Re: pdflush stuck in D state with v2.6.24-rc1-192-gef49c32

[Fengguang Wu <wfg@mail.ustc.edu.cn>]

On Wed, Oct 31, 2007 at 12:53:18PM -0500, Florin Iucha wrote:
> On Wed, Oct 31, 2007 at 07:16:06AM -0500, Florin Iucha wrote:
> > On Wed, Oct 31, 2007 at 02:53:25PM +0800, Fengguang Wu wrote:
> > > On Tue, Oct 30, 2007 at 10:52:45PM -0500, Florin Iucha wrote:
> > > > On Tue, Oct 30, 2007 at 07:02:42PM -0500, Florin Iucha wrote:
> > > > > I have added the patches and started a linux kernel compilation, and
> > > > > something really interesting happens.  I run the build with the
> > > > > equivalent of "make -j3" and in a separate console I am watching the
> > > > > build with 'top'.  The build consumes 98% of both CPUs.  If I stop the
> > > > > output in the build console with "Ctrl-S", one core goes to idle,
> > > > > while the other is in 50% waiting, then goes to 75% waiting.  When I
> > > > > resume the build with "Ctrl-Q", the build starts to use both CPUs at
> > > > > 98-99%.  The NFS4 use was minimal, as I did not login with Gnome, but
> > > > > just logged on the console.  Also, the CPU that is in 75% waiting
> > > > > state changes occasionally.  'Top' shows pdflush in D state, using
> > > > > 5-6% of CPU.
> > > > 
> > > > I forgot the traces:
> > > > 
> > > >    http://iucha.net/2.6.24-rc1/fw.1.gz
> > > >    http://iucha.net/2.6.24-rc1/fw.2.gz
> > > >    http://iucha.net/2.6.24-rc1/fw.3.gz
> > > 
> > > Sorry for the delay - I've been fixing our server today.
> > > 
> > > [  263.685691] mm/page-writeback.c 655 wb_kupdate: pdflush(248) 24235 global 4593 0 0 wc _M tw 1024 sk 0
> > > [  263.789648] requeue_io 301: inode 4031199 size 562 at 08:07(sda7)
> > > [  263.789656] requeue_io 301: inode 4031231 size 329 at 08:07(sda7)
> > > [  263.789660] requeue_io 301: inode 4031255 size 177 at 08:07(sda7)
> > > [  263.789664] requeue_io 301: inode 4031268 size 94 at 08:07(sda7)
> > > [  263.789667] requeue_io 301: inode 4031329 size 88 at 08:07(sda7)
> > > [  263.789671] requeue_io 301: inode 4031351 size 74 at 08:07(sda7)
> > > [  263.789674] requeue_io 301: inode 4031408 size 175 at 08:07(sda7)
> > > [  263.789678] requeue_io 301: inode 4031413 size 129 at 08:07(sda7)
> > > [  263.789681] requeue_io 301: inode 4031415 size 391 at 08:07(sda7)
> > > [  263.789690] mm/page-writeback.c 655 wb_kupdate: pdflush(248) 24235 global 4593 0 0 wc _M tw 1024 sk 0
> > > [  263.890184] requeue_io 301: inode 4031199 size 562 at 08:07(sda7)
> > > [  263.890191] requeue_io 301: inode 4031231 size 329 at 08:07(sda7)
> > > [  263.890195] requeue_io 301: inode 4031255 size 177 at 08:07(sda7)
> > > [  263.890198] requeue_io 301: inode 4031268 size 94 at 08:07(sda7)
> > > [  263.890202] requeue_io 301: inode 4031329 size 88 at 08:07(sda7)
> > > [  263.890205] requeue_io 301: inode 4031351 size 74 at 08:07(sda7)
> > > [  263.890208] requeue_io 301: inode 4031408 size 175 at 08:07(sda7)
> > > [  263.890212] requeue_io 301: inode 4031413 size 129 at 08:07(sda7)
> > > [  263.890215] requeue_io 301: inode 4031415 size 391 at 08:07(sda7)
> > > [  263.890223] mm/page-writeback.c 655 wb_kupdate: pdflush(248) 24235 global 4593 0 0 wc _M tw 1024 sk 0
> > > 
> > > It's about sda7, not NFSv4.
> > > 
> > > Is it a Reiserfs? We have a fresh fix for it: http://lkml.org/lkml/2007/10/23/93
> > 
> > Yes, it is a Reiserfs.  Incidentally it is the partition that holds
> > the kernel sources and build directory.  The message states that the
> > same bug exists in 2.6.23 but I do not see the same behavior in
> > 2.6.23.  Anyway, I will apply the patch and see what I get.
> 
> Fengguang,
> 
> This patch does not fix anything for me.  Even such light use of the
> reiserfs filesystem as pulling the linux-2.6 git tree updates caused
> one CPU to go to 75% iowait.
 
Thank you, Florin. Could you provide more details about sda7, such as
the mount option and output of `reiserfstune /dev/sda7`? I'll try to
reproduce it before asking for your help.

Fengguang