extent_io hooks for data IO
This code handles most of the data IO processing. It provides the
hooks used by extent_io.c to do readpages/writepages and other basic
page operations.
It also does delayed allocation processing and metadata updates when
IO is complete for data=ordered.
Compression is started done when delayed allocation is filled. The
pages are sent from extent_io.c locked with the extent_state locked as well.
This code will try to compress the pages via the async work queues and
start IO either on compressed extents or as plain data.
Inline extents are inserted as delayed allocation time as well. Doing it
this late allows us to avoid inefficient inline extent manipulation at
file write time (for example appending to the file one byte at a time).
Signed-off-by: Chris Mason <chris.mason@xxxxxxxxxx>
--- a/fs/btrfs/inode.c 2009-01-08 09:43:29.596585385 -0500
+++ b/fs/btrfs/inode.c 2009-01-08 09:46:16.272224440 -0500
@@ -85,6 +85,7 @@
};
static void btrfs_truncate(struct inode *inode);
+static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
static noinline int cow_file_range(struct inode *inode,
struct page *locked_page,
u64 start, u64 end, int *page_started,
@@ -118,6 +119,1323 @@
return ret;
}
+/*
+ * this does all the hard work for inserting an inline extent into
+ * the btree. The caller should have done a btrfs_drop_extents so that
+ * no overlapping inline items exist in the btree
+ */
+static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, struct inode *inode,
+ u64 start, size_t size, size_t compressed_size,
+ struct page **compressed_pages)
+{
+ struct btrfs_key key;
+ struct btrfs_path *path;
+ struct extent_buffer *leaf;
+ struct page *page = NULL;
+ char *kaddr;
+ unsigned long ptr;
+ struct btrfs_file_extent_item *ei;
+ int err = 0;
+ int ret;
+ size_t cur_size = size;
+ size_t datasize;
+ unsigned long offset;
+ int use_compress = 0;
+
+ if (compressed_size && compressed_pages) {
+ use_compress = 1;
+ cur_size = compressed_size;
+ }
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ btrfs_set_trans_block_group(trans, inode);
+
+ key.objectid = inode->i_ino;
+ key.offset = start;
+ btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
+ datasize = btrfs_file_extent_calc_inline_size(cur_size);
+
+ inode_add_bytes(inode, size);
+ ret = btrfs_insert_empty_item(trans, root, path, &key,
+ datasize);
+ BUG_ON(ret);
+ if (ret) {
+ err = ret;
+ goto fail;
+ }
+ leaf = path->nodes[0];
+ ei = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_file_extent_item);
+ btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+ btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
+ btrfs_set_file_extent_encryption(leaf, ei, 0);
+ btrfs_set_file_extent_other_encoding(leaf, ei, 0);
+ btrfs_set_file_extent_ram_bytes(leaf, ei, size);
+ ptr = btrfs_file_extent_inline_start(ei);
+
+ if (use_compress) {
+ struct page *cpage;
+ int i = 0;
+ while (compressed_size > 0) {
+ cpage = compressed_pages[i];
+ cur_size = min_t(unsigned long, compressed_size,
+ PAGE_CACHE_SIZE);
+
+ kaddr = kmap(cpage);
+ write_extent_buffer(leaf, kaddr, ptr, cur_size);
+ kunmap(cpage);
+
+ i++;
+ ptr += cur_size;
+ compressed_size -= cur_size;
+ }
+ btrfs_set_file_extent_compression(leaf, ei,
+ BTRFS_COMPRESS_ZLIB);
+ } else {
+ page = find_get_page(inode->i_mapping,
+ start >> PAGE_CACHE_SHIFT);
+ btrfs_set_file_extent_compression(leaf, ei, 0);
+ kaddr = kmap_atomic(page, KM_USER0);
+ offset = start & (PAGE_CACHE_SIZE - 1);
+ write_extent_buffer(leaf, kaddr + offset, ptr, size);
+ kunmap_atomic(kaddr, KM_USER0);
+ page_cache_release(page);
+ }
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_free_path(path);
+
+ BTRFS_I(inode)->disk_i_size = inode->i_size;
+ btrfs_update_inode(trans, root, inode);
+ return 0;
+fail:
+ btrfs_free_path(path);
+ return err;
+}
+
+
+/*
+ * conditionally insert an inline extent into the file. This
+ * does the checks required to make sure the data is small enough
+ * to fit as an inline extent.
+ */
+static int cow_file_range_inline(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct inode *inode, u64 start, u64 end,
+ size_t compressed_size,
+ struct page **compressed_pages)
+{
+ u64 isize = i_size_read(inode);
+ u64 actual_end = min(end + 1, isize);
+ u64 inline_len = actual_end - start;
+ u64 aligned_end = (end + root->sectorsize - 1) &
+ ~((u64)root->sectorsize - 1);
+ u64 hint_byte;
+ u64 data_len = inline_len;
+ int ret;
+
+ if (compressed_size)
+ data_len = compressed_size;
+
+ if (start > 0 ||
+ actual_end >= PAGE_CACHE_SIZE ||
+ data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
+ (!compressed_size &&
+ (actual_end & (root->sectorsize - 1)) == 0) ||
+ end + 1 < isize ||
+ data_len > root->fs_info->max_inline) {
+ return 1;
+ }
+
+ ret = btrfs_drop_extents(trans, root, inode, start,
+ aligned_end, start, &hint_byte);
+ BUG_ON(ret);
+
+ if (isize > actual_end)
+ inline_len = min_t(u64, isize, actual_end);
+ ret = insert_inline_extent(trans, root, inode, start,
+ inline_len, compressed_size,
+ compressed_pages);
+ BUG_ON(ret);
+ btrfs_drop_extent_cache(inode, start, aligned_end, 0);
+ return 0;
+}
+
+struct async_extent {
+ u64 start;
+ u64 ram_size;
+ u64 compressed_size;
+ struct page **pages;
+ unsigned long nr_pages;
+ struct list_head list;
+};
+
+struct async_cow {
+ struct inode *inode;
+ struct btrfs_root *root;
+ struct page *locked_page;
+ u64 start;
+ u64 end;
+ struct list_head extents;
+ struct btrfs_work work;
+};
+
+static noinline int add_async_extent(struct async_cow *cow,
+ u64 start, u64 ram_size,
+ u64 compressed_size,
+ struct page **pages,
+ unsigned long nr_pages)
+{
+ struct async_extent *async_extent;
+
+ async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
+ async_extent->start = start;
+ async_extent->ram_size = ram_size;
+ async_extent->compressed_size = compressed_size;
+ async_extent->pages = pages;
+ async_extent->nr_pages = nr_pages;
+ list_add_tail(&async_extent->list, &cow->extents);
+ return 0;
+}
+
+/*
+ * we create compressed extents in two phases. The first
+ * phase compresses a range of pages that have already been
+ * locked (both pages and state bits are locked).
+ *
+ * This is done inside an ordered work queue, and the compression
+ * is spread across many cpus. The actual IO submission is step
+ * two, and the ordered work queue takes care of making sure that
+ * happens in the same order things were put onto the queue by
+ * writepages and friends.
