X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbuckets.c;h=c0dac04253f7c8787570fdb50f137b0ae20ba26a;hb=e962ffb76959b40016b8b26e366bcb30f4a40bdb;hp=b4be2122c2d5eb72f8d6112e1a5d7f3696cab97d;hpb=ddac1641ee1e2686c2211a8d671ea723634dfc89;p=bcachefs-tools-debian diff --git a/libbcachefs/buckets.c b/libbcachefs/buckets.c index b4be212..c0dac04 100644 --- a/libbcachefs/buckets.c +++ b/libbcachefs/buckets.c @@ -21,9 +21,9 @@ #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, @@ -47,27 +47,23 @@ static inline void fs_usage_data_type_to_base(struct bch_fs_usage *fs_usage, 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 + @@ -89,32 +85,17 @@ static inline struct bch_dev_usage *dev_usage_ptr(struct bch_dev *ca, : 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) -{ - percpu_rwsem_assert_held(&c->mark_lock); - BUG_ON(!gc && !journal_seq); - - 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) @@ -140,25 +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; @@ -166,8 +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)); @@ -179,7 +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, @@ -222,7 +209,7 @@ void bch2_fs_usage_to_text(struct printbuf *out, } 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); prt_printf(out, "\t"); @@ -285,12 +272,28 @@ void bch2_dev_usage_init(struct bch_dev *ca) ca->usage_base->d[BCH_DATA_free].buckets = ca->mi.nbuckets - ca->mi.first_bucket; } -static inline int bucket_sectors_fragmented(struct bch_dev *ca, - struct bch_alloc_v4 a) +void bch2_dev_usage_to_text(struct printbuf *out, struct bch_dev_usage *usage) { - return a.dirty_sectors - ? max(0, (int) ca->mi.bucket_size - (int) a.dirty_sectors) - : 0; + 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); + + 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, @@ -314,46 +317,41 @@ static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca, u->d[old.data_type].buckets--; u->d[new.data_type].buckets++; - u->buckets_ec -= (int) !!old.stripe; - u->buckets_ec += (int) !!new.stripe; - - u->d[old.data_type].sectors -= old.dirty_sectors; - u->d[new.data_type].sectors += new.dirty_sectors; + u->d[old.data_type].sectors -= bch2_bucket_sectors_dirty(old); + u->d[new.data_type].sectors += bch2_bucket_sectors_dirty(new); 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 void bch2_dev_usage_update_m(struct bch_fs *c, struct bch_dev *ca, - struct bucket old, struct bucket new, - u64 journal_seq, bool gc) +static inline struct bch_alloc_v4 bucket_m_to_alloc(struct bucket b) { - struct bch_alloc_v4 old_a = { - .gen = old.gen, - .data_type = old.data_type, - .dirty_sectors = old.dirty_sectors, - .cached_sectors = old.cached_sectors, - .stripe = old.stripe, - }; - struct bch_alloc_v4 new_a = { - .gen = new.gen, - .data_type = new.data_type, - .dirty_sectors = new.dirty_sectors, - .cached_sectors = new.cached_sectors, - .stripe = new.stripe, + 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, }; +} - bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, gc); +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); @@ -367,20 +365,19 @@ static inline int __update_replicas(struct bch_fs *c, } static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k, - struct bch_replicas_entry *r, s64 sectors, + struct bch_replicas_entry_v1 *r, s64 sectors, unsigned journal_seq, bool gc) { - struct bch_fs_usage __percpu *fs_usage; + struct bch_fs_usage *fs_usage; int idx, ret = 0; struct printbuf buf = PRINTBUF; percpu_down_read(&c->mark_lock); - buf.atomic++; idx = bch2_replicas_entry_idx(c, r); if (idx < 0 && - fsck_err(c, "no replicas entry\n" - " while marking %s", + 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); @@ -419,8 +416,8 @@ static inline int update_cached_sectors(struct bch_fs *c, 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; @@ -429,12 +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) @@ -448,49 +449,64 @@ 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((void *) n + offsetof(struct replicas_delta, 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); + return