what's the incompatibility between hibernation and hotplug memory

what's the incompatibility between hibernation and hotplug memory

[Jeremy Fitzhardinge]

Hi,

I noticed that Kconfig warns that hibernation and hotplug memory are not 
compatible.  What's the issue?

Thanks,
    J

Re: what's the incompatibility between hibernation and hotplug memory

[Rafael J. Wysocki]

On Thursday, 27 of March 2008, Jeremy Fitzhardinge wrote:
> Hi,

Hi,

> I noticed that Kconfig warns that hibernation and hotplug memory are not 
> compatible.  What's the issue?

Generally, we might not be able to handle memory unplugged at a wrong time
(eg. after we've created the image).

Which part of Kconfig says that exactly?

Rafael

Re: what's the incompatibility between hibernation and hotplug memory

[Jeremy Fitzhardinge]

Rafael J. Wysocki wrote:
> Generally, we might not be able to handle memory unplugged at a wrong time
> (eg. after we've created the image).
>   

Hot memory unplugging is distinct from memory hotplug.

> Which part of Kconfig says that exactly?
>   

mm/Kconfig:

config MEMORY_HOTPLUG
	bool "Allow for memory hot-add"
	depends on SPARSEMEM || X86_64_ACPI_NUMA
	depends on HOTPLUG && !HIBERNATION && ARCH_ENABLE_MEMORY_HOTPLUG
	depends on (IA64 || X86 || PPC64 || SUPERH)

comment "Memory hotplug is currently incompatible with Software Suspend"
	depends on SPARSEMEM && HOTPLUG && HIBERNATION


    J

Re: what's the incompatibility between hibernation and hotplug memory

[Pavel Machek]

Hi!

> > I noticed that Kconfig warns that hibernation and hotplug memory are not 
> > compatible.  What's the issue?
> 
> Generally, we might not be able to handle memory unplugged at a wrong time
> (eg. after we've created the image).
> 
> Which part of Kconfig says that exactly?

Actually I believe it is simpler than that. We were not prepared to
deal with memory holes in hibernation code, so we just disabled it in
Kconfig...
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
pomozte zachranit klanovicky les:  http://www.ujezdskystrom.info/

Re: what's the incompatibility between hibernation and hotplug memory

[Jeremy Fitzhardinge]

Pavel Machek wrote:
> Hi!
>
>   
>>> I noticed that Kconfig warns that hibernation and hotplug memory are not 
>>> compatible.  What's the issue?
>>>       
>> Generally, we might not be able to handle memory unplugged at a wrong time
>> (eg. after we've created the image).
>>
>> Which part of Kconfig says that exactly?
>>     
>
> Actually I believe it is simpler than that. We were not prepared to
> deal with memory holes in hibernation code, so we just disabled it in
> Kconfig...

Doesn't that mean its generally incompatible with non-flatmem memory 
arrangements rather than hotplug memory specifically?

    J

Re: what's the incompatibility between hibernation and hotplug memory

[Rafael J. Wysocki]

On Thursday, 27 of March 2008, Jeremy Fitzhardinge wrote:
> Pavel Machek wrote:
> > Hi!
> >
> >   
> >>> I noticed that Kconfig warns that hibernation and hotplug memory are not 
> >>> compatible.  What's the issue?
> >>>       
> >> Generally, we might not be able to handle memory unplugged at a wrong time
> >> (eg. after we've created the image).
> >>
> >> Which part of Kconfig says that exactly?
> >>     
> >
> > Actually I believe it is simpler than that. We were not prepared to
> > deal with memory holes in hibernation code, so we just disabled it in
> > Kconfig...
> 
> Doesn't that mean its generally incompatible with non-flatmem memory 
> arrangements rather than hotplug memory specifically?

No, it should work with NUMA nowadays, actually.

Perhaps it's worth trying it with hotplug memory and see what problems arise?

