X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;ds=sidebyside;f=libbcachefs%2Freplicas.c;h=92ba56ef1fc89690656e9625871ecd7ee38b5f9b;hb=b5fd066153c40a70a29caa1ea7987723ab687763;hp=a7a4e280840b90c2ce755f34758b615b0c64cfe0;hpb=cc6479303f0672d5e07899009c64eb32821e170b;p=bcachefs-tools-debian

diff --git a/libbcachefs/replicas.c b/libbcachefs/replicas.c
index a7a4e28..92ba56e 100644
--- a/libbcachefs/replicas.c
+++ b/libbcachefs/replicas.c
@@ -1,378 +1,658 @@
+// SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
+#include "buckets.h"
+#include "journal.h"
 #include "replicas.h"
 #include "super-io.h"
 
-struct bch_replicas_entry_padded {
-	struct bch_replicas_entry	e;
-	u8				pad[BCH_SB_MEMBERS_MAX];
-};
-
 static int bch2_cpu_replicas_to_sb_replicas(struct bch_fs *,
 					    struct bch_replicas_cpu *);
 
 /* Replicas tracking - in memory: */
 
-static inline int u8_cmp(u8 l, u8 r)
+static void verify_replicas_entry(struct bch_replicas_entry_v1 *e)
 {
-	return (l > r) - (l < r);
+#ifdef CONFIG_BCACHEFS_DEBUG
+	unsigned i;
+
+	BUG_ON(e->data_type >= BCH_DATA_NR);
+	BUG_ON(!e->nr_devs);
+	BUG_ON(e->nr_required > 1 &&
+	       e->nr_required >= e->nr_devs);
+
+	for (i = 0; i + 1 < e->nr_devs; i++)
+		BUG_ON(e->devs[i] >= e->devs[i + 1]);
+#endif
 }
 
-static void replicas_entry_sort(struct bch_replicas_entry *e)
+void bch2_replicas_entry_sort(struct bch_replicas_entry_v1 *e)
 {
 	bubble_sort(e->devs, e->nr_devs, u8_cmp);
 }
 
-#define for_each_cpu_replicas_entry(_r, _i)				\
-	for (_i = (_r)->entries;					\
-	     (void *) (_i) < (void *) (_r)->entries + (_r)->nr * (_r)->entry_size;\
-	     _i = (void *) (_i) + (_r)->entry_size)
-
-static inline struct bch_replicas_entry *
-cpu_replicas_entry(struct bch_replicas_cpu *r, unsigned i)
+static void bch2_cpu_replicas_sort(struct bch_replicas_cpu *r)
 {
-	return (void *) r->entries + r->entry_size * i;
+	eytzinger0_sort(r->entries, r->nr, r->entry_size, memcmp, NULL);
 }
 
-static void bch2_cpu_replicas_sort(struct bch_replicas_cpu *r)
+static void bch2_replicas_entry_v0_to_text(struct printbuf *out,
+					   struct bch_replicas_entry_v0 *e)
 {
-	eytzinger0_sort(r->entries, r->nr, r->entry_size, memcmp, NULL);
+	unsigned i;
+
+	if (e->data_type < BCH_DATA_NR)
+		prt_printf(out, "%s", bch2_data_types[e->data_type]);
+	else
+		prt_printf(out, "(invalid data type %u)", e->data_type);
+
+	prt_printf(out, ": %u [", e->nr_devs);
+	for (i = 0; i < e->nr_devs; i++)
+		prt_printf(out, i ? " %u" : "%u", e->devs[i]);
+	prt_printf(out, "]");
 }
 
-static void replicas_entry_to_text(struct printbuf *out,
-				  struct bch_replicas_entry *e)
+void bch2_replicas_entry_to_text(struct printbuf *out,
+				 struct bch_replicas_entry_v1 *e)
 {
 	unsigned i;
 
-	pr_buf(out, "%s: %u/%u [",
-	       bch2_data_types[e->data_type],
-	       e->nr_required,
-	       e->nr_devs);
+	if (e->data_type < BCH_DATA_NR)
+		prt_printf(out, "%s", bch2_data_types[e->data_type]);
+	else
+		prt_printf(out, "(invalid data type %u)", e->data_type);
 
+	prt_printf(out, ": %u/%u [", e->nr_required, e->nr_devs);
 	for (i = 0; i < e->nr_devs; i++)
-		pr_buf(out, i ? " %u" : "%u", e->devs[i]);
-	pr_buf(out, "]");
+		prt_printf(out, i ? " %u" : "%u", e->devs[i]);
+	prt_printf(out, "]");
+}
+
+int bch2_replicas_entry_validate(struct bch_replicas_entry_v1 *r,
+				 struct bch_sb *sb,
+				 struct printbuf *err)
+{
+	if (!r->nr_devs) {
+		prt_printf(err, "no devices in entry ");
+		goto bad;
+	}
+
+	if (r->nr_required > 1 &&
+	    r->nr_required >= r->nr_devs) {
+		prt_printf(err, "bad nr_required in entry ");
+		goto bad;
+	}
+
+	for (unsigned i = 0; i < r->nr_devs; i++)
+		if (!bch2_dev_exists(sb, r->devs[i])) {
+			prt_printf(err, "invalid device %u in entry ", r->devs[i]);
+			goto bad;
+		}
+
+	return 0;
+bad:
+	bch2_replicas_entry_to_text(err, r);
+	return -BCH_ERR_invalid_replicas_entry;
 }
 
 void bch2_cpu_replicas_to_text(struct printbuf *out,
-			      struct bch_replicas_cpu *r)
+			       struct bch_replicas_cpu *r)
 {
-	struct bch_replicas_entry *e;
+	struct bch_replicas_entry_v1 *e;
 	bool first = true;
 
 	for_each_cpu_replicas_entry(r, e) {
 		if (!first)
-			pr_buf(out, " ");
+			prt_printf(out, " ");
 		first = false;
 
-		replicas_entry_to_text(out, e);
+		bch2_replicas_entry_to_text(out, e);
 	}
 }
 
 static void extent_to_replicas(struct bkey_s_c k,
-			       struct bch_replicas_entry *r)
+			       struct bch_replicas_entry_v1 *r)
 {
-	if (bkey_extent_is_data(k.k)) {
-		struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
-		const union bch_extent_entry *entry;
-		struct extent_ptr_decoded p;
+	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+	const union bch_extent_entry *entry;
+	struct extent_ptr_decoded p;
+
+	r->nr_required	= 1;
 
-		r->nr_required	= 1;
+	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
+		if (p.ptr.cached)
+			continue;
 
-		extent_for_each_ptr_decode(e, p, entry)
-			if (!p.ptr.cached)
-				r->devs[r->nr_devs++] = p.ptr.dev;
+		if (!p.has_ec)
+			r->devs[r->nr_devs++] = p.ptr.dev;
+		else
+			r->nr_required = 0;
 	}
 }
 
