]> 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
policy.c: Fix for compiler error
[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
DW		 98
99/* Disk configuration info. */
100#define IMSM_MAX_DEVICES 255
101struct imsm_disk {
102 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 103 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 104 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 105#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
106#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
107#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 108#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 109 __u32 status; /* 0xF0 - 0xF3 */
1011e834 110 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 111 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
112#define IMSM_DISK_FILLERS 3
113 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc
DW		 114};
115
3b451610
AK		 116/* map selector for map managment
117 */
238c0a71
AK		 118#define MAP_0 0
119#define MAP_1 1
120#define MAP_X -1
3b451610 121
cdddbdbc
DW		 122/* RAID map configuration infos. */
123struct imsm_map {
5551b113
CA		 124 __u32 pba_of_lba0_lo; /* start address of partition */
125 __u32 blocks_per_member_lo;/* blocks per member */
126 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 127 __u16 blocks_per_strip;
128 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
129#define IMSM_T_STATE_NORMAL 0
130#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 131#define IMSM_T_STATE_DEGRADED 2
132#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 133 __u8 raid_level;
134#define IMSM_T_RAID0 0
135#define IMSM_T_RAID1 1
136#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
137 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 138 __u8 num_domains; /* number of parity domains */
139 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 140 __u8 ddf;
5551b113
CA		 141 __u32 pba_of_lba0_hi;
142 __u32 blocks_per_member_hi;
143 __u32 num_data_stripes_hi;
144 __u32 filler[4]; /* expansion area */
7eef0453 145#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 146 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 147 * top byte contains some flags
148 */
cdddbdbc
DW		 149} __attribute__ ((packed));
150
151struct imsm_vol {
f8f603f1 152 __u32 curr_migr_unit;
fe7ed8cb 153 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 154 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 155#define MIGR_INIT 0
156#define MIGR_REBUILD 1
157#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
158#define MIGR_GEN_MIGR 3
159#define MIGR_STATE_CHANGE 4
1484e727 160#define MIGR_REPAIR 5
cdddbdbc 161 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 162#define RAIDVOL_CLEAN 0
163#define RAIDVOL_DIRTY 1
164#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 165 __u8 dirty;
fe7ed8cb
DW		 166 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
167 __u16 verify_errors; /* number of mismatches */
168 __u16 bad_blocks; /* number of bad blocks during verify */
169 __u32 filler[4];
cdddbdbc
DW		 170 struct imsm_map map[1];
171 /* here comes another one if migr_state */
172} __attribute__ ((packed));
173
174struct imsm_dev {
fe7ed8cb 175 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 176 __u32 size_low;
177 __u32 size_high;
fe7ed8cb
DW		 178#define DEV_BOOTABLE __cpu_to_le32(0x01)
179#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
180#define DEV_READ_COALESCING __cpu_to_le32(0x04)
181#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
182#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
183#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
184#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
185#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
186#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
187#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
188#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
189#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
190#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 191 __u32 status; /* Persistent RaidDev status */
192 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 193 __u8 migr_priority;
194 __u8 num_sub_vols;
195 __u8 tid;
196 __u8 cng_master_disk;
197 __u16 cache_policy;
198 __u8 cng_state;
199 __u8 cng_sub_state;
2432ce9b
AP		 200 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
201
202 /* NVM_EN */
203 __u8 nv_cache_mode;
204 __u8 nv_cache_flags;
205
206 /* Unique Volume Id of the NvCache Volume associated with this volume */
207 __u32 nvc_vol_orig_family_num;
208 __u16 nvc_vol_raid_dev_num;
209
210#define RWH_OFF 0
211#define RWH_DISTRIBUTED 1
212#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 213#define RWH_MULTIPLE_DISTRIBUTED 3
214#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
215#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 216 __u8 rwh_policy; /* Raid Write Hole Policy */
217 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
218 __u8 filler1;
219
220#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 221 __u32 filler[IMSM_DEV_FILLERS];
222 struct imsm_vol vol;
223} __attribute__ ((packed));
224
225struct imsm_super {
226 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
227 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
228 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
229 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
230 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 231 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
232 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 233 __u8 num_disks; /* 0x38 Number of configured disks */
234 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 235 __u8 error_log_pos; /* 0x3A */
236 __u8 fill[1]; /* 0x3B */
237 __u32 cache_size; /* 0x3c - 0x40 in mb */
238 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
239 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
240 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 241 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
242 * volume IDs for raid_dev created in this array
243 * (starts at 1)
244 */
245 __u16 filler1; /* 0x4E - 0x4F */
246#define IMSM_FILLERS 34
247 __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 248 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
249 /* here comes imsm_dev[num_raid_devs] */
604b746f 250 /* here comes BBM logs */
cdddbdbc
DW		 251} __attribute__ ((packed));
252
604b746f 253#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 254#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
255#define BBM_LOG_SIGNATURE 0xabadb10c
256
257struct bbm_log_block_addr {
258 __u16 w1;
259 __u32 dw1;
260} __attribute__ ((__packed__));
604b746f
JD		 261
262struct bbm_log_entry {
8d67477f
TM		 263 __u8 marked_count; /* Number of blocks marked - 1 */
264 __u8 disk_ordinal; /* Disk entry within the imsm_super */
265 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 266} __attribute__ ((__packed__));
267
268struct bbm_log {
269 __u32 signature; /* 0xABADB10C */
270 __u32 entry_count;
8d67477f 271 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
604b746f
JD		 272} __attribute__ ((__packed__));
273
cdddbdbc 274static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 275
b53bfba6
TM		 276#define BLOCKS_PER_KB (1024/512)
277
8e59f3d8
AK		 278#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
279
280#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
281
de44e46f
PB		 282#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
283#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
284 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 285 */
286
8e59f3d8
AK		 287#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
288 * be recovered using srcMap */
289#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
290 * already been migrated and must
291 * be recovered from checkpoint area */
2432ce9b 292
c2462068 293#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 294
8e59f3d8
AK		 295struct migr_record {
296 __u32 rec_status; /* Status used to determine how to restart
297 * migration in case it aborts
298 * in some fashion */
299 __u32 curr_migr_unit; /* 0..numMigrUnits-1 */
300 __u32 family_num; /* Family number of MPB
301 * containing the RaidDev
302 * that is migrating */
303 __u32 ascending_migr; /* True if migrating in increasing
304 * order of lbas */
305 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
306 __u32 dest_depth_per_unit; /* Num member blocks each destMap
307 * member disk
308 * advances per unit-of-operation */
309 __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */
310 __u32 dest_1st_member_lba; /* First member lba on first
311 * stripe of destination */
312 __u32 num_migr_units; /* Total num migration units-of-op */
313 __u32 post_migr_vol_cap; /* Size of volume after
314 * migration completes */
315 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
316 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
317 * migration ckpt record was read from
318 * (for recovered migrations) */
319} __attribute__ ((__packed__));
320
ec50f7b6
LM		 321struct md_list {
322 /* usage marker:
323 * 1: load metadata
324 * 2: metadata does not match
325 * 4: already checked
326 */
327 int used;
328 char *devname;
329 int found;
330 int container;
331 dev_t st_rdev;
332 struct md_list *next;
333};
334
e7b84f9d 335#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 336
1484e727
DW		 337static __u8 migr_type(struct imsm_dev *dev)
338{
339 if (dev->vol.migr_type == MIGR_VERIFY &&
340 dev->status & DEV_VERIFY_AND_FIX)
341 return MIGR_REPAIR;
342 else
343 return dev->vol.migr_type;
344}
345
346static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
347{
348 /* for compatibility with older oroms convert MIGR_REPAIR, into
349 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
350 */
351 if (migr_type == MIGR_REPAIR) {
352 dev->vol.migr_type = MIGR_VERIFY;
353 dev->status |= DEV_VERIFY_AND_FIX;
354 } else {
355 dev->vol.migr_type = migr_type;
356 dev->status &= ~DEV_VERIFY_AND_FIX;
357 }
358}
359
f36a9ecd 360static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 361{
f36a9ecd 362 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 363}
cdddbdbc 364
f36a9ecd
PB		 365static unsigned int mpb_sectors(struct imsm_super *mpb,
366 unsigned int sector_size)
87eb16df 367{
f36a9ecd 368 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 369}
370
ba2de7ba
DW		 371struct intel_dev {
372 struct imsm_dev *dev;
373 struct intel_dev *next;
f21e18ca 374 unsigned index;
ba2de7ba
DW		 375};
376
88654014
LM		 377struct intel_hba {
378 enum sys_dev_type type;
379 char *path;
380 char *pci_id;
381 struct intel_hba *next;
382};
383
1a64be56
LM		 384enum action {
385 DISK_REMOVE = 1,
386 DISK_ADD
387};
cdddbdbc
DW		 388/* internal representation of IMSM metadata */
389struct intel_super {
390 union {
949c47a0
DW		 391 void *buf; /* O_DIRECT buffer for reading/writing metadata */
392 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 393 };
8e59f3d8
AK		 394 union {
395 void *migr_rec_buf; /* buffer for I/O operations */
396 struct migr_record *migr_rec; /* migration record */
397 };
51d83f5d
AK		 398 int clean_migration_record_by_mdmon; /* when reshape is switched to next
399 array, it indicates that mdmon is allowed to clean migration
400 record */
949c47a0 401 size_t len; /* size of the 'buf' allocation */
bbab0940 402 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 403 void *next_buf; /* for realloc'ing buf from the manager */
404 size_t next_len;
c2c087e6 405 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 406 int current_vol; /* index of raid device undergoing creation */
5551b113 407 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 408 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 409 struct intel_dev *devlist;
fa7bb6f8 410 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 411 struct dl {
412 struct dl *next;
413 int index;
414 __u8 serial[MAX_RAID_SERIAL_LEN];
415 int major, minor;
416 char *devname;
b9f594fe 417 struct imsm_disk disk;
cdddbdbc 418 int fd;
0dcecb2e
DW		 419 int extent_cnt;
420 struct extent *e; /* for determining freespace @ create */
efb30e7f 421 int raiddisk; /* slot to fill in autolayout */
1a64be56 422 enum action action;
ca0748fa 423 } *disks, *current_disk;
1a64be56
LM		 424 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
425 active */
47ee5a45 426 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 427 struct bbm_log *bbm_log;
88654014 428 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 429 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 430 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 431 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 432};
433
434struct intel_disk {
435 struct imsm_disk disk;
436 #define IMSM_UNKNOWN_OWNER (-1)
437 int owner;
438 struct intel_disk *next;
cdddbdbc
DW		 439};
440
c2c087e6
DW		 441struct extent {
442 unsigned long long start, size;
443};
444
694575e7
KW		 445/* definitions of reshape process types */
446enum imsm_reshape_type {
447 CH_TAKEOVER,
b5347799 448 CH_MIGRATION,
7abc9871 449 CH_ARRAY_SIZE,
694575e7
KW		 450};
451
88758e9d
DW		 452/* definition of messages passed to imsm_process_update */
453enum imsm_update_type {
454 update_activate_spare,
8273f55e 455 update_create_array,
33414a01 456 update_kill_array,
aa534678 457 update_rename_array,
1a64be56 458 update_add_remove_disk,
78b10e66 459 update_reshape_container_disks,
48c5303a 460 update_reshape_migration,
2d40f3a1
AK		 461 update_takeover,
462 update_general_migration_checkpoint,
f3871fdc 463 update_size_change,
bbab0940 464 update_prealloc_badblocks_mem,
e6e9dd3f 465 update_rwh_policy,
88758e9d
DW		 466};
467
468struct imsm_update_activate_spare {
469 enum imsm_update_type type;
d23fe947 470 struct dl *dl;
88758e9d
DW		 471 int slot;
472 int array;
473 struct imsm_update_activate_spare *next;
474};
475
78b10e66 476struct geo_params {
4dd2df09 477 char devnm[32];
78b10e66 478 char *dev_name;
d04f65f4 479 unsigned long long size;
78b10e66
N		 480 int level;
481 int layout;
482 int chunksize;
483 int raid_disks;
484};
485
bb025c2f
KW		 486enum takeover_direction {
487 R10_TO_R0,
488 R0_TO_R10
489};
490struct imsm_update_takeover {
491 enum imsm_update_type type;
492 int subarray;
493 enum takeover_direction direction;
494};
78b10e66
N		 495
496struct imsm_update_reshape {
497 enum imsm_update_type type;
498 int old_raid_disks;
499 int new_raid_disks;
48c5303a
PC		 500
501 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
502};
503
504struct imsm_update_reshape_migration {
505 enum imsm_update_type type;
506 int old_raid_disks;
507 int new_raid_disks;
508 /* fields for array migration changes
509 */
510 int subdev;
511 int new_level;
512 int new_layout;
4bba0439 513 int new_chunksize;
48c5303a 514
d195167d 515 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 516};
517
f3871fdc
AK		 518struct imsm_update_size_change {
519 enum imsm_update_type type;
520 int subdev;
521 long long new_size;
522};
523
2d40f3a1
AK		 524struct imsm_update_general_migration_checkpoint {
525 enum imsm_update_type type;
526 __u32 curr_migr_unit;
527};
528
54c2c1ea
DW		 529struct disk_info {
530 __u8 serial[MAX_RAID_SERIAL_LEN];
531};
532
8273f55e
DW		 533struct imsm_update_create_array {
534 enum imsm_update_type type;
8273f55e 535 int dev_idx;
6a3e913e 536 struct imsm_dev dev;
8273f55e
DW		 537};
538
33414a01
DW		 539struct imsm_update_kill_array {
540 enum imsm_update_type type;
541 int dev_idx;
542};
543
aa534678
DW		 544struct imsm_update_rename_array {
545 enum imsm_update_type type;
546 __u8 name[MAX_RAID_SERIAL_LEN];
547 int dev_idx;
548};
549
1a64be56 550struct imsm_update_add_remove_disk {
43dad3d6
DW		 551 enum imsm_update_type type;
552};
553
bbab0940
TM		 554struct imsm_update_prealloc_bb_mem {
555 enum imsm_update_type type;
556};
557
e6e9dd3f
AP		 558struct imsm_update_rwh_policy {
559 enum imsm_update_type type;
560 int new_policy;
561 int dev_idx;
562};
563
88654014
LM		 564static const char *_sys_dev_type[] = {
565 [SYS_DEV_UNKNOWN] = "Unknown",
566 [SYS_DEV_SAS] = "SAS",
614902f6 567 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 568 [SYS_DEV_NVME] = "NVMe",
569 [SYS_DEV_VMD] = "VMD"
88654014
LM		 570};
571
572const char *get_sys_dev_type(enum sys_dev_type type)
573{
574 if (type >= SYS_DEV_MAX)
575 type = SYS_DEV_UNKNOWN;
576
577 return _sys_dev_type[type];
578}
579
580static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
581{
503975b9
N		 582 struct intel_hba *result = xmalloc(sizeof(*result));
583
584 result->type = device->type;
585 result->path = xstrdup(device->path);
586 result->next = NULL;
587 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
588 result->pci_id++;
589
88654014
LM		 590 return result;
591}
592
593static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
594{
594dc1b8
JS		 595 struct intel_hba *result;
596
88654014
LM		 597 for (result = hba; result; result = result->next) {
598 if (result->type == device->type && strcmp(result->path, device->path) == 0)
599 break;
600 }
601 return result;
602}
603
b4cf4cba 604static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 605{
606 struct intel_hba *hba;
607
608 /* check if disk attached to Intel HBA */
609 hba = find_intel_hba(super->hba, device);
610 if (hba != NULL)
611 return 1;
612 /* Check if HBA is already attached to super */
613 if (super->hba == NULL) {
614 super->hba = alloc_intel_hba(device);
615 return 1;
6b781d33
AP		 616 }
617
618 hba = super->hba;
619 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 620 * Do not support HBA types mixing
6b781d33
AP		 621 */
622 if (device->type != hba->type)
88654014 623 return 2;
6b781d33
AP		 624
625 /* Multiple same type HBAs can be used if they share the same OROM */
626 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
627
628 if (device_orom != super->orom)
629 return 2;
630
631 while (hba->next)
632 hba = hba->next;
633
634 hba->next = alloc_intel_hba(device);
635 return 1;
88654014
LM		 636}
637
638static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
639{
9bc4ae77 640 struct sys_dev *list, *elem;
88654014
LM		 641 char *disk_path;
642
643 if ((list = find_intel_devices()) == NULL)
644 return 0;
645
646 if (fd < 0)
647 disk_path = (char *) devname;
648 else
649 disk_path = diskfd_to_devpath(fd);
650
9bc4ae77 651 if (!disk_path)
88654014 652 return 0;
88654014 653
9bc4ae77
N		 654 for (elem = list; elem; elem = elem->next)
655 if (path_attached_to_hba(disk_path, elem->path))
88654014 656 return elem;
9bc4ae77 657
88654014
LM		 658 if (disk_path != devname)
659 free(disk_path);
88654014
LM		 660
661 return NULL;
662}
663
d424212e
N		 664static int find_intel_hba_capability(int fd, struct intel_super *super,
665 char *devname);
f2f5c343 666
cdddbdbc
DW		 667static struct supertype *match_metadata_desc_imsm(char *arg)
668{
669 struct supertype *st;
670
671 if (strcmp(arg, "imsm") != 0 &&
672 strcmp(arg, "default") != 0
673 )
674 return NULL;
675
503975b9 676 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 677 st->ss = &super_imsm;
678 st->max_devs = IMSM_MAX_DEVICES;
679 st->minor_version = 0;
680 st->sb = NULL;
681 return st;
682}
683
cdddbdbc
DW		 684static __u8 *get_imsm_version(struct imsm_super *mpb)
685{
686 return &mpb->sig[MPB_SIG_LEN];
687}
688
949c47a0
DW		 689/* retrieve a disk directly from the anchor when the anchor is known to be
690 * up-to-date, currently only at load time
691 */
692static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 693{
949c47a0 694 if (index >= mpb->num_disks)
cdddbdbc
DW		 695 return NULL;
696 return &mpb->disk[index];
697}
698
95d07a2c
LM		 699/* retrieve the disk description based on a index of the disk
700 * in the sub-array
701 */
702static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 703{
b9f594fe
DW		 704 struct dl *d;
705
706 for (d = super->disks; d; d = d->next)
707 if (d->index == index)
95d07a2c
LM		 708 return d;
709
710 return NULL;
711}
712/* retrieve a disk from the parsed metadata */
713static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
714{
715 struct dl *dl;
716
717 dl = get_imsm_dl_disk(super, index);
718 if (dl)
719 return &dl->disk;
720
b9f594fe 721 return NULL;
949c47a0
DW		 722}
723
724/* generate a checksum directly from the anchor when the anchor is known to be
725 * up-to-date, currently only at load or write_super after coalescing
726 */
727static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 728{
729 __u32 end = mpb->mpb_size / sizeof(end);
730 __u32 *p = (__u32 *) mpb;
731 __u32 sum = 0;
732
5d500228
N		 733 while (end--) {
734 sum += __le32_to_cpu(*p);
97f734fd
N		 735 p++;
736 }
cdddbdbc 737
5d500228 738 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 739}
740
a965f303
DW		 741static size_t sizeof_imsm_map(struct imsm_map *map)
742{
743 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
744}
745
746struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 747{
5e7b0330
AK		 748 /* A device can have 2 maps if it is in the middle of a migration.
749 * If second_map is:
238c0a71
AK		 750 * MAP_0 - we return the first map
751 * MAP_1 - we return the second map if it exists, else NULL
752 * MAP_X - we return the second map if it exists, else the first
5e7b0330 753 */
a965f303 754 struct imsm_map *map = &dev->vol.map[0];
9535fc47 755 struct imsm_map *map2 = NULL;
a965f303 756
9535fc47
AK		 757 if (dev->vol.migr_state)
758 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 759
9535fc47 760 switch (second_map) {
3b451610 761 case MAP_0:
9535fc47 762 break;
3b451610 763 case MAP_1:
9535fc47
AK		 764 map = map2;
765 break;
238c0a71 766 case MAP_X:
9535fc47
AK		 767 if (map2)
768 map = map2;
769 break;
9535fc47
AK		 770 default:
771 map = NULL;
772 }
773 return map;
5e7b0330 774
a965f303 775}
cdddbdbc 776
3393c6af
DW		 777/* return the size of the device.
778 * migr_state increases the returned size if map[0] were to be duplicated
779 */
780static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 781{
782 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 783 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 784
785 /* migrating means an additional map */
a965f303 786 if (dev->vol.migr_state)
238c0a71 787 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 788 else if (migr_state)
238c0a71 789 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 790
791 return size;
792}
793
54c2c1ea
DW		 794/* retrieve disk serial number list from a metadata update */
795static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
796{
797 void *u = update;
798 struct disk_info *inf;
799
800 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
801 sizeof_imsm_dev(&update->dev, 0);
802
803 return inf;
804}
54c2c1ea 805
949c47a0 806static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 807{
808 int offset;
809 int i;
810 void *_mpb = mpb;
811
949c47a0 812 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 813 return NULL;
814
815 /* devices start after all disks */
816 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
817
818 for (i = 0; i <= index; i++)
819 if (i == index)
820 return _mpb + offset;
821 else
3393c6af 822 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 823
824 return NULL;
825}
826
949c47a0
DW		 827static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
828{
ba2de7ba
DW		 829 struct intel_dev *dv;
830
949c47a0
DW		 831 if (index >= super->anchor->num_raid_devs)
832 return NULL;
ba2de7ba
DW		 833 for (dv = super->devlist; dv; dv = dv->next)
834 if (dv->index == index)
835 return dv->dev;
836 return NULL;
949c47a0
DW		 837}
838
8d67477f
TM		 839static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
840 *addr)
841{
842 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
843 __le16_to_cpu(addr->w1));
844}
845
846static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
847{
848 struct bbm_log_block_addr addr;
849
850 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
851 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
852 return addr;
853}
854
8d67477f
TM		 855/* get size of the bbm log */
856static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
857{
858 if (!log || log->entry_count == 0)
859 return 0;
860
861 return sizeof(log->signature) +
862 sizeof(log->entry_count) +
863 log->entry_count * sizeof(struct bbm_log_entry);
864}
6f50473f
TM		 865
866/* check if bad block is not partially stored in bbm log */
867static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
868 long long sector, const int length, __u32 *pos)
869{
870 __u32 i;
871
872 for (i = *pos; i < log->entry_count; i++) {
873 struct bbm_log_entry *entry = &log->marked_block_entries[i];
874 unsigned long long bb_start;
875 unsigned long long bb_end;
876
877 bb_start = __le48_to_cpu(&entry->defective_block_start);
878 bb_end = bb_start + (entry->marked_count + 1);
879
880 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
881 (bb_end <= sector + length)) {
882 *pos = i;
883 return 1;
884 }
885 }
886 return 0;
887}
888
889/* record new bad block in bbm log */
890static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
891 long long sector, int length)
892{
893 int new_bb = 0;
894 __u32 pos = 0;
895 struct bbm_log_entry *entry = NULL;
896
897 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
898 struct bbm_log_entry *e = &log->marked_block_entries[pos];
899
900 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
901 (__le48_to_cpu(&e->defective_block_start) == sector)) {
902 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
903 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
904 pos = pos + 1;
905 continue;
906 }
907 entry = e;
908 break;
909 }
910
911 if (entry) {
912 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
913 BBM_LOG_MAX_LBA_ENTRY_VAL;
914 entry->defective_block_start = __cpu_to_le48(sector);
915 entry->marked_count = cnt - 1;
916 if (cnt == length)
917 return 1;
918 sector += cnt;
919 length -= cnt;
920 }
921
922 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
923 BBM_LOG_MAX_LBA_ENTRY_VAL;
924 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
925 return 0;
926
927 while (length > 0) {
928 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
929 BBM_LOG_MAX_LBA_ENTRY_VAL;
930 struct bbm_log_entry *entry =
931 &log->marked_block_entries[log->entry_count];
932
933 entry->defective_block_start = __cpu_to_le48(sector);
934 entry->marked_count = cnt - 1;
935 entry->disk_ordinal = idx;
936
937 sector += cnt;
938 length -= cnt;
939
940 log->entry_count++;
941 }
942
943 return new_bb;
944}
c07a5a4f 945
4c9e8c1e
TM		 946/* clear all bad blocks for given disk */
947static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
948{
949 __u32 i = 0;
950
951 while (i < log->entry_count) {
952 struct bbm_log_entry *entries = log->marked_block_entries;
953
954 if (entries[i].disk_ordinal == idx) {
955 if (i < log->entry_count - 1)
956 entries[i] = entries[log->entry_count - 1];
957 log->entry_count--;
958 } else {
959 i++;
960 }
961 }
962}
963
c07a5a4f
TM		 964/* clear given bad block */
965static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
966 long long sector, const int length) {
967 __u32 i = 0;
968
969 while (i < log->entry_count) {
970 struct bbm_log_entry *entries = log->marked_block_entries;
971
972 if ((entries[i].disk_ordinal == idx) &&
973 (__le48_to_cpu(&entries[i].defective_block_start) ==
974 sector) && (entries[i].marked_count + 1 == length)) {
975 if (i < log->entry_count - 1)
976 entries[i] = entries[log->entry_count - 1];
977 log->entry_count--;
978 break;
979 }
980 i++;
981 }
982
983 return 1;
984}
8d67477f
TM		 985
986/* allocate and load BBM log from metadata */
987static int load_bbm_log(struct intel_super *super)
988{
989 struct imsm_super *mpb = super->anchor;
990 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
991
992 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
993 if (!super->bbm_log)
994 return 1;
995
996 if (bbm_log_size) {
997 struct bbm_log *log = (void *)mpb +
998 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
999
1000 __u32 entry_count;
1001
1002 if (bbm_log_size < sizeof(log->signature) +
1003 sizeof(log->entry_count))
1004 return 2;
1005
1006 entry_count = __le32_to_cpu(log->entry_count);
1007 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1008 (entry_count > BBM_LOG_MAX_ENTRIES))
1009 return 3;
1010
1011 if (bbm_log_size !=
1012 sizeof(log->signature) + sizeof(log->entry_count) +
1013 entry_count * sizeof(struct bbm_log_entry))
1014 return 4;
1015
1016 memcpy(super->bbm_log, log, bbm_log_size);
1017 } else {
1018 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1019 super->bbm_log->entry_count = 0;
1020 }
1021
1022 return 0;
1023}
1024
b12796be
TM		 1025/* checks if bad block is within volume boundaries */
1026static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1027 const unsigned long long start_sector,
1028 const unsigned long long size)
1029{
1030 unsigned long long bb_start;
1031 unsigned long long bb_end;
1032
1033 bb_start = __le48_to_cpu(&entry->defective_block_start);
1034 bb_end = bb_start + (entry->marked_count + 1);
1035
1036 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1037 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1038 return 1;
1039
1040 return 0;
1041}
1042
1043/* get list of bad blocks on a drive for a volume */
1044static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1045 const unsigned long long start_sector,
1046 const unsigned long long size,
1047 struct md_bb *bbs)
1048{
1049 __u32 count = 0;
1050 __u32 i;
1051
1052 for (i = 0; i < log->entry_count; i++) {
1053 const struct bbm_log_entry *ent =
1054 &log->marked_block_entries[i];
1055 struct md_bb_entry *bb;
1056
1057 if ((ent->disk_ordinal == idx) &&
1058 is_bad_block_in_volume(ent, start_sector, size)) {
1059
1060 if (!bbs->entries) {
1061 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1062 sizeof(*bb));
1063 if (!bbs->entries)
1064 break;
1065 }
1066
1067 bb = &bbs->entries[count++];
1068 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1069 bb->length = ent->marked_count + 1;
1070 }
1071 }
1072 bbs->count = count;
1073}
1074
98130f40
AK		 1075/*
1076 * for second_map:
238c0a71
AK		 1077 * == MAP_0 get first map
1078 * == MAP_1 get second map
1079 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1080 */
1081static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1082 int slot,
1083 int second_map)
7eef0453
DW		 1084{
1085 struct imsm_map *map;
1086
5e7b0330 1087 map = get_imsm_map(dev, second_map);
7eef0453 1088
ff077194
DW		 1089 /* top byte identifies disk under rebuild */
1090 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1091}
1092
1093#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1094static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1095{
98130f40 1096 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1097
1098 return ord_to_idx(ord);
7eef0453
DW		 1099}
1100
be73972f
DW		 1101static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1102{
1103 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1104}
1105
f21e18ca 1106static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1107{
1108 int slot;
1109 __u32 ord;
1110
1111 for (slot = 0; slot < map->num_members; slot++) {
1112 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1113 if (ord_to_idx(ord) == idx)
1114 return slot;
1115 }
1116
1117 return -1;
1118}
1119
cdddbdbc
DW		 1120static int get_imsm_raid_level(struct imsm_map *map)
1121{
1122 if (map->raid_level == 1) {
1123 if (map->num_members == 2)
1124 return 1;
1125 else
1126 return 10;
1127 }
1128
1129 return map->raid_level;
1130}
1131
c2c087e6
DW		 1132static int cmp_extent(const void *av, const void *bv)
1133{
1134 const struct extent *a = av;
1135 const struct extent *b = bv;
1136 if (a->start < b->start)
1137 return -1;
1138 if (a->start > b->start)
1139 return 1;
1140 return 0;
1141}
1142
0dcecb2e 1143static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1144{
c2c087e6 1145 int memberships = 0;
620b1713 1146 int i;
c2c087e6 1147
949c47a0
DW		 1148 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1149 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1150 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1151
620b1713
DW		 1152 if (get_imsm_disk_slot(map, dl->index) >= 0)
1153 memberships++;
c2c087e6 1154 }
0dcecb2e
DW		 1155
1156 return memberships;
1157}
1158
b81221b7
CA		 1159static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1160
5551b113
CA		 1161static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi)
1162{
1163 if (lo == 0 || hi == 0)
1164 return 1;
1165 *lo = __le32_to_cpu((unsigned)n);
1166 *hi = __le32_to_cpu((unsigned)(n >> 32));
1167 return 0;
1168}
1169
1170static unsigned long long join_u32(__u32 lo, __u32 hi)
1171{
1172 return (unsigned long long)__le32_to_cpu(lo) |
1173 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1174}
1175
1176static unsigned long long total_blocks(struct imsm_disk *disk)
1177{
1178 if (disk == NULL)
1179 return 0;
1180 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1181}
1182
1183static unsigned long long pba_of_lba0(struct imsm_map *map)
1184{
1185 if (map == NULL)
1186 return 0;
1187 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1188}
1189
1190static unsigned long long blocks_per_member(struct imsm_map *map)
1191{
1192 if (map == NULL)
1193 return 0;
1194 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1195}
1196
1197static unsigned long long num_data_stripes(struct imsm_map *map)
1198{
1199 if (map == NULL)
1200 return 0;
1201 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1202}
1203
fcc2c9da
MD		 1204static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1205{
1206 if (dev == NULL)
1207 return 0;
1208 return join_u32(dev->size_low, dev->size_high);
1209}
1210
5551b113
CA		 1211static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1212{
1213 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1214}
1215
1216static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1217{
1218 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1219}
1220
1221static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1222{
1223 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1224}
1225
1226static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1227{
1228 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1229}
1230
fcc2c9da
MD		 1231static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1232{
1233 split_ull(n, &dev->size_low, &dev->size_high);
1234}
1235
44490938
MD		 1236static unsigned long long per_dev_array_size(struct imsm_map *map)
1237{
1238 unsigned long long array_size = 0;
1239
1240 if (map == NULL)
1241 return array_size;
1242
1243 array_size = num_data_stripes(map) * map->blocks_per_strip;
1244 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1245 array_size *= 2;
1246
1247 return array_size;
1248}
1249
0dcecb2e
DW		 1250static struct extent *get_extents(struct intel_super *super, struct dl *dl)
1251{
1252 /* find a list of used extents on the given physical device */
1253 struct extent *rv, *e;
620b1713 1254 int i;
0dcecb2e 1255 int memberships = count_memberships(dl, super);
b276dd33
DW		 1256 __u32 reservation;
1257
1258 /* trim the reserved area for spares, so they can join any array
1259 * regardless of whether the OROM has assigned sectors from the
1260 * IMSM_RESERVED_SECTORS region
1261 */
1262 if (dl->index == -1)
b81221b7 1263 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1264 else
1265 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1266
503975b9 1267 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1268 e = rv;
1269
949c47a0
DW		 1270 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1271 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1272 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1273
620b1713 1274 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1275 e->start = pba_of_lba0(map);
44490938 1276 e->size = per_dev_array_size(map);
620b1713 1277 e++;
c2c087e6
DW		 1278 }
1279 }
1280 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1281
1011e834 1282 /* determine the start of the metadata
14e8215b
DW		 1283 * when no raid devices are defined use the default
1284 * ...otherwise allow the metadata to truncate the value
1285 * as is the case with older versions of imsm
1286 */
1287 if (memberships) {
1288 struct extent *last = &rv[memberships - 1];
5551b113 1289 unsigned long long remainder;
14e8215b 1290
5551b113 1291 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1292 /* round down to 1k block to satisfy precision of the kernel
1293 * 'size' interface
1294 */
1295 remainder &= ~1UL;
1296 /* make sure remainder is still sane */
f21e18ca 1297 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1298 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1299 if (reservation > remainder)
1300 reservation = remainder;
1301 }
5551b113 1302 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1303 e->size = 0;
1304 return rv;
1305}
1306
14e8215b
DW		 1307/* try to determine how much space is reserved for metadata from
1308 * the last get_extents() entry, otherwise fallback to the
1309 * default
1310 */
1311static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1312{
1313 struct extent *e;
1314 int i;
1315 __u32 rv;
1316
1317 /* for spares just return a minimal reservation which will grow
1318 * once the spare is picked up by an array
1319 */
1320 if (dl->index == -1)
1321 return MPB_SECTOR_CNT;
1322
1323 e = get_extents(super, dl);
1324 if (!e)
1325 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1326
1327 /* scroll to last entry */
1328 for (i = 0; e[i].size; i++)
1329 continue;
1330
5551b113 1331 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1332
1333 free(e);
1334
1335 return rv;
1336}
1337
25ed7e59
DW		 1338static int is_spare(struct imsm_disk *disk)
1339{
1340 return (disk->status & SPARE_DISK) == SPARE_DISK;
1341}
1342
1343static int is_configured(struct imsm_disk *disk)
1344{
1345 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1346}
1347
1348static int is_failed(struct imsm_disk *disk)
1349{
1350 return (disk->status & FAILED_DISK) == FAILED_DISK;
1351}
1352
2432ce9b
AP		 1353static int is_journal(struct imsm_disk *disk)
1354{
1355 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1356}
1357
b53bfba6
TM		 1358/* round array size down to closest MB and ensure it splits evenly
1359 * between members
1360 */
1361static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1362 disk_count)
1363{
1364 size /= disk_count;
1365 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1366 size *= disk_count;
1367
1368 return size;
1369}
1370
8b9cd157
MK		 1371static int able_to_resync(int raid_level, int missing_disks)
1372{
1373 int max_missing_disks = 0;
1374
1375 switch (raid_level) {
1376 case 10:
1377 max_missing_disks = 1;
1378 break;
1379 default:
1380 max_missing_disks = 0;
1381 }
1382 return missing_disks <= max_missing_disks;
1383}
1384
b81221b7
CA		 1385/* try to determine how much space is reserved for metadata from
1386 * the last get_extents() entry on the smallest active disk,
1387 * otherwise fallback to the default
1388 */
1389static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1390{
1391 struct extent *e;
1392 int i;
5551b113
CA		 1393 unsigned long long min_active;
1394 __u32 remainder;
b81221b7
CA		 1395 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1396 struct dl *dl, *dl_min = NULL;
1397
1398 if (!super)
1399 return rv;
1400
1401 min_active = 0;
1402 for (dl = super->disks; dl; dl = dl->next) {
1403 if (dl->index < 0)
1404 continue;
5551b113
CA		 1405 unsigned long long blocks = total_blocks(&dl->disk);
1406 if (blocks < min_active || min_active == 0) {
b81221b7 1407 dl_min = dl;
5551b113 1408 min_active = blocks;
b81221b7
CA		 1409 }
1410 }
1411 if (!dl_min)
1412 return rv;
1413
1414 /* find last lba used by subarrays on the smallest active disk */
1415 e = get_extents(super, dl_min);
1416 if (!e)
1417 return rv;
1418 for (i = 0; e[i].size; i++)
1419 continue;
1420
1421 remainder = min_active - e[i].start;
1422 free(e);
1423
1424 /* to give priority to recovery we should not require full
1425 IMSM_RESERVED_SECTORS from the spare */
1426 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1427
1428 /* if real reservation is smaller use that value */
1429 return (remainder < rv) ? remainder : rv;
1430}
1431
fbfdcb06
AO		 1432/*
1433 * Return minimum size of a spare and sector size
1434 * that can be used in this array
1435 */
1436int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1437{
1438 struct intel_super *super = st->sb;
1439 struct dl *dl;
1440 struct extent *e;
1441 int i;
fbfdcb06
AO		 1442 unsigned long long size = 0;
1443
1444 c->min_size = 0;
4b57ecf6 1445 c->sector_size = 0;
80e7f8c3
AC		 1446
1447 if (!super)
fbfdcb06 1448 return -EINVAL;
80e7f8c3
AC		 1449 /* find first active disk in array */
1450 dl = super->disks;
1451 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1452 dl = dl->next;
1453 if (!dl)
fbfdcb06 1454 return -EINVAL;
80e7f8c3
AC		 1455 /* find last lba used by subarrays */
1456 e = get_extents(super, dl);
1457 if (!e)
fbfdcb06 1458 return -EINVAL;
80e7f8c3
AC		 1459 for (i = 0; e[i].size; i++)
1460 continue;
1461 if (i > 0)
fbfdcb06 1462 size = e[i-1].start + e[i-1].size;
80e7f8c3 1463 free(e);
b81221b7 1464
80e7f8c3 1465 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1466 size += imsm_min_reserved_sectors(super);
1467
1468 c->min_size = size * 512;
4b57ecf6 1469 c->sector_size = super->sector_size;
b81221b7 1470
fbfdcb06 1471 return 0;
80e7f8c3
AC		 1472}
1473
d1e02575
AK		 1474static int is_gen_migration(struct imsm_dev *dev);
1475
f36a9ecd
PB		 1476#define IMSM_4K_DIV 8
1477
c47b0ff6
AK		 1478static __u64 blocks_per_migr_unit(struct intel_super *super,
1479 struct imsm_dev *dev);
1e5c6983 1480
c47b0ff6
AK		 1481static void print_imsm_dev(struct intel_super *super,
1482 struct imsm_dev *dev,
1483 char *uuid,
1484 int disk_idx)
cdddbdbc
DW		 1485{
1486 __u64 sz;
0d80bb2f 1487 int slot, i;
238c0a71
AK		 1488 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1489 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1490 __u32 ord;
cdddbdbc
DW		 1491
1492 printf("\n");
1e7bc0ed 1493 printf("[%.16s]:\n", dev->volume);
44470971 1494 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1495 printf(" RAID Level : %d", get_imsm_raid_level(map));
1496 if (map2)
1497 printf(" <-- %d", get_imsm_raid_level(map2));
1498 printf("\n");
1499 printf(" Members : %d", map->num_members);
1500 if (map2)
1501 printf(" <-- %d", map2->num_members);
1502 printf("\n");
0d80bb2f
DW		 1503 printf(" Slots : [");
1504 for (i = 0; i < map->num_members; i++) {
238c0a71 1505 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1506 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1507 }
dd8bcb3b
AK		 1508 printf("]");
1509 if (map2) {
1510 printf(" <-- [");
1511 for (i = 0; i < map2->num_members; i++) {
238c0a71 1512 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1513 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1514 }
1515 printf("]");
1516 }
1517 printf("\n");
7095bccb
AK		 1518 printf(" Failed disk : ");
1519 if (map->failed_disk_num == 0xff)
1520 printf("none");
1521 else
1522 printf("%i", map->failed_disk_num);
1523 printf("\n");
620b1713
DW		 1524 slot = get_imsm_disk_slot(map, disk_idx);
1525 if (slot >= 0) {
238c0a71 1526 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1527 printf(" This Slot : %d%s\n", slot,
1528 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1529 } else
cdddbdbc 1530 printf(" This Slot : ?\n");
84918897 1531 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1532 sz = imsm_dev_size(dev);
84918897
MK		 1533 printf(" Array Size : %llu%s\n",
1534 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1535 human_size(sz * 512));
5551b113 1536 sz = blocks_per_member(map);
84918897
MK		 1537 printf(" Per Dev Size : %llu%s\n",
1538 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1539 human_size(sz * 512));
5551b113
CA		 1540 printf(" Sector Offset : %llu\n",
1541 pba_of_lba0(map));
1542 printf(" Num Stripes : %llu\n",
1543 num_data_stripes(map));
dd8bcb3b 1544 printf(" Chunk Size : %u KiB",
cdddbdbc 1545 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1546 if (map2)
1547 printf(" <-- %u KiB",
1548 __le16_to_cpu(map2->blocks_per_strip) / 2);
1549 printf("\n");
cdddbdbc 1550 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1551 printf(" Migrate State : ");
1484e727
DW		 1552 if (dev->vol.migr_state) {
1553 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1554 printf("initialize\n");
1484e727 1555 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1556 printf("rebuild\n");
1484e727 1557 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1558 printf("check\n");
1484e727 1559 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1560 printf("general migration\n");
1484e727 1561 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1562 printf("state change\n");
1484e727 1563 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1564 printf("repair\n");
1484e727 1565 else
8655a7b1
DW		 1566 printf("<unknown:%d>\n", migr_type(dev));
1567 } else
1568 printf("idle\n");
3393c6af
DW		 1569 printf(" Map State : %s", map_state_str[map->map_state]);
1570 if (dev->vol.migr_state) {
238c0a71 1571 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1572
b10b37b8 1573 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1574 printf("\n Checkpoint : %u ",
1575 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1576 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1577 printf("(N/A)");
1578 else
1579 printf("(%llu)", (unsigned long long)
1580 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1581 }
1582 printf("\n");
2432ce9b
AP		 1583 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1584 "dirty" : "clean");
1585 printf(" RWH Policy : ");
c2462068 1586 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1587 printf("off\n");
1588 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1589 printf("PPL distributed\n");
1590 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1591 printf("PPL journaling drive\n");
c2462068
PB		 1592 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1593 printf("Multiple distributed PPLs\n");
1594 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1595 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1596 else
1597 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1598}
1599
ef5c214e
MK		 1600static void print_imsm_disk(struct imsm_disk *disk,
1601 int index,
1602 __u32 reserved,
1603 unsigned int sector_size) {
1f24f035 1604 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1605 __u64 sz;
1606
0ec1f4e8 1607 if (index < -1 || !disk)
e9d82038
DW		 1608 return;
1609
cdddbdbc 1610 printf("\n");
1f24f035 1611 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1612 if (index >= 0)
1613 printf(" Disk%02d Serial : %s\n", index, str);
1614 else
1615 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1616 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1617 is_configured(disk) ? " active" : "",
1618 is_failed(disk) ? " failed" : "",
1619 is_journal(disk) ? " journal" : "");
cdddbdbc 1620 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1621 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1622 printf(" Usable Size : %llu%s\n",
1623 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1624 human_size(sz * 512));
1625}
1626
de44e46f
PB		 1627void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1628{
1629 struct migr_record *migr_rec = super->migr_rec;
1630
1631 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
1632 migr_rec->ckpt_area_pba /= IMSM_4K_DIV;
1633 migr_rec->dest_1st_member_lba /= IMSM_4K_DIV;
1634 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1635 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1636 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1637 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
1638}
1639
f36a9ecd
PB		 1640void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1641{
1642 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1643}
1644
1645void convert_to_4k(struct intel_super *super)
1646{
1647 struct imsm_super *mpb = super->anchor;
1648 struct imsm_disk *disk;
1649 int i;
e4467bc7 1650 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1651
1652 for (i = 0; i < mpb->num_disks ; i++) {
1653 disk = __get_imsm_disk(mpb, i);
1654 /* disk */
1655 convert_to_4k_imsm_disk(disk);
1656 }
1657 for (i = 0; i < mpb->num_raid_devs; i++) {
1658 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1659 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1660 /* dev */
fcc2c9da 1661 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1662 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1663
1664 /* map0 */
1665 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1666 map->blocks_per_strip /= IMSM_4K_DIV;
1667 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1668
1669 if (dev->vol.migr_state) {
1670 /* map1 */
1671 map = get_imsm_map(dev, MAP_1);
1672 set_blocks_per_member(map,
1673 blocks_per_member(map)/IMSM_4K_DIV);
1674 map->blocks_per_strip /= IMSM_4K_DIV;
1675 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1676 }
1677 }
e4467bc7
TM		 1678 if (bbm_log_size) {
1679 struct bbm_log *log = (void *)mpb +
1680 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1681 __u32 i;
1682
1683 for (i = 0; i < log->entry_count; i++) {
1684 struct bbm_log_entry *entry =
1685 &log->marked_block_entries[i];
1686
1687 __u8 count = entry->marked_count + 1;
1688 unsigned long long sector =
1689 __le48_to_cpu(&entry->defective_block_start);
1690
1691 entry->defective_block_start =
1692 __cpu_to_le48(sector/IMSM_4K_DIV);
1693 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1694 }
1695 }
f36a9ecd
PB		 1696
1697 mpb->check_sum = __gen_imsm_checksum(mpb);
1698}
1699
520e69e2
AK		 1700void examine_migr_rec_imsm(struct intel_super *super)
1701{
1702 struct migr_record *migr_rec = super->migr_rec;
1703 struct imsm_super *mpb = super->anchor;
1704 int i;
1705
1706 for (i = 0; i < mpb->num_raid_devs; i++) {
1707 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1708 struct imsm_map *map;
b4ab44d8 1709 int slot = -1;
3136abe5 1710
520e69e2
AK		 1711 if (is_gen_migration(dev) == 0)
1712 continue;
1713
1714 printf("\nMigration Record Information:");
3136abe5 1715
44bfe6df
AK		 1716 /* first map under migration */
1717 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1718 if (map)
1719 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1720 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1721 printf(" Empty\n ");
1722 printf("Examine one of first two disks in array\n");
1723 break;
1724 }
1725 printf("\n Status : ");
1726 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1727 printf("Normal\n");
1728 else
1729 printf("Contains Data\n");
1730 printf(" Current Unit : %u\n",
1731 __le32_to_cpu(migr_rec->curr_migr_unit));
1732 printf(" Family : %u\n",
1733 __le32_to_cpu(migr_rec->family_num));
1734 printf(" Ascending : %u\n",
1735 __le32_to_cpu(migr_rec->ascending_migr));
1736 printf(" Blocks Per Unit : %u\n",
1737 __le32_to_cpu(migr_rec->blocks_per_unit));
1738 printf(" Dest. Depth Per Unit : %u\n",
1739 __le32_to_cpu(migr_rec->dest_depth_per_unit));
1740 printf(" Checkpoint Area pba : %u\n",
1741 __le32_to_cpu(migr_rec->ckpt_area_pba));
1742 printf(" First member lba : %u\n",
1743 __le32_to_cpu(migr_rec->dest_1st_member_lba));
1744 printf(" Total Number of Units : %u\n",
1745 __le32_to_cpu(migr_rec->num_migr_units));
1746 printf(" Size of volume : %u\n",
1747 __le32_to_cpu(migr_rec->post_migr_vol_cap));
1748 printf(" Expansion space for LBA64 : %u\n",
1749 __le32_to_cpu(migr_rec->post_migr_vol_cap_hi));
1750 printf(" Record was read from : %u\n",
1751 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1752
1753 break;
1754 }
1755}
f36a9ecd 1756
de44e46f
PB		 1757void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1758{
1759 struct migr_record *migr_rec = super->migr_rec;
1760
1761 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
1762 migr_rec->ckpt_area_pba *= IMSM_4K_DIV;
1763 migr_rec->dest_1st_member_lba *= IMSM_4K_DIV;
1764 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1765 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1766 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1767 &migr_rec->post_migr_vol_cap,
1768 &migr_rec->post_migr_vol_cap_hi);
1769}
1770
f36a9ecd
PB		 1771void convert_from_4k(struct intel_super *super)
1772{
1773 struct imsm_super *mpb = super->anchor;
1774 struct imsm_disk *disk;
1775 int i;
e4467bc7 1776 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1777
1778 for (i = 0; i < mpb->num_disks ; i++) {
1779 disk = __get_imsm_disk(mpb, i);
1780 /* disk */
1781 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1782 }
1783
1784 for (i = 0; i < mpb->num_raid_devs; i++) {
1785 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1786 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1787 /* dev */
fcc2c9da 1788 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1789 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1790
1791 /* map0 */
1792 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1793 map->blocks_per_strip *= IMSM_4K_DIV;
1794 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1795
1796 if (dev->vol.migr_state) {
1797 /* map1 */
1798 map = get_imsm_map(dev, MAP_1);
1799 set_blocks_per_member(map,
1800 blocks_per_member(map)*IMSM_4K_DIV);
1801 map->blocks_per_strip *= IMSM_4K_DIV;
1802 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1803 }
1804 }
e4467bc7
TM		 1805 if (bbm_log_size) {
1806 struct bbm_log *log = (void *)mpb +
1807 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1808 __u32 i;
1809
1810 for (i = 0; i < log->entry_count; i++) {
1811 struct bbm_log_entry *entry =
1812 &log->marked_block_entries[i];
1813
1814 __u8 count = entry->marked_count + 1;
1815 unsigned long long sector =
1816 __le48_to_cpu(&entry->defective_block_start);
1817
1818 entry->defective_block_start =
1819 __cpu_to_le48(sector*IMSM_4K_DIV);
1820 entry->marked_count = count*IMSM_4K_DIV - 1;
1821 }
1822 }
f36a9ecd
PB		 1823
1824 mpb->check_sum = __gen_imsm_checksum(mpb);
1825}
1826
19482bcc
AK		 1827/*******************************************************************************
1828 * function: imsm_check_attributes
1829 * Description: Function checks if features represented by attributes flags
1011e834 1830 * are supported by mdadm.
