]> 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
Make target to install binaries only
[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);
a8f3cfd5
MT		 2367 if (!imsm_is_nvme_supported(fd, 0)) {
2368 if (fd >= 0)
2369 close(fd);
6da53c0e
BK		 2370 free(rp);
2371 continue;
2372 }
60f0f54d 2373
6da53c0e
BK		 2374 fd2devname(fd, buf);
2375 if (hba->type == SYS_DEV_VMD)
2376 printf(" NVMe under VMD : %s", buf);
2377 else if (hba->type == SYS_DEV_NVME)
2378 printf(" NVMe Device : %s", buf);
2379 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2380 sizeof(buf)))
2381 printf(" (%s)\n", buf);
2382 else
2383 printf("()\n");
2384 close(fd);
2385 }
2386 free(rp);
60f0f54d 2387 }
60f0f54d
PB		 2388 }
2389
b9135011 2390 closedir(dir);
b5eece69 2391 return 0;
60f0f54d
PB		 2392}
2393
120dc887
LM		 2394static void print_found_intel_controllers(struct sys_dev *elem)
2395{
2396 for (; elem; elem = elem->next) {
e7b84f9d 2397 pr_err("found Intel(R) ");
120dc887
LM		 2398 if (elem->type == SYS_DEV_SATA)
2399 fprintf(stderr, "SATA ");
155cbb4c
LM		 2400 else if (elem->type == SYS_DEV_SAS)
2401 fprintf(stderr, "SAS ");
0858eccf
AP		 2402 else if (elem->type == SYS_DEV_NVME)
2403 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2404
2405 if (elem->type == SYS_DEV_VMD)
2406 fprintf(stderr, "VMD domain");
2407 else
2408 fprintf(stderr, "RAID controller");
2409
120dc887
LM		 2410 if (elem->pci_id)
2411 fprintf(stderr, " at %s", elem->pci_id);
2412 fprintf(stderr, ".\n");
2413 }
2414 fflush(stderr);
2415}
2416
120dc887
LM		 2417static int ahci_get_port_count(const char *hba_path, int *port_count)
2418{
2419 struct dirent *ent;
2420 DIR *dir;
2421 int host_base = -1;
2422
2423 *port_count = 0;
2424 if ((dir = opendir(hba_path)) == NULL)
2425 return -1;
2426
2427 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2428 int host;
2429
0858eccf
AP		 2430 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2431 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2432 continue;
2433 if (*port_count == 0)
2434 host_base = host;
2435 else if (host < host_base)
2436 host_base = host;
2437
2438 if (host + 1 > *port_count + host_base)
2439 *port_count = host + 1 - host_base;
2440 }
2441 closedir(dir);
2442 return host_base;
2443}
2444
a891a3c2
LM		 2445static void print_imsm_capability(const struct imsm_orom *orom)
2446{
0858eccf
AP		 2447 printf(" Platform : Intel(R) ");
2448 if (orom->capabilities == 0 && orom->driver_features == 0)
2449 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2450 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2451 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2452 else
2453 printf("Rapid Storage Technology%s\n",
2454 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2455 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2456 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2457 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2458 printf(" RAID Levels :%s%s%s%s%s\n",
2459 imsm_orom_has_raid0(orom) ? " raid0" : "",
2460 imsm_orom_has_raid1(orom) ? " raid1" : "",
2461 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2462 imsm_orom_has_raid10(orom) ? " raid10" : "",
2463 imsm_orom_has_raid5(orom) ? " raid5" : "");
2464 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2465 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2466 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2467 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2468 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2469 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2470 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2471 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2472 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2473 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2474 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2475 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2476 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2477 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2478 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2479 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2480 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2481 printf(" 2TB volumes :%s supported\n",
2482 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2483 printf(" 2TB disks :%s supported\n",
2484 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2485 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2486 printf(" Max Volumes : %d per array, %d per %s\n",
2487 orom->vpa, orom->vphba,
2488 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2489 return;
2490}
2491
e50cf220
MN		 2492static void print_imsm_capability_export(const struct imsm_orom *orom)
2493{
2494 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2495 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2496 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2497 orom->hotfix_ver, orom->build);
e50cf220
MN		 2498 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2499 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2500 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2501 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2502 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2503 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2504 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2505 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2506 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2507 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2508 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2509 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2510 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2511 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2512 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2513 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2514 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2515 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2516 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2517 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2518 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2519 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2520 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2521 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2522 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2523 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2524 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2525 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2526}
2527
9eafa1de 2528static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2529{
2530 /* There are two components to imsm platform support, the ahci SATA
2531 * controller and the option-rom. To find the SATA controller we
2532 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2533 * controller with the Intel vendor id is present. This approach
2534 * allows mdadm to leverage the kernel's ahci detection logic, with the
2535 * caveat that if ahci.ko is not loaded mdadm will not be able to
2536 * detect platform raid capabilities. The option-rom resides in a
2537 * platform "Adapter ROM". We scan for its signature to retrieve the
2538 * platform capabilities. If raid support is disabled in the BIOS the
2539 * option-rom capability structure will not be available.
2540 */
d665cc31 2541 struct sys_dev *list, *hba;
d665cc31
DW		 2542 int host_base = 0;
2543 int port_count = 0;
9eafa1de 2544 int result=1;
d665cc31 2545
5615172f 2546 if (enumerate_only) {
a891a3c2 2547 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2548 return 0;
a891a3c2
LM		 2549 list = find_intel_devices();
2550 if (!list)
2551 return 2;
2552 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2553 if (find_imsm_capability(hba)) {
2554 result = 0;
a891a3c2
LM		 2555 break;
2556 }
9eafa1de 2557 else
6b781d33 2558 result = 2;
a891a3c2 2559 }
a891a3c2 2560 return result;
5615172f
DW		 2561 }
2562
155cbb4c
LM		 2563 list = find_intel_devices();
2564 if (!list) {
ba728be7 2565 if (verbose > 0)
7a862a02 2566 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2567 return 2;
ba728be7 2568 } else if (verbose > 0)
155cbb4c 2569 print_found_intel_controllers(list);
d665cc31 2570
a891a3c2 2571 for (hba = list; hba; hba = hba->next) {
0858eccf 2572 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2573 continue;
0858eccf 2574 if (!find_imsm_capability(hba)) {
60f0f54d 2575 char buf[PATH_MAX];
e7b84f9d 2576 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2577 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2578 get_sys_dev_type(hba->type));
0858eccf
AP		 2579 continue;
2580 }
2581 result = 0;
2582 }
2583
2584 if (controller_path && result == 1) {
2585 pr_err("no active Intel(R) RAID controller found under %s\n",
2586 controller_path);
2587 return result;
2588 }
2589
5e1d6128 2590 const struct orom_entry *entry;
0858eccf 2591
5e1d6128 2592 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2593 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2594 print_imsm_capability(&entry->orom);
32716c51
PB		 2595 printf(" 3rd party NVMe :%s supported\n",
2596 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2597 for (hba = list; hba; hba = hba->next) {
2598 if (hba->type == SYS_DEV_VMD) {
2599 char buf[PATH_MAX];
60f0f54d
PB		 2600 printf(" I/O Controller : %s (%s)\n",
2601 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2602 if (print_nvme_info(hba)) {
b5eece69
PB		 2603 if (verbose > 0)
2604 pr_err("failed to get devices attached to VMD domain.\n");
2605 result |= 2;
2606 }
60f0f54d
PB		 2607 }
2608 }
07cb1e57 2609 printf("\n");
60f0f54d
PB		 2610 continue;
2611 }
0858eccf 2612
60f0f54d
PB		 2613 print_imsm_capability(&entry->orom);
2614 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2615 for (hba = list; hba; hba = hba->next) {
2616 if (hba->type == SYS_DEV_NVME)
6da53c0e 2617 print_nvme_info(hba);
0858eccf 2618 }
60f0f54d 2619 printf("\n");
0858eccf
AP		 2620 continue;
2621 }
2622
2623 struct devid_list *devid;
5e1d6128 2624 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2625 hba = device_by_id(devid->devid);
2626 if (!hba)
2627 continue;
2628
9eafa1de
MN		 2629 printf(" I/O Controller : %s (%s)\n",
2630 hba->path, get_sys_dev_type(hba->type));
2631 if (hba->type == SYS_DEV_SATA) {
2632 host_base = ahci_get_port_count(hba->path, &port_count);
2633 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2634 if (verbose > 0)
7a862a02 2635 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2636 result |= 2;
2637 }
120dc887
LM		 2638 }
2639 }
0858eccf 2640 printf("\n");
d665cc31 2641 }
155cbb4c 2642
120dc887 2643 return result;
d665cc31 2644}
e50cf220 2645
9eafa1de 2646static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2647{
e50cf220
MN		 2648 struct sys_dev *list, *hba;
2649 int result=1;
2650
2651 list = find_intel_devices();
2652 if (!list) {
2653 if (verbose > 0)
2654 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2655 result = 2;
e50cf220
MN		 2656 return result;
2657 }
2658
2659 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2660 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2661 continue;
60f0f54d
PB		 2662 if (!find_imsm_capability(hba) && verbose > 0) {
2663 char buf[PATH_MAX];
2664 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2665 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2666 }
0858eccf 2667 else
e50cf220 2668 result = 0;
e50cf220
MN		 2669 }
2670
5e1d6128 2671 const struct orom_entry *entry;
0858eccf 2672
60f0f54d
PB		 2673 for (entry = orom_entries; entry; entry = entry->next) {
2674 if (entry->type == SYS_DEV_VMD) {
2675 for (hba = list; hba; hba = hba->next)
2676 print_imsm_capability_export(&entry->orom);
2677 continue;
2678 }
5e1d6128 2679 print_imsm_capability_export(&entry->orom);
60f0f54d 2680 }
0858eccf 2681
e50cf220
MN		 2682 return result;
2683}
2684
cdddbdbc
DW		 2685static int match_home_imsm(struct supertype *st, char *homehost)
2686{
5115ca67
DW		 2687 /* the imsm metadata format does not specify any host
2688 * identification information. We return -1 since we can never
2689 * confirm nor deny whether a given array is "meant" for this
148acb7b 2690 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2691 * exclude member disks that do not belong, and we rely on
2692 * mdadm.conf to specify the arrays that should be assembled.
2693 * Auto-assembly may still pick up "foreign" arrays.
2694 */
cdddbdbc 2695
9362c1c8 2696 return -1;
cdddbdbc
DW		 2697}
2698
2699static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2700{
51006d85
N		 2701 /* The uuid returned here is used for:
2702 * uuid to put into bitmap file (Create, Grow)
2703 * uuid for backup header when saving critical section (Grow)
2704 * comparing uuids when re-adding a device into an array
2705 * In these cases the uuid required is that of the data-array,
2706 * not the device-set.
2707 * uuid to recognise same set when adding a missing device back
2708 * to an array. This is a uuid for the device-set.
1011e834 2709 *
51006d85
N		 2710 * For each of these we can make do with a truncated
2711 * or hashed uuid rather than the original, as long as
2712 * everyone agrees.
2713 * In each case the uuid required is that of the data-array,
2714 * not the device-set.
43dad3d6 2715 */
51006d85
N		 2716 /* imsm does not track uuid's so we synthesis one using sha1 on
2717 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2718 * - the orig_family_num of the container
51006d85
N		 2719 * - the index number of the volume
2720 * - the 'serial' number of the volume.
2721 * Hopefully these are all constant.
2722 */
2723 struct intel_super *super = st->sb;
43dad3d6 2724
51006d85
N		 2725 char buf[20];
2726 struct sha1_ctx ctx;
2727 struct imsm_dev *dev = NULL;
148acb7b 2728 __u32 family_num;
51006d85 2729
148acb7b
DW		 2730 /* some mdadm versions failed to set ->orig_family_num, in which
2731 * case fall back to ->family_num. orig_family_num will be
2732 * fixed up with the first metadata update.
2733 */
2734 family_num = super->anchor->orig_family_num;
2735 if (family_num == 0)
2736 family_num = super->anchor->family_num;
51006d85 2737 sha1_init_ctx(&ctx);
92bd8f8d 2738 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2739 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2740 if (super->current_vol >= 0)
2741 dev = get_imsm_dev(super, super->current_vol);
2742 if (dev) {
2743 __u32 vol = super->current_vol;
2744 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2745 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2746 }
2747 sha1_finish_ctx(&ctx, buf);
2748 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2749}
2750
0d481d37 2751#if 0
4f5bc454
DW		 2752static void
2753get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2754{
cdddbdbc
DW		 2755 __u8 *v = get_imsm_version(mpb);
2756 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2757 char major[] = { 0, 0, 0 };
2758 char minor[] = { 0 ,0, 0 };
2759 char patch[] = { 0, 0, 0 };
2760 char *ver_parse[] = { major, minor, patch };
2761 int i, j;
2762
2763 i = j = 0;
2764 while (*v != '\0' && v < end) {
2765 if (*v != '.' && j < 2)
2766 ver_parse[i][j++] = *v;
2767 else {
2768 i++;
2769 j = 0;
2770 }
2771 v++;
2772 }
2773
4f5bc454
DW		 2774 *m = strtol(minor, NULL, 0);
2775 *p = strtol(patch, NULL, 0);
2776}
0d481d37 2777#endif
4f5bc454 2778
1e5c6983
DW		 2779static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2780{
2781 /* migr_strip_size when repairing or initializing parity */
238c0a71 2782 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2783 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2784
2785 switch (get_imsm_raid_level(map)) {
2786 case 5:
2787 case 10:
2788 return chunk;
2789 default:
2790 return 128*1024 >> 9;
2791 }
2792}
2793
2794static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2795{
2796 /* migr_strip_size when rebuilding a degraded disk, no idea why
2797 * this is different than migr_strip_size_resync(), but it's good
2798 * to be compatible
2799 */
238c0a71 2800 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2801 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2802
2803 switch (get_imsm_raid_level(map)) {
2804 case 1:
2805 case 10:
2806 if (map->num_members % map->num_domains == 0)
2807 return 128*1024 >> 9;
2808 else
2809 return chunk;
2810 case 5:
2811 return max((__u32) 64*1024 >> 9, chunk);
2812 default:
2813 return 128*1024 >> 9;
2814 }
2815}
2816
2817static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2818{
238c0a71
AK		 2819 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2820 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2821 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2822 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2823
2824 return max((__u32) 1, hi_chunk / lo_chunk);
2825}
2826
2827static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2828{
238c0a71 2829 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2830 int level = get_imsm_raid_level(lo);
2831
2832 if (level == 1 || level == 10) {
238c0a71 2833 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2834
2835 return hi->num_domains;
2836 } else
2837 return num_stripes_per_unit_resync(dev);
2838}
2839
9529d343 2840static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2841{
2842 /* named 'imsm_' because raid0, raid1 and raid10
2843 * counter-intuitively have the same number of data disks
2844 */
1e5c6983
DW		 2845 switch (get_imsm_raid_level(map)) {
2846 case 0:
36fd8ccc
AK		 2847 return map->num_members;
2848 break;
1e5c6983
DW		 2849 case 1:
2850 case 10:
36fd8ccc 2851 return map->num_members/2;
1e5c6983
DW		 2852 case 5:
2853 return map->num_members - 1;
2854 default:
1ade5cc1 2855 dprintf("unsupported raid level\n");
1e5c6983
DW		 2856 return 0;
2857 }
2858}
2859
44490938
MD		 2860static unsigned long long calc_component_size(struct imsm_map *map,
2861 struct imsm_dev *dev)
2862{
2863 unsigned long long component_size;
2864 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2865 long long calc_dev_size = 0;
44490938
MD		 2866 unsigned int member_disks = imsm_num_data_members(map);
2867
2868 if (member_disks == 0)
2869 return 0;
2870
2871 component_size = per_dev_array_size(map);
2872 calc_dev_size = component_size * member_disks;
2873
2874 /* Component size is rounded to 1MB so difference between size from
2875 * metadata and size calculated from num_data_stripes equals up to
2876 * 2048 blocks per each device. If the difference is higher it means
2877 * that array size was expanded and num_data_stripes was not updated.
2878 */
a4f7290c 2879 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2880 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2881 component_size = dev_size / member_disks;
2882 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2883 component_size / map->blocks_per_strip,
2884 num_data_stripes(map));
2885 }
2886
2887 return component_size;
2888}
2889
1e5c6983
DW		 2890static __u32 parity_segment_depth(struct imsm_dev *dev)
2891{
238c0a71 2892 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2893 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2894
2895 switch(get_imsm_raid_level(map)) {
2896 case 1:
2897 case 10:
2898 return chunk * map->num_domains;
2899 case 5:
2900 return chunk * map->num_members;
2901 default:
2902 return chunk;
2903 }
2904}
2905
2906static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2907{
238c0a71 2908 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2909 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2910 __u32 strip = block / chunk;
2911
2912 switch (get_imsm_raid_level(map)) {
2913 case 1:
2914 case 10: {
2915 __u32 vol_strip = (strip * map->num_domains) + 1;
2916 __u32 vol_stripe = vol_strip / map->num_members;
2917
2918 return vol_stripe * chunk + block % chunk;
2919 } case 5: {
2920 __u32 stripe = strip / (map->num_members - 1);
2921
2922 return stripe * chunk + block % chunk;
2923 }
2924 default:
2925 return 0;
2926 }
2927}
2928
c47b0ff6
AK		 2929static __u64 blocks_per_migr_unit(struct intel_super *super,
2930 struct imsm_dev *dev)
1e5c6983
DW		 2931{
2932 /* calculate the conversion factor between per member 'blocks'
2933 * (md/{resync,rebuild}_start) and imsm migration units, return
2934 * 0 for the 'not migrating' and 'unsupported migration' cases
2935 */
2936 if (!dev->vol.migr_state)
2937 return 0;
2938
2939 switch (migr_type(dev)) {
c47b0ff6
AK		 2940 case MIGR_GEN_MIGR: {
2941 struct migr_record *migr_rec = super->migr_rec;
2942 return __le32_to_cpu(migr_rec->blocks_per_unit);
2943 }
1e5c6983
DW		 2944 case MIGR_VERIFY:
2945 case MIGR_REPAIR:
2946 case MIGR_INIT: {
238c0a71 2947 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2948 __u32 stripes_per_unit;
2949 __u32 blocks_per_unit;
2950 __u32 parity_depth;
2951 __u32 migr_chunk;
2952 __u32 block_map;
2953 __u32 block_rel;
2954 __u32 segment;
2955 __u32 stripe;
2956 __u8 disks;
2957
2958 /* yes, this is really the translation of migr_units to
2959 * per-member blocks in the 'resync' case
2960 */
2961 stripes_per_unit = num_stripes_per_unit_resync(dev);
2962 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2963 disks = imsm_num_data_members(map);
1e5c6983 2964 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2965 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2966 segment = blocks_per_unit / stripe;
2967 block_rel = blocks_per_unit - segment * stripe;
2968 parity_depth = parity_segment_depth(dev);
2969 block_map = map_migr_block(dev, block_rel);
2970 return block_map + parity_depth * segment;
2971 }
2972 case MIGR_REBUILD: {
2973 __u32 stripes_per_unit;
2974 __u32 migr_chunk;
2975
2976 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2977 migr_chunk = migr_strip_blocks_rebuild(dev);
2978 return migr_chunk * stripes_per_unit;
2979 }
1e5c6983
DW		 2980 case MIGR_STATE_CHANGE:
2981 default:
2982 return 0;
2983 }
2984}
2985
c2c087e6
DW		 2986static int imsm_level_to_layout(int level)
2987{
2988 switch (level) {
2989 case 0:
2990 case 1:
2991 return 0;
2992 case 5:
2993 case 6:
a380c027 2994 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2995 case 10:
c92a2527 2996 return 0x102;
c2c087e6 2997 }
a18a888e 2998 return UnSet;
c2c087e6
DW		 2999}
3000
8e59f3d8
AK		 3001/*******************************************************************************
3002 * Function: read_imsm_migr_rec
3003 * Description: Function reads imsm migration record from last sector of disk
3004 * Parameters:
3005 * fd : disk descriptor
3006 * super : metadata info
3007 * Returns:
3008 * 0 : success,
3009 * -1 : fail
3010 ******************************************************************************/
3011static int read_imsm_migr_rec(int fd, struct intel_super *super)
3012{
3013 int ret_val = -1;
de44e46f 3014 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3015 unsigned long long dsize;
3016
3017 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3018 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3019 SEEK_SET) < 0) {
e7b84f9d
N		 3020 pr_err("Cannot seek to anchor block: %s\n",
3021 strerror(errno));
8e59f3d8
AK		 3022 goto out;
3023 }
466070ad 3024 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3025 MIGR_REC_BUF_SECTORS*sector_size) !=
3026 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3027 pr_err("Cannot read migr record block: %s\n",
3028 strerror(errno));
8e59f3d8
AK		 3029 goto out;
3030 }
3031 ret_val = 0;
de44e46f
PB		 3032 if (sector_size == 4096)
3033 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3034
3035out:
3036 return ret_val;
3037}
3038
3136abe5
AK		 3039static struct imsm_dev *imsm_get_device_during_migration(
3040 struct intel_super *super)
3041{
3042
3043 struct intel_dev *dv;
3044
3045 for (dv = super->devlist; dv; dv = dv->next) {
3046 if (is_gen_migration(dv->dev))
3047 return dv->dev;
3048 }
3049 return NULL;
3050}
3051
8e59f3d8
AK		 3052/*******************************************************************************
3053 * Function: load_imsm_migr_rec
3054 * Description: Function reads imsm migration record (it is stored at the last
3055 * sector of disk)
3056 * Parameters:
3057 * super : imsm internal array info
3058 * info : general array info
3059 * Returns:
3060 * 0 : success
3061 * -1 : fail
4c965cc9 3062 * -2 : no migration in progress
8e59f3d8
AK		 3063 ******************************************************************************/
3064static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3065{
3066 struct mdinfo *sd;
594dc1b8 3067 struct dl *dl;
8e59f3d8
AK		 3068 char nm[30];
3069 int retval = -1;
3070 int fd = -1;
3136abe5 3071 struct imsm_dev *dev;
594dc1b8 3072 struct imsm_map *map;
b4ab44d8 3073 int slot = -1;
3136abe5
AK		 3074
3075 /* find map under migration */
3076 dev = imsm_get_device_during_migration(super);
3077 /* nothing to load,no migration in progress?
3078 */
3079 if (dev == NULL)
4c965cc9 3080 return -2;
8e59f3d8
AK		 3081
3082 if (info) {
3083 for (sd = info->devs ; sd ; sd = sd->next) {
3084 /* read only from one of the first two slots */
12fe93e9
TM		 3085 if ((sd->disk.raid_disk < 0) ||
3086 (sd->disk.raid_disk > 1))
8e59f3d8 3087 continue;
3136abe5 3088
8e59f3d8
AK		 3089 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3090 fd = dev_open(nm, O_RDONLY);
3091 if (fd >= 0)
3092 break;
3093 }
3094 }
3095 if (fd < 0) {
12fe93e9 3096 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3097 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3098 /* skip spare and failed disks
3099 */
3100 if (dl->index < 0)
3101 continue;
8e59f3d8 3102 /* read only from one of the first two slots */
3136abe5
AK		 3103 if (map)
3104 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3105 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3106 continue;
3107 sprintf(nm, "%d:%d", dl->major, dl->minor);
3108 fd = dev_open(nm, O_RDONLY);
3109 if (fd >= 0)
3110 break;
3111 }
3112 }
3113 if (fd < 0)
3114 goto out;
3115 retval = read_imsm_migr_rec(fd, super);
3116
3117out:
3118 if (fd >= 0)
3119 close(fd);
3120 return retval;
3121}
3122
c17608ea
AK		 3123/*******************************************************************************
3124 * function: imsm_create_metadata_checkpoint_update
3125 * Description: It creates update for checkpoint change.
3126 * Parameters:
3127 * super : imsm internal array info
3128 * u : pointer to prepared update
3129 * Returns:
3130 * Uptate length.
3131 * If length is equal to 0, input pointer u contains no update
3132 ******************************************************************************/
3133static int imsm_create_metadata_checkpoint_update(
3134 struct intel_super *super,
3135 struct imsm_update_general_migration_checkpoint **u)
3136{
3137
3138 int update_memory_size = 0;
3139
1ade5cc1 3140 dprintf("(enter)\n");
c17608ea
AK		 3141
3142 if (u == NULL)
3143 return 0;
3144 *u = NULL;
3145
3146 /* size of all update data without anchor */
3147 update_memory_size =
3148 sizeof(struct imsm_update_general_migration_checkpoint);
3149
503975b9 3150 *u = xcalloc(1, update_memory_size);
c17608ea 3151 if (*u == NULL) {
1ade5cc1 3152 dprintf("error: cannot get memory\n");
c17608ea
AK		 3153 return 0;
3154 }
3155 (*u)->type = update_general_migration_checkpoint;
9f421827 3156 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3157 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3158
3159 return update_memory_size;
3160}
3161
c17608ea
AK		 3162static void imsm_update_metadata_locally(struct supertype *st,
3163 void *buf, int len);
3164
687629c2
AK		 3165/*******************************************************************************
3166 * Function: write_imsm_migr_rec
3167 * Description: Function writes imsm migration record
3168 * (at the last sector of disk)
3169 * Parameters:
3170 * super : imsm internal array info
3171 * Returns:
3172 * 0 : success
3173 * -1 : if fail
3174 ******************************************************************************/
3175static int write_imsm_migr_rec(struct supertype *st)
3176{
3177 struct intel_super *super = st->sb;
de44e46f 3178 unsigned int sector_size = super->sector_size;
687629c2
AK		 3179 unsigned long long dsize;
3180 char nm[30];
3181 int fd = -1;
3182 int retval = -1;
3183 struct dl *sd;
c17608ea
AK		 3184 int len;
3185 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3186 struct imsm_dev *dev;
594dc1b8 3187 struct imsm_map *map;
3136abe5
AK		 3188
3189 /* find map under migration */
3190 dev = imsm_get_device_during_migration(super);
3191 /* if no migration, write buffer anyway to clear migr_record
3192 * on disk based on first available device
3193 */
3194 if (dev == NULL)
3195 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3196 super->current_vol);
3197
44bfe6df 3198 map = get_imsm_map(dev, MAP_0);
687629c2 3199
de44e46f
PB		 3200 if (sector_size == 4096)
3201 convert_to_4k_imsm_migr_rec(super);
687629c2 3202 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3203 int slot = -1;
3136abe5
AK		 3204
3205 /* skip failed and spare devices */
3206 if (sd->index < 0)
3207 continue;
687629c2 3208 /* write to 2 first slots only */
3136abe5
AK		 3209 if (map)
3210 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3211 if (map == NULL || slot > 1 || slot < 0)
687629c2 3212 continue;
3136abe5 3213
687629c2
AK		 3214 sprintf(nm, "%d:%d", sd->major, sd->minor);
3215 fd = dev_open(nm, O_RDWR);
3216 if (fd < 0)
3217 continue;
3218 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3219 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3220 SEEK_SET) < 0) {
e7b84f9d
N		 3221 pr_err("Cannot seek to anchor block: %s\n",
3222 strerror(errno));
687629c2
AK		 3223 goto out;
3224 }
466070ad 3225 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3226 MIGR_REC_BUF_SECTORS*sector_size) !=
3227 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3228 pr_err("Cannot write migr record block: %s\n",
3229 strerror(errno));
687629c2
AK		 3230 goto out;
3231 }
3232 close(fd);
3233 fd = -1;
3234 }
de44e46f
PB		 3235 if (sector_size == 4096)
3236 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3237 /* update checkpoint information in metadata */
3238 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3239 if (len <= 0) {
3240 dprintf("imsm: Cannot prepare update\n");
3241 goto out;
3242 }
3243 /* update metadata locally */
3244 imsm_update_metadata_locally(st, u, len);
3245 /* and possibly remotely */
3246 if (st->update_tail) {
3247 append_metadata_update(st, u, len);
3248 /* during reshape we do all work inside metadata handler
3249 * manage_reshape(), so metadata update has to be triggered
3250 * insida it
3251 */
3252 flush_metadata_updates(st);
3253 st->update_tail = &st->updates;
3254 } else
3255 free(u);
687629c2
AK		 3256
3257 retval = 0;
3258 out:
3259 if (fd >= 0)
3260 close(fd);
3261 return retval;
3262}
3263
e2962bfc
AK		 3264/* spare/missing disks activations are not allowe when
3265 * array/container performs reshape operation, because
3266 * all arrays in container works on the same disks set
3267 */
3268int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3269{
3270 int rv = 0;
3271 struct intel_dev *i_dev;
3272 struct imsm_dev *dev;
3273
3274 /* check whole container
3275 */
3276 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3277 dev = i_dev->dev;
3ad25638 3278 if (is_gen_migration(dev)) {
e2962bfc
AK		 3279 /* No repair during any migration in container
3280 */
3281 rv = 1;
3282 break;
3283 }
3284 }
3285 return rv;
3286}
3e684231 3287static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3288 int chunk_size,
f36a9ecd 3289 unsigned int sector_size,
c41e00b2
AK		 3290 unsigned long long component_size)
3291{
3e684231 3292 unsigned int component_size_alignment;
c41e00b2 3293
3e684231 3294 /* check component size alignment
c41e00b2 3295 */
3e684231 3296 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3297
3e684231 3298 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3299 level, chunk_size, component_size,
3e684231 3300 component_size_alignment);
c41e00b2 3301
3e684231
MZ		 3302 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3303 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3304 component_size);
3e684231 3305 component_size -= component_size_alignment;
1ade5cc1 3306 dprintf_cont("to %llu (%i).\n",
3e684231 3307 component_size, component_size_alignment);
c41e00b2
AK		 3308 }
3309
3310 return component_size;
3311}
e2962bfc 3312
2432ce9b
AP		 3313static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3314{
3315 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3316 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3317
3318 return pba_of_lba0(map) +
3319 (num_data_stripes(map) * map->blocks_per_strip);
3320}
3321
a5d85af7 3322static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3323{
3324 struct intel_super *super = st->sb;
c47b0ff6 3325 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3326 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3327 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3328 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3329 struct imsm_map *map_to_analyse = map;
efb30e7f 3330 struct dl *dl;
a5d85af7 3331 int map_disks = info->array.raid_disks;
bf5a934a 3332
95eeceeb 3333 memset(info, 0, sizeof(*info));
b335e593
AK		 3334 if (prev_map)
3335 map_to_analyse = prev_map;
3336
ca0748fa 3337 dl = super->current_disk;
9894ec0d 3338
bf5a934a 3339 info->container_member = super->current_vol;
cd0430a1 3340 info->array.raid_disks = map->num_members;
b335e593 3341 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3342 info->array.layout = imsm_level_to_layout(info->array.level);
3343 info->array.md_minor = -1;
3344 info->array.ctime = 0;
3345 info->array.utime = 0;
b335e593
AK		 3346 info->array.chunk_size =
3347 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3348 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3349 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3350 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3351
3f510843 3352 if (is_gen_migration(dev)) {
3f83228a 3353 info->reshape_active = 1;
b335e593
AK		 3354 info->new_level = get_imsm_raid_level(map);
3355 info->new_layout = imsm_level_to_layout(info->new_level);
3356 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3357 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3358 if (info->delta_disks) {
3359 /* this needs to be applied to every array
3360 * in the container.
3361 */
81219e70 3362 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3363 }
3f83228a
N		 3364 /* We shape information that we give to md might have to be
3365 * modify to cope with md's requirement for reshaping arrays.
3366 * For example, when reshaping a RAID0, md requires it to be
3367 * presented as a degraded RAID4.
3368 * Also if a RAID0 is migrating to a RAID5 we need to specify
3369 * the array as already being RAID5, but the 'before' layout
3370 * is a RAID4-like layout.
3371 */
3372 switch (info->array.level) {
3373 case 0:
3374 switch(info->new_level) {
3375 case 0:
3376 /* conversion is happening as RAID4 */
3377 info->array.level = 4;
3378 info->array.raid_disks += 1;
3379 break;
3380 case 5:
3381 /* conversion is happening as RAID5 */
3382 info->array.level = 5;
3383 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3384 info->delta_disks -= 1;
3385 break;
3386 default:
3387 /* FIXME error message */
3388 info->array.level = UnSet;
3389 break;
3390 }
3391 break;
3392 }
b335e593
AK		 3393 } else {
3394 info->new_level = UnSet;
3395 info->new_layout = UnSet;
3396 info->new_chunk = info->array.chunk_size;
3f83228a 3397 info->delta_disks = 0;
b335e593 3398 }
ca0748fa 3399
efb30e7f
DW		 3400 if (dl) {
3401 info->disk.major = dl->major;
3402 info->disk.minor = dl->minor;
ca0748fa 3403 info->disk.number = dl->index;
656b6b5a
N		 3404 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3405 dl->index);
efb30e7f 3406 }
bf5a934a 3407
5551b113 3408 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3409 info->component_size = calc_component_size(map, dev);
3e684231 3410 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3411 info->array.level,
3412 info->array.chunk_size,
f36a9ecd 3413 super->sector_size,
c41e00b2 3414 info->component_size);
5e46202e 3415 info->bb.supported = 1;
139dae11 3416
301406c9 3417 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3418 info->recovery_start = MaxSector;
bf5a934a 3419
c2462068
PB		 3420 if (info->array.level == 5 &&
3421 (dev->rwh_policy == RWH_DISTRIBUTED ||
3422 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3423 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3424 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3425 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3426 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3427 else
3428 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3429 >> 9;
2432ce9b
AP		 3430 } else if (info->array.level <= 0) {
3431 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3432 } else {
3433 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3434 }
3435
d2e6d5d6 3436 info->reshape_progress = 0;
b6796ce1 3437 info->resync_start = MaxSector;
b9172665 3438 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3439 !(info->array.state & 1)) &&
b9172665 3440 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3441 info->resync_start = 0;
b6796ce1
AK		 3442 }
3443 if (dev->vol.migr_state) {
1e5c6983
DW		 3444 switch (migr_type(dev)) {
3445 case MIGR_REPAIR:
3446 case MIGR_INIT: {
c47b0ff6
AK		 3447 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3448 dev);
1e5c6983
DW		 3449 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3450
3451 info->resync_start = blocks_per_unit * units;
3452 break;
3453 }
d2e6d5d6 3454 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3455 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3456 dev);
9f421827 3457 __u64 units = current_migr_unit(migr_rec);
04fa9523 3458 int used_disks;
d2e6d5d6 3459
befb629b
AK		 3460 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3461 (units <
9f421827 3462 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3463 (super->migr_rec->rec_status ==
3464 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3465 units++;
3466
d2e6d5d6 3467 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3468
7a862a02 3469 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3470 (unsigned long long)units,
3471 (unsigned long long)blocks_per_unit,
3472 info->reshape_progress);
75156c46 3473
9529d343 3474 used_disks = imsm_num_data_members(prev_map);
75156c46 3475 if (used_disks > 0) {
895ffd99 3476 info->custom_array_size = per_dev_array_size(map) *
75156c46 3477 used_disks;
75156c46 3478 }
d2e6d5d6 3479 }
1e5c6983
DW		 3480 case MIGR_VERIFY:
3481 /* we could emulate the checkpointing of
3482 * 'sync_action=check' migrations, but for now
3483 * we just immediately complete them
3484 */
3485 case MIGR_REBUILD:
3486 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3487 case MIGR_STATE_CHANGE:
3488 /* FIXME handle other migrations */
3489 default:
3490 /* we are not dirty, so... */
3491 info->resync_start = MaxSector;
3492 }
b6796ce1 3493 }
301406c9
DW		 3494
3495 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3496 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3497
f35f2525
N		 3498 info->array.major_version = -1;
3499 info->array.minor_version = -2;
4dd2df09 3500 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3501 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3502 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3503
3504 if (dmap) {
3505 int i, j;
3506 for (i=0; i<map_disks; i++) {
3507 dmap[i] = 0;
3508 if (i < info->array.raid_disks) {
3509 struct imsm_disk *dsk;
238c0a71 3510 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3511 dsk = get_imsm_disk(super, j);
3512 if (dsk && (dsk->status & CONFIGURED_DISK))
3513 dmap[i] = 1;
3514 }
3515 }
3516 }
81ac8b4d 3517}
bf5a934a 3518
3b451610
AK		 3519static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3520 int failed, int look_in_map);
3521
3522static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3523 int look_in_map);
3524
3525static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3526{
3527 if (is_gen_migration(dev)) {
3528 int failed;
3529 __u8 map_state;
3530 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3531
3532 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3533 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3534 if (map2->map_state != map_state) {
3535 map2->map_state = map_state;
3536 super->updates_pending++;
3537 }
3538 }
3539}
97b4d0e9
DW		 3540
3541static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3542{
3543 struct dl *d;
3544
3545 for (d = super->missing; d; d = d->next)
3546 if (d->index == index)
3547 return &d->disk;
3548 return NULL;
3549}
3550
a5d85af7 3551static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3552{
3553 struct intel_super *super = st->sb;
4f5bc454 3554 struct imsm_disk *disk;
a5d85af7 3555 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3556 int max_enough = -1;
3557 int i;
3558 struct imsm_super *mpb;
4f5bc454 3559
bf5a934a 3560 if (super->current_vol >= 0) {
a5d85af7 3561 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3562 return;
3563 }
95eeceeb 3564 memset(info, 0, sizeof(*info));
d23fe947
DW		 3565
3566 /* Set raid_disks to zero so that Assemble will always pull in valid
3567 * spares
3568 */
3569 info->array.raid_disks = 0;
cdddbdbc
DW		 3570 info->array.level = LEVEL_CONTAINER;
3571 info->array.layout = 0;
3572 info->array.md_minor = -1;
1011e834 3573 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3574 info->array.utime = 0;
3575 info->array.chunk_size = 0;
3576
3577 info->disk.major = 0;
3578 info->disk.minor = 0;
cdddbdbc 3579 info->disk.raid_disk = -1;
c2c087e6 3580 info->reshape_active = 0;
f35f2525
N		 3581 info->array.major_version = -1;
3582 info->array.minor_version = -2;
c2c087e6 3583 strcpy(info->text_version, "imsm");
a67dd8cc 3584 info->safe_mode_delay = 0;
c2c087e6
DW		 3585 info->disk.number = -1;
3586 info->disk.state = 0;
c5afc314 3587 info->name[0] = 0;
921d9e16 3588 info->recovery_start = MaxSector;
3ad25638 3589 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3590 info->bb.supported = 1;
c2c087e6 3591
97b4d0e9 3592 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3593 mpb = super->anchor;
b7d81a38 3594 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3595
ab3cb6b3
N		 3596 for (i = 0; i < mpb->num_raid_devs; i++) {
3597 struct imsm_dev *dev = get_imsm_dev(super, i);
3598 int failed, enough, j, missing = 0;
3599 struct imsm_map *map;
3600 __u8 state;
97b4d0e9 3601
3b451610
AK		 3602 failed = imsm_count_failed(super, dev, MAP_0);
3603 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3604 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3605
3606 /* any newly missing disks?
3607 * (catches single-degraded vs double-degraded)
3608 */
3609 for (j = 0; j < map->num_members; j++) {
238c0a71 3610 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3611 __u32 idx = ord_to_idx(ord);
3612
20dc76d1
MT		 3613 if (super->disks && super->disks->index == (int)idx)
3614 info->disk.raid_disk = j;
3615
ab3cb6b3
N		 3616 if (!(ord & IMSM_ORD_REBUILD) &&
3617 get_imsm_missing(super, idx)) {
3618 missing = 1;
3619 break;
3620 }
97b4d0e9 3621 }
ab3cb6b3
N		 3622
3623 if (state == IMSM_T_STATE_FAILED)
3624 enough = -1;
3625 else if (state == IMSM_T_STATE_DEGRADED &&
3626 (state != map->map_state || missing))
3627 enough = 0;
3628 else /* we're normal, or already degraded */
3629 enough = 1;
d2bde6d3
AK		 3630 if (is_gen_migration(dev) && missing) {
3631 /* during general migration we need all disks
3632 * that process is running on.
