Btrfs: Add back metadata checksumming
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index d1c6f02..1af0a96 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -310,9 +310,11 @@
struct extent_buffer *sb_buffer;
struct super_block *sb;
struct inode *btree_inode;
+ spinlock_t hash_lock;
struct mutex trans_mutex;
struct mutex fs_mutex;
struct list_head trans_list;
+ struct list_head hashers;
struct list_head dead_roots;
struct delayed_work trans_work;
struct kobject super_kobj;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 16f0260..1176e54 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -60,7 +60,7 @@
struct extent_buffer *eb;
eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
- bytenr, blocksize, GFP_NOFS);
+ bytenr, blocksize, NULL, GFP_NOFS);
return eb;
}
@@ -99,10 +99,102 @@
return em;
}
+u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
+{
+ return crc32c(seed, data, len);
+}
+
+void btrfs_csum_final(u32 crc, char *result)
+{
+ *(__le32 *)result = ~cpu_to_le32(crc);
+}
+
+static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
+ int verify)
+{
+ char result[BTRFS_CRC32_SIZE];
+ unsigned long len;
+ unsigned long cur_len;
+ unsigned long offset = BTRFS_CSUM_SIZE;
+ char *map_token = NULL;
+ char *kaddr;
+ unsigned long map_start;
+ unsigned long map_len;
+ int err;
+ u32 crc = ~(u32)0;
+
+ len = buf->len - offset;
+ while(len > 0) {
+ err = map_private_extent_buffer(buf, offset, 32,
+ &map_token, &kaddr,
+ &map_start, &map_len, KM_USER0);
+ if (err) {
+ printk("failed to map extent buffer! %lu\n",
+ offset);
+ return 1;
+ }
+ cur_len = min(len, map_len - (offset - map_start));
+ crc = btrfs_csum_data(root, kaddr + offset - map_start,
+ crc, cur_len);
+ len -= cur_len;
+ offset += cur_len;
+ unmap_extent_buffer(buf, map_token, KM_USER0);
+ }
+ btrfs_csum_final(crc, result);
+
+ if (verify) {
+ if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
+ printk("btrfs: %s checksum verify failed on %llu\n",
+ root->fs_info->sb->s_id,
+ buf->start);
+ return 1;
+ }
+ } else {
+ write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
+ }
+ return 0;
+}
+
+
+int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
+{
+ struct extent_map_tree *tree;
+ u64 start = page->index << PAGE_CACHE_SHIFT;
+ u64 found_start;
+ int found_level;
+ unsigned long len;
+ struct extent_buffer *eb;
+ tree = &BTRFS_I(page->mapping->host)->extent_tree;
+
+ if (page->private == EXTENT_PAGE_PRIVATE)
+ goto out;
+ if (!page->private)
+ goto out;
+ len = page->private >> 2;
+ if (len == 0) {
+ WARN_ON(1);
+ }
+ eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
+ read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1);
+ found_start = btrfs_header_bytenr(eb);
+ if (found_start != start) {
+ printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
+ start, found_start, len);
+ }
+ found_level = btrfs_header_level(eb);
+ csum_tree_block(root, eb, 0);
+ free_extent_buffer(eb);
+out:
+ return 0;
+}
+
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
struct extent_map_tree *tree;
+ struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
tree = &BTRFS_I(page->mapping->host)->extent_tree;
+
+ csum_dirty_buffer(root, page);
return extent_write_full_page(tree, page, btree_get_extent, wbc);
}
int btree_readpage(struct file *file, struct page *page)
@@ -117,7 +209,6 @@
struct extent_map_tree *tree;
int ret;
- BUG_ON(page->private != 1);
tree = &BTRFS_I(page->mapping->host)->extent_tree;
ret = try_release_extent_mapping(tree, page);
if (ret == 1) {
@@ -136,46 +227,6 @@
btree_releasepage(page, GFP_NOFS);
}
-int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
- char *result)
-{
- return 0;
-#if 0
- u32 crc;
- crc = crc32c(0, data, len);
- memcpy(result, &crc, BTRFS_CRC32_SIZE);
- return 0;
-#endif
-}
-
-#if 0
-static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
- int verify)
-{
- return 0;
- char result[BTRFS_CRC32_SIZE];
- int ret;
- struct btrfs_node *node;
-
- 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_CRC32_SIZE)) {
- printk("btrfs: %s checksum verify failed on %llu\n",
- root->fs_info->sb->s_id,
- (unsigned long long)bh_blocknr(bh));
- return 1;
- }
- } else {
- node = btrfs_buffer_node(bh);
- memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
- }
- return 0;
-}
-#endif
-
#if 0
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
@@ -215,7 +266,7 @@
if (!buf)
return 0;
read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
- buf, 0);
+ buf, 0, 0);
free_extent_buffer(buf);
return ret;
}
@@ -225,12 +276,29 @@
{
struct extent_buffer *buf = NULL;
struct inode *btree_inode = root->fs_info->btree_inode;
+ struct extent_map_tree *extent_tree;
+ int ret;
+
+ extent_tree = &BTRFS_I(btree_inode)->extent_tree;
buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
if (!buf)
return NULL;
read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
- buf, 1);
+ buf, 0, 1);
+ if (buf->flags & EXTENT_CSUM) {
+ return buf;
+ }
+ if (test_range_bit(extent_tree, buf->start, buf->start + buf->len - 1,
+ EXTENT_CSUM, 1)) {
+ buf->flags |= EXTENT_CSUM;
+ return buf;
+ }
+ ret = csum_tree_block(root, buf, 1);
+ set_extent_bits(extent_tree, buf->start,
+ buf->start + buf->len - 1,
+ EXTENT_CSUM, GFP_NOFS);
+ buf->flags |= EXTENT_CSUM;
return buf;
}
@@ -251,13 +319,6 @@
return 0;
}
-int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
-{
- struct inode *btree_inode = root->fs_info->btree_inode;
- set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
- return 0;
-}
-
static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
struct btrfs_root *root,
struct btrfs_fs_info *fs_info,
@@ -416,7 +477,24 @@
return root;
}
+#if 0
+static int add_hasher(struct btrfs_fs_info *info, char *type) {
+ struct btrfs_hasher *hasher;
+ hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
+ if (!hasher)
+ return -ENOMEM;
+ hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
+ if (!hasher->hash_tfm) {
+ kfree(hasher);
+ return -EINVAL;
+ }
+ spin_lock(&info->hash_lock);
+ list_add(&hasher->list, &info->hashers);
+ spin_unlock(&info->hash_lock);
+ return 0;
+}
+#endif
struct btrfs_root *open_ctree(struct super_block *sb)
{
u32 sectorsize;
@@ -440,6 +518,9 @@
INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
INIT_LIST_HEAD(&fs_info->trans_list);
INIT_LIST_HEAD(&fs_info->dead_roots);
+ INIT_LIST_HEAD(&fs_info->hashers);
+ spin_lock_init(&fs_info->hash_lock);
+
memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
init_completion(&fs_info->kobj_unregister);
sb_set_blocksize(sb, 4096);
@@ -479,6 +560,14 @@
mutex_init(&fs_info->trans_mutex);
mutex_init(&fs_info->fs_mutex);
+#if 0
+ ret = add_hasher(fs_info, "crc32c");
+ if (ret) {
+ printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
+ err = -ENOMEM;
+ goto fail_iput;
+ }
+#endif
__setup_root(512, 512, 512, tree_root,
fs_info, BTRFS_ROOT_TREE_OBJECTID);
@@ -509,25 +598,21 @@
i_size_write(fs_info->btree_inode,
btrfs_super_total_bytes(disk_super));
-
if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
sizeof(disk_super->magic))) {
printk("btrfs: valid FS not found on %s\n", sb->s_id);
goto fail_sb_buffer;
}
+
blocksize = btrfs_level_size(tree_root,
btrfs_super_root_level(disk_super));
+
tree_root->node = read_tree_block(tree_root,
btrfs_super_root(disk_super),
blocksize);
if (!tree_root->node)
goto fail_sb_buffer;
-#if 0
- btrfs_print_leaf(tree_root, tree_root->node);
- err = -EIO;
- goto fail_tree_root;
-#endif
mutex_lock(&fs_info->fs_mutex);
ret = find_and_setup_root(tree_root, fs_info,
@@ -634,9 +719,19 @@
btrfs_free_block_groups(root->fs_info);
del_fs_roots(fs_info);
- extent_map_tree_cleanup(&BTRFS_I(fs_info->btree_inode)->extent_tree);
+ extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree);
truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
