#include "ioctl.h"
#include "print-tree.h"
#include "volumes.h"
+#include "ordered-data.h"
struct btrfs_iget_args {
u64 ino;
u64 num_bytes;
u64 cur_alloc_size;
u64 blocksize = root->sectorsize;
- u64 orig_start = start;
u64 orig_num_bytes;
struct btrfs_key ins;
- int ret;
+ struct extent_map *em;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ int ret = 0;
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
BUG_ON(!trans);
btrfs_set_trans_block_group(trans, inode);
num_bytes = (end - start + blocksize) & ~(blocksize - 1);
num_bytes = max(blocksize, num_bytes);
- ret = btrfs_drop_extents(trans, root, inode,
- start, start + num_bytes, start, &alloc_hint);
orig_num_bytes = num_bytes;
if (alloc_hint == EXTENT_MAP_INLINE)
goto out;
BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
+ mutex_lock(&BTRFS_I(inode)->extent_mutex);
+ btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
+ mutex_unlock(&BTRFS_I(inode)->extent_mutex);
while(num_bytes > 0) {
cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
- ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
- root->sectorsize,
- root->root_key.objectid,
- trans->transid,
- inode->i_ino, start, 0,
- alloc_hint, (u64)-1, &ins, 1);
+ ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
+ root->sectorsize, 0, 0,
+ (u64)-1, &ins, 1);
if (ret) {
WARN_ON(1);
goto out;
}
+ em = alloc_extent_map(GFP_NOFS);
+ em->start = start;
+ em->len = ins.offset;
+ em->block_start = ins.objectid;
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ mutex_lock(&BTRFS_I(inode)->extent_mutex);
+ set_bit(EXTENT_FLAG_PINNED, &em->flags);
+ while(1) {
+ spin_lock(&em_tree->lock);
+ ret = add_extent_mapping(em_tree, em);
+ spin_unlock(&em_tree->lock);
+ if (ret != -EEXIST) {
+ free_extent_map(em);
+ break;
+ }
+ btrfs_drop_extent_cache(inode, start,
+ start + ins.offset - 1);
+ }
+ mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+
cur_alloc_size = ins.offset;
- ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
- start, ins.objectid, ins.offset,
- ins.offset, 0);
- inode->i_blocks += ins.offset >> 9;
- btrfs_check_file(root, inode);
+ ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
+ ins.offset);
+ BUG_ON(ret);
if (num_bytes < cur_alloc_size) {
printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
cur_alloc_size);
alloc_hint = ins.objectid + ins.offset;
start += cur_alloc_size;
}
- btrfs_drop_extent_cache(inode, orig_start,
- orig_start + orig_num_bytes - 1);
- btrfs_add_ordered_inode(inode);
- btrfs_update_inode(trans, root, inode);
out:
btrfs_end_transaction(trans, root);
return ret;
int mirror_num)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
- struct btrfs_trans_handle *trans;
int ret = 0;
- char *sums = NULL;
-
- ret = btrfs_csum_one_bio(root, bio, &sums);
- BUG_ON(ret);
-
- trans = btrfs_start_transaction(root, 1);
-
- btrfs_set_trans_block_group(trans, inode);
- btrfs_csum_file_blocks(trans, root, inode, bio, sums);
- ret = btrfs_end_transaction(trans, root);
+ ret = btrfs_csum_one_bio(root, inode, bio);
BUG_ON(ret);
- kfree(sums);
-
return btrfs_map_bio(root, rw, bio, mirror_num, 1);
}
struct btrfs_root *root = BTRFS_I(inode)->root;
int ret = 0;
- if (!(rw & (1 << BIO_RW))) {
- ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
- BUG_ON(ret);
- goto mapit;
- }
+ ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+ BUG_ON(ret);
- if (btrfs_test_opt(root, NODATASUM) ||
- btrfs_test_flag(inode, NODATASUM)) {
+ if (!(rw & (1 << BIO_RW))) {
goto mapit;
}
return btrfs_map_bio(root, rw, bio, mirror_num, 0);
}
+static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
+ struct inode *inode, u64 file_offset,
+ struct list_head *list)
+{
+ struct list_head *cur;
+ struct btrfs_ordered_sum *sum;
