#undef pr_fmt
#define pr_fmt(fmt) "bcachefs: %s() " fmt "\n", __func__
+#include <linux/backing-dev-defs.h>
#include <linux/bug.h>
#include <linux/bio.h>
#include <linux/closure.h>
#include <linux/percpu-rwsem.h>
#include <linux/rhashtable.h>
#include <linux/rwsem.h>
+#include <linux/semaphore.h>
#include <linux/seqlock.h>
#include <linux/shrinker.h>
+#include <linux/srcu.h>
#include <linux/types.h>
#include <linux/workqueue.h>
#include <linux/zstd.h>
#include "bcachefs_format.h"
+#include "errcode.h"
#include "fifo.h"
#include "opts.h"
#include "util.h"
-#include <linux/dynamic_fault.h>
+#define dynamic_fault(...) 0
+#define race_fault(...) 0
#define bch2_fs_init_fault(name) \
dynamic_fault("bcachefs:bch_fs_init:" name)
dynamic_fault("bcachefs:meta:write:" name)
#ifdef __KERNEL__
-#define bch2_fmt(_c, fmt) "bcachefs (%s): " fmt "\n", ((_c)->name)
+#define bch2_fmt(_c, fmt) "bcachefs (%s): " fmt "\n", ((_c)->name)
+#define bch2_fmt_inum(_c, _inum, fmt) "bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
#else
-#define bch2_fmt(_c, fmt) fmt "\n"
+#define bch2_fmt(_c, fmt) "%s: " fmt "\n", ((_c)->name)
+#define bch2_fmt_inum(_c, _inum, fmt) "%s inum %llu: " fmt "\n", ((_c)->name), (_inum)
#endif
#define bch_info(c, fmt, ...) \
printk_ratelimited(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
#define bch_err(c, fmt, ...) \
printk(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
+
#define bch_err_ratelimited(c, fmt, ...) \
printk_ratelimited(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
+#define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
+ printk_ratelimited(KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
#define bch_verbose(c, fmt, ...) \
do { \
BCH_DEBUG_PARAM(btree_gc_rewrite_disabled, \
"Disables rewriting of btree nodes during mark and sweep")\
BCH_DEBUG_PARAM(btree_shrinker_disabled, \
- "Disables the shrinker callback for the btree node cache")
+ "Disables the shrinker callback for the btree node cache")\
+ BCH_DEBUG_PARAM(verify_btree_ondisk, \
+ "Reread btree nodes at various points to verify the " \
+ "mergesort in the read path against modifications " \
+ "done in memory") \
+ BCH_DEBUG_PARAM(verify_all_btree_replicas, \
+ "When reading btree nodes, read all replicas and " \
+ "compare them")
/* Parameters that should only be compiled in in debug mode: */
#define BCH_DEBUG_PARAMS_DEBUG() \
BCH_DEBUG_PARAM(debug_check_bkeys, \
"Run bkey_debugcheck (primarily checking GC/allocation "\
"information) when iterating over keys") \
- BCH_DEBUG_PARAM(verify_btree_ondisk, \
- "Reread btree nodes at various points to verify the " \
- "mergesort in the read path against modifications " \
- "done in memory") \
+ BCH_DEBUG_PARAM(debug_check_btree_accounting, \
+ "Verify btree accounting for keys within a node") \
BCH_DEBUG_PARAM(journal_seq_verify, \
"Store the journal sequence number in the version " \
"number of every btree key, and verify that btree " \
"Force reads to use the reconstruct path, when reading" \
"from erasure coded extents") \
BCH_DEBUG_PARAM(test_restart_gc, \
- "Test restarting mark and sweep gc when bucket gens change")\
- BCH_DEBUG_PARAM(test_reconstruct_alloc, \
- "Test reconstructing the alloc btree")
+ "Test restarting mark and sweep gc when bucket gens change")