-static void bkey_to_replicas(enum bkey_type type,
-			     struct bkey_s_c k,
-			     struct bch_replicas_entry *e)
+static void stripe_to_replicas(struct bkey_s_c k,
+			       struct bch_replicas_entry_v1 *r)
+{
+	struct bkey_s_c_stripe s = bkey_s_c_to_stripe(k);
+	const struct bch_extent_ptr *ptr;
+
+	r->nr_required	= s.v->nr_blocks - s.v->nr_redundant;
+
+	for (ptr = s.v->ptrs;
+	     ptr < s.v->ptrs + s.v->nr_blocks;
+	     ptr++)
+		r->devs[r->nr_devs++] = ptr->dev;
+}
+
+void bch2_bkey_to_replicas(struct bch_replicas_entry_v1 *e,
+			   struct bkey_s_c k)
 {
 	e->nr_devs = 0;
 
-	switch (type) {
-	case BKEY_TYPE_BTREE:
-		e->data_type = BCH_DATA_BTREE;
+	switch (k.k->type) {
+	case KEY_TYPE_btree_ptr:
+	case KEY_TYPE_btree_ptr_v2:
+		e->data_type = BCH_DATA_btree;
 		extent_to_replicas(k, e);
 		break;
-	case BKEY_TYPE_EXTENTS:
-		e->data_type = BCH_DATA_USER;
+	case KEY_TYPE_extent:
+	case KEY_TYPE_reflink_v:
+		e->data_type = BCH_DATA_user;
 		extent_to_replicas(k, e);
 		break;
-	default:
+	case KEY_TYPE_stripe:
+		e->data_type = BCH_DATA_parity;
+		stripe_to_replicas(k, e);
 		break;
 	}
 
-	replicas_entry_sort(e);
+	bch2_replicas_entry_sort(e);
 }
 
-static inline void devlist_to_replicas(struct bch_devs_list devs,
-				       enum bch_data_type data_type,
-				       struct bch_replicas_entry *e)
+void bch2_devlist_to_replicas(struct bch_replicas_entry_v1 *e,
+			      enum bch_data_type data_type,
+			      struct bch_devs_list devs)
 {
-	unsigned i;
-
 	BUG_ON(!data_type ||
-	       data_type == BCH_DATA_SB ||
+	       data_type == BCH_DATA_sb ||
 	       data_type >= BCH_DATA_NR);
 
 	e->data_type	= data_type;
 	e->nr_devs	= 0;
 	e->nr_required	= 1;
 
-	for (i = 0; i < devs.nr; i++)
-		e->devs[e->nr_devs++] = devs.devs[i];
+	darray_for_each(devs, i)
+		e->devs[e->nr_devs++] = *i;
 
-	replicas_entry_sort(e);
+	bch2_replicas_entry_sort(e);
 }
 
-static struct bch_replicas_cpu *
-cpu_replicas_add_entry(struct bch_replicas_cpu *old,
-		       struct bch_replicas_entry *new_entry)
+static struct bch_replicas_cpu
+cpu_replicas_add_entry(struct bch_fs *c,
+		       struct bch_replicas_cpu *old,
+		       struct bch_replicas_entry_v1 *new_entry)
 {
-	struct bch_replicas_cpu *new;
-	unsigned i, nr, entry_size;
+	unsigned i;
+	struct bch_replicas_cpu new = {
+		.nr		= old->nr + 1,
+		.entry_size	= max_t(unsigned, old->entry_size,
+					replicas_entry_bytes(new_entry)),
+	};
 
-	entry_size = max_t(unsigned, old->entry_size,
-			   replicas_entry_bytes(new_entry));
-	nr = old->nr + 1;
+	for (i = 0; i < new_entry->nr_devs; i++)
+		BUG_ON(!bch2_dev_exists2(c, new_entry->devs[i]));
 
-	new = kzalloc(sizeof(struct bch_replicas_cpu) +
-		      nr * entry_size, GFP_NOIO);
-	if (!new)
-		return NULL;
+	BUG_ON(!new_entry->data_type);
+	verify_replicas_entry(new_entry);
 
-	new->nr		= nr;
-	new->entry_size	= entry_size;
+	new.entries = kcalloc(new.nr, new.entry_size, GFP_KERNEL);
+	if (!new.entries)
+		return new;
 
 	for (i = 0; i < old->nr; i++)
-		memcpy(cpu_replicas_entry(new, i),
+		memcpy(cpu_replicas_entry(&new, i),
 		       cpu_replicas_entry(old, i),
 		       old->entry_size);
 
-	memcpy(cpu_replicas_entry(new, old->nr),
+	memcpy(cpu_replicas_entry(&new, old->nr),
 	       new_entry,
 	       replicas_entry_bytes(new_entry));
 
-	bch2_cpu_replicas_sort(new);
+	bch2_cpu_replicas_sort(&new);
 	return new;
 }
 
+static inline int __replicas_entry_idx(struct bch_replicas_cpu *r,
+				       struct bch_replicas_entry_v1 *search)
+{
+	int idx, entry_size = replicas_entry_bytes(search);
+
+	if (unlikely(entry_size > r->entry_size))
+		return -1;
+
+	verify_replicas_entry(search);
+
+#define entry_cmp(_l, _r, size)	memcmp(_l, _r, entry_size)
+	idx = eytzinger0_find(r->entries, r->nr, r->entry_size,
+			      entry_cmp, search);
+#undef entry_cmp
+
+	return idx < r->nr ? idx : -1;
+}
+
+int bch2_replicas_entry_idx(struct bch_fs *c,
+			    struct bch_replicas_entry_v1 *search)
+{
+	bch2_replicas_entry_sort(search);
+
+	return __replicas_entry_idx(&c->replicas, search);
+}
+
 static bool __replicas_has_entry(struct bch_replicas_cpu *r,
-				 struct bch_replicas_entry *search)
+				 struct bch_replicas_entry_v1 *search)
 {
-	return replicas_entry_bytes(search) <= r->entry_size &&
-		eytzinger0_find(r->entries, r->nr,
-				r->entry_size,
-				memcmp, search) < r->nr;
+	return __replicas_entry_idx(r, search) >= 0;
 }
 
-static bool replicas_has_entry(struct bch_fs *c,
-			       struct bch_replicas_entry *search,
-			       bool check_gc_replicas)
+bool bch2_replicas_marked(struct bch_fs *c,
+			  struct bch_replicas_entry_v1 *search)
 {
-	struct bch_replicas_cpu *r, *gc_r;
 	bool marked;
 
-	rcu_read_lock();
-	r = rcu_dereference(c->replicas);
-	marked = __replicas_has_entry(r, search) &&
-		(!check_gc_replicas ||
-		 likely(!(gc_r = rcu_dereference(c->replicas_gc))) ||
-		 __replicas_has_entry(gc_r, search));
-	rcu_read_unlock();
+	if (!search->nr_devs)
+		return true;
+
+	verify_replicas_entry(search);
+
+	percpu_down_read(&c->mark_lock);
+	marked = __replicas_has_entry(&c->replicas, search) &&
+		(likely((!c->replicas_gc.entries)) ||
+		 __replicas_has_entry(&c->replicas_gc, search));
+	percpu_up_read(&c->mark_lock);
 
 	return marked;
 }
 
+static void __replicas_table_update(struct bch_fs_usage *dst,
+				    struct bch_replicas_cpu *dst_r,
+				    struct bch_fs_usage *src,
+				    struct bch_replicas_cpu *src_r)
+{
+	int src_idx, dst_idx;
+
+	*dst = *src;
+
+	for (src_idx = 0; src_idx < src_r->nr; src_idx++) {
+		if (!src->replicas[src_idx])
+			continue;
+
+		dst_idx = __replicas_entry_idx(dst_r,
+				cpu_replicas_entry(src_r, src_idx));
+		BUG_ON(dst_idx < 0);
+
+		dst->replicas[dst_idx] = src->replicas[src_idx];
+	}
+}
+
+static void __replicas_table_update_pcpu(struct bch_fs_usage __percpu *dst_p,
+				    struct bch_replicas_cpu *dst_r,
+				    struct bch_fs_usage __percpu *src_p,
+				    struct bch_replicas_cpu *src_r)
+{
+	unsigned src_nr = sizeof(struct bch_fs_usage) / sizeof(u64) + src_r->nr;
+	struct bch_fs_usage *dst, *src = (void *)
+		bch2_acc_percpu_u64s((u64 __percpu *) src_p, src_nr);
+
+	preempt_disable();
+	dst = this_cpu_ptr(dst_p);
+	preempt_enable();
+
+	__replicas_table_update(dst, dst_r, src, src_r);
+}
+
+/*
+ * Resize filesystem accounting:
+ */
+static int replicas_table_update(struct bch_fs *c,
+				 struct bch_replicas_cpu *new_r)
+{
+	struct bch_fs_usage __percpu *new_usage[JOURNAL_BUF_NR];
+	struct bch_fs_usage_online *new_scratch = NULL;
+	struct bch_fs_usage __percpu *new_gc = NULL;
+	struct bch_fs_usage *new_base = NULL;
+	unsigned i, bytes = sizeof(struct bch_fs_usage) +
+		sizeof(u64) * new_r->nr;
+	unsigned scratch_bytes = sizeof(struct bch_fs_usage_online) +
+		sizeof(u64) * new_r->nr;
+	int ret = 0;
+
+	memset(new_usage, 0, sizeof(new_usage));
+
+	for (i = 0; i < ARRAY_SIZE(new_usage); i++)
+		if (!(new_usage[i] = __alloc_percpu_gfp(bytes,
+					sizeof(u64), GFP_KERNEL)))
+			goto err;
+
+	if (!(new_base = kzalloc(bytes, GFP_KERNEL)) ||
+	    !(new_scratch  = kmalloc(scratch_bytes, GFP_KERNEL)) ||
+	    (c->usage_gc &&
+	     !(new_gc = __alloc_percpu_gfp(bytes, sizeof(u64), GFP_KERNEL))))
+		goto err;
+
+	for (i = 0; i < ARRAY_SIZE(new_usage); i++)
+		if (c->usage[i])
+			__replicas_table_update_pcpu(new_usage[i], new_r,
+						     c->usage[i], &c->replicas);
+	if (c->usage_base)
+		__replicas_table_update(new_base,		new_r,
+					c->usage_base,		&c->replicas);
+	if (c->usage_gc)
+		__replicas_table_update_pcpu(new_gc,		new_r,
+					     c->usage_gc,	&c->replicas);
+
+	for (i = 0; i < ARRAY_SIZE(new_usage); i++)
+		swap(c->usage[i],	new_usage[i]);
+	swap(c->usage_base,	new_base);
+	swap(c->usage_scratch,	new_scratch);
+	swap(c->usage_gc,	new_gc);
+	swap(c->replicas,	*new_r);
+out:
+	free_percpu(new_gc);
+	kfree(new_scratch);
+	for (i = 0; i < ARRAY_SIZE(new_usage); i++)
+		free_percpu(new_usage[i]);
+	kfree(new_base);
+	return ret;
+err:
+	bch_err(c, "error updating replicas table: memory allocation failure");
+	ret = -BCH_ERR_ENOMEM_replicas_table;
+	goto out;
+}
+
+static unsigned reserve_journal_replicas(struct bch_fs *c,
+				     struct bch_replicas_cpu *r)
+{
+	struct bch_replicas_entry_v1 *e;
+	unsigned journal_res_u64s = 0;
+
+	/* nr_inodes: */
+	journal_res_u64s +=
+		DIV_ROUND_UP(sizeof(struct jset_entry_usage), sizeof(u64));
+
+	/* key_version: */
+	journal_res_u64s +=
+		DIV_ROUND_UP(sizeof(struct jset_entry_usage), sizeof(u64));
+
+	/* persistent_reserved: */
+	journal_res_u64s +=
+		DIV_ROUND_UP(sizeof(struct jset_entry_usage), sizeof(u64)) *
+		BCH_REPLICAS_MAX;
+
+	for_each_cpu_replicas_entry(r, e)
+		journal_res_u64s +=
+			DIV_ROUND_UP(sizeof(struct jset_entry_data_usage) +
+				     e->nr_devs, sizeof(u64));
+	return journal_res_u64s;
+}
+
 noinline
 static int bch2_mark_replicas_slowpath(struct bch_fs *c,
-				struct bch_replicas_entry *new_entry)
+				struct bch_replicas_entry_v1 *new_entry)
 {
-	struct bch_replicas_cpu *old_gc, *new_gc = NULL, *old_r, *new_r = NULL;
-	int ret = -ENOMEM;
+	struct bch_replicas_cpu new_r, new_gc;
+	int ret = 0;
+
+	verify_replicas_entry(new_entry);
+
+	memset(&new_r, 0, sizeof(new_r));
+	memset(&new_gc, 0, sizeof(new_gc));
 
 	mutex_lock(&c->sb_lock);
 
-	old_gc = rcu_dereference_protected(c->replicas_gc,
-					   lockdep_is_held(&c->sb_lock));
-	if (old_gc && !__replicas_has_entry(old_gc, new_entry)) {
-		new_gc = cpu_replicas_add_entry(old_gc, new_entry);
-		if (!new_gc)
+	if (c->replicas_gc.entries &&
+	    !__replicas_has_entry(&c->replicas_gc, new_entry)) {
+		new_gc = cpu_replicas_add_entry(c, &c->replicas_gc, new_entry);
+		if (!new_gc.entries) {
+			ret = -BCH_ERR_ENOMEM_cpu_replicas;
 			goto err;
+		}
 	}
 
-	old_r = rcu_dereference_protected(c->replicas,
-					  lockdep_is_held(&c->sb_lock));
-	if (!__replicas_has_entry(old_r, new_entry)) {
-		new_r = cpu_replicas_add_entry(old_r, new_entry);
-		if (!new_r)
+	if (!__replicas_has_entry(&c->replicas, new_entry)) {
+		new_r = cpu_replicas_add_entry(c, &c->replicas, new_entry);
+		if (!new_r.entries) {
+			ret = -BCH_ERR_ENOMEM_cpu_replicas;
 			goto err;
+		}
 
-		ret = bch2_cpu_replicas_to_sb_replicas(c, new_r);
+		ret = bch2_cpu_replicas_to_sb_replicas(c, &new_r);
 		if (ret)
 			goto err;
+
+		bch2_journal_entry_res_resize(&c->journal,
+				&c->replicas_journal_res,
+				reserve_journal_replicas(c, &new_r));
 	}
 
+	if (!new_r.entries &&
+	    !new_gc.entries)
+		goto out;
+
 	/* allocations done, now commit: */
 
-	if (new_r)
+	if (new_r.entries)
 		bch2_write_super(c);
 
