[Btrfs-devel] cloning file data
Sage Weil
sage at newdream.net
Thu Apr 24 15:47:08 PDT 2008
Hi-
I'm working on a clone ioctl that will quickly and efficiently duplicate
the contents of a file, e.g.
int main(int argc, const char **argv)
{
int in = open(argv[1], O_RDONLY);
int out = open(argv[2], O_CREAT|O_TRUNC|O_WRONLY, 0644);
ioctl(out, BTRFS_IOC_CLONE, in);
close(in);
close(out);
return 0;
}
I've probably got the locking order a bit wrong, lots of error handling is
missing, and I suspect there's a cleaner way to do the target inode size
update, but it behaves well enough in my (limited :) testing.
Oh, and I wasn't certain the 'offset' in file_extent_item could be safely
ignored when duplicating the extent reference. My assumption was that it
is orthogonal to extent allocation and isn't related to the backref.
However, btrfs_insert_file_extent() always set offset=0. I'm guessing I
need to add an argument there and fix up the other callers?
Anyway, any comments or suggestions (on the interface or implemantation)
are welcome.. :)
sage
diff -r 1791a620d509 inode.c
--- a/inode.c Thu Apr 24 13:43:27 2008 -0700
+++ b/inode.c Thu Apr 24 15:10:17 2008 -0700
@@ -18,6 +18,7 @@
#include <linux/bio.h>
#include <linux/buffer_head.h>
+#include <linux/file.h>
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
@@ -2726,6 +2727,158 @@ long btrfs_ioctl_trans_end(struct file *
return 0;
}
+void dup_item_to_inode(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct extent_buffer *leaf,
+ int slot,
+ struct btrfs_key *key,
+ u64 destino)
+{
+ struct btrfs_path *cpath = btrfs_alloc_path();
+ int len = btrfs_item_size_nr(leaf, slot);
+ int dstoff;
+ struct btrfs_key ckey = *key;
+ int ret;
+
+ ckey.objectid = destino;
+ ret = btrfs_insert_empty_item(trans, root, cpath, &ckey, len);
+ dstoff = btrfs_item_ptr_offset(cpath->nodes[0], cpath->slots[0]);
+ copy_extent_buffer(cpath->nodes[0], leaf, dstoff,
+ btrfs_item_ptr_offset(leaf, slot),
+ len);
+ btrfs_release_path(root, cpath);
+}
+
+long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
+{
+ struct inode *inode = fdentry(file)->d_inode;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct file *src_file;
+ struct inode *src;
+ struct btrfs_trans_handle *trans;
+ int ret;
+ u64 pos;
+ struct btrfs_path *path;
+ struct btrfs_key key;
+ struct extent_buffer *leaf;
+ u32 nritems;
+ int nextret;
+ int slot;
+
+ src_file = fget(src_fd);
+ if (!src_file)
+ return -EBADF;
+ src = src_file->f_dentry->d_inode;
+
+ ret = -EXDEV;
+ if (src->i_sb != inode->i_sb)
+ goto out_fput;
+
+ if (inode < src) {
+ mutex_lock(&inode->i_mutex);
+ mutex_lock(&src->i_mutex);
+ } else {
+ mutex_lock(&src->i_mutex);
+ mutex_lock(&inode->i_mutex);
+ }
+
+ ret = -ENOTEMPTY;
+ if (inode->i_size)
+ goto out_unlock;
+
+ /* do any pending delalloc/csum calc on src, one way or another */
+ filemap_write_and_wait(src->i_mapping);
+
+ mutex_lock(&root->fs_info->fs_mutex);
+ trans = btrfs_start_transaction(root, 0);
+ path = btrfs_alloc_path();
+ pos = 0;
+ while (1) {
+ ret = btrfs_lookup_file_extent(trans, root, path, src->i_ino,
+ pos, 0);
+ if (ret < 0)
+ goto out;
+ if (ret > 0) {
+ if (path->slots[0] == 0) {
+ ret = 0;
+ goto out;
+ }
+ path->slots[0]--;
+ }
+ next_slot:
+ leaf = path->nodes[0];
+ slot = path->slots[0];
+ btrfs_item_key_to_cpu(leaf, &key, slot);
+
+ printk("key(%llu %x %llu)\n",
+ key.objectid, key.type, key.offset);
+ if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY ||
+ key.objectid != src->i_ino)
+ goto out;
+ if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
+ struct btrfs_file_extent_item *extent;
+ int found_type;
+ u64 len;
+ pos = key.offset;
+ extent = btrfs_item_ptr(leaf, slot,
+ struct btrfs_file_extent_item);
+ found_type = btrfs_file_extent_type(leaf, extent);
+ len = btrfs_file_extent_num_bytes(leaf, extent);
+ if (found_type == BTRFS_FILE_EXTENT_REG) {
+ u64 ds = btrfs_file_extent_disk_bytenr(leaf,
+ extent);
+ u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
+ extent);
+ printk(" %llu~%llu disk %llu~%llu off %llu\n",
+ pos, len, ds, dl,
+ btrfs_file_extent_offset(leaf, extent));
+ btrfs_insert_file_extent(trans, root,
+ inode->i_ino, pos,
+ ds, dl, len);
+ btrfs_inc_extent_ref(trans, root, ds, dl,
+ root->root_key.objectid,
+ trans->transid,
+ inode->i_ino, pos);
+ } else if (found_type == BTRFS_FILE_EXTENT_INLINE)
+ dup_item_to_inode(trans, root, path, leaf, slot,
+ &key, inode->i_ino);
+ pos = key.offset + len;
+ } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY)
+ dup_item_to_inode(trans, root, path, leaf, slot, &key,
+ inode->i_ino);
+
+ nritems = btrfs_header_nritems(leaf);
+ if (slot >= nritems - 1) {
+ nextret = btrfs_next_leaf(root, path);
+ if (nextret)
+ goto out;
+ } else {
+ path->slots[0]++;
+ }
+ goto next_slot;
+ }
+
+out:
+ ret = 0;
+ mutex_unlock(&root->fs_info->fs_mutex);
+
+ i_size_write(inode, src->i_size);
+ inode->i_blocks = src->i_blocks;
+ mark_inode_dirty(inode);
+
+ mutex_lock(&root->fs_info->fs_mutex);
+ btrfs_end_transaction(trans, root);
+ mutex_unlock(&root->fs_info->fs_mutex);
+
+out_unlock:
+ mutex_unlock(&src->i_mutex);
+ mutex_unlock(&inode->i_mutex);
+out_fput:
+ fput(src_file);
+ return ret;
+}
+
long btrfs_ioctl(struct file *file, unsigned int
cmd, unsigned long arg)
{
@@ -2744,6 +2897,9 @@ long btrfs_ioctl(struct file *file, unsi
return btrfs_ioctl_trans_end(file);
case BTRFS_IOC_SYNC:
btrfs_sync_fs(file->f_dentry->d_sb, 1);
+ return 0;
+ case BTRFS_IOC_CLONE:
+ btrfs_ioctl_clone(file, arg);
return 0;
}
diff -r 1791a620d509 ioctl.h
--- a/ioctl.h Thu Apr 24 13:43:27 2008 -0700
+++ b/ioctl.h Thu Apr 24 15:10:17 2008 -0700
@@ -36,5 +36,6 @@ struct btrfs_ioctl_vol_args {
#define BTRFS_IOC_TRANS_START _IO(BTRFS_IOCTL_MAGIC, 6)
#define BTRFS_IOC_TRANS_END _IO(BTRFS_IOCTL_MAGIC, 7)
#define BTRFS_IOC_SYNC _IO(BTRFS_IOCTL_MAGIC, 8)
+#define BTRFS_IOC_CLONE _IOW(BTRFS_IOCTL_MAGIC, 9, int)
#endif
More information about the Btrfs-devel
mailing list