Linux-Fsdevel Archive on lore.kernel.org
 help / color / Atom feed
From: ebiederm@xmission.com (Eric W. Biederman)
To: Andrew Morton <akpm@linux-foundation.org>
Cc: <linux-kernel@vger.kernel.org>, <linux-pci@vger.kernel.org>,
	<linux-mm@kvack.org>, <linux-fsdevel@vger.kernel.org>,
	Al Viro <viro@ZenIV.linux.org.uk>,
	Hugh Dickins <hugh@veritas.com>, Tejun Heo <tj@kernel.org>,
	Alexey Dobriyan <adobriyan@gmail.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Alan Cox <alan@lxorguk.ukuu.org.uk>,
	Greg Kroah-Hartman <gregkh@suse.de>
Subject: [RFC][PATCH 7/9] vfs: Optimize fops_read_lock
Date: Sat, 11 Apr 2009 05:11:59 -0700
Message-ID: <m1ws9r5r00.fsf@fess.ebiederm.org> (raw)
In-Reply-To: <m1skkf761y.fsf@fess.ebiederm.org> (Eric W. Biederman's message of "Sat\, 11 Apr 2009 05\:01\:29 -0700")


After seeing fget_light and the justification for it in the commit
log I did not want to introduce something into the common read/write
path of file descriptors that could have a significant measurable impact
on I/O speed.

commit f6435db01709533f270b2dce1e5914770dbc65de
Author: akpm <akpm>
Date:   Thu May 8 05:19:50 2003 +0000

    [PATCH] reduced overheads in fget/fput

    From: Dipankar Sarma <dipankar@in.ibm.com>

    fget() shows up on profiles, especially on SMP.  Dipankar's patch
    special-cases the situation wherein there are no sharers of current->files.

    In this situation we know that no other process can close this file, so it
    is not necessary to increment the file's refcount.

    It's ugly as sin, but makes a substantial difference.

    The test is

        dd if=/dev/zero of=foo bs=1 count=1M

    On 4CPU P3 xeon with 1MB L2 cache and 512MB ram:

        kernel           sys time     std-dev
        ------------     --------     -------

        UP - vanilla     2.104        0.028
        UP - file        1.867        0.019

        SMP - vanilla    2.976        0.023
        SMP - file       2.719        0.026

    BKrev: 3eb9e8f6Db0nMWoSx5IdHx6SBal8aw

My test case was:
    dd if=/dev/zero of=/dev/null bs=1 count=1M.

As writing to a real file turned out to cover the cost.

Without the optimization I am seeing 2.4 - 2.5 MB/s on my
idle core2 single socket quad core.

With this optimization I am seeing 2.9 - 3.0 MB/s on the
same machine.  Maybe 2% slower than before I introduced
fops_read_lock.

The common case is that there is only one thread and so
the fget_light optimization applies and f_count remains
at 1.  Which implies that there is only a single process
performing operations through the file descriptor.

In that case because there is no possible contention
it is possible safely skip the atomic operations, gaining all
of the benefits of rcu without requiring a per cpu variable.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/file_table.c |   18 ++++++++++++++----
 1 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/fs/file_table.c b/fs/file_table.c
index d216557..634d44c 100644
--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -495,15 +495,25 @@ void file_kill(struct file *file)
 int fops_read_lock(struct file *file)
 {
 	int revoked = (file->f_mode & FMODE_REVOKED);
-	if (likely(!revoked))
-		atomic_long_inc(&file->f_use);
+	if (likely(!revoked)) {
+		if (likely(atomic_long_read(&file->f_count) == 1))
+			atomic_long_set(&file->f_use,
+				atomic_long_read(&file->f_use) + 1);
+		else
+			atomic_long_inc(&file->f_use);
+	}
 	return revoked;
 }
 
 void fops_read_unlock(struct file *file, int revoked)
 {
-	if (likely(!revoked))
-		atomic_long_dec(&file->f_use);
+	if (likely(!revoked)) {
+		if (likely(atomic_long_read(&file->f_count) == 1))
+			atomic_long_set(&file->f_use,
+				atomic_long_read(&file->f_use) - 1);
+		else
+			atomic_long_dec(&file->f_use);
+	}
 }
 
