linux-kernel - Re: [PATCH -mmotm 1/5] memcg: disable irq at page cgroup lock

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]

Message-Id: <20100415091444.aa743668.kamezawa.hiroyu@jp.fujitsu.com>
Date:	Thu, 15 Apr 2010 09:14:44 +0900
From:	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
To:	Vivek Goyal <vgoyal@...hat.com>
Cc:	Greg Thelen <gthelen@...gle.com>, balbir@...ux.vnet.ibm.com,
	Andrea Righi <arighi@...eler.com>,
	Daisuke Nishimura <nishimura@....nes.nec.co.jp>,
	Peter Zijlstra <peterz@...radead.org>,
	Trond Myklebust <trond.myklebust@....uio.no>,
	Suleiman Souhlal <suleiman@...gle.com>,
	"Kirill A. Shutemov" <kirill@...temov.name>,
	Andrew Morton <akpm@...ux-foundation.org>,
	containers@...ts.linux-foundation.org,
	linux-kernel@...r.kernel.org, linux-mm@...ck.org
Subject: Re: [PATCH -mmotm 1/5] memcg: disable irq at page cgroup lock

On Wed, 14 Apr 2010 10:04:30 -0400
Vivek Goyal <vgoyal@...hat.com> wrote:

> On Wed, Apr 14, 2010 at 06:29:04PM +0900, KAMEZAWA Hiroyuki wrote:

> Hi Kame-san,
> 
> May be I am missing something but how does it solve the issue of making sure
> lock_page_cgroup() is not held in interrupt context? IIUC, above code will
> make sure that for file cache accouting, lock_page_cgroup() is taken only
> if task migration is on. But say task migration is on, and then some IO
> completes and we update WRITEBACK stat (i think this is the one which can
> be called from interrupt context), then we will still take the
> lock_page_cgroup() and again run into the issue of deadlocks?
> 
Yes and No.

At "Set", IIRC, almost all updates against DIRTY and WRITBACK accountings
can be done under mapping->tree_lock, which disables IRQ always.
(ex. I don't mention TestSetPageWriteback but account_page_diritied().)
Then, save/restore irq flags is just a cost and no benefit, in such cases.
Of course, there are cases irqs doesn't enabled.

About FILE_MAPPED, it's not updated under mapping->tree_lock. 
So, we'll have race with charge/uncharge. We have to take lock_page_cgroup(), always.


So, I think we'll have 2 or 3 interfaces, finally.
	mem_cgroup_update_stat_fast()  // the caller must disable IRQ and lock_page()
and
	mem_cgroup_update_stat_locked() // the caller has lock_page().
and
	mem_cgroup_update_stat_safe()  // the caller don't have to do anything.

Why update_stat_fast() is for avoiding _unnecessary_ costs.
When we lock a page and disables IRQ, we don't have to do anything.
There are no races. 

But yes, it's complicated. I'd like to see what we can do.


Thanks,
-Kame


--
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/