*
* BTREE NODES:
*
- * Our unit of allocation is a bucket, and we we can't arbitrarily allocate and
+ * Our unit of allocation is a bucket, and we can't arbitrarily allocate and
* free smaller than a bucket - so, that's how big our btree nodes are.
*
* (If buckets are really big we'll only use part of the bucket for a btree node
#include "bcachefs_format.h"
#include "errcode.h"
#include "fifo.h"
+#include "nocow_locking_types.h"
#include "opts.h"
+#include "seqmutex.h"
#include "util.h"
+#ifdef CONFIG_BCACHEFS_DEBUG
+#define BCH_WRITE_REF_DEBUG
+#endif
+
+#ifndef dynamic_fault
#define dynamic_fault(...) 0
-#define race_fault(...) 0
+#endif
+
+#define race_fault(...) dynamic_fault("bcachefs:race")
#define trace_and_count(_c, _name, ...) \
do { \
dynamic_fault("bcachefs:meta:write:" name)
#ifdef __KERNEL__
-#define bch2_log_msg(_c, fmt) "bcachefs (%s): " fmt, ((_c)->name)
-#define bch2_fmt(_c, fmt) bch2_log_msg(_c, fmt "\n")
-#define bch2_fmt_inum(_c, _inum, fmt) "bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
+#define BCACHEFS_LOG_PREFIX
+#endif
+
+#ifdef BCACHEFS_LOG_PREFIX
+
+#define bch2_log_msg(_c, fmt) "bcachefs (%s): " fmt, ((_c)->name)
+#define bch2_fmt_dev(_ca, fmt) "bcachefs (%s): " fmt "\n", ((_ca)->name)
+#define bch2_fmt_dev_offset(_ca, _offset, fmt) "bcachefs (%s sector %llu): " fmt "\n", ((_ca)->name), (_offset)
+#define bch2_fmt_inum(_c, _inum, fmt) "bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
+#define bch2_fmt_inum_offset(_c, _inum, _offset, fmt) \
+ "bcachefs (%s inum %llu offset %llu): " fmt "\n", ((_c)->name), (_inum), (_offset)
+
#else
-#define bch2_log_msg(_c, fmt) fmt
-#define bch2_fmt(_c, fmt) fmt "\n"
-#define bch2_fmt_inum(_c, _inum, fmt) "inum %llu: " fmt "\n", (_inum)
+
+#define bch2_log_msg(_c, fmt) fmt
+#define bch2_fmt_dev(_ca, fmt) "%s: " fmt "\n", ((_ca)->name)
+#define bch2_fmt_dev_offset(_ca, _offset, fmt) "%s sector %llu: " fmt "\n", ((_ca)->name), (_offset)
+#define bch2_fmt_inum(_c, _inum, fmt) "inum %llu: " fmt "\n", (_inum)
+#define bch2_fmt_inum_offset(_c, _inum, _offset, fmt) \
+ "inum %llu offset %llu: " fmt "\n", (_inum), (_offset)
+
#endif
+#define bch2_fmt(_c, fmt) bch2_log_msg(_c, fmt "\n")
+
#define bch_info(c, fmt, ...) \
printk(KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
#define bch_notice(c, fmt, ...) \
printk(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
#define bch_warn_ratelimited(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_dev(ca, fmt, ...) \
+ printk(KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
+#define bch_err_dev_offset(ca, _offset, fmt, ...) \
+ printk(KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
+#define bch_err_inum(c, _inum, fmt, ...) \
+ printk(KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
+#define bch_err_inum_offset(c, _inum, _offset, fmt, ...) \
+ printk(KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
#define bch_err_ratelimited(c, fmt, ...) \
