Btrfs: still corruption hunting
[safe/jmp/linux-2.6] / fs / btrfs / disk-io.c
index 14955e4..2dbd550 100644 (file)
-#define _XOPEN_SOURCE 500
-#include <stdio.h>
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <unistd.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/blkdev.h>
+#include <linux/crypto.h>
+#include <linux/scatterlist.h>
+#include <linux/swap.h>
 #include "ctree.h"
 #include "disk-io.h"
+#include "transaction.h"
 
-static int allocated_blocks = 0;
-
-static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
-       if (buf->blocknr != buf->node.header.blocknr)
+       struct btrfs_node *node = btrfs_buffer_node(buf);
+       if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
                BUG();
-       if (root->node && buf->node.header.parentid != root->node->node.header.parentid)
+       }
+       if (root->node && btrfs_header_parentid(&node->header) !=
+           btrfs_header_parentid(btrfs_buffer_header(root->node))) {
                BUG();
+       }
        return 0;
 }
 
-struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
+struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-       struct tree_buffer *buf;
-       int ret;
-       buf = malloc(sizeof(struct tree_buffer));
-       if (!buf)
-               return buf;
-       allocated_blocks++;
-       buf->blocknr = blocknr;
-       buf->count = 1;
-       radix_tree_preload(GFP_KERNEL);
-       ret = radix_tree_insert(&root->cache_radix, blocknr, buf);
-       radix_tree_preload_end();
-       if (ret) {
-               free(buf);
+       return sb_find_get_block(root->fs_info->sb, blocknr);
+#if 0
+       struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+       int blockbits = root->fs_info->sb->s_blocksize_bits;
+       unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+       struct page *page;
+       struct buffer_head *bh;
+       struct buffer_head *head;
+       struct buffer_head *ret = NULL;
+
+
+       page = find_lock_page(mapping, index);
+       if (!page)
                return NULL;
+
+       if (!page_has_buffers(page))
+               goto out_unlock;
+
+       head = page_buffers(page);
+       bh = head;
+       do {
+               if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
+                       ret = bh;
+                       get_bh(bh);
+                       goto out_unlock;
+               }
+               bh = bh->b_this_page;
+       } while (bh != head);
+out_unlock:
+       unlock_page(page);
+       if (ret) {
+               touch_buffer(ret);
        }
-       return buf;
+       page_cache_release(page);
+       return ret;
+#endif
 }
 
-struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr)
+struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
+                                                u64 blocknr)
 {
-       struct tree_buffer *buf;
-       buf = radix_tree_lookup(&root->cache_radix, blocknr);
-       if (buf) {
-               buf->count++;
-       } else {
-               buf = alloc_tree_block(root, blocknr);
-               if (!buf) {
-                       BUG();
-                       return NULL;
+       return sb_getblk(root->fs_info->sb, blocknr);
+#if 0
+       struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+       int blockbits = root->fs_info->sb->s_blocksize_bits;
+       unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+       struct page *page;
+       struct buffer_head *bh;
+       struct buffer_head *head;
+       struct buffer_head *ret = NULL;
+       u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
+
+       page = grab_cache_page(mapping, index);
+       if (!page)
+               return NULL;
+
+       if (!page_has_buffers(page))
+               create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
+       head = page_buffers(page);
+       bh = head;
+       do {
+               if (!buffer_mapped(bh)) {
+                       bh->b_bdev = root->fs_info->sb->s_bdev;
+                       bh->b_blocknr = first_block;
+                       set_buffer_mapped(bh);
+               }
+               if (bh->b_blocknr == blocknr) {
+                       ret = bh;
+                       get_bh(bh);
+                       goto out_unlock;
                }
+               bh = bh->b_this_page;
+               first_block++;
+       } while (bh != head);
+out_unlock:
+       unlock_page(page);
+       if (ret)
+               touch_buffer(ret);
+       page_cache_release(page);
+       return ret;
+#endif
+}
+
+static sector_t max_block(struct block_device *bdev)
+{
+       sector_t retval = ~((sector_t)0);
+       loff_t sz = i_size_read(bdev->bd_inode);
+
+       if (sz) {
+               unsigned int size = block_size(bdev);
+               unsigned int sizebits = blksize_bits(size);
+               retval = (sz >> sizebits);
        }
-       return buf;
+       return retval;
 }
 
+static int btree_get_block(struct inode *inode, sector_t iblock,
+                          struct buffer_head *bh, int create)
+{
+       if (iblock >= max_block(inode->i_sb->s_bdev)) {
+               if (create)
+                       return -EIO;
 
-struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
+               /*
+                * for reads, we're just trying to fill a partial page.
+                * return a hole, they will have to call get_block again
+                * before they can fill it, and they will get -EIO at that
+                * time
+                */
+               return 0;
+       }
+       bh->b_bdev = inode->i_sb->s_bdev;
+       bh->b_blocknr = iblock;
+       set_buffer_mapped(bh);
+       return 0;
+}
+
+int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
+                   char *result)
 {
-       loff_t offset = blocknr * CTREE_BLOCKSIZE;
-       struct tree_buffer *buf;
+       struct scatterlist sg;
+       struct crypto_hash *tfm = root->fs_info->hash_tfm;
+       struct hash_desc desc;
        int ret;
 
-       buf = radix_tree_lookup(&root->cache_radix, blocknr);
-       if (buf) {
-               buf->count++;
-       } else {
-               buf = alloc_tree_block(root, blocknr);
-               if (!buf)
-                       return NULL;
-               ret = pread(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
-               if (ret != CTREE_BLOCKSIZE) {
-                       free(buf);
-                       return NULL;
+       desc.tfm = tfm;
+       desc.flags = 0;
+       sg_init_one(&sg, data, len);
+       spin_lock(&root->fs_info->hash_lock);
+       ret = crypto_hash_digest(&desc, &sg, 1, result);
+       spin_unlock(&root->fs_info->hash_lock);
+       if (ret) {
+               printk("sha256 digest failed\n");
+       }
+       return ret;
+}
+static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
+                          int verify)
+{
+       char result[BTRFS_CSUM_SIZE];
+       int ret;
+       struct btrfs_node *node;
+
+       return 0;
+       ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
+                             bh->b_size - BTRFS_CSUM_SIZE, result);
+       if (ret)
+               return ret;
+       if (verify) {
+               if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
+                       printk("checksum verify failed on %lu\n",
+                              bh->b_blocknr);
+                       return 1;
                }
+       } else {
+               node = btrfs_buffer_node(bh);
+               memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
        }
-       if (check_tree_block(root, buf))
-               BUG();
-       return buf;
+       return 0;
 }
 
-int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
-       u64 blocknr = buf->blocknr;
-       loff_t offset = blocknr * CTREE_BLOCKSIZE;
-       int ret;
+#if 0
+       struct buffer_head *bh;
+       struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
+       struct buffer_head *head;
+       if (!page_has_buffers(page)) {
+               create_empty_buffers(page, root->fs_info->sb->s_blocksize,
+                                       (1 << BH_Dirty)|(1 << BH_Uptodate));
+       }
+       head = page_buffers(page);
+       bh = head;
+       do {
+               if (buffer_dirty(bh))
+                       csum_tree_block(root, bh, 0);
+               bh = bh->b_this_page;
+       } while (bh != head);
+#endif
+       return block_write_full_page(page, btree_get_block, wbc);
+}
+
+static int btree_readpage(struct file * file, struct page * page)
+{
+       return block_read_full_page(page, btree_get_block);
+}
+
+static struct address_space_operations btree_aops = {
+       .readpage       = btree_readpage,
+       .writepage      = btree_writepage,
+       .sync_page      = block_sync_page,
+};
 
-       if (buf->blocknr != buf->node.header.blocknr)
+struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
+{
+       return sb_bread(root->fs_info->sb, blocknr);
+#if 0
+       struct buffer_head *bh = NULL;
+
+       bh = btrfs_find_create_tree_block(root, blocknr);
+       if (!bh)
+               return bh;
+       lock_buffer(bh);
+       if (!buffer_uptodate(bh)) {
+               get_bh(bh);
+               bh->b_end_io = end_buffer_read_sync;
+               submit_bh(READ, bh);
+               wait_on_buffer(bh);
+               if (!buffer_uptodate(bh))
+                       goto fail;
+               csum_tree_block(root, bh, 1);
+       } else {
+               unlock_buffer(bh);
+       }
+       if (check_tree_block(root, bh))
                BUG();
-       ret = pwrite(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
-       if (ret != CTREE_BLOCKSIZE)
-               return ret;
+       return bh;
+fail:
+       brelse(bh);
+       return NULL;
+
+#endif
+}
+
+int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+                    struct buffer_head *buf)
+{
+       WARN_ON(atomic_read(&buf->b_count) == 0);
+       mark_buffer_dirty(buf);
+       return 0;
+}
+
+int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+                    struct buffer_head *buf)
+{
+       WARN_ON(atomic_read(&buf->b_count) == 0);
+       clear_buffer_dirty(buf);
        return 0;
 }
 
