]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
imsm: add generic method to resolve "device" links
[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
fbc42556
JR		 99/*
100 * Internal Write-intent bitmap is stored in the same area where PPL.
101 * Both features are mutually exclusive, so it is not an issue.
102 * The first 8KiB of the area are reserved and shall not be used.
103 */
104#define IMSM_BITMAP_AREA_RESERVED_SIZE 8192
105
106#define IMSM_BITMAP_HEADER_OFFSET (IMSM_BITMAP_AREA_RESERVED_SIZE)
107#define IMSM_BITMAP_HEADER_SIZE MAX_SECTOR_SIZE
108
109#define IMSM_BITMAP_START_OFFSET (IMSM_BITMAP_HEADER_OFFSET + IMSM_BITMAP_HEADER_SIZE)
110#define IMSM_BITMAP_AREA_SIZE (MULTIPLE_PPL_AREA_SIZE_IMSM - IMSM_BITMAP_START_OFFSET)
111#define IMSM_BITMAP_AND_HEADER_SIZE (IMSM_BITMAP_AREA_SIZE + IMSM_BITMAP_HEADER_SIZE)
112
113#define IMSM_DEFAULT_BITMAP_CHUNKSIZE (64 * 1024 * 1024)
114#define IMSM_DEFAULT_BITMAP_DAEMON_SLEEP 5
115
761e3bd9
N		 116/*
117 * This macro let's us ensure that no-one accidentally
118 * changes the size of a struct
119 */
120#define ASSERT_SIZE(_struct, size) \
121static inline void __assert_size_##_struct(void) \
122{ \
123 switch (0) { \
124 case 0: break; \
125 case (sizeof(struct _struct) == size): break; \
126 } \
127}
128
cdddbdbc
DW		 129/* Disk configuration info. */
130#define IMSM_MAX_DEVICES 255
131struct imsm_disk {
132 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 133 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 134 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 135#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
136#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
137#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 138#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 139 __u32 status; /* 0xF0 - 0xF3 */
1011e834 140 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 141 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
142#define IMSM_DISK_FILLERS 3
143 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 144};
761e3bd9 145ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 146
3b451610
AK		 147/* map selector for map managment
148 */
238c0a71
AK		 149#define MAP_0 0
150#define MAP_1 1
151#define MAP_X -1
3b451610 152
cdddbdbc
DW		 153/* RAID map configuration infos. */
154struct imsm_map {
5551b113
CA		 155 __u32 pba_of_lba0_lo; /* start address of partition */
156 __u32 blocks_per_member_lo;/* blocks per member */
157 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 158 __u16 blocks_per_strip;
159 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
160#define IMSM_T_STATE_NORMAL 0
161#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 162#define IMSM_T_STATE_DEGRADED 2
163#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 164 __u8 raid_level;
165#define IMSM_T_RAID0 0
166#define IMSM_T_RAID1 1
167#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
168 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 169 __u8 num_domains; /* number of parity domains */
170 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 171 __u8 ddf;
5551b113
CA		 172 __u32 pba_of_lba0_hi;
173 __u32 blocks_per_member_hi;
174 __u32 num_data_stripes_hi;
175 __u32 filler[4]; /* expansion area */
7eef0453 176#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 177 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 178 * top byte contains some flags
179 */
761e3bd9
N		 180};
181ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 182
183struct imsm_vol {
4036e7ee 184 __u32 curr_migr_unit_lo;
fe7ed8cb 185 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 186 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 187#define MIGR_INIT 0
188#define MIGR_REBUILD 1
189#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
190#define MIGR_GEN_MIGR 3
191#define MIGR_STATE_CHANGE 4
1484e727 192#define MIGR_REPAIR 5
cdddbdbc 193 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 194#define RAIDVOL_CLEAN 0
195#define RAIDVOL_DIRTY 1
196#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 197 __u8 dirty;
fe7ed8cb
DW		 198 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
199 __u16 verify_errors; /* number of mismatches */
200 __u16 bad_blocks; /* number of bad blocks during verify */
4036e7ee
MT		 201 __u32 curr_migr_unit_hi;
202 __u32 filler[3];
cdddbdbc
DW		 203 struct imsm_map map[1];
204 /* here comes another one if migr_state */
761e3bd9
N		 205};
206ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 207
208struct imsm_dev {
fe7ed8cb 209 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 210 __u32 size_low;
211 __u32 size_high;
fe7ed8cb
DW		 212#define DEV_BOOTABLE __cpu_to_le32(0x01)
213#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
214#define DEV_READ_COALESCING __cpu_to_le32(0x04)
215#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
216#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
217#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
218#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
219#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
220#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
221#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
222#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
223#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
224#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 225 __u32 status; /* Persistent RaidDev status */
226 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 227 __u8 migr_priority;
228 __u8 num_sub_vols;
229 __u8 tid;
230 __u8 cng_master_disk;
231 __u16 cache_policy;
232 __u8 cng_state;
233 __u8 cng_sub_state;
2432ce9b
AP		 234 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
235
236 /* NVM_EN */
237 __u8 nv_cache_mode;
238 __u8 nv_cache_flags;
239
240 /* Unique Volume Id of the NvCache Volume associated with this volume */
241 __u32 nvc_vol_orig_family_num;
242 __u16 nvc_vol_raid_dev_num;
243
244#define RWH_OFF 0
245#define RWH_DISTRIBUTED 1
246#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 247#define RWH_MULTIPLE_DISTRIBUTED 3
248#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
249#define RWH_MULTIPLE_OFF 5
fbc42556 250#define RWH_BITMAP 6
2432ce9b
AP		 251 __u8 rwh_policy; /* Raid Write Hole Policy */
252 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
253 __u8 filler1;
254
255#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 256 __u32 filler[IMSM_DEV_FILLERS];
257 struct imsm_vol vol;
761e3bd9
N		 258};
259ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 260
261struct imsm_super {
262 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
263 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
264 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
265 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
266 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 267 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
268 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 269 __u8 num_disks; /* 0x38 Number of configured disks */
270 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 271 __u8 error_log_pos; /* 0x3A */
272 __u8 fill[1]; /* 0x3B */
273 __u32 cache_size; /* 0x3c - 0x40 in mb */
274 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
275 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
276 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 277 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
278 * volume IDs for raid_dev created in this array
279 * (starts at 1)
280 */
281 __u16 filler1; /* 0x4E - 0x4F */
e48aed3c
AP		 282 __u64 creation_time; /* 0x50 - 0x57 Array creation time */
283#define IMSM_FILLERS 32
284 __u32 filler[IMSM_FILLERS]; /* 0x58 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 285 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
286 /* here comes imsm_dev[num_raid_devs] */
604b746f 287 /* here comes BBM logs */
761e3bd9
N		 288};
289ASSERT_SIZE(imsm_super, 264)
cdddbdbc 290
604b746f 291#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 292#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
293#define BBM_LOG_SIGNATURE 0xabadb10c
294
295struct bbm_log_block_addr {
296 __u16 w1;
297 __u32 dw1;
298} __attribute__ ((__packed__));
604b746f
JD		 299
300struct bbm_log_entry {
8d67477f
TM		 301 __u8 marked_count; /* Number of blocks marked - 1 */
302 __u8 disk_ordinal; /* Disk entry within the imsm_super */
303 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 304} __attribute__ ((__packed__));
305
306struct bbm_log {
307 __u32 signature; /* 0xABADB10C */
308 __u32 entry_count;
8d67477f 309 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 310};
311ASSERT_SIZE(bbm_log, 2040)
604b746f 312
cdddbdbc 313static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 314
b53bfba6
TM		 315#define BLOCKS_PER_KB (1024/512)
316
8e59f3d8
AK		 317#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
318
319#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
320
de44e46f
PB		 321#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
322#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
323 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 324 */
325
8e59f3d8
AK		 326#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
327 * be recovered using srcMap */
328#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
329 * already been migrated and must
330 * be recovered from checkpoint area */
2432ce9b 331
c2462068 332#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 333
8e59f3d8
AK		 334struct migr_record {
335 __u32 rec_status; /* Status used to determine how to restart
336 * migration in case it aborts
337 * in some fashion */
9f421827 338 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 339 __u32 family_num; /* Family number of MPB
340 * containing the RaidDev
341 * that is migrating */
342 __u32 ascending_migr; /* True if migrating in increasing
343 * order of lbas */
344 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
345 __u32 dest_depth_per_unit; /* Num member blocks each destMap
346 * member disk
347 * advances per unit-of-operation */
9f421827
PB		 348 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
349 __u32 dest_1st_member_lba_lo; /* First member lba on first
350 * stripe of destination */
351 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 352 __u32 post_migr_vol_cap; /* Size of volume after
353 * migration completes */
354 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
355 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
356 * migration ckpt record was read from
357 * (for recovered migrations) */
9f421827
PB		 358 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
359 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
360 * high order 32 bits */
361 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
362 * destination - high order 32 bits */
363 __u32 num_migr_units_hi; /* Total num migration units-of-op
364 * high order 32 bits */
4036e7ee 365 __u32 filler[16];
761e3bd9 366};
4036e7ee 367ASSERT_SIZE(migr_record, 128)
8e59f3d8 368
ec50f7b6
LM		 369struct md_list {
370 /* usage marker:
371 * 1: load metadata
372 * 2: metadata does not match
373 * 4: already checked
374 */
375 int used;
376 char *devname;
377 int found;
378 int container;
379 dev_t st_rdev;
380 struct md_list *next;
381};
382
e7b84f9d 383#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 384
1484e727
DW		 385static __u8 migr_type(struct imsm_dev *dev)
386{
387 if (dev->vol.migr_type == MIGR_VERIFY &&
388 dev->status & DEV_VERIFY_AND_FIX)
389 return MIGR_REPAIR;
390 else
391 return dev->vol.migr_type;
392}
393
394static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
395{
396 /* for compatibility with older oroms convert MIGR_REPAIR, into
397 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
398 */
399 if (migr_type == MIGR_REPAIR) {
400 dev->vol.migr_type = MIGR_VERIFY;
401 dev->status |= DEV_VERIFY_AND_FIX;
402 } else {
403 dev->vol.migr_type = migr_type;
404 dev->status &= ~DEV_VERIFY_AND_FIX;
405 }
406}
407
f36a9ecd 408static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 409{
f36a9ecd 410 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 411}
cdddbdbc 412
f36a9ecd
PB		 413static unsigned int mpb_sectors(struct imsm_super *mpb,
414 unsigned int sector_size)
87eb16df 415{
f36a9ecd 416 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 417}
418
ba2de7ba
DW		 419struct intel_dev {
420 struct imsm_dev *dev;
421 struct intel_dev *next;
f21e18ca 422 unsigned index;
ba2de7ba
DW		 423};
424
88654014
LM		 425struct intel_hba {
426 enum sys_dev_type type;
427 char *path;
428 char *pci_id;
429 struct intel_hba *next;
430};
431
1a64be56
LM		 432enum action {
433 DISK_REMOVE = 1,
434 DISK_ADD
435};
cdddbdbc
DW		 436/* internal representation of IMSM metadata */
437struct intel_super {
438 union {
949c47a0
DW		 439 void *buf; /* O_DIRECT buffer for reading/writing metadata */
440 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 441 };
8e59f3d8
AK		 442 union {
443 void *migr_rec_buf; /* buffer for I/O operations */
444 struct migr_record *migr_rec; /* migration record */
445 };
51d83f5d
AK		 446 int clean_migration_record_by_mdmon; /* when reshape is switched to next
447 array, it indicates that mdmon is allowed to clean migration
448 record */
949c47a0 449 size_t len; /* size of the 'buf' allocation */
bbab0940 450 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 451 void *next_buf; /* for realloc'ing buf from the manager */
452 size_t next_len;
c2c087e6 453 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 454 int current_vol; /* index of raid device undergoing creation */
5551b113 455 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 456 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 457 struct intel_dev *devlist;
fa7bb6f8 458 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 459 struct dl {
460 struct dl *next;
461 int index;
462 __u8 serial[MAX_RAID_SERIAL_LEN];
463 int major, minor;
464 char *devname;
b9f594fe 465 struct imsm_disk disk;
cdddbdbc 466 int fd;
0dcecb2e
DW		 467 int extent_cnt;
468 struct extent *e; /* for determining freespace @ create */
efb30e7f 469 int raiddisk; /* slot to fill in autolayout */
1a64be56 470 enum action action;
ca0748fa 471 } *disks, *current_disk;
1a64be56
LM		 472 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
473 active */
47ee5a45 474 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 475 struct bbm_log *bbm_log;
88654014 476 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 477 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 478 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 479 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 480};
481
482struct intel_disk {
483 struct imsm_disk disk;
484 #define IMSM_UNKNOWN_OWNER (-1)
485 int owner;
486 struct intel_disk *next;
cdddbdbc
DW		 487};
488
c2c087e6
DW		 489struct extent {
490 unsigned long long start, size;
491};
492
694575e7
KW		 493/* definitions of reshape process types */
494enum imsm_reshape_type {
495 CH_TAKEOVER,
b5347799 496 CH_MIGRATION,
7abc9871 497 CH_ARRAY_SIZE,
694575e7
KW		 498};
499
88758e9d
DW		 500/* definition of messages passed to imsm_process_update */
501enum imsm_update_type {
502 update_activate_spare,
8273f55e 503 update_create_array,
33414a01 504 update_kill_array,
aa534678 505 update_rename_array,
1a64be56 506 update_add_remove_disk,
78b10e66 507 update_reshape_container_disks,
48c5303a 508 update_reshape_migration,
2d40f3a1
AK		 509 update_takeover,
510 update_general_migration_checkpoint,
f3871fdc 511 update_size_change,
bbab0940 512 update_prealloc_badblocks_mem,
e6e9dd3f 513 update_rwh_policy,
88758e9d
DW		 514};
515
516struct imsm_update_activate_spare {
517 enum imsm_update_type type;
d23fe947 518 struct dl *dl;
88758e9d
DW		 519 int slot;
520 int array;
521 struct imsm_update_activate_spare *next;
522};
523
78b10e66 524struct geo_params {
4dd2df09 525 char devnm[32];
78b10e66 526 char *dev_name;
d04f65f4 527 unsigned long long size;
78b10e66
N		 528 int level;
529 int layout;
530 int chunksize;
531 int raid_disks;
532};
533
bb025c2f
KW		 534enum takeover_direction {
535 R10_TO_R0,
536 R0_TO_R10
537};
538struct imsm_update_takeover {
539 enum imsm_update_type type;
540 int subarray;
541 enum takeover_direction direction;
542};
78b10e66
N		 543
544struct imsm_update_reshape {
545 enum imsm_update_type type;
546 int old_raid_disks;
547 int new_raid_disks;
48c5303a
PC		 548
549 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
550};
551
552struct imsm_update_reshape_migration {
553 enum imsm_update_type type;
554 int old_raid_disks;
555 int new_raid_disks;
556 /* fields for array migration changes
557 */
558 int subdev;
559 int new_level;
560 int new_layout;
4bba0439 561 int new_chunksize;
48c5303a 562
d195167d 563 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 564};
565
f3871fdc
AK		 566struct imsm_update_size_change {
567 enum imsm_update_type type;
568 int subdev;
569 long long new_size;
570};
571
2d40f3a1
AK		 572struct imsm_update_general_migration_checkpoint {
573 enum imsm_update_type type;
4036e7ee 574 __u64 curr_migr_unit;
2d40f3a1
AK		 575};
576
54c2c1ea
DW		 577struct disk_info {
578 __u8 serial[MAX_RAID_SERIAL_LEN];
579};
580
8273f55e
DW		 581struct imsm_update_create_array {
582 enum imsm_update_type type;
8273f55e 583 int dev_idx;
6a3e913e 584 struct imsm_dev dev;
8273f55e
DW		 585};
586
33414a01
DW		 587struct imsm_update_kill_array {
588 enum imsm_update_type type;
589 int dev_idx;
590};
591
aa534678
DW		 592struct imsm_update_rename_array {
593 enum imsm_update_type type;
594 __u8 name[MAX_RAID_SERIAL_LEN];
595 int dev_idx;
596};
597
1a64be56 598struct imsm_update_add_remove_disk {
43dad3d6
DW		 599 enum imsm_update_type type;
600};
601
bbab0940
TM		 602struct imsm_update_prealloc_bb_mem {
603 enum imsm_update_type type;
604};
605
e6e9dd3f
AP		 606struct imsm_update_rwh_policy {
607 enum imsm_update_type type;
608 int new_policy;
609 int dev_idx;
610};
611
88654014
LM		 612static const char *_sys_dev_type[] = {
613 [SYS_DEV_UNKNOWN] = "Unknown",
614 [SYS_DEV_SAS] = "SAS",
614902f6 615 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 616 [SYS_DEV_NVME] = "NVMe",
617 [SYS_DEV_VMD] = "VMD"
88654014
LM		 618};
619
620const char *get_sys_dev_type(enum sys_dev_type type)
621{
622 if (type >= SYS_DEV_MAX)
623 type = SYS_DEV_UNKNOWN;
624
625 return _sys_dev_type[type];
626}
627
628static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
629{
503975b9
N		 630 struct intel_hba *result = xmalloc(sizeof(*result));
631
632 result->type = device->type;
633 result->path = xstrdup(device->path);
634 result->next = NULL;
635 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
636 result->pci_id++;
637
88654014
LM		 638 return result;
639}
640
641static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
642{
594dc1b8
JS		 643 struct intel_hba *result;
644
88654014
LM		 645 for (result = hba; result; result = result->next) {
646 if (result->type == device->type && strcmp(result->path, device->path) == 0)
647 break;
648 }
649 return result;
650}
651
b4cf4cba 652static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 653{
654 struct intel_hba *hba;
655
656 /* check if disk attached to Intel HBA */
657 hba = find_intel_hba(super->hba, device);
658 if (hba != NULL)
659 return 1;
660 /* Check if HBA is already attached to super */
661 if (super->hba == NULL) {
662 super->hba = alloc_intel_hba(device);
663 return 1;
6b781d33
AP		 664 }
665
666 hba = super->hba;
667 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 668 * Do not support HBA types mixing
6b781d33
AP		 669 */
670 if (device->type != hba->type)
88654014 671 return 2;
6b781d33
AP		 672
673 /* Multiple same type HBAs can be used if they share the same OROM */
674 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
675
676 if (device_orom != super->orom)
677 return 2;
678
679 while (hba->next)
680 hba = hba->next;
681
682 hba->next = alloc_intel_hba(device);
683 return 1;
88654014
LM		 684}
685
686static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
687{
9bc4ae77 688 struct sys_dev *list, *elem;
88654014
LM		 689 char *disk_path;
690
691 if ((list = find_intel_devices()) == NULL)
692 return 0;
693
694 if (fd < 0)
695 disk_path = (char *) devname;
696 else
7c798f87 697 disk_path = diskfd_to_devpath(fd, 1, NULL);
88654014 698
9bc4ae77 699 if (!disk_path)
88654014 700 return 0;
88654014 701
9bc4ae77
N		 702 for (elem = list; elem; elem = elem->next)
703 if (path_attached_to_hba(disk_path, elem->path))
88654014 704 return elem;
9bc4ae77 705
88654014
LM		 706 if (disk_path != devname)
707 free(disk_path);
88654014
LM		 708
709 return NULL;
710}
711
d424212e
N		 712static int find_intel_hba_capability(int fd, struct intel_super *super,
713 char *devname);
f2f5c343 714
cdddbdbc
DW		 715static struct supertype *match_metadata_desc_imsm(char *arg)
716{
717 struct supertype *st;
718
719 if (strcmp(arg, "imsm") != 0 &&
720 strcmp(arg, "default") != 0
721 )
722 return NULL;
723
503975b9 724 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 725 st->ss = &super_imsm;
726 st->max_devs = IMSM_MAX_DEVICES;
727 st->minor_version = 0;
728 st->sb = NULL;
729 return st;
730}
731
cdddbdbc
DW		 732static __u8 *get_imsm_version(struct imsm_super *mpb)
733{
734 return &mpb->sig[MPB_SIG_LEN];
735}
736
949c47a0
DW		 737/* retrieve a disk directly from the anchor when the anchor is known to be
738 * up-to-date, currently only at load time
739 */
740static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 741{
949c47a0 742 if (index >= mpb->num_disks)
cdddbdbc
DW		 743 return NULL;
744 return &mpb->disk[index];
745}
746
95d07a2c
LM		 747/* retrieve the disk description based on a index of the disk
748 * in the sub-array
749 */
750static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 751{
b9f594fe
DW		 752 struct dl *d;
753
754 for (d = super->disks; d; d = d->next)
755 if (d->index == index)
95d07a2c
LM		 756 return d;
757
758 return NULL;
759}
760/* retrieve a disk from the parsed metadata */
761static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
762{
763 struct dl *dl;
764
765 dl = get_imsm_dl_disk(super, index);
766 if (dl)
767 return &dl->disk;
768
b9f594fe 769 return NULL;
949c47a0
DW		 770}
771
772/* generate a checksum directly from the anchor when the anchor is known to be
773 * up-to-date, currently only at load or write_super after coalescing
774 */
775static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 776{
777 __u32 end = mpb->mpb_size / sizeof(end);
778 __u32 *p = (__u32 *) mpb;
779 __u32 sum = 0;
780
5d500228
N		 781 while (end--) {
782 sum += __le32_to_cpu(*p);
97f734fd
N		 783 p++;
784 }
cdddbdbc 785
5d500228 786 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 787}
788
a965f303
DW		 789static size_t sizeof_imsm_map(struct imsm_map *map)
790{
791 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
792}
793
794struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 795{
5e7b0330
AK		 796 /* A device can have 2 maps if it is in the middle of a migration.
797 * If second_map is:
238c0a71
AK		 798 * MAP_0 - we return the first map
799 * MAP_1 - we return the second map if it exists, else NULL
800 * MAP_X - we return the second map if it exists, else the first
5e7b0330 801 */
a965f303 802 struct imsm_map *map = &dev->vol.map[0];
9535fc47 803 struct imsm_map *map2 = NULL;
a965f303 804
9535fc47
AK		 805 if (dev->vol.migr_state)
806 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 807
9535fc47 808 switch (second_map) {
3b451610 809 case MAP_0:
9535fc47 810 break;
3b451610 811 case MAP_1:
9535fc47
AK		 812 map = map2;
813 break;
238c0a71 814 case MAP_X:
9535fc47
AK		 815 if (map2)
816 map = map2;
817 break;
9535fc47
AK		 818 default:
819 map = NULL;
820 }
821 return map;
5e7b0330 822
a965f303 823}
cdddbdbc 824
3393c6af
DW		 825/* return the size of the device.
826 * migr_state increases the returned size if map[0] were to be duplicated
827 */
828static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 829{
830 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 831 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 832
833 /* migrating means an additional map */
a965f303 834 if (dev->vol.migr_state)
238c0a71 835 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 836 else if (migr_state)
238c0a71 837 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 838
839 return size;
840}
841
54c2c1ea
DW		 842/* retrieve disk serial number list from a metadata update */
843static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
844{
845 void *u = update;
846 struct disk_info *inf;
847
848 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
849 sizeof_imsm_dev(&update->dev, 0);
850
851 return inf;
852}
54c2c1ea 853
949c47a0 854static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 855{
856 int offset;
857 int i;
858 void *_mpb = mpb;
859
949c47a0 860 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 861 return NULL;
862
863 /* devices start after all disks */
864 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
865
866 for (i = 0; i <= index; i++)
867 if (i == index)
868 return _mpb + offset;
869 else
3393c6af 870 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 871
872 return NULL;
873}
874
949c47a0
DW		 875static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
876{
ba2de7ba
DW		 877 struct intel_dev *dv;
878
949c47a0
DW		 879 if (index >= super->anchor->num_raid_devs)
880 return NULL;
ba2de7ba
DW		 881 for (dv = super->devlist; dv; dv = dv->next)
882 if (dv->index == index)
883 return dv->dev;
884 return NULL;
949c47a0
DW		 885}
886
8d67477f
TM		 887static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
888 *addr)
889{
890 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
891 __le16_to_cpu(addr->w1));
892}
893
894static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
895{
896 struct bbm_log_block_addr addr;
897
898 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
899 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
900 return addr;
901}
902
8d67477f
TM		 903/* get size of the bbm log */
904static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
905{
906 if (!log || log->entry_count == 0)
907 return 0;
908
909 return sizeof(log->signature) +
910 sizeof(log->entry_count) +
911 log->entry_count * sizeof(struct bbm_log_entry);
912}
6f50473f
TM		 913
914/* check if bad block is not partially stored in bbm log */
915static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
916 long long sector, const int length, __u32 *pos)
917{
918 __u32 i;
919
920 for (i = *pos; i < log->entry_count; i++) {
921 struct bbm_log_entry *entry = &log->marked_block_entries[i];
922 unsigned long long bb_start;
923 unsigned long long bb_end;
924
925 bb_start = __le48_to_cpu(&entry->defective_block_start);
926 bb_end = bb_start + (entry->marked_count + 1);
927
928 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
929 (bb_end <= sector + length)) {
930 *pos = i;
931 return 1;
932 }
933 }
934 return 0;
935}
936
937/* record new bad block in bbm log */
938static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
939 long long sector, int length)
940{
941 int new_bb = 0;
942 __u32 pos = 0;
943 struct bbm_log_entry *entry = NULL;
944
945 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
946 struct bbm_log_entry *e = &log->marked_block_entries[pos];
947
948 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
949 (__le48_to_cpu(&e->defective_block_start) == sector)) {
950 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
951 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
952 pos = pos + 1;
953 continue;
954 }
955 entry = e;
956 break;
957 }
958
959 if (entry) {
960 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
961 BBM_LOG_MAX_LBA_ENTRY_VAL;
962 entry->defective_block_start = __cpu_to_le48(sector);
963 entry->marked_count = cnt - 1;
964 if (cnt == length)
965 return 1;
966 sector += cnt;
967 length -= cnt;
968 }
969
970 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
971 BBM_LOG_MAX_LBA_ENTRY_VAL;
972 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
973 return 0;
974
975 while (length > 0) {
976 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
977 BBM_LOG_MAX_LBA_ENTRY_VAL;
978 struct bbm_log_entry *entry =
979 &log->marked_block_entries[log->entry_count];
980
981 entry->defective_block_start = __cpu_to_le48(sector);
982 entry->marked_count = cnt - 1;
983 entry->disk_ordinal = idx;
984
985 sector += cnt;
986 length -= cnt;
987
988 log->entry_count++;
989 }
990
991 return new_bb;
992}
c07a5a4f 993
4c9e8c1e
TM		 994/* clear all bad blocks for given disk */
995static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
996{
997 __u32 i = 0;
998
999 while (i < log->entry_count) {
1000 struct bbm_log_entry *entries = log->marked_block_entries;
1001
1002 if (entries[i].disk_ordinal == idx) {
1003 if (i < log->entry_count - 1)
1004 entries[i] = entries[log->entry_count - 1];
1005 log->entry_count--;
1006 } else {
1007 i++;
1008 }
1009 }
1010}
1011
c07a5a4f
TM		 1012/* clear given bad block */
1013static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
1014 long long sector, const int length) {
1015 __u32 i = 0;
1016
1017 while (i < log->entry_count) {
1018 struct bbm_log_entry *entries = log->marked_block_entries;
1019
1020 if ((entries[i].disk_ordinal == idx) &&
1021 (__le48_to_cpu(&entries[i].defective_block_start) ==
1022 sector) && (entries[i].marked_count + 1 == length)) {
1023 if (i < log->entry_count - 1)
1024 entries[i] = entries[log->entry_count - 1];
1025 log->entry_count--;
1026 break;
1027 }
1028 i++;
1029 }
1030
1031 return 1;
1032}
8d67477f
TM		 1033
1034/* allocate and load BBM log from metadata */
1035static int load_bbm_log(struct intel_super *super)
1036{
1037 struct imsm_super *mpb = super->anchor;
1038 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1039
1040 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1041 if (!super->bbm_log)
1042 return 1;
1043
1044 if (bbm_log_size) {
1045 struct bbm_log *log = (void *)mpb +
1046 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1047
1048 __u32 entry_count;
1049
1050 if (bbm_log_size < sizeof(log->signature) +
1051 sizeof(log->entry_count))
1052 return 2;
1053
1054 entry_count = __le32_to_cpu(log->entry_count);
1055 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1056 (entry_count > BBM_LOG_MAX_ENTRIES))
1057 return 3;
1058
1059 if (bbm_log_size !=
1060 sizeof(log->signature) + sizeof(log->entry_count) +
1061 entry_count * sizeof(struct bbm_log_entry))
1062 return 4;
1063
1064 memcpy(super->bbm_log, log, bbm_log_size);
1065 } else {
1066 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1067 super->bbm_log->entry_count = 0;
1068 }
1069
1070 return 0;
1071}
1072
b12796be
TM		 1073/* checks if bad block is within volume boundaries */
1074static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1075 const unsigned long long start_sector,
1076 const unsigned long long size)
1077{
1078 unsigned long long bb_start;
1079 unsigned long long bb_end;
1080
1081 bb_start = __le48_to_cpu(&entry->defective_block_start);
1082 bb_end = bb_start + (entry->marked_count + 1);
1083
1084 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1085 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1086 return 1;
1087
1088 return 0;
1089}
1090
1091/* get list of bad blocks on a drive for a volume */
1092static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1093 const unsigned long long start_sector,
1094 const unsigned long long size,
1095 struct md_bb *bbs)
1096{
1097 __u32 count = 0;
1098 __u32 i;
1099
1100 for (i = 0; i < log->entry_count; i++) {
1101 const struct bbm_log_entry *ent =
1102 &log->marked_block_entries[i];
1103 struct md_bb_entry *bb;
1104
1105 if ((ent->disk_ordinal == idx) &&
1106 is_bad_block_in_volume(ent, start_sector, size)) {
1107
1108 if (!bbs->entries) {
1109 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1110 sizeof(*bb));
1111 if (!bbs->entries)
1112 break;
1113 }
1114
1115 bb = &bbs->entries[count++];
1116 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1117 bb->length = ent->marked_count + 1;
1118 }
1119 }
1120 bbs->count = count;
1121}
1122
98130f40
AK		 1123/*
1124 * for second_map:
238c0a71
AK		 1125 * == MAP_0 get first map
1126 * == MAP_1 get second map
1127 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1128 */
1129static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1130 int slot,
1131 int second_map)
7eef0453
DW		 1132{
1133 struct imsm_map *map;
1134
5e7b0330 1135 map = get_imsm_map(dev, second_map);
7eef0453 1136
ff077194
DW		 1137 /* top byte identifies disk under rebuild */
1138 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1139}
1140
1141#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1142static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1143{
98130f40 1144 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1145
1146 return ord_to_idx(ord);
7eef0453
DW		 1147}
1148
be73972f
DW		 1149static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1150{
1151 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1152}
1153
f21e18ca 1154static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1155{
1156 int slot;
1157 __u32 ord;
1158
1159 for (slot = 0; slot < map->num_members; slot++) {
1160 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1161 if (ord_to_idx(ord) == idx)
1162 return slot;
1163 }
1164
1165 return -1;
1166}
1167
cdddbdbc
DW		 1168static int get_imsm_raid_level(struct imsm_map *map)
1169{
1170 if (map->raid_level == 1) {
1171 if (map->num_members == 2)
1172 return 1;
1173 else
1174 return 10;
1175 }
1176
1177 return map->raid_level;
1178}
1179
c2c087e6
DW		 1180static int cmp_extent(const void *av, const void *bv)
1181{
1182 const struct extent *a = av;
1183 const struct extent *b = bv;
1184 if (a->start < b->start)
1185 return -1;
1186 if (a->start > b->start)
1187 return 1;
1188 return 0;
1189}
1190
0dcecb2e 1191static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1192{
c2c087e6 1193 int memberships = 0;
620b1713 1194 int i;
c2c087e6 1195
949c47a0
DW		 1196 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1197 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1198 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1199
620b1713
DW		 1200 if (get_imsm_disk_slot(map, dl->index) >= 0)
1201 memberships++;
c2c087e6 1202 }
0dcecb2e
DW		 1203
1204 return memberships;
1205}
1206
b81221b7
CA		 1207static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1208
486720e0 1209static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1210{
1211 if (lo == 0 || hi == 0)
1212 return 1;
486720e0
JS		 1213 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1214 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1215 return 0;
1216}
1217
1218static unsigned long long join_u32(__u32 lo, __u32 hi)
1219{
1220 return (unsigned long long)__le32_to_cpu(lo) |
1221 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1222}
1223
1224static unsigned long long total_blocks(struct imsm_disk *disk)
1225{
1226 if (disk == NULL)
1227 return 0;
1228 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1229}
1230
1231static unsigned long long pba_of_lba0(struct imsm_map *map)
1232{
1233 if (map == NULL)
1234 return 0;
1235 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1236}
1237
1238static unsigned long long blocks_per_member(struct imsm_map *map)
1239{
1240 if (map == NULL)
1241 return 0;
1242 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1243}
1244
1245static unsigned long long num_data_stripes(struct imsm_map *map)
1246{
1247 if (map == NULL)
1248 return 0;
1249 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1250}
1251
4036e7ee
MT		 1252static unsigned long long vol_curr_migr_unit(struct imsm_dev *dev)
1253{
1254 if (dev == NULL)
1255 return 0;
1256
1257 return join_u32(dev->vol.curr_migr_unit_lo, dev->vol.curr_migr_unit_hi);
1258}
1259
fcc2c9da
MD		 1260static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1261{
1262 if (dev == NULL)
1263 return 0;
1264 return join_u32(dev->size_low, dev->size_high);
1265}
1266
9f421827
PB		 1267static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1268{
1269 if (migr_rec == NULL)
1270 return 0;
1271 return join_u32(migr_rec->ckpt_area_pba_lo,
1272 migr_rec->ckpt_area_pba_hi);
1273}
1274
1275static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1276{
1277 if (migr_rec == NULL)
1278 return 0;
1279 return join_u32(migr_rec->curr_migr_unit_lo,
1280 migr_rec->curr_migr_unit_hi);
1281}
1282
1283static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1284{
1285 if (migr_rec == NULL)
1286 return 0;
1287 return join_u32(migr_rec->dest_1st_member_lba_lo,
1288 migr_rec->dest_1st_member_lba_hi);
1289}
1290
1291static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1292{
1293 if (migr_rec == NULL)
1294 return 0;
1295 return join_u32(migr_rec->num_migr_units_lo,
1296 migr_rec->num_migr_units_hi);
1297}
1298
5551b113
CA		 1299static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1300{
1301 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1302}
1303
1304static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1305{
1306 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1307}
1308
1309static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1310{
1311 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1312}
1313
1314static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1315{
1316 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1317}
1318
4036e7ee
MT		 1319static void set_vol_curr_migr_unit(struct imsm_dev *dev, unsigned long long n)
1320{
1321 if (dev == NULL)
1322 return;
1323
1324 split_ull(n, &dev->vol.curr_migr_unit_lo, &dev->vol.curr_migr_unit_hi);
1325}
1326
fcc2c9da
MD		 1327static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1328{
1329 split_ull(n, &dev->size_low, &dev->size_high);
1330}
1331
9f421827
PB		 1332static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1333 unsigned long long n)
1334{
1335 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1336}
1337
1338static void set_current_migr_unit(struct migr_record *migr_rec,
1339 unsigned long long n)
1340{
1341 split_ull(n, &migr_rec->curr_migr_unit_lo,
1342 &migr_rec->curr_migr_unit_hi);
1343}
1344
1345static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1346 unsigned long long n)
1347{
1348 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1349 &migr_rec->dest_1st_member_lba_hi);
1350}
1351
1352static void set_num_migr_units(struct migr_record *migr_rec,
1353 unsigned long long n)
1354{
1355 split_ull(n, &migr_rec->num_migr_units_lo,
1356 &migr_rec->num_migr_units_hi);
1357}
1358
44490938
MD		 1359static unsigned long long per_dev_array_size(struct imsm_map *map)
1360{
1361 unsigned long long array_size = 0;
1362
1363 if (map == NULL)
1364 return array_size;
1365
1366 array_size = num_data_stripes(map) * map->blocks_per_strip;
1367 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1368 array_size *= 2;
1369
1370 return array_size;
1371}
1372
05501181
PB		 1373static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1374 int get_minimal_reservation)
0dcecb2e
DW		 1375{
1376 /* find a list of used extents on the given physical device */
1377 struct extent *rv, *e;
620b1713 1378 int i;
0dcecb2e 1379 int memberships = count_memberships(dl, super);
b276dd33
DW		 1380 __u32 reservation;
1381
1382 /* trim the reserved area for spares, so they can join any array
1383 * regardless of whether the OROM has assigned sectors from the
1384 * IMSM_RESERVED_SECTORS region
1385 */
05501181 1386 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1387 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1388 else
1389 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1390
503975b9 1391 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1392 e = rv;
1393
949c47a0
DW		 1394 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1395 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1396 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1397
620b1713 1398 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1399 e->start = pba_of_lba0(map);
44490938 1400 e->size = per_dev_array_size(map);
620b1713 1401 e++;
c2c087e6
DW		 1402 }
1403 }
1404 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1405
1011e834 1406 /* determine the start of the metadata
14e8215b
DW		 1407 * when no raid devices are defined use the default
1408 * ...otherwise allow the metadata to truncate the value
1409 * as is the case with older versions of imsm
1410 */
1411 if (memberships) {
1412 struct extent *last = &rv[memberships - 1];
5551b113 1413 unsigned long long remainder;
14e8215b 1414
5551b113 1415 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1416 /* round down to 1k block to satisfy precision of the kernel
1417 * 'size' interface
1418 */
1419 remainder &= ~1UL;
1420 /* make sure remainder is still sane */
f21e18ca 1421 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1422 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1423 if (reservation > remainder)
1424 reservation = remainder;
1425 }
5551b113 1426 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1427 e->size = 0;
1428 return rv;
1429}
1430
14e8215b
DW		 1431/* try to determine how much space is reserved for metadata from
1432 * the last get_extents() entry, otherwise fallback to the
1433 * default
1434 */
1435static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1436{
1437 struct extent *e;
1438 int i;
1439 __u32 rv;
1440
1441 /* for spares just return a minimal reservation which will grow
1442 * once the spare is picked up by an array
1443 */
1444 if (dl->index == -1)
1445 return MPB_SECTOR_CNT;
1446
05501181 1447 e = get_extents(super, dl, 0);
14e8215b
DW		 1448 if (!e)
1449 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1450
1451 /* scroll to last entry */
1452 for (i = 0; e[i].size; i++)
1453 continue;
1454
5551b113 1455 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1456
1457 free(e);
1458
1459 return rv;
1460}
1461
25ed7e59
DW		 1462static int is_spare(struct imsm_disk *disk)
1463{
1464 return (disk->status & SPARE_DISK) == SPARE_DISK;
1465}
1466
1467static int is_configured(struct imsm_disk *disk)
1468{
1469 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1470}
1471
1472static int is_failed(struct imsm_disk *disk)
1473{
1474 return (disk->status & FAILED_DISK) == FAILED_DISK;
1475}
1476
2432ce9b
AP		 1477static int is_journal(struct imsm_disk *disk)
1478{
1479 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1480}
1481
b53bfba6
TM		 1482/* round array size down to closest MB and ensure it splits evenly
1483 * between members
1484 */
1485static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1486 disk_count)
1487{
1488 size /= disk_count;
1489 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1490 size *= disk_count;
1491
1492 return size;
1493}
1494
8b9cd157
MK		 1495static int able_to_resync(int raid_level, int missing_disks)
1496{
1497 int max_missing_disks = 0;
1498
1499 switch (raid_level) {
1500 case 10:
1501 max_missing_disks = 1;
1502 break;
1503 default:
1504 max_missing_disks = 0;
1505 }
1506 return missing_disks <= max_missing_disks;
1507}
1508
b81221b7
CA		 1509/* try to determine how much space is reserved for metadata from
1510 * the last get_extents() entry on the smallest active disk,
1511 * otherwise fallback to the default
1512 */
1513static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1514{
1515 struct extent *e;
1516 int i;
5551b113
CA		 1517 unsigned long long min_active;
1518 __u32 remainder;
b81221b7
CA		 1519 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1520 struct dl *dl, *dl_min = NULL;
1521
1522 if (!super)
1523 return rv;
1524
1525 min_active = 0;
1526 for (dl = super->disks; dl; dl = dl->next) {
1527 if (dl->index < 0)
1528 continue;
5551b113
CA		 1529 unsigned long long blocks = total_blocks(&dl->disk);
1530 if (blocks < min_active || min_active == 0) {
b81221b7 1531 dl_min = dl;
5551b113 1532 min_active = blocks;
b81221b7
CA		 1533 }
1534 }
1535 if (!dl_min)
1536 return rv;
1537
1538 /* find last lba used by subarrays on the smallest active disk */
05501181 1539 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1540 if (!e)
1541 return rv;
1542 for (i = 0; e[i].size; i++)
1543 continue;
1544
1545 remainder = min_active - e[i].start;
1546 free(e);
1547
1548 /* to give priority to recovery we should not require full
1549 IMSM_RESERVED_SECTORS from the spare */
1550 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1551
1552 /* if real reservation is smaller use that value */
1553 return (remainder < rv) ? remainder : rv;
1554}
1555
fbfdcb06
AO		 1556/*
1557 * Return minimum size of a spare and sector size
1558 * that can be used in this array
1559 */
1560int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1561{
1562 struct intel_super *super = st->sb;
1563 struct dl *dl;
1564 struct extent *e;
1565 int i;
fbfdcb06
AO		 1566 unsigned long long size = 0;
1567
1568 c->min_size = 0;
4b57ecf6 1569 c->sector_size = 0;
80e7f8c3
AC		 1570
1571 if (!super)
fbfdcb06 1572 return -EINVAL;
80e7f8c3
AC		 1573 /* find first active disk in array */
1574 dl = super->disks;
1575 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1576 dl = dl->next;
1577 if (!dl)
fbfdcb06 1578 return -EINVAL;
80e7f8c3 1579 /* find last lba used by subarrays */
05501181 1580 e = get_extents(super, dl, 0);
80e7f8c3 1581 if (!e)
fbfdcb06 1582 return -EINVAL;
80e7f8c3
AC		 1583 for (i = 0; e[i].size; i++)
1584 continue;
1585 if (i > 0)
fbfdcb06 1586 size = e[i-1].start + e[i-1].size;
80e7f8c3 1587 free(e);
b81221b7 1588
80e7f8c3 1589 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1590 size += imsm_min_reserved_sectors(super);
1591
1592 c->min_size = size * 512;
4b57ecf6 1593 c->sector_size = super->sector_size;
b81221b7 1594
fbfdcb06 1595 return 0;
80e7f8c3
AC		 1596}
1597
d1e02575
AK		 1598static int is_gen_migration(struct imsm_dev *dev);
1599
f36a9ecd
PB		 1600#define IMSM_4K_DIV 8
1601
c47b0ff6
AK		 1602static __u64 blocks_per_migr_unit(struct intel_super *super,
1603 struct imsm_dev *dev);
1e5c6983 1604
c47b0ff6
AK		 1605static void print_imsm_dev(struct intel_super *super,
1606 struct imsm_dev *dev,
1607 char *uuid,
1608 int disk_idx)
cdddbdbc
DW		 1609{
1610 __u64 sz;
0d80bb2f 1611 int slot, i;
238c0a71
AK		 1612 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1613 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1614 __u32 ord;
cdddbdbc
DW		 1615
1616 printf("\n");
1e7bc0ed 1617 printf("[%.16s]:\n", dev->volume);
ba1b3bc8 1618 printf(" Subarray : %d\n", super->current_vol);
44470971 1619 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1620 printf(" RAID Level : %d", get_imsm_raid_level(map));
1621 if (map2)
1622 printf(" <-- %d", get_imsm_raid_level(map2));
1623 printf("\n");
1624 printf(" Members : %d", map->num_members);
1625 if (map2)
1626 printf(" <-- %d", map2->num_members);
1627 printf("\n");
0d80bb2f
DW		 1628 printf(" Slots : [");
1629 for (i = 0; i < map->num_members; i++) {
238c0a71 1630 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1631 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1632 }
dd8bcb3b
AK		 1633 printf("]");
1634 if (map2) {
1635 printf(" <-- [");
1636 for (i = 0; i < map2->num_members; i++) {
238c0a71 1637 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1638 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1639 }
1640 printf("]");
1641 }
1642 printf("\n");
7095bccb
AK		 1643 printf(" Failed disk : ");
1644 if (map->failed_disk_num == 0xff)
1645 printf("none");
1646 else
1647 printf("%i", map->failed_disk_num);
1648 printf("\n");
620b1713
DW		 1649 slot = get_imsm_disk_slot(map, disk_idx);
1650 if (slot >= 0) {
238c0a71 1651 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1652 printf(" This Slot : %d%s\n", slot,
1653 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1654 } else
cdddbdbc 1655 printf(" This Slot : ?\n");
84918897 1656 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1657 sz = imsm_dev_size(dev);
84918897
MK		 1658 printf(" Array Size : %llu%s\n",
1659 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1660 human_size(sz * 512));
5551b113 1661 sz = blocks_per_member(map);
84918897
MK		 1662 printf(" Per Dev Size : %llu%s\n",
1663 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1664 human_size(sz * 512));
5551b113
CA		 1665 printf(" Sector Offset : %llu\n",
1666 pba_of_lba0(map));
1667 printf(" Num Stripes : %llu\n",
1668 num_data_stripes(map));
dd8bcb3b 1669 printf(" Chunk Size : %u KiB",
cdddbdbc 1670 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1671 if (map2)
1672 printf(" <-- %u KiB",
1673 __le16_to_cpu(map2->blocks_per_strip) / 2);
1674 printf("\n");
cdddbdbc 1675 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1676 printf(" Migrate State : ");
1484e727
DW		 1677 if (dev->vol.migr_state) {
1678 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1679 printf("initialize\n");
1484e727 1680 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1681 printf("rebuild\n");
1484e727 1682 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1683 printf("check\n");
1484e727 1684 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1685 printf("general migration\n");
1484e727 1686 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1687 printf("state change\n");
1484e727 1688 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1689 printf("repair\n");
1484e727 1690 else
8655a7b1
DW		 1691 printf("<unknown:%d>\n", migr_type(dev));
1692 } else
1693 printf("idle\n");
3393c6af
DW		 1694 printf(" Map State : %s", map_state_str[map->map_state]);
1695 if (dev->vol.migr_state) {
238c0a71 1696 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1697
b10b37b8 1698 printf(" <-- %s", map_state_str[map->map_state]);
4036e7ee 1699 printf("\n Checkpoint : %llu ", vol_curr_migr_unit(dev));
089f9d79 1700 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1701 printf("(N/A)");
1702 else
1703 printf("(%llu)", (unsigned long long)
1704 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1705 }
1706 printf("\n");
2432ce9b
AP		 1707 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1708 "dirty" : "clean");
1709 printf(" RWH Policy : ");
c2462068 1710 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1711 printf("off\n");
1712 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1713 printf("PPL distributed\n");
1714 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1715 printf("PPL journaling drive\n");
c2462068
PB		 1716 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1717 printf("Multiple distributed PPLs\n");
1718 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1719 printf("Multiple PPLs on journaling drive\n");
fbc42556
JR		 1720 else if (dev->rwh_policy == RWH_BITMAP)
1721 printf("Write-intent bitmap\n");
2432ce9b
AP		 1722 else
1723 printf("<unknown:%d>\n", dev->rwh_policy);
ba1b3bc8
AP		 1724
1725 printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num);
cdddbdbc
DW		 1726}
1727
ef5c214e
MK		 1728static void print_imsm_disk(struct imsm_disk *disk,
1729 int index,
1730 __u32 reserved,
1731 unsigned int sector_size) {
1f24f035 1732 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1733 __u64 sz;
1734
0ec1f4e8 1735 if (index < -1 || !disk)
e9d82038
DW		 1736 return;
1737
cdddbdbc 1738 printf("\n");
1f24f035 1739 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1740 if (index >= 0)
1741 printf(" Disk%02d Serial : %s\n", index, str);
1742 else
1743 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1744 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1745 is_configured(disk) ? " active" : "",
1746 is_failed(disk) ? " failed" : "",
1747 is_journal(disk) ? " journal" : "");
cdddbdbc 1748 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1749 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1750 printf(" Usable Size : %llu%s\n",
1751 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1752 human_size(sz * 512));
1753}
1754
de44e46f
PB		 1755void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1756{
1757 struct migr_record *migr_rec = super->migr_rec;
1758
1759 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1760 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1761 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1762 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1763 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1764 set_migr_chkp_area_pba(migr_rec,
1765 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1766 set_migr_dest_1st_member_lba(migr_rec,
1767 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1768}
1769
f36a9ecd
PB		 1770void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1771{
1772 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1773}
1774
1775void convert_to_4k(struct intel_super *super)
1776{
1777 struct imsm_super *mpb = super->anchor;
1778 struct imsm_disk *disk;
1779 int i;
e4467bc7 1780 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1781
1782 for (i = 0; i < mpb->num_disks ; i++) {
1783 disk = __get_imsm_disk(mpb, i);
1784 /* disk */
1785 convert_to_4k_imsm_disk(disk);
1786 }
1787 for (i = 0; i < mpb->num_raid_devs; i++) {
1788 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1789 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1790 /* dev */
fcc2c9da 1791 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
4036e7ee
MT		 1792 set_vol_curr_migr_unit(dev,
1793 vol_curr_migr_unit(dev) / IMSM_4K_DIV);
f36a9ecd
PB		 1794
1795 /* map0 */
1796 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1797 map->blocks_per_strip /= IMSM_4K_DIV;
1798 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1799
1800 if (dev->vol.migr_state) {
1801 /* map1 */
1802 map = get_imsm_map(dev, MAP_1);
1803 set_blocks_per_member(map,
1804 blocks_per_member(map)/IMSM_4K_DIV);
1805 map->blocks_per_strip /= IMSM_4K_DIV;
1806 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1807 }
1808 }
e4467bc7
TM		 1809 if (bbm_log_size) {
1810 struct bbm_log *log = (void *)mpb +
1811 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1812 __u32 i;
1813
1814 for (i = 0; i < log->entry_count; i++) {
1815 struct bbm_log_entry *entry =
1816 &log->marked_block_entries[i];
1817
1818 __u8 count = entry->marked_count + 1;
1819 unsigned long long sector =
1820 __le48_to_cpu(&entry->defective_block_start);
1821
1822 entry->defective_block_start =
1823 __cpu_to_le48(sector/IMSM_4K_DIV);
1824 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1825 }
1826 }
f36a9ecd
PB		 1827
1828 mpb->check_sum = __gen_imsm_checksum(mpb);
1829}
1830
520e69e2
AK		 1831void examine_migr_rec_imsm(struct intel_super *super)
1832{
1833 struct migr_record *migr_rec = super->migr_rec;
1834 struct imsm_super *mpb = super->anchor;
1835 int i;
1836
1837 for (i = 0; i < mpb->num_raid_devs; i++) {
1838 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1839 struct imsm_map *map;
b4ab44d8 1840 int slot = -1;
3136abe5 1841
520e69e2
AK		 1842 if (is_gen_migration(dev) == 0)
1843 continue;
1844
1845 printf("\nMigration Record Information:");
3136abe5 1846
44bfe6df
AK		 1847 /* first map under migration */
1848 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1849 if (map)
1850 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1851 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1852 printf(" Empty\n ");
1853 printf("Examine one of first two disks in array\n");
1854 break;
1855 }
1856 printf("\n Status : ");
1857 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1858 printf("Normal\n");
1859 else
1860 printf("Contains Data\n");
9f421827
PB		 1861 printf(" Current Unit : %llu\n",
1862 current_migr_unit(migr_rec));
520e69e2
AK		 1863 printf(" Family : %u\n",
1864 __le32_to_cpu(migr_rec->family_num));
1865 printf(" Ascending : %u\n",
1866 __le32_to_cpu(migr_rec->ascending_migr));
1867 printf(" Blocks Per Unit : %u\n",
1868 __le32_to_cpu(migr_rec->blocks_per_unit));
1869 printf(" Dest. Depth Per Unit : %u\n",
1870 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1871 printf(" Checkpoint Area pba : %llu\n",
1872 migr_chkp_area_pba(migr_rec));
1873 printf(" First member lba : %llu\n",
1874 migr_dest_1st_member_lba(migr_rec));
1875 printf(" Total Number of Units : %llu\n",
1876 get_num_migr_units(migr_rec));
1877 printf(" Size of volume : %llu\n",
1878 join_u32(migr_rec->post_migr_vol_cap,
1879 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1880 printf(" Record was read from : %u\n",
1881 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1882
1883 break;
1884 }
1885}
f36a9ecd 1886
de44e46f
PB		 1887void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1888{
1889 struct migr_record *migr_rec = super->migr_rec;
1890
1891 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1892 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1893 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1894 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1895 &migr_rec->post_migr_vol_cap,
1896 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1897 set_migr_chkp_area_pba(migr_rec,
1898 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1899 set_migr_dest_1st_member_lba(migr_rec,
1900 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1901}
1902
f36a9ecd
PB		 1903void convert_from_4k(struct intel_super *super)
1904{
1905 struct imsm_super *mpb = super->anchor;
1906 struct imsm_disk *disk;
1907 int i;
e4467bc7 1908 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1909
1910 for (i = 0; i < mpb->num_disks ; i++) {
1911 disk = __get_imsm_disk(mpb, i);
1912 /* disk */
1913 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1914 }
1915
1916 for (i = 0; i < mpb->num_raid_devs; i++) {
1917 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1918 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1919 /* dev */
fcc2c9da 1920 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
4036e7ee
MT		 1921 set_vol_curr_migr_unit(dev,
1922 vol_curr_migr_unit(dev) * IMSM_4K_DIV);
f36a9ecd
PB		 1923
1924 /* map0 */
1925 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1926 map->blocks_per_strip *= IMSM_4K_DIV;
1927 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1928
1929 if (dev->vol.migr_state) {
1930 /* map1 */
1931 map = get_imsm_map(dev, MAP_1);
1932 set_blocks_per_member(map,
1933 blocks_per_member(map)*IMSM_4K_DIV);
1934 map->blocks_per_strip *= IMSM_4K_DIV;
1935 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1936 }
1937 }
e4467bc7
TM		 1938 if (bbm_log_size) {
1939 struct bbm_log *log = (void *)mpb +
1940 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1941 __u32 i;
1942
1943 for (i = 0; i < log->entry_count; i++) {
1944 struct bbm_log_entry *entry =
1945 &log->marked_block_entries[i];
1946
1947 __u8 count = entry->marked_count + 1;
1948 unsigned long long sector =
1949 __le48_to_cpu(&entry->defective_block_start);
1950
1951 entry->defective_block_start =
1952 __cpu_to_le48(sector*IMSM_4K_DIV);
1953 entry->marked_count = count*IMSM_4K_DIV - 1;
1954 }
1955 }
f36a9ecd
PB		 1956
1957 mpb->check_sum = __gen_imsm_checksum(mpb);
1958}
1959
19482bcc
AK		 1960/*******************************************************************************
1961 * function: imsm_check_attributes
1962 * Description: Function checks if features represented by attributes flags
1011e834 1963 * are supported by mdadm.
