Re: [Xen-devel] frame table setup for memory hotplug

["Jiang, Yunhong" <yunhong.jiang@xxxxxxxxx>]

Jan, sorry for slow response.

IIRC, the reason we do this is because in memory hotplug situation, there will 
be a very big hole between the address of the memory populated before hot-plug 
and the memory populated by hot-added memory. (i.e. the added memory started at 
very high-end address). So instead of setup the frame table for the whole 
address space, we expand the frame table dynamically after hotplug.

We have the memory hotplug environment, so if you have any patch, I'm glad to 
test it, or have my colleagues help to test it.

Thanks
--jyh

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: Saturday, December 08, 2012 12:16 AM
> To: Jiang, Yunhong
> Cc: xen-devel
> Subject: frame table setup for memory hotplug
> 
> Yunhong,
> 
> in c/s 20617:283a5357d196 you modified init_frametable() to
> populate the frame table slightly differently for the hotplug
> case. I wonder why you did that, because (apart from the bug
> already fixed, and the off-by-one bugs I'm having a fix pending
> for) I fear you didn't pay attention to the fact that using
> pdx_to_page() on something that doesn't really represent the
> PDX for a valid page may return a value not validly usable here.
> 
> Do you happen to recall what it was that caused you to do that
> adjustment in the first place? If you don't, do you have an
> environment where you would be able to test an eventual
> change of mine (effectively undoing that part of said c/s)?
> 
> Thanks, Jan
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel