bcachefs: Convert io paths for snapshots

This plumbs around the subvolume ID as was done previously for other
filesystem code, but now for the IO paths - the control flow in the IO
paths is trickier so the changes in this patch are more involved.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

fs/bcachefs/fs-io.c

@@ -769,23 +769,35 @@ static void readpage_bio_extend(struct readpages_iter *iter,
 	}
 }
 
-static void bchfs_read(struct btree_trans *trans, struct btree_iter *iter,
-		       struct bch_read_bio *rbio, u64 inum,
+static void bchfs_read(struct btree_trans *trans,
+		       struct bch_read_bio *rbio,
+		       subvol_inum inum,
 		       struct readpages_iter *readpages_iter)
 {
 	struct bch_fs *c = trans->c;
+	struct btree_iter iter;
 	struct bkey_buf sk;
 	int flags = BCH_READ_RETRY_IF_STALE|
 		BCH_READ_MAY_PROMOTE;
+	u32 snapshot;
 	int ret = 0;
 
 	rbio->c = c;
 	rbio->start_time = local_clock();
+	rbio->subvol = inum.subvol;
 
 	bch2_bkey_buf_init(&sk);
 retry:
 	bch2_trans_begin(trans);
+	iter = (struct btree_iter) { NULL };
 
+	ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+	if (ret)
+		goto err;
+
+	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
+			     SPOS(inum.inum, rbio->bio.bi_iter.bi_sector, snapshot),
+			     BTREE_ITER_SLOTS|BTREE_ITER_FILTER_SNAPSHOTS);
 	while (1) {
 		struct bkey_s_c k;
 		unsigned bytes, sectors, offset_into_extent;
@@ -800,15 +812,15 @@ static void bchfs_read(struct btree_trans *trans, struct btree_iter *iter,
 			break;
 		}
 
-		bch2_btree_iter_set_pos(iter,
-				POS(inum, rbio->bio.bi_iter.bi_sector));
+		bch2_btree_iter_set_pos(&iter,
+				POS(inum.inum, rbio->bio.bi_iter.bi_sector));
 
-		k = bch2_btree_iter_peek_slot(iter);
+		k = bch2_btree_iter_peek_slot(&iter);
 		ret = bkey_err(k);
 		if (ret)
 			break;
 
-		offset_into_extent = iter->pos.offset -
+		offset_into_extent = iter.pos.offset -
 			bkey_start_offset(k.k);
 		sectors = k.k->size - offset_into_extent;
 
@@ -838,7 +850,7 @@ static void bchfs_read(struct btree_trans *trans, struct btree_iter *iter,
 		if (bkey_extent_is_allocation(k.k))
 			bch2_add_page_sectors(&rbio->bio, k);
 
-		bch2_read_extent(trans, rbio, iter->pos,
+		bch2_read_extent(trans, rbio, iter.pos,
 				 data_btree, k, offset_into_extent, flags);
 
 		if (flags & BCH_READ_LAST_FRAGMENT)
@@ -847,12 +859,14 @@ static void bchfs_read(struct btree_trans *trans, struct btree_iter *iter,
 		swap(rbio->bio.bi_iter.bi_size, bytes);
 		bio_advance(&rbio->bio, bytes);
 	}
+err:
+	bch2_trans_iter_exit(trans, &iter);
 
 	if (ret == -EINTR)
 		goto retry;
 
 	if (ret) {
-		bch_err_inum_ratelimited(c, inum,
+		bch_err_inum_ratelimited(c, inum.inum,
 				"read error %i from btree lookup", ret);
 		rbio->bio.bi_status = BLK_STS_IOERR;
 		bio_endio(&rbio->bio);
@@ -867,7 +881,6 @@ void bch2_readahead(struct readahead_control *ractl)
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	struct bch_io_opts opts = io_opts(c, &inode->ei_inode);
 	struct btree_trans trans;
-	struct btree_iter iter;
 	struct page *page;
 	struct readpages_iter readpages_iter;
 	int ret;
@@ -876,8 +889,6 @@ void bch2_readahead(struct readahead_control *ractl)
 	BUG_ON(ret);
 
 	bch2_trans_init(&trans, c, 0, 0);
-	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents, POS_MIN,
-			     BTREE_ITER_SLOTS);
 
 	bch2_pagecache_add_get(&inode->ei_pagecache_lock);
 
@@ -898,22 +909,20 @@ void bch2_readahead(struct readahead_control *ractl)
 		rbio->bio.bi_end_io = bch2_readpages_end_io;
 		BUG_ON(!bio_add_page(&rbio->bio, page, PAGE_SIZE, 0));
 
-		bchfs_read(&trans, &iter, rbio, inode->v.i_ino,
+		bchfs_read(&trans, rbio, inode_inum(inode),
 			   &readpages_iter);
 	}
 
 	bch2_pagecache_add_put(&inode->ei_pagecache_lock);
 
-	bch2_trans_iter_exit(&trans, &iter);
 	bch2_trans_exit(&trans);
 	kfree(readpages_iter.pages);
 }
 
 static void __bchfs_readpage(struct bch_fs *c, struct bch_read_bio *rbio,
-			     u64 inum, struct page *page)
+			     subvol_inum inum, struct page *page)
 {
 	struct btree_trans trans;
-	struct btree_iter iter;
 
 	bch2_page_state_create(page, __GFP_NOFAIL);
 
@@ -923,12 +932,7 @@ static void __bchfs_readpage(struct bch_fs *c, struct bch_read_bio *rbio,
 	BUG_ON(!bio_add_page(&rbio->bio, page, PAGE_SIZE, 0));
 
 	bch2_trans_init(&trans, c, 0, 0);
-	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents, POS_MIN,
-			     BTREE_ITER_SLOTS);
-
-	bchfs_read(&trans, &iter, rbio, inum, NULL);
-
-	bch2_trans_iter_exit(&trans, &iter);
+	bchfs_read(&trans, rbio, inum, NULL);
 	bch2_trans_exit(&trans);
 }
 
@@ -951,7 +955,7 @@ static int bch2_read_single_page(struct page *page,
 	rbio->bio.bi_private = &done;
 	rbio->bio.bi_end_io = bch2_read_single_page_end_io;
 
-	__bchfs_readpage(c, rbio, inode->v.i_ino, page);
+	__bchfs_readpage(c, rbio, inode_inum(inode), page);
 	wait_for_completion(&done);
 
 	ret = blk_status_to_errno(rbio->bio.bi_status);
@@ -1096,6 +1100,7 @@ static void bch2_writepage_io_alloc(struct bch_fs *c,
 	op->nr_replicas		= nr_replicas;
 	op->res.nr_replicas	= nr_replicas;
 	op->write_point		= writepoint_hashed(inode->ei_last_dirtied);
+	op->subvol		= inode->ei_subvol;
 	op->pos			= POS(inode->v.i_ino, sector);
 	op->end_io		= bch2_writepage_io_done;
 	op->wbio.bio.bi_iter.bi_sector = sector;
@@ -1733,7 +1738,7 @@ static int bch2_direct_IO_read(struct kiocb *req, struct iov_iter *iter)
 		if (iter->count)
 			closure_get(&dio->cl);
 
-		bch2_read(c, rbio_init(bio, opts), inode->v.i_ino);
+		bch2_read(c, rbio_init(bio, opts), inode_inum(inode));
 	}
 
 	iter->count += shorten;
@@ -1816,7 +1821,8 @@ static bool bch2_check_range_allocated(struct bch_fs *c, subvol_inum inum,
 		if (bkey_cmp(bkey_start_pos(k.k), POS(inum.inum, end)) >= 0)
 			break;
 
-		if (nr_replicas > bch2_bkey_replicas(c, k) ||
+		if (k.k->p.snapshot != snapshot ||
+		    nr_replicas > bch2_bkey_replicas(c, k) ||
 		    (!compressed && bch2_bkey_sectors_compressed(k))) {
 			ret = false;
 			break;
@@ -1944,6 +1950,7 @@ static long bch2_dio_write_loop(struct dio_write *dio)
 		op_journal_seq_set(&dio->op, &inode->ei_journal_seq);
 		dio->op.write_point	= writepoint_hashed((unsigned long) current);
 		dio->op.nr_replicas	= dio->op.opts.data_replicas;
+		dio->op.subvol		= inode->ei_subvol;
 		dio->op.pos		= POS(inode->v.i_ino, (u64) req->ki_pos >> 9);
 
 		if ((req->ki_flags & IOCB_DSYNC) &&
@@ -2438,7 +2445,7 @@ int bch2_truncate(struct mnt_idmap *idmap,
 
 	truncate_setsize(&inode->v, iattr->ia_size);
 