 	/* don't update in memory replicas until changes are persistent */
+	percpu_down_write(&c->mark_lock);
+	if (new_r.entries)
+		ret = replicas_table_update(c, &new_r);
+	if (new_gc.entries)
+		swap(new_gc, c->replicas_gc);
+	percpu_up_write(&c->mark_lock);
+out:
+	mutex_unlock(&c->sb_lock);
 
-	if (new_gc) {
-		rcu_assign_pointer(c->replicas_gc, new_gc);
-		kfree_rcu(old_gc, rcu);
-	}
-
-	if (new_r) {
-		rcu_assign_pointer(c->replicas, new_r);
-		kfree_rcu(old_r, rcu);
-	}
+	kfree(new_r.entries);
+	kfree(new_gc.entries);
 
-	mutex_unlock(&c->sb_lock);
-	return 0;
-err:
-	mutex_unlock(&c->sb_lock);
-	kfree(new_gc);
-	kfree(new_r);
 	return ret;
+err:
+	bch_err_msg(c, ret, "adding replicas entry");
+	goto out;
 }
 
-static int __bch2_mark_replicas(struct bch_fs *c,
-				struct bch_replicas_entry *devs)
+int bch2_mark_replicas(struct bch_fs *c, struct bch_replicas_entry_v1 *r)
 {
-	return likely(replicas_has_entry(c, devs, true))
-		? 0
-		: bch2_mark_replicas_slowpath(c, devs);
+	return likely(bch2_replicas_marked(c, r))
+		? 0 : bch2_mark_replicas_slowpath(c, r);
 }
 
-int bch2_mark_replicas(struct bch_fs *c,
-		       enum bch_data_type data_type,
-		       struct bch_devs_list devs)
-{
-	struct bch_replicas_entry_padded search;
+/* replicas delta list: */
 
-	if (!devs.nr)
-		return 0;
+int bch2_replicas_delta_list_mark(struct bch_fs *c,
+				  struct replicas_delta_list *r)
+{
+	struct replicas_delta *d = r->d;
+	struct replicas_delta *top = (void *) r->d + r->used;
+	int ret = 0;
 
-	memset(&search, 0, sizeof(search));
+	for (d = r->d; !ret && d != top; d = replicas_delta_next(d))
+		ret = bch2_mark_replicas(c, &d->r);
+	return ret;
+}
 
-	BUG_ON(devs.nr >= BCH_REPLICAS_MAX);
+/*
+ * Old replicas_gc mechanism: only used for journal replicas entries now, should
+ * die at some point:
+ */
 
-	devlist_to_replicas(devs, data_type, &search.e);
+int bch2_replicas_gc_end(struct bch_fs *c, int ret)
+{
+	lockdep_assert_held(&c->replicas_gc_lock);
 
-	return __bch2_mark_replicas(c, &search.e);
-}
+	mutex_lock(&c->sb_lock);
+	percpu_down_write(&c->mark_lock);
 
-int bch2_mark_bkey_replicas(struct bch_fs *c,
-			    enum bkey_type type,
-			    struct bkey_s_c k)
-{
-	struct bch_replicas_entry_padded search;
-	int ret;
+	ret =   ret ?:
+		bch2_cpu_replicas_to_sb_replicas(c, &c->replicas_gc) ?:
+		replicas_table_update(c, &c->replicas_gc);
 
-	memset(&search, 0, sizeof(search));
+	kfree(c->replicas_gc.entries);
+	c->replicas_gc.entries = NULL;
 
-	if (type == BKEY_TYPE_EXTENTS) {
-		struct bch_devs_list cached = bch2_bkey_cached_devs(k);
-		unsigned i;
+	percpu_up_write(&c->mark_lock);
 
-		for (i = 0; i < cached.nr; i++)
-			if ((ret = bch2_mark_replicas(c, BCH_DATA_CACHED,
-						bch2_dev_list_single(cached.devs[i]))))
-				return ret;
-	}
+	if (!ret)
+		bch2_write_super(c);
 
-	bkey_to_replicas(type, k, &search.e);
+	mutex_unlock(&c->sb_lock);
 
-	return search.e.nr_devs
-		? __bch2_mark_replicas(c, &search.e)
-		: 0;
+	return ret;
 }
 
-int bch2_replicas_gc_end(struct bch_fs *c, int ret)
+int bch2_replicas_gc_start(struct bch_fs *c, unsigned typemask)
 {
-	struct bch_replicas_cpu *new_r, *old_r;
+	struct bch_replicas_entry_v1 *e;
+	unsigned i = 0;
 
 	lockdep_assert_held(&c->replicas_gc_lock);
 
 	mutex_lock(&c->sb_lock);
+	BUG_ON(c->replicas_gc.entries);
 
-	new_r = rcu_dereference_protected(c->replicas_gc,
-					  lockdep_is_held(&c->sb_lock));
-	rcu_assign_pointer(c->replicas_gc, NULL);
+	c->replicas_gc.nr		= 0;
+	c->replicas_gc.entry_size	= 0;
 
-	if (ret)
-		goto err;
+	for_each_cpu_replicas_entry(&c->replicas, e)
+		if (!((1 << e->data_type) & typemask)) {
+			c->replicas_gc.nr++;
+			c->replicas_gc.entry_size =
+				max_t(unsigned, c->replicas_gc.entry_size,
+				      replicas_entry_bytes(e));
+		}
 
-	if (bch2_cpu_replicas_to_sb_replicas(c, new_r)) {
-		ret = -ENOSPC;
-		goto err;
+	c->replicas_gc.entries = kcalloc(c->replicas_gc.nr,
+					 c->replicas_gc.entry_size,
+					 GFP_KERNEL);
+	if (!c->replicas_gc.entries) {
+		mutex_unlock(&c->sb_lock);
+		bch_err(c, "error allocating c->replicas_gc");
+		return -BCH_ERR_ENOMEM_replicas_gc;
 	}
 
-	bch2_write_super(c);
-
-	/* don't update in memory replicas until changes are persistent */
-
-	old_r = rcu_dereference_protected(c->replicas,
-					  lockdep_is_held(&c->sb_lock));
+	for_each_cpu_replicas_entry(&c->replicas, e)
+		if (!((1 << e->data_type) & typemask))
+			memcpy(cpu_replicas_entry(&c->replicas_gc, i++),
+			       e, c->replicas_gc.entry_size);
 
-	rcu_assign_pointer(c->replicas, new_r);
-	kfree_rcu(old_r, rcu);
-out:
+	bch2_cpu_replicas_sort(&c->replicas_gc);
 	mutex_unlock(&c->sb_lock);
-	return ret;
-err:
-	kfree_rcu(new_r, rcu);
-	goto out;
+
+	return 0;
 }
 
-int bch2_replicas_gc_start(struct bch_fs *c, unsigned typemask)
+/*
+ * New much simpler mechanism for clearing out unneeded replicas entries - drop
+ * replicas entries that have 0 sectors used.
+ *
+ * However, we don't track sector counts for journal usage, so this doesn't drop
+ * any BCH_DATA_journal entries; the old bch2_replicas_gc_(start|end) mechanism
+ * is retained for that.
+ */
+int bch2_replicas_gc2(struct bch_fs *c)
 {
-	struct bch_replicas_cpu *dst, *src;
-	struct bch_replicas_entry *e;
-
-	lockdep_assert_held(&c->replicas_gc_lock);
+	struct bch_replicas_cpu new = { 0 };
+	unsigned i, nr;
+	int ret = 0;
+
+	bch2_journal_meta(&c->journal);
+retry:
+	nr		= READ_ONCE(c->replicas.nr);
+	new.entry_size	= READ_ONCE(c->replicas.entry_size);
+	new.entries	= kcalloc(nr, new.entry_size, GFP_KERNEL);
+	if (!new.entries) {
+		bch_err(c, "error allocating c->replicas_gc");
+		return -BCH_ERR_ENOMEM_replicas_gc;
+	}
 
 	mutex_lock(&c->sb_lock);
-	BUG_ON(c->replicas_gc);
+	percpu_down_write(&c->mark_lock);
 
-	src = rcu_dereference_protected(c->replicas,
-					lockdep_is_held(&c->sb_lock));
-
-	dst = kzalloc(sizeof(struct bch_replicas_cpu) +
-		      src->nr * src->entry_size, GFP_NOIO);
-	if (!dst) {
+	if (nr			!= c->replicas.nr ||
+	    new.entry_size	!= c->replicas.entry_size) {
+		percpu_up_write(&c->mark_lock);
 		mutex_unlock(&c->sb_lock);
-		return -ENOMEM;
+		kfree(new.entries);
+		goto retry;
+	}
+
+	for (i = 0; i < c->replicas.nr; i++) {
+		struct bch_replicas_entry_v1 *e =
+			cpu_replicas_entry(&c->replicas, i);
+
+		if (e->data_type == BCH_DATA_journal ||
+		    c->usage_base->replicas[i] ||
+		    percpu_u64_get(&c->usage[0]->replicas[i]) ||
+		    percpu_u64_get(&c->usage[1]->replicas[i]) ||
+		    percpu_u64_get(&c->usage[2]->replicas[i]) ||
+		    percpu_u64_get(&c->usage[3]->replicas[i]))
+			memcpy(cpu_replicas_entry(&new, new.nr++),
+			       e, new.entry_size);
 	}
 
-	dst->nr		= 0;
-	dst->entry_size	= src->entry_size;
+	bch2_cpu_replicas_sort(&new);
 
-	for_each_cpu_replicas_entry(src, e)
-		if (!((1 << e->data_type) & typemask))
-			memcpy(cpu_replicas_entry(dst, dst->nr++),
-			       e, dst->entry_size);
+	ret =   bch2_cpu_replicas_to_sb_replicas(c, &new) ?:
+		replicas_table_update(c, &new);
 
-	bch2_cpu_replicas_sort(dst);
+	kfree(new.entries);
+
+	percpu_up_write(&c->mark_lock);
+
+	if (!ret)
+		bch2_write_super(c);
 
-	rcu_assign_pointer(c->replicas_gc, dst);
 	mutex_unlock(&c->sb_lock);
 
+	return ret;
+}
+
+int bch2_replicas_set_usage(struct bch_fs *c,
+			    struct bch_replicas_entry_v1 *r,
+			    u64 sectors)
+{
+	int ret, idx = bch2_replicas_entry_idx(c, r);
+
+	if (idx < 0) {
+		struct bch_replicas_cpu n;
+
+		n = cpu_replicas_add_entry(c, &c->replicas, r);
+		if (!n.entries)
+			return -BCH_ERR_ENOMEM_cpu_replicas;
+
+		ret = replicas_table_update(c, &n);
+		if (ret)
+			return ret;
+
+		kfree(n.entries);
+
+		idx = bch2_replicas_entry_idx(c, r);
+		BUG_ON(ret < 0);
+	}
+
+	c->usage_base->replicas[idx] = sectors;
+
 	return 0;
 }
 
 /* Replicas tracking - superblock: */
 
-static struct bch_replicas_cpu *
-__bch2_sb_replicas_to_cpu_replicas(struct bch_sb_field_replicas *sb_r)
+static int
+__bch2_sb_replicas_to_cpu_replicas(struct bch_sb_field_replicas *sb_r,
+				   struct bch_replicas_cpu *cpu_r)
 {
-	struct bch_replicas_entry *e, *dst;
-	struct bch_replicas_cpu *cpu_r;
+	struct bch_replicas_entry_v1 *e, *dst;
 	unsigned nr = 0, entry_size = 0, idx = 0;
 
 	for_each_replicas_entry(sb_r, e) {
@@ -381,10 +661,9 @@ __bch2_sb_replicas_to_cpu_replicas(struct bch_sb_field_replicas *sb_r)
 		nr++;
 	}
 
-	cpu_r = kzalloc(sizeof(struct bch_replicas_cpu) +
-			nr * entry_size, GFP_NOIO);
-	if (!cpu_r)
-		return NULL;
+	cpu_r->entries = kcalloc(nr, entry_size, GFP_KERNEL);
+	if (!cpu_r->entries)
+		return -BCH_ERR_ENOMEM_cpu_replicas;
 
 	cpu_r->nr		= nr;
 	cpu_r->entry_size	= entry_size;
@@ -392,17 +671,17 @@ __bch2_sb_replicas_to_cpu_replicas(struct bch_sb_field_replicas *sb_r)
 	for_each_replicas_entry(sb_r, e) {
 		dst = cpu_replicas_entry(cpu_r, idx++);
 		memcpy(dst, e, replicas_entry_bytes(e));
-		replicas_entry_sort(dst);
+		bch2_replicas_entry_sort(dst);
 	}
 
-	return cpu_r;
+	return 0;
 }
 
-static struct bch_replicas_cpu *
-__bch2_sb_replicas_v0_to_cpu_replicas(struct bch_sb_field_replicas_v0 *sb_r)
+static int
+__bch2_sb_replicas_v0_to_cpu_replicas(struct bch_sb_field_replicas_v0 *sb_r,
+				      struct bch_replicas_cpu *cpu_r)
 {
 	struct bch_replicas_entry_v0 *e;
-	struct bch_replicas_cpu *cpu_r;
 	unsigned nr = 0, entry_size = 0, idx = 0;
 
 	for_each_replicas_entry(sb_r, e) {
@@ -411,53 +690,52 @@ __bch2_sb_replicas_v0_to_cpu_replicas(struct bch_sb_field_replicas_v0 *sb_r)
 		nr++;
 	}
 
-	entry_size += sizeof(struct bch_replicas_entry) -
+	entry_size += sizeof(struct bch_replicas_entry_v1) -
 		sizeof(struct bch_replicas_entry_v0);
 
-	cpu_r = kzalloc(sizeof(struct bch_replicas_cpu) +
-			nr * entry_size, GFP_NOIO);
-	if (!cpu_r)
-		return NULL;
+	cpu_r->entries = kcalloc(nr, entry_size, GFP_KERNEL);
+	if (!cpu_r->entries)
+		return -BCH_ERR_ENOMEM_cpu_replicas;
 
 	cpu_r->nr		= nr;
 	cpu_r->entry_size	= entry_size;
 
 	for_each_replicas_entry(sb_r, e) {
-		struct bch_replicas_entry *dst =
+		struct bch_replicas_entry_v1 *dst =
 			cpu_replicas_entry(cpu_r, idx++);
 
 		dst->data_type	= e->data_type;
 		dst->nr_devs	= e->nr_devs;
 		dst->nr_required = 1;
 		memcpy(dst->devs, e->devs, e->nr_devs);
-		replicas_entry_sort(dst);
+		bch2_replicas_entry_sort(dst);
 	}
 
