From: "Yan, Zheng" <zheng.yan@oracle.com>
To: linux-btrfs@vger.kernel.org, Chris Mason <chris.mason@oracle.com>
Subject: [PATCH 1/8] Avoid superfluous tree-log writeout
Date: Mon, 02 Nov 2009 17:19:45 +0800 [thread overview]
Message-ID: <4AEEA431.1070301@oracle.com> (raw)
We allow two log transactions at a time, but use same flag
to mark dirty tree-log btree blocks. So we may flush dirty
blocks belonging to newer log transaction when committing a
log transaction. This patch fixes the issue by using two
flags to mark dirty tree-log btree blocks.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
---
diff -urp 1/fs/btrfs/disk-io.c 3/fs/btrfs/disk-io.c
--- 1/fs/btrfs/disk-io.c 2009-10-15 09:33:17.811423754 +0800
+++ 3/fs/btrfs/disk-io.c 2009-10-14 14:41:34.000000000 +0800
@@ -982,12 +982,12 @@ int btrfs_free_log_root_tree(struct btrf
while (1) {
ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
- 0, &start, &end, EXTENT_DIRTY);
+ 0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
if (ret)
break;
- clear_extent_dirty(&log_root_tree->dirty_log_pages,
- start, end, GFP_NOFS);
+ clear_extent_bits(&log_root_tree->dirty_log_pages, start, end,
+ EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
}
eb = fs_info->log_root_tree->node;
diff -urp 1/fs/btrfs/extent-tree.c 3/fs/btrfs/extent-tree.c
--- 1/fs/btrfs/extent-tree.c 2009-10-21 20:42:19.806223422 +0800
+++ 3/fs/btrfs/extent-tree.c 2009-10-14 13:27:51.000000000 +0800
@@ -4850,8 +4850,16 @@ struct extent_buffer *btrfs_init_new_buf
btrfs_set_buffer_uptodate(buf);
if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
- set_extent_dirty(&root->dirty_log_pages, buf->start,
- buf->start + buf->len - 1, GFP_NOFS);
+ /*
+ * we allow two log transactions at a time, use different
+ * EXENT bit to differentiate dirty pages.
+ */
+ if (root->log_transid % 2 == 0)
+ set_extent_dirty(&root->dirty_log_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);
+ else
+ set_extent_new(&root->dirty_log_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);
} else {
set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
buf->start + buf->len - 1, GFP_NOFS);
diff -urp 1/fs/btrfs/transaction.c 3/fs/btrfs/transaction.c
--- 1/fs/btrfs/transaction.c 2009-10-15 09:33:17.893417499 +0800
+++ 3/fs/btrfs/transaction.c 2009-10-14 13:27:51.000000000 +0800
@@ -347,7 +347,7 @@ int btrfs_end_transaction_throttle(struc
* those extents are sent to disk but does not wait on them
*/
int btrfs_write_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages)
+ struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int err = 0;
@@ -360,7 +360,7 @@ int btrfs_write_marked_extents(struct bt
while (1) {
ret = find_first_extent_bit(dirty_pages, start, &start, &end,
- EXTENT_DIRTY);
+ mark);
if (ret)
break;
while (start <= end) {
@@ -406,7 +406,7 @@ int btrfs_write_marked_extents(struct bt
* on all the pages and clear them from the dirty pages state tree
*/
int btrfs_wait_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages)
+ struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int err = 0;
@@ -418,12 +418,12 @@ int btrfs_wait_marked_extents(struct btr
unsigned long index;
while (1) {
- ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
- EXTENT_DIRTY);
+ ret = find_first_extent_bit(dirty_pages, start, &start, &end,
+ mark);
if (ret)
break;
- clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+ clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
while (start <= end) {
index = start >> PAGE_CACHE_SHIFT;
start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
@@ -453,13 +453,13 @@ int btrfs_wait_marked_extents(struct btr
* those extents are on disk for transaction or log commit
*/
int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages)
+ struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int ret2;
- ret = btrfs_write_marked_extents(root, dirty_pages);
- ret2 = btrfs_wait_marked_extents(root, dirty_pages);
+ ret = btrfs_write_marked_extents(root, dirty_pages, mark);
+ ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
return ret || ret2;
}
@@ -472,7 +472,8 @@ int btrfs_write_and_wait_transaction(str
return filemap_write_and_wait(btree_inode->i_mapping);
}
return btrfs_write_and_wait_marked_extents(root,
