Re: Bug in kernel 2.6.31, Slow wb_kupdate writeout

[Wu Fengguang <fengguang.wu@intel.com>]

On Thu, Jul 30, 2009 at 10:59:09AM +0800, Martin Bligh wrote:
> On Wed, Jul 29, 2009 at 6:57 PM, Wu Fengguang<fengguang.wu@intel.com> wrote:
> > On Thu, Jul 30, 2009 at 09:12:26AM +0800, Martin Bligh wrote:
> >> > I agree on the unification of kupdate and sync paths. In fact I had a
> >> > patch for doing this. And I'd recommend to do it in two patches:
> >> > one to fix the congestion case, another to do the code unification.
> >> >
> >> > The sync path don't care whether requeue_io() or redirty_tail() is
> >> > used, because they disregard the time stamps totally - only order of
> >> > inodes matters (ie. starvation), which is same for requeue_io()/redirty_tail().
> >>
> >> But, as I understand it, both paths share the same lists, so we still have
> >> to be consistent?
> >
> > Then let's first unify the code, then fix the congestion case? :)
> 
> OK, I will send it out as separate patches. I am just finishing up the testing
> first.

Note that this is a simple fix that may have suboptimal write performance.
Here is an old reasoning:

        http://lkml.org/lkml/2009/3/28/235

Thanks,
Fengguang