#define BCH_DEBUG_PARAMS_ALL() BCH_DEBUG_PARAMS_ALWAYS() BCH_DEBUG_PARAMS_DEBUG()
#define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALWAYS()
#endif
+#define BCH_DEBUG_PARAM(name, description) extern bool bch2_##name;
+BCH_DEBUG_PARAMS()
+#undef BCH_DEBUG_PARAM
+
+#ifndef CONFIG_BCACHEFS_DEBUG
+#define BCH_DEBUG_PARAM(name, description) static const bool bch2_##name;
+BCH_DEBUG_PARAMS_DEBUG()
+#undef BCH_DEBUG_PARAM
+#endif
+
#define BCH_TIME_STATS() \
x(btree_node_mem_alloc) \
x(btree_node_split) \
+ x(btree_node_compact) \
+ x(btree_node_merge) \
x(btree_node_sort) \
x(btree_node_read) \
+ x(btree_interior_update_foreground) \
+ x(btree_interior_update_total) \
x(btree_gc) \
- x(btree_update) \
x(btree_lock_contended_read) \
x(btree_lock_contended_intent) \
x(btree_lock_contended_write) \
x(data_write) \
x(data_read) \
x(data_promote) \
- x(journal_write) \
- x(journal_delay) \
+ x(journal_flush_write) \
+ x(journal_noflush_write) \
x(journal_flush_seq) \
x(blocked_journal) \
x(blocked_allocate) \
#include "quota_types.h"
#include "rebalance_types.h"
#include "replicas_types.h"
+#include "subvolume_types.h"
#include "super_types.h"
/* Number of nodes btree coalesce will try to coalesce at once */
#define BTREE_RESERVE_MAX (BTREE_MAX_DEPTH + (BTREE_MAX_DEPTH - 1))
/* Size of the freelist we allocate btree nodes from: */
-#define BTREE_NODE_RESERVE BTREE_RESERVE_MAX
+#define BTREE_NODE_RESERVE (BTREE_RESERVE_MAX * 4)
#define BTREE_NODE_OPEN_BUCKET_RESERVE (BTREE_RESERVE_MAX * BCH_REPLICAS_MAX)
GC_PHASE_START,
GC_PHASE_SB,
- GC_PHASE_BTREE_EC,
- GC_PHASE_BTREE_EXTENTS,
- GC_PHASE_BTREE_INODES,
- GC_PHASE_BTREE_DIRENTS,
- GC_PHASE_BTREE_XATTRS,
- GC_PHASE_BTREE_ALLOC,
- GC_PHASE_BTREE_QUOTAS,
- GC_PHASE_BTREE_REFLINK,
+ GC_PHASE_BTREE_stripes,
+ GC_PHASE_BTREE_extents,
+ GC_PHASE_BTREE_inodes,
+ GC_PHASE_BTREE_dirents,
+ GC_PHASE_BTREE_xattrs,
+ GC_PHASE_BTREE_alloc,
+ GC_PHASE_BTREE_quotas,
+ GC_PHASE_BTREE_reflink,
+ GC_PHASE_BTREE_subvolumes,
+ GC_PHASE_BTREE_snapshots,
GC_PHASE_PENDING_DELETE,
- GC_PHASE_ALLOC,
};
struct gc_pos {
unsigned level;
};
+struct reflink_gc {
+ u64 offset;
+ u32 size;
+ u32 refcount;
+};
+
+typedef GENRADIX(struct reflink_gc) reflink_gc_table;
+
struct io_count {
u64 sectors[2][BCH_DATA_NR];
};
* Or rcu_read_lock(), but only for ptr_stale():
*/
struct bucket_array __rcu *buckets[2];
+ struct bucket_gens *bucket_gens;
unsigned long *buckets_nouse;
- unsigned long *buckets_written;
struct rw_semaphore bucket_lock;
- struct bch_dev_usage __percpu *usage[2];
+ struct bch_dev_usage *usage_base;
+ struct bch_dev_usage __percpu *usage[JOURNAL_BUF_NR];
+ struct bch_dev_usage __percpu *usage_gc;
/* Allocator: */
+ u64 new_fs_bucket_idx;
struct task_struct __rcu *alloc_thread;
/*
*/
alloc_fifo free[RESERVE_NR];
alloc_fifo free_inc;
- spinlock_t freelist_lock;
+ unsigned nr_open_buckets;
- u8 open_buckets_partial[OPEN_BUCKETS_COUNT];
- unsigned open_buckets_partial_nr;
+ open_bucket_idx_t open_buckets_partial[OPEN_BUCKETS_COUNT];
+ open_bucket_idx_t open_buckets_partial_nr;
size_t fifo_last_bucket;
- /* last calculated minimum prio */
- u16 max_last_bucket_io[2];
-
size_t inc_gen_needs_gc;
size_t inc_gen_really_needs_gc;
- /*
- * XXX: this should be an enum for allocator state, so as to include
- * error state
- */
- enum {
