Reid Sutherland <[email protected]> writes:

> On 2026-07-27 3:47 p.m., Sam James wrote:
>> Reid Sutherland <[email protected]> writes:
>>
>>> On 2026-07-27 12:25 p.m., Solar Designer wrote:
>>>> On Mon, Jul 27, 2026 at 09:46:12AM -0400, Reid Sutherland wrote:
>>>>> For your information.
>>>>>
>>>>> https://git.kernel.org/pub/scm/virt/kvm/kvm.git/commit/?id=a204badd8432f93b7e862e7dac6db0fe3d65f370
>>>> Thanks, but can you please explain why exactly you think this is
>>>> noteworthy for oss-security?
>>>>
>>>> I see there's a recent Phoronix story:
>>>>
>>>> https://www.phoronix.com/news/KVM-Chainsaw-Linux-7.3
>>>>
>>>> and the patch series had been tracked and archived by LWN.
>>>>
>>>> This is definitely noteworthy for KVM project development, but even
>>>> seeing all those other resources, I do not see why bring this in here?
>>>
>>> Core changes to virtualization should receive examination. Changing
>>> data structures around because they're too big.. it's concerning,
>>> following that patch is tricky because of its size. Would feel better
>>> about it if changes like this were passed through a reliable LLM, but
>>> even that requires relevant experience.  And naming it chainsaw, what
>>> purpose does that serve?  I have very little trust.
>> Paolo has been a maintainer for a long time and in FOSS even longer than
>> that. Big refactoring can introduce bugs but I've no reason to doubt him
>> and I don't really see why this change is any different from any other
>> refactoring.
>
>
> The refactor doesn't seem necessary, the original data structure
> appears to be fine, now we have surface risk for something that
> doesn't solve a bug or implement a feature.
>
> I don't care what someone's street cred is, it's their result.
>
>>>
>>>> In general, posting a link without explanation is inappropriate here.
>>>
>>> The 16+ year local root vulnerabilities in the kernel is enough of a
>>> bruising to be extra cautious.
>> This could be used to post all sorts of large kernel changes here,
>> though, and I don't think we want to do that.
>>
>> Now, with regard to what you said above: I don't particularly want to
>> encourage people to just put things into an LLM and see what they say,
>> but if you were to ask an LLM, find it had something useful to say (and
>> you're qualified to assess that), then sharing it here might be
>> appropriate depending on context.
>
>
> Again, steering an LLM to properly search for bugs in C is for the
> experienced.  If you have a problem with LLMs, you're fighting the
> calculator.  It's very clear that LLMs are finding bugs that have gone
> undiscovered for almost two decades.

I think speculation to this degree is offtopic for the list.

Attachment: signature.asc
Description: PGP signature

Reply via email to