 int fs_may_remount_ro(struct super_block *sb)
-- 
1.6.1.2.350.g88cc

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  parent reply index

Thread overview: 99+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-04-11 12:01 [RFC][PATCH 0/9] File descriptor hot-unplug support Eric W. Biederman
2009-04-11 12:03 ` [RFC][PATCH 1/9] mm: Introduce remap_file_mappings Eric W. Biederman
2009-04-11 12:05 ` [RFC][PATCH 2/9] mm: Implement generic support for revoking a mapping Eric W. Biederman
2009-04-11 12:06 ` [RFC][PATCH 3/9] sysfs: Use remap_file_mappings Eric W. Biederman
2009-04-11 12:07 ` [RFC][PATCH 4/9] vfs: Generalize the file_list Eric W. Biederman
2009-04-11 12:08 ` [RFC][PATCH 5/9] vfs: Introduce basic infrastructure for revoking a file Eric W. Biederman
2009-04-14 22:12   ` Jonathan Corbet
2009-04-15  2:55     ` Eric W. Biederman
2009-04-11 12:10 ` [RFC][PATCH 6/9] vfs: Utilize fops_read_lock where appropriate Eric W. Biederman
2009-04-11 12:11 ` Eric W. Biederman [this message]
2009-04-11 12:13 ` [RFC][PATCH 8/9] vfs: Implement generic revoked file operations Eric W. Biederman
2009-04-12 18:56   ` Jamie Lokier
2009-04-12 20:04     ` Eric W. Biederman
2009-04-12 20:31       ` Jamie Lokier
2009-04-12 21:53         ` Eric W. Biederman
2009-04-12 20:54       ` Eric W. Biederman
2009-04-12 21:02         ` Jamie Lokier
2009-04-12 23:06           ` Eric W. Biederman
2009-04-11 12:14 ` [RFC][PATCH 9/9] proc: Use the generic vfs revoke facility that now exists Eric W. Biederman
2009-04-11 15:58 ` [RFC][PATCH 0/9] File descriptor hot-unplug support Al Viro
2009-04-11 16:49   ` Eric W. Biederman
2009-04-11 16:56     ` Al Viro
2009-04-11 23:57       ` Eric W. Biederman
2009-04-12 20:21       ` Eric W. Biederman
2009-04-14  3:16 ` Tejun Heo
2009-04-14  7:39   ` Eric W. Biederman
2009-04-14  7:45     ` Tejun Heo
2009-04-14  8:27       ` Eric W. Biederman
2009-04-14  8:49         ` Tejun Heo
2009-04-14 15:07         ` Jamie Lokier
2009-04-14 19:09           ` Eric W. Biederman
2009-06-01 21:45 ` [PATCH 0/23] File descriptor hot-unplug support v2 Eric W. Biederman
2009-06-01 21:50   ` [PATCH 01/23] mm: Introduce revoke_file_mappings Eric W. Biederman
2009-06-01 22:25     ` Andrew Morton
2009-06-02  0:12       ` Eric W. Biederman
2009-06-01 21:50   ` [PATCH 02/23] vfs: Implement unpoll_file Eric W. Biederman
2009-06-06  8:08     ` Al Viro
2009-06-01 21:50   ` [PATCH 03/23] vfs: Generalize the file_list Eric W. Biederman
2009-06-02  7:06     ` Nick Piggin
2009-06-05 19:33       ` Eric W. Biederman
2009-06-09 10:38         ` Nick Piggin
2009-06-09 18:38           ` Eric W. Biederman
2009-06-10  6:05             ` Nick Piggin
2009-06-01 21:50   ` [PATCH 04/23] vfs: Introduce infrastructure for revoking a file Eric W. Biederman
2009-06-02  5:16     ` Pekka Enberg
2009-06-02  6:51       ` Eric W. Biederman
2009-06-02  7:08         ` Pekka Enberg
2009-06-02  7:14     ` Nick Piggin
2009-06-02 17:06       ` Linus Torvalds
2009-06-02 20:52         ` Eric W. Biederman
2009-06-03  6:37           ` Nick Piggin
2009-06-02 22:56       ` Eric W. Biederman
2009-06-03  6:38         ` Nick Piggin
2009-06-05  9:03     ` Miklos Szeredi
2009-06-05 19:06       ` Eric W. Biederman
2009-06-01 21:50   ` [PATCH 05/23] vfs: Teach lseek to use file_hotplug_lock Eric W. Biederman
2009-06-01 21:50   ` [PATCH 06/23] vfs: Teach read/write to use file_hotplug_read_lock Eric W. Biederman
2009-06-01 21:50   ` [PATCH 07/23] vfs: Teach sendfile,splice,tee,and vmsplice to use file_hotplug_lock Eric W. Biederman
2009-06-03 23:39     ` Badari Pulavarty
2009-06-05 19:37       ` Eric W. Biederman
2009-06-01 21:50   ` [PATCH 08/23] vfs: Teach readdir " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 09/23] vfs: Teach poll and select " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 10/23] vfs: Teach do_path_lookup " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 11/23] mm: Teach mmap " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 12/23] vfs: Teach fcntl " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 13/23] vfs: Teach ioctl " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 14/23] vfs: Teach flock " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 15/23] vfs: Teach fallocate, and filp_close " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 16/23] vfs: Teach fstatfs, fstatfs64, ftruncate, fchdir, fchmod, fchown " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 17/23] proc: Teach /proc/<pid>/fd " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 18/23] vfs: Teach epoll " Eric W. Biederman
2009-06-02 16:51     ` Davide Libenzi
2009-06-02 21:23       ` Eric W. Biederman
2009-06-02 21:52         ` Davide Libenzi
2009-06-02 22:51           ` Eric W. Biederman
2009-06-03 14:57             ` Davide Libenzi
2009-06-03 20:53               ` Eric W. Biederman
2009-06-04  0:50                 ` Davide Libenzi
2009-06-04  1:42                   ` Eric W. Biederman
2009-06-01 21:50   ` [PATCH 19/23] eventpoll: Fix comment Eric W. Biederman
2009-06-01 21:50   ` [PATCH 20/23] vfs: Teach aio to use file_hotplug_lock Eric W. Biederman
2009-06-01 21:50   ` [PATCH 21/23] vfs: Teach fsync " Eric W. Biederman
2009-06-01 21:50   ` [PATCH 22/23] vfs: Teach fadvice to file_hotplug_lock Eric W. Biederman
2009-06-01 21:50   ` [PATCH 23/23] vfs: Teach readahead to use the file_hotplug_lock Eric W. Biederman
2009-06-03 23:25     ` Badari Pulavarty
2009-06-06  8:03   ` [PATCH 0/23] File descriptor hot-unplug support v2 Al Viro
2009-06-08  9:41     ` Miklos Szeredi
2009-06-08 10:24       ` Jamie Lokier
2009-06-08 16:29       ` Al Viro
2009-06-08 16:44         ` Miklos Szeredi
2009-06-08 17:50           ` Al Viro
2009-06-08 18:01             ` Linus Torvalds
2009-06-08 18:50               ` Al Viro
2009-06-08 19:18                 ` Linus Torvalds
2009-06-09  6:42                   ` Eric W. Biederman
2009-06-09 10:52                     ` Nick Piggin
2009-06-09  5:50             ` Miklos Szeredi
2009-06-09  6:31               ` Eric W. Biederman
2009-06-09  6:22     ` Eric W. Biederman

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=m1ws9r5r00.fsf@fess.ebiederm.org \
    --to=ebiederm@xmission.com \
    --cc=adobriyan@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=alan@lxorguk.ukuu.org.uk \
    --cc=gregkh@suse.de \
    --cc=hugh@veritas.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=tj@kernel.org \
    --cc=torvalds@linux-foundation.org \
    --cc=viro@ZenIV.linux.org.uk \
    /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

Linux-Fsdevel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-fsdevel/0 linux-fsdevel/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-fsdevel linux-fsdevel/ https://lore.kernel.org/linux-fsdevel \
		linux-fsdevel@vger.kernel.org
	public-inbox-index linux-fsdevel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-fsdevel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git