+
+ btrfs_set_trans_block_group(trans, inode);
+ list_for_each(cur, list) {
+ sum = list_entry(cur, struct btrfs_ordered_sum, list);
+ mutex_lock(&BTRFS_I(inode)->csum_mutex);
+ btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
+ inode, sum);
+ mutex_unlock(&BTRFS_I(inode)->csum_mutex);
+ }
+ return 0;
+}
+
+struct btrfs_writepage_fixup {
+ struct page *page;
+ struct btrfs_work work;
+};
+
+/* see btrfs_writepage_start_hook for details on why this is required */
+void btrfs_writepage_fixup_worker(struct btrfs_work *work)
+{
+ struct btrfs_writepage_fixup *fixup;
+ struct btrfs_ordered_extent *ordered;
+ struct page *page;
+ struct inode *inode;
+ u64 page_start;
+ u64 page_end;
+
+ fixup = container_of(work, struct btrfs_writepage_fixup, work);
+ page = fixup->page;
+again:
+ lock_page(page);
+ if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
+ ClearPageChecked(page);
+ goto out_page;
+ }
+
+ inode = page->mapping->host;
+ page_start = page_offset(page);
+ page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
+
+ lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
+
+ /* already ordered? We're done */
+ if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
+ EXTENT_ORDERED, 0)) {
+ goto out;
+ }
+
+ ordered = btrfs_lookup_ordered_extent(inode, page_start);
+ if (ordered) {
+ unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
+ page_end, GFP_NOFS);
+ unlock_page(page);
+ btrfs_start_ordered_extent(inode, ordered, 1);
+ goto again;
+ }
+
+ set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
+ GFP_NOFS);
+ ClearPageChecked(page);
+out:
+ unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
+out_page:
+ unlock_page(page);
+ page_cache_release(page);
+}
+
+/*
+ * There are a few paths in the higher layers of the kernel that directly
+ * set the page dirty bit without asking the filesystem if it is a
+ * good idea. This causes problems because we want to make sure COW
+ * properly happens and the data=ordered rules are followed.
+ *
+ * In our case any range that doesn't have the EXTENT_ORDERED bit set
+ * hasn't been properly setup for IO. We kick off an async process
+ * to fix it up. The async helper will wait for ordered extents, set
+ * the delalloc bit and make it safe to write the page.
+ */
+int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
+{
+ struct inode *inode = page->mapping->host;
+ struct btrfs_writepage_fixup *fixup;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ int ret;
+
+ ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
+ EXTENT_ORDERED, 0);
+ if (ret)
+ return 0;
+
+ if (PageChecked(page))
+ return -EAGAIN;
+
+ fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
+ if (!fixup)
+ return -EAGAIN;
+printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page));
+ SetPageChecked(page);
+ page_cache_get(page);
+ fixup->work.func = btrfs_writepage_fixup_worker;
+ fixup->page = page;
+ btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
+ return -EAGAIN;
+}
+
+static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_trans_handle *trans;
+ struct btrfs_ordered_extent *ordered_extent;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct extent_map *em;
+ u64 alloc_hint = 0;
+ u64 clear_start;
+ u64 clear_end;
+ struct list_head list;
+ struct btrfs_key ins;
+ int ret;
+
+ ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
+ if (!ret)
+ return 0;
+
+ trans = btrfs_join_transaction(root, 1);
+
+ ordered_extent = btrfs_lookup_ordered_extent(inode, start);
+ BUG_ON(!ordered_extent);
+
+ lock_extent(io_tree, ordered_extent->file_offset,
+ ordered_extent->file_offset + ordered_extent->len - 1,
+ GFP_NOFS);
+
+ INIT_LIST_HEAD(&list);
+
+ ins.objectid = ordered_extent->start;