3633 * No new missing disk is allowed.
3634 */
3635 max_enough = -1;
3636 enough = -1;
3637 /* no more checks necessary
3638 */
3639 break;
3640 }
ab3cb6b3
N		 3641 /* in the missing/failed disk case check to see
3642 * if at least one array is runnable
3643 */
3644 max_enough = max(max_enough, enough);
3645 }
1ade5cc1 3646 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3647 info->container_enough = max_enough;
97b4d0e9 3648
4a04ec6c 3649 if (super->disks) {
14e8215b
DW		 3650 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3651
b9f594fe 3652 disk = &super->disks->disk;
5551b113 3653 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3654 info->component_size = reserved;
25ed7e59 3655 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3656 /* we don't change info->disk.raid_disk here because
3657 * this state will be finalized in mdmon after we have
3658 * found the 'most fresh' version of the metadata
3659 */
25ed7e59 3660 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3661 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3662 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3663 }
a575e2a7
DW		 3664
3665 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3666 * ->compare_super may have updated the 'num_raid_devs' field for spares
3667 */
3668 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3669 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3670 else
3671 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3672
3673 /* I don't know how to compute 'map' on imsm, so use safe default */
3674 if (map) {
3675 int i;
3676 for (i = 0; i < map_disks; i++)
3677 map[i] = 1;
3678 }
3679
cdddbdbc
DW		 3680}
3681
5c4cd5da
AC		 3682/* allocates memory and fills disk in mdinfo structure
3683 * for each disk in array */
3684struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3685{
594dc1b8 3686 struct mdinfo *mddev;
5c4cd5da
AC		 3687 struct intel_super *super = st->sb;
3688 struct imsm_disk *disk;
3689 int count = 0;
3690 struct dl *dl;
3691 if (!super || !super->disks)
3692 return NULL;
3693 dl = super->disks;
503975b9 3694 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3695 while (dl) {
3696 struct mdinfo *tmp;
3697 disk = &dl->disk;
503975b9 3698 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3699 if (mddev->devs)
3700 tmp->next = mddev->devs;
3701 mddev->devs = tmp;
3702 tmp->disk.number = count++;
3703 tmp->disk.major = dl->major;
3704 tmp->disk.minor = dl->minor;
3705 tmp->disk.state = is_configured(disk) ?
3706 (1 << MD_DISK_ACTIVE) : 0;
3707 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3708 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3709 tmp->disk.raid_disk = -1;
3710 dl = dl->next;
3711 }
3712 return mddev;
3713}
3714
cdddbdbc
DW		 3715static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3716 char *update, char *devname, int verbose,
3717 int uuid_set, char *homehost)
3718{
f352c545
DW		 3719 /* For 'assemble' and 'force' we need to return non-zero if any
3720 * change was made. For others, the return value is ignored.
3721 * Update options are:
3722 * force-one : This device looks a bit old but needs to be included,
3723 * update age info appropriately.
3724 * assemble: clear any 'faulty' flag to allow this device to
3725 * be assembled.
3726 * force-array: Array is degraded but being forced, mark it clean
3727 * if that will be needed to assemble it.
3728 *
3729 * newdev: not used ????
3730 * grow: Array has gained a new device - this is currently for
3731 * linear only
3732 * resync: mark as dirty so a resync will happen.
3733 * name: update the name - preserving the homehost
6e46bf34 3734 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3735 *
3736 * Following are not relevant for this imsm:
3737 * sparc2.2 : update from old dodgey metadata
3738 * super-minor: change the preferred_minor number
3739 * summaries: update redundant counters.
f352c545
DW		 3740 * homehost: update the recorded homehost
3741 * _reshape_progress: record new reshape_progress position.
3742 */
6e46bf34
DW		 3743 int rv = 1;
3744 struct intel_super *super = st->sb;
3745 struct imsm_super *mpb;
f352c545 3746
6e46bf34
DW		 3747 /* we can only update container info */
3748 if (!super || super->current_vol >= 0 || !super->anchor)
3749 return 1;
3750
3751 mpb = super->anchor;
3752
81a5b4f5
N		 3753 if (strcmp(update, "uuid") == 0) {
3754 /* We take this to mean that the family_num should be updated.
3755 * However that is much smaller than the uuid so we cannot really
3756 * allow an explicit uuid to be given. And it is hard to reliably
3757 * know if one was.
3758 * So if !uuid_set we know the current uuid is random and just used
3759 * the first 'int' and copy it to the other 3 positions.
3760 * Otherwise we require the 4 'int's to be the same as would be the
3761 * case if we are using a random uuid. So an explicit uuid will be
3762 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3763 */
81a5b4f5
N		 3764 if (!uuid_set) {
3765 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3766 rv = 0;
81a5b4f5
N		 3767 } else {
3768 if (info->uuid[0] != info->uuid[1] ||
3769 info->uuid[1] != info->uuid[2] ||
3770 info->uuid[2] != info->uuid[3])
3771 rv = -1;
3772 else
3773 rv = 0;
6e46bf34 3774 }
81a5b4f5
N		 3775 if (rv == 0)
3776 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3777 } else if (strcmp(update, "assemble") == 0)
3778 rv = 0;
3779 else
1e2b2765 3780 rv = -1;
f352c545 3781
6e46bf34
DW		 3782 /* successful update? recompute checksum */
3783 if (rv == 0)
3784 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3785
3786 return rv;
cdddbdbc
DW		 3787}
3788
c2c087e6 3789static size_t disks_to_mpb_size(int disks)
cdddbdbc 3790{
c2c087e6 3791 size_t size;
cdddbdbc 3792
c2c087e6
DW		 3793 size = sizeof(struct imsm_super);
3794 size += (disks - 1) * sizeof(struct imsm_disk);
3795 size += 2 * sizeof(struct imsm_dev);
3796 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3797 size += (4 - 2) * sizeof(struct imsm_map);
3798 /* 4 possible disk_ord_tbl's */
3799 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3800 /* maximum bbm log */
3801 size += sizeof(struct bbm_log);
c2c087e6
DW		 3802
3803 return size;
3804}
3805
387fcd59
N		 3806static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3807 unsigned long long data_offset)
c2c087e6
DW		 3808{
3809 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3810 return 0;
3811
3812 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3813}
3814
ba2de7ba
DW		 3815static void free_devlist(struct intel_super *super)
3816{
3817 struct intel_dev *dv;
3818
3819 while (super->devlist) {
3820 dv = super->devlist->next;
3821 free(super->devlist->dev);
3822 free(super->devlist);
3823 super->devlist = dv;
3824 }
3825}
3826
3827static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3828{
3829 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3830}
3831
cdddbdbc
DW		 3832static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3833{
3834 /*
3835 * return:
3836 * 0 same, or first was empty, and second was copied
3837 * 1 second had wrong number
3838 * 2 wrong uuid
3839 * 3 wrong other info
3840 */
3841 struct intel_super *first = st->sb;
3842 struct intel_super *sec = tst->sb;
3843
5d500228
N		 3844 if (!first) {
3845 st->sb = tst->sb;
3846 tst->sb = NULL;
3847 return 0;
3848 }
8603ea6f
LM		 3849 /* in platform dependent environment test if the disks
3850 * use the same Intel hba
cb8f6859 3851 * If not on Intel hba at all, allow anything.
8603ea6f 3852 */
6b781d33
AP		 3853 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3854 if (first->hba->type != sec->hba->type) {
8603ea6f 3855 fprintf(stderr,
6b781d33
AP		 3856 "HBAs of devices do not match %s != %s\n",
3857 get_sys_dev_type(first->hba->type),
3858 get_sys_dev_type(sec->hba->type));
3859 return 3;
3860 }
3861 if (first->orom != sec->orom) {
3862 fprintf(stderr,
3863 "HBAs of devices do not match %s != %s\n",
3864 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3865 return 3;
3866 }
3867 }
cdddbdbc 3868
d23fe947
DW		 3869 /* if an anchor does not have num_raid_devs set then it is a free
3870 * floating spare
3871 */
3872 if (first->anchor->num_raid_devs > 0 &&
3873 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3874 /* Determine if these disks might ever have been
3875 * related. Further disambiguation can only take place
3876 * in load_super_imsm_all
3877 */
3878 __u32 first_family = first->anchor->orig_family_num;
3879 __u32 sec_family = sec->anchor->orig_family_num;
3880
f796af5d
DW		 3881 if (memcmp(first->anchor->sig, sec->anchor->sig,
3882 MAX_SIGNATURE_LENGTH) != 0)
3883 return 3;
3884
a2b97981
DW		 3885 if (first_family == 0)
3886 first_family = first->anchor->family_num;
3887 if (sec_family == 0)
3888 sec_family = sec->anchor->family_num;
3889
3890 if (first_family != sec_family)
d23fe947 3891 return 3;
f796af5d 3892
d23fe947 3893 }
cdddbdbc 3894
3e372e5a
DW		 3895 /* if 'first' is a spare promote it to a populated mpb with sec's
3896 * family number
3897 */
3898 if (first->anchor->num_raid_devs == 0 &&
3899 sec->anchor->num_raid_devs > 0) {
78d30f94 3900 int i;
ba2de7ba
DW		 3901 struct intel_dev *dv;
3902 struct imsm_dev *dev;
78d30f94
DW		 3903
3904 /* we need to copy raid device info from sec if an allocation
3905 * fails here we don't associate the spare
3906 */
3907 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3908 dv = xmalloc(sizeof(*dv));
3909 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3910 dv->dev = dev;
3911 dv->index = i;
3912 dv->next = first->devlist;
3913 first->devlist = dv;
78d30f94 3914 }
709743c5 3915 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3916 /* allocation failure */
3917 free_devlist(first);
e12b3daa 3918 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3919 return 3;
78d30f94 3920 }
3e372e5a 3921 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3922 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3923 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3924 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3925 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3926 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3927 }
3928
cdddbdbc
DW		 3929 return 0;
3930}
3931
0030e8d6
DW		 3932static void fd2devname(int fd, char *name)
3933{
3934 struct stat st;
3935 char path[256];
33a6535d 3936 char dname[PATH_MAX];
0030e8d6
DW		 3937 char *nm;
3938 int rv;
3939
3940 name[0] = '\0';
3941 if (fstat(fd, &st) != 0)
3942 return;
3943 sprintf(path, "/sys/dev/block/%d:%d",
3944 major(st.st_rdev), minor(st.st_rdev));
3945
9cf014ec 3946 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3947 if (rv <= 0)
3948 return;
9587c373 3949
0030e8d6
DW		 3950 dname[rv] = '\0';
3951 nm = strrchr(dname, '/');
7897de29
JS		 3952 if (nm) {
3953 nm++;
3954 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3955 }
0030e8d6
DW		 3956}
3957
21e9380b
AP		 3958static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3959{
3960 char path[60];
3961 char *name = fd2kname(fd);
3962
3963 if (!name)
3964 return 1;
3965
3966 if (strncmp(name, "nvme", 4) != 0)
3967 return 1;
3968
3969 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3970
3971 return load_sys(path, buf, buf_len);
3972}
3973
cdddbdbc
DW		 3974extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3975
3976static int imsm_read_serial(int fd, char *devname,
6da53c0e 3977 __u8 *serial, size_t serial_buf_len)
cdddbdbc 3978{
21e9380b 3979 char buf[50];
cdddbdbc 3980 int rv;
6da53c0e 3981 size_t len;
316e2bf4
DW		 3982 char *dest;
3983 char *src;
21e9380b
AP		 3984 unsigned int i;
3985
3986 memset(buf, 0, sizeof(buf));
cdddbdbc 3987
21e9380b 3988 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3989
21e9380b
AP		 3990 if (rv)
3991 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3992
40ebbb9c 3993 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3994 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3995 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3996 return 0;
3997 }
3998
cdddbdbc
DW		 3999 if (rv != 0) {
4000 if (devname)
e7b84f9d
N		 4001 pr_err("Failed to retrieve serial for %s\n",
4002 devname);
cdddbdbc
DW		 4003 return rv;
4004 }
4005
316e2bf4
DW		 4006 /* trim all whitespace and non-printable characters and convert
4007 * ':' to ';'
4008 */
21e9380b
AP		 4009 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4010 src = &buf[i];
316e2bf4
DW		 4011 if (*src > 0x20) {
4012 /* ':' is reserved for use in placeholder serial
4013 * numbers for missing disks
4014 */
4015 if (*src == ':')
4016 *dest++ = ';';
4017 else
4018 *dest++ = *src;
4019 }
4020 }
21e9380b
AP		 4021 len = dest - buf;
4022 dest = buf;
316e2bf4 4023
6da53c0e
BK		 4024 if (len > serial_buf_len) {
4025 /* truncate leading characters */
4026 dest += len - serial_buf_len;
4027 len = serial_buf_len;
316e2bf4 4028 }
5c3db629 4029
6da53c0e 4030 memset(serial, 0, serial_buf_len);
316e2bf4 4031 memcpy(serial, dest, len);
cdddbdbc
DW		 4032
4033 return 0;
4034}
4035
1f24f035
DW		 4036static int serialcmp(__u8 *s1, __u8 *s2)
4037{
4038 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4039}
4040
4041static void serialcpy(__u8 *dest, __u8 *src)
4042{
4043 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4044}
4045
54c2c1ea
DW		 4046static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4047{
4048 struct dl *dl;
4049
4050 for (dl = super->disks; dl; dl = dl->next)
4051 if (serialcmp(dl->serial, serial) == 0)
4052 break;
4053
4054 return dl;
4055}
4056
a2b97981
DW		 4057static struct imsm_disk *
4058__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4059{
4060 int i;
4061
4062 for (i = 0; i < mpb->num_disks; i++) {
4063 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4064
4065 if (serialcmp(disk->serial, serial) == 0) {
4066 if (idx)
4067 *idx = i;
4068 return disk;
4069 }
4070 }
4071
4072 return NULL;
4073}
4074
cdddbdbc
DW		 4075static int
4076load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4077{
a2b97981 4078 struct imsm_disk *disk;
cdddbdbc
DW		 4079 struct dl *dl;
4080 struct stat stb;
cdddbdbc 4081 int rv;
a2b97981 4082 char name[40];
d23fe947
DW		 4083 __u8 serial[MAX_RAID_SERIAL_LEN];
4084
6da53c0e 4085 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4086
4087 if (rv != 0)
4088 return 2;
4089
503975b9 4090 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4091
a2b97981
DW		 4092 fstat(fd, &stb);
4093 dl->major = major(stb.st_rdev);
4094 dl->minor = minor(stb.st_rdev);
4095 dl->next = super->disks;
4096 dl->fd = keep_fd ? fd : -1;
4097 assert(super->disks == NULL);
4098 super->disks = dl;
4099 serialcpy(dl->serial, serial);
4100 dl->index = -2;
4101 dl->e = NULL;
4102 fd2devname(fd, name);
4103 if (devname)
503975b9 4104 dl->devname = xstrdup(devname);
a2b97981 4105 else
503975b9 4106 dl->devname = xstrdup(name);
cdddbdbc 4107
d23fe947 4108 /* look up this disk's index in the current anchor */
a2b97981
DW		 4109 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4110 if (disk) {
4111 dl->disk = *disk;
4112 /* only set index on disks that are a member of a
4113 * populated contianer, i.e. one with raid_devs
4114 */
4115 if (is_failed(&dl->disk))
3f6efecc 4116 dl->index = -2;
2432ce9b 4117 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4118 dl->index = -1;
3f6efecc
DW		 4119 }
4120
949c47a0
DW		 4121 return 0;
4122}
4123
0c046afd
DW		 4124/* When migrating map0 contains the 'destination' state while map1
4125 * contains the current state. When not migrating map0 contains the
4126 * current state. This routine assumes that map[0].map_state is set to
4127 * the current array state before being called.
4128 *
4129 * Migration is indicated by one of the following states
4130 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4131 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4132 * map1state=unitialized)
1484e727 4133 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4134 * map1state=normal)
e3bba0e0 4135 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4136 * map1state=degraded)
8e59f3d8
AK		 4137 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4138 * map1state=normal)
0c046afd 4139 */
8e59f3d8
AK		 4140static void migrate(struct imsm_dev *dev, struct intel_super *super,
4141 __u8 to_state, int migr_type)
3393c6af 4142{
0c046afd 4143 struct imsm_map *dest;
238c0a71 4144 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4145
0c046afd 4146 dev->vol.migr_state = 1;
1484e727 4147 set_migr_type(dev, migr_type);
f8f603f1 4148 dev->vol.curr_migr_unit = 0;
238c0a71 4149 dest = get_imsm_map(dev, MAP_1);
0c046afd 4150
0556e1a2 4151 /* duplicate and then set the target end state in map[0] */
3393c6af 4152 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4153 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4154 __u32 ord;
4155 int i;
4156
4157 for (i = 0; i < src->num_members; i++) {
4158 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4159 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4160 }
4161 }
4162
8e59f3d8
AK		 4163 if (migr_type == MIGR_GEN_MIGR)
4164 /* Clear migration record */
4165 memset(super->migr_rec, 0, sizeof(struct migr_record));
4166
0c046afd 4167 src->map_state = to_state;
949c47a0 4168}
f8f603f1 4169
809da78e
AK		 4170static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4171 __u8 map_state)
f8f603f1 4172{
238c0a71
AK		 4173 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4174 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4175 MAP_0 : MAP_1);
28bce06f 4176 int i, j;
0556e1a2
DW		 4177
4178 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4179 * completed in the last migration.
4180 *
28bce06f 4181 * FIXME add support for raid-level-migration
0556e1a2 4182 */
089f9d79
JS		 4183 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4184 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4185 /* when final map state is other than expected
4186 * merge maps (not for migration)
4187 */
4188 int failed;
4189
4190 for (i = 0; i < prev->num_members; i++)
4191 for (j = 0; j < map->num_members; j++)
4192 /* during online capacity expansion
4193 * disks position can be changed
4194 * if takeover is used
4195 */
4196 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4197 ord_to_idx(prev->disk_ord_tbl[i])) {
4198 map->disk_ord_tbl[j] |=
4199 prev->disk_ord_tbl[i];
4200 break;
4201 }
4202 failed = imsm_count_failed(super, dev, MAP_0);
4203 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4204 }
f8f603f1
DW		 4205
4206 dev->vol.migr_state = 0;
ea672ee1 4207 set_migr_type(dev, 0);
f8f603f1
DW		 4208 dev->vol.curr_migr_unit = 0;
4209 map->map_state = map_state;
4210}
949c47a0
DW		 4211
4212static int parse_raid_devices(struct intel_super *super)
4213{
4214 int i;
4215 struct imsm_dev *dev_new;
4d7b1503 4216 size_t len, len_migr;
401d313b 4217 size_t max_len = 0;
4d7b1503
DW		 4218 size_t space_needed = 0;
4219 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4220
4221 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4222 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4223 struct intel_dev *dv;
949c47a0 4224
4d7b1503
DW		 4225 len = sizeof_imsm_dev(dev_iter, 0);
4226 len_migr = sizeof_imsm_dev(dev_iter, 1);
4227 if (len_migr > len)
4228 space_needed += len_migr - len;
ca9de185 4229
503975b9 4230 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4231 if (max_len < len_migr)
4232 max_len = len_migr;
4233 if (max_len > len_migr)
4234 space_needed += max_len - len_migr;
503975b9 4235 dev_new = xmalloc(max_len);
949c47a0 4236 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4237 dv->dev = dev_new;
4238 dv->index = i;
4239 dv->next = super->devlist;
4240 super->devlist = dv;
949c47a0 4241 }
cdddbdbc 4242
4d7b1503
DW		 4243 /* ensure that super->buf is large enough when all raid devices
4244 * are migrating
4245 */
4246 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4247 void *buf;
4248
f36a9ecd
PB		 4249 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4250 super->sector_size);
4251 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4252 return 1;
4253
1f45a8ad
DW		 4254 memcpy(buf, super->buf, super->len);
4255 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4256 free(super->buf);
4257 super->buf = buf;
4258 super->len = len;
4259 }
ca9de185 4260
bbab0940
TM		 4261 super->extra_space += space_needed;
4262
cdddbdbc
DW		 4263 return 0;
4264}
4265
e2f41b2c
AK		 4266/*******************************************************************************
4267 * Function: check_mpb_migr_compatibility
4268 * Description: Function checks for unsupported migration features:
4269 * - migration optimization area (pba_of_lba0)
4270 * - descending reshape (ascending_migr)
4271 * Parameters:
4272 * super : imsm metadata information
4273 * Returns:
4274 * 0 : migration is compatible
4275 * -1 : migration is not compatible
4276 ******************************************************************************/
4277int check_mpb_migr_compatibility(struct intel_super *super)
4278{
4279 struct imsm_map *map0, *map1;
4280 struct migr_record *migr_rec = super->migr_rec;
4281 int i;
4282
4283 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4284 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4285
4286 if (dev_iter &&
4287 dev_iter->vol.migr_state == 1 &&
4288 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4289 /* This device is migrating */
238c0a71
AK		 4290 map0 = get_imsm_map(dev_iter, MAP_0);
4291 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4292 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4293 /* migration optimization area was used */
4294 return -1;
fc54fe7a
JS		 4295 if (migr_rec->ascending_migr == 0 &&
4296 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4297 /* descending reshape not supported yet */
4298 return -1;
4299 }
4300 }
4301 return 0;
4302}
4303
d23fe947 4304static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4305
cdddbdbc 4306/* load_imsm_mpb - read matrix metadata
f2f5c343 4307 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4308 */
4309static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4310{
4311 unsigned long long dsize;
cdddbdbc 4312 unsigned long long sectors;
f36a9ecd 4313 unsigned int sector_size = super->sector_size;
cdddbdbc 4314 struct stat;
6416d527 4315 struct imsm_super *anchor;
cdddbdbc
DW		 4316 __u32 check_sum;
4317
cdddbdbc 4318 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4319 if (dsize < 2*sector_size) {
64436f06 4320 if (devname)
e7b84f9d
N		 4321 pr_err("%s: device to small for imsm\n",
4322 devname);
64436f06
N		 4323 return 1;
4324 }
cdddbdbc 4325
f36a9ecd 4326 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4327 if (devname)
e7b84f9d
N		 4328 pr_err("Cannot seek to anchor block on %s: %s\n",
4329 devname, strerror(errno));
cdddbdbc
DW		 4330 return 1;
4331 }
4332
f36a9ecd 4333 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4334 if (devname)
7a862a02 4335 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4336 return 1;
4337 }
466070ad 4338 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4339 if (devname)
e7b84f9d
N		 4340 pr_err("Cannot read anchor block on %s: %s\n",
4341 devname, strerror(errno));
6416d527 4342 free(anchor);
cdddbdbc
DW		 4343 return 1;
4344 }
4345
6416d527 4346 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4347 if (devname)
e7b84f9d 4348 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4349 free(anchor);
cdddbdbc
DW		 4350 return 2;
4351 }
4352
d23fe947 4353 __free_imsm(super, 0);
f2f5c343
LM		 4354 /* reload capability and hba */
4355
4356 /* capability and hba must be updated with new super allocation */
d424212e 4357 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4358 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4359 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4360 if (devname)
e7b84f9d
N		 4361 pr_err("unable to allocate %zu byte mpb buffer\n",
4362 super->len);
6416d527 4363 free(anchor);
cdddbdbc
DW		 4364 return 2;
4365 }
f36a9ecd 4366 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4367
f36a9ecd 4368 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4369 free(anchor);
8e59f3d8 4370
85337573
AO		 4371 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4372 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4373 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4374 free(super->buf);
4375 return 2;
4376 }
51d83f5d 4377 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4378
949c47a0 4379 if (!sectors) {
ecf45690
DW		 4380 check_sum = __gen_imsm_checksum(super->anchor);
4381 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4382 if (devname)
e7b84f9d
N		 4383 pr_err("IMSM checksum %x != %x on %s\n",
4384 check_sum,
4385 __le32_to_cpu(super->anchor->check_sum),
4386 devname);
ecf45690
DW		 4387 return 2;
4388 }
4389
a2b97981 4390 return 0;
949c47a0 4391 }
cdddbdbc
DW		 4392
4393 /* read the extended mpb */
f36a9ecd 4394 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4395 if (devname)
e7b84f9d
N		 4396 pr_err("Cannot seek to extended mpb on %s: %s\n",
4397 devname, strerror(errno));
cdddbdbc
DW		 4398 return 1;
4399 }
4400
f36a9ecd
PB		 4401 if ((unsigned int)read(fd, super->buf + sector_size,
4402 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4403 if (devname)
e7b84f9d
N		 4404 pr_err("Cannot read extended mpb on %s: %s\n",
4405 devname, strerror(errno));
cdddbdbc
DW		 4406 return 2;
4407 }
4408
949c47a0
DW		 4409 check_sum = __gen_imsm_checksum(super->anchor);
4410 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4411 if (devname)
e7b84f9d
N		 4412 pr_err("IMSM checksum %x != %x on %s\n",
4413 check_sum, __le32_to_cpu(super->anchor->check_sum),
4414 devname);
db575f3b 4415 return 3;
cdddbdbc
DW		 4416 }
4417
a2b97981
DW		 4418 return 0;
4419}
4420
8e59f3d8
AK		 4421static int read_imsm_migr_rec(int fd, struct intel_super *super);
4422
97f81ee2
CA		 4423/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4424static void clear_hi(struct intel_super *super)
4425{
4426 struct imsm_super *mpb = super->anchor;
4427 int i, n;
4428 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4429 return;
4430 for (i = 0; i < mpb->num_disks; ++i) {
4431 struct imsm_disk *disk = &mpb->disk[i];
4432 disk->total_blocks_hi = 0;
4433 }
4434 for (i = 0; i < mpb->num_raid_devs; ++i) {
4435 struct imsm_dev *dev = get_imsm_dev(super, i);
4436 if (!dev)
4437 return;
4438 for (n = 0; n < 2; ++n) {
4439 struct imsm_map *map = get_imsm_map(dev, n);
4440 if (!map)
4441 continue;
4442 map->pba_of_lba0_hi = 0;
4443 map->blocks_per_member_hi = 0;
4444 map->num_data_stripes_hi = 0;
4445 }
4446 }
4447}
4448
a2b97981
DW		 4449static int
4450load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4451{
4452 int err;
4453
4454 err = load_imsm_mpb(fd, super, devname);
4455 if (err)
4456 return err;
f36a9ecd
PB		 4457 if (super->sector_size == 4096)
4458 convert_from_4k(super);
a2b97981
DW		 4459 err = load_imsm_disk(fd, super, devname, keep_fd);
4460 if (err)
4461 return err;
4462 err = parse_raid_devices(super);
8d67477f
TM		 4463 if (err)
4464 return err;
4465 err = load_bbm_log(super);
97f81ee2 4466 clear_hi(super);
a2b97981 4467 return err;
cdddbdbc
DW		 4468}
4469
ae6aad82
DW		 4470static void __free_imsm_disk(struct dl *d)
4471{
4472 if (d->fd >= 0)
4473 close(d->fd);
4474 if (d->devname)
4475 free(d->devname);
0dcecb2e
DW		 4476 if (d->e)
4477 free(d->e);
ae6aad82
DW		 4478 free(d);
4479
4480}
1a64be56 4481
cdddbdbc
DW		 4482static void free_imsm_disks(struct intel_super *super)
4483{
47ee5a45 4484 struct dl *d;
cdddbdbc 4485
47ee5a45
DW		 4486 while (super->disks) {
4487 d = super->disks;
cdddbdbc 4488 super->disks = d->next;
ae6aad82 4489 __free_imsm_disk(d);
cdddbdbc 4490 }
cb82edca
AK		 4491 while (super->disk_mgmt_list) {
4492 d = super->disk_mgmt_list;
4493 super->disk_mgmt_list = d->next;
4494 __free_imsm_disk(d);
4495 }
47ee5a45
DW		 4496 while (super->missing) {
4497 d = super->missing;
4498 super->missing = d->next;
4499 __free_imsm_disk(d);
4500 }
4501
cdddbdbc
DW		 4502}
4503
9ca2c81c 4504/* free all the pieces hanging off of a super pointer */
d23fe947 4505static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4506{
88654014
LM		 4507 struct intel_hba *elem, *next;
4508
9ca2c81c 4509 if (super->buf) {
949c47a0 4510 free(super->buf);
9ca2c81c
DW		 4511 super->buf = NULL;
4512 }
f2f5c343
LM		 4513 /* unlink capability description */
4514 super->orom = NULL;
8e59f3d8
AK		 4515 if (super->migr_rec_buf) {
4516 free(super->migr_rec_buf);
4517 super->migr_rec_buf = NULL;
4518 }
d23fe947
DW		 4519 if (free_disks)
4520 free_imsm_disks(super);
ba2de7ba 4521 free_devlist(super);
88654014
LM		 4522 elem = super->hba;
4523 while (elem) {
4524 if (elem->path)
4525 free((void *)elem->path);
4526 next = elem->next;
4527 free(elem);
4528 elem = next;
88c32bb1 4529 }
8d67477f
TM		 4530 if (super->bbm_log)
4531 free(super->bbm_log);
88654014 4532 super->hba = NULL;
cdddbdbc
DW		 4533}
4534
9ca2c81c
DW		 4535static void free_imsm(struct intel_super *super)
4536{
d23fe947 4537 __free_imsm(super, 1);
928f1424 4538 free(super->bb.entries);
9ca2c81c
DW		 4539 free(super);
4540}
cdddbdbc
DW		 4541
4542static void free_super_imsm(struct supertype *st)
4543{
4544 struct intel_super *super = st->sb;
4545
4546 if (!super)
4547 return;
4548
4549 free_imsm(super);
4550 st->sb = NULL;
4551}
4552
49133e57 4553static struct intel_super *alloc_super(void)
c2c087e6 4554{
503975b9 4555 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4556
503975b9
N		 4557 super->current_vol = -1;
4558 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4559
4560 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4561 sizeof(struct md_bb_entry));
4562 if (!super->bb.entries) {
4563 free(super);
4564 return NULL;
4565 }
4566
c2c087e6
DW		 4567 return super;
4568}
4569
f0f5a016
LM		 4570/*
4571 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4572 */
d424212e 4573static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4574{
4575 struct sys_dev *hba_name;
4576 int rv = 0;
4577
3a30e28e
MT		 4578 if (fd >= 0 && test_partition(fd)) {
4579 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4580 devname);
4581 return 1;
4582 }
089f9d79 4583 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4584 super->orom = NULL;
f0f5a016
LM		 4585 super->hba = NULL;
4586 return 0;
4587 }
4588 hba_name = find_disk_attached_hba(fd, NULL);
4589 if (!hba_name) {
d424212e 4590 if (devname)
e7b84f9d
N		 4591 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4592 devname);
f0f5a016
LM		 4593 return 1;
4594 }
4595 rv = attach_hba_to_super(super, hba_name);
4596 if (rv == 2) {
d424212e
N		 4597 if (devname) {
4598 struct intel_hba *hba = super->hba;
f0f5a016 4599
60f0f54d
PB		 4600 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4601 " but the container is assigned to Intel(R) %s %s (",
d424212e 4602 devname,
614902f6 4603 get_sys_dev_type(hba_name->type),
60f0f54d 4604 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4605 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4606 get_sys_dev_type(super->hba->type),
4607 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4608
f0f5a016
LM		 4609 while (hba) {
4610 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4611 if (hba->next)
4612 fprintf(stderr, ", ");
4613 hba = hba->next;
4614 }
6b781d33 4615 fprintf(stderr, ").\n"
cca67208 4616 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4617 }
f0f5a016
LM		 4618 return 2;
4619 }
6b781d33 4620 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4621 if (!super->orom)
4622 return 3;
614902f6 4623
f0f5a016
LM		 4624 return 0;
4625}
4626
47ee5a45
DW		 4627/* find_missing - helper routine for load_super_imsm_all that identifies
4628 * disks that have disappeared from the system. This routine relies on
4629 * the mpb being uptodate, which it is at load time.
4630 */
4631static int find_missing(struct intel_super *super)
4632{
4633 int i;
4634 struct imsm_super *mpb = super->anchor;
4635 struct dl *dl;
4636 struct imsm_disk *disk;
47ee5a45
DW		 4637
4638 for (i = 0; i < mpb->num_disks; i++) {
4639 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4640 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4641 if (dl)
4642 continue;
47ee5a45 4643
503975b9 4644 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4645 dl->major = 0;
4646 dl->minor = 0;
4647 dl->fd = -1;
503975b9 4648 dl->devname = xstrdup("missing");
47ee5a45
DW		 4649 dl->index = i;
4650 serialcpy(dl->serial, disk->serial);
4651 dl->disk = *disk;
689c9bf3 4652 dl->e = NULL;
47ee5a45
DW		 4653 dl->next = super->missing;
4654 super->missing = dl;
4655 }
4656
4657 return 0;
4658}
4659
a2b97981
DW		 4660static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4661{
4662 struct intel_disk *idisk = disk_list;
4663
4664 while (idisk) {
4665 if (serialcmp(idisk->disk.serial, serial) == 0)
4666 break;
4667 idisk = idisk->next;
4668 }
4669
4670 return idisk;
4671}
4672
4673static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4674 struct intel_super *super,
4675 struct intel_disk **disk_list)
4676{
4677 struct imsm_disk *d = &super->disks->disk;
4678 struct imsm_super *mpb = super->anchor;
4679 int i, j;
4680
4681 for (i = 0; i < tbl_size; i++) {
4682 struct imsm_super *tbl_mpb = table[i]->anchor;
4683 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4684
4685 if (tbl_mpb->family_num == mpb->family_num) {
4686 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4687 dprintf("mpb from %d:%d matches %d:%d\n",
4688 super->disks->major,
a2b97981
DW		 4689 super->disks->minor,
4690 table[i]->disks->major,
4691 table[i]->disks->minor);
4692 break;
4693 }
4694
4695 if (((is_configured(d) && !is_configured(tbl_d)) ||
4696 is_configured(d) == is_configured(tbl_d)) &&
4697 tbl_mpb->generation_num < mpb->generation_num) {
4698 /* current version of the mpb is a
4699 * better candidate than the one in
4700 * super_table, but copy over "cross
4701 * generational" status
4702 */
4703 struct intel_disk *idisk;
4704
1ade5cc1
N		 4705 dprintf("mpb from %d:%d replaces %d:%d\n",
4706 super->disks->major,
a2b97981
DW		 4707 super->disks->minor,
4708 table[i]->disks->major,
4709 table[i]->disks->minor);
4710
4711 idisk = disk_list_get(tbl_d->serial, *disk_list);
4712 if (idisk && is_failed(&idisk->disk))
4713 tbl_d->status |= FAILED_DISK;
4714 break;
4715 } else {
4716 struct intel_disk *idisk;
4717 struct imsm_disk *disk;
4718
4719 /* tbl_mpb is more up to date, but copy
4720 * over cross generational status before
4721 * returning
4722 */
4723 disk = __serial_to_disk(d->serial, mpb, NULL);
4724 if (disk && is_failed(disk))
4725 d->status |= FAILED_DISK;
4726
4727 idisk = disk_list_get(d->serial, *disk_list);
4728 if (idisk) {
4729 idisk->owner = i;
4730 if (disk && is_configured(disk))
4731 idisk->disk.status |= CONFIGURED_DISK;
4732 }
4733
1ade5cc1
N		 4734 dprintf("mpb from %d:%d prefer %d:%d\n",
4735 super->disks->major,
a2b97981
DW		 4736 super->disks->minor,
4737 table[i]->disks->major,
4738 table[i]->disks->minor);
4739
4740 return tbl_size;
4741 }
4742 }
4743 }
4744
4745 if (i >= tbl_size)
4746 table[tbl_size++] = super;
4747 else
4748 table[i] = super;
4749
4750 /* update/extend the merged list of imsm_disk records */
4751 for (j = 0; j < mpb->num_disks; j++) {
4752 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4753 struct intel_disk *idisk;
4754
4755 idisk = disk_list_get(disk->serial, *disk_list);
4756 if (idisk) {
4757 idisk->disk.status |= disk->status;
4758 if (is_configured(&idisk->disk) ||
4759 is_failed(&idisk->disk))
4760 idisk->disk.status &= ~(SPARE_DISK);
4761 } else {
503975b9 4762 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4763 idisk->owner = IMSM_UNKNOWN_OWNER;
4764 idisk->disk = *disk;
4765 idisk->next = *disk_list;
4766 *disk_list = idisk;
4767 }
4768
4769 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4770 idisk->owner = i;
4771 }
4772
4773 return tbl_size;
4774}
4775
4776static struct intel_super *
4777validate_members(struct intel_super *super, struct intel_disk *disk_list,
4778 const int owner)
4779{
4780 struct imsm_super *mpb = super->anchor;
4781 int ok_count = 0;
4782 int i;
4783
4784 for (i = 0; i < mpb->num_disks; i++) {
4785 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4786 struct intel_disk *idisk;
4787
4788 idisk = disk_list_get(disk->serial, disk_list);
4789 if (idisk) {
4790 if (idisk->owner == owner ||
4791 idisk->owner == IMSM_UNKNOWN_OWNER)
4792 ok_count++;
4793 else
1ade5cc1
N		 4794 dprintf("'%.16s' owner %d != %d\n",
4795 disk->serial, idisk->owner,
a2b97981
DW		 4796 owner);
4797 } else {
1ade5cc1
N		 4798 dprintf("unknown disk %x [%d]: %.16s\n",
4799 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4800 disk->serial);
4801 break;
4802 }
4803 }
4804
4805 if (ok_count == mpb->num_disks)
4806 return super;
4807 return NULL;
4808}
4809
4810static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4811{
4812 struct intel_super *s;
4813
4814 for (s = super_list; s; s = s->next) {
4815 if (family_num != s->anchor->family_num)
4816 continue;
e12b3daa 4817 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4818 __le32_to_cpu(family_num), s->disks->devname);
4819 }
4820}
4821
4822static struct intel_super *
4823imsm_thunderdome(struct intel_super **super_list, int len)
4824{
4825 struct intel_super *super_table[len];
4826 struct intel_disk *disk_list = NULL;
4827 struct intel_super *champion, *spare;
4828 struct intel_super *s, **del;
4829 int tbl_size = 0;
4830 int conflict;
4831 int i;
4832
4833 memset(super_table, 0, sizeof(super_table));
4834 for (s = *super_list; s; s = s->next)
4835 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4836
4837 for (i = 0; i < tbl_size; i++) {
4838 struct imsm_disk *d;
4839 struct intel_disk *idisk;
4840 struct imsm_super *mpb = super_table[i]->anchor;
4841
4842 s = super_table[i];
4843 d = &s->disks->disk;
4844
4845 /* 'd' must appear in merged disk list for its
4846 * configuration to be valid
4847 */
4848 idisk = disk_list_get(d->serial, disk_list);
4849 if (idisk && idisk->owner == i)
4850 s = validate_members(s, disk_list, i);
4851 else
4852 s = NULL;
4853
4854 if (!s)
1ade5cc1
N		 4855 dprintf("marking family: %#x from %d:%d offline\n",
4856 mpb->family_num,
a2b97981
DW		 4857 super_table[i]->disks->major,
4858 super_table[i]->disks->minor);
4859 super_table[i] = s;
4860 }
4861
4862 /* This is where the mdadm implementation differs from the Windows
4863 * driver which has no strict concept of a container. We can only
4864 * assemble one family from a container, so when returning a prodigal
4865 * array member to this system the code will not be able to disambiguate
4866 * the container contents that should be assembled ("foreign" versus
4867 * "local"). It requires user intervention to set the orig_family_num
4868 * to a new value to establish a new container. The Windows driver in
4869 * this situation fixes up the volume name in place and manages the
4870 * foreign array as an independent entity.
