X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbuckets.c;h=c0dac04253f7c8787570fdb50f137b0ae20ba26a;hb=3070f45d968575efd42ecf44f94427203b2ab62e;hp=d5ec4d727d0e56ad2e35cdc8832b781aa1358dc9;hpb=06a4dda333b7ae0d3678b0ae61200edb5d0fd40b;p=bcachefs-tools-debian diff --git a/libbcachefs/buckets.c b/libbcachefs/buckets.c index d5ec4d7..c0dac04 100644 --- a/libbcachefs/buckets.c +++ b/libbcachefs/buckets.c @@ -7,20 +7,23 @@ #include "bcachefs.h" #include "alloc_background.h" +#include "backpointers.h" #include "bset.h" #include "btree_gc.h" #include "btree_update.h" #include "buckets.h" +#include "buckets_waiting_for_journal.h" #include "ec.h" #include "error.h" +#include "inode.h" #include "movinggc.h" #include "recovery.h" #include "reflink.h" #include "replicas.h" #include "subvolume.h" +#include "trace.h" #include -#include static inline void fs_usage_data_type_to_base(struct bch_fs_usage *fs_usage, enum bch_data_type data_type, @@ -42,66 +45,25 @@ static inline void fs_usage_data_type_to_base(struct bch_fs_usage *fs_usage, } } -/* - * Clear journal_seq_valid for buckets for which it's not needed, to prevent - * wraparound: - */ -void bch2_bucket_seq_cleanup(struct bch_fs *c) -{ - u64 journal_seq = atomic64_read(&c->journal.seq); - u16 last_seq_ondisk = c->journal.last_seq_ondisk; - struct bch_dev *ca; - struct bucket_array *buckets; - struct bucket *g; - struct bucket_mark m; - unsigned i; - - if (journal_seq - c->last_bucket_seq_cleanup < - (1U << (BUCKET_JOURNAL_SEQ_BITS - 2))) - return; - - c->last_bucket_seq_cleanup = journal_seq; - - for_each_member_device(ca, c, i) { - down_read(&ca->bucket_lock); - buckets = bucket_array(ca); - - for_each_bucket(g, buckets) { - bucket_cmpxchg(g, m, ({ - if (!m.journal_seq_valid || - bucket_needs_journal_commit(m, last_seq_ondisk)) - break; - - m.journal_seq_valid = 0; - })); - } - up_read(&ca->bucket_lock); - } -} - void bch2_fs_usage_initialize(struct bch_fs *c) { - struct bch_fs_usage *usage; - struct bch_dev *ca; - unsigned i; - percpu_down_write(&c->mark_lock); - usage = c->usage_base; + struct bch_fs_usage *usage = c->usage_base; - for (i = 0; i < ARRAY_SIZE(c->usage); i++) + for (unsigned i = 0; i < ARRAY_SIZE(c->usage); i++) bch2_fs_usage_acc_to_base(c, i); - for (i = 0; i < BCH_REPLICAS_MAX; i++) + for (unsigned i = 0; i < BCH_REPLICAS_MAX; i++) usage->reserved += usage->persistent_reserved[i]; - for (i = 0; i < c->replicas.nr; i++) { - struct bch_replicas_entry *e = + for (unsigned i = 0; i < c->replicas.nr; i++) { + struct bch_replicas_entry_v1 *e = cpu_replicas_entry(&c->replicas, i); fs_usage_data_type_to_base(usage, e->data_type, usage->replicas[i]); } - for_each_member_device(ca, c, i) { + for_each_member_device(c, ca) { struct bch_dev_usage dev = bch2_dev_usage_read(ca); usage->hidden += (dev.d[BCH_DATA_sb].buckets + @@ -116,34 +78,24 @@ static inline struct bch_dev_usage *dev_usage_ptr(struct bch_dev *ca, unsigned journal_seq, bool gc) { + BUG_ON(!gc && !journal_seq); + return this_cpu_ptr(gc ? ca->usage_gc : ca->usage[journal_seq & JOURNAL_BUF_MASK]); } -struct bch_dev_usage bch2_dev_usage_read(struct bch_dev *ca) +void bch2_dev_usage_read_fast(struct bch_dev *ca, struct bch_dev_usage *usage) { struct bch_fs *c = ca->fs; - struct bch_dev_usage ret; unsigned seq, i, u64s = dev_usage_u64s(); do { seq = read_seqcount_begin(&c->usage_lock); - memcpy(&ret, ca->usage_base, u64s * sizeof(u64)); + memcpy(usage, ca->usage_base, u64s * sizeof(u64)); for (i = 0; i < ARRAY_SIZE(ca->usage); i++) - acc_u64s_percpu((u64 *) &ret, (u64 __percpu *) ca->usage[i], u64s); + acc_u64s_percpu((u64 *) usage, (u64 __percpu *) ca->usage[i], u64s); } while (read_seqcount_retry(&c->usage_lock, seq)); - - return ret; -} - -static inline struct bch_fs_usage *fs_usage_ptr(struct bch_fs *c, - unsigned journal_seq, - bool gc) -{ - return this_cpu_ptr(gc - ? c->usage_gc - : c->usage[journal_seq & JOURNAL_BUF_MASK]); } u64 bch2_fs_usage_read_one(struct bch_fs *c, u64 *v) @@ -169,25 +121,32 @@ u64 bch2_fs_usage_read_one(struct bch_fs *c, u64 *v) struct bch_fs_usage_online *bch2_fs_usage_read(struct bch_fs *c) { struct bch_fs_usage_online *ret; - unsigned seq, i, u64s; + unsigned nr_replicas = READ_ONCE(c->replicas.nr); + unsigned seq, i; +retry: + ret = kmalloc(__fs_usage_online_u64s(nr_replicas) * sizeof(u64), GFP_KERNEL); + if (unlikely(!ret)) + return NULL; percpu_down_read(&c->mark_lock); - ret = kmalloc(sizeof(struct bch_fs_usage_online) + - sizeof(u64) * c->replicas.nr, GFP_NOFS); - if (unlikely(!ret)) { + if (nr_replicas != c->replicas.nr) { + nr_replicas = c->replicas.nr; percpu_up_read(&c->mark_lock); - return NULL; + kfree(ret); + goto retry; } ret->online_reserved = percpu_u64_get(c->online_reserved); - u64s = fs_usage_u64s(c); do { seq = read_seqcount_begin(&c->usage_lock); - memcpy(&ret->u, c->usage_base, u64s * sizeof(u64)); + unsafe_memcpy(&ret->u, c->usage_base, + __fs_usage_u64s(nr_replicas) * sizeof(u64), + "embedded variable length struct"); for (i = 0; i < ARRAY_SIZE(c->usage); i++) - acc_u64s_percpu((u64 *) &ret->u, (u64 __percpu *) c->usage[i], u64s); + acc_u64s_percpu((u64 *) &ret->u, (u64 __percpu *) c->usage[i], + __fs_usage_u64s(nr_replicas)); } while (read_seqcount_retry(&c->usage_lock, seq)); return ret; @@ -195,8 +154,7 @@ struct bch_fs_usage_online *bch2_fs_usage_read(struct bch_fs *c) void bch2_fs_usage_acc_to_base(struct bch_fs *c, unsigned idx) { - struct bch_dev *ca; - unsigned i, u64s = fs_usage_u64s(c); + unsigned u64s = fs_usage_u64s(c); BUG_ON(idx >= ARRAY_SIZE(c->usage)); @@ -208,7 +166,7 @@ void bch2_fs_usage_acc_to_base(struct bch_fs *c, unsigned idx) percpu_memset(c->usage[idx], 0, u64s * sizeof(u64)); rcu_read_lock(); - for_each_member_device_rcu(ca, c, i, NULL) { + for_each_member_device_rcu(c, ca, NULL) { u64s = dev_usage_u64s(); acc_u64s_percpu((u64 *) ca->usage_base, @@ -227,36 +185,36 @@ void bch2_fs_usage_to_text(struct printbuf *out, { unsigned i; - pr_buf(out, "capacity:\t\t\t%llu\n", c->capacity); + prt_printf(out, "capacity:\t\t\t%llu\n", c->capacity); - pr_buf(out, "hidden:\t\t\t\t%llu\n", + prt_printf(out, "hidden:\t\t\t\t%llu\n", fs_usage->u.hidden); - pr_buf(out, "data:\t\t\t\t%llu\n", + prt_printf(out, "data:\t\t\t\t%llu\n", fs_usage->u.data); - pr_buf(out, "cached:\t\t\t\t%llu\n", + prt_printf(out, "cached:\t\t\t\t%llu\n", fs_usage->u.cached); - pr_buf(out, "reserved:\t\t\t%llu\n", + prt_printf(out, "reserved:\t\t\t%llu\n", fs_usage->u.reserved); - pr_buf(out, "nr_inodes:\t\t\t%llu\n", + prt_printf(out, "nr_inodes:\t\t\t%llu\n", fs_usage->u.nr_inodes); - pr_buf(out, "online reserved:\t\t%llu\n", + prt_printf(out, "online reserved:\t\t%llu\n", fs_usage->online_reserved); for (i = 0; i < ARRAY_SIZE(fs_usage->u.persistent_reserved); i++) { - pr_buf(out, "%u replicas:\n", i + 1); - pr_buf(out, "\treserved:\t\t%llu\n", + prt_printf(out, "%u replicas:\n", i + 1); + prt_printf(out, "\treserved:\t\t%llu\n", fs_usage->u.persistent_reserved[i]); } for (i = 0; i < c->replicas.nr; i++) { - struct bch_replicas_entry *e = + struct bch_replicas_entry_v1 *e = cpu_replicas_entry(&c->replicas, i); - pr_buf(out, "\t"); + prt_printf(out, "\t"); bch2_replicas_entry_to_text(out, e); - pr_buf(out, ":\t%llu\n", fs_usage->u.replicas[i]); + prt_printf(out, ":\t%llu\n", fs_usage->u.replicas[i]); } } @@ -309,91 +267,91 @@ bch2_fs_usage_read_short(struct bch_fs *c) return ret; } -static inline int is_unavailable_bucket(struct bucket_mark m) -{ - return !is_available_bucket(m); -} - -static inline int bucket_sectors_fragmented(struct bch_dev *ca, - struct bucket_mark m) -{ - return bucket_sectors_used(m) - ? max(0, (int) ca->mi.bucket_size - (int) bucket_sectors_used(m)) - : 0; -} - -static inline int is_stripe_data_bucket(struct bucket_mark m) -{ - return m.stripe && m.data_type != BCH_DATA_parity; -} - -static inline enum bch_data_type bucket_type(struct bucket_mark m) -{ - return m.cached_sectors && !m.dirty_sectors - ? BCH_DATA_cached - : m.data_type; -} - -static bool bucket_became_unavailable(struct bucket_mark old, - struct bucket_mark new) +void bch2_dev_usage_init(struct bch_dev *ca) { - return is_available_bucket(old) && - !is_available_bucket(new); + ca->usage_base->d[BCH_DATA_free].buckets = ca->mi.nbuckets - ca->mi.first_bucket; } -static inline void account_bucket(struct bch_fs_usage *fs_usage, - struct bch_dev_usage *dev_usage, - enum bch_data_type type, - int nr, s64 size) +void bch2_dev_usage_to_text(struct printbuf *out, struct bch_dev_usage *usage) { - if (type == BCH_DATA_sb || type == BCH_DATA_journal) - fs_usage->hidden += size; + prt_tab(out); + prt_str(out, "buckets"); + prt_tab_rjust(out); + prt_str(out, "sectors"); + prt_tab_rjust(out); + prt_str(out, "fragmented"); + prt_tab_rjust(out); + prt_newline(out); - dev_usage->d[type].buckets += nr; + for (unsigned i = 0; i < BCH_DATA_NR; i++) { + prt_str(out, bch2_data_types[i]); + prt_tab(out); + prt_u64(out, usage->d[i].buckets); + prt_tab_rjust(out); + prt_u64(out, usage->d[i].sectors); + prt_tab_rjust(out); + prt_u64(out, usage->d[i].fragmented); + prt_tab_rjust(out); + prt_newline(out); + } } static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca, - struct bucket_mark old, struct bucket_mark new, + struct bch_alloc_v4 old, + struct bch_alloc_v4 new, u64 journal_seq, bool gc) { struct bch_fs_usage *fs_usage; struct bch_dev_usage *u; - percpu_rwsem_assert_held(&c->mark_lock); - preempt_disable(); fs_usage = fs_usage_ptr(c, journal_seq, gc); - u = dev_usage_ptr(ca, journal_seq, gc); - if (bucket_type(old)) - account_bucket(fs_usage, u, bucket_type(old), - -1, -ca->mi.bucket_size); + if (data_type_is_hidden(old.data_type)) + fs_usage->hidden -= ca->mi.bucket_size; + if (data_type_is_hidden(new.data_type)) + fs_usage->hidden += ca->mi.bucket_size; - if (bucket_type(new)) - account_bucket(fs_usage, u, bucket_type(new), - 1, ca->mi.bucket_size); + u = dev_usage_ptr(ca, journal_seq, gc); + + u->d[old.data_type].buckets--; + u->d[new.data_type].buckets++; - u->buckets_ec += (int) new.stripe - (int) old.stripe; - u->buckets_unavailable += - is_unavailable_bucket(new) - is_unavailable_bucket(old); + u->d[old.data_type].sectors -= bch2_bucket_sectors_dirty(old); + u->d[new.data_type].sectors += bch2_bucket_sectors_dirty(new); - u->d[old.data_type].sectors -= old.dirty_sectors; - u->d[new.data_type].sectors += new.dirty_sectors; - u->d[BCH_DATA_cached].sectors += - (int) new.cached_sectors - (int) old.cached_sectors; + u->d[BCH_DATA_cached].sectors += new.cached_sectors; + u->d[BCH_DATA_cached].sectors -= old.cached_sectors; - u->d[old.data_type].fragmented -= bucket_sectors_fragmented(ca, old); - u->d[new.data_type].fragmented += bucket_sectors_fragmented(ca, new); + u->d[old.data_type].fragmented -= bch2_bucket_sectors_fragmented(ca, old); + u->d[new.data_type].fragmented += bch2_bucket_sectors_fragmented(ca, new); preempt_enable(); +} + +static inline struct bch_alloc_v4 bucket_m_to_alloc(struct bucket b) +{ + return (struct bch_alloc_v4) { + .gen = b.gen, + .data_type = b.data_type, + .dirty_sectors = b.dirty_sectors, + .cached_sectors = b.cached_sectors, + .stripe = b.stripe, + }; +} - if (!is_available_bucket(old) && is_available_bucket(new)) - bch2_wake_allocator(ca); +static void bch2_dev_usage_update_m(struct bch_fs *c, struct bch_dev *ca, + struct bucket old, struct bucket new) +{ + bch2_dev_usage_update(c, ca, + bucket_m_to_alloc(old), + bucket_m_to_alloc(new), + 0, true); } static inline int __update_replicas(struct bch_fs *c, struct bch_fs_usage *fs_usage, - struct bch_replicas_entry *r, + struct bch_replicas_entry_v1 *r, s64 sectors) { int idx = bch2_replicas_entry_idx(c, r); @@ -406,25 +364,48 @@ static inline int __update_replicas(struct bch_fs *c, return 0; } -static inline int update_replicas(struct bch_fs *c, - struct bch_replicas_entry *r, s64 sectors, +static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k, + struct bch_replicas_entry_v1 *r, s64 sectors, unsigned journal_seq, bool gc) { - struct bch_fs_usage __percpu *fs_usage; - int idx = bch2_replicas_entry_idx(c, r); + struct bch_fs_usage *fs_usage; + int idx, ret = 0; + struct printbuf buf = PRINTBUF; - if (idx < 0) - return -1; + percpu_down_read(&c->mark_lock); + + idx = bch2_replicas_entry_idx(c, r); + if (idx < 0 && + fsck_err(c, ptr_to_missing_replicas_entry, + "no replicas entry\n while marking %s", + (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) { + percpu_up_read(&c->mark_lock); + ret = bch2_mark_replicas(c, r); + percpu_down_read(&c->mark_lock); + + if (ret) + goto err; + idx = bch2_replicas_entry_idx(c, r); + } + if (idx < 0) { + ret = -1; + goto err; + } preempt_disable(); fs_usage = fs_usage_ptr(c, journal_seq, gc); fs_usage_data_type_to_base(fs_usage, r->data_type, sectors); fs_usage->replicas[idx] += sectors; preempt_enable(); - return 0; +err: +fsck_err: + percpu_up_read(&c->mark_lock); + printbuf_exit(&buf); + return ret; } static inline int update_cached_sectors(struct bch_fs *c, + struct bkey_s_c k, unsigned dev, s64 sectors, unsigned journal_seq, bool gc) { @@ -432,11 +413,11 @@ static inline int update_cached_sectors(struct bch_fs *c, bch2_replicas_entry_cached(&r.e, dev); - return update_replicas(c, &r.e, sectors, journal_seq, gc); + return update_replicas(c, k, &r.e, sectors, journal_seq, gc); } -static struct replicas_delta_list * -replicas_deltas_realloc(struct btree_trans *trans, unsigned more) +static int __replicas_deltas_realloc(struct btree_trans *trans, unsigned more, + gfp_t gfp) { struct replicas_delta_list *d = trans->fs_usage_deltas; unsigned new_size = d ? (d->size + more) * 2 : 128; @@ -445,12 +426,16 @@ replicas_deltas_realloc(struct btree_trans *trans, unsigned more) WARN_ON_ONCE(alloc_size > REPLICAS_DELTA_LIST_MAX); if (!d || d->used + more > d->size) { - d = krealloc(d, alloc_size, GFP_NOIO|__GFP_ZERO); + d = krealloc(d, alloc_size, gfp|__GFP_ZERO); + + if (unlikely(!d)) { + if (alloc_size > REPLICAS_DELTA_LIST_MAX) + return -ENOMEM; - BUG_ON(!d && alloc_size > REPLICAS_DELTA_LIST_MAX); + d = mempool_alloc(&trans->c->replicas_delta_pool, gfp); + if (!d) + return -ENOMEM; - if (!d) { - d = mempool_alloc(&trans->c->replicas_delta_pool, GFP_NOIO); memset(d, 0, REPLICAS_DELTA_LIST_MAX); if (trans->fs_usage_deltas) @@ -464,80 +449,66 @@ replicas_deltas_realloc(struct btree_trans *trans, unsigned more) d->size = new_size; trans->fs_usage_deltas = d; } - return d; + + return 0; } -static inline void update_replicas_list(struct btree_trans *trans, - struct bch_replicas_entry *r, - s64 sectors) +int bch2_replicas_deltas_realloc(struct btree_trans *trans, unsigned more) +{ + return allocate_dropping_locks_errcode(trans, + __replicas_deltas_realloc(trans, more, _gfp)); +} + +int bch2_update_replicas_list(struct btree_trans *trans, + struct bch_replicas_entry_v1 *r, + s64 sectors) { struct replicas_delta_list *d; struct replicas_delta *n; unsigned b; + int ret; if (!sectors) - return; + return 0; b = replicas_entry_bytes(r) + 8; - d = replicas_deltas_realloc(trans, b); + ret = bch2_replicas_deltas_realloc(trans, b); + if (ret) + return ret; + d = trans->fs_usage_deltas; n = (void *) d->d + d->used; n->delta = sectors; - memcpy(&n->r, r, replicas_entry_bytes(r)); + unsafe_memcpy((void *) n + offsetof(struct replicas_delta, r), + r, replicas_entry_bytes(r), + "flexible array member embedded in strcuct with padding"); bch2_replicas_entry_sort(&n->r); d->used += b; + return 0; } -static inline void update_cached_sectors_list(struct btree_trans *trans, - unsigned dev, s64 sectors) +int bch2_update_cached_sectors_list(struct btree_trans *trans, unsigned dev, s64 sectors) { struct bch_replicas_padded r; bch2_replicas_entry_cached(&r.e, dev); - update_replicas_list(trans, &r.e, sectors); -} - -#define do_mark_fn(fn, c, pos, flags, ...) \ -({ \ - int gc, ret = 0; \ - \ - percpu_rwsem_assert_held(&c->mark_lock); \ - \ - for (gc = 0; gc < 2 && !ret; gc++) \ - if (!gc == !(flags & BTREE_TRIGGER_GC) || \ - (gc && gc_visited(c, pos))) \ - ret = fn(c, __VA_ARGS__, gc); \ - ret; \ -}) - -void bch2_mark_alloc_bucket(struct bch_fs *c, struct bch_dev *ca, - size_t b, bool owned_by_allocator) -{ - struct bucket *g = bucket(ca, b); - struct bucket_mark old, new; - - old = bucket_cmpxchg(g, new, ({ - new.owned_by_allocator = owned_by_allocator; - })); - - BUG_ON(owned_by_allocator == old.owned_by_allocator); + return bch2_update_replicas_list(trans, &r.e, sectors); } -static int bch2_mark_alloc(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - u64 journal_seq, unsigned flags) +int bch2_mark_alloc(struct btree_trans *trans, + enum btree_id btree, unsigned level, + struct bkey_s_c old, struct bkey_s_c new, + unsigned flags) { bool gc = flags & BTREE_TRIGGER_GC; - struct bkey_alloc_unpacked u; + u64 journal_seq = trans->journal_res.seq; + u64 bucket_journal_seq; + struct bch_fs *c = trans->c; + struct bch_alloc_v4 old_a_convert, new_a_convert; + const struct bch_alloc_v4 *old_a, *new_a; struct bch_dev *ca; - struct bucket *g; - struct bucket_mark old_m, m; - - /* We don't do anything for deletions - do we?: */ - if (new.k->type != KEY_TYPE_alloc && - new.k->type != KEY_TYPE_alloc_v2) - return 0; + int ret = 0; /* * alloc btree is read in by bch2_alloc_read, not gc: @@ -546,340 +517,397 @@ static int bch2_mark_alloc(struct bch_fs *c, !(flags & BTREE_TRIGGER_BUCKET_INVALIDATE)) return 0; + if (bch2_trans_inconsistent_on(!bch2_dev_bucket_exists(c, new.k->p), trans, + "alloc key for invalid device or bucket")) + return -EIO; + ca = bch_dev_bkey_exists(c, new.k->p.inode); - if (new.k->p.offset >= ca->mi.nbuckets) - return 0; + old_a = bch2_alloc_to_v4(old, &old_a_convert); + new_a = bch2_alloc_to_v4(new, &new_a_convert); - g = __bucket(ca, new.k->p.offset, gc); - u = bch2_alloc_unpack(new); + bucket_journal_seq = new_a->journal_seq; - old_m = bucket_cmpxchg(g, m, ({ - m.gen = u.gen; - m.data_type = u.data_type; - m.dirty_sectors = u.dirty_sectors; - m.cached_sectors = u.cached_sectors; - m.stripe = u.stripe != 0; + if ((flags & BTREE_TRIGGER_INSERT) && + data_type_is_empty(old_a->data_type) != + data_type_is_empty(new_a->data_type) && + new.k->type == KEY_TYPE_alloc_v4) { + struct bch_alloc_v4 *v = (struct bch_alloc_v4 *) new.v; - if (journal_seq) { - m.journal_seq_valid = 1; - m.journal_seq = journal_seq; - } - })); + EBUG_ON(!journal_seq); - bch2_dev_usage_update(c, ca, old_m, m, journal_seq, gc); + /* + * If the btree updates referring to a bucket weren't flushed + * before the bucket became empty again, then the we don't have + * to wait on a journal flush before we can reuse the bucket: + */ + v->journal_seq = bucket_journal_seq = + data_type_is_empty(new_a->data_type) && + (journal_seq == v->journal_seq || + bch2_journal_noflush_seq(&c->journal, v->journal_seq)) + ? 