-	ret = bch2_fpunch(c, inode->v.i_ino,
+	ret = bch2_fpunch(c, inode_inum(inode),
 			round_up(iattr->ia_size, block_bytes(c)) >> 9,
 			U64_MAX, &inode->ei_journal_seq, &i_sectors_delta);
 	i_sectors_acct(c, inode, NULL, i_sectors_delta);
@@ -2498,7 +2505,7 @@ static long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len
 	if (discard_start < discard_end) {
 		s64 i_sectors_delta = 0;
 
-		ret = bch2_fpunch(c, inode->v.i_ino,
+		ret = bch2_fpunch(c, inode_inum(inode),
 				  discard_start, discard_end,
 				  &inode->ei_journal_seq,
 				  &i_sectors_delta);
@@ -2577,7 +2584,7 @@ static long bchfs_fcollapse_finsert(struct bch_inode_info *inode,
 	} else {
 		s64 i_sectors_delta = 0;
 
-		ret = bch2_fpunch(c, inode->v.i_ino,
+		ret = bch2_fpunch(c, inode_inum(inode),
 				  offset >> 9, (offset + len) >> 9,
 				  &inode->ei_journal_seq,
 				  &i_sectors_delta);
@@ -2793,7 +2800,8 @@ static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
 			reservation.v.nr_replicas = disk_res.nr_replicas;
 		}
 
-		ret = bch2_extent_update(&trans, &iter, &reservation.k_i,
+		ret = bch2_extent_update(&trans, inode_inum(inode), &iter,
+					 &reservation.k_i,
 				&disk_res, &inode->ei_journal_seq,
 				0, &i_sectors_delta, true);
 		i_sectors_acct(c, inode, &quota_res, i_sectors_delta);

fs/bcachefs/io.c

@@ -27,6 +27,7 @@
 #include "keylist.h"
 #include "move.h"
 #include "rebalance.h"
+#include "subvolume.h"
 #include "super.h"
 #include "super-io.h"
 #include "trace.h"
@@ -230,7 +231,8 @@ int bch2_sum_sector_overwrites(struct btree_trans *trans,
 			: 0;
 
 		if (!*usage_increasing &&
-		    (new_replicas > bch2_bkey_replicas(c, old) ||
+		    (new->k.p.snapshot != old.k->p.snapshot ||
+		     new_replicas > bch2_bkey_replicas(c, old) ||
 		     (!new_compressed && bch2_bkey_sectors_compressed(old))))
 			*usage_increasing = true;
 
@@ -266,6 +268,7 @@ int bch2_sum_sector_overwrites(struct btree_trans *trans,
 }
 
 int bch2_extent_update(struct btree_trans *trans,
+		       subvol_inum inum,
 		       struct btree_iter *iter,
 		       struct bkey_i *k,
 		       struct disk_reservation *disk_res,
@@ -324,11 +327,8 @@ int bch2_extent_update(struct btree_trans *trans,
 		struct btree_iter inode_iter;
 		struct bch_inode_unpacked inode_u;
 
-		ret = bch2_inode_peek(trans, &inode_iter, &inode_u,
-				      (subvol_inum) {
-				      .subvol = BCACHEFS_ROOT_SUBVOL,
-				      .inum = k->k.p.inode,
-				      }, BTREE_ITER_INTENT);
+		ret = bch2_inode_peek(trans, &inode_iter, &inode_u, inum,
+				      BTREE_ITER_INTENT);
 		if (ret)
 			return ret;
 
@@ -384,22 +384,37 @@ int bch2_extent_update(struct btree_trans *trans,
 	return 0;
 }
 