4871 */
4872 s = NULL;
4873 spare = NULL;
4874 conflict = 0;
4875 for (i = 0; i < tbl_size; i++) {
4876 struct intel_super *tbl_ent = super_table[i];
4877 int is_spare = 0;
4878
4879 if (!tbl_ent)
4880 continue;
4881
4882 if (tbl_ent->anchor->num_raid_devs == 0) {
4883 spare = tbl_ent;
4884 is_spare = 1;
4885 }
4886
4887 if (s && !is_spare) {
4888 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4889 conflict++;
4890 } else if (!s && !is_spare)
4891 s = tbl_ent;
4892 }
4893
4894 if (!s)
4895 s = spare;
4896 if (!s) {
4897 champion = NULL;
4898 goto out;
4899 }
4900 champion = s;
4901
4902 if (conflict)
7a862a02 4903 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4904 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4905
4906 /* collect all dl's onto 'champion', and update them to
4907 * champion's version of the status
4908 */
4909 for (s = *super_list; s; s = s->next) {
4910 struct imsm_super *mpb = champion->anchor;
4911 struct dl *dl = s->disks;
4912
4913 if (s == champion)
4914 continue;
4915
5d7b407a
CA		 4916 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4917
a2b97981
DW		 4918 for (i = 0; i < mpb->num_disks; i++) {
4919 struct imsm_disk *disk;
4920
4921 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4922 if (disk) {
4923 dl->disk = *disk;
4924 /* only set index on disks that are a member of
4925 * a populated contianer, i.e. one with
4926 * raid_devs
4927 */
4928 if (is_failed(&dl->disk))
4929 dl->index = -2;
4930 else if (is_spare(&dl->disk))
4931 dl->index = -1;
4932 break;
4933 }
4934 }
4935
4936 if (i >= mpb->num_disks) {
4937 struct intel_disk *idisk;
4938
4939 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4940 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4941 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4942 dl->index = -1;
4943 else {
4944 dl->index = -2;
4945 continue;
4946 }
4947 }
4948
4949 dl->next = champion->disks;
4950 champion->disks = dl;
4951 s->disks = NULL;
4952 }
4953
4954 /* delete 'champion' from super_list */
4955 for (del = super_list; *del; ) {
4956 if (*del == champion) {
4957 *del = (*del)->next;
4958 break;
4959 } else
4960 del = &(*del)->next;
4961 }
4962 champion->next = NULL;
4963
4964 out:
4965 while (disk_list) {
4966 struct intel_disk *idisk = disk_list;
4967
4968 disk_list = disk_list->next;
4969 free(idisk);
4970 }
4971
4972 return champion;
4973}
4974
9587c373
LM		 4975static int
4976get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4977static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4978 int major, int minor, int keep_fd);
ec50f7b6
LM		 4979static int
4980get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4981 int *max, int keep_fd);
4982
cdddbdbc 4983static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4984 char *devname, struct md_list *devlist,
4985 int keep_fd)
cdddbdbc 4986{
a2b97981
DW		 4987 struct intel_super *super_list = NULL;
4988 struct intel_super *super = NULL;
a2b97981 4989 int err = 0;
9587c373 4990 int i = 0;
dab4a513 4991
9587c373
LM		 4992 if (fd >= 0)
4993 /* 'fd' is an opened container */
4994 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4995 else
ec50f7b6
LM		 4996 /* get super block from devlist devices */
4997 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4998 if (err)
1602d52c 4999 goto error;
a2b97981
DW		 5000 /* all mpbs enter, maybe one leaves */
5001 super = imsm_thunderdome(&super_list, i);
5002 if (!super) {
5003 err = 1;
5004 goto error;
cdddbdbc
DW		 5005 }
5006
47ee5a45
DW		 5007 if (find_missing(super) != 0) {
5008 free_imsm(super);
a2b97981
DW		 5009 err = 2;
5010 goto error;
47ee5a45 5011 }
8e59f3d8
AK		 5012
5013 /* load migration record */
5014 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 5015 if (err == -1) {
5016 /* migration is in progress,
5017 * but migr_rec cannot be loaded,
5018 */
8e59f3d8
AK		 5019 err = 4;
5020 goto error;
5021 }
e2f41b2c
AK		 5022
5023 /* Check migration compatibility */
089f9d79 5024 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5025 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5026 if (devname)
5027 fprintf(stderr, " on %s\n", devname);
5028 else
5029 fprintf(stderr, " (IMSM).\n");
5030
5031 err = 5;
5032 goto error;
5033 }
5034
a2b97981
DW		 5035 err = 0;
5036
5037 error:
5038 while (super_list) {
5039 struct intel_super *s = super_list;
5040
5041 super_list = super_list->next;
5042 free_imsm(s);
5043 }
9587c373 5044
a2b97981
DW		 5045 if (err)
5046 return err;
f7e7067b 5047
cdddbdbc 5048 *sbp = super;
9587c373 5049 if (fd >= 0)
4dd2df09 5050 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5051 else
4dd2df09 5052 st->container_devnm[0] = 0;
a2b97981 5053 if (err == 0 && st->ss == NULL) {
bf5a934a 5054 st->ss = &super_imsm;
cdddbdbc
DW		 5055 st->minor_version = 0;
5056 st->max_devs = IMSM_MAX_DEVICES;
5057 }
cdddbdbc
DW		 5058 return 0;
5059}
2b959fbf 5060
ec50f7b6
LM		 5061static int
5062get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5063 int *max, int keep_fd)
5064{
5065 struct md_list *tmpdev;
5066 int err = 0;
5067 int i = 0;
9587c373 5068
ec50f7b6
LM		 5069 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5070 if (tmpdev->used != 1)
5071 continue;
5072 if (tmpdev->container == 1) {
ca9de185 5073 int lmax = 0;
ec50f7b6
LM		 5074 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5075 if (fd < 0) {
e7b84f9d 5076 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5077 tmpdev->devname, strerror(errno));
5078 err = 8;
5079 goto error;
5080 }
5081 err = get_sra_super_block(fd, super_list,
5082 tmpdev->devname, &lmax,
5083 keep_fd);
5084 i += lmax;
5085 close(fd);
5086 if (err) {
5087 err = 7;
5088 goto error;
5089 }
5090 } else {
5091 int major = major(tmpdev->st_rdev);
5092 int minor = minor(tmpdev->st_rdev);
5093 err = get_super_block(super_list,
4dd2df09 5094 NULL,
ec50f7b6
LM		 5095 tmpdev->devname,
5096 major, minor,
5097 keep_fd);
5098 i++;
5099 if (err) {
5100 err = 6;
5101 goto error;
5102 }
5103 }
5104 }
5105 error:
5106 *max = i;
5107 return err;
5108}
9587c373 5109
4dd2df09 5110static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5111 int major, int minor, int keep_fd)
5112{
594dc1b8 5113 struct intel_super *s;
9587c373
LM		 5114 char nm[32];
5115 int dfd = -1;
9587c373
LM		 5116 int err = 0;
5117 int retry;
5118
5119 s = alloc_super();
5120 if (!s) {
5121 err = 1;
5122 goto error;
5123 }
5124
5125 sprintf(nm, "%d:%d", major, minor);
5126 dfd = dev_open(nm, O_RDWR);
5127 if (dfd < 0) {
5128 err = 2;
5129 goto error;
5130 }
5131
fa7bb6f8 5132 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5133 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5134 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5135
5136 /* retry the load if we might have raced against mdmon */
4dd2df09 5137 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5138 for (retry = 0; retry < 3; retry++) {
5139 usleep(3000);
5140 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5141 if (err != 3)
5142 break;
5143 }
5144 error:
5145 if (!err) {
5146 s->next = *super_list;
5147 *super_list = s;
5148 } else {
5149 if (s)
8d67477f 5150 free_imsm(s);
36614e95 5151 if (dfd >= 0)
9587c373
LM		 5152 close(dfd);
5153 }
089f9d79 5154 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5155 close(dfd);
5156 return err;
5157
5158}
5159
5160static int
5161get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5162{
5163 struct mdinfo *sra;
4dd2df09 5164 char *devnm;
9587c373
LM		 5165 struct mdinfo *sd;
5166 int err = 0;
5167 int i = 0;
4dd2df09 5168 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5169 if (!sra)
5170 return 1;
5171
5172 if (sra->array.major_version != -1 ||
5173 sra->array.minor_version != -2 ||
5174 strcmp(sra->text_version, "imsm") != 0) {
5175 err = 1;
5176 goto error;
5177 }
5178 /* load all mpbs */
4dd2df09 5179 devnm = fd2devnm(fd);
9587c373 5180 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5181 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5182 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5183 err = 7;
5184 goto error;
5185 }
5186 }
5187 error:
5188 sysfs_free(sra);
5189 *max = i;
5190 return err;
5191}
5192
2b959fbf
N		 5193static int load_container_imsm(struct supertype *st, int fd, char *devname)
5194{
ec50f7b6 5195 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5196}
cdddbdbc
DW		 5197
5198static int load_super_imsm(struct supertype *st, int fd, char *devname)
5199{
5200 struct intel_super *super;
5201 int rv;
8a3544f8 5202 int retry;
cdddbdbc 5203
357ac106 5204 if (test_partition(fd))
691c6ee1
N		 5205 /* IMSM not allowed on partitions */
5206 return 1;
5207
37424f13
DW		 5208 free_super_imsm(st);
5209
49133e57 5210 super = alloc_super();
fa7bb6f8 5211 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5212 if (!super)
5213 return 1;
ea2bc72b
LM		 5214 /* Load hba and capabilities if they exist.
5215 * But do not preclude loading metadata in case capabilities or hba are
5216 * non-compliant and ignore_hw_compat is set.
5217 */
d424212e 5218 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5219 /* no orom/efi or non-intel hba of the disk */
089f9d79 5220 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5221 if (devname)
e7b84f9d 5222 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5223 free_imsm(super);
5224 return 2;
5225 }
a2b97981 5226 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5227
8a3544f8
AP		 5228 /* retry the load if we might have raced against mdmon */
5229 if (rv == 3) {
f96b1302
AP		 5230 struct mdstat_ent *mdstat = NULL;
5231 char *name = fd2kname(fd);
5232
5233 if (name)
5234 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5235
5236 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5237 for (retry = 0; retry < 3; retry++) {
5238 usleep(3000);
5239 rv = load_and_parse_mpb(fd, super, devname, 0);
5240 if (rv != 3)
5241 break;
5242 }
5243 }
5244
5245 free_mdstat(mdstat);
5246 }
5247
cdddbdbc
DW		 5248 if (rv) {
5249 if (devname)
7a862a02 5250 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5251 free_imsm(super);
5252 return rv;
5253 }
5254
5255 st->sb = super;
5256 if (st->ss == NULL) {
5257 st->ss = &super_imsm;
5258 st->minor_version = 0;
5259 st->max_devs = IMSM_MAX_DEVICES;
5260 }
8e59f3d8
AK		 5261
5262 /* load migration record */
2e062e82
AK		 5263 if (load_imsm_migr_rec(super, NULL) == 0) {
5264 /* Check for unsupported migration features */
5265 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5266 pr_err("Unsupported migration detected");
2e062e82
AK		 5267 if (devname)
5268 fprintf(stderr, " on %s\n", devname);
5269 else
5270 fprintf(stderr, " (IMSM).\n");
5271 return 3;
5272 }
e2f41b2c
AK		 5273 }
5274
cdddbdbc
DW		 5275 return 0;
5276}
5277
ef6ffade
DW		 5278static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5279{
5280 if (info->level == 1)
5281 return 128;
5282 return info->chunk_size >> 9;
5283}
5284
5551b113
CA		 5285static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5286 unsigned long long size)
fcfd9599 5287{
4025c288 5288 if (info->level == 1)
5551b113 5289 return size * 2;
4025c288 5290 else
5551b113 5291 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5292}
5293
4d1313e9
DW		 5294static void imsm_update_version_info(struct intel_super *super)
5295{
5296 /* update the version and attributes */
5297 struct imsm_super *mpb = super->anchor;
5298 char *version;
5299 struct imsm_dev *dev;
5300 struct imsm_map *map;
5301 int i;
5302
5303 for (i = 0; i < mpb->num_raid_devs; i++) {
5304 dev = get_imsm_dev(super, i);
238c0a71 5305 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5306 if (__le32_to_cpu(dev->size_high) > 0)
5307 mpb->attributes |= MPB_ATTRIB_2TB;
5308
5309 /* FIXME detect when an array spans a port multiplier */
5310 #if 0
5311 mpb->attributes |= MPB_ATTRIB_PM;
5312 #endif
5313
5314 if (mpb->num_raid_devs > 1 ||
5315 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5316 version = MPB_VERSION_ATTRIBS;
5317 switch (get_imsm_raid_level(map)) {
5318 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5319 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5320 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5321 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5322 }
5323 } else {
5324 if (map->num_members >= 5)
5325 version = MPB_VERSION_5OR6_DISK_ARRAY;
5326 else if (dev->status == DEV_CLONE_N_GO)
5327 version = MPB_VERSION_CNG;
5328 else if (get_imsm_raid_level(map) == 5)
5329 version = MPB_VERSION_RAID5;
5330 else if (map->num_members >= 3)
5331 version = MPB_VERSION_3OR4_DISK_ARRAY;
5332 else if (get_imsm_raid_level(map) == 1)
5333 version = MPB_VERSION_RAID1;
5334 else
5335 version = MPB_VERSION_RAID0;
5336 }
5337 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5338 }
5339}
5340
aa534678
DW		 5341static int check_name(struct intel_super *super, char *name, int quiet)
5342{
5343 struct imsm_super *mpb = super->anchor;
5344 char *reason = NULL;
9bd99a90
RS		 5345 char *start = name;
5346 size_t len = strlen(name);
aa534678
DW		 5347 int i;
5348
9bd99a90
RS		 5349 if (len > 0) {
5350 while (isspace(start[len - 1]))
5351 start[--len] = 0;
5352 while (*start && isspace(*start))
5353 ++start, --len;
5354 memmove(name, start, len + 1);
5355 }
5356
5357 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5358 reason = "must be 16 characters or less";
9bd99a90
RS		 5359 else if (len == 0)
5360 reason = "must be a non-empty string";
aa534678
DW		 5361
5362 for (i = 0; i < mpb->num_raid_devs; i++) {
5363 struct imsm_dev *dev = get_imsm_dev(super, i);
5364
5365 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5366 reason = "already exists";
5367 break;
5368 }
5369 }
5370
5371 if (reason && !quiet)
e7b84f9d 5372 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5373
5374 return !reason;
5375}
5376
8b353278 5377static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5378 struct shape *s, char *name,
83cd1e97
N		 5379 char *homehost, int *uuid,
5380 long long data_offset)
cdddbdbc 5381{
c2c087e6
DW		 5382 /* We are creating a volume inside a pre-existing container.
5383 * so st->sb is already set.
5384 */
5385 struct intel_super *super = st->sb;
f36a9ecd 5386 unsigned int sector_size = super->sector_size;
949c47a0 5387 struct imsm_super *mpb = super->anchor;
ba2de7ba 5388 struct intel_dev *dv;
c2c087e6
DW		 5389 struct imsm_dev *dev;
5390 struct imsm_vol *vol;
5391 struct imsm_map *map;
5392 int idx = mpb->num_raid_devs;
5393 int i;
760365f9 5394 int namelen;
c2c087e6 5395 unsigned long long array_blocks;
2c092cad 5396 size_t size_old, size_new;
5551b113 5397 unsigned long long num_data_stripes;
b53bfba6
TM		 5398 unsigned int data_disks;
5399 unsigned long long size_per_member;
cdddbdbc 5400
88c32bb1 5401 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5402 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5403 return 0;
5404 }
5405
2c092cad
DW		 5406 /* ensure the mpb is large enough for the new data */
5407 size_old = __le32_to_cpu(mpb->mpb_size);
5408 size_new = disks_to_mpb_size(info->nr_disks);
5409 if (size_new > size_old) {
5410 void *mpb_new;
f36a9ecd 5411 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5412
f36a9ecd 5413 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5414 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5415 return 0;
5416 }
85337573
AO		 5417 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5418 MIGR_REC_BUF_SECTORS*
5419 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5420 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5421 free(super->buf);
5422 free(super);
ea944c8f 5423 free(mpb_new);
8e59f3d8
AK		 5424 return 0;
5425 }
2c092cad
DW		 5426 memcpy(mpb_new, mpb, size_old);
5427 free(mpb);
5428 mpb = mpb_new;
949c47a0 5429 super->anchor = mpb_new;
2c092cad
DW		 5430 mpb->mpb_size = __cpu_to_le32(size_new);
5431 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5432 super->len = size_round;
2c092cad 5433 }
bf5a934a 5434 super->current_vol = idx;
3960e579
DW		 5435
5436 /* handle 'failed_disks' by either:
5437 * a) create dummy disk entries in the table if this the first
5438 * volume in the array. We add them here as this is the only
5439 * opportunity to add them. add_to_super_imsm_volume()
5440 * handles the non-failed disks and continues incrementing
5441 * mpb->num_disks.
5442 * b) validate that 'failed_disks' matches the current number
5443 * of missing disks if the container is populated
d23fe947 5444 */
3960e579 5445 if (super->current_vol == 0) {
d23fe947 5446 mpb->num_disks = 0;
3960e579
DW		 5447 for (i = 0; i < info->failed_disks; i++) {
5448 struct imsm_disk *disk;
5449
5450 mpb->num_disks++;
5451 disk = __get_imsm_disk(mpb, i);
5452 disk->status = CONFIGURED_DISK | FAILED_DISK;
5453 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5454 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5455 "missing:%d", (__u8)i);
3960e579
DW		 5456 }
5457 find_missing(super);
5458 } else {
5459 int missing = 0;
5460 struct dl *d;
5461
5462 for (d = super->missing; d; d = d->next)
5463 missing++;
5464 if (info->failed_disks > missing) {
e7b84f9d 5465 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5466 return 0;
5467 }
5468 }
5a038140 5469
aa534678
DW		 5470 if (!check_name(super, name, 0))
5471 return 0;
503975b9
N		 5472 dv = xmalloc(sizeof(*dv));
5473 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5474 /*
5475 * Explicitly allow truncating to not confuse gcc's
5476 * -Werror=stringop-truncation
5477 */
5478 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5479 memcpy(dev->volume, name, namelen);
e03640bd 5480 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5481 info->layout, info->chunk_size,
b53bfba6
TM		 5482 s->size * BLOCKS_PER_KB);
5483 data_disks = get_data_disks(info->level, info->layout,
5484 info->raid_disks);
5485 array_blocks = round_size_to_mb(array_blocks, data_disks);
5486 size_per_member = array_blocks / data_disks;
979d38be 5487
fcc2c9da 5488 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5489 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5490 vol = &dev->vol;
5491 vol->migr_state = 0;
1484e727 5492 set_migr_type(dev, MIGR_INIT);
3960e579 5493 vol->dirty = !info->state;
f8f603f1 5494 vol->curr_migr_unit = 0;
238c0a71 5495 map = get_imsm_map(dev, MAP_0);
5551b113 5496 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5497 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5498 map->failed_disk_num = ~0;
bf4442ab 5499 if (info->level > 0)
fffaf1ff
N		 5500 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5501 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5502 else
5503 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5504 IMSM_T_STATE_NORMAL;
252d23c0 5505 map->ddf = 1;
ef6ffade
DW		 5506
5507 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5508 free(dev);
5509 free(dv);
7a862a02 5510 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5511 return 0;
5512 }
81062a36
DW		 5513
5514 map->raid_level = info->level;
4d1313e9 5515 if (info->level == 10) {
c2c087e6 5516 map->raid_level = 1;
4d1313e9 5517 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5518 } else if (info->level == 1)
5519 map->num_domains = info->raid_disks;
5520 else
ff596308 5521 map->num_domains = 1;
81062a36 5522
5551b113 5523 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5524 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5525 num_data_stripes /= map->num_domains;
5526 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5527
44490938
MD		 5528 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5529 set_blocks_per_member(map, info_to_blocks_per_member(info,
5530 size_per_member /
5531 BLOCKS_PER_KB));
5532
c2c087e6
DW		 5533 map->num_members = info->raid_disks;
5534 for (i = 0; i < map->num_members; i++) {
5535 /* initialized in add_to_super */
4eb26970 5536 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5537 }
949c47a0 5538 mpb->num_raid_devs++;
2a24dc1b
PB		 5539 mpb->num_raid_devs_created++;
5540 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5541
b7580566 5542 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5543 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5544 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5545 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5546 } else {
5547 free(dev);
5548 free(dv);
5549 pr_err("imsm does not support consistency policy %s\n",
5550 map_num(consistency_policies, s->consistency_policy));
5551 return 0;
5552 }
5553
ba2de7ba
DW		 5554 dv->dev = dev;
5555 dv->index = super->current_vol;
5556 dv->next = super->devlist;
5557 super->devlist = dv;
c2c087e6 5558
4d1313e9
DW		 5559 imsm_update_version_info(super);
5560
c2c087e6 5561 return 1;
cdddbdbc
DW		 5562}
5563
bf5a934a 5564static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5565 struct shape *s, char *name,
83cd1e97
N		 5566 char *homehost, int *uuid,
5567 unsigned long long data_offset)
bf5a934a
DW		 5568{
5569 /* This is primarily called by Create when creating a new array.
5570 * We will then get add_to_super called for each component, and then
5571 * write_init_super called to write it out to each device.
5572 * For IMSM, Create can create on fresh devices or on a pre-existing
5573 * array.
5574 * To create on a pre-existing array a different method will be called.
5575 * This one is just for fresh drives.
5576 */
5577 struct intel_super *super;
5578 struct imsm_super *mpb;
5579 size_t mpb_size;
4d1313e9 5580 char *version;
bf5a934a 5581
83cd1e97 5582 if (data_offset != INVALID_SECTORS) {
ed503f89 5583 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5584 return 0;
5585 }
5586
bf5a934a 5587 if (st->sb)
5308f117 5588 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5589 data_offset);
e683ca88
DW		 5590
5591 if (info)
5592 mpb_size = disks_to_mpb_size(info->nr_disks);
5593 else
f36a9ecd 5594 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5595
49133e57 5596 super = alloc_super();
f36a9ecd
PB		 5597 if (super &&
5598 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5599 free_imsm(super);
e683ca88
DW		 5600 super = NULL;
5601 }
5602 if (!super) {
1ade5cc1 5603 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5604 return 0;
5605 }
de44e46f
PB		 5606 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5607 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5608 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5609 free(super->buf);
8d67477f 5610 free_imsm(super);
8e59f3d8
AK		 5611 return 0;
5612 }
e683ca88 5613 memset(super->buf, 0, mpb_size);
ef649044 5614 mpb = super->buf;
e683ca88
DW		 5615 mpb->mpb_size = __cpu_to_le32(mpb_size);
5616 st->sb = super;
5617
5618 if (info == NULL) {
5619 /* zeroing superblock */
5620 return 0;
5621 }
bf5a934a 5622
4d1313e9
DW		 5623 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5624
5625 version = (char *) mpb->sig;
5626 strcpy(version, MPB_SIGNATURE);
5627 version += strlen(MPB_SIGNATURE);
5628 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5629
bf5a934a
DW		 5630 return 1;
5631}
5632
f2cc4f7d
AO		 5633static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5634{
5635 unsigned int member_sector_size;
5636
5637 if (dl->fd < 0) {
5638 pr_err("Invalid file descriptor for %s\n", dl->devname);
5639 return 0;
5640 }
5641
5642 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5643 return 0;
5644 if (member_sector_size != super->sector_size)
5645 return 0;
5646 return 1;
5647}
5648
f20c3968 5649static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5650 int fd, char *devname)
5651{
5652 struct intel_super *super = st->sb;
d23fe947 5653 struct imsm_super *mpb = super->anchor;
3960e579 5654 struct imsm_disk *_disk;
bf5a934a
DW		 5655 struct imsm_dev *dev;
5656 struct imsm_map *map;
3960e579 5657 struct dl *dl, *df;
4eb26970 5658 int slot;
bf5a934a 5659
949c47a0 5660 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5661 map = get_imsm_map(dev, MAP_0);
bf5a934a 5662
208933a7 5663 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5664 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5665 devname);
5666 return 1;
5667 }
5668
efb30e7f
DW		 5669 if (fd == -1) {
5670 /* we're doing autolayout so grab the pre-marked (in
5671 * validate_geometry) raid_disk
5672 */
5673 for (dl = super->disks; dl; dl = dl->next)
5674 if (dl->raiddisk == dk->raid_disk)
5675 break;
5676 } else {
5677 for (dl = super->disks; dl ; dl = dl->next)
5678 if (dl->major == dk->major &&
5679 dl->minor == dk->minor)
5680 break;
5681 }
d23fe947 5682
208933a7 5683 if (!dl) {
e7b84f9d 5684 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5685 return 1;
208933a7 5686 }
bf5a934a 5687
59632db9
MZ		 5688 if (mpb->num_disks == 0)
5689 if (!get_dev_sector_size(dl->fd, dl->devname,
5690 &super->sector_size))
5691 return 1;
5692
f2cc4f7d
AO		 5693 if (!drive_validate_sector_size(super, dl)) {
5694 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5695 return 1;
5696 }
5697
d23fe947
DW		 5698 /* add a pristine spare to the metadata */
5699 if (dl->index < 0) {
5700 dl->index = super->anchor->num_disks;
5701 super->anchor->num_disks++;
5702 }
4eb26970
DW		 5703 /* Check the device has not already been added */
5704 slot = get_imsm_disk_slot(map, dl->index);
5705 if (slot >= 0 &&
238c0a71 5706 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5707 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5708 devname);
5709 return 1;
5710 }
656b6b5a 5711 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5712 dl->disk.status = CONFIGURED_DISK;
d23fe947 5713
3960e579
DW		 5714 /* update size of 'missing' disks to be at least as large as the
5715 * largest acitve member (we only have dummy missing disks when
5716 * creating the first volume)
5717 */
5718 if (super->current_vol == 0) {
5719 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5720 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5721 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5722 _disk = __get_imsm_disk(mpb, df->index);
5723 *_disk = df->disk;
5724 }
5725 }
5726
5727 /* refresh unset/failed slots to point to valid 'missing' entries */
5728 for (df = super->missing; df; df = df->next)
5729 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5730 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5731
5732 if ((ord & IMSM_ORD_REBUILD) == 0)
5733 continue;
5734 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5735 if (is_gen_migration(dev)) {
238c0a71
AK		 5736 struct imsm_map *map2 = get_imsm_map(dev,
5737 MAP_1);
0a108d63 5738 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5739 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5740 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5741 slot2,
5742 MAP_1);
1ace8403
AK		 5743 if ((unsigned)df->index ==
5744 ord_to_idx(ord2))
5745 set_imsm_ord_tbl_ent(map2,
0a108d63 5746 slot2,
1ace8403
AK		 5747 df->index |
5748 IMSM_ORD_REBUILD);
5749 }
5750 }
3960e579
DW		 5751 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5752 break;
5753 }
5754
d23fe947
DW		 5755 /* if we are creating the first raid device update the family number */
5756 if (super->current_vol == 0) {
5757 __u32 sum;
5758 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5759
3960e579 5760 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5761 if (!_dev || !_disk) {
e7b84f9d 5762 pr_err("BUG mpb setup error\n");
791b666a
AW		 5763 return 1;
5764 }
d23fe947
DW		 5765 *_dev = *dev;
5766 *_disk = dl->disk;
148acb7b
DW		 5767 sum = random32();
5768 sum += __gen_imsm_checksum(mpb);
d23fe947 5769 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5770 mpb->orig_family_num = mpb->family_num;
e48aed3c 5771 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5772 }
ca0748fa 5773 super->current_disk = dl;
f20c3968 5774 return 0;
bf5a934a
DW		 5775}
5776
a8619d23
AK		 5777/* mark_spare()
5778 * Function marks disk as spare and restores disk serial
5779 * in case it was previously marked as failed by takeover operation
5780 * reruns:
5781 * -1 : critical error
5782 * 0 : disk is marked as spare but serial is not set
5783 * 1 : success
5784 */
5785int mark_spare(struct dl *disk)
5786{
5787 __u8 serial[MAX_RAID_SERIAL_LEN];
5788 int ret_val = -1;
5789
5790 if (!disk)
5791 return ret_val;
5792
5793 ret_val = 0;
6da53c0e 5794 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5795 /* Restore disk serial number, because takeover marks disk
5796 * as failed and adds to serial ':0' before it becomes
5797 * a spare disk.
5798 */
5799 serialcpy(disk->serial, serial);
5800 serialcpy(disk->disk.serial, serial);
5801 ret_val = 1;
5802 }
5803 disk->disk.status = SPARE_DISK;
5804 disk->index = -1;
5805
5806 return ret_val;
5807}
88654014 5808
12724c01
TM		 5809
5810static int write_super_imsm_spare(struct intel_super *super, struct dl *d);
5811
f20c3968 5812static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5813 int fd, char *devname,
5814 unsigned long long data_offset)
cdddbdbc 5815{
c2c087e6 5816 struct intel_super *super = st->sb;
c2c087e6
DW		 5817 struct dl *dd;
5818 unsigned long long size;
fa7bb6f8 5819 unsigned int member_sector_size;
f2f27e63 5820 __u32 id;
c2c087e6
DW		 5821 int rv;
5822 struct stat stb;
5823
88654014
LM		 5824 /* If we are on an RAID enabled platform check that the disk is
5825 * attached to the raid controller.
5826 * We do not need to test disks attachment for container based additions,
5827 * they shall be already tested when container was created/assembled.
88c32bb1 5828 */
d424212e 5829 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5830 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5831 if (rv != 0) {
5832 dprintf("capability: %p fd: %d ret: %d\n",
5833 super->orom, fd, rv);
5834 return 1;
88c32bb1
DW		 5835 }
5836
f20c3968
DW		 5837 if (super->current_vol >= 0)
5838 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5839
c2c087e6 5840 fstat(fd, &stb);
503975b9 5841 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5842 dd->major = major(stb.st_rdev);
5843 dd->minor = minor(stb.st_rdev);
503975b9 5844 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5845 dd->fd = fd;
689c9bf3 5846 dd->e = NULL;
1a64be56 5847 dd->action = DISK_ADD;
6da53c0e 5848 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5849 if (rv) {
e7b84f9d 5850 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5851 if (dd->devname)
5852 free(dd->devname);
949c47a0 5853 free(dd);
0030e8d6 5854 abort();
c2c087e6 5855 }
20bee0f8
PB		 5856 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5857 (super->hba->type == SYS_DEV_VMD))) {
5858 int i;
5859 char *devpath = diskfd_to_devpath(fd);
5860 char controller_path[PATH_MAX];
5861
5862 if (!devpath) {
5863 pr_err("failed to get devpath, aborting\n");
5864 if (dd->devname)
5865 free(dd->devname);
5866 free(dd);
5867 return 1;
5868 }
5869
5870 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5871 free(devpath);
5872
a8f3cfd5
MT		 5873 if (!imsm_is_nvme_supported(dd->fd, 1)) {
5874 if (dd->devname)
5875 free(dd->devname);
5876 free(dd);
5877 return 1;
5878 }
5879
20bee0f8
PB		 5880 if (devpath_to_vendor(controller_path) == 0x8086) {
5881 /*
5882 * If Intel's NVMe drive has serial ended with
5883 * "-A","-B","-1" or "-2" it means that this is "x8"
5884 * device (double drive on single PCIe card).
5885 * User should be warned about potential data loss.
5886 */
5887 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5888 /* Skip empty character at the end */
5889 if (dd->serial[i] == 0)
5890 continue;
5891
5892 if (((dd->serial[i] == 'A') ||
5893 (dd->serial[i] == 'B') ||
5894 (dd->serial[i] == '1') ||
5895 (dd->serial[i] == '2')) &&
5896 (dd->serial[i-1] == '-'))
5897 pr_err("\tThe action you are about to take may put your data at risk.\n"
5898 "\tPlease note that x8 devices may consist of two separate x4 devices "
5899 "located on a single PCIe port.\n"
5900 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5901 break;
5902 }
32716c51
PB		 5903 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5904 !imsm_orom_has_tpv_support(super->orom)) {
5905 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5906 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5907 free(dd->devname);
5908 free(dd);
5909 return 1;
20bee0f8
PB		 5910 }
5911 }
c2c087e6 5912
c2c087e6 5913 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5914 get_dev_sector_size(fd, NULL, &member_sector_size);
5915
5916 if (super->sector_size == 0) {
5917 /* this a first device, so sector_size is not set yet */
5918 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5919 }
5920
71e5411e 5921 /* clear migr_rec when adding disk to container */
85337573
AO		 5922 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5923 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5924 SEEK_SET) >= 0) {
466070ad 5925 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5926 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5927 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5928 perror("Write migr_rec failed");
5929 }
5930
c2c087e6 5931 size /= 512;
1f24f035 5932 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5933 set_total_blocks(&dd->disk, size);
5934 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5935 struct imsm_super *mpb = super->anchor;
5936 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5937 }
a8619d23 5938 mark_spare(dd);
c2c087e6 5939 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5940 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5941 else
b9f594fe 5942 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5943
5944 if (st->update_tail) {
1a64be56
LM		 5945 dd->next = super->disk_mgmt_list;
5946 super->disk_mgmt_list = dd;
43dad3d6 5947 } else {
12724c01
TM		 5948 /* this is called outside of mdmon
5949 * write initial spare metadata
5950 * mdmon will overwrite it.
5951 */
43dad3d6
DW		 5952 dd->next = super->disks;
5953 super->disks = dd;
12724c01 5954 write_super_imsm_spare(super, dd);
43dad3d6 5955 }
f20c3968
DW		 5956
5957 return 0;
cdddbdbc
DW		 5958}
5959
1a64be56
LM		 5960static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5961{
5962 struct intel_super *super = st->sb;
5963 struct dl *dd;
5964
5965 /* remove from super works only in mdmon - for communication
5966 * manager - monitor. Check if communication memory buffer
5967 * is prepared.
5968 */
5969 if (!st->update_tail) {
1ade5cc1 5970 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5971 return 1;
5972 }
503975b9 5973 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5974 dd->major = dk->major;
5975 dd->minor = dk->minor;
1a64be56 5976 dd->fd = -1;
a8619d23 5977 mark_spare(dd);
1a64be56
LM		 5978 dd->action = DISK_REMOVE;
5979
5980 dd->next = super->disk_mgmt_list;
5981 super->disk_mgmt_list = dd;
5982
1a64be56
LM		 5983 return 0;
5984}
5985
f796af5d
DW		 5986static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5987
5988static union {
f36a9ecd 5989 char buf[MAX_SECTOR_SIZE];
f796af5d 5990 struct imsm_super anchor;
f36a9ecd 5991} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5992
12724c01
TM		 5993
5994static int write_super_imsm_spare(struct intel_super *super, struct dl *d)
d23fe947 5995{
d23fe947 5996 struct imsm_super *mpb = super->anchor;
f796af5d 5997 struct imsm_super *spare = &spare_record.anchor;
d23fe947 5998 __u32 sum;
12724c01
TM		 5999
6000 if (d->index != -1)
6001 return 1;
d23fe947 6002
68641cdb
JS		 6003 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6004 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6005 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6006 spare->num_disks = 1;
6007 spare->num_raid_devs = 0;
6008 spare->cache_size = mpb->cache_size;
6009 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6010
6011 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6012 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947 6013
12724c01
TM		 6014 spare->disk[0] = d->disk;
6015 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6016 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6017
6018 if (super->sector_size == 4096)
6019 convert_to_4k_imsm_disk(&spare->disk[0]);
d23fe947 6020
12724c01
TM		 6021 sum = __gen_imsm_checksum(spare);
6022 spare->family_num = __cpu_to_le32(sum);
6023 spare->orig_family_num = 0;
6024 sum = __gen_imsm_checksum(spare);
6025 spare->check_sum = __cpu_to_le32(sum);
027c374f 6026
12724c01
TM		 6027 if (store_imsm_mpb(d->fd, spare)) {
6028 pr_err("failed for device %d:%d %s\n",
6029 d->major, d->minor, strerror(errno));
6030 return 1;
6031 }
6032
6033 return 0;
6034}
6035/* spare records have their own family number and do not have any defined raid
6036 * devices
6037 */
6038static int write_super_imsm_spares(struct intel_super *super, int doclose)
6039{
6040 struct dl *d;
f36a9ecd 6041
12724c01
TM		 6042 for (d = super->disks; d; d = d->next) {
6043 if (d->index != -1)
6044 continue;
d23fe947 6045
12724c01 6046 if (write_super_imsm_spare(super, d))
e74255d9 6047 return 1;
12724c01 6048
d23fe947
DW		 6049 if (doclose) {
6050 close(d->fd);
6051 d->fd = -1;
6052 }
6053 }
6054
e74255d9 6055 return 0;
d23fe947
DW		 6056}
6057
36988a3d 6058static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6059{
36988a3d 6060 struct intel_super *super = st->sb;
f36a9ecd 6061 unsigned int sector_size = super->sector_size;
949c47a0 6062 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6063 struct dl *d;
6064 __u32 generation;
6065 __u32 sum;
d23fe947 6066 int spares = 0;
949c47a0 6067 int i;
a48ac0a8 6068 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6069 int num_disks = 0;
146c6260 6070 int clear_migration_record = 1;
bbab0940 6071 __u32 bbm_log_size;
cdddbdbc 6072
c2c087e6
DW		 6073 /* 'generation' is incremented everytime the metadata is written */
6074 generation = __le32_to_cpu(mpb->generation_num);
6075 generation++;
6076 mpb->generation_num = __cpu_to_le32(generation);
6077
148acb7b
DW		 6078 /* fix up cases where previous mdadm releases failed to set
6079 * orig_family_num
6080 */
6081 if (mpb->orig_family_num == 0)
6082 mpb->orig_family_num = mpb->family_num;
6083
d23fe947 6084 for (d = super->disks; d; d = d->next) {
8796fdc4 6085 if (d->index == -1)
d23fe947 6086 spares++;
36988a3d 6087 else {
d23fe947 6088 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6089 num_disks++;
6090 }
d23fe947 6091 }
36988a3d 6092 for (d = super->missing; d; d = d->next) {
47ee5a45 6093 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6094 num_disks++;
6095 }
6096 mpb->num_disks = num_disks;
6097 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6098
949c47a0
DW		 6099 for (i = 0; i < mpb->num_raid_devs; i++) {
6100 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6101 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6102 if (dev && dev2) {
6103 imsm_copy_dev(dev, dev2);
6104 mpb_size += sizeof_imsm_dev(dev, 0);
6105 }
146c6260
AK		 6106 if (is_gen_migration(dev2))
6107 clear_migration_record = 0;
949c47a0 6108 }
bbab0940
TM		 6109
6110 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6111
6112 if (bbm_log_size) {
6113 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6114 mpb->attributes |= MPB_ATTRIB_BBM;
6115 } else
6116 mpb->attributes &= ~MPB_ATTRIB_BBM;
6117
6118 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6119 mpb_size += bbm_log_size;
a48ac0a8 6120 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6121
bbab0940
TM		 6122#ifdef DEBUG
6123 assert(super->len == 0 || mpb_size <= super->len);
6124#endif
6125
c2c087e6 6126 /* recalculate checksum */
949c47a0 6127 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6128 mpb->check_sum = __cpu_to_le32(sum);
6129
51d83f5d
AK		 6130 if (super->clean_migration_record_by_mdmon) {
6131 clear_migration_record = 1;
6132 super->clean_migration_record_by_mdmon = 0;
6133 }
146c6260 6134 if (clear_migration_record)
de44e46f 6135 memset(super->migr_rec_buf, 0,
85337573 6136 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6137
f36a9ecd
PB		 6138 if (sector_size == 4096)
6139 convert_to_4k(super);
6140
d23fe947 6141 /* write the mpb for disks that compose raid devices */
c2c087e6 6142 for (d = super->disks; d ; d = d->next) {
86c54047 6143 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6144 continue;
30602f53 6145
146c6260
AK		 6146 if (clear_migration_record) {
6147 unsigned long long dsize;
6148
6149 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6150 if (lseek64(d->fd, dsize - sector_size,
6151 SEEK_SET) >= 0) {
466070ad
PB		 6152 if ((unsigned int)write(d->fd,
6153 super->migr_rec_buf,
de44e46f
PB		 6154 MIGR_REC_BUF_SECTORS*sector_size) !=
6155 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6156 perror("Write migr_rec failed");
146c6260
AK		 6157 }
6158 }
51d83f5d
AK		 6159
6160 if (store_imsm_mpb(d->fd, mpb))
6161 fprintf(stderr,
1ade5cc1
N		 6162 "failed for device %d:%d (fd: %d)%s\n",
6163 d->major, d->minor,
51d83f5d
AK		 6164 d->fd, strerror(errno));
6165
c2c087e6
DW		 6166 if (doclose) {
6167 close(d->fd);
6168 d->fd = -1;
6169 }
6170 }
6171
d23fe947
DW		 6172 if (spares)
6173 return write_super_imsm_spares(super, doclose);
6174
e74255d9 6175 return 0;
c2c087e6
DW		 6176}
6177
9b1fb677 6178static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6179{
6180 size_t len;
6181 struct imsm_update_create_array *u;
6182 struct intel_super *super = st->sb;
9b1fb677 6183 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6184 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6185 struct disk_info *inf;
6186 struct imsm_disk *disk;
6187 int i;
43dad3d6 6188
54c2c1ea
DW		 6189 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6190 sizeof(*inf) * map->num_members;
503975b9 6191 u = xmalloc(len);
43dad3d6 6192 u->type = update_create_array;
9b1fb677 6193 u->dev_idx = dev_idx;
43dad3d6 6194 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6195 inf = get_disk_info(u);
6196 for (i = 0; i < map->num_members; i++) {
238c0a71 6197 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6198
54c2c1ea 6199 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6200 if (!disk)
6201 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6202 serialcpy(inf[i].serial, disk->serial);
6203 }
43dad3d6
DW		 6204 append_metadata_update(st, u, len);
6205
6206 return 0;
6207}
6208
1a64be56 6209static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6210{
6211 struct intel_super *super = st->sb;
6212 size_t len;
1a64be56 6213 struct imsm_update_add_remove_disk *u;
43dad3d6 6214
1a64be56 6215 if (!super->disk_mgmt_list)
43dad3d6
DW		 6216 return 0;
6217
6218 len = sizeof(*u);
503975b9 6219 u = xmalloc(len);
1a64be56 6220 u->type = update_add_remove_disk;
43dad3d6
DW		 6221 append_metadata_update(st, u, len);
6222
6223 return 0;
6224}
2432ce9b
AP		 6225
6226__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6227
e397cefe
AP		 6228static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6229{
6230 struct ppl_header *ppl_hdr = buf;
6231 int ret;
6232
6233 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6234
6235 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6236 ret = -errno;
6237 perror("Failed to seek to PPL header location");
6238 return ret;
6239 }
6240
6241 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6242 ret = -errno;
6243 perror("Write PPL header failed");
6244 return ret;
6245 }
6246
6247 fsync(fd);
6248
6249 return 0;
6250}
6251
2432ce9b
AP		 6252static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6253{
6254 struct intel_super *super = st->sb;
6255 void *buf;
6256 struct ppl_header *ppl_hdr;
6257 int ret;
6258
b2514242
PB		 6259 /* first clear entire ppl space */
6260 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6261 if (ret)
6262 return ret;
6263
6264 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6265 if (ret) {
6266 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6267 return -ret;
2432ce9b
AP		 6268 }
6269
6270 memset(buf, 0, PPL_HEADER_SIZE);
6271 ppl_hdr = buf;
6272 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6273 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6274
6275 if (info->mismatch_cnt) {
6276 /*
6277 * We are overwriting an invalid ppl. Make one entry with wrong
6278 * checksum to prevent the kernel from skipping resync.
