[PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Scott Wood]

From: Liu Yu <yu.liu@freescale.com>

This patch utilize PID1 to map guest ID to shadow ID,
so that we can combine all needs to invalidate host TLB together,
and therefor reduce the frequency of calling _tlbia().

Signed-off-by: Liu Yu <yu.liu@freescale.com>
Signed-off-by: Scott Wood <scottwood@freescale.com>
---
 arch/powerpc/include/asm/kvm_e500.h |    8 +-
 arch/powerpc/include/asm/kvm_host.h |    1 +
 arch/powerpc/kernel/asm-offsets.c   |    1 +
 arch/powerpc/kvm/44x_tlb.c          |    4 +-
 arch/powerpc/kvm/booke.c            |   11 +-
 arch/powerpc/kvm/booke.h            |    1 +
 arch/powerpc/kvm/booke_interrupts.S |   11 ++
 arch/powerpc/kvm/e500_tlb.c         |  275 +++++++++++++++++++++++++++++++---
 arch/powerpc/kvm/e500_tlb.h         |   13 ++-
 9 files changed, 293 insertions(+), 32 deletions(-)

diff --git a/arch/powerpc/include/asm/kvm_e500.h b/arch/powerpc/include/asm/kvm_e500.h
index 4a6d77a..adbfca9 100644
--- a/arch/powerpc/include/asm/kvm_e500.h
+++ b/arch/powerpc/include/asm/kvm_e500.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 Freescale Semiconductor, Inc. All rights reserved.
+ * Copyright (C) 2008-2011 Freescale Semiconductor, Inc. All rights reserved.
  *
  * Author: Yu Liu, <yu.liu@freescale.com>
  *
@@ -37,6 +37,8 @@ struct tlbe_priv {
 	unsigned int flags; /* E500_TLB_* */
 };
 
+struct vcpu_id_table;
+
 struct kvmppc_vcpu_e500 {
 	/* Unmodified copy of the guest's TLB. */
 	struct tlbe *gtlb_arch[E500_TLB_NUM];
@@ -59,6 +61,10 @@ struct kvmppc_vcpu_e500 {
 	u32 mas5;
 	u32 mas6;
 	u32 mas7;
+
+	/* vcpu id table */
+	struct vcpu_id_table *idt;
+
 	u32 l1csr0;
 	u32 l1csr1;
 	u32 hid0;
diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
index c4ce105..6e05b2d 100644
--- a/arch/powerpc/include/asm/kvm_host.h
+++ b/arch/powerpc/include/asm/kvm_host.h
@@ -255,6 +255,7 @@ struct kvm_vcpu_arch {
 	u32 pvr;
 
 	u32 shadow_pid;
+	u32 shadow_pid1;
 	u32 pid;
 	u32 swap_pid;
 
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 7b9e35e..d6c439c 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -401,6 +401,7 @@ int main(void)
 	DEFINE(VCPU_SPRG6, offsetof(struct kvm_vcpu, arch.sprg6));
 	DEFINE(VCPU_SPRG7, offsetof(struct kvm_vcpu, arch.sprg7));
 	DEFINE(VCPU_SHADOW_PID, offsetof(struct kvm_vcpu, arch.shadow_pid));
+	DEFINE(VCPU_SHADOW_PID1, offsetof(struct kvm_vcpu, arch.shadow_pid1));
 	DEFINE(VCPU_SHARED, offsetof(struct kvm_vcpu, arch.shared));
 	DEFINE(VCPU_SHARED_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
 	DEFINE(VCPU_SHADOW_MSR, offsetof(struct kvm_vcpu, arch.shadow_msr));
diff --git a/arch/powerpc/kvm/44x_tlb.c b/arch/powerpc/kvm/44x_tlb.c
index 5f3cff8..33aa715 100644
--- a/arch/powerpc/kvm/44x_tlb.c
+++ b/arch/powerpc/kvm/44x_tlb.c
@@ -387,8 +387,10 @@ static void kvmppc_44x_invalidate(struct kvm_vcpu *vcpu,
 	}
 }
 
-void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
 {
+	int usermode = vcpu->arch.shared->msr & MSR_PR;
+
 	vcpu->arch.shadow_pid = !usermode;
 }
 
diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
index 036625e..1de0dc5 100644
--- a/arch/powerpc/kvm/booke.c
+++ b/arch/powerpc/kvm/booke.c
@@ -113,15 +113,18 @@ static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 }
 #endif
 
-/* Helper function for "full" MSR writes. No need to call this if only EE is
- * changing. */
+/*
+ * Helper function for "full" MSR writes.  No need to call this if only
+ * EE/CE/ME/DE/RI are changing.
+ */
 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 {
-	if ((new_msr & MSR_PR) != (vcpu->arch.shared->msr & MSR_PR))
-		kvmppc_mmu_priv_switch(vcpu, new_msr & MSR_PR);
+	u32 old_msr = vcpu->arch.shared->msr;
 
 	vcpu->arch.shared->msr = new_msr;
 
+	kvmppc_mmu_msr_notify(vcpu, old_msr);
+
 	if (vcpu->arch.shared->msr & MSR_WE) {
 		kvm_vcpu_block(vcpu);
 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
diff --git a/arch/powerpc/kvm/booke.h b/arch/powerpc/kvm/booke.h
index 0fa1732..8e1fe33 100644
--- a/arch/powerpc/kvm/booke.h
+++ b/arch/powerpc/kvm/booke.h
@@ -53,6 +53,7 @@
 extern unsigned long kvmppc_booke_handlers;
 
 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr);
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr);
 
 int kvmppc_booke_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
                             unsigned int inst, int *advance);
diff --git a/arch/powerpc/kvm/booke_interrupts.S b/arch/powerpc/kvm/booke_interrupts.S
index 8cb3dfe..42f2fb1 100644
--- a/arch/powerpc/kvm/booke_interrupts.S
+++ b/arch/powerpc/kvm/booke_interrupts.S
@@ -191,6 +191,12 @@ _GLOBAL(kvmppc_resume_host)
 	lwz	r3, VCPU_HOST_PID(r4)
 	mtspr	SPRN_PID, r3
 
