linux-kernel - Re: [PATCH 3/7] Make TestSetPageDirty and dirty page accounting in one func

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Sat, 7 Jul 2012 22:42:28 +0800
From:	Fengguang Wu <fengguang.wu@...el.com>
To:	Kamezawa Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
Cc:	Sha Zhengju <handai.szj@...il.com>, linux-mm@...ck.org,
	cgroups@...r.kernel.org, gthelen@...gle.com, yinghan@...gle.com,
	akpm@...ux-foundation.org, mhocko@...e.cz,
	linux-kernel@...r.kernel.org, torvalds@...ux-foundation.org,
	viro@...iv.linux.org.uk, linux-fsdevel@...r.kernel.org,
	Sha Zhengju <handai.szj@...bao.com>
Subject: Re: [PATCH 3/7] Make TestSetPageDirty and dirty page accounting in
 one func

On Mon, Jul 02, 2012 at 08:14:02PM +0900, KAMEZAWA Hiroyuki wrote:
> (2012/06/28 20:01), Sha Zhengju wrote:
> > From: Sha Zhengju <handai.szj@...bao.com>
> > 
> > Commit a8e7d49a(Fix race in create_empty_buffers() vs __set_page_dirty_buffers())
> > extracts TestSetPageDirty from __set_page_dirty and is far away from
> > account_page_dirtied.But it's better to make the two operations in one single
> > function to keep modular.So in order to avoid the potential race mentioned in
> > commit a8e7d49a, we can hold private_lock until __set_page_dirty completes.
> > I guess there's no deadlock between ->private_lock and ->tree_lock by quick look.
> > 
> > It's a prepare patch for following memcg dirty page accounting patches.
> > 
> > Signed-off-by: Sha Zhengju <handai.szj@...bao.com>
> 
> I think there is no problem with the lock order.

Me think so, too.

> My small concern is the impact on the performance. IIUC, lock contention here can be
> seen if multiple threads write to the same file in parallel.
> Do you have any numbers before/after the patch ?

That would be a worthwhile test. The patch moves ->tree_lock and
->i_lock into ->private_lock, these are often contented locks..

For example, in the below case of 12 hard disks, each running 1 dd
write, the ->tree_lock and ->private_lock have the top #1 and #2
contentions.

lkp-nex04/JBOD-12HDD-thresh=1000M/ext4-1dd-1-3.3.0/lock_stat
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
                              class name    con-bounces    contentions   waittime-min   waittime-max waittime-total    acq-bounces   acquisitions   holdtime-min   holdtime-max holdtime-total
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

           &(&mapping->tree_lock)->rlock:      18629034       19138284           0.09        1029.32    24353812.07       49650988      482883410           0.11         186.88   260706119.09
           -----------------------------
           &(&mapping->tree_lock)->rlock            783          [<ffffffff81109267>] tag_pages_for_writeback+0x2b/0x9d
           &(&mapping->tree_lock)->rlock        3195817          [<ffffffff81100d6c>] add_to_page_cache_locked+0xa3/0x119
           &(&mapping->tree_lock)->rlock        3863710          [<ffffffff81108df7>] test_set_page_writeback+0x63/0x140
           &(&mapping->tree_lock)->rlock        3311518          [<ffffffff81172ade>] __set_page_dirty+0x25/0xa5
           -----------------------------
           &(&mapping->tree_lock)->rlock        3450725          [<ffffffff81100d6c>] add_to_page_cache_locked+0xa3/0x119
           &(&mapping->tree_lock)->rlock        3225542          [<ffffffff81172ade>] __set_page_dirty+0x25/0xa5
           &(&mapping->tree_lock)->rlock        2241958          [<ffffffff81108df7>] test_set_page_writeback+0x63/0x140
           &(&mapping->tree_lock)->rlock        7339603          [<ffffffff8110ac33>] test_clear_page_writeback+0x64/0x155

...............................................................................................................................................................................................

        &(&mapping->private_lock)->rlock:       1165199        1191201           0.11        2843.25     1621608.38       13341420      152761848           0.10        3727.92    33559035.07
        --------------------------------
        &(&mapping->private_lock)->rlock              1          [<ffffffff81172913>] __find_get_block_slow+0x5a/0x135
        &(&mapping->private_lock)->rlock         385576          [<ffffffff811735d6>] create_empty_buffers+0x48/0xbf
        &(&mapping->private_lock)->rlock         805624          [<ffffffff8117346d>] try_to_free_buffers+0x57/0xaa
        --------------------------------
        &(&mapping->private_lock)->rlock              1          [<ffffffff811746dd>] __getblk+0x1b8/0x257
        &(&mapping->private_lock)->rlock         952718          [<ffffffff8117346d>] try_to_free_buffers+0x57/0xaa
        &(&mapping->private_lock)->rlock         238482          [<ffffffff811735d6>] create_empty_buffers+0x48/0xbf

Thanks,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/