]> git.sesse.net Git - bcachefs-tools-debian/blobdiff - libbcachefs/buckets.c
Update bcachefs sources to bee7b5a4fa21 bcachefs: Pin btree cache in ram for random...
[bcachefs-tools-debian] / libbcachefs / buckets.c
index a8af803e7289fec94dfcc03929b1fcc1c94d5b98..c2f46b267b3ad50c796690320f0a700411940931 100644 (file)
@@ -25,7 +25,7 @@
 
 #include <linux/preempt.h>
 
-static inline void fs_usage_data_type_to_base(struct bch_fs_usage *fs_usage,
+static inline void fs_usage_data_type_to_base(struct bch_fs_usage_base *fs_usage,
                                              enum bch_data_type data_type,
                                              s64 sectors)
 {
@@ -47,31 +47,27 @@ 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++)
-               usage->reserved += usage->persistent_reserved[i];
+       for (unsigned i = 0; i < BCH_REPLICAS_MAX; i++)
+               usage->b.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]);
+               fs_usage_data_type_to_base(&usage->b, 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 +
-                                 dev.d[BCH_DATA_journal].buckets) *
+               usage->b.hidden += (dev.d[BCH_DATA_sb].buckets +
+                                   dev.d[BCH_DATA_journal].buckets) *
                        ca->mi.bucket_size;
        }
 
@@ -158,8 +154,7 @@ retry:
 
 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));
 
@@ -171,7 +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,
@@ -193,15 +188,15 @@ void bch2_fs_usage_to_text(struct printbuf *out,
        prt_printf(out, "capacity:\t\t\t%llu\n", c->capacity);
 
        prt_printf(out, "hidden:\t\t\t\t%llu\n",
-              fs_usage->u.hidden);
+              fs_usage->u.b.hidden);
        prt_printf(out, "data:\t\t\t\t%llu\n",
-              fs_usage->u.data);
+              fs_usage->u.b.data);
        prt_printf(out, "cached:\t\t\t\t%llu\n",
-              fs_usage->u.cached);
+              fs_usage->u.b.cached);
        prt_printf(out, "reserved:\t\t\t%llu\n",
-              fs_usage->u.reserved);
+              fs_usage->u.b.reserved);
        prt_printf(out, "nr_inodes:\t\t\t%llu\n",
-              fs_usage->u.nr_inodes);
+              fs_usage->u.b.nr_inodes);
        prt_printf(out, "online reserved:\t\t%llu\n",
               fs_usage->online_reserved);
 
@@ -214,7 +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");
@@ -230,10 +225,10 @@ static u64 reserve_factor(u64 r)
 
 u64 bch2_fs_sectors_used(struct bch_fs *c, struct bch_fs_usage_online *fs_usage)
 {
-       return min(fs_usage->u.hidden +
-                  fs_usage->u.btree +
-                  fs_usage->u.data +
-                  reserve_factor(fs_usage->u.reserved +
+       return min(fs_usage->u.b.hidden +
+                  fs_usage->u.b.btree +
+                  fs_usage->u.b.data +
+                  reserve_factor(fs_usage->u.b.reserved +
                                  fs_usage->online_reserved),
                   c->capacity);
 }
@@ -245,17 +240,17 @@ __bch2_fs_usage_read_short(struct bch_fs *c)
        u64 data, reserved;
 
        ret.capacity = c->capacity -
-               bch2_fs_usage_read_one(c, &c->usage_base->hidden);
+               bch2_fs_usage_read_one(c, &c->usage_base->b.hidden);
 
-       data            = bch2_fs_usage_read_one(c, &c->usage_base->data) +
-               bch2_fs_usage_read_one(c, &c->usage_base->btree);
-       reserved        = bch2_fs_usage_read_one(c, &c->usage_base->reserved) +
+       data            = bch2_fs_usage_read_one(c, &c->usage_base->b.data) +
+               bch2_fs_usage_read_one(c, &c->usage_base->b.btree);
+       reserved        = bch2_fs_usage_read_one(c, &c->usage_base->b.reserved) +
                percpu_u64_get(c->online_reserved);
 
        ret.used        = min(ret.capacity, data + reserve_factor(reserved));
        ret.free        = ret.capacity - ret.used;
 
-       ret.nr_inodes   = bch2_fs_usage_read_one(c, &c->usage_base->nr_inodes);
+       ret.nr_inodes   = bch2_fs_usage_read_one(c, &c->usage_base->b.nr_inodes);
 
        return ret;
 }
@@ -277,18 +272,34 @@ 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++) {
+               bch2_prt_data_type(out, 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 bch_alloc_v4 old,
-                                 struct bch_alloc_v4 new,
-                                 u64 journal_seq, bool gc)
+void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
+                          const struct bch_alloc_v4 *old,
+                          const struct bch_alloc_v4 *new,
+                          u64 journal_seq, bool gc)
 {
        struct bch_fs_usage *fs_usage;
        struct bch_dev_usage *u;
@@ -296,56 +307,51 @@ static void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
        preempt_disable();
        fs_usage = fs_usage_ptr(c, journal_seq, gc);
 
-       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 (data_type_is_hidden(old->data_type))
+               fs_usage->b.hidden -= ca->mi.bucket_size;
+       if (data_type_is_hidden(new->data_type))
+               fs_usage->b.hidden += 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->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 -= 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 += new->cached_sectors;
+       u->d[BCH_DATA_cached].sectors -= 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 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,
        };
+}
+
+void bch2_dev_usage_update_m(struct bch_fs *c, struct bch_dev *ca,
+                            struct bucket *old, struct bucket *new)
+{
+       struct bch_alloc_v4 old_a = bucket_m_to_alloc(*old);
+       struct bch_alloc_v4 new_a = bucket_m_to_alloc(*new);
 