- &trans->transaction->dirty_pages);
+ &trans->transaction->dirty_pages,
+ EXTENT_DIRTY);
}
/*
diff -urp 1/fs/btrfs/transaction.h 3/fs/btrfs/transaction.h
--- 1/fs/btrfs/transaction.h 2009-10-15 09:33:17.894417497 +0800
+++ 3/fs/btrfs/transaction.h 2009-10-14 13:27:51.000000000 +0800
@@ -107,10 +107,10 @@ void btrfs_throttle(struct btrfs_root *r
int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
struct btrfs_root *root);
int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages);
+ struct extent_io_tree *dirty_pages, int mark);
int btrfs_write_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages);
+ struct extent_io_tree *dirty_pages, int mark);
int btrfs_wait_marked_extents(struct btrfs_root *root,
- struct extent_io_tree *dirty_pages);
+ struct extent_io_tree *dirty_pages, int mark);
int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
#endif
diff -urp 1/fs/btrfs/tree-log.c 3/fs/btrfs/tree-log.c
--- 1/fs/btrfs/tree-log.c 2009-10-15 09:33:17.896417074 +0800
+++ 3/fs/btrfs/tree-log.c 2009-10-14 14:42:01.000000000 +0800
@@ -1977,10 +1977,11 @@ int btrfs_sync_log(struct btrfs_trans_ha
{
int index1;
int index2;
+ int mark;
int ret;
struct btrfs_root *log = root->log_root;
struct btrfs_root *log_root_tree = root->fs_info->log_root_tree;
- u64 log_transid = 0;
+ unsigned long log_transid = 0;
mutex_lock(&root->log_mutex);
index1 = root->log_transid % 2;
@@ -2014,24 +2015,29 @@ int btrfs_sync_log(struct btrfs_trans_ha
goto out;
}
+ log_transid = root->log_transid;
+ if (log_transid % 2 == 0)
+ mark = EXTENT_DIRTY;
+ else
+ mark = EXTENT_NEW;
+
/* we start IO on all the marked extents here, but we don't actually
* wait for them until later.
*/
- ret = btrfs_write_marked_extents(log, &log->dirty_log_pages);
+ ret = btrfs_write_marked_extents(log, &log->dirty_log_pages, mark);
BUG_ON(ret);
btrfs_set_root_node(&log->root_item, log->node);
root->log_batch = 0;
- log_transid = root->log_transid;
root->log_transid++;
log->log_transid = root->log_transid;
root->log_start_pid = 0;
smp_mb();
/*
- * log tree has been flushed to disk, new modifications of
- * the log will be written to new positions. so it's safe to
- * allow log writers to go in.
+ * IO has been started, blocks of the log tree have WRITTEN flag set
+ * in their headers. new modifications of the log will be written to
+ * new positions. so it's safe to allow log writers to go in.
*/
mutex_unlock(&root->log_mutex);
@@ -2052,7 +2058,7 @@ int btrfs_sync_log(struct btrfs_trans_ha
index2 = log_root_tree->log_transid % 2;
if (atomic_read(&log_root_tree->log_commit[index2])) {
- btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+ btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
wait_log_commit(trans, log_root_tree,
log_root_tree->log_transid);
mutex_unlock(&log_root_tree->log_mutex);
@@ -2072,16 +2078,17 @@ int btrfs_sync_log(struct btrfs_trans_ha
* check the full commit flag again
*/
if (root->fs_info->last_trans_log_full_commit == trans->transid) {
- btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+ btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
mutex_unlock(&log_root_tree->log_mutex);
ret = -EAGAIN;
goto out_wake_log_root;
}
ret = btrfs_write_and_wait_marked_extents(log_root_tree,
- &log_root_tree->dirty_log_pages);
+ &log_root_tree->dirty_log_pages,
+ EXTENT_DIRTY | EXTENT_NEW);
BUG_ON(ret);
- btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+ btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
btrfs_set_super_log_root(&root->fs_info->super_for_commit,
log_root_tree->node->start);
@@ -2147,12 +2154,12 @@ int btrfs_free_log(struct btrfs_trans_ha
while (1) {
ret = find_first_extent_bit(&log->dirty_log_pages,
- 0, &start, &end, EXTENT_DIRTY);
+ 0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
if (ret)
break;
- clear_extent_dirty(&log->dirty_log_pages,
- start, end, GFP_NOFS);
+ clear_extent_bits(&log->dirty_log_pages, start, end,
+ EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
}
if (log->log_transid > 0) {
reply other threads:[~2009-11-02 9:19 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4AEEA431.1070301@oracle.com \
--to=zheng.yan@oracle.com \
--cc=chris.mason@oracle.com \
--cc=linux-btrfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.