+/*
+ * Returns -EINTR if we had to drop locks:
+ */
 int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
-		   struct bpos end, u64 *journal_seq,
-		   s64 *i_sectors_delta)
+		   subvol_inum inum, u64 end,
+		   u64 *journal_seq, s64 *i_sectors_delta)
 {
 	struct bch_fs *c	= trans->c;
 	unsigned max_sectors	= KEY_SIZE_MAX & (~0 << c->block_bits);
+	struct bpos end_pos = POS(inum.inum, end);
 	struct bkey_s_c k;
 	int ret = 0, ret2 = 0;
+	u32 snapshot;
 
-	while ((bch2_trans_begin(trans),
-		(k = bch2_btree_iter_peek(iter)).k) &&
-	       bkey_cmp(iter->pos, end) < 0) {
+	while (1) {
 		struct disk_reservation disk_res =
 			bch2_disk_reservation_init(c, 0);
 		struct bkey_i delete;
 
+		bch2_trans_begin(trans);
+
+		ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+		if (ret)
+			goto btree_err;
+
+		bch2_btree_iter_set_snapshot(iter, snapshot);
+
+		k = bch2_btree_iter_peek(iter);
+		if (bkey_cmp(iter->pos, end_pos) >= 0)
+			break;
+
 		ret = bkey_err(k);
 		if (ret)
 			goto btree_err;
@@ -409,9 +424,9 @@ int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
 
 		/* create the biggest key we can */
 		bch2_key_resize(&delete.k, max_sectors);
-		bch2_cut_back(end, &delete);
+		bch2_cut_back(end_pos, &delete);
 
-		ret = bch2_extent_update(trans, iter, &delete,
+		ret = bch2_extent_update(trans, inum, iter, &delete,
 				&disk_res, journal_seq,
 				0, i_sectors_delta, false);
 		bch2_disk_reservation_put(c, &disk_res);
@@ -424,36 +439,31 @@ int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
 			break;
 	}
 
-	if (bkey_cmp(iter->pos, end) > 0) {
-		bch2_btree_iter_set_pos(iter, end);
-		ret = bch2_btree_iter_traverse(iter);
-	}
+	if (bkey_cmp(iter->pos, end_pos) > 0)
+		bch2_btree_iter_set_pos(iter, end_pos);
 
 	return ret ?: ret2;
 }
 
-int bch2_fpunch(struct bch_fs *c, u64 inum, u64 start, u64 end,
+int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
 		u64 *journal_seq, s64 *i_sectors_delta)
 {
 	struct btree_trans trans;
 	struct btree_iter iter;
-	int ret = 0;
+	int ret;
 
 	bch2_trans_init(&trans, c, BTREE_ITER_MAX, 1024);
 	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
-				   POS(inum, start),
-				   BTREE_ITER_INTENT);
+			     POS(inum.inum, start),
+			     BTREE_ITER_INTENT);
 
-	ret = bch2_fpunch_at(&trans, &iter, POS(inum, end),
+	ret = bch2_fpunch_at(&trans, &iter, inum, end,
 			     journal_seq, i_sectors_delta);
 
 	bch2_trans_iter_exit(&trans, &iter);
 	bch2_trans_exit(&trans);
 
-	if (ret == -EINTR)
-		ret = 0;
-
-	return ret;
+	return ret == -EINTR ? 0 : ret;
 }
 
 static int bch2_write_index_default(struct bch_write_op *op)
@@ -464,40 +474,51 @@ static int bch2_write_index_default(struct bch_write_op *op)
 	struct bkey_i *k = bch2_keylist_front(keys);
 	struct btree_trans trans;
 	struct btree_iter iter;
+	subvol_inum inum = {
+		.subvol = op->subvol,
+		.inum	= k->k.p.inode,
+	};
 	int ret;
 
+	BUG_ON(!inum.subvol);
+
 	bch2_bkey_buf_init(&sk);
 	bch2_trans_init(&trans, c, BTREE_ITER_MAX, 1024);
 
-	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
-			     bkey_start_pos(&k->k),
-			     BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
-
 	do {
 		bch2_trans_begin(&trans);
 
 		k = bch2_keylist_front(keys);
+		bch2_bkey_buf_copy(&sk, c, k);
 