0 : journal_seq; + } - g->io_time[READ] = u.read_time; - g->io_time[WRITE] = u.write_time; - g->oldest_gen = u.oldest_gen; - g->gen_valid = 1; - g->stripe = u.stripe; - g->stripe_redundancy = u.stripe_redundancy; + if (!data_type_is_empty(old_a->data_type) && + data_type_is_empty(new_a->data_type) && + bucket_journal_seq) { + ret = bch2_set_bucket_needs_journal_commit(&c->buckets_waiting_for_journal, + c->journal.flushed_seq_ondisk, + new.k->p.inode, new.k->p.offset, + bucket_journal_seq); + if (ret) { + bch2_fs_fatal_error(c, + "error setting bucket_needs_journal_commit: %i", ret); + return ret; + } + } - /* - * need to know if we're getting called from the invalidate path or - * not: - */ + percpu_down_read(&c->mark_lock); + if (!gc && new_a->gen != old_a->gen) + *bucket_gen(ca, new.k->p.offset) = new_a->gen; - if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) && - old_m.cached_sectors) { - if (update_cached_sectors(c, ca->dev_idx, -old_m.cached_sectors, - journal_seq, gc)) { - bch2_fs_fatal_error(c, "bch2_mark_alloc(): no replicas entry while updating cached sectors"); - return -1; - } + bch2_dev_usage_update(c, ca, *old_a, *new_a, journal_seq, gc); - trace_invalidate(ca, bucket_to_sector(ca, new.k->p.offset), - old_m.cached_sectors); - } + if (gc) { + struct bucket *g = gc_bucket(ca, new.k->p.offset); - return 0; -} + bucket_lock(g); -#define checked_add(a, b) \ -({ \ - unsigned _res = (unsigned) (a) + (b); \ - bool overflow = _res > U16_MAX; \ - if (overflow) \ - _res = U16_MAX; \ - (a) = _res; \ - overflow; \ -}) - -static int __bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca, - size_t b, enum bch_data_type data_type, - unsigned sectors, bool gc) -{ - struct bucket *g = __bucket(ca, b, gc); - struct bucket_mark old, new; - bool overflow; + g->gen_valid = 1; + g->gen = new_a->gen; + g->data_type = new_a->data_type; + g->stripe = new_a->stripe; + g->stripe_redundancy = new_a->stripe_redundancy; + g->dirty_sectors = new_a->dirty_sectors; + g->cached_sectors = new_a->cached_sectors; - BUG_ON(data_type != BCH_DATA_sb && - data_type != BCH_DATA_journal); + bucket_unlock(g); + } + percpu_up_read(&c->mark_lock); - old = bucket_cmpxchg(g, new, ({ - new.data_type = data_type; - overflow = checked_add(new.dirty_sectors, sectors); - })); + if (new_a->data_type == BCH_DATA_free && + (!new_a->journal_seq || new_a->journal_seq < c->journal.flushed_seq_ondisk)) + closure_wake_up(&c->freelist_wait); - bch2_fs_inconsistent_on(old.data_type && - old.data_type != data_type, c, - "different types of data in same bucket: %s, %s", - bch2_data_types[old.data_type], - bch2_data_types[data_type]); + if (new_a->data_type == BCH_DATA_need_discard && + (!bucket_journal_seq || bucket_journal_seq < c->journal.flushed_seq_ondisk)) + bch2_do_discards(c); - bch2_fs_inconsistent_on(overflow, c, - "bucket %u:%zu gen %u data type %s sector count overflow: %u + %u > U16_MAX", - ca->dev_idx, b, new.gen, - bch2_data_types[old.data_type ?: data_type], - old.dirty_sectors, sectors); + if (old_a->data_type != BCH_DATA_cached && + new_a->data_type == BCH_DATA_cached && + should_invalidate_buckets(ca, bch2_dev_usage_read(ca))) + bch2_do_invalidates(c); - if (c) - bch2_dev_usage_update(c, ca, old, new, 0, gc); + if (new_a->data_type == BCH_DATA_need_gc_gens) + bch2_do_gc_gens(c); return 0; } -void bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca, - size_t b, enum bch_data_type type, - unsigned sectors, struct gc_pos pos, - unsigned flags) +int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca, + size_t b, enum bch_data_type data_type, + unsigned sectors, struct gc_pos pos, + unsigned flags) { - BUG_ON(type != BCH_DATA_sb && - type != BCH_DATA_journal); + struct bucket old, new, *g; + int ret = 0; + + BUG_ON(!(flags & BTREE_TRIGGER_GC)); + BUG_ON(data_type != BCH_DATA_sb && + data_type != BCH_DATA_journal); /* * Backup superblock might be past the end of our normal usable space: */ if (b >= ca->mi.nbuckets) - return; + return 0; - if (likely(c)) { - do_mark_fn(__bch2_mark_metadata_bucket, c, pos, flags, - ca, b, type, sectors); - } else { - __bch2_mark_metadata_bucket(c, ca, b, type, sectors, 0); + percpu_down_read(&c->mark_lock); + g = gc_bucket(ca, b); + + bucket_lock(g); + old = *g; + + if (bch2_fs_inconsistent_on(g->data_type && + g->data_type != data_type, c, + "different types of data in same bucket: %s, %s", + bch2_data_types[g->data_type], + bch2_data_types[data_type])) { + ret = -EIO; + goto err; } -} -static s64 ptr_disk_sectors(s64 sectors, struct extent_ptr_decoded p) -{ - EBUG_ON(sectors < 0); + if (bch2_fs_inconsistent_on((u64) g->dirty_sectors + sectors > ca->mi.bucket_size, c, + "bucket %u:%zu gen %u data type %s sector count overflow: %u + %u > bucket size", + ca->dev_idx, b, g->gen, + bch2_data_types[g->data_type ?: data_type], + g->dirty_sectors, sectors)) { + ret = -EIO; + goto err; + } - return p.crc.compression_type && - p.crc.compression_type != BCH_COMPRESSION_TYPE_incompressible - ? DIV_ROUND_UP_ULL(sectors * p.crc.compressed_size, - p.crc.uncompressed_size) - : sectors; + g->data_type = data_type; + g->dirty_sectors += sectors; + new = *g; +err: + bucket_unlock(g); + if (!ret) + bch2_dev_usage_update_m(c, ca, old, new); + percpu_up_read(&c->mark_lock); + return ret; } -static int check_bucket_ref(struct bch_fs *c, struct bkey_s_c k, +static int check_bucket_ref(struct btree_trans *trans, + struct bkey_s_c k, const struct bch_extent_ptr *ptr, s64 sectors, enum bch_data_type ptr_data_type, - u8 bucket_gen, u8 bucket_data_type, - u16 dirty_sectors, u16 cached_sectors) + u8 b_gen, u8 bucket_data_type, + u32 bucket_sectors) { - size_t bucket_nr = PTR_BUCKET_NR(bch_dev_bkey_exists(c, ptr->dev), ptr); - u16 bucket_sectors = !ptr->cached - ? dirty_sectors - : cached_sectors; - char buf[200]; + struct bch_fs *c = trans->c; + struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); + size_t bucket_nr = PTR_BUCKET_NR(ca, ptr); + struct printbuf buf = PRINTBUF; + int ret = 0; + + if (bucket_data_type == BCH_DATA_cached) + bucket_data_type = BCH_DATA_user; - if (gen_after(ptr->gen, bucket_gen)) { + if ((bucket_data_type == BCH_DATA_stripe && ptr_data_type == BCH_DATA_user) || + (bucket_data_type == BCH_DATA_user && ptr_data_type == BCH_DATA_stripe)) + bucket_data_type = ptr_data_type = BCH_DATA_stripe; + + if (gen_after(ptr->gen, b_gen)) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, + BCH_FSCK_ERR_ptr_gen_newer_than_bucket_gen, "bucket %u:%zu gen %u data type %s: ptr gen %u newer than bucket gen\n" "while marking %s", - ptr->dev, bucket_nr, bucket_gen, + ptr->dev, bucket_nr, b_gen, bch2_data_types[bucket_data_type ?: ptr_data_type], ptr->gen, - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EIO; + (bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EIO; + goto err; } - if (gen_cmp(bucket_gen, ptr->gen) > BUCKET_GC_GEN_MAX) { + if (gen_cmp(b_gen, ptr->gen) > BUCKET_GC_GEN_MAX) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, + BCH_FSCK_ERR_ptr_too_stale, "bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n" "while marking %s", - ptr->dev, bucket_nr, bucket_gen, + ptr->dev, bucket_nr, b_gen, bch2_data_types[bucket_data_type ?: ptr_data_type], ptr->gen, - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EIO; + (printbuf_reset(&buf), + bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EIO; + goto err; } - if (bucket_gen != ptr->gen && !ptr->cached) { + if (b_gen != ptr->gen && !ptr->cached) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, - "bucket %u:%zu gen %u data type %s: stale dirty ptr (gen %u)\n" + BCH_FSCK_ERR_stale_dirty_ptr, + "bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n" "while marking %s", - ptr->dev, bucket_nr, bucket_gen, + ptr->dev, bucket_nr, b_gen, + *bucket_gen(ca, bucket_nr), bch2_data_types[bucket_data_type ?: ptr_data_type], ptr->gen, - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EIO; + (printbuf_reset(&buf), + bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EIO; + goto err; } - if (bucket_gen != ptr->gen) - return 1; + if (b_gen != ptr->gen) { + ret = 1; + goto out; + } - if (bucket_data_type && ptr_data_type && + if (!data_type_is_empty(bucket_data_type) && + ptr_data_type && bucket_data_type != ptr_data_type) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, + BCH_FSCK_ERR_ptr_bucket_data_type_mismatch, "bucket %u:%zu gen %u different types of data in same bucket: %s, %s\n" "while marking %s", - ptr->dev, bucket_nr, bucket_gen, + ptr->dev, bucket_nr, b_gen, bch2_data_types[bucket_data_type], bch2_data_types[ptr_data_type], - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EIO; + (printbuf_reset(&buf), + bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EIO; + goto err; } - if ((unsigned) (bucket_sectors + sectors) > U16_MAX) { + if ((u64) bucket_sectors + sectors > U32_MAX) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, - "bucket %u:%zu gen %u data type %s sector count overflow: %u + %lli > U16_MAX\n" + BCH_FSCK_ERR_bucket_sector_count_overflow, + "bucket %u:%zu gen %u data type %s sector count overflow: %u + %lli > U32_MAX\n" "while marking %s", - ptr->dev, bucket_nr, bucket_gen, + ptr->dev, bucket_nr, b_gen, bch2_data_types[bucket_data_type ?: ptr_data_type], bucket_sectors, sectors, - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EIO; + (printbuf_reset(&buf), + bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EIO; + goto err; } - - return 0; +out: + printbuf_exit(&buf); + return ret; +err: + bch2_dump_trans_updates(trans); + goto out; } -static int mark_stripe_bucket(struct bch_fs *c, struct bkey_s_c k, - unsigned ptr_idx, - u64 journal_seq, unsigned flags) +static int mark_stripe_bucket(struct btree_trans *trans, + struct bkey_s_c k, + unsigned ptr_idx, + unsigned flags) { + struct bch_fs *c = trans->c; const struct bch_stripe *s = bkey_s_c_to_stripe(k).v; unsigned nr_data = s->nr_blocks - s->nr_redundant; bool parity = ptr_idx >= nr_data; + enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe; + s64 sectors = parity ? le16_to_cpu(s->sectors) : 0; const struct bch_extent_ptr *ptr = s->ptrs + ptr_idx; - bool gc = flags & BTREE_TRIGGER_GC; struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); - struct bucket *g = PTR_BUCKET(ca, ptr, gc); - struct bucket_mark new, old; - char buf[200]; - int ret; + struct bucket old, new, *g; + struct printbuf buf = PRINTBUF; + int ret = 0; + + BUG_ON(!