19482bcc
AK		 1831 * Parameters:
1832 * attributes - Attributes read from metadata
1833 * Returns:
1011e834
N		 1834 * 0 - passed attributes contains unsupported features flags
1835 * 1 - all features are supported
19482bcc
AK		 1836 ******************************************************************************/
1837static int imsm_check_attributes(__u32 attributes)
1838{
1839 int ret_val = 1;
418f9b36
N		 1840 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1841
1842 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1843
1844 not_supported &= attributes;
1845 if (not_supported) {
e7b84f9d 1846 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1847 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1848 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1849 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1850 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1851 }
1852 if (not_supported & MPB_ATTRIB_2TB) {
1853 dprintf("\t\tMPB_ATTRIB_2TB\n");
1854 not_supported ^= MPB_ATTRIB_2TB;
1855 }
1856 if (not_supported & MPB_ATTRIB_RAID0) {
1857 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1858 not_supported ^= MPB_ATTRIB_RAID0;
1859 }
1860 if (not_supported & MPB_ATTRIB_RAID1) {
1861 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1862 not_supported ^= MPB_ATTRIB_RAID1;
1863 }
1864 if (not_supported & MPB_ATTRIB_RAID10) {
1865 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1866 not_supported ^= MPB_ATTRIB_RAID10;
1867 }
1868 if (not_supported & MPB_ATTRIB_RAID1E) {
1869 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1870 not_supported ^= MPB_ATTRIB_RAID1E;
1871 }
1872 if (not_supported & MPB_ATTRIB_RAID5) {
1873 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1874 not_supported ^= MPB_ATTRIB_RAID5;
1875 }
1876 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1877 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1878 not_supported ^= MPB_ATTRIB_RAIDCNG;
1879 }
1880 if (not_supported & MPB_ATTRIB_BBM) {
1881 dprintf("\t\tMPB_ATTRIB_BBM\n");
1882 not_supported ^= MPB_ATTRIB_BBM;
1883 }
1884 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1885 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1886 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1887 }
1888 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1889 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1890 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1891 }
1892 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1893 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1894 not_supported ^= MPB_ATTRIB_2TB_DISK;
1895 }
1896 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1897 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1898 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1899 }
1900 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1901 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1902 not_supported ^= MPB_ATTRIB_NEVER_USE;
1903 }
1904
1905 if (not_supported)
1ade5cc1 1906 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1907
1908 ret_val = 0;
1909 }
1910
1911 return ret_val;
1912}
1913
a5d85af7 1914static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 1915
cdddbdbc
DW		 1916static void examine_super_imsm(struct supertype *st, char *homehost)
1917{
1918 struct intel_super *super = st->sb;
949c47a0 1919 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 1920 char str[MAX_SIGNATURE_LENGTH];
1921 int i;
27fd6274
DW		 1922 struct mdinfo info;
1923 char nbuf[64];
cdddbdbc 1924 __u32 sum;
14e8215b 1925 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 1926 struct dl *dl;
27fd6274 1927
618f4e6d
XN		 1928 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
1929 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 1930 printf(" Magic : %s\n", str);
cdddbdbc 1931 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 1932 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 1933 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
1934 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 1935 printf(" Attributes : ");
1936 if (imsm_check_attributes(mpb->attributes))
1937 printf("All supported\n");
1938 else
1939 printf("not supported\n");
a5d85af7 1940 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1941 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 1942 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 1943 sum = __le32_to_cpu(mpb->check_sum);
1944 printf(" Checksum : %08x %s\n", sum,
949c47a0 1945 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 1946 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 1947 printf(" Disks : %d\n", mpb->num_disks);
1948 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 1949 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
1950 super->disks->index, reserved, super->sector_size);
8d67477f 1951 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 1952 struct bbm_log *log = super->bbm_log;
1953
1954 printf("\n");
1955 printf("Bad Block Management Log:\n");
1956 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
1957 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
1958 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 1959 }
44470971
DW		 1960 for (i = 0; i < mpb->num_raid_devs; i++) {
1961 struct mdinfo info;
1962 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1963
1964 super->current_vol = i;
a5d85af7 1965 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1966 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 1967 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 1968 }
cdddbdbc
DW		 1969 for (i = 0; i < mpb->num_disks; i++) {
1970 if (i == super->disks->index)
1971 continue;
ef5c214e
MK		 1972 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
1973 super->sector_size);
cdddbdbc 1974 }
94827db3 1975
0ec1f4e8
DW		 1976 for (dl = super->disks; dl; dl = dl->next)
1977 if (dl->index == -1)
ef5c214e
MK		 1978 print_imsm_disk(&dl->disk, -1, reserved,
1979 super->sector_size);
520e69e2
AK		 1980
1981 examine_migr_rec_imsm(super);
cdddbdbc
DW		 1982}
1983
061f2c6a 1984static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 1985{
27fd6274 1986 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 1987 struct mdinfo info;
1988 char nbuf[64];
1e7bc0ed 1989 struct intel_super *super = st->sb;
1e7bc0ed 1990
0d5a423f
DW		 1991 if (!super->anchor->num_raid_devs) {
1992 printf("ARRAY metadata=imsm\n");
1e7bc0ed 1993 return;
0d5a423f 1994 }
ff54de6e 1995
a5d85af7 1996 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 1997 fname_from_uuid(st, &info, nbuf, ':');
1998 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
1999}
2000
2001static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2002{
2003 /* We just write a generic IMSM ARRAY entry */
2004 struct mdinfo info;
2005 char nbuf[64];
2006 char nbuf1[64];
2007 struct intel_super *super = st->sb;
2008 int i;
2009
2010 if (!super->anchor->num_raid_devs)
2011 return;
2012
a5d85af7 2013 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2014 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2015 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2016 struct imsm_dev *dev = get_imsm_dev(super, i);
2017
2018 super->current_vol = i;
a5d85af7 2019 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2020 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2021 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2022 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2023 }
cdddbdbc
DW		 2024}
2025
9d84c8ea
DW		 2026static void export_examine_super_imsm(struct supertype *st)
2027{
2028 struct intel_super *super = st->sb;
2029 struct imsm_super *mpb = super->anchor;
2030 struct mdinfo info;
2031 char nbuf[64];
2032
a5d85af7 2033 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2034 fname_from_uuid(st, &info, nbuf, ':');
2035 printf("MD_METADATA=imsm\n");
2036 printf("MD_LEVEL=container\n");
2037 printf("MD_UUID=%s\n", nbuf+5);
2038 printf("MD_DEVICES=%u\n", mpb->num_disks);
2039}
2040
74db60b0
N		 2041static int copy_metadata_imsm(struct supertype *st, int from, int to)
2042{
f36a9ecd 2043 /* The second last sector of the device contains
74db60b0
N		 2044 * the "struct imsm_super" metadata.
2045 * This contains mpb_size which is the size in bytes of the
2046 * extended metadata. This is located immediately before
2047 * the imsm_super.
2048 * We want to read all that, plus the last sector which
2049 * may contain a migration record, and write it all
2050 * to the target.
2051 */
2052 void *buf;
2053 unsigned long long dsize, offset;
2054 int sectors;
2055 struct imsm_super *sb;
f36a9ecd
PB		 2056 struct intel_super *super = st->sb;
2057 unsigned int sector_size = super->sector_size;
2058 unsigned int written = 0;
74db60b0 2059
de44e46f 2060 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 2061 return 1;
2062
2063 if (!get_dev_size(from, NULL, &dsize))
2064 goto err;
2065
f36a9ecd 2066 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 2067 goto err;
466070ad 2068 if ((unsigned int)read(from, buf, sector_size) != sector_size)
74db60b0
N		 2069 goto err;
2070 sb = buf;
2071 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
2072 goto err;
2073
f36a9ecd
PB		 2074 sectors = mpb_sectors(sb, sector_size) + 2;
2075 offset = dsize - sectors * sector_size;
74db60b0
N		 2076 if (lseek64(from, offset, 0) < 0 ||
2077 lseek64(to, offset, 0) < 0)
2078 goto err;
f36a9ecd
PB		 2079 while (written < sectors * sector_size) {
2080 int n = sectors*sector_size - written;
74db60b0
N		 2081 if (n > 4096)
2082 n = 4096;
2083 if (read(from, buf, n) != n)
2084 goto err;
2085 if (write(to, buf, n) != n)
2086 goto err;
2087 written += n;
2088 }
2089 free(buf);
2090 return 0;
2091err:
2092 free(buf);
2093 return 1;
2094}
2095
cdddbdbc
DW		 2096static void detail_super_imsm(struct supertype *st, char *homehost)
2097{
3ebe00a1
DW		 2098 struct mdinfo info;
2099 char nbuf[64];
2100
a5d85af7 2101 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2102 fname_from_uuid(st, &info, nbuf, ':');
65884368 2103 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2104}
2105
2106static void brief_detail_super_imsm(struct supertype *st)
2107{
ff54de6e
N		 2108 struct mdinfo info;
2109 char nbuf[64];
a5d85af7 2110 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2111 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2112 printf(" UUID=%s", nbuf + 5);
cdddbdbc 2113}
d665cc31
DW		 2114
2115static int imsm_read_serial(int fd, char *devname, __u8 *serial);
2116static void fd2devname(int fd, char *name);
2117
120dc887 2118static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2119{
120dc887
LM		 2120 /* dump an unsorted list of devices attached to AHCI Intel storage
2121 * controller, as well as non-connected ports
d665cc31
DW		 2122 */
2123 int hba_len = strlen(hba_path) + 1;
2124 struct dirent *ent;
2125 DIR *dir;
2126 char *path = NULL;
2127 int err = 0;
2128 unsigned long port_mask = (1 << port_count) - 1;
2129
f21e18ca 2130 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2131 if (verbose > 0)
e7b84f9d 2132 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2133 return 2;
2134 }
2135
2136 /* scroll through /sys/dev/block looking for devices attached to
2137 * this hba
2138 */
2139 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2140 if (!dir)
2141 return 1;
2142
2143 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2144 int fd;
2145 char model[64];
2146 char vendor[64];
2147 char buf[1024];
2148 int major, minor;
2149 char *device;
2150 char *c;
2151 int port;
2152 int type;
2153
2154 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2155 continue;
2156 path = devt_to_devpath(makedev(major, minor));
2157 if (!path)
2158 continue;
2159 if (!path_attached_to_hba(path, hba_path)) {
2160 free(path);
2161 path = NULL;
2162 continue;
2163 }
2164
2165 /* retrieve the scsi device type */
2166 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2167 if (verbose > 0)
e7b84f9d 2168 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2169 err = 2;
2170 break;
2171 }
2172 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2173 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2174 if (verbose > 0)
e7b84f9d 2175 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2176 path);
2177 err = 2;
2178 free(device);
2179 break;
2180 }
2181 type = strtoul(buf, NULL, 10);
2182
2183 /* if it's not a disk print the vendor and model */
2184 if (!(type == 0 || type == 7 || type == 14)) {
2185 vendor[0] = '\0';
2186 model[0] = '\0';
2187 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2188 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2189 strncpy(vendor, buf, sizeof(vendor));
2190 vendor[sizeof(vendor) - 1] = '\0';
2191 c = (char *) &vendor[sizeof(vendor) - 1];
2192 while (isspace(*c) || *c == '\0')
2193 *c-- = '\0';
2194
2195 }
2196 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2197 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2198 strncpy(model, buf, sizeof(model));
2199 model[sizeof(model) - 1] = '\0';
2200 c = (char *) &model[sizeof(model) - 1];
2201 while (isspace(*c) || *c == '\0')
2202 *c-- = '\0';
2203 }
2204
2205 if (vendor[0] && model[0])
2206 sprintf(buf, "%.64s %.64s", vendor, model);
2207 else
2208 switch (type) { /* numbers from hald/linux/device.c */
2209 case 1: sprintf(buf, "tape"); break;
2210 case 2: sprintf(buf, "printer"); break;
2211 case 3: sprintf(buf, "processor"); break;
2212 case 4:
2213 case 5: sprintf(buf, "cdrom"); break;
2214 case 6: sprintf(buf, "scanner"); break;
2215 case 8: sprintf(buf, "media_changer"); break;
2216 case 9: sprintf(buf, "comm"); break;
2217 case 12: sprintf(buf, "raid"); break;
2218 default: sprintf(buf, "unknown");
2219 }
2220 } else
2221 buf[0] = '\0';
2222 free(device);
2223
2224 /* chop device path to 'host%d' and calculate the port number */
2225 c = strchr(&path[hba_len], '/');
4e5e717d 2226 if (!c) {
ba728be7 2227 if (verbose > 0)
e7b84f9d 2228 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2229 err = 2;
2230 break;
2231 }
d665cc31 2232 *c = '\0';
0858eccf
AP		 2233 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2234 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2235 port -= host_base;
2236 else {
ba728be7 2237 if (verbose > 0) {
d665cc31 2238 *c = '/'; /* repair the full string */
e7b84f9d 2239 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2240 path);
2241 }
2242 err = 2;
2243 break;
2244 }
2245
2246 /* mark this port as used */
2247 port_mask &= ~(1 << port);
2248
2249 /* print out the device information */
2250 if (buf[0]) {
2251 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2252 continue;
2253 }
2254
2255 fd = dev_open(ent->d_name, O_RDONLY);
2256 if (fd < 0)
2257 printf(" Port%d : - disk info unavailable -\n", port);
2258 else {
2259 fd2devname(fd, buf);
2260 printf(" Port%d : %s", port, buf);
2261 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2262 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2263 else
664d5325 2264 printf(" ()\n");
4dab422a 2265 close(fd);
d665cc31 2266 }
d665cc31
DW		 2267 free(path);
2268 path = NULL;
2269 }
2270 if (path)
2271 free(path);
2272 if (dir)
2273 closedir(dir);
2274 if (err == 0) {
2275 int i;
2276
2277 for (i = 0; i < port_count; i++)
2278 if (port_mask & (1 << i))
2279 printf(" Port%d : - no device attached -\n", i);
2280 }
2281
2282 return err;
2283}
2284
b5eece69 2285static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2286{
2287 struct dirent *ent;
2288 DIR *dir;
2289 char path[292];
2290 char link[256];
2291 char *c, *rp;
2292
2293 if (hba->type != SYS_DEV_VMD)
b5eece69 2294 return 1;
60f0f54d
PB		 2295
2296 /* scroll through /sys/dev/block looking for devices attached to
2297 * this hba
2298 */
2299 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2300 if (!dir)
b5eece69 2301 return 1;
b9135011
JS		 2302
2303 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2304 int n;
2305
2306 /* is 'ent' a device? check that the 'subsystem' link exists and
2307 * that its target matches 'bus'
2308 */
2309 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2310 ent->d_name);
2311 n = readlink(path, link, sizeof(link));
2312 if (n < 0 || n >= (int)sizeof(link))
2313 continue;
2314 link[n] = '\0';
2315 c = strrchr(link, '/');
2316 if (!c)
2317 continue;
2318 if (strncmp("pci", c+1, strlen("pci")) != 0)
2319 continue;
2320
2321 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2322
2323 rp = realpath(path, NULL);
2324 if (!rp)
2325 continue;
2326
2327 if (path_attached_to_hba(rp, hba->path)) {
2328 printf(" NVMe under VMD : %s\n", rp);
2329 }
2330 free(rp);
2331 }
2332
b9135011 2333 closedir(dir);
b5eece69 2334 return 0;
60f0f54d
PB		 2335}
2336
120dc887
LM		 2337static void print_found_intel_controllers(struct sys_dev *elem)
2338{
2339 for (; elem; elem = elem->next) {
e7b84f9d 2340 pr_err("found Intel(R) ");
120dc887
LM		 2341 if (elem->type == SYS_DEV_SATA)
2342 fprintf(stderr, "SATA ");
155cbb4c
LM		 2343 else if (elem->type == SYS_DEV_SAS)
2344 fprintf(stderr, "SAS ");
0858eccf
AP		 2345 else if (elem->type == SYS_DEV_NVME)
2346 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2347
2348 if (elem->type == SYS_DEV_VMD)
2349 fprintf(stderr, "VMD domain");
2350 else
2351 fprintf(stderr, "RAID controller");
2352
120dc887
LM		 2353 if (elem->pci_id)
2354 fprintf(stderr, " at %s", elem->pci_id);
2355 fprintf(stderr, ".\n");
2356 }
2357 fflush(stderr);
2358}
2359
120dc887
LM		 2360static int ahci_get_port_count(const char *hba_path, int *port_count)
2361{
2362 struct dirent *ent;
2363 DIR *dir;
2364 int host_base = -1;
2365
2366 *port_count = 0;
2367 if ((dir = opendir(hba_path)) == NULL)
2368 return -1;
2369
2370 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2371 int host;
2372
0858eccf
AP		 2373 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2374 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2375 continue;
2376 if (*port_count == 0)
2377 host_base = host;
2378 else if (host < host_base)
2379 host_base = host;
2380
2381 if (host + 1 > *port_count + host_base)
2382 *port_count = host + 1 - host_base;
2383 }
2384 closedir(dir);
2385 return host_base;
2386}
2387
a891a3c2
LM		 2388static void print_imsm_capability(const struct imsm_orom *orom)
2389{
0858eccf
AP		 2390 printf(" Platform : Intel(R) ");
2391 if (orom->capabilities == 0 && orom->driver_features == 0)
2392 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2393 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2394 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2395 else
2396 printf("Rapid Storage Technology%s\n",
2397 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2398 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2399 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2400 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2401 printf(" RAID Levels :%s%s%s%s%s\n",
2402 imsm_orom_has_raid0(orom) ? " raid0" : "",
2403 imsm_orom_has_raid1(orom) ? " raid1" : "",
2404 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2405 imsm_orom_has_raid10(orom) ? " raid10" : "",
2406 imsm_orom_has_raid5(orom) ? " raid5" : "");
2407 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2408 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2409 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2410 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2411 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2412 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2413 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2414 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2415 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2416 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2417 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2418 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2419 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2420 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2421 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2422 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2423 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2424 printf(" 2TB volumes :%s supported\n",
2425 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2426 printf(" 2TB disks :%s supported\n",
2427 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2428 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2429 printf(" Max Volumes : %d per array, %d per %s\n",
2430 orom->vpa, orom->vphba,
2431 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2432 return;
2433}
2434
e50cf220
MN		 2435static void print_imsm_capability_export(const struct imsm_orom *orom)
2436{
2437 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2438 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2439 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2440 orom->hotfix_ver, orom->build);
e50cf220
MN		 2441 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2442 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2443 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2444 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2445 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2446 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2447 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2448 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2449 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2450 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2451 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2452 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2453 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2454 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2455 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2456 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2457 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2458 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2459 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2460 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2461 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2462 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2463 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2464 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2465 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2466 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2467 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2468 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2469}
2470
9eafa1de 2471static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2472{
2473 /* There are two components to imsm platform support, the ahci SATA
2474 * controller and the option-rom. To find the SATA controller we
2475 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2476 * controller with the Intel vendor id is present. This approach
2477 * allows mdadm to leverage the kernel's ahci detection logic, with the
2478 * caveat that if ahci.ko is not loaded mdadm will not be able to
2479 * detect platform raid capabilities. The option-rom resides in a
2480 * platform "Adapter ROM". We scan for its signature to retrieve the
2481 * platform capabilities. If raid support is disabled in the BIOS the
2482 * option-rom capability structure will not be available.
2483 */
d665cc31 2484 struct sys_dev *list, *hba;
d665cc31
DW		 2485 int host_base = 0;
2486 int port_count = 0;
9eafa1de 2487 int result=1;
d665cc31 2488
5615172f 2489 if (enumerate_only) {
a891a3c2 2490 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2491 return 0;
a891a3c2
LM		 2492 list = find_intel_devices();
2493 if (!list)
2494 return 2;
2495 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2496 if (find_imsm_capability(hba)) {
2497 result = 0;
a891a3c2
LM		 2498 break;
2499 }
9eafa1de 2500 else
6b781d33 2501 result = 2;
a891a3c2 2502 }
a891a3c2 2503 return result;
5615172f
DW		 2504 }
2505
155cbb4c
LM		 2506 list = find_intel_devices();
2507 if (!list) {
ba728be7 2508 if (verbose > 0)
7a862a02 2509 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2510 return 2;
ba728be7 2511 } else if (verbose > 0)
155cbb4c 2512 print_found_intel_controllers(list);
d665cc31 2513
a891a3c2 2514 for (hba = list; hba; hba = hba->next) {
0858eccf 2515 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2516 continue;
0858eccf 2517 if (!find_imsm_capability(hba)) {
60f0f54d 2518 char buf[PATH_MAX];
e7b84f9d 2519 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2520 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2521 get_sys_dev_type(hba->type));
0858eccf
AP		 2522 continue;
2523 }
2524 result = 0;
2525 }
2526
2527 if (controller_path && result == 1) {
2528 pr_err("no active Intel(R) RAID controller found under %s\n",
2529 controller_path);
2530 return result;
2531 }
2532
5e1d6128 2533 const struct orom_entry *entry;
0858eccf 2534
5e1d6128 2535 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2536 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2537 print_imsm_capability(&entry->orom);
32716c51
PB		 2538 printf(" 3rd party NVMe :%s supported\n",
2539 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2540 for (hba = list; hba; hba = hba->next) {
2541 if (hba->type == SYS_DEV_VMD) {
2542 char buf[PATH_MAX];
60f0f54d
PB		 2543 printf(" I/O Controller : %s (%s)\n",
2544 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2545 if (print_vmd_attached_devs(hba)) {
2546 if (verbose > 0)
2547 pr_err("failed to get devices attached to VMD domain.\n");
2548 result |= 2;
2549 }
60f0f54d
PB		 2550 }
2551 }
07cb1e57 2552 printf("\n");
60f0f54d
PB		 2553 continue;
2554 }
0858eccf 2555
60f0f54d
PB		 2556 print_imsm_capability(&entry->orom);
2557 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2558 for (hba = list; hba; hba = hba->next) {
2559 if (hba->type == SYS_DEV_NVME)
2560 printf(" NVMe Device : %s\n", hba->path);
2561 }
60f0f54d 2562 printf("\n");
0858eccf
AP		 2563 continue;
2564 }
2565
2566 struct devid_list *devid;
5e1d6128 2567 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2568 hba = device_by_id(devid->devid);
2569 if (!hba)
2570 continue;
2571
9eafa1de
MN		 2572 printf(" I/O Controller : %s (%s)\n",
2573 hba->path, get_sys_dev_type(hba->type));
2574 if (hba->type == SYS_DEV_SATA) {
2575 host_base = ahci_get_port_count(hba->path, &port_count);
2576 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2577 if (verbose > 0)
7a862a02 2578 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2579 result |= 2;
2580 }
120dc887
LM		 2581 }
2582 }
0858eccf 2583 printf("\n");
d665cc31 2584 }
155cbb4c 2585
120dc887 2586 return result;
d665cc31 2587}
e50cf220 2588
9eafa1de 2589static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2590{
e50cf220
MN		 2591 struct sys_dev *list, *hba;
2592 int result=1;
2593
2594 list = find_intel_devices();
2595 if (!list) {
2596 if (verbose > 0)
2597 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2598 result = 2;
e50cf220
MN		 2599 return result;
2600 }
2601
2602 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2603 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2604 continue;
60f0f54d
PB		 2605 if (!find_imsm_capability(hba) && verbose > 0) {
2606 char buf[PATH_MAX];
2607 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2608 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2609 }
0858eccf 2610 else
e50cf220 2611 result = 0;
e50cf220
MN		 2612 }
2613
5e1d6128 2614 const struct orom_entry *entry;
0858eccf 2615
60f0f54d
PB		 2616 for (entry = orom_entries; entry; entry = entry->next) {
2617 if (entry->type == SYS_DEV_VMD) {
2618 for (hba = list; hba; hba = hba->next)
2619 print_imsm_capability_export(&entry->orom);
2620 continue;
2621 }
5e1d6128 2622 print_imsm_capability_export(&entry->orom);
60f0f54d 2623 }
0858eccf 2624
e50cf220
MN		 2625 return result;
2626}
2627
cdddbdbc
DW		 2628static int match_home_imsm(struct supertype *st, char *homehost)
2629{
5115ca67
DW		 2630 /* the imsm metadata format does not specify any host
2631 * identification information. We return -1 since we can never
2632 * confirm nor deny whether a given array is "meant" for this
148acb7b 2633 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2634 * exclude member disks that do not belong, and we rely on
2635 * mdadm.conf to specify the arrays that should be assembled.
2636 * Auto-assembly may still pick up "foreign" arrays.
2637 */
cdddbdbc 2638
9362c1c8 2639 return -1;
cdddbdbc
DW		 2640}
2641
2642static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2643{
51006d85
N		 2644 /* The uuid returned here is used for:
2645 * uuid to put into bitmap file (Create, Grow)
2646 * uuid for backup header when saving critical section (Grow)
2647 * comparing uuids when re-adding a device into an array
2648 * In these cases the uuid required is that of the data-array,
2649 * not the device-set.
2650 * uuid to recognise same set when adding a missing device back
2651 * to an array. This is a uuid for the device-set.
1011e834 2652 *
51006d85
N		 2653 * For each of these we can make do with a truncated
2654 * or hashed uuid rather than the original, as long as
2655 * everyone agrees.
2656 * In each case the uuid required is that of the data-array,
2657 * not the device-set.
43dad3d6 2658 */
51006d85
N		 2659 /* imsm does not track uuid's so we synthesis one using sha1 on
2660 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2661 * - the orig_family_num of the container
51006d85
N		 2662 * - the index number of the volume
2663 * - the 'serial' number of the volume.
2664 * Hopefully these are all constant.
2665 */
2666 struct intel_super *super = st->sb;
43dad3d6 2667
51006d85
N		 2668 char buf[20];
2669 struct sha1_ctx ctx;
2670 struct imsm_dev *dev = NULL;
148acb7b 2671 __u32 family_num;
51006d85 2672
148acb7b
DW		 2673 /* some mdadm versions failed to set ->orig_family_num, in which
2674 * case fall back to ->family_num. orig_family_num will be
2675 * fixed up with the first metadata update.
2676 */
2677 family_num = super->anchor->orig_family_num;
2678 if (family_num == 0)
2679 family_num = super->anchor->family_num;
51006d85 2680 sha1_init_ctx(&ctx);
92bd8f8d 2681 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2682 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2683 if (super->current_vol >= 0)
2684 dev = get_imsm_dev(super, super->current_vol);
2685 if (dev) {
2686 __u32 vol = super->current_vol;
2687 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2688 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2689 }
2690 sha1_finish_ctx(&ctx, buf);
2691 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2692}
2693
0d481d37 2694#if 0
4f5bc454
DW		 2695static void
2696get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2697{
cdddbdbc
DW		 2698 __u8 *v = get_imsm_version(mpb);
2699 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2700 char major[] = { 0, 0, 0 };
2701 char minor[] = { 0 ,0, 0 };
2702 char patch[] = { 0, 0, 0 };
2703 char *ver_parse[] = { major, minor, patch };
2704 int i, j;
2705
2706 i = j = 0;
2707 while (*v != '\0' && v < end) {
2708 if (*v != '.' && j < 2)
2709 ver_parse[i][j++] = *v;
2710 else {
2711 i++;
2712 j = 0;
2713 }
2714 v++;
2715 }
2716
4f5bc454
DW		 2717 *m = strtol(minor, NULL, 0);
2718 *p = strtol(patch, NULL, 0);
2719}
0d481d37 2720#endif
4f5bc454 2721
1e5c6983
DW		 2722static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2723{
2724 /* migr_strip_size when repairing or initializing parity */
238c0a71 2725 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2726 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2727
2728 switch (get_imsm_raid_level(map)) {
2729 case 5:
2730 case 10:
2731 return chunk;
2732 default:
2733 return 128*1024 >> 9;
2734 }
2735}
2736
2737static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2738{
2739 /* migr_strip_size when rebuilding a degraded disk, no idea why
2740 * this is different than migr_strip_size_resync(), but it's good
2741 * to be compatible
2742 */
238c0a71 2743 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2744 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2745
2746 switch (get_imsm_raid_level(map)) {
2747 case 1:
2748 case 10:
2749 if (map->num_members % map->num_domains == 0)
2750 return 128*1024 >> 9;
2751 else
2752 return chunk;
2753 case 5:
2754 return max((__u32) 64*1024 >> 9, chunk);
2755 default:
2756 return 128*1024 >> 9;
2757 }
2758}
2759
2760static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2761{
238c0a71
AK		 2762 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2763 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2764 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2765 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2766
2767 return max((__u32) 1, hi_chunk / lo_chunk);
2768}
2769
2770static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2771{
238c0a71 2772 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2773 int level = get_imsm_raid_level(lo);
2774
2775 if (level == 1 || level == 10) {
238c0a71 2776 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2777
2778 return hi->num_domains;
2779 } else
2780 return num_stripes_per_unit_resync(dev);
2781}
2782
9529d343 2783static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2784{
2785 /* named 'imsm_' because raid0, raid1 and raid10
2786 * counter-intuitively have the same number of data disks
2787 */
1e5c6983
DW		 2788 switch (get_imsm_raid_level(map)) {
2789 case 0:
36fd8ccc
AK		 2790 return map->num_members;
2791 break;
1e5c6983
DW		 2792 case 1:
2793 case 10:
36fd8ccc 2794 return map->num_members/2;
1e5c6983
DW		 2795 case 5:
2796 return map->num_members - 1;
2797 default:
1ade5cc1 2798 dprintf("unsupported raid level\n");
1e5c6983
DW		 2799 return 0;
2800 }
2801}
2802
44490938
MD		 2803static unsigned long long calc_component_size(struct imsm_map *map,
2804 struct imsm_dev *dev)
2805{
2806 unsigned long long component_size;
2807 unsigned long long dev_size = imsm_dev_size(dev);
2808 unsigned long long calc_dev_size = 0;
2809 unsigned int member_disks = imsm_num_data_members(map);
2810
2811 if (member_disks == 0)
2812 return 0;
2813
2814 component_size = per_dev_array_size(map);
2815 calc_dev_size = component_size * member_disks;
2816
2817 /* Component size is rounded to 1MB so difference between size from
2818 * metadata and size calculated from num_data_stripes equals up to
2819 * 2048 blocks per each device. If the difference is higher it means
2820 * that array size was expanded and num_data_stripes was not updated.
2821 */
2822 if ((unsigned int)abs(calc_dev_size - dev_size) >
2823 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2824 component_size = dev_size / member_disks;
2825 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2826 component_size / map->blocks_per_strip,
2827 num_data_stripes(map));
2828 }
2829
2830 return component_size;
2831}
2832
1e5c6983
DW		 2833static __u32 parity_segment_depth(struct imsm_dev *dev)
2834{
238c0a71 2835 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2836 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2837
2838 switch(get_imsm_raid_level(map)) {
2839 case 1:
2840 case 10:
2841 return chunk * map->num_domains;
2842 case 5:
2843 return chunk * map->num_members;
2844 default:
2845 return chunk;
2846 }
2847}
2848
2849static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2850{
238c0a71 2851 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2852 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2853 __u32 strip = block / chunk;
2854
2855 switch (get_imsm_raid_level(map)) {
2856 case 1:
2857 case 10: {
2858 __u32 vol_strip = (strip * map->num_domains) + 1;
2859 __u32 vol_stripe = vol_strip / map->num_members;
2860
2861 return vol_stripe * chunk + block % chunk;
2862 } case 5: {
2863 __u32 stripe = strip / (map->num_members - 1);
2864
2865 return stripe * chunk + block % chunk;
2866 }
2867 default:
2868 return 0;
2869 }
2870}
2871
c47b0ff6
AK		 2872static __u64 blocks_per_migr_unit(struct intel_super *super,
2873 struct imsm_dev *dev)
1e5c6983
DW		 2874{
2875 /* calculate the conversion factor between per member 'blocks'
2876 * (md/{resync,rebuild}_start) and imsm migration units, return
2877 * 0 for the 'not migrating' and 'unsupported migration' cases
2878 */
2879 if (!dev->vol.migr_state)
2880 return 0;
2881
2882 switch (migr_type(dev)) {
c47b0ff6
AK		 2883 case MIGR_GEN_MIGR: {
2884 struct migr_record *migr_rec = super->migr_rec;
2885 return __le32_to_cpu(migr_rec->blocks_per_unit);
2886 }
1e5c6983
DW		 2887 case MIGR_VERIFY:
2888 case MIGR_REPAIR:
2889 case MIGR_INIT: {
238c0a71 2890 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2891 __u32 stripes_per_unit;
2892 __u32 blocks_per_unit;
2893 __u32 parity_depth;
2894 __u32 migr_chunk;
2895 __u32 block_map;
2896 __u32 block_rel;
2897 __u32 segment;
2898 __u32 stripe;
2899 __u8 disks;
2900
2901 /* yes, this is really the translation of migr_units to
2902 * per-member blocks in the 'resync' case
2903 */
2904 stripes_per_unit = num_stripes_per_unit_resync(dev);
2905 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2906 disks = imsm_num_data_members(map);
1e5c6983 2907 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2908 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2909 segment = blocks_per_unit / stripe;
2910 block_rel = blocks_per_unit - segment * stripe;
2911 parity_depth = parity_segment_depth(dev);
2912 block_map = map_migr_block(dev, block_rel);
2913 return block_map + parity_depth * segment;
2914 }
2915 case MIGR_REBUILD: {
2916 __u32 stripes_per_unit;
2917 __u32 migr_chunk;
2918
2919 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2920 migr_chunk = migr_strip_blocks_rebuild(dev);
2921 return migr_chunk * stripes_per_unit;
2922 }
1e5c6983
DW		 2923 case MIGR_STATE_CHANGE:
2924 default:
2925 return 0;
2926 }
2927}
2928
c2c087e6
DW		 2929static int imsm_level_to_layout(int level)
2930{
2931 switch (level) {
2932 case 0:
2933 case 1:
2934 return 0;
2935 case 5:
2936 case 6:
a380c027 2937 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2938 case 10:
c92a2527 2939 return 0x102;
c2c087e6 2940 }
a18a888e 2941 return UnSet;
c2c087e6
DW		 2942}
2943
8e59f3d8
AK		 2944/*******************************************************************************
2945 * Function: read_imsm_migr_rec
2946 * Description: Function reads imsm migration record from last sector of disk
2947 * Parameters:
2948 * fd : disk descriptor
2949 * super : metadata info
2950 * Returns:
2951 * 0 : success,
2952 * -1 : fail
2953 ******************************************************************************/
2954static int read_imsm_migr_rec(int fd, struct intel_super *super)
2955{
2956 int ret_val = -1;
de44e46f 2957 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 2958 unsigned long long dsize;
2959
2960 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 2961 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
2962 SEEK_SET) < 0) {
e7b84f9d
N		 2963 pr_err("Cannot seek to anchor block: %s\n",
2964 strerror(errno));
8e59f3d8
AK		 2965 goto out;
2966 }
466070ad 2967 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 2968 MIGR_REC_BUF_SECTORS*sector_size) !=
2969 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 2970 pr_err("Cannot read migr record block: %s\n",
2971 strerror(errno));
8e59f3d8
AK		 2972 goto out;
2973 }
2974 ret_val = 0;
de44e46f
PB		 2975 if (sector_size == 4096)
2976 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 2977
2978out:
2979 return ret_val;
2980}
2981
3136abe5
AK		 2982static struct imsm_dev *imsm_get_device_during_migration(
2983 struct intel_super *super)
2984{
2985
2986 struct intel_dev *dv;
2987
2988 for (dv = super->devlist; dv; dv = dv->next) {
2989 if (is_gen_migration(dv->dev))
2990 return dv->dev;
2991 }
2992 return NULL;
2993}
2994
8e59f3d8
AK		 2995/*******************************************************************************
2996 * Function: load_imsm_migr_rec
2997 * Description: Function reads imsm migration record (it is stored at the last
2998 * sector of disk)
2999 * Parameters:
3000 * super : imsm internal array info
3001 * info : general array info
3002 * Returns:
3003 * 0 : success
3004 * -1 : fail
4c965cc9 3005 * -2 : no migration in progress
8e59f3d8
AK		 3006 ******************************************************************************/
3007static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3008{
3009 struct mdinfo *sd;
594dc1b8 3010 struct dl *dl;
8e59f3d8
AK		 3011 char nm[30];
3012 int retval = -1;
3013 int fd = -1;
3136abe5 3014 struct imsm_dev *dev;
594dc1b8 3015 struct imsm_map *map;
b4ab44d8 3016 int slot = -1;
3136abe5
AK		 3017
3018 /* find map under migration */
3019 dev = imsm_get_device_during_migration(super);
3020 /* nothing to load,no migration in progress?
3021 */
3022 if (dev == NULL)
4c965cc9 3023 return -2;
8e59f3d8
AK		 3024
3025 if (info) {
3026 for (sd = info->devs ; sd ; sd = sd->next) {
3027 /* read only from one of the first two slots */
12fe93e9
TM		 3028 if ((sd->disk.raid_disk < 0) ||
3029 (sd->disk.raid_disk > 1))
8e59f3d8 3030 continue;
3136abe5 3031
8e59f3d8
AK		 3032 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3033 fd = dev_open(nm, O_RDONLY);
3034 if (fd >= 0)
3035 break;
3036 }
3037 }
3038 if (fd < 0) {
12fe93e9 3039 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3040 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3041 /* skip spare and failed disks
3042 */
3043 if (dl->index < 0)
3044 continue;
8e59f3d8 3045 /* read only from one of the first two slots */
3136abe5
AK		 3046 if (map)
3047 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3048 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3049 continue;
3050 sprintf(nm, "%d:%d", dl->major, dl->minor);
3051 fd = dev_open(nm, O_RDONLY);
3052 if (fd >= 0)
3053 break;
3054 }
3055 }
3056 if (fd < 0)
3057 goto out;
3058 retval = read_imsm_migr_rec(fd, super);
3059
3060out:
3061 if (fd >= 0)
3062 close(fd);
3063 return retval;
3064}
3065
c17608ea
AK		 3066/*******************************************************************************
3067 * function: imsm_create_metadata_checkpoint_update
3068 * Description: It creates update for checkpoint change.
3069 * Parameters:
3070 * super : imsm internal array info
3071 * u : pointer to prepared update
3072 * Returns:
3073 * Uptate length.
3074 * If length is equal to 0, input pointer u contains no update
3075 ******************************************************************************/
3076static int imsm_create_metadata_checkpoint_update(
3077 struct intel_super *super,
3078 struct imsm_update_general_migration_checkpoint **u)
3079{
3080
3081 int update_memory_size = 0;
3082
1ade5cc1 3083 dprintf("(enter)\n");
c17608ea
AK		 3084
3085 if (u == NULL)
3086 return 0;
3087 *u = NULL;
3088
3089 /* size of all update data without anchor */
3090 update_memory_size =
3091 sizeof(struct imsm_update_general_migration_checkpoint);
3092
503975b9 3093 *u = xcalloc(1, update_memory_size);
c17608ea 3094 if (*u == NULL) {
1ade5cc1 3095 dprintf("error: cannot get memory\n");
c17608ea
AK		 3096 return 0;
3097 }
3098 (*u)->type = update_general_migration_checkpoint;
3099 (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit);
1ade5cc1 3100 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3101
3102 return update_memory_size;
3103}
3104
c17608ea
AK		 3105static void imsm_update_metadata_locally(struct supertype *st,
3106 void *buf, int len);
3107
687629c2
AK		 3108/*******************************************************************************
3109 * Function: write_imsm_migr_rec
3110 * Description: Function writes imsm migration record
3111 * (at the last sector of disk)
3112 * Parameters:
3113 * super : imsm internal array info
3114 * Returns:
3115 * 0 : success
3116 * -1 : if fail
3117 ******************************************************************************/
3118static int write_imsm_migr_rec(struct supertype *st)
3119{
3120 struct intel_super *super = st->sb;
de44e46f 3121 unsigned int sector_size = super->sector_size;
687629c2
AK		 3122 unsigned long long dsize;
3123 char nm[30];
3124 int fd = -1;
3125 int retval = -1;
3126 struct dl *sd;
c17608ea
AK		 3127 int len;
3128 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3129 struct imsm_dev *dev;
594dc1b8 3130 struct imsm_map *map;
3136abe5
AK		 3131
3132 /* find map under migration */
3133 dev = imsm_get_device_during_migration(super);
3134 /* if no migration, write buffer anyway to clear migr_record
3135 * on disk based on first available device
3136 */
3137 if (dev == NULL)
3138 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3139 super->current_vol);
3140
44bfe6df 3141 map = get_imsm_map(dev, MAP_0);
687629c2 3142
de44e46f
PB		 3143 if (sector_size == 4096)
3144 convert_to_4k_imsm_migr_rec(super);
687629c2 3145 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3146 int slot = -1;
3136abe5
AK		 3147
3148 /* skip failed and spare devices */
3149 if (sd->index < 0)
3150 continue;
687629c2 3151 /* write to 2 first slots only */
3136abe5
AK		 3152 if (map)
3153 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3154 if (map == NULL || slot > 1 || slot < 0)
687629c2 3155 continue;
3136abe5 3156
687629c2
AK		 3157 sprintf(nm, "%d:%d", sd->major, sd->minor);
3158 fd = dev_open(nm, O_RDWR);
3159 if (fd < 0)
3160 continue;
3161 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3162 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3163 SEEK_SET) < 0) {
e7b84f9d
N		 3164 pr_err("Cannot seek to anchor block: %s\n",
3165 strerror(errno));
687629c2
AK		 3166 goto out;
3167 }
466070ad 3168 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3169 MIGR_REC_BUF_SECTORS*sector_size) !=
3170 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3171 pr_err("Cannot write migr record block: %s\n",
3172 strerror(errno));
687629c2
AK		 3173 goto out;
3174 }
3175 close(fd);
3176 fd = -1;
3177 }
de44e46f
PB		 3178 if (sector_size == 4096)
3179 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3180 /* update checkpoint information in metadata */
3181 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3182 if (len <= 0) {
3183 dprintf("imsm: Cannot prepare update\n");
3184 goto out;
3185 }
3186 /* update metadata locally */
3187 imsm_update_metadata_locally(st, u, len);
3188 /* and possibly remotely */
3189 if (st->update_tail) {
3190 append_metadata_update(st, u, len);
3191 /* during reshape we do all work inside metadata handler
3192 * manage_reshape(), so metadata update has to be triggered
3193 * insida it
3194 */
3195 flush_metadata_updates(st);
3196 st->update_tail = &st->updates;
3197 } else
3198 free(u);
687629c2
AK		 3199
3200 retval = 0;
3201 out:
3202 if (fd >= 0)
3203 close(fd);
3204 return retval;
3205}
3206
e2962bfc
AK		 3207/* spare/missing disks activations are not allowe when
3208 * array/container performs reshape operation, because
3209 * all arrays in container works on the same disks set
3210 */
3211int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3212{
3213 int rv = 0;
3214 struct intel_dev *i_dev;
3215 struct imsm_dev *dev;
3216
3217 /* check whole container
3218 */
3219 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3220 dev = i_dev->dev;
3ad25638 3221 if (is_gen_migration(dev)) {
e2962bfc
AK		 3222 /* No repair during any migration in container
3223 */
3224 rv = 1;
3225 break;
3226 }
3227 }
3228 return rv;
3229}
3e684231 3230static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3231 int chunk_size,
f36a9ecd 3232 unsigned int sector_size,
c41e00b2
AK		 3233 unsigned long long component_size)
3234{
3e684231 3235 unsigned int component_size_alignment;
c41e00b2 3236
3e684231 3237 /* check component size alignment
c41e00b2 3238 */
3e684231 3239 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3240
3e684231 3241 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3242 level, chunk_size, component_size,
3e684231 3243 component_size_alignment);
c41e00b2 3244
3e684231
MZ		 3245 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3246 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3247 component_size);
3e684231 3248 component_size -= component_size_alignment;
1ade5cc1 3249 dprintf_cont("to %llu (%i).\n",
3e684231 3250 component_size, component_size_alignment);
c41e00b2
AK		 3251 }
3252
3253 return component_size;
3254}
e2962bfc 3255
2432ce9b
AP		 3256static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3257{
3258 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3259 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3260
3261 return pba_of_lba0(map) +
3262 (num_data_stripes(map) * map->blocks_per_strip);
3263}
3264
a5d85af7 3265static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3266{
3267 struct intel_super *super = st->sb;
c47b0ff6 3268 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3269 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3270 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3271 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3272 struct imsm_map *map_to_analyse = map;
efb30e7f 3273 struct dl *dl;
a5d85af7 3274 int map_disks = info->array.raid_disks;
bf5a934a 3275
95eeceeb 3276 memset(info, 0, sizeof(*info));
b335e593
AK		 3277 if (prev_map)
3278 map_to_analyse = prev_map;
3279
ca0748fa 3280 dl = super->current_disk;
9894ec0d 3281
bf5a934a 3282 info->container_member = super->current_vol;
cd0430a1 3283 info->array.raid_disks = map->num_members;
b335e593 3284 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3285 info->array.layout = imsm_level_to_layout(info->array.level);
3286 info->array.md_minor = -1;
3287 info->array.ctime = 0;
3288 info->array.utime = 0;
b335e593
AK		 3289 info->array.chunk_size =
3290 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3291 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3292 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3293 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3294
3f510843 3295 if (is_gen_migration(dev)) {
3f83228a 3296 info->reshape_active = 1;
b335e593
AK		 3297 info->new_level = get_imsm_raid_level(map);
3298 info->new_layout = imsm_level_to_layout(info->new_level);
3299 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3300 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3301 if (info->delta_disks) {
3302 /* this needs to be applied to every array
3303 * in the container.
