]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
mdadm.conf.5: typo: of -> or
[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 | \
418f9b36
N
84 MPB_ATTRIB_EXP_STRIPE_SIZE)
85
86/* Define attributes that are unused but not harmful */
87#define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE)
fe7ed8cb 88
8e59f3d8 89#define MPB_SECTOR_CNT 2210
c2c087e6 90#define IMSM_RESERVED_SECTORS 4096
b81221b7 91#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056
979d38be 92#define SECT_PER_MB_SHIFT 11
cdddbdbc
DW
93
94/* Disk configuration info. */
95#define IMSM_MAX_DEVICES 255
96struct imsm_disk {
97 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 98 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 99 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW
100#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
101#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
102#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
cdddbdbc 103 __u32 status; /* 0xF0 - 0xF3 */
fe7ed8cb 104 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA
105 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
106#define IMSM_DISK_FILLERS 3
107 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc
DW
108};
109
3b451610
AK
110/* map selector for map managment
111 */
238c0a71
AK
112#define MAP_0 0
113#define MAP_1 1
114#define MAP_X -1
3b451610 115
cdddbdbc
DW
116/* RAID map configuration infos. */
117struct imsm_map {
5551b113
CA
118 __u32 pba_of_lba0_lo; /* start address of partition */
119 __u32 blocks_per_member_lo;/* blocks per member */
120 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW
121 __u16 blocks_per_strip;
122 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
123#define IMSM_T_STATE_NORMAL 0
124#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW
125#define IMSM_T_STATE_DEGRADED 2
126#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW
127 __u8 raid_level;
128#define IMSM_T_RAID0 0
129#define IMSM_T_RAID1 1
130#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
131 __u8 num_members; /* number of member disks */
fe7ed8cb
DW
132 __u8 num_domains; /* number of parity domains */
133 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 134 __u8 ddf;
5551b113
CA
135 __u32 pba_of_lba0_hi;
136 __u32 blocks_per_member_hi;
137 __u32 num_data_stripes_hi;
138 __u32 filler[4]; /* expansion area */
7eef0453 139#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 140 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW
141 * top byte contains some flags
142 */
cdddbdbc
DW
143} __attribute__ ((packed));
144
145struct imsm_vol {
f8f603f1 146 __u32 curr_migr_unit;
fe7ed8cb 147 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 148 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW
149#define MIGR_INIT 0
150#define MIGR_REBUILD 1
151#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
152#define MIGR_GEN_MIGR 3
153#define MIGR_STATE_CHANGE 4
1484e727 154#define MIGR_REPAIR 5
cdddbdbc
DW
155 __u8 migr_type; /* Initializing, Rebuilding, ... */
156 __u8 dirty;
fe7ed8cb
DW
157 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
158 __u16 verify_errors; /* number of mismatches */
159 __u16 bad_blocks; /* number of bad blocks during verify */
160 __u32 filler[4];
cdddbdbc
DW
161 struct imsm_map map[1];
162 /* here comes another one if migr_state */
163} __attribute__ ((packed));
164
165struct imsm_dev {
fe7ed8cb 166 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW
167 __u32 size_low;
168 __u32 size_high;
fe7ed8cb
DW
169#define DEV_BOOTABLE __cpu_to_le32(0x01)
170#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
171#define DEV_READ_COALESCING __cpu_to_le32(0x04)
172#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
173#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
174#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
175#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
176#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
177#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
178#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
179#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
180#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
181#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW
182 __u32 status; /* Persistent RaidDev status */
183 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW
184 __u8 migr_priority;
185 __u8 num_sub_vols;
186 __u8 tid;
187 __u8 cng_master_disk;
188 __u16 cache_policy;
189 __u8 cng_state;
190 __u8 cng_sub_state;
191#define IMSM_DEV_FILLERS 10
cdddbdbc
DW
192 __u32 filler[IMSM_DEV_FILLERS];
193 struct imsm_vol vol;
194} __attribute__ ((packed));
195
196struct imsm_super {
197 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
198 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
199 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
200 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
201 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD
202 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
203 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW
204 __u8 num_disks; /* 0x38 Number of configured disks */
205 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD
206 __u8 error_log_pos; /* 0x3A */
207 __u8 fill[1]; /* 0x3B */
208 __u32 cache_size; /* 0x3c - 0x40 in mb */
209 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
210 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
211 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
212#define IMSM_FILLERS 35
213 __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW
214 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
215 /* here comes imsm_dev[num_raid_devs] */
604b746f 216 /* here comes BBM logs */
cdddbdbc
DW
217} __attribute__ ((packed));
218
604b746f
JD
219#define BBM_LOG_MAX_ENTRIES 254
220
221struct bbm_log_entry {
222 __u64 defective_block_start;
223#define UNREADABLE 0xFFFFFFFF
224 __u32 spare_block_offset;
225 __u16 remapped_marked_count;
226 __u16 disk_ordinal;
227} __attribute__ ((__packed__));
228
229struct bbm_log {
230 __u32 signature; /* 0xABADB10C */
231 __u32 entry_count;
232 __u32 reserved_spare_block_count; /* 0 */
233 __u32 reserved; /* 0xFFFF */
234 __u64 first_spare_lba;
235 struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES];
236} __attribute__ ((__packed__));
237
238
cdddbdbc
DW
239#ifndef MDASSEMBLE
240static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
241#endif
242
8e59f3d8
AK
243#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
244
245#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
246
17a4eaf9
AK
247#define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */
248#define MIGR_REC_POSITION 512 /* migr_record position offset on disk,
249 * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION
250 */
251
252
8e59f3d8
AK
253#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
254 * be recovered using srcMap */
255#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
256 * already been migrated and must
257 * be recovered from checkpoint area */
258struct migr_record {
259 __u32 rec_status; /* Status used to determine how to restart
260 * migration in case it aborts
261 * in some fashion */
262 __u32 curr_migr_unit; /* 0..numMigrUnits-1 */
263 __u32 family_num; /* Family number of MPB
264 * containing the RaidDev
265 * that is migrating */
266 __u32 ascending_migr; /* True if migrating in increasing
267 * order of lbas */
268 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
269 __u32 dest_depth_per_unit; /* Num member blocks each destMap
270 * member disk
271 * advances per unit-of-operation */
272 __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */
273 __u32 dest_1st_member_lba; /* First member lba on first
274 * stripe of destination */
275 __u32 num_migr_units; /* Total num migration units-of-op */
276 __u32 post_migr_vol_cap; /* Size of volume after
277 * migration completes */
278 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
279 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
280 * migration ckpt record was read from
281 * (for recovered migrations) */
282} __attribute__ ((__packed__));
283
ec50f7b6
LM
284struct md_list {
285 /* usage marker:
286 * 1: load metadata
287 * 2: metadata does not match
288 * 4: already checked
289 */
290 int used;
291 char *devname;
292 int found;
293 int container;
294 dev_t st_rdev;
295 struct md_list *next;
296};
297
298#define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
299
1484e727
DW
300static __u8 migr_type(struct imsm_dev *dev)
301{
302 if (dev->vol.migr_type == MIGR_VERIFY &&
303 dev->status & DEV_VERIFY_AND_FIX)
304 return MIGR_REPAIR;
305 else
306 return dev->vol.migr_type;
307}
308
309static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
310{
311 /* for compatibility with older oroms convert MIGR_REPAIR, into
312 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
313 */
314 if (migr_type == MIGR_REPAIR) {
315 dev->vol.migr_type = MIGR_VERIFY;
316 dev->status |= DEV_VERIFY_AND_FIX;
317 } else {
318 dev->vol.migr_type = migr_type;
319 dev->status &= ~DEV_VERIFY_AND_FIX;
320 }
321}
322
87eb16df 323static unsigned int sector_count(__u32 bytes)
cdddbdbc 324{
654a3817 325 return ROUND_UP(bytes, 512) / 512;
87eb16df 326}
cdddbdbc 327
87eb16df
DW
328static unsigned int mpb_sectors(struct imsm_super *mpb)
329{
330 return sector_count(__le32_to_cpu(mpb->mpb_size));
cdddbdbc
DW
331}
332
ba2de7ba
DW
333struct intel_dev {
334 struct imsm_dev *dev;
335 struct intel_dev *next;
f21e18ca 336 unsigned index;
ba2de7ba
DW
337};
338
88654014
LM
339struct intel_hba {
340 enum sys_dev_type type;
341 char *path;
342 char *pci_id;
343 struct intel_hba *next;
344};
345
1a64be56
LM
346enum action {
347 DISK_REMOVE = 1,
348 DISK_ADD
349};
cdddbdbc
DW
350/* internal representation of IMSM metadata */
351struct intel_super {
352 union {
949c47a0
DW
353 void *buf; /* O_DIRECT buffer for reading/writing metadata */
354 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 355 };
8e59f3d8
AK
356 union {
357 void *migr_rec_buf; /* buffer for I/O operations */
358 struct migr_record *migr_rec; /* migration record */
359 };
51d83f5d
AK
360 int clean_migration_record_by_mdmon; /* when reshape is switched to next
361 array, it indicates that mdmon is allowed to clean migration
362 record */
949c47a0 363 size_t len; /* size of the 'buf' allocation */
4d7b1503
DW
364 void *next_buf; /* for realloc'ing buf from the manager */
365 size_t next_len;
c2c087e6 366 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 367 int current_vol; /* index of raid device undergoing creation */
5551b113 368 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 369 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 370 struct intel_dev *devlist;
cdddbdbc
DW
371 struct dl {
372 struct dl *next;
373 int index;
374 __u8 serial[MAX_RAID_SERIAL_LEN];
375 int major, minor;
376 char *devname;
b9f594fe 377 struct imsm_disk disk;
cdddbdbc 378 int fd;
0dcecb2e
DW
379 int extent_cnt;
380 struct extent *e; /* for determining freespace @ create */
efb30e7f 381 int raiddisk; /* slot to fill in autolayout */
1a64be56 382 enum action action;
ca0748fa 383 } *disks, *current_disk;
1a64be56
LM
384 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
385 active */
47ee5a45 386 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 387 struct bbm_log *bbm_log;
88654014 388 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 389 const struct imsm_orom *orom; /* platform firmware support */
a2b97981
DW
390 struct intel_super *next; /* (temp) list for disambiguating family_num */
391};
392
393struct intel_disk {
394 struct imsm_disk disk;
395 #define IMSM_UNKNOWN_OWNER (-1)
396 int owner;
397 struct intel_disk *next;
cdddbdbc
DW
398};
399
c2c087e6
DW
400struct extent {
401 unsigned long long start, size;
402};
403
694575e7
KW
404/* definitions of reshape process types */
405enum imsm_reshape_type {
406 CH_TAKEOVER,
b5347799 407 CH_MIGRATION,
7abc9871 408 CH_ARRAY_SIZE,
694575e7
KW
409};
410
88758e9d
DW
411/* definition of messages passed to imsm_process_update */
412enum imsm_update_type {
413 update_activate_spare,
8273f55e 414 update_create_array,
33414a01 415 update_kill_array,
aa534678 416 update_rename_array,
1a64be56 417 update_add_remove_disk,
78b10e66 418 update_reshape_container_disks,
48c5303a 419 update_reshape_migration,
2d40f3a1
AK
420 update_takeover,
421 update_general_migration_checkpoint,
f3871fdc 422 update_size_change,
88758e9d
DW
423};
424
425struct imsm_update_activate_spare {
426 enum imsm_update_type type;
d23fe947 427 struct dl *dl;
88758e9d
DW
428 int slot;
429 int array;
430 struct imsm_update_activate_spare *next;
431};
432
78b10e66
N
433struct geo_params {
434 int dev_id;
435 char *dev_name;
436 long long size;
437 int level;
438 int layout;
439 int chunksize;
440 int raid_disks;
441};
442
bb025c2f
KW
443enum takeover_direction {
444 R10_TO_R0,
445 R0_TO_R10
446};
447struct imsm_update_takeover {
448 enum imsm_update_type type;
449 int subarray;
450 enum takeover_direction direction;
451};
78b10e66
N
452
453struct imsm_update_reshape {
454 enum imsm_update_type type;
455 int old_raid_disks;
456 int new_raid_disks;
48c5303a
PC
457
458 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
459};
460
461struct imsm_update_reshape_migration {
462 enum imsm_update_type type;
463 int old_raid_disks;
464 int new_raid_disks;
465 /* fields for array migration changes
466 */
467 int subdev;
468 int new_level;
469 int new_layout;
4bba0439 470 int new_chunksize;
48c5303a 471
d195167d 472 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N
473};
474
f3871fdc
AK
475struct imsm_update_size_change {
476 enum imsm_update_type type;
477 int subdev;
478 long long new_size;
479};
480
2d40f3a1
AK
481struct imsm_update_general_migration_checkpoint {
482 enum imsm_update_type type;
483 __u32 curr_migr_unit;
484};
485
54c2c1ea
DW
486struct disk_info {
487 __u8 serial[MAX_RAID_SERIAL_LEN];
488};
489
8273f55e
DW
490struct imsm_update_create_array {
491 enum imsm_update_type type;
8273f55e 492 int dev_idx;
6a3e913e 493 struct imsm_dev dev;
8273f55e
DW
494};
495
33414a01
DW
496struct imsm_update_kill_array {
497 enum imsm_update_type type;
498 int dev_idx;
499};
500
aa534678
DW
501struct imsm_update_rename_array {
502 enum imsm_update_type type;
503 __u8 name[MAX_RAID_SERIAL_LEN];
504 int dev_idx;
505};
506
1a64be56 507struct imsm_update_add_remove_disk {
43dad3d6
DW
508 enum imsm_update_type type;
509};
510
88654014
LM
511
512static const char *_sys_dev_type[] = {
513 [SYS_DEV_UNKNOWN] = "Unknown",
514 [SYS_DEV_SAS] = "SAS",
515 [SYS_DEV_SATA] = "SATA"
516};
517
518const char *get_sys_dev_type(enum sys_dev_type type)
519{
520 if (type >= SYS_DEV_MAX)
521 type = SYS_DEV_UNKNOWN;
522
523 return _sys_dev_type[type];
524}
525
526static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
527{
528 struct intel_hba *result = malloc(sizeof(*result));
529 if (result) {
530 result->type = device->type;
531 result->path = strdup(device->path);
532 result->next = NULL;
533 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
534 result->pci_id++;
535 }
536 return result;
537}
538
539static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
540{
541 struct intel_hba *result=NULL;
542 for (result = hba; result; result = result->next) {
543 if (result->type == device->type && strcmp(result->path, device->path) == 0)
544 break;
545 }
546 return result;
547}
548
b4cf4cba 549static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM
550{
551 struct intel_hba *hba;
552
553 /* check if disk attached to Intel HBA */
554 hba = find_intel_hba(super->hba, device);
555 if (hba != NULL)
556 return 1;
557 /* Check if HBA is already attached to super */
558 if (super->hba == NULL) {
559 super->hba = alloc_intel_hba(device);
560 return 1;
561 }
562
563 hba = super->hba;
564 /* Intel metadata allows for all disks attached to the same type HBA.
565 * Do not sypport odf HBA types mixing
566 */
567 if (device->type != hba->type)
568 return 2;
569
570 while (hba->next)
571 hba = hba->next;
572
573 hba->next = alloc_intel_hba(device);
574 return 1;
575}
576
577static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
578{
579 struct sys_dev *list, *elem, *prev;
580 char *disk_path;
581
582 if ((list = find_intel_devices()) == NULL)
583 return 0;
584
585 if (fd < 0)
586 disk_path = (char *) devname;
587 else
588 disk_path = diskfd_to_devpath(fd);
589
590 if (!disk_path) {
591 free_sys_dev(&list);
592 return 0;
593 }
594
595 for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) {
596 if (path_attached_to_hba(disk_path, elem->path)) {
597 if (prev == NULL)
598 list = list->next;
599 else
600 prev->next = elem->next;
601 elem->next = NULL;
602 if (disk_path != devname)
603 free(disk_path);
604 free_sys_dev(&list);
605 return elem;
606 }
607 }
608 if (disk_path != devname)
609 free(disk_path);
610 free_sys_dev(&list);
611
612 return NULL;
613}
614
615
d424212e
N
616static int find_intel_hba_capability(int fd, struct intel_super *super,
617 char *devname);
f2f5c343 618
cdddbdbc
DW
619static struct supertype *match_metadata_desc_imsm(char *arg)
620{
621 struct supertype *st;
622
623 if (strcmp(arg, "imsm") != 0 &&
624 strcmp(arg, "default") != 0
625 )
626 return NULL;
627
628 st = malloc(sizeof(*st));
4e9d2186
AW
629 if (!st)
630 return NULL;
ef609477 631 memset(st, 0, sizeof(*st));
d1d599ea 632 st->container_dev = NoMdDev;
cdddbdbc
DW
633 st->ss = &super_imsm;
634 st->max_devs = IMSM_MAX_DEVICES;
635 st->minor_version = 0;
636 st->sb = NULL;
637 return st;
638}
639
0e600426 640#ifndef MDASSEMBLE
cdddbdbc
DW
641static __u8 *get_imsm_version(struct imsm_super *mpb)
642{
643 return &mpb->sig[MPB_SIG_LEN];
644}
9e2d750d 645#endif
cdddbdbc 646
949c47a0
DW
647/* retrieve a disk directly from the anchor when the anchor is known to be
648 * up-to-date, currently only at load time
649 */
650static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 651{
949c47a0 652 if (index >= mpb->num_disks)
cdddbdbc
DW
653 return NULL;
654 return &mpb->disk[index];
655}
656
95d07a2c
LM
657/* retrieve the disk description based on a index of the disk
658 * in the sub-array
659 */
660static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 661{
b9f594fe
DW
662 struct dl *d;
663
664 for (d = super->disks; d; d = d->next)
665 if (d->index == index)
95d07a2c
LM
666 return d;
667
668 return NULL;
669}
670/* retrieve a disk from the parsed metadata */
671static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
672{
673 struct dl *dl;
674
675 dl = get_imsm_dl_disk(super, index);
676 if (dl)
677 return &dl->disk;
678
b9f594fe 679 return NULL;
949c47a0
DW
680}
681
682/* generate a checksum directly from the anchor when the anchor is known to be
683 * up-to-date, currently only at load or write_super after coalescing
684 */
685static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW
686{
687 __u32 end = mpb->mpb_size / sizeof(end);
688 __u32 *p = (__u32 *) mpb;
689 __u32 sum = 0;
690
97f734fd
N
691 while (end--) {
692 sum += __le32_to_cpu(*p);
693 p++;
694 }
cdddbdbc
DW
695
696 return sum - __le32_to_cpu(mpb->check_sum);
697}
698
a965f303
DW
699static size_t sizeof_imsm_map(struct imsm_map *map)
700{
701 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
702}
703
704struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 705{
5e7b0330
AK
706 /* A device can have 2 maps if it is in the middle of a migration.
707 * If second_map is:
238c0a71
AK
708 * MAP_0 - we return the first map
709 * MAP_1 - we return the second map if it exists, else NULL
710 * MAP_X - we return the second map if it exists, else the first
5e7b0330 711 */
a965f303 712 struct imsm_map *map = &dev->vol.map[0];
9535fc47 713 struct imsm_map *map2 = NULL;
a965f303 714
9535fc47
AK
715 if (dev->vol.migr_state)
716 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 717
9535fc47 718 switch (second_map) {
3b451610 719 case MAP_0:
9535fc47 720 break;
3b451610 721 case MAP_1:
9535fc47
AK
722 map = map2;
723 break;
238c0a71 724 case MAP_X:
9535fc47
AK
725 if (map2)
726 map = map2;
727 break;
9535fc47
AK
728 default:
729 map = NULL;
730 }
731 return map;
5e7b0330 732
a965f303 733}
cdddbdbc 734
3393c6af
DW
735/* return the size of the device.
736 * migr_state increases the returned size if map[0] were to be duplicated
737 */
738static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW
739{
740 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 741 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW
742
743 /* migrating means an additional map */
a965f303 744 if (dev->vol.migr_state)
238c0a71 745 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 746 else if (migr_state)
238c0a71 747 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW
748
749 return size;
750}
751
54c2c1ea
DW
752#ifndef MDASSEMBLE
753/* retrieve disk serial number list from a metadata update */
754static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
755{
756 void *u = update;
757 struct disk_info *inf;
758
759 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
760 sizeof_imsm_dev(&update->dev, 0);
761
762 return inf;
763}
764#endif
765
949c47a0 766static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW
767{
768 int offset;
769 int i;
770 void *_mpb = mpb;
771
949c47a0 772 if (index >= mpb->num_raid_devs)
cdddbdbc
DW
773 return NULL;
774
775 /* devices start after all disks */
776 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
777
778 for (i = 0; i <= index; i++)
779 if (i == index)
780 return _mpb + offset;
781 else
3393c6af 782 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW
783
784 return NULL;
785}
786
949c47a0
DW
787static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
788{
ba2de7ba
DW
789 struct intel_dev *dv;
790
949c47a0
DW
791 if (index >= super->anchor->num_raid_devs)
792 return NULL;
ba2de7ba
DW
793 for (dv = super->devlist; dv; dv = dv->next)
794 if (dv->index == index)
795 return dv->dev;
796 return NULL;
949c47a0
DW
797}
798
98130f40
AK
799/*
800 * for second_map:
238c0a71
AK
801 * == MAP_0 get first map
802 * == MAP_1 get second map
803 * == MAP_X than get map according to the current migr_state
98130f40
AK
804 */
805static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
806 int slot,
807 int second_map)
7eef0453
DW
808{
809 struct imsm_map *map;
810
5e7b0330 811 map = get_imsm_map(dev, second_map);
7eef0453 812
ff077194
DW
813 /* top byte identifies disk under rebuild */
814 return __le32_to_cpu(map->disk_ord_tbl[slot]);
815}
816
817#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 818static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 819{
98130f40 820 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW
821
822 return ord_to_idx(ord);
7eef0453
DW
823}
824
be73972f
DW
825static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
826{
827 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
828}
829
f21e18ca 830static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW
831{
832 int slot;
833 __u32 ord;
834
835 for (slot = 0; slot < map->num_members; slot++) {
836 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
837 if (ord_to_idx(ord) == idx)
838 return slot;
839 }
840
841 return -1;
842}
843
cdddbdbc
DW
844static int get_imsm_raid_level(struct imsm_map *map)
845{
846 if (map->raid_level == 1) {
847 if (map->num_members == 2)
848 return 1;
849 else
850 return 10;
851 }
852
853 return map->raid_level;
854}
855
c2c087e6
DW
856static int cmp_extent(const void *av, const void *bv)
857{
858 const struct extent *a = av;
859 const struct extent *b = bv;
860 if (a->start < b->start)
861 return -1;
862 if (a->start > b->start)
863 return 1;
864 return 0;
865}
866
0dcecb2e 867static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 868{
c2c087e6 869 int memberships = 0;
620b1713 870 int i;
c2c087e6 871
949c47a0
DW
872 for (i = 0; i < super->anchor->num_raid_devs; i++) {
873 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 874 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 875
620b1713
DW
876 if (get_imsm_disk_slot(map, dl->index) >= 0)
877 memberships++;
c2c087e6 878 }
0dcecb2e
DW
879
880 return memberships;
881}
882
b81221b7
CA
883static __u32 imsm_min_reserved_sectors(struct intel_super *super);
884
5551b113
CA
885static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi)
886{
887 if (lo == 0 || hi == 0)
888 return 1;
889 *lo = __le32_to_cpu((unsigned)n);
890 *hi = __le32_to_cpu((unsigned)(n >> 32));
891 return 0;
892}
893
894static unsigned long long join_u32(__u32 lo, __u32 hi)
895{
896 return (unsigned long long)__le32_to_cpu(lo) |
897 (((unsigned long long)__le32_to_cpu(hi)) << 32);
898}
899
900static unsigned long long total_blocks(struct imsm_disk *disk)
901{
902 if (disk == NULL)
903 return 0;
904 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
905}
906
907static unsigned long long pba_of_lba0(struct imsm_map *map)
908{
909 if (map == NULL)
910 return 0;
911 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
912}
913
914static unsigned long long blocks_per_member(struct imsm_map *map)
915{
916 if (map == NULL)
917 return 0;
918 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
919}
920
ae2416e7 921#ifndef MDASSEMBLE
5551b113
CA
922static unsigned long long num_data_stripes(struct imsm_map *map)
923{
924 if (map == NULL)
925 return 0;
926 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
927}
928
929static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
930{
931 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
932}
ae2416e7 933#endif
5551b113
CA
934
935static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
936{
937 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
938}
939
940static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
941{
942 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
943}
944
945static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
946{
947 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
948}
949
0dcecb2e
DW
950static struct extent *get_extents(struct intel_super *super, struct dl *dl)
951{
952 /* find a list of used extents on the given physical device */
953 struct extent *rv, *e;
620b1713 954 int i;
0dcecb2e 955 int memberships = count_memberships(dl, super);
b276dd33
DW
956 __u32 reservation;
957
958 /* trim the reserved area for spares, so they can join any array
959 * regardless of whether the OROM has assigned sectors from the
960 * IMSM_RESERVED_SECTORS region
961 */
962 if (dl->index == -1)
b81221b7 963 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW
964 else
965 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 966
c2c087e6
DW
967 rv = malloc(sizeof(struct extent) * (memberships + 1));
968 if (!rv)
969 return NULL;
970 e = rv;
971
949c47a0
DW
972 for (i = 0; i < super->anchor->num_raid_devs; i++) {
973 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 974 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 975
620b1713 976 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113
CA
977 e->start = pba_of_lba0(map);
978 e->size = blocks_per_member(map);
620b1713 979 e++;
c2c087e6
DW
980 }
981 }
982 qsort(rv, memberships, sizeof(*rv), cmp_extent);
983
14e8215b
DW
984 /* determine the start of the metadata
985 * when no raid devices are defined use the default
986 * ...otherwise allow the metadata to truncate the value
987 * as is the case with older versions of imsm
988 */
989 if (memberships) {
990 struct extent *last = &rv[memberships - 1];
5551b113 991 unsigned long long remainder;
14e8215b 992
5551b113 993 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW
994 /* round down to 1k block to satisfy precision of the kernel
995 * 'size' interface
996 */
997 remainder &= ~1UL;
998 /* make sure remainder is still sane */
f21e18ca 999 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1000 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW
1001 if (reservation > remainder)
1002 reservation = remainder;
1003 }
5551b113 1004 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW
1005 e->size = 0;
1006 return rv;
1007}
1008
14e8215b
DW
1009/* try to determine how much space is reserved for metadata from
1010 * the last get_extents() entry, otherwise fallback to the
1011 * default
1012 */
1013static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1014{
1015 struct extent *e;
1016 int i;
1017 __u32 rv;
1018
1019 /* for spares just return a minimal reservation which will grow
1020 * once the spare is picked up by an array
1021 */
1022 if (dl->index == -1)
1023 return MPB_SECTOR_CNT;
1024
1025 e = get_extents(super, dl);
1026 if (!e)
1027 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1028
1029 /* scroll to last entry */
1030 for (i = 0; e[i].size; i++)
1031 continue;
1032
5551b113 1033 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW
1034
1035 free(e);
1036
1037 return rv;
1038}
1039
25ed7e59
DW
1040static int is_spare(struct imsm_disk *disk)
1041{
1042 return (disk->status & SPARE_DISK) == SPARE_DISK;
1043}
1044
1045static int is_configured(struct imsm_disk *disk)
1046{
1047 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1048}
1049
1050static int is_failed(struct imsm_disk *disk)
1051{
1052 return (disk->status & FAILED_DISK) == FAILED_DISK;
1053}
1054
b81221b7
CA
1055/* try to determine how much space is reserved for metadata from
1056 * the last get_extents() entry on the smallest active disk,
1057 * otherwise fallback to the default
1058 */
1059static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1060{
1061 struct extent *e;
1062 int i;
5551b113
CA
1063 unsigned long long min_active;
1064 __u32 remainder;
b81221b7
CA
1065 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1066 struct dl *dl, *dl_min = NULL;
1067
1068 if (!super)
1069 return rv;
1070
1071 min_active = 0;
1072 for (dl = super->disks; dl; dl = dl->next) {
1073 if (dl->index < 0)
1074 continue;
5551b113
CA
1075 unsigned long long blocks = total_blocks(&dl->disk);
1076 if (blocks < min_active || min_active == 0) {
b81221b7 1077 dl_min = dl;
5551b113 1078 min_active = blocks;
b81221b7
CA
1079 }
1080 }
1081 if (!dl_min)
1082 return rv;
1083
1084 /* find last lba used by subarrays on the smallest active disk */
1085 e = get_extents(super, dl_min);
1086 if (!e)
1087 return rv;
1088 for (i = 0; e[i].size; i++)
1089 continue;
1090
1091 remainder = min_active - e[i].start;
1092 free(e);
1093
1094 /* to give priority to recovery we should not require full
1095 IMSM_RESERVED_SECTORS from the spare */
1096 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1097
1098 /* if real reservation is smaller use that value */
1099 return (remainder < rv) ? remainder : rv;
1100}
1101
80e7f8c3
AC
1102/* Return minimum size of a spare that can be used in this array*/
1103static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st)
1104{
1105 struct intel_super *super = st->sb;
1106 struct dl *dl;
1107 struct extent *e;
1108 int i;
1109 unsigned long long rv = 0;
1110
1111 if (!super)
1112 return rv;
1113 /* find first active disk in array */
1114 dl = super->disks;
1115 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1116 dl = dl->next;
1117 if (!dl)
1118 return rv;
1119 /* find last lba used by subarrays */
1120 e = get_extents(super, dl);
1121 if (!e)
1122 return rv;
1123 for (i = 0; e[i].size; i++)
1124 continue;
1125 if (i > 0)
1126 rv = e[i-1].start + e[i-1].size;
1127 free(e);
b81221b7 1128
80e7f8c3 1129 /* add the amount of space needed for metadata */
b81221b7
CA
1130 rv = rv + imsm_min_reserved_sectors(super);
1131
80e7f8c3
AC
1132 return rv * 512;
1133}
1134
d1e02575
AK
1135static int is_gen_migration(struct imsm_dev *dev);
1136
1799c9e8 1137#ifndef MDASSEMBLE
c47b0ff6
AK
1138static __u64 blocks_per_migr_unit(struct intel_super *super,
1139 struct imsm_dev *dev);
1e5c6983 1140
c47b0ff6
AK
1141static void print_imsm_dev(struct intel_super *super,
1142 struct imsm_dev *dev,
1143 char *uuid,
1144 int disk_idx)
cdddbdbc
DW
1145{
1146 __u64 sz;
0d80bb2f 1147 int slot, i;
238c0a71
AK
1148 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1149 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1150 __u32 ord;
cdddbdbc
DW
1151
1152 printf("\n");
1e7bc0ed 1153 printf("[%.16s]:\n", dev->volume);
44470971 1154 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK
1155 printf(" RAID Level : %d", get_imsm_raid_level(map));
1156 if (map2)
1157 printf(" <-- %d", get_imsm_raid_level(map2));
1158 printf("\n");
1159 printf(" Members : %d", map->num_members);
1160 if (map2)
1161 printf(" <-- %d", map2->num_members);
1162 printf("\n");
0d80bb2f
DW
1163 printf(" Slots : [");
1164 for (i = 0; i < map->num_members; i++) {
238c0a71 1165 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW
1166 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1167 }
dd8bcb3b
AK
1168 printf("]");
1169 if (map2) {
1170 printf(" <-- [");
1171 for (i = 0; i < map2->num_members; i++) {
238c0a71 1172 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK
1173 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1174 }
1175 printf("]");
1176 }
1177 printf("\n");
7095bccb
AK
1178 printf(" Failed disk : ");
1179 if (map->failed_disk_num == 0xff)
1180 printf("none");
1181 else
1182 printf("%i", map->failed_disk_num);
1183 printf("\n");
620b1713
DW
1184 slot = get_imsm_disk_slot(map, disk_idx);
1185 if (slot >= 0) {
238c0a71 1186 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW
1187 printf(" This Slot : %d%s\n", slot,
1188 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1189 } else
cdddbdbc
DW
1190 printf(" This Slot : ?\n");
1191 sz = __le32_to_cpu(dev->size_high);
1192 sz <<= 32;
1193 sz += __le32_to_cpu(dev->size_low);
1194 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
1195 human_size(sz * 512));
5551b113 1196 sz = blocks_per_member(map);
cdddbdbc
DW
1197 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
1198 human_size(sz * 512));
5551b113
CA
1199 printf(" Sector Offset : %llu\n",
1200 pba_of_lba0(map));
1201 printf(" Num Stripes : %llu\n",
1202 num_data_stripes(map));
dd8bcb3b 1203 printf(" Chunk Size : %u KiB",
cdddbdbc 1204 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK
1205 if (map2)
1206 printf(" <-- %u KiB",
1207 __le16_to_cpu(map2->blocks_per_strip) / 2);
1208 printf("\n");
cdddbdbc 1209 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1210 printf(" Migrate State : ");
1484e727
DW
1211 if (dev->vol.migr_state) {
1212 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1213 printf("initialize\n");
1484e727 1214 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1215 printf("rebuild\n");
1484e727 1216 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1217 printf("check\n");
1484e727 1218 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1219 printf("general migration\n");
1484e727 1220 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1221 printf("state change\n");
1484e727 1222 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1223 printf("repair\n");
1484e727 1224 else
8655a7b1
DW
1225 printf("<unknown:%d>\n", migr_type(dev));
1226 } else
1227 printf("idle\n");
3393c6af
DW
1228 printf(" Map State : %s", map_state_str[map->map_state]);
1229 if (dev->vol.migr_state) {
238c0a71 1230 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1231
b10b37b8 1232 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD
1233 printf("\n Checkpoint : %u ",
1234 __le32_to_cpu(dev->vol.curr_migr_unit));
3136abe5 1235 if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0)))
464d40e8
LD
1236 printf("(N/A)");
1237 else
1238 printf("(%llu)", (unsigned long long)
1239 blocks_per_migr_unit(super, dev));
3393c6af
DW
1240 }
1241 printf("\n");
cdddbdbc 1242 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
cdddbdbc
DW
1243}
1244
0ec1f4e8 1245static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved)
cdddbdbc 1246{
1f24f035 1247 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW
1248 __u64 sz;
1249
0ec1f4e8 1250 if (index < -1 || !disk)
e9d82038
DW
1251 return;
1252
cdddbdbc 1253 printf("\n");
1f24f035 1254 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW
1255 if (index >= 0)
1256 printf(" Disk%02d Serial : %s\n", index, str);
1257 else
1258 printf(" Disk Serial : %s\n", str);
25ed7e59
DW
1259 printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "",
1260 is_configured(disk) ? " active" : "",
1261 is_failed(disk) ? " failed" : "");
cdddbdbc 1262 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1263 sz = total_blocks(disk) - reserved;
cdddbdbc
DW
1264 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
1265 human_size(sz * 512));
1266}
1267
520e69e2
AK
1268void examine_migr_rec_imsm(struct intel_super *super)
1269{
1270 struct migr_record *migr_rec = super->migr_rec;
1271 struct imsm_super *mpb = super->anchor;
1272 int i;
1273
1274 for (i = 0; i < mpb->num_raid_devs; i++) {
1275 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1276 struct imsm_map *map;
b4ab44d8 1277 int slot = -1;
3136abe5 1278
520e69e2
AK
1279 if (is_gen_migration(dev) == 0)
1280 continue;
1281
1282 printf("\nMigration Record Information:");
3136abe5 1283
44bfe6df
AK
1284 /* first map under migration */
1285 map = get_imsm_map(dev, MAP_0);
3136abe5
AK
1286 if (map)
1287 slot = get_imsm_disk_slot(map, super->disks->index);
1288 if ((map == NULL) || (slot > 1) || (slot < 0)) {
520e69e2
AK
1289 printf(" Empty\n ");
1290 printf("Examine one of first two disks in array\n");
1291 break;
1292 }
1293 printf("\n Status : ");
1294 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1295 printf("Normal\n");
1296 else
1297 printf("Contains Data\n");
1298 printf(" Current Unit : %u\n",
1299 __le32_to_cpu(migr_rec->curr_migr_unit));
1300 printf(" Family : %u\n",
1301 __le32_to_cpu(migr_rec->family_num));
1302 printf(" Ascending : %u\n",
1303 __le32_to_cpu(migr_rec->ascending_migr));
1304 printf(" Blocks Per Unit : %u\n",
1305 __le32_to_cpu(migr_rec->blocks_per_unit));
1306 printf(" Dest. Depth Per Unit : %u\n",
1307 __le32_to_cpu(migr_rec->dest_depth_per_unit));
1308 printf(" Checkpoint Area pba : %u\n",
1309 __le32_to_cpu(migr_rec->ckpt_area_pba));
1310 printf(" First member lba : %u\n",
1311 __le32_to_cpu(migr_rec->dest_1st_member_lba));
1312 printf(" Total Number of Units : %u\n",
1313 __le32_to_cpu(migr_rec->num_migr_units));
1314 printf(" Size of volume : %u\n",
1315 __le32_to_cpu(migr_rec->post_migr_vol_cap));
1316 printf(" Expansion space for LBA64 : %u\n",
1317 __le32_to_cpu(migr_rec->post_migr_vol_cap_hi));
1318 printf(" Record was read from : %u\n",
1319 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1320
1321 break;
1322 }
1323}
9e2d750d 1324#endif /* MDASSEMBLE */
19482bcc
AK
1325/*******************************************************************************
1326 * function: imsm_check_attributes
1327 * Description: Function checks if features represented by attributes flags
1328 * are supported by mdadm.
