bcachefs: Snapshot creationbcachefs-v5.10-snapshots

7 files changed, 259 insertions, 37 deletions

diff --git a/fs/bcachefs/chardev.c b/fs/bcachefs/chardev.c
index c61601476c0d..0370640efb9f 100644
--- a/fs/bcachefs/chardev.c
+++ b/fs/bcachefs/chardev.c
@@ -5,6 +5,9 @@
 #include "bcachefs_ioctl.h"
 #include "buckets.h"
 #include "chardev.h"
+#include "dirent.h"
+#include "fs.h"
+#include "fs-common.h"
 #include "journal.h"
 #include "move.h"
 #include "replicas.h"
@@ -16,12 +19,16 @@
 #include <linux/device.h>
 #include <linux/file.h>
 #include <linux/fs.h>
+#include <linux/fsnotify.h>
 #include <linux/ioctl.h>
 #include <linux/kthread.h>
 #include <linux/major.h>
+#include <linux/namei.h>
 #include <linux/sched/task.h>
+#include <linux/security.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
+#include <linux/writeback.h>
 
 /* returns with ref on ca->ref */
 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
@@ -585,6 +592,136 @@ static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
 	return ret;
 }
 
+static long bch2_ioctl_subvolume_create(struct bch_fs *c,
+				struct bch_ioctl_subvolume arg)
+{
+	struct inode *dir;
+	struct bch_inode_info *inode;
+	struct user_namespace *s_user_ns;
+	struct dentry *dst_dentry;
+	struct path src_path, dst_path;
+	int how = LOOKUP_FOLLOW;
+	int error;
+	subvol_inum snapshot_src = { 0 };
+	unsigned lookup_flags = 0;
+	unsigned create_flags = BCH_CREATE_SUBVOL;
+
+	if (arg.flags & ~(BCH_SUBVOL_SNAPSHOT_CREATE|
+			  BCH_SUBVOL_SNAPSHOT_RO))
+		return -EINVAL;
+
+	if (!(arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE) &&
+	    (arg.src_ptr ||
+	     (arg.flags & BCH_SUBVOL_SNAPSHOT_RO)))
+		return -EINVAL;
+
+	if (arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE)
+		create_flags |= BCH_CREATE_SNAPSHOT;
+
+	if (arg.flags & BCH_SUBVOL_SNAPSHOT_RO)
+		create_flags |= BCH_CREATE_SNAPSHOT_RO;
+
+	down_read(&c->vfs_sb->s_umount);
+
+	if (arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE)
+		sync_inodes_sb(c->vfs_sb);
+retry:
+	if (arg.src_ptr) {
+		error = user_path_at(arg.dirfd,
+				(const char __user *)(unsigned long)arg.src_ptr,
+				how, &src_path);
+		if (error)
+			goto err1;
+
+		if (src_path.dentry->d_sb->s_fs_info != c) {
+			path_put(&src_path);
+			error = -EXDEV;
+			goto err1;
+		}
+
+		snapshot_src = inode_inum(to_bch_ei(src_path.dentry->d_inode));
+	}
+
+	dst_dentry = user_path_create(arg.dirfd,
+			(const char __user *)(unsigned long)arg.dst_ptr,
+			&dst_path, lookup_flags);
+	error = PTR_ERR_OR_ZERO(dst_dentry);
+	if (error)
+		goto err2;
+
+	if (dst_dentry->d_sb->s_fs_info != c) {
+		error = -EXDEV;
+		goto err3;
+	}
+
+	if (dst_dentry->d_inode) {
+		error = -EEXIST;
+		goto err3;
+	}
+
+	dir = dst_path.dentry->d_inode;
+	if (IS_DEADDIR(dir)) {
+		error = -ENOENT;
+		goto err3;
+	}
+
+	s_user_ns = dir->i_sb->s_user_ns;
+	if (!kuid_has_mapping(s_user_ns, current_fsuid()) ||
+	    !kgid_has_mapping(s_user_ns, current_fsgid())) {
+		error = -EOVERFLOW;
+		goto err3;
+	}
+
+	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
+	if (error)
+		goto err3;
+
+	if (!IS_POSIXACL(dir))
+		arg.mode &= ~current_umask();
+
+	error = security_path_mkdir(&dst_path, dst_dentry, arg.mode);
+	if (error)
+		goto err3;
+
+	if ((arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE) &&
+	    !arg.src_ptr)
+		snapshot_src.subvol = to_bch_ei(dir)->ei_inode.bi_subvol;
+
+	inode = __bch2_create(to_bch_ei(dir), dst_dentry, arg.mode|S_IFDIR,
+			      0, snapshot_src, create_flags);
+	error = PTR_ERR_OR_ZERO(inode);
+	if (error)
+		goto err3;
+
+	d_instantiate(dst_dentry, &inode->v);
+	fsnotify_mkdir(dir, dst_dentry);
+err3:
+	done_path_create(&dst_path, dst_dentry);
+err2:
+	if (arg.src_ptr)
+		path_put(&src_path);
+
+	if (retry_estale(error, lookup_flags)) {
+		lookup_flags |= LOOKUP_REVAL;
+		goto retry;
+	}
+err1:
+	up_read(&c->vfs_sb->s_umount);
+
+	return error;
+}
+
+static long bch2_ioctl_subvolume_destroy(struct bch_fs *c,
+				struct bch_ioctl_subvolume arg)
+{
+	int ret = 0;
+
+	if (arg.flags)
+		return -EINVAL;
+
+	return ret;
+}
+
 #define BCH_IOCTL(_name, _argtype)					\
 do {									\
 	_argtype i;							\
@@ -644,6 +781,11 @@ long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
 	case BCH_IOCTL_DISK_RESIZE_JOURNAL:
 		BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
 
