bcachefs: bch2_accounting_mem_gc()

Add a new helper to free zeroed out accounting entries, and use it in
bch2_replicas_gc2(); bch2_replicas_gc2() was killing superblock replicas
entries if their corresponding accounting counters were nonzero, but
that's incorrect - the superblock replicas entry needs to exist if the
accounting entry exists, not if it's nonzero, because we check and
create the replicas entry when creating the new accounting entry - we
don't know when it's becoming nonzero.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

3 files changed, 42 insertions, 6 deletions

diff --git a/fs/bcachefs/disk_accounting.c b/fs/bcachefs/disk_accounting.c
index 3327d465908d..5b1546d1a23d 100644
--- a/fs/bcachefs/disk_accounting.c
+++ b/fs/bcachefs/disk_accounting.c
@@ -273,6 +273,38 @@ int bch2_accounting_mem_insert(struct bch_fs *c, struct bkey_s_c_accounting a, b
 	return ret;
 }
 
+static bool accounting_mem_entry_is_zero(struct accounting_mem_entry *e)
+{
+	for (unsigned i = 0; i < e->nr_counters; i++)
+		if (percpu_u64_get(e->v[0] + i) ||
+		    (e->v[1] &&
+		     percpu_u64_get(e->v[1] + i)))
+			return false;
+	return true;
+}
+
+void bch2_accounting_mem_gc(struct bch_fs *c)
+{
+	struct bch_accounting_mem *acc = &c->accounting;
+
+	percpu_down_write(&c->mark_lock);
+	struct accounting_mem_entry *dst = acc->k.data;
+
+	darray_for_each(acc->k, src) {
+		if (accounting_mem_entry_is_zero(src)) {
+			free_percpu(src->v[0]);
+			free_percpu(src->v[1]);
+		} else {
+			*dst++ = *src;
+		}
+	}
+
+	acc->k.nr = dst - acc->k.data;
+	eytzinger0_sort(acc->k.data, acc->k.nr, sizeof(acc->k.data[0]),
+			accounting_pos_cmp, NULL);
+	percpu_up_write(&c->mark_lock);
+}
+
 /*
  * Read out accounting keys for replicas entries, as an array of
  * bch_replicas_usage entries.
diff --git a/fs/bcachefs/disk_accounting.h b/fs/bcachefs/disk_accounting.h
index 81dab01d1eb8..3d3f25e08b69 100644
--- a/fs/bcachefs/disk_accounting.h
+++ b/fs/bcachefs/disk_accounting.h
@@ -105,6 +105,7 @@ static inline int accounting_pos_cmp(const void *_l, const void *_r)
 }
 
 int bch2_accounting_mem_insert(struct bch_fs *, struct bkey_s_c_accounting, bool);
+void bch2_accounting_mem_gc(struct bch_fs *);
 
 static inline int __bch2_accounting_mem_mod(struct bch_fs *c, struct bkey_s_c_accounting a, bool gc)
 {
diff --git a/fs/bcachefs/replicas.c b/fs/bcachefs/replicas.c
index 06f6d48f74c0..10c96cb2047a 100644
--- a/fs/bcachefs/replicas.c
+++ b/fs/bcachefs/replicas.c
@@ -420,10 +420,10 @@ int bch2_replicas_gc_start(struct bch_fs *c, unsigned typemask)
 int bch2_replicas_gc2(struct bch_fs *c)
 {
 	struct bch_replicas_cpu new = { 0 };
-	unsigned i, nr;
+	unsigned nr;
 	int ret = 0;
 
-	bch2_journal_meta(&c->journal);
+	bch2_accounting_mem_gc(c);
 retry:
 	nr		= READ_ONCE(c->replicas.nr);
 	new.entry_size	= READ_ONCE(c->replicas.entry_size);
@@ -444,7 +444,7 @@ retry:
 		goto retry;
 	}
 
-	for (i = 0; i < c->replicas.nr; i++) {
+	for (unsigned i = 0; i < c->replicas.nr; i++) {
 		struct bch_replicas_entry_v1 *e =
 			cpu_replicas_entry(&c->replicas, i);
 
@@ -454,10 +454,13 @@ retry:
 
 		memcpy(&k.replicas, e, replicas_entry_bytes(e));
 
-		u64 v = 0;
-		bch2_accounting_mem_read(c, disk_accounting_pos_to_bpos(&k), &v, 1);
+		struct bpos p = disk_accounting_pos_to_bpos(&k);
 
-		if (e->data_type == BCH_DATA_journal || v)
+		struct bch_accounting_mem *acc = &c->accounting;
+		bool kill = eytzinger0_find(acc->k.data, acc->k.nr, sizeof(acc->k.data[0]),
+					    accounting_pos_cmp, &p) >= acc->k.nr;
+
+		if (e->data_type == BCH_DATA_journal || !kill)
 			memcpy(cpu_replicas_entry(&new, new.nr++),
 			       e, new.entry_size);
 	}