19888c340SDavid Sterba /* SPDX-License-Identifier: GPL-2.0 */ 20b86a832SChris Mason /* 30b86a832SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 40b86a832SChris Mason */ 50b86a832SChris Mason 69888c340SDavid Sterba #ifndef BTRFS_VOLUMES_H 79888c340SDavid Sterba #define BTRFS_VOLUMES_H 88790d502SChris Mason 9cea9e445SChris Mason #include <linux/bio.h> 10b2117a39SMiao Xie #include <linux/sort.h> 1155e301fdSFilipe Brandenburger #include <linux/btrfs.h> 128b712842SChris Mason #include "async-thread.h" 13cea9e445SChris Mason 14fce466eaSQu Wenruo #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G) 15fce466eaSQu Wenruo 1667a2c45eSMiao Xie extern struct mutex uuid_mutex; 1767a2c45eSMiao Xie 18ee22184bSByongho Lee #define BTRFS_STRIPE_LEN SZ_64K 19b2117a39SMiao Xie 20719fae89SQu Wenruo /* Used by sanity check for btrfs_raid_types. */ 21719fae89SQu Wenruo #define const_ffs(n) (__builtin_ctzll(n) + 1) 22719fae89SQu Wenruo 23719fae89SQu Wenruo /* 24719fae89SQu Wenruo * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires 25719fae89SQu Wenruo * RAID0 always to be the lowest profile bit. 26719fae89SQu Wenruo * Although it's part of on-disk format and should never change, do extra 27719fae89SQu Wenruo * compile-time sanity checks. 28719fae89SQu Wenruo */ 29719fae89SQu Wenruo static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) < 30719fae89SQu Wenruo const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0)); 31719fae89SQu Wenruo static_assert(const_ilog2(BTRFS_BLOCK_GROUP_RAID0) > 32719fae89SQu Wenruo ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK)); 33719fae89SQu Wenruo 34719fae89SQu Wenruo /* ilog2() can handle both constants and variables */ 35719fae89SQu Wenruo #define BTRFS_BG_FLAG_TO_INDEX(profile) \ 36719fae89SQu Wenruo ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1)) 37719fae89SQu Wenruo 38f04fbcc6SQu Wenruo enum btrfs_raid_types { 39719fae89SQu Wenruo /* SINGLE is the special one as it doesn't have on-disk bit. */ 40719fae89SQu Wenruo BTRFS_RAID_SINGLE = 0, 41719fae89SQu Wenruo 42719fae89SQu Wenruo BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0), 43719fae89SQu Wenruo BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1), 44719fae89SQu Wenruo BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP), 45719fae89SQu Wenruo BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10), 46719fae89SQu Wenruo BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5), 47719fae89SQu Wenruo BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6), 48719fae89SQu Wenruo BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3), 49719fae89SQu Wenruo BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4), 50719fae89SQu Wenruo 51f04fbcc6SQu Wenruo BTRFS_NR_RAID_TYPES 52f04fbcc6SQu Wenruo }; 53f04fbcc6SQu Wenruo 545f141126SNikolay Borisov struct btrfs_io_geometry { 555f141126SNikolay Borisov /* remaining bytes before crossing a stripe */ 565f141126SNikolay Borisov u64 len; 575f141126SNikolay Borisov /* offset of logical address in chunk */ 585f141126SNikolay Borisov u64 offset; 595f141126SNikolay Borisov /* length of single IO stripe */ 60cc353a8bSQu Wenruo u32 stripe_len; 61cc353a8bSQu Wenruo /* offset of address in stripe */ 62cc353a8bSQu Wenruo u32 stripe_offset; 635f141126SNikolay Borisov /* number of stripe where address falls */ 645f141126SNikolay Borisov u64 stripe_nr; 655f141126SNikolay Borisov /* offset of raid56 stripe into the chunk */ 665f141126SNikolay Borisov u64 raid56_stripe_offset; 675f141126SNikolay Borisov }; 685f141126SNikolay Borisov 697cc8e58dSMiao Xie /* 707cc8e58dSMiao Xie * Use sequence counter to get consistent device stat data on 717cc8e58dSMiao Xie * 32-bit processors. 727cc8e58dSMiao Xie */ 737cc8e58dSMiao Xie #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 747cc8e58dSMiao Xie #include <linux/seqlock.h> 757cc8e58dSMiao Xie #define __BTRFS_NEED_DEVICE_DATA_ORDERED 76c41ec452SSu Yue #define btrfs_device_data_ordered_init(device) \ 77c41ec452SSu Yue seqcount_init(&device->data_seqcount) 787cc8e58dSMiao Xie #else 79c41ec452SSu Yue #define btrfs_device_data_ordered_init(device) do { } while (0) 807cc8e58dSMiao Xie #endif 817cc8e58dSMiao Xie 82ebbede42SAnand Jain #define BTRFS_DEV_STATE_WRITEABLE (0) 83e12c9621SAnand Jain #define BTRFS_DEV_STATE_IN_FS_METADATA (1) 84e6e674bdSAnand Jain #define BTRFS_DEV_STATE_MISSING (2) 85401e29c1SAnand Jain #define BTRFS_DEV_STATE_REPLACE_TGT (3) 861c3063b6SAnand Jain #define