6279 */
6280 ppl_hdr->entries_count = __cpu_to_le32(1);
6281 ppl_hdr->entries[0].checksum = ~0;
6282 }
6283
e397cefe 6284 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6285
6286 free(buf);
6287 return ret;
6288}
6289
e397cefe
AP		 6290static int is_rebuilding(struct imsm_dev *dev);
6291
2432ce9b
AP		 6292static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6293 struct mdinfo *disk)
6294{
6295 struct intel_super *super = st->sb;
6296 struct dl *d;
e397cefe 6297 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6298 int ret = 0;
e397cefe 6299 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6300 __u32 crc;
6301 struct imsm_dev *dev;
2432ce9b 6302 __u32 idx;
44b6b876
PB		 6303 unsigned int i;
6304 unsigned long long ppl_offset = 0;
6305 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6306
6307 if (disk->disk.raid_disk < 0)
6308 return 0;
6309
2432ce9b 6310 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6311 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6312 d = get_imsm_dl_disk(super, idx);
6313
6314 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6315 return 0;
6316
6317 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6318 pr_err("Failed to allocate PPL header buffer\n");
6319 return -1;
6320 }
6321 buf = buf_orig;
2432ce9b 6322
44b6b876
PB		 6323 ret = 1;
6324 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6325 void *tmp;
6326
44b6b876 6327 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6328
44b6b876
PB		 6329 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6330 SEEK_SET) < 0) {
6331 perror("Failed to seek to PPL header location");
6332 ret = -1;
e397cefe 6333 break;
44b6b876 6334 }
2432ce9b 6335
44b6b876
PB		 6336 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6337 perror("Read PPL header failed");
6338 ret = -1;
e397cefe 6339 break;
44b6b876 6340 }
2432ce9b 6341
44b6b876 6342 ppl_hdr = buf;
2432ce9b 6343
44b6b876
PB		 6344 crc = __le32_to_cpu(ppl_hdr->checksum);
6345 ppl_hdr->checksum = 0;
6346
6347 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6348 dprintf("Wrong PPL header checksum on %s\n",
6349 d->devname);
e397cefe 6350 break;
44b6b876
PB		 6351 }
6352
6353 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6354 /* previous was newest, it was already checked */
e397cefe 6355 break;
44b6b876
PB		 6356 }
6357
6358 if ((__le32_to_cpu(ppl_hdr->signature) !=
6359 super->anchor->orig_family_num)) {
6360 dprintf("Wrong PPL header signature on %s\n",
6361 d->devname);
6362 ret = 1;
e397cefe 6363 break;
44b6b876
PB		 6364 }
6365
6366 ret = 0;
6367 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6368
44b6b876
PB		 6369 ppl_offset += PPL_HEADER_SIZE;
6370 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6371 ppl_offset +=
6372 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6373
6374 if (!buf_prev)
6375 buf_prev = buf + PPL_HEADER_SIZE;
6376 tmp = buf_prev;
6377 buf_prev = buf;
6378 buf = tmp;
2432ce9b
AP		 6379 }
6380
e397cefe
AP		 6381 if (buf_prev) {
6382 buf = buf_prev;
6383 ppl_hdr = buf_prev;
6384 }
2432ce9b 6385
54148aba
PB		 6386 /*
6387 * Update metadata to use mutliple PPLs area (1MB).
6388 * This is done once for all RAID members
6389 */
6390 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6391 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6392 char subarray[20];
6393 struct mdinfo *member_dev;
6394
6395 sprintf(subarray, "%d", info->container_member);
6396
6397 if (mdmon_running(st->container_devnm))
6398 st->update_tail = &st->updates;
6399
6400 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6401 pr_err("Failed to update subarray %s\n",
6402 subarray);
6403 } else {
6404 if (st->update_tail)
6405 flush_metadata_updates(st);
6406 else
6407 st->ss->sync_metadata(st);
6408 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6409 for (member_dev = info->devs; member_dev;
6410 member_dev = member_dev->next)
6411 member_dev->ppl_size =
6412 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6413 }
6414 }
6415
b23d0750 6416 if (ret == 1) {
2fc0fc63
AP		 6417 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6418
50b9c10d
PB		 6419 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6420 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6421 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6422 (is_rebuilding(dev) &&
2ec9d182
AP		 6423 dev->vol.curr_migr_unit == 0 &&
6424 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6425 ret = st->ss->write_init_ppl(st, info, d->fd);
6426 else
6427 info->mismatch_cnt++;
e397cefe
AP		 6428 } else if (ret == 0 &&
6429 ppl_hdr->entries_count == 0 &&
6430 is_rebuilding(dev) &&
6431 info->resync_start == 0) {
6432 /*
6433 * The header has no entries - add a single empty entry and
6434 * rewrite the header to prevent the kernel from going into
6435 * resync after an interrupted rebuild.
6436 */
6437 ppl_hdr->entries_count = __cpu_to_le32(1);
6438 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6439 }
2432ce9b 6440
e397cefe
AP		 6441 free(buf_orig);
6442
2432ce9b
AP		 6443 return ret;
6444}
6445
2432ce9b
AP		 6446static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6447{
6448 struct intel_super *super = st->sb;
6449 struct dl *d;
6450 int ret = 0;
6451
6452 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6453 info->array.level != 5)
6454 return 0;
6455
6456 for (d = super->disks; d ; d = d->next) {
6457 if (d->index < 0 || is_failed(&d->disk))
6458 continue;
6459
6460 ret = st->ss->write_init_ppl(st, info, d->fd);
6461 if (ret)
6462 break;
6463 }
6464
6465 return ret;
6466}
43dad3d6 6467
c2c087e6
DW		 6468static int write_init_super_imsm(struct supertype *st)
6469{
9b1fb677
DW		 6470 struct intel_super *super = st->sb;
6471 int current_vol = super->current_vol;
2432ce9b
AP		 6472 int rv = 0;
6473 struct mdinfo info;
6474
6475 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6476
6477 /* we are done with current_vol reset it to point st at the container */
6478 super->current_vol = -1;
6479
8273f55e 6480 if (st->update_tail) {
43dad3d6
DW		 6481 /* queue the recently created array / added disk
6482 * as a metadata update */
8273f55e 6483
43dad3d6 6484 /* determine if we are creating a volume or adding a disk */
9b1fb677 6485 if (current_vol < 0) {
1a64be56
LM		 6486 /* in the mgmt (add/remove) disk case we are running
6487 * in mdmon context, so don't close fd's
43dad3d6 6488 */
2432ce9b
AP		 6489 rv = mgmt_disk(st);
6490 } else {
6491 rv = write_init_ppl_imsm_all(st, &info);
6492 if (!rv)
6493 rv = create_array(st, current_vol);
6494 }
d682f344
N		 6495 } else {
6496 struct dl *d;
6497 for (d = super->disks; d; d = d->next)
ba728be7 6498 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6499 if (current_vol >= 0)
6500 rv = write_init_ppl_imsm_all(st, &info);
6501 if (!rv)
6502 rv = write_super_imsm(st, 1);
d682f344 6503 }
2432ce9b
AP		 6504
6505 return rv;
cdddbdbc
DW		 6506}
6507
e683ca88 6508static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6509{
e683ca88
DW		 6510 struct intel_super *super = st->sb;
6511 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6512
e683ca88 6513 if (!mpb)
ad97895e
DW		 6514 return 1;
6515
f36a9ecd
PB		 6516 if (super->sector_size == 4096)
6517 convert_to_4k(super);
e683ca88 6518 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6519}
6520
cdddbdbc
DW		 6521static int validate_geometry_imsm_container(struct supertype *st, int level,
6522 int layout, int raiddisks, int chunk,
af4348dd
N		 6523 unsigned long long size,
6524 unsigned long long data_offset,
6525 char *dev,
2c514b71
NB		 6526 unsigned long long *freesize,
6527 int verbose)
cdddbdbc 6528{
c2c087e6
DW		 6529 int fd;
6530 unsigned long long ldsize;
594dc1b8 6531 struct intel_super *super;
f2f5c343 6532 int rv = 0;
cdddbdbc 6533
c2c087e6
DW		 6534 if (level != LEVEL_CONTAINER)
6535 return 0;
6536 if (!dev)
6537 return 1;
6538
6539 fd = open(dev, O_RDONLY|O_EXCL, 0);
6540 if (fd < 0) {
ba728be7 6541 if (verbose > 0)
e7b84f9d 6542 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6543 dev, strerror(errno));
c2c087e6
DW		 6544 return 0;
6545 }
6546 if (!get_dev_size(fd, dev, &ldsize)) {
6547 close(fd);
6548 return 0;
6549 }
f2f5c343
LM		 6550
6551 /* capabilities retrieve could be possible
6552 * note that there is no fd for the disks in array.
6553 */
6554 super = alloc_super();
8d67477f
TM		 6555 if (!super) {
6556 close(fd);
6557 return 0;
6558 }
fa7bb6f8
PB		 6559 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6560 close(fd);
6561 free_imsm(super);
6562 return 0;
6563 }
6564
ba728be7 6565 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6566 if (rv != 0) {
6567#if DEBUG
6568 char str[256];
6569 fd2devname(fd, str);
1ade5cc1 6570 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6571 fd, str, super->orom, rv, raiddisks);
6572#endif
6573 /* no orom/efi or non-intel hba of the disk */
6574 close(fd);
6575 free_imsm(super);
6576 return 0;
6577 }
c2c087e6 6578 close(fd);
9126b9a8
CA		 6579 if (super->orom) {
6580 if (raiddisks > super->orom->tds) {
6581 if (verbose)
7a862a02 6582 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6583 raiddisks, super->orom->tds);
6584 free_imsm(super);
6585 return 0;
6586 }
6587 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6588 (ldsize >> 9) >> 32 > 0) {
6589 if (verbose)
e7b84f9d 6590 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6591 free_imsm(super);
6592 return 0;
6593 }
f2f5c343 6594 }
c2c087e6 6595
af4348dd 6596 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6597 free_imsm(super);
c2c087e6
DW		 6598
6599 return 1;
cdddbdbc
DW		 6600}
6601
0dcecb2e
DW		 6602static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6603{
6604 const unsigned long long base_start = e[*idx].start;
6605 unsigned long long end = base_start + e[*idx].size;
6606 int i;
6607
6608 if (base_start == end)
6609 return 0;
6610
6611 *idx = *idx + 1;
6612 for (i = *idx; i < num_extents; i++) {
6613 /* extend overlapping extents */
6614 if (e[i].start >= base_start &&
6615 e[i].start <= end) {
6616 if (e[i].size == 0)
6617 return 0;
6618 if (e[i].start + e[i].size > end)
6619 end = e[i].start + e[i].size;
6620 } else if (e[i].start > end) {
6621 *idx = i;
6622 break;
6623 }
6624 }
6625
6626 return end - base_start;
6627}
6628
6629static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6630{
6631 /* build a composite disk with all known extents and generate a new
6632 * 'maxsize' given the "all disks in an array must share a common start
6633 * offset" constraint
6634 */
503975b9 6635 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6636 struct dl *dl;
6637 int i, j;
6638 int start_extent;
6639 unsigned long long pos;
b9d77223 6640 unsigned long long start = 0;
0dcecb2e
DW		 6641 unsigned long long maxsize;
6642 unsigned long reserve;
6643
0dcecb2e
DW		 6644 /* coalesce and sort all extents. also, check to see if we need to
6645 * reserve space between member arrays
6646 */
6647 j = 0;
6648 for (dl = super->disks; dl; dl = dl->next) {
6649 if (!dl->e)
6650 continue;
6651 for (i = 0; i < dl->extent_cnt; i++)
6652 e[j++] = dl->e[i];
6653 }
6654 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6655
6656 /* merge extents */
6657 i = 0;
6658 j = 0;
6659 while (i < sum_extents) {
6660 e[j].start = e[i].start;
6661 e[j].size = find_size(e, &i, sum_extents);
6662 j++;
6663 if (e[j-1].size == 0)
6664 break;
6665 }
6666
6667 pos = 0;
6668 maxsize = 0;
6669 start_extent = 0;
6670 i = 0;
6671 do {
6672 unsigned long long esize;
6673
6674 esize = e[i].start - pos;
6675 if (esize >= maxsize) {
6676 maxsize = esize;
6677 start = pos;
6678 start_extent = i;
6679 }
6680 pos = e[i].start + e[i].size;
6681 i++;
6682 } while (e[i-1].size);
6683 free(e);
6684
a7dd165b
DW		 6685 if (maxsize == 0)
6686 return 0;
6687
6688 /* FIXME assumes volume at offset 0 is the first volume in a
6689 * container
6690 */
0dcecb2e
DW		 6691 if (start_extent > 0)
6692 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6693 else
6694 reserve = 0;
6695
6696 if (maxsize < reserve)
a7dd165b 6697 return 0;
0dcecb2e 6698
5551b113 6699 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6700 if (start + reserve > super->create_offset)
a7dd165b 6701 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6702 super->create_offset = start + reserve;
6703
6704 return maxsize - reserve;
6705}
6706
88c32bb1
DW		 6707static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6708{
6709 if (level < 0 || level == 6 || level == 4)
6710 return 0;
6711
6712 /* if we have an orom prevent invalid raid levels */
6713 if (orom)
6714 switch (level) {
6715 case 0: return imsm_orom_has_raid0(orom);
6716 case 1:
6717 if (raiddisks > 2)
6718 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6719 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6720 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6721 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6722 }
6723 else
6724 return 1; /* not on an Intel RAID platform so anything goes */
6725
6726 return 0;
6727}
6728
ca9de185
LM		 6729static int
6730active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6731 int dpa, int verbose)
6732{
6733 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6734 struct mdstat_ent *memb;
ca9de185
LM		 6735 int count = 0;
6736 int num = 0;
594dc1b8 6737 struct md_list *dv;
ca9de185
LM		 6738 int found;
6739
6740 for (memb = mdstat ; memb ; memb = memb->next) {
6741 if (memb->metadata_version &&
fc54fe7a 6742 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6743 (strcmp(&memb->metadata_version[9], name) == 0) &&
6744 !is_subarray(memb->metadata_version+9) &&
6745 memb->members) {
6746 struct dev_member *dev = memb->members;
6747 int fd = -1;
6748 while(dev && (fd < 0)) {
503975b9
N		 6749 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6750 num = sprintf(path, "%s%s", "/dev/", dev->name);
6751 if (num > 0)
6752 fd = open(path, O_RDONLY, 0);
089f9d79 6753 if (num <= 0 || fd < 0) {
676e87a8 6754 pr_vrb("Cannot open %s: %s\n",
503975b9 6755 dev->name, strerror(errno));
ca9de185 6756 }
503975b9 6757 free(path);
ca9de185
LM		 6758 dev = dev->next;
6759 }
6760 found = 0;
089f9d79 6761 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6762 struct mdstat_ent *vol;
6763 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6764 if (vol->active > 0 &&
ca9de185 6765 vol->metadata_version &&
9581efb1 6766 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6767 found++;
6768 count++;
6769 }
6770 }
6771 if (*devlist && (found < dpa)) {
503975b9 6772 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6773 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6774 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6775 dv->found = found;
6776 dv->used = 0;
6777 dv->next = *devlist;
6778 *devlist = dv;
ca9de185
LM		 6779 }
6780 }
6781 if (fd >= 0)
6782 close(fd);
6783 }
6784 }
6785 free_mdstat(mdstat);
6786 return count;
6787}
6788
6789#ifdef DEBUG_LOOP
6790static struct md_list*
6791get_loop_devices(void)
6792{
6793 int i;
6794 struct md_list *devlist = NULL;
594dc1b8 6795 struct md_list *dv;
ca9de185
LM		 6796
6797 for(i = 0; i < 12; i++) {
503975b9
N		 6798 dv = xcalloc(1, sizeof(*dv));
6799 dv->devname = xmalloc(40);
ca9de185
LM		 6800 sprintf(dv->devname, "/dev/loop%d", i);
6801 dv->next = devlist;
6802 devlist = dv;
6803 }
6804 return devlist;
6805}
6806#endif
6807
6808static struct md_list*
6809get_devices(const char *hba_path)
6810{
6811 struct md_list *devlist = NULL;
594dc1b8 6812 struct md_list *dv;
ca9de185
LM		 6813 struct dirent *ent;
6814 DIR *dir;
6815 int err = 0;
6816
6817#if DEBUG_LOOP
6818 devlist = get_loop_devices();
6819 return devlist;
6820#endif
6821 /* scroll through /sys/dev/block looking for devices attached to
6822 * this hba
6823 */
6824 dir = opendir("/sys/dev/block");
6825 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6826 int fd;
6827 char buf[1024];
6828 int major, minor;
6829 char *path = NULL;
6830 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6831 continue;
6832 path = devt_to_devpath(makedev(major, minor));
6833 if (!path)
6834 continue;
6835 if (!path_attached_to_hba(path, hba_path)) {
6836 free(path);
6837 path = NULL;
6838 continue;
6839 }
6840 free(path);
6841 path = NULL;
6842 fd = dev_open(ent->d_name, O_RDONLY);
6843 if (fd >= 0) {
6844 fd2devname(fd, buf);
6845 close(fd);
6846 } else {
e7b84f9d 6847 pr_err("cannot open device: %s\n",
ca9de185
LM		 6848 ent->d_name);
6849 continue;
6850 }
6851
503975b9
N		 6852 dv = xcalloc(1, sizeof(*dv));
6853 dv->devname = xstrdup(buf);
ca9de185
LM		 6854 dv->next = devlist;
6855 devlist = dv;
6856 }
6857 if (err) {
6858 while(devlist) {
6859 dv = devlist;
6860 devlist = devlist->next;
6861 free(dv->devname);
6862 free(dv);
6863 }
6864 }
562aa102 6865 closedir(dir);
ca9de185
LM		 6866 return devlist;
6867}
6868
6869static int
6870count_volumes_list(struct md_list *devlist, char *homehost,
6871 int verbose, int *found)
6872{
6873 struct md_list *tmpdev;
6874 int count = 0;
594dc1b8 6875 struct supertype *st;
ca9de185
LM		 6876
6877 /* first walk the list of devices to find a consistent set
6878 * that match the criterea, if that is possible.
6879 * We flag the ones we like with 'used'.
6880 */
6881 *found = 0;
6882 st = match_metadata_desc_imsm("imsm");
6883 if (st == NULL) {
676e87a8 6884 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6885 return 0;
6886 }
6887
6888 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6889 char *devname = tmpdev->devname;
0a6bff09 6890 dev_t rdev;
ca9de185
LM		 6891 struct supertype *tst;
6892 int dfd;
6893 if (tmpdev->used > 1)
6894 continue;
6895 tst = dup_super(st);
6896 if (tst == NULL) {
676e87a8 6897 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6898 goto err_1;
6899 }
6900 tmpdev->container = 0;
6901 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6902 if (dfd < 0) {
1ade5cc1 6903 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6904 devname, strerror(errno));
6905 tmpdev->used = 2;
0a6bff09 6906 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6907 tmpdev->used = 2;
6908 } else if (must_be_container(dfd)) {
6909 struct supertype *cst;
6910 cst = super_by_fd(dfd, NULL);
6911 if (cst == NULL) {
1ade5cc1 6912 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6913 devname);
6914 tmpdev->used = 2;
6915 } else if (tst->ss != st->ss) {
1ade5cc1 6916 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6917 devname);
6918 tmpdev->used = 2;
6919 } else if (!tst->ss->load_container ||
6920 tst->ss->load_container(tst, dfd, NULL))
6921 tmpdev->used = 2;
6922 else {
6923 tmpdev->container = 1;
6924 }
6925 if (cst)
6926 cst->ss->free_super(cst);
6927 } else {
0a6bff09 6928 tmpdev->st_rdev = rdev;
ca9de185 6929 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6930 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6931 devname);
6932 tmpdev->used = 2;
6933 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6934 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6935 tst->ss->name, devname);
6936 tmpdev->used = 2;
6937 }
6938 }
6939 if (dfd >= 0)
6940 close(dfd);
6941 if (tmpdev->used == 2 || tmpdev->used == 4) {
6942 /* Ignore unrecognised devices during auto-assembly */
6943 goto loop;
6944 }
6945 else {
6946 struct mdinfo info;
6947 tst->ss->getinfo_super(tst, &info, NULL);
6948
6949 if (st->minor_version == -1)
6950 st->minor_version = tst->minor_version;
6951
6952 if (memcmp(info.uuid, uuid_zero,
6953 sizeof(int[4])) == 0) {
6954 /* this is a floating spare. It cannot define
6955 * an array unless there are no more arrays of
6956 * this type to be found. It can be included
6957 * in an array of this type though.
6958 */
6959 tmpdev->used = 3;
6960 goto loop;
6961 }
6962
6963 if (st->ss != tst->ss ||
6964 st->minor_version != tst->minor_version ||
6965 st->ss->compare_super(st, tst) != 0) {
6966 /* Some mismatch. If exactly one array matches this host,
6967 * we can resolve on that one.
6968 * Or, if we are auto assembling, we just ignore the second
6969 * for now.
6970 */
1ade5cc1 6971 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6972 devname);
6973 goto loop;
6974 }
6975 tmpdev->used = 1;
6976 *found = 1;
6977 dprintf("found: devname: %s\n", devname);
6978 }
6979 loop:
6980 if (tst)
6981 tst->ss->free_super(tst);
6982 }
6983 if (*found != 0) {
6984 int err;
6985 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6986 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6987 for (iter = head; iter; iter = iter->next) {
6988 dprintf("content->text_version: %s vol\n",
6989 iter->text_version);
6990 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6991 /* do not assemble arrays with unsupported
6992 configurations */
1ade5cc1 6993 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6994 iter->text_version);
6995 } else
6996 count++;
6997 }
6998 sysfs_free(head);
6999
7000 } else {
1ade5cc1 7001 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7002 err, st->sb);
7003 }
7004 } else {
1ade5cc1 7005 dprintf("no more devices to examine\n");
ca9de185
LM		 7006 }
7007
7008 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7009 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7010 if (count) {
7011 if (count < tmpdev->found)
7012 count = 0;
7013 else
7014 count -= tmpdev->found;
7015 }
7016 }
7017 if (tmpdev->used == 1)
7018 tmpdev->used = 4;
7019 }
7020 err_1:
7021 if (st)
7022 st->ss->free_super(st);
7023 return count;
7024}
7025
d3c11416
AO		 7026static int __count_volumes(char *hba_path, int dpa, int verbose,
7027 int cmp_hba_path)
ca9de185 7028{
72a45777 7029 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7030 int count = 0;
72a45777
PB		 7031 const struct orom_entry *entry;
7032 struct devid_list *dv, *devid_list;
ca9de185 7033
d3c11416 7034 if (!hba_path)
ca9de185
LM		 7035 return 0;
7036
72a45777 7037 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7038 if (strstr(idev->path, hba_path))
7039 break;
72a45777
PB		 7040 }
7041
7042 if (!idev || !idev->dev_id)
ca9de185 7043 return 0;
72a45777
PB		 7044
7045 entry = get_orom_entry_by_device_id(idev->dev_id);
7046
7047 if (!entry || !entry->devid_list)
7048 return 0;
7049
7050 devid_list = entry->devid_list;
7051 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7052 struct md_list *devlist;
d3c11416
AO		 7053 struct sys_dev *device = NULL;
7054 char *hpath;
72a45777
PB		 7055 int found = 0;
7056
d3c11416
AO		 7057 if (cmp_hba_path)
7058 device = device_by_id_and_path(dv->devid, hba_path);
7059 else
7060 device = device_by_id(dv->devid);
7061
72a45777 7062 if (device)
d3c11416 7063 hpath = device->path;
72a45777
PB		 7064 else
7065 return 0;
7066
d3c11416 7067 devlist = get_devices(hpath);
72a45777
PB		 7068 /* if no intel devices return zero volumes */
7069 if (devlist == NULL)
7070 return 0;
7071
d3c11416
AO		 7072 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7073 verbose);
7074 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7075 if (devlist == NULL)
7076 return 0;
7077 do {
7078 found = 0;
7079 count += count_volumes_list(devlist,
7080 NULL,
7081 verbose,
7082 &found);
7083 dprintf("found %d count: %d\n", found, count);
7084 } while (found);
7085
d3c11416 7086 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7087
7088 while (devlist) {
7089 struct md_list *dv = devlist;
7090 devlist = devlist->next;
7091 free(dv->devname);
7092 free(dv);
7093 }
ca9de185
LM		 7094 }
7095 return count;
7096}
7097
d3c11416
AO		 7098static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7099{
7100 if (!hba)
7101 return 0;
7102 if (hba->type == SYS_DEV_VMD) {
7103 struct sys_dev *dev;
7104 int count = 0;
7105
7106 for (dev = find_intel_devices(); dev; dev = dev->next) {
7107 if (dev->type == SYS_DEV_VMD)
7108 count += __count_volumes(dev->path, dpa,
7109 verbose, 1);
7110 }
7111 return count;
7112 }
7113 return __count_volumes(hba->path, dpa, verbose, 0);
7114}
7115
cd9d1ac7
DW		 7116static int imsm_default_chunk(const struct imsm_orom *orom)
7117{
7118 /* up to 512 if the plaform supports it, otherwise the platform max.
7119 * 128 if no platform detected
7120 */
7121 int fs = max(7, orom ? fls(orom->sss) : 0);
7122
7123 return min(512, (1 << fs));
7124}
73408129 7125
6592ce37
DW		 7126static int
7127validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7128 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7129{
660260d0
DW		 7130 /* check/set platform and metadata limits/defaults */
7131 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7132 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7133 super->orom->dpa);
73408129
LM		 7134 return 0;
7135 }
7136
5d500228 7137 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7138 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7139 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7140 level, raiddisks, raiddisks > 1 ? "s" : "");
7141 return 0;
7142 }
cd9d1ac7 7143
7ccc4cc4 7144 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7145 *chunk = imsm_default_chunk(super->orom);
7146
7ccc4cc4 7147 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7148 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7149 return 0;
6592ce37 7150 }
cd9d1ac7 7151
6592ce37
DW		 7152 if (layout != imsm_level_to_layout(level)) {
7153 if (level == 5)
676e87a8 7154 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7155 else if (level == 10)
676e87a8 7156 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7157 else
676e87a8 7158 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7159 layout, level);
7160 return 0;
7161 }
2cc699af 7162
7ccc4cc4 7163 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7164 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7165 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7166 return 0;
7167 }
614902f6 7168
6592ce37
DW		 7169 return 1;
7170}
7171
1011e834 7172/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7173 * FIX ME add ahci details
7174 */
8b353278 7175static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7176 int layout, int raiddisks, int *chunk,
af4348dd
N		 7177 unsigned long long size,
7178 unsigned long long data_offset,
7179 char *dev,
2c514b71
NB		 7180 unsigned long long *freesize,
7181 int verbose)
cdddbdbc 7182{
9e04ac1c 7183 dev_t rdev;
c2c087e6 7184 struct intel_super *super = st->sb;
b2916f25 7185 struct imsm_super *mpb;
c2c087e6
DW		 7186 struct dl *dl;
7187 unsigned long long pos = 0;
7188 unsigned long long maxsize;
7189 struct extent *e;
7190 int i;
cdddbdbc 7191
88c32bb1
DW		 7192 /* We must have the container info already read in. */
7193 if (!super)
c2c087e6
DW		 7194 return 0;
7195
b2916f25
JS		 7196 mpb = super->anchor;
7197
2cc699af 7198 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7199 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7200 return 0;
d54559f0 7201 }
c2c087e6
DW		 7202 if (!dev) {
7203 /* General test: make sure there is space for
2da8544a
DW		 7204 * 'raiddisks' device extents of size 'size' at a given
7205 * offset
c2c087e6 7206 */
e46273eb 7207 unsigned long long minsize = size;
b7528a20 7208 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7209 int dcnt = 0;
7210 if (minsize == 0)
7211 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7212 for (dl = super->disks; dl ; dl = dl->next) {
7213 int found = 0;
7214
bf5a934a 7215 pos = 0;
c2c087e6 7216 i = 0;
05501181 7217 e = get_extents(super, dl, 0);
c2c087e6
DW		 7218 if (!e) continue;
7219 do {
7220 unsigned long long esize;
7221 esize = e[i].start - pos;
7222 if (esize >= minsize)
7223 found = 1;
b7528a20 7224 if (found && start_offset == MaxSector) {
2da8544a
DW		 7225 start_offset = pos;
7226 break;
7227 } else if (found && pos != start_offset) {
7228 found = 0;
7229 break;
7230 }
c2c087e6
DW		 7231 pos = e[i].start + e[i].size;
7232 i++;
7233 } while (e[i-1].size);
7234 if (found)
7235 dcnt++;
7236 free(e);
7237 }
7238 if (dcnt < raiddisks) {
2c514b71 7239 if (verbose)
7a862a02 7240 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7241 dcnt, raiddisks);
c2c087e6
DW		 7242 return 0;
7243 }
7244 return 1;
7245 }
0dcecb2e 7246
c2c087e6 7247 /* This device must be a member of the set */
9e04ac1c 7248 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7249 return 0;
7250 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7251 if (dl->major == (int)major(rdev) &&
7252 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7253 break;
7254 }
7255 if (!dl) {
2c514b71 7256 if (verbose)
7a862a02 7257 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7258 return 0;
a20d2ba5
DW		 7259 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7260 /* If a volume is present then the current creation attempt
7261 * cannot incorporate new spares because the orom may not
7262 * understand this configuration (all member disks must be
7263 * members of each array in the container).
7264 */
7a862a02
N		 7265 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7266 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7267 return 0;
5fe62b94
WD		 7268 } else if (super->orom && mpb->num_raid_devs > 0 &&
7269 mpb->num_disks != raiddisks) {
7a862a02 7270 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7271 return 0;
c2c087e6 7272 }
0dcecb2e
DW		 7273
7274 /* retrieve the largest free space block */
05501181 7275 e = get_extents(super, dl, 0);
c2c087e6
DW		 7276 maxsize = 0;
7277 i = 0;
0dcecb2e
DW		 7278 if (e) {
7279 do {
7280 unsigned long long esize;
7281
7282 esize = e[i].start - pos;
7283 if (esize >= maxsize)
7284 maxsize = esize;
7285 pos = e[i].start + e[i].size;
7286 i++;
7287 } while (e[i-1].size);
7288 dl->e = e;
7289 dl->extent_cnt = i;
7290 } else {
7291 if (verbose)
e7b84f9d 7292 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7293 dev);
7294 return 0;
7295 }
7296 if (maxsize < size) {
7297 if (verbose)
e7b84f9d 7298 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7299 dev, maxsize, size);
7300 return 0;
7301 }
7302
7303 /* count total number of extents for merge */
7304 i = 0;
7305 for (dl = super->disks; dl; dl = dl->next)
7306 if (dl->e)
7307 i += dl->extent_cnt;
7308
7309 maxsize = merge_extents(super, i);
3baa56ab 7310
1a1ced1e
KS		 7311 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7312 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7313
a7dd165b 7314 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7315 if (verbose) {
7316 if (maxsize == 0)
7a862a02 7317 pr_err("no free space left on device. Aborting...\n");
b3071342 7318 else
7a862a02 7319 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7320 maxsize, size);
7321 }
0dcecb2e 7322 return 0;
0dcecb2e
DW		 7323 }
7324
c2c087e6
DW		 7325 *freesize = maxsize;
7326
ca9de185 7327 if (super->orom) {
72a45777 7328 int count = count_volumes(super->hba,
ca9de185
LM		 7329 super->orom->dpa, verbose);
7330 if (super->orom->vphba <= count) {
676e87a8 7331 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7332 super->orom->vphba);
7333 return 0;
7334 }
7335 }
c2c087e6 7336 return 1;
cdddbdbc
DW		 7337}
7338
13bcac90 7339static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7340 unsigned long long size, int chunk,
7341 unsigned long long *freesize)
7342{
7343 struct intel_super *super = st->sb;
7344 struct imsm_super *mpb = super->anchor;
7345 struct dl *dl;
7346 int i;
7347 int extent_cnt;
7348 struct extent *e;
7349 unsigned long long maxsize;
7350 unsigned long long minsize;
7351 int cnt;
7352 int used;
7353
7354 /* find the largest common start free region of the possible disks */
7355 used = 0;
7356 extent_cnt = 0;
7357 cnt = 0;
7358 for (dl = super->disks; dl; dl = dl->next) {
7359 dl->raiddisk = -1;
7360
7361 if (dl->index >= 0)
7362 used++;
7363
7364 /* don't activate new spares if we are orom constrained
7365 * and there is already a volume active in the container
7366 */
7367 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7368 continue;
7369
05501181 7370 e = get_extents(super, dl, 0);
efb30e7f
DW		 7371 if (!e)
7372 continue;
7373 for (i = 1; e[i-1].size; i++)
7374 ;
7375 dl->e = e;
7376 dl->extent_cnt = i;
7377 extent_cnt += i;
7378 cnt++;
7379 }
7380
7381 maxsize = merge_extents(super, extent_cnt);
7382 minsize = size;
7383 if (size == 0)
612e59d8
CA		 7384 /* chunk is in K */
7385 minsize = chunk * 2;
efb30e7f
DW		 7386
7387 if (cnt < raiddisks ||
7388 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7389 maxsize < minsize ||
7390 maxsize == 0) {
e7b84f9d 7391 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7392 return 0; /* No enough free spaces large enough */
7393 }
7394
7395 if (size == 0) {
7396 size = maxsize;
7397 if (chunk) {
612e59d8
CA		 7398 size /= 2 * chunk;
7399 size *= 2 * chunk;
efb30e7f 7400 }
f878b242
LM		 7401 maxsize = size;
7402 }
1a1ced1e
KS		 7403 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7404 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7405 cnt = 0;
7406 for (dl = super->disks; dl; dl = dl->next)
7407 if (dl->e)
7408 dl->raiddisk = cnt++;
7409
7410 *freesize = size;
7411
13bcac90
AK		 7412 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7413
efb30e7f
DW		 7414 return 1;
7415}
7416
13bcac90
AK		 7417static int reserve_space(struct supertype *st, int raiddisks,
7418 unsigned long long size, int chunk,
7419 unsigned long long *freesize)
7420{
7421 struct intel_super *super = st->sb;
7422 struct dl *dl;
7423 int cnt;
7424 int rv = 0;
7425
7426 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7427 if (rv) {
7428 cnt = 0;
7429 for (dl = super->disks; dl; dl = dl->next)
7430 if (dl->e)
7431 dl->raiddisk = cnt++;
7432 rv = 1;
7433 }
7434
7435 return rv;
7436}
7437
bf5a934a 7438static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7439 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7440 unsigned long long data_offset,
bf5a934a 7441 char *dev, unsigned long long *freesize,
5308f117 7442 int consistency_policy, int verbose)
bf5a934a
DW		 7443{
7444 int fd, cfd;
7445 struct mdinfo *sra;
20cbe8d2 7446 int is_member = 0;
bf5a934a 7447
d54559f0
LM		 7448 /* load capability
7449 * if given unused devices create a container
bf5a934a
DW		 7450 * if given given devices in a container create a member volume
7451 */
7452 if (level == LEVEL_CONTAINER) {
7453 /* Must be a fresh device to add to a container */
7454 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7455 raiddisks,
7ccc4cc4 7456 *chunk,
af4348dd 7457 size, data_offset,
bf5a934a
DW		 7458 dev, freesize,
7459 verbose);
7460 }
9587c373 7461
06a6101c
BK		 7462 /*
7463 * Size is given in sectors.
