linux: enable BPF_UNPRIV_DEFAULT_OFF between 5.10 and 5.15

Disable unprivileged access to BPF syscalls to prevent denial of service
and privilege escalation via

a) potential speculative execution side-channel-attacks on unmitigated
hardware[0]

or

b) unvalidated memory access in ringbuffer helper functions[1].

Fixes: CVE-2021-4204, CVE-2022-23222

[0] https://ebpf.io/summit-2021-slides/eBPF_Summit_2021-Keynote-Daniel_Borkmann-BPF_and_Spectre.pdf
[1] https://www.openwall.com/lists/oss-security/2022/01/13/1

1 files changed, 9 insertions, 0 deletions

diff --git a/nixos/doc/manual/from_md/release-notes/rl-2111.section.xml b/nixos/doc/manual/from_md/release-notes/rl-2111.section.xml
index 6b706e4aeaa..59da373f38e 100644
--- a/nixos/doc/manual/from_md/release-notes/rl-2111.section.xml
+++ b/nixos/doc/manual/from_md/release-notes/rl-2111.section.xml
@@ -1420,6 +1420,15 @@ Superuser created successfully.
           for those who want to have all RetroArch cores available.
         </para>
       </listitem>
+      <listitem>
+        <para>
+          The Linux kernel for security reasons now restricts access to
+          BPF syscalls via <literal>BPF_UNPRIV_DEFAULT_OFF=y</literal>.
+          Unprivileged access can be reenabled via the
+          <literal>kernel.unprivileged_bpf_disabled</literal> sysctl
+          knob.
+        </para>
+      </listitem>
     </itemizedlist>
   </section>
   <section xml:id="sec-release-21.11-notable-changes">