#include <linux/radix-tree.h>
#include <linux/writeback.h>
#include <linux/buffer_head.h> // for block_sync_page
+#include <linux/workqueue.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#endif
static struct extent_io_ops btree_extent_io_ops;
+static struct workqueue_struct *end_io_workqueue;
+
+struct end_io_wq {
+ struct bio *bio;
+ bio_end_io_t *end_io;
+ void *private;
+ struct btrfs_fs_info *info;
+ int error;
+ int metadata;
+ struct list_head list;
+};
struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
size_t page_offset, u64 start, u64 len,
memcpy(&found, result, BTRFS_CRC32_SIZE);
read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
- WARN_ON(1);
printk("btrfs: %s checksum verify failed on %llu "
"wanted %X found %X from_this_trans %d "
"level %d\n",
return 0;
}
+static int btree_read_extent_buffer_pages(struct btrfs_root *root,
+ struct extent_buffer *eb,
+ u64 start)
+{
+ struct extent_io_tree *io_tree;
+ int ret;
+ int num_copies = 0;
+ int mirror_num = 0;
+
+ io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
+ while (1) {
+ ret = read_extent_buffer_pages(io_tree, eb, start, 1,
+ btree_get_extent, mirror_num);
+ if (!ret) {
+ if (mirror_num)
+printk("good read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
+ return ret;
+ }
+ num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
+ eb->start, eb->len);
+printk("failed to read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
+ if (num_copies == 1) {
+printk("reading %Lu failed only one copy\n", eb->start);
+ return ret;
+ }
+ mirror_num++;
+ if (mirror_num > num_copies) {
+printk("bailing at mirror %d of %d\n", mirror_num, num_copies);
+ return ret;
+ }
+ }
+printk("read extent buffer page last\n");
+ return -EIO;
+}
int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
{
int found_level;
unsigned long len;
struct extent_buffer *eb;
+ int ret;
+
tree = &BTRFS_I(page->mapping->host)->io_tree;
if (page->private == EXTENT_PAGE_PRIVATE)
WARN_ON(1);
}
eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
- read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
- btree_get_extent);
+ ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE);
+ BUG_ON(ret);
btrfs_clear_buffer_defrag(eb);
found_start = btrfs_header_bytenr(eb);
if (found_start != start) {
return 0;
}
-static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
+int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
+ struct extent_state *state)
+{
+ struct extent_io_tree *tree;
+ u64 found_start;
+ int found_level;
+ unsigned long len;
+ struct extent_buffer *eb;
+ struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+ int ret = 0;
+
+ tree = &BTRFS_I(page->mapping->host)->io_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);
+
+ btrfs_clear_buffer_defrag(eb);
+ found_start = btrfs_header_bytenr(eb);
+ if (found_start != start) {
+printk("bad start on %Lu found %Lu\n", eb->start, found_start);
+ ret = -EIO;
+ goto err;
+ }
+ if (eb->first_page != page) {
+ printk("bad first page %lu %lu\n", eb->first_page->index,
+ page->index);
+ WARN_ON(1);
+ ret = -EIO;
+ goto err;
+ }
+ found_level = btrfs_header_level(eb);
+
+ ret = csum_tree_block(root, eb, 1);
+ if (ret)
+ ret = -EIO;
+
+ end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
+ end = eb->start + end - 1;
+ release_extent_buffer_tail_pages(eb);
+err:
+ free_extent_buffer(eb);
+out:
+ return ret;
+}
+
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_workqueue_bio(struct bio *bio, int err)
+#else
+static int end_workqueue_bio(struct bio *bio,
+ unsigned int bytes_done, int err)
+#endif
+{
+ struct end_io_wq *end_io_wq = bio->bi_private;
+ struct btrfs_fs_info *fs_info;
+ unsigned long flags;
