Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH 1/5] mm: Add support for unaccepted memory
To:     Joerg Roedel <jroedel@suse.de>
Cc:     Dave Hansen <dave.hansen@intel.com>,
        Andi Kleen <ak@linux.intel.com>,
        "Kirill A. Shutemov" <kirill@shutemov.name>,
        Borislav Petkov <bp@alien8.de>,
        Andy Lutomirski <luto@kernel.org>,
        Sean Christopherson <seanjc@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Kuppuswamy Sathyanarayanan 
        <sathyanarayanan.kuppuswamy@linux.intel.com>,
        David Rientjes <rientjes@google.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Tom Lendacky <thomas.lendacky@amd.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Ingo Molnar <mingo@redhat.com>,
        Varad Gautam <varad.gautam@suse.com>,
        Dario Faggioli <dfaggioli@suse.com>, x86@kernel.org,
        linux-mm@kvack.org, linux-coco@lists.linux.dev,
        linux-kernel@vger.kernel.org,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
References: <d091b333-9ef8-ac32-58c5-c325d29f26d7@intel.com>
 <9748c07c-4e59-89d0-f425-c57f778d1b42@linux.intel.com>
 <17b6a3a3-bd7d-f57e-8762-96258b16247a@intel.com>
 <796a4b20-7fa3-3086-efa0-2f728f35ae06@linux.intel.com>
 <f445da8b-c044-3765-65f2-f911dbf6a507@intel.com>
 <3caf5e73-c104-0057-680c-7851476e67ac@linux.intel.com>
 <YRTZr/Pxyb8fsV58@suse.de> <25312492-5d67-e5b0-1a51-b6880f45a550@intel.com>
 <YRaGch+D5HPtUk9l@suse.de> <b0228048-d2c2-6081-cfae-147e17acf785@redhat.com>
 <YR4qexszQ+Wxqe5r@suse.de>
From:   David Hildenbrand <david@redhat.com>
Organization: Red Hat
Message-ID: <39d09b80-f79e-0dab-2be7-31d4db43ab77@redhat.com>
Date:   Thu, 19 Aug 2021 12:06:40 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <YR4qexszQ+Wxqe5r@suse.de>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 19.08.21 11:55, Joerg Roedel wrote:
> Hi David,
> 
> On Tue, Aug 17, 2021 at 05:00:55PM +0200, David Hildenbrand wrote:
>> Not sure if already discussed, but what about making sure that free pages
>> are not a mixture (partially unaccepted, partially accepted).
>>
>> You'd have to expose the pages in that granularity to the buddy
>> (__free_pages_core), indicating the state. You'd have to reject merging
>> pages of differing acceptance state.
>>
>> Accepting a page would then be handled outside of the zone lock, completely
>> controlled by the state.
>>
>> So a page in the buddy would either be completely accepted or completely
>> unaccepted, signaled e.g., by PageOffline().
>>
>> Consequently, when allocating a 4KiB page, you'd split an unaccepted 2MiB
>> page into separate unaccepted pages. You'd grab one of the unaccepted 4KiB
>> pages and accept it before initializing it and handing it out.
> 
> Yes, that is the alternative to over-accepting memory on allocation. But
> the problem here is that accepting/validating memory is an expensive
> operation which also requires a hypercall. The hypercalls on SNP and TDX
> can accept/validate multiple pages in one call. So the recommendation is
> to accept memory in bigger chunks, like the 2MB that have been proposed.
> 

The general idea would be to have one thread scanning the free page list 
and accepting pages in the background. Take a page out, accept it, 
release it back to the buddy. If we place accepted pages to the front of 
the free list, allocations would be quite fast.

Sure, you'd have some slow early allocations, but I'd guess it really 
wouldn't make a big impact overall. We'd have to try.

It's quite similar to the free page reporting infrastructure, except 
that with free page reporting we want to place pages to the tail, not to 
the front. That would be easy to add.


> Only accepting memory in allocation size might be too slow, as there is
> a lot of code doing order-0 allocations. I think this approach will also
> be more intrusive to the page alloctor, as it needs more changes on the
> free path to check for acceptance states before pages can be merged.

That's already mostly done in this patch IIRC.

-- 
Thanks,

David / dhildenb