[PULL,15/17] KVM: PPC: Book3S PR: Don't corrupt guest state when kernel uses VMX

Message ID 1377771735-3289-16-git-send-email-agraf@suse.de
State New, archived
Headers show

Commit Message

Alexander Graf Aug. 29, 2013, 10:22 a.m.
From: Paul Mackerras <paulus@samba.org>

Currently the code assumes that once we load up guest FP/VSX or VMX
state into the CPU, it stays valid in the CPU registers until we
explicitly flush it to the thread_struct.  However, on POWER7,
copy_page() and memcpy() can use VMX.  These functions do flush the
VMX state to the thread_struct before using VMX instructions, but if
this happens while we have guest state in the VMX registers, and we
then re-enter the guest, we don't reload the VMX state from the
thread_struct, leading to guest corruption.  This has been observed
to cause guest processes to segfault.

To fix this, we check before re-entering the guest that all of the
bits corresponding to facilities owned by the guest, as expressed
in vcpu->arch.guest_owned_ext, are set in current->thread.regs->msr.
Any bits that have been cleared correspond to facilities that have
been used by kernel code and thus flushed to the thread_struct, so
for them we reload the state from the thread_struct.

We also need to check current->thread.regs->msr before calling
giveup_fpu() or giveup_altivec(), since if the relevant bit is
clear, the state has already been flushed to the thread_struct and
to flush it again would corrupt it.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
 arch/powerpc/kvm/book3s_pr.c | 29 +++++++++++++++++++++++++----
 1 file changed, 25 insertions(+), 4 deletions(-)


diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
index 19cbac6..983e5ed 100644
--- a/arch/powerpc/kvm/book3s_pr.c
+++ b/arch/powerpc/kvm/book3s_pr.c
@@ -468,7 +468,8 @@  void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
 		 * both the traditional FP registers and the added VSX
 		 * registers into thread.fpr[].
-		giveup_fpu(current);
+		if (current->thread.regs->msr & MSR_FP)
+			giveup_fpu(current);
 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
 			vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
@@ -483,7 +484,8 @@  void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
 	if (msr & MSR_VEC) {
-		giveup_altivec(current);
+		if (current->thread.regs->msr & MSR_VEC)
+			giveup_altivec(current);
 		memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
 		vcpu->arch.vscr = t->vscr;
@@ -575,8 +577,6 @@  static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
 	printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
-	current->thread.regs->msr |= msr;
 	if (msr & MSR_FP) {
 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
 			thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
@@ -598,12 +598,32 @@  static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
+	current->thread.regs->msr |= msr;
 	vcpu->arch.guest_owned_ext |= msr;
 	return RESUME_GUEST;
+ * Kernel code using FP or VMX could have flushed guest state to
+ * the thread_struct; if so, get it back now.
+ */
+static void kvmppc_handle_lost_ext(struct kvm_vcpu *vcpu)
+	unsigned long lost_ext;
+	lost_ext = vcpu->arch.guest_owned_ext & ~current->thread.regs->msr;
+	if (!lost_ext)
+		return;
+	if (lost_ext & MSR_FP)
+		kvmppc_load_up_fpu();
+	if (lost_ext & MSR_VEC)
+		kvmppc_load_up_altivec();
+	current->thread.regs->msr |= lost_ext;
 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
                        unsigned int exit_nr)
@@ -892,6 +912,7 @@  program_interrupt:
 		} else {
+		kvmppc_handle_lost_ext(vcpu);
 	trace_kvm_book3s_reenter(r, vcpu);