btrfs-progs/utils.c

823 lines
23 KiB
C
Raw Normal View History

/*
* Copyright (C) 2007 Oracle. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License v2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
#define _XOPEN_SOURCE 600
#define __USE_XOPEN2K
#include <stdio.h>
#include <stdlib.h>
2008-03-24 19:04:49 +00:00
#ifndef __CHECKER__
#include <sys/ioctl.h>
#include <sys/mount.h>
#endif
#include <sys/types.h>
#include <sys/stat.h>
#include <uuid/uuid.h>
#include <dirent.h>
#include <fcntl.h>
#include <unistd.h>
#include <mntent.h>
#include "kerncompat.h"
#include "radix-tree.h"
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "crc32c.h"
#include "utils.h"
2008-03-24 19:04:49 +00:00
#include "volumes.h"
#include "ioctl.h"
2008-03-24 19:04:49 +00:00
#ifdef __CHECKER__
#define BLKGETSIZE64 0
static inline int ioctl(int fd, int define, u64 *size) { return 0; }
#endif
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
static u64 reference_root_table[] = {
[1] = BTRFS_ROOT_TREE_OBJECTID,
[2] = BTRFS_EXTENT_TREE_OBJECTID,
[3] = BTRFS_CHUNK_TREE_OBJECTID,
[4] = BTRFS_DEV_TREE_OBJECTID,
[5] = BTRFS_FS_TREE_OBJECTID,
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
[6] = BTRFS_CSUM_TREE_OBJECTID,
};
int make_btrfs(int fd, const char *device, const char *label,
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
u64 blocks[7], u64 num_bytes, u32 nodesize,
u32 leafsize, u32 sectorsize, u32 stripesize)
{
struct btrfs_super_block super;
struct extent_buffer *buf;
struct btrfs_root_item root_item;
struct btrfs_disk_key disk_key;
struct btrfs_extent_item *extent_item;
struct btrfs_inode_item *inode_item;
struct btrfs_chunk *chunk;
struct btrfs_dev_item *dev_item;
struct btrfs_dev_extent *dev_extent;
u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
u8 *ptr;
int i;
int ret;
u32 itemoff;
u32 nritems = 0;
u64 first_free;
u64 ref_root;
u32 array_size;
u32 item_size;
first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
first_free &= ~((u64)sectorsize - 1);
memset(&super, 0, sizeof(super));
num_bytes = (num_bytes / sectorsize) * sectorsize;
uuid_generate(super.fsid);
uuid_generate(super.dev_item.uuid);
uuid_generate(chunk_tree_uuid);
btrfs_set_super_bytenr(&super, blocks[0]);
2008-03-24 19:04:49 +00:00
btrfs_set_super_num_devices(&super, 1);
strncpy((char *)&super.magic, BTRFS_MAGIC, sizeof(super.magic));
btrfs_set_super_generation(&super, 1);
btrfs_set_super_root(&super, blocks[1]);
btrfs_set_super_chunk_root(&super, blocks[3]);
btrfs_set_super_total_bytes(&super, num_bytes);
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
btrfs_set_super_bytes_used(&super, 6 * leafsize);
btrfs_set_super_sectorsize(&super, sectorsize);
btrfs_set_super_leafsize(&super, leafsize);
btrfs_set_super_nodesize(&super, nodesize);
btrfs_set_super_stripesize(&super, stripesize);
btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
btrfs_set_super_chunk_root_generation(&super, 1);
if (label)
strcpy(super.label, label);
buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
/* create the tree of root objects */
memset(buf->data, 0, leafsize);
buf->len = leafsize;
btrfs_set_header_bytenr(buf, blocks[1]);
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
btrfs_set_header_nritems(buf, 4);
btrfs_set_header_generation(buf, 1);
btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
write_extent_buffer(buf, super.fsid, (unsigned long)
btrfs_header_fsid(buf), BTRFS_FSID_SIZE);
write_extent_buffer(buf, chunk_tree_uuid, (unsigned long)
btrfs_header_chunk_tree_uuid(buf),
BTRFS_UUID_SIZE);
/* create the items for the root tree */
memset(&root_item, 0, sizeof(root_item));
inode_item = &root_item.inode;
btrfs_set_stack_inode_generation(inode_item, 1);
btrfs_set_stack_inode_size(inode_item, 3);
btrfs_set_stack_inode_nlink(inode_item, 1);
btrfs_set_stack_inode_nbytes(inode_item, leafsize);
btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
btrfs_set_root_refs(&root_item, 1);
btrfs_set_root_used(&root_item, leafsize);
