On Tue, May 19, 2026 at 05:11:23PM +0100, Daniel P. Berrangé wrote:
> On Tue, May 19, 2026 at 11:18:31AM -0400, Michael S. Tsirkin wrote:
> > On Tue, May 19, 2026 at 03:26:51PM +0100, Daniel P. Berrangé wrote:
> > > The qemu-security mailing list was created several years back now and
> > > traditionally saw 1-2 disclosures a month at worst. This was manageable.
> > > 
> > > Since approx March 1st, the new normal is to see as many as 20 disclosures
> > > in one single day, more than 200 in total now. This is unsustainable.
> > > I was thinking we needed more people on qemu-security to triage, but IMHO
> > > this won't really fix the problem.
> > > 
> > > This needs an issue tracker to cope with & email is not an issue tracker.
> > > We faked an issue tracker with a shared spreadsheet to prevent us drowning
> > > these past few months, but this is still not sustainable & probably won't
> > > ever be.
> > > 
> > > The traditional POV was that security disclosures must be dealt with on
> > > a strict "need to know" basis until there was a concious decision to make
> > > them public. Typically that happened when a patch was published by the
> > > maintainer on qemu-devel. Rarely have we applied an embargo period after
> > > a maintainer had a patch ready to post, as most bugs were not severe
> > > enough. Also in typical usage Libvirt has SELinux/AppArmour to mitigate
> > > impact of a guest ecsape into QEMU.
> > > 
> > > QEMU is not alone in this chaos, to quote Linus
> > > 
> > >    https://lkml.org/lkml/2026/5/17/896
> > > 
> > > [quote]
> > >   the continued flood of AI reports has basically made the security list
> > >   almost entirely unmanageable, with enormous duplication due to
> > >   different people finding the same things with the same tools. People
> > >   spend all their time just forwarding things to the right people or
> > >   saying "that was already fixed a week/month ago" and pointing to the
> > >   public discussion.
> > > [/quote]
> > > 
> > > They have explicitly changed their policy for disclosure now:
> > > 
> > >   
> > > https://github.com/torvalds/linux/commit/a03ef333fbd6cd861c8457c3d055ee3643a9baad
> > > 
> > > [quote]
> > >   If you resorted to AI assistance to identify a bug, you must treat
> > >   it as public. While you may have valid reasons to believe it is not,
> > >   the security team's experience shows that bugs discovered this way
> > >   systematically surface simultaneously across multiple researchers,
> > >   often on the same day
> > > [/quote]
> > > 
> > > IME for qemu-security we've definitely seen the same bugs reported by
> > > different people, within days of each other, a couple of times even
> > > the same day. Dupes are not the majority  - QEMU has so much low
> > > hanging fruit - but they're a sizeable chunk.
> > > 
> > > 
> > > Because we don't have a good automated way to publish disclosures
> > > from qemu-security, we don't provide contributors any way to check
> > > for duplicates before submission. The burden of checking for dupes
> > > thus falls on the qemu-security triage people, or the maintainers
> > > we forward issues to.
> > > 
> > > For disclosures in the non-virtualization use case, we ask the
> > > reporter to file a gitlab issue or send a patch. They almost never
> > > do this. We're lucky to have any reply to emails at all.
> > > 
> > > 
> > > IMHO we need to move to an issue tracker. GitLab was originally
> > > discounted in favour of qemu-security list since "Confidential"
> > > issues cannot have visibility managed in a fine grained way.
> > > They are visible to everyone with "Reporter" role or higher,
> > > which is 70+ people for QEMU.
> > > 
> > > If we accept the kernel's view that AI discovered issues should be
> > > considered public immediately because of the high chance of parallel
> > > "discovery", then GitLab's limitations don't seem to bad.
> > > 
> > > It is also of note that we've had some people ignoring qemu-security
> > > policy and directly filing public gitlab issues for stuff discovered
> > > via AI/LLM agents already. Those are not getting attention for CVE
> > > assignment, should it be needed which and can't be cross-referenced
> > > by general maintaniers to stuff we have received via qemu-security.
> > > 
> > > We have some options IMHO
> > > 
> > >  1. Move all security disclosure to GitLab confidential issues
> > >     no disclosures via email
> > > 
> > >  2. Move AI/fuzzer assisted disclosures to GitLab confidential
> > >     issues, keep human discovered issues on qemu-security list
> > > 
> > >  3. Move AI/fuzzer assisted disclosures to GitLab public
> > >     issues, keep human discovered issues on qemu-security list
> > 
> > or 4. do it like linux did, and move ai assisted disclosured
> > to the public ML. Fuzzer assisted can stay with human discovered.
> 
> Linux could do that because of the way they issue CVEs. IIUC, every
> bug that goes into stable kernels gets evaluated for whether a CVE
> is required. Thus they don't need to do CVE allocation in alignment
> with any bug disclosure process.
> 
> While our stable branch process for QEMU is healthier than it has
> ever been, it is still essentially a 1 person effort looking for
> candidate patches, not a strong enough culture of maintainers
> sending patches to stable proactively.  So I don't think we're
> in a position to replicate Linux's approach to CVE assignment.
> 
> So IMHO we still need to have some issue tracking mechanism to
> tag incoming disclosures as needing CVE assignment or not. If
> all disclosures end up on qemu-devel and untracked we'll just
> never end up getting CVEs for most of them. That doesn't feel
> like a viable approach
> 
> 
> With regards,
> Daniel
> -- 
> |: https://berrange.com       ~~        https://hachyderm.io/@berrange :|
> |: https://libvirt.org          ~~          https://entangle-photo.org :|
> |: https://pixelfed.art/berrange   ~~    https://fstop138.berrange.com :|



Oh I certainly do not object to that! can be public gitlab+public ML.

-- 
MST


Reply via email to