3304 */
81219e70 3305 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3306 }
3f83228a
N		 3307 /* We shape information that we give to md might have to be
3308 * modify to cope with md's requirement for reshaping arrays.
3309 * For example, when reshaping a RAID0, md requires it to be
3310 * presented as a degraded RAID4.
3311 * Also if a RAID0 is migrating to a RAID5 we need to specify
3312 * the array as already being RAID5, but the 'before' layout
3313 * is a RAID4-like layout.
3314 */
3315 switch (info->array.level) {
3316 case 0:
3317 switch(info->new_level) {
3318 case 0:
3319 /* conversion is happening as RAID4 */
3320 info->array.level = 4;
3321 info->array.raid_disks += 1;
3322 break;
3323 case 5:
3324 /* conversion is happening as RAID5 */
3325 info->array.level = 5;
3326 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3327 info->delta_disks -= 1;
3328 break;
3329 default:
3330 /* FIXME error message */
3331 info->array.level = UnSet;
3332 break;
3333 }
3334 break;
3335 }
b335e593
AK		 3336 } else {
3337 info->new_level = UnSet;
3338 info->new_layout = UnSet;
3339 info->new_chunk = info->array.chunk_size;
3f83228a 3340 info->delta_disks = 0;
b335e593 3341 }
ca0748fa 3342
efb30e7f
DW		 3343 if (dl) {
3344 info->disk.major = dl->major;
3345 info->disk.minor = dl->minor;
ca0748fa 3346 info->disk.number = dl->index;
656b6b5a
N		 3347 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3348 dl->index);
efb30e7f 3349 }
bf5a934a 3350
5551b113 3351 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3352 info->component_size = calc_component_size(map, dev);
3e684231 3353 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3354 info->array.level,
3355 info->array.chunk_size,
f36a9ecd 3356 super->sector_size,
c41e00b2 3357 info->component_size);
5e46202e 3358 info->bb.supported = 1;
139dae11 3359
301406c9 3360 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3361 info->recovery_start = MaxSector;
bf5a934a 3362
c2462068
PB		 3363 if (info->array.level == 5 &&
3364 (dev->rwh_policy == RWH_DISTRIBUTED ||
3365 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3366 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3367 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3368 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3369 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3370 else
3371 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3372 >> 9;
2432ce9b
AP		 3373 } else if (info->array.level <= 0) {
3374 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3375 } else {
3376 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3377 }
3378
d2e6d5d6 3379 info->reshape_progress = 0;
b6796ce1 3380 info->resync_start = MaxSector;
b9172665 3381 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3382 !(info->array.state & 1)) &&
b9172665 3383 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3384 info->resync_start = 0;
b6796ce1
AK		 3385 }
3386 if (dev->vol.migr_state) {
1e5c6983
DW		 3387 switch (migr_type(dev)) {
3388 case MIGR_REPAIR:
3389 case MIGR_INIT: {
c47b0ff6
AK		 3390 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3391 dev);
1e5c6983
DW		 3392 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3393
3394 info->resync_start = blocks_per_unit * units;
3395 break;
3396 }
d2e6d5d6 3397 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3398 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3399 dev);
3400 __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit);
04fa9523
AK		 3401 unsigned long long array_blocks;
3402 int used_disks;
d2e6d5d6 3403
befb629b
AK		 3404 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3405 (units <
3406 (__le32_to_cpu(migr_rec->num_migr_units)-1)) &&
3407 (super->migr_rec->rec_status ==
3408 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3409 units++;
3410
d2e6d5d6 3411 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3412
7a862a02 3413 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3414 (unsigned long long)units,
3415 (unsigned long long)blocks_per_unit,
3416 info->reshape_progress);
75156c46 3417
9529d343 3418 used_disks = imsm_num_data_members(prev_map);
75156c46 3419 if (used_disks > 0) {
44490938 3420 array_blocks = per_dev_array_size(map) *
75156c46 3421 used_disks;
b53bfba6
TM		 3422 info->custom_array_size =
3423 round_size_to_mb(array_blocks,
3424 used_disks);
3425
75156c46 3426 }
d2e6d5d6 3427 }
1e5c6983
DW		 3428 case MIGR_VERIFY:
3429 /* we could emulate the checkpointing of
3430 * 'sync_action=check' migrations, but for now
3431 * we just immediately complete them
3432 */
3433 case MIGR_REBUILD:
3434 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3435 case MIGR_STATE_CHANGE:
3436 /* FIXME handle other migrations */
3437 default:
3438 /* we are not dirty, so... */
3439 info->resync_start = MaxSector;
3440 }
b6796ce1 3441 }
301406c9
DW		 3442
3443 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3444 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3445
f35f2525
N		 3446 info->array.major_version = -1;
3447 info->array.minor_version = -2;
4dd2df09 3448 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3449 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3450 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3451
3452 if (dmap) {
3453 int i, j;
3454 for (i=0; i<map_disks; i++) {
3455 dmap[i] = 0;
3456 if (i < info->array.raid_disks) {
3457 struct imsm_disk *dsk;
238c0a71 3458 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3459 dsk = get_imsm_disk(super, j);
3460 if (dsk && (dsk->status & CONFIGURED_DISK))
3461 dmap[i] = 1;
3462 }
3463 }
3464 }
81ac8b4d 3465}
bf5a934a 3466
3b451610
AK		 3467static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3468 int failed, int look_in_map);
3469
3470static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3471 int look_in_map);
3472
3473static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3474{
3475 if (is_gen_migration(dev)) {
3476 int failed;
3477 __u8 map_state;
3478 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3479
3480 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3481 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3482 if (map2->map_state != map_state) {
3483 map2->map_state = map_state;
3484 super->updates_pending++;
3485 }
3486 }
3487}
97b4d0e9
DW		 3488
3489static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3490{
3491 struct dl *d;
3492
3493 for (d = super->missing; d; d = d->next)
3494 if (d->index == index)
3495 return &d->disk;
3496 return NULL;
3497}
3498
a5d85af7 3499static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3500{
3501 struct intel_super *super = st->sb;
4f5bc454 3502 struct imsm_disk *disk;
a5d85af7 3503 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3504 int max_enough = -1;
3505 int i;
3506 struct imsm_super *mpb;
4f5bc454 3507
bf5a934a 3508 if (super->current_vol >= 0) {
a5d85af7 3509 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3510 return;
3511 }
95eeceeb 3512 memset(info, 0, sizeof(*info));
d23fe947
DW		 3513
3514 /* Set raid_disks to zero so that Assemble will always pull in valid
3515 * spares
3516 */
3517 info->array.raid_disks = 0;
cdddbdbc
DW		 3518 info->array.level = LEVEL_CONTAINER;
3519 info->array.layout = 0;
3520 info->array.md_minor = -1;
1011e834 3521 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3522 info->array.utime = 0;
3523 info->array.chunk_size = 0;
3524
3525 info->disk.major = 0;
3526 info->disk.minor = 0;
cdddbdbc 3527 info->disk.raid_disk = -1;
c2c087e6 3528 info->reshape_active = 0;
f35f2525
N		 3529 info->array.major_version = -1;
3530 info->array.minor_version = -2;
c2c087e6 3531 strcpy(info->text_version, "imsm");
a67dd8cc 3532 info->safe_mode_delay = 0;
c2c087e6
DW		 3533 info->disk.number = -1;
3534 info->disk.state = 0;
c5afc314 3535 info->name[0] = 0;
921d9e16 3536 info->recovery_start = MaxSector;
3ad25638 3537 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3538 info->bb.supported = 1;
c2c087e6 3539
97b4d0e9 3540 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3541 mpb = super->anchor;
b7d81a38 3542 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3543
ab3cb6b3
N		 3544 for (i = 0; i < mpb->num_raid_devs; i++) {
3545 struct imsm_dev *dev = get_imsm_dev(super, i);
3546 int failed, enough, j, missing = 0;
3547 struct imsm_map *map;
3548 __u8 state;
97b4d0e9 3549
3b451610
AK		 3550 failed = imsm_count_failed(super, dev, MAP_0);
3551 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3552 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3553
3554 /* any newly missing disks?
3555 * (catches single-degraded vs double-degraded)
3556 */
3557 for (j = 0; j < map->num_members; j++) {
238c0a71 3558 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3559 __u32 idx = ord_to_idx(ord);
3560
20dc76d1
MT		 3561 if (super->disks && super->disks->index == (int)idx)
3562 info->disk.raid_disk = j;
3563
ab3cb6b3
N		 3564 if (!(ord & IMSM_ORD_REBUILD) &&
3565 get_imsm_missing(super, idx)) {
3566 missing = 1;
3567 break;
3568 }
97b4d0e9 3569 }
ab3cb6b3
N		 3570
3571 if (state == IMSM_T_STATE_FAILED)
3572 enough = -1;
3573 else if (state == IMSM_T_STATE_DEGRADED &&
3574 (state != map->map_state || missing))
3575 enough = 0;
3576 else /* we're normal, or already degraded */
3577 enough = 1;
d2bde6d3
AK		 3578 if (is_gen_migration(dev) && missing) {
3579 /* during general migration we need all disks
3580 * that process is running on.
3581 * No new missing disk is allowed.
3582 */
3583 max_enough = -1;
3584 enough = -1;
3585 /* no more checks necessary
3586 */
3587 break;
3588 }
ab3cb6b3
N		 3589 /* in the missing/failed disk case check to see
3590 * if at least one array is runnable
3591 */
3592 max_enough = max(max_enough, enough);
3593 }
1ade5cc1 3594 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3595 info->container_enough = max_enough;
97b4d0e9 3596
4a04ec6c 3597 if (super->disks) {
14e8215b
DW		 3598 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3599
b9f594fe 3600 disk = &super->disks->disk;
5551b113 3601 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3602 info->component_size = reserved;
25ed7e59 3603 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3604 /* we don't change info->disk.raid_disk here because
3605 * this state will be finalized in mdmon after we have
3606 * found the 'most fresh' version of the metadata
3607 */
25ed7e59 3608 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3609 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3610 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3611 }
a575e2a7
DW		 3612
3613 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3614 * ->compare_super may have updated the 'num_raid_devs' field for spares
3615 */
3616 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3617 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3618 else
3619 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3620
3621 /* I don't know how to compute 'map' on imsm, so use safe default */
3622 if (map) {
3623 int i;
3624 for (i = 0; i < map_disks; i++)
3625 map[i] = 1;
3626 }
3627
cdddbdbc
DW		 3628}
3629
5c4cd5da
AC		 3630/* allocates memory and fills disk in mdinfo structure
3631 * for each disk in array */
3632struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3633{
594dc1b8 3634 struct mdinfo *mddev;
5c4cd5da
AC		 3635 struct intel_super *super = st->sb;
3636 struct imsm_disk *disk;
3637 int count = 0;
3638 struct dl *dl;
3639 if (!super || !super->disks)
3640 return NULL;
3641 dl = super->disks;
503975b9 3642 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3643 while (dl) {
3644 struct mdinfo *tmp;
3645 disk = &dl->disk;
503975b9 3646 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3647 if (mddev->devs)
3648 tmp->next = mddev->devs;
3649 mddev->devs = tmp;
3650 tmp->disk.number = count++;
3651 tmp->disk.major = dl->major;
3652 tmp->disk.minor = dl->minor;
3653 tmp->disk.state = is_configured(disk) ?
3654 (1 << MD_DISK_ACTIVE) : 0;
3655 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3656 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3657 tmp->disk.raid_disk = -1;
3658 dl = dl->next;
3659 }
3660 return mddev;
3661}
3662
cdddbdbc
DW		 3663static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3664 char *update, char *devname, int verbose,
3665 int uuid_set, char *homehost)
3666{
f352c545
DW		 3667 /* For 'assemble' and 'force' we need to return non-zero if any
3668 * change was made. For others, the return value is ignored.
3669 * Update options are:
3670 * force-one : This device looks a bit old but needs to be included,
3671 * update age info appropriately.
3672 * assemble: clear any 'faulty' flag to allow this device to
3673 * be assembled.
3674 * force-array: Array is degraded but being forced, mark it clean
3675 * if that will be needed to assemble it.
3676 *
3677 * newdev: not used ????
3678 * grow: Array has gained a new device - this is currently for
3679 * linear only
3680 * resync: mark as dirty so a resync will happen.
3681 * name: update the name - preserving the homehost
6e46bf34 3682 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3683 *
3684 * Following are not relevant for this imsm:
3685 * sparc2.2 : update from old dodgey metadata
3686 * super-minor: change the preferred_minor number
3687 * summaries: update redundant counters.
f352c545
DW		 3688 * homehost: update the recorded homehost
3689 * _reshape_progress: record new reshape_progress position.
3690 */
6e46bf34
DW		 3691 int rv = 1;
3692 struct intel_super *super = st->sb;
3693 struct imsm_super *mpb;
f352c545 3694
6e46bf34
DW		 3695 /* we can only update container info */
3696 if (!super || super->current_vol >= 0 || !super->anchor)
3697 return 1;
3698
3699 mpb = super->anchor;
3700
81a5b4f5
N		 3701 if (strcmp(update, "uuid") == 0) {
3702 /* We take this to mean that the family_num should be updated.
3703 * However that is much smaller than the uuid so we cannot really
3704 * allow an explicit uuid to be given. And it is hard to reliably
3705 * know if one was.
3706 * So if !uuid_set we know the current uuid is random and just used
3707 * the first 'int' and copy it to the other 3 positions.
3708 * Otherwise we require the 4 'int's to be the same as would be the
3709 * case if we are using a random uuid. So an explicit uuid will be
3710 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3711 */
81a5b4f5
N		 3712 if (!uuid_set) {
3713 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3714 rv = 0;
81a5b4f5
N		 3715 } else {
3716 if (info->uuid[0] != info->uuid[1] ||
3717 info->uuid[1] != info->uuid[2] ||
3718 info->uuid[2] != info->uuid[3])
3719 rv = -1;
3720 else
3721 rv = 0;
6e46bf34 3722 }
81a5b4f5
N		 3723 if (rv == 0)
3724 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3725 } else if (strcmp(update, "assemble") == 0)
3726 rv = 0;
3727 else
1e2b2765 3728 rv = -1;
f352c545 3729
6e46bf34
DW		 3730 /* successful update? recompute checksum */
3731 if (rv == 0)
3732 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3733
3734 return rv;
cdddbdbc
DW		 3735}
3736
c2c087e6 3737static size_t disks_to_mpb_size(int disks)
cdddbdbc 3738{
c2c087e6 3739 size_t size;
cdddbdbc 3740
c2c087e6
DW		 3741 size = sizeof(struct imsm_super);
3742 size += (disks - 1) * sizeof(struct imsm_disk);
3743 size += 2 * sizeof(struct imsm_dev);
3744 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3745 size += (4 - 2) * sizeof(struct imsm_map);
3746 /* 4 possible disk_ord_tbl's */
3747 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3748 /* maximum bbm log */
3749 size += sizeof(struct bbm_log);
c2c087e6
DW		 3750
3751 return size;
3752}
3753
387fcd59
N		 3754static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3755 unsigned long long data_offset)
c2c087e6
DW		 3756{
3757 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3758 return 0;
3759
3760 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3761}
3762
ba2de7ba
DW		 3763static void free_devlist(struct intel_super *super)
3764{
3765 struct intel_dev *dv;
3766
3767 while (super->devlist) {
3768 dv = super->devlist->next;
3769 free(super->devlist->dev);
3770 free(super->devlist);
3771 super->devlist = dv;
3772 }
3773}
3774
3775static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3776{
3777 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3778}
3779
cdddbdbc
DW		 3780static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3781{
3782 /*
3783 * return:
3784 * 0 same, or first was empty, and second was copied
3785 * 1 second had wrong number
3786 * 2 wrong uuid
3787 * 3 wrong other info
3788 */
3789 struct intel_super *first = st->sb;
3790 struct intel_super *sec = tst->sb;
3791
5d500228
N		 3792 if (!first) {
3793 st->sb = tst->sb;
3794 tst->sb = NULL;
3795 return 0;
3796 }
8603ea6f
LM		 3797 /* in platform dependent environment test if the disks
3798 * use the same Intel hba
cb8f6859 3799 * If not on Intel hba at all, allow anything.
8603ea6f 3800 */
6b781d33
AP		 3801 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3802 if (first->hba->type != sec->hba->type) {
8603ea6f 3803 fprintf(stderr,
6b781d33
AP		 3804 "HBAs of devices do not match %s != %s\n",
3805 get_sys_dev_type(first->hba->type),
3806 get_sys_dev_type(sec->hba->type));
3807 return 3;
3808 }
3809 if (first->orom != sec->orom) {
3810 fprintf(stderr,
3811 "HBAs of devices do not match %s != %s\n",
3812 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3813 return 3;
3814 }
3815 }
cdddbdbc 3816
d23fe947
DW		 3817 /* if an anchor does not have num_raid_devs set then it is a free
3818 * floating spare
3819 */
3820 if (first->anchor->num_raid_devs > 0 &&
3821 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3822 /* Determine if these disks might ever have been
3823 * related. Further disambiguation can only take place
3824 * in load_super_imsm_all
3825 */
3826 __u32 first_family = first->anchor->orig_family_num;
3827 __u32 sec_family = sec->anchor->orig_family_num;
3828
f796af5d
DW		 3829 if (memcmp(first->anchor->sig, sec->anchor->sig,
3830 MAX_SIGNATURE_LENGTH) != 0)
3831 return 3;
3832
a2b97981
DW		 3833 if (first_family == 0)
3834 first_family = first->anchor->family_num;
3835 if (sec_family == 0)
3836 sec_family = sec->anchor->family_num;
3837
3838 if (first_family != sec_family)
d23fe947 3839 return 3;
f796af5d 3840
d23fe947 3841 }
cdddbdbc 3842
3e372e5a
DW		 3843 /* if 'first' is a spare promote it to a populated mpb with sec's
3844 * family number
3845 */
3846 if (first->anchor->num_raid_devs == 0 &&
3847 sec->anchor->num_raid_devs > 0) {
78d30f94 3848 int i;
ba2de7ba
DW		 3849 struct intel_dev *dv;
3850 struct imsm_dev *dev;
78d30f94
DW		 3851
3852 /* we need to copy raid device info from sec if an allocation
3853 * fails here we don't associate the spare
3854 */
3855 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3856 dv = xmalloc(sizeof(*dv));
3857 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3858 dv->dev = dev;
3859 dv->index = i;
3860 dv->next = first->devlist;
3861 first->devlist = dv;
78d30f94 3862 }
709743c5 3863 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3864 /* allocation failure */
3865 free_devlist(first);
e12b3daa 3866 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3867 return 3;
78d30f94 3868 }
3e372e5a 3869 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3870 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3871 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3872 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3873 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3874 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3875 }
3876
cdddbdbc
DW		 3877 return 0;
3878}
3879
0030e8d6
DW		 3880static void fd2devname(int fd, char *name)
3881{
3882 struct stat st;
3883 char path[256];
33a6535d 3884 char dname[PATH_MAX];
0030e8d6
DW		 3885 char *nm;
3886 int rv;
3887
3888 name[0] = '\0';
3889 if (fstat(fd, &st) != 0)
3890 return;
3891 sprintf(path, "/sys/dev/block/%d:%d",
3892 major(st.st_rdev), minor(st.st_rdev));
3893
9cf014ec 3894 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3895 if (rv <= 0)
3896 return;
9587c373 3897
0030e8d6
DW		 3898 dname[rv] = '\0';
3899 nm = strrchr(dname, '/');
7897de29
JS		 3900 if (nm) {
3901 nm++;
3902 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3903 }
0030e8d6
DW		 3904}
3905
21e9380b
AP		 3906static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3907{
3908 char path[60];
3909 char *name = fd2kname(fd);
3910
3911 if (!name)
3912 return 1;
3913
3914 if (strncmp(name, "nvme", 4) != 0)
3915 return 1;
3916
3917 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3918
3919 return load_sys(path, buf, buf_len);
3920}
3921
cdddbdbc
DW		 3922extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3923
3924static int imsm_read_serial(int fd, char *devname,
3925 __u8 serial[MAX_RAID_SERIAL_LEN])
3926{
21e9380b 3927 char buf[50];
cdddbdbc 3928 int rv;
1f24f035 3929 int len;
316e2bf4
DW		 3930 char *dest;
3931 char *src;
21e9380b
AP		 3932 unsigned int i;
3933
3934 memset(buf, 0, sizeof(buf));
cdddbdbc 3935
21e9380b 3936 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3937
21e9380b
AP		 3938 if (rv)
3939 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3940
40ebbb9c 3941 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3942 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3943 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3944 return 0;
3945 }
3946
cdddbdbc
DW		 3947 if (rv != 0) {
3948 if (devname)
e7b84f9d
N		 3949 pr_err("Failed to retrieve serial for %s\n",
3950 devname);
cdddbdbc
DW		 3951 return rv;
3952 }
3953
316e2bf4
DW		 3954 /* trim all whitespace and non-printable characters and convert
3955 * ':' to ';'
3956 */
21e9380b
AP		 3957 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
3958 src = &buf[i];
316e2bf4
DW		 3959 if (*src > 0x20) {
3960 /* ':' is reserved for use in placeholder serial
3961 * numbers for missing disks
3962 */
3963 if (*src == ':')
3964 *dest++ = ';';
3965 else
3966 *dest++ = *src;
3967 }
3968 }
21e9380b
AP		 3969 len = dest - buf;
3970 dest = buf;
316e2bf4
DW		 3971
3972 /* truncate leading characters */
3973 if (len > MAX_RAID_SERIAL_LEN) {
3974 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 3975 len = MAX_RAID_SERIAL_LEN;
316e2bf4 3976 }
5c3db629 3977
5c3db629 3978 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 3979 memcpy(serial, dest, len);
cdddbdbc
DW		 3980
3981 return 0;
3982}
3983
1f24f035
DW		 3984static int serialcmp(__u8 *s1, __u8 *s2)
3985{
3986 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
3987}
3988
3989static void serialcpy(__u8 *dest, __u8 *src)
3990{
3991 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
3992}
3993
54c2c1ea
DW		 3994static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
3995{
3996 struct dl *dl;
3997
3998 for (dl = super->disks; dl; dl = dl->next)
3999 if (serialcmp(dl->serial, serial) == 0)
4000 break;
4001
4002 return dl;
4003}
4004
a2b97981
DW		 4005static struct imsm_disk *
4006__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4007{
4008 int i;
4009
4010 for (i = 0; i < mpb->num_disks; i++) {
4011 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4012
4013 if (serialcmp(disk->serial, serial) == 0) {
4014 if (idx)
4015 *idx = i;
4016 return disk;
4017 }
4018 }
4019
4020 return NULL;
4021}
4022
cdddbdbc
DW		 4023static int
4024load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4025{
a2b97981 4026 struct imsm_disk *disk;
cdddbdbc
DW		 4027 struct dl *dl;
4028 struct stat stb;
cdddbdbc 4029 int rv;
a2b97981 4030 char name[40];
d23fe947
DW		 4031 __u8 serial[MAX_RAID_SERIAL_LEN];
4032
4033 rv = imsm_read_serial(fd, devname, serial);
4034
4035 if (rv != 0)
4036 return 2;
4037
503975b9 4038 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4039
a2b97981
DW		 4040 fstat(fd, &stb);
4041 dl->major = major(stb.st_rdev);
4042 dl->minor = minor(stb.st_rdev);
4043 dl->next = super->disks;
4044 dl->fd = keep_fd ? fd : -1;
4045 assert(super->disks == NULL);
4046 super->disks = dl;
4047 serialcpy(dl->serial, serial);
4048 dl->index = -2;
4049 dl->e = NULL;
4050 fd2devname(fd, name);
4051 if (devname)
503975b9 4052 dl->devname = xstrdup(devname);
a2b97981 4053 else
503975b9 4054 dl->devname = xstrdup(name);
cdddbdbc 4055
d23fe947 4056 /* look up this disk's index in the current anchor */
a2b97981
DW		 4057 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4058 if (disk) {
4059 dl->disk = *disk;
4060 /* only set index on disks that are a member of a
4061 * populated contianer, i.e. one with raid_devs
4062 */
4063 if (is_failed(&dl->disk))
3f6efecc 4064 dl->index = -2;
2432ce9b 4065 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4066 dl->index = -1;
3f6efecc
DW		 4067 }
4068
949c47a0
DW		 4069 return 0;
4070}
4071
0c046afd
DW		 4072/* When migrating map0 contains the 'destination' state while map1
4073 * contains the current state. When not migrating map0 contains the
4074 * current state. This routine assumes that map[0].map_state is set to
4075 * the current array state before being called.
4076 *
4077 * Migration is indicated by one of the following states
4078 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4079 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4080 * map1state=unitialized)
1484e727 4081 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4082 * map1state=normal)
e3bba0e0 4083 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4084 * map1state=degraded)
8e59f3d8
AK		 4085 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4086 * map1state=normal)
0c046afd 4087 */
8e59f3d8
AK		 4088static void migrate(struct imsm_dev *dev, struct intel_super *super,
4089 __u8 to_state, int migr_type)
3393c6af 4090{
0c046afd 4091 struct imsm_map *dest;
238c0a71 4092 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4093
0c046afd 4094 dev->vol.migr_state = 1;
1484e727 4095 set_migr_type(dev, migr_type);
f8f603f1 4096 dev->vol.curr_migr_unit = 0;
238c0a71 4097 dest = get_imsm_map(dev, MAP_1);
0c046afd 4098
0556e1a2 4099 /* duplicate and then set the target end state in map[0] */
3393c6af 4100 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4101 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4102 __u32 ord;
4103 int i;
4104
4105 for (i = 0; i < src->num_members; i++) {
4106 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4107 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4108 }
4109 }
4110
8e59f3d8
AK		 4111 if (migr_type == MIGR_GEN_MIGR)
4112 /* Clear migration record */
4113 memset(super->migr_rec, 0, sizeof(struct migr_record));
4114
0c046afd 4115 src->map_state = to_state;
949c47a0 4116}
f8f603f1 4117
809da78e
AK		 4118static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4119 __u8 map_state)
f8f603f1 4120{
238c0a71
AK		 4121 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4122 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4123 MAP_0 : MAP_1);
28bce06f 4124 int i, j;
0556e1a2
DW		 4125
4126 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4127 * completed in the last migration.
4128 *
28bce06f 4129 * FIXME add support for raid-level-migration
0556e1a2 4130 */
089f9d79
JS		 4131 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4132 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4133 /* when final map state is other than expected
4134 * merge maps (not for migration)
4135 */
4136 int failed;
4137
4138 for (i = 0; i < prev->num_members; i++)
4139 for (j = 0; j < map->num_members; j++)
4140 /* during online capacity expansion
4141 * disks position can be changed
4142 * if takeover is used
4143 */
4144 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4145 ord_to_idx(prev->disk_ord_tbl[i])) {
4146 map->disk_ord_tbl[j] |=
4147 prev->disk_ord_tbl[i];
4148 break;
4149 }
4150 failed = imsm_count_failed(super, dev, MAP_0);
4151 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4152 }
f8f603f1
DW		 4153
4154 dev->vol.migr_state = 0;
ea672ee1 4155 set_migr_type(dev, 0);
f8f603f1
DW		 4156 dev->vol.curr_migr_unit = 0;
4157 map->map_state = map_state;
4158}
949c47a0
DW		 4159
4160static int parse_raid_devices(struct intel_super *super)
4161{
4162 int i;
4163 struct imsm_dev *dev_new;
4d7b1503 4164 size_t len, len_migr;
401d313b 4165 size_t max_len = 0;
4d7b1503
DW		 4166 size_t space_needed = 0;
4167 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4168
4169 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4170 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4171 struct intel_dev *dv;
949c47a0 4172
4d7b1503
DW		 4173 len = sizeof_imsm_dev(dev_iter, 0);
4174 len_migr = sizeof_imsm_dev(dev_iter, 1);
4175 if (len_migr > len)
4176 space_needed += len_migr - len;
ca9de185 4177
503975b9 4178 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4179 if (max_len < len_migr)
4180 max_len = len_migr;
4181 if (max_len > len_migr)
4182 space_needed += max_len - len_migr;
503975b9 4183 dev_new = xmalloc(max_len);
949c47a0 4184 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4185 dv->dev = dev_new;
4186 dv->index = i;
4187 dv->next = super->devlist;
4188 super->devlist = dv;
949c47a0 4189 }
cdddbdbc 4190
4d7b1503
DW		 4191 /* ensure that super->buf is large enough when all raid devices
4192 * are migrating
4193 */
4194 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4195 void *buf;
4196
f36a9ecd
PB		 4197 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4198 super->sector_size);
4199 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4200 return 1;
4201
1f45a8ad
DW		 4202 memcpy(buf, super->buf, super->len);
4203 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4204 free(super->buf);
4205 super->buf = buf;
4206 super->len = len;
4207 }
ca9de185 4208
bbab0940
TM		 4209 super->extra_space += space_needed;
4210
cdddbdbc
DW		 4211 return 0;
4212}
4213
e2f41b2c
AK		 4214/*******************************************************************************
4215 * Function: check_mpb_migr_compatibility
4216 * Description: Function checks for unsupported migration features:
4217 * - migration optimization area (pba_of_lba0)
4218 * - descending reshape (ascending_migr)
4219 * Parameters:
4220 * super : imsm metadata information
4221 * Returns:
4222 * 0 : migration is compatible
4223 * -1 : migration is not compatible
4224 ******************************************************************************/
4225int check_mpb_migr_compatibility(struct intel_super *super)
4226{
4227 struct imsm_map *map0, *map1;
4228 struct migr_record *migr_rec = super->migr_rec;
4229 int i;
4230
4231 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4232 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4233
4234 if (dev_iter &&
4235 dev_iter->vol.migr_state == 1 &&
4236 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4237 /* This device is migrating */
238c0a71
AK		 4238 map0 = get_imsm_map(dev_iter, MAP_0);
4239 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4240 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4241 /* migration optimization area was used */
4242 return -1;
fc54fe7a
JS		 4243 if (migr_rec->ascending_migr == 0 &&
4244 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4245 /* descending reshape not supported yet */
4246 return -1;
4247 }
4248 }
4249 return 0;
4250}
4251
d23fe947 4252static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4253
cdddbdbc 4254/* load_imsm_mpb - read matrix metadata
f2f5c343 4255 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4256 */
4257static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4258{
4259 unsigned long long dsize;
cdddbdbc 4260 unsigned long long sectors;
f36a9ecd 4261 unsigned int sector_size = super->sector_size;
cdddbdbc 4262 struct stat;
6416d527 4263 struct imsm_super *anchor;
cdddbdbc
DW		 4264 __u32 check_sum;
4265
cdddbdbc 4266 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4267 if (dsize < 2*sector_size) {
64436f06 4268 if (devname)
e7b84f9d
N		 4269 pr_err("%s: device to small for imsm\n",
4270 devname);
64436f06
N		 4271 return 1;
4272 }
cdddbdbc 4273
f36a9ecd 4274 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4275 if (devname)
e7b84f9d
N		 4276 pr_err("Cannot seek to anchor block on %s: %s\n",
4277 devname, strerror(errno));
cdddbdbc
DW		 4278 return 1;
4279 }
4280
f36a9ecd 4281 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4282 if (devname)
7a862a02 4283 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4284 return 1;
4285 }
466070ad 4286 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4287 if (devname)
e7b84f9d
N		 4288 pr_err("Cannot read anchor block on %s: %s\n",
4289 devname, strerror(errno));
6416d527 4290 free(anchor);
cdddbdbc
DW		 4291 return 1;
4292 }
4293
6416d527 4294 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4295 if (devname)
e7b84f9d 4296 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4297 free(anchor);
cdddbdbc
DW		 4298 return 2;
4299 }
4300
d23fe947 4301 __free_imsm(super, 0);
f2f5c343
LM		 4302 /* reload capability and hba */
4303
4304 /* capability and hba must be updated with new super allocation */
d424212e 4305 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4306 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4307 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4308 if (devname)
e7b84f9d
N		 4309 pr_err("unable to allocate %zu byte mpb buffer\n",
4310 super->len);
6416d527 4311 free(anchor);
cdddbdbc
DW		 4312 return 2;
4313 }
f36a9ecd 4314 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4315
f36a9ecd 4316 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4317 free(anchor);
8e59f3d8 4318
85337573
AO		 4319 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4320 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4321 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4322 free(super->buf);
4323 return 2;
4324 }
51d83f5d 4325 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4326
949c47a0 4327 if (!sectors) {
ecf45690
DW		 4328 check_sum = __gen_imsm_checksum(super->anchor);
4329 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4330 if (devname)
e7b84f9d
N		 4331 pr_err("IMSM checksum %x != %x on %s\n",
4332 check_sum,
4333 __le32_to_cpu(super->anchor->check_sum),
4334 devname);
ecf45690
DW		 4335 return 2;
4336 }
4337
a2b97981 4338 return 0;
949c47a0 4339 }
cdddbdbc
DW		 4340
4341 /* read the extended mpb */
f36a9ecd 4342 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4343 if (devname)
e7b84f9d
N		 4344 pr_err("Cannot seek to extended mpb on %s: %s\n",
4345 devname, strerror(errno));
cdddbdbc
DW		 4346 return 1;
4347 }
4348
f36a9ecd
PB		 4349 if ((unsigned int)read(fd, super->buf + sector_size,
4350 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4351 if (devname)
e7b84f9d
N		 4352 pr_err("Cannot read extended mpb on %s: %s\n",
4353 devname, strerror(errno));
cdddbdbc
DW		 4354 return 2;
4355 }
4356
949c47a0
DW		 4357 check_sum = __gen_imsm_checksum(super->anchor);
4358 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4359 if (devname)
e7b84f9d
N		 4360 pr_err("IMSM checksum %x != %x on %s\n",
4361 check_sum, __le32_to_cpu(super->anchor->check_sum),
4362 devname);
db575f3b 4363 return 3;
cdddbdbc
DW		 4364 }
4365
a2b97981
DW		 4366 return 0;
4367}
4368
8e59f3d8
AK		 4369static int read_imsm_migr_rec(int fd, struct intel_super *super);
4370
97f81ee2
CA		 4371/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4372static void clear_hi(struct intel_super *super)
4373{
4374 struct imsm_super *mpb = super->anchor;
4375 int i, n;
4376 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4377 return;
4378 for (i = 0; i < mpb->num_disks; ++i) {
4379 struct imsm_disk *disk = &mpb->disk[i];
4380 disk->total_blocks_hi = 0;
4381 }
4382 for (i = 0; i < mpb->num_raid_devs; ++i) {
4383 struct imsm_dev *dev = get_imsm_dev(super, i);
4384 if (!dev)
4385 return;
4386 for (n = 0; n < 2; ++n) {
4387 struct imsm_map *map = get_imsm_map(dev, n);
4388 if (!map)
4389 continue;
4390 map->pba_of_lba0_hi = 0;
4391 map->blocks_per_member_hi = 0;
4392 map->num_data_stripes_hi = 0;
4393 }
4394 }
4395}
4396
a2b97981
DW		 4397static int
4398load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4399{
4400 int err;
4401
4402 err = load_imsm_mpb(fd, super, devname);
4403 if (err)
4404 return err;
f36a9ecd
PB		 4405 if (super->sector_size == 4096)
4406 convert_from_4k(super);
a2b97981
DW		 4407 err = load_imsm_disk(fd, super, devname, keep_fd);
4408 if (err)
4409 return err;
4410 err = parse_raid_devices(super);
8d67477f
TM		 4411 if (err)
4412 return err;
4413 err = load_bbm_log(super);
97f81ee2 4414 clear_hi(super);
a2b97981 4415 return err;
cdddbdbc
DW		 4416}
4417
ae6aad82
DW		 4418static void __free_imsm_disk(struct dl *d)
4419{
4420 if (d->fd >= 0)
4421 close(d->fd);
4422 if (d->devname)
4423 free(d->devname);
0dcecb2e
DW		 4424 if (d->e)
4425 free(d->e);
ae6aad82
DW		 4426 free(d);
4427
4428}
1a64be56 4429
cdddbdbc
DW		 4430static void free_imsm_disks(struct intel_super *super)
4431{
47ee5a45 4432 struct dl *d;
cdddbdbc 4433
47ee5a45
DW		 4434 while (super->disks) {
4435 d = super->disks;
cdddbdbc 4436 super->disks = d->next;
ae6aad82 4437 __free_imsm_disk(d);
cdddbdbc 4438 }
cb82edca
AK		 4439 while (super->disk_mgmt_list) {
4440 d = super->disk_mgmt_list;
4441 super->disk_mgmt_list = d->next;
4442 __free_imsm_disk(d);
4443 }
47ee5a45
DW		 4444 while (super->missing) {
4445 d = super->missing;
4446 super->missing = d->next;
4447 __free_imsm_disk(d);
4448 }
4449
cdddbdbc
DW		 4450}
4451
9ca2c81c 4452/* free all the pieces hanging off of a super pointer */
d23fe947 4453static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4454{
88654014
LM		 4455 struct intel_hba *elem, *next;
4456
9ca2c81c 4457 if (super->buf) {
949c47a0 4458 free(super->buf);
9ca2c81c
DW		 4459 super->buf = NULL;
4460 }
f2f5c343
LM		 4461 /* unlink capability description */
4462 super->orom = NULL;
8e59f3d8
AK		 4463 if (super->migr_rec_buf) {
4464 free(super->migr_rec_buf);
4465 super->migr_rec_buf = NULL;
4466 }
d23fe947
DW		 4467 if (free_disks)
4468 free_imsm_disks(super);
ba2de7ba 4469 free_devlist(super);
88654014
LM		 4470 elem = super->hba;
4471 while (elem) {
4472 if (elem->path)
4473 free((void *)elem->path);
4474 next = elem->next;
4475 free(elem);
4476 elem = next;
88c32bb1 4477 }
8d67477f
TM		 4478 if (super->bbm_log)
4479 free(super->bbm_log);
88654014 4480 super->hba = NULL;
cdddbdbc
DW		 4481}
4482
9ca2c81c
DW		 4483static void free_imsm(struct intel_super *super)
4484{
d23fe947 4485 __free_imsm(super, 1);
928f1424 4486 free(super->bb.entries);
9ca2c81c
DW		 4487 free(super);
4488}
cdddbdbc
DW		 4489
4490static void free_super_imsm(struct supertype *st)
4491{
4492 struct intel_super *super = st->sb;
4493
4494 if (!super)
4495 return;
4496
4497 free_imsm(super);
4498 st->sb = NULL;
4499}
4500
49133e57 4501static struct intel_super *alloc_super(void)
c2c087e6 4502{
503975b9 4503 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4504
503975b9
N		 4505 super->current_vol = -1;
4506 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4507
4508 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4509 sizeof(struct md_bb_entry));
4510 if (!super->bb.entries) {
4511 free(super);
4512 return NULL;
4513 }
4514
c2c087e6
DW		 4515 return super;
4516}
4517
f0f5a016
LM		 4518/*
4519 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4520 */
d424212e 4521static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4522{
4523 struct sys_dev *hba_name;
4524 int rv = 0;
4525
3a30e28e
MT		 4526 if (fd >= 0 && test_partition(fd)) {
4527 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4528 devname);
4529 return 1;
4530 }
089f9d79 4531 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4532 super->orom = NULL;
f0f5a016
LM		 4533 super->hba = NULL;
4534 return 0;
4535 }
4536 hba_name = find_disk_attached_hba(fd, NULL);
4537 if (!hba_name) {
d424212e 4538 if (devname)
e7b84f9d
N		 4539 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4540 devname);
f0f5a016
LM		 4541 return 1;
4542 }
4543 rv = attach_hba_to_super(super, hba_name);
4544 if (rv == 2) {
d424212e
N		 4545 if (devname) {
4546 struct intel_hba *hba = super->hba;
f0f5a016 4547
60f0f54d
PB		 4548 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4549 " but the container is assigned to Intel(R) %s %s (",
d424212e 4550 devname,
614902f6 4551 get_sys_dev_type(hba_name->type),
60f0f54d 4552 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4553 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4554 get_sys_dev_type(super->hba->type),
4555 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4556
f0f5a016
LM		 4557 while (hba) {
4558 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4559 if (hba->next)
4560 fprintf(stderr, ", ");
4561 hba = hba->next;
4562 }
6b781d33 4563 fprintf(stderr, ").\n"
cca67208 4564 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4565 }
f0f5a016
LM		 4566 return 2;
4567 }
6b781d33 4568 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4569 if (!super->orom)
4570 return 3;
614902f6 4571
f0f5a016
LM		 4572 return 0;
4573}
4574
47ee5a45
DW		 4575/* find_missing - helper routine for load_super_imsm_all that identifies
4576 * disks that have disappeared from the system. This routine relies on
4577 * the mpb being uptodate, which it is at load time.
4578 */
4579static int find_missing(struct intel_super *super)
4580{
4581 int i;
4582 struct imsm_super *mpb = super->anchor;
4583 struct dl *dl;
4584 struct imsm_disk *disk;
47ee5a45
DW		 4585
4586 for (i = 0; i < mpb->num_disks; i++) {
4587 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4588 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4589 if (dl)
4590 continue;
47ee5a45 4591
503975b9 4592 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4593 dl->major = 0;
4594 dl->minor = 0;
4595 dl->fd = -1;
503975b9 4596 dl->devname = xstrdup("missing");
47ee5a45
DW		 4597 dl->index = i;
4598 serialcpy(dl->serial, disk->serial);
4599 dl->disk = *disk;
689c9bf3 4600 dl->e = NULL;
47ee5a45
DW		 4601 dl->next = super->missing;
4602 super->missing = dl;
4603 }
4604
4605 return 0;
4606}
4607
a2b97981
DW		 4608static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4609{
4610 struct intel_disk *idisk = disk_list;
4611
4612 while (idisk) {
4613 if (serialcmp(idisk->disk.serial, serial) == 0)
4614 break;
4615 idisk = idisk->next;
4616 }
4617
4618 return idisk;
4619}
4620
4621static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4622 struct intel_super *super,
4623 struct intel_disk **disk_list)
4624{
4625 struct imsm_disk *d = &super->disks->disk;
4626 struct imsm_super *mpb = super->anchor;
4627 int i, j;
4628
4629 for (i = 0; i < tbl_size; i++) {
4630 struct imsm_super *tbl_mpb = table[i]->anchor;
4631 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4632
4633 if (tbl_mpb->family_num == mpb->family_num) {
4634 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4635 dprintf("mpb from %d:%d matches %d:%d\n",
4636 super->disks->major,
a2b97981
DW		 4637 super->disks->minor,
4638 table[i]->disks->major,
4639 table[i]->disks->minor);
4640 break;
4641 }
4642
4643 if (((is_configured(d) && !is_configured(tbl_d)) ||
4644 is_configured(d) == is_configured(tbl_d)) &&
4645 tbl_mpb->generation_num < mpb->generation_num) {
4646 /* current version of the mpb is a
4647 * better candidate than the one in
4648 * super_table, but copy over "cross
4649 * generational" status
4650 */
4651 struct intel_disk *idisk;
4652
1ade5cc1
N		 4653 dprintf("mpb from %d:%d replaces %d:%d\n",
4654 super->disks->major,
a2b97981
DW		 4655 super->disks->minor,
4656 table[i]->disks->major,
4657 table[i]->disks->minor);
4658
4659 idisk = disk_list_get(tbl_d->serial, *disk_list);
4660 if (idisk && is_failed(&idisk->disk))
4661 tbl_d->status |= FAILED_DISK;
4662 break;
4663 } else {
4664 struct intel_disk *idisk;
4665 struct imsm_disk *disk;
4666
4667 /* tbl_mpb is more up to date, but copy
4668 * over cross generational status before
4669 * returning
4670 */
4671 disk = __serial_to_disk(d->serial, mpb, NULL);
4672 if (disk && is_failed(disk))
4673 d->status |= FAILED_DISK;
4674
4675 idisk = disk_list_get(d->serial, *disk_list);
4676 if (idisk) {
4677 idisk->owner = i;
4678 if (disk && is_configured(disk))
4679 idisk->disk.status |= CONFIGURED_DISK;
4680 }
4681
1ade5cc1
N		 4682 dprintf("mpb from %d:%d prefer %d:%d\n",
4683 super->disks->major,
a2b97981
DW		 4684 super->disks->minor,
4685 table[i]->disks->major,
4686 table[i]->disks->minor);
4687
4688 return tbl_size;
4689 }
4690 }
4691 }
4692
4693 if (i >= tbl_size)
4694 table[tbl_size++] = super;
4695 else
4696 table[i] = super;
4697
4698 /* update/extend the merged list of imsm_disk records */
4699 for (j = 0; j < mpb->num_disks; j++) {
4700 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4701 struct intel_disk *idisk;
4702
4703 idisk = disk_list_get(disk->serial, *disk_list);
4704 if (idisk) {
4705 idisk->disk.status |= disk->status;
4706 if (is_configured(&idisk->disk) ||
4707 is_failed(&idisk->disk))
4708 idisk->disk.status &= ~(SPARE_DISK);
4709 } else {
503975b9 4710 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4711 idisk->owner = IMSM_UNKNOWN_OWNER;
4712 idisk->disk = *disk;
4713 idisk->next = *disk_list;
4714 *disk_list = idisk;
4715 }
4716
4717 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4718 idisk->owner = i;
4719 }
4720
4721 return tbl_size;
4722}
4723
4724static struct intel_super *
4725validate_members(struct intel_super *super, struct intel_disk *disk_list,
4726 const int owner)
4727{
4728 struct imsm_super *mpb = super->anchor;
4729 int ok_count = 0;
4730 int i;
4731
4732 for (i = 0; i < mpb->num_disks; i++) {
4733 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4734 struct intel_disk *idisk;
4735
4736 idisk = disk_list_get(disk->serial, disk_list);
4737 if (idisk) {
4738 if (idisk->owner == owner ||
4739 idisk->owner == IMSM_UNKNOWN_OWNER)
4740 ok_count++;
4741 else
1ade5cc1
N		 4742 dprintf("'%.16s' owner %d != %d\n",
4743 disk->serial, idisk->owner,
a2b97981
DW		 4744 owner);
4745 } else {
1ade5cc1
N		 4746 dprintf("unknown disk %x [%d]: %.16s\n",
4747 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4748 disk->serial);
4749 break;
4750 }
4751 }
4752
4753 if (ok_count == mpb->num_disks)
4754 return super;
4755 return NULL;
4756}
4757
4758static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4759{
4760 struct intel_super *s;
4761
4762 for (s = super_list; s; s = s->next) {
4763 if (family_num != s->anchor->family_num)
4764 continue;
e12b3daa 4765 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4766 __le32_to_cpu(family_num), s->disks->devname);
4767 }
4768}
4769
4770static struct intel_super *
4771imsm_thunderdome(struct intel_super **super_list, int len)
4772{
4773 struct intel_super *super_table[len];
4774 struct intel_disk *disk_list = NULL;
4775 struct intel_super *champion, *spare;
4776 struct intel_super *s, **del;
4777 int tbl_size = 0;
4778 int conflict;
4779 int i;
4780
4781 memset(super_table, 0, sizeof(super_table));
4782 for (s = *super_list; s; s = s->next)
4783 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4784
4785 for (i = 0; i < tbl_size; i++) {
4786 struct imsm_disk *d;
4787 struct intel_disk *idisk;
4788 struct imsm_super *mpb = super_table[i]->anchor;
4789
4790 s = super_table[i];
4791 d = &s->disks->disk;
4792
4793 /* 'd' must appear in merged disk list for its
4794 * configuration to be valid
4795 */
4796 idisk = disk_list_get(d->serial, disk_list);
4797 if (idisk && idisk->owner == i)
4798 s = validate_members(s, disk_list, i);
4799 else
4800 s = NULL;
4801
4802 if (!s)
1ade5cc1
N		 4803 dprintf("marking family: %#x from %d:%d offline\n",
4804 mpb->family_num,
a2b97981
DW		 4805 super_table[i]->disks->major,
4806 super_table[i]->disks->minor);
4807 super_table[i] = s;
4808 }
4809
4810 /* This is where the mdadm implementation differs from the Windows
4811 * driver which has no strict concept of a container. We can only
4812 * assemble one family from a container, so when returning a prodigal
4813 * array member to this system the code will not be able to disambiguate
4814 * the container contents that should be assembled ("foreign" versus
4815 * "local"). It requires user intervention to set the orig_family_num
4816 * to a new value to establish a new container. The Windows driver in
4817 * this situation fixes up the volume name in place and manages the
4818 * foreign array as an independent entity.