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+ if (bio->bi_size)
+ return 1;
+#endif
+
+ fs_info = end_io_wq->info;
+ spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+ end_io_wq->error = err;
+ list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list);
+ spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
+ queue_work(end_io_workqueue, &fs_info->end_io_work);
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+ return 0;
+#endif
+}
+
+int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
+ int metadata)
+{
+ struct end_io_wq *end_io_wq;
+ end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
+ if (!end_io_wq)
+ return -ENOMEM;
+
+ end_io_wq->private = bio->bi_private;
+ end_io_wq->end_io = bio->bi_end_io;
+ end_io_wq->info = info;
+ end_io_wq->error = 0;
+ end_io_wq->bio = bio;
+ end_io_wq->metadata = metadata;
+
+ bio->bi_private = end_io_wq;
+ bio->bi_end_io = end_workqueue_bio;
+ return 0;
+}
+
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+ int mirror_num)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
u64 offset;
+ int ret;
+
offset = bio->bi_sector << 9;
+
+ if (rw & (1 << BIO_RW)) {
+ return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
+ }
+
+ ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
+ BUG_ON(ret);
+
if (offset == BTRFS_SUPER_INFO_OFFSET) {
- bio->bi_bdev = root->fs_info->sb->s_bdev;
+ bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
submit_bio(rw, bio);
return 0;
}
- return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
+ return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
}
static int btree_writepage(struct page *page, struct writeback_control *wbc)
struct extent_map_tree *map;
int ret;
+ if (page_count(page) > 3) {
+ /* once for page->private, once for the caller, once
+ * once for the page cache
+ */
+ return 0;
+ }
tree = &BTRFS_I(page->mapping->host)->io_tree;
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);
if (!buf)
return 0;
read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
- buf, 0, 0, btree_get_extent);
+ buf, 0, 0, btree_get_extent, 0);
free_extent_buffer(buf);
return ret;
}
int btrfs_verify_block_csum(struct btrfs_root *root,
struct extent_buffer *buf)
{
- struct extent_io_tree *io_tree;
- u64 end;
- int ret;
-
- io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
- if (buf->flags & EXTENT_CSUM)
- return 0;
-
- end = min_t(u64, buf->len, PAGE_CACHE_SIZE);
- end = buf->start + end - 1;
- if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
- buf->flags |= EXTENT_CSUM;
- return 0;
- }
-
- lock_extent(io_tree, buf->start, end, GFP_NOFS);
-
- if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
- buf->flags |= EXTENT_CSUM;
- ret = 0;
- goto out_unlock;
- }
-
- ret = csum_tree_block(root, buf, 1);
- set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS);
- buf->flags |= EXTENT_CSUM;
-
-out_unlock:
- unlock_extent(io_tree, buf->start, end, GFP_NOFS);
- return ret;
+ return btrfs_buffer_uptodate(buf);
}
struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
if (!buf)
return NULL;
- read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
- btree_get_extent);
- ret = btrfs_verify_block_csum(root, buf);
+ ret = btree_read_extent_buffer_pages(root, buf, 0);
+
+ if (ret == 0) {
+ buf->flags |= EXTENT_UPTODATE;
+ }
return buf;
+
}
int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
return 0;
}
+static int bio_ready_for_csum(struct bio *bio)
+{
+ u64 length = 0;
+ u64 buf_len = 0;
+ u64 start = 0;
+ struct page *page;
+ struct extent_io_tree *io_tree = NULL;
+ struct btrfs_fs_info *info = NULL;
+ struct bio_vec *bvec;
+ int i;
+ int ret;
+
+ bio_for_each_segment(bvec, bio, i) {
+ page = bvec->bv_page;
+ if (page->private == EXTENT_PAGE_PRIVATE) {