-	return cpu_r;
+	return 0;
 }
 
 int bch2_sb_replicas_to_cpu_replicas(struct bch_fs *c)
 {
 	struct bch_sb_field_replicas *sb_v1;
 	struct bch_sb_field_replicas_v0 *sb_v0;
-	struct bch_replicas_cpu *cpu_r, *old_r;
+	struct bch_replicas_cpu new_r = { 0, 0, NULL };
+	int ret = 0;
 
-	if ((sb_v1 = bch2_sb_get_replicas(c->disk_sb.sb)))
-		cpu_r = __bch2_sb_replicas_to_cpu_replicas(sb_v1);
-	else if ((sb_v0 = bch2_sb_get_replicas_v0(c->disk_sb.sb)))
-		cpu_r = __bch2_sb_replicas_v0_to_cpu_replicas(sb_v0);
-	else
-		cpu_r = kzalloc(sizeof(struct bch_replicas_cpu), GFP_NOIO);
+	if ((sb_v1 = bch2_sb_field_get(c->disk_sb.sb, replicas)))
+		ret = __bch2_sb_replicas_to_cpu_replicas(sb_v1, &new_r);
+	else if ((sb_v0 = bch2_sb_field_get(c->disk_sb.sb, replicas_v0)))
+		ret = __bch2_sb_replicas_v0_to_cpu_replicas(sb_v0, &new_r);
+	if (ret)
+		return ret;
+
+	bch2_cpu_replicas_sort(&new_r);
 
-	if (!cpu_r)
-		return -ENOMEM;
+	percpu_down_write(&c->mark_lock);
 
-	bch2_cpu_replicas_sort(cpu_r);
+	ret = replicas_table_update(c, &new_r);
+	percpu_up_write(&c->mark_lock);
 
-	old_r = rcu_dereference_check(c->replicas, lockdep_is_held(&c->sb_lock));
-	rcu_assign_pointer(c->replicas, cpu_r);
-	if (old_r)
-		kfree_rcu(old_r, rcu);
+	kfree(new_r.entries);
 
 	return 0;
 }
@@ -467,7 +745,7 @@ static int bch2_cpu_replicas_to_sb_replicas_v0(struct bch_fs *c,
 {
 	struct bch_sb_field_replicas_v0 *sb_r;
 	struct bch_replicas_entry_v0 *dst;
-	struct bch_replicas_entry *src;
+	struct bch_replicas_entry_v1 *src;
 	size_t bytes;
 
 	bytes = sizeof(struct bch_sb_field_replicas);
@@ -475,13 +753,13 @@ static int bch2_cpu_replicas_to_sb_replicas_v0(struct bch_fs *c,
 	for_each_cpu_replicas_entry(r, src)
 		bytes += replicas_entry_bytes(src) - 1;
 
-	sb_r = bch2_sb_resize_replicas_v0(&c->disk_sb,
+	sb_r = bch2_sb_field_resize(&c->disk_sb, replicas_v0,
 			DIV_ROUND_UP(bytes, sizeof(u64)));
 	if (!sb_r)
-		return -ENOSPC;
+		return -BCH_ERR_ENOSPC_sb_replicas;
 
 	bch2_sb_field_delete(&c->disk_sb, BCH_SB_FIELD_replicas);
-	sb_r = bch2_sb_get_replicas_v0(c->disk_sb.sb);
+	sb_r = bch2_sb_field_get(c->disk_sb.sb, replicas_v0);
 
 	memset(&sb_r->entries, 0,
 	       vstruct_end(&sb_r->field) -
@@ -505,7 +783,7 @@ static int bch2_cpu_replicas_to_sb_replicas(struct bch_fs *c,
 					    struct bch_replicas_cpu *r)
 {
 	struct bch_sb_field_replicas *sb_r;
-	struct bch_replicas_entry *dst, *src;
+	struct bch_replicas_entry_v1 *dst, *src;
 	bool need_v1 = false;
 	size_t bytes;
 
@@ -520,13 +798,13 @@ static int bch2_cpu_replicas_to_sb_replicas(struct bch_fs *c,
 	if (!need_v1)
 		return bch2_cpu_replicas_to_sb_replicas_v0(c, r);
 
-	sb_r = bch2_sb_resize_replicas(&c->disk_sb,
+	sb_r = bch2_sb_field_resize(&c->disk_sb, replicas,
 			DIV_ROUND_UP(bytes, sizeof(u64)));
 	if (!sb_r)
-		return -ENOSPC;
+		return -BCH_ERR_ENOSPC_sb_replicas;
 
 	bch2_sb_field_delete(&c->disk_sb, BCH_SB_FIELD_replicas_v0);
-	sb_r = bch2_sb_get_replicas(c->disk_sb.sb);
+	sb_r = bch2_sb_field_get(c->disk_sb.sb, replicas);
 
 	memset(&sb_r->entries, 0,
 	       vstruct_end(&sb_r->field) -
@@ -544,7 +822,9 @@ static int bch2_cpu_replicas_to_sb_replicas(struct bch_fs *c,
 	return 0;
 }
 
-static const char *check_dup_replicas_entries(struct bch_replicas_cpu *cpu_r)
+static int bch2_cpu_replicas_validate(struct bch_replicas_cpu *cpu_r,
+				      struct bch_sb *sb,
+				      struct printbuf *err)
 {
 	unsigned i;
 
@@ -553,60 +833,45 @@ static const char *check_dup_replicas_entries(struct bch_replicas_cpu *cpu_r)
 		      cpu_r->entry_size,
 		      memcmp, NULL);
 
-	for (i = 0; i + 1 < cpu_r->nr; i++) {
-		struct bch_replicas_entry *l =
+	for (i = 0; i < cpu_r->nr; i++) {
+		struct bch_replicas_entry_v1 *e =
 			cpu_replicas_entry(cpu_r, i);
-		struct bch_replicas_entry *r =
-			cpu_replicas_entry(cpu_r, i + 1);
 
-		BUG_ON(memcmp(l, r, cpu_r->entry_size) > 0);
+		int ret = bch2_replicas_entry_validate(e, sb, err);
+		if (ret)
+			return ret;
+
+		if (i + 1 < cpu_r->nr) {
+			struct bch_replicas_entry_v1 *n =
+				cpu_replicas_entry(cpu_r, i + 1);
 
-		if (!memcmp(l, r, cpu_r->entry_size))
-			return "duplicate replicas entry";
+			BUG_ON(memcmp(e, n, cpu_r->entry_size) > 0);
+
+			if (!memcmp(e, n, cpu_r->entry_size)) {
+				prt_printf(err, "duplicate replicas entry ");
+				bch2_replicas_entry_to_text(err, e);
+				return -BCH_ERR_invalid_sb_replicas;
+			}
+		}
 	}
 
-	return NULL;
+	return 0;
 }
 
-static const char *bch2_sb_validate_replicas(struct bch_sb *sb, struct bch_sb_field *f)
+static int bch2_sb_replicas_validate(struct bch_sb *sb, struct bch_sb_field *f,
+				     struct printbuf *err)
 {
 	struct bch_sb_field_replicas *sb_r = field_to_type(f, replicas);
-	struct bch_sb_field_members *mi = bch2_sb_get_members(sb);
-	struct bch_replicas_cpu *cpu_r = NULL;
-	struct bch_replicas_entry *e;
-	const char *err;
-	unsigned i;
-
-	for_each_replicas_entry(sb_r, e) {
-		err = "invalid replicas entry: invalid data type";
-		if (e->data_type >= BCH_DATA_NR)
-			goto err;
-
-		err = "invalid replicas entry: no devices";
-		if (!e->nr_devs)
-			goto err;
-
-		err = "invalid replicas entry: bad nr_required";
-		if (!e->nr_required ||
-		    (e->nr_required > 1 &&
-		     e->nr_required >= e->nr_devs))
-			goto err;
-
-		err = "invalid replicas entry: invalid device";
-		for (i = 0; i < e->nr_devs; i++)
-			if (!bch2_dev_exists(sb, mi, e->devs[i]))
-				goto err;
-	}
+	struct bch_replicas_cpu cpu_r;
+	int ret;
 
-	err = "cannot allocate memory";
-	cpu_r = __bch2_sb_replicas_to_cpu_replicas(sb_r);
-	if (!cpu_r)
-		goto err;
+	ret = __bch2_sb_replicas_to_cpu_replicas(sb_r, &cpu_r);
+	if (ret)
+		return ret;
 
-	err = check_dup_replicas_entries(cpu_r);
-err:
-	kfree(cpu_r);
-	return err;
+	ret = bch2_cpu_replicas_validate(&cpu_r, sb, err);
+	kfree(cpu_r.entries);
+	return ret;
 }
 
 static void bch2_sb_replicas_to_text(struct printbuf *out,
@@ -614,212 +879,179 @@ static void bch2_sb_replicas_to_text(struct printbuf *out,
 				     struct bch_sb_field *f)
 {
 	struct bch_sb_field_replicas *r = field_to_type(f, replicas);
-	struct bch_replicas_entry *e;
+	struct bch_replicas_entry_v1 *e;
 	bool first = true;
 
 	for_each_replicas_entry(r, e) {
 		if (!first)
-			pr_buf(out, " ");
+			prt_printf(out, " ");
 		first = false;
 
-		replicas_entry_to_text(out, e);
+		bch2_replicas_entry_to_text(out, e);
 	}
+	prt_newline(out);
 }
 
 const struct bch_sb_field_ops bch_sb_field_ops_replicas = {
-	.validate	= bch2_sb_validate_replicas,
+	.validate	= bch2_sb_replicas_validate,
 	.to_text	= bch2_sb_replicas_to_text,
 };
 
-static const char *bch2_sb_validate_replicas_v0(struct bch_sb *sb, struct bch_sb_field *f)
+static int bch2_sb_replicas_v0_validate(struct bch_sb *sb, struct bch_sb_field *f,
+					struct printbuf *err)
 {
 	struct bch_sb_field_replicas_v0 *sb_r = field_to_type(f, replicas_v0);
-	struct bch_sb_field_members *mi = bch2_sb_get_members(sb);
-	struct bch_replicas_cpu *cpu_r = NULL;
-	struct bch_replicas_entry_v0 *e;
-	const char *err;
-	unsigned i;
+	struct bch_replicas_cpu cpu_r;
+	int ret;
 
-	for_each_replicas_entry_v0(sb_r, e) {
-		err = "invalid replicas entry: invalid data type";
-		if (e->data_type >= BCH_DATA_NR)
-			goto err;
+	ret = __bch2_sb_replicas_v0_to_cpu_replicas(sb_r, &cpu_r);
+	if (ret)
+		return ret;
 
-		err = "invalid replicas entry: no devices";
-		if (!e->nr_devs)
-			goto err;
+	ret = bch2_cpu_replicas_validate(&cpu_r, sb, err);
+	kfree(cpu_r.entries);
+	return ret;
+}
 
-		err = "invalid replicas entry: invalid device";
-		for (i = 0; i < e->nr_devs; i++)
-			if (!bch2_dev_exists(sb, mi, e->devs[i]))
-				goto err;
-	}
+static void bch2_sb_replicas_v0_to_text(struct printbuf *out,
+					struct bch_sb *sb,
+					struct bch_sb_field *f)
+{
+	struct bch_sb_field_replicas_v0 *sb_r = field_to_type(f, replicas_v0);
+	struct bch_replicas_entry_v0 *e;
+	bool first = true;
 
-	err = "cannot allocate memory";
-	cpu_r = __bch2_sb_replicas_v0_to_cpu_replicas(sb_r);
-	if (!cpu_r)
-		goto err;
+	for_each_replicas_entry(sb_r, e) {
+		if (!first)
+			prt_printf(out, " ");
+		first = false;
 
-	err = check_dup_replicas_entries(cpu_r);
-err:
-	kfree(cpu_r);
-	return err;
+		bch2_replicas_entry_v0_to_text(out, e);
+	}
+	prt_newline(out);
 }
 
 const struct bch_sb_field_ops bch_sb_field_ops_replicas_v0 = {
-	.validate	= bch2_sb_validate_replicas_v0,
+	.validate	= bch2_sb_replicas_v0_validate,
+	.to_text	= bch2_sb_replicas_v0_to_text,
 };
 
 /* Query replicas: */
 
-bool bch2_replicas_marked(struct bch_fs *c,
-			  enum bch_data_type data_type,
-			  struct bch_devs_list devs,
-			  bool check_gc_replicas)
-{
-	struct bch_replicas_entry_padded search;
-
-	if (!devs.nr)
-		return true;
-
-	memset(&search, 0, sizeof(search));
-
-	devlist_to_replicas(devs, data_type, &search.e);
-
-	return replicas_has_entry(c, &search.e, check_gc_replicas);
-}
-
-bool bch2_bkey_replicas_marked(struct bch_fs *c,
-			       enum bkey_type type,
-			       struct bkey_s_c k,
-			       bool check_gc_replicas)
-{
-	struct bch_replicas_entry_padded search;
-
-	memset(&search, 0, sizeof(search));
-
-	if (type == BKEY_TYPE_EXTENTS) {
-		struct bch_devs_list cached = bch2_bkey_cached_devs(k);
-		unsigned i;
-
-		for (i = 0; i < cached.nr; i++)
-			if (!bch2_replicas_marked(c, BCH_DATA_CACHED,
-					bch2_dev_list_single(cached.devs[i]),
-					check_gc_replicas))
-				return false;
-	}
-
-	bkey_to_replicas(type, k, &search.e);
-
-	return search.e.nr_devs
-		? replicas_has_entry(c, &search.e, check_gc_replicas)
-		: true;
-}
-
-struct replicas_status __bch2_replicas_status(struct bch_fs *c,
-					      struct bch_devs_mask online_devs)
+bool bch2_have_enough_devs(struct bch_fs *c, struct bch_devs_mask devs,
+			   unsigned flags, bool print)
 {
-	struct bch_sb_field_members *mi;
-	struct bch_replicas_entry *e;
-	struct bch_replicas_cpu *r;
-	unsigned i, nr_online, nr_offline;
-	struct replicas_status ret;
-
-	memset(&ret, 0, sizeof(ret));
+	struct bch_replicas_entry_v1 *e;
+	bool ret = true;
 