-static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
-                       struct ctree_root_info *info, int fp)
+static int __setup_root(int blocksize,
+                       struct btrfs_root *root,
+                       struct btrfs_fs_info *fs_info,
+                       u64 objectid)
 {
-       root->fp = fp;
        root->node = NULL;
-       root->node = read_tree_block(root, info->tree_root);
-       root->extent_root = extent_root;
+       root->commit_root = NULL;
+       root->blocksize = blocksize;
+       root->ref_cows = 0;
+       root->fs_info = fs_info;
+       memset(&root->root_key, 0, sizeof(root->root_key));
+       memset(&root->root_item, 0, sizeof(root->root_item));
        return 0;
 }
 
-struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
+static int find_and_setup_root(int blocksize,
+                              struct btrfs_root *tree_root,
+                              struct btrfs_fs_info *fs_info,
+                              u64 objectid,
+                              struct btrfs_root *root)
 {
-       struct ctree_root *root = malloc(sizeof(struct ctree_root));
-       struct ctree_root *extent_root = malloc(sizeof(struct ctree_root));
-       int fp;
        int ret;
 
-       fp = open(filename, O_CREAT | O_RDWR);
-       if (fp < 0) {
-               free(root);
+       __setup_root(blocksize, root, fs_info, objectid);
+       ret = btrfs_find_last_root(tree_root, objectid,
+                                  &root->root_item, &root->root_key);
+       BUG_ON(ret);
+
+       root->node = read_tree_block(root,
+                                    btrfs_root_blocknr(&root->root_item));
+       BUG_ON(!root->node);
+       return 0;
+}
+
+struct btrfs_root *open_ctree(struct super_block *sb)
+{
+       struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
+                                         GFP_NOFS);
+       struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
+                                                GFP_NOFS);
+       struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
+                                              GFP_NOFS);
+       struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
+                                               GFP_NOFS);
+       struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
+                                               GFP_NOFS);
+       int ret;
+       struct btrfs_super_block *disk_super;
+
+       init_bit_radix(&fs_info->pinned_radix);
+       init_bit_radix(&fs_info->pending_del_radix);
+       sb_set_blocksize(sb, 4096);
+       fs_info->running_transaction = NULL;
+       fs_info->fs_root = root;
+       fs_info->tree_root = tree_root;
+       fs_info->extent_root = extent_root;
+       fs_info->inode_root = inode_root;
+       fs_info->last_inode_alloc = 0;
+       fs_info->last_inode_alloc_dirid = 0;
+       fs_info->sb = sb;
+       fs_info->btree_inode = NULL;
+#if 0
+       fs_info->btree_inode = new_inode(sb);
+       fs_info->btree_inode->i_ino = 1;
+       fs_info->btree_inode->i_nlink = 1;
+       fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
+       fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+       insert_inode_hash(fs_info->btree_inode);
+       mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+#endif
+       fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
+       spin_lock_init(&fs_info->hash_lock);
+       if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
+               printk("failed to allocate sha256 hash\n");
+               return NULL;
+       }
+       mutex_init(&fs_info->trans_mutex);
+       mutex_init(&fs_info->fs_mutex);
+       memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
+       memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
+
+       __setup_root(sb->s_blocksize, tree_root,
+                    fs_info, BTRFS_ROOT_TREE_OBJECTID);
+       fs_info->sb_buffer = read_tree_block(tree_root,
+                                            BTRFS_SUPER_INFO_OFFSET /
+                                            sb->s_blocksize);
+
+       if (!fs_info->sb_buffer) {
+printk("failed2\n");
                return NULL;
        }
-       INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
-       INIT_RADIX_TREE(&extent_root->cache_radix, GFP_KERNEL);
-       ret = pread(fp, super, sizeof(struct ctree_super_block),
-                    CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
-       if (ret == 0 || super->root_info.tree_root == 0) {
-               printf("making new FS!\n");
-               ret = mkfs(fp);
-               if (ret)
-                       return NULL;
-               ret = pread(fp, super, sizeof(struct ctree_super_block),
-                            CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
-               if (ret != sizeof(struct ctree_super_block))
-                       return NULL;
+       disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
+       if (!btrfs_super_root(disk_super)) {
+               return NULL;
        }
-       BUG_ON(ret < 0);
-       __setup_root(root, extent_root, &super->root_info, fp);
-       __setup_root(extent_root, extent_root, &super->extent_info, fp);
+       fs_info->disk_super = disk_super;
+       tree_root->node = read_tree_block(tree_root,
+                                         btrfs_super_root(disk_super));
+       BUG_ON(!tree_root->node);
+
+       mutex_lock(&fs_info->fs_mutex);
+       ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
+                                 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
+       BUG_ON(ret);
+
+       ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
+                                 BTRFS_INODE_MAP_OBJECTID, inode_root);
+       BUG_ON(ret);
+
+       ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
+                                 BTRFS_FS_TREE_OBJECTID, root);
+       mutex_unlock(&fs_info->fs_mutex);
+       BUG_ON(ret);
+       root->commit_root = root->node;
+       get_bh(root->node);
+       root->ref_cows = 1;
+       root->fs_info->generation = root->root_key.offset + 1;
        return root;
 }
 
-static int __update_root(struct ctree_root *root, struct ctree_root_info *info)
+int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
+                     *root)
 {
-       info->tree_root = root->node->blocknr;
-       return 0;
-}
+       struct buffer_head *bh = root->fs_info->sb_buffer;
 
-int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s)
-{
-       int ret;
-       __update_root(root, &s->root_info);
-       __update_root(root->extent_root, &s->extent_info);
-       ret = pwrite(root->fp, s, sizeof(*s), CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
-       if (ret != sizeof(*s)) {
-               fprintf(stderr, "failed to write new super block err %d\n", ret);
-               return ret;
+       btrfs_set_super_root(root->fs_info->disk_super,
+                            root->fs_info->tree_root->node->b_blocknr);
+       lock_buffer(bh);
+       WARN_ON(atomic_read(&bh->b_count) < 1);
+       clear_buffer_dirty(bh);
+       csum_tree_block(root, bh, 0);
+       bh->b_end_io = end_buffer_write_sync;
+       get_bh(bh);
+       submit_bh(WRITE, bh);
+       wait_on_buffer(bh);
+       if (!buffer_uptodate(bh)) {
+               WARN_ON(1);
+               return -EIO;
        }
        return 0;
 }
 
-int close_ctree(struct ctree_root *root)
+int close_ctree(struct btrfs_root *root)
 {
-       close(root->fp);
+       int ret;
+       struct btrfs_trans_handle *trans;
+
+       mutex_lock(&root->fs_info->fs_mutex);
+       trans = btrfs_start_transaction(root, 1);
+       btrfs_commit_transaction(trans, root);
+       /* run commit again to  drop the original snapshot */
+       trans = btrfs_start_transaction(root, 1);
+       btrfs_commit_transaction(trans, root);
+       ret = btrfs_write_and_wait_transaction(NULL, root);
+       BUG_ON(ret);
+       write_ctree_super(NULL, root);
+       mutex_unlock(&root->fs_info->fs_mutex);
+
        if (root->node)
-               tree_block_release(root, root->node);
-       if (root->extent_root->node)
-               tree_block_release(root->extent_root, root->extent_root->node);
-       free(root);
-       printf("on close %d blocks are allocated\n", allocated_blocks);
+               btrfs_block_release(root, root->node);
+       if (root->fs_info->extent_root->node)
+               btrfs_block_release(root->fs_info->extent_root,
+                                   root->fs_info->extent_root->node);
+       if (root->fs_info->inode_root->node)
+               btrfs_block_release(root->fs_info->inode_root,
+                                   root->fs_info->inode_root->node);
+       if (root->fs_info->tree_root->node)
+               btrfs_block_release(root->fs_info->tree_root,
+                                   root->fs_info->tree_root->node);
+       btrfs_block_release(root, root->commit_root);
+       btrfs_block_release(root, root->fs_info->sb_buffer);
+       crypto_free_hash(root->fs_info->hash_tfm);
+       // truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
+       // iput(root->fs_info->btree_inode);
+       kfree(root->fs_info->extent_root);
+       kfree(root->fs_info->inode_root);
+       kfree(root->fs_info->tree_root);
+       kfree(root->fs_info);
+       kfree(root);
        return 0;
 }
 
-void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
+void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 {
-       buf->count--;
-       write_tree_block(root, buf);
-       if (buf->count < 0)
-               BUG();
-       if (buf->count == 0) {
-               if (!radix_tree_lookup(&root->cache_radix, buf->blocknr))
-                       BUG();
-               radix_tree_delete(&root->cache_radix, buf->blocknr);
-               memset(buf, 0, sizeof(*buf));
-               free(buf);
-               BUG_ON(allocated_blocks == 0);
-               allocated_blocks--;
-       }
+       // brelse(buf);
 }