From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dmitry Monakhov Subject: [PATCH 3/3] ext4: optimize quota write locking Date: Sat, 27 Mar 2010 15:15:40 +0300 Message-ID: <1269692140-5375-4-git-send-email-dmonakhov@openvz.org> References: <1269692140-5375-1-git-send-email-dmonakhov@openvz.org> <1269692140-5375-2-git-send-email-dmonakhov@openvz.org> <1269692140-5375-3-git-send-email-dmonakhov@openvz.org> Cc: linux-fsdevel@vger.kernel.org, jack@suse.cz, Dmitry Monakhov To: linux-ext4@vger.kernel.org Return-path: Received: from mail-bw0-f209.google.com ([209.85.218.209]:61402 "EHLO mail-bw0-f209.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753026Ab0C0MP6 (ORCPT ); Sat, 27 Mar 2010 08:15:58 -0400 In-Reply-To: <1269692140-5375-3-git-send-email-dmonakhov@openvz.org> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Quota inode is append only. In most cases writes performed to allocated blocks before i_size. Currently we grup i_mutex for each quota write to protect i_size and i_mtime. But in most cases i_size not changed. And i_mtime is not that important as for regular files. So let's update it only when quota file increased. This allow us to avoid i_mutex in most cases. Some day quota file will be hiden completly, so i_mtime hack will disappear. Signed-off-by: Dmitry Monakhov --- fs/ext4/super.c | 22 +++++++++++++++------- 1 files changed, 15 insertions(+), 7 deletions(-) diff --git a/fs/ext4/super.c b/fs/ext4/super.c index b7b5707..5a378e3 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -4039,6 +4039,7 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type, ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); int err = 0; int offset = off & (sb->s_blocksize - 1); + int is_locked = 0; int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL; struct buffer_head *bh; handle_t *handle = journal_current_handle(); @@ -4059,8 +4060,14 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type, (unsigned long long)off, (unsigned long long)len); return -EIO; } - - mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); + /* + * Inode file is append only. It is not necessery to hold i_mutex + * if quota block is allocated already. Optimize that case. + */ + if (off + len > i_size_read(inode)) { + mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); + is_locked = 1; + } bh = ext4_bread(handle, inode, blk, 1, &err); if (!bh) goto out; @@ -4085,16 +4092,17 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type, brelse(bh); out: if (err) { - mutex_unlock(&inode->i_mutex); + if (is_locked) + mutex_unlock(&inode->i_mutex); return err; } - if (inode->i_size < off + len) { + if (is_locked) { i_size_write(inode, off + len); EXT4_I(inode)->i_disksize = inode->i_size; + inode->i_mtime = inode->i_ctime = CURRENT_TIME; + ext4_mark_inode_dirty(handle, inode); + mutex_unlock(&inode->i_mutex); } - inode->i_mtime = inode->i_ctime = CURRENT_TIME; - ext4_mark_inode_dirty(handle, inode); - mutex_unlock(&inode->i_mutex); return len; } -- 1.6.6.1