19482bcc
AK		 1964 * Parameters:
1965 * attributes - Attributes read from metadata
1966 * Returns:
1011e834
N		 1967 * 0 - passed attributes contains unsupported features flags
1968 * 1 - all features are supported
19482bcc
AK		 1969 ******************************************************************************/
1970static int imsm_check_attributes(__u32 attributes)
1971{
1972 int ret_val = 1;
418f9b36
N		 1973 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1974
1975 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1976
1977 not_supported &= attributes;
1978 if (not_supported) {
e7b84f9d 1979 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1980 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1981 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1982 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1983 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1984 }
1985 if (not_supported & MPB_ATTRIB_2TB) {
1986 dprintf("\t\tMPB_ATTRIB_2TB\n");
1987 not_supported ^= MPB_ATTRIB_2TB;
1988 }
1989 if (not_supported & MPB_ATTRIB_RAID0) {
1990 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1991 not_supported ^= MPB_ATTRIB_RAID0;
1992 }
1993 if (not_supported & MPB_ATTRIB_RAID1) {
1994 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1995 not_supported ^= MPB_ATTRIB_RAID1;
1996 }
1997 if (not_supported & MPB_ATTRIB_RAID10) {
1998 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1999 not_supported ^= MPB_ATTRIB_RAID10;
2000 }
2001 if (not_supported & MPB_ATTRIB_RAID1E) {
2002 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
2003 not_supported ^= MPB_ATTRIB_RAID1E;
2004 }
2005 if (not_supported & MPB_ATTRIB_RAID5) {
2006 dprintf("\t\tMPB_ATTRIB_RAID5\n");
2007 not_supported ^= MPB_ATTRIB_RAID5;
2008 }
2009 if (not_supported & MPB_ATTRIB_RAIDCNG) {
2010 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
2011 not_supported ^= MPB_ATTRIB_RAIDCNG;
2012 }
2013 if (not_supported & MPB_ATTRIB_BBM) {
2014 dprintf("\t\tMPB_ATTRIB_BBM\n");
2015 not_supported ^= MPB_ATTRIB_BBM;
2016 }
2017 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
2018 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
2019 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
2020 }
2021 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
2022 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
2023 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
2024 }
2025 if (not_supported & MPB_ATTRIB_2TB_DISK) {
2026 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
2027 not_supported ^= MPB_ATTRIB_2TB_DISK;
2028 }
2029 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
2030 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
2031 not_supported ^= MPB_ATTRIB_NEVER_USE2;
2032 }
2033 if (not_supported & MPB_ATTRIB_NEVER_USE) {
2034 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
2035 not_supported ^= MPB_ATTRIB_NEVER_USE;
2036 }
2037
2038 if (not_supported)
1ade5cc1 2039 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 2040
2041 ret_val = 0;
2042 }
2043
2044 return ret_val;
2045}
2046
a5d85af7 2047static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2048
cdddbdbc
DW		 2049static void examine_super_imsm(struct supertype *st, char *homehost)
2050{
2051 struct intel_super *super = st->sb;
949c47a0 2052 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2053 char str[MAX_SIGNATURE_LENGTH];
2054 int i;
27fd6274
DW		 2055 struct mdinfo info;
2056 char nbuf[64];
cdddbdbc 2057 __u32 sum;
14e8215b 2058 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2059 struct dl *dl;
e48aed3c 2060 time_t creation_time;
27fd6274 2061
618f4e6d
XN		 2062 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2063 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2064 printf(" Magic : %s\n", str);
cdddbdbc 2065 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2066 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2067 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2068 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
e48aed3c
AP		 2069 creation_time = __le64_to_cpu(mpb->creation_time);
2070 printf(" Creation Time : %.24s\n",
2071 creation_time ? ctime(&creation_time) : "Unknown");
19482bcc
AK		 2072 printf(" Attributes : ");
2073 if (imsm_check_attributes(mpb->attributes))
2074 printf("All supported\n");
2075 else
2076 printf("not supported\n");
a5d85af7 2077 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2078 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2079 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2080 sum = __le32_to_cpu(mpb->check_sum);
2081 printf(" Checksum : %08x %s\n", sum,
949c47a0 2082 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2083 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2084 printf(" Disks : %d\n", mpb->num_disks);
2085 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2086 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2087 super->disks->index, reserved, super->sector_size);
8d67477f 2088 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2089 struct bbm_log *log = super->bbm_log;
2090
2091 printf("\n");
2092 printf("Bad Block Management Log:\n");
2093 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2094 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2095 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2096 }
44470971
DW		 2097 for (i = 0; i < mpb->num_raid_devs; i++) {
2098 struct mdinfo info;
2099 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2100
2101 super->current_vol = i;
a5d85af7 2102 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2103 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2104 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2105 }
cdddbdbc
DW		 2106 for (i = 0; i < mpb->num_disks; i++) {
2107 if (i == super->disks->index)
2108 continue;
ef5c214e
MK		 2109 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2110 super->sector_size);
cdddbdbc 2111 }
94827db3 2112
0ec1f4e8
DW		 2113 for (dl = super->disks; dl; dl = dl->next)
2114 if (dl->index == -1)
ef5c214e
MK		 2115 print_imsm_disk(&dl->disk, -1, reserved,
2116 super->sector_size);
520e69e2
AK		 2117
2118 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2119}
2120
061f2c6a 2121static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2122{
27fd6274 2123 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2124 struct mdinfo info;
2125 char nbuf[64];
1e7bc0ed 2126 struct intel_super *super = st->sb;
1e7bc0ed 2127
0d5a423f
DW		 2128 if (!super->anchor->num_raid_devs) {
2129 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2130 return;
0d5a423f 2131 }
ff54de6e 2132
a5d85af7 2133 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2134 fname_from_uuid(st, &info, nbuf, ':');
2135 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2136}
2137
2138static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2139{
2140 /* We just write a generic IMSM ARRAY entry */
2141 struct mdinfo info;
2142 char nbuf[64];
2143 char nbuf1[64];
2144 struct intel_super *super = st->sb;
2145 int i;
2146
2147 if (!super->anchor->num_raid_devs)
2148 return;
2149
a5d85af7 2150 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2151 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2152 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2153 struct imsm_dev *dev = get_imsm_dev(super, i);
2154
2155 super->current_vol = i;
a5d85af7 2156 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2157 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2158 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2159 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2160 }
cdddbdbc
DW		 2161}
2162
9d84c8ea
DW		 2163static void export_examine_super_imsm(struct supertype *st)
2164{
2165 struct intel_super *super = st->sb;
2166 struct imsm_super *mpb = super->anchor;
2167 struct mdinfo info;
2168 char nbuf[64];
2169
a5d85af7 2170 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2171 fname_from_uuid(st, &info, nbuf, ':');
2172 printf("MD_METADATA=imsm\n");
2173 printf("MD_LEVEL=container\n");
2174 printf("MD_UUID=%s\n", nbuf+5);
2175 printf("MD_DEVICES=%u\n", mpb->num_disks);
e48aed3c 2176 printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time));
9d84c8ea
DW		 2177}
2178
b771faef
BK		 2179static void detail_super_imsm(struct supertype *st, char *homehost,
2180 char *subarray)
cdddbdbc 2181{
3ebe00a1
DW		 2182 struct mdinfo info;
2183 char nbuf[64];
b771faef
BK		 2184 struct intel_super *super = st->sb;
2185 int temp_vol = super->current_vol;
2186
2187 if (subarray)
2188 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2189
a5d85af7 2190 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2191 fname_from_uuid(st, &info, nbuf, ':');
65884368 2192 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2193
2194 super->current_vol = temp_vol;
cdddbdbc
DW		 2195}
2196
b771faef 2197static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2198{
ff54de6e
N		 2199 struct mdinfo info;
2200 char nbuf[64];
b771faef
BK		 2201 struct intel_super *super = st->sb;
2202 int temp_vol = super->current_vol;
2203
2204 if (subarray)
2205 super->current_vol = strtoul(subarray, NULL, 10);
2206
a5d85af7 2207 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2208 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2209 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2210
2211 super->current_vol = temp_vol;
cdddbdbc 2212}
d665cc31 2213
6da53c0e
BK		 2214static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2215 size_t serial_buf_len);
d665cc31
DW		 2216static void fd2devname(int fd, char *name);
2217
120dc887 2218static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2219{
120dc887
LM		 2220 /* dump an unsorted list of devices attached to AHCI Intel storage
2221 * controller, as well as non-connected ports
d665cc31
DW		 2222 */
2223 int hba_len = strlen(hba_path) + 1;
2224 struct dirent *ent;
2225 DIR *dir;
2226 char *path = NULL;
2227 int err = 0;
2228 unsigned long port_mask = (1 << port_count) - 1;
2229
f21e18ca 2230 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2231 if (verbose > 0)
e7b84f9d 2232 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2233 return 2;
2234 }
2235
2236 /* scroll through /sys/dev/block looking for devices attached to
2237 * this hba
2238 */
2239 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2240 if (!dir)
2241 return 1;
2242
2243 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2244 int fd;
2245 char model[64];
2246 char vendor[64];
2247 char buf[1024];
2248 int major, minor;
2249 char *device;
2250 char *c;
2251 int port;
2252 int type;
2253
2254 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2255 continue;
7c798f87 2256 path = devt_to_devpath(makedev(major, minor), 1, NULL);
d665cc31
DW		 2257 if (!path)
2258 continue;
2259 if (!path_attached_to_hba(path, hba_path)) {
2260 free(path);
2261 path = NULL;
2262 continue;
2263 }
2264
2265 /* retrieve the scsi device type */
2266 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2267 if (verbose > 0)
e7b84f9d 2268 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2269 err = 2;
2270 break;
2271 }
2272 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2273 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2274 if (verbose > 0)
e7b84f9d 2275 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2276 path);
2277 err = 2;
2278 free(device);
2279 break;
2280 }
2281 type = strtoul(buf, NULL, 10);
2282
2283 /* if it's not a disk print the vendor and model */
2284 if (!(type == 0 || type == 7 || type == 14)) {
2285 vendor[0] = '\0';
2286 model[0] = '\0';
2287 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2288 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2289 strncpy(vendor, buf, sizeof(vendor));
2290 vendor[sizeof(vendor) - 1] = '\0';
2291 c = (char *) &vendor[sizeof(vendor) - 1];
2292 while (isspace(*c) || *c == '\0')
2293 *c-- = '\0';
2294
2295 }
2296 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2297 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2298 strncpy(model, buf, sizeof(model));
2299 model[sizeof(model) - 1] = '\0';
2300 c = (char *) &model[sizeof(model) - 1];
2301 while (isspace(*c) || *c == '\0')
2302 *c-- = '\0';
2303 }
2304
2305 if (vendor[0] && model[0])
2306 sprintf(buf, "%.64s %.64s", vendor, model);
2307 else
2308 switch (type) { /* numbers from hald/linux/device.c */
2309 case 1: sprintf(buf, "tape"); break;
2310 case 2: sprintf(buf, "printer"); break;
2311 case 3: sprintf(buf, "processor"); break;
2312 case 4:
2313 case 5: sprintf(buf, "cdrom"); break;
2314 case 6: sprintf(buf, "scanner"); break;
2315 case 8: sprintf(buf, "media_changer"); break;
2316 case 9: sprintf(buf, "comm"); break;
2317 case 12: sprintf(buf, "raid"); break;
2318 default: sprintf(buf, "unknown");
2319 }
2320 } else
2321 buf[0] = '\0';
2322 free(device);
2323
2324 /* chop device path to 'host%d' and calculate the port number */
2325 c = strchr(&path[hba_len], '/');
4e5e717d 2326 if (!c) {
ba728be7 2327 if (verbose > 0)
e7b84f9d 2328 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2329 err = 2;
2330 break;
2331 }
d665cc31 2332 *c = '\0';
0858eccf
AP		 2333 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2334 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2335 port -= host_base;
2336 else {
ba728be7 2337 if (verbose > 0) {
d665cc31 2338 *c = '/'; /* repair the full string */
e7b84f9d 2339 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2340 path);
2341 }
2342 err = 2;
2343 break;
2344 }
2345
2346 /* mark this port as used */
2347 port_mask &= ~(1 << port);
2348
2349 /* print out the device information */
2350 if (buf[0]) {
2351 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2352 continue;
2353 }
2354
2355 fd = dev_open(ent->d_name, O_RDONLY);
2356 if (fd < 0)
2357 printf(" Port%d : - disk info unavailable -\n", port);
2358 else {
2359 fd2devname(fd, buf);
2360 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2361 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2362 sizeof(buf)) == 0)
2363 printf(" (%s)\n", buf);
d665cc31 2364 else
664d5325 2365 printf(" ()\n");
4dab422a 2366 close(fd);
d665cc31 2367 }
d665cc31
DW		 2368 free(path);
2369 path = NULL;
2370 }
2371 if (path)
2372 free(path);
2373 if (dir)
2374 closedir(dir);
2375 if (err == 0) {
2376 int i;
2377
2378 for (i = 0; i < port_count; i++)
2379 if (port_mask & (1 << i))
2380 printf(" Port%d : - no device attached -\n", i);
2381 }
2382
2383 return err;
2384}
2385
6da53c0e 2386static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2387{
6da53c0e 2388 char buf[1024];
d835518b 2389 char *device_path;
60f0f54d
PB		 2390 struct dirent *ent;
2391 DIR *dir;
6da53c0e 2392 int fd;
60f0f54d 2393
6da53c0e 2394 dir = opendir("/sys/block/");
b9135011 2395 if (!dir)
b5eece69 2396 return 1;
b9135011
JS		 2397
2398 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e 2399 if (strstr(ent->d_name, "nvme")) {
d835518b
BK		 2400 fd = open_dev(ent->d_name);
2401 if (fd < 0)
6da53c0e 2402 continue;
60f0f54d 2403
d835518b
BK		 2404 if (!imsm_is_nvme_supported(fd, 0)) {
2405 if (fd >= 0)
2406 close(fd);
2407 continue;
2408 }
2409
7c798f87 2410 device_path = diskfd_to_devpath(fd, 1, NULL);
d835518b
BK		 2411 if (!device_path) {
2412 close(fd);
2413 continue;
2414 }
2415
2416 if (path_attached_to_hba(device_path, hba->path)) {
6da53c0e
BK		 2417 fd2devname(fd, buf);
2418 if (hba->type == SYS_DEV_VMD)
2419 printf(" NVMe under VMD : %s", buf);
2420 else if (hba->type == SYS_DEV_NVME)
2421 printf(" NVMe Device : %s", buf);
2422 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2423 sizeof(buf)))
2424 printf(" (%s)\n", buf);
2425 else
2426 printf("()\n");
6da53c0e 2427 }
d835518b
BK		 2428 free(device_path);
2429 close(fd);
60f0f54d 2430 }
60f0f54d
PB		 2431 }
2432
b9135011 2433 closedir(dir);
b5eece69 2434 return 0;
60f0f54d
PB		 2435}
2436
120dc887
LM		 2437static void print_found_intel_controllers(struct sys_dev *elem)
2438{
2439 for (; elem; elem = elem->next) {
e7b84f9d 2440 pr_err("found Intel(R) ");
120dc887
LM		 2441 if (elem->type == SYS_DEV_SATA)
2442 fprintf(stderr, "SATA ");
155cbb4c
LM		 2443 else if (elem->type == SYS_DEV_SAS)
2444 fprintf(stderr, "SAS ");
0858eccf
AP		 2445 else if (elem->type == SYS_DEV_NVME)
2446 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2447
2448 if (elem->type == SYS_DEV_VMD)
2449 fprintf(stderr, "VMD domain");
2450 else
2451 fprintf(stderr, "RAID controller");
2452
120dc887
LM		 2453 if (elem->pci_id)
2454 fprintf(stderr, " at %s", elem->pci_id);
2455 fprintf(stderr, ".\n");
2456 }
2457 fflush(stderr);
2458}
2459
120dc887
LM		 2460static int ahci_get_port_count(const char *hba_path, int *port_count)
2461{
2462 struct dirent *ent;
2463 DIR *dir;
2464 int host_base = -1;
2465
2466 *port_count = 0;
2467 if ((dir = opendir(hba_path)) == NULL)
2468 return -1;
2469
2470 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2471 int host;
2472
0858eccf
AP		 2473 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2474 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2475 continue;
2476 if (*port_count == 0)
2477 host_base = host;
2478 else if (host < host_base)
2479 host_base = host;
2480
2481 if (host + 1 > *port_count + host_base)
2482 *port_count = host + 1 - host_base;
2483 }
2484 closedir(dir);
2485 return host_base;
2486}
2487
a891a3c2
LM		 2488static void print_imsm_capability(const struct imsm_orom *orom)
2489{
0858eccf
AP		 2490 printf(" Platform : Intel(R) ");
2491 if (orom->capabilities == 0 && orom->driver_features == 0)
2492 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2493 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2494 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2495 else
2496 printf("Rapid Storage Technology%s\n",
2497 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2498 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2499 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2500 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2501 printf(" RAID Levels :%s%s%s%s%s\n",
2502 imsm_orom_has_raid0(orom) ? " raid0" : "",
2503 imsm_orom_has_raid1(orom) ? " raid1" : "",
2504 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2505 imsm_orom_has_raid10(orom) ? " raid10" : "",
2506 imsm_orom_has_raid5(orom) ? " raid5" : "");
2507 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2508 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2509 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2510 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2511 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2512 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2513 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2514 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2515 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2516 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2517 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2518 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2519 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2520 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2521 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2522 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2523 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2524 printf(" 2TB volumes :%s supported\n",
2525 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2526 printf(" 2TB disks :%s supported\n",
2527 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2528 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2529 printf(" Max Volumes : %d per array, %d per %s\n",
2530 orom->vpa, orom->vphba,
2531 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2532 return;
2533}
2534
e50cf220
MN		 2535static void print_imsm_capability_export(const struct imsm_orom *orom)
2536{
2537 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2538 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2539 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2540 orom->hotfix_ver, orom->build);
e50cf220
MN		 2541 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2542 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2543 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2544 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2545 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2546 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2547 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2548 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2549 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2550 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2551 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2552 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2553 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2554 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2555 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2556 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2557 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2558 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2559 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2560 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2561 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2562 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2563 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2564 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2565 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2566 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2567 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2568 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2569}
2570
9eafa1de 2571static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2572{
2573 /* There are two components to imsm platform support, the ahci SATA
2574 * controller and the option-rom. To find the SATA controller we
2575 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2576 * controller with the Intel vendor id is present. This approach
2577 * allows mdadm to leverage the kernel's ahci detection logic, with the
2578 * caveat that if ahci.ko is not loaded mdadm will not be able to
2579 * detect platform raid capabilities. The option-rom resides in a
2580 * platform "Adapter ROM". We scan for its signature to retrieve the
2581 * platform capabilities. If raid support is disabled in the BIOS the
2582 * option-rom capability structure will not be available.
2583 */
d665cc31 2584 struct sys_dev *list, *hba;
d665cc31
DW		 2585 int host_base = 0;
2586 int port_count = 0;
9eafa1de 2587 int result=1;
d665cc31 2588
5615172f 2589 if (enumerate_only) {
a891a3c2 2590 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2591 return 0;
a891a3c2
LM		 2592 list = find_intel_devices();
2593 if (!list)
2594 return 2;
2595 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2596 if (find_imsm_capability(hba)) {
2597 result = 0;
a891a3c2
LM		 2598 break;
2599 }
9eafa1de 2600 else
6b781d33 2601 result = 2;
a891a3c2 2602 }
a891a3c2 2603 return result;
5615172f
DW		 2604 }
2605
155cbb4c
LM		 2606 list = find_intel_devices();
2607 if (!list) {
ba728be7 2608 if (verbose > 0)
7a862a02 2609 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2610 return 2;
ba728be7 2611 } else if (verbose > 0)
155cbb4c 2612 print_found_intel_controllers(list);
d665cc31 2613
a891a3c2 2614 for (hba = list; hba; hba = hba->next) {
0858eccf 2615 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2616 continue;
0858eccf 2617 if (!find_imsm_capability(hba)) {
60f0f54d 2618 char buf[PATH_MAX];
e7b84f9d 2619 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2620 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2621 get_sys_dev_type(hba->type));
0858eccf
AP		 2622 continue;
2623 }
2624 result = 0;
2625 }
2626
2627 if (controller_path && result == 1) {
2628 pr_err("no active Intel(R) RAID controller found under %s\n",
2629 controller_path);
2630 return result;
2631 }
2632
5e1d6128 2633 const struct orom_entry *entry;
0858eccf 2634
5e1d6128 2635 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2636 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2637 print_imsm_capability(&entry->orom);
32716c51
PB		 2638 printf(" 3rd party NVMe :%s supported\n",
2639 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2640 for (hba = list; hba; hba = hba->next) {
2641 if (hba->type == SYS_DEV_VMD) {
2642 char buf[PATH_MAX];
60f0f54d
PB		 2643 printf(" I/O Controller : %s (%s)\n",
2644 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2645 if (print_nvme_info(hba)) {
b5eece69
PB		 2646 if (verbose > 0)
2647 pr_err("failed to get devices attached to VMD domain.\n");
2648 result |= 2;
2649 }
60f0f54d
PB		 2650 }
2651 }
07cb1e57 2652 printf("\n");
60f0f54d
PB		 2653 continue;
2654 }
0858eccf 2655
60f0f54d
PB		 2656 print_imsm_capability(&entry->orom);
2657 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2658 for (hba = list; hba; hba = hba->next) {
2659 if (hba->type == SYS_DEV_NVME)
6da53c0e 2660 print_nvme_info(hba);
0858eccf 2661 }
60f0f54d 2662 printf("\n");
0858eccf
AP		 2663 continue;
2664 }
2665
2666 struct devid_list *devid;
5e1d6128 2667 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2668 hba = device_by_id(devid->devid);
2669 if (!hba)
2670 continue;
2671
9eafa1de
MN		 2672 printf(" I/O Controller : %s (%s)\n",
2673 hba->path, get_sys_dev_type(hba->type));
2674 if (hba->type == SYS_DEV_SATA) {
2675 host_base = ahci_get_port_count(hba->path, &port_count);
2676 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2677 if (verbose > 0)
7a862a02 2678 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2679 result |= 2;
2680 }
120dc887
LM		 2681 }
2682 }
0858eccf 2683 printf("\n");
d665cc31 2684 }
155cbb4c 2685
120dc887 2686 return result;
d665cc31 2687}
e50cf220 2688
9eafa1de 2689static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2690{
e50cf220
MN		 2691 struct sys_dev *list, *hba;
2692 int result=1;
2693
2694 list = find_intel_devices();
2695 if (!list) {
2696 if (verbose > 0)
2697 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2698 result = 2;
e50cf220
MN		 2699 return result;
2700 }
2701
2702 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2703 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2704 continue;
60f0f54d
PB		 2705 if (!find_imsm_capability(hba) && verbose > 0) {
2706 char buf[PATH_MAX];
2707 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2708 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2709 }
0858eccf 2710 else
e50cf220 2711 result = 0;
e50cf220
MN		 2712 }
2713
5e1d6128 2714 const struct orom_entry *entry;
0858eccf 2715
60f0f54d
PB		 2716 for (entry = orom_entries; entry; entry = entry->next) {
2717 if (entry->type == SYS_DEV_VMD) {
2718 for (hba = list; hba; hba = hba->next)
2719 print_imsm_capability_export(&entry->orom);
2720 continue;
2721 }
5e1d6128 2722 print_imsm_capability_export(&entry->orom);
60f0f54d 2723 }
0858eccf 2724
e50cf220
MN		 2725 return result;
2726}
2727
cdddbdbc
DW		 2728static int match_home_imsm(struct supertype *st, char *homehost)
2729{
5115ca67
DW		 2730 /* the imsm metadata format does not specify any host
2731 * identification information. We return -1 since we can never
2732 * confirm nor deny whether a given array is "meant" for this
148acb7b 2733 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2734 * exclude member disks that do not belong, and we rely on
2735 * mdadm.conf to specify the arrays that should be assembled.
2736 * Auto-assembly may still pick up "foreign" arrays.
2737 */
cdddbdbc 2738
9362c1c8 2739 return -1;
cdddbdbc
DW		 2740}
2741
2742static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2743{
51006d85
N		 2744 /* The uuid returned here is used for:
2745 * uuid to put into bitmap file (Create, Grow)
2746 * uuid for backup header when saving critical section (Grow)
2747 * comparing uuids when re-adding a device into an array
2748 * In these cases the uuid required is that of the data-array,
2749 * not the device-set.
2750 * uuid to recognise same set when adding a missing device back
2751 * to an array. This is a uuid for the device-set.
1011e834 2752 *
51006d85
N		 2753 * For each of these we can make do with a truncated
2754 * or hashed uuid rather than the original, as long as
2755 * everyone agrees.
2756 * In each case the uuid required is that of the data-array,
2757 * not the device-set.
43dad3d6 2758 */
51006d85
N		 2759 /* imsm does not track uuid's so we synthesis one using sha1 on
2760 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2761 * - the orig_family_num of the container
51006d85
N		 2762 * - the index number of the volume
2763 * - the 'serial' number of the volume.
2764 * Hopefully these are all constant.
2765 */
2766 struct intel_super *super = st->sb;
43dad3d6 2767
51006d85
N		 2768 char buf[20];
2769 struct sha1_ctx ctx;
2770 struct imsm_dev *dev = NULL;
148acb7b 2771 __u32 family_num;
51006d85 2772
148acb7b
DW		 2773 /* some mdadm versions failed to set ->orig_family_num, in which
2774 * case fall back to ->family_num. orig_family_num will be
2775 * fixed up with the first metadata update.
2776 */
2777 family_num = super->anchor->orig_family_num;
2778 if (family_num == 0)
2779 family_num = super->anchor->family_num;
51006d85 2780 sha1_init_ctx(&ctx);
92bd8f8d 2781 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2782 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2783 if (super->current_vol >= 0)
2784 dev = get_imsm_dev(super, super->current_vol);
2785 if (dev) {
2786 __u32 vol = super->current_vol;
2787 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2788 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2789 }
2790 sha1_finish_ctx(&ctx, buf);
2791 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2792}
2793
0d481d37 2794#if 0
4f5bc454
DW		 2795static void
2796get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2797{
cdddbdbc
DW		 2798 __u8 *v = get_imsm_version(mpb);
2799 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2800 char major[] = { 0, 0, 0 };
2801 char minor[] = { 0 ,0, 0 };
2802 char patch[] = { 0, 0, 0 };
2803 char *ver_parse[] = { major, minor, patch };
2804 int i, j;
2805
2806 i = j = 0;
2807 while (*v != '\0' && v < end) {
2808 if (*v != '.' && j < 2)
2809 ver_parse[i][j++] = *v;
2810 else {
2811 i++;
2812 j = 0;
2813 }
2814 v++;
2815 }
2816
4f5bc454
DW		 2817 *m = strtol(minor, NULL, 0);
2818 *p = strtol(patch, NULL, 0);
2819}
0d481d37 2820#endif
4f5bc454 2821
1e5c6983
DW		 2822static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2823{
2824 /* migr_strip_size when repairing or initializing parity */
238c0a71 2825 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2826 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2827
2828 switch (get_imsm_raid_level(map)) {
2829 case 5:
2830 case 10:
2831 return chunk;
2832 default:
2833 return 128*1024 >> 9;
2834 }
2835}
2836
2837static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2838{
2839 /* migr_strip_size when rebuilding a degraded disk, no idea why
2840 * this is different than migr_strip_size_resync(), but it's good
2841 * to be compatible
2842 */
238c0a71 2843 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2844 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2845
2846 switch (get_imsm_raid_level(map)) {
2847 case 1:
2848 case 10:
2849 if (map->num_members % map->num_domains == 0)
2850 return 128*1024 >> 9;
2851 else
2852 return chunk;
2853 case 5:
2854 return max((__u32) 64*1024 >> 9, chunk);
2855 default:
2856 return 128*1024 >> 9;
2857 }
2858}
2859
2860static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2861{
238c0a71
AK		 2862 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2863 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2864 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2865 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2866
2867 return max((__u32) 1, hi_chunk / lo_chunk);
2868}
2869
2870static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2871{
238c0a71 2872 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2873 int level = get_imsm_raid_level(lo);
2874
2875 if (level == 1 || level == 10) {
238c0a71 2876 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2877
2878 return hi->num_domains;
2879 } else
2880 return num_stripes_per_unit_resync(dev);
2881}
2882
9529d343 2883static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2884{
2885 /* named 'imsm_' because raid0, raid1 and raid10
2886 * counter-intuitively have the same number of data disks
2887 */
1e5c6983
DW		 2888 switch (get_imsm_raid_level(map)) {
2889 case 0:
36fd8ccc
AK		 2890 return map->num_members;
2891 break;
1e5c6983
DW		 2892 case 1:
2893 case 10:
36fd8ccc 2894 return map->num_members/2;
1e5c6983
DW		 2895 case 5:
2896 return map->num_members - 1;
2897 default:
1ade5cc1 2898 dprintf("unsupported raid level\n");
1e5c6983
DW		 2899 return 0;
2900 }
2901}
2902
44490938
MD		 2903static unsigned long long calc_component_size(struct imsm_map *map,
2904 struct imsm_dev *dev)
2905{
2906 unsigned long long component_size;
2907 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2908 long long calc_dev_size = 0;
44490938
MD		 2909 unsigned int member_disks = imsm_num_data_members(map);
2910
2911 if (member_disks == 0)
2912 return 0;
2913
2914 component_size = per_dev_array_size(map);
2915 calc_dev_size = component_size * member_disks;
2916
2917 /* Component size is rounded to 1MB so difference between size from
2918 * metadata and size calculated from num_data_stripes equals up to
2919 * 2048 blocks per each device. If the difference is higher it means
2920 * that array size was expanded and num_data_stripes was not updated.
2921 */
a4f7290c 2922 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2923 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2924 component_size = dev_size / member_disks;
2925 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2926 component_size / map->blocks_per_strip,
2927 num_data_stripes(map));
2928 }
2929
2930 return component_size;
2931}
2932
1e5c6983
DW		 2933static __u32 parity_segment_depth(struct imsm_dev *dev)
2934{
238c0a71 2935 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2936 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2937
2938 switch(get_imsm_raid_level(map)) {
2939 case 1:
2940 case 10:
2941 return chunk * map->num_domains;
2942 case 5:
2943 return chunk * map->num_members;
2944 default:
2945 return chunk;
2946 }
2947}
2948
2949static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2950{
238c0a71 2951 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2952 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2953 __u32 strip = block / chunk;
2954
2955 switch (get_imsm_raid_level(map)) {
2956 case 1:
2957 case 10: {
2958 __u32 vol_strip = (strip * map->num_domains) + 1;
2959 __u32 vol_stripe = vol_strip / map->num_members;
2960
2961 return vol_stripe * chunk + block % chunk;
2962 } case 5: {
2963 __u32 stripe = strip / (map->num_members - 1);
2964
2965 return stripe * chunk + block % chunk;
2966 }
2967 default:
2968 return 0;
2969 }
2970}
2971
c47b0ff6
AK		 2972static __u64 blocks_per_migr_unit(struct intel_super *super,
2973 struct imsm_dev *dev)
1e5c6983
DW		 2974{
2975 /* calculate the conversion factor between per member 'blocks'
2976 * (md/{resync,rebuild}_start) and imsm migration units, return
2977 * 0 for the 'not migrating' and 'unsupported migration' cases
2978 */
2979 if (!dev->vol.migr_state)
2980 return 0;
2981
2982 switch (migr_type(dev)) {
c47b0ff6
AK		 2983 case MIGR_GEN_MIGR: {
2984 struct migr_record *migr_rec = super->migr_rec;
2985 return __le32_to_cpu(migr_rec->blocks_per_unit);
2986 }
1e5c6983
DW		 2987 case MIGR_VERIFY:
2988 case MIGR_REPAIR:
2989 case MIGR_INIT: {
238c0a71 2990 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2991 __u32 stripes_per_unit;
2992 __u32 blocks_per_unit;
2993 __u32 parity_depth;
2994 __u32 migr_chunk;
2995 __u32 block_map;
2996 __u32 block_rel;
2997 __u32 segment;
2998 __u32 stripe;
2999 __u8 disks;
3000
3001 /* yes, this is really the translation of migr_units to
3002 * per-member blocks in the 'resync' case
3003 */
3004 stripes_per_unit = num_stripes_per_unit_resync(dev);
3005 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 3006 disks = imsm_num_data_members(map);
1e5c6983 3007 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 3008 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 3009 segment = blocks_per_unit / stripe;
3010 block_rel = blocks_per_unit - segment * stripe;
3011 parity_depth = parity_segment_depth(dev);
3012 block_map = map_migr_block(dev, block_rel);
3013 return block_map + parity_depth * segment;
3014 }
3015 case MIGR_REBUILD: {
3016 __u32 stripes_per_unit;
3017 __u32 migr_chunk;
3018
3019 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
3020 migr_chunk = migr_strip_blocks_rebuild(dev);
3021 return migr_chunk * stripes_per_unit;
3022 }
1e5c6983
DW		 3023 case MIGR_STATE_CHANGE:
3024 default:
3025 return 0;
3026 }
3027}
3028
c2c087e6
DW		 3029static int imsm_level_to_layout(int level)
3030{
3031 switch (level) {
3032 case 0:
3033 case 1:
3034 return 0;
3035 case 5:
3036 case 6:
a380c027 3037 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 3038 case 10:
c92a2527 3039 return 0x102;
c2c087e6 3040 }
a18a888e 3041 return UnSet;
c2c087e6
DW		 3042}
3043
8e59f3d8
AK		 3044/*******************************************************************************
3045 * Function: read_imsm_migr_rec
3046 * Description: Function reads imsm migration record from last sector of disk
3047 * Parameters:
3048 * fd : disk descriptor
3049 * super : metadata info
3050 * Returns:
3051 * 0 : success,
3052 * -1 : fail
3053 ******************************************************************************/
3054static int read_imsm_migr_rec(int fd, struct intel_super *super)
3055{
3056 int ret_val = -1;
de44e46f 3057 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3058 unsigned long long dsize;
3059
3060 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3061 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3062 SEEK_SET) < 0) {
e7b84f9d
N		 3063 pr_err("Cannot seek to anchor block: %s\n",
3064 strerror(errno));
8e59f3d8
AK		 3065 goto out;
3066 }
466070ad 3067 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3068 MIGR_REC_BUF_SECTORS*sector_size) !=
3069 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3070 pr_err("Cannot read migr record block: %s\n",
3071 strerror(errno));
8e59f3d8
AK		 3072 goto out;
3073 }
3074 ret_val = 0;
de44e46f
PB		 3075 if (sector_size == 4096)
3076 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3077
3078out:
3079 return ret_val;
3080}
3081
3136abe5
AK		 3082static struct imsm_dev *imsm_get_device_during_migration(
3083 struct intel_super *super)
3084{
3085
3086 struct intel_dev *dv;
3087
3088 for (dv = super->devlist; dv; dv = dv->next) {
3089 if (is_gen_migration(dv->dev))
3090 return dv->dev;
3091 }
3092 return NULL;
3093}
3094
8e59f3d8
AK		 3095/*******************************************************************************
3096 * Function: load_imsm_migr_rec
3097 * Description: Function reads imsm migration record (it is stored at the last
3098 * sector of disk)
3099 * Parameters:
3100 * super : imsm internal array info
8e59f3d8
AK		 3101 * Returns:
3102 * 0 : success
3103 * -1 : fail
4c965cc9 3104 * -2 : no migration in progress
8e59f3d8 3105 ******************************************************************************/
2f86fda3 3106static int load_imsm_migr_rec(struct intel_super *super)
8e59f3d8 3107{
594dc1b8 3108 struct dl *dl;
8e59f3d8
AK		 3109 char nm[30];
3110 int retval = -1;
3111 int fd = -1;
3136abe5 3112 struct imsm_dev *dev;
594dc1b8 3113 struct imsm_map *map;
b4ab44d8 3114 int slot = -1;
2f86fda3 3115 int keep_fd = 1;
3136abe5
AK		 3116
3117 /* find map under migration */
3118 dev = imsm_get_device_during_migration(super);
3119 /* nothing to load,no migration in progress?
3120 */
3121 if (dev == NULL)
4c965cc9 3122 return -2;
8e59f3d8 3123
2f86fda3
MT		 3124 map = get_imsm_map(dev, MAP_0);
3125 if (!map)
3126 return -1;
3136abe5 3127
2f86fda3
MT		 3128 for (dl = super->disks; dl; dl = dl->next) {
3129 /* skip spare and failed disks
3130 */
3131 if (dl->index < 0)
3132 continue;
3133 /* read only from one of the first two slots
3134 */
3135 slot = get_imsm_disk_slot(map, dl->index);
3136 if (slot > 1 || slot < 0)
3137 continue;
3138
3139 if (dl->fd < 0) {
8e59f3d8
AK		 3140 sprintf(nm, "%d:%d", dl->major, dl->minor);
3141 fd = dev_open(nm, O_RDONLY);
2f86fda3
MT		 3142 if (fd >= 0) {
3143 keep_fd = 0;
8e59f3d8 3144 break;
2f86fda3
MT		 3145 }
3146 } else {
3147 fd = dl->fd;
3148 break;
8e59f3d8
AK		 3149 }
3150 }
2f86fda3 3151
8e59f3d8 3152 if (fd < 0)
2f86fda3 3153 return retval;
8e59f3d8 3154 retval = read_imsm_migr_rec(fd, super);
2f86fda3 3155 if (!keep_fd)
8e59f3d8 3156 close(fd);
2f86fda3 3157
8e59f3d8
AK		 3158 return retval;
3159}
3160
c17608ea
AK		 3161/*******************************************************************************
3162 * function: imsm_create_metadata_checkpoint_update
3163 * Description: It creates update for checkpoint change.
3164 * Parameters:
3165 * super : imsm internal array info
3166 * u : pointer to prepared update
3167 * Returns:
3168 * Uptate length.
3169 * If length is equal to 0, input pointer u contains no update
3170 ******************************************************************************/
3171static int imsm_create_metadata_checkpoint_update(
3172 struct intel_super *super,
3173 struct imsm_update_general_migration_checkpoint **u)
3174{
3175
3176 int update_memory_size = 0;
3177
1ade5cc1 3178 dprintf("(enter)\n");
c17608ea
AK		 3179
3180 if (u == NULL)
3181 return 0;
3182 *u = NULL;
3183
3184 /* size of all update data without anchor */
3185 update_memory_size =
3186 sizeof(struct imsm_update_general_migration_checkpoint);
3187
503975b9 3188 *u = xcalloc(1, update_memory_size);
c17608ea 3189 if (*u == NULL) {
1ade5cc1 3190 dprintf("error: cannot get memory\n");
c17608ea
AK		 3191 return 0;
3192 }
3193 (*u)->type = update_general_migration_checkpoint;
9f421827 3194 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
83b3de77 3195 dprintf("prepared for %llu\n", (unsigned long long)(*u)->curr_migr_unit);
c17608ea
AK		 3196
3197 return update_memory_size;
3198}
3199
c17608ea
AK		 3200static void imsm_update_metadata_locally(struct supertype *st,
3201 void *buf, int len);
3202
687629c2
AK		 3203/*******************************************************************************
3204 * Function: write_imsm_migr_rec
3205 * Description: Function writes imsm migration record
3206 * (at the last sector of disk)
3207 * Parameters:
3208 * super : imsm internal array info
3209 * Returns:
3210 * 0 : success
3211 * -1 : if fail
3212 ******************************************************************************/
3213static int write_imsm_migr_rec(struct supertype *st)
3214{
3215 struct intel_super *super = st->sb;
de44e46f 3216 unsigned int sector_size = super->sector_size;
687629c2 3217 unsigned long long dsize;
687629c2
AK		 3218 int retval = -1;
3219 struct dl *sd;
c17608ea
AK		 3220 int len;
3221 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3222 struct imsm_dev *dev;
594dc1b8 3223 struct imsm_map *map;
3136abe5
AK		 3224
3225 /* find map under migration */
3226 dev = imsm_get_device_during_migration(super);
3227 /* if no migration, write buffer anyway to clear migr_record
3228 * on disk based on first available device
3229 */
3230 if (dev == NULL)
3231 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3232 super->current_vol);
3233
44bfe6df 3234 map = get_imsm_map(dev, MAP_0);
687629c2 3235
de44e46f
PB		 3236 if (sector_size == 4096)
3237 convert_to_4k_imsm_migr_rec(super);
687629c2 3238 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3239 int slot = -1;
3136abe5
AK		 3240
3241 /* skip failed and spare devices */
3242 if (sd->index < 0)
3243 continue;
687629c2 3244 /* write to 2 first slots only */
3136abe5
AK		 3245 if (map)
3246 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3247 if (map == NULL || slot > 1 || slot < 0)
687629c2 3248 continue;
3136abe5 3249
2f86fda3
MT		 3250 get_dev_size(sd->fd, NULL, &dsize);
3251 if (lseek64(sd->fd, dsize - (MIGR_REC_SECTOR_POSITION *
3252 sector_size),
de44e46f 3253 SEEK_SET) < 0) {
e7b84f9d
N		 3254 pr_err("Cannot seek to anchor block: %s\n",
3255 strerror(errno));
687629c2
AK		 3256 goto out;
3257 }
2f86fda3 3258 if ((unsigned int)write(sd->fd, super->migr_rec_buf,
de44e46f
PB		 3259 MIGR_REC_BUF_SECTORS*sector_size) !=
3260 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3261 pr_err("Cannot write migr record block: %s\n",
3262 strerror(errno));
687629c2
AK		 3263 goto out;
3264 }
687629c2 3265 }
de44e46f
PB		 3266 if (sector_size == 4096)
3267 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3268 /* update checkpoint information in metadata */
3269 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3270 if (len <= 0) {
3271 dprintf("imsm: Cannot prepare update\n");
3272 goto out;
3273 }
3274 /* update metadata locally */
3275 imsm_update_metadata_locally(st, u, len);
3276 /* and possibly remotely */
3277 if (st->update_tail) {
3278 append_metadata_update(st, u, len);
3279 /* during reshape we do all work inside metadata handler
3280 * manage_reshape(), so metadata update has to be triggered
3281 * insida it
3282 */
3283 flush_metadata_updates(st);
3284 st->update_tail = &st->updates;
3285 } else
3286 free(u);
687629c2
AK		 3287
3288 retval = 0;
3289 out:
687629c2
AK		 3290 return retval;
3291}
3292
e2962bfc
AK		 3293/* spare/missing disks activations are not allowe when
3294 * array/container performs reshape operation, because
3295 * all arrays in container works on the same disks set
3296 */
3297int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3298{
3299 int rv = 0;
3300 struct intel_dev *i_dev;
3301 struct imsm_dev *dev;
3302
3303 /* check whole container
3304 */
3305 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3306 dev = i_dev->dev;
3ad25638 3307 if (is_gen_migration(dev)) {
e2962bfc
AK		 3308 /* No repair during any migration in container
3309 */
3310 rv = 1;
3311 break;
3312 }
3313 }
3314 return rv;
3315}
3e684231 3316static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3317 int chunk_size,
f36a9ecd 3318 unsigned int sector_size,
c41e00b2
AK		 3319 unsigned long long component_size)
3320{
3e684231 3321 unsigned int component_size_alignment;
c41e00b2 3322
3e684231 3323 /* check component size alignment
c41e00b2 3324 */
3e684231 3325 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3326
3e684231 3327 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3328 level, chunk_size, component_size,
3e684231 3329 component_size_alignment);
c41e00b2 3330
3e684231
MZ		 3331 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3332 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3333 component_size);
3e684231 3334 component_size -= component_size_alignment;
1ade5cc1 3335 dprintf_cont("to %llu (%i).\n",
3e684231 3336 component_size, component_size_alignment);
c41e00b2
AK		 3337 }
3338
3339 return component_size;
3340}
e2962bfc 3341
fbc42556
JR		 3342/*******************************************************************************
3343 * Function: get_bitmap_header_sector
3344 * Description: Returns the sector where the bitmap header is placed.
3345 * Parameters:
3346 * st : supertype information
3347 * dev_idx : index of the device with bitmap
3348 *
3349 * Returns:
3350 * The sector where the bitmap header is placed
3351 ******************************************************************************/
3352static unsigned long long get_bitmap_header_sector(struct intel_super *super,
3353 int dev_idx)
3354{
3355 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3356 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3357
3358 if (!super->sector_size) {
3359 dprintf("sector size is not set\n");
3360 return 0;
3361 }
3362
3363 return pba_of_lba0(map) + calc_component_size(map, dev) +
3364 (IMSM_BITMAP_HEADER_OFFSET / super->sector_size);
3365}
3366
3367/*******************************************************************************
3368 * Function: get_bitmap_sector
3369 * Description: Returns the sector where the bitmap is placed.
3370 * Parameters:
3371 * st : supertype information
3372 * dev_idx : index of the device with bitmap
3373 *
3374 * Returns:
3375 * The sector where the bitmap is placed
3376 ******************************************************************************/
3377static unsigned long long get_bitmap_sector(struct intel_super *super,
3378 int dev_idx)
3379{
3380 if (!super->sector_size) {
3381 dprintf("sector size is not set\n");
3382 return 0;
3383 }
3384
3385 return get_bitmap_header_sector(super, dev_idx) +
3386 (IMSM_BITMAP_HEADER_SIZE / super->sector_size);
3387}
3388
2432ce9b
AP		 3389static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3390{
3391 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3392 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3393
3394 return pba_of_lba0(map) +
3395 (num_data_stripes(map) * map->blocks_per_strip);
3396}
3397
a5d85af7 3398static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3399{
3400 struct intel_super *super = st->sb;
c47b0ff6 3401 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3402 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3403 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3404 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3405 struct imsm_map *map_to_analyse = map;
efb30e7f 3406 struct dl *dl;
a5d85af7 3407 int map_disks = info->array.raid_disks;
bf5a934a 3408
95eeceeb 3409 memset(info, 0, sizeof(*info));
b335e593
AK		 3410 if (prev_map)
3411 map_to_analyse = prev_map;
3412
ca0748fa 3413 dl = super->current_disk;
9894ec0d 3414
bf5a934a 3415 info->container_member = super->current_vol;
cd0430a1 3416 info->array.raid_disks = map->num_members;
b335e593 3417 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3418 info->array.layout = imsm_level_to_layout(info->array.level);
3419 info->array.md_minor = -1;
3420 info->array.ctime = 0;
3421 info->array.utime = 0;
b335e593
AK		 3422 info->array.chunk_size =
3423 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3424 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3425 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3426 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3427
3f510843 3428 if (is_gen_migration(dev)) {
3f83228a 3429 info->reshape_active = 1;
b335e593
AK		 3430 info->new_level = get_imsm_raid_level(map);
3431 info->new_layout = imsm_level_to_layout(info->new_level);
3432 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3433 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3434 if (info->delta_disks) {
3435 /* this needs to be applied to every array
3436 * in the container.
3437 */
81219e70 3438 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3439 }
3f83228a
N		 3440 /* We shape information that we give to md might have to be
3441 * modify to cope with md's requirement for reshaping arrays.
3442 * For example, when reshaping a RAID0, md requires it to be
3443 * presented as a degraded RAID4.
3444 * Also if a RAID0 is migrating to a RAID5 we need to specify
3445 * the array as already being RAID5, but the 'before' layout
3446 * is a RAID4-like layout.