-		k->k.p.snapshot = iter.snapshot;
+		ret = bch2_subvolume_get_snapshot(&trans, inum.subvol,
+						  &sk.k->k.p.snapshot);
+		if (ret == -EINTR)
+			continue;
+		if (ret)
+			break;
 
-		bch2_bkey_buf_realloc(&sk, c, k->k.u64s);
-		bkey_copy(sk.k, k);
-		bch2_cut_front(iter.pos, sk.k);
+		bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
+				     bkey_start_pos(&sk.k->k),
+				     BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
 
-		ret = bch2_extent_update(&trans, &iter, sk.k,
+		ret = bch2_extent_update(&trans, inum, &iter, sk.k,
 					 &op->res, op_journal_seq(op),
 					 op->new_i_size, &op->i_sectors_delta,
 					 op->flags & BCH_WRITE_CHECK_ENOSPC);
+		bch2_trans_iter_exit(&trans, &iter);
+
 		if (ret == -EINTR)
 			continue;
 		if (ret)
 			break;
 
 		if (bkey_cmp(iter.pos, k->k.p) >= 0)
-			bch2_keylist_pop_front(keys);
+			bch2_keylist_pop_front(&op->insert_keys);
+		else
+			bch2_cut_front(iter.pos, k);
 	} while (!bch2_keylist_empty(keys));
 
-	bch2_trans_iter_exit(&trans, &iter);
 	bch2_trans_exit(&trans);
 	bch2_bkey_buf_exit(&sk, c);
 
@@ -1645,7 +1666,7 @@ static void bch2_rbio_done(struct bch_read_bio *rbio)
 }
 
 static void bch2_read_retry_nodecode(struct bch_fs *c, struct bch_read_bio *rbio,
-				     struct bvec_iter bvec_iter, u64 inode,
+				     struct bvec_iter bvec_iter,
 				     struct bch_io_failures *failed,
 				     unsigned flags)
 {
@@ -1709,7 +1730,10 @@ static void bch2_rbio_retry(struct work_struct *work)
 	struct bch_fs *c	= rbio->c;
 	struct bvec_iter iter	= rbio->bvec_iter;
 	unsigned flags		= rbio->flags;
-	u64 inode		= rbio->read_pos.inode;
+	subvol_inum inum = {
+		.subvol = rbio->subvol,
+		.inum	= rbio->read_pos.inode,
+	};
 	struct bch_io_failures failed = { .nr = 0 };
 
 	trace_read_retry(&rbio->bio);
@@ -1725,12 +1749,12 @@ static void bch2_rbio_retry(struct work_struct *work)
 	flags &= ~BCH_READ_MAY_PROMOTE;
 
 	if (flags & BCH_READ_NODECODE) {
-		bch2_read_retry_nodecode(c, rbio, iter, inode, &failed, flags);
+		bch2_read_retry_nodecode(c, rbio, iter, &failed, flags);
 	} else {
 		flags &= ~BCH_READ_LAST_FRAGMENT;
 		flags |= BCH_READ_MUST_CLONE;
 
-		__bch2_read(c, rbio, iter, inode, &failed, flags);
+		__bch2_read(c, rbio, iter, inum, &failed, flags);
 	}
 }
 
@@ -2174,6 +2198,7 @@ int __bch2_read_extent(struct btree_trans *trans, struct bch_read_bio *orig,
 	/* XXX: only initialize this if needed */
 	rbio->devs_have		= bch2_bkey_devs(k);
 	rbio->pick		= pick;
+	rbio->subvol		= orig->subvol;
 	rbio->read_pos		= read_pos;
 	rbio->data_btree	= data_btree;
 	rbio->data_pos		= data_pos;
@@ -2281,25 +2306,31 @@ int __bch2_read_extent(struct btree_trans *trans, struct bch_read_bio *orig,
 }
 
 void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
-		 struct bvec_iter bvec_iter, u64 inode,
+		 struct bvec_iter bvec_iter, subvol_inum inum,
 		 struct bch_io_failures *failed, unsigned flags)
 {
 	struct btree_trans trans;
 	struct btree_iter iter;
 	struct bkey_buf sk;
 	struct bkey_s_c k;
+	u32 snapshot;
 	int ret;
 
