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2250 strncpy(vendor, buf, sizeof(vendor));
2251 vendor[sizeof(vendor) - 1] = '\0';
2252 c = (char *) &vendor[sizeof(vendor) - 1];
2253 while (isspace(*c) || *c == '\0')
2254 *c-- = '\0';
2255
2256 }
2257 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2258 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2259 strncpy(model, buf, sizeof(model));
2260 model[sizeof(model) - 1] = '\0';
2261 c = (char *) &model[sizeof(model) - 1];
2262 while (isspace(*c) || *c == '\0')
2263 *c-- = '\0';
2264 }
2265
2266 if (vendor[0] && model[0])
2267 sprintf(buf, "%.64s %.64s", vendor, model);
2268 else
2269 switch (type) { /* numbers from hald/linux/device.c */
2270 case 1: sprintf(buf, "tape"); break;
2271 case 2: sprintf(buf, "printer"); break;
2272 case 3: sprintf(buf, "processor"); break;
2273 case 4:
2274 case 5: sprintf(buf, "cdrom"); break;
2275 case 6: sprintf(buf, "scanner"); break;
2276 case 8: sprintf(buf, "media_changer"); break;
2277 case 9: sprintf(buf, "comm"); break;
2278 case 12: sprintf(buf, "raid"); break;
2279 default: sprintf(buf, "unknown");
2280 }
2281 } else
2282 buf[0] = '\0';
2283 free(device);
2284
2285 /* chop device path to 'host%d' and calculate the port number */
2286 c = strchr(&path[hba_len], '/');
4e5e717d 2287 if (!c) {
ba728be7 2288 if (verbose > 0)
e7b84f9d 2289 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2290 err = 2;
2291 break;
2292 }
d665cc31 2293 *c = '\0';
0858eccf
AP		 2294 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2295 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2296 port -= host_base;
2297 else {
ba728be7 2298 if (verbose > 0) {
d665cc31 2299 *c = '/'; /* repair the full string */
e7b84f9d 2300 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2301 path);
2302 }
2303 err = 2;
2304 break;
2305 }
2306
2307 /* mark this port as used */
2308 port_mask &= ~(1 << port);
2309
2310 /* print out the device information */
2311 if (buf[0]) {
2312 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2313 continue;
2314 }
2315
2316 fd = dev_open(ent->d_name, O_RDONLY);
2317 if (fd < 0)
2318 printf(" Port%d : - disk info unavailable -\n", port);
2319 else {
2320 fd2devname(fd, buf);
2321 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2322 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2323 sizeof(buf)) == 0)
2324 printf(" (%s)\n", buf);
d665cc31 2325 else
664d5325 2326 printf(" ()\n");
4dab422a 2327 close(fd);
d665cc31 2328 }
d665cc31
DW		 2329 free(path);
2330 path = NULL;
2331 }
2332 if (path)
2333 free(path);
2334 if (dir)
2335 closedir(dir);
2336 if (err == 0) {
2337 int i;
2338
2339 for (i = 0; i < port_count; i++)
2340 if (port_mask & (1 << i))
2341 printf(" Port%d : - no device attached -\n", i);
2342 }
2343
2344 return err;
2345}
2346
6da53c0e 2347static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2348{
6da53c0e 2349 char buf[1024];
60f0f54d
PB		 2350 struct dirent *ent;
2351 DIR *dir;
6da53c0e
BK		 2352 char *rp;
2353 int fd;
60f0f54d 2354
6da53c0e 2355 dir = opendir("/sys/block/");
b9135011 2356 if (!dir)
b5eece69 2357 return 1;
b9135011
JS		 2358
2359 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e
BK		 2360 if (strstr(ent->d_name, "nvme")) {
2361 sprintf(buf, "/sys/block/%s", ent->d_name);
2362 rp = realpath(buf, NULL);
2363 if (!rp)
2364 continue;
2365 if (path_attached_to_hba(rp, hba->path)) {
2366 fd = open_dev(ent->d_name);
2367 if (fd < 0) {
2368 free(rp);
2369 continue;
2370 }
60f0f54d 2371
6da53c0e
BK		 2372 fd2devname(fd, buf);
2373 if (hba->type == SYS_DEV_VMD)
2374 printf(" NVMe under VMD : %s", buf);
2375 else if (hba->type == SYS_DEV_NVME)
2376 printf(" NVMe Device : %s", buf);
2377 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2378 sizeof(buf)))
2379 printf(" (%s)\n", buf);
2380 else
2381 printf("()\n");
2382 close(fd);
2383 }
2384 free(rp);
60f0f54d 2385 }
60f0f54d
PB		 2386 }
2387
b9135011 2388 closedir(dir);
b5eece69 2389 return 0;
60f0f54d
PB		 2390}
2391
120dc887
LM		 2392static void print_found_intel_controllers(struct sys_dev *elem)
2393{
2394 for (; elem; elem = elem->next) {
e7b84f9d 2395 pr_err("found Intel(R) ");
120dc887
LM		 2396 if (elem->type == SYS_DEV_SATA)
2397 fprintf(stderr, "SATA ");
155cbb4c
LM		 2398 else if (elem->type == SYS_DEV_SAS)
2399 fprintf(stderr, "SAS ");
0858eccf
AP		 2400 else if (elem->type == SYS_DEV_NVME)
2401 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2402
2403 if (elem->type == SYS_DEV_VMD)
2404 fprintf(stderr, "VMD domain");
2405 else
2406 fprintf(stderr, "RAID controller");
2407
120dc887
LM		 2408 if (elem->pci_id)
2409 fprintf(stderr, " at %s", elem->pci_id);
2410 fprintf(stderr, ".\n");
2411 }
2412 fflush(stderr);
2413}
2414
120dc887
LM		 2415static int ahci_get_port_count(const char *hba_path, int *port_count)
2416{
2417 struct dirent *ent;
2418 DIR *dir;
2419 int host_base = -1;
2420
2421 *port_count = 0;
2422 if ((dir = opendir(hba_path)) == NULL)
2423 return -1;
2424
2425 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2426 int host;
2427
0858eccf
AP		 2428 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2429 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2430 continue;
2431 if (*port_count == 0)
2432 host_base = host;
2433 else if (host < host_base)
2434 host_base = host;
2435
2436 if (host + 1 > *port_count + host_base)
2437 *port_count = host + 1 - host_base;
2438 }
2439 closedir(dir);
2440 return host_base;
2441}
2442
a891a3c2
LM		 2443static void print_imsm_capability(const struct imsm_orom *orom)
2444{
0858eccf
AP		 2445 printf(" Platform : Intel(R) ");
2446 if (orom->capabilities == 0 && orom->driver_features == 0)
2447 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2448 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2449 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2450 else
2451 printf("Rapid Storage Technology%s\n",
2452 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2453 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2454 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2455 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2456 printf(" RAID Levels :%s%s%s%s%s\n",
2457 imsm_orom_has_raid0(orom) ? " raid0" : "",
2458 imsm_orom_has_raid1(orom) ? " raid1" : "",
2459 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2460 imsm_orom_has_raid10(orom) ? " raid10" : "",
2461 imsm_orom_has_raid5(orom) ? " raid5" : "");
2462 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2463 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2464 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2465 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2466 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2467 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2468 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2469 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2470 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2471 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2472 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2473 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2474 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2475 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2476 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2477 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2478 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2479 printf(" 2TB volumes :%s supported\n",
2480 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2481 printf(" 2TB disks :%s supported\n",
2482 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2483 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2484 printf(" Max Volumes : %d per array, %d per %s\n",
2485 orom->vpa, orom->vphba,
2486 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2487 return;
2488}
2489
e50cf220
MN		 2490static void print_imsm_capability_export(const struct imsm_orom *orom)
2491{
2492 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2493 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2494 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2495 orom->hotfix_ver, orom->build);
e50cf220
MN		 2496 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2497 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2498 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2499 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2500 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2501 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2502 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2503 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2504 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2505 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2506 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2507 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2508 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2509 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2510 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2511 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2512 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2513 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2514 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2515 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2516 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2517 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2518 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2519 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2520 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2521 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2522 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2523 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2524}
2525
9eafa1de 2526static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2527{
2528 /* There are two components to imsm platform support, the ahci SATA
2529 * controller and the option-rom. To find the SATA controller we
2530 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2531 * controller with the Intel vendor id is present. This approach
2532 * allows mdadm to leverage the kernel's ahci detection logic, with the
2533 * caveat that if ahci.ko is not loaded mdadm will not be able to
2534 * detect platform raid capabilities. The option-rom resides in a
2535 * platform "Adapter ROM". We scan for its signature to retrieve the
2536 * platform capabilities. If raid support is disabled in the BIOS the
2537 * option-rom capability structure will not be available.
2538 */
d665cc31 2539 struct sys_dev *list, *hba;
d665cc31
DW		 2540 int host_base = 0;
2541 int port_count = 0;
9eafa1de 2542 int result=1;
d665cc31 2543
5615172f 2544 if (enumerate_only) {
a891a3c2 2545 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2546 return 0;
a891a3c2
LM		 2547 list = find_intel_devices();
2548 if (!list)
2549 return 2;
2550 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2551 if (find_imsm_capability(hba)) {
2552 result = 0;
a891a3c2
LM		 2553 break;
2554 }
9eafa1de 2555 else
6b781d33 2556 result = 2;
a891a3c2 2557 }
a891a3c2 2558 return result;
5615172f
DW		 2559 }
2560
155cbb4c
LM		 2561 list = find_intel_devices();
2562 if (!list) {
ba728be7 2563 if (verbose > 0)
7a862a02 2564 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2565 return 2;
ba728be7 2566 } else if (verbose > 0)
155cbb4c 2567 print_found_intel_controllers(list);
d665cc31 2568
a891a3c2 2569 for (hba = list; hba; hba = hba->next) {
0858eccf 2570 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2571 continue;
0858eccf 2572 if (!find_imsm_capability(hba)) {
60f0f54d 2573 char buf[PATH_MAX];
e7b84f9d 2574 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2575 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2576 get_sys_dev_type(hba->type));
0858eccf
AP		 2577 continue;
2578 }
2579 result = 0;
2580 }
2581
2582 if (controller_path && result == 1) {
2583 pr_err("no active Intel(R) RAID controller found under %s\n",
2584 controller_path);
2585 return result;
2586 }
2587
5e1d6128 2588 const struct orom_entry *entry;
0858eccf 2589
5e1d6128 2590 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2591 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2592 print_imsm_capability(&entry->orom);
32716c51
PB		 2593 printf(" 3rd party NVMe :%s supported\n",
2594 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2595 for (hba = list; hba; hba = hba->next) {
2596 if (hba->type == SYS_DEV_VMD) {
2597 char buf[PATH_MAX];
60f0f54d
PB		 2598 printf(" I/O Controller : %s (%s)\n",
2599 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2600 if (print_nvme_info(hba)) {
b5eece69
PB		 2601 if (verbose > 0)
2602 pr_err("failed to get devices attached to VMD domain.\n");
2603 result |= 2;
2604 }
60f0f54d
PB		 2605 }
2606 }
07cb1e57 2607 printf("\n");
60f0f54d
PB		 2608 continue;
2609 }
0858eccf 2610
60f0f54d
PB		 2611 print_imsm_capability(&entry->orom);
2612 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2613 for (hba = list; hba; hba = hba->next) {
2614 if (hba->type == SYS_DEV_NVME)
6da53c0e 2615 print_nvme_info(hba);
0858eccf 2616 }
60f0f54d 2617 printf("\n");
0858eccf
AP		 2618 continue;
2619 }
2620
2621 struct devid_list *devid;
5e1d6128 2622 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2623 hba = device_by_id(devid->devid);
2624 if (!hba)
2625 continue;
2626
9eafa1de
MN		 2627 printf(" I/O Controller : %s (%s)\n",
2628 hba->path, get_sys_dev_type(hba->type));
2629 if (hba->type == SYS_DEV_SATA) {
2630 host_base = ahci_get_port_count(hba->path, &port_count);
2631 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2632 if (verbose > 0)
7a862a02 2633 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2634 result |= 2;
2635 }
120dc887
LM		 2636 }
2637 }
0858eccf 2638 printf("\n");
d665cc31 2639 }
155cbb4c 2640
120dc887 2641 return result;
d665cc31 2642}
e50cf220 2643
9eafa1de 2644static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2645{
e50cf220
MN		 2646 struct sys_dev *list, *hba;
2647 int result=1;
2648
2649 list = find_intel_devices();
2650 if (!list) {
2651 if (verbose > 0)
2652 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2653 result = 2;
e50cf220
MN		 2654 return result;
2655 }
2656
2657 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2658 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2659 continue;
60f0f54d
PB		 2660 if (!find_imsm_capability(hba) && verbose > 0) {
2661 char buf[PATH_MAX];
2662 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2663 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2664 }
0858eccf 2665 else
e50cf220 2666 result = 0;
e50cf220
MN		 2667 }
2668
5e1d6128 2669 const struct orom_entry *entry;
0858eccf 2670
60f0f54d
PB		 2671 for (entry = orom_entries; entry; entry = entry->next) {
2672 if (entry->type == SYS_DEV_VMD) {
2673 for (hba = list; hba; hba = hba->next)
2674 print_imsm_capability_export(&entry->orom);
2675 continue;
2676 }
5e1d6128 2677 print_imsm_capability_export(&entry->orom);
60f0f54d 2678 }
0858eccf 2679
e50cf220
MN		 2680 return result;
2681}
2682
cdddbdbc
DW		 2683static int match_home_imsm(struct supertype *st, char *homehost)
2684{
5115ca67
DW		 2685 /* the imsm metadata format does not specify any host
2686 * identification information. We return -1 since we can never
2687 * confirm nor deny whether a given array is "meant" for this
148acb7b 2688 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2689 * exclude member disks that do not belong, and we rely on
2690 * mdadm.conf to specify the arrays that should be assembled.
2691 * Auto-assembly may still pick up "foreign" arrays.
2692 */
cdddbdbc 2693
9362c1c8 2694 return -1;
cdddbdbc
DW		 2695}
2696
2697static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2698{
51006d85
N		 2699 /* The uuid returned here is used for:
2700 * uuid to put into bitmap file (Create, Grow)
2701 * uuid for backup header when saving critical section (Grow)
2702 * comparing uuids when re-adding a device into an array
2703 * In these cases the uuid required is that of the data-array,
2704 * not the device-set.
2705 * uuid to recognise same set when adding a missing device back
2706 * to an array. This is a uuid for the device-set.
1011e834 2707 *
51006d85
N		 2708 * For each of these we can make do with a truncated
2709 * or hashed uuid rather than the original, as long as
2710 * everyone agrees.
2711 * In each case the uuid required is that of the data-array,
2712 * not the device-set.
43dad3d6 2713 */
51006d85
N		 2714 /* imsm does not track uuid's so we synthesis one using sha1 on
2715 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2716 * - the orig_family_num of the container
51006d85
N		 2717 * - the index number of the volume
2718 * - the 'serial' number of the volume.
2719 * Hopefully these are all constant.
2720 */
2721 struct intel_super *super = st->sb;
43dad3d6 2722
51006d85
N		 2723 char buf[20];
2724 struct sha1_ctx ctx;
2725 struct imsm_dev *dev = NULL;
148acb7b 2726 __u32 family_num;
51006d85 2727
148acb7b
DW		 2728 /* some mdadm versions failed to set ->orig_family_num, in which
2729 * case fall back to ->family_num. orig_family_num will be
2730 * fixed up with the first metadata update.
2731 */
2732 family_num = super->anchor->orig_family_num;
2733 if (family_num == 0)
2734 family_num = super->anchor->family_num;
51006d85 2735 sha1_init_ctx(&ctx);
92bd8f8d 2736 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2737 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2738 if (super->current_vol >= 0)
2739 dev = get_imsm_dev(super, super->current_vol);
2740 if (dev) {
2741 __u32 vol = super->current_vol;
2742 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2743 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2744 }
2745 sha1_finish_ctx(&ctx, buf);
2746 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2747}
2748
0d481d37 2749#if 0
4f5bc454
DW		 2750static void
2751get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2752{
cdddbdbc
DW		 2753 __u8 *v = get_imsm_version(mpb);
2754 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2755 char major[] = { 0, 0, 0 };
2756 char minor[] = { 0 ,0, 0 };
2757 char patch[] = { 0, 0, 0 };
2758 char *ver_parse[] = { major, minor, patch };
2759 int i, j;
2760
2761 i = j = 0;
2762 while (*v != '\0' && v < end) {
2763 if (*v != '.' && j < 2)
2764 ver_parse[i][j++] = *v;
2765 else {
2766 i++;
2767 j = 0;
2768 }
2769 v++;
2770 }
2771
4f5bc454
DW		 2772 *m = strtol(minor, NULL, 0);
2773 *p = strtol(patch, NULL, 0);
2774}
0d481d37 2775#endif
4f5bc454 2776
1e5c6983
DW		 2777static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2778{
2779 /* migr_strip_size when repairing or initializing parity */
238c0a71 2780 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2781 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2782
2783 switch (get_imsm_raid_level(map)) {
2784 case 5:
2785 case 10:
2786 return chunk;
2787 default:
2788 return 128*1024 >> 9;
2789 }
2790}
2791
2792static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2793{
2794 /* migr_strip_size when rebuilding a degraded disk, no idea why
2795 * this is different than migr_strip_size_resync(), but it's good
2796 * to be compatible
2797 */
238c0a71 2798 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2799 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2800
2801 switch (get_imsm_raid_level(map)) {
2802 case 1:
2803 case 10:
2804 if (map->num_members % map->num_domains == 0)
2805 return 128*1024 >> 9;
2806 else
2807 return chunk;
2808 case 5:
2809 return max((__u32) 64*1024 >> 9, chunk);
2810 default:
2811 return 128*1024 >> 9;
2812 }
2813}
2814
2815static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2816{
238c0a71
AK		 2817 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2818 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2819 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2820 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2821
2822 return max((__u32) 1, hi_chunk / lo_chunk);
2823}
2824
2825static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2826{
238c0a71 2827 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2828 int level = get_imsm_raid_level(lo);
2829
2830 if (level == 1 || level == 10) {
238c0a71 2831 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2832
2833 return hi->num_domains;
2834 } else
2835 return num_stripes_per_unit_resync(dev);
2836}
2837
9529d343 2838static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2839{
2840 /* named 'imsm_' because raid0, raid1 and raid10
2841 * counter-intuitively have the same number of data disks
2842 */
1e5c6983
DW		 2843 switch (get_imsm_raid_level(map)) {
2844 case 0:
36fd8ccc
AK		 2845 return map->num_members;
2846 break;
1e5c6983
DW		 2847 case 1:
2848 case 10:
36fd8ccc 2849 return map->num_members/2;
1e5c6983
DW		 2850 case 5:
2851 return map->num_members - 1;
2852 default:
1ade5cc1 2853 dprintf("unsupported raid level\n");
1e5c6983
DW		 2854 return 0;
2855 }
2856}
2857
44490938
MD		 2858static unsigned long long calc_component_size(struct imsm_map *map,
2859 struct imsm_dev *dev)
2860{
2861 unsigned long long component_size;
2862 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2863 long long calc_dev_size = 0;
44490938
MD		 2864 unsigned int member_disks = imsm_num_data_members(map);
2865
2866 if (member_disks == 0)
2867 return 0;
2868
2869 component_size = per_dev_array_size(map);
2870 calc_dev_size = component_size * member_disks;
2871
2872 /* Component size is rounded to 1MB so difference between size from
2873 * metadata and size calculated from num_data_stripes equals up to
2874 * 2048 blocks per each device. If the difference is higher it means
2875 * that array size was expanded and num_data_stripes was not updated.
2876 */
a4f7290c 2877 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2878 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2879 component_size = dev_size / member_disks;
2880 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2881 component_size / map->blocks_per_strip,
2882 num_data_stripes(map));
2883 }
2884
2885 return component_size;
2886}
2887
1e5c6983
DW		 2888static __u32 parity_segment_depth(struct imsm_dev *dev)
2889{
238c0a71 2890 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2891 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2892
2893 switch(get_imsm_raid_level(map)) {
2894 case 1:
2895 case 10:
2896 return chunk * map->num_domains;
2897 case 5:
2898 return chunk * map->num_members;
2899 default:
2900 return chunk;
2901 }
2902}
2903
2904static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2905{
238c0a71 2906 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2907 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2908 __u32 strip = block / chunk;
2909
2910 switch (get_imsm_raid_level(map)) {
2911 case 1:
2912 case 10: {
2913 __u32 vol_strip = (strip * map->num_domains) + 1;
2914 __u32 vol_stripe = vol_strip / map->num_members;
2915
2916 return vol_stripe * chunk + block % chunk;
2917 } case 5: {
2918 __u32 stripe = strip / (map->num_members - 1);
2919
2920 return stripe * chunk + block % chunk;
2921 }
2922 default:
2923 return 0;
2924 }
2925}
2926
c47b0ff6
AK		 2927static __u64 blocks_per_migr_unit(struct intel_super *super,
2928 struct imsm_dev *dev)
1e5c6983
DW		 2929{
2930 /* calculate the conversion factor between per member 'blocks'
2931 * (md/{resync,rebuild}_start) and imsm migration units, return
2932 * 0 for the 'not migrating' and 'unsupported migration' cases
2933 */
2934 if (!dev->vol.migr_state)
2935 return 0;
2936
2937 switch (migr_type(dev)) {
c47b0ff6
AK		 2938 case MIGR_GEN_MIGR: {
2939 struct migr_record *migr_rec = super->migr_rec;
2940 return __le32_to_cpu(migr_rec->blocks_per_unit);
2941 }
1e5c6983
DW		 2942 case MIGR_VERIFY:
2943 case MIGR_REPAIR:
2944 case MIGR_INIT: {
238c0a71 2945 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2946 __u32 stripes_per_unit;
2947 __u32 blocks_per_unit;
2948 __u32 parity_depth;
2949 __u32 migr_chunk;
2950 __u32 block_map;
2951 __u32 block_rel;
2952 __u32 segment;
2953 __u32 stripe;
2954 __u8 disks;
2955
2956 /* yes, this is really the translation of migr_units to
2957 * per-member blocks in the 'resync' case
2958 */
2959 stripes_per_unit = num_stripes_per_unit_resync(dev);
2960 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2961 disks = imsm_num_data_members(map);
1e5c6983 2962 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2963 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2964 segment = blocks_per_unit / stripe;
2965 block_rel = blocks_per_unit - segment * stripe;
2966 parity_depth = parity_segment_depth(dev);
2967 block_map = map_migr_block(dev, block_rel);
2968 return block_map + parity_depth * segment;
2969 }
2970 case MIGR_REBUILD: {
2971 __u32 stripes_per_unit;
2972 __u32 migr_chunk;
2973
2974 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2975 migr_chunk = migr_strip_blocks_rebuild(dev);
2976 return migr_chunk * stripes_per_unit;
2977 }
1e5c6983
DW		 2978 case MIGR_STATE_CHANGE:
2979 default:
2980 return 0;
2981 }
2982}
2983
c2c087e6
DW		 2984static int imsm_level_to_layout(int level)
2985{
2986 switch (level) {
2987 case 0:
2988 case 1:
2989 return 0;
2990 case 5:
2991 case 6:
a380c027 2992 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2993 case 10:
c92a2527 2994 return 0x102;
c2c087e6 2995 }
a18a888e 2996 return UnSet;
c2c087e6
DW		 2997}
2998
8e59f3d8
AK		 2999/*******************************************************************************
3000 * Function: read_imsm_migr_rec
3001 * Description: Function reads imsm migration record from last sector of disk
3002 * Parameters:
3003 * fd : disk descriptor
3004 * super : metadata info
3005 * Returns:
3006 * 0 : success,
3007 * -1 : fail
3008 ******************************************************************************/
3009static int read_imsm_migr_rec(int fd, struct intel_super *super)
3010{
3011 int ret_val = -1;
de44e46f 3012 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3013 unsigned long long dsize;
3014
3015 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3016 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3017 SEEK_SET) < 0) {
e7b84f9d
N		 3018 pr_err("Cannot seek to anchor block: %s\n",
3019 strerror(errno));
8e59f3d8
AK		 3020 goto out;
3021 }
466070ad 3022 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3023 MIGR_REC_BUF_SECTORS*sector_size) !=
3024 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3025 pr_err("Cannot read migr record block: %s\n",
3026 strerror(errno));
8e59f3d8
AK		 3027 goto out;
3028 }
3029 ret_val = 0;
de44e46f
PB		 3030 if (sector_size == 4096)
3031 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3032
3033out:
3034 return ret_val;
3035}
3036
3136abe5
AK		 3037static struct imsm_dev *imsm_get_device_during_migration(
3038 struct intel_super *super)
3039{
3040
3041 struct intel_dev *dv;
3042
3043 for (dv = super->devlist; dv; dv = dv->next) {
3044 if (is_gen_migration(dv->dev))
3045 return dv->dev;
3046 }
3047 return NULL;
3048}
3049
8e59f3d8
AK		 3050/*******************************************************************************
3051 * Function: load_imsm_migr_rec
3052 * Description: Function reads imsm migration record (it is stored at the last
3053 * sector of disk)
3054 * Parameters:
3055 * super : imsm internal array info
3056 * info : general array info
3057 * Returns:
3058 * 0 : success
3059 * -1 : fail
4c965cc9 3060 * -2 : no migration in progress
8e59f3d8
AK		 3061 ******************************************************************************/
3062static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3063{
3064 struct mdinfo *sd;
594dc1b8 3065 struct dl *dl;
8e59f3d8
AK		 3066 char nm[30];
3067 int retval = -1;
3068 int fd = -1;
3136abe5 3069 struct imsm_dev *dev;
594dc1b8 3070 struct imsm_map *map;
b4ab44d8 3071 int slot = -1;
3136abe5
AK		 3072
3073 /* find map under migration */
3074 dev = imsm_get_device_during_migration(super);
3075 /* nothing to load,no migration in progress?
3076 */
3077 if (dev == NULL)
4c965cc9 3078 return -2;
8e59f3d8
AK		 3079
3080 if (info) {
3081 for (sd = info->devs ; sd ; sd = sd->next) {
3082 /* read only from one of the first two slots */
12fe93e9
TM		 3083 if ((sd->disk.raid_disk < 0) ||
3084 (sd->disk.raid_disk > 1))
8e59f3d8 3085 continue;
3136abe5 3086
8e59f3d8
AK		 3087 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3088 fd = dev_open(nm, O_RDONLY);
3089 if (fd >= 0)
3090 break;
3091 }
3092 }
3093 if (fd < 0) {
12fe93e9 3094 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3095 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3096 /* skip spare and failed disks
3097 */
3098 if (dl->index < 0)
3099 continue;
8e59f3d8 3100 /* read only from one of the first two slots */
3136abe5
AK		 3101 if (map)
3102 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3103 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3104 continue;
3105 sprintf(nm, "%d:%d", dl->major, dl->minor);
3106 fd = dev_open(nm, O_RDONLY);
3107 if (fd >= 0)
3108 break;
3109 }
3110 }
3111 if (fd < 0)
3112 goto out;
3113 retval = read_imsm_migr_rec(fd, super);
3114
3115out:
3116 if (fd >= 0)
3117 close(fd);
3118 return retval;
3119}
3120
c17608ea
AK		 3121/*******************************************************************************
3122 * function: imsm_create_metadata_checkpoint_update
3123 * Description: It creates update for checkpoint change.
3124 * Parameters:
3125 * super : imsm internal array info
3126 * u : pointer to prepared update
3127 * Returns:
3128 * Uptate length.
3129 * If length is equal to 0, input pointer u contains no update
3130 ******************************************************************************/
3131static int imsm_create_metadata_checkpoint_update(
3132 struct intel_super *super,
3133 struct imsm_update_general_migration_checkpoint **u)
3134{
3135
3136 int update_memory_size = 0;
3137
1ade5cc1 3138 dprintf("(enter)\n");
c17608ea
AK		 3139
3140 if (u == NULL)
3141 return 0;
3142 *u = NULL;
3143
3144 /* size of all update data without anchor */
3145 update_memory_size =
3146 sizeof(struct imsm_update_general_migration_checkpoint);
3147
503975b9 3148 *u = xcalloc(1, update_memory_size);
c17608ea 3149 if (*u == NULL) {
1ade5cc1 3150 dprintf("error: cannot get memory\n");
c17608ea
AK		 3151 return 0;
3152 }
3153 (*u)->type = update_general_migration_checkpoint;
9f421827 3154 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3155 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3156
3157 return update_memory_size;
3158}
3159
c17608ea
AK		 3160static void imsm_update_metadata_locally(struct supertype *st,
3161 void *buf, int len);
3162
687629c2
AK		 3163/*******************************************************************************
3164 * Function: write_imsm_migr_rec
3165 * Description: Function writes imsm migration record
3166 * (at the last sector of disk)
3167 * Parameters:
3168 * super : imsm internal array info
3169 * Returns:
3170 * 0 : success
3171 * -1 : if fail
3172 ******************************************************************************/
3173static int write_imsm_migr_rec(struct supertype *st)
3174{
3175 struct intel_super *super = st->sb;
de44e46f 3176 unsigned int sector_size = super->sector_size;
687629c2
AK		 3177 unsigned long long dsize;
3178 char nm[30];
3179 int fd = -1;
3180 int retval = -1;
3181 struct dl *sd;
c17608ea
AK		 3182 int len;
3183 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3184 struct imsm_dev *dev;
594dc1b8 3185 struct imsm_map *map;
3136abe5
AK		 3186
3187 /* find map under migration */
3188 dev = imsm_get_device_during_migration(super);
3189 /* if no migration, write buffer anyway to clear migr_record
3190 * on disk based on first available device
3191 */
3192 if (dev == NULL)
3193 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3194 super->current_vol);
3195
44bfe6df 3196 map = get_imsm_map(dev, MAP_0);
687629c2 3197
de44e46f
PB		 3198 if (sector_size == 4096)
3199 convert_to_4k_imsm_migr_rec(super);
687629c2 3200 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3201 int slot = -1;
3136abe5
AK		 3202
3203 /* skip failed and spare devices */
3204 if (sd->index < 0)
3205 continue;
687629c2 3206 /* write to 2 first slots only */
3136abe5
AK		 3207 if (map)
3208 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3209 if (map == NULL || slot > 1 || slot < 0)
687629c2 3210 continue;
3136abe5 3211
687629c2
AK		 3212 sprintf(nm, "%d:%d", sd->major, sd->minor);
3213 fd = dev_open(nm, O_RDWR);
3214 if (fd < 0)
3215 continue;
3216 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3217 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3218 SEEK_SET) < 0) {
e7b84f9d
N		 3219 pr_err("Cannot seek to anchor block: %s\n",
3220 strerror(errno));
687629c2
AK		 3221 goto out;
3222 }
466070ad 3223 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3224 MIGR_REC_BUF_SECTORS*sector_size) !=
3225 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3226 pr_err("Cannot write migr record block: %s\n",
3227 strerror(errno));
687629c2
AK		 3228 goto out;
3229 }
3230 close(fd);
3231 fd = -1;
3232 }
de44e46f
PB		 3233 if (sector_size == 4096)
3234 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3235 /* update checkpoint information in metadata */
3236 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3237 if (len <= 0) {
3238 dprintf("imsm: Cannot prepare update\n");
3239 goto out;
3240 }
3241 /* update metadata locally */
3242 imsm_update_metadata_locally(st, u, len);
3243 /* and possibly remotely */
3244 if (st->update_tail) {
3245 append_metadata_update(st, u, len);
3246 /* during reshape we do all work inside metadata handler
3247 * manage_reshape(), so metadata update has to be triggered
3248 * insida it
3249 */
3250 flush_metadata_updates(st);
3251 st->update_tail = &st->updates;
3252 } else
3253 free(u);
687629c2
AK		 3254
3255 retval = 0;
3256 out:
3257 if (fd >= 0)
3258 close(fd);
3259 return retval;
3260}
3261
e2962bfc
AK		 3262/* spare/missing disks activations are not allowe when
3263 * array/container performs reshape operation, because
3264 * all arrays in container works on the same disks set
3265 */
3266int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3267{
3268 int rv = 0;
3269 struct intel_dev *i_dev;
3270 struct imsm_dev *dev;
3271
3272 /* check whole container
3273 */
3274 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3275 dev = i_dev->dev;
3ad25638 3276 if (is_gen_migration(dev)) {
e2962bfc
AK		 3277 /* No repair during any migration in container
3278 */
3279 rv = 1;
3280 break;
3281 }
3282 }
3283 return rv;
3284}
3e684231 3285static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3286 int chunk_size,
f36a9ecd 3287 unsigned int sector_size,
c41e00b2
AK		 3288 unsigned long long component_size)
3289{
3e684231 3290 unsigned int component_size_alignment;
c41e00b2 3291
3e684231 3292 /* check component size alignment
c41e00b2 3293 */
3e684231 3294 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3295
3e684231 3296 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3297 level, chunk_size, component_size,
3e684231 3298 component_size_alignment);
c41e00b2 3299
3e684231
MZ		 3300 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3301 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3302 component_size);
3e684231 3303 component_size -= component_size_alignment;
1ade5cc1 3304 dprintf_cont("to %llu (%i).\n",
3e684231 3305 component_size, component_size_alignment);
c41e00b2
AK		 3306 }
3307
3308 return component_size;
3309}
e2962bfc 3310
2432ce9b
AP		 3311static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3312{
3313 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3314 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3315
3316 return pba_of_lba0(map) +
3317 (num_data_stripes(map) * map->blocks_per_strip);
3318}
3319
a5d85af7 3320static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3321{
3322 struct intel_super *super = st->sb;
c47b0ff6 3323 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3324 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3325 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3326 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3327 struct imsm_map *map_to_analyse = map;
efb30e7f 3328 struct dl *dl;
a5d85af7 3329 int map_disks = info->array.raid_disks;
bf5a934a 3330
95eeceeb 3331 memset(info, 0, sizeof(*info));
b335e593
AK		 3332 if (prev_map)
3333 map_to_analyse = prev_map;
3334
ca0748fa 3335 dl = super->current_disk;
9894ec0d 3336
bf5a934a 3337 info->container_member = super->current_vol;
cd0430a1 3338 info->array.raid_disks = map->num_members;
b335e593 3339 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3340 info->array.layout = imsm_level_to_layout(info->array.level);
3341 info->array.md_minor = -1;
3342 info->array.ctime = 0;
3343 info->array.utime = 0;
b335e593
AK		 3344 info->array.chunk_size =
3345 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3346 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3347 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3348 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3349
3f510843 3350 if (is_gen_migration(dev)) {
3f83228a 3351 info->reshape_active = 1;
b335e593
AK		 3352 info->new_level = get_imsm_raid_level(map);
3353 info->new_layout = imsm_level_to_layout(info->new_level);
3354 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3355 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3356 if (info->delta_disks) {
3357 /* this needs to be applied to every array
3358 * in the container.
3359 */
81219e70 3360 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3361 }
3f83228a
N		 3362 /* We shape information that we give to md might have to be
3363 * modify to cope with md's requirement for reshaping arrays.
3364 * For example, when reshaping a RAID0, md requires it to be
3365 * presented as a degraded RAID4.
3366 * Also if a RAID0 is migrating to a RAID5 we need to specify
3367 * the array as already being RAID5, but the 'before' layout
3368 * is a RAID4-like layout.
3369 */
3370 switch (info->array.level) {
3371 case 0:
3372 switch(info->new_level) {
3373 case 0:
3374 /* conversion is happening as RAID4 */
3375 info->array.level = 4;
3376 info->array.raid_disks += 1;
3377 break;
3378 case 5:
3379 /* conversion is happening as RAID5 */
3380 info->array.level = 5;
3381 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3382 info->delta_disks -= 1;
3383 break;
3384 default:
3385 /* FIXME error message */
3386 info->array.level = UnSet;
3387 break;
3388 }
3389 break;
3390 }
b335e593
AK		 3391 } else {
3392 info->new_level = UnSet;
3393 info->new_layout = UnSet;
3394 info->new_chunk = info->array.chunk_size;
3f83228a 3395 info->delta_disks = 0;
b335e593 3396 }
ca0748fa 3397
efb30e7f
DW		 3398 if (dl) {
3399 info->disk.major = dl->major;
3400 info->disk.minor = dl->minor;
ca0748fa 3401 info->disk.number = dl->index;
656b6b5a
N		 3402 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3403 dl->index);
efb30e7f 3404 }
bf5a934a 3405
5551b113 3406 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3407 info->component_size = calc_component_size(map, dev);
3e684231 3408 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3409 info->array.level,
3410 info->array.chunk_size,
f36a9ecd 3411 super->sector_size,
c41e00b2 3412 info->component_size);
5e46202e 3413 info->bb.supported = 1;
139dae11 3414
301406c9 3415 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3416 info->recovery_start = MaxSector;
bf5a934a 3417
c2462068
PB		 3418 if (info->array.level == 5 &&
3419 (dev->rwh_policy == RWH_DISTRIBUTED ||
3420 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3421 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3422 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3423 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3424 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3425 else
3426 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3427 >> 9;
2432ce9b
AP		 3428 } else if (info->array.level <= 0) {
3429 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3430 } else {
3431 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3432 }
3433
d2e6d5d6 3434 info->reshape_progress = 0;
b6796ce1 3435 info->resync_start = MaxSector;
b9172665 3436 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3437 !(info->array.state & 1)) &&
b9172665 3438 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3439 info->resync_start = 0;
b6796ce1
AK		 3440 }
3441 if (dev->vol.migr_state) {
1e5c6983
DW		 3442 switch (migr_type(dev)) {
3443 case MIGR_REPAIR:
3444 case MIGR_INIT: {
c47b0ff6
AK		 3445 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3446 dev);
1e5c6983
DW		 3447 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3448
3449 info->resync_start = blocks_per_unit * units;
3450 break;
3451 }
d2e6d5d6 3452 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3453 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3454 dev);
9f421827 3455 __u64 units = current_migr_unit(migr_rec);
04fa9523
AK		 3456 unsigned long long array_blocks;
3457 int used_disks;
d2e6d5d6 3458
befb629b
AK		 3459 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3460 (units <
9f421827 3461 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3462 (super->migr_rec->rec_status ==
3463 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3464 units++;
3465
d2e6d5d6 3466 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3467
7a862a02 3468 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3469 (unsigned long long)units,
3470 (unsigned long long)blocks_per_unit,
3471 info->reshape_progress);
75156c46 3472
9529d343 3473 used_disks = imsm_num_data_members(prev_map);
75156c46 3474 if (used_disks > 0) {
44490938 3475 array_blocks = per_dev_array_size(map) *
75156c46 3476 used_disks;
b53bfba6
TM		 3477 info->custom_array_size =
3478 round_size_to_mb(array_blocks,
3479 used_disks);
3480
75156c46 3481 }
d2e6d5d6 3482 }
1e5c6983
DW		 3483 case MIGR_VERIFY:
3484 /* we could emulate the checkpointing of
3485 * 'sync_action=check' migrations, but for now
3486 * we just immediately complete them
3487 */
3488 case MIGR_REBUILD:
3489 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3490 case MIGR_STATE_CHANGE:
3491 /* FIXME handle other migrations */
3492 default:
3493 /* we are not dirty, so... */
3494 info->resync_start = MaxSector;
3495 }
b6796ce1 3496 }
301406c9
DW		 3497
3498 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3499 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3500
f35f2525
N		 3501 info->array.major_version = -1;
3502 info->array.minor_version = -2;
4dd2df09 3503 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3504 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3505 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3506
3507 if (dmap) {
3508 int i, j;
3509 for (i=0; i<map_disks; i++) {
3510 dmap[i] = 0;
3511 if (i < info->array.raid_disks) {
3512 struct imsm_disk *dsk;
238c0a71 3513 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3514 dsk = get_imsm_disk(super, j);
3515 if (dsk && (dsk->status & CONFIGURED_DISK))
3516 dmap[i] = 1;
3517 }
3518 }
3519 }
81ac8b4d 3520}
bf5a934a 3521
3b451610
AK		 3522static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3523 int failed, int look_in_map);
3524
3525static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3526 int look_in_map);
3527
3528static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3529{
3530 if (is_gen_migration(dev)) {
3531 int failed;
3532 __u8 map_state;
3533 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3534
3535 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3536 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3537 if (map2->map_state != map_state) {
3538 map2->map_state = map_state;
3539 super->updates_pending++;
3540 }
3541 }
3542}
97b4d0e9
DW		 3543
3544static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3545{
3546 struct dl *d;
3547
3548 for (d = super->missing; d; d = d->next)
3549 if (d->index == index)
3550 return &d->disk;
3551 return NULL;
3552}
3553
a5d85af7 3554static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3555{
3556 struct intel_super *super = st->sb;
4f5bc454 3557 struct imsm_disk *disk;
a5d85af7 3558 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3559 int max_enough = -1;
3560 int i;
3561 struct imsm_super *mpb;
4f5bc454 3562
bf5a934a 3563 if (super->current_vol >= 0) {
a5d85af7 3564 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3565 return;
3566 }
95eeceeb 3567 memset(info, 0, sizeof(*info));
d23fe947
DW		 3568
3569 /* Set raid_disks to zero so that Assemble will always pull in valid
3570 * spares
3571 */
3572 info->array.raid_disks = 0;
cdddbdbc
DW		 3573 info->array.level = LEVEL_CONTAINER;
3574 info->array.layout = 0;
3575 info->array.md_minor = -1;
1011e834 3576 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3577 info->array.utime = 0;
3578 info->array.chunk_size = 0;
3579
3580 info->disk.major = 0;
3581 info->disk.minor = 0;
cdddbdbc 3582 info->disk.raid_disk = -1;
c2c087e6 3583 info->reshape_active = 0;
f35f2525
N		 3584 info->array.major_version = -1;
3585 info->array.minor_version = -2;
c2c087e6 3586 strcpy(info->text_version, "imsm");
a67dd8cc 3587 info->safe_mode_delay = 0;
c2c087e6
DW		 3588 info->disk.number = -1;
3589 info->disk.state = 0;
c5afc314 3590 info->name[0] = 0;
921d9e16 3591 info->recovery_start = MaxSector;
3ad25638 3592 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3593 info->bb.supported = 1;
c2c087e6 3594
97b4d0e9 3595 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3596 mpb = super->anchor;
b7d81a38 3597 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3598
ab3cb6b3
N		 3599 for (i = 0; i < mpb->num_raid_devs; i++) {
3600 struct imsm_dev *dev = get_imsm_dev(super, i);
3601 int failed, enough, j, missing = 0;
3602 struct imsm_map *map;
3603 __u8 state;
97b4d0e9 3604
3b451610
AK		 3605 failed = imsm_count_failed(super, dev, MAP_0);
3606 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3607 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3608
3609 /* any newly missing disks?