+#ifdef CONFIG_FSL_BOOKE
+	/* we cheat and know that Linux doesn't use PID1 which is always 0 */
+	lis	r3, 0
+	mtspr	SPRN_PID1, r3
+#endif
+
 	/* Restore host IVPR before re-enabling interrupts. We cheat and know
 	 * that Linux IVPR is always 0xc0000000. */
 	lis	r3, 0xc000
@@ -365,6 +371,11 @@ lightweight_exit:
 	lwz	r3, VCPU_SHADOW_PID(r4)
 	mtspr	SPRN_PID, r3
 
+#ifdef CONFIG_FSL_BOOKE
+	lwz	r3, VCPU_SHADOW_PID1(r4)
+	mtspr	SPRN_PID1, r3
+#endif
+
 #ifdef CONFIG_44x
 	iccci	0, 0 /* XXX hack */
 #endif
diff --git a/arch/powerpc/kvm/e500_tlb.c b/arch/powerpc/kvm/e500_tlb.c
index 9d1e28d..e7f1951 100644
--- a/arch/powerpc/kvm/e500_tlb.c
+++ b/arch/powerpc/kvm/e500_tlb.c
@@ -28,8 +28,167 @@
 
 #define to_htlb1_esel(esel) (tlb1_entry_num - (esel) - 1)
 
+struct id {
+	unsigned long val;
+	struct id **pentry;
+};
+
+/*
+ * This table provide mappings from:
+ * (guestAS,guestTID,guestPR) --> ID of physical cpu
+ * guestAS	[0..1]
+ * guestTID	[0..255]
+ * guestPR	[0..1]
+ * ID		[1..255]
+ * Each vcpu keeps one vcpu_id_table.
+ */
+struct vcpu_id_table {
+	struct id id[2][256][2];
+};
+
+/*
+ * This table provide reversed mappings of vcpu_id_table:
+ * ID --> address of vcpu_id_table item.
+ * Each physical core has one pcpu_id_table.
+ */
+struct pcpu_id_table {
+	struct id *entry[256];
+};
+
+static DEFINE_PER_CPU(struct pcpu_id_table, pcpu_sids);
+
+/* This variable keeps last used shadow ID on local core.
+ * The valid range of shadow ID is [1..255] */
+static DEFINE_PER_CPU(unsigned long, pcpu_last_used_sid);
+
 static unsigned int tlb1_entry_num;
 
+/*
+ * Allocate a free shadow id and setup a valid sid mapping in given entry.
+ * A mapping is only valid when vcpu_id_table and pcpu_id_table are match.
+ */
+static inline int local_sid_setup_one(struct id *entry)
+{
+	unsigned long sid;
+	int ret = -1;
+
+	sid = ++(__get_cpu_var(pcpu_last_used_sid));
+	if (sid < 256) {
+		__get_cpu_var(pcpu_sids).entry[sid] = entry;
+		entry->val = sid;
+		entry->pentry = &__get_cpu_var(pcpu_sids).entry[sid];
+		ret = sid;
+	}
+
+	WARN_ON(sid > 256);
+
+	return ret;
+}
+
+/*
+ * Check if given entry contain a valid shadow id mapping.
+ * An ID mapping is considered valid only if
+ * both vcpu and pcpu know this mapping.
+ */
+static inline int local_sid_lookup(struct id *entry)
+{
+	if (entry && entry->val != 0 &&
+	    __get_cpu_var(pcpu_sids).entry[entry->val] = entry &&
+	    entry->pentry = &__get_cpu_var(pcpu_sids).entry[entry->val])
+		return entry->val;
+	return -1;
+}
+
+/* Invalidate all id mappings on local core */
+static inline void local_sid_destroy_all(void)
+{
+	preempt_disable();
+	__get_cpu_var(pcpu_last_used_sid) = 0;
+	memset(&__get_cpu_var(pcpu_sids), 0, sizeof(__get_cpu_var(pcpu_sids)));
+	preempt_enable();
+}
+
+static void *kvmppc_e500_id_table_alloc(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	vcpu_e500->idt = kzalloc(sizeof(struct vcpu_id_table), GFP_KERNEL);
+	return vcpu_e500->idt;
+}
+
+static void kvmppc_e500_id_table_free(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	kfree(vcpu_e500->idt);
+}
+
+/* Invalidate all mappings on vcpu */
+static void kvmppc_e500_id_table_reset_all(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	memset(vcpu_e500->idt, 0, sizeof(struct vcpu_id_table));
+
+	/* Update shadow pid when mappings are changed */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+}
+
+/* Invalidate one ID mapping on vcpu */
+static inline void kvmppc_e500_id_table_reset_one(
+			       struct kvmppc_vcpu_e500 *vcpu_e500,
+			       int as, int pid, int pr)
+{
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+
+	idt->id[as][pid][pr].val = 0;
+	idt->id[as][pid][pr].pentry = NULL;
+
+	/* Update shadow pid when mappings are changed */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+}
+
+/*
+ * Map guest (vcpu,AS,ID,PR) to physical core shadow id.
+ * This function first lookup if a valid mapping exists,
+ * if not, then creates a new one.
+ */
+static unsigned int kvmppc_e500_get_sid(struct kvmppc_vcpu_e500 *vcpu_e500,
+					int as, int gid, int pr,
+					int avoid_recursion)
+{
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+	int sid;
+
+	sid = local_sid_lookup(&idt->id[as][gid][pr]);
+
+	while (sid <= 0) {
+		/* No mapping yet */
+		sid = local_sid_setup_one(&idt->id[as][gid][pr]);
+		if (sid <= 0) {
+			_tlbil_all();
+			local_sid_destroy_all();
+		}
+
+		/* Update shadow pid when mappings are changed */
+		if (!avoid_recursion)
+			kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+	}
+
+	return sid;
+}
+
+/* Map guest pid to shadow.
+ * We use PID to keep shadow of current guest non-zero PID,
+ * and use PID1 to keep shadow of guest zero PID.
+ * So that guest tlbe with TID=0 can be accessed at any time */
+void kvmppc_e500_recalc_shadow_pid(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	preempt_disable();
+	vcpu_e500->vcpu.arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500,
+			get_cur_as(&vcpu_e500->vcpu),
+			get_cur_pid(&vcpu_e500->vcpu),
+			get_cur_pr(&vcpu_e500->vcpu), 1);
+	vcpu_e500->vcpu.arch.shadow_pid1 = kvmppc_e500_get_sid(vcpu_e500,
+			get_cur_as(&vcpu_e500->vcpu), 0,
+			get_cur_pr(&vcpu_e500->vcpu), 1);
+	preempt_enable();
+}
+
 void kvmppc_dump_tlbs(struct kvm_vcpu *vcpu)
 {
 	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
@@ -134,14 +293,19 @@ static inline void write_host_tlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
 