+ ins.offset = ordered_extent->len;
+ ins.type = BTRFS_EXTENT_ITEM_KEY;
+
+ ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
+ trans->transid, inode->i_ino,
+ ordered_extent->file_offset, &ins);
+ BUG_ON(ret);
+
+ mutex_lock(&BTRFS_I(inode)->extent_mutex);
+
+ ret = btrfs_drop_extents(trans, root, inode,
+ ordered_extent->file_offset,
+ ordered_extent->file_offset +
+ ordered_extent->len,
+ ordered_extent->file_offset, &alloc_hint);
+ BUG_ON(ret);
+ ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
+ ordered_extent->file_offset,
+ ordered_extent->start,
+ ordered_extent->len,
+ ordered_extent->len, 0);
+ BUG_ON(ret);
+
+ spin_lock(&em_tree->lock);
+ clear_start = ordered_extent->file_offset;
+ clear_end = ordered_extent->file_offset + ordered_extent->len;
+ while(clear_start < clear_end) {
+ em = lookup_extent_mapping(em_tree, clear_start,
+ clear_end - clear_start);
+ if (em) {
+ clear_bit(EXTENT_FLAG_PINNED, &em->flags);
+ clear_start = em->start + em->len;
+ free_extent_map(em);
+ } else {
+ break;
+ }
+ }
+ spin_unlock(&em_tree->lock);
+
+ btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
+ ordered_extent->file_offset +
+ ordered_extent->len - 1);
+ mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+
+ inode->i_blocks += ordered_extent->len >> 9;
+ unlock_extent(io_tree, ordered_extent->file_offset,
+ ordered_extent->file_offset + ordered_extent->len - 1,
+ GFP_NOFS);
+ add_pending_csums(trans, inode, ordered_extent->file_offset,
+ &ordered_extent->list);
+
+ btrfs_ordered_update_i_size(inode, ordered_extent);
+ btrfs_remove_ordered_extent(inode, ordered_extent);
+
+ /* once for us */
+ btrfs_put_ordered_extent(ordered_extent);
+ /* once for the tree */
+ btrfs_put_ordered_extent(ordered_extent);
+
+ btrfs_update_inode(trans, root, inode);
+ btrfs_end_transaction(trans, root);
+ return 0;
+}
+
+int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
+ struct extent_state *state, int uptodate)
+{
+ return btrfs_finish_ordered_io(page->mapping->host, start, end);
+}
+
int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
{
int ret = 0;
path = btrfs_alloc_path();
item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
if (IS_ERR(item)) {
+ /*
+ * It is possible there is an ordered extent that has
+ * not yet finished for this range in the file. If so,
+ * that extent will have a csum cached, and it will insert
+ * the sum after all the blocks in the extent are fully
+ * on disk. So, look for an ordered extent and use the
+ * sum if found.
+ */
+ ret = btrfs_find_ordered_sum(inode, start, &csum);
+ if (ret == 0)
+ goto found;
+
ret = PTR_ERR(item);
/* a csum that isn't present is a preallocated region. */
if (ret == -ENOENT || ret == -EFBIG)
ret = 0;
csum = 0;
- printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
+ printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
+ start);
goto out;
}
read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
BTRFS_CRC32_SIZE);
+found:
set_state_private(io_tree, start, csum);
out:
if (path)
inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
inode->i_uid = btrfs_inode_uid(leaf, inode_item);
inode->i_gid = btrfs_inode_gid(leaf, inode_item);
- inode->i_size = btrfs_inode_size(leaf, inode_item);
+ btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
tspec = btrfs_inode_atime(inode_item);
inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
{
btrfs_set_inode_uid(leaf, item, inode->i_uid);
btrfs_set_inode_gid(leaf, item, inode->i_gid);
- btrfs_set_inode_size(leaf, item, inode->i_size);
+ btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
btrfs_set_inode_mode(leaf, item, inode->i_mode);
btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