-       bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, gc);
+       bch2_dev_usage_update(c, ca, &old_a, &new_a, 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);
@@ -353,14 +359,14 @@ static inline int __update_replicas(struct bch_fs *c,
        if (idx < 0)
                return -1;
 
-       fs_usage_data_type_to_base(fs_usage, r->data_type, sectors);
+       fs_usage_data_type_to_base(&fs_usage->b, r->data_type, sectors);
        fs_usage->replicas[idx]         += sectors;
        return 0;
 }
 
-static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k,
-                       struct bch_replicas_entry *r, s64 sectors,
-                       unsigned journal_seq, bool gc)
+int bch2_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 *fs_usage;
        int idx, ret = 0;
@@ -370,8 +376,8 @@ static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k,
 
        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);
@@ -388,7 +394,7 @@ static inline int update_replicas(struct bch_fs *c, struct bkey_s_c k,
 
        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_data_type_to_base(&fs_usage->b, r->data_type, sectors);
        fs_usage->replicas[idx]         += sectors;
        preempt_enable();
 err:
@@ -407,7 +413,7 @@ static inline int update_cached_sectors(struct bch_fs *c,
 
        bch2_replicas_entry_cached(&r.e, dev);
 
-       return update_replicas(c, k, &r.e, sectors, journal_seq, gc);
+       return bch2_update_replicas(c, k, &r.e, sectors, journal_seq, gc);
 }
 
 static int __replicas_deltas_realloc(struct btree_trans *trans, unsigned more,
@@ -453,9 +459,9 @@ int bch2_replicas_deltas_realloc(struct btree_trans *trans, unsigned more)
                                __replicas_deltas_realloc(trans, more, _gfp));
 }
 
-static inline int update_replicas_list(struct btree_trans *trans,
-                                       struct bch_replicas_entry *r,
-                                       s64 sectors)
+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;
@@ -481,139 +487,13 @@ static inline int update_replicas_list(struct btree_trans *trans,
        return 0;
 }
 
-static inline int 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);
 
-       return 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_convert, new_a_convert;
-       const struct bch_alloc_v4 *old_a, *new_a;
-       struct bch_dev *ca;
-       int ret = 0;
-
-       /*
-        * alloc btree is read in by bch2_alloc_read, not gc:
-        */
-       if ((flags & BTREE_TRIGGER_GC) &&
-           !(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);
-
-       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) &&
-           new.k->type == KEY_TYPE_alloc_v4) {
-               struct bch_alloc_v4 *v = (struct bch_alloc_v4 *) new.v;
-
-               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:
-                */
-               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;
-       }
-
-       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;
-               }
-       }
-
-       percpu_down_read(&c->mark_lock);
-       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);
-
-       if (gc) {
-               struct bucket *g = gc_bucket(ca, new.k->p.offset);
-
-               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;
-
-               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,
-                                           -((s64) old_a->cached_sectors),
-                                           journal_seq, gc);
-               if (ret) {
-                       bch2_fs_fatal_error(c, "%s(): no replicas entry while updating cached sectors",
-                                           __func__);
-                       return ret;
-               }
-       }
-
-       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 &&
-           (!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 &&
-           should_invalidate_buckets(ca, bch2_dev_usage_read(ca)))
-               bch2_do_invalidates(c);
-
-       if (new_a->data_type == BCH_DATA_need_gc_gens)
-               bch2_do_gc_gens(c);
-
-       return 0;
+       return bch2_update_replicas_list(trans, &r.e, sectors);
 }
 
 int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
@@ -643,8 +523,8 @@ int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
        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])) {
+                       bch2_data_type_str(g->data_type),
+                       bch2_data_type_str(data_type))) {
                ret = -EIO;
                goto err;
        }
@@ -652,37 +532,33 @@ int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
        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],
+                       bch2_data_type_str(g->data_type ?: data_type),
                        g->dirty_sectors, sectors)) {
                ret = -EIO;
                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 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)
+int bch2_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 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);
-       u32 bucket_sectors = !ptr->cached
-               ? dirty_sectors
-               : cached_sectors;
        struct printbuf buf = PRINTBUF;
        int ret = 0;
 
@@ -695,10 +571,11 @@ static int check_bucket_ref(struct btree_trans *trans,
 
        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,
-                       bch2_data_types[bucket_data_type ?: ptr_data_type],
+                       bch2_data_type_str(bucket_data_type ?: ptr_data_type),
                        ptr->gen,
                        (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
                ret = -EIO;
@@ -707,10 +584,11 @@ static int check_bucket_ref(struct btree_trans *trans,
 
        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,
-                       bch2_data_types[bucket_data_type ?: ptr_data_type],
+                       bch2_data_type_str(bucket_data_type ?: ptr_data_type),
                        ptr->gen,
                        (printbuf_reset(&buf),
                         bch2_bkey_val_to_text(&buf, c, k), buf.buf));
@@ -720,11 +598,12 @@ static int check_bucket_ref(struct btree_trans *trans,
 