BTRFS_DEV_STATE_FLUSH_SENT (4) 8766d204a1SFilipe Manana #define BTRFS_DEV_STATE_NO_READA (5) 88ebbede42SAnand Jain 895b316468SNaohiro Aota struct btrfs_zoned_device_info; 905b316468SNaohiro Aota 910b86a832SChris Mason struct btrfs_device { 920b6f5d40SNikolay Borisov struct list_head dev_list; /* device_list_mutex */ 930b6f5d40SNikolay Borisov struct list_head dev_alloc_list; /* chunk mutex */ 94bbbf7243SNikolay Borisov struct list_head post_commit_list; /* chunk mutex */ 952b82032cSYan Zheng struct btrfs_fs_devices *fs_devices; 96fb456252SJeff Mahoney struct btrfs_fs_info *fs_info; 97ffbd517dSChris Mason 988d1a7aaeSMadhuparna Bhowmik struct rcu_string __rcu *name; 99d5ee37bcSMiao Xie 100d5ee37bcSMiao Xie u64 generation; 101d5ee37bcSMiao Xie 102d5ee37bcSMiao Xie struct block_device *bdev; 103d5ee37bcSMiao Xie 1045b316468SNaohiro Aota struct btrfs_zoned_device_info *zone_info; 1055b316468SNaohiro Aota 106d5ee37bcSMiao Xie /* the mode sent to blkdev_get */ 107d5ee37bcSMiao Xie fmode_t mode; 108d5ee37bcSMiao Xie 1094889bc05SAnand Jain /* 1104889bc05SAnand Jain * Device's major-minor number. Must be set even if the device is not 1114889bc05SAnand Jain * opened (bdev == NULL), unless the device is missing. 1124889bc05SAnand Jain */ 1134889bc05SAnand Jain dev_t devt; 114ebbede42SAnand Jain unsigned long dev_state; 11558efbc9fSOmar Sandoval blk_status_t last_flush_error; 116b3075717SChris Mason 1177cc8e58dSMiao Xie #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 118c41ec452SSu Yue seqcount_t data_seqcount; 1197cc8e58dSMiao Xie #endif 1207cc8e58dSMiao Xie 1210b86a832SChris Mason /* the internal btrfs device id */ 1220b86a832SChris Mason u64 devid; 1230b86a832SChris Mason 1246ba40b61SMiao Xie /* size of the device in memory */ 1250b86a832SChris Mason u64 total_bytes; 1260b86a832SChris Mason 1276ba40b61SMiao Xie /* size of the device on disk */ 128d6397baeSChris Ball u64 disk_total_bytes; 129d6397baeSChris Ball 1300b86a832SChris Mason /* bytes used */ 1310b86a832SChris Mason u64 bytes_used; 1320b86a832SChris Mason 1330b86a832SChris Mason /* optimal io alignment for this device */ 1340b86a832SChris Mason u32 io_align; 1350b86a832SChris Mason 1360b86a832SChris Mason /* optimal io width for this device */ 1370b86a832SChris Mason u32 io_width; 1383c45bfc1SDulshani Gunawardhana /* type and info about this device */ 1393c45bfc1SDulshani Gunawardhana u64 type; 1400b86a832SChris Mason 1410b86a832SChris Mason /* minimal io size for this device */ 1420b86a832SChris Mason u32 sector_size; 1430b86a832SChris Mason 1440b86a832SChris Mason /* physical drive uuid (or lvm uuid) */ 145e17cade2SChris Mason u8 uuid[BTRFS_UUID_SIZE]; 1468b712842SChris Mason 147935e5cc9SMiao Xie /* 148935e5cc9SMiao Xie * size of the device on the current transaction 149935e5cc9SMiao Xie * 150935e5cc9SMiao Xie * This variant is update when committing the transaction, 151bbbf7243SNikolay Borisov * and protected by chunk mutex 152935e5cc9SMiao Xie */ 153935e5cc9SMiao Xie u64 commit_total_bytes; 154935e5cc9SMiao Xie 155ce7213c7SMiao Xie /* bytes used on the current transaction */ 156ce7213c7SMiao Xie u64 commit_bytes_used; 157935e5cc9SMiao Xie 158f9e69aa9SChristoph Hellwig /* Bio used for flushing device barriers */ 159f9e69aa9SChristoph Hellwig struct bio flush_bio; 1603c45bfc1SDulshani Gunawardhana struct completion flush_wait; 1613c45bfc1SDulshani Gunawardhana 162a2de733cSArne Jansen /* per-device scrub information */ 163cadbc0a0SAnand Jain struct scrub_ctx *scrub_ctx; 164a2de733cSArne Jansen 165442a4f63SStefan Behrens /* disk I/O failure stats. For detailed description refer to 166442a4f63SStefan Behrens * enum btrfs_dev_stat_values in ioctl.h */ 167733f4fbbSStefan Behrens int dev_stats_valid; 168addc3fa7SMiao Xie 169addc3fa7SMiao Xie /* Counter to record the change of device stats */ 170addc3fa7SMiao Xie atomic_t dev_stats_ccnt; 171442a4f63SStefan Behrens atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 1721c11b63eSJeff Mahoney 1731c11b63eSJeff Mahoney struct extent_io_tree alloc_state; 174668e48afSAnand Jain 175668e48afSAnand Jain struct completion kobj_unregister; 176668e48afSAnand Jain /* For sysfs/FSID/devinfo/devid/ */ 177668e48afSAnand Jain struct kobject devid_kobj; 178eb3b5053SDavid Sterba 179eb3b5053SDavid Sterba /* Bandwidth limit for scrub, in bytes */ 180eb3b5053SDavid Sterba u64 scrub_speed_max; 1810b86a832SChris Mason }; 1820b86a832SChris Mason 1837cc8e58dSMiao Xie /* 1847cc8e58dSMiao Xie * If we read