iput(fs_info->btree_inode);
+#if 0
+ while(!list_empty(&fs_info->hashers)) {
+ struct btrfs_hasher *hasher;
+ hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
+ hashers);
+ list_del(&hasher->hashers);
+ crypto_free_hash(&fs_info->hash_tfm);
+ kfree(hasher);
+ }
+#endif
kfree(fs_info->extent_root);
kfree(fs_info->tree_root);
return 0;
@@ -733,5 +828,5 @@
struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
struct inode *btree_inode = root->fs_info->btree_inode;
return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
- buf, 1);
+ buf, 0, 1);
}
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index e4e68ea..6b2f2b4 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1178,6 +1178,10 @@
btrfs_set_buffer_uptodate(buf);
set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
buf->start + buf->len - 1, GFP_NOFS);
+ set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->extent_tree,
+ buf->start, buf->start + buf->len - 1,
+ EXTENT_CSUM, GFP_NOFS);
+ buf->flags |= EXTENT_CSUM;
btrfs_set_buffer_defrag(buf);
trans->blocks_used++;
return buf;
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index f8aaba8..2a8bc4b 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -81,7 +81,7 @@
}
EXPORT_SYMBOL(extent_map_tree_init);
-void extent_map_tree_cleanup(struct extent_map_tree *tree)
+void extent_map_tree_empty_lru(struct extent_map_tree *tree)
{
struct extent_buffer *eb;
while(!list_empty(&tree->buffer_lru)) {
@@ -91,7 +91,7 @@
free_extent_buffer(eb);
}
}
-EXPORT_SYMBOL(extent_map_tree_cleanup);
+EXPORT_SYMBOL(extent_map_tree_empty_lru);
struct extent_map *alloc_extent_map(gfp_t mask)
{
@@ -1464,7 +1464,7 @@
if (!PagePrivate(page)) {
SetPagePrivate(page);
WARN_ON(!page->mapping->a_ops->invalidatepage);
- set_page_private(page, 1);
+ set_page_private(page, EXTENT_PAGE_PRIVATE);
page_cache_get(page);
}
}
@@ -1979,8 +1979,9 @@
spin_lock(&tree->lru_lock);
eb = find_lru(tree, start, len);
- if (eb)
+ if (eb) {
goto lru_add;
+ }
spin_unlock(&tree->lru_lock);
if (eb) {
@@ -2007,6 +2008,7 @@
struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
u64 start, unsigned long len,
+ struct page *page0,
gfp_t mask)
{
unsigned long num_pages = num_extent_pages(start, len);
@@ -2024,7 +2026,18 @@
if (eb->flags & EXTENT_BUFFER_FILLED)
return eb;
- for (i = 0; i < num_pages; i++, index++) {
+ if (page0) {
+ eb->first_page = page0;
+ i = 1;
+ index++;
+ page_cache_get(page0);
+ set_page_extent_mapped(page0);
+ set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ i = 0;
+ }
+ for (; i < num_pages; i++, index++) {
p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
if (!p) {
WARN_ON(1);
@@ -2036,8 +2049,13 @@
goto fail;
}
set_page_extent_mapped(p);
- if (i == 0)
+ if (i == 0) {
eb->first_page = p;
+ set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ set_page_private(p, EXTENT_PAGE_PRIVATE);
+ }
if (!PageUptodate(p))
uptodate = 0;
unlock_page(p);
@@ -2057,8 +2075,7 @@
gfp_t mask)
{
unsigned long num_pages = num_extent_pages(start, len);
- unsigned long i;
- unsigned long index = start >> PAGE_CACHE_SHIFT;
+ unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT;
struct extent_buffer *eb;
struct page *p;
struct address_space *mapping = tree->mapping;
@@ -2082,8 +2099,15 @@
goto fail;
}
set_page_extent_mapped(p);
- if (i == 0)
+
+ if (i == 0) {
eb->first_page = p;
+ set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ len << 2);
+ } else {
+ set_page_private(p, EXTENT_PAGE_PRIVATE);
+ }
+
if (!PageUptodate(p))
uptodate = 0;
unlock_page(p);
@@ -2174,7 +2198,21 @@
num_pages = num_extent_pages(eb->start, eb->len);
for (i = 0; i < num_pages; i++) {
+ struct page *page = extent_buffer_page(eb, i);
+ /* writepage may need to do something special for the
+ * first page, we have to make sure page->private is
+ * properly set. releasepage may drop page->private
+ * on us if the page isn't already dirty.
