On Thursday, April 30th, 2026 at 7:13 AM, Russell Senior <[email protected]> 
wrote:

> 
> 
> On 4/30/26 06:42, Ted Mittelstaedt wrote:
> > Note that Ubuntu 26.04 was released on the 23rd of April, and its NOT 
> > vulnerable.  I suspect that there is a connection here and that the 26.04 
> > release date was
> > Advanced.
> 
> I don't think the Ubuntu 26.04 release schedule was advanced. The
> release date is consistent with past releases, see here:
> 
> https://documentation.ubuntu.com/project/release-team/list-of-releases/
> 
> The reason it isn't vulnerable is that the fix got into v7.0 and (I'm
> not sure of the Ubuntu policy, but guessing) because v7.0 was released
> before Ubuntu 26.04 was released, they went with it.
> 
> The thing that kind of surprises me is that the major distributions
> didn't have the fix in by the disclosure day. ArchLinux was also not
> vulnerable, if you update reasonably regularly because they stay pretty
> close to upstream stable kernels and so had the fix as a matter of
> course. Debian and Ubuntu (and Fedora?) seem to have been caught a bit
> flat footed.
> 
> The thing I haven't seen reported yet is: "are non-x86/ architectures
> also affected?" You would guess so, since this was apparently a logical
> error, but the published python script exploit doesn't work on them to
> test, and I haven't seen anyone say. An exploit tuned for ARM, might.
> 
> --
> Russell Senior
> [email protected]
> 

The PoC script makes assumptions about system configuration that result in it 
failing on certain distributions. 

The algif_aead kernel module needs to be loaded and su needs to readable by 
non-root users. (-rwsr-xr-x). The python script is pretty simple so chances are 
some systems are not as directly vulnerable.

It's still serious, since the exploit involves poisoning the page cache of a 
binary in order to trick users into running malicious code. Privilege 
escalation is a side effect of the exploit implementation, not the exploit 
itself.

A more direct PoC demonstrating the bug on an arbitrary binary would be much 
better IMO.
-Ben

Reply via email to