4819 */
4820 s = NULL;
4821 spare = NULL;
4822 conflict = 0;
4823 for (i = 0; i < tbl_size; i++) {
4824 struct intel_super *tbl_ent = super_table[i];
4825 int is_spare = 0;
4826
4827 if (!tbl_ent)
4828 continue;
4829
4830 if (tbl_ent->anchor->num_raid_devs == 0) {
4831 spare = tbl_ent;
4832 is_spare = 1;
4833 }
4834
4835 if (s && !is_spare) {
4836 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4837 conflict++;
4838 } else if (!s && !is_spare)
4839 s = tbl_ent;
4840 }
4841
4842 if (!s)
4843 s = spare;
4844 if (!s) {
4845 champion = NULL;
4846 goto out;
4847 }
4848 champion = s;
4849
4850 if (conflict)
7a862a02 4851 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4852 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4853
4854 /* collect all dl's onto 'champion', and update them to
4855 * champion's version of the status
4856 */
4857 for (s = *super_list; s; s = s->next) {
4858 struct imsm_super *mpb = champion->anchor;
4859 struct dl *dl = s->disks;
4860
4861 if (s == champion)
4862 continue;
4863
5d7b407a
CA		 4864 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4865
a2b97981
DW		 4866 for (i = 0; i < mpb->num_disks; i++) {
4867 struct imsm_disk *disk;
4868
4869 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4870 if (disk) {
4871 dl->disk = *disk;
4872 /* only set index on disks that are a member of
4873 * a populated contianer, i.e. one with
4874 * raid_devs
4875 */
4876 if (is_failed(&dl->disk))
4877 dl->index = -2;
4878 else if (is_spare(&dl->disk))
4879 dl->index = -1;
4880 break;
4881 }
4882 }
4883
4884 if (i >= mpb->num_disks) {
4885 struct intel_disk *idisk;
4886
4887 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4888 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4889 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4890 dl->index = -1;
4891 else {
4892 dl->index = -2;
4893 continue;
4894 }
4895 }
4896
4897 dl->next = champion->disks;
4898 champion->disks = dl;
4899 s->disks = NULL;
4900 }
4901
4902 /* delete 'champion' from super_list */
4903 for (del = super_list; *del; ) {
4904 if (*del == champion) {
4905 *del = (*del)->next;
4906 break;
4907 } else
4908 del = &(*del)->next;
4909 }
4910 champion->next = NULL;
4911
4912 out:
4913 while (disk_list) {
4914 struct intel_disk *idisk = disk_list;
4915
4916 disk_list = disk_list->next;
4917 free(idisk);
4918 }
4919
4920 return champion;
4921}
4922
9587c373
LM		 4923static int
4924get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4925static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4926 int major, int minor, int keep_fd);
ec50f7b6
LM		 4927static int
4928get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4929 int *max, int keep_fd);
4930
cdddbdbc 4931static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4932 char *devname, struct md_list *devlist,
4933 int keep_fd)
cdddbdbc 4934{
a2b97981
DW		 4935 struct intel_super *super_list = NULL;
4936 struct intel_super *super = NULL;
a2b97981 4937 int err = 0;
9587c373 4938 int i = 0;
dab4a513 4939
9587c373
LM		 4940 if (fd >= 0)
4941 /* 'fd' is an opened container */
4942 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4943 else
ec50f7b6
LM		 4944 /* get super block from devlist devices */
4945 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4946 if (err)
1602d52c 4947 goto error;
a2b97981
DW		 4948 /* all mpbs enter, maybe one leaves */
4949 super = imsm_thunderdome(&super_list, i);
4950 if (!super) {
4951 err = 1;
4952 goto error;
cdddbdbc
DW		 4953 }
4954
47ee5a45
DW		 4955 if (find_missing(super) != 0) {
4956 free_imsm(super);
a2b97981
DW		 4957 err = 2;
4958 goto error;
47ee5a45 4959 }
8e59f3d8
AK		 4960
4961 /* load migration record */
4962 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 4963 if (err == -1) {
4964 /* migration is in progress,
4965 * but migr_rec cannot be loaded,
4966 */
8e59f3d8
AK		 4967 err = 4;
4968 goto error;
4969 }
e2f41b2c
AK		 4970
4971 /* Check migration compatibility */
089f9d79 4972 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 4973 pr_err("Unsupported migration detected");
e2f41b2c
AK		 4974 if (devname)
4975 fprintf(stderr, " on %s\n", devname);
4976 else
4977 fprintf(stderr, " (IMSM).\n");
4978
4979 err = 5;
4980 goto error;
4981 }
4982
a2b97981
DW		 4983 err = 0;
4984
4985 error:
4986 while (super_list) {
4987 struct intel_super *s = super_list;
4988
4989 super_list = super_list->next;
4990 free_imsm(s);
4991 }
9587c373 4992
a2b97981
DW		 4993 if (err)
4994 return err;
f7e7067b 4995
cdddbdbc 4996 *sbp = super;
9587c373 4997 if (fd >= 0)
4dd2df09 4998 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 4999 else
4dd2df09 5000 st->container_devnm[0] = 0;
a2b97981 5001 if (err == 0 && st->ss == NULL) {
bf5a934a 5002 st->ss = &super_imsm;
cdddbdbc
DW		 5003 st->minor_version = 0;
5004 st->max_devs = IMSM_MAX_DEVICES;
5005 }
cdddbdbc
DW		 5006 return 0;
5007}
2b959fbf 5008
ec50f7b6
LM		 5009static int
5010get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5011 int *max, int keep_fd)
5012{
5013 struct md_list *tmpdev;
5014 int err = 0;
5015 int i = 0;
9587c373 5016
ec50f7b6
LM		 5017 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5018 if (tmpdev->used != 1)
5019 continue;
5020 if (tmpdev->container == 1) {
ca9de185 5021 int lmax = 0;
ec50f7b6
LM		 5022 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5023 if (fd < 0) {
e7b84f9d 5024 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5025 tmpdev->devname, strerror(errno));
5026 err = 8;
5027 goto error;
5028 }
5029 err = get_sra_super_block(fd, super_list,
5030 tmpdev->devname, &lmax,
5031 keep_fd);
5032 i += lmax;
5033 close(fd);
5034 if (err) {
5035 err = 7;
5036 goto error;
5037 }
5038 } else {
5039 int major = major(tmpdev->st_rdev);
5040 int minor = minor(tmpdev->st_rdev);
5041 err = get_super_block(super_list,
4dd2df09 5042 NULL,
ec50f7b6
LM		 5043 tmpdev->devname,
5044 major, minor,
5045 keep_fd);
5046 i++;
5047 if (err) {
5048 err = 6;
5049 goto error;
5050 }
5051 }
5052 }
5053 error:
5054 *max = i;
5055 return err;
5056}
9587c373 5057
4dd2df09 5058static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5059 int major, int minor, int keep_fd)
5060{
594dc1b8 5061 struct intel_super *s;
9587c373
LM		 5062 char nm[32];
5063 int dfd = -1;
9587c373
LM		 5064 int err = 0;
5065 int retry;
5066
5067 s = alloc_super();
5068 if (!s) {
5069 err = 1;
5070 goto error;
5071 }
5072
5073 sprintf(nm, "%d:%d", major, minor);
5074 dfd = dev_open(nm, O_RDWR);
5075 if (dfd < 0) {
5076 err = 2;
5077 goto error;
5078 }
5079
fa7bb6f8 5080 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5081 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5082 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5083
5084 /* retry the load if we might have raced against mdmon */
4dd2df09 5085 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5086 for (retry = 0; retry < 3; retry++) {
5087 usleep(3000);
5088 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5089 if (err != 3)
5090 break;
5091 }
5092 error:
5093 if (!err) {
5094 s->next = *super_list;
5095 *super_list = s;
5096 } else {
5097 if (s)
8d67477f 5098 free_imsm(s);
36614e95 5099 if (dfd >= 0)
9587c373
LM		 5100 close(dfd);
5101 }
089f9d79 5102 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5103 close(dfd);
5104 return err;
5105
5106}
5107
5108static int
5109get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5110{
5111 struct mdinfo *sra;
4dd2df09 5112 char *devnm;
9587c373
LM		 5113 struct mdinfo *sd;
5114 int err = 0;
5115 int i = 0;
4dd2df09 5116 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5117 if (!sra)
5118 return 1;
5119
5120 if (sra->array.major_version != -1 ||
5121 sra->array.minor_version != -2 ||
5122 strcmp(sra->text_version, "imsm") != 0) {
5123 err = 1;
5124 goto error;
5125 }
5126 /* load all mpbs */
4dd2df09 5127 devnm = fd2devnm(fd);
9587c373 5128 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5129 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5130 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5131 err = 7;
5132 goto error;
5133 }
5134 }
5135 error:
5136 sysfs_free(sra);
5137 *max = i;
5138 return err;
5139}
5140
2b959fbf
N		 5141static int load_container_imsm(struct supertype *st, int fd, char *devname)
5142{
ec50f7b6 5143 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5144}
cdddbdbc
DW		 5145
5146static int load_super_imsm(struct supertype *st, int fd, char *devname)
5147{
5148 struct intel_super *super;
5149 int rv;
8a3544f8 5150 int retry;
cdddbdbc 5151
357ac106 5152 if (test_partition(fd))
691c6ee1
N		 5153 /* IMSM not allowed on partitions */
5154 return 1;
5155
37424f13
DW		 5156 free_super_imsm(st);
5157
49133e57 5158 super = alloc_super();
fa7bb6f8 5159 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5160 if (!super)
5161 return 1;
ea2bc72b
LM		 5162 /* Load hba and capabilities if they exist.
5163 * But do not preclude loading metadata in case capabilities or hba are
5164 * non-compliant and ignore_hw_compat is set.
5165 */
d424212e 5166 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5167 /* no orom/efi or non-intel hba of the disk */
089f9d79 5168 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5169 if (devname)
e7b84f9d 5170 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5171 free_imsm(super);
5172 return 2;
5173 }
a2b97981 5174 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5175
8a3544f8
AP		 5176 /* retry the load if we might have raced against mdmon */
5177 if (rv == 3) {
f96b1302
AP		 5178 struct mdstat_ent *mdstat = NULL;
5179 char *name = fd2kname(fd);
5180
5181 if (name)
5182 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5183
5184 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5185 for (retry = 0; retry < 3; retry++) {
5186 usleep(3000);
5187 rv = load_and_parse_mpb(fd, super, devname, 0);
5188 if (rv != 3)
5189 break;
5190 }
5191 }
5192
5193 free_mdstat(mdstat);
5194 }
5195
cdddbdbc
DW		 5196 if (rv) {
5197 if (devname)
7a862a02 5198 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5199 free_imsm(super);
5200 return rv;
5201 }
5202
5203 st->sb = super;
5204 if (st->ss == NULL) {
5205 st->ss = &super_imsm;
5206 st->minor_version = 0;
5207 st->max_devs = IMSM_MAX_DEVICES;
5208 }
8e59f3d8
AK		 5209
5210 /* load migration record */
2e062e82
AK		 5211 if (load_imsm_migr_rec(super, NULL) == 0) {
5212 /* Check for unsupported migration features */
5213 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5214 pr_err("Unsupported migration detected");
2e062e82
AK		 5215 if (devname)
5216 fprintf(stderr, " on %s\n", devname);
5217 else
5218 fprintf(stderr, " (IMSM).\n");
5219 return 3;
5220 }
e2f41b2c
AK		 5221 }
5222
cdddbdbc
DW		 5223 return 0;
5224}
5225
ef6ffade
DW		 5226static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5227{
5228 if (info->level == 1)
5229 return 128;
5230 return info->chunk_size >> 9;
5231}
5232
5551b113
CA		 5233static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5234 unsigned long long size)
fcfd9599 5235{
4025c288 5236 if (info->level == 1)
5551b113 5237 return size * 2;
4025c288 5238 else
5551b113 5239 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5240}
5241
4d1313e9
DW		 5242static void imsm_update_version_info(struct intel_super *super)
5243{
5244 /* update the version and attributes */
5245 struct imsm_super *mpb = super->anchor;
5246 char *version;
5247 struct imsm_dev *dev;
5248 struct imsm_map *map;
5249 int i;
5250
5251 for (i = 0; i < mpb->num_raid_devs; i++) {
5252 dev = get_imsm_dev(super, i);
238c0a71 5253 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5254 if (__le32_to_cpu(dev->size_high) > 0)
5255 mpb->attributes |= MPB_ATTRIB_2TB;
5256
5257 /* FIXME detect when an array spans a port multiplier */
5258 #if 0
5259 mpb->attributes |= MPB_ATTRIB_PM;
5260 #endif
5261
5262 if (mpb->num_raid_devs > 1 ||
5263 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5264 version = MPB_VERSION_ATTRIBS;
5265 switch (get_imsm_raid_level(map)) {
5266 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5267 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5268 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5269 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5270 }
5271 } else {
5272 if (map->num_members >= 5)
5273 version = MPB_VERSION_5OR6_DISK_ARRAY;
5274 else if (dev->status == DEV_CLONE_N_GO)
5275 version = MPB_VERSION_CNG;
5276 else if (get_imsm_raid_level(map) == 5)
5277 version = MPB_VERSION_RAID5;
5278 else if (map->num_members >= 3)
5279 version = MPB_VERSION_3OR4_DISK_ARRAY;
5280 else if (get_imsm_raid_level(map) == 1)
5281 version = MPB_VERSION_RAID1;
5282 else
5283 version = MPB_VERSION_RAID0;
5284 }
5285 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5286 }
5287}
5288
aa534678
DW		 5289static int check_name(struct intel_super *super, char *name, int quiet)
5290{
5291 struct imsm_super *mpb = super->anchor;
5292 char *reason = NULL;
9bd99a90
RS		 5293 char *start = name;
5294 size_t len = strlen(name);
aa534678
DW		 5295 int i;
5296
9bd99a90
RS		 5297 if (len > 0) {
5298 while (isspace(start[len - 1]))
5299 start[--len] = 0;
5300 while (*start && isspace(*start))
5301 ++start, --len;
5302 memmove(name, start, len + 1);
5303 }
5304
5305 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5306 reason = "must be 16 characters or less";
9bd99a90
RS		 5307 else if (len == 0)
5308 reason = "must be a non-empty string";
aa534678
DW		 5309
5310 for (i = 0; i < mpb->num_raid_devs; i++) {
5311 struct imsm_dev *dev = get_imsm_dev(super, i);
5312
5313 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5314 reason = "already exists";
5315 break;
5316 }
5317 }
5318
5319 if (reason && !quiet)
e7b84f9d 5320 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5321
5322 return !reason;
5323}
5324
8b353278 5325static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5326 struct shape *s, char *name,
83cd1e97
N		 5327 char *homehost, int *uuid,
5328 long long data_offset)
cdddbdbc 5329{
c2c087e6
DW		 5330 /* We are creating a volume inside a pre-existing container.
5331 * so st->sb is already set.
5332 */
5333 struct intel_super *super = st->sb;
f36a9ecd 5334 unsigned int sector_size = super->sector_size;
949c47a0 5335 struct imsm_super *mpb = super->anchor;
ba2de7ba 5336 struct intel_dev *dv;
c2c087e6
DW		 5337 struct imsm_dev *dev;
5338 struct imsm_vol *vol;
5339 struct imsm_map *map;
5340 int idx = mpb->num_raid_devs;
5341 int i;
760365f9 5342 int namelen;
c2c087e6 5343 unsigned long long array_blocks;
2c092cad 5344 size_t size_old, size_new;
5551b113 5345 unsigned long long num_data_stripes;
b53bfba6
TM		 5346 unsigned int data_disks;
5347 unsigned long long size_per_member;
cdddbdbc 5348
88c32bb1 5349 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5350 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5351 return 0;
5352 }
5353
2c092cad
DW		 5354 /* ensure the mpb is large enough for the new data */
5355 size_old = __le32_to_cpu(mpb->mpb_size);
5356 size_new = disks_to_mpb_size(info->nr_disks);
5357 if (size_new > size_old) {
5358 void *mpb_new;
f36a9ecd 5359 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5360
f36a9ecd 5361 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5362 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5363 return 0;
5364 }
85337573
AO		 5365 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5366 MIGR_REC_BUF_SECTORS*
5367 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5368 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5369 free(super->buf);
5370 free(super);
ea944c8f 5371 free(mpb_new);
8e59f3d8
AK		 5372 return 0;
5373 }
2c092cad
DW		 5374 memcpy(mpb_new, mpb, size_old);
5375 free(mpb);
5376 mpb = mpb_new;
949c47a0 5377 super->anchor = mpb_new;
2c092cad
DW		 5378 mpb->mpb_size = __cpu_to_le32(size_new);
5379 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5380 super->len = size_round;
2c092cad 5381 }
bf5a934a 5382 super->current_vol = idx;
3960e579
DW		 5383
5384 /* handle 'failed_disks' by either:
5385 * a) create dummy disk entries in the table if this the first
5386 * volume in the array. We add them here as this is the only
5387 * opportunity to add them. add_to_super_imsm_volume()
5388 * handles the non-failed disks and continues incrementing
5389 * mpb->num_disks.
5390 * b) validate that 'failed_disks' matches the current number
5391 * of missing disks if the container is populated
d23fe947 5392 */
3960e579 5393 if (super->current_vol == 0) {
d23fe947 5394 mpb->num_disks = 0;
3960e579
DW		 5395 for (i = 0; i < info->failed_disks; i++) {
5396 struct imsm_disk *disk;
5397
5398 mpb->num_disks++;
5399 disk = __get_imsm_disk(mpb, i);
5400 disk->status = CONFIGURED_DISK | FAILED_DISK;
5401 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5402 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5403 "missing:%d", (__u8)i);
3960e579
DW		 5404 }
5405 find_missing(super);
5406 } else {
5407 int missing = 0;
5408 struct dl *d;
5409
5410 for (d = super->missing; d; d = d->next)
5411 missing++;
5412 if (info->failed_disks > missing) {
e7b84f9d 5413 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5414 return 0;
5415 }
5416 }
5a038140 5417
aa534678
DW		 5418 if (!check_name(super, name, 0))
5419 return 0;
503975b9
N		 5420 dv = xmalloc(sizeof(*dv));
5421 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5422 /*
5423 * Explicitly allow truncating to not confuse gcc's
5424 * -Werror=stringop-truncation
5425 */
5426 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5427 memcpy(dev->volume, name, namelen);
e03640bd 5428 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5429 info->layout, info->chunk_size,
b53bfba6
TM		 5430 s->size * BLOCKS_PER_KB);
5431 data_disks = get_data_disks(info->level, info->layout,
5432 info->raid_disks);
5433 array_blocks = round_size_to_mb(array_blocks, data_disks);
5434 size_per_member = array_blocks / data_disks;
979d38be 5435
fcc2c9da 5436 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5437 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5438 vol = &dev->vol;
5439 vol->migr_state = 0;
1484e727 5440 set_migr_type(dev, MIGR_INIT);
3960e579 5441 vol->dirty = !info->state;
f8f603f1 5442 vol->curr_migr_unit = 0;
238c0a71 5443 map = get_imsm_map(dev, MAP_0);
5551b113 5444 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5445 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5446 map->failed_disk_num = ~0;
bf4442ab 5447 if (info->level > 0)
fffaf1ff
N		 5448 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5449 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5450 else
5451 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5452 IMSM_T_STATE_NORMAL;
252d23c0 5453 map->ddf = 1;
ef6ffade
DW		 5454
5455 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5456 free(dev);
5457 free(dv);
7a862a02 5458 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5459 return 0;
5460 }
81062a36
DW		 5461
5462 map->raid_level = info->level;
4d1313e9 5463 if (info->level == 10) {
c2c087e6 5464 map->raid_level = 1;
4d1313e9 5465 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5466 } else if (info->level == 1)
5467 map->num_domains = info->raid_disks;
5468 else
ff596308 5469 map->num_domains = 1;
81062a36 5470
5551b113 5471 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5472 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5473 num_data_stripes /= map->num_domains;
5474 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5475
44490938
MD		 5476 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5477 set_blocks_per_member(map, info_to_blocks_per_member(info,
5478 size_per_member /
5479 BLOCKS_PER_KB));
5480
c2c087e6
DW		 5481 map->num_members = info->raid_disks;
5482 for (i = 0; i < map->num_members; i++) {
5483 /* initialized in add_to_super */
4eb26970 5484 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5485 }
949c47a0 5486 mpb->num_raid_devs++;
2a24dc1b
PB		 5487 mpb->num_raid_devs_created++;
5488 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5489
b7580566 5490 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5491 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5492 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5493 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5494 } else {
5495 free(dev);
5496 free(dv);
5497 pr_err("imsm does not support consistency policy %s\n",
5498 map_num(consistency_policies, s->consistency_policy));
5499 return 0;
5500 }
5501
ba2de7ba
DW		 5502 dv->dev = dev;
5503 dv->index = super->current_vol;
5504 dv->next = super->devlist;
5505 super->devlist = dv;
c2c087e6 5506
4d1313e9
DW		 5507 imsm_update_version_info(super);
5508
c2c087e6 5509 return 1;
cdddbdbc
DW		 5510}
5511
bf5a934a 5512static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5513 struct shape *s, char *name,
83cd1e97
N		 5514 char *homehost, int *uuid,
5515 unsigned long long data_offset)
bf5a934a
DW		 5516{
5517 /* This is primarily called by Create when creating a new array.
5518 * We will then get add_to_super called for each component, and then
5519 * write_init_super called to write it out to each device.
5520 * For IMSM, Create can create on fresh devices or on a pre-existing
5521 * array.
5522 * To create on a pre-existing array a different method will be called.
5523 * This one is just for fresh drives.
5524 */
5525 struct intel_super *super;
5526 struct imsm_super *mpb;
5527 size_t mpb_size;
4d1313e9 5528 char *version;
bf5a934a 5529
83cd1e97 5530 if (data_offset != INVALID_SECTORS) {
ed503f89 5531 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5532 return 0;
5533 }
5534
bf5a934a 5535 if (st->sb)
5308f117 5536 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5537 data_offset);
e683ca88
DW		 5538
5539 if (info)
5540 mpb_size = disks_to_mpb_size(info->nr_disks);
5541 else
f36a9ecd 5542 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5543
49133e57 5544 super = alloc_super();
f36a9ecd
PB		 5545 if (super &&
5546 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5547 free_imsm(super);
e683ca88
DW		 5548 super = NULL;
5549 }
5550 if (!super) {
1ade5cc1 5551 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5552 return 0;
5553 }
de44e46f
PB		 5554 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5555 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5556 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5557 free(super->buf);
8d67477f 5558 free_imsm(super);
8e59f3d8
AK		 5559 return 0;
5560 }
e683ca88 5561 memset(super->buf, 0, mpb_size);
ef649044 5562 mpb = super->buf;
e683ca88
DW		 5563 mpb->mpb_size = __cpu_to_le32(mpb_size);
5564 st->sb = super;
5565
5566 if (info == NULL) {
5567 /* zeroing superblock */
5568 return 0;
5569 }
bf5a934a 5570
4d1313e9
DW		 5571 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5572
5573 version = (char *) mpb->sig;
5574 strcpy(version, MPB_SIGNATURE);
5575 version += strlen(MPB_SIGNATURE);
5576 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5577
bf5a934a
DW		 5578 return 1;
5579}
5580
f2cc4f7d
AO		 5581static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5582{
5583 unsigned int member_sector_size;
5584
5585 if (dl->fd < 0) {
5586 pr_err("Invalid file descriptor for %s\n", dl->devname);
5587 return 0;
5588 }
5589
5590 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5591 return 0;
5592 if (member_sector_size != super->sector_size)
5593 return 0;
5594 return 1;
5595}
5596
f20c3968 5597static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5598 int fd, char *devname)
5599{
5600 struct intel_super *super = st->sb;
d23fe947 5601 struct imsm_super *mpb = super->anchor;
3960e579 5602 struct imsm_disk *_disk;
bf5a934a
DW		 5603 struct imsm_dev *dev;
5604 struct imsm_map *map;
3960e579 5605 struct dl *dl, *df;
4eb26970 5606 int slot;
bf5a934a 5607
949c47a0 5608 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5609 map = get_imsm_map(dev, MAP_0);
bf5a934a 5610
208933a7 5611 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5612 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5613 devname);
5614 return 1;
5615 }
5616
efb30e7f
DW		 5617 if (fd == -1) {
5618 /* we're doing autolayout so grab the pre-marked (in
5619 * validate_geometry) raid_disk
5620 */
5621 for (dl = super->disks; dl; dl = dl->next)
5622 if (dl->raiddisk == dk->raid_disk)
5623 break;
5624 } else {
5625 for (dl = super->disks; dl ; dl = dl->next)
5626 if (dl->major == dk->major &&
5627 dl->minor == dk->minor)
5628 break;
5629 }
d23fe947 5630
208933a7 5631 if (!dl) {
e7b84f9d 5632 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5633 return 1;
208933a7 5634 }
bf5a934a 5635
59632db9
MZ		 5636 if (mpb->num_disks == 0)
5637 if (!get_dev_sector_size(dl->fd, dl->devname,
5638 &super->sector_size))
5639 return 1;
5640
f2cc4f7d
AO		 5641 if (!drive_validate_sector_size(super, dl)) {
5642 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5643 return 1;
5644 }
5645
d23fe947
DW		 5646 /* add a pristine spare to the metadata */
5647 if (dl->index < 0) {
5648 dl->index = super->anchor->num_disks;
5649 super->anchor->num_disks++;
5650 }
4eb26970
DW		 5651 /* Check the device has not already been added */
5652 slot = get_imsm_disk_slot(map, dl->index);
5653 if (slot >= 0 &&
238c0a71 5654 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5655 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5656 devname);
5657 return 1;
5658 }
656b6b5a 5659 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5660 dl->disk.status = CONFIGURED_DISK;
d23fe947 5661
3960e579
DW		 5662 /* update size of 'missing' disks to be at least as large as the
5663 * largest acitve member (we only have dummy missing disks when
5664 * creating the first volume)
5665 */
5666 if (super->current_vol == 0) {
5667 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5668 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5669 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5670 _disk = __get_imsm_disk(mpb, df->index);
5671 *_disk = df->disk;
5672 }
5673 }
5674
5675 /* refresh unset/failed slots to point to valid 'missing' entries */
5676 for (df = super->missing; df; df = df->next)
5677 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5678 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5679
5680 if ((ord & IMSM_ORD_REBUILD) == 0)
5681 continue;
5682 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5683 if (is_gen_migration(dev)) {
238c0a71
AK		 5684 struct imsm_map *map2 = get_imsm_map(dev,
5685 MAP_1);
0a108d63 5686 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5687 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5688 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5689 slot2,
5690 MAP_1);
1ace8403
AK		 5691 if ((unsigned)df->index ==
5692 ord_to_idx(ord2))
5693 set_imsm_ord_tbl_ent(map2,
0a108d63 5694 slot2,
1ace8403
AK		 5695 df->index |
5696 IMSM_ORD_REBUILD);
5697 }
5698 }
3960e579
DW		 5699 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5700 break;
5701 }
5702
d23fe947
DW		 5703 /* if we are creating the first raid device update the family number */
5704 if (super->current_vol == 0) {
5705 __u32 sum;
5706 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5707
3960e579 5708 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5709 if (!_dev || !_disk) {
e7b84f9d 5710 pr_err("BUG mpb setup error\n");
791b666a
AW		 5711 return 1;
5712 }
d23fe947
DW		 5713 *_dev = *dev;
5714 *_disk = dl->disk;
148acb7b
DW		 5715 sum = random32();
5716 sum += __gen_imsm_checksum(mpb);
d23fe947 5717 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5718 mpb->orig_family_num = mpb->family_num;
d23fe947 5719 }
ca0748fa 5720 super->current_disk = dl;
f20c3968 5721 return 0;
bf5a934a
DW		 5722}
5723
a8619d23
AK		 5724/* mark_spare()
5725 * Function marks disk as spare and restores disk serial
5726 * in case it was previously marked as failed by takeover operation
5727 * reruns:
5728 * -1 : critical error
5729 * 0 : disk is marked as spare but serial is not set
5730 * 1 : success
5731 */
5732int mark_spare(struct dl *disk)
5733{
5734 __u8 serial[MAX_RAID_SERIAL_LEN];
5735 int ret_val = -1;
5736
5737 if (!disk)
5738 return ret_val;
5739
5740 ret_val = 0;
5741 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5742 /* Restore disk serial number, because takeover marks disk
5743 * as failed and adds to serial ':0' before it becomes
5744 * a spare disk.
5745 */
5746 serialcpy(disk->serial, serial);
5747 serialcpy(disk->disk.serial, serial);
5748 ret_val = 1;
5749 }
5750 disk->disk.status = SPARE_DISK;
5751 disk->index = -1;
5752
5753 return ret_val;
5754}
88654014 5755
f20c3968 5756static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5757 int fd, char *devname,
5758 unsigned long long data_offset)
cdddbdbc 5759{
c2c087e6 5760 struct intel_super *super = st->sb;
c2c087e6
DW		 5761 struct dl *dd;
5762 unsigned long long size;
fa7bb6f8 5763 unsigned int member_sector_size;
f2f27e63 5764 __u32 id;
c2c087e6
DW		 5765 int rv;
5766 struct stat stb;
5767
88654014
LM		 5768 /* If we are on an RAID enabled platform check that the disk is
5769 * attached to the raid controller.
5770 * We do not need to test disks attachment for container based additions,
5771 * they shall be already tested when container was created/assembled.
88c32bb1 5772 */
d424212e 5773 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5774 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5775 if (rv != 0) {
5776 dprintf("capability: %p fd: %d ret: %d\n",
5777 super->orom, fd, rv);
5778 return 1;
88c32bb1
DW		 5779 }
5780
f20c3968
DW		 5781 if (super->current_vol >= 0)
5782 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5783
c2c087e6 5784 fstat(fd, &stb);
503975b9 5785 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5786 dd->major = major(stb.st_rdev);
5787 dd->minor = minor(stb.st_rdev);
503975b9 5788 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5789 dd->fd = fd;
689c9bf3 5790 dd->e = NULL;
1a64be56 5791 dd->action = DISK_ADD;
c2c087e6 5792 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5793 if (rv) {
e7b84f9d 5794 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5795 if (dd->devname)
5796 free(dd->devname);
949c47a0 5797 free(dd);
0030e8d6 5798 abort();
c2c087e6 5799 }
20bee0f8
PB		 5800 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5801 (super->hba->type == SYS_DEV_VMD))) {
5802 int i;
5803 char *devpath = diskfd_to_devpath(fd);
5804 char controller_path[PATH_MAX];
5805
5806 if (!devpath) {
5807 pr_err("failed to get devpath, aborting\n");
5808 if (dd->devname)
5809 free(dd->devname);
5810 free(dd);
5811 return 1;
5812 }
5813
5814 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5815 free(devpath);
5816
5817 if (devpath_to_vendor(controller_path) == 0x8086) {
5818 /*
5819 * If Intel's NVMe drive has serial ended with
5820 * "-A","-B","-1" or "-2" it means that this is "x8"
5821 * device (double drive on single PCIe card).
5822 * User should be warned about potential data loss.
5823 */
5824 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5825 /* Skip empty character at the end */
5826 if (dd->serial[i] == 0)
5827 continue;
5828
5829 if (((dd->serial[i] == 'A') ||
5830 (dd->serial[i] == 'B') ||
5831 (dd->serial[i] == '1') ||
5832 (dd->serial[i] == '2')) &&
5833 (dd->serial[i-1] == '-'))
5834 pr_err("\tThe action you are about to take may put your data at risk.\n"
5835 "\tPlease note that x8 devices may consist of two separate x4 devices "
5836 "located on a single PCIe port.\n"
5837 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5838 break;
5839 }
32716c51
PB		 5840 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5841 !imsm_orom_has_tpv_support(super->orom)) {
5842 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5843 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5844 free(dd->devname);
5845 free(dd);
5846 return 1;
20bee0f8
PB		 5847 }
5848 }
c2c087e6 5849
c2c087e6 5850 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5851 get_dev_sector_size(fd, NULL, &member_sector_size);
5852
5853 if (super->sector_size == 0) {
5854 /* this a first device, so sector_size is not set yet */
5855 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5856 }
5857
71e5411e 5858 /* clear migr_rec when adding disk to container */
85337573
AO		 5859 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5860 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5861 SEEK_SET) >= 0) {
466070ad 5862 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5863 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5864 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5865 perror("Write migr_rec failed");
5866 }
5867
c2c087e6 5868 size /= 512;
1f24f035 5869 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5870 set_total_blocks(&dd->disk, size);
5871 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5872 struct imsm_super *mpb = super->anchor;
5873 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5874 }
a8619d23 5875 mark_spare(dd);
c2c087e6 5876 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5877 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5878 else
b9f594fe 5879 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5880
5881 if (st->update_tail) {
1a64be56
LM		 5882 dd->next = super->disk_mgmt_list;
5883 super->disk_mgmt_list = dd;
43dad3d6
DW		 5884 } else {
5885 dd->next = super->disks;
5886 super->disks = dd;
ceaf0ee1 5887 super->updates_pending++;
43dad3d6 5888 }
f20c3968
DW		 5889
5890 return 0;
cdddbdbc
DW		 5891}
5892
1a64be56
LM		 5893static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5894{
5895 struct intel_super *super = st->sb;
5896 struct dl *dd;
5897
5898 /* remove from super works only in mdmon - for communication
5899 * manager - monitor. Check if communication memory buffer
5900 * is prepared.
5901 */
5902 if (!st->update_tail) {
1ade5cc1 5903 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5904 return 1;
5905 }
503975b9 5906 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5907 dd->major = dk->major;
5908 dd->minor = dk->minor;
1a64be56 5909 dd->fd = -1;
a8619d23 5910 mark_spare(dd);
1a64be56
LM		 5911 dd->action = DISK_REMOVE;
5912
5913 dd->next = super->disk_mgmt_list;
5914 super->disk_mgmt_list = dd;
5915
1a64be56
LM		 5916 return 0;
5917}
5918
f796af5d
DW		 5919static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5920
5921static union {
f36a9ecd 5922 char buf[MAX_SECTOR_SIZE];
f796af5d 5923 struct imsm_super anchor;
f36a9ecd 5924} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5925
d23fe947
DW		 5926/* spare records have their own family number and do not have any defined raid
5927 * devices
5928 */
5929static int write_super_imsm_spares(struct intel_super *super, int doclose)
5930{
d23fe947 5931 struct imsm_super *mpb = super->anchor;
f796af5d 5932 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5933 __u32 sum;
5934 struct dl *d;
5935
68641cdb
JS		 5936 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5937 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5938 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5939 spare->num_disks = 1;
5940 spare->num_raid_devs = 0;
5941 spare->cache_size = mpb->cache_size;
5942 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5943
5944 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5945 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5946
5947 for (d = super->disks; d; d = d->next) {
8796fdc4 5948 if (d->index != -1)
d23fe947
DW		 5949 continue;
5950
f796af5d 5951 spare->disk[0] = d->disk;
027c374f
CA		 5952 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5953 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5954
f36a9ecd
PB		 5955 if (super->sector_size == 4096)
5956 convert_to_4k_imsm_disk(&spare->disk[0]);
5957
f796af5d
DW		 5958 sum = __gen_imsm_checksum(spare);
5959 spare->family_num = __cpu_to_le32(sum);
5960 spare->orig_family_num = 0;
5961 sum = __gen_imsm_checksum(spare);
5962 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5963
f796af5d 5964 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 5965 pr_err("failed for device %d:%d %s\n",
5966 d->major, d->minor, strerror(errno));
e74255d9 5967 return 1;
d23fe947
DW		 5968 }
5969 if (doclose) {
5970 close(d->fd);
5971 d->fd = -1;
5972 }
5973 }
5974
e74255d9 5975 return 0;
d23fe947
DW		 5976}
5977
36988a3d 5978static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5979{
36988a3d 5980 struct intel_super *super = st->sb;
f36a9ecd 5981 unsigned int sector_size = super->sector_size;
949c47a0 5982 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 5983 struct dl *d;
5984 __u32 generation;
5985 __u32 sum;
d23fe947 5986 int spares = 0;
949c47a0 5987 int i;
a48ac0a8 5988 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5989 int num_disks = 0;
146c6260 5990 int clear_migration_record = 1;
bbab0940 5991 __u32 bbm_log_size;
cdddbdbc 5992
c2c087e6
DW		 5993 /* 'generation' is incremented everytime the metadata is written */
5994 generation = __le32_to_cpu(mpb->generation_num);
5995 generation++;
5996 mpb->generation_num = __cpu_to_le32(generation);
5997
148acb7b
DW		 5998 /* fix up cases where previous mdadm releases failed to set
5999 * orig_family_num
6000 */
6001 if (mpb->orig_family_num == 0)
6002 mpb->orig_family_num = mpb->family_num;
6003
d23fe947 6004 for (d = super->disks; d; d = d->next) {
8796fdc4 6005 if (d->index == -1)
d23fe947 6006 spares++;
36988a3d 6007 else {
d23fe947 6008 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6009 num_disks++;
6010 }
d23fe947 6011 }
36988a3d 6012 for (d = super->missing; d; d = d->next) {
47ee5a45 6013 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6014 num_disks++;
6015 }
6016 mpb->num_disks = num_disks;
6017 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6018
949c47a0
DW		 6019 for (i = 0; i < mpb->num_raid_devs; i++) {
6020 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6021 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6022 if (dev && dev2) {
6023 imsm_copy_dev(dev, dev2);
6024 mpb_size += sizeof_imsm_dev(dev, 0);
6025 }
146c6260
AK		 6026 if (is_gen_migration(dev2))
6027 clear_migration_record = 0;
949c47a0 6028 }
bbab0940
TM		 6029
6030 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6031
6032 if (bbm_log_size) {
6033 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6034 mpb->attributes |= MPB_ATTRIB_BBM;
6035 } else
6036 mpb->attributes &= ~MPB_ATTRIB_BBM;
6037
6038 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6039 mpb_size += bbm_log_size;
a48ac0a8 6040 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6041
bbab0940
TM		 6042#ifdef DEBUG
6043 assert(super->len == 0 || mpb_size <= super->len);
6044#endif
6045
c2c087e6 6046 /* recalculate checksum */
949c47a0 6047 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6048 mpb->check_sum = __cpu_to_le32(sum);
6049
51d83f5d
AK		 6050 if (super->clean_migration_record_by_mdmon) {
6051 clear_migration_record = 1;
6052 super->clean_migration_record_by_mdmon = 0;
6053 }
146c6260 6054 if (clear_migration_record)
de44e46f 6055 memset(super->migr_rec_buf, 0,
85337573 6056 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6057
f36a9ecd
PB		 6058 if (sector_size == 4096)
6059 convert_to_4k(super);
6060
d23fe947 6061 /* write the mpb for disks that compose raid devices */
c2c087e6 6062 for (d = super->disks; d ; d = d->next) {
86c54047 6063 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6064 continue;
30602f53 6065
146c6260
AK		 6066 if (clear_migration_record) {
6067 unsigned long long dsize;
6068
6069 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6070 if (lseek64(d->fd, dsize - sector_size,
6071 SEEK_SET) >= 0) {
466070ad
PB		 6072 if ((unsigned int)write(d->fd,
6073 super->migr_rec_buf,
de44e46f
PB		 6074 MIGR_REC_BUF_SECTORS*sector_size) !=
6075 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6076 perror("Write migr_rec failed");
146c6260
AK		 6077 }
6078 }
51d83f5d
AK		 6079
6080 if (store_imsm_mpb(d->fd, mpb))
6081 fprintf(stderr,
1ade5cc1
N		 6082 "failed for device %d:%d (fd: %d)%s\n",
6083 d->major, d->minor,
51d83f5d
AK		 6084 d->fd, strerror(errno));
6085
c2c087e6
DW		 6086 if (doclose) {
6087 close(d->fd);
6088 d->fd = -1;
6089 }
6090 }
6091
d23fe947
DW		 6092 if (spares)
6093 return write_super_imsm_spares(super, doclose);
6094
e74255d9 6095 return 0;
c2c087e6
DW		 6096}
6097
9b1fb677 6098static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6099{
6100 size_t len;
6101 struct imsm_update_create_array *u;
6102 struct intel_super *super = st->sb;
9b1fb677 6103 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6104 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6105 struct disk_info *inf;
6106 struct imsm_disk *disk;
6107 int i;
43dad3d6 6108
54c2c1ea
DW		 6109 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6110 sizeof(*inf) * map->num_members;
503975b9 6111 u = xmalloc(len);
43dad3d6 6112 u->type = update_create_array;
9b1fb677 6113 u->dev_idx = dev_idx;
43dad3d6 6114 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6115 inf = get_disk_info(u);
6116 for (i = 0; i < map->num_members; i++) {
238c0a71 6117 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6118
54c2c1ea 6119 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6120 if (!disk)
6121 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6122 serialcpy(inf[i].serial, disk->serial);
6123 }
43dad3d6
DW		 6124 append_metadata_update(st, u, len);
6125
6126 return 0;
6127}
6128
1a64be56 6129static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6130{
6131 struct intel_super *super = st->sb;
6132 size_t len;
1a64be56 6133 struct imsm_update_add_remove_disk *u;
43dad3d6 6134
1a64be56 6135 if (!super->disk_mgmt_list)
43dad3d6
DW		 6136 return 0;
6137
6138 len = sizeof(*u);
503975b9 6139 u = xmalloc(len);
1a64be56 6140 u->type = update_add_remove_disk;
43dad3d6
DW		 6141 append_metadata_update(st, u, len);
6142
6143 return 0;
6144}
2432ce9b
AP		 6145
6146__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6147
e397cefe
AP		 6148static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6149{
6150 struct ppl_header *ppl_hdr = buf;
6151 int ret;
6152
6153 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6154
6155 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6156 ret = -errno;
6157 perror("Failed to seek to PPL header location");
6158 return ret;
6159 }
6160
6161 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6162 ret = -errno;
6163 perror("Write PPL header failed");
6164 return ret;
6165 }
6166
6167 fsync(fd);
6168
6169 return 0;
6170}
6171
2432ce9b
AP		 6172static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6173{
6174 struct intel_super *super = st->sb;
6175 void *buf;
6176 struct ppl_header *ppl_hdr;
6177 int ret;
6178
b2514242
PB		 6179 /* first clear entire ppl space */
6180 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6181 if (ret)
6182 return ret;
6183
6184 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6185 if (ret) {
6186 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6187 return -ret;
2432ce9b
AP		 6188 }
6189
6190 memset(buf, 0, PPL_HEADER_SIZE);
6191 ppl_hdr = buf;
6192 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6193 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6194
6195 if (info->mismatch_cnt) {
6196 /*
6197 * We are overwriting an invalid ppl. Make one entry with wrong
6198 * checksum to prevent the kernel from skipping resync.
6199 */
6200 ppl_hdr->entries_count = __cpu_to_le32(1);
6201 ppl_hdr->entries[0].checksum = ~0;
6202 }
6203
e397cefe 6204 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6205
6206 free(buf);
6207 return ret;
6208}
6209
e397cefe
AP		 6210static int is_rebuilding(struct imsm_dev *dev);
6211
2432ce9b
AP		 6212static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6213 struct mdinfo *disk)
6214{
6215 struct intel_super *super = st->sb;
6216 struct dl *d;
e397cefe 6217 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6218 int ret = 0;
e397cefe 6219 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6220 __u32 crc;
6221 struct imsm_dev *dev;
2432ce9b 6222 __u32 idx;
44b6b876
PB		 6223 unsigned int i;
6224 unsigned long long ppl_offset = 0;
6225 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6226
6227 if (disk->disk.raid_disk < 0)
6228 return 0;
6229
2432ce9b 6230 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6231 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6232 d = get_imsm_dl_disk(super, idx);
6233
6234 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6235 return 0;
6236
6237 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6238 pr_err("Failed to allocate PPL header buffer\n");
6239 return -1;
6240 }
6241 buf = buf_orig;
2432ce9b 6242
44b6b876
PB		 6243 ret = 1;
6244 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6245 void *tmp;
6246
44b6b876 6247 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6248
44b6b876
PB		 6249 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6250 SEEK_SET) < 0) {
6251 perror("Failed to seek to PPL header location");
6252 ret = -1;
e397cefe 6253 break;
44b6b876 6254 }
2432ce9b 6255
44b6b876
PB		 6256 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6257 perror("Read PPL header failed");
6258 ret = -1;
e397cefe 6259 break;
44b6b876 6260 }
2432ce9b 6261
44b6b876 6262 ppl_hdr = buf;
2432ce9b 6263
44b6b876
PB		 6264 crc = __le32_to_cpu(ppl_hdr->checksum);
6265 ppl_hdr->checksum = 0;
6266
6267 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6268 dprintf("Wrong PPL header checksum on %s\n",
6269 d->devname);
e397cefe 6270 break;
44b6b876
PB		 6271 }
6272
6273 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6274 /* previous was newest, it was already checked */
e397cefe 6275 break;
44b6b876
PB		 6276 }
6277
6278 if ((__le32_to_cpu(ppl_hdr->signature) !=
6279 super->anchor->orig_family_num)) {
6280 dprintf("Wrong PPL header signature on %s\n",
6281 d->devname);
6282 ret = 1;
e397cefe 6283 break;
44b6b876
PB		 6284 }
6285
6286 ret = 0;
6287 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6288
44b6b876
PB		 6289 ppl_offset += PPL_HEADER_SIZE;
6290 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6291 ppl_offset +=
6292 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6293
6294 if (!buf_prev)
6295 buf_prev = buf + PPL_HEADER_SIZE;
6296 tmp = buf_prev;
6297 buf_prev = buf;
6298 buf = tmp;
2432ce9b
AP		 6299 }
6300
e397cefe
AP		 6301 if (buf_prev) {
6302 buf = buf_prev;
6303 ppl_hdr = buf_prev;
6304 }
2432ce9b 6305
54148aba
PB		 6306 /*
6307 * Update metadata to use mutliple PPLs area (1MB).