(flags & BTREE_TRIGGER_GC)); + + /* * XXX doesn't handle deletion */ + + percpu_down_read(&c->mark_lock); + g = PTR_GC_BUCKET(ca, ptr); - if (g->stripe && g->stripe != k.k->p.offset) { + if (g->dirty_sectors || + (g->stripe && g->stripe != k.k->p.offset)) { bch2_fs_inconsistent(c, "bucket %u:%zu gen %u: multiple stripes using same bucket\n%s", - ptr->dev, PTR_BUCKET_NR(ca, ptr), g->mark.gen, - (bch2_bkey_val_to_text(&PBUF(buf), c, k), buf)); - return -EINVAL; + ptr->dev, PTR_BUCKET_NR(ca, ptr), g->gen, + (bch2_bkey_val_to_text(&buf, c, k), buf.buf)); + ret = -EINVAL; + goto err; } - old = bucket_cmpxchg(g, new, ({ - ret = check_bucket_ref(c, k, ptr, 0, 0, new.gen, new.data_type, - new.dirty_sectors, new.cached_sectors); - if (ret) - return ret; + bucket_lock(g); + old = *g; - if (parity) { - new.data_type = BCH_DATA_parity; - new.dirty_sectors = le16_to_cpu(s->sectors); - } + ret = check_bucket_ref(trans, k, ptr, sectors, data_type, + g->gen, g->data_type, + g->dirty_sectors); + if (ret) + goto err; - if (journal_seq) { - new.journal_seq_valid = 1; - new.journal_seq = journal_seq; - } - })); + g->data_type = data_type; + g->dirty_sectors += sectors; g->stripe = k.k->p.offset; g->stripe_redundancy = s->nr_redundant; - - bch2_dev_usage_update(c, ca, old, new, journal_seq, gc); - return 0; + new = *g; +err: + bucket_unlock(g); + if (!ret) + bch2_dev_usage_update_m(c, ca, old, new); + percpu_up_read(&c->mark_lock); + printbuf_exit(&buf); + return ret; } -static int __mark_pointer(struct bch_fs *c, struct bkey_s_c k, +static int __mark_pointer(struct btree_trans *trans, + struct bkey_s_c k, const struct bch_extent_ptr *ptr, s64 sectors, enum bch_data_type ptr_data_type, u8 bucket_gen, u8 *bucket_data_type, - u16 *dirty_sectors, u16 *cached_sectors) + u32 *dirty_sectors, u32 *cached_sectors) { - u16 *dst_sectors = !ptr->cached + u32 *dst_sectors = !ptr->cached ? dirty_sectors : cached_sectors; - int ret = check_bucket_ref(c, k, ptr, sectors, ptr_data_type, - bucket_gen, *bucket_data_type, - *dirty_sectors, *cached_sectors); + int ret = check_bucket_ref(trans, k, ptr, sectors, ptr_data_type, + bucket_gen, *bucket_data_type, *dst_sectors); if (ret) return ret; *dst_sectors += sectors; - *bucket_data_type = *dirty_sectors || *cached_sectors - ? ptr_data_type : 0; + + if (!*dirty_sectors && !*cached_sectors) + *bucket_data_type = 0; + else if (*bucket_data_type != BCH_DATA_stripe) + *bucket_data_type = ptr_data_type; + return 0; } -static int bch2_mark_pointer(struct bch_fs *c, struct bkey_s_c k, +static int bch2_mark_pointer(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, struct extent_ptr_decoded p, - s64 sectors, enum bch_data_type data_type, - u64 journal_seq, unsigned flags) + s64 sectors, + unsigned flags) { - bool gc = flags & BTREE_TRIGGER_GC; - struct bucket_mark old, new; + struct bch_fs *c = trans->c; struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); - struct bucket *g = PTR_BUCKET(ca, &p.ptr, gc); + struct bucket old, new, *g; + enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p); u8 bucket_data_type; - u64 v; - int ret; - - v = atomic64_read(&g->_mark.v); - do { - new.v.counter = old.v.counter = v; - bucket_data_type = new.data_type; - - ret = __mark_pointer(c, k, &p.ptr, sectors, data_type, new.gen, - &bucket_data_type, - &new.dirty_sectors, - &new.cached_sectors); - if (ret) - return ret; - - new.data_type = bucket_data_type; - - if (journal_seq) { - new.journal_seq_valid = 1; - new.journal_seq = journal_seq; - } - - if (flags & BTREE_TRIGGER_NOATOMIC) { - g->_mark = new; - break; - } - } while ((v = atomic64_cmpxchg(&g->_mark.v, - old.v.counter, - new.v.counter)) != old.v.counter); + int ret = 0; - bch2_dev_usage_update(c, ca, old, new, journal_seq, gc); + BUG_ON(!(flags & BTREE_TRIGGER_GC)); - BUG_ON(!gc && bucket_became_unavailable(old, new)); + percpu_down_read(&c->mark_lock); + g = PTR_GC_BUCKET(ca, &p.ptr); + bucket_lock(g); + old = *g; + + bucket_data_type = g->data_type; + ret = __mark_pointer(trans, k, &p.ptr, sectors, + data_type, g->gen, + &bucket_data_type, + &g->dirty_sectors, + &g->cached_sectors); + if (!ret) + g->data_type = bucket_data_type; + + new = *g; + bucket_unlock(g); + if (!ret) + bch2_dev_usage_update_m(c, ca, old, new); + percpu_up_read(&c->mark_lock); - return 0; + return ret; } -static int bch2_mark_stripe_ptr(struct bch_fs *c, +static int bch2_mark_stripe_ptr(struct btree_trans *trans, + struct bkey_s_c k, struct bch_extent_stripe_ptr p, enum bch_data_type data_type, s64 sectors, - unsigned journal_seq, unsigned flags) + unsigned flags) { - bool gc = flags & BTREE_TRIGGER_GC; + struct bch_fs *c = trans->c; struct bch_replicas_padded r; - struct stripe *m; - unsigned i, blocks_nonempty = 0; + struct gc_stripe *m; + + BUG_ON(!(flags & BTREE_TRIGGER_GC)); - m = genradix_ptr(&c->stripes[gc], p.idx); + m = genradix_ptr_alloc(&c->gc_stripes, p.idx, GFP_KERNEL); + if (!m) { + bch_err(c, "error allocating memory for gc_stripes, idx %llu", + (u64) p.idx); + return -BCH_ERR_ENOMEM_mark_stripe_ptr; + } - spin_lock(&c->ec_stripes_heap_lock); + mutex_lock(&c->ec_stripes_heap_lock); if (!m || !m->alive) { - spin_unlock(&c->ec_stripes_heap_lock); + mutex_unlock(&c->ec_stripes_heap_lock); bch_err_ratelimited(c, "pointer to nonexistent stripe %llu", (u64) p.idx); bch2_inconsistent_error(c); @@ -889,30 +917,20 @@ static int bch2_mark_stripe_ptr(struct bch_fs *c, m->block_sectors[p.block] += sectors; r = m->r; - - for (i = 0; i < m->nr_blocks; i++) - blocks_nonempty += m->block_sectors[i] != 0; - - if (m->blocks_nonempty != blocks_nonempty) { - m->blocks_nonempty = blocks_nonempty; - if (!gc) - bch2_stripes_heap_update(c, m, p.idx); - } - - spin_unlock(&c->ec_stripes_heap_lock); + mutex_unlock(&c->ec_stripes_heap_lock); r.e.data_type = data_type; - update_replicas(c, &r.e, sectors, journal_seq, gc); + update_replicas(c, k, &r.e, sectors, trans->journal_res.seq, true); return 0; } -static int bch2_mark_extent(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - unsigned journal_seq, unsigned flags) +static int __mark_extent(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, unsigned flags) { - bool gc = flags & BTREE_TRIGGER_GC; - struct bkey_s_c k = flags & BTREE_TRIGGER_INSERT ? new : old; + u64 journal_seq = trans->journal_res.seq; + struct bch_fs *c = trans->c; struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); const union bch_extent_entry *entry; struct extent_ptr_decoded p; @@ -921,14 +939,13 @@ static int bch2_mark_extent(struct bch_fs *c, ? BCH_DATA_btree : BCH_DATA_user; s64 sectors = bkey_is_btree_ptr(k.k) - ? c->opts.btree_node_size + ? btree_sectors(c) : k.k->size; s64 dirty_sectors = 0; bool stale; int ret; - BUG_ON((flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)) == - (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)); + BUG_ON(!(flags & BTREE_TRIGGER_GC)); r.e.data_type = data_type; r.e.nr_devs = 0; @@ -940,27 +957,28 @@ static int bch2_mark_extent(struct bch_fs *c, if (flags & BTREE_TRIGGER_OVERWRITE) disk_sectors = -disk_sectors; - ret = bch2_mark_pointer(c, k, p, disk_sectors, data_type, - journal_seq, flags); + ret = bch2_mark_pointer(trans, btree_id, level, k, p, disk_sectors, flags); if (ret < 0) return ret; stale = ret > 0; if (p.ptr.cached) { - if (!stale) - if (update_cached_sectors(c, p.ptr.dev, disk_sectors, - journal_seq, gc)) { - bch2_fs_fatal_error(c, "bch2_mark_extent(): no replicas entry while updating cached sectors"); - return -1; - + if (!stale) { + ret = update_cached_sectors(c, k, p.ptr.dev, + disk_sectors, journal_seq, true); + if (ret) { + bch2_fs_fatal_error(c, "%s(): no replicas entry while updating cached sectors", + __func__); + return ret; } + } } else if (!p.has_ec) { dirty_sectors += disk_sectors; r.e.devs[r.e.nr_devs++] = p.ptr.dev; } else { - ret = bch2_mark_stripe_ptr(c, p.ec, data_type, - disk_sectors, journal_seq, flags); + ret = bch2_mark_stripe_ptr(trans, k, p.ec, data_type, + disk_sectors, flags); if (ret) return ret; @@ -974,372 +992,228 @@ static int bch2_mark_extent(struct bch_fs *c, } if (r.e.nr_devs) { - if (update_replicas(c, &r.e, dirty_sectors, journal_seq, gc)) { - char buf[200]; + ret = update_replicas(c, k, &r.e, dirty_sectors, journal_seq, true); + if (ret) { + struct printbuf buf = PRINTBUF; - bch2_bkey_val_to_text(&PBUF(buf), c, k); - bch2_fs_fatal_error(c, "no replicas entry for %s", buf); - return -1; + bch2_bkey_val_to_text(&buf, c, k); + bch2_fs_fatal_error(c, "%s(): no replicas entry for %s", __func__, buf.buf); + printbuf_exit(&buf); + return ret; } } return 0; } -static int bch2_mark_stripe(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - u64 journal_seq, unsigned flags) +int bch2_mark_extent(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_s_c new, + unsigned flags) +{ + return mem_trigger_run_overwrite_then_insert(__mark_extent, trans, btree_id, level, old, new, flags); +} + +int bch2_mark_stripe(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_s_c new, + unsigned flags) { bool gc = flags & BTREE_TRIGGER_GC; - size_t idx = new.k->p.offset; + u64 journal_seq = trans->journal_res.seq; + struct bch_fs *c = trans->c; + u64 idx = new.k->p.offset; const struct bch_stripe *old_s = old.k->type == KEY_TYPE_stripe ? bkey_s_c_to_stripe(old).v : NULL; const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe ? bkey_s_c_to_stripe(new).v : NULL; - struct stripe *m = genradix_ptr(&c->stripes[gc], idx); unsigned i; int