3447 */
3448 switch (info->array.level) {
3449 case 0:
3450 switch(info->new_level) {
3451 case 0:
3452 /* conversion is happening as RAID4 */
3453 info->array.level = 4;
3454 info->array.raid_disks += 1;
3455 break;
3456 case 5:
3457 /* conversion is happening as RAID5 */
3458 info->array.level = 5;
3459 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3460 info->delta_disks -= 1;
3461 break;
3462 default:
3463 /* FIXME error message */
3464 info->array.level = UnSet;
3465 break;
3466 }
3467 break;
3468 }
b335e593
AK		 3469 } else {
3470 info->new_level = UnSet;
3471 info->new_layout = UnSet;
3472 info->new_chunk = info->array.chunk_size;
3f83228a 3473 info->delta_disks = 0;
b335e593 3474 }
ca0748fa 3475
efb30e7f
DW		 3476 if (dl) {
3477 info->disk.major = dl->major;
3478 info->disk.minor = dl->minor;
ca0748fa 3479 info->disk.number = dl->index;
656b6b5a
N		 3480 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3481 dl->index);
efb30e7f 3482 }
bf5a934a 3483
5551b113 3484 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3485 info->component_size = calc_component_size(map, dev);
3e684231 3486 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3487 info->array.level,
3488 info->array.chunk_size,
f36a9ecd 3489 super->sector_size,
c41e00b2 3490 info->component_size);
5e46202e 3491 info->bb.supported = 1;
139dae11 3492
301406c9 3493 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3494 info->recovery_start = MaxSector;
bf5a934a 3495
c2462068
PB		 3496 if (info->array.level == 5 &&
3497 (dev->rwh_policy == RWH_DISTRIBUTED ||
3498 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3499 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3500 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3501 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3502 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3503 else
3504 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3505 >> 9;
2432ce9b
AP		 3506 } else if (info->array.level <= 0) {
3507 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3508 } else {
fbc42556
JR		 3509 if (dev->rwh_policy == RWH_BITMAP) {
3510 info->bitmap_offset = get_bitmap_sector(super, super->current_vol);
3511 info->consistency_policy = CONSISTENCY_POLICY_BITMAP;
3512 } else {
3513 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3514 }
2432ce9b
AP		 3515 }
3516
d2e6d5d6 3517 info->reshape_progress = 0;
b6796ce1 3518 info->resync_start = MaxSector;
b9172665 3519 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3520 !(info->array.state & 1)) &&
b9172665 3521 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3522 info->resync_start = 0;
b6796ce1
AK		 3523 }
3524 if (dev->vol.migr_state) {
1e5c6983
DW		 3525 switch (migr_type(dev)) {
3526 case MIGR_REPAIR:
3527 case MIGR_INIT: {
c47b0ff6
AK		 3528 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3529 dev);
4036e7ee 3530 __u64 units = vol_curr_migr_unit(dev);
1e5c6983
DW		 3531
3532 info->resync_start = blocks_per_unit * units;
3533 break;
3534 }
d2e6d5d6 3535 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3536 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3537 dev);
9f421827 3538 __u64 units = current_migr_unit(migr_rec);
04fa9523 3539 int used_disks;
d2e6d5d6 3540
befb629b
AK		 3541 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3542 (units <
9f421827 3543 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3544 (super->migr_rec->rec_status ==
3545 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3546 units++;
3547
d2e6d5d6 3548 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3549
7a862a02 3550 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3551 (unsigned long long)units,
3552 (unsigned long long)blocks_per_unit,
3553 info->reshape_progress);
75156c46 3554
9529d343 3555 used_disks = imsm_num_data_members(prev_map);
75156c46 3556 if (used_disks > 0) {
895ffd99 3557 info->custom_array_size = per_dev_array_size(map) *
75156c46 3558 used_disks;
75156c46 3559 }
d2e6d5d6 3560 }
1e5c6983
DW		 3561 case MIGR_VERIFY:
3562 /* we could emulate the checkpointing of
3563 * 'sync_action=check' migrations, but for now
3564 * we just immediately complete them
3565 */
3566 case MIGR_REBUILD:
3567 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3568 case MIGR_STATE_CHANGE:
3569 /* FIXME handle other migrations */
3570 default:
3571 /* we are not dirty, so... */
3572 info->resync_start = MaxSector;
3573 }
b6796ce1 3574 }
301406c9
DW		 3575
3576 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3577 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3578
f35f2525
N		 3579 info->array.major_version = -1;
3580 info->array.minor_version = -2;
4dd2df09 3581 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3582 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3583 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3584
3585 if (dmap) {
3586 int i, j;
3587 for (i=0; i<map_disks; i++) {
3588 dmap[i] = 0;
3589 if (i < info->array.raid_disks) {
3590 struct imsm_disk *dsk;
238c0a71 3591 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3592 dsk = get_imsm_disk(super, j);
3593 if (dsk && (dsk->status & CONFIGURED_DISK))
3594 dmap[i] = 1;
3595 }
3596 }
3597 }
81ac8b4d 3598}
bf5a934a 3599
3b451610
AK		 3600static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3601 int failed, int look_in_map);
3602
3603static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3604 int look_in_map);
3605
3606static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3607{
3608 if (is_gen_migration(dev)) {
3609 int failed;
3610 __u8 map_state;
3611 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3612
3613 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3614 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3615 if (map2->map_state != map_state) {
3616 map2->map_state = map_state;
3617 super->updates_pending++;
3618 }
3619 }
3620}
97b4d0e9
DW		 3621
3622static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3623{
3624 struct dl *d;
3625
3626 for (d = super->missing; d; d = d->next)
3627 if (d->index == index)
3628 return &d->disk;
3629 return NULL;
3630}
3631
a5d85af7 3632static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3633{
3634 struct intel_super *super = st->sb;
4f5bc454 3635 struct imsm_disk *disk;
a5d85af7 3636 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3637 int max_enough = -1;
3638 int i;
3639 struct imsm_super *mpb;
4f5bc454 3640
bf5a934a 3641 if (super->current_vol >= 0) {
a5d85af7 3642 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3643 return;
3644 }
95eeceeb 3645 memset(info, 0, sizeof(*info));
d23fe947
DW		 3646
3647 /* Set raid_disks to zero so that Assemble will always pull in valid
3648 * spares
3649 */
3650 info->array.raid_disks = 0;
cdddbdbc
DW		 3651 info->array.level = LEVEL_CONTAINER;
3652 info->array.layout = 0;
3653 info->array.md_minor = -1;
1011e834 3654 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3655 info->array.utime = 0;
3656 info->array.chunk_size = 0;
3657
3658 info->disk.major = 0;
3659 info->disk.minor = 0;
cdddbdbc 3660 info->disk.raid_disk = -1;
c2c087e6 3661 info->reshape_active = 0;
f35f2525
N		 3662 info->array.major_version = -1;
3663 info->array.minor_version = -2;
c2c087e6 3664 strcpy(info->text_version, "imsm");
a67dd8cc 3665 info->safe_mode_delay = 0;
c2c087e6
DW		 3666 info->disk.number = -1;
3667 info->disk.state = 0;
c5afc314 3668 info->name[0] = 0;
921d9e16 3669 info->recovery_start = MaxSector;
3ad25638 3670 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3671 info->bb.supported = 1;
c2c087e6 3672
97b4d0e9 3673 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3674 mpb = super->anchor;
b7d81a38 3675 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3676
ab3cb6b3
N		 3677 for (i = 0; i < mpb->num_raid_devs; i++) {
3678 struct imsm_dev *dev = get_imsm_dev(super, i);
3679 int failed, enough, j, missing = 0;
3680 struct imsm_map *map;
3681 __u8 state;
97b4d0e9 3682
3b451610
AK		 3683 failed = imsm_count_failed(super, dev, MAP_0);
3684 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3685 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3686
3687 /* any newly missing disks?
3688 * (catches single-degraded vs double-degraded)
3689 */
3690 for (j = 0; j < map->num_members; j++) {
238c0a71 3691 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3692 __u32 idx = ord_to_idx(ord);
3693
20dc76d1
MT		 3694 if (super->disks && super->disks->index == (int)idx)
3695 info->disk.raid_disk = j;
3696
ab3cb6b3
N		 3697 if (!(ord & IMSM_ORD_REBUILD) &&
3698 get_imsm_missing(super, idx)) {
3699 missing = 1;
3700 break;
3701 }
97b4d0e9 3702 }
ab3cb6b3
N		 3703
3704 if (state == IMSM_T_STATE_FAILED)
3705 enough = -1;
3706 else if (state == IMSM_T_STATE_DEGRADED &&
3707 (state != map->map_state || missing))
3708 enough = 0;
3709 else /* we're normal, or already degraded */
3710 enough = 1;
d2bde6d3
AK		 3711 if (is_gen_migration(dev) && missing) {
3712 /* during general migration we need all disks
3713 * that process is running on.
3714 * No new missing disk is allowed.
3715 */
3716 max_enough = -1;
3717 enough = -1;
3718 /* no more checks necessary
3719 */
3720 break;
3721 }
ab3cb6b3
N		 3722 /* in the missing/failed disk case check to see
3723 * if at least one array is runnable
3724 */
3725 max_enough = max(max_enough, enough);
3726 }
1ade5cc1 3727 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3728 info->container_enough = max_enough;
97b4d0e9 3729
4a04ec6c 3730 if (super->disks) {
14e8215b
DW		 3731 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3732
b9f594fe 3733 disk = &super->disks->disk;
5551b113 3734 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3735 info->component_size = reserved;
25ed7e59 3736 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3737 /* we don't change info->disk.raid_disk here because
3738 * this state will be finalized in mdmon after we have
3739 * found the 'most fresh' version of the metadata
3740 */
25ed7e59 3741 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3742 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3743 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3744 }
a575e2a7
DW		 3745
3746 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3747 * ->compare_super may have updated the 'num_raid_devs' field for spares
3748 */
3749 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3750 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3751 else
3752 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3753
3754 /* I don't know how to compute 'map' on imsm, so use safe default */
3755 if (map) {
3756 int i;
3757 for (i = 0; i < map_disks; i++)
3758 map[i] = 1;
3759 }
3760
cdddbdbc
DW		 3761}
3762
5c4cd5da
AC		 3763/* allocates memory and fills disk in mdinfo structure
3764 * for each disk in array */
3765struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3766{
594dc1b8 3767 struct mdinfo *mddev;
5c4cd5da
AC		 3768 struct intel_super *super = st->sb;
3769 struct imsm_disk *disk;
3770 int count = 0;
3771 struct dl *dl;
3772 if (!super || !super->disks)
3773 return NULL;
3774 dl = super->disks;
503975b9 3775 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3776 while (dl) {
3777 struct mdinfo *tmp;
3778 disk = &dl->disk;
503975b9 3779 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3780 if (mddev->devs)
3781 tmp->next = mddev->devs;
3782 mddev->devs = tmp;
3783 tmp->disk.number = count++;
3784 tmp->disk.major = dl->major;
3785 tmp->disk.minor = dl->minor;
3786 tmp->disk.state = is_configured(disk) ?
3787 (1 << MD_DISK_ACTIVE) : 0;
3788 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3789 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3790 tmp->disk.raid_disk = -1;
3791 dl = dl->next;
3792 }
3793 return mddev;
3794}
3795
cdddbdbc
DW		 3796static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3797 char *update, char *devname, int verbose,
3798 int uuid_set, char *homehost)
3799{
f352c545
DW		 3800 /* For 'assemble' and 'force' we need to return non-zero if any
3801 * change was made. For others, the return value is ignored.
3802 * Update options are:
3803 * force-one : This device looks a bit old but needs to be included,
3804 * update age info appropriately.
3805 * assemble: clear any 'faulty' flag to allow this device to
3806 * be assembled.
3807 * force-array: Array is degraded but being forced, mark it clean
3808 * if that will be needed to assemble it.
3809 *
3810 * newdev: not used ????
3811 * grow: Array has gained a new device - this is currently for
3812 * linear only
3813 * resync: mark as dirty so a resync will happen.
3814 * name: update the name - preserving the homehost
6e46bf34 3815 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3816 *
3817 * Following are not relevant for this imsm:
3818 * sparc2.2 : update from old dodgey metadata
3819 * super-minor: change the preferred_minor number
3820 * summaries: update redundant counters.
f352c545
DW		 3821 * homehost: update the recorded homehost
3822 * _reshape_progress: record new reshape_progress position.
3823 */
6e46bf34
DW		 3824 int rv = 1;
3825 struct intel_super *super = st->sb;
3826 struct imsm_super *mpb;
f352c545 3827
6e46bf34
DW		 3828 /* we can only update container info */
3829 if (!super || super->current_vol >= 0 || !super->anchor)
3830 return 1;
3831
3832 mpb = super->anchor;
3833
81a5b4f5
N		 3834 if (strcmp(update, "uuid") == 0) {
3835 /* We take this to mean that the family_num should be updated.
3836 * However that is much smaller than the uuid so we cannot really
3837 * allow an explicit uuid to be given. And it is hard to reliably
3838 * know if one was.
3839 * So if !uuid_set we know the current uuid is random and just used
3840 * the first 'int' and copy it to the other 3 positions.
3841 * Otherwise we require the 4 'int's to be the same as would be the
3842 * case if we are using a random uuid. So an explicit uuid will be
3843 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3844 */
81a5b4f5
N		 3845 if (!uuid_set) {
3846 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3847 rv = 0;
81a5b4f5
N		 3848 } else {
3849 if (info->uuid[0] != info->uuid[1] ||
3850 info->uuid[1] != info->uuid[2] ||
3851 info->uuid[2] != info->uuid[3])
3852 rv = -1;
3853 else
3854 rv = 0;
6e46bf34 3855 }
81a5b4f5
N		 3856 if (rv == 0)
3857 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3858 } else if (strcmp(update, "assemble") == 0)
3859 rv = 0;
3860 else
1e2b2765 3861 rv = -1;
f352c545 3862
6e46bf34
DW		 3863 /* successful update? recompute checksum */
3864 if (rv == 0)
3865 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3866
3867 return rv;
cdddbdbc
DW		 3868}
3869
c2c087e6 3870static size_t disks_to_mpb_size(int disks)
cdddbdbc 3871{
c2c087e6 3872 size_t size;
cdddbdbc 3873
c2c087e6
DW		 3874 size = sizeof(struct imsm_super);
3875 size += (disks - 1) * sizeof(struct imsm_disk);
3876 size += 2 * sizeof(struct imsm_dev);
3877 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3878 size += (4 - 2) * sizeof(struct imsm_map);
3879 /* 4 possible disk_ord_tbl's */
3880 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3881 /* maximum bbm log */
3882 size += sizeof(struct bbm_log);
c2c087e6
DW		 3883
3884 return size;
3885}
3886
387fcd59
N		 3887static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3888 unsigned long long data_offset)
c2c087e6
DW		 3889{
3890 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3891 return 0;
3892
3893 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3894}
3895
ba2de7ba
DW		 3896static void free_devlist(struct intel_super *super)
3897{
3898 struct intel_dev *dv;
3899
3900 while (super->devlist) {
3901 dv = super->devlist->next;
3902 free(super->devlist->dev);
3903 free(super->devlist);
3904 super->devlist = dv;
3905 }
3906}
3907
3908static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3909{
3910 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3911}
3912
c7b8547c
MT		 3913static int compare_super_imsm(struct supertype *st, struct supertype *tst,
3914 int verbose)
cdddbdbc
DW		 3915{
3916 /*
3917 * return:
3918 * 0 same, or first was empty, and second was copied
3919 * 1 second had wrong number
3920 * 2 wrong uuid
3921 * 3 wrong other info
3922 */
3923 struct intel_super *first = st->sb;
3924 struct intel_super *sec = tst->sb;
3925
5d500228
N		 3926 if (!first) {
3927 st->sb = tst->sb;
3928 tst->sb = NULL;
3929 return 0;
3930 }
8603ea6f
LM		 3931 /* in platform dependent environment test if the disks
3932 * use the same Intel hba
cb8f6859 3933 * If not on Intel hba at all, allow anything.
8603ea6f 3934 */
6b781d33
AP		 3935 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3936 if (first->hba->type != sec->hba->type) {
c7b8547c
MT		 3937 if (verbose)
3938 pr_err("HBAs of devices do not match %s != %s\n",
3939 get_sys_dev_type(first->hba->type),
3940 get_sys_dev_type(sec->hba->type));
6b781d33
AP		 3941 return 3;
3942 }
c7b8547c 3943
6b781d33 3944 if (first->orom != sec->orom) {
c7b8547c
MT		 3945 if (verbose)
3946 pr_err("HBAs of devices do not match %s != %s\n",
3947 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3948 return 3;
3949 }
c7b8547c 3950
8603ea6f 3951 }
cdddbdbc 3952
d23fe947
DW		 3953 /* if an anchor does not have num_raid_devs set then it is a free
3954 * floating spare
3955 */
3956 if (first->anchor->num_raid_devs > 0 &&
3957 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3958 /* Determine if these disks might ever have been
3959 * related. Further disambiguation can only take place
3960 * in load_super_imsm_all
3961 */
3962 __u32 first_family = first->anchor->orig_family_num;
3963 __u32 sec_family = sec->anchor->orig_family_num;
3964
f796af5d
DW		 3965 if (memcmp(first->anchor->sig, sec->anchor->sig,
3966 MAX_SIGNATURE_LENGTH) != 0)
3967 return 3;
3968
a2b97981
DW		 3969 if (first_family == 0)
3970 first_family = first->anchor->family_num;
3971 if (sec_family == 0)
3972 sec_family = sec->anchor->family_num;
3973
3974 if (first_family != sec_family)
d23fe947 3975 return 3;
f796af5d 3976
d23fe947 3977 }
cdddbdbc 3978
3e372e5a
DW		 3979 /* if 'first' is a spare promote it to a populated mpb with sec's
3980 * family number
3981 */
3982 if (first->anchor->num_raid_devs == 0 &&
3983 sec->anchor->num_raid_devs > 0) {
78d30f94 3984 int i;
ba2de7ba
DW		 3985 struct intel_dev *dv;
3986 struct imsm_dev *dev;
78d30f94
DW		 3987
3988 /* we need to copy raid device info from sec if an allocation
3989 * fails here we don't associate the spare
3990 */
3991 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3992 dv = xmalloc(sizeof(*dv));
3993 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3994 dv->dev = dev;
3995 dv->index = i;
3996 dv->next = first->devlist;
3997 first->devlist = dv;
78d30f94 3998 }
709743c5 3999 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 4000 /* allocation failure */
4001 free_devlist(first);
e12b3daa 4002 pr_err("imsm: failed to associate spare\n");
ba2de7ba 4003 return 3;
78d30f94 4004 }
3e372e5a 4005 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 4006 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 4007 first->anchor->family_num = sec->anchor->family_num;
ac6449be 4008 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 4009 for (i = 0; i < sec->anchor->num_raid_devs; i++)
4010 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 4011 }
4012
cdddbdbc
DW		 4013 return 0;
4014}
4015
0030e8d6
DW		 4016static void fd2devname(int fd, char *name)
4017{
0030e8d6 4018 char *nm;
0030e8d6 4019
7c798f87
MT		 4020 nm = fd2kname(fd);
4021 if (!nm)
0030e8d6 4022 return;
9587c373 4023
7c798f87 4024 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
0030e8d6
DW		 4025}
4026
21e9380b
AP		 4027static int nvme_get_serial(int fd, void *buf, size_t buf_len)
4028{
4029 char path[60];
4030 char *name = fd2kname(fd);
4031
4032 if (!name)
4033 return 1;
4034
4035 if (strncmp(name, "nvme", 4) != 0)
4036 return 1;
4037
4038 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
4039
4040 return load_sys(path, buf, buf_len);
4041}
4042
cdddbdbc
DW		 4043extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
4044
4045static int imsm_read_serial(int fd, char *devname,
6da53c0e 4046 __u8 *serial, size_t serial_buf_len)
cdddbdbc 4047{
21e9380b 4048 char buf[50];
cdddbdbc 4049 int rv;
6da53c0e 4050 size_t len;
316e2bf4
DW		 4051 char *dest;
4052 char *src;
21e9380b
AP		 4053 unsigned int i;
4054
4055 memset(buf, 0, sizeof(buf));
cdddbdbc 4056
21e9380b 4057 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 4058
21e9380b
AP		 4059 if (rv)
4060 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 4061
40ebbb9c 4062 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 4063 memset(serial, 0, MAX_RAID_SERIAL_LEN);
4064 fd2devname(fd, (char *) serial);
0030e8d6
DW		 4065 return 0;
4066 }
4067
cdddbdbc
DW		 4068 if (rv != 0) {
4069 if (devname)
e7b84f9d
N		 4070 pr_err("Failed to retrieve serial for %s\n",
4071 devname);
cdddbdbc
DW		 4072 return rv;
4073 }
4074
316e2bf4
DW		 4075 /* trim all whitespace and non-printable characters and convert
4076 * ':' to ';'
4077 */
21e9380b
AP		 4078 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4079 src = &buf[i];
316e2bf4
DW		 4080 if (*src > 0x20) {
4081 /* ':' is reserved for use in placeholder serial
4082 * numbers for missing disks
4083 */
4084 if (*src == ':')
4085 *dest++ = ';';
4086 else
4087 *dest++ = *src;
4088 }
4089 }
21e9380b
AP		 4090 len = dest - buf;
4091 dest = buf;
316e2bf4 4092
6da53c0e
BK		 4093 if (len > serial_buf_len) {
4094 /* truncate leading characters */
4095 dest += len - serial_buf_len;
4096 len = serial_buf_len;
316e2bf4 4097 }
5c3db629 4098
6da53c0e 4099 memset(serial, 0, serial_buf_len);
316e2bf4 4100 memcpy(serial, dest, len);
cdddbdbc
DW		 4101
4102 return 0;
4103}
4104
1f24f035
DW		 4105static int serialcmp(__u8 *s1, __u8 *s2)
4106{
4107 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4108}
4109
4110static void serialcpy(__u8 *dest, __u8 *src)
4111{
4112 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4113}
4114
54c2c1ea
DW		 4115static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4116{
4117 struct dl *dl;
4118
4119 for (dl = super->disks; dl; dl = dl->next)
4120 if (serialcmp(dl->serial, serial) == 0)
4121 break;
4122
4123 return dl;
4124}
4125
a2b97981
DW		 4126static struct imsm_disk *
4127__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4128{
4129 int i;
4130
4131 for (i = 0; i < mpb->num_disks; i++) {
4132 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4133
4134 if (serialcmp(disk->serial, serial) == 0) {
4135 if (idx)
4136 *idx = i;
4137 return disk;
4138 }
4139 }
4140
4141 return NULL;
4142}
4143
cdddbdbc
DW		 4144static int
4145load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4146{
a2b97981 4147 struct imsm_disk *disk;
cdddbdbc
DW		 4148 struct dl *dl;
4149 struct stat stb;
cdddbdbc 4150 int rv;
a2b97981 4151 char name[40];
d23fe947
DW		 4152 __u8 serial[MAX_RAID_SERIAL_LEN];
4153
6da53c0e 4154 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4155
4156 if (rv != 0)
4157 return 2;
4158
503975b9 4159 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4160
a2b97981
DW		 4161 fstat(fd, &stb);
4162 dl->major = major(stb.st_rdev);
4163 dl->minor = minor(stb.st_rdev);
4164 dl->next = super->disks;
4165 dl->fd = keep_fd ? fd : -1;
4166 assert(super->disks == NULL);
4167 super->disks = dl;
4168 serialcpy(dl->serial, serial);
4169 dl->index = -2;
4170 dl->e = NULL;
4171 fd2devname(fd, name);
4172 if (devname)
503975b9 4173 dl->devname = xstrdup(devname);
a2b97981 4174 else
503975b9 4175 dl->devname = xstrdup(name);
cdddbdbc 4176
d23fe947 4177 /* look up this disk's index in the current anchor */
a2b97981
DW		 4178 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4179 if (disk) {
4180 dl->disk = *disk;
4181 /* only set index on disks that are a member of a
4182 * populated contianer, i.e. one with raid_devs
4183 */
4184 if (is_failed(&dl->disk))
3f6efecc 4185 dl->index = -2;
2432ce9b 4186 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4187 dl->index = -1;
3f6efecc
DW		 4188 }
4189
949c47a0
DW		 4190 return 0;
4191}
4192
0c046afd
DW		 4193/* When migrating map0 contains the 'destination' state while map1
4194 * contains the current state. When not migrating map0 contains the
4195 * current state. This routine assumes that map[0].map_state is set to
4196 * the current array state before being called.
4197 *
4198 * Migration is indicated by one of the following states
4199 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4200 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4201 * map1state=unitialized)
1484e727 4202 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4203 * map1state=normal)
e3bba0e0 4204 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4205 * map1state=degraded)
8e59f3d8
AK		 4206 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4207 * map1state=normal)
0c046afd 4208 */
8e59f3d8
AK		 4209static void migrate(struct imsm_dev *dev, struct intel_super *super,
4210 __u8 to_state, int migr_type)
3393c6af 4211{
0c046afd 4212 struct imsm_map *dest;
238c0a71 4213 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4214
0c046afd 4215 dev->vol.migr_state = 1;
1484e727 4216 set_migr_type(dev, migr_type);
4036e7ee 4217 set_vol_curr_migr_unit(dev, 0);
238c0a71 4218 dest = get_imsm_map(dev, MAP_1);
0c046afd 4219
0556e1a2 4220 /* duplicate and then set the target end state in map[0] */
3393c6af 4221 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4222 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4223 __u32 ord;
4224 int i;
4225
4226 for (i = 0; i < src->num_members; i++) {
4227 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4228 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4229 }
4230 }
4231
8e59f3d8
AK		 4232 if (migr_type == MIGR_GEN_MIGR)
4233 /* Clear migration record */
4234 memset(super->migr_rec, 0, sizeof(struct migr_record));
4235
0c046afd 4236 src->map_state = to_state;
949c47a0 4237}
f8f603f1 4238
809da78e
AK		 4239static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4240 __u8 map_state)
f8f603f1 4241{
238c0a71
AK		 4242 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4243 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4244 MAP_0 : MAP_1);
28bce06f 4245 int i, j;
0556e1a2
DW		 4246
4247 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4248 * completed in the last migration.
4249 *
28bce06f 4250 * FIXME add support for raid-level-migration
0556e1a2 4251 */
089f9d79
JS		 4252 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4253 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4254 /* when final map state is other than expected
4255 * merge maps (not for migration)
4256 */
4257 int failed;
4258
4259 for (i = 0; i < prev->num_members; i++)
4260 for (j = 0; j < map->num_members; j++)
4261 /* during online capacity expansion
4262 * disks position can be changed
4263 * if takeover is used
4264 */
4265 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4266 ord_to_idx(prev->disk_ord_tbl[i])) {
4267 map->disk_ord_tbl[j] |=
4268 prev->disk_ord_tbl[i];
4269 break;
4270 }
4271 failed = imsm_count_failed(super, dev, MAP_0);
4272 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4273 }
f8f603f1
DW		 4274
4275 dev->vol.migr_state = 0;
ea672ee1 4276 set_migr_type(dev, 0);
4036e7ee 4277 set_vol_curr_migr_unit(dev, 0);
f8f603f1
DW		 4278 map->map_state = map_state;
4279}
949c47a0
DW		 4280
4281static int parse_raid_devices(struct intel_super *super)
4282{
4283 int i;
4284 struct imsm_dev *dev_new;
4d7b1503 4285 size_t len, len_migr;
401d313b 4286 size_t max_len = 0;
4d7b1503
DW		 4287 size_t space_needed = 0;
4288 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4289
4290 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4291 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4292 struct intel_dev *dv;
949c47a0 4293
4d7b1503
DW		 4294 len = sizeof_imsm_dev(dev_iter, 0);
4295 len_migr = sizeof_imsm_dev(dev_iter, 1);
4296 if (len_migr > len)
4297 space_needed += len_migr - len;
ca9de185 4298
503975b9 4299 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4300 if (max_len < len_migr)
4301 max_len = len_migr;
4302 if (max_len > len_migr)
4303 space_needed += max_len - len_migr;
503975b9 4304 dev_new = xmalloc(max_len);
949c47a0 4305 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4306 dv->dev = dev_new;
4307 dv->index = i;
4308 dv->next = super->devlist;
4309 super->devlist = dv;
949c47a0 4310 }
cdddbdbc 4311
4d7b1503
DW		 4312 /* ensure that super->buf is large enough when all raid devices
4313 * are migrating
4314 */
4315 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4316 void *buf;
4317
f36a9ecd
PB		 4318 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4319 super->sector_size);
4320 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4321 return 1;
4322
1f45a8ad
DW		 4323 memcpy(buf, super->buf, super->len);
4324 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4325 free(super->buf);
4326 super->buf = buf;
4327 super->len = len;
4328 }
ca9de185 4329
bbab0940
TM		 4330 super->extra_space += space_needed;
4331
cdddbdbc
DW		 4332 return 0;
4333}
4334
e2f41b2c
AK		 4335/*******************************************************************************
4336 * Function: check_mpb_migr_compatibility
4337 * Description: Function checks for unsupported migration features:
4338 * - migration optimization area (pba_of_lba0)
4339 * - descending reshape (ascending_migr)
4340 * Parameters:
4341 * super : imsm metadata information
4342 * Returns:
4343 * 0 : migration is compatible
4344 * -1 : migration is not compatible
4345 ******************************************************************************/
4346int check_mpb_migr_compatibility(struct intel_super *super)
4347{
4348 struct imsm_map *map0, *map1;
4349 struct migr_record *migr_rec = super->migr_rec;
4350 int i;
4351
4352 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4353 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4354
4355 if (dev_iter &&
4356 dev_iter->vol.migr_state == 1 &&
4357 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4358 /* This device is migrating */
238c0a71
AK		 4359 map0 = get_imsm_map(dev_iter, MAP_0);
4360 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4361 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4362 /* migration optimization area was used */
4363 return -1;
fc54fe7a
JS		 4364 if (migr_rec->ascending_migr == 0 &&
4365 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4366 /* descending reshape not supported yet */
4367 return -1;
4368 }
4369 }
4370 return 0;
4371}
4372
d23fe947 4373static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4374
cdddbdbc 4375/* load_imsm_mpb - read matrix metadata
f2f5c343 4376 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4377 */
4378static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4379{
4380 unsigned long long dsize;
cdddbdbc 4381 unsigned long long sectors;
f36a9ecd 4382 unsigned int sector_size = super->sector_size;
cdddbdbc 4383 struct stat;
6416d527 4384 struct imsm_super *anchor;
cdddbdbc
DW		 4385 __u32 check_sum;
4386
cdddbdbc 4387 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4388 if (dsize < 2*sector_size) {
64436f06 4389 if (devname)
e7b84f9d
N		 4390 pr_err("%s: device to small for imsm\n",
4391 devname);
64436f06
N		 4392 return 1;
4393 }
cdddbdbc 4394
f36a9ecd 4395 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4396 if (devname)
e7b84f9d
N		 4397 pr_err("Cannot seek to anchor block on %s: %s\n",
4398 devname, strerror(errno));
cdddbdbc
DW		 4399 return 1;
4400 }
4401
f36a9ecd 4402 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4403 if (devname)
7a862a02 4404 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4405 return 1;
4406 }
466070ad 4407 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4408 if (devname)
e7b84f9d
N		 4409 pr_err("Cannot read anchor block on %s: %s\n",
4410 devname, strerror(errno));
6416d527 4411 free(anchor);
cdddbdbc
DW		 4412 return 1;
4413 }
4414
6416d527 4415 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4416 if (devname)
e7b84f9d 4417 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4418 free(anchor);
cdddbdbc
DW		 4419 return 2;
4420 }
4421
d23fe947 4422 __free_imsm(super, 0);
f2f5c343
LM		 4423 /* reload capability and hba */
4424
4425 /* capability and hba must be updated with new super allocation */
d424212e 4426 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4427 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4428 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4429 if (devname)
e7b84f9d
N		 4430 pr_err("unable to allocate %zu byte mpb buffer\n",
4431 super->len);
6416d527 4432 free(anchor);
cdddbdbc
DW		 4433 return 2;
4434 }
f36a9ecd 4435 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4436
f36a9ecd 4437 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4438 free(anchor);
8e59f3d8 4439
85337573
AO		 4440 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4441 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4442 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4443 free(super->buf);
4444 return 2;
4445 }
51d83f5d 4446 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4447
949c47a0 4448 if (!sectors) {
ecf45690
DW		 4449 check_sum = __gen_imsm_checksum(super->anchor);
4450 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4451 if (devname)
e7b84f9d
N		 4452 pr_err("IMSM checksum %x != %x on %s\n",
4453 check_sum,
4454 __le32_to_cpu(super->anchor->check_sum),
4455 devname);
ecf45690
DW		 4456 return 2;
4457 }
4458
a2b97981 4459 return 0;
949c47a0 4460 }
cdddbdbc
DW		 4461
4462 /* read the extended mpb */
f36a9ecd 4463 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4464 if (devname)
e7b84f9d
N		 4465 pr_err("Cannot seek to extended mpb on %s: %s\n",
4466 devname, strerror(errno));
cdddbdbc
DW		 4467 return 1;
4468 }
4469
f36a9ecd
PB		 4470 if ((unsigned int)read(fd, super->buf + sector_size,
4471 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4472 if (devname)
e7b84f9d
N		 4473 pr_err("Cannot read extended mpb on %s: %s\n",
4474 devname, strerror(errno));
cdddbdbc
DW		 4475 return 2;
4476 }
4477
949c47a0
DW		 4478 check_sum = __gen_imsm_checksum(super->anchor);
4479 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4480 if (devname)
e7b84f9d
N		 4481 pr_err("IMSM checksum %x != %x on %s\n",
4482 check_sum, __le32_to_cpu(super->anchor->check_sum),
4483 devname);
db575f3b 4484 return 3;
cdddbdbc
DW		 4485 }
4486
a2b97981
DW		 4487 return 0;
4488}
4489
8e59f3d8
AK		 4490static int read_imsm_migr_rec(int fd, struct intel_super *super);
4491
97f81ee2
CA		 4492/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4493static void clear_hi(struct intel_super *super)
4494{
4495 struct imsm_super *mpb = super->anchor;
4496 int i, n;
4497 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4498 return;
4499 for (i = 0; i < mpb->num_disks; ++i) {
4500 struct imsm_disk *disk = &mpb->disk[i];
4501 disk->total_blocks_hi = 0;
4502 }
4503 for (i = 0; i < mpb->num_raid_devs; ++i) {
4504 struct imsm_dev *dev = get_imsm_dev(super, i);
4505 if (!dev)
4506 return;
4507 for (n = 0; n < 2; ++n) {
4508 struct imsm_map *map = get_imsm_map(dev, n);
4509 if (!map)
4510 continue;
4511 map->pba_of_lba0_hi = 0;
4512 map->blocks_per_member_hi = 0;
4513 map->num_data_stripes_hi = 0;
4514 }
4515 }
4516}
4517
a2b97981
DW		 4518static int
4519load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4520{
4521 int err;
4522
4523 err = load_imsm_mpb(fd, super, devname);
4524 if (err)
4525 return err;
f36a9ecd
PB		 4526 if (super->sector_size == 4096)
4527 convert_from_4k(super);
a2b97981
DW		 4528 err = load_imsm_disk(fd, super, devname, keep_fd);
4529 if (err)
4530 return err;
4531 err = parse_raid_devices(super);
8d67477f
TM		 4532 if (err)
4533 return err;
4534 err = load_bbm_log(super);
97f81ee2 4535 clear_hi(super);
a2b97981 4536 return err;
cdddbdbc
DW		 4537}
4538
ae6aad82
DW		 4539static void __free_imsm_disk(struct dl *d)
4540{
4541 if (d->fd >= 0)
4542 close(d->fd);
4543 if (d->devname)
4544 free(d->devname);
0dcecb2e
DW		 4545 if (d->e)
4546 free(d->e);
ae6aad82
DW		 4547 free(d);
4548
4549}
1a64be56 4550
cdddbdbc
DW		 4551static void free_imsm_disks(struct intel_super *super)
4552{
47ee5a45 4553 struct dl *d;
cdddbdbc 4554
47ee5a45
DW		 4555 while (super->disks) {
4556 d = super->disks;
cdddbdbc 4557 super->disks = d->next;
ae6aad82 4558 __free_imsm_disk(d);
cdddbdbc 4559 }
cb82edca
AK		 4560 while (super->disk_mgmt_list) {
4561 d = super->disk_mgmt_list;
4562 super->disk_mgmt_list = d->next;
4563 __free_imsm_disk(d);
4564 }
47ee5a45
DW		 4565 while (super->missing) {
4566 d = super->missing;
4567 super->missing = d->next;
4568 __free_imsm_disk(d);
4569 }
4570
cdddbdbc
DW		 4571}
4572
9ca2c81c 4573/* free all the pieces hanging off of a super pointer */
d23fe947 4574static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4575{
88654014
LM		 4576 struct intel_hba *elem, *next;
4577
9ca2c81c 4578 if (super->buf) {
949c47a0 4579 free(super->buf);
9ca2c81c
DW		 4580 super->buf = NULL;
4581 }
f2f5c343
LM		 4582 /* unlink capability description */
4583 super->orom = NULL;
8e59f3d8
AK		 4584 if (super->migr_rec_buf) {
4585 free(super->migr_rec_buf);
4586 super->migr_rec_buf = NULL;
4587 }
d23fe947
DW		 4588 if (free_disks)
4589 free_imsm_disks(super);
ba2de7ba 4590 free_devlist(super);
88654014
LM		 4591 elem = super->hba;
4592 while (elem) {
4593 if (elem->path)
4594 free((void *)elem->path);
4595 next = elem->next;
4596 free(elem);
4597 elem = next;
88c32bb1 4598 }
8d67477f
TM		 4599 if (super->bbm_log)
4600 free(super->bbm_log);
88654014 4601 super->hba = NULL;
cdddbdbc
DW		 4602}
4603
9ca2c81c
DW		 4604static void free_imsm(struct intel_super *super)
4605{
d23fe947 4606 __free_imsm(super, 1);
928f1424 4607 free(super->bb.entries);
9ca2c81c
DW		 4608 free(super);
4609}
cdddbdbc
DW		 4610
4611static void free_super_imsm(struct supertype *st)
4612{
4613 struct intel_super *super = st->sb;
4614
4615 if (!super)
4616 return;
4617
4618 free_imsm(super);
4619 st->sb = NULL;
4620}
4621
49133e57 4622static struct intel_super *alloc_super(void)
c2c087e6 4623{
503975b9 4624 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4625
503975b9
N		 4626 super->current_vol = -1;
4627 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4628
4629 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4630 sizeof(struct md_bb_entry));
4631 if (!super->bb.entries) {
4632 free(super);
4633 return NULL;
4634 }
4635
c2c087e6
DW		 4636 return super;
4637}
4638
f0f5a016
LM		 4639/*
4640 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4641 */
d424212e 4642static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4643{
4644 struct sys_dev *hba_name;
4645 int rv = 0;
4646
3a30e28e
MT		 4647 if (fd >= 0 && test_partition(fd)) {
4648 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4649 devname);
4650 return 1;
4651 }
089f9d79 4652 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4653 super->orom = NULL;
f0f5a016
LM		 4654 super->hba = NULL;
4655 return 0;
4656 }
4657 hba_name = find_disk_attached_hba(fd, NULL);
4658 if (!hba_name) {
d424212e 4659 if (devname)
e7b84f9d
N		 4660 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4661 devname);
f0f5a016
LM		 4662 return 1;
4663 }
4664 rv = attach_hba_to_super(super, hba_name);
4665 if (rv == 2) {
d424212e
N		 4666 if (devname) {
4667 struct intel_hba *hba = super->hba;
f0f5a016 4668
60f0f54d
PB		 4669 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4670 " but the container is assigned to Intel(R) %s %s (",
d424212e 4671 devname,
614902f6 4672 get_sys_dev_type(hba_name->type),
60f0f54d 4673 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4674 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4675 get_sys_dev_type(super->hba->type),
4676 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4677
f0f5a016
LM		 4678 while (hba) {
4679 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4680 if (hba->next)
4681 fprintf(stderr, ", ");
4682 hba = hba->next;
4683 }
6b781d33 4684 fprintf(stderr, ").\n"
cca67208 4685 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4686 }
f0f5a016
LM		 4687 return 2;
4688 }
6b781d33 4689 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4690 if (!super->orom)
4691 return 3;
614902f6 4692
f0f5a016
LM		 4693 return 0;
4694}
4695
47ee5a45
DW		 4696/* find_missing - helper routine for load_super_imsm_all that identifies
4697 * disks that have disappeared from the system. This routine relies on
4698 * the mpb being uptodate, which it is at load time.
4699 */
4700static int find_missing(struct intel_super *super)
4701{
4702 int i;
4703 struct imsm_super *mpb = super->anchor;
4704 struct dl *dl;
4705 struct imsm_disk *disk;
47ee5a45
DW		 4706
4707 for (i = 0; i < mpb->num_disks; i++) {
4708 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4709 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4710 if (dl)
4711 continue;
47ee5a45 4712
503975b9 4713 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4714 dl->major = 0;
4715 dl->minor = 0;
4716 dl->fd = -1;
503975b9 4717 dl->devname = xstrdup("missing");
47ee5a45
DW		 4718 dl->index = i;
4719 serialcpy(dl->serial, disk->serial);
4720 dl->disk = *disk;
689c9bf3 4721 dl->e = NULL;
47ee5a45
DW		 4722 dl->next = super->missing;
4723 super->missing = dl;
4724 }
4725
4726 return 0;
4727}
4728
a2b97981
DW		 4729static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4730{
4731 struct intel_disk *idisk = disk_list;
4732
4733 while (idisk) {
4734 if (serialcmp(idisk->disk.serial, serial) == 0)
4735 break;
4736 idisk = idisk->next;
4737 }
4738
4739 return idisk;
4740}
4741
4742static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4743 struct intel_super *super,
4744 struct intel_disk **disk_list)
4745{
4746 struct imsm_disk *d = &super->disks->disk;
4747 struct imsm_super *mpb = super->anchor;
4748 int i, j;
4749
4750 for (i = 0; i < tbl_size; i++) {
4751 struct imsm_super *tbl_mpb = table[i]->anchor;
4752 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4753
4754 if (tbl_mpb->family_num == mpb->family_num) {
4755 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4756 dprintf("mpb from %d:%d matches %d:%d\n",
4757 super->disks->major,
a2b97981
DW		 4758 super->disks->minor,
4759 table[i]->disks->major,
4760 table[i]->disks->minor);
4761 break;
4762 }
4763
4764 if (((is_configured(d) && !is_configured(tbl_d)) ||
4765 is_configured(d) == is_configured(tbl_d)) &&
4766 tbl_mpb->generation_num < mpb->generation_num) {
4767 /* current version of the mpb is a
4768 * better candidate than the one in
4769 * super_table, but copy over "cross
4770 * generational" status
4771 */
4772 struct intel_disk *idisk;
4773
1ade5cc1
N		 4774 dprintf("mpb from %d:%d replaces %d:%d\n",
4775 super->disks->major,
a2b97981
DW		 4776 super->disks->minor,
4777 table[i]->disks->major,
4778 table[i]->disks->minor);
4779
4780 idisk = disk_list_get(tbl_d->serial, *disk_list);
4781 if (idisk && is_failed(&idisk->disk))
4782 tbl_d->status |= FAILED_DISK;
4783 break;
4784 } else {
4785 struct intel_disk *idisk;
4786 struct imsm_disk *disk;
4787
4788 /* tbl_mpb is more up to date, but copy
4789 * over cross generational status before
4790 * returning
4791 */
4792 disk = __serial_to_disk(d->serial, mpb, NULL);
4793 if (disk && is_failed(disk))
4794 d->status |= FAILED_DISK;
4795
4796 idisk = disk_list_get(d->serial, *disk_list);
4797 if (idisk) {
4798 idisk->owner = i;
4799 if (disk && is_configured(disk))
4800 idisk->disk.status |= CONFIGURED_DISK;
4801 }
4802
1ade5cc1
N		 4803 dprintf("mpb from %d:%d prefer %d:%d\n",
4804 super->disks->major,
a2b97981
DW		 4805 super->disks->minor,
4806 table[i]->disks->major,
4807 table[i]->disks->minor);
4808
4809 return tbl_size;
4810 }
4811 }
4812 }
4813
4814 if (i >= tbl_size)
4815 table[tbl_size++] = super;
4816 else
4817 table[i] = super;
4818
4819 /* update/extend the merged list of imsm_disk records */
4820 for (j = 0; j < mpb->num_disks; j++) {
4821 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4822 struct intel_disk *idisk;
4823
4824 idisk = disk_list_get(disk->serial, *disk_list);
4825 if (idisk) {
4826 idisk->disk.status |= disk->status;
4827 if (is_configured(&idisk->disk) ||
4828 is_failed(&idisk->disk))
4829 idisk->disk.status &= ~(SPARE_DISK);
4830 } else {
503975b9 4831 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4832 idisk->owner = IMSM_UNKNOWN_OWNER;
4833 idisk->disk = *disk;
4834 idisk->next = *disk_list;
4835 *disk_list = idisk;
4836 }
4837
4838 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4839 idisk->owner = i;
4840 }
4841
4842 return tbl_size;
4843}
4844
4845static struct intel_super *
4846validate_members(struct intel_super *super, struct intel_disk *disk_list,
4847 const int owner)
4848{
4849 struct imsm_super *mpb = super->anchor;
4850 int ok_count = 0;
4851 int i;
4852
4853 for (i = 0; i < mpb->num_disks; i++) {
4854 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4855 struct intel_disk *idisk;
4856
4857 idisk = disk_list_get(disk->serial, disk_list);
4858 if (idisk) {
4859 if (idisk->owner == owner ||
4860 idisk->owner == IMSM_UNKNOWN_OWNER)
4861 ok_count++;
4862 else
1ade5cc1
N		 4863 dprintf("'%.16s' owner %d != %d\n",
4864 disk->serial, idisk->owner,
a2b97981
DW		 4865 owner);
4866 } else {
1ade5cc1
N		 4867 dprintf("unknown disk %x [%d]: %.16s\n",
4868 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4869 disk->serial);
4870 break;
4871 }
4872 }
4873
4874 if (ok_count == mpb->num_disks)
4875 return super;
4876 return NULL;
4877}
4878
4879static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4880{
4881 struct intel_super *s;
4882
4883 for (s = super_list; s; s = s->next) {
4884 if (family_num != s->anchor->family_num)
4885 continue;
e12b3daa 4886 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4887 __le32_to_cpu(family_num), s->disks->devname);
4888 }
4889}
4890
4891static struct intel_super *
4892imsm_thunderdome(struct intel_super **super_list, int len)
4893{
4894 struct intel_super *super_table[len];
4895 struct intel_disk *disk_list = NULL;
4896 struct intel_super *champion, *spare;
4897 struct intel_super *s, **del;
4898 int tbl_size = 0;
4899 int conflict;
4900 int i;
4901
4902 memset(super_table, 0, sizeof(super_table));
4903 for (s = *super_list; s; s = s->next)
4904 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4905
4906 for (i = 0; i < tbl_size; i++) {
4907 struct imsm_disk *d;
4908 struct intel_disk *idisk;
4909 struct imsm_super *mpb = super_table[i]->anchor;
4910
4911 s = super_table[i];
4912 d = &s->disks->disk;
4913
4914 /* 'd' must appear in merged disk list for its
4915 * configuration to be valid
4916 */
4917 idisk = disk_list_get(d->serial, disk_list);
4918 if (idisk && idisk->owner == i)
4919 s = validate_members(s, disk_list, i);
4920 else
4921 s = NULL;
4922
4923 if (!s)
1ade5cc1
N		 4924 dprintf("marking family: %#x from %d:%d offline\n",
4925 mpb->family_num,
a2b97981
DW		 4926 super_table[i]->disks->major,
4927 super_table[i]->disks->minor);
4928 super_table[i] = s;
4929 }
4930
4931 /* This is where the mdadm implementation differs from the Windows
4932 * driver which has no strict concept of a container. We can only
4933 * assemble one family from a container, so when returning a prodigal
4934 * array member to this system the code will not be able to disambiguate
4935 * the container contents that should be assembled ("foreign" versus
4936 * "local"). It requires user intervention to set the orig_family_num
4937 * to a new value to establish a new container. The Windows driver in
4938 * this situation fixes up the volume name in place and manages the
4939 * foreign array as an independent entity.