3610 * (catches single-degraded vs double-degraded)
3611 */
3612 for (j = 0; j < map->num_members; j++) {
238c0a71 3613 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3614 __u32 idx = ord_to_idx(ord);
3615
20dc76d1
MT		 3616 if (super->disks && super->disks->index == (int)idx)
3617 info->disk.raid_disk = j;
3618
ab3cb6b3
N		 3619 if (!(ord & IMSM_ORD_REBUILD) &&
3620 get_imsm_missing(super, idx)) {
3621 missing = 1;
3622 break;
3623 }
97b4d0e9 3624 }
ab3cb6b3
N		 3625
3626 if (state == IMSM_T_STATE_FAILED)
3627 enough = -1;
3628 else if (state == IMSM_T_STATE_DEGRADED &&
3629 (state != map->map_state || missing))
3630 enough = 0;
3631 else /* we're normal, or already degraded */
3632 enough = 1;
d2bde6d3
AK		 3633 if (is_gen_migration(dev) && missing) {
3634 /* during general migration we need all disks
3635 * that process is running on.
3636 * No new missing disk is allowed.
3637 */
3638 max_enough = -1;
3639 enough = -1;
3640 /* no more checks necessary
3641 */
3642 break;
3643 }
ab3cb6b3
N		 3644 /* in the missing/failed disk case check to see
3645 * if at least one array is runnable
3646 */
3647 max_enough = max(max_enough, enough);
3648 }
1ade5cc1 3649 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3650 info->container_enough = max_enough;
97b4d0e9 3651
4a04ec6c 3652 if (super->disks) {
14e8215b
DW		 3653 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3654
b9f594fe 3655 disk = &super->disks->disk;
5551b113 3656 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3657 info->component_size = reserved;
25ed7e59 3658 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3659 /* we don't change info->disk.raid_disk here because
3660 * this state will be finalized in mdmon after we have
3661 * found the 'most fresh' version of the metadata
3662 */
25ed7e59 3663 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3664 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3665 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3666 }
a575e2a7
DW		 3667
3668 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3669 * ->compare_super may have updated the 'num_raid_devs' field for spares
3670 */
3671 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3672 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3673 else
3674 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3675
3676 /* I don't know how to compute 'map' on imsm, so use safe default */
3677 if (map) {
3678 int i;
3679 for (i = 0; i < map_disks; i++)
3680 map[i] = 1;
3681 }
3682
cdddbdbc
DW		 3683}
3684
5c4cd5da
AC		 3685/* allocates memory and fills disk in mdinfo structure
3686 * for each disk in array */
3687struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3688{
594dc1b8 3689 struct mdinfo *mddev;
5c4cd5da
AC		 3690 struct intel_super *super = st->sb;
3691 struct imsm_disk *disk;
3692 int count = 0;
3693 struct dl *dl;
3694 if (!super || !super->disks)
3695 return NULL;
3696 dl = super->disks;
503975b9 3697 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3698 while (dl) {
3699 struct mdinfo *tmp;
3700 disk = &dl->disk;
503975b9 3701 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3702 if (mddev->devs)
3703 tmp->next = mddev->devs;
3704 mddev->devs = tmp;
3705 tmp->disk.number = count++;
3706 tmp->disk.major = dl->major;
3707 tmp->disk.minor = dl->minor;
3708 tmp->disk.state = is_configured(disk) ?
3709 (1 << MD_DISK_ACTIVE) : 0;
3710 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3711 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3712 tmp->disk.raid_disk = -1;
3713 dl = dl->next;
3714 }
3715 return mddev;
3716}
3717
cdddbdbc
DW		 3718static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3719 char *update, char *devname, int verbose,
3720 int uuid_set, char *homehost)
3721{
f352c545
DW		 3722 /* For 'assemble' and 'force' we need to return non-zero if any
3723 * change was made. For others, the return value is ignored.
3724 * Update options are:
3725 * force-one : This device looks a bit old but needs to be included,
3726 * update age info appropriately.
3727 * assemble: clear any 'faulty' flag to allow this device to
3728 * be assembled.
3729 * force-array: Array is degraded but being forced, mark it clean
3730 * if that will be needed to assemble it.
3731 *
3732 * newdev: not used ????
3733 * grow: Array has gained a new device - this is currently for
3734 * linear only
3735 * resync: mark as dirty so a resync will happen.
3736 * name: update the name - preserving the homehost
6e46bf34 3737 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3738 *
3739 * Following are not relevant for this imsm:
3740 * sparc2.2 : update from old dodgey metadata
3741 * super-minor: change the preferred_minor number
3742 * summaries: update redundant counters.
f352c545
DW		 3743 * homehost: update the recorded homehost
3744 * _reshape_progress: record new reshape_progress position.
3745 */
6e46bf34
DW		 3746 int rv = 1;
3747 struct intel_super *super = st->sb;
3748 struct imsm_super *mpb;
f352c545 3749
6e46bf34
DW		 3750 /* we can only update container info */
3751 if (!super || super->current_vol >= 0 || !super->anchor)
3752 return 1;
3753
3754 mpb = super->anchor;
3755
81a5b4f5
N		 3756 if (strcmp(update, "uuid") == 0) {
3757 /* We take this to mean that the family_num should be updated.
3758 * However that is much smaller than the uuid so we cannot really
3759 * allow an explicit uuid to be given. And it is hard to reliably
3760 * know if one was.
3761 * So if !uuid_set we know the current uuid is random and just used
3762 * the first 'int' and copy it to the other 3 positions.
3763 * Otherwise we require the 4 'int's to be the same as would be the
3764 * case if we are using a random uuid. So an explicit uuid will be
3765 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3766 */
81a5b4f5
N		 3767 if (!uuid_set) {
3768 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3769 rv = 0;
81a5b4f5
N		 3770 } else {
3771 if (info->uuid[0] != info->uuid[1] ||
3772 info->uuid[1] != info->uuid[2] ||
3773 info->uuid[2] != info->uuid[3])
3774 rv = -1;
3775 else
3776 rv = 0;
6e46bf34 3777 }
81a5b4f5
N		 3778 if (rv == 0)
3779 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3780 } else if (strcmp(update, "assemble") == 0)
3781 rv = 0;
3782 else
1e2b2765 3783 rv = -1;
f352c545 3784
6e46bf34
DW		 3785 /* successful update? recompute checksum */
3786 if (rv == 0)
3787 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3788
3789 return rv;
cdddbdbc
DW		 3790}
3791
c2c087e6 3792static size_t disks_to_mpb_size(int disks)
cdddbdbc 3793{
c2c087e6 3794 size_t size;
cdddbdbc 3795
c2c087e6
DW		 3796 size = sizeof(struct imsm_super);
3797 size += (disks - 1) * sizeof(struct imsm_disk);
3798 size += 2 * sizeof(struct imsm_dev);
3799 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3800 size += (4 - 2) * sizeof(struct imsm_map);
3801 /* 4 possible disk_ord_tbl's */
3802 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3803 /* maximum bbm log */
3804 size += sizeof(struct bbm_log);
c2c087e6
DW		 3805
3806 return size;
3807}
3808
387fcd59
N		 3809static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3810 unsigned long long data_offset)
c2c087e6
DW		 3811{
3812 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3813 return 0;
3814
3815 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3816}
3817
ba2de7ba
DW		 3818static void free_devlist(struct intel_super *super)
3819{
3820 struct intel_dev *dv;
3821
3822 while (super->devlist) {
3823 dv = super->devlist->next;
3824 free(super->devlist->dev);
3825 free(super->devlist);
3826 super->devlist = dv;
3827 }
3828}
3829
3830static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3831{
3832 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3833}
3834
cdddbdbc
DW		 3835static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3836{
3837 /*
3838 * return:
3839 * 0 same, or first was empty, and second was copied
3840 * 1 second had wrong number
3841 * 2 wrong uuid
3842 * 3 wrong other info
3843 */
3844 struct intel_super *first = st->sb;
3845 struct intel_super *sec = tst->sb;
3846
5d500228
N		 3847 if (!first) {
3848 st->sb = tst->sb;
3849 tst->sb = NULL;
3850 return 0;
3851 }
8603ea6f
LM		 3852 /* in platform dependent environment test if the disks
3853 * use the same Intel hba
cb8f6859 3854 * If not on Intel hba at all, allow anything.
8603ea6f 3855 */
6b781d33
AP		 3856 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3857 if (first->hba->type != sec->hba->type) {
8603ea6f 3858 fprintf(stderr,
6b781d33
AP		 3859 "HBAs of devices do not match %s != %s\n",
3860 get_sys_dev_type(first->hba->type),
3861 get_sys_dev_type(sec->hba->type));
3862 return 3;
3863 }
3864 if (first->orom != sec->orom) {
3865 fprintf(stderr,
3866 "HBAs of devices do not match %s != %s\n",
3867 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3868 return 3;
3869 }
3870 }
cdddbdbc 3871
d23fe947
DW		 3872 /* if an anchor does not have num_raid_devs set then it is a free
3873 * floating spare
3874 */
3875 if (first->anchor->num_raid_devs > 0 &&
3876 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3877 /* Determine if these disks might ever have been
3878 * related. Further disambiguation can only take place
3879 * in load_super_imsm_all
3880 */
3881 __u32 first_family = first->anchor->orig_family_num;
3882 __u32 sec_family = sec->anchor->orig_family_num;
3883
f796af5d
DW		 3884 if (memcmp(first->anchor->sig, sec->anchor->sig,
3885 MAX_SIGNATURE_LENGTH) != 0)
3886 return 3;
3887
a2b97981
DW		 3888 if (first_family == 0)
3889 first_family = first->anchor->family_num;
3890 if (sec_family == 0)
3891 sec_family = sec->anchor->family_num;
3892
3893 if (first_family != sec_family)
d23fe947 3894 return 3;
f796af5d 3895
d23fe947 3896 }
cdddbdbc 3897
3e372e5a
DW		 3898 /* if 'first' is a spare promote it to a populated mpb with sec's
3899 * family number
3900 */
3901 if (first->anchor->num_raid_devs == 0 &&
3902 sec->anchor->num_raid_devs > 0) {
78d30f94 3903 int i;
ba2de7ba
DW		 3904 struct intel_dev *dv;
3905 struct imsm_dev *dev;
78d30f94
DW		 3906
3907 /* we need to copy raid device info from sec if an allocation
3908 * fails here we don't associate the spare
3909 */
3910 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3911 dv = xmalloc(sizeof(*dv));
3912 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3913 dv->dev = dev;
3914 dv->index = i;
3915 dv->next = first->devlist;
3916 first->devlist = dv;
78d30f94 3917 }
709743c5 3918 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3919 /* allocation failure */
3920 free_devlist(first);
e12b3daa 3921 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3922 return 3;
78d30f94 3923 }
3e372e5a 3924 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3925 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3926 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3927 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3928 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3929 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3930 }
3931
cdddbdbc
DW		 3932 return 0;
3933}
3934
0030e8d6
DW		 3935static void fd2devname(int fd, char *name)
3936{
3937 struct stat st;
3938 char path[256];
33a6535d 3939 char dname[PATH_MAX];
0030e8d6
DW		 3940 char *nm;
3941 int rv;
3942
3943 name[0] = '\0';
3944 if (fstat(fd, &st) != 0)
3945 return;
3946 sprintf(path, "/sys/dev/block/%d:%d",
3947 major(st.st_rdev), minor(st.st_rdev));
3948
9cf014ec 3949 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3950 if (rv <= 0)
3951 return;
9587c373 3952
0030e8d6
DW		 3953 dname[rv] = '\0';
3954 nm = strrchr(dname, '/');
7897de29
JS		 3955 if (nm) {
3956 nm++;
3957 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3958 }
0030e8d6
DW		 3959}
3960
21e9380b
AP		 3961static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3962{
3963 char path[60];
3964 char *name = fd2kname(fd);
3965
3966 if (!name)
3967 return 1;
3968
3969 if (strncmp(name, "nvme", 4) != 0)
3970 return 1;
3971
3972 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3973
3974 return load_sys(path, buf, buf_len);
3975}
3976
cdddbdbc
DW		 3977extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3978
3979static int imsm_read_serial(int fd, char *devname,
6da53c0e 3980 __u8 *serial, size_t serial_buf_len)
cdddbdbc 3981{
21e9380b 3982 char buf[50];
cdddbdbc 3983 int rv;
6da53c0e 3984 size_t len;
316e2bf4
DW		 3985 char *dest;
3986 char *src;
21e9380b
AP		 3987 unsigned int i;
3988
3989 memset(buf, 0, sizeof(buf));
cdddbdbc 3990
21e9380b 3991 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3992
21e9380b
AP		 3993 if (rv)
3994 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3995
40ebbb9c 3996 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3997 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3998 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3999 return 0;
4000 }
4001
cdddbdbc
DW		 4002 if (rv != 0) {
4003 if (devname)
e7b84f9d
N		 4004 pr_err("Failed to retrieve serial for %s\n",
4005 devname);
cdddbdbc
DW		 4006 return rv;
4007 }
4008
316e2bf4
DW		 4009 /* trim all whitespace and non-printable characters and convert
4010 * ':' to ';'
4011 */
21e9380b
AP		 4012 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4013 src = &buf[i];
316e2bf4
DW		 4014 if (*src > 0x20) {
4015 /* ':' is reserved for use in placeholder serial
4016 * numbers for missing disks
4017 */
4018 if (*src == ':')
4019 *dest++ = ';';
4020 else
4021 *dest++ = *src;
4022 }
4023 }
21e9380b
AP		 4024 len = dest - buf;
4025 dest = buf;
316e2bf4 4026
6da53c0e
BK		 4027 if (len > serial_buf_len) {
4028 /* truncate leading characters */
4029 dest += len - serial_buf_len;
4030 len = serial_buf_len;
316e2bf4 4031 }
5c3db629 4032
6da53c0e 4033 memset(serial, 0, serial_buf_len);
316e2bf4 4034 memcpy(serial, dest, len);
cdddbdbc
DW		 4035
4036 return 0;
4037}
4038
1f24f035
DW		 4039static int serialcmp(__u8 *s1, __u8 *s2)
4040{
4041 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4042}
4043
4044static void serialcpy(__u8 *dest, __u8 *src)
4045{
4046 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4047}
4048
54c2c1ea
DW		 4049static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4050{
4051 struct dl *dl;
4052
4053 for (dl = super->disks; dl; dl = dl->next)
4054 if (serialcmp(dl->serial, serial) == 0)
4055 break;
4056
4057 return dl;
4058}
4059
a2b97981
DW		 4060static struct imsm_disk *
4061__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4062{
4063 int i;
4064
4065 for (i = 0; i < mpb->num_disks; i++) {
4066 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4067
4068 if (serialcmp(disk->serial, serial) == 0) {
4069 if (idx)
4070 *idx = i;
4071 return disk;
4072 }
4073 }
4074
4075 return NULL;
4076}
4077
cdddbdbc
DW		 4078static int
4079load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4080{
a2b97981 4081 struct imsm_disk *disk;
cdddbdbc
DW		 4082 struct dl *dl;
4083 struct stat stb;
cdddbdbc 4084 int rv;
a2b97981 4085 char name[40];
d23fe947
DW		 4086 __u8 serial[MAX_RAID_SERIAL_LEN];
4087
6da53c0e 4088 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4089
4090 if (rv != 0)
4091 return 2;
4092
503975b9 4093 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4094
a2b97981
DW		 4095 fstat(fd, &stb);
4096 dl->major = major(stb.st_rdev);
4097 dl->minor = minor(stb.st_rdev);
4098 dl->next = super->disks;
4099 dl->fd = keep_fd ? fd : -1;
4100 assert(super->disks == NULL);
4101 super->disks = dl;
4102 serialcpy(dl->serial, serial);
4103 dl->index = -2;
4104 dl->e = NULL;
4105 fd2devname(fd, name);
4106 if (devname)
503975b9 4107 dl->devname = xstrdup(devname);
a2b97981 4108 else
503975b9 4109 dl->devname = xstrdup(name);
cdddbdbc 4110
d23fe947 4111 /* look up this disk's index in the current anchor */
a2b97981
DW		 4112 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4113 if (disk) {
4114 dl->disk = *disk;
4115 /* only set index on disks that are a member of a
4116 * populated contianer, i.e. one with raid_devs
4117 */
4118 if (is_failed(&dl->disk))
3f6efecc 4119 dl->index = -2;
2432ce9b 4120 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4121 dl->index = -1;
3f6efecc
DW		 4122 }
4123
949c47a0
DW		 4124 return 0;
4125}
4126
0c046afd
DW		 4127/* When migrating map0 contains the 'destination' state while map1
4128 * contains the current state. When not migrating map0 contains the
4129 * current state. This routine assumes that map[0].map_state is set to
4130 * the current array state before being called.
4131 *
4132 * Migration is indicated by one of the following states
4133 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4134 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4135 * map1state=unitialized)
1484e727 4136 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4137 * map1state=normal)
e3bba0e0 4138 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4139 * map1state=degraded)
8e59f3d8
AK		 4140 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4141 * map1state=normal)
0c046afd 4142 */
8e59f3d8
AK		 4143static void migrate(struct imsm_dev *dev, struct intel_super *super,
4144 __u8 to_state, int migr_type)
3393c6af 4145{
0c046afd 4146 struct imsm_map *dest;
238c0a71 4147 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4148
0c046afd 4149 dev->vol.migr_state = 1;
1484e727 4150 set_migr_type(dev, migr_type);
f8f603f1 4151 dev->vol.curr_migr_unit = 0;
238c0a71 4152 dest = get_imsm_map(dev, MAP_1);
0c046afd 4153
0556e1a2 4154 /* duplicate and then set the target end state in map[0] */
3393c6af 4155 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4156 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4157 __u32 ord;
4158 int i;
4159
4160 for (i = 0; i < src->num_members; i++) {
4161 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4162 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4163 }
4164 }
4165
8e59f3d8
AK		 4166 if (migr_type == MIGR_GEN_MIGR)
4167 /* Clear migration record */
4168 memset(super->migr_rec, 0, sizeof(struct migr_record));
4169
0c046afd 4170 src->map_state = to_state;
949c47a0 4171}
f8f603f1 4172
809da78e
AK		 4173static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4174 __u8 map_state)
f8f603f1 4175{
238c0a71
AK		 4176 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4177 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4178 MAP_0 : MAP_1);
28bce06f 4179 int i, j;
0556e1a2
DW		 4180
4181 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4182 * completed in the last migration.
4183 *
28bce06f 4184 * FIXME add support for raid-level-migration
0556e1a2 4185 */
089f9d79
JS		 4186 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4187 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4188 /* when final map state is other than expected
4189 * merge maps (not for migration)
4190 */
4191 int failed;
4192
4193 for (i = 0; i < prev->num_members; i++)
4194 for (j = 0; j < map->num_members; j++)
4195 /* during online capacity expansion
4196 * disks position can be changed
4197 * if takeover is used
4198 */
4199 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4200 ord_to_idx(prev->disk_ord_tbl[i])) {
4201 map->disk_ord_tbl[j] |=
4202 prev->disk_ord_tbl[i];
4203 break;
4204 }
4205 failed = imsm_count_failed(super, dev, MAP_0);
4206 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4207 }
f8f603f1
DW		 4208
4209 dev->vol.migr_state = 0;
ea672ee1 4210 set_migr_type(dev, 0);
f8f603f1
DW		 4211 dev->vol.curr_migr_unit = 0;
4212 map->map_state = map_state;
4213}
949c47a0
DW		 4214
4215static int parse_raid_devices(struct intel_super *super)
4216{
4217 int i;
4218 struct imsm_dev *dev_new;
4d7b1503 4219 size_t len, len_migr;
401d313b 4220 size_t max_len = 0;
4d7b1503
DW		 4221 size_t space_needed = 0;
4222 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4223
4224 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4225 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4226 struct intel_dev *dv;
949c47a0 4227
4d7b1503
DW		 4228 len = sizeof_imsm_dev(dev_iter, 0);
4229 len_migr = sizeof_imsm_dev(dev_iter, 1);
4230 if (len_migr > len)
4231 space_needed += len_migr - len;
ca9de185 4232
503975b9 4233 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4234 if (max_len < len_migr)
4235 max_len = len_migr;
4236 if (max_len > len_migr)
4237 space_needed += max_len - len_migr;
503975b9 4238 dev_new = xmalloc(max_len);
949c47a0 4239 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4240 dv->dev = dev_new;
4241 dv->index = i;
4242 dv->next = super->devlist;
4243 super->devlist = dv;
949c47a0 4244 }
cdddbdbc 4245
4d7b1503
DW		 4246 /* ensure that super->buf is large enough when all raid devices
4247 * are migrating
4248 */
4249 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4250 void *buf;
4251
f36a9ecd
PB		 4252 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4253 super->sector_size);
4254 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4255 return 1;
4256
1f45a8ad
DW		 4257 memcpy(buf, super->buf, super->len);
4258 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4259 free(super->buf);
4260 super->buf = buf;
4261 super->len = len;
4262 }
ca9de185 4263
bbab0940
TM		 4264 super->extra_space += space_needed;
4265
cdddbdbc
DW		 4266 return 0;
4267}
4268
e2f41b2c
AK		 4269/*******************************************************************************
4270 * Function: check_mpb_migr_compatibility
4271 * Description: Function checks for unsupported migration features:
4272 * - migration optimization area (pba_of_lba0)
4273 * - descending reshape (ascending_migr)
4274 * Parameters:
4275 * super : imsm metadata information
4276 * Returns:
4277 * 0 : migration is compatible
4278 * -1 : migration is not compatible
4279 ******************************************************************************/
4280int check_mpb_migr_compatibility(struct intel_super *super)
4281{
4282 struct imsm_map *map0, *map1;
4283 struct migr_record *migr_rec = super->migr_rec;
4284 int i;
4285
4286 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4287 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4288
4289 if (dev_iter &&
4290 dev_iter->vol.migr_state == 1 &&
4291 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4292 /* This device is migrating */
238c0a71
AK		 4293 map0 = get_imsm_map(dev_iter, MAP_0);
4294 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4295 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4296 /* migration optimization area was used */
4297 return -1;
fc54fe7a
JS		 4298 if (migr_rec->ascending_migr == 0 &&
4299 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4300 /* descending reshape not supported yet */
4301 return -1;
4302 }
4303 }
4304 return 0;
4305}
4306
d23fe947 4307static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4308
cdddbdbc 4309/* load_imsm_mpb - read matrix metadata
f2f5c343 4310 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4311 */
4312static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4313{
4314 unsigned long long dsize;
cdddbdbc 4315 unsigned long long sectors;
f36a9ecd 4316 unsigned int sector_size = super->sector_size;
cdddbdbc 4317 struct stat;
6416d527 4318 struct imsm_super *anchor;
cdddbdbc
DW		 4319 __u32 check_sum;
4320
cdddbdbc 4321 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4322 if (dsize < 2*sector_size) {
64436f06 4323 if (devname)
e7b84f9d
N		 4324 pr_err("%s: device to small for imsm\n",
4325 devname);
64436f06
N		 4326 return 1;
4327 }
cdddbdbc 4328
f36a9ecd 4329 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4330 if (devname)
e7b84f9d
N		 4331 pr_err("Cannot seek to anchor block on %s: %s\n",
4332 devname, strerror(errno));
cdddbdbc
DW		 4333 return 1;
4334 }
4335
f36a9ecd 4336 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4337 if (devname)
7a862a02 4338 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4339 return 1;
4340 }
466070ad 4341 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4342 if (devname)
e7b84f9d
N		 4343 pr_err("Cannot read anchor block on %s: %s\n",
4344 devname, strerror(errno));
6416d527 4345 free(anchor);
cdddbdbc
DW		 4346 return 1;
4347 }
4348
6416d527 4349 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4350 if (devname)
e7b84f9d 4351 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4352 free(anchor);
cdddbdbc
DW		 4353 return 2;
4354 }
4355
d23fe947 4356 __free_imsm(super, 0);
f2f5c343
LM		 4357 /* reload capability and hba */
4358
4359 /* capability and hba must be updated with new super allocation */
d424212e 4360 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4361 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4362 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4363 if (devname)
e7b84f9d
N		 4364 pr_err("unable to allocate %zu byte mpb buffer\n",
4365 super->len);
6416d527 4366 free(anchor);
cdddbdbc
DW		 4367 return 2;
4368 }
f36a9ecd 4369 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4370
f36a9ecd 4371 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4372 free(anchor);
8e59f3d8 4373
85337573
AO		 4374 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4375 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4376 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4377 free(super->buf);
4378 return 2;
4379 }
51d83f5d 4380 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4381
949c47a0 4382 if (!sectors) {
ecf45690
DW		 4383 check_sum = __gen_imsm_checksum(super->anchor);
4384 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4385 if (devname)
e7b84f9d
N		 4386 pr_err("IMSM checksum %x != %x on %s\n",
4387 check_sum,
4388 __le32_to_cpu(super->anchor->check_sum),
4389 devname);
ecf45690
DW		 4390 return 2;
4391 }
4392
a2b97981 4393 return 0;
949c47a0 4394 }
cdddbdbc
DW		 4395
4396 /* read the extended mpb */
f36a9ecd 4397 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4398 if (devname)
e7b84f9d
N		 4399 pr_err("Cannot seek to extended mpb on %s: %s\n",
4400 devname, strerror(errno));
cdddbdbc
DW		 4401 return 1;
4402 }
4403
f36a9ecd
PB		 4404 if ((unsigned int)read(fd, super->buf + sector_size,
4405 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4406 if (devname)
e7b84f9d
N		 4407 pr_err("Cannot read extended mpb on %s: %s\n",
4408 devname, strerror(errno));
cdddbdbc
DW		 4409 return 2;
4410 }
4411
949c47a0
DW		 4412 check_sum = __gen_imsm_checksum(super->anchor);
4413 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4414 if (devname)
e7b84f9d
N		 4415 pr_err("IMSM checksum %x != %x on %s\n",
4416 check_sum, __le32_to_cpu(super->anchor->check_sum),
4417 devname);
db575f3b 4418 return 3;
cdddbdbc
DW		 4419 }
4420
a2b97981
DW		 4421 return 0;
4422}
4423
8e59f3d8
AK		 4424static int read_imsm_migr_rec(int fd, struct intel_super *super);
4425
97f81ee2
CA		 4426/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4427static void clear_hi(struct intel_super *super)
4428{
4429 struct imsm_super *mpb = super->anchor;
4430 int i, n;
4431 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4432 return;
4433 for (i = 0; i < mpb->num_disks; ++i) {
4434 struct imsm_disk *disk = &mpb->disk[i];
4435 disk->total_blocks_hi = 0;
4436 }
4437 for (i = 0; i < mpb->num_raid_devs; ++i) {
4438 struct imsm_dev *dev = get_imsm_dev(super, i);
4439 if (!dev)
4440 return;
4441 for (n = 0; n < 2; ++n) {
4442 struct imsm_map *map = get_imsm_map(dev, n);
4443 if (!map)
4444 continue;
4445 map->pba_of_lba0_hi = 0;
4446 map->blocks_per_member_hi = 0;
4447 map->num_data_stripes_hi = 0;
4448 }
4449 }
4450}
4451
a2b97981
DW		 4452static int
4453load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4454{
4455 int err;
4456
4457 err = load_imsm_mpb(fd, super, devname);
4458 if (err)
4459 return err;
f36a9ecd
PB		 4460 if (super->sector_size == 4096)
4461 convert_from_4k(super);
a2b97981
DW		 4462 err = load_imsm_disk(fd, super, devname, keep_fd);
4463 if (err)
4464 return err;
4465 err = parse_raid_devices(super);
8d67477f
TM		 4466 if (err)
4467 return err;
4468 err = load_bbm_log(super);
97f81ee2 4469 clear_hi(super);
a2b97981 4470 return err;
cdddbdbc
DW		 4471}
4472
ae6aad82
DW		 4473static void __free_imsm_disk(struct dl *d)
4474{
4475 if (d->fd >= 0)
4476 close(d->fd);
4477 if (d->devname)
4478 free(d->devname);
0dcecb2e
DW		 4479 if (d->e)
4480 free(d->e);
ae6aad82
DW		 4481 free(d);
4482
4483}
1a64be56 4484
cdddbdbc
DW		 4485static void free_imsm_disks(struct intel_super *super)
4486{
47ee5a45 4487 struct dl *d;
cdddbdbc 4488
47ee5a45
DW		 4489 while (super->disks) {
4490 d = super->disks;
cdddbdbc 4491 super->disks = d->next;
ae6aad82 4492 __free_imsm_disk(d);
cdddbdbc 4493 }
cb82edca
AK		 4494 while (super->disk_mgmt_list) {
4495 d = super->disk_mgmt_list;
4496 super->disk_mgmt_list = d->next;
4497 __free_imsm_disk(d);
4498 }
47ee5a45
DW		 4499 while (super->missing) {
4500 d = super->missing;
4501 super->missing = d->next;
4502 __free_imsm_disk(d);
4503 }
4504
cdddbdbc
DW		 4505}
4506
9ca2c81c 4507/* free all the pieces hanging off of a super pointer */
d23fe947 4508static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4509{
88654014
LM		 4510 struct intel_hba *elem, *next;
4511
9ca2c81c 4512 if (super->buf) {
949c47a0 4513 free(super->buf);
9ca2c81c
DW		 4514 super->buf = NULL;
4515 }
f2f5c343
LM		 4516 /* unlink capability description */
4517 super->orom = NULL;
8e59f3d8
AK		 4518 if (super->migr_rec_buf) {
4519 free(super->migr_rec_buf);
4520 super->migr_rec_buf = NULL;
4521 }
d23fe947
DW		 4522 if (free_disks)
4523 free_imsm_disks(super);
ba2de7ba 4524 free_devlist(super);
88654014
LM		 4525 elem = super->hba;
4526 while (elem) {
4527 if (elem->path)
4528 free((void *)elem->path);
4529 next = elem->next;
4530 free(elem);
4531 elem = next;
88c32bb1 4532 }
8d67477f
TM		 4533 if (super->bbm_log)
4534 free(super->bbm_log);
88654014 4535 super->hba = NULL;
cdddbdbc
DW		 4536}
4537
9ca2c81c
DW		 4538static void free_imsm(struct intel_super *super)
4539{
d23fe947 4540 __free_imsm(super, 1);
928f1424 4541 free(super->bb.entries);
9ca2c81c
DW		 4542 free(super);
4543}
cdddbdbc
DW		 4544
4545static void free_super_imsm(struct supertype *st)
4546{
4547 struct intel_super *super = st->sb;
4548
4549 if (!super)
4550 return;
4551
4552 free_imsm(super);
4553 st->sb = NULL;
4554}
4555
49133e57 4556static struct intel_super *alloc_super(void)
c2c087e6 4557{
503975b9 4558 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4559
503975b9
N		 4560 super->current_vol = -1;
4561 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4562
4563 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4564 sizeof(struct md_bb_entry));
4565 if (!super->bb.entries) {
4566 free(super);
4567 return NULL;
4568 }
4569
c2c087e6
DW		 4570 return super;
4571}
4572
f0f5a016
LM		 4573/*
4574 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4575 */
d424212e 4576static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4577{
4578 struct sys_dev *hba_name;
4579 int rv = 0;
4580
3a30e28e
MT		 4581 if (fd >= 0 && test_partition(fd)) {
4582 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4583 devname);
4584 return 1;
4585 }
089f9d79 4586 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4587 super->orom = NULL;
f0f5a016
LM		 4588 super->hba = NULL;
4589 return 0;
4590 }
4591 hba_name = find_disk_attached_hba(fd, NULL);
4592 if (!hba_name) {
d424212e 4593 if (devname)
e7b84f9d
N		 4594 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4595 devname);
f0f5a016
LM		 4596 return 1;
4597 }
4598 rv = attach_hba_to_super(super, hba_name);
4599 if (rv == 2) {
d424212e
N		 4600 if (devname) {
4601 struct intel_hba *hba = super->hba;
f0f5a016 4602
60f0f54d
PB		 4603 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4604 " but the container is assigned to Intel(R) %s %s (",
d424212e 4605 devname,
614902f6 4606 get_sys_dev_type(hba_name->type),
60f0f54d 4607 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4608 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4609 get_sys_dev_type(super->hba->type),
4610 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4611
f0f5a016
LM		 4612 while (hba) {
4613 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4614 if (hba->next)
4615 fprintf(stderr, ", ");
4616 hba = hba->next;
4617 }
6b781d33 4618 fprintf(stderr, ").\n"
cca67208 4619 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4620 }
f0f5a016
LM		 4621 return 2;
4622 }
6b781d33 4623 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4624 if (!super->orom)
4625 return 3;
614902f6 4626
f0f5a016
LM		 4627 return 0;
4628}
4629
47ee5a45
DW		 4630/* find_missing - helper routine for load_super_imsm_all that identifies
4631 * disks that have disappeared from the system. This routine relies on
4632 * the mpb being uptodate, which it is at load time.
4633 */
4634static int find_missing(struct intel_super *super)
4635{
4636 int i;
4637 struct imsm_super *mpb = super->anchor;
4638 struct dl *dl;
4639 struct imsm_disk *disk;
47ee5a45
DW		 4640
4641 for (i = 0; i < mpb->num_disks; i++) {
4642 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4643 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4644 if (dl)
4645 continue;
47ee5a45 4646
503975b9 4647 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4648 dl->major = 0;
4649 dl->minor = 0;
4650 dl->fd = -1;
503975b9 4651 dl->devname = xstrdup("missing");
47ee5a45
DW		 4652 dl->index = i;
4653 serialcpy(dl->serial, disk->serial);
4654 dl->disk = *disk;
689c9bf3 4655 dl->e = NULL;
47ee5a45
DW		 4656 dl->next = super->missing;
4657 super->missing = dl;
4658 }
4659
4660 return 0;
4661}
4662
a2b97981
DW		 4663static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4664{
4665 struct intel_disk *idisk = disk_list;
4666
4667 while (idisk) {
4668 if (serialcmp(idisk->disk.serial, serial) == 0)
4669 break;
4670 idisk = idisk->next;
4671 }
4672
4673 return idisk;
4674}
4675
4676static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4677 struct intel_super *super,
4678 struct intel_disk **disk_list)
4679{
4680 struct imsm_disk *d = &super->disks->disk;
4681 struct imsm_super *mpb = super->anchor;
4682 int i, j;
4683
4684 for (i = 0; i < tbl_size; i++) {
4685 struct imsm_super *tbl_mpb = table[i]->anchor;
4686 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4687
4688 if (tbl_mpb->family_num == mpb->family_num) {
4689 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4690 dprintf("mpb from %d:%d matches %d:%d\n",
4691 super->disks->major,
a2b97981
DW		 4692 super->disks->minor,
4693 table[i]->disks->major,
4694 table[i]->disks->minor);
4695 break;
4696 }
4697
4698 if (((is_configured(d) && !is_configured(tbl_d)) ||
4699 is_configured(d) == is_configured(tbl_d)) &&
4700 tbl_mpb->generation_num < mpb->generation_num) {
4701 /* current version of the mpb is a
4702 * better candidate than the one in
4703 * super_table, but copy over "cross
4704 * generational" status
4705 */
4706 struct intel_disk *idisk;
4707
1ade5cc1
N		 4708 dprintf("mpb from %d:%d replaces %d:%d\n",
4709 super->disks->major,
a2b97981
DW		 4710 super->disks->minor,
4711 table[i]->disks->major,
4712 table[i]->disks->minor);
4713
4714 idisk = disk_list_get(tbl_d->serial, *disk_list);
4715 if (idisk && is_failed(&idisk->disk))
4716 tbl_d->status |= FAILED_DISK;
4717 break;
4718 } else {
4719 struct intel_disk *idisk;
4720 struct imsm_disk *disk;
4721
4722 /* tbl_mpb is more up to date, but copy
4723 * over cross generational status before
4724 * returning
4725 */
4726 disk = __serial_to_disk(d->serial, mpb, NULL);
4727 if (disk && is_failed(disk))
4728 d->status |= FAILED_DISK;
4729
4730 idisk = disk_list_get(d->serial, *disk_list);
4731 if (idisk) {
4732 idisk->owner = i;
4733 if (disk && is_configured(disk))
4734 idisk->disk.status |= CONFIGURED_DISK;
4735 }
4736
1ade5cc1
N		 4737 dprintf("mpb from %d:%d prefer %d:%d\n",
4738 super->disks->major,
a2b97981
DW		 4739 super->disks->minor,
4740 table[i]->disks->major,
4741 table[i]->disks->minor);
4742
4743 return tbl_size;
4744 }
4745 }
4746 }
4747
4748 if (i >= tbl_size)
4749 table[tbl_size++] = super;
4750 else
4751 table[i] = super;
4752
4753 /* update/extend the merged list of imsm_disk records */
4754 for (j = 0; j < mpb->num_disks; j++) {
4755 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4756 struct intel_disk *idisk;
4757
4758 idisk = disk_list_get(disk->serial, *disk_list);
4759 if (idisk) {
4760 idisk->disk.status |= disk->status;
4761 if (is_configured(&idisk->disk) ||
4762 is_failed(&idisk->disk))
4763 idisk->disk.status &= ~(SPARE_DISK);
4764 } else {
503975b9 4765 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4766 idisk->owner = IMSM_UNKNOWN_OWNER;
4767 idisk->disk = *disk;
4768 idisk->next = *disk_list;
4769 *disk_list = idisk;
4770 }
4771
4772 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4773 idisk->owner = i;
4774 }
4775
4776 return tbl_size;
4777}
4778
4779static struct intel_super *
4780validate_members(struct intel_super *super, struct intel_disk *disk_list,
4781 const int owner)
4782{
4783 struct imsm_super *mpb = super->anchor;
4784 int ok_count = 0;
4785 int i;
4786
4787 for (i = 0; i < mpb->num_disks; i++) {
4788 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4789 struct intel_disk *idisk;
4790
4791 idisk = disk_list_get(disk->serial, disk_list);
4792 if (idisk) {
4793 if (idisk->owner == owner ||
4794 idisk->owner == IMSM_UNKNOWN_OWNER)
4795 ok_count++;
4796 else
1ade5cc1
N		 4797 dprintf("'%.16s' owner %d != %d\n",
4798 disk->serial, idisk->owner,
a2b97981
DW		 4799 owner);
4800 } else {
1ade5cc1
N		 4801 dprintf("unknown disk %x [%d]: %.16s\n",
4802 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4803 disk->serial);
4804 break;
4805 }
4806 }
4807
4808 if (ok_count == mpb->num_disks)
4809 return super;
4810 return NULL;
4811}
4812
4813static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4814{
4815 struct intel_super *s;
4816
4817 for (s = super_list; s; s = s->next) {
4818 if (family_num != s->anchor->family_num)
4819 continue;
e12b3daa 4820 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4821 __le32_to_cpu(family_num), s->disks->devname);
4822 }
4823}
4824
4825static struct intel_super *
4826imsm_thunderdome(struct intel_super **super_list, int len)
4827{
4828 struct intel_super *super_table[len];
4829 struct intel_disk *disk_list = NULL;
4830 struct intel_super *champion, *spare;
4831 struct intel_super *s, **del;
4832 int tbl_size = 0;
4833 int conflict;
4834 int i;
4835
4836 memset(super_table, 0, sizeof(super_table));
4837 for (s = *super_list; s; s = s->next)
4838 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4839
4840 for (i = 0; i < tbl_size; i++) {
4841 struct imsm_disk *d;
4842 struct intel_disk *idisk;
4843 struct imsm_super *mpb = super_table[i]->anchor;
4844
4845 s = super_table[i];
4846 d = &s->disks->disk;
4847
4848 /* 'd' must appear in merged disk list for its
4849 * configuration to be valid
4850 */
4851 idisk = disk_list_get(d->serial, disk_list);
4852 if (idisk && idisk->owner == i)
4853 s = validate_members(s, disk_list, i);
4854 else
4855 s = NULL;
4856
4857 if (!s)
1ade5cc1
N		 4858 dprintf("marking family: %#x from %d:%d offline\n",
4859 mpb->family_num,
a2b97981
DW		 4860 super_table[i]->disks->major,
4861 super_table[i]->disks->minor);
4862 super_table[i] = s;
4863 }
4864
4865 /* This is where the mdadm implementation differs from the Windows
4866 * driver which has no strict concept of a container. We can only
4867 * assemble one family from a container, so when returning a prodigal
4868 * array member to this system the code will not be able to disambiguate
4869 * the container contents that should be assembled ("foreign" versus
4870 * "local"). It requires user intervention to set the orig_family_num
4871 * to a new value to establish a new container. The Windows driver in
4872 * this situation fixes up the volume name in place and manages the
4873 * foreign array as an independent entity.
4874 */
4875 s = NULL;
4876 spare = NULL;
4877 conflict = 0;
4878 for (i = 0; i < tbl_size; i++) {
4879 struct intel_super *tbl_ent = super_table[i];
4880 int is_spare = 0;
4881
4882 if (!tbl_ent)
4883 continue;
4884
4885 if (tbl_ent->anchor->num_raid_devs == 0) {
4886 spare = tbl_ent;
4887 is_spare = 1;
4888 }
4889
4890 if (s && !is_spare) {
4891 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4892 conflict++;
4893 } else if (!s && !is_spare)
4894 s = tbl_ent;
4895 }
4896
4897 if (!s)
4898 s = spare;
4899 if (!s) {
4900 champion = NULL;
4901 goto out;
4902 }
4903 champion = s;
4904
4905 if (conflict)
7a862a02 4906 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4907 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4908
4909 /* collect all dl's onto 'champion', and update them to
4910 * champion's version of the status
4911 */
4912 for (s = *super_list; s; s = s->next) {
4913 struct imsm_super *mpb = champion->anchor;
4914 struct dl *dl = s->disks;
4915
4916 if (s == champion)
4917 continue;
4918
5d7b407a
CA		 4919 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4920
a2b97981
DW		 4921 for (i = 0; i < mpb->num_disks; i++) {
4922 struct imsm_disk *disk;
4923
4924 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4925 if (disk) {
4926 dl->disk = *disk;
4927 /* only set index on disks that are a member of
4928 * a populated contianer, i.e. one with
4929 * raid_devs
4930 */
4931 if (is_failed(&dl->disk))
4932 dl->index = -2;
4933 else if (is_spare(&dl->disk))
4934 dl->index = -1;
4935 break;
4936 }
4937 }
4938
4939 if (i >= mpb->num_disks) {
4940 struct intel_disk *idisk;
4941
4942 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4943 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4944 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4945 dl->index = -1;
4946 else {
4947 dl->index = -2;
4948 continue;
4949 }
4950 }
4951
4952 dl->next = champion->disks;
4953 champion->disks = dl;
4954 s->disks = NULL;
4955 }
4956
4957 /* delete 'champion' from super_list */
4958 for (del = super_list; *del; ) {
4959 if (*del == champion) {
4960 *del = (*del)->next;
4961 break;
4962 } else
4963 del = &(*del)->next;
4964 }
4965 champion->next = NULL;
4966
4967 out:
4968 while (disk_list) {
4969 struct intel_disk *idisk = disk_list;
4970
4971 disk_list = disk_list->next;
4972 free(idisk);
4973 }
4974
4975 return champion;
4976}
4977
9587c373
LM		 4978static int
4979get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4980static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4981 int major, int minor, int keep_fd);
ec50f7b6
LM		 4982static int
4983get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4984 int *max, int keep_fd);
4985
cdddbdbc 4986static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4987 char *devname, struct md_list *devlist,
4988 int keep_fd)
cdddbdbc 4989{
a2b97981
DW		 4990 struct intel_super *super_list = NULL;
4991 struct intel_super *super = NULL;
a2b97981 4992 int err = 0;
9587c373 4993 int i = 0;
dab4a513 4994
9587c373
LM		 4995 if (fd >= 0)
4996 /* 'fd' is an opened container */
4997 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4998 else
ec50f7b6
LM		 4999 /* get super block from devlist devices */
5000 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5001 if (err)
1602d52c 5002 goto error;
a2b97981
DW		 5003 /* all mpbs enter, maybe one leaves */
5004 super = imsm_thunderdome(&super_list, i);
5005 if (!super) {
5006 err = 1;
5007 goto error;
cdddbdbc
DW		 5008 }
5009
47ee5a45
DW		 5010 if (find_missing(super) != 0) {
5011 free_imsm(super);
a2b97981
DW		 5012 err = 2;
5013 goto error;
47ee5a45 5014 }
8e59f3d8
AK		 5015
5016 /* load migration record */
5017 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5018 if (err == -1) {
5019 /* migration is in progress,
5020 * but migr_rec cannot be loaded,
5021 */
8e59f3d8
AK		 5022 err = 4;
5023 goto error;
5024 }
e2f41b2c
AK		 5025
5026 /* Check migration compatibility */
089f9d79 5027 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5028 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5029 if (devname)
5030 fprintf(stderr, " on %s\n", devname);
5031 else
5032 fprintf(stderr, " (IMSM).\n");
5033
5034 err = 5;
5035 goto error;
5036 }
5037
a2b97981
DW		 5038 err = 0;
5039
5040 error:
5041 while (super_list) {
5042 struct intel_super *s = super_list;
5043
5044 super_list = super_list->next;
5045 free_imsm(s);
5046 }
9587c373 5047
a2b97981
DW		 5048 if (err)
5049 return err;
f7e7067b 5050
cdddbdbc 5051 *sbp = super;
9587c373 5052 if (fd >= 0)
4dd2df09 5053 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5054 else
4dd2df09 5055 st->container_devnm[0] = 0;
a2b97981 5056 if (err == 0 && st->ss == NULL) {
bf5a934a 5057 st->ss = &super_imsm;
cdddbdbc
DW		 5058 st->minor_version = 0;
5059 st->max_devs = IMSM_MAX_DEVICES;
5060 }
cdddbdbc
DW		 5061 return 0;
5062}
2b959fbf 5063
ec50f7b6
LM		 5064static int
5065get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5066 int *max, int keep_fd)
5067{
5068 struct md_list *tmpdev;
5069 int err = 0;
5070 int i = 0;
9587c373 5071
ec50f7b6
LM		 5072 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5073 if (tmpdev->used != 1)
5074 continue;
5075 if (tmpdev->container == 1) {
ca9de185 5076 int lmax = 0;
ec50f7b6
LM		 5077 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5078 if (fd < 0) {
e7b84f9d 5079 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5080 tmpdev->devname, strerror(errno));
5081 err = 8;
5082 goto error;
5083 }
5084 err = get_sra_super_block(fd, super_list,
5085 tmpdev->devname, &lmax,
5086 keep_fd);
5087 i += lmax;
5088 close(fd);
5089 if (err) {
5090 err = 7;
5091 goto error;
5092 }
5093 } else {
5094 int major = major(tmpdev->st_rdev);
5095 int minor = minor(tmpdev->st_rdev);
5096 err = get_super_block(super_list,
4dd2df09 5097 NULL,
ec50f7b6
LM		 5098 tmpdev->devname,
5099 major, minor,
5100 keep_fd);
5101 i++;
5102 if (err) {
5103 err = 6;
5104 goto error;
5105 }
5106 }
5107 }
5108 error:
5109 *max = i;
5110 return err;
5111}
9587c373 5112
4dd2df09 5113static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5114 int major, int minor, int keep_fd)
5115{
594dc1b8 5116 struct intel_super *s;
9587c373
LM		 5117 char nm[32];
5118 int dfd = -1;
9587c373
LM		 5119 int err = 0;
5120 int retry;
5121
5122 s = alloc_super();
5123 if (!s) {
5124 err = 1;
5125 goto error;
5126 }
5127
5128 sprintf(nm, "%d:%d", major, minor);
5129 dfd = dev_open(nm, O_RDWR);
5130 if (dfd < 0) {
5131 err = 2;
5132 goto error;
5133 }
5134
fa7bb6f8 5135 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5136 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5137 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5138
5139 /* retry the load if we might have raced against mdmon */
4dd2df09 5140 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5141 for (retry = 0; retry < 3; retry++) {
5142 usleep(3000);
5143 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5144 if (err != 3)
5145 break;
5146 }
5147 error:
5148 if (!err) {
5149 s->next = *super_list;
5150 *super_list = s;
5151 } else {
5152 if (s)
8d67477f 5153 free_imsm(s);
36614e95 5154 if (dfd >= 0)
9587c373
LM		 5155 close(dfd);
5156 }
089f9d79 5157 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5158 close(dfd);
5159 return err;
5160
5161}
5162
5163static int
5164get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5165{
5166 struct mdinfo *sra;
4dd2df09 5167 char *devnm;
9587c373
LM		 5168 struct mdinfo *sd;
5169 int err = 0;
5170 int i = 0;
4dd2df09 5171 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5172 if (!sra)
5173 return 1;
5174
5175 if (sra->array.major_version != -1 ||
5176 sra->array.minor_version != -2 ||
5177 strcmp(sra->text_version, "imsm") != 0) {
5178 err = 1;
5179 goto error;
5180 }
5181 /* load all mpbs */
4dd2df09 5182 devnm = fd2devnm(fd);
9587c373 5183 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5184 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5185 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5186 err = 7;
5187 goto error;
5188 }
5189 }
5190 error:
5191 sysfs_free(sra);
5192 *max = i;
5193 return err;
5194}
5195
2b959fbf
N		 5196static int load_container_imsm(struct supertype *st, int fd, char *devname)
5197{
ec50f7b6 5198 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5199}
cdddbdbc
DW		 5200
5201static int load_super_imsm(struct supertype *st, int fd, char *devname)
5202{
5203 struct intel_super *super;
5204 int rv;
8a3544f8 5205 int retry;
cdddbdbc 5206
357ac106 5207 if (test_partition(fd))
691c6ee1
N		 5208 /* IMSM not allowed on partitions */
5209 return 1;
5210
37424f13
DW		 5211 free_super_imsm(st);
5212
49133e57 5213 super = alloc_super();
fa7bb6f8 5214 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5215 if (!super)
5216 return 1;
ea2bc72b
LM		 5217 /* Load hba and capabilities if they exist.
5218 * But do not preclude loading metadata in case capabilities or hba are
5219 * non-compliant and ignore_hw_compat is set.
5220 */
d424212e 5221 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5222 /* no orom/efi or non-intel hba of the disk */
089f9d79 5223 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5224 if (devname)
e7b84f9d 5225 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5226 free_imsm(super);
5227 return 2;
5228 }
a2b97981 5229 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5230
8a3544f8
AP		 5231 /* retry the load if we might have raced against mdmon */
5232 if (rv == 3) {
f96b1302
AP		 5233 struct mdstat_ent *mdstat = NULL;
5234 char *name = fd2kname(fd);
5235
5236 if (name)
5237 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5238
5239 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5240 for (retry = 0; retry < 3; retry++) {
5241 usleep(3000);
5242 rv = load_and_parse_mpb(fd, super, devname, 0);
5243 if (rv != 3)
5244 break;
5245 }
5246 }
5247
5248 free_mdstat(mdstat);
5249 }
5250
cdddbdbc
DW		 5251 if (rv) {
5252 if (devname)
7a862a02 5253 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5254 free_imsm(super);
5255 return rv;
5256 }
5257
5258 st->sb = super;
5259 if (st->ss == NULL) {
5260 st->ss = &super_imsm;
5261 st->minor_version = 0;
5262 st->max_devs = IMSM_MAX_DEVICES;
5263 }
8e59f3d8
AK		 5264
5265 /* load migration record */
2e062e82
AK		 5266 if (load_imsm_migr_rec(super, NULL) == 0) {
5267 /* Check for unsupported migration features */
5268 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5269 pr_err("Unsupported migration detected");
2e062e82
AK		 5270 if (devname)
5271 fprintf(stderr, " on %s\n", devname);
5272 else
5273 fprintf(stderr, " (IMSM).\n");
5274 return 3;
5275 }
e2f41b2c
AK		 5276 }
5277
cdddbdbc
DW		 5278 return 0;
5279}
5280
ef6ffade
DW		 5281static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5282{
5283 if (info->level == 1)
5284 return 128;
5285 return info->chunk_size >> 9;
5286}
5287
5551b113
CA		 5288static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5289 unsigned long long size)
fcfd9599 5290{
4025c288 5291 if (info->level == 1)
5551b113 5292 return size * 2;
4025c288 5293 else
5551b113 5294 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5295}
5296
4d1313e9
DW		 5297static void imsm_update_version_info(struct intel_super *super)
5298{
5299 /* update the version and attributes */
5300 struct imsm_super *mpb = super->anchor;
5301 char *version;
5302 struct imsm_dev *dev;
5303 struct imsm_map *map;
5304 int i;
5305
5306 for (i = 0; i < mpb->num_raid_devs; i++) {
5307 dev = get_imsm_dev(super, i);
238c0a71 5308 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5309 if (__le32_to_cpu(dev->size_high) > 0)
5310 mpb->attributes |= MPB_ATTRIB_2TB;
5311
5312 /* FIXME detect when an array spans a port multiplier */
5313 #if 0
5314 mpb->attributes |= MPB_ATTRIB_PM;
5315 #endif
5316
5317 if (mpb->num_raid_devs > 1 ||
5318 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5319 version = MPB_VERSION_ATTRIBS;
5320 switch (get_imsm_raid_level(map)) {
5321 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5322 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5323 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5324 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5325 }
5326 } else {
5327 if (map->num_members >= 5)
5328 version = MPB_VERSION_5OR6_DISK_ARRAY;
5329 else if (dev->status == DEV_CLONE_N_GO)
5330 version = MPB_VERSION_CNG;
5331 else if (get_imsm_raid_level(map) == 5)
5332 version = MPB_VERSION_RAID5;
5333 else if (map->num_members >= 3)
5334 version = MPB_VERSION_3OR4_DISK_ARRAY;
5335 else if (get_imsm_raid_level(map) == 1)
5336 version = MPB_VERSION_RAID1;
5337 else
5338 version = MPB_VERSION_RAID0;
5339 }
5340 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5341 }
5342}
5343
aa534678
DW		 5344static int check_name(struct intel_super *super, char *name, int quiet)
5345{
5346 struct imsm_super *mpb = super->anchor;
5347 char *reason = NULL;
9bd99a90
RS		 5348 char *start = name;
5349 size_t len = strlen(name);
aa534678
DW		 5350 int i;
5351
9bd99a90
RS		 5352 if (len > 0) {
5353 while (isspace(start[len - 1]))
5354 start[--len] = 0;
5355 while (*start && isspace(*start))
5356 ++start, --len;
5357 memmove(name, start, len + 1);
5358 }
5359
5360 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5361 reason = "must be 16 characters or less";
9bd99a90
RS		 5362 else if (len == 0)
5363 reason = "must be a non-empty string";
aa534678
DW		 5364
5365 for (i = 0; i < mpb->num_raid_devs; i++) {
5366 struct imsm_dev *dev = get_imsm_dev(super, i);
5367
5368 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5369 reason = "already exists";
5370 break;
5371 }
5372 }
5373
5374 if (reason && !quiet)
e7b84f9d 5375 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5376
5377 return !reason;
5378}
5379
8b353278 5380static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5381 struct shape *s, char *name,
83cd1e97
N		 5382 char *homehost, int *uuid,
5383 long long data_offset)
cdddbdbc 5384{
c2c087e6
DW		 5385 /* We are creating a volume inside a pre-existing container.
5386 * so st->sb is already set.
5387 */
5388 struct intel_super *super = st->sb;
f36a9ecd 5389 unsigned int sector_size = super->sector_size;
949c47a0 5390 struct imsm_super *mpb = super->anchor;
ba2de7ba 5391 struct intel_dev *dv;
c2c087e6
DW		 5392 struct imsm_dev *dev;
5393 struct imsm_vol *vol;
5394 struct imsm_map *map;
5395 int idx = mpb->num_raid_devs;
5396 int i;
760365f9 5397 int namelen;
c2c087e6 5398 unsigned long long array_blocks;
2c092cad 5399 size_t size_old, size_new;
5551b113 5400 unsigned long long num_data_stripes;
b53bfba6
TM		 5401 unsigned int data_disks;
5402 unsigned long long size_per_member;
cdddbdbc 5403
88c32bb1 5404 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5405 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5406 return 0;
5407 }
5408
2c092cad
DW		 5409 /* ensure the mpb is large enough for the new data */
5410 size_old = __le32_to_cpu(mpb->mpb_size);
5411 size_new = disks_to_mpb_size(info->nr_disks);
5412 if (size_new > size_old) {
5413 void *mpb_new;
f36a9ecd 5414 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5415
f36a9ecd 5416 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5417 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5418 return 0;
5419 }
85337573
AO		 5420 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5421 MIGR_REC_BUF_SECTORS*
5422 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5423 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5424 free(super->buf);
5425 free(super);
ea944c8f 5426 free(mpb_new);
8e59f3d8
AK		 5427 return 0;
5428 }
2c092cad
DW		 5429 memcpy(mpb_new, mpb, size_old);
5430 free(mpb);
5431 mpb = mpb_new;
949c47a0 5432 super->anchor = mpb_new;
2c092cad
DW		 5433 mpb->mpb_size = __cpu_to_le32(size_new);
5434 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5435 super->len = size_round;
2c092cad 5436 }
bf5a934a 5437 super->current_vol = idx;
3960e579
DW		 5438
5439 /* handle 'failed_disks' by either:
5440 * a) create dummy disk entries in the table if this the first
5441 * volume in the array. We add them here as this is the only
5442 * opportunity to add them. add_to_super_imsm_volume()
5443 * handles the non-failed disks and continues incrementing
5444 * mpb->num_disks.
5445 * b) validate that 'failed_disks' matches the current number
5446 * of missing disks if the container is populated
d23fe947 5447 */
3960e579 5448 if (super->current_vol == 0) {
d23fe947 5449 mpb->num_disks = 0;
3960e579
DW		 5450 for (i = 0; i < info->failed_disks; i++) {
5451 struct imsm_disk *disk;
5452
5453 mpb->num_disks++;
5454 disk = __get_imsm_disk(mpb, i);
5455 disk->status = CONFIGURED_DISK | FAILED_DISK;
5456 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5457 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5458 "missing:%d", (__u8)i);
3960e579
DW		 5459 }
5460 find_missing(super);
5461 } else {
5462 int missing = 0;
5463 struct dl *d;
5464
5465 for (d = super->missing; d; d = d->next)
5466 missing++;
5467 if (info->failed_disks > missing) {
e7b84f9d 5468 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5469 return 0;
5470 }
5471 }
5a038140 5472
aa534678
DW		 5473 if (!check_name(super, name, 0))
5474 return 0;
503975b9
N		 5475 dv = xmalloc(sizeof(*dv));
5476 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5477 /*
5478 * Explicitly allow truncating to not confuse gcc's
5479 * -Werror=stringop-truncation
5480 */
5481 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5482 memcpy(dev->volume, name, namelen);
e03640bd 5483 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5484 info->layout, info->chunk_size,
b53bfba6
TM		 5485 s->size * BLOCKS_PER_KB);
5486 data_disks = get_data_disks(info->level, info->layout,
5487 info->raid_disks);
5488 array_blocks = round_size_to_mb(array_blocks, data_disks);
5489 size_per_member = array_blocks / data_disks;
979d38be 5490
fcc2c9da 5491 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5492 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5493 vol = &dev->vol;
5494 vol->migr_state = 0;
1484e727 5495 set_migr_type(dev, MIGR_INIT);
3960e579 5496 vol->dirty = !info->state;
f8f603f1 5497 vol->curr_migr_unit = 0;
238c0a71 5498 map = get_imsm_map(dev, MAP_0);
5551b113 5499 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5500 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5501 map->failed_disk_num = ~0;
bf4442ab 5502 if (info->level > 0)
fffaf1ff
N		 5503 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5504 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5505 else
5506 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5507 IMSM_T_STATE_NORMAL;
252d23c0 5508 map->ddf = 1;
ef6ffade
DW		 5509
5510 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5511 free(dev);
5512 free(dv);
7a862a02 5513 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5514 return 0;
5515 }
81062a36
DW		 5516
5517 map->raid_level = info->level;
4d1313e9 5518 if (info->level == 10) {
c2c087e6 5519 map->raid_level = 1;
4d1313e9 5520 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5521 } else if (info->level == 1)
5522 map->num_domains = info->raid_disks;
5523 else
ff596308 5524 map->num_domains = 1;
81062a36 5525
5551b113 5526 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5527 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5528 num_data_stripes /= map->num_domains;
5529 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5530
44490938
MD		 5531 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5532 set_blocks_per_member(map, info_to_blocks_per_member(info,
5533 size_per_member /
5534 BLOCKS_PER_KB));
5535
c2c087e6
DW		 5536 map->num_members = info->raid_disks;
5537 for (i = 0; i < map->num_members; i++) {
5538 /* initialized in add_to_super */
4eb26970 5539 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5540 }
949c47a0 5541 mpb->num_raid_devs++;
2a24dc1b
PB		 5542 mpb->num_raid_devs_created++;
5543 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5544
b7580566 5545 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5546 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5547 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5548 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5549 } else {
5550 free(dev);
5551 free(dv);
5552 pr_err("imsm does not support consistency policy %s\n",
5553 map_num(consistency_policies, s->consistency_policy));
5554 return 0;
5555 }
5556
ba2de7ba
DW		 5557 dv->dev = dev;
5558 dv->index = super->current_vol;
5559 dv->next = super->devlist;
5560 super->devlist = dv;
c2c087e6 5561
4d1313e9
DW		 5562 imsm_update_version_info(super);
5563
c2c087e6 5564 return 1;
cdddbdbc
DW		 5565}
5566
bf5a934a 5567static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5568 struct shape *s, char *name,
83cd1e97
N		 5569 char *homehost, int *uuid,
5570 unsigned long long data_offset)
bf5a934a
DW		 5571{
5572 /* This is primarily called by Create when creating a new array.
5573 * We will then get add_to_super called for each component, and then
5574 * write_init_super called to write it out to each device.
5575 * For IMSM, Create can create on fresh devices or on a pre-existing
5576 * array.
5577 * To create on a pre-existing array a different method will be called.
5578 * This one is just for fresh drives.
5579 */
5580 struct intel_super *super;
5581 struct imsm_super *mpb;
5582 size_t mpb_size;
4d1313e9 5583 char *version;
bf5a934a 5584
83cd1e97 5585 if (data_offset != INVALID_SECTORS) {
ed503f89 5586 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5587 return 0;
5588 }
5589
bf5a934a 5590 if (st->sb)
5308f117 5591 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5592 data_offset);
e683ca88
DW		 5593
5594 if (info)
5595 mpb_size = disks_to_mpb_size(info->nr_disks);
5596 else
f36a9ecd 5597 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5598
49133e57 5599 super = alloc_super();
f36a9ecd
PB		 5600 if (super &&
5601 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5602 free_imsm(super);
e683ca88
DW		 5603 super = NULL;
5604 }
5605 if (!super) {
1ade5cc1 5606 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5607 return 0;
5608 }
de44e46f
PB		 5609 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5610 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5611 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5612 free(super->buf);
8d67477f 5613 free_imsm(super);
8e59f3d8
AK		 5614 return 0;
5615 }
e683ca88 5616 memset(super->buf, 0, mpb_size);
ef649044 5617 mpb = super->buf;
e683ca88
DW		 5618 mpb->mpb_size = __cpu_to_le32(mpb_size);
5619 st->sb = super;
5620
5621 if (info == NULL) {
5622 /* zeroing superblock */
5623 return 0;
5624 }
bf5a934a 5625
4d1313e9
DW		 5626 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5627
5628 version = (char *) mpb->sig;
5629 strcpy(version, MPB_SIGNATURE);
5630 version += strlen(MPB_SIGNATURE);
5631 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5632
bf5a934a
DW		 5633 return 1;
5634}
5635
f2cc4f7d
AO		 5636static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5637{
5638 unsigned int member_sector_size;
5639
5640 if (dl->fd < 0) {
5641 pr_err("Invalid file descriptor for %s\n", dl->devname);
5642 return 0;
5643 }
5644
5645 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5646 return 0;
5647 if (member_sector_size != super->sector_size)
5648 return 0;
5649 return 1;
5650}
5651
f20c3968 5652static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5653 int fd, char *devname)
5654{
5655 struct intel_super *super = st->sb;
d23fe947 5656 struct imsm_super *mpb = super->anchor;
3960e579 5657 struct imsm_disk *_disk;
bf5a934a
DW		 5658 struct imsm_dev *dev;
5659 struct imsm_map *map;
3960e579 5660 struct dl *dl, *df;
4eb26970 5661 int slot;
bf5a934a 5662
949c47a0 5663 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5664 map = get_imsm_map(dev, MAP_0);
bf5a934a 5665
208933a7 5666 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5667 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5668 devname);
5669 return 1;
5670 }
5671
efb30e7f
DW		 5672 if (fd == -1) {
5673 /* we're doing autolayout so grab the pre-marked (in
5674 * validate_geometry) raid_disk
5675 */
5676 for (dl = super->disks; dl; dl = dl->next)
5677 if (dl->raiddisk == dk->raid_disk)
5678 break;
5679 } else {
5680 for (dl = super->disks; dl ; dl = dl->next)
5681 if (dl->major == dk->major &&
5682 dl->minor == dk->minor)
5683 break;
5684 }
d23fe947 5685
208933a7 5686 if (!dl) {
e7b84f9d 5687 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5688 return 1;
208933a7 5689 }
bf5a934a 5690
59632db9
MZ		 5691 if (mpb->num_disks == 0)
5692 if (!get_dev_sector_size(dl->fd, dl->devname,
5693 &super->sector_size))
5694 return 1;
5695
f2cc4f7d
AO		 5696 if (!drive_validate_sector_size(super, dl)) {
5697 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5698 return 1;
5699 }
5700
d23fe947
DW		 5701 /* add a pristine spare to the metadata */
5702 if (dl->index < 0) {
5703 dl->index = super->anchor->num_disks;
5704 super->anchor->num_disks++;
5705 }
4eb26970
DW		 5706 /* Check the device has not already been added */
5707 slot = get_imsm_disk_slot(map, dl->index);
5708 if (slot >= 0 &&
238c0a71 5709 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5710 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5711 devname);
5712 return 1;
5713 }
656b6b5a 5714 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5715 dl->disk.status = CONFIGURED_DISK;
d23fe947 5716
3960e579
DW		 5717 /* update size of 'missing' disks to be at least as large as the
5718 * largest acitve member (we only have dummy missing disks when
5719 * creating the first volume)
5720 */
5721 if (super->current_vol == 0) {
5722 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5723 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5724 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5725 _disk = __get_imsm_disk(mpb, df->index);
5726 *_disk = df->disk;
5727 }
5728 }
5729
5730 /* refresh unset/failed slots to point to valid 'missing' entries */
5731 for (df = super->missing; df; df = df->next)
5732 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5733 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5734
5735 if ((ord & IMSM_ORD_REBUILD) == 0)
5736 continue;
5737 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5738 if (is_gen_migration(dev)) {
238c0a71
AK		 5739 struct imsm_map *map2 = get_imsm_map(dev,
5740 MAP_1);
0a108d63 5741 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5742 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5743 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5744 slot2,
5745 MAP_1);
1ace8403
AK		 5746 if ((unsigned)df->index ==
5747 ord_to_idx(ord2))
5748 set_imsm_ord_tbl_ent(map2,
0a108d63 5749 slot2,
1ace8403
AK		 5750 df->index |
5751 IMSM_ORD_REBUILD);
5752 }
5753 }
3960e579
DW		 5754 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5755 break;
5756 }
5757
d23fe947
DW		 5758 /* if we are creating the first raid device update the family number */
5759 if (super->current_vol == 0) {
5760 __u32 sum;
5761 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5762
3960e579 5763 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5764 if (!_dev || !_disk) {
e7b84f9d 5765 pr_err("BUG mpb setup error\n");
791b666a
AW		 5766 return 1;
5767 }
d23fe947
DW		 5768 *_dev = *dev;
5769 *_disk = dl->disk;
148acb7b
DW		 5770 sum = random32();
5771 sum += __gen_imsm_checksum(mpb);
d23fe947 5772 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5773 mpb->orig_family_num = mpb->family_num;
e48aed3c 5774 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5775 }
ca0748fa 5776 super->current_disk = dl;
f20c3968 5777 return 0;
bf5a934a
DW		 5778}
5779
a8619d23
AK		 5780/* mark_spare()
5781 * Function marks disk as spare and restores disk serial
5782 * in case it was previously marked as failed by takeover operation
5783 * reruns:
5784 * -1 : critical error
5785 * 0 : disk is marked as spare but serial is not set
5786 * 1 : success
5787 */
5788int mark_spare(struct dl *disk)
5789{
5790 __u8 serial[MAX_RAID_SERIAL_LEN];
5791 int ret_val = -1;
5792
5793 if (!disk)
5794 return ret_val;
5795
5796 ret_val = 0;
6da53c0e 5797 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5798 /* Restore disk serial number, because takeover marks disk
5799 * as failed and adds to serial ':0' before it becomes
5800 * a spare disk.
5801 */
5802 serialcpy(disk->serial, serial);
5803 serialcpy(disk->disk.serial, serial);
5804 ret_val = 1;
5805 }
5806 disk->disk.status = SPARE_DISK;
5807 disk->index = -1;
5808
5809 return ret_val;
5810}
88654014 5811
12724c01
TM		 5812
5813static int write_super_imsm_spare(struct intel_super *super, struct dl *d);
5814
f20c3968 5815static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5816 int fd, char *devname,
5817 unsigned long long data_offset)
cdddbdbc 5818{
c2c087e6 5819 struct intel_super *super = st->sb;
c2c087e6
DW		 5820 struct dl *dd;
5821 unsigned long long size;
fa7bb6f8 5822 unsigned int member_sector_size;
f2f27e63 5823 __u32 id;
c2c087e6
DW		 5824 int rv;
5825 struct stat stb;
5826
88654014
LM		 5827 /* If we are on an RAID enabled platform check that the disk is
5828 * attached to the raid controller.
5829 * We do not need to test disks attachment for container based additions,
5830 * they shall be already tested when container was created/assembled.
88c32bb1 5831 */
d424212e 5832 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5833 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5834 if (rv != 0) {
5835 dprintf("capability: %p fd: %d ret: %d\n",
5836 super->orom, fd, rv);
5837 return 1;
88c32bb1
DW		 5838 }
5839
f20c3968
DW		 5840 if (super->current_vol >= 0)
5841 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5842
c2c087e6 5843 fstat(fd, &stb);
503975b9 5844 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5845 dd->major = major(stb.st_rdev);
5846 dd->minor = minor(stb.st_rdev);
503975b9 5847 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5848 dd->fd = fd;
689c9bf3 5849 dd->e = NULL;
1a64be56 5850 dd->action = DISK_ADD;
6da53c0e 5851 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5852 if (rv) {
e7b84f9d 5853 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5854 if (dd->devname)
5855 free(dd->devname);
949c47a0 5856 free(dd);
0030e8d6 5857 abort();
c2c087e6 5858 }
20bee0f8
PB		 5859 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5860 (super->hba->type == SYS_DEV_VMD))) {
5861 int i;
5862 char *devpath = diskfd_to_devpath(fd);
5863 char controller_path[PATH_MAX];
5864
5865 if (!devpath) {
5866 pr_err("failed to get devpath, aborting\n");
5867 if (dd->devname)
5868 free(dd->devname);
5869 free(dd);
5870 return 1;
5871 }
5872
5873 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5874 free(devpath);
5875
5876 if (devpath_to_vendor(controller_path) == 0x8086) {
5877 /*
5878 * If Intel's NVMe drive has serial ended with
5879 * "-A","-B","-1" or "-2" it means that this is "x8"
5880 * device (double drive on single PCIe card).
5881 * User should be warned about potential data loss.
5882 */
5883 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5884 /* Skip empty character at the end */
5885 if (dd->serial[i] == 0)
5886 continue;
5887
5888 if (((dd->serial[i] == 'A') ||
5889 (dd->serial[i] == 'B') ||
5890 (dd->serial[i] == '1') ||
5891 (dd->serial[i] == '2')) &&
5892 (dd->serial[i-1] == '-'))
5893 pr_err("\tThe action you are about to take may put your data at risk.\n"
5894 "\tPlease note that x8 devices may consist of two separate x4 devices "
5895 "located on a single PCIe port.\n"
5896 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5897 break;
5898 }
32716c51
PB		 5899 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5900 !imsm_orom_has_tpv_support(super->orom)) {
5901 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5902 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5903 free(dd->devname);
5904 free(dd);
5905 return 1;
20bee0f8
PB		 5906 }
5907 }
c2c087e6 5908
c2c087e6 5909 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5910 get_dev_sector_size(fd, NULL, &member_sector_size);
5911
5912 if (super->sector_size == 0) {
5913 /* this a first device, so sector_size is not set yet */
5914 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5915 }
5916
71e5411e 5917 /* clear migr_rec when adding disk to container */
85337573
AO		 5918 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5919 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5920 SEEK_SET) >= 0) {
466070ad 5921 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5922 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5923 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5924 perror("Write migr_rec failed");
5925 }
5926
c2c087e6 5927 size /= 512;
1f24f035 5928 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5929 set_total_blocks(&dd->disk, size);
5930 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5931 struct imsm_super *mpb = super->anchor;
5932 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5933 }
a8619d23 5934 mark_spare(dd);
c2c087e6 5935 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5936 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5937 else
b9f594fe 5938 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5939
5940 if (st->update_tail) {
1a64be56
LM		 5941 dd->next = super->disk_mgmt_list;
5942 super->disk_mgmt_list = dd;
43dad3d6 5943 } else {
12724c01
TM		 5944 /* this is called outside of mdmon
5945 * write initial spare metadata
5946 * mdmon will overwrite it.
5947 */
43dad3d6
DW		 5948 dd->next = super->disks;
5949 super->disks = dd;
12724c01 5950 write_super_imsm_spare(super, dd);
43dad3d6 5951 }
f20c3968
DW		 5952
5953 return 0;
cdddbdbc
DW		 5954}
5955
1a64be56
LM		 5956static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5957{
5958 struct intel_super *super = st->sb;
5959 struct dl *dd;
5960
5961 /* remove from super works only in mdmon - for communication
5962 * manager - monitor. Check if communication memory buffer
5963 * is prepared.
5964 */
5965 if (!st->update_tail) {
1ade5cc1 5966 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5967 return 1;
5968 }
503975b9 5969 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5970 dd->major = dk->major;
5971 dd->minor = dk->minor;
1a64be56 5972 dd->fd = -1;
a8619d23 5973 mark_spare(dd);
1a64be56
LM		 5974 dd->action = DISK_REMOVE;
5975
5976 dd->next = super->disk_mgmt_list;
5977 super->disk_mgmt_list = dd;
5978
1a64be56
LM		 5979 return 0;
5980}
5981
f796af5d
DW		 5982static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5983
5984static union {
f36a9ecd 5985 char buf[MAX_SECTOR_SIZE];
f796af5d 5986 struct imsm_super anchor;
f36a9ecd 5987} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5988
12724c01
TM		 5989
5990static int write_super_imsm_spare(struct intel_super *super, struct dl *d)
d23fe947 5991{
d23fe947 5992 struct imsm_super *mpb = super->anchor;
f796af5d 5993 struct imsm_super *spare = &spare_record.anchor;
d23fe947 5994 __u32 sum;
12724c01
TM		 5995
5996 if (d->index != -1)
5997 return 1;
d23fe947 5998
68641cdb
JS		 5999 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6000 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6001 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6002 spare->num_disks = 1;
6003 spare->num_raid_devs = 0;
6004 spare->cache_size = mpb->cache_size;
6005 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6006
6007 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6008 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947 6009
12724c01
TM		 6010 spare->disk[0] = d->disk;
6011 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6012 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6013
6014 if (super->sector_size == 4096)
6015 convert_to_4k_imsm_disk(&spare->disk[0]);
d23fe947 6016
12724c01
TM		 6017 sum = __gen_imsm_checksum(spare);
6018 spare->family_num = __cpu_to_le32(sum);
6019 spare->orig_family_num = 0;
6020 sum = __gen_imsm_checksum(spare);
6021 spare->check_sum = __cpu_to_le32(sum);
027c374f 6022
12724c01
TM		 6023 if (store_imsm_mpb(d->fd, spare)) {
6024 pr_err("failed for device %d:%d %s\n",
6025 d->major, d->minor, strerror(errno));
6026 return 1;
6027 }
6028
6029 return 0;
6030}
6031/* spare records have their own family number and do not have any defined raid
6032 * devices
6033 */
6034static int write_super_imsm_spares(struct intel_super *super, int doclose)
6035{
6036 struct dl *d;
f36a9ecd 6037
12724c01
TM		 6038 for (d = super->disks; d; d = d->next) {
6039 if (d->index != -1)
6040 continue;
d23fe947 6041
12724c01 6042 if (write_super_imsm_spare(super, d))
e74255d9 6043 return 1;
12724c01 6044
d23fe947
DW		 6045 if (doclose) {
6046 close(d->fd);
6047 d->fd = -1;
6048 }
6049 }
6050
e74255d9 6051 return 0;
d23fe947
DW		 6052}
6053
36988a3d 6054static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6055{
36988a3d 6056 struct intel_super *super = st->sb;
f36a9ecd 6057 unsigned int sector_size = super->sector_size;
949c47a0 6058 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6059 struct dl *d;
6060 __u32 generation;
6061 __u32 sum;
d23fe947 6062 int spares = 0;
949c47a0 6063 int i;
a48ac0a8 6064 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6065 int num_disks = 0;
146c6260 6066 int clear_migration_record = 1;
bbab0940 6067 __u32 bbm_log_size;
cdddbdbc 6068
c2c087e6
DW		 6069 /* 'generation' is incremented everytime the metadata is written */
6070 generation = __le32_to_cpu(mpb->generation_num);
6071 generation++;
6072 mpb->generation_num = __cpu_to_le32(generation);
6073
148acb7b
DW		 6074 /* fix up cases where previous mdadm releases failed to set
6075 * orig_family_num
6076 */
6077 if (mpb->orig_family_num == 0)
6078 mpb->orig_family_num = mpb->family_num;
6079
d23fe947 6080 for (d = super->disks; d; d = d->next) {
8796fdc4 6081 if (d->index == -1)
d23fe947 6082 spares++;
36988a3d 6083 else {
d23fe947 6084 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6085 num_disks++;
6086 }
d23fe947 6087 }
36988a3d 6088 for (d = super->missing; d; d = d->next) {
47ee5a45 6089 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6090 num_disks++;
6091 }
6092 mpb->num_disks = num_disks;
6093 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6094
949c47a0
DW		 6095 for (i = 0; i < mpb->num_raid_devs; i++) {
6096 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6097 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6098 if (dev && dev2) {
6099 imsm_copy_dev(dev, dev2);
6100 mpb_size += sizeof_imsm_dev(dev, 0);
6101 }
146c6260
AK		 6102 if (is_gen_migration(dev2))
6103 clear_migration_record = 0;
949c47a0 6104 }
bbab0940
TM		 6105
6106 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6107
6108 if (bbm_log_size) {
6109 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6110 mpb->attributes |= MPB_ATTRIB_BBM;
6111 } else
6112 mpb->attributes &= ~MPB_ATTRIB_BBM;
6113
6114 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6115 mpb_size += bbm_log_size;
a48ac0a8 6116 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6117
bbab0940
TM		 6118#ifdef DEBUG
6119 assert(super->len == 0 || mpb_size <= super->len);
6120#endif
6121
c2c087e6 6122 /* recalculate checksum */
949c47a0 6123 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6124 mpb->check_sum = __cpu_to_le32(sum);
6125
51d83f5d
AK		 6126 if (super->clean_migration_record_by_mdmon) {
6127 clear_migration_record = 1;
6128 super->clean_migration_record_by_mdmon = 0;
6129 }
146c6260 6130 if (clear_migration_record)
de44e46f 6131 memset(super->migr_rec_buf, 0,
85337573 6132 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6133
f36a9ecd
PB		 6134 if (sector_size == 4096)
6135 convert_to_4k(super);
6136
d23fe947 6137 /* write the mpb for disks that compose raid devices */
c2c087e6 6138 for (d = super->disks; d ; d = d->next) {
86c54047 6139 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6140 continue;
30602f53 6141
146c6260
AK		 6142 if (clear_migration_record) {
6143 unsigned long long dsize;
6144
6145 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6146 if (lseek64(d->fd, dsize - sector_size,
6147 SEEK_SET) >= 0) {
466070ad
PB		 6148 if ((unsigned int)write(d->fd,
6149 super->migr_rec_buf,
de44e46f
PB		 6150 MIGR_REC_BUF_SECTORS*sector_size) !=
6151 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6152 perror("Write migr_rec failed");
146c6260
AK		 6153 }
6154 }
51d83f5d
AK		 6155
6156 if (store_imsm_mpb(d->fd, mpb))
6157 fprintf(stderr,
1ade5cc1
N		 6158 "failed for device %d:%d (fd: %d)%s\n",
6159 d->major, d->minor,
51d83f5d
AK		 6160 d->fd, strerror(errno));
6161
c2c087e6
DW		 6162 if (doclose) {
6163 close(d->fd);
6164 d->fd = -1;
6165 }
6166 }
6167
d23fe947
DW		 6168 if (spares)
6169 return write_super_imsm_spares(super, doclose);
6170
e74255d9 6171 return 0;
c2c087e6
DW		 6172}
6173
9b1fb677 6174static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6175{
6176 size_t len;
6177 struct imsm_update_create_array *u;
6178 struct intel_super *super = st->sb;
9b1fb677 6179 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6180 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6181 struct disk_info *inf;
6182 struct imsm_disk *disk;
6183 int i;
43dad3d6 6184
54c2c1ea
DW		 6185 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6186 sizeof(*inf) * map->num_members;
503975b9 6187 u = xmalloc(len);
43dad3d6 6188 u->type = update_create_array;
9b1fb677 6189 u->dev_idx = dev_idx;
43dad3d6 6190 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6191 inf = get_disk_info(u);
6192 for (i = 0; i < map->num_members; i++) {
238c0a71 6193 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6194
54c2c1ea 6195 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6196 if (!disk)
6197 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6198 serialcpy(inf[i].serial, disk->serial);
6199 }
43dad3d6
DW		 6200 append_metadata_update(st, u, len);
6201
6202 return 0;
6203}
6204
1a64be56 6205static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6206{
6207 struct intel_super *super = st->sb;
6208 size_t len;
1a64be56 6209 struct imsm_update_add_remove_disk *u;
43dad3d6 6210
1a64be56 6211 if (!super->disk_mgmt_list)
43dad3d6
DW		 6212 return 0;
6213
6214 len = sizeof(*u);
503975b9 6215 u = xmalloc(len);
1a64be56 6216 u->type = update_add_remove_disk;
43dad3d6
DW		 6217 append_metadata_update(st, u, len);
6218
6219 return 0;
6220}
2432ce9b
AP		 6221
6222__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6223
e397cefe
AP		 6224static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6225{
6226 struct ppl_header *ppl_hdr = buf;
6227 int ret;
6228
6229 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6230
6231 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6232 ret = -errno;
6233 perror("Failed to seek to PPL header location");
6234 return ret;
6235 }
6236
6237 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6238 ret = -errno;
6239 perror("Write PPL header failed");
6240 return ret;
6241 }
6242
6243 fsync(fd);
6244
6245 return 0;
6246}
6247
2432ce9b
AP		 6248static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6249{
6250 struct intel_super *super = st->sb;
6251 void *buf;
6252 struct ppl_header *ppl_hdr;
6253 int ret;
6254
b2514242
PB		 6255 /* first clear entire ppl space */
6256 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6257 if (ret)
6258 return ret;
6259
6260 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6261 if (ret) {
6262 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6263 return -ret;
2432ce9b
AP		 6264 }
6265
6266 memset(buf, 0, PPL_HEADER_SIZE);
6267 ppl_hdr = buf;
6268 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6269 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6270
6271 if (info->mismatch_cnt) {
6272 /*
6273 * We are overwriting an invalid ppl. Make one entry with wrong
6274 * checksum to prevent the kernel from skipping resync.
6275 */
6276 ppl_hdr->entries_count = __cpu_to_le32(1);
6277 ppl_hdr->entries[0].checksum = ~0;
6278 }
6279
e397cefe 6280 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6281
6282 free(buf);
6283 return ret;
6284}
6285
e397cefe
AP		 6286static int is_rebuilding(struct imsm_dev *dev);
6287
2432ce9b
AP		 6288static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6289 struct mdinfo *disk)
6290{
6291 struct intel_super *super = st->sb;
6292 struct dl *d;
e397cefe 6293 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6294 int ret = 0;
e397cefe 6295 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6296 __u32 crc;
6297 struct imsm_dev *dev;
2432ce9b 6298 __u32 idx;
44b6b876
PB		 6299 unsigned int i;
6300 unsigned long long ppl_offset = 0;
6301 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6302
6303 if (disk->disk.raid_disk < 0)
6304 return 0;
6305
2432ce9b 6306 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6307 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6308 d = get_imsm_dl_disk(super, idx);
6309
6310 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6311 return 0;
6312
6313 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6314 pr_err("Failed to allocate PPL header buffer\n");
6315 return -1;
6316 }
6317 buf = buf_orig;
2432ce9b 6318
44b6b876
PB		 6319 ret = 1;
6320 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6321 void *tmp;
6322
44b6b876 6323 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6324
44b6b876
PB		 6325 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6326 SEEK_SET) < 0) {
6327 perror("Failed to seek to PPL header location");
6328 ret = -1;
e397cefe 6329 break;
44b6b876 6330 }
2432ce9b 6331
44b6b876
PB		 6332 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6333 perror("Read PPL header failed");
6334 ret = -1;
e397cefe 6335 break;
44b6b876 6336 }
2432ce9b 6337
44b6b876 6338 ppl_hdr = buf;
2432ce9b 6339
44b6b876
PB		 6340 crc = __le32_to_cpu(ppl_hdr->checksum);
6341 ppl_hdr->checksum = 0;
6342
6343 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6344 dprintf("Wrong PPL header checksum on %s\n",
6345 d->devname);
e397cefe 6346 break;
44b6b876
PB		 6347 }
6348
6349 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6350 /* previous was newest, it was already checked */
e397cefe 6351 break;
44b6b876
PB		 6352 }
6353
6354 if ((__le32_to_cpu(ppl_hdr->signature) !=
6355 super->anchor->orig_family_num)) {
6356 dprintf("Wrong PPL header signature on %s\n",
6357 d->devname);
6358 ret = 1;
e397cefe 6359 break;
44b6b876
PB		 6360 }
6361
6362 ret = 0;
6363 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6364
44b6b876
PB		 6365 ppl_offset += PPL_HEADER_SIZE;
6366 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6367 ppl_offset +=
6368 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6369
6370 if (!buf_prev)
6371 buf_prev = buf + PPL_HEADER_SIZE;
6372 tmp = buf_prev;
6373 buf_prev = buf;
6374 buf = tmp;
2432ce9b
AP		 6375 }
6376
e397cefe
AP		 6377 if (buf_prev) {
6378 buf = buf_prev;
6379 ppl_hdr = buf_prev;
6380 }
2432ce9b 6381
54148aba
PB		 6382 /*
6383 * Update metadata to use mutliple PPLs area (1MB).
6384 * This is done once for all RAID members
6385 */
6386 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6387 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6388 char subarray[20];
6389 struct mdinfo *member_dev;
6390
6391 sprintf(subarray, "%d", info->container_member);
6392
6393 if (mdmon_running(st->container_devnm))
6394 st->update_tail = &st->updates;
6395
6396 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6397 pr_err("Failed to update subarray %s\n",
6398 subarray);
6399 } else {
6400 if (st->update_tail)
6401 flush_metadata_updates(st);
6402 else
6403 st->ss->sync_metadata(st);
6404 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6405 for (member_dev = info->devs; member_dev;
6406 member_dev = member_dev->next)
6407 member_dev->ppl_size =
6408 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6409 }
6410 }
6411
b23d0750 6412 if (ret == 1) {
2fc0fc63
AP		 6413 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6414
50b9c10d
PB		 6415 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6416 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6417 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6418 (is_rebuilding(dev) &&
2ec9d182
AP		 6419 dev->vol.curr_migr_unit == 0 &&
6420 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6421 ret = st->ss->write_init_ppl(st, info, d->fd);
6422 else
6423 info->mismatch_cnt++;
e397cefe
AP		 6424 } else if (ret == 0 &&
6425 ppl_hdr->entries_count == 0 &&
6426 is_rebuilding(dev) &&
6427 info->resync_start == 0) {
6428 /*
6429 * The header has no entries - add a single empty entry and
6430 * rewrite the header to prevent the kernel from going into
6431 * resync after an interrupted rebuild.
6432 */
6433 ppl_hdr->entries_count = __cpu_to_le32(1);
6434 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6435 }
2432ce9b 6436
e397cefe
AP		 6437 free(buf_orig);
6438
2432ce9b
AP		 6439 return ret;
6440}
6441
2432ce9b
AP		 6442static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6443{
6444 struct intel_super *super = st->sb;
6445 struct dl *d;
6446 int ret = 0;
6447
6448 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6449 info->array.level != 5)
6450 return 0;
6451
6452 for (d = super->disks; d ; d = d->next) {
6453 if (d->index < 0 || is_failed(&d->disk))
6454 continue;
6455
6456 ret = st->ss->write_init_ppl(st, info, d->fd);
6457 if (ret)
6458 break;
6459 }
6460
6461 return ret;
6462}
43dad3d6 6463
c2c087e6
DW		 6464static int write_init_super_imsm(struct supertype *st)
6465{
9b1fb677
DW		 6466 struct intel_super *super = st->sb;
6467 int current_vol = super->current_vol;
2432ce9b
AP		 6468 int rv = 0;
6469 struct mdinfo info;
6470
6471 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6472
6473 /* we are done with current_vol reset it to point st at the container */
6474 super->current_vol = -1;
6475
8273f55e 6476 if (st->update_tail) {
43dad3d6
DW		 6477 /* queue the recently created array / added disk
6478 * as a metadata update */
8273f55e 6479
43dad3d6 6480 /* determine if we are creating a volume or adding a disk */
9b1fb677 6481 if (current_vol < 0) {
1a64be56
LM		 6482 /* in the mgmt (add/remove) disk case we are running
6483 * in mdmon context, so don't close fd's
43dad3d6 6484 */
2432ce9b
AP		 6485 rv = mgmt_disk(st);
6486 } else {
6487 rv = write_init_ppl_imsm_all(st, &info);
6488 if (!rv)
6489 rv = create_array(st, current_vol);
6490 }
d682f344
N		 6491 } else {
6492 struct dl *d;
6493 for (d = super->disks; d; d = d->next)
ba728be7 6494 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6495 if (current_vol >= 0)
6496 rv = write_init_ppl_imsm_all(st, &info);
6497 if (!rv)
6498 rv = write_super_imsm(st, 1);
d682f344 6499 }
2432ce9b
AP		 6500
6501 return rv;
cdddbdbc
DW		 6502}
6503
e683ca88 6504static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6505{
e683ca88
DW		 6506 struct intel_super *super = st->sb;
6507 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6508
e683ca88 6509 if (!mpb)
ad97895e
DW		 6510 return 1;
6511
f36a9ecd
PB		 6512 if (super->sector_size == 4096)
6513 convert_to_4k(super);
e683ca88 6514 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6515}
6516
cdddbdbc
DW		 6517static int validate_geometry_imsm_container(struct supertype *st, int level,
6518 int layout, int raiddisks, int chunk,
af4348dd
N		 6519 unsigned long long size,
6520 unsigned long long data_offset,
6521 char *dev,
2c514b71
NB		 6522 unsigned long long *freesize,
6523 int verbose)
cdddbdbc 6524{
c2c087e6
DW		 6525 int fd;
6526 unsigned long long ldsize;
594dc1b8 6527 struct intel_super *super;
f2f5c343 6528 int rv = 0;
cdddbdbc 6529
c2c087e6
DW		 6530 if (level != LEVEL_CONTAINER)
6531 return 0;
6532 if (!dev)
6533 return 1;
6534
6535 fd = open(dev, O_RDONLY|O_EXCL, 0);
6536 if (fd < 0) {
ba728be7 6537 if (verbose > 0)
e7b84f9d 6538 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6539 dev, strerror(errno));
c2c087e6
DW		 6540 return 0;
6541 }
6542 if (!get_dev_size(fd, dev, &ldsize)) {
6543 close(fd);
6544 return 0;
6545 }
f2f5c343
LM		 6546
6547 /* capabilities retrieve could be possible
6548 * note that there is no fd for the disks in array.
6549 */
6550 super = alloc_super();
8d67477f
TM		 6551 if (!super) {
6552 close(fd);
6553 return 0;
6554 }
fa7bb6f8
PB		 6555 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6556 close(fd);
6557 free_imsm(super);
6558 return 0;
6559 }
6560
ba728be7 6561 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6562 if (rv != 0) {
6563#if DEBUG
6564 char str[256];
6565 fd2devname(fd, str);
1ade5cc1 6566 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6567 fd, str, super->orom, rv, raiddisks);
6568#endif
6569 /* no orom/efi or non-intel hba of the disk */
6570 close(fd);
6571 free_imsm(super);
6572 return 0;
6573 }
c2c087e6 6574 close(fd);
9126b9a8
CA		 6575 if (super->orom) {
6576 if (raiddisks > super->orom->tds) {
6577 if (verbose)
7a862a02 6578 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6579 raiddisks, super->orom->tds);
6580 free_imsm(super);
6581 return 0;
6582 }
6583 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6584 (ldsize >> 9) >> 32 > 0) {
6585 if (verbose)
e7b84f9d 6586 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6587 free_imsm(super);
6588 return 0;
6589 }
f2f5c343 6590 }
c2c087e6 6591
af4348dd 6592 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6593 free_imsm(super);
c2c087e6
DW		 6594
6595 return 1;
cdddbdbc
DW		 6596}
6597
0dcecb2e
DW		 6598static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6599{
6600 const unsigned long long base_start = e[*idx].start;
6601 unsigned long long end = base_start + e[*idx].size;
6602 int i;
6603
6604 if (base_start == end)
6605 return 0;
6606
6607 *idx = *idx + 1;
6608 for (i = *idx; i < num_extents; i++) {
6609 /* extend overlapping extents */
6610 if (e[i].start >= base_start &&
6611 e[i].start <= end) {
6612 if (e[i].size == 0)
6613 return 0;
6614 if (e[i].start + e[i].size > end)
6615 end = e[i].start + e[i].size;
6616 } else if (e[i].start > end) {
6617 *idx = i;
6618 break;
6619 }
6620 }
6621
6622 return end - base_start;
6623}
6624
6625static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6626{
6627 /* build a composite disk with all known extents and generate a new
6628 * 'maxsize' given the "all disks in an array must share a common start
6629 * offset" constraint
6630 */
503975b9 6631 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6632 struct dl *dl;
6633 int i, j;
6634 int start_extent;
6635 unsigned long long pos;
b9d77223 6636 unsigned long long start = 0;
0dcecb2e
DW		 6637 unsigned long long maxsize;
6638 unsigned long reserve;
6639
0dcecb2e
DW		 6640 /* coalesce and sort all extents. also, check to see if we need to
6641 * reserve space between member arrays
6642 */
6643 j = 0;
6644 for (dl = super->disks; dl; dl = dl->next) {
6645 if (!dl->e)
6646 continue;
6647 for (i = 0; i < dl->extent_cnt; i++)
6648 e[j++] = dl->e[i];
6649 }
6650 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6651
6652 /* merge extents */
6653 i = 0;
6654 j = 0;
6655 while (i < sum_extents) {
6656 e[j].start = e[i].start;
6657 e[j].size = find_size(e, &i, sum_extents);
6658 j++;
6659 if (e[j-1].size == 0)
6660 break;
6661 }
6662
6663 pos = 0;
6664 maxsize = 0;
6665 start_extent = 0;
6666 i = 0;
6667 do {
6668 unsigned long long esize;
6669
6670 esize = e[i].start - pos;
6671 if (esize >= maxsize) {
6672 maxsize = esize;
6673 start = pos;
6674 start_extent = i;
6675 }
6676 pos = e[i].start + e[i].size;
6677 i++;
6678 } while (e[i-1].size);
6679 free(e);
6680
a7dd165b
DW		 6681 if (maxsize == 0)
6682 return 0;
6683
6684 /* FIXME assumes volume at offset 0 is the first volume in a
6685 * container
6686 */
0dcecb2e
DW		 6687 if (start_extent > 0)
6688 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6689 else
6690 reserve = 0;
6691
6692 if (maxsize < reserve)
a7dd165b 6693 return 0;
0dcecb2e 6694
5551b113 6695 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6696 if (start + reserve > super->create_offset)
a7dd165b 6697 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6698 super->create_offset = start + reserve;
6699
6700 return maxsize - reserve;
6701}
6702
88c32bb1
DW		 6703static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6704{
6705 if (level < 0 || level == 6 || level == 4)
6706 return 0;
6707
6708 /* if we have an orom prevent invalid raid levels */
6709 if (orom)
6710 switch (level) {
6711 case 0: return imsm_orom_has_raid0(orom);
6712 case 1:
6713 if (raiddisks > 2)
6714 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6715 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6716 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6717 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6718 }
6719 else
6720 return 1; /* not on an Intel RAID platform so anything goes */
6721
6722 return 0;
6723}
6724
ca9de185
LM		 6725static int
6726active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6727 int dpa, int verbose)
6728{
6729 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6730 struct mdstat_ent *memb;
ca9de185
LM		 6731 int count = 0;
6732 int num = 0;
594dc1b8 6733 struct md_list *dv;
ca9de185
LM		 6734 int found;
6735
6736 for (memb = mdstat ; memb ; memb = memb->next) {
6737 if (memb->metadata_version &&
fc54fe7a 6738 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6739 (strcmp(&memb->metadata_version[9], name) == 0) &&
6740 !is_subarray(memb->metadata_version+9) &&
6741 memb->members) {
6742 struct dev_member *dev = memb->members;
6743 int fd = -1;
6744 while(dev && (fd < 0)) {
503975b9
N		 6745 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6746 num = sprintf(path, "%s%s", "/dev/", dev->name);
6747 if (num > 0)
6748 fd = open(path, O_RDONLY, 0);
089f9d79 6749 if (num <= 0 || fd < 0) {
676e87a8 6750 pr_vrb("Cannot open %s: %s\n",
503975b9 6751 dev->name, strerror(errno));
ca9de185 6752 }
503975b9 6753 free(path);
ca9de185
LM		 6754 dev = dev->next;
6755 }
6756 found = 0;
089f9d79 6757 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6758 struct mdstat_ent *vol;
6759 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6760 if (vol->active > 0 &&
ca9de185 6761 vol->metadata_version &&
9581efb1 6762 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6763 found++;
6764 count++;
6765 }
6766 }
6767 if (*devlist && (found < dpa)) {
503975b9 6768 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6769 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6770 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6771 dv->found = found;
6772 dv->used = 0;
6773 dv->next = *devlist;
6774 *devlist = dv;
ca9de185
LM		 6775 }
6776 }
6777 if (fd >= 0)
6778 close(fd);
6779 }
6780 }
6781 free_mdstat(mdstat);
6782 return count;
6783}
6784
6785#ifdef DEBUG_LOOP
6786static struct md_list*
6787get_loop_devices(void)
6788{
6789 int i;
6790 struct md_list *devlist = NULL;
594dc1b8 6791 struct md_list *dv;
ca9de185
LM		 6792
6793 for(i = 0; i < 12; i++) {
503975b9
N		 6794 dv = xcalloc(1, sizeof(*dv));
6795 dv->devname = xmalloc(40);
ca9de185
LM		 6796 sprintf(dv->devname, "/dev/loop%d", i);
6797 dv->next = devlist;
6798 devlist = dv;
6799 }
6800 return devlist;
6801}
6802#endif
6803
6804static struct md_list*
6805get_devices(const char *hba_path)
6806{
6807 struct md_list *devlist = NULL;
594dc1b8 6808 struct md_list *dv;
ca9de185
LM		 6809 struct dirent *ent;
6810 DIR *dir;
6811 int err = 0;
6812
6813#if DEBUG_LOOP
6814 devlist = get_loop_devices();
6815 return devlist;
6816#endif
6817 /* scroll through /sys/dev/block looking for devices attached to
6818 * this hba
6819 */
6820 dir = opendir("/sys/dev/block");
6821 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6822 int fd;
6823 char buf[1024];
6824 int major, minor;
6825 char *path = NULL;
6826 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6827 continue;
6828 path = devt_to_devpath(makedev(major, minor));
6829 if (!path)
6830 continue;
6831 if (!path_attached_to_hba(path, hba_path)) {
6832 free(path);
6833 path = NULL;
6834 continue;
6835 }
6836 free(path);
6837 path = NULL;
6838 fd = dev_open(ent->d_name, O_RDONLY);
6839 if (fd >= 0) {
6840 fd2devname(fd, buf);
6841 close(fd);
6842 } else {
e7b84f9d 6843 pr_err("cannot open device: %s\n",
ca9de185
LM		 6844 ent->d_name);
6845 continue;
6846 }
6847
503975b9
N		 6848 dv = xcalloc(1, sizeof(*dv));
6849 dv->devname = xstrdup(buf);
ca9de185
LM		 6850 dv->next = devlist;
6851 devlist = dv;
6852 }
6853 if (err) {
6854 while(devlist) {
6855 dv = devlist;
6856 devlist = devlist->next;
6857 free(dv->devname);
6858 free(dv);
6859 }
6860 }
562aa102 6861 closedir(dir);
ca9de185
LM		 6862 return devlist;
6863}
6864
6865static int
6866count_volumes_list(struct md_list *devlist, char *homehost,
6867 int verbose, int *found)
6868{
6869 struct md_list *tmpdev;
6870 int count = 0;
594dc1b8 6871 struct supertype *st;
ca9de185
LM		 6872
6873 /* first walk the list of devices to find a consistent set
6874 * that match the criterea, if that is possible.
6875 * We flag the ones we like with 'used'.
6876 */
6877 *found = 0;
6878 st = match_metadata_desc_imsm("imsm");
6879 if (st == NULL) {
676e87a8 6880 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6881 return 0;
6882 }
6883
6884 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6885 char *devname = tmpdev->devname;
0a6bff09 6886 dev_t rdev;
ca9de185
LM		 6887 struct supertype *tst;
6888 int dfd;
6889 if (tmpdev->used > 1)
6890 continue;
6891 tst = dup_super(st);
6892 if (tst == NULL) {
676e87a8 6893 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6894 goto err_1;
6895 }
6896 tmpdev->container = 0;
6897 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6898 if (dfd < 0) {
1ade5cc1 6899 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6900 devname, strerror(errno));
6901 tmpdev->used = 2;
0a6bff09 6902 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6903 tmpdev->used = 2;
6904 } else if (must_be_container(dfd)) {
6905 struct supertype *cst;
6906 cst = super_by_fd(dfd, NULL);
6907 if (cst == NULL) {
1ade5cc1 6908 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6909 devname);
6910 tmpdev->used = 2;
6911 } else if (tst->ss != st->ss) {
1ade5cc1 6912 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6913 devname);
6914 tmpdev->used = 2;
6915 } else if (!tst->ss->load_container ||
6916 tst->ss->load_container(tst, dfd, NULL))
6917 tmpdev->used = 2;
6918 else {
6919 tmpdev->container = 1;
6920 }
6921 if (cst)
6922 cst->ss->free_super(cst);
6923 } else {
0a6bff09 6924 tmpdev->st_rdev = rdev;
ca9de185 6925 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6926 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6927 devname);
6928 tmpdev->used = 2;
6929 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6930 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6931 tst->ss->name, devname);
6932 tmpdev->used = 2;
6933 }
6934 }
6935 if (dfd >= 0)
6936 close(dfd);
6937 if (tmpdev->used == 2 || tmpdev->used == 4) {
6938 /* Ignore unrecognised devices during auto-assembly */
6939 goto loop;
6940 }
6941 else {
6942 struct mdinfo info;
6943 tst->ss->getinfo_super(tst, &info, NULL);
6944
6945 if (st->minor_version == -1)
6946 st->minor_version = tst->minor_version;
6947
6948 if (memcmp(info.uuid, uuid_zero,
6949 sizeof(int[4])) == 0) {
6950 /* this is a floating spare. It cannot define
6951 * an array unless there are no more arrays of
6952 * this type to be found. It can be included
6953 * in an array of this type though.
6954 */
6955 tmpdev->used = 3;
6956 goto loop;
6957 }
6958
6959 if (st->ss != tst->ss ||
6960 st->minor_version != tst->minor_version ||
6961 st->ss->compare_super(st, tst) != 0) {
6962 /* Some mismatch. If exactly one array matches this host,
6963 * we can resolve on that one.
6964 * Or, if we are auto assembling, we just ignore the second
6965 * for now.
6966 */
1ade5cc1 6967 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6968 devname);
6969 goto loop;
6970 }
6971 tmpdev->used = 1;
6972 *found = 1;
6973 dprintf("found: devname: %s\n", devname);
6974 }
6975 loop:
6976 if (tst)
6977 tst->ss->free_super(tst);
6978 }
6979 if (*found != 0) {
6980 int err;
6981 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6982 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6983 for (iter = head; iter; iter = iter->next) {
6984 dprintf("content->text_version: %s vol\n",
6985 iter->text_version);
6986 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6987 /* do not assemble arrays with unsupported
6988 configurations */
1ade5cc1 6989 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6990 iter->text_version);
6991 } else
6992 count++;
6993 }
6994 sysfs_free(head);
6995
6996 } else {
1ade5cc1 6997 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6998 err, st->sb);
6999 }
7000 } else {
1ade5cc1 7001 dprintf("no more devices to examine\n");
ca9de185
LM		 7002 }
7003
7004 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7005 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7006 if (count) {
7007 if (count < tmpdev->found)
7008 count = 0;
7009 else
7010 count -= tmpdev->found;
7011 }
7012 }
7013 if (tmpdev->used == 1)
7014 tmpdev->used = 4;
7015 }
7016 err_1:
7017 if (st)
7018 st->ss->free_super(st);
7019 return count;
7020}
7021
d3c11416
AO		 7022static int __count_volumes(char *hba_path, int dpa, int verbose,
7023 int cmp_hba_path)
ca9de185 7024{
72a45777 7025 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7026 int count = 0;
72a45777
PB		 7027 const struct orom_entry *entry;
7028 struct devid_list *dv, *devid_list;
ca9de185 7029
d3c11416 7030 if (!hba_path)
ca9de185
LM		 7031 return 0;
7032
72a45777 7033 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7034 if (strstr(idev->path, hba_path))
7035 break;
72a45777
PB		 7036 }
7037
7038 if (!idev || !idev->dev_id)
ca9de185 7039 return 0;
72a45777
PB		 7040
7041 entry = get_orom_entry_by_device_id(idev->dev_id);
7042
7043 if (!entry || !entry->devid_list)
7044 return 0;
7045
7046 devid_list = entry->devid_list;
7047 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7048 struct md_list *devlist;
d3c11416
AO		 7049 struct sys_dev *device = NULL;
7050 char *hpath;
72a45777
PB		 7051 int found = 0;
7052
d3c11416
AO		 7053 if (cmp_hba_path)
7054 device = device_by_id_and_path(dv->devid, hba_path);
7055 else
7056 device = device_by_id(dv->devid);
7057
72a45777 7058 if (device)
d3c11416 7059 hpath = device->path;
72a45777
PB		 7060 else
7061 return 0;
7062
d3c11416 7063 devlist = get_devices(hpath);
72a45777
PB		 7064 /* if no intel devices return zero volumes */
7065 if (devlist == NULL)
7066 return 0;
7067
d3c11416
AO		 7068 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7069 verbose);
7070 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7071 if (devlist == NULL)
7072 return 0;
7073 do {
7074 found = 0;
7075 count += count_volumes_list(devlist,
7076 NULL,
7077 verbose,
7078 &found);
7079 dprintf("found %d count: %d\n", found, count);
7080 } while (found);
7081
d3c11416 7082 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7083
7084 while (devlist) {
7085 struct md_list *dv = devlist;
7086 devlist = devlist->next;
7087 free(dv->devname);
7088 free(dv);
7089 }
ca9de185
LM		 7090 }
7091 return count;
7092}
7093
d3c11416
AO		 7094static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7095{
7096 if (!hba)
7097 return 0;
7098 if (hba->type == SYS_DEV_VMD) {
7099 struct sys_dev *dev;
7100 int count = 0;
7101
7102 for (dev = find_intel_devices(); dev; dev = dev->next) {
7103 if (dev->type == SYS_DEV_VMD)
7104 count += __count_volumes(dev->path, dpa,
7105 verbose, 1);
7106 }
7107 return count;
7108 }
7109 return __count_volumes(hba->path, dpa, verbose, 0);
7110}
7111
cd9d1ac7
DW		 7112static int imsm_default_chunk(const struct imsm_orom *orom)
7113{
7114 /* up to 512 if the plaform supports it, otherwise the platform max.
7115 * 128 if no platform detected
7116 */
7117 int fs = max(7, orom ? fls(orom->sss) : 0);
7118
7119 return min(512, (1 << fs));
7120}
73408129 7121
6592ce37
DW		 7122static int
7123validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7124 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7125{
660260d0
DW		 7126 /* check/set platform and metadata limits/defaults */
7127 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7128 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7129 super->orom->dpa);
73408129
LM		 7130 return 0;
7131 }
7132
5d500228 7133 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7134 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7135 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7136 level, raiddisks, raiddisks > 1 ? "s" : "");
7137 return 0;
7138 }
cd9d1ac7 7139
7ccc4cc4 7140 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7141 *chunk = imsm_default_chunk(super->orom);
7142
7ccc4cc4 7143 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7144 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7145 return 0;
6592ce37 7146 }
cd9d1ac7 7147
6592ce37
DW		 7148 if (layout != imsm_level_to_layout(level)) {
7149 if (level == 5)
676e87a8 7150 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7151 else if (level == 10)
676e87a8 7152 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7153 else
676e87a8 7154 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7155 layout, level);
7156 return 0;
7157 }
2cc699af 7158
7ccc4cc4 7159 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7160 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7161 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7162 return 0;
7163 }
614902f6 7164
6592ce37
DW		 7165 return 1;
7166}
7167
1011e834 7168/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7169 * FIX ME add ahci details
7170 */
8b353278 7171static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7172 int layout, int raiddisks, int *chunk,
af4348dd
N		 7173 unsigned long long size,
7174 unsigned long long data_offset,
7175 char *dev,
2c514b71
NB		 7176 unsigned long long *freesize,
7177 int verbose)
cdddbdbc 7178{
9e04ac1c 7179 dev_t rdev;
c2c087e6 7180 struct intel_super *super = st->sb;
b2916f25 7181 struct imsm_super *mpb;
c2c087e6
DW		 7182 struct dl *dl;
7183 unsigned long long pos = 0;
7184 unsigned long long maxsize;
7185 struct extent *e;
7186 int i;
cdddbdbc 7187
88c32bb1
DW		 7188 /* We must have the container info already read in. */
7189 if (!super)
c2c087e6
DW		 7190 return 0;
7191
b2916f25
JS		 7192 mpb = super->anchor;
7193
2cc699af 7194 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7195 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7196 return 0;
d54559f0 7197 }
c2c087e6
DW		 7198 if (!dev) {
7199 /* General test: make sure there is space for
2da8544a
DW		 7200 * 'raiddisks' device extents of size 'size' at a given
7201 * offset
c2c087e6 7202 */
e46273eb 7203 unsigned long long minsize = size;
b7528a20 7204 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7205 int dcnt = 0;
7206 if (minsize == 0)
7207 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7208 for (dl = super->disks; dl ; dl = dl->next) {
7209 int found = 0;
7210
bf5a934a 7211 pos = 0;
c2c087e6 7212 i = 0;
05501181 7213 e = get_extents(super, dl, 0);
c2c087e6
DW		 7214 if (!e) continue;
7215 do {
7216 unsigned long long esize;
7217 esize = e[i].start - pos;
7218 if (esize >= minsize)
7219 found = 1;
b7528a20 7220 if (found && start_offset == MaxSector) {
2da8544a
DW		 7221 start_offset = pos;
7222 break;
7223 } else if (found && pos != start_offset) {
7224 found = 0;
7225 break;
7226 }
c2c087e6
DW		 7227 pos = e[i].start + e[i].size;
7228 i++;
7229 } while (e[i-1].size);
7230 if (found)
7231 dcnt++;
7232 free(e);
7233 }
7234 if (dcnt < raiddisks) {
2c514b71 7235 if (verbose)
7a862a02 7236 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7237 dcnt, raiddisks);
c2c087e6
DW		 7238 return 0;
7239 }
7240 return 1;
7241 }
0dcecb2e 7242
c2c087e6 7243 /* This device must be a member of the set */
9e04ac1c 7244 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7245 return 0;
7246 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7247 if (dl->major == (int)major(rdev) &&
7248 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7249 break;
7250 }
7251 if (!dl) {
2c514b71 7252 if (verbose)
7a862a02 7253 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7254 return 0;
a20d2ba5
DW		 7255 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7256 /* If a volume is present then the current creation attempt
7257 * cannot incorporate new spares because the orom may not
7258 * understand this configuration (all member disks must be
7259 * members of each array in the container).
7260 */
7a862a02
N		 7261 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7262 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7263 return 0;
5fe62b94
WD		 7264 } else if (super->orom && mpb->num_raid_devs > 0 &&
7265 mpb->num_disks != raiddisks) {
7a862a02 7266 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7267 return 0;
c2c087e6 7268 }
0dcecb2e
DW		 7269
7270 /* retrieve the largest free space block */
05501181 7271 e = get_extents(super, dl, 0);
c2c087e6
DW		 7272 maxsize = 0;
7273 i = 0;
0dcecb2e
DW		 7274 if (e) {
7275 do {
7276 unsigned long long esize;
7277
7278 esize = e[i].start - pos;
7279 if (esize >= maxsize)
7280 maxsize = esize;
7281 pos = e[i].start + e[i].size;
7282 i++;
7283 } while (e[i-1].size);
7284 dl->e = e;
7285 dl->extent_cnt = i;
7286 } else {
7287 if (verbose)
e7b84f9d 7288 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7289 dev);
7290 return 0;
7291 }
7292 if (maxsize < size) {
7293 if (verbose)
e7b84f9d 7294 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7295 dev, maxsize, size);
7296 return 0;
7297 }
7298
7299 /* count total number of extents for merge */
7300 i = 0;
7301 for (dl = super->disks; dl; dl = dl->next)
7302 if (dl->e)
7303 i += dl->extent_cnt;
7304
7305 maxsize = merge_extents(super, i);
3baa56ab 7306
1a1ced1e
KS		 7307 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7308 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7309
a7dd165b 7310 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7311 if (verbose) {
7312 if (maxsize == 0)
7a862a02 7313 pr_err("no free space left on device. Aborting...\n");
b3071342 7314 else
7a862a02 7315 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7316 maxsize, size);
7317 }
0dcecb2e 7318 return 0;
0dcecb2e
DW		 7319 }
7320
c2c087e6
DW		 7321 *freesize = maxsize;
7322
ca9de185 7323 if (super->orom) {
72a45777 7324 int count = count_volumes(super->hba,
ca9de185
LM		 7325 super->orom->dpa, verbose);
7326 if (super->orom->vphba <= count) {
676e87a8 7327 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7328 super->orom->vphba);
7329 return 0;
7330 }
7331 }
c2c087e6 7332 return 1;
cdddbdbc
DW		 7333}
7334
13bcac90 7335static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7336 unsigned long long size, int chunk,
7337 unsigned long long *freesize)
7338{
7339 struct intel_super *super = st->sb;
7340 struct imsm_super *mpb = super->anchor;
7341 struct dl *dl;
7342 int i;
7343 int extent_cnt;
7344 struct extent *e;
7345 unsigned long long maxsize;
7346 unsigned long long minsize;
7347 int cnt;
7348 int used;
7349
7350 /* find the largest common start free region of the possible disks */
7351 used = 0;
7352 extent_cnt = 0;
7353 cnt = 0;
7354 for (dl = super->disks; dl; dl = dl->next) {
7355 dl->raiddisk = -1;
7356
7357 if (dl->index >= 0)
7358 used++;
7359
7360 /* don't activate new spares if we are orom constrained
7361 * and there is already a volume active in the container
7362 */
7363 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7364 continue;
7365
05501181 7366 e = get_extents(super, dl, 0);
efb30e7f
DW		 7367 if (!e)
7368 continue;
7369 for (i = 1; e[i-1].size; i++)
7370 ;
7371 dl->e = e;
7372 dl->extent_cnt = i;
7373 extent_cnt += i;
7374 cnt++;
7375 }
7376
7377 maxsize = merge_extents(super, extent_cnt);
7378 minsize = size;
7379 if (size == 0)
612e59d8
CA		 7380 /* chunk is in K */
7381 minsize = chunk * 2;
efb30e7f
DW		 7382
7383 if (cnt < raiddisks ||
7384 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7385 maxsize < minsize ||
7386 maxsize == 0) {
e7b84f9d 7387 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7388 return 0; /* No enough free spaces large enough */
7389 }
7390
7391 if (size == 0) {
7392 size = maxsize;
7393 if (chunk) {
612e59d8
CA		 7394 size /= 2 * chunk;
7395 size *= 2 * chunk;
efb30e7f 7396 }
f878b242
LM		 7397 maxsize = size;
7398 }
1a1ced1e
KS		 7399 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7400 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7401 cnt = 0;
7402 for (dl = super->disks; dl; dl = dl->next)
7403 if (dl->e)
7404 dl->raiddisk = cnt++;
7405
7406 *freesize = size;
7407
13bcac90
AK		 7408 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7409
efb30e7f
DW		 7410 return 1;
7411}
7412
13bcac90
AK		 7413static int reserve_space(struct supertype *st, int raiddisks,
7414 unsigned long long size, int chunk,
7415 unsigned long long *freesize)
7416{
7417 struct intel_super *super = st->sb;
7418 struct dl *dl;
7419 int cnt;
7420 int rv = 0;
7421
7422 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7423 if (rv) {
7424 cnt = 0;
7425 for (dl = super->disks; dl; dl = dl->next)
7426 if (dl->e)
7427 dl->raiddisk = cnt++;
7428 rv = 1;
7429 }
7430
7431 return rv;
7432}
7433
bf5a934a 7434static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7435 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7436 unsigned long long data_offset,
bf5a934a 7437 char *dev, unsigned long long *freesize,
5308f117 7438 int consistency_policy, int verbose)
bf5a934a
DW		 7439{
7440 int fd, cfd;
7441 struct mdinfo *sra;
20cbe8d2 7442 int is_member = 0;
bf5a934a 7443
d54559f0
LM		 7444 /* load capability
7445 * if given unused devices create a container
bf5a934a
DW		 7446 * if given given devices in a container create a member volume
7447 */
7448 if (level == LEVEL_CONTAINER) {
7449 /* Must be a fresh device to add to a container */
7450 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7451 raiddisks,
7ccc4cc4 7452 *chunk,
af4348dd 7453 size, data_offset,
bf5a934a
DW		 7454 dev, freesize,
7455 verbose);
7456 }
9587c373 7457
06a6101c
BK		 7458 /*
7459 * Size is given in sectors.
7460 */
7461 if (size && (size < 2048)) {
22dc741f 7462 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7463 /* Depends on algorithm in Create.c :
7464 * if container was given (dev == NULL) return -1,
7465 * if block device was given ( dev != NULL) return 0.
7466 */
7467 return dev ? -1 : 0;
7468 }
7469
8592f29d 7470 if (!dev) {
e91a3bad 7471 if (st->sb) {
ca9de185 7472 struct intel_super *super = st->sb;
e91a3bad 7473 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7474 raiddisks, chunk, size,
e91a3bad
LM		 7475 verbose))
7476 return 0;
efb30e7f
DW		 7477 /* we are being asked to automatically layout a
7478 * new volume based on the current contents of
7479 * the container. If the the parameters can be
7480 * satisfied reserve_space will record the disks,
7481 * start offset, and size of the volume to be
7482 * created. add_to_super and getinfo_super
7483 * detect when autolayout is in progress.
7484 */
ca9de185
LM		 7485 /* assuming that freesize is always given when array is
7486 created */
7487 if (super->orom && freesize) {
7488 int count;
72a45777 7489 count = count_volumes(super->hba,
ca9de185
LM		 7490 super->orom->dpa, verbose);
7491 if (super->orom->vphba <= count) {
676e87a8 7492 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7493 super->orom->vphba);
7494 return 0;
7495 }
7496 }
e91a3bad
LM		 7497 if (freesize)
7498 return reserve_space(st, raiddisks, size,
7ccc4cc4 7499 *chunk, freesize);
8592f29d
N		 7500 }
7501 return 1;
7502 }
bf5a934a
DW		 7503 if (st->sb) {
7504 /* creating in a given container */
7505 return validate_geometry_imsm_volume(st, level, layout,
7506 raiddisks, chunk, size,
af4348dd 7507 data_offset,
bf5a934a
DW		 7508 dev, freesize, verbose);
7509 }
7510
bf5a934a
DW		 7511 /* This device needs to be a device in an 'imsm' container */
7512 fd = open(dev, O_RDONLY|O_EXCL, 0);
7513 if (fd >= 0) {
7514 if (verbose)
e7b84f9d
N		 7515 pr_err("Cannot create this array on device %s\n",
7516 dev);
bf5a934a
DW		 7517 close(fd);
7518 return 0;
7519 }
7520 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7521 if (verbose)
e7b84f9d 7522 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7523 dev, strerror(errno));
7524 return 0;
7525 }
7526 /* Well, it is in use by someone, maybe an 'imsm' container. */
7527 cfd = open_container(fd);
20cbe8d2 7528 close(fd);
bf5a934a 7529 if (cfd < 0) {
bf5a934a 7530 if (verbose)
e7b84f9d 7531 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7532 dev);
7533 return 0;
7534 }
4dd2df09 7535 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7536 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7537 strcmp(sra->text_version, "imsm") == 0)
7538 is_member = 1;
7539 sysfs_free(sra);
7540 if (is_member) {
bf5a934a
DW		 7541 /* This is a member of a imsm container. Load the container
7542 * and try to create a volume
7543 */
7544 struct intel_super *super;
7545
ec50f7b6 7546 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7547 st->sb = super;
4dd2df09 7548 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7549 close(cfd);
7550 return validate_geometry_imsm_volume(st, level, layout,
7551 raiddisks, chunk,
af4348dd 7552 size, data_offset, dev,
ecbd9e81
N		 7553 freesize, 1)
7554 ? 1 : -1;
bf5a934a 7555 }
20cbe8d2 7556 }
bf5a934a 7557
20cbe8d2 7558 if (verbose)
e7b84f9d 7559 pr_err("failed container membership check\n");
20cbe8d2
AW		 7560
7561 close(cfd);
7562 return 0;
bf5a934a 7563}
0bd16cf2 7564
30f58b22 7565static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7566{
7567 struct intel_super *super = st->sb;
7568
30f58b22
DW		 7569 if (level && *level == UnSet)
7570 *level = LEVEL_CONTAINER;
7571
7572 if (level && layout && *layout == UnSet)
7573 *layout = imsm_level_to_layout(*level);
0bd16cf2 7574
cd9d1ac7
DW		 7575 if (chunk && (*chunk == UnSet || *chunk == 0))
7576 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7577}
7578
33414a01
DW		 7579static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7580
3364781b 7581static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7582{
3364781b 7583 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7584 __u8 i;
7585 struct intel_dev **dp;
7586 struct intel_super *super = st->sb;
3364781b 7587 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7588 struct imsm_super *mpb = super->anchor;
7589
3364781b 7590 if (mpb->num_raid_devs == 0)
33414a01 7591 return 2;
33414a01
DW		 7592
7593 /* block deletions that would change the uuid of active subarrays
7594 *
7595 * FIXME when immutable ids are available, but note that we'll
7596 * also need to fixup the invalidated/active subarray indexes in
7597 * mdstat
7598 */
7599 for (i = 0; i < mpb->num_raid_devs; i++) {
7600 char subarray[4];
7601
7602 if (i < current_vol)
7603 continue;
7604 sprintf(subarray, "%u", i);
4dd2df09 7605 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7606 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7607 current_vol, i);
33414a01
DW		 7608
7609 return 2;
7610 }
7611 }
7612
7613 if (st->update_tail) {
503975b9 7614 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7615
33414a01
DW		 7616 u->type = update_kill_array;
7617 u->dev_idx = current_vol;
7618 append_metadata_update(st, u, sizeof(*u));
7619
7620 return 0;
7621 }
7622
7623 for (dp = &super->devlist; *dp;)
7624 if ((*dp)->index == current_vol) {
7625 *dp = (*dp)->next;
7626 } else {
7627 handle_missing(super, (*dp)->dev);
7628 if ((*dp)->index > current_vol)
7629 (*dp)->index--;
7630 dp = &(*dp)->next;
7631 }
7632
7633 /* no more raid devices, all active components are now spares,
7634 * but of course failed are still failed
7635 */
7636 if (--mpb->num_raid_devs == 0) {
7637 struct dl *d;
7638
7639 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7640 if (d->index > -2)
7641 mark_spare(d);
33414a01
DW		 7642 }
7643
7644 super->updates_pending++;
7645
7646 return 0;
7647}
aa534678 7648
a951a4f7 7649static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7650 char *update, struct mddev_ident *ident)
aa534678
DW		 7651{
7652 /* update the subarray currently referenced by ->current_vol */
7653 struct intel_super *super = st->sb;
7654 struct imsm_super *mpb = super->anchor;
7655
aa534678
DW		 7656 if (strcmp(update, "name") == 0) {
7657 char *name = ident->name;
a951a4f7
N		 7658 char *ep;
7659 int vol;
aa534678 7660
4dd2df09 7661 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7662 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7663 return 2;
7664 }
7665
7666 if (!check_name(super, name, 0))
7667 return 2;
7668
a951a4f7
N		 7669 vol = strtoul(subarray, &ep, 10);
7670 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7671 return 2;
7672
aa534678 7673 if (st->update_tail) {
503975b9 7674 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7675
aa534678 7676 u->type = update_rename_array;
a951a4f7 7677 u->dev_idx = vol;
618f4e6d
XN		 7678 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7679 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7680 append_metadata_update(st, u, sizeof(*u));
7681 } else {
7682 struct imsm_dev *dev;
ebad3af2 7683 int i, namelen;
aa534678 7684
a951a4f7 7685 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7686 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7687 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7688 memcpy(dev->volume, name, namelen);
aa534678
DW		 7689 for (i = 0; i < mpb->num_raid_devs; i++) {
7690 dev = get_imsm_dev(super, i);
7691 handle_missing(super, dev);
7692 }
7693 super->updates_pending++;
7694 }
e6e9dd3f
AP		 7695 } else if (strcmp(update, "ppl") == 0 ||
7696 strcmp(update, "no-ppl") == 0) {
7697 int new_policy;
7698 char *ep;
7699 int vol = strtoul(subarray, &ep, 10);
7700
7701 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7702 return 2;
7703
7704 if (strcmp(update, "ppl") == 0)
c2462068 7705 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7706 else
c2462068 7707 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7708
7709 if (st->update_tail) {
7710 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7711
7712 u->type = update_rwh_policy;
7713 u->dev_idx = vol;
7714 u->new_policy = new_policy;
7715 append_metadata_update(st, u, sizeof(*u));
7716 } else {
7717 struct imsm_dev *dev;
7718
7719 dev = get_imsm_dev(super, vol);
7720 dev->rwh_policy = new_policy;
7721 super->updates_pending++;
7722 }
aa534678
DW		 7723 } else
7724 return 2;
7725
7726 return 0;
7727}
bf5a934a 7728
28bce06f
AK		 7729static int is_gen_migration(struct imsm_dev *dev)
7730{
7534230b
AK		 7731 if (dev == NULL)
7732 return 0;
7733
28bce06f
AK		 7734 if (!dev->vol.migr_state)
7735 return 0;
7736
7737 if (migr_type(dev) == MIGR_GEN_MIGR)
7738 return 1;
7739
7740 return 0;
7741}
7742
1e5c6983
DW		 7743static int is_rebuilding(struct imsm_dev *dev)
7744{
7745 struct imsm_map *migr_map;
7746
7747 if (!dev->vol.migr_state)
7748 return 0;
7749
7750 if (migr_type(dev) != MIGR_REBUILD)
7751 return 0;
7752
238c0a71 7753 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7754
7755 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7756 return 1;
7757 else
7758 return 0;
7759}
7760
6ce1fbf1
AK		 7761static int is_initializing(struct imsm_dev *dev)
7762{
7763 struct imsm_map *migr_map;
7764
7765 if (!dev->vol.migr_state)
7766 return 0;
7767
7768 if (migr_type(dev) != MIGR_INIT)
7769 return 0;
7770
238c0a71 7771 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7772
7773 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7774 return 1;
7775
7776 return 0;
6ce1fbf1
AK		 7777}
7778
c47b0ff6
AK		 7779static void update_recovery_start(struct intel_super *super,
7780 struct imsm_dev *dev,
7781 struct mdinfo *array)
1e5c6983
DW		 7782{
7783 struct mdinfo *rebuild = NULL;
7784 struct mdinfo *d;
7785 __u32 units;
7786
7787 if (!is_rebuilding(dev))
7788 return;
7789
7790 /* Find the rebuild target, but punt on the dual rebuild case */
7791 for (d = array->devs; d; d = d->next)
7792 if (d->recovery_start == 0) {
7793 if (rebuild)
7794 return;
7795 rebuild = d;
7796 }
7797
4363fd80
DW		 7798 if (!rebuild) {
7799 /* (?) none of the disks are marked with
7800 * IMSM_ORD_REBUILD, so assume they are missing and the
7801 * disk_ord_tbl was not correctly updated
7802 */
1ade5cc1 7803 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7804 return;
7805 }
7806
1e5c6983 7807 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7808 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7809}
7810
276d77db 7811static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7812
00bbdbda 7813static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7814{
4f5bc454
DW		 7815 /* Given a container loaded by load_super_imsm_all,
7816 * extract information about all the arrays into
7817 * an mdinfo tree.
00bbdbda 7818 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7819 *
7820 * For each imsm_dev create an mdinfo, fill it in,
7821 * then look for matching devices in super->disks
7822 * and create appropriate device mdinfo.
7823 */
7824 struct intel_super *super = st->sb;
949c47a0 7825 struct imsm_super *mpb = super->anchor;
4f5bc454 7826 struct mdinfo *rest = NULL;
00bbdbda 7827 unsigned int i;
81219e70 7828 int sb_errors = 0;
abef11a3
AK		 7829 struct dl *d;
7830 int spare_disks = 0;
b6180160 7831 int current_vol = super->current_vol;
cdddbdbc 7832
19482bcc
AK		 7833 /* do not assemble arrays when not all attributes are supported */
7834 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7835 sb_errors = 1;
7a862a02 7836 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7837 }
7838
abef11a3
AK		 7839 /* count spare devices, not used in maps
7840 */
7841 for (d = super->disks; d; d = d->next)
7842 if (d->index == -1)
7843 spare_disks++;
7844
4f5bc454 7845 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7846 struct imsm_dev *dev;
7847 struct imsm_map *map;
86e3692b 7848 struct imsm_map *map2;
4f5bc454 7849 struct mdinfo *this;
a6482415 7850 int slot;
a6482415 7851 int chunk;
00bbdbda 7852 char *ep;
8b9cd157 7853 int level;
00bbdbda
N		 7854
7855 if (subarray &&
7856 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7857 continue;
7858
7859 dev = get_imsm_dev(super, i);
238c0a71
AK		 7860 map = get_imsm_map(dev, MAP_0);
7861 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7862 level = get_imsm_raid_level(map);
4f5bc454 7863
1ce0101c
DW		 7864 /* do not publish arrays that are in the middle of an
7865 * unsupported migration
7866 */
7867 if (dev->vol.migr_state &&
28bce06f 7868 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7869 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7870 dev->volume);
7871 continue;
7872 }
2db86302
LM		 7873 /* do not publish arrays that are not support by controller's
7874 * OROM/EFI
7875 */
1ce0101c 7876
503975b9 7877 this = xmalloc(sizeof(*this));
4f5bc454 7878
301406c9 7879 super->current_vol = i;
a5d85af7 7880 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7881 this->next = rest;
a6482415 7882 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7883 /* mdadm does not support all metadata features- set the bit in all arrays state */
7884 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7885 level, /* RAID level */
7886 imsm_level_to_layout(level),
81219e70 7887 map->num_members, /* raid disks */
fcc2c9da 7888 &chunk, imsm_dev_size(dev),
81219e70 7889 1 /* verbose */)) {
7a862a02 7890 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7891 dev->volume);
7892 this->array.state |=
7893 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7894 (1<<MD_SB_BLOCK_VOLUME);
7895 }
81219e70
LM		 7896
7897 /* if array has bad blocks, set suitable bit in all arrays state */
7898 if (sb_errors)
7899 this->array.state |=
7900 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7901 (1<<MD_SB_BLOCK_VOLUME);
7902
4f5bc454 7903 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7904 unsigned long long recovery_start;
4f5bc454
DW		 7905 struct mdinfo *info_d;
7906 struct dl *d;
7907 int idx;
9a1608e5 7908 int skip;
7eef0453 7909 __u32 ord;
8b9cd157 7910 int missing = 0;
4f5bc454 7911
9a1608e5 7912 skip = 0;
238c0a71
AK		 7913 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7914 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7915 for (d = super->disks; d ; d = d->next)
7916 if (d->index == idx)
0fbd635c 7917 break;
4f5bc454 7918
1e5c6983 7919 recovery_start = MaxSector;
4f5bc454 7920 if (d == NULL)
9a1608e5 7921 skip = 1;
25ed7e59 7922 if (d && is_failed(&d->disk))
9a1608e5 7923 skip = 1;
8b9cd157 7924 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7925 recovery_start = 0;
1e93d0d1
BK		 7926 if (!(ord & IMSM_ORD_REBUILD))
7927 this->array.working_disks++;
1011e834 7928 /*
9a1608e5 7929 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7930 * reset resync start to avoid a dirty-degraded
7931 * situation when performing the intial sync
9a1608e5 7932 */
8b9cd157
MK		 7933 if (skip)
7934 missing++;
7935
7936 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7937 if ((!able_to_resync(level, missing) ||
7938 recovery_start == 0))
7939 this->resync_start = MaxSector;
7940 } else {
7941 /*
7942 * FIXME handle dirty degraded
7943 */
7944 }
7945
9a1608e5
DW		 7946 if (skip)
7947 continue;
4f5bc454 7948
503975b9 7949 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7950 info_d->next = this->devs;
7951 this->devs = info_d;
7952
4f5bc454
DW		 7953 info_d->disk.number = d->index;
7954 info_d->disk.major = d->major;
7955 info_d->disk.minor = d->minor;
7956 info_d->disk.raid_disk = slot;
1e5c6983 7957 info_d->recovery_start = recovery_start;
86e3692b
AK		 7958 if (map2) {
7959 if (slot < map2->num_members)
7960 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7961 else
7962 this->array.spare_disks++;
86e3692b
AK		 7963 } else {
7964 if (slot < map->num_members)
7965 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7966 else
7967 this->array.spare_disks++;
86e3692b 7968 }
4f5bc454
DW		 7969
7970 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7971 info_d->data_offset = pba_of_lba0(map);
44490938 7972 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7973
7974 if (map->raid_level == 5) {
2432ce9b
AP		 7975 info_d->ppl_sector = this->ppl_sector;
7976 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7977 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7978 recovery_start == 0)
7979 this->resync_start = 0;
06fb291a 7980 }
b12796be 7981
5e46202e 7982 info_d->bb.supported = 1;
b12796be
TM		 7983 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7984 info_d->data_offset,
7985 info_d->component_size,
7986 &info_d->bb);
4f5bc454 7987 }
1e5c6983 7988 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7989 update_recovery_start(super, dev, this);
abef11a3 7990 this->array.spare_disks += spare_disks;
276d77db
AK		 7991
7992 /* check for reshape */
7993 if (this->reshape_active == 1)
7994 recover_backup_imsm(st, this);
9a1608e5 7995 rest = this;
4f5bc454
DW		 7996 }
7997
b6180160 7998 super->current_vol = current_vol;
4f5bc454 7999 return rest;
cdddbdbc
DW		 8000}
8001
3b451610
AK		 8002static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8003 int failed, int look_in_map)
c2a1e7da 8004{
3b451610
AK		 8005 struct imsm_map *map;
8006
8007 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8008
8009 if (!failed)
1011e834 8010 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8011 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8012
8013 switch (get_imsm_raid_level(map)) {
8014 case 0:
8015 return IMSM_T_STATE_FAILED;
8016 break;
8017 case 1:
8018 if (failed < map->num_members)
8019 return IMSM_T_STATE_DEGRADED;
8020 else
8021 return IMSM_T_STATE_FAILED;
8022 break;
8023 case 10:
8024 {
8025 /**
c92a2527
DW		 8026 * check to see if any mirrors have failed, otherwise we
8027 * are degraded. Even numbered slots are mirrored on
8028 * slot+1
c2a1e7da 8029 */
c2a1e7da 8030 int i;
d9b420a5
N		 8031 /* gcc -Os complains that this is unused */
8032 int insync = insync;
c2a1e7da
DW		 8033
8034 for (i = 0; i < map->num_members; i++) {
238c0a71 8035 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8036 int idx = ord_to_idx(ord);
8037 struct imsm_disk *disk;
c2a1e7da 8038
c92a2527 8039 /* reset the potential in-sync count on even-numbered
1011e834 8040 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8041 */
8042 if ((i & 1) == 0)
8043 insync = 2;
c2a1e7da 8044
c92a2527 8045 disk = get_imsm_disk(super, idx);
25ed7e59 8046 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8047 insync--;
c2a1e7da 8048
c92a2527
DW		 8049 /* no in-sync disks left in this mirror the
8050 * array has failed
8051 */
8052 if (insync == 0)
8053 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8054 }
8055
8056 return IMSM_T_STATE_DEGRADED;
8057 }
8058 case 5:
8059 if (failed < 2)
8060 return IMSM_T_STATE_DEGRADED;
8061 else
8062 return IMSM_T_STATE_FAILED;
8063 break;
8064 default:
8065 break;
8066 }
8067
8068 return map->map_state;
8069}
8070
3b451610
AK		 8071static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8072 int look_in_map)
c2a1e7da
DW		 8073{
8074 int i;
8075 int failed = 0;
8076 struct imsm_disk *disk;
d5985138
AK		 8077 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8078 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8079 struct imsm_map *map_for_loop;
0556e1a2
DW		 8080 __u32 ord;
8081 int idx;
d5985138 8082 int idx_1;
c2a1e7da 8083
0556e1a2
DW		 8084 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8085 * disks that are being rebuilt. New failures are recorded to
8086 * map[0]. So we look through all the disks we started with and
8087 * see if any failures are still present, or if any new ones
8088 * have arrived
0556e1a2 8089 */
d5985138
AK		 8090 map_for_loop = map;
8091 if (prev && (map->num_members < prev->num_members))
8092 map_for_loop = prev;
68fe4598
LD		 8093
8094 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8095 idx_1 = -255;
238c0a71
AK		 8096 /* when MAP_X is passed both maps failures are counted
8097 */
d5985138 8098 if (prev &&
089f9d79
JS		 8099 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8100 i < prev->num_members) {
d5985138
AK		 8101 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8102 idx_1 = ord_to_idx(ord);
c2a1e7da 8103
d5985138
AK		 8104 disk = get_imsm_disk(super, idx_1);
8105 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8106 failed++;
8107 }
089f9d79
JS		 8108 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8109 i < map->num_members) {
d5985138
AK		 8110 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8111 idx = ord_to_idx(ord);
8112
8113 if (idx != idx_1) {
8114 disk = get_imsm_disk(super, idx);
8115 if (!disk || is_failed(disk) ||
8116 ord & IMSM_ORD_REBUILD)
8117 failed++;
8118 }
8119 }
c2a1e7da
DW		 8120 }
8121
8122 return failed;
845dea95
NB		 8123}
8124
97b4d0e9
DW		 8125static int imsm_open_new(struct supertype *c, struct active_array *a,
8126 char *inst)
8127{
8128 struct intel_super *super = c->sb;
8129 struct imsm_super *mpb = super->anchor;
bbab0940 8130 struct imsm_update_prealloc_bb_mem u;
9587c373 8131
97b4d0e9 8132 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8133 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8134 return -ENODEV;
8135 }
8136
8137 dprintf("imsm: open_new %s\n", inst);
8138 a->info.container_member = atoi(inst);
bbab0940
TM		 8139
8140 u.type = update_prealloc_badblocks_mem;
8141 imsm_update_metadata_locally(c, &u, sizeof(u));
8142
97b4d0e9
DW		 8143 return 0;
8144}
8145
0c046afd
DW		 8146static int is_resyncing(struct imsm_dev *dev)
8147{
8148 struct imsm_map *migr_map;
8149
8150 if (!dev->vol.migr_state)
8151 return 0;
8152
1484e727
DW		 8153 if (migr_type(dev) == MIGR_INIT ||
8154 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8155 return 1;
8156
4c9bc37b
AK		 8157 if (migr_type(dev) == MIGR_GEN_MIGR)
8158 return 0;
8159
238c0a71 8160 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8161
089f9d79
JS		 8162 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8163 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8164 return 1;
8165 else
8166 return 0;
8167}
8168
0556e1a2 8169/* return true if we recorded new information */
4c9e8c1e
TM		 8170static int mark_failure(struct intel_super *super,
8171 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8172{
0556e1a2
DW		 8173 __u32 ord;
8174 int slot;
8175 struct imsm_map *map;
86c54047
DW		 8176 char buf[MAX_RAID_SERIAL_LEN+3];
8177 unsigned int len, shift = 0;
0556e1a2
DW		 8178
8179 /* new failures are always set in map[0] */
238c0a71 8180 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8181
8182 slot = get_imsm_disk_slot(map, idx);
8183 if (slot < 0)
8184 return 0;
8185
8186 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8187 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8188 return 0;
8189
7d0c5e24
LD		 8190 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8191 buf[MAX_RAID_SERIAL_LEN] = '\000';
8192 strcat(buf, ":0");
86c54047
DW		 8193 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8194 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8195 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8196
f2f27e63 8197 disk->status |= FAILED_DISK;
0556e1a2 8198 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8199 /* mark failures in second map if second map exists and this disk
8200 * in this slot.
8201 * This is valid for migration, initialization and rebuild
8202 */
8203 if (dev->vol.migr_state) {
238c0a71 8204 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8205 int slot2 = get_imsm_disk_slot(map2, idx);
8206
089f9d79 8207 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8208 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8209 idx | IMSM_ORD_REBUILD);
8210 }
d7a1fda2
MT		 8211 if (map->failed_disk_num == 0xff ||
8212 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8213 map->failed_disk_num = slot;
4c9e8c1e
TM		 8214
8215 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8216
0556e1a2
DW		 8217 return 1;
8218}
8219
4c9e8c1e
TM		 8220static void mark_missing(struct intel_super *super,
8221 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8222{
4c9e8c1e 8223 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8224
8225 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8226 return;
8227
47ee5a45
DW		 8228 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8229 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8230}
8231
33414a01
DW		 8232static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8233{
33414a01 8234 struct dl *dl;
33414a01
DW		 8235
8236 if (!super->missing)
8237 return;
33414a01 8238
79b68f1b
PC		 8239 /* When orom adds replacement for missing disk it does
8240 * not remove entry of missing disk, but just updates map with
8241 * new added disk. So it is not enough just to test if there is
8242 * any missing disk, we have to look if there are any failed disks
8243 * in map to stop migration */
8244
33414a01 8245 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8246 /* end process for initialization and rebuild only
8247 */
8248 if (is_gen_migration(dev) == 0) {
fb12a745 8249 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8250
fb12a745
TM		 8251 if (failed) {
8252 __u8 map_state;
8253 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8254 struct imsm_map *map1;
8255 int i, ord, ord_map1;
8256 int rebuilt = 1;
3d59f0c0 8257
fb12a745
TM		 8258 for (i = 0; i < map->num_members; i++) {
8259 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8260 if (!(ord & IMSM_ORD_REBUILD))
8261 continue;
8262
8263 map1 = get_imsm_map(dev, MAP_1);
8264 if (!map1)
8265 continue;
8266
8267 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8268 if (ord_map1 & IMSM_ORD_REBUILD)
8269 rebuilt = 0;
8270 }
8271
8272 if (rebuilt) {
8273 map_state = imsm_check_degraded(super, dev,
8274 failed, MAP_0);
8275 end_migration(dev, super, map_state);
8276 }
8277 }
3d59f0c0 8278 }
33414a01 8279 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8280 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8281 super->updates_pending++;
8282}
8283
f3871fdc
AK		 8284static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8285 long long new_size)
70bdf0dc 8286{
70bdf0dc 8287 unsigned long long array_blocks;
9529d343
MD		 8288 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8289 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8290
8291 if (used_disks == 0) {
8292 /* when problems occures
8293 * return current array_blocks value
8294 */
fcc2c9da 8295 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8296
8297 return array_blocks;
8298 }
8299
8300 /* set array size in metadata
8301 */
9529d343 8302 if (new_size <= 0)
f3871fdc
AK		 8303 /* OLCE size change is caused by added disks
8304 */
44490938 8305 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8306 else
f3871fdc
AK		 8307 /* Online Volume Size Change
8308 * Using available free space
8309 */
8310 array_blocks = new_size;
70bdf0dc 8311
b53bfba6 8312 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8313 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8314
8315 return array_blocks;
8316}
8317
28bce06f
AK		 8318static void imsm_set_disk(struct active_array *a, int n, int state);
8319
0e2d1a4e
AK		 8320static void imsm_progress_container_reshape(struct intel_super *super)
8321{
8322 /* if no device has a migr_state, but some device has a
8323 * different number of members than the previous device, start
8324 * changing the number of devices in this device to match
8325 * previous.
8326 */
8327 struct imsm_super *mpb = super->anchor;
8328 int prev_disks = -1;
8329 int i;
1dfaa380 8330 int copy_map_size;
0e2d1a4e
AK		 8331
8332 for (i = 0; i < mpb->num_raid_devs; i++) {
8333 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8334 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8335 struct imsm_map *map2;
8336 int prev_num_members;
0e2d1a4e
AK		 8337
8338 if (dev->vol.migr_state)
8339 return;
8340
8341 if (prev_disks == -1)
8342 prev_disks = map->num_members;
8343 if (prev_disks == map->num_members)
8344 continue;
8345
8346 /* OK, this array needs to enter reshape mode.
8347 * i.e it needs a migr_state
8348 */
8349
1dfaa380 8350 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8351 prev_num_members = map->num_members;
8352 map->num_members = prev_disks;
8353 dev->vol.migr_state = 1;
8354 dev->vol.curr_migr_unit = 0;
ea672ee1 8355 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8356 for (i = prev_num_members;
8357 i < map->num_members; i++)
8358 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8359 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8360 /* Copy the current map */
1dfaa380 8361 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8362 map2->num_members = prev_num_members;
8363
f3871fdc 8364 imsm_set_array_size(dev, -1);
51d83f5d 8365 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8366 super->updates_pending++;
8367 }
8368}
8369
aad6f216 8370/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8371 * states are handled in imsm_set_disk() with one exception, when a
8372 * resync is stopped due to a new failure this routine will set the
8373 * 'degraded' state for the array.
8374 */
01f157d7 8375static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8376{
8377 int inst = a->info.container_member;
8378 struct intel_super *super = a->container->sb;
949c47a0 8379 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8380 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8381 int failed = imsm_count_failed(super, dev, MAP_0);
8382 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8383 __u32 blocks_per_unit;
a862209d 8384
1af97990
AK		 8385 if (dev->vol.migr_state &&
8386 dev->vol.migr_type == MIGR_GEN_MIGR) {
8387 /* array state change is blocked due to reshape action
aad6f216
N		 8388 * We might need to
8389 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8390 * - finish the reshape (if last_checkpoint is big and action != reshape)
8391 * - update curr_migr_unit
1af97990 8392 */
aad6f216
N		 8393 if (a->curr_action == reshape) {
8394 /* still reshaping, maybe update curr_migr_unit */
633b5610 8395 goto mark_checkpoint;
aad6f216
N		 8396 } else {
8397 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8398 /* for some reason we aborted the reshape.
