bcachefs: KEY_TYPE_alloc_v2

This introduces a new version of KEY_TYPE_alloc, which uses the new
varint encoding introduced for inodes. This means we'll eventually be
able to support much larger bucket sizes (for SMR devices), and the
read/write time fields are expanded to 64 bits - which will be used in
the next patch to get rid of the periodic rescaling of those fields.

Also, for buckets that are members of erasure coded stripes, this adds
persistent fields for the index of the stripe they're members of and the
stripe redundancy. This is part of work to get rid of having to scan and
read into memory the alloc and stripes btrees at mount time.

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

1 files changed, 25 insertions, 10 deletions

diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index 78bea3e5fa9a..a32d399e5b6f 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -105,6 +105,9 @@ const char *bch2_stripe_invalid(const struct bch_fs *c, struct bkey_s_c k)
 {
 	const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
 
+	if (!bkey_cmp(k.k->p, POS_MIN))
+		return "stripe at pos 0";
+
 	if (k.k->p.inode)
 		return "invalid stripe key";
 
@@ -279,10 +282,14 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
 			struct bch_csum got = ec_block_checksum(buf, i, offset);
 
 			if (bch2_crc_cmp(want, got)) {
+				char buf2[200];
+
+				bch2_bkey_val_to_text(&PBUF(buf2), c, bkey_i_to_s_c(&buf->key.k_i));
+
 				bch_err_ratelimited(c,
-					"stripe checksum error at %u:%u: csum type %u, expected %llx got %llx",
-					i, j, v->csum_type,
-					want.lo, got.lo);
+					"stripe checksum error for %ps at %u:%u: csum type %u, expected %llx got %llx\n%s",
+					(void *) _RET_IP_, i, j, v->csum_type,
+					want.lo, got.lo, buf2);
 				clear_bit(i, buf->valid);
 				break;
 			}
@@ -335,6 +342,8 @@ static int ec_do_recov(struct bch_fs *c, struct ec_stripe_buf *buf)
 static void ec_block_endio(struct bio *bio)
 {
 	struct ec_bio *ec_bio = container_of(bio, struct ec_bio, bio);
+	struct bch_stripe *v = &ec_bio->buf->key.v;
+	struct bch_extent_ptr *ptr = &v->ptrs[ec_bio->idx];
 	struct bch_dev *ca = ec_bio->ca;
 	struct closure *cl = bio->bi_private;
 
@@ -343,6 +352,13 @@ static void ec_block_endio(struct bio *bio)
 			       bch2_blk_status_to_str(bio->bi_status)))
 		clear_bit(ec_bio->idx, ec_bio->buf->valid);
 
+	if (ptr_stale(ca, ptr)) {
+		bch_err_ratelimited(ca->fs,
+				    "error %s stripe: stale pointer after io",
+				    bio_data_dir(bio) == READ ? "reading from" : "writing to");
+		clear_bit(ec_bio->idx, ec_bio->buf->valid);
+	}
+
 	bio_put(&ec_bio->bio);
 	percpu_ref_put(&ca->io_ref);
 	closure_put(cl);
@@ -652,7 +668,6 @@ void bch2_stripes_heap_update(struct bch_fs *c,
 
 static int ec_stripe_delete(struct bch_fs *c, size_t idx)
 {
-	//pr_info("deleting stripe %zu", idx);
 	return bch2_btree_delete_range(c, BTREE_ID_EC,
 				       POS(0, idx),
 				       POS(0, idx + 1),
@@ -795,6 +810,7 @@ static void extent_stripe_ptr_add(struct bkey_s_extent e,
 	*dst = (struct bch_extent_stripe_ptr) {
 		.type = 1 << BCH_EXTENT_ENTRY_stripe_ptr,
 		.block		= block,
+		.redundancy	= s->key.v.nr_redundant,
 		.idx		= s->key.k.p.offset,
 	};
 }
@@ -1054,8 +1070,6 @@ void bch2_ec_add_backpointer(struct bch_fs *c, struct write_point *wp,
 	if (!ob)
 		return;
 
-	//pr_info("adding backpointer at %llu:%llu", pos.inode, pos.offset);
-
 	ec = ob->ec;
 	mutex_lock(&ec->lock);
 
@@ -1348,12 +1362,14 @@ static s64 get_existing_stripe(struct bch_fs *c,
 	struct stripe *m;
 	size_t heap_idx;
 	u64 stripe_idx;
+	s64 ret = -1;
 
 	if (may_create_new_stripe(c))
 		return -1;
 
 	spin_lock(&c->ec_stripes_heap_lock);
 	for (heap_idx = 0; heap_idx < h->used; heap_idx++) {
+		/* No blocks worth reusing, stripe will just be deleted: */
 		if (!h->data[heap_idx].blocks_nonempty)
 			continue;
 
@@ -1365,13 +1381,12 @@ static s64 get_existing_stripe(struct bch_fs *c,
 		    m->sectors		== head->blocksize &&
 		    m->blocks_nonempty	< m->nr_blocks - m->nr_redundant) {
 			bch2_stripes_heap_del(c, m, stripe_idx);
-			spin_unlock(&c->ec_stripes_heap_lock);
-			return stripe_idx;
+			ret = stripe_idx;
+			break;
 		}
 	}
-
 	spin_unlock(&c->ec_stripes_heap_lock);
-	return -1;
+	return ret;
 }
 
 struct ec_stripe_head *bch2_ec_stripe_head_get(struct bch_fs *c,