printk_ratelimited(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
+#define bch_err_dev_ratelimited(ca, fmt, ...) \
+ printk_ratelimited(KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
+#define bch_err_dev_offset_ratelimited(ca, _offset, fmt, ...) \
+ printk_ratelimited(KERN_ERR bch2_fmt_dev_offset(ca, _offset, 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_err_inum_offset_ratelimited(c, _inum, _offset, fmt, ...) \
+ printk_ratelimited(KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
+
+#define bch_err_fn(_c, _ret) \
+ bch_err(_c, "%s(): error %s", __func__, bch2_err_str(_ret))
+#define bch_err_msg(_c, _ret, _msg) \
+ bch_err(_c, "%s(): error " _msg " %s", __func__, bch2_err_str(_ret))
#define bch_verbose(c, fmt, ...) \
do { \
"done in memory") \
BCH_DEBUG_PARAM(verify_all_btree_replicas, \
"When reading btree nodes, read all replicas and " \
- "compare them")
+ "compare them") \
+ BCH_DEBUG_PARAM(backpointers_no_use_write_buffer, \
+ "Don't use the write buffer for backpointers, enabling "\
+ "extra runtime checks")
-/* Parameters that should only be compiled in in debug mode: */
+/* Parameters that should only be compiled in debug mode: */
#define BCH_DEBUG_PARAMS_DEBUG() \
BCH_DEBUG_PARAM(expensive_debug_checks, \
"Enables various runtime debugging checks that " \
x(journal_flush_seq) \
x(blocked_journal) \
x(blocked_allocate) \
- x(blocked_allocate_open_bucket)
+ x(blocked_allocate_open_bucket) \
+ x(nocow_lock_contended)
enum bch_time_stats {
#define x(name) BCH_TIME_##name,
#include "alloc_types.h"
#include "btree_types.h"
+#include "btree_write_buffer_types.h"
#include "buckets_types.h"
#include "buckets_waiting_for_journal_types.h"
#include "clock_types.h"
GC_PHASE_BTREE_freespace,
GC_PHASE_BTREE_need_discard,
GC_PHASE_BTREE_backpointers,
+ GC_PHASE_BTREE_bucket_gens,
+ GC_PHASE_BTREE_snapshot_trees,
GC_PHASE_PENDING_DELETE,
};
* Committed by bch2_write_super() -> bch_fs_mi_update()
*/
struct bch_member_cpu mi;
- uuid_le uuid;
+ __uuid_t uuid;
char name[BDEVNAME_SIZE];
struct bch_sb_handle disk_sb;
struct bch_sb *sb_read_scratch;
int sb_write_error;
dev_t dev;
+ atomic_t flush_seq;
struct bch_devs_mask self;
/* Allocator: */
u64 new_fs_bucket_idx;
- u64 bucket_alloc_trans_early_cursor;
+ u64 alloc_cursor;
unsigned nr_open_buckets;
unsigned nr_btree_reserve;
- open_bucket_idx_t open_buckets_partial[OPEN_BUCKETS_COUNT];
- open_bucket_idx_t open_buckets_partial_nr;
-
size_t inc_gen_needs_gc;
size_t inc_gen_really_needs_gc;
size_t buckets_waiting_on_journal;
/* The rest of this all shows up in sysfs */
atomic64_t cur_latency[2];
- struct time_stats io_latency[2];
+ struct bch2_time_stats io_latency[2];
#define CONGESTED_MAX 1024
atomic_t congested;
/* shutdown: */
BCH_FS_STOPPING,
BCH_FS_EMERGENCY_RO,
+ BCH_FS_GOING_RO,
BCH_FS_WRITE_DISABLE_COMPLETE,
BCH_FS_CLEAN_SHUTDOWN,
/* fsck passes: */
- BCH_FS_TOPOLOGY_REPAIR_DONE,
- BCH_FS_INITIAL_GC_DONE, /* kill when we enumerate fsck passes */