those variants at the context of their own lock, we needn't 1857cc8e58dSMiao Xie * use the following helpers, reading them directly is safe. 1867cc8e58dSMiao Xie */ 1877cc8e58dSMiao Xie #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 1887cc8e58dSMiao Xie #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 1897cc8e58dSMiao Xie static inline u64 \ 1907cc8e58dSMiao Xie btrfs_device_get_##name(const struct btrfs_device *dev) \ 1917cc8e58dSMiao Xie { \ 1927cc8e58dSMiao Xie u64 size; \ 1937cc8e58dSMiao Xie unsigned int seq; \ 1947cc8e58dSMiao Xie \ 1957cc8e58dSMiao Xie do { \ 1967cc8e58dSMiao Xie seq = read_seqcount_begin(&dev->data_seqcount); \ 1977cc8e58dSMiao Xie size = dev->name; \ 1987cc8e58dSMiao Xie } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 1997cc8e58dSMiao Xie return size; \ 2007cc8e58dSMiao Xie } \ 2017cc8e58dSMiao Xie \ 2027cc8e58dSMiao Xie static inline void \ 2037cc8e58dSMiao Xie btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 2047cc8e58dSMiao Xie { \ 205c41ec452SSu Yue preempt_disable(); \ 2067cc8e58dSMiao Xie write_seqcount_begin(&dev->data_seqcount); \ 2077cc8e58dSMiao Xie dev->name = size; \ 2087cc8e58dSMiao Xie write_seqcount_end(&dev->data_seqcount); \ 209c41ec452SSu Yue preempt_enable(); \ 2107cc8e58dSMiao Xie } 21194545870SThomas Gleixner #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 2127cc8e58dSMiao Xie #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 2137cc8e58dSMiao Xie static inline u64 \ 2147cc8e58dSMiao Xie btrfs_device_get_##name(const struct btrfs_device *dev) \ 2157cc8e58dSMiao Xie { \ 2167cc8e58dSMiao Xie u64 size; \ 2177cc8e58dSMiao Xie \ 2187cc8e58dSMiao Xie preempt_disable(); \ 2197cc8e58dSMiao Xie size = dev->name; \ 2207cc8e58dSMiao Xie preempt_enable(); \ 2217cc8e58dSMiao Xie return size; \ 2227cc8e58dSMiao Xie } \ 2237cc8e58dSMiao Xie \ 2247cc8e58dSMiao Xie static inline void \ 2257cc8e58dSMiao Xie btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 2267cc8e58dSMiao Xie { \ 2277cc8e58dSMiao Xie preempt_disable(); \ 2287cc8e58dSMiao Xie dev->name = size; \ 2297cc8e58dSMiao Xie preempt_enable(); \ 2307cc8e58dSMiao Xie } 2317cc8e58dSMiao Xie #else 2327cc8e58dSMiao Xie #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 2337cc8e58dSMiao Xie static inline u64 \ 2347cc8e58dSMiao Xie btrfs_device_get_##name(const struct btrfs_device *dev) \ 2357cc8e58dSMiao Xie { \ 2367cc8e58dSMiao Xie return dev->name; \ 2377cc8e58dSMiao Xie } \ 2387cc8e58dSMiao Xie \ 2397cc8e58dSMiao Xie static inline void \ 2407cc8e58dSMiao Xie btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 2417cc8e58dSMiao Xie { \ 2427cc8e58dSMiao Xie dev->name = size; \ 2437cc8e58dSMiao Xie } 2447cc8e58dSMiao Xie #endif 2457cc8e58dSMiao Xie 2467cc8e58dSMiao Xie BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 2477cc8e58dSMiao Xie BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 2487cc8e58dSMiao Xie BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 2497cc8e58dSMiao Xie 250c4a816c6SNaohiro Aota enum btrfs_chunk_allocation_policy { 251c4a816c6SNaohiro Aota BTRFS_CHUNK_ALLOC_REGULAR, 2521cd6121fSNaohiro Aota BTRFS_CHUNK_ALLOC_ZONED, 253c4a816c6SNaohiro Aota }; 254c4a816c6SNaohiro Aota 25533fd2f71SAnand Jain /* 25633fd2f71SAnand Jain * Read policies for mirrored block group profiles, read picks the stripe based 25733fd2f71SAnand Jain * on these policies. 25833fd2f71SAnand Jain */ 25933fd2f71SAnand Jain enum btrfs_read_policy { 26033fd2f71SAnand Jain /* Use process PID to choose the stripe */ 26133fd2f71SAnand Jain BTRFS_READ_POLICY_PID, 26233fd2f71SAnand Jain BTRFS_NR_READ_POLICY, 26333fd2f71SAnand Jain }; 26433fd2f71SAnand Jain 2658a4b83ccSChris Mason struct btrfs_fs_devices { 2668a4b83ccSChris Mason u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 2677239ff4bSNikolay Borisov u8 metadata_uuid[BTRFS_FSID_SIZE]; 268d1a63002SNikolay Borisov bool fsid_change; 269c4babc5eSAnand Jain struct list_head fs_list; 2708a4b83ccSChris Mason 271add9745aSAnand Jain /* 272add9745aSAnand Jain * Number of devices under this fsid including missing and 273add9745aSAnand Jain * replace-target device and excludes seed devices. 274add9745aSAnand Jain */ 2758a4b83ccSChris Mason u64 num_devices; 276add9745aSAnand Jain 277add9745aSAnand Jain /* 278add9745aSAnand Jain * The number of devices that successfully opened, including 279add9745aSAnand Jain * replace-target, excludes seed devices. 