KVM: PPC: Book3S HV: Always save guest pmu for guest capable of nesting
authorSuraj Jitindar Singh <sjitindarsingh@gmail.com>
Wed, 3 Jul 2019 01:20:20 +0000 (11:20 +1000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 31 Jul 2019 05:24:57 +0000 (07:24 +0200)
commit08ab7cccafa6c1d022b9a35f7641d0bdaca64b0f
tree7374d01f5df7f8607c0f83f32802d423b73890ae
parentca7e6b286333749abd7ccc3f413b518a5405f512
KVM: PPC: Book3S HV: Always save guest pmu for guest capable of nesting

commit 63279eeb7f93abb1692573c26f1e038e1a87358b upstream.

The performance monitoring unit (PMU) registers are saved on guest
exit when the guest has set the pmcregs_in_use flag in its lppaca, if
it exists, or unconditionally if it doesn't. If a nested guest is
being run then the hypervisor doesn't, and in most cases can't, know
if the PMU registers are in use since it doesn't know the location of
the lppaca for the nested guest, although it may have one for its
immediate guest. This results in the values of these registers being
lost across nested guest entry and exit in the case where the nested
guest was making use of the performance monitoring facility while it's
nested guest hypervisor wasn't.

Further more the hypervisor could interrupt a guest hypervisor between
when it has loaded up the PMU registers and it calling H_ENTER_NESTED
or between returning from the nested guest to the guest hypervisor and
the guest hypervisor reading the PMU registers, in
kvmhv_p9_guest_entry(). This means that it isn't sufficient to just
save the PMU registers when entering or exiting a nested guest, but
that it is necessary to always save the PMU registers whenever a guest
is capable of running nested guests to ensure the register values
aren't lost in the context switch.

Ensure the PMU register values are preserved by always saving their
value into the vcpu struct when a guest is capable of running nested
guests.

This should have minimal performance impact however any impact can be
avoided by booting a guest with "-machine pseries,cap-nested-hv=false"
on the qemu commandline.

Fixes: 95a6432ce903 ("KVM: PPC: Book3S HV: Streamlined guest entry/exit path on P9 for radix guests")
Cc: stable@vger.kernel.org # v4.20+
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190703012022.15644-1-sjitindarsingh@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/powerpc/kvm/book3s_hv.c