b66e591b
AK		 8399 *
8400 * disable automatic metadata rollback
8401 * user action is required to recover process
aad6f216 8402 */
b66e591b 8403 if (0) {
238c0a71
AK		 8404 struct imsm_map *map2 =
8405 get_imsm_map(dev, MAP_1);
8406 dev->vol.migr_state = 0;
8407 set_migr_type(dev, 0);
8408 dev->vol.curr_migr_unit = 0;
8409 memcpy(map, map2,
8410 sizeof_imsm_map(map2));
8411 super->updates_pending++;
b66e591b 8412 }
aad6f216
N		 8413 }
8414 if (a->last_checkpoint >= a->info.component_size) {
8415 unsigned long long array_blocks;
8416 int used_disks;
e154ced3 8417 struct mdinfo *mdi;
aad6f216 8418
9529d343 8419 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8420 if (used_disks > 0) {
8421 array_blocks =
44490938 8422 per_dev_array_size(map) *
d55adef9 8423 used_disks;
b53bfba6
TM		 8424 array_blocks =
8425 round_size_to_mb(array_blocks,
8426 used_disks);
d55adef9
AK		 8427 a->info.custom_array_size = array_blocks;
8428 /* encourage manager to update array
8429 * size
8430 */
e154ced3 8431
d55adef9 8432 a->check_reshape = 1;
633b5610 8433 }
e154ced3
AK		 8434 /* finalize online capacity expansion/reshape */
8435 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8436 imsm_set_disk(a,
8437 mdi->disk.raid_disk,
8438 mdi->curr_state);
8439
0e2d1a4e 8440 imsm_progress_container_reshape(super);
e154ced3 8441 }
aad6f216 8442 }
1af97990
AK		 8443 }
8444
47ee5a45 8445 /* before we activate this array handle any missing disks */
33414a01
DW		 8446 if (consistent == 2)
8447 handle_missing(super, dev);
1e5c6983 8448
0c046afd 8449 if (consistent == 2 &&
b7941fd6 8450 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8451 map_state != IMSM_T_STATE_NORMAL ||
8452 dev->vol.migr_state))
01f157d7 8453 consistent = 0;
272906ef 8454
b7941fd6 8455 if (is_resync_complete(&a->info)) {
0c046afd 8456 /* complete intialization / resync,
0556e1a2
DW		 8457 * recovery and interrupted recovery is completed in
8458 * ->set_disk
0c046afd
DW		 8459 */
8460 if (is_resyncing(dev)) {
8461 dprintf("imsm: mark resync done\n");
809da78e 8462 end_migration(dev, super, map_state);
115c3803 8463 super->updates_pending++;
484240d8 8464 a->last_checkpoint = 0;
115c3803 8465 }
b9172665
AK		 8466 } else if ((!is_resyncing(dev) && !failed) &&
8467 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8468 /* mark the start of the init process if nothing is failed */
b7941fd6 8469 dprintf("imsm: mark resync start\n");
1484e727 8470 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8471 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8472 else
8e59f3d8 8473 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8474 super->updates_pending++;
115c3803 8475 }
a862209d 8476
633b5610 8477mark_checkpoint:
5b83bacf
AK		 8478 /* skip checkpointing for general migration,
8479 * it is controlled in mdadm
8480 */
8481 if (is_gen_migration(dev))
8482 goto skip_mark_checkpoint;
8483
1e5c6983 8484 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8485 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8486 if (blocks_per_unit) {
1e5c6983
DW		 8487 __u32 units32;
8488 __u64 units;
8489
4f0a7acc 8490 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8491 units32 = units;
8492
8493 /* check that we did not overflow 32-bits, and that
8494 * curr_migr_unit needs updating
8495 */
8496 if (units32 == units &&
bfd80a56 8497 units32 != 0 &&
1e5c6983
DW		 8498 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8499 dprintf("imsm: mark checkpoint (%u)\n", units32);
8500 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8501 super->updates_pending++;
8502 }
8503 }
f8f603f1 8504
5b83bacf 8505skip_mark_checkpoint:
3393c6af 8506 /* mark dirty / clean */
2432ce9b
AP		 8507 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8508 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8509 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8510 if (consistent) {
8511 dev->vol.dirty = RAIDVOL_CLEAN;
8512 } else {
8513 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8514 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8515 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8516 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8517 }
a862209d
DW		 8518 super->updates_pending++;
8519 }
28bce06f 8520
01f157d7 8521 return consistent;
a862209d
DW		 8522}
8523
6f50473f
TM		 8524static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8525{
8526 int inst = a->info.container_member;
8527 struct intel_super *super = a->container->sb;
8528 struct imsm_dev *dev = get_imsm_dev(super, inst);
8529 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8530
8531 if (slot > map->num_members) {
8532 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8533 slot, map->num_members - 1);
8534 return -1;
8535 }
8536
8537 if (slot < 0)
8538 return -1;
8539
8540 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8541}
8542
8d45d196 8543static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8544{
8d45d196
DW		 8545 int inst = a->info.container_member;
8546 struct intel_super *super = a->container->sb;
949c47a0 8547 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8548 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8549 struct imsm_disk *disk;
7ce05701
LD		 8550 struct mdinfo *mdi;
8551 int recovery_not_finished = 0;
0c046afd 8552 int failed;
6f50473f 8553 int ord;
0c046afd 8554 __u8 map_state;
fb12a745
TM		 8555 int rebuild_done = 0;
8556 int i;
8d45d196 8557
fb12a745 8558 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8559 if (ord < 0)
8d45d196
DW		 8560 return;
8561
4e6e574a 8562 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8563 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8564
5802a811 8565 /* check for new failures */
ae7d61e3 8566 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8567 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8568 super->updates_pending++;
8d45d196 8569 }
47ee5a45 8570
19859edc 8571 /* check if in_sync */
0556e1a2 8572 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8573 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8574
8575 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8576 rebuild_done = 1;
19859edc
DW		 8577 super->updates_pending++;
8578 }
8d45d196 8579
3b451610
AK		 8580 failed = imsm_count_failed(super, dev, MAP_0);
8581 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8582
0c046afd 8583 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8584 dprintf("imsm: Detected transition to state ");
8585 switch (map_state) {
8586 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8587 dprintf("normal: ");
8588 if (is_rebuilding(dev)) {
1ade5cc1 8589 dprintf_cont("while rebuilding");
7ce05701
LD		 8590 /* check if recovery is really finished */
8591 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8592 if (mdi->recovery_start != MaxSector) {
8593 recovery_not_finished = 1;
8594 break;
8595 }
8596 if (recovery_not_finished) {
1ade5cc1
N		 8597 dprintf_cont("\n");
8598 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8599 if (a->last_checkpoint < mdi->recovery_start) {
8600 a->last_checkpoint = mdi->recovery_start;
8601 super->updates_pending++;
8602 }
8603 break;
8604 }
94002678 8605 end_migration(dev, super, map_state);
238c0a71 8606 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8607 map->failed_disk_num = ~0;
8608 super->updates_pending++;
8609 a->last_checkpoint = 0;
8610 break;
8611 }
8612 if (is_gen_migration(dev)) {
1ade5cc1 8613 dprintf_cont("while general migration");
bf2f0071 8614 if (a->last_checkpoint >= a->info.component_size)
809da78e 8615 end_migration(dev, super, map_state);
94002678
AK		 8616 else
8617 map->map_state = map_state;
238c0a71 8618 map = get_imsm_map(dev, MAP_0);
28bce06f 8619 map->failed_disk_num = ~0;
94002678 8620 super->updates_pending++;
bf2f0071 8621 break;
94002678
AK		 8622 }
8623 break;
8624 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8625 dprintf_cont("degraded: ");
089f9d79 8626 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8627 dprintf_cont("mark degraded");
94002678
AK		 8628 map->map_state = map_state;
8629 super->updates_pending++;
8630 a->last_checkpoint = 0;
8631 break;
8632 }
8633 if (is_rebuilding(dev)) {
d7a1fda2 8634 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8635 if (state & DS_FAULTY) {
8636 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8637 if (n == map->failed_disk_num) {
8638 dprintf_cont("end migration");
8639 end_migration(dev, super, map_state);
a4e96fd8 8640 a->last_checkpoint = 0;
d7a1fda2 8641 } else {
a4e96fd8 8642 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8643 map->map_state = map_state;
8644 }
94002678 8645 super->updates_pending++;
fb12a745
TM		 8646 }
8647
a4e96fd8
MT		 8648 if (!rebuild_done)
8649 break;
8650
fb12a745
TM		 8651 /* check if recovery is really finished */
8652 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8653 if (mdi->recovery_start != MaxSector) {
8654 recovery_not_finished = 1;
8655 break;
8656 }
8657 if (recovery_not_finished) {
8658 dprintf_cont("\n");
a4e96fd8 8659 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8660 if (a->last_checkpoint < mdi->recovery_start) {
8661 a->last_checkpoint =
8662 mdi->recovery_start;
8663 super->updates_pending++;
8664 }
8665 break;
94002678 8666 }
fb12a745
TM		 8667
8668 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8669 end_migration(dev, super, map_state);
8670 a->last_checkpoint = 0;
8671 super->updates_pending++;
fb12a745
TM		 8672
8673 for (i = 0; i < map->num_members; i++) {
8674 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8675
8676 if (idx & IMSM_ORD_REBUILD)
8677 map->failed_disk_num = i;
8678 }
8679 super->updates_pending++;
94002678
AK		 8680 break;
8681 }
8682 if (is_gen_migration(dev)) {
1ade5cc1 8683 dprintf_cont("while general migration");
bf2f0071 8684 if (a->last_checkpoint >= a->info.component_size)
809da78e 8685 end_migration(dev, super, map_state);
94002678
AK		 8686 else {
8687 map->map_state = map_state;
3b451610 8688 manage_second_map(super, dev);
94002678
AK		 8689 }
8690 super->updates_pending++;
bf2f0071 8691 break;
28bce06f 8692 }
6ce1fbf1 8693 if (is_initializing(dev)) {
1ade5cc1 8694 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8695 map->map_state = map_state;
8696 super->updates_pending++;
8697 break;
8698 }
94002678
AK		 8699 break;
8700 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8701 dprintf_cont("failed: ");
94002678 8702 if (is_gen_migration(dev)) {
1ade5cc1 8703 dprintf_cont("while general migration");
94002678
AK		 8704 map->map_state = map_state;
8705 super->updates_pending++;
8706 break;
8707 }
8708 if (map->map_state != map_state) {
1ade5cc1 8709 dprintf_cont("mark failed");
94002678
AK		 8710 end_migration(dev, super, map_state);
8711 super->updates_pending++;
8712 a->last_checkpoint = 0;
8713 break;
8714 }
8715 break;
8716 default:
1ade5cc1 8717 dprintf_cont("state %i\n", map_state);
5802a811 8718 }
1ade5cc1 8719 dprintf_cont("\n");
845dea95
NB		 8720}
8721
f796af5d 8722static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8723{
f796af5d 8724 void *buf = mpb;
c2a1e7da
DW		 8725 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8726 unsigned long long dsize;
8727 unsigned long long sectors;
f36a9ecd 8728 unsigned int sector_size;
c2a1e7da 8729
f36a9ecd 8730 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8731 get_dev_size(fd, NULL, &dsize);
8732
f36a9ecd 8733 if (mpb_size > sector_size) {
272f648f 8734 /* -1 to account for anchor */
f36a9ecd 8735 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8736
272f648f 8737 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8738 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8739 SEEK_SET) < 0)
272f648f 8740 return 1;
c2a1e7da 8741
f36a9ecd
PB		 8742 if ((unsigned long long)write(fd, buf + sector_size,
8743 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8744 return 1;
8745 }
c2a1e7da 8746
272f648f 8747 /* first block is stored on second to last sector of the disk */
f36a9ecd 8748 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8749 return 1;
8750
466070ad 8751 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8752 return 1;
8753
c2a1e7da
DW		 8754 return 0;
8755}
8756
2e735d19 8757static void imsm_sync_metadata(struct supertype *container)
845dea95 8758{
2e735d19 8759 struct intel_super *super = container->sb;
c2a1e7da 8760
1a64be56 8761 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8762 if (!super->updates_pending)
8763 return;
8764
36988a3d 8765 write_super_imsm(container, 0);
c2a1e7da
DW		 8766
8767 super->updates_pending = 0;
845dea95
NB		 8768}
8769
272906ef
DW		 8770static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8771{
8772 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8773 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8774 struct dl *dl;
8775
8776 for (dl = super->disks; dl; dl = dl->next)
8777 if (dl->index == i)
8778 break;
8779
25ed7e59 8780 if (dl && is_failed(&dl->disk))
272906ef
DW		 8781 dl = NULL;
8782
8783 if (dl)
1ade5cc1 8784 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8785
8786 return dl;
8787}
8788
a20d2ba5 8789static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8790 struct active_array *a, int activate_new,
8791 struct mdinfo *additional_test_list)
272906ef
DW		 8792{
8793 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8794 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8795 struct imsm_super *mpb = super->anchor;
8796 struct imsm_map *map;
272906ef
DW		 8797 unsigned long long pos;
8798 struct mdinfo *d;
8799 struct extent *ex;
a20d2ba5 8800 int i, j;
272906ef 8801 int found;
569cc43f
DW		 8802 __u32 array_start = 0;
8803 __u32 array_end = 0;
272906ef 8804 struct dl *dl;
6c932028 8805 struct mdinfo *test_list;
272906ef
DW		 8806
8807 for (dl = super->disks; dl; dl = dl->next) {
8808 /* If in this array, skip */
8809 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8810 if (d->state_fd >= 0 &&
8811 d->disk.major == dl->major &&
272906ef 8812 d->disk.minor == dl->minor) {
8ba77d32
AK		 8813 dprintf("%x:%x already in array\n",
8814 dl->major, dl->minor);
272906ef
DW		 8815 break;
8816 }
8817 if (d)
8818 continue;
6c932028
AK		 8819 test_list = additional_test_list;
8820 while (test_list) {
8821 if (test_list->disk.major == dl->major &&
8822 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8823 dprintf("%x:%x already in additional test list\n",
8824 dl->major, dl->minor);
8825 break;
8826 }
6c932028 8827 test_list = test_list->next;
8ba77d32 8828 }
6c932028 8829 if (test_list)
8ba77d32 8830 continue;
272906ef 8831
e553d2a4 8832 /* skip in use or failed drives */
25ed7e59 8833 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8834 dl->index == -2) {
8835 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8836 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8837 continue;
8838 }
8839
a20d2ba5
DW		 8840 /* skip pure spares when we are looking for partially
8841 * assimilated drives
8842 */
8843 if (dl->index == -1 && !activate_new)
8844 continue;
8845
f2cc4f7d
AO		 8846 if (!drive_validate_sector_size(super, dl))
8847 continue;
8848
272906ef 8849 /* Does this unused device have the requisite free space?
a20d2ba5 8850 * It needs to be able to cover all member volumes
272906ef 8851 */
05501181 8852 ex = get_extents(super, dl, 1);
272906ef
DW		 8853 if (!ex) {
8854 dprintf("cannot get extents\n");
8855 continue;
8856 }
a20d2ba5
DW		 8857 for (i = 0; i < mpb->num_raid_devs; i++) {
8858 dev = get_imsm_dev(super, i);
238c0a71 8859 map = get_imsm_map(dev, MAP_0);
272906ef 8860
a20d2ba5
DW		 8861 /* check if this disk is already a member of
8862 * this array
272906ef 8863 */
620b1713 8864 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8865 continue;
8866
8867 found = 0;
8868 j = 0;
8869 pos = 0;
5551b113 8870 array_start = pba_of_lba0(map);
329c8278 8871 array_end = array_start +
44490938 8872 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8873
8874 do {
8875 /* check that we can start at pba_of_lba0 with
44490938 8876 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8877 */
329c8278 8878 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8879 found = 1;
8880 break;
8881 }
8882 pos = ex[j].start + ex[j].size;
8883 j++;
8884 } while (ex[j-1].size);
8885
8886 if (!found)
272906ef 8887 break;
a20d2ba5 8888 }
272906ef
DW		 8889
8890 free(ex);
a20d2ba5 8891 if (i < mpb->num_raid_devs) {
329c8278
DW		 8892 dprintf("%x:%x does not have %u to %u available\n",
8893 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8894 /* No room */
8895 continue;
a20d2ba5
DW		 8896 }
8897 return dl;
272906ef
DW		 8898 }
8899
8900 return dl;
8901}
8902
95d07a2c
LM		 8903static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8904{
8905 struct imsm_dev *dev2;
8906 struct imsm_map *map;
8907 struct dl *idisk;
8908 int slot;
8909 int idx;
8910 __u8 state;
8911
8912 dev2 = get_imsm_dev(cont->sb, dev_idx);
8913 if (dev2) {
238c0a71 8914 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8915 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8916 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8917 if (!map)
8918 return 1;
8919 for (slot = 0; slot < map->num_members; slot++) {
8920 /*
8921 * Check if failed disks are deleted from intel
8922 * disk list or are marked to be deleted
8923 */
238c0a71 8924 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8925 idisk = get_imsm_dl_disk(cont->sb, idx);
8926 /*
8927 * Do not rebuild the array if failed disks
8928 * from failed sub-array are not removed from
8929 * container.
8930 */
8931 if (idisk &&
8932 is_failed(&idisk->disk) &&
8933 (idisk->action != DISK_REMOVE))
8934 return 0;
8935 }
8936 }
8937 }
8938 return 1;
8939}
8940
88758e9d
DW		 8941static struct mdinfo *imsm_activate_spare(struct active_array *a,
8942 struct metadata_update **updates)
8943{
8944 /**
d23fe947
DW		 8945 * Find a device with unused free space and use it to replace a
8946 * failed/vacant region in an array. We replace failed regions one a
8947 * array at a time. The result is that a new spare disk will be added
8948 * to the first failed array and after the monitor has finished
8949 * propagating failures the remainder will be consumed.
88758e9d 8950 *
d23fe947
DW		 8951 * FIXME add a capability for mdmon to request spares from another
8952 * container.
88758e9d
DW		 8953 */
8954
8955 struct intel_super *super = a->container->sb;
88758e9d 8956 int inst = a->info.container_member;
949c47a0 8957 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8958 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8959 int failed = a->info.array.raid_disks;
8960 struct mdinfo *rv = NULL;
8961 struct mdinfo *d;
8962 struct mdinfo *di;
8963 struct metadata_update *mu;
8964 struct dl *dl;
8965 struct imsm_update_activate_spare *u;
8966 int num_spares = 0;
8967 int i;
95d07a2c 8968 int allowed;
88758e9d
DW		 8969
8970 for (d = a->info.devs ; d ; d = d->next) {
8971 if ((d->curr_state & DS_FAULTY) &&
8972 d->state_fd >= 0)
8973 /* wait for Removal to happen */
8974 return NULL;
8975 if (d->state_fd >= 0)
8976 failed--;
8977 }
8978
8979 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8980 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8981
e2962bfc
AK		 8982 if (imsm_reshape_blocks_arrays_changes(super))
8983 return NULL;
1af97990 8984
fc8ca064
AK		 8985 /* Cannot activate another spare if rebuild is in progress already
8986 */
8987 if (is_rebuilding(dev)) {
7a862a02 8988 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8989 return NULL;
8990 }
8991
89c67882
AK		 8992 if (a->info.array.level == 4)
8993 /* No repair for takeovered array
8994 * imsm doesn't support raid4
8995 */
8996 return NULL;
8997
3b451610
AK		 8998 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8999 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9000 return NULL;
9001
83ca7d45
AP		 9002 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9003 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9004 return NULL;
9005 }
9006
95d07a2c
LM		 9007 /*
9008 * If there are any failed disks check state of the other volume.
9009 * Block rebuild if the another one is failed until failed disks
9010 * are removed from container.
9011 */
9012 if (failed) {
7a862a02 9013 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9014 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9015 /* check if states of the other volumes allow for rebuild */
9016 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9017 if (i != inst) {
9018 allowed = imsm_rebuild_allowed(a->container,
9019 i, failed);
9020 if (!allowed)
9021 return NULL;
9022 }
9023 }
9024 }
9025
88758e9d 9026 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9027 for (i = 0; i < a->info.array.raid_disks; i++) {
9028 for (d = a->info.devs ; d ; d = d->next)
9029 if (d->disk.raid_disk == i)
9030 break;
9031 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9032 if (d && (d->state_fd >= 0))
9033 continue;
9034
272906ef 9035 /*
a20d2ba5
DW		 9036 * OK, this device needs recovery. Try to re-add the
9037 * previous occupant of this slot, if this fails see if
9038 * we can continue the assimilation of a spare that was
9039 * partially assimilated, finally try to activate a new
9040 * spare.
272906ef
DW		 9041 */
9042 dl = imsm_readd(super, i, a);
9043 if (!dl)
b303fe21 9044 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9045 if (!dl)
b303fe21 9046 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9047 if (!dl)
9048 continue;
1011e834 9049
272906ef 9050 /* found a usable disk with enough space */
503975b9 9051 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9052
9053 /* dl->index will be -1 in the case we are activating a
9054 * pristine spare. imsm_process_update() will create a
9055 * new index in this case. Once a disk is found to be
9056 * failed in all member arrays it is kicked from the
9057 * metadata
9058 */
9059 di->disk.number = dl->index;
d23fe947 9060
272906ef
DW		 9061 /* (ab)use di->devs to store a pointer to the device
9062 * we chose
9063 */
9064 di->devs = (struct mdinfo *) dl;
9065
9066 di->disk.raid_disk = i;
9067 di->disk.major = dl->major;
9068 di->disk.minor = dl->minor;
9069 di->disk.state = 0;
d23534e4 9070 di->recovery_start = 0;
5551b113 9071 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9072 di->component_size = a->info.component_size;
9073 di->container_member = inst;
5e46202e 9074 di->bb.supported = 1;
2c8890e9 9075 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9076 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9077 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9078 }
148acb7b 9079 super->random = random32();
272906ef
DW		 9080 di->next = rv;
9081 rv = di;
9082 num_spares++;
9083 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9084 i, di->data_offset);
88758e9d
DW		 9085 }
9086
9087 if (!rv)
9088 /* No spares found */
9089 return rv;
9090 /* Now 'rv' has a list of devices to return.
9091 * Create a metadata_update record to update the
9092 * disk_ord_tbl for the array
9093 */
503975b9 9094 mu = xmalloc(sizeof(*mu));
1011e834 9095 mu->buf = xcalloc(num_spares,
503975b9 9096 sizeof(struct imsm_update_activate_spare));
88758e9d 9097 mu->space = NULL;
cb23f1f4 9098 mu->space_list = NULL;
88758e9d
DW		 9099 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9100 mu->next = *updates;
9101 u = (struct imsm_update_activate_spare *) mu->buf;
9102
9103 for (di = rv ; di ; di = di->next) {
9104 u->type = update_activate_spare;
d23fe947
DW		 9105 u->dl = (struct dl *) di->devs;
9106 di->devs = NULL;
88758e9d
DW		 9107 u->slot = di->disk.raid_disk;
9108 u->array = inst;
9109 u->next = u + 1;
9110 u++;
9111 }
9112 (u-1)->next = NULL;
9113 *updates = mu;
9114
9115 return rv;
9116}
9117
54c2c1ea 9118static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9119{
54c2c1ea 9120 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9121 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9122 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9123 struct disk_info *inf = get_disk_info(u);
9124 struct imsm_disk *disk;
8273f55e
DW		 9125 int i;
9126 int j;
8273f55e 9127
54c2c1ea 9128 for (i = 0; i < map->num_members; i++) {
238c0a71 9129 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9130 for (j = 0; j < new_map->num_members; j++)
9131 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9132 return 1;
9133 }
9134
9135 return 0;
9136}
9137
1a64be56
LM		 9138static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9139{
594dc1b8
JS		 9140 struct dl *dl;
9141
1a64be56 9142 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9143 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9144 return dl;
9145 return NULL;
9146}
9147
9148static int remove_disk_super(struct intel_super *super, int major, int minor)
9149{
594dc1b8 9150 struct dl *prev;
1a64be56
LM		 9151 struct dl *dl;
9152
9153 prev = NULL;
9154 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9155 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9156 /* remove */
9157 if (prev)
9158 prev->next = dl->next;
9159 else
9160 super->disks = dl->next;
9161 dl->next = NULL;
9162 __free_imsm_disk(dl);
1ade5cc1 9163 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9164 break;
9165 }
9166 prev = dl;
9167 }
9168 return 0;
9169}
9170
f21e18ca 9171static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9172
1a64be56
LM		 9173static int add_remove_disk_update(struct intel_super *super)
9174{
9175 int check_degraded = 0;
594dc1b8
JS		 9176 struct dl *disk;
9177
1a64be56
LM		 9178 /* add/remove some spares to/from the metadata/contrainer */
9179 while (super->disk_mgmt_list) {
9180 struct dl *disk_cfg;
9181
9182 disk_cfg = super->disk_mgmt_list;
9183 super->disk_mgmt_list = disk_cfg->next;
9184 disk_cfg->next = NULL;
9185
9186 if (disk_cfg->action == DISK_ADD) {
9187 disk_cfg->next = super->disks;
9188 super->disks = disk_cfg;
9189 check_degraded = 1;
1ade5cc1
N		 9190 dprintf("added %x:%x\n",
9191 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9192 } else if (disk_cfg->action == DISK_REMOVE) {
9193 dprintf("Disk remove action processed: %x.%x\n",
9194 disk_cfg->major, disk_cfg->minor);
9195 disk = get_disk_super(super,
9196 disk_cfg->major,
9197 disk_cfg->minor);
9198 if (disk) {
9199 /* store action status */
9200 disk->action = DISK_REMOVE;
9201 /* remove spare disks only */
9202 if (disk->index == -1) {
9203 remove_disk_super(super,
9204 disk_cfg->major,
9205 disk_cfg->minor);
91c97c54
MT		 9206 } else {
9207 disk_cfg->fd = disk->fd;
9208 disk->fd = -1;
1a64be56
LM		 9209 }
9210 }
9211 /* release allocate disk structure */
9212 __free_imsm_disk(disk_cfg);
9213 }
9214 }
9215 return check_degraded;
9216}
9217
a29911da
PC		 9218static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9219 struct intel_super *super,
9220 void ***space_list)
9221{
9222 struct intel_dev *id;
9223 void **tofree = NULL;
9224 int ret_val = 0;
9225
1ade5cc1 9226 dprintf("(enter)\n");
089f9d79 9227 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9228 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9229 return ret_val;
9230 }
089f9d79 9231 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9232 dprintf("imsm: Error: Memory is not allocated\n");
9233 return ret_val;
9234 }
9235
9236 for (id = super->devlist ; id; id = id->next) {
9237 if (id->index == (unsigned)u->subdev) {
9238 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9239 struct imsm_map *map;
9240 struct imsm_dev *new_dev =
9241 (struct imsm_dev *)*space_list;
238c0a71 9242 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9243 int to_state;
9244 struct dl *new_disk;
9245
9246 if (new_dev == NULL)
9247 return ret_val;
9248 *space_list = **space_list;
9249 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9250 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9251 if (migr_map) {
9252 dprintf("imsm: Error: migration in progress");
9253 return ret_val;
9254 }
9255
9256 to_state = map->map_state;
9257 if ((u->new_level == 5) && (map->raid_level == 0)) {
9258 map->num_members++;
9259 /* this should not happen */
9260 if (u->new_disks[0] < 0) {
9261 map->failed_disk_num =
9262 map->num_members - 1;
9263 to_state = IMSM_T_STATE_DEGRADED;
9264 } else
9265 to_state = IMSM_T_STATE_NORMAL;
9266 }
8e59f3d8 9267 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9268 if (u->new_level > -1)
9269 map->raid_level = u->new_level;
238c0a71 9270 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9271 if ((u->new_level == 5) &&
9272 (migr_map->raid_level == 0)) {
9273 int ord = map->num_members - 1;
9274 migr_map->num_members--;
9275 if (u->new_disks[0] < 0)
9276 ord |= IMSM_ORD_REBUILD;
9277 set_imsm_ord_tbl_ent(map,
9278 map->num_members - 1,
9279 ord);
9280 }
9281 id->dev = new_dev;
9282 tofree = (void **)dev;
9283
4bba0439
PC		 9284 /* update chunk size
9285 */
06fb291a
PB		 9286 if (u->new_chunksize > 0) {
9287 unsigned long long num_data_stripes;
9529d343
MD		 9288 struct imsm_map *dest_map =
9289 get_imsm_map(dev, MAP_0);
06fb291a 9290 int used_disks =
9529d343 9291 imsm_num_data_members(dest_map);
06fb291a
PB		 9292
9293 if (used_disks == 0)
9294 return ret_val;
9295
4bba0439
PC		 9296 map->blocks_per_strip =
9297 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9298 num_data_stripes =
fcc2c9da 9299 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9300 num_data_stripes /= map->blocks_per_strip;
9301 num_data_stripes /= map->num_domains;
9302 set_num_data_stripes(map, num_data_stripes);
9303 }
4bba0439 9304
44490938
MD		 9305 /* ensure blocks_per_member has valid value
9306 */
9307 set_blocks_per_member(map,
9308 per_dev_array_size(map) +
9309 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9310
a29911da
PC		 9311 /* add disk
9312 */
089f9d79
JS		 9313 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9314 migr_map->raid_level == map->raid_level)
a29911da
PC		 9315 goto skip_disk_add;
9316
9317 if (u->new_disks[0] >= 0) {
9318 /* use passes spare
9319 */
9320 new_disk = get_disk_super(super,
9321 major(u->new_disks[0]),
9322 minor(u->new_disks[0]));
7a862a02 9323 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9324 major(u->new_disks[0]),
9325 minor(u->new_disks[0]),
9326 new_disk, new_disk->index);
9327 if (new_disk == NULL)
9328 goto error_disk_add;
9329
9330 new_disk->index = map->num_members - 1;
9331 /* slot to fill in autolayout
9332 */
9333 new_disk->raiddisk = new_disk->index;
9334 new_disk->disk.status |= CONFIGURED_DISK;
9335 new_disk->disk.status &= ~SPARE_DISK;
9336 } else
9337 goto error_disk_add;
9338
9339skip_disk_add:
9340 *tofree = *space_list;
9341 /* calculate new size
9342 */
f3871fdc 9343 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9344
9345 ret_val = 1;
9346 }
9347 }
9348
9349 if (tofree)
9350 *space_list = tofree;
9351 return ret_val;
9352
9353error_disk_add:
9354 dprintf("Error: imsm: Cannot find disk.\n");
9355 return ret_val;
9356}
9357
f3871fdc
AK		 9358static int apply_size_change_update(struct imsm_update_size_change *u,
9359 struct intel_super *super)
9360{
9361 struct intel_dev *id;
9362 int ret_val = 0;
9363
1ade5cc1 9364 dprintf("(enter)\n");
089f9d79 9365 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9366 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9367 return ret_val;
9368 }
9369
9370 for (id = super->devlist ; id; id = id->next) {
9371 if (id->index == (unsigned)u->subdev) {
9372 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9373 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9374 int used_disks = imsm_num_data_members(map);
f3871fdc 9375 unsigned long long blocks_per_member;
06fb291a 9376 unsigned long long num_data_stripes;
44490938
MD		 9377 unsigned long long new_size_per_disk;
9378
9379 if (used_disks == 0)
9380 return 0;
f3871fdc
AK		 9381
9382 /* calculate new size
9383 */
44490938
MD		 9384 new_size_per_disk = u->new_size / used_disks;
9385 blocks_per_member = new_size_per_disk +
9386 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9387 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9388 map->blocks_per_strip;
9389 num_data_stripes /= map->num_domains;
9390 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9391 u->new_size, new_size_per_disk,
06fb291a 9392 num_data_stripes);
f3871fdc 9393 set_blocks_per_member(map, blocks_per_member);
06fb291a 9394 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9395 imsm_set_array_size(dev, u->new_size);
9396
9397 ret_val = 1;
9398 break;
9399 }
9400 }
9401
9402 return ret_val;
9403}
9404
061d7da3 9405static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9406 struct intel_super *super,
061d7da3
LO		 9407 struct active_array *active_array)
9408{
9409 struct imsm_super *mpb = super->anchor;
9410 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9411 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9412 struct imsm_map *migr_map;
9413 struct active_array *a;
9414 struct imsm_disk *disk;
9415 __u8 to_state;
9416 struct dl *dl;
9417 unsigned int found;
9418 int failed;
5961eeec 9419 int victim;
061d7da3 9420 int i;
5961eeec 9421 int second_map_created = 0;
061d7da3 9422
5961eeec 9423 for (; u; u = u->next) {
238c0a71 9424 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9425
5961eeec 9426 if (victim < 0)
9427 return 0;
061d7da3 9428
5961eeec 9429 for (dl = super->disks; dl; dl = dl->next)
9430 if (dl == u->dl)
9431 break;
061d7da3 9432
5961eeec 9433 if (!dl) {
7a862a02 9434 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9435 u->dl->index);
9436 return 0;
9437 }
061d7da3 9438
5961eeec 9439 /* count failures (excluding rebuilds and the victim)
9440 * to determine map[0] state
9441 */
9442 failed = 0;
9443 for (i = 0; i < map->num_members; i++) {
9444 if (i == u->slot)
9445 continue;
9446 disk = get_imsm_disk(super,
238c0a71 9447 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9448 if (!disk || is_failed(disk))
9449 failed++;
9450 }
061d7da3 9451
5961eeec 9452 /* adding a pristine spare, assign a new index */
9453 if (dl->index < 0) {
9454 dl->index = super->anchor->num_disks;
9455 super->anchor->num_disks++;
9456 }
9457 disk = &dl->disk;
9458 disk->status |= CONFIGURED_DISK;
9459 disk->status &= ~SPARE_DISK;
9460
9461 /* mark rebuild */
238c0a71 9462 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9463 if (!second_map_created) {
9464 second_map_created = 1;
9465 map->map_state = IMSM_T_STATE_DEGRADED;
9466 migrate(dev, super, to_state, MIGR_REBUILD);
9467 } else
9468 map->map_state = to_state;
238c0a71 9469 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9470 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9471 set_imsm_ord_tbl_ent(migr_map, u->slot,
9472 dl->index | IMSM_ORD_REBUILD);
9473
9474 /* update the family_num to mark a new container
9475 * generation, being careful to record the existing
9476 * family_num in orig_family_num to clean up after
9477 * earlier mdadm versions that neglected to set it.
9478 */
9479 if (mpb->orig_family_num == 0)
9480 mpb->orig_family_num = mpb->family_num;
9481 mpb->family_num += super->random;
9482
9483 /* count arrays using the victim in the metadata */
9484 found = 0;
9485 for (a = active_array; a ; a = a->next) {
9486 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9487 map = get_imsm_map(dev, MAP_0);
061d7da3 9488
5961eeec 9489 if (get_imsm_disk_slot(map, victim) >= 0)
9490 found++;
9491 }
061d7da3 9492
5961eeec 9493 /* delete the victim if it is no longer being
9494 * utilized anywhere
061d7da3 9495 */
5961eeec 9496 if (!found) {
9497 struct dl **dlp;
061d7da3 9498
5961eeec 9499 /* We know that 'manager' isn't touching anything,
9500 * so it is safe to delete
9501 */
9502 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9503 if ((*dlp)->index == victim)
9504 break;
5961eeec 9505
9506 /* victim may be on the missing list */
9507 if (!*dlp)
9508 for (dlp = &super->missing; *dlp;
9509 dlp = &(*dlp)->next)
9510 if ((*dlp)->index == victim)
9511 break;
9512 imsm_delete(super, dlp, victim);
9513 }
061d7da3
LO		 9514 }
9515
9516 return 1;
9517}
a29911da 9518
2e5dc010
N		 9519static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9520 struct intel_super *super,
9521 void ***space_list)
9522{
9523 struct dl *new_disk;
9524 struct intel_dev *id;
9525 int i;
9526 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9527 int disk_count = u->old_raid_disks;
2e5dc010
N		 9528 void **tofree = NULL;
9529 int devices_to_reshape = 1;
9530 struct imsm_super *mpb = super->anchor;
9531 int ret_val = 0;
d098291a 9532 unsigned int dev_id;
2e5dc010 9533
1ade5cc1 9534 dprintf("(enter)\n");
2e5dc010
N		 9535
9536 /* enable spares to use in array */
9537 for (i = 0; i < delta_disks; i++) {
9538 new_disk = get_disk_super(super,
9539 major(u->new_disks[i]),
9540 minor(u->new_disks[i]));
7a862a02 9541 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9542 major(u->new_disks[i]), minor(u->new_disks[i]),
9543 new_disk, new_disk->index);
089f9d79
JS		 9544 if (new_disk == NULL ||
9545 (new_disk->index >= 0 &&
9546 new_disk->index < u->old_raid_disks))
2e5dc010 9547 goto update_reshape_exit;
ee4beede 9548 new_disk->index = disk_count++;
2e5dc010
N		 9549 /* slot to fill in autolayout
9550 */
9551 new_disk->raiddisk = new_disk->index;
9552 new_disk->disk.status |=
9553 CONFIGURED_DISK;
9554 new_disk->disk.status &= ~SPARE_DISK;
9555 }
9556
ed7333bd
AK		 9557 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9558 mpb->num_raid_devs);
2e5dc010
N		 9559 /* manage changes in volume
9560 */
d098291a 9561 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9562 void **sp = *space_list;
9563 struct imsm_dev *newdev;
9564 struct imsm_map *newmap, *oldmap;
9565
d098291a
AK		 9566 for (id = super->devlist ; id; id = id->next) {
9567 if (id->index == dev_id)
9568 break;
9569 }
9570 if (id == NULL)
9571 break;
2e5dc010
N		 9572 if (!sp)
9573 continue;
9574 *space_list = *sp;
9575 newdev = (void*)sp;
9576 /* Copy the dev, but not (all of) the map */
9577 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9578 oldmap = get_imsm_map(id->dev, MAP_0);
9579 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9580 /* Copy the current map */
9581 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9582 /* update one device only
9583 */
9584 if (devices_to_reshape) {
ed7333bd
AK		 9585 dprintf("imsm: modifying subdev: %i\n",
9586 id->index);
2e5dc010
N		 9587 devices_to_reshape--;
9588 newdev->vol.migr_state = 1;
9589 newdev->vol.curr_migr_unit = 0;
ea672ee1 9590 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9591 newmap->num_members = u->new_raid_disks;
9592 for (i = 0; i < delta_disks; i++) {
9593 set_imsm_ord_tbl_ent(newmap,
9594 u->old_raid_disks + i,
9595 u->old_raid_disks + i);
9596 }
9597 /* New map is correct, now need to save old map
9598 */
238c0a71 9599 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9600 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9601
f3871fdc 9602 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9603 }
9604
9605 sp = (void **)id->dev;
9606 id->dev = newdev;
9607 *sp = tofree;
9608 tofree = sp;
8e59f3d8
AK		 9609
9610 /* Clear migration record */
9611 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9612 }
819bc634
AK		 9613 if (tofree)
9614 *space_list = tofree;
2e5dc010
N		 9615 ret_val = 1;
9616
9617update_reshape_exit:
9618
9619 return ret_val;
9620}
9621
bb025c2f 9622static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9623 struct intel_super *super,
9624 void ***space_list)
bb025c2f
KW		 9625{
9626 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9627 struct intel_dev *dv;
9628 struct imsm_dev *dev_new;
bb025c2f
KW		 9629 struct imsm_map *map;
9630 struct dl *dm, *du;
8ca6df95 9631 int i;
bb025c2f
KW		 9632
9633 for (dv = super->devlist; dv; dv = dv->next)
9634 if (dv->index == (unsigned int)u->subarray) {
9635 dev = dv->dev;
9636 break;
9637 }
9638
9639 if (dev == NULL)
9640 return 0;
9641
238c0a71 9642 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9643
9644 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9645 unsigned long long num_data_stripes;
9646
43d5ec18 9647 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9648 if (imsm_count_failed(super, dev, MAP_0) !=
9649 (map->num_members / 2))
43d5ec18
KW		 9650 return 0;
9651
bb025c2f
KW		 9652 /* iterate through devices to mark removed disks as spare */
9653 for (dm = super->disks; dm; dm = dm->next) {
9654 if (dm->disk.status & FAILED_DISK) {
9655 int idx = dm->index;
9656 /* update indexes on the disk list */
9657/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9658 the index values will end up being correct.... NB */
9659 for (du = super->disks; du; du = du->next)
9660 if (du->index > idx)
9661 du->index--;
9662 /* mark as spare disk */
a8619d23 9663 mark_spare(dm);
bb025c2f
KW		 9664 }
9665 }
bb025c2f
KW		 9666 /* update map */
9667 map->num_members = map->num_members / 2;
9668 map->map_state = IMSM_T_STATE_NORMAL;
9669 map->num_domains = 1;
9670 map->raid_level = 0;
9671 map->failed_disk_num = -1;
4a353e6e
RS		 9672 num_data_stripes = imsm_dev_size(dev) / 2;
9673 num_data_stripes /= map->blocks_per_strip;
9674 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9675 }
9676
8ca6df95
KW		 9677 if (u->direction == R0_TO_R10) {
9678 void **space;
4a353e6e
RS		 9679 unsigned long long num_data_stripes;
9680
8ca6df95
KW		 9681 /* update slots in current disk list */
9682 for (dm = super->disks; dm; dm = dm->next) {
9683 if (dm->index >= 0)
9684 dm->index *= 2;
9685 }
9686 /* create new *missing* disks */
9687 for (i = 0; i < map->num_members; i++) {
9688 space = *space_list;
9689 if (!space)
9690 continue;
9691 *space_list = *space;
9692 du = (void *)space;
9693 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9694 du->fd = -1;
9695 du->minor = 0;
9696 du->major = 0;
9697 du->index = (i * 2) + 1;
9698 sprintf((char *)du->disk.serial,
9699 " MISSING_%d", du->index);
9700 sprintf((char *)du->serial,
9701 "MISSING_%d", du->index);
9702 du->next = super->missing;
9703 super->missing = du;
9704 }
9705 /* create new dev and map */
9706 space = *space_list;
9707 if (!space)
9708 return 0;
9709 *space_list = *space;
9710 dev_new = (void *)space;
9711 memcpy(dev_new, dev, sizeof(*dev));
9712 /* update new map */
238c0a71 9713 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9714 map->num_members = map->num_members * 2;
1a2487c2 9715 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9716 map->num_domains = 2;
9717 map->raid_level = 1;
4a353e6e
RS		 9718 num_data_stripes = imsm_dev_size(dev) / 2;
9719 num_data_stripes /= map->blocks_per_strip;
9720 num_data_stripes /= map->num_domains;
9721 set_num_data_stripes(map, num_data_stripes);
9722
8ca6df95
KW		 9723 /* replace dev<->dev_new */
9724 dv->dev = dev_new;
9725 }
bb025c2f
KW		 9726 /* update disk order table */
9727 for (du = super->disks; du; du = du->next)
9728 if (du->index >= 0)
9729 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9730 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9731 if (du->index >= 0) {
9732 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9733 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9734 }
bb025c2f
KW		 9735
9736 return 1;
9737}
9738
e8319a19
DW		 9739static void imsm_process_update(struct supertype *st,
9740 struct metadata_update *update)
9741{
9742 /**
9743 * crack open the metadata_update envelope to find the update record
9744 * update can be one of:
d195167d
AK		 9745 * update_reshape_container_disks - all the arrays in the container
9746 * are being reshaped to have more devices. We need to mark
9747 * the arrays for general migration and convert selected spares
9748 * into active devices.
9749 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9750 * device in an array, update the disk_ord_tbl. If this disk is
9751 * present in all member arrays then also clear the SPARE_DISK
9752 * flag
d195167d
AK		 9753 * update_create_array
9754 * update_kill_array
9755 * update_rename_array
9756 * update_add_remove_disk
e8319a19
DW		 9757 */
9758 struct intel_super *super = st->sb;
4d7b1503 9759 struct imsm_super *mpb;
e8319a19
DW		 9760 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9761
4d7b1503
DW		 9762 /* update requires a larger buf but the allocation failed */
9763 if (super->next_len && !super->next_buf) {
9764 super->next_len = 0;
9765 return;
9766 }
9767
9768 if (super->next_buf) {
9769 memcpy(super->next_buf, super->buf, super->len);
9770 free(super->buf);
9771 super->len = super->next_len;
9772 super->buf = super->next_buf;
9773
9774 super->next_len = 0;
9775 super->next_buf = NULL;
9776 }
9777
9778 mpb = super->anchor;
9779
e8319a19 9780 switch (type) {
0ec5d470
AK		 9781 case update_general_migration_checkpoint: {
9782 struct intel_dev *id;
9783 struct imsm_update_general_migration_checkpoint *u =
9784 (void *)update->buf;
9785
1ade5cc1 9786 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9787
9788 /* find device under general migration */
9789 for (id = super->devlist ; id; id = id->next) {
9790 if (is_gen_migration(id->dev)) {
9791 id->dev->vol.curr_migr_unit =
9792 __cpu_to_le32(u->curr_migr_unit);
9793 super->updates_pending++;
9794 }
9795 }
9796 break;
9797 }
bb025c2f
KW		 9798 case update_takeover: {
9799 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9800 if (apply_takeover_update(u, super, &update->space_list)) {
9801 imsm_update_version_info(super);
bb025c2f 9802 super->updates_pending++;
1a2487c2 9803 }
bb025c2f
KW		 9804 break;
9805 }
9806
78b10e66 9807 case update_reshape_container_disks: {
d195167d 9808 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9809 if (apply_reshape_container_disks_update(
9810 u, super, &update->space_list))
9811 super->updates_pending++;
78b10e66
N		 9812 break;
9813 }
48c5303a 9814 case update_reshape_migration: {
a29911da
PC		 9815 struct imsm_update_reshape_migration *u = (void *)update->buf;
9816 if (apply_reshape_migration_update(
9817 u, super, &update->space_list))
9818 super->updates_pending++;
48c5303a
PC		 9819 break;
9820 }
f3871fdc
AK		 9821 case update_size_change: {
9822 struct imsm_update_size_change *u = (void *)update->buf;
9823 if (apply_size_change_update(u, super))
9824 super->updates_pending++;
9825 break;
9826 }
e8319a19 9827 case update_activate_spare: {
1011e834 9828 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9829 if (apply_update_activate_spare(u, super, st->arrays))
9830 super->updates_pending++;
8273f55e
DW		 9831 break;
9832 }
9833 case update_create_array: {
9834 /* someone wants to create a new array, we need to be aware of
9835 * a few races/collisions:
9836 * 1/ 'Create' called by two separate instances of mdadm
9837 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9838 * devices that have since been assimilated via
9839 * activate_spare.
9840 * In the event this update can not be carried out mdadm will
9841 * (FIX ME) notice that its update did not take hold.