+ *
+ * If this code finds it can't get good compression, it puts an
+ * entry onto the work queue to write the uncompressed bytes. This
+ * makes sure that both compressed inodes and uncompressed inodes
+ * are written in the same order that pdflush sent them down.
+ */
+static noinline int compress_file_range(struct inode *inode,
+ struct page *locked_page,
+ u64 start, u64 end,
+ struct async_cow *async_cow,
+ int *num_added)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_trans_handle *trans;
+ u64 num_bytes;
+ u64 orig_start;
+ u64 disk_num_bytes;
+ u64 blocksize = root->sectorsize;
+ u64 actual_end;
+ u64 isize = i_size_read(inode);
+ int ret = 0;
+ struct page **pages = NULL;
+ unsigned long nr_pages;
+ unsigned long nr_pages_ret = 0;
+ unsigned long total_compressed = 0;
+ unsigned long total_in = 0;
+ unsigned long max_compressed = 128 * 1024;
+ unsigned long max_uncompressed = 128 * 1024;
+ int i;
+ int will_compress;
+
+ orig_start = start;
+
+ actual_end = min_t(u64, isize, end + 1);
+again:
+ will_compress = 0;
+ nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1;
+ nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE);
+
+ total_compressed = actual_end - start;
+
+ /* we want to make sure that amount of ram required to uncompress
+ * an extent is reasonable, so we limit the total size in ram
+ * of a compressed extent to 128k. This is a crucial number
+ * because it also controls how easily we can spread reads across
+ * cpus for decompression.
+ *
+ * We also want to make sure the amount of IO required to do
+ * a random read is reasonably small, so we limit the size of
+ * a compressed extent to 128k.
+ */
+ total_compressed = min(total_compressed, max_uncompressed);
+ num_bytes = (end - start + blocksize) & ~(blocksize - 1);
+ num_bytes = max(blocksize, num_bytes);
+ disk_num_bytes = num_bytes;
+ total_in = 0;
+ ret = 0;
+
+ /*
+ * we do compression for mount -o compress and when the
+ * inode has not been flagged as nocompress. This flag can
+ * change at any time if we discover bad compression ratios.
+ */
+ if (!btrfs_test_flag(inode, NOCOMPRESS) &&
+ btrfs_test_opt(root, COMPRESS)) {
+ WARN_ON(pages);
+ pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
+
+ ret = btrfs_zlib_compress_pages(inode->i_mapping, start,
+ total_compressed, pages,
+ nr_pages, &nr_pages_ret,
+ &total_in,
+ &total_compressed,
+ max_compressed);
+
+ if (!ret) {
+ unsigned long offset = total_compressed &
+ (PAGE_CACHE_SIZE - 1);
+ struct page *page = pages[nr_pages_ret - 1];
+ char *kaddr;
+
+ /* zero the tail end of the last page, we might be
+ * sending it down to disk
+ */
+ if (offset) {
+ kaddr = kmap_atomic(page, KM_USER0);
+ memset(kaddr + offset, 0,
+ PAGE_CACHE_SIZE - offset);
+ kunmap_atomic(kaddr, KM_USER0);
+ }
+ will_compress = 1;
+ }
+ }
+ if (start == 0) {
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(!trans);
+ btrfs_set_trans_block_group(trans, inode);
+
+ /* lets try to make an inline extent */
+ if (ret || total_in < (actual_end - start)) {
+ /* we didn't compress the entire range, try
+ * to make an uncompressed inline extent.
+ */
+ ret = cow_file_range_inline(trans, root, inode,
+ start, end, 0, NULL);
+ } else {
+ /* try making a compressed inline extent */
+ ret = cow_file_range_inline(trans, root, inode,
+ start, end,
+ total_compressed, pages);
+ }
+ btrfs_end_transaction(trans, root);
+ if (ret == 0) {
+ /*
+ * inline extent creation worked, we don't need
+ * to create any more async work items. Unlock
+ * and free up our temp pages.
+ */
+ extent_clear_unlock_delalloc(inode,
+ &BTRFS_I(inode)->io_tree,
+ start, end, NULL, 1, 0,
+ 0, 1, 1, 1);
+ ret = 0;
+ goto free_pages_out;
+ }
+ }
+
+ if (will_compress) {
+ /*
+ * we aren't doing an inline extent round the compressed size
+ * up to a block size boundary so the allocator does sane
+ * things
+ */
+ total_compressed = (total_compressed + blocksize - 1) &
+ ~(blocksize - 1);
+
+ /*
+ * one last check to make sure the compression is really a
+ * win, compare the page count read with the blocks on disk
+ */
+ total_in = (total_in + PAGE_CACHE_SIZE - 1) &
+ ~(PAGE_CACHE_SIZE - 1);
+ if (total_compressed >= total_in) {
+ will_compress = 0;
+ } else {
+ disk_num_bytes = total_compressed;
+ num_bytes = total_in;
+ }
+ }
+ if (!will_compress && pages) {
+ /*
+ * the compression code ran but failed to make things smaller,
+ * free any pages it allocated and our page pointer array
+ */
+ for (i = 0; i < nr_pages_ret; i++) {
+ WARN_ON(pages[i]->mapping);
+ page_cache_release(pages[i]);
+ }
+ kfree(pages);
+ pages = NULL;
+ total_compressed = 0;
+ nr_pages_ret = 0;
+
+ /* flag the file so we don't compress in the future */
+ btrfs_set_flag(inode, NOCOMPRESS);
+ }
+ if (will_compress) {
+ *num_added += 1;
+
+ /* the async work queues will take care of doing actual
+ * allocation on disk for these compressed pages,
+ * and will submit them to the elevator.
+ */
+ add_async_extent(async_cow, start, num_bytes,
+ total_compressed, pages, nr_pages_ret);
+
+ if (start + num_bytes < end && start + num_bytes < actual_end) {
+ start += num_bytes;
+ pages = NULL;
+ cond_resched();
+ goto again;
+ }
+ } else {
+ /*
+ * No compression, but we still need to write the pages in
+ * the file we've been given so far. redirty the locked
+ * page if it corresponds to our extent and set things up
+ * for the async work queue to run cow_file_range to do
+ * the normal delalloc dance
+ */
+ if (page_offset(locked_page) >= start &&
+ page_offset(locked_page) <= end) {
+ __set_page_dirty_nobuffers(locked_page);
+ /* unlocked later on in the async handlers */
+ }
+ add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0);
+ *num_added += 1;
+ }
+
+out:
+ return 0;
+
+free_pages_out:
+ for (i = 0; i < nr_pages_ret; i++) {
+ WARN_ON(pages[i]->mapping);
+ page_cache_release(pages[i]);
+ }
+ kfree(pages);
+
+ goto out;
+}
+
+/*
+ * phase two of compressed writeback. This is the ordered portion
+ * of the code, which only gets called in the order the work was
+ * queued. We walk all the async extents created by compress_file_range
+ * and send them down to the disk.