bch2_update_replicas_list(trans, &r.e, sectors); } 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; u64 journal_seq = trans->journal_res.seq; + u64 bucket_journal_seq; struct bch_fs *c = trans->c; - struct bch_alloc_v4 old_a, new_a; + struct bch_alloc_v4 old_a_convert, new_a_convert; + const struct bch_alloc_v4 *old_a, *new_a; struct bch_dev *ca; int ret = 0; @@ -507,36 +523,38 @@ int bch2_mark_alloc(struct btree_trans *trans, ca = bch_dev_bkey_exists(c, new.k->p.inode); - bch2_alloc_to_v4(old, &old_a); - bch2_alloc_to_v4(new, &new_a); + old_a = bch2_alloc_to_v4(old, &old_a_convert); + new_a = bch2_alloc_to_v4(new, &new_a_convert); + + bucket_journal_seq = new_a->journal_seq; if ((flags & BTREE_TRIGGER_INSERT) && - data_type_is_empty(old_a.data_type) != - data_type_is_empty(new_a.data_type) && + 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; - BUG_ON(!journal_seq); + EBUG_ON(!journal_seq); /* * 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: */ - new_a.journal_seq = data_type_is_empty(new_a.data_type) && + 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; - v->journal_seq = new_a.journal_seq; } - if (!data_type_is_empty(old_a.data_type) && - data_type_is_empty(new_a.data_type) && - new_a.journal_seq) { + 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, - new_a.journal_seq); + bucket_journal_seq); if (ret) { bch2_fs_fatal_error(c, "error setting bucket_needs_journal_commit: %i", ret); @@ -545,10 +563,10 @@ int bch2_mark_alloc(struct btree_trans *trans, } 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 (!gc && new_a->gen != old_a->gen) + *bucket_gen(ca, new.k->p.offset) = new_a->gen; - bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, gc); + bch2_dev_usage_update(c, ca, *old_a, *new_a, journal_seq, gc); if (gc) { struct bucket *g = gc_bucket(ca, new.k->p.offset); @@ -556,47 +574,31 @@ int bch2_mark_alloc(struct btree_trans *trans, bucket_lock(g); 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; + 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; bucket_unlock(g); } percpu_up_read(&c->mark_lock); - /* - * need to know if we're getting called from the invalidate path or - * not: - */ - - if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) && - old_a.cached_sectors) { - ret = update_cached_sectors(c, new, ca->dev_idx, - -old_a.cached_sectors, - journal_seq, gc); - if (ret) { - bch2_fs_fatal_error(c, "bch2_mark_alloc(): no replicas entry while updating cached sectors"); - return ret; - } - } - - if (new_a.data_type == BCH_DATA_free && - (!new_a.journal_seq || new_a.journal_seq < c->journal.flushed_seq_ondisk)) + 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); - if (new_a.data_type == BCH_DATA_need_discard && - (!new_a.journal_seq || new_a.journal_seq < c->journal.flushed_seq_ondisk)) + if (new_a->data_type == BCH_DATA_need_discard && + (!bucket_journal_seq || bucket_journal_seq < c->journal.flushed_seq_ondisk)) bch2_do_discards(c); - if (old_a.data_type != BCH_DATA_cached && - new_a.data_type == BCH_DATA_cached && + 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 (new_a.data_type == BCH_DATA_need_gc_gens) + if (new_a->data_type == BCH_DATA_need_gc_gens) bch2_do_gc_gens(c); return 0; @@ -644,38 +646,40 @@ int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca, goto err; } - 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, 0, true); + 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, +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 b_gen, u8 bucket_data_type, - u32 dirty_sectors, u32 cached_sectors) + u32 bucket_sectors) { + 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); - u16 bucket_sectors = !ptr->cached - ? dirty_sectors - : cached_sectors; struct printbuf buf = PRINTBUF; int ret = 0; if (bucket_data_type == BCH_DATA_cached) bucket_data_type = BCH_DATA_user; + 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, b_gen, @@ -688,6 +692,7 @@ static int check_bucket_ref(struct bch_fs *c, 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, b_gen, @@ -701,6 +706,7 @@ static int check_bucket_ref(struct bch_fs *c, if (b_gen != ptr->gen && !ptr->cached) { bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK, + 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, b_gen, @@ -715,13 +721,14 @@ static int check_bucket_ref(struct bch_fs *c, if (b_gen != ptr->gen) { ret = 1; - goto err; + goto out; } 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, b_gen, @@ -733,9 +740,10 @@ static int check_bucket_ref(struct bch_fs *c, goto err; } - if ((unsigned) (bucket_sectors + sectors) > U32_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, b_gen, bch2_data_types[bucket_data_type ?: ptr_data_type], @@ -745,9 +753,12 @@ static int check_bucket_ref(struct bch_fs *c, ret = -EIO; goto err; } -err: +out: printbuf_exit(&buf); return ret; +err: + bch2_dump_trans_updates(trans); + goto out; } static int mark_stripe_bucket(struct btree_trans *trans, @@ -756,11 +767,10 @@ static int mark_stripe_bucket(struct btree_trans *trans, unsigned flags) { struct bch_fs *c = trans->c; - u64 journal_seq = trans->journal_res.seq; 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 : 0; + 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; struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); @@ -773,7 +783,6 @@ static int mark_stripe_bucket(struct btree_trans *trans, /* * XXX doesn't handle deletion */ percpu_down_read(&c->mark_lock); - buf.atomic++; g = PTR_GC_BUCKET(ca, ptr); if (g->dirty_sectors || @@ -789,14 +798,13 @@ static int mark_stripe_bucket(struct btree_trans *trans, bucket_lock(g); old = *g; - ret = check_bucket_ref(c, k, ptr, sectors, data_type, + ret = check_bucket_ref(trans, k, ptr, sectors, data_type, g->gen, g->data_type, - g->dirty_sectors, g->cached_sectors); + g->dirty_sectors); if (ret) goto err; - if (data_type) - g->data_type = data_type; + g->data_type = data_type; g->dirty_sectors += sectors; g->stripe = k.k->p.offset; @@ -805,7 +813,7 @@ static int mark_stripe_bucket(struct btree_trans *trans, err: bucket_unlock(g); if (!ret) - bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true); + bch2_dev_usage_update_m(c, ca, old, new); percpu_up_read(&c->mark_lock); printbuf_exit(&buf); return ret; @@ -821,29 +829,33 @@ static int __mark_pointer(struct btree_trans *trans, u32 *dst_sectors = !ptr->cached ? dirty_sectors : cached_sectors; - int ret = check_bucket_ref(trans->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 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, + s64 sectors, unsigned flags) { - u64 journal_seq = trans->journal_res.seq; struct bch_fs *c = trans->c; struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); struct bucket old, new, *g; + enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p); u8 bucket_data_type; int ret = 0; @@ -866,7 +878,7 @@ static int bch2_mark_pointer(struct btree_trans *trans, new = *g; bucket_unlock(g); if (!ret) - bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true); + bch2_dev_usage_update_m(c, ca, old, new); percpu_up_read(&c->mark_lock); return ret; @@ -889,13 +901,13 @@ static int bch2_mark_stripe_ptr(struct btree_trans *trans, if (!m) { bch_err(c, "error allocating memory for gc_stripes, idx %llu", (u64) p.idx); - return -ENOMEM; + 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); @@ -905,7 +917,7 @@ static int bch2_mark_stripe_ptr(struct btree_trans *trans, m->block_sectors[p.block] += sectors; r = m->r; - spin_unlock(&c->ec_stripes_heap_lock); + mutex_unlock(&c->ec_stripes_heap_lock); r.e.data_type = data_type; update_replicas(c, k, &r.e, sectors, trans->journal_res.seq, true); @@ -913,13 +925,12 @@ static int bch2_mark_stripe_ptr(struct btree_trans *trans, return 0; } -int bch2_mark_extent(struct btree_trans *trans, - struct bkey_s_c old, struct bkey_s_c new, - unsigned flags) +static int __mark_extent(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c k, unsigned flags) { u64 journal_seq = trans->journal_res.seq; struct bch_fs *c = trans->c; - struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new; struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); const union bch_extent_entry *entry; struct extent_ptr_decoded p; @@ -946,8 +957,7 @@ int bch2_mark_extent(struct btree_trans *trans, if (flags & BTREE_TRIGGER_OVERWRITE) disk_sectors = -disk_sectors; - ret = bch2_mark_pointer(trans, k, p, disk_sectors, - data_type, flags); + ret = bch2_mark_pointer(trans, btree_id, level, k, p, disk_sectors, flags); if (ret < 0) return ret; @@ -958,7 +968,8 @@ int bch2_mark_extent(struct btree_trans *trans, ret = update_cached_sectors(c, k, p.ptr.dev, disk_sectors, journal_seq, true); if (ret) { - bch2_fs_fatal_error(c, "bch2_mark_extent(): no replicas entry while updating cached sectors"); + bch2_fs_fatal_error(c, "%s(): no replicas entry while updating cached sectors", + __func__); return ret; } } @@ -986,7 +997,7 @@ int bch2_mark_extent(struct btree_trans *trans, struct printbuf buf = PRINTBUF; bch2_bkey_val_to_text(&buf, c, k); - bch2_fs_fatal_error(c, "no replicas entry for %s", buf.buf); + bch2_fs_fatal_error(c, "%s(): no replicas entry for %s", __func__, buf.buf); printbuf_exit(&buf); return ret; } @@ -995,7 +1006,16 @@ int bch2_mark_extent(struct btree_trans *trans, return 0; } +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) { @@ -1015,7 +1035,7 @@ int bch2_mark_stripe(struct btree_trans *trans, if (!gc) { struct stripe *m = genradix_ptr(&c->stripes, idx); - if (!m || (old_s && !m->alive)) { + if (!m) { struct printbuf buf1 = PRINTBUF; struct printbuf buf2 = PRINTBUF; @@ -1031,13 +1051,10 @@ int bch2_mark_stripe(struct btree_trans *trans, } if (!new_s) { - spin_lock(&c->ec_stripes_heap_lock); bch2_stripes_heap_del(c, m, idx); - spin_unlock(&c->ec_stripes_heap_lock); memset(m, 0, sizeof(*m)); } else { - m->alive = true; m->sectors = le16_to_cpu(new_s->sectors); m->algorithm = new_s->algorithm; m->nr_blocks = new_s->nr_blocks; @@ -1047,9 +1064,10 @@ int bch2_mark_stripe(struct btree_trans *trans, for (i = 0; i < new_s->nr_blocks; i++) m->blocks_nonempty += !!stripe_blockcount_get(new_s, i); - spin_lock(&c->ec_stripes_heap_lock); - bch2_stripes_heap_update(c, m, idx); - spin_unlock(&c->ec_stripes_heap_lock); + if (!old_s) + bch2_stripes_heap_insert(c, m, idx); + else + bch2_stripes_heap_update(c, m, idx); } } else { struct gc_stripe *m = @@ -1058,7 +1076,7 @@ int bch2_mark_stripe(struct btree_trans *trans, if (!m) { bch_err(c, "error allocating memory for gc_stripes, idx %llu", idx); - return -ENOMEM; + return -BCH_ERR_ENOMEM_mark_stripe; } /* * This will be wrong when we bring back runtime gc: we should @@ -1102,44 +1120,12 @@ int bch2_mark_stripe(struct btree_trans *trans, return 0; } -int bch2_mark_inode(struct btree_trans *trans, - struct bkey_s_c old, struct bkey_s_c new, - unsigned flags) -{ - struct bch_fs *c = trans->c; - struct bch_fs_usage __percpu *fs_usage; - u64 journal_seq = trans->journal_res.seq; - - if (flags & BTREE_TRIGGER_INSERT) { - struct bch_inode_v2 *v = (struct bch_inode_v2 *) new.v; - - BUG_ON(!journal_seq); - BUG_ON(new.k->type != KEY_TYPE_inode_v2); - - v->bi_journal_seq = cpu_to_le64(journal_seq); - } - - if (flags & BTREE_TRIGGER_GC) { - percpu_down_read(&c->mark_lock); - preempt_disable(); - - fs_usage = fs_usage_ptr(c, journal_seq, flags & BTREE_TRIGGER_GC); - fs_usage->nr_inodes += bkey_is_inode(new.k); - fs_usage->nr_inodes -= bkey_is_inode(old.k); - - preempt_enable(); - percpu_up_read(&c->mark_lock); - } - return 0; -} - -int bch2_mark_reservation(struct btree_trans *trans, - struct bkey_s_c old, struct bkey_s_c new, - 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 bch_fs *c = trans->c; - struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new; - struct bch_fs_usage __percpu *fs_usage; + struct bch_fs_usage *fs_usage; unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas; s64 sectors = (s64) k.k->size; @@ -1165,119 +1151,53 @@ int bch2_mark_reservation(struct btree_trans *trans, return 0; } -static s64 __bch2_mark_reflink_p(struct btree_trans *trans, - struct bkey_s_c_reflink_p p, - u64 start, u64 end, - u64 *idx, unsigned flags, size_t r_idx) +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 bch_fs *c = trans->c; - struct reflink_gc *r; - int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1; - u64 