bcachefs: journal_replay_early()

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

fs/bcachefs/btree_gc.c

@@ -1249,19 +1249,3 @@ int bch2_gc_thread_start(struct bch_fs *c)
 	wake_up_process(p);
 	return 0;
 }
-
-/* Initial GC computes bucket marks during startup */
-
-int bch2_initial_gc(struct bch_fs *c, struct list_head *journal)
-{
-	int ret = bch2_gc(c, journal, true);
-
-	/*
-	 * Skip past versions that might have possibly been used (as nonces),
-	 * but hadn't had their pointers written:
-	 */
-	if (c->sb.encryption_type)
-		atomic64_add(1 << 16, &c->key_version);
-
-	return ret;
-}

fs/bcachefs/btree_gc.h

@@ -8,7 +8,6 @@ void bch2_coalesce(struct bch_fs *);
 int bch2_gc(struct bch_fs *, struct list_head *, bool);
 void bch2_gc_thread_stop(struct bch_fs *);
 int bch2_gc_thread_start(struct bch_fs *);
-int bch2_initial_gc(struct bch_fs *, struct list_head *);
 void bch2_mark_dev_superblock(struct bch_fs *, struct bch_dev *, unsigned);
 
 /*

fs/bcachefs/btree_types.h

@@ -475,6 +475,7 @@ struct btree_root {
 	__BKEY_PADDED(key, BKEY_BTREE_PTR_VAL_U64s_MAX);
 	u8			level;
 	u8			alive;
+	s8			error;
 };
 
 /*

fs/bcachefs/btree_update_interior.c

@@ -2122,7 +2122,6 @@ void bch2_btree_set_root_for_read(struct bch_fs *c, struct btree *b)
 	BUG_ON(btree_node_root(c, b));
 
 	__bch2_btree_set_root_inmem(c, b);
-	bch2_btree_set_root_ondisk(c, b, READ);
 }
 
 void bch2_btree_root_alloc(struct bch_fs *c, enum btree_id id)

fs/bcachefs/recovery.c

@@ -13,16 +13,17 @@
 #include "journal_io.h"
 #include "quota.h"
 #include "recovery.h"
+#include "replicas.h"
 #include "super-io.h"
 
 #include <linux/stat.h>
 
 #define QSTR(n) { { { .len = strlen(n) } }, .name = n }
 
-struct bkey_i *btree_root_find(struct bch_fs *c,
-			       struct bch_sb_field_clean *clean,
-			       struct jset *j,
-			       enum btree_id id, unsigned *level)
+static struct bkey_i *btree_root_find(struct bch_fs *c,
+				      struct bch_sb_field_clean *clean,
+				      struct jset *j,
+				      enum btree_id id, unsigned *level)
 {
 	struct bkey_i *k;
 	struct jset_entry *entry, *start, *end;
@@ -50,6 +51,51 @@ struct bkey_i *btree_root_find(struct bch_fs *c,
 	return k;
 }
 
+static int journal_replay_entry_early(struct bch_fs *c,
+				      struct jset_entry *entry)
+{
+	int ret = 0;
+
+	switch (entry->type) {
+	case BCH_JSET_ENTRY_btree_root: {
+		struct btree_root *r = &c->btree_roots[entry->btree_id];
+
+		if (entry->u64s) {
+			r->level = entry->level;
+			bkey_copy(&r->key, &entry->start[0]);
+			r->error = 0;
+		} else {
+			r->error = -EIO;
+		}
+		r->alive = true;
+		break;
+	}
+	case BCH_JSET_ENTRY_usage: {
+		struct jset_entry_usage *u =
+			container_of(entry, struct jset_entry_usage, entry);
+
+		switch (u->type) {
+		case FS_USAGE_REPLICAS:
+			ret = bch2_replicas_set_usage(c, &u->r,
+						le64_to_cpu(u->sectors));
+			break;
+		case FS_USAGE_INODES:
+			percpu_u64_set(&c->usage[0]->s.nr_inodes,
+						le64_to_cpu(u->sectors));
+			break;
+		case FS_USAGE_KEY_VERSION:
+			atomic64_set(&c->key_version,
+				     le64_to_cpu(u->sectors));
+			break;
+		}
+
+		break;
+	}
+	}
+
+	return ret;
+}
+
 static int verify_superblock_clean(struct bch_fs *c,
 				   struct bch_sb_field_clean *clean,
 				   struct jset *j)
@@ -126,6 +172,7 @@ int bch2_fs_recovery(struct bch_fs *c)
 {
 	const char *err = "cannot allocate memory";
 	struct bch_sb_field_clean *clean = NULL, *sb_clean = NULL;
+	struct jset_entry *entry;
 	LIST_HEAD(journal);
 	struct jset *j = NULL;
 	unsigned i;
@@ -178,28 +225,44 @@ int bch2_fs_recovery(struct bch_fs *c)
 	fsck_err_on(clean && !journal_empty(&journal), c,
 		    "filesystem marked clean but journal not empty");
 
