KVM: PPC: e500mc: Add support for single threaded vcpus on e6500 core

ePAPR represents hardware threads as cpu node properties in device tree.
So with existing QEMU, hardware threads are simply exposed as vcpus with
one hardware thread.

The e6500 core shares TLBs between hardware threads. Without tlb write
conditional instruction, the Linux kernel uses per core mechanisms to
protect against duplicate TLB entries.

The guest is unable to detect real siblings threads, so it can't use the
TLB protection mechanism. An alternative solution is to use the hypervisor
to allocate different lpids to guest's vcpus that runs simultaneous on real
siblings threads. On systems with two threads per core this patch halves
the size of the lpid pool that the allocator sees and use two lpids per VM.
Use even numbers to speedup vcpu lpid computation with consecutive lpids
per VM: vm1 will use lpids 2 and 3, vm2 lpids 4 and 5, and so on.

Signed-off-by: Mihai Caraman <mihai.caraman@freescale.com>
[agraf: fix spelling]
Signed-off-by: Alexander Graf <agraf@suse.de>

4 files changed, 65 insertions, 24 deletions

diff --git a/arch/powerpc/include/asm/kvm_booke.h b/arch/powerpc/include/asm/kvm_booke.h
index f7aa5cc395c4..630134d17f5d 100644
--- a/arch/powerpc/include/asm/kvm_booke.h
+++ b/arch/powerpc/include/asm/kvm_booke.h
@@ -23,7 +23,10 @@
 #include <linux/types.h>
 #include <linux/kvm_host.h>
 
-/* LPIDs we support with this build -- runtime limit may be lower */
+/*
+ * Number of available lpids. Only the low-order 6 bits of LPID rgister are
+ * implemented on e500mc+ cores.
+ */
 #define KVMPPC_NR_LPIDS                        64
 
 #define KVMPPC_INST_EHPRIV		0x7c00021c
diff --git a/arch/powerpc/kvm/e500.h b/arch/powerpc/kvm/e500.h
index a326178bdea5..72920bed3ac6 100644
--- a/arch/powerpc/kvm/e500.h
+++ b/arch/powerpc/kvm/e500.h
@@ -22,6 +22,7 @@
 #include <linux/kvm_host.h>
 #include <asm/mmu-book3e.h>
 #include <asm/tlb.h>
+#include <asm/cputhreads.h>
 
 enum vcpu_ftr {
 	VCPU_FTR_MMU_V2
@@ -289,6 +290,25 @@ void kvmppc_e500_tlbil_all(struct kvmppc_vcpu_e500 *vcpu_e500);
 #define kvmppc_e500_get_tlb_stid(vcpu, gtlbe)       get_tlb_tid(gtlbe)
 #define get_tlbmiss_tid(vcpu)           get_cur_pid(vcpu)
 #define get_tlb_sts(gtlbe)              (gtlbe->mas1 & MAS1_TS)
+
+/*
+ * These functions should be called with preemption disabled
+ * and the returned value is valid only in that context
+ */
+static inline int get_thread_specific_lpid(int vm_lpid)
+{
+	int vcpu_lpid = vm_lpid;
+
+	if (threads_per_core == 2)
+		vcpu_lpid |= smp_processor_id() & 1;
+
+	return vcpu_lpid;
+}
+
+static inline int get_lpid(struct kvm_vcpu *vcpu)
+{
+	return get_thread_specific_lpid(vcpu->kvm->arch.lpid);
+}
 #else
 unsigned int kvmppc_e500_get_tlb_stid(struct kvm_vcpu *vcpu,
 				      struct kvm_book3e_206_tlb_entry *gtlbe);
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
index 08f14bb57897..c8795a64e935 100644
--- a/arch/powerpc/kvm/e500_mmu_host.c
+++ b/arch/powerpc/kvm/e500_mmu_host.c
@@ -69,7 +69,8 @@ static inline u32 e500_shadow_mas3_attrib(u32 mas3, int usermode)
  * writing shadow tlb entry to host TLB
  */
 static inline void __write_host_tlbe(struct kvm_book3e_206_tlb_entry *stlbe,
-				     uint32_t mas0)
+				     uint32_t mas0,
+				     uint32_t lpid)
 {
 	unsigned long flags;
 
@@ -80,7 +81,7 @@ static inline void __write_host_tlbe(struct kvm_book3e_206_tlb_entry *stlbe,
 	mtspr(SPRN_MAS3, (u32)stlbe->mas7_3);
 	mtspr(SPRN_MAS7, (u32)(stlbe->mas7_3 >> 32));
 #ifdef CONFIG_KVM_BOOKE_HV
-	mtspr(SPRN_MAS8, stlbe->mas8);
+	mtspr(SPRN_MAS8, MAS8_TGS | get_thread_specific_lpid(lpid));
 #endif
 	asm volatile("isync; tlbwe" : : : "memory");
 