6308 * This is done once for all RAID members
6309 */
6310 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6311 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6312 char subarray[20];
6313 struct mdinfo *member_dev;
6314
6315 sprintf(subarray, "%d", info->container_member);
6316
6317 if (mdmon_running(st->container_devnm))
6318 st->update_tail = &st->updates;
6319
6320 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6321 pr_err("Failed to update subarray %s\n",
6322 subarray);
6323 } else {
6324 if (st->update_tail)
6325 flush_metadata_updates(st);
6326 else
6327 st->ss->sync_metadata(st);
6328 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6329 for (member_dev = info->devs; member_dev;
6330 member_dev = member_dev->next)
6331 member_dev->ppl_size =
6332 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6333 }
6334 }
6335
b23d0750 6336 if (ret == 1) {
2fc0fc63
AP		 6337 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6338
50b9c10d
PB		 6339 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6340 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6341 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6342 (is_rebuilding(dev) &&
2ec9d182
AP		 6343 dev->vol.curr_migr_unit == 0 &&
6344 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6345 ret = st->ss->write_init_ppl(st, info, d->fd);
6346 else
6347 info->mismatch_cnt++;
e397cefe
AP		 6348 } else if (ret == 0 &&
6349 ppl_hdr->entries_count == 0 &&
6350 is_rebuilding(dev) &&
6351 info->resync_start == 0) {
6352 /*
6353 * The header has no entries - add a single empty entry and
6354 * rewrite the header to prevent the kernel from going into
6355 * resync after an interrupted rebuild.
6356 */
6357 ppl_hdr->entries_count = __cpu_to_le32(1);
6358 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6359 }
2432ce9b 6360
e397cefe
AP		 6361 free(buf_orig);
6362
2432ce9b
AP		 6363 return ret;
6364}
6365
2432ce9b
AP		 6366static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6367{
6368 struct intel_super *super = st->sb;
6369 struct dl *d;
6370 int ret = 0;
6371
6372 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6373 info->array.level != 5)
6374 return 0;
6375
6376 for (d = super->disks; d ; d = d->next) {
6377 if (d->index < 0 || is_failed(&d->disk))
6378 continue;
6379
6380 ret = st->ss->write_init_ppl(st, info, d->fd);
6381 if (ret)
6382 break;
6383 }
6384
6385 return ret;
6386}
43dad3d6 6387
c2c087e6
DW		 6388static int write_init_super_imsm(struct supertype *st)
6389{
9b1fb677
DW		 6390 struct intel_super *super = st->sb;
6391 int current_vol = super->current_vol;
2432ce9b
AP		 6392 int rv = 0;
6393 struct mdinfo info;
6394
6395 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6396
6397 /* we are done with current_vol reset it to point st at the container */
6398 super->current_vol = -1;
6399
8273f55e 6400 if (st->update_tail) {
43dad3d6
DW		 6401 /* queue the recently created array / added disk
6402 * as a metadata update */
8273f55e 6403
43dad3d6 6404 /* determine if we are creating a volume or adding a disk */
9b1fb677 6405 if (current_vol < 0) {
1a64be56
LM		 6406 /* in the mgmt (add/remove) disk case we are running
6407 * in mdmon context, so don't close fd's
43dad3d6 6408 */
2432ce9b
AP		 6409 rv = mgmt_disk(st);
6410 } else {
6411 rv = write_init_ppl_imsm_all(st, &info);
6412 if (!rv)
6413 rv = create_array(st, current_vol);
6414 }
d682f344
N		 6415 } else {
6416 struct dl *d;
6417 for (d = super->disks; d; d = d->next)
ba728be7 6418 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6419 if (current_vol >= 0)
6420 rv = write_init_ppl_imsm_all(st, &info);
6421 if (!rv)
6422 rv = write_super_imsm(st, 1);
d682f344 6423 }
2432ce9b
AP		 6424
6425 return rv;
cdddbdbc
DW		 6426}
6427
e683ca88 6428static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6429{
e683ca88
DW		 6430 struct intel_super *super = st->sb;
6431 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6432
e683ca88 6433 if (!mpb)
ad97895e
DW		 6434 return 1;
6435
f36a9ecd
PB		 6436 if (super->sector_size == 4096)
6437 convert_to_4k(super);
e683ca88 6438 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6439}
6440
cdddbdbc
DW		 6441static int validate_geometry_imsm_container(struct supertype *st, int level,
6442 int layout, int raiddisks, int chunk,
af4348dd
N		 6443 unsigned long long size,
6444 unsigned long long data_offset,
6445 char *dev,
2c514b71
NB		 6446 unsigned long long *freesize,
6447 int verbose)
cdddbdbc 6448{
c2c087e6
DW		 6449 int fd;
6450 unsigned long long ldsize;
594dc1b8 6451 struct intel_super *super;
f2f5c343 6452 int rv = 0;
cdddbdbc 6453
c2c087e6
DW		 6454 if (level != LEVEL_CONTAINER)
6455 return 0;
6456 if (!dev)
6457 return 1;
6458
6459 fd = open(dev, O_RDONLY|O_EXCL, 0);
6460 if (fd < 0) {
ba728be7 6461 if (verbose > 0)
e7b84f9d 6462 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6463 dev, strerror(errno));
c2c087e6
DW		 6464 return 0;
6465 }
6466 if (!get_dev_size(fd, dev, &ldsize)) {
6467 close(fd);
6468 return 0;
6469 }
f2f5c343
LM		 6470
6471 /* capabilities retrieve could be possible
6472 * note that there is no fd for the disks in array.
6473 */
6474 super = alloc_super();
8d67477f
TM		 6475 if (!super) {
6476 close(fd);
6477 return 0;
6478 }
fa7bb6f8
PB		 6479 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6480 close(fd);
6481 free_imsm(super);
6482 return 0;
6483 }
6484
ba728be7 6485 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6486 if (rv != 0) {
6487#if DEBUG
6488 char str[256];
6489 fd2devname(fd, str);
1ade5cc1 6490 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6491 fd, str, super->orom, rv, raiddisks);
6492#endif
6493 /* no orom/efi or non-intel hba of the disk */
6494 close(fd);
6495 free_imsm(super);
6496 return 0;
6497 }
c2c087e6 6498 close(fd);
9126b9a8
CA		 6499 if (super->orom) {
6500 if (raiddisks > super->orom->tds) {
6501 if (verbose)
7a862a02 6502 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6503 raiddisks, super->orom->tds);
6504 free_imsm(super);
6505 return 0;
6506 }
6507 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6508 (ldsize >> 9) >> 32 > 0) {
6509 if (verbose)
e7b84f9d 6510 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6511 free_imsm(super);
6512 return 0;
6513 }
f2f5c343 6514 }
c2c087e6 6515
af4348dd 6516 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6517 free_imsm(super);
c2c087e6
DW		 6518
6519 return 1;
cdddbdbc
DW		 6520}
6521
0dcecb2e
DW		 6522static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6523{
6524 const unsigned long long base_start = e[*idx].start;
6525 unsigned long long end = base_start + e[*idx].size;
6526 int i;
6527
6528 if (base_start == end)
6529 return 0;
6530
6531 *idx = *idx + 1;
6532 for (i = *idx; i < num_extents; i++) {
6533 /* extend overlapping extents */
6534 if (e[i].start >= base_start &&
6535 e[i].start <= end) {
6536 if (e[i].size == 0)
6537 return 0;
6538 if (e[i].start + e[i].size > end)
6539 end = e[i].start + e[i].size;
6540 } else if (e[i].start > end) {
6541 *idx = i;
6542 break;
6543 }
6544 }
6545
6546 return end - base_start;
6547}
6548
6549static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6550{
6551 /* build a composite disk with all known extents and generate a new
6552 * 'maxsize' given the "all disks in an array must share a common start
6553 * offset" constraint
6554 */
503975b9 6555 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6556 struct dl *dl;
6557 int i, j;
6558 int start_extent;
6559 unsigned long long pos;
b9d77223 6560 unsigned long long start = 0;
0dcecb2e
DW		 6561 unsigned long long maxsize;
6562 unsigned long reserve;
6563
0dcecb2e
DW		 6564 /* coalesce and sort all extents. also, check to see if we need to
6565 * reserve space between member arrays
6566 */
6567 j = 0;
6568 for (dl = super->disks; dl; dl = dl->next) {
6569 if (!dl->e)
6570 continue;
6571 for (i = 0; i < dl->extent_cnt; i++)
6572 e[j++] = dl->e[i];
6573 }
6574 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6575
6576 /* merge extents */
6577 i = 0;
6578 j = 0;
6579 while (i < sum_extents) {
6580 e[j].start = e[i].start;
6581 e[j].size = find_size(e, &i, sum_extents);
6582 j++;
6583 if (e[j-1].size == 0)
6584 break;
6585 }
6586
6587 pos = 0;
6588 maxsize = 0;
6589 start_extent = 0;
6590 i = 0;
6591 do {
6592 unsigned long long esize;
6593
6594 esize = e[i].start - pos;
6595 if (esize >= maxsize) {
6596 maxsize = esize;
6597 start = pos;
6598 start_extent = i;
6599 }
6600 pos = e[i].start + e[i].size;
6601 i++;
6602 } while (e[i-1].size);
6603 free(e);
6604
a7dd165b
DW		 6605 if (maxsize == 0)
6606 return 0;
6607
6608 /* FIXME assumes volume at offset 0 is the first volume in a
6609 * container
6610 */
0dcecb2e
DW		 6611 if (start_extent > 0)
6612 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6613 else
6614 reserve = 0;
6615
6616 if (maxsize < reserve)
a7dd165b 6617 return 0;
0dcecb2e 6618
5551b113 6619 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6620 if (start + reserve > super->create_offset)
a7dd165b 6621 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6622 super->create_offset = start + reserve;
6623
6624 return maxsize - reserve;
6625}
6626
88c32bb1
DW		 6627static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6628{
6629 if (level < 0 || level == 6 || level == 4)
6630 return 0;
6631
6632 /* if we have an orom prevent invalid raid levels */
6633 if (orom)
6634 switch (level) {
6635 case 0: return imsm_orom_has_raid0(orom);
6636 case 1:
6637 if (raiddisks > 2)
6638 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6639 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6640 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6641 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6642 }
6643 else
6644 return 1; /* not on an Intel RAID platform so anything goes */
6645
6646 return 0;
6647}
6648
ca9de185
LM		 6649static int
6650active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6651 int dpa, int verbose)
6652{
6653 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6654 struct mdstat_ent *memb;
ca9de185
LM		 6655 int count = 0;
6656 int num = 0;
594dc1b8 6657 struct md_list *dv;
ca9de185
LM		 6658 int found;
6659
6660 for (memb = mdstat ; memb ; memb = memb->next) {
6661 if (memb->metadata_version &&
fc54fe7a 6662 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6663 (strcmp(&memb->metadata_version[9], name) == 0) &&
6664 !is_subarray(memb->metadata_version+9) &&
6665 memb->members) {
6666 struct dev_member *dev = memb->members;
6667 int fd = -1;
6668 while(dev && (fd < 0)) {
503975b9
N		 6669 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6670 num = sprintf(path, "%s%s", "/dev/", dev->name);
6671 if (num > 0)
6672 fd = open(path, O_RDONLY, 0);
089f9d79 6673 if (num <= 0 || fd < 0) {
676e87a8 6674 pr_vrb("Cannot open %s: %s\n",
503975b9 6675 dev->name, strerror(errno));
ca9de185 6676 }
503975b9 6677 free(path);
ca9de185
LM		 6678 dev = dev->next;
6679 }
6680 found = 0;
089f9d79 6681 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6682 struct mdstat_ent *vol;
6683 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6684 if (vol->active > 0 &&
ca9de185 6685 vol->metadata_version &&
9581efb1 6686 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6687 found++;
6688 count++;
6689 }
6690 }
6691 if (*devlist && (found < dpa)) {
503975b9 6692 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6693 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6694 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6695 dv->found = found;
6696 dv->used = 0;
6697 dv->next = *devlist;
6698 *devlist = dv;
ca9de185
LM		 6699 }
6700 }
6701 if (fd >= 0)
6702 close(fd);
6703 }
6704 }
6705 free_mdstat(mdstat);
6706 return count;
6707}
6708
6709#ifdef DEBUG_LOOP
6710static struct md_list*
6711get_loop_devices(void)
6712{
6713 int i;
6714 struct md_list *devlist = NULL;
594dc1b8 6715 struct md_list *dv;
ca9de185
LM		 6716
6717 for(i = 0; i < 12; i++) {
503975b9
N		 6718 dv = xcalloc(1, sizeof(*dv));
6719 dv->devname = xmalloc(40);
ca9de185
LM		 6720 sprintf(dv->devname, "/dev/loop%d", i);
6721 dv->next = devlist;
6722 devlist = dv;
6723 }
6724 return devlist;
6725}
6726#endif
6727
6728static struct md_list*
6729get_devices(const char *hba_path)
6730{
6731 struct md_list *devlist = NULL;
594dc1b8 6732 struct md_list *dv;
ca9de185
LM		 6733 struct dirent *ent;
6734 DIR *dir;
6735 int err = 0;
6736
6737#if DEBUG_LOOP
6738 devlist = get_loop_devices();
6739 return devlist;
6740#endif
6741 /* scroll through /sys/dev/block looking for devices attached to
6742 * this hba
6743 */
6744 dir = opendir("/sys/dev/block");
6745 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6746 int fd;
6747 char buf[1024];
6748 int major, minor;
6749 char *path = NULL;
6750 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6751 continue;
6752 path = devt_to_devpath(makedev(major, minor));
6753 if (!path)
6754 continue;
6755 if (!path_attached_to_hba(path, hba_path)) {
6756 free(path);
6757 path = NULL;
6758 continue;
6759 }
6760 free(path);
6761 path = NULL;
6762 fd = dev_open(ent->d_name, O_RDONLY);
6763 if (fd >= 0) {
6764 fd2devname(fd, buf);
6765 close(fd);
6766 } else {
e7b84f9d 6767 pr_err("cannot open device: %s\n",
ca9de185
LM		 6768 ent->d_name);
6769 continue;
6770 }
6771
503975b9
N		 6772 dv = xcalloc(1, sizeof(*dv));
6773 dv->devname = xstrdup(buf);
ca9de185
LM		 6774 dv->next = devlist;
6775 devlist = dv;
6776 }
6777 if (err) {
6778 while(devlist) {
6779 dv = devlist;
6780 devlist = devlist->next;
6781 free(dv->devname);
6782 free(dv);
6783 }
6784 }
562aa102 6785 closedir(dir);
ca9de185
LM		 6786 return devlist;
6787}
6788
6789static int
6790count_volumes_list(struct md_list *devlist, char *homehost,
6791 int verbose, int *found)
6792{
6793 struct md_list *tmpdev;
6794 int count = 0;
594dc1b8 6795 struct supertype *st;
ca9de185
LM		 6796
6797 /* first walk the list of devices to find a consistent set
6798 * that match the criterea, if that is possible.
6799 * We flag the ones we like with 'used'.
6800 */
6801 *found = 0;
6802 st = match_metadata_desc_imsm("imsm");
6803 if (st == NULL) {
676e87a8 6804 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6805 return 0;
6806 }
6807
6808 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6809 char *devname = tmpdev->devname;
0a6bff09 6810 dev_t rdev;
ca9de185
LM		 6811 struct supertype *tst;
6812 int dfd;
6813 if (tmpdev->used > 1)
6814 continue;
6815 tst = dup_super(st);
6816 if (tst == NULL) {
676e87a8 6817 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6818 goto err_1;
6819 }
6820 tmpdev->container = 0;
6821 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6822 if (dfd < 0) {
1ade5cc1 6823 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6824 devname, strerror(errno));
6825 tmpdev->used = 2;
0a6bff09 6826 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6827 tmpdev->used = 2;
6828 } else if (must_be_container(dfd)) {
6829 struct supertype *cst;
6830 cst = super_by_fd(dfd, NULL);
6831 if (cst == NULL) {
1ade5cc1 6832 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6833 devname);
6834 tmpdev->used = 2;
6835 } else if (tst->ss != st->ss) {
1ade5cc1 6836 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6837 devname);
6838 tmpdev->used = 2;
6839 } else if (!tst->ss->load_container ||
6840 tst->ss->load_container(tst, dfd, NULL))
6841 tmpdev->used = 2;
6842 else {
6843 tmpdev->container = 1;
6844 }
6845 if (cst)
6846 cst->ss->free_super(cst);
6847 } else {
0a6bff09 6848 tmpdev->st_rdev = rdev;
ca9de185 6849 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6850 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6851 devname);
6852 tmpdev->used = 2;
6853 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6854 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6855 tst->ss->name, devname);
6856 tmpdev->used = 2;
6857 }
6858 }
6859 if (dfd >= 0)
6860 close(dfd);
6861 if (tmpdev->used == 2 || tmpdev->used == 4) {
6862 /* Ignore unrecognised devices during auto-assembly */
6863 goto loop;
6864 }
6865 else {
6866 struct mdinfo info;
6867 tst->ss->getinfo_super(tst, &info, NULL);
6868
6869 if (st->minor_version == -1)
6870 st->minor_version = tst->minor_version;
6871
6872 if (memcmp(info.uuid, uuid_zero,
6873 sizeof(int[4])) == 0) {
6874 /* this is a floating spare. It cannot define
6875 * an array unless there are no more arrays of
6876 * this type to be found. It can be included
6877 * in an array of this type though.
6878 */
6879 tmpdev->used = 3;
6880 goto loop;
6881 }
6882
6883 if (st->ss != tst->ss ||
6884 st->minor_version != tst->minor_version ||
6885 st->ss->compare_super(st, tst) != 0) {
6886 /* Some mismatch. If exactly one array matches this host,
6887 * we can resolve on that one.
6888 * Or, if we are auto assembling, we just ignore the second
6889 * for now.
6890 */
1ade5cc1 6891 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6892 devname);
6893 goto loop;
6894 }
6895 tmpdev->used = 1;
6896 *found = 1;
6897 dprintf("found: devname: %s\n", devname);
6898 }
6899 loop:
6900 if (tst)
6901 tst->ss->free_super(tst);
6902 }
6903 if (*found != 0) {
6904 int err;
6905 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6906 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6907 for (iter = head; iter; iter = iter->next) {
6908 dprintf("content->text_version: %s vol\n",
6909 iter->text_version);
6910 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6911 /* do not assemble arrays with unsupported
6912 configurations */
1ade5cc1 6913 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6914 iter->text_version);
6915 } else
6916 count++;
6917 }
6918 sysfs_free(head);
6919
6920 } else {
1ade5cc1 6921 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6922 err, st->sb);
6923 }
6924 } else {
1ade5cc1 6925 dprintf("no more devices to examine\n");
ca9de185
LM		 6926 }
6927
6928 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6929 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6930 if (count) {
6931 if (count < tmpdev->found)
6932 count = 0;
6933 else
6934 count -= tmpdev->found;
6935 }
6936 }
6937 if (tmpdev->used == 1)
6938 tmpdev->used = 4;
6939 }
6940 err_1:
6941 if (st)
6942 st->ss->free_super(st);
6943 return count;
6944}
6945
d3c11416
AO		 6946static int __count_volumes(char *hba_path, int dpa, int verbose,
6947 int cmp_hba_path)
ca9de185 6948{
72a45777 6949 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6950 int count = 0;
72a45777
PB		 6951 const struct orom_entry *entry;
6952 struct devid_list *dv, *devid_list;
ca9de185 6953
d3c11416 6954 if (!hba_path)
ca9de185
LM		 6955 return 0;
6956
72a45777 6957 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 6958 if (strstr(idev->path, hba_path))
6959 break;
72a45777
PB		 6960 }
6961
6962 if (!idev || !idev->dev_id)
ca9de185 6963 return 0;
72a45777
PB		 6964
6965 entry = get_orom_entry_by_device_id(idev->dev_id);
6966
6967 if (!entry || !entry->devid_list)
6968 return 0;
6969
6970 devid_list = entry->devid_list;
6971 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 6972 struct md_list *devlist;
d3c11416
AO		 6973 struct sys_dev *device = NULL;
6974 char *hpath;
72a45777
PB		 6975 int found = 0;
6976
d3c11416
AO		 6977 if (cmp_hba_path)
6978 device = device_by_id_and_path(dv->devid, hba_path);
6979 else
6980 device = device_by_id(dv->devid);
6981
72a45777 6982 if (device)
d3c11416 6983 hpath = device->path;
72a45777
PB		 6984 else
6985 return 0;
6986
d3c11416 6987 devlist = get_devices(hpath);
72a45777
PB		 6988 /* if no intel devices return zero volumes */
6989 if (devlist == NULL)
6990 return 0;
6991
d3c11416
AO		 6992 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
6993 verbose);
6994 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 6995 if (devlist == NULL)
6996 return 0;
6997 do {
6998 found = 0;
6999 count += count_volumes_list(devlist,
7000 NULL,
7001 verbose,
7002 &found);
7003 dprintf("found %d count: %d\n", found, count);
7004 } while (found);
7005
d3c11416 7006 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7007
7008 while (devlist) {
7009 struct md_list *dv = devlist;
7010 devlist = devlist->next;
7011 free(dv->devname);
7012 free(dv);
7013 }
ca9de185
LM		 7014 }
7015 return count;
7016}
7017
d3c11416
AO		 7018static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7019{
7020 if (!hba)
7021 return 0;
7022 if (hba->type == SYS_DEV_VMD) {
7023 struct sys_dev *dev;
7024 int count = 0;
7025
7026 for (dev = find_intel_devices(); dev; dev = dev->next) {
7027 if (dev->type == SYS_DEV_VMD)
7028 count += __count_volumes(dev->path, dpa,
7029 verbose, 1);
7030 }
7031 return count;
7032 }
7033 return __count_volumes(hba->path, dpa, verbose, 0);
7034}
7035
cd9d1ac7
DW		 7036static int imsm_default_chunk(const struct imsm_orom *orom)
7037{
7038 /* up to 512 if the plaform supports it, otherwise the platform max.
7039 * 128 if no platform detected
7040 */
7041 int fs = max(7, orom ? fls(orom->sss) : 0);
7042
7043 return min(512, (1 << fs));
7044}
73408129 7045
6592ce37
DW		 7046static int
7047validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7048 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7049{
660260d0
DW		 7050 /* check/set platform and metadata limits/defaults */
7051 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7052 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7053 super->orom->dpa);
73408129
LM		 7054 return 0;
7055 }
7056
5d500228 7057 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7058 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7059 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7060 level, raiddisks, raiddisks > 1 ? "s" : "");
7061 return 0;
7062 }
cd9d1ac7 7063
7ccc4cc4 7064 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7065 *chunk = imsm_default_chunk(super->orom);
7066
7ccc4cc4 7067 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7068 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7069 return 0;
6592ce37 7070 }
cd9d1ac7 7071
6592ce37
DW		 7072 if (layout != imsm_level_to_layout(level)) {
7073 if (level == 5)
676e87a8 7074 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7075 else if (level == 10)
676e87a8 7076 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7077 else
676e87a8 7078 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7079 layout, level);
7080 return 0;
7081 }
2cc699af 7082
7ccc4cc4 7083 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7084 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7085 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7086 return 0;
7087 }
614902f6 7088
6592ce37
DW		 7089 return 1;
7090}
7091
1011e834 7092/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7093 * FIX ME add ahci details
7094 */
8b353278 7095static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7096 int layout, int raiddisks, int *chunk,
af4348dd
N		 7097 unsigned long long size,
7098 unsigned long long data_offset,
7099 char *dev,
2c514b71
NB		 7100 unsigned long long *freesize,
7101 int verbose)
cdddbdbc 7102{
9e04ac1c 7103 dev_t rdev;
c2c087e6 7104 struct intel_super *super = st->sb;
b2916f25 7105 struct imsm_super *mpb;
c2c087e6
DW		 7106 struct dl *dl;
7107 unsigned long long pos = 0;
7108 unsigned long long maxsize;
7109 struct extent *e;
7110 int i;
cdddbdbc 7111
88c32bb1
DW		 7112 /* We must have the container info already read in. */
7113 if (!super)
c2c087e6
DW		 7114 return 0;
7115
b2916f25
JS		 7116 mpb = super->anchor;
7117
2cc699af 7118 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7119 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7120 return 0;
d54559f0 7121 }
c2c087e6
DW		 7122 if (!dev) {
7123 /* General test: make sure there is space for
2da8544a
DW		 7124 * 'raiddisks' device extents of size 'size' at a given
7125 * offset
c2c087e6 7126 */
e46273eb 7127 unsigned long long minsize = size;
b7528a20 7128 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7129 int dcnt = 0;
7130 if (minsize == 0)
7131 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7132 for (dl = super->disks; dl ; dl = dl->next) {
7133 int found = 0;
7134
bf5a934a 7135 pos = 0;
c2c087e6
DW		 7136 i = 0;
7137 e = get_extents(super, dl);
7138 if (!e) continue;
7139 do {
7140 unsigned long long esize;
7141 esize = e[i].start - pos;
7142 if (esize >= minsize)
7143 found = 1;
b7528a20 7144 if (found && start_offset == MaxSector) {
2da8544a
DW		 7145 start_offset = pos;
7146 break;
7147 } else if (found && pos != start_offset) {
7148 found = 0;
7149 break;
7150 }
c2c087e6
DW		 7151 pos = e[i].start + e[i].size;
7152 i++;
7153 } while (e[i-1].size);
7154 if (found)
7155 dcnt++;
7156 free(e);
7157 }
7158 if (dcnt < raiddisks) {
2c514b71 7159 if (verbose)
7a862a02 7160 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7161 dcnt, raiddisks);
c2c087e6
DW		 7162 return 0;
7163 }
7164 return 1;
7165 }
0dcecb2e 7166
c2c087e6 7167 /* This device must be a member of the set */
9e04ac1c 7168 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7169 return 0;
7170 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7171 if (dl->major == (int)major(rdev) &&
7172 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7173 break;
7174 }
7175 if (!dl) {
2c514b71 7176 if (verbose)
7a862a02 7177 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7178 return 0;
a20d2ba5
DW		 7179 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7180 /* If a volume is present then the current creation attempt
7181 * cannot incorporate new spares because the orom may not
7182 * understand this configuration (all member disks must be
7183 * members of each array in the container).
7184 */
7a862a02
N		 7185 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7186 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7187 return 0;
5fe62b94
WD		 7188 } else if (super->orom && mpb->num_raid_devs > 0 &&
7189 mpb->num_disks != raiddisks) {
7a862a02 7190 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7191 return 0;
c2c087e6 7192 }
0dcecb2e
DW		 7193
7194 /* retrieve the largest free space block */
c2c087e6
DW		 7195 e = get_extents(super, dl);
7196 maxsize = 0;
7197 i = 0;
0dcecb2e
DW		 7198 if (e) {
7199 do {
7200 unsigned long long esize;
7201
7202 esize = e[i].start - pos;
7203 if (esize >= maxsize)
7204 maxsize = esize;
7205 pos = e[i].start + e[i].size;
7206 i++;
7207 } while (e[i-1].size);
7208 dl->e = e;
7209 dl->extent_cnt = i;
7210 } else {
7211 if (verbose)
e7b84f9d 7212 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7213 dev);
7214 return 0;
7215 }
7216 if (maxsize < size) {
7217 if (verbose)
e7b84f9d 7218 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7219 dev, maxsize, size);
7220 return 0;
7221 }
7222
7223 /* count total number of extents for merge */
7224 i = 0;
7225 for (dl = super->disks; dl; dl = dl->next)
7226 if (dl->e)
7227 i += dl->extent_cnt;
7228
7229 maxsize = merge_extents(super, i);
3baa56ab
LO		 7230
7231 if (!check_env("IMSM_NO_PLATFORM") &&
7232 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7233 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
3baa56ab
LO		 7234 return 0;
7235 }
7236
a7dd165b 7237 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7238 if (verbose) {
7239 if (maxsize == 0)
7a862a02 7240 pr_err("no free space left on device. Aborting...\n");
b3071342 7241 else
7a862a02 7242 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7243 maxsize, size);
7244 }
0dcecb2e 7245 return 0;
0dcecb2e
DW		 7246 }
7247
c2c087e6
DW		 7248 *freesize = maxsize;
7249
ca9de185 7250 if (super->orom) {
72a45777 7251 int count = count_volumes(super->hba,
ca9de185
LM		 7252 super->orom->dpa, verbose);
7253 if (super->orom->vphba <= count) {
676e87a8 7254 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7255 super->orom->vphba);
7256 return 0;
7257 }
7258 }
c2c087e6 7259 return 1;
cdddbdbc
DW		 7260}
7261
13bcac90 7262static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7263 unsigned long long size, int chunk,
7264 unsigned long long *freesize)
7265{
7266 struct intel_super *super = st->sb;
7267 struct imsm_super *mpb = super->anchor;
7268 struct dl *dl;
7269 int i;
7270 int extent_cnt;
7271 struct extent *e;
7272 unsigned long long maxsize;
7273 unsigned long long minsize;
7274 int cnt;
7275 int used;
7276
7277 /* find the largest common start free region of the possible disks */
7278 used = 0;
7279 extent_cnt = 0;
7280 cnt = 0;
7281 for (dl = super->disks; dl; dl = dl->next) {
7282 dl->raiddisk = -1;
7283
7284 if (dl->index >= 0)
7285 used++;
7286
7287 /* don't activate new spares if we are orom constrained
7288 * and there is already a volume active in the container
7289 */
7290 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7291 continue;
7292
7293 e = get_extents(super, dl);
7294 if (!e)
7295 continue;
7296 for (i = 1; e[i-1].size; i++)
7297 ;
7298 dl->e = e;
7299 dl->extent_cnt = i;
7300 extent_cnt += i;
7301 cnt++;
7302 }
7303
7304 maxsize = merge_extents(super, extent_cnt);
7305 minsize = size;
7306 if (size == 0)
612e59d8
CA		 7307 /* chunk is in K */
7308 minsize = chunk * 2;
efb30e7f
DW		 7309
7310 if (cnt < raiddisks ||
7311 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7312 maxsize < minsize ||
7313 maxsize == 0) {
e7b84f9d 7314 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7315 return 0; /* No enough free spaces large enough */
7316 }
7317
7318 if (size == 0) {
7319 size = maxsize;
7320 if (chunk) {
612e59d8
CA		 7321 size /= 2 * chunk;
7322 size *= 2 * chunk;
efb30e7f 7323 }
f878b242
LM		 7324 maxsize = size;
7325 }
7326 if (!check_env("IMSM_NO_PLATFORM") &&
7327 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7328 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
f878b242 7329 return 0;
efb30e7f 7330 }
efb30e7f
DW		 7331 cnt = 0;
7332 for (dl = super->disks; dl; dl = dl->next)
7333 if (dl->e)
7334 dl->raiddisk = cnt++;
7335
7336 *freesize = size;
7337
13bcac90
AK		 7338 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7339
efb30e7f
DW		 7340 return 1;
7341}
7342
13bcac90
AK		 7343static int reserve_space(struct supertype *st, int raiddisks,
7344 unsigned long long size, int chunk,
7345 unsigned long long *freesize)
7346{
7347 struct intel_super *super = st->sb;
7348 struct dl *dl;
7349 int cnt;
7350 int rv = 0;
7351
7352 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7353 if (rv) {
7354 cnt = 0;
7355 for (dl = super->disks; dl; dl = dl->next)
7356 if (dl->e)
7357 dl->raiddisk = cnt++;
7358 rv = 1;
7359 }
7360
7361 return rv;
7362}
7363
bf5a934a 7364static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7365 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7366 unsigned long long data_offset,
bf5a934a 7367 char *dev, unsigned long long *freesize,
5308f117 7368 int consistency_policy, int verbose)
bf5a934a
DW		 7369{
7370 int fd, cfd;
7371 struct mdinfo *sra;
20cbe8d2 7372 int is_member = 0;
bf5a934a 7373
d54559f0
LM		 7374 /* load capability
7375 * if given unused devices create a container
bf5a934a
DW		 7376 * if given given devices in a container create a member volume
7377 */
7378 if (level == LEVEL_CONTAINER) {
7379 /* Must be a fresh device to add to a container */
7380 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7381 raiddisks,
7ccc4cc4 7382 *chunk,
af4348dd 7383 size, data_offset,
bf5a934a
DW		 7384 dev, freesize,
7385 verbose);
7386 }
9587c373 7387
54865c30
RS		 7388 if (size && ((size < 1024) || (*chunk != UnSet &&
7389 size < (unsigned long long) *chunk))) {
7390 pr_err("Given size must be greater than 1M and chunk size.\n");
7391 /* Depends on algorithm in Create.c :
7392 * if container was given (dev == NULL) return -1,
7393 * if block device was given ( dev != NULL) return 0.
7394 */
7395 return dev ? -1 : 0;
7396 }
7397
8592f29d 7398 if (!dev) {
e91a3bad 7399 if (st->sb) {
ca9de185 7400 struct intel_super *super = st->sb;
e91a3bad 7401 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7402 raiddisks, chunk, size,
e91a3bad
LM		 7403 verbose))
7404 return 0;
efb30e7f
DW		 7405 /* we are being asked to automatically layout a
7406 * new volume based on the current contents of
7407 * the container. If the the parameters can be
7408 * satisfied reserve_space will record the disks,
7409 * start offset, and size of the volume to be
7410 * created. add_to_super and getinfo_super
7411 * detect when autolayout is in progress.
7412 */
ca9de185
LM		 7413 /* assuming that freesize is always given when array is
7414 created */
7415 if (super->orom && freesize) {
7416 int count;
72a45777 7417 count = count_volumes(super->hba,
ca9de185
LM		 7418 super->orom->dpa, verbose);
7419 if (super->orom->vphba <= count) {
676e87a8 7420 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7421 super->orom->vphba);
7422 return 0;
7423 }
7424 }
e91a3bad
LM		 7425 if (freesize)
7426 return reserve_space(st, raiddisks, size,
7ccc4cc4 7427 *chunk, freesize);
8592f29d
N		 7428 }
7429 return 1;
7430 }
bf5a934a
DW		 7431 if (st->sb) {
7432 /* creating in a given container */
7433 return validate_geometry_imsm_volume(st, level, layout,
7434 raiddisks, chunk, size,
af4348dd 7435 data_offset,
bf5a934a
DW		 7436 dev, freesize, verbose);
7437 }
7438
bf5a934a
DW		 7439 /* This device needs to be a device in an 'imsm' container */
7440 fd = open(dev, O_RDONLY|O_EXCL, 0);
7441 if (fd >= 0) {
7442 if (verbose)
e7b84f9d
N		 7443 pr_err("Cannot create this array on device %s\n",
7444 dev);
bf5a934a
DW		 7445 close(fd);
7446 return 0;
7447 }
7448 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7449 if (verbose)
e7b84f9d 7450 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7451 dev, strerror(errno));
7452 return 0;
7453 }
7454 /* Well, it is in use by someone, maybe an 'imsm' container. */
7455 cfd = open_container(fd);
20cbe8d2 7456 close(fd);
bf5a934a 7457 if (cfd < 0) {
bf5a934a 7458 if (verbose)
e7b84f9d 7459 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7460 dev);
7461 return 0;
7462 }
4dd2df09 7463 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7464 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7465 strcmp(sra->text_version, "imsm") == 0)
7466 is_member = 1;
7467 sysfs_free(sra);
7468 if (is_member) {
bf5a934a
DW		 7469 /* This is a member of a imsm container. Load the container
7470 * and try to create a volume
7471 */
7472 struct intel_super *super;
7473
ec50f7b6 7474 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7475 st->sb = super;
4dd2df09 7476 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7477 close(cfd);
7478 return validate_geometry_imsm_volume(st, level, layout,
7479 raiddisks, chunk,
af4348dd 7480 size, data_offset, dev,
ecbd9e81
N		 7481 freesize, 1)
7482 ? 1 : -1;
bf5a934a 7483 }
20cbe8d2 7484 }
bf5a934a 7485
20cbe8d2 7486 if (verbose)
e7b84f9d 7487 pr_err("failed container membership check\n");
20cbe8d2
AW		 7488
7489 close(cfd);
7490 return 0;
bf5a934a 7491}
0bd16cf2 7492
30f58b22 7493static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7494{
7495 struct intel_super *super = st->sb;
7496
30f58b22
DW		 7497 if (level && *level == UnSet)
7498 *level = LEVEL_CONTAINER;
7499
7500 if (level && layout && *layout == UnSet)
7501 *layout = imsm_level_to_layout(*level);
0bd16cf2 7502
cd9d1ac7
DW		 7503 if (chunk && (*chunk == UnSet || *chunk == 0))
7504 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7505}
7506
33414a01
DW		 7507static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7508
7509static int kill_subarray_imsm(struct supertype *st)
7510{
7511 /* remove the subarray currently referenced by ->current_vol */
7512 __u8 i;
7513 struct intel_dev **dp;
7514 struct intel_super *super = st->sb;
7515 __u8 current_vol = super->current_vol;
7516 struct imsm_super *mpb = super->anchor;
7517
7518 if (super->current_vol < 0)
7519 return 2;
7520 super->current_vol = -1; /* invalidate subarray cursor */
7521
7522 /* block deletions that would change the uuid of active subarrays
7523 *
7524 * FIXME when immutable ids are available, but note that we'll
7525 * also need to fixup the invalidated/active subarray indexes in
7526 * mdstat
7527 */
7528 for (i = 0; i < mpb->num_raid_devs; i++) {
7529 char subarray[4];
7530
7531 if (i < current_vol)
7532 continue;
7533 sprintf(subarray, "%u", i);
4dd2df09 7534 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7535 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7536 current_vol, i);
33414a01
DW		 7537
7538 return 2;
7539 }
7540 }
7541
7542 if (st->update_tail) {
503975b9 7543 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7544
33414a01
DW		 7545 u->type = update_kill_array;
7546 u->dev_idx = current_vol;
7547 append_metadata_update(st, u, sizeof(*u));
7548
7549 return 0;
7550 }
7551
7552 for (dp = &super->devlist; *dp;)
7553 if ((*dp)->index == current_vol) {
7554 *dp = (*dp)->next;
7555 } else {
7556 handle_missing(super, (*dp)->dev);
7557 if ((*dp)->index > current_vol)
7558 (*dp)->index--;
7559 dp = &(*dp)->next;
7560 }
7561
7562 /* no more raid devices, all active components are now spares,
7563 * but of course failed are still failed
7564 */
7565 if (--mpb->num_raid_devs == 0) {
7566 struct dl *d;
7567
7568 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7569 if (d->index > -2)
7570 mark_spare(d);
33414a01
DW		 7571 }
7572
7573 super->updates_pending++;
7574
7575 return 0;
7576}
aa534678 7577
a951a4f7 7578static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7579 char *update, struct mddev_ident *ident)
aa534678
DW		 7580{
7581 /* update the subarray currently referenced by ->current_vol */
7582 struct intel_super *super = st->sb;
7583 struct imsm_super *mpb = super->anchor;
7584
aa534678
DW		 7585 if (strcmp(update, "name") == 0) {
7586 char *name = ident->name;
a951a4f7
N		 7587 char *ep;
7588 int vol;
aa534678 7589
4dd2df09 7590 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7591 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7592 return 2;
7593 }
7594
7595 if (!check_name(super, name, 0))
7596 return 2;
7597
a951a4f7
N		 7598 vol = strtoul(subarray, &ep, 10);
7599 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7600 return 2;
7601
aa534678 7602 if (st->update_tail) {
503975b9 7603 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7604
aa534678 7605 u->type = update_rename_array;
a951a4f7 7606 u->dev_idx = vol;
618f4e6d
XN		 7607 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7608 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7609 append_metadata_update(st, u, sizeof(*u));
7610 } else {
7611 struct imsm_dev *dev;
ebad3af2 7612 int i, namelen;
aa534678 7613
a951a4f7 7614 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7615 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7616 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7617 memcpy(dev->volume, name, namelen);
aa534678
DW		 7618 for (i = 0; i < mpb->num_raid_devs; i++) {
7619 dev = get_imsm_dev(super, i);
7620 handle_missing(super, dev);
7621 }
7622 super->updates_pending++;
7623 }
e6e9dd3f
AP		 7624 } else if (strcmp(update, "ppl") == 0 ||
7625 strcmp(update, "no-ppl") == 0) {
7626 int new_policy;
7627 char *ep;
7628 int vol = strtoul(subarray, &ep, 10);
7629
7630 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7631 return 2;
7632
7633 if (strcmp(update, "ppl") == 0)
c2462068 7634 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7635 else
c2462068 7636 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7637
7638 if (st->update_tail) {
7639 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7640
7641 u->type = update_rwh_policy;
7642 u->dev_idx = vol;
7643 u->new_policy = new_policy;
7644 append_metadata_update(st, u, sizeof(*u));
7645 } else {
7646 struct imsm_dev *dev;
7647
7648 dev = get_imsm_dev(super, vol);
7649 dev->rwh_policy = new_policy;
7650 super->updates_pending++;
7651 }
aa534678
DW		 7652 } else
7653 return 2;
7654
7655 return 0;
7656}
bf5a934a 7657
28bce06f
AK		 7658static int is_gen_migration(struct imsm_dev *dev)
7659{
7534230b
AK		 7660 if (dev == NULL)
7661 return 0;
7662
28bce06f
AK		 7663 if (!dev->vol.migr_state)
7664 return 0;
7665
7666 if (migr_type(dev) == MIGR_GEN_MIGR)
7667 return 1;
7668
7669 return 0;
7670}
7671
1e5c6983
DW		 7672static int is_rebuilding(struct imsm_dev *dev)
7673{
7674 struct imsm_map *migr_map;
7675
7676 if (!dev->vol.migr_state)
7677 return 0;
7678
7679 if (migr_type(dev) != MIGR_REBUILD)
7680 return 0;
7681
238c0a71 7682 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7683
7684 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7685 return 1;
7686 else
7687 return 0;
7688}
7689
6ce1fbf1
AK		 7690static int is_initializing(struct imsm_dev *dev)
7691{
7692 struct imsm_map *migr_map;
7693
7694 if (!dev->vol.migr_state)
7695 return 0;
7696
7697 if (migr_type(dev) != MIGR_INIT)
7698 return 0;
7699
238c0a71 7700 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7701
7702 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7703 return 1;
7704
7705 return 0;
6ce1fbf1
AK		 7706}
7707
c47b0ff6
AK		 7708static void update_recovery_start(struct intel_super *super,
7709 struct imsm_dev *dev,
7710 struct mdinfo *array)
1e5c6983
DW		 7711{
7712 struct mdinfo *rebuild = NULL;
7713 struct mdinfo *d;
7714 __u32 units;
7715
7716 if (!is_rebuilding(dev))
7717 return;
7718
7719 /* Find the rebuild target, but punt on the dual rebuild case */
7720 for (d = array->devs; d; d = d->next)
7721 if (d->recovery_start == 0) {
7722 if (rebuild)
7723 return;
7724 rebuild = d;
7725 }
7726
4363fd80
DW		 7727 if (!rebuild) {
7728 /* (?) none of the disks are marked with
7729 * IMSM_ORD_REBUILD, so assume they are missing and the
7730 * disk_ord_tbl was not correctly updated
7731 */
1ade5cc1 7732 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7733 return;
7734 }
7735
1e5c6983 7736 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7737 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7738}
7739
276d77db 7740static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7741
00bbdbda 7742static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7743{
4f5bc454
DW		 7744 /* Given a container loaded by load_super_imsm_all,
7745 * extract information about all the arrays into
7746 * an mdinfo tree.
00bbdbda 7747 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7748 *
7749 * For each imsm_dev create an mdinfo, fill it in,
7750 * then look for matching devices in super->disks
7751 * and create appropriate device mdinfo.
