On 13.04.2018 19:11, Matthew Wilcox wrote:
> On Fri, Apr 13, 2018 at 03:16:26PM +0200, David Hildenbrand wrote:
>> online_pages()/offline_pages() theoretically allows us to work on
>> sub-section sizes. This is especially relevant in the context of
>> virtualization. It e.g. allows us to add/remove memory to Linux in a VM in
>> 4MB chunks.
>> While the whole section is marked as online/offline, we have to know
>> the state of each page. E.g. to not read memory that is not online
>> during kexec() or to properly mark a section as offline as soon as all
>> contained pages are offline.
> Can you not use PG_reserved for this purpose?
>> + * PG_offline indicates that a page is offline and the backing storage
>> + * might already have been removed (virtualization). Don't touch!
> * PG_reserved is set for special pages, which can never be swapped out. Some
> * of them might not even exist...
> They seem pretty congruent to me.
Can we really go ahead and make dump tools exclude any PG_reserved page
from a memory dump? While it might be true for ballooned pages, I doubt
that this assumption holds in general. ("cannot be swapped out" doesn't
imply "content should never be read/dumped")
I need PG_offline right now for two reasons:
1. Make kdump skip these pages (like PG_hwpoison), because they might
not even be readable anymore as the hypervisor might have restricted
memory access completely.
2. Detect when all pages of a memory section are offline, so we can mark
the section as offline and eventually remove it.
A clear point speaking against using PG_reserved for 2. is the following
Let's assume we use virtio-balloon and inflated some chunk of memory in
a section (let's say 4MB). Now we offline (using the new driver) all
other chunks in a section, except the memory allocated by
virtio-balloon. We would suddenly mark the section as offline and
eventually remove it. This is of course very bad.
I think using PG_reserved for 1. is wrong. PG_reserved is usually used
for pages _after_ coming from an allocator. Using PG_reserved for 2.
will not work.
An ugly way for 2. would be, remembering for each section which pages
are actually online, but I would like to avoid that, especially as it
only solves part of a problem.
David / dhildenb