ret; BUG_ON(gc && old_s); - if (!m || (old_s && !m->alive)) { - bch_err_ratelimited(c, "error marking nonexistent stripe %zu", - idx); - bch2_inconsistent_error(c); - return -1; - } + if (!gc) { + struct stripe *m = genradix_ptr(&c->stripes, idx); + + if (!m) { + struct printbuf buf1 = PRINTBUF; + struct printbuf buf2 = PRINTBUF; + + bch2_bkey_val_to_text(&buf1, c, old); + bch2_bkey_val_to_text(&buf2, c, new); + bch_err_ratelimited(c, "error marking nonexistent stripe %llu while marking\n" + "old %s\n" + "new %s", idx, buf1.buf, buf2.buf); + printbuf_exit(&buf2); + printbuf_exit(&buf1); + bch2_inconsistent_error(c); + return -1; + } - if (!new_s) { - spin_lock(&c->ec_stripes_heap_lock); - bch2_stripes_heap_del(c, m, idx); - spin_unlock(&c->ec_stripes_heap_lock); + if (!new_s) { + bch2_stripes_heap_del(c, m, idx); - memset(m, 0, sizeof(*m)); + memset(m, 0, sizeof(*m)); + } else { + m->sectors = le16_to_cpu(new_s->sectors); + m->algorithm = new_s->algorithm; + m->nr_blocks = new_s->nr_blocks; + m->nr_redundant = new_s->nr_redundant; + m->blocks_nonempty = 0; + + for (i = 0; i < new_s->nr_blocks; i++) + m->blocks_nonempty += !!stripe_blockcount_get(new_s, i); + + if (!old_s) + bch2_stripes_heap_insert(c, m, idx); + else + bch2_stripes_heap_update(c, m, idx); + } } else { + struct gc_stripe *m = + genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL); + + if (!m) { + bch_err(c, "error allocating memory for gc_stripes, idx %llu", + idx); + return -BCH_ERR_ENOMEM_mark_stripe; + } + /* + * This will be wrong when we bring back runtime gc: we should + * be unmarking the old key and then marking the new key + */ m->alive = true; m->sectors = le16_to_cpu(new_s->sectors); - m->algorithm = new_s->algorithm; m->nr_blocks = new_s->nr_blocks; m->nr_redundant = new_s->nr_redundant; - m->blocks_nonempty = 0; - - for (i = 0; i < new_s->nr_blocks; i++) { - m->block_sectors[i] = - stripe_blockcount_get(new_s, i); - m->blocks_nonempty += !!m->block_sectors[i]; + for (i = 0; i < new_s->nr_blocks; i++) m->ptrs[i] = new_s->ptrs[i]; - } bch2_bkey_to_replicas(&m->r.e, new); - if (!gc) { - spin_lock(&c->ec_stripes_heap_lock); - bch2_stripes_heap_update(c, m, idx); - spin_unlock(&c->ec_stripes_heap_lock); - } - } - - if (gc) { /* * gc recalculates this field from stripe ptr * references: */ memset(m->block_sectors, 0, sizeof(m->block_sectors)); - m->blocks_nonempty = 0; for (i = 0; i < new_s->nr_blocks; i++) { - ret = mark_stripe_bucket(c, new, i, journal_seq, flags); + ret = mark_stripe_bucket(trans, new, i, flags); if (ret) return ret; } - if (update_replicas(c, &m->r.e, - ((s64) m->sectors * m->nr_redundant), - journal_seq, gc)) { - char buf[200]; + ret = update_replicas(c, new, &m->r.e, + ((s64) m->sectors * m->nr_redundant), + journal_seq, gc); + if (ret) { + struct printbuf buf = PRINTBUF; - bch2_bkey_val_to_text(&PBUF(buf), c, new); - bch2_fs_fatal_error(c, "no replicas entry for %s", buf); - return -1; + bch2_bkey_val_to_text(&buf, c, new); + bch2_fs_fatal_error(c, "no replicas entry for %s", buf.buf); + printbuf_exit(&buf); + return ret; } } return 0; } -static int bch2_mark_inode(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - u64 journal_seq, unsigned flags) -{ - struct bch_fs_usage __percpu *fs_usage; - - preempt_disable(); - fs_usage = fs_usage_ptr(c, journal_seq, flags & BTREE_TRIGGER_GC); - fs_usage->nr_inodes += new.k->type == KEY_TYPE_inode; - fs_usage->nr_inodes -= old.k->type == KEY_TYPE_inode; - preempt_enable(); - return 0; -} - -static int bch2_mark_reservation(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - u64 journal_seq, unsigned flags) +static int __mark_reservation(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, unsigned flags) { - struct bkey_s_c k = flags & BTREE_TRIGGER_INSERT ? new : old; - struct bch_fs_usage __percpu *fs_usage; + struct bch_fs *c = trans->c; + struct bch_fs_usage *fs_usage; unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas; s64 sectors = (s64) k.k->size; + BUG_ON(!(flags & BTREE_TRIGGER_GC)); + if (flags & BTREE_TRIGGER_OVERWRITE) sectors = -sectors; sectors *= replicas; + percpu_down_read(&c->mark_lock); preempt_disable(); - fs_usage = fs_usage_ptr(c, journal_seq, flags & BTREE_TRIGGER_GC); + + fs_usage = fs_usage_ptr(c, trans->journal_res.seq, flags & BTREE_TRIGGER_GC); replicas = clamp_t(unsigned, replicas, 1, ARRAY_SIZE(fs_usage->persistent_reserved)); fs_usage->reserved += sectors; fs_usage->persistent_reserved[replicas - 1] += sectors; - preempt_enable(); - return 0; -} - -static s64 __bch2_mark_reflink_p(struct bch_fs *c, struct bkey_s_c_reflink_p p, - u64 idx, unsigned flags, size_t *r_idx) -{ - struct reflink_gc *r; - int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1; - s64 ret = 0; - - while (*r_idx < c->reflink_gc_nr) { - r = genradix_ptr(&c->reflink_gc_table, *r_idx); - BUG_ON(!r); - - if (idx < r->offset) - break; - (*r_idx)++; - } - - if (*r_idx >= c->reflink_gc_nr || - idx < r->offset - r->size) { - ret = p.k->size; - goto not_found; - } - - BUG_ON((s64) r->refcount + add < 0); - - r->refcount += add; - return r->offset - idx; -not_found: - if ((flags & BTREE_TRIGGER_GC) && - (flags & BTREE_TRIGGER_NOATOMIC)) { - /* - * XXX: we're replacing the entire reflink pointer with an error - * key, we should just be replacing the part that was missing: - */ - if (fsck_err(c, "%llu:%llu len %u points to nonexistent indirect extent %llu", - p.k->p.inode, p.k->p.offset, p.k->size, idx)) { - struct bkey_i_error *new; - - new = kmalloc(sizeof(*new), GFP_KERNEL); - if (!new) { - bch_err(c, "%s: error allocating new key", __func__); - return -ENOMEM; - } - - bkey_init(&new->k); - new->k.type = KEY_TYPE_error; - new->k.p = p.k->p; - new->k.size = p.k->size; - ret = bch2_journal_key_insert(c, BTREE_ID_extents, 0, &new->k_i); - - } - } else { - bch2_fs_inconsistent(c, - "%llu:%llu len %u points to nonexistent indirect extent %llu", - p.k->p.inode, p.k->p.offset, p.k->size, idx); - bch2_inconsistent_error(c); - ret = -EIO; - } -fsck_err: - return ret; -} - -static int bch2_mark_reflink_p(struct bch_fs *c, - struct bkey_s_c old, struct bkey_s_c new, - u64 journal_seq, unsigned flags) -{ - struct bkey_s_c k = flags & BTREE_TRIGGER_INSERT ? new : old; - struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k); - struct reflink_gc *ref; - size_t l, r, m; - u64 idx = le64_to_cpu(p.v->idx); - unsigned sectors = p.k->size; - s64 ret = 0; - - BUG_ON((flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)) == - (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)); - - l = 0; - r = c->reflink_gc_nr; - while (l < r) { - m = l + (r - l) / 2; - - ref = genradix_ptr(&c->reflink_gc_table, m); - if (ref->offset <= idx) - l = m + 1; - else - r = m; - } - - while (sectors) { - ret = __bch2_mark_reflink_p(c, p, idx, flags, &l); - if (ret <= 0) - return ret; - - ret = min_t(s64, ret, sectors); - idx += ret; - sectors -= ret; - } + preempt_enable(); + percpu_up_read(&c->mark_lock); return 0; } -static int bch2_mark_key_locked(struct bch_fs *c, - struct bkey_s_c old, - struct bkey_s_c new, - u64 journal_seq, unsigned flags) +int bch2_mark_reservation(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_s_c new, + unsigned flags) { - struct bkey_s_c k = flags & BTREE_TRIGGER_INSERT ? new : old; - - BUG_ON(!(flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE))); - - switch (k.k->type) { - case KEY_TYPE_alloc: - case KEY_TYPE_alloc_v2: - return bch2_mark_alloc(c, old, new, journal_seq, flags); - case KEY_TYPE_btree_ptr: - case KEY_TYPE_btree_ptr_v2: - case KEY_TYPE_extent: - case KEY_TYPE_reflink_v: - return bch2_mark_extent(c, old, new, journal_seq, flags); - case KEY_TYPE_stripe: - return bch2_mark_stripe(c, old, new, journal_seq, flags); - case KEY_TYPE_inode: - return bch2_mark_inode(c, old, new, journal_seq, flags); - case KEY_TYPE_reservation: - return bch2_mark_reservation(c, old, new, journal_seq, flags); - case KEY_TYPE_reflink_p: - return bch2_mark_reflink_p(c, old, new, journal_seq, flags); - case KEY_TYPE_snapshot: - return bch2_mark_snapshot(c, old, new, journal_seq, flags); - default: - return 0; - } + return mem_trigger_run_overwrite_then_insert(__mark_reservation, trans, btree_id, level, old, new, flags); } -int bch2_mark_key(struct bch_fs *c, struct bkey_s_c new, unsigned flags) +void bch2_trans_fs_usage_revert(struct btree_trans *trans, + struct replicas_delta_list *deltas) { - struct bkey deleted = KEY(0, 0, 0); - struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL }; - int ret; + struct bch_fs *c = trans->c; + struct bch_fs_usage *dst; + struct replicas_delta *d, *top = (void *) deltas->d + deltas->used; + s64 added = 0; + unsigned i; percpu_down_read(&c->mark_lock); - ret = bch2_mark_key_locked(c, old, new, 0, flags); - percpu_up_read(&c->mark_lock); - - return ret; -} - -int bch2_mark_update(struct btree_trans *trans, struct btree_path *path, - struct bkey_i *new, unsigned flags) -{ - struct bch_fs *c = trans->c; - struct bkey _deleted = KEY(0, 0, 0); - struct bkey_s_c deleted = (struct bkey_s_c) { &_deleted, NULL }; - struct bkey_s_c old; - struct bkey unpacked; - int ret; - - if (unlikely(flags & BTREE_TRIGGER_NORUN)) - return 0; - - if (!btree_node_type_needs_gc(path->btree_id)) - return 0; - - old = bch2_btree_path_peek_slot(path, &unpacked); + preempt_disable(); + dst = fs_usage_ptr(c, trans->journal_res.seq, false); - if (old.k->type == new->k.type && - ((1U << old.k->type) & BTREE_TRIGGER_WANTS_OLD_AND_NEW)) { - ret = bch2_mark_key_locked(c, old, bkey_i_to_s_c(new), - trans->journal_res.seq, - BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE|flags); - } else { - ret = bch2_mark_key_locked(c, deleted, bkey_i_to_s_c(new), - trans->journal_res.seq, - BTREE_TRIGGER_INSERT|flags) ?: - bch2_mark_key_locked(c, old, deleted, - trans->journal_res.seq, - BTREE_TRIGGER_OVERWRITE|flags); + /* revert changes: */ + for (d = deltas->d; d != top; d = replicas_delta_next(d)) { + switch (d->r.data_type) { + case BCH_DATA_btree: + case BCH_DATA_user: + case BCH_DATA_parity: + added += d->delta; + } + BUG_ON(__update_replicas(c, dst, &d->r, -d->delta)); } - return ret; -} - -static noinline __cold -void fs_usage_apply_warn(struct btree_trans *trans, - unsigned disk_res_sectors, - s64 should_not_have_added) -{ - struct bch_fs *c = trans->c; - struct btree_insert_entry *i; - char buf[200]; - - bch_err(c, "disk usage increased %lli more than %u sectors reserved", - should_not_have_added, disk_res_sectors); + dst->nr_inodes -= deltas->nr_inodes; - trans_for_each_update(trans, i) { - pr_err("while inserting"); - bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(i->k)); - pr_err("%s", buf); - pr_err("overlapping with"); - - if (!i->cached) { - struct bkey u; - struct bkey_s_c k = bch2_btree_path_peek_slot(i->path, &u); - - bch2_bkey_val_to_text(&PBUF(buf), c, k); - pr_err("%s", buf); - } else { - struct bkey_cached *ck = (void *) i->path->l[0].b; + for (i = 0; i < BCH_REPLICAS_MAX; i++) { + added -= deltas->persistent_reserved[i]; + dst->reserved -= deltas->persistent_reserved[i]; + dst->persistent_reserved[i] -= deltas->persistent_reserved[i]; + } - if (ck->valid) { - bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(ck->k)); - pr_err("%s", buf); - } - } + if (added > 0) { + trans->disk_res->sectors += added; + this_cpu_add(*c->online_reserved, added); } - __WARN(); + + preempt_enable(); + percpu_up_read(&c->mark_lock); } -void bch2_trans_fs_usage_apply(struct btree_trans *trans, - struct replicas_delta_list *deltas) +int bch2_trans_fs_usage_apply(struct btree_trans *trans, + struct replicas_delta_list *deltas) { struct bch_fs *c = trans->c; static int warned_disk_usage = 0; bool warn = false; - unsigned disk_res_sectors = trans->disk_res ? trans->disk_res->sectors : 0; - struct replicas_delta *d = deltas->d; + u64 disk_res_sectors = trans->disk_res ? trans->disk_res->sectors : 0; + struct replicas_delta *d, *d2; struct replicas_delta *top = (void *) deltas->d + deltas->used; struct bch_fs_usage *dst; s64 added = 0, should_not_have_added; unsigned i; - percpu_rwsem_assert_held(&c->mark_lock); - + percpu_down_read(&c->mark_lock); preempt_disable(); dst = fs_usage_ptr(c, trans->journal_res.seq, false); @@ -1351,7 +1225,8 @@ void bch2_trans_fs_usage_apply(struct btree_trans *trans, added += d->delta; } - BUG_ON(__update_replicas(c, dst, &d->r, d->delta)); + if (__update_replicas(c, dst, &d->r, d->delta)) + goto need_mark; } dst->nr_inodes += deltas->nr_inodes; @@ -1386,135 +1261,108 @@ void bch2_trans_fs_usage_apply(struct btree_trans *trans, } preempt_enable(); + percpu_up_read(&c->mark_lock); if (unlikely(warn) && !xchg(&warned_disk_usage, 1)) - fs_usage_apply_warn(trans, disk_res_sectors, should_not_have_added); + bch2_trans_inconsistent(trans, + "disk usage increased %lli more than %llu sectors reserved)", + should_not_have_added, disk_res_sectors); + return 0; +need_mark: + /* revert changes: */ + for (d2 = deltas->d; d2 != d; d2 = replicas_delta_next(d2)) + BUG_ON(__update_replicas(c, dst, &d2->r, -d2->delta)); + + preempt_enable(); + percpu_up_read(&c->mark_lock); + return -1; } /* trans_mark: */ -static struct bkey_alloc_buf * -bch2_trans_start_alloc_update(struct btree_trans *trans, struct btree_iter *iter, - const struct bch_extent_ptr *ptr, - struct bkey_alloc_unpacked *u) -{ - struct bch_fs *c = trans->c; - struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); - struct bpos pos = POS(ptr->dev, PTR_BUCKET_NR(ca, ptr)); - struct bucket *g; - struct bkey_alloc_buf *a; - struct bkey_i *update = btree_trans_peek_updates(trans, BTREE_ID_alloc, pos); - int ret; - - a = bch2_trans_kmalloc(trans, sizeof(struct bkey_alloc_buf)); - if (IS_ERR(a)) - return a; - - bch2_trans_iter_init(trans, iter, BTREE_ID_alloc, pos, - BTREE_ITER_CACHED| - BTREE_ITER_CACHED_NOFILL| - BTREE_ITER_INTENT); - ret = bch2_btree_iter_traverse(iter); - if (ret) { - bch2_trans_iter_exit(trans, iter); - return ERR_PTR(ret); - } - - if (update && !bpos_cmp(update->k.p, pos)) { - *u = bch2_alloc_unpack(bkey_i_to_s_c(update)); - } else { - percpu_down_read(&c->mark_lock); - g = bucket(ca, pos.offset); - *u = alloc_mem_to_key(iter, g, READ_ONCE(g->mark)); - percpu_up_read(&c->mark_lock); - } - - return a; -} - -static int bch2_trans_mark_pointer(struct btree_trans *trans, - struct bkey_s_c k, struct extent_ptr_decoded p, - s64 sectors, enum bch_data_type data_type) +static inline int bch2_trans_mark_pointer(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, struct extent_ptr_decoded p, + unsigned flags) { - struct bch_fs *c = trans->c; + bool insert = !(flags & BTREE_TRIGGER_OVERWRITE); struct btree_iter iter; - struct bkey_alloc_unpacked u; - struct bkey_alloc_buf *a; + struct bkey_i_alloc_v4 *a; + struct bpos bucket; + struct bch_backpointer bp; + s64 sectors; int ret; - a = bch2_trans_start_alloc_update(trans, &iter, &p.ptr, &u); + bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, &bucket, &bp); + sectors = bp.bucket_len; + if (!insert) + sectors = -sectors; + + a = bch2_trans_start_alloc_update(trans, &iter, bucket); if (IS_ERR(a)) return PTR_ERR(a); - ret = __mark_pointer(c, k, &p.ptr, sectors, data_type, u.gen, &u.data_type, - &u.dirty_sectors, &u.cached_sectors); + ret = __mark_pointer(trans, k, &p.ptr, sectors, bp.data_type, + a->v.gen, &a->v.data_type, + &a->v.dirty_sectors, &a->v.cached_sectors) ?: + bch2_trans_update(trans, &iter, &a->k_i, 0); + bch2_trans_iter_exit(trans, &iter); + if (ret) - goto out; + return ret; - bch2_alloc_pack(c, a, u); - bch2_trans_update(trans, &iter, &a->k, 0); -out: - bch2_trans_iter_exit(trans, &iter); - return ret; + if (!p.ptr.cached) { + ret = bch2_bucket_backpointer_mod(trans, bucket, bp, k, insert); + if (ret) + return ret; + } + + return 0; } static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans, struct extent_ptr_decoded p, s64 sectors, enum bch_data_type data_type) { - struct bch_fs *c = trans->c; struct btree_iter iter; - struct bkey_s_c k; struct bkey_i_stripe *s; struct bch_replicas_padded r; int ret = 0; - bch2_trans_iter_init(trans, &iter, BTREE_ID_stripes, POS(0, p.ec.idx), - BTREE_ITER_INTENT| - BTREE_ITER_WITH_UPDATES); - k = bch2_btree_iter_peek_slot(&iter); - ret = bkey_err(k); - if (ret) - goto err; - - if (k.k->type != KEY_TYPE_stripe) { - bch2_fs_inconsistent(c, + s = bch2_bkey_get_mut_typed(trans, &iter, + BTREE_ID_stripes, POS(0, p.ec.idx), + BTREE_ITER_WITH_UPDATES, stripe); + ret = PTR_ERR_OR_ZERO(s); + if (unlikely(ret)) { + bch2_trans_inconsistent_on(bch2_err_matches(ret, ENOENT), trans, "pointer to nonexistent stripe %llu", (u64) p.ec.idx); - bch2_inconsistent_error(c); - ret = -EIO; goto err; } - if (!bch2_ptr_matches_stripe(bkey_s_c_to_stripe(k).v, p)) { - bch2_fs_inconsistent(c, + if (!bch2_ptr_matches_stripe(&s->v, p)) { + bch2_trans_inconsistent(trans, "stripe pointer doesn't match stripe %llu", (u64) p.ec.idx); ret = -EIO; goto err; } - s = bch2_trans_kmalloc(trans, bkey_bytes(k.k)); - ret = PTR_ERR_OR_ZERO(s); - if (ret) - goto err; - - bkey_reassemble(&s->k_i, k); stripe_blockcount_set(&s->v, p.ec.block, stripe_blockcount_get(&s->v, p.ec.block) + sectors); - bch2_trans_update(trans, &iter, &s->k_i, 0); bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(&s->k_i)); r.e.data_type = data_type; - update_replicas_list(trans, &r.e, sectors); + ret = bch2_update_replicas_list(trans, &r.e, sectors); err: bch2_trans_iter_exit(trans, &iter); return ret; } -static int bch2_trans_mark_extent(struct btree_trans *trans, - struct bkey_s_c k, unsigned flags) +static int __trans_mark_extent(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, unsigned flags) { struct bch_fs *c = trans->c; struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); @@ -1525,14 +1373,11 @@ static int bch2_trans_mark_extent(struct btree_trans *trans, ? BCH_DATA_btree : BCH_DATA_user; s64 sectors = bkey_is_btree_ptr(k.k) - ? c->opts.btree_node_size + ? btree_sectors(c) : k.k->size; s64 dirty_sectors = 0; bool stale; - int ret; - - BUG_ON((flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)) == - (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)); + int ret = 0; r.e.data_type = data_type; r.e.nr_devs = 0; @@ -1544,17 +1389,19 @@ static int bch2_trans_mark_extent(struct btree_trans *trans, if (flags & BTREE_TRIGGER_OVERWRITE) disk_sectors = -disk_sectors; - ret = bch2_trans_mark_pointer(trans, k, p, - disk_sectors, data_type); + ret = bch2_trans_mark_pointer(trans, btree_id, level, k, p, flags); if (ret < 0) return ret; stale = ret > 0; if (p.ptr.cached) { - if (!stale) - update_cached_sectors_list(trans, p.ptr.dev, - disk_sectors); + if (!stale) { + ret = bch2_update_cached_sectors_list(trans, p.ptr.dev, + disk_sectors); + if (ret) + return ret; + } } else if (!p.has_ec) { dirty_sectors += disk_sectors; r.e.devs[r.e.nr_devs++] = p.ptr.dev; @@ -1569,151 +1416,200 @@ static int bch2_trans_mark_extent(struct btree_trans *trans, } if (r.e.nr_devs) - update_replicas_list(trans, &r.e, dirty_sectors); + ret = bch2_update_replicas_list(trans, &r.e, dirty_sectors); - return 0; + return ret; } -static int bch2_trans_mark_stripe_alloc_ref(struct btree_trans *trans, - struct bkey_s_c_stripe s, - unsigned idx, bool deleting) +int bch2_trans_mark_extent(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_i *new, + unsigned