280add9745aSAnand Jain */ 281a0af469bSChris Mason u64 open_devices; 282add9745aSAnand Jain 283add9745aSAnand Jain /* The number of devices that are under the chunk allocation list. */ 2842b82032cSYan Zheng u64 rw_devices; 285add9745aSAnand Jain 286add9745aSAnand Jain /* Count of missing devices under this fsid excluding seed device. */ 287cd02dca5SChris Mason u64 missing_devices; 2882b82032cSYan Zheng u64 total_rw_bytes; 289add9745aSAnand Jain 290add9745aSAnand Jain /* 291add9745aSAnand Jain * Count of devices from btrfs_super_block::num_devices for this fsid, 292add9745aSAnand Jain * which includes the seed device, excludes the transient replace-target 293add9745aSAnand Jain * device. 294add9745aSAnand Jain */ 29502db0844SJosef Bacik u64 total_devices; 296d1a63002SNikolay Borisov 297d1a63002SNikolay Borisov /* Highest generation number of seen devices */ 298d1a63002SNikolay Borisov u64 latest_generation; 299d1a63002SNikolay Borisov 300d24fa5c1SAnand Jain /* 301d24fa5c1SAnand Jain * The mount device or a device with highest generation after removal 302d24fa5c1SAnand Jain * or replace. 303d24fa5c1SAnand Jain */ 304d24fa5c1SAnand Jain struct btrfs_device *latest_dev; 305e5e9a520SChris Mason 306e5e9a520SChris Mason /* all of the devices in the FS, protected by a mutex 307e5e9a520SChris Mason * so we can safely walk it to write out the supers without 3089b011adfSWang Shilong * worrying about add/remove by the multi-device code. 3099b011adfSWang Shilong * Scrubbing super can kick off supers writing by holding 3109b011adfSWang Shilong * this mutex lock. 311e5e9a520SChris Mason */ 312e5e9a520SChris Mason struct mutex device_list_mutex; 3130b6f5d40SNikolay Borisov 3140b6f5d40SNikolay Borisov /* List of all devices, protected by device_list_mutex */ 3158a4b83ccSChris Mason struct list_head devices; 316b3075717SChris Mason 3170b6f5d40SNikolay Borisov /* 3180b6f5d40SNikolay Borisov * Devices which can satisfy space allocation. Protected by 3190b6f5d40SNikolay Borisov * chunk_mutex 3200b6f5d40SNikolay Borisov */ 321b3075717SChris Mason struct list_head alloc_list; 3222b82032cSYan Zheng 323944d3f9fSNikolay Borisov struct list_head seed_list; 3240395d84fSJohannes Thumshirn bool seeding; 3252b82032cSYan Zheng 3262b82032cSYan Zheng int opened; 327c289811cSChris Mason 328c289811cSChris Mason /* set when we find or add a device that doesn't have the 329c289811cSChris Mason * nonrot flag set 330c289811cSChris Mason */ 3317f0432d0SJohannes Thumshirn bool rotating; 3322e7910d6SAnand Jain 3335a13f430SAnand Jain struct btrfs_fs_info *fs_info; 3342e7910d6SAnand Jain /* sysfs kobjects */ 335c1b7e474SAnand Jain struct kobject fsid_kobj; 336b5501504SAnand Jain struct kobject *devices_kobj; 337a013d141SAnand Jain struct kobject *devinfo_kobj; 3382e7910d6SAnand Jain struct completion kobj_unregister; 339c4a816c6SNaohiro Aota 340c4a816c6SNaohiro Aota enum btrfs_chunk_allocation_policy chunk_alloc_policy; 34133fd2f71SAnand Jain 34233fd2f71SAnand Jain /* Policy used to read the mirrored stripes */ 34333fd2f71SAnand Jain enum btrfs_read_policy read_policy; 3448a4b83ccSChris Mason }; 3458a4b83ccSChris Mason 346facc8a22SMiao Xie #define BTRFS_BIO_INLINE_CSUM_SIZE 64 347facc8a22SMiao Xie 348ab4ba2e1SQu Wenruo #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \ 349ab4ba2e1SQu Wenruo - sizeof(struct btrfs_chunk)) \ 350ab4ba2e1SQu Wenruo / sizeof(struct btrfs_stripe) + 1) 351ab4ba2e1SQu Wenruo 352ab4ba2e1SQu Wenruo #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \ 353ab4ba2e1SQu Wenruo - 2 * sizeof(struct btrfs_disk_key) \ 354ab4ba2e1SQu Wenruo - 2 * sizeof(struct btrfs_chunk)) \ 355ab4ba2e1SQu Wenruo / sizeof(struct btrfs_stripe) + 1) 356ab4ba2e1SQu Wenruo 3579be3395bSChris Mason /* 358c3a3b19bSQu Wenruo * Additional info to pass along bio. 359c3a3b19bSQu Wenruo * 360c3a3b19bSQu Wenruo * Mostly for btrfs specific features like csum and mirror_num. 3619be3395bSChris Mason */ 362c3a3b19bSQu Wenruo struct btrfs_bio { 363c1dc0896SMiao Xie unsigned int mirror_num; 364c3a3b19bSQu Wenruo 36500d82525SChristoph Hellwig /* for direct I/O */ 36600d82525SChristoph Hellwig u64 file_offset; 36700d82525SChristoph Hellwig 368c3a3b19bSQu Wenruo /* @device is for stripe IO submission. */ 369c31efbdfSNikolay Borisov struct btrfs_device *device; 370facc8a22SMiao Xie u8 *csum; 371facc8a22SMiao Xie u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE]; 37217347cecSLiu Bo struct bvec_iter iter; 373c3a3b19bSQu Wenruo 374*d7b9416fSChristoph Hellwig /* For read end I/O handling */ 