@@ -129,11 +130,12 @@ static inline void write_host_tlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
 
 	if (tlbsel == 0) {
 		mas0 = get_host_mas0(stlbe->mas2);
-		__write_host_tlbe(stlbe, mas0);
+		__write_host_tlbe(stlbe, mas0, vcpu_e500->vcpu.kvm->arch.lpid);
 	} else {
 		__write_host_tlbe(stlbe,
 				  MAS0_TLBSEL(1) |
-				  MAS0_ESEL(to_htlb1_esel(sesel)));
+				  MAS0_ESEL(to_htlb1_esel(sesel)),
+				  vcpu_e500->vcpu.kvm->arch.lpid);
 	}
 }
 
@@ -176,7 +178,7 @@ void kvmppc_map_magic(struct kvm_vcpu *vcpu)
 		       MAS3_SW | MAS3_SR | MAS3_UW | MAS3_UR;
 	magic.mas8 = 0;
 
-	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index));
+	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index), 0);
 	preempt_enable();
 }
 #endif
@@ -317,10 +319,6 @@ static void kvmppc_e500_setup_stlbe(
 	stlbe->mas2 = (gvaddr & MAS2_EPN) | (ref->flags & E500_TLB_MAS2_ATTR);
 	stlbe->mas7_3 = ((u64)pfn << PAGE_SHIFT) |
 			e500_shadow_mas3_attrib(gtlbe->mas7_3, pr);
-
-#ifdef CONFIG_KVM_BOOKE_HV
-	stlbe->mas8 = MAS8_TGS | vcpu->kvm->arch.lpid;
-#endif
 }
 
 static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
@@ -633,7 +631,7 @@ int kvmppc_load_last_inst(struct kvm_vcpu *vcpu, enum instruction_type type,
 
 	local_irq_save(flags);
 	mtspr(SPRN_MAS6, (vcpu->arch.pid << MAS6_SPID_SHIFT) | addr_space);
-	mtspr(SPRN_MAS5, MAS5_SGS | vcpu->kvm->arch.lpid);
+	mtspr(SPRN_MAS5, MAS5_SGS | get_lpid(vcpu));
 	asm volatile("tlbsx 0, %[geaddr]\n" : :
 		     [geaddr] "r" (geaddr));
 	mtspr(SPRN_MAS5, 0);
diff --git a/arch/powerpc/kvm/e500mc.c b/arch/powerpc/kvm/e500mc.c
index 454934990672..bf8f99f6676a 100644
--- a/arch/powerpc/kvm/e500mc.c
+++ b/arch/powerpc/kvm/e500mc.c
@@ -48,10 +48,11 @@ void kvmppc_set_pending_interrupt(struct kvm_vcpu *vcpu, enum int_class type)
 		return;
 	}
 
-
-	tag = PPC_DBELL_LPID(vcpu->kvm->arch.lpid) | vcpu->vcpu_id;
+	preempt_disable();
+	tag = PPC_DBELL_LPID(get_lpid(vcpu)) | vcpu->vcpu_id;
 	mb();
 	ppc_msgsnd(dbell_type, 0, tag);
+	preempt_enable();
 }
 
 /* gtlbe must not be mapped by more than one host tlb entry */
@@ -60,12 +61,11 @@ void kvmppc_e500_tlbil_one(struct kvmppc_vcpu_e500 *vcpu_e500,
 {
 	unsigned int tid, ts;
 	gva_t eaddr;
-	u32 val, lpid;
+	u32 val;
 	unsigned long flags;
 
 	ts = get_tlb_ts(gtlbe);
 	tid = get_tlb_tid(gtlbe);
-	lpid = vcpu_e500->vcpu.kvm->arch.lpid;
 
 	/* We search the host TLB to invalidate its shadow TLB entry */
 	val = (tid << 16) | ts;
@@ -74,7 +74,7 @@ void kvmppc_e500_tlbil_one(struct kvmppc_vcpu_e500 *vcpu_e500,
 	local_irq_save(flags);
 
 	mtspr(SPRN_MAS6, val);
-	mtspr(SPRN_MAS5, MAS5_SGS | lpid);
+	mtspr(SPRN_MAS5, MAS5_SGS | get_lpid(&vcpu_e500->vcpu));
 
 	asm volatile("tlbsx 0, %[eaddr]\n" : : [eaddr] "r" (eaddr));
 	val = mfspr(SPRN_MAS1);
@@ -95,7 +95,7 @@ void kvmppc_e500_tlbil_all(struct kvmppc_vcpu_e500 *vcpu_e500)
 	unsigned long flags;
 
 	local_irq_save(flags);
-	mtspr(SPRN_MAS5, MAS5_SGS | vcpu_e500->vcpu.kvm->arch.lpid);
+	mtspr(SPRN_MAS5, MAS5_SGS | get_lpid(&vcpu_e500->vcpu));
 	asm volatile("tlbilxlpid");
 	mtspr(SPRN_MAS5, 0);
 	local_irq_restore(flags);
@@ -110,6 +110,7 @@ void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
 {
 }
 