BTRFS_I(inode)->block_group->key.objectid);
}
-int btrfs_update_inode(struct btrfs_trans_handle *trans,
+int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct inode *inode)
{
err:
btrfs_free_path(path);
if (!ret) {
- dir->i_size -= name_len * 2;
+ btrfs_i_size_write(dir, dir->i_size - name_len * 2);
dir->i_mtime = dir->i_ctime = CURRENT_TIME;
btrfs_update_inode(trans, root, dir);
#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
{
struct btrfs_root *root;
struct btrfs_trans_handle *trans;
- struct inode *inode = dentry->d_inode;
int ret;
unsigned long nr = 0;
ret = btrfs_unlink_trans(trans, root, dir, dentry);
nr = trans->blocks_used;
- if (inode->i_nlink == 0) {
- /* if the inode isn't linked anywhere,
- * we don't need to worry about
- * data=ordered
- */
- btrfs_del_ordered_inode(inode);
- }
-
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
fail:
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return ret;
}
/* now the directory is empty */
err = btrfs_unlink_trans(trans, root, dir, dentry);
if (!err) {
- inode->i_size = 0;
+ btrfs_i_size_write(inode, 0);
}
nr = trans->blocks_used;
- ret = btrfs_end_transaction(trans, root);
+ ret = btrfs_end_transaction_throttle(trans, root);
fail:
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
if (ret && !err)
err = ret;
return ret;
}
-static int btrfs_cow_one_page(struct inode *inode, struct page *page,
- size_t zero_start)
-{
- char *kaddr;
- struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
- u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
- u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
- int ret = 0;
-
- WARN_ON(!PageLocked(page));
- set_page_extent_mapped(page);
-
- lock_extent(io_tree, page_start, page_end, GFP_NOFS);
- set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
- page_end, GFP_NOFS);
-
- if (zero_start != PAGE_CACHE_SIZE) {
- kaddr = kmap(page);
- memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
- flush_dcache_page(page);
- kunmap(page);
- }
- set_page_dirty(page);
- unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
-
- return ret;
-}
-
/*
* taken from block_truncate_page, but does cow as it zeros out
* any bytes left in the last page in the file.
{
struct inode *inode = mapping->host;
struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ struct btrfs_ordered_extent *ordered;
+ char *kaddr;
u32 blocksize = root->sectorsize;
pgoff_t index = from >> PAGE_CACHE_SHIFT;
unsigned offset = from & (PAGE_CACHE_SIZE-1);
struct page *page;
int ret = 0;
u64 page_start;
+ u64 page_end;
if ((offset & (blocksize - 1)) == 0)
goto out;
page = grab_cache_page(mapping, index);
if (!page)
goto out;
+
+ page_start = page_offset(page);
+ page_end = page_start + PAGE_CACHE_SIZE - 1;
+
if (!PageUptodate(page)) {
ret = btrfs_readpage(NULL, page);
lock_page(page);
goto out;
}
}
-
- page_start = (u64)page->index << PAGE_CACHE_SHIFT;
wait_on_page_writeback(page);
- ret = btrfs_cow_one_page(inode, page, offset);
+
+ lock_extent(io_tree, page_start, page_end, GFP_NOFS);
+ set_page_extent_mapped(page);
+
+ ordered = btrfs_lookup_ordered_extent(inode, page_start);
+ if (ordered) {
+ unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
+ unlock_page(page);
+ page_cache_release(page);
+ btrfs_start_ordered_extent(inode, ordered, 1);
+ btrfs_put_ordered_extent(ordered);
+ goto again;
+ }
+
+ set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
+ page_end, GFP_NOFS);
+ ret = 0;
+ if (offset != PAGE_CACHE_SIZE) {
+ kaddr = kmap(page);
+ memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
+ flush_dcache_page(page);
+ kunmap(page);
+ }
+ ClearPageChecked(page);
+ set_page_dirty(page);
+ unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
unlock_page(page);
