Re: CVE-2026-31431: CopyFail: linux local privilege scalation

[Aaron Rainbolt <arraybolt3 () gmail com>]

On Wed, 29 Apr 2026 23:39:44 +0100
Sam James <sam () gentoo org> wrote:

> Jan Schaumann <jschauma () netmeister org> writes:
>
> > Hi,
> >
> > This is currently making the rounds and looks pretty
> > severe:
> >
> > https://copy.fail/
> >
> > A local privilege escalation vulnerability with a
> > working PoC python script exploiting a logic flaw in
> > the kernel crypto API (AF_ALG) affecting most Linux
> > distributions.
> >
> > More detailed write-up:
> > https://xint.io/blog/copy-fail-linux-distributions
> >
> > [...]  
>
> > Affected and fixed versions
> > ===========================
> >
> > Issue introduced in 4.14 with commit
> > 72548b093ee38a6d4f2a19e6ef1948ae05c181f7 and fixed in
> > 6.18.22 with commit
> > fafe0fa2995a0f7073c1c358d7d3145bcc9aedd8
> >
> > Issue introduced in 4.14 with commit
> > 72548b093ee38a6d4f2a19e6ef1948ae05c181f7 and fixed in
> > 6.19.12 with commit
> > ce42ee423e58dffa5ec03524054c9d8bfd4f6237
> >
> > Issue introduced in 4.14 with commit
> > 72548b093ee38a6d4f2a19e6ef1948ae05c181f7 and fixed in
> > 7.0 with commit
> > a664bf3d603dc3bdcf9ae47cc21e0daec706d7a5  
>
> To say it explicitly, all the kernels in-between before 6.18 aren't
> fixed.
>
> > https://git.kernel.org/stable/c/fafe0fa2995a0f7073c1c358d7d3145bcc9aedd8
> > https://git.kernel.org/stable/c/ce42ee423e58dffa5ec03524054c9d8bfd4f6237
> > https://git.kernel.org/stable/c/a664bf3d603dc3bdcf9ae47cc21e0daec706d7a5
> >
> > ----
> >
> > PoC:
> > https://github.com/theori-io/copy-fail-CVE-2026-31431/blob/main/copy_fail_exp.py
> >
> >
> > Mitigation:
> >
> > # echo "install algif_aead /bin/false" >
> > /etc/modprobe.d/disable-algif.conf # rmmod algif_aead   
>
> Brad Spengler has been pointing out that this won't work on a few
> common enterprise kernels where CONFIG_CRYPTO_USER_API_AEAD=y (rather
> than m).

I'd like to also point out that the copy.fail website may underplay the
impact this mitigation has on userspace. It's not the easiest thing in
the world to find out everywhere where an AF_ALG socket is opened and
then set up in AEAD mode in Debian, but so far I've found that at least
bluez, cryptsetup, iwd, and stress-ng contain code that does this, as
does a particular test in some Rust code in rustc, firefox-esr, and
thunderbird. libkcapi is also a thing, anything that does AEAD with it
will be affected. Simply nuking this part of the kernel API from orbit
is not a universally safe operation.

(Note that I've not yet made any of these applications misbehave by
applying the mitigation, I tried some experiments with cryptsetup and
the results were inconclusive. So this might be safe enough for most
people.)

--
Aaron

Attachment: _bin
Description: OpenPGP digital signature