Thanks,
Rafael

Re: what's the incompatibility between hibernation and hotplug?memory

[Pavel Machek]

On Thu 2008-03-27 21:55:01, Rafael J. Wysocki wrote:
> On Thursday, 27 of March 2008, Jeremy Fitzhardinge wrote:
> > Pavel Machek wrote:
> > > Hi!
> > >
> > >   
> > >>> I noticed that Kconfig warns that hibernation and hotplug memory are not 
> > >>> compatible.  What's the issue?
> > >>>       
> > >> Generally, we might not be able to handle memory unplugged at a wrong time
> > >> (eg. after we've created the image).
> > >>
> > >> Which part of Kconfig says that exactly?
> > >>     
> > >
> > > Actually I believe it is simpler than that. We were not prepared to
> > > deal with memory holes in hibernation code, so we just disabled it in
> > > Kconfig...
> > 
> > Doesn't that mean its generally incompatible with non-flatmem memory 
> > arrangements rather than hotplug memory specifically?
> 
> No, it should work with NUMA nowadays, actually.
> 
> Perhaps it's worth trying it with hotplug memory and see what problems arise?

Good. I believe it is good chance it will work with hotplug, too... as
long as memory is not inserted/removed during suspend/resume.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
pomozte zachranit klanovicky les:  http://www.ujezdskystrom.info/

Re: what's the incompatibility between hibernation and hotplug?memory

[Rafael J. Wysocki]

On Friday, 28 of March 2008, Pavel Machek wrote:
> On Thu 2008-03-27 21:55:01, Rafael J. Wysocki wrote:
> > On Thursday, 27 of March 2008, Jeremy Fitzhardinge wrote:
> > > Pavel Machek wrote:
> > > > Hi!
> > > >
> > > >   
> > > >>> I noticed that Kconfig warns that hibernation and hotplug memory are not 
> > > >>> compatible.  What's the issue?
> > > >>>       
> > > >> Generally, we might not be able to handle memory unplugged at a wrong time
> > > >> (eg. after we've created the image).
> > > >>
> > > >> Which part of Kconfig says that exactly?
> > > >>     
> > > >
> > > > Actually I believe it is simpler than that. We were not prepared to
> > > > deal with memory holes in hibernation code, so we just disabled it in
> > > > Kconfig...
> > > 
> > > Doesn't that mean its generally incompatible with non-flatmem memory 
> > > arrangements rather than hotplug memory specifically?
> > 
> > No, it should work with NUMA nowadays, actually.
> > 
> > Perhaps it's worth trying it with hotplug memory and see what problems arise?
> 
> Good. I believe it is good chance it will work with hotplug, too... as
> long as memory is not inserted/removed during suspend/resume.

Well, there may be a problem if it's inserted after we've created the image
and before powering off, because the resumed kernel will not have any idea of
the new memory.  It probably will require us to add some resume callbacks
for the memory hotplugging to handle this cleanly.

Thanks,
Rafael

Re: what's the incompatibility between hibernation and hotplug?memory

[Pavel Machek]

On Fri 2008-03-28 00:44:00, Rafael J. Wysocki wrote:
> On Friday, 28 of March 2008, Pavel Machek wrote:
> > On Thu 2008-03-27 21:55:01, Rafael J. Wysocki wrote:
> > > On Thursday, 27 of March 2008, Jeremy Fitzhardinge wrote:
> > > > Pavel Machek wrote:
> > > > > Hi!
> > > > >
> > > > >   
> > > > >>> I noticed that Kconfig warns that hibernation and hotplug memory are not 
> > > > >>> compatible.  What's the issue?
> > > > >>>       
> > > > >> Generally, we might not be able to handle memory unplugged at a wrong time
> > > > >> (eg. after we've created the image).
> > > > >>
> > > > >> Which part of Kconfig says that exactly?
> > > > >>     
> > > > >
> > > > > Actually I believe it is simpler than that. We were not prepared to
> > > > > deal with memory holes in hibernation code, so we just disabled it in
> > > > > Kconfig...
> > > > 
> > > > Doesn't that mean its generally incompatible with non-flatmem memory 
> > > > arrangements rather than hotplug memory specifically?
> > > 
> > > No, it should work with NUMA nowadays, actually.
> > > 
> > > Perhaps it's worth trying it with hotplug memory and see what problems arise?
> > 
> > Good. I believe it is good chance it will work with hotplug, too... as
> > long as memory is not inserted/removed during suspend/resume.
> 
> Well, there may be a problem if it's inserted after we've created the image
> and before powering off, because the resumed kernel will not have any idea of
> the new memory.  It probably will require us to add some resume callbacks
> for the memory hotplugging to handle this cleanly.

Well, "not having any idea of new memory" should not lead to big
problems. 

"Not being aware that memory was removed" would be worse ;-).

And I think that we have the ammount of memory check, and will refuse
the resume which is good failsafe behaviour.

...so someone interested should just test it ;-).
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html