linux-kernel - Re: [PATCH 00/17] per device dirty throttling -v7

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:	Tue, 17 Jul 2007 12:10:52 +0200
From:	Miklos Szeredi <miklos@...redi.hu>
To:	a.p.zijlstra@...llo.nl
CC:	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	miklos@...redi.hu, akpm@...ux-foundation.org, neilb@...e.de,
	dgc@....com, tomoki.sekiyama.qu@...achi.com,
	a.p.zijlstra@...llo.nl, nikita@...sterfs.com,
	trond.myklebust@....uio.no, yingchao.zhou@...il.com, andrea@...e.de
Subject: Re: [PATCH 00/17] per device dirty throttling -v7

> Latest version of the per bdi dirty throttling patches.
> 
> Most of the changes since last time are little cleanups and more
> detail in the split out of the floating proportion into their
> own little lib.
> 
> Patches are against 2.6.22-rc4-mm2
> 
> A rollup of all this against 2.6.21 is available here:
>   http://programming.kicks-ass.net/kernel-patches/balance_dirty_pages/2.6.21-per_bdi_dirty_pages.patch
> 
> This patch-set passes the starve an USB stick test..

I've done some testing of several problem cases.

1) fuse writable mmap patches + bash_shared_mapping
2) writes in a setup involving a loop dev
  a) ext3 over loop over ext3
  b) ext3 over loop over fuse-passthrough over ext3
  c) ext3 over loop over ntfs-3g

Without the patch, in all the cases I've seen deadlocks or long
stalls.  With the patch, I could not reproduce this in any of the
cases.  As predicted, the patch is performing well in this respect :)

2a is the simplest to reproduce (2.6.22, dual core, 1GB ram)

  dd if=/dev/zero of=/tmp/p5 bs=1M seek=4999 count=1
  mkfs.ext3 -F /tmp/p5
  mkdir /tmp/m5
  mount -oloop /tmp/p5 /tmp/m5
  dd if=/dev/zero of=/tmp/m5/foo bs=1M count=4000

The second dd can stall for indefinite amounts of time.  Kicking it
with sync can get it moving, but it relapses after some time.

Even with the per-device-throttling patch, case 2 shows an nr_dirty
elevated far above the 10% limit, reaching 40% or higher.  I believe,
this is due to a missing balance_dirty_pages() call in the loop
device.  And indeed the anomaly can be solved by adding this patch:

  http://lkml.org/lkml/2007/3/24/101

Miklos
-
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/