Re: [2.6.18-rc7-mm1] slow boot

[Andrew Morton <akpm@osdl.org>]

On Wed, 27 Sep 2006 01:48:12 +0200
Tilman Schmidt <tilman@imap.cc> wrote:

> On 24.09.2006 23:53, Andrew Morton wrote:
> > make-ext3-mount-default-to-barrier=1.patch takes my laptop's bootup time
> > from 53 seconds to 68, which is rather painful.  In fact I'm inclined to
> > drop the patch because of this, and I'd also be quite concerned about the
> > similar reiserfs patch, make-reiserfs-default-to-barrier=flush.patch.
> [...]
> > Do you have the time to go through the
> > http://www.zip.com.au/~akpm/linux/patches/stuff/bisecting-mm-trees.txt
> > process?
> 
> Ok, so far I've narrowed it down to the section between
> #X64_64-START
> and
> #X64_64-END

argh.

> which I guess lets make-{ext3-mount,reiserfs}-default-to-barrier=1.patch
> off the hook for now.
> 
> Trying to bisect further into that section now,

Thanks.  You may find that none of it compiles, and you'll need to take the
four or five patches immediately after #X64_64-END (ie: fixes against the
x86_64 tree) and place them at the appropriate places immediately after the
x86_64-mm-<whatever>.patch which they fix.

Specifically, put fix-x86_64-mm-i386-pda-smp-processorid.patch immediately
after x86_64-mm-i386-pda-smp-processorid.patch and put
fix-x86_64-mm-spinlock-cleanup.patch immediately after
x86_64-mm-spinlock-cleanup.patch.

> but perhaps that'll
> already trigger some thoughts?

Nope, there's a huge amount of stuff in there.  And it's pretty much all in
mainline as of a couple of hours ago, so bisecting the tree which you have
there is increasingly valuable.

Thanks for persisting with this.