7752 */
7753 struct intel_super *super = st->sb;
949c47a0 7754 struct imsm_super *mpb = super->anchor;
4f5bc454 7755 struct mdinfo *rest = NULL;
00bbdbda 7756 unsigned int i;
81219e70 7757 int sb_errors = 0;
abef11a3
AK		 7758 struct dl *d;
7759 int spare_disks = 0;
cdddbdbc 7760
19482bcc
AK		 7761 /* do not assemble arrays when not all attributes are supported */
7762 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7763 sb_errors = 1;
7a862a02 7764 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7765 }
7766
abef11a3
AK		 7767 /* count spare devices, not used in maps
7768 */
7769 for (d = super->disks; d; d = d->next)
7770 if (d->index == -1)
7771 spare_disks++;
7772
4f5bc454 7773 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7774 struct imsm_dev *dev;
7775 struct imsm_map *map;
86e3692b 7776 struct imsm_map *map2;
4f5bc454 7777 struct mdinfo *this;
a6482415 7778 int slot;
a6482415 7779 int chunk;
00bbdbda 7780 char *ep;
8b9cd157 7781 int level;
00bbdbda
N		 7782
7783 if (subarray &&
7784 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7785 continue;
7786
7787 dev = get_imsm_dev(super, i);
238c0a71
AK		 7788 map = get_imsm_map(dev, MAP_0);
7789 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7790 level = get_imsm_raid_level(map);
4f5bc454 7791
1ce0101c
DW		 7792 /* do not publish arrays that are in the middle of an
7793 * unsupported migration
7794 */
7795 if (dev->vol.migr_state &&
28bce06f 7796 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7797 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7798 dev->volume);
7799 continue;
7800 }
2db86302
LM		 7801 /* do not publish arrays that are not support by controller's
7802 * OROM/EFI
7803 */
1ce0101c 7804
503975b9 7805 this = xmalloc(sizeof(*this));
4f5bc454 7806
301406c9 7807 super->current_vol = i;
a5d85af7 7808 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7809 this->next = rest;
a6482415 7810 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7811 /* mdadm does not support all metadata features- set the bit in all arrays state */
7812 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7813 level, /* RAID level */
7814 imsm_level_to_layout(level),
81219e70 7815 map->num_members, /* raid disks */
fcc2c9da 7816 &chunk, imsm_dev_size(dev),
81219e70 7817 1 /* verbose */)) {
7a862a02 7818 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7819 dev->volume);
7820 this->array.state |=
7821 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7822 (1<<MD_SB_BLOCK_VOLUME);
7823 }
81219e70
LM		 7824
7825 /* if array has bad blocks, set suitable bit in all arrays state */
7826 if (sb_errors)
7827 this->array.state |=
7828 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7829 (1<<MD_SB_BLOCK_VOLUME);
7830
4f5bc454 7831 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7832 unsigned long long recovery_start;
4f5bc454
DW		 7833 struct mdinfo *info_d;
7834 struct dl *d;
7835 int idx;
9a1608e5 7836 int skip;
7eef0453 7837 __u32 ord;
8b9cd157 7838 int missing = 0;
4f5bc454 7839
9a1608e5 7840 skip = 0;
238c0a71
AK		 7841 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7842 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7843 for (d = super->disks; d ; d = d->next)
7844 if (d->index == idx)
0fbd635c 7845 break;
4f5bc454 7846
1e5c6983 7847 recovery_start = MaxSector;
4f5bc454 7848 if (d == NULL)
9a1608e5 7849 skip = 1;
25ed7e59 7850 if (d && is_failed(&d->disk))
9a1608e5 7851 skip = 1;
8b9cd157 7852 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7853 recovery_start = 0;
9a1608e5 7854
1011e834 7855 /*
9a1608e5 7856 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7857 * reset resync start to avoid a dirty-degraded
7858 * situation when performing the intial sync
9a1608e5 7859 */
8b9cd157
MK		 7860 if (skip)
7861 missing++;
7862
7863 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7864 if ((!able_to_resync(level, missing) ||
7865 recovery_start == 0))
7866 this->resync_start = MaxSector;
7867 } else {
7868 /*
7869 * FIXME handle dirty degraded
7870 */
7871 }
7872
9a1608e5
DW		 7873 if (skip)
7874 continue;
4f5bc454 7875
503975b9 7876 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7877 info_d->next = this->devs;
7878 this->devs = info_d;
7879
4f5bc454
DW		 7880 info_d->disk.number = d->index;
7881 info_d->disk.major = d->major;
7882 info_d->disk.minor = d->minor;
7883 info_d->disk.raid_disk = slot;
1e5c6983 7884 info_d->recovery_start = recovery_start;
86e3692b
AK		 7885 if (map2) {
7886 if (slot < map2->num_members)
7887 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7888 else
7889 this->array.spare_disks++;
86e3692b
AK		 7890 } else {
7891 if (slot < map->num_members)
7892 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7893 else
7894 this->array.spare_disks++;
86e3692b 7895 }
1e5c6983
DW		 7896 if (info_d->recovery_start == MaxSector)
7897 this->array.working_disks++;
4f5bc454
DW		 7898
7899 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7900 info_d->data_offset = pba_of_lba0(map);
44490938 7901 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7902
7903 if (map->raid_level == 5) {
2432ce9b
AP		 7904 info_d->ppl_sector = this->ppl_sector;
7905 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7906 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7907 recovery_start == 0)
7908 this->resync_start = 0;
06fb291a 7909 }
b12796be 7910
5e46202e 7911 info_d->bb.supported = 1;
b12796be
TM		 7912 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7913 info_d->data_offset,
7914 info_d->component_size,
7915 &info_d->bb);
4f5bc454 7916 }
1e5c6983 7917 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7918 update_recovery_start(super, dev, this);
abef11a3 7919 this->array.spare_disks += spare_disks;
276d77db
AK		 7920
7921 /* check for reshape */
7922 if (this->reshape_active == 1)
7923 recover_backup_imsm(st, this);
9a1608e5 7924 rest = this;
4f5bc454
DW		 7925 }
7926
7927 return rest;
cdddbdbc
DW		 7928}
7929
3b451610
AK		 7930static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7931 int failed, int look_in_map)
c2a1e7da 7932{
3b451610
AK		 7933 struct imsm_map *map;
7934
7935 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7936
7937 if (!failed)
1011e834 7938 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7939 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7940
7941 switch (get_imsm_raid_level(map)) {
7942 case 0:
7943 return IMSM_T_STATE_FAILED;
7944 break;
7945 case 1:
7946 if (failed < map->num_members)
7947 return IMSM_T_STATE_DEGRADED;
7948 else
7949 return IMSM_T_STATE_FAILED;
7950 break;
7951 case 10:
7952 {
7953 /**
c92a2527
DW		 7954 * check to see if any mirrors have failed, otherwise we
7955 * are degraded. Even numbered slots are mirrored on
7956 * slot+1
c2a1e7da 7957 */
c2a1e7da 7958 int i;
d9b420a5
N		 7959 /* gcc -Os complains that this is unused */
7960 int insync = insync;
c2a1e7da
DW		 7961
7962 for (i = 0; i < map->num_members; i++) {
238c0a71 7963 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 7964 int idx = ord_to_idx(ord);
7965 struct imsm_disk *disk;
c2a1e7da 7966
c92a2527 7967 /* reset the potential in-sync count on even-numbered
1011e834 7968 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 7969 */
7970 if ((i & 1) == 0)
7971 insync = 2;
c2a1e7da 7972
c92a2527 7973 disk = get_imsm_disk(super, idx);
25ed7e59 7974 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 7975 insync--;
c2a1e7da 7976
c92a2527
DW		 7977 /* no in-sync disks left in this mirror the
7978 * array has failed
7979 */
7980 if (insync == 0)
7981 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 7982 }
7983
7984 return IMSM_T_STATE_DEGRADED;
7985 }
7986 case 5:
7987 if (failed < 2)
7988 return IMSM_T_STATE_DEGRADED;
7989 else
7990 return IMSM_T_STATE_FAILED;
7991 break;
7992 default:
7993 break;
7994 }
7995
7996 return map->map_state;
7997}
7998
3b451610
AK		 7999static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8000 int look_in_map)
c2a1e7da
DW		 8001{
8002 int i;
8003 int failed = 0;
8004 struct imsm_disk *disk;
d5985138
AK		 8005 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8006 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8007 struct imsm_map *map_for_loop;
0556e1a2
DW		 8008 __u32 ord;
8009 int idx;
d5985138 8010 int idx_1;
c2a1e7da 8011
0556e1a2
DW		 8012 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8013 * disks that are being rebuilt. New failures are recorded to
8014 * map[0]. So we look through all the disks we started with and
8015 * see if any failures are still present, or if any new ones
8016 * have arrived
0556e1a2 8017 */
d5985138
AK		 8018 map_for_loop = map;
8019 if (prev && (map->num_members < prev->num_members))
8020 map_for_loop = prev;
68fe4598
LD		 8021
8022 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8023 idx_1 = -255;
238c0a71
AK		 8024 /* when MAP_X is passed both maps failures are counted
8025 */
d5985138 8026 if (prev &&
089f9d79
JS		 8027 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8028 i < prev->num_members) {
d5985138
AK		 8029 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8030 idx_1 = ord_to_idx(ord);
c2a1e7da 8031
d5985138
AK		 8032 disk = get_imsm_disk(super, idx_1);
8033 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8034 failed++;
8035 }
089f9d79
JS		 8036 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8037 i < map->num_members) {
d5985138
AK		 8038 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8039 idx = ord_to_idx(ord);
8040
8041 if (idx != idx_1) {
8042 disk = get_imsm_disk(super, idx);
8043 if (!disk || is_failed(disk) ||
8044 ord & IMSM_ORD_REBUILD)
8045 failed++;
8046 }
8047 }
c2a1e7da
DW		 8048 }
8049
8050 return failed;
845dea95
NB		 8051}
8052
97b4d0e9
DW		 8053static int imsm_open_new(struct supertype *c, struct active_array *a,
8054 char *inst)
8055{
8056 struct intel_super *super = c->sb;
8057 struct imsm_super *mpb = super->anchor;
bbab0940 8058 struct imsm_update_prealloc_bb_mem u;
9587c373 8059
97b4d0e9 8060 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8061 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8062 return -ENODEV;
8063 }
8064
8065 dprintf("imsm: open_new %s\n", inst);
8066 a->info.container_member = atoi(inst);
bbab0940
TM		 8067
8068 u.type = update_prealloc_badblocks_mem;
8069 imsm_update_metadata_locally(c, &u, sizeof(u));
8070
97b4d0e9
DW		 8071 return 0;
8072}
8073
0c046afd
DW		 8074static int is_resyncing(struct imsm_dev *dev)
8075{
8076 struct imsm_map *migr_map;
8077
8078 if (!dev->vol.migr_state)
8079 return 0;
8080
1484e727
DW		 8081 if (migr_type(dev) == MIGR_INIT ||
8082 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8083 return 1;
8084
4c9bc37b
AK		 8085 if (migr_type(dev) == MIGR_GEN_MIGR)
8086 return 0;
8087
238c0a71 8088 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8089
089f9d79
JS		 8090 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8091 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8092 return 1;
8093 else
8094 return 0;
8095}
8096
0556e1a2 8097/* return true if we recorded new information */
4c9e8c1e
TM		 8098static int mark_failure(struct intel_super *super,
8099 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8100{
0556e1a2
DW		 8101 __u32 ord;
8102 int slot;
8103 struct imsm_map *map;
86c54047
DW		 8104 char buf[MAX_RAID_SERIAL_LEN+3];
8105 unsigned int len, shift = 0;
0556e1a2
DW		 8106
8107 /* new failures are always set in map[0] */
238c0a71 8108 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8109
8110 slot = get_imsm_disk_slot(map, idx);
8111 if (slot < 0)
8112 return 0;
8113
8114 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8115 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8116 return 0;
8117
7d0c5e24
LD		 8118 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8119 buf[MAX_RAID_SERIAL_LEN] = '\000';
8120 strcat(buf, ":0");
86c54047
DW		 8121 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8122 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8123 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8124
f2f27e63 8125 disk->status |= FAILED_DISK;
0556e1a2 8126 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8127 /* mark failures in second map if second map exists and this disk
8128 * in this slot.
8129 * This is valid for migration, initialization and rebuild
8130 */
8131 if (dev->vol.migr_state) {
238c0a71 8132 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8133 int slot2 = get_imsm_disk_slot(map2, idx);
8134
089f9d79 8135 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8136 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8137 idx | IMSM_ORD_REBUILD);
8138 }
d7a1fda2
MT		 8139 if (map->failed_disk_num == 0xff ||
8140 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8141 map->failed_disk_num = slot;
4c9e8c1e
TM		 8142
8143 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8144
0556e1a2
DW		 8145 return 1;
8146}
8147
4c9e8c1e
TM		 8148static void mark_missing(struct intel_super *super,
8149 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8150{
4c9e8c1e 8151 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8152
8153 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8154 return;
8155
47ee5a45
DW		 8156 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8157 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8158}
8159
33414a01
DW		 8160static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8161{
33414a01 8162 struct dl *dl;
33414a01
DW		 8163
8164 if (!super->missing)
8165 return;
33414a01 8166
79b68f1b
PC		 8167 /* When orom adds replacement for missing disk it does
8168 * not remove entry of missing disk, but just updates map with
8169 * new added disk. So it is not enough just to test if there is
8170 * any missing disk, we have to look if there are any failed disks
8171 * in map to stop migration */
8172
33414a01 8173 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8174 /* end process for initialization and rebuild only
8175 */
8176 if (is_gen_migration(dev) == 0) {
fb12a745 8177 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8178
fb12a745
TM		 8179 if (failed) {
8180 __u8 map_state;
8181 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8182 struct imsm_map *map1;
8183 int i, ord, ord_map1;
8184 int rebuilt = 1;
3d59f0c0 8185
fb12a745
TM		 8186 for (i = 0; i < map->num_members; i++) {
8187 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8188 if (!(ord & IMSM_ORD_REBUILD))
8189 continue;
8190
8191 map1 = get_imsm_map(dev, MAP_1);
8192 if (!map1)
8193 continue;
8194
8195 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8196 if (ord_map1 & IMSM_ORD_REBUILD)
8197 rebuilt = 0;
8198 }
8199
8200 if (rebuilt) {
8201 map_state = imsm_check_degraded(super, dev,
8202 failed, MAP_0);
8203 end_migration(dev, super, map_state);
8204 }
8205 }
3d59f0c0 8206 }
33414a01 8207 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8208 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8209 super->updates_pending++;
8210}
8211
f3871fdc
AK		 8212static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8213 long long new_size)
70bdf0dc 8214{
70bdf0dc 8215 unsigned long long array_blocks;
9529d343
MD		 8216 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8217 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8218
8219 if (used_disks == 0) {
8220 /* when problems occures
8221 * return current array_blocks value
8222 */
fcc2c9da 8223 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8224
8225 return array_blocks;
8226 }
8227
8228 /* set array size in metadata
8229 */
9529d343 8230 if (new_size <= 0)
f3871fdc
AK		 8231 /* OLCE size change is caused by added disks
8232 */
44490938 8233 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8234 else
f3871fdc
AK		 8235 /* Online Volume Size Change
8236 * Using available free space
8237 */
8238 array_blocks = new_size;
70bdf0dc 8239
b53bfba6 8240 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8241 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8242
8243 return array_blocks;
8244}
8245
28bce06f
AK		 8246static void imsm_set_disk(struct active_array *a, int n, int state);
8247
0e2d1a4e
AK		 8248static void imsm_progress_container_reshape(struct intel_super *super)
8249{
8250 /* if no device has a migr_state, but some device has a
8251 * different number of members than the previous device, start
8252 * changing the number of devices in this device to match
8253 * previous.
8254 */
8255 struct imsm_super *mpb = super->anchor;
8256 int prev_disks = -1;
8257 int i;
1dfaa380 8258 int copy_map_size;
0e2d1a4e
AK		 8259
8260 for (i = 0; i < mpb->num_raid_devs; i++) {
8261 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8262 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8263 struct imsm_map *map2;
8264 int prev_num_members;
0e2d1a4e
AK		 8265
8266 if (dev->vol.migr_state)
8267 return;
8268
8269 if (prev_disks == -1)
8270 prev_disks = map->num_members;
8271 if (prev_disks == map->num_members)
8272 continue;
8273
8274 /* OK, this array needs to enter reshape mode.
8275 * i.e it needs a migr_state
8276 */
8277
1dfaa380 8278 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8279 prev_num_members = map->num_members;
8280 map->num_members = prev_disks;
8281 dev->vol.migr_state = 1;
8282 dev->vol.curr_migr_unit = 0;
ea672ee1 8283 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8284 for (i = prev_num_members;
8285 i < map->num_members; i++)
8286 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8287 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8288 /* Copy the current map */
1dfaa380 8289 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8290 map2->num_members = prev_num_members;
8291
f3871fdc 8292 imsm_set_array_size(dev, -1);
51d83f5d 8293 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8294 super->updates_pending++;
8295 }
8296}
8297
aad6f216 8298/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8299 * states are handled in imsm_set_disk() with one exception, when a
8300 * resync is stopped due to a new failure this routine will set the
8301 * 'degraded' state for the array.
8302 */
01f157d7 8303static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8304{
8305 int inst = a->info.container_member;
8306 struct intel_super *super = a->container->sb;
949c47a0 8307 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8308 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8309 int failed = imsm_count_failed(super, dev, MAP_0);
8310 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8311 __u32 blocks_per_unit;
a862209d 8312
1af97990
AK		 8313 if (dev->vol.migr_state &&
8314 dev->vol.migr_type == MIGR_GEN_MIGR) {
8315 /* array state change is blocked due to reshape action
aad6f216
N		 8316 * We might need to
8317 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8318 * - finish the reshape (if last_checkpoint is big and action != reshape)
8319 * - update curr_migr_unit
1af97990 8320 */
aad6f216
N		 8321 if (a->curr_action == reshape) {
8322 /* still reshaping, maybe update curr_migr_unit */
633b5610 8323 goto mark_checkpoint;
aad6f216
N		 8324 } else {
8325 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8326 /* for some reason we aborted the reshape.
b66e591b
AK		 8327 *
8328 * disable automatic metadata rollback
8329 * user action is required to recover process
aad6f216 8330 */
b66e591b 8331 if (0) {
238c0a71
AK		 8332 struct imsm_map *map2 =
8333 get_imsm_map(dev, MAP_1);
8334 dev->vol.migr_state = 0;
8335 set_migr_type(dev, 0);
8336 dev->vol.curr_migr_unit = 0;
8337 memcpy(map, map2,
8338 sizeof_imsm_map(map2));
8339 super->updates_pending++;
b66e591b 8340 }
aad6f216
N		 8341 }
8342 if (a->last_checkpoint >= a->info.component_size) {
8343 unsigned long long array_blocks;
8344 int used_disks;
e154ced3 8345 struct mdinfo *mdi;
aad6f216 8346
9529d343 8347 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8348 if (used_disks > 0) {
8349 array_blocks =
44490938 8350 per_dev_array_size(map) *
d55adef9 8351 used_disks;
b53bfba6
TM		 8352 array_blocks =
8353 round_size_to_mb(array_blocks,
8354 used_disks);
d55adef9
AK		 8355 a->info.custom_array_size = array_blocks;
8356 /* encourage manager to update array
8357 * size
8358 */
e154ced3 8359
d55adef9 8360 a->check_reshape = 1;
633b5610 8361 }
e154ced3
AK		 8362 /* finalize online capacity expansion/reshape */
8363 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8364 imsm_set_disk(a,
8365 mdi->disk.raid_disk,
8366 mdi->curr_state);
8367
0e2d1a4e 8368 imsm_progress_container_reshape(super);
e154ced3 8369 }
aad6f216 8370 }
1af97990
AK		 8371 }
8372
47ee5a45 8373 /* before we activate this array handle any missing disks */
33414a01
DW		 8374 if (consistent == 2)
8375 handle_missing(super, dev);
1e5c6983 8376
0c046afd 8377 if (consistent == 2 &&
b7941fd6 8378 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8379 map_state != IMSM_T_STATE_NORMAL ||
8380 dev->vol.migr_state))
01f157d7 8381 consistent = 0;
272906ef 8382
b7941fd6 8383 if (is_resync_complete(&a->info)) {
0c046afd 8384 /* complete intialization / resync,
0556e1a2
DW		 8385 * recovery and interrupted recovery is completed in
8386 * ->set_disk
0c046afd
DW		 8387 */
8388 if (is_resyncing(dev)) {
8389 dprintf("imsm: mark resync done\n");
809da78e 8390 end_migration(dev, super, map_state);
115c3803 8391 super->updates_pending++;
484240d8 8392 a->last_checkpoint = 0;
115c3803 8393 }
b9172665
AK		 8394 } else if ((!is_resyncing(dev) && !failed) &&
8395 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8396 /* mark the start of the init process if nothing is failed */
b7941fd6 8397 dprintf("imsm: mark resync start\n");
1484e727 8398 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8399 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8400 else
8e59f3d8 8401 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8402 super->updates_pending++;
115c3803 8403 }
a862209d 8404
633b5610 8405mark_checkpoint:
5b83bacf
AK		 8406 /* skip checkpointing for general migration,
8407 * it is controlled in mdadm
8408 */
8409 if (is_gen_migration(dev))
8410 goto skip_mark_checkpoint;
8411
1e5c6983 8412 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8413 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8414 if (blocks_per_unit) {
1e5c6983
DW		 8415 __u32 units32;
8416 __u64 units;
8417
4f0a7acc 8418 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8419 units32 = units;
8420
8421 /* check that we did not overflow 32-bits, and that
8422 * curr_migr_unit needs updating
8423 */
8424 if (units32 == units &&
bfd80a56 8425 units32 != 0 &&
1e5c6983
DW		 8426 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8427 dprintf("imsm: mark checkpoint (%u)\n", units32);
8428 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8429 super->updates_pending++;
8430 }
8431 }
f8f603f1 8432
5b83bacf 8433skip_mark_checkpoint:
3393c6af 8434 /* mark dirty / clean */
2432ce9b
AP		 8435 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8436 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8437 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8438 if (consistent) {
8439 dev->vol.dirty = RAIDVOL_CLEAN;
8440 } else {
8441 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8442 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8443 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8444 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8445 }
a862209d
DW		 8446 super->updates_pending++;
8447 }
28bce06f 8448
01f157d7 8449 return consistent;
a862209d
DW		 8450}
8451
6f50473f
TM		 8452static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8453{
8454 int inst = a->info.container_member;
8455 struct intel_super *super = a->container->sb;
8456 struct imsm_dev *dev = get_imsm_dev(super, inst);
8457 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8458
8459 if (slot > map->num_members) {
8460 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8461 slot, map->num_members - 1);
8462 return -1;
8463 }
8464
8465 if (slot < 0)
8466 return -1;
8467
8468 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8469}
8470
8d45d196 8471static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8472{
8d45d196
DW		 8473 int inst = a->info.container_member;
8474 struct intel_super *super = a->container->sb;
949c47a0 8475 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8476 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8477 struct imsm_disk *disk;
7ce05701
LD		 8478 struct mdinfo *mdi;
8479 int recovery_not_finished = 0;
0c046afd 8480 int failed;
6f50473f 8481 int ord;
0c046afd 8482 __u8 map_state;
fb12a745
TM		 8483 int rebuild_done = 0;
8484 int i;
8d45d196 8485
fb12a745 8486 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8487 if (ord < 0)
8d45d196
DW		 8488 return;
8489
4e6e574a 8490 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8491 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8492
5802a811 8493 /* check for new failures */
0556e1a2 8494 if (state & DS_FAULTY) {
4c9e8c1e 8495 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8496 super->updates_pending++;
8d45d196 8497 }
47ee5a45 8498
19859edc 8499 /* check if in_sync */
0556e1a2 8500 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8501 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8502
8503 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8504 rebuild_done = 1;
19859edc
DW		 8505 super->updates_pending++;
8506 }
8d45d196 8507
3b451610
AK		 8508 failed = imsm_count_failed(super, dev, MAP_0);
8509 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8510
0c046afd 8511 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8512 dprintf("imsm: Detected transition to state ");
8513 switch (map_state) {
8514 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8515 dprintf("normal: ");
8516 if (is_rebuilding(dev)) {
1ade5cc1 8517 dprintf_cont("while rebuilding");
7ce05701
LD		 8518 /* check if recovery is really finished */
8519 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8520 if (mdi->recovery_start != MaxSector) {
8521 recovery_not_finished = 1;
8522 break;
8523 }
8524 if (recovery_not_finished) {
1ade5cc1
N		 8525 dprintf_cont("\n");
8526 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8527 if (a->last_checkpoint < mdi->recovery_start) {
8528 a->last_checkpoint = mdi->recovery_start;
8529 super->updates_pending++;
8530 }
8531 break;
8532 }
94002678 8533 end_migration(dev, super, map_state);
238c0a71 8534 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8535 map->failed_disk_num = ~0;
8536 super->updates_pending++;
8537 a->last_checkpoint = 0;
8538 break;
8539 }
8540 if (is_gen_migration(dev)) {
1ade5cc1 8541 dprintf_cont("while general migration");
bf2f0071 8542 if (a->last_checkpoint >= a->info.component_size)
809da78e 8543 end_migration(dev, super, map_state);
94002678
AK		 8544 else
8545 map->map_state = map_state;
238c0a71 8546 map = get_imsm_map(dev, MAP_0);
28bce06f 8547 map->failed_disk_num = ~0;
94002678 8548 super->updates_pending++;
bf2f0071 8549 break;
94002678
AK		 8550 }
8551 break;
8552 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8553 dprintf_cont("degraded: ");
089f9d79 8554 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8555 dprintf_cont("mark degraded");
94002678
AK		 8556 map->map_state = map_state;
8557 super->updates_pending++;
8558 a->last_checkpoint = 0;
8559 break;
8560 }
8561 if (is_rebuilding(dev)) {
d7a1fda2 8562 dprintf_cont("while rebuilding ");
94002678 8563 if (map->map_state != map_state) {
d7a1fda2
MT		 8564 dprintf_cont("map state change ");
8565 if (n == map->failed_disk_num) {
8566 dprintf_cont("end migration");
8567 end_migration(dev, super, map_state);
8568 } else {
8569 dprintf_cont("raid10 double degradation, map state change");
8570 map->map_state = map_state;
8571 }
94002678 8572 super->updates_pending++;
d7a1fda2 8573 } else if (!rebuild_done)
fb12a745 8574 break;
d7a1fda2
MT		 8575 else if (n == map->failed_disk_num) {
8576 /* r10 double degraded to degraded transition */
8577 dprintf_cont("raid10 double degradation end migration");
8578 end_migration(dev, super, map_state);
8579 a->last_checkpoint = 0;
8580 super->updates_pending++;
fb12a745
TM		 8581 }
8582
8583 /* check if recovery is really finished */
8584 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8585 if (mdi->recovery_start != MaxSector) {
8586 recovery_not_finished = 1;
8587 break;
8588 }
8589 if (recovery_not_finished) {
8590 dprintf_cont("\n");
d7a1fda2 8591 dprintf_cont("Rebuild has not finished yet, map state changes only if raid10 double degradation happens");
fb12a745
TM		 8592 if (a->last_checkpoint < mdi->recovery_start) {
8593 a->last_checkpoint =
8594 mdi->recovery_start;
8595 super->updates_pending++;
8596 }
8597 break;
94002678 8598 }
fb12a745
TM		 8599
8600 dprintf_cont(" Rebuild done, still degraded");
8601 dev->vol.migr_state = 0;
8602 set_migr_type(dev, 0);
8603 dev->vol.curr_migr_unit = 0;
8604
8605 for (i = 0; i < map->num_members; i++) {
8606 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8607
8608 if (idx & IMSM_ORD_REBUILD)
8609 map->failed_disk_num = i;
8610 }
8611 super->updates_pending++;
94002678
AK		 8612 break;
8613 }
8614 if (is_gen_migration(dev)) {
1ade5cc1 8615 dprintf_cont("while general migration");
bf2f0071 8616 if (a->last_checkpoint >= a->info.component_size)
809da78e 8617 end_migration(dev, super, map_state);
94002678
AK		 8618 else {
8619 map->map_state = map_state;
3b451610 8620 manage_second_map(super, dev);
94002678
AK		 8621 }
8622 super->updates_pending++;
bf2f0071 8623 break;
28bce06f 8624 }
6ce1fbf1 8625 if (is_initializing(dev)) {
1ade5cc1 8626 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8627 map->map_state = map_state;
8628 super->updates_pending++;
8629 break;
8630 }
94002678
AK		 8631 break;
8632 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8633 dprintf_cont("failed: ");
94002678 8634 if (is_gen_migration(dev)) {
1ade5cc1 8635 dprintf_cont("while general migration");
94002678
AK		 8636 map->map_state = map_state;
8637 super->updates_pending++;
8638 break;
8639 }
8640 if (map->map_state != map_state) {
1ade5cc1 8641 dprintf_cont("mark failed");
94002678
AK		 8642 end_migration(dev, super, map_state);
8643 super->updates_pending++;
8644 a->last_checkpoint = 0;
8645 break;
8646 }
8647 break;
8648 default:
1ade5cc1 8649 dprintf_cont("state %i\n", map_state);
5802a811 8650 }
1ade5cc1 8651 dprintf_cont("\n");
845dea95
NB		 8652}
8653
f796af5d 8654static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8655{
f796af5d 8656 void *buf = mpb;
c2a1e7da
DW		 8657 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8658 unsigned long long dsize;
8659 unsigned long long sectors;
f36a9ecd 8660 unsigned int sector_size;
c2a1e7da 8661
f36a9ecd 8662 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8663 get_dev_size(fd, NULL, &dsize);
8664
f36a9ecd 8665 if (mpb_size > sector_size) {
272f648f 8666 /* -1 to account for anchor */
f36a9ecd 8667 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8668
272f648f 8669 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8670 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8671 SEEK_SET) < 0)
272f648f 8672 return 1;
c2a1e7da 8673
f36a9ecd
PB		 8674 if ((unsigned long long)write(fd, buf + sector_size,
8675 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8676 return 1;
8677 }
c2a1e7da 8678
272f648f 8679 /* first block is stored on second to last sector of the disk */
f36a9ecd 8680 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8681 return 1;
8682
466070ad 8683 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8684 return 1;
8685
c2a1e7da
DW		 8686 return 0;
8687}
8688
2e735d19 8689static void imsm_sync_metadata(struct supertype *container)
845dea95 8690{
2e735d19 8691 struct intel_super *super = container->sb;
c2a1e7da 8692
1a64be56 8693 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8694 if (!super->updates_pending)
8695 return;
8696
36988a3d 8697 write_super_imsm(container, 0);
c2a1e7da
DW		 8698
8699 super->updates_pending = 0;
845dea95
NB		 8700}
8701
272906ef
DW		 8702static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8703{
8704 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8705 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8706 struct dl *dl;
8707
8708 for (dl = super->disks; dl; dl = dl->next)
8709 if (dl->index == i)
8710 break;
8711
25ed7e59 8712 if (dl && is_failed(&dl->disk))
272906ef
DW		 8713 dl = NULL;
8714
8715 if (dl)
1ade5cc1 8716 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8717
8718 return dl;
8719}
8720
a20d2ba5 8721static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8722 struct active_array *a, int activate_new,
8723 struct mdinfo *additional_test_list)
272906ef
DW		 8724{
8725 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8726 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8727 struct imsm_super *mpb = super->anchor;
8728 struct imsm_map *map;
272906ef
DW		 8729 unsigned long long pos;
8730 struct mdinfo *d;
8731 struct extent *ex;
a20d2ba5 8732 int i, j;
272906ef 8733 int found;
569cc43f
DW		 8734 __u32 array_start = 0;
8735 __u32 array_end = 0;
272906ef 8736 struct dl *dl;
6c932028 8737 struct mdinfo *test_list;
272906ef
DW		 8738
8739 for (dl = super->disks; dl; dl = dl->next) {
8740 /* If in this array, skip */
8741 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8742 if (d->state_fd >= 0 &&
8743 d->disk.major == dl->major &&
272906ef 8744 d->disk.minor == dl->minor) {
8ba77d32
AK		 8745 dprintf("%x:%x already in array\n",
8746 dl->major, dl->minor);
272906ef
DW		 8747 break;
8748 }
8749 if (d)
8750 continue;
6c932028
AK		 8751 test_list = additional_test_list;
8752 while (test_list) {
8753 if (test_list->disk.major == dl->major &&
8754 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8755 dprintf("%x:%x already in additional test list\n",
8756 dl->major, dl->minor);
8757 break;
8758 }
6c932028 8759 test_list = test_list->next;
8ba77d32 8760 }
6c932028 8761 if (test_list)
8ba77d32 8762 continue;
272906ef 8763
e553d2a4 8764 /* skip in use or failed drives */
25ed7e59 8765 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8766 dl->index == -2) {
8767 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8768 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8769 continue;
8770 }
8771
a20d2ba5
DW		 8772 /* skip pure spares when we are looking for partially
8773 * assimilated drives
8774 */
8775 if (dl->index == -1 && !activate_new)
8776 continue;
8777
f2cc4f7d
AO		 8778 if (!drive_validate_sector_size(super, dl))
8779 continue;
8780
272906ef 8781 /* Does this unused device have the requisite free space?
a20d2ba5 8782 * It needs to be able to cover all member volumes
272906ef
DW		 8783 */
8784 ex = get_extents(super, dl);
8785 if (!ex) {
8786 dprintf("cannot get extents\n");
8787 continue;
8788 }
a20d2ba5
DW		 8789 for (i = 0; i < mpb->num_raid_devs; i++) {
8790 dev = get_imsm_dev(super, i);
238c0a71 8791 map = get_imsm_map(dev, MAP_0);
272906ef 8792
a20d2ba5
DW		 8793 /* check if this disk is already a member of
8794 * this array
272906ef 8795 */
620b1713 8796 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8797 continue;
8798
8799 found = 0;
8800 j = 0;
8801 pos = 0;
5551b113 8802 array_start = pba_of_lba0(map);
329c8278 8803 array_end = array_start +
44490938 8804 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8805
8806 do {
8807 /* check that we can start at pba_of_lba0 with
44490938 8808 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8809 */
329c8278 8810 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8811 found = 1;
8812 break;
8813 }
8814 pos = ex[j].start + ex[j].size;
8815 j++;
8816 } while (ex[j-1].size);
8817
8818 if (!found)
272906ef 8819 break;
a20d2ba5 8820 }
272906ef
DW		 8821
8822 free(ex);
a20d2ba5 8823 if (i < mpb->num_raid_devs) {
329c8278
DW		 8824 dprintf("%x:%x does not have %u to %u available\n",
8825 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8826 /* No room */
8827 continue;
a20d2ba5
DW		 8828 }
8829 return dl;
272906ef
DW		 8830 }
8831
8832 return dl;
8833}
8834
95d07a2c
LM		 8835static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8836{
8837 struct imsm_dev *dev2;
8838 struct imsm_map *map;
8839 struct dl *idisk;
8840 int slot;
8841 int idx;
8842 __u8 state;
8843
8844 dev2 = get_imsm_dev(cont->sb, dev_idx);
8845 if (dev2) {
238c0a71 8846 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8847 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8848 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8849 if (!map)
8850 return 1;
8851 for (slot = 0; slot < map->num_members; slot++) {
8852 /*
8853 * Check if failed disks are deleted from intel
8854 * disk list or are marked to be deleted
8855 */
238c0a71 8856 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8857 idisk = get_imsm_dl_disk(cont->sb, idx);
8858 /*
8859 * Do not rebuild the array if failed disks
8860 * from failed sub-array are not removed from
8861 * container.
8862 */
8863 if (idisk &&
8864 is_failed(&idisk->disk) &&
8865 (idisk->action != DISK_REMOVE))
8866 return 0;
8867 }
8868 }
8869 }
8870 return 1;
8871}
8872
88758e9d
DW		 8873static struct mdinfo *imsm_activate_spare(struct active_array *a,
8874 struct metadata_update **updates)
8875{
8876 /**
d23fe947
DW		 8877 * Find a device with unused free space and use it to replace a
8878 * failed/vacant region in an array. We replace failed regions one a
8879 * array at a time. The result is that a new spare disk will be added
8880 * to the first failed array and after the monitor has finished
8881 * propagating failures the remainder will be consumed.
88758e9d 8882 *
d23fe947
DW		 8883 * FIXME add a capability for mdmon to request spares from another
8884 * container.
88758e9d
DW		 8885 */
8886
8887 struct intel_super *super = a->container->sb;
88758e9d 8888 int inst = a->info.container_member;
949c47a0 8889 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8890 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8891 int failed = a->info.array.raid_disks;
8892 struct mdinfo *rv = NULL;
8893 struct mdinfo *d;
8894 struct mdinfo *di;
8895 struct metadata_update *mu;
8896 struct dl *dl;
8897 struct imsm_update_activate_spare *u;
8898 int num_spares = 0;
8899 int i;
95d07a2c 8900 int allowed;
88758e9d
DW		 8901
8902 for (d = a->info.devs ; d ; d = d->next) {
8903 if ((d->curr_state & DS_FAULTY) &&
8904 d->state_fd >= 0)
8905 /* wait for Removal to happen */
8906 return NULL;
8907 if (d->state_fd >= 0)
8908 failed--;
8909 }
8910
8911 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8912 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8913
e2962bfc
AK		 8914 if (imsm_reshape_blocks_arrays_changes(super))
8915 return NULL;
1af97990 8916
fc8ca064
AK		 8917 /* Cannot activate another spare if rebuild is in progress already
8918 */
8919 if (is_rebuilding(dev)) {
7a862a02 8920 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8921 return NULL;
8922 }
8923
89c67882
AK		 8924 if (a->info.array.level == 4)
8925 /* No repair for takeovered array
8926 * imsm doesn't support raid4
8927 */
8928 return NULL;
8929
3b451610
AK		 8930 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8931 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8932 return NULL;
8933
83ca7d45
AP		 8934 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8935 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8936 return NULL;
8937 }
8938
95d07a2c
LM		 8939 /*
8940 * If there are any failed disks check state of the other volume.
8941 * Block rebuild if the another one is failed until failed disks
8942 * are removed from container.
8943 */
8944 if (failed) {
7a862a02 8945 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8946 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8947 /* check if states of the other volumes allow for rebuild */
8948 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8949 if (i != inst) {
8950 allowed = imsm_rebuild_allowed(a->container,
8951 i, failed);
8952 if (!allowed)
8953 return NULL;
8954 }
8955 }
8956 }
8957
88758e9d 8958 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8959 for (i = 0; i < a->info.array.raid_disks; i++) {
8960 for (d = a->info.devs ; d ; d = d->next)
8961 if (d->disk.raid_disk == i)
8962 break;
8963 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
8964 if (d && (d->state_fd >= 0))
8965 continue;
8966
272906ef 8967 /*
a20d2ba5
DW		 8968 * OK, this device needs recovery. Try to re-add the
8969 * previous occupant of this slot, if this fails see if
8970 * we can continue the assimilation of a spare that was
8971 * partially assimilated, finally try to activate a new
8972 * spare.
272906ef
DW		 8973 */
8974 dl = imsm_readd(super, i, a);
8975 if (!dl)
b303fe21 8976 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 8977 if (!dl)
b303fe21 8978 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 8979 if (!dl)
8980 continue;
1011e834 8981
272906ef 8982 /* found a usable disk with enough space */
503975b9 8983 di = xcalloc(1, sizeof(*di));
272906ef
DW		 8984
8985 /* dl->index will be -1 in the case we are activating a
8986 * pristine spare. imsm_process_update() will create a
8987 * new index in this case. Once a disk is found to be
8988 * failed in all member arrays it is kicked from the
8989 * metadata
8990 */
8991 di->disk.number = dl->index;
d23fe947 8992
272906ef
DW		 8993 /* (ab)use di->devs to store a pointer to the device
8994 * we chose
8995 */
8996 di->devs = (struct mdinfo *) dl;
8997
8998 di->disk.raid_disk = i;
8999 di->disk.major = dl->major;
9000 di->disk.minor = dl->minor;
9001 di->disk.state = 0;
d23534e4 9002 di->recovery_start = 0;
5551b113 9003 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9004 di->component_size = a->info.component_size;
9005 di->container_member = inst;
5e46202e 9006 di->bb.supported = 1;
2c8890e9 9007 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9008 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9009 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9010 }
148acb7b 9011 super->random = random32();
272906ef
DW		 9012 di->next = rv;
9013 rv = di;
9014 num_spares++;
9015 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9016 i, di->data_offset);
88758e9d
DW		 9017 }
9018
9019 if (!rv)
9020 /* No spares found */
9021 return rv;
9022 /* Now 'rv' has a list of devices to return.
9023 * Create a metadata_update record to update the
9024 * disk_ord_tbl for the array
9025 */
503975b9 9026 mu = xmalloc(sizeof(*mu));
1011e834 9027 mu->buf = xcalloc(num_spares,
503975b9 9028 sizeof(struct imsm_update_activate_spare));
88758e9d 9029 mu->space = NULL;
cb23f1f4 9030 mu->space_list = NULL;
88758e9d
DW		 9031 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9032 mu->next = *updates;
9033 u = (struct imsm_update_activate_spare *) mu->buf;
9034
9035 for (di = rv ; di ; di = di->next) {
9036 u->type = update_activate_spare;
d23fe947
DW		 9037 u->dl = (struct dl *) di->devs;
9038 di->devs = NULL;
88758e9d
DW		 9039 u->slot = di->disk.raid_disk;
9040 u->array = inst;
9041 u->next = u + 1;
9042 u++;
9043 }
9044 (u-1)->next = NULL;
9045 *updates = mu;
9046
9047 return rv;
9048}
9049
54c2c1ea 9050static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9051{
54c2c1ea 9052 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9053 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9054 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9055 struct disk_info *inf = get_disk_info(u);
9056 struct imsm_disk *disk;
8273f55e
DW		 9057 int i;
9058 int j;
8273f55e 9059
54c2c1ea 9060 for (i = 0; i < map->num_members; i++) {
238c0a71 9061 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9062 for (j = 0; j < new_map->num_members; j++)
9063 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9064 return 1;
9065 }
9066
9067 return 0;
9068}
9069
1a64be56
LM		 9070static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9071{
594dc1b8
JS		 9072 struct dl *dl;
9073
1a64be56 9074 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9075 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9076 return dl;
9077 return NULL;
9078}
9079
9080static int remove_disk_super(struct intel_super *super, int major, int minor)
9081{
594dc1b8 9082 struct dl *prev;
1a64be56
LM		 9083 struct dl *dl;
9084
9085 prev = NULL;
9086 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9087 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9088 /* remove */
9089 if (prev)
9090 prev->next = dl->next;
9091 else
9092 super->disks = dl->next;
9093 dl->next = NULL;
9094 __free_imsm_disk(dl);
1ade5cc1 9095 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9096 break;
9097 }
9098 prev = dl;
9099 }
9100 return 0;
9101}
9102
f21e18ca 9103static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9104
1a64be56
LM		 9105static int add_remove_disk_update(struct intel_super *super)
9106{
9107 int check_degraded = 0;
594dc1b8
JS		 9108 struct dl *disk;
9109
1a64be56
LM		 9110 /* add/remove some spares to/from the metadata/contrainer */
9111 while (super->disk_mgmt_list) {
9112 struct dl *disk_cfg;
9113
9114 disk_cfg = super->disk_mgmt_list;
9115 super->disk_mgmt_list = disk_cfg->next;
9116 disk_cfg->next = NULL;
9117
9118 if (disk_cfg->action == DISK_ADD) {
9119 disk_cfg->next = super->disks;
9120 super->disks = disk_cfg;
9121 check_degraded = 1;
1ade5cc1
N		 9122 dprintf("added %x:%x\n",
9123 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9124 } else if (disk_cfg->action == DISK_REMOVE) {
9125 dprintf("Disk remove action processed: %x.%x\n",
9126 disk_cfg->major, disk_cfg->minor);
9127 disk = get_disk_super(super,
9128 disk_cfg->major,
9129 disk_cfg->minor);
9130 if (disk) {
9131 /* store action status */
9132 disk->action = DISK_REMOVE;
9133 /* remove spare disks only */
9134 if (disk->index == -1) {
9135 remove_disk_super(super,
9136 disk_cfg->major,
9137 disk_cfg->minor);
9138 }
9139 }
9140 /* release allocate disk structure */
9141 __free_imsm_disk(disk_cfg);
9142 }
9143 }
9144 return check_degraded;
9145}
9146
a29911da
PC		 9147static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9148 struct intel_super *super,
9149 void ***space_list)
9150{
9151 struct intel_dev *id;
9152 void **tofree = NULL;
9153 int ret_val = 0;
9154
1ade5cc1 9155 dprintf("(enter)\n");
089f9d79 9156 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9157 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9158 return ret_val;
9159 }
089f9d79 9160 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9161 dprintf("imsm: Error: Memory is not allocated\n");
9162 return ret_val;
9163 }
9164
9165 for (id = super->devlist ; id; id = id->next) {
9166 if (id->index == (unsigned)u->subdev) {
9167 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9168 struct imsm_map *map;
9169 struct imsm_dev *new_dev =
9170 (struct imsm_dev *)*space_list;
238c0a71 9171 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9172 int to_state;
9173 struct dl *new_disk;
9174
9175 if (new_dev == NULL)
9176 return ret_val;
9177 *space_list = **space_list;
9178 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9179 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9180 if (migr_map) {
9181 dprintf("imsm: Error: migration in progress");
9182 return ret_val;
9183 }
9184
9185 to_state = map->map_state;
9186 if ((u->new_level == 5) && (map->raid_level == 0)) {
9187 map->num_members++;
9188 /* this should not happen */
9189 if (u->new_disks[0] < 0) {
9190 map->failed_disk_num =
9191 map->num_members - 1;
9192 to_state = IMSM_T_STATE_DEGRADED;
9193 } else
9194 to_state = IMSM_T_STATE_NORMAL;
9195 }
8e59f3d8 9196 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9197 if (u->new_level > -1)
9198 map->raid_level = u->new_level;
238c0a71 9199 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9200 if ((u->new_level == 5) &&
9201 (migr_map->raid_level == 0)) {
9202 int ord = map->num_members - 1;
9203 migr_map->num_members--;
9204 if (u->new_disks[0] < 0)
9205 ord |= IMSM_ORD_REBUILD;
9206 set_imsm_ord_tbl_ent(map,
9207 map->num_members - 1,
9208 ord);
9209 }
9210 id->dev = new_dev;
9211 tofree = (void **)dev;
9212
4bba0439
PC		 9213 /* update chunk size
9214 */
06fb291a
PB		 9215 if (u->new_chunksize > 0) {
9216 unsigned long long num_data_stripes;
9529d343
MD		 9217 struct imsm_map *dest_map =
9218 get_imsm_map(dev, MAP_0);
06fb291a 9219 int used_disks =
9529d343 9220 imsm_num_data_members(dest_map);
06fb291a
PB		 9221
9222 if (used_disks == 0)
9223 return ret_val;
9224
4bba0439
PC		 9225 map->blocks_per_strip =
9226 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9227 num_data_stripes =
fcc2c9da 9228 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9229 num_data_stripes /= map->blocks_per_strip;
9230 num_data_stripes /= map->num_domains;
9231 set_num_data_stripes(map, num_data_stripes);
9232 }
4bba0439 9233
44490938
MD		 9234 /* ensure blocks_per_member has valid value
9235 */
9236 set_blocks_per_member(map,
9237 per_dev_array_size(map) +
9238 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9239
a29911da
PC		 9240 /* add disk
9241 */
089f9d79
JS		 9242 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9243 migr_map->raid_level == map->raid_level)
a29911da
PC		 9244 goto skip_disk_add;
9245
9246 if (u->new_disks[0] >= 0) {
9247 /* use passes spare
9248 */
9249 new_disk = get_disk_super(super,
9250 major(u->new_disks[0]),
9251 minor(u->new_disks[0]));
7a862a02 9252 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9253 major(u->new_disks[0]),
9254 minor(u->new_disks[0]),
9255 new_disk, new_disk->index);
9256 if (new_disk == NULL)
9257 goto error_disk_add;
9258
9259 new_disk->index = map->num_members - 1;
9260 /* slot to fill in autolayout
9261 */
9262 new_disk->raiddisk = new_disk->index;
9263 new_disk->disk.status |= CONFIGURED_DISK;
9264 new_disk->disk.status &= ~SPARE_DISK;
9265 } else
9266 goto error_disk_add;
9267
9268skip_disk_add:
9269 *tofree = *space_list;
9270 /* calculate new size
9271 */
f3871fdc 9272 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9273
9274 ret_val = 1;
9275 }
9276 }
9277
9278 if (tofree)
9279 *space_list = tofree;
9280 return ret_val;
9281
9282error_disk_add:
9283 dprintf("Error: imsm: Cannot find disk.\n");
9284 return ret_val;
9285}
9286
f3871fdc
AK		 9287static int apply_size_change_update(struct imsm_update_size_change *u,
9288 struct intel_super *super)
9289{
9290 struct intel_dev *id;
9291 int ret_val = 0;
9292
1ade5cc1 9293 dprintf("(enter)\n");
089f9d79 9294 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9295 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9296 return ret_val;
9297 }
9298
9299 for (id = super->devlist ; id; id = id->next) {
9300 if (id->index == (unsigned)u->subdev) {
9301 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9302 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9303 int used_disks = imsm_num_data_members(map);
f3871fdc 9304 unsigned long long blocks_per_member;
06fb291a 9305 unsigned long long num_data_stripes;
44490938
MD		 9306 unsigned long long new_size_per_disk;
9307
9308 if (used_disks == 0)
9309 return 0;
f3871fdc
AK		 9310
9311 /* calculate new size
9312 */
44490938
MD		 9313 new_size_per_disk = u->new_size / used_disks;
9314 blocks_per_member = new_size_per_disk +
9315 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9316 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9317 map->blocks_per_strip;
9318 num_data_stripes /= map->num_domains;
9319 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9320 u->new_size, new_size_per_disk,
06fb291a 9321 num_data_stripes);
f3871fdc 9322 set_blocks_per_member(map, blocks_per_member);
06fb291a 9323 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9324 imsm_set_array_size(dev, u->new_size);
9325
9326 ret_val = 1;
9327 break;
9328 }
9329 }
9330
9331 return ret_val;
9332}
9333
061d7da3 9334static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9335 struct intel_super *super,
061d7da3
LO		 9336 struct active_array *active_array)
9337{
9338 struct imsm_super *mpb = super->anchor;
9339 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9340 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9341 struct imsm_map *migr_map;
9342 struct active_array *a;
9343 struct imsm_disk *disk;
9344 __u8 to_state;
9345 struct dl *dl;
9346 unsigned int found;
9347 int failed;
5961eeec 9348 int victim;
061d7da3 9349 int i;
5961eeec 9350 int second_map_created = 0;
061d7da3 9351
5961eeec 9352 for (; u; u = u->next) {
238c0a71 9353 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9354
5961eeec 9355 if (victim < 0)
9356 return 0;
061d7da3 9357
5961eeec 9358 for (dl = super->disks; dl; dl = dl->next)
9359 if (dl == u->dl)
9360 break;
061d7da3 9361
5961eeec 9362 if (!dl) {
7a862a02 9363 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9364 u->dl->index);
9365 return 0;
9366 }
061d7da3 9367
5961eeec 9368 /* count failures (excluding rebuilds and the victim)
9369 * to determine map[0] state
9370 */
9371 failed = 0;
9372 for (i = 0; i < map->num_members; i++) {
9373 if (i == u->slot)
9374 continue;
9375 disk = get_imsm_disk(super,
238c0a71 9376 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9377 if (!disk || is_failed(disk))
9378 failed++;
9379 }
061d7da3 9380
5961eeec 9381 /* adding a pristine spare, assign a new index */
9382 if (dl->index < 0) {
9383 dl->index = super->anchor->num_disks;
9384 super->anchor->num_disks++;
9385 }
9386 disk = &dl->disk;
9387 disk->status |= CONFIGURED_DISK;
9388 disk->status &= ~SPARE_DISK;
9389
9390 /* mark rebuild */
238c0a71 9391 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9392 if (!second_map_created) {
9393 second_map_created = 1;
9394 map->map_state = IMSM_T_STATE_DEGRADED;
9395 migrate(dev, super, to_state, MIGR_REBUILD);
9396 } else
9397 map->map_state = to_state;
238c0a71 9398 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9399 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9400 set_imsm_ord_tbl_ent(migr_map, u->slot,
9401 dl->index | IMSM_ORD_REBUILD);
9402
9403 /* update the family_num to mark a new container
9404 * generation, being careful to record the existing
9405 * family_num in orig_family_num to clean up after
9406 * earlier mdadm versions that neglected to set it.
