X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbcachefs.h;h=3f88e7eac17c13e7e9b41e55c458763b948fa336;hb=5966e59eaeb4033f40ddd84fbe6445598a908163;hp=33186fa82682e037af78dcb3543d5bbf3049e26b;hpb=188b6d0c8ef1c02462a744b176557c27220112c9;p=bcachefs-tools-debian diff --git a/libbcachefs/bcachefs.h b/libbcachefs/bcachefs.h index 33186fa..3f88e7e 100644 --- a/libbcachefs/bcachefs.h +++ b/libbcachefs/bcachefs.h @@ -206,9 +206,14 @@ #include "bcachefs_format.h" #include "errcode.h" #include "fifo.h" +#include "nocow_locking_types.h" #include "opts.h" #include "util.h" +#ifdef CONFIG_BCACHEFS_DEBUG +#define BCH_WRITE_REF_DEBUG +#endif + #define dynamic_fault(...) 0 #define race_fault(...) 0 @@ -226,15 +231,31 @@ 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, ...) \ @@ -243,13 +264,28 @@ do { \ 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 { \ @@ -280,9 +316,12 @@ 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 " \ @@ -345,7 +384,8 @@ BCH_DEBUG_PARAMS_DEBUG() 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, @@ -356,6 +396,7 @@ enum bch_time_stats { #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" @@ -400,6 +441,7 @@ enum gc_phase { GC_PHASE_BTREE_freespace, GC_PHASE_BTREE_need_discard, GC_PHASE_BTREE_backpointers, + GC_PHASE_BTREE_bucket_gens, GC_PHASE_PENDING_DELETE, }; @@ -444,6 +486,7 @@ struct bch_dev { struct bch_sb *sb_read_scratch; int sb_write_error; dev_t dev; + atomic_t flush_seq; struct bch_devs_mask self; @@ -468,14 +511,11 @@ struct bch_dev { /* 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; @@ -489,7 +529,7 @@ struct bch_dev { /* 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; @@ -508,12 +548,14 @@ enum { /* 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_ALLOC_DONE, BCH_FS_CHECK_LRUS_DONE, BCH_FS_CHECK_BACKPOINTERS_DONE, BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, @@ -537,9 +579,10 @@ struct btree_debug { #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; }; @@ -598,6 +641,29 @@ typedef struct { #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(move) \ + 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, +}; + struct bch_fs { struct closure cl; @@ -619,7 +685,11 @@ struct bch_fs { 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]; @@ -707,6 +777,16 @@ struct bch_fs { 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 list_head btree_trans_list; @@ -720,6 +800,8 @@ struct bch_fs { 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.. */ @@ -774,6 +856,9 @@ struct bch_fs { 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; @@ -814,11 +899,14 @@ struct bch_fs { 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]; @@ -839,7 +927,6 @@ struct bch_fs { /* COPYGC */ struct task_struct *copygc_thread; - copygc_heap copygc_heap; struct write_point copygc_write_point; s64 copygc_wait; bool copygc_running; @@ -853,8 +940,11 @@ struct bch_fs { 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; @@ -880,11 +970,7 @@ struct bch_fs { 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; @@ -928,11 +1014,51 @@ struct bch_fs { 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]; 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