+ */
+static noinline int submit_compressed_extents(struct inode *inode,
+ struct async_cow *async_cow)
+{
+ struct async_extent *async_extent;
+ u64 alloc_hint = 0;
+ struct btrfs_trans_handle *trans;
+ struct btrfs_key ins;
+ struct extent_map *em;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct extent_io_tree *io_tree;
+ int ret;
+
+ if (list_empty(&async_cow->extents))
+ return 0;
+
+ trans = btrfs_join_transaction(root, 1);
+
+ while (!list_empty(&async_cow->extents)) {
+ async_extent = list_entry(async_cow->extents.next,
+ struct async_extent, list);
+ list_del(&async_extent->list);
+
+ io_tree = &BTRFS_I(inode)->io_tree;
+
+ /* did the compression code fall back to uncompressed IO? */
+ if (!async_extent->pages) {
+ int page_started = 0;
+ unsigned long nr_written = 0;
+
+ lock_extent(io_tree, async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1, GFP_NOFS);
+
+ /* allocate blocks */
+ cow_file_range(inode, async_cow->locked_page,
+ async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1,
+ &page_started, &nr_written, 0);
+
+ /*
+ * if page_started, cow_file_range inserted an
+ * inline extent and took care of all the unlocking
+ * and IO for us. Otherwise, we need to submit
+ * all those pages down to the drive.
+ */
+ if (!page_started)
+ extent_write_locked_range(io_tree,
+ inode, async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1,
+ btrfs_get_extent,
+ WB_SYNC_ALL);
+ kfree(async_extent);
+ cond_resched();
+ continue;
+ }
+
+ lock_extent(io_tree, async_extent->start,
+ async_extent->start + async_extent->ram_size - 1,
+ GFP_NOFS);
+ /*
+ * here we're doing allocation and writeback of the
+ * compressed pages
+ */
+ btrfs_drop_extent_cache(inode, async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1, 0);
+
+ ret = btrfs_reserve_extent(trans, root,
+ async_extent->compressed_size,
+ async_extent->compressed_size,
+ 0, alloc_hint,
+ (u64)-1, &ins, 1);
+ BUG_ON(ret);
+ em = alloc_extent_map(GFP_NOFS);
+ em->start = async_extent->start;
+ em->len = async_extent->ram_size;
+ em->orig_start = em->start;
+
+ em->block_start = ins.objectid;
+ em->block_len = ins.offset;
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ set_bit(EXTENT_FLAG_PINNED, &em->flags);
+ set_bit(EXTENT_FLAG_COMPRESSED, &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, async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1, 0);
+ }
+
+ ret = btrfs_add_ordered_extent(inode, async_extent->start,
+ ins.objectid,
+ async_extent->ram_size,
+ ins.offset,
+ BTRFS_ORDERED_COMPRESSED);
+ BUG_ON(ret);
+
+ btrfs_end_transaction(trans, root);
+
+ /*
+ * clear dirty, set writeback and unlock the pages.
+ */
+ extent_clear_unlock_delalloc(inode,
+ &BTRFS_I(inode)->io_tree,
+ async_extent->start,
+ async_extent->start +
+ async_extent->ram_size - 1,
+ NULL, 1, 1, 0, 1, 1, 0);
+
+ ret = btrfs_submit_compressed_write(inode,
+ async_extent->start,
+ async_extent->ram_size,
+ ins.objectid,
+ ins.offset, async_extent->pages,
+ async_extent->nr_pages);
+
+ BUG_ON(ret);
+ trans = btrfs_join_transaction(root, 1);
+ alloc_hint = ins.objectid + ins.offset;
+ kfree(async_extent);
+ cond_resched();
+ }
+
+ btrfs_end_transaction(trans, root);
+ return 0;
+}
+
+/*
+ * when extent_io.c finds a delayed allocation range in the file,
+ * the call backs end up in this code. The basic idea is to
+ * allocate extents on disk for the range, and create ordered data structs
+ * in ram to track those extents.
+ *
+ * locked_page is the page that writepage had locked already. We use
+ * it to make sure we don't do extra locks or unlocks.
+ *
+ * *page_started is set to one if we unlock locked_page and do everything
+ * required to start IO on it. It may be clean and already done with
+ * IO when we return.
+ */
+static noinline int cow_file_range(struct inode *inode,
+ struct page *locked_page,
+ u64 start, u64 end, int *page_started,
+ unsigned long *nr_written,
+ int unlock)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_trans_handle *trans;
+ u64 alloc_hint = 0;
+ u64 num_bytes;
+ unsigned long ram_size;
+ u64 disk_num_bytes;
+ u64 cur_alloc_size;
+ u64 blocksize = root->sectorsize;
+ u64 actual_end;
+ u64 isize = i_size_read(inode);
+ struct btrfs_key ins;
+ struct extent_map *em;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ int ret = 0;
+
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(!trans);
+ btrfs_set_trans_block_group(trans, inode);
+
+ actual_end = min_t(u64, isize, end + 1);
+
+ num_bytes = (end - start + blocksize) & ~(blocksize - 1);
+ num_bytes = max(blocksize, num_bytes);
+ disk_num_bytes = num_bytes;
+ ret = 0;
+
+ if (start == 0) {
+ /* lets try to make an inline extent */
+ ret = cow_file_range_inline(trans, root, inode,
+ start, end, 0, NULL);
+ if (ret == 0) {
+ extent_clear_unlock_delalloc(inode,
+ &BTRFS_I(inode)->io_tree,
+ start, end, NULL, 1, 1,
+ 1, 1, 1, 1);
+ *nr_written = *nr_written +
+ (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
+ *page_started = 1;
+ ret = 0;
+ goto out;
+ }
+ }
+
+ BUG_ON(disk_num_bytes >
+ btrfs_super_total_bytes(&root->fs_info->super_copy));
+
+ btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
+
+ while (disk_num_bytes > 0) {
+ cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
+ ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
+ root->sectorsize, 0, alloc_hint,
+ (u64)-1, &ins, 1);
+ BUG_ON(ret);
+
+ em = alloc_extent_map(GFP_NOFS);
+ em->start = start;
+ em->orig_start = em->start;
+
+ ram_size = ins.offset;
+ em->len = ins.offset;
+
+ em->block_start = ins.objectid;
+ em->block_len = ins.offset;
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ 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 + ram_size - 1, 0);
+ }
+
+ cur_alloc_size = ins.offset;
+ ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
+ ram_size, cur_alloc_size, 0);
+ BUG_ON(ret);
+
+ if (root->root_key.objectid ==
+ BTRFS_DATA_RELOC_TREE_OBJECTID) {
+ ret = btrfs_reloc_clone_csums(inode, start,
+ cur_alloc_size);
+ BUG_ON(ret);
+ }
+
+ if (disk_num_bytes < cur_alloc_size)
+ break;
+
+ /* we're not doing compressed IO, don't unlock the first
+ * page (which the caller expects to stay locked), don't
+ * clear any dirty bits and don't set any writeback bits
+ */
+ extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
+ start, start + ram_size - 1,
+ locked_page, unlock, 1,
+ 1, 0, 0, 0);
+ disk_num_bytes -= cur_alloc_size;
+ num_bytes -= cur_alloc_size;
+ alloc_hint = ins.objectid + ins.offset;
+ start += cur_alloc_size;
+ }
+out:
+ ret = 0;
+ btrfs_end_transaction(trans, root);
+
+ return ret;