page_cache_release(page);
btrfs_truncate_page(inode->i_mapping, inode->i_size);
- lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
hole_size = block_end - hole_start;
+ btrfs_wait_ordered_range(inode, hole_start, hole_size);
+ lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
+ mutex_lock(&BTRFS_I(inode)->extent_mutex);
err = btrfs_drop_extents(trans, root, inode,
hole_start, block_end, hole_start,
&alloc_hint);
(u64)-1);
btrfs_check_file(root, inode);
}
+ mutex_unlock(&BTRFS_I(inode)->extent_mutex);
btrfs_end_transaction(trans, root);
unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
if (err)
if (is_bad_inode(inode)) {
goto no_delete;
}
+ btrfs_wait_ordered_range(inode, 0, (u64)-1);
- inode->i_size = 0;
+ btrfs_i_size_write(inode, 0);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
btrfs_end_transaction(trans, root);
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return;
no_delete_lock:
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
no_delete:
clear_inode(inode);
}
inode->i_ino = args->ino;
BTRFS_I(inode)->root = args->root;
BTRFS_I(inode)->delalloc_bytes = 0;
+ BTRFS_I(inode)->disk_i_size = 0;
extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
extent_io_tree_init(&BTRFS_I(inode)->io_tree,
inode->i_mapping, GFP_NOFS);
extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
inode->i_mapping, GFP_NOFS);
- atomic_set(&BTRFS_I(inode)->ordered_writeback, 0);
+ btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
+ mutex_init(&BTRFS_I(inode)->csum_mutex);
+ mutex_init(&BTRFS_I(inode)->extent_mutex);
return 0;
}
int ret = 0;
if (wait) {
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
ret = btrfs_commit_transaction(trans, root);
}
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_trans_handle *trans;
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
btrfs_update_inode(trans, root, inode);
btrfs_end_transaction(trans, root);
inode->i_mapping, GFP_NOFS);
extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
inode->i_mapping, GFP_NOFS);
- atomic_set(&BTRFS_I(inode)->ordered_writeback, 0);
+ btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
+ mutex_init(&BTRFS_I(inode)->csum_mutex);
+ mutex_init(&BTRFS_I(inode)->extent_mutex);
BTRFS_I(inode)->delalloc_bytes = 0;
+ BTRFS_I(inode)->disk_i_size = 0;
BTRFS_I(inode)->root = root;
if (mode & S_IFDIR)
dentry->d_parent->d_inode->i_ino);
}
parent_inode = dentry->d_parent->d_inode;
- parent_inode->i_size += dentry->d_name.len * 2;
+ btrfs_i_size_write(parent_inode, parent_inode->i_size +
+ dentry->d_name.len * 2);
parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
ret = btrfs_update_inode(trans, root,
dentry->d_parent->d_inode);
btrfs_update_inode_block_group(trans, dir);
out_unlock:
nr = trans->blocks_used;
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
fail:
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
}
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return err;
}
inode->i_mapping, GFP_NOFS);
extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
inode->i_mapping, GFP_NOFS);
+ mutex_init(&BTRFS_I(inode)->csum_mutex);
+ mutex_init(&BTRFS_I(inode)->extent_mutex);
BTRFS_I(inode)->delalloc_bytes = 0;
- atomic_set(&BTRFS_I(inode)->ordered_writeback, 0);
+ BTRFS_I(inode)->disk_i_size = 0;
BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
+ btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
}
dir->i_sb->s_dirt = 1;
btrfs_update_inode_block_group(trans, inode);
btrfs_update_inode_block_group(trans, dir);
out_unlock:
nr = trans->blocks_used;
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
fail:
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
}
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return err;
}
