Re: CVE-2024-49993: iommu/vt-d: Fix potential lockup if qi_submit_sync called with 0 count

Re: CVE-2024-49993: iommu/vt-d: Fix potential lockup if qi_submit_sync called with 0 count

[Brendan Jackman]

Hi Greg,

> Currently, there is no impact
> by this bug on the existing users because no callers are submitting
> invalidations with 0 descriptors.

I think this CVE could be discarded, the count arg is always hard-coded to 1.
The buggy function isn't even exposed to modules so I think even if we care
about out-of-tree code we should be OK here. (But based on [1] it sounds like
out-of-tree code is probably out-of-scope for kernel CVEs anyway?)

[1] https://docs.kernel.org/process/cve.html#invalid-cves

FWIW, I don't have any burning desire to kill this CVE in particular, I'm just
testing the water to see if this is one reasonable way we could share some
triage effort among consumers of kernel CVEs...

Cheers,
Brendan

Re: CVE-2024-49993: iommu/vt-d: Fix potential lockup if qi_submit_sync called with 0 count

[Greg KH]

On Tue, Oct 29, 2024 at 11:40:08AM +0000, Brendan Jackman wrote:
> Hi Greg,
> 
> > Currently, there is no impact
> > by this bug on the existing users because no callers are submitting
> > invalidations with 0 descriptors.
> 
> I think this CVE could be discarded, the count arg is always hard-coded to 1.
> The buggy function isn't even exposed to modules so I think even if we care
> about out-of-tree code we should be OK here. (But based on [1] it sounds like
> out-of-tree code is probably out-of-scope for kernel CVEs anyway?)
> 
> [1] https://docs.kernel.org/process/cve.html#invalid-cves

Yes, out-of-tree code is on its own, for obvious reasons (i.e. we have
no idea what they are doing, and they know exactly what we are doing...)

> FWIW, I don't have any burning desire to kill this CVE in particular, I'm just
> testing the water to see if this is one reasonable way we could share some
> triage effort among consumers of kernel CVEs...

Yes, you are right, this one should be rejected, and that's now done,
thanks for the review.

greg k-h