1329 * Parameters:
1330 * attributes - Attributes read from metadata
1331 * Returns:
1332 * 0 - passed attributes contains unsupported features flags
1333 * 1 - all features are supported
1334 ******************************************************************************/
1335static int imsm_check_attributes(__u32 attributes)
1336{
1337 int ret_val = 1;
418f9b36
N
1338 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1339
1340 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK
1341
1342 not_supported &= attributes;
1343 if (not_supported) {
418f9b36
N
1344 fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n",
1345 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK
1346 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1347 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1348 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1349 }
1350 if (not_supported & MPB_ATTRIB_2TB) {
1351 dprintf("\t\tMPB_ATTRIB_2TB\n");
1352 not_supported ^= MPB_ATTRIB_2TB;
1353 }
1354 if (not_supported & MPB_ATTRIB_RAID0) {
1355 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1356 not_supported ^= MPB_ATTRIB_RAID0;
1357 }
1358 if (not_supported & MPB_ATTRIB_RAID1) {
1359 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1360 not_supported ^= MPB_ATTRIB_RAID1;
1361 }
1362 if (not_supported & MPB_ATTRIB_RAID10) {
1363 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1364 not_supported ^= MPB_ATTRIB_RAID10;
1365 }
1366 if (not_supported & MPB_ATTRIB_RAID1E) {
1367 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1368 not_supported ^= MPB_ATTRIB_RAID1E;
1369 }
1370 if (not_supported & MPB_ATTRIB_RAID5) {
1371 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1372 not_supported ^= MPB_ATTRIB_RAID5;
1373 }
1374 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1375 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1376 not_supported ^= MPB_ATTRIB_RAIDCNG;
1377 }
1378 if (not_supported & MPB_ATTRIB_BBM) {
1379 dprintf("\t\tMPB_ATTRIB_BBM\n");
1380 not_supported ^= MPB_ATTRIB_BBM;
1381 }
1382 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1383 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1384 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1385 }
1386 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1387 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1388 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1389 }
1390 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1391 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1392 not_supported ^= MPB_ATTRIB_2TB_DISK;
1393 }
1394 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1395 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1396 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1397 }
1398 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1399 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1400 not_supported ^= MPB_ATTRIB_NEVER_USE;
1401 }
1402
1403 if (not_supported)
1404 dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported);
1405
1406 ret_val = 0;
1407 }
1408
1409 return ret_val;
1410}
1411
9e2d750d 1412#ifndef MDASSEMBLE
a5d85af7 1413static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 1414
cdddbdbc
DW
1415static void examine_super_imsm(struct supertype *st, char *homehost)
1416{
1417 struct intel_super *super = st->sb;
949c47a0 1418 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW
1419 char str[MAX_SIGNATURE_LENGTH];
1420 int i;
27fd6274
DW
1421 struct mdinfo info;
1422 char nbuf[64];
cdddbdbc 1423 __u32 sum;
14e8215b 1424 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 1425 struct dl *dl;
27fd6274 1426
cdddbdbc
DW
1427 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
1428 printf(" Magic : %s\n", str);
1429 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
1430 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 1431 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW
1432 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
1433 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK
1434 printf(" Attributes : ");
1435 if (imsm_check_attributes(mpb->attributes))
1436 printf("All supported\n");
1437 else
1438 printf("not supported\n");
a5d85af7 1439 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1440 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 1441 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW
1442 sum = __le32_to_cpu(mpb->check_sum);
1443 printf(" Checksum : %08x %s\n", sum,
949c47a0 1444 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
87eb16df 1445 printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
cdddbdbc
DW
1446 printf(" Disks : %d\n", mpb->num_disks);
1447 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
0ec1f4e8 1448 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved);
604b746f
JD
1449 if (super->bbm_log) {
1450 struct bbm_log *log = super->bbm_log;
1451
1452 printf("\n");
1453 printf("Bad Block Management Log:\n");
1454 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
1455 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
1456 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
1457 printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count));
13a3b65d
N
1458 printf(" First Spare : %llx\n",
1459 (unsigned long long) __le64_to_cpu(log->first_spare_lba));
604b746f 1460 }
44470971
DW
1461 for (i = 0; i < mpb->num_raid_devs; i++) {
1462 struct mdinfo info;
1463 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1464
1465 super->current_vol = i;
a5d85af7 1466 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1467 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 1468 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 1469 }
cdddbdbc
DW
1470 for (i = 0; i < mpb->num_disks; i++) {
1471 if (i == super->disks->index)
1472 continue;
0ec1f4e8 1473 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved);
cdddbdbc 1474 }
94827db3 1475
0ec1f4e8
DW
1476 for (dl = super->disks; dl; dl = dl->next)
1477 if (dl->index == -1)
1478 print_imsm_disk(&dl->disk, -1, reserved);
520e69e2
AK
1479
1480 examine_migr_rec_imsm(super);
cdddbdbc
DW
1481}
1482
061f2c6a 1483static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 1484{
27fd6274 1485 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N
1486 struct mdinfo info;
1487 char nbuf[64];
1e7bc0ed 1488 struct intel_super *super = st->sb;
1e7bc0ed 1489
0d5a423f
DW
1490 if (!super->anchor->num_raid_devs) {
1491 printf("ARRAY metadata=imsm\n");
1e7bc0ed 1492 return;
0d5a423f 1493 }
ff54de6e 1494
a5d85af7 1495 getinfo_super_imsm(st, &info, NULL);
4737ae25
N
1496 fname_from_uuid(st, &info, nbuf, ':');
1497 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
1498}
1499
1500static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
1501{
1502 /* We just write a generic IMSM ARRAY entry */
1503 struct mdinfo info;
1504 char nbuf[64];
1505 char nbuf1[64];
1506 struct intel_super *super = st->sb;
1507 int i;
1508
1509 if (!super->anchor->num_raid_devs)
1510 return;
1511
a5d85af7 1512 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1513 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW
1514 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1515 struct imsm_dev *dev = get_imsm_dev(super, i);
1516
1517 super->current_vol = i;
a5d85af7 1518 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1519 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 1520 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 1521 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 1522 }
cdddbdbc
DW
1523}
1524
9d84c8ea
DW
1525static void export_examine_super_imsm(struct supertype *st)
1526{
1527 struct intel_super *super = st->sb;
1528 struct imsm_super *mpb = super->anchor;
1529 struct mdinfo info;
1530 char nbuf[64];
1531
a5d85af7 1532 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW
1533 fname_from_uuid(st, &info, nbuf, ':');
1534 printf("MD_METADATA=imsm\n");
1535 printf("MD_LEVEL=container\n");
1536 printf("MD_UUID=%s\n", nbuf+5);
1537 printf("MD_DEVICES=%u\n", mpb->num_disks);
1538}
1539
cdddbdbc
DW
1540static void detail_super_imsm(struct supertype *st, char *homehost)
1541{
3ebe00a1
DW
1542 struct mdinfo info;
1543 char nbuf[64];
1544
a5d85af7 1545 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1546 fname_from_uuid(st, &info, nbuf, ':');
3ebe00a1 1547 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW
1548}
1549
1550static void brief_detail_super_imsm(struct supertype *st)
1551{
ff54de6e
N
1552 struct mdinfo info;
1553 char nbuf[64];
a5d85af7 1554 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1555 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 1556 printf(" UUID=%s", nbuf + 5);
cdddbdbc 1557}
d665cc31
DW
1558
1559static int imsm_read_serial(int fd, char *devname, __u8 *serial);
1560static void fd2devname(int fd, char *name);
1561
120dc887 1562static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 1563{
120dc887
LM
1564 /* dump an unsorted list of devices attached to AHCI Intel storage
1565 * controller, as well as non-connected ports
d665cc31
DW
1566 */
1567 int hba_len = strlen(hba_path) + 1;
1568 struct dirent *ent;
1569 DIR *dir;
1570 char *path = NULL;
1571 int err = 0;
1572 unsigned long port_mask = (1 << port_count) - 1;
1573
f21e18ca 1574 if (port_count > (int)sizeof(port_mask) * 8) {
d665cc31
DW
1575 if (verbose)
1576 fprintf(stderr, Name ": port_count %d out of range\n", port_count);
1577 return 2;
1578 }
1579
1580 /* scroll through /sys/dev/block looking for devices attached to
1581 * this hba
1582 */
1583 dir = opendir("/sys/dev/block");
1584 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
1585 int fd;
1586 char model[64];
1587 char vendor[64];
1588 char buf[1024];
1589 int major, minor;
1590 char *device;
1591 char *c;
1592 int port;
1593 int type;
1594
1595 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
1596 continue;
1597 path = devt_to_devpath(makedev(major, minor));
1598 if (!path)
1599 continue;
1600 if (!path_attached_to_hba(path, hba_path)) {
1601 free(path);
1602 path = NULL;
1603 continue;
1604 }
1605
1606 /* retrieve the scsi device type */
1607 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
1608 if (verbose)
1609 fprintf(stderr, Name ": failed to allocate 'device'\n");
1610 err = 2;
1611 break;
1612 }
1613 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
1614 if (load_sys(device, buf) != 0) {
1615 if (verbose)
1616 fprintf(stderr, Name ": failed to read device type for %s\n",
1617 path);
1618 err = 2;
1619 free(device);
1620 break;
1621 }
1622 type = strtoul(buf, NULL, 10);
1623
1624 /* if it's not a disk print the vendor and model */
1625 if (!(type == 0 || type == 7 || type == 14)) {
1626 vendor[0] = '\0';
1627 model[0] = '\0';
1628 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
1629 if (load_sys(device, buf) == 0) {
1630 strncpy(vendor, buf, sizeof(vendor));
1631 vendor[sizeof(vendor) - 1] = '\0';
1632 c = (char *) &vendor[sizeof(vendor) - 1];
1633 while (isspace(*c) || *c == '\0')
1634 *c-- = '\0';
1635
1636 }
1637 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
1638 if (load_sys(device, buf) == 0) {
1639 strncpy(model, buf, sizeof(model));
1640 model[sizeof(model) - 1] = '\0';
1641 c = (char *) &model[sizeof(model) - 1];
1642 while (isspace(*c) || *c == '\0')
1643 *c-- = '\0';
1644 }
1645
1646 if (vendor[0] && model[0])
1647 sprintf(buf, "%.64s %.64s", vendor, model);
1648 else
1649 switch (type) { /* numbers from hald/linux/device.c */
1650 case 1: sprintf(buf, "tape"); break;
1651 case 2: sprintf(buf, "printer"); break;
1652 case 3: sprintf(buf, "processor"); break;
1653 case 4:
1654 case 5: sprintf(buf, "cdrom"); break;
1655 case 6: sprintf(buf, "scanner"); break;
1656 case 8: sprintf(buf, "media_changer"); break;
1657 case 9: sprintf(buf, "comm"); break;
1658 case 12: sprintf(buf, "raid"); break;
1659 default: sprintf(buf, "unknown");
1660 }
1661 } else
1662 buf[0] = '\0';
1663 free(device);
1664
1665 /* chop device path to 'host%d' and calculate the port number */
1666 c = strchr(&path[hba_len], '/');
4e5e717d
AW
1667 if (!c) {
1668 if (verbose)
1669 fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len);
1670 err = 2;
1671 break;
1672 }
d665cc31
DW
1673 *c = '\0';
1674 if (sscanf(&path[hba_len], "host%d", &port) == 1)
1675 port -= host_base;
1676 else {
1677 if (verbose) {
1678 *c = '/'; /* repair the full string */
1679 fprintf(stderr, Name ": failed to determine port number for %s\n",
1680 path);
1681 }
1682 err = 2;
1683 break;
1684 }
1685
1686 /* mark this port as used */
1687 port_mask &= ~(1 << port);
1688
1689 /* print out the device information */
1690 if (buf[0]) {
1691 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
1692 continue;
1693 }
1694
1695 fd = dev_open(ent->d_name, O_RDONLY);
1696 if (fd < 0)
1697 printf(" Port%d : - disk info unavailable -\n", port);
1698 else {
1699 fd2devname(fd, buf);
1700 printf(" Port%d : %s", port, buf);
1701 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 1702 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 1703 else
664d5325 1704 printf(" ()\n");
4dab422a 1705 close(fd);
d665cc31 1706 }
d665cc31
DW
1707 free(path);
1708 path = NULL;
1709 }
1710 if (path)
1711 free(path);
1712 if (dir)
1713 closedir(dir);
1714 if (err == 0) {
1715 int i;
1716
1717 for (i = 0; i < port_count; i++)
1718 if (port_mask & (1 << i))
1719 printf(" Port%d : - no device attached -\n", i);
1720 }
1721
1722 return err;
1723}
1724
120dc887
LM
1725static void print_found_intel_controllers(struct sys_dev *elem)
1726{
1727 for (; elem; elem = elem->next) {
1728 fprintf(stderr, Name ": found Intel(R) ");
1729 if (elem->type == SYS_DEV_SATA)
1730 fprintf(stderr, "SATA ");
155cbb4c
LM
1731 else if (elem->type == SYS_DEV_SAS)
1732 fprintf(stderr, "SAS ");
120dc887
LM
1733 fprintf(stderr, "RAID controller");
1734 if (elem->pci_id)
1735 fprintf(stderr, " at %s", elem->pci_id);
1736 fprintf(stderr, ".\n");
1737 }
1738 fflush(stderr);
1739}
1740
120dc887
LM
1741static int ahci_get_port_count(const char *hba_path, int *port_count)
1742{
1743 struct dirent *ent;
1744 DIR *dir;
1745 int host_base = -1;
1746
1747 *port_count = 0;
1748 if ((dir = opendir(hba_path)) == NULL)
1749 return -1;
1750
1751 for (ent = readdir(dir); ent; ent = readdir(dir)) {
1752 int host;
1753
1754 if (sscanf(ent->d_name, "host%d", &host) != 1)
1755 continue;
1756 if (*port_count == 0)
1757 host_base = host;
1758 else if (host < host_base)
1759 host_base = host;
1760
1761 if (host + 1 > *port_count + host_base)
1762 *port_count = host + 1 - host_base;
1763 }
1764 closedir(dir);
1765 return host_base;
1766}
1767
a891a3c2
LM
1768static void print_imsm_capability(const struct imsm_orom *orom)
1769{
1770 printf(" Platform : Intel(R) Matrix Storage Manager\n");
1771 printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
1772 orom->hotfix_ver, orom->build);
1773 printf(" RAID Levels :%s%s%s%s%s\n",
1774 imsm_orom_has_raid0(orom) ? " raid0" : "",
1775 imsm_orom_has_raid1(orom) ? " raid1" : "",
1776 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
1777 imsm_orom_has_raid10(orom) ? " raid10" : "",
1778 imsm_orom_has_raid5(orom) ? " raid5" : "");
1779 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
1780 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
1781 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
1782 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
1783 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
1784 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
1785 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
1786 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
1787 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
1788 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
1789 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
1790 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
1791 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
1792 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
1793 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
1794 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
1795 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA
1796 printf(" 2TB volumes :%s supported\n",
1797 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
1798 printf(" 2TB disks :%s supported\n",
1799 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8
LM
1800 printf(" Max Disks : %d\n", orom->tds);
1801 printf(" Max Volumes : %d per array, %d per controller\n",
1802 orom->vpa, orom->vphba);
a891a3c2
LM
1803 return;
1804}
1805
5615172f 1806static int detail_platform_imsm(int verbose, int enumerate_only)
d665cc31
DW
1807{
1808 /* There are two components to imsm platform support, the ahci SATA
1809 * controller and the option-rom. To find the SATA controller we
1810 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
1811 * controller with the Intel vendor id is present. This approach
1812 * allows mdadm to leverage the kernel's ahci detection logic, with the
1813 * caveat that if ahci.ko is not loaded mdadm will not be able to
1814 * detect platform raid capabilities. The option-rom resides in a
1815 * platform "Adapter ROM". We scan for its signature to retrieve the
1816 * platform capabilities. If raid support is disabled in the BIOS the
1817 * option-rom capability structure will not be available.
1818 */
1819 const struct imsm_orom *orom;
1820 struct sys_dev *list, *hba;
d665cc31
DW
1821 int host_base = 0;
1822 int port_count = 0;
120dc887 1823 int result=0;
d665cc31 1824
5615172f 1825 if (enumerate_only) {
a891a3c2 1826 if (check_env("IMSM_NO_PLATFORM"))
5615172f 1827 return 0;
a891a3c2
LM
1828 list = find_intel_devices();
1829 if (!list)
1830 return 2;
1831 for (hba = list; hba; hba = hba->next) {
1832 orom = find_imsm_capability(hba->type);
1833 if (!orom) {
1834 result = 2;
1835 break;
1836 }
1837 }
1838 free_sys_dev(&list);
1839 return result;
5615172f
DW
1840 }
1841
155cbb4c
LM
1842 list = find_intel_devices();
1843 if (!list) {
d665cc31 1844 if (verbose)
155cbb4c
LM
1845 fprintf(stderr, Name ": no active Intel(R) RAID "
1846 "controller found.\n");
d665cc31
DW
1847 free_sys_dev(&list);
1848 return 2;
1849 } else if (verbose)
155cbb4c 1850 print_found_intel_controllers(list);
d665cc31 1851
a891a3c2
LM
1852 for (hba = list; hba; hba = hba->next) {
1853 orom = find_imsm_capability(hba->type);
1854 if (!orom)
1855 fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n",
1856 hba->path, get_sys_dev_type(hba->type));
1857 else
1858 print_imsm_capability(orom);
d665cc31
DW
1859 }
1860
120dc887
LM
1861 for (hba = list; hba; hba = hba->next) {
1862 printf(" I/O Controller : %s (%s)\n",
1863 hba->path, get_sys_dev_type(hba->type));
d665cc31 1864
120dc887
LM
1865 if (hba->type == SYS_DEV_SATA) {
1866 host_base = ahci_get_port_count(hba->path, &port_count);
1867 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
1868 if (verbose)
1869 fprintf(stderr, Name ": failed to enumerate "
1870 "ports on SATA controller at %s.", hba->pci_id);
1871 result |= 2;
1872 }
1873 }
d665cc31 1874 }
155cbb4c 1875
120dc887
LM
1876 free_sys_dev(&list);
1877 return result;
d665cc31 1878}
cdddbdbc
DW
1879#endif
1880
1881static int match_home_imsm(struct supertype *st, char *homehost)
1882{
5115ca67
DW
1883 /* the imsm metadata format does not specify any host
1884 * identification information. We return -1 since we can never
1885 * confirm nor deny whether a given array is "meant" for this
148acb7b 1886 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW
1887 * exclude member disks that do not belong, and we rely on
1888 * mdadm.conf to specify the arrays that should be assembled.
1889 * Auto-assembly may still pick up "foreign" arrays.
1890 */
cdddbdbc 1891
9362c1c8 1892 return -1;
cdddbdbc
DW
1893}
1894
1895static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
1896{
51006d85
N
1897 /* The uuid returned here is used for:
1898 * uuid to put into bitmap file (Create, Grow)
1899 * uuid for backup header when saving critical section (Grow)
1900 * comparing uuids when re-adding a device into an array
1901 * In these cases the uuid required is that of the data-array,
1902 * not the device-set.
1903 * uuid to recognise same set when adding a missing device back
1904 * to an array. This is a uuid for the device-set.
1905 *
1906 * For each of these we can make do with a truncated
1907 * or hashed uuid rather than the original, as long as
1908 * everyone agrees.
1909 * In each case the uuid required is that of the data-array,
1910 * not the device-set.
43dad3d6 1911 */
51006d85
N
1912 /* imsm does not track uuid's so we synthesis one using sha1 on
1913 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 1914 * - the orig_family_num of the container
51006d85
N
1915 * - the index number of the volume
1916 * - the 'serial' number of the volume.
1917 * Hopefully these are all constant.
1918 */
1919 struct intel_super *super = st->sb;
43dad3d6 1920
51006d85
N
1921 char buf[20];
1922 struct sha1_ctx ctx;
1923 struct imsm_dev *dev = NULL;
148acb7b 1924 __u32 family_num;
51006d85 1925
148acb7b
DW
1926 /* some mdadm versions failed to set ->orig_family_num, in which
1927 * case fall back to ->family_num. orig_family_num will be
1928 * fixed up with the first metadata update.
1929 */
1930 family_num = super->anchor->orig_family_num;
1931 if (family_num == 0)
1932 family_num = super->anchor->family_num;
51006d85 1933 sha1_init_ctx(&ctx);
92bd8f8d 1934 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 1935 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N
1936 if (super->current_vol >= 0)
1937 dev = get_imsm_dev(super, super->current_vol);
1938 if (dev) {
1939 __u32 vol = super->current_vol;
1940 sha1_process_bytes(&vol, sizeof(vol), &ctx);
1941 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
1942 }
1943 sha1_finish_ctx(&ctx, buf);
1944 memcpy(uuid, buf, 4*4);
cdddbdbc
DW
1945}
1946
0d481d37 1947#if 0
4f5bc454
DW
1948static void
1949get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 1950{
cdddbdbc
DW
1951 __u8 *v = get_imsm_version(mpb);
1952 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
1953 char major[] = { 0, 0, 0 };
1954 char minor[] = { 0 ,0, 0 };
1955 char patch[] = { 0, 0, 0 };
1956 char *ver_parse[] = { major, minor, patch };
1957 int i, j;
1958
1959 i = j = 0;
1960 while (*v != '\0' && v < end) {
1961 if (*v != '.' && j < 2)
1962 ver_parse[i][j++] = *v;
1963 else {
1964 i++;
1965 j = 0;
1966 }
1967 v++;
1968 }
1969
4f5bc454
DW
1970 *m = strtol(minor, NULL, 0);
1971 *p = strtol(patch, NULL, 0);
1972}
0d481d37 1973#endif
4f5bc454 1974
1e5c6983
DW
1975static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
1976{
1977 /* migr_strip_size when repairing or initializing parity */
238c0a71 1978 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW
1979 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
1980
1981 switch (get_imsm_raid_level(map)) {
1982 case 5:
1983 case 10:
1984 return chunk;
1985 default:
1986 return 128*1024 >> 9;
1987 }
1988}
1989
1990static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
1991{
1992 /* migr_strip_size when rebuilding a degraded disk, no idea why
1993 * this is different than migr_strip_size_resync(), but it's good
1994 * to be compatible
1995 */
238c0a71 1996 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW
1997 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
1998
1999 switch (get_imsm_raid_level(map)) {
2000 case 1:
2001 case 10:
2002 if (map->num_members % map->num_domains == 0)
2003 return 128*1024 >> 9;
2004 else
2005 return chunk;
2006 case 5:
2007 return max((__u32) 64*1024 >> 9, chunk);
2008 default:
2009 return 128*1024 >> 9;
2010 }
2011}
2012
2013static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2014{
238c0a71
AK
2015 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2016 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW
2017 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2018 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2019
2020 return max((__u32) 1, hi_chunk / lo_chunk);
2021}
2022
2023static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2024{
238c0a71 2025 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW
2026 int level = get_imsm_raid_level(lo);
2027
2028 if (level == 1 || level == 10) {
238c0a71 2029 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW
2030
2031 return hi->num_domains;
2032 } else
2033 return num_stripes_per_unit_resync(dev);
2034}
2035
98130f40 2036static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map)
1e5c6983
DW
2037{
2038 /* named 'imsm_' because raid0, raid1 and raid10
2039 * counter-intuitively have the same number of data disks
2040 */
98130f40 2041 struct imsm_map *map = get_imsm_map(dev, second_map);
1e5c6983
DW
2042
2043 switch (get_imsm_raid_level(map)) {
2044 case 0:
36fd8ccc
AK
2045 return map->num_members;
2046 break;
1e5c6983
DW
2047 case 1:
2048 case 10:
36fd8ccc 2049 return map->num_members/2;
1e5c6983
DW
2050 case 5:
2051 return map->num_members - 1;
2052 default:
2053 dprintf("%s: unsupported raid level\n", __func__);
2054 return 0;
2055 }
2056}
2057
2058static __u32 parity_segment_depth(struct imsm_dev *dev)
2059{
238c0a71 2060 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW
2061 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2062
2063 switch(get_imsm_raid_level(map)) {
2064 case 1:
2065 case 10:
2066 return chunk * map->num_domains;
2067 case 5:
2068 return chunk * map->num_members;
2069 default:
2070 return chunk;
2071 }
2072}
2073
2074static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2075{
238c0a71 2076 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW
2077 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2078 __u32 strip = block / chunk;
2079
2080 switch (get_imsm_raid_level(map)) {
2081 case 1:
2082 case 10: {
2083 __u32 vol_strip = (strip * map->num_domains) + 1;
2084 __u32 vol_stripe = vol_strip / map->num_members;
2085
2086 return vol_stripe * chunk + block % chunk;
2087 } case 5: {
2088 __u32 stripe = strip / (map->num_members - 1);
2089
2090 return stripe * chunk + block % chunk;
2091 }
2092 default:
2093 return 0;
2094 }
2095}
2096
c47b0ff6
AK
2097static __u64 blocks_per_migr_unit(struct intel_super *super,
2098 struct imsm_dev *dev)
1e5c6983
DW
2099{
2100 /* calculate the conversion factor between per member 'blocks'
2101 * (md/{resync,rebuild}_start) and imsm migration units, return
2102 * 0 for the 'not migrating' and 'unsupported migration' cases
2103 */
2104 if (!dev->vol.migr_state)
2105 return 0;
2106
2107 switch (migr_type(dev)) {
c47b0ff6
AK
2108 case MIGR_GEN_MIGR: {
2109 struct migr_record *migr_rec = super->migr_rec;
2110 return __le32_to_cpu(migr_rec->blocks_per_unit);
2111 }
1e5c6983
DW
2112 case MIGR_VERIFY:
2113 case MIGR_REPAIR:
2114 case MIGR_INIT: {
238c0a71 2115 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW
2116 __u32 stripes_per_unit;
2117 __u32 blocks_per_unit;
2118 __u32 parity_depth;
2119 __u32 migr_chunk;
2120 __u32 block_map;
2121 __u32 block_rel;
2122 __u32 segment;
2123 __u32 stripe;
2124 __u8 disks;
2125
2126 /* yes, this is really the translation of migr_units to
2127 * per-member blocks in the 'resync' case
2128 */
2129 stripes_per_unit = num_stripes_per_unit_resync(dev);
2130 migr_chunk = migr_strip_blocks_resync(dev);
238c0a71 2131 disks = imsm_num_data_members(dev, MAP_0);
1e5c6983 2132 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2133 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW
2134 segment = blocks_per_unit / stripe;
2135 block_rel = blocks_per_unit - segment * stripe;
2136 parity_depth = parity_segment_depth(dev);
2137 block_map = map_migr_block(dev, block_rel);
2138 return block_map + parity_depth * segment;
2139 }
2140 case MIGR_REBUILD: {
2141 __u32 stripes_per_unit;
2142 __u32 migr_chunk;
2143
2144 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2145 migr_chunk = migr_strip_blocks_rebuild(dev);
2146 return migr_chunk * stripes_per_unit;
2147 }
1e5c6983
DW
2148 case MIGR_STATE_CHANGE:
2149 default:
2150 return 0;
2151 }
2152}
2153
c2c087e6
DW
2154static int imsm_level_to_layout(int level)
2155{
2156 switch (level) {
2157 case 0:
2158 case 1:
2159 return 0;
2160 case 5:
2161 case 6:
a380c027 2162 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2163 case 10:
c92a2527 2164 return 0x102;
c2c087e6 2165 }
a18a888e 2166 return UnSet;
c2c087e6
DW
2167}
2168
8e59f3d8
AK
2169/*******************************************************************************
2170 * Function: read_imsm_migr_rec
2171 * Description: Function reads imsm migration record from last sector of disk
2172 * Parameters:
2173 * fd : disk descriptor
2174 * super : metadata info
2175 * Returns:
2176 * 0 : success,
2177 * -1 : fail
2178 ******************************************************************************/
2179static int read_imsm_migr_rec(int fd, struct intel_super *super)
2180{
2181 int ret_val = -1;
2182 unsigned long long dsize;
2183
2184 get_dev_size(fd, NULL, &dsize);
17a4eaf9 2185 if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) {
8e59f3d8
AK
2186 fprintf(stderr,
2187 Name ": Cannot seek to anchor block: %s\n",
2188 strerror(errno));
2189 goto out;
2190 }
17a4eaf9
AK
2191 if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) !=
2192 MIGR_REC_BUF_SIZE) {
8e59f3d8
AK
2193 fprintf(stderr,
2194 Name ": Cannot read migr record block: %s\n",
2195 strerror(errno));
2196 goto out;
2197 }
2198 ret_val = 0;
2199
2200out:
2201 return ret_val;
2202}
2203
3136abe5
AK
2204static struct imsm_dev *imsm_get_device_during_migration(
2205 struct intel_super *super)
2206{
2207
2208 struct intel_dev *dv;
2209
2210 for (dv = super->devlist; dv; dv = dv->next) {
2211 if (is_gen_migration(dv->dev))
2212 return dv->dev;
2213 }
2214 return NULL;
2215}
2216
8e59f3d8
AK
2217/*******************************************************************************
2218 * Function: load_imsm_migr_rec
2219 * Description: Function reads imsm migration record (it is stored at the last
2220 * sector of disk)
2221 * Parameters:
2222 * super : imsm internal array info
2223 * info : general array info
2224 * Returns:
2225 * 0 : success
2226 * -1 : fail
4c965cc9 2227 * -2 : no migration in progress
8e59f3d8
AK
2228 ******************************************************************************/
2229static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
2230{
2231 struct mdinfo *sd;
2232 struct dl *dl = NULL;
2233 char nm[30];
2234 int retval = -1;
2235 int fd = -1;
3136abe5
AK
2236 struct imsm_dev *dev;
2237 struct imsm_map *map = NULL;
b4ab44d8 2238 int slot = -1;
3136abe5
AK
2239
2240 /* find map under migration */
2241 dev = imsm_get_device_during_migration(super);
2242 /* nothing to load,no migration in progress?
2243 */
2244 if (dev == NULL)
4c965cc9 2245 return -2;
44bfe6df 2246 map = get_imsm_map(dev, MAP_0);
8e59f3d8
AK
2247
2248 if (info) {
2249 for (sd = info->devs ; sd ; sd = sd->next) {
3136abe5
AK
2250 /* skip spare and failed disks
2251 */
2252 if (sd->disk.raid_disk < 0)
2253 continue;
8e59f3d8 2254 /* read only from one of the first two slots */
3136abe5
AK
2255 if (map)
2256 slot = get_imsm_disk_slot(map,
2257 sd->disk.raid_disk);
2258 if ((map == NULL) || (slot > 1) || (slot < 0))
8e59f3d8 2259 continue;
3136abe5 2260
8e59f3d8
AK
2261 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2262 fd = dev_open(nm, O_RDONLY);
2263 if (fd >= 0)
2264 break;
2265 }
2266 }
2267 if (fd < 0) {
2268 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK
2269 /* skip spare and failed disks
2270 */
2271 if (dl->index < 0)
2272 continue;
8e59f3d8 2273 /* read only from one of the first two slots */
3136abe5
AK
2274 if (map)
2275 slot = get_imsm_disk_slot(map, dl->index);
2276 if ((map == NULL) || (slot > 1) || (slot < 0))
8e59f3d8
AK
2277 continue;
2278 sprintf(nm, "%d:%d", dl->major, dl->minor);
2279 fd = dev_open(nm, O_RDONLY);
2280 if (fd >= 0)
2281 break;
2282 }
2283 }
2284 if (fd < 0)
2285 goto out;
2286 retval = read_imsm_migr_rec(fd, super);
2287
2288out:
2289 if (fd >= 0)
2290 close(fd);
2291 return retval;
2292}
2293
9e2d750d 2294#ifndef MDASSEMBLE
c17608ea
AK
2295/*******************************************************************************
2296 * function: imsm_create_metadata_checkpoint_update
2297 * Description: It creates update for checkpoint change.
2298 * Parameters:
2299 * super : imsm internal array info
2300 * u : pointer to prepared update
2301 * Returns:
2302 * Uptate length.
2303 * If length is equal to 0, input pointer u contains no update
2304 ******************************************************************************/
2305static int imsm_create_metadata_checkpoint_update(
2306 struct intel_super *super,
2307 struct imsm_update_general_migration_checkpoint **u)
2308{
2309
2310 int update_memory_size = 0;
2311
2312 dprintf("imsm_create_metadata_checkpoint_update(enter)\n");
2313
2314 if (u == NULL)
2315 return 0;
2316 *u = NULL;
2317
2318 /* size of all update data without anchor */
2319 update_memory_size =
2320 sizeof(struct imsm_update_general_migration_checkpoint);
2321
2322 *u = calloc(1, update_memory_size);
2323 if (*u == NULL) {
2324 dprintf("error: cannot get memory for "
2325 "imsm_create_metadata_checkpoint_update update\n");
2326 return 0;
2327 }
2328 (*u)->type = update_general_migration_checkpoint;
2329 (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit);
2330 dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n",
2331 (*u)->curr_migr_unit);
2332
2333 return update_memory_size;
2334}
2335
2336
2337static void imsm_update_metadata_locally(struct supertype *st,
2338 void *buf, int len);
2339
687629c2
AK
2340/*******************************************************************************
2341 * Function: write_imsm_migr_rec
2342 * Description: Function writes imsm migration record
2343 * (at the last sector of disk)
2344 * Parameters:
2345 * super : imsm internal array info
2346 * Returns:
2347 * 0 : success
2348 * -1 : if fail
2349 ******************************************************************************/
2350static int write_imsm_migr_rec(struct supertype *st)
2351{
2352 struct intel_super *super = st->sb;
2353 unsigned long long dsize;
2354 char nm[30];
2355 int fd = -1;
2356 int retval = -1;
2357 struct dl *sd;
c17608ea
AK
2358 int len;
2359 struct imsm_update_general_migration_checkpoint *u;
3136abe5
AK
2360 struct imsm_dev *dev;
2361 struct imsm_map *map = NULL;
2362
2363 /* find map under migration */
2364 dev = imsm_get_device_during_migration(super);
2365 /* if no migration, write buffer anyway to clear migr_record
2366 * on disk based on first available device
2367 */
2368 if (dev == NULL)
2369 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
2370 super->current_vol);
2371
44bfe6df 2372 map = get_imsm_map(dev, MAP_0);
687629c2
AK
2373
2374 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 2375 int slot = -1;
3136abe5
AK
2376
2377 /* skip failed and spare devices */
2378 if (sd->index < 0)
2379 continue;
687629c2 2380 /* write to 2 first slots only */
3136abe5
AK
2381 if (map)
2382 slot = get_imsm_disk_slot(map, sd->index);
2383 if ((map == NULL) || (slot > 1) || (slot < 0))
687629c2 2384 continue;
3136abe5 2385
687629c2
AK
2386 sprintf(nm, "%d:%d", sd->major, sd->minor);
2387 fd = dev_open(nm, O_RDWR);
2388 if (fd < 0)
2389 continue;
2390 get_dev_size(fd, NULL, &dsize);
17a4eaf9 2391 if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) {
687629c2
AK
2392 fprintf(stderr,
2393 Name ": Cannot seek to anchor block: %s\n",
2394 strerror(errno));
2395 goto out;
2396 }
17a4eaf9
AK
2397 if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) !=
2398 MIGR_REC_BUF_SIZE) {
687629c2
AK
2399 fprintf(stderr,
2400 Name ": Cannot write migr record block: %s\n",
2401 strerror(errno));
2402 goto out;
2403 }
2404 close(fd);
2405 fd = -1;
2406 }
c17608ea
AK
2407 /* update checkpoint information in metadata */
2408 len = imsm_create_metadata_checkpoint_update(super, &u);
2409
2410 if (len <= 0) {
2411 dprintf("imsm: Cannot prepare update\n");
2412 goto out;
2413 }
2414 /* update metadata locally */
2415 imsm_update_metadata_locally(st, u, len);
2416 /* and possibly remotely */
2417 if (st->update_tail) {
2418 append_metadata_update(st, u, len);
2419 /* during reshape we do all work inside metadata handler
2420 * manage_reshape(), so metadata update has to be triggered
2421 * insida it
2422 */
2423 flush_metadata_updates(st);
2424 st->update_tail = &st->updates;
2425 } else
2426 free(u);
687629c2
AK
2427
2428 retval = 0;
2429 out:
2430 if (fd >= 0)
2431 close(fd);
2432 return retval;
2433}
9e2d750d 2434#endif /* MDASSEMBLE */
687629c2 2435
e2962bfc
AK
2436/* spare/missing disks activations are not allowe when
2437 * array/container performs reshape operation, because
2438 * all arrays in container works on the same disks set
2439 */
2440int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
2441{
2442 int rv = 0;
2443 struct intel_dev *i_dev;
2444 struct imsm_dev *dev;
2445
2446 /* check whole container
2447 */
2448 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
2449 dev = i_dev->dev;
3ad25638 2450 if (is_gen_migration(dev)) {
e2962bfc
AK
2451 /* No repair during any migration in container
2452 */
2453 rv = 1;
2454 break;
2455 }
2456 }
2457 return rv;
2458}
c41e00b2
AK
2459static unsigned long long imsm_component_size_aligment_check(int level,
2460 int chunk_size,
2461 unsigned long long component_size)
2462{
2463 unsigned int component_size_alligment;
2464
2465 /* check component size aligment
2466 */
2467 component_size_alligment = component_size % (chunk_size/512);
2468
2469 dprintf("imsm_component_size_aligment_check(Level: %i, "
2470 "chunk_size = %i, component_size = %llu), "
2471 "component_size_alligment = %u\n",
2472 level, chunk_size, component_size,
2473 component_size_alligment);
2474
2475 if (component_size_alligment && (level != 1) && (level != UnSet)) {
2476 dprintf("imsm: reported component size alligned from %llu ",
2477 component_size);
2478 component_size -= component_size_alligment;
2479 dprintf("to %llu (%i).\n",
2480 component_size, component_size_alligment);
2481 }
2482
2483 return component_size;
2484}
e2962bfc 2485
a5d85af7 2486static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW
2487{
2488 struct intel_super *super = st->sb;
c47b0ff6 2489 struct migr_record *migr_rec = super->migr_rec;
949c47a0 2490 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK
2491 struct imsm_map *map = get_imsm_map(dev, MAP_0);
2492 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 2493 struct imsm_map *map_to_analyse = map;
efb30e7f 2494 struct dl *dl;
e207da2f 2495 char *devname;
a5d85af7 2496 int map_disks = info->array.raid_disks;
bf5a934a 2497
95eeceeb 2498 memset(info, 0, sizeof(*info));
b335e593
AK
2499 if (prev_map)
2500 map_to_analyse = prev_map;
2501
ca0748fa 2502 dl = super->current_disk;
9894ec0d 2503
bf5a934a 2504 info->container_member = super->current_vol;
cd0430a1 2505 info->array.raid_disks = map->num_members;
b335e593 2506 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW
2507 info->array.layout = imsm_level_to_layout(info->array.level);
2508 info->array.md_minor = -1;
2509 info->array.ctime = 0;
2510 info->array.utime = 0;
b335e593
AK
2511 info->array.chunk_size =
2512 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
301406c9 2513 info->array.state = !dev->vol.dirty;
da9b4a62
DW
2514 info->custom_array_size = __le32_to_cpu(dev->size_high);
2515 info->custom_array_size <<= 32;
2516 info->custom_array_size |= __le32_to_cpu(dev->size_low);
3ad25638
AK
2517 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
2518
3f510843 2519 if (is_gen_migration(dev)) {
3f83228a 2520 info->reshape_active = 1;
b335e593
AK
2521 info->new_level = get_imsm_raid_level(map);
2522 info->new_layout = imsm_level_to_layout(info->new_level);
2523 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 2524 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N
2525 if (info->delta_disks) {
2526 /* this needs to be applied to every array
2527 * in the container.
2528 */
81219e70 2529 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 2530 }
3f83228a
N
2531 /* We shape information that we give to md might have to be
2532 * modify to cope with md's requirement for reshaping arrays.
2533 * For example, when reshaping a RAID0, md requires it to be
2534 * presented as a degraded RAID4.
2535 * Also if a RAID0 is migrating to a RAID5 we need to specify
2536 * the array as already being RAID5, but the 'before' layout
2537 * is a RAID4-like layout.
2538 */
2539 switch (info->array.level) {
2540 case 0:
2541 switch(info->new_level) {
2542 case 0:
2543 /* conversion is happening as RAID4 */
2544 info->array.level = 4;
2545 info->array.raid_disks += 1;
2546 break;
2547 case 5:
2548 /* conversion is happening as RAID5 */
2549 info->array.level = 5;
2550 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N
2551 info->delta_disks -= 1;
2552 break;
2553 default:
2554 /* FIXME error message */
2555 info->array.level = UnSet;
2556 break;
2557 }
2558 break;
2559 }
b335e593
AK
2560 } else {
2561 info->new_level = UnSet;
2562 info->new_layout = UnSet;
2563 info->new_chunk = info->array.chunk_size;
3f83228a 2564 info->delta_disks = 0;
b335e593 2565 }
ca0748fa 2566
efb30e7f
DW
2567 if (dl) {
2568 info->disk.major = dl->major;
2569 info->disk.minor = dl->minor;
ca0748fa 2570 info->disk.number = dl->index;
656b6b5a
N
2571 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
2572 dl->index);
efb30e7f 2573 }
bf5a934a 2574
5551b113
CA
2575 info->data_offset = pba_of_lba0(map_to_analyse);
2576 info->component_size = blocks_per_member(map_to_analyse);
139dae11 2577
c41e00b2
AK
2578 info->component_size = imsm_component_size_aligment_check(
2579 info->array.level,
2580 info->array.chunk_size,
2581 info->component_size);
139dae11 2582
301406c9 2583 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 2584 info->recovery_start = MaxSector;
bf5a934a 2585
d2e6d5d6 2586 info->reshape_progress = 0;
b6796ce1 2587 info->resync_start = MaxSector;
b9172665
AK
2588 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2589 dev->vol.dirty) &&
2590 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 2591 info->resync_start = 0;
b6796ce1
AK
2592 }
2593 if (dev->vol.migr_state) {
1e5c6983
DW
2594 switch (migr_type(dev)) {
2595 case MIGR_REPAIR:
2596 case MIGR_INIT: {
c47b0ff6
AK
2597 __u64 blocks_per_unit = blocks_per_migr_unit(super,
2598 dev);
1e5c6983
DW
2599 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
2600
2601 info->resync_start = blocks_per_unit * units;
2602 break;
2603 }
d2e6d5d6 2604 case MIGR_GEN_MIGR: {
c47b0ff6
AK
2605 __u64 blocks_per_unit = blocks_per_migr_unit(super,
2606 dev);
2607 __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit);
04fa9523
AK
2608 unsigned long long array_blocks;
2609 int used_disks;
d2e6d5d6 2610
befb629b
AK
2611 if (__le32_to_cpu(migr_rec->ascending_migr) &&
2612 (units <
2613 (__le32_to_cpu(migr_rec->num_migr_units)-1)) &&
2614 (super->migr_rec->rec_status ==
2615 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
2616 units++;
2617
d2e6d5d6 2618 info->reshape_progress = blocks_per_unit * units;
6289d1e0 2619
d2e6d5d6
AK
2620 dprintf("IMSM: General Migration checkpoint : %llu "
2621 "(%llu) -> read reshape progress : %llu\n",
19986c72
MB
2622 (unsigned long long)units,
2623 (unsigned long long)blocks_per_unit,
2624 info->reshape_progress);
75156c46 2625
238c0a71 2626 used_disks = imsm_num_data_members(dev, MAP_1);
75156c46 2627 if (used_disks > 0) {
5551b113 2628 array_blocks = blocks_per_member(map) *
75156c46
AK
2629 used_disks;
2630 /* round array size down to closest MB
2631 */
2632 info->custom_array_size = (array_blocks
2633 >> SECT_PER_MB_SHIFT)
2634 << SECT_PER_MB_SHIFT;
2635 }
d2e6d5d6 2636 }
1e5c6983
DW
2637 case MIGR_VERIFY:
2638 /* we could emulate the checkpointing of
2639 * 'sync_action=check' migrations, but for now
2640 * we just immediately complete them
2641 */
2642 case MIGR_REBUILD:
2643 /* this is handled by container_content_imsm() */
1e5c6983
DW
2644 case MIGR_STATE_CHANGE:
2645 /* FIXME handle other migrations */
2646 default:
2647 /* we are not dirty, so... */
2648 info->resync_start = MaxSector;
2649 }
b6796ce1 2650 }
301406c9
DW
2651
2652 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
2653 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 2654
f35f2525
N
2655 info->array.major_version = -1;
2656 info->array.minor_version = -2;
e207da2f
AW
2657 devname = devnum2devname(st->container_dev);
2658 *info->text_version = '\0';
2659 if (devname)
2660 sprintf(info->text_version, "/%s/%d", devname, info->container_member);
2661 free(devname);
a67dd8cc 2662 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 2663 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N
2664
2665 if (dmap) {
2666 int i, j;
2667 for (i=0; i<map_disks; i++) {
2668 dmap[i] = 0;
2669 if (i < info->array.raid_disks) {
2670 struct imsm_disk *dsk;
238c0a71 2671 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N
2672 dsk = get_imsm_disk(super, j);
2673 if (dsk && (dsk->status & CONFIGURED_DISK))
2674 dmap[i] = 1;
2675 }
2676 }
2677 }
81ac8b4d 2678}
bf5a934a 2679
3b451610
AK
2680static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
2681 int failed, int look_in_map);
2682
2683static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
2684 int look_in_map);
2685
b4ab44d8
AK
2686
2687#ifndef MDASSEMBLE
3b451610
AK
2688static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
2689{
2690 if (is_gen_migration(dev)) {
2691 int failed;
2692 __u8 map_state;
2693 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
2694
2695 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 2696 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK
2697 if (map2->map_state != map_state) {
2698 map2->map_state = map_state;
2699 super->updates_pending++;
2700 }
2701 }
2702}
b4ab44d8 2703#endif
97b4d0e9
DW
2704
2705static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
2706{
2707 struct dl *d;
2708
2709 for (d = super->missing; d; d = d->next)
2710 if (d->index == index)
2711 return &d->disk;
2712 return NULL;
2713}
2714
a5d85af7 2715static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW
2716{
2717 struct intel_super *super = st->sb;
4f5bc454 2718 struct imsm_disk *disk;
a5d85af7 2719 int map_disks = info->array.raid_disks;
ab3cb6b3
N
2720 int max_enough = -1;
2721 int i;
2722 struct imsm_super *mpb;
4f5bc454 2723
bf5a934a 2724 if (super->current_vol >= 0) {
a5d85af7 2725 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW
2726 return;
2727 }
95eeceeb 2728 memset(info, 0, sizeof(*info));
d23fe947
DW
2729
2730 /* Set raid_disks to zero so that Assemble will always pull in valid
2731 * spares
2732 */
2733 info->array.raid_disks = 0;
cdddbdbc
DW
2734 info->array.level = LEVEL_CONTAINER;
2735 info->array.layout = 0;
2736 info->array.md_minor = -1;
c2c087e6 2737 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW
2738 info->array.utime = 0;
2739 info->array.chunk_size = 0;
2740
2741 info->disk.major = 0;
2742 info->disk.minor = 0;
cdddbdbc 2743 info->disk.raid_disk = -1;
c2c087e6 2744 info->reshape_active = 0;
f35f2525
N
2745 info->array.major_version = -1;
2746 info->array.minor_version = -2;
c2c087e6 2747 strcpy(info->text_version, "imsm");
a67dd8cc 2748 info->safe_mode_delay = 0;
c2c087e6
DW
2749 info->disk.number = -1;
2750 info->disk.state = 0;
c5afc314 2751 info->name[0] = 0;
921d9e16 2752 info->recovery_start = MaxSector;
3ad25638 2753 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
c2c087e6 2754
97b4d0e9 2755 /* do we have the all the insync disks that we expect? */
ab3cb6b3 2756 mpb = super->anchor;
97b4d0e9 2757
ab3cb6b3
N
2758 for (i = 0; i < mpb->num_raid_devs; i++) {
2759 struct imsm_dev *dev = get_imsm_dev(super, i);
2760 int failed, enough, j, missing = 0;
2761 struct imsm_map *map;
2762 __u8 state;
97b4d0e9 2763
3b451610
AK
2764 failed = imsm_count_failed(super, dev, MAP_0);
2765 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 2766 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N
2767
2768 /* any newly missing disks?
2769 * (catches single-degraded vs double-degraded)
2770 */
2771 for (j = 0; j < map->num_members; j++) {
238c0a71 2772 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N
2773 __u32 idx = ord_to_idx(ord);
2774
2775 if (!(ord & IMSM_ORD_REBUILD) &&
2776 get_imsm_missing(super, idx)) {
2777 missing = 1;
2778 break;
2779 }
97b4d0e9 2780 }
ab3cb6b3
N
2781
2782 if (state == IMSM_T_STATE_FAILED)
2783 enough = -1;
2784 else if (state == IMSM_T_STATE_DEGRADED &&
2785 (state != map->map_state || missing))
2786 enough = 0;
2787 else /* we're normal, or already degraded */
2788 enough = 1;
d2bde6d3
AK
2789 if (is_gen_migration(dev) && missing) {
2790 /* during general migration we need all disks
2791 * that process is running on.
2792 * No new missing disk is allowed.
2793 */
2794 max_enough = -1;
2795 enough = -1;
2796 /* no more checks necessary
2797 */
2798 break;
2799 }
ab3cb6b3
N
2800 /* in the missing/failed disk case check to see
2801 * if at least one array is runnable
2802 */
2803 max_enough = max(max_enough, enough);
2804 }
2805 dprintf("%s: enough: %d\n", __func__, max_enough);
2806 info->container_enough = max_enough;
97b4d0e9 2807
4a04ec6c 2808 if (super->disks) {
14e8215b
DW
2809 __u32 reserved = imsm_reserved_sectors(super, super->disks);
2810
b9f594fe 2811 disk = &super->disks->disk;
5551b113 2812 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 2813 info->component_size = reserved;
25ed7e59 2814 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW
2815 /* we don't change info->disk.raid_disk here because
2816 * this state will be finalized in mdmon after we have
2817 * found the 'most fresh' version of the metadata
2818 */
25ed7e59
DW
2819 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2820 info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
cdddbdbc 2821 }
a575e2a7
DW
2822
2823 /* only call uuid_from_super_imsm when this disk is part of a populated container,
2824 * ->compare_super may have updated the 'num_raid_devs' field for spares
2825 */
2826 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 2827 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC
2828 else
2829 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N
2830
2831 /* I don't know how to compute 'map' on imsm, so use safe default */
2832 if (map) {
2833 int i;
2834 for (i = 0; i < map_disks; i++)
2835 map[i] = 1;
2836 }
2837
cdddbdbc
DW
2838}
2839
5c4cd5da
AC
2840/* allocates memory and fills disk in mdinfo structure
2841 * for each disk in array */
2842struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
2843{
2844 struct mdinfo *mddev = NULL;
2845 struct intel_super *super = st->sb;
2846 struct imsm_disk *disk;
2847 int count = 0;
2848 struct dl *dl;
2849 if (!super || !super->disks)
2850 return NULL;
2851 dl = super->disks;
2852 mddev = malloc(sizeof(*mddev));
2853 if (!mddev) {
2854 fprintf(stderr, Name ": Failed to allocate memory.\n");
2855 return NULL;
2856 }
2857 memset(mddev, 0, sizeof(*mddev));
2858 while (dl) {
2859 struct mdinfo *tmp;
2860 disk = &dl->disk;
2861 tmp = malloc(sizeof(*tmp));
2862 if (!tmp) {
2863 fprintf(stderr, Name ": Failed to allocate memory.\n");
2864 if (mddev)
2865 sysfs_free(mddev);
2866 return NULL;
2867 }
2868 memset(tmp, 0, sizeof(*tmp));
2869 if (mddev->devs)
2870 tmp->next = mddev->devs;
2871 mddev->devs = tmp;
2872 tmp->disk.number = count++;
2873 tmp->disk.major = dl->major;
2874 tmp->disk.minor = dl->minor;
2875 tmp->disk.state = is_configured(disk) ?