7464 */
7465 if (size && (size < 2048)) {
22dc741f 7466 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7467 /* Depends on algorithm in Create.c :
7468 * if container was given (dev == NULL) return -1,
7469 * if block device was given ( dev != NULL) return 0.
7470 */
7471 return dev ? -1 : 0;
7472 }
7473
8592f29d 7474 if (!dev) {
e91a3bad 7475 if (st->sb) {
ca9de185 7476 struct intel_super *super = st->sb;
e91a3bad 7477 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7478 raiddisks, chunk, size,
e91a3bad
LM		 7479 verbose))
7480 return 0;
efb30e7f
DW		 7481 /* we are being asked to automatically layout a
7482 * new volume based on the current contents of
7483 * the container. If the the parameters can be
7484 * satisfied reserve_space will record the disks,
7485 * start offset, and size of the volume to be
7486 * created. add_to_super and getinfo_super
7487 * detect when autolayout is in progress.
7488 */
ca9de185
LM		 7489 /* assuming that freesize is always given when array is
7490 created */
7491 if (super->orom && freesize) {
7492 int count;
72a45777 7493 count = count_volumes(super->hba,
ca9de185
LM		 7494 super->orom->dpa, verbose);
7495 if (super->orom->vphba <= count) {
676e87a8 7496 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7497 super->orom->vphba);
7498 return 0;
7499 }
7500 }
e91a3bad
LM		 7501 if (freesize)
7502 return reserve_space(st, raiddisks, size,
7ccc4cc4 7503 *chunk, freesize);
8592f29d
N		 7504 }
7505 return 1;
7506 }
bf5a934a
DW		 7507 if (st->sb) {
7508 /* creating in a given container */
7509 return validate_geometry_imsm_volume(st, level, layout,
7510 raiddisks, chunk, size,
af4348dd 7511 data_offset,
bf5a934a
DW		 7512 dev, freesize, verbose);
7513 }
7514
bf5a934a
DW		 7515 /* This device needs to be a device in an 'imsm' container */
7516 fd = open(dev, O_RDONLY|O_EXCL, 0);
7517 if (fd >= 0) {
7518 if (verbose)
e7b84f9d
N		 7519 pr_err("Cannot create this array on device %s\n",
7520 dev);
bf5a934a
DW		 7521 close(fd);
7522 return 0;
7523 }
7524 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7525 if (verbose)
e7b84f9d 7526 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7527 dev, strerror(errno));
7528 return 0;
7529 }
7530 /* Well, it is in use by someone, maybe an 'imsm' container. */
7531 cfd = open_container(fd);
20cbe8d2 7532 close(fd);
bf5a934a 7533 if (cfd < 0) {
bf5a934a 7534 if (verbose)
e7b84f9d 7535 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7536 dev);
7537 return 0;
7538 }
4dd2df09 7539 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7540 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7541 strcmp(sra->text_version, "imsm") == 0)
7542 is_member = 1;
7543 sysfs_free(sra);
7544 if (is_member) {
bf5a934a
DW		 7545 /* This is a member of a imsm container. Load the container
7546 * and try to create a volume
7547 */
7548 struct intel_super *super;
7549
ec50f7b6 7550 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7551 st->sb = super;
4dd2df09 7552 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7553 close(cfd);
7554 return validate_geometry_imsm_volume(st, level, layout,
7555 raiddisks, chunk,
af4348dd 7556 size, data_offset, dev,
ecbd9e81
N		 7557 freesize, 1)
7558 ? 1 : -1;
bf5a934a 7559 }
20cbe8d2 7560 }
bf5a934a 7561
20cbe8d2 7562 if (verbose)
e7b84f9d 7563 pr_err("failed container membership check\n");
20cbe8d2
AW		 7564
7565 close(cfd);
7566 return 0;
bf5a934a 7567}
0bd16cf2 7568
30f58b22 7569static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7570{
7571 struct intel_super *super = st->sb;
7572
30f58b22
DW		 7573 if (level && *level == UnSet)
7574 *level = LEVEL_CONTAINER;
7575
7576 if (level && layout && *layout == UnSet)
7577 *layout = imsm_level_to_layout(*level);
0bd16cf2 7578
cd9d1ac7
DW		 7579 if (chunk && (*chunk == UnSet || *chunk == 0))
7580 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7581}
7582
33414a01
DW		 7583static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7584
3364781b 7585static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7586{
3364781b 7587 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7588 __u8 i;
7589 struct intel_dev **dp;
7590 struct intel_super *super = st->sb;
3364781b 7591 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7592 struct imsm_super *mpb = super->anchor;
7593
3364781b 7594 if (mpb->num_raid_devs == 0)
33414a01 7595 return 2;
33414a01
DW		 7596
7597 /* block deletions that would change the uuid of active subarrays
7598 *
7599 * FIXME when immutable ids are available, but note that we'll
7600 * also need to fixup the invalidated/active subarray indexes in
7601 * mdstat
7602 */
7603 for (i = 0; i < mpb->num_raid_devs; i++) {
7604 char subarray[4];
7605
7606 if (i < current_vol)
7607 continue;
7608 sprintf(subarray, "%u", i);
4dd2df09 7609 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7610 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7611 current_vol, i);
33414a01
DW		 7612
7613 return 2;
7614 }
7615 }
7616
7617 if (st->update_tail) {
503975b9 7618 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7619
33414a01
DW		 7620 u->type = update_kill_array;
7621 u->dev_idx = current_vol;
7622 append_metadata_update(st, u, sizeof(*u));
7623
7624 return 0;
7625 }
7626
7627 for (dp = &super->devlist; *dp;)
7628 if ((*dp)->index == current_vol) {
7629 *dp = (*dp)->next;
7630 } else {
7631 handle_missing(super, (*dp)->dev);
7632 if ((*dp)->index > current_vol)
7633 (*dp)->index--;
7634 dp = &(*dp)->next;
7635 }
7636
7637 /* no more raid devices, all active components are now spares,
7638 * but of course failed are still failed
7639 */
7640 if (--mpb->num_raid_devs == 0) {
7641 struct dl *d;
7642
7643 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7644 if (d->index > -2)
7645 mark_spare(d);
33414a01
DW		 7646 }
7647
7648 super->updates_pending++;
7649
7650 return 0;
7651}
aa534678 7652
a951a4f7 7653static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7654 char *update, struct mddev_ident *ident)
aa534678
DW		 7655{
7656 /* update the subarray currently referenced by ->current_vol */
7657 struct intel_super *super = st->sb;
7658 struct imsm_super *mpb = super->anchor;
7659
aa534678
DW		 7660 if (strcmp(update, "name") == 0) {
7661 char *name = ident->name;
a951a4f7
N		 7662 char *ep;
7663 int vol;
aa534678 7664
4dd2df09 7665 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7666 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7667 return 2;
7668 }
7669
7670 if (!check_name(super, name, 0))
7671 return 2;
7672
a951a4f7
N		 7673 vol = strtoul(subarray, &ep, 10);
7674 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7675 return 2;
7676
aa534678 7677 if (st->update_tail) {
503975b9 7678 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7679
aa534678 7680 u->type = update_rename_array;
a951a4f7 7681 u->dev_idx = vol;
618f4e6d
XN		 7682 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7683 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7684 append_metadata_update(st, u, sizeof(*u));
7685 } else {
7686 struct imsm_dev *dev;
ebad3af2 7687 int i, namelen;
aa534678 7688
a951a4f7 7689 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7690 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7691 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7692 memcpy(dev->volume, name, namelen);
aa534678
DW		 7693 for (i = 0; i < mpb->num_raid_devs; i++) {
7694 dev = get_imsm_dev(super, i);
7695 handle_missing(super, dev);
7696 }
7697 super->updates_pending++;
7698 }
e6e9dd3f
AP		 7699 } else if (strcmp(update, "ppl") == 0 ||
7700 strcmp(update, "no-ppl") == 0) {
7701 int new_policy;
7702 char *ep;
7703 int vol = strtoul(subarray, &ep, 10);
7704
7705 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7706 return 2;
7707
7708 if (strcmp(update, "ppl") == 0)
c2462068 7709 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7710 else
c2462068 7711 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7712
7713 if (st->update_tail) {
7714 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7715
7716 u->type = update_rwh_policy;
7717 u->dev_idx = vol;
7718 u->new_policy = new_policy;
7719 append_metadata_update(st, u, sizeof(*u));
7720 } else {
7721 struct imsm_dev *dev;
7722
7723 dev = get_imsm_dev(super, vol);
7724 dev->rwh_policy = new_policy;
7725 super->updates_pending++;
7726 }
aa534678
DW		 7727 } else
7728 return 2;
7729
7730 return 0;
7731}
bf5a934a 7732
28bce06f
AK		 7733static int is_gen_migration(struct imsm_dev *dev)
7734{
7534230b
AK		 7735 if (dev == NULL)
7736 return 0;
7737
28bce06f
AK		 7738 if (!dev->vol.migr_state)
7739 return 0;
7740
7741 if (migr_type(dev) == MIGR_GEN_MIGR)
7742 return 1;
7743
7744 return 0;
7745}
7746
1e5c6983
DW		 7747static int is_rebuilding(struct imsm_dev *dev)
7748{
7749 struct imsm_map *migr_map;
7750
7751 if (!dev->vol.migr_state)
7752 return 0;
7753
7754 if (migr_type(dev) != MIGR_REBUILD)
7755 return 0;
7756
238c0a71 7757 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7758
7759 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7760 return 1;
7761 else
7762 return 0;
7763}
7764
6ce1fbf1
AK		 7765static int is_initializing(struct imsm_dev *dev)
7766{
7767 struct imsm_map *migr_map;
7768
7769 if (!dev->vol.migr_state)
7770 return 0;
7771
7772 if (migr_type(dev) != MIGR_INIT)
7773 return 0;
7774
238c0a71 7775 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7776
7777 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7778 return 1;
7779
7780 return 0;
6ce1fbf1
AK		 7781}
7782
c47b0ff6
AK		 7783static void update_recovery_start(struct intel_super *super,
7784 struct imsm_dev *dev,
7785 struct mdinfo *array)
1e5c6983
DW		 7786{
7787 struct mdinfo *rebuild = NULL;
7788 struct mdinfo *d;
7789 __u32 units;
7790
7791 if (!is_rebuilding(dev))
7792 return;
7793
7794 /* Find the rebuild target, but punt on the dual rebuild case */
7795 for (d = array->devs; d; d = d->next)
7796 if (d->recovery_start == 0) {
7797 if (rebuild)
7798 return;
7799 rebuild = d;
7800 }
7801
4363fd80
DW		 7802 if (!rebuild) {
7803 /* (?) none of the disks are marked with
7804 * IMSM_ORD_REBUILD, so assume they are missing and the
7805 * disk_ord_tbl was not correctly updated
7806 */
1ade5cc1 7807 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7808 return;
7809 }
7810
1e5c6983 7811 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7812 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7813}
7814
276d77db 7815static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7816
00bbdbda 7817static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7818{
4f5bc454
DW		 7819 /* Given a container loaded by load_super_imsm_all,
7820 * extract information about all the arrays into
7821 * an mdinfo tree.
00bbdbda 7822 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7823 *
7824 * For each imsm_dev create an mdinfo, fill it in,
7825 * then look for matching devices in super->disks
7826 * and create appropriate device mdinfo.
7827 */
7828 struct intel_super *super = st->sb;
949c47a0 7829 struct imsm_super *mpb = super->anchor;
4f5bc454 7830 struct mdinfo *rest = NULL;
00bbdbda 7831 unsigned int i;
81219e70 7832 int sb_errors = 0;
abef11a3
AK		 7833 struct dl *d;
7834 int spare_disks = 0;
b6180160 7835 int current_vol = super->current_vol;
cdddbdbc 7836
19482bcc
AK		 7837 /* do not assemble arrays when not all attributes are supported */
7838 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7839 sb_errors = 1;
7a862a02 7840 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7841 }
7842
abef11a3
AK		 7843 /* count spare devices, not used in maps
7844 */
7845 for (d = super->disks; d; d = d->next)
7846 if (d->index == -1)
7847 spare_disks++;
7848
4f5bc454 7849 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7850 struct imsm_dev *dev;
7851 struct imsm_map *map;
86e3692b 7852 struct imsm_map *map2;
4f5bc454 7853 struct mdinfo *this;
a6482415 7854 int slot;
a6482415 7855 int chunk;
00bbdbda 7856 char *ep;
8b9cd157 7857 int level;
00bbdbda
N		 7858
7859 if (subarray &&
7860 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7861 continue;
7862
7863 dev = get_imsm_dev(super, i);
238c0a71
AK		 7864 map = get_imsm_map(dev, MAP_0);
7865 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7866 level = get_imsm_raid_level(map);
4f5bc454 7867
1ce0101c
DW		 7868 /* do not publish arrays that are in the middle of an
7869 * unsupported migration
7870 */
7871 if (dev->vol.migr_state &&
28bce06f 7872 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7873 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7874 dev->volume);
7875 continue;
7876 }
2db86302
LM		 7877 /* do not publish arrays that are not support by controller's
7878 * OROM/EFI
7879 */
1ce0101c 7880
503975b9 7881 this = xmalloc(sizeof(*this));
4f5bc454 7882
301406c9 7883 super->current_vol = i;
a5d85af7 7884 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7885 this->next = rest;
a6482415 7886 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7887 /* mdadm does not support all metadata features- set the bit in all arrays state */
7888 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7889 level, /* RAID level */
7890 imsm_level_to_layout(level),
81219e70 7891 map->num_members, /* raid disks */
fcc2c9da 7892 &chunk, imsm_dev_size(dev),
81219e70 7893 1 /* verbose */)) {
7a862a02 7894 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7895 dev->volume);
7896 this->array.state |=
7897 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7898 (1<<MD_SB_BLOCK_VOLUME);
7899 }
81219e70
LM		 7900
7901 /* if array has bad blocks, set suitable bit in all arrays state */
7902 if (sb_errors)
7903 this->array.state |=
7904 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7905 (1<<MD_SB_BLOCK_VOLUME);
7906
4f5bc454 7907 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7908 unsigned long long recovery_start;
4f5bc454
DW		 7909 struct mdinfo *info_d;
7910 struct dl *d;
7911 int idx;
9a1608e5 7912 int skip;
7eef0453 7913 __u32 ord;
8b9cd157 7914 int missing = 0;
4f5bc454 7915
9a1608e5 7916 skip = 0;
238c0a71
AK		 7917 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7918 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7919 for (d = super->disks; d ; d = d->next)
7920 if (d->index == idx)
0fbd635c 7921 break;
4f5bc454 7922
1e5c6983 7923 recovery_start = MaxSector;
4f5bc454 7924 if (d == NULL)
9a1608e5 7925 skip = 1;
25ed7e59 7926 if (d && is_failed(&d->disk))
9a1608e5 7927 skip = 1;
8b9cd157 7928 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7929 recovery_start = 0;
1e93d0d1
BK		 7930 if (!(ord & IMSM_ORD_REBUILD))
7931 this->array.working_disks++;
1011e834 7932 /*
9a1608e5 7933 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7934 * reset resync start to avoid a dirty-degraded
7935 * situation when performing the intial sync
9a1608e5 7936 */
8b9cd157
MK		 7937 if (skip)
7938 missing++;
7939
7940 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7941 if ((!able_to_resync(level, missing) ||
7942 recovery_start == 0))
7943 this->resync_start = MaxSector;
7944 } else {
7945 /*
7946 * FIXME handle dirty degraded
7947 */
7948 }
7949
9a1608e5
DW		 7950 if (skip)
7951 continue;
4f5bc454 7952
503975b9 7953 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7954 info_d->next = this->devs;
7955 this->devs = info_d;
7956
4f5bc454
DW		 7957 info_d->disk.number = d->index;
7958 info_d->disk.major = d->major;
7959 info_d->disk.minor = d->minor;
7960 info_d->disk.raid_disk = slot;
1e5c6983 7961 info_d->recovery_start = recovery_start;
86e3692b
AK		 7962 if (map2) {
7963 if (slot < map2->num_members)
7964 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7965 else
7966 this->array.spare_disks++;
86e3692b
AK		 7967 } else {
7968 if (slot < map->num_members)
7969 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7970 else
7971 this->array.spare_disks++;
86e3692b 7972 }
4f5bc454
DW		 7973
7974 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7975 info_d->data_offset = pba_of_lba0(map);
44490938 7976 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7977
7978 if (map->raid_level == 5) {
2432ce9b
AP		 7979 info_d->ppl_sector = this->ppl_sector;
7980 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7981 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7982 recovery_start == 0)
7983 this->resync_start = 0;
06fb291a 7984 }
b12796be 7985
5e46202e 7986 info_d->bb.supported = 1;
b12796be
TM		 7987 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7988 info_d->data_offset,
7989 info_d->component_size,
7990 &info_d->bb);
4f5bc454 7991 }
1e5c6983 7992 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7993 update_recovery_start(super, dev, this);
abef11a3 7994 this->array.spare_disks += spare_disks;
276d77db
AK		 7995
7996 /* check for reshape */
7997 if (this->reshape_active == 1)
7998 recover_backup_imsm(st, this);
9a1608e5 7999 rest = this;
4f5bc454
DW		 8000 }
8001
b6180160 8002 super->current_vol = current_vol;
4f5bc454 8003 return rest;
cdddbdbc
DW		 8004}
8005
3b451610
AK		 8006static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8007 int failed, int look_in_map)
c2a1e7da 8008{
3b451610
AK		 8009 struct imsm_map *map;
8010
8011 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8012
8013 if (!failed)
1011e834 8014 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8015 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8016
8017 switch (get_imsm_raid_level(map)) {
8018 case 0:
8019 return IMSM_T_STATE_FAILED;
8020 break;
8021 case 1:
8022 if (failed < map->num_members)
8023 return IMSM_T_STATE_DEGRADED;
8024 else
8025 return IMSM_T_STATE_FAILED;
8026 break;
8027 case 10:
8028 {
8029 /**
c92a2527
DW		 8030 * check to see if any mirrors have failed, otherwise we
8031 * are degraded. Even numbered slots are mirrored on
8032 * slot+1
c2a1e7da 8033 */
c2a1e7da 8034 int i;
d9b420a5
N		 8035 /* gcc -Os complains that this is unused */
8036 int insync = insync;
c2a1e7da
DW		 8037
8038 for (i = 0; i < map->num_members; i++) {
238c0a71 8039 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8040 int idx = ord_to_idx(ord);
8041 struct imsm_disk *disk;
c2a1e7da 8042
c92a2527 8043 /* reset the potential in-sync count on even-numbered
1011e834 8044 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8045 */
8046 if ((i & 1) == 0)
8047 insync = 2;
c2a1e7da 8048
c92a2527 8049 disk = get_imsm_disk(super, idx);
25ed7e59 8050 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8051 insync--;
c2a1e7da 8052
c92a2527
DW		 8053 /* no in-sync disks left in this mirror the
8054 * array has failed
8055 */
8056 if (insync == 0)
8057 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8058 }
8059
8060 return IMSM_T_STATE_DEGRADED;
8061 }
8062 case 5:
8063 if (failed < 2)
8064 return IMSM_T_STATE_DEGRADED;
8065 else
8066 return IMSM_T_STATE_FAILED;
8067 break;
8068 default:
8069 break;
8070 }
8071
8072 return map->map_state;
8073}
8074
3b451610
AK		 8075static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8076 int look_in_map)
c2a1e7da
DW		 8077{
8078 int i;
8079 int failed = 0;
8080 struct imsm_disk *disk;
d5985138
AK		 8081 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8082 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8083 struct imsm_map *map_for_loop;
0556e1a2
DW		 8084 __u32 ord;
8085 int idx;
d5985138 8086 int idx_1;
c2a1e7da 8087
0556e1a2
DW		 8088 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8089 * disks that are being rebuilt. New failures are recorded to
8090 * map[0]. So we look through all the disks we started with and
8091 * see if any failures are still present, or if any new ones
8092 * have arrived
0556e1a2 8093 */
d5985138
AK		 8094 map_for_loop = map;
8095 if (prev && (map->num_members < prev->num_members))
8096 map_for_loop = prev;
68fe4598
LD		 8097
8098 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8099 idx_1 = -255;
238c0a71
AK		 8100 /* when MAP_X is passed both maps failures are counted
8101 */
d5985138 8102 if (prev &&
089f9d79
JS		 8103 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8104 i < prev->num_members) {
d5985138
AK		 8105 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8106 idx_1 = ord_to_idx(ord);
c2a1e7da 8107
d5985138
AK		 8108 disk = get_imsm_disk(super, idx_1);
8109 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8110 failed++;
8111 }
089f9d79
JS		 8112 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8113 i < map->num_members) {
d5985138
AK		 8114 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8115 idx = ord_to_idx(ord);
8116
8117 if (idx != idx_1) {
8118 disk = get_imsm_disk(super, idx);
8119 if (!disk || is_failed(disk) ||
8120 ord & IMSM_ORD_REBUILD)
8121 failed++;
8122 }
8123 }
c2a1e7da
DW		 8124 }
8125
8126 return failed;
845dea95
NB		 8127}
8128
97b4d0e9
DW		 8129static int imsm_open_new(struct supertype *c, struct active_array *a,
8130 char *inst)
8131{
8132 struct intel_super *super = c->sb;
8133 struct imsm_super *mpb = super->anchor;
bbab0940 8134 struct imsm_update_prealloc_bb_mem u;
9587c373 8135
97b4d0e9 8136 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8137 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8138 return -ENODEV;
8139 }
8140
8141 dprintf("imsm: open_new %s\n", inst);
8142 a->info.container_member = atoi(inst);
bbab0940
TM		 8143
8144 u.type = update_prealloc_badblocks_mem;
8145 imsm_update_metadata_locally(c, &u, sizeof(u));
8146
97b4d0e9
DW		 8147 return 0;
8148}
8149
0c046afd
DW		 8150static int is_resyncing(struct imsm_dev *dev)
8151{
8152 struct imsm_map *migr_map;
8153
8154 if (!dev->vol.migr_state)
8155 return 0;
8156
1484e727
DW		 8157 if (migr_type(dev) == MIGR_INIT ||
8158 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8159 return 1;
8160
4c9bc37b
AK		 8161 if (migr_type(dev) == MIGR_GEN_MIGR)
8162 return 0;
8163
238c0a71 8164 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8165
089f9d79
JS		 8166 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8167 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8168 return 1;
8169 else
8170 return 0;
8171}
8172
0556e1a2 8173/* return true if we recorded new information */
4c9e8c1e
TM		 8174static int mark_failure(struct intel_super *super,
8175 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8176{
0556e1a2
DW		 8177 __u32 ord;
8178 int slot;
8179 struct imsm_map *map;
86c54047
DW		 8180 char buf[MAX_RAID_SERIAL_LEN+3];
8181 unsigned int len, shift = 0;
0556e1a2
DW		 8182
8183 /* new failures are always set in map[0] */
238c0a71 8184 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8185
8186 slot = get_imsm_disk_slot(map, idx);
8187 if (slot < 0)
8188 return 0;
8189
8190 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8191 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8192 return 0;
8193
7d0c5e24
LD		 8194 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8195 buf[MAX_RAID_SERIAL_LEN] = '\000';
8196 strcat(buf, ":0");
86c54047
DW		 8197 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8198 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8199 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8200
f2f27e63 8201 disk->status |= FAILED_DISK;
0556e1a2 8202 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8203 /* mark failures in second map if second map exists and this disk
8204 * in this slot.
8205 * This is valid for migration, initialization and rebuild
8206 */
8207 if (dev->vol.migr_state) {
238c0a71 8208 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8209 int slot2 = get_imsm_disk_slot(map2, idx);
8210
089f9d79 8211 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8212 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8213 idx | IMSM_ORD_REBUILD);
8214 }
d7a1fda2
MT		 8215 if (map->failed_disk_num == 0xff ||
8216 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8217 map->failed_disk_num = slot;
4c9e8c1e
TM		 8218
8219 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8220
0556e1a2
DW		 8221 return 1;
8222}
8223
4c9e8c1e
TM		 8224static void mark_missing(struct intel_super *super,
8225 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8226{
4c9e8c1e 8227 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8228
8229 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8230 return;
8231
47ee5a45
DW		 8232 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8233 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8234}
8235
33414a01
DW		 8236static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8237{
33414a01 8238 struct dl *dl;
33414a01
DW		 8239
8240 if (!super->missing)
8241 return;
33414a01 8242
79b68f1b
PC		 8243 /* When orom adds replacement for missing disk it does
8244 * not remove entry of missing disk, but just updates map with
8245 * new added disk. So it is not enough just to test if there is
8246 * any missing disk, we have to look if there are any failed disks
8247 * in map to stop migration */
8248
33414a01 8249 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8250 /* end process for initialization and rebuild only
8251 */
8252 if (is_gen_migration(dev) == 0) {
fb12a745 8253 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8254
fb12a745
TM		 8255 if (failed) {
8256 __u8 map_state;
8257 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8258 struct imsm_map *map1;
8259 int i, ord, ord_map1;
8260 int rebuilt = 1;
3d59f0c0 8261
fb12a745
TM		 8262 for (i = 0; i < map->num_members; i++) {
8263 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8264 if (!(ord & IMSM_ORD_REBUILD))
8265 continue;
8266
8267 map1 = get_imsm_map(dev, MAP_1);
8268 if (!map1)
8269 continue;
8270
8271 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8272 if (ord_map1 & IMSM_ORD_REBUILD)
8273 rebuilt = 0;
8274 }
8275
8276 if (rebuilt) {
8277 map_state = imsm_check_degraded(super, dev,
8278 failed, MAP_0);
8279 end_migration(dev, super, map_state);
8280 }
8281 }
3d59f0c0 8282 }
33414a01 8283 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8284 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8285 super->updates_pending++;
8286}
8287
f3871fdc
AK		 8288static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8289 long long new_size)
70bdf0dc 8290{
70bdf0dc 8291 unsigned long long array_blocks;
9529d343
MD		 8292 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8293 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8294
8295 if (used_disks == 0) {
8296 /* when problems occures
8297 * return current array_blocks value
8298 */
fcc2c9da 8299 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8300
8301 return array_blocks;
8302 }
8303
8304 /* set array size in metadata
8305 */
9529d343 8306 if (new_size <= 0)
f3871fdc
AK		 8307 /* OLCE size change is caused by added disks
8308 */
44490938 8309 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8310 else
f3871fdc
AK		 8311 /* Online Volume Size Change
8312 * Using available free space
8313 */
8314 array_blocks = new_size;
70bdf0dc 8315
b53bfba6 8316 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8317 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8318
8319 return array_blocks;
8320}
8321
28bce06f
AK		 8322static void imsm_set_disk(struct active_array *a, int n, int state);
8323
0e2d1a4e
AK		 8324static void imsm_progress_container_reshape(struct intel_super *super)
8325{
8326 /* if no device has a migr_state, but some device has a
8327 * different number of members than the previous device, start
8328 * changing the number of devices in this device to match
8329 * previous.
8330 */
8331 struct imsm_super *mpb = super->anchor;
8332 int prev_disks = -1;
8333 int i;
1dfaa380 8334 int copy_map_size;
0e2d1a4e
AK		 8335
8336 for (i = 0; i < mpb->num_raid_devs; i++) {
8337 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8338 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8339 struct imsm_map *map2;
8340 int prev_num_members;
0e2d1a4e
AK		 8341
8342 if (dev->vol.migr_state)
8343 return;
8344
8345 if (prev_disks == -1)
8346 prev_disks = map->num_members;
8347 if (prev_disks == map->num_members)
8348 continue;
8349
8350 /* OK, this array needs to enter reshape mode.
8351 * i.e it needs a migr_state
8352 */
8353
1dfaa380 8354 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8355 prev_num_members = map->num_members;
8356 map->num_members = prev_disks;
8357 dev->vol.migr_state = 1;
8358 dev->vol.curr_migr_unit = 0;
ea672ee1 8359 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8360 for (i = prev_num_members;
8361 i < map->num_members; i++)
8362 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8363 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8364 /* Copy the current map */
1dfaa380 8365 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8366 map2->num_members = prev_num_members;
8367
f3871fdc 8368 imsm_set_array_size(dev, -1);
51d83f5d 8369 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8370 super->updates_pending++;
8371 }
8372}
8373
aad6f216 8374/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8375 * states are handled in imsm_set_disk() with one exception, when a
8376 * resync is stopped due to a new failure this routine will set the
8377 * 'degraded' state for the array.
8378 */
01f157d7 8379static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8380{
8381 int inst = a->info.container_member;
8382 struct intel_super *super = a->container->sb;
949c47a0 8383 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8384 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8385 int failed = imsm_count_failed(super, dev, MAP_0);
8386 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8387 __u32 blocks_per_unit;
a862209d 8388
1af97990
AK		 8389 if (dev->vol.migr_state &&
8390 dev->vol.migr_type == MIGR_GEN_MIGR) {
8391 /* array state change is blocked due to reshape action
aad6f216
N		 8392 * We might need to
8393 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8394 * - finish the reshape (if last_checkpoint is big and action != reshape)
8395 * - update curr_migr_unit
1af97990 8396 */
aad6f216
N		 8397 if (a->curr_action == reshape) {
8398 /* still reshaping, maybe update curr_migr_unit */
633b5610 8399 goto mark_checkpoint;
aad6f216
N		 8400 } else {
8401 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8402 /* for some reason we aborted the reshape.
b66e591b
AK		 8403 *
8404 * disable automatic metadata rollback
8405 * user action is required to recover process
aad6f216 8406 */
b66e591b 8407 if (0) {
238c0a71
AK		 8408 struct imsm_map *map2 =
8409 get_imsm_map(dev, MAP_1);
8410 dev->vol.migr_state = 0;
8411 set_migr_type(dev, 0);
8412 dev->vol.curr_migr_unit = 0;
8413 memcpy(map, map2,
8414 sizeof_imsm_map(map2));
8415 super->updates_pending++;
b66e591b 8416 }
aad6f216
N		 8417 }
8418 if (a->last_checkpoint >= a->info.component_size) {
8419 unsigned long long array_blocks;
8420 int used_disks;
e154ced3 8421 struct mdinfo *mdi;
aad6f216 8422
9529d343 8423 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8424 if (used_disks > 0) {
8425 array_blocks =
44490938 8426 per_dev_array_size(map) *
d55adef9 8427 used_disks;
b53bfba6
TM		 8428 array_blocks =
8429 round_size_to_mb(array_blocks,
8430 used_disks);
d55adef9
AK		 8431 a->info.custom_array_size = array_blocks;
8432 /* encourage manager to update array
8433 * size
8434 */
e154ced3 8435
d55adef9 8436 a->check_reshape = 1;
633b5610 8437 }
e154ced3
AK		 8438 /* finalize online capacity expansion/reshape */
8439 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8440 imsm_set_disk(a,
8441 mdi->disk.raid_disk,
8442 mdi->curr_state);
8443
0e2d1a4e 8444 imsm_progress_container_reshape(super);
e154ced3 8445 }
aad6f216 8446 }
1af97990
AK		 8447 }
8448
47ee5a45 8449 /* before we activate this array handle any missing disks */
33414a01
DW		 8450 if (consistent == 2)
8451 handle_missing(super, dev);
1e5c6983 8452
0c046afd 8453 if (consistent == 2 &&
b7941fd6 8454 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8455 map_state != IMSM_T_STATE_NORMAL ||
8456 dev->vol.migr_state))
01f157d7 8457 consistent = 0;
272906ef 8458
b7941fd6 8459 if (is_resync_complete(&a->info)) {
0c046afd 8460 /* complete intialization / resync,
0556e1a2
DW		 8461 * recovery and interrupted recovery is completed in
8462 * ->set_disk
0c046afd
DW		 8463 */
8464 if (is_resyncing(dev)) {
8465 dprintf("imsm: mark resync done\n");
809da78e 8466 end_migration(dev, super, map_state);
115c3803 8467 super->updates_pending++;
484240d8 8468 a->last_checkpoint = 0;
115c3803 8469 }
b9172665
AK		 8470 } else if ((!is_resyncing(dev) && !failed) &&
8471 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8472 /* mark the start of the init process if nothing is failed */
b7941fd6 8473 dprintf("imsm: mark resync start\n");
1484e727 8474 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8475 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8476 else
8e59f3d8 8477 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8478 super->updates_pending++;
115c3803 8479 }
a862209d 8480
633b5610 8481mark_checkpoint:
5b83bacf
AK		 8482 /* skip checkpointing for general migration,
8483 * it is controlled in mdadm
8484 */
8485 if (is_gen_migration(dev))
8486 goto skip_mark_checkpoint;
8487
1e5c6983 8488 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8489 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8490 if (blocks_per_unit) {
1e5c6983
DW		 8491 __u32 units32;
8492 __u64 units;
8493
4f0a7acc 8494 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8495 units32 = units;
8496
8497 /* check that we did not overflow 32-bits, and that
8498 * curr_migr_unit needs updating
8499 */
8500 if (units32 == units &&
bfd80a56 8501 units32 != 0 &&
1e5c6983
DW		 8502 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8503 dprintf("imsm: mark checkpoint (%u)\n", units32);
8504 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8505 super->updates_pending++;
8506 }
8507 }
f8f603f1 8508
5b83bacf 8509skip_mark_checkpoint:
3393c6af 8510 /* mark dirty / clean */
2432ce9b
AP		 8511 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8512 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8513 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8514 if (consistent) {
8515 dev->vol.dirty = RAIDVOL_CLEAN;
8516 } else {
8517 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8518 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8519 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8520 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8521 }
a862209d
DW		 8522 super->updates_pending++;
8523 }
28bce06f 8524
01f157d7 8525 return consistent;
a862209d
DW		 8526}
8527
6f50473f
TM		 8528static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8529{
8530 int inst = a->info.container_member;
8531 struct intel_super *super = a->container->sb;
8532 struct imsm_dev *dev = get_imsm_dev(super, inst);
8533 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8534
8535 if (slot > map->num_members) {
8536 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8537 slot, map->num_members - 1);
8538 return -1;
8539 }
8540
8541 if (slot < 0)
8542 return -1;
8543
8544 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8545}
8546
8d45d196 8547static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8548{
8d45d196
DW		 8549 int inst = a->info.container_member;
8550 struct intel_super *super = a->container->sb;
949c47a0 8551 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8552 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8553 struct imsm_disk *disk;
7ce05701
LD		 8554 struct mdinfo *mdi;
8555 int recovery_not_finished = 0;
0c046afd 8556 int failed;
6f50473f 8557 int ord;
0c046afd 8558 __u8 map_state;
fb12a745
TM		 8559 int rebuild_done = 0;
8560 int i;
8d45d196 8561
fb12a745 8562 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8563 if (ord < 0)
8d45d196
DW		 8564 return;
8565
4e6e574a 8566 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8567 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8568
5802a811 8569 /* check for new failures */
ae7d61e3 8570 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8571 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8572 super->updates_pending++;
8d45d196 8573 }
47ee5a45 8574
19859edc 8575 /* check if in_sync */
0556e1a2 8576 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8577 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8578
8579 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8580 rebuild_done = 1;
19859edc
DW		 8581 super->updates_pending++;
8582 }
8d45d196 8583
3b451610
AK		 8584 failed = imsm_count_failed(super, dev, MAP_0);
8585 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8586
0c046afd 8587 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8588 dprintf("imsm: Detected transition to state ");
8589 switch (map_state) {
8590 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8591 dprintf("normal: ");
8592 if (is_rebuilding(dev)) {
1ade5cc1 8593 dprintf_cont("while rebuilding");
7ce05701
LD		 8594 /* check if recovery is really finished */
8595 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8596 if (mdi->recovery_start != MaxSector) {
8597 recovery_not_finished = 1;
8598 break;
8599 }
8600 if (recovery_not_finished) {
1ade5cc1
N		 8601 dprintf_cont("\n");
8602 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8603 if (a->last_checkpoint < mdi->recovery_start) {
8604 a->last_checkpoint = mdi->recovery_start;
8605 super->updates_pending++;
8606 }
8607 break;
8608 }
94002678 8609 end_migration(dev, super, map_state);
94002678
AK		 8610 map->failed_disk_num = ~0;
8611 super->updates_pending++;
8612 a->last_checkpoint = 0;
8613 break;
8614 }
8615 if (is_gen_migration(dev)) {
1ade5cc1 8616 dprintf_cont("while general migration");
bf2f0071 8617 if (a->last_checkpoint >= a->info.component_size)
809da78e 8618 end_migration(dev, super, map_state);
94002678
AK		 8619 else
8620 map->map_state = map_state;
28bce06f 8621 map->failed_disk_num = ~0;
94002678 8622 super->updates_pending++;
bf2f0071 8623 break;
94002678
AK		 8624 }
8625 break;
8626 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8627 dprintf_cont("degraded: ");
089f9d79 8628 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8629 dprintf_cont("mark degraded");
94002678
AK		 8630 map->map_state = map_state;
8631 super->updates_pending++;
8632 a->last_checkpoint = 0;
8633 break;
8634 }
8635 if (is_rebuilding(dev)) {
d7a1fda2 8636 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8637 if (state & DS_FAULTY) {
8638 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8639 if (n == map->failed_disk_num) {
8640 dprintf_cont("end migration");
8641 end_migration(dev, super, map_state);
a4e96fd8 8642 a->last_checkpoint = 0;
d7a1fda2 8643 } else {
a4e96fd8 8644 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8645 map->map_state = map_state;
8646 }
94002678 8647 super->updates_pending++;
fb12a745
TM		 8648 }
8649
a4e96fd8
MT		 8650 if (!rebuild_done)
8651 break;
8652
fb12a745
TM		 8653 /* check if recovery is really finished */
8654 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8655 if (mdi->recovery_start != MaxSector) {
8656 recovery_not_finished = 1;
8657 break;
8658 }
8659 if (recovery_not_finished) {
8660 dprintf_cont("\n");
a4e96fd8 8661 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8662 if (a->last_checkpoint < mdi->recovery_start) {
8663 a->last_checkpoint =
8664 mdi->recovery_start;
8665 super->updates_pending++;
8666 }
8667 break;
94002678 8668 }
fb12a745
TM		 8669
8670 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8671 end_migration(dev, super, map_state);
8672 a->last_checkpoint = 0;
8673 super->updates_pending++;
fb12a745
TM		 8674
8675 for (i = 0; i < map->num_members; i++) {
8676 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8677
8678 if (idx & IMSM_ORD_REBUILD)
8679 map->failed_disk_num = i;
8680 }
8681 super->updates_pending++;
94002678
AK		 8682 break;
8683 }
8684 if (is_gen_migration(dev)) {
1ade5cc1 8685 dprintf_cont("while general migration");
bf2f0071 8686 if (a->last_checkpoint >= a->info.component_size)
809da78e 8687 end_migration(dev, super, map_state);
94002678
AK		 8688 else {
8689 map->map_state = map_state;
3b451610 8690 manage_second_map(super, dev);
94002678
AK		 8691 }
8692 super->updates_pending++;
bf2f0071 8693 break;
28bce06f 8694 }
6ce1fbf1 8695 if (is_initializing(dev)) {
1ade5cc1 8696 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8697 map->map_state = map_state;
8698 super->updates_pending++;
8699 break;
8700 }
94002678
AK		 8701 break;
8702 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8703 dprintf_cont("failed: ");
94002678 8704 if (is_gen_migration(dev)) {
1ade5cc1 8705 dprintf_cont("while general migration");
94002678
AK		 8706 map->map_state = map_state;
8707 super->updates_pending++;
8708 break;
8709 }
8710 if (map->map_state != map_state) {
1ade5cc1 8711 dprintf_cont("mark failed");
94002678
AK		 8712 end_migration(dev, super, map_state);
8713 super->updates_pending++;
8714 a->last_checkpoint = 0;
8715 break;
8716 }
8717 break;
8718 default:
1ade5cc1 8719 dprintf_cont("state %i\n", map_state);
5802a811 8720 }
1ade5cc1 8721 dprintf_cont("\n");
845dea95
NB		 8722}
8723
f796af5d 8724static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8725{
f796af5d 8726 void *buf = mpb;
c2a1e7da
DW		 8727 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8728 unsigned long long dsize;
8729 unsigned long long sectors;
f36a9ecd 8730 unsigned int sector_size;
c2a1e7da 8731
f36a9ecd 8732 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8733 get_dev_size(fd, NULL, &dsize);
8734
f36a9ecd 8735 if (mpb_size > sector_size) {
272f648f 8736 /* -1 to account for anchor */
f36a9ecd 8737 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8738
272f648f 8739 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8740 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8741 SEEK_SET) < 0)
272f648f 8742 return 1;
c2a1e7da 8743
f36a9ecd
PB		 8744 if ((unsigned long long)write(fd, buf + sector_size,
8745 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8746 return 1;
8747 }
c2a1e7da 8748
272f648f 8749 /* first block is stored on second to last sector of the disk */
f36a9ecd 8750 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8751 return 1;
8752
466070ad 8753 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8754 return 1;
8755
c2a1e7da
DW		 8756 return 0;
8757}
8758
2e735d19 8759static void imsm_sync_metadata(struct supertype *container)
845dea95 8760{
2e735d19 8761 struct intel_super *super = container->sb;
c2a1e7da 8762
1a64be56 8763 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8764 if (!super->updates_pending)
8765 return;
8766
36988a3d 8767 write_super_imsm(container, 0);
c2a1e7da
DW		 8768
8769 super->updates_pending = 0;
845dea95
NB		 8770}
8771
272906ef
DW		 8772static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8773{
8774 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8775 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8776 struct dl *dl;
8777
8778 for (dl = super->disks; dl; dl = dl->next)
8779 if (dl->index == i)
8780 break;
8781
25ed7e59 8782 if (dl && is_failed(&dl->disk))
272906ef
DW		 8783 dl = NULL;
8784
8785 if (dl)
1ade5cc1 8786 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8787
8788 return dl;
8789}
8790
a20d2ba5 8791static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8792 struct active_array *a, int activate_new,
8793 struct mdinfo *additional_test_list)
272906ef
DW		 8794{
8795 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8796 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8797 struct imsm_super *mpb = super->anchor;
8798 struct imsm_map *map;
272906ef
DW		 8799 unsigned long long pos;
8800 struct mdinfo *d;
8801 struct extent *ex;
a20d2ba5 8802 int i, j;
272906ef 8803 int found;
569cc43f
DW		 8804 __u32 array_start = 0;
8805 __u32 array_end = 0;
272906ef 8806 struct dl *dl;
6c932028 8807 struct mdinfo *test_list;
272906ef
DW		 8808
8809 for (dl = super->disks; dl; dl = dl->next) {
8810 /* If in this array, skip */
8811 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8812 if (d->state_fd >= 0 &&
8813 d->disk.major == dl->major &&
272906ef 8814 d->disk.minor == dl->minor) {
8ba77d32
AK		 8815 dprintf("%x:%x already in array\n",
8816 dl->major, dl->minor);
272906ef
DW		 8817 break;
8818 }
8819 if (d)
8820 continue;
6c932028
AK		 8821 test_list = additional_test_list;
8822 while (test_list) {
8823 if (test_list->disk.major == dl->major &&
8824 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8825 dprintf("%x:%x already in additional test list\n",
8826 dl->major, dl->minor);
8827 break;
8828 }
6c932028 8829 test_list = test_list->next;
8ba77d32 8830 }
6c932028 8831 if (test_list)
8ba77d32 8832 continue;
272906ef 8833
e553d2a4 8834 /* skip in use or failed drives */
25ed7e59 8835 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8836 dl->index == -2) {
8837 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8838 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8839 continue;
8840 }
8841
a20d2ba5
DW		 8842 /* skip pure spares when we are looking for partially
8843 * assimilated drives
8844 */
8845 if (dl->index == -1 && !activate_new)
8846 continue;
8847
f2cc4f7d
AO		 8848 if (!drive_validate_sector_size(super, dl))
8849 continue;
8850
272906ef 8851 /* Does this unused device have the requisite free space?