 	BUG_ON(flags & BCH_READ_NODECODE);
 
 	bch2_bkey_buf_init(&sk);
 	bch2_trans_init(&trans, c, 0, 0);
-	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
-			     POS(inode, bvec_iter.bi_sector),
-			     BTREE_ITER_SLOTS);
 retry:
 	bch2_trans_begin(&trans);
+	iter = (struct btree_iter) { NULL };
+
+	ret = bch2_subvolume_get_snapshot(&trans, inum.subvol, &snapshot);
+	if (ret)
+		goto err;
 
+	bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
+			     SPOS(inum.inum, bvec_iter.bi_sector, snapshot),
+			     BTREE_ITER_SLOTS|BTREE_ITER_FILTER_SNAPSHOTS);
 	while (1) {
 		unsigned bytes, sectors, offset_into_extent;
 		enum btree_id data_btree = BTREE_ID_extents;
@@ -2314,7 +2345,7 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 		}
 
 		bch2_btree_iter_set_pos(&iter,
-				POS(inode, bvec_iter.bi_sector));
+				POS(inum.inum, bvec_iter.bi_sector));
 
 		k = bch2_btree_iter_peek_slot(&iter);
 		ret = bkey_err(k);
@@ -2364,16 +2395,17 @@ void __bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 		swap(bvec_iter.bi_size, bytes);
 		bio_advance_iter(&rbio->bio, &bvec_iter, bytes);
 	}
+err:
+	bch2_trans_iter_exit(&trans, &iter);
 
 	if (ret == -EINTR || ret == READ_RETRY || ret == READ_RETRY_AVOID)
 		goto retry;
 
-	bch2_trans_iter_exit(&trans, &iter);
 	bch2_trans_exit(&trans);
 	bch2_bkey_buf_exit(&sk, c);
 
 	if (ret) {
-		bch_err_inum_ratelimited(c, inode,
+		bch_err_inum_ratelimited(c, inum.inum,
 					 "read error %i from btree lookup", ret);
 		rbio->bio.bi_status = BLK_STS_IOERR;
 		bch2_rbio_done(rbio);

fs/bcachefs/io.h

@@ -83,12 +83,13 @@ static inline struct workqueue_struct *index_update_wq(struct bch_write_op *op)
 
 int bch2_sum_sector_overwrites(struct btree_trans *, struct btree_iter *,
 			       struct bkey_i *, bool *, bool *, s64 *, s64 *);
-int bch2_extent_update(struct btree_trans *, struct btree_iter *,
-		       struct bkey_i *, struct disk_reservation *,
-		       u64 *, u64, s64 *, bool);
+int bch2_extent_update(struct btree_trans *, subvol_inum,
+		       struct btree_iter *, struct bkey_i *,
+		       struct disk_reservation *, u64 *, u64, s64 *, bool);
+
 int bch2_fpunch_at(struct btree_trans *, struct btree_iter *,
-		   struct bpos, u64 *, s64 *);
-int bch2_fpunch(struct bch_fs *c, u64, u64, u64, u64 *, s64 *);
+		   subvol_inum, u64, u64 *, s64 *);
+int bch2_fpunch(struct bch_fs *c, subvol_inum, u64, u64, u64 *, s64 *);
 
 static inline void bch2_write_op_init(struct bch_write_op *op, struct bch_fs *c,
 				      struct bch_io_opts opts)
@@ -108,6 +109,7 @@ static inline void bch2_write_op_init(struct bch_write_op *op, struct bch_fs *c,
 	op->devs_have.nr	= 0;
 	op->target		= 0;
 	op->opts		= opts;
+	op->subvol		= 0;
 	op->pos			= POS_MAX;
 	op->version		= ZERO_VERSION;
 	op->write_point		= (struct write_point_specifier) { 0 };
@@ -174,10 +176,10 @@ static inline void bch2_read_extent(struct btree_trans *trans,
 }
 
 void __bch2_read(struct bch_fs *, struct bch_read_bio *, struct bvec_iter,
-		 u64, struct bch_io_failures *, unsigned flags);
+		 subvol_inum, struct bch_io_failures *, unsigned flags);
 
 static inline void bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
-			     u64 inode)
+			     subvol_inum inum)
 {
 	struct bch_io_failures failed = { .nr = 0 };
 