drop_inode = 1;
nr = trans->blocks_used;
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
fail:
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
}
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return err;
}
inode->i_fop = &btrfs_dir_file_operations;
btrfs_set_trans_block_group(trans, inode);
- inode->i_size = 0;
+ btrfs_i_size_write(inode, 0);
err = btrfs_update_inode(trans, root, inode);
if (err)
goto out_fail;
out_fail:
nr = trans->blocks_used;
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
out_unlock:
if (drop_on_err)
iput(inode);
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return err;
}
static int merge_extent_mapping(struct extent_map_tree *em_tree,
struct extent_map *existing,
- struct extent_map *em)
+ struct extent_map *em,
+ u64 map_start, u64 map_len)
{
u64 start_diff;
- u64 new_end;
- int ret = 0;
- int real_blocks = existing->block_start < EXTENT_MAP_LAST_BYTE;
-
- if (real_blocks && em->block_start >= EXTENT_MAP_LAST_BYTE)
- goto invalid;
-
- if (!real_blocks && em->block_start != existing->block_start)
- goto invalid;
-
- new_end = max(existing->start + existing->len, em->start + em->len);
-
- if (existing->start >= em->start) {
- if (em->start + em->len < existing->start)
- goto invalid;
-
- start_diff = existing->start - em->start;
- if (real_blocks && em->block_start + start_diff !=
- existing->block_start)
- goto invalid;
-
- em->len = new_end - em->start;
-
- remove_extent_mapping(em_tree, existing);
- /* free for the tree */
- free_extent_map(existing);
- ret = add_extent_mapping(em_tree, em);
- } else if (em->start > existing->start) {
-
- if (existing->start + existing->len < em->start)
- goto invalid;
-
- start_diff = em->start - existing->start;
- if (real_blocks && existing->block_start + start_diff !=
- em->block_start)
- goto invalid;
-
- remove_extent_mapping(em_tree, existing);
- em->block_start = existing->block_start;
- em->start = existing->start;
- em->len = new_end - existing->start;
- free_extent_map(existing);
-
- ret = add_extent_mapping(em_tree, em);
- } else {
- goto invalid;
- }
- return ret;
-
-invalid:
- printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n",
- existing->start, existing->len, existing->block_start,
- em->start, em->len, em->block_start);
- return -EIO;
+ BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
+ start_diff = map_start - em->start;
+ em->start = map_start;
+ em->len = map_len;
+ if (em->block_start < EXTENT_MAP_LAST_BYTE)
+ em->block_start += start_diff;
+ return add_extent_mapping(em_tree, em);
}
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
err = -ENOMEM;
goto out;
}
-
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
em->start = EXTENT_MAP_HOLE;
em->len = (u64)-1;
- em->bdev = root->fs_info->fs_devices->latest_bdev;
ret = btrfs_lookup_file_extent(trans, root, path,
objectid, start, trans != NULL);
if (ret < 0) {
free_extent_map(em);
em = NULL;
btrfs_release_path(root, path);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
goto again;
}
write_extent_buffer(leaf, map + pg_offset, ptr,
*/
if (ret == -EEXIST) {
struct extent_map *existing;
+
+ ret = 0;
+
existing = lookup_extent_mapping(em_tree, start, len);
if (existing && (existing->start > start ||
existing->start + existing->len <= start)) {
em->len);
if (existing) {
err = merge_extent_mapping(em_tree, existing,
- em);
+ em, start,
+ root->sectorsize);
free_extent_map(existing);
if (err) {
free_extent_map(em);
} else {
free_extent_map(em);
em = existing;
+ err = 0;
}
}
spin_unlock(&em_tree->lock);
btrfs_free_path(path);
if (trans) {
ret = btrfs_end_transaction(trans, root);
- if (!err)
+ if (!err) {
err = ret;
+ }
}
if (err) {
free_extent_map(em);
return extent_readpages(tree, mapping, pages, nr_pages,