+ */
+ if (i == 0) {
+ lock_page(page);
+ set_page_private(page,
+ EXTENT_PAGE_PRIVATE_FIRST_PAGE |
+ eb->len << 2);
+ }
__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
+ if (i == 0)
+ unlock_page(page);
}
return set_extent_dirty(tree, eb->start,
eb->start + eb->len - 1, GFP_NOFS);
@@ -2217,9 +2255,12 @@
EXPORT_SYMBOL(extent_buffer_uptodate);
int read_extent_buffer_pages(struct extent_map_tree *tree,
- struct extent_buffer *eb, int wait)
+ struct extent_buffer *eb,
+ u64 start,
+ int wait)
{
unsigned long i;
+ unsigned long start_i;
struct page *page;
int err;
int ret = 0;
@@ -2232,9 +2273,16 @@
EXTENT_UPTODATE, 1)) {
return 0;
}
+ if (start) {
+ WARN_ON(start < eb->start);
+ start_i = (start >> PAGE_CACHE_SHIFT) -
+ (eb->start >> PAGE_CACHE_SHIFT);
+ } else {
+ start_i = 0;
+ }
num_pages = num_extent_pages(eb->start, eb->len);
- for (i = 0; i < num_pages; i++) {
+ for (i = start_i; i < num_pages; i++) {
page = extent_buffer_page(eb, i);
if (PageUptodate(page)) {
continue;
@@ -2260,7 +2308,7 @@
return ret;
}
- for (i = 0; i < num_pages; i++) {
+ for (i = start_i; i < num_pages; i++) {
page = extent_buffer_page(eb, i);
wait_on_page_locked(page);
if (!PageUptodate(page)) {
@@ -2314,7 +2362,7 @@
}
EXPORT_SYMBOL(read_extent_buffer);
-static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
unsigned long min_len, char **token, char **map,
unsigned long *map_start,
unsigned long *map_len, int km)
@@ -2337,6 +2385,10 @@
offset = 0;
*map_start = (i << PAGE_CACHE_SHIFT) - start_offset;
}
+ if (start + min_len >= eb->len) {
+printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len);
+ WARN_ON(1);
+ }
p = extent_buffer_page(eb, i);
WARN_ON(!PageUptodate(p));
@@ -2346,6 +2398,7 @@
*map_len = PAGE_CACHE_SIZE - offset;
return 0;
}
+EXPORT_SYMBOL(map_private_extent_buffer);
int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
unsigned long min_len,
@@ -2360,8 +2413,8 @@
eb->map_token = NULL;
save = 1;
}
- err = __map_extent_buffer(eb, start, min_len, token, map,
- map_start, map_len, km);
+ err = map_private_extent_buffer(eb, start, min_len, token, map,
+ map_start, map_len, km);
if (!err && save) {
eb->map_token = *token;
eb->kaddr = *map;
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index f1dc28d..39d78d3 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -17,8 +17,17 @@
#define EXTENT_DEFRAG (1 << 6)
#define EXTENT_DEFRAG_DONE (1 << 7)
#define EXTENT_BUFFER_FILLED (1 << 8)
+#define EXTENT_CSUM (1 << 9)
#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+/*
+ * page->private values. Every page that is controlled by the extent
+ * map has page->private set to one.
+ */
+
+#define EXTENT_PAGE_PRIVATE 1
+#define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3
+
struct extent_map_ops {
int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
@@ -89,7 +98,7 @@
void extent_map_tree_init(struct extent_map_tree *tree,
struct address_space *mapping, gfp_t mask);
-void extent_map_tree_cleanup(struct extent_map_tree *tree);
+void extent_map_tree_empty_lru(struct extent_map_tree *tree);
struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
u64 start, u64 end);
int add_extent_mapping(struct extent_map_tree *tree,
@@ -143,13 +152,14 @@
struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
u64 start, unsigned long len,
+ struct page *page0,
gfp_t mask);
struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
u64 start, unsigned long len,
gfp_t mask);
void free_extent_buffer(struct extent_buffer *eb);
int read_extent_buffer_pages(struct extent_map_tree *tree,
- struct extent_buffer *eb, int wait);
+ struct extent_buffer *eb, u64 start, int wait);
static inline void extent_buffer_get(struct extent_buffer *eb)
{
@@ -187,5 +197,9 @@
unsigned long min_len, char **token, char **map,
unsigned long *map_start,
unsigned long *map_len, int km);
+int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
+ unsigned long min_len, char **token, char **map,
+ unsigned long *map_start,
+ unsigned long *map_len, int km);
void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
#endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 84f496c..80f84b4 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1713,10 +1713,6 @@
struct extent_map_tree *tree;
int ret;
- if (page->private != 1) {
- WARN_ON(1);
- return try_to_free_buffers(page);
- }
tree = &BTRFS_I(page->mapping->host)->extent_tree;
ret = try_release_extent_mapping(tree, page);
if (ret == 1) {