+}
+
+/*
+ * work queue call back to started compression on a file and pages
+ */
+static noinline void async_cow_start(struct btrfs_work *work)
+{
+ struct async_cow *async_cow;
+ int num_added = 0;
+ async_cow = container_of(work, struct async_cow, work);
+
+ compress_file_range(async_cow->inode, async_cow->locked_page,
+ async_cow->start, async_cow->end, async_cow,
+ &num_added);
+ if (num_added == 0)
+ async_cow->inode = NULL;
+}
+
+/*
+ * work queue call back to submit previously compressed pages
+ */
+static noinline void async_cow_submit(struct btrfs_work *work)
+{
+ struct async_cow *async_cow;
+ struct btrfs_root *root;
+ unsigned long nr_pages;
+
+ async_cow = container_of(work, struct async_cow, work);
+
+ root = async_cow->root;
+ nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
+ PAGE_CACHE_SHIFT;
+
+ atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages);
+
+ if (atomic_read(&root->fs_info->async_delalloc_pages) <
+ 5 * 1042 * 1024 &&
+ waitqueue_active(&root->fs_info->async_submit_wait))
+ wake_up(&root->fs_info->async_submit_wait);
+
+ if (async_cow->inode)
+ submit_compressed_extents(async_cow->inode, async_cow);
+}
+
+static noinline void async_cow_free(struct btrfs_work *work)
+{
+ struct async_cow *async_cow;
+ async_cow = container_of(work, struct async_cow, work);
+ kfree(async_cow);
+}
+
+static int cow_file_range_async(struct inode *inode, struct page *locked_page,
+ u64 start, u64 end, int *page_started,
+ unsigned long *nr_written)
+{
+ struct async_cow *async_cow;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ unsigned long nr_pages;
+ u64 cur_end;
+ int limit = 10 * 1024 * 1042;
+
+ if (!btrfs_test_opt(root, COMPRESS)) {
+ return cow_file_range(inode, locked_page, start, end,
+ page_started, nr_written, 1);
+ }
+
+ clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED |
+ EXTENT_DELALLOC, 1, 0, GFP_NOFS);
+ while (start < end) {
+ async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
+ async_cow->inode = inode;
+ async_cow->root = root;
+ async_cow->locked_page = locked_page;
+ async_cow->start = start;
+
+ if (btrfs_test_flag(inode, NOCOMPRESS))
+ cur_end = end;
+ else
+ cur_end = min(end, start + 512 * 1024 - 1);
+
+ async_cow->end = cur_end;
+ INIT_LIST_HEAD(&async_cow->extents);
+
+ async_cow->work.func = async_cow_start;
+ async_cow->work.ordered_func = async_cow_submit;
+ async_cow->work.ordered_free = async_cow_free;
+ async_cow->work.flags = 0;
+
+ nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >>
+ PAGE_CACHE_SHIFT;
+ atomic_add(nr_pages, &root->fs_info->async_delalloc_pages);
+
+ btrfs_queue_worker(&root->fs_info->delalloc_workers,
+ &async_cow->work);
+
+ if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) {
+ wait_event(root->fs_info->async_submit_wait,
+ (atomic_read(&root->fs_info->async_delalloc_pages) <
+ limit));
+ }
+
+ while (atomic_read(&root->fs_info->async_submit_draining) &&
+ atomic_read(&root->fs_info->async_delalloc_pages)) {
+ wait_event(root->fs_info->async_submit_wait,
+ (atomic_read(&root->fs_info->async_delalloc_pages) ==
+ 0));
+ }
+
+ *nr_written += nr_pages;
+ start = cur_end + 1;
+ }
+ *page_started = 1;
+ return 0;
+}
+
+static noinline int csum_exist_in_range(struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes)
+{
+ int ret;
+ struct btrfs_ordered_sum *sums;
+ LIST_HEAD(list);
+
+ ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
+ bytenr + num_bytes - 1, &list);
+ if (ret == 0 && list_empty(&list))
+ return 0;
+
+ while (!list_empty(&list)) {
+ sums = list_entry(list.next, struct btrfs_ordered_sum, list);
+ list_del(&sums->list);
+ kfree(sums);
+ }
+ return 1;
+}
+
+/*
+ * when nowcow writeback call back. This checks for snapshots or COW copies
+ * of the extents that exist in the file, and COWs the file as required.
+ *
+ * If no cow copies or snapshots exist, we write directly to the existing
+ * blocks on disk
+ */
+static int run_delalloc_nocow(struct inode *inode, struct page *locked_page,
+ u64 start, u64 end, int *page_started, int force,
+ unsigned long *nr_written)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_trans_handle *trans;
+ struct extent_buffer *leaf;
+ struct btrfs_path *path;
+ struct btrfs_file_extent_item *fi;
+ struct btrfs_key found_key;
+ u64 cow_start;
+ u64 cur_offset;
+ u64 extent_end;
+ u64 disk_bytenr;
+ u64 num_bytes;
+ int extent_type;
+ int ret;
+ int type;
+ int nocow;
+ int check_prev = 1;
+
+ path = btrfs_alloc_path();
+ BUG_ON(!path);
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(!trans);
+
+ cow_start = (u64)-1;
+ cur_offset = start;
+ while (1) {
+ ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
+ cur_offset, 0);
+ BUG_ON(ret < 0);
+ if (ret > 0 && path->slots[0] > 0 && check_prev) {
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &found_key,
+ path->slots[0] - 1);
+ if (found_key.objectid == inode->i_ino &&
+ found_key.type == BTRFS_EXTENT_DATA_KEY)
+ path->slots[0]--;
+ }
+ check_prev = 0;
+next_slot:
+ leaf = path->nodes[0];
+ if (path->slots[0] >= btrfs_header_nritems(leaf)) {
+ ret = btrfs_next_leaf(root, path);
+ if (ret < 0)
+ BUG_ON(1);
+ if (ret > 0)
+ break;
+ leaf = path->nodes[0];
+ }
+
+ nocow = 0;
+ disk_bytenr = 0;
+ num_bytes = 0;
+ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+
+ if (found_key.objectid > inode->i_ino ||
+ found_key.type > BTRFS_EXTENT_DATA_KEY ||
+ found_key.offset > end)
+ break;
+
+ if (found_key.offset > cur_offset) {
+ extent_end = found_key.offset;
+ goto out_check;
+ }
+
+ fi = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_file_extent_item);
+ extent_type = btrfs_file_extent_type(leaf, fi);
+
+ if (extent_type == BTRFS_FILE_EXTENT_REG ||
+ extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
+ disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+ extent_end = found_key.offset +
+ btrfs_file_extent_num_bytes(leaf, fi);
+ if (extent_end <= start) {
+ path->slots[0]++;
+ goto next_slot;
+ }
+ if (disk_bytenr == 0)
+ goto out_check;
+ if (btrfs_file_extent_compression(leaf, fi) ||
+ btrfs_file_extent_encryption(leaf, fi) ||
+ btrfs_file_extent_other_encoding(leaf, fi))
+ goto out_check;
+ if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
+ goto out_check;
+ if (btrfs_extent_readonly(root, disk_bytenr))
+ goto out_check;
+ if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
+ disk_bytenr))
+ goto out_check;
+ disk_bytenr += btrfs_file_extent_offset(leaf, fi);
+ disk_bytenr += cur_offset - found_key.offset;
+ num_bytes = min(end + 1, extent_end) - cur_offset;
+ /*
+ * force cow if csum exists in the range.