375*d7b9416fSChristoph Hellwig struct work_struct end_io_work; 376*d7b9416fSChristoph Hellwig 377fa1bcbe0SDavid Sterba /* 378fa1bcbe0SDavid Sterba * This member must come last, bio_alloc_bioset will allocate enough 379c3a3b19bSQu Wenruo * bytes for entire btrfs_bio but relies on bio being last. 380fa1bcbe0SDavid Sterba */ 3819be3395bSChris Mason struct bio bio; 3829be3395bSChris Mason }; 3839be3395bSChris Mason 384c3a3b19bSQu Wenruo static inline struct btrfs_bio *btrfs_bio(struct bio *bio) 3859be3395bSChris Mason { 386c3a3b19bSQu Wenruo return container_of(bio, struct btrfs_bio, bio); 3879be3395bSChris Mason } 3889be3395bSChris Mason 389c3a3b19bSQu Wenruo static inline void btrfs_bio_free_csum(struct btrfs_bio *bbio) 390b3a0dd50SDavid Sterba { 391c3a3b19bSQu Wenruo if (bbio->csum != bbio->csum_inline) { 392c3a3b19bSQu Wenruo kfree(bbio->csum); 393c3a3b19bSQu Wenruo bbio->csum = NULL; 394b3a0dd50SDavid Sterba } 395b3a0dd50SDavid Sterba } 396b3a0dd50SDavid Sterba 397261d812bSQu Wenruo /* 398261d812bSQu Wenruo * Iterate through a btrfs_bio (@bbio) on a per-sector basis. 399261d812bSQu Wenruo * 400261d812bSQu Wenruo * bvl - struct bio_vec 401261d812bSQu Wenruo * bbio - struct btrfs_bio 402261d812bSQu Wenruo * iters - struct bvec_iter 403261d812bSQu Wenruo * bio_offset - unsigned int 404261d812bSQu Wenruo */ 405261d812bSQu Wenruo #define btrfs_bio_for_each_sector(fs_info, bvl, bbio, iter, bio_offset) \ 406261d812bSQu Wenruo for ((iter) = (bbio)->iter, (bio_offset) = 0; \ 407261d812bSQu Wenruo (iter).bi_size && \ 408261d812bSQu Wenruo (((bvl) = bio_iter_iovec((&(bbio)->bio), (iter))), 1); \ 409261d812bSQu Wenruo (bio_offset) += fs_info->sectorsize, \ 410261d812bSQu Wenruo bio_advance_iter_single(&(bbio)->bio, &(iter), \ 411261d812bSQu Wenruo (fs_info)->sectorsize)) 412261d812bSQu Wenruo 4134c664611SQu Wenruo struct btrfs_io_stripe { 414cea9e445SChris Mason struct btrfs_device *dev; 415cea9e445SChris Mason u64 physical; 416fce3bb9aSLi Dongyang u64 length; /* only used for discard mappings */ 417cea9e445SChris Mason }; 418cea9e445SChris Mason 4194c664611SQu Wenruo /* 4204c664611SQu Wenruo * Context for IO subsmission for device stripe. 4214c664611SQu Wenruo * 4224c664611SQu Wenruo * - Track the unfinished mirrors for mirror based profiles 4234c664611SQu Wenruo * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. 4244c664611SQu Wenruo * 4254c664611SQu Wenruo * - Contain the logical -> physical mapping info 4264c664611SQu Wenruo * Used by submit_stripe_bio() for mapping logical bio 4274c664611SQu Wenruo * into physical device address. 4284c664611SQu Wenruo * 4294c664611SQu Wenruo * - Contain device replace info 4304c664611SQu Wenruo * Used by handle_ops_on_dev_replace() to copy logical bios 4314c664611SQu Wenruo * into the new device. 4324c664611SQu Wenruo * 4334c664611SQu Wenruo * - Contain RAID56 full stripe logical bytenrs 4344c664611SQu Wenruo */ 4354c664611SQu Wenruo struct btrfs_io_context { 436140475aeSElena Reshetova refcount_t refs; 437cea9e445SChris Mason atomic_t stripes_pending; 438c404e0dcSMiao Xie struct btrfs_fs_info *fs_info; 43910f11900SZhao Lei u64 map_type; /* get from map_lookup->type */ 440cea9e445SChris Mason bio_end_io_t *end_io; 4417d2b4daaSChris Mason struct bio *orig_bio; 442cea9e445SChris Mason void *private; 443a236aed1SChris Mason atomic_t error; 444a236aed1SChris Mason int max_errors; 445cea9e445SChris Mason int num_stripes; 446a1d3c478SJan Schmidt int mirror_num; 4472c8cdd6eSMiao Xie int num_tgtdevs; 4482c8cdd6eSMiao Xie int *tgtdev_map; 4498e5cfb55SZhao Lei /* 4508e5cfb55SZhao Lei * logical block numbers for the start of each stripe 4518e5cfb55SZhao Lei * The last one or two are p/q. These are sorted, 4528e5cfb55SZhao Lei * so raid_map[0] is the start of our full stripe 4538e5cfb55SZhao Lei */ 4548e5cfb55SZhao Lei u64 *raid_map; 4554c664611SQu Wenruo struct btrfs_io_stripe stripes[]; 456cea9e445SChris Mason }; 457cea9e445SChris Mason 458b2117a39SMiao Xie struct btrfs_device_info { 459b2117a39SMiao Xie struct btrfs_device *dev; 460b2117a39SMiao Xie u64 dev_offset; 461b2117a39SMiao Xie u64 max_avail; 46273c5de00SArne Jansen u64 total_avail; 463b2117a39SMiao Xie }; 464b2117a39SMiao Xie 46531e50229SLiu Bo struct btrfs_raid_attr { 4668c3e3582SDavid Sterba u8 sub_stripes; /* sub_stripes info for map */ 4678c3e3582SDavid Sterba u8 dev_stripes; /* stripes per dev */ 4688c3e3582SDavid Sterba u8 devs_max; /* max devs to use */ 4698c3e3582SDavid Sterba u8 