a20d2ba5 8852 * It needs to be able to cover all member volumes
272906ef 8853 */
05501181 8854 ex = get_extents(super, dl, 1);
272906ef
DW		 8855 if (!ex) {
8856 dprintf("cannot get extents\n");
8857 continue;
8858 }
a20d2ba5
DW		 8859 for (i = 0; i < mpb->num_raid_devs; i++) {
8860 dev = get_imsm_dev(super, i);
238c0a71 8861 map = get_imsm_map(dev, MAP_0);
272906ef 8862
a20d2ba5
DW		 8863 /* check if this disk is already a member of
8864 * this array
272906ef 8865 */
620b1713 8866 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8867 continue;
8868
8869 found = 0;
8870 j = 0;
8871 pos = 0;
5551b113 8872 array_start = pba_of_lba0(map);
329c8278 8873 array_end = array_start +
44490938 8874 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8875
8876 do {
8877 /* check that we can start at pba_of_lba0 with
44490938 8878 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8879 */
329c8278 8880 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8881 found = 1;
8882 break;
8883 }
8884 pos = ex[j].start + ex[j].size;
8885 j++;
8886 } while (ex[j-1].size);
8887
8888 if (!found)
272906ef 8889 break;
a20d2ba5 8890 }
272906ef
DW		 8891
8892 free(ex);
a20d2ba5 8893 if (i < mpb->num_raid_devs) {
329c8278
DW		 8894 dprintf("%x:%x does not have %u to %u available\n",
8895 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8896 /* No room */
8897 continue;
a20d2ba5
DW		 8898 }
8899 return dl;
272906ef
DW		 8900 }
8901
8902 return dl;
8903}
8904
95d07a2c
LM		 8905static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8906{
8907 struct imsm_dev *dev2;
8908 struct imsm_map *map;
8909 struct dl *idisk;
8910 int slot;
8911 int idx;
8912 __u8 state;
8913
8914 dev2 = get_imsm_dev(cont->sb, dev_idx);
8915 if (dev2) {
238c0a71 8916 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8917 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8918 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8919 if (!map)
8920 return 1;
8921 for (slot = 0; slot < map->num_members; slot++) {
8922 /*
8923 * Check if failed disks are deleted from intel
8924 * disk list or are marked to be deleted
8925 */
238c0a71 8926 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8927 idisk = get_imsm_dl_disk(cont->sb, idx);
8928 /*
8929 * Do not rebuild the array if failed disks
8930 * from failed sub-array are not removed from
8931 * container.
8932 */
8933 if (idisk &&
8934 is_failed(&idisk->disk) &&
8935 (idisk->action != DISK_REMOVE))
8936 return 0;
8937 }
8938 }
8939 }
8940 return 1;
8941}
8942
88758e9d
DW		 8943static struct mdinfo *imsm_activate_spare(struct active_array *a,
8944 struct metadata_update **updates)
8945{
8946 /**
d23fe947
DW		 8947 * Find a device with unused free space and use it to replace a
8948 * failed/vacant region in an array. We replace failed regions one a
8949 * array at a time. The result is that a new spare disk will be added
8950 * to the first failed array and after the monitor has finished
8951 * propagating failures the remainder will be consumed.
88758e9d 8952 *
d23fe947
DW		 8953 * FIXME add a capability for mdmon to request spares from another
8954 * container.
88758e9d
DW		 8955 */
8956
8957 struct intel_super *super = a->container->sb;
88758e9d 8958 int inst = a->info.container_member;
949c47a0 8959 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8960 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8961 int failed = a->info.array.raid_disks;
8962 struct mdinfo *rv = NULL;
8963 struct mdinfo *d;
8964 struct mdinfo *di;
8965 struct metadata_update *mu;
8966 struct dl *dl;
8967 struct imsm_update_activate_spare *u;
8968 int num_spares = 0;
8969 int i;
95d07a2c 8970 int allowed;
88758e9d
DW		 8971
8972 for (d = a->info.devs ; d ; d = d->next) {
8973 if ((d->curr_state & DS_FAULTY) &&
8974 d->state_fd >= 0)
8975 /* wait for Removal to happen */
8976 return NULL;
8977 if (d->state_fd >= 0)
8978 failed--;
8979 }
8980
8981 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8982 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8983
e2962bfc
AK		 8984 if (imsm_reshape_blocks_arrays_changes(super))
8985 return NULL;
1af97990 8986
fc8ca064
AK		 8987 /* Cannot activate another spare if rebuild is in progress already
8988 */
8989 if (is_rebuilding(dev)) {
7a862a02 8990 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8991 return NULL;
8992 }
8993
89c67882
AK		 8994 if (a->info.array.level == 4)
8995 /* No repair for takeovered array
8996 * imsm doesn't support raid4
8997 */
8998 return NULL;
8999
3b451610
AK		 9000 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9001 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9002 return NULL;
9003
83ca7d45
AP		 9004 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9005 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9006 return NULL;
9007 }
9008
95d07a2c
LM		 9009 /*
9010 * If there are any failed disks check state of the other volume.
9011 * Block rebuild if the another one is failed until failed disks
9012 * are removed from container.
9013 */
9014 if (failed) {
7a862a02 9015 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9016 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9017 /* check if states of the other volumes allow for rebuild */
9018 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9019 if (i != inst) {
9020 allowed = imsm_rebuild_allowed(a->container,
9021 i, failed);
9022 if (!allowed)
9023 return NULL;
9024 }
9025 }
9026 }
9027
88758e9d 9028 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9029 for (i = 0; i < a->info.array.raid_disks; i++) {
9030 for (d = a->info.devs ; d ; d = d->next)
9031 if (d->disk.raid_disk == i)
9032 break;
9033 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9034 if (d && (d->state_fd >= 0))
9035 continue;
9036
272906ef 9037 /*
a20d2ba5
DW		 9038 * OK, this device needs recovery. Try to re-add the
9039 * previous occupant of this slot, if this fails see if
9040 * we can continue the assimilation of a spare that was
9041 * partially assimilated, finally try to activate a new
9042 * spare.
272906ef
DW		 9043 */
9044 dl = imsm_readd(super, i, a);
9045 if (!dl)
b303fe21 9046 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9047 if (!dl)
b303fe21 9048 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9049 if (!dl)
9050 continue;
1011e834 9051
272906ef 9052 /* found a usable disk with enough space */
503975b9 9053 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9054
9055 /* dl->index will be -1 in the case we are activating a
9056 * pristine spare. imsm_process_update() will create a
9057 * new index in this case. Once a disk is found to be
9058 * failed in all member arrays it is kicked from the
9059 * metadata
9060 */
9061 di->disk.number = dl->index;
d23fe947 9062
272906ef
DW		 9063 /* (ab)use di->devs to store a pointer to the device
9064 * we chose
9065 */
9066 di->devs = (struct mdinfo *) dl;
9067
9068 di->disk.raid_disk = i;
9069 di->disk.major = dl->major;
9070 di->disk.minor = dl->minor;
9071 di->disk.state = 0;
d23534e4 9072 di->recovery_start = 0;
5551b113 9073 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9074 di->component_size = a->info.component_size;
9075 di->container_member = inst;
5e46202e 9076 di->bb.supported = 1;
2c8890e9 9077 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9078 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9079 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9080 }
148acb7b 9081 super->random = random32();
272906ef
DW		 9082 di->next = rv;
9083 rv = di;
9084 num_spares++;
9085 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9086 i, di->data_offset);
88758e9d
DW		 9087 }
9088
9089 if (!rv)
9090 /* No spares found */
9091 return rv;
9092 /* Now 'rv' has a list of devices to return.
9093 * Create a metadata_update record to update the
9094 * disk_ord_tbl for the array
9095 */
503975b9 9096 mu = xmalloc(sizeof(*mu));
1011e834 9097 mu->buf = xcalloc(num_spares,
503975b9 9098 sizeof(struct imsm_update_activate_spare));
88758e9d 9099 mu->space = NULL;
cb23f1f4 9100 mu->space_list = NULL;
88758e9d
DW		 9101 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9102 mu->next = *updates;
9103 u = (struct imsm_update_activate_spare *) mu->buf;
9104
9105 for (di = rv ; di ; di = di->next) {
9106 u->type = update_activate_spare;
d23fe947
DW		 9107 u->dl = (struct dl *) di->devs;
9108 di->devs = NULL;
88758e9d
DW		 9109 u->slot = di->disk.raid_disk;
9110 u->array = inst;
9111 u->next = u + 1;
9112 u++;
9113 }
9114 (u-1)->next = NULL;
9115 *updates = mu;
9116
9117 return rv;
9118}
9119
54c2c1ea 9120static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9121{
54c2c1ea 9122 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9123 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9124 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9125 struct disk_info *inf = get_disk_info(u);
9126 struct imsm_disk *disk;
8273f55e
DW		 9127 int i;
9128 int j;
8273f55e 9129
54c2c1ea 9130 for (i = 0; i < map->num_members; i++) {
238c0a71 9131 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9132 for (j = 0; j < new_map->num_members; j++)
9133 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9134 return 1;
9135 }
9136
9137 return 0;
9138}
9139
1a64be56
LM		 9140static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9141{
594dc1b8
JS		 9142 struct dl *dl;
9143
1a64be56 9144 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9145 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9146 return dl;
9147 return NULL;
9148}
9149
9150static int remove_disk_super(struct intel_super *super, int major, int minor)
9151{
594dc1b8 9152 struct dl *prev;
1a64be56
LM		 9153 struct dl *dl;
9154
9155 prev = NULL;
9156 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9157 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9158 /* remove */
9159 if (prev)
9160 prev->next = dl->next;
9161 else
9162 super->disks = dl->next;
9163 dl->next = NULL;
9164 __free_imsm_disk(dl);
1ade5cc1 9165 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9166 break;
9167 }
9168 prev = dl;
9169 }
9170 return 0;
9171}
9172
f21e18ca 9173static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9174
1a64be56
LM		 9175static int add_remove_disk_update(struct intel_super *super)
9176{
9177 int check_degraded = 0;
594dc1b8
JS		 9178 struct dl *disk;
9179
1a64be56
LM		 9180 /* add/remove some spares to/from the metadata/contrainer */
9181 while (super->disk_mgmt_list) {
9182 struct dl *disk_cfg;
9183
9184 disk_cfg = super->disk_mgmt_list;
9185 super->disk_mgmt_list = disk_cfg->next;
9186 disk_cfg->next = NULL;
9187
9188 if (disk_cfg->action == DISK_ADD) {
9189 disk_cfg->next = super->disks;
9190 super->disks = disk_cfg;
9191 check_degraded = 1;
1ade5cc1
N		 9192 dprintf("added %x:%x\n",
9193 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9194 } else if (disk_cfg->action == DISK_REMOVE) {
9195 dprintf("Disk remove action processed: %x.%x\n",
9196 disk_cfg->major, disk_cfg->minor);
9197 disk = get_disk_super(super,
9198 disk_cfg->major,
9199 disk_cfg->minor);
9200 if (disk) {
9201 /* store action status */
9202 disk->action = DISK_REMOVE;
9203 /* remove spare disks only */
9204 if (disk->index == -1) {
9205 remove_disk_super(super,
9206 disk_cfg->major,
9207 disk_cfg->minor);
91c97c54
MT		 9208 } else {
9209 disk_cfg->fd = disk->fd;
9210 disk->fd = -1;
1a64be56
LM		 9211 }
9212 }
9213 /* release allocate disk structure */
9214 __free_imsm_disk(disk_cfg);
9215 }
9216 }
9217 return check_degraded;
9218}
9219
a29911da
PC		 9220static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9221 struct intel_super *super,
9222 void ***space_list)
9223{
9224 struct intel_dev *id;
9225 void **tofree = NULL;
9226 int ret_val = 0;
9227
1ade5cc1 9228 dprintf("(enter)\n");
089f9d79 9229 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9230 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9231 return ret_val;
9232 }
089f9d79 9233 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9234 dprintf("imsm: Error: Memory is not allocated\n");
9235 return ret_val;
9236 }
9237
9238 for (id = super->devlist ; id; id = id->next) {
9239 if (id->index == (unsigned)u->subdev) {
9240 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9241 struct imsm_map *map;
9242 struct imsm_dev *new_dev =
9243 (struct imsm_dev *)*space_list;
238c0a71 9244 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9245 int to_state;
9246 struct dl *new_disk;
9247
9248 if (new_dev == NULL)
9249 return ret_val;
9250 *space_list = **space_list;
9251 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9252 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9253 if (migr_map) {
9254 dprintf("imsm: Error: migration in progress");
9255 return ret_val;
9256 }
9257
9258 to_state = map->map_state;
9259 if ((u->new_level == 5) && (map->raid_level == 0)) {
9260 map->num_members++;
9261 /* this should not happen */
9262 if (u->new_disks[0] < 0) {
9263 map->failed_disk_num =
9264 map->num_members - 1;
9265 to_state = IMSM_T_STATE_DEGRADED;
9266 } else
9267 to_state = IMSM_T_STATE_NORMAL;
9268 }
8e59f3d8 9269 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9270 if (u->new_level > -1)
9271 map->raid_level = u->new_level;
238c0a71 9272 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9273 if ((u->new_level == 5) &&
9274 (migr_map->raid_level == 0)) {
9275 int ord = map->num_members - 1;
9276 migr_map->num_members--;
9277 if (u->new_disks[0] < 0)
9278 ord |= IMSM_ORD_REBUILD;
9279 set_imsm_ord_tbl_ent(map,
9280 map->num_members - 1,
9281 ord);
9282 }
9283 id->dev = new_dev;
9284 tofree = (void **)dev;
9285
4bba0439
PC		 9286 /* update chunk size
9287 */
06fb291a
PB		 9288 if (u->new_chunksize > 0) {
9289 unsigned long long num_data_stripes;
9529d343
MD		 9290 struct imsm_map *dest_map =
9291 get_imsm_map(dev, MAP_0);
06fb291a 9292 int used_disks =
9529d343 9293 imsm_num_data_members(dest_map);
06fb291a
PB		 9294
9295 if (used_disks == 0)
9296 return ret_val;
9297
4bba0439
PC		 9298 map->blocks_per_strip =
9299 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9300 num_data_stripes =
fcc2c9da 9301 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9302 num_data_stripes /= map->blocks_per_strip;
9303 num_data_stripes /= map->num_domains;
9304 set_num_data_stripes(map, num_data_stripes);
9305 }
4bba0439 9306
44490938
MD		 9307 /* ensure blocks_per_member has valid value
9308 */
9309 set_blocks_per_member(map,
9310 per_dev_array_size(map) +
9311 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9312
a29911da
PC		 9313 /* add disk
9314 */
089f9d79
JS		 9315 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9316 migr_map->raid_level == map->raid_level)
a29911da
PC		 9317 goto skip_disk_add;
9318
9319 if (u->new_disks[0] >= 0) {
9320 /* use passes spare
9321 */
9322 new_disk = get_disk_super(super,
9323 major(u->new_disks[0]),
9324 minor(u->new_disks[0]));
7a862a02 9325 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9326 major(u->new_disks[0]),
9327 minor(u->new_disks[0]),
9328 new_disk, new_disk->index);
9329 if (new_disk == NULL)
9330 goto error_disk_add;
9331
9332 new_disk->index = map->num_members - 1;
9333 /* slot to fill in autolayout
9334 */
9335 new_disk->raiddisk = new_disk->index;
9336 new_disk->disk.status |= CONFIGURED_DISK;
9337 new_disk->disk.status &= ~SPARE_DISK;
9338 } else
9339 goto error_disk_add;
9340
9341skip_disk_add:
9342 *tofree = *space_list;
9343 /* calculate new size
9344 */
f3871fdc 9345 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9346
9347 ret_val = 1;
9348 }
9349 }
9350
9351 if (tofree)
9352 *space_list = tofree;
9353 return ret_val;
9354
9355error_disk_add:
9356 dprintf("Error: imsm: Cannot find disk.\n");
9357 return ret_val;
9358}
9359
f3871fdc
AK		 9360static int apply_size_change_update(struct imsm_update_size_change *u,
9361 struct intel_super *super)
9362{
9363 struct intel_dev *id;
9364 int ret_val = 0;
9365
1ade5cc1 9366 dprintf("(enter)\n");
089f9d79 9367 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9368 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9369 return ret_val;
9370 }
9371
9372 for (id = super->devlist ; id; id = id->next) {
9373 if (id->index == (unsigned)u->subdev) {
9374 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9375 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9376 int used_disks = imsm_num_data_members(map);
f3871fdc 9377 unsigned long long blocks_per_member;
06fb291a 9378 unsigned long long num_data_stripes;
44490938
MD		 9379 unsigned long long new_size_per_disk;
9380
9381 if (used_disks == 0)
9382 return 0;
f3871fdc
AK		 9383
9384 /* calculate new size
9385 */
44490938
MD		 9386 new_size_per_disk = u->new_size / used_disks;
9387 blocks_per_member = new_size_per_disk +
9388 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9389 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9390 map->blocks_per_strip;
9391 num_data_stripes /= map->num_domains;
9392 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9393 u->new_size, new_size_per_disk,
06fb291a 9394 num_data_stripes);
f3871fdc 9395 set_blocks_per_member(map, blocks_per_member);
06fb291a 9396 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9397 imsm_set_array_size(dev, u->new_size);
9398
9399 ret_val = 1;
9400 break;
9401 }
9402 }
9403
9404 return ret_val;
9405}
9406
061d7da3 9407static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9408 struct intel_super *super,
061d7da3
LO		 9409 struct active_array *active_array)
9410{
9411 struct imsm_super *mpb = super->anchor;
9412 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9413 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9414 struct imsm_map *migr_map;
9415 struct active_array *a;
9416 struct imsm_disk *disk;
9417 __u8 to_state;
9418 struct dl *dl;
9419 unsigned int found;
9420 int failed;
5961eeec 9421 int victim;
061d7da3 9422 int i;
5961eeec 9423 int second_map_created = 0;
061d7da3 9424
5961eeec 9425 for (; u; u = u->next) {
238c0a71 9426 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9427
5961eeec 9428 if (victim < 0)
9429 return 0;
061d7da3 9430
5961eeec 9431 for (dl = super->disks; dl; dl = dl->next)
9432 if (dl == u->dl)
9433 break;
061d7da3 9434
5961eeec 9435 if (!dl) {
7a862a02 9436 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9437 u->dl->index);
9438 return 0;
9439 }
061d7da3 9440
5961eeec 9441 /* count failures (excluding rebuilds and the victim)
9442 * to determine map[0] state
9443 */
9444 failed = 0;
9445 for (i = 0; i < map->num_members; i++) {
9446 if (i == u->slot)
9447 continue;
9448 disk = get_imsm_disk(super,
238c0a71 9449 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9450 if (!disk || is_failed(disk))
9451 failed++;
9452 }
061d7da3 9453
5961eeec 9454 /* adding a pristine spare, assign a new index */
9455 if (dl->index < 0) {
9456 dl->index = super->anchor->num_disks;
9457 super->anchor->num_disks++;
9458 }
9459 disk = &dl->disk;
9460 disk->status |= CONFIGURED_DISK;
9461 disk->status &= ~SPARE_DISK;
9462
9463 /* mark rebuild */
238c0a71 9464 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9465 if (!second_map_created) {
9466 second_map_created = 1;
9467 map->map_state = IMSM_T_STATE_DEGRADED;
9468 migrate(dev, super, to_state, MIGR_REBUILD);
9469 } else
9470 map->map_state = to_state;
238c0a71 9471 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9472 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9473 set_imsm_ord_tbl_ent(migr_map, u->slot,
9474 dl->index | IMSM_ORD_REBUILD);
9475
9476 /* update the family_num to mark a new container
9477 * generation, being careful to record the existing
9478 * family_num in orig_family_num to clean up after
9479 * earlier mdadm versions that neglected to set it.
9480 */
9481 if (mpb->orig_family_num == 0)
9482 mpb->orig_family_num = mpb->family_num;
9483 mpb->family_num += super->random;
9484
9485 /* count arrays using the victim in the metadata */
9486 found = 0;
9487 for (a = active_array; a ; a = a->next) {
9488 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9489 map = get_imsm_map(dev, MAP_0);
061d7da3 9490
5961eeec 9491 if (get_imsm_disk_slot(map, victim) >= 0)
9492 found++;
9493 }
061d7da3 9494
5961eeec 9495 /* delete the victim if it is no longer being
9496 * utilized anywhere
061d7da3 9497 */
5961eeec 9498 if (!found) {
9499 struct dl **dlp;
061d7da3 9500
5961eeec 9501 /* We know that 'manager' isn't touching anything,
9502 * so it is safe to delete
9503 */
9504 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9505 if ((*dlp)->index == victim)
9506 break;
5961eeec 9507
9508 /* victim may be on the missing list */
9509 if (!*dlp)
9510 for (dlp = &super->missing; *dlp;
9511 dlp = &(*dlp)->next)
9512 if ((*dlp)->index == victim)
9513 break;
9514 imsm_delete(super, dlp, victim);
9515 }
061d7da3
LO		 9516 }
9517
9518 return 1;
9519}
a29911da 9520
2e5dc010
N		 9521static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9522 struct intel_super *super,
9523 void ***space_list)
9524{
9525 struct dl *new_disk;
9526 struct intel_dev *id;
9527 int i;
9528 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9529 int disk_count = u->old_raid_disks;
2e5dc010
N		 9530 void **tofree = NULL;
9531 int devices_to_reshape = 1;
9532 struct imsm_super *mpb = super->anchor;
9533 int ret_val = 0;
d098291a 9534 unsigned int dev_id;
2e5dc010 9535
1ade5cc1 9536 dprintf("(enter)\n");
2e5dc010
N		 9537
9538 /* enable spares to use in array */
9539 for (i = 0; i < delta_disks; i++) {
9540 new_disk = get_disk_super(super,
9541 major(u->new_disks[i]),
9542 minor(u->new_disks[i]));
7a862a02 9543 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9544 major(u->new_disks[i]), minor(u->new_disks[i]),
9545 new_disk, new_disk->index);
089f9d79
JS		 9546 if (new_disk == NULL ||
9547 (new_disk->index >= 0 &&
9548 new_disk->index < u->old_raid_disks))
2e5dc010 9549 goto update_reshape_exit;
ee4beede 9550 new_disk->index = disk_count++;
2e5dc010
N		 9551 /* slot to fill in autolayout
9552 */
9553 new_disk->raiddisk = new_disk->index;
9554 new_disk->disk.status |=
9555 CONFIGURED_DISK;
9556 new_disk->disk.status &= ~SPARE_DISK;
9557 }
9558
ed7333bd
AK		 9559 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9560 mpb->num_raid_devs);
2e5dc010
N		 9561 /* manage changes in volume
9562 */
d098291a 9563 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9564 void **sp = *space_list;
9565 struct imsm_dev *newdev;
9566 struct imsm_map *newmap, *oldmap;
9567
d098291a
AK		 9568 for (id = super->devlist ; id; id = id->next) {
9569 if (id->index == dev_id)
9570 break;
9571 }
9572 if (id == NULL)
9573 break;
2e5dc010
N		 9574 if (!sp)
9575 continue;
9576 *space_list = *sp;
9577 newdev = (void*)sp;
9578 /* Copy the dev, but not (all of) the map */
9579 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9580 oldmap = get_imsm_map(id->dev, MAP_0);
9581 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9582 /* Copy the current map */
9583 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9584 /* update one device only
9585 */
9586 if (devices_to_reshape) {
ed7333bd
AK		 9587 dprintf("imsm: modifying subdev: %i\n",
9588 id->index);
2e5dc010
N		 9589 devices_to_reshape--;
9590 newdev->vol.migr_state = 1;
9591 newdev->vol.curr_migr_unit = 0;
ea672ee1 9592 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9593 newmap->num_members = u->new_raid_disks;
9594 for (i = 0; i < delta_disks; i++) {
9595 set_imsm_ord_tbl_ent(newmap,
9596 u->old_raid_disks + i,
9597 u->old_raid_disks + i);
9598 }
9599 /* New map is correct, now need to save old map
9600 */
238c0a71 9601 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9602 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9603
f3871fdc 9604 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9605 }
9606
9607 sp = (void **)id->dev;
9608 id->dev = newdev;
9609 *sp = tofree;
9610 tofree = sp;
8e59f3d8
AK		 9611
9612 /* Clear migration record */
9613 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9614 }
819bc634
AK		 9615 if (tofree)
9616 *space_list = tofree;
2e5dc010
N		 9617 ret_val = 1;
9618
9619update_reshape_exit:
9620
9621 return ret_val;
9622}
9623
bb025c2f 9624static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9625 struct intel_super *super,
9626 void ***space_list)
bb025c2f
KW		 9627{
9628 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9629 struct intel_dev *dv;
9630 struct imsm_dev *dev_new;
bb025c2f
KW		 9631 struct imsm_map *map;
9632 struct dl *dm, *du;
8ca6df95 9633 int i;
bb025c2f
KW		 9634
9635 for (dv = super->devlist; dv; dv = dv->next)
9636 if (dv->index == (unsigned int)u->subarray) {
9637 dev = dv->dev;
9638 break;
9639 }
9640
9641 if (dev == NULL)
9642 return 0;
9643
238c0a71 9644 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9645
9646 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9647 unsigned long long num_data_stripes;
9648
43d5ec18 9649 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9650 if (imsm_count_failed(super, dev, MAP_0) !=
9651 (map->num_members / 2))
43d5ec18
KW		 9652 return 0;
9653
bb025c2f
KW		 9654 /* iterate through devices to mark removed disks as spare */
9655 for (dm = super->disks; dm; dm = dm->next) {
9656 if (dm->disk.status & FAILED_DISK) {
9657 int idx = dm->index;
9658 /* update indexes on the disk list */
9659/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9660 the index values will end up being correct.... NB */
9661 for (du = super->disks; du; du = du->next)
9662 if (du->index > idx)
9663 du->index--;
9664 /* mark as spare disk */
a8619d23 9665 mark_spare(dm);
bb025c2f
KW		 9666 }
9667 }
bb025c2f
KW		 9668 /* update map */
9669 map->num_members = map->num_members / 2;
9670 map->map_state = IMSM_T_STATE_NORMAL;
9671 map->num_domains = 1;
9672 map->raid_level = 0;
9673 map->failed_disk_num = -1;
4a353e6e
RS		 9674 num_data_stripes = imsm_dev_size(dev) / 2;
9675 num_data_stripes /= map->blocks_per_strip;
9676 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9677 }
9678
8ca6df95
KW		 9679 if (u->direction == R0_TO_R10) {
9680 void **space;
4a353e6e
RS		 9681 unsigned long long num_data_stripes;
9682
8ca6df95
KW		 9683 /* update slots in current disk list */
9684 for (dm = super->disks; dm; dm = dm->next) {
9685 if (dm->index >= 0)
9686 dm->index *= 2;
9687 }
9688 /* create new *missing* disks */
9689 for (i = 0; i < map->num_members; i++) {
9690 space = *space_list;
9691 if (!space)
9692 continue;
9693 *space_list = *space;
9694 du = (void *)space;
9695 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9696 du->fd = -1;
9697 du->minor = 0;
9698 du->major = 0;
9699 du->index = (i * 2) + 1;
9700 sprintf((char *)du->disk.serial,
9701 " MISSING_%d", du->index);
9702 sprintf((char *)du->serial,
9703 "MISSING_%d", du->index);
9704 du->next = super->missing;
9705 super->missing = du;
9706 }
9707 /* create new dev and map */
9708 space = *space_list;
9709 if (!space)
9710 return 0;
9711 *space_list = *space;
9712 dev_new = (void *)space;
9713 memcpy(dev_new, dev, sizeof(*dev));
9714 /* update new map */
238c0a71 9715 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9716 map->num_members = map->num_members * 2;
1a2487c2 9717 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9718 map->num_domains = 2;
9719 map->raid_level = 1;
4a353e6e
RS		 9720 num_data_stripes = imsm_dev_size(dev) / 2;
9721 num_data_stripes /= map->blocks_per_strip;
9722 num_data_stripes /= map->num_domains;
9723 set_num_data_stripes(map, num_data_stripes);
9724
8ca6df95
KW		 9725 /* replace dev<->dev_new */
9726 dv->dev = dev_new;
9727 }
bb025c2f
KW		 9728 /* update disk order table */
9729 for (du = super->disks; du; du = du->next)
9730 if (du->index >= 0)
9731 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9732 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9733 if (du->index >= 0) {
9734 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9735 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9736 }
bb025c2f
KW		 9737
9738 return 1;
9739}
9740
e8319a19
DW		 9741static void imsm_process_update(struct supertype *st,
9742 struct metadata_update *update)
9743{
9744 /**
9745 * crack open the metadata_update envelope to find the update record
9746 * update can be one of:
d195167d
AK		 9747 * update_reshape_container_disks - all the arrays in the container
9748 * are being reshaped to have more devices. We need to mark
9749 * the arrays for general migration and convert selected spares
9750 * into active devices.
9751 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9752 * device in an array, update the disk_ord_tbl. If this disk is
9753 * present in all member arrays then also clear the SPARE_DISK
9754 * flag
d195167d
AK		 9755 * update_create_array
9756 * update_kill_array
9757 * update_rename_array
9758 * update_add_remove_disk
e8319a19
DW		 9759 */
9760 struct intel_super *super = st->sb;
4d7b1503 9761 struct imsm_super *mpb;
e8319a19
DW		 9762 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9763
4d7b1503
DW		 9764 /* update requires a larger buf but the allocation failed */
9765 if (super->next_len && !super->next_buf) {
9766 super->next_len = 0;
9767 return;
9768 }
9769
9770 if (super->next_buf) {
9771 memcpy(super->next_buf, super->buf, super->len);
9772 free(super->buf);
9773 super->len = super->next_len;
9774 super->buf = super->next_buf;
9775
9776 super->next_len = 0;
9777 super->next_buf = NULL;
9778 }
9779
9780 mpb = super->anchor;
9781
e8319a19 9782 switch (type) {
0ec5d470
AK		 9783 case update_general_migration_checkpoint: {
9784 struct intel_dev *id;
9785 struct imsm_update_general_migration_checkpoint *u =
9786 (void *)update->buf;
9787
1ade5cc1 9788 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9789
9790 /* find device under general migration */
9791 for (id = super->devlist ; id; id = id->next) {
9792 if (is_gen_migration(id->dev)) {
9793 id->dev->vol.curr_migr_unit =
9794 __cpu_to_le32(u->curr_migr_unit);
9795 super->updates_pending++;
9796 }
9797 }
9798 break;
9799 }
bb025c2f
KW		 9800 case update_takeover: {
9801 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9802 if (apply_takeover_update(u, super, &update->space_list)) {
9803 imsm_update_version_info(super);
bb025c2f 9804 super->updates_pending++;
1a2487c2 9805 }
bb025c2f
KW		 9806 break;
9807 }
9808
78b10e66 9809 case update_reshape_container_disks: {
d195167d 9810 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9811 if (apply_reshape_container_disks_update(
9812 u, super, &update->space_list))
9813 super->updates_pending++;
78b10e66
N		 9814 break;
9815 }
48c5303a 9816 case update_reshape_migration: {
a29911da
PC		 9817 struct imsm_update_reshape_migration *u = (void *)update->buf;
9818 if (apply_reshape_migration_update(
9819 u, super, &update->space_list))
9820 super->updates_pending++;
48c5303a
PC		 9821 break;
9822 }
f3871fdc
AK		 9823 case update_size_change: {
9824 struct imsm_update_size_change *u = (void *)update->buf;
9825 if (apply_size_change_update(u, super))
9826 super->updates_pending++;
9827 break;
9828 }
e8319a19 9829 case update_activate_spare: {
1011e834 9830 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9831 if (apply_update_activate_spare(u, super, st->arrays))
9832 super->updates_pending++;
8273f55e
DW		 9833 break;
9834 }
9835 case update_create_array: {
9836 /* someone wants to create a new array, we need to be aware of
9837 * a few races/collisions:
9838 * 1/ 'Create' called by two separate instances of mdadm
9839 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9840 * devices that have since been assimilated via
9841 * activate_spare.
9842 * In the event this update can not be carried out mdadm will
9843 * (FIX ME) notice that its update did not take hold.
9844 */
9845 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9846 struct intel_dev *dv;
8273f55e
DW		 9847 struct imsm_dev *dev;
9848 struct imsm_map *map, *new_map;
9849 unsigned long long start, end;
9850 unsigned long long new_start, new_end;
9851 int i;
54c2c1ea
DW		 9852 struct disk_info *inf;
9853 struct dl *dl;
8273f55e
DW		 9854
9855 /* handle racing creates: first come first serve */
9856 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9857 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9858 goto create_error;
8273f55e
DW		 9859 }
9860
9861 /* check update is next in sequence */
9862 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9863 dprintf("can not create array %d expected index %d\n",
9864 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9865 goto create_error;
8273f55e
DW		 9866 }
9867
238c0a71 9868 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9869 new_start = pba_of_lba0(new_map);
44490938 9870 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9871 inf = get_disk_info(u);
8273f55e
DW		 9872
9873 /* handle activate_spare versus create race:
9874 * check to make sure that overlapping arrays do not include
9875 * overalpping disks
9876 */
9877 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9878 dev = get_imsm_dev(super, i);
238c0a71 9879 map = get_imsm_map(dev, MAP_0);
5551b113 9880 start = pba_of_lba0(map);
44490938 9881 end = start + per_dev_array_size(map);
8273f55e
DW		 9882 if ((new_start >= start && new_start <= end) ||
9883 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9884 /* overlap */;
9885 else
9886 continue;
9887
9888 if (disks_overlap(super, i, u)) {
1ade5cc1 9889 dprintf("arrays overlap\n");
ba2de7ba 9890 goto create_error;
8273f55e
DW		 9891 }
9892 }
8273f55e 9893
949c47a0
DW		 9894 /* check that prepare update was successful */
9895 if (!update->space) {
1ade5cc1 9896 dprintf("prepare update failed\n");
ba2de7ba 9897 goto create_error;
949c47a0
DW		 9898 }
9899
54c2c1ea
DW		 9900 /* check that all disks are still active before committing
9901 * changes. FIXME: could we instead handle this by creating a
9902 * degraded array? That's probably not what the user expects,
9903 * so better to drop this update on the floor.