 void kvmppc_map_magic(struct kvm_vcpu *vcpu)
 {
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
 	struct tlbe magic;
 	ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
+	unsigned int stid;
 	pfn_t pfn;
 
 	pfn = (pfn_t)virt_to_phys((void *)shared_page) >> PAGE_SHIFT;
 	get_page(pfn_to_page(pfn));
 
-	magic.mas1 = MAS1_VALID | MAS1_TS |
+	preempt_disable();
+	stid = kvmppc_e500_get_sid(vcpu_e500, 0, 0, 0, 0);
+
+	magic.mas1 = MAS1_VALID | MAS1_TS | MAS1_TID(stid) |
 		     MAS1_TSIZE(BOOK3E_PAGESZ_4K);
 	magic.mas2 = vcpu->arch.magic_page_ea | MAS2_M;
 	magic.mas3 = (pfn << PAGE_SHIFT) |
@@ -149,15 +313,68 @@ void kvmppc_map_magic(struct kvm_vcpu *vcpu)
 	magic.mas7 = pfn >> (32 - PAGE_SHIFT);
 
 	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index));
+	preempt_enable();
 }
 
 void kvmppc_e500_tlb_load(struct kvm_vcpu *vcpu, int cpu)
 {
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+
+	/* Shadow PID may be expired on local core */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
 }
 
 void kvmppc_e500_tlb_put(struct kvm_vcpu *vcpu)
 {
-	_tlbil_all();
+}
+
+static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
+					 int tlbsel, int esel)
+{
+	struct tlbe *gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+	unsigned int pr, tid, ts, pid;
+	u32 val, eaddr;
+	unsigned long flags;
+
+	ts = get_tlb_ts(gtlbe);
+	tid = get_tlb_tid(gtlbe);
+
+	/* One guest ID may be mapped to two shadow IDs */
+	for (pr = 0; pr < 2; pr++) {
+		/*
+		 * If we don't find a valid ID mapping, maybe means
+		 * it doesn't have a valid ID mapping on local core,
+		 * and has valid ID mapping on other cores. We should
+		 * clear this ID mapping to disable related shadow TLBe.
+		 *
+		 * And we always clear corresponding ID mapping
+		 * if guest is invalidating a TLB1 entry.
+		 */
+		if (tlbsel = 1 ||
+		    (pid = local_sid_lookup(&idt->id[ts][tid][pr])) <= 0) {
+			kvmppc_e500_id_table_reset_one(vcpu_e500, ts, tid, pr);
+			continue;
+		}
+
+		/* Guest is invalidating a TLB0 entry
+		 * which has a valid ID mapping.
+		 * We search host TLB0 to invalidate it's shadow TLBe */
+		val = (pid << 16) | MAS6_SAS;
+		eaddr = get_tlb_eaddr(gtlbe);
+
+		local_irq_save(flags);
+
+		mtspr(SPRN_MAS6, val);
+		asm volatile("tlbsx 0, %[eaddr]" : : [eaddr] "r" (eaddr));
+		val = mfspr(SPRN_MAS1);
+		if (val & MAS1_VALID) {
+			mtspr(SPRN_MAS1, val & ~MAS1_VALID);
+			asm volatile("tlbwe");
+		}
+
+		local_irq_restore(flags);
+	}
 }
 
 /* Search the guest TLB for a matching entry. */
@@ -216,12 +433,6 @@ static inline void kvmppc_e500_priv_release(struct tlbe_priv *priv)
 	}
 }
 
-static void kvmppc_e500_tlb1_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
-					int esel)
-{
-	mtspr(SPRN_MMUCSR0, MMUCSR0_TLB1FI);
-}
-
 static inline void kvmppc_e500_deliver_tlb_miss(struct kvm_vcpu *vcpu,
 		unsigned int eaddr, int as)
 {
@@ -255,10 +466,15 @@ static inline void kvmppc_e500_setup_stlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
 					   u64 gvaddr, struct tlbe *stlbe)
 {
 	pfn_t pfn = priv->pfn;
+	unsigned int stid;
+
+	stid = kvmppc_e500_get_sid(vcpu_e500, get_tlb_ts(gtlbe),
+				   get_tlb_tid(gtlbe),
+				   get_cur_pr(&vcpu_e500->vcpu), 0);
 
 	/* Force TS=1 IPROT=0 for all guest mappings. */
 	stlbe->mas1 = MAS1_TSIZE(tsize)
-		| MAS1_TID(get_tlb_tid(gtlbe)) | MAS1_TS | MAS1_VALID;
+		| MAS1_TID(stid) | MAS1_TS | MAS1_VALID;
 	stlbe->mas2 = (gvaddr & MAS2_EPN)
 		| e500_shadow_mas2_attrib(gtlbe->mas2,
 				vcpu_e500->vcpu.arch.shared->msr & MSR_PR);
@@ -414,14 +630,12 @@ static int kvmppc_e500_tlb1_map(struct kvmppc_vcpu_e500 *vcpu_e500,
 	return victim;
 }
 
-/* Invalidate all guest kernel mappings when enter usermode,
- * so that when they fault back in they will get the
- * proper permission bits. */
-void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
 {
-	if (usermode) {
-		_tlbil_all();
-	}
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+
+	/* Recalc shadow pid since MSR changes */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
 }
 
 static inline int kvmppc_e500_gtlbe_invalidate(
@@ -449,7 +663,8 @@ int kvmppc_e500_emul_mt_mmucsr0(struct kvmppc_vcpu_e500 *vcpu_e500, ulong value)
 		for (esel = 0; esel < vcpu_e500->gtlb_size[1]; esel++)
 			kvmppc_e500_gtlbe_invalidate(vcpu_e500, 1, esel);
 
