LKML Archive on
help / color / mirror / Atom feed
From: Peter Zijlstra <>
To: Andrew Morton <>
Cc: Miklos Szeredi <>,,
Subject: Re: [patch 2/3] only allow nonlinear vmas for ram backed filesystems
Date: Mon, 26 Mar 2007 08:57:41 +0200	[thread overview]
Message-ID: <1174892261.6792.9.camel@twins> (raw)
In-Reply-To: <>

On Sun, 2007-03-25 at 16:00 -0800, Andrew Morton wrote:
> On Sat, 24 Mar 2007 23:09:19 +0100 Miklos Szeredi <> wrote:
> > Dirty page accounting/limiting doesn't work for nonlinear mappings,
> Doesn't it?  iirc the problem is that we don't correctly re-clean the ptes
> while starting writeout.  And the dirty-page accounting is in fact correct
> (it'd darn well better be).

If we do not re-protect the pages on writeout, we'll decrement the dirty
count but not get a fault on re-dirty. Hence the dirty count will
actually skew.

In order to make page_mkclean() work for nonlinear vmas we need to do a
full pte scan for each invocation (we could perhaps only scan 1 in n
times to try and limit the damage) and that hurts. This will basically
render it useless.

The other solution is adding rmap information to nonlinear vmas but
doubling the memory overhead for nonlinear mappings was not deemed a
good idea.

> > so
> > for non-ram backed filesystems emulate with linear mappings.  This
> > retains ABI compatibility with previous kernels at minimal code cost.
> > 
> > All known users of nonlinear mappings actually use tmpfs, so this
> > shouldn't have any negative effect.
> Unless someone is using remap_file_pages() against an ext3 file, in which
> case their application stops working?

it'll work up to a certain point (when you hit the max vma count). When
this issue first came up a few weeks ago, nobody knew of any non tmpfs

> That would be a problem.  These guys:
>, for example, will be in
> for a little surprise.

Its an example, it doesn't show if there are actually users of this, but
point taken. We actually could make this very example work by noting
that they map PROT_READ only.

  reply	other threads:[~2007-03-26  6:57 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-03-24 22:07 [patch 1/3] split mmap Miklos Szeredi
2007-03-24 22:09 ` [patch 2/3] only allow nonlinear vmas for ram backed filesystems Miklos Szeredi
2007-03-25 12:12   ` Peter Zijlstra
2007-03-25 15:51     ` Matt Mackall
2007-03-27  0:02       ` William Lee Irwin III
2007-03-26  0:00   ` Andrew Morton
2007-03-26  6:57     ` Peter Zijlstra [this message]
2007-03-24 22:11 ` [patch 3/3] update ctime and mtime for mmaped write Miklos Szeredi
2007-03-25 12:12   ` Peter Zijlstra
2007-03-25 21:08     ` Miklos Szeredi
2007-03-25 12:12 ` [patch 1/3] split mmap Peter Zijlstra

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:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1174892261.6792.9.camel@twins \ \ \ \ \ \
    --subject='Re: [patch 2/3] only allow nonlinear vmas for ram backed filesystems' \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).