9904 */
9905 for (i = 0; i < new_map->num_members; i++) {
9906 dl = serial_to_dl(inf[i].serial, super);
9907 if (!dl) {
1ade5cc1 9908 dprintf("disk disappeared\n");
ba2de7ba 9909 goto create_error;
54c2c1ea 9910 }
949c47a0
DW		 9911 }
9912
8273f55e 9913 super->updates_pending++;
54c2c1ea
DW		 9914
9915 /* convert spares to members and fixup ord_tbl */
9916 for (i = 0; i < new_map->num_members; i++) {
9917 dl = serial_to_dl(inf[i].serial, super);
9918 if (dl->index == -1) {
9919 dl->index = mpb->num_disks;
9920 mpb->num_disks++;
9921 dl->disk.status |= CONFIGURED_DISK;
9922 dl->disk.status &= ~SPARE_DISK;
9923 }
9924 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9925 }
9926
ba2de7ba
DW		 9927 dv = update->space;
9928 dev = dv->dev;
949c47a0
DW		 9929 update->space = NULL;
9930 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9931 dv->index = u->dev_idx;
9932 dv->next = super->devlist;
9933 super->devlist = dv;
8273f55e 9934 mpb->num_raid_devs++;
8273f55e 9935
4d1313e9 9936 imsm_update_version_info(super);
8273f55e 9937 break;
ba2de7ba
DW		 9938 create_error:
9939 /* mdmon knows how to release update->space, but not
9940 * ((struct intel_dev *) update->space)->dev
9941 */
9942 if (update->space) {
9943 dv = update->space;
9944 free(dv->dev);
9945 }
8273f55e 9946 break;
e8319a19 9947 }
33414a01
DW		 9948 case update_kill_array: {
9949 struct imsm_update_kill_array *u = (void *) update->buf;
9950 int victim = u->dev_idx;
9951 struct active_array *a;
9952 struct intel_dev **dp;
9953 struct imsm_dev *dev;
9954
9955 /* sanity check that we are not affecting the uuid of
9956 * active arrays, or deleting an active array
9957 *
9958 * FIXME when immutable ids are available, but note that
9959 * we'll also need to fixup the invalidated/active
9960 * subarray indexes in mdstat
9961 */
9962 for (a = st->arrays; a; a = a->next)
9963 if (a->info.container_member >= victim)
9964 break;
9965 /* by definition if mdmon is running at least one array
9966 * is active in the container, so checking
9967 * mpb->num_raid_devs is just extra paranoia
9968 */
9969 dev = get_imsm_dev(super, victim);
9970 if (a || !dev || mpb->num_raid_devs == 1) {
9971 dprintf("failed to delete subarray-%d\n", victim);
9972 break;
9973 }
9974
9975 for (dp = &super->devlist; *dp;)
f21e18ca 9976 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9977 *dp = (*dp)->next;
9978 } else {
f21e18ca 9979 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9980 (*dp)->index--;
9981 dp = &(*dp)->next;
9982 }
9983 mpb->num_raid_devs--;
9984 super->updates_pending++;
9985 break;
9986 }
aa534678
DW		 9987 case update_rename_array: {
9988 struct imsm_update_rename_array *u = (void *) update->buf;
9989 char name[MAX_RAID_SERIAL_LEN+1];
9990 int target = u->dev_idx;
9991 struct active_array *a;
9992 struct imsm_dev *dev;
9993
9994 /* sanity check that we are not affecting the uuid of
9995 * an active array
9996 */
40659392 9997 memset(name, 0, sizeof(name));
aa534678
DW		 9998 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9999 name[MAX_RAID_SERIAL_LEN] = '\0';
10000 for (a = st->arrays; a; a = a->next)
10001 if (a->info.container_member == target)
10002 break;
10003 dev = get_imsm_dev(super, u->dev_idx);
10004 if (a || !dev || !check_name(super, name, 1)) {
10005 dprintf("failed to rename subarray-%d\n", target);
10006 break;
10007 }
10008
40659392 10009 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10010 super->updates_pending++;
10011 break;
10012 }
1a64be56 10013 case update_add_remove_disk: {
43dad3d6 10014 /* we may be able to repair some arrays if disks are
095b8088 10015 * being added, check the status of add_remove_disk
1a64be56
LM		 10016 * if discs has been added.
10017 */
10018 if (add_remove_disk_update(super)) {
43dad3d6 10019 struct active_array *a;
072b727f
DW		 10020
10021 super->updates_pending++;
1a64be56 10022 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10023 a->check_degraded = 1;
10024 }
43dad3d6 10025 break;
e8319a19 10026 }
bbab0940
TM		 10027 case update_prealloc_badblocks_mem:
10028 break;
e6e9dd3f
AP		 10029 case update_rwh_policy: {
10030 struct imsm_update_rwh_policy *u = (void *)update->buf;
10031 int target = u->dev_idx;
10032 struct imsm_dev *dev = get_imsm_dev(super, target);
10033 if (!dev) {
10034 dprintf("could not find subarray-%d\n", target);
10035 break;
10036 }
10037
10038 if (dev->rwh_policy != u->new_policy) {
10039 dev->rwh_policy = u->new_policy;
10040 super->updates_pending++;
10041 }
10042 break;
10043 }
1a64be56 10044 default:
ebf3be99 10045 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10046 }
e8319a19 10047}
88758e9d 10048
bc0b9d34
PC		 10049static struct mdinfo *get_spares_for_grow(struct supertype *st);
10050
5fe6f031
N		 10051static int imsm_prepare_update(struct supertype *st,
10052 struct metadata_update *update)
8273f55e 10053{
949c47a0 10054 /**
4d7b1503
DW		 10055 * Allocate space to hold new disk entries, raid-device entries or a new
10056 * mpb if necessary. The manager synchronously waits for updates to
10057 * complete in the monitor, so new mpb buffers allocated here can be
10058 * integrated by the monitor thread without worrying about live pointers
10059 * in the manager thread.
8273f55e 10060 */
095b8088 10061 enum imsm_update_type type;
4d7b1503 10062 struct intel_super *super = st->sb;
f36a9ecd 10063 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10064 struct imsm_super *mpb = super->anchor;
10065 size_t buf_len;
10066 size_t len = 0;
949c47a0 10067
095b8088
N		 10068 if (update->len < (int)sizeof(type))
10069 return 0;
10070
10071 type = *(enum imsm_update_type *) update->buf;
10072
949c47a0 10073 switch (type) {
0ec5d470 10074 case update_general_migration_checkpoint:
095b8088
N		 10075 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10076 return 0;
1ade5cc1 10077 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10078 break;
abedf5fc
KW		 10079 case update_takeover: {
10080 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10081 if (update->len < (int)sizeof(*u))
10082 return 0;
abedf5fc
KW		 10083 if (u->direction == R0_TO_R10) {
10084 void **tail = (void **)&update->space_list;
10085 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10086 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10087 int num_members = map->num_members;
10088 void *space;
10089 int size, i;
abedf5fc
KW		 10090 /* allocate memory for added disks */
10091 for (i = 0; i < num_members; i++) {
10092 size = sizeof(struct dl);
503975b9 10093 space = xmalloc(size);
abedf5fc
KW		 10094 *tail = space;
10095 tail = space;
10096 *tail = NULL;
10097 }
10098 /* allocate memory for new device */
10099 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10100 (num_members * sizeof(__u32));
503975b9
N		 10101 space = xmalloc(size);
10102 *tail = space;
10103 tail = space;
10104 *tail = NULL;
10105 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10106 }
10107
10108 break;
10109 }
78b10e66 10110 case update_reshape_container_disks: {
d195167d
AK		 10111 /* Every raid device in the container is about to
10112 * gain some more devices, and we will enter a
10113 * reconfiguration.
10114 * So each 'imsm_map' will be bigger, and the imsm_vol
10115 * will now hold 2 of them.
10116 * Thus we need new 'struct imsm_dev' allocations sized
10117 * as sizeof_imsm_dev but with more devices in both maps.
10118 */
10119 struct imsm_update_reshape *u = (void *)update->buf;
10120 struct intel_dev *dl;
10121 void **space_tail = (void**)&update->space_list;
10122
095b8088
N		 10123 if (update->len < (int)sizeof(*u))
10124 return 0;
10125
1ade5cc1 10126 dprintf("for update_reshape\n");
d195167d
AK		 10127
10128 for (dl = super->devlist; dl; dl = dl->next) {
10129 int size = sizeof_imsm_dev(dl->dev, 1);
10130 void *s;
d677e0b8
AK		 10131 if (u->new_raid_disks > u->old_raid_disks)
10132 size += sizeof(__u32)*2*
10133 (u->new_raid_disks - u->old_raid_disks);
503975b9 10134 s = xmalloc(size);
d195167d
AK		 10135 *space_tail = s;
10136 space_tail = s;
10137 *space_tail = NULL;
10138 }
10139
10140 len = disks_to_mpb_size(u->new_raid_disks);
10141 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10142 break;
10143 }
48c5303a 10144 case update_reshape_migration: {
bc0b9d34
PC		 10145 /* for migration level 0->5 we need to add disks
10146 * so the same as for container operation we will copy
10147 * device to the bigger location.
10148 * in memory prepared device and new disk area are prepared
10149 * for usage in process update
10150 */
10151 struct imsm_update_reshape_migration *u = (void *)update->buf;
10152 struct intel_dev *id;
10153 void **space_tail = (void **)&update->space_list;
10154 int size;
10155 void *s;
10156 int current_level = -1;
10157
095b8088
N		 10158 if (update->len < (int)sizeof(*u))
10159 return 0;
10160
1ade5cc1 10161 dprintf("for update_reshape\n");
bc0b9d34
PC		 10162
10163 /* add space for bigger array in update
10164 */
10165 for (id = super->devlist; id; id = id->next) {
10166 if (id->index == (unsigned)u->subdev) {
10167 size = sizeof_imsm_dev(id->dev, 1);
10168 if (u->new_raid_disks > u->old_raid_disks)
10169 size += sizeof(__u32)*2*
10170 (u->new_raid_disks - u->old_raid_disks);
503975b9 10171 s = xmalloc(size);
bc0b9d34
PC		 10172 *space_tail = s;
10173 space_tail = s;
10174 *space_tail = NULL;
10175 break;
10176 }
10177 }
10178 if (update->space_list == NULL)
10179 break;
10180
10181 /* add space for disk in update
10182 */
10183 size = sizeof(struct dl);
503975b9 10184 s = xmalloc(size);
bc0b9d34
PC		 10185 *space_tail = s;
10186 space_tail = s;
10187 *space_tail = NULL;
10188
10189 /* add spare device to update
10190 */
10191 for (id = super->devlist ; id; id = id->next)
10192 if (id->index == (unsigned)u->subdev) {
10193 struct imsm_dev *dev;
10194 struct imsm_map *map;
10195
10196 dev = get_imsm_dev(super, u->subdev);
238c0a71 10197 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10198 current_level = map->raid_level;
10199 break;
10200 }
089f9d79 10201 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10202 struct mdinfo *spares;
10203
10204 spares = get_spares_for_grow(st);
10205 if (spares) {
10206 struct dl *dl;
10207 struct mdinfo *dev;
10208
10209 dev = spares->devs;
10210 if (dev) {
10211 u->new_disks[0] =
10212 makedev(dev->disk.major,
10213 dev->disk.minor);
10214 dl = get_disk_super(super,
10215 dev->disk.major,
10216 dev->disk.minor);
10217 dl->index = u->old_raid_disks;
10218 dev = dev->next;
10219 }
10220 sysfs_free(spares);
10221 }
10222 }
10223 len = disks_to_mpb_size(u->new_raid_disks);
10224 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10225 break;
10226 }
f3871fdc 10227 case update_size_change: {
095b8088
N		 10228 if (update->len < (int)sizeof(struct imsm_update_size_change))
10229 return 0;
10230 break;
10231 }
10232 case update_activate_spare: {
10233 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10234 return 0;
f3871fdc
AK		 10235 break;
10236 }
949c47a0
DW		 10237 case update_create_array: {
10238 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10239 struct intel_dev *dv;
54c2c1ea 10240 struct imsm_dev *dev = &u->dev;
238c0a71 10241 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10242 struct dl *dl;
10243 struct disk_info *inf;
10244 int i;
10245 int activate = 0;
949c47a0 10246
095b8088
N		 10247 if (update->len < (int)sizeof(*u))
10248 return 0;
10249
54c2c1ea
DW		 10250 inf = get_disk_info(u);
10251 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10252 /* allocate a new super->devlist entry */
503975b9
N		 10253 dv = xmalloc(sizeof(*dv));
10254 dv->dev = xmalloc(len);
10255 update->space = dv;
949c47a0 10256
54c2c1ea
DW		 10257 /* count how many spares will be converted to members */
10258 for (i = 0; i < map->num_members; i++) {
10259 dl = serial_to_dl(inf[i].serial, super);
10260 if (!dl) {
10261 /* hmm maybe it failed?, nothing we can do about
10262 * it here
10263 */
10264 continue;
10265 }
10266 if (count_memberships(dl, super) == 0)
10267 activate++;
10268 }
10269 len += activate * sizeof(struct imsm_disk);
949c47a0 10270 break;
095b8088
N		 10271 }
10272 case update_kill_array: {
10273 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10274 return 0;
949c47a0
DW		 10275 break;
10276 }
095b8088
N		 10277 case update_rename_array: {
10278 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10279 return 0;
10280 break;
10281 }
10282 case update_add_remove_disk:
10283 /* no update->len needed */
10284 break;
bbab0940
TM		 10285 case update_prealloc_badblocks_mem:
10286 super->extra_space += sizeof(struct bbm_log) -
10287 get_imsm_bbm_log_size(super->bbm_log);
10288 break;
e6e9dd3f
AP		 10289 case update_rwh_policy: {
10290 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10291 return 0;
10292 break;
10293 }
095b8088
N		 10294 default:
10295 return 0;
949c47a0 10296 }
8273f55e 10297
4d7b1503
DW		 10298 /* check if we need a larger metadata buffer */
10299 if (super->next_buf)
10300 buf_len = super->next_len;
10301 else
10302 buf_len = super->len;
10303
bbab0940 10304 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10305 /* ok we need a larger buf than what is currently allocated
10306 * if this allocation fails process_update will notice that
10307 * ->next_len is set and ->next_buf is NULL
10308 */
bbab0940
TM		 10309 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10310 super->extra_space + len, sector_size);
4d7b1503
DW		 10311 if (super->next_buf)
10312 free(super->next_buf);
10313
10314 super->next_len = buf_len;
f36a9ecd 10315 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10316 memset(super->next_buf, 0, buf_len);
10317 else
4d7b1503
DW		 10318 super->next_buf = NULL;
10319 }
5fe6f031 10320 return 1;
8273f55e
DW		 10321}
10322
ae6aad82 10323/* must be called while manager is quiesced */
f21e18ca 10324static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10325{
10326 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10327 struct dl *iter;
10328 struct imsm_dev *dev;
10329 struct imsm_map *map;
4c9e8c1e 10330 unsigned int i, j, num_members;
fb12a745 10331 __u32 ord, ord_map0;
4c9e8c1e 10332 struct bbm_log *log = super->bbm_log;
ae6aad82 10333
1ade5cc1 10334 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10335
10336 /* shift all indexes down one */
10337 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10338 if (iter->index > (int)index)
ae6aad82 10339 iter->index--;
47ee5a45 10340 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10341 if (iter->index > (int)index)
47ee5a45 10342 iter->index--;
ae6aad82
DW		 10343
10344 for (i = 0; i < mpb->num_raid_devs; i++) {
10345 dev = get_imsm_dev(super, i);
238c0a71 10346 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10347 num_members = map->num_members;
10348 for (j = 0; j < num_members; j++) {
10349 /* update ord entries being careful not to propagate
10350 * ord-flags to the first map
10351 */
238c0a71 10352 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10353 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10354
24565c9a
DW		 10355 if (ord_to_idx(ord) <= index)
10356 continue;
ae6aad82 10357
238c0a71 10358 map = get_imsm_map(dev, MAP_0);
fb12a745 10359 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10360 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10361 if (map)
10362 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10363 }
10364 }
10365
4c9e8c1e
TM		 10366 for (i = 0; i < log->entry_count; i++) {
10367 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10368
10369 if (entry->disk_ordinal <= index)
10370 continue;
10371 entry->disk_ordinal--;
10372 }
10373
ae6aad82
DW		 10374 mpb->num_disks--;
10375 super->updates_pending++;
24565c9a
DW		 10376 if (*dlp) {
10377 struct dl *dl = *dlp;
10378
10379 *dlp = (*dlp)->next;
10380 __free_imsm_disk(dl);
10381 }
ae6aad82 10382}
9a717282
AK		 10383
10384static void close_targets(int *targets, int new_disks)
10385{
10386 int i;
10387
10388 if (!targets)
10389 return;
10390
10391 for (i = 0; i < new_disks; i++) {
10392 if (targets[i] >= 0) {
10393 close(targets[i]);
10394 targets[i] = -1;
10395 }
10396 }
10397}
10398
10399static int imsm_get_allowed_degradation(int level, int raid_disks,
10400 struct intel_super *super,
10401 struct imsm_dev *dev)
10402{
10403 switch (level) {
bf5cf7c7 10404 case 1:
9a717282
AK		 10405 case 10:{
10406 int ret_val = 0;
10407 struct imsm_map *map;
10408 int i;
10409
10410 ret_val = raid_disks/2;
10411 /* check map if all disks pairs not failed
10412 * in both maps
10413 */
238c0a71 10414 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10415 for (i = 0; i < ret_val; i++) {
10416 int degradation = 0;
10417 if (get_imsm_disk(super, i) == NULL)
10418 degradation++;
10419 if (get_imsm_disk(super, i + 1) == NULL)
10420 degradation++;
10421 if (degradation == 2)
10422 return 0;
10423 }
238c0a71 10424 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10425 /* if there is no second map
10426 * result can be returned
10427 */
10428 if (map == NULL)
10429 return ret_val;
10430 /* check degradation in second map
10431 */
10432 for (i = 0; i < ret_val; i++) {
10433 int degradation = 0;
10434 if (get_imsm_disk(super, i) == NULL)
10435 degradation++;
10436 if (get_imsm_disk(super, i + 1) == NULL)
10437 degradation++;
10438 if (degradation == 2)
10439 return 0;
10440 }
10441 return ret_val;
10442 }
10443 case 5:
10444 return 1;
10445 case 6:
10446 return 2;
10447 default:
10448 return 0;
10449 }
10450}
10451
687629c2
AK		 10452/*******************************************************************************
10453 * Function: open_backup_targets
10454 * Description: Function opens file descriptors for all devices given in
10455 * info->devs
10456 * Parameters:
10457 * info : general array info
10458 * raid_disks : number of disks
10459 * raid_fds : table of device's file descriptors
9a717282
AK		 10460 * super : intel super for raid10 degradation check
10461 * dev : intel device for raid10 degradation check
687629c2
AK		 10462 * Returns:
10463 * 0 : success
10464 * -1 : fail
10465 ******************************************************************************/
9a717282
AK		 10466int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10467 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10468{
10469 struct mdinfo *sd;
f627f5ad 10470 int i;
9a717282 10471 int opened = 0;
f627f5ad
AK		 10472
10473 for (i = 0; i < raid_disks; i++)
10474 raid_fds[i] = -1;
687629c2
AK		 10475
10476 for (sd = info->devs ; sd ; sd = sd->next) {
10477 char *dn;
10478
10479 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10480 dprintf("disk is faulty!!\n");
10481 continue;
10482 }
10483
089f9d79 10484 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10485 continue;
10486
10487 dn = map_dev(sd->disk.major,
10488 sd->disk.minor, 1);
10489 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10490 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10491 pr_err("cannot open component\n");
9a717282 10492 continue;
687629c2 10493 }
9a717282
AK		 10494 opened++;
10495 }
10496 /* check if maximum array degradation level is not exceeded
10497 */
10498 if ((raid_disks - opened) >
089f9d79
JS		 10499 imsm_get_allowed_degradation(info->new_level, raid_disks,
10500 super, dev)) {
e12b3daa 10501 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10502 close_targets(raid_fds, raid_disks);
10503 return -2;
687629c2
AK		 10504 }
10505 return 0;
10506}
10507
d31ad643
PB		 10508/*******************************************************************************
10509 * Function: validate_container_imsm
10510 * Description: This routine validates container after assemble,
10511 * eg. if devices in container are under the same controller.
10512 *
10513 * Parameters:
10514 * info : linked list with info about devices used in array
10515 * Returns:
10516 * 1 : HBA mismatch
10517 * 0 : Success
10518 ******************************************************************************/
10519int validate_container_imsm(struct mdinfo *info)
10520{
6b781d33
AP		 10521 if (check_env("IMSM_NO_PLATFORM"))
10522 return 0;
d31ad643 10523
6b781d33
AP		 10524 struct sys_dev *idev;
10525 struct sys_dev *hba = NULL;
10526 struct sys_dev *intel_devices = find_intel_devices();
10527 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10528 info->disk.minor));
10529
10530 for (idev = intel_devices; idev; idev = idev->next) {
10531 if (dev_path && strstr(dev_path, idev->path)) {
10532 hba = idev;
10533 break;
d31ad643 10534 }
6b781d33
AP		 10535 }
10536 if (dev_path)
d31ad643
PB		 10537 free(dev_path);
10538
6b781d33
AP		 10539 if (!hba) {
10540 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10541 devid2kname(makedev(info->disk.major, info->disk.minor)));
10542 return 1;
10543 }
10544
10545 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10546 struct mdinfo *dev;
10547
10548 for (dev = info->next; dev; dev = dev->next) {
10549 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10550
10551 struct sys_dev *hba2 = NULL;
10552 for (idev = intel_devices; idev; idev = idev->next) {
10553 if (dev_path && strstr(dev_path, idev->path)) {
10554 hba2 = idev;
10555 break;
d31ad643
PB		 10556 }
10557 }
6b781d33
AP		 10558 if (dev_path)
10559 free(dev_path);
10560
10561 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10562 get_orom_by_device_id(hba2->dev_id);
10563
10564 if (hba2 && hba->type != hba2->type) {
10565 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10566 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10567 return 1;
10568 }
10569
07cb1e57 10570 if (orom != orom2) {
6b781d33
AP		 10571 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10572 " This operation is not supported and can lead to data loss.\n");
10573 return 1;
10574 }
10575
10576 if (!orom) {
10577 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10578 " This operation is not supported and can lead to data loss.\n");
10579 return 1;
10580 }
d31ad643 10581 }
6b781d33 10582
d31ad643
PB		 10583 return 0;
10584}
32141c17 10585
6f50473f
TM		 10586/*******************************************************************************
10587* Function: imsm_record_badblock
10588* Description: This routine stores new bad block record in BBM log
10589*
10590* Parameters:
10591* a : array containing a bad block
10592* slot : disk number containing a bad block
10593* sector : bad block sector
10594* length : bad block sectors range
10595* Returns:
10596* 1 : Success
10597* 0 : Error
10598******************************************************************************/
10599static int imsm_record_badblock(struct active_array *a, int slot,
10600 unsigned long long sector, int length)
10601{
10602 struct intel_super *super = a->container->sb;
10603 int ord;
10604 int ret;
10605
10606 ord = imsm_disk_slot_to_ord(a, slot);
10607 if (ord < 0)
10608 return 0;
10609
10610 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10611 length);
10612 if (ret)
10613 super->updates_pending++;
10614
10615 return ret;
10616}
c07a5a4f
TM		 10617/*******************************************************************************
10618* Function: imsm_clear_badblock
10619* Description: This routine clears bad block record from BBM log
10620*
10621* Parameters:
10622* a : array containing a bad block
10623* slot : disk number containing a bad block
10624* sector : bad block sector
10625* length : bad block sectors range
10626* Returns:
10627* 1 : Success
10628* 0 : Error
10629******************************************************************************/
10630static int imsm_clear_badblock(struct active_array *a, int slot,
10631 unsigned long long sector, int length)
10632{
10633 struct intel_super *super = a->container->sb;
10634 int ord;
10635 int ret;
10636
10637 ord = imsm_disk_slot_to_ord(a, slot);
10638 if (ord < 0)
10639 return 0;
10640
10641 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10642 if (ret)
10643 super->updates_pending++;
10644
10645 return ret;
10646}
928f1424
TM		 10647/*******************************************************************************
10648* Function: imsm_get_badblocks
10649* Description: This routine get list of bad blocks for an array
10650*
10651* Parameters:
10652* a : array
10653* slot : disk number
10654* Returns:
10655* bb : structure containing bad blocks
10656* NULL : error
10657******************************************************************************/
10658static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10659{
10660 int inst = a->info.container_member;
10661 struct intel_super *super = a->container->sb;
10662 struct imsm_dev *dev = get_imsm_dev(super, inst);
10663 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10664 int ord;
10665
10666 ord = imsm_disk_slot_to_ord(a, slot);
10667 if (ord < 0)
10668 return NULL;
10669
10670 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10671 per_dev_array_size(map), &super->bb);
928f1424
TM		 10672
10673 return &super->bb;
10674}
27156a57
TM		 10675/*******************************************************************************
10676* Function: examine_badblocks_imsm
10677* Description: Prints list of bad blocks on a disk to the standard output
10678*
10679* Parameters:
10680* st : metadata handler
10681* fd : open file descriptor for device
10682* devname : device name
10683* Returns:
10684* 0 : Success
10685* 1 : Error
10686******************************************************************************/
10687static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10688{
10689 struct intel_super *super = st->sb;
10690 struct bbm_log *log = super->bbm_log;
10691 struct dl *d = NULL;
10692 int any = 0;
10693
10694 for (d = super->disks; d ; d = d->next) {
10695 if (strcmp(d->devname, devname) == 0)
10696 break;
10697 }
10698
10699 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10700 pr_err("%s doesn't appear to be part of a raid array\n",
10701 devname);
10702 return 1;
10703 }
10704
10705 if (log != NULL) {
10706 unsigned int i;
10707 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10708
10709 for (i = 0; i < log->entry_count; i++) {
10710 if (entry[i].disk_ordinal == d->index) {
10711 unsigned long long sector = __le48_to_cpu(
10712 &entry[i].defective_block_start);
10713 int cnt = entry[i].marked_count + 1;
10714
10715 if (!any) {
10716 printf("Bad-blocks on %s:\n", devname);
10717 any = 1;
10718 }
10719
10720 printf("%20llu for %d sectors\n", sector, cnt);
10721 }
10722 }
10723 }
10724
10725 if (!any)
10726 printf("No bad-blocks list configured on %s\n", devname);
10727
10728 return 0;
10729}
687629c2
AK		 10730/*******************************************************************************
10731 * Function: init_migr_record_imsm
10732 * Description: Function inits imsm migration record
10733 * Parameters:
10734 * super : imsm internal array info
10735 * dev : device under migration
10736 * info : general array info to find the smallest device
10737 * Returns:
10738 * none
10739 ******************************************************************************/
10740void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10741 struct mdinfo *info)
10742{
10743 struct intel_super *super = st->sb;
10744 struct migr_record *migr_rec = super->migr_rec;
10745 int new_data_disks;
10746 unsigned long long dsize, dev_sectors;
10747 long long unsigned min_dev_sectors = -1LLU;
10748 struct mdinfo *sd;
10749 char nm[30];
10750 int fd;
238c0a71
AK		 10751 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10752 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10753 unsigned long long num_migr_units;
3ef4403c 10754 unsigned long long array_blocks;
687629c2
AK		 10755
10756 memset(migr_rec, 0, sizeof(struct migr_record));
10757 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10758
10759 /* only ascending reshape supported now */
10760 migr_rec->ascending_migr = __cpu_to_le32(1);
10761
10762 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10763 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10764 migr_rec->dest_depth_per_unit *=
10765 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10766 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10767 migr_rec->blocks_per_unit =
10768 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10769 migr_rec->dest_depth_per_unit =
10770 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10771 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10772 num_migr_units =
10773 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10774
10775 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10776 num_migr_units++;
9f421827 10777 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10778
10779 migr_rec->post_migr_vol_cap = dev->size_low;
10780 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10781
687629c2
AK		 10782 /* Find the smallest dev */
10783 for (sd = info->devs ; sd ; sd = sd->next) {
10784 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10785 fd = dev_open(nm, O_RDONLY);
10786 if (fd < 0)
10787 continue;
10788 get_dev_size(fd, NULL, &dsize);
10789 dev_sectors = dsize / 512;
10790 if (dev_sectors < min_dev_sectors)
10791 min_dev_sectors = dev_sectors;
10792 close(fd);
10793 }
9f421827 10794 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10795 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10796
10797 write_imsm_migr_rec(st);
10798
10799 return;
10800}
10801
10802/*******************************************************************************
10803 * Function: save_backup_imsm
10804 * Description: Function saves critical data stripes to Migration Copy Area
10805 * and updates the current migration unit status.
10806 * Use restore_stripes() to form a destination stripe,
10807 * and to write it to the Copy Area.
10808 * Parameters:
10809 * st : supertype information
aea93171 10810 * dev : imsm device that backup is saved for
687629c2
AK		 10811 * info : general array info
10812 * buf : input buffer
687629c2
AK		 10813 * length : length of data to backup (blocks_per_unit)
10814 * Returns:
10815 * 0 : success
10816 *, -1 : fail
10817 ******************************************************************************/
10818int save_backup_imsm(struct supertype *st,
10819 struct imsm_dev *dev,
10820 struct mdinfo *info,
10821 void *buf,
687629c2
AK		 10822 int length)
10823{
10824 int rv = -1;
10825 struct intel_super *super = st->sb;
594dc1b8
JS		 10826 unsigned long long *target_offsets;
10827 int *targets;
687629c2 10828 int i;
238c0a71 10829 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10830 int new_disks = map_dest->num_members;
ab724b98
AK		 10831 int dest_layout = 0;
10832 int dest_chunk;
d1877f69 10833 unsigned long long start;
9529d343 10834 int data_disks = imsm_num_data_members(map_dest);
687629c2 10835
503975b9 10836 targets = xmalloc(new_disks * sizeof(int));
687629c2 10837
7e45b550
AK		 10838 for (i = 0; i < new_disks; i++)
10839 targets[i] = -1;
10840
503975b9 10841 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10842
d1877f69 10843 start = info->reshape_progress * 512;
687629c2 10844 for (i = 0; i < new_disks; i++) {
9f421827 10845 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10846 /* move back copy area adderss, it will be moved forward
10847 * in restore_stripes() using start input variable
10848 */
10849 target_offsets[i] -= start/data_disks;
687629c2
AK		 10850 }
10851
9a717282
AK		 10852 if (open_backup_targets(info, new_disks, targets,
10853 super, dev))
687629c2
AK		 10854 goto abort;
10855
68eb8bc6 10856 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10857 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10858
687629c2
AK		 10859 if (restore_stripes(targets, /* list of dest devices */
10860 target_offsets, /* migration record offsets */
10861 new_disks,
ab724b98
AK		 10862 dest_chunk,
10863 map_dest->raid_level,
10864 dest_layout,
10865 -1, /* source backup file descriptor */
10866 0, /* input buf offset
10867 * always 0 buf is already offseted */
d1877f69 10868 start,
687629c2
AK		 10869 length,
10870 buf) != 0) {
e7b84f9d 10871 pr_err("Error restoring stripes\n");
687629c2
AK		 10872 goto abort;
10873 }
10874
10875 rv = 0;
10876
10877abort:
10878 if (targets) {
9a717282 10879 close_targets(targets, new_disks);
687629c2
AK		 10880 free(targets);
10881 }
10882 free(target_offsets);
10883
10884 return rv;
10885}
10886
10887/*******************************************************************************
10888 * Function: save_checkpoint_imsm
10889 * Description: Function called for current unit status update
10890 * in the migration record. It writes it to disk.
10891 * Parameters:
10892 * super : imsm internal array info
10893 * info : general array info
10894 * Returns:
10895 * 0: success
10896 * 1: failure
0228d92c
AK		 10897 * 2: failure, means no valid migration record
10898 * / no general migration in progress /
687629c2
AK		 10899 ******************************************************************************/
10900int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10901{
10902 struct intel_super *super = st->sb;
f8b72ef5
AK		 10903 unsigned long long blocks_per_unit;
10904 unsigned long long curr_migr_unit;
10905
2e062e82 10906 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10907 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10908 return 1;
10909 }
10910
f8b72ef5
AK		 10911 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10912 if (blocks_per_unit == 0) {
0228d92c
AK		 10913 dprintf("imsm: no migration in progress.\n");
10914 return 2;
687629c2 10915 }
f8b72ef5
AK		 10916 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10917 /* check if array is alligned to copy area
10918 * if it is not alligned, add one to current migration unit value
10919 * this can happend on array reshape finish only
10920 */
10921 if (info->reshape_progress % blocks_per_unit)
10922 curr_migr_unit++;
687629c2 10923
9f421827 10924 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10925 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10926 set_migr_dest_1st_member_lba(super->migr_rec,
10927 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10928
687629c2 10929 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10930 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10931 return 1;
10932 }
10933
10934 return 0;
10935}
10936
276d77db
AK		 10937/*******************************************************************************
10938 * Function: recover_backup_imsm
10939 * Description: Function recovers critical data from the Migration Copy Area
10940 * while assembling an array.
10941 * Parameters:
10942 * super : imsm internal array info
10943 * info : general array info
10944 * Returns:
10945 * 0 : success (or there is no data to recover)
10946 * 1 : fail
10947 ******************************************************************************/
10948int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10949{
10950 struct intel_super *super = st->sb;
10951 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10952 struct imsm_map *map_dest;
276d77db
AK		 10953 struct intel_dev *id = NULL;
10954 unsigned long long read_offset;
10955 unsigned long long write_offset;
10956 unsigned unit_len;
10957 int *targets = NULL;
10958 int new_disks, i, err;
10959 char *buf = NULL;
10960 int retval = 1;
f36a9ecd 10961 unsigned int sector_size = super->sector_size;
9f421827
PB		 10962 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10963 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10964 char buffer[20];
6c3560c0 10965 int skipped_disks = 0;
276d77db
AK		 10966
10967 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10968 if (err < 1)
10969 return 1;
10970
10971 /* recover data only during assemblation */
10972 if (strncmp(buffer, "inactive", 8) != 0)
10973 return 0;
10974 /* no data to recover */
10975 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10976 return 0;
10977 if (curr_migr_unit >= num_migr_units)
10978 return 1;
10979
10980 /* find device during reshape */
10981 for (id = super->devlist; id; id = id->next)
10982 if (is_gen_migration(id->dev))
10983 break;
10984 if (id == NULL)
10985 return 1;
10986
238c0a71 10987 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10988 new_disks = map_dest->num_members;
10989
9f421827 10990 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 10991
9f421827 10992 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 10993 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10994
10995 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10996 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10997 goto abort;
503975b9 10998 targets = xcalloc(new_disks, sizeof(int));
276d77db 10999
9a717282 11000 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 11001 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 11002 goto abort;
11003 }
276d77db
AK		 11004
11005 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 11006 if (targets[i] < 0) {
11007 skipped_disks++;
11008 continue;
11009 }
276d77db 11010 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11011 pr_err("Cannot seek to block: %s\n",
11012 strerror(errno));
137debce
AK		 11013 skipped_disks++;
11014 continue;
276d77db 11015 }
9ec11d1a 11016 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11017 pr_err("Cannot read copy area block: %s\n",
11018 strerror(errno));
137debce
AK		 11019 skipped_disks++;
11020 continue;
276d77db
AK		 11021 }
11022 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11023 pr_err("Cannot seek to block: %s\n",
11024 strerror(errno));
137debce
AK		 11025 skipped_disks++;
11026 continue;
276d77db 11027 }
9ec11d1a 11028 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 11029 pr_err("Cannot restore block: %s\n",
11030 strerror(errno));
137debce
AK		 11031 skipped_disks++;
11032 continue;
276d77db
AK		 11033 }
11034 }
11035
137debce
AK		 11036 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11037 new_disks,
11038 super,
11039 id->dev)) {
7a862a02 11040 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11041 goto abort;
11042 }
11043
befb629b
AK		 11044 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11045 /* ignore error == 2, this can mean end of reshape here
11046 */
7a862a02 11047 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11048 } else
276d77db 11049 retval = 0;
276d77db
AK		 11050
11051abort:
11052 if (targets) {
11053 for (i = 0; i < new_disks; i++)
11054 if (targets[i])
11055 close(targets[i]);
11056 free(targets);
11057 }
11058 free(buf);
11059 return retval;
11060}
11061
2cda7640
ML		 11062static char disk_by_path[] = "/dev/disk/by-path/";
11063
11064static const char *imsm_get_disk_controller_domain(const char *path)
11065{
2cda7640 11066 char disk_path[PATH_MAX];
96234762
LM		 11067 char *drv=NULL;
11068 struct stat st;
2cda7640 11069
6d8d290a 11070 strcpy(disk_path, disk_by_path);
96234762
LM		 11071 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11072 if (stat(disk_path, &st) == 0) {
11073 struct sys_dev* hba;
594dc1b8 11074 char *path;
96234762
LM		 11075
11076 path = devt_to_devpath(st.st_rdev);
11077 if (path == NULL)
11078 return "unknown";
11079 hba = find_disk_attached_hba(-1, path);
11080 if (hba && hba->type == SYS_DEV_SAS)
11081 drv = "isci";
11082 else if (hba && hba->type == SYS_DEV_SATA)
11083 drv = "ahci";
c6839718
MT		 11084 else if (hba && hba->type == SYS_DEV_VMD)
11085 drv = "vmd";
11086 else if (hba && hba->type == SYS_DEV_NVME)
11087 drv = "nvme";
1011e834 11088 else
96234762
LM		 11089 drv = "unknown";
11090 dprintf("path: %s hba: %s attached: %s\n",
11091 path, (hba) ? hba->path : "NULL", drv);
11092 free(path);
2cda7640 11093 }
96234762 11094 return drv;
2cda7640
ML		 11095}
11096
4dd2df09 11097static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11098{
4dd2df09 11099 static char devnm[32];
78b10e66
N		 11100 char subdev_name[20];
11101 struct mdstat_ent *mdstat;
11102
11103 sprintf(subdev_name, "%d", subdev);
11104 mdstat = mdstat_by_subdev(subdev_name, container);
11105 if (!mdstat)
4dd2df09 11106 return NULL;
78b10e66 11107
4dd2df09 11108 strcpy(devnm, mdstat->devnm);
78b10e66 11109 free_mdstat(mdstat);
4dd2df09 11110 return devnm;
78b10e66
N		 11111}
11112
11113static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11114 struct geo_params *geo,
fbf3d202
AK		 11115 int *old_raid_disks,
11116 int direction)
78b10e66 11117{
694575e7
KW		 11118 /* currently we only support increasing the number of devices
11119 * for a container. This increases the number of device for each
11120 * member array. They must all be RAID0 or RAID5.
11121 */
78b10e66
N		 11122 int ret_val = 0;
11123 struct mdinfo *info, *member;
11124 int devices_that_can_grow = 0;
11125
7a862a02 11126 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11127
d04f65f4 11128 if (geo->size > 0 ||
78b10e66
N		 11129 geo->level != UnSet ||
11130 geo->layout != UnSet ||
11131 geo->chunksize != 0 ||
11132 geo->raid_disks == UnSet) {
7a862a02 11133 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11134 return ret_val;
11135 }
11136
fbf3d202 11137 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11138 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11139 return ret_val;
11140 }
11141
78b10e66
N		 11142 info = container_content_imsm(st, NULL);
11143 for (member = info; member; member = member->next) {
4dd2df09 11144 char *result;
78b10e66
N		 11145
11146 dprintf("imsm: checking device_num: %i\n",
11147 member->container_member);
11148
d7d205bd 11149 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11150 /* we work on container for Online Capacity Expansion
11151 * only so raid_disks has to grow
11152 */
7a862a02 11153 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11154 break;
11155 }
11156
089f9d79 11157 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11158 /* we cannot use this container with other raid level
11159 */
7a862a02 11160 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11161 info->array.level);
11162 break;
11163 } else {
11164 /* check for platform support
11165 * for this raid level configuration
11166 */
11167 struct intel_super *super = st->sb;
11168 if (!is_raid_level_supported(super->orom,
11169 member->array.level,
11170 geo->raid_disks)) {
7a862a02 11171 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11172 info->array.level,
11173 geo->raid_disks,
11174 geo->raid_disks > 1 ? "s" : "");
11175 break;
11176 }
2a4a08e7
AK		 11177 /* check if component size is aligned to chunk size
11178 */
11179 if (info->component_size %
11180 (info->array.chunk_size/512)) {
7a862a02 11181 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11182 break;
11183 }
78b10e66
N		 11184 }
11185
11186 if (*old_raid_disks &&
11187 info->array.raid_disks != *old_raid_disks)
11188 break;
11189 *old_raid_disks = info->array.raid_disks;
11190
11191 /* All raid5 and raid0 volumes in container
11192 * have to be ready for Online Capacity Expansion
11193 * so they need to be assembled. We have already
11194 * checked that no recovery etc is happening.