9842 */
9843 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9844 struct intel_dev *dv;
8273f55e
DW		 9845 struct imsm_dev *dev;
9846 struct imsm_map *map, *new_map;
9847 unsigned long long start, end;
9848 unsigned long long new_start, new_end;
9849 int i;
54c2c1ea
DW		 9850 struct disk_info *inf;
9851 struct dl *dl;
8273f55e
DW		 9852
9853 /* handle racing creates: first come first serve */
9854 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9855 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9856 goto create_error;
8273f55e
DW		 9857 }
9858
9859 /* check update is next in sequence */
9860 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9861 dprintf("can not create array %d expected index %d\n",
9862 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9863 goto create_error;
8273f55e
DW		 9864 }
9865
238c0a71 9866 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9867 new_start = pba_of_lba0(new_map);
44490938 9868 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9869 inf = get_disk_info(u);
8273f55e
DW		 9870
9871 /* handle activate_spare versus create race:
9872 * check to make sure that overlapping arrays do not include
9873 * overalpping disks
9874 */
9875 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9876 dev = get_imsm_dev(super, i);
238c0a71 9877 map = get_imsm_map(dev, MAP_0);
5551b113 9878 start = pba_of_lba0(map);
44490938 9879 end = start + per_dev_array_size(map);
8273f55e
DW		 9880 if ((new_start >= start && new_start <= end) ||
9881 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9882 /* overlap */;
9883 else
9884 continue;
9885
9886 if (disks_overlap(super, i, u)) {
1ade5cc1 9887 dprintf("arrays overlap\n");
ba2de7ba 9888 goto create_error;
8273f55e
DW		 9889 }
9890 }
8273f55e 9891
949c47a0
DW		 9892 /* check that prepare update was successful */
9893 if (!update->space) {
1ade5cc1 9894 dprintf("prepare update failed\n");
ba2de7ba 9895 goto create_error;
949c47a0
DW		 9896 }
9897
54c2c1ea
DW		 9898 /* check that all disks are still active before committing
9899 * changes. FIXME: could we instead handle this by creating a
9900 * degraded array? That's probably not what the user expects,
9901 * so better to drop this update on the floor.
9902 */
9903 for (i = 0; i < new_map->num_members; i++) {
9904 dl = serial_to_dl(inf[i].serial, super);
9905 if (!dl) {
1ade5cc1 9906 dprintf("disk disappeared\n");
ba2de7ba 9907 goto create_error;
54c2c1ea 9908 }
949c47a0
DW		 9909 }
9910
8273f55e 9911 super->updates_pending++;
54c2c1ea
DW		 9912
9913 /* convert spares to members and fixup ord_tbl */
9914 for (i = 0; i < new_map->num_members; i++) {
9915 dl = serial_to_dl(inf[i].serial, super);
9916 if (dl->index == -1) {
9917 dl->index = mpb->num_disks;
9918 mpb->num_disks++;
9919 dl->disk.status |= CONFIGURED_DISK;
9920 dl->disk.status &= ~SPARE_DISK;
9921 }
9922 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9923 }
9924
ba2de7ba
DW		 9925 dv = update->space;
9926 dev = dv->dev;
949c47a0
DW		 9927 update->space = NULL;
9928 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9929 dv->index = u->dev_idx;
9930 dv->next = super->devlist;
9931 super->devlist = dv;
8273f55e 9932 mpb->num_raid_devs++;
8273f55e 9933
4d1313e9 9934 imsm_update_version_info(super);
8273f55e 9935 break;
ba2de7ba
DW		 9936 create_error:
9937 /* mdmon knows how to release update->space, but not
9938 * ((struct intel_dev *) update->space)->dev
9939 */
9940 if (update->space) {
9941 dv = update->space;
9942 free(dv->dev);
9943 }
8273f55e 9944 break;
e8319a19 9945 }
33414a01
DW		 9946 case update_kill_array: {
9947 struct imsm_update_kill_array *u = (void *) update->buf;
9948 int victim = u->dev_idx;
9949 struct active_array *a;
9950 struct intel_dev **dp;
9951 struct imsm_dev *dev;
9952
9953 /* sanity check that we are not affecting the uuid of
9954 * active arrays, or deleting an active array
9955 *
9956 * FIXME when immutable ids are available, but note that
9957 * we'll also need to fixup the invalidated/active
9958 * subarray indexes in mdstat
9959 */
9960 for (a = st->arrays; a; a = a->next)
9961 if (a->info.container_member >= victim)
9962 break;
9963 /* by definition if mdmon is running at least one array
9964 * is active in the container, so checking
9965 * mpb->num_raid_devs is just extra paranoia
9966 */
9967 dev = get_imsm_dev(super, victim);
9968 if (a || !dev || mpb->num_raid_devs == 1) {
9969 dprintf("failed to delete subarray-%d\n", victim);
9970 break;
9971 }
9972
9973 for (dp = &super->devlist; *dp;)
f21e18ca 9974 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9975 *dp = (*dp)->next;
9976 } else {
f21e18ca 9977 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9978 (*dp)->index--;
9979 dp = &(*dp)->next;
9980 }
9981 mpb->num_raid_devs--;
9982 super->updates_pending++;
9983 break;
9984 }
aa534678
DW		 9985 case update_rename_array: {
9986 struct imsm_update_rename_array *u = (void *) update->buf;
9987 char name[MAX_RAID_SERIAL_LEN+1];
9988 int target = u->dev_idx;
9989 struct active_array *a;
9990 struct imsm_dev *dev;
9991
9992 /* sanity check that we are not affecting the uuid of
9993 * an active array
9994 */
40659392 9995 memset(name, 0, sizeof(name));
aa534678
DW		 9996 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9997 name[MAX_RAID_SERIAL_LEN] = '\0';
9998 for (a = st->arrays; a; a = a->next)
9999 if (a->info.container_member == target)
10000 break;
10001 dev = get_imsm_dev(super, u->dev_idx);
10002 if (a || !dev || !check_name(super, name, 1)) {
10003 dprintf("failed to rename subarray-%d\n", target);
10004 break;
10005 }
10006
40659392 10007 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10008 super->updates_pending++;
10009 break;
10010 }
1a64be56 10011 case update_add_remove_disk: {
43dad3d6 10012 /* we may be able to repair some arrays if disks are
095b8088 10013 * being added, check the status of add_remove_disk
1a64be56
LM		 10014 * if discs has been added.
10015 */
10016 if (add_remove_disk_update(super)) {
43dad3d6 10017 struct active_array *a;
072b727f
DW		 10018
10019 super->updates_pending++;
1a64be56 10020 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10021 a->check_degraded = 1;
10022 }
43dad3d6 10023 break;
e8319a19 10024 }
bbab0940
TM		 10025 case update_prealloc_badblocks_mem:
10026 break;
e6e9dd3f
AP		 10027 case update_rwh_policy: {
10028 struct imsm_update_rwh_policy *u = (void *)update->buf;
10029 int target = u->dev_idx;
10030 struct imsm_dev *dev = get_imsm_dev(super, target);
10031 if (!dev) {
10032 dprintf("could not find subarray-%d\n", target);
10033 break;
10034 }
10035
10036 if (dev->rwh_policy != u->new_policy) {
10037 dev->rwh_policy = u->new_policy;
10038 super->updates_pending++;
10039 }
10040 break;
10041 }
1a64be56 10042 default:
ebf3be99 10043 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10044 }
e8319a19 10045}
88758e9d 10046
bc0b9d34
PC		 10047static struct mdinfo *get_spares_for_grow(struct supertype *st);
10048
5fe6f031
N		 10049static int imsm_prepare_update(struct supertype *st,
10050 struct metadata_update *update)
8273f55e 10051{
949c47a0 10052 /**
4d7b1503
DW		 10053 * Allocate space to hold new disk entries, raid-device entries or a new
10054 * mpb if necessary. The manager synchronously waits for updates to
10055 * complete in the monitor, so new mpb buffers allocated here can be
10056 * integrated by the monitor thread without worrying about live pointers
10057 * in the manager thread.
8273f55e 10058 */
095b8088 10059 enum imsm_update_type type;
4d7b1503 10060 struct intel_super *super = st->sb;
f36a9ecd 10061 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10062 struct imsm_super *mpb = super->anchor;
10063 size_t buf_len;
10064 size_t len = 0;
949c47a0 10065
095b8088
N		 10066 if (update->len < (int)sizeof(type))
10067 return 0;
10068
10069 type = *(enum imsm_update_type *) update->buf;
10070
949c47a0 10071 switch (type) {
0ec5d470 10072 case update_general_migration_checkpoint:
095b8088
N		 10073 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10074 return 0;
1ade5cc1 10075 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10076 break;
abedf5fc
KW		 10077 case update_takeover: {
10078 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10079 if (update->len < (int)sizeof(*u))
10080 return 0;
abedf5fc
KW		 10081 if (u->direction == R0_TO_R10) {
10082 void **tail = (void **)&update->space_list;
10083 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10084 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10085 int num_members = map->num_members;
10086 void *space;
10087 int size, i;
abedf5fc
KW		 10088 /* allocate memory for added disks */
10089 for (i = 0; i < num_members; i++) {
10090 size = sizeof(struct dl);
503975b9 10091 space = xmalloc(size);
abedf5fc
KW		 10092 *tail = space;
10093 tail = space;
10094 *tail = NULL;
10095 }
10096 /* allocate memory for new device */
10097 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10098 (num_members * sizeof(__u32));
503975b9
N		 10099 space = xmalloc(size);
10100 *tail = space;
10101 tail = space;
10102 *tail = NULL;
10103 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10104 }
10105
10106 break;
10107 }
78b10e66 10108 case update_reshape_container_disks: {
d195167d
AK		 10109 /* Every raid device in the container is about to
10110 * gain some more devices, and we will enter a
10111 * reconfiguration.
10112 * So each 'imsm_map' will be bigger, and the imsm_vol
10113 * will now hold 2 of them.
10114 * Thus we need new 'struct imsm_dev' allocations sized
10115 * as sizeof_imsm_dev but with more devices in both maps.
10116 */
10117 struct imsm_update_reshape *u = (void *)update->buf;
10118 struct intel_dev *dl;
10119 void **space_tail = (void**)&update->space_list;
10120
095b8088
N		 10121 if (update->len < (int)sizeof(*u))
10122 return 0;
10123
1ade5cc1 10124 dprintf("for update_reshape\n");
d195167d
AK		 10125
10126 for (dl = super->devlist; dl; dl = dl->next) {
10127 int size = sizeof_imsm_dev(dl->dev, 1);
10128 void *s;
d677e0b8
AK		 10129 if (u->new_raid_disks > u->old_raid_disks)
10130 size += sizeof(__u32)*2*
10131 (u->new_raid_disks - u->old_raid_disks);
503975b9 10132 s = xmalloc(size);
d195167d
AK		 10133 *space_tail = s;
10134 space_tail = s;
10135 *space_tail = NULL;
10136 }
10137
10138 len = disks_to_mpb_size(u->new_raid_disks);
10139 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10140 break;
10141 }
48c5303a 10142 case update_reshape_migration: {
bc0b9d34
PC		 10143 /* for migration level 0->5 we need to add disks
10144 * so the same as for container operation we will copy
10145 * device to the bigger location.
10146 * in memory prepared device and new disk area are prepared
10147 * for usage in process update
10148 */
10149 struct imsm_update_reshape_migration *u = (void *)update->buf;
10150 struct intel_dev *id;
10151 void **space_tail = (void **)&update->space_list;
10152 int size;
10153 void *s;
10154 int current_level = -1;
10155
095b8088
N		 10156 if (update->len < (int)sizeof(*u))
10157 return 0;
10158
1ade5cc1 10159 dprintf("for update_reshape\n");
bc0b9d34
PC		 10160
10161 /* add space for bigger array in update
10162 */
10163 for (id = super->devlist; id; id = id->next) {
10164 if (id->index == (unsigned)u->subdev) {
10165 size = sizeof_imsm_dev(id->dev, 1);
10166 if (u->new_raid_disks > u->old_raid_disks)
10167 size += sizeof(__u32)*2*
10168 (u->new_raid_disks - u->old_raid_disks);
503975b9 10169 s = xmalloc(size);
bc0b9d34
PC		 10170 *space_tail = s;
10171 space_tail = s;
10172 *space_tail = NULL;
10173 break;
10174 }
10175 }
10176 if (update->space_list == NULL)
10177 break;
10178
10179 /* add space for disk in update
10180 */
10181 size = sizeof(struct dl);
503975b9 10182 s = xmalloc(size);
bc0b9d34
PC		 10183 *space_tail = s;
10184 space_tail = s;
10185 *space_tail = NULL;
10186
10187 /* add spare device to update
10188 */
10189 for (id = super->devlist ; id; id = id->next)
10190 if (id->index == (unsigned)u->subdev) {
10191 struct imsm_dev *dev;
10192 struct imsm_map *map;
10193
10194 dev = get_imsm_dev(super, u->subdev);
238c0a71 10195 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10196 current_level = map->raid_level;
10197 break;
10198 }
089f9d79 10199 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10200 struct mdinfo *spares;
10201
10202 spares = get_spares_for_grow(st);
10203 if (spares) {
10204 struct dl *dl;
10205 struct mdinfo *dev;
10206
10207 dev = spares->devs;
10208 if (dev) {
10209 u->new_disks[0] =
10210 makedev(dev->disk.major,
10211 dev->disk.minor);
10212 dl = get_disk_super(super,
10213 dev->disk.major,
10214 dev->disk.minor);
10215 dl->index = u->old_raid_disks;
10216 dev = dev->next;
10217 }
10218 sysfs_free(spares);
10219 }
10220 }
10221 len = disks_to_mpb_size(u->new_raid_disks);
10222 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10223 break;
10224 }
f3871fdc 10225 case update_size_change: {
095b8088
N		 10226 if (update->len < (int)sizeof(struct imsm_update_size_change))
10227 return 0;
10228 break;
10229 }
10230 case update_activate_spare: {
10231 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10232 return 0;
f3871fdc
AK		 10233 break;
10234 }
949c47a0
DW		 10235 case update_create_array: {
10236 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10237 struct intel_dev *dv;
54c2c1ea 10238 struct imsm_dev *dev = &u->dev;
238c0a71 10239 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10240 struct dl *dl;
10241 struct disk_info *inf;
10242 int i;
10243 int activate = 0;
949c47a0 10244
095b8088
N		 10245 if (update->len < (int)sizeof(*u))
10246 return 0;
10247
54c2c1ea
DW		 10248 inf = get_disk_info(u);
10249 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10250 /* allocate a new super->devlist entry */
503975b9
N		 10251 dv = xmalloc(sizeof(*dv));
10252 dv->dev = xmalloc(len);
10253 update->space = dv;
949c47a0 10254
54c2c1ea
DW		 10255 /* count how many spares will be converted to members */
10256 for (i = 0; i < map->num_members; i++) {
10257 dl = serial_to_dl(inf[i].serial, super);
10258 if (!dl) {
10259 /* hmm maybe it failed?, nothing we can do about
10260 * it here
10261 */
10262 continue;
10263 }
10264 if (count_memberships(dl, super) == 0)
10265 activate++;
10266 }
10267 len += activate * sizeof(struct imsm_disk);
949c47a0 10268 break;
095b8088
N		 10269 }
10270 case update_kill_array: {
10271 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10272 return 0;
949c47a0
DW		 10273 break;
10274 }
095b8088
N		 10275 case update_rename_array: {
10276 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10277 return 0;
10278 break;
10279 }
10280 case update_add_remove_disk:
10281 /* no update->len needed */
10282 break;
bbab0940
TM		 10283 case update_prealloc_badblocks_mem:
10284 super->extra_space += sizeof(struct bbm_log) -
10285 get_imsm_bbm_log_size(super->bbm_log);
10286 break;
e6e9dd3f
AP		 10287 case update_rwh_policy: {
10288 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10289 return 0;
10290 break;
10291 }
095b8088
N		 10292 default:
10293 return 0;
949c47a0 10294 }
8273f55e 10295
4d7b1503
DW		 10296 /* check if we need a larger metadata buffer */
10297 if (super->next_buf)
10298 buf_len = super->next_len;
10299 else
10300 buf_len = super->len;
10301
bbab0940 10302 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10303 /* ok we need a larger buf than what is currently allocated
10304 * if this allocation fails process_update will notice that
10305 * ->next_len is set and ->next_buf is NULL
10306 */
bbab0940
TM		 10307 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10308 super->extra_space + len, sector_size);
4d7b1503
DW		 10309 if (super->next_buf)
10310 free(super->next_buf);
10311
10312 super->next_len = buf_len;
f36a9ecd 10313 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10314 memset(super->next_buf, 0, buf_len);
10315 else
4d7b1503
DW		 10316 super->next_buf = NULL;
10317 }
5fe6f031 10318 return 1;
8273f55e
DW		 10319}
10320
ae6aad82 10321/* must be called while manager is quiesced */
f21e18ca 10322static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10323{
10324 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10325 struct dl *iter;
10326 struct imsm_dev *dev;
10327 struct imsm_map *map;
4c9e8c1e 10328 unsigned int i, j, num_members;
fb12a745 10329 __u32 ord, ord_map0;
4c9e8c1e 10330 struct bbm_log *log = super->bbm_log;
ae6aad82 10331
1ade5cc1 10332 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10333
10334 /* shift all indexes down one */
10335 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10336 if (iter->index > (int)index)
ae6aad82 10337 iter->index--;
47ee5a45 10338 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10339 if (iter->index > (int)index)
47ee5a45 10340 iter->index--;
ae6aad82
DW		 10341
10342 for (i = 0; i < mpb->num_raid_devs; i++) {
10343 dev = get_imsm_dev(super, i);
238c0a71 10344 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10345 num_members = map->num_members;
10346 for (j = 0; j < num_members; j++) {
10347 /* update ord entries being careful not to propagate
10348 * ord-flags to the first map
10349 */
238c0a71 10350 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10351 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10352
24565c9a
DW		 10353 if (ord_to_idx(ord) <= index)
10354 continue;
ae6aad82 10355
238c0a71 10356 map = get_imsm_map(dev, MAP_0);
fb12a745 10357 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10358 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10359 if (map)
10360 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10361 }
10362 }
10363
4c9e8c1e
TM		 10364 for (i = 0; i < log->entry_count; i++) {
10365 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10366
10367 if (entry->disk_ordinal <= index)
10368 continue;
10369 entry->disk_ordinal--;
10370 }
10371
ae6aad82
DW		 10372 mpb->num_disks--;
10373 super->updates_pending++;
24565c9a
DW		 10374 if (*dlp) {
10375 struct dl *dl = *dlp;
10376
10377 *dlp = (*dlp)->next;
10378 __free_imsm_disk(dl);
10379 }
ae6aad82 10380}
9a717282
AK		 10381
10382static void close_targets(int *targets, int new_disks)
10383{
10384 int i;
10385
10386 if (!targets)
10387 return;
10388
10389 for (i = 0; i < new_disks; i++) {
10390 if (targets[i] >= 0) {
10391 close(targets[i]);
10392 targets[i] = -1;
10393 }
10394 }
10395}
10396
10397static int imsm_get_allowed_degradation(int level, int raid_disks,
10398 struct intel_super *super,
10399 struct imsm_dev *dev)
10400{
10401 switch (level) {
bf5cf7c7 10402 case 1:
9a717282
AK		 10403 case 10:{
10404 int ret_val = 0;
10405 struct imsm_map *map;
10406 int i;
10407
10408 ret_val = raid_disks/2;
10409 /* check map if all disks pairs not failed
10410 * in both maps
10411 */
238c0a71 10412 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10413 for (i = 0; i < ret_val; i++) {
10414 int degradation = 0;
10415 if (get_imsm_disk(super, i) == NULL)
10416 degradation++;
10417 if (get_imsm_disk(super, i + 1) == NULL)
10418 degradation++;
10419 if (degradation == 2)
10420 return 0;
10421 }
238c0a71 10422 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10423 /* if there is no second map
10424 * result can be returned
10425 */
10426 if (map == NULL)
10427 return ret_val;
10428 /* check degradation in second map
10429 */
10430 for (i = 0; i < ret_val; i++) {
10431 int degradation = 0;
10432 if (get_imsm_disk(super, i) == NULL)
10433 degradation++;
10434 if (get_imsm_disk(super, i + 1) == NULL)
10435 degradation++;
10436 if (degradation == 2)
10437 return 0;
10438 }
10439 return ret_val;
10440 }
10441 case 5:
10442 return 1;
10443 case 6:
10444 return 2;
10445 default:
10446 return 0;
10447 }
10448}
10449
687629c2
AK		 10450/*******************************************************************************
10451 * Function: open_backup_targets
10452 * Description: Function opens file descriptors for all devices given in
10453 * info->devs
10454 * Parameters:
10455 * info : general array info
10456 * raid_disks : number of disks
10457 * raid_fds : table of device's file descriptors
9a717282
AK		 10458 * super : intel super for raid10 degradation check
10459 * dev : intel device for raid10 degradation check
687629c2
AK		 10460 * Returns:
10461 * 0 : success
10462 * -1 : fail
10463 ******************************************************************************/
9a717282
AK		 10464int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10465 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10466{
10467 struct mdinfo *sd;
f627f5ad 10468 int i;
9a717282 10469 int opened = 0;
f627f5ad
AK		 10470
10471 for (i = 0; i < raid_disks; i++)
10472 raid_fds[i] = -1;
687629c2
AK		 10473
10474 for (sd = info->devs ; sd ; sd = sd->next) {
10475 char *dn;
10476
10477 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10478 dprintf("disk is faulty!!\n");
10479 continue;
10480 }
10481
089f9d79 10482 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10483 continue;
10484
10485 dn = map_dev(sd->disk.major,
10486 sd->disk.minor, 1);
10487 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10488 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10489 pr_err("cannot open component\n");
9a717282 10490 continue;
687629c2 10491 }
9a717282
AK		 10492 opened++;
10493 }
10494 /* check if maximum array degradation level is not exceeded
10495 */
10496 if ((raid_disks - opened) >
089f9d79
JS		 10497 imsm_get_allowed_degradation(info->new_level, raid_disks,
10498 super, dev)) {
e12b3daa 10499 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10500 close_targets(raid_fds, raid_disks);
10501 return -2;
687629c2
AK		 10502 }
10503 return 0;
10504}
10505
d31ad643
PB		 10506/*******************************************************************************
10507 * Function: validate_container_imsm
10508 * Description: This routine validates container after assemble,
10509 * eg. if devices in container are under the same controller.
10510 *
10511 * Parameters:
10512 * info : linked list with info about devices used in array
10513 * Returns:
10514 * 1 : HBA mismatch
10515 * 0 : Success
10516 ******************************************************************************/
10517int validate_container_imsm(struct mdinfo *info)
10518{
6b781d33
AP		 10519 if (check_env("IMSM_NO_PLATFORM"))
10520 return 0;
d31ad643 10521
6b781d33
AP		 10522 struct sys_dev *idev;
10523 struct sys_dev *hba = NULL;
10524 struct sys_dev *intel_devices = find_intel_devices();
10525 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10526 info->disk.minor));
10527
10528 for (idev = intel_devices; idev; idev = idev->next) {
10529 if (dev_path && strstr(dev_path, idev->path)) {
10530 hba = idev;
10531 break;
d31ad643 10532 }
6b781d33
AP		 10533 }
10534 if (dev_path)
d31ad643
PB		 10535 free(dev_path);
10536
6b781d33
AP		 10537 if (!hba) {
10538 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10539 devid2kname(makedev(info->disk.major, info->disk.minor)));
10540 return 1;
10541 }
10542
10543 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10544 struct mdinfo *dev;
10545
10546 for (dev = info->next; dev; dev = dev->next) {
10547 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10548
10549 struct sys_dev *hba2 = NULL;
10550 for (idev = intel_devices; idev; idev = idev->next) {
10551 if (dev_path && strstr(dev_path, idev->path)) {
10552 hba2 = idev;
10553 break;
d31ad643
PB		 10554 }
10555 }
6b781d33
AP		 10556 if (dev_path)
10557 free(dev_path);
10558
10559 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10560 get_orom_by_device_id(hba2->dev_id);
10561
10562 if (hba2 && hba->type != hba2->type) {
10563 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10564 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10565 return 1;
10566 }
10567
07cb1e57 10568 if (orom != orom2) {
6b781d33
AP		 10569 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10570 " This operation is not supported and can lead to data loss.\n");
10571 return 1;
10572 }
10573
10574 if (!orom) {
10575 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10576 " This operation is not supported and can lead to data loss.\n");
10577 return 1;
10578 }
d31ad643 10579 }
6b781d33 10580
d31ad643
PB		 10581 return 0;
10582}
32141c17 10583
6f50473f
TM		 10584/*******************************************************************************
10585* Function: imsm_record_badblock
10586* Description: This routine stores new bad block record in BBM log
10587*
10588* Parameters:
10589* a : array containing a bad block
10590* slot : disk number containing a bad block
10591* sector : bad block sector
10592* length : bad block sectors range
10593* Returns:
10594* 1 : Success
10595* 0 : Error
10596******************************************************************************/
10597static int imsm_record_badblock(struct active_array *a, int slot,
10598 unsigned long long sector, int length)
10599{
10600 struct intel_super *super = a->container->sb;
10601 int ord;
10602 int ret;
10603
10604 ord = imsm_disk_slot_to_ord(a, slot);
10605 if (ord < 0)
10606 return 0;
10607
10608 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10609 length);
10610 if (ret)
10611 super->updates_pending++;
10612
10613 return ret;
10614}
c07a5a4f
TM		 10615/*******************************************************************************
10616* Function: imsm_clear_badblock
10617* Description: This routine clears bad block record from BBM log
10618*
10619* Parameters:
10620* a : array containing a bad block
10621* slot : disk number containing a bad block
10622* sector : bad block sector
10623* length : bad block sectors range
10624* Returns:
10625* 1 : Success
10626* 0 : Error
10627******************************************************************************/
10628static int imsm_clear_badblock(struct active_array *a, int slot,
10629 unsigned long long sector, int length)
10630{
10631 struct intel_super *super = a->container->sb;
10632 int ord;
10633 int ret;
10634
10635 ord = imsm_disk_slot_to_ord(a, slot);
10636 if (ord < 0)
10637 return 0;
10638
10639 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10640 if (ret)
10641 super->updates_pending++;
10642
10643 return ret;
10644}
928f1424
TM		 10645/*******************************************************************************
10646* Function: imsm_get_badblocks
10647* Description: This routine get list of bad blocks for an array
10648*
10649* Parameters:
10650* a : array
10651* slot : disk number
10652* Returns:
10653* bb : structure containing bad blocks
10654* NULL : error
10655******************************************************************************/
10656static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10657{
10658 int inst = a->info.container_member;
10659 struct intel_super *super = a->container->sb;
10660 struct imsm_dev *dev = get_imsm_dev(super, inst);
10661 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10662 int ord;
10663
10664 ord = imsm_disk_slot_to_ord(a, slot);
10665 if (ord < 0)
10666 return NULL;
10667
10668 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10669 per_dev_array_size(map), &super->bb);
928f1424
TM		 10670
10671 return &super->bb;
10672}
27156a57
TM		 10673/*******************************************************************************
10674* Function: examine_badblocks_imsm
10675* Description: Prints list of bad blocks on a disk to the standard output
10676*
10677* Parameters:
10678* st : metadata handler
10679* fd : open file descriptor for device
10680* devname : device name
10681* Returns:
10682* 0 : Success
10683* 1 : Error
10684******************************************************************************/
10685static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10686{
10687 struct intel_super *super = st->sb;
10688 struct bbm_log *log = super->bbm_log;
10689 struct dl *d = NULL;
10690 int any = 0;
10691
10692 for (d = super->disks; d ; d = d->next) {
10693 if (strcmp(d->devname, devname) == 0)
10694 break;
10695 }
10696
10697 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10698 pr_err("%s doesn't appear to be part of a raid array\n",
10699 devname);
10700 return 1;
10701 }
10702
10703 if (log != NULL) {
10704 unsigned int i;
10705 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10706
10707 for (i = 0; i < log->entry_count; i++) {
10708 if (entry[i].disk_ordinal == d->index) {
10709 unsigned long long sector = __le48_to_cpu(
10710 &entry[i].defective_block_start);
10711 int cnt = entry[i].marked_count + 1;
10712
10713 if (!any) {
10714 printf("Bad-blocks on %s:\n", devname);
10715 any = 1;
10716 }
10717
10718 printf("%20llu for %d sectors\n", sector, cnt);
10719 }
10720 }
10721 }
10722
10723 if (!any)
10724 printf("No bad-blocks list configured on %s\n", devname);
10725
10726 return 0;
10727}
687629c2
AK		 10728/*******************************************************************************
10729 * Function: init_migr_record_imsm
10730 * Description: Function inits imsm migration record
10731 * Parameters:
10732 * super : imsm internal array info
10733 * dev : device under migration
10734 * info : general array info to find the smallest device
10735 * Returns:
10736 * none
10737 ******************************************************************************/
10738void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10739 struct mdinfo *info)
10740{
10741 struct intel_super *super = st->sb;
10742 struct migr_record *migr_rec = super->migr_rec;
10743 int new_data_disks;
10744 unsigned long long dsize, dev_sectors;
10745 long long unsigned min_dev_sectors = -1LLU;
10746 struct mdinfo *sd;
10747 char nm[30];
10748 int fd;
238c0a71
AK		 10749 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10750 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10751 unsigned long long num_migr_units;
3ef4403c 10752 unsigned long long array_blocks;
687629c2
AK		 10753
10754 memset(migr_rec, 0, sizeof(struct migr_record));
10755 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10756
10757 /* only ascending reshape supported now */
10758 migr_rec->ascending_migr = __cpu_to_le32(1);
10759
10760 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10761 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10762 migr_rec->dest_depth_per_unit *=
10763 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10764 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10765 migr_rec->blocks_per_unit =
10766 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10767 migr_rec->dest_depth_per_unit =
10768 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10769 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10770 num_migr_units =
10771 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10772
10773 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10774 num_migr_units++;
9f421827 10775 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10776
10777 migr_rec->post_migr_vol_cap = dev->size_low;
10778 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10779
687629c2
AK		 10780 /* Find the smallest dev */
10781 for (sd = info->devs ; sd ; sd = sd->next) {
10782 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10783 fd = dev_open(nm, O_RDONLY);
10784 if (fd < 0)
10785 continue;
10786 get_dev_size(fd, NULL, &dsize);
10787 dev_sectors = dsize / 512;
10788 if (dev_sectors < min_dev_sectors)
10789 min_dev_sectors = dev_sectors;
10790 close(fd);
10791 }
9f421827 10792 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10793 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10794
10795 write_imsm_migr_rec(st);
10796
10797 return;
10798}
10799
10800/*******************************************************************************
10801 * Function: save_backup_imsm
10802 * Description: Function saves critical data stripes to Migration Copy Area
10803 * and updates the current migration unit status.
10804 * Use restore_stripes() to form a destination stripe,
10805 * and to write it to the Copy Area.
10806 * Parameters:
10807 * st : supertype information
aea93171 10808 * dev : imsm device that backup is saved for
687629c2
AK		 10809 * info : general array info
10810 * buf : input buffer
687629c2
AK		 10811 * length : length of data to backup (blocks_per_unit)
10812 * Returns:
10813 * 0 : success
10814 *, -1 : fail
10815 ******************************************************************************/
10816int save_backup_imsm(struct supertype *st,
10817 struct imsm_dev *dev,
10818 struct mdinfo *info,
10819 void *buf,
687629c2
AK		 10820 int length)
10821{
10822 int rv = -1;
10823 struct intel_super *super = st->sb;
594dc1b8
JS		 10824 unsigned long long *target_offsets;
10825 int *targets;
687629c2 10826 int i;
238c0a71 10827 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10828 int new_disks = map_dest->num_members;
ab724b98
AK		 10829 int dest_layout = 0;
10830 int dest_chunk;
d1877f69 10831 unsigned long long start;
9529d343 10832 int data_disks = imsm_num_data_members(map_dest);
687629c2 10833
503975b9 10834 targets = xmalloc(new_disks * sizeof(int));
687629c2 10835
7e45b550
AK		 10836 for (i = 0; i < new_disks; i++)
10837 targets[i] = -1;
10838
503975b9 10839 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10840
d1877f69 10841 start = info->reshape_progress * 512;
687629c2 10842 for (i = 0; i < new_disks; i++) {
9f421827 10843 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10844 /* move back copy area adderss, it will be moved forward
10845 * in restore_stripes() using start input variable
10846 */
10847 target_offsets[i] -= start/data_disks;
687629c2
AK		 10848 }
10849
9a717282
AK		 10850 if (open_backup_targets(info, new_disks, targets,
10851 super, dev))
687629c2
AK		 10852 goto abort;
10853
68eb8bc6 10854 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10855 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10856
687629c2
AK		 10857 if (restore_stripes(targets, /* list of dest devices */
10858 target_offsets, /* migration record offsets */
10859 new_disks,
ab724b98
AK		 10860 dest_chunk,
10861 map_dest->raid_level,
10862 dest_layout,
10863 -1, /* source backup file descriptor */
10864 0, /* input buf offset
10865 * always 0 buf is already offseted */
d1877f69 10866 start,
687629c2
AK		 10867 length,
10868 buf) != 0) {
e7b84f9d 10869 pr_err("Error restoring stripes\n");
687629c2
AK		 10870 goto abort;
10871 }
10872
10873 rv = 0;
10874
10875abort:
10876 if (targets) {
9a717282 10877 close_targets(targets, new_disks);
687629c2
AK		 10878 free(targets);
10879 }
10880 free(target_offsets);
10881
10882 return rv;
10883}
10884
10885/*******************************************************************************
10886 * Function: save_checkpoint_imsm
10887 * Description: Function called for current unit status update
10888 * in the migration record. It writes it to disk.
10889 * Parameters:
10890 * super : imsm internal array info
10891 * info : general array info
10892 * Returns:
10893 * 0: success
10894 * 1: failure
0228d92c
AK		 10895 * 2: failure, means no valid migration record
10896 * / no general migration in progress /
687629c2
AK		 10897 ******************************************************************************/
10898int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10899{
10900 struct intel_super *super = st->sb;
f8b72ef5
AK		 10901 unsigned long long blocks_per_unit;
10902 unsigned long long curr_migr_unit;
10903
2e062e82 10904 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10905 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10906 return 1;
10907 }
10908
f8b72ef5
AK		 10909 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10910 if (blocks_per_unit == 0) {
0228d92c
AK		 10911 dprintf("imsm: no migration in progress.\n");
10912 return 2;
687629c2 10913 }
f8b72ef5
AK		 10914 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10915 /* check if array is alligned to copy area
10916 * if it is not alligned, add one to current migration unit value
10917 * this can happend on array reshape finish only
10918 */
10919 if (info->reshape_progress % blocks_per_unit)
10920 curr_migr_unit++;
687629c2 10921
9f421827 10922 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10923 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10924 set_migr_dest_1st_member_lba(super->migr_rec,
10925 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10926
687629c2 10927 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10928 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10929 return 1;
10930 }
10931
10932 return 0;
10933}
10934
276d77db
AK		 10935/*******************************************************************************
10936 * Function: recover_backup_imsm
10937 * Description: Function recovers critical data from the Migration Copy Area
10938 * while assembling an array.
10939 * Parameters:
10940 * super : imsm internal array info
10941 * info : general array info
10942 * Returns:
10943 * 0 : success (or there is no data to recover)
10944 * 1 : fail
10945 ******************************************************************************/
10946int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10947{
10948 struct intel_super *super = st->sb;
10949 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10950 struct imsm_map *map_dest;
276d77db
AK		 10951 struct intel_dev *id = NULL;
10952 unsigned long long read_offset;
10953 unsigned long long write_offset;
10954 unsigned unit_len;
10955 int *targets = NULL;
10956 int new_disks, i, err;
10957 char *buf = NULL;
10958 int retval = 1;
f36a9ecd 10959 unsigned int sector_size = super->sector_size;
9f421827
PB		 10960 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10961 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10962 char buffer[20];
6c3560c0 10963 int skipped_disks = 0;
276d77db
AK		 10964
10965 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10966 if (err < 1)
10967 return 1;
10968
10969 /* recover data only during assemblation */
10970 if (strncmp(buffer, "inactive", 8) != 0)
10971 return 0;
10972 /* no data to recover */
10973 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10974 return 0;
10975 if (curr_migr_unit >= num_migr_units)
10976 return 1;
10977
10978 /* find device during reshape */
10979 for (id = super->devlist; id; id = id->next)
10980 if (is_gen_migration(id->dev))
10981 break;
10982 if (id == NULL)
10983 return 1;
10984
238c0a71 10985 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10986 new_disks = map_dest->num_members;
10987
9f421827 10988 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 10989
9f421827 10990 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 10991 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10992
10993 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10994 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10995 goto abort;
503975b9 10996 targets = xcalloc(new_disks, sizeof(int));
276d77db 10997
9a717282 10998 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10999 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 11000 goto abort;
11001 }
276d77db
AK		 11002
11003 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 11004 if (targets[i] < 0) {
11005 skipped_disks++;
11006 continue;
11007 }
276d77db 11008 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11009 pr_err("Cannot seek to block: %s\n",
11010 strerror(errno));
137debce
AK		 11011 skipped_disks++;
11012 continue;
276d77db 11013 }
9ec11d1a 11014 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11015 pr_err("Cannot read copy area block: %s\n",
11016 strerror(errno));
137debce
AK		 11017 skipped_disks++;
11018 continue;
276d77db
AK		 11019 }
11020 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11021 pr_err("Cannot seek to block: %s\n",
11022 strerror(errno));
137debce
AK		 11023 skipped_disks++;
11024 continue;
276d77db 11025 }
9ec11d1a 11026 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11027 pr_err("Cannot restore block: %s\n",
11028 strerror(errno));
137debce
AK		 11029 skipped_disks++;
11030 continue;
276d77db
AK		 11031 }
11032 }
11033
137debce
AK		 11034 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11035 new_disks,
11036 super,
11037 id->dev)) {
7a862a02 11038 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11039 goto abort;
11040 }
11041
befb629b
AK		 11042 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11043 /* ignore error == 2, this can mean end of reshape here
11044 */
7a862a02 11045 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11046 } else
276d77db 11047 retval = 0;
276d77db
AK		 11048
11049abort:
11050 if (targets) {
11051 for (i = 0; i < new_disks; i++)
11052 if (targets[i])
11053 close(targets[i]);
11054 free(targets);
11055 }
11056 free(buf);
11057 return retval;
11058}
11059
2cda7640
ML		 11060static char disk_by_path[] = "/dev/disk/by-path/";
11061
11062static const char *imsm_get_disk_controller_domain(const char *path)
11063{
2cda7640 11064 char disk_path[PATH_MAX];
96234762
LM		 11065 char *drv=NULL;
11066 struct stat st;
2cda7640 11067
6d8d290a 11068 strcpy(disk_path, disk_by_path);
96234762
LM		 11069 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11070 if (stat(disk_path, &st) == 0) {
11071 struct sys_dev* hba;
594dc1b8 11072 char *path;
96234762
LM		 11073
11074 path = devt_to_devpath(st.st_rdev);
11075 if (path == NULL)
11076 return "unknown";
11077 hba = find_disk_attached_hba(-1, path);
11078 if (hba && hba->type == SYS_DEV_SAS)
11079 drv = "isci";
11080 else if (hba && hba->type == SYS_DEV_SATA)
11081 drv = "ahci";
c6839718
MT		 11082 else if (hba && hba->type == SYS_DEV_VMD)
11083 drv = "vmd";
11084 else if (hba && hba->type == SYS_DEV_NVME)
11085 drv = "nvme";
1011e834 11086 else
96234762
LM		 11087 drv = "unknown";
11088 dprintf("path: %s hba: %s attached: %s\n",
11089 path, (hba) ? hba->path : "NULL", drv);
11090 free(path);
2cda7640 11091 }
96234762 11092 return drv;
2cda7640
ML		 11093}
11094
4dd2df09 11095static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11096{
4dd2df09 11097 static char devnm[32];
78b10e66
N		 11098 char subdev_name[20];
11099 struct mdstat_ent *mdstat;
11100
11101 sprintf(subdev_name, "%d", subdev);
11102 mdstat = mdstat_by_subdev(subdev_name, container);
11103 if (!mdstat)
4dd2df09 11104 return NULL;
78b10e66 11105
4dd2df09 11106 strcpy(devnm, mdstat->devnm);
78b10e66 11107 free_mdstat(mdstat);
4dd2df09 11108 return devnm;
78b10e66
N		 11109}
11110
11111static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11112 struct geo_params *geo,
fbf3d202
AK		 11113 int *old_raid_disks,
11114 int direction)
78b10e66 11115{
694575e7
KW		 11116 /* currently we only support increasing the number of devices
11117 * for a container. This increases the number of device for each
11118 * member array. They must all be RAID0 or RAID5.