- ALLOCATOR_STOPPED,
- ALLOCATOR_RUNNING,
- ALLOCATOR_BLOCKED,
- ALLOCATOR_BLOCKED_FULL,
- } allocator_state;
+ enum allocator_states allocator_state;
alloc_heap alloc_heap;
- /* Copying GC: */
- struct task_struct *copygc_thread;
- copygc_heap copygc_heap;
- struct bch_pd_controller copygc_pd;
- struct write_point copygc_write_point;
- u64 copygc_threshold;
-
atomic64_t rebalance_work;
struct journal_device journal;
+ u64 prev_journal_sector;
struct work_struct io_error_work;
enum {
/* startup: */
+ BCH_FS_INITIALIZED,
BCH_FS_ALLOC_READ_DONE,
- BCH_FS_ALLOCATOR_STARTED,
+ BCH_FS_ALLOC_CLEAN,
BCH_FS_ALLOCATOR_RUNNING,
+ BCH_FS_ALLOCATOR_STOPPING,
BCH_FS_INITIAL_GC_DONE,
+ BCH_FS_INITIAL_GC_UNFIXED,
+ BCH_FS_TOPOLOGY_REPAIR_DONE,
BCH_FS_FSCK_DONE,
BCH_FS_STARTED,
BCH_FS_RW,
+ BCH_FS_WAS_RW,
/* shutdown: */
BCH_FS_STOPPING,
/* errors: */
BCH_FS_ERROR,
+ BCH_FS_TOPOLOGY_ERROR,
BCH_FS_ERRORS_FIXED,
+ BCH_FS_ERRORS_NOT_FIXED,
/* misc: */
- BCH_FS_BDEV_MOUNTED,
- BCH_FS_FIXED_GENS,
+ BCH_FS_NEED_ANOTHER_GC,
+ BCH_FS_DELETED_NODES,
BCH_FS_REBUILD_REPLICAS,
BCH_FS_HOLD_BTREE_WRITES,
};
} entries[0];
};
+struct journal_keys {
+ struct journal_key {
+ enum btree_id btree_id:8;
+ unsigned level:8;
+ bool allocated;
+ bool overwritten;
+ struct bkey_i *k;
+ u32 journal_seq;
+ u32 journal_offset;
+ } *d;
+ size_t nr;
+ size_t size;
+ u64 journal_seq_base;
+};
+
+struct btree_path_buf {
+ struct btree_path *path;
+};
+
+#define REPLICAS_DELTA_LIST_MAX (1U << 16)
+
+struct snapshot_t {
+ u32 parent;
+ u32 children[2];
+ u32 subvol; /* Nonzero only if a subvolume points to this node: */
+ u32 equiv;
+};
+
+typedef struct {
+ u32 subvol;
+ u64 inum;
+} subvol_inum;
+
+#define BCACHEFS_ROOT_SUBVOL_INUM \
+ ((subvol_inum) { BCACHEFS_ROOT_SUBVOL, BCACHEFS_ROOT_INO })
+
struct bch_fs {
struct closure cl;
int minor;
struct device *chardev;
struct super_block *vfs_sb;
+ dev_t dev;
char name[40];
- /* ro/rw, add/remove devices: */
- struct mutex state_lock;
+ /* ro/rw, add/remove/resize devices: */
+ struct rw_semaphore state_lock;
/* Counts outstanding writes, for clean transition to read-only */
struct percpu_ref writes;
struct bch_replicas_cpu replicas;
struct bch_replicas_cpu replicas_gc;
struct mutex replicas_gc_lock;
+ mempool_t replicas_delta_pool;
+ struct journal_entry_res btree_root_journal_res;
struct journal_entry_res replicas_journal_res;
+ struct journal_entry_res clock_journal_res;
+ struct journal_entry_res dev_usage_journal_res;
struct bch_disk_groups_cpu __rcu *disk_groups;
uuid_le user_uuid;
u16 version;
- u16 encoded_extent_max;
+ u16 version_min;
u8 nr_devices;
u8 clean;
u64 time_base_lo;
u32 time_base_hi;
- u32 time_precision;
+ unsigned time_units_per_sec;
+ unsigned nsec_per_time_unit;
u64 features;
u64 compat;
} sb;
+
struct bch_sb_handle disk_sb;
unsigned short block_bits; /* ilog2(block_size) */
struct closure sb_write;
struct mutex sb_lock;
+ /* snapshot.c: */
+ GENRADIX(struct snapshot_t) snapshots;
+ struct bch_snapshot_table __rcu *snapshot_table;
+ struct mutex snapshot_table_lock;
+ struct work_struct snapshot_delete_work;
+ struct work_struct snapshot_wait_for_pagecache_and_delete_work;
+ struct snapshot_id_list snapshots_unlinked;
+ struct mutex snapshots_unlinked_lock;
+
/* BTREE CACHE */
struct bio_set btree_bio;
+ struct workqueue_struct *io_complete_wq;
struct btree_root btree_roots[BTREE_ID_NR];
- bool btree_roots_dirty;
struct mutex btree_root_lock;
struct btree_cache btree_cache;
- mempool_t btree_reserve_pool;
-
/*
* Cache of allocated btree nodes - if we allocate a btree node and
* don't use it, if we free it that space can't be reused until going
mempool_t btree_interior_update_pool;
struct list_head btree_interior_update_list;
+ struct list_head btree_interior_updates_unwritten;
struct mutex btree_interior_update_lock;
struct closure_waitlist btree_interior_update_wait;
- mempool_t btree_iters_pool;
+ struct workqueue_struct *btree_interior_update_worker;
+ struct work_struct btree_interior_update_work;
- struct workqueue_struct *wq;
+ /* btree_iter.c: */
+ struct mutex btree_trans_lock;
+ struct list_head btree_trans_list;
+ mempool_t btree_paths_pool;
+ mempool_t btree_trans_mem_pool;
+ struct btree_path_buf __percpu *btree_paths_bufs;
+
+ struct srcu_struct btree_trans_barrier;
+ bool btree_trans_barrier_initialized;
+
+ struct btree_key_cache btree_key_cache;
+
+ struct workqueue_struct *btree_update_wq;
+ struct workqueue_struct *btree_io_complete_wq;
/* copygc needs its own workqueue for index updates.. */
struct workqueue_struct *copygc_wq;
- struct workqueue_struct *journal_reclaim_wq;
/* ALLOCATION */
- struct delayed_work pd_controllers_update;
- unsigned pd_controllers_update_seconds;
-
struct bch_devs_mask rw_devs[BCH_DATA_NR];
u64 capacity; /* sectors */
unsigned bucket_size_max;
atomic64_t sectors_available;
+ struct mutex sectors_available_lock;
struct bch_fs_pcpu __percpu *pcpu;
seqcount_t usage_lock;
struct bch_fs_usage *usage_base;
- struct bch_fs_usage __percpu *usage[2];
+ struct bch_fs_usage __percpu *usage[JOURNAL_BUF_NR];
struct bch_fs_usage __percpu *usage_gc;
+ u64 __percpu *online_reserved;
/* single element mempool: */
struct mutex usage_scratch_lock;
- struct bch_fs_usage *usage_scratch;
-
- /*
- * When we invalidate buckets, we use both the priority and the amount
- * of good data to determine which buckets to reuse first - to weight
- * those together consistently we keep track of the smallest nonzero
- * priority of any bucket.
- */
- struct bucket_clock bucket_clock[2];
+ struct bch_fs_usage_online *usage_scratch;
struct io_clock io_clock[2];
/* JOURNAL SEQ BLACKLIST */
struct journal_seq_blacklist_table *
journal_seq_blacklist_table;
- struct work_struct journal_seq_blacklist_gc_work;
/* ALLOCATOR */
spinlock_t freelist_lock;
struct closure_waitlist freelist_wait;
u64 blocked_allocate;
u64 blocked_allocate_open_bucket;
- u8 open_buckets_freelist;
- u8 open_buckets_nr_free;
+
+ open_bucket_idx_t open_buckets_freelist;
+ open_bucket_idx_t open_buckets_nr_free;
struct closure_waitlist open_buckets_wait;
struct open_bucket open_buckets[OPEN_BUCKETS_COUNT];
+ open_bucket_idx_t open_buckets_hash[OPEN_BUCKETS_COUNT];
struct write_point btree_write_point;
struct write_point rebalance_write_point;
atomic_t kick_gc;
unsigned long gc_count;
+ enum btree_id gc_gens_btree;
+ struct bpos gc_gens_pos;
+
/*
* Tracks GC's progress - everything in the range [ZERO_KEY..gc_cur_pos]
* has been marked by GC.
*
- * gc_cur_phase is a superset of btree_ids (BTREE_ID_EXTENTS etc.)
+ * gc_cur_phase is a superset of btree_ids (BTREE_ID_extents etc.)