        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,
                        *bucket_gen(ca, bucket_nr),
-                       bch2_data_types[bucket_data_type ?: ptr_data_type],
+                       bch2_data_type_str(bucket_data_type ?: ptr_data_type),
                        ptr->gen,
                        (printbuf_reset(&buf),
                         bch2_bkey_val_to_text(&buf, c, k), buf.buf));
@@ -741,11 +620,12 @@ static int check_bucket_ref(struct btree_trans *trans,
            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,
-                       bch2_data_types[bucket_data_type],
-                       bch2_data_types[ptr_data_type],
+                       bch2_data_type_str(bucket_data_type),
+                       bch2_data_type_str(ptr_data_type),
                        (printbuf_reset(&buf),
                         bch2_bkey_val_to_text(&buf, c, k), buf.buf));
                ret = -EIO;
@@ -754,10 +634,11 @@ static int check_bucket_ref(struct btree_trans *trans,
 
        if ((u64) bucket_sectors + sectors > U32_MAX) {
                bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+                             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],
+                       bch2_data_type_str(bucket_data_type ?: ptr_data_type),
                        bucket_sectors, sectors,
                        (printbuf_reset(&buf),
                         bch2_bkey_val_to_text(&buf, c, k), buf.buf));
@@ -772,65 +653,139 @@ err:
        goto out;
 }
 
-static int mark_stripe_bucket(struct btree_trans *trans,
-                             struct bkey_s_c k,
-                             unsigned ptr_idx,
-                             unsigned flags)
+void bch2_trans_fs_usage_revert(struct btree_trans *trans,
+                               struct replicas_delta_list *deltas)
 {
        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 : 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);
-       struct bucket old, new, *g;
-       struct printbuf buf = PRINTBUF;
-       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);
+
+       /* 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));
+       }
+
+       dst->b.nr_inodes -= deltas->nr_inodes;
+
+       for (i = 0; i < BCH_REPLICAS_MAX; i++) {
+               added                           -= deltas->persistent_reserved[i];
+               dst->b.reserved                 -= deltas->persistent_reserved[i];
+               dst->persistent_reserved[i]     -= deltas->persistent_reserved[i];
+       }
 
-       /* * XXX doesn't handle deletion */
+       if (added > 0) {
+               trans->disk_res->sectors += added;
+               this_cpu_add(*c->online_reserved, added);
+       }
+
+       preempt_enable();
+       percpu_up_read(&c->mark_lock);
+}
+
+void bch2_trans_account_disk_usage_change(struct btree_trans *trans)
+{
+       struct bch_fs *c = trans->c;
+       u64 disk_res_sectors = trans->disk_res ? trans->disk_res->sectors : 0;
+       static int warned_disk_usage = 0;
+       bool warn = false;
 
        percpu_down_read(&c->mark_lock);
-       g = PTR_GC_BUCKET(ca, ptr);
-
-       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->gen,
-                             (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-               ret = -EINVAL;
-               goto err;
+       preempt_disable();
+       struct bch_fs_usage_base *dst = &fs_usage_ptr(c, trans->journal_res.seq, false)->b;
+       struct bch_fs_usage_base *src = &trans->fs_usage_delta;
+
+       s64 added = src->btree + src->data + src->reserved;
+
+       /*
+        * Not allowed to reduce sectors_available except by getting a
+        * reservation:
+        */
+       s64 should_not_have_added = added - (s64) disk_res_sectors;
+       if (unlikely(should_not_have_added > 0)) {
+               u64 old, new, v = atomic64_read(&c->sectors_available);
+
+               do {
+                       old = v;
+                       new = max_t(s64, 0, old - should_not_have_added);
+               } while ((v = atomic64_cmpxchg(&c->sectors_available,
+                                              old, new)) != old);
+
+               added -= should_not_have_added;
+               warn = true;
        }
 
-       bucket_lock(g);
-       old = *g;
+       if (added > 0) {
+               trans->disk_res->sectors -= added;
+               this_cpu_sub(*c->online_reserved, added);
+       }
 
-       ret = check_bucket_ref(trans, k, ptr, sectors, data_type,
-                              g->gen, g->data_type,
-                              g->dirty_sectors, g->cached_sectors);
-       if (ret)
-               goto err;
+       dst->hidden     += src->hidden;
+       dst->btree      += src->btree;
+       dst->data       += src->data;
+       dst->cached     += src->cached;
+       dst->reserved   += src->reserved;
+       dst->nr_inodes  += src->nr_inodes;
 
-       g->data_type = data_type;
-       g->dirty_sectors += sectors;
+       preempt_enable();
+       percpu_up_read(&c->mark_lock);
 
-       g->stripe               = k.k->p.offset;
-       g->stripe_redundancy    = s->nr_redundant;
-       new = *g;
-err:
-       bucket_unlock(g);
-       if (!ret)
-               bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true);
+       if (unlikely(warn) && !xchg(&warned_disk_usage, 1))
+               bch2_trans_inconsistent(trans,
+                                       "disk usage increased %lli more than %llu sectors reserved)",
+                                       should_not_have_added, disk_res_sectors);
+}
+
+int bch2_trans_fs_usage_apply(struct btree_trans *trans,
+                             struct replicas_delta_list *deltas)
+{
+       struct bch_fs *c = trans->c;
+       struct replicas_delta *d, *d2;
+       struct replicas_delta *top = (void *) deltas->d + deltas->used;
+       struct bch_fs_usage *dst;
+       unsigned i;
+
+       percpu_down_read(&c->mark_lock);
+       preempt_disable();
+       dst = fs_usage_ptr(c, trans->journal_res.seq, false);
+
+       for (d = deltas->d; d != top; d = replicas_delta_next(d))
+               if (__update_replicas(c, dst, &d->r, d->delta))
+                       goto need_mark;
+
+       dst->b.nr_inodes += deltas->nr_inodes;
+
+       for (i = 0; i < BCH_REPLICAS_MAX; i++) {
+               dst->b.reserved                 += deltas->persistent_reserved[i];
+               dst->persistent_reserved[i]     += deltas->persistent_reserved[i];
+       }
+
+       preempt_enable();
        percpu_up_read(&c->mark_lock);
-       printbuf_exit(&buf);
-       return ret;
+       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;
 }
 
+/* KEY_TYPE_extent: */
+
 static int __mark_pointer(struct btree_trans *trans,
                          struct bkey_s_c k,
                          const struct bch_extent_ptr *ptr,
@@ -841,105 +796,172 @@ static int __mark_pointer(struct btree_trans *trans,
        u32 *dst_sectors = !ptr->cached
                ? dirty_sectors
                : cached_sectors;
-       int ret = check_bucket_ref(trans, k, ptr, sectors, ptr_data_type,
-                                  bucket_gen, *bucket_data_type,
-                                  *dirty_sectors, *cached_sectors);
+       int ret = bch2_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,
-                            unsigned flags)
+static int bch2_trigger_pointer(struct btree_trans *trans,
+                       enum btree_id btree_id, unsigned level,
+                       struct bkey_s_c k, struct extent_ptr_decoded p,
+                       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;
+       bool insert = !(flags & BTREE_TRIGGER_OVERWRITE);
+       struct bpos bucket;
+       struct bch_backpointer bp;
 
-       BUG_ON(!(flags & BTREE_TRIGGER_GC));
+       bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, &bucket, &bp);
+       *sectors = insert ? bp.bucket_len : -((s64) bp.bucket_len);
 
-       percpu_down_read(&c->mark_lock);
-       g = PTR_GC_BUCKET(ca, &p.ptr);
-       bucket_lock(g);
-       old = *g;
+       if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+               struct btree_iter iter;
+               struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update(trans, &iter, bucket);
+               int ret = PTR_ERR_OR_ZERO(a);
+               if (ret)
+                       return ret;
 
-       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;
+               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);
 
-       new = *g;
-       bucket_unlock(g);
-       if (!ret)
-               bch2_dev_usage_update_m(c, ca, old, new, journal_seq, true);
-       percpu_up_read(&c->mark_lock);
+               if (ret)
+                       return ret;
 
-       return ret;
+               if (!p.ptr.cached) {
+                       ret = bch2_bucket_backpointer_mod(trans, bucket, bp, k, insert);
+                       if (ret)
+                               return ret;
+               }
+       }
+
+       if (flags & BTREE_TRIGGER_GC) {
+               struct bch_fs *c = trans->c;
+               struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
+               enum bch_data_type data_type = bkey_ptr_data_type(btree_id, level, k, p);
+
+               percpu_down_read(&c->mark_lock);
+               struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
+               bucket_lock(g);
+               struct bucket old = *g;
+
+               u8 bucket_data_type = g->data_type;
+               int ret = __mark_pointer(trans, k, &p.ptr, *sectors,
+                                    data_type, g->gen,
+                                    &bucket_data_type,
+                                    &g->dirty_sectors,
+                                    &g->cached_sectors);
+               if (ret) {
+                       bucket_unlock(g);
+                       percpu_up_read(&c->mark_lock);
+                       return ret;
+               }
+
+               g->data_type = bucket_data_type;
+               struct bucket new = *g;
+               bucket_unlock(g);
+               bch2_dev_usage_update_m(c, ca, &old, &new);
+               percpu_up_read(&c->mark_lock);
+       }
+
+       return 0;
 }
 
-static int bch2_mark_stripe_ptr(struct btree_trans *trans,
+static int bch2_trigger_stripe_ptr(struct btree_trans *trans,
                                struct bkey_s_c k,
-                               struct bch_extent_stripe_ptr p,
+                               struct extent_ptr_decoded p,
                                enum bch_data_type data_type,
-                               s64 sectors,
-                               unsigned flags)
+                               s64 sectors, unsigned flags)
 {
-       struct bch_fs *c = trans->c;
-       struct bch_replicas_padded r;
-       struct gc_stripe *m;
+       if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+               struct btree_iter iter;
+               struct bkey_i_stripe *s = bch2_bkey_get_mut_typed(trans, &iter,
+                               BTREE_ID_stripes, POS(0, p.ec.idx),
+                               BTREE_ITER_WITH_UPDATES, stripe);
+               int 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);
+                       goto err;
+               }
 
-       BUG_ON(!(flags & BTREE_TRIGGER_GC));
+               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;
+               }
+
+               stripe_blockcount_set(&s->v, p.ec.block,
+                       stripe_blockcount_get(&s->v, p.ec.block) +
+                       sectors);
 
-       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;
+               struct bch_replicas_padded r;
+               bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(&s->k_i));
+               r.e.data_type = data_type;
+               ret = bch2_update_replicas_list(trans, &r.e, sectors);
+err:
+               bch2_trans_iter_exit(trans, &iter);
+               return ret;
        }
 
-       mutex_lock(&c->ec_stripes_heap_lock);
+       if (flags & BTREE_TRIGGER_GC) {
+               struct bch_fs *c = trans->c;
 
-       if (!m || !m->alive) {
-               mutex_unlock(&c->ec_stripes_heap_lock);
-               bch_err_ratelimited(c, "pointer to nonexistent stripe %llu",
-                                   (u64) p.idx);
-               bch2_inconsistent_error(c);
-               return -EIO;
-       }
+               BUG_ON(!(flags & BTREE_TRIGGER_GC));
 
-       m->block_sectors[p.block] += sectors;
+               struct gc_stripe *m = genradix_ptr_alloc(&c->gc_stripes, p.ec.idx, GFP_KERNEL);
+               if (!m) {
+                       bch_err(c, "error allocating memory for gc_stripes, idx %llu",
+                               (u64) p.ec.idx);
+                       return -BCH_ERR_ENOMEM_mark_stripe_ptr;
+               }
+
+               mutex_lock(&c->ec_stripes_heap_lock);
+
+               if (!m || !m->alive) {
+                       mutex_unlock(&c->ec_stripes_heap_lock);
+                       struct printbuf buf = PRINTBUF;
+                       bch2_bkey_val_to_text(&buf, c, k);
+                       bch_err_ratelimited(c, "pointer to nonexistent stripe %llu\n  while marking %s",
+                                           (u64) p.ec.idx, buf.buf);
+                       printbuf_exit(&buf);
+                       bch2_inconsistent_error(c);
+                       return -EIO;
+               }
 
-       r = m->r;
-       mutex_unlock(&c->ec_stripes_heap_lock);
+               m->block_sectors[p.ec.block] += sectors;
+
+               struct bch_replicas_padded r = m->r;
+               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);
+               r.e.data_type = data_type;
+               bch2_update_replicas(c, k, &r.e, sectors, trans->journal_res.seq, true);
+       }
 
        return 0;
 }
 
-static int __mark_extent(struct btree_trans *trans,
-                        enum btree_id btree_id, unsigned level,
-                        struct bkey_s_c k, unsigned flags)
+static int __trigger_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;
+       bool gc = flags & BTREE_TRIGGER_GC;
        struct bch_fs *c = trans->c;
        struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
        const union bch_extent_entry *entry;
@@ -948,47 +970,36 @@ static int __mark_extent(struct btree_trans *trans,
        enum bch_data_type data_type = bkey_is_btree_ptr(k.k)
                ? BCH_DATA_btree
                : BCH_DATA_user;
-       s64 sectors = bkey_is_btree_ptr(k.k)
-               ? btree_sectors(c)
-               : k.k->size;
        s64 dirty_sectors = 0;
-       bool stale;
-       int ret;
-
-       BUG_ON(!(flags & BTREE_TRIGGER_GC));
+       int ret = 0;
 
        r.e.data_type   = data_type;
        r.e.nr_devs     = 0;
        r.e.nr_required = 1;
 
        bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
-               s64 disk_sectors = ptr_disk_sectors(sectors, p);
-
-               if (flags & BTREE_TRIGGER_OVERWRITE)
-                       disk_sectors = -disk_sectors;
-
-               ret = bch2_mark_pointer(trans, btree_id, level, k, p, disk_sectors, flags);
+               s64 disk_sectors;
+               ret = bch2_trigger_pointer(trans, btree_id, level, k, p, &disk_sectors, flags);
                if (ret < 0)
                        return ret;
 
-               stale = ret > 0;
+               bool stale = ret > 0;
 
                if (p.ptr.cached) {
                        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__);
+                               ret = !gc
+                                       ? bch2_update_cached_sectors_list(trans, p.ptr.dev, disk_sectors)
+                                       : update_cached_sectors(c, k, p.ptr.dev, disk_sectors, 0, true);
+                               bch2_fs_fatal_err_on(ret && gc, c, "%s(): no replicas entry while updating cached sectors",
+                                                    __func__);
+                               if (ret)
                                        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(trans, k, p.ec, data_type,
-                                       disk_sectors, flags);
+                       ret = bch2_trigger_stripe_ptr(trans, k, p, data_type, disk_sectors, flags);
                        if (ret)
                                return ret;
 
@@ -1002,838 +1013,109 @@ static int __mark_extent(struct btree_trans *trans,
        }
 
        if (r.e.nr_devs) {
-               ret = update_replicas(c, k, &r.e, dirty_sectors, journal_seq, true);
-               if (ret) {
+               ret = !gc
+                       ? bch2_update_replicas_list(trans, &r.e, dirty_sectors)
+                       : bch2_update_replicas(c, k, &r.e, dirty_sectors, 0, true);
+               if (unlikely(ret && gc)) {
                        struct printbuf buf = PRINTBUF;
 
                        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;
-}
-
-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;
-       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;
-       unsigned i;
-       int ret;
-
-       BUG_ON(gc && old_s);
-
-       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) {
-                       bch2_stripes_heap_del(c, m, idx);
-
-                       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->nr_blocks    = new_s->nr_blocks;
-               m->nr_redundant = new_s->nr_redundant;
-
-               for (i = 0; i < new_s->nr_blocks; i++)
-                       m->ptrs[i] = new_s->ptrs[i];
-
-               bch2_bkey_to_replicas(&m->r.e, new);
-
-               /*
-                * gc recalculates this field from stripe ptr
-                * references:
-                */
-               memset(m->block_sectors, 0, sizeof(m->block_sectors));
-
-               for (i = 0; i < new_s->nr_blocks; i++) {
-                       ret = mark_stripe_bucket(trans, new, i, flags);
-                       if (ret)
-                               return ret;
-               }
-
-               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(&buf, c, new);
-                       bch2_fs_fatal_error(c, "no replicas entry for %s", buf.buf);
-                       printbuf_exit(&buf);
+               if (ret)
                        return ret;
-               }
        }
 
        return 0;
 }
 
-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 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, 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();
-       percpu_up_read(&c->mark_lock);
-
-       return 0;
-}
-
-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)
-{
-       return mem_trigger_run_overwrite_then_insert(__mark_reservation, trans, btree_id, level, old, new, flags);
-}
-
-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)
-{
-       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;
-
-               new = bch2_trans_kmalloc(trans, sizeof(*new));
-               ret = PTR_ERR_OR_ZERO(new);
-               if (ret)
-                       goto err;
-
-               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(trans, BTREE_ID_extents, &new->k_i,
-                                         BTREE_TRIGGER_NORUN);
-       }
-
-       *idx = next_idx;
-err:
-fsck_err:
-       printbuf_exit(&buf);
-       return ret;
-}
-
-static int __mark_reflink_p(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_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;
-
-       BUG_ON(!(flags & BTREE_TRIGGER_GC));
-
-       if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_reflink_p_fix) {
-               idx -= le32_to_cpu(p.v->front_pad);
-               end += le32_to_cpu(p.v->back_pad);
-       }
-
-       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 (idx < end && !ret)
-               ret = __bch2_mark_reflink_p(trans, p, start, end,
-                                           &idx, flags, l++);
-
-       return ret;
-}
-
-int bch2_mark_reflink_p(struct btree_trans *trans,
+int bch2_trigger_extent(struct btree_trans *trans,
                        enum btree_id btree_id, unsigned level,
-                       struct bkey_s_c old, struct bkey_s_c new,
+                       struct bkey_s_c old, struct bkey_s new,
                        unsigned flags)
 {
-       return mem_trigger_run_overwrite_then_insert(__mark_reflink_p, trans, btree_id, level, old, new, flags);
-}
-
-void bch2_trans_fs_usage_revert(struct btree_trans *trans,
-                               struct replicas_delta_list *deltas)
-{
-       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);
-       preempt_disable();
-       dst = fs_usage_ptr(c, trans->journal_res.seq, false);
-
-       /* 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));
-       }
-
-       dst->nr_inodes -= deltas->nr_inodes;
-
-       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 (added > 0) {
-               trans->disk_res->sectors += added;
-               this_cpu_add(*c->online_reserved, added);
-       }
-
-       preempt_enable();
-       percpu_up_read(&c->mark_lock);
-}
-
-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, *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_down_read(&c->mark_lock);
-       preempt_disable();
-       dst = fs_usage_ptr(c, trans->journal_res.seq, false);
-
-       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;
-               }
-
-               if (__update_replicas(c, dst, &d->r, d->delta))
-                       goto need_mark;
-       }
-
-       dst->nr_inodes += deltas->nr_inodes;
-
-       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];
-       }
-
-       /*
-        * Not allowed to reduce sectors_available except by getting a
-        * reservation:
-        */
-       should_not_have_added = added - (s64) disk_res_sectors;
-       if (unlikely(should_not_have_added > 0)) {
-               u64 old, new, v = atomic64_read(&c->sectors_available);
-
-               do {
-                       old = v;
-                       new = max_t(s64, 0, old - should_not_have_added);
-               } while ((v = atomic64_cmpxchg(&c->sectors_available,
-                                              old, new)) != old);
-
-               added -= should_not_have_added;
-               warn = true;
-       }
-
-       if (added > 0) {
-               trans->disk_res->sectors -= added;
-               this_cpu_sub(*c->online_reserved, added);
-       }
-
-       preempt_enable();
-       percpu_up_read(&c->mark_lock);
-
-       if (unlikely(warn) && !xchg(&warned_disk_usage, 1))
-               bch2_trans_inconsistent(trans,
-                                       "disk usage increased %lli more than %u 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 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)
-{
-       bool insert = !(flags & BTREE_TRIGGER_OVERWRITE);
-       struct btree_iter iter;
-       struct bkey_i_alloc_v4 *a;
-       struct bpos bucket;
-       struct bch_backpointer bp;
-       s64 sectors;
-       int ret;
-
-       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(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)
-               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 btree_iter iter;
-       struct bkey_i_stripe *s;
-       struct bch_replicas_padded r;
-       int ret = 0;
-
-       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);
-               goto err;
-       }
-
-       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;
-       }
-
-       stripe_blockcount_set(&s->v, p.ec.block,
-               stripe_blockcount_get(&s->v, p.ec.block) +
-               sectors);
-
-       bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(&s->k_i));
-       r.e.data_type = data_type;
-       ret = update_replicas_list(trans, &r.e, sectors);
-err:
-       bch2_trans_iter_exit(trans, &iter);
-       return ret;
-}
-
-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);
-       const union bch_extent_entry *entry;
-       struct extent_ptr_decoded p;
-       struct bch_replicas_padded r;
-       enum bch_data_type data_type = bkey_is_btree_ptr(k.k)
-               ? BCH_DATA_btree
-               : BCH_DATA_user;
-       s64 sectors = bkey_is_btree_ptr(k.k)
-               ? btree_sectors(c)
-               : k.k->size;
-       s64 dirty_sectors = 0;
-       bool stale;
-       int ret = 0;
-
-       r.e.data_type   = data_type;
-       r.e.nr_devs     = 0;
-       r.e.nr_required = 1;
-
-       bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
-               s64 disk_sectors = ptr_disk_sectors(sectors, p);
-
-               if (flags & BTREE_TRIGGER_OVERWRITE)
-                       disk_sectors = -disk_sectors;
-
-               ret = bch2_trans_mark_pointer(trans, btree_id, level, k, p, flags);
-               if (ret < 0)
-                       return ret;
+       struct bkey_ptrs_c new_ptrs = bch2_bkey_ptrs_c(new.s_c);
+       struct bkey_ptrs_c old_ptrs = bch2_bkey_ptrs_c(old);
+       unsigned new_ptrs_bytes = (void *) new_ptrs.end - (void *) new_ptrs.start;
+       unsigned old_ptrs_bytes = (void *) old_ptrs.end - (void *) old_ptrs.start;
+
+       /* if pointers aren't changing - nothing to do: */
+       if (new_ptrs_bytes == old_ptrs_bytes &&
+           !memcmp(new_ptrs.start,
+                   old_ptrs.start,
+                   new_ptrs_bytes))
+               return 0;
 
