Re: [RFC PATCH v11 00/29] KVM: guest_memfd() and per-page attributes

From: Vlastimil Babka <vbabka@suse.cz>
To: Chao Peng <chao.p.peng@linux.intel.com>,
	Sean Christopherson <seanjc@google.com>
Cc: kvm@vger.kernel.org, Fuad Tabba <tabba@google.com>,
	Yu Zhang <yu.c.zhang@linux.intel.com>,
	Vishal Annapurve <vannapurve@google.com>,
	Ackerley Tng <ackerleytng@google.com>,
	Maciej Szmigiero <mail@maciej.szmigiero.name>,
	Quentin Perret <qperret@google.com>,
	Michael Roth <michael.roth@amd.com>, Wang <wei.w.wang@intel.com>,
	Liam Merwick <liam.merwick@oracle.com>,
	Isaku Yamahata <isaku.yamahata@gmail.com>,
	"Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>,
	Paolo Bonzini <pbonzini@redhat.com>, Jorg Rodel <jroedel@suse.de>,
	Vitaly Kuznetsov <vkuznets@redhat.com>
Subject: Re: [RFC PATCH v11 00/29] KVM: guest_memfd() and per-page attributes
Date: Fri, 1 Sep 2023 10:26:51 +0200	[thread overview]
Message-ID: <030ede02-074d-98f5-ca71-5540f5b1fbb6@suse.cz> (raw)
In-Reply-To: <20230901011711.GA673287@chaop.bj.intel.com>

On 9/1/23 03:17, Chao Peng wrote:
> On Thu, Aug 31, 2023 at 11:29:22AM -0700, Sean Christopherson wrote:
>> On Tue, Aug 29, 2023, Chao Peng wrote:
>> > On Fri, Aug 25, 2023 at 10:47:12AM -0700, Sean Christopherson wrote:
>> > > 
>> > > 
>> > > Filemap vs. xarray
>> > > ------------------
>> > > This is the main item that needs attention.  I don't want to merge guest_memfd()
>> > > without doing this comparison, as not using filemap means we don't need AS_UNMOVABLE.
>> > > Arguably we could merge a filemap implementation without AS_UNMOVABLE and just eat
>> > > the suboptimal behavior, but not waiting a little while longer to do everything we
>> > > can to get this right the first time seems ridiculous after we've been working on
>> > > this for literally years.
>> > > 
>> > > Paolo was going to work on an axarray implementation, but AFAIK he hasn't done
>> > > anything yet.  We (Google) don't have anyone available to work on an xarray
>> > > implementation for several weeks (at best), so if anyone has the bandwidth and
>> > > desire to take stab at an xarray implementation, please speak up.
>> > 
>> > I can do some experiments in the following weeks on the xarray
>> > direction. I'm not quite confident I understood all what Paolo
>> > originally wanted to do, so questions may have.
>> 
>> FYI, I jumped the gun, sounds like Paolo got far enough along to form a strong
>> opinion[*].
> 
> Yeah, I see that, that is a good news actually, then we can go ahead with
> the current filemap one. I personally think these mm touchpoints are not
> a big deal when compared to previous versions, most part we are just using
> the APIs.

I also agree with the outcome, I think staying will filemap will be more
future proof e.g. if migration and swapout becomes possible in the future. I
don't think having to touch some mm/ code is that much of a problem.
Hopefully the AS_UNMOVABLE issue will be sufficiently addressed by this:
https://lore.kernel.org/all/20230901082025.20548-2-vbabka@suse.cz/

>> 
>> Thanks for volunteering though, much appreciated!
> 
> NP, any collaboration is to make this lasting years series merge earlier.
> 
> Chao
>> 
>> [*] https://lore.kernel.org/all/CABgObfay4FKV=foWLZzAWaC2kVHRnF1ib+6NC058QVZVFhGeyA@mail.gmail.com