+	err = "insufficient memory";
 	if (clean) {
 		c->bucket_clock[READ].hand = le16_to_cpu(clean->read_clock);
 		c->bucket_clock[WRITE].hand = le16_to_cpu(clean->write_clock);
+
+		for (entry = clean->start;
+		     entry != vstruct_end(&clean->field);
+		     entry = vstruct_next(entry)) {
+			ret = journal_replay_entry_early(c, entry);
+			if (ret)
+				goto err;
+		}
 	} else {
+		struct journal_replay *i;
+
 		c->bucket_clock[READ].hand = le16_to_cpu(j->read_clock);
 		c->bucket_clock[WRITE].hand = le16_to_cpu(j->write_clock);
+
+		list_for_each_entry(i, &journal, list)
+			vstruct_for_each(&i->j, entry) {
+				ret = journal_replay_entry_early(c, entry);
+				if (ret)
+					goto err;
+			}
 	}
 
 	for (i = 0; i < BTREE_ID_NR; i++) {
-		unsigned level;
-		struct bkey_i *k;
+		struct btree_root *r = &c->btree_roots[i];
 
-		k = btree_root_find(c, clean, j, i, &level);
-		if (!k)
+		if (!r->alive)
 			continue;
 
 		err = "invalid btree root pointer";
-		if (IS_ERR(k))
+		if (r->error)
 			goto err;
 
 		err = "error reading btree root";
-		if (bch2_btree_root_read(c, i, k, level)) {
+		if (bch2_btree_root_read(c, i, &r->key, r->level)) {
 			if (i != BTREE_ID_ALLOC)
 				goto err;
 
@@ -226,13 +289,20 @@ int bch2_fs_recovery(struct bch_fs *c)
 
 	bch_verbose(c, "starting mark and sweep:");
 	err = "error in recovery";
-	ret = bch2_initial_gc(c, &journal);
+	ret = bch2_gc(c, &journal, true);
 	if (ret)
 		goto err;
 	bch_verbose(c, "mark and sweep done");
 
 	clear_bit(BCH_FS_REBUILD_REPLICAS, &c->flags);
 
+	/*
+	 * Skip past versions that might have possibly been used (as nonces),
+	 * but hadn't had their pointers written:
+	 */
+	if (c->sb.encryption_type && !c->sb.clean)
+		atomic64_add(1 << 16, &c->key_version);
+
 	if (c->opts.noreplay)
 		goto out;
 
@@ -319,7 +389,7 @@ int bch2_fs_initialize(struct bch_fs *c)
 	for (i = 0; i < BTREE_ID_NR; i++)
 		bch2_btree_root_alloc(c, i);
 
-	ret = bch2_initial_gc(c, &journal);
+	ret = bch2_gc(c, &journal, true);
 	if (ret)
 		goto err;
 

fs/bcachefs/replicas.c

@@ -530,6 +530,34 @@ int bch2_replicas_gc_start(struct bch_fs *c, unsigned typemask)
 	return 0;
 }
 
+int bch2_replicas_set_usage(struct bch_fs *c,
+			    struct bch_replicas_entry *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->replicas, r);
+		if (!n.entries)
+			return -ENOMEM;
+
+		ret = replicas_table_update(c, &n);
+		if (ret)
+			return ret;
+
+		kfree(n.entries);
+
+		idx = bch2_replicas_entry_idx(c, r);
+		BUG_ON(ret < 0);
+	}
+
+	percpu_u64_set(&c->usage[0]->data[idx], sectors);
+
+	return 0;
+}
+
 /* Replicas tracking - superblock: */
 
 static int

fs/bcachefs/replicas.h

@@ -57,6 +57,10 @@ unsigned bch2_dev_has_data(struct bch_fs *, struct bch_dev *);
 int bch2_replicas_gc_end(struct bch_fs *, int);
 int bch2_replicas_gc_start(struct bch_fs *, unsigned);
 
+int bch2_replicas_set_usage(struct bch_fs *,
+			    struct bch_replicas_entry *,
+			    u64);
+
 #define for_each_cpu_replicas_entry(_r, _i)				\
 	for (_i = (_r)->entries;					\
 	     (void *) (_i) < (void *) (_r)->entries + (_r)->nr * (_r)->entry_size;\