2876 (1 << MD_DISK_ACTIVE) : 0;
2877 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2878 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
2879 tmp->disk.raid_disk = -1;
2880 dl = dl->next;
2881 }
2882 return mddev;
2883}
2884
cdddbdbc
DW
2885static int update_super_imsm(struct supertype *st, struct mdinfo *info,
2886 char *update, char *devname, int verbose,
2887 int uuid_set, char *homehost)
2888{
f352c545
DW
2889 /* For 'assemble' and 'force' we need to return non-zero if any
2890 * change was made. For others, the return value is ignored.
2891 * Update options are:
2892 * force-one : This device looks a bit old but needs to be included,
2893 * update age info appropriately.
2894 * assemble: clear any 'faulty' flag to allow this device to
2895 * be assembled.
2896 * force-array: Array is degraded but being forced, mark it clean
2897 * if that will be needed to assemble it.
2898 *
2899 * newdev: not used ????
2900 * grow: Array has gained a new device - this is currently for
2901 * linear only
2902 * resync: mark as dirty so a resync will happen.
2903 * name: update the name - preserving the homehost
6e46bf34 2904 * uuid: Change the uuid of the array to match watch is given
f352c545
DW
2905 *
2906 * Following are not relevant for this imsm:
2907 * sparc2.2 : update from old dodgey metadata
2908 * super-minor: change the preferred_minor number
2909 * summaries: update redundant counters.
f352c545
DW
2910 * homehost: update the recorded homehost
2911 * _reshape_progress: record new reshape_progress position.
2912 */
6e46bf34
DW
2913 int rv = 1;
2914 struct intel_super *super = st->sb;
2915 struct imsm_super *mpb;
f352c545 2916
6e46bf34
DW
2917 /* we can only update container info */
2918 if (!super || super->current_vol >= 0 || !super->anchor)
2919 return 1;
2920
2921 mpb = super->anchor;
2922
81a5b4f5
N
2923 if (strcmp(update, "uuid") == 0) {
2924 /* We take this to mean that the family_num should be updated.
2925 * However that is much smaller than the uuid so we cannot really
2926 * allow an explicit uuid to be given. And it is hard to reliably
2927 * know if one was.
2928 * So if !uuid_set we know the current uuid is random and just used
2929 * the first 'int' and copy it to the other 3 positions.
2930 * Otherwise we require the 4 'int's to be the same as would be the
2931 * case if we are using a random uuid. So an explicit uuid will be
2932 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 2933 */
81a5b4f5
N
2934 if (!uuid_set) {
2935 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 2936 rv = 0;
81a5b4f5
N
2937 } else {
2938 if (info->uuid[0] != info->uuid[1] ||
2939 info->uuid[1] != info->uuid[2] ||
2940 info->uuid[2] != info->uuid[3])
2941 rv = -1;
2942 else
2943 rv = 0;
6e46bf34 2944 }
81a5b4f5
N
2945 if (rv == 0)
2946 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW
2947 } else if (strcmp(update, "assemble") == 0)
2948 rv = 0;
2949 else
1e2b2765 2950 rv = -1;
f352c545 2951
6e46bf34
DW
2952 /* successful update? recompute checksum */
2953 if (rv == 0)
2954 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW
2955
2956 return rv;
cdddbdbc
DW
2957}
2958
c2c087e6 2959static size_t disks_to_mpb_size(int disks)
cdddbdbc 2960{
c2c087e6 2961 size_t size;
cdddbdbc 2962
c2c087e6
DW
2963 size = sizeof(struct imsm_super);
2964 size += (disks - 1) * sizeof(struct imsm_disk);
2965 size += 2 * sizeof(struct imsm_dev);
2966 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
2967 size += (4 - 2) * sizeof(struct imsm_map);
2968 /* 4 possible disk_ord_tbl's */
2969 size += 4 * (disks - 1) * sizeof(__u32);
2970
2971 return size;
2972}
2973
2974static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
2975{
2976 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
2977 return 0;
2978
2979 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW
2980}
2981
ba2de7ba
DW
2982static void free_devlist(struct intel_super *super)
2983{
2984 struct intel_dev *dv;
2985
2986 while (super->devlist) {
2987 dv = super->devlist->next;
2988 free(super->devlist->dev);
2989 free(super->devlist);
2990 super->devlist = dv;
2991 }
2992}
2993
2994static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
2995{
2996 memcpy(dest, src, sizeof_imsm_dev(src, 0));
2997}
2998
cdddbdbc
DW
2999static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3000{
3001 /*
3002 * return:
3003 * 0 same, or first was empty, and second was copied
3004 * 1 second had wrong number
3005 * 2 wrong uuid
3006 * 3 wrong other info
3007 */
3008 struct intel_super *first = st->sb;
3009 struct intel_super *sec = tst->sb;
3010
3011 if (!first) {
3012 st->sb = tst->sb;
3013 tst->sb = NULL;
3014 return 0;
3015 }
8603ea6f
LM
3016 /* in platform dependent environment test if the disks
3017 * use the same Intel hba
3018 */
3019 if (!check_env("IMSM_NO_PLATFORM")) {
ea2bc72b
LM
3020 if (!first->hba || !sec->hba ||
3021 (first->hba->type != sec->hba->type)) {
8603ea6f
LM
3022 fprintf(stderr,
3023 "HBAs of devices does not match %s != %s\n",
ea2bc72b
LM
3024 first->hba ? get_sys_dev_type(first->hba->type) : NULL,
3025 sec->hba ? get_sys_dev_type(sec->hba->type) : NULL);
8603ea6f
LM
3026 return 3;
3027 }
3028 }
cdddbdbc 3029
d23fe947
DW
3030 /* if an anchor does not have num_raid_devs set then it is a free
3031 * floating spare
3032 */
3033 if (first->anchor->num_raid_devs > 0 &&
3034 sec->anchor->num_raid_devs > 0) {
a2b97981
DW
3035 /* Determine if these disks might ever have been
3036 * related. Further disambiguation can only take place
3037 * in load_super_imsm_all
3038 */
3039 __u32 first_family = first->anchor->orig_family_num;
3040 __u32 sec_family = sec->anchor->orig_family_num;
3041
f796af5d
DW
3042 if (memcmp(first->anchor->sig, sec->anchor->sig,
3043 MAX_SIGNATURE_LENGTH) != 0)
3044 return 3;
3045
a2b97981
DW
3046 if (first_family == 0)
3047 first_family = first->anchor->family_num;
3048 if (sec_family == 0)
3049 sec_family = sec->anchor->family_num;
3050
3051 if (first_family != sec_family)
d23fe947 3052 return 3;
f796af5d 3053
d23fe947 3054 }
cdddbdbc 3055
f796af5d 3056
3e372e5a
DW
3057 /* if 'first' is a spare promote it to a populated mpb with sec's
3058 * family number
3059 */
3060 if (first->anchor->num_raid_devs == 0 &&
3061 sec->anchor->num_raid_devs > 0) {
78d30f94 3062 int i;
ba2de7ba
DW
3063 struct intel_dev *dv;
3064 struct imsm_dev *dev;
78d30f94
DW
3065
3066 /* we need to copy raid device info from sec if an allocation
3067 * fails here we don't associate the spare
3068 */
3069 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
ba2de7ba
DW
3070 dv = malloc(sizeof(*dv));
3071 if (!dv)
3072 break;
3073 dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
3074 if (!dev) {
3075 free(dv);
3076 break;
78d30f94 3077 }
ba2de7ba
DW
3078 dv->dev = dev;
3079 dv->index = i;
3080 dv->next = first->devlist;
3081 first->devlist = dv;
78d30f94 3082 }
709743c5 3083 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW
3084 /* allocation failure */
3085 free_devlist(first);
3086 fprintf(stderr, "imsm: failed to associate spare\n");
3087 return 3;
78d30f94 3088 }
3e372e5a 3089 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3090 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3091 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3092 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW
3093 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3094 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW
3095 }
3096
cdddbdbc
DW
3097 return 0;
3098}
3099
0030e8d6
DW
3100static void fd2devname(int fd, char *name)
3101{
3102 struct stat st;
3103 char path[256];
33a6535d 3104 char dname[PATH_MAX];
0030e8d6
DW
3105 char *nm;
3106 int rv;
3107
3108 name[0] = '\0';
3109 if (fstat(fd, &st) != 0)
3110 return;
3111 sprintf(path, "/sys/dev/block/%d:%d",
3112 major(st.st_rdev), minor(st.st_rdev));
3113
9cf014ec 3114 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW
3115 if (rv <= 0)
3116 return;
9587c373 3117
0030e8d6
DW
3118 dname[rv] = '\0';
3119 nm = strrchr(dname, '/');
7897de29
JS
3120 if (nm) {
3121 nm++;
3122 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3123 }
0030e8d6
DW
3124}
3125
cdddbdbc
DW
3126extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3127
3128static int imsm_read_serial(int fd, char *devname,
3129 __u8 serial[MAX_RAID_SERIAL_LEN])
3130{
3131 unsigned char scsi_serial[255];
cdddbdbc
DW
3132 int rv;
3133 int rsp_len;
1f24f035 3134 int len;
316e2bf4
DW
3135 char *dest;
3136 char *src;
3137 char *rsp_buf;
3138 int i;
cdddbdbc
DW
3139
3140 memset(scsi_serial, 0, sizeof(scsi_serial));
cdddbdbc 3141
f9ba0ff1
DW
3142 rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
3143
40ebbb9c 3144 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW
3145 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3146 fd2devname(fd, (char *) serial);
0030e8d6
DW
3147 return 0;
3148 }
3149
cdddbdbc
DW
3150 if (rv != 0) {
3151 if (devname)
3152 fprintf(stderr,
3153 Name ": Failed to retrieve serial for %s\n",
3154 devname);
3155 return rv;
3156 }
3157
3158 rsp_len = scsi_serial[3];
03cd4cc8
DW
3159 if (!rsp_len) {
3160 if (devname)
3161 fprintf(stderr,
3162 Name ": Failed to retrieve serial for %s\n",
3163 devname);
3164 return 2;
3165 }
1f24f035 3166 rsp_buf = (char *) &scsi_serial[4];
5c3db629 3167
316e2bf4
DW
3168 /* trim all whitespace and non-printable characters and convert
3169 * ':' to ';'
3170 */
3171 for (i = 0, dest = rsp_buf; i < rsp_len; i++) {
3172 src = &rsp_buf[i];
3173 if (*src > 0x20) {
3174 /* ':' is reserved for use in placeholder serial
3175 * numbers for missing disks
3176 */
3177 if (*src == ':')
3178 *dest++ = ';';
3179 else
3180 *dest++ = *src;
3181 }
3182 }
3183 len = dest - rsp_buf;
3184 dest = rsp_buf;
3185
3186 /* truncate leading characters */
3187 if (len > MAX_RAID_SERIAL_LEN) {
3188 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 3189 len = MAX_RAID_SERIAL_LEN;
316e2bf4 3190 }
5c3db629 3191
5c3db629 3192 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 3193 memcpy(serial, dest, len);
cdddbdbc
DW
3194
3195 return 0;
3196}
3197
1f24f035
DW
3198static int serialcmp(__u8 *s1, __u8 *s2)
3199{
3200 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
3201}
3202
3203static void serialcpy(__u8 *dest, __u8 *src)
3204{
3205 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
3206}
3207
54c2c1ea
DW
3208static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
3209{
3210 struct dl *dl;
3211
3212 for (dl = super->disks; dl; dl = dl->next)
3213 if (serialcmp(dl->serial, serial) == 0)
3214 break;
3215
3216 return dl;
3217}
3218
a2b97981
DW
3219static struct imsm_disk *
3220__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
3221{
3222 int i;
3223
3224 for (i = 0; i < mpb->num_disks; i++) {
3225 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
3226
3227 if (serialcmp(disk->serial, serial) == 0) {
3228 if (idx)
3229 *idx = i;
3230 return disk;
3231 }
3232 }
3233
3234 return NULL;
3235}
3236
cdddbdbc
DW
3237static int
3238load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
3239{
a2b97981 3240 struct imsm_disk *disk;
cdddbdbc
DW
3241 struct dl *dl;
3242 struct stat stb;
cdddbdbc 3243 int rv;
a2b97981 3244 char name[40];
d23fe947
DW
3245 __u8 serial[MAX_RAID_SERIAL_LEN];
3246
3247 rv = imsm_read_serial(fd, devname, serial);
3248
3249 if (rv != 0)
3250 return 2;
3251
a2b97981 3252 dl = calloc(1, sizeof(*dl));
b9f594fe 3253 if (!dl) {
cdddbdbc
DW
3254 if (devname)
3255 fprintf(stderr,
3256 Name ": failed to allocate disk buffer for %s\n",
3257 devname);
3258 return 2;
3259 }
cdddbdbc 3260
a2b97981
DW
3261 fstat(fd, &stb);
3262 dl->major = major(stb.st_rdev);
3263 dl->minor = minor(stb.st_rdev);
3264 dl->next = super->disks;
3265 dl->fd = keep_fd ? fd : -1;
3266 assert(super->disks == NULL);
3267 super->disks = dl;
3268 serialcpy(dl->serial, serial);
3269 dl->index = -2;
3270 dl->e = NULL;
3271 fd2devname(fd, name);
3272 if (devname)
3273 dl->devname = strdup(devname);
3274 else
3275 dl->devname = strdup(name);
cdddbdbc 3276
d23fe947 3277 /* look up this disk's index in the current anchor */
a2b97981
DW
3278 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
3279 if (disk) {
3280 dl->disk = *disk;
3281 /* only set index on disks that are a member of a
3282 * populated contianer, i.e. one with raid_devs
3283 */
3284 if (is_failed(&dl->disk))
3f6efecc 3285 dl->index = -2;
a2b97981
DW
3286 else if (is_spare(&dl->disk))
3287 dl->index = -1;
3f6efecc
DW
3288 }
3289
949c47a0
DW
3290 return 0;
3291}
3292
0e600426 3293#ifndef MDASSEMBLE
0c046afd
DW
3294/* When migrating map0 contains the 'destination' state while map1
3295 * contains the current state. When not migrating map0 contains the
3296 * current state. This routine assumes that map[0].map_state is set to
3297 * the current array state before being called.
3298 *
3299 * Migration is indicated by one of the following states
3300 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 3301 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 3302 * map1state=unitialized)
1484e727 3303 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 3304 * map1state=normal)
e3bba0e0 3305 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 3306 * map1state=degraded)
8e59f3d8
AK
3307 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
3308 * map1state=normal)
0c046afd 3309 */
8e59f3d8
AK
3310static void migrate(struct imsm_dev *dev, struct intel_super *super,
3311 __u8 to_state, int migr_type)
3393c6af 3312{
0c046afd 3313 struct imsm_map *dest;
238c0a71 3314 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 3315
0c046afd 3316 dev->vol.migr_state = 1;
1484e727 3317 set_migr_type(dev, migr_type);
f8f603f1 3318 dev->vol.curr_migr_unit = 0;
238c0a71 3319 dest = get_imsm_map(dev, MAP_1);
0c046afd 3320
0556e1a2 3321 /* duplicate and then set the target end state in map[0] */
3393c6af 3322 memcpy(dest, src, sizeof_imsm_map(src));
28bce06f
AK
3323 if ((migr_type == MIGR_REBUILD) ||
3324 (migr_type == MIGR_GEN_MIGR)) {
0556e1a2
DW
3325 __u32 ord;
3326 int i;
3327
3328 for (i = 0; i < src->num_members; i++) {
3329 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
3330 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
3331 }
3332 }
3333
8e59f3d8
AK
3334 if (migr_type == MIGR_GEN_MIGR)
3335 /* Clear migration record */
3336 memset(super->migr_rec, 0, sizeof(struct migr_record));
3337
0c046afd 3338 src->map_state = to_state;
949c47a0 3339}
f8f603f1 3340
809da78e
AK
3341static void end_migration(struct imsm_dev *dev, struct intel_super *super,
3342 __u8 map_state)
f8f603f1 3343{
238c0a71
AK
3344 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3345 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
3346 MAP_0 : MAP_1);
28bce06f 3347 int i, j;
0556e1a2
DW
3348
3349 /* merge any IMSM_ORD_REBUILD bits that were not successfully
3350 * completed in the last migration.
3351 *
28bce06f 3352 * FIXME add support for raid-level-migration
0556e1a2 3353 */
809da78e
AK
3354 if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) &&
3355 (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) {
3356 /* when final map state is other than expected
3357 * merge maps (not for migration)
3358 */
3359 int failed;
3360
3361 for (i = 0; i < prev->num_members; i++)
3362 for (j = 0; j < map->num_members; j++)
3363 /* during online capacity expansion
3364 * disks position can be changed
3365 * if takeover is used
3366 */
3367 if (ord_to_idx(map->disk_ord_tbl[j]) ==
3368 ord_to_idx(prev->disk_ord_tbl[i])) {
3369 map->disk_ord_tbl[j] |=
3370 prev->disk_ord_tbl[i];
3371 break;
3372 }
3373 failed = imsm_count_failed(super, dev, MAP_0);
3374 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
3375 }
f8f603f1
DW
3376
3377 dev->vol.migr_state = 0;
ea672ee1 3378 set_migr_type(dev, 0);
f8f603f1
DW
3379 dev->vol.curr_migr_unit = 0;
3380 map->map_state = map_state;
3381}
0e600426 3382#endif
949c47a0
DW
3383
3384static int parse_raid_devices(struct intel_super *super)
3385{
3386 int i;
3387 struct imsm_dev *dev_new;
4d7b1503 3388 size_t len, len_migr;
401d313b 3389 size_t max_len = 0;
4d7b1503
DW
3390 size_t space_needed = 0;
3391 struct imsm_super *mpb = super->anchor;
949c47a0
DW
3392
3393 for (i = 0; i < super->anchor->num_raid_devs; i++) {
3394 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 3395 struct intel_dev *dv;
949c47a0 3396
4d7b1503
DW
3397 len = sizeof_imsm_dev(dev_iter, 0);
3398 len_migr = sizeof_imsm_dev(dev_iter, 1);
3399 if (len_migr > len)
3400 space_needed += len_migr - len;
ca9de185 3401
ba2de7ba
DW
3402 dv = malloc(sizeof(*dv));
3403 if (!dv)
3404 return 1;
401d313b
AK
3405 if (max_len < len_migr)
3406 max_len = len_migr;
3407 if (max_len > len_migr)
3408 space_needed += max_len - len_migr;
3409 dev_new = malloc(max_len);
ba2de7ba
DW
3410 if (!dev_new) {
3411 free(dv);
949c47a0 3412 return 1;
ba2de7ba 3413 }
949c47a0 3414 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW
3415 dv->dev = dev_new;
3416 dv->index = i;
3417 dv->next = super->devlist;
3418 super->devlist = dv;
949c47a0 3419 }
cdddbdbc 3420
4d7b1503
DW
3421 /* ensure that super->buf is large enough when all raid devices
3422 * are migrating
3423 */
3424 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
3425 void *buf;
3426
3427 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512);
3428 if (posix_memalign(&buf, 512, len) != 0)
3429 return 1;
3430
1f45a8ad
DW
3431 memcpy(buf, super->buf, super->len);
3432 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW
3433 free(super->buf);
3434 super->buf = buf;
3435 super->len = len;
3436 }
ca9de185 3437
cdddbdbc
DW
3438 return 0;
3439}
3440
604b746f
JD
3441/* retrieve a pointer to the bbm log which starts after all raid devices */
3442struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb)
3443{
3444 void *ptr = NULL;
3445
3446 if (__le32_to_cpu(mpb->bbm_log_size)) {
3447 ptr = mpb;
3448 ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size);
3449 }
3450
3451 return ptr;
3452}
3453
e2f41b2c
AK
3454/*******************************************************************************
3455 * Function: check_mpb_migr_compatibility
3456 * Description: Function checks for unsupported migration features:
3457 * - migration optimization area (pba_of_lba0)
3458 * - descending reshape (ascending_migr)
3459 * Parameters:
3460 * super : imsm metadata information
3461 * Returns:
3462 * 0 : migration is compatible
3463 * -1 : migration is not compatible
3464 ******************************************************************************/
3465int check_mpb_migr_compatibility(struct intel_super *super)
3466{
3467 struct imsm_map *map0, *map1;
3468 struct migr_record *migr_rec = super->migr_rec;
3469 int i;
3470
3471 for (i = 0; i < super->anchor->num_raid_devs; i++) {
3472 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
3473
3474 if (dev_iter &&
3475 dev_iter->vol.migr_state == 1 &&
3476 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
3477 /* This device is migrating */
238c0a71
AK
3478 map0 = get_imsm_map(dev_iter, MAP_0);
3479 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 3480 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK
3481 /* migration optimization area was used */
3482 return -1;
3483 if (migr_rec->ascending_migr == 0
3484 && migr_rec->dest_depth_per_unit > 0)
3485 /* descending reshape not supported yet */
3486 return -1;
3487 }
3488 }
3489 return 0;
3490}
3491
d23fe947 3492static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 3493
cdddbdbc 3494/* load_imsm_mpb - read matrix metadata
f2f5c343 3495 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW
3496 */
3497static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
3498{
3499 unsigned long long dsize;
cdddbdbc
DW
3500 unsigned long long sectors;
3501 struct stat;
6416d527 3502 struct imsm_super *anchor;
cdddbdbc
DW
3503 __u32 check_sum;
3504
cdddbdbc 3505 get_dev_size(fd, NULL, &dsize);
64436f06
N
3506 if (dsize < 1024) {
3507 if (devname)
3508 fprintf(stderr,
3509 Name ": %s: device to small for imsm\n",
3510 devname);
3511 return 1;
3512 }
cdddbdbc
DW
3513
3514 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
3515 if (devname)
2e062e82
AK
3516 fprintf(stderr, Name
3517 ": Cannot seek to anchor block on %s: %s\n",
cdddbdbc
DW
3518 devname, strerror(errno));
3519 return 1;
3520 }
3521
949c47a0 3522 if (posix_memalign((void**)&anchor, 512, 512) != 0) {
ad97895e
DW
3523 if (devname)
3524 fprintf(stderr,
3525 Name ": Failed to allocate imsm anchor buffer"
3526 " on %s\n", devname);
3527 return 1;
3528 }
949c47a0 3529 if (read(fd, anchor, 512) != 512) {
cdddbdbc
DW
3530 if (devname)
3531 fprintf(stderr,
3532 Name ": Cannot read anchor block on %s: %s\n",
3533 devname, strerror(errno));
6416d527 3534 free(anchor);
cdddbdbc
DW
3535 return 1;
3536 }
3537
6416d527 3538 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc
DW
3539 if (devname)
3540 fprintf(stderr,
3541 Name ": no IMSM anchor on %s\n", devname);
6416d527 3542 free(anchor);
cdddbdbc
DW
3543 return 2;
3544 }
3545
d23fe947 3546 __free_imsm(super, 0);
f2f5c343
LM
3547 /* reload capability and hba */
3548
3549 /* capability and hba must be updated with new super allocation */
d424212e 3550 find_intel_hba_capability(fd, super, devname);
949c47a0
DW
3551 super->len = ROUND_UP(anchor->mpb_size, 512);
3552 if (posix_memalign(&super->buf, 512, super->len) != 0) {
cdddbdbc
DW
3553 if (devname)
3554 fprintf(stderr,
3555 Name ": unable to allocate %zu byte mpb buffer\n",
949c47a0 3556 super->len);
6416d527 3557 free(anchor);
cdddbdbc
DW
3558 return 2;
3559 }
949c47a0 3560 memcpy(super->buf, anchor, 512);
cdddbdbc 3561
6416d527
NB
3562 sectors = mpb_sectors(anchor) - 1;
3563 free(anchor);
8e59f3d8 3564
17a4eaf9 3565 if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) {
8e59f3d8
AK
3566 fprintf(stderr, Name
3567 ": %s could not allocate migr_rec buffer\n", __func__);
3568 free(super->buf);
3569 return 2;
3570 }
51d83f5d 3571 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 3572
949c47a0 3573 if (!sectors) {
ecf45690
DW
3574 check_sum = __gen_imsm_checksum(super->anchor);
3575 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
3576 if (devname)
3577 fprintf(stderr,
3578 Name ": IMSM checksum %x != %x on %s\n",
3579 check_sum,
3580 __le32_to_cpu(super->anchor->check_sum),
3581 devname);
3582 return 2;
3583 }
3584
a2b97981 3585 return 0;
949c47a0 3586 }
cdddbdbc
DW
3587
3588 /* read the extended mpb */
3589 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
3590 if (devname)
3591 fprintf(stderr,
3592 Name ": Cannot seek to extended mpb on %s: %s\n",
3593 devname, strerror(errno));
3594 return 1;
3595 }
3596
f21e18ca 3597 if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) {
cdddbdbc
DW
3598 if (devname)
3599 fprintf(stderr,
3600 Name ": Cannot read extended mpb on %s: %s\n",
3601 devname, strerror(errno));
3602 return 2;
3603 }
3604
949c47a0
DW
3605 check_sum = __gen_imsm_checksum(super->anchor);
3606 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc
DW
3607 if (devname)
3608 fprintf(stderr,
3609 Name ": IMSM checksum %x != %x on %s\n",
949c47a0 3610 check_sum, __le32_to_cpu(super->anchor->check_sum),
cdddbdbc 3611 devname);
db575f3b 3612 return 3;
cdddbdbc
DW
3613 }
3614
604b746f
JD
3615 /* FIXME the BBM log is disk specific so we cannot use this global
3616 * buffer for all disks. Ok for now since we only look at the global
3617 * bbm_log_size parameter to gate assembly
3618 */
3619 super->bbm_log = __get_imsm_bbm_log(super->anchor);
3620
a2b97981
DW
3621 return 0;
3622}
3623
8e59f3d8
AK
3624static int read_imsm_migr_rec(int fd, struct intel_super *super);
3625
97f81ee2
CA
3626/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
3627static void clear_hi(struct intel_super *super)
3628{
3629 struct imsm_super *mpb = super->anchor;
3630 int i, n;
3631 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
3632 return;
3633 for (i = 0; i < mpb->num_disks; ++i) {
3634 struct imsm_disk *disk = &mpb->disk[i];
3635 disk->total_blocks_hi = 0;
3636 }
3637 for (i = 0; i < mpb->num_raid_devs; ++i) {
3638 struct imsm_dev *dev = get_imsm_dev(super, i);
3639 if (!dev)
3640 return;
3641 for (n = 0; n < 2; ++n) {
3642 struct imsm_map *map = get_imsm_map(dev, n);
3643 if (!map)
3644 continue;
3645 map->pba_of_lba0_hi = 0;
3646 map->blocks_per_member_hi = 0;
3647 map->num_data_stripes_hi = 0;
3648 }
3649 }
3650}
3651
a2b97981
DW
3652static int
3653load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
3654{
3655 int err;
3656
3657 err = load_imsm_mpb(fd, super, devname);
3658 if (err)
3659 return err;
3660 err = load_imsm_disk(fd, super, devname, keep_fd);
3661 if (err)
3662 return err;
3663 err = parse_raid_devices(super);
97f81ee2 3664 clear_hi(super);
a2b97981 3665 return err;
cdddbdbc
DW
3666}
3667
ae6aad82
DW
3668static void __free_imsm_disk(struct dl *d)
3669{
3670 if (d->fd >= 0)
3671 close(d->fd);
3672 if (d->devname)
3673 free(d->devname);
0dcecb2e
DW
3674 if (d->e)
3675 free(d->e);
ae6aad82
DW
3676 free(d);
3677
3678}
1a64be56 3679
cdddbdbc
DW
3680static void free_imsm_disks(struct intel_super *super)
3681{
47ee5a45 3682 struct dl *d;
cdddbdbc 3683
47ee5a45
DW
3684 while (super->disks) {
3685 d = super->disks;
cdddbdbc 3686 super->disks = d->next;
ae6aad82 3687 __free_imsm_disk(d);
cdddbdbc 3688 }
cb82edca
AK
3689 while (super->disk_mgmt_list) {
3690 d = super->disk_mgmt_list;
3691 super->disk_mgmt_list = d->next;
3692 __free_imsm_disk(d);
3693 }
47ee5a45
DW
3694 while (super->missing) {
3695 d = super->missing;
3696 super->missing = d->next;
3697 __free_imsm_disk(d);
3698 }
3699
cdddbdbc
DW
3700}
3701
9ca2c81c 3702/* free all the pieces hanging off of a super pointer */
d23fe947 3703static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 3704{
88654014
LM
3705 struct intel_hba *elem, *next;
3706
9ca2c81c 3707 if (super->buf) {
949c47a0 3708 free(super->buf);
9ca2c81c
DW
3709 super->buf = NULL;
3710 }
f2f5c343
LM
3711 /* unlink capability description */
3712 super->orom = NULL;
8e59f3d8
AK
3713 if (super->migr_rec_buf) {
3714 free(super->migr_rec_buf);
3715 super->migr_rec_buf = NULL;
3716 }
d23fe947
DW
3717 if (free_disks)
3718 free_imsm_disks(super);
ba2de7ba 3719 free_devlist(super);
88654014
LM
3720 elem = super->hba;
3721 while (elem) {
3722 if (elem->path)
3723 free((void *)elem->path);
3724 next = elem->next;
3725 free(elem);
3726 elem = next;
88c32bb1 3727 }
88654014 3728 super->hba = NULL;
cdddbdbc
DW
3729}
3730
9ca2c81c
DW
3731static void free_imsm(struct intel_super *super)
3732{
d23fe947 3733 __free_imsm(super, 1);
9ca2c81c
DW
3734 free(super);
3735}
cdddbdbc
DW
3736
3737static void free_super_imsm(struct supertype *st)
3738{
3739 struct intel_super *super = st->sb;
3740
3741 if (!super)
3742 return;
3743
3744 free_imsm(super);
3745 st->sb = NULL;
3746}
3747
49133e57 3748static struct intel_super *alloc_super(void)
c2c087e6
DW
3749{
3750 struct intel_super *super = malloc(sizeof(*super));
3751
3752 if (super) {
3753 memset(super, 0, sizeof(*super));
bf5a934a 3754 super->current_vol = -1;
5551b113 3755 super->create_offset = ~((unsigned long long) 0);
c2c087e6 3756 }
c2c087e6
DW
3757 return super;
3758}
3759
f0f5a016
LM
3760/*
3761 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
3762 */
d424212e 3763static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM
3764{
3765 struct sys_dev *hba_name;
3766 int rv = 0;
3767
3768 if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 3769 super->orom = NULL;
f0f5a016
LM
3770 super->hba = NULL;
3771 return 0;
3772 }
3773 hba_name = find_disk_attached_hba(fd, NULL);
3774 if (!hba_name) {
d424212e 3775 if (devname)
f0f5a016
LM
3776 fprintf(stderr,
3777 Name ": %s is not attached to Intel(R) RAID controller.\n",
d424212e 3778 devname);
f0f5a016
LM
3779 return 1;
3780 }
3781 rv = attach_hba_to_super(super, hba_name);
3782 if (rv == 2) {
d424212e
N
3783 if (devname) {
3784 struct intel_hba *hba = super->hba;
f0f5a016 3785
f0f5a016
LM
3786 fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID "
3787 "controller (%s),\n"
3788 " but the container is assigned to Intel(R) "
3789 "%s RAID controller (",
d424212e 3790 devname,
f0f5a016
LM
3791 hba_name->path,
3792 hba_name->pci_id ? : "Err!",
3793 get_sys_dev_type(hba_name->type));
3794
f0f5a016
LM
3795 while (hba) {
3796 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
3797 if (hba->next)
3798 fprintf(stderr, ", ");
3799 hba = hba->next;
3800 }
3801
3802 fprintf(stderr, ").\n"
3803 " Mixing devices attached to different controllers "
3804 "is not allowed.\n");
3805 }
3806 free_sys_dev(&hba_name);
3807 return 2;
3808 }
f2f5c343 3809 super->orom = find_imsm_capability(hba_name->type);
f0f5a016 3810 free_sys_dev(&hba_name);
f2f5c343
LM
3811 if (!super->orom)
3812 return 3;
f0f5a016
LM
3813 return 0;
3814}
3815
47ee5a45
DW
3816/* find_missing - helper routine for load_super_imsm_all that identifies
3817 * disks that have disappeared from the system. This routine relies on
3818 * the mpb being uptodate, which it is at load time.
3819 */
3820static int find_missing(struct intel_super *super)
3821{
3822 int i;
3823 struct imsm_super *mpb = super->anchor;
3824 struct dl *dl;
3825 struct imsm_disk *disk;
47ee5a45
DW
3826
3827 for (i = 0; i < mpb->num_disks; i++) {
3828 disk = __get_imsm_disk(mpb, i);
54c2c1ea 3829 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW
3830 if (dl)
3831 continue;
47ee5a45
DW
3832
3833 dl = malloc(sizeof(*dl));
3834 if (!dl)
3835 return 1;
3836 dl->major = 0;
3837 dl->minor = 0;
3838 dl->fd = -1;
3839 dl->devname = strdup("missing");
3840 dl->index = i;
3841 serialcpy(dl->serial, disk->serial);
3842 dl->disk = *disk;
689c9bf3 3843 dl->e = NULL;
47ee5a45
DW
3844 dl->next = super->missing;
3845 super->missing = dl;
3846 }
3847
3848 return 0;
3849}
3850
3960e579 3851#ifndef MDASSEMBLE
a2b97981
DW
3852static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
3853{
3854 struct intel_disk *idisk = disk_list;
3855
3856 while (idisk) {
3857 if (serialcmp(idisk->disk.serial, serial) == 0)
3858 break;
3859 idisk = idisk->next;
3860 }
3861
3862 return idisk;
3863}
3864
3865static int __prep_thunderdome(struct intel_super **table, int tbl_size,
3866 struct intel_super *super,
3867 struct intel_disk **disk_list)
3868{
3869 struct imsm_disk *d = &super->disks->disk;
3870 struct imsm_super *mpb = super->anchor;
3871 int i, j;
3872
3873 for (i = 0; i < tbl_size; i++) {
3874 struct imsm_super *tbl_mpb = table[i]->anchor;
3875 struct imsm_disk *tbl_d = &table[i]->disks->disk;
3876
3877 if (tbl_mpb->family_num == mpb->family_num) {
3878 if (tbl_mpb->check_sum == mpb->check_sum) {
3879 dprintf("%s: mpb from %d:%d matches %d:%d\n",
3880 __func__, super->disks->major,
3881 super->disks->minor,
3882 table[i]->disks->major,
3883 table[i]->disks->minor);
3884 break;
3885 }
3886
3887 if (((is_configured(d) && !is_configured(tbl_d)) ||
3888 is_configured(d) == is_configured(tbl_d)) &&
3889 tbl_mpb->generation_num < mpb->generation_num) {
3890 /* current version of the mpb is a
3891 * better candidate than the one in
3892 * super_table, but copy over "cross
3893 * generational" status
3894 */
3895 struct intel_disk *idisk;
3896
3897 dprintf("%s: mpb from %d:%d replaces %d:%d\n",
3898 __func__, super->disks->major,
3899 super->disks->minor,
3900 table[i]->disks->major,
3901 table[i]->disks->minor);
3902
3903 idisk = disk_list_get(tbl_d->serial, *disk_list);
3904 if (idisk && is_failed(&idisk->disk))
3905 tbl_d->status |= FAILED_DISK;
3906 break;
3907 } else {
3908 struct intel_disk *idisk;
3909 struct imsm_disk *disk;
3910
3911 /* tbl_mpb is more up to date, but copy
3912 * over cross generational status before
3913 * returning
3914 */
3915 disk = __serial_to_disk(d->serial, mpb, NULL);
3916 if (disk && is_failed(disk))
3917 d->status |= FAILED_DISK;
3918
3919 idisk = disk_list_get(d->serial, *disk_list);
3920 if (idisk) {
3921 idisk->owner = i;
3922 if (disk && is_configured(disk))
3923 idisk->disk.status |= CONFIGURED_DISK;
3924 }
3925
3926 dprintf("%s: mpb from %d:%d prefer %d:%d\n",
3927 __func__, super->disks->major,
3928 super->disks->minor,
3929 table[i]->disks->major,
3930 table[i]->disks->minor);
3931
3932 return tbl_size;
3933 }
3934 }
3935 }
3936
3937 if (i >= tbl_size)
3938 table[tbl_size++] = super;
3939 else
3940 table[i] = super;
3941
3942 /* update/extend the merged list of imsm_disk records */
3943 for (j = 0; j < mpb->num_disks; j++) {
3944 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
3945 struct intel_disk *idisk;
3946
3947 idisk = disk_list_get(disk->serial, *disk_list);
3948 if (idisk) {
3949 idisk->disk.status |= disk->status;
3950 if (is_configured(&idisk->disk) ||
3951 is_failed(&idisk->disk))
3952 idisk->disk.status &= ~(SPARE_DISK);
3953 } else {
3954 idisk = calloc(1, sizeof(*idisk));
3955 if (!idisk)
3956 return -1;
3957 idisk->owner = IMSM_UNKNOWN_OWNER;
3958 idisk->disk = *disk;
3959 idisk->next = *disk_list;
3960 *disk_list = idisk;
3961 }
3962
3963 if (serialcmp(idisk->disk.serial, d->serial) == 0)
3964 idisk->owner = i;
3965 }
3966
3967 return tbl_size;
3968}
3969
3970static struct intel_super *
3971validate_members(struct intel_super *super, struct intel_disk *disk_list,
3972 const int owner)
3973{
3974 struct imsm_super *mpb = super->anchor;
3975 int ok_count = 0;
3976 int i;
3977
3978 for (i = 0; i < mpb->num_disks; i++) {
3979 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
3980 struct intel_disk *idisk;
3981
3982 idisk = disk_list_get(disk->serial, disk_list);
3983 if (idisk) {
3984 if (idisk->owner == owner ||
3985 idisk->owner == IMSM_UNKNOWN_OWNER)
3986 ok_count++;
3987 else
3988 dprintf("%s: '%.16s' owner %d != %d\n",
3989 __func__, disk->serial, idisk->owner,
3990 owner);
3991 } else {
3992 dprintf("%s: unknown disk %x [%d]: %.16s\n",
3993 __func__, __le32_to_cpu(mpb->family_num), i,
3994 disk->serial);
3995 break;
3996 }
3997 }
3998
3999 if (ok_count == mpb->num_disks)
4000 return super;
4001 return NULL;
4002}
4003
4004static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4005{
4006 struct intel_super *s;
4007
4008 for (s = super_list; s; s = s->next) {
4009 if (family_num != s->anchor->family_num)
4010 continue;
4011 fprintf(stderr, "Conflict, offlining family %#x on '%s'\n",
4012 __le32_to_cpu(family_num), s->disks->devname);
4013 }
4014}
4015
4016static struct intel_super *
4017imsm_thunderdome(struct intel_super **super_list, int len)
4018{
4019 struct intel_super *super_table[len];
4020 struct intel_disk *disk_list = NULL;
4021 struct intel_super *champion, *spare;
4022 struct intel_super *s, **del;
4023 int tbl_size = 0;
4024 int conflict;
4025 int i;
4026
4027 memset(super_table, 0, sizeof(super_table));
4028 for (s = *super_list; s; s = s->next)
4029 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4030
4031 for (i = 0; i < tbl_size; i++) {
4032 struct imsm_disk *d;
4033 struct intel_disk *idisk;
4034 struct imsm_super *mpb = super_table[i]->anchor;
4035
4036 s = super_table[i];
4037 d = &s->disks->disk;
4038
4039 /* 'd' must appear in merged disk list for its
4040 * configuration to be valid
4041 */
4042 idisk = disk_list_get(d->serial, disk_list);
4043 if (idisk && idisk->owner == i)
4044 s = validate_members(s, disk_list, i);
4045 else
4046 s = NULL;
4047
4048 if (!s)
4049 dprintf("%s: marking family: %#x from %d:%d offline\n",
4050 __func__, mpb->family_num,
4051 super_table[i]->disks->major,
4052 super_table[i]->disks->minor);
4053 super_table[i] = s;
4054 }
4055
4056 /* This is where the mdadm implementation differs from the Windows
4057 * driver which has no strict concept of a container. We can only
4058 * assemble one family from a container, so when returning a prodigal
4059 * array member to this system the code will not be able to disambiguate
4060 * the container contents that should be assembled ("foreign" versus
4061 * "local"). It requires user intervention to set the orig_family_num
4062 * to a new value to establish a new container. The Windows driver in
4063 * this situation fixes up the volume name in place and manages the
4064 * foreign array as an independent entity.