+ * this ensure that csum for a given extent are
+ * either valid or do not exist.
+ */
+ if (csum_exist_in_range(root, disk_bytenr, num_bytes))
+ goto out_check;
+ nocow = 1;
+ } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+ extent_end = found_key.offset +
+ btrfs_file_extent_inline_len(leaf, fi);
+ extent_end = ALIGN(extent_end, root->sectorsize);
+ } else {
+ BUG_ON(1);
+ }
+out_check:
+ if (extent_end <= start) {
+ path->slots[0]++;
+ goto next_slot;
+ }
+ if (!nocow) {
+ if (cow_start == (u64)-1)
+ cow_start = cur_offset;
+ cur_offset = extent_end;
+ if (cur_offset > end)
+ break;
+ path->slots[0]++;
+ goto next_slot;
+ }
+
+ btrfs_release_path(root, path);
+ if (cow_start != (u64)-1) {
+ ret = cow_file_range(inode, locked_page, cow_start,
+ found_key.offset - 1, page_started,
+ nr_written, 1);
+ BUG_ON(ret);
+ cow_start = (u64)-1;
+ }
+
+ if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
+ struct extent_map *em;
+ struct extent_map_tree *em_tree;
+ em_tree = &BTRFS_I(inode)->extent_tree;
+ em = alloc_extent_map(GFP_NOFS);
+ em->start = cur_offset;
+ em->orig_start = em->start;
+ em->len = num_bytes;
+ em->block_len = num_bytes;
+ em->block_start = disk_bytenr;
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ 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, em->start,
+ em->start + em->len - 1, 0);
+ }
+ type = BTRFS_ORDERED_PREALLOC;
+ } else {
+ type = BTRFS_ORDERED_NOCOW;
+ }
+
+ ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
+ num_bytes, num_bytes, type);
+ BUG_ON(ret);
+
+ extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
+ cur_offset, cur_offset + num_bytes - 1,
+ locked_page, 1, 1, 1, 0, 0, 0);
+ cur_offset = extent_end;
+ if (cur_offset > end)
+ break;
+ }
+ btrfs_release_path(root, path);
+
+ if (cur_offset <= end && cow_start == (u64)-1)
+ cow_start = cur_offset;
+ if (cow_start != (u64)-1) {
+ ret = cow_file_range(inode, locked_page, cow_start, end,
+ page_started, nr_written, 1);
+ BUG_ON(ret);
+ }
+
+ ret = btrfs_end_transaction(trans, root);
+ BUG_ON(ret);
+ btrfs_free_path(path);
+ return 0;
+}
+
+/*
+ * extent_io.c call back to do delayed allocation processing
+ */
+static int run_delalloc_range(struct inode *inode, struct page *locked_page,
+ u64 start, u64 end, int *page_started,
+ unsigned long *nr_written)
+{
+ int ret;
+
+ if (btrfs_test_flag(inode, NODATACOW))
+ ret = run_delalloc_nocow(inode, locked_page, start, end,
+ page_started, 1, nr_written);
+ else if (btrfs_test_flag(inode, PREALLOC))
+ ret = run_delalloc_nocow(inode, locked_page, start, end,
+ page_started, 0, nr_written);
+ else
+ ret = cow_file_range_async(inode, locked_page, start, end,
+ page_started, nr_written);
+
+ return ret;
+}
+
+/*
+ * extent_io.c set_bit_hook, used to track delayed allocation
+ * bytes in this file, and to maintain the list of inodes that
+ * have pending delalloc work to be done.
+ */
+static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
+ unsigned long old, unsigned long bits)
+{
+ /*
+ * set_bit and clear bit hooks normally require _irqsave/restore
+ * but in this case, we are only testeing for the DELALLOC
+ * bit, which is only set or cleared with irqs on
+ */
+ if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ spin_lock(&root->fs_info->delalloc_lock);
+ BTRFS_I(inode)->delalloc_bytes += end - start + 1;
+ root->fs_info->delalloc_bytes += end - start + 1;
+ if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
+ list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
+ &root->fs_info->delalloc_inodes);
+ }
+ spin_unlock(&root->fs_info->delalloc_lock);
+ }
+ return 0;
+}
+
+/*
+ * extent_io.c clear_bit_hook, see set_bit_hook for why
+ */
+static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
+ unsigned long old, unsigned long bits)
+{
+ /*
+ * set_bit and clear bit hooks normally require _irqsave/restore
+ * but in this case, we are only testeing for the DELALLOC
+ * bit, which is only set or cleared with irqs on
+ */
+ if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+
+ spin_lock(&root->fs_info->delalloc_lock);
+ if (end - start + 1 > root->fs_info->delalloc_bytes) {
+ printk(KERN_INFO "btrfs warning: delalloc account "
+ "%llu %llu\n",
+ (unsigned long long)end - start + 1,
+ (unsigned long long)
+ root->fs_info->delalloc_bytes);
+ root->fs_info->delalloc_bytes = 0;
+ BTRFS_I(inode)->delalloc_bytes = 0;
+ } else {
+ root->fs_info->delalloc_bytes -= end - start + 1;
+ BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
+ }
+ if (BTRFS_I(inode)->delalloc_bytes == 0 &&
+ !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
+ list_del_init(&BTRFS_I(inode)->delalloc_inodes);
+ }
+ spin_unlock(&root->fs_info->delalloc_lock);
+ }
+ return 0;
+}
+
+/*
+ * extent_io.c merge_bio_hook, this must check the chunk tree to make sure
+ * we don't create bios that span stripes or chunks
+ */
+int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
+ size_t size, struct bio *bio,
+ unsigned long bio_flags)
+{
+ struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+ struct btrfs_mapping_tree *map_tree;
+ u64 logical = (u64)bio->bi_sector << 9;
+ u64 length = 0;
+ u64 map_length;
+ int ret;
+
+ if (bio_flags & EXTENT_BIO_COMPRESSED)
+ return 0;
+
+ length = bio->bi_size;
+ map_tree = &root->fs_info->mapping_tree;
+ map_length = length;
+ ret = btrfs_map_block(map_tree, READ, logical,
+ &map_length, NULL, 0);
+
+ if (map_length < length + size)
+ return 1;
+ return 0;
+}
+
+/*
+ * in order to insert checksums into the metadata in large chunks,
+ * we wait until bio submission time. All the pages in the bio are
+ * checksummed and sums are attached onto the ordered extent record.
+ *
+ * At IO completion time the cums attached on the ordered extent record
+ * are inserted into the btree
+ */
+static int __btrfs_submit_bio_start(struct inode *inode, int rw,
+ struct bio *bio, int mirror_num,
+ unsigned long bio_flags)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ int ret = 0;
+
+ ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
+ BUG_ON(ret);
+ return 0;
+}
+
+/*
+ * in order to insert checksums into the metadata in large chunks,
+ * we wait until bio submission time. All the pages in the bio are
+ * checksummed and sums are attached onto the ordered extent record.