9407 */
9408 if (mpb->orig_family_num == 0)
9409 mpb->orig_family_num = mpb->family_num;
9410 mpb->family_num += super->random;
9411
9412 /* count arrays using the victim in the metadata */
9413 found = 0;
9414 for (a = active_array; a ; a = a->next) {
9415 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9416 map = get_imsm_map(dev, MAP_0);
061d7da3 9417
5961eeec 9418 if (get_imsm_disk_slot(map, victim) >= 0)
9419 found++;
9420 }
061d7da3 9421
5961eeec 9422 /* delete the victim if it is no longer being
9423 * utilized anywhere
061d7da3 9424 */
5961eeec 9425 if (!found) {
9426 struct dl **dlp;
061d7da3 9427
5961eeec 9428 /* We know that 'manager' isn't touching anything,
9429 * so it is safe to delete
9430 */
9431 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9432 if ((*dlp)->index == victim)
9433 break;
5961eeec 9434
9435 /* victim may be on the missing list */
9436 if (!*dlp)
9437 for (dlp = &super->missing; *dlp;
9438 dlp = &(*dlp)->next)
9439 if ((*dlp)->index == victim)
9440 break;
9441 imsm_delete(super, dlp, victim);
9442 }
061d7da3
LO		 9443 }
9444
9445 return 1;
9446}
a29911da 9447
2e5dc010
N		 9448static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9449 struct intel_super *super,
9450 void ***space_list)
9451{
9452 struct dl *new_disk;
9453 struct intel_dev *id;
9454 int i;
9455 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9456 int disk_count = u->old_raid_disks;
2e5dc010
N		 9457 void **tofree = NULL;
9458 int devices_to_reshape = 1;
9459 struct imsm_super *mpb = super->anchor;
9460 int ret_val = 0;
d098291a 9461 unsigned int dev_id;
2e5dc010 9462
1ade5cc1 9463 dprintf("(enter)\n");
2e5dc010
N		 9464
9465 /* enable spares to use in array */
9466 for (i = 0; i < delta_disks; i++) {
9467 new_disk = get_disk_super(super,
9468 major(u->new_disks[i]),
9469 minor(u->new_disks[i]));
7a862a02 9470 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9471 major(u->new_disks[i]), minor(u->new_disks[i]),
9472 new_disk, new_disk->index);
089f9d79
JS		 9473 if (new_disk == NULL ||
9474 (new_disk->index >= 0 &&
9475 new_disk->index < u->old_raid_disks))
2e5dc010 9476 goto update_reshape_exit;
ee4beede 9477 new_disk->index = disk_count++;
2e5dc010
N		 9478 /* slot to fill in autolayout
9479 */
9480 new_disk->raiddisk = new_disk->index;
9481 new_disk->disk.status |=
9482 CONFIGURED_DISK;
9483 new_disk->disk.status &= ~SPARE_DISK;
9484 }
9485
ed7333bd
AK		 9486 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9487 mpb->num_raid_devs);
2e5dc010
N		 9488 /* manage changes in volume
9489 */
d098291a 9490 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9491 void **sp = *space_list;
9492 struct imsm_dev *newdev;
9493 struct imsm_map *newmap, *oldmap;
9494
d098291a
AK		 9495 for (id = super->devlist ; id; id = id->next) {
9496 if (id->index == dev_id)
9497 break;
9498 }
9499 if (id == NULL)
9500 break;
2e5dc010
N		 9501 if (!sp)
9502 continue;
9503 *space_list = *sp;
9504 newdev = (void*)sp;
9505 /* Copy the dev, but not (all of) the map */
9506 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9507 oldmap = get_imsm_map(id->dev, MAP_0);
9508 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9509 /* Copy the current map */
9510 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9511 /* update one device only
9512 */
9513 if (devices_to_reshape) {
ed7333bd
AK		 9514 dprintf("imsm: modifying subdev: %i\n",
9515 id->index);
2e5dc010
N		 9516 devices_to_reshape--;
9517 newdev->vol.migr_state = 1;
9518 newdev->vol.curr_migr_unit = 0;
ea672ee1 9519 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9520 newmap->num_members = u->new_raid_disks;
9521 for (i = 0; i < delta_disks; i++) {
9522 set_imsm_ord_tbl_ent(newmap,
9523 u->old_raid_disks + i,
9524 u->old_raid_disks + i);
9525 }
9526 /* New map is correct, now need to save old map
9527 */
238c0a71 9528 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9529 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9530
f3871fdc 9531 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9532 }
9533
9534 sp = (void **)id->dev;
9535 id->dev = newdev;
9536 *sp = tofree;
9537 tofree = sp;
8e59f3d8
AK		 9538
9539 /* Clear migration record */
9540 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9541 }
819bc634
AK		 9542 if (tofree)
9543 *space_list = tofree;
2e5dc010
N		 9544 ret_val = 1;
9545
9546update_reshape_exit:
9547
9548 return ret_val;
9549}
9550
bb025c2f 9551static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9552 struct intel_super *super,
9553 void ***space_list)
bb025c2f
KW		 9554{
9555 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9556 struct intel_dev *dv;
9557 struct imsm_dev *dev_new;
bb025c2f
KW		 9558 struct imsm_map *map;
9559 struct dl *dm, *du;
8ca6df95 9560 int i;
bb025c2f
KW		 9561
9562 for (dv = super->devlist; dv; dv = dv->next)
9563 if (dv->index == (unsigned int)u->subarray) {
9564 dev = dv->dev;
9565 break;
9566 }
9567
9568 if (dev == NULL)
9569 return 0;
9570
238c0a71 9571 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9572
9573 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9574 unsigned long long num_data_stripes;
9575
43d5ec18 9576 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9577 if (imsm_count_failed(super, dev, MAP_0) !=
9578 (map->num_members / 2))
43d5ec18
KW		 9579 return 0;
9580
bb025c2f
KW		 9581 /* iterate through devices to mark removed disks as spare */
9582 for (dm = super->disks; dm; dm = dm->next) {
9583 if (dm->disk.status & FAILED_DISK) {
9584 int idx = dm->index;
9585 /* update indexes on the disk list */
9586/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9587 the index values will end up being correct.... NB */
9588 for (du = super->disks; du; du = du->next)
9589 if (du->index > idx)
9590 du->index--;
9591 /* mark as spare disk */
a8619d23 9592 mark_spare(dm);
bb025c2f
KW		 9593 }
9594 }
bb025c2f
KW		 9595 /* update map */
9596 map->num_members = map->num_members / 2;
9597 map->map_state = IMSM_T_STATE_NORMAL;
9598 map->num_domains = 1;
9599 map->raid_level = 0;
9600 map->failed_disk_num = -1;
4a353e6e
RS		 9601 num_data_stripes = imsm_dev_size(dev) / 2;
9602 num_data_stripes /= map->blocks_per_strip;
9603 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9604 }
9605
8ca6df95
KW		 9606 if (u->direction == R0_TO_R10) {
9607 void **space;
4a353e6e
RS		 9608 unsigned long long num_data_stripes;
9609
8ca6df95
KW		 9610 /* update slots in current disk list */
9611 for (dm = super->disks; dm; dm = dm->next) {
9612 if (dm->index >= 0)
9613 dm->index *= 2;
9614 }
9615 /* create new *missing* disks */
9616 for (i = 0; i < map->num_members; i++) {
9617 space = *space_list;
9618 if (!space)
9619 continue;
9620 *space_list = *space;
9621 du = (void *)space;
9622 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9623 du->fd = -1;
9624 du->minor = 0;
9625 du->major = 0;
9626 du->index = (i * 2) + 1;
9627 sprintf((char *)du->disk.serial,
9628 " MISSING_%d", du->index);
9629 sprintf((char *)du->serial,
9630 "MISSING_%d", du->index);
9631 du->next = super->missing;
9632 super->missing = du;
9633 }
9634 /* create new dev and map */
9635 space = *space_list;
9636 if (!space)
9637 return 0;
9638 *space_list = *space;
9639 dev_new = (void *)space;
9640 memcpy(dev_new, dev, sizeof(*dev));
9641 /* update new map */
238c0a71 9642 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9643 map->num_members = map->num_members * 2;
1a2487c2 9644 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9645 map->num_domains = 2;
9646 map->raid_level = 1;
4a353e6e
RS		 9647 num_data_stripes = imsm_dev_size(dev) / 2;
9648 num_data_stripes /= map->blocks_per_strip;
9649 num_data_stripes /= map->num_domains;
9650 set_num_data_stripes(map, num_data_stripes);
9651
8ca6df95
KW		 9652 /* replace dev<->dev_new */
9653 dv->dev = dev_new;
9654 }
bb025c2f
KW		 9655 /* update disk order table */
9656 for (du = super->disks; du; du = du->next)
9657 if (du->index >= 0)
9658 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9659 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9660 if (du->index >= 0) {
9661 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9662 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9663 }
bb025c2f
KW		 9664
9665 return 1;
9666}
9667
e8319a19
DW		 9668static void imsm_process_update(struct supertype *st,
9669 struct metadata_update *update)
9670{
9671 /**
9672 * crack open the metadata_update envelope to find the update record
9673 * update can be one of:
d195167d
AK		 9674 * update_reshape_container_disks - all the arrays in the container
9675 * are being reshaped to have more devices. We need to mark
9676 * the arrays for general migration and convert selected spares
9677 * into active devices.
9678 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9679 * device in an array, update the disk_ord_tbl. If this disk is
9680 * present in all member arrays then also clear the SPARE_DISK
9681 * flag
d195167d
AK		 9682 * update_create_array
9683 * update_kill_array
9684 * update_rename_array
9685 * update_add_remove_disk
e8319a19
DW		 9686 */
9687 struct intel_super *super = st->sb;
4d7b1503 9688 struct imsm_super *mpb;
e8319a19
DW		 9689 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9690
4d7b1503
DW		 9691 /* update requires a larger buf but the allocation failed */
9692 if (super->next_len && !super->next_buf) {
9693 super->next_len = 0;
9694 return;
9695 }
9696
9697 if (super->next_buf) {
9698 memcpy(super->next_buf, super->buf, super->len);
9699 free(super->buf);
9700 super->len = super->next_len;
9701 super->buf = super->next_buf;
9702
9703 super->next_len = 0;
9704 super->next_buf = NULL;
9705 }
9706
9707 mpb = super->anchor;
9708
e8319a19 9709 switch (type) {
0ec5d470
AK		 9710 case update_general_migration_checkpoint: {
9711 struct intel_dev *id;
9712 struct imsm_update_general_migration_checkpoint *u =
9713 (void *)update->buf;
9714
1ade5cc1 9715 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9716
9717 /* find device under general migration */
9718 for (id = super->devlist ; id; id = id->next) {
9719 if (is_gen_migration(id->dev)) {
9720 id->dev->vol.curr_migr_unit =
9721 __cpu_to_le32(u->curr_migr_unit);
9722 super->updates_pending++;
9723 }
9724 }
9725 break;
9726 }
bb025c2f
KW		 9727 case update_takeover: {
9728 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9729 if (apply_takeover_update(u, super, &update->space_list)) {
9730 imsm_update_version_info(super);
bb025c2f 9731 super->updates_pending++;
1a2487c2 9732 }
bb025c2f
KW		 9733 break;
9734 }
9735
78b10e66 9736 case update_reshape_container_disks: {
d195167d 9737 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9738 if (apply_reshape_container_disks_update(
9739 u, super, &update->space_list))
9740 super->updates_pending++;
78b10e66
N		 9741 break;
9742 }
48c5303a 9743 case update_reshape_migration: {
a29911da
PC		 9744 struct imsm_update_reshape_migration *u = (void *)update->buf;
9745 if (apply_reshape_migration_update(
9746 u, super, &update->space_list))
9747 super->updates_pending++;
48c5303a
PC		 9748 break;
9749 }
f3871fdc
AK		 9750 case update_size_change: {
9751 struct imsm_update_size_change *u = (void *)update->buf;
9752 if (apply_size_change_update(u, super))
9753 super->updates_pending++;
9754 break;
9755 }
e8319a19 9756 case update_activate_spare: {
1011e834 9757 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9758 if (apply_update_activate_spare(u, super, st->arrays))
9759 super->updates_pending++;
8273f55e
DW		 9760 break;
9761 }
9762 case update_create_array: {
9763 /* someone wants to create a new array, we need to be aware of
9764 * a few races/collisions:
9765 * 1/ 'Create' called by two separate instances of mdadm
9766 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9767 * devices that have since been assimilated via
9768 * activate_spare.
9769 * In the event this update can not be carried out mdadm will
9770 * (FIX ME) notice that its update did not take hold.
9771 */
9772 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9773 struct intel_dev *dv;
8273f55e
DW		 9774 struct imsm_dev *dev;
9775 struct imsm_map *map, *new_map;
9776 unsigned long long start, end;
9777 unsigned long long new_start, new_end;
9778 int i;
54c2c1ea
DW		 9779 struct disk_info *inf;
9780 struct dl *dl;
8273f55e
DW		 9781
9782 /* handle racing creates: first come first serve */
9783 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9784 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9785 goto create_error;
8273f55e
DW		 9786 }
9787
9788 /* check update is next in sequence */
9789 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9790 dprintf("can not create array %d expected index %d\n",
9791 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9792 goto create_error;
8273f55e
DW		 9793 }
9794
238c0a71 9795 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9796 new_start = pba_of_lba0(new_map);
44490938 9797 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9798 inf = get_disk_info(u);
8273f55e
DW		 9799
9800 /* handle activate_spare versus create race:
9801 * check to make sure that overlapping arrays do not include
9802 * overalpping disks
9803 */
9804 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9805 dev = get_imsm_dev(super, i);
238c0a71 9806 map = get_imsm_map(dev, MAP_0);
5551b113 9807 start = pba_of_lba0(map);
44490938 9808 end = start + per_dev_array_size(map);
8273f55e
DW		 9809 if ((new_start >= start && new_start <= end) ||
9810 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9811 /* overlap */;
9812 else
9813 continue;
9814
9815 if (disks_overlap(super, i, u)) {
1ade5cc1 9816 dprintf("arrays overlap\n");
ba2de7ba 9817 goto create_error;
8273f55e
DW		 9818 }
9819 }
8273f55e 9820
949c47a0
DW		 9821 /* check that prepare update was successful */
9822 if (!update->space) {
1ade5cc1 9823 dprintf("prepare update failed\n");
ba2de7ba 9824 goto create_error;
949c47a0
DW		 9825 }
9826
54c2c1ea
DW		 9827 /* check that all disks are still active before committing
9828 * changes. FIXME: could we instead handle this by creating a
9829 * degraded array? That's probably not what the user expects,
9830 * so better to drop this update on the floor.
9831 */
9832 for (i = 0; i < new_map->num_members; i++) {
9833 dl = serial_to_dl(inf[i].serial, super);
9834 if (!dl) {
1ade5cc1 9835 dprintf("disk disappeared\n");
ba2de7ba 9836 goto create_error;
54c2c1ea 9837 }
949c47a0
DW		 9838 }
9839
8273f55e 9840 super->updates_pending++;
54c2c1ea
DW		 9841
9842 /* convert spares to members and fixup ord_tbl */
9843 for (i = 0; i < new_map->num_members; i++) {
9844 dl = serial_to_dl(inf[i].serial, super);
9845 if (dl->index == -1) {
9846 dl->index = mpb->num_disks;
9847 mpb->num_disks++;
9848 dl->disk.status |= CONFIGURED_DISK;
9849 dl->disk.status &= ~SPARE_DISK;
9850 }
9851 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9852 }
9853
ba2de7ba
DW		 9854 dv = update->space;
9855 dev = dv->dev;
949c47a0
DW		 9856 update->space = NULL;
9857 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9858 dv->index = u->dev_idx;
9859 dv->next = super->devlist;
9860 super->devlist = dv;
8273f55e 9861 mpb->num_raid_devs++;
8273f55e 9862
4d1313e9 9863 imsm_update_version_info(super);
8273f55e 9864 break;
ba2de7ba
DW		 9865 create_error:
9866 /* mdmon knows how to release update->space, but not
9867 * ((struct intel_dev *) update->space)->dev
9868 */
9869 if (update->space) {
9870 dv = update->space;
9871 free(dv->dev);
9872 }
8273f55e 9873 break;
e8319a19 9874 }
33414a01
DW		 9875 case update_kill_array: {
9876 struct imsm_update_kill_array *u = (void *) update->buf;
9877 int victim = u->dev_idx;
9878 struct active_array *a;
9879 struct intel_dev **dp;
9880 struct imsm_dev *dev;
9881
9882 /* sanity check that we are not affecting the uuid of
9883 * active arrays, or deleting an active array
9884 *
9885 * FIXME when immutable ids are available, but note that
9886 * we'll also need to fixup the invalidated/active
9887 * subarray indexes in mdstat
9888 */
9889 for (a = st->arrays; a; a = a->next)
9890 if (a->info.container_member >= victim)
9891 break;
9892 /* by definition if mdmon is running at least one array
9893 * is active in the container, so checking
9894 * mpb->num_raid_devs is just extra paranoia
9895 */
9896 dev = get_imsm_dev(super, victim);
9897 if (a || !dev || mpb->num_raid_devs == 1) {
9898 dprintf("failed to delete subarray-%d\n", victim);
9899 break;
9900 }
9901
9902 for (dp = &super->devlist; *dp;)
f21e18ca 9903 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9904 *dp = (*dp)->next;
9905 } else {
f21e18ca 9906 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9907 (*dp)->index--;
9908 dp = &(*dp)->next;
9909 }
9910 mpb->num_raid_devs--;
9911 super->updates_pending++;
9912 break;
9913 }
aa534678
DW		 9914 case update_rename_array: {
9915 struct imsm_update_rename_array *u = (void *) update->buf;
9916 char name[MAX_RAID_SERIAL_LEN+1];
9917 int target = u->dev_idx;
9918 struct active_array *a;
9919 struct imsm_dev *dev;
9920
9921 /* sanity check that we are not affecting the uuid of
9922 * an active array
9923 */
40659392 9924 memset(name, 0, sizeof(name));
aa534678
DW		 9925 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9926 name[MAX_RAID_SERIAL_LEN] = '\0';
9927 for (a = st->arrays; a; a = a->next)
9928 if (a->info.container_member == target)
9929 break;
9930 dev = get_imsm_dev(super, u->dev_idx);
9931 if (a || !dev || !check_name(super, name, 1)) {
9932 dprintf("failed to rename subarray-%d\n", target);
9933 break;
9934 }
9935
40659392 9936 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 9937 super->updates_pending++;
9938 break;
9939 }
1a64be56 9940 case update_add_remove_disk: {
43dad3d6 9941 /* we may be able to repair some arrays if disks are
095b8088 9942 * being added, check the status of add_remove_disk
1a64be56
LM		 9943 * if discs has been added.
9944 */
9945 if (add_remove_disk_update(super)) {
43dad3d6 9946 struct active_array *a;
072b727f
DW		 9947
9948 super->updates_pending++;
1a64be56 9949 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9950 a->check_degraded = 1;
9951 }
43dad3d6 9952 break;
e8319a19 9953 }
bbab0940
TM		 9954 case update_prealloc_badblocks_mem:
9955 break;
e6e9dd3f
AP		 9956 case update_rwh_policy: {
9957 struct imsm_update_rwh_policy *u = (void *)update->buf;
9958 int target = u->dev_idx;
9959 struct imsm_dev *dev = get_imsm_dev(super, target);
9960 if (!dev) {
9961 dprintf("could not find subarray-%d\n", target);
9962 break;
9963 }
9964
9965 if (dev->rwh_policy != u->new_policy) {
9966 dev->rwh_policy = u->new_policy;
9967 super->updates_pending++;
9968 }
9969 break;
9970 }
1a64be56 9971 default:
7a862a02 9972 pr_err("error: unsuported process update type:(type: %d)\n", type);
1a64be56 9973 }
e8319a19 9974}
88758e9d 9975
bc0b9d34
PC		 9976static struct mdinfo *get_spares_for_grow(struct supertype *st);
9977
5fe6f031
N		 9978static int imsm_prepare_update(struct supertype *st,
9979 struct metadata_update *update)
8273f55e 9980{
949c47a0 9981 /**
4d7b1503
DW		 9982 * Allocate space to hold new disk entries, raid-device entries or a new
9983 * mpb if necessary. The manager synchronously waits for updates to
9984 * complete in the monitor, so new mpb buffers allocated here can be
9985 * integrated by the monitor thread without worrying about live pointers
9986 * in the manager thread.
8273f55e 9987 */
095b8088 9988 enum imsm_update_type type;
4d7b1503 9989 struct intel_super *super = st->sb;
f36a9ecd 9990 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 9991 struct imsm_super *mpb = super->anchor;
9992 size_t buf_len;
9993 size_t len = 0;
949c47a0 9994
095b8088
N		 9995 if (update->len < (int)sizeof(type))
9996 return 0;
9997
9998 type = *(enum imsm_update_type *) update->buf;
9999
949c47a0 10000 switch (type) {
0ec5d470 10001 case update_general_migration_checkpoint:
095b8088
N		 10002 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10003 return 0;
1ade5cc1 10004 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10005 break;
abedf5fc
KW		 10006 case update_takeover: {
10007 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10008 if (update->len < (int)sizeof(*u))
10009 return 0;
abedf5fc
KW		 10010 if (u->direction == R0_TO_R10) {
10011 void **tail = (void **)&update->space_list;
10012 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10013 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10014 int num_members = map->num_members;
10015 void *space;
10016 int size, i;
abedf5fc
KW		 10017 /* allocate memory for added disks */
10018 for (i = 0; i < num_members; i++) {
10019 size = sizeof(struct dl);
503975b9 10020 space = xmalloc(size);
abedf5fc
KW		 10021 *tail = space;
10022 tail = space;
10023 *tail = NULL;
10024 }
10025 /* allocate memory for new device */
10026 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10027 (num_members * sizeof(__u32));
503975b9
N		 10028 space = xmalloc(size);
10029 *tail = space;
10030 tail = space;
10031 *tail = NULL;
10032 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10033 }
10034
10035 break;
10036 }
78b10e66 10037 case update_reshape_container_disks: {
d195167d
AK		 10038 /* Every raid device in the container is about to
10039 * gain some more devices, and we will enter a
10040 * reconfiguration.
10041 * So each 'imsm_map' will be bigger, and the imsm_vol
10042 * will now hold 2 of them.
10043 * Thus we need new 'struct imsm_dev' allocations sized
10044 * as sizeof_imsm_dev but with more devices in both maps.
10045 */
10046 struct imsm_update_reshape *u = (void *)update->buf;
10047 struct intel_dev *dl;
10048 void **space_tail = (void**)&update->space_list;
10049
095b8088
N		 10050 if (update->len < (int)sizeof(*u))
10051 return 0;
10052
1ade5cc1 10053 dprintf("for update_reshape\n");
d195167d
AK		 10054
10055 for (dl = super->devlist; dl; dl = dl->next) {
10056 int size = sizeof_imsm_dev(dl->dev, 1);
10057 void *s;
d677e0b8
AK		 10058 if (u->new_raid_disks > u->old_raid_disks)
10059 size += sizeof(__u32)*2*
10060 (u->new_raid_disks - u->old_raid_disks);
503975b9 10061 s = xmalloc(size);
d195167d
AK		 10062 *space_tail = s;
10063 space_tail = s;
10064 *space_tail = NULL;
10065 }
10066
10067 len = disks_to_mpb_size(u->new_raid_disks);
10068 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10069 break;
10070 }
48c5303a 10071 case update_reshape_migration: {
bc0b9d34
PC		 10072 /* for migration level 0->5 we need to add disks
10073 * so the same as for container operation we will copy
10074 * device to the bigger location.
10075 * in memory prepared device and new disk area are prepared
10076 * for usage in process update
10077 */
10078 struct imsm_update_reshape_migration *u = (void *)update->buf;
10079 struct intel_dev *id;
10080 void **space_tail = (void **)&update->space_list;
10081 int size;
10082 void *s;
10083 int current_level = -1;
10084
095b8088
N		 10085 if (update->len < (int)sizeof(*u))
10086 return 0;
10087
1ade5cc1 10088 dprintf("for update_reshape\n");
bc0b9d34
PC		 10089
10090 /* add space for bigger array in update
10091 */
10092 for (id = super->devlist; id; id = id->next) {
10093 if (id->index == (unsigned)u->subdev) {
10094 size = sizeof_imsm_dev(id->dev, 1);
10095 if (u->new_raid_disks > u->old_raid_disks)
10096 size += sizeof(__u32)*2*
10097 (u->new_raid_disks - u->old_raid_disks);
503975b9 10098 s = xmalloc(size);
bc0b9d34
PC		 10099 *space_tail = s;
10100 space_tail = s;
10101 *space_tail = NULL;
10102 break;
10103 }
10104 }
10105 if (update->space_list == NULL)
10106 break;
10107
10108 /* add space for disk in update
10109 */
10110 size = sizeof(struct dl);
503975b9 10111 s = xmalloc(size);
bc0b9d34
PC		 10112 *space_tail = s;
10113 space_tail = s;
10114 *space_tail = NULL;
10115
10116 /* add spare device to update
10117 */
10118 for (id = super->devlist ; id; id = id->next)
10119 if (id->index == (unsigned)u->subdev) {
10120 struct imsm_dev *dev;
10121 struct imsm_map *map;
10122
10123 dev = get_imsm_dev(super, u->subdev);
238c0a71 10124 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10125 current_level = map->raid_level;
10126 break;
10127 }
089f9d79 10128 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10129 struct mdinfo *spares;
10130
10131 spares = get_spares_for_grow(st);
10132 if (spares) {
10133 struct dl *dl;
10134 struct mdinfo *dev;
10135
10136 dev = spares->devs;
10137 if (dev) {
10138 u->new_disks[0] =
10139 makedev(dev->disk.major,
10140 dev->disk.minor);
10141 dl = get_disk_super(super,
10142 dev->disk.major,
10143 dev->disk.minor);
10144 dl->index = u->old_raid_disks;
10145 dev = dev->next;
10146 }
10147 sysfs_free(spares);
10148 }
10149 }
10150 len = disks_to_mpb_size(u->new_raid_disks);
10151 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10152 break;
10153 }
f3871fdc 10154 case update_size_change: {
095b8088
N		 10155 if (update->len < (int)sizeof(struct imsm_update_size_change))
10156 return 0;
10157 break;
10158 }
10159 case update_activate_spare: {
10160 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10161 return 0;
f3871fdc
AK		 10162 break;
10163 }
949c47a0
DW		 10164 case update_create_array: {
10165 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10166 struct intel_dev *dv;
54c2c1ea 10167 struct imsm_dev *dev = &u->dev;
238c0a71 10168 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10169 struct dl *dl;
10170 struct disk_info *inf;
10171 int i;
10172 int activate = 0;
949c47a0 10173
095b8088
N		 10174 if (update->len < (int)sizeof(*u))
10175 return 0;
10176
54c2c1ea
DW		 10177 inf = get_disk_info(u);
10178 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10179 /* allocate a new super->devlist entry */
503975b9
N		 10180 dv = xmalloc(sizeof(*dv));
10181 dv->dev = xmalloc(len);
10182 update->space = dv;
949c47a0 10183
54c2c1ea
DW		 10184 /* count how many spares will be converted to members */
10185 for (i = 0; i < map->num_members; i++) {
10186 dl = serial_to_dl(inf[i].serial, super);
10187 if (!dl) {
10188 /* hmm maybe it failed?, nothing we can do about
10189 * it here
10190 */
10191 continue;
10192 }
10193 if (count_memberships(dl, super) == 0)
10194 activate++;
10195 }
10196 len += activate * sizeof(struct imsm_disk);
949c47a0 10197 break;
095b8088
N		 10198 }
10199 case update_kill_array: {
10200 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10201 return 0;
949c47a0
DW		 10202 break;
10203 }
095b8088
N		 10204 case update_rename_array: {
10205 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10206 return 0;
10207 break;
10208 }
10209 case update_add_remove_disk:
10210 /* no update->len needed */
10211 break;
bbab0940
TM		 10212 case update_prealloc_badblocks_mem:
10213 super->extra_space += sizeof(struct bbm_log) -
10214 get_imsm_bbm_log_size(super->bbm_log);
10215 break;
e6e9dd3f
AP		 10216 case update_rwh_policy: {
10217 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10218 return 0;
10219 break;
10220 }
095b8088
N		 10221 default:
10222 return 0;
949c47a0 10223 }
8273f55e 10224
4d7b1503
DW		 10225 /* check if we need a larger metadata buffer */
10226 if (super->next_buf)
10227 buf_len = super->next_len;
10228 else
10229 buf_len = super->len;
10230
bbab0940 10231 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10232 /* ok we need a larger buf than what is currently allocated
10233 * if this allocation fails process_update will notice that
10234 * ->next_len is set and ->next_buf is NULL
10235 */
bbab0940
TM		 10236 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10237 super->extra_space + len, sector_size);
4d7b1503
DW		 10238 if (super->next_buf)
10239 free(super->next_buf);
10240
10241 super->next_len = buf_len;
f36a9ecd 10242 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10243 memset(super->next_buf, 0, buf_len);
10244 else
4d7b1503
DW		 10245 super->next_buf = NULL;
10246 }
5fe6f031 10247 return 1;
8273f55e
DW		 10248}
10249
ae6aad82 10250/* must be called while manager is quiesced */
f21e18ca 10251static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10252{
10253 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10254 struct dl *iter;
10255 struct imsm_dev *dev;
10256 struct imsm_map *map;
4c9e8c1e 10257 unsigned int i, j, num_members;
fb12a745 10258 __u32 ord, ord_map0;
4c9e8c1e 10259 struct bbm_log *log = super->bbm_log;
ae6aad82 10260
1ade5cc1 10261 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10262
10263 /* shift all indexes down one */
10264 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10265 if (iter->index > (int)index)
ae6aad82 10266 iter->index--;
47ee5a45 10267 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10268 if (iter->index > (int)index)
47ee5a45 10269 iter->index--;
ae6aad82
DW		 10270
10271 for (i = 0; i < mpb->num_raid_devs; i++) {
10272 dev = get_imsm_dev(super, i);
238c0a71 10273 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10274 num_members = map->num_members;
10275 for (j = 0; j < num_members; j++) {
10276 /* update ord entries being careful not to propagate
10277 * ord-flags to the first map
10278 */
238c0a71 10279 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10280 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10281
24565c9a
DW		 10282 if (ord_to_idx(ord) <= index)
10283 continue;
ae6aad82 10284
238c0a71 10285 map = get_imsm_map(dev, MAP_0);
fb12a745 10286 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10287 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10288 if (map)
10289 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10290 }
10291 }
10292
4c9e8c1e
TM		 10293 for (i = 0; i < log->entry_count; i++) {
10294 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10295
10296 if (entry->disk_ordinal <= index)
10297 continue;
10298 entry->disk_ordinal--;
10299 }
10300
ae6aad82
DW		 10301 mpb->num_disks--;
10302 super->updates_pending++;
24565c9a
DW		 10303 if (*dlp) {
10304 struct dl *dl = *dlp;
10305
10306 *dlp = (*dlp)->next;
10307 __free_imsm_disk(dl);
10308 }
ae6aad82 10309}
9a717282
AK		 10310
10311static void close_targets(int *targets, int new_disks)
10312{
10313 int i;
10314
10315 if (!targets)
10316 return;
10317
10318 for (i = 0; i < new_disks; i++) {
10319 if (targets[i] >= 0) {
10320 close(targets[i]);
10321 targets[i] = -1;
10322 }
10323 }
10324}
10325
10326static int imsm_get_allowed_degradation(int level, int raid_disks,
10327 struct intel_super *super,
10328 struct imsm_dev *dev)
10329{
10330 switch (level) {
bf5cf7c7 10331 case 1:
9a717282
AK		 10332 case 10:{
10333 int ret_val = 0;
10334 struct imsm_map *map;
10335 int i;
10336
10337 ret_val = raid_disks/2;
10338 /* check map if all disks pairs not failed
10339 * in both maps
10340 */
238c0a71 10341 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10342 for (i = 0; i < ret_val; i++) {
10343 int degradation = 0;
10344 if (get_imsm_disk(super, i) == NULL)
10345 degradation++;
10346 if (get_imsm_disk(super, i + 1) == NULL)
10347 degradation++;
10348 if (degradation == 2)
10349 return 0;
10350 }
238c0a71 10351 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10352 /* if there is no second map
10353 * result can be returned
10354 */
10355 if (map == NULL)
10356 return ret_val;
10357 /* check degradation in second map
10358 */
10359 for (i = 0; i < ret_val; i++) {
10360 int degradation = 0;
10361 if (get_imsm_disk(super, i) == NULL)
10362 degradation++;
10363 if (get_imsm_disk(super, i + 1) == NULL)
10364 degradation++;
10365 if (degradation == 2)
10366 return 0;
10367 }
10368 return ret_val;
10369 }
10370 case 5:
10371 return 1;
10372 case 6:
10373 return 2;
10374 default:
10375 return 0;
10376 }
10377}
10378
687629c2
AK		 10379/*******************************************************************************
10380 * Function: open_backup_targets
10381 * Description: Function opens file descriptors for all devices given in
10382 * info->devs
10383 * Parameters:
10384 * info : general array info
10385 * raid_disks : number of disks
10386 * raid_fds : table of device's file descriptors
9a717282
AK		 10387 * super : intel super for raid10 degradation check
10388 * dev : intel device for raid10 degradation check
687629c2
AK		 10389 * Returns:
10390 * 0 : success
10391 * -1 : fail
10392 ******************************************************************************/
9a717282
AK		 10393int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10394 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10395{
10396 struct mdinfo *sd;
f627f5ad 10397 int i;
9a717282 10398 int opened = 0;
f627f5ad
AK		 10399
10400 for (i = 0; i < raid_disks; i++)
10401 raid_fds[i] = -1;
687629c2
AK		 10402
10403 for (sd = info->devs ; sd ; sd = sd->next) {
10404 char *dn;
10405
10406 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10407 dprintf("disk is faulty!!\n");
10408 continue;
10409 }
10410
089f9d79 10411 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10412 continue;
10413
10414 dn = map_dev(sd->disk.major,
10415 sd->disk.minor, 1);
10416 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10417 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10418 pr_err("cannot open component\n");
9a717282 10419 continue;
687629c2 10420 }
9a717282
AK		 10421 opened++;
10422 }
10423 /* check if maximum array degradation level is not exceeded
10424 */
10425 if ((raid_disks - opened) >
089f9d79
JS		 10426 imsm_get_allowed_degradation(info->new_level, raid_disks,
10427 super, dev)) {
e12b3daa 10428 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10429 close_targets(raid_fds, raid_disks);
10430 return -2;
687629c2
AK		 10431 }
10432 return 0;
10433}
10434
d31ad643
PB		 10435/*******************************************************************************
10436 * Function: validate_container_imsm
10437 * Description: This routine validates container after assemble,
10438 * eg. if devices in container are under the same controller.
10439 *
10440 * Parameters:
10441 * info : linked list with info about devices used in array
10442 * Returns:
10443 * 1 : HBA mismatch
10444 * 0 : Success
10445 ******************************************************************************/
10446int validate_container_imsm(struct mdinfo *info)
10447{
6b781d33
AP		 10448 if (check_env("IMSM_NO_PLATFORM"))
10449 return 0;
d31ad643 10450
6b781d33
AP		 10451 struct sys_dev *idev;
10452 struct sys_dev *hba = NULL;
10453 struct sys_dev *intel_devices = find_intel_devices();
10454 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10455 info->disk.minor));
10456
10457 for (idev = intel_devices; idev; idev = idev->next) {
10458 if (dev_path && strstr(dev_path, idev->path)) {
10459 hba = idev;
10460 break;
d31ad643 10461 }
6b781d33
AP		 10462 }
10463 if (dev_path)
d31ad643
PB		 10464 free(dev_path);
10465
6b781d33
AP		 10466 if (!hba) {
10467 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10468 devid2kname(makedev(info->disk.major, info->disk.minor)));
10469 return 1;
10470 }
10471
10472 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10473 struct mdinfo *dev;
10474
10475 for (dev = info->next; dev; dev = dev->next) {
10476 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10477
10478 struct sys_dev *hba2 = NULL;
10479 for (idev = intel_devices; idev; idev = idev->next) {
10480 if (dev_path && strstr(dev_path, idev->path)) {
10481 hba2 = idev;
10482 break;
d31ad643
PB		 10483 }
10484 }
6b781d33
AP		 10485 if (dev_path)
10486 free(dev_path);
10487
10488 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10489 get_orom_by_device_id(hba2->dev_id);
10490
10491 if (hba2 && hba->type != hba2->type) {
10492 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10493 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10494 return 1;
10495 }
10496
07cb1e57 10497 if (orom != orom2) {
6b781d33
AP		 10498 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10499 " This operation is not supported and can lead to data loss.\n");
10500 return 1;
10501 }
10502
10503 if (!orom) {
10504 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10505 " This operation is not supported and can lead to data loss.\n");
10506 return 1;
10507 }
d31ad643 10508 }
6b781d33 10509
d31ad643
PB		 10510 return 0;
10511}
32141c17 10512
6f50473f
TM		 10513/*******************************************************************************
10514* Function: imsm_record_badblock
10515* Description: This routine stores new bad block record in BBM log
10516*
10517* Parameters:
10518* a : array containing a bad block
10519* slot : disk number containing a bad block
10520* sector : bad block sector
10521* length : bad block sectors range
10522* Returns:
10523* 1 : Success
10524* 0 : Error
10525******************************************************************************/
10526static int imsm_record_badblock(struct active_array *a, int slot,
10527 unsigned long long sector, int length)
10528{
10529 struct intel_super *super = a->container->sb;
10530 int ord;
10531 int ret;
10532
10533 ord = imsm_disk_slot_to_ord(a, slot);
10534 if (ord < 0)
10535 return 0;
10536
10537 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10538 length);
10539 if (ret)
10540 super->updates_pending++;
10541
10542 return ret;
10543}
c07a5a4f
TM		 10544/*******************************************************************************
10545* Function: imsm_clear_badblock
10546* Description: This routine clears bad block record from BBM log
10547*
10548* Parameters:
10549* a : array containing a bad block
10550* slot : disk number containing a bad block
10551* sector : bad block sector
10552* length : bad block sectors range
10553* Returns:
10554* 1 : Success
10555* 0 : Error
10556******************************************************************************/
10557static int imsm_clear_badblock(struct active_array *a, int slot,
10558 unsigned long long sector, int length)
10559{
10560 struct intel_super *super = a->container->sb;
10561 int ord;
10562 int ret;
10563
10564 ord = imsm_disk_slot_to_ord(a, slot);
10565 if (ord < 0)
10566 return 0;
10567
10568 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10569 if (ret)
10570 super->updates_pending++;
10571
10572 return ret;
10573}
928f1424
TM		 10574/*******************************************************************************
10575* Function: imsm_get_badblocks
10576* Description: This routine get list of bad blocks for an array
10577*
10578* Parameters:
10579* a : array
10580* slot : disk number
10581* Returns:
10582* bb : structure containing bad blocks
10583* NULL : error
10584******************************************************************************/
10585static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10586{
10587 int inst = a->info.container_member;
10588 struct intel_super *super = a->container->sb;
10589 struct imsm_dev *dev = get_imsm_dev(super, inst);
10590 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10591 int ord;
10592
10593 ord = imsm_disk_slot_to_ord(a, slot);
10594 if (ord < 0)
10595 return NULL;
10596
10597 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10598 per_dev_array_size(map), &super->bb);
928f1424
TM		 10599
10600 return &super->bb;
10601}
27156a57
TM		 10602/*******************************************************************************
10603* Function: examine_badblocks_imsm
10604* Description: Prints list of bad blocks on a disk to the standard output
10605*
10606* Parameters:
10607* st : metadata handler
10608* fd : open file descriptor for device
10609* devname : device name
10610* Returns:
10611* 0 : Success
10612* 1 : Error
10613******************************************************************************/
10614static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10615{
10616 struct intel_super *super = st->sb;
10617 struct bbm_log *log = super->bbm_log;
10618 struct dl *d = NULL;
10619 int any = 0;
10620
10621 for (d = super->disks; d ; d = d->next) {
10622 if (strcmp(d->devname, devname) == 0)
10623 break;
10624 }
10625
10626 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10627 pr_err("%s doesn't appear to be part of a raid array\n",
10628 devname);
10629 return 1;
10630 }
10631
10632 if (log != NULL) {
10633 unsigned int i;
10634 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10635
10636 for (i = 0; i < log->entry_count; i++) {
10637 if (entry[i].disk_ordinal == d->index) {
10638 unsigned long long sector = __le48_to_cpu(
10639 &entry[i].defective_block_start);
10640 int cnt = entry[i].marked_count + 1;
10641
10642 if (!any) {
10643 printf("Bad-blocks on %s:\n", devname);
10644 any = 1;
10645 }
10646
10647 printf("%20llu for %d sectors\n", sector, cnt);
10648 }
10649 }
10650 }
10651
10652 if (!any)
10653 printf("No bad-blocks list configured on %s\n", devname);
10654
10655 return 0;
10656}
687629c2
AK		 10657/*******************************************************************************
10658 * Function: init_migr_record_imsm
10659 * Description: Function inits imsm migration record
10660 * Parameters:
10661 * super : imsm internal array info
10662 * dev : device under migration
10663 * info : general array info to find the smallest device
10664 * Returns:
10665 * none
10666 ******************************************************************************/
10667void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10668 struct mdinfo *info)
10669{
10670 struct intel_super *super = st->sb;
10671 struct migr_record *migr_rec = super->migr_rec;
10672 int new_data_disks;
10673 unsigned long long dsize, dev_sectors;
10674 long long unsigned min_dev_sectors = -1LLU;
10675 struct mdinfo *sd;
10676 char nm[30];
10677 int fd;
238c0a71
AK		 10678 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10679 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10680 unsigned long long num_migr_units;
3ef4403c 10681 unsigned long long array_blocks;
687629c2
AK		 10682
10683 memset(migr_rec, 0, sizeof(struct migr_record));
10684 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10685
10686 /* only ascending reshape supported now */
10687 migr_rec->ascending_migr = __cpu_to_le32(1);
10688
10689 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10690 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10691 migr_rec->dest_depth_per_unit *=
10692 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10693 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10694 migr_rec->blocks_per_unit =
10695 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10696 migr_rec->dest_depth_per_unit =
10697 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10698 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10699 num_migr_units =
10700 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10701
10702 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10703 num_migr_units++;
10704 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
10705
10706 migr_rec->post_migr_vol_cap = dev->size_low;
10707 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10708
687629c2
AK		 10709 /* Find the smallest dev */
10710 for (sd = info->devs ; sd ; sd = sd->next) {
10711 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10712 fd = dev_open(nm, O_RDONLY);
10713 if (fd < 0)
10714 continue;
10715 get_dev_size(fd, NULL, &dsize);
10716 dev_sectors = dsize / 512;
10717 if (dev_sectors < min_dev_sectors)
10718 min_dev_sectors = dev_sectors;
10719 close(fd);
10720 }
10721 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
10722 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10723
10724 write_imsm_migr_rec(st);
10725
10726 return;
10727}
10728
10729/*******************************************************************************
10730 * Function: save_backup_imsm
10731 * Description: Function saves critical data stripes to Migration Copy Area
10732 * and updates the current migration unit status.
10733 * Use restore_stripes() to form a destination stripe,
10734 * and to write it to the Copy Area.
10735 * Parameters:
10736 * st : supertype information
aea93171 10737 * dev : imsm device that backup is saved for
687629c2
AK		 10738 * info : general array info
10739 * buf : input buffer
687629c2
AK		 10740 * length : length of data to backup (blocks_per_unit)
10741 * Returns:
10742 * 0 : success
10743 *, -1 : fail
10744 ******************************************************************************/
10745int save_backup_imsm(struct supertype *st,
10746 struct imsm_dev *dev,
10747 struct mdinfo *info,
10748 void *buf,
687629c2
AK		 10749 int length)
10750{
10751 int rv = -1;
10752 struct intel_super *super = st->sb;
594dc1b8
JS		 10753 unsigned long long *target_offsets;
10754 int *targets;
687629c2 10755 int i;
238c0a71 10756 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10757 int new_disks = map_dest->num_members;
ab724b98
AK		 10758 int dest_layout = 0;
10759 int dest_chunk;
d1877f69 10760 unsigned long long start;
9529d343 10761 int data_disks = imsm_num_data_members(map_dest);
687629c2 10762
503975b9 10763 targets = xmalloc(new_disks * sizeof(int));
687629c2 10764
7e45b550
AK		 10765 for (i = 0; i < new_disks; i++)
10766 targets[i] = -1;
10767
503975b9 10768 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10769
d1877f69 10770 start = info->reshape_progress * 512;
687629c2 10771 for (i = 0; i < new_disks; i++) {
687629c2
AK		 10772 target_offsets[i] = (unsigned long long)
10773 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK		 10774 /* move back copy area adderss, it will be moved forward
10775 * in restore_stripes() using start input variable
10776 */
10777 target_offsets[i] -= start/data_disks;
687629c2
AK		 10778 }
10779
9a717282
AK		 10780 if (open_backup_targets(info, new_disks, targets,
10781 super, dev))
687629c2
AK		 10782 goto abort;
10783
68eb8bc6 10784 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10785 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10786
687629c2
AK		 10787 if (restore_stripes(targets, /* list of dest devices */
10788 target_offsets, /* migration record offsets */
10789 new_disks,
ab724b98
AK		 10790 dest_chunk,
10791 map_dest->raid_level,
10792 dest_layout,
10793 -1, /* source backup file descriptor */
10794 0, /* input buf offset
10795 * always 0 buf is already offseted */
d1877f69 10796 start,
687629c2
AK		 10797 length,
10798 buf) != 0) {
e7b84f9d 10799 pr_err("Error restoring stripes\n");
687629c2
AK		 10800 goto abort;
10801 }
10802
10803 rv = 0;
10804
10805abort:
10806 if (targets) {
9a717282 10807 close_targets(targets, new_disks);
687629c2
AK		 10808 free(targets);
10809 }
10810 free(target_offsets);
10811
10812 return rv;
10813}
10814
10815/*******************************************************************************
10816 * Function: save_checkpoint_imsm
10817 * Description: Function called for current unit status update
10818 * in the migration record. It writes it to disk.