4940 */
4941 s = NULL;
4942 spare = NULL;
4943 conflict = 0;
4944 for (i = 0; i < tbl_size; i++) {
4945 struct intel_super *tbl_ent = super_table[i];
4946 int is_spare = 0;
4947
4948 if (!tbl_ent)
4949 continue;
4950
4951 if (tbl_ent->anchor->num_raid_devs == 0) {
4952 spare = tbl_ent;
4953 is_spare = 1;
4954 }
4955
4956 if (s && !is_spare) {
4957 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4958 conflict++;
4959 } else if (!s && !is_spare)
4960 s = tbl_ent;
4961 }
4962
4963 if (!s)
4964 s = spare;
4965 if (!s) {
4966 champion = NULL;
4967 goto out;
4968 }
4969 champion = s;
4970
4971 if (conflict)
7a862a02 4972 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4973 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4974
4975 /* collect all dl's onto 'champion', and update them to
4976 * champion's version of the status
4977 */
4978 for (s = *super_list; s; s = s->next) {
4979 struct imsm_super *mpb = champion->anchor;
4980 struct dl *dl = s->disks;
4981
4982 if (s == champion)
4983 continue;
4984
5d7b407a
CA		 4985 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4986
a2b97981
DW		 4987 for (i = 0; i < mpb->num_disks; i++) {
4988 struct imsm_disk *disk;
4989
4990 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4991 if (disk) {
4992 dl->disk = *disk;
4993 /* only set index on disks that are a member of
4994 * a populated contianer, i.e. one with
4995 * raid_devs
4996 */
4997 if (is_failed(&dl->disk))
4998 dl->index = -2;
4999 else if (is_spare(&dl->disk))
5000 dl->index = -1;
5001 break;
5002 }
5003 }
5004
5005 if (i >= mpb->num_disks) {
5006 struct intel_disk *idisk;
5007
5008 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 5009 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 5010 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
5011 dl->index = -1;
5012 else {
5013 dl->index = -2;
5014 continue;
5015 }
5016 }
5017
5018 dl->next = champion->disks;
5019 champion->disks = dl;
5020 s->disks = NULL;
5021 }
5022
5023 /* delete 'champion' from super_list */
5024 for (del = super_list; *del; ) {
5025 if (*del == champion) {
5026 *del = (*del)->next;
5027 break;
5028 } else
5029 del = &(*del)->next;
5030 }
5031 champion->next = NULL;
5032
5033 out:
5034 while (disk_list) {
5035 struct intel_disk *idisk = disk_list;
5036
5037 disk_list = disk_list->next;
5038 free(idisk);
5039 }
5040
5041 return champion;
5042}
5043
9587c373
LM		 5044static int
5045get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 5046static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 5047 int major, int minor, int keep_fd);
ec50f7b6
LM		 5048static int
5049get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5050 int *max, int keep_fd);
5051
cdddbdbc 5052static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 5053 char *devname, struct md_list *devlist,
5054 int keep_fd)
cdddbdbc 5055{
a2b97981
DW		 5056 struct intel_super *super_list = NULL;
5057 struct intel_super *super = NULL;
a2b97981 5058 int err = 0;
9587c373 5059 int i = 0;
dab4a513 5060
9587c373
LM		 5061 if (fd >= 0)
5062 /* 'fd' is an opened container */
5063 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
5064 else
ec50f7b6
LM		 5065 /* get super block from devlist devices */
5066 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5067 if (err)
1602d52c 5068 goto error;
a2b97981
DW		 5069 /* all mpbs enter, maybe one leaves */
5070 super = imsm_thunderdome(&super_list, i);
5071 if (!super) {
5072 err = 1;
5073 goto error;
cdddbdbc
DW		 5074 }
5075
47ee5a45
DW		 5076 if (find_missing(super) != 0) {
5077 free_imsm(super);
a2b97981
DW		 5078 err = 2;
5079 goto error;
47ee5a45 5080 }
8e59f3d8
AK		 5081
5082 /* load migration record */
2f86fda3 5083 err = load_imsm_migr_rec(super);
4c965cc9
AK		 5084 if (err == -1) {
5085 /* migration is in progress,
5086 * but migr_rec cannot be loaded,
5087 */
8e59f3d8
AK		 5088 err = 4;
5089 goto error;
5090 }
e2f41b2c
AK		 5091
5092 /* Check migration compatibility */
089f9d79 5093 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5094 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5095 if (devname)
5096 fprintf(stderr, " on %s\n", devname);
5097 else
5098 fprintf(stderr, " (IMSM).\n");
5099
5100 err = 5;
5101 goto error;
5102 }
5103
a2b97981
DW		 5104 err = 0;
5105
5106 error:
5107 while (super_list) {
5108 struct intel_super *s = super_list;
5109
5110 super_list = super_list->next;
5111 free_imsm(s);
5112 }
9587c373 5113
a2b97981
DW		 5114 if (err)
5115 return err;
f7e7067b 5116
cdddbdbc 5117 *sbp = super;
9587c373 5118 if (fd >= 0)
4dd2df09 5119 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5120 else
4dd2df09 5121 st->container_devnm[0] = 0;
a2b97981 5122 if (err == 0 && st->ss == NULL) {
bf5a934a 5123 st->ss = &super_imsm;
cdddbdbc
DW		 5124 st->minor_version = 0;
5125 st->max_devs = IMSM_MAX_DEVICES;
5126 }
cdddbdbc
DW		 5127 return 0;
5128}
2b959fbf 5129
ec50f7b6
LM		 5130static int
5131get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5132 int *max, int keep_fd)
5133{
5134 struct md_list *tmpdev;
5135 int err = 0;
5136 int i = 0;
9587c373 5137
ec50f7b6
LM		 5138 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5139 if (tmpdev->used != 1)
5140 continue;
5141 if (tmpdev->container == 1) {
ca9de185 5142 int lmax = 0;
ec50f7b6
LM		 5143 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5144 if (fd < 0) {
e7b84f9d 5145 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5146 tmpdev->devname, strerror(errno));
5147 err = 8;
5148 goto error;
5149 }
5150 err = get_sra_super_block(fd, super_list,
5151 tmpdev->devname, &lmax,
5152 keep_fd);
5153 i += lmax;
5154 close(fd);
5155 if (err) {
5156 err = 7;
5157 goto error;
5158 }
5159 } else {
5160 int major = major(tmpdev->st_rdev);
5161 int minor = minor(tmpdev->st_rdev);
5162 err = get_super_block(super_list,
4dd2df09 5163 NULL,
ec50f7b6
LM		 5164 tmpdev->devname,
5165 major, minor,
5166 keep_fd);
5167 i++;
5168 if (err) {
5169 err = 6;
5170 goto error;
5171 }
5172 }
5173 }
5174 error:
5175 *max = i;
5176 return err;
5177}
9587c373 5178
4dd2df09 5179static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5180 int major, int minor, int keep_fd)
5181{
594dc1b8 5182 struct intel_super *s;
9587c373
LM		 5183 char nm[32];
5184 int dfd = -1;
9587c373
LM		 5185 int err = 0;
5186 int retry;
5187
5188 s = alloc_super();
5189 if (!s) {
5190 err = 1;
5191 goto error;
5192 }
5193
5194 sprintf(nm, "%d:%d", major, minor);
5195 dfd = dev_open(nm, O_RDWR);
5196 if (dfd < 0) {
5197 err = 2;
5198 goto error;
5199 }
5200
aec01630
JS		 5201 if (!get_dev_sector_size(dfd, NULL, &s->sector_size)) {
5202 err = 2;
5203 goto error;
5204 }
cb8f6859 5205 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5206 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5207
5208 /* retry the load if we might have raced against mdmon */
4dd2df09 5209 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5210 for (retry = 0; retry < 3; retry++) {
5211 usleep(3000);
5212 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5213 if (err != 3)
5214 break;
5215 }
5216 error:
5217 if (!err) {
5218 s->next = *super_list;
5219 *super_list = s;
5220 } else {
5221 if (s)
8d67477f 5222 free_imsm(s);
36614e95 5223 if (dfd >= 0)
9587c373
LM		 5224 close(dfd);
5225 }
089f9d79 5226 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5227 close(dfd);
5228 return err;
5229
5230}
5231
5232static int
5233get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5234{
5235 struct mdinfo *sra;
4dd2df09 5236 char *devnm;
9587c373
LM		 5237 struct mdinfo *sd;
5238 int err = 0;
5239 int i = 0;
4dd2df09 5240 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5241 if (!sra)
5242 return 1;
5243
5244 if (sra->array.major_version != -1 ||
5245 sra->array.minor_version != -2 ||
5246 strcmp(sra->text_version, "imsm") != 0) {
5247 err = 1;
5248 goto error;
5249 }
5250 /* load all mpbs */
4dd2df09 5251 devnm = fd2devnm(fd);
9587c373 5252 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5253 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5254 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5255 err = 7;
5256 goto error;
5257 }
5258 }
5259 error:
5260 sysfs_free(sra);
5261 *max = i;
5262 return err;
5263}
5264
2b959fbf
N		 5265static int load_container_imsm(struct supertype *st, int fd, char *devname)
5266{
ec50f7b6 5267 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5268}
cdddbdbc
DW		 5269
5270static int load_super_imsm(struct supertype *st, int fd, char *devname)
5271{
5272 struct intel_super *super;
5273 int rv;
8a3544f8 5274 int retry;
cdddbdbc 5275
357ac106 5276 if (test_partition(fd))
691c6ee1
N		 5277 /* IMSM not allowed on partitions */
5278 return 1;
5279
37424f13
DW		 5280 free_super_imsm(st);
5281
49133e57 5282 super = alloc_super();
aec01630
JS		 5283 if (!get_dev_sector_size(fd, NULL, &super->sector_size))
5284 return 1;
8d67477f
TM		 5285 if (!super)
5286 return 1;
ea2bc72b
LM		 5287 /* Load hba and capabilities if they exist.
5288 * But do not preclude loading metadata in case capabilities or hba are
5289 * non-compliant and ignore_hw_compat is set.
5290 */
d424212e 5291 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5292 /* no orom/efi or non-intel hba of the disk */
089f9d79 5293 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5294 if (devname)
e7b84f9d 5295 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5296 free_imsm(super);
5297 return 2;
5298 }
a2b97981 5299 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5300
8a3544f8
AP		 5301 /* retry the load if we might have raced against mdmon */
5302 if (rv == 3) {
f96b1302
AP		 5303 struct mdstat_ent *mdstat = NULL;
5304 char *name = fd2kname(fd);
5305
5306 if (name)
5307 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5308
5309 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5310 for (retry = 0; retry < 3; retry++) {
5311 usleep(3000);
5312 rv = load_and_parse_mpb(fd, super, devname, 0);
5313 if (rv != 3)
5314 break;
5315 }
5316 }
5317
5318 free_mdstat(mdstat);
5319 }
5320
cdddbdbc
DW		 5321 if (rv) {
5322 if (devname)
7a862a02 5323 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5324 free_imsm(super);
5325 return rv;
5326 }
5327
5328 st->sb = super;
5329 if (st->ss == NULL) {
5330 st->ss = &super_imsm;
5331 st->minor_version = 0;
5332 st->max_devs = IMSM_MAX_DEVICES;
5333 }
8e59f3d8
AK		 5334
5335 /* load migration record */
2f86fda3 5336 if (load_imsm_migr_rec(super) == 0) {
2e062e82
AK		 5337 /* Check for unsupported migration features */
5338 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5339 pr_err("Unsupported migration detected");
2e062e82
AK		 5340 if (devname)
5341 fprintf(stderr, " on %s\n", devname);
5342 else
5343 fprintf(stderr, " (IMSM).\n");
5344 return 3;
5345 }
e2f41b2c
AK		 5346 }
5347
cdddbdbc
DW		 5348 return 0;
5349}
5350
ef6ffade
DW		 5351static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5352{
5353 if (info->level == 1)
5354 return 128;
5355 return info->chunk_size >> 9;
5356}
5357
5551b113
CA		 5358static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5359 unsigned long long size)
fcfd9599 5360{
4025c288 5361 if (info->level == 1)
5551b113 5362 return size * 2;
4025c288 5363 else
5551b113 5364 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5365}
5366
4d1313e9
DW		 5367static void imsm_update_version_info(struct intel_super *super)
5368{
5369 /* update the version and attributes */
5370 struct imsm_super *mpb = super->anchor;
5371 char *version;
5372 struct imsm_dev *dev;
5373 struct imsm_map *map;
5374 int i;
5375
5376 for (i = 0; i < mpb->num_raid_devs; i++) {
5377 dev = get_imsm_dev(super, i);
238c0a71 5378 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5379 if (__le32_to_cpu(dev->size_high) > 0)
5380 mpb->attributes |= MPB_ATTRIB_2TB;
5381
5382 /* FIXME detect when an array spans a port multiplier */
5383 #if 0
5384 mpb->attributes |= MPB_ATTRIB_PM;
5385 #endif
5386
5387 if (mpb->num_raid_devs > 1 ||
5388 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5389 version = MPB_VERSION_ATTRIBS;
5390 switch (get_imsm_raid_level(map)) {
5391 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5392 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5393 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5394 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5395 }
5396 } else {
5397 if (map->num_members >= 5)
5398 version = MPB_VERSION_5OR6_DISK_ARRAY;
5399 else if (dev->status == DEV_CLONE_N_GO)
5400 version = MPB_VERSION_CNG;
5401 else if (get_imsm_raid_level(map) == 5)
5402 version = MPB_VERSION_RAID5;
5403 else if (map->num_members >= 3)
5404 version = MPB_VERSION_3OR4_DISK_ARRAY;
5405 else if (get_imsm_raid_level(map) == 1)
5406 version = MPB_VERSION_RAID1;
5407 else
5408 version = MPB_VERSION_RAID0;
5409 }
5410 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5411 }
5412}
5413
aa534678
DW		 5414static int check_name(struct intel_super *super, char *name, int quiet)
5415{
5416 struct imsm_super *mpb = super->anchor;
5417 char *reason = NULL;
9bd99a90
RS		 5418 char *start = name;
5419 size_t len = strlen(name);
aa534678
DW		 5420 int i;
5421
9bd99a90
RS		 5422 if (len > 0) {
5423 while (isspace(start[len - 1]))
5424 start[--len] = 0;
5425 while (*start && isspace(*start))
5426 ++start, --len;
5427 memmove(name, start, len + 1);
5428 }
5429
5430 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5431 reason = "must be 16 characters or less";
9bd99a90
RS		 5432 else if (len == 0)
5433 reason = "must be a non-empty string";
aa534678
DW		 5434
5435 for (i = 0; i < mpb->num_raid_devs; i++) {
5436 struct imsm_dev *dev = get_imsm_dev(super, i);
5437
5438 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5439 reason = "already exists";
5440 break;
5441 }
5442 }
5443
5444 if (reason && !quiet)
e7b84f9d 5445 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5446
5447 return !reason;
5448}
5449
8b353278 5450static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5451 struct shape *s, char *name,
83cd1e97
N		 5452 char *homehost, int *uuid,
5453 long long data_offset)
cdddbdbc 5454{
c2c087e6
DW		 5455 /* We are creating a volume inside a pre-existing container.
5456 * so st->sb is already set.
5457 */
5458 struct intel_super *super = st->sb;
f36a9ecd 5459 unsigned int sector_size = super->sector_size;
949c47a0 5460 struct imsm_super *mpb = super->anchor;
ba2de7ba 5461 struct intel_dev *dv;
c2c087e6
DW		 5462 struct imsm_dev *dev;
5463 struct imsm_vol *vol;
5464 struct imsm_map *map;
5465 int idx = mpb->num_raid_devs;
5466 int i;
760365f9 5467 int namelen;
c2c087e6 5468 unsigned long long array_blocks;
2c092cad 5469 size_t size_old, size_new;
5551b113 5470 unsigned long long num_data_stripes;
b53bfba6
TM		 5471 unsigned int data_disks;
5472 unsigned long long size_per_member;
cdddbdbc 5473
88c32bb1 5474 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5475 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5476 return 0;
5477 }
5478
2c092cad
DW		 5479 /* ensure the mpb is large enough for the new data */
5480 size_old = __le32_to_cpu(mpb->mpb_size);
5481 size_new = disks_to_mpb_size(info->nr_disks);
5482 if (size_new > size_old) {
5483 void *mpb_new;
f36a9ecd 5484 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5485
f36a9ecd 5486 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5487 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5488 return 0;
5489 }
85337573
AO		 5490 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5491 MIGR_REC_BUF_SECTORS*
5492 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5493 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5494 free(super->buf);
5495 free(super);
ea944c8f 5496 free(mpb_new);
8e59f3d8
AK		 5497 return 0;
5498 }
2c092cad
DW		 5499 memcpy(mpb_new, mpb, size_old);
5500 free(mpb);
5501 mpb = mpb_new;
949c47a0 5502 super->anchor = mpb_new;
2c092cad
DW		 5503 mpb->mpb_size = __cpu_to_le32(size_new);
5504 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5505 super->len = size_round;
2c092cad 5506 }
bf5a934a 5507 super->current_vol = idx;
3960e579
DW		 5508
5509 /* handle 'failed_disks' by either:
5510 * a) create dummy disk entries in the table if this the first
5511 * volume in the array. We add them here as this is the only
5512 * opportunity to add them. add_to_super_imsm_volume()
5513 * handles the non-failed disks and continues incrementing
5514 * mpb->num_disks.
5515 * b) validate that 'failed_disks' matches the current number
5516 * of missing disks if the container is populated
d23fe947 5517 */
3960e579 5518 if (super->current_vol == 0) {
d23fe947 5519 mpb->num_disks = 0;
3960e579
DW		 5520 for (i = 0; i < info->failed_disks; i++) {
5521 struct imsm_disk *disk;
5522
5523 mpb->num_disks++;
5524 disk = __get_imsm_disk(mpb, i);
5525 disk->status = CONFIGURED_DISK | FAILED_DISK;
5526 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5527 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5528 "missing:%d", (__u8)i);
3960e579
DW		 5529 }
5530 find_missing(super);
5531 } else {
5532 int missing = 0;
5533 struct dl *d;
5534
5535 for (d = super->missing; d; d = d->next)
5536 missing++;
5537 if (info->failed_disks > missing) {
e7b84f9d 5538 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5539 return 0;
5540 }
5541 }
5a038140 5542
aa534678
DW		 5543 if (!check_name(super, name, 0))
5544 return 0;
503975b9
N		 5545 dv = xmalloc(sizeof(*dv));
5546 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5547 /*
5548 * Explicitly allow truncating to not confuse gcc's
5549 * -Werror=stringop-truncation
5550 */
5551 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5552 memcpy(dev->volume, name, namelen);
e03640bd 5553 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5554 info->layout, info->chunk_size,
b53bfba6
TM		 5555 s->size * BLOCKS_PER_KB);
5556 data_disks = get_data_disks(info->level, info->layout,
5557 info->raid_disks);
5558 array_blocks = round_size_to_mb(array_blocks, data_disks);
5559 size_per_member = array_blocks / data_disks;
979d38be 5560
fcc2c9da 5561 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5562 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5563 vol = &dev->vol;
5564 vol->migr_state = 0;
1484e727 5565 set_migr_type(dev, MIGR_INIT);
3960e579 5566 vol->dirty = !info->state;
4036e7ee 5567 set_vol_curr_migr_unit(dev, 0);
238c0a71 5568 map = get_imsm_map(dev, MAP_0);
5551b113 5569 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5570 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5571 map->failed_disk_num = ~0;
bf4442ab 5572 if (info->level > 0)
fffaf1ff
N		 5573 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5574 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5575 else
5576 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5577 IMSM_T_STATE_NORMAL;
252d23c0 5578 map->ddf = 1;
ef6ffade
DW		 5579
5580 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5581 free(dev);
5582 free(dv);
7a862a02 5583 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5584 return 0;
5585 }
81062a36
DW		 5586
5587 map->raid_level = info->level;
4d1313e9 5588 if (info->level == 10) {
c2c087e6 5589 map->raid_level = 1;
4d1313e9 5590 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5591 } else if (info->level == 1)
5592 map->num_domains = info->raid_disks;
5593 else
ff596308 5594 map->num_domains = 1;
81062a36 5595
5551b113 5596 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5597 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5598 num_data_stripes /= map->num_domains;
5599 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5600
44490938
MD		 5601 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5602 set_blocks_per_member(map, info_to_blocks_per_member(info,
5603 size_per_member /
5604 BLOCKS_PER_KB));
5605
c2c087e6
DW		 5606 map->num_members = info->raid_disks;
5607 for (i = 0; i < map->num_members; i++) {
5608 /* initialized in add_to_super */
4eb26970 5609 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5610 }
949c47a0 5611 mpb->num_raid_devs++;
2a24dc1b
PB		 5612 mpb->num_raid_devs_created++;
5613 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5614
b7580566 5615 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5616 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5617 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5618 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5619 } else {
5620 free(dev);
5621 free(dv);
5622 pr_err("imsm does not support consistency policy %s\n",
5623 map_num(consistency_policies, s->consistency_policy));
5624 return 0;
5625 }
5626
ba2de7ba
DW		 5627 dv->dev = dev;
5628 dv->index = super->current_vol;
5629 dv->next = super->devlist;
5630 super->devlist = dv;
c2c087e6 5631
4d1313e9
DW		 5632 imsm_update_version_info(super);
5633
c2c087e6 5634 return 1;
cdddbdbc
DW		 5635}
5636
bf5a934a 5637static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5638 struct shape *s, char *name,
83cd1e97
N		 5639 char *homehost, int *uuid,
5640 unsigned long long data_offset)
bf5a934a
DW		 5641{
5642 /* This is primarily called by Create when creating a new array.
5643 * We will then get add_to_super called for each component, and then
5644 * write_init_super called to write it out to each device.
5645 * For IMSM, Create can create on fresh devices or on a pre-existing
5646 * array.
5647 * To create on a pre-existing array a different method will be called.
5648 * This one is just for fresh drives.
5649 */
5650 struct intel_super *super;
5651 struct imsm_super *mpb;
5652 size_t mpb_size;
4d1313e9 5653 char *version;
bf5a934a 5654
83cd1e97 5655 if (data_offset != INVALID_SECTORS) {
ed503f89 5656 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5657 return 0;
5658 }
5659
bf5a934a 5660 if (st->sb)
5308f117 5661 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5662 data_offset);
e683ca88
DW		 5663
5664 if (info)
5665 mpb_size = disks_to_mpb_size(info->nr_disks);
5666 else
f36a9ecd 5667 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5668
49133e57 5669 super = alloc_super();
f36a9ecd
PB		 5670 if (super &&
5671 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5672 free_imsm(super);
e683ca88
DW		 5673 super = NULL;
5674 }
5675 if (!super) {
1ade5cc1 5676 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5677 return 0;
5678 }
de44e46f
PB		 5679 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5680 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5681 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5682 free(super->buf);
8d67477f 5683 free_imsm(super);
8e59f3d8
AK		 5684 return 0;
5685 }
e683ca88 5686 memset(super->buf, 0, mpb_size);
ef649044 5687 mpb = super->buf;
e683ca88
DW		 5688 mpb->mpb_size = __cpu_to_le32(mpb_size);
5689 st->sb = super;
5690
5691 if (info == NULL) {
5692 /* zeroing superblock */
5693 return 0;
5694 }
bf5a934a 5695
4d1313e9
DW		 5696 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5697
5698 version = (char *) mpb->sig;
5699 strcpy(version, MPB_SIGNATURE);
5700 version += strlen(MPB_SIGNATURE);
5701 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5702
bf5a934a
DW		 5703 return 1;
5704}
5705
f2cc4f7d
AO		 5706static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5707{
5708 unsigned int member_sector_size;
5709
5710 if (dl->fd < 0) {
5711 pr_err("Invalid file descriptor for %s\n", dl->devname);
5712 return 0;
5713 }
5714
5715 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5716 return 0;
5717 if (member_sector_size != super->sector_size)
5718 return 0;
5719 return 1;
5720}
5721
f20c3968 5722static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5723 int fd, char *devname)
5724{
5725 struct intel_super *super = st->sb;
d23fe947 5726 struct imsm_super *mpb = super->anchor;
3960e579 5727 struct imsm_disk *_disk;
bf5a934a
DW		 5728 struct imsm_dev *dev;
5729 struct imsm_map *map;
3960e579 5730 struct dl *dl, *df;
4eb26970 5731 int slot;
bf5a934a 5732
949c47a0 5733 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5734 map = get_imsm_map(dev, MAP_0);
bf5a934a 5735
208933a7 5736 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5737 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5738 devname);
5739 return 1;
5740 }
5741
efb30e7f
DW		 5742 if (fd == -1) {
5743 /* we're doing autolayout so grab the pre-marked (in
5744 * validate_geometry) raid_disk
5745 */
5746 for (dl = super->disks; dl; dl = dl->next)
5747 if (dl->raiddisk == dk->raid_disk)
5748 break;
5749 } else {
5750 for (dl = super->disks; dl ; dl = dl->next)
5751 if (dl->major == dk->major &&
5752 dl->minor == dk->minor)
5753 break;
5754 }
d23fe947 5755
208933a7 5756 if (!dl) {
e7b84f9d 5757 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5758 return 1;
208933a7 5759 }
bf5a934a 5760
59632db9
MZ		 5761 if (mpb->num_disks == 0)
5762 if (!get_dev_sector_size(dl->fd, dl->devname,
5763 &super->sector_size))
5764 return 1;
5765
f2cc4f7d
AO		 5766 if (!drive_validate_sector_size(super, dl)) {
5767 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5768 return 1;
5769 }
5770
d23fe947
DW		 5771 /* add a pristine spare to the metadata */
5772 if (dl->index < 0) {
5773 dl->index = super->anchor->num_disks;
5774 super->anchor->num_disks++;
5775 }
4eb26970
DW		 5776 /* Check the device has not already been added */
5777 slot = get_imsm_disk_slot(map, dl->index);
5778 if (slot >= 0 &&
238c0a71 5779 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5780 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5781 devname);
5782 return 1;
5783 }
656b6b5a 5784 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5785 dl->disk.status = CONFIGURED_DISK;
d23fe947 5786
3960e579
DW		 5787 /* update size of 'missing' disks to be at least as large as the
5788 * largest acitve member (we only have dummy missing disks when
5789 * creating the first volume)
5790 */
5791 if (super->current_vol == 0) {
5792 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5793 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5794 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5795 _disk = __get_imsm_disk(mpb, df->index);
5796 *_disk = df->disk;
5797 }
5798 }
5799
5800 /* refresh unset/failed slots to point to valid 'missing' entries */
5801 for (df = super->missing; df; df = df->next)
5802 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5803 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5804
5805 if ((ord & IMSM_ORD_REBUILD) == 0)
5806 continue;
5807 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5808 if (is_gen_migration(dev)) {
238c0a71
AK		 5809 struct imsm_map *map2 = get_imsm_map(dev,
5810 MAP_1);
0a108d63 5811 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5812 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5813 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5814 slot2,
5815 MAP_1);
1ace8403
AK		 5816 if ((unsigned)df->index ==
5817 ord_to_idx(ord2))
5818 set_imsm_ord_tbl_ent(map2,
0a108d63 5819 slot2,
1ace8403
AK		 5820 df->index |
5821 IMSM_ORD_REBUILD);
5822 }
5823 }
3960e579
DW		 5824 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5825 break;
5826 }
5827
d23fe947
DW		 5828 /* if we are creating the first raid device update the family number */
5829 if (super->current_vol == 0) {
5830 __u32 sum;
5831 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5832
3960e579 5833 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5834 if (!_dev || !_disk) {
e7b84f9d 5835 pr_err("BUG mpb setup error\n");
791b666a
AW		 5836 return 1;
5837 }
d23fe947
DW		 5838 *_dev = *dev;
5839 *_disk = dl->disk;
148acb7b
DW		 5840 sum = random32();
5841 sum += __gen_imsm_checksum(mpb);
d23fe947 5842 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5843 mpb->orig_family_num = mpb->family_num;
e48aed3c 5844 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5845 }
ca0748fa 5846 super->current_disk = dl;
f20c3968 5847 return 0;
bf5a934a
DW		 5848}
5849
a8619d23
AK		 5850/* mark_spare()
5851 * Function marks disk as spare and restores disk serial
5852 * in case it was previously marked as failed by takeover operation
5853 * reruns:
5854 * -1 : critical error
5855 * 0 : disk is marked as spare but serial is not set
5856 * 1 : success
5857 */
5858int mark_spare(struct dl *disk)
5859{
5860 __u8 serial[MAX_RAID_SERIAL_LEN];
5861 int ret_val = -1;
5862
5863 if (!disk)
5864 return ret_val;
5865
5866 ret_val = 0;
6da53c0e 5867 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5868 /* Restore disk serial number, because takeover marks disk
5869 * as failed and adds to serial ':0' before it becomes
5870 * a spare disk.
5871 */
5872 serialcpy(disk->serial, serial);
5873 serialcpy(disk->disk.serial, serial);
5874 ret_val = 1;
5875 }
5876 disk->disk.status = SPARE_DISK;
5877 disk->index = -1;
5878
5879 return ret_val;
5880}
88654014 5881
12724c01
TM		 5882
5883static int write_super_imsm_spare(struct intel_super *super, struct dl *d);
5884
f20c3968 5885static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5886 int fd, char *devname,
5887 unsigned long long data_offset)
cdddbdbc 5888{
c2c087e6 5889 struct intel_super *super = st->sb;
c2c087e6
DW		 5890 struct dl *dd;
5891 unsigned long long size;
fa7bb6f8 5892 unsigned int member_sector_size;
f2f27e63 5893 __u32 id;
c2c087e6
DW		 5894 int rv;
5895 struct stat stb;
5896
88654014
LM		 5897 /* If we are on an RAID enabled platform check that the disk is
5898 * attached to the raid controller.
5899 * We do not need to test disks attachment for container based additions,
5900 * they shall be already tested when container was created/assembled.
88c32bb1 5901 */
d424212e 5902 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5903 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5904 if (rv != 0) {
5905 dprintf("capability: %p fd: %d ret: %d\n",
5906 super->orom, fd, rv);
5907 return 1;
88c32bb1
DW		 5908 }
5909
f20c3968
DW		 5910 if (super->current_vol >= 0)
5911 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5912
c2c087e6 5913 fstat(fd, &stb);
503975b9 5914 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5915 dd->major = major(stb.st_rdev);
5916 dd->minor = minor(stb.st_rdev);
503975b9 5917 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5918 dd->fd = fd;
689c9bf3 5919 dd->e = NULL;
1a64be56 5920 dd->action = DISK_ADD;
6da53c0e 5921 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5922 if (rv) {
e7b84f9d 5923 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5924 if (dd->devname)
5925 free(dd->devname);
949c47a0 5926 free(dd);
0030e8d6 5927 abort();
c2c087e6 5928 }
7c798f87 5929
20bee0f8
PB		 5930 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5931 (super->hba->type == SYS_DEV_VMD))) {
5932 int i;
7c798f87
MT		 5933 char cntrl_path[PATH_MAX];
5934 char *cntrl_name;
5935 char pci_dev_path[PATH_MAX];
20bee0f8 5936
7c798f87
MT		 5937 if (!diskfd_to_devpath(fd, 2, pci_dev_path) ||
5938 !diskfd_to_devpath(fd, 1, cntrl_path)) {
5939 pr_err("failed to get dev_path, aborting\n");
20bee0f8
PB		 5940 if (dd->devname)
5941 free(dd->devname);
5942 free(dd);
5943 return 1;
5944 }
5945
a8f3cfd5
MT		 5946 if (!imsm_is_nvme_supported(dd->fd, 1)) {
5947 if (dd->devname)
5948 free(dd->devname);
5949 free(dd);
5950 return 1;
5951 }
5952
7c798f87
MT		 5953 cntrl_name = basename(cntrl_path);
5954 if (is_multipath_nvme(fd))
5955 pr_err("%s controller supports Multi-Path I/O, Intel (R) VROC does not support multipathing\n",
5956 cntrl_name);
5957
5958 if (devpath_to_vendor(pci_dev_path) == 0x8086) {
20bee0f8
PB		 5959 /*
5960 * If Intel's NVMe drive has serial ended with
5961 * "-A","-B","-1" or "-2" it means that this is "x8"
5962 * device (double drive on single PCIe card).
5963 * User should be warned about potential data loss.
5964 */
5965 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5966 /* Skip empty character at the end */
5967 if (dd->serial[i] == 0)
5968 continue;
5969
5970 if (((dd->serial[i] == 'A') ||
5971 (dd->serial[i] == 'B') ||
5972 (dd->serial[i] == '1') ||
5973 (dd->serial[i] == '2')) &&
5974 (dd->serial[i-1] == '-'))
5975 pr_err("\tThe action you are about to take may put your data at risk.\n"
5976 "\tPlease note that x8 devices may consist of two separate x4 devices "
5977 "located on a single PCIe port.\n"
5978 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5979 break;
5980 }
32716c51
PB		 5981 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5982 !imsm_orom_has_tpv_support(super->orom)) {
5983 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5984 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5985 free(dd->devname);
5986 free(dd);
5987 return 1;
20bee0f8
PB		 5988 }
5989 }
c2c087e6 5990
c2c087e6 5991 get_dev_size(fd, NULL, &size);
aec01630
JS		 5992 if (!get_dev_sector_size(fd, NULL, &member_sector_size))
5993 return 1;
fa7bb6f8
PB		 5994
5995 if (super->sector_size == 0) {
5996 /* this a first device, so sector_size is not set yet */
5997 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5998 }
5999
71e5411e 6000 /* clear migr_rec when adding disk to container */
85337573
AO		 6001 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
6002 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 6003 SEEK_SET) >= 0) {
466070ad 6004 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 6005 MIGR_REC_BUF_SECTORS*member_sector_size) !=
6006 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 6007 perror("Write migr_rec failed");
6008 }
6009
c2c087e6 6010 size /= 512;
1f24f035 6011 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 6012 set_total_blocks(&dd->disk, size);
6013 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
6014 struct imsm_super *mpb = super->anchor;
6015 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
6016 }
a8619d23 6017 mark_spare(dd);
c2c087e6 6018 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 6019 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 6020 else
b9f594fe 6021 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 6022
6023 if (st->update_tail) {
1a64be56
LM		 6024 dd->next = super->disk_mgmt_list;
6025 super->disk_mgmt_list = dd;
43dad3d6 6026 } else {
12724c01
TM		 6027 /* this is called outside of mdmon
6028 * write initial spare metadata
6029 * mdmon will overwrite it.
6030 */
43dad3d6
DW		 6031 dd->next = super->disks;
6032 super->disks = dd;
12724c01 6033 write_super_imsm_spare(super, dd);
43dad3d6 6034 }
f20c3968
DW		 6035
6036 return 0;
cdddbdbc
DW		 6037}
6038
1a64be56
LM		 6039static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
6040{
6041 struct intel_super *super = st->sb;
6042 struct dl *dd;
6043
6044 /* remove from super works only in mdmon - for communication
6045 * manager - monitor. Check if communication memory buffer
6046 * is prepared.
6047 */
6048 if (!st->update_tail) {
1ade5cc1 6049 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 6050 return 1;
6051 }
503975b9 6052 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 6053 dd->major = dk->major;
6054 dd->minor = dk->minor;
1a64be56 6055 dd->fd = -1;
a8619d23 6056 mark_spare(dd);
1a64be56
LM		 6057 dd->action = DISK_REMOVE;
6058
6059 dd->next = super->disk_mgmt_list;
6060 super->disk_mgmt_list = dd;
6061
1a64be56
LM		 6062 return 0;
6063}
6064
f796af5d
DW		 6065static int store_imsm_mpb(int fd, struct imsm_super *mpb);
6066
6067static union {
f36a9ecd 6068 char buf[MAX_SECTOR_SIZE];
f796af5d 6069 struct imsm_super anchor;
f36a9ecd 6070} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 6071
12724c01
TM		 6072
6073static int write_super_imsm_spare(struct intel_super *super, struct dl *d)
d23fe947 6074{
d23fe947 6075 struct imsm_super *mpb = super->anchor;
f796af5d 6076 struct imsm_super *spare = &spare_record.anchor;
d23fe947 6077 __u32 sum;
12724c01
TM		 6078
6079 if (d->index != -1)
6080 return 1;
d23fe947 6081
68641cdb
JS		 6082 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6083 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6084 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6085 spare->num_disks = 1;
6086 spare->num_raid_devs = 0;
6087 spare->cache_size = mpb->cache_size;
6088 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6089
6090 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6091 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947 6092
12724c01
TM		 6093 spare->disk[0] = d->disk;
6094 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6095 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6096
6097 if (super->sector_size == 4096)
6098 convert_to_4k_imsm_disk(&spare->disk[0]);
d23fe947 6099
12724c01
TM		 6100 sum = __gen_imsm_checksum(spare);
6101 spare->family_num = __cpu_to_le32(sum);
6102 spare->orig_family_num = 0;
6103 sum = __gen_imsm_checksum(spare);
6104 spare->check_sum = __cpu_to_le32(sum);
027c374f 6105
12724c01
TM		 6106 if (store_imsm_mpb(d->fd, spare)) {
6107 pr_err("failed for device %d:%d %s\n",
6108 d->major, d->minor, strerror(errno));
6109 return 1;
6110 }
6111
6112 return 0;
6113}
6114/* spare records have their own family number and do not have any defined raid
6115 * devices
6116 */
6117static int write_super_imsm_spares(struct intel_super *super, int doclose)
6118{
6119 struct dl *d;
f36a9ecd 6120
12724c01
TM		 6121 for (d = super->disks; d; d = d->next) {
6122 if (d->index != -1)
6123 continue;
d23fe947 6124
12724c01 6125 if (write_super_imsm_spare(super, d))
e74255d9 6126 return 1;
12724c01 6127
d23fe947
DW		 6128 if (doclose) {
6129 close(d->fd);
6130 d->fd = -1;
6131 }
6132 }
6133
e74255d9 6134 return 0;
d23fe947
DW		 6135}
6136
36988a3d 6137static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6138{
36988a3d 6139 struct intel_super *super = st->sb;
f36a9ecd 6140 unsigned int sector_size = super->sector_size;
949c47a0 6141 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6142 struct dl *d;
6143 __u32 generation;
6144 __u32 sum;
d23fe947 6145 int spares = 0;
949c47a0 6146 int i;
a48ac0a8 6147 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6148 int num_disks = 0;
146c6260 6149 int clear_migration_record = 1;
bbab0940 6150 __u32 bbm_log_size;
cdddbdbc 6151
c2c087e6
DW		 6152 /* 'generation' is incremented everytime the metadata is written */
6153 generation = __le32_to_cpu(mpb->generation_num);
6154 generation++;
6155 mpb->generation_num = __cpu_to_le32(generation);
6156
148acb7b
DW		 6157 /* fix up cases where previous mdadm releases failed to set
6158 * orig_family_num
6159 */
6160 if (mpb->orig_family_num == 0)
6161 mpb->orig_family_num = mpb->family_num;
6162
d23fe947 6163 for (d = super->disks; d; d = d->next) {
8796fdc4 6164 if (d->index == -1)
d23fe947 6165 spares++;
36988a3d 6166 else {
d23fe947 6167 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6168 num_disks++;
6169 }
d23fe947 6170 }
36988a3d 6171 for (d = super->missing; d; d = d->next) {
47ee5a45 6172 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6173 num_disks++;
6174 }
6175 mpb->num_disks = num_disks;
6176 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6177
949c47a0
DW		 6178 for (i = 0; i < mpb->num_raid_devs; i++) {
6179 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6180 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6181 if (dev && dev2) {
6182 imsm_copy_dev(dev, dev2);
6183 mpb_size += sizeof_imsm_dev(dev, 0);
6184 }
146c6260
AK		 6185 if (is_gen_migration(dev2))
6186 clear_migration_record = 0;
949c47a0 6187 }
bbab0940
TM		 6188
6189 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6190
6191 if (bbm_log_size) {
6192 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6193 mpb->attributes |= MPB_ATTRIB_BBM;
6194 } else
6195 mpb->attributes &= ~MPB_ATTRIB_BBM;
6196
6197 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6198 mpb_size += bbm_log_size;
a48ac0a8 6199 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6200
bbab0940
TM		 6201#ifdef DEBUG
6202 assert(super->len == 0 || mpb_size <= super->len);
6203#endif
6204
c2c087e6 6205 /* recalculate checksum */
949c47a0 6206 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6207 mpb->check_sum = __cpu_to_le32(sum);
6208
51d83f5d
AK		 6209 if (super->clean_migration_record_by_mdmon) {
6210 clear_migration_record = 1;
6211 super->clean_migration_record_by_mdmon = 0;
6212 }
146c6260 6213 if (clear_migration_record)
de44e46f 6214 memset(super->migr_rec_buf, 0,
85337573 6215 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6216
f36a9ecd
PB		 6217 if (sector_size == 4096)
6218 convert_to_4k(super);
6219
d23fe947 6220 /* write the mpb for disks that compose raid devices */
c2c087e6 6221 for (d = super->disks; d ; d = d->next) {
86c54047 6222 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6223 continue;
30602f53 6224
146c6260
AK		 6225 if (clear_migration_record) {
6226 unsigned long long dsize;
6227
6228 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6229 if (lseek64(d->fd, dsize - sector_size,
6230 SEEK_SET) >= 0) {
466070ad
PB		 6231 if ((unsigned int)write(d->fd,
6232 super->migr_rec_buf,
de44e46f
PB		 6233 MIGR_REC_BUF_SECTORS*sector_size) !=
6234 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6235 perror("Write migr_rec failed");
146c6260
AK		 6236 }
6237 }
51d83f5d
AK		 6238
6239 if (store_imsm_mpb(d->fd, mpb))
6240 fprintf(stderr,
1ade5cc1
N		 6241 "failed for device %d:%d (fd: %d)%s\n",
6242 d->major, d->minor,
51d83f5d
AK		 6243 d->fd, strerror(errno));
6244
c2c087e6
DW		 6245 if (doclose) {
6246 close(d->fd);
6247 d->fd = -1;
6248 }
6249 }
6250
d23fe947
DW		 6251 if (spares)
6252 return write_super_imsm_spares(super, doclose);
6253
e74255d9 6254 return 0;
c2c087e6
DW		 6255}
6256
9b1fb677 6257static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6258{
6259 size_t len;
6260 struct imsm_update_create_array *u;
6261 struct intel_super *super = st->sb;
9b1fb677 6262 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6263 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6264 struct disk_info *inf;
6265 struct imsm_disk *disk;
6266 int i;
43dad3d6 6267
54c2c1ea
DW		 6268 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6269 sizeof(*inf) * map->num_members;
503975b9 6270 u = xmalloc(len);
43dad3d6 6271 u->type = update_create_array;
9b1fb677 6272 u->dev_idx = dev_idx;
43dad3d6 6273 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6274 inf = get_disk_info(u);
6275 for (i = 0; i < map->num_members; i++) {
238c0a71 6276 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6277
54c2c1ea 6278 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6279 if (!disk)
6280 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6281 serialcpy(inf[i].serial, disk->serial);
6282 }
43dad3d6
DW		 6283 append_metadata_update(st, u, len);
6284
6285 return 0;
6286}
6287
1a64be56 6288static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6289{
6290 struct intel_super *super = st->sb;
6291 size_t len;
1a64be56 6292 struct imsm_update_add_remove_disk *u;
43dad3d6 6293
1a64be56 6294 if (!super->disk_mgmt_list)
43dad3d6
DW		 6295 return 0;
6296
6297 len = sizeof(*u);
503975b9 6298 u = xmalloc(len);
1a64be56 6299 u->type = update_add_remove_disk;
43dad3d6
DW		 6300 append_metadata_update(st, u, len);
6301
6302 return 0;
6303}
2432ce9b
AP		 6304
6305__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6306
e397cefe
AP		 6307static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6308{
6309 struct ppl_header *ppl_hdr = buf;
6310 int ret;
6311
6312 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6313
6314 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6315 ret = -errno;
6316 perror("Failed to seek to PPL header location");
6317 return ret;
6318 }
6319
6320 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6321 ret = -errno;
6322 perror("Write PPL header failed");
6323 return ret;
6324 }
6325
6326 fsync(fd);
6327
6328 return 0;
6329}
6330
2432ce9b
AP		 6331static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6332{
6333 struct intel_super *super = st->sb;
6334 void *buf;
6335 struct ppl_header *ppl_hdr;
6336 int ret;
6337
b2514242
PB		 6338 /* first clear entire ppl space */
6339 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6340 if (ret)
6341 return ret;
6342
6343 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6344 if (ret) {
6345 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6346 return -ret;
2432ce9b
AP		 6347 }
6348
6349 memset(buf, 0, PPL_HEADER_SIZE);
6350 ppl_hdr = buf;
6351 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6352 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6353
6354 if (info->mismatch_cnt) {
6355 /*
6356 * We are overwriting an invalid ppl. Make one entry with wrong
6357 * checksum to prevent the kernel from skipping resync.
6358 */
6359 ppl_hdr->entries_count = __cpu_to_le32(1);
6360 ppl_hdr->entries[0].checksum = ~0;
6361 }
6362
e397cefe 6363 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6364
6365 free(buf);
6366 return ret;
6367}
6368
e397cefe
AP		 6369static int is_rebuilding(struct imsm_dev *dev);
6370
2432ce9b
AP		 6371static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6372 struct mdinfo *disk)
6373{
6374 struct intel_super *super = st->sb;
6375 struct dl *d;
e397cefe 6376 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6377 int ret = 0;
e397cefe 6378 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6379 __u32 crc;
6380 struct imsm_dev *dev;
2432ce9b 6381 __u32 idx;
44b6b876
PB		 6382 unsigned int i;
6383 unsigned long long ppl_offset = 0;
6384 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6385
6386 if (disk->disk.raid_disk < 0)
6387 return 0;
6388
2432ce9b 6389 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6390 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6391 d = get_imsm_dl_disk(super, idx);
6392
6393 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6394 return 0;
6395
6396 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6397 pr_err("Failed to allocate PPL header buffer\n");
6398 return -1;
6399 }
6400 buf = buf_orig;
2432ce9b 6401
44b6b876
PB		 6402 ret = 1;
6403 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6404 void *tmp;
6405
44b6b876 6406 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6407
44b6b876
PB		 6408 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6409 SEEK_SET) < 0) {
6410 perror("Failed to seek to PPL header location");
6411 ret = -1;
e397cefe 6412 break;
44b6b876 6413 }
2432ce9b 6414
44b6b876
PB		 6415 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6416 perror("Read PPL header failed");
6417 ret = -1;
e397cefe 6418 break;
44b6b876 6419 }
2432ce9b 6420
44b6b876 6421 ppl_hdr = buf;
2432ce9b 6422
44b6b876
PB		 6423 crc = __le32_to_cpu(ppl_hdr->checksum);
6424 ppl_hdr->checksum = 0;
6425
6426 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6427 dprintf("Wrong PPL header checksum on %s\n",
6428 d->devname);
e397cefe 6429 break;
44b6b876
PB		 6430 }
6431
6432 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6433 /* previous was newest, it was already checked */
e397cefe 6434 break;
44b6b876
PB		 6435 }
6436
6437 if ((__le32_to_cpu(ppl_hdr->signature) !=
6438 super->anchor->orig_family_num)) {
6439 dprintf("Wrong PPL header signature on %s\n",
6440 d->devname);
6441 ret = 1;
e397cefe 6442 break;
44b6b876
PB		 6443 }
6444
6445 ret = 0;
6446 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6447
44b6b876
PB		 6448 ppl_offset += PPL_HEADER_SIZE;
6449 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6450 ppl_offset +=
6451 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6452
6453 if (!buf_prev)
6454 buf_prev = buf + PPL_HEADER_SIZE;
6455 tmp = buf_prev;
6456 buf_prev = buf;
6457 buf = tmp;
2432ce9b
AP		 6458 }
6459
e397cefe
AP		 6460 if (buf_prev) {
6461 buf = buf_prev;
6462 ppl_hdr = buf_prev;
6463 }
2432ce9b 6464
54148aba
PB		 6465 /*
6466 * Update metadata to use mutliple PPLs area (1MB).
6467 * This is done once for all RAID members
6468 */
6469 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6470 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6471 char subarray[20];
6472 struct mdinfo *member_dev;
6473
6474 sprintf(subarray, "%d", info->container_member);
6475
6476 if (mdmon_running(st->container_devnm))
6477 st->update_tail = &st->updates;
6478
6479 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6480 pr_err("Failed to update subarray %s\n",
6481 subarray);
6482 } else {
6483 if (st->update_tail)
6484 flush_metadata_updates(st);
6485 else
6486 st->ss->sync_metadata(st);
6487 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6488 for (member_dev = info->devs; member_dev;
6489 member_dev = member_dev->next)
6490 member_dev->ppl_size =
6491 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6492 }
6493 }
6494
b23d0750 6495 if (ret == 1) {
2fc0fc63
AP		 6496 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6497
50b9c10d
PB		 6498 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6499 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6500 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6501 (is_rebuilding(dev) &&
4036e7ee 6502 vol_curr_migr_unit(dev) == 0 &&
2ec9d182 6503 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6504 ret = st->ss->write_init_ppl(st, info, d->fd);
6505 else
6506 info->mismatch_cnt++;
e397cefe
AP		 6507 } else if (ret == 0 &&
6508 ppl_hdr->entries_count == 0 &&
6509 is_rebuilding(dev) &&
6510 info->resync_start == 0) {
6511 /*
6512 * The header has no entries - add a single empty entry and
6513 * rewrite the header to prevent the kernel from going into
6514 * resync after an interrupted rebuild.
6515 */
6516 ppl_hdr->entries_count = __cpu_to_le32(1);
6517 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6518 }
2432ce9b 6519
e397cefe
AP		 6520 free(buf_orig);
6521
2432ce9b
AP		 6522 return ret;
6523}
6524
2432ce9b
AP		 6525static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6526{
6527 struct intel_super *super = st->sb;
6528 struct dl *d;
6529 int ret = 0;
6530
6531 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6532 info->array.level != 5)
6533 return 0;
6534
6535 for (d = super->disks; d ; d = d->next) {
6536 if (d->index < 0 || is_failed(&d->disk))
6537 continue;
6538
6539 ret = st->ss->write_init_ppl(st, info, d->fd);
6540 if (ret)
6541 break;
6542 }
6543
6544 return ret;
6545}
43dad3d6 6546
fbc42556
JR		 6547/*******************************************************************************
6548 * Function: write_init_bitmap_imsm_vol
6549 * Description: Write a bitmap header and prepares the area for the bitmap.
6550 * Parameters:
6551 * st : supertype information
6552 * vol_idx : the volume index to use
6553 *
6554 * Returns:
6555 * 0 : success
6556 * -1 : fail
6557 ******************************************************************************/
6558static int write_init_bitmap_imsm_vol(struct supertype *st, int vol_idx)
6559{
6560 struct intel_super *super = st->sb;
6561 int prev_current_vol = super->current_vol;
6562 struct dl *d;
6563 int ret = 0;
6564
6565 super->current_vol = vol_idx;
6566 for (d = super->disks; d; d = d->next) {
6567 if (d->index < 0 || is_failed(&d->disk))
6568 continue;
6569 ret = st->ss->write_bitmap(st, d->fd, NoUpdate);
6570 if (ret)
6571 break;
6572 }
6573 super->current_vol = prev_current_vol;
6574 return ret;
6575}
6576
6577/*******************************************************************************
6578 * Function: write_init_bitmap_imsm_all
6579 * Description: Write a bitmap header and prepares the area for the bitmap.
6580 * Operation is executed for volumes with CONSISTENCY_POLICY_BITMAP.
6581 * Parameters:
6582 * st : supertype information
6583 * info : info about the volume where the bitmap should be written
6584 * vol_idx : the volume index to use
6585 *
6586 * Returns:
6587 * 0 : success
6588 * -1 : fail
6589 ******************************************************************************/
6590static int write_init_bitmap_imsm_all(struct supertype *st, struct mdinfo *info,
6591 int vol_idx)
6592{
6593 int ret = 0;
6594
6595 if (info && (info->consistency_policy == CONSISTENCY_POLICY_BITMAP))
6596 ret = write_init_bitmap_imsm_vol(st, vol_idx);
6597
6598 return ret;
6599}
6600
c2c087e6
DW		 6601static int write_init_super_imsm(struct supertype *st)
6602{
9b1fb677
DW		 6603 struct intel_super *super = st->sb;
6604 int current_vol = super->current_vol;
2432ce9b
AP		 6605 int rv = 0;
6606 struct mdinfo info;
6607
6608 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6609
6610 /* we are done with current_vol reset it to point st at the container */
6611 super->current_vol = -1;
6612
8273f55e 6613 if (st->update_tail) {
43dad3d6
DW		 6614 /* queue the recently created array / added disk
6615 * as a metadata update */
8273f55e 6616
43dad3d6 6617 /* determine if we are creating a volume or adding a disk */
9b1fb677 6618 if (current_vol < 0) {
1a64be56
LM		 6619 /* in the mgmt (add/remove) disk case we are running
6620 * in mdmon context, so don't close fd's
43dad3d6 6621 */
2432ce9b
AP		 6622 rv = mgmt_disk(st);
6623 } else {
fbc42556 6624 /* adding the second volume to the array */
2432ce9b 6625 rv = write_init_ppl_imsm_all(st, &info);
fbc42556
JR		 6626 if (!rv)
6627 rv = write_init_bitmap_imsm_all(st, &info, current_vol);
2432ce9b
AP		 6628 if (!rv)
6629 rv = create_array(st, current_vol);
6630 }
d682f344
N		 6631 } else {
6632 struct dl *d;
6633 for (d = super->disks; d; d = d->next)
ba728be7 6634 Kill(d->devname, NULL, 0, -1, 1);
fbc42556 6635 if (current_vol >= 0) {
2432ce9b 6636 rv = write_init_ppl_imsm_all(st, &info);
fbc42556
JR		 6637 if (!rv)
6638 rv = write_init_bitmap_imsm_all(st, &info, current_vol);
6639 }
6640
2432ce9b
AP		 6641 if (!rv)
6642 rv = write_super_imsm(st, 1);
d682f344 6643 }
2432ce9b
AP		 6644
6645 return rv;
cdddbdbc
DW		 6646}
6647
e683ca88 6648static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6649{
e683ca88
DW		 6650 struct intel_super *super = st->sb;
6651 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6652
e683ca88 6653 if (!mpb)
ad97895e
DW		 6654 return 1;
6655
f36a9ecd
PB		 6656 if (super->sector_size == 4096)
6657 convert_to_4k(super);
e683ca88 6658 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6659}
6660
cdddbdbc
DW		 6661static int validate_geometry_imsm_container(struct supertype *st, int level,
6662 int layout, int raiddisks, int chunk,
af4348dd
N		 6663 unsigned long long size,
6664 unsigned long long data_offset,
6665 char *dev,
2c514b71
NB		 6666 unsigned long long *freesize,
6667 int verbose)
cdddbdbc 6668{
c2c087e6
DW		 6669 int fd;
6670 unsigned long long ldsize;
594dc1b8 6671 struct intel_super *super;
f2f5c343 6672 int rv = 0;
cdddbdbc 6673
c2c087e6
DW		 6674 if (level != LEVEL_CONTAINER)
6675 return 0;
6676 if (!dev)
6677 return 1;
6678
6679 fd = open(dev, O_RDONLY|O_EXCL, 0);
6680 if (fd < 0) {
ba728be7 6681 if (verbose > 0)
e7b84f9d 6682 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6683 dev, strerror(errno));
c2c087e6
DW		 6684 return 0;
6685 }
6686 if (!get_dev_size(fd, dev, &ldsize)) {
6687 close(fd);
6688 return 0;
6689 }
f2f5c343
LM		 6690
6691 /* capabilities retrieve could be possible
6692 * note that there is no fd for the disks in array.