- BCH_FS_CHECK_LRUS_DONE,
- BCH_FS_CHECK_BACKPOINTERS_DONE,
- BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE,
BCH_FS_FSCK_DONE,
BCH_FS_INITIAL_GC_UNFIXED, /* kill when we enumerate fsck errors */
BCH_FS_NEED_ANOTHER_GC,
#define BCH_TRANSACTIONS_NR 128
struct btree_transaction_stats {
+ struct bch2_time_stats lock_hold_times;
struct mutex lock;
- struct time_stats lock_hold_times;
unsigned nr_max_paths;
+ unsigned wb_updates_size;
unsigned max_mem;
char *max_paths_text;
};
#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 })
+#define BCH_WRITE_REFS() \
+ x(trans) \
+ x(write) \
+ x(promote) \
+ x(node_rewrite) \
+ x(stripe_create) \
+ x(stripe_delete) \
+ x(reflink) \
+ x(fallocate) \
+ x(discard) \
+ x(invalidate) \
+ x(delete_dead_snapshots) \
+ x(snapshot_delete_pagecache) \
+ x(sysfs)
+
+enum bch_write_ref {
+#define x(n) BCH_WRITE_REF_##n,
+ BCH_WRITE_REFS()
+#undef x
+ BCH_WRITE_REF_NR,
+};
+
+#define PASS_SILENT BIT(0)
+#define PASS_FSCK BIT(1)
+#define PASS_UNCLEAN BIT(2)
+#define PASS_ALWAYS BIT(3)
+
+#define BCH_RECOVERY_PASSES() \
+ x(alloc_read, PASS_ALWAYS) \
+ x(stripes_read, PASS_ALWAYS) \
+ x(initialize_subvolumes, 0) \
+ x(snapshots_read, PASS_ALWAYS) \
+ x(check_topology, 0) \
+ x(check_allocations, PASS_FSCK) \
+ x(set_may_go_rw, PASS_ALWAYS|PASS_SILENT) \
+ x(journal_replay, PASS_ALWAYS) \
+ x(check_alloc_info, PASS_FSCK) \
+ x(check_lrus, PASS_FSCK) \
+ x(check_btree_backpointers, PASS_FSCK) \
+ x(check_backpointers_to_extents,PASS_FSCK) \
+ x(check_extents_to_backpointers,PASS_FSCK) \
+ x(check_alloc_to_lru_refs, PASS_FSCK) \
+ x(fs_freespace_init, PASS_ALWAYS|PASS_SILENT) \
+ x(bucket_gens_init, 0) \
+ x(check_snapshot_trees, PASS_FSCK) \
+ x(check_snapshots, PASS_FSCK) \
+ x(check_subvols, PASS_FSCK) \
+ x(delete_dead_snapshots, PASS_FSCK|PASS_UNCLEAN) \
+ x(fs_upgrade_for_subvolumes, 0) \
+ x(check_inodes, PASS_FSCK|PASS_UNCLEAN) \
+ x(check_extents, PASS_FSCK) \
+ x(check_dirents, PASS_FSCK) \
+ x(check_xattrs, PASS_FSCK) \
+ x(check_root, PASS_FSCK) \
+ x(check_directory_structure, PASS_FSCK) \
+ x(check_nlinks, PASS_FSCK) \
+ x(fix_reflink_p, 0) \
+
+enum bch_recovery_pass {
+#define x(n, when) BCH_RECOVERY_PASS_##n,
+ BCH_RECOVERY_PASSES()
+#undef x
+};
+
struct bch_fs {
struct closure cl;
struct rw_semaphore state_lock;
/* Counts outstanding writes, for clean transition to read-only */
+#ifdef BCH_WRITE_REF_DEBUG
+ atomic_long_t writes[BCH_WRITE_REF_NR];
+#else
struct percpu_ref writes;
+#endif
struct work_struct read_only_work;
struct bch_dev __rcu *devs[BCH_SB_MEMBERS_MAX];
/* Updated by bch2_sb_update():*/
struct {
- uuid_le uuid;
- uuid_le user_uuid;
+ __uuid_t uuid;
+ __uuid_t user_uuid;
u16 version;
u16 version_min;
+ u16 version_upgrade_complete;
u8 nr_devices;
u8 clean;
struct mutex sb_lock;
/* snapshot.c: */
- GENRADIX(struct snapshot_t) snapshots;
- struct bch_snapshot_table __rcu *snapshot_table;