next_idx = end; - s64 ret = 0; - struct printbuf buf = PRINTBUF; - - if (r_idx >= c->reflink_gc_nr) - goto not_found; - - r = genradix_ptr(&c->reflink_gc_table, r_idx); - next_idx = min(next_idx, r->offset - r->size); - if (*idx < next_idx) - goto not_found; - - BUG_ON((s64) r->refcount + add < 0); - - r->refcount += add; - *idx = r->offset; - return 0; -not_found: - if (fsck_err(c, "pointer to missing indirect extent\n" - " %s\n" - " missing range %llu-%llu", - (bch2_bkey_val_to_text(&buf, c, p.s_c), buf.buf), - *idx, next_idx)) { - struct bkey_i_error new; - - bkey_init(&new.k); - new.k.type = KEY_TYPE_error; - new.k.p = bkey_start_pos(p.k); - new.k.p.offset += *idx - start; - bch2_key_resize(&new.k, next_idx - *idx); - ret = __bch2_btree_insert(trans, BTREE_ID_extents, &new.k_i); - } - - *idx = next_idx; -fsck_err: - printbuf_exit(&buf); - return ret; + return mem_trigger_run_overwrite_then_insert(__mark_reservation, trans, btree_id, level, old, new, flags); } -int bch2_mark_reflink_p(struct btree_trans *trans, - struct bkey_s_c old, struct bkey_s_c new, - unsigned flags) +void bch2_trans_fs_usage_revert(struct btree_trans *trans, + struct replicas_delta_list *deltas) { struct bch_fs *c = trans->c; - struct bkey_s_c k = flags & BTREE_TRIGGER_OVERWRITE ? old: new; - 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), start = idx; - u64 end = le64_to_cpu(p.v->idx) + p.k->size; - int ret = 0; + struct bch_fs_usage *dst; + struct replicas_delta *d, *top = (void *) deltas->d + deltas->used; + s64 added = 0; + unsigned i; - BUG_ON(!(flags & BTREE_TRIGGER_GC)); + percpu_down_read(&c->mark_lock); + preempt_disable(); + dst = fs_usage_ptr(c, trans->journal_res.seq, false); - if (c->sb.version >= bcachefs_metadata_version_reflink_p_fix) { - idx -= le32_to_cpu(p.v->front_pad); - end += le32_to_cpu(p.v->back_pad); + /* 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)); } - l = 0; - r = c->reflink_gc_nr; - while (l < r) { - m = l + (r - l) / 2; + dst->nr_inodes -= deltas->nr_inodes; - ref = genradix_ptr(&c->reflink_gc_table, m); - if (ref->offset <= idx) - l = m + 1; - else - r = m; + 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]; } - while (idx < end && !ret) - ret = __bch2_mark_reflink_p(trans, p, start, end, - &idx, flags, l++); - - 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; - struct printbuf buf = PRINTBUF; - - bch_err(c, "disk usage increased %lli more than %u sectors reserved", - should_not_have_added, disk_res_sectors); - - trans_for_each_update(trans, i) { - struct bkey_s_c old = { &i->old_k, i->old_v }; - - pr_err("while inserting"); - printbuf_reset(&buf); - bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(i->k)); - pr_err(" %s", buf.buf); - pr_err("overlapping with"); - printbuf_reset(&buf); - bch2_bkey_val_to_text(&buf, c, old); - pr_err(" %s", buf.buf); + if (added > 0) { + trans->disk_res->sectors += added; + this_cpu_add(*c->online_reserved, added); } - __WARN(); - printbuf_exit(&buf); + preempt_enable(); + percpu_up_read(&c->mark_lock); } int bch2_trans_fs_usage_apply(struct btree_trans *trans, @@ -1286,8 +1206,8 @@ int bch2_trans_fs_usage_apply(struct btree_trans *trans, 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, *d2; + 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; @@ -1344,7 +1264,9 @@ int bch2_trans_fs_usage_apply(struct btree_trans *trans, 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: */ @@ -1358,7 +1280,7 @@ need_mark: /* trans_mark: */ -static int bch2_trans_mark_pointer(struct btree_trans *trans, +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) @@ -1366,38 +1288,36 @@ static int bch2_trans_mark_pointer(struct btree_trans *trans, bool insert = !