6693 */
6694 super = alloc_super();
8d67477f
TM		 6695 if (!super) {
6696 close(fd);
6697 return 0;
6698 }
fa7bb6f8
PB		 6699 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6700 close(fd);
6701 free_imsm(super);
6702 return 0;
6703 }
6704
ba728be7 6705 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6706 if (rv != 0) {
6707#if DEBUG
6708 char str[256];
6709 fd2devname(fd, str);
1ade5cc1 6710 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6711 fd, str, super->orom, rv, raiddisks);
6712#endif
6713 /* no orom/efi or non-intel hba of the disk */
6714 close(fd);
6715 free_imsm(super);
6716 return 0;
6717 }
c2c087e6 6718 close(fd);
9126b9a8
CA		 6719 if (super->orom) {
6720 if (raiddisks > super->orom->tds) {
6721 if (verbose)
7a862a02 6722 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6723 raiddisks, super->orom->tds);
6724 free_imsm(super);
6725 return 0;
6726 }
6727 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6728 (ldsize >> 9) >> 32 > 0) {
6729 if (verbose)
e7b84f9d 6730 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6731 free_imsm(super);
6732 return 0;
6733 }
f2f5c343 6734 }
c2c087e6 6735
af4348dd 6736 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6737 free_imsm(super);
c2c087e6
DW		 6738
6739 return 1;
cdddbdbc
DW		 6740}
6741
0dcecb2e
DW		 6742static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6743{
6744 const unsigned long long base_start = e[*idx].start;
6745 unsigned long long end = base_start + e[*idx].size;
6746 int i;
6747
6748 if (base_start == end)
6749 return 0;
6750
6751 *idx = *idx + 1;
6752 for (i = *idx; i < num_extents; i++) {
6753 /* extend overlapping extents */
6754 if (e[i].start >= base_start &&
6755 e[i].start <= end) {
6756 if (e[i].size == 0)
6757 return 0;
6758 if (e[i].start + e[i].size > end)
6759 end = e[i].start + e[i].size;
6760 } else if (e[i].start > end) {
6761 *idx = i;
6762 break;
6763 }
6764 }
6765
6766 return end - base_start;
6767}
6768
6769static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6770{
6771 /* build a composite disk with all known extents and generate a new
6772 * 'maxsize' given the "all disks in an array must share a common start
6773 * offset" constraint
6774 */
503975b9 6775 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6776 struct dl *dl;
6777 int i, j;
6778 int start_extent;
6779 unsigned long long pos;
b9d77223 6780 unsigned long long start = 0;
0dcecb2e
DW		 6781 unsigned long long maxsize;
6782 unsigned long reserve;
6783
0dcecb2e
DW		 6784 /* coalesce and sort all extents. also, check to see if we need to
6785 * reserve space between member arrays
6786 */
6787 j = 0;
6788 for (dl = super->disks; dl; dl = dl->next) {
6789 if (!dl->e)
6790 continue;
6791 for (i = 0; i < dl->extent_cnt; i++)
6792 e[j++] = dl->e[i];
6793 }
6794 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6795
6796 /* merge extents */
6797 i = 0;
6798 j = 0;
6799 while (i < sum_extents) {
6800 e[j].start = e[i].start;
6801 e[j].size = find_size(e, &i, sum_extents);
6802 j++;
6803 if (e[j-1].size == 0)
6804 break;
6805 }
6806
6807 pos = 0;
6808 maxsize = 0;
6809 start_extent = 0;
6810 i = 0;
6811 do {
6812 unsigned long long esize;
6813
6814 esize = e[i].start - pos;
6815 if (esize >= maxsize) {
6816 maxsize = esize;
6817 start = pos;
6818 start_extent = i;
6819 }
6820 pos = e[i].start + e[i].size;
6821 i++;
6822 } while (e[i-1].size);
6823 free(e);
6824
a7dd165b
DW		 6825 if (maxsize == 0)
6826 return 0;
6827
6828 /* FIXME assumes volume at offset 0 is the first volume in a
6829 * container
6830 */
0dcecb2e
DW		 6831 if (start_extent > 0)
6832 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6833 else
6834 reserve = 0;
6835
6836 if (maxsize < reserve)
a7dd165b 6837 return 0;
0dcecb2e 6838
5551b113 6839 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6840 if (start + reserve > super->create_offset)
a7dd165b 6841 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6842 super->create_offset = start + reserve;
6843
6844 return maxsize - reserve;
6845}
6846
88c32bb1
DW		 6847static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6848{
6849 if (level < 0 || level == 6 || level == 4)
6850 return 0;
6851
6852 /* if we have an orom prevent invalid raid levels */
6853 if (orom)
6854 switch (level) {
6855 case 0: return imsm_orom_has_raid0(orom);
6856 case 1:
6857 if (raiddisks > 2)
6858 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6859 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6860 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6861 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6862 }
6863 else
6864 return 1; /* not on an Intel RAID platform so anything goes */
6865
6866 return 0;
6867}
6868
ca9de185
LM		 6869static int
6870active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6871 int dpa, int verbose)
6872{
6873 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6874 struct mdstat_ent *memb;
ca9de185
LM		 6875 int count = 0;
6876 int num = 0;
594dc1b8 6877 struct md_list *dv;
ca9de185
LM		 6878 int found;
6879
6880 for (memb = mdstat ; memb ; memb = memb->next) {
6881 if (memb->metadata_version &&
fc54fe7a 6882 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6883 (strcmp(&memb->metadata_version[9], name) == 0) &&
6884 !is_subarray(memb->metadata_version+9) &&
6885 memb->members) {
6886 struct dev_member *dev = memb->members;
6887 int fd = -1;
6888 while(dev && (fd < 0)) {
503975b9
N		 6889 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6890 num = sprintf(path, "%s%s", "/dev/", dev->name);
6891 if (num > 0)
6892 fd = open(path, O_RDONLY, 0);
089f9d79 6893 if (num <= 0 || fd < 0) {
676e87a8 6894 pr_vrb("Cannot open %s: %s\n",
503975b9 6895 dev->name, strerror(errno));
ca9de185 6896 }
503975b9 6897 free(path);
ca9de185
LM		 6898 dev = dev->next;
6899 }
6900 found = 0;
089f9d79 6901 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6902 struct mdstat_ent *vol;
6903 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6904 if (vol->active > 0 &&
ca9de185 6905 vol->metadata_version &&
9581efb1 6906 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6907 found++;
6908 count++;
6909 }
6910 }
6911 if (*devlist && (found < dpa)) {
503975b9 6912 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6913 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6914 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6915 dv->found = found;
6916 dv->used = 0;
6917 dv->next = *devlist;
6918 *devlist = dv;
ca9de185
LM		 6919 }
6920 }
6921 if (fd >= 0)
6922 close(fd);
6923 }
6924 }
6925 free_mdstat(mdstat);
6926 return count;
6927}
6928
6929#ifdef DEBUG_LOOP
6930static struct md_list*
6931get_loop_devices(void)
6932{
6933 int i;
6934 struct md_list *devlist = NULL;
594dc1b8 6935 struct md_list *dv;
ca9de185
LM		 6936
6937 for(i = 0; i < 12; i++) {
503975b9
N		 6938 dv = xcalloc(1, sizeof(*dv));
6939 dv->devname = xmalloc(40);
ca9de185
LM		 6940 sprintf(dv->devname, "/dev/loop%d", i);
6941 dv->next = devlist;
6942 devlist = dv;
6943 }
6944 return devlist;
6945}
6946#endif
6947
6948static struct md_list*
6949get_devices(const char *hba_path)
6950{
6951 struct md_list *devlist = NULL;
594dc1b8 6952 struct md_list *dv;
ca9de185
LM		 6953 struct dirent *ent;
6954 DIR *dir;
6955 int err = 0;
6956
6957#if DEBUG_LOOP
6958 devlist = get_loop_devices();
6959 return devlist;
6960#endif
6961 /* scroll through /sys/dev/block looking for devices attached to
6962 * this hba
6963 */
6964 dir = opendir("/sys/dev/block");
6965 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6966 int fd;
6967 char buf[1024];
6968 int major, minor;
6969 char *path = NULL;
6970 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6971 continue;
7c798f87 6972 path = devt_to_devpath(makedev(major, minor), 1, NULL);
ca9de185
LM		 6973 if (!path)
6974 continue;
6975 if (!path_attached_to_hba(path, hba_path)) {
6976 free(path);
6977 path = NULL;
6978 continue;
6979 }
6980 free(path);
6981 path = NULL;
6982 fd = dev_open(ent->d_name, O_RDONLY);
6983 if (fd >= 0) {
6984 fd2devname(fd, buf);
6985 close(fd);
6986 } else {
e7b84f9d 6987 pr_err("cannot open device: %s\n",
ca9de185
LM		 6988 ent->d_name);
6989 continue;
6990 }
6991
503975b9
N		 6992 dv = xcalloc(1, sizeof(*dv));
6993 dv->devname = xstrdup(buf);
ca9de185
LM		 6994 dv->next = devlist;
6995 devlist = dv;
6996 }
6997 if (err) {
6998 while(devlist) {
6999 dv = devlist;
7000 devlist = devlist->next;
7001 free(dv->devname);
7002 free(dv);
7003 }
7004 }
562aa102 7005 closedir(dir);
ca9de185
LM		 7006 return devlist;
7007}
7008
7009static int
7010count_volumes_list(struct md_list *devlist, char *homehost,
7011 int verbose, int *found)
7012{
7013 struct md_list *tmpdev;
7014 int count = 0;
594dc1b8 7015 struct supertype *st;
ca9de185
LM		 7016
7017 /* first walk the list of devices to find a consistent set
7018 * that match the criterea, if that is possible.
7019 * We flag the ones we like with 'used'.
7020 */
7021 *found = 0;
7022 st = match_metadata_desc_imsm("imsm");
7023 if (st == NULL) {
676e87a8 7024 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 7025 return 0;
7026 }
7027
7028 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
7029 char *devname = tmpdev->devname;
0a6bff09 7030 dev_t rdev;
ca9de185
LM		 7031 struct supertype *tst;
7032 int dfd;
7033 if (tmpdev->used > 1)
7034 continue;
7035 tst = dup_super(st);
7036 if (tst == NULL) {
676e87a8 7037 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 7038 goto err_1;
7039 }
7040 tmpdev->container = 0;
7041 dfd = dev_open(devname, O_RDONLY|O_EXCL);
7042 if (dfd < 0) {
1ade5cc1 7043 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 7044 devname, strerror(errno));
7045 tmpdev->used = 2;
0a6bff09 7046 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 7047 tmpdev->used = 2;
7048 } else if (must_be_container(dfd)) {
7049 struct supertype *cst;
7050 cst = super_by_fd(dfd, NULL);
7051 if (cst == NULL) {
1ade5cc1 7052 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 7053 devname);
7054 tmpdev->used = 2;
7055 } else if (tst->ss != st->ss) {
1ade5cc1 7056 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 7057 devname);
7058 tmpdev->used = 2;
7059 } else if (!tst->ss->load_container ||
7060 tst->ss->load_container(tst, dfd, NULL))
7061 tmpdev->used = 2;
7062 else {
7063 tmpdev->container = 1;
7064 }
7065 if (cst)
7066 cst->ss->free_super(cst);
7067 } else {
0a6bff09 7068 tmpdev->st_rdev = rdev;
ca9de185 7069 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 7070 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 7071 devname);
7072 tmpdev->used = 2;
7073 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 7074 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 7075 tst->ss->name, devname);
7076 tmpdev->used = 2;
7077 }
7078 }
7079 if (dfd >= 0)
7080 close(dfd);
7081 if (tmpdev->used == 2 || tmpdev->used == 4) {
7082 /* Ignore unrecognised devices during auto-assembly */
7083 goto loop;
7084 }
7085 else {
7086 struct mdinfo info;
7087 tst->ss->getinfo_super(tst, &info, NULL);
7088
7089 if (st->minor_version == -1)
7090 st->minor_version = tst->minor_version;
7091
7092 if (memcmp(info.uuid, uuid_zero,
7093 sizeof(int[4])) == 0) {
7094 /* this is a floating spare. It cannot define
7095 * an array unless there are no more arrays of
7096 * this type to be found. It can be included
7097 * in an array of this type though.
7098 */
7099 tmpdev->used = 3;
7100 goto loop;
7101 }
7102
7103 if (st->ss != tst->ss ||
7104 st->minor_version != tst->minor_version ||
c7b8547c 7105 st->ss->compare_super(st, tst, 1) != 0) {
ca9de185
LM		 7106 /* Some mismatch. If exactly one array matches this host,
7107 * we can resolve on that one.
7108 * Or, if we are auto assembling, we just ignore the second
7109 * for now.
7110 */
1ade5cc1 7111 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 7112 devname);
7113 goto loop;
7114 }
7115 tmpdev->used = 1;
7116 *found = 1;
7117 dprintf("found: devname: %s\n", devname);
7118 }
7119 loop:
7120 if (tst)
7121 tst->ss->free_super(tst);
7122 }
7123 if (*found != 0) {
7124 int err;
7125 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
7126 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
7127 for (iter = head; iter; iter = iter->next) {
7128 dprintf("content->text_version: %s vol\n",
7129 iter->text_version);
7130 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
7131 /* do not assemble arrays with unsupported
7132 configurations */
1ade5cc1 7133 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 7134 iter->text_version);
7135 } else
7136 count++;
7137 }
7138 sysfs_free(head);
7139
7140 } else {
1ade5cc1 7141 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7142 err, st->sb);
7143 }
7144 } else {
1ade5cc1 7145 dprintf("no more devices to examine\n");
ca9de185
LM		 7146 }
7147
7148 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7149 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7150 if (count) {
7151 if (count < tmpdev->found)
7152 count = 0;
7153 else
7154 count -= tmpdev->found;
7155 }
7156 }
7157 if (tmpdev->used == 1)
7158 tmpdev->used = 4;
7159 }
7160 err_1:
7161 if (st)
7162 st->ss->free_super(st);
7163 return count;
7164}
7165
d3c11416
AO		 7166static int __count_volumes(char *hba_path, int dpa, int verbose,
7167 int cmp_hba_path)
ca9de185 7168{
72a45777 7169 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7170 int count = 0;
72a45777
PB		 7171 const struct orom_entry *entry;
7172 struct devid_list *dv, *devid_list;
ca9de185 7173
d3c11416 7174 if (!hba_path)
ca9de185
LM		 7175 return 0;
7176
72a45777 7177 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7178 if (strstr(idev->path, hba_path))
7179 break;
72a45777
PB		 7180 }
7181
7182 if (!idev || !idev->dev_id)
ca9de185 7183 return 0;
72a45777
PB		 7184
7185 entry = get_orom_entry_by_device_id(idev->dev_id);
7186
7187 if (!entry || !entry->devid_list)
7188 return 0;
7189
7190 devid_list = entry->devid_list;
7191 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7192 struct md_list *devlist;
d3c11416
AO		 7193 struct sys_dev *device = NULL;
7194 char *hpath;
72a45777
PB		 7195 int found = 0;
7196
d3c11416
AO		 7197 if (cmp_hba_path)
7198 device = device_by_id_and_path(dv->devid, hba_path);
7199 else
7200 device = device_by_id(dv->devid);
7201
72a45777 7202 if (device)
d3c11416 7203 hpath = device->path;
72a45777
PB		 7204 else
7205 return 0;
7206
d3c11416 7207 devlist = get_devices(hpath);
72a45777
PB		 7208 /* if no intel devices return zero volumes */
7209 if (devlist == NULL)
7210 return 0;
7211
d3c11416
AO		 7212 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7213 verbose);
7214 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7215 if (devlist == NULL)
7216 return 0;
7217 do {
7218 found = 0;
7219 count += count_volumes_list(devlist,
7220 NULL,
7221 verbose,
7222 &found);
7223 dprintf("found %d count: %d\n", found, count);
7224 } while (found);
7225
d3c11416 7226 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7227
7228 while (devlist) {
7229 struct md_list *dv = devlist;
7230 devlist = devlist->next;
7231 free(dv->devname);
7232 free(dv);
7233 }
ca9de185
LM		 7234 }
7235 return count;
7236}
7237
d3c11416
AO		 7238static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7239{
7240 if (!hba)
7241 return 0;
7242 if (hba->type == SYS_DEV_VMD) {
7243 struct sys_dev *dev;
7244 int count = 0;
7245
7246 for (dev = find_intel_devices(); dev; dev = dev->next) {
7247 if (dev->type == SYS_DEV_VMD)
7248 count += __count_volumes(dev->path, dpa,
7249 verbose, 1);
7250 }
7251 return count;
7252 }
7253 return __count_volumes(hba->path, dpa, verbose, 0);
7254}
7255
cd9d1ac7
DW		 7256static int imsm_default_chunk(const struct imsm_orom *orom)
7257{
7258 /* up to 512 if the plaform supports it, otherwise the platform max.
7259 * 128 if no platform detected
7260 */
7261 int fs = max(7, orom ? fls(orom->sss) : 0);
7262
7263 return min(512, (1 << fs));
7264}
73408129 7265
6592ce37
DW		 7266static int
7267validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7268 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7269{
660260d0
DW		 7270 /* check/set platform and metadata limits/defaults */
7271 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7272 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7273 super->orom->dpa);
73408129
LM		 7274 return 0;
7275 }
7276
5d500228 7277 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7278 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7279 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7280 level, raiddisks, raiddisks > 1 ? "s" : "");
7281 return 0;
7282 }
cd9d1ac7 7283
7ccc4cc4 7284 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7285 *chunk = imsm_default_chunk(super->orom);
7286
7ccc4cc4 7287 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7288 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7289 return 0;
6592ce37 7290 }
cd9d1ac7 7291
6592ce37
DW		 7292 if (layout != imsm_level_to_layout(level)) {
7293 if (level == 5)
676e87a8 7294 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7295 else if (level == 10)
676e87a8 7296 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7297 else
676e87a8 7298 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7299 layout, level);
7300 return 0;
7301 }
2cc699af 7302
7ccc4cc4 7303 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7304 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7305 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7306 return 0;
7307 }
614902f6 7308
6592ce37
DW		 7309 return 1;
7310}
7311
1011e834 7312/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7313 * FIX ME add ahci details
7314 */
8b353278 7315static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7316 int layout, int raiddisks, int *chunk,
af4348dd
N		 7317 unsigned long long size,
7318 unsigned long long data_offset,
7319 char *dev,
2c514b71
NB		 7320 unsigned long long *freesize,
7321 int verbose)
cdddbdbc 7322{
9e04ac1c 7323 dev_t rdev;
c2c087e6 7324 struct intel_super *super = st->sb;
b2916f25 7325 struct imsm_super *mpb;
c2c087e6
DW		 7326 struct dl *dl;
7327 unsigned long long pos = 0;
7328 unsigned long long maxsize;
7329 struct extent *e;
7330 int i;
cdddbdbc 7331
88c32bb1
DW		 7332 /* We must have the container info already read in. */
7333 if (!super)
c2c087e6
DW		 7334 return 0;
7335
b2916f25
JS		 7336 mpb = super->anchor;
7337
2cc699af 7338 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7339 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7340 return 0;
d54559f0 7341 }
c2c087e6
DW		 7342 if (!dev) {
7343 /* General test: make sure there is space for
2da8544a
DW		 7344 * 'raiddisks' device extents of size 'size' at a given
7345 * offset
c2c087e6 7346 */
e46273eb 7347 unsigned long long minsize = size;
b7528a20 7348 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7349 int dcnt = 0;
7350 if (minsize == 0)
7351 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7352 for (dl = super->disks; dl ; dl = dl->next) {
7353 int found = 0;
7354
bf5a934a 7355 pos = 0;
c2c087e6 7356 i = 0;
05501181 7357 e = get_extents(super, dl, 0);
c2c087e6
DW		 7358 if (!e) continue;
7359 do {
7360 unsigned long long esize;
7361 esize = e[i].start - pos;
7362 if (esize >= minsize)
7363 found = 1;
b7528a20 7364 if (found && start_offset == MaxSector) {
2da8544a
DW		 7365 start_offset = pos;
7366 break;
7367 } else if (found && pos != start_offset) {
7368 found = 0;
7369 break;
7370 }
c2c087e6
DW		 7371 pos = e[i].start + e[i].size;
7372 i++;
7373 } while (e[i-1].size);
7374 if (found)
7375 dcnt++;
7376 free(e);
7377 }
7378 if (dcnt < raiddisks) {
2c514b71 7379 if (verbose)
7a862a02 7380 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7381 dcnt, raiddisks);
c2c087e6
DW		 7382 return 0;
7383 }
7384 return 1;
7385 }
0dcecb2e 7386
c2c087e6 7387 /* This device must be a member of the set */
9e04ac1c 7388 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7389 return 0;
7390 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7391 if (dl->major == (int)major(rdev) &&
7392 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7393 break;
7394 }
7395 if (!dl) {
2c514b71 7396 if (verbose)
7a862a02 7397 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7398 return 0;
a20d2ba5
DW		 7399 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7400 /* If a volume is present then the current creation attempt
7401 * cannot incorporate new spares because the orom may not
7402 * understand this configuration (all member disks must be
7403 * members of each array in the container).
7404 */
7a862a02
N		 7405 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7406 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7407 return 0;
5fe62b94
WD		 7408 } else if (super->orom && mpb->num_raid_devs > 0 &&
7409 mpb->num_disks != raiddisks) {
7a862a02 7410 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7411 return 0;
c2c087e6 7412 }
0dcecb2e
DW		 7413
7414 /* retrieve the largest free space block */
05501181 7415 e = get_extents(super, dl, 0);
c2c087e6
DW		 7416 maxsize = 0;
7417 i = 0;
0dcecb2e
DW		 7418 if (e) {
7419 do {
7420 unsigned long long esize;
7421
7422 esize = e[i].start - pos;
7423 if (esize >= maxsize)
7424 maxsize = esize;
7425 pos = e[i].start + e[i].size;
7426 i++;
7427 } while (e[i-1].size);
7428 dl->e = e;
7429 dl->extent_cnt = i;
7430 } else {
7431 if (verbose)
e7b84f9d 7432 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7433 dev);
7434 return 0;
7435 }
7436 if (maxsize < size) {
7437 if (verbose)
e7b84f9d 7438 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7439 dev, maxsize, size);
7440 return 0;
7441 }
7442
7443 /* count total number of extents for merge */
7444 i = 0;
7445 for (dl = super->disks; dl; dl = dl->next)
7446 if (dl->e)
7447 i += dl->extent_cnt;
7448
7449 maxsize = merge_extents(super, i);
3baa56ab 7450
1a1ced1e
KS		 7451 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7452 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7453
a7dd165b 7454 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7455 if (verbose) {
7456 if (maxsize == 0)
7a862a02 7457 pr_err("no free space left on device. Aborting...\n");
b3071342 7458 else
7a862a02 7459 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7460 maxsize, size);
7461 }
0dcecb2e 7462 return 0;
0dcecb2e
DW		 7463 }
7464
c2c087e6
DW		 7465 *freesize = maxsize;
7466
ca9de185 7467 if (super->orom) {
72a45777 7468 int count = count_volumes(super->hba,
ca9de185
LM		 7469 super->orom->dpa, verbose);
7470 if (super->orom->vphba <= count) {
676e87a8 7471 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7472 super->orom->vphba);
7473 return 0;
7474 }
7475 }
c2c087e6 7476 return 1;
cdddbdbc
DW		 7477}
7478
13bcac90 7479static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7480 unsigned long long size, int chunk,
7481 unsigned long long *freesize)
7482{
7483 struct intel_super *super = st->sb;
7484 struct imsm_super *mpb = super->anchor;
7485 struct dl *dl;
7486 int i;
7487 int extent_cnt;
7488 struct extent *e;
7489 unsigned long long maxsize;
7490 unsigned long long minsize;
7491 int cnt;
7492 int used;
7493
7494 /* find the largest common start free region of the possible disks */
7495 used = 0;
7496 extent_cnt = 0;
7497 cnt = 0;
7498 for (dl = super->disks; dl; dl = dl->next) {
7499 dl->raiddisk = -1;
7500
7501 if (dl->index >= 0)
7502 used++;
7503
7504 /* don't activate new spares if we are orom constrained
7505 * and there is already a volume active in the container
7506 */
7507 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7508 continue;
7509
05501181 7510 e = get_extents(super, dl, 0);
efb30e7f
DW		 7511 if (!e)
7512 continue;
7513 for (i = 1; e[i-1].size; i++)
7514 ;
7515 dl->e = e;
7516 dl->extent_cnt = i;
7517 extent_cnt += i;
7518 cnt++;
7519 }
7520
7521 maxsize = merge_extents(super, extent_cnt);
7522 minsize = size;
7523 if (size == 0)
612e59d8
CA		 7524 /* chunk is in K */
7525 minsize = chunk * 2;
efb30e7f
DW		 7526
7527 if (cnt < raiddisks ||
7528 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7529 maxsize < minsize ||
7530 maxsize == 0) {
e7b84f9d 7531 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7532 return 0; /* No enough free spaces large enough */
7533 }
7534
7535 if (size == 0) {
7536 size = maxsize;
7537 if (chunk) {
612e59d8
CA		 7538 size /= 2 * chunk;
7539 size *= 2 * chunk;
efb30e7f 7540 }
f878b242
LM		 7541 maxsize = size;
7542 }
1a1ced1e
KS		 7543 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7544 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7545 cnt = 0;
7546 for (dl = super->disks; dl; dl = dl->next)
7547 if (dl->e)
7548 dl->raiddisk = cnt++;
7549
7550 *freesize = size;
7551
13bcac90
AK		 7552 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7553
efb30e7f
DW		 7554 return 1;
7555}
7556
13bcac90
AK		 7557static int reserve_space(struct supertype *st, int raiddisks,
7558 unsigned long long size, int chunk,
7559 unsigned long long *freesize)
7560{
7561 struct intel_super *super = st->sb;
7562 struct dl *dl;
7563 int cnt;
7564 int rv = 0;
7565
7566 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7567 if (rv) {
7568 cnt = 0;
7569 for (dl = super->disks; dl; dl = dl->next)
7570 if (dl->e)
7571 dl->raiddisk = cnt++;
7572 rv = 1;
7573 }
7574
7575 return rv;
7576}
7577
bf5a934a 7578static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7579 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7580 unsigned long long data_offset,
bf5a934a 7581 char *dev, unsigned long long *freesize,
5308f117 7582 int consistency_policy, int verbose)
bf5a934a
DW		 7583{
7584 int fd, cfd;
7585 struct mdinfo *sra;
20cbe8d2 7586 int is_member = 0;
bf5a934a 7587
d54559f0
LM		 7588 /* load capability
7589 * if given unused devices create a container
bf5a934a
DW		 7590 * if given given devices in a container create a member volume
7591 */
7592 if (level == LEVEL_CONTAINER) {
7593 /* Must be a fresh device to add to a container */
7594 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7595 raiddisks,
7ccc4cc4 7596 *chunk,
af4348dd 7597 size, data_offset,
bf5a934a
DW		 7598 dev, freesize,
7599 verbose);
7600 }
9587c373 7601
06a6101c
BK		 7602 /*
7603 * Size is given in sectors.
7604 */
7605 if (size && (size < 2048)) {
22dc741f 7606 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7607 /* Depends on algorithm in Create.c :
7608 * if container was given (dev == NULL) return -1,
7609 * if block device was given ( dev != NULL) return 0.
7610 */
7611 return dev ? -1 : 0;
7612 }
7613
8592f29d 7614 if (!dev) {
e91a3bad 7615 if (st->sb) {
ca9de185 7616 struct intel_super *super = st->sb;
e91a3bad 7617 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7618 raiddisks, chunk, size,
e91a3bad
LM		 7619 verbose))
7620 return 0;
efb30e7f
DW		 7621 /* we are being asked to automatically layout a
7622 * new volume based on the current contents of
7623 * the container. If the the parameters can be
7624 * satisfied reserve_space will record the disks,
7625 * start offset, and size of the volume to be
7626 * created. add_to_super and getinfo_super
7627 * detect when autolayout is in progress.
7628 */
ca9de185
LM		 7629 /* assuming that freesize is always given when array is
7630 created */
7631 if (super->orom && freesize) {
7632 int count;
72a45777 7633 count = count_volumes(super->hba,
ca9de185
LM		 7634 super->orom->dpa, verbose);
7635 if (super->orom->vphba <= count) {
676e87a8 7636 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7637 super->orom->vphba);
7638 return 0;
7639 }
7640 }
e91a3bad
LM		 7641 if (freesize)
7642 return reserve_space(st, raiddisks, size,
7ccc4cc4 7643 *chunk, freesize);
8592f29d
N		 7644 }
7645 return 1;
7646 }
bf5a934a
DW		 7647 if (st->sb) {
7648 /* creating in a given container */
7649 return validate_geometry_imsm_volume(st, level, layout,
7650 raiddisks, chunk, size,
af4348dd 7651 data_offset,
bf5a934a
DW		 7652 dev, freesize, verbose);
7653 }
7654
bf5a934a
DW		 7655 /* This device needs to be a device in an 'imsm' container */
7656 fd = open(dev, O_RDONLY|O_EXCL, 0);
7657 if (fd >= 0) {
7658 if (verbose)
e7b84f9d
N		 7659 pr_err("Cannot create this array on device %s\n",
7660 dev);
bf5a934a
DW		 7661 close(fd);
7662 return 0;
7663 }
7664 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7665 if (verbose)
e7b84f9d 7666 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7667 dev, strerror(errno));
7668 return 0;
7669 }
7670 /* Well, it is in use by someone, maybe an 'imsm' container. */
7671 cfd = open_container(fd);
20cbe8d2 7672 close(fd);
bf5a934a 7673 if (cfd < 0) {
bf5a934a 7674 if (verbose)
e7b84f9d 7675 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7676 dev);
7677 return 0;
7678 }
4dd2df09 7679 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7680 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7681 strcmp(sra->text_version, "imsm") == 0)
7682 is_member = 1;
7683 sysfs_free(sra);
7684 if (is_member) {
bf5a934a
DW		 7685 /* This is a member of a imsm container. Load the container
7686 * and try to create a volume
7687 */
7688 struct intel_super *super;
7689
ec50f7b6 7690 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7691 st->sb = super;
4dd2df09 7692 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7693 close(cfd);
7694 return validate_geometry_imsm_volume(st, level, layout,
7695 raiddisks, chunk,
af4348dd 7696 size, data_offset, dev,
ecbd9e81
N		 7697 freesize, 1)
7698 ? 1 : -1;
bf5a934a 7699 }
20cbe8d2 7700 }
bf5a934a 7701
20cbe8d2 7702 if (verbose)
e7b84f9d 7703 pr_err("failed container membership check\n");
20cbe8d2
AW		 7704
7705 close(cfd);
7706 return 0;
bf5a934a 7707}
0bd16cf2 7708
30f58b22 7709static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7710{
7711 struct intel_super *super = st->sb;
7712
30f58b22
DW		 7713 if (level && *level == UnSet)
7714 *level = LEVEL_CONTAINER;
7715
7716 if (level && layout && *layout == UnSet)
7717 *layout = imsm_level_to_layout(*level);
0bd16cf2 7718
cd9d1ac7
DW		 7719 if (chunk && (*chunk == UnSet || *chunk == 0))
7720 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7721}
7722
33414a01
DW		 7723static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7724
3364781b 7725static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7726{
3364781b 7727 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7728 __u8 i;
7729 struct intel_dev **dp;
7730 struct intel_super *super = st->sb;
3364781b 7731 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7732 struct imsm_super *mpb = super->anchor;
7733
3364781b 7734 if (mpb->num_raid_devs == 0)
33414a01 7735 return 2;
33414a01
DW		 7736
7737 /* block deletions that would change the uuid of active subarrays
7738 *
7739 * FIXME when immutable ids are available, but note that we'll
7740 * also need to fixup the invalidated/active subarray indexes in
7741 * mdstat
7742 */
7743 for (i = 0; i < mpb->num_raid_devs; i++) {
7744 char subarray[4];
7745
7746 if (i < current_vol)
7747 continue;
7748 sprintf(subarray, "%u", i);
4dd2df09 7749 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7750 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7751 current_vol, i);
33414a01
DW		 7752
7753 return 2;
7754 }
7755 }
7756
7757 if (st->update_tail) {
503975b9 7758 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7759
33414a01
DW		 7760 u->type = update_kill_array;
7761 u->dev_idx = current_vol;
7762 append_metadata_update(st, u, sizeof(*u));
7763
7764 return 0;
7765 }
7766
7767 for (dp = &super->devlist; *dp;)
7768 if ((*dp)->index == current_vol) {
7769 *dp = (*dp)->next;
7770 } else {
7771 handle_missing(super, (*dp)->dev);
7772 if ((*dp)->index > current_vol)
7773 (*dp)->index--;
7774 dp = &(*dp)->next;
7775 }
7776
7777 /* no more raid devices, all active components are now spares,
7778 * but of course failed are still failed
7779 */
7780 if (--mpb->num_raid_devs == 0) {
7781 struct dl *d;
7782
7783 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7784 if (d->index > -2)
7785 mark_spare(d);
33414a01
DW		 7786 }
7787
7788 super->updates_pending++;
7789
7790 return 0;
7791}
aa534678 7792
19ad203e
JR		 7793static int get_rwh_policy_from_update(char *update)
7794{
7795 if (strcmp(update, "ppl") == 0)
7796 return RWH_MULTIPLE_DISTRIBUTED;
7797 else if (strcmp(update, "no-ppl") == 0)
7798 return RWH_MULTIPLE_OFF;
7799 else if (strcmp(update, "bitmap") == 0)
7800 return RWH_BITMAP;
7801 else if (strcmp(update, "no-bitmap") == 0)
7802 return RWH_OFF;
7803 return -1;
7804}
7805
a951a4f7 7806static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7807 char *update, struct mddev_ident *ident)
aa534678
DW		 7808{
7809 /* update the subarray currently referenced by ->current_vol */
7810 struct intel_super *super = st->sb;
7811 struct imsm_super *mpb = super->anchor;
7812
aa534678
DW		 7813 if (strcmp(update, "name") == 0) {
7814 char *name = ident->name;
a951a4f7
N		 7815 char *ep;
7816 int vol;
aa534678 7817
4dd2df09 7818 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7819 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7820 return 2;
7821 }
7822
7823 if (!check_name(super, name, 0))
7824 return 2;
7825
a951a4f7
N		 7826 vol = strtoul(subarray, &ep, 10);
7827 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7828 return 2;
7829
aa534678 7830 if (st->update_tail) {
503975b9 7831 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7832
aa534678 7833 u->type = update_rename_array;
a951a4f7 7834 u->dev_idx = vol;
618f4e6d
XN		 7835 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7836 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7837 append_metadata_update(st, u, sizeof(*u));
7838 } else {
7839 struct imsm_dev *dev;
ebad3af2 7840 int i, namelen;
aa534678 7841
a951a4f7 7842 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7843 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7844 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7845 memcpy(dev->volume, name, namelen);
aa534678
DW		 7846 for (i = 0; i < mpb->num_raid_devs; i++) {
7847 dev = get_imsm_dev(super, i);
7848 handle_missing(super, dev);
7849 }
7850 super->updates_pending++;
7851 }
19ad203e 7852 } else if (get_rwh_policy_from_update(update) != -1) {
e6e9dd3f
AP		 7853 int new_policy;
7854 char *ep;
7855 int vol = strtoul(subarray, &ep, 10);
7856
7857 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7858 return 2;
7859
19ad203e 7860 new_policy = get_rwh_policy_from_update(update);
e6e9dd3f
AP		 7861
7862 if (st->update_tail) {
7863 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7864
7865 u->type = update_rwh_policy;
7866 u->dev_idx = vol;
7867 u->new_policy = new_policy;
7868 append_metadata_update(st, u, sizeof(*u));
7869 } else {
7870 struct imsm_dev *dev;
7871
7872 dev = get_imsm_dev(super, vol);
7873 dev->rwh_policy = new_policy;
7874 super->updates_pending++;
7875 }
19ad203e
JR		 7876 if (new_policy == RWH_BITMAP)
7877 return write_init_bitmap_imsm_vol(st, vol);
aa534678
DW		 7878 } else
7879 return 2;
7880
7881 return 0;
7882}
bf5a934a 7883
28bce06f
AK		 7884static int is_gen_migration(struct imsm_dev *dev)
7885{
7534230b
AK		 7886 if (dev == NULL)
7887 return 0;
7888
28bce06f
AK		 7889 if (!dev->vol.migr_state)
7890 return 0;
7891
7892 if (migr_type(dev) == MIGR_GEN_MIGR)
7893 return 1;
7894
7895 return 0;
7896}
7897
1e5c6983
DW		 7898static int is_rebuilding(struct imsm_dev *dev)
7899{
7900 struct imsm_map *migr_map;
7901
7902 if (!dev->vol.migr_state)
7903 return 0;
7904
7905 if (migr_type(dev) != MIGR_REBUILD)
7906 return 0;
7907
238c0a71 7908 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7909
7910 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7911 return 1;
7912 else
7913 return 0;
7914}
7915
6ce1fbf1
AK		 7916static int is_initializing(struct imsm_dev *dev)
7917{
7918 struct imsm_map *migr_map;
7919
7920 if (!dev->vol.migr_state)
7921 return 0;
7922
7923 if (migr_type(dev) != MIGR_INIT)
7924 return 0;
7925
238c0a71 7926 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7927
7928 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7929 return 1;
7930
7931 return 0;
6ce1fbf1
AK		 7932}
7933
c47b0ff6
AK		 7934static void update_recovery_start(struct intel_super *super,
7935 struct imsm_dev *dev,
7936 struct mdinfo *array)
1e5c6983
DW		 7937{
7938 struct mdinfo *rebuild = NULL;
7939 struct mdinfo *d;
7940 __u32 units;
7941
7942 if (!is_rebuilding(dev))
7943 return;
7944
7945 /* Find the rebuild target, but punt on the dual rebuild case */
7946 for (d = array->devs; d; d = d->next)
7947 if (d->recovery_start == 0) {
7948 if (rebuild)
7949 return;
7950 rebuild = d;
7951 }
7952
4363fd80
DW		 7953 if (!rebuild) {
7954 /* (?) none of the disks are marked with
7955 * IMSM_ORD_REBUILD, so assume they are missing and the
7956 * disk_ord_tbl was not correctly updated
7957 */
1ade5cc1 7958 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7959 return;
7960 }
7961
4036e7ee 7962 units = vol_curr_migr_unit(dev);
c47b0ff6 7963 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7964}
7965
276d77db 7966static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7967
00bbdbda 7968static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7969{
4f5bc454
DW		 7970 /* Given a container loaded by load_super_imsm_all,
7971 * extract information about all the arrays into
7972 * an mdinfo tree.
00bbdbda 7973 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7974 *
7975 * For each imsm_dev create an mdinfo, fill it in,
7976 * then look for matching devices in super->disks
7977 * and create appropriate device mdinfo.
7978 */
7979 struct intel_super *super = st->sb;
949c47a0 7980 struct imsm_super *mpb = super->anchor;
4f5bc454 7981 struct mdinfo *rest = NULL;
00bbdbda 7982 unsigned int i;
81219e70 7983 int sb_errors = 0;
abef11a3
AK		 7984 struct dl *d;
7985 int spare_disks = 0;
b6180160 7986 int current_vol = super->current_vol;
cdddbdbc 7987
19482bcc
AK		 7988 /* do not assemble arrays when not all attributes are supported */
7989 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7990 sb_errors = 1;
7a862a02 7991 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7992 }
7993
abef11a3
AK		 7994 /* count spare devices, not used in maps
7995 */
7996 for (d = super->disks; d; d = d->next)
7997 if (d->index == -1)
7998 spare_disks++;
7999
4f5bc454 8000 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 8001 struct imsm_dev *dev;
8002 struct imsm_map *map;
86e3692b 8003 struct imsm_map *map2;
4f5bc454 8004 struct mdinfo *this;
a6482415 8005 int slot;
a6482415 8006 int chunk;
00bbdbda 8007 char *ep;
8b9cd157 8008 int level;
00bbdbda
N		 8009
8010 if (subarray &&
8011 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
8012 continue;
8013
8014 dev = get_imsm_dev(super, i);
238c0a71
AK		 8015 map = get_imsm_map(dev, MAP_0);
8016 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 8017 level = get_imsm_raid_level(map);
4f5bc454 8018
1ce0101c
DW		 8019 /* do not publish arrays that are in the middle of an
8020 * unsupported migration
8021 */
8022 if (dev->vol.migr_state &&
28bce06f 8023 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 8024 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 8025 dev->volume);
8026 continue;
8027 }
2db86302
LM		 8028 /* do not publish arrays that are not support by controller's
8029 * OROM/EFI
8030 */
1ce0101c 8031
503975b9 8032 this = xmalloc(sizeof(*this));
4f5bc454 8033
301406c9 8034 super->current_vol = i;
a5d85af7 8035 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 8036 this->next = rest;
a6482415 8037 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 8038 /* mdadm does not support all metadata features- set the bit in all arrays state */
8039 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 8040 level, /* RAID level */
8041 imsm_level_to_layout(level),
81219e70 8042 map->num_members, /* raid disks */
fcc2c9da 8043 &chunk, imsm_dev_size(dev),
81219e70 8044 1 /* verbose */)) {
7a862a02 8045 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 8046 dev->volume);
8047 this->array.state |=
8048 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
8049 (1<<MD_SB_BLOCK_VOLUME);
8050 }
81219e70
LM		 8051
8052 /* if array has bad blocks, set suitable bit in all arrays state */
8053 if (sb_errors)
8054 this->array.state |=
8055 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
8056 (1<<MD_SB_BLOCK_VOLUME);
8057
4f5bc454 8058 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 8059 unsigned long long recovery_start;
4f5bc454
DW		 8060 struct mdinfo *info_d;
8061 struct dl *d;
8062 int idx;
9a1608e5 8063 int skip;
7eef0453 8064 __u32 ord;
8b9cd157 8065 int missing = 0;
4f5bc454 8066
9a1608e5 8067 skip = 0;
238c0a71
AK		 8068 idx = get_imsm_disk_idx(dev, slot, MAP_0);
8069 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 8070 for (d = super->disks; d ; d = d->next)
8071 if (d->index == idx)
0fbd635c 8072 break;
4f5bc454 8073
1e5c6983 8074 recovery_start = MaxSector;
4f5bc454 8075 if (d == NULL)
9a1608e5 8076 skip = 1;
25ed7e59 8077 if (d && is_failed(&d->disk))
9a1608e5 8078 skip = 1;
8b9cd157 8079 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 8080 recovery_start = 0;
1e93d0d1
BK		 8081 if (!(ord & IMSM_ORD_REBUILD))
8082 this->array.working_disks++;
1011e834 8083 /*
9a1608e5 8084 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 8085 * reset resync start to avoid a dirty-degraded
8086 * situation when performing the intial sync
9a1608e5 8087 */
8b9cd157
MK		 8088 if (skip)
8089 missing++;
8090
8091 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
8092 if ((!able_to_resync(level, missing) ||
8093 recovery_start == 0))
8094 this->resync_start = MaxSector;
8095 } else {
8096 /*
8097 * FIXME handle dirty degraded
8098 */
8099 }
8100
9a1608e5
DW		 8101 if (skip)
8102 continue;
4f5bc454 8103
503975b9 8104 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 8105 info_d->next = this->devs;
8106 this->devs = info_d;
8107
4f5bc454
DW		 8108 info_d->disk.number = d->index;
8109 info_d->disk.major = d->major;
8110 info_d->disk.minor = d->minor;
8111 info_d->disk.raid_disk = slot;
1e5c6983 8112 info_d->recovery_start = recovery_start;
86e3692b
AK		 8113 if (map2) {
8114 if (slot < map2->num_members)
8115 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 8116 else
8117 this->array.spare_disks++;
86e3692b
AK		 8118 } else {
8119 if (slot < map->num_members)
8120 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 8121 else
8122 this->array.spare_disks++;
86e3692b 8123 }
4f5bc454
DW		 8124
8125 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 8126 info_d->data_offset = pba_of_lba0(map);
44490938 8127 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 8128
8129 if (map->raid_level == 5) {
2432ce9b
AP		 8130 info_d->ppl_sector = this->ppl_sector;
8131 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 8132 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
8133 recovery_start == 0)
8134 this->resync_start = 0;
06fb291a 8135 }
b12796be 8136
5e46202e 8137 info_d->bb.supported = 1;
b12796be
TM		 8138 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
8139 info_d->data_offset,
8140 info_d->component_size,
8141 &info_d->bb);
4f5bc454 8142 }
1e5c6983 8143 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 8144 update_recovery_start(super, dev, this);
abef11a3 8145 this->array.spare_disks += spare_disks;
276d77db
AK		 8146
8147 /* check for reshape */
8148 if (this->reshape_active == 1)
8149 recover_backup_imsm(st, this);
9a1608e5 8150 rest = this;
4f5bc454
DW		 8151 }
8152
b6180160 8153 super->current_vol = current_vol;
4f5bc454 8154 return rest;
cdddbdbc
DW		 8155}
8156
3b451610
AK		 8157static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8158 int failed, int look_in_map)
c2a1e7da 8159{
3b451610
AK		 8160 struct imsm_map *map;
8161
8162 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8163
8164 if (!failed)
1011e834 8165 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8166 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8167
8168 switch (get_imsm_raid_level(map)) {
8169 case 0:
8170 return IMSM_T_STATE_FAILED;
8171 break;
8172 case 1:
8173 if (failed < map->num_members)
8174 return IMSM_T_STATE_DEGRADED;
8175 else
8176 return IMSM_T_STATE_FAILED;
8177 break;
8178 case 10:
8179 {
8180 /**
c92a2527
DW		 8181 * check to see if any mirrors have failed, otherwise we
8182 * are degraded. Even numbered slots are mirrored on
8183 * slot+1
c2a1e7da 8184 */
c2a1e7da 8185 int i;
d9b420a5
N		 8186 /* gcc -Os complains that this is unused */
8187 int insync = insync;
c2a1e7da
DW		 8188
8189 for (i = 0; i < map->num_members; i++) {
238c0a71 8190 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8191 int idx = ord_to_idx(ord);
8192 struct imsm_disk *disk;
c2a1e7da 8193
c92a2527 8194 /* reset the potential in-sync count on even-numbered
1011e834 8195 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8196 */
8197 if ((i & 1) == 0)
8198 insync = 2;
c2a1e7da 8199
c92a2527 8200 disk = get_imsm_disk(super, idx);
25ed7e59 8201 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8202 insync--;
c2a1e7da 8203
c92a2527
DW		 8204 /* no in-sync disks left in this mirror the
8205 * array has failed
8206 */
8207 if (insync == 0)
8208 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8209 }
8210
8211 return IMSM_T_STATE_DEGRADED;
8212 }
8213 case 5:
8214 if (failed < 2)
8215 return IMSM_T_STATE_DEGRADED;
8216 else
8217 return IMSM_T_STATE_FAILED;
8218 break;
8219 default:
8220 break;
8221 }
8222
8223 return map->map_state;
8224}
8225
3b451610
AK		 8226static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8227 int look_in_map)
c2a1e7da
DW		 8228{
8229 int i;
8230 int failed = 0;
8231 struct imsm_disk *disk;
d5985138
AK		 8232 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8233 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8234 struct imsm_map *map_for_loop;
0556e1a2
DW		 8235 __u32 ord;
8236 int idx;
d5985138 8237 int idx_1;
c2a1e7da 8238
0556e1a2
DW		 8239 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8240 * disks that are being rebuilt. New failures are recorded to
8241 * map[0]. So we look through all the disks we started with and
8242 * see if any failures are still present, or if any new ones
8243 * have arrived
0556e1a2 8244 */
d5985138
AK		 8245 map_for_loop = map;
8246 if (prev && (map->num_members < prev->num_members))
8247 map_for_loop = prev;
68fe4598
LD		 8248
8249 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8250 idx_1 = -255;
238c0a71
AK		 8251 /* when MAP_X is passed both maps failures are counted
8252 */
d5985138 8253 if (prev &&
089f9d79
JS		 8254 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8255 i < prev->num_members) {
d5985138
AK		 8256 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8257 idx_1 = ord_to_idx(ord);
c2a1e7da 8258
d5985138
AK		 8259 disk = get_imsm_disk(super, idx_1);
8260 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8261 failed++;
8262 }
089f9d79
JS		 8263 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8264 i < map->num_members) {
d5985138
AK		 8265 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8266 idx = ord_to_idx(ord);
8267
8268 if (idx != idx_1) {
8269 disk = get_imsm_disk(super, idx);
8270 if (!disk || is_failed(disk) ||
8271 ord & IMSM_ORD_REBUILD)
8272 failed++;
8273 }
8274 }
c2a1e7da
DW		 8275 }
8276
8277 return failed;
845dea95
NB		 8278}
8279
97b4d0e9
DW		 8280static int imsm_open_new(struct supertype *c, struct active_array *a,
8281 char *inst)
8282{
8283 struct intel_super *super = c->sb;
8284 struct imsm_super *mpb = super->anchor;
bbab0940 8285 struct imsm_update_prealloc_bb_mem u;
9587c373 8286
97b4d0e9 8287 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8288 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8289 return -ENODEV;
8290 }
8291
8292 dprintf("imsm: open_new %s\n", inst);
8293 a->info.container_member = atoi(inst);
bbab0940
TM		 8294
8295 u.type = update_prealloc_badblocks_mem;
8296 imsm_update_metadata_locally(c, &u, sizeof(u));
8297
97b4d0e9
DW		 8298 return 0;
8299}
8300
0c046afd
DW		 8301static int is_resyncing(struct imsm_dev *dev)
8302{
8303 struct imsm_map *migr_map;
8304
8305 if (!dev->vol.migr_state)
8306 return 0;
8307
1484e727
DW		 8308 if (migr_type(dev) == MIGR_INIT ||
8309 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8310 return 1;
8311
4c9bc37b
AK		 8312 if (migr_type(dev) == MIGR_GEN_MIGR)
8313 return 0;
8314
238c0a71 8315 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8316
089f9d79
JS		 8317 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8318 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8319 return 1;
8320 else
8321 return 0;
8322}
8323
0556e1a2 8324/* return true if we recorded new information */
4c9e8c1e
TM		 8325static int mark_failure(struct intel_super *super,
8326 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8327{
0556e1a2
DW		 8328 __u32 ord;
8329 int slot;
8330 struct imsm_map *map;
86c54047
DW		 8331 char buf[MAX_RAID_SERIAL_LEN+3];
8332 unsigned int len, shift = 0;
0556e1a2
DW		 8333
8334 /* new failures are always set in map[0] */
238c0a71 8335 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8336
8337 slot = get_imsm_disk_slot(map, idx);
8338 if (slot < 0)
8339 return 0;
8340
8341 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8342 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8343 return 0;
8344
7d0c5e24
LD		 8345 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8346 buf[MAX_RAID_SERIAL_LEN] = '\000';
8347 strcat(buf, ":0");
86c54047
DW		 8348 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8349 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8350 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8351
f2f27e63 8352 disk->status |= FAILED_DISK;
0556e1a2 8353 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8354 /* mark failures in second map if second map exists and this disk
8355 * in this slot.