-	_tlbil_all();
+	/* Invalidate all vcpu id mappings */
+	kvmppc_e500_id_table_reset_all(vcpu_e500);
 
 	return EMULATE_DONE;
 }
@@ -480,7 +695,8 @@ int kvmppc_e500_emul_tlbivax(struct kvm_vcpu *vcpu, int ra, int rb)
 			kvmppc_e500_gtlbe_invalidate(vcpu_e500, tlbsel, esel);
 	}
 
-	_tlbil_all();
+	/* Invalidate all vcpu id mappings */
+	kvmppc_e500_id_table_reset_all(vcpu_e500);
 
 	return EMULATE_DONE;
 }
@@ -564,8 +780,8 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 
 	gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
 
-	if (get_tlb_v(gtlbe) && tlbsel = 1)
-		kvmppc_e500_tlb1_invalidate(vcpu_e500, esel);
+	if (get_tlb_v(gtlbe))
+		kvmppc_e500_stlbe_invalidate(vcpu_e500, tlbsel, esel);
 
 	gtlbe->mas1 = vcpu_e500->mas1;
 	gtlbe->mas2 = vcpu_e500->mas2;
@@ -582,6 +798,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 		u64 eaddr;
 		u64 raddr;
 
+		preempt_disable();
 		switch (tlbsel) {
 		case 0:
 			/* TLB0 */
@@ -611,6 +828,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 			BUG();
 		}
 		write_host_tlbe(vcpu_e500, stlbsel, sesel, &stlbe);
+		preempt_enable();
 	}
 
 	kvmppc_set_exit_type(vcpu, EMULATED_TLBWE_EXITS);
@@ -672,6 +890,7 @@ void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 eaddr, gpa_t gpaddr,
 
 	gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
 
+	preempt_disable();
 	switch (tlbsel) {
 	case 0:
 		stlbsel = 0;
@@ -697,6 +916,7 @@ void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 eaddr, gpa_t gpaddr,
 	}
 
 	write_host_tlbe(vcpu_e500, stlbsel, sesel, &stlbe);
+	preempt_enable();
 }
 
 int kvmppc_e500_tlb_search(struct kvm_vcpu *vcpu,
@@ -718,8 +938,10 @@ void kvmppc_set_pid(struct kvm_vcpu *vcpu, u32 pid)
 {
 	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
 
-	vcpu_e500->pid[0] = vcpu->arch.shadow_pid -		vcpu->arch.pid = pid;
+	if (vcpu->arch.pid != pid) {
+		vcpu_e500->pid[0] = vcpu->arch.pid = pid;
+		kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+	}
 }
 
 void kvmppc_e500_tlb_setup(struct kvmppc_vcpu_e500 *vcpu_e500)
@@ -767,6 +989,9 @@ int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
 	if (vcpu_e500->gtlb_priv[1] = NULL)
 		goto err_out_priv0;
 
+	if (kvmppc_e500_id_table_alloc(vcpu_e500) = NULL)
+		goto err_out_priv1;
+
 	/* Init TLB configuration register */
 	vcpu_e500->tlb0cfg = mfspr(SPRN_TLB0CFG) & ~0xfffUL;
 	vcpu_e500->tlb0cfg |= vcpu_e500->gtlb_size[0];
@@ -775,6 +1000,8 @@ int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
 
 	return 0;
 
+err_out_priv1:
+	kfree(vcpu_e500->gtlb_priv[1]);
 err_out_priv0:
 	kfree(vcpu_e500->gtlb_priv[0]);
 err_out_guest1:
@@ -797,9 +1024,7 @@ void kvmppc_e500_tlb_uninit(struct kvmppc_vcpu_e500 *vcpu_e500)
 			kvmppc_e500_priv_release(priv);
 		}
 
-	/* discard all guest mapping */
-	_tlbil_all();
-
+	kvmppc_e500_id_table_free(vcpu_e500);
 	kfree(vcpu_e500->gtlb_arch[1]);
 	kfree(vcpu_e500->gtlb_arch[0]);
 }
diff --git a/arch/powerpc/kvm/e500_tlb.h b/arch/powerpc/kvm/e500_tlb.h
index 458946b..59b88e9 100644
--- a/arch/powerpc/kvm/e500_tlb.h
+++ b/arch/powerpc/kvm/e500_tlb.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 Freescale Semiconductor, Inc. All rights reserved.
+ * Copyright (C) 2008-2011 Freescale Semiconductor, Inc. All rights reserved.
  *
  * Author: Yu Liu, yu.liu@freescale.com
  *
@@ -55,6 +55,7 @@ extern void kvmppc_e500_tlb_load(struct kvm_vcpu *, int);
 extern int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *);
 extern void kvmppc_e500_tlb_uninit(struct kvmppc_vcpu_e500 *);
 extern void kvmppc_e500_tlb_setup(struct kvmppc_vcpu_e500 *);
+extern void kvmppc_e500_recalc_shadow_pid(struct kvmppc_vcpu_e500 *);
 
 /* TLB helper functions */
 static inline unsigned int get_tlb_size(const struct tlbe *tlbe)
@@ -110,6 +111,16 @@ static inline unsigned int get_cur_pid(struct kvm_vcpu *vcpu)
 	return vcpu->arch.pid & 0xff;
 }
 
+static inline unsigned int get_cur_as(struct kvm_vcpu *vcpu)
+{
+	return !!(vcpu->arch.shared->msr & (MSR_IS | MSR_DS));
+}
+
+static inline unsigned int get_cur_pr(struct kvm_vcpu *vcpu)
+{
+	return !!(vcpu->arch.shared->msr & MSR_PR);
+}
+
 static inline unsigned int get_cur_spid(
 		const struct kvmppc_vcpu_e500 *vcpu_e500)
 {
-- 
1.7.4.1

Re: [PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Alexander Graf]