+ length += bvec->bv_len;
+ continue;
+ }
+ if (!page->private) {
+ length += bvec->bv_len;
+ continue;
+ }
+ length = bvec->bv_len;
+ buf_len = page->private >> 2;
+ start = page_offset(page) + bvec->bv_offset;
+ io_tree = &BTRFS_I(page->mapping->host)->io_tree;
+ info = BTRFS_I(page->mapping->host)->root->fs_info;
+ }
+ /* are we fully contained in this bio? */
+ if (buf_len <= length)
+ return 1;
+
+ ret = extent_range_uptodate(io_tree, start + length,
+ start + buf_len - 1);
+ if (ret == 1)
+ return ret;
+ return ret;
+}
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+void btrfs_end_io_csum(void *p)
+#else
+void btrfs_end_io_csum(struct work_struct *work)
+#endif
+{
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+ struct btrfs_fs_info *fs_info = p;
+#else
+ struct btrfs_fs_info *fs_info = container_of(work,
+ struct btrfs_fs_info,
+ end_io_work);
+#endif
+ unsigned long flags;
+ struct end_io_wq *end_io_wq;
+ struct bio *bio;
+ struct list_head *next;
+ int error;
+ int was_empty;
+
+ while(1) {
+ spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+ if (list_empty(&fs_info->end_io_work_list)) {
+ spin_unlock_irqrestore(&fs_info->end_io_work_lock,
+ flags);
+ return;
+ }
+ next = fs_info->end_io_work_list.next;
+ list_del(next);
+ spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
+
+ end_io_wq = list_entry(next, struct end_io_wq, list);
+
+ bio = end_io_wq->bio;
+ if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
+ spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+ was_empty = list_empty(&fs_info->end_io_work_list);
+ list_add_tail(&end_io_wq->list,
+ &fs_info->end_io_work_list);
+ spin_unlock_irqrestore(&fs_info->end_io_work_lock,
+ flags);
+ if (was_empty)
+ return;
+ continue;
+ }
+ error = end_io_wq->error;
+ bio->bi_private = end_io_wq->private;
+ bio->bi_end_io = end_io_wq->end_io;
+ kfree(end_io_wq);
+ bio_endio(bio, error);
+ }
+}
+
+
struct btrfs_root *open_ctree(struct super_block *sb,
struct btrfs_fs_devices *fs_devices)
{
err = -ENOMEM;
goto fail;
}
+ end_io_workqueue = create_workqueue("btrfs-end-io");
+ BUG_ON(!end_io_workqueue);
+
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);
+ INIT_LIST_HEAD(&fs_info->end_io_work_list);
spin_lock_init(&fs_info->hash_lock);
+ spin_lock_init(&fs_info->end_io_work_lock);
spin_lock_init(&fs_info->delalloc_lock);
spin_lock_init(&fs_info->new_trans_lock);
init_completion(&fs_info->kobj_unregister);
- sb_set_blocksize(sb, 4096);
+ sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE);
fs_info->tree_root = tree_root;
fs_info->extent_root = extent_root;
fs_info->chunk_root = chunk_root;
fs_info->btree_inode->i_mapping, GFP_NOFS);
fs_info->do_barriers = 1;
+ INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
#else
btrfs_read_block_groups(extent_root);
fs_info->generation = btrfs_super_generation(disk_super) + 1;
- if (btrfs_super_num_devices(disk_super) > 0) {
- fs_info->data_alloc_profile = BTRFS_BLOCK_GROUP_RAID0;
- fs_info->metadata_alloc_profile = BTRFS_BLOCK_GROUP_RAID1;
- fs_info->system_alloc_profile = BTRFS_BLOCK_GROUP_RAID0;
- }
+ fs_info->data_alloc_profile = (u64)-1;
+ fs_info->metadata_alloc_profile = (u64)-1;
+ fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
+
mutex_unlock(&fs_info->fs_mutex);
return tree_root;
return ERR_PTR(err);
}
+static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
+{
+ char b[BDEVNAME_SIZE];
+
+ if (uptodate) {
+ set_buffer_uptodate(bh);
+ } else {
+ if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
+ printk(KERN_WARNING "lost page write due to "
+ "I/O error on %s\n",
+ bdevname(bh->b_bdev, b));
+ }
+ set_buffer_write_io_error(bh);
+ clear_buffer_uptodate(bh);
+ }
+ unlock_buffer(bh);
+ put_bh(bh);
