bcachefs: Update export_operations for snapshots

When support for snapshots was merged, export operations weren't
updated yet. This patch adds new filehandle types for bcachefs that
include the subvolume ID and updates export operations for subvolumes -
and also .get_parent, support for which was added just prior to
snapshots.

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

4 files changed, 218 insertions, 25 deletions

diff --git a/fs/bcachefs/dirent.c b/fs/bcachefs/dirent.c
index 5db1426faaf3..4dfcc955675b 100644
--- a/fs/bcachefs/dirent.c
+++ b/fs/bcachefs/dirent.c
@@ -197,8 +197,8 @@ static void dirent_copy_target(struct bkey_i_dirent *dst,
 	dst->v.d_type = src.v->d_type;
 }
 
-static int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir,
-				   struct bkey_s_c_dirent d, subvol_inum *target)
+int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir,
+			    struct bkey_s_c_dirent d, subvol_inum *target)
 {
 	struct bch_subvolume s;
 	int ret = 0;
diff --git a/fs/bcachefs/dirent.h b/fs/bcachefs/dirent.h
index 8ae407765fe4..1bb4d802bc1d 100644
--- a/fs/bcachefs/dirent.h
+++ b/fs/bcachefs/dirent.h
@@ -29,6 +29,9 @@ static inline unsigned dirent_val_u64s(unsigned len)
 			    sizeof(u64));
 }
 
+int bch2_dirent_read_target(struct btree_trans *, subvol_inum,
+			    struct bkey_s_c_dirent, subvol_inum *);
+
 int bch2_dirent_create(struct btree_trans *, subvol_inum,
 		       const struct bch_hash_info *, u8,
 		       const struct qstr *, u64, u64 *, int);
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 4561c60c95e3..61027d349cd8 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -1124,46 +1124,230 @@ static const struct address_space_operations bch_address_space_operations = {
 	.error_remove_page = generic_error_remove_page,
 };
 
-#if 0
-static struct inode *bch2_nfs_get_inode(struct super_block *sb,
-		u64 ino, u32 generation)
+struct bcachefs_fid {
+	u64		inum;
+	u32		subvol;
+	u32		gen;
+} __packed;
+
+struct bcachefs_fid_with_parent {
+	struct bcachefs_fid	fid;
+	struct bcachefs_fid	dir;
+} __packed;
+
+static int bcachefs_fid_valid(int fh_len, int fh_type)
 {
-	struct bch_fs *c = sb->s_fs_info;
-	struct inode *vinode;
+	switch (fh_type) {
+	case FILEID_BCACHEFS_WITHOUT_PARENT:
+		return fh_len == sizeof(struct bcachefs_fid) / sizeof(u32);
+	case FILEID_BCACHEFS_WITH_PARENT:
+		return fh_len == sizeof(struct bcachefs_fid_with_parent) / sizeof(u32);
+	default:
+		return false;
+	}
+}
+
+static struct bcachefs_fid bch2_inode_to_fid(struct bch_inode_info *inode)
+{
+	return (struct bcachefs_fid) {
+		.inum	= inode->ei_inode.bi_inum,
+		.subvol	= inode->ei_subvol,
+		.gen	= inode->ei_inode.bi_generation,
+	};
+}
+
+static int bch2_encode_fh(struct inode *vinode, u32 *fh, int *len,
+			  struct inode *vdir)
+{
+	struct bch_inode_info *inode	= to_bch_ei(vinode);
+	struct bch_inode_info *dir	= to_bch_ei(vdir);
+
+	if (*len < sizeof(struct bcachefs_fid_with_parent) / sizeof(u32))
+		return FILEID_INVALID;
+
+	if (!S_ISDIR(inode->v.i_mode) && dir) {
+		struct bcachefs_fid_with_parent *fid = (void *) fh;
+
+		fid->fid = bch2_inode_to_fid(inode);
+		fid->dir = bch2_inode_to_fid(dir);
+
+		*len = sizeof(*fid) / sizeof(u32);
+		return FILEID_BCACHEFS_WITH_PARENT;
+	} else {
+		struct bcachefs_fid *fid = (void *) fh;
 