On 05/18/2011 01:42 AM, Scott Wood wrote:
> From: Liu Yu<yu.liu@freescale.com>
>
> This patch utilize PID1 to map guest ID to shadow ID,
> so that we can combine all needs to invalidate host TLB together,
> and therefor reduce the frequency of calling _tlbia().
>
> Signed-off-by: Liu Yu<yu.liu@freescale.com>
> Signed-off-by: Scott Wood<scottwood@freescale.com>
> ---
>   arch/powerpc/include/asm/kvm_e500.h |    8 +-
>   arch/powerpc/include/asm/kvm_host.h |    1 +
>   arch/powerpc/kernel/asm-offsets.c   |    1 +
>   arch/powerpc/kvm/44x_tlb.c          |    4 +-
>   arch/powerpc/kvm/booke.c            |   11 +-
>   arch/powerpc/kvm/booke.h            |    1 +
>   arch/powerpc/kvm/booke_interrupts.S |   11 ++
>   arch/powerpc/kvm/e500_tlb.c         |  275 +++++++++++++++++++++++++++++++---
>   arch/powerpc/kvm/e500_tlb.h         |   13 ++-
>   9 files changed, 293 insertions(+), 32 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/kvm_e500.h b/arch/powerpc/include/asm/kvm_e500.h
> index 4a6d77a..adbfca9 100644
> --- a/arch/powerpc/include/asm/kvm_e500.h
> +++ b/arch/powerpc/include/asm/kvm_e500.h
> @@ -1,5 +1,5 @@
>   /*
> - * Copyright (C) 2008 Freescale Semiconductor, Inc. All rights reserved.
> + * Copyright (C) 2008-2011 Freescale Semiconductor, Inc. All rights reserved.
>    *
>    * Author: Yu Liu,<yu.liu@freescale.com>
>    *
> @@ -37,6 +37,8 @@ struct tlbe_priv {
>   	unsigned int flags; /* E500_TLB_* */
>   };
>
> +struct vcpu_id_table;
> +
>   struct kvmppc_vcpu_e500 {
>   	/* Unmodified copy of the guest's TLB. */
>   	struct tlbe *gtlb_arch[E500_TLB_NUM];
> @@ -59,6 +61,10 @@ struct kvmppc_vcpu_e500 {
>   	u32 mas5;
>   	u32 mas6;
>   	u32 mas7;
> +
> +	/* vcpu id table */
> +	struct vcpu_id_table *idt;
> +
>   	u32 l1csr0;
>   	u32 l1csr1;
>   	u32 hid0;
> diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
> index c4ce105..6e05b2d 100644
> --- a/arch/powerpc/include/asm/kvm_host.h
> +++ b/arch/powerpc/include/asm/kvm_host.h
> @@ -255,6 +255,7 @@ struct kvm_vcpu_arch {
>   	u32 pvr;
>
>   	u32 shadow_pid;
> +	u32 shadow_pid1;
>   	u32 pid;
>   	u32 swap_pid;
>
> diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
> index 7b9e35e..d6c439c 100644
> --- a/arch/powerpc/kernel/asm-offsets.c
> +++ b/arch/powerpc/kernel/asm-offsets.c
> @@ -401,6 +401,7 @@ int main(void)
>   	DEFINE(VCPU_SPRG6, offsetof(struct kvm_vcpu, arch.sprg6));
>   	DEFINE(VCPU_SPRG7, offsetof(struct kvm_vcpu, arch.sprg7));
>   	DEFINE(VCPU_SHADOW_PID, offsetof(struct kvm_vcpu, arch.shadow_pid));
> +	DEFINE(VCPU_SHADOW_PID1, offsetof(struct kvm_vcpu, arch.shadow_pid1));
>   	DEFINE(VCPU_SHARED, offsetof(struct kvm_vcpu, arch.shared));
>   	DEFINE(VCPU_SHARED_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
>   	DEFINE(VCPU_SHADOW_MSR, offsetof(struct kvm_vcpu, arch.shadow_msr));
> diff --git a/arch/powerpc/kvm/44x_tlb.c b/arch/powerpc/kvm/44x_tlb.c
> index 5f3cff8..33aa715 100644
> --- a/arch/powerpc/kvm/44x_tlb.c
> +++ b/arch/powerpc/kvm/44x_tlb.c
> @@ -387,8 +387,10 @@ static void kvmppc_44x_invalidate(struct kvm_vcpu *vcpu,
>   	}
>   }
>
> -void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
> +void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
>   {
> +	int usermode = vcpu->arch.shared->msr&  MSR_PR;
> +
>   	vcpu->arch.shadow_pid = !usermode;
>   }
>
> diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
> index 036625e..1de0dc5 100644
> --- a/arch/powerpc/kvm/booke.c
> +++ b/arch/powerpc/kvm/booke.c
> @@ -113,15 +113,18 @@ static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
>   }
>   #endif
>
> -/* Helper function for "full" MSR writes. No need to call this if only EE is
> - * changing. */
> +/*
> + * Helper function for "full" MSR writes.  No need to call this if only
> + * EE/CE/ME/DE/RI are changing.
> + */
>   void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
>   {
> -	if ((new_msr&  MSR_PR) != (vcpu->arch.shared->msr&  MSR_PR))
> -		kvmppc_mmu_priv_switch(vcpu, new_msr&  MSR_PR);
> +	u32 old_msr = vcpu->arch.shared->msr;
>
>   	vcpu->arch.shared->msr = new_msr;
>
> +	kvmppc_mmu_msr_notify(vcpu, old_msr);
> +
>   	if (vcpu->arch.shared->msr&  MSR_WE) {
>   		kvm_vcpu_block(vcpu);
>   		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
> diff --git a/arch/powerpc/kvm/booke.h b/arch/powerpc/kvm/booke.h
> index 0fa1732..8e1fe33 100644
> --- a/arch/powerpc/kvm/booke.h
> +++ b/arch/powerpc/kvm/booke.h
> @@ -53,6 +53,7 @@
>   extern unsigned long kvmppc_booke_handlers;
>
>   void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr);
> +void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr);
>
>   int kvmppc_booke_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
>                               unsigned int inst, int *advance);
> diff --git a/arch/powerpc/kvm/booke_interrupts.S b/arch/powerpc/kvm/booke_interrupts.S
> index 8cb3dfe..42f2fb1 100644
> --- a/arch/powerpc/kvm/booke_interrupts.S
> +++ b/arch/powerpc/kvm/booke_interrupts.S
> @@ -191,6 +191,12 @@ _GLOBAL(kvmppc_resume_host)
>   	lwz	r3, VCPU_HOST_PID(r4)
>   	mtspr	SPRN_PID, r3
>
> +#ifdef CONFIG_FSL_BOOKE
> +	/* we cheat and know that Linux doesn't use PID1 which is always 0 */
> +	lis	r3, 0
> +	mtspr	SPRN_PID1, r3
> +#endif
> +
>   	/* Restore host IVPR before re-enabling interrupts. We cheat and know
>   	 * that Linux IVPR is always 0xc0000000. */
>   	lis	r3, 0xc000
> @@ -365,6 +371,11 @@ lightweight_exit:
>   	lwz	r3, VCPU_SHADOW_PID(r4)
>   	mtspr	SPRN_PID, r3
>
> +#ifdef CONFIG_FSL_BOOKE
> +	lwz	r3, VCPU_SHADOW_PID1(r4)
> +	mtspr	SPRN_PID1, r3
> +#endif
> +
>   #ifdef CONFIG_44x
>   	iccci	0, 0 /* XXX hack */
>   #endif
> diff --git a/arch/powerpc/kvm/e500_tlb.c b/arch/powerpc/kvm/e500_tlb.c
> index 9d1e28d..e7f1951 100644
> --- a/arch/powerpc/kvm/e500_tlb.c
> +++ b/arch/powerpc/kvm/e500_tlb.c
> @@ -28,8 +28,167 @@
>
>   #define to_htlb1_esel(esel) (tlb1_entry_num - (esel) - 1)
>
> +struct id {
> +	unsigned long val;
> +	struct id **pentry;
> +};
> +
> +/*
> + * This table provide mappings from:
> + * (guestAS,guestTID,guestPR) -->  ID of physical cpu
> + * guestAS	[0..1]
> + * guestTID	[0..255]
> + * guestPR	[0..1]
> + * ID		[1..255]
> + * Each vcpu keeps one vcpu_id_table.
> + */
> +struct vcpu_id_table {
> +	struct id id[2][256][2];
> +};
> +
> +/*
> + * This table provide reversed mappings of vcpu_id_table:
> + * ID -->  address of vcpu_id_table item.
> + * Each physical core has one pcpu_id_table.
> + */
> +struct pcpu_id_table {
> +	struct id *entry[256];
> +};
> +
> +static DEFINE_PER_CPU(struct pcpu_id_table, pcpu_sids);
> +
> +/* This variable keeps last used shadow ID on local core.
> + * The valid range of shadow ID is [1..255] */
> +static DEFINE_PER_CPU(unsigned long, pcpu_last_used_sid);
> +
>   static unsigned int tlb1_entry_num;
>
> +/*
> + * Allocate a free shadow id and setup a valid sid mapping in given entry.
> + * A mapping is only valid when vcpu_id_table and pcpu_id_table are match.
> + */

Please add in the comment that this runs with preemption disabled. Same 
for the other helpers below.

> +static inline int local_sid_setup_one(struct id *entry)
> +{
> +	unsigned long sid;
> +	int ret = -1;
> +
> +	sid = ++(__get_cpu_var(pcpu_last_used_sid));
> +	if (sid<  256) {
> +		__get_cpu_var(pcpu_sids).entry[sid] = entry;
> +		entry->val = sid;
> +		entry->pentry =&__get_cpu_var(pcpu_sids).entry[sid];
> +		ret = sid;
> +	}
> +
> +	WARN_ON(sid>  256);

Why > 256?

> +
> +	return ret;
> +}
> +
> +/*
> + * Check if given entry contain a valid shadow id mapping.
> + * An ID mapping is considered valid only if
> + * both vcpu and pcpu know this mapping.
> + */
> +static inline int local_sid_lookup(struct id *entry)
> +{
> +	if (entry&&  entry->val != 0&&
> +	    __get_cpu_var(pcpu_sids).entry[entry->val] = entry&&
> +	    entry->pentry =&__get_cpu_var(pcpu_sids).entry[entry->val])
> +		return entry->val;
> +	return -1;
> +}
> +
> +/* Invalidate all id mappings on local core */
> +static inline void local_sid_destroy_all(void)
> +{
> +	preempt_disable();
> +	__get_cpu_var(pcpu_last_used_sid) = 0;
> +	memset(&__get_cpu_var(pcpu_sids), 0, sizeof(__get_cpu_var(pcpu_sids)));
> +	preempt_enable();
> +}
> +
> +static void *kvmppc_e500_id_table_alloc(struct kvmppc_vcpu_e500 *vcpu_e500)
> +{
> +	vcpu_e500->idt = kzalloc(sizeof(struct vcpu_id_table), GFP_KERNEL);
> +	return vcpu_e500->idt;
> +}
> +
> +static void kvmppc_e500_id_table_free(struct kvmppc_vcpu_e500 *vcpu_e500)
> +{
> +	kfree(vcpu_e500->idt);
> +}
> +
> +/* Invalidate all mappings on vcpu */
> +static void kvmppc_e500_id_table_reset_all(struct kvmppc_vcpu_e500 *vcpu_e500)
> +{
> +	memset(vcpu_e500->idt, 0, sizeof(struct vcpu_id_table));
> +
> +	/* Update shadow pid when mappings are changed */
> +	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
> +}
> +
> +/* Invalidate one ID mapping on vcpu */
> +static inline void kvmppc_e500_id_table_reset_one(
> +			       struct kvmppc_vcpu_e500 *vcpu_e500,
> +			       int as, int pid, int pr)
> +{
> +	struct vcpu_id_table *idt = vcpu_e500->idt;
> +
> +	idt->id[as][pid][pr].val = 0;
> +	idt->id[as][pid][pr].pentry = NULL;
> +
> +	/* Update shadow pid when mappings are changed */
> +	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
> +}
> +
> +/*
> + * Map guest (vcpu,AS,ID,PR) to physical core shadow id.
> + * This function first lookup if a valid mapping exists,
> + * if not, then creates a new one.
> + */
> +static unsigned int kvmppc_e500_get_sid(struct kvmppc_vcpu_e500 *vcpu_e500,
> +					int as, int gid, int pr,
> +					int avoid_recursion)
> +{
> +	struct vcpu_id_table *idt = vcpu_e500->idt;
> +	int sid;
> +

Please check for as,gid,pr boundaries with BUG_ON.

