KVM: PPC: Book3S HV: Handle guest-caused machine checks on POWER7 without panicking

Currently, if a machine check interrupt happens while we are in the
guest, we exit the guest and call the host's machine check handler,
which tends to cause the host to panic.  Some machine checks can be
triggered by the guest; for example, if the guest creates two entries
in the SLB that map the same effective address, and then accesses that
effective address, the CPU will take a machine check interrupt.

To handle this better, when a machine check happens inside the guest,
we call a new function, kvmppc_realmode_machine_check(), while still in
real mode before exiting the guest.  On POWER7, it handles the cases
that the guest can trigger, either by flushing and reloading the SLB,
or by flushing the TLB, and then it delivers the machine check interrupt
directly to the guest without going back to the host.  On POWER7, the
OPAL firmware patches the machine check interrupt vector so that it
gets control first, and it leaves behind its analysis of the situation
in a structure pointed to by the opal_mc_evt field of the paca.  The
kvmppc_realmode_machine_check() function looks at this, and if OPAL
reports that there was no error, or that it has handled the error, we
also go straight back to the guest with a machine check.  We have to
deliver a machine check to the guest since the machine check interrupt
might have trashed valid values in SRR0/1.

If the machine check is one we can't handle in real mode, and one that
OPAL hasn't already handled, or on PPC970, we exit the guest and call
the host's machine check handler.  We do this by jumping to the
machine_check_fwnmi label, rather than absolute address 0x200, because
we don't want to re-execute OPAL's handler on POWER7.  On PPC970, the
two are equivalent because address 0x200 just contains a branch.

Then, if the host machine check handler decides that the system can
continue executing, kvmppc_handle_exit() delivers a machine check
interrupt to the guest -- once again to let the guest know that SRR0/1
have been modified.

Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix checkpatch warnings]
Signed-off-by: Alexander Graf <agraf@suse.de>

5 files changed, 213 insertions, 28 deletions

diff --git a/arch/powerpc/include/asm/mmu-hash64.h b/arch/powerpc/include/asm/mmu-hash64.h
index 9673f73eb8db..2fdb47a19efd 100644
--- a/arch/powerpc/include/asm/mmu-hash64.h
+++ b/arch/powerpc/include/asm/mmu-hash64.h
@@ -121,6 +121,16 @@ extern char initial_stab[];
 #define PP_RXRX 3	/* Supervisor read,       User read */
 #define PP_RXXX	(HPTE_R_PP0 | 2)	/* Supervisor read, user none */
 
+/* Fields for tlbiel instruction in architecture 2.06 */
+#define TLBIEL_INVAL_SEL_MASK	0xc00	/* invalidation selector */
+#define  TLBIEL_INVAL_PAGE	0x000	/* invalidate a single page */
+#define  TLBIEL_INVAL_SET_LPID	0x800	/* invalidate a set for current LPID */
+#define  TLBIEL_INVAL_SET	0xc00	/* invalidate a set for all LPIDs */
+#define TLBIEL_INVAL_SET_MASK	0xfff000	/* set number to inval. */
+#define TLBIEL_INVAL_SET_SHIFT	12
+
+#define POWER7_TLB_SETS		128	/* # sets in POWER7 TLB */
+
 #ifndef __ASSEMBLY__
 
 struct hash_pte {
diff --git a/arch/powerpc/kvm/Makefile b/arch/powerpc/kvm/Makefile
index cd8965828676..1e473d46322c 100644
--- a/arch/powerpc/kvm/Makefile
+++ b/arch/powerpc/kvm/Makefile
@@ -73,6 +73,7 @@ kvm-book3s_64-builtin-objs-$(CONFIG_KVM_BOOK3S_64_HV) := \
 	book3s_hv_rmhandlers.o \
 	book3s_hv_rm_mmu.o \
 	book3s_64_vio_hv.o \
+	book3s_hv_ras.o \
 	book3s_hv_builtin.o
 