4065 */
4066 s = NULL;
4067 spare = NULL;
4068 conflict = 0;
4069 for (i = 0; i < tbl_size; i++) {
4070 struct intel_super *tbl_ent = super_table[i];
4071 int is_spare = 0;
4072
4073 if (!tbl_ent)
4074 continue;
4075
4076 if (tbl_ent->anchor->num_raid_devs == 0) {
4077 spare = tbl_ent;
4078 is_spare = 1;
4079 }
4080
4081 if (s && !is_spare) {
4082 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4083 conflict++;
4084 } else if (!s && !is_spare)
4085 s = tbl_ent;
4086 }
4087
4088 if (!s)
4089 s = spare;
4090 if (!s) {
4091 champion = NULL;
4092 goto out;
4093 }
4094 champion = s;
4095
4096 if (conflict)
4097 fprintf(stderr, "Chose family %#x on '%s', "
4098 "assemble conflicts to new container with '--update=uuid'\n",
4099 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4100
4101 /* collect all dl's onto 'champion', and update them to
4102 * champion's version of the status
4103 */
4104 for (s = *super_list; s; s = s->next) {
4105 struct imsm_super *mpb = champion->anchor;
4106 struct dl *dl = s->disks;
4107
4108 if (s == champion)
4109 continue;
4110
5d7b407a
CA
4111 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4112
a2b97981
DW
4113 for (i = 0; i < mpb->num_disks; i++) {
4114 struct imsm_disk *disk;
4115
4116 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4117 if (disk) {
4118 dl->disk = *disk;
4119 /* only set index on disks that are a member of
4120 * a populated contianer, i.e. one with
4121 * raid_devs
4122 */
4123 if (is_failed(&dl->disk))
4124 dl->index = -2;
4125 else if (is_spare(&dl->disk))
4126 dl->index = -1;
4127 break;
4128 }
4129 }
4130
4131 if (i >= mpb->num_disks) {
4132 struct intel_disk *idisk;
4133
4134 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4135 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW
4136 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4137 dl->index = -1;
4138 else {
4139 dl->index = -2;
4140 continue;
4141 }
4142 }
4143
4144 dl->next = champion->disks;
4145 champion->disks = dl;
4146 s->disks = NULL;
4147 }
4148
4149 /* delete 'champion' from super_list */
4150 for (del = super_list; *del; ) {
4151 if (*del == champion) {
4152 *del = (*del)->next;
4153 break;
4154 } else
4155 del = &(*del)->next;
4156 }
4157 champion->next = NULL;
4158
4159 out:
4160 while (disk_list) {
4161 struct intel_disk *idisk = disk_list;
4162
4163 disk_list = disk_list->next;
4164 free(idisk);
4165 }
4166
4167 return champion;
4168}
4169
9587c373
LM
4170
4171static int
4172get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
9587c373
LM
4173static int get_super_block(struct intel_super **super_list, int devnum, char *devname,
4174 int major, int minor, int keep_fd);
ec50f7b6
LM
4175static int
4176get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4177 int *max, int keep_fd);
4178
9587c373 4179
cdddbdbc 4180static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM
4181 char *devname, struct md_list *devlist,
4182 int keep_fd)
cdddbdbc 4183{
a2b97981
DW
4184 struct intel_super *super_list = NULL;
4185 struct intel_super *super = NULL;
a2b97981 4186 int err = 0;
9587c373 4187 int i = 0;
dab4a513 4188
9587c373
LM
4189 if (fd >= 0)
4190 /* 'fd' is an opened container */
4191 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4192 else
ec50f7b6
LM
4193 /* get super block from devlist devices */
4194 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4195 if (err)
1602d52c 4196 goto error;
a2b97981
DW
4197 /* all mpbs enter, maybe one leaves */
4198 super = imsm_thunderdome(&super_list, i);
4199 if (!super) {
4200 err = 1;
4201 goto error;
cdddbdbc
DW
4202 }
4203
47ee5a45
DW
4204 if (find_missing(super) != 0) {
4205 free_imsm(super);
a2b97981
DW
4206 err = 2;
4207 goto error;
47ee5a45 4208 }
8e59f3d8
AK
4209
4210 /* load migration record */
4211 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK
4212 if (err == -1) {
4213 /* migration is in progress,
4214 * but migr_rec cannot be loaded,
4215 */
8e59f3d8
AK
4216 err = 4;
4217 goto error;
4218 }
e2f41b2c
AK
4219
4220 /* Check migration compatibility */
4c965cc9 4221 if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) {
e2f41b2c
AK
4222 fprintf(stderr, Name ": Unsupported migration detected");
4223 if (devname)
4224 fprintf(stderr, " on %s\n", devname);
4225 else
4226 fprintf(stderr, " (IMSM).\n");
4227
4228 err = 5;
4229 goto error;
4230 }
4231
a2b97981
DW
4232 err = 0;
4233
4234 error:
4235 while (super_list) {
4236 struct intel_super *s = super_list;
4237
4238 super_list = super_list->next;
4239 free_imsm(s);
4240 }
9587c373 4241
a2b97981
DW
4242
4243 if (err)
4244 return err;
f7e7067b 4245
cdddbdbc 4246 *sbp = super;
9587c373
LM
4247 if (fd >= 0)
4248 st->container_dev = fd2devnum(fd);
4249 else
4250 st->container_dev = NoMdDev;
a2b97981 4251 if (err == 0 && st->ss == NULL) {
bf5a934a 4252 st->ss = &super_imsm;
cdddbdbc
DW
4253 st->minor_version = 0;
4254 st->max_devs = IMSM_MAX_DEVICES;
4255 }
cdddbdbc
DW
4256 return 0;
4257}
2b959fbf 4258
9587c373 4259
ec50f7b6
LM
4260static int
4261get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4262 int *max, int keep_fd)
4263{
4264 struct md_list *tmpdev;
4265 int err = 0;
4266 int i = 0;
9587c373 4267
ec50f7b6
LM
4268 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
4269 if (tmpdev->used != 1)
4270 continue;
4271 if (tmpdev->container == 1) {
ca9de185 4272 int lmax = 0;
ec50f7b6
LM
4273 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
4274 if (fd < 0) {
4275 fprintf(stderr, Name ": cannot open device %s: %s\n",
4276 tmpdev->devname, strerror(errno));
4277 err = 8;
4278 goto error;
4279 }
4280 err = get_sra_super_block(fd, super_list,
4281 tmpdev->devname, &lmax,
4282 keep_fd);
4283 i += lmax;
4284 close(fd);
4285 if (err) {
4286 err = 7;
4287 goto error;
4288 }
4289 } else {
4290 int major = major(tmpdev->st_rdev);
4291 int minor = minor(tmpdev->st_rdev);
4292 err = get_super_block(super_list,
4293 -1,
4294 tmpdev->devname,
4295 major, minor,
4296 keep_fd);
4297 i++;
4298 if (err) {
4299 err = 6;
4300 goto error;
4301 }
4302 }
4303 }
4304 error:
4305 *max = i;
4306 return err;
4307}
9587c373
LM
4308
4309static int get_super_block(struct intel_super **super_list, int devnum, char *devname,
4310 int major, int minor, int keep_fd)
4311{
4312 struct intel_super*s = NULL;
4313 char nm[32];
4314 int dfd = -1;
4315 int rv;
4316 int err = 0;
4317 int retry;
4318
4319 s = alloc_super();
4320 if (!s) {
4321 err = 1;
4322 goto error;
4323 }
4324
4325 sprintf(nm, "%d:%d", major, minor);
4326 dfd = dev_open(nm, O_RDWR);
4327 if (dfd < 0) {
4328 err = 2;
4329 goto error;
4330 }
4331
4332 rv = find_intel_hba_capability(dfd, s, devname);
4333 /* no orom/efi or non-intel hba of the disk */
4334 if (rv != 0) {
4335 err = 4;
4336 goto error;
4337 }
4338
4339 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4340
4341 /* retry the load if we might have raced against mdmon */
4342 if (err == 3 && (devnum != -1) && mdmon_running(devnum))
4343 for (retry = 0; retry < 3; retry++) {
4344 usleep(3000);
4345 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4346 if (err != 3)
4347 break;
4348 }
4349 error:
4350 if (!err) {
4351 s->next = *super_list;
4352 *super_list = s;
4353 } else {
4354 if (s)
4355 free(s);
36614e95 4356 if (dfd >= 0)
9587c373
LM
4357 close(dfd);
4358 }
4359 if ((dfd >= 0) && (!keep_fd))
4360 close(dfd);
4361 return err;
4362
4363}
4364
4365static int
4366get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
4367{
4368 struct mdinfo *sra;
4369 int devnum;
4370 struct mdinfo *sd;
4371 int err = 0;
4372 int i = 0;
4373 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
4374 if (!sra)
4375 return 1;
4376
4377 if (sra->array.major_version != -1 ||
4378 sra->array.minor_version != -2 ||
4379 strcmp(sra->text_version, "imsm") != 0) {
4380 err = 1;
4381 goto error;
4382 }
4383 /* load all mpbs */
4384 devnum = fd2devnum(fd);
4385 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4386 if (get_super_block(super_list, devnum, devname,
4387 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
4388 err = 7;
4389 goto error;
4390 }
4391 }
4392 error:
4393 sysfs_free(sra);
4394 *max = i;
4395 return err;
4396}
4397
2b959fbf
N
4398static int load_container_imsm(struct supertype *st, int fd, char *devname)
4399{
ec50f7b6 4400 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 4401}
cdddbdbc
DW
4402#endif
4403
4404static int load_super_imsm(struct supertype *st, int fd, char *devname)
4405{
4406 struct intel_super *super;
4407 int rv;
4408
691c6ee1
N
4409 if (test_partition(fd))
4410 /* IMSM not allowed on partitions */
4411 return 1;
4412
37424f13
DW
4413 free_super_imsm(st);
4414
49133e57 4415 super = alloc_super();
cdddbdbc
DW
4416 if (!super) {
4417 fprintf(stderr,
4418 Name ": malloc of %zu failed.\n",
4419 sizeof(*super));
4420 return 1;
4421 }
ea2bc72b
LM
4422 /* Load hba and capabilities if they exist.
4423 * But do not preclude loading metadata in case capabilities or hba are
4424 * non-compliant and ignore_hw_compat is set.
4425 */
d424212e 4426 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 4427 /* no orom/efi or non-intel hba of the disk */
ea2bc72b 4428 if ((rv != 0) && (st->ignore_hw_compat == 0)) {
f2f5c343
LM
4429 if (devname)
4430 fprintf(stderr,
4431 Name ": No OROM/EFI properties for %s\n", devname);
4432 free_imsm(super);
4433 return 2;
4434 }
a2b97981 4435 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc
DW
4436
4437 if (rv) {
4438 if (devname)
4439 fprintf(stderr,
4440 Name ": Failed to load all information "
4441 "sections on %s\n", devname);
4442 free_imsm(super);
4443 return rv;
4444 }
4445
4446 st->sb = super;
4447 if (st->ss == NULL) {
4448 st->ss = &super_imsm;
4449 st->minor_version = 0;
4450 st->max_devs = IMSM_MAX_DEVICES;
4451 }
8e59f3d8
AK
4452
4453 /* load migration record */
2e062e82
AK
4454 if (load_imsm_migr_rec(super, NULL) == 0) {
4455 /* Check for unsupported migration features */
4456 if (check_mpb_migr_compatibility(super) != 0) {
4457 fprintf(stderr,
4458 Name ": Unsupported migration detected");
4459 if (devname)
4460 fprintf(stderr, " on %s\n", devname);
4461 else
4462 fprintf(stderr, " (IMSM).\n");
4463 return 3;
4464 }
e2f41b2c
AK
4465 }
4466
cdddbdbc
DW
4467 return 0;
4468}
4469
ef6ffade
DW
4470static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
4471{
4472 if (info->level == 1)
4473 return 128;
4474 return info->chunk_size >> 9;
4475}
4476
5551b113
CA
4477static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
4478 unsigned long long size)
fcfd9599 4479{
4025c288 4480 if (info->level == 1)
5551b113 4481 return size * 2;
4025c288 4482 else
5551b113 4483 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW
4484}
4485
4d1313e9
DW
4486static void imsm_update_version_info(struct intel_super *super)
4487{
4488 /* update the version and attributes */
4489 struct imsm_super *mpb = super->anchor;
4490 char *version;
4491 struct imsm_dev *dev;
4492 struct imsm_map *map;
4493 int i;
4494
4495 for (i = 0; i < mpb->num_raid_devs; i++) {
4496 dev = get_imsm_dev(super, i);
238c0a71 4497 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW
4498 if (__le32_to_cpu(dev->size_high) > 0)
4499 mpb->attributes |= MPB_ATTRIB_2TB;
4500
4501 /* FIXME detect when an array spans a port multiplier */
4502 #if 0
4503 mpb->attributes |= MPB_ATTRIB_PM;
4504 #endif
4505
4506 if (mpb->num_raid_devs > 1 ||
4507 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
4508 version = MPB_VERSION_ATTRIBS;
4509 switch (get_imsm_raid_level(map)) {
4510 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
4511 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
4512 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
4513 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
4514 }
4515 } else {
4516 if (map->num_members >= 5)
4517 version = MPB_VERSION_5OR6_DISK_ARRAY;
4518 else if (dev->status == DEV_CLONE_N_GO)
4519 version = MPB_VERSION_CNG;
4520 else if (get_imsm_raid_level(map) == 5)
4521 version = MPB_VERSION_RAID5;
4522 else if (map->num_members >= 3)
4523 version = MPB_VERSION_3OR4_DISK_ARRAY;
4524 else if (get_imsm_raid_level(map) == 1)
4525 version = MPB_VERSION_RAID1;
4526 else
4527 version = MPB_VERSION_RAID0;
4528 }
4529 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
4530 }
4531}
4532
aa534678
DW
4533static int check_name(struct intel_super *super, char *name, int quiet)
4534{
4535 struct imsm_super *mpb = super->anchor;
4536 char *reason = NULL;
4537 int i;
4538
4539 if (strlen(name) > MAX_RAID_SERIAL_LEN)
4540 reason = "must be 16 characters or less";
4541
4542 for (i = 0; i < mpb->num_raid_devs; i++) {
4543 struct imsm_dev *dev = get_imsm_dev(super, i);
4544
4545 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
4546 reason = "already exists";
4547 break;
4548 }
4549 }
4550
4551 if (reason && !quiet)
4552 fprintf(stderr, Name ": imsm volume name %s\n", reason);
4553
4554 return !reason;
4555}
4556
8b353278
DW
4557static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
4558 unsigned long long size, char *name,
4559 char *homehost, int *uuid)
cdddbdbc 4560{
c2c087e6
DW
4561 /* We are creating a volume inside a pre-existing container.
4562 * so st->sb is already set.
4563 */
4564 struct intel_super *super = st->sb;
949c47a0 4565 struct imsm_super *mpb = super->anchor;
ba2de7ba 4566 struct intel_dev *dv;
c2c087e6
DW
4567 struct imsm_dev *dev;
4568 struct imsm_vol *vol;
4569 struct imsm_map *map;
4570 int idx = mpb->num_raid_devs;
4571 int i;
4572 unsigned long long array_blocks;
2c092cad 4573 size_t size_old, size_new;
5551b113 4574 unsigned long long num_data_stripes;
cdddbdbc 4575
88c32bb1 4576 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
c2c087e6 4577 fprintf(stderr, Name": This imsm-container already has the "
88c32bb1 4578 "maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW
4579 return 0;
4580 }
4581
2c092cad
DW
4582 /* ensure the mpb is large enough for the new data */
4583 size_old = __le32_to_cpu(mpb->mpb_size);
4584 size_new = disks_to_mpb_size(info->nr_disks);
4585 if (size_new > size_old) {
4586 void *mpb_new;
4587 size_t size_round = ROUND_UP(size_new, 512);
4588
4589 if (posix_memalign(&mpb_new, 512, size_round) != 0) {
4590 fprintf(stderr, Name": could not allocate new mpb\n");
4591 return 0;
4592 }
17a4eaf9
AK
4593 if (posix_memalign(&super->migr_rec_buf, 512,
4594 MIGR_REC_BUF_SIZE) != 0) {
8e59f3d8
AK
4595 fprintf(stderr, Name
4596 ": %s could not allocate migr_rec buffer\n",
4597 __func__);
4598 free(super->buf);
4599 free(super);
ea944c8f 4600 free(mpb_new);
8e59f3d8
AK
4601 return 0;
4602 }
2c092cad
DW
4603 memcpy(mpb_new, mpb, size_old);
4604 free(mpb);
4605 mpb = mpb_new;
949c47a0 4606 super->anchor = mpb_new;
2c092cad
DW
4607 mpb->mpb_size = __cpu_to_le32(size_new);
4608 memset(mpb_new + size_old, 0, size_round - size_old);
4609 }
bf5a934a 4610 super->current_vol = idx;
3960e579
DW
4611
4612 /* handle 'failed_disks' by either:
4613 * a) create dummy disk entries in the table if this the first
4614 * volume in the array. We add them here as this is the only
4615 * opportunity to add them. add_to_super_imsm_volume()
4616 * handles the non-failed disks and continues incrementing
4617 * mpb->num_disks.
4618 * b) validate that 'failed_disks' matches the current number
4619 * of missing disks if the container is populated
d23fe947 4620 */
3960e579 4621 if (super->current_vol == 0) {
d23fe947 4622 mpb->num_disks = 0;
3960e579
DW
4623 for (i = 0; i < info->failed_disks; i++) {
4624 struct imsm_disk *disk;
4625
4626 mpb->num_disks++;
4627 disk = __get_imsm_disk(mpb, i);
4628 disk->status = CONFIGURED_DISK | FAILED_DISK;
4629 disk->scsi_id = __cpu_to_le32(~(__u32)0);
4630 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
4631 "missing:%d", i);
4632 }
4633 find_missing(super);
4634 } else {
4635 int missing = 0;
4636 struct dl *d;
4637
4638 for (d = super->missing; d; d = d->next)
4639 missing++;
4640 if (info->failed_disks > missing) {
4641 fprintf(stderr, Name": unable to add 'missing' disk to container\n");
4642 return 0;
4643 }
4644 }
5a038140 4645
aa534678
DW
4646 if (!check_name(super, name, 0))
4647 return 0;
ba2de7ba
DW
4648 dv = malloc(sizeof(*dv));
4649 if (!dv) {
4650 fprintf(stderr, Name ": failed to allocate device list entry\n");
4651 return 0;
4652 }
1a2487c2 4653 dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
949c47a0 4654 if (!dev) {
ba2de7ba 4655 free(dv);
949c47a0
DW
4656 fprintf(stderr, Name": could not allocate raid device\n");
4657 return 0;
4658 }
1a2487c2 4659
c2c087e6 4660 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
e03640bd 4661 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 4662 info->layout, info->chunk_size,
5551b113 4663 size * 2);
979d38be
DW
4664 /* round array size down to closest MB */
4665 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
4666
c2c087e6
DW
4667 dev->size_low = __cpu_to_le32((__u32) array_blocks);
4668 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1a2487c2 4669 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW
4670 vol = &dev->vol;
4671 vol->migr_state = 0;
1484e727 4672 set_migr_type(dev, MIGR_INIT);
3960e579 4673 vol->dirty = !info->state;
f8f603f1 4674 vol->curr_migr_unit = 0;
238c0a71 4675 map = get_imsm_map(dev, MAP_0);
5551b113
CA
4676 set_pba_of_lba0(map, super->create_offset);
4677 set_blocks_per_member(map, info_to_blocks_per_member(info, size));
ef6ffade 4678 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 4679 map->failed_disk_num = ~0;
bf4442ab
AK
4680 if (info->level > 0)
4681 map->map_state = IMSM_T_STATE_UNINITIALIZED;
4682 else
4683 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
4684 IMSM_T_STATE_NORMAL;
252d23c0 4685 map->ddf = 1;
ef6ffade
DW
4686
4687 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW
4688 free(dev);
4689 free(dv);
ef6ffade
DW
4690 fprintf(stderr, Name": imsm does not support more than 2 disks"
4691 "in a raid1 volume\n");
4692 return 0;
4693 }
81062a36
DW
4694
4695 map->raid_level = info->level;
4d1313e9 4696 if (info->level == 10) {
c2c087e6 4697 map->raid_level = 1;
4d1313e9 4698 map->num_domains = info->raid_disks / 2;
81062a36
DW
4699 } else if (info->level == 1)
4700 map->num_domains = info->raid_disks;
4701 else
ff596308 4702 map->num_domains = 1;
81062a36 4703
5551b113
CA
4704 /* info->size is only int so use the 'size' parameter instead */
4705 num_data_stripes = (size * 2) / info_to_blocks_per_strip(info);
4706 num_data_stripes /= map->num_domains;
4707 set_num_data_stripes(map, num_data_stripes);
ef6ffade 4708
c2c087e6
DW
4709 map->num_members = info->raid_disks;
4710 for (i = 0; i < map->num_members; i++) {
4711 /* initialized in add_to_super */
4eb26970 4712 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 4713 }
949c47a0 4714 mpb->num_raid_devs++;
ba2de7ba
DW
4715
4716 dv->dev = dev;
4717 dv->index = super->current_vol;
4718 dv->next = super->devlist;
4719 super->devlist = dv;
c2c087e6 4720
4d1313e9
DW
4721 imsm_update_version_info(super);
4722
c2c087e6 4723 return 1;
cdddbdbc
DW
4724}
4725
bf5a934a
DW
4726static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
4727 unsigned long long size, char *name,
4728 char *homehost, int *uuid)
4729{
4730 /* This is primarily called by Create when creating a new array.
4731 * We will then get add_to_super called for each component, and then
4732 * write_init_super called to write it out to each device.
4733 * For IMSM, Create can create on fresh devices or on a pre-existing
4734 * array.
4735 * To create on a pre-existing array a different method will be called.
4736 * This one is just for fresh drives.
4737 */
4738 struct intel_super *super;
4739 struct imsm_super *mpb;
4740 size_t mpb_size;
4d1313e9 4741 char *version;
bf5a934a 4742
bf5a934a 4743 if (st->sb)
e683ca88
DW
4744 return init_super_imsm_volume(st, info, size, name, homehost, uuid);
4745
4746 if (info)
4747 mpb_size = disks_to_mpb_size(info->nr_disks);
4748 else
4749 mpb_size = 512;
bf5a934a 4750
49133e57 4751 super = alloc_super();
e683ca88 4752 if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) {
bf5a934a 4753 free(super);
e683ca88
DW
4754 super = NULL;
4755 }
4756 if (!super) {
4757 fprintf(stderr, Name
4758 ": %s could not allocate superblock\n", __func__);
bf5a934a
DW
4759 return 0;
4760 }
17a4eaf9 4761 if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) {
8e59f3d8
AK
4762 fprintf(stderr, Name
4763 ": %s could not allocate migr_rec buffer\n", __func__);
4764 free(super->buf);
4765 free(super);
4766 return 0;
4767 }
e683ca88 4768 memset(super->buf, 0, mpb_size);
ef649044 4769 mpb = super->buf;
e683ca88
DW
4770 mpb->mpb_size = __cpu_to_le32(mpb_size);
4771 st->sb = super;
4772
4773 if (info == NULL) {
4774 /* zeroing superblock */
4775 return 0;
4776 }
bf5a934a 4777
4d1313e9
DW
4778 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
4779
4780 version = (char *) mpb->sig;
4781 strcpy(version, MPB_SIGNATURE);
4782 version += strlen(MPB_SIGNATURE);
4783 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 4784
bf5a934a
DW
4785 return 1;
4786}
4787
0e600426 4788#ifndef MDASSEMBLE
f20c3968 4789static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW
4790 int fd, char *devname)
4791{
4792 struct intel_super *super = st->sb;
d23fe947 4793 struct imsm_super *mpb = super->anchor;
3960e579 4794 struct imsm_disk *_disk;
bf5a934a
DW
4795 struct imsm_dev *dev;
4796 struct imsm_map *map;
3960e579 4797 struct dl *dl, *df;
4eb26970 4798 int slot;
bf5a934a 4799
949c47a0 4800 dev = get_imsm_dev(super, super->current_vol);
238c0a71 4801 map = get_imsm_map(dev, MAP_0);
bf5a934a 4802
208933a7
N
4803 if (! (dk->state & (1<<MD_DISK_SYNC))) {
4804 fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n",
4805 devname);
4806 return 1;
4807 }
4808
efb30e7f
DW
4809 if (fd == -1) {
4810 /* we're doing autolayout so grab the pre-marked (in
4811 * validate_geometry) raid_disk
4812 */
4813 for (dl = super->disks; dl; dl = dl->next)
4814 if (dl->raiddisk == dk->raid_disk)
4815 break;
4816 } else {
4817 for (dl = super->disks; dl ; dl = dl->next)
4818 if (dl->major == dk->major &&
4819 dl->minor == dk->minor)
4820 break;
4821 }
d23fe947 4822
208933a7
N
4823 if (!dl) {
4824 fprintf(stderr, Name ": %s is not a member of the same container\n", devname);
f20c3968 4825 return 1;
208933a7 4826 }
bf5a934a 4827
d23fe947
DW
4828 /* add a pristine spare to the metadata */
4829 if (dl->index < 0) {
4830 dl->index = super->anchor->num_disks;
4831 super->anchor->num_disks++;
4832 }
4eb26970
DW
4833 /* Check the device has not already been added */
4834 slot = get_imsm_disk_slot(map, dl->index);
4835 if (slot >= 0 &&
238c0a71 4836 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
4eb26970
DW
4837 fprintf(stderr, Name ": %s has been included in this array twice\n",
4838 devname);
4839 return 1;
4840 }
656b6b5a 4841 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 4842 dl->disk.status = CONFIGURED_DISK;
d23fe947 4843
3960e579
DW
4844 /* update size of 'missing' disks to be at least as large as the
4845 * largest acitve member (we only have dummy missing disks when
4846 * creating the first volume)
4847 */
4848 if (super->current_vol == 0) {
4849 for (df = super->missing; df; df = df->next) {
5551b113
CA
4850 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
4851 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW
4852 _disk = __get_imsm_disk(mpb, df->index);
4853 *_disk = df->disk;
4854 }
4855 }
4856
4857 /* refresh unset/failed slots to point to valid 'missing' entries */
4858 for (df = super->missing; df; df = df->next)
4859 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 4860 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW
4861
4862 if ((ord & IMSM_ORD_REBUILD) == 0)
4863 continue;
4864 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 4865 if (is_gen_migration(dev)) {
238c0a71
AK
4866 struct imsm_map *map2 = get_imsm_map(dev,
4867 MAP_1);
0a108d63
AK
4868 int slot2 = get_imsm_disk_slot(map2, df->index);
4869 if ((slot2 < map2->num_members) &&
4870 (slot2 >= 0)) {
1ace8403 4871 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK
4872 slot2,
4873 MAP_1);
1ace8403
AK
4874 if ((unsigned)df->index ==
4875 ord_to_idx(ord2))
4876 set_imsm_ord_tbl_ent(map2,
0a108d63 4877 slot2,
1ace8403
AK
4878 df->index |
4879 IMSM_ORD_REBUILD);
4880 }
4881 }
3960e579
DW
4882 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
4883 break;
4884 }
4885
d23fe947
DW
4886 /* if we are creating the first raid device update the family number */
4887 if (super->current_vol == 0) {
4888 __u32 sum;
4889 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 4890
3960e579 4891 _disk = __get_imsm_disk(mpb, dl->index);
791b666a
AW
4892 if (!_dev || !_disk) {
4893 fprintf(stderr, Name ": BUG mpb setup error\n");
4894 return 1;
4895 }
d23fe947
DW
4896 *_dev = *dev;
4897 *_disk = dl->disk;
148acb7b
DW
4898 sum = random32();
4899 sum += __gen_imsm_checksum(mpb);
d23fe947 4900 mpb->family_num = __cpu_to_le32(sum);
148acb7b 4901 mpb->orig_family_num = mpb->family_num;
d23fe947 4902 }
ca0748fa 4903 super->current_disk = dl;
f20c3968 4904 return 0;
bf5a934a
DW
4905}
4906
a8619d23
AK
4907/* mark_spare()
4908 * Function marks disk as spare and restores disk serial
4909 * in case it was previously marked as failed by takeover operation
4910 * reruns:
4911 * -1 : critical error
4912 * 0 : disk is marked as spare but serial is not set
4913 * 1 : success
4914 */
4915int mark_spare(struct dl *disk)
4916{
4917 __u8 serial[MAX_RAID_SERIAL_LEN];
4918 int ret_val = -1;
4919
4920 if (!disk)
4921 return ret_val;
4922
4923 ret_val = 0;
4924 if (!imsm_read_serial(disk->fd, NULL, serial)) {
4925 /* Restore disk serial number, because takeover marks disk
4926 * as failed and adds to serial ':0' before it becomes
4927 * a spare disk.
4928 */
4929 serialcpy(disk->serial, serial);
4930 serialcpy(disk->disk.serial, serial);
4931 ret_val = 1;
4932 }
4933 disk->disk.status = SPARE_DISK;
4934 disk->index = -1;
4935
4936 return ret_val;
4937}
88654014 4938
f20c3968 4939static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
88654014 4940 int fd, char *devname)
cdddbdbc 4941{
c2c087e6 4942 struct intel_super *super = st->sb;
c2c087e6
DW
4943 struct dl *dd;
4944 unsigned long long size;
f2f27e63 4945 __u32 id;
c2c087e6
DW
4946 int rv;
4947 struct stat stb;
4948
88654014
LM
4949 /* If we are on an RAID enabled platform check that the disk is
4950 * attached to the raid controller.
4951 * We do not need to test disks attachment for container based additions,
4952 * they shall be already tested when container was created/assembled.
88c32bb1 4953 */
d424212e 4954 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 4955 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM
4956 if (rv != 0) {
4957 dprintf("capability: %p fd: %d ret: %d\n",
4958 super->orom, fd, rv);
4959 return 1;
88c32bb1
DW
4960 }
4961
f20c3968
DW
4962 if (super->current_vol >= 0)
4963 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 4964
c2c087e6
DW
4965 fstat(fd, &stb);
4966 dd = malloc(sizeof(*dd));
b9f594fe 4967 if (!dd) {
c2c087e6
DW
4968 fprintf(stderr,
4969 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
f20c3968 4970 return 1;
c2c087e6
DW
4971 }
4972 memset(dd, 0, sizeof(*dd));
4973 dd->major = major(stb.st_rdev);
4974 dd->minor = minor(stb.st_rdev);
c2c087e6 4975 dd->devname = devname ? strdup(devname) : NULL;
c2c087e6 4976 dd->fd = fd;
689c9bf3 4977 dd->e = NULL;
1a64be56 4978 dd->action = DISK_ADD;
c2c087e6 4979 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 4980 if (rv) {
c2c087e6 4981 fprintf(stderr,
0030e8d6 4982 Name ": failed to retrieve scsi serial, aborting\n");
949c47a0 4983 free(dd);
0030e8d6 4984 abort();
c2c087e6
DW
4985 }
4986
c2c087e6
DW
4987 get_dev_size(fd, NULL, &size);
4988 size /= 512;
1f24f035 4989 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA
4990 set_total_blocks(&dd->disk, size);
4991 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
4992 struct imsm_super *mpb = super->anchor;
4993 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
4994 }
a8619d23 4995 mark_spare(dd);
c2c087e6 4996 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 4997 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 4998 else
b9f594fe 4999 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW
5000
5001 if (st->update_tail) {
1a64be56
LM
5002 dd->next = super->disk_mgmt_list;
5003 super->disk_mgmt_list = dd;
43dad3d6
DW
5004 } else {
5005 dd->next = super->disks;
5006 super->disks = dd;
ceaf0ee1 5007 super->updates_pending++;
43dad3d6 5008 }
f20c3968
DW
5009
5010 return 0;
cdddbdbc
DW
5011}
5012
1a64be56
LM
5013
5014static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5015{
5016 struct intel_super *super = st->sb;
5017 struct dl *dd;
5018
5019 /* remove from super works only in mdmon - for communication
5020 * manager - monitor. Check if communication memory buffer
5021 * is prepared.
5022 */
5023 if (!st->update_tail) {
5024 fprintf(stderr,
5025 Name ": %s shall be used in mdmon context only"
5026 "(line %d).\n", __func__, __LINE__);
5027 return 1;
5028 }
5029 dd = malloc(sizeof(*dd));
5030 if (!dd) {
5031 fprintf(stderr,
5032 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
5033 return 1;
5034 }
5035 memset(dd, 0, sizeof(*dd));
5036 dd->major = dk->major;
5037 dd->minor = dk->minor;
1a64be56 5038 dd->fd = -1;
a8619d23 5039 mark_spare(dd);
1a64be56
LM
5040 dd->action = DISK_REMOVE;
5041
5042 dd->next = super->disk_mgmt_list;
5043 super->disk_mgmt_list = dd;
5044
5045
5046 return 0;
5047}
5048
f796af5d
DW
5049static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5050
5051static union {
5052 char buf[512];
5053 struct imsm_super anchor;
5054} spare_record __attribute__ ((aligned(512)));
c2c087e6 5055
d23fe947
DW
5056/* spare records have their own family number and do not have any defined raid
5057 * devices
5058 */
5059static int write_super_imsm_spares(struct intel_super *super, int doclose)
5060{
d23fe947 5061 struct imsm_super *mpb = super->anchor;
f796af5d 5062 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW
5063 __u32 sum;
5064 struct dl *d;
5065
f796af5d
DW
5066 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)),
5067 spare->generation_num = __cpu_to_le32(1UL),
5068 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5069 spare->num_disks = 1,
5070 spare->num_raid_devs = 0,
5071 spare->cache_size = mpb->cache_size,
5072 spare->pwr_cycle_count = __cpu_to_le32(1),
5073
5074 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5075 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW
5076
5077 for (d = super->disks; d; d = d->next) {
8796fdc4 5078 if (d->index != -1)
d23fe947
DW
5079 continue;
5080
f796af5d 5081 spare->disk[0] = d->disk;
027c374f
CA
5082 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5083 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5084
f796af5d
DW
5085 sum = __gen_imsm_checksum(spare);
5086 spare->family_num = __cpu_to_le32(sum);
5087 spare->orig_family_num = 0;
5088 sum = __gen_imsm_checksum(spare);
5089 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5090
f796af5d 5091 if (store_imsm_mpb(d->fd, spare)) {
d23fe947
DW
5092 fprintf(stderr, "%s: failed for device %d:%d %s\n",
5093 __func__, d->major, d->minor, strerror(errno));
e74255d9 5094 return 1;
d23fe947
DW
5095 }
5096 if (doclose) {
5097 close(d->fd);
5098 d->fd = -1;
5099 }
5100 }
5101
e74255d9 5102 return 0;
d23fe947
DW
5103}
5104
36988a3d 5105static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5106{
36988a3d 5107 struct intel_super *super = st->sb;
949c47a0 5108 struct imsm_super *mpb = super->anchor;
c2c087e6
DW
5109 struct dl *d;
5110 __u32 generation;
5111 __u32 sum;
d23fe947 5112 int spares = 0;
949c47a0 5113 int i;
a48ac0a8 5114 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5115 int num_disks = 0;
146c6260 5116 int clear_migration_record = 1;
cdddbdbc 5117
c2c087e6
DW
5118 /* 'generation' is incremented everytime the metadata is written */
5119 generation = __le32_to_cpu(mpb->generation_num);
5120 generation++;
5121 mpb->generation_num = __cpu_to_le32(generation);
5122
148acb7b
DW
5123 /* fix up cases where previous mdadm releases failed to set
5124 * orig_family_num
5125 */
5126 if (mpb->orig_family_num == 0)
5127 mpb->orig_family_num = mpb->family_num;
5128
d23fe947 5129 for (d = super->disks; d; d = d->next) {
8796fdc4 5130 if (d->index == -1)
d23fe947 5131 spares++;
36988a3d 5132 else {
d23fe947 5133 mpb->disk[d->index] = d->disk;
36988a3d
AK
5134 num_disks++;
5135 }
d23fe947 5136 }
36988a3d 5137 for (d = super->missing; d; d = d->next) {
47ee5a45 5138 mpb->disk[d->index] = d->disk;
36988a3d
AK
5139 num_disks++;
5140 }
5141 mpb->num_disks = num_disks;
5142 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 5143
949c47a0
DW
5144 for (i = 0; i < mpb->num_raid_devs; i++) {
5145 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK
5146 struct imsm_dev *dev2 = get_imsm_dev(super, i);
5147 if (dev && dev2) {
5148 imsm_copy_dev(dev, dev2);
5149 mpb_size += sizeof_imsm_dev(dev, 0);
5150 }
146c6260
AK
5151 if (is_gen_migration(dev2))
5152 clear_migration_record = 0;
949c47a0 5153 }
a48ac0a8
DW
5154 mpb_size += __le32_to_cpu(mpb->bbm_log_size);
5155 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 5156
c2c087e6 5157 /* recalculate checksum */
949c47a0 5158 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW
5159 mpb->check_sum = __cpu_to_le32(sum);
5160
51d83f5d
AK
5161 if (super->clean_migration_record_by_mdmon) {
5162 clear_migration_record = 1;
5163 super->clean_migration_record_by_mdmon = 0;
5164 }
146c6260 5165 if (clear_migration_record)
17a4eaf9 5166 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE);
146c6260 5167
d23fe947 5168 /* write the mpb for disks that compose raid devices */
c2c087e6 5169 for (d = super->disks; d ; d = d->next) {
86c54047 5170 if (d->index < 0 || is_failed(&d->disk))
d23fe947 5171 continue;
30602f53 5172
146c6260
AK
5173 if (clear_migration_record) {
5174 unsigned long long dsize;
5175
5176 get_dev_size(d->fd, NULL, &dsize);
5177 if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) {
17a4eaf9
AK
5178 if (write(d->fd, super->migr_rec_buf,
5179 MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE)
9e2d750d 5180 perror("Write migr_rec failed");
146c6260
AK
5181 }
5182 }
51d83f5d
AK
5183
5184 if (store_imsm_mpb(d->fd, mpb))
5185 fprintf(stderr,
5186 "%s: failed for device %d:%d (fd: %d)%s\n",
5187 __func__, d->major, d->minor,
5188 d->fd, strerror(errno));
5189
c2c087e6
DW
5190 if (doclose) {
5191 close(d->fd);
5192 d->fd = -1;
5193 }
5194 }
5195
d23fe947
DW
5196 if (spares)
5197 return write_super_imsm_spares(super, doclose);
5198
e74255d9 5199 return 0;
c2c087e6
DW
5200}
5201
0e600426 5202
9b1fb677 5203static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW
5204{
5205 size_t len;
5206 struct imsm_update_create_array *u;
5207 struct intel_super *super = st->sb;
9b1fb677 5208 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 5209 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW
5210 struct disk_info *inf;
5211 struct imsm_disk *disk;
5212 int i;
43dad3d6 5213
54c2c1ea
DW
5214 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
5215 sizeof(*inf) * map->num_members;
43dad3d6
DW
5216 u = malloc(len);
5217 if (!u) {
5218 fprintf(stderr, "%s: failed to allocate update buffer\n",
5219 __func__);
5220 return 1;
5221 }
5222
5223 u->type = update_create_array;
9b1fb677 5224 u->dev_idx = dev_idx;
43dad3d6 5225 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW
5226 inf = get_disk_info(u);
5227 for (i = 0; i < map->num_members; i++) {
238c0a71 5228 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 5229
54c2c1ea
DW
5230 disk = get_imsm_disk(super, idx);
5231 serialcpy(inf[i].serial, disk->serial);
5232 }
43dad3d6
DW
5233 append_metadata_update(st, u, len);
5234
5235 return 0;
5236}
5237
1a64be56 5238static int mgmt_disk(struct supertype *st)
43dad3d6
DW
5239{
5240 struct intel_super *super = st->sb;
5241 size_t len;
1a64be56 5242 struct imsm_update_add_remove_disk *u;
43dad3d6 5243
1a64be56 5244 if (!super->disk_mgmt_list)
43dad3d6
DW
5245 return 0;
5246
5247 len = sizeof(*u);
5248 u = malloc(len);
5249 if (!u) {
5250 fprintf(stderr, "%s: failed to allocate update buffer\n",
5251 __func__);
5252 return 1;
5253 }
5254
1a64be56 5255 u->type = update_add_remove_disk;
43dad3d6
DW
5256 append_metadata_update(st, u, len);
5257
5258 return 0;
5259}
5260
c2c087e6
DW
5261static int write_init_super_imsm(struct supertype *st)
5262{
9b1fb677
DW
5263 struct intel_super *super = st->sb;
5264 int current_vol = super->current_vol;
5265
5266 /* we are done with current_vol reset it to point st at the container */
5267 super->current_vol = -1;
5268
8273f55e 5269 if (st->update_tail) {
43dad3d6
DW
5270 /* queue the recently created array / added disk
5271 * as a metadata update */
43dad3d6 5272 int rv;
8273f55e 5273
43dad3d6 5274 /* determine if we are creating a volume or adding a disk */
9b1fb677 5275 if (current_vol < 0) {
1a64be56
LM
5276 /* in the mgmt (add/remove) disk case we are running
5277 * in mdmon context, so don't close fd's
43dad3d6 5278 */
1a64be56 5279 return mgmt_disk(st);
43dad3d6 5280 } else
9b1fb677 5281 rv = create_array(st, current_vol);
8273f55e 5282
43dad3d6 5283 return rv;
d682f344
N
5284 } else {
5285 struct dl *d;
5286 for (d = super->disks; d; d = d->next)
5287 Kill(d->devname, NULL, 0, 1, 1);
36988a3d 5288 return write_super_imsm(st, 1);
d682f344 5289 }
cdddbdbc 5290}
0e600426 5291#endif
cdddbdbc 5292
e683ca88 5293static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 5294{
e683ca88
DW
5295 struct intel_super *super = st->sb;
5296 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 5297
e683ca88 5298 if (!mpb)
ad97895e
DW
5299 return 1;
5300
1799c9e8 5301#ifndef MDASSEMBLE
e683ca88 5302 return store_imsm_mpb(fd, mpb);
1799c9e8
N
5303#else
5304 return 1;
5305#endif
cdddbdbc
DW
5306}
5307
0e600426
N
5308static int imsm_bbm_log_size(struct imsm_super *mpb)
5309{
5310 return __le32_to_cpu(mpb->bbm_log_size);
5311}
5312
5313#ifndef MDASSEMBLE
cdddbdbc
DW
5314static int validate_geometry_imsm_container(struct supertype *st, int level,
5315 int layout, int raiddisks, int chunk,
c2c087e6 5316 unsigned long long size, char *dev,
2c514b71
NB
5317 unsigned long long *freesize,
5318 int verbose)
cdddbdbc 5319{
c2c087e6
DW
5320 int fd;
5321 unsigned long long ldsize;
f2f5c343
LM
5322 struct intel_super *super=NULL;
5323 int rv = 0;
cdddbdbc 5324
c2c087e6
DW
5325 if (level != LEVEL_CONTAINER)
5326 return 0;
5327 if (!dev)
5328 return 1;
5329
5330 fd = open(dev, O_RDONLY|O_EXCL, 0);
5331 if (fd < 0) {
2c514b71
NB
5332 if (verbose)
5333 fprintf(stderr, Name ": imsm: Cannot open %s: %s\n",
5334 dev, strerror(errno));
c2c087e6
DW
5335 return 0;
5336 }
5337 if (!get_dev_size(fd, dev, &ldsize)) {
5338 close(fd);
5339 return 0;
5340 }
f2f5c343
LM
5341
5342 /* capabilities retrieve could be possible
5343 * note that there is no fd for the disks in array.