10819 * Parameters:
10820 * super : imsm internal array info
10821 * info : general array info
10822 * Returns:
10823 * 0: success
10824 * 1: failure
0228d92c
AK		 10825 * 2: failure, means no valid migration record
10826 * / no general migration in progress /
687629c2
AK		 10827 ******************************************************************************/
10828int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10829{
10830 struct intel_super *super = st->sb;
f8b72ef5
AK		 10831 unsigned long long blocks_per_unit;
10832 unsigned long long curr_migr_unit;
10833
2e062e82 10834 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10835 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10836 return 1;
10837 }
10838
f8b72ef5
AK		 10839 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10840 if (blocks_per_unit == 0) {
0228d92c
AK		 10841 dprintf("imsm: no migration in progress.\n");
10842 return 2;
687629c2 10843 }
f8b72ef5
AK		 10844 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10845 /* check if array is alligned to copy area
10846 * if it is not alligned, add one to current migration unit value
10847 * this can happend on array reshape finish only
10848 */
10849 if (info->reshape_progress % blocks_per_unit)
10850 curr_migr_unit++;
687629c2
AK		 10851
10852 super->migr_rec->curr_migr_unit =
f8b72ef5 10853 __cpu_to_le32(curr_migr_unit);
687629c2
AK		 10854 super->migr_rec->rec_status = __cpu_to_le32(state);
10855 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK		 10856 __cpu_to_le32(curr_migr_unit *
10857 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2 10858 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10859 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10860 return 1;
10861 }
10862
10863 return 0;
10864}
10865
276d77db
AK		 10866/*******************************************************************************
10867 * Function: recover_backup_imsm
10868 * Description: Function recovers critical data from the Migration Copy Area
10869 * while assembling an array.
10870 * Parameters:
10871 * super : imsm internal array info
10872 * info : general array info
10873 * Returns:
10874 * 0 : success (or there is no data to recover)
10875 * 1 : fail
10876 ******************************************************************************/
10877int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10878{
10879 struct intel_super *super = st->sb;
10880 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10881 struct imsm_map *map_dest;
276d77db
AK		 10882 struct intel_dev *id = NULL;
10883 unsigned long long read_offset;
10884 unsigned long long write_offset;
10885 unsigned unit_len;
10886 int *targets = NULL;
10887 int new_disks, i, err;
10888 char *buf = NULL;
10889 int retval = 1;
f36a9ecd 10890 unsigned int sector_size = super->sector_size;
276d77db
AK		 10891 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
10892 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 10893 char buffer[20];
6c3560c0 10894 int skipped_disks = 0;
276d77db
AK		 10895
10896 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10897 if (err < 1)
10898 return 1;
10899
10900 /* recover data only during assemblation */
10901 if (strncmp(buffer, "inactive", 8) != 0)
10902 return 0;
10903 /* no data to recover */
10904 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10905 return 0;
10906 if (curr_migr_unit >= num_migr_units)
10907 return 1;
10908
10909 /* find device during reshape */
10910 for (id = super->devlist; id; id = id->next)
10911 if (is_gen_migration(id->dev))
10912 break;
10913 if (id == NULL)
10914 return 1;
10915
238c0a71 10916 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10917 new_disks = map_dest->num_members;
10918
10919 read_offset = (unsigned long long)
10920 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
10921
10922 write_offset = ((unsigned long long)
10923 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 10924 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10925
10926 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10927 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10928 goto abort;
503975b9 10929 targets = xcalloc(new_disks, sizeof(int));
276d77db 10930
9a717282 10931 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10932 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10933 goto abort;
10934 }
276d77db
AK		 10935
10936 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10937 if (targets[i] < 0) {
10938 skipped_disks++;
10939 continue;
10940 }
276d77db 10941 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10942 pr_err("Cannot seek to block: %s\n",
10943 strerror(errno));
137debce
AK		 10944 skipped_disks++;
10945 continue;
276d77db 10946 }
9ec11d1a 10947 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10948 pr_err("Cannot read copy area block: %s\n",
10949 strerror(errno));
137debce
AK		 10950 skipped_disks++;
10951 continue;
276d77db
AK		 10952 }
10953 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10954 pr_err("Cannot seek to block: %s\n",
10955 strerror(errno));
137debce
AK		 10956 skipped_disks++;
10957 continue;
276d77db 10958 }
9ec11d1a 10959 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10960 pr_err("Cannot restore block: %s\n",
10961 strerror(errno));
137debce
AK		 10962 skipped_disks++;
10963 continue;
276d77db
AK		 10964 }
10965 }
10966
137debce
AK		 10967 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10968 new_disks,
10969 super,
10970 id->dev)) {
7a862a02 10971 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10972 goto abort;
10973 }
10974
befb629b
AK		 10975 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10976 /* ignore error == 2, this can mean end of reshape here
10977 */
7a862a02 10978 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10979 } else
276d77db 10980 retval = 0;
276d77db
AK		 10981
10982abort:
10983 if (targets) {
10984 for (i = 0; i < new_disks; i++)
10985 if (targets[i])
10986 close(targets[i]);
10987 free(targets);
10988 }
10989 free(buf);
10990 return retval;
10991}
10992
2cda7640
ML		 10993static char disk_by_path[] = "/dev/disk/by-path/";
10994
10995static const char *imsm_get_disk_controller_domain(const char *path)
10996{
2cda7640 10997 char disk_path[PATH_MAX];
96234762
LM		 10998 char *drv=NULL;
10999 struct stat st;
2cda7640 11000
6d8d290a 11001 strcpy(disk_path, disk_by_path);
96234762
LM		 11002 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11003 if (stat(disk_path, &st) == 0) {
11004 struct sys_dev* hba;
594dc1b8 11005 char *path;
96234762
LM		 11006
11007 path = devt_to_devpath(st.st_rdev);
11008 if (path == NULL)
11009 return "unknown";
11010 hba = find_disk_attached_hba(-1, path);
11011 if (hba && hba->type == SYS_DEV_SAS)
11012 drv = "isci";
11013 else if (hba && hba->type == SYS_DEV_SATA)
11014 drv = "ahci";
c6839718
MT		 11015 else if (hba && hba->type == SYS_DEV_VMD)
11016 drv = "vmd";
11017 else if (hba && hba->type == SYS_DEV_NVME)
11018 drv = "nvme";
1011e834 11019 else
96234762
LM		 11020 drv = "unknown";
11021 dprintf("path: %s hba: %s attached: %s\n",
11022 path, (hba) ? hba->path : "NULL", drv);
11023 free(path);
2cda7640 11024 }
96234762 11025 return drv;
2cda7640
ML		 11026}
11027
4dd2df09 11028static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11029{
4dd2df09 11030 static char devnm[32];
78b10e66
N		 11031 char subdev_name[20];
11032 struct mdstat_ent *mdstat;
11033
11034 sprintf(subdev_name, "%d", subdev);
11035 mdstat = mdstat_by_subdev(subdev_name, container);
11036 if (!mdstat)
4dd2df09 11037 return NULL;
78b10e66 11038
4dd2df09 11039 strcpy(devnm, mdstat->devnm);
78b10e66 11040 free_mdstat(mdstat);
4dd2df09 11041 return devnm;
78b10e66
N		 11042}
11043
11044static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11045 struct geo_params *geo,
fbf3d202
AK		 11046 int *old_raid_disks,
11047 int direction)
78b10e66 11048{
694575e7
KW		 11049 /* currently we only support increasing the number of devices
11050 * for a container. This increases the number of device for each
11051 * member array. They must all be RAID0 or RAID5.
11052 */
78b10e66
N		 11053 int ret_val = 0;
11054 struct mdinfo *info, *member;
11055 int devices_that_can_grow = 0;
11056
7a862a02 11057 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11058
d04f65f4 11059 if (geo->size > 0 ||
78b10e66
N		 11060 geo->level != UnSet ||
11061 geo->layout != UnSet ||
11062 geo->chunksize != 0 ||
11063 geo->raid_disks == UnSet) {
7a862a02 11064 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11065 return ret_val;
11066 }
11067
fbf3d202 11068 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11069 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11070 return ret_val;
11071 }
11072
78b10e66
N		 11073 info = container_content_imsm(st, NULL);
11074 for (member = info; member; member = member->next) {
4dd2df09 11075 char *result;
78b10e66
N		 11076
11077 dprintf("imsm: checking device_num: %i\n",
11078 member->container_member);
11079
d7d205bd 11080 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11081 /* we work on container for Online Capacity Expansion
11082 * only so raid_disks has to grow
11083 */
7a862a02 11084 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11085 break;
11086 }
11087
089f9d79 11088 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11089 /* we cannot use this container with other raid level
11090 */
7a862a02 11091 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11092 info->array.level);
11093 break;
11094 } else {
11095 /* check for platform support
11096 * for this raid level configuration
11097 */
11098 struct intel_super *super = st->sb;
11099 if (!is_raid_level_supported(super->orom,
11100 member->array.level,
11101 geo->raid_disks)) {
7a862a02 11102 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11103 info->array.level,
11104 geo->raid_disks,
11105 geo->raid_disks > 1 ? "s" : "");
11106 break;
11107 }
2a4a08e7
AK		 11108 /* check if component size is aligned to chunk size
11109 */
11110 if (info->component_size %
11111 (info->array.chunk_size/512)) {
7a862a02 11112 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11113 break;
11114 }
78b10e66
N		 11115 }
11116
11117 if (*old_raid_disks &&
11118 info->array.raid_disks != *old_raid_disks)
11119 break;
11120 *old_raid_disks = info->array.raid_disks;
11121
11122 /* All raid5 and raid0 volumes in container
11123 * have to be ready for Online Capacity Expansion
11124 * so they need to be assembled. We have already
11125 * checked that no recovery etc is happening.
11126 */
4dd2df09
N		 11127 result = imsm_find_array_devnm_by_subdev(member->container_member,
11128 st->container_devnm);
11129 if (result == NULL) {
78b10e66
N		 11130 dprintf("imsm: cannot find array\n");
11131 break;
11132 }
11133 devices_that_can_grow++;
11134 }
11135 sysfs_free(info);
11136 if (!member && devices_that_can_grow)
11137 ret_val = 1;
11138
11139 if (ret_val)
1ade5cc1 11140 dprintf("Container operation allowed\n");
78b10e66 11141 else
1ade5cc1 11142 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11143
11144 return ret_val;
11145}
11146
11147/* Function: get_spares_for_grow
11148 * Description: Allocates memory and creates list of spare devices
1011e834 11149 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11150 * Parameters: Pointer to the supertype structure
11151 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11152 * NULL if fail
78b10e66
N		 11153 */
11154static struct mdinfo *get_spares_for_grow(struct supertype *st)
11155{
fbfdcb06
AO		 11156 struct spare_criteria sc;
11157
11158 get_spare_criteria_imsm(st, &sc);
11159 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11160}
11161
11162/******************************************************************************
11163 * function: imsm_create_metadata_update_for_reshape
11164 * Function creates update for whole IMSM container.
11165 *
11166 ******************************************************************************/
11167static int imsm_create_metadata_update_for_reshape(
11168 struct supertype *st,
11169 struct geo_params *geo,
11170 int old_raid_disks,
11171 struct imsm_update_reshape **updatep)
11172{
11173 struct intel_super *super = st->sb;
11174 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11175 int update_memory_size;
11176 struct imsm_update_reshape *u;
11177 struct mdinfo *spares;
78b10e66 11178 int i;
594dc1b8 11179 int delta_disks;
bbd24d86 11180 struct mdinfo *dev;
78b10e66 11181
1ade5cc1 11182 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11183
11184 delta_disks = geo->raid_disks - old_raid_disks;
11185
11186 /* size of all update data without anchor */
11187 update_memory_size = sizeof(struct imsm_update_reshape);
11188
11189 /* now add space for spare disks that we need to add. */
11190 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11191
503975b9 11192 u = xcalloc(1, update_memory_size);
78b10e66
N		 11193 u->type = update_reshape_container_disks;
11194 u->old_raid_disks = old_raid_disks;
11195 u->new_raid_disks = geo->raid_disks;
11196
11197 /* now get spare disks list
11198 */
11199 spares = get_spares_for_grow(st);
11200
d7be7d87 11201 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11202 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11203 i = -1;
78b10e66
N		 11204 goto abort;
11205 }
11206
11207 /* we have got spares
11208 * update disk list in imsm_disk list table in anchor
11209 */
11210 dprintf("imsm: %i spares are available.\n\n",
11211 spares->array.spare_disks);
11212
bbd24d86 11213 dev = spares->devs;
78b10e66 11214 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11215 struct dl *dl;
11216
bbd24d86
AK		 11217 if (dev == NULL)
11218 break;
78b10e66
N		 11219 u->new_disks[i] = makedev(dev->disk.major,
11220 dev->disk.minor);
11221 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11222 dl->index = mpb->num_disks;
11223 mpb->num_disks++;
bbd24d86 11224 dev = dev->next;
78b10e66 11225 }
78b10e66
N		 11226
11227abort:
11228 /* free spares
11229 */
11230 sysfs_free(spares);
11231
d677e0b8 11232 dprintf("imsm: reshape update preparation :");
78b10e66 11233 if (i == delta_disks) {
1ade5cc1 11234 dprintf_cont(" OK\n");
78b10e66
N		 11235 *updatep = u;
11236 return update_memory_size;
11237 }
11238 free(u);
1ade5cc1 11239 dprintf_cont(" Error\n");
78b10e66
N		 11240
11241 return 0;
11242}
11243
f3871fdc
AK		 11244/******************************************************************************
11245 * function: imsm_create_metadata_update_for_size_change()
11246 * Creates update for IMSM array for array size change.
11247 *
11248 ******************************************************************************/
11249static int imsm_create_metadata_update_for_size_change(
11250 struct supertype *st,
11251 struct geo_params *geo,
11252 struct imsm_update_size_change **updatep)
11253{
11254 struct intel_super *super = st->sb;
594dc1b8
JS		 11255 int update_memory_size;
11256 struct imsm_update_size_change *u;
f3871fdc 11257
1ade5cc1 11258 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11259
11260 /* size of all update data without anchor */
11261 update_memory_size = sizeof(struct imsm_update_size_change);
11262
503975b9 11263 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11264 u->type = update_size_change;
11265 u->subdev = super->current_vol;
11266 u->new_size = geo->size;
11267
11268 dprintf("imsm: reshape update preparation : OK\n");
11269 *updatep = u;
11270
11271 return update_memory_size;
11272}
11273
48c5303a
PC		 11274/******************************************************************************
11275 * function: imsm_create_metadata_update_for_migration()
11276 * Creates update for IMSM array.
11277 *
11278 ******************************************************************************/
11279static int imsm_create_metadata_update_for_migration(
11280 struct supertype *st,
11281 struct geo_params *geo,
11282 struct imsm_update_reshape_migration **updatep)
11283{
11284 struct intel_super *super = st->sb;
594dc1b8
JS		 11285 int update_memory_size;
11286 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11287 struct imsm_dev *dev;
11288 int previous_level = -1;
11289
1ade5cc1 11290 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11291
11292 /* size of all update data without anchor */
11293 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11294
503975b9 11295 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11296 u->type = update_reshape_migration;
11297 u->subdev = super->current_vol;
11298 u->new_level = geo->level;
11299 u->new_layout = geo->layout;
11300 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11301 u->new_disks[0] = -1;
4bba0439 11302 u->new_chunksize = -1;
48c5303a
PC		 11303
11304 dev = get_imsm_dev(super, u->subdev);
11305 if (dev) {
11306 struct imsm_map *map;
11307
238c0a71 11308 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11309 if (map) {
11310 int current_chunk_size =
11311 __le16_to_cpu(map->blocks_per_strip) / 2;
11312
11313 if (geo->chunksize != current_chunk_size) {
11314 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11315 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11316 current_chunk_size, u->new_chunksize);
11317 }
48c5303a 11318 previous_level = map->raid_level;
4bba0439 11319 }
48c5303a 11320 }
089f9d79 11321 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11322 struct mdinfo *spares = NULL;
11323
11324 u->new_raid_disks++;
11325 spares = get_spares_for_grow(st);
089f9d79 11326 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11327 free(u);
11328 sysfs_free(spares);
11329 update_memory_size = 0;
565cc99e 11330 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11331 return 0;
11332 }
11333 sysfs_free(spares);
11334 }
11335 dprintf("imsm: reshape update preparation : OK\n");
11336 *updatep = u;
11337
11338 return update_memory_size;
11339}
11340
8dd70bce
AK		 11341static void imsm_update_metadata_locally(struct supertype *st,
11342 void *buf, int len)
11343{
11344 struct metadata_update mu;
11345
11346 mu.buf = buf;
11347 mu.len = len;
11348 mu.space = NULL;
11349 mu.space_list = NULL;
11350 mu.next = NULL;
5fe6f031
N		 11351 if (imsm_prepare_update(st, &mu))
11352 imsm_process_update(st, &mu);
8dd70bce
AK		 11353
11354 while (mu.space_list) {
11355 void **space = mu.space_list;
11356 mu.space_list = *space;
11357 free(space);
11358 }
11359}
78b10e66 11360
471bceb6 11361/***************************************************************************
694575e7 11362* Function: imsm_analyze_change
471bceb6 11363* Description: Function analyze change for single volume
1011e834 11364* and validate if transition is supported
fbf3d202
AK		 11365* Parameters: Geometry parameters, supertype structure,
11366* metadata change direction (apply/rollback)
694575e7 11367* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11368****************************************************************************/
11369enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11370 struct geo_params *geo,
11371 int direction)
694575e7 11372{
471bceb6
KW		 11373 struct mdinfo info;
11374 int change = -1;
11375 int check_devs = 0;
c21e737b 11376 int chunk;
67a2db32
AK		 11377 /* number of added/removed disks in operation result */
11378 int devNumChange = 0;
11379 /* imsm compatible layout value for array geometry verification */
11380 int imsm_layout = -1;
7abc9871
AK		 11381 int data_disks;
11382 struct imsm_dev *dev;
9529d343 11383 struct imsm_map *map;
7abc9871 11384 struct intel_super *super;
d04f65f4 11385 unsigned long long current_size;
65d38cca 11386 unsigned long long free_size;
d04f65f4 11387 unsigned long long max_size;
65d38cca 11388 int rv;
471bceb6
KW		 11389
11390 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11391 if (geo->level != info.array.level && geo->level >= 0 &&
11392 geo->level != UnSet) {
471bceb6
KW		 11393 switch (info.array.level) {
11394 case 0:
11395 if (geo->level == 5) {
b5347799 11396 change = CH_MIGRATION;
e13ce846 11397 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11398 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11399 change = -1;
11400 goto analyse_change_exit;
11401 }
67a2db32 11402 imsm_layout = geo->layout;
471bceb6 11403 check_devs = 1;
e91a3bad
LM		 11404 devNumChange = 1; /* parity disk added */
11405 } else if (geo->level == 10) {
471bceb6
KW		 11406 change = CH_TAKEOVER;
11407 check_devs = 1;
e91a3bad 11408 devNumChange = 2; /* two mirrors added */
67a2db32 11409 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11410 }
dfe77a9e
KW		 11411 break;
11412 case 1:
471bceb6
KW		 11413 case 10:
11414 if (geo->level == 0) {
11415 change = CH_TAKEOVER;
11416 check_devs = 1;
e91a3bad 11417 devNumChange = -(geo->raid_disks/2);
67a2db32 11418 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11419 }
11420 break;
11421 }
11422 if (change == -1) {
7a862a02 11423 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11424 info.array.level, geo->level);
471bceb6
KW		 11425 goto analyse_change_exit;
11426 }
11427 } else
11428 geo->level = info.array.level;
11429
089f9d79
JS		 11430 if (geo->layout != info.array.layout &&
11431 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11432 change = CH_MIGRATION;
089f9d79
JS		 11433 if (info.array.layout == 0 && info.array.level == 5 &&
11434 geo->layout == 5) {
471bceb6 11435 /* reshape 5 -> 4 */
089f9d79
JS		 11436 } else if (info.array.layout == 5 && info.array.level == 5 &&
11437 geo->layout == 0) {
471bceb6
KW		 11438 /* reshape 4 -> 5 */
11439 geo->layout = 0;
11440 geo->level = 5;
11441 } else {
7a862a02 11442 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11443 info.array.layout, geo->layout);
471bceb6
KW		 11444 change = -1;
11445 goto analyse_change_exit;
11446 }
67a2db32 11447 } else {
471bceb6 11448 geo->layout = info.array.layout;
67a2db32
AK		 11449 if (imsm_layout == -1)
11450 imsm_layout = info.array.layout;
11451 }
471bceb6 11452
089f9d79
JS		 11453 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11454 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11455 if (info.array.level == 10) {
11456 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11457 change = -1;
11458 goto analyse_change_exit;
1e9b2c3f
PB		 11459 } else if (info.component_size % (geo->chunksize/512)) {
11460 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11461 geo->chunksize/1024, info.component_size/2);
11462 change = -1;
11463 goto analyse_change_exit;
2d2b0eb7 11464 }
b5347799 11465 change = CH_MIGRATION;
2d2b0eb7 11466 } else {
471bceb6 11467 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11468 }
471bceb6 11469
c21e737b 11470 chunk = geo->chunksize / 1024;
7abc9871
AK		 11471
11472 super = st->sb;
11473 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11474 map = get_imsm_map(dev, MAP_0);
11475 data_disks = imsm_num_data_members(map);
c41e00b2 11476 /* compute current size per disk member
7abc9871 11477 */
c41e00b2
AK		 11478 current_size = info.custom_array_size / data_disks;
11479
089f9d79 11480 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11481 /* align component size
11482 */
3e684231 11483 geo->size = imsm_component_size_alignment_check(
c41e00b2 11484 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11485 chunk * 1024, super->sector_size,
c41e00b2 11486 geo->size * 2);
65d0b4ce 11487 if (geo->size == 0) {
7a862a02 11488 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11489 current_size);
11490 goto analyse_change_exit;
11491 }
c41e00b2 11492 }
7abc9871 11493
089f9d79 11494 if (current_size != geo->size && geo->size > 0) {
7abc9871 11495 if (change != -1) {
7a862a02 11496 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11497 change = -1;
11498 goto analyse_change_exit;
11499 }
11500 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11501 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11502 super->current_vol, st->devnm);
7abc9871
AK		 11503 goto analyse_change_exit;
11504 }
65d38cca
LD		 11505 /* check the maximum available size
11506 */
11507 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11508 0, chunk, &free_size);
11509 if (rv == 0)
11510 /* Cannot find maximum available space
11511 */
11512 max_size = 0;
11513 else {
11514 max_size = free_size + current_size;
11515 /* align component size
11516 */
3e684231 11517 max_size = imsm_component_size_alignment_check(
65d38cca 11518 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11519 chunk * 1024, super->sector_size,
65d38cca
LD		 11520 max_size);
11521 }
d04f65f4 11522 if (geo->size == MAX_SIZE) {
b130333f
AK		 11523 /* requested size change to the maximum available size
11524 */
65d38cca 11525 if (max_size == 0) {
7a862a02 11526 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11527 change = -1;
11528 goto analyse_change_exit;
65d38cca
LD		 11529 } else
11530 geo->size = max_size;
c41e00b2 11531 }
b130333f 11532
681b7ae2 11533 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11534 /* accept size for rollback only
11535 */
11536 } else {
11537 /* round size due to metadata compatibility
11538 */
11539 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11540 << SECT_PER_MB_SHIFT;
11541 dprintf("Prepare update for size change to %llu\n",
11542 geo->size );
11543 if (current_size >= geo->size) {
7a862a02 11544 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11545 current_size, geo->size);
fbf3d202
AK		 11546 goto analyse_change_exit;
11547 }
65d38cca 11548 if (max_size && geo->size > max_size) {
7a862a02 11549 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11550 max_size, geo->size);
65d38cca
LD		 11551 goto analyse_change_exit;
11552 }
7abc9871
AK		 11553 }
11554 geo->size *= data_disks;
11555 geo->raid_disks = dev->vol.map->num_members;
11556 change = CH_ARRAY_SIZE;
11557 }
471bceb6
KW		 11558 if (!validate_geometry_imsm(st,
11559 geo->level,
67a2db32 11560 imsm_layout,
e91a3bad 11561 geo->raid_disks + devNumChange,
c21e737b 11562 &chunk,
af4348dd 11563 geo->size, INVALID_SECTORS,
5308f117 11564 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11565 change = -1;
11566
11567 if (check_devs) {
11568 struct intel_super *super = st->sb;
11569 struct imsm_super *mpb = super->anchor;
11570
11571 if (mpb->num_raid_devs > 1) {
7a862a02 11572 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11573 geo->dev_name);
471bceb6
KW		 11574 change = -1;
11575 }
11576 }
11577
11578analyse_change_exit:
089f9d79
JS		 11579 if (direction == ROLLBACK_METADATA_CHANGES &&
11580 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11581 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11582 change = -1;
11583 }
471bceb6 11584 return change;
694575e7
KW		 11585}
11586
bb025c2f
KW		 11587int imsm_takeover(struct supertype *st, struct geo_params *geo)
11588{
11589 struct intel_super *super = st->sb;
11590 struct imsm_update_takeover *u;
11591
503975b9 11592 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11593
11594 u->type = update_takeover;
11595 u->subarray = super->current_vol;
11596
11597 /* 10->0 transition */
11598 if (geo->level == 0)
11599 u->direction = R10_TO_R0;
11600
0529c688
KW		 11601 /* 0->10 transition */
11602 if (geo->level == 10)
11603 u->direction = R0_TO_R10;
11604
bb025c2f
KW		 11605 /* update metadata locally */
11606 imsm_update_metadata_locally(st, u,
11607 sizeof(struct imsm_update_takeover));
11608 /* and possibly remotely */
11609 if (st->update_tail)
11610 append_metadata_update(st, u,
11611 sizeof(struct imsm_update_takeover));
11612 else
11613 free(u);
11614
11615 return 0;
11616}
11617
d04f65f4
N		 11618static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11619 int level,
78b10e66 11620 int layout, int chunksize, int raid_disks,
41784c88 11621 int delta_disks, char *backup, char *dev,
016e00f5 11622 int direction, int verbose)
78b10e66 11623{
78b10e66
N		 11624 int ret_val = 1;
11625 struct geo_params geo;
11626
1ade5cc1 11627 dprintf("(enter)\n");
78b10e66 11628
71204a50 11629 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11630
11631 geo.dev_name = dev;
4dd2df09 11632 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11633 geo.size = size;
11634 geo.level = level;
11635 geo.layout = layout;
11636 geo.chunksize = chunksize;
11637 geo.raid_disks = raid_disks;
41784c88
AK		 11638 if (delta_disks != UnSet)
11639 geo.raid_disks += delta_disks;
78b10e66 11640
1ade5cc1
N		 11641 dprintf("for level : %i\n", geo.level);
11642 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11643
4dd2df09 11644 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11645 /* On container level we can only increase number of devices. */
11646 dprintf("imsm: info: Container operation\n");
78b10e66 11647 int old_raid_disks = 0;
6dc0be30 11648
78b10e66 11649 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11650 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11651 struct imsm_update_reshape *u = NULL;
11652 int len;
11653
11654 len = imsm_create_metadata_update_for_reshape(
11655 st, &geo, old_raid_disks, &u);
11656
ed08d51c
AK		 11657 if (len <= 0) {
11658 dprintf("imsm: Cannot prepare update\n");
11659 goto exit_imsm_reshape_super;
11660 }
11661
8dd70bce
AK		 11662 ret_val = 0;
11663 /* update metadata locally */
11664 imsm_update_metadata_locally(st, u, len);
11665 /* and possibly remotely */
11666 if (st->update_tail)
11667 append_metadata_update(st, u, len);
11668 else
ed08d51c 11669 free(u);
8dd70bce 11670
694575e7 11671 } else {
7a862a02 11672 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11673 }
11674 } else {
11675 /* On volume level we support following operations
471bceb6
KW		 11676 * - takeover: raid10 -> raid0; raid0 -> raid10
11677 * - chunk size migration
11678 * - migration: raid5 -> raid0; raid0 -> raid5
11679 */
11680 struct intel_super *super = st->sb;
11681 struct intel_dev *dev = super->devlist;
4dd2df09 11682 int change;
694575e7 11683 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11684 /* find requested device */
11685 while (dev) {
1011e834 11686 char *devnm =
4dd2df09
N		 11687 imsm_find_array_devnm_by_subdev(
11688 dev->index, st->container_devnm);
11689 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11690 break;
11691 dev = dev->next;
11692 }
11693 if (dev == NULL) {
4dd2df09
N		 11694 pr_err("Cannot find %s (%s) subarray\n",
11695 geo.dev_name, geo.devnm);
471bceb6
KW		 11696 goto exit_imsm_reshape_super;
11697 }
11698 super->current_vol = dev->index;
fbf3d202 11699 change = imsm_analyze_change(st, &geo, direction);
694575e7 11700 switch (change) {
471bceb6 11701 case CH_TAKEOVER:
bb025c2f 11702 ret_val = imsm_takeover(st, &geo);
694575e7 11703 break;
48c5303a
PC		 11704 case CH_MIGRATION: {
11705 struct imsm_update_reshape_migration *u = NULL;
11706 int len =
11707 imsm_create_metadata_update_for_migration(
11708 st, &geo, &u);
11709 if (len < 1) {
7a862a02 11710 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11711 break;
11712 }
471bceb6 11713 ret_val = 0;
48c5303a
PC		 11714 /* update metadata locally */
11715 imsm_update_metadata_locally(st, u, len);
11716 /* and possibly remotely */
11717 if (st->update_tail)
11718 append_metadata_update(st, u, len);
11719 else
11720 free(u);
11721 }
11722 break;
7abc9871 11723 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11724 struct imsm_update_size_change *u = NULL;
11725 int len =
11726 imsm_create_metadata_update_for_size_change(
11727 st, &geo, &u);
11728 if (len < 1) {
7a862a02 11729 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11730 break;
11731 }
11732 ret_val = 0;
11733 /* update metadata locally */
11734 imsm_update_metadata_locally(st, u, len);
11735 /* and possibly remotely */
11736 if (st->update_tail)
11737 append_metadata_update(st, u, len);
11738 else
11739 free(u);
7abc9871
AK		 11740 }
11741 break;
471bceb6
KW		 11742 default:
11743 ret_val = 1;
694575e7 11744 }
694575e7 11745 }
78b10e66 11746
ed08d51c 11747exit_imsm_reshape_super:
78b10e66
N		 11748 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11749 return ret_val;
11750}
2cda7640 11751
0febb20c
AO		 11752#define COMPLETED_OK 0
11753#define COMPLETED_NONE 1
11754#define COMPLETED_DELAYED 2
11755
11756static int read_completed(int fd, unsigned long long *val)
11757{
11758 int ret;
11759 char buf[50];
11760
11761 ret = sysfs_fd_get_str(fd, buf, 50);
11762 if (ret < 0)
11763 return ret;
11764
11765 ret = COMPLETED_OK;
11766 if (strncmp(buf, "none", 4) == 0) {
11767 ret = COMPLETED_NONE;
11768 } else if (strncmp(buf, "delayed", 7) == 0) {
11769 ret = COMPLETED_DELAYED;
11770 } else {
11771 char *ep;
11772 *val = strtoull(buf, &ep, 0);
11773 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11774 ret = -1;
11775 }
11776 return ret;
11777}
11778
eee67a47
AK		 11779/*******************************************************************************
11780 * Function: wait_for_reshape_imsm
11781 * Description: Function writes new sync_max value and waits until
11782 * reshape process reach new position
11783 * Parameters:
11784 * sra : general array info
eee67a47
AK		 11785 * ndata : number of disks in new array's layout
11786 * Returns:
11787 * 0 : success,
11788 * 1 : there is no reshape in progress,
11789 * -1 : fail
11790 ******************************************************************************/
ae9f01f8 11791int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11792{
85ca499c 11793 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11794 int retry = 3;
eee67a47 11795 unsigned long long completed;
ae9f01f8
AK		 11796 /* to_complete : new sync_max position */
11797 unsigned long long to_complete = sra->reshape_progress;
11798 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11799
ae9f01f8 11800 if (fd < 0) {
1ade5cc1 11801 dprintf("cannot open reshape_position\n");
eee67a47 11802 return 1;
ae9f01f8 11803 }
eee67a47 11804
df2647fa
PB		 11805 do {
11806 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11807 if (!retry) {
11808 dprintf("cannot read reshape_position (no reshape in progres)\n");
11809 close(fd);
11810 return 1;
11811 }
11812 usleep(30000);
11813 } else
11814 break;
11815 } while (retry--);
eee67a47 11816
85ca499c 11817 if (completed > position_to_set) {
1ade5cc1 11818 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11819 to_complete, position_to_set);
ae9f01f8
AK		 11820 close(fd);
11821 return -1;
11822 }
11823 dprintf("Position set: %llu\n", position_to_set);
11824 if (sysfs_set_num(sra, NULL, "sync_max",
11825 position_to_set) != 0) {
1ade5cc1 11826 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11827 position_to_set);
11828 close(fd);
11829 return -1;
eee67a47
AK		 11830 }
11831
eee67a47 11832 do {
0febb20c 11833 int rc;
eee67a47 11834 char action[20];
5ff3a780 11835 int timeout = 3000;
0febb20c 11836
5ff3a780 11837 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11838 if (sysfs_get_str(sra, NULL, "sync_action",
11839 action, 20) > 0 &&
d7d3809a 11840 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11841 if (strncmp(action, "idle", 4) == 0)
11842 break;
d7d3809a
AP		 11843 close(fd);
11844 return -1;
11845 }
0febb20c
AO		 11846
11847 rc = read_completed(fd, &completed);
11848 if (rc < 0) {
1ade5cc1 11849 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11850 close(fd);
11851 return 1;
0febb20c
AO		 11852 } else if (rc == COMPLETED_NONE)
11853 break;
85ca499c 11854 } while (completed < position_to_set);
b2be2b62 11855
eee67a47
AK		 11856 close(fd);
11857 return 0;
eee67a47
AK		 11858}
11859
b915c95f
AK		 11860/*******************************************************************************
11861 * Function: check_degradation_change
11862 * Description: Check that array hasn't become failed.
11863 * Parameters:
11864 * info : for sysfs access
11865 * sources : source disks descriptors
11866 * degraded: previous degradation level
11867 * Returns:
11868 * degradation level
11869 ******************************************************************************/
11870int check_degradation_change(struct mdinfo *info,
11871 int *sources,
11872 int degraded)
11873{
11874 unsigned long long new_degraded;
e1993023
LD		 11875 int rv;
11876
11877 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11878 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11879 /* check each device to ensure it is still working */
11880 struct mdinfo *sd;
11881 new_degraded = 0;
11882 for (sd = info->devs ; sd ; sd = sd->next) {
11883 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11884 continue;
11885 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11886 char sbuf[100];
11887
b915c95f 11888 if (sysfs_get_str(info,
cf52eff5 11889 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11890 strstr(sbuf, "faulty") ||
11891 strstr(sbuf, "in_sync") == NULL) {
11892 /* this device is dead */
11893 sd->disk.state = (1<<MD_DISK_FAULTY);
11894 if (sd->disk.raid_disk >= 0 &&
11895 sources[sd->disk.raid_disk] >= 0) {
11896 close(sources[
11897 sd->disk.raid_disk]);
11898 sources[sd->disk.raid_disk] =
11899 -1;
11900 }
11901 new_degraded++;
11902 }
11903 }
11904 }
11905 }
11906
11907 return new_degraded;
11908}
11909
10f22854
AK		 11910/*******************************************************************************
11911 * Function: imsm_manage_reshape
11912 * Description: Function finds array under reshape and it manages reshape
11913 * process. It creates stripes backups (if required) and sets
942e1cdb 11914 * checkpoints.
10f22854
AK		 11915 * Parameters:
11916 * afd : Backup handle (nattive) - not used
11917 * sra : general array info
11918 * reshape : reshape parameters - not used
11919 * st : supertype structure
11920 * blocks : size of critical section [blocks]
11921 * fds : table of source device descriptor
11922 * offsets : start of array (offest per devices)
11923 * dests : not used
11924 * destfd : table of destination device descriptor
11925 * destoffsets : table of destination offsets (per device)
11926 * Returns:
11927 * 1 : success, reshape is done
11928 * 0 : fail
11929 ******************************************************************************/
999b4972
N		 11930static int imsm_manage_reshape(
11931 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11932 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11933 int *fds, unsigned long long *offsets,
11934 int dests, int *destfd, unsigned long long *destoffsets)
11935{
10f22854
AK		 11936 int ret_val = 0;
11937 struct intel_super *super = st->sb;
594dc1b8 11938 struct intel_dev *dv;
de44e46f 11939 unsigned int sector_size = super->sector_size;
10f22854 11940 struct imsm_dev *dev = NULL;
9529d343 11941 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11942 int migr_vol_qan = 0;
11943 int ndata, odata; /* [bytes] */
11944 int chunk; /* [bytes] */
11945 struct migr_record *migr_rec;
11946 char *buf = NULL;
11947 unsigned int buf_size; /* [bytes] */
11948 unsigned long long max_position; /* array size [bytes] */
11949 unsigned long long next_step; /* [blocks]/[bytes] */
11950 unsigned long long old_data_stripe_length;
10f22854
AK		 11951 unsigned long long start_src; /* [bytes] */
11952 unsigned long long start; /* [bytes] */
11953 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11954 int degraded = 0;
ab724b98 11955 int source_layout = 0;
10f22854 11956
79a16a9b
JS		 11957 if (!sra)
11958 return ret_val;
11959
11960 if (!fds || !offsets)
10f22854
AK		 11961 goto abort;
11962
11963 /* Find volume during the reshape */
11964 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 11965 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
11966 dv->dev->vol.migr_state == 1) {
10f22854
AK		 11967 dev = dv->dev;
11968 migr_vol_qan++;
11969 }
11970 }
11971 /* Only one volume can migrate at the same time */
11972 if (migr_vol_qan != 1) {
676e87a8 11973 pr_err("%s", migr_vol_qan ?
10f22854
AK		 11974 "Number of migrating volumes greater than 1\n" :
11975 "There is no volume during migrationg\n");
11976 goto abort;
11977 }
11978
9529d343 11979 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 11980 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 11981 if (map_src == NULL)
11982 goto abort;
10f22854 11983
9529d343
MD		 11984 ndata = imsm_num_data_members(map_dest);
11985 odata = imsm_num_data_members(map_src);
10f22854 11986
7b1ab482 11987 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 11988 old_data_stripe_length = odata * chunk;
11989
11990 migr_rec = super->migr_rec;
11991
10f22854
AK		 11992 /* initialize migration record for start condition */
11993 if (sra->reshape_progress == 0)
11994 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 11995 else {
11996 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 11997 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 11998 goto abort;
11999 }
6a75c8ca
AK		 12000 /* Save checkpoint to update migration record for current
12001 * reshape position (in md). It can be farther than current
12002 * reshape position in metadata.
12003 */
12004 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12005 /* ignore error == 2, this can mean end of reshape here
12006 */
7a862a02 12007 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12008 goto abort;
12009 }
b2c59438 12010 }
10f22854
AK		 12011
12012 /* size for data */
12013 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12014 /* extend buffer size for parity disk */
12015 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12016 /* add space for stripe alignment */
10f22854 12017 buf_size += old_data_stripe_length;
de44e46f
PB		 12018 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12019 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12020 goto abort;
12021 }
12022
3ef4403c 12023 max_position = sra->component_size * ndata;
68eb8bc6 12024 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK		 12025
12026 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
12027 __le32_to_cpu(migr_rec->num_migr_units)) {
12028 /* current reshape position [blocks] */
12029 unsigned long long current_position =
12030 __le32_to_cpu(migr_rec->blocks_per_unit)
12031 * __le32_to_cpu(migr_rec->curr_migr_unit);
12032 unsigned long long border;
12033
b915c95f
AK		 12034 /* Check that array hasn't become failed.
12035 */
12036 degraded = check_degradation_change(sra, fds, degraded);
12037 if (degraded > 1) {
7a862a02 12038 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12039 goto abort;
12040 }
12041
10f22854
AK		 12042 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12043
12044 if ((current_position + next_step) > max_position)
12045 next_step = max_position - current_position;
12046
92144abf 12047 start = current_position * 512;
10f22854 12048
942e1cdb 12049 /* align reading start to old geometry */
10f22854
AK		 12050 start_buf_shift = start % old_data_stripe_length;
12051 start_src = start - start_buf_shift;
12052
12053 border = (start_src / odata) - (start / ndata);
12054 border /= 512;
12055 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12056 /* save critical stripes to buf
12057 * start - start address of current unit
12058 * to backup [bytes]
12059 * start_src - start address of current unit
12060 * to backup alligned to source array
12061 * [bytes]
12062 */
594dc1b8 12063 unsigned long long next_step_filler;
10f22854
AK		 12064 unsigned long long copy_length = next_step * 512;
12065
12066 /* allign copy area length to stripe in old geometry */
12067 next_step_filler = ((copy_length + start_buf_shift)
12068 % old_data_stripe_length);
12069 if (next_step_filler)
12070 next_step_filler = (old_data_stripe_length
12071 - next_step_filler);
7a862a02 12072 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		 12073 start, start_src, copy_length,
12074 start_buf_shift, next_step_filler);
12075
12076 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12077 chunk, map_src->raid_level,
12078 source_layout, 0, NULL, start_src,
10f22854
AK		 12079 copy_length +
12080 next_step_filler + start_buf_shift,
12081 buf)) {
7a862a02 12082 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12083 goto abort;
12084 }
12085 /* Convert data to destination format and store it
12086 * in backup general migration area
12087 */
12088 if (save_backup_imsm(st, dev, sra,
aea93171 12089 buf + start_buf_shift, copy_length)) {
7a862a02 12090 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12091 goto abort;
12092 }
12093 if (save_checkpoint_imsm(st, sra,
12094 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12095 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12096 goto abort;
12097 }
8016a6d4
AK		 12098 } else {
12099 /* set next step to use whole border area */
12100 border /= next_step;
12101 if (border > 1)
12102 next_step *= border;
10f22854
AK		 12103 }
12104 /* When data backed up, checkpoint stored,
12105 * kick the kernel to reshape unit of data
12106 */
12107 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12108 /* limit next step to array max position */
12109 if (next_step > max_position)
12110 next_step = max_position;
10f22854
AK		 12111 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12112 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12113 sra->reshape_progress = next_step;
10f22854
AK		 12114
12115 /* wait until reshape finish */
c85338c6 12116 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12117 dprintf("wait_for_reshape_imsm returned error!\n");
12118 goto abort;
12119 }
84d11e6c
N		 12120 if (sigterm)
12121 goto abort;
10f22854 12122
0228d92c
AK		 12123 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12124 /* ignore error == 2, this can mean end of reshape here
12125 */
7a862a02 12126 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12127 goto abort;
12128 }
12129
12130 }
12131
71e5411e
PB		 12132 /* clear migr_rec on disks after successful migration */
12133 struct dl *d;
12134
85337573 12135 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12136 for (d = super->disks; d; d = d->next) {
12137 if (d->index < 0 || is_failed(&d->disk))
12138 continue;
12139 unsigned long long dsize;
12140
12141 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12142 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12143 SEEK_SET) >= 0) {
466070ad 12144 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12145 MIGR_REC_BUF_SECTORS*sector_size) !=
12146 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12147 perror("Write migr_rec failed");
12148 }
12149 }
12150
10f22854
AK		 12151 /* return '1' if done */
12152 ret_val = 1;
12153abort:
12154 free(buf);
942e1cdb
N		 12155 /* See Grow.c: abort_reshape() for further explanation */
12156 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12157 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12158 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12159
12160 return ret_val;
999b4972 12161}
0c21b485 12162
cdddbdbc 12163struct superswitch super_imsm = {
cdddbdbc
DW		 12164 .examine_super = examine_super_imsm,
12165 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12166 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12167 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12168 .detail_super = detail_super_imsm,
12169 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12170 .write_init_super = write_init_super_imsm,
0e600426
N		 12171 .validate_geometry = validate_geometry_imsm,
12172 .add_to_super = add_to_super_imsm,
1a64be56 12173 .remove_from_super = remove_from_super_imsm,
d665cc31 12174 .detail_platform = detail_platform_imsm,
e50cf220 12175 .export_detail_platform = export_detail_platform_imsm,
33414a01 12176 .kill_subarray = kill_subarray_imsm,
aa534678 12177 .update_subarray = update_subarray_imsm,
2b959fbf 12178 .load_container = load_container_imsm,
71204a50
N		 12179 .default_geometry = default_geometry_imsm,
12180 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12181 .reshape_super = imsm_reshape_super,
12182 .manage_reshape = imsm_manage_reshape,
9e2d750d 12183 .recover_backup = recover_backup_imsm,
74db60b0 12184 .copy_metadata = copy_metadata_imsm,
27156a57 12185 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12186 .match_home = match_home_imsm,
12187 .uuid_from_super= uuid_from_super_imsm,
12188 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12189 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12190 .update_super = update_super_imsm,
12191
12192 .avail_size = avail_size_imsm,
fbfdcb06 12193 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12194
12195 .compare_super = compare_super_imsm,
12196
12197 .load_super = load_super_imsm,
bf5a934a 12198 .init_super = init_super_imsm,
e683ca88 12199 .store_super = store_super_imsm,
cdddbdbc
DW		 12200 .free_super = free_super_imsm,
12201 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12202 .container_content = container_content_imsm,
0c21b485 12203 .validate_container = validate_container_imsm,
cdddbdbc 12204
2432ce9b
AP		 12205 .write_init_ppl = write_init_ppl_imsm,
12206 .validate_ppl = validate_ppl_imsm,
12207
cdddbdbc 12208 .external = 1,
4cce4069 12209 .name = "imsm",
845dea95
NB		 12210
12211/* for mdmon */
12212 .open_new = imsm_open_new,
ed9d66aa 12213 .set_array_state= imsm_set_array_state,
845dea95
NB		 12214 .set_disk = imsm_set_disk,
12215 .sync_metadata = imsm_sync_metadata,
88758e9d 12216 .activate_spare = imsm_activate_spare,
e8319a19 12217 .process_update = imsm_process_update,
8273f55e 12218 .prepare_update = imsm_prepare_update,
6f50473f 12219 .record_bad_block = imsm_record_badblock,
c07a5a4f 12220 .clear_bad_block = imsm_clear_badblock,
928f1424 12221 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12222};
Unnamed repository; edit this file 'description' to name the repository.