 kvm-book3s_64-module-objs := \
diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
index ddbec60cb0d2..71d0c90b62bf 100644
--- a/arch/powerpc/kvm/book3s_hv.c
+++ b/arch/powerpc/kvm/book3s_hv.c
@@ -545,6 +545,17 @@ static int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 	case BOOK3S_INTERRUPT_PERFMON:
 		r = RESUME_GUEST;
 		break;
+	case BOOK3S_INTERRUPT_MACHINE_CHECK:
+		/*
+		 * Deliver a machine check interrupt to the guest.
+		 * We have to do this, even if the host has handled the
+		 * machine check, because machine checks use SRR0/1 and
+		 * the interrupt might have trashed guest state in them.
+		 */
+		kvmppc_book3s_queue_irqprio(vcpu,
+					    BOOK3S_INTERRUPT_MACHINE_CHECK);
+		r = RESUME_GUEST;
+		break;
 	case BOOK3S_INTERRUPT_PROGRAM:
 	{
 		ulong flags;
diff --git a/arch/powerpc/kvm/book3s_hv_ras.c b/arch/powerpc/kvm/book3s_hv_ras.c
new file mode 100644
index 000000000000..35f3cf0269b3
--- /dev/null
+++ b/arch/powerpc/kvm/book3s_hv_ras.c
@@ -0,0 +1,144 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * Copyright 2012 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
+ */
+
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/kvm.h>
+#include <linux/kvm_host.h>
+#include <linux/kernel.h>
+#include <asm/opal.h>
+
+/* SRR1 bits for machine check on POWER7 */
+#define SRR1_MC_LDSTERR		(1ul << (63-42))
+#define SRR1_MC_IFETCH_SH	(63-45)
+#define SRR1_MC_IFETCH_MASK	0x7
+#define SRR1_MC_IFETCH_SLBPAR		2	/* SLB parity error */
+#define SRR1_MC_IFETCH_SLBMULTI		3	/* SLB multi-hit */
+#define SRR1_MC_IFETCH_SLBPARMULTI	4	/* SLB parity + multi-hit */
+#define SRR1_MC_IFETCH_TLBMULTI		5	/* I-TLB multi-hit */
+
+/* DSISR bits for machine check on POWER7 */
+#define DSISR_MC_DERAT_MULTI	0x800		/* D-ERAT multi-hit */
+#define DSISR_MC_TLB_MULTI	0x400		/* D-TLB multi-hit */
+#define DSISR_MC_SLB_PARITY	0x100		/* SLB parity error */
+#define DSISR_MC_SLB_MULTI	0x080		/* SLB multi-hit */
+#define DSISR_MC_SLB_PARMULTI	0x040		/* SLB parity + multi-hit */
+
+/* POWER7 SLB flush and reload */
+static void reload_slb(struct kvm_vcpu *vcpu)
+{
+	struct slb_shadow *slb;
+	unsigned long i, n;
+
+	/* First clear out SLB */
+	asm volatile("slbmte %0,%0; slbia" : : "r" (0));
+
+	/* Do they have an SLB shadow buffer registered? */
+	slb = vcpu->arch.slb_shadow.pinned_addr;
+	if (!slb)
+		return;
+
+	/* Sanity check */
+	n = min_t(u32, slb->persistent, SLB_MIN_SIZE);
+	if ((void *) &slb->save_area[n] > vcpu->arch.slb_shadow.pinned_end)
+		return;
+
+	/* Load up the SLB from that */
+	for (i = 0; i < n; ++i) {
+		unsigned long rb = slb->save_area[i].esid;
+		unsigned long rs = slb->save_area[i].vsid;
+
+		rb = (rb & ~0xFFFul) | i;	/* insert entry number */