5344 */
5345 super = alloc_super();
5346 if (!super) {
5347 fprintf(stderr,
5348 Name ": malloc of %zu failed.\n",
5349 sizeof(*super));
5350 close(fd);
5351 return 0;
5352 }
5353
d424212e 5354 rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL);
f2f5c343
LM
5355 if (rv != 0) {
5356#if DEBUG
5357 char str[256];
5358 fd2devname(fd, str);
5359 dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n",
5360 fd, str, super->orom, rv, raiddisks);
5361#endif
5362 /* no orom/efi or non-intel hba of the disk */
5363 close(fd);
5364 free_imsm(super);
5365 return 0;
5366 }
c2c087e6 5367 close(fd);
9126b9a8
CA
5368 if (super->orom) {
5369 if (raiddisks > super->orom->tds) {
5370 if (verbose)
5371 fprintf(stderr, Name ": %d exceeds maximum number of"
5372 " platform supported disks: %d\n",
5373 raiddisks, super->orom->tds);
5374 free_imsm(super);
5375 return 0;
5376 }
5377 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
5378 (ldsize >> 9) >> 32 > 0) {
5379 if (verbose)
5380 fprintf(stderr, Name ": %s exceeds maximum platform supported size\n", dev);
5381 free_imsm(super);
5382 return 0;
5383 }
f2f5c343 5384 }
c2c087e6
DW
5385
5386 *freesize = avail_size_imsm(st, ldsize >> 9);
f2f5c343 5387 free_imsm(super);
c2c087e6
DW
5388
5389 return 1;
cdddbdbc
DW
5390}
5391
0dcecb2e
DW
5392static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
5393{
5394 const unsigned long long base_start = e[*idx].start;
5395 unsigned long long end = base_start + e[*idx].size;
5396 int i;
5397
5398 if (base_start == end)
5399 return 0;
5400
5401 *idx = *idx + 1;
5402 for (i = *idx; i < num_extents; i++) {
5403 /* extend overlapping extents */
5404 if (e[i].start >= base_start &&
5405 e[i].start <= end) {
5406 if (e[i].size == 0)
5407 return 0;
5408 if (e[i].start + e[i].size > end)
5409 end = e[i].start + e[i].size;
5410 } else if (e[i].start > end) {
5411 *idx = i;
5412 break;
5413 }
5414 }
5415
5416 return end - base_start;
5417}
5418
5419static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
5420{
5421 /* build a composite disk with all known extents and generate a new
5422 * 'maxsize' given the "all disks in an array must share a common start
5423 * offset" constraint
5424 */
5425 struct extent *e = calloc(sum_extents, sizeof(*e));
5426 struct dl *dl;
5427 int i, j;
5428 int start_extent;
5429 unsigned long long pos;
b9d77223 5430 unsigned long long start = 0;
0dcecb2e
DW
5431 unsigned long long maxsize;
5432 unsigned long reserve;
5433
5434 if (!e)
a7dd165b 5435 return 0;
0dcecb2e
DW
5436
5437 /* coalesce and sort all extents. also, check to see if we need to
5438 * reserve space between member arrays
5439 */
5440 j = 0;
5441 for (dl = super->disks; dl; dl = dl->next) {
5442 if (!dl->e)
5443 continue;
5444 for (i = 0; i < dl->extent_cnt; i++)
5445 e[j++] = dl->e[i];
5446 }
5447 qsort(e, sum_extents, sizeof(*e), cmp_extent);
5448
5449 /* merge extents */
5450 i = 0;
5451 j = 0;
5452 while (i < sum_extents) {
5453 e[j].start = e[i].start;
5454 e[j].size = find_size(e, &i, sum_extents);
5455 j++;
5456 if (e[j-1].size == 0)
5457 break;
5458 }
5459
5460 pos = 0;
5461 maxsize = 0;
5462 start_extent = 0;
5463 i = 0;
5464 do {
5465 unsigned long long esize;
5466
5467 esize = e[i].start - pos;
5468 if (esize >= maxsize) {
5469 maxsize = esize;
5470 start = pos;
5471 start_extent = i;
5472 }
5473 pos = e[i].start + e[i].size;
5474 i++;
5475 } while (e[i-1].size);
5476 free(e);
5477
a7dd165b
DW
5478 if (maxsize == 0)
5479 return 0;
5480
5481 /* FIXME assumes volume at offset 0 is the first volume in a
5482 * container
5483 */
0dcecb2e
DW
5484 if (start_extent > 0)
5485 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
5486 else
5487 reserve = 0;
5488
5489 if (maxsize < reserve)
a7dd165b 5490 return 0;
0dcecb2e 5491
5551b113 5492 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 5493 if (start + reserve > super->create_offset)
a7dd165b 5494 return 0; /* start overflows create_offset */
0dcecb2e
DW
5495 super->create_offset = start + reserve;
5496
5497 return maxsize - reserve;
5498}
5499
88c32bb1
DW
5500static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
5501{
5502 if (level < 0 || level == 6 || level == 4)
5503 return 0;
5504
5505 /* if we have an orom prevent invalid raid levels */
5506 if (orom)
5507 switch (level) {
5508 case 0: return imsm_orom_has_raid0(orom);
5509 case 1:
5510 if (raiddisks > 2)
5511 return imsm_orom_has_raid1e(orom);
1c556e92
DW
5512 return imsm_orom_has_raid1(orom) && raiddisks == 2;
5513 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
5514 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW
5515 }
5516 else
5517 return 1; /* not on an Intel RAID platform so anything goes */
5518
5519 return 0;
5520}
5521
ca9de185
LM
5522
5523static int
5524active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
5525 int dpa, int verbose)
5526{
5527 struct mdstat_ent *mdstat = mdstat_read(0, 0);
5528 struct mdstat_ent *memb = NULL;
5529 int count = 0;
5530 int num = 0;
5531 struct md_list *dv = NULL;
5532 int found;
5533
5534 for (memb = mdstat ; memb ; memb = memb->next) {
5535 if (memb->metadata_version &&
5536 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
5537 (strcmp(&memb->metadata_version[9], name) == 0) &&
5538 !is_subarray(memb->metadata_version+9) &&
5539 memb->members) {
5540 struct dev_member *dev = memb->members;
5541 int fd = -1;
5542 while(dev && (fd < 0)) {
5543 char *path = malloc(strlen(dev->name) + strlen("/dev/") + 1);
5544 if (path) {
5545 num = sprintf(path, "%s%s", "/dev/", dev->name);
5546 if (num > 0)
5547 fd = open(path, O_RDONLY, 0);
5548 if ((num <= 0) || (fd < 0)) {
5549 pr_vrb(": Cannot open %s: %s\n",
5550 dev->name, strerror(errno));
5551 }
5552 free(path);
5553 }
5554 dev = dev->next;
5555 }
5556 found = 0;
5557 if ((fd >= 0) && disk_attached_to_hba(fd, hba)) {
5558 struct mdstat_ent *vol;
5559 for (vol = mdstat ; vol ; vol = vol->next) {
5560 if ((vol->active > 0) &&
5561 vol->metadata_version &&
5562 is_container_member(vol, memb->dev)) {
5563 found++;
5564 count++;
5565 }
5566 }
5567 if (*devlist && (found < dpa)) {
5568 dv = calloc(1, sizeof(*dv));
5569 if (dv == NULL)
5570 fprintf(stderr, Name ": calloc failed\n");
5571 else {
5572 dv->devname = malloc(strlen(memb->dev) + strlen("/dev/") + 1);
5573 if (dv->devname != NULL) {
5574 sprintf(dv->devname, "%s%s", "/dev/", memb->dev);
5575 dv->found = found;
5576 dv->used = 0;
5577 dv->next = *devlist;
5578 *devlist = dv;
5579 } else
5580 free(dv);
5581 }
5582 }
5583 }
5584 if (fd >= 0)
5585 close(fd);
5586 }
5587 }
5588 free_mdstat(mdstat);
5589 return count;
5590}
5591
5592#ifdef DEBUG_LOOP
5593static struct md_list*
5594get_loop_devices(void)
5595{
5596 int i;
5597 struct md_list *devlist = NULL;
5598 struct md_list *dv = NULL;
5599
5600 for(i = 0; i < 12; i++) {
5601 dv = calloc(1, sizeof(*dv));
5602 if (dv == NULL) {
5603 fprintf(stderr, Name ": calloc failed\n");
5604 break;
5605 }
5606 dv->devname = malloc(40);
5607 if (dv->devname == NULL) {
5608 fprintf(stderr, Name ": malloc failed\n");
5609 free(dv);
5610 break;
5611 }
5612 sprintf(dv->devname, "/dev/loop%d", i);
5613 dv->next = devlist;
5614 devlist = dv;
5615 }
5616 return devlist;
5617}
5618#endif
5619
5620static struct md_list*
5621get_devices(const char *hba_path)
5622{
5623 struct md_list *devlist = NULL;
5624 struct md_list *dv = NULL;
5625 struct dirent *ent;
5626 DIR *dir;
5627 int err = 0;
5628
5629#if DEBUG_LOOP
5630 devlist = get_loop_devices();
5631 return devlist;
5632#endif
5633 /* scroll through /sys/dev/block looking for devices attached to
5634 * this hba
5635 */
5636 dir = opendir("/sys/dev/block");
5637 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
5638 int fd;
5639 char buf[1024];
5640 int major, minor;
5641 char *path = NULL;
5642 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
5643 continue;
5644 path = devt_to_devpath(makedev(major, minor));
5645 if (!path)
5646 continue;
5647 if (!path_attached_to_hba(path, hba_path)) {
5648 free(path);
5649 path = NULL;
5650 continue;
5651 }
5652 free(path);
5653 path = NULL;
5654 fd = dev_open(ent->d_name, O_RDONLY);
5655 if (fd >= 0) {
5656 fd2devname(fd, buf);
5657 close(fd);
5658 } else {
5659 fprintf(stderr, Name ": cannot open device: %s\n",
5660 ent->d_name);
5661 continue;
5662 }
5663
5664
5665 dv = calloc(1, sizeof(*dv));
5666 if (dv == NULL) {
5667 fprintf(stderr, Name ": malloc failed\n");
5668 err = 1;
5669 break;
5670 }
5671 dv->devname = strdup(buf);
5672 if (dv->devname == NULL) {
5673 fprintf(stderr, Name ": malloc failed\n");
5674 err = 1;
5675 free(dv);
5676 break;
5677 }
5678 dv->next = devlist;
5679 devlist = dv;
5680 }
5681 if (err) {
5682 while(devlist) {
5683 dv = devlist;
5684 devlist = devlist->next;
5685 free(dv->devname);
5686 free(dv);
5687 }
5688 }
562aa102 5689 closedir(dir);
ca9de185
LM
5690 return devlist;
5691}
5692
5693static int
5694count_volumes_list(struct md_list *devlist, char *homehost,
5695 int verbose, int *found)
5696{
5697 struct md_list *tmpdev;
5698 int count = 0;
5699 struct supertype *st = NULL;
5700
5701 /* first walk the list of devices to find a consistent set
5702 * that match the criterea, if that is possible.
5703 * We flag the ones we like with 'used'.
5704 */
5705 *found = 0;
5706 st = match_metadata_desc_imsm("imsm");
5707 if (st == NULL) {
5708 pr_vrb(": cannot allocate memory for imsm supertype\n");
5709 return 0;
5710 }
5711
5712 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5713 char *devname = tmpdev->devname;
5714 struct stat stb;
5715 struct supertype *tst;
5716 int dfd;
5717 if (tmpdev->used > 1)
5718 continue;
5719 tst = dup_super(st);
5720 if (tst == NULL) {
5721 pr_vrb(": cannot allocate memory for imsm supertype\n");
5722 goto err_1;
5723 }
5724 tmpdev->container = 0;
5725 dfd = dev_open(devname, O_RDONLY|O_EXCL);
5726 if (dfd < 0) {
5727 dprintf(": cannot open device %s: %s\n",
5728 devname, strerror(errno));
5729 tmpdev->used = 2;
5730 } else if (fstat(dfd, &stb)< 0) {
5731 /* Impossible! */
5732 dprintf(": fstat failed for %s: %s\n",
5733 devname, strerror(errno));
5734 tmpdev->used = 2;
5735 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
5736 dprintf(": %s is not a block device.\n",
5737 devname);
5738 tmpdev->used = 2;
5739 } else if (must_be_container(dfd)) {
5740 struct supertype *cst;
5741 cst = super_by_fd(dfd, NULL);
5742 if (cst == NULL) {
5743 dprintf(": cannot recognize container type %s\n",
5744 devname);
5745 tmpdev->used = 2;
5746 } else if (tst->ss != st->ss) {
5747 dprintf(": non-imsm container - ignore it: %s\n",
5748 devname);
5749 tmpdev->used = 2;
5750 } else if (!tst->ss->load_container ||
5751 tst->ss->load_container(tst, dfd, NULL))
5752 tmpdev->used = 2;
5753 else {
5754 tmpdev->container = 1;
5755 }
5756 if (cst)
5757 cst->ss->free_super(cst);
5758 } else {
5759 tmpdev->st_rdev = stb.st_rdev;
5760 if (tst->ss->load_super(tst,dfd, NULL)) {
5761 dprintf(": no RAID superblock on %s\n",
5762 devname);
5763 tmpdev->used = 2;
5764 } else if (tst->ss->compare_super == NULL) {
5765 dprintf(": Cannot assemble %s metadata on %s\n",
5766 tst->ss->name, devname);
5767 tmpdev->used = 2;
5768 }
5769 }
5770 if (dfd >= 0)
5771 close(dfd);
5772 if (tmpdev->used == 2 || tmpdev->used == 4) {
5773 /* Ignore unrecognised devices during auto-assembly */
5774 goto loop;
5775 }
5776 else {
5777 struct mdinfo info;
5778 tst->ss->getinfo_super(tst, &info, NULL);
5779
5780 if (st->minor_version == -1)
5781 st->minor_version = tst->minor_version;
5782
5783 if (memcmp(info.uuid, uuid_zero,
5784 sizeof(int[4])) == 0) {
5785 /* this is a floating spare. It cannot define
5786 * an array unless there are no more arrays of
5787 * this type to be found. It can be included
5788 * in an array of this type though.
5789 */
5790 tmpdev->used = 3;
5791 goto loop;
5792 }
5793
5794 if (st->ss != tst->ss ||
5795 st->minor_version != tst->minor_version ||
5796 st->ss->compare_super(st, tst) != 0) {
5797 /* Some mismatch. If exactly one array matches this host,
5798 * we can resolve on that one.
5799 * Or, if we are auto assembling, we just ignore the second
5800 * for now.
5801 */
5802 dprintf(": superblock on %s doesn't match others - assembly aborted\n",
5803 devname);
5804 goto loop;
5805 }
5806 tmpdev->used = 1;
5807 *found = 1;
5808 dprintf("found: devname: %s\n", devname);
5809 }
5810 loop:
5811 if (tst)
5812 tst->ss->free_super(tst);
5813 }
5814 if (*found != 0) {
5815 int err;
5816 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
5817 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
5818 for (iter = head; iter; iter = iter->next) {
5819 dprintf("content->text_version: %s vol\n",
5820 iter->text_version);
5821 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
5822 /* do not assemble arrays with unsupported
5823 configurations */
5824 dprintf(": Cannot activate member %s.\n",
5825 iter->text_version);
5826 } else
5827 count++;
5828 }
5829 sysfs_free(head);
5830
5831 } else {
5832 dprintf(" no valid super block on device list: err: %d %p\n",
5833 err, st->sb);
5834 }
5835 } else {
5836 dprintf(" no more devices to examin\n");
5837 }
5838
5839 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5840 if ((tmpdev->used == 1) && (tmpdev->found)) {
5841 if (count) {
5842 if (count < tmpdev->found)
5843 count = 0;
5844 else
5845 count -= tmpdev->found;
5846 }
5847 }
5848 if (tmpdev->used == 1)
5849 tmpdev->used = 4;
5850 }
5851 err_1:
5852 if (st)
5853 st->ss->free_super(st);
5854 return count;
5855}
5856
5857
5858static int
5859count_volumes(char *hba, int dpa, int verbose)
5860{
5861 struct md_list *devlist = NULL;
5862 int count = 0;
5863 int found = 0;;
5864
5865 devlist = get_devices(hba);
5866 /* if no intel devices return zero volumes */
5867 if (devlist == NULL)
5868 return 0;
5869
5870 count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose);
5871 dprintf(" path: %s active arrays: %d\n", hba, count);
5872 if (devlist == NULL)
5873 return 0;
5874 do {
5875 found = 0;
5876 count += count_volumes_list(devlist,
5877 NULL,
5878 verbose,
5879 &found);
5880 dprintf("found %d count: %d\n", found, count);
5881 } while (found);
5882
5883 dprintf("path: %s total number of volumes: %d\n", hba, count);
5884
5885 while(devlist) {
5886 struct md_list *dv = devlist;
5887 devlist = devlist->next;
5888 free(dv->devname);
5889 free(dv);
5890 }
5891 return count;
5892}
5893
cd9d1ac7
DW
5894static int imsm_default_chunk(const struct imsm_orom *orom)
5895{
5896 /* up to 512 if the plaform supports it, otherwise the platform max.
5897 * 128 if no platform detected
5898 */
5899 int fs = max(7, orom ? fls(orom->sss) : 0);
5900
5901 return min(512, (1 << fs));
5902}
73408129 5903
6592ce37
DW
5904static int
5905validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 5906 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 5907{
660260d0
DW
5908 /* check/set platform and metadata limits/defaults */
5909 if (super->orom && raiddisks > super->orom->dpa) {
5910 pr_vrb(": platform supports a maximum of %d disks per array\n",
5911 super->orom->dpa);
73408129
LM
5912 return 0;
5913 }
5914
5915 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 5916 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
6592ce37
DW
5917 pr_vrb(": platform does not support raid%d with %d disk%s\n",
5918 level, raiddisks, raiddisks > 1 ? "s" : "");
5919 return 0;
5920 }
cd9d1ac7
DW
5921
5922 if (chunk && (*chunk == 0 || *chunk == UnSet))
5923 *chunk = imsm_default_chunk(super->orom);
5924
5925 if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) {
5926 pr_vrb(": platform does not support a chunk size of: "
5927 "%d\n", *chunk);
5928 return 0;
6592ce37 5929 }
cd9d1ac7 5930
6592ce37
DW
5931 if (layout != imsm_level_to_layout(level)) {
5932 if (level == 5)
5933 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
5934 else if (level == 10)
5935 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
5936 else
5937 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
5938 layout, level);
5939 return 0;
5940 }
2cc699af
CA
5941
5942 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && chunk &&
5943 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
5944 pr_vrb(": platform does not support a volume size over 2TB\n");
5945 return 0;
5946 }
6592ce37
DW
5947 return 1;
5948}
5949
c2c087e6
DW
5950/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
5951 * FIX ME add ahci details
5952 */
8b353278 5953static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 5954 int layout, int raiddisks, int *chunk,
c2c087e6 5955 unsigned long long size, char *dev,
2c514b71
NB
5956 unsigned long long *freesize,
5957 int verbose)
cdddbdbc 5958{
c2c087e6
DW
5959 struct stat stb;
5960 struct intel_super *super = st->sb;
b2916f25 5961 struct imsm_super *mpb;
c2c087e6
DW
5962 struct dl *dl;
5963 unsigned long long pos = 0;
5964 unsigned long long maxsize;
5965 struct extent *e;
5966 int i;
cdddbdbc 5967
88c32bb1
DW
5968 /* We must have the container info already read in. */
5969 if (!super)
c2c087e6
DW
5970 return 0;
5971
b2916f25
JS
5972 mpb = super->anchor;
5973
2cc699af 5974 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
d54559f0
LM
5975 fprintf(stderr, Name ": RAID gemetry validation failed. "
5976 "Cannot proceed with the action(s).\n");
c2c087e6 5977 return 0;
d54559f0 5978 }
c2c087e6
DW
5979 if (!dev) {
5980 /* General test: make sure there is space for
2da8544a
DW
5981 * 'raiddisks' device extents of size 'size' at a given
5982 * offset
c2c087e6 5983 */
e46273eb 5984 unsigned long long minsize = size;
b7528a20 5985 unsigned long long start_offset = MaxSector;
c2c087e6
DW
5986 int dcnt = 0;
5987 if (minsize == 0)
5988 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
5989 for (dl = super->disks; dl ; dl = dl->next) {
5990 int found = 0;
5991
bf5a934a 5992 pos = 0;
c2c087e6
DW
5993 i = 0;
5994 e = get_extents(super, dl);
5995 if (!e) continue;
5996 do {
5997 unsigned long long esize;
5998 esize = e[i].start - pos;
5999 if (esize >= minsize)
6000 found = 1;
b7528a20 6001 if (found && start_offset == MaxSector) {
2da8544a
DW
6002 start_offset = pos;
6003 break;
6004 } else if (found && pos != start_offset) {
6005 found = 0;
6006 break;
6007 }
c2c087e6
DW
6008 pos = e[i].start + e[i].size;
6009 i++;
6010 } while (e[i-1].size);
6011 if (found)
6012 dcnt++;
6013 free(e);
6014 }
6015 if (dcnt < raiddisks) {
2c514b71
NB
6016 if (verbose)
6017 fprintf(stderr, Name ": imsm: Not enough "
6018 "devices with space for this array "
6019 "(%d < %d)\n",
6020 dcnt, raiddisks);
c2c087e6
DW
6021 return 0;
6022 }
6023 return 1;
6024 }
0dcecb2e 6025
c2c087e6
DW
6026 /* This device must be a member of the set */
6027 if (stat(dev, &stb) < 0)
6028 return 0;
6029 if ((S_IFMT & stb.st_mode) != S_IFBLK)
6030 return 0;
6031 for (dl = super->disks ; dl ; dl = dl->next) {
f21e18ca
N
6032 if (dl->major == (int)major(stb.st_rdev) &&
6033 dl->minor == (int)minor(stb.st_rdev))
c2c087e6
DW
6034 break;
6035 }
6036 if (!dl) {
2c514b71
NB
6037 if (verbose)
6038 fprintf(stderr, Name ": %s is not in the "
6039 "same imsm set\n", dev);
c2c087e6 6040 return 0;
a20d2ba5
DW
6041 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
6042 /* If a volume is present then the current creation attempt
6043 * cannot incorporate new spares because the orom may not
6044 * understand this configuration (all member disks must be
6045 * members of each array in the container).
6046 */
6047 fprintf(stderr, Name ": %s is a spare and a volume"
6048 " is already defined for this container\n", dev);
6049 fprintf(stderr, Name ": The option-rom requires all member"
6050 " disks to be a member of all volumes\n");
6051 return 0;
5fe62b94
WD
6052 } else if (super->orom && mpb->num_raid_devs > 0 &&
6053 mpb->num_disks != raiddisks) {
6054 fprintf(stderr, Name ": The option-rom requires all member"
6055 " disks to be a member of all volumes\n");
6056 return 0;
c2c087e6 6057 }
0dcecb2e
DW
6058
6059 /* retrieve the largest free space block */
c2c087e6
DW
6060 e = get_extents(super, dl);
6061 maxsize = 0;
6062 i = 0;
0dcecb2e
DW
6063 if (e) {
6064 do {
6065 unsigned long long esize;
6066
6067 esize = e[i].start - pos;
6068 if (esize >= maxsize)
6069 maxsize = esize;
6070 pos = e[i].start + e[i].size;
6071 i++;
6072 } while (e[i-1].size);
6073 dl->e = e;
6074 dl->extent_cnt = i;
6075 } else {
6076 if (verbose)
6077 fprintf(stderr, Name ": unable to determine free space for: %s\n",
6078 dev);
6079 return 0;
6080 }
6081 if (maxsize < size) {
6082 if (verbose)
6083 fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n",
6084 dev, maxsize, size);
6085 return 0;
6086 }
6087
6088 /* count total number of extents for merge */
6089 i = 0;
6090 for (dl = super->disks; dl; dl = dl->next)
6091 if (dl->e)
6092 i += dl->extent_cnt;
6093
6094 maxsize = merge_extents(super, i);
3baa56ab
LO
6095
6096 if (!check_env("IMSM_NO_PLATFORM") &&
6097 mpb->num_raid_devs > 0 && size && size != maxsize) {
6098 fprintf(stderr, Name ": attempting to create a second "
6099 "volume with size less then remaining space. "
6100 "Aborting...\n");
6101 return 0;
6102 }
6103
a7dd165b 6104 if (maxsize < size || maxsize == 0) {
b3071342
LD
6105 if (verbose) {
6106 if (maxsize == 0)
6107 fprintf(stderr, Name ": no free space"
6108 " left on device. Aborting...\n");
6109 else
6110 fprintf(stderr, Name ": not enough space"
6111 " to create volume of given size"
6112 " (%llu < %llu). Aborting...\n",
6113 maxsize, size);
6114 }
0dcecb2e 6115 return 0;
0dcecb2e
DW
6116 }
6117
c2c087e6
DW
6118 *freesize = maxsize;
6119
ca9de185
LM
6120 if (super->orom) {
6121 int count = count_volumes(super->hba->path,
6122 super->orom->dpa, verbose);
6123 if (super->orom->vphba <= count) {
40110b90 6124 pr_vrb(": platform does not support more than %d raid volumes.\n",
ca9de185
LM
6125 super->orom->vphba);
6126 return 0;
6127 }
6128 }
c2c087e6 6129 return 1;
cdddbdbc
DW
6130}
6131
13bcac90 6132static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW
6133 unsigned long long size, int chunk,
6134 unsigned long long *freesize)
6135{
6136 struct intel_super *super = st->sb;
6137 struct imsm_super *mpb = super->anchor;
6138 struct dl *dl;
6139 int i;
6140 int extent_cnt;
6141 struct extent *e;
6142 unsigned long long maxsize;
6143 unsigned long long minsize;
6144 int cnt;
6145 int used;
6146
6147 /* find the largest common start free region of the possible disks */
6148 used = 0;
6149 extent_cnt = 0;
6150 cnt = 0;
6151 for (dl = super->disks; dl; dl = dl->next) {
6152 dl->raiddisk = -1;
6153
6154 if (dl->index >= 0)
6155 used++;
6156
6157 /* don't activate new spares if we are orom constrained
6158 * and there is already a volume active in the container
6159 */
6160 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
6161 continue;
6162
6163 e = get_extents(super, dl);
6164 if (!e)
6165 continue;
6166 for (i = 1; e[i-1].size; i++)
6167 ;
6168 dl->e = e;
6169 dl->extent_cnt = i;
6170 extent_cnt += i;
6171 cnt++;
6172 }
6173
6174 maxsize = merge_extents(super, extent_cnt);
6175 minsize = size;
6176 if (size == 0)
612e59d8
CA
6177 /* chunk is in K */
6178 minsize = chunk * 2;
efb30e7f
DW
6179
6180 if (cnt < raiddisks ||
6181 (super->orom && used && used != raiddisks) ||
a7dd165b
DW
6182 maxsize < minsize ||
6183 maxsize == 0) {
efb30e7f
DW
6184 fprintf(stderr, Name ": not enough devices with space to create array.\n");
6185 return 0; /* No enough free spaces large enough */
6186 }
6187
6188 if (size == 0) {
6189 size = maxsize;
6190 if (chunk) {
612e59d8
CA
6191 size /= 2 * chunk;
6192 size *= 2 * chunk;
efb30e7f 6193 }
f878b242
LM
6194 maxsize = size;
6195 }
6196 if (!check_env("IMSM_NO_PLATFORM") &&
6197 mpb->num_raid_devs > 0 && size && size != maxsize) {
6198 fprintf(stderr, Name ": attempting to create a second "
6199 "volume with size less then remaining space. "
6200 "Aborting...\n");
6201 return 0;
efb30e7f 6202 }
efb30e7f
DW
6203 cnt = 0;
6204 for (dl = super->disks; dl; dl = dl->next)
6205 if (dl->e)
6206 dl->raiddisk = cnt++;
6207
6208 *freesize = size;
6209
13bcac90
AK
6210 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
6211
efb30e7f
DW
6212 return 1;
6213}
6214
13bcac90
AK
6215static int reserve_space(struct supertype *st, int raiddisks,
6216 unsigned long long size, int chunk,
6217 unsigned long long *freesize)
6218{
6219 struct intel_super *super = st->sb;
6220 struct dl *dl;
6221 int cnt;
6222 int rv = 0;
6223
6224 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
6225 if (rv) {
6226 cnt = 0;
6227 for (dl = super->disks; dl; dl = dl->next)
6228 if (dl->e)
6229 dl->raiddisk = cnt++;
6230 rv = 1;
6231 }
6232
6233 return rv;
6234}
6235
bf5a934a 6236static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 6237 int raiddisks, int *chunk, unsigned long long size,
bf5a934a
DW
6238 char *dev, unsigned long long *freesize,
6239 int verbose)
6240{
6241 int fd, cfd;
6242 struct mdinfo *sra;
20cbe8d2 6243 int is_member = 0;
bf5a934a 6244
d54559f0
LM
6245 /* load capability
6246 * if given unused devices create a container
bf5a934a
DW
6247 * if given given devices in a container create a member volume
6248 */
6249 if (level == LEVEL_CONTAINER) {
6250 /* Must be a fresh device to add to a container */
6251 return validate_geometry_imsm_container(st, level, layout,
c21e737b
CA
6252 raiddisks,
6253 chunk?*chunk:0, size,
bf5a934a
DW
6254 dev, freesize,
6255 verbose);
6256 }
9587c373 6257
8592f29d 6258 if (!dev) {
e91a3bad 6259 if (st->sb) {
ca9de185 6260 struct intel_super *super = st->sb;
e91a3bad 6261 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 6262 raiddisks, chunk, size,
e91a3bad
LM
6263 verbose))
6264 return 0;
efb30e7f
DW
6265 /* we are being asked to automatically layout a
6266 * new volume based on the current contents of
6267 * the container. If the the parameters can be
6268 * satisfied reserve_space will record the disks,
6269 * start offset, and size of the volume to be
6270 * created. add_to_super and getinfo_super
6271 * detect when autolayout is in progress.
6272 */
ca9de185
LM
6273 /* assuming that freesize is always given when array is
6274 created */
6275 if (super->orom && freesize) {
6276 int count;
6277 count = count_volumes(super->hba->path,
6278 super->orom->dpa, verbose);
6279 if (super->orom->vphba <= count) {
6280 pr_vrb(": platform does not support more"
40110b90 6281 " than %d raid volumes.\n",
ca9de185
LM
6282 super->orom->vphba);
6283 return 0;
6284 }
6285 }
e91a3bad
LM
6286 if (freesize)
6287 return reserve_space(st, raiddisks, size,
6288 chunk?*chunk:0, freesize);
8592f29d
N
6289 }
6290 return 1;
6291 }
bf5a934a
DW
6292 if (st->sb) {
6293 /* creating in a given container */
6294 return validate_geometry_imsm_volume(st, level, layout,
6295 raiddisks, chunk, size,
6296 dev, freesize, verbose);
6297 }
6298
bf5a934a
DW
6299 /* This device needs to be a device in an 'imsm' container */
6300 fd = open(dev, O_RDONLY|O_EXCL, 0);
6301 if (fd >= 0) {
6302 if (verbose)
6303 fprintf(stderr,
6304 Name ": Cannot create this array on device %s\n",
6305 dev);
6306 close(fd);
6307 return 0;
6308 }
6309 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
6310 if (verbose)
6311 fprintf(stderr, Name ": Cannot open %s: %s\n",
6312 dev, strerror(errno));
6313 return 0;
6314 }
6315 /* Well, it is in use by someone, maybe an 'imsm' container. */
6316 cfd = open_container(fd);
20cbe8d2 6317 close(fd);
bf5a934a 6318 if (cfd < 0) {
bf5a934a
DW
6319 if (verbose)
6320 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
6321 dev);
6322 return 0;
6323 }
6324 sra = sysfs_read(cfd, 0, GET_VERSION);
bf5a934a 6325 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW
6326 strcmp(sra->text_version, "imsm") == 0)
6327 is_member = 1;
6328 sysfs_free(sra);
6329 if (is_member) {
bf5a934a
DW
6330 /* This is a member of a imsm container. Load the container
6331 * and try to create a volume
6332 */
6333 struct intel_super *super;
6334
ec50f7b6 6335 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a
DW
6336 st->sb = super;
6337 st->container_dev = fd2devnum(cfd);
6338 close(cfd);
6339 return validate_geometry_imsm_volume(st, level, layout,
6340 raiddisks, chunk,
6341 size, dev,
ecbd9e81
N
6342 freesize, 1)
6343 ? 1 : -1;
bf5a934a 6344 }
20cbe8d2 6345 }
bf5a934a 6346
20cbe8d2
AW
6347 if (verbose)
6348 fprintf(stderr, Name ": failed container membership check\n");
6349
6350 close(cfd);
6351 return 0;
bf5a934a 6352}
0bd16cf2 6353
30f58b22 6354static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ
6355{
6356 struct intel_super *super = st->sb;
6357
30f58b22
DW
6358 if (level && *level == UnSet)
6359 *level = LEVEL_CONTAINER;
6360
6361 if (level && layout && *layout == UnSet)
6362 *layout = imsm_level_to_layout(*level);
0bd16cf2 6363
cd9d1ac7
DW
6364 if (chunk && (*chunk == UnSet || *chunk == 0))
6365 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ
6366}
6367
33414a01
DW
6368static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
6369
6370static int kill_subarray_imsm(struct supertype *st)
6371{
6372 /* remove the subarray currently referenced by ->current_vol */
6373 __u8 i;
6374 struct intel_dev **dp;
6375 struct intel_super *super = st->sb;
6376 __u8 current_vol = super->current_vol;
6377 struct imsm_super *mpb = super->anchor;
6378
6379 if (super->current_vol < 0)
6380 return 2;
6381 super->current_vol = -1; /* invalidate subarray cursor */
6382
6383 /* block deletions that would change the uuid of active subarrays
6384 *
6385 * FIXME when immutable ids are available, but note that we'll
6386 * also need to fixup the invalidated/active subarray indexes in
6387 * mdstat
6388 */
6389 for (i = 0; i < mpb->num_raid_devs; i++) {
6390 char subarray[4];
6391
6392 if (i < current_vol)
6393 continue;
6394 sprintf(subarray, "%u", i);
6395 if (is_subarray_active(subarray, st->devname)) {
6396 fprintf(stderr,
6397 Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
6398 current_vol, i);
6399
6400 return 2;
6401 }
6402 }
6403
6404 if (st->update_tail) {
6405 struct imsm_update_kill_array *u = malloc(sizeof(*u));
6406
6407 if (!u)
6408 return 2;
6409 u->type = update_kill_array;
6410 u->dev_idx = current_vol;
6411 append_metadata_update(st, u, sizeof(*u));
6412
6413 return 0;
6414 }
6415
6416 for (dp = &super->devlist; *dp;)
6417 if ((*dp)->index == current_vol) {
6418 *dp = (*dp)->next;
6419 } else {
6420 handle_missing(super, (*dp)->dev);
6421 if ((*dp)->index > current_vol)
6422 (*dp)->index--;
6423 dp = &(*dp)->next;
6424 }
6425
6426 /* no more raid devices, all active components are now spares,
6427 * but of course failed are still failed
6428 */
6429 if (--mpb->num_raid_devs == 0) {
6430 struct dl *d;
6431
6432 for (d = super->disks; d; d = d->next)
a8619d23
AK
6433 if (d->index > -2)
6434 mark_spare(d);
33414a01
DW
6435 }
6436
6437 super->updates_pending++;
6438
6439 return 0;
6440}
aa534678 6441
a951a4f7 6442static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 6443 char *update, struct mddev_ident *ident)
aa534678
DW
6444{
6445 /* update the subarray currently referenced by ->current_vol */
6446 struct intel_super *super = st->sb;
6447 struct imsm_super *mpb = super->anchor;
6448
aa534678
DW
6449 if (strcmp(update, "name") == 0) {
6450 char *name = ident->name;
a951a4f7
N
6451 char *ep;
6452 int vol;
aa534678 6453
a951a4f7 6454 if (is_subarray_active(subarray, st->devname)) {
aa534678
DW
6455 fprintf(stderr,
6456 Name ": Unable to update name of active subarray\n");
6457 return 2;
6458 }
6459
6460 if (!check_name(super, name, 0))
6461 return 2;
6462
a951a4f7
N
6463 vol = strtoul(subarray, &ep, 10);
6464 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
6465 return 2;
6466
aa534678
DW
6467 if (st->update_tail) {
6468 struct imsm_update_rename_array *u = malloc(sizeof(*u));
6469
6470 if (!u)
6471 return 2;
6472 u->type = update_rename_array;
a951a4f7 6473 u->dev_idx = vol;
aa534678
DW
6474 snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name);
6475 append_metadata_update(st, u, sizeof(*u));
6476 } else {
6477 struct imsm_dev *dev;
6478 int i;
6479
a951a4f7 6480 dev = get_imsm_dev(super, vol);
aa534678
DW
6481 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
6482 for (i = 0; i < mpb->num_raid_devs; i++) {
6483 dev = get_imsm_dev(super, i);
6484 handle_missing(super, dev);
6485 }
6486 super->updates_pending++;
6487 }
6488 } else
6489 return 2;
6490
6491 return 0;
6492}
d1e02575 6493#endif /* MDASSEMBLE */
bf5a934a 6494
28bce06f
AK
6495static int is_gen_migration(struct imsm_dev *dev)
6496{
7534230b
AK
6497 if (dev == NULL)
6498 return 0;
6499
28bce06f
AK
6500 if (!dev->vol.migr_state)
6501 return 0;
6502
6503 if (migr_type(dev) == MIGR_GEN_MIGR)
6504 return 1;
6505
6506 return 0;
6507}
6508
1e5c6983
DW
6509static int is_rebuilding(struct imsm_dev *dev)
6510{
6511 struct imsm_map *migr_map;
6512
6513 if (!dev->vol.migr_state)
6514 return 0;
6515
6516 if (migr_type(dev) != MIGR_REBUILD)
6517 return 0;
6518
238c0a71 6519 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW
6520
6521 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
6522 return 1;
6523 else
6524 return 0;
6525}
6526
b4ab44d8 6527#ifndef MDASSEMBLE
6ce1fbf1
AK
6528static int is_initializing(struct imsm_dev *dev)
6529{
6530 struct imsm_map *migr_map;
6531
6532 if (!dev->vol.migr_state)
6533 return 0;
6534
6535 if (migr_type(dev) != MIGR_INIT)
6536 return 0;
6537
238c0a71 6538 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK
6539
6540 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
6541 return 1;
6542
6543 return 0;
6ce1fbf1 6544}
b4ab44d8 6545#endif
6ce1fbf1 6546
c47b0ff6
AK
6547static void update_recovery_start(struct intel_super *super,
6548 struct imsm_dev *dev,
6549 struct mdinfo *array)
1e5c6983
DW
6550{
6551 struct mdinfo *rebuild = NULL;
6552 struct mdinfo *d;
6553 __u32 units;
6554
6555 if (!is_rebuilding(dev))
6556 return;
6557
6558 /* Find the rebuild target, but punt on the dual rebuild case */
6559 for (d = array->devs; d; d = d->next)
6560 if (d->recovery_start == 0) {
6561 if (rebuild)
6562 return;
6563 rebuild = d;
6564 }
6565
4363fd80
DW
6566 if (!rebuild) {
6567 /* (?) none of the disks are marked with
6568 * IMSM_ORD_REBUILD, so assume they are missing and the
6569 * disk_ord_tbl was not correctly updated
6570 */
6571 dprintf("%s: failed to locate out-of-sync disk\n", __func__);
6572 return;
6573 }
6574
1e5c6983 6575 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 6576 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW
6577}
6578
9e2d750d 6579#ifndef MDASSEMBLE
276d77db 6580static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
9e2d750d 6581#endif
1e5c6983 6582
00bbdbda 6583static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 6584{
4f5bc454
DW
6585 /* Given a container loaded by load_super_imsm_all,
6586 * extract information about all the arrays into
6587 * an mdinfo tree.
00bbdbda 6588 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW
6589 *
6590 * For each imsm_dev create an mdinfo, fill it in,
6591 * then look for matching devices in super->disks
6592 * and create appropriate device mdinfo.
6593 */
6594 struct intel_super *super = st->sb;
949c47a0 6595 struct imsm_super *mpb = super->anchor;
4f5bc454 6596 struct mdinfo *rest = NULL;
00bbdbda 6597 unsigned int i;
81219e70 6598 int sb_errors = 0;
abef11a3
AK
6599 struct dl *d;
6600 int spare_disks = 0;
cdddbdbc 6601
19482bcc
AK
6602 /* do not assemble arrays when not all attributes are supported */
6603 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70
LM
6604 sb_errors = 1;
6605 fprintf(stderr, Name ": Unsupported attributes in IMSM metadata."
6606 "Arrays activation is blocked.\n");
19482bcc
AK
6607 }
6608
a06d022d 6609 /* check for bad blocks */
81219e70
LM
6610 if (imsm_bbm_log_size(super->anchor)) {
6611 fprintf(stderr, Name ": BBM log found in IMSM metadata."
6612 "Arrays activation is blocked.\n");
6613 sb_errors = 1;
6614 }
6615
604b746f 6616
abef11a3
AK
6617 /* count spare devices, not used in maps
6618 */
6619 for (d = super->disks; d; d = d->next)
6620 if (d->index == -1)
6621 spare_disks++;
6622
4f5bc454 6623 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N
6624 struct imsm_dev *dev;
6625 struct imsm_map *map;
86e3692b 6626 struct imsm_map *map2;
4f5bc454 6627 struct mdinfo *this;
a6482415
N
6628 int slot;
6629#ifndef MDASSEMBLE
6630 int chunk;
6631#endif
00bbdbda
N
6632 char *ep;
6633
6634 if (subarray &&
6635 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
6636 continue;
6637
6638 dev = get_imsm_dev(super, i);
238c0a71
AK
6639 map = get_imsm_map(dev, MAP_0);
6640 map2 = get_imsm_map(dev, MAP_1);
4f5bc454 6641
1ce0101c
DW
6642 /* do not publish arrays that are in the middle of an
6643 * unsupported migration
6644 */
6645 if (dev->vol.migr_state &&
28bce06f 6646 (migr_type(dev) == MIGR_STATE_CHANGE)) {
1ce0101c
DW
6647 fprintf(stderr, Name ": cannot assemble volume '%.16s':"
6648 " unsupported migration in progress\n",
6649 dev->volume);
6650 continue;
6651 }
2db86302
LM
6652 /* do not publish arrays that are not support by controller's
6653 * OROM/EFI
6654 */
1ce0101c 6655
4f5bc454 6656 this = malloc(sizeof(*this));
0fbd635c 6657 if (!this) {
cf1be220 6658 fprintf(stderr, Name ": failed to allocate %zu bytes\n",
0fbd635c
AW
6659 sizeof(*this));
6660 break;
6661 }
4f5bc454 6662
301406c9 6663 super->current_vol = i;
a5d85af7 6664 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 6665 this->next = rest;
81219e70 6666#ifndef MDASSEMBLE
a6482415 6667 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM
6668 /* mdadm does not support all metadata features- set the bit in all arrays state */
6669 if (!validate_geometry_imsm_orom(super,
6670 get_imsm_raid_level(map), /* RAID level */
6671 imsm_level_to_layout(get_imsm_raid_level(map)),
6672 map->num_members, /* raid disks */
2cc699af 6673 &chunk, join_u32(dev->size_low, dev->size_high),
81219e70 6674 1 /* verbose */)) {
446894ea
N
6675 fprintf(stderr, Name ": IMSM RAID geometry validation"
6676 " failed. Array %s activation is blocked.\n",
81219e70
LM
6677 dev->volume);
6678 this->array.state |=
6679 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
6680 (1<<MD_SB_BLOCK_VOLUME);
6681 }
6682#endif
6683
6684 /* if array has bad blocks, set suitable bit in all arrays state */
6685 if (sb_errors)
6686 this->array.state |=
6687 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
6688 (1<<MD_SB_BLOCK_VOLUME);
6689
4f5bc454 6690 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 6691 unsigned long long recovery_start;
4f5bc454
DW
6692 struct mdinfo *info_d;
6693 struct dl *d;
6694 int idx;
9a1608e5 6695 int skip;
7eef0453 6696 __u32 ord;
4f5bc454 6697
9a1608e5 6698 skip = 0;
238c0a71
AK
6699 idx = get_imsm_disk_idx(dev, slot, MAP_0);
6700 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW
6701 for (d = super->disks; d ; d = d->next)
6702 if (d->index == idx)
0fbd635c 6703 break;
4f5bc454 6704
1e5c6983 6705 recovery_start = MaxSector;
4f5bc454 6706 if (d == NULL)
9a1608e5 6707 skip = 1;
25ed7e59 6708 if (d && is_failed(&d->disk))
9a1608e5 6709 skip = 1;
7eef0453 6710 if (ord & IMSM_ORD_REBUILD)
1e5c6983 6711 recovery_start = 0;
9a1608e5
DW
6712
6713 /*
6714 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW
6715 * reset resync start to avoid a dirty-degraded
6716 * situation when performing the intial sync
9a1608e5
DW
6717 *
6718 * FIXME handle dirty degraded
6719 */
1e5c6983 6720 if ((skip || recovery_start == 0) && !dev->vol.dirty)
b7528a20 6721 this->resync_start = MaxSector;
9a1608e5
DW
6722 if (skip)
6723 continue;
4f5bc454 6724
1e5c6983 6725 info_d = calloc(1, sizeof(*info_d));
9a1608e5
DW
6726 if (!info_d) {
6727 fprintf(stderr, Name ": failed to allocate disk"
1ce0101c 6728 " for volume %.16s\n", dev->volume);
1e5c6983
DW
6729 info_d = this->devs;
6730 while (info_d) {
6731 struct mdinfo *d = info_d->next;
6732
6733 free(info_d);
6734 info_d = d;
6735 }
9a1608e5
DW
6736 free(this);
6737 this = rest;
6738 break;
6739 }
4f5bc454
DW
6740 info_d->next = this->devs;
6741 this->devs = info_d;
6742
4f5bc454
DW
6743 info_d->disk.number = d->index;
6744 info_d->disk.major = d->major;
6745 info_d->disk.minor = d->minor;
6746 info_d->disk.raid_disk = slot;
1e5c6983 6747 info_d->recovery_start = recovery_start;
86e3692b
AK
6748 if (map2) {
6749 if (slot < map2->num_members)
6750 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK
6751 else
6752 this->array.spare_disks++;
86e3692b
AK
6753 } else {
6754 if (slot < map->num_members)
6755 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK
6756 else
6757 this->array.spare_disks++;
86e3692b 6758 }
1e5c6983
DW
6759 if (info_d->recovery_start == MaxSector)
6760 this->array.working_disks++;
4f5bc454
DW
6761
6762 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113
CA
6763 info_d->data_offset = pba_of_lba0(map);
6764 info_d->component_size = blocks_per_member(map);
4f5bc454 6765 }
1e5c6983 6766 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 6767 update_recovery_start(super, dev, this);
abef11a3 6768 this->array.spare_disks += spare_disks;
276d77db 6769
9e2d750d 6770#ifndef MDASSEMBLE
276d77db
AK
6771 /* check for reshape */
6772 if (this->reshape_active == 1)
6773 recover_backup_imsm(st, this);
9e2d750d 6774#endif
9a1608e5 6775 rest = this;
4f5bc454
DW
6776 }
6777
6778 return rest;
cdddbdbc
DW
6779}
6780
845dea95 6781
3b451610
AK
6782static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
6783 int failed, int look_in_map)
c2a1e7da 6784{
3b451610
AK
6785 struct imsm_map *map;
6786
6787 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW
6788
6789 if (!failed)
3393c6af
DW
6790 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
6791 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW
6792
6793 switch (get_imsm_raid_level(map)) {
6794 case 0:
6795 return IMSM_T_STATE_FAILED;
6796 break;
6797 case 1:
6798 if (failed < map->num_members)
6799 return IMSM_T_STATE_DEGRADED;
6800 else
6801 return IMSM_T_STATE_FAILED;
6802 break;
6803 case 10:
6804 {
6805 /**
c92a2527
DW
6806 * check to see if any mirrors have failed, otherwise we
6807 * are degraded. Even numbered slots are mirrored on
6808 * slot+1
c2a1e7da 6809 */
c2a1e7da 6810 int i;
d9b420a5
N
6811 /* gcc -Os complains that this is unused */
6812 int insync = insync;
c2a1e7da
DW
6813
6814 for (i = 0; i < map->num_members; i++) {
238c0a71 6815 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW
6816 int idx = ord_to_idx(ord);
6817 struct imsm_disk *disk;
c2a1e7da 6818
c92a2527
DW
6819 /* reset the potential in-sync count on even-numbered
6820 * slots. num_copies is always 2 for imsm raid10
6821 */
6822 if ((i & 1) == 0)
6823 insync = 2;
c2a1e7da 6824
c92a2527 6825 disk = get_imsm_disk(super, idx);
25ed7e59 6826 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 6827 insync--;
c2a1e7da 6828
c92a2527
DW
6829 /* no in-sync disks left in this mirror the
6830 * array has failed
6831 */
6832 if (insync == 0)
6833 return IMSM_T_STATE_FAILED;
c2a1e7da
DW
6834 }
6835
6836 return IMSM_T_STATE_DEGRADED;
6837 }
6838 case 5:
6839 if (failed < 2)
6840 return IMSM_T_STATE_DEGRADED;
6841 else
6842 return IMSM_T_STATE_FAILED;
6843 break;
6844 default:
6845 break;
6846 }
6847
6848 return map->map_state;
6849}
6850
3b451610
AK
6851static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
6852 int look_in_map)
c2a1e7da
DW
6853{
6854 int i;
6855 int failed = 0;
6856 struct imsm_disk *disk;
d5985138
AK
6857 struct imsm_map *map = get_imsm_map(dev, MAP_0);
6858 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 6859 struct imsm_map *map_for_loop;
0556e1a2
DW
6860 __u32 ord;
6861 int idx;
d5985138 6862 int idx_1;
c2a1e7da 6863
0556e1a2
DW
6864 /* at the beginning of migration we set IMSM_ORD_REBUILD on
6865 * disks that are being rebuilt. New failures are recorded to
6866 * map[0]. So we look through all the disks we started with and
6867 * see if any failures are still present, or if any new ones
6868 * have arrived
0556e1a2 6869 */
d5985138
AK
6870 map_for_loop = map;
6871 if (prev && (map->num_members < prev->num_members))
6872 map_for_loop = prev;
68fe4598
LD
6873
6874 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 6875 idx_1 = -255;
238c0a71
AK
6876 /* when MAP_X is passed both maps failures are counted
6877 */
d5985138 6878 if (prev &&
238c0a71
AK
6879 ((look_in_map == MAP_1) || (look_in_map == MAP_X)) &&
6880 (i < prev->num_members)) {
d5985138
AK
6881 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
6882 idx_1 = ord_to_idx(ord);
c2a1e7da 6883
d5985138
AK
6884 disk = get_imsm_disk(super, idx_1);
6885 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
6886 failed++;
6887 }
238c0a71
AK
6888 if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) &&
6889 (i < map->num_members)) {
d5985138
AK
6890 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
6891 idx = ord_to_idx(ord);
6892
6893 if (idx != idx_1) {
6894 disk = get_imsm_disk(super, idx);
6895 if (!disk || is_failed(disk) ||
6896 ord & IMSM_ORD_REBUILD)
6897 failed++;
6898 }
6899 }
c2a1e7da
DW
6900 }
6901
6902 return failed;
845dea95
NB
6903}
6904
97b4d0e9
DW
6905#ifndef MDASSEMBLE
6906static int imsm_open_new(struct supertype *c, struct active_array *a,
6907 char *inst)
6908{
6909 struct intel_super *super = c->sb;
6910 struct imsm_super *mpb = super->anchor;
9587c373 6911
97b4d0e9
DW
6912 if (atoi(inst) >= mpb->num_raid_devs) {
6913 fprintf(stderr, "%s: subarry index %d, out of range\n",
6914 __func__, atoi(inst));
6915 return -ENODEV;
6916 }
6917
6918 dprintf("imsm: open_new %s\n", inst);
6919 a->info.container_member = atoi(inst);
6920 return 0;
6921}
6922
0c046afd
DW
6923static int is_resyncing(struct imsm_dev *dev)
6924{
6925 struct imsm_map *migr_map;
6926
6927 if (!dev->vol.migr_state)
6928 return 0;
6929
1484e727
DW
6930 if (migr_type(dev) == MIGR_INIT ||
6931 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW
6932 return 1;
6933
4c9bc37b
AK
6934 if (migr_type(dev) == MIGR_GEN_MIGR)
6935 return 0;
6936
238c0a71 6937 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 6938
4c9bc37b
AK
6939 if ((migr_map->map_state == IMSM_T_STATE_NORMAL) &&
6940 (dev->vol.migr_type != MIGR_GEN_MIGR))
0c046afd
DW
6941 return 1;
6942 else
6943 return 0;
6944}
6945
0556e1a2
DW
6946/* return true if we recorded new information */
6947static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 6948{
0556e1a2
DW
6949 __u32 ord;
6950 int slot;
6951 struct imsm_map *map;
86c54047
DW
6952 char buf[MAX_RAID_SERIAL_LEN+3];
6953 unsigned int len, shift = 0;
0556e1a2
DW
6954
6955 /* new failures are always set in map[0] */
238c0a71 6956 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW
6957
6958 slot = get_imsm_disk_slot(map, idx);
6959 if (slot < 0)
6960 return 0;
6961
6962 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 6963 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW
6964 return 0;
6965
7d0c5e24
LD
6966 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
6967 buf[MAX_RAID_SERIAL_LEN] = '\000';
6968 strcat(buf, ":0");
86c54047
DW
6969 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
6970 shift = len - MAX_RAID_SERIAL_LEN + 1;
6971 strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN);
6972
f2f27e63 6973 disk->status |= FAILED_DISK;
0556e1a2 6974 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK
6975 /* mark failures in second map if second map exists and this disk
6976 * in this slot.
6977 * This is valid for migration, initialization and rebuild
6978 */
6979 if (dev->vol.migr_state) {
238c0a71 6980 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK
6981 int slot2 = get_imsm_disk_slot(map2, idx);
6982
6983 if ((slot2 < map2->num_members) &&
6984 (slot2 >= 0))
6985 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK
6986 idx | IMSM_ORD_REBUILD);
6987 }
f21e18ca 6988 if (map->failed_disk_num == 0xff)
0556e1a2
DW
6989 map->failed_disk_num = slot;
6990 return 1;
6991}
6992
6993static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx)
6994{
6995 mark_failure(dev, disk, idx);
6996
6997 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
6998 return;
6999
47ee5a45
DW
7000 disk->scsi_id = __cpu_to_le32(~(__u32)0);
7001 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
7002}
7003
33414a01
DW
7004static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
7005{
33414a01 7006 struct dl *dl;
33414a01
DW
7007
7008 if (!super->missing)
7009 return;
33414a01
DW
7010
7011 dprintf("imsm: mark missing\n");
3d59f0c0
AK
7012 /* end process for initialization and rebuild only
7013 */
7014 if (is_gen_migration(dev) == 0) {
7015 __u8 map_state;
7016 int failed;
7017
7018 failed = imsm_count_failed(super, dev, MAP_0);
7019 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
7020
7021 end_migration(dev, super, map_state);
7022 }
33414a01
DW
7023 for (dl = super->missing; dl; dl = dl->next)
7024 mark_missing(dev, &dl->disk, dl->index);
7025 super->updates_pending++;
7026}
7027
f3871fdc
AK
7028static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
7029 long long new_size)
70bdf0dc 7030{
238c0a71 7031 int used_disks = imsm_num_data_members(dev, MAP_0);
70bdf0dc
AK
7032 unsigned long long array_blocks;
7033 struct imsm_map *map;
7034
7035 if (used_disks == 0) {
7036 /* when problems occures
7037 * return current array_blocks value
7038 */
7039 array_blocks = __le32_to_cpu(dev->size_high);
7040 array_blocks = array_blocks << 32;
7041 array_blocks += __le32_to_cpu(dev->size_low);
7042
7043 return array_blocks;
7044 }
7045
7046 /* set array size in metadata
7047 */
f3871fdc
AK
7048 if (new_size <= 0) {
7049 /* OLCE size change is caused by added disks
7050 */
7051 map = get_imsm_map(dev, MAP_0);
7052 array_blocks = blocks_per_member(map) * used_disks;
7053 } else {
7054 /* Online Volume Size Change
7055 * Using available free space
7056 */
7057 array_blocks = new_size;
7058 }
70bdf0dc
AK
7059
7060 /* round array size down to closest MB
7061 */
7062 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
7063 dev->size_low = __cpu_to_le32((__u32)array_blocks);
7064 dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32));
7065
7066 return array_blocks;
7067}
7068
28bce06f
AK
7069static void imsm_set_disk(struct active_array *a, int n, int state);
7070
0e2d1a4e
AK
7071static void imsm_progress_container_reshape(struct intel_super *super)
7072{
7073 /* if no device has a migr_state, but some device has a
7074 * different number of members than the previous device, start
7075 * changing the number of devices in this device to match
7076 * previous.
7077 */
7078 struct imsm_super *mpb = super->anchor;
7079 int prev_disks = -1;
7080 int i;
1dfaa380 7081 int copy_map_size;
0e2d1a4e
AK
7082
7083 for (i = 0; i < mpb->num_raid_devs; i++) {
7084 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 7085 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK
7086 struct imsm_map *map2;
7087 int prev_num_members;
0e2d1a4e
AK
7088
7089 if (dev->vol.migr_state)
7090 return;
7091
7092 if (prev_disks == -1)
7093 prev_disks = map->num_members;
7094 if (prev_disks == map->num_members)
7095 continue;
7096
7097 /* OK, this array needs to enter reshape mode.
7098 * i.e it needs a migr_state
7099 */
7100
1dfaa380 7101 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK
7102 prev_num_members = map->num_members;
7103 map->num_members = prev_disks;
7104 dev->vol.migr_state = 1;
7105 dev->vol.curr_migr_unit = 0;
ea672ee1 7106 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK
7107 for (i = prev_num_members;
7108 i < map->num_members; i++)
7109 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 7110 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 7111 /* Copy the current map */
1dfaa380 7112 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK
7113 map2->num_members = prev_num_members;
7114
f3871fdc 7115 imsm_set_array_size(dev, -1);
51d83f5d 7116 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK
7117 super->updates_pending++;
7118 }
7119}
7120
aad6f216 7121/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW
7122 * states are handled in imsm_set_disk() with one exception, when a
7123 * resync is stopped due to a new failure this routine will set the
7124 * 'degraded' state for the array.
7125 */
01f157d7 7126static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW
7127{
7128 int inst = a->info.container_member;
7129 struct intel_super *super = a->container->sb;
949c47a0 7130 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 7131 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK
7132 int failed = imsm_count_failed(super, dev, MAP_0);
7133 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 7134 __u32 blocks_per_unit;
a862209d 7135
1af97990
AK
7136 if (dev->vol.migr_state &&
7137 dev->vol.migr_type == MIGR_GEN_MIGR) {
7138 /* array state change is blocked due to reshape action
aad6f216
N
7139 * We might need to
7140 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
7141 * - finish the reshape (if last_checkpoint is big and action != reshape)
7142 * - update curr_migr_unit
1af97990 7143 */
aad6f216
N
7144 if (a->curr_action == reshape) {
7145 /* still reshaping, maybe update curr_migr_unit */
633b5610 7146 goto mark_checkpoint;
aad6f216
N
7147 } else {
7148 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
7149 /* for some reason we aborted the reshape.
b66e591b
AK
7150 *
7151 * disable automatic metadata rollback
7152 * user action is required to recover process
aad6f216 7153 */
b66e591b 7154 if (0) {
238c0a71
AK
7155 struct imsm_map *map2 =
7156 get_imsm_map(dev, MAP_1);
7157 dev->vol.migr_state = 0;
7158 set_migr_type(dev, 0);
7159 dev->vol.curr_migr_unit = 0;
7160 memcpy(map, map2,
7161 sizeof_imsm_map(map2));
7162 super->updates_pending++;
b66e591b 7163 }
aad6f216
N
7164 }
7165 if (a->last_checkpoint >= a->info.component_size) {
7166 unsigned long long array_blocks;
7167 int used_disks;
e154ced3 7168 struct mdinfo *mdi;
aad6f216 7169
238c0a71 7170 used_disks = imsm_num_data_members(dev, MAP_0);
d55adef9
AK
7171 if (used_disks > 0) {
7172 array_blocks =
5551b113 7173 blocks_per_member(map) *
d55adef9
AK
7174 used_disks;
7175 /* round array size down to closest MB
7176 */
7177 array_blocks = (array_blocks
7178 >> SECT_PER_MB_SHIFT)
7179 << SECT_PER_MB_SHIFT;
d55adef9
AK
7180 a->info.custom_array_size = array_blocks;
7181 /* encourage manager to update array
7182 * size
7183 */
e154ced3 7184
d55adef9 7185 a->check_reshape = 1;
633b5610 7186 }
e154ced3
AK
7187 /* finalize online capacity expansion/reshape */
7188 for (mdi = a->info.devs; mdi; mdi = mdi->next)
7189 imsm_set_disk(a,
7190 mdi->disk.raid_disk,
7191 mdi->curr_state);
7192
0e2d1a4e 7193 imsm_progress_container_reshape(super);
e154ced3 7194 }
aad6f216 7195 }
1af97990
AK
7196 }
7197
47ee5a45 7198 /* before we activate this array handle any missing disks */
33414a01
DW
7199 if (consistent == 2)
7200 handle_missing(super, dev);
1e5c6983 7201
0c046afd 7202 if (consistent == 2 &&
b7941fd6 7203 (!is_resync_complete(&a->info) ||
0c046afd
DW
7204 map_state != IMSM_T_STATE_NORMAL ||
7205 dev->vol.migr_state))
01f157d7 7206 consistent = 0;
272906ef 7207
b7941fd6 7208 if (is_resync_complete(&a->info)) {
0c046afd 7209 /* complete intialization / resync,
0556e1a2
DW
7210 * recovery and interrupted recovery is completed in
7211 * ->set_disk
0c046afd
DW
7212 */
7213 if (is_resyncing(dev)) {
7214 dprintf("imsm: mark resync done\n");
809da78e 7215 end_migration(dev, super, map_state);
115c3803 7216 super->updates_pending++;
484240d8 7217 a->last_checkpoint = 0;
115c3803 7218 }
b9172665
AK
7219 } else if ((!is_resyncing(dev) && !failed) &&
7220 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 7221 /* mark the start of the init process if nothing is failed */
b7941fd6 7222 dprintf("imsm: mark resync start\n");
1484e727 7223 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 7224 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 7225 else
8e59f3d8 7226 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 7227 super->updates_pending++;
115c3803 7228 }
a862209d 7229
633b5610 7230mark_checkpoint:
5b83bacf
AK
7231 /* skip checkpointing for general migration,
7232 * it is controlled in mdadm
7233 */
7234 if (is_gen_migration(dev))
7235 goto skip_mark_checkpoint;
7236
1e5c6983 7237 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 7238 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 7239 if (blocks_per_unit) {
1e5c6983
DW
7240 __u32 units32;
7241 __u64 units;
7242
4f0a7acc 7243 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW
7244 units32 = units;
7245
7246 /* check that we did not overflow 32-bits, and that
7247 * curr_migr_unit needs updating
7248 */
7249 if (units32 == units &&
bfd80a56 7250 units32 != 0 &&
1e5c6983
DW
7251 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
7252 dprintf("imsm: mark checkpoint (%u)\n", units32);
7253 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
7254 super->updates_pending++;
7255 }
7256 }
f8f603f1 7257
5b83bacf 7258skip_mark_checkpoint:
3393c6af 7259 /* mark dirty / clean */
0c046afd 7260 if (dev->vol.dirty != !consistent) {
b7941fd6 7261 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
0c046afd
DW
7262 if (consistent)
7263 dev->vol.dirty = 0;
7264 else
7265 dev->vol.dirty = 1;
a862209d
DW
7266 super->updates_pending++;
7267 }
28bce06f 7268
01f157d7 7269 return consistent;
a862209d
DW
7270}
7271
8d45d196 7272static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 7273{
8d45d196
DW
7274 int inst = a->info.container_member;
7275 struct intel_super *super = a->container->sb;
949c47a0 7276 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 7277 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 7278 struct imsm_disk *disk;
7ce05701
LD
7279 struct mdinfo *mdi;
7280 int recovery_not_finished = 0;
0c046afd 7281 int failed;
b10b37b8 7282 __u32 ord;
0c046afd 7283 __u8 map_state;
8d45d196
DW
7284
7285 if (n > map->num_members)
7286 fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
7287 n, map->num_members - 1);
7288
7289 if (n < 0)
7290 return;
7291
4e6e574a 7292 dprintf("imsm: set_disk %d:%x\n", n, state);
8d45d196 7293
238c0a71 7294 ord = get_imsm_ord_tbl_ent(dev, n, MAP_0);
b10b37b8 7295 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 7296
5802a811 7297 /* check for new failures */
0556e1a2
DW
7298 if (state & DS_FAULTY) {
7299 if (mark_failure(dev, disk, ord_to_idx(ord)))
7300 super->updates_pending++;
8d45d196 7301 }
47ee5a45 7302
19859edc 7303 /* check if in_sync */
0556e1a2 7304 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 7305 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW
7306
7307 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
19859edc
DW
7308 super->updates_pending++;
7309 }
8d45d196 7310
3b451610
AK
7311 failed = imsm_count_failed(super, dev, MAP_0);
7312 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 7313
0c046afd 7314 /* check if recovery complete, newly degraded, or failed */
94002678
AK
7315 dprintf("imsm: Detected transition to state ");
7316 switch (map_state) {
7317 case IMSM_T_STATE_NORMAL: /* transition to normal state */
7318 dprintf("normal: ");
7319 if (is_rebuilding(dev)) {
7320 dprintf("while rebuilding");
7ce05701
LD
7321 /* check if recovery is really finished */
7322 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
7323 if (mdi->recovery_start != MaxSector) {
7324 recovery_not_finished = 1;
7325 break;
7326 }
7327 if (recovery_not_finished) {
7328 dprintf("\nimsm: Rebuild has not finished yet, "
7329 "state not changed");
7330 if (a->last_checkpoint < mdi->recovery_start) {
7331 a->last_checkpoint = mdi->recovery_start;
7332 super->updates_pending++;
7333 }
7334 break;
7335 }
94002678 7336 end_migration(dev, super, map_state);
238c0a71 7337 map = get_imsm_map(dev, MAP_0);
94002678
AK
7338 map->failed_disk_num = ~0;
7339 super->updates_pending++;
7340 a->last_checkpoint = 0;
7341 break;
7342 }
7343 if (is_gen_migration(dev)) {
7344 dprintf("while general migration");
bf2f0071 7345 if (a->last_checkpoint >= a->info.component_size)
809da78e 7346 end_migration(dev, super, map_state);
94002678
AK
7347 else
7348 map->map_state = map_state;
238c0a71 7349 map = get_imsm_map(dev, MAP_0);
28bce06f 7350 map->failed_disk_num = ~0;
94002678 7351 super->updates_pending++;
bf2f0071 7352 break;
94002678
AK
7353 }
7354 break;
7355 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
7356 dprintf("degraded: ");
7357 if ((map->map_state != map_state) &&
7358 !dev->vol.migr_state) {
7359 dprintf("mark degraded");
7360 map->map_state = map_state;
7361 super->updates_pending++;
7362 a->last_checkpoint = 0;
7363 break;
7364 }
7365 if (is_rebuilding(dev)) {
7366 dprintf("while rebuilding.");
7367 if (map->map_state != map_state) {
7368 dprintf(" Map state change");
7369 end_migration(dev, super, map_state);
7370 super->updates_pending++;
7371 }
7372 break;
7373 }
7374 if (is_gen_migration(dev)) {
7375 dprintf("while general migration");
bf2f0071 7376 if (a->last_checkpoint >= a->info.component_size)
809da78e 7377 end_migration(dev, super, map_state);
94002678
AK
7378 else {
7379 map->map_state = map_state;
3b451610 7380 manage_second_map(super, dev);
94002678
AK
7381 }
7382 super->updates_pending++;
bf2f0071 7383 break;
28bce06f 7384 }
6ce1fbf1
AK
7385 if (is_initializing(dev)) {
7386 dprintf("while initialization.");
7387 map->map_state = map_state;
7388 super->updates_pending++;
7389 break;
7390 }
94002678
AK
7391 break;
7392 case IMSM_T_STATE_FAILED: /* transition to failed state */
7393 dprintf("failed: ");
7394 if (is_gen_migration(dev)) {
7395 dprintf("while general migration");
7396 map->map_state = map_state;
7397 super->updates_pending++;
7398 break;
7399 }
7400 if (map->map_state != map_state) {
7401 dprintf("mark failed");
7402 end_migration(dev, super, map_state);
7403 super->updates_pending++;
7404 a->last_checkpoint = 0;
7405 break;
7406 }
7407 break;
7408 default:
7409 dprintf("state %i\n", map_state);
5802a811 7410 }
94002678
AK
7411 dprintf("\n");
7412
845dea95
NB
7413}
7414
f796af5d 7415static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 7416{
f796af5d 7417 void *buf = mpb;
c2a1e7da
DW
7418 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
7419 unsigned long long dsize;
7420 unsigned long long sectors;
7421
7422 get_dev_size(fd, NULL, &dsize);
7423
272f648f
DW
7424 if (mpb_size > 512) {
7425 /* -1 to account for anchor */
7426 sectors = mpb_sectors(mpb) - 1;
c2a1e7da 7427
272f648f
DW
7428 /* write the extended mpb to the sectors preceeding the anchor */
7429 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0)
7430 return 1;
c2a1e7da 7431
f21e18ca
N
7432 if ((unsigned long long)write(fd, buf + 512, 512 * sectors)
7433 != 512 * sectors)
272f648f
DW
7434 return 1;
7435 }
c2a1e7da 7436
272f648f
DW
7437 /* first block is stored on second to last sector of the disk */
7438 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
c2a1e7da
DW
7439 return 1;
7440
f796af5d 7441 if (write(fd, buf, 512) != 512)
c2a1e7da
DW
7442 return 1;
7443
c2a1e7da
DW
7444 return 0;
7445}
7446
2e735d19 7447static void imsm_sync_metadata(struct supertype *container)
845dea95 7448{
2e735d19 7449 struct intel_super *super = container->sb;
c2a1e7da 7450
1a64be56 7451 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW
7452 if (!super->updates_pending)
7453 return;
7454
36988a3d 7455 write_super_imsm(container, 0);
c2a1e7da
DW
7456
7457 super->updates_pending = 0;
845dea95
NB
7458}
7459
272906ef
DW
7460static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
7461{
7462 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 7463 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW
7464 struct dl *dl;
7465
7466 for (dl = super->disks; dl; dl = dl->next)
7467 if (dl->index == i)
7468 break;
7469
25ed7e59 7470 if (dl && is_failed(&dl->disk))
272906ef
DW
7471 dl = NULL;
7472
7473 if (dl)
7474 dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor);
7475
7476 return dl;
7477}
7478
a20d2ba5 7479static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK
7480 struct active_array *a, int activate_new,
7481 struct mdinfo *additional_test_list)
272906ef
DW
7482{
7483 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 7484 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW
7485 struct imsm_super *mpb = super->anchor;
7486 struct imsm_map *map;
272906ef
DW
7487 unsigned long long pos;
7488 struct mdinfo *d;
7489 struct extent *ex;
a20d2ba5 7490 int i, j;
272906ef 7491 int found;
569cc43f
DW
7492 __u32 array_start = 0;
7493 __u32 array_end = 0;
272906ef 7494 struct dl *dl;
6c932028 7495 struct mdinfo *test_list;
272906ef
DW
7496
7497 for (dl = super->disks; dl; dl = dl->next) {
7498 /* If in this array, skip */
7499 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW
7500 if (d->state_fd >= 0 &&
7501 d->disk.major == dl->major &&
272906ef 7502 d->disk.minor == dl->minor) {
8ba77d32
AK
7503 dprintf("%x:%x already in array\n",
7504 dl->major, dl->minor);
272906ef
DW
7505 break;
7506 }
7507 if (d)
7508 continue;
6c932028
AK
7509 test_list = additional_test_list;
7510 while (test_list) {
7511 if (test_list->disk.major == dl->major &&
7512 test_list->disk.minor == dl->minor) {
8ba77d32
AK
7513 dprintf("%x:%x already in additional test list\n",
7514 dl->major, dl->minor);
7515 break;
7516 }
6c932028 7517 test_list = test_list->next;
8ba77d32 7518 }
6c932028 7519 if (test_list)
8ba77d32 7520 continue;
272906ef 7521
e553d2a4 7522 /* skip in use or failed drives */
25ed7e59 7523 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW
7524 dl->index == -2) {
7525 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 7526 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW
7527 continue;
7528 }
7529
a20d2ba5
DW
7530 /* skip pure spares when we are looking for partially
7531 * assimilated drives
7532 */
7533 if (dl->index == -1 && !activate_new)
7534 continue;
7535
272906ef 7536 /* Does this unused device have the requisite free space?
a20d2ba5 7537 * It needs to be able to cover all member volumes
272906ef
DW
7538 */
7539 ex = get_extents(super, dl);
7540 if (!ex) {
7541 dprintf("cannot get extents\n");
7542 continue;
7543 }
a20d2ba5
DW
7544 for (i = 0; i < mpb->num_raid_devs; i++) {
7545 dev = get_imsm_dev(super, i);
238c0a71 7546 map = get_imsm_map(dev, MAP_0);
272906ef 7547
a20d2ba5
DW
7548 /* check if this disk is already a member of
7549 * this array
272906ef 7550 */
620b1713 7551 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW
7552 continue;
7553
7554 found = 0;
7555 j = 0;
7556 pos = 0;
5551b113 7557 array_start = pba_of_lba0(map);
329c8278 7558 array_end = array_start +
5551b113 7559 blocks_per_member(map) - 1;
a20d2ba5
DW
7560
7561 do {
7562 /* check that we can start at pba_of_lba0 with
7563 * blocks_per_member of space
7564 */
329c8278 7565 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW
7566 found = 1;
7567 break;
7568 }
7569 pos = ex[j].start + ex[j].size;
7570 j++;
7571 } while (ex[j-1].size);
7572
7573 if (!found)
272906ef 7574 break;
a20d2ba5 7575 }
272906ef
DW
7576
7577 free(ex);
a20d2ba5 7578 if (i < mpb->num_raid_devs) {
329c8278
DW
7579 dprintf("%x:%x does not have %u to %u available\n",
7580 dl->major, dl->minor, array_start, array_end);
272906ef
DW
7581 /* No room */
7582 continue;
a20d2ba5
DW
7583 }
7584 return dl;
272906ef
DW
7585 }
7586
7587 return dl;
7588}
7589
95d07a2c
LM
7590
7591static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
7592{
7593 struct imsm_dev *dev2;
7594 struct imsm_map *map;
7595 struct dl *idisk;
7596 int slot;
7597 int idx;
7598 __u8 state;
7599
7600 dev2 = get_imsm_dev(cont->sb, dev_idx);
7601 if (dev2) {
238c0a71 7602 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 7603 if (state == IMSM_T_STATE_FAILED) {
238c0a71 7604 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM
7605 if (!map)
7606 return 1;
7607 for (slot = 0; slot < map->num_members; slot++) {
7608 /*
7609 * Check if failed disks are deleted from intel
7610 * disk list or are marked to be deleted
7611 */
238c0a71 7612 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM
7613 idisk = get_imsm_dl_disk(cont->sb, idx);
7614 /*
7615 * Do not rebuild the array if failed disks
7616 * from failed sub-array are not removed from
7617 * container.
7618 */
7619 if (idisk &&
7620 is_failed(&idisk->disk) &&
7621 (idisk->action != DISK_REMOVE))
7622 return 0;
7623 }
7624 }
7625 }
7626 return 1;
7627}
7628
88758e9d
DW
7629static struct mdinfo *imsm_activate_spare(struct active_array *a,
7630 struct metadata_update **updates)
7631{
7632 /**
d23fe947
DW
7633 * Find a device with unused free space and use it to replace a
7634 * failed/vacant region in an array. We replace failed regions one a
7635 * array at a time. The result is that a new spare disk will be added
7636 * to the first failed array and after the monitor has finished
7637 * propagating failures the remainder will be consumed.
88758e9d 7638 *
d23fe947
DW
7639 * FIXME add a capability for mdmon to request spares from another
7640 * container.
88758e9d
DW
7641 */
7642
7643 struct intel_super *super = a->container->sb;
88758e9d 7644 int inst = a->info.container_member;
949c47a0 7645 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 7646 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW
7647 int failed = a->info.array.raid_disks;
7648 struct mdinfo *rv = NULL;
7649 struct mdinfo *d;
7650 struct mdinfo *di;
7651 struct metadata_update *mu;
7652 struct dl *dl;
7653 struct imsm_update_activate_spare *u;
7654 int num_spares = 0;
7655 int i;
95d07a2c 7656 int allowed;
88758e9d
DW
7657
7658 for (d = a->info.devs ; d ; d = d->next) {
7659 if ((d->curr_state & DS_FAULTY) &&
7660 d->state_fd >= 0)
7661 /* wait for Removal to happen */
7662 return NULL;
7663 if (d->state_fd >= 0)
7664 failed--;
7665 }
7666
7667 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
7668 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 7669
e2962bfc
AK
7670 if (imsm_reshape_blocks_arrays_changes(super))
7671 return NULL;
1af97990 7672
fc8ca064
AK
7673 /* Cannot activate another spare if rebuild is in progress already
7674 */
7675 if (is_rebuilding(dev)) {
7676 dprintf("imsm: No spare activation allowed. "
7677 "Rebuild in progress already.\n");
7678 return NULL;
7679 }
7680
89c67882
AK
7681 if (a->info.array.level == 4)
7682 /* No repair for takeovered array
7683 * imsm doesn't support raid4
7684 */
7685 return NULL;
7686
3b451610
AK
7687 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
7688 IMSM_T_STATE_DEGRADED)
88758e9d
DW
7689 return NULL;
7690
95d07a2c
LM
7691 /*
7692 * If there are any failed disks check state of the other volume.
7693 * Block rebuild if the another one is failed until failed disks
7694 * are removed from container.
7695 */
7696 if (failed) {
c4acd1e5 7697 dprintf("found failed disks in %.*s, check if there another"
95d07a2c 7698 "failed sub-array.\n",
c4acd1e5 7699 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM
7700 /* check if states of the other volumes allow for rebuild */
7701 for (i = 0; i < super->anchor->num_raid_devs; i++) {
7702 if (i != inst) {
7703 allowed = imsm_rebuild_allowed(a->container,
7704 i, failed);
7705 if (!allowed)
7706 return NULL;
7707 }
7708 }
7709 }
7710
88758e9d 7711 /* For each slot, if it is not working, find a spare */
88758e9d
DW
7712 for (i = 0; i < a->info.array.raid_disks; i++) {
7713 for (d = a->info.devs ; d ; d = d->next)
7714 if (d->disk.raid_disk == i)
7715 break;
7716 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7717 if (d && (d->state_fd >= 0))
7718 continue;
7719
272906ef 7720 /*
a20d2ba5
DW
7721 * OK, this device needs recovery. Try to re-add the
7722 * previous occupant of this slot, if this fails see if
7723 * we can continue the assimilation of a spare that was
7724 * partially assimilated, finally try to activate a new
7725 * spare.
272906ef
DW
7726 */
7727 dl = imsm_readd(super, i, a);
7728 if (!dl)
b303fe21 7729 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 7730 if (!dl)
b303fe21 7731 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW
7732 if (!dl)
7733 continue;
7734
7735 /* found a usable disk with enough space */
7736 di = malloc(sizeof(*di));
79244939
DW
7737 if (!di)
7738 continue;
272906ef
DW
7739 memset(di, 0, sizeof(*di));
7740
7741 /* dl->index will be -1 in the case we are activating a
7742 * pristine spare. imsm_process_update() will create a
7743 * new index in this case. Once a disk is found to be
7744 * failed in all member arrays it is kicked from the
7745 * metadata
7746 */
7747 di->disk.number = dl->index;
d23fe947 7748
272906ef
DW
7749 /* (ab)use di->devs to store a pointer to the device
7750 * we chose
7751 */
7752 di->devs = (struct mdinfo *) dl;
7753
7754 di->disk.raid_disk = i;
7755 di->disk.major = dl->major;
7756 di->disk.minor = dl->minor;
7757 di->disk.state = 0;
d23534e4 7758 di->recovery_start = 0;
5551b113 7759 di->data_offset = pba_of_lba0(map);
272906ef
DW
7760 di->component_size = a->info.component_size;
7761 di->container_member = inst;
148acb7b 7762 super->random = random32();
272906ef
DW
7763 di->next = rv;
7764 rv = di;
7765 num_spares++;
7766 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
7767 i, di->data_offset);
88758e9d
DW
7768 }
7769
7770 if (!rv)
7771 /* No spares found */
7772 return rv;
7773 /* Now 'rv' has a list of devices to return.
7774 * Create a metadata_update record to update the
7775 * disk_ord_tbl for the array
7776 */
7777 mu = malloc(sizeof(*mu));
79244939
DW
7778 if (mu) {
7779 mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares);
7780 if (mu->buf == NULL) {
7781 free(mu);
7782 mu = NULL;
7783 }
7784 }
7785 if (!mu) {
7786 while (rv) {
7787 struct mdinfo *n = rv->next;
7788
7789 free(rv);
7790 rv = n;
7791 }
7792 return NULL;
7793 }
ca9de185 7794
88758e9d 7795 mu->space = NULL;
cb23f1f4 7796 mu->space_list = NULL;
88758e9d
DW
7797 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
7798 mu->next = *updates;
7799 u = (struct imsm_update_activate_spare *) mu->buf;
7800
7801 for (di = rv ; di ; di = di->next) {
7802 u->type = update_activate_spare;
d23fe947
DW
7803 u->dl = (struct dl *) di->devs;
7804 di->devs = NULL;
88758e9d
DW
7805 u->slot = di->disk.raid_disk;
7806 u->array = inst;
7807 u->next = u + 1;
7808 u++;
7809 }
7810 (u-1)->next = NULL;
7811 *updates = mu;
7812
7813 return rv;
7814}
7815
54c2c1ea 7816static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 7817{
54c2c1ea 7818 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK
7819 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7820 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW
7821 struct disk_info *inf = get_disk_info(u);
7822 struct imsm_disk *disk;
8273f55e
DW
7823 int i;
7824 int j;
8273f55e 7825
54c2c1ea 7826 for (i = 0; i < map->num_members; i++) {
238c0a71 7827 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW
7828 for (j = 0; j < new_map->num_members; j++)
7829 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW
7830 return 1;
7831 }
7832
7833 return 0;
7834}
7835
1a64be56
LM
7836
7837static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
7838{
7839 struct dl *dl = NULL;
7840 for (dl = super->disks; dl; dl = dl->next)
7841 if ((dl->major == major) && (dl->minor == minor))
7842 return dl;
7843 return NULL;
7844}
7845
7846static int remove_disk_super(struct intel_super *super, int major, int minor)
7847{
7848 struct dl *prev = NULL;
7849 struct dl *dl;
7850
7851 prev = NULL;
7852 for (dl = super->disks; dl; dl = dl->next) {
7853 if ((dl->major == major) && (dl->minor == minor)) {
7854 /* remove */
7855 if (prev)
7856 prev->next = dl->next;
7857 else
7858 super->disks = dl->next;
7859 dl->next = NULL;
7860 __free_imsm_disk(dl);
7861 dprintf("%s: removed %x:%x\n",
7862 __func__, major, minor);
7863 break;
7864 }
7865 prev = dl;
7866 }
7867 return 0;
7868}
7869
f21e18ca 7870static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 7871
1a64be56
LM
7872static int add_remove_disk_update(struct intel_super *super)
7873{
7874 int check_degraded = 0;
7875 struct dl *disk = NULL;
7876 /* add/remove some spares to/from the metadata/contrainer */
7877 while (super->disk_mgmt_list) {
7878 struct dl *disk_cfg;
7879
7880 disk_cfg = super->disk_mgmt_list;
7881 super->disk_mgmt_list = disk_cfg->next;
7882 disk_cfg->next = NULL;
7883
7884 if (disk_cfg->action == DISK_ADD) {
7885 disk_cfg->next = super->disks;
7886 super->disks = disk_cfg;
7887 check_degraded = 1;
7888 dprintf("%s: added %x:%x\n",
7889 __func__, disk_cfg->major,
7890 disk_cfg->minor);
7891 } else if (disk_cfg->action == DISK_REMOVE) {
7892 dprintf("Disk remove action processed: %x.%x\n",
7893 disk_cfg->major, disk_cfg->minor);
7894 disk = get_disk_super(super,
7895 disk_cfg->major,
7896 disk_cfg->minor);
7897 if (disk) {
7898 /* store action status */
7899 disk->action = DISK_REMOVE;
7900 /* remove spare disks only */
7901 if (disk->index == -1) {
7902 remove_disk_super(super,
7903 disk_cfg->major,
7904 disk_cfg->minor);
7905 }
7906 }
7907 /* release allocate disk structure */
7908 __free_imsm_disk(disk_cfg);
7909 }
7910 }
7911 return check_degraded;
7912}
7913
a29911da
PC
7914
7915static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
7916 struct intel_super *super,
7917 void ***space_list)
7918{
7919 struct intel_dev *id;
7920 void **tofree = NULL;
7921 int ret_val = 0;
7922
7923 dprintf("apply_reshape_migration_update()\n");
7924 if ((u->subdev < 0) ||
7925 (u->subdev > 1)) {
7926 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
7927 return ret_val;
7928 }
7929 if ((space_list == NULL) || (*space_list == NULL)) {
7930 dprintf("imsm: Error: Memory is not allocated\n");
7931 return ret_val;
7932 }
7933
7934 for (id = super->devlist ; id; id = id->next) {
7935 if (id->index == (unsigned)u->subdev) {
7936 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
7937 struct imsm_map *map;
7938 struct imsm_dev *new_dev =
7939 (struct imsm_dev *)*space_list;
238c0a71 7940 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC
7941 int to_state;
7942 struct dl *new_disk;
7943
7944 if (new_dev == NULL)
7945 return ret_val;
7946 *space_list = **space_list;
7947 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 7948 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC
7949 if (migr_map) {
7950 dprintf("imsm: Error: migration in progress");
7951 return ret_val;
7952 }
7953
7954 to_state = map->map_state;
7955 if ((u->new_level == 5) && (map->raid_level == 0)) {
7956 map->num_members++;
7957 /* this should not happen */
7958 if (u->new_disks[0] < 0) {
7959 map->failed_disk_num =
7960 map->num_members - 1;
7961 to_state = IMSM_T_STATE_DEGRADED;
7962 } else
7963 to_state = IMSM_T_STATE_NORMAL;
7964 }
8e59f3d8 7965 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC
7966 if (u->new_level > -1)
7967 map->raid_level = u->new_level;
238c0a71 7968 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC
7969 if ((u->new_level == 5) &&
7970 (migr_map->raid_level == 0)) {
7971 int ord = map->num_members - 1;
7972 migr_map->num_members--;
7973 if (u->new_disks[0] < 0)
7974 ord |= IMSM_ORD_REBUILD;
7975 set_imsm_ord_tbl_ent(map,
7976 map->num_members - 1,
7977 ord);
7978 }
7979 id->dev = new_dev;
7980 tofree = (void **)dev;
7981
4bba0439
PC
7982 /* update chunk size
7983 */
7984 if (u->new_chunksize > 0)
7985 map->blocks_per_strip =
7986 __cpu_to_le16(u->new_chunksize * 2);
7987
a29911da
PC
7988 /* add disk
7989 */
7990 if ((u->new_level != 5) ||
7991 (migr_map->raid_level != 0) ||
7992 (migr_map->raid_level == map->raid_level))
7993 goto skip_disk_add;
7994
7995 if (u->new_disks[0] >= 0) {
7996 /* use passes spare
7997 */
7998 new_disk = get_disk_super(super,
7999 major(u->new_disks[0]),
8000 minor(u->new_disks[0]));
8001 dprintf("imsm: new disk for reshape is: %i:%i "
8002 "(%p, index = %i)\n",
8003 major(u->new_disks[0]),
8004 minor(u->new_disks[0]),
8005 new_disk, new_disk->index);
8006 if (new_disk == NULL)
8007 goto error_disk_add;
8008
8009 new_disk->index = map->num_members - 1;
8010 /* slot to fill in autolayout
8011 */
8012 new_disk->raiddisk = new_disk->index;
8013 new_disk->disk.status |= CONFIGURED_DISK;
8014 new_disk->disk.status &= ~SPARE_DISK;
8015 } else
8016 goto error_disk_add;
8017
8018skip_disk_add:
8019 *tofree = *space_list;
8020 /* calculate new size
8021 */
f3871fdc 8022 imsm_set_array_size(new_dev, -1);
a29911da
PC
8023
8024 ret_val = 1;
8025 }
8026 }
8027
8028 if (tofree)
8029 *space_list = tofree;
8030 return ret_val;
8031
8032error_disk_add:
8033 dprintf("Error: imsm: Cannot find disk.\n");
8034 return ret_val;
8035}
8036
f3871fdc
AK
8037static int apply_size_change_update(struct imsm_update_size_change *u,
8038 struct intel_super *super)
8039{
8040 struct intel_dev *id;
8041 int ret_val = 0;
8042
8043 dprintf("apply_size_change_update()\n");
8044 if ((u->subdev < 0) ||
8045 (u->subdev > 1)) {
8046 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
8047 return ret_val;
8048 }
8049
8050 for (id = super->devlist ; id; id = id->next) {
8051 if (id->index == (unsigned)u->subdev) {
8052 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
8053 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8054 int used_disks = imsm_num_data_members(dev, MAP_0);
8055 unsigned long long blocks_per_member;
8056
8057 /* calculate new size
8058 */
8059 blocks_per_member = u->new_size / used_disks;
8060 dprintf("imsm: apply_size_change_update(size: %llu, "
8061 "blocks per member: %llu)\n",
8062 u->new_size, blocks_per_member);
8063 set_blocks_per_member(map, blocks_per_member);
8064 imsm_set_array_size(dev, u->new_size);
8065
8066 ret_val = 1;
8067 break;
8068 }
8069 }
8070
8071 return ret_val;
8072}
8073
8074
061d7da3 8075static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 8076 struct intel_super *super,
061d7da3
LO
8077 struct active_array *active_array)
8078{
8079 struct imsm_super *mpb = super->anchor;
8080 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 8081 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO
8082 struct imsm_map *migr_map;
8083 struct active_array *a;
8084 struct imsm_disk *disk;
8085 __u8 to_state;
8086 struct dl *dl;
8087 unsigned int found;
8088 int failed;
5961eeec 8089 int victim;
061d7da3 8090 int i;
5961eeec 8091 int second_map_created = 0;
061d7da3 8092
5961eeec 8093 for (; u; u = u->next) {
238c0a71 8094 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 8095
5961eeec 8096 if (victim < 0)
8097 return 0;
061d7da3 8098
5961eeec 8099 for (dl = super->disks; dl; dl = dl->next)
8100 if (dl == u->dl)
8101 break;
061d7da3 8102
5961eeec 8103 if (!dl) {
8104 fprintf(stderr, "error: imsm_activate_spare passed "
8105 "an unknown disk (index: %d)\n",
8106 u->dl->index);
8107 return 0;
8108 }
061d7da3 8109
5961eeec 8110 /* count failures (excluding rebuilds and the victim)
8111 * to determine map[0] state
8112 */
8113 failed = 0;
8114 for (i = 0; i < map->num_members; i++) {
8115 if (i == u->slot)
8116 continue;
8117 disk = get_imsm_disk(super,
238c0a71 8118 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 8119 if (!disk || is_failed(disk))
8120 failed++;
8121 }
061d7da3 8122
5961eeec 8123 /* adding a pristine spare, assign a new index */
8124 if (dl->index < 0) {
8125 dl->index = super->anchor->num_disks;
8126 super->anchor->num_disks++;
8127 }
8128 disk = &dl->disk;
8129 disk->status |= CONFIGURED_DISK;
8130 disk->status &= ~SPARE_DISK;
8131
8132 /* mark rebuild */
238c0a71 8133 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 8134 if (!second_map_created) {
8135 second_map_created = 1;
8136 map->map_state = IMSM_T_STATE_DEGRADED;
8137 migrate(dev, super, to_state, MIGR_REBUILD);
8138 } else
8139 map->map_state = to_state;
238c0a71 8140 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 8141 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
8142 set_imsm_ord_tbl_ent(migr_map, u->slot,
8143 dl->index | IMSM_ORD_REBUILD);
8144
8145 /* update the family_num to mark a new container
8146 * generation, being careful to record the existing
8147 * family_num in orig_family_num to clean up after
8148 * earlier mdadm versions that neglected to set it.
