From 0ae45f63d4ef8d8eeec49c7d8b44a1775fff13e8 Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Mon, 2 Feb 2015 00:37:00 -0500 Subject: vfs: add support for a lazytime mount option Add a new mount option which enables a new "lazytime" mode. This mode causes atime, mtime, and ctime updates to only be made to the in-memory version of the inode. The on-disk times will only get updated when (a) if the inode needs to be updated for some non-time related change, (b) if userspace calls fsync(), syncfs() or sync(), or (c) just before an undeleted inode is evicted from memory. This is OK according to POSIX because there are no guarantees after a crash unless userspace explicitly requests via a fsync(2) call. For workloads which feature a large number of random write to a preallocated file, the lazytime mount option significantly reduces writes to the inode table. The repeated 4k writes to a single block will result in undesirable stress on flash devices and SMR disk drives. Even on conventional HDD's, the repeated writes to the inode table block will trigger Adjacent Track Interference (ATI) remediation latencies, which very negatively impact long tail latencies --- which is a very big deal for web serving tiers (for example). Google-Bug-Id: 18297052 Signed-off-by: Theodore Ts'o Signed-off-by: Al Viro --- fs/ext4/inode.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs/ext4') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 5653fa42930b..628df5ba44a6 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -4840,11 +4840,17 @@ int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode) * If the inode is marked synchronous, we don't honour that here - doing * so would cause a commit on atime updates, which we don't bother doing. * We handle synchronous inodes at the highest possible level. + * + * If only the I_DIRTY_TIME flag is set, we can skip everything. If + * I_DIRTY_TIME and I_DIRTY_SYNC is set, the only inode fields we need + * to copy into the on-disk inode structure are the timestamp files. */ void ext4_dirty_inode(struct inode *inode, int flags) { handle_t *handle; + if (flags == I_DIRTY_TIME) + return; handle = ext4_journal_start(inode, EXT4_HT_INODE, 2); if (IS_ERR(handle)) goto out; -- cgit v1.2.3 From a26f49926da938f47561f386be56a83dd37a496d Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Mon, 2 Feb 2015 00:37:02 -0500 Subject: ext4: add optimization for the lazytime mount option Add an optimization for the MS_LAZYTIME mount option so that we will opportunistically write out any inodes with the I_DIRTY_TIME flag set in a particular inode table block when we need to update some inode in that inode table block anyway. Also add some temporary code so that we can set the lazytime mount option without needing a modified /sbin/mount program which can set MS_LAZYTIME. We can eventually make this go away once util-linux has added support. Google-Bug-Id: 18297052 Signed-off-by: Theodore Ts'o Signed-off-by: Al Viro --- fs/ext4/inode.c | 64 +++++++++++++++++++++++++++++++++++++++++++++++++++++++-- fs/ext4/super.c | 10 +++++++++ 2 files changed, 72 insertions(+), 2 deletions(-) (limited to 'fs/ext4') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 628df5ba44a6..9193ea130dcb 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -4139,6 +4139,65 @@ static int ext4_inode_blocks_set(handle_t *handle, return 0; } +struct other_inode { + unsigned long orig_ino; + struct ext4_inode *raw_inode; +}; + +static int other_inode_match(struct inode * inode, unsigned long ino, + void *data) +{ + struct other_inode *oi = (struct other_inode *) data; + + if ((inode->i_ino != ino) || + (inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW | + I_DIRTY_SYNC | I_DIRTY_DATASYNC)) || + ((inode->i_state & I_DIRTY_TIME) == 0)) + return 0; + spin_lock(&inode->i_lock); + if (((inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW | + I_DIRTY_SYNC | I_DIRTY_DATASYNC)) == 0) && + (inode->i_state & I_DIRTY_TIME)) { + struct ext4_inode_info *ei = EXT4_I(inode); + + inode->i_state &= ~(I_DIRTY_TIME | I_DIRTY_TIME_EXPIRED); + spin_unlock(&inode->i_lock); + + spin_lock(&ei->i_raw_lock); + EXT4_INODE_SET_XTIME(i_ctime, inode, oi->raw_inode); + EXT4_INODE_SET_XTIME(i_mtime, inode, oi->raw_inode); + EXT4_INODE_SET_XTIME(i_atime, inode, oi->raw_inode); + ext4_inode_csum_set(inode, oi->raw_inode, ei); + spin_unlock(&ei->i_raw_lock); + trace_ext4_other_inode_update_time(inode, oi->orig_ino); + return -1; + } + spin_unlock(&inode->i_lock); + return -1; +} + +/* + * Opportunistically update the other time fields for other inodes in + * the same inode table block. + */ +static void ext4_update_other_inodes_time(struct super_block *sb, + unsigned long orig_ino, char *buf) +{ + struct other_inode oi; + unsigned long ino; + int i, inodes_per_block = EXT4_SB(sb)->s_inodes_per_block; + int inode_size = EXT4_INODE_SIZE(sb); + + oi.orig_ino = orig_ino; + ino = orig_ino & ~(inodes_per_block - 1); + for (i = 0; i < inodes_per_block; i++, ino++, buf += inode_size) { + if (ino == orig_ino) + continue; + oi.raw_inode = (struct ext4_inode *) buf; + (void) find_inode_nowait(sb, ino, other_inode_match, &oi); + } +} + /* * Post the struct inode info into an on-disk inode location in the * buffer-cache. This gobbles the caller's reference to the @@ -4248,10 +4307,11 @@ static int ext4_do_update_inode(handle_t *handle, cpu_to_le16(ei->i_extra_isize); } } - ext4_inode_csum_set(inode, raw_inode, ei); - spin_unlock(&ei->i_raw_lock); + if (inode->i_sb->s_flags & MS_LAZYTIME) + ext4_update_other_inodes_time(inode->i_sb, inode->i_ino, + bh->b_data); BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); rc = ext4_handle_dirty_metadata(handle, NULL, bh); diff --git a/fs/ext4/super.c b/fs/ext4/super.c index 74c5f53595fb..362b23c8497a 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -1139,6 +1139,7 @@ enum { Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, + Opt_lazytime, Opt_nolazytime, Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, Opt_inode_readahead_blks, Opt_journal_ioprio, Opt_dioread_nolock, Opt_dioread_lock, @@ -1202,6 +1203,8 @@ static const match_table_t tokens = { {Opt_i_version, "i_version"}, {Opt_stripe, "stripe=%u"}, {Opt_delalloc, "delalloc"}, + {Opt_lazytime, "lazytime"}, + {Opt_nolazytime, "nolazytime"}, {Opt_nodelalloc, "nodelalloc"}, {Opt_removed, "mblk_io_submit"}, {Opt_removed, "nomblk_io_submit"}, @@ -1459,6 +1462,12 @@ static int handle_mount_opt(struct super_block *sb, char *opt, int token, case Opt_i_version: sb->s_flags |= MS_I_VERSION; return 1; + case Opt_lazytime: + sb->s_flags |= MS_LAZYTIME; + return 1; + case Opt_nolazytime: + sb->s_flags &= ~MS_LAZYTIME; + return 1; } for (m = ext4_mount_opts; m->token != Opt_err; m++) @@ -5020,6 +5029,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data) } #endif + *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME); ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); kfree(orig_data); return 0; -- cgit v1.2.3