devs_min; /* min devs needed */ 4708c3e3582SDavid Sterba u8 tolerated_failures; /* max tolerated fail devs */ 4718c3e3582SDavid Sterba u8 devs_increment; /* ndevs has to be a multiple of this */ 4728c3e3582SDavid Sterba u8 ncopies; /* how many copies to data has */ 4738c3e3582SDavid Sterba u8 nparity; /* number of stripes worth of bytes to store 474b50836edSHans van Kranenburg * parity information */ 4758c3e3582SDavid Sterba u8 mindev_error; /* error code if min devs requisite is unmet */ 476ed23467bSAnand Jain const char raid_name[8]; /* name of the raid */ 47741a6e891SAnand Jain u64 bg_flag; /* block group flag of the raid */ 47831e50229SLiu Bo }; 47931e50229SLiu Bo 480af902047SZhao Lei extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 481af902047SZhao Lei 4821abe9b8aSliubo struct map_lookup { 4831abe9b8aSliubo u64 type; 4841abe9b8aSliubo int io_align; 4851abe9b8aSliubo int io_width; 486cc353a8bSQu Wenruo u32 stripe_len; 4871abe9b8aSliubo int num_stripes; 4881abe9b8aSliubo int sub_stripes; 489cf90d884SQu Wenruo int verified_stripes; /* For mount time dev extent verification */ 4904c664611SQu Wenruo struct btrfs_io_stripe stripes[]; 4911abe9b8aSliubo }; 4921abe9b8aSliubo 493a2de733cSArne Jansen #define map_lookup_size(n) (sizeof(struct map_lookup) + \ 4944c664611SQu Wenruo (sizeof(struct btrfs_io_stripe) * (n))) 495a2de733cSArne Jansen 496c9e9f97bSIlya Dryomov struct btrfs_balance_args; 49719a39dceSIlya Dryomov struct btrfs_balance_progress; 498c9e9f97bSIlya Dryomov struct btrfs_balance_control { 499c9e9f97bSIlya Dryomov struct btrfs_balance_args data; 500c9e9f97bSIlya Dryomov struct btrfs_balance_args meta; 501c9e9f97bSIlya Dryomov struct btrfs_balance_args sys; 502c9e9f97bSIlya Dryomov 503c9e9f97bSIlya Dryomov u64 flags; 50419a39dceSIlya Dryomov 50519a39dceSIlya Dryomov struct btrfs_balance_progress stat; 506c9e9f97bSIlya Dryomov }; 507c9e9f97bSIlya Dryomov 508562d7b15SJosef Bacik /* 509562d7b15SJosef Bacik * Search for a given device by the set parameters 510562d7b15SJosef Bacik */ 511562d7b15SJosef Bacik struct btrfs_dev_lookup_args { 512562d7b15SJosef Bacik u64 devid; 513562d7b15SJosef Bacik u8 *uuid; 514562d7b15SJosef Bacik u8 *fsid; 515562d7b15SJosef Bacik bool missing; 516562d7b15SJosef Bacik }; 517562d7b15SJosef Bacik 518562d7b15SJosef Bacik /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */ 519562d7b15SJosef Bacik #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 } 520562d7b15SJosef Bacik 521562d7b15SJosef Bacik #define BTRFS_DEV_LOOKUP_ARGS(name) \ 522562d7b15SJosef Bacik struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT 523562d7b15SJosef Bacik 524cf8cddd3SChristoph Hellwig enum btrfs_map_op { 525cf8cddd3SChristoph Hellwig BTRFS_MAP_READ, 526cf8cddd3SChristoph Hellwig BTRFS_MAP_WRITE, 527cf8cddd3SChristoph Hellwig BTRFS_MAP_DISCARD, 528cf8cddd3SChristoph Hellwig BTRFS_MAP_GET_READ_MIRRORS, 529cf8cddd3SChristoph Hellwig }; 530cf8cddd3SChristoph Hellwig 531cf8cddd3SChristoph Hellwig static inline enum btrfs_map_op btrfs_op(struct bio *bio) 532cf8cddd3SChristoph Hellwig { 533cf8cddd3SChristoph Hellwig switch (bio_op(bio)) { 534cf8cddd3SChristoph Hellwig case REQ_OP_DISCARD: 535cf8cddd3SChristoph Hellwig return BTRFS_MAP_DISCARD; 536cf8cddd3SChristoph Hellwig case REQ_OP_WRITE: 537cfe94440SNaohiro Aota case REQ_OP_ZONE_APPEND: 538cf8cddd3SChristoph Hellwig return BTRFS_MAP_WRITE; 539cf8cddd3SChristoph Hellwig default: 540cf8cddd3SChristoph Hellwig WARN_ON_ONCE(1); 541c730ae0cSMarcos Paulo de Souza fallthrough; 542cf8cddd3SChristoph Hellwig case REQ_OP_READ: 543cf8cddd3SChristoph Hellwig return BTRFS_MAP_READ; 544cf8cddd3SChristoph Hellwig } 545cf8cddd3SChristoph Hellwig } 546cf8cddd3SChristoph Hellwig 5474c664611SQu Wenruo void btrfs_get_bioc(struct btrfs_io_context *bioc); 5484c664611SQu Wenruo void btrfs_put_bioc(struct btrfs_io_context *bioc); 549cf8cddd3SChristoph Hellwig int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 550cea9e445SChris Mason u64 logical, u64 *length, 5514c664611SQu Wenruo struct btrfs_io_context **bioc_ret, int mirror_num); 552cf8cddd3SChristoph Hellwig int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 553af8e2d1dSMiao Xie u64 logical, u64 *length, 5544c664611SQu Wenruo struct btrfs_io_context **bioc_ret); 55542034313SMichal Rostecki int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, struct extent_map *map, 55643c0d1a5SQu Wenruo enum btrfs_map_op op, u64 logical, 55742034313SMichal Rostecki struct