+	case BCH_IOCTL_SUBVOLUME_CREATE:
+		BCH_IOCTL(subvolume_create, struct bch_ioctl_subvolume);
+	case BCH_IOCTL_SUBVOLUME_DESTROY:
+		BCH_IOCTL(subvolume_destroy, struct bch_ioctl_subvolume);
+
 	default:
 		return -ENOTTY;
 	}
diff --git a/fs/bcachefs/fs-common.c b/fs/bcachefs/fs-common.c
index 310ec36c4df4..9d605df70c14 100644
--- a/fs/bcachefs/fs-common.c
+++ b/fs/bcachefs/fs-common.c
@@ -19,6 +19,7 @@ int bch2_create_trans(struct btree_trans *trans,
 		      uid_t uid, gid_t gid, umode_t mode, dev_t rdev,
 		      struct posix_acl *default_acl,
 		      struct posix_acl *acl,
+		      subvol_inum snapshot_src,
 		      unsigned flags)
 {
 	struct bch_fs *c = trans->c;
@@ -26,10 +27,9 @@ int bch2_create_trans(struct btree_trans *trans,
 	struct btree_iter *inode_iter = NULL;
 	subvol_inum new_inum = dir;
 	u64 now = bch2_current_time(c);
-	u64 dir_offset = 0;
 	u64 dir_target;
 	u32 snapshot;
-	unsigned dir_type;
+	unsigned dir_type = mode_to_type(mode);
 	int ret;
 
 	ret = bch2_subvolume_get_snapshot(trans, dir.subvol, &snapshot);
@@ -41,36 +41,116 @@ int bch2_create_trans(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	bch2_inode_init_late(new_inode, now, uid, gid, mode, rdev, dir_u);
+	if (!(flags & BCH_CREATE_SNAPSHOT)) {
+		/* Normal create path - allocate a new inode: */
+		bch2_inode_init_late(new_inode, now, uid, gid, mode, rdev, dir_u);
 
-	if (!name)
-		new_inode->bi_flags |= BCH_INODE_UNLINKED;
+		if (flags & BCH_CREATE_TMPFILE)
+			new_inode->bi_flags |= BCH_INODE_UNLINKED;
 
-	inode_iter = bch2_inode_create(trans, new_inode, snapshot);
-	ret = PTR_ERR_OR_ZERO(inode_iter);
-	if (ret)
-		goto err;
+		inode_iter = bch2_inode_create(trans, new_inode, snapshot);
+		ret = PTR_ERR_OR_ZERO(inode_iter);
+		if (ret)
+			goto err;
+
+		snapshot_src = (subvol_inum) { 0 };
+	} else {
+		/*
+		 * Creating a snapshot - we're not allocating a new inode, but
+		 * we do have to lookup the root inode of the subvolume we're
+		 * snapshotting and update it (in the new snapshot):
+		 */
+
+		if (!snapshot_src.inum) {
+			/* Inode wasn't specified, just snapshot: */
+			struct btree_iter *subvol_iter =
+				bch2_trans_get_iter(trans, BTREE_ID_subvolumes,
+						    POS(0, snapshot_src.subvol), 0);
+			struct bkey_s_c k = bch2_btree_iter_peek_slot(subvol_iter);
+
+			ret = bkey_err(k);
+			if (!ret && k.k->type != KEY_TYPE_subvolume) {
+				bch_err(c, "subvolume %u not found",
+					snapshot_src.subvol);
+				ret = -ENOENT;
+			}
+
+			if (!ret)
+				snapshot_src.inum = le64_to_cpu(bkey_s_c_to_subvolume(k).v->inode);
+			bch2_trans_iter_put(trans, subvol_iter);
+
+			if (ret)
+				goto err;
+		}
+
+		inode_iter = bch2_inode_peek(trans, new_inode, snapshot_src,
+					     BTREE_ITER_INTENT);
+		ret = PTR_ERR_OR_ZERO(inode_iter);
+		if (ret)
+			goto err;
+
+		if (new_inode->bi_subvol != snapshot_src.subvol) {
+			/* Not a subvolume root: */
+			ret = -EINVAL;
+			goto err;
+		}
+
+		/*
+		 * If we're not root, we have to own the subvolume being
+		 * snapshotted:
+		 */
+		if (uid && new_inode->bi_uid != uid) {
+			ret = -EPERM;
+			goto err;
+		}
+
+		flags |= BCH_CREATE_SUBVOL;
+	}
 