@@ -185,8 +187,9 @@ static inline void bch2_read(struct bch_fs *c, struct bch_read_bio *rbio,
 
 	rbio->c = c;
 	rbio->start_time = local_clock();
+	rbio->subvol = inum.subvol;
 
-	__bch2_read(c, rbio, rbio->bio.bi_iter, inode, &failed,
+	__bch2_read(c, rbio, rbio->bio.bi_iter, inum, &failed,
 		    BCH_READ_RETRY_IF_STALE|
 		    BCH_READ_MAY_PROMOTE|
 		    BCH_READ_USER_MAPPED);

fs/bcachefs/io_types.h

@@ -62,6 +62,7 @@ struct bch_read_bio {
 	/*
 	 * pos we read from - different from data_pos for indirect extents:
 	 */
+	u32			subvol;
 	struct bpos		read_pos;
 
 	/*
@@ -124,6 +125,7 @@ struct bch_write_op {
 	u16			nonce;
 	struct bch_io_opts	opts;
 
+	u32			subvol;
 	struct bpos		pos;
 	struct bversion		version;
 

fs/bcachefs/reflink.c

@@ -212,6 +212,7 @@ s64 bch2_remap_range(struct bch_fs *c,
 	struct bpos dst_end = dst_start, src_end = src_start;
 	struct bpos src_want;
 	u64 dst_done;
+	u32 dst_snapshot, src_snapshot;
 	int ret = 0, ret2 = 0;
 
 	if (!percpu_ref_tryget(&c->writes))
@@ -243,15 +244,19 @@ s64 bch2_remap_range(struct bch_fs *c,
 		}
 
 		ret = bch2_subvolume_get_snapshot(&trans, src_inum.subvol,
-						  &src_iter.snapshot);
+						  &src_snapshot);
 		if (ret)
 			continue;
 
+		bch2_btree_iter_set_snapshot(&src_iter, src_snapshot);
+
 		ret = bch2_subvolume_get_snapshot(&trans, dst_inum.subvol,
-						  &dst_iter.snapshot);
+						  &dst_snapshot);
 		if (ret)
 			continue;
 
+		bch2_btree_iter_set_snapshot(&dst_iter, dst_snapshot);
+
 		dst_done = dst_iter.pos.offset - dst_start.offset;
 		src_want = POS(src_start.inode, src_start.offset + dst_done);
 		bch2_btree_iter_set_pos(&src_iter, src_want);
@@ -262,11 +267,11 @@ s64 bch2_remap_range(struct bch_fs *c,
 			continue;
 
 		if (bkey_cmp(src_want, src_iter.pos) < 0) {
-			ret = bch2_fpunch_at(&trans, &dst_iter,
-					bpos_min(dst_end,
-						 POS(dst_iter.pos.inode, dst_iter.pos.offset +
-						     src_iter.pos.offset - src_want.offset)),
-						 journal_seq, i_sectors_delta);
+			ret = bch2_fpunch_at(&trans, &dst_iter, dst_inum,
+					min(dst_end.offset,
+					    dst_iter.pos.offset +
+					    src_iter.pos.offset - src_want.offset),
+					journal_seq, i_sectors_delta);
 			continue;
 		}
 
@@ -303,8 +308,9 @@ s64 bch2_remap_range(struct bch_fs *c,
 		bch2_key_resize(&new_dst.k->k,
 				min(src_k.k->p.offset - src_want.offset,
 				    dst_end.offset - dst_iter.pos.offset));
-		ret = bch2_extent_update(&trans, &dst_iter, new_dst.k,
-					 &disk_res, journal_seq,
+
+		ret = bch2_extent_update(&trans, dst_inum, &dst_iter,
+					 new_dst.k, &disk_res, journal_seq,
 					 new_i_size, i_sectors_delta,
 					 true);
 		bch2_disk_reservation_put(c, &disk_res);