+ *
+ * At IO completion time the cums attached on the ordered extent record
+ * are inserted into the btree
+ */
+static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
+ int mirror_num, unsigned long bio_flags)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ return btrfs_map_bio(root, rw, bio, mirror_num, 1);
+}
+
+/*
+ * extent_io.c submission hook. This does the right thing for csum calculation
+ * on write, or reading the csums from the tree before a read
+ */
+static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+ int mirror_num, unsigned long bio_flags)
+{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ int ret = 0;
+ int skip_sum;
+
+ skip_sum = btrfs_test_flag(inode, NODATASUM);
+
+ ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+ BUG_ON(ret);
+
+ if (!(rw & (1 << BIO_RW))) {
+ if (bio_flags & EXTENT_BIO_COMPRESSED) {
+ return btrfs_submit_compressed_read(inode, bio,
+ mirror_num, bio_flags);
+ } else if (!skip_sum)
+ btrfs_lookup_bio_sums(root, inode, bio, NULL);
+ goto mapit;
+ } else if (!skip_sum) {
+ /* csum items have already been cloned */
+ if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
+ goto mapit;
+ /* we're doing a write, do the async checksumming */
+ return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+ inode, rw, bio, mirror_num,
+ bio_flags, __btrfs_submit_bio_start,
+ __btrfs_submit_bio_done);
+ }
+
+mapit:
+ return btrfs_map_bio(root, rw, bio, mirror_num, 0);
+}
+
+/*
+ * given a list of ordered sums record them in the inode. This happens
+ * at IO completion time based on sums calculated at bio submission time.
+ */
+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);
+ btrfs_csum_file_blocks(trans,
+ BTRFS_I(inode)->root->fs_info->csum_root, sum);
+ }
+ return 0;
+}
+
+int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end)
+{
+ if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
+ WARN_ON(1);
+ return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
+ GFP_NOFS);
+}
+
+/* see btrfs_writepage_start_hook for details on why this is required */
+struct btrfs_writepage_fixup {
+ struct page *page;
+ struct btrfs_work work;
+};
+
+static 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;
+ }
+
+ btrfs_set_extent_delalloc(inode, page_start, page_end);
+ 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 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.
+ */
+static 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;
+
+ 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 insert_reserved_file_extent(struct btrfs_trans_handle *trans,
struct inode *inode, u64 file_pos,
u64 disk_bytenr, u64 disk_num_bytes,
@@ -175,6 +1493,291 @@
return 0;
}
+/* as ordered data IO finishes, this gets called so we can finish
+ * an ordered extent if the range of bytes in the file it covers are
+ * fully written.
+ */
+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;
+ int compressed = 0;
+ 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);
+ if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags))
+ goto nocow;
+
+ lock_extent(io_tree, ordered_extent->file_offset,
+ ordered_extent->file_offset + ordered_extent->len - 1,
+ GFP_NOFS);
+
+ if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
+ compressed = 1;
+ if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
+ BUG_ON(compressed);
+ ret = btrfs_mark_extent_written(trans, root, inode,
+ ordered_extent->file_offset,
+ ordered_extent->file_offset +
+ ordered_extent->len);
+ BUG_ON(ret);
+ } else {
+ ret = insert_reserved_file_extent(trans, inode,
+ ordered_extent->file_offset,
+ ordered_extent->start,
+ ordered_extent->disk_len,
+ ordered_extent->len,
+ ordered_extent->len,
+ compressed, 0, 0,
+ BTRFS_FILE_EXTENT_REG);
+ BUG_ON(ret);
+ }
+ unlock_extent(io_tree, ordered_extent->file_offset,
+ ordered_extent->file_offset + ordered_extent->len - 1,
+ GFP_NOFS);
+nocow:
+ add_pending_csums(trans, inode, ordered_extent->file_offset,
+ &ordered_extent->list);
+
+ mutex_lock(&BTRFS_I(inode)->extent_mutex);
+ btrfs_ordered_update_i_size(inode, ordered_extent);
+ btrfs_update_inode(trans, root, inode);
+ btrfs_remove_ordered_extent(inode, ordered_extent);
+ mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+
+ /* once for us */
+ btrfs_put_ordered_extent(ordered_extent);
+ /* once for the tree */
+ btrfs_put_ordered_extent(ordered_extent);
+
+ btrfs_end_transaction(trans, root);
+ return 0;
+}
+
+static 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);
+}
+
+/*
+ * When IO fails, either with EIO or csum verification fails, we
+ * try other mirrors that might have a good copy of the data. This
+ * io_failure_record is used to record state as we go through all the
+ * mirrors. If another mirror has good data, the page is set up to date
+ * and things continue. If a good mirror can't be found, the original
+ * bio end_io callback is called to indicate things have failed.
+ */
+struct io_failure_record {
+ struct page *page;
+ u64 start;
+ u64 len;
+ u64 logical;
+ unsigned long bio_flags;
+ int last_mirror;
+};
+
+static int btrfs_io_failed_hook(struct bio *failed_bio,
+ struct page *page, u64 start, u64 end,
+ struct extent_state *state)
+{
+ struct io_failure_record *failrec = NULL;
+ u64 private;
+ struct extent_map *em;
+ struct inode *inode = page->mapping->host;
+ struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct bio *bio;
+ int num_copies;
+ int ret;
+ int rw;
+ u64 logical;
+
+ ret = get_state_private(failure_tree, start, &private);
+ if (ret) {
+ failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
+ if (!failrec)
+ return -ENOMEM;
+ failrec->start = start;
+ failrec->len = end - start + 1;
+ failrec->last_mirror = 0;
+ failrec->bio_flags = 0;
+
+ spin_lock(&em_tree->lock);
+ em = lookup_extent_mapping(em_tree, start, failrec->len);
+ if (em->start > start || em->start + em->len < start) {
+ free_extent_map(em);
+ em = NULL;
+ }
+ spin_unlock(&em_tree->lock);
+
+ if (!em || IS_ERR(em)) {
+ kfree(failrec);
+ return -EIO;
+ }
+ logical = start - em->start;
+ logical = em->block_start + logical;
+ if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
+ logical = em->block_start;
+ failrec->bio_flags = EXTENT_BIO_COMPRESSED;
+ }
+ failrec->logical = logical;
+ free_extent_map(em);
+ set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
+ EXTENT_DIRTY, GFP_NOFS);
+ set_state_private(failure_tree, start,
+ (u64)(unsigned long)failrec);
+ } else {
+ failrec = (struct io_failure_record *)(unsigned long)private;
+ }
+ num_copies = btrfs_num_copies(
+ &BTRFS_I(inode)->root->fs_info->mapping_tree,
+ failrec->logical, failrec->len);
+ failrec->last_mirror++;
+ if (!state) {
+ spin_lock(&BTRFS_I(inode)->io_tree.lock);
+ state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
+ failrec->start,
+ EXTENT_LOCKED);
+ if (state && state->start != failrec->start)
+ state = NULL;
+ spin_unlock(&BTRFS_I(inode)->io_tree.lock);
+ }
+ if (!state || failrec->last_mirror > num_copies) {
+ set_state_private(failure_tree, failrec->start, 0);
+ clear_extent_bits(failure_tree, failrec->start,
+ failrec->start + failrec->len - 1,
+ EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
+ kfree(failrec);
+ return -EIO;
+ }
+ bio = bio_alloc(GFP_NOFS, 1);
+ bio->bi_private = state;
+ bio->bi_end_io = failed_bio->bi_end_io;
+ bio->bi_sector = failrec->logical >> 9;
+ bio->bi_bdev = failed_bio->bi_bdev;
+ bio->bi_size = 0;
+
+ bio_add_page(bio, page, failrec->len, start - page_offset(page));
+ if (failed_bio->bi_rw & (1 << BIO_RW))
+ rw = WRITE;
+ else
+ rw = READ;
+
+ BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
+ failrec->last_mirror,
+ failrec->bio_flags);
+ return 0;
+}
+
+/*
+ * each time an IO finishes, we do a fast check in the IO failure tree
+ * to see if we need to process or clean up an io_failure_record
+ */
+static int btrfs_clean_io_failures(struct inode *inode, u64 start)
+{
+ u64 private;
+ u64 private_failure;
+ struct io_failure_record *failure;
+ int ret;
+
+ private = 0;
+ if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
+ (u64)-1, 1, EXTENT_DIRTY)) {
+ ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
+ start, &private_failure);
+ if (ret == 0) {
+ failure = (struct io_failure_record *)(unsigned long)
+ private_failure;
+ set_state_private(&BTRFS_I(inode)->io_failure_tree,
+ failure->start, 0);
+ clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
+ failure->start,
+ failure->start + failure->len - 1,
+ EXTENT_DIRTY | EXTENT_LOCKED,
+ GFP_NOFS);
+ kfree(failure);
+ }
+ }
+ return 0;
+}
+
+/*
+ * when reads are done, we need to check csums to verify the data is correct
+ * if there's a match, we allow the bio to finish. If not, we go through
+ * the io_failure_record routines to find good copies
+ */
+static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
+ struct extent_state *state)
+{
+ size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
+ struct inode *inode = page->mapping->host;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ char *kaddr;
+ u64 private = ~(u32)0;
+ int ret;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ u32 csum = ~(u32)0;
+
+ if (PageChecked(page)) {
+ ClearPageChecked(page);
+ goto good;
+ }
+ if (btrfs_test_flag(inode, NODATASUM))
+ return 0;
+
+ if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
+ test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1)) {
+ clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
+ GFP_NOFS);
+ return 0;
+ }
+
+ if (state && state->start == start) {
+ private = state->private;
+ ret = 0;
+ } else {
+ ret = get_state_private(io_tree, start, &private);
+ }
+ kaddr = kmap_atomic(page, KM_USER0);
+ if (ret)
+ goto zeroit;
+
+ csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
+ btrfs_csum_final(csum, (char *)&csum);
+ if (csum != private)
+ goto zeroit;
+
+ kunmap_atomic(kaddr, KM_USER0);
+good:
+ /* if the io failure tree for this inode is non-empty,
+ * check to see if we've recovered from a failed IO
+ */
+ btrfs_clean_io_failures(inode, start);
+ return 0;
+
+zeroit:
+ printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
+ "private %llu\n", page->mapping->host->i_ino,
+ (unsigned long long)start, csum,
+ (unsigned long long)private);
+ memset(kaddr + offset, 1, end - start + 1);
+ flush_dcache_page(page);
+ kunmap_atomic(kaddr, KM_USER0);
+ if (private == 0)
+ return 0;
+ return -EIO;
+}
+
/*
* This creates an orphan entry for the given inode in case something goes
* wrong in the middle of an unlink/truncate.
@@ -2035,6 +3638,349 @@
return err;
}
+/* helper for btfs_get_extent. Given an existing extent in the tree,
+ * and an extent that you want to insert, deal with overlap and insert
+ * the new extent into the tree.
+ */
+static int merge_extent_mapping(struct extent_map_tree *em_tree,
+ struct extent_map *existing,
+ struct extent_map *em,
+ u64 map_start, u64 map_len)
+{
+ u64 start_diff;
+
+ 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 &&
+ !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
+ em->block_start += start_diff;
+ em->block_len -= start_diff;
+ }
+ return add_extent_mapping(em_tree, em);
+}
+
+static noinline int uncompress_inline(struct btrfs_path *path,
+ struct inode *inode, struct page *page,
+ size_t pg_offset, u64 extent_offset,
+ struct btrfs_file_extent_item *item)
+{
+ int ret;
+ struct extent_buffer *leaf = path->nodes[0];
+ char *tmp;
+ size_t max_size;
+ unsigned long inline_size;
+ unsigned long ptr;
+
+ WARN_ON(pg_offset != 0);
+ max_size = btrfs_file_extent_ram_bytes(leaf, item);
+ inline_size = btrfs_file_extent_inline_item_len(leaf,
+ btrfs_item_nr(leaf, path->slots[0]));
+ tmp = kmalloc(inline_size, GFP_NOFS);
+ ptr = btrfs_file_extent_inline_start(item);
+
+ read_extent_buffer(leaf, tmp, ptr, inline_size);
+
+ max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
+ ret = btrfs_zlib_decompress(tmp, page, extent_offset,
+ inline_size, max_size);
+ if (ret) {
+ char *kaddr = kmap_atomic(page, KM_USER0);
+ unsigned long copy_size = min_t(u64,
+ PAGE_CACHE_SIZE - pg_offset,
+ max_size - extent_offset);
+ memset(kaddr + pg_offset, 0, copy_size);
+ kunmap_atomic(kaddr, KM_USER0);
+ }
+ kfree(tmp);
+ return 0;
+}
+
+/*
+ * a bit scary, this does extent mapping from logical file offset to the disk.
+ * the ugly parts come from merging extents from the disk with the in-ram
+ * representation. This gets more complex because of the data=ordered code,
+ * where the in-ram extents might be locked pending data=ordered completion.
+ *
+ * This also copies inline extents directly into the page.
