On Tuesday, 11 of November 2008, Pavel Machek wrote:
>
> > > > It is. However, the problem is 100% reproducible on any 32-bit single-node
> > > > system with CONFIG_NUMA set, from what I can tell.
> > > >
> > > > It doesn't happen if the kernel is booted with highmem=0, so it looks like
> > > > the code that saves highmem causes the problem to happen. However, this
> > > > same code works well for all of the !CONFIG_NUMA cases and practically only
> > > > the only non-open-coded it uses is kmap_atomic().
> > > >
> > > > > One possibility would be to bisect if it ever worked?
> > > >
> > > > Not sure it did, probably not. :-(
> > >
> > > Well, interesting point would be just before this commit:
> > >
> > >
> > > commit 8357376d3df21b7d6f857931a57ac50da9c66e26
> > > tree daf2c369e9b79d24c1666323b3ae75189e482a4a
> > > parent bf73bae6ba0dc4bd4f1e570feb34a06b72725af6
> > > author Rafael J. Wysocki Wed, 06 Dec 2006 20:34:18 -0800
> > > committer Linus Torvalds Thu, 07 Dec 2006
> > > 08:39:27 -0800
> > >
> > > [PATCH] swsusp: Improve handling of highmem
> > >
> > > Currently swsusp saves the contents of highmem pages by copying
> > > them to the
> > > normal zone which is quite inefficient (eg. it requires two
> > > normal pages
> > > to be used for saving one highmem page). This may be improved by
> > > using
> > > highmem for saving the contents of saveable highmem pages.
> > >
> > > ...highmem handling was way simpler in those good old days ;-)

> >
> > Please stop kidding, this is a serious issue.
> >
> > The hibernation code _works_ with all kinds of highmem when CONFIG_NUMA is
> > unset.

>
> And it does not work with single highmem page when NUMA is set... I
> went through the highmem saving code, and it depends on highmem not
> changing from under it (right?) and is generally quite tricky ('if
> they are both in highmem do this, else if one of them is do that, else
> do something else')


It actually is quite simple, if you know the idea.

> and it changes page protections on the fly, etc.


No, it doesn't do that, at least for pages it hasn't allocated itself. I don't
think it changes anything like page protections at all, though.

> I'm not saying the bug is in that code, but before that commit we had
> very stupid --- but very robust -- code. I'll try if that one works
> with config_numa, perhaps we can get some debug info that way.


if you can do that, it actually may be valuable information.

Thanks,
Rafael
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/