(flags & BTREE_TRIGGER_OVERWRITE); struct btree_iter iter; struct bkey_i_alloc_v4 *a; - struct bpos bucket_pos; + struct bpos bucket; struct bch_backpointer bp; s64 sectors; int ret; - bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, &bucket_pos, &bp); + 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_pos); + a = bch2_trans_start_alloc_update(trans, &iter, bucket); if (IS_ERR(a)) return PTR_ERR(a); 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); + &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 err; + return ret; if (!p.ptr.cached) { - ret = insert - ? bch2_bucket_backpointer_add(trans, a, bp, k) - : bch2_bucket_backpointer_del(trans, a, bp, k); + ret = bch2_bucket_backpointer_mod(trans, bucket, bp, k, insert); if (ret) - goto err; + return ret; } - ret = bch2_trans_update(trans, &iter, &a->k_i, 0); -err: - bch2_trans_iter_exit(trans, &iter); - return ret; + return 0; } static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans, @@ -1405,28 +1325,22 @@ static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans, s64 sectors, enum bch_data_type data_type) { 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_trans_inconsistent(trans, + 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); - ret = -EIO; goto err; } - if (!bch2_ptr_matches_stripe(bkey_s_c_to_stripe(k).v, p)) { + if (!bch2_ptr_matches_stripe(&s->v, p)) { bch2_trans_inconsistent(trans, "stripe pointer doesn't match stripe %llu", (u64) p.ec.idx); @@ -1434,37 +1348,23 @@ static int bch2_trans_mark_stripe_ptr(struct btree_trans *trans, 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); - ret = bch2_trans_update(trans, &iter, &s->k_i, 0); - if (ret) - goto err; - 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; } -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) +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_s_c k = flags & BTREE_TRIGGER_OVERWRITE - ? old - : bkey_i_to_s_c(new); struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); const union bch_extent_entry *entry; struct extent_ptr_decoded p; @@ -1477,7 +1377,7 @@ int bch2_trans_mark_extent(struct btree_trans *trans, : k.k->size; s64 dirty_sectors = 0; bool stale; - int ret; + int ret = 0; r.e.data_type = data_type; r.e.nr_devs = 0; @@ -1496,9 +1396,12 @@ int bch2_trans_mark_extent(struct btree_trans *trans, 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; @@ -1513,9 +1416,27 @@ 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; +} + +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, @@ -1538,9 +1459,9 @@ static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans, if (IS_ERR(a)) return PTR_ERR(a); - ret = check_bucket_ref(c, s.s_c, ptr, sectors, data_type, + ret = check_bucket_ref(trans, s.s_c, ptr, sectors, data_type, a->v.gen, a->v.data_type, - a->v.dirty_sectors, a->v.cached_sectors); + a->v.dirty_sectors); if (ret) goto err; @@ -1568,6 +1489,7 @@ static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans, a->v.stripe = s.k->p.offset; a->v.stripe_redundancy = s.v->nr_redundant; + a->v.data_type = BCH_DATA_stripe; } else { if (bch2_trans_inconsistent_on(a->v.stripe != s.k->p.offset || a->v.stripe_redundancy != s.v->nr_redundant, trans, @@ -1580,6 +1502,7 @@ static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans, a->v.stripe = 0; a->v.stripe_redundancy = 0; + a->v.data_type = alloc_data_type(a->v, BCH_DATA_user); } a->v.dirty_sectors += sectors; @@ -1630,14 +1553,18 @@ int bch2_trans_mark_stripe(struct btree_trans *trans, s64 sectors = le16_to_cpu(new_s->sectors); bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(new)); - update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant); + ret = bch2_update_replicas_list(trans, &r.e, sectors * new_s->nr_redundant); + if (ret) + return ret; } 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->nr_redundant); + ret = bch2_update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant); + if (ret) + return ret; } for (i = 0; i < nr_blocks; i++) { @@ -1665,42 +1592,24 @@ int bch2_trans_mark_stripe(struct btree_trans *trans, return ret; } -int bch2_trans_mark_inode(struct btree_trans *trans, - enum btree_id btree_id, unsigned level, - struct bkey_s_c old, - struct bkey_i *new, - unsigned flags) -{ - int nr = bkey_is_inode(&new->k) - bkey_is_inode(old.k); - - if (nr) { - struct replicas_delta_list *d = - replicas_deltas_realloc(trans, 0); - d->nr_inodes += nr; - } - - return 