CVE-2020-8834: Linux kernel Power8 conflicting use of HSTATE_HOST_R1 vulnerability

[Steve Beattie <steve.beattie () canonical com>]

Hello,

KVM in the Linux kernel on Power8 processors has a conflicting
use of HSTATE_HOST_R1 to store r1 state in kvmppc_hv_entry plus in
kvmppc_{save,restore}_tm, leading to a stack corruption. Because
of this, an attacker with the ability run code in kernel space of a
guest VM can cause the host kernel to panic.

There were two commits that, according to the reporter, introduced the
vulnerability:

  f024ee098476 ("KVM: PPC: Book3S HV: Pull out TM state save/restore
                into separate procedures")
  87a11bb6a7f7 ("KVM: PPC: Book3S HV: Work around XER[SO] bug in fake
                suspend mode")

The former landed in 4.8, the latter in 4.17. This was fixed without
realizing the impact in 4.18 with the following three commits, though
it's believed the first is the only strictly necessary commit:

  6f597c6b63b6 ("KVM: PPC: Book3S PR: Add guest MSR parameter for
                kvmppc_save_tm()/kvmppc_restore_tm()")
  7b0e827c6970 ("KVM: PPC: Book3S HV: Factor fake-suspend handling out
                of kvmppc_save/restore_tm")
  009c872a8bc4 ("KVM: PPC: Book3S PR: Move kvmppc_save_tm/kvmppc_restore_tm
                to separate file")

Thus, the only upstream kernel affected is the 4.17 kernel -- Ubuntu's
4.15 kernels are affected because we backported 87a11bb6a7f7 to our 4.15
kernels. It does not appear to have been cherrypicked to any of the
upstream stable kernels.

Discovered by Gustavo Romero and Paul Mackerras.
Ref: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1867717
CVE: CVE-2020-8834

-- 
Steve Beattie
<sbeattie () ubuntu com>
http://NxNW.org/~steve/

Attachment: signature.asc
Description: