Messages in this thread | | | Date | Mon, 05 Aug 2013 21:38:25 +0800 | From | Tang Chen <> | Subject | Re: [PATCH v2 00/18] Arrange hotpluggable memory as ZONE_MOVABLE. |
| |
Hi hpa,
I'm sorry but I don't quite following it.
On 08/05/2013 09:07 PM, H. Peter Anvin wrote: > On 08/01/2013 12:06 AM, Tang Chen wrote: >> This patch-set aims to solve some problems at system boot time >> to enhance memory hotplug functionality. >> >> [Background] >> >> The Linux kernel cannot migrate pages used by the kernel because >> of the kernel direct mapping. Since va = pa + PAGE_OFFSET, if the >> physical address is changed, we cannot simply update the kernel >> pagetable. On the contrary, we have to update all the pointers >> pointing to the virtual address, which is very difficult to do. >> > > It does beg the question if that "since" statement should be changed ...
Do you mean we are going to do kernel page migration in the future ?
> we already have it handled differently on Xen PV, but that is kind of > "special". There are a whole bunch of other issues with moving kernel > memory around: you have to worry what might have a physical address > cached somewhere and what might be in active use and so on...
The current solution is to hotplug ZONE_MOVABLE, which the kernel won't use. So do you mean if I want to do kernel page migration (which I'm not doing), I need to worry about what you said above ?
>I am not > really suggesting it as anything but food for thought at this time.
Sorry for my poor English, and I really cannot understand this one.
Thanks.
| |