8149 */
8150 if (mpb->orig_family_num == 0)
8151 mpb->orig_family_num = mpb->family_num;
8152 mpb->family_num += super->random;
8153
8154 /* count arrays using the victim in the metadata */
8155 found = 0;
8156 for (a = active_array; a ; a = a->next) {
8157 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8158 map = get_imsm_map(dev, MAP_0);
061d7da3 8159
5961eeec 8160 if (get_imsm_disk_slot(map, victim) >= 0)
8161 found++;
8162 }
061d7da3 8163
5961eeec 8164 /* delete the victim if it is no longer being
8165 * utilized anywhere
061d7da3 8166 */
5961eeec 8167 if (!found) {
8168 struct dl **dlp;
061d7da3 8169
5961eeec 8170 /* We know that 'manager' isn't touching anything,
8171 * so it is safe to delete
8172 */
8173 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO
8174 if ((*dlp)->index == victim)
8175 break;
5961eeec 8176
8177 /* victim may be on the missing list */
8178 if (!*dlp)
8179 for (dlp = &super->missing; *dlp;
8180 dlp = &(*dlp)->next)
8181 if ((*dlp)->index == victim)
8182 break;
8183 imsm_delete(super, dlp, victim);
8184 }
061d7da3
LO
8185 }
8186
8187 return 1;
8188}
a29911da 8189
2e5dc010
N
8190static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
8191 struct intel_super *super,
8192 void ***space_list)
8193{
8194 struct dl *new_disk;
8195 struct intel_dev *id;
8196 int i;
8197 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 8198 int disk_count = u->old_raid_disks;
2e5dc010
N
8199 void **tofree = NULL;
8200 int devices_to_reshape = 1;
8201 struct imsm_super *mpb = super->anchor;
8202 int ret_val = 0;
d098291a 8203 unsigned int dev_id;
2e5dc010 8204
ed7333bd 8205 dprintf("imsm: apply_reshape_container_disks_update()\n");
2e5dc010
N
8206
8207 /* enable spares to use in array */
8208 for (i = 0; i < delta_disks; i++) {
8209 new_disk = get_disk_super(super,
8210 major(u->new_disks[i]),
8211 minor(u->new_disks[i]));
ed7333bd
AK
8212 dprintf("imsm: new disk for reshape is: %i:%i "
8213 "(%p, index = %i)\n",
2e5dc010
N
8214 major(u->new_disks[i]), minor(u->new_disks[i]),
8215 new_disk, new_disk->index);
8216 if ((new_disk == NULL) ||
8217 ((new_disk->index >= 0) &&
8218 (new_disk->index < u->old_raid_disks)))
8219 goto update_reshape_exit;
ee4beede 8220 new_disk->index = disk_count++;
2e5dc010
N
8221 /* slot to fill in autolayout
8222 */
8223 new_disk->raiddisk = new_disk->index;
8224 new_disk->disk.status |=
8225 CONFIGURED_DISK;
8226 new_disk->disk.status &= ~SPARE_DISK;
8227 }
8228
ed7333bd
AK
8229 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
8230 mpb->num_raid_devs);
2e5dc010
N
8231 /* manage changes in volume
8232 */
d098291a 8233 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N
8234 void **sp = *space_list;
8235 struct imsm_dev *newdev;
8236 struct imsm_map *newmap, *oldmap;
8237
d098291a
AK
8238 for (id = super->devlist ; id; id = id->next) {
8239 if (id->index == dev_id)
8240 break;
8241 }
8242 if (id == NULL)
8243 break;
2e5dc010
N
8244 if (!sp)
8245 continue;
8246 *space_list = *sp;
8247 newdev = (void*)sp;
8248 /* Copy the dev, but not (all of) the map */
8249 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK
8250 oldmap = get_imsm_map(id->dev, MAP_0);
8251 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N
8252 /* Copy the current map */
8253 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
8254 /* update one device only
8255 */
8256 if (devices_to_reshape) {
ed7333bd
AK
8257 dprintf("imsm: modifying subdev: %i\n",
8258 id->index);
2e5dc010
N
8259 devices_to_reshape--;
8260 newdev->vol.migr_state = 1;
8261 newdev->vol.curr_migr_unit = 0;
ea672ee1 8262 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N
8263 newmap->num_members = u->new_raid_disks;
8264 for (i = 0; i < delta_disks; i++) {
8265 set_imsm_ord_tbl_ent(newmap,
8266 u->old_raid_disks + i,
8267 u->old_raid_disks + i);
8268 }
8269 /* New map is correct, now need to save old map
8270 */
238c0a71 8271 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N
8272 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
8273
f3871fdc 8274 imsm_set_array_size(newdev, -1);
2e5dc010
N
8275 }
8276
8277 sp = (void **)id->dev;
8278 id->dev = newdev;
8279 *sp = tofree;
8280 tofree = sp;
8e59f3d8
AK
8281
8282 /* Clear migration record */
8283 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 8284 }
819bc634
AK
8285 if (tofree)
8286 *space_list = tofree;
2e5dc010
N
8287 ret_val = 1;
8288
8289update_reshape_exit:
8290
8291 return ret_val;
8292}
8293
bb025c2f 8294static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW
8295 struct intel_super *super,
8296 void ***space_list)
bb025c2f
KW
8297{
8298 struct imsm_dev *dev = NULL;
8ca6df95
KW
8299 struct intel_dev *dv;
8300 struct imsm_dev *dev_new;
bb025c2f
KW
8301 struct imsm_map *map;
8302 struct dl *dm, *du;
8ca6df95 8303 int i;
bb025c2f
KW
8304
8305 for (dv = super->devlist; dv; dv = dv->next)
8306 if (dv->index == (unsigned int)u->subarray) {
8307 dev = dv->dev;
8308 break;
8309 }
8310
8311 if (dev == NULL)
8312 return 0;
8313
238c0a71 8314 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW
8315
8316 if (u->direction == R10_TO_R0) {
43d5ec18 8317 /* Number of failed disks must be half of initial disk number */
3b451610
AK
8318 if (imsm_count_failed(super, dev, MAP_0) !=
8319 (map->num_members / 2))
43d5ec18
KW
8320 return 0;
8321
bb025c2f
KW
8322 /* iterate through devices to mark removed disks as spare */
8323 for (dm = super->disks; dm; dm = dm->next) {
8324 if (dm->disk.status & FAILED_DISK) {
8325 int idx = dm->index;
8326 /* update indexes on the disk list */
8327/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
8328 the index values will end up being correct.... NB */
8329 for (du = super->disks; du; du = du->next)
8330 if (du->index > idx)
8331 du->index--;
8332 /* mark as spare disk */
a8619d23 8333 mark_spare(dm);
bb025c2f
KW
8334 }
8335 }
bb025c2f
KW
8336 /* update map */
8337 map->num_members = map->num_members / 2;
8338 map->map_state = IMSM_T_STATE_NORMAL;
8339 map->num_domains = 1;
8340 map->raid_level = 0;
8341 map->failed_disk_num = -1;
8342 }
8343
8ca6df95
KW
8344 if (u->direction == R0_TO_R10) {
8345 void **space;
8346 /* update slots in current disk list */
8347 for (dm = super->disks; dm; dm = dm->next) {
8348 if (dm->index >= 0)
8349 dm->index *= 2;
8350 }
8351 /* create new *missing* disks */
8352 for (i = 0; i < map->num_members; i++) {
8353 space = *space_list;
8354 if (!space)
8355 continue;
8356 *space_list = *space;
8357 du = (void *)space;
8358 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW
8359 du->fd = -1;
8360 du->minor = 0;
8361 du->major = 0;
8362 du->index = (i * 2) + 1;
8363 sprintf((char *)du->disk.serial,
8364 " MISSING_%d", du->index);
8365 sprintf((char *)du->serial,
8366 "MISSING_%d", du->index);
8367 du->next = super->missing;
8368 super->missing = du;
8369 }
8370 /* create new dev and map */
8371 space = *space_list;
8372 if (!space)
8373 return 0;
8374 *space_list = *space;
8375 dev_new = (void *)space;
8376 memcpy(dev_new, dev, sizeof(*dev));
8377 /* update new map */
238c0a71 8378 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 8379 map->num_members = map->num_members * 2;
1a2487c2 8380 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW
8381 map->num_domains = 2;
8382 map->raid_level = 1;
8383 /* replace dev<->dev_new */
8384 dv->dev = dev_new;
8385 }
bb025c2f
KW
8386 /* update disk order table */
8387 for (du = super->disks; du; du = du->next)
8388 if (du->index >= 0)
8389 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 8390 for (du = super->missing; du; du = du->next)
1a2487c2
KW
8391 if (du->index >= 0) {
8392 set_imsm_ord_tbl_ent(map, du->index, du->index);
e4c72d1d 8393 mark_missing(dv->dev, &du->disk, du->index);
1a2487c2 8394 }
bb025c2f
KW
8395
8396 return 1;
8397}
8398
e8319a19
DW
8399static void imsm_process_update(struct supertype *st,
8400 struct metadata_update *update)
8401{
8402 /**
8403 * crack open the metadata_update envelope to find the update record
8404 * update can be one of:
d195167d
AK
8405 * update_reshape_container_disks - all the arrays in the container
8406 * are being reshaped to have more devices. We need to mark
8407 * the arrays for general migration and convert selected spares
8408 * into active devices.
8409 * update_activate_spare - a spare device has replaced a failed
e8319a19
DW
8410 * device in an array, update the disk_ord_tbl. If this disk is
8411 * present in all member arrays then also clear the SPARE_DISK
8412 * flag
d195167d
AK
8413 * update_create_array
8414 * update_kill_array
8415 * update_rename_array
8416 * update_add_remove_disk
e8319a19
DW
8417 */
8418 struct intel_super *super = st->sb;
4d7b1503 8419 struct imsm_super *mpb;
e8319a19
DW
8420 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
8421
4d7b1503
DW
8422 /* update requires a larger buf but the allocation failed */
8423 if (super->next_len && !super->next_buf) {
8424 super->next_len = 0;
8425 return;
8426 }
8427
8428 if (super->next_buf) {
8429 memcpy(super->next_buf, super->buf, super->len);
8430 free(super->buf);
8431 super->len = super->next_len;
8432 super->buf = super->next_buf;
8433
8434 super->next_len = 0;
8435 super->next_buf = NULL;
8436 }
8437
8438 mpb = super->anchor;
8439
e8319a19 8440 switch (type) {
0ec5d470
AK
8441 case update_general_migration_checkpoint: {
8442 struct intel_dev *id;
8443 struct imsm_update_general_migration_checkpoint *u =
8444 (void *)update->buf;
8445
8446 dprintf("imsm: process_update() "
8447 "for update_general_migration_checkpoint called\n");
8448
8449 /* find device under general migration */
8450 for (id = super->devlist ; id; id = id->next) {
8451 if (is_gen_migration(id->dev)) {
8452 id->dev->vol.curr_migr_unit =
8453 __cpu_to_le32(u->curr_migr_unit);
8454 super->updates_pending++;
8455 }
8456 }
8457 break;
8458 }
bb025c2f
KW
8459 case update_takeover: {
8460 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW
8461 if (apply_takeover_update(u, super, &update->space_list)) {
8462 imsm_update_version_info(super);
bb025c2f 8463 super->updates_pending++;
1a2487c2 8464 }
bb025c2f
KW
8465 break;
8466 }
8467
78b10e66 8468 case update_reshape_container_disks: {
d195167d 8469 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N
8470 if (apply_reshape_container_disks_update(
8471 u, super, &update->space_list))
8472 super->updates_pending++;
78b10e66
N
8473 break;
8474 }
48c5303a 8475 case update_reshape_migration: {
a29911da
PC
8476 struct imsm_update_reshape_migration *u = (void *)update->buf;
8477 if (apply_reshape_migration_update(
8478 u, super, &update->space_list))
8479 super->updates_pending++;
48c5303a
PC
8480 break;
8481 }
f3871fdc
AK
8482 case update_size_change: {
8483 struct imsm_update_size_change *u = (void *)update->buf;
8484 if (apply_size_change_update(u, super))
8485 super->updates_pending++;
8486 break;
8487 }
e8319a19
DW
8488 case update_activate_spare: {
8489 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO
8490 if (apply_update_activate_spare(u, super, st->arrays))
8491 super->updates_pending++;
8273f55e
DW
8492 break;
8493 }
8494 case update_create_array: {
8495 /* someone wants to create a new array, we need to be aware of
8496 * a few races/collisions:
8497 * 1/ 'Create' called by two separate instances of mdadm
8498 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
8499 * devices that have since been assimilated via
8500 * activate_spare.
8501 * In the event this update can not be carried out mdadm will
8502 * (FIX ME) notice that its update did not take hold.
8503 */
8504 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 8505 struct intel_dev *dv;
8273f55e
DW
8506 struct imsm_dev *dev;
8507 struct imsm_map *map, *new_map;
8508 unsigned long long start, end;
8509 unsigned long long new_start, new_end;
8510 int i;
54c2c1ea
DW
8511 struct disk_info *inf;
8512 struct dl *dl;
8273f55e
DW
8513
8514 /* handle racing creates: first come first serve */
8515 if (u->dev_idx < mpb->num_raid_devs) {
8516 dprintf("%s: subarray %d already defined\n",
8517 __func__, u->dev_idx);
ba2de7ba 8518 goto create_error;
8273f55e
DW
8519 }
8520
8521 /* check update is next in sequence */
8522 if (u->dev_idx != mpb->num_raid_devs) {
6a3e913e
DW
8523 dprintf("%s: can not create array %d expected index %d\n",
8524 __func__, u->dev_idx, mpb->num_raid_devs);
ba2de7ba 8525 goto create_error;
8273f55e
DW
8526 }
8527
238c0a71 8528 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113
CA
8529 new_start = pba_of_lba0(new_map);
8530 new_end = new_start + blocks_per_member(new_map);
54c2c1ea 8531 inf = get_disk_info(u);
8273f55e
DW
8532
8533 /* handle activate_spare versus create race:
8534 * check to make sure that overlapping arrays do not include
8535 * overalpping disks
8536 */
8537 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 8538 dev = get_imsm_dev(super, i);
238c0a71 8539 map = get_imsm_map(dev, MAP_0);
5551b113
CA
8540 start = pba_of_lba0(map);
8541 end = start + blocks_per_member(map);
8273f55e
DW
8542 if ((new_start >= start && new_start <= end) ||
8543 (start >= new_start && start <= new_end))
54c2c1ea
DW
8544 /* overlap */;
8545 else
8546 continue;
8547
8548 if (disks_overlap(super, i, u)) {
8273f55e 8549 dprintf("%s: arrays overlap\n", __func__);
ba2de7ba 8550 goto create_error;
8273f55e
DW
8551 }
8552 }
8273f55e 8553
949c47a0
DW
8554 /* check that prepare update was successful */
8555 if (!update->space) {
8556 dprintf("%s: prepare update failed\n", __func__);
ba2de7ba 8557 goto create_error;
949c47a0
DW
8558 }
8559
54c2c1ea
DW
8560 /* check that all disks are still active before committing
8561 * changes. FIXME: could we instead handle this by creating a
8562 * degraded array? That's probably not what the user expects,
8563 * so better to drop this update on the floor.
8564 */
8565 for (i = 0; i < new_map->num_members; i++) {
8566 dl = serial_to_dl(inf[i].serial, super);
8567 if (!dl) {
8568 dprintf("%s: disk disappeared\n", __func__);
ba2de7ba 8569 goto create_error;
54c2c1ea 8570 }
949c47a0
DW
8571 }
8572
8273f55e 8573 super->updates_pending++;
54c2c1ea
DW
8574
8575 /* convert spares to members and fixup ord_tbl */
8576 for (i = 0; i < new_map->num_members; i++) {
8577 dl = serial_to_dl(inf[i].serial, super);
8578 if (dl->index == -1) {
8579 dl->index = mpb->num_disks;
8580 mpb->num_disks++;
8581 dl->disk.status |= CONFIGURED_DISK;
8582 dl->disk.status &= ~SPARE_DISK;
8583 }
8584 set_imsm_ord_tbl_ent(new_map, i, dl->index);
8585 }
8586
ba2de7ba
DW
8587 dv = update->space;
8588 dev = dv->dev;
949c47a0
DW
8589 update->space = NULL;
8590 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW
8591 dv->index = u->dev_idx;
8592 dv->next = super->devlist;
8593 super->devlist = dv;
8273f55e 8594 mpb->num_raid_devs++;
8273f55e 8595
4d1313e9 8596 imsm_update_version_info(super);
8273f55e 8597 break;
ba2de7ba
DW
8598 create_error:
8599 /* mdmon knows how to release update->space, but not
8600 * ((struct intel_dev *) update->space)->dev
8601 */
8602 if (update->space) {
8603 dv = update->space;
8604 free(dv->dev);
8605 }
8273f55e 8606 break;
e8319a19 8607 }
33414a01
DW
8608 case update_kill_array: {
8609 struct imsm_update_kill_array *u = (void *) update->buf;
8610 int victim = u->dev_idx;
8611 struct active_array *a;
8612 struct intel_dev **dp;
8613 struct imsm_dev *dev;
8614
8615 /* sanity check that we are not affecting the uuid of
8616 * active arrays, or deleting an active array
8617 *
8618 * FIXME when immutable ids are available, but note that
8619 * we'll also need to fixup the invalidated/active
8620 * subarray indexes in mdstat
8621 */
8622 for (a = st->arrays; a; a = a->next)
8623 if (a->info.container_member >= victim)
8624 break;
8625 /* by definition if mdmon is running at least one array
8626 * is active in the container, so checking
8627 * mpb->num_raid_devs is just extra paranoia
8628 */
8629 dev = get_imsm_dev(super, victim);
8630 if (a || !dev || mpb->num_raid_devs == 1) {
8631 dprintf("failed to delete subarray-%d\n", victim);
8632 break;
8633 }
8634
8635 for (dp = &super->devlist; *dp;)
f21e18ca 8636 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW
8637 *dp = (*dp)->next;
8638 } else {
f21e18ca 8639 if ((*dp)->index > (unsigned)victim)
33414a01
DW
8640 (*dp)->index--;
8641 dp = &(*dp)->next;
8642 }
8643 mpb->num_raid_devs--;
8644 super->updates_pending++;
8645 break;
8646 }
aa534678
DW
8647 case update_rename_array: {
8648 struct imsm_update_rename_array *u = (void *) update->buf;
8649 char name[MAX_RAID_SERIAL_LEN+1];
8650 int target = u->dev_idx;
8651 struct active_array *a;
8652 struct imsm_dev *dev;
8653
8654 /* sanity check that we are not affecting the uuid of
8655 * an active array
8656 */
8657 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
8658 name[MAX_RAID_SERIAL_LEN] = '\0';
8659 for (a = st->arrays; a; a = a->next)
8660 if (a->info.container_member == target)
8661 break;
8662 dev = get_imsm_dev(super, u->dev_idx);
8663 if (a || !dev || !check_name(super, name, 1)) {
8664 dprintf("failed to rename subarray-%d\n", target);
8665 break;
8666 }
8667
cdbe98cd 8668 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
aa534678
DW
8669 super->updates_pending++;
8670 break;
8671 }
1a64be56 8672 case update_add_remove_disk: {
43dad3d6 8673 /* we may be able to repair some arrays if disks are
1a64be56
LM
8674 * being added, check teh status of add_remove_disk
8675 * if discs has been added.
8676 */
8677 if (add_remove_disk_update(super)) {
43dad3d6 8678 struct active_array *a;
072b727f
DW
8679
8680 super->updates_pending++;
1a64be56 8681 for (a = st->arrays; a; a = a->next)
43dad3d6
DW
8682 a->check_degraded = 1;
8683 }
43dad3d6 8684 break;
e8319a19 8685 }
1a64be56
LM
8686 default:
8687 fprintf(stderr, "error: unsuported process update type:"
8688 "(type: %d)\n", type);
8689 }
e8319a19 8690}
88758e9d 8691
bc0b9d34
PC
8692static struct mdinfo *get_spares_for_grow(struct supertype *st);
8693
8273f55e
DW
8694static void imsm_prepare_update(struct supertype *st,
8695 struct metadata_update *update)
8696{
949c47a0 8697 /**
4d7b1503
DW
8698 * Allocate space to hold new disk entries, raid-device entries or a new
8699 * mpb if necessary. The manager synchronously waits for updates to
8700 * complete in the monitor, so new mpb buffers allocated here can be
8701 * integrated by the monitor thread without worrying about live pointers
8702 * in the manager thread.
8273f55e 8703 */
949c47a0 8704 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
4d7b1503
DW
8705 struct intel_super *super = st->sb;
8706 struct imsm_super *mpb = super->anchor;
8707 size_t buf_len;
8708 size_t len = 0;
949c47a0
DW
8709
8710 switch (type) {
0ec5d470
AK
8711 case update_general_migration_checkpoint:
8712 dprintf("imsm: prepare_update() "
8713 "for update_general_migration_checkpoint called\n");
8714 break;
abedf5fc
KW
8715 case update_takeover: {
8716 struct imsm_update_takeover *u = (void *)update->buf;
8717 if (u->direction == R0_TO_R10) {
8718 void **tail = (void **)&update->space_list;
8719 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 8720 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW
8721 int num_members = map->num_members;
8722 void *space;
8723 int size, i;
8724 int err = 0;
8725 /* allocate memory for added disks */
8726 for (i = 0; i < num_members; i++) {
8727 size = sizeof(struct dl);
8728 space = malloc(size);
8729 if (!space) {
8730 err++;
8731 break;
8732 }
8733 *tail = space;
8734 tail = space;
8735 *tail = NULL;
8736 }
8737 /* allocate memory for new device */
8738 size = sizeof_imsm_dev(super->devlist->dev, 0) +
8739 (num_members * sizeof(__u32));
8740 space = malloc(size);
8741 if (!space)
8742 err++;
8743 else {
8744 *tail = space;
8745 tail = space;
8746 *tail = NULL;
8747 }
8748 if (!err) {
8749 len = disks_to_mpb_size(num_members * 2);
8750 } else {
8751 /* if allocation didn't success, free buffer */
8752 while (update->space_list) {
8753 void **sp = update->space_list;
8754 update->space_list = *sp;
8755 free(sp);
8756 }
8757 }
8758 }
8759
8760 break;
8761 }
78b10e66 8762 case update_reshape_container_disks: {
d195167d
AK
8763 /* Every raid device in the container is about to
8764 * gain some more devices, and we will enter a
8765 * reconfiguration.
8766 * So each 'imsm_map' will be bigger, and the imsm_vol
8767 * will now hold 2 of them.
8768 * Thus we need new 'struct imsm_dev' allocations sized
8769 * as sizeof_imsm_dev but with more devices in both maps.
8770 */
8771 struct imsm_update_reshape *u = (void *)update->buf;
8772 struct intel_dev *dl;
8773 void **space_tail = (void**)&update->space_list;
8774
8775 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
8776
8777 for (dl = super->devlist; dl; dl = dl->next) {
8778 int size = sizeof_imsm_dev(dl->dev, 1);
8779 void *s;
d677e0b8
AK
8780 if (u->new_raid_disks > u->old_raid_disks)
8781 size += sizeof(__u32)*2*
8782 (u->new_raid_disks - u->old_raid_disks);
d195167d
AK
8783 s = malloc(size);
8784 if (!s)
8785 break;
8786 *space_tail = s;
8787 space_tail = s;
8788 *space_tail = NULL;
8789 }
8790
8791 len = disks_to_mpb_size(u->new_raid_disks);
8792 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N
8793 break;
8794 }
48c5303a 8795 case update_reshape_migration: {
bc0b9d34
PC
8796 /* for migration level 0->5 we need to add disks
8797 * so the same as for container operation we will copy
8798 * device to the bigger location.
8799 * in memory prepared device and new disk area are prepared
8800 * for usage in process update
8801 */
8802 struct imsm_update_reshape_migration *u = (void *)update->buf;
8803 struct intel_dev *id;
8804 void **space_tail = (void **)&update->space_list;
8805 int size;
8806 void *s;
8807 int current_level = -1;
8808
8809 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
8810
8811 /* add space for bigger array in update
8812 */
8813 for (id = super->devlist; id; id = id->next) {
8814 if (id->index == (unsigned)u->subdev) {
8815 size = sizeof_imsm_dev(id->dev, 1);
8816 if (u->new_raid_disks > u->old_raid_disks)
8817 size += sizeof(__u32)*2*
8818 (u->new_raid_disks - u->old_raid_disks);
8819 s = malloc(size);
8820 if (!s)
8821 break;
8822 *space_tail = s;
8823 space_tail = s;
8824 *space_tail = NULL;
8825 break;
8826 }
8827 }
8828 if (update->space_list == NULL)
8829 break;
8830
8831 /* add space for disk in update
8832 */
8833 size = sizeof(struct dl);
8834 s = malloc(size);
8835 if (!s) {
8836 free(update->space_list);
8837 update->space_list = NULL;
8838 break;
8839 }
8840 *space_tail = s;
8841 space_tail = s;
8842 *space_tail = NULL;
8843
8844 /* add spare device to update
8845 */
8846 for (id = super->devlist ; id; id = id->next)
8847 if (id->index == (unsigned)u->subdev) {
8848 struct imsm_dev *dev;
8849 struct imsm_map *map;
8850
8851 dev = get_imsm_dev(super, u->subdev);
238c0a71 8852 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC
8853 current_level = map->raid_level;
8854 break;
8855 }
8856 if ((u->new_level == 5) && (u->new_level != current_level)) {
8857 struct mdinfo *spares;
8858
8859 spares = get_spares_for_grow(st);
8860 if (spares) {
8861 struct dl *dl;
8862 struct mdinfo *dev;
8863
8864 dev = spares->devs;
8865 if (dev) {
8866 u->new_disks[0] =
8867 makedev(dev->disk.major,
8868 dev->disk.minor);
8869 dl = get_disk_super(super,
8870 dev->disk.major,
8871 dev->disk.minor);
8872 dl->index = u->old_raid_disks;
8873 dev = dev->next;
8874 }
8875 sysfs_free(spares);
8876 }
8877 }
8878 len = disks_to_mpb_size(u->new_raid_disks);
8879 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC
8880 break;
8881 }
f3871fdc
AK
8882 case update_size_change: {
8883 break;
8884 }
949c47a0
DW
8885 case update_create_array: {
8886 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 8887 struct intel_dev *dv;
54c2c1ea 8888 struct imsm_dev *dev = &u->dev;
238c0a71 8889 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW
8890 struct dl *dl;
8891 struct disk_info *inf;
8892 int i;
8893 int activate = 0;
949c47a0 8894
54c2c1ea
DW
8895 inf = get_disk_info(u);
8896 len = sizeof_imsm_dev(dev, 1);
ba2de7ba
DW
8897 /* allocate a new super->devlist entry */
8898 dv = malloc(sizeof(*dv));
8899 if (dv) {
8900 dv->dev = malloc(len);
8901 if (dv->dev)
8902 update->space = dv;
8903 else {
8904 free(dv);
8905 update->space = NULL;
8906 }
8907 }
949c47a0 8908
54c2c1ea
DW
8909 /* count how many spares will be converted to members */
8910 for (i = 0; i < map->num_members; i++) {
8911 dl = serial_to_dl(inf[i].serial, super);
8912 if (!dl) {
8913 /* hmm maybe it failed?, nothing we can do about
8914 * it here
8915 */
8916 continue;
8917 }
8918 if (count_memberships(dl, super) == 0)
8919 activate++;
8920 }
8921 len += activate * sizeof(struct imsm_disk);
949c47a0
DW
8922 break;
8923 default:
8924 break;
8925 }
8926 }
8273f55e 8927
4d7b1503
DW
8928 /* check if we need a larger metadata buffer */
8929 if (super->next_buf)
8930 buf_len = super->next_len;
8931 else
8932 buf_len = super->len;
8933
8934 if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) {
8935 /* ok we need a larger buf than what is currently allocated
8936 * if this allocation fails process_update will notice that
8937 * ->next_len is set and ->next_buf is NULL
8938 */
8939 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512);
8940 if (super->next_buf)
8941 free(super->next_buf);
8942
8943 super->next_len = buf_len;
1f45a8ad
DW
8944 if (posix_memalign(&super->next_buf, 512, buf_len) == 0)
8945 memset(super->next_buf, 0, buf_len);
8946 else
4d7b1503
DW
8947 super->next_buf = NULL;
8948 }
8273f55e
DW
8949}
8950
ae6aad82 8951/* must be called while manager is quiesced */
f21e18ca 8952static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW
8953{
8954 struct imsm_super *mpb = super->anchor;
ae6aad82
DW
8955 struct dl *iter;
8956 struct imsm_dev *dev;
8957 struct imsm_map *map;
24565c9a
DW
8958 int i, j, num_members;
8959 __u32 ord;
ae6aad82 8960
24565c9a
DW
8961 dprintf("%s: deleting device[%d] from imsm_super\n",
8962 __func__, index);
ae6aad82
DW
8963
8964 /* shift all indexes down one */
8965 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 8966 if (iter->index > (int)index)
ae6aad82 8967 iter->index--;
47ee5a45 8968 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 8969 if (iter->index > (int)index)
47ee5a45 8970 iter->index--;
ae6aad82
DW
8971
8972 for (i = 0; i < mpb->num_raid_devs; i++) {
8973 dev = get_imsm_dev(super, i);
238c0a71 8974 map = get_imsm_map(dev, MAP_0);
24565c9a
DW
8975 num_members = map->num_members;
8976 for (j = 0; j < num_members; j++) {
8977 /* update ord entries being careful not to propagate
8978 * ord-flags to the first map
8979 */
238c0a71 8980 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
ae6aad82 8981
24565c9a
DW
8982 if (ord_to_idx(ord) <= index)
8983 continue;
ae6aad82 8984
238c0a71 8985 map = get_imsm_map(dev, MAP_0);
24565c9a 8986 set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1));
238c0a71 8987 map = get_imsm_map(dev, MAP_1);
24565c9a
DW
8988 if (map)
8989 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW
8990 }
8991 }
8992
8993 mpb->num_disks--;
8994 super->updates_pending++;
24565c9a
DW
8995 if (*dlp) {
8996 struct dl *dl = *dlp;
8997
8998 *dlp = (*dlp)->next;
8999 __free_imsm_disk(dl);
9000 }
ae6aad82 9001}
9e2d750d 9002#endif /* MDASSEMBLE */
9a717282
AK
9003
9004static void close_targets(int *targets, int new_disks)
9005{
9006 int i;
9007
9008 if (!targets)
9009 return;
9010
9011 for (i = 0; i < new_disks; i++) {
9012 if (targets[i] >= 0) {
9013 close(targets[i]);
9014 targets[i] = -1;
9015 }
9016 }
9017}
9018
9019static int imsm_get_allowed_degradation(int level, int raid_disks,
9020 struct intel_super *super,
9021 struct imsm_dev *dev)
9022{
9023 switch (level) {
bf5cf7c7 9024 case 1:
9a717282
AK
9025 case 10:{
9026 int ret_val = 0;
9027 struct imsm_map *map;
9028 int i;
9029
9030 ret_val = raid_disks/2;
9031 /* check map if all disks pairs not failed
9032 * in both maps
9033 */
238c0a71 9034 map = get_imsm_map(dev, MAP_0);
9a717282
AK
9035 for (i = 0; i < ret_val; i++) {
9036 int degradation = 0;
9037 if (get_imsm_disk(super, i) == NULL)
9038 degradation++;
9039 if (get_imsm_disk(super, i + 1) == NULL)
9040 degradation++;
9041 if (degradation == 2)
9042 return 0;
9043 }
238c0a71 9044 map = get_imsm_map(dev, MAP_1);
9a717282
AK
9045 /* if there is no second map
9046 * result can be returned
9047 */
9048 if (map == NULL)
9049 return ret_val;
9050 /* check degradation in second map
9051 */
9052 for (i = 0; i < ret_val; i++) {
9053 int degradation = 0;
9054 if (get_imsm_disk(super, i) == NULL)
9055 degradation++;
9056 if (get_imsm_disk(super, i + 1) == NULL)
9057 degradation++;
9058 if (degradation == 2)
9059 return 0;
9060 }
9061 return ret_val;
9062 }
9063 case 5:
9064 return 1;
9065 case 6:
9066 return 2;
9067 default:
9068 return 0;
9069 }
9070}
9071
9072
687629c2
AK
9073/*******************************************************************************
9074 * Function: open_backup_targets
9075 * Description: Function opens file descriptors for all devices given in
9076 * info->devs
9077 * Parameters:
9078 * info : general array info
9079 * raid_disks : number of disks
9080 * raid_fds : table of device's file descriptors
9a717282
AK
9081 * super : intel super for raid10 degradation check
9082 * dev : intel device for raid10 degradation check
687629c2
AK
9083 * Returns:
9084 * 0 : success
9085 * -1 : fail
9086 ******************************************************************************/
9a717282
AK
9087int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
9088 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK
9089{
9090 struct mdinfo *sd;
f627f5ad 9091 int i;
9a717282 9092 int opened = 0;
f627f5ad
AK
9093
9094 for (i = 0; i < raid_disks; i++)
9095 raid_fds[i] = -1;
687629c2
AK
9096
9097 for (sd = info->devs ; sd ; sd = sd->next) {
9098 char *dn;
9099
9100 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
9101 dprintf("disk is faulty!!\n");
9102 continue;
9103 }
9104
9105 if ((sd->disk.raid_disk >= raid_disks) ||
9106 (sd->disk.raid_disk < 0))
9107 continue;
9108
9109 dn = map_dev(sd->disk.major,
9110 sd->disk.minor, 1);
9111 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
9112 if (raid_fds[sd->disk.raid_disk] < 0) {
9113 fprintf(stderr, "cannot open component\n");
9a717282 9114 continue;
687629c2 9115 }
9a717282
AK
9116 opened++;
9117 }
9118 /* check if maximum array degradation level is not exceeded
9119 */
9120 if ((raid_disks - opened) >
9121 imsm_get_allowed_degradation(info->new_level,
9122 raid_disks,
9123 super, dev)) {
9124 fprintf(stderr, "Not enough disks can be opened.\n");
9125 close_targets(raid_fds, raid_disks);
9126 return -2;
687629c2
AK
9127 }
9128 return 0;
9129}
9130
9e2d750d 9131#ifndef MDASSEMBLE
687629c2
AK
9132/*******************************************************************************
9133 * Function: init_migr_record_imsm
9134 * Description: Function inits imsm migration record
9135 * Parameters:
9136 * super : imsm internal array info
9137 * dev : device under migration
9138 * info : general array info to find the smallest device
9139 * Returns:
9140 * none
9141 ******************************************************************************/
9142void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
9143 struct mdinfo *info)
9144{
9145 struct intel_super *super = st->sb;
9146 struct migr_record *migr_rec = super->migr_rec;
9147 int new_data_disks;
9148 unsigned long long dsize, dev_sectors;
9149 long long unsigned min_dev_sectors = -1LLU;
9150 struct mdinfo *sd;
9151 char nm[30];
9152 int fd;
238c0a71
AK
9153 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
9154 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 9155 unsigned long long num_migr_units;
3ef4403c 9156 unsigned long long array_blocks;
687629c2
AK
9157
9158 memset(migr_rec, 0, sizeof(struct migr_record));
9159 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
9160
9161 /* only ascending reshape supported now */
9162 migr_rec->ascending_migr = __cpu_to_le32(1);
9163
9164 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
9165 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK
9166 migr_rec->dest_depth_per_unit *=
9167 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
238c0a71 9168 new_data_disks = imsm_num_data_members(dev, MAP_0);
687629c2
AK
9169 migr_rec->blocks_per_unit =
9170 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
9171 migr_rec->dest_depth_per_unit =
9172 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 9173 array_blocks = info->component_size * new_data_disks;
687629c2
AK
9174 num_migr_units =
9175 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
9176
9177 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
9178 num_migr_units++;
9179 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
9180
9181 migr_rec->post_migr_vol_cap = dev->size_low;
9182 migr_rec->post_migr_vol_cap_hi = dev->size_high;
9183
9184
9185 /* Find the smallest dev */
9186 for (sd = info->devs ; sd ; sd = sd->next) {
9187 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
9188 fd = dev_open(nm, O_RDONLY);
9189 if (fd < 0)
9190 continue;
9191 get_dev_size(fd, NULL, &dsize);
9192 dev_sectors = dsize / 512;
9193 if (dev_sectors < min_dev_sectors)
9194 min_dev_sectors = dev_sectors;
9195 close(fd);
9196 }
9197 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
9198 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
9199
9200 write_imsm_migr_rec(st);
9201
9202 return;
9203}
9204
9205/*******************************************************************************
9206 * Function: save_backup_imsm
9207 * Description: Function saves critical data stripes to Migration Copy Area
9208 * and updates the current migration unit status.
9209 * Use restore_stripes() to form a destination stripe,
9210 * and to write it to the Copy Area.
9211 * Parameters:
9212 * st : supertype information
aea93171 9213 * dev : imsm device that backup is saved for
687629c2
AK
9214 * info : general array info
9215 * buf : input buffer
687629c2
AK
9216 * length : length of data to backup (blocks_per_unit)
9217 * Returns:
9218 * 0 : success
9219 *, -1 : fail
9220 ******************************************************************************/
9221int save_backup_imsm(struct supertype *st,
9222 struct imsm_dev *dev,
9223 struct mdinfo *info,
9224 void *buf,
687629c2
AK
9225 int length)
9226{
9227 int rv = -1;
9228 struct intel_super *super = st->sb;
9229 unsigned long long *target_offsets = NULL;
9230 int *targets = NULL;
9231 int i;
238c0a71 9232 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 9233 int new_disks = map_dest->num_members;
ab724b98
AK
9234 int dest_layout = 0;
9235 int dest_chunk;
d1877f69 9236 unsigned long long start;
238c0a71 9237 int data_disks = imsm_num_data_members(dev, MAP_0);
687629c2
AK
9238
9239 targets = malloc(new_disks * sizeof(int));
9240 if (!targets)
9241 goto abort;
9242
7e45b550
AK
9243 for (i = 0; i < new_disks; i++)
9244 targets[i] = -1;
9245
687629c2
AK
9246 target_offsets = malloc(new_disks * sizeof(unsigned long long));
9247 if (!target_offsets)
9248 goto abort;
9249
d1877f69 9250 start = info->reshape_progress * 512;
687629c2 9251 for (i = 0; i < new_disks; i++) {
687629c2
AK
9252 target_offsets[i] = (unsigned long long)
9253 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK
9254 /* move back copy area adderss, it will be moved forward
9255 * in restore_stripes() using start input variable
9256 */
9257 target_offsets[i] -= start/data_disks;
687629c2
AK
9258 }
9259
9a717282
AK
9260 if (open_backup_targets(info, new_disks, targets,
9261 super, dev))
687629c2
AK
9262 goto abort;
9263
68eb8bc6 9264 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK
9265 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
9266
687629c2
AK
9267 if (restore_stripes(targets, /* list of dest devices */
9268 target_offsets, /* migration record offsets */
9269 new_disks,
ab724b98
AK
9270 dest_chunk,
9271 map_dest->raid_level,
9272 dest_layout,
9273 -1, /* source backup file descriptor */
9274 0, /* input buf offset
9275 * always 0 buf is already offseted */
d1877f69 9276 start,
687629c2
AK
9277 length,
9278 buf) != 0) {
9279 fprintf(stderr, Name ": Error restoring stripes\n");
9280 goto abort;
9281 }
9282
9283 rv = 0;
9284
9285abort:
9286 if (targets) {
9a717282 9287 close_targets(targets, new_disks);
687629c2
AK
9288 free(targets);
9289 }
9290 free(target_offsets);
9291
9292 return rv;
9293}
9294
9295/*******************************************************************************
9296 * Function: save_checkpoint_imsm
9297 * Description: Function called for current unit status update
9298 * in the migration record. It writes it to disk.
9299 * Parameters:
9300 * super : imsm internal array info
9301 * info : general array info
9302 * Returns:
9303 * 0: success
9304 * 1: failure
0228d92c
AK
9305 * 2: failure, means no valid migration record
9306 * / no general migration in progress /
687629c2
AK
9307 ******************************************************************************/
9308int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
9309{
9310 struct intel_super *super = st->sb;
f8b72ef5
AK
9311 unsigned long long blocks_per_unit;
9312 unsigned long long curr_migr_unit;
9313
2e062e82
AK
9314 if (load_imsm_migr_rec(super, info) != 0) {
9315 dprintf("imsm: ERROR: Cannot read migration record "
9316 "for checkpoint save.\n");
9317 return 1;
9318 }
9319
f8b72ef5
AK
9320 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
9321 if (blocks_per_unit == 0) {
0228d92c
AK
9322 dprintf("imsm: no migration in progress.\n");
9323 return 2;
687629c2 9324 }
f8b72ef5
AK
9325 curr_migr_unit = info->reshape_progress / blocks_per_unit;
9326 /* check if array is alligned to copy area
9327 * if it is not alligned, add one to current migration unit value
9328 * this can happend on array reshape finish only
9329 */
9330 if (info->reshape_progress % blocks_per_unit)
9331 curr_migr_unit++;
687629c2
AK
9332
9333 super->migr_rec->curr_migr_unit =
f8b72ef5 9334 __cpu_to_le32(curr_migr_unit);
687629c2
AK
9335 super->migr_rec->rec_status = __cpu_to_le32(state);
9336 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK
9337 __cpu_to_le32(curr_migr_unit *
9338 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2
AK
9339 if (write_imsm_migr_rec(st) < 0) {
9340 dprintf("imsm: Cannot write migration record "
9341 "outside backup area\n");
9342 return 1;
9343 }
9344
9345 return 0;
9346}
9347
276d77db
AK
9348/*******************************************************************************
9349 * Function: recover_backup_imsm
9350 * Description: Function recovers critical data from the Migration Copy Area
9351 * while assembling an array.
9352 * Parameters:
9353 * super : imsm internal array info
9354 * info : general array info
9355 * Returns:
9356 * 0 : success (or there is no data to recover)
9357 * 1 : fail
9358 ******************************************************************************/
9359int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
9360{
9361 struct intel_super *super = st->sb;
9362 struct migr_record *migr_rec = super->migr_rec;
9363 struct imsm_map *map_dest = NULL;
9364 struct intel_dev *id = NULL;
9365 unsigned long long read_offset;
9366 unsigned long long write_offset;
9367 unsigned unit_len;
9368 int *targets = NULL;
9369 int new_disks, i, err;
9370 char *buf = NULL;
9371 int retval = 1;
9372 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
9373 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 9374 char buffer[20];
6c3560c0 9375 int skipped_disks = 0;
276d77db
AK
9376
9377 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
9378 if (err < 1)
9379 return 1;
9380
9381 /* recover data only during assemblation */
9382 if (strncmp(buffer, "inactive", 8) != 0)
9383 return 0;
9384 /* no data to recover */
9385 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
9386 return 0;
9387 if (curr_migr_unit >= num_migr_units)
9388 return 1;
9389
9390 /* find device during reshape */
9391 for (id = super->devlist; id; id = id->next)
9392 if (is_gen_migration(id->dev))
9393 break;
9394 if (id == NULL)
9395 return 1;
9396
238c0a71 9397 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK
9398 new_disks = map_dest->num_members;
9399
9400 read_offset = (unsigned long long)
9401 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
9402
9403 write_offset = ((unsigned long long)
9404 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 9405 pba_of_lba0(map_dest)) * 512;
276d77db
AK
9406
9407 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
9408 if (posix_memalign((void **)&buf, 512, unit_len) != 0)
9409 goto abort;
9410 targets = malloc(new_disks * sizeof(int));
9411 if (!targets)
9412 goto abort;
9413
9a717282 9414 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
f627f5ad
AK
9415 fprintf(stderr,
9416 Name ": Cannot open some devices belonging to array.\n");
9417 goto abort;
9418 }
276d77db
AK
9419
9420 for (i = 0; i < new_disks; i++) {
6c3560c0
AK
9421 if (targets[i] < 0) {
9422 skipped_disks++;
9423 continue;
9424 }
276d77db
AK
9425 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
9426 fprintf(stderr,
9427 Name ": Cannot seek to block: %s\n",
9428 strerror(errno));
137debce
AK
9429 skipped_disks++;
9430 continue;
276d77db 9431 }
9ec11d1a 9432 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
276d77db
AK
9433 fprintf(stderr,
9434 Name ": Cannot read copy area block: %s\n",
9435 strerror(errno));
137debce
AK
9436 skipped_disks++;
9437 continue;
276d77db
AK
9438 }
9439 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
9440 fprintf(stderr,
9441 Name ": Cannot seek to block: %s\n",
9442 strerror(errno));
137debce
AK
9443 skipped_disks++;
9444 continue;
276d77db 9445 }
9ec11d1a 9446 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
276d77db
AK
9447 fprintf(stderr,
9448 Name ": Cannot restore block: %s\n",
9449 strerror(errno));
137debce
AK
9450 skipped_disks++;
9451 continue;
276d77db
AK
9452 }
9453 }
9454
137debce
AK
9455 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
9456 new_disks,
9457 super,
9458 id->dev)) {
6c3560c0
AK
9459 fprintf(stderr,
9460 Name ": Cannot restore data from backup."
9461 " Too many failed disks\n");
9462 goto abort;
9463 }
9464
befb629b
AK
9465 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
9466 /* ignore error == 2, this can mean end of reshape here
9467 */
9468 dprintf("imsm: Cannot write checkpoint to "
9469 "migration record (UNIT_SRC_NORMAL) during restart\n");
9470 } else
276d77db 9471 retval = 0;
276d77db
AK
9472
9473abort:
9474 if (targets) {
9475 for (i = 0; i < new_disks; i++)
9476 if (targets[i])
9477 close(targets[i]);
9478 free(targets);
9479 }
9480 free(buf);
9481 return retval;
9482}
9483
2cda7640
ML
9484static char disk_by_path[] = "/dev/disk/by-path/";
9485
9486static const char *imsm_get_disk_controller_domain(const char *path)
9487{
2cda7640 9488 char disk_path[PATH_MAX];
96234762
LM
9489 char *drv=NULL;
9490 struct stat st;
2cda7640 9491
96234762
LM
9492 strncpy(disk_path, disk_by_path, PATH_MAX - 1);
9493 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
9494 if (stat(disk_path, &st) == 0) {
9495 struct sys_dev* hba;
9496 char *path=NULL;
9497
9498 path = devt_to_devpath(st.st_rdev);
9499 if (path == NULL)
9500 return "unknown";
9501 hba = find_disk_attached_hba(-1, path);
9502 if (hba && hba->type == SYS_DEV_SAS)
9503 drv = "isci";
9504 else if (hba && hba->type == SYS_DEV_SATA)
9505 drv = "ahci";
9506 else
9507 drv = "unknown";
9508 dprintf("path: %s hba: %s attached: %s\n",
9509 path, (hba) ? hba->path : "NULL", drv);
9510 free(path);
9511 if (hba)
9512 free_sys_dev(&hba);
2cda7640 9513 }
96234762 9514 return drv;
2cda7640
ML
9515}
9516
78b10e66
N
9517static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor)
9518{
9519 char subdev_name[20];
9520 struct mdstat_ent *mdstat;
9521
9522 sprintf(subdev_name, "%d", subdev);
9523 mdstat = mdstat_by_subdev(subdev_name, container);
9524 if (!mdstat)
9525 return -1;
9526
9527 *minor = mdstat->devnum;
9528 free_mdstat(mdstat);
9529 return 0;
9530}
9531
9532static int imsm_reshape_is_allowed_on_container(struct supertype *st,
9533 struct geo_params *geo,
fbf3d202
AK
9534 int *old_raid_disks,
9535 int direction)
78b10e66 9536{
694575e7
KW
9537 /* currently we only support increasing the number of devices
9538 * for a container. This increases the number of device for each
9539 * member array. They must all be RAID0 or RAID5.