+ */
+
+struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
+ size_t pg_offset, u64 start, u64 len,
+ int create)
+{
+ int ret;
+ int err = 0;
+ u64 bytenr;
+ u64 extent_start = 0;
+ u64 extent_end = 0;
+ u64 objectid = inode->i_ino;
+ u32 found_type;
+ struct btrfs_path *path = NULL;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_file_extent_item *item;
+ struct extent_buffer *leaf;
+ struct btrfs_key found_key;
+ struct extent_map *em = NULL;
+ struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ struct btrfs_trans_handle *trans = NULL;
+ int compressed;
+
+again:
+ spin_lock(&em_tree->lock);
+ em = lookup_extent_mapping(em_tree, start, len);
+ if (em)
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ spin_unlock(&em_tree->lock);
+
+ if (em) {
+ if (em->start > start || em->start + em->len <= start)
+ free_extent_map(em);
+ else if (em->block_start == EXTENT_MAP_INLINE && page)
+ free_extent_map(em);
+ else
+ goto out;
+ }
+ em = alloc_extent_map(GFP_NOFS);
+ if (!em) {
+ err = -ENOMEM;
+ goto out;
+ }
+ em->bdev = root->fs_info->fs_devices->latest_bdev;
+ em->start = EXTENT_MAP_HOLE;
+ em->orig_start = EXTENT_MAP_HOLE;
+ em->len = (u64)-1;
+ em->block_len = (u64)-1;
+
+ if (!path) {
+ path = btrfs_alloc_path();
+ BUG_ON(!path);
+ }
+
+ ret = btrfs_lookup_file_extent(trans, root, path,
+ objectid, start, trans != NULL);
+ if (ret < 0) {
+ err = ret;
+ goto out;
+ }
+
+ if (ret != 0) {
+ if (path->slots[0] == 0)
+ goto not_found;
+ path->slots[0]--;
+ }
+
+ leaf = path->nodes[0];
+ item = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_file_extent_item);
+ /* are we inside the extent that was found? */
+ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+ found_type = btrfs_key_type(&found_key);
+ if (found_key.objectid != objectid ||
+ found_type != BTRFS_EXTENT_DATA_KEY) {
+ goto not_found;
+ }
+
+ found_type = btrfs_file_extent_type(leaf, item);
+ extent_start = found_key.offset;
+ compressed = btrfs_file_extent_compression(leaf, item);
+ if (found_type == BTRFS_FILE_EXTENT_REG ||
+ found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+ extent_end = extent_start +
+ btrfs_file_extent_num_bytes(leaf, item);
+ } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+ size_t size;
+ size = btrfs_file_extent_inline_len(leaf, item);
+ extent_end = (extent_start + size + root->sectorsize - 1) &
+ ~((u64)root->sectorsize - 1);
+ }
+
+ if (start >= extent_end) {
+ path->slots[0]++;
+ if (path->slots[0] >= btrfs_header_nritems(leaf)) {
+ ret = btrfs_next_leaf(root, path);
+ if (ret < 0) {
+ err = ret;
+ goto out;
+ }
+ if (ret > 0)
+ goto not_found;
+ leaf = path->nodes[0];
+ }
+ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+ if (found_key.objectid != objectid ||
+ found_key.type != BTRFS_EXTENT_DATA_KEY)
+ goto not_found;
+ if (start + len <= found_key.offset)
+ goto not_found;
+ em->start = start;
+ em->len = found_key.offset - start;
+ goto not_found_em;
+ }
+
+ if (found_type == BTRFS_FILE_EXTENT_REG ||
+ found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+ em->start = extent_start;
+ em->len = extent_end - extent_start;
+ em->orig_start = extent_start -
+ btrfs_file_extent_offset(leaf, item);
+ bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
+ if (bytenr == 0) {
+ em->block_start = EXTENT_MAP_HOLE;
+ goto insert;
+ }
+ if (compressed) {
+ set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+ em->block_start = bytenr;
+ em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
+ item);
+ } else {
+ bytenr += btrfs_file_extent_offset(leaf, item);
+ em->block_start = bytenr;
+ em->block_len = em->len;
+ if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
+ set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
+ }
+ goto insert;
+ } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+ unsigned long ptr;
+ char *map;
+ size_t size;
+ size_t extent_offset;
+ size_t copy_size;
+
+ em->block_start = EXTENT_MAP_INLINE;
+ if (!page || create) {
+ em->start = extent_start;
+ em->len = extent_end - extent_start;
+ goto out;
+ }
+
+ size = btrfs_file_extent_inline_len(leaf, item);
+ extent_offset = page_offset(page) + pg_offset - extent_start;
+ copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
+ size - extent_offset);
+ em->start = extent_start + extent_offset;
+ em->len = (copy_size + root->sectorsize - 1) &
+ ~((u64)root->sectorsize - 1);
+ em->orig_start = EXTENT_MAP_INLINE;
+ if (compressed)
+ set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+ ptr = btrfs_file_extent_inline_start(item) + extent_offset;
+ if (create == 0 && !PageUptodate(page)) {
+ if (btrfs_file_extent_compression(leaf, item) ==
+ BTRFS_COMPRESS_ZLIB) {
+ ret = uncompress_inline(path, inode, page,
+ pg_offset,
+ extent_offset, item);
+ BUG_ON(ret);
+ } else {
+ map = kmap(page);
+ read_extent_buffer(leaf, map + pg_offset, ptr,
+ copy_size);
+ kunmap(page);
+ }
+ flush_dcache_page(page);
+ } else if (create && PageUptodate(page)) {
+ if (!trans) {
+ kunmap(page);
+ free_extent_map(em);
+ em = NULL;
+ btrfs_release_path(root, path);
+ trans = btrfs_join_transaction(root, 1);
+ goto again;
+ }
+ map = kmap(page);
+ write_extent_buffer(leaf, map + pg_offset, ptr,
+ copy_size);
+ kunmap(page);
+ btrfs_mark_buffer_dirty(leaf);
+ }
+ set_extent_uptodate(io_tree, em->start,
+ extent_map_end(em) - 1, GFP_NOFS);
+ goto insert;
+ } else {
+ printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
+ WARN_ON(1);
+ }
+not_found:
+ em->start = start;
+ em->len = len;
+not_found_em:
+ em->block_start = EXTENT_MAP_HOLE;
+ set_bit(EXTENT_FLAG_VACANCY, &em->flags);
+insert:
+ btrfs_release_path(root, path);
+ if (em->start > start || extent_map_end(em) <= start) {
+ printk(KERN_ERR "Btrfs: bad extent! em: [%llu %llu] passed "
+ "[%llu %llu]\n", (unsigned long long)em->start,
+ (unsigned long long)em->len,
+ (unsigned long long)start,
+ (unsigned long long)len);
+ err = -EIO;
+ goto out;
+ }
+
+ err = 0;
+ spin_lock(&em_tree->lock);
+ ret = add_extent_mapping(em_tree, em);
+ /* it is possible that someone inserted the extent into the tree
+ * while we had the lock dropped. It is also possible that
+ * an overlapping map exists in the tree
+ */
+ 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)) {
+ free_extent_map(existing);
+ existing = NULL;
+ }
+ if (!existing) {
+ existing = lookup_extent_mapping(em_tree, em->start,
+ em->len);
+ if (existing) {
+ err = merge_extent_mapping(em_tree, existing,
+ em, start,
+ root->sectorsize);
+ free_extent_map(existing);
+ if (err) {
+ free_extent_map(em);
+ em = NULL;
+ }
+ } else {
+ err = -EIO;
+ free_extent_map(em);
+ em = NULL;
+ }
+ } else {
+ free_extent_map(em);
+ em = existing;
+ err = 0;
+ }
+ }
+ spin_unlock(&em_tree->lock);
+out:
+ if (path)
+ btrfs_free_path(path);
+ if (trans) {
+ ret = btrfs_end_transaction(trans, root);
+ if (!err)
+ err = ret;
+ }
+ if (err) {
+ free_extent_map(em);
+ WARN_ON(1);
+ return ERR_PTR(err);
+ }
+ return em;
+}
+
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
const struct iovec *iov, loff_t offset,
unsigned long nr_segs)
@@ -2859,6 +4805,18 @@
.fsync = btrfs_sync_file,
};
+static struct extent_io_ops btrfs_extent_io_ops = {
+ .fill_delalloc = run_delalloc_range,
+ .submit_bio_hook = btrfs_submit_bio_hook,
+ .merge_bio_hook = btrfs_merge_bio_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,
+};
+
static struct address_space_operations btrfs_aops = {
.readpage = btrfs_readpage,
.writepage = btrfs_writepage,
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