btrfs_io_geometry *io_geom); 5586bccf3abSJeff Mahoney int btrfs_read_sys_array(struct btrfs_fs_info *fs_info); 5595b4aacefSJeff Mahoney int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info); 560f6f39f7aSNikolay Borisov struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans, 56179bd3712SFilipe Manana u64 type); 562c8bf1b67SDavid Sterba void btrfs_mapping_tree_free(struct extent_map_tree *tree); 56358efbc9fSOmar Sandoval blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 56408635baeSChris Mason int mirror_num); 5658a4b83ccSChris Mason int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 56697288f2cSChristoph Hellwig fmode_t flags, void *holder); 56736350e95SGu Jinxiang struct btrfs_device *btrfs_scan_one_device(const char *path, 56836350e95SGu Jinxiang fmode_t flags, void *holder); 56916cab91aSAnand Jain int btrfs_forget_devices(dev_t devt); 57054eed6aeSNikolay Borisov void btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 571bacce86aSAnand Jain void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices); 572d6507cf1SNikolay Borisov void btrfs_assign_next_active_device(struct btrfs_device *device, 573d6507cf1SNikolay Borisov struct btrfs_device *this_dev); 574a27a94c2SNikolay Borisov struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, 575a27a94c2SNikolay Borisov u64 devid, 576a27a94c2SNikolay Borisov const char *devpath); 577faa775c4SJosef Bacik int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info, 578faa775c4SJosef Bacik struct btrfs_dev_lookup_args *args, 579faa775c4SJosef Bacik const char *path); 58012bd2fc0SIlya Dryomov struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 58112bd2fc0SIlya Dryomov const u64 *devid, 58212bd2fc0SIlya Dryomov const u8 *uuid); 583faa775c4SJosef Bacik void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args); 584a425f9d4SDavid Sterba void btrfs_free_device(struct btrfs_device *device); 5852ff7e61eSJeff Mahoney int btrfs_rm_device(struct btrfs_fs_info *fs_info, 5861a15eb72SJosef Bacik struct btrfs_dev_lookup_args *args, 5873fa421deSJosef Bacik struct block_device **bdev, fmode_t *mode); 588ffc5a379SDavid Sterba void __exit btrfs_cleanup_fs_uuids(void); 5895d964051SStefan Behrens int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 5908f18cf13SChris Mason int btrfs_grow_device(struct btrfs_trans_handle *trans, 5918f18cf13SChris Mason struct btrfs_device *device, u64 new_size); 592562d7b15SJosef Bacik struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices, 593562d7b15SJosef Bacik const struct btrfs_dev_lookup_args *args); 5948f18cf13SChris Mason int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 595da353f6bSDavid Sterba int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path); 5966fcf6e2bSDavid Sterba int btrfs_balance(struct btrfs_fs_info *fs_info, 5976fcf6e2bSDavid Sterba struct btrfs_balance_control *bctl, 598c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs); 599f89e09cfSAnand Jain void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf); 6002b6ba629SIlya Dryomov int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 60168310a5eSIlya Dryomov int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 602837d5b6eSIlya Dryomov int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 60318bb8bbfSJohannes Thumshirn int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset); 604a7e99c69SIlya Dryomov int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 605f7a81ea4SStefan Behrens int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info); 60697f4dd09SNikolay Borisov int btrfs_uuid_scan_kthread(void *data); 607a09f23c3SAnand Jain bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset); 60860dfdf25SNikolay Borisov int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes, 609ba1bf481SJosef Bacik u64 *start, u64 *max_avail); 610442a4f63SStefan Behrens void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 6112ff7e61eSJeff Mahoney int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, 612b27f7c0cSDavid Sterba struct btrfs_ioctl_get_dev_stats *stats); 613cb517eabSMiao Xie void btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 614733f4fbbSStefan Behrens int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 615196c9d8dSDavid Sterba int btrfs_run_dev_stats(struct btrfs_trans_handle *trans); 61668a9db5fSNikolay Borisov void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev); 61765237ee3SDavid Sterba void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev); 