-	if (ino < BCACHEFS_ROOT_INO)
-		return ERR_PTR(-ESTALE);
+		*fid = bch2_inode_to_fid(inode);
 
-	vinode = bch2_vfs_inode_get(c, ino);
-	if (IS_ERR(vinode))
-		return ERR_CAST(vinode);
-	if (generation && vinode->i_generation != generation) {
-		/* we didn't find the right inode.. */
+		*len = sizeof(*fid) / sizeof(u32);
+		return FILEID_BCACHEFS_WITHOUT_PARENT;
+	}
+}
+
+static struct inode *bch2_nfs_get_inode(struct super_block *sb,
+					struct bcachefs_fid fid)
+{
+	struct bch_fs *c = sb->s_fs_info;
+	struct inode *vinode = bch2_vfs_inode_get(c, (subvol_inum) {
+				    .subvol = fid.subvol,
+				    .inum = fid.inum,
+	});
+	if (!IS_ERR(vinode) && vinode->i_generation != fid.gen) {
 		iput(vinode);
-		return ERR_PTR(-ESTALE);
+		vinode = ERR_PTR(-ESTALE);
 	}
 	return vinode;
 }
 
-static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *fid,
+static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *_fid,
 		int fh_len, int fh_type)
 {
-	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
-				    bch2_nfs_get_inode);
+	struct bcachefs_fid *fid = (void *) _fid;
+
+	if (!bcachefs_fid_valid(fh_len, fh_type))
+		return NULL;
+
+	return d_obtain_alias(bch2_nfs_get_inode(sb, *fid));
 }
 