+		asm volatile("slbmte %0,%1" : : "r" (rs), "r" (rb));
+	}
+}
+
+/* POWER7 TLB flush */
+static void flush_tlb_power7(struct kvm_vcpu *vcpu)
+{
+	unsigned long i, rb;
+
+	rb = TLBIEL_INVAL_SET_LPID;
+	for (i = 0; i < POWER7_TLB_SETS; ++i) {
+		asm volatile("tlbiel %0" : : "r" (rb));
+		rb += 1 << TLBIEL_INVAL_SET_SHIFT;
+	}
+}
+
+/*
+ * On POWER7, see if we can handle a machine check that occurred inside
+ * the guest in real mode, without switching to the host partition.
+ *
+ * Returns: 0 => exit guest, 1 => deliver machine check to guest
+ */
+static long kvmppc_realmode_mc_power7(struct kvm_vcpu *vcpu)
+{
+	unsigned long srr1 = vcpu->arch.shregs.msr;
+	struct opal_machine_check_event *opal_evt;
+	long handled = 1;
+
+	if (srr1 & SRR1_MC_LDSTERR) {
+		/* error on load/store */
+		unsigned long dsisr = vcpu->arch.shregs.dsisr;
+
+		if (dsisr & (DSISR_MC_SLB_PARMULTI | DSISR_MC_SLB_MULTI |
+			     DSISR_MC_SLB_PARITY | DSISR_MC_DERAT_MULTI)) {
+			/* flush and reload SLB; flushes D-ERAT too */
+			reload_slb(vcpu);
+			dsisr &= ~(DSISR_MC_SLB_PARMULTI | DSISR_MC_SLB_MULTI |
+				   DSISR_MC_SLB_PARITY | DSISR_MC_DERAT_MULTI);
+		}
+		if (dsisr & DSISR_MC_TLB_MULTI) {
+			flush_tlb_power7(vcpu);
+			dsisr &= ~DSISR_MC_TLB_MULTI;
+		}
+		/* Any other errors we don't understand? */
+		if (dsisr & 0xffffffffUL)
+			handled = 0;
+	}
+
+	switch ((srr1 >> SRR1_MC_IFETCH_SH) & SRR1_MC_IFETCH_MASK) {
+	case 0:
+		break;
+	case SRR1_MC_IFETCH_SLBPAR:
+	case SRR1_MC_IFETCH_SLBMULTI:
+	case SRR1_MC_IFETCH_SLBPARMULTI:
+		reload_slb(vcpu);
+		break;
+	case SRR1_MC_IFETCH_TLBMULTI:
+		flush_tlb_power7(vcpu);
+		break;
+	default:
+		handled = 0;
+	}
+
+	/*
+	 * See if OPAL has already handled the condition.
+	 * We assume that if the condition is recovered then OPAL
+	 * will have generated an error log event that we will pick
+	 * up and log later.
+	 */
+	opal_evt = local_paca->opal_mc_evt;
+	if (opal_evt->version == OpalMCE_V1 &&
+	    (opal_evt->severity == OpalMCE_SEV_NO_ERROR ||
+	     opal_evt->disposition == OpalMCE_DISPOSITION_RECOVERED))
+		handled = 1;
+
+	if (handled)
+		opal_evt->in_use = 0;
+
+	return handled;
+}
+
+long kvmppc_realmode_machine_check(struct kvm_vcpu *vcpu)
+{
+	if (cpu_has_feature(CPU_FTR_ARCH_206))
+		return kvmppc_realmode_mc_power7(vcpu);
+
+	return 0;
+}
diff --git a/arch/powerpc/kvm/book3s_hv_rmhandlers.S b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
index b48bd53dd771..10b6c358dd77 100644
--- a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
+++ b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
@@ -27,6 +27,7 @@
 #include <asm/asm-offsets.h>
 #include <asm/exception-64s.h>
 #include <asm/kvm_book3s_asm.h>
+#include <asm/mmu-hash64.h>
 
 /*****************************************************************************
  *                                                                           *
@@ -678,8 +679,7 @@ BEGIN_FTR_SECTION
 1:
 END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
 
-nohpte_cont:
-hcall_real_cont:		/* r9 = vcpu, r12 = trap, r13 = paca */
+guest_exit_cont:		/* r9 = vcpu, r12 = trap, r13 = paca */
 	/* Save DEC */
 	mfspr	r5,SPRN_DEC
 	mftb	r6
@@ -700,6 +700,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
 	std	r6, VCPU_FAULT_DAR(r9)
 	stw	r7, VCPU_FAULT_DSISR(r9)
 
+	/* See if it is a machine check */
+	cmpwi	r12, BOOK3S_INTERRUPT_MACHINE_CHECK
+	beq	machine_check_realmode
+mc_cont:
+
 	/* Save guest CTRL register, set runlatch to 1 */
 6:	mfspr	r6,SPRN_CTRLF
 	stw	r6,VCPU_CTRL(r9)
@@ -1112,38 +1117,41 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
 	/*
 	 * For external and machine check interrupts, we need
 	 * to call the Linux handler to process the interrupt.
-	 * We do that by jumping to the interrupt vector address
-	 * which we have in r12.  The [h]rfid at the end of the
+	 * We do that by jumping to absolute address 0x500 for
+	 * external interrupts, or the machine_check_fwnmi label
+	 * for machine checks (since firmware might have patched
+	 * the vector area at 0x200).  The [h]rfid at the end of the
 	 * handler will return to the book3s_hv_interrupts.S code.
 	 * For other interrupts we do the rfid to get back
-	 * to the book3s_interrupts.S code here.
+	 * to the book3s_hv_interrupts.S code here.
 	 */
 	ld	r8, HSTATE_VMHANDLER(r13)
 	ld	r7, HSTATE_HOST_MSR(r13)
 
+	cmpwi	cr1, r12, BOOK3S_INTERRUPT_MACHINE_CHECK
 	cmpwi	r12, BOOK3S_INTERRUPT_EXTERNAL
+BEGIN_FTR_SECTION
 	beq	11f
-	cmpwi	r12, BOOK3S_INTERRUPT_MACHINE_CHECK
+END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
 
 	/* RFI into the highmem handler, or branch to interrupt handler */
-12:	mfmsr	r6
-	mtctr	r12
+	mfmsr	r6
 	li	r0, MSR_RI
 	andc	r6, r6, r0
 	mtmsrd	r6, 1			/* Clear RI in MSR */
 	mtsrr0	r8
 	mtsrr1	r7
-	beqctr
+	beqa	0x500			/* external interrupt (PPC970) */
+	beq	cr1, 13f		/* machine check */
 	RFI
 
-11:
-BEGIN_FTR_SECTION
-	b	12b
-END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
-	mtspr	SPRN_HSRR0, r8
+	/* On POWER7, we have external interrupts set to use HSRR0/1 */
+11:	mtspr	SPRN_HSRR0, r8
 	mtspr	SPRN_HSRR1, r7
 	ba	0x500
 
+13:	b	machine_check_fwnmi
+
 /*
  * Check whether an HDSI is an HPTE not found fault or something else.
  * If it is an HPTE not found fault that is due to the guest accessing
@@ -1176,7 +1184,7 @@ kvmppc_hdsi:
 	cmpdi	r3, 0			/* retry the instruction */
 	beq	6f
 	cmpdi	r3, -1			/* handle in kernel mode */
-	beq	nohpte_cont
+	beq	guest_exit_cont
 	cmpdi	r3, -2			/* MMIO emulation; need instr word */
 	beq	2f
 
@@ -1190,6 +1198,7 @@ kvmppc_hdsi:
 	li	r10, BOOK3S_INTERRUPT_DATA_STORAGE
 	li	r11, (MSR_ME << 1) | 1	/* synthesize MSR_SF | MSR_ME */
 	rotldi	r11, r11, 63
+fast_interrupt_c_return:
 6:	ld	r7, VCPU_CTR(r9)
 	lwz	r8, VCPU_XER(r9)
 	mtctr	r7
@@ -1222,7 +1231,7 @@ kvmppc_hdsi:
 	/* Unset guest mode. */
 	li	r0, KVM_GUEST_MODE_NONE
 	stb	r0, HSTATE_IN_GUEST(r13)
-	b	nohpte_cont
+	b	guest_exit_cont
 
