*
* 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 "util.h"
#define dynamic_fault(...) 0
#define race_fault(...) 0
+#define trace_and_count(_c, _name, ...) \
+do { \
+ this_cpu_inc((_c)->counters[BCH_COUNTER_##_name]); \
+ trace_##_name(__VA_ARGS__); \
+} while (0)
+
#define bch2_fs_init_fault(name) \
dynamic_fault("bcachefs:bch_fs_init:" name)
#define bch2_meta_read_fault(name) \
dynamic_fault("bcachefs:meta:write:" name)
#ifdef __KERNEL__
-#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)
+#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_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_verbose(c, fmt, ...) \
do { \
"When reading btree nodes, read all replicas and " \
"compare them")
-/* 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 " \
"significantly affect performance") \
BCH_DEBUG_PARAM(debug_check_iterators, \
"Enables extra verification for btree iterators") \
- BCH_DEBUG_PARAM(debug_check_bkeys, \
- "Run bkey_debugcheck (primarily checking GC/allocation "\
- "information) when iterating over keys") \
BCH_DEBUG_PARAM(debug_check_btree_accounting, \
"Verify btree accounting for keys within a node") \
BCH_DEBUG_PARAM(journal_seq_verify, \
x(btree_interior_update_foreground) \
x(btree_interior_update_total) \
x(btree_gc) \
- 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_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,
GC_PHASE_BTREE_reflink,
GC_PHASE_BTREE_subvolumes,
GC_PHASE_BTREE_snapshots,
+ GC_PHASE_BTREE_lru,
+ GC_PHASE_BTREE_freespace,
+ GC_PHASE_BTREE_need_discard,
+ GC_PHASE_BTREE_backpointers,
+ GC_PHASE_BTREE_bucket_gens,
GC_PHASE_PENDING_DELETE,
};
struct bch_sb *sb_read_scratch;
int sb_write_error;
dev_t dev;
+ atomic_t flush_seq;
struct bch_devs_mask self;
* gc_lock, for device resize - holding any is sufficient for access:
* Or rcu_read_lock(), but only for ptr_stale():
*/
- struct bucket_array __rcu *buckets[2];
+ struct bucket_array __rcu *buckets_gc;
struct bucket_gens __rcu *bucket_gens;
u8 *oldest_gen;
unsigned long *buckets_nouse;
/* Allocator: */
u64 new_fs_bucket_idx;
- struct task_struct __rcu *alloc_thread;
+ u64 bucket_alloc_trans_early_cursor;
- /*
- * free: Buckets that are ready to be used
- *
- * free_inc: Incoming buckets - these are buckets that currently have
- * cached data in them, and we can't reuse them until after we write
- * their new gen to disk. After prio_write() finishes writing the new
- * gens/prios, they'll be moved to the free list (and possibly discarded
- * in the process)
- */
- alloc_fifo free[RESERVE_NR];
- alloc_fifo free_inc;
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 fifo_last_bucket;
-
size_t inc_gen_needs_gc;
size_t inc_gen_really_needs_gc;
size_t buckets_waiting_on_journal;
- enum allocator_states allocator_state;
-
- alloc_heap alloc_heap;
-
atomic64_t rebalance_work;
struct journal_device journal;
enum {
/* startup: */
- BCH_FS_INITIALIZED,
- BCH_FS_ALLOC_READ_DONE,
- 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_MAY_GO_RW,
BCH_FS_RW,
BCH_FS_WAS_RW,
BCH_FS_STOPPING,
BCH_FS_EMERGENCY_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_ALLOC_DONE,
+ 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,
+
+ BCH_FS_HAVE_DELETED_SNAPSHOTS,
/* errors: */
BCH_FS_ERROR,
BCH_FS_TOPOLOGY_ERROR,
BCH_FS_ERRORS_FIXED,
BCH_FS_ERRORS_NOT_FIXED,
-
- /* misc: */
- BCH_FS_NEED_ANOTHER_GC,
- BCH_FS_DELETED_NODES,
- BCH_FS_REBUILD_REPLICAS,
- BCH_FS_HOLD_BTREE_WRITES,
};
struct btree_debug {
unsigned id;
- struct dentry *btree;
- struct dentry *btree_format;
- struct dentry *failed;
+};
+
+#define BCH_TRANSACTIONS_NR 128
+
+struct btree_transaction_stats {
+ struct mutex lock;
+ struct time_stats lock_hold_times;
+ unsigned nr_max_paths;
+ unsigned max_mem;
+ char *max_paths_text;
};
struct bch_fs_pcpu {
struct journal_keys {
struct journal_key {
+ u64 journal_seq;
+ u32 journal_offset;
enum btree_id btree_id:8;
unsigned level:8;
bool allocated;
bool overwritten;
struct bkey_i *k;
- u32 journal_seq;
- u32 journal_offset;
} *d;
+ /*
+ * Gap buffer: instead of all the empty space in the array being at the
+ * end of the buffer - from @nr to @size - the empty space is at @gap.
+ * This means that sequential insertions are O(n) instead of O(n^2).
+ */
+ size_t gap;
size_t nr;
size_t size;
- u64 journal_seq_base;
};
struct btree_path_buf {
struct list_head list;
struct kobject kobj;
+ struct kobject counters_kobj;
struct kobject internal;
struct kobject opts_dir;
struct kobject time_stats;
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;
+ snapshot_id_list snapshots_unlinked;
struct mutex snapshots_unlinked_lock;
/* BTREE CACHE */
struct workqueue_struct *btree_interior_update_worker;
struct work_struct btree_interior_update_work;
+ /* 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 list_head btree_trans_list;
unsigned write_points_nr;
struct buckets_waiting_for_journal buckets_waiting_for_journal;
+ struct work_struct discard_work;
+ struct work_struct invalidate_work;
/* GARBAGE COLLECTION */
struct task_struct *gc_thread;
* it's not while a gc is in progress.
*/
struct rw_semaphore gc_lock;
+ 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];
copygc_heap copygc_heap;
struct write_point copygc_write_point;
s64 copygc_wait;
+ bool copygc_running;
+ wait_queue_head_t copygc_running_wq;
/* DATA PROGRESS STATS */
struct list_head data_progress_list;
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;
struct bch_memquota_type quotas[QTYP_NR];
/* DEBUG JUNK */
- struct dentry *debug;
+ struct dentry *fs_debug_dir;
+ struct dentry *btree_debug_dir;
struct btree_debug btree_debug[BTREE_ID_NR];
struct btree *verify_data;
struct btree_node *verify_ondisk;
mempool_t btree_bounce_pool;
struct journal journal;
- struct list_head journal_entries;
+ GENRADIX(struct journal_replay *) journal_entries;
+ u64 journal_entries_base_seq;
struct journal_keys journal_keys;
struct list_head journal_iters;
u64 last_bucket_seq_cleanup;
- /* The rest of this all shows up in sysfs */
- atomic_long_t read_realloc_races;
- atomic_long_t extent_migrate_done;
- atomic_long_t extent_migrate_raced;
+ u64 counters_on_mount[BCH_COUNTER_NR];
+ u64 __percpu *counters;
unsigned btree_gc_periodic:1;
unsigned copy_gc_enabled:1;
bool promote_whole_extents;
struct time_stats times[BCH_TIME_STAT_NR];
+
+ struct btree_transaction_stats btree_transaction_stats[BCH_TRANSACTIONS_NR];
};
static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
projects / bcachefs-tools-debian / blobdiff