OSDN Git Service

arm64: KVM: Guests can skip __install_bp_hardening_cb()s HYP work
authorJames Morse <james.morse@arm.com>
Fri, 21 Sep 2018 20:49:19 +0000 (21:49 +0100)
committerCatalin Marinas <catalin.marinas@arm.com>
Fri, 19 Oct 2018 14:37:25 +0000 (15:37 +0100)
enable_smccc_arch_workaround_1() passes NULL as the hyp_vecs start and
end if the HVC conduit is in use, and ARM_SMCCC_ARCH_WORKAROUND_1 is
detected.

If the guest kernel happened to be built with KVM_INDIRECT_VECTORS,
we go on to allocate a slot, memcpy() the empty workaround in and
do the appropriate cache maintenance.

This works as we always tell memcpy() the range is 0, so it never
accesses the NULL src pointer, but we still do the cache maintenance.

If hyp_vecs_start is NULL we know we're a guest, just update the fn
like the !KVM_INDIRECT_VECTORS version.

Reviewed-by: Julien Thierry <julien.thierry@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
arch/arm64/kernel/cpu_errata.c

index 9af8df9..a509e35 100644 (file)
@@ -138,6 +138,15 @@ static void __install_bp_hardening_cb(bp_hardening_cb_t fn,
        static DEFINE_SPINLOCK(bp_lock);
        int cpu, slot = -1;
 
+       /*
+        * enable_smccc_arch_workaround_1() passes NULL for the hyp_vecs
+        * start/end if we're a guest. Skip the hyp-vectors work.
+        */
+       if (!hyp_vecs_start) {
+               __this_cpu_write(bp_hardening_data.fn, fn);
+               return;
+       }
+
        spin_lock(&bp_lock);
        for_each_possible_cpu(cpu) {
                if (per_cpu(bp_hardening_data.fn, cpu) == fn) {