btrfs_get_extent);
}
-
-static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
+static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
struct extent_io_tree *tree;
struct extent_map_tree *map;
map = &BTRFS_I(page->mapping->host)->extent_tree;
ret = try_release_extent_mapping(map, tree, page, gfp_flags);
if (ret == 1) {
- invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
ClearPagePrivate(page);
set_page_private(page, 0);
page_cache_release(page);
return ret;
}
+static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
+{
+ return __btrfs_releasepage(page, gfp_flags);
+}
+
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
{
struct extent_io_tree *tree;
+ struct btrfs_ordered_extent *ordered;
+ u64 page_start = page_offset(page);
+ u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
+ wait_on_page_writeback(page);
tree = &BTRFS_I(page->mapping->host)->io_tree;
- extent_invalidatepage(tree, page, offset);
- btrfs_releasepage(page, GFP_NOFS);
+ if (offset) {
+ btrfs_releasepage(page, GFP_NOFS);
+ return;
+ }
+
+ lock_extent(tree, page_start, page_end, GFP_NOFS);
+ ordered = btrfs_lookup_ordered_extent(page->mapping->host,
+ page_offset(page));
+ if (ordered) {
+ /*
+ * IO on this page will never be started, so we need
+ * to account for any ordered extents now
+ */
+ clear_extent_bit(tree, page_start, page_end,
+ EXTENT_DIRTY | EXTENT_DELALLOC |
+ EXTENT_LOCKED, 1, 0, GFP_NOFS);
+ btrfs_finish_ordered_io(page->mapping->host,
+ page_start, page_end);
+ btrfs_put_ordered_extent(ordered);
+ lock_extent(tree, page_start, page_end, GFP_NOFS);
+ }
+ clear_extent_bit(tree, page_start, page_end,
+ EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
+ EXTENT_ORDERED,
+ 1, 1, GFP_NOFS);
+ __btrfs_releasepage(page, GFP_NOFS);
+
+ ClearPageChecked(page);
if (PagePrivate(page)) {
- invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
ClearPagePrivate(page);
set_page_private(page, 0);
page_cache_release(page);
{
struct inode *inode = fdentry(vma->vm_file)->d_inode;
struct btrfs_root *root = BTRFS_I(inode)->root;
- unsigned long end;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ struct btrfs_ordered_extent *ordered;
+ char *kaddr;
+ unsigned long zero_start;
loff_t size;
int ret;
u64 page_start;
+ u64 page_end;
ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
if (ret)
goto out;
ret = -EINVAL;
-
+again:
lock_page(page);
- wait_on_page_writeback(page);
size = i_size_read(inode);
- page_start = (u64)page->index << PAGE_CACHE_SHIFT;
+ page_start = page_offset(page);
+ page_end = page_start + PAGE_CACHE_SIZE - 1;
if ((page->mapping != inode->i_mapping) ||
- (page_start > size)) {
+ (page_start >= size)) {
/* page got truncated out from underneath us */
goto out_unlock;
}
+ wait_on_page_writeback(page);
+
+ lock_extent(io_tree, page_start, page_end, GFP_NOFS);
+ set_page_extent_mapped(page);
+
+ /*
+ * we can't set the delalloc bits if there are pending ordered
+ * extents. Drop our locks and wait for them to finish
+ */
+ ordered = btrfs_lookup_ordered_extent(inode, page_start);
+ if (ordered) {
+ unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
+ unlock_page(page);
+ btrfs_start_ordered_extent(inode, ordered, 1);
+ btrfs_put_ordered_extent(ordered);
+ goto again;
+ }
+
+ set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
+ page_end, GFP_NOFS);
+ ret = 0;
/* page is wholly or partially inside EOF */
if (page_start + PAGE_CACHE_SIZE > size)
- end = size & ~PAGE_CACHE_MASK;
+ zero_start = size & ~PAGE_CACHE_MASK;
else
- end = PAGE_CACHE_SIZE;
+ zero_start = PAGE_CACHE_SIZE;
- ret = btrfs_cow_one_page(inode, page, end);
+ if (zero_start != PAGE_CACHE_SIZE) {
+ kaddr = kmap(page);
+ memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
+ flush_dcache_page(page);
+ kunmap(page);
+ }
+ ClearPageChecked(page);
+ set_page_dirty(page);
+ unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
out_unlock:
unlock_page(page);
int ret;
struct btrfs_trans_handle *trans;
unsigned long nr;
+ u64 mask = root->sectorsize - 1;
if (!S_ISREG(inode->i_mode))
return;
return;
btrfs_truncate_page(inode->i_mapping, inode->i_size);
+ btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
+ btrfs_i_size_write(inode, inode->i_size);
/* FIXME, add redo link to tree so we don't leak on crash */
ret = btrfs_truncate_in_trans(trans, root, inode,
btrfs_update_inode(trans, root, inode);
nr = trans->blocks_used;
- ret = btrfs_end_transaction(trans, root);
+ ret = btrfs_end_transaction_throttle(trans, root);
BUG_ON(ret);
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
}
/*
ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
new_dirid);
inode->i_nlink = 1;
- inode->i_size = 0;
+ btrfs_i_size_write(inode, 0);
return btrfs_update_inode(trans, new_root, inode);
}
if (!ei)
return NULL;
ei->last_trans = 0;
- ei->ordered_trans = 0;
+ btrfs_ordered_inode_tree_init(&ei->ordered_tree);
return &ei->vfs_inode;
}
void btrfs_destroy_inode(struct inode *inode)
{
+ struct btrfs_ordered_extent *ordered;
WARN_ON(!list_empty(&inode->i_dentry));
WARN_ON(inode->i_data.nrpages);
+ while(1) {
+ ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
+ if (!ordered)
+ break;
+ else {
+ printk("found ordered extent %Lu %Lu\n",
+ ordered->file_offset, ordered->len);
+ btrfs_remove_ordered_extent(inode, ordered);
+ btrfs_put_ordered_extent(ordered);
+ btrfs_put_ordered_extent(ordered);
+ }
+ }
btrfs_drop_extent_cache(inode, 0, (u64)-1);
kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}
inode->i_mapping, GFP_NOFS);
extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
inode->i_mapping, GFP_NOFS);
+ mutex_init(&BTRFS_I(inode)->csum_mutex);
+ mutex_init(&BTRFS_I(inode)->extent_mutex);
BTRFS_I(inode)->delalloc_bytes = 0;
- atomic_set(&BTRFS_I(inode)->ordered_writeback, 0);
+ BTRFS_I(inode)->disk_i_size = 0;
BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
+ btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
}
dir->i_sb->s_dirt = 1;
btrfs_update_inode_block_group(trans, inode);
inode->i_op = &btrfs_symlink_inode_operations;
inode->i_mapping->a_ops = &btrfs_symlink_aops;
inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
- inode->i_size = name_len - 1;
+ btrfs_i_size_write(inode, name_len - 1);
err = btrfs_update_inode(trans, root, inode);
if (err)
drop_inode = 1;
out_unlock:
nr = trans->blocks_used;
- btrfs_end_transaction(trans, root);
+ btrfs_end_transaction_throttle(trans, root);
out_fail:
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
}
btrfs_btree_balance_dirty(root, nr);
- btrfs_throttle(root);
return err;
}
+static int btrfs_set_page_dirty(struct page *page)
+{
+ return __set_page_dirty_nobuffers(page);
+}
+
static int btrfs_permission(struct inode *inode, int mask,
struct nameidata *nd)
{
.merge_bio_hook = btrfs_merge_bio_hook,
.readpage_io_hook = btrfs_readpage_io_hook,
.readpage_end_io_hook = btrfs_readpage_end_io_hook,
+ .writepage_end_io_hook = btrfs_writepage_end_io_hook,
+ .writepage_start_hook = btrfs_writepage_start_hook,
.readpage_io_failed_hook = btrfs_io_failed_hook,
.set_bit_hook = btrfs_set_bit_hook,
.clear_bit_hook = btrfs_clear_bit_hook,
.direct_IO = btrfs_direct_IO,
.invalidatepage = btrfs_invalidatepage,
.releasepage = btrfs_releasepage,
- .set_page_dirty = __set_page_dirty_nobuffers,
+ .set_page_dirty = btrfs_set_page_dirty,
};
static struct address_space_operations btrfs_symlink_aops = {
git.samba.org / sfrench / cifs-2.6.git / blobdiff