-               stale = ret > 0;
+       if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+               struct bch_fs *c = trans->c;
+               int mod = (int) bch2_bkey_needs_rebalance(c, new.s_c) -
+                         (int) bch2_bkey_needs_rebalance(c, old);
 
-               if (p.ptr.cached) {
-                       if (!stale) {
-                               ret = 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;
-               } else {
-                       ret = bch2_trans_mark_stripe_ptr(trans, p,
-                                       disk_sectors, data_type);
+               if (mod) {
+                       int ret = bch2_btree_bit_mod_buffered(trans, BTREE_ID_rebalance_work,
+                                                             new.k->p, mod > 0);
                        if (ret)
                                return ret;
-
-                       r.e.nr_required = 0;
                }
        }
 
-       if (r.e.nr_devs)
-               ret = update_replicas_list(trans, &r.e, dirty_sectors);
+       if (flags & (BTREE_TRIGGER_TRANSACTIONAL|BTREE_TRIGGER_GC))
+               return trigger_run_overwrite_then_insert(__trigger_extent, trans, btree_id, level, old, new, flags);
 
-       return ret;
+       return 0;
 }
 
-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);
-}
+/* KEY_TYPE_reservation */
 
-static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans,
-                                        struct bkey_s_c_stripe s,
-                                        unsigned idx, bool deleting)
+static int __trigger_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;
-       const struct bch_extent_ptr *ptr = &s.v->ptrs[idx];
-       struct btree_iter iter;
-       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;
-
-       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);
-
-       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);
-       if (ret)
-               goto err;
-
-       if (!deleting) {
-               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;
-               }
-
-               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,
-                               "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);
-       }
-
-       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;
-}
-
-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)
-{
-       const struct bch_stripe *old_s = NULL;
-       struct bch_stripe *new_s = NULL;
-       struct bch_replicas_padded r;
-       unsigned i, nr_blocks;
-       int ret = 0;
-
-       if (old.k->type == KEY_TYPE_stripe)
-               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 && 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;
-
-       BUG_ON(new_s && old_s &&
-              (new_s->nr_blocks        != old_s->nr_blocks ||
-               new_s->nr_redundant     != old_s->nr_redundant));
-
-       nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
-
-       if (new_s) {
-               s64 sectors = le16_to_cpu(new_s->sectors);
-
-               bch2_bkey_to_replicas(&r.e, bkey_i_to_s_c(new));
-               ret = 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);
-               ret = update_replicas_list(trans, &r.e, sectors * old_s->nr_redundant);
-               if (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;
-
-               if (new_s) {
-                       ret = bch2_trans_mark_stripe_bucket(trans,
-                                       bkey_i_to_s_c_stripe(new), i, false);
-                       if (ret)
-                               break;
-               }
-
-               if (old_s) {
-                       ret = bch2_trans_mark_stripe_bucket(trans,
-                                       bkey_s_c_to_stripe(old), i, true);
-                       if (ret)
-                               break;
-               }
-       }
-
-       return ret;
-}
-
-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;
-       int ret;
+       s64 sectors = (s64) k.k->size * replicas;
 
        if (flags & BTREE_TRIGGER_OVERWRITE)
                sectors = -sectors;
-       sectors *= replicas;
-
-       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));
-
-       d->persistent_reserved[replicas - 1] += sectors;
-       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)
-{
-       return trigger_run_overwrite_then_insert(__trans_mark_reservation, trans, btree_id, level, old, new, flags);
-}
+       if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+               int ret = bch2_replicas_deltas_realloc(trans, 0);
+               if (ret)
+                       return ret;
 
-static int trans_mark_reflink_p_segment(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_i *k;
-       __le64 *refcount;
-       int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
-       struct printbuf buf = PRINTBUF;
-       int ret;
+               struct replicas_delta_list *d = trans->fs_usage_deltas;
+               replicas = min(replicas, ARRAY_SIZE(d->persistent_reserved));
 
-       k = bch2_bkey_get_mut_noupdate(trans, &iter,
-                       BTREE_ID_reflink, POS(0, *idx),
-                       BTREE_ITER_WITH_UPDATES);
-       ret = PTR_ERR_OR_ZERO(k);
-       if (ret)
-               goto err;
-
-       refcount = bkey_refcount(k);
-       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;
+               d->persistent_reserved[replicas - 1] += sectors;
        }
 
-       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_GC) {
+               percpu_down_read(&c->mark_lock);
+               preempt_disable();
 
-       if (flags & BTREE_TRIGGER_INSERT) {
-               struct bch_reflink_p *v = (struct bch_reflink_p *) p.v;
-               u64 pad;
+               struct bch_fs_usage *fs_usage = this_cpu_ptr(c->usage_gc);
 
-               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);
+               replicas = min(replicas, ARRAY_SIZE(fs_usage->persistent_reserved));
+               fs_usage->b.reserved                            += sectors;
+               fs_usage->persistent_reserved[replicas - 1]     += sectors;
 
-               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);
+               preempt_enable();
+               percpu_up_read(&c->mark_lock);
        }
 
-       le64_add_cpu(refcount, add);
-
-       bch2_btree_iter_set_pos_to_extent_start(&iter);
-       ret = bch2_trans_update(trans, &iter, k, 0);
-       if (ret)
-               goto err;
-
-       *idx = k->k.p.offset;
-err:
-       bch2_trans_iter_exit(trans, &iter);
-       printbuf_exit(&buf);
-       return ret;
+       return 0;
 }
 
-static int __trans_mark_reflink_p(struct btree_trans *trans,
-                               enum btree_id btree_id, unsigned level,
-                               struct bkey_s_c k, unsigned flags)
+int bch2_trigger_reservation(struct btree_trans *trans,
+                         enum btree_id btree_id, unsigned level,
+                         struct bkey_s_c old, struct bkey_s new,
+                         unsigned flags)
 {
-       struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
-       u64 idx, end_idx;
-       int ret = 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 = trans_mark_reflink_p_segment(trans, p, &idx, flags);
-       return ret;
+       return trigger_run_overwrite_then_insert(__trigger_reservation, trans, btree_id, level, old, new, flags);
 }
 
-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)
-{
-       if (flags & BTREE_TRIGGER_INSERT) {
-               struct bch_reflink_p *v = &bkey_i_to_reflink_p(new)->v;
-
-               v->front_pad = v->back_pad = 0;
-       }
-
-       return trigger_run_overwrite_then_insert(__trans_mark_reflink_p, trans, btree_id, level, old, new, flags);
-}
+/* Mark superblocks: */
 
 static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
                                    struct bch_dev *ca, size_t b,
@@ -1857,12 +1139,13 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 
        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,
-                       bch2_data_types[a->v.data_type],
-                       bch2_data_types[type],
-                       bch2_data_types[type]);
+                       bch2_data_type_str(a->v.data_type),
+                       bch2_data_type_str(type),
+                       bch2_data_type_str(type));
                ret = -EIO;
                goto err;
        }
@@ -1963,17 +1246,13 @@ int bch2_trans_mark_dev_sb(struct bch_fs *c, struct bch_dev *ca)
 {
        int ret = bch2_trans_run(c, __bch2_trans_mark_dev_sb(trans, ca));
 
-       if (ret)
-               bch_err_fn(c, ret);
+       bch_err_fn(c, ret);
        return ret;
 }
 
 int bch2_trans_mark_dev_sbs(struct bch_fs *c)
 {
-       struct bch_dev *ca;
-       unsigned i;
-
-       for_each_online_member(ca, c, i) {
+       for_each_online_member(c, ca) {
                int ret = bch2_trans_mark_dev_sb(c, ca);
                if (ret) {
                        percpu_ref_put(&ca->ref);
@@ -2057,7 +1336,7 @@ static void bucket_gens_free_rcu(struct rcu_head *rcu)
        struct bucket_gens *buckets =
                container_of(rcu, struct bucket_gens, rcu);
 
-       kvpfree(buckets, sizeof(*buckets) + buckets->nbuckets);
+       kvfree(buckets);
 }
 
 int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
@@ -2067,16 +1346,16 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
        bool resize = ca->bucket_gens != NULL;
        int ret;
 
-       if (!(bucket_gens       = kvpmalloc(sizeof(struct bucket_gens) + nbuckets,
-                                           GFP_KERNEL|__GFP_ZERO))) {
+       if (!(bucket_gens       = kvmalloc(sizeof(struct bucket_gens) + nbuckets,
+                                          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)))) {
+            !(buckets_nouse    = kvmalloc(BITS_TO_LONGS(nbuckets) *
+                                          sizeof(unsigned long),
+                                          GFP_KERNEL|__GFP_ZERO)))) {
                ret = -BCH_ERR_ENOMEM_buckets_nouse;
                goto err;
        }
@@ -2084,8 +1363,6 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
        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);
@@ -2121,8 +1398,7 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 
        ret = 0;
 err:
-       kvpfree(buckets_nouse,
-               BITS_TO_LONGS(nbuckets) * sizeof(unsigned long));
+       kvfree(buckets_nouse);
        if (bucket_gens)
                call_rcu(&bucket_gens->rcu, bucket_gens_free_rcu);
 
@@ -2131,27 +1407,21 @@ err:
 
 void bch2_dev_buckets_free(struct bch_dev *ca)
 {
-       unsigned i;
+       kvfree(ca->buckets_nouse);
+       kvfree(rcu_dereference_protected(ca->bucket_gens, 1));
 
-       kvpfree(ca->buckets_nouse,
-               BITS_TO_LONGS(ca->mi.nbuckets) * sizeof(unsigned long));
-       kvpfree(rcu_dereference_protected(ca->bucket_gens, 1),
-               sizeof(struct bucket_gens) + ca->mi.nbuckets);
-
-       for (i = 0; i < ARRAY_SIZE(ca->usage); i++)
+       for (unsigned i = 0; i < ARRAY_SIZE(ca->usage); i++)
                free_percpu(ca->usage[i]);
        kfree(ca->usage_base);
 }
 
 int bch2_dev_buckets_alloc(struct bch_fs *c, struct bch_dev *ca)
 {
-       unsigned i;
-
        ca->usage_base = kzalloc(sizeof(struct bch_dev_usage), GFP_KERNEL);
        if (!ca->usage_base)
                return -BCH_ERR_ENOMEM_usage_init;
 
-       for (i = 0; i < ARRAY_SIZE(ca->usage); i++) {
+       for (unsigned i = 0; i < ARRAY_SIZE(ca->usage); i++) {
                ca->usage[i] = alloc_percpu(struct bch_dev_usage);
                if (!ca->usage[i])
                        return -BCH_ERR_ENOMEM_usage_init;