 /*
  * Similarly for an HISI, reflect it to the guest as an ISI unless
@@ -1248,9 +1257,9 @@ kvmppc_hisi:
 	ld	r11, VCPU_MSR(r9)
 	li	r12, BOOK3S_INTERRUPT_H_INST_STORAGE
 	cmpdi	r3, 0			/* retry the instruction */
-	beq	6f
+	beq	fast_interrupt_c_return
 	cmpdi	r3, -1			/* handle in kernel mode */
-	beq	nohpte_cont
+	beq	guest_exit_cont
 
 	/* Synthesize an ISI for the guest */
 	mr	r11, r3
@@ -1259,12 +1268,7 @@ kvmppc_hisi:
 	li	r10, BOOK3S_INTERRUPT_INST_STORAGE
 	li	r11, (MSR_ME << 1) | 1	/* synthesize MSR_SF | MSR_ME */
 	rotldi	r11, r11, 63
-6:	ld	r7, VCPU_CTR(r9)
-	lwz	r8, VCPU_XER(r9)
-	mtctr	r7
-	mtxer	r8
-	mr	r4, r9
-	b	fast_guest_return
+	b	fast_interrupt_c_return
 
 3:	ld	r6, VCPU_KVM(r9)	/* not relocated, use VRMA */
 	ld	r5, KVM_VRMA_SLB_V(r6)
@@ -1280,14 +1284,14 @@ kvmppc_hisi:
 hcall_try_real_mode:
 	ld	r3,VCPU_GPR(R3)(r9)
 	andi.	r0,r11,MSR_PR
-	bne	hcall_real_cont
+	bne	guest_exit_cont
 	clrrdi	r3,r3,2
 	cmpldi	r3,hcall_real_table_end - hcall_real_table
-	bge	hcall_real_cont
+	bge	guest_exit_cont
 	LOAD_REG_ADDR(r4, hcall_real_table)
 	lwzx	r3,r3,r4
 	cmpwi	r3,0
-	beq	hcall_real_cont
+	beq	guest_exit_cont
 	add	r3,r3,r4
 	mtctr	r3
 	mr	r3,r9		/* get vcpu pointer */
@@ -1308,7 +1312,7 @@ hcall_real_fallback:
 	li	r12,BOOK3S_INTERRUPT_SYSCALL
 	ld	r9, HSTATE_KVM_VCPU(r13)
 
-	b	hcall_real_cont
+	b	guest_exit_cont
 
 	.globl	hcall_real_table
 hcall_real_table:
@@ -1567,6 +1571,21 @@ kvm_cede_exit:
 	li	r3,H_TOO_HARD
 	blr
 
+	/* Try to handle a machine check in real mode */
+machine_check_realmode:
+	mr	r3, r9		/* get vcpu pointer */
+	bl	.kvmppc_realmode_machine_check
+	nop
+	cmpdi	r3, 0		/* continue exiting from guest? */
+	ld	r9, HSTATE_KVM_VCPU(r13)
+	li	r12, BOOK3S_INTERRUPT_MACHINE_CHECK
+	beq	mc_cont
+	/* If not, deliver a machine check.  SRR0/1 are already set */
+	li	r10, BOOK3S_INTERRUPT_MACHINE_CHECK
+	li	r11, (MSR_ME << 1) | 1	/* synthesize MSR_SF | MSR_ME */
+	rotldi	r11, r11, 63
+	b	fast_interrupt_c_return
+
 secondary_too_late:
 	ld	r5,HSTATE_KVM_VCORE(r13)
 	HMT_LOW