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

[Reid Sutherland <reid () thirddimension net>]

On Fri, 2026-05-01 at 18:52 +0200, Justin Swartz wrote:
> On Fri, 2026-05-01 at 11:08 -0400, Reid Sutherland wrote:
> > Does anything load the vulnerable module by default or not?  If
> > not,
> > this should be low-rated IMO.
>
> An unprivileged user requesting an AF_ALG socket will trigger the
> kernel
> module autoloader:
>
> $ su -l
> Password:
>
> # cat > /root/modprobe << "EOF"
> #!/bin/sh
> echo "$(date -u) modprobe $@" >> /tmp/modprobe.log
> exec /sbin/modprobe "$@"
> EOF
>
> # chmod 700 /root/modprobe
>
> # cat /proc/sys/kernel/modprobe
> /sbin/modprobe
>
> # echo "/root/modprobe" > /proc/sys/kernel/modprobe
> # cat /proc/sys/kernel/modprobe
> /root/modprobe
>
> # exit
>
> $ lsmod | grep aead | wc -l
> 0
>
> $ date -u && ./copy_fail_exp.py
> Fri 01 May 2026 16:08:24 UTC
>
> # cat /tmp/modprobe.log
> Fri May  1 16:08:24 UTC 2026 modprobe -q -- net-pf-38
> Fri May  1 16:08:24 UTC 2026 modprobe -q -- algif-aead
>
> # lsmod | grep aead
> algif_aead             16384  0
> af_alg                 36864  1 algif_aead
>
> # echo "/sbin/modprobe" > /proc/sys/kernel/modprobe
> # cat /proc/sys/kernel/modprobe
> /sbin/modprobe
>
> # exit


Why is userspace allowed to load modules in any capacity?  Why do we
need kernel modules for math?

I'm assuming any thoroughly qualified platform engineer compiles the
host kernel without module support.  At least, that needs to make a
comeback, bring back applying grsec patches and make menuconfig..

I just finished defending the kernel on LinkedIn too, in that kernel
exploit attack surface is a non-issue if you trust how it's maintained.
Massive torpedo to my reputation two days later.