-static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *fid,
+static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *_fid,
 		int fh_len, int fh_type)
 {
-	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
-				    bch2_nfs_get_inode);
+	struct bcachefs_fid_with_parent *fid = (void *) _fid;
+
+	if (!bcachefs_fid_valid(fh_len, fh_type) ||
+	    fh_type != FILEID_BCACHEFS_WITH_PARENT)
+		return NULL;
+
+	return d_obtain_alias(bch2_nfs_get_inode(sb, fid->dir));
+}
+
+static struct dentry *bch2_get_parent(struct dentry *child)
+{
+	struct bch_inode_info *inode = to_bch_ei(child->d_inode);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+	subvol_inum parent_inum = {
+		.subvol = inode->ei_inode.bi_parent_subvol ?:
+			inode->ei_subvol,
+		.inum = inode->ei_inode.bi_dir,
+	};
+
+	if (!parent_inum.inum)
+		return NULL;
+
+	return d_obtain_alias(bch2_vfs_inode_get(c, parent_inum));
+}
+
+static int bch2_get_name(struct dentry *parent, char *name, struct dentry *child)
+{
+	struct bch_inode_info *inode	= to_bch_ei(child->d_inode);
+	struct bch_inode_info *dir	= to_bch_ei(parent->d_inode);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+	struct btree_trans trans;
+	struct btree_iter iter1;
+	struct btree_iter iter2;
+	struct bkey_s_c k;
+	struct bkey_s_c_dirent d;
+	struct bch_inode_unpacked inode_u;
+	subvol_inum target;
+	u32 snapshot;
+	unsigned name_len;
+	int ret;
+
+	if (!S_ISDIR(dir->v.i_mode))
+		return -EINVAL;
+
+	bch2_trans_init(&trans, c, 0, 0);
+
+	bch2_trans_iter_init(&trans, &iter1, BTREE_ID_dirents,
+			     POS(dir->ei_inode.bi_inum, 0), 0);
+	bch2_trans_iter_init(&trans, &iter2, BTREE_ID_dirents,
+			     POS(dir->ei_inode.bi_inum, 0), 0);
+retry:
+	bch2_trans_begin(&trans);
+
+	ret = bch2_subvolume_get_snapshot(&trans, dir->ei_subvol, &snapshot);
+	if (ret)
+		goto err;
+
+	bch2_btree_iter_set_snapshot(&iter1, snapshot);
+	bch2_btree_iter_set_snapshot(&iter2, snapshot);
+
+	ret = bch2_inode_find_by_inum_trans(&trans, inode_inum(inode), &inode_u);
+	if (ret)
+		goto err;
+
+	if (inode_u.bi_dir == dir->ei_inode.bi_inum) {
+		bch2_btree_iter_set_pos(&iter1, POS(inode_u.bi_dir, inode_u.bi_dir_offset));
+
+		k = bch2_btree_iter_peek_slot(&iter1);
+		ret = bkey_err(k);
+		if (ret)
+			goto err;
+
+		if (k.k->type != KEY_TYPE_dirent) {
+			ret = -ENOENT;
+			goto err;
+		}
+
+		d = bkey_s_c_to_dirent(k);
+		ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
+		if (ret > 0)
+			ret = -ENOENT;
+		if (ret)
+			goto err;
+
+		if (target.subvol	== inode->ei_subvol &&
+		    target.inum		== inode->ei_inode.bi_inum)
+			goto found;
+	} else {
+		/*
+		 * File with multiple hardlinks and our backref is to the wrong
+		 * directory - linear search:
+		 */
+		for_each_btree_key_continue_norestart(iter2, 0, k, ret) {
+			if (k.k->p.inode > dir->ei_inode.bi_inum)
+				break;
+
+			if (k.k->type != KEY_TYPE_dirent)
+				continue;
+
+			d = bkey_s_c_to_dirent(k);
+			ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
+			if (ret < 0)
+				break;
+			if (ret)
+				continue;
+
+			if (target.subvol	== inode->ei_subvol &&
+			    target.inum		== inode->ei_inode.bi_inum)
+				goto found;
+		}
+	}
+
+	ret = -ENOENT;
+	goto err;
+found:
+	name_len = min_t(unsigned, bch2_dirent_name_bytes(d), NAME_MAX);
+
+	memcpy(name, d.v->d_name, name_len);
+	name[name_len] = '\0';
+err:
+	if (ret == -EINTR)
+		goto retry;
+
+	bch2_trans_iter_exit(&trans, &iter1);
+	bch2_trans_iter_exit(&trans, &iter2);
+	bch2_trans_exit(&trans);
+
+	return ret;
 }
-#endif
 
 static const struct export_operations bch_export_ops = {
-	//.fh_to_dentry	= bch2_fh_to_dentry,
-	//.fh_to_parent	= bch2_fh_to_parent,
-	//.get_parent	= bch2_get_parent,
+	.encode_fh	= bch2_encode_fh,
+	.fh_to_dentry	= bch2_fh_to_dentry,
+	.fh_to_parent	= bch2_fh_to_parent,
+	.get_parent	= bch2_get_parent,
+	.get_name	= bch2_get_name,
 };
 
 static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
diff --git a/include/linux/exportfs.h b/include/linux/exportfs.h
index 11fbd0ee1370..f49a7d31167e 100644
--- a/include/linux/exportfs.h
+++ b/include/linux/exportfs.h
@@ -99,6 +99,12 @@ enum fid_type {
 	FILEID_FAT_WITH_PARENT = 0x72,
 
 	/*
+	 * 64 bit inode number, 32 bit subvolume, 32 bit generation number:
+	 */
+	FILEID_BCACHEFS_WITHOUT_PARENT = 0x80,
+	FILEID_BCACHEFS_WITH_PARENT = 0x81,
+
+	/*
 	 * 128 bit child FID (struct lu_fid)
 	 * 128 bit parent FID (struct lu_fid)
 	 */