[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <alpine.LNX.2.00.1307181254520.1047@eggly.anvils>
Date: Thu, 18 Jul 2013 12:59:10 -0700 (PDT)
From: Hugh Dickins <hughd@...gle.com>
To: Dave Jones <davej@...hat.com>
cc: Dave Hansen <dave@...1.net>,
Andrew Morton <akpm@...ux-foundation.org>,
"Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
Andrea Arcangeli <aarcange@...hat.com>,
Al Viro <viro@...iv.linux.org.uk>,
Wu Fengguang <fengguang.wu@...el.com>, Jan Kara <jack@...e.cz>,
Mel Gorman <mgorman@...e.de>, linux-mm@...ck.org,
Andi Kleen <ak@...ux.intel.com>,
Matthew Wilcox <willy@...ux.intel.com>,
"Kirill A. Shutemov" <kirill@...temov.name>,
Hillf Danton <dhillf@...il.com>, linux-fsdevel@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 5/8] thp, mm: locking tail page is a bug
On Wed, 17 Jul 2013, Dave Jones wrote:
> On Wed, Jul 17, 2013 at 05:58:13PM -0700, Hugh Dickins wrote:
> > On Wed, 17 Jul 2013, Dave Hansen wrote:
> > > On 07/17/2013 02:09 PM, Andrew Morton wrote:
> > > > lock_page() is a pretty commonly called function, and I assume quite a
> > > > lot of people run with CONFIG_DEBUG_VM=y.
> > > >
> > > > Is the overhead added by this patch really worthwhile?
> > >
> > > I always thought of it as a developer-only thing. I don't think any of
> > > the big distros turn it on by default.
> >
> > That's how I think of it too (and the problem is often that too few mm
> > developers turn it on); but Dave Jones did confirm last November that
> > Fedora turns it on.
> >
> > I believe Fedora turns it on to help us all, and wouldn't mind a mere
> > VM_BUG_ON(PageTail(page)) in __lock_page() if it's helpful to Kirill.
> >
> > But if VM_BUG_ONs become expensive, I do think it's for Fedora to
> > turn off CONFIG_DEBUG_VM, rather than for mm developers to avoid it.
>
> I'm ambivalent about whether we keep it on or off, we have no shortage
> of bugs to fix already, though I think as mentioned above, very few people
> actually enable it, so we're going to lose a lot of testing.
>
> Another idea, perhaps is an extra config option for more expensive debug options ?
I can't think of any expensive debug under CONFIG_DEBUG_VM at present,
just a little overhead as you'd expect. I don't think checking a page
flag or two in __lock_page() will change that much.
We do tend to make specific debug options for the more expensive ones,
rather than a generic heavy debug option. For example CONFIG_DEBUG_VM_RB,
which checks an mm's entire vma tree whenever a change is made there.
We could group the heavy debug options, like that and CONFIG_PROVE_LOCKING,
into a menu of their own; but I've no appetite for such a change myself!
Hugh
--
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/
Powered by blists - more mailing lists