0; -} - -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) +static int __trans_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_OVERWRITE - ? old - : bkey_i_to_s_c(new); unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas; s64 sectors = (s64) k.k->size; struct replicas_delta_list *d; + 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)); @@ -1708,109 +1617,13 @@ 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; - struct printbuf buf = PRINTBUF; - int 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_bkey_val_to_text(&buf, c, p.s_c); - bch2_trans_inconsistent(trans, - "nonexistent indirect extent at %llu while marking\n %s", - *idx, buf.buf); - ret = -EIO; - goto err; - } - - if (!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)) { - bch2_bkey_val_to_text(&buf, c, p.s_c); - bch2_trans_inconsistent(trans, - "indirect extent refcount underflow at %llu while marking\n %s", - *idx, buf.buf); - ret = -EIO; - goto err; - } - - if (flags & BTREE_TRIGGER_INSERT) { - struct bch_reflink_p *v = (struct bch_reflink_p *) p.v; - u64 pad; - - pad = max_t(s64, le32_to_cpu(v->front_pad), - le64_to_cpu(v->idx) - bkey_start_offset(k.k)); - BUG_ON(pad > U32_MAX); - v->front_pad = cpu_to_le32(pad); - - pad = max_t(s64, le32_to_cpu(v->back_pad), - k.k->p.offset - p.k->size - le64_to_cpu(v->idx)); - BUG_ON(pad > U32_MAX); - v->back_pad = cpu_to_le32(pad); - } - - le64_add_cpu(refcount, add); - - bch2_btree_iter_set_pos_to_extent_start(&iter); - ret = bch2_trans_update(trans, &iter, n, 0); - if (ret) - goto err; - - *idx = k.k->p.offset; -err: - bch2_trans_iter_exit(trans, &iter); - printbuf_exit(&buf); - return ret; -} - -int bch2_trans_mark_reflink_p(struct btree_trans *trans, - enum btree_id btree_id, unsigned level, - struct bkey_s_c old, - struct bkey_i *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_OVERWRITE - ? old - : bkey_i_to_s_c(new); - struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k); - u64 idx, end_idx; - int ret = 0; - - if (flags & BTREE_TRIGGER_INSERT) { - struct bch_reflink_p *v = (struct bch_reflink_p *) p.v; - - v->front_pad = v->back_pad = 0; - } - - idx = le64_to_cpu(p.v->idx) - le32_to_cpu(p.v->front_pad); - end_idx = le64_to_cpu(p.v->idx) + p.k->size + - le32_to_cpu(p.v->back_pad); - - while (idx < end_idx && !ret) - ret = __bch2_trans_mark_reflink_p(trans, p, &idx, 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, @@ -1833,8 +1646,9 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans, if (IS_ERR(a)) return PTR_ERR(a); - if (a->v.data_type && a->v.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, a->v.gen, @@ -1842,16 +1656,16 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans, bch2_data_types[type], bch2_data_types[type]); ret = -EIO; - goto out; + goto err; } - a->v.data_type = type; - a->v.dirty_sectors = sectors; - - ret = bch2_trans_update(trans, &iter, &a->k_i, 0); - if (ret) - goto out; -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; } @@ -1939,14 +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_run(c, __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; @@ -1999,7 +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); @@ -2023,21 +1853,25 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets) struct bucket_gens *bucket_gens = NULL, *old_bucket_gens = NULL; unsigned long *buckets_nouse = NULL; bool resize = ca->bucket_gens != NULL; - int ret = -ENOMEM; + int ret; if (!(bucket_gens = kvpmalloc(sizeof(struct bucket_gens) + nbuckets, - GFP_KERNEL|__GFP_ZERO)) || - (c->opts.buckets_nouse && + GFP_KERNEL|__GFP_ZERO))) { + ret = -BCH_ERR_ENOMEM_bucket_gens; + goto err; + } + + if ((c->opts.buckets_nouse && !(buckets_nouse = kvpmalloc(BITS_TO_LONGS(nbuckets) * sizeof(unsigned long), - GFP_KERNEL|__GFP_ZERO)))) + GFP_KERNEL|__GFP_ZERO)))) { + ret = -BCH_ERR_ENOMEM_buckets_nouse; goto err; + } bucket_gens->first_bucket = ca->mi.first_bucket; bucket_gens->nbuckets = nbuckets; - bch2_copygc_stop(c); - if (resize) { down_write(&c->gc_lock); down_write(&ca->bucket_lock); @@ -2101,13 +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); }