Re: New CVE entries this week


Pavel Machek
 

Hi!

** Updated CVEs
CVE-2021-22543: v4.19 and v5.10 are fixed. v4.4 uses another way to
get pfn. If v4.4 is vulnerable it needs to write its own patch.
4.4 is very different in that area, and KVM is not exactly our
focus. A lot of research would be needed. I guess we can simply ignore
this one.

* CVE detail

CVE-2021-35477: unprivileged BPF program can obtain sensitive
information from kernel memory via a speculative store bypass
side-channel attack because the technique used by the BPF verifier to
manage speculation is unreliable

CVE-2021-34556 and CVE-2021-35477 are fixed by the same commits.
commit 2039f26f3aca fixes af86ca4e3088(introduced by v4.17-rc7) and
f7cf25b2026d(introduced by v5.3-rc1).

Fixed status
mainline: [f5e81d1117501546b7be050c5fbafa6efd2c722c,
2039f26f3aca5b0e419b98f65dd36481337b86ee]
stable/5.10: [bea9e2fd180892eba2574711b05b794f1d0e7b73,
0e9280654aa482088ee6ef3deadef331f5ac5fb0]
stable/5.13: [ddab060f996e17b38bb181c5fd11a83fd1bfa0df,
0b27bdf02c400684225ee5ee99970bcbf5082282]
Yes, speculation is huge problem, and getting BPF right with broken
CPUs will be hard. I'd hope CIP people are not using untrusted BTF
programs, and that we can ignore it.

CVE-2021-3669: reading /proc/sysvipc/shm does not scale with large
shared memory segment counts

According to redhat bugzilla, it said "Not reported upstream, patches
are being worked on. It is not considered high impact because of the
requirements and need to have massive amount of shm (usually well
above ulimits) ".

https://bugzilla.redhat.com/show_bug.cgi?id=1986473#c10
DoS only, and only in unusual configuration. I believe we can ignore
this one.

CVE-2021-37159: hso_free_net_device in drivers/net/usb/hso.c in the
Linux kernel through 5.13.4 calls unregister_netdev without checking
for the NETREG_REGISTERED state, leading to a use-after-free and a
double free.

The mainline, 5.10, 5.13 are fixed.

Fixed status
mainline: [a6ecfb39ba9d7316057cea823b196b734f6b18ca]
stable/5.10: [115e4f5b64ae8d9dd933167cafe2070aaac45849]
stable/5.13: [eeaa4b8d1e2e6f10362673d283a97dccc7275afa]
I guess we could try to rework the function in similar way 5.10 did,
but... we are not using HSO in our configs, and I have hard time
imagining how "attacker" would trigger it. So this is... just a
bug. I'd suggest ignoring.

Best regards,
Pavel

--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

Join cip-dev@lists.cip-project.org to automatically receive all group messages.