> +	sid = local_sid_lookup(&idt->id[as][gid][pr]);
> +
> +	while (sid<= 0) {
> +		/* No mapping yet */
> +		sid = local_sid_setup_one(&idt->id[as][gid][pr]);
> +		if (sid<= 0) {
> +			_tlbil_all();
> +			local_sid_destroy_all();
> +		}
> +
> +		/* Update shadow pid when mappings are changed */
> +		if (!avoid_recursion)
> +			kvmppc_e500_recalc_shadow_pid(vcpu_e500);
> +	}
> +
> +	return sid;
> +}
> +
> +/* Map guest pid to shadow.
> + * We use PID to keep shadow of current guest non-zero PID,

Since the MMU code expects PID1 and PID2 to be zero (which breaks nested 
KVM!), please make sure to properly abort in the SPR emulation code if 
anyone tries to write to either of them.

> + * and use PID1 to keep shadow of guest zero PID.
> + * So that guest tlbe with TID=0 can be accessed at any time */
> +void kvmppc_e500_recalc_shadow_pid(struct kvmppc_vcpu_e500 *vcpu_e500)
> +{
> +	preempt_disable();
> +	vcpu_e500->vcpu.arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500,
> +			get_cur_as(&vcpu_e500->vcpu),
> +			get_cur_pid(&vcpu_e500->vcpu),
> +			get_cur_pr(&vcpu_e500->vcpu), 1);
> +	vcpu_e500->vcpu.arch.shadow_pid1 = kvmppc_e500_get_sid(vcpu_e500,
> +			get_cur_as(&vcpu_e500->vcpu), 0,
> +			get_cur_pr(&vcpu_e500->vcpu), 1);
> +	preempt_enable();
> +}
> +
>   void kvmppc_dump_tlbs(struct kvm_vcpu *vcpu)
>   {
>   	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
> @@ -134,14 +293,19 @@ static inline void write_host_tlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
>
>   void kvmppc_map_magic(struct kvm_vcpu *vcpu)
>   {
> +	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
>   	struct tlbe magic;
>   	ulong shared_page = ((ulong)vcpu->arch.shared)&  PAGE_MASK;
> +	unsigned int stid;
>   	pfn_t pfn;
>
>   	pfn = (pfn_t)virt_to_phys((void *)shared_page)>>  PAGE_SHIFT;
>   	get_page(pfn_to_page(pfn));
>
> -	magic.mas1 = MAS1_VALID | MAS1_TS |
> +	preempt_disable();
> +	stid = kvmppc_e500_get_sid(vcpu_e500, 0, 0, 0, 0);
> +
> +	magic.mas1 = MAS1_VALID | MAS1_TS | MAS1_TID(stid) |
>   		     MAS1_TSIZE(BOOK3E_PAGESZ_4K);
>   	magic.mas2 = vcpu->arch.magic_page_ea | MAS2_M;
>   	magic.mas3 = (pfn<<  PAGE_SHIFT) |
> @@ -149,15 +313,68 @@ void kvmppc_map_magic(struct kvm_vcpu *vcpu)
>   	magic.mas7 = pfn>>  (32 - PAGE_SHIFT);
>
>   	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index));
> +	preempt_enable();
>   }
>
>   void kvmppc_e500_tlb_load(struct kvm_vcpu *vcpu, int cpu)
>   {
> +	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
> +
> +	/* Shadow PID may be expired on local core */
> +	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
>   }
>
>   void kvmppc_e500_tlb_put(struct kvm_vcpu *vcpu)
>   {
> -	_tlbil_all();
> +}
> +
> +static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
> +					 int tlbsel, int esel)
> +{
> +	struct tlbe *gtlbe =&vcpu_e500->gtlb_arch[tlbsel][esel];
> +	struct vcpu_id_table *idt = vcpu_e500->idt;
> +	unsigned int pr, tid, ts, pid;
> +	u32 val, eaddr;
> +	unsigned long flags;
> +
> +	ts = get_tlb_ts(gtlbe);
> +	tid = get_tlb_tid(gtlbe);
> +
> +	/* One guest ID may be mapped to two shadow IDs */
> +	for (pr = 0; pr<  2; pr++) {
> +		/*
> +		 * If we don't find a valid ID mapping, maybe means
> +		 * it doesn't have a valid ID mapping on local core,
> +		 * and has valid ID mapping on other cores. We should
> +		 * clear this ID mapping to disable related shadow TLBe.
> +		 *
> +		 * And we always clear corresponding ID mapping
> +		 * if guest is invalidating a TLB1 entry.

Where?

> +		 */
> +		if (tlbsel = 1 ||
> +		    (pid = local_sid_lookup(&idt->id[ts][tid][pr]))<= 0) {
> +			kvmppc_e500_id_table_reset_one(vcpu_e500, ts, tid, pr);
> +			continue;
> +		}
> +
> +		/* Guest is invalidating a TLB0 entry
> +		 * which has a valid ID mapping.
> +		 * We search host TLB0 to invalidate it's shadow TLBe */
> +		val = (pid<<  16) | MAS6_SAS;

Is there no define for the shift? If not, please create one.

> +		eaddr = get_tlb_eaddr(gtlbe);
> +
> +		local_irq_save(flags);
> +
> +		mtspr(SPRN_MAS6, val);
> +		asm volatile("tlbsx 0, %[eaddr]" : : [eaddr] "r" (eaddr));
> +		val = mfspr(SPRN_MAS1);
> +		if (val&  MAS1_VALID) {
> +			mtspr(SPRN_MAS1, val&  ~MAS1_VALID);
> +			asm volatile("tlbwe");
> +		}

Is all this faster than tlbivax? Sure, you might flush other guest's and 
PID's addresses as well, but save here. Might be worth measuring.