8356 * This is valid for migration, initialization and rebuild
8357 */
8358 if (dev->vol.migr_state) {
238c0a71 8359 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8360 int slot2 = get_imsm_disk_slot(map2, idx);
8361
089f9d79 8362 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8363 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8364 idx | IMSM_ORD_REBUILD);
8365 }
d7a1fda2
MT		 8366 if (map->failed_disk_num == 0xff ||
8367 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8368 map->failed_disk_num = slot;
4c9e8c1e
TM		 8369
8370 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8371
0556e1a2
DW		 8372 return 1;
8373}
8374
4c9e8c1e
TM		 8375static void mark_missing(struct intel_super *super,
8376 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8377{
4c9e8c1e 8378 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8379
8380 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8381 return;
8382
47ee5a45
DW		 8383 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8384 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8385}
8386
33414a01
DW		 8387static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8388{
33414a01 8389 struct dl *dl;
33414a01
DW		 8390
8391 if (!super->missing)
8392 return;
33414a01 8393
79b68f1b
PC		 8394 /* When orom adds replacement for missing disk it does
8395 * not remove entry of missing disk, but just updates map with
8396 * new added disk. So it is not enough just to test if there is
8397 * any missing disk, we have to look if there are any failed disks
8398 * in map to stop migration */
8399
33414a01 8400 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8401 /* end process for initialization and rebuild only
8402 */
8403 if (is_gen_migration(dev) == 0) {
fb12a745 8404 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8405
fb12a745
TM		 8406 if (failed) {
8407 __u8 map_state;
8408 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8409 struct imsm_map *map1;
8410 int i, ord, ord_map1;
8411 int rebuilt = 1;
3d59f0c0 8412
fb12a745
TM		 8413 for (i = 0; i < map->num_members; i++) {
8414 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8415 if (!(ord & IMSM_ORD_REBUILD))
8416 continue;
8417
8418 map1 = get_imsm_map(dev, MAP_1);
8419 if (!map1)
8420 continue;
8421
8422 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8423 if (ord_map1 & IMSM_ORD_REBUILD)
8424 rebuilt = 0;
8425 }
8426
8427 if (rebuilt) {
8428 map_state = imsm_check_degraded(super, dev,
8429 failed, MAP_0);
8430 end_migration(dev, super, map_state);
8431 }
8432 }
3d59f0c0 8433 }
33414a01 8434 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8435 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8436 super->updates_pending++;
8437}
8438
f3871fdc
AK		 8439static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8440 long long new_size)
70bdf0dc 8441{
70bdf0dc 8442 unsigned long long array_blocks;
9529d343
MD		 8443 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8444 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8445
8446 if (used_disks == 0) {
8447 /* when problems occures
8448 * return current array_blocks value
8449 */
fcc2c9da 8450 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8451
8452 return array_blocks;
8453 }
8454
8455 /* set array size in metadata
8456 */
9529d343 8457 if (new_size <= 0)
f3871fdc
AK		 8458 /* OLCE size change is caused by added disks
8459 */
44490938 8460 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8461 else
f3871fdc
AK		 8462 /* Online Volume Size Change
8463 * Using available free space
8464 */
8465 array_blocks = new_size;
70bdf0dc 8466
b53bfba6 8467 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8468 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8469
8470 return array_blocks;
8471}
8472
28bce06f
AK		 8473static void imsm_set_disk(struct active_array *a, int n, int state);
8474
0e2d1a4e
AK		 8475static void imsm_progress_container_reshape(struct intel_super *super)
8476{
8477 /* if no device has a migr_state, but some device has a
8478 * different number of members than the previous device, start
8479 * changing the number of devices in this device to match
8480 * previous.
8481 */
8482 struct imsm_super *mpb = super->anchor;
8483 int prev_disks = -1;
8484 int i;
1dfaa380 8485 int copy_map_size;
0e2d1a4e
AK		 8486
8487 for (i = 0; i < mpb->num_raid_devs; i++) {
8488 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8489 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8490 struct imsm_map *map2;
8491 int prev_num_members;
0e2d1a4e
AK		 8492
8493 if (dev->vol.migr_state)
8494 return;
8495
8496 if (prev_disks == -1)
8497 prev_disks = map->num_members;
8498 if (prev_disks == map->num_members)
8499 continue;
8500
8501 /* OK, this array needs to enter reshape mode.
8502 * i.e it needs a migr_state
8503 */
8504
1dfaa380 8505 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8506 prev_num_members = map->num_members;
8507 map->num_members = prev_disks;
8508 dev->vol.migr_state = 1;
4036e7ee 8509 set_vol_curr_migr_unit(dev, 0);
ea672ee1 8510 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8511 for (i = prev_num_members;
8512 i < map->num_members; i++)
8513 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8514 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8515 /* Copy the current map */
1dfaa380 8516 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8517 map2->num_members = prev_num_members;
8518
f3871fdc 8519 imsm_set_array_size(dev, -1);
51d83f5d 8520 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8521 super->updates_pending++;
8522 }
8523}
8524
aad6f216 8525/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8526 * states are handled in imsm_set_disk() with one exception, when a
8527 * resync is stopped due to a new failure this routine will set the
8528 * 'degraded' state for the array.
8529 */
01f157d7 8530static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8531{
8532 int inst = a->info.container_member;
8533 struct intel_super *super = a->container->sb;
949c47a0 8534 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8535 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8536 int failed = imsm_count_failed(super, dev, MAP_0);
8537 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8538 __u32 blocks_per_unit;
a862209d 8539
1af97990
AK		 8540 if (dev->vol.migr_state &&
8541 dev->vol.migr_type == MIGR_GEN_MIGR) {
8542 /* array state change is blocked due to reshape action
aad6f216
N		 8543 * We might need to
8544 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8545 * - finish the reshape (if last_checkpoint is big and action != reshape)
4036e7ee 8546 * - update vol_curr_migr_unit
1af97990 8547 */
aad6f216 8548 if (a->curr_action == reshape) {
4036e7ee 8549 /* still reshaping, maybe update vol_curr_migr_unit */
633b5610 8550 goto mark_checkpoint;
aad6f216
N		 8551 } else {
8552 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8553 /* for some reason we aborted the reshape.
b66e591b
AK		 8554 *
8555 * disable automatic metadata rollback
8556 * user action is required to recover process
aad6f216 8557 */
b66e591b 8558 if (0) {
238c0a71
AK		 8559 struct imsm_map *map2 =
8560 get_imsm_map(dev, MAP_1);
8561 dev->vol.migr_state = 0;
8562 set_migr_type(dev, 0);
4036e7ee 8563 set_vol_curr_migr_unit(dev, 0);
238c0a71
AK		 8564 memcpy(map, map2,
8565 sizeof_imsm_map(map2));
8566 super->updates_pending++;
b66e591b 8567 }
aad6f216
N		 8568 }
8569 if (a->last_checkpoint >= a->info.component_size) {
8570 unsigned long long array_blocks;
8571 int used_disks;
e154ced3 8572 struct mdinfo *mdi;
aad6f216 8573
9529d343 8574 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8575 if (used_disks > 0) {
8576 array_blocks =
44490938 8577 per_dev_array_size(map) *
d55adef9 8578 used_disks;
b53bfba6
TM		 8579 array_blocks =
8580 round_size_to_mb(array_blocks,
8581 used_disks);
d55adef9
AK		 8582 a->info.custom_array_size = array_blocks;
8583 /* encourage manager to update array
8584 * size
8585 */
e154ced3 8586
d55adef9 8587 a->check_reshape = 1;
633b5610 8588 }
e154ced3
AK		 8589 /* finalize online capacity expansion/reshape */
8590 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8591 imsm_set_disk(a,
8592 mdi->disk.raid_disk,
8593 mdi->curr_state);
8594
0e2d1a4e 8595 imsm_progress_container_reshape(super);
e154ced3 8596 }
aad6f216 8597 }
1af97990
AK		 8598 }
8599
47ee5a45 8600 /* before we activate this array handle any missing disks */
33414a01
DW		 8601 if (consistent == 2)
8602 handle_missing(super, dev);
1e5c6983 8603
0c046afd 8604 if (consistent == 2 &&
b7941fd6 8605 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8606 map_state != IMSM_T_STATE_NORMAL ||
8607 dev->vol.migr_state))
01f157d7 8608 consistent = 0;
272906ef 8609
b7941fd6 8610 if (is_resync_complete(&a->info)) {
0c046afd 8611 /* complete intialization / resync,
0556e1a2
DW		 8612 * recovery and interrupted recovery is completed in
8613 * ->set_disk
0c046afd
DW		 8614 */
8615 if (is_resyncing(dev)) {
8616 dprintf("imsm: mark resync done\n");
809da78e 8617 end_migration(dev, super, map_state);
115c3803 8618 super->updates_pending++;
484240d8 8619 a->last_checkpoint = 0;
115c3803 8620 }
b9172665
AK		 8621 } else if ((!is_resyncing(dev) && !failed) &&
8622 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8623 /* mark the start of the init process if nothing is failed */
b7941fd6 8624 dprintf("imsm: mark resync start\n");
1484e727 8625 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8626 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8627 else
8e59f3d8 8628 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8629 super->updates_pending++;
115c3803 8630 }
a862209d 8631
633b5610 8632mark_checkpoint:
5b83bacf
AK		 8633 /* skip checkpointing for general migration,
8634 * it is controlled in mdadm
8635 */
8636 if (is_gen_migration(dev))
8637 goto skip_mark_checkpoint;
8638
4036e7ee
MT		 8639 /* check if we can update vol_curr_migr_unit from resync_start,
8640 * recovery_start
8641 */
c47b0ff6 8642 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8643 if (blocks_per_unit) {
4036e7ee
MT		 8644 set_vol_curr_migr_unit(dev,
8645 a->last_checkpoint / blocks_per_unit);
8646 dprintf("imsm: mark checkpoint (%llu)\n",
8647 vol_curr_migr_unit(dev));
8648 super->updates_pending++;
1e5c6983 8649 }
f8f603f1 8650
5b83bacf 8651skip_mark_checkpoint:
3393c6af 8652 /* mark dirty / clean */
2432ce9b
AP		 8653 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8654 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8655 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8656 if (consistent) {
8657 dev->vol.dirty = RAIDVOL_CLEAN;
8658 } else {
8659 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8660 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8661 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8662 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8663 }
a862209d
DW		 8664 super->updates_pending++;
8665 }
28bce06f 8666
01f157d7 8667 return consistent;
a862209d
DW		 8668}
8669
6f50473f
TM		 8670static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8671{
8672 int inst = a->info.container_member;
8673 struct intel_super *super = a->container->sb;
8674 struct imsm_dev *dev = get_imsm_dev(super, inst);
8675 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8676
8677 if (slot > map->num_members) {
8678 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8679 slot, map->num_members - 1);
8680 return -1;
8681 }
8682
8683 if (slot < 0)
8684 return -1;
8685
8686 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8687}
8688
8d45d196 8689static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8690{
8d45d196
DW		 8691 int inst = a->info.container_member;
8692 struct intel_super *super = a->container->sb;
949c47a0 8693 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8694 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8695 struct imsm_disk *disk;
7ce05701
LD		 8696 struct mdinfo *mdi;
8697 int recovery_not_finished = 0;
0c046afd 8698 int failed;
6f50473f 8699 int ord;
0c046afd 8700 __u8 map_state;
fb12a745
TM		 8701 int rebuild_done = 0;
8702 int i;
8d45d196 8703
fb12a745 8704 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8705 if (ord < 0)
8d45d196
DW		 8706 return;
8707
4e6e574a 8708 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8709 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8710
5802a811 8711 /* check for new failures */
ae7d61e3 8712 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8713 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8714 super->updates_pending++;
8d45d196 8715 }
47ee5a45 8716
19859edc 8717 /* check if in_sync */
0556e1a2 8718 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8719 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8720
8721 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8722 rebuild_done = 1;
19859edc
DW		 8723 super->updates_pending++;
8724 }
8d45d196 8725
3b451610
AK		 8726 failed = imsm_count_failed(super, dev, MAP_0);
8727 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8728
0c046afd 8729 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8730 dprintf("imsm: Detected transition to state ");
8731 switch (map_state) {
8732 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8733 dprintf("normal: ");
8734 if (is_rebuilding(dev)) {
1ade5cc1 8735 dprintf_cont("while rebuilding");
7ce05701
LD		 8736 /* check if recovery is really finished */
8737 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8738 if (mdi->recovery_start != MaxSector) {
8739 recovery_not_finished = 1;
8740 break;
8741 }
8742 if (recovery_not_finished) {
1ade5cc1
N		 8743 dprintf_cont("\n");
8744 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8745 if (a->last_checkpoint < mdi->recovery_start) {
8746 a->last_checkpoint = mdi->recovery_start;
8747 super->updates_pending++;
8748 }
8749 break;
8750 }
94002678 8751 end_migration(dev, super, map_state);
94002678
AK		 8752 map->failed_disk_num = ~0;
8753 super->updates_pending++;
8754 a->last_checkpoint = 0;
8755 break;
8756 }
8757 if (is_gen_migration(dev)) {
1ade5cc1 8758 dprintf_cont("while general migration");
bf2f0071 8759 if (a->last_checkpoint >= a->info.component_size)
809da78e 8760 end_migration(dev, super, map_state);
94002678
AK		 8761 else
8762 map->map_state = map_state;
28bce06f 8763 map->failed_disk_num = ~0;
94002678 8764 super->updates_pending++;
bf2f0071 8765 break;
94002678
AK		 8766 }
8767 break;
8768 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8769 dprintf_cont("degraded: ");
089f9d79 8770 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8771 dprintf_cont("mark degraded");
94002678
AK		 8772 map->map_state = map_state;
8773 super->updates_pending++;
8774 a->last_checkpoint = 0;
8775 break;
8776 }
8777 if (is_rebuilding(dev)) {
d7a1fda2 8778 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8779 if (state & DS_FAULTY) {
8780 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8781 if (n == map->failed_disk_num) {
8782 dprintf_cont("end migration");
8783 end_migration(dev, super, map_state);
a4e96fd8 8784 a->last_checkpoint = 0;
d7a1fda2 8785 } else {
a4e96fd8 8786 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8787 map->map_state = map_state;
8788 }
94002678 8789 super->updates_pending++;
fb12a745
TM		 8790 }
8791
a4e96fd8
MT		 8792 if (!rebuild_done)
8793 break;
8794
fb12a745
TM		 8795 /* check if recovery is really finished */
8796 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8797 if (mdi->recovery_start != MaxSector) {
8798 recovery_not_finished = 1;
8799 break;
8800 }
8801 if (recovery_not_finished) {
8802 dprintf_cont("\n");
a4e96fd8 8803 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8804 if (a->last_checkpoint < mdi->recovery_start) {
8805 a->last_checkpoint =
8806 mdi->recovery_start;
8807 super->updates_pending++;
8808 }
8809 break;
94002678 8810 }
fb12a745
TM		 8811
8812 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8813 end_migration(dev, super, map_state);
8814 a->last_checkpoint = 0;
8815 super->updates_pending++;
fb12a745
TM		 8816
8817 for (i = 0; i < map->num_members; i++) {
8818 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8819
8820 if (idx & IMSM_ORD_REBUILD)
8821 map->failed_disk_num = i;
8822 }
8823 super->updates_pending++;
94002678
AK		 8824 break;
8825 }
8826 if (is_gen_migration(dev)) {
1ade5cc1 8827 dprintf_cont("while general migration");
bf2f0071 8828 if (a->last_checkpoint >= a->info.component_size)
809da78e 8829 end_migration(dev, super, map_state);
94002678
AK		 8830 else {
8831 map->map_state = map_state;
3b451610 8832 manage_second_map(super, dev);
94002678
AK		 8833 }
8834 super->updates_pending++;
bf2f0071 8835 break;
28bce06f 8836 }
6ce1fbf1 8837 if (is_initializing(dev)) {
1ade5cc1 8838 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8839 map->map_state = map_state;
8840 super->updates_pending++;
8841 break;
8842 }
94002678
AK		 8843 break;
8844 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8845 dprintf_cont("failed: ");
94002678 8846 if (is_gen_migration(dev)) {
1ade5cc1 8847 dprintf_cont("while general migration");
94002678
AK		 8848 map->map_state = map_state;
8849 super->updates_pending++;
8850 break;
8851 }
8852 if (map->map_state != map_state) {
1ade5cc1 8853 dprintf_cont("mark failed");
94002678
AK		 8854 end_migration(dev, super, map_state);
8855 super->updates_pending++;
8856 a->last_checkpoint = 0;
8857 break;
8858 }
8859 break;
8860 default:
1ade5cc1 8861 dprintf_cont("state %i\n", map_state);
5802a811 8862 }
1ade5cc1 8863 dprintf_cont("\n");
845dea95
NB		 8864}
8865
f796af5d 8866static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8867{
f796af5d 8868 void *buf = mpb;
c2a1e7da
DW		 8869 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8870 unsigned long long dsize;
8871 unsigned long long sectors;
f36a9ecd 8872 unsigned int sector_size;
c2a1e7da 8873
aec01630
JS		 8874 if (!get_dev_sector_size(fd, NULL, &sector_size))
8875 return 1;
c2a1e7da
DW		 8876 get_dev_size(fd, NULL, &dsize);
8877
f36a9ecd 8878 if (mpb_size > sector_size) {
272f648f 8879 /* -1 to account for anchor */
f36a9ecd 8880 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8881
272f648f 8882 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8883 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8884 SEEK_SET) < 0)
272f648f 8885 return 1;
c2a1e7da 8886
f36a9ecd
PB		 8887 if ((unsigned long long)write(fd, buf + sector_size,
8888 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8889 return 1;
8890 }
c2a1e7da 8891
272f648f 8892 /* first block is stored on second to last sector of the disk */
f36a9ecd 8893 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8894 return 1;
8895
466070ad 8896 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8897 return 1;
8898
c2a1e7da
DW		 8899 return 0;
8900}
8901
2e735d19 8902static void imsm_sync_metadata(struct supertype *container)
845dea95 8903{
2e735d19 8904 struct intel_super *super = container->sb;
c2a1e7da 8905
1a64be56 8906 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8907 if (!super->updates_pending)
8908 return;
8909
36988a3d 8910 write_super_imsm(container, 0);
c2a1e7da
DW		 8911
8912 super->updates_pending = 0;
845dea95
NB		 8913}
8914
272906ef
DW		 8915static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8916{
8917 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8918 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8919 struct dl *dl;
8920
8921 for (dl = super->disks; dl; dl = dl->next)
8922 if (dl->index == i)
8923 break;
8924
25ed7e59 8925 if (dl && is_failed(&dl->disk))
272906ef
DW		 8926 dl = NULL;
8927
8928 if (dl)
1ade5cc1 8929 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8930
8931 return dl;
8932}
8933
a20d2ba5 8934static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8935 struct active_array *a, int activate_new,
8936 struct mdinfo *additional_test_list)
272906ef
DW		 8937{
8938 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8939 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8940 struct imsm_super *mpb = super->anchor;
8941 struct imsm_map *map;
272906ef
DW		 8942 unsigned long long pos;
8943 struct mdinfo *d;
8944 struct extent *ex;
a20d2ba5 8945 int i, j;
272906ef 8946 int found;
569cc43f
DW		 8947 __u32 array_start = 0;
8948 __u32 array_end = 0;
272906ef 8949 struct dl *dl;
6c932028 8950 struct mdinfo *test_list;
272906ef
DW		 8951
8952 for (dl = super->disks; dl; dl = dl->next) {
8953 /* If in this array, skip */
8954 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8955 if (d->state_fd >= 0 &&
8956 d->disk.major == dl->major &&
272906ef 8957 d->disk.minor == dl->minor) {
8ba77d32
AK		 8958 dprintf("%x:%x already in array\n",
8959 dl->major, dl->minor);
272906ef
DW		 8960 break;
8961 }
8962 if (d)
8963 continue;
6c932028
AK		 8964 test_list = additional_test_list;
8965 while (test_list) {
8966 if (test_list->disk.major == dl->major &&
8967 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8968 dprintf("%x:%x already in additional test list\n",
8969 dl->major, dl->minor);
8970 break;
8971 }
6c932028 8972 test_list = test_list->next;
8ba77d32 8973 }
6c932028 8974 if (test_list)
8ba77d32 8975 continue;
272906ef 8976
e553d2a4 8977 /* skip in use or failed drives */
25ed7e59 8978 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8979 dl->index == -2) {
8980 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8981 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8982 continue;
8983 }
8984
a20d2ba5
DW		 8985 /* skip pure spares when we are looking for partially
8986 * assimilated drives
8987 */
8988 if (dl->index == -1 && !activate_new)
8989 continue;
8990
f2cc4f7d
AO		 8991 if (!drive_validate_sector_size(super, dl))
8992 continue;
8993
272906ef 8994 /* Does this unused device have the requisite free space?
a20d2ba5 8995 * It needs to be able to cover all member volumes
272906ef 8996 */
05501181 8997 ex = get_extents(super, dl, 1);
272906ef
DW		 8998 if (!ex) {
8999 dprintf("cannot get extents\n");
9000 continue;
9001 }
a20d2ba5
DW		 9002 for (i = 0; i < mpb->num_raid_devs; i++) {
9003 dev = get_imsm_dev(super, i);
238c0a71 9004 map = get_imsm_map(dev, MAP_0);
272906ef 9005
a20d2ba5
DW		 9006 /* check if this disk is already a member of
9007 * this array
272906ef 9008 */
620b1713 9009 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 9010 continue;
9011
9012 found = 0;
9013 j = 0;
9014 pos = 0;
5551b113 9015 array_start = pba_of_lba0(map);
329c8278 9016 array_end = array_start +
44490938 9017 per_dev_array_size(map) - 1;
a20d2ba5
DW		 9018
9019 do {
9020 /* check that we can start at pba_of_lba0 with
44490938 9021 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 9022 */
329c8278 9023 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 9024 found = 1;
9025 break;
9026 }
9027 pos = ex[j].start + ex[j].size;
9028 j++;
9029 } while (ex[j-1].size);
9030
9031 if (!found)
272906ef 9032 break;
a20d2ba5 9033 }
272906ef
DW		 9034
9035 free(ex);
a20d2ba5 9036 if (i < mpb->num_raid_devs) {
329c8278
DW		 9037 dprintf("%x:%x does not have %u to %u available\n",
9038 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 9039 /* No room */
9040 continue;
a20d2ba5
DW		 9041 }
9042 return dl;
272906ef
DW		 9043 }
9044
9045 return dl;
9046}
9047
95d07a2c
LM		 9048static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
9049{
9050 struct imsm_dev *dev2;
9051 struct imsm_map *map;
9052 struct dl *idisk;
9053 int slot;
9054 int idx;
9055 __u8 state;
9056
9057 dev2 = get_imsm_dev(cont->sb, dev_idx);
9058 if (dev2) {
238c0a71 9059 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 9060 if (state == IMSM_T_STATE_FAILED) {
238c0a71 9061 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 9062 if (!map)
9063 return 1;
9064 for (slot = 0; slot < map->num_members; slot++) {
9065 /*
9066 * Check if failed disks are deleted from intel
9067 * disk list or are marked to be deleted
9068 */
238c0a71 9069 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 9070 idisk = get_imsm_dl_disk(cont->sb, idx);
9071 /*
9072 * Do not rebuild the array if failed disks
9073 * from failed sub-array are not removed from
9074 * container.
9075 */
9076 if (idisk &&
9077 is_failed(&idisk->disk) &&
9078 (idisk->action != DISK_REMOVE))
9079 return 0;
9080 }
9081 }
9082 }
9083 return 1;
9084}
9085
88758e9d
DW		 9086static struct mdinfo *imsm_activate_spare(struct active_array *a,
9087 struct metadata_update **updates)
9088{
9089 /**
d23fe947
DW		 9090 * Find a device with unused free space and use it to replace a
9091 * failed/vacant region in an array. We replace failed regions one a
9092 * array at a time. The result is that a new spare disk will be added
9093 * to the first failed array and after the monitor has finished
9094 * propagating failures the remainder will be consumed.
88758e9d 9095 *
d23fe947
DW		 9096 * FIXME add a capability for mdmon to request spares from another
9097 * container.
88758e9d
DW		 9098 */
9099
9100 struct intel_super *super = a->container->sb;
88758e9d 9101 int inst = a->info.container_member;
949c47a0 9102 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 9103 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 9104 int failed = a->info.array.raid_disks;
9105 struct mdinfo *rv = NULL;
9106 struct mdinfo *d;
9107 struct mdinfo *di;
9108 struct metadata_update *mu;
9109 struct dl *dl;
9110 struct imsm_update_activate_spare *u;
9111 int num_spares = 0;
9112 int i;
95d07a2c 9113 int allowed;
88758e9d
DW		 9114
9115 for (d = a->info.devs ; d ; d = d->next) {
9116 if ((d->curr_state & DS_FAULTY) &&
9117 d->state_fd >= 0)
9118 /* wait for Removal to happen */
9119 return NULL;
9120 if (d->state_fd >= 0)
9121 failed--;
9122 }
9123
9124 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
9125 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 9126
e2962bfc
AK		 9127 if (imsm_reshape_blocks_arrays_changes(super))
9128 return NULL;
1af97990 9129
fc8ca064
AK		 9130 /* Cannot activate another spare if rebuild is in progress already
9131 */
9132 if (is_rebuilding(dev)) {
7a862a02 9133 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 9134 return NULL;
9135 }
9136
89c67882
AK		 9137 if (a->info.array.level == 4)
9138 /* No repair for takeovered array
9139 * imsm doesn't support raid4
9140 */
9141 return NULL;
9142
3b451610
AK		 9143 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9144 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9145 return NULL;
9146
83ca7d45
AP		 9147 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9148 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9149 return NULL;
9150 }
9151
95d07a2c
LM		 9152 /*
9153 * If there are any failed disks check state of the other volume.
9154 * Block rebuild if the another one is failed until failed disks
9155 * are removed from container.
9156 */
9157 if (failed) {
7a862a02 9158 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9159 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9160 /* check if states of the other volumes allow for rebuild */
9161 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9162 if (i != inst) {
9163 allowed = imsm_rebuild_allowed(a->container,
9164 i, failed);
9165 if (!allowed)
9166 return NULL;
9167 }
9168 }
9169 }
9170
88758e9d 9171 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9172 for (i = 0; i < a->info.array.raid_disks; i++) {
9173 for (d = a->info.devs ; d ; d = d->next)
9174 if (d->disk.raid_disk == i)
9175 break;
9176 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9177 if (d && (d->state_fd >= 0))
9178 continue;
9179
272906ef 9180 /*
a20d2ba5
DW		 9181 * OK, this device needs recovery. Try to re-add the
9182 * previous occupant of this slot, if this fails see if
9183 * we can continue the assimilation of a spare that was
9184 * partially assimilated, finally try to activate a new
9185 * spare.
272906ef
DW		 9186 */
9187 dl = imsm_readd(super, i, a);
9188 if (!dl)
b303fe21 9189 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9190 if (!dl)
b303fe21 9191 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9192 if (!dl)
9193 continue;
1011e834 9194
272906ef 9195 /* found a usable disk with enough space */
503975b9 9196 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9197
9198 /* dl->index will be -1 in the case we are activating a
9199 * pristine spare. imsm_process_update() will create a
9200 * new index in this case. Once a disk is found to be
9201 * failed in all member arrays it is kicked from the
9202 * metadata
9203 */
9204 di->disk.number = dl->index;
d23fe947 9205
272906ef
DW		 9206 /* (ab)use di->devs to store a pointer to the device
9207 * we chose
9208 */
9209 di->devs = (struct mdinfo *) dl;
9210
9211 di->disk.raid_disk = i;
9212 di->disk.major = dl->major;
9213 di->disk.minor = dl->minor;
9214 di->disk.state = 0;
d23534e4 9215 di->recovery_start = 0;
5551b113 9216 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9217 di->component_size = a->info.component_size;
9218 di->container_member = inst;
5e46202e 9219 di->bb.supported = 1;
2c8890e9 9220 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9221 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9222 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9223 }
148acb7b 9224 super->random = random32();
272906ef
DW		 9225 di->next = rv;
9226 rv = di;
9227 num_spares++;
9228 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9229 i, di->data_offset);
88758e9d
DW		 9230 }
9231
9232 if (!rv)
9233 /* No spares found */
9234 return rv;
9235 /* Now 'rv' has a list of devices to return.
9236 * Create a metadata_update record to update the
9237 * disk_ord_tbl for the array
9238 */
503975b9 9239 mu = xmalloc(sizeof(*mu));
1011e834 9240 mu->buf = xcalloc(num_spares,
503975b9 9241 sizeof(struct imsm_update_activate_spare));
88758e9d 9242 mu->space = NULL;
cb23f1f4 9243 mu->space_list = NULL;
88758e9d
DW		 9244 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9245 mu->next = *updates;
9246 u = (struct imsm_update_activate_spare *) mu->buf;
9247
9248 for (di = rv ; di ; di = di->next) {
9249 u->type = update_activate_spare;
d23fe947
DW		 9250 u->dl = (struct dl *) di->devs;
9251 di->devs = NULL;
88758e9d
DW		 9252 u->slot = di->disk.raid_disk;
9253 u->array = inst;
9254 u->next = u + 1;
9255 u++;
9256 }
9257 (u-1)->next = NULL;
9258 *updates = mu;
9259
9260 return rv;
9261}
9262
54c2c1ea 9263static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9264{
54c2c1ea 9265 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9266 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9267 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9268 struct disk_info *inf = get_disk_info(u);
9269 struct imsm_disk *disk;
8273f55e
DW		 9270 int i;
9271 int j;
8273f55e 9272
54c2c1ea 9273 for (i = 0; i < map->num_members; i++) {
238c0a71 9274 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9275 for (j = 0; j < new_map->num_members; j++)
9276 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9277 return 1;
9278 }
9279
9280 return 0;
9281}
9282
1a64be56
LM		 9283static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9284{
594dc1b8
JS		 9285 struct dl *dl;
9286
1a64be56 9287 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9288 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9289 return dl;
9290 return NULL;
9291}
9292
9293static int remove_disk_super(struct intel_super *super, int major, int minor)
9294{
594dc1b8 9295 struct dl *prev;
1a64be56
LM		 9296 struct dl *dl;
9297
9298 prev = NULL;
9299 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9300 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9301 /* remove */
9302 if (prev)
9303 prev->next = dl->next;
9304 else
9305 super->disks = dl->next;
9306 dl->next = NULL;
9307 __free_imsm_disk(dl);
1ade5cc1 9308 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9309 break;
9310 }
9311 prev = dl;
9312 }
9313 return 0;
9314}
9315
f21e18ca 9316static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9317
1a64be56
LM		 9318static int add_remove_disk_update(struct intel_super *super)
9319{
9320 int check_degraded = 0;
594dc1b8
JS		 9321 struct dl *disk;
9322
1a64be56
LM		 9323 /* add/remove some spares to/from the metadata/contrainer */
9324 while (super->disk_mgmt_list) {
9325 struct dl *disk_cfg;
9326
9327 disk_cfg = super->disk_mgmt_list;
9328 super->disk_mgmt_list = disk_cfg->next;
9329 disk_cfg->next = NULL;
9330
9331 if (disk_cfg->action == DISK_ADD) {
9332 disk_cfg->next = super->disks;
9333 super->disks = disk_cfg;
9334 check_degraded = 1;
1ade5cc1
N		 9335 dprintf("added %x:%x\n",
9336 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9337 } else if (disk_cfg->action == DISK_REMOVE) {
9338 dprintf("Disk remove action processed: %x.%x\n",
9339 disk_cfg->major, disk_cfg->minor);
9340 disk = get_disk_super(super,
9341 disk_cfg->major,
9342 disk_cfg->minor);
9343 if (disk) {
9344 /* store action status */
9345 disk->action = DISK_REMOVE;
9346 /* remove spare disks only */
9347 if (disk->index == -1) {
9348 remove_disk_super(super,
9349 disk_cfg->major,
9350 disk_cfg->minor);
91c97c54
MT		 9351 } else {
9352 disk_cfg->fd = disk->fd;
9353 disk->fd = -1;
1a64be56
LM		 9354 }
9355 }
9356 /* release allocate disk structure */
9357 __free_imsm_disk(disk_cfg);
9358 }
9359 }
9360 return check_degraded;
9361}
9362
a29911da
PC		 9363static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9364 struct intel_super *super,
9365 void ***space_list)
9366{
9367 struct intel_dev *id;
9368 void **tofree = NULL;
9369 int ret_val = 0;
9370
1ade5cc1 9371 dprintf("(enter)\n");
089f9d79 9372 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9373 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9374 return ret_val;
9375 }
089f9d79 9376 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9377 dprintf("imsm: Error: Memory is not allocated\n");
9378 return ret_val;
9379 }
9380
9381 for (id = super->devlist ; id; id = id->next) {
9382 if (id->index == (unsigned)u->subdev) {
9383 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9384 struct imsm_map *map;
9385 struct imsm_dev *new_dev =
9386 (struct imsm_dev *)*space_list;
238c0a71 9387 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9388 int to_state;
9389 struct dl *new_disk;
9390
9391 if (new_dev == NULL)
9392 return ret_val;
9393 *space_list = **space_list;
9394 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9395 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9396 if (migr_map) {
9397 dprintf("imsm: Error: migration in progress");
9398 return ret_val;
9399 }
9400
9401 to_state = map->map_state;
9402 if ((u->new_level == 5) && (map->raid_level == 0)) {
9403 map->num_members++;
9404 /* this should not happen */
9405 if (u->new_disks[0] < 0) {
9406 map->failed_disk_num =
9407 map->num_members - 1;
9408 to_state = IMSM_T_STATE_DEGRADED;
9409 } else
9410 to_state = IMSM_T_STATE_NORMAL;
9411 }
8e59f3d8 9412 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9413 if (u->new_level > -1)
9414 map->raid_level = u->new_level;
238c0a71 9415 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9416 if ((u->new_level == 5) &&
9417 (migr_map->raid_level == 0)) {
9418 int ord = map->num_members - 1;
9419 migr_map->num_members--;
9420 if (u->new_disks[0] < 0)
9421 ord |= IMSM_ORD_REBUILD;
9422 set_imsm_ord_tbl_ent(map,
9423 map->num_members - 1,
9424 ord);
9425 }
9426 id->dev = new_dev;
9427 tofree = (void **)dev;
9428
4bba0439
PC		 9429 /* update chunk size
9430 */
06fb291a
PB		 9431 if (u->new_chunksize > 0) {
9432 unsigned long long num_data_stripes;
9529d343
MD		 9433 struct imsm_map *dest_map =
9434 get_imsm_map(dev, MAP_0);
06fb291a 9435 int used_disks =
9529d343 9436 imsm_num_data_members(dest_map);
06fb291a
PB		 9437
9438 if (used_disks == 0)
9439 return ret_val;
9440
4bba0439
PC		 9441 map->blocks_per_strip =
9442 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9443 num_data_stripes =
fcc2c9da 9444 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9445 num_data_stripes /= map->blocks_per_strip;
9446 num_data_stripes /= map->num_domains;
9447 set_num_data_stripes(map, num_data_stripes);
9448 }
4bba0439 9449
44490938
MD		 9450 /* ensure blocks_per_member has valid value
9451 */
9452 set_blocks_per_member(map,
9453 per_dev_array_size(map) +
9454 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9455
a29911da
PC		 9456 /* add disk
9457 */
089f9d79
JS		 9458 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9459 migr_map->raid_level == map->raid_level)
a29911da
PC		 9460 goto skip_disk_add;
9461
9462 if (u->new_disks[0] >= 0) {
9463 /* use passes spare
9464 */
9465 new_disk = get_disk_super(super,
9466 major(u->new_disks[0]),
9467 minor(u->new_disks[0]));
7a862a02 9468 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9469 major(u->new_disks[0]),
9470 minor(u->new_disks[0]),
9471 new_disk, new_disk->index);
9472 if (new_disk == NULL)
9473 goto error_disk_add;
9474
9475 new_disk->index = map->num_members - 1;
9476 /* slot to fill in autolayout
9477 */
9478 new_disk->raiddisk = new_disk->index;
9479 new_disk->disk.status |= CONFIGURED_DISK;
9480 new_disk->disk.status &= ~SPARE_DISK;
9481 } else
9482 goto error_disk_add;
9483
9484skip_disk_add:
9485 *tofree = *space_list;
9486 /* calculate new size
9487 */
f3871fdc 9488 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9489
9490 ret_val = 1;
9491 }
9492 }
9493
9494 if (tofree)
9495 *space_list = tofree;
9496 return ret_val;
9497
9498error_disk_add:
9499 dprintf("Error: imsm: Cannot find disk.\n");
9500 return ret_val;
9501}
9502
f3871fdc
AK		 9503static int apply_size_change_update(struct imsm_update_size_change *u,
9504 struct intel_super *super)
9505{
9506 struct intel_dev *id;
9507 int ret_val = 0;
9508
1ade5cc1 9509 dprintf("(enter)\n");
089f9d79 9510 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9511 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9512 return ret_val;
9513 }
9514
9515 for (id = super->devlist ; id; id = id->next) {
9516 if (id->index == (unsigned)u->subdev) {
9517 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9518 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9519 int used_disks = imsm_num_data_members(map);
f3871fdc 9520 unsigned long long blocks_per_member;
06fb291a 9521 unsigned long long num_data_stripes;
44490938
MD		 9522 unsigned long long new_size_per_disk;
9523
9524 if (used_disks == 0)
9525 return 0;
f3871fdc
AK		 9526
9527 /* calculate new size
9528 */
44490938
MD		 9529 new_size_per_disk = u->new_size / used_disks;
9530 blocks_per_member = new_size_per_disk +
9531 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9532 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9533 map->blocks_per_strip;
9534 num_data_stripes /= map->num_domains;
9535 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9536 u->new_size, new_size_per_disk,
06fb291a 9537 num_data_stripes);
f3871fdc 9538 set_blocks_per_member(map, blocks_per_member);
06fb291a 9539 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9540 imsm_set_array_size(dev, u->new_size);
9541
9542 ret_val = 1;
9543 break;
9544 }
9545 }
9546
9547 return ret_val;
9548}
9549
69d40de4
JR		 9550static int prepare_spare_to_activate(struct supertype *st,
9551 struct imsm_update_activate_spare *u)
9552{
9553 struct intel_super *super = st->sb;
9554 int prev_current_vol = super->current_vol;
9555 struct active_array *a;
9556 int ret = 1;
9557
9558 for (a = st->arrays; a; a = a->next)
9559 /*
9560 * Additional initialization (adding bitmap header, filling
9561 * the bitmap area with '1's to force initial rebuild for a whole
9562 * data-area) is required when adding the spare to the volume
9563 * with write-intent bitmap.
9564 */
9565 if (a->info.container_member == u->array &&
9566 a->info.consistency_policy == CONSISTENCY_POLICY_BITMAP) {
9567 struct dl *dl;
9568
9569 for (dl = super->disks; dl; dl = dl->next)
9570 if (dl == u->dl)
9571 break;
9572 if (!dl)
9573 break;
9574
9575 super->current_vol = u->array;
9576 if (st->ss->write_bitmap(st, dl->fd, NoUpdate))
9577 ret = 0;
9578 super->current_vol = prev_current_vol;
9579 }
9580 return ret;
9581}
9582
061d7da3 9583static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9584 struct intel_super *super,
061d7da3
LO		 9585 struct active_array *active_array)
9586{
9587 struct imsm_super *mpb = super->anchor;
9588 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9589 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9590 struct imsm_map *migr_map;
9591 struct active_array *a;
9592 struct imsm_disk *disk;
9593 __u8 to_state;
9594 struct dl *dl;
9595 unsigned int found;
9596 int failed;
5961eeec 9597 int victim;
061d7da3 9598 int i;
5961eeec 9599 int second_map_created = 0;
061d7da3 9600
5961eeec 9601 for (; u; u = u->next) {
238c0a71 9602 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9603
5961eeec 9604 if (victim < 0)
9605 return 0;
061d7da3 9606
5961eeec 9607 for (dl = super->disks; dl; dl = dl->next)
9608 if (dl == u->dl)
9609 break;
061d7da3 9610
5961eeec 9611 if (!dl) {
7a862a02 9612 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9613 u->dl->index);
9614 return 0;
9615 }
061d7da3 9616
5961eeec 9617 /* count failures (excluding rebuilds and the victim)
9618 * to determine map[0] state
9619 */
9620 failed = 0;
9621 for (i = 0; i < map->num_members; i++) {
9622 if (i == u->slot)
9623 continue;
9624 disk = get_imsm_disk(super,
238c0a71 9625 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9626 if (!disk || is_failed(disk))
9627 failed++;
9628 }
061d7da3 9629
5961eeec 9630 /* adding a pristine spare, assign a new index */
9631 if (dl->index < 0) {
9632 dl->index = super->anchor->num_disks;
9633 super->anchor->num_disks++;
9634 }
9635 disk = &dl->disk;
9636 disk->status |= CONFIGURED_DISK;
9637 disk->status &= ~SPARE_DISK;
9638
9639 /* mark rebuild */
238c0a71 9640 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9641 if (!second_map_created) {
9642 second_map_created = 1;
9643 map->map_state = IMSM_T_STATE_DEGRADED;
9644 migrate(dev, super, to_state, MIGR_REBUILD);
9645 } else
9646 map->map_state = to_state;
238c0a71 9647 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9648 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9649 set_imsm_ord_tbl_ent(migr_map, u->slot,
9650 dl->index | IMSM_ORD_REBUILD);
9651
9652 /* update the family_num to mark a new container
9653 * generation, being careful to record the existing
9654 * family_num in orig_family_num to clean up after
9655 * earlier mdadm versions that neglected to set it.
9656 */
9657 if (mpb->orig_family_num == 0)
9658 mpb->orig_family_num = mpb->family_num;
9659 mpb->family_num += super->random;
9660
9661 /* count arrays using the victim in the metadata */
9662 found = 0;
9663 for (a = active_array; a ; a = a->next) {
9664 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9665 map = get_imsm_map(dev, MAP_0);
061d7da3 9666
5961eeec 9667 if (get_imsm_disk_slot(map, victim) >= 0)
9668 found++;
9669 }
061d7da3 9670
5961eeec 9671 /* delete the victim if it is no longer being
9672 * utilized anywhere
061d7da3 9673 */
5961eeec 9674 if (!found) {
9675 struct dl **dlp;
061d7da3 9676
5961eeec 9677 /* We know that 'manager' isn't touching anything,
9678 * so it is safe to delete
9679 */
9680 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9681 if ((*dlp)->index == victim)
9682 break;
5961eeec 9683
9684 /* victim may be on the missing list */
9685 if (!*dlp)
9686 for (dlp = &super->missing; *dlp;
9687 dlp = &(*dlp)->next)
9688 if ((*dlp)->index == victim)
9689 break;
9690 imsm_delete(super, dlp, victim);
9691 }
061d7da3
LO		 9692 }
9693
9694 return 1;
9695}
a29911da 9696
2e5dc010
N		 9697static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9698 struct intel_super *super,
9699 void ***space_list)
9700{
9701 struct dl *new_disk;
9702 struct intel_dev *id;
9703 int i;
9704 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9705 int disk_count = u->old_raid_disks;
2e5dc010
N		 9706 void **tofree = NULL;
9707 int devices_to_reshape = 1;
9708 struct imsm_super *mpb = super->anchor;
9709 int ret_val = 0;
d098291a 9710 unsigned int dev_id;
2e5dc010 9711
1ade5cc1 9712 dprintf("(enter)\n");
2e5dc010
N		 9713
9714 /* enable spares to use in array */
9715 for (i = 0; i < delta_disks; i++) {
9716 new_disk = get_disk_super(super,
9717 major(u->new_disks[i]),
9718 minor(u->new_disks[i]));
7a862a02 9719 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9720 major(u->new_disks[i]), minor(u->new_disks[i]),
9721 new_disk, new_disk->index);
089f9d79
JS		 9722 if (new_disk == NULL ||
9723 (new_disk->index >= 0 &&
9724 new_disk->index < u->old_raid_disks))
2e5dc010 9725 goto update_reshape_exit;
ee4beede 9726 new_disk->index = disk_count++;
2e5dc010
N		 9727 /* slot to fill in autolayout
9728 */
9729 new_disk->raiddisk = new_disk->index;
9730 new_disk->disk.status |=
9731 CONFIGURED_DISK;
9732 new_disk->disk.status &= ~SPARE_DISK;
9733 }
9734
ed7333bd
AK		 9735 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9736 mpb->num_raid_devs);
2e5dc010
N		 9737 /* manage changes in volume
9738 */
d098291a 9739 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9740 void **sp = *space_list;
9741 struct imsm_dev *newdev;
9742 struct imsm_map *newmap, *oldmap;
9743
d098291a
AK		 9744 for (id = super->devlist ; id; id = id->next) {
9745 if (id->index == dev_id)
9746 break;
9747 }
9748 if (id == NULL)
9749 break;
2e5dc010
N		 9750 if (!sp)
9751 continue;
9752 *space_list = *sp;
9753 newdev = (void*)sp;
9754 /* Copy the dev, but not (all of) the map */
9755 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9756 oldmap = get_imsm_map(id->dev, MAP_0);
9757 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9758 /* Copy the current map */
9759 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9760 /* update one device only
9761 */
9762 if (devices_to_reshape) {
ed7333bd
AK		 9763 dprintf("imsm: modifying subdev: %i\n",
9764 id->index);
2e5dc010
N		 9765 devices_to_reshape--;
9766 newdev->vol.migr_state = 1;
4036e7ee 9767 set_vol_curr_migr_unit(newdev, 0);
ea672ee1 9768 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9769 newmap->num_members = u->new_raid_disks;
9770 for (i = 0; i < delta_disks; i++) {
9771 set_imsm_ord_tbl_ent(newmap,
9772 u->old_raid_disks + i,
9773 u->old_raid_disks + i);
9774 }
9775 /* New map is correct, now need to save old map
9776 */
238c0a71 9777 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9778 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9779
f3871fdc 9780 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9781 }
9782
9783 sp = (void **)id->dev;
9784 id->dev = newdev;
9785 *sp = tofree;
9786 tofree = sp;
8e59f3d8
AK		 9787
9788 /* Clear migration record */
9789 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9790 }
819bc634
AK		 9791 if (tofree)
9792 *space_list = tofree;
2e5dc010
N		 9793 ret_val = 1;
9794
9795update_reshape_exit:
9796
9797 return ret_val;
9798}
9799
bb025c2f 9800static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9801 struct intel_super *super,
9802 void ***space_list)
bb025c2f
KW		 9803{
9804 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9805 struct intel_dev *dv;
9806 struct imsm_dev *dev_new;
bb025c2f
KW		 9807 struct imsm_map *map;
9808 struct dl *dm, *du;
8ca6df95 9809 int i;
bb025c2f
KW		 9810
9811 for (dv = super->devlist; dv; dv = dv->next)
9812 if (dv->index == (unsigned int)u->subarray) {
9813 dev = dv->dev;
9814 break;
9815 }
9816
9817 if (dev == NULL)
9818 return 0;
9819
238c0a71 9820 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9821
9822 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9823 unsigned long long num_data_stripes;
9824
43d5ec18 9825 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9826 if (imsm_count_failed(super, dev, MAP_0) !=
9827 (map->num_members / 2))
43d5ec18
KW		 9828 return 0;
9829
bb025c2f
KW		 9830 /* iterate through devices to mark removed disks as spare */
9831 for (dm = super->disks; dm; dm = dm->next) {
9832 if (dm->disk.status & FAILED_DISK) {
9833 int idx = dm->index;
9834 /* update indexes on the disk list */
9835/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9836 the index values will end up being correct.... NB */
9837 for (du = super->disks; du; du = du->next)
9838 if (du->index > idx)
9839 du->index--;
9840 /* mark as spare disk */
a8619d23 9841 mark_spare(dm);
bb025c2f
KW		 9842 }
9843 }
bb025c2f
KW		 9844 /* update map */
9845 map->num_members = map->num_members / 2;
9846 map->map_state = IMSM_T_STATE_NORMAL;
9847 map->num_domains = 1;
9848 map->raid_level = 0;
9849 map->failed_disk_num = -1;
4a353e6e
RS		 9850 num_data_stripes = imsm_dev_size(dev) / 2;
9851 num_data_stripes /= map->blocks_per_strip;
9852 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9853 }
9854
8ca6df95
KW		 9855 if (u->direction == R0_TO_R10) {
9856 void **space;
4a353e6e
RS		 9857 unsigned long long num_data_stripes;
9858
8ca6df95
KW		 9859 /* update slots in current disk list */
9860 for (dm = super->disks; dm; dm = dm->next) {
9861 if (dm->index >= 0)
9862 dm->index *= 2;
9863 }
9864 /* create new *missing* disks */
9865 for (i = 0; i < map->num_members; i++) {
9866 space = *space_list;
9867 if (!space)
9868 continue;
9869 *space_list = *space;
9870 du = (void *)space;
9871 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9872 du->fd = -1;
9873 du->minor = 0;
9874 du->major = 0;
9875 du->index = (i * 2) + 1;
9876 sprintf((char *)du->disk.serial,
9877 " MISSING_%d", du->index);
9878 sprintf((char *)du->serial,
9879 "MISSING_%d", du->index);
9880 du->next = super->missing;
9881 super->missing = du;
9882 }
9883 /* create new dev and map */
9884 space = *space_list;
9885 if (!space)
9886 return 0;
9887 *space_list = *space;
9888 dev_new = (void *)space;
9889 memcpy(dev_new, dev, sizeof(*dev));
9890 /* update new map */
238c0a71 9891 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9892 map->num_members = map->num_members * 2;
1a2487c2 9893 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9894 map->num_domains = 2;
9895 map->raid_level = 1;
4a353e6e
RS		 9896 num_data_stripes = imsm_dev_size(dev) / 2;
9897 num_data_stripes /= map->blocks_per_strip;
9898 num_data_stripes /= map->num_domains;
9899 set_num_data_stripes(map, num_data_stripes);
9900
8ca6df95
KW		 9901 /* replace dev<->dev_new */
9902 dv->dev = dev_new;
9903 }
bb025c2f
KW		 9904 /* update disk order table */
9905 for (du = super->disks; du; du = du->next)
9906 if (du->index >= 0)
9907 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9908 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9909 if (du->index >= 0) {
9910 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9911 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9912 }
bb025c2f
KW		 9913
9914 return 1;
9915}
9916
e8319a19
DW		 9917static void imsm_process_update(struct supertype *st,
9918 struct metadata_update *update)
9919{
9920 /**
9921 * crack open the metadata_update envelope to find the update record
9922 * update can be one of:
d195167d
AK		 9923 * update_reshape_container_disks - all the arrays in the container
9924 * are being reshaped to have more devices. We need to mark
9925 * the arrays for general migration and convert selected spares
9926 * into active devices.