+ struct snapshot_table __rcu *snapshots;
+ size_t snapshot_table_size;
struct mutex snapshot_table_lock;
+
struct work_struct snapshot_delete_work;
struct work_struct snapshot_wait_for_pagecache_and_delete_work;
snapshot_id_list snapshots_unlinked;
struct bio_set btree_bio;
struct workqueue_struct *io_complete_wq;
- struct btree_root btree_roots[BTREE_ID_NR];
+ struct btree_root btree_roots_known[BTREE_ID_NR];
+ DARRAY(struct btree_root) btree_roots_extra;
struct mutex btree_root_lock;
struct btree_cache btree_cache;
struct workqueue_struct *btree_interior_update_worker;
struct work_struct btree_interior_update_work;
+ struct list_head pending_node_rewrites;
+ struct mutex pending_node_rewrites_lock;
+
+ /* btree_io.c: */
+ spinlock_t btree_write_error_lock;
+ struct btree_write_stats {
+ atomic64_t nr;
+ atomic64_t bytes;
+ } btree_write_stats[BTREE_WRITE_TYPE_NR];
+
/* btree_iter.c: */
- struct mutex btree_trans_lock;
+ struct seqmutex btree_trans_lock;
struct list_head btree_trans_list;
mempool_t btree_paths_pool;
mempool_t btree_trans_mem_pool;
struct btree_key_cache btree_key_cache;
unsigned btree_key_cache_btrees;
+ struct btree_write_buffer btree_write_buffer;
+
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;
+ /*
+ * Use a dedicated wq for write ref holder tasks. Required to avoid
+ * dependency problems with other wq tasks that can block on ref
+ * draining, such as read-only transition.
+ */
+ struct workqueue_struct *write_ref_wq;
/* ALLOCATION */
struct bch_devs_mask rw_devs[BCH_DATA_NR];
struct open_bucket open_buckets[OPEN_BUCKETS_COUNT];
open_bucket_idx_t open_buckets_hash[OPEN_BUCKETS_COUNT];
+ open_bucket_idx_t open_buckets_partial[OPEN_BUCKETS_COUNT];
+ open_bucket_idx_t open_buckets_partial_nr;
+
struct write_point btree_write_point;
struct write_point rebalance_write_point;
struct mutex gc_gens_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 mutex bio_bounce_pages_lock;
mempool_t bio_bounce_pages;
+ struct bucket_nocow_lock_table
+ nocow_locks;
struct rhashtable promote_table;
mempool_t compression_bounce[2];
mempool_t large_bkey_pool;
+ /* MOVE.C */
+ struct list_head moving_context_list;
+ struct mutex moving_context_lock;
+
+ struct list_head data_progress_list;
+ struct mutex data_progress_lock;
+
/* 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_at;
s64 copygc_wait;
bool copygc_running;
wait_queue_head_t copygc_running_wq;
- /* DATA PROGRESS STATS */
- struct list_head data_progress_list;
- struct mutex data_progress_lock;
-
/* STRIPES: */
GENRADIX(struct stripe) stripes;
GENRADIX(struct gc_stripe) gc_stripes;
+ struct hlist_head ec_stripes_new[32];
+ spinlock_t ec_stripes_new_lock;
+
ec_stripes_heap ec_stripes_heap;
- spinlock_t ec_stripes_heap_lock;
+ struct mutex ec_stripes_heap_lock;
/* ERASURE CODING */
struct list_head ec_stripe_head_list;
struct list_head ec_stripe_new_list;
struct mutex ec_stripe_new_lock;
+ wait_queue_head_t ec_stripe_new_wait;
struct work_struct ec_stripe_create_work;
u64 ec_stripe_hint;
- struct bio_set ec_bioset;
-
struct work_struct ec_stripe_delete_work;
- struct llist_head ec_stripe_delete_list;
+
+ struct bio_set ec_bioset;
/* REFLINK */
- u64 reflink_hint;
reflink_gc_table reflink_gc_table;
size_t reflink_gc_nr;
+ /* fs.c */
+ struct list_head vfs_inodes_list;
+ struct mutex vfs_inodes_lock;
+
/* VFS IO PATH - fs-io.c */
struct bio_set writepage_bioset;
struct bio_set dio_write_bioset;
struct bio_set dio_read_bioset;
-
-
- atomic64_t btree_writes_nr;
- atomic64_t btree_writes_sectors;
- spinlock_t btree_write_error_lock;
+ struct bio_set nocow_flush_bioset;
/* ERRORS */
struct list_head fsck_errors;
/* QUOTAS */
struct bch_memquota_type quotas[QTYP_NR];
+ /* RECOVERY */
+ u64 journal_replay_seq_start;
+ u64 journal_replay_seq_end;
+ enum bch_recovery_pass curr_recovery_pass;
+ /* bitmap of explicitly enabled recovery passes: */
+ u64 recovery_passes_explicit;
+
/* DEBUG JUNK */
struct dentry *fs_debug_dir;
struct dentry *btree_debug_dir;
unsigned copy_gc_enabled:1;
bool promote_whole_extents;
- struct time_stats times[BCH_TIME_STAT_NR];
+ struct bch2_time_stats times[BCH_TIME_STAT_NR];
- const char *btree_transaction_fns[BCH_TRANSACTIONS_NR];
struct btree_transaction_stats btree_transaction_stats[BCH_TRANSACTIONS_NR];
};
+extern struct wait_queue_head bch2_read_only_wait;
+
+static inline void bch2_write_ref_get(struct bch_fs *c, enum bch_write_ref ref)
+{
+#ifdef BCH_WRITE_REF_DEBUG
+ atomic_long_inc(&c->writes[ref]);
+#else
+ percpu_ref_get(&c->writes);
+#endif
+}
+
+static inline bool bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
+{
+#ifdef BCH_WRITE_REF_DEBUG
+ return !test_bit(BCH_FS_GOING_RO, &c->flags) &&
+ atomic_long_inc_not_zero(&c->writes[ref]);
+#else
+ return percpu_ref_tryget_live(&c->writes);
+#endif
+}
+
+static inline void bch2_write_ref_put(struct bch_fs *c, enum bch_write_ref ref)
+{
+#ifdef BCH_WRITE_REF_DEBUG
+ long v = atomic_long_dec_return(&c->writes[ref]);
+
+ BUG_ON(v < 0);
+ if (v)
+ return;
+ for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++)
+ if (atomic_long_read(&c->writes[i]))
+ return;
+
+ set_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
+ wake_up(&bch2_read_only_wait);
+#else
+ percpu_ref_put(&c->writes);
+#endif
+}
+
static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
{
#ifndef NO_BCACHEFS_FS
return dev < c->sb.nr_devices && c->devs[dev];
}
+/*
+ * For when we need to rewind recovery passes and run a pass we skipped:
+ */
+static inline int bch2_run_explicit_recovery_pass(struct bch_fs *c,
+ enum bch_recovery_pass pass)
+{
+ c->recovery_passes_explicit |= BIT_ULL(pass);
+
+ if (c->curr_recovery_pass >= pass) {
+ c->curr_recovery_pass = pass;
+ return -BCH_ERR_restart_recovery;
+ } else {
+ return 0;
+ }
+}
+
+#define BKEY_PADDED_ONSTACK(key, pad) \
+ struct { struct bkey_i key; __u64 key ## _pad[pad]; }
+
#endif /* _BCACHEFS_H */