*
* Protected by gc_pos_lock. Only written to by GC thread, so GC thread
* can read without a lock.
struct rw_semaphore gc_lock;
/* IO PATH */
+ struct semaphore io_in_flight;
struct bio_set bio_read;
struct bio_set bio_read_split;
struct bio_set bio_write;
struct rhashtable promote_table;
mempool_t compression_bounce[2];
- mempool_t compress_workspace[BCH_COMPRESSION_NR];
+ mempool_t compress_workspace[BCH_COMPRESSION_TYPE_NR];
mempool_t decompress_workspace;
ZSTD_parameters zstd_params;
atomic64_t key_version;
+ mempool_t large_bkey_pool;
+
/* REBALANCE */
struct bch_fs_rebalance rebalance;
+ /* COPYGC */
+ struct task_struct *copygc_thread;
+ copygc_heap copygc_heap;
+ struct write_point copygc_write_point;
+ s64 copygc_wait;
+
+ /* DATA PROGRESS STATS */
+ struct list_head data_progress_list;
+ struct mutex data_progress_lock;
+
/* STRIPES: */
- GENRADIX(struct stripe) stripes[2];
- struct mutex ec_stripe_create_lock;
+ GENRADIX(struct stripe) stripes;
+ GENRADIX(struct gc_stripe) gc_stripes;
ec_stripes_heap ec_stripes_heap;
spinlock_t ec_stripes_heap_lock;
/* ERASURE CODING */
- struct list_head ec_new_stripe_list;
- struct mutex ec_new_stripe_lock;
+ struct list_head ec_stripe_head_list;
+ struct mutex ec_stripe_head_lock;
+
+ struct list_head ec_stripe_new_list;
+ struct mutex ec_stripe_new_lock;
+
+ struct work_struct ec_stripe_create_work;
u64 ec_stripe_hint;
struct bio_set ec_bioset;
/* REFLINK */
u64 reflink_hint;
+ reflink_gc_table reflink_gc_table;
+ size_t reflink_gc_nr;
/* VFS IO PATH - fs-io.c */
struct bio_set writepage_bioset;
struct bio_set dio_write_bioset;
struct bio_set dio_read_bioset;
- struct bio_list btree_write_error_list;
- struct work_struct btree_write_error_work;
+
+ atomic64_t btree_writes_nr;
+ atomic64_t btree_writes_sectors;
spinlock_t btree_write_error_lock;
/* ERRORS */
/* DEBUG JUNK */
struct dentry *debug;
struct btree_debug btree_debug[BTREE_ID_NR];
-#ifdef CONFIG_BCACHEFS_DEBUG
struct btree *verify_data;
struct btree_node *verify_ondisk;
struct mutex verify_lock;
-#endif
- u64 unused_inode_hint;
+ u64 *unused_inode_hints;
+ unsigned inode_shard_bits;
/*
* A btree node on disk could have too many bsets for an iterator to fit
mempool_t btree_bounce_pool;
struct journal journal;
+ struct list_head journal_entries;
+ struct journal_keys journal_keys;
+ struct list_head journal_iters;
u64 last_bucket_seq_cleanup;
unsigned copy_gc_enabled:1;
bool promote_whole_extents;
-#define BCH_DEBUG_PARAM(name, description) bool name;
- BCH_DEBUG_PARAMS_ALL()
-#undef BCH_DEBUG_PARAM
-
struct time_stats times[BCH_TIME_STAT_NR];
};
static inline unsigned block_bytes(const struct bch_fs *c)
{
- return c->opts.block_size << 9;
+ return c->opts.block_size;
+}
+
+static inline unsigned block_sectors(const struct bch_fs *c)
+{
+ return c->opts.block_size >> 9;
}
-static inline struct timespec64 bch2_time_to_timespec(struct bch_fs *c, u64 time)
+static inline size_t btree_sectors(const struct bch_fs *c)
{
- return ns_to_timespec64(time * c->sb.time_precision + c->sb.time_base_lo);
+ return c->opts.btree_node_size >> 9;
}
-static inline s64 timespec_to_bch2_time(struct bch_fs *c, struct timespec64 ts)
+static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
{
- s64 ns = timespec64_to_ns(&ts) - c->sb.time_base_lo;
+ struct timespec64 t;
+ s32 rem;
+
+ time += c->sb.time_base_lo;
- if (c->sb.time_precision == 1)
- return ns;
+ t.tv_sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
+ t.tv_nsec = rem * c->sb.nsec_per_time_unit;
+ return t;
+}
- return div_s64(ns, c->sb.time_precision);
+static inline s64 timespec_to_bch2_time(const struct bch_fs *c, struct timespec64 ts)
+{
+ return (ts.tv_sec * c->sb.time_units_per_sec +
+ (int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
}
-static inline s64 bch2_current_time(struct bch_fs *c)
+static inline s64 bch2_current_time(const struct bch_fs *c)
{
struct timespec64 now;