6184f5ad7bdSNikolay Borisov void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev); 619592d92eeSLiu Bo int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info, 620e4ff5fb5SNikolay Borisov u64 logical, u64 len); 6212ff7e61eSJeff Mahoney unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info, 62253b381b3SDavid Woodhouse u64 logical); 62379bd3712SFilipe Manana int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans, 62479bd3712SFilipe Manana struct btrfs_block_group *bg); 62597aff912SNikolay Borisov int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset); 62660ca842eSOmar Sandoval struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info, 62760ca842eSOmar Sandoval u64 logical, u64 length); 6288f32380dSJohannes Thumshirn void btrfs_release_disk_super(struct btrfs_super_block *super); 629addc3fa7SMiao Xie 630442a4f63SStefan Behrens static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 631442a4f63SStefan Behrens int index) 632442a4f63SStefan Behrens { 633442a4f63SStefan Behrens atomic_inc(dev->dev_stat_values + index); 6349deae968SNikolay Borisov /* 6359deae968SNikolay Borisov * This memory barrier orders stores updating statistics before stores 6369deae968SNikolay Borisov * updating dev_stats_ccnt. 6379deae968SNikolay Borisov * 6389deae968SNikolay Borisov * It pairs with smp_rmb() in btrfs_run_dev_stats(). 6399deae968SNikolay Borisov */ 640addc3fa7SMiao Xie smp_mb__before_atomic(); 641addc3fa7SMiao Xie atomic_inc(&dev->dev_stats_ccnt); 642442a4f63SStefan Behrens } 643442a4f63SStefan Behrens 644442a4f63SStefan Behrens static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 645442a4f63SStefan Behrens int index) 646442a4f63SStefan Behrens { 647442a4f63SStefan Behrens return atomic_read(dev->dev_stat_values + index); 648442a4f63SStefan Behrens } 649442a4f63SStefan Behrens 650442a4f63SStefan Behrens static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 651442a4f63SStefan Behrens int index) 652442a4f63SStefan Behrens { 653442a4f63SStefan Behrens int ret; 654442a4f63SStefan Behrens 655442a4f63SStefan Behrens ret = atomic_xchg(dev->dev_stat_values + index, 0); 6564660c49fSNikolay Borisov /* 6574660c49fSNikolay Borisov * atomic_xchg implies a full memory barriers as per atomic_t.txt: 6584660c49fSNikolay Borisov * - RMW operations that have a return value are fully ordered; 6594660c49fSNikolay Borisov * 6604660c49fSNikolay Borisov * This implicit memory barriers is paired with the smp_rmb in 6614660c49fSNikolay Borisov * btrfs_run_dev_stats 6624660c49fSNikolay Borisov */ 663addc3fa7SMiao Xie atomic_inc(&dev->dev_stats_ccnt); 664442a4f63SStefan Behrens return ret; 665442a4f63SStefan Behrens } 666442a4f63SStefan Behrens 667442a4f63SStefan Behrens static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 668442a4f63SStefan Behrens int index, unsigned long val) 669442a4f63SStefan Behrens { 670442a4f63SStefan Behrens atomic_set(dev->dev_stat_values + index, val); 6719deae968SNikolay Borisov /* 6729deae968SNikolay Borisov * This memory barrier orders stores updating statistics before stores 6739deae968SNikolay Borisov * updating dev_stats_ccnt. 6749deae968SNikolay Borisov * 6759deae968SNikolay Borisov * It pairs with smp_rmb() in btrfs_run_dev_stats(). 6769deae968SNikolay Borisov */ 677addc3fa7SMiao Xie smp_mb__before_atomic(); 678addc3fa7SMiao Xie atomic_inc(&dev->dev_stats_ccnt); 679442a4f63SStefan Behrens } 680442a4f63SStefan Behrens 681bbbf7243SNikolay Borisov void btrfs_commit_device_sizes(struct btrfs_transaction *trans); 68204216820SFilipe Manana 6834143cb8bSDavid Sterba struct list_head * __attribute_const__ btrfs_get_fs_uuids(void); 6846528b99dSAnand Jain bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info, 6856528b99dSAnand Jain struct btrfs_device *failing_dev); 686313b0858SJosef Bacik void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, 687313b0858SJosef Bacik struct block_device *bdev, 688313b0858SJosef Bacik const char *device_path); 68921634a19SQu Wenruo 690500a44c9SDavid Sterba enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags); 69146df06b8SDavid Sterba int btrfs_bg_type_to_factor(u64 flags); 692158da513SDavid Sterba const char *btrfs_bg_type_to_raid_name(u64 flags); 693cf90d884SQu Wenruo int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info); 694554aed7dSJohannes Thumshirn bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical); 69546df06b8SDavid Sterba 6960b86a832SChris Mason #endif 697