 	new_inum.inum	= new_inode->bi_inum;
 	dir_target	= new_inode->bi_inum;
-	dir_type	= mode_to_type(new_inode->bi_mode);
 
-	if (default_acl) {
-		ret = bch2_set_acl_trans(trans, new_inum, new_inode,
-					 default_acl, ACL_TYPE_DEFAULT);
+	if (flags & BCH_CREATE_SUBVOL) {
+		u32 new_subvol, dir_snapshot;
+
+		ret = bch2_subvolume_create(trans, new_inode->bi_inum,
+					    snapshot_src.subvol,
+					    &new_subvol, &snapshot,
+					    (flags & BCH_CREATE_SNAPSHOT_RO) != 0);
 		if (ret)
 			goto err;
-	}
 
-	if (acl) {
-		ret = bch2_set_acl_trans(trans, new_inum, new_inode,
-					 acl, ACL_TYPE_ACCESS);
+		new_inode->bi_parent_subvol	= dir.subvol;
+		new_inode->bi_subvol		= new_subvol;
+		new_inum.subvol			= new_subvol;
+		dir_target			= new_subvol;
+		dir_type			= DT_SUBVOL;
+
+		ret = bch2_subvolume_get_snapshot(trans, dir.subvol, &dir_snapshot);
 		if (ret)
 			goto err;
+		bch2_btree_iter_set_snapshot(dir_iter, dir_snapshot);
 	}
 
-	if (name) {
+	if (!(flags & BCH_CREATE_SNAPSHOT)) {
+		if (default_acl) {
+			ret = bch2_set_acl_trans(trans, new_inum, new_inode,
+						 default_acl, ACL_TYPE_DEFAULT);
+			if (ret)
+				goto err;
+		}
+
+		if (acl) {
+			ret = bch2_set_acl_trans(trans, new_inum, new_inode,
+						 acl, ACL_TYPE_ACCESS);
+			if (ret)
+				goto err;
+		}
+	}
+
+	if (!(flags & BCH_CREATE_TMPFILE)) {
 		struct bch_hash_info dir_hash = bch2_hash_info_init(c, dir_u);
+		u64 dir_offset;
 
 		if (S_ISDIR(new_inode->bi_mode))
 			dir_u->bi_nlink++;
@@ -88,11 +168,11 @@ int bch2_create_trans(struct btree_trans *trans,
 					 BCH_HASH_SET_MUST_CREATE);
 		if (ret)
 			goto err;
-	}
 
-	if (c->sb.version >= bcachefs_metadata_version_inode_backpointers) {
-		new_inode->bi_dir		= dir_u->bi_inum;
-		new_inode->bi_dir_offset	= dir_offset;
+		if (c->sb.version >= bcachefs_metadata_version_inode_backpointers) {
+			new_inode->bi_dir		= dir_u->bi_inum;
+			new_inode->bi_dir_offset	= dir_offset;
+		}
 	}
 
 	inode_iter->flags &= ~BTREE_ITER_ALL_SNAPSHOTS;
diff --git a/fs/bcachefs/fs-common.h b/fs/bcachefs/fs-common.h
index 1bb2ac4dc13a..e25ef56147ac 100644
--- a/fs/bcachefs/fs-common.h
+++ b/fs/bcachefs/fs-common.h
@@ -5,6 +5,9 @@
 struct posix_acl;
 
 #define BCH_CREATE_TMPFILE		(1U << 0)
+#define BCH_CREATE_SUBVOL		(1U << 1)
+#define BCH_CREATE_SNAPSHOT		(1U << 2)
+#define BCH_CREATE_SNAPSHOT_RO		(1U << 3)
 
 int bch2_create_trans(struct btree_trans *, subvol_inum,
 		      struct bch_inode_unpacked *,
@@ -13,7 +16,7 @@ int bch2_create_trans(struct btree_trans *, subvol_inum,
 		      uid_t, gid_t, umode_t, dev_t,
 		      struct posix_acl *,
 		      struct posix_acl *,
-		      unsigned);
+		      subvol_inum, unsigned);
 
 int bch2_link_trans(struct btree_trans *,
 		    subvol_inum, struct bch_inode_unpacked *,
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index d944f2292663..73f93f1ac35f 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -239,12 +239,6 @@ struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
 	struct bch_inode_info *inode;
 	int ret;
 
-	/*
-	 * debug assert, to be removed when we start creating
-	 * subvolumes/snapshots:
-	 */
-	BUG_ON(inum.subvol != BCACHEFS_ROOT_SUBVOL);
-
 	inode = to_bch_ei(iget5_locked(c->vfs_sb,
 				       bch2_inode_hash(inum),
 				       bch2_iget5_test,
@@ -272,7 +266,8 @@ struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
 
 struct bch_inode_info *
 __bch2_create(struct bch_inode_info *dir, struct dentry *dentry,
-	      umode_t mode, dev_t rdev, unsigned flags)
+	      umode_t mode, dev_t rdev, subvol_inum snapshot_src,
+	      unsigned flags)
 {
 	struct bch_fs *c = dir->v.i_sb->s_fs_info;
 	struct user_namespace *ns = dir->v.i_sb->s_user_ns;
@@ -318,7 +313,7 @@ retry:
 				  from_kuid(ns, current_fsuid()),
 				  from_kgid(ns, current_fsgid()),
 				  mode, rdev,
-				  default_acl, acl, flags) ?:
+				  default_acl, acl, snapshot_src, flags) ?:
 		bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
 				KEY_TYPE_QUOTA_PREALLOC);
 	if (unlikely(ret))
@@ -425,7 +420,8 @@ static int bch2_mknod(struct inode *vdir, struct dentry *dentry,
 		      umode_t mode, dev_t rdev)
 {
 	struct bch_inode_info *inode =
-		__bch2_create(to_bch_ei(vdir), dentry, mode, rdev, 0);
+		__bch2_create(to_bch_ei(vdir), dentry, mode, rdev,
+			      (subvol_inum) { 0 }, 0);
 
 	if (IS_ERR(inode))
 		return PTR_ERR(inode);
@@ -535,7 +531,7 @@ static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
 	int ret;
 
 	inode = __bch2_create(dir, dentry, S_IFLNK|S_IRWXUGO, 0,
-			      BCH_CREATE_TMPFILE);
+			      (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
 	if (unlikely(IS_ERR(inode)))
 		return PTR_ERR(inode);
 
@@ -846,7 +842,7 @@ static int bch2_tmpfile(struct inode *vdir, struct dentry *dentry, umode_t mode)
 {
 	struct bch_inode_info *inode =
 		__bch2_create(to_bch_ei(vdir), dentry, mode, 0,
-			      BCH_CREATE_TMPFILE);
+			      (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
 
 	if (IS_ERR(inode))
 		return PTR_ERR(inode);
diff --git a/fs/bcachefs/fs.h b/fs/bcachefs/fs.h
index 066fd1c68c3c..e65513c819f0 100644
--- a/fs/bcachefs/fs.h
+++ b/fs/bcachefs/fs.h
@@ -147,7 +147,7 @@ struct bch_inode_unpacked;
 
 struct bch_inode_info *
 __bch2_create(struct bch_inode_info *, struct dentry *,
-	      umode_t, dev_t, unsigned);
+	      umode_t, dev_t, subvol_inum, unsigned);
 
 int bch2_fs_quota_transfer(struct bch_fs *,
 			   struct bch_inode_info *,
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index bfc27233d9c5..e286ae0b9a39 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -289,7 +289,8 @@ create_lostfound:
 				      BTREE_INSERT_LAZY_RW,
 			bch2_create_trans(trans, root_inum, &root,
 					  lostfound, &lostfound_str,
-					  0, 0, S_IFDIR|0700, 0, NULL, NULL, 0));
+					  0, 0, S_IFDIR|0700, 0, NULL, NULL,
+					  (subvol_inum) { }, 0));
 		if (ret)
 			bch_err(c, "error creating lost+found: %i", ret);
 	}
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 072e45481c37..176ab38699dd 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -1407,7 +1407,7 @@ int bch2_fs_initialize(struct bch_fs *c)
 				  &root_inode, &lostfound_inode,
 				  &lostfound,
 				  0, 0, S_IFDIR|0700, 0,
-				  NULL, NULL, 0));
+				  NULL, NULL, (subvol_inum) { 0 }, 0));
 	if (ret) {
 		bch_err(c, "error creating lost+found");
 		goto err;