+}
+
+int write_all_supers(struct btrfs_root *root)
+{
+ struct list_head *cur;
+ struct list_head *head = &root->fs_info->fs_devices->devices;
+ struct btrfs_device *dev;
+ struct extent_buffer *sb;
+ struct btrfs_dev_item *dev_item;
+ struct buffer_head *bh;
+ int ret;
+ int do_barriers;
+
+ do_barriers = !btrfs_test_opt(root, NOBARRIER);
+
+ sb = root->fs_info->sb_buffer;
+ dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
+ dev_item);
+ list_for_each(cur, head) {
+ dev = list_entry(cur, struct btrfs_device, dev_list);
+ btrfs_set_device_type(sb, dev_item, dev->type);
+ btrfs_set_device_id(sb, dev_item, dev->devid);
+ btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes);
+ btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used);
+ btrfs_set_device_io_align(sb, dev_item, dev->io_align);
+ btrfs_set_device_io_width(sb, dev_item, dev->io_width);
+ btrfs_set_device_sector_size(sb, dev_item, dev->sector_size);
+ write_extent_buffer(sb, dev->uuid,
+ (unsigned long)btrfs_device_uuid(dev_item),
+ BTRFS_DEV_UUID_SIZE);
+
+ btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN);
+ csum_tree_block(root, sb, 0);
+
+ bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET /
+ root->fs_info->sb->s_blocksize,
+ BTRFS_SUPER_INFO_SIZE);
+
+ read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE);
+ dev->pending_io = bh;
+
+ get_bh(bh);
+ set_buffer_uptodate(bh);
+ lock_buffer(bh);
+ bh->b_end_io = btrfs_end_buffer_write_sync;
+
+ if (do_barriers && dev->barriers) {
+ ret = submit_bh(WRITE_BARRIER, bh);
+ if (ret == -EOPNOTSUPP) {
+ printk("btrfs: disabling barriers on dev %s\n",
+ dev->name);
+ set_buffer_uptodate(bh);
+ dev->barriers = 0;
+ get_bh(bh);
+ lock_buffer(bh);
+ ret = submit_bh(WRITE, bh);
+ }
+ } else {
+ ret = submit_bh(WRITE, bh);
+ }
+ BUG_ON(ret);
+ }
+
+ list_for_each(cur, head) {
+ dev = list_entry(cur, struct btrfs_device, dev_list);
+ BUG_ON(!dev->pending_io);
+ bh = dev->pending_io;
+ wait_on_buffer(bh);
+ if (!buffer_uptodate(dev->pending_io)) {
+ if (do_barriers && dev->barriers) {
+ printk("btrfs: disabling barriers on dev %s\n",
+ dev->name);
+ set_buffer_uptodate(bh);
+ get_bh(bh);
+ lock_buffer(bh);
+ dev->barriers = 0;
+ ret = submit_bh(WRITE, bh);
+ BUG_ON(ret);
+ wait_on_buffer(bh);
+ BUG_ON(!buffer_uptodate(bh));
+ } else {
+ BUG();
+ }
+
+ }
+ dev->pending_io = NULL;
+ brelse(bh);
+ }
+ return 0;
+}
+
int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
*root)
{
int ret;
- struct extent_buffer *super = root->fs_info->sb_buffer;
- struct inode *btree_inode = root->fs_info->btree_inode;
- struct super_block *sb = root->fs_info->sb;
+ ret = write_all_supers(root);
+#if 0
if (!btrfs_test_opt(root, NOBARRIER))
blkdev_issue_flush(sb->s_bdev, NULL);
set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
super->start, super->len);
if (!btrfs_test_opt(root, NOBARRIER))
blkdev_issue_flush(sb->s_bdev, NULL);
+#endif
return ret;
}
extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+ flush_workqueue(end_io_workqueue);
+ destroy_workqueue(end_io_workqueue);
iput(fs_info->btree_inode);
#if 0
int btrfs_read_buffer(struct extent_buffer *buf)
{
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)->io_tree,
- buf, 0, 1, btree_get_extent);
+ int ret;
+ ret = btree_read_extent_buffer_pages(root, buf, 0);
+ if (ret == 0) {
+ buf->flags |= EXTENT_UPTODATE;
+ }
+ return ret;
}
static struct extent_io_ops btree_extent_io_ops = {
.writepage_io_hook = btree_writepage_io_hook,
+ .readpage_end_io_hook = btree_readpage_end_io_hook,
.submit_bio_hook = btree_submit_bio_hook,
/* note we're sharing with inode.c for the merge bio hook */
.merge_bio_hook = btrfs_merge_bio_hook,
projects / deliverable / linux.git / blobdiff