11195 */
4dd2df09
N		 11196 result = imsm_find_array_devnm_by_subdev(member->container_member,
11197 st->container_devnm);
11198 if (result == NULL) {
78b10e66
N		 11199 dprintf("imsm: cannot find array\n");
11200 break;
11201 }
11202 devices_that_can_grow++;
11203 }
11204 sysfs_free(info);
11205 if (!member && devices_that_can_grow)
11206 ret_val = 1;
11207
11208 if (ret_val)
1ade5cc1 11209 dprintf("Container operation allowed\n");
78b10e66 11210 else
1ade5cc1 11211 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11212
11213 return ret_val;
11214}
11215
11216/* Function: get_spares_for_grow
11217 * Description: Allocates memory and creates list of spare devices
1011e834 11218 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11219 * Parameters: Pointer to the supertype structure
11220 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11221 * NULL if fail
78b10e66
N		 11222 */
11223static struct mdinfo *get_spares_for_grow(struct supertype *st)
11224{
fbfdcb06
AO		 11225 struct spare_criteria sc;
11226
11227 get_spare_criteria_imsm(st, &sc);
11228 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11229}
11230
11231/******************************************************************************
11232 * function: imsm_create_metadata_update_for_reshape
11233 * Function creates update for whole IMSM container.
11234 *
11235 ******************************************************************************/
11236static int imsm_create_metadata_update_for_reshape(
11237 struct supertype *st,
11238 struct geo_params *geo,
11239 int old_raid_disks,
11240 struct imsm_update_reshape **updatep)
11241{
11242 struct intel_super *super = st->sb;
11243 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11244 int update_memory_size;
11245 struct imsm_update_reshape *u;
11246 struct mdinfo *spares;
78b10e66 11247 int i;
594dc1b8 11248 int delta_disks;
bbd24d86 11249 struct mdinfo *dev;
78b10e66 11250
1ade5cc1 11251 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11252
11253 delta_disks = geo->raid_disks - old_raid_disks;
11254
11255 /* size of all update data without anchor */
11256 update_memory_size = sizeof(struct imsm_update_reshape);
11257
11258 /* now add space for spare disks that we need to add. */
11259 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11260
503975b9 11261 u = xcalloc(1, update_memory_size);
78b10e66
N		 11262 u->type = update_reshape_container_disks;
11263 u->old_raid_disks = old_raid_disks;
11264 u->new_raid_disks = geo->raid_disks;
11265
11266 /* now get spare disks list
11267 */
11268 spares = get_spares_for_grow(st);
11269
d7be7d87 11270 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11271 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11272 i = -1;
78b10e66
N		 11273 goto abort;
11274 }
11275
11276 /* we have got spares
11277 * update disk list in imsm_disk list table in anchor
11278 */
11279 dprintf("imsm: %i spares are available.\n\n",
11280 spares->array.spare_disks);
11281
bbd24d86 11282 dev = spares->devs;
78b10e66 11283 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11284 struct dl *dl;
11285
bbd24d86
AK		 11286 if (dev == NULL)
11287 break;
78b10e66
N		 11288 u->new_disks[i] = makedev(dev->disk.major,
11289 dev->disk.minor);
11290 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11291 dl->index = mpb->num_disks;
11292 mpb->num_disks++;
bbd24d86 11293 dev = dev->next;
78b10e66 11294 }
78b10e66
N		 11295
11296abort:
11297 /* free spares
11298 */
11299 sysfs_free(spares);
11300
d677e0b8 11301 dprintf("imsm: reshape update preparation :");
78b10e66 11302 if (i == delta_disks) {
1ade5cc1 11303 dprintf_cont(" OK\n");
78b10e66
N		 11304 *updatep = u;
11305 return update_memory_size;
11306 }
11307 free(u);
1ade5cc1 11308 dprintf_cont(" Error\n");
78b10e66
N		 11309
11310 return 0;
11311}
11312
f3871fdc
AK		 11313/******************************************************************************
11314 * function: imsm_create_metadata_update_for_size_change()
11315 * Creates update for IMSM array for array size change.
11316 *
11317 ******************************************************************************/
11318static int imsm_create_metadata_update_for_size_change(
11319 struct supertype *st,
11320 struct geo_params *geo,
11321 struct imsm_update_size_change **updatep)
11322{
11323 struct intel_super *super = st->sb;
594dc1b8
JS		 11324 int update_memory_size;
11325 struct imsm_update_size_change *u;
f3871fdc 11326
1ade5cc1 11327 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11328
11329 /* size of all update data without anchor */
11330 update_memory_size = sizeof(struct imsm_update_size_change);
11331
503975b9 11332 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11333 u->type = update_size_change;
11334 u->subdev = super->current_vol;
11335 u->new_size = geo->size;
11336
11337 dprintf("imsm: reshape update preparation : OK\n");
11338 *updatep = u;
11339
11340 return update_memory_size;
11341}
11342
48c5303a
PC		 11343/******************************************************************************
11344 * function: imsm_create_metadata_update_for_migration()
11345 * Creates update for IMSM array.
11346 *
11347 ******************************************************************************/
11348static int imsm_create_metadata_update_for_migration(
11349 struct supertype *st,
11350 struct geo_params *geo,
11351 struct imsm_update_reshape_migration **updatep)
11352{
11353 struct intel_super *super = st->sb;
594dc1b8
JS		 11354 int update_memory_size;
11355 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11356 struct imsm_dev *dev;
11357 int previous_level = -1;
11358
1ade5cc1 11359 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11360
11361 /* size of all update data without anchor */
11362 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11363
503975b9 11364 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11365 u->type = update_reshape_migration;
11366 u->subdev = super->current_vol;
11367 u->new_level = geo->level;
11368 u->new_layout = geo->layout;
11369 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11370 u->new_disks[0] = -1;
4bba0439 11371 u->new_chunksize = -1;
48c5303a
PC		 11372
11373 dev = get_imsm_dev(super, u->subdev);
11374 if (dev) {
11375 struct imsm_map *map;
11376
238c0a71 11377 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11378 if (map) {
11379 int current_chunk_size =
11380 __le16_to_cpu(map->blocks_per_strip) / 2;
11381
11382 if (geo->chunksize != current_chunk_size) {
11383 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11384 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11385 current_chunk_size, u->new_chunksize);
11386 }
48c5303a 11387 previous_level = map->raid_level;
4bba0439 11388 }
48c5303a 11389 }
089f9d79 11390 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11391 struct mdinfo *spares = NULL;
11392
11393 u->new_raid_disks++;
11394 spares = get_spares_for_grow(st);
089f9d79 11395 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11396 free(u);
11397 sysfs_free(spares);
11398 update_memory_size = 0;
565cc99e 11399 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11400 return 0;
11401 }
11402 sysfs_free(spares);
11403 }
11404 dprintf("imsm: reshape update preparation : OK\n");
11405 *updatep = u;
11406
11407 return update_memory_size;
11408}
11409
8dd70bce
AK		 11410static void imsm_update_metadata_locally(struct supertype *st,
11411 void *buf, int len)
11412{
11413 struct metadata_update mu;
11414
11415 mu.buf = buf;
11416 mu.len = len;
11417 mu.space = NULL;
11418 mu.space_list = NULL;
11419 mu.next = NULL;
5fe6f031
N		 11420 if (imsm_prepare_update(st, &mu))
11421 imsm_process_update(st, &mu);
8dd70bce
AK		 11422
11423 while (mu.space_list) {
11424 void **space = mu.space_list;
11425 mu.space_list = *space;
11426 free(space);
11427 }
11428}
78b10e66 11429
471bceb6 11430/***************************************************************************
694575e7 11431* Function: imsm_analyze_change
471bceb6 11432* Description: Function analyze change for single volume
1011e834 11433* and validate if transition is supported
fbf3d202
AK		 11434* Parameters: Geometry parameters, supertype structure,
11435* metadata change direction (apply/rollback)
694575e7 11436* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11437****************************************************************************/
11438enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11439 struct geo_params *geo,
11440 int direction)
694575e7 11441{
471bceb6
KW		 11442 struct mdinfo info;
11443 int change = -1;
11444 int check_devs = 0;
c21e737b 11445 int chunk;
67a2db32
AK		 11446 /* number of added/removed disks in operation result */
11447 int devNumChange = 0;
11448 /* imsm compatible layout value for array geometry verification */
11449 int imsm_layout = -1;
7abc9871
AK		 11450 int data_disks;
11451 struct imsm_dev *dev;
9529d343 11452 struct imsm_map *map;
7abc9871 11453 struct intel_super *super;
d04f65f4 11454 unsigned long long current_size;
65d38cca 11455 unsigned long long free_size;
d04f65f4 11456 unsigned long long max_size;
65d38cca 11457 int rv;
471bceb6
KW		 11458
11459 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11460 if (geo->level != info.array.level && geo->level >= 0 &&
11461 geo->level != UnSet) {
471bceb6
KW		 11462 switch (info.array.level) {
11463 case 0:
11464 if (geo->level == 5) {
b5347799 11465 change = CH_MIGRATION;
e13ce846 11466 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11467 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11468 change = -1;
11469 goto analyse_change_exit;
11470 }
67a2db32 11471 imsm_layout = geo->layout;
471bceb6 11472 check_devs = 1;
e91a3bad
LM		 11473 devNumChange = 1; /* parity disk added */
11474 } else if (geo->level == 10) {
471bceb6
KW		 11475 change = CH_TAKEOVER;
11476 check_devs = 1;
e91a3bad 11477 devNumChange = 2; /* two mirrors added */
67a2db32 11478 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11479 }
dfe77a9e
KW		 11480 break;
11481 case 1:
471bceb6
KW		 11482 case 10:
11483 if (geo->level == 0) {
11484 change = CH_TAKEOVER;
11485 check_devs = 1;
e91a3bad 11486 devNumChange = -(geo->raid_disks/2);
67a2db32 11487 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11488 }
11489 break;
11490 }
11491 if (change == -1) {
7a862a02 11492 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11493 info.array.level, geo->level);
471bceb6
KW		 11494 goto analyse_change_exit;
11495 }
11496 } else
11497 geo->level = info.array.level;
11498
089f9d79
JS		 11499 if (geo->layout != info.array.layout &&
11500 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11501 change = CH_MIGRATION;
089f9d79
JS		 11502 if (info.array.layout == 0 && info.array.level == 5 &&
11503 geo->layout == 5) {
471bceb6 11504 /* reshape 5 -> 4 */
089f9d79
JS		 11505 } else if (info.array.layout == 5 && info.array.level == 5 &&
11506 geo->layout == 0) {
471bceb6
KW		 11507 /* reshape 4 -> 5 */
11508 geo->layout = 0;
11509 geo->level = 5;
11510 } else {
7a862a02 11511 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11512 info.array.layout, geo->layout);
471bceb6
KW		 11513 change = -1;
11514 goto analyse_change_exit;
11515 }
67a2db32 11516 } else {
471bceb6 11517 geo->layout = info.array.layout;
67a2db32
AK		 11518 if (imsm_layout == -1)
11519 imsm_layout = info.array.layout;
11520 }
471bceb6 11521
089f9d79
JS		 11522 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11523 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11524 if (info.array.level == 10) {
11525 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11526 change = -1;
11527 goto analyse_change_exit;
1e9b2c3f
PB		 11528 } else if (info.component_size % (geo->chunksize/512)) {
11529 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11530 geo->chunksize/1024, info.component_size/2);
11531 change = -1;
11532 goto analyse_change_exit;
2d2b0eb7 11533 }
b5347799 11534 change = CH_MIGRATION;
2d2b0eb7 11535 } else {
471bceb6 11536 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11537 }
471bceb6 11538
c21e737b 11539 chunk = geo->chunksize / 1024;
7abc9871
AK		 11540
11541 super = st->sb;
11542 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11543 map = get_imsm_map(dev, MAP_0);
11544 data_disks = imsm_num_data_members(map);
c41e00b2 11545 /* compute current size per disk member
7abc9871 11546 */
c41e00b2
AK		 11547 current_size = info.custom_array_size / data_disks;
11548
089f9d79 11549 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11550 /* align component size
11551 */
3e684231 11552 geo->size = imsm_component_size_alignment_check(
c41e00b2 11553 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11554 chunk * 1024, super->sector_size,
c41e00b2 11555 geo->size * 2);
65d0b4ce 11556 if (geo->size == 0) {
7a862a02 11557 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11558 current_size);
11559 goto analyse_change_exit;
11560 }
c41e00b2 11561 }
7abc9871 11562
089f9d79 11563 if (current_size != geo->size && geo->size > 0) {
7abc9871 11564 if (change != -1) {
7a862a02 11565 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11566 change = -1;
11567 goto analyse_change_exit;
11568 }
11569 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11570 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11571 super->current_vol, st->devnm);
7abc9871
AK		 11572 goto analyse_change_exit;
11573 }
65d38cca
LD		 11574 /* check the maximum available size
11575 */
11576 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11577 0, chunk, &free_size);
11578 if (rv == 0)
11579 /* Cannot find maximum available space
11580 */
11581 max_size = 0;
11582 else {
11583 max_size = free_size + current_size;
11584 /* align component size
11585 */
3e684231 11586 max_size = imsm_component_size_alignment_check(
65d38cca 11587 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11588 chunk * 1024, super->sector_size,
65d38cca
LD		 11589 max_size);
11590 }
d04f65f4 11591 if (geo->size == MAX_SIZE) {
b130333f
AK		 11592 /* requested size change to the maximum available size
11593 */
65d38cca 11594 if (max_size == 0) {
7a862a02 11595 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11596 change = -1;
11597 goto analyse_change_exit;
65d38cca
LD		 11598 } else
11599 geo->size = max_size;
c41e00b2 11600 }
b130333f 11601
681b7ae2 11602 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11603 /* accept size for rollback only
11604 */
11605 } else {
11606 /* round size due to metadata compatibility
11607 */
11608 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11609 << SECT_PER_MB_SHIFT;
11610 dprintf("Prepare update for size change to %llu\n",
11611 geo->size );
11612 if (current_size >= geo->size) {
7a862a02 11613 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11614 current_size, geo->size);
fbf3d202
AK		 11615 goto analyse_change_exit;
11616 }
65d38cca 11617 if (max_size && geo->size > max_size) {
7a862a02 11618 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11619 max_size, geo->size);
65d38cca
LD		 11620 goto analyse_change_exit;
11621 }
7abc9871
AK		 11622 }
11623 geo->size *= data_disks;
11624 geo->raid_disks = dev->vol.map->num_members;
11625 change = CH_ARRAY_SIZE;
11626 }
471bceb6
KW		 11627 if (!validate_geometry_imsm(st,
11628 geo->level,
67a2db32 11629 imsm_layout,
e91a3bad 11630 geo->raid_disks + devNumChange,
c21e737b 11631 &chunk,
af4348dd 11632 geo->size, INVALID_SECTORS,
5308f117 11633 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11634 change = -1;
11635
11636 if (check_devs) {
11637 struct intel_super *super = st->sb;
11638 struct imsm_super *mpb = super->anchor;
11639
11640 if (mpb->num_raid_devs > 1) {
7a862a02 11641 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11642 geo->dev_name);
471bceb6
KW		 11643 change = -1;
11644 }
11645 }
11646
11647analyse_change_exit:
089f9d79
JS		 11648 if (direction == ROLLBACK_METADATA_CHANGES &&
11649 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11650 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11651 change = -1;
11652 }
471bceb6 11653 return change;
694575e7
KW		 11654}
11655
bb025c2f
KW		 11656int imsm_takeover(struct supertype *st, struct geo_params *geo)
11657{
11658 struct intel_super *super = st->sb;
11659 struct imsm_update_takeover *u;
11660
503975b9 11661 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11662
11663 u->type = update_takeover;
11664 u->subarray = super->current_vol;
11665
11666 /* 10->0 transition */
11667 if (geo->level == 0)
11668 u->direction = R10_TO_R0;
11669
0529c688
KW		 11670 /* 0->10 transition */
11671 if (geo->level == 10)
11672 u->direction = R0_TO_R10;
11673
bb025c2f
KW		 11674 /* update metadata locally */
11675 imsm_update_metadata_locally(st, u,
11676 sizeof(struct imsm_update_takeover));
11677 /* and possibly remotely */
11678 if (st->update_tail)
11679 append_metadata_update(st, u,
11680 sizeof(struct imsm_update_takeover));
11681 else
11682 free(u);
11683
11684 return 0;
11685}
11686
895ffd99
MT		 11687/* Flush size update if size calculated by num_data_stripes is higher than
11688 * imsm_dev_size to eliminate differences during reshape.
11689 * Mdmon will recalculate them correctly.
11690 * If subarray index is not set then check whole container.
11691 * Returns:
11692 * 0 - no error occurred
11693 * 1 - error detected
11694 */
11695static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index)
11696{
11697 struct intel_super *super = st->sb;
11698 int tmp = super->current_vol;
11699 int ret_val = 1;
11700 int i;
11701
11702 for (i = 0; i < super->anchor->num_raid_devs; i++) {
11703 if (subarray_index >= 0 && i != subarray_index)
11704 continue;
11705 super->current_vol = i;
11706 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
11707 struct imsm_map *map = get_imsm_map(dev, MAP_0);
11708 unsigned int disc_count = imsm_num_data_members(map);
11709 struct geo_params geo;
11710 struct imsm_update_size_change *update;
11711 unsigned long long calc_size = per_dev_array_size(map) * disc_count;
11712 unsigned long long d_size = imsm_dev_size(dev);
11713 int u_size;
11714
11715 if (calc_size == d_size || dev->vol.migr_type == MIGR_GEN_MIGR)
11716 continue;
11717
11718 /* There is a difference, verify that imsm_dev_size is
11719 * rounded correctly and push update.
11720 */
11721 if (d_size != round_size_to_mb(d_size, disc_count)) {
11722 dprintf("imsm: Size of volume %d is not rounded correctly\n",
11723 i);
11724 goto exit;
11725 }
11726 memset(&geo, 0, sizeof(struct geo_params));
11727 geo.size = d_size;
11728 u_size = imsm_create_metadata_update_for_size_change(st, &geo,
11729 &update);
11730 if (u_size < 1) {
11731 dprintf("imsm: Cannot prepare size change update\n");
11732 goto exit;
11733 }
11734 imsm_update_metadata_locally(st, update, u_size);
11735 if (st->update_tail) {
11736 append_metadata_update(st, update, u_size);
11737 flush_metadata_updates(st);
11738 st->update_tail = &st->updates;
11739 } else {
11740 imsm_sync_metadata(st);
11741 }
11742 }
11743 ret_val = 0;
11744exit:
11745 super->current_vol = tmp;
11746 return ret_val;
11747}
11748
d04f65f4
N		 11749static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11750 int level,
78b10e66 11751 int layout, int chunksize, int raid_disks,
41784c88 11752 int delta_disks, char *backup, char *dev,
016e00f5 11753 int direction, int verbose)
78b10e66 11754{
78b10e66
N		 11755 int ret_val = 1;
11756 struct geo_params geo;
11757
1ade5cc1 11758 dprintf("(enter)\n");
78b10e66 11759
71204a50 11760 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11761
11762 geo.dev_name = dev;
4dd2df09 11763 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11764 geo.size = size;
11765 geo.level = level;
11766 geo.layout = layout;
11767 geo.chunksize = chunksize;
11768 geo.raid_disks = raid_disks;
41784c88
AK		 11769 if (delta_disks != UnSet)
11770 geo.raid_disks += delta_disks;
78b10e66 11771
1ade5cc1
N		 11772 dprintf("for level : %i\n", geo.level);
11773 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11774
4dd2df09 11775 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11776 /* On container level we can only increase number of devices. */
11777 dprintf("imsm: info: Container operation\n");
78b10e66 11778 int old_raid_disks = 0;
6dc0be30 11779
78b10e66 11780 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11781 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11782 struct imsm_update_reshape *u = NULL;
11783 int len;
11784
895ffd99
MT		 11785 if (imsm_fix_size_mismatch(st, -1)) {
11786 dprintf("imsm: Cannot fix size mismatch\n");
11787 goto exit_imsm_reshape_super;
11788 }
11789
78b10e66
N		 11790 len = imsm_create_metadata_update_for_reshape(
11791 st, &geo, old_raid_disks, &u);
11792
ed08d51c
AK		 11793 if (len <= 0) {
11794 dprintf("imsm: Cannot prepare update\n");
11795 goto exit_imsm_reshape_super;
11796 }
11797
8dd70bce
AK		 11798 ret_val = 0;
11799 /* update metadata locally */
11800 imsm_update_metadata_locally(st, u, len);
11801 /* and possibly remotely */
11802 if (st->update_tail)
11803 append_metadata_update(st, u, len);
11804 else
ed08d51c 11805 free(u);
8dd70bce 11806
694575e7 11807 } else {
7a862a02 11808 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11809 }
11810 } else {
11811 /* On volume level we support following operations
471bceb6
KW		 11812 * - takeover: raid10 -> raid0; raid0 -> raid10
11813 * - chunk size migration
11814 * - migration: raid5 -> raid0; raid0 -> raid5
11815 */
11816 struct intel_super *super = st->sb;
11817 struct intel_dev *dev = super->devlist;
4dd2df09 11818 int change;
694575e7 11819 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11820 /* find requested device */
11821 while (dev) {
1011e834 11822 char *devnm =
4dd2df09
N		 11823 imsm_find_array_devnm_by_subdev(
11824 dev->index, st->container_devnm);
11825 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11826 break;
11827 dev = dev->next;
11828 }
11829 if (dev == NULL) {
4dd2df09
N		 11830 pr_err("Cannot find %s (%s) subarray\n",
11831 geo.dev_name, geo.devnm);
471bceb6
KW		 11832 goto exit_imsm_reshape_super;
11833 }
11834 super->current_vol = dev->index;
fbf3d202 11835 change = imsm_analyze_change(st, &geo, direction);
694575e7 11836 switch (change) {
471bceb6 11837 case CH_TAKEOVER:
bb025c2f 11838 ret_val = imsm_takeover(st, &geo);
694575e7 11839 break;
48c5303a
PC		 11840 case CH_MIGRATION: {
11841 struct imsm_update_reshape_migration *u = NULL;
11842 int len =
11843 imsm_create_metadata_update_for_migration(
11844 st, &geo, &u);
11845 if (len < 1) {
7a862a02 11846 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11847 break;
11848 }
471bceb6 11849 ret_val = 0;
48c5303a
PC		 11850 /* update metadata locally */
11851 imsm_update_metadata_locally(st, u, len);
11852 /* and possibly remotely */
11853 if (st->update_tail)
11854 append_metadata_update(st, u, len);
11855 else
11856 free(u);
11857 }
11858 break;
7abc9871 11859 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11860 struct imsm_update_size_change *u = NULL;
11861 int len =
11862 imsm_create_metadata_update_for_size_change(
11863 st, &geo, &u);
11864 if (len < 1) {
7a862a02 11865 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11866 break;
11867 }
11868 ret_val = 0;
11869 /* update metadata locally */
11870 imsm_update_metadata_locally(st, u, len);
11871 /* and possibly remotely */
11872 if (st->update_tail)
11873 append_metadata_update(st, u, len);
11874 else
11875 free(u);
7abc9871
AK		 11876 }
11877 break;
471bceb6
KW		 11878 default:
11879 ret_val = 1;
694575e7 11880 }
694575e7 11881 }
78b10e66 11882
ed08d51c 11883exit_imsm_reshape_super:
78b10e66
N		 11884 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11885 return ret_val;
11886}
2cda7640 11887
0febb20c
AO		 11888#define COMPLETED_OK 0
11889#define COMPLETED_NONE 1
11890#define COMPLETED_DELAYED 2
11891
11892static int read_completed(int fd, unsigned long long *val)
11893{
11894 int ret;
11895 char buf[50];
11896
11897 ret = sysfs_fd_get_str(fd, buf, 50);
11898 if (ret < 0)
11899 return ret;
11900
11901 ret = COMPLETED_OK;
11902 if (strncmp(buf, "none", 4) == 0) {
11903 ret = COMPLETED_NONE;
11904 } else if (strncmp(buf, "delayed", 7) == 0) {
11905 ret = COMPLETED_DELAYED;
11906 } else {
11907 char *ep;
11908 *val = strtoull(buf, &ep, 0);
11909 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11910 ret = -1;
11911 }
11912 return ret;
11913}
11914
eee67a47
AK		 11915/*******************************************************************************
11916 * Function: wait_for_reshape_imsm
11917 * Description: Function writes new sync_max value and waits until
11918 * reshape process reach new position
11919 * Parameters:
11920 * sra : general array info
eee67a47
AK		 11921 * ndata : number of disks in new array's layout
11922 * Returns:
11923 * 0 : success,
11924 * 1 : there is no reshape in progress,
11925 * -1 : fail
11926 ******************************************************************************/
ae9f01f8 11927int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11928{
85ca499c 11929 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11930 int retry = 3;
eee67a47 11931 unsigned long long completed;
ae9f01f8
AK		 11932 /* to_complete : new sync_max position */
11933 unsigned long long to_complete = sra->reshape_progress;
11934 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11935
ae9f01f8 11936 if (fd < 0) {
1ade5cc1 11937 dprintf("cannot open reshape_position\n");
eee67a47 11938 return 1;
ae9f01f8 11939 }
eee67a47 11940
df2647fa
PB		 11941 do {
11942 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11943 if (!retry) {
11944 dprintf("cannot read reshape_position (no reshape in progres)\n");
11945 close(fd);
11946 return 1;
11947 }
11948 usleep(30000);
11949 } else
11950 break;
11951 } while (retry--);
eee67a47 11952
85ca499c 11953 if (completed > position_to_set) {
1ade5cc1 11954 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11955 to_complete, position_to_set);
ae9f01f8
AK		 11956 close(fd);
11957 return -1;
11958 }
11959 dprintf("Position set: %llu\n", position_to_set);
11960 if (sysfs_set_num(sra, NULL, "sync_max",
11961 position_to_set) != 0) {
1ade5cc1 11962 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11963 position_to_set);
11964 close(fd);
11965 return -1;
eee67a47
AK		 11966 }
11967
eee67a47 11968 do {
0febb20c 11969 int rc;
eee67a47 11970 char action[20];
5ff3a780 11971 int timeout = 3000;
0febb20c 11972
5ff3a780 11973 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11974 if (sysfs_get_str(sra, NULL, "sync_action",
11975 action, 20) > 0 &&
d7d3809a 11976 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11977 if (strncmp(action, "idle", 4) == 0)
11978 break;
d7d3809a
AP		 11979 close(fd);
11980 return -1;
11981 }
0febb20c
AO		 11982
11983 rc = read_completed(fd, &completed);
11984 if (rc < 0) {
1ade5cc1 11985 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11986 close(fd);
11987 return 1;
0febb20c
AO		 11988 } else if (rc == COMPLETED_NONE)
11989 break;
85ca499c 11990 } while (completed < position_to_set);
b2be2b62 11991
eee67a47
AK		 11992 close(fd);
11993 return 0;
eee67a47
AK		 11994}
11995
b915c95f
AK		 11996/*******************************************************************************
11997 * Function: check_degradation_change
11998 * Description: Check that array hasn't become failed.
11999 * Parameters:
12000 * info : for sysfs access
12001 * sources : source disks descriptors
12002 * degraded: previous degradation level
12003 * Returns:
12004 * degradation level
12005 ******************************************************************************/
12006int check_degradation_change(struct mdinfo *info,
12007 int *sources,
12008 int degraded)
12009{
12010 unsigned long long new_degraded;
e1993023
LD		 12011 int rv;
12012
12013 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 12014 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 12015 /* check each device to ensure it is still working */
12016 struct mdinfo *sd;
12017 new_degraded = 0;
12018 for (sd = info->devs ; sd ; sd = sd->next) {
12019 if (sd->disk.state & (1<<MD_DISK_FAULTY))
12020 continue;
12021 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 12022 char sbuf[100];
12023
b915c95f 12024 if (sysfs_get_str(info,
cf52eff5 12025 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 12026 strstr(sbuf, "faulty") ||
12027 strstr(sbuf, "in_sync") == NULL) {
12028 /* this device is dead */
12029 sd->disk.state = (1<<MD_DISK_FAULTY);
12030 if (sd->disk.raid_disk >= 0 &&
12031 sources[sd->disk.raid_disk] >= 0) {
12032 close(sources[
12033 sd->disk.raid_disk]);
12034 sources[sd->disk.raid_disk] =
12035 -1;
12036 }
12037 new_degraded++;
12038 }
12039 }
12040 }
12041 }
12042
12043 return new_degraded;
12044}
12045
10f22854
AK		 12046/*******************************************************************************
12047 * Function: imsm_manage_reshape
12048 * Description: Function finds array under reshape and it manages reshape
12049 * process. It creates stripes backups (if required) and sets
942e1cdb 12050 * checkpoints.
10f22854
AK		 12051 * Parameters:
12052 * afd : Backup handle (nattive) - not used
12053 * sra : general array info
12054 * reshape : reshape parameters - not used
12055 * st : supertype structure
12056 * blocks : size of critical section [blocks]
12057 * fds : table of source device descriptor
12058 * offsets : start of array (offest per devices)
12059 * dests : not used
12060 * destfd : table of destination device descriptor
12061 * destoffsets : table of destination offsets (per device)
12062 * Returns:
12063 * 1 : success, reshape is done
12064 * 0 : fail
12065 ******************************************************************************/
999b4972
N		 12066static int imsm_manage_reshape(
12067 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12068 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12069 int *fds, unsigned long long *offsets,
12070 int dests, int *destfd, unsigned long long *destoffsets)
12071{
10f22854
AK		 12072 int ret_val = 0;
12073 struct intel_super *super = st->sb;
594dc1b8 12074 struct intel_dev *dv;
de44e46f 12075 unsigned int sector_size = super->sector_size;
10f22854 12076 struct imsm_dev *dev = NULL;
9529d343 12077 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12078 int migr_vol_qan = 0;
12079 int ndata, odata; /* [bytes] */
12080 int chunk; /* [bytes] */
12081 struct migr_record *migr_rec;
12082 char *buf = NULL;
12083 unsigned int buf_size; /* [bytes] */
12084 unsigned long long max_position; /* array size [bytes] */
12085 unsigned long long next_step; /* [blocks]/[bytes] */
12086 unsigned long long old_data_stripe_length;
10f22854
AK		 12087 unsigned long long start_src; /* [bytes] */
12088 unsigned long long start; /* [bytes] */
12089 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12090 int degraded = 0;
ab724b98 12091 int source_layout = 0;
895ffd99 12092 int subarray_index = -1;
10f22854 12093
79a16a9b
JS		 12094 if (!sra)
12095 return ret_val;
12096
12097 if (!fds || !offsets)
10f22854
AK		 12098 goto abort;
12099
12100 /* Find volume during the reshape */
12101 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12102 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12103 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12104 dev = dv->dev;
12105 migr_vol_qan++;
895ffd99 12106 subarray_index = dv->index;
10f22854
AK		 12107 }
12108 }
12109 /* Only one volume can migrate at the same time */
12110 if (migr_vol_qan != 1) {
676e87a8 12111 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12112 "Number of migrating volumes greater than 1\n" :
12113 "There is no volume during migrationg\n");
12114 goto abort;
12115 }
12116
9529d343 12117 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12118 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12119 if (map_src == NULL)
12120 goto abort;
10f22854 12121
9529d343
MD		 12122 ndata = imsm_num_data_members(map_dest);
12123 odata = imsm_num_data_members(map_src);
10f22854 12124
7b1ab482 12125 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12126 old_data_stripe_length = odata * chunk;
12127
12128 migr_rec = super->migr_rec;
12129
10f22854
AK		 12130 /* initialize migration record for start condition */
12131 if (sra->reshape_progress == 0)
12132 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12133 else {
12134 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12135 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12136 goto abort;
12137 }
6a75c8ca
AK		 12138 /* Save checkpoint to update migration record for current
12139 * reshape position (in md). It can be farther than current
12140 * reshape position in metadata.
12141 */
12142 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12143 /* ignore error == 2, this can mean end of reshape here
12144 */
7a862a02 12145 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12146 goto abort;
12147 }
b2c59438 12148 }
10f22854
AK		 12149
12150 /* size for data */
12151 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12152 /* extend buffer size for parity disk */
12153 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12154 /* add space for stripe alignment */
10f22854 12155 buf_size += old_data_stripe_length;
de44e46f
PB		 12156 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12157 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12158 goto abort;
12159 }
12160
3ef4403c 12161 max_position = sra->component_size * ndata;
68eb8bc6 12162 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12163
9f421827
PB		 12164 while (current_migr_unit(migr_rec) <
12165 get_num_migr_units(migr_rec)) {
10f22854
AK		 12166 /* current reshape position [blocks] */
12167 unsigned long long current_position =
12168 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12169 * current_migr_unit(migr_rec);
10f22854
AK		 12170 unsigned long long border;
12171
b915c95f
AK		 12172 /* Check that array hasn't become failed.
12173 */
12174 degraded = check_degradation_change(sra, fds, degraded);
12175 if (degraded > 1) {
7a862a02 12176 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12177 goto abort;
12178 }
12179
10f22854
AK		 12180 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12181
12182 if ((current_position + next_step) > max_position)
12183 next_step = max_position - current_position;
12184
92144abf 12185 start = current_position * 512;
10f22854 12186
942e1cdb 12187 /* align reading start to old geometry */
10f22854
AK		 12188 start_buf_shift = start % old_data_stripe_length;
12189 start_src = start - start_buf_shift;
12190
12191 border = (start_src / odata) - (start / ndata);
12192 border /= 512;
12193 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12194 /* save critical stripes to buf
12195 * start - start address of current unit
12196 * to backup [bytes]
12197 * start_src - start address of current unit
12198 * to backup alligned to source array
12199 * [bytes]
12200 */
594dc1b8 12201 unsigned long long next_step_filler;
10f22854
AK		 12202 unsigned long long copy_length = next_step * 512;
12203
12204 /* allign copy area length to stripe in old geometry */
12205 next_step_filler = ((copy_length + start_buf_shift)
12206 % old_data_stripe_length);
12207 if (next_step_filler)
12208 next_step_filler = (old_data_stripe_length
12209 - next_step_filler);
7a862a02 12210 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		 12211 start, start_src, copy_length,
12212 start_buf_shift, next_step_filler);
12213
12214 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12215 chunk, map_src->raid_level,
12216 source_layout, 0, NULL, start_src,
10f22854
AK		 12217 copy_length +
12218 next_step_filler + start_buf_shift,
12219 buf)) {
7a862a02 12220 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12221 goto abort;
12222 }
12223 /* Convert data to destination format and store it
12224 * in backup general migration area
12225 */
12226 if (save_backup_imsm(st, dev, sra,
aea93171 12227 buf + start_buf_shift, copy_length)) {
7a862a02 12228 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12229 goto abort;
12230 }
12231 if (save_checkpoint_imsm(st, sra,
12232 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12233 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12234 goto abort;
12235 }
8016a6d4
AK		 12236 } else {
12237 /* set next step to use whole border area */
12238 border /= next_step;
12239 if (border > 1)
12240 next_step *= border;
10f22854
AK		 12241 }
12242 /* When data backed up, checkpoint stored,
12243 * kick the kernel to reshape unit of data
12244 */
12245 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12246 /* limit next step to array max position */
12247 if (next_step > max_position)
12248 next_step = max_position;
10f22854
AK		 12249 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12250 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12251 sra->reshape_progress = next_step;
10f22854
AK		 12252
12253 /* wait until reshape finish */
c85338c6 12254 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12255 dprintf("wait_for_reshape_imsm returned error!\n");
12256 goto abort;
12257 }
84d11e6c
N		 12258 if (sigterm)
12259 goto abort;
10f22854 12260
0228d92c
AK		 12261 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12262 /* ignore error == 2, this can mean end of reshape here
12263 */
7a862a02 12264 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12265 goto abort;
12266 }
12267
12268 }
12269
71e5411e
PB		 12270 /* clear migr_rec on disks after successful migration */
12271 struct dl *d;
12272
85337573 12273 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12274 for (d = super->disks; d; d = d->next) {
12275 if (d->index < 0 || is_failed(&d->disk))
12276 continue;
12277 unsigned long long dsize;
12278
12279 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12280 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12281 SEEK_SET) >= 0) {
466070ad 12282 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12283 MIGR_REC_BUF_SECTORS*sector_size) !=
12284 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12285 perror("Write migr_rec failed");
12286 }
12287 }
12288
10f22854
AK		 12289 /* return '1' if done */
12290 ret_val = 1;
895ffd99
MT		 12291
12292 /* After the reshape eliminate size mismatch in metadata.
12293 * Don't update md/component_size here, volume hasn't
12294 * to take whole space. It is allowed by kernel.
12295 * md/component_size will be set propoperly after next assembly.
12296 */
12297 imsm_fix_size_mismatch(st, subarray_index);
12298
10f22854
AK		 12299abort:
12300 free(buf);
942e1cdb
N		 12301 /* See Grow.c: abort_reshape() for further explanation */
12302 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12303 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12304 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12305
12306 return ret_val;
999b4972 12307}
0c21b485 12308
cdddbdbc 12309struct superswitch super_imsm = {
cdddbdbc
DW		 12310 .examine_super = examine_super_imsm,
12311 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12312 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12313 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12314 .detail_super = detail_super_imsm,
12315 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12316 .write_init_super = write_init_super_imsm,
0e600426
N		 12317 .validate_geometry = validate_geometry_imsm,
12318 .add_to_super = add_to_super_imsm,
1a64be56 12319 .remove_from_super = remove_from_super_imsm,
d665cc31 12320 .detail_platform = detail_platform_imsm,
e50cf220 12321 .export_detail_platform = export_detail_platform_imsm,
33414a01 12322 .kill_subarray = kill_subarray_imsm,
aa534678 12323 .update_subarray = update_subarray_imsm,
2b959fbf 12324 .load_container = load_container_imsm,
71204a50
N		 12325 .default_geometry = default_geometry_imsm,
12326 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12327 .reshape_super = imsm_reshape_super,
12328 .manage_reshape = imsm_manage_reshape,
9e2d750d 12329 .recover_backup = recover_backup_imsm,
27156a57 12330 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12331 .match_home = match_home_imsm,
12332 .uuid_from_super= uuid_from_super_imsm,
12333 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12334 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12335 .update_super = update_super_imsm,
12336
12337 .avail_size = avail_size_imsm,
fbfdcb06 12338 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12339
12340 .compare_super = compare_super_imsm,
12341
12342 .load_super = load_super_imsm,
bf5a934a 12343 .init_super = init_super_imsm,
e683ca88 12344 .store_super = store_super_imsm,
cdddbdbc
DW		 12345 .free_super = free_super_imsm,
12346 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12347 .container_content = container_content_imsm,
0c21b485 12348 .validate_container = validate_container_imsm,
cdddbdbc 12349
2432ce9b
AP		 12350 .write_init_ppl = write_init_ppl_imsm,
12351 .validate_ppl = validate_ppl_imsm,
12352
cdddbdbc 12353 .external = 1,
4cce4069 12354 .name = "imsm",
845dea95
NB		 12355
12356/* for mdmon */
12357 .open_new = imsm_open_new,
ed9d66aa 12358 .set_array_state= imsm_set_array_state,
845dea95
NB		 12359 .set_disk = imsm_set_disk,
12360 .sync_metadata = imsm_sync_metadata,
88758e9d 12361 .activate_spare = imsm_activate_spare,
e8319a19 12362 .process_update = imsm_process_update,
8273f55e 12363 .prepare_update = imsm_prepare_update,
6f50473f 12364 .record_bad_block = imsm_record_badblock,
c07a5a4f 12365 .clear_bad_block = imsm_clear_badblock,
928f1424 12366 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12367};
Unnamed repository; edit this file 'description' to name the repository.