flags) +{ + struct bch_fs *c = trans->c; + int mod = (int) bch2_bkey_needs_rebalance(c, bkey_i_to_s_c(new)) - + (int) bch2_bkey_needs_rebalance(c, old); + + if (mod) { + int ret = bch2_btree_bit_mod(trans, BTREE_ID_rebalance_work, new->k.p, mod > 0); + if (ret) + return ret; + } + + return trigger_run_overwrite_then_insert(__trans_mark_extent, trans, btree_id, level, old, new, flags); +} + +static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans, + struct bkey_s_c_stripe s, + unsigned idx, bool deleting) { struct bch_fs *c = trans->c; const struct bch_extent_ptr *ptr = &s.v->ptrs[idx]; - struct bkey_alloc_buf *a; struct btree_iter iter; - struct bkey_alloc_unpacked u; - bool parity = idx >= s.v->nr_blocks - s.v->nr_redundant; + struct bkey_i_alloc_v4 *a; + enum bch_data_type data_type = idx >= s.v->nr_blocks - s.v->nr_redundant + ? BCH_DATA_parity : 0; + s64 sectors = data_type ? le16_to_cpu(s.v->sectors) : 0; int ret = 0; - a = bch2_trans_start_alloc_update(trans, &iter, ptr, &u); + if (deleting) + sectors = -sectors; + + a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(c, ptr)); if (IS_ERR(a)) return PTR_ERR(a); - if (parity) { - s64 sectors = le16_to_cpu(s.v->sectors); - - if (deleting) - sectors = -sectors; - - u.dirty_sectors += sectors; - u.data_type = u.dirty_sectors - ? BCH_DATA_parity - : 0; - } + ret = check_bucket_ref(trans, s.s_c, ptr, sectors, data_type, + a->v.gen, a->v.data_type, + a->v.dirty_sectors); + if (ret) + goto err; if (!deleting) { - if (bch2_fs_inconsistent_on(u.stripe && u.stripe != s.k->p.offset, c, - "bucket %llu:%llu gen %u: multiple stripes using same bucket (%u, %llu)", - iter.pos.inode, iter.pos.offset, u.gen, - u.stripe, s.k->p.offset)) { + if (bch2_trans_inconsistent_on(a->v.stripe || + a->v.stripe_redundancy, trans, + "bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)", + iter.pos.inode, iter.pos.offset, a->v.gen, + bch2_data_types[a->v.data_type], + a->v.dirty_sectors, + a->v.stripe, s.k->p.offset)) { + ret = -EIO; + goto err; + } + + if (bch2_trans_inconsistent_on(data_type && a->v.dirty_sectors, trans, + "bucket %llu:%llu gen %u data type %s dirty_sectors %u: data already in stripe bucket %llu", + iter.pos.inode, iter.pos.offset, a->v.gen, + bch2_data_types[a->v.data_type], + a->v.dirty_sectors, + s.k->p.offset)) { ret = -EIO; goto err; } - u.stripe = s.k->p.offset; - u.stripe_redundancy = s.v->nr_redundant; + a->v.stripe = s.k->p.offset; + a->v.stripe_redundancy = s.v->nr_redundant; + a->v.data_type = BCH_DATA_stripe; } else { - u.stripe = 0; - u.stripe_redundancy = 0; + if (bch2_trans_inconsistent_on(a->v.stripe != s.k->p.offset || + a->v.stripe_redundancy != s.v->nr_redundant, trans, + "bucket %llu:%llu gen %u: not marked as stripe when deleting stripe %llu (got %u)", + iter.pos.inode, iter.pos.offset, a->v.gen, + s.k->p.offset, a->v.stripe)) { + ret = -EIO; + goto err; + } + + a->v.stripe = 0; + a->v.stripe_redundancy = 0; + a->v.data_type = alloc_data_type(a->v, BCH_DATA_user); } - bch2_alloc_pack(c, a, u); - bch2_trans_update(trans, &iter, &a->k, 0); + a->v.dirty_sectors += sectors; + if (data_type) + a->v.data_type = !deleting ? data_type : 0; + + ret = bch2_trans_update(trans, &iter, &a->k_i, 0); + if (ret) + goto err; err: bch2_trans_iter_exit(trans, &iter); return ret; } -static int bch2_trans_mark_stripe(struct btree_trans *trans, - struct bkey_s_c old, struct bkey_s_c new, - unsigned flags) +int bch2_trans_mark_stripe(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_i *new, + unsigned flags) { - struct bkey_s_c_stripe old_s = { .k = NULL }; - struct bkey_s_c_stripe new_s = { .k = NULL }; + const struct bch_stripe *old_s = NULL; + struct bch_stripe *new_s = NULL; struct bch_replicas_padded r; - unsigned i; + unsigned i, nr_blocks; int ret = 0; if (old.k->type == KEY_TYPE_stripe) - old_s = bkey_s_c_to_stripe(old); - if (new.k->type == KEY_TYPE_stripe) - new_s = bkey_s_c_to_stripe(new); + old_s = bkey_s_c_to_stripe(old).v; + if (new->k.type == KEY_TYPE_stripe) + new_s = &bkey_i_to_stripe(new)->v; /* * If the pointers aren't changing, we don't need to do anything: */ - if (new_s.k && old_s.k && - new_s.v->nr_blocks == old_s.v->nr_blocks && - new_s.v->nr_redundant == old_s.v->nr_redundant && - !memcmp(old_s.v->ptrs, new_s.v->ptrs, - new_s.v->nr_blocks * sizeof(struct bch_extent_ptr))) + if (new_s && old_s && + new_s->nr_blocks == old_s->nr_blocks && + new_s->nr_redundant == old_s->nr_redundant && + !memcmp(old_s->ptrs, new_s->ptrs, + new_s->nr_blocks * sizeof(struct bch_extent_ptr))) return 0; - if (new_s.k) { - s64 sectors = le16_to_cpu(new_s.v->sectors); + BUG_ON(new_s && old_s && + (new_s->nr_blocks != old_s->nr_blocks || + new_s->nr_redundant != old_s->nr_redundant)); - bch2_bkey_to_replicas(&r.e, new); - update_replicas_list(trans, &r.e, sectors * new_s.v->nr_redundant); + nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks; - for (i = 0; i < new_s.v->nr_blocks; i++) { - ret = bch2_trans_mark_stripe_alloc_ref(trans, new_s, - i, false); - if (ret) - return ret; - } + if (new_s) { + s64 sectors = le16_to_cpu(new_s->sectors); + + bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(new)); + ret = bch2_update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant); + if (ret) + return ret; } - if (old_s.k) { - s64 sectors = -((s64) le16_to_cpu(old_s.v->sectors)); + if (old_s) { + s64 sectors = -((s64) le16_to_cpu(old_s->sectors)); bch2_bkey_to_replicas(&r.e, old); - update_replicas_list(trans, &r.e, sectors * old_s.v->nr_redundant); - - for (i = 0; i < old_s.v->nr_blocks; i++) { - ret = bch2_trans_mark_stripe_alloc_ref(trans, old_s, - i, true); - if (ret) - return ret; - } + ret = bch2_update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant); + if (ret) + return ret; } - return ret; -} + for (i = 0; i < nr_blocks; i++) { + if (new_s && old_s && + !memcmp(&new_s->ptrs[i], + &old_s->ptrs[i], + sizeof(new_s->ptrs[i]))) + continue; -static int bch2_trans_mark_inode(struct btree_trans *trans, - struct bkey_s_c old, - struct bkey_s_c new, - unsigned flags) -{ - int nr = (new.k->type == KEY_TYPE_inode) - - (old.k->type == KEY_TYPE_inode); + if (new_s) { + ret = bch2_trans_mark_stripe_bucket(trans, + bkey_i_to_s_c_stripe(new), i, false); + if (ret) + break; + } - if (nr) { - struct replicas_delta_list *d = - replicas_deltas_realloc(trans, 0); - d->nr_inodes += nr; + if (old_s) { + ret = bch2_trans_mark_stripe_bucket(trans, + bkey_s_c_to_stripe(old), i, true); + if (ret) + break; + } } - return 0; + return ret; } -static int bch2_trans_mark_reservation(struct btree_trans *trans, - struct bkey_s_c k, unsigned flags) +static int __trans_mark_reservation(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, unsigned flags) { unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas; s64 sectors = (s64) k.k->size; struct replicas_delta_list *d; - - BUG_ON((flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)) == - (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE)); + int ret; if (flags & BTREE_TRIGGER_OVERWRITE) sectors = -sectors; sectors *= replicas; - d = replicas_deltas_realloc(trans, 0); + ret = bch2_replicas_deltas_realloc(trans, 0); + if (ret) + return ret; + d = trans->fs_usage_deltas; replicas = clamp_t(unsigned, replicas, 1, ARRAY_SIZE(d->persistent_reserved)); @@ -1721,140 +1617,13 @@ static int bch2_trans_mark_reservation(struct btree_trans *trans, return 0; } -static int __bch2_trans_mark_reflink_p(struct btree_trans *trans, - struct bkey_s_c_reflink_p p, - u64 idx, unsigned flags) -{ - struct bch_fs *c = trans->c; - struct btree_iter iter; - struct bkey_s_c k; - struct bkey_i *n; - __le64 *refcount; - int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1; - s64 ret; - - bch2_trans_iter_init(trans, &iter, BTREE_ID_reflink, POS(0, idx), - BTREE_ITER_INTENT| - BTREE_ITER_WITH_UPDATES); - k = bch2_btree_iter_peek_slot(&iter); - ret = bkey_err(k); - if (ret) - goto err; - - n = bch2_trans_kmalloc(trans, bkey_bytes(k.k)); - ret = PTR_ERR_OR_ZERO(n); - if (ret) - goto err; - - bkey_reassemble(n, k); - - refcount = bkey_refcount(n); - if (!refcount) { - bch2_fs_inconsistent(c, - "%llu:%llu len %u points to nonexistent indirect extent %llu", - p.k->p.inode, p.k->p.offset, p.k->size, idx); - bch2_inconsistent_error(c); - ret = -EIO; - goto err; - } - - BUG_ON(!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)); - le64_add_cpu(refcount, add); - - if (!*refcount) { - n->k.type = KEY_TYPE_deleted; - set_bkey_val_u64s(&n->k, 0); - } - - bch2_btree_iter_set_pos_to_extent_start(&iter); - ret = bch2_trans_update(trans, &iter, n, 0); - if (ret) - goto err; - - ret = k.k->p.offset - idx; -err: - bch2_trans_iter_exit(trans, &iter); - return ret; -} - -static int bch2_trans_mark_reflink_p(struct btree_trans *trans, - struct bkey_s_c k, unsigned flags) -{ - struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k); - u64 idx = le64_to_cpu(p.v->idx); - unsigned sectors = p.k->size; - s64 ret = 0; - - while (sectors) { - ret = __bch2_trans_mark_reflink_p(trans, p, idx, flags); - if (ret < 0) - return ret; - - ret = min_t(s64, ret, sectors); - idx += ret; - sectors -= ret; - } - - return 0; -} - -int bch2_trans_mark_key(struct btree_trans *trans, struct bkey_s_c old, - struct bkey_s_c new, unsigned flags) +int bch2_trans_mark_reservation(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, + struct bkey_i *new, + unsigned flags) { - struct bkey_s_c k = flags & BTREE_TRIGGER_INSERT ? new : old; - - BUG_ON(!(flags & (BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE))); - - switch (k.k->type) { - case KEY_TYPE_btree_ptr: - case KEY_TYPE_btree_ptr_v2: - case KEY_TYPE_extent: - case KEY_TYPE_reflink_v: - return bch2_trans_mark_extent(trans, k, flags); - case KEY_TYPE_stripe: - return bch2_trans_mark_stripe(trans, old, new, flags); - case KEY_TYPE_inode: - return bch2_trans_mark_inode(trans, old, new, flags); - case KEY_TYPE_reservation: - return bch2_trans_mark_reservation(trans, k, flags); - case KEY_TYPE_reflink_p: - return bch2_trans_mark_reflink_p(trans, k, flags); - default: - return 0; - } -} - -int bch2_trans_mark_update(struct btree_trans *trans, - struct btree_path *path, - struct bkey_i *new, - unsigned flags) -{ - struct bkey _deleted = KEY(0, 0, 0); - struct bkey_s_c deleted = (struct bkey_s_c) { &_deleted, NULL }; - struct bkey_s_c old; - struct bkey unpacked; - int ret; - - if (unlikely(flags & BTREE_TRIGGER_NORUN)) - return 0; - - if (!btree_node_type_needs_gc(path->btree_id)) - return 0; - - old = bch2_btree_path_peek_slot(path, &unpacked); - - if (old.k->type == new->k.type && - ((1U << old.k->type) & BTREE_TRIGGER_WANTS_OLD_AND_NEW)) { - ret = bch2_trans_mark_key(trans, old, bkey_i_to_s_c(new), - BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE|flags); - } else { - ret = bch2_trans_mark_key(trans, deleted, bkey_i_to_s_c(new), - BTREE_TRIGGER_INSERT|flags) ?: - bch2_trans_mark_key(trans, old, deleted, - BTREE_TRIGGER_OVERWRITE|flags); - } - - return ret; + return trigger_run_overwrite_then_insert(__trans_mark_reservation, trans, btree_id, level, old, new, flags); } static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans, @@ -1864,12 +1633,7 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans, { struct bch_fs *c = trans->c; struct btree_iter iter; - struct bkey_alloc_unpacked u; - struct bkey_alloc_buf *a; - struct bch_extent_ptr ptr = { - .dev = ca->dev_idx, - .offset = bucket_to_sector(ca, b), - }; + struct bkey_i_alloc_v4 *a; int ret = 0; /* @@ -1878,28 +1642,30 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans, if (b >= ca->mi.nbuckets) return 0; - a = bch2_trans_start_alloc_update(trans, &iter, &ptr, &u); + a = bch2_trans_start_alloc_update(trans, &iter, POS(ca->dev_idx, b)); if (IS_ERR(a)) return PTR_ERR(a); - if (u.data_type && u.data_type != type) { + if (a->v.data_type && type && a->v.data_type != type) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, + BCH_FSCK_ERR_bucket_metadata_type_mismatch, "bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n" "while marking %s", - iter.pos.inode, iter.pos.offset, u.gen, - bch2_data_types[u.data_type], + iter.pos.inode, iter.pos.offset, a->v.gen, + bch2_data_types[a->v.data_type], bch2_data_types[type], bch2_data_types[type]); ret = -EIO; - goto out; + goto err; } - u.data_type = type; - u.dirty_sectors = sectors; - - bch2_alloc_pack(c, a, u); - bch2_trans_update(trans, &iter, &a->k, 0); -out: + if (a->v.data_type != type || + a->v.dirty_sectors != sectors) { + a->v.data_type = type; + a->v.dirty_sectors = sectors; + ret = bch2_trans_update(trans, &iter, &a->k_i, 0); + } +err: bch2_trans_iter_exit(trans, &iter); return ret; } @@ -1909,7 +1675,7 @@ int bch2_trans_mark_metadata_bucket(struct btree_trans *trans, enum bch_data_type type, unsigned sectors) { - return __bch2_trans_do(trans, NULL, NULL, 0, + return commit_do(trans, NULL, NULL, 0, __bch2_trans_mark_metadata_bucket(trans, ca, b, type, sectors)); } @@ -1987,15 +1753,30 @@ static int __bch2_trans_mark_dev_sb(struct btree_trans *trans, int bch2_trans_mark_dev_sb(struct bch_fs *c, struct bch_dev *ca) { - return bch2_trans_do(c, NULL, NULL, BTREE_INSERT_LAZY_RW, - __bch2_trans_mark_dev_sb(&trans, ca)); + int ret = bch2_trans_run(c, __bch2_trans_mark_dev_sb(trans, ca)); + + bch_err_fn(c, ret); + return ret; +} + +int bch2_trans_mark_dev_sbs(struct bch_fs *c) +{ + for_each_online_member(c, ca) { + int ret = bch2_trans_mark_dev_sb(c, ca); + if (ret) { + percpu_ref_put(&ca->ref); + return ret; + } + } + + return 0; } /* Disk reservations: */ #define SECTORS_CACHE 1024 -int bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res, +int __bch2_disk_reservation_add(struct bch_fs *c, struct disk_reservation *res, u64 sectors, int flags) { struct bch_fs_pcpu *pcpu; @@ -2048,7 +1829,7 @@ recalculate: ret = 0; } else { atomic64_set(&c->sectors_available, sectors_available); - ret = -ENOSPC; + ret = -BCH_ERR_ENOSPC_disk_reservation; } mutex_unlock(&c->sectors_available_lock); @@ -2059,56 +1840,37 @@ recalculate: /* Startup/shutdown: */ -static void buckets_free_rcu(struct rcu_head *rcu) +static void bucket_gens_free_rcu(struct rcu_head *rcu) { - struct bucket_array *buckets = - container_of(rcu, struct bucket_array, rcu); + struct bucket_gens *buckets = + container_of(rcu, struct bucket_gens, rcu); - kvpfree(buckets, - sizeof(struct bucket_array) + - buckets->nbuckets * sizeof(struct bucket)); + kvpfree(buckets, sizeof(*buckets) + buckets->nbuckets); } int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets) { - struct bucket_array *buckets = NULL, *old_buckets = NULL; + struct bucket_gens *bucket_gens = NULL, *old_bucket_gens = NULL; unsigned long *buckets_nouse = NULL; - alloc_fifo free[RESERVE_NR]; - alloc_fifo free_inc; - alloc_heap alloc_heap; - - size_t btree_reserve = DIV_ROUND_UP(BTREE_NODE_RESERVE, - ca->mi.bucket_size / c->opts.btree_node_size); - /* XXX: these should be tunable */ - size_t reserve_none = max_t(size_t, 1, nbuckets >> 9); - size_t copygc_reserve = max_t(size_t, 2, nbuckets >> 6); - size_t free_inc_nr = max(max_t(size_t, 1, nbuckets >> 12), - btree_reserve * 2); - bool resize = ca->buckets[0] != NULL; - int ret = -ENOMEM; - unsigned i; + bool resize = ca->bucket_gens != NULL; + int ret; - memset(&free, 0, sizeof(free)); - memset(&free_inc, 0, sizeof(free_inc)); - memset(&alloc_heap, 0, sizeof(alloc_heap)); + if (!(bucket_gens = kvpmalloc(sizeof(struct bucket_gens) + nbuckets, + GFP_KERNEL|__GFP_ZERO))) { + ret = -BCH_ERR_ENOMEM_bucket_gens; + goto err; + } - if (!(buckets = kvpmalloc(sizeof(struct bucket_array) + - nbuckets * sizeof(struct bucket), - GFP_KERNEL|__GFP_ZERO)) || - !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) * + if ((c->opts.buckets_nouse && + !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) * sizeof(unsigned long), - GFP_KERNEL|__GFP_ZERO)) || - !init_fifo(&free[RESERVE_MOVINGGC], - copygc_reserve, GFP_KERNEL) || - !init_fifo(&free[RESERVE_NONE], reserve_none, GFP_KERNEL) || - !init_fifo(&free_inc, free_inc_nr, GFP_KERNEL) || - !init_heap(&alloc_heap, ALLOC_SCAN_BATCH(ca) << 1, GFP_KERNEL)) + GFP_KERNEL|__GFP_ZERO)))) { + ret = -BCH_ERR_ENOMEM_buckets_nouse; goto err; + } - buckets->first_bucket = ca->mi.first_bucket; - buckets->nbuckets = nbuckets; - - bch2_copygc_stop(c); + bucket_gens->first_bucket = ca->mi.first_bucket; + bucket_gens->nbuckets = nbuckets; if (resize) { down_write(&c->gc_lock); @@ -2116,56 +1878,39 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets) percpu_down_write(&c->mark_lock); } - old_buckets = bucket_array(ca); + old_bucket_gens = rcu_dereference_protected(ca->bucket_gens, 1); if (resize) { - size_t n = min(buckets->nbuckets, old_buckets->nbuckets); - - memcpy(buckets->b, - old_buckets->b, - n * sizeof(struct bucket)); - memcpy(buckets_nouse, - ca->buckets_nouse, - BITS_TO_LONGS(n) * sizeof(unsigned long)); + size_t n = min(bucket_gens->nbuckets, old_bucket_gens->nbuckets); + + memcpy(bucket_gens->b, + old_bucket_gens->b, + n); + if (buckets_nouse) + memcpy(buckets_nouse, + ca->buckets_nouse, + BITS_TO_LONGS(n) * sizeof(unsigned long)); } - rcu_assign_pointer(ca->buckets[0], buckets); - buckets = old_buckets; + rcu_assign_pointer(ca->bucket_gens, bucket_gens); + bucket_gens = old_bucket_gens; swap(ca->buckets_nouse, buckets_nouse); + nbuckets = ca->mi.nbuckets; + if (resize) { percpu_up_write(&c->mark_lock); + up_write(&ca->bucket_lock); up_write(&c->gc_lock); } - spin_lock(&c->freelist_lock); - for (i = 0; i < RESERVE_NR; i++) { - fifo_move(&free[i], &ca->free[i]); - swap(ca->free[i], free[i]); - } - fifo_move(&free_inc, &ca->free_inc); - swap(ca->free_inc, free_inc); - spin_unlock(&c->freelist_lock); - - /* with gc lock held, alloc_heap can't be in use: */ - swap(ca->alloc_heap, alloc_heap); - - nbuckets = ca->mi.nbuckets; - - if (resize) - up_write(&ca->bucket_lock); - ret = 0; err: - free_heap(&alloc_heap); - free_fifo(&free_inc); - for (i = 0; i < RESERVE_NR; i++) - free_fifo(&free[i]); kvpfree(buckets_nouse, BITS_TO_LONGS(nbuckets) * sizeof(unsigned long)); - if (buckets) - call_rcu(&old_buckets->rcu, buckets_free_rcu); + if (bucket_gens) + call_rcu(&bucket_gens->rcu, bucket_gens_free_rcu); return ret; } @@ -2174,15 +1919,10 @@ void bch2_dev_buckets_free(struct bch_dev *ca) { unsigned i; - free_heap(&ca->alloc_heap); - free_fifo(&ca->free_inc); - for (i = 0; i < RESERVE_NR; i++) - free_fifo(&ca->free[i]); kvpfree(ca->buckets_nouse, BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long)); - kvpfree(rcu_dereference_protected(ca->buckets[0], 1), - sizeof(struct bucket_array) + - ca->mi.nbuckets * sizeof(struct bucket)); + kvpfree(rcu_dereference_protected(ca->bucket_gens, 1), + sizeof(struct bucket_gens) + ca->mi.nbuckets); for (i = 0; i < ARRAY_SIZE(ca->usage); i++) free_percpu(ca->usage[i]); @@ -2195,13 +1935,13 @@ int bch2_dev_buckets_alloc(struct bch_fs *c, struct bch_dev *ca) ca->usage_base = kzalloc(sizeof(struct bch_dev_usage), GFP_KERNEL); if (!ca->usage_base) - return -ENOMEM; + return -BCH_ERR_ENOMEM_usage_init; for (i = 0; i < ARRAY_SIZE(ca->usage); i++) { ca->usage[i] = alloc_percpu(struct bch_dev_usage); if (!ca->usage[i]) - return -ENOMEM; + return -BCH_ERR_ENOMEM_usage_init; } - return bch2_dev_buckets_resize(c, ca, ca->mi.nbuckets);; + return bch2_dev_buckets_resize(c, ca, ca->mi.nbuckets); }