9927 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9928 * device in an array, update the disk_ord_tbl. If this disk is
9929 * present in all member arrays then also clear the SPARE_DISK
9930 * flag
d195167d
AK		 9931 * update_create_array
9932 * update_kill_array
9933 * update_rename_array
9934 * update_add_remove_disk
e8319a19
DW		 9935 */
9936 struct intel_super *super = st->sb;
4d7b1503 9937 struct imsm_super *mpb;
e8319a19
DW		 9938 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9939
4d7b1503
DW		 9940 /* update requires a larger buf but the allocation failed */
9941 if (super->next_len && !super->next_buf) {
9942 super->next_len = 0;
9943 return;
9944 }
9945
9946 if (super->next_buf) {
9947 memcpy(super->next_buf, super->buf, super->len);
9948 free(super->buf);
9949 super->len = super->next_len;
9950 super->buf = super->next_buf;
9951
9952 super->next_len = 0;
9953 super->next_buf = NULL;
9954 }
9955
9956 mpb = super->anchor;
9957
e8319a19 9958 switch (type) {
0ec5d470
AK		 9959 case update_general_migration_checkpoint: {
9960 struct intel_dev *id;
9961 struct imsm_update_general_migration_checkpoint *u =
9962 (void *)update->buf;
9963
1ade5cc1 9964 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9965
9966 /* find device under general migration */
9967 for (id = super->devlist ; id; id = id->next) {
9968 if (is_gen_migration(id->dev)) {
4036e7ee
MT		 9969 set_vol_curr_migr_unit(id->dev,
9970 u->curr_migr_unit);
0ec5d470
AK		 9971 super->updates_pending++;
9972 }
9973 }
9974 break;
9975 }
bb025c2f
KW		 9976 case update_takeover: {
9977 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9978 if (apply_takeover_update(u, super, &update->space_list)) {
9979 imsm_update_version_info(super);
bb025c2f 9980 super->updates_pending++;
1a2487c2 9981 }
bb025c2f
KW		 9982 break;
9983 }
9984
78b10e66 9985 case update_reshape_container_disks: {
d195167d 9986 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9987 if (apply_reshape_container_disks_update(
9988 u, super, &update->space_list))
9989 super->updates_pending++;
78b10e66
N		 9990 break;
9991 }
48c5303a 9992 case update_reshape_migration: {
a29911da
PC		 9993 struct imsm_update_reshape_migration *u = (void *)update->buf;
9994 if (apply_reshape_migration_update(
9995 u, super, &update->space_list))
9996 super->updates_pending++;
48c5303a
PC		 9997 break;
9998 }
f3871fdc
AK		 9999 case update_size_change: {
10000 struct imsm_update_size_change *u = (void *)update->buf;
10001 if (apply_size_change_update(u, super))
10002 super->updates_pending++;
10003 break;
10004 }
e8319a19 10005 case update_activate_spare: {
1011e834 10006 struct imsm_update_activate_spare *u = (void *) update->buf;
69d40de4
JR		 10007
10008 if (prepare_spare_to_activate(st, u) &&
10009 apply_update_activate_spare(u, super, st->arrays))
061d7da3 10010 super->updates_pending++;
8273f55e
DW		 10011 break;
10012 }
10013 case update_create_array: {
10014 /* someone wants to create a new array, we need to be aware of
10015 * a few races/collisions:
10016 * 1/ 'Create' called by two separate instances of mdadm
10017 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
10018 * devices that have since been assimilated via
10019 * activate_spare.
10020 * In the event this update can not be carried out mdadm will
10021 * (FIX ME) notice that its update did not take hold.
10022 */
10023 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10024 struct intel_dev *dv;
8273f55e
DW		 10025 struct imsm_dev *dev;
10026 struct imsm_map *map, *new_map;
10027 unsigned long long start, end;
10028 unsigned long long new_start, new_end;
10029 int i;
54c2c1ea
DW		 10030 struct disk_info *inf;
10031 struct dl *dl;
8273f55e
DW		 10032
10033 /* handle racing creates: first come first serve */
10034 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 10035 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 10036 goto create_error;
8273f55e
DW		 10037 }
10038
10039 /* check update is next in sequence */
10040 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 10041 dprintf("can not create array %d expected index %d\n",
10042 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 10043 goto create_error;
8273f55e
DW		 10044 }
10045
238c0a71 10046 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 10047 new_start = pba_of_lba0(new_map);
44490938 10048 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 10049 inf = get_disk_info(u);
8273f55e
DW		 10050
10051 /* handle activate_spare versus create race:
10052 * check to make sure that overlapping arrays do not include
10053 * overalpping disks
10054 */
10055 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 10056 dev = get_imsm_dev(super, i);
238c0a71 10057 map = get_imsm_map(dev, MAP_0);
5551b113 10058 start = pba_of_lba0(map);
44490938 10059 end = start + per_dev_array_size(map);
8273f55e
DW		 10060 if ((new_start >= start && new_start <= end) ||
10061 (start >= new_start && start <= new_end))
54c2c1ea
DW		 10062 /* overlap */;
10063 else
10064 continue;
10065
10066 if (disks_overlap(super, i, u)) {
1ade5cc1 10067 dprintf("arrays overlap\n");
ba2de7ba 10068 goto create_error;
8273f55e
DW		 10069 }
10070 }
8273f55e 10071
949c47a0
DW		 10072 /* check that prepare update was successful */
10073 if (!update->space) {
1ade5cc1 10074 dprintf("prepare update failed\n");
ba2de7ba 10075 goto create_error;
949c47a0
DW		 10076 }
10077
54c2c1ea
DW		 10078 /* check that all disks are still active before committing
10079 * changes. FIXME: could we instead handle this by creating a
10080 * degraded array? That's probably not what the user expects,
10081 * so better to drop this update on the floor.
10082 */
10083 for (i = 0; i < new_map->num_members; i++) {
10084 dl = serial_to_dl(inf[i].serial, super);
10085 if (!dl) {
1ade5cc1 10086 dprintf("disk disappeared\n");
ba2de7ba 10087 goto create_error;
54c2c1ea 10088 }
949c47a0
DW		 10089 }
10090
8273f55e 10091 super->updates_pending++;
54c2c1ea
DW		 10092
10093 /* convert spares to members and fixup ord_tbl */
10094 for (i = 0; i < new_map->num_members; i++) {
10095 dl = serial_to_dl(inf[i].serial, super);
10096 if (dl->index == -1) {
10097 dl->index = mpb->num_disks;
10098 mpb->num_disks++;
10099 dl->disk.status |= CONFIGURED_DISK;
10100 dl->disk.status &= ~SPARE_DISK;
10101 }
10102 set_imsm_ord_tbl_ent(new_map, i, dl->index);
10103 }
10104
ba2de7ba
DW		 10105 dv = update->space;
10106 dev = dv->dev;
949c47a0
DW		 10107 update->space = NULL;
10108 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 10109 dv->index = u->dev_idx;
10110 dv->next = super->devlist;
10111 super->devlist = dv;
8273f55e 10112 mpb->num_raid_devs++;
8273f55e 10113
4d1313e9 10114 imsm_update_version_info(super);
8273f55e 10115 break;
ba2de7ba
DW		 10116 create_error:
10117 /* mdmon knows how to release update->space, but not
10118 * ((struct intel_dev *) update->space)->dev
10119 */
10120 if (update->space) {
10121 dv = update->space;
10122 free(dv->dev);
10123 }
8273f55e 10124 break;
e8319a19 10125 }
33414a01
DW		 10126 case update_kill_array: {
10127 struct imsm_update_kill_array *u = (void *) update->buf;
10128 int victim = u->dev_idx;
10129 struct active_array *a;
10130 struct intel_dev **dp;
10131 struct imsm_dev *dev;
10132
10133 /* sanity check that we are not affecting the uuid of
10134 * active arrays, or deleting an active array
10135 *
10136 * FIXME when immutable ids are available, but note that
10137 * we'll also need to fixup the invalidated/active
10138 * subarray indexes in mdstat
10139 */
10140 for (a = st->arrays; a; a = a->next)
10141 if (a->info.container_member >= victim)
10142 break;
10143 /* by definition if mdmon is running at least one array
10144 * is active in the container, so checking
10145 * mpb->num_raid_devs is just extra paranoia
10146 */
10147 dev = get_imsm_dev(super, victim);
10148 if (a || !dev || mpb->num_raid_devs == 1) {
10149 dprintf("failed to delete subarray-%d\n", victim);
10150 break;
10151 }
10152
10153 for (dp = &super->devlist; *dp;)
f21e18ca 10154 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 10155 *dp = (*dp)->next;
10156 } else {
f21e18ca 10157 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 10158 (*dp)->index--;
10159 dp = &(*dp)->next;
10160 }
10161 mpb->num_raid_devs--;
10162 super->updates_pending++;
10163 break;
10164 }
aa534678
DW		 10165 case update_rename_array: {
10166 struct imsm_update_rename_array *u = (void *) update->buf;
10167 char name[MAX_RAID_SERIAL_LEN+1];
10168 int target = u->dev_idx;
10169 struct active_array *a;
10170 struct imsm_dev *dev;
10171
10172 /* sanity check that we are not affecting the uuid of
10173 * an active array
10174 */
40659392 10175 memset(name, 0, sizeof(name));
aa534678
DW		 10176 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
10177 name[MAX_RAID_SERIAL_LEN] = '\0';
10178 for (a = st->arrays; a; a = a->next)
10179 if (a->info.container_member == target)
10180 break;
10181 dev = get_imsm_dev(super, u->dev_idx);
10182 if (a || !dev || !check_name(super, name, 1)) {
10183 dprintf("failed to rename subarray-%d\n", target);
10184 break;
10185 }
10186
40659392 10187 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10188 super->updates_pending++;
10189 break;
10190 }
1a64be56 10191 case update_add_remove_disk: {
43dad3d6 10192 /* we may be able to repair some arrays if disks are
095b8088 10193 * being added, check the status of add_remove_disk
1a64be56
LM		 10194 * if discs has been added.
10195 */
10196 if (add_remove_disk_update(super)) {
43dad3d6 10197 struct active_array *a;
072b727f
DW		 10198
10199 super->updates_pending++;
1a64be56 10200 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10201 a->check_degraded = 1;
10202 }
43dad3d6 10203 break;
e8319a19 10204 }
bbab0940
TM		 10205 case update_prealloc_badblocks_mem:
10206 break;
e6e9dd3f
AP		 10207 case update_rwh_policy: {
10208 struct imsm_update_rwh_policy *u = (void *)update->buf;
10209 int target = u->dev_idx;
10210 struct imsm_dev *dev = get_imsm_dev(super, target);
10211 if (!dev) {
10212 dprintf("could not find subarray-%d\n", target);
10213 break;
10214 }
10215
10216 if (dev->rwh_policy != u->new_policy) {
10217 dev->rwh_policy = u->new_policy;
10218 super->updates_pending++;
10219 }
10220 break;
10221 }
1a64be56 10222 default:
ebf3be99 10223 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10224 }
e8319a19 10225}
88758e9d 10226
bc0b9d34
PC		 10227static struct mdinfo *get_spares_for_grow(struct supertype *st);
10228
5fe6f031
N		 10229static int imsm_prepare_update(struct supertype *st,
10230 struct metadata_update *update)
8273f55e 10231{
949c47a0 10232 /**
4d7b1503
DW		 10233 * Allocate space to hold new disk entries, raid-device entries or a new
10234 * mpb if necessary. The manager synchronously waits for updates to
10235 * complete in the monitor, so new mpb buffers allocated here can be
10236 * integrated by the monitor thread without worrying about live pointers
10237 * in the manager thread.
8273f55e 10238 */
095b8088 10239 enum imsm_update_type type;
4d7b1503 10240 struct intel_super *super = st->sb;
f36a9ecd 10241 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10242 struct imsm_super *mpb = super->anchor;
10243 size_t buf_len;
10244 size_t len = 0;
949c47a0 10245
095b8088
N		 10246 if (update->len < (int)sizeof(type))
10247 return 0;
10248
10249 type = *(enum imsm_update_type *) update->buf;
10250
949c47a0 10251 switch (type) {
0ec5d470 10252 case update_general_migration_checkpoint:
095b8088
N		 10253 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10254 return 0;
1ade5cc1 10255 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10256 break;
abedf5fc
KW		 10257 case update_takeover: {
10258 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10259 if (update->len < (int)sizeof(*u))
10260 return 0;
abedf5fc
KW		 10261 if (u->direction == R0_TO_R10) {
10262 void **tail = (void **)&update->space_list;
10263 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10264 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10265 int num_members = map->num_members;
10266 void *space;
10267 int size, i;
abedf5fc
KW		 10268 /* allocate memory for added disks */
10269 for (i = 0; i < num_members; i++) {
10270 size = sizeof(struct dl);
503975b9 10271 space = xmalloc(size);
abedf5fc
KW		 10272 *tail = space;
10273 tail = space;
10274 *tail = NULL;
10275 }
10276 /* allocate memory for new device */
10277 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10278 (num_members * sizeof(__u32));
503975b9
N		 10279 space = xmalloc(size);
10280 *tail = space;
10281 tail = space;
10282 *tail = NULL;
10283 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10284 }
10285
10286 break;
10287 }
78b10e66 10288 case update_reshape_container_disks: {
d195167d
AK		 10289 /* Every raid device in the container is about to
10290 * gain some more devices, and we will enter a
10291 * reconfiguration.
10292 * So each 'imsm_map' will be bigger, and the imsm_vol
10293 * will now hold 2 of them.
10294 * Thus we need new 'struct imsm_dev' allocations sized
10295 * as sizeof_imsm_dev but with more devices in both maps.
10296 */
10297 struct imsm_update_reshape *u = (void *)update->buf;
10298 struct intel_dev *dl;
10299 void **space_tail = (void**)&update->space_list;
10300
095b8088
N		 10301 if (update->len < (int)sizeof(*u))
10302 return 0;
10303
1ade5cc1 10304 dprintf("for update_reshape\n");
d195167d
AK		 10305
10306 for (dl = super->devlist; dl; dl = dl->next) {
10307 int size = sizeof_imsm_dev(dl->dev, 1);
10308 void *s;
d677e0b8
AK		 10309 if (u->new_raid_disks > u->old_raid_disks)
10310 size += sizeof(__u32)*2*
10311 (u->new_raid_disks - u->old_raid_disks);
503975b9 10312 s = xmalloc(size);
d195167d
AK		 10313 *space_tail = s;
10314 space_tail = s;
10315 *space_tail = NULL;
10316 }
10317
10318 len = disks_to_mpb_size(u->new_raid_disks);
10319 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10320 break;
10321 }
48c5303a 10322 case update_reshape_migration: {
bc0b9d34
PC		 10323 /* for migration level 0->5 we need to add disks
10324 * so the same as for container operation we will copy
10325 * device to the bigger location.
10326 * in memory prepared device and new disk area are prepared
10327 * for usage in process update
10328 */
10329 struct imsm_update_reshape_migration *u = (void *)update->buf;
10330 struct intel_dev *id;
10331 void **space_tail = (void **)&update->space_list;
10332 int size;
10333 void *s;
10334 int current_level = -1;
10335
095b8088
N		 10336 if (update->len < (int)sizeof(*u))
10337 return 0;
10338
1ade5cc1 10339 dprintf("for update_reshape\n");
bc0b9d34
PC		 10340
10341 /* add space for bigger array in update
10342 */
10343 for (id = super->devlist; id; id = id->next) {
10344 if (id->index == (unsigned)u->subdev) {
10345 size = sizeof_imsm_dev(id->dev, 1);
10346 if (u->new_raid_disks > u->old_raid_disks)
10347 size += sizeof(__u32)*2*
10348 (u->new_raid_disks - u->old_raid_disks);
503975b9 10349 s = xmalloc(size);
bc0b9d34
PC		 10350 *space_tail = s;
10351 space_tail = s;
10352 *space_tail = NULL;
10353 break;
10354 }
10355 }
10356 if (update->space_list == NULL)
10357 break;
10358
10359 /* add space for disk in update
10360 */
10361 size = sizeof(struct dl);
503975b9 10362 s = xmalloc(size);
bc0b9d34
PC		 10363 *space_tail = s;
10364 space_tail = s;
10365 *space_tail = NULL;
10366
10367 /* add spare device to update
10368 */
10369 for (id = super->devlist ; id; id = id->next)
10370 if (id->index == (unsigned)u->subdev) {
10371 struct imsm_dev *dev;
10372 struct imsm_map *map;
10373
10374 dev = get_imsm_dev(super, u->subdev);
238c0a71 10375 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10376 current_level = map->raid_level;
10377 break;
10378 }
089f9d79 10379 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10380 struct mdinfo *spares;
10381
10382 spares = get_spares_for_grow(st);
10383 if (spares) {
10384 struct dl *dl;
10385 struct mdinfo *dev;
10386
10387 dev = spares->devs;
10388 if (dev) {
10389 u->new_disks[0] =
10390 makedev(dev->disk.major,
10391 dev->disk.minor);
10392 dl = get_disk_super(super,
10393 dev->disk.major,
10394 dev->disk.minor);
10395 dl->index = u->old_raid_disks;
10396 dev = dev->next;
10397 }
10398 sysfs_free(spares);
10399 }
10400 }
10401 len = disks_to_mpb_size(u->new_raid_disks);
10402 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10403 break;
10404 }
f3871fdc 10405 case update_size_change: {
095b8088
N		 10406 if (update->len < (int)sizeof(struct imsm_update_size_change))
10407 return 0;
10408 break;
10409 }
10410 case update_activate_spare: {
10411 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10412 return 0;
f3871fdc
AK		 10413 break;
10414 }
949c47a0
DW		 10415 case update_create_array: {
10416 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10417 struct intel_dev *dv;
54c2c1ea 10418 struct imsm_dev *dev = &u->dev;
238c0a71 10419 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10420 struct dl *dl;
10421 struct disk_info *inf;
10422 int i;
10423 int activate = 0;
949c47a0 10424
095b8088
N		 10425 if (update->len < (int)sizeof(*u))
10426 return 0;
10427
54c2c1ea
DW		 10428 inf = get_disk_info(u);
10429 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10430 /* allocate a new super->devlist entry */
503975b9
N		 10431 dv = xmalloc(sizeof(*dv));
10432 dv->dev = xmalloc(len);
10433 update->space = dv;
949c47a0 10434
54c2c1ea
DW		 10435 /* count how many spares will be converted to members */
10436 for (i = 0; i < map->num_members; i++) {
10437 dl = serial_to_dl(inf[i].serial, super);
10438 if (!dl) {
10439 /* hmm maybe it failed?, nothing we can do about
10440 * it here
10441 */
10442 continue;
10443 }
10444 if (count_memberships(dl, super) == 0)
10445 activate++;
10446 }
10447 len += activate * sizeof(struct imsm_disk);
949c47a0 10448 break;
095b8088
N		 10449 }
10450 case update_kill_array: {
10451 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10452 return 0;
949c47a0
DW		 10453 break;
10454 }
095b8088
N		 10455 case update_rename_array: {
10456 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10457 return 0;
10458 break;
10459 }
10460 case update_add_remove_disk:
10461 /* no update->len needed */
10462 break;
bbab0940
TM		 10463 case update_prealloc_badblocks_mem:
10464 super->extra_space += sizeof(struct bbm_log) -
10465 get_imsm_bbm_log_size(super->bbm_log);
10466 break;
e6e9dd3f
AP		 10467 case update_rwh_policy: {
10468 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10469 return 0;
10470 break;
10471 }
095b8088
N		 10472 default:
10473 return 0;
949c47a0 10474 }
8273f55e 10475
4d7b1503
DW		 10476 /* check if we need a larger metadata buffer */
10477 if (super->next_buf)
10478 buf_len = super->next_len;
10479 else
10480 buf_len = super->len;
10481
bbab0940 10482 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10483 /* ok we need a larger buf than what is currently allocated
10484 * if this allocation fails process_update will notice that
10485 * ->next_len is set and ->next_buf is NULL
10486 */
bbab0940
TM		 10487 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10488 super->extra_space + len, sector_size);
4d7b1503
DW		 10489 if (super->next_buf)
10490 free(super->next_buf);
10491
10492 super->next_len = buf_len;
f36a9ecd 10493 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10494 memset(super->next_buf, 0, buf_len);
10495 else
4d7b1503
DW		 10496 super->next_buf = NULL;
10497 }
5fe6f031 10498 return 1;
8273f55e
DW		 10499}
10500
ae6aad82 10501/* must be called while manager is quiesced */
f21e18ca 10502static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10503{
10504 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10505 struct dl *iter;
10506 struct imsm_dev *dev;
10507 struct imsm_map *map;
4c9e8c1e 10508 unsigned int i, j, num_members;
fb12a745 10509 __u32 ord, ord_map0;
4c9e8c1e 10510 struct bbm_log *log = super->bbm_log;
ae6aad82 10511
1ade5cc1 10512 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10513
10514 /* shift all indexes down one */
10515 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10516 if (iter->index > (int)index)
ae6aad82 10517 iter->index--;
47ee5a45 10518 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10519 if (iter->index > (int)index)
47ee5a45 10520 iter->index--;
ae6aad82
DW		 10521
10522 for (i = 0; i < mpb->num_raid_devs; i++) {
10523 dev = get_imsm_dev(super, i);
238c0a71 10524 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10525 num_members = map->num_members;
10526 for (j = 0; j < num_members; j++) {
10527 /* update ord entries being careful not to propagate
10528 * ord-flags to the first map
10529 */
238c0a71 10530 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10531 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10532
24565c9a
DW		 10533 if (ord_to_idx(ord) <= index)
10534 continue;
ae6aad82 10535
238c0a71 10536 map = get_imsm_map(dev, MAP_0);
fb12a745 10537 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10538 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10539 if (map)
10540 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10541 }
10542 }
10543
4c9e8c1e
TM		 10544 for (i = 0; i < log->entry_count; i++) {
10545 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10546
10547 if (entry->disk_ordinal <= index)
10548 continue;
10549 entry->disk_ordinal--;
10550 }
10551
ae6aad82
DW		 10552 mpb->num_disks--;
10553 super->updates_pending++;
24565c9a
DW		 10554 if (*dlp) {
10555 struct dl *dl = *dlp;
10556
10557 *dlp = (*dlp)->next;
10558 __free_imsm_disk(dl);
10559 }
ae6aad82 10560}
9a717282 10561
9a717282
AK		 10562static int imsm_get_allowed_degradation(int level, int raid_disks,
10563 struct intel_super *super,
10564 struct imsm_dev *dev)
10565{
10566 switch (level) {
bf5cf7c7 10567 case 1:
9a717282
AK		 10568 case 10:{
10569 int ret_val = 0;
10570 struct imsm_map *map;
10571 int i;
10572
10573 ret_val = raid_disks/2;
10574 /* check map if all disks pairs not failed
10575 * in both maps
10576 */
238c0a71 10577 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10578 for (i = 0; i < ret_val; i++) {
10579 int degradation = 0;
10580 if (get_imsm_disk(super, i) == NULL)
10581 degradation++;
10582 if (get_imsm_disk(super, i + 1) == NULL)
10583 degradation++;
10584 if (degradation == 2)
10585 return 0;
10586 }
238c0a71 10587 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10588 /* if there is no second map
10589 * result can be returned
10590 */
10591 if (map == NULL)
10592 return ret_val;
10593 /* check degradation in second map
10594 */
10595 for (i = 0; i < ret_val; i++) {
10596 int degradation = 0;
10597 if (get_imsm_disk(super, i) == NULL)
10598 degradation++;
10599 if (get_imsm_disk(super, i + 1) == NULL)
10600 degradation++;
10601 if (degradation == 2)
10602 return 0;
10603 }
10604 return ret_val;
10605 }
10606 case 5:
10607 return 1;
10608 case 6:
10609 return 2;
10610 default:
10611 return 0;
10612 }
10613}
10614
d31ad643
PB		 10615/*******************************************************************************
10616 * Function: validate_container_imsm
10617 * Description: This routine validates container after assemble,
10618 * eg. if devices in container are under the same controller.
10619 *
10620 * Parameters:
10621 * info : linked list with info about devices used in array
10622 * Returns:
10623 * 1 : HBA mismatch
10624 * 0 : Success
10625 ******************************************************************************/
10626int validate_container_imsm(struct mdinfo *info)
10627{
6b781d33
AP		 10628 if (check_env("IMSM_NO_PLATFORM"))
10629 return 0;
d31ad643 10630
6b781d33
AP		 10631 struct sys_dev *idev;
10632 struct sys_dev *hba = NULL;
10633 struct sys_dev *intel_devices = find_intel_devices();
10634 char *dev_path = devt_to_devpath(makedev(info->disk.major,
7c798f87 10635 info->disk.minor), 1, NULL);
6b781d33
AP		 10636
10637 for (idev = intel_devices; idev; idev = idev->next) {
10638 if (dev_path && strstr(dev_path, idev->path)) {
10639 hba = idev;
10640 break;
d31ad643 10641 }
6b781d33
AP		 10642 }
10643 if (dev_path)
d31ad643
PB		 10644 free(dev_path);
10645
6b781d33
AP		 10646 if (!hba) {
10647 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10648 devid2kname(makedev(info->disk.major, info->disk.minor)));
10649 return 1;
10650 }
10651
10652 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10653 struct mdinfo *dev;
10654
10655 for (dev = info->next; dev; dev = dev->next) {
7c798f87
MT		 10656 dev_path = devt_to_devpath(makedev(dev->disk.major,
10657 dev->disk.minor), 1, NULL);
6b781d33
AP		 10658
10659 struct sys_dev *hba2 = NULL;
10660 for (idev = intel_devices; idev; idev = idev->next) {
10661 if (dev_path && strstr(dev_path, idev->path)) {
10662 hba2 = idev;
10663 break;
d31ad643
PB		 10664 }
10665 }
6b781d33
AP		 10666 if (dev_path)
10667 free(dev_path);
10668
10669 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10670 get_orom_by_device_id(hba2->dev_id);
10671
10672 if (hba2 && hba->type != hba2->type) {
10673 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10674 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10675 return 1;
10676 }
10677
07cb1e57 10678 if (orom != orom2) {
6b781d33
AP		 10679 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10680 " This operation is not supported and can lead to data loss.\n");
10681 return 1;
10682 }
10683
10684 if (!orom) {
10685 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10686 " This operation is not supported and can lead to data loss.\n");
10687 return 1;
10688 }
d31ad643 10689 }
6b781d33 10690
d31ad643
PB		 10691 return 0;
10692}
32141c17 10693
6f50473f
TM		 10694/*******************************************************************************
10695* Function: imsm_record_badblock
10696* Description: This routine stores new bad block record in BBM log
10697*
10698* Parameters:
10699* a : array containing a bad block
10700* slot : disk number containing a bad block
10701* sector : bad block sector
10702* length : bad block sectors range
10703* Returns:
10704* 1 : Success
10705* 0 : Error
10706******************************************************************************/
10707static int imsm_record_badblock(struct active_array *a, int slot,
10708 unsigned long long sector, int length)
10709{
10710 struct intel_super *super = a->container->sb;
10711 int ord;
10712 int ret;
10713
10714 ord = imsm_disk_slot_to_ord(a, slot);
10715 if (ord < 0)
10716 return 0;
10717
10718 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10719 length);
10720 if (ret)
10721 super->updates_pending++;
10722
10723 return ret;
10724}
c07a5a4f
TM		 10725/*******************************************************************************
10726* Function: imsm_clear_badblock
10727* Description: This routine clears bad block record from BBM log
10728*
10729* Parameters:
10730* a : array containing a bad block
10731* slot : disk number containing a bad block
10732* sector : bad block sector
10733* length : bad block sectors range
10734* Returns:
10735* 1 : Success
10736* 0 : Error
10737******************************************************************************/
10738static int imsm_clear_badblock(struct active_array *a, int slot,
10739 unsigned long long sector, int length)
10740{
10741 struct intel_super *super = a->container->sb;
10742 int ord;
10743 int ret;
10744
10745 ord = imsm_disk_slot_to_ord(a, slot);
10746 if (ord < 0)
10747 return 0;
10748
10749 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10750 if (ret)
10751 super->updates_pending++;
10752
10753 return ret;
10754}
928f1424
TM		 10755/*******************************************************************************
10756* Function: imsm_get_badblocks
10757* Description: This routine get list of bad blocks for an array
10758*
10759* Parameters:
10760* a : array
10761* slot : disk number
10762* Returns:
10763* bb : structure containing bad blocks
10764* NULL : error
10765******************************************************************************/
10766static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10767{
10768 int inst = a->info.container_member;
10769 struct intel_super *super = a->container->sb;
10770 struct imsm_dev *dev = get_imsm_dev(super, inst);
10771 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10772 int ord;
10773
10774 ord = imsm_disk_slot_to_ord(a, slot);
10775 if (ord < 0)
10776 return NULL;
10777
10778 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10779 per_dev_array_size(map), &super->bb);
928f1424
TM		 10780
10781 return &super->bb;
10782}
27156a57
TM		 10783/*******************************************************************************
10784* Function: examine_badblocks_imsm
10785* Description: Prints list of bad blocks on a disk to the standard output
10786*
10787* Parameters:
10788* st : metadata handler
10789* fd : open file descriptor for device
10790* devname : device name
10791* Returns:
10792* 0 : Success
10793* 1 : Error
10794******************************************************************************/
10795static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10796{
10797 struct intel_super *super = st->sb;
10798 struct bbm_log *log = super->bbm_log;
10799 struct dl *d = NULL;
10800 int any = 0;
10801
10802 for (d = super->disks; d ; d = d->next) {
10803 if (strcmp(d->devname, devname) == 0)
10804 break;
10805 }
10806
10807 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10808 pr_err("%s doesn't appear to be part of a raid array\n",
10809 devname);
10810 return 1;
10811 }
10812
10813 if (log != NULL) {
10814 unsigned int i;
10815 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10816
10817 for (i = 0; i < log->entry_count; i++) {
10818 if (entry[i].disk_ordinal == d->index) {
10819 unsigned long long sector = __le48_to_cpu(
10820 &entry[i].defective_block_start);
10821 int cnt = entry[i].marked_count + 1;
10822
10823 if (!any) {
10824 printf("Bad-blocks on %s:\n", devname);
10825 any = 1;
10826 }
10827
10828 printf("%20llu for %d sectors\n", sector, cnt);
10829 }
10830 }
10831 }
10832
10833 if (!any)
10834 printf("No bad-blocks list configured on %s\n", devname);
10835
10836 return 0;
10837}
687629c2
AK		 10838/*******************************************************************************
10839 * Function: init_migr_record_imsm
10840 * Description: Function inits imsm migration record
10841 * Parameters:
10842 * super : imsm internal array info
10843 * dev : device under migration
10844 * info : general array info to find the smallest device
10845 * Returns:
10846 * none
10847 ******************************************************************************/
10848void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10849 struct mdinfo *info)
10850{
10851 struct intel_super *super = st->sb;
10852 struct migr_record *migr_rec = super->migr_rec;
10853 int new_data_disks;
10854 unsigned long long dsize, dev_sectors;
10855 long long unsigned min_dev_sectors = -1LLU;
238c0a71
AK		 10856 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10857 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10858 unsigned long long num_migr_units;
3ef4403c 10859 unsigned long long array_blocks;
2f86fda3 10860 struct dl *dl_disk = NULL;
687629c2
AK		 10861
10862 memset(migr_rec, 0, sizeof(struct migr_record));
10863 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10864
10865 /* only ascending reshape supported now */
10866 migr_rec->ascending_migr = __cpu_to_le32(1);
10867
10868 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10869 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10870 migr_rec->dest_depth_per_unit *=
10871 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10872 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10873 migr_rec->blocks_per_unit =
10874 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10875 migr_rec->dest_depth_per_unit =
10876 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10877 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10878 num_migr_units =
10879 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10880
10881 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10882 num_migr_units++;
9f421827 10883 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10884
10885 migr_rec->post_migr_vol_cap = dev->size_low;
10886 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10887
687629c2 10888 /* Find the smallest dev */
2f86fda3
MT		 10889 for (dl_disk = super->disks; dl_disk ; dl_disk = dl_disk->next) {
10890 /* ignore spares in container */
10891 if (dl_disk->index < 0)
687629c2 10892 continue;
2f86fda3 10893 get_dev_size(dl_disk->fd, NULL, &dsize);
687629c2
AK		 10894 dev_sectors = dsize / 512;
10895 if (dev_sectors < min_dev_sectors)
10896 min_dev_sectors = dev_sectors;
687629c2 10897 }
9f421827 10898 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10899 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10900
10901 write_imsm_migr_rec(st);
10902
10903 return;
10904}
10905
10906/*******************************************************************************
10907 * Function: save_backup_imsm
10908 * Description: Function saves critical data stripes to Migration Copy Area
10909 * and updates the current migration unit status.
10910 * Use restore_stripes() to form a destination stripe,
10911 * and to write it to the Copy Area.
10912 * Parameters:
10913 * st : supertype information
aea93171 10914 * dev : imsm device that backup is saved for
687629c2
AK		 10915 * info : general array info
10916 * buf : input buffer
687629c2
AK		 10917 * length : length of data to backup (blocks_per_unit)
10918 * Returns:
10919 * 0 : success
10920 *, -1 : fail
10921 ******************************************************************************/
10922int save_backup_imsm(struct supertype *st,
10923 struct imsm_dev *dev,
10924 struct mdinfo *info,
10925 void *buf,
687629c2
AK		 10926 int length)
10927{
10928 int rv = -1;
10929 struct intel_super *super = st->sb;
594dc1b8
JS		 10930 unsigned long long *target_offsets;
10931 int *targets;
687629c2 10932 int i;
238c0a71 10933 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10934 int new_disks = map_dest->num_members;
ab724b98
AK		 10935 int dest_layout = 0;
10936 int dest_chunk;
d1877f69 10937 unsigned long long start;
9529d343 10938 int data_disks = imsm_num_data_members(map_dest);
687629c2 10939
503975b9 10940 targets = xmalloc(new_disks * sizeof(int));
687629c2 10941
2f86fda3
MT		 10942 for (i = 0; i < new_disks; i++) {
10943 struct dl *dl_disk = get_imsm_dl_disk(super, i);
10944
10945 targets[i] = dl_disk->fd;
10946 }
7e45b550 10947
503975b9 10948 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10949
d1877f69 10950 start = info->reshape_progress * 512;
687629c2 10951 for (i = 0; i < new_disks; i++) {
9f421827 10952 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10953 /* move back copy area adderss, it will be moved forward
10954 * in restore_stripes() using start input variable
10955 */
10956 target_offsets[i] -= start/data_disks;
687629c2
AK		 10957 }
10958
68eb8bc6 10959 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10960 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10961
687629c2
AK		 10962 if (restore_stripes(targets, /* list of dest devices */
10963 target_offsets, /* migration record offsets */
10964 new_disks,
ab724b98
AK		 10965 dest_chunk,
10966 map_dest->raid_level,
10967 dest_layout,
10968 -1, /* source backup file descriptor */
10969 0, /* input buf offset
10970 * always 0 buf is already offseted */
d1877f69 10971 start,
687629c2
AK		 10972 length,
10973 buf) != 0) {
e7b84f9d 10974 pr_err("Error restoring stripes\n");
687629c2
AK		 10975 goto abort;
10976 }
10977
10978 rv = 0;
10979
10980abort:
10981 if (targets) {
687629c2
AK		 10982 free(targets);
10983 }
10984 free(target_offsets);
10985
10986 return rv;
10987}
10988
10989/*******************************************************************************
10990 * Function: save_checkpoint_imsm
10991 * Description: Function called for current unit status update
10992 * in the migration record. It writes it to disk.
10993 * Parameters:
10994 * super : imsm internal array info
10995 * info : general array info
10996 * Returns:
10997 * 0: success
10998 * 1: failure
0228d92c
AK		 10999 * 2: failure, means no valid migration record
11000 * / no general migration in progress /
687629c2
AK		 11001 ******************************************************************************/
11002int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
11003{
11004 struct intel_super *super = st->sb;
f8b72ef5
AK		 11005 unsigned long long blocks_per_unit;
11006 unsigned long long curr_migr_unit;
11007
2f86fda3 11008 if (load_imsm_migr_rec(super) != 0) {
7a862a02 11009 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 11010 return 1;
11011 }
11012
f8b72ef5
AK		 11013 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
11014 if (blocks_per_unit == 0) {
0228d92c
AK		 11015 dprintf("imsm: no migration in progress.\n");
11016 return 2;
687629c2 11017 }
f8b72ef5
AK		 11018 curr_migr_unit = info->reshape_progress / blocks_per_unit;
11019 /* check if array is alligned to copy area
11020 * if it is not alligned, add one to current migration unit value
11021 * this can happend on array reshape finish only
11022 */
11023 if (info->reshape_progress % blocks_per_unit)
11024 curr_migr_unit++;
687629c2 11025
9f421827 11026 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 11027 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 11028 set_migr_dest_1st_member_lba(super->migr_rec,
11029 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
11030
687629c2 11031 if (write_imsm_migr_rec(st) < 0) {
7a862a02 11032 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 11033 return 1;
11034 }
11035
11036 return 0;
11037}
11038
276d77db
AK		 11039/*******************************************************************************
11040 * Function: recover_backup_imsm
11041 * Description: Function recovers critical data from the Migration Copy Area
11042 * while assembling an array.
11043 * Parameters:
11044 * super : imsm internal array info
11045 * info : general array info
11046 * Returns:
11047 * 0 : success (or there is no data to recover)
11048 * 1 : fail
11049 ******************************************************************************/
11050int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
11051{
11052 struct intel_super *super = st->sb;
11053 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 11054 struct imsm_map *map_dest;
276d77db
AK		 11055 struct intel_dev *id = NULL;
11056 unsigned long long read_offset;
11057 unsigned long long write_offset;
11058 unsigned unit_len;
2f86fda3 11059 int new_disks, err;
276d77db
AK		 11060 char *buf = NULL;
11061 int retval = 1;
f36a9ecd 11062 unsigned int sector_size = super->sector_size;
4036e7ee
MT		 11063 unsigned long long curr_migr_unit = current_migr_unit(migr_rec);
11064 unsigned long long num_migr_units = get_num_migr_units(migr_rec);
276d77db 11065 char buffer[20];
6c3560c0 11066 int skipped_disks = 0;
2f86fda3 11067 struct dl *dl_disk;
276d77db
AK		 11068
11069 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
11070 if (err < 1)
11071 return 1;
11072
11073 /* recover data only during assemblation */
11074 if (strncmp(buffer, "inactive", 8) != 0)
11075 return 0;
11076 /* no data to recover */
11077 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
11078 return 0;
11079 if (curr_migr_unit >= num_migr_units)
11080 return 1;
11081
11082 /* find device during reshape */
11083 for (id = super->devlist; id; id = id->next)
11084 if (is_gen_migration(id->dev))
11085 break;
11086 if (id == NULL)
11087 return 1;
11088
238c0a71 11089 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 11090 new_disks = map_dest->num_members;
11091
9f421827 11092 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 11093
9f421827 11094 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 11095 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 11096
11097 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 11098 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 11099 goto abort;
276d77db 11100
2f86fda3
MT		 11101 for (dl_disk = super->disks; dl_disk; dl_disk = dl_disk->next) {
11102 if (dl_disk->index < 0)
11103 continue;
276d77db 11104
2f86fda3 11105 if (dl_disk->fd < 0) {
6c3560c0
AK		 11106 skipped_disks++;
11107 continue;
11108 }
2f86fda3 11109 if (lseek64(dl_disk->fd, read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11110 pr_err("Cannot seek to block: %s\n",
11111 strerror(errno));
137debce
AK		 11112 skipped_disks++;
11113 continue;
276d77db 11114 }
83b3de77 11115 if (read(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) {
e7b84f9d
N		 11116 pr_err("Cannot read copy area block: %s\n",
11117 strerror(errno));
137debce
AK		 11118 skipped_disks++;
11119 continue;
276d77db 11120 }
2f86fda3 11121 if (lseek64(dl_disk->fd, write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11122 pr_err("Cannot seek to block: %s\n",
11123 strerror(errno));
137debce
AK		 11124 skipped_disks++;
11125 continue;
276d77db 11126 }
83b3de77 11127 if (write(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) {
e7b84f9d
N		 11128 pr_err("Cannot restore block: %s\n",
11129 strerror(errno));
137debce
AK		 11130 skipped_disks++;
11131 continue;
276d77db
AK		 11132 }
11133 }
11134
137debce
AK		 11135 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11136 new_disks,
11137 super,
11138 id->dev)) {
7a862a02 11139 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11140 goto abort;
11141 }
11142
befb629b
AK		 11143 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11144 /* ignore error == 2, this can mean end of reshape here
11145 */
7a862a02 11146 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11147 } else
276d77db 11148 retval = 0;
276d77db
AK		 11149
11150abort:
276d77db
AK		 11151 free(buf);
11152 return retval;
11153}
11154
2cda7640
ML		 11155static char disk_by_path[] = "/dev/disk/by-path/";
11156
11157static const char *imsm_get_disk_controller_domain(const char *path)
11158{
2cda7640 11159 char disk_path[PATH_MAX];
96234762
LM		 11160 char *drv=NULL;
11161 struct stat st;
2cda7640 11162
6d8d290a 11163 strcpy(disk_path, disk_by_path);
96234762
LM		 11164 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11165 if (stat(disk_path, &st) == 0) {
11166 struct sys_dev* hba;
594dc1b8 11167 char *path;
96234762 11168
7c798f87 11169 path = devt_to_devpath(st.st_rdev, 1, NULL);
96234762
LM		 11170 if (path == NULL)
11171 return "unknown";
11172 hba = find_disk_attached_hba(-1, path);
11173 if (hba && hba->type == SYS_DEV_SAS)
11174 drv = "isci";
11175 else if (hba && hba->type == SYS_DEV_SATA)
11176 drv = "ahci";
c6839718
MT		 11177 else if (hba && hba->type == SYS_DEV_VMD)
11178 drv = "vmd";
11179 else if (hba && hba->type == SYS_DEV_NVME)
11180 drv = "nvme";
1011e834 11181 else
96234762
LM		 11182 drv = "unknown";
11183 dprintf("path: %s hba: %s attached: %s\n",
11184 path, (hba) ? hba->path : "NULL", drv);
11185 free(path);
2cda7640 11186 }
96234762 11187 return drv;
2cda7640
ML		 11188}
11189
4dd2df09 11190static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11191{
4dd2df09 11192 static char devnm[32];
78b10e66
N		 11193 char subdev_name[20];
11194 struct mdstat_ent *mdstat;
11195
11196 sprintf(subdev_name, "%d", subdev);
11197 mdstat = mdstat_by_subdev(subdev_name, container);
11198 if (!mdstat)
4dd2df09 11199 return NULL;
78b10e66 11200
4dd2df09 11201 strcpy(devnm, mdstat->devnm);
78b10e66 11202 free_mdstat(mdstat);
4dd2df09 11203 return devnm;
78b10e66
N		 11204}
11205
11206static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11207 struct geo_params *geo,
fbf3d202
AK		 11208 int *old_raid_disks,
11209 int direction)
78b10e66 11210{
694575e7
KW		 11211 /* currently we only support increasing the number of devices
11212 * for a container. This increases the number of device for each
11213 * member array. They must all be RAID0 or RAID5.
11214 */
78b10e66
N		 11215 int ret_val = 0;
11216 struct mdinfo *info, *member;
11217 int devices_that_can_grow = 0;
11218
7a862a02 11219 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11220
d04f65f4 11221 if (geo->size > 0 ||
78b10e66
N		 11222 geo->level != UnSet ||
11223 geo->layout != UnSet ||
11224 geo->chunksize != 0 ||
11225 geo->raid_disks == UnSet) {
7a862a02 11226 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11227 return ret_val;
11228 }
11229
fbf3d202 11230 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11231 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11232 return ret_val;
11233 }
11234
78b10e66
N		 11235 info = container_content_imsm(st, NULL);
11236 for (member = info; member; member = member->next) {
4dd2df09 11237 char *result;
78b10e66
N		 11238
11239 dprintf("imsm: checking device_num: %i\n",
11240 member->container_member);
11241
d7d205bd 11242 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11243 /* we work on container for Online Capacity Expansion
11244 * only so raid_disks has to grow
11245 */
7a862a02 11246 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11247 break;
11248 }
11249
089f9d79 11250 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11251 /* we cannot use this container with other raid level
11252 */
7a862a02 11253 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11254 info->array.level);
11255 break;
11256 } else {
11257 /* check for platform support
11258 * for this raid level configuration
11259 */
11260 struct intel_super *super = st->sb;
11261 if (!is_raid_level_supported(super->orom,
11262 member->array.level,
11263 geo->raid_disks)) {
7a862a02 11264 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11265 info->array.level,
11266 geo->raid_disks,
11267 geo->raid_disks > 1 ? "s" : "");
11268 break;
11269 }
2a4a08e7
AK		 11270 /* check if component size is aligned to chunk size
11271 */
11272 if (info->component_size %
11273 (info->array.chunk_size/512)) {
7a862a02 11274 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11275 break;
11276 }
78b10e66
N		 11277 }
11278
11279 if (*old_raid_disks &&
11280 info->array.raid_disks != *old_raid_disks)
11281 break;
11282 *old_raid_disks = info->array.raid_disks;
11283
11284 /* All raid5 and raid0 volumes in container
11285 * have to be ready for Online Capacity Expansion
11286 * so they need to be assembled. We have already
11287 * checked that no recovery etc is happening.
11288 */
4dd2df09
N		 11289 result = imsm_find_array_devnm_by_subdev(member->container_member,
11290 st->container_devnm);
11291 if (result == NULL) {
78b10e66
N		 11292 dprintf("imsm: cannot find array\n");
11293 break;
11294 }
11295 devices_that_can_grow++;
11296 }
11297 sysfs_free(info);
11298 if (!member && devices_that_can_grow)
11299 ret_val = 1;
11300
11301 if (ret_val)
1ade5cc1 11302 dprintf("Container operation allowed\n");
78b10e66 11303 else
1ade5cc1 11304 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11305
11306 return ret_val;
11307}
11308
11309/* Function: get_spares_for_grow
11310 * Description: Allocates memory and creates list of spare devices
1011e834 11311 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11312 * Parameters: Pointer to the supertype structure
11313 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11314 * NULL if fail
78b10e66
N		 11315 */
11316static struct mdinfo *get_spares_for_grow(struct supertype *st)
11317{
fbfdcb06
AO		 11318 struct spare_criteria sc;
11319
11320 get_spare_criteria_imsm(st, &sc);
11321 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11322}
11323
11324/******************************************************************************
11325 * function: imsm_create_metadata_update_for_reshape
11326 * Function creates update for whole IMSM container.
11327 *
11328 ******************************************************************************/
11329static int imsm_create_metadata_update_for_reshape(
11330 struct supertype *st,
11331 struct geo_params *geo,
11332 int old_raid_disks,
11333 struct imsm_update_reshape **updatep)
11334{
11335 struct intel_super *super = st->sb;
11336 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11337 int update_memory_size;
11338 struct imsm_update_reshape *u;
11339 struct mdinfo *spares;
78b10e66 11340 int i;
594dc1b8 11341 int delta_disks;
bbd24d86 11342 struct mdinfo *dev;
78b10e66 11343
1ade5cc1 11344 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11345
11346 delta_disks = geo->raid_disks - old_raid_disks;
11347
11348 /* size of all update data without anchor */
11349 update_memory_size = sizeof(struct imsm_update_reshape);
11350
11351 /* now add space for spare disks that we need to add. */
11352 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11353
503975b9 11354 u = xcalloc(1, update_memory_size);
78b10e66
N		 11355 u->type = update_reshape_container_disks;
11356 u->old_raid_disks = old_raid_disks;
11357 u->new_raid_disks = geo->raid_disks;
11358
11359 /* now get spare disks list
11360 */
11361 spares = get_spares_for_grow(st);
11362
d7be7d87 11363 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11364 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11365 i = -1;
78b10e66
N		 11366 goto abort;
11367 }
11368
11369 /* we have got spares
11370 * update disk list in imsm_disk list table in anchor
11371 */
11372 dprintf("imsm: %i spares are available.\n\n",
11373 spares->array.spare_disks);
11374
bbd24d86 11375 dev = spares->devs;
78b10e66 11376 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11377 struct dl *dl;
11378
bbd24d86
AK		 11379 if (dev == NULL)
11380 break;
78b10e66
N		 11381 u->new_disks[i] = makedev(dev->disk.major,
11382 dev->disk.minor);
11383 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11384 dl->index = mpb->num_disks;
11385 mpb->num_disks++;
bbd24d86 11386 dev = dev->next;
78b10e66 11387 }
78b10e66
N		 11388
11389abort:
11390 /* free spares
11391 */
11392 sysfs_free(spares);
11393
d677e0b8 11394 dprintf("imsm: reshape update preparation :");
78b10e66 11395 if (i == delta_disks) {
1ade5cc1 11396 dprintf_cont(" OK\n");
78b10e66
N		 11397 *updatep = u;
11398 return update_memory_size;
11399 }
11400 free(u);
1ade5cc1 11401 dprintf_cont(" Error\n");
78b10e66
N		 11402
11403 return 0;
11404}
11405
f3871fdc
AK		 11406/******************************************************************************
11407 * function: imsm_create_metadata_update_for_size_change()
11408 * Creates update for IMSM array for array size change.
11409 *
11410 ******************************************************************************/
11411static int imsm_create_metadata_update_for_size_change(
11412 struct supertype *st,
11413 struct geo_params *geo,
11414 struct imsm_update_size_change **updatep)
11415{
11416 struct intel_super *super = st->sb;
594dc1b8
JS		 11417 int update_memory_size;
11418 struct imsm_update_size_change *u;
f3871fdc 11419
1ade5cc1 11420 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11421
11422 /* size of all update data without anchor */
11423 update_memory_size = sizeof(struct imsm_update_size_change);
11424
503975b9 11425 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11426 u->type = update_size_change;
11427 u->subdev = super->current_vol;
11428 u->new_size = geo->size;
11429
11430 dprintf("imsm: reshape update preparation : OK\n");
11431 *updatep = u;
11432
11433 return update_memory_size;
11434}
11435
48c5303a
PC		 11436/******************************************************************************
11437 * function: imsm_create_metadata_update_for_migration()
11438 * Creates update for IMSM array.