11119 */
78b10e66
N		 11120 int ret_val = 0;
11121 struct mdinfo *info, *member;
11122 int devices_that_can_grow = 0;
11123
7a862a02 11124 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11125
d04f65f4 11126 if (geo->size > 0 ||
78b10e66
N		 11127 geo->level != UnSet ||
11128 geo->layout != UnSet ||
11129 geo->chunksize != 0 ||
11130 geo->raid_disks == UnSet) {
7a862a02 11131 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11132 return ret_val;
11133 }
11134
fbf3d202 11135 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11136 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11137 return ret_val;
11138 }
11139
78b10e66
N		 11140 info = container_content_imsm(st, NULL);
11141 for (member = info; member; member = member->next) {
4dd2df09 11142 char *result;
78b10e66
N		 11143
11144 dprintf("imsm: checking device_num: %i\n",
11145 member->container_member);
11146
d7d205bd 11147 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11148 /* we work on container for Online Capacity Expansion
11149 * only so raid_disks has to grow
11150 */
7a862a02 11151 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11152 break;
11153 }
11154
089f9d79 11155 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11156 /* we cannot use this container with other raid level
11157 */
7a862a02 11158 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11159 info->array.level);
11160 break;
11161 } else {
11162 /* check for platform support
11163 * for this raid level configuration
11164 */
11165 struct intel_super *super = st->sb;
11166 if (!is_raid_level_supported(super->orom,
11167 member->array.level,
11168 geo->raid_disks)) {
7a862a02 11169 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11170 info->array.level,
11171 geo->raid_disks,
11172 geo->raid_disks > 1 ? "s" : "");
11173 break;
11174 }
2a4a08e7
AK		 11175 /* check if component size is aligned to chunk size
11176 */
11177 if (info->component_size %
11178 (info->array.chunk_size/512)) {
7a862a02 11179 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11180 break;
11181 }
78b10e66
N		 11182 }
11183
11184 if (*old_raid_disks &&
11185 info->array.raid_disks != *old_raid_disks)
11186 break;
11187 *old_raid_disks = info->array.raid_disks;
11188
11189 /* All raid5 and raid0 volumes in container
11190 * have to be ready for Online Capacity Expansion
11191 * so they need to be assembled. We have already
11192 * checked that no recovery etc is happening.
11193 */
4dd2df09
N		 11194 result = imsm_find_array_devnm_by_subdev(member->container_member,
11195 st->container_devnm);
11196 if (result == NULL) {
78b10e66
N		 11197 dprintf("imsm: cannot find array\n");
11198 break;
11199 }
11200 devices_that_can_grow++;
11201 }
11202 sysfs_free(info);
11203 if (!member && devices_that_can_grow)
11204 ret_val = 1;
11205
11206 if (ret_val)
1ade5cc1 11207 dprintf("Container operation allowed\n");
78b10e66 11208 else
1ade5cc1 11209 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11210
11211 return ret_val;
11212}
11213
11214/* Function: get_spares_for_grow
11215 * Description: Allocates memory and creates list of spare devices
1011e834 11216 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11217 * Parameters: Pointer to the supertype structure
11218 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11219 * NULL if fail
78b10e66
N		 11220 */
11221static struct mdinfo *get_spares_for_grow(struct supertype *st)
11222{
fbfdcb06
AO		 11223 struct spare_criteria sc;
11224
11225 get_spare_criteria_imsm(st, &sc);
11226 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11227}
11228
11229/******************************************************************************
11230 * function: imsm_create_metadata_update_for_reshape
11231 * Function creates update for whole IMSM container.
11232 *
11233 ******************************************************************************/
11234static int imsm_create_metadata_update_for_reshape(
11235 struct supertype *st,
11236 struct geo_params *geo,
11237 int old_raid_disks,
11238 struct imsm_update_reshape **updatep)
11239{
11240 struct intel_super *super = st->sb;
11241 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11242 int update_memory_size;
11243 struct imsm_update_reshape *u;
11244 struct mdinfo *spares;
78b10e66 11245 int i;
594dc1b8 11246 int delta_disks;
bbd24d86 11247 struct mdinfo *dev;
78b10e66 11248
1ade5cc1 11249 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11250
11251 delta_disks = geo->raid_disks - old_raid_disks;
11252
11253 /* size of all update data without anchor */
11254 update_memory_size = sizeof(struct imsm_update_reshape);
11255
11256 /* now add space for spare disks that we need to add. */
11257 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11258
503975b9 11259 u = xcalloc(1, update_memory_size);
78b10e66
N		 11260 u->type = update_reshape_container_disks;
11261 u->old_raid_disks = old_raid_disks;
11262 u->new_raid_disks = geo->raid_disks;
11263
11264 /* now get spare disks list
11265 */
11266 spares = get_spares_for_grow(st);
11267
d7be7d87 11268 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11269 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11270 i = -1;
78b10e66
N		 11271 goto abort;
11272 }
11273
11274 /* we have got spares
11275 * update disk list in imsm_disk list table in anchor
11276 */
11277 dprintf("imsm: %i spares are available.\n\n",
11278 spares->array.spare_disks);
11279
bbd24d86 11280 dev = spares->devs;
78b10e66 11281 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11282 struct dl *dl;
11283
bbd24d86
AK		 11284 if (dev == NULL)
11285 break;
78b10e66
N		 11286 u->new_disks[i] = makedev(dev->disk.major,
11287 dev->disk.minor);
11288 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11289 dl->index = mpb->num_disks;
11290 mpb->num_disks++;
bbd24d86 11291 dev = dev->next;
78b10e66 11292 }
78b10e66
N		 11293
11294abort:
11295 /* free spares
11296 */
11297 sysfs_free(spares);
11298
d677e0b8 11299 dprintf("imsm: reshape update preparation :");
78b10e66 11300 if (i == delta_disks) {
1ade5cc1 11301 dprintf_cont(" OK\n");
78b10e66
N		 11302 *updatep = u;
11303 return update_memory_size;
11304 }
11305 free(u);
1ade5cc1 11306 dprintf_cont(" Error\n");
78b10e66
N		 11307
11308 return 0;
11309}
11310
f3871fdc
AK		 11311/******************************************************************************
11312 * function: imsm_create_metadata_update_for_size_change()
11313 * Creates update for IMSM array for array size change.
11314 *
11315 ******************************************************************************/
11316static int imsm_create_metadata_update_for_size_change(
11317 struct supertype *st,
11318 struct geo_params *geo,
11319 struct imsm_update_size_change **updatep)
11320{
11321 struct intel_super *super = st->sb;
594dc1b8
JS		 11322 int update_memory_size;
11323 struct imsm_update_size_change *u;
f3871fdc 11324
1ade5cc1 11325 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11326
11327 /* size of all update data without anchor */
11328 update_memory_size = sizeof(struct imsm_update_size_change);
11329
503975b9 11330 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11331 u->type = update_size_change;
11332 u->subdev = super->current_vol;
11333 u->new_size = geo->size;
11334
11335 dprintf("imsm: reshape update preparation : OK\n");
11336 *updatep = u;
11337
11338 return update_memory_size;
11339}
11340
48c5303a
PC		 11341/******************************************************************************
11342 * function: imsm_create_metadata_update_for_migration()
11343 * Creates update for IMSM array.
11344 *
11345 ******************************************************************************/
11346static int imsm_create_metadata_update_for_migration(
11347 struct supertype *st,
11348 struct geo_params *geo,
11349 struct imsm_update_reshape_migration **updatep)
11350{
11351 struct intel_super *super = st->sb;
594dc1b8
JS		 11352 int update_memory_size;
11353 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11354 struct imsm_dev *dev;
11355 int previous_level = -1;
11356
1ade5cc1 11357 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11358
11359 /* size of all update data without anchor */
11360 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11361
503975b9 11362 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11363 u->type = update_reshape_migration;
11364 u->subdev = super->current_vol;
11365 u->new_level = geo->level;
11366 u->new_layout = geo->layout;
11367 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11368 u->new_disks[0] = -1;
4bba0439 11369 u->new_chunksize = -1;
48c5303a
PC		 11370
11371 dev = get_imsm_dev(super, u->subdev);
11372 if (dev) {
11373 struct imsm_map *map;
11374
238c0a71 11375 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11376 if (map) {
11377 int current_chunk_size =
11378 __le16_to_cpu(map->blocks_per_strip) / 2;
11379
11380 if (geo->chunksize != current_chunk_size) {
11381 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11382 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11383 current_chunk_size, u->new_chunksize);
11384 }
48c5303a 11385 previous_level = map->raid_level;
4bba0439 11386 }
48c5303a 11387 }
089f9d79 11388 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11389 struct mdinfo *spares = NULL;
11390
11391 u->new_raid_disks++;
11392 spares = get_spares_for_grow(st);
089f9d79 11393 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11394 free(u);
11395 sysfs_free(spares);
11396 update_memory_size = 0;
565cc99e 11397 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11398 return 0;
11399 }
11400 sysfs_free(spares);
11401 }
11402 dprintf("imsm: reshape update preparation : OK\n");
11403 *updatep = u;
11404
11405 return update_memory_size;
11406}
11407
8dd70bce
AK		 11408static void imsm_update_metadata_locally(struct supertype *st,
11409 void *buf, int len)
11410{
11411 struct metadata_update mu;
11412
11413 mu.buf = buf;
11414 mu.len = len;
11415 mu.space = NULL;
11416 mu.space_list = NULL;
11417 mu.next = NULL;
5fe6f031
N		 11418 if (imsm_prepare_update(st, &mu))
11419 imsm_process_update(st, &mu);
8dd70bce
AK		 11420
11421 while (mu.space_list) {
11422 void **space = mu.space_list;
11423 mu.space_list = *space;
11424 free(space);
11425 }
11426}
78b10e66 11427
471bceb6 11428/***************************************************************************
694575e7 11429* Function: imsm_analyze_change
471bceb6 11430* Description: Function analyze change for single volume
1011e834 11431* and validate if transition is supported
fbf3d202
AK		 11432* Parameters: Geometry parameters, supertype structure,
11433* metadata change direction (apply/rollback)
694575e7 11434* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11435****************************************************************************/
11436enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11437 struct geo_params *geo,
11438 int direction)
694575e7 11439{
471bceb6
KW		 11440 struct mdinfo info;
11441 int change = -1;
11442 int check_devs = 0;
c21e737b 11443 int chunk;
67a2db32
AK		 11444 /* number of added/removed disks in operation result */
11445 int devNumChange = 0;
11446 /* imsm compatible layout value for array geometry verification */
11447 int imsm_layout = -1;
7abc9871
AK		 11448 int data_disks;
11449 struct imsm_dev *dev;
9529d343 11450 struct imsm_map *map;
7abc9871 11451 struct intel_super *super;
d04f65f4 11452 unsigned long long current_size;
65d38cca 11453 unsigned long long free_size;
d04f65f4 11454 unsigned long long max_size;
65d38cca 11455 int rv;
471bceb6
KW		 11456
11457 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11458 if (geo->level != info.array.level && geo->level >= 0 &&
11459 geo->level != UnSet) {
471bceb6
KW		 11460 switch (info.array.level) {
11461 case 0:
11462 if (geo->level == 5) {
b5347799 11463 change = CH_MIGRATION;
e13ce846 11464 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11465 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11466 change = -1;
11467 goto analyse_change_exit;
11468 }
67a2db32 11469 imsm_layout = geo->layout;
471bceb6 11470 check_devs = 1;
e91a3bad
LM		 11471 devNumChange = 1; /* parity disk added */
11472 } else if (geo->level == 10) {
471bceb6
KW		 11473 change = CH_TAKEOVER;
11474 check_devs = 1;
e91a3bad 11475 devNumChange = 2; /* two mirrors added */
67a2db32 11476 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11477 }
dfe77a9e
KW		 11478 break;
11479 case 1:
471bceb6
KW		 11480 case 10:
11481 if (geo->level == 0) {
11482 change = CH_TAKEOVER;
11483 check_devs = 1;
e91a3bad 11484 devNumChange = -(geo->raid_disks/2);
67a2db32 11485 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11486 }
11487 break;
11488 }
11489 if (change == -1) {
7a862a02 11490 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11491 info.array.level, geo->level);
471bceb6
KW		 11492 goto analyse_change_exit;
11493 }
11494 } else
11495 geo->level = info.array.level;
11496
089f9d79
JS		 11497 if (geo->layout != info.array.layout &&
11498 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11499 change = CH_MIGRATION;
089f9d79
JS		 11500 if (info.array.layout == 0 && info.array.level == 5 &&
11501 geo->layout == 5) {
471bceb6 11502 /* reshape 5 -> 4 */
089f9d79
JS		 11503 } else if (info.array.layout == 5 && info.array.level == 5 &&
11504 geo->layout == 0) {
471bceb6
KW		 11505 /* reshape 4 -> 5 */
11506 geo->layout = 0;
11507 geo->level = 5;
11508 } else {
7a862a02 11509 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11510 info.array.layout, geo->layout);
471bceb6
KW		 11511 change = -1;
11512 goto analyse_change_exit;
11513 }
67a2db32 11514 } else {
471bceb6 11515 geo->layout = info.array.layout;
67a2db32
AK		 11516 if (imsm_layout == -1)
11517 imsm_layout = info.array.layout;
11518 }
471bceb6 11519
089f9d79
JS		 11520 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11521 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11522 if (info.array.level == 10) {
11523 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11524 change = -1;
11525 goto analyse_change_exit;
1e9b2c3f
PB		 11526 } else if (info.component_size % (geo->chunksize/512)) {
11527 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11528 geo->chunksize/1024, info.component_size/2);
11529 change = -1;
11530 goto analyse_change_exit;
2d2b0eb7 11531 }
b5347799 11532 change = CH_MIGRATION;
2d2b0eb7 11533 } else {
471bceb6 11534 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11535 }
471bceb6 11536
c21e737b 11537 chunk = geo->chunksize / 1024;
7abc9871
AK		 11538
11539 super = st->sb;
11540 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11541 map = get_imsm_map(dev, MAP_0);
11542 data_disks = imsm_num_data_members(map);
c41e00b2 11543 /* compute current size per disk member
7abc9871 11544 */
c41e00b2
AK		 11545 current_size = info.custom_array_size / data_disks;
11546
089f9d79 11547 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11548 /* align component size
11549 */
3e684231 11550 geo->size = imsm_component_size_alignment_check(
c41e00b2 11551 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11552 chunk * 1024, super->sector_size,
c41e00b2 11553 geo->size * 2);
65d0b4ce 11554 if (geo->size == 0) {
7a862a02 11555 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11556 current_size);
11557 goto analyse_change_exit;
11558 }
c41e00b2 11559 }
7abc9871 11560
089f9d79 11561 if (current_size != geo->size && geo->size > 0) {
7abc9871 11562 if (change != -1) {
7a862a02 11563 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11564 change = -1;
11565 goto analyse_change_exit;
11566 }
11567 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11568 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11569 super->current_vol, st->devnm);
7abc9871
AK		 11570 goto analyse_change_exit;
11571 }
65d38cca
LD		 11572 /* check the maximum available size
11573 */
11574 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11575 0, chunk, &free_size);
11576 if (rv == 0)
11577 /* Cannot find maximum available space
11578 */
11579 max_size = 0;
11580 else {
11581 max_size = free_size + current_size;
11582 /* align component size
11583 */
3e684231 11584 max_size = imsm_component_size_alignment_check(
65d38cca 11585 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11586 chunk * 1024, super->sector_size,
65d38cca
LD		 11587 max_size);
11588 }
d04f65f4 11589 if (geo->size == MAX_SIZE) {
b130333f
AK		 11590 /* requested size change to the maximum available size
11591 */
65d38cca 11592 if (max_size == 0) {
7a862a02 11593 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11594 change = -1;
11595 goto analyse_change_exit;
65d38cca
LD		 11596 } else
11597 geo->size = max_size;
c41e00b2 11598 }
b130333f 11599
681b7ae2 11600 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11601 /* accept size for rollback only
11602 */
11603 } else {
11604 /* round size due to metadata compatibility
11605 */
11606 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11607 << SECT_PER_MB_SHIFT;
11608 dprintf("Prepare update for size change to %llu\n",
11609 geo->size );
11610 if (current_size >= geo->size) {
7a862a02 11611 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11612 current_size, geo->size);
fbf3d202
AK		 11613 goto analyse_change_exit;
11614 }
65d38cca 11615 if (max_size && geo->size > max_size) {
7a862a02 11616 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11617 max_size, geo->size);
65d38cca
LD		 11618 goto analyse_change_exit;
11619 }
7abc9871
AK		 11620 }
11621 geo->size *= data_disks;
11622 geo->raid_disks = dev->vol.map->num_members;
11623 change = CH_ARRAY_SIZE;
11624 }
471bceb6
KW		 11625 if (!validate_geometry_imsm(st,
11626 geo->level,
67a2db32 11627 imsm_layout,
e91a3bad 11628 geo->raid_disks + devNumChange,
c21e737b 11629 &chunk,
af4348dd 11630 geo->size, INVALID_SECTORS,
5308f117 11631 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11632 change = -1;
11633
11634 if (check_devs) {
11635 struct intel_super *super = st->sb;
11636 struct imsm_super *mpb = super->anchor;
11637
11638 if (mpb->num_raid_devs > 1) {
7a862a02 11639 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11640 geo->dev_name);
471bceb6
KW		 11641 change = -1;
11642 }
11643 }
11644
11645analyse_change_exit:
089f9d79
JS		 11646 if (direction == ROLLBACK_METADATA_CHANGES &&
11647 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11648 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11649 change = -1;
11650 }
471bceb6 11651 return change;
694575e7
KW		 11652}
11653
bb025c2f
KW		 11654int imsm_takeover(struct supertype *st, struct geo_params *geo)
11655{
11656 struct intel_super *super = st->sb;
11657 struct imsm_update_takeover *u;
11658
503975b9 11659 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11660
11661 u->type = update_takeover;
11662 u->subarray = super->current_vol;
11663
11664 /* 10->0 transition */
11665 if (geo->level == 0)
11666 u->direction = R10_TO_R0;
11667
0529c688
KW		 11668 /* 0->10 transition */
11669 if (geo->level == 10)
11670 u->direction = R0_TO_R10;
11671
bb025c2f
KW		 11672 /* update metadata locally */
11673 imsm_update_metadata_locally(st, u,
11674 sizeof(struct imsm_update_takeover));
11675 /* and possibly remotely */
11676 if (st->update_tail)
11677 append_metadata_update(st, u,
11678 sizeof(struct imsm_update_takeover));
11679 else
11680 free(u);
11681
11682 return 0;
11683}
11684
d04f65f4
N		 11685static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11686 int level,
78b10e66 11687 int layout, int chunksize, int raid_disks,
41784c88 11688 int delta_disks, char *backup, char *dev,
016e00f5 11689 int direction, int verbose)
78b10e66 11690{
78b10e66
N		 11691 int ret_val = 1;
11692 struct geo_params geo;
11693
1ade5cc1 11694 dprintf("(enter)\n");
78b10e66 11695
71204a50 11696 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11697
11698 geo.dev_name = dev;
4dd2df09 11699 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11700 geo.size = size;
11701 geo.level = level;
11702 geo.layout = layout;
11703 geo.chunksize = chunksize;
11704 geo.raid_disks = raid_disks;
41784c88
AK		 11705 if (delta_disks != UnSet)
11706 geo.raid_disks += delta_disks;
78b10e66 11707
1ade5cc1
N		 11708 dprintf("for level : %i\n", geo.level);
11709 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11710
4dd2df09 11711 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11712 /* On container level we can only increase number of devices. */
11713 dprintf("imsm: info: Container operation\n");
78b10e66 11714 int old_raid_disks = 0;
6dc0be30 11715
78b10e66 11716 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11717 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11718 struct imsm_update_reshape *u = NULL;
11719 int len;
11720
11721 len = imsm_create_metadata_update_for_reshape(
11722 st, &geo, old_raid_disks, &u);
11723
ed08d51c
AK		 11724 if (len <= 0) {
11725 dprintf("imsm: Cannot prepare update\n");
11726 goto exit_imsm_reshape_super;
11727 }
11728
8dd70bce
AK		 11729 ret_val = 0;
11730 /* update metadata locally */
11731 imsm_update_metadata_locally(st, u, len);
11732 /* and possibly remotely */
11733 if (st->update_tail)
11734 append_metadata_update(st, u, len);
11735 else
ed08d51c 11736 free(u);
8dd70bce 11737
694575e7 11738 } else {
7a862a02 11739 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11740 }
11741 } else {
11742 /* On volume level we support following operations
471bceb6
KW		 11743 * - takeover: raid10 -> raid0; raid0 -> raid10
11744 * - chunk size migration
11745 * - migration: raid5 -> raid0; raid0 -> raid5
11746 */
11747 struct intel_super *super = st->sb;
11748 struct intel_dev *dev = super->devlist;
4dd2df09 11749 int change;
694575e7 11750 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11751 /* find requested device */
11752 while (dev) {
1011e834 11753 char *devnm =
4dd2df09
N		 11754 imsm_find_array_devnm_by_subdev(
11755 dev->index, st->container_devnm);
11756 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11757 break;
11758 dev = dev->next;
11759 }
11760 if (dev == NULL) {
4dd2df09
N		 11761 pr_err("Cannot find %s (%s) subarray\n",
11762 geo.dev_name, geo.devnm);
471bceb6
KW		 11763 goto exit_imsm_reshape_super;
11764 }
11765 super->current_vol = dev->index;
fbf3d202 11766 change = imsm_analyze_change(st, &geo, direction);
694575e7 11767 switch (change) {
471bceb6 11768 case CH_TAKEOVER:
bb025c2f 11769 ret_val = imsm_takeover(st, &geo);
694575e7 11770 break;
48c5303a
PC		 11771 case CH_MIGRATION: {
11772 struct imsm_update_reshape_migration *u = NULL;
11773 int len =
11774 imsm_create_metadata_update_for_migration(
11775 st, &geo, &u);
11776 if (len < 1) {
7a862a02 11777 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11778 break;
11779 }
471bceb6 11780 ret_val = 0;
48c5303a
PC		 11781 /* update metadata locally */
11782 imsm_update_metadata_locally(st, u, len);
11783 /* and possibly remotely */
11784 if (st->update_tail)
11785 append_metadata_update(st, u, len);
11786 else
11787 free(u);
11788 }
11789 break;
7abc9871 11790 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11791 struct imsm_update_size_change *u = NULL;
11792 int len =
11793 imsm_create_metadata_update_for_size_change(
11794 st, &geo, &u);
11795 if (len < 1) {
7a862a02 11796 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11797 break;
11798 }
11799 ret_val = 0;
11800 /* update metadata locally */
11801 imsm_update_metadata_locally(st, u, len);
11802 /* and possibly remotely */
11803 if (st->update_tail)
11804 append_metadata_update(st, u, len);
11805 else
11806 free(u);
7abc9871
AK		 11807 }
11808 break;
471bceb6
KW		 11809 default:
11810 ret_val = 1;
694575e7 11811 }
694575e7 11812 }
78b10e66 11813
ed08d51c 11814exit_imsm_reshape_super:
78b10e66
N		 11815 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11816 return ret_val;
11817}
2cda7640 11818
0febb20c
AO		 11819#define COMPLETED_OK 0
11820#define COMPLETED_NONE 1
11821#define COMPLETED_DELAYED 2
11822
11823static int read_completed(int fd, unsigned long long *val)
11824{
11825 int ret;
11826 char buf[50];
11827
11828 ret = sysfs_fd_get_str(fd, buf, 50);
11829 if (ret < 0)
11830 return ret;
11831
11832 ret = COMPLETED_OK;
11833 if (strncmp(buf, "none", 4) == 0) {
11834 ret = COMPLETED_NONE;
11835 } else if (strncmp(buf, "delayed", 7) == 0) {
11836 ret = COMPLETED_DELAYED;
11837 } else {
11838 char *ep;
11839 *val = strtoull(buf, &ep, 0);
11840 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11841 ret = -1;
11842 }
11843 return ret;
11844}
11845
eee67a47
AK		 11846/*******************************************************************************
11847 * Function: wait_for_reshape_imsm
11848 * Description: Function writes new sync_max value and waits until
11849 * reshape process reach new position
11850 * Parameters:
11851 * sra : general array info
eee67a47
AK		 11852 * ndata : number of disks in new array's layout
11853 * Returns:
11854 * 0 : success,
11855 * 1 : there is no reshape in progress,
11856 * -1 : fail
11857 ******************************************************************************/
ae9f01f8 11858int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11859{
85ca499c 11860 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11861 int retry = 3;
eee67a47 11862 unsigned long long completed;
ae9f01f8
AK		 11863 /* to_complete : new sync_max position */
11864 unsigned long long to_complete = sra->reshape_progress;
11865 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11866
ae9f01f8 11867 if (fd < 0) {
1ade5cc1 11868 dprintf("cannot open reshape_position\n");
eee67a47 11869 return 1;
ae9f01f8 11870 }
eee67a47 11871
df2647fa
PB		 11872 do {
11873 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11874 if (!retry) {
11875 dprintf("cannot read reshape_position (no reshape in progres)\n");
11876 close(fd);
11877 return 1;
11878 }
11879 usleep(30000);
11880 } else
11881 break;
11882 } while (retry--);
eee67a47 11883
85ca499c 11884 if (completed > position_to_set) {
1ade5cc1 11885 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11886 to_complete, position_to_set);
ae9f01f8
AK		 11887 close(fd);
11888 return -1;
11889 }
11890 dprintf("Position set: %llu\n", position_to_set);
11891 if (sysfs_set_num(sra, NULL, "sync_max",
11892 position_to_set) != 0) {
1ade5cc1 11893 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11894 position_to_set);
11895 close(fd);
11896 return -1;
eee67a47
AK		 11897 }
11898
eee67a47 11899 do {
0febb20c 11900 int rc;
eee67a47 11901 char action[20];
5ff3a780 11902 int timeout = 3000;
0febb20c 11903
5ff3a780 11904 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11905 if (sysfs_get_str(sra, NULL, "sync_action",
11906 action, 20) > 0 &&
d7d3809a 11907 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11908 if (strncmp(action, "idle", 4) == 0)
11909 break;
d7d3809a
AP		 11910 close(fd);
11911 return -1;
11912 }
0febb20c
AO		 11913
11914 rc = read_completed(fd, &completed);
11915 if (rc < 0) {
1ade5cc1 11916 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11917 close(fd);
11918 return 1;
0febb20c
AO		 11919 } else if (rc == COMPLETED_NONE)
11920 break;
85ca499c 11921 } while (completed < position_to_set);
b2be2b62 11922
eee67a47
AK		 11923 close(fd);
11924 return 0;
eee67a47
AK		 11925}
11926
b915c95f
AK		 11927/*******************************************************************************
11928 * Function: check_degradation_change
11929 * Description: Check that array hasn't become failed.
11930 * Parameters:
11931 * info : for sysfs access
11932 * sources : source disks descriptors
11933 * degraded: previous degradation level
11934 * Returns:
11935 * degradation level
11936 ******************************************************************************/
11937int check_degradation_change(struct mdinfo *info,
11938 int *sources,
11939 int degraded)
11940{
11941 unsigned long long new_degraded;
e1993023
LD		 11942 int rv;
11943
11944 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11945 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11946 /* check each device to ensure it is still working */
11947 struct mdinfo *sd;
11948 new_degraded = 0;
11949 for (sd = info->devs ; sd ; sd = sd->next) {
11950 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11951 continue;
11952 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11953 char sbuf[100];
11954
b915c95f 11955 if (sysfs_get_str(info,
cf52eff5 11956 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11957 strstr(sbuf, "faulty") ||
11958 strstr(sbuf, "in_sync") == NULL) {
11959 /* this device is dead */
11960 sd->disk.state = (1<<MD_DISK_FAULTY);
11961 if (sd->disk.raid_disk >= 0 &&
11962 sources[sd->disk.raid_disk] >= 0) {
11963 close(sources[
11964 sd->disk.raid_disk]);
11965 sources[sd->disk.raid_disk] =
11966 -1;
11967 }
11968 new_degraded++;
11969 }
11970 }
11971 }
11972 }
11973
11974 return new_degraded;
11975}
11976
10f22854
AK		 11977/*******************************************************************************
11978 * Function: imsm_manage_reshape
11979 * Description: Function finds array under reshape and it manages reshape
11980 * process. It creates stripes backups (if required) and sets
942e1cdb 11981 * checkpoints.
10f22854
AK		 11982 * Parameters:
11983 * afd : Backup handle (nattive) - not used
11984 * sra : general array info
11985 * reshape : reshape parameters - not used
11986 * st : supertype structure
11987 * blocks : size of critical section [blocks]
11988 * fds : table of source device descriptor
11989 * offsets : start of array (offest per devices)
11990 * dests : not used
11991 * destfd : table of destination device descriptor
11992 * destoffsets : table of destination offsets (per device)
11993 * Returns:
11994 * 1 : success, reshape is done
11995 * 0 : fail
11996 ******************************************************************************/
999b4972
N		 11997static int imsm_manage_reshape(
11998 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11999 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12000 int *fds, unsigned long long *offsets,
12001 int dests, int *destfd, unsigned long long *destoffsets)
12002{
10f22854
AK		 12003 int ret_val = 0;
12004 struct intel_super *super = st->sb;
594dc1b8 12005 struct intel_dev *dv;
de44e46f 12006 unsigned int sector_size = super->sector_size;
10f22854 12007 struct imsm_dev *dev = NULL;
9529d343 12008 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12009 int migr_vol_qan = 0;
12010 int ndata, odata; /* [bytes] */
12011 int chunk; /* [bytes] */
12012 struct migr_record *migr_rec;
12013 char *buf = NULL;
12014 unsigned int buf_size; /* [bytes] */
12015 unsigned long long max_position; /* array size [bytes] */
12016 unsigned long long next_step; /* [blocks]/[bytes] */
12017 unsigned long long old_data_stripe_length;
10f22854
AK		 12018 unsigned long long start_src; /* [bytes] */
12019 unsigned long long start; /* [bytes] */
12020 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12021 int degraded = 0;
ab724b98 12022 int source_layout = 0;
10f22854 12023
79a16a9b
JS		 12024 if (!sra)
12025 return ret_val;
12026
12027 if (!fds || !offsets)
10f22854
AK		 12028 goto abort;
12029
12030 /* Find volume during the reshape */
12031 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12032 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12033 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12034 dev = dv->dev;
12035 migr_vol_qan++;
12036 }
12037 }
12038 /* Only one volume can migrate at the same time */
12039 if (migr_vol_qan != 1) {
676e87a8 12040 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12041 "Number of migrating volumes greater than 1\n" :
12042 "There is no volume during migrationg\n");
12043 goto abort;
12044 }
12045
9529d343 12046 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12047 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12048 if (map_src == NULL)
12049 goto abort;
10f22854 12050
9529d343
MD		 12051 ndata = imsm_num_data_members(map_dest);
12052 odata = imsm_num_data_members(map_src);
10f22854 12053
7b1ab482 12054 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12055 old_data_stripe_length = odata * chunk;
12056
12057 migr_rec = super->migr_rec;
12058
10f22854
AK		 12059 /* initialize migration record for start condition */
12060 if (sra->reshape_progress == 0)
12061 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12062 else {
12063 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12064 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12065 goto abort;
12066 }
6a75c8ca
AK		 12067 /* Save checkpoint to update migration record for current
12068 * reshape position (in md). It can be farther than current
12069 * reshape position in metadata.
12070 */
12071 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12072 /* ignore error == 2, this can mean end of reshape here
12073 */
7a862a02 12074 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12075 goto abort;
12076 }
b2c59438 12077 }
10f22854
AK		 12078
12079 /* size for data */
12080 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12081 /* extend buffer size for parity disk */
12082 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12083 /* add space for stripe alignment */
10f22854 12084 buf_size += old_data_stripe_length;
de44e46f
PB		 12085 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12086 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12087 goto abort;
12088 }
12089
3ef4403c 12090 max_position = sra->component_size * ndata;
68eb8bc6 12091 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12092
9f421827
PB		 12093 while (current_migr_unit(migr_rec) <
12094 get_num_migr_units(migr_rec)) {
10f22854
AK		 12095 /* current reshape position [blocks] */
12096 unsigned long long current_position =
12097 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12098 * current_migr_unit(migr_rec);
10f22854
AK		 12099 unsigned long long border;
12100
b915c95f
AK		 12101 /* Check that array hasn't become failed.
12102 */
12103 degraded = check_degradation_change(sra, fds, degraded);
12104 if (degraded > 1) {
7a862a02 12105 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12106 goto abort;
12107 }
12108
10f22854
AK		 12109 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12110
12111 if ((current_position + next_step) > max_position)
12112 next_step = max_position - current_position;
12113
92144abf 12114 start = current_position * 512;
10f22854 12115
942e1cdb 12116 /* align reading start to old geometry */
10f22854
AK		 12117 start_buf_shift = start % old_data_stripe_length;
12118 start_src = start - start_buf_shift;
12119
12120 border = (start_src / odata) - (start / ndata);
12121 border /= 512;
12122 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12123 /* save critical stripes to buf
12124 * start - start address of current unit
12125 * to backup [bytes]
12126 * start_src - start address of current unit
12127 * to backup alligned to source array
12128 * [bytes]
12129 */
594dc1b8 12130 unsigned long long next_step_filler;
10f22854
AK		 12131 unsigned long long copy_length = next_step * 512;
12132
12133 /* allign copy area length to stripe in old geometry */
12134 next_step_filler = ((copy_length + start_buf_shift)
12135 % old_data_stripe_length);
12136 if (next_step_filler)
12137 next_step_filler = (old_data_stripe_length
12138 - next_step_filler);
7a862a02 12139 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		 12140 start, start_src, copy_length,
12141 start_buf_shift, next_step_filler);
12142
12143 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12144 chunk, map_src->raid_level,
12145 source_layout, 0, NULL, start_src,
10f22854
AK		 12146 copy_length +
12147 next_step_filler + start_buf_shift,
12148 buf)) {
7a862a02 12149 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12150 goto abort;
12151 }
12152 /* Convert data to destination format and store it
12153 * in backup general migration area
12154 */
12155 if (save_backup_imsm(st, dev, sra,
aea93171 12156 buf + start_buf_shift, copy_length)) {
7a862a02 12157 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12158 goto abort;
12159 }
12160 if (save_checkpoint_imsm(st, sra,
12161 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12162 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12163 goto abort;
12164 }
8016a6d4
AK		 12165 } else {
12166 /* set next step to use whole border area */
12167 border /= next_step;
12168 if (border > 1)
12169 next_step *= border;
10f22854
AK		 12170 }
12171 /* When data backed up, checkpoint stored,
12172 * kick the kernel to reshape unit of data
12173 */
12174 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12175 /* limit next step to array max position */
12176 if (next_step > max_position)
12177 next_step = max_position;
10f22854
AK		 12178 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12179 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12180 sra->reshape_progress = next_step;
10f22854
AK		 12181
12182 /* wait until reshape finish */
c85338c6 12183 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12184 dprintf("wait_for_reshape_imsm returned error!\n");
12185 goto abort;
12186 }
84d11e6c
N		 12187 if (sigterm)
12188 goto abort;
10f22854 12189
0228d92c
AK		 12190 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12191 /* ignore error == 2, this can mean end of reshape here
12192 */
7a862a02 12193 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12194 goto abort;
12195 }
12196
12197 }
12198
71e5411e
PB		 12199 /* clear migr_rec on disks after successful migration */
12200 struct dl *d;
12201
85337573 12202 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12203 for (d = super->disks; d; d = d->next) {
12204 if (d->index < 0 || is_failed(&d->disk))
12205 continue;
12206 unsigned long long dsize;
12207
12208 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12209 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12210 SEEK_SET) >= 0) {
466070ad 12211 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12212 MIGR_REC_BUF_SECTORS*sector_size) !=
12213 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12214 perror("Write migr_rec failed");
12215 }
12216 }
12217
10f22854
AK		 12218 /* return '1' if done */
12219 ret_val = 1;
12220abort:
12221 free(buf);
942e1cdb
N		 12222 /* See Grow.c: abort_reshape() for further explanation */
12223 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12224 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12225 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12226
12227 return ret_val;
999b4972 12228}
0c21b485 12229
cdddbdbc 12230struct superswitch super_imsm = {
cdddbdbc
DW		 12231 .examine_super = examine_super_imsm,
12232 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12233 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12234 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12235 .detail_super = detail_super_imsm,
12236 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12237 .write_init_super = write_init_super_imsm,
0e600426
N		 12238 .validate_geometry = validate_geometry_imsm,
12239 .add_to_super = add_to_super_imsm,
1a64be56 12240 .remove_from_super = remove_from_super_imsm,
d665cc31 12241 .detail_platform = detail_platform_imsm,
e50cf220 12242 .export_detail_platform = export_detail_platform_imsm,
33414a01 12243 .kill_subarray = kill_subarray_imsm,
aa534678 12244 .update_subarray = update_subarray_imsm,
2b959fbf 12245 .load_container = load_container_imsm,
71204a50
N		 12246 .default_geometry = default_geometry_imsm,
12247 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12248 .reshape_super = imsm_reshape_super,
12249 .manage_reshape = imsm_manage_reshape,
9e2d750d 12250 .recover_backup = recover_backup_imsm,
27156a57 12251 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12252 .match_home = match_home_imsm,
12253 .uuid_from_super= uuid_from_super_imsm,
12254 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12255 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12256 .update_super = update_super_imsm,
12257
12258 .avail_size = avail_size_imsm,
fbfdcb06 12259 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12260
12261 .compare_super = compare_super_imsm,
12262
12263 .load_super = load_super_imsm,
bf5a934a 12264 .init_super = init_super_imsm,
e683ca88 12265 .store_super = store_super_imsm,
cdddbdbc
DW		 12266 .free_super = free_super_imsm,
12267 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12268 .container_content = container_content_imsm,
0c21b485 12269 .validate_container = validate_container_imsm,
cdddbdbc 12270
2432ce9b
AP		 12271 .write_init_ppl = write_init_ppl_imsm,
12272 .validate_ppl = validate_ppl_imsm,
12273
cdddbdbc 12274 .external = 1,
4cce4069 12275 .name = "imsm",
845dea95
NB		 12276
12277/* for mdmon */
12278 .open_new = imsm_open_new,
ed9d66aa 12279 .set_array_state= imsm_set_array_state,
845dea95
NB		 12280 .set_disk = imsm_set_disk,
12281 .sync_metadata = imsm_sync_metadata,
88758e9d 12282 .activate_spare = imsm_activate_spare,
e8319a19 12283 .process_update = imsm_process_update,
8273f55e 12284 .prepare_update = imsm_prepare_update,
6f50473f 12285 .record_bad_block = imsm_record_badblock,
c07a5a4f 12286 .clear_bad_block = imsm_clear_badblock,
928f1424 12287 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12288};
Unnamed repository; edit this file 'description' to name the repository.