Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
Message-ID: <de4b831a-6649-2462-e4cf-fcefa6e912be@gmail.com>
Date:   Thu, 14 Sep 2023 11:24:35 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.15.1
Subject: Re: [PATCH v4] KVM: x86/tsc: Don't sync user changes to TSC with
 KVM-initiated change
To:     Sean Christopherson <seanjc@google.com>,
        David Woodhouse <dwmw2@infradead.org>
Cc:     Paolo Bonzini <pbonzini@redhat.com>,
        Oliver Upton <oliver.upton@linux.dev>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20230801034524.64007-1-likexu@tencent.com>
 <ZNa9QyRmuAjNAonC@google.com>
 <055482bec09cae1ea56f979893c6b67e9d6b26a2.camel@infradead.org>
 <ZQHMM8/7xXReZHdD@google.com>
 <b83a52bf72b951e69d3df23fff144899b0d6c11d.camel@infradead.org>
 <ZQHSJ9Epx1oNTZGE@google.com>
 <814044c62f8804866a6dc4c523797c06d73c82f1.camel@infradead.org>
Content-Language: en-US
From:   Like Xu <like.xu.linux@gmail.com>
In-Reply-To: <814044c62f8804866a6dc4c523797c06d73c82f1.camel@infradead.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 13/9/2023 11:24 pm, David Woodhouse wrote:
> On Wed, 2023-09-13 at 15:15 +0000, Sean Christopherson wrote:
>>>> e.g. if userspace writes '0' immediately after creating, and then later writes a
>>>> small delta, the v6 code wouldn't trigger synchronization because "user_set_tsc"
>>>> would be left unseft by the write of '0'.
>>>
>>> True, but that's the existing behaviour,
>>
>> No?  The existing code will fall into the "conditionally sync" logic for any
>> non-zero value.
> 
> Yeah, OK. This isn't one of the cases we set out to deliberately
> change, but it would be changed by v6 of the patch, and I suppose
> you're right that we should accept a small amount of extra code
> complexity just to avoid making any changes we don't *need* to, even
> for stupid cases like this.
> 
> 
>> I don't care (in the Tommy Lee Jones[*] sense).  All I care about is minimizing
>> the probability of breaking userspace, which means making the smallest possible
>> change to KVM's ABI.  For me, whether or not userspace is doing something sensible
>> doesn't factor into that equation.
> 
> Ack.

If we combine the v5 code diff (the u64 *user_value proposal) with the refined 
changelog in v6,
it seems like we've reached a point of equilibrium on this issue, doesn't it ?

Please let me know you have more concerns.

> 
> Although there's a strong argument that adding further warts to an
> already fundamentally broken API probably isn't a great idea in the
> first place. Just deprecate it and use the saner replacement API...
> which I just realised we don't have (qv). Ooops :)
>