9540 */
78b10e66
N
9541 int ret_val = 0;
9542 struct mdinfo *info, *member;
9543 int devices_that_can_grow = 0;
9544
9545 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): "
9546 "st->devnum = (%i)\n",
9547 st->devnum);
9548
9549 if (geo->size != -1 ||
9550 geo->level != UnSet ||
9551 geo->layout != UnSet ||
9552 geo->chunksize != 0 ||
9553 geo->raid_disks == UnSet) {
9554 dprintf("imsm: Container operation is allowed for "
9555 "raid disks number change only.\n");
9556 return ret_val;
9557 }
9558
fbf3d202
AK
9559 if (direction == ROLLBACK_METADATA_CHANGES) {
9560 dprintf("imsm: Metadata changes rollback is not supported for "
9561 "container operation.\n");
9562 return ret_val;
9563 }
9564
78b10e66
N
9565 info = container_content_imsm(st, NULL);
9566 for (member = info; member; member = member->next) {
9567 int result;
9568 int minor;
9569
9570 dprintf("imsm: checking device_num: %i\n",
9571 member->container_member);
9572
d7d205bd 9573 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N
9574 /* we work on container for Online Capacity Expansion
9575 * only so raid_disks has to grow
9576 */
9577 dprintf("imsm: for container operation raid disks "
9578 "increase is required\n");
9579 break;
9580 }
9581
9582 if ((info->array.level != 0) &&
9583 (info->array.level != 5)) {
9584 /* we cannot use this container with other raid level
9585 */
690aae1a 9586 dprintf("imsm: for container operation wrong"
78b10e66
N
9587 " raid level (%i) detected\n",
9588 info->array.level);
9589 break;
9590 } else {
9591 /* check for platform support
9592 * for this raid level configuration
9593 */
9594 struct intel_super *super = st->sb;
9595 if (!is_raid_level_supported(super->orom,
9596 member->array.level,
9597 geo->raid_disks)) {
690aae1a 9598 dprintf("platform does not support raid%d with"
78b10e66
N
9599 " %d disk%s\n",
9600 info->array.level,
9601 geo->raid_disks,
9602 geo->raid_disks > 1 ? "s" : "");
9603 break;
9604 }
2a4a08e7
AK
9605 /* check if component size is aligned to chunk size
9606 */
9607 if (info->component_size %
9608 (info->array.chunk_size/512)) {
9609 dprintf("Component size is not aligned to "
9610 "chunk size\n");
9611 break;
9612 }
78b10e66
N
9613 }
9614
9615 if (*old_raid_disks &&
9616 info->array.raid_disks != *old_raid_disks)
9617 break;
9618 *old_raid_disks = info->array.raid_disks;
9619
9620 /* All raid5 and raid0 volumes in container
9621 * have to be ready for Online Capacity Expansion
9622 * so they need to be assembled. We have already
9623 * checked that no recovery etc is happening.
9624 */
9625 result = imsm_find_array_minor_by_subdev(member->container_member,
9626 st->container_dev,
9627 &minor);
9628 if (result < 0) {
9629 dprintf("imsm: cannot find array\n");
9630 break;
9631 }
9632 devices_that_can_grow++;
9633 }
9634 sysfs_free(info);
9635 if (!member && devices_that_can_grow)
9636 ret_val = 1;
9637
9638 if (ret_val)
9639 dprintf("\tContainer operation allowed\n");
9640 else
9641 dprintf("\tError: %i\n", ret_val);
9642
9643 return ret_val;
9644}
9645
9646/* Function: get_spares_for_grow
9647 * Description: Allocates memory and creates list of spare devices
9648 * avaliable in container. Checks if spare drive size is acceptable.
9649 * Parameters: Pointer to the supertype structure
9650 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
9651 * NULL if fail
9652 */
9653static struct mdinfo *get_spares_for_grow(struct supertype *st)
9654{
78b10e66 9655 unsigned long long min_size = min_acceptable_spare_size_imsm(st);
326727d9 9656 return container_choose_spares(st, min_size, NULL, NULL, NULL, 0);
78b10e66
N
9657}
9658
9659/******************************************************************************
9660 * function: imsm_create_metadata_update_for_reshape
9661 * Function creates update for whole IMSM container.
9662 *
9663 ******************************************************************************/
9664static int imsm_create_metadata_update_for_reshape(
9665 struct supertype *st,
9666 struct geo_params *geo,
9667 int old_raid_disks,
9668 struct imsm_update_reshape **updatep)
9669{
9670 struct intel_super *super = st->sb;
9671 struct imsm_super *mpb = super->anchor;
9672 int update_memory_size = 0;
9673 struct imsm_update_reshape *u = NULL;
9674 struct mdinfo *spares = NULL;
9675 int i;
9676 int delta_disks = 0;
bbd24d86 9677 struct mdinfo *dev;
78b10e66
N
9678
9679 dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n",
9680 geo->raid_disks);
9681
9682 delta_disks = geo->raid_disks - old_raid_disks;
9683
9684 /* size of all update data without anchor */
9685 update_memory_size = sizeof(struct imsm_update_reshape);
9686
9687 /* now add space for spare disks that we need to add. */
9688 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
9689
9690 u = calloc(1, update_memory_size);
9691 if (u == NULL) {
9692 dprintf("error: "
9693 "cannot get memory for imsm_update_reshape update\n");
9694 return 0;
9695 }
9696 u->type = update_reshape_container_disks;
9697 u->old_raid_disks = old_raid_disks;
9698 u->new_raid_disks = geo->raid_disks;
9699
9700 /* now get spare disks list
9701 */
9702 spares = get_spares_for_grow(st);
9703
9704 if (spares == NULL
9705 || delta_disks > spares->array.spare_disks) {
e14e5960
KW
9706 fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices "
9707 "for %s.\n", geo->dev_name);
e4c72d1d 9708 i = -1;
78b10e66
N
9709 goto abort;
9710 }
9711
9712 /* we have got spares
9713 * update disk list in imsm_disk list table in anchor
9714 */
9715 dprintf("imsm: %i spares are available.\n\n",
9716 spares->array.spare_disks);
9717
bbd24d86 9718 dev = spares->devs;
78b10e66 9719 for (i = 0; i < delta_disks; i++) {
78b10e66
N
9720 struct dl *dl;
9721
bbd24d86
AK
9722 if (dev == NULL)
9723 break;
78b10e66
N
9724 u->new_disks[i] = makedev(dev->disk.major,
9725 dev->disk.minor);
9726 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK
9727 dl->index = mpb->num_disks;
9728 mpb->num_disks++;
bbd24d86 9729 dev = dev->next;
78b10e66 9730 }
78b10e66
N
9731
9732abort:
9733 /* free spares
9734 */
9735 sysfs_free(spares);
9736
d677e0b8 9737 dprintf("imsm: reshape update preparation :");
78b10e66 9738 if (i == delta_disks) {
d677e0b8 9739 dprintf(" OK\n");
78b10e66
N
9740 *updatep = u;
9741 return update_memory_size;
9742 }
9743 free(u);
d677e0b8 9744 dprintf(" Error\n");
78b10e66
N
9745
9746 return 0;
9747}
9748
f3871fdc
AK
9749
9750/******************************************************************************
9751 * function: imsm_create_metadata_update_for_size_change()
9752 * Creates update for IMSM array for array size change.
9753 *
9754 ******************************************************************************/
9755static int imsm_create_metadata_update_for_size_change(
9756 struct supertype *st,
9757 struct geo_params *geo,
9758 struct imsm_update_size_change **updatep)
9759{
9760 struct intel_super *super = st->sb;
9761 int update_memory_size = 0;
9762 struct imsm_update_size_change *u = NULL;
9763
9764 dprintf("imsm_create_metadata_update_for_size_change(enter)"
9765 " New size = %llu\n", geo->size);
9766
9767 /* size of all update data without anchor */
9768 update_memory_size = sizeof(struct imsm_update_size_change);
9769
9770 u = calloc(1, update_memory_size);
9771 if (u == NULL) {
9772 dprintf("error: cannot get memory for "
9773 "imsm_create_metadata_update_for_size_change\n");
9774 return 0;
9775 }
9776 u->type = update_size_change;
9777 u->subdev = super->current_vol;
9778 u->new_size = geo->size;
9779
9780 dprintf("imsm: reshape update preparation : OK\n");
9781 *updatep = u;
9782
9783 return update_memory_size;
9784}
9785
48c5303a
PC
9786/******************************************************************************
9787 * function: imsm_create_metadata_update_for_migration()
9788 * Creates update for IMSM array.
9789 *
9790 ******************************************************************************/
9791static int imsm_create_metadata_update_for_migration(
9792 struct supertype *st,
9793 struct geo_params *geo,
9794 struct imsm_update_reshape_migration **updatep)
9795{
9796 struct intel_super *super = st->sb;
9797 int update_memory_size = 0;
9798 struct imsm_update_reshape_migration *u = NULL;
9799 struct imsm_dev *dev;
9800 int previous_level = -1;
9801
9802 dprintf("imsm_create_metadata_update_for_migration(enter)"
9803 " New Level = %i\n", geo->level);
9804
9805 /* size of all update data without anchor */
9806 update_memory_size = sizeof(struct imsm_update_reshape_migration);
9807
9808 u = calloc(1, update_memory_size);
9809 if (u == NULL) {
9810 dprintf("error: cannot get memory for "
9811 "imsm_create_metadata_update_for_migration\n");
9812 return 0;
9813 }
9814 u->type = update_reshape_migration;
9815 u->subdev = super->current_vol;
9816 u->new_level = geo->level;
9817 u->new_layout = geo->layout;
9818 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
9819 u->new_disks[0] = -1;
4bba0439 9820 u->new_chunksize = -1;
48c5303a
PC
9821
9822 dev = get_imsm_dev(super, u->subdev);
9823 if (dev) {
9824 struct imsm_map *map;
9825
238c0a71 9826 map = get_imsm_map(dev, MAP_0);
4bba0439
PC
9827 if (map) {
9828 int current_chunk_size =
9829 __le16_to_cpu(map->blocks_per_strip) / 2;
9830
9831 if (geo->chunksize != current_chunk_size) {
9832 u->new_chunksize = geo->chunksize / 1024;
9833 dprintf("imsm: "
9834 "chunk size change from %i to %i\n",
9835 current_chunk_size, u->new_chunksize);
9836 }
48c5303a 9837 previous_level = map->raid_level;
4bba0439 9838 }
48c5303a
PC
9839 }
9840 if ((geo->level == 5) && (previous_level == 0)) {
9841 struct mdinfo *spares = NULL;
9842
9843 u->new_raid_disks++;
9844 spares = get_spares_for_grow(st);
9845 if ((spares == NULL) || (spares->array.spare_disks < 1)) {
9846 free(u);
9847 sysfs_free(spares);
9848 update_memory_size = 0;
9849 dprintf("error: cannot get spare device "
9850 "for requested migration");
9851 return 0;
9852 }
9853 sysfs_free(spares);
9854 }
9855 dprintf("imsm: reshape update preparation : OK\n");
9856 *updatep = u;
9857
9858 return update_memory_size;
9859}
9860
8dd70bce
AK
9861static void imsm_update_metadata_locally(struct supertype *st,
9862 void *buf, int len)
9863{
9864 struct metadata_update mu;
9865
9866 mu.buf = buf;
9867 mu.len = len;
9868 mu.space = NULL;
9869 mu.space_list = NULL;
9870 mu.next = NULL;
9871 imsm_prepare_update(st, &mu);
9872 imsm_process_update(st, &mu);
9873
9874 while (mu.space_list) {
9875 void **space = mu.space_list;
9876 mu.space_list = *space;
9877 free(space);
9878 }
9879}
78b10e66 9880
471bceb6 9881/***************************************************************************
694575e7 9882* Function: imsm_analyze_change
471bceb6
KW
9883* Description: Function analyze change for single volume
9884* and validate if transition is supported
fbf3d202
AK
9885* Parameters: Geometry parameters, supertype structure,
9886* metadata change direction (apply/rollback)
694575e7 9887* Returns: Operation type code on success, -1 if fail
471bceb6
KW
9888****************************************************************************/
9889enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK
9890 struct geo_params *geo,
9891 int direction)
694575e7 9892{
471bceb6
KW
9893 struct mdinfo info;
9894 int change = -1;
9895 int check_devs = 0;
c21e737b 9896 int chunk;
67a2db32
AK
9897 /* number of added/removed disks in operation result */
9898 int devNumChange = 0;
9899 /* imsm compatible layout value for array geometry verification */
9900 int imsm_layout = -1;
7abc9871
AK
9901 int data_disks;
9902 struct imsm_dev *dev;
9903 struct intel_super *super;
9904 long long current_size;
471bceb6
KW
9905
9906 getinfo_super_imsm_volume(st, &info, NULL);
471bceb6
KW
9907 if ((geo->level != info.array.level) &&
9908 (geo->level >= 0) &&
9909 (geo->level != UnSet)) {
9910 switch (info.array.level) {
9911 case 0:
9912 if (geo->level == 5) {
b5347799 9913 change = CH_MIGRATION;
e13ce846
AK
9914 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
9915 fprintf(stderr,
9916 Name " Error. Requested Layout "
9917 "not supported (left-asymmetric layout "
9918 "is supported only)!\n");
9919 change = -1;
9920 goto analyse_change_exit;
9921 }
67a2db32 9922 imsm_layout = geo->layout;
471bceb6 9923 check_devs = 1;
e91a3bad
LM
9924 devNumChange = 1; /* parity disk added */
9925 } else if (geo->level == 10) {
471bceb6
KW
9926 change = CH_TAKEOVER;
9927 check_devs = 1;
e91a3bad 9928 devNumChange = 2; /* two mirrors added */
67a2db32 9929 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 9930 }
dfe77a9e
KW
9931 break;
9932 case 1:
471bceb6
KW
9933 case 10:
9934 if (geo->level == 0) {
9935 change = CH_TAKEOVER;
9936 check_devs = 1;
e91a3bad 9937 devNumChange = -(geo->raid_disks/2);
67a2db32 9938 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW
9939 }
9940 break;
9941 }
9942 if (change == -1) {
9943 fprintf(stderr,
9944 Name " Error. Level Migration from %d to %d "
9945 "not supported!\n",
9946 info.array.level, geo->level);
9947 goto analyse_change_exit;
9948 }
9949 } else
9950 geo->level = info.array.level;
9951
9952 if ((geo->layout != info.array.layout)
9953 && ((geo->layout != UnSet) && (geo->layout != -1))) {
b5347799 9954 change = CH_MIGRATION;
471bceb6
KW
9955 if ((info.array.layout == 0)
9956 && (info.array.level == 5)
9957 && (geo->layout == 5)) {
9958 /* reshape 5 -> 4 */
9959 } else if ((info.array.layout == 5)
9960 && (info.array.level == 5)
9961 && (geo->layout == 0)) {
9962 /* reshape 4 -> 5 */
9963 geo->layout = 0;
9964 geo->level = 5;
9965 } else {
9966 fprintf(stderr,
9967 Name " Error. Layout Migration from %d to %d "
9968 "not supported!\n",
9969 info.array.layout, geo->layout);
9970 change = -1;
9971 goto analyse_change_exit;
9972 }
67a2db32 9973 } else {
471bceb6 9974 geo->layout = info.array.layout;
67a2db32
AK
9975 if (imsm_layout == -1)
9976 imsm_layout = info.array.layout;
9977 }
471bceb6
KW
9978
9979 if ((geo->chunksize > 0) && (geo->chunksize != UnSet)
9980 && (geo->chunksize != info.array.chunk_size))
b5347799 9981 change = CH_MIGRATION;
471bceb6
KW
9982 else
9983 geo->chunksize = info.array.chunk_size;
9984
c21e737b 9985 chunk = geo->chunksize / 1024;
7abc9871
AK
9986
9987 super = st->sb;
9988 dev = get_imsm_dev(super, super->current_vol);
9989 data_disks = imsm_num_data_members(dev , MAP_0);
c41e00b2 9990 /* compute current size per disk member
7abc9871 9991 */
c41e00b2
AK
9992 current_size = info.custom_array_size / data_disks;
9993
9994 if (geo->size > 0) {
9995 /* align component size
9996 */
9997 geo->size = imsm_component_size_aligment_check(
9998 get_imsm_raid_level(dev->vol.map),
9999 chunk * 1024,
10000 geo->size * 2);
10001 }
7abc9871 10002
b130333f 10003 if ((current_size != geo->size) && (geo->size >= 0)) {
7abc9871
AK
10004 if (change != -1) {
10005 fprintf(stderr,
10006 Name " Error. Size change should be the only "
10007 "one at a time.\n");
10008 change = -1;
10009 goto analyse_change_exit;
10010 }
10011 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
10012 fprintf(stderr,
10013 Name " Error. The last volume in container "
10014 "can be expanded only (%i/%i).\n",
10015 super->current_vol, st->devnum);
10016 goto analyse_change_exit;
10017 }
b130333f
AK
10018 if (geo->size == 0) {
10019 /* requested size change to the maximum available size
10020 */
10021 unsigned long long freesize;
10022 int rv;
10023
10024 rv = imsm_get_free_size(st, dev->vol.map->num_members,
10025 0, chunk, &freesize);
10026 if (rv == 0) {
10027 fprintf(stderr, Name " Error. Cannot find "
10028 "maximum available space.\n");
10029 change = -1;
10030 goto analyse_change_exit;
10031 }
10032 geo->size = freesize + current_size;
10033
c41e00b2
AK
10034 /* align component size
10035 */
10036 geo->size = imsm_component_size_aligment_check(
10037 get_imsm_raid_level(dev->vol.map),
10038 chunk * 1024,
10039 geo->size);
10040 }
b130333f 10041
fbf3d202
AK
10042 if ((direction == ROLLBACK_METADATA_CHANGES)) {
10043 /* accept size for rollback only
10044 */
10045 } else {
10046 /* round size due to metadata compatibility
10047 */
10048 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
10049 << SECT_PER_MB_SHIFT;
10050 dprintf("Prepare update for size change to %llu\n",
10051 geo->size );
10052 if (current_size >= geo->size) {
10053 fprintf(stderr,
10054 Name " Error. Size expanssion is "
10055 "supported only (current size is %llu, "
10056 "requested size /rounded/ is %llu).\n",
10057 current_size, geo->size);
10058 goto analyse_change_exit;
10059 }
7abc9871
AK
10060 }
10061 geo->size *= data_disks;
10062 geo->raid_disks = dev->vol.map->num_members;
10063 change = CH_ARRAY_SIZE;
10064 }
471bceb6
KW
10065 if (!validate_geometry_imsm(st,
10066 geo->level,
67a2db32 10067 imsm_layout,
e91a3bad 10068 geo->raid_disks + devNumChange,
c21e737b 10069 &chunk,
471bceb6
KW
10070 geo->size,
10071 0, 0, 1))
10072 change = -1;
10073
10074 if (check_devs) {
10075 struct intel_super *super = st->sb;
10076 struct imsm_super *mpb = super->anchor;
10077
10078 if (mpb->num_raid_devs > 1) {
10079 fprintf(stderr,
10080 Name " Error. Cannot perform operation on %s"
10081 "- for this operation it MUST be single "
10082 "array in container\n",
10083 geo->dev_name);
10084 change = -1;
10085 }
10086 }
10087
10088analyse_change_exit:
fbf3d202
AK
10089 if ((direction == ROLLBACK_METADATA_CHANGES) &&
10090 ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) {
10091 dprintf("imsm: Metadata changes rollback is not supported for "
10092 "migration and takeover operations.\n");
10093 change = -1;
10094 }
471bceb6 10095 return change;
694575e7
KW
10096}
10097
bb025c2f
KW
10098int imsm_takeover(struct supertype *st, struct geo_params *geo)
10099{
10100 struct intel_super *super = st->sb;
10101 struct imsm_update_takeover *u;
10102
10103 u = malloc(sizeof(struct imsm_update_takeover));
10104 if (u == NULL)
10105 return 1;
10106
10107 u->type = update_takeover;
10108 u->subarray = super->current_vol;
10109
10110 /* 10->0 transition */
10111 if (geo->level == 0)
10112 u->direction = R10_TO_R0;
10113
0529c688
KW
10114 /* 0->10 transition */
10115 if (geo->level == 10)
10116 u->direction = R0_TO_R10;
10117
bb025c2f
KW
10118 /* update metadata locally */
10119 imsm_update_metadata_locally(st, u,
10120 sizeof(struct imsm_update_takeover));
10121 /* and possibly remotely */
10122 if (st->update_tail)
10123 append_metadata_update(st, u,
10124 sizeof(struct imsm_update_takeover));
10125 else
10126 free(u);
10127
10128 return 0;
10129}
10130
78b10e66
N
10131static int imsm_reshape_super(struct supertype *st, long long size, int level,
10132 int layout, int chunksize, int raid_disks,
41784c88 10133 int delta_disks, char *backup, char *dev,
016e00f5 10134 int direction, int verbose)
78b10e66 10135{
78b10e66
N
10136 int ret_val = 1;
10137 struct geo_params geo;
10138
10139 dprintf("imsm: reshape_super called.\n");
10140
71204a50 10141 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N
10142
10143 geo.dev_name = dev;
694575e7 10144 geo.dev_id = st->devnum;
78b10e66
N
10145 geo.size = size;
10146 geo.level = level;
10147 geo.layout = layout;
10148 geo.chunksize = chunksize;
10149 geo.raid_disks = raid_disks;
41784c88
AK
10150 if (delta_disks != UnSet)
10151 geo.raid_disks += delta_disks;
78b10e66
N
10152
10153 dprintf("\tfor level : %i\n", geo.level);
10154 dprintf("\tfor raid_disks : %i\n", geo.raid_disks);
10155
10156 if (experimental() == 0)
10157 return ret_val;
10158
78b10e66 10159 if (st->container_dev == st->devnum) {
694575e7
KW
10160 /* On container level we can only increase number of devices. */
10161 dprintf("imsm: info: Container operation\n");
78b10e66 10162 int old_raid_disks = 0;
6dc0be30 10163
78b10e66 10164 if (imsm_reshape_is_allowed_on_container(
fbf3d202 10165 st, &geo, &old_raid_disks, direction)) {
78b10e66
N
10166 struct imsm_update_reshape *u = NULL;
10167 int len;
10168
10169 len = imsm_create_metadata_update_for_reshape(
10170 st, &geo, old_raid_disks, &u);
10171
ed08d51c
AK
10172 if (len <= 0) {
10173 dprintf("imsm: Cannot prepare update\n");
10174 goto exit_imsm_reshape_super;
10175 }
10176
8dd70bce
AK
10177 ret_val = 0;
10178 /* update metadata locally */
10179 imsm_update_metadata_locally(st, u, len);
10180 /* and possibly remotely */
10181 if (st->update_tail)
10182 append_metadata_update(st, u, len);
10183 else
ed08d51c 10184 free(u);
8dd70bce 10185
694575e7 10186 } else {
e7ff7e40
AK
10187 fprintf(stderr, Name ": (imsm) Operation "
10188 "is not allowed on this container\n");
694575e7
KW
10189 }
10190 } else {
10191 /* On volume level we support following operations
471bceb6
KW
10192 * - takeover: raid10 -> raid0; raid0 -> raid10
10193 * - chunk size migration
10194 * - migration: raid5 -> raid0; raid0 -> raid5
10195 */
10196 struct intel_super *super = st->sb;
10197 struct intel_dev *dev = super->devlist;
10198 int change, devnum;
694575e7 10199 dprintf("imsm: info: Volume operation\n");
471bceb6
KW
10200 /* find requested device */
10201 while (dev) {
19986c72
MB
10202 if (imsm_find_array_minor_by_subdev(
10203 dev->index, st->container_dev, &devnum) == 0
10204 && devnum == geo.dev_id)
471bceb6
KW
10205 break;
10206 dev = dev->next;
10207 }
10208 if (dev == NULL) {
10209 fprintf(stderr, Name " Cannot find %s (%i) subarray\n",
10210 geo.dev_name, geo.dev_id);
10211 goto exit_imsm_reshape_super;
10212 }
10213 super->current_vol = dev->index;
fbf3d202 10214 change = imsm_analyze_change(st, &geo, direction);
694575e7 10215 switch (change) {
471bceb6 10216 case CH_TAKEOVER:
bb025c2f 10217 ret_val = imsm_takeover(st, &geo);
694575e7 10218 break;
48c5303a
PC
10219 case CH_MIGRATION: {
10220 struct imsm_update_reshape_migration *u = NULL;
10221 int len =
10222 imsm_create_metadata_update_for_migration(
10223 st, &geo, &u);
10224 if (len < 1) {
10225 dprintf("imsm: "
10226 "Cannot prepare update\n");
10227 break;
10228 }
471bceb6 10229 ret_val = 0;
48c5303a
PC
10230 /* update metadata locally */
10231 imsm_update_metadata_locally(st, u, len);
10232 /* and possibly remotely */
10233 if (st->update_tail)
10234 append_metadata_update(st, u, len);
10235 else
10236 free(u);
10237 }
10238 break;
7abc9871 10239 case CH_ARRAY_SIZE: {
f3871fdc
AK
10240 struct imsm_update_size_change *u = NULL;
10241 int len =
10242 imsm_create_metadata_update_for_size_change(
10243 st, &geo, &u);
10244 if (len < 1) {
10245 dprintf("imsm: "
10246 "Cannot prepare update\n");
10247 break;
10248 }
10249 ret_val = 0;
10250 /* update metadata locally */
10251 imsm_update_metadata_locally(st, u, len);
10252 /* and possibly remotely */
10253 if (st->update_tail)
10254 append_metadata_update(st, u, len);
10255 else
10256 free(u);
7abc9871
AK
10257 }
10258 break;
471bceb6
KW
10259 default:
10260 ret_val = 1;
694575e7 10261 }
694575e7 10262 }
78b10e66 10263
ed08d51c 10264exit_imsm_reshape_super:
78b10e66
N
10265 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
10266 return ret_val;
10267}
2cda7640 10268
eee67a47
AK
10269/*******************************************************************************
10270 * Function: wait_for_reshape_imsm
10271 * Description: Function writes new sync_max value and waits until
10272 * reshape process reach new position
10273 * Parameters:
10274 * sra : general array info
eee67a47
AK
10275 * ndata : number of disks in new array's layout
10276 * Returns:
10277 * 0 : success,
10278 * 1 : there is no reshape in progress,
10279 * -1 : fail
10280 ******************************************************************************/
ae9f01f8 10281int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47
AK
10282{
10283 int fd = sysfs_get_fd(sra, NULL, "reshape_position");
10284 unsigned long long completed;
ae9f01f8
AK
10285 /* to_complete : new sync_max position */
10286 unsigned long long to_complete = sra->reshape_progress;
10287 unsigned long long position_to_set = to_complete / ndata;
eee67a47 10288
ae9f01f8
AK
10289 if (fd < 0) {
10290 dprintf("imsm: wait_for_reshape_imsm() "
10291 "cannot open reshape_position\n");
eee67a47 10292 return 1;
ae9f01f8 10293 }
eee67a47 10294
ae9f01f8
AK
10295 if (sysfs_fd_get_ll(fd, &completed) < 0) {
10296 dprintf("imsm: wait_for_reshape_imsm() "
10297 "cannot read reshape_position (no reshape in progres)\n");
10298 close(fd);
10299 return 0;
10300 }
eee67a47 10301
ae9f01f8
AK
10302 if (completed > to_complete) {
10303 dprintf("imsm: wait_for_reshape_imsm() "
10304 "wrong next position to set %llu (%llu)\n",
10305 to_complete, completed);
10306 close(fd);
10307 return -1;
10308 }
10309 dprintf("Position set: %llu\n", position_to_set);
10310 if (sysfs_set_num(sra, NULL, "sync_max",
10311 position_to_set) != 0) {
10312 dprintf("imsm: wait_for_reshape_imsm() "
10313 "cannot set reshape position to %llu\n",
10314 position_to_set);
10315 close(fd);
10316 return -1;
eee67a47
AK
10317 }
10318
eee67a47
AK
10319 do {
10320 char action[20];
10321 fd_set rfds;
10322 FD_ZERO(&rfds);
10323 FD_SET(fd, &rfds);
a47e44fb
AK
10324 select(fd+1, &rfds, NULL, NULL, NULL);
10325 if (sysfs_get_str(sra, NULL, "sync_action",
10326 action, 20) > 0 &&
10327 strncmp(action, "reshape", 7) != 0)
10328 break;
eee67a47 10329 if (sysfs_fd_get_ll(fd, &completed) < 0) {
ae9f01f8
AK
10330 dprintf("imsm: wait_for_reshape_imsm() "
10331 "cannot read reshape_position (in loop)\n");
eee67a47
AK
10332 close(fd);
10333 return 1;
10334 }
eee67a47
AK
10335 } while (completed < to_complete);
10336 close(fd);
10337 return 0;
10338
10339}
10340
b915c95f
AK
10341/*******************************************************************************
10342 * Function: check_degradation_change
10343 * Description: Check that array hasn't become failed.
10344 * Parameters:
10345 * info : for sysfs access
10346 * sources : source disks descriptors
10347 * degraded: previous degradation level
10348 * Returns:
10349 * degradation level
10350 ******************************************************************************/
10351int check_degradation_change(struct mdinfo *info,
10352 int *sources,
10353 int degraded)
10354{
10355 unsigned long long new_degraded;
10356 sysfs_get_ll(info, NULL, "degraded", &new_degraded);
10357 if (new_degraded != (unsigned long long)degraded) {
10358 /* check each device to ensure it is still working */
10359 struct mdinfo *sd;
10360 new_degraded = 0;
10361 for (sd = info->devs ; sd ; sd = sd->next) {
10362 if (sd->disk.state & (1<<MD_DISK_FAULTY))
10363 continue;
10364 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
10365 char sbuf[20];
10366 if (sysfs_get_str(info,
10367 sd, "state", sbuf, 20) < 0 ||
10368 strstr(sbuf, "faulty") ||
10369 strstr(sbuf, "in_sync") == NULL) {
10370 /* this device is dead */
10371 sd->disk.state = (1<<MD_DISK_FAULTY);
10372 if (sd->disk.raid_disk >= 0 &&
10373 sources[sd->disk.raid_disk] >= 0) {
10374 close(sources[
10375 sd->disk.raid_disk]);
10376 sources[sd->disk.raid_disk] =
10377 -1;
10378 }
10379 new_degraded++;
10380 }
10381 }
10382 }
10383 }
10384
10385 return new_degraded;
10386}
10387
10f22854
AK
10388/*******************************************************************************
10389 * Function: imsm_manage_reshape
10390 * Description: Function finds array under reshape and it manages reshape
10391 * process. It creates stripes backups (if required) and sets
10392 * checheckpoits.
10393 * Parameters:
10394 * afd : Backup handle (nattive) - not used
10395 * sra : general array info
10396 * reshape : reshape parameters - not used
10397 * st : supertype structure
10398 * blocks : size of critical section [blocks]
10399 * fds : table of source device descriptor
10400 * offsets : start of array (offest per devices)
10401 * dests : not used
10402 * destfd : table of destination device descriptor
10403 * destoffsets : table of destination offsets (per device)
10404 * Returns:
10405 * 1 : success, reshape is done
10406 * 0 : fail
10407 ******************************************************************************/
999b4972
N
10408static int imsm_manage_reshape(
10409 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 10410 struct supertype *st, unsigned long backup_blocks,
999b4972
N
10411 int *fds, unsigned long long *offsets,
10412 int dests, int *destfd, unsigned long long *destoffsets)
10413{
10f22854
AK
10414 int ret_val = 0;
10415 struct intel_super *super = st->sb;
10416 struct intel_dev *dv = NULL;
10417 struct imsm_dev *dev = NULL;
a6b6d984 10418 struct imsm_map *map_src;
10f22854
AK
10419 int migr_vol_qan = 0;
10420 int ndata, odata; /* [bytes] */
10421 int chunk; /* [bytes] */
10422 struct migr_record *migr_rec;
10423 char *buf = NULL;
10424 unsigned int buf_size; /* [bytes] */
10425 unsigned long long max_position; /* array size [bytes] */
10426 unsigned long long next_step; /* [blocks]/[bytes] */
10427 unsigned long long old_data_stripe_length;
10f22854
AK
10428 unsigned long long start_src; /* [bytes] */
10429 unsigned long long start; /* [bytes] */
10430 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 10431 int degraded = 0;
ab724b98 10432 int source_layout = 0;
10f22854 10433
1ab242d8 10434 if (!fds || !offsets || !sra)
10f22854
AK
10435 goto abort;
10436
10437 /* Find volume during the reshape */
10438 for (dv = super->devlist; dv; dv = dv->next) {
10439 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR
10440 && dv->dev->vol.migr_state == 1) {
10441 dev = dv->dev;
10442 migr_vol_qan++;
10443 }
10444 }
10445 /* Only one volume can migrate at the same time */
10446 if (migr_vol_qan != 1) {
10447 fprintf(stderr, Name " : %s", migr_vol_qan ?
10448 "Number of migrating volumes greater than 1\n" :
10449 "There is no volume during migrationg\n");
10450 goto abort;
10451 }
10452
238c0a71 10453 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK
10454 if (map_src == NULL)
10455 goto abort;
10f22854 10456
238c0a71
AK
10457 ndata = imsm_num_data_members(dev, MAP_0);
10458 odata = imsm_num_data_members(dev, MAP_1);
10f22854 10459
7b1ab482 10460 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK
10461 old_data_stripe_length = odata * chunk;
10462
10463 migr_rec = super->migr_rec;
10464
10f22854
AK
10465 /* initialize migration record for start condition */
10466 if (sra->reshape_progress == 0)
10467 init_migr_record_imsm(st, dev, sra);
b2c59438
AK
10468 else {
10469 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
10470 dprintf("imsm: cannot restart migration when data "
10471 "are present in copy area.\n");
10472 goto abort;
10473 }
6a75c8ca
AK
10474 /* Save checkpoint to update migration record for current
10475 * reshape position (in md). It can be farther than current
10476 * reshape position in metadata.
10477 */
10478 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
10479 /* ignore error == 2, this can mean end of reshape here
10480 */
10481 dprintf("imsm: Cannot write checkpoint to "
10482 "migration record (UNIT_SRC_NORMAL, "
10483 "initial save)\n");
10484 goto abort;
10485 }
b2c59438 10486 }
10f22854
AK
10487
10488 /* size for data */
10489 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
10490 /* extend buffer size for parity disk */
10491 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
10492 /* add space for stripe aligment */
10493 buf_size += old_data_stripe_length;
10494 if (posix_memalign((void **)&buf, 4096, buf_size)) {
10495 dprintf("imsm: Cannot allocate checpoint buffer\n");
10496 goto abort;
10497 }
10498
3ef4403c 10499 max_position = sra->component_size * ndata;
68eb8bc6 10500 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK
10501
10502 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
10503 __le32_to_cpu(migr_rec->num_migr_units)) {
10504 /* current reshape position [blocks] */
10505 unsigned long long current_position =
10506 __le32_to_cpu(migr_rec->blocks_per_unit)
10507 * __le32_to_cpu(migr_rec->curr_migr_unit);
10508 unsigned long long border;
10509
b915c95f
AK
10510 /* Check that array hasn't become failed.
10511 */
10512 degraded = check_degradation_change(sra, fds, degraded);
10513 if (degraded > 1) {
10514 dprintf("imsm: Abort reshape due to degradation"
10515 " level (%i)\n", degraded);
10516 goto abort;
10517 }
10518
10f22854
AK
10519 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
10520
10521 if ((current_position + next_step) > max_position)
10522 next_step = max_position - current_position;
10523
92144abf 10524 start = current_position * 512;
10f22854
AK
10525
10526 /* allign reading start to old geometry */
10527 start_buf_shift = start % old_data_stripe_length;
10528 start_src = start - start_buf_shift;
10529
10530 border = (start_src / odata) - (start / ndata);
10531 border /= 512;
10532 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
10533 /* save critical stripes to buf
10534 * start - start address of current unit
10535 * to backup [bytes]
10536 * start_src - start address of current unit
10537 * to backup alligned to source array
10538 * [bytes]
10539 */
10540 unsigned long long next_step_filler = 0;
10541 unsigned long long copy_length = next_step * 512;
10542
10543 /* allign copy area length to stripe in old geometry */
10544 next_step_filler = ((copy_length + start_buf_shift)
10545 % old_data_stripe_length);
10546 if (next_step_filler)
10547 next_step_filler = (old_data_stripe_length
10548 - next_step_filler);
10549 dprintf("save_stripes() parameters: start = %llu,"
10550 "\tstart_src = %llu,\tnext_step*512 = %llu,"
10551 "\tstart_in_buf_shift = %llu,"
10552 "\tnext_step_filler = %llu\n",
10553 start, start_src, copy_length,
10554 start_buf_shift, next_step_filler);
10555
10556 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK
10557 chunk, map_src->raid_level,
10558 source_layout, 0, NULL, start_src,
10f22854
AK
10559 copy_length +
10560 next_step_filler + start_buf_shift,
10561 buf)) {
10562 dprintf("imsm: Cannot save stripes"
10563 " to buffer\n");
10564 goto abort;
10565 }
10566 /* Convert data to destination format and store it
10567 * in backup general migration area
10568 */
10569 if (save_backup_imsm(st, dev, sra,
aea93171 10570 buf + start_buf_shift, copy_length)) {
10f22854
AK
10571 dprintf("imsm: Cannot save stripes to "
10572 "target devices\n");
10573 goto abort;
10574 }
10575 if (save_checkpoint_imsm(st, sra,
10576 UNIT_SRC_IN_CP_AREA)) {
10577 dprintf("imsm: Cannot write checkpoint to "
10578 "migration record (UNIT_SRC_IN_CP_AREA)\n");
10579 goto abort;
10580 }
8016a6d4
AK
10581 } else {
10582 /* set next step to use whole border area */
10583 border /= next_step;
10584 if (border > 1)
10585 next_step *= border;
10f22854
AK
10586 }
10587 /* When data backed up, checkpoint stored,
10588 * kick the kernel to reshape unit of data
10589 */
10590 next_step = next_step + sra->reshape_progress;
8016a6d4
AK
10591 /* limit next step to array max position */
10592 if (next_step > max_position)
10593 next_step = max_position;
10f22854
AK
10594 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
10595 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 10596 sra->reshape_progress = next_step;
10f22854
AK
10597
10598 /* wait until reshape finish */
ae9f01f8 10599 if (wait_for_reshape_imsm(sra, ndata) < 0) {
c47b0ff6
AK
10600 dprintf("wait_for_reshape_imsm returned error!\n");
10601 goto abort;
10602 }
10f22854 10603
0228d92c
AK
10604 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
10605 /* ignore error == 2, this can mean end of reshape here
10606 */
10f22854
AK
10607 dprintf("imsm: Cannot write checkpoint to "
10608 "migration record (UNIT_SRC_NORMAL)\n");
10609 goto abort;
10610 }
10611
10612 }
10613
10614 /* return '1' if done */
10615 ret_val = 1;
10616abort:
10617 free(buf);
10618 abort_reshape(sra);
10619
10620 return ret_val;
999b4972 10621}
71204a50 10622#endif /* MDASSEMBLE */
999b4972 10623
cdddbdbc
DW
10624struct superswitch super_imsm = {
10625#ifndef MDASSEMBLE
10626 .examine_super = examine_super_imsm,
10627 .brief_examine_super = brief_examine_super_imsm,
4737ae25 10628 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 10629 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW
10630 .detail_super = detail_super_imsm,
10631 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 10632 .write_init_super = write_init_super_imsm,
0e600426
N
10633 .validate_geometry = validate_geometry_imsm,
10634 .add_to_super = add_to_super_imsm,
1a64be56 10635 .remove_from_super = remove_from_super_imsm,
d665cc31 10636 .detail_platform = detail_platform_imsm,
33414a01 10637 .kill_subarray = kill_subarray_imsm,
aa534678 10638 .update_subarray = update_subarray_imsm,
2b959fbf 10639 .load_container = load_container_imsm,
71204a50
N
10640 .default_geometry = default_geometry_imsm,
10641 .get_disk_controller_domain = imsm_get_disk_controller_domain,
10642 .reshape_super = imsm_reshape_super,
10643 .manage_reshape = imsm_manage_reshape,
9e2d750d 10644 .recover_backup = recover_backup_imsm,
cdddbdbc
DW
10645#endif
10646 .match_home = match_home_imsm,
10647 .uuid_from_super= uuid_from_super_imsm,
10648 .getinfo_super = getinfo_super_imsm,
5c4cd5da 10649 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW
10650 .update_super = update_super_imsm,
10651
10652 .avail_size = avail_size_imsm,
80e7f8c3 10653 .min_acceptable_spare_size = min_acceptable_spare_size_imsm,
cdddbdbc
DW
10654
10655 .compare_super = compare_super_imsm,
10656
10657 .load_super = load_super_imsm,
bf5a934a 10658 .init_super = init_super_imsm,
e683ca88 10659 .store_super = store_super_imsm,
cdddbdbc
DW
10660 .free_super = free_super_imsm,
10661 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 10662 .container_content = container_content_imsm,
cdddbdbc 10663
276d77db 10664
cdddbdbc 10665 .external = 1,
4cce4069 10666 .name = "imsm",
845dea95 10667
0e600426 10668#ifndef MDASSEMBLE
845dea95
NB
10669/* for mdmon */
10670 .open_new = imsm_open_new,
ed9d66aa 10671 .set_array_state= imsm_set_array_state,
845dea95
NB
10672 .set_disk = imsm_set_disk,
10673 .sync_metadata = imsm_sync_metadata,
88758e9d 10674 .activate_spare = imsm_activate_spare,
e8319a19 10675 .process_update = imsm_process_update,
8273f55e 10676 .prepare_update = imsm_prepare_update,
0e600426 10677#endif /* MDASSEMBLE */
cdddbdbc 10678};