Re: [PATCH] hibernation should work ok with memory hotplug

[Rafael J. Wysocki]

On Monday, 3 of November 2008, Dave Hansen wrote:
> On Mon, 2008-11-03 at 12:51 -0800, Andrew Morton wrote:
> > On Wed, 29 Oct 2008 13:25:00 +0100
> > "Rafael J. Wysocki" <rjw@xxxxxxx> wrote:
> > > On Wednesday, 29 of October 2008, Pavel Machek wrote:
> > > > 
> > > > hibernation + memory hotplug was disabled in kconfig because we could
> > > > not handle hibernation + sparse mem at some point. It seems to work
> > > > now, so I guess we can enable it.
> > > 
> > > OK, if "it seems to work now" means that it has been tested and confirmed to
> > > work, no objection from me.
> > 
> > yes, that was not a terribly confidence-inspiring commit message.
> > 
> > 3947be1969a9ce455ec30f60ef51efb10e4323d1 said "For now, disable memory
> > hotplug when swsusp is enabled.  There's a lot of churn there right
> > now.  We'll fix it up properly once it calms down." which is also
> > rather rubbery.  
> > 
> > Cough up, guys: what was the issue with memory hotplug and swsusp, and
> > is it indeed now fixed?
> 
> I suck.  That commit message was horrid and I'm racking my brain now to
> remember what I meant.  Don't end up like me, kids.
> 
> I've attached the message that I sent to the swsusp folks.  I never got
> a reply from that as far as I can tell.
> 
> http://sourceforge.net/mailarchive/forum.php?thread_name=1118682535.22631.22.camel%40localhost&forum_name=lhms-devel
> 
> As I look at it now, it hasn't improved much since 2005.  Take a look at
> kernel/power/snapshot.c::copy_data_pages().  It still assumes that the
> list of zones that a system has is static.  Memory hotplug needs to be
> excluded while that operation is going on.

This operation is carried out on one CPU with interrupts disabled.  Is that
not enough?

> page_is_saveable() checks for pfn_valid().  But, with memory hotplug,
> things can become invalid at any time since no references are held or
> taken on the page.  Or, a page that *was* invalid may become valid and
> get missed.

Can that really happen given the conditions above?

> The "missing a page" thing is probably correctable via the
> zone_span_seqbegin() locks.  The "page becoming invalid" thing is
> probably mostly fixable by acquiring a reference to the page itself.
> I'd need to look how the locking on the hot remove side is working these
> days to be much more constructive than that.

Well, I don't think any locking is necessary for the image creation.

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