11439 *
11440 ******************************************************************************/
11441static int imsm_create_metadata_update_for_migration(
11442 struct supertype *st,
11443 struct geo_params *geo,
11444 struct imsm_update_reshape_migration **updatep)
11445{
11446 struct intel_super *super = st->sb;
594dc1b8
JS		 11447 int update_memory_size;
11448 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11449 struct imsm_dev *dev;
11450 int previous_level = -1;
11451
1ade5cc1 11452 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11453
11454 /* size of all update data without anchor */
11455 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11456
503975b9 11457 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11458 u->type = update_reshape_migration;
11459 u->subdev = super->current_vol;
11460 u->new_level = geo->level;
11461 u->new_layout = geo->layout;
11462 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11463 u->new_disks[0] = -1;
4bba0439 11464 u->new_chunksize = -1;
48c5303a
PC		 11465
11466 dev = get_imsm_dev(super, u->subdev);
11467 if (dev) {
11468 struct imsm_map *map;
11469
238c0a71 11470 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11471 if (map) {
11472 int current_chunk_size =
11473 __le16_to_cpu(map->blocks_per_strip) / 2;
11474
11475 if (geo->chunksize != current_chunk_size) {
11476 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11477 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11478 current_chunk_size, u->new_chunksize);
11479 }
48c5303a 11480 previous_level = map->raid_level;
4bba0439 11481 }
48c5303a 11482 }
089f9d79 11483 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11484 struct mdinfo *spares = NULL;
11485
11486 u->new_raid_disks++;
11487 spares = get_spares_for_grow(st);
089f9d79 11488 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11489 free(u);
11490 sysfs_free(spares);
11491 update_memory_size = 0;
565cc99e 11492 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11493 return 0;
11494 }
11495 sysfs_free(spares);
11496 }
11497 dprintf("imsm: reshape update preparation : OK\n");
11498 *updatep = u;
11499
11500 return update_memory_size;
11501}
11502
8dd70bce
AK		 11503static void imsm_update_metadata_locally(struct supertype *st,
11504 void *buf, int len)
11505{
11506 struct metadata_update mu;
11507
11508 mu.buf = buf;
11509 mu.len = len;
11510 mu.space = NULL;
11511 mu.space_list = NULL;
11512 mu.next = NULL;
5fe6f031
N		 11513 if (imsm_prepare_update(st, &mu))
11514 imsm_process_update(st, &mu);
8dd70bce
AK		 11515
11516 while (mu.space_list) {
11517 void **space = mu.space_list;
11518 mu.space_list = *space;
11519 free(space);
11520 }
11521}
78b10e66 11522
471bceb6 11523/***************************************************************************
694575e7 11524* Function: imsm_analyze_change
471bceb6 11525* Description: Function analyze change for single volume
1011e834 11526* and validate if transition is supported
fbf3d202
AK		 11527* Parameters: Geometry parameters, supertype structure,
11528* metadata change direction (apply/rollback)
694575e7 11529* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11530****************************************************************************/
11531enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11532 struct geo_params *geo,
11533 int direction)
694575e7 11534{
471bceb6
KW		 11535 struct mdinfo info;
11536 int change = -1;
11537 int check_devs = 0;
c21e737b 11538 int chunk;
67a2db32
AK		 11539 /* number of added/removed disks in operation result */
11540 int devNumChange = 0;
11541 /* imsm compatible layout value for array geometry verification */
11542 int imsm_layout = -1;
7abc9871
AK		 11543 int data_disks;
11544 struct imsm_dev *dev;
9529d343 11545 struct imsm_map *map;
7abc9871 11546 struct intel_super *super;
d04f65f4 11547 unsigned long long current_size;
65d38cca 11548 unsigned long long free_size;
d04f65f4 11549 unsigned long long max_size;
65d38cca 11550 int rv;
471bceb6
KW		 11551
11552 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11553 if (geo->level != info.array.level && geo->level >= 0 &&
11554 geo->level != UnSet) {
471bceb6
KW		 11555 switch (info.array.level) {
11556 case 0:
11557 if (geo->level == 5) {
b5347799 11558 change = CH_MIGRATION;
e13ce846 11559 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11560 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11561 change = -1;
11562 goto analyse_change_exit;
11563 }
67a2db32 11564 imsm_layout = geo->layout;
471bceb6 11565 check_devs = 1;
e91a3bad
LM		 11566 devNumChange = 1; /* parity disk added */
11567 } else if (geo->level == 10) {
471bceb6
KW		 11568 change = CH_TAKEOVER;
11569 check_devs = 1;
e91a3bad 11570 devNumChange = 2; /* two mirrors added */
67a2db32 11571 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11572 }
dfe77a9e
KW		 11573 break;
11574 case 1:
471bceb6
KW		 11575 case 10:
11576 if (geo->level == 0) {
11577 change = CH_TAKEOVER;
11578 check_devs = 1;
e91a3bad 11579 devNumChange = -(geo->raid_disks/2);
67a2db32 11580 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11581 }
11582 break;
11583 }
11584 if (change == -1) {
7a862a02 11585 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11586 info.array.level, geo->level);
471bceb6
KW		 11587 goto analyse_change_exit;
11588 }
11589 } else
11590 geo->level = info.array.level;
11591
089f9d79
JS		 11592 if (geo->layout != info.array.layout &&
11593 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11594 change = CH_MIGRATION;
089f9d79
JS		 11595 if (info.array.layout == 0 && info.array.level == 5 &&
11596 geo->layout == 5) {
471bceb6 11597 /* reshape 5 -> 4 */
089f9d79
JS		 11598 } else if (info.array.layout == 5 && info.array.level == 5 &&
11599 geo->layout == 0) {
471bceb6
KW		 11600 /* reshape 4 -> 5 */
11601 geo->layout = 0;
11602 geo->level = 5;
11603 } else {
7a862a02 11604 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11605 info.array.layout, geo->layout);
471bceb6
KW		 11606 change = -1;
11607 goto analyse_change_exit;
11608 }
67a2db32 11609 } else {
471bceb6 11610 geo->layout = info.array.layout;
67a2db32
AK		 11611 if (imsm_layout == -1)
11612 imsm_layout = info.array.layout;
11613 }
471bceb6 11614
089f9d79
JS		 11615 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11616 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11617 if (info.array.level == 10) {
11618 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11619 change = -1;
11620 goto analyse_change_exit;
1e9b2c3f
PB		 11621 } else if (info.component_size % (geo->chunksize/512)) {
11622 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11623 geo->chunksize/1024, info.component_size/2);
11624 change = -1;
11625 goto analyse_change_exit;
2d2b0eb7 11626 }
b5347799 11627 change = CH_MIGRATION;
2d2b0eb7 11628 } else {
471bceb6 11629 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11630 }
471bceb6 11631
c21e737b 11632 chunk = geo->chunksize / 1024;
7abc9871
AK		 11633
11634 super = st->sb;
11635 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11636 map = get_imsm_map(dev, MAP_0);
11637 data_disks = imsm_num_data_members(map);
c41e00b2 11638 /* compute current size per disk member
7abc9871 11639 */
c41e00b2
AK		 11640 current_size = info.custom_array_size / data_disks;
11641
089f9d79 11642 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11643 /* align component size
11644 */
3e684231 11645 geo->size = imsm_component_size_alignment_check(
c41e00b2 11646 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11647 chunk * 1024, super->sector_size,
c41e00b2 11648 geo->size * 2);
65d0b4ce 11649 if (geo->size == 0) {
7a862a02 11650 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11651 current_size);
11652 goto analyse_change_exit;
11653 }
c41e00b2 11654 }
7abc9871 11655
089f9d79 11656 if (current_size != geo->size && geo->size > 0) {
7abc9871 11657 if (change != -1) {
7a862a02 11658 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11659 change = -1;
11660 goto analyse_change_exit;
11661 }
11662 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11663 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11664 super->current_vol, st->devnm);
7abc9871
AK		 11665 goto analyse_change_exit;
11666 }
65d38cca
LD		 11667 /* check the maximum available size
11668 */
11669 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11670 0, chunk, &free_size);
11671 if (rv == 0)
11672 /* Cannot find maximum available space
11673 */
11674 max_size = 0;
11675 else {
11676 max_size = free_size + current_size;
11677 /* align component size
11678 */
3e684231 11679 max_size = imsm_component_size_alignment_check(
65d38cca 11680 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11681 chunk * 1024, super->sector_size,
65d38cca
LD		 11682 max_size);
11683 }
d04f65f4 11684 if (geo->size == MAX_SIZE) {
b130333f
AK		 11685 /* requested size change to the maximum available size
11686 */
65d38cca 11687 if (max_size == 0) {
7a862a02 11688 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11689 change = -1;
11690 goto analyse_change_exit;
65d38cca
LD		 11691 } else
11692 geo->size = max_size;
c41e00b2 11693 }
b130333f 11694
681b7ae2 11695 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11696 /* accept size for rollback only
11697 */
11698 } else {
11699 /* round size due to metadata compatibility
11700 */
11701 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11702 << SECT_PER_MB_SHIFT;
11703 dprintf("Prepare update for size change to %llu\n",
11704 geo->size );
11705 if (current_size >= geo->size) {
7a862a02 11706 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11707 current_size, geo->size);
fbf3d202
AK		 11708 goto analyse_change_exit;
11709 }
65d38cca 11710 if (max_size && geo->size > max_size) {
7a862a02 11711 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11712 max_size, geo->size);
65d38cca
LD		 11713 goto analyse_change_exit;
11714 }
7abc9871
AK		 11715 }
11716 geo->size *= data_disks;
11717 geo->raid_disks = dev->vol.map->num_members;
11718 change = CH_ARRAY_SIZE;
11719 }
471bceb6
KW		 11720 if (!validate_geometry_imsm(st,
11721 geo->level,
67a2db32 11722 imsm_layout,
e91a3bad 11723 geo->raid_disks + devNumChange,
c21e737b 11724 &chunk,
af4348dd 11725 geo->size, INVALID_SECTORS,
5308f117 11726 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11727 change = -1;
11728
11729 if (check_devs) {
11730 struct intel_super *super = st->sb;
11731 struct imsm_super *mpb = super->anchor;
11732
11733 if (mpb->num_raid_devs > 1) {
7a862a02 11734 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11735 geo->dev_name);
471bceb6
KW		 11736 change = -1;
11737 }
11738 }
11739
11740analyse_change_exit:
089f9d79
JS		 11741 if (direction == ROLLBACK_METADATA_CHANGES &&
11742 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11743 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11744 change = -1;
11745 }
471bceb6 11746 return change;
694575e7
KW		 11747}
11748
bb025c2f
KW		 11749int imsm_takeover(struct supertype *st, struct geo_params *geo)
11750{
11751 struct intel_super *super = st->sb;
11752 struct imsm_update_takeover *u;
11753
503975b9 11754 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11755
11756 u->type = update_takeover;
11757 u->subarray = super->current_vol;
11758
11759 /* 10->0 transition */
11760 if (geo->level == 0)
11761 u->direction = R10_TO_R0;
11762
0529c688
KW		 11763 /* 0->10 transition */
11764 if (geo->level == 10)
11765 u->direction = R0_TO_R10;
11766
bb025c2f
KW		 11767 /* update metadata locally */
11768 imsm_update_metadata_locally(st, u,
11769 sizeof(struct imsm_update_takeover));
11770 /* and possibly remotely */
11771 if (st->update_tail)
11772 append_metadata_update(st, u,
11773 sizeof(struct imsm_update_takeover));
11774 else
11775 free(u);
11776
11777 return 0;
11778}
11779
895ffd99
MT		 11780/* Flush size update if size calculated by num_data_stripes is higher than
11781 * imsm_dev_size to eliminate differences during reshape.
11782 * Mdmon will recalculate them correctly.
11783 * If subarray index is not set then check whole container.
11784 * Returns:
11785 * 0 - no error occurred
11786 * 1 - error detected
11787 */
11788static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index)
11789{
11790 struct intel_super *super = st->sb;
11791 int tmp = super->current_vol;
11792 int ret_val = 1;
11793 int i;
11794
11795 for (i = 0; i < super->anchor->num_raid_devs; i++) {
11796 if (subarray_index >= 0 && i != subarray_index)
11797 continue;
11798 super->current_vol = i;
11799 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
11800 struct imsm_map *map = get_imsm_map(dev, MAP_0);
11801 unsigned int disc_count = imsm_num_data_members(map);
11802 struct geo_params geo;
11803 struct imsm_update_size_change *update;
11804 unsigned long long calc_size = per_dev_array_size(map) * disc_count;
11805 unsigned long long d_size = imsm_dev_size(dev);
11806 int u_size;
11807
11808 if (calc_size == d_size || dev->vol.migr_type == MIGR_GEN_MIGR)
11809 continue;
11810
ff904202
MT		 11811 /* There is a difference, confirm that imsm_dev_size is
11812 * smaller and push update.
895ffd99 11813 */
ff904202
MT		 11814 if (d_size > calc_size) {
11815 pr_err("imsm: dev size of subarray %d is incorrect\n",
11816 i);
895ffd99
MT		 11817 goto exit;
11818 }
11819 memset(&geo, 0, sizeof(struct geo_params));
11820 geo.size = d_size;
11821 u_size = imsm_create_metadata_update_for_size_change(st, &geo,
11822 &update);
11823 if (u_size < 1) {
11824 dprintf("imsm: Cannot prepare size change update\n");
11825 goto exit;
11826 }
11827 imsm_update_metadata_locally(st, update, u_size);
11828 if (st->update_tail) {
11829 append_metadata_update(st, update, u_size);
11830 flush_metadata_updates(st);
11831 st->update_tail = &st->updates;
11832 } else {
11833 imsm_sync_metadata(st);
11834 }
11835 }
11836 ret_val = 0;
11837exit:
11838 super->current_vol = tmp;
11839 return ret_val;
11840}
11841
d04f65f4
N		 11842static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11843 int level,
78b10e66 11844 int layout, int chunksize, int raid_disks,
41784c88 11845 int delta_disks, char *backup, char *dev,
016e00f5 11846 int direction, int verbose)
78b10e66 11847{
78b10e66
N		 11848 int ret_val = 1;
11849 struct geo_params geo;
11850
1ade5cc1 11851 dprintf("(enter)\n");
78b10e66 11852
71204a50 11853 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11854
11855 geo.dev_name = dev;
4dd2df09 11856 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11857 geo.size = size;
11858 geo.level = level;
11859 geo.layout = layout;
11860 geo.chunksize = chunksize;
11861 geo.raid_disks = raid_disks;
41784c88
AK		 11862 if (delta_disks != UnSet)
11863 geo.raid_disks += delta_disks;
78b10e66 11864
1ade5cc1
N		 11865 dprintf("for level : %i\n", geo.level);
11866 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11867
4dd2df09 11868 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11869 /* On container level we can only increase number of devices. */
11870 dprintf("imsm: info: Container operation\n");
78b10e66 11871 int old_raid_disks = 0;
6dc0be30 11872
78b10e66 11873 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11874 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11875 struct imsm_update_reshape *u = NULL;
11876 int len;
11877
895ffd99
MT		 11878 if (imsm_fix_size_mismatch(st, -1)) {
11879 dprintf("imsm: Cannot fix size mismatch\n");
11880 goto exit_imsm_reshape_super;
11881 }
11882
78b10e66
N		 11883 len = imsm_create_metadata_update_for_reshape(
11884 st, &geo, old_raid_disks, &u);
11885
ed08d51c
AK		 11886 if (len <= 0) {
11887 dprintf("imsm: Cannot prepare update\n");
11888 goto exit_imsm_reshape_super;
11889 }
11890
8dd70bce
AK		 11891 ret_val = 0;
11892 /* update metadata locally */
11893 imsm_update_metadata_locally(st, u, len);
11894 /* and possibly remotely */
11895 if (st->update_tail)
11896 append_metadata_update(st, u, len);
11897 else
ed08d51c 11898 free(u);
8dd70bce 11899
694575e7 11900 } else {
7a862a02 11901 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11902 }
11903 } else {
11904 /* On volume level we support following operations
471bceb6
KW		 11905 * - takeover: raid10 -> raid0; raid0 -> raid10
11906 * - chunk size migration
11907 * - migration: raid5 -> raid0; raid0 -> raid5
11908 */
11909 struct intel_super *super = st->sb;
11910 struct intel_dev *dev = super->devlist;
4dd2df09 11911 int change;
694575e7 11912 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11913 /* find requested device */
11914 while (dev) {
1011e834 11915 char *devnm =
4dd2df09
N		 11916 imsm_find_array_devnm_by_subdev(
11917 dev->index, st->container_devnm);
11918 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11919 break;
11920 dev = dev->next;
11921 }
11922 if (dev == NULL) {
4dd2df09
N		 11923 pr_err("Cannot find %s (%s) subarray\n",
11924 geo.dev_name, geo.devnm);
471bceb6
KW		 11925 goto exit_imsm_reshape_super;
11926 }
11927 super->current_vol = dev->index;
fbf3d202 11928 change = imsm_analyze_change(st, &geo, direction);
694575e7 11929 switch (change) {
471bceb6 11930 case CH_TAKEOVER:
bb025c2f 11931 ret_val = imsm_takeover(st, &geo);
694575e7 11932 break;
48c5303a
PC		 11933 case CH_MIGRATION: {
11934 struct imsm_update_reshape_migration *u = NULL;
11935 int len =
11936 imsm_create_metadata_update_for_migration(
11937 st, &geo, &u);
11938 if (len < 1) {
7a862a02 11939 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11940 break;
11941 }
471bceb6 11942 ret_val = 0;
48c5303a
PC		 11943 /* update metadata locally */
11944 imsm_update_metadata_locally(st, u, len);
11945 /* and possibly remotely */
11946 if (st->update_tail)
11947 append_metadata_update(st, u, len);
11948 else
11949 free(u);
11950 }
11951 break;
7abc9871 11952 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11953 struct imsm_update_size_change *u = NULL;
11954 int len =
11955 imsm_create_metadata_update_for_size_change(
11956 st, &geo, &u);
11957 if (len < 1) {
7a862a02 11958 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11959 break;
11960 }
11961 ret_val = 0;
11962 /* update metadata locally */
11963 imsm_update_metadata_locally(st, u, len);
11964 /* and possibly remotely */
11965 if (st->update_tail)
11966 append_metadata_update(st, u, len);
11967 else
11968 free(u);
7abc9871
AK		 11969 }
11970 break;
471bceb6
KW		 11971 default:
11972 ret_val = 1;
694575e7 11973 }
694575e7 11974 }
78b10e66 11975
ed08d51c 11976exit_imsm_reshape_super:
78b10e66
N		 11977 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11978 return ret_val;
11979}
2cda7640 11980
0febb20c
AO		 11981#define COMPLETED_OK 0
11982#define COMPLETED_NONE 1
11983#define COMPLETED_DELAYED 2
11984
11985static int read_completed(int fd, unsigned long long *val)
11986{
11987 int ret;
11988 char buf[50];
11989
11990 ret = sysfs_fd_get_str(fd, buf, 50);
11991 if (ret < 0)
11992 return ret;
11993
11994 ret = COMPLETED_OK;
11995 if (strncmp(buf, "none", 4) == 0) {
11996 ret = COMPLETED_NONE;
11997 } else if (strncmp(buf, "delayed", 7) == 0) {
11998 ret = COMPLETED_DELAYED;
11999 } else {
12000 char *ep;
12001 *val = strtoull(buf, &ep, 0);
12002 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
12003 ret = -1;
12004 }
12005 return ret;
12006}
12007
eee67a47
AK		 12008/*******************************************************************************
12009 * Function: wait_for_reshape_imsm
12010 * Description: Function writes new sync_max value and waits until
12011 * reshape process reach new position
12012 * Parameters:
12013 * sra : general array info
eee67a47
AK		 12014 * ndata : number of disks in new array's layout
12015 * Returns:
12016 * 0 : success,
12017 * 1 : there is no reshape in progress,
12018 * -1 : fail
12019 ******************************************************************************/
ae9f01f8 12020int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 12021{
85ca499c 12022 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 12023 int retry = 3;
eee67a47 12024 unsigned long long completed;
ae9f01f8
AK		 12025 /* to_complete : new sync_max position */
12026 unsigned long long to_complete = sra->reshape_progress;
12027 unsigned long long position_to_set = to_complete / ndata;
eee67a47 12028
ae9f01f8 12029 if (fd < 0) {
1ade5cc1 12030 dprintf("cannot open reshape_position\n");
eee67a47 12031 return 1;
ae9f01f8 12032 }
eee67a47 12033
df2647fa
PB		 12034 do {
12035 if (sysfs_fd_get_ll(fd, &completed) < 0) {
12036 if (!retry) {
12037 dprintf("cannot read reshape_position (no reshape in progres)\n");
12038 close(fd);
12039 return 1;
12040 }
12041 usleep(30000);
12042 } else
12043 break;
12044 } while (retry--);
eee67a47 12045
85ca499c 12046 if (completed > position_to_set) {
1ade5cc1 12047 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 12048 to_complete, position_to_set);
ae9f01f8
AK		 12049 close(fd);
12050 return -1;
12051 }
12052 dprintf("Position set: %llu\n", position_to_set);
12053 if (sysfs_set_num(sra, NULL, "sync_max",
12054 position_to_set) != 0) {
1ade5cc1 12055 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 12056 position_to_set);
12057 close(fd);
12058 return -1;
eee67a47
AK		 12059 }
12060
eee67a47 12061 do {
0febb20c 12062 int rc;
eee67a47 12063 char action[20];
5ff3a780 12064 int timeout = 3000;
0febb20c 12065
5ff3a780 12066 sysfs_wait(fd, &timeout);
a47e44fb
AK		 12067 if (sysfs_get_str(sra, NULL, "sync_action",
12068 action, 20) > 0 &&
d7d3809a 12069 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 12070 if (strncmp(action, "idle", 4) == 0)
12071 break;
d7d3809a
AP		 12072 close(fd);
12073 return -1;
12074 }
0febb20c
AO		 12075
12076 rc = read_completed(fd, &completed);
12077 if (rc < 0) {
1ade5cc1 12078 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 12079 close(fd);
12080 return 1;
0febb20c
AO		 12081 } else if (rc == COMPLETED_NONE)
12082 break;
85ca499c 12083 } while (completed < position_to_set);
b2be2b62 12084
eee67a47
AK		 12085 close(fd);
12086 return 0;
eee67a47
AK		 12087}
12088
b915c95f
AK		 12089/*******************************************************************************
12090 * Function: check_degradation_change
12091 * Description: Check that array hasn't become failed.
12092 * Parameters:
12093 * info : for sysfs access
12094 * sources : source disks descriptors
12095 * degraded: previous degradation level
12096 * Returns:
12097 * degradation level
12098 ******************************************************************************/
12099int check_degradation_change(struct mdinfo *info,
12100 int *sources,
12101 int degraded)
12102{
12103 unsigned long long new_degraded;
e1993023
LD		 12104 int rv;
12105
12106 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 12107 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 12108 /* check each device to ensure it is still working */
12109 struct mdinfo *sd;
12110 new_degraded = 0;
12111 for (sd = info->devs ; sd ; sd = sd->next) {
12112 if (sd->disk.state & (1<<MD_DISK_FAULTY))
12113 continue;
12114 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 12115 char sbuf[100];
12116
b915c95f 12117 if (sysfs_get_str(info,
cf52eff5 12118 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 12119 strstr(sbuf, "faulty") ||
12120 strstr(sbuf, "in_sync") == NULL) {
12121 /* this device is dead */
12122 sd->disk.state = (1<<MD_DISK_FAULTY);
12123 if (sd->disk.raid_disk >= 0 &&
12124 sources[sd->disk.raid_disk] >= 0) {
12125 close(sources[
12126 sd->disk.raid_disk]);
12127 sources[sd->disk.raid_disk] =
12128 -1;
12129 }
12130 new_degraded++;
12131 }
12132 }
12133 }
12134 }
12135
12136 return new_degraded;
12137}
12138
10f22854
AK		 12139/*******************************************************************************
12140 * Function: imsm_manage_reshape
12141 * Description: Function finds array under reshape and it manages reshape
12142 * process. It creates stripes backups (if required) and sets
942e1cdb 12143 * checkpoints.
10f22854
AK		 12144 * Parameters:
12145 * afd : Backup handle (nattive) - not used
12146 * sra : general array info
12147 * reshape : reshape parameters - not used
12148 * st : supertype structure
12149 * blocks : size of critical section [blocks]
12150 * fds : table of source device descriptor
12151 * offsets : start of array (offest per devices)
12152 * dests : not used
12153 * destfd : table of destination device descriptor
12154 * destoffsets : table of destination offsets (per device)
12155 * Returns:
12156 * 1 : success, reshape is done
12157 * 0 : fail
12158 ******************************************************************************/
999b4972
N		 12159static int imsm_manage_reshape(
12160 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12161 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12162 int *fds, unsigned long long *offsets,
12163 int dests, int *destfd, unsigned long long *destoffsets)
12164{
10f22854
AK		 12165 int ret_val = 0;
12166 struct intel_super *super = st->sb;
594dc1b8 12167 struct intel_dev *dv;
de44e46f 12168 unsigned int sector_size = super->sector_size;
10f22854 12169 struct imsm_dev *dev = NULL;
9529d343 12170 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12171 int migr_vol_qan = 0;
12172 int ndata, odata; /* [bytes] */
12173 int chunk; /* [bytes] */
12174 struct migr_record *migr_rec;
12175 char *buf = NULL;
12176 unsigned int buf_size; /* [bytes] */
12177 unsigned long long max_position; /* array size [bytes] */
12178 unsigned long long next_step; /* [blocks]/[bytes] */
12179 unsigned long long old_data_stripe_length;
10f22854
AK		 12180 unsigned long long start_src; /* [bytes] */
12181 unsigned long long start; /* [bytes] */
12182 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12183 int degraded = 0;
ab724b98 12184 int source_layout = 0;
895ffd99 12185 int subarray_index = -1;
10f22854 12186
79a16a9b
JS		 12187 if (!sra)
12188 return ret_val;
12189
12190 if (!fds || !offsets)
10f22854
AK		 12191 goto abort;
12192
12193 /* Find volume during the reshape */
12194 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12195 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12196 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12197 dev = dv->dev;
12198 migr_vol_qan++;
895ffd99 12199 subarray_index = dv->index;
10f22854
AK		 12200 }
12201 }
12202 /* Only one volume can migrate at the same time */
12203 if (migr_vol_qan != 1) {
676e87a8 12204 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12205 "Number of migrating volumes greater than 1\n" :
12206 "There is no volume during migrationg\n");
12207 goto abort;
12208 }
12209
9529d343 12210 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12211 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12212 if (map_src == NULL)
12213 goto abort;
10f22854 12214
9529d343
MD		 12215 ndata = imsm_num_data_members(map_dest);
12216 odata = imsm_num_data_members(map_src);
10f22854 12217
7b1ab482 12218 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12219 old_data_stripe_length = odata * chunk;
12220
12221 migr_rec = super->migr_rec;
12222
10f22854
AK		 12223 /* initialize migration record for start condition */
12224 if (sra->reshape_progress == 0)
12225 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12226 else {
12227 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12228 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12229 goto abort;
12230 }
6a75c8ca
AK		 12231 /* Save checkpoint to update migration record for current
12232 * reshape position (in md). It can be farther than current
12233 * reshape position in metadata.
12234 */
12235 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12236 /* ignore error == 2, this can mean end of reshape here
12237 */
7a862a02 12238 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12239 goto abort;
12240 }
b2c59438 12241 }
10f22854
AK		 12242
12243 /* size for data */
12244 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12245 /* extend buffer size for parity disk */
12246 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12247 /* add space for stripe alignment */
10f22854 12248 buf_size += old_data_stripe_length;
de44e46f
PB		 12249 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12250 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12251 goto abort;
12252 }
12253
3ef4403c 12254 max_position = sra->component_size * ndata;
68eb8bc6 12255 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12256
9f421827
PB		 12257 while (current_migr_unit(migr_rec) <
12258 get_num_migr_units(migr_rec)) {
10f22854
AK		 12259 /* current reshape position [blocks] */
12260 unsigned long long current_position =
12261 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12262 * current_migr_unit(migr_rec);
10f22854
AK		 12263 unsigned long long border;
12264
b915c95f
AK		 12265 /* Check that array hasn't become failed.
12266 */
12267 degraded = check_degradation_change(sra, fds, degraded);
12268 if (degraded > 1) {
7a862a02 12269 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12270 goto abort;
12271 }
12272
10f22854
AK		 12273 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12274
12275 if ((current_position + next_step) > max_position)
12276 next_step = max_position - current_position;
12277
92144abf 12278 start = current_position * 512;
10f22854 12279
942e1cdb 12280 /* align reading start to old geometry */
10f22854
AK		 12281 start_buf_shift = start % old_data_stripe_length;
12282 start_src = start - start_buf_shift;
12283
12284 border = (start_src / odata) - (start / ndata);
12285 border /= 512;
12286 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12287 /* save critical stripes to buf
12288 * start - start address of current unit
12289 * to backup [bytes]
12290 * start_src - start address of current unit
12291 * to backup alligned to source array
12292 * [bytes]
12293 */
594dc1b8 12294 unsigned long long next_step_filler;
10f22854
AK		 12295 unsigned long long copy_length = next_step * 512;
12296
12297 /* allign copy area length to stripe in old geometry */
12298 next_step_filler = ((copy_length + start_buf_shift)
12299 % old_data_stripe_length);
12300 if (next_step_filler)
12301 next_step_filler = (old_data_stripe_length
12302 - next_step_filler);
7a862a02 12303 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		 12304 start, start_src, copy_length,
12305 start_buf_shift, next_step_filler);
12306
12307 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12308 chunk, map_src->raid_level,
12309 source_layout, 0, NULL, start_src,
10f22854
AK		 12310 copy_length +
12311 next_step_filler + start_buf_shift,
12312 buf)) {
7a862a02 12313 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12314 goto abort;
12315 }
12316 /* Convert data to destination format and store it
12317 * in backup general migration area
12318 */
12319 if (save_backup_imsm(st, dev, sra,
aea93171 12320 buf + start_buf_shift, copy_length)) {
7a862a02 12321 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12322 goto abort;
12323 }
12324 if (save_checkpoint_imsm(st, sra,
12325 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12326 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12327 goto abort;
12328 }
8016a6d4
AK		 12329 } else {
12330 /* set next step to use whole border area */
12331 border /= next_step;
12332 if (border > 1)
12333 next_step *= border;
10f22854
AK		 12334 }
12335 /* When data backed up, checkpoint stored,
12336 * kick the kernel to reshape unit of data
12337 */
12338 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12339 /* limit next step to array max position */
12340 if (next_step > max_position)
12341 next_step = max_position;
10f22854
AK		 12342 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12343 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12344 sra->reshape_progress = next_step;
10f22854
AK		 12345
12346 /* wait until reshape finish */
c85338c6 12347 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12348 dprintf("wait_for_reshape_imsm returned error!\n");
12349 goto abort;
12350 }
84d11e6c
N		 12351 if (sigterm)
12352 goto abort;
10f22854 12353
0228d92c
AK		 12354 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12355 /* ignore error == 2, this can mean end of reshape here
12356 */
7a862a02 12357 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12358 goto abort;
12359 }
12360
12361 }
12362
71e5411e
PB		 12363 /* clear migr_rec on disks after successful migration */
12364 struct dl *d;
12365
85337573 12366 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12367 for (d = super->disks; d; d = d->next) {
12368 if (d->index < 0 || is_failed(&d->disk))
12369 continue;
12370 unsigned long long dsize;
12371
12372 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12373 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12374 SEEK_SET) >= 0) {
466070ad 12375 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12376 MIGR_REC_BUF_SECTORS*sector_size) !=
12377 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12378 perror("Write migr_rec failed");
12379 }
12380 }
12381
10f22854
AK		 12382 /* return '1' if done */
12383 ret_val = 1;
895ffd99
MT		 12384
12385 /* After the reshape eliminate size mismatch in metadata.
12386 * Don't update md/component_size here, volume hasn't
12387 * to take whole space. It is allowed by kernel.
12388 * md/component_size will be set propoperly after next assembly.
12389 */
12390 imsm_fix_size_mismatch(st, subarray_index);
12391
10f22854
AK		 12392abort:
12393 free(buf);
942e1cdb
N		 12394 /* See Grow.c: abort_reshape() for further explanation */
12395 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12396 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12397 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12398
12399 return ret_val;
999b4972 12400}
0c21b485 12401
fbc42556
JR		 12402/*******************************************************************************
12403 * Function: calculate_bitmap_min_chunksize
12404 * Description: Calculates the minimal valid bitmap chunk size
12405 * Parameters:
12406 * max_bits : indicate how many bits can be used for the bitmap
12407 * data_area_size : the size of the data area covered by the bitmap
12408 *
12409 * Returns:
12410 * The bitmap chunk size
12411 ******************************************************************************/
12412static unsigned long long
12413calculate_bitmap_min_chunksize(unsigned long long max_bits,
12414 unsigned long long data_area_size)
12415{
12416 unsigned long long min_chunk =
12417 4096; /* sub-page chunks don't work yet.. */
12418 unsigned long long bits = data_area_size / min_chunk + 1;
12419
12420 while (bits > max_bits) {
12421 min_chunk *= 2;
12422 bits = (bits + 1) / 2;
12423 }
12424 return min_chunk;
12425}
12426
12427/*******************************************************************************
12428 * Function: calculate_bitmap_chunksize
12429 * Description: Calculates the bitmap chunk size for the given device
12430 * Parameters:
12431 * st : supertype information
12432 * dev : device for the bitmap
12433 *
12434 * Returns:
12435 * The bitmap chunk size
12436 ******************************************************************************/
12437static unsigned long long calculate_bitmap_chunksize(struct supertype *st,
12438 struct imsm_dev *dev)
12439{
12440 struct intel_super *super = st->sb;
12441 unsigned long long min_chunksize;
12442 unsigned long long result = IMSM_DEFAULT_BITMAP_CHUNKSIZE;
12443 size_t dev_size = imsm_dev_size(dev);
12444
12445 min_chunksize = calculate_bitmap_min_chunksize(
12446 IMSM_BITMAP_AREA_SIZE * super->sector_size, dev_size);
12447
12448 if (result < min_chunksize)
12449 result = min_chunksize;
12450
12451 return result;
12452}
12453
12454/*******************************************************************************
12455 * Function: init_bitmap_header
12456 * Description: Initialize the bitmap header structure
12457 * Parameters:
12458 * st : supertype information
12459 * bms : bitmap header struct to initialize
12460 * dev : device for the bitmap
12461 *
12462 * Returns:
12463 * 0 : success
12464 * -1 : fail
12465 ******************************************************************************/
12466static int init_bitmap_header(struct supertype *st, struct bitmap_super_s *bms,
12467 struct imsm_dev *dev)
12468{
12469 int vol_uuid[4];
12470
12471 if (!bms || !dev)
12472 return -1;
12473
12474 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
12475 bms->version = __cpu_to_le32(BITMAP_MAJOR_HI);
12476 bms->daemon_sleep = __cpu_to_le32(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP);
12477 bms->sync_size = __cpu_to_le64(IMSM_BITMAP_AREA_SIZE);
12478 bms->write_behind = __cpu_to_le32(0);
12479
12480 uuid_from_super_imsm(st, vol_uuid);
12481 memcpy(bms->uuid, vol_uuid, 16);
12482
12483 bms->chunksize = calculate_bitmap_chunksize(st, dev);
12484
12485 return 0;
12486}
12487
12488/*******************************************************************************
12489 * Function: validate_internal_bitmap_for_drive
12490 * Description: Verify if the bitmap header for a given drive.
12491 * Parameters:
12492 * st : supertype information
12493 * offset : The offset from the beginning of the drive where to look for
12494 * the bitmap header.
12495 * d : the drive info
12496 *
12497 * Returns:
12498 * 0 : success
12499 * -1 : fail
12500 ******************************************************************************/
12501static int validate_internal_bitmap_for_drive(struct supertype *st,
12502 unsigned long long offset,
12503 struct dl *d)
12504{
12505 struct intel_super *super = st->sb;
12506 int ret = -1;
12507 int vol_uuid[4];
12508 bitmap_super_t *bms;
12509 int fd;
12510
12511 if (!d)
12512 return -1;
12513
12514 void *read_buf;
12515
12516 if (posix_memalign(&read_buf, MAX_SECTOR_SIZE, IMSM_BITMAP_HEADER_SIZE))
12517 return -1;
12518
12519 fd = d->fd;
12520 if (fd < 0) {
12521 fd = open(d->devname, O_RDONLY, 0);
12522 if (fd < 0) {
12523 dprintf("cannot open the device %s\n", d->devname);
12524 goto abort;
12525 }
12526 }
12527
12528 if (lseek64(fd, offset * super->sector_size, SEEK_SET) < 0)
12529 goto abort;
12530 if (read(fd, read_buf, IMSM_BITMAP_HEADER_SIZE) !=
12531 IMSM_BITMAP_HEADER_SIZE)
12532 goto abort;
12533
12534 uuid_from_super_imsm(st, vol_uuid);
12535
12536 bms = read_buf;
12537 if ((bms->magic != __cpu_to_le32(BITMAP_MAGIC)) ||
12538 (bms->version != __cpu_to_le32(BITMAP_MAJOR_HI)) ||
12539 (!same_uuid((int *)bms->uuid, vol_uuid, st->ss->swapuuid))) {
12540 dprintf("wrong bitmap header detected\n");
12541 goto abort;
12542 }
12543
12544 ret = 0;
12545abort:
12546 if ((d->fd < 0) && (fd >= 0))
12547 close(fd);
12548 if (read_buf)
12549 free(read_buf);
12550
12551 return ret;
12552}
12553
12554/*******************************************************************************
12555 * Function: validate_internal_bitmap_imsm
12556 * Description: Verify if the bitmap header is in place and with proper data.
12557 * Parameters:
12558 * st : supertype information
12559 *
12560 * Returns:
12561 * 0 : success or device w/o RWH_BITMAP
12562 * -1 : fail
12563 ******************************************************************************/
12564static int validate_internal_bitmap_imsm(struct supertype *st)
12565{
12566 struct intel_super *super = st->sb;
12567 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
12568 unsigned long long offset;
12569 struct dl *d;
12570
12571 if (!dev)
12572 return -1;
12573
12574 if (dev->rwh_policy != RWH_BITMAP)
12575 return 0;
12576
12577 offset = get_bitmap_header_sector(super, super->current_vol);
12578 for (d = super->disks; d; d = d->next) {
12579 if (d->index < 0 || is_failed(&d->disk))
12580 continue;
12581
12582 if (validate_internal_bitmap_for_drive(st, offset, d)) {
12583 pr_err("imsm: bitmap validation failed\n");
12584 return -1;
12585 }
12586 }
12587 return 0;
12588}
12589
12590/*******************************************************************************
12591 * Function: add_internal_bitmap_imsm
12592 * Description: Mark the volume to use the bitmap and updates the chunk size value.
12593 * Parameters:
12594 * st : supertype information
12595 * chunkp : bitmap chunk size
12596 * delay : not used for imsm
12597 * write_behind : not used for imsm
12598 * size : not used for imsm
12599 * may_change : not used for imsm
12600 * amajor : not used for imsm
12601 *
12602 * Returns:
12603 * 0 : success
12604 * -1 : fail
12605 ******************************************************************************/
12606static int add_internal_bitmap_imsm(struct supertype *st, int *chunkp,
12607 int delay, int write_behind,
12608 unsigned long long size, int may_change,
12609 int amajor)
12610{
12611 struct intel_super *super = st->sb;
12612 int vol_idx = super->current_vol;
12613 struct imsm_dev *dev;
12614
12615 if (!super->devlist || vol_idx == -1 || !chunkp)
12616 return -1;
12617
12618 dev = get_imsm_dev(super, vol_idx);
12619
12620 if (!dev) {
12621 dprintf("cannot find the device for volume index %d\n",
12622 vol_idx);
12623 return -1;
12624 }
12625 dev->rwh_policy = RWH_BITMAP;
12626
12627 *chunkp = calculate_bitmap_chunksize(st, dev);
12628
12629 return 0;
12630}
12631
12632/*******************************************************************************
12633 * Function: locate_bitmap_imsm
12634 * Description: Seek 'fd' to start of write-intent-bitmap.
12635 * Parameters:
12636 * st : supertype information
12637 * fd : file descriptor for the device
12638 * node_num : not used for imsm
12639 *
12640 * Returns:
12641 * 0 : success
12642 * -1 : fail
12643 ******************************************************************************/
12644static int locate_bitmap_imsm(struct supertype *st, int fd, int node_num)
12645{
12646 struct intel_super *super = st->sb;
12647 unsigned long long offset;
12648 int vol_idx = super->current_vol;
12649
12650 if (!super->devlist || vol_idx == -1)
12651 return -1;
12652
12653 offset = get_bitmap_header_sector(super, super->current_vol);
12654 dprintf("bitmap header offset is %llu\n", offset);
12655
12656 lseek64(fd, offset << 9, 0);
12657
12658 return 0;
12659}
12660
12661/*******************************************************************************
12662 * Function: write_init_bitmap_imsm
12663 * Description: Write a bitmap header and prepares the area for the bitmap.
12664 * Parameters:
12665 * st : supertype information
12666 * fd : file descriptor for the device
12667 * update : not used for imsm
12668 *
12669 * Returns:
12670 * 0 : success
12671 * -1 : fail
12672 ******************************************************************************/
12673static int write_init_bitmap_imsm(struct supertype *st, int fd,
12674 enum bitmap_update update)
12675{
12676 struct intel_super *super = st->sb;
12677 int vol_idx = super->current_vol;
12678 int ret = 0;
12679 unsigned long long offset;
12680 bitmap_super_t bms = { 0 };
12681 size_t written = 0;
12682 size_t to_write;
12683 ssize_t rv_num;
12684 void *buf;
12685
12686 if (!super->devlist || !super->sector_size || vol_idx == -1)
12687 return -1;
12688
12689 struct imsm_dev *dev = get_imsm_dev(super, vol_idx);
12690
12691 /* first clear the space for bitmap header */
12692 unsigned long long bitmap_area_start =
12693 get_bitmap_header_sector(super, vol_idx);
12694
12695 dprintf("zeroing area start (%llu) and size (%u)\n", bitmap_area_start,
12696 IMSM_BITMAP_AND_HEADER_SIZE / super->sector_size);
12697 if (zero_disk_range(fd, bitmap_area_start,
12698 IMSM_BITMAP_HEADER_SIZE / super->sector_size)) {
12699 pr_err("imsm: cannot zeroing the space for the bitmap\n");
12700 return -1;
12701 }
12702
12703 /* The bitmap area should be filled with "1"s to perform initial
12704 * synchronization.
12705 */
12706 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE))
12707 return -1;
12708 memset(buf, 0xFF, MAX_SECTOR_SIZE);
12709 offset = get_bitmap_sector(super, vol_idx);
12710 lseek64(fd, offset << 9, 0);
12711 while (written < IMSM_BITMAP_AREA_SIZE) {
12712 to_write = IMSM_BITMAP_AREA_SIZE - written;
12713 if (to_write > MAX_SECTOR_SIZE)
12714 to_write = MAX_SECTOR_SIZE;
12715 rv_num = write(fd, buf, MAX_SECTOR_SIZE);
12716 if (rv_num != MAX_SECTOR_SIZE) {
12717 ret = -1;
12718 dprintf("cannot initialize bitmap area\n");
12719 goto abort;
12720 }
12721 written += rv_num;
12722 }
12723
12724 /* write a bitmap header */
12725 init_bitmap_header(st, &bms, dev);
12726 memset(buf, 0, MAX_SECTOR_SIZE);
12727 memcpy(buf, &bms, sizeof(bitmap_super_t));
12728 if (locate_bitmap_imsm(st, fd, 0)) {
12729 ret = -1;
12730 dprintf("cannot locate the bitmap\n");
12731 goto abort;
12732 }
12733 if (write(fd, buf, MAX_SECTOR_SIZE) != MAX_SECTOR_SIZE) {
12734 ret = -1;
12735 dprintf("cannot write the bitmap header\n");
12736 goto abort;
12737 }
12738 fsync(fd);
12739
12740abort:
12741 free(buf);
12742
12743 return ret;
12744}
12745
12746/*******************************************************************************
12747 * Function: is_vol_to_setup_bitmap
12748 * Description: Checks if a bitmap should be activated on the dev.
12749 * Parameters:
12750 * info : info about the volume to setup the bitmap
12751 * dev : the device to check against bitmap creation
12752 *
12753 * Returns:
12754 * 0 : bitmap should be set up on the device
12755 * -1 : otherwise
12756 ******************************************************************************/
12757static int is_vol_to_setup_bitmap(struct mdinfo *info, struct imsm_dev *dev)
12758{
12759 if (!dev || !info)
12760 return -1;
12761
12762 if ((strcmp((char *)dev->volume, info->name) == 0) &&
12763 (dev->rwh_policy == RWH_BITMAP))
12764 return -1;
12765
12766 return 0;
12767}
12768
12769/*******************************************************************************
12770 * Function: set_bitmap_sysfs
12771 * Description: Set the sysfs atributes of a given volume to activate the bitmap.
12772 * Parameters:
12773 * info : info about the volume where the bitmap should be setup
12774 * chunksize : bitmap chunk size
12775 * location : location of the bitmap
12776 *
12777 * Returns:
12778 * 0 : success
12779 * -1 : fail
12780 ******************************************************************************/
12781static int set_bitmap_sysfs(struct mdinfo *info, unsigned long long chunksize,
12782 char *location)
12783{
12784 /* The bitmap/metadata is set to external to allow changing of value for
12785 * bitmap/location. When external is used, the kernel will treat an offset
12786 * related to the device's first lba (in opposition to the "internal" case
12787 * when this value is related to the beginning of the superblock).
12788 */
12789 if (sysfs_set_str(info, NULL, "bitmap/metadata", "external")) {
12790 dprintf("failed to set bitmap/metadata\n");
12791 return -1;
12792 }
12793
12794 /* It can only be changed when no bitmap is active.
12795 * Should be bigger than 512 and must be power of 2.
12796 * It is expecting the value in bytes.
12797 */
12798 if (sysfs_set_num(info, NULL, "bitmap/chunksize",
12799 __cpu_to_le32(chunksize))) {
12800 dprintf("failed to set bitmap/chunksize\n");
12801 return -1;
12802 }
12803
12804 /* It is expecting the value in sectors. */
12805 if (sysfs_set_num(info, NULL, "bitmap/space",
12806 __cpu_to_le64(IMSM_BITMAP_AREA_SIZE))) {
12807 dprintf("failed to set bitmap/space\n");
12808 return -1;
12809 }
12810
12811 /* Determines the delay between the bitmap updates.
12812 * It is expecting the value in seconds.
12813 */
12814 if (sysfs_set_num(info, NULL, "bitmap/time_base",
12815 __cpu_to_le64(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP))) {
12816 dprintf("failed to set bitmap/time_base\n");
12817 return -1;
12818 }
12819
12820 /* It is expecting the value in sectors with a sign at the beginning. */
12821 if (sysfs_set_str(info, NULL, "bitmap/location", location)) {
12822 dprintf("failed to set bitmap/location\n");
12823 return -1;
12824 }
12825
12826 return 0;
12827}
12828
12829/*******************************************************************************
12830 * Function: set_bitmap_imsm
12831 * Description: Setup the bitmap for the given volume
12832 * Parameters:
12833 * st : supertype information
12834 * info : info about the volume where the bitmap should be setup
12835 *
12836 * Returns:
12837 * 0 : success
12838 * -1 : fail
12839 ******************************************************************************/
12840static int set_bitmap_imsm(struct supertype *st, struct mdinfo *info)
12841{
12842 struct intel_super *super = st->sb;
12843 int prev_current_vol = super->current_vol;
12844 struct imsm_dev *dev;
12845 int ret = -1;
12846 char location[16] = "";
12847 unsigned long long chunksize;
12848 struct intel_dev *dev_it;
12849
12850 for (dev_it = super->devlist; dev_it; dev_it = dev_it->next) {
12851 super->current_vol = dev_it->index;
12852 dev = get_imsm_dev(super, super->current_vol);
12853
12854 if (is_vol_to_setup_bitmap(info, dev)) {
12855 if (validate_internal_bitmap_imsm(st)) {
12856 dprintf("bitmap header validation failed\n");
12857 goto abort;
12858 }
12859
12860 chunksize = calculate_bitmap_chunksize(st, dev);
12861 dprintf("chunk size is %llu\n", chunksize);
12862
12863 snprintf(location, sizeof(location), "+%llu",
12864 get_bitmap_sector(super, super->current_vol));
12865 dprintf("bitmap offset is %s\n", location);
12866
12867 if (set_bitmap_sysfs(info, chunksize, location)) {
12868 dprintf("cannot setup the bitmap\n");
12869 goto abort;
12870 }
12871 }
12872 }
12873 ret = 0;
12874abort:
12875 super->current_vol = prev_current_vol;
12876 return ret;
12877}
12878
cdddbdbc 12879struct superswitch super_imsm = {
cdddbdbc
DW		 12880 .examine_super = examine_super_imsm,
12881 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12882 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12883 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12884 .detail_super = detail_super_imsm,
12885 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12886 .write_init_super = write_init_super_imsm,
0e600426
N		 12887 .validate_geometry = validate_geometry_imsm,
12888 .add_to_super = add_to_super_imsm,
1a64be56 12889 .remove_from_super = remove_from_super_imsm,
d665cc31 12890 .detail_platform = detail_platform_imsm,
e50cf220 12891 .export_detail_platform = export_detail_platform_imsm,
33414a01 12892 .kill_subarray = kill_subarray_imsm,
aa534678 12893 .update_subarray = update_subarray_imsm,
2b959fbf 12894 .load_container = load_container_imsm,
71204a50
N		 12895 .default_geometry = default_geometry_imsm,
12896 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12897 .reshape_super = imsm_reshape_super,
12898 .manage_reshape = imsm_manage_reshape,
9e2d750d 12899 .recover_backup = recover_backup_imsm,
27156a57 12900 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12901 .match_home = match_home_imsm,
12902 .uuid_from_super= uuid_from_super_imsm,
12903 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12904 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12905 .update_super = update_super_imsm,
12906
12907 .avail_size = avail_size_imsm,
fbfdcb06 12908 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12909
12910 .compare_super = compare_super_imsm,
12911
12912 .load_super = load_super_imsm,
bf5a934a 12913 .init_super = init_super_imsm,
e683ca88 12914 .store_super = store_super_imsm,
cdddbdbc
DW		 12915 .free_super = free_super_imsm,
12916 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12917 .container_content = container_content_imsm,
0c21b485 12918 .validate_container = validate_container_imsm,
cdddbdbc 12919
fbc42556
JR		 12920 .add_internal_bitmap = add_internal_bitmap_imsm,
12921 .locate_bitmap = locate_bitmap_imsm,
12922 .write_bitmap = write_init_bitmap_imsm,
12923 .set_bitmap = set_bitmap_imsm,
12924
2432ce9b
AP		 12925 .write_init_ppl = write_init_ppl_imsm,
12926 .validate_ppl = validate_ppl_imsm,
12927
cdddbdbc 12928 .external = 1,
4cce4069 12929 .name = "imsm",
845dea95
NB		 12930
12931/* for mdmon */
12932 .open_new = imsm_open_new,
ed9d66aa 12933 .set_array_state= imsm_set_array_state,
845dea95
NB		 12934 .set_disk = imsm_set_disk,
12935 .sync_metadata = imsm_sync_metadata,
88758e9d 12936 .activate_spare = imsm_activate_spare,
e8319a19 12937 .process_update = imsm_process_update,
8273f55e 12938 .prepare_update = imsm_prepare_update,
6f50473f 12939 .record_bad_block = imsm_record_badblock,
c07a5a4f 12940 .clear_bad_block = imsm_clear_badblock,
928f1424 12941 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12942};
Unnamed repository; edit this file 'description' to name the repository.