+/* We use two lpids per VM */
 static DEFINE_PER_CPU(struct kvm_vcpu *[KVMPPC_NR_LPIDS], last_vcpu_of_lpid);
 
 static void kvmppc_core_vcpu_load_e500mc(struct kvm_vcpu *vcpu, int cpu)
@@ -118,10 +119,12 @@ static void kvmppc_core_vcpu_load_e500mc(struct kvm_vcpu *vcpu, int cpu)
 
 	kvmppc_booke_vcpu_load(vcpu, cpu);
 
-	mtspr(SPRN_LPID, vcpu->kvm->arch.lpid);
+	mtspr(SPRN_LPID, get_lpid(vcpu));
 	mtspr(SPRN_EPCR, vcpu->arch.shadow_epcr);
 	mtspr(SPRN_GPIR, vcpu->vcpu_id);
 	mtspr(SPRN_MSRP, vcpu->arch.shadow_msrp);
+	vcpu->arch.eplc = EPC_EGS | (get_lpid(vcpu) << EPC_ELPID_SHIFT);
+	vcpu->arch.epsc = vcpu->arch.eplc;
 	mtspr(SPRN_EPLC, vcpu->arch.eplc);
 	mtspr(SPRN_EPSC, vcpu->arch.epsc);
 
@@ -141,9 +144,9 @@ static void kvmppc_core_vcpu_load_e500mc(struct kvm_vcpu *vcpu, int cpu)
 	mtspr(SPRN_GESR, vcpu->arch.shared->esr);
 
 	if (vcpu->arch.oldpir != mfspr(SPRN_PIR) ||
-	    __get_cpu_var(last_vcpu_of_lpid)[vcpu->kvm->arch.lpid] != vcpu) {
+	    __get_cpu_var(last_vcpu_of_lpid)[get_lpid(vcpu)] != vcpu) {
 		kvmppc_e500_tlbil_all(vcpu_e500);
-		__get_cpu_var(last_vcpu_of_lpid)[vcpu->kvm->arch.lpid] = vcpu;
+		__get_cpu_var(last_vcpu_of_lpid)[get_lpid(vcpu)] = vcpu;
 	}
 }
 
@@ -193,8 +196,6 @@ int kvmppc_core_vcpu_setup(struct kvm_vcpu *vcpu)
 	vcpu->arch.shadow_epcr |= SPRN_EPCR_ICM;
 #endif
 	vcpu->arch.shadow_msrp = MSRP_UCLEP | MSRP_PMMP;
-	vcpu->arch.eplc = EPC_EGS | (vcpu->kvm->arch.lpid << EPC_ELPID_SHIFT);
-	vcpu->arch.epsc = vcpu->arch.eplc;
 
 	vcpu->arch.pvr = mfspr(SPRN_PVR);
 	vcpu_e500->svr = mfspr(SPRN_SVR);
@@ -354,13 +355,26 @@ static int kvmppc_core_init_vm_e500mc(struct kvm *kvm)
 	if (lpid < 0)
 		return lpid;
 
+	/*
+	 * Use two lpids per VM on cores with two threads like e6500. Use
+	 * even numbers to speedup vcpu lpid computation with consecutive lpids
+	 * per VM. vm1 will use lpids 2 and 3, vm2 lpids 4 and 5, and so on.
+	 */
+	if (threads_per_core == 2)
+		lpid <<= 1;
+
 	kvm->arch.lpid = lpid;
 	return 0;
 }
 
 static void kvmppc_core_destroy_vm_e500mc(struct kvm *kvm)
 {
-	kvmppc_free_lpid(kvm->arch.lpid);
+	int lpid = kvm->arch.lpid;
+
+	if (threads_per_core == 2)
+		lpid >>= 1;
+
+	kvmppc_free_lpid(lpid);
 }
 
 static struct kvmppc_ops kvm_ops_e500mc = {
@@ -388,7 +402,13 @@ static int __init kvmppc_e500mc_init(void)
 	if (r)
 		goto err_out;
 
-	kvmppc_init_lpid(64);
+	/*
+	 * Use two lpids per VM on dual threaded processors like e6500
+	 * to workarround the lack of tlb write conditional instruction.
+	 * Expose half the number of available hardware lpids to the lpid
+	 * allocator.
+	 */
+	kvmppc_init_lpid(KVMPPC_NR_LPIDS/threads_per_core);
 	kvmppc_claim_lpid(0); /* host */
 
 	r = kvm_init(NULL, sizeof(struct kvmppc_vcpu_e500), 0, THIS_MODULE);