-	for (i = 0; i < ARRAY_SIZE(ret.replicas); i++)
-		ret.replicas[i].redundancy = INT_MAX;
+	percpu_down_read(&c->mark_lock);
+	for_each_cpu_replicas_entry(&c->replicas, e) {
+		unsigned i, nr_online = 0, nr_failed = 0, dflags = 0;
+		bool metadata = e->data_type < BCH_DATA_user;
 
-	mi = bch2_sb_get_members(c->disk_sb.sb);
-	rcu_read_lock();
-	r = rcu_dereference(c->replicas);
-
-	for_each_cpu_replicas_entry(r, e) {
-		if (e->data_type >= ARRAY_SIZE(ret.replicas))
-			panic("e %p data_type %u\n", e, e->data_type);
-
-		nr_online = nr_offline = 0;
+		if (e->data_type == BCH_DATA_cached)
+			continue;
 
 		for (i = 0; i < e->nr_devs; i++) {
-			BUG_ON(!bch2_dev_exists(c->disk_sb.sb, mi,
-						e->devs[i]));
+			struct bch_dev *ca = bch_dev_bkey_exists(c, e->devs[i]);
 
-			if (test_bit(e->devs[i], online_devs.d))
-				nr_online++;
-			else
-				nr_offline++;
+			nr_online += test_bit(e->devs[i], devs.d);
+			nr_failed += ca->mi.state == BCH_MEMBER_STATE_failed;
 		}
 
-		ret.replicas[e->data_type].redundancy =
-			min(ret.replicas[e->data_type].redundancy,
-			    (int) nr_online - (int) e->nr_required);
+		if (nr_failed == e->nr_devs)
+			continue;
+
+		if (nr_online < e->nr_required)
+			dflags |= metadata
+				? BCH_FORCE_IF_METADATA_LOST
+				: BCH_FORCE_IF_DATA_LOST;
+
+		if (nr_online < e->nr_devs)
+			dflags |= metadata
+				? BCH_FORCE_IF_METADATA_DEGRADED
+				: BCH_FORCE_IF_DATA_DEGRADED;
+
+		if (dflags & ~flags) {
+			if (print) {
+				struct printbuf buf = PRINTBUF;
+
+				bch2_replicas_entry_to_text(&buf, e);
+				bch_err(c, "insufficient devices online (%u) for replicas entry %s",
+					nr_online, buf.buf);
+				printbuf_exit(&buf);
+			}
+			ret = false;
+			break;
+		}
 
-		ret.replicas[e->data_type].nr_offline =
-			max(ret.replicas[e->data_type].nr_offline,
-			    nr_offline);
 	}
-
-	rcu_read_unlock();
-
-	for (i = 0; i < ARRAY_SIZE(ret.replicas); i++)
-		if (ret.replicas[i].redundancy == INT_MAX)
-			ret.replicas[i].redundancy = 0;
+	percpu_up_read(&c->mark_lock);
 
 	return ret;
 }
 
-struct replicas_status bch2_replicas_status(struct bch_fs *c)
+unsigned bch2_sb_dev_has_data(struct bch_sb *sb, unsigned dev)
 {
-	return __bch2_replicas_status(c, bch2_online_devs(c));
-}
+	struct bch_sb_field_replicas *replicas;
+	struct bch_sb_field_replicas_v0 *replicas_v0;
+	unsigned i, data_has = 0;
+
+	replicas = bch2_sb_field_get(sb, replicas);
+	replicas_v0 = bch2_sb_field_get(sb, replicas_v0);
+
+	if (replicas) {
+		struct bch_replicas_entry_v1 *r;
+
+		for_each_replicas_entry(replicas, r)
+			for (i = 0; i < r->nr_devs; i++)
+				if (r->devs[i] == dev)
+					data_has |= 1 << r->data_type;
+	} else if (replicas_v0) {
+		struct bch_replicas_entry_v0 *r;
+
+		for_each_replicas_entry_v0(replicas_v0, r)
+			for (i = 0; i < r->nr_devs; i++)
+				if (r->devs[i] == dev)
+					data_has |= 1 << r->data_type;
+	}
 
-static bool have_enough_devs(struct replicas_status s,
-			     enum bch_data_type type,
-			     bool force_if_degraded,
-			     bool force_if_lost)
-{
-	return (!s.replicas[type].nr_offline || force_if_degraded) &&
-		(s.replicas[type].redundancy >= 0 || force_if_lost);
-}
 
-bool bch2_have_enough_devs(struct replicas_status s, unsigned flags)
-{
-	return (have_enough_devs(s, BCH_DATA_JOURNAL,
-				 flags & BCH_FORCE_IF_METADATA_DEGRADED,
-				 flags & BCH_FORCE_IF_METADATA_LOST) &&
-		have_enough_devs(s, BCH_DATA_BTREE,
-				 flags & BCH_FORCE_IF_METADATA_DEGRADED,
-				 flags & BCH_FORCE_IF_METADATA_LOST) &&
-		have_enough_devs(s, BCH_DATA_USER,
-				 flags & BCH_FORCE_IF_DATA_DEGRADED,
-				 flags & BCH_FORCE_IF_DATA_LOST));
+	return data_has;
 }
 
-int bch2_replicas_online(struct bch_fs *c, bool meta)
+unsigned bch2_dev_has_data(struct bch_fs *c, struct bch_dev *ca)
 {
-	struct replicas_status s = bch2_replicas_status(c);
+	unsigned ret;
+
+	mutex_lock(&c->sb_lock);
+	ret = bch2_sb_dev_has_data(c->disk_sb.sb, ca->dev_idx);
+	mutex_unlock(&c->sb_lock);
 
-	return (meta
-		? min(s.replicas[BCH_DATA_JOURNAL].redundancy,
-		      s.replicas[BCH_DATA_BTREE].redundancy)
-		: s.replicas[BCH_DATA_USER].redundancy) + 1;
+	return ret;
 }
 
-unsigned bch2_dev_has_data(struct bch_fs *c, struct bch_dev *ca)
+void bch2_fs_replicas_exit(struct bch_fs *c)
 {
-	struct bch_replicas_entry *e;
-	struct bch_replicas_cpu *r;
-	unsigned i, ret = 0;
+	unsigned i;
 
-	rcu_read_lock();
-	r = rcu_dereference(c->replicas);
+	kfree(c->usage_scratch);
+	for (i = 0; i < ARRAY_SIZE(c->usage); i++)
+		free_percpu(c->usage[i]);
+	kfree(c->usage_base);
+	kfree(c->replicas.entries);
+	kfree(c->replicas_gc.entries);
 
-	for_each_cpu_replicas_entry(r, e)
-		for (i = 0; i < e->nr_devs; i++)
-			if (e->devs[i] == ca->dev_idx)
-				ret |= 1 << e->data_type;
+	mempool_exit(&c->replicas_delta_pool);
+}
 
-	rcu_read_unlock();
+int bch2_fs_replicas_init(struct bch_fs *c)
+{
+	bch2_journal_entry_res_resize(&c->journal,
+			&c->replicas_journal_res,
+			reserve_journal_replicas(c, &c->replicas));
 
-	return ret;
+	return mempool_init_kmalloc_pool(&c->replicas_delta_pool, 1,
+					 REPLICAS_DELTA_LIST_MAX) ?:
+		replicas_table_update(c, &c->replicas);
 }