Re: [PATCH] Add a page cache-backed balloon device driver.

From: Frank Swiderski <fes@google.com>
To: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>,
	riel@redhat.com, Andrea Arcangeli <aarcange@redhat.com>,
	virtualization@lists.linux-foundation.org,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
	mikew@google.com
Subject: Re: [PATCH] Add a page cache-backed balloon device driver.
Date: Wed, 27 Jun 2012 08:48:55 -0700	[thread overview]
Message-ID: <CAK+C7kWRsDcsB-W9m=Hn65xekvb-uOZC4oAMT0z48CC5q00oJw@mail.gmail.com> (raw)
In-Reply-To: <87lij91myw.fsf@rustcorp.com.au>

On Tue, Jun 26, 2012 at 7:56 PM, Rusty Russell <rusty@rustcorp.com.au> wrote:
> On Wed, 27 Jun 2012 00:41:06 +0300, "Michael S. Tsirkin" <mst@redhat.com> wrote:
>> On Tue, Jun 26, 2012 at 01:32:58PM -0700, Frank Swiderski wrote:
>> > This implementation of a virtio balloon driver uses the page cache to
>> > "store" pages that have been released to the host.  The communication
>> > (outside of target counts) is one way--the guest notifies the host when
>> > it adds a page to the page cache, allowing the host to madvise(2) with
>> > MADV_DONTNEED.  Reclaim in the guest is therefore automatic and implicit
>> > (via the regular page reclaim).  This means that inflating the balloon
>> > is similar to the existing balloon mechanism, but the deflate is
>> > different--it re-uses existing Linux kernel functionality to
>> > automatically reclaim.
>> >
>> > Signed-off-by: Frank Swiderski <fes@google.com>
>>
>> I'm pondering this:
>>
>> Should it really be a separate driver/device ID?
>> If it behaves the same from host POV, maybe it
>> should be up to the guest how to inflate/deflate
>> the balloon internally?
>
> Well, it shouldn't steal ID 10, either way :)  Either use a completely
> bogus number, or ask for an id.
>
> But AFAICT this should be a an alternate driver of for the same device:
> it's not really a separate device, is it?
>
> Cheers,
> Rusty.

Apologies, Rusty.  Asking for an ID is in the virtio spec, and I
completely neglected that step.  Though as you and others have pointed
out, this probably fits better as a different driver for the same
device.  Since it changes whether or not the deflate operation is
necessary, it also seems that how this should look is different
behavior based on a feature bit in the device.

If that sounds reasonable, then what I'll do with this patch is merge
it with the existing virtio balloon driver with a feature bit for
determining which behavior to use.

I also think the idea of a generic balloon that the different balloon
drivers use for the inflate/deflate operations is interesting and
useful, though I think the suggestion of pending that until later is
correct.

Sounds reasonable?

Regards,
fes