btrfs_set_root_generation(&root_item, 1);
memset(&disk_key, 0, sizeof(disk_key));
btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
btrfs_set_disk_key_offset(&disk_key, 0);
nritems = 0;
itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
btrfs_set_root_bytenr(&root_item, blocks[2]);
btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(root_item));
write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
nritems), sizeof(root_item));
nritems++;
itemoff = itemoff - sizeof(root_item);
btrfs_set_root_bytenr(&root_item, blocks[4]);
btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(root_item));
write_extent_buffer(buf, &root_item,
btrfs_item_ptr_offset(buf, nritems),
sizeof(root_item));
nritems++;
itemoff = itemoff - sizeof(root_item);
btrfs_set_root_bytenr(&root_item, blocks[5]);
btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(root_item));
write_extent_buffer(buf, &root_item,
btrfs_item_ptr_offset(buf, nritems),
sizeof(root_item));
nritems++;
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
itemoff = itemoff - sizeof(root_item);
btrfs_set_root_bytenr(&root_item, blocks[6]);
btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(root_item));
write_extent_buffer(buf, &root_item,
btrfs_item_ptr_offset(buf, nritems),
sizeof(root_item));
nritems++;
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[1]);
BUG_ON(ret != leafsize);
/* create the items for the extent tree */
nritems = 0;
itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
for (i = 1; i < 7; i++) {
BUG_ON(blocks[i] < first_free);
BUG_ON(blocks[i] < blocks[i - 1]);
/* create extent item */
itemoff -= sizeof(struct btrfs_extent_item) +
sizeof(struct btrfs_tree_block_info);
btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
btrfs_set_disk_key_offset(&disk_key, leafsize);
btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(struct btrfs_extent_item) +
sizeof(struct btrfs_tree_block_info));
extent_item = btrfs_item_ptr(buf, nritems,
struct btrfs_extent_item);
btrfs_set_extent_refs(buf, extent_item, 1);
btrfs_set_extent_generation(buf, extent_item, 1);
btrfs_set_extent_flags(buf, extent_item,
BTRFS_EXTENT_FLAG_TREE_BLOCK);
nritems++;
/* create extent ref */
ref_root = reference_root_table[i];
btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
btrfs_set_disk_key_offset(&disk_key, ref_root);
btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), 0);
nritems++;
}
btrfs_set_header_bytenr(buf, blocks[2]);
btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
btrfs_set_header_nritems(buf, nritems);
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[2]);
BUG_ON(ret != leafsize);
/* create the chunk tree */
nritems = 0;
item_size = sizeof(*dev_item);
itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
/* first device 1 (there is no device 0) */
btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
btrfs_set_disk_key_offset(&disk_key, 1);
btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
btrfs_set_device_id(buf, dev_item, 1);
btrfs_set_device_generation(buf, dev_item, 0);
btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
btrfs_set_device_bytes_used(buf, dev_item,
BTRFS_MKFS_SYSTEM_GROUP_SIZE);
btrfs_set_device_io_align(buf, dev_item, sectorsize);
btrfs_set_device_io_width(buf, dev_item, sectorsize);
btrfs_set_device_sector_size(buf, dev_item, sectorsize);
btrfs_set_device_type(buf, dev_item, 0);
write_extent_buffer(buf, super.dev_item.uuid,
(unsigned long)btrfs_device_uuid(dev_item),
BTRFS_UUID_SIZE);
write_extent_buffer(buf, super.fsid,
(unsigned long)btrfs_device_fsid(dev_item),
BTRFS_UUID_SIZE);
read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
sizeof(*dev_item));
nritems++;
item_size = btrfs_chunk_item_size(1);
itemoff = itemoff - item_size;
/* then we have chunk 0 */
btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
btrfs_set_disk_key_offset(&disk_key, 0);
btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
btrfs_set_chunk_io_align(buf, chunk, sectorsize);
btrfs_set_chunk_io_width(buf, chunk, sectorsize);
btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
btrfs_set_chunk_num_stripes(buf, chunk, 1);
btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
nritems++;
write_extent_buffer(buf, super.dev_item.uuid,
(unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
BTRFS_UUID_SIZE);
/* copy the key for the chunk to the system array */
ptr = super.sys_chunk_array;
array_size = sizeof(disk_key);
memcpy(ptr, &disk_key, sizeof(disk_key));
ptr += sizeof(disk_key);
/* copy the chunk to the system array */
read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
array_size += item_size;
ptr += item_size;
btrfs_set_super_sys_array_size(&super, array_size);
btrfs_set_header_bytenr(buf, blocks[3]);
btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
btrfs_set_header_nritems(buf, nritems);
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[3]);
/* create the device tree */
nritems = 0;
itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
sizeof(struct btrfs_dev_extent);
btrfs_set_disk_key_objectid(&disk_key, 1);
btrfs_set_disk_key_offset(&disk_key, 0);
btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
btrfs_set_item_key(buf, &disk_key, nritems);
btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
sizeof(struct btrfs_dev_extent));
dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
BTRFS_CHUNK_TREE_OBJECTID);
btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
BTRFS_FIRST_CHUNK_TREE_OBJECTID);
btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
write_extent_buffer(buf, chunk_tree_uuid,
(unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
BTRFS_UUID_SIZE);
btrfs_set_dev_extent_length(buf, dev_extent,
BTRFS_MKFS_SYSTEM_GROUP_SIZE);
nritems++;
btrfs_set_header_bytenr(buf, blocks[4]);
btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
btrfs_set_header_nritems(buf, nritems);
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[4]);
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
/* create the FS root */
btrfs_set_header_bytenr(buf, blocks[5]);
btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
btrfs_set_header_nritems(buf, 0);
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[5]);
BUG_ON(ret != leafsize);
Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:00:31 +00:00
/* finally create the csum root */
btrfs_set_header_bytenr(buf, blocks[6]);
btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
btrfs_set_header_nritems(buf, 0);
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, leafsize, blocks[6]);
BUG_ON(ret != leafsize);
/* and write out the super block */
BUG_ON(sizeof(super) > sectorsize);
memset(buf->data, 0, sectorsize);
memcpy(buf->data, &super, sizeof(super));
buf->len = sectorsize;
csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
BUG_ON(ret != sectorsize);
free(buf);
return 0;
}
2008-03-24 19:04:49 +00:00
static u64 device_size(int fd, struct stat *st)
{
u64 size;
if (S_ISREG(st->st_mode)) {
return st->st_size;
}
if (!S_ISBLK(st->st_mode)) {
return 0;
}
if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
return size;
}
return 0;
}
static int zero_blocks(int fd, off_t start, size_t len)
{
char *buf = malloc(len);
int ret = 0;
ssize_t written;
if (!buf)
return -ENOMEM;
memset(buf, 0, len);
written = pwrite(fd, buf, len, start);
if (written != len)
ret = -EIO;
free(buf);
return ret;
}
static int zero_dev_start(int fd)
{
off_t start = 0;
size_t len = 2 * 1024 * 1024;
#ifdef __sparc__
/* don't overwrite the disk labels on sparc */
start = 1024;
len -= 1024;
#endif
return zero_blocks(fd, start, len);
}
static int zero_dev_end(int fd, u64 dev_size)
{
size_t len = 2 * 1024 * 1024;
off_t start = dev_size - len;
return zero_blocks(fd, start, len);
}
int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
struct btrfs_root *root, int fd, char *path,
u64 block_count, u32 io_width, u32 io_align,
u32 sectorsize)
2008-03-24 19:04:49 +00:00
{
struct btrfs_super_block *disk_super;
struct btrfs_super_block *super = &root->fs_info->super_copy;
struct btrfs_device *device;
2008-03-24 19:04:49 +00:00
struct btrfs_dev_item *dev_item;
char *buf;
u64 total_bytes;
u64 num_devs;
int ret;
device = kmalloc(sizeof(*device), GFP_NOFS);
if (!device)
return -ENOMEM;
buf = kmalloc(sectorsize, GFP_NOFS);
if (!buf) {
kfree(device);
return -ENOMEM;
}
2008-03-24 19:04:49 +00:00
BUG_ON(sizeof(*disk_super) > sectorsize);
memset(buf, 0, sectorsize);
disk_super = (struct btrfs_super_block *)buf;
dev_item = &disk_super->dev_item;
uuid_generate(device->uuid);
device->devid = 0;
device->type = 0;
device->io_width = io_width;
device->io_align = io_align;
device->sector_size = sectorsize;
device->fd = fd;
device->writeable = 1;
device->total_bytes = block_count;
device->bytes_used = 0;
device->total_ios = 0;
device->dev_root = root->fs_info->dev_root;
ret = btrfs_add_device(trans, root, device);
2008-03-24 19:04:49 +00:00
BUG_ON(ret);
total_bytes = btrfs_super_total_bytes(super) + block_count;
btrfs_set_super_total_bytes(super, total_bytes);
num_devs = btrfs_super_num_devices(super) + 1;
btrfs_set_super_num_devices(super, num_devs);
memcpy(disk_super, super, sizeof(*disk_super));
printf("adding device %s id %llu\n", path,
(unsigned long long)device->devid);
btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
btrfs_set_stack_device_id(dev_item, device->devid);
btrfs_set_stack_device_type(dev_item, device->type);
btrfs_set_stack_device_io_align(dev_item, device->io_align);
btrfs_set_stack_device_io_width(dev_item, device->io_width);
btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
2008-03-24 19:04:49 +00:00
ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
BUG_ON(ret != sectorsize);
kfree(buf);
list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
device->fs_devices = root->fs_info->fs_devices;
2008-03-24 19:04:49 +00:00
return 0;
}
int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret)
{
u64 block_count;
u64 bytenr;
2008-03-24 19:04:49 +00:00
struct stat st;
int i, ret;
2008-03-24 19:04:49 +00:00
ret = fstat(fd, &st);
if (ret < 0) {
fprintf(stderr, "unable to stat %s\n", file);
exit(1);
}
block_count = device_size(fd, &st);
if (block_count == 0) {
fprintf(stderr, "unable to find %s size\n", file);
exit(1);
}
zero_end = 1;
if (block_count < 256 * 1024 * 1024) {
fprintf(stderr, "device %s is too small "
"(must be at least 256 MB)\n", file);
2008-03-24 19:04:49 +00:00
exit(1);
}
ret = zero_dev_start(fd);
if (ret) {
fprintf(stderr, "failed to zero device start %d\n", ret);
exit(1);
}
for (i = 0 ; i < BTRFS_SUPER_MIRROR_MAX; i++) {
bytenr = btrfs_sb_offset(i);
if (bytenr >= block_count)
break;
zero_blocks(fd, bytenr, BTRFS_SUPER_INFO_SIZE);
}
2008-03-24 19:04:49 +00:00
if (zero_end) {
ret = zero_dev_end(fd, block_count);
if (ret) {
fprintf(stderr, "failed to zero device end %d\n", ret);
exit(1);
}
}
*block_count_ret = block_count;
return 0;
}
int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 objectid)
{
int ret;
struct btrfs_inode_item inode_item;
memset(&inode_item, 0, sizeof(inode_item));
btrfs_set_stack_inode_generation(&inode_item, trans->transid);
btrfs_set_stack_inode_size(&inode_item, 0);
btrfs_set_stack_inode_nlink(&inode_item, 1);
btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0555);
if (root->fs_info->tree_root == root)
btrfs_set_super_root_dir(&root->fs_info->super_copy, objectid);
ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
if (ret)
goto error;
2008-07-24 16:13:32 +00:00
ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
if (ret)
goto error;
btrfs_set_root_dirid(&root->root_item, objectid);
ret = 0;
error:
return ret;
}
/*
* returns 1 if the device was mounted, < 0 on error or 0 if everything
* is safe to continue. TODO, this should also scan multi-device filesystems
*/
int check_mounted(char *file)
{
struct mntent *mnt;
struct stat st_buf;
dev_t file_dev = 0;
dev_t file_rdev = 0;
ino_t file_ino = 0;
FILE *f;
int ret = 0;
if ((f = setmntent ("/proc/mounts", "r")) == NULL)
return -errno;
if (stat(file, &st_buf) < 0) {
return -errno;
} else {
if (S_ISBLK(st_buf.st_mode)) {
file_rdev = st_buf.st_rdev;
} else {
file_dev = st_buf.st_dev;
file_ino = st_buf.st_ino;
}
}
while ((mnt = getmntent (f)) != NULL) {
if (strcmp(file, mnt->mnt_fsname) == 0)
break;
if (stat(mnt->mnt_fsname, &st_buf) == 0) {
if (S_ISBLK(st_buf.st_mode)) {
if (file_rdev && (file_rdev == st_buf.st_rdev))
break;
} else if (file_dev && ((file_dev == st_buf.st_dev) &&
(file_ino == st_buf.st_ino))) {
break;
}
}
}
if (mnt) {
/* found an entry in mnt table */
ret = 1;
}
endmntent (f);
return ret;
}
struct pending_dir {
struct list_head list;
char name[256];
};
void btrfs_register_one_device(char *fname)
{
struct btrfs_ioctl_vol_args args;
int fd;
int ret;
fd = open("/dev/btrfs-control", O_RDONLY);
if (fd < 0) {
fprintf(stderr, "failed to open /dev/btrfs-control "
"skipping device registration\n");
return;
}
strcpy(args.name, fname);
ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
close(fd);
}
int btrfs_scan_one_dir(char *dirname, int run_ioctl)
{
DIR *dirp = NULL;
struct dirent *dirent;
struct pending_dir *pending;
struct stat st;
int ret;
int fd;
int dirname_len;
int pathlen;
char *fullpath;
struct list_head pending_list;
struct btrfs_fs_devices *tmp_devices;
u64 num_devices;
INIT_LIST_HEAD(&pending_list);
pending = malloc(sizeof(*pending));
if (!pending)
return -ENOMEM;
strcpy(pending->name, dirname);
again:
dirname_len = strlen(pending->name);
pathlen = 1024;
fullpath = malloc(pathlen);
dirname = pending->name;
if (!fullpath) {
ret = -ENOMEM;
goto fail;
}
dirp = opendir(dirname);
if (!dirp) {
fprintf(stderr, "Unable to open /sys/block for scanning\n");
return -ENOENT;
}
while(1) {
dirent = readdir(dirp);
if (!dirent)
break;
if (dirent->d_name[0] == '.')
continue;
if (dirname_len + strlen(dirent->d_name) + 2 > pathlen) {
ret = -EFAULT;
goto fail;
}
snprintf(fullpath, pathlen, "%s/%s", dirname, dirent->d_name);
ret = lstat(fullpath, &st);
if (ret < 0) {
fprintf(stderr, "failed to stat %s\n", fullpath);
continue;
}
if (S_ISLNK(st.st_mode))
continue;
if (S_ISDIR(st.st_mode)) {
struct pending_dir *next = malloc(sizeof(*next));
if (!next) {
ret = -ENOMEM;
goto fail;
}
strcpy(next->name, fullpath);
list_add_tail(&next->list, &pending_list);
}
if (!S_ISBLK(st.st_mode)) {
continue;
}
fd = open(fullpath, O_RDONLY);
if (fd < 0) {
fprintf(stderr, "failed to read %s\n", fullpath);
continue;
}
ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
&num_devices,
BTRFS_SUPER_INFO_OFFSET);
if (ret == 0 && run_ioctl > 0) {
btrfs_register_one_device(fullpath);
}
close(fd);
}
if (!list_empty(&pending_list)) {
free(pending);
pending = list_entry(pending_list.next, struct pending_dir,
list);
list_del(&pending->list);
closedir(dirp);
goto again;
}
ret = 0;
fail:
free(pending);
if (dirp)
closedir(dirp);
return ret;
}
int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs,
int run_ioctls)
{
return btrfs_scan_one_dir("/dev", run_ioctls);
}
int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
int super_offset)
{
struct btrfs_super_block *disk_super;
char *buf;
int ret = 0;
buf = malloc(BTRFS_SUPER_INFO_SIZE);
if (!buf) {
ret = -ENOMEM;
goto out;
}
ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
if (ret != BTRFS_SUPER_INFO_SIZE)
goto brelse;
ret = 0;
disk_super = (struct btrfs_super_block *)buf;
if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
sizeof(disk_super->magic)))
goto brelse;
if (!memcmp(disk_super->fsid, root->fs_info->super_copy.fsid,
BTRFS_FSID_SIZE))
ret = 1;
brelse:
free(buf);
out:
return ret;
}
static char *size_strs[] = { "", "KB", "MB", "GB", "TB",
"PB", "EB", "ZB", "YB"};
char *pretty_sizes(u64 size)
{
int num_divs = 0;
u64 last_size = size;
u64 fract_size = size;
float fraction;
char *pretty;
while(size > 0) {
fract_size = last_size;
last_size = size;
size /= 1024;
num_divs++;
}
if (num_divs == 0)
num_divs = 1;
if (num_divs > ARRAY_SIZE(size_strs))
return NULL;
fraction = (float)fract_size / 1024;
pretty = malloc(16);
sprintf(pretty, "%.2f%s", fraction, size_strs[num_divs-1]);
return pretty;
}