> +
> +		local_irq_restore(flags);
> +	}
>   }
>
>   /* Search the guest TLB for a matching entry. */
> @@ -216,12 +433,6 @@ static inline void kvmppc_e500_priv_release(struct tlbe_priv *priv)
>   	}
>   }
>
> -static void kvmppc_e500_tlb1_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
> -					int esel)
> -{
> -	mtspr(SPRN_MMUCSR0, MMUCSR0_TLB1FI);
> -}
> -
>   static inline void kvmppc_e500_deliver_tlb_miss(struct kvm_vcpu *vcpu,
>   		unsigned int eaddr, int as)
>   {
> @@ -255,10 +466,15 @@ static inline void kvmppc_e500_setup_stlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
>   					   u64 gvaddr, struct tlbe *stlbe)
>   {
>   	pfn_t pfn = priv->pfn;
> +	unsigned int stid;
> +
> +	stid = kvmppc_e500_get_sid(vcpu_e500, get_tlb_ts(gtlbe),
> +				   get_tlb_tid(gtlbe),
> +				   get_cur_pr(&vcpu_e500->vcpu), 0);
>
>   	/* Force TS=1 IPROT=0 for all guest mappings. */
>   	stlbe->mas1 = MAS1_TSIZE(tsize)
> -		| MAS1_TID(get_tlb_tid(gtlbe)) | MAS1_TS | MAS1_VALID;
> +		| MAS1_TID(stid) | MAS1_TS | MAS1_VALID;
>   	stlbe->mas2 = (gvaddr&  MAS2_EPN)
>   		| e500_shadow_mas2_attrib(gtlbe->mas2,
>   				vcpu_e500->vcpu.arch.shared->msr&  MSR_PR);
> @@ -414,14 +630,12 @@ static int kvmppc_e500_tlb1_map(struct kvmppc_vcpu_e500 *vcpu_e500,
>   	return victim;
>   }
>
> -/* Invalidate all guest kernel mappings when enter usermode,
> - * so that when they fault back in they will get the
> - * proper permission bits. */
> -void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
> +void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
>   {
> -	if (usermode) {
> -		_tlbil_all();
> -	}
> +	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
> +
> +	/* Recalc shadow pid since MSR changes */
> +	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
>   }
>
>   static inline int kvmppc_e500_gtlbe_invalidate(
> @@ -449,7 +663,8 @@ int kvmppc_e500_emul_mt_mmucsr0(struct kvmppc_vcpu_e500 *vcpu_e500, ulong value)
>   		for (esel = 0; esel<  vcpu_e500->gtlb_size[1]; esel++)
>   			kvmppc_e500_gtlbe_invalidate(vcpu_e500, 1, esel);
>
> -	_tlbil_all();
> +	/* Invalidate all vcpu id mappings */
> +	kvmppc_e500_id_table_reset_all(vcpu_e500);
>
>   	return EMULATE_DONE;
>   }
> @@ -480,7 +695,8 @@ int kvmppc_e500_emul_tlbivax(struct kvm_vcpu *vcpu, int ra, int rb)
>   			kvmppc_e500_gtlbe_invalidate(vcpu_e500, tlbsel, esel);
>   	}
>
> -	_tlbil_all();
> +	/* Invalidate all vcpu id mappings */
> +	kvmppc_e500_id_table_reset_all(vcpu_e500);
>
>   	return EMULATE_DONE;
>   }
> @@ -564,8 +780,8 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
>
>   	gtlbe =&vcpu_e500->gtlb_arch[tlbsel][esel];
>
> -	if (get_tlb_v(gtlbe)&&  tlbsel = 1)
> -		kvmppc_e500_tlb1_invalidate(vcpu_e500, esel);
> +	if (get_tlb_v(gtlbe))
> +		kvmppc_e500_stlbe_invalidate(vcpu_e500, tlbsel, esel);
>
>   	gtlbe->mas1 = vcpu_e500->mas1;
>   	gtlbe->mas2 = vcpu_e500->mas2;
> @@ -582,6 +798,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
>   		u64 eaddr;
>   		u64 raddr;
>
> +		preempt_disable();

Are those preempt disables related or just happen to land in this patch? 
I'd assume the actual tlb modification functions to care about preemption?


Alex

Re: [PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Scott Wood]

On Thu, 19 May 2011 15:18:30 +0200
Alexander Graf <agraf@suse.de> wrote:
	
> On 05/18/2011 01:42 AM, Scott Wood wrote:
> > +static inline int local_sid_setup_one(struct id *entry)
> > +{
> > +	unsigned long sid;
> > +	int ret = -1;
> > +
> > +	sid = ++(__get_cpu_var(pcpu_last_used_sid));
> > +	if (sid<  256) {
> > +		__get_cpu_var(pcpu_sids).entry[sid] = entry;
> > +		entry->val = sid;
> > +		entry->pentry =&__get_cpu_var(pcpu_sids).entry[sid];
> > +		ret = sid;
> > +	}
> > +
> > +	WARN_ON(sid>  256);
> 
> Why > 256?

If sid = 256, we've run out of sids, this function will return -1, all
the sids will be dumped, and pcpu_last_used_sid will be reset to zero.

sid > 256 indicates a race.

> Since the MMU code expects PID1 and PID2 to be zero (which breaks nested 
> KVM!), please make sure to properly abort in the SPR emulation code if 
> anyone tries to write to either of them.

Yeah, I figured someone would bring up nested KVM. :-)

It should be possible to handle guest PID1/PID2 by dynamically assigning
shadow PIDs to hardware PID registers, rather than fixed PID0/PID1 usage.

> > +static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
> > +					 int tlbsel, int esel)
> > +{
> > +	struct tlbe *gtlbe =&vcpu_e500->gtlb_arch[tlbsel][esel];
> > +	struct vcpu_id_table *idt = vcpu_e500->idt;
> > +	unsigned int pr, tid, ts, pid;
> > +	u32 val, eaddr;
> > +	unsigned long flags;
> > +
> > +	ts = get_tlb_ts(gtlbe);
> > +	tid = get_tlb_tid(gtlbe);
> > +
> > +	/* One guest ID may be mapped to two shadow IDs */
> > +	for (pr = 0; pr<  2; pr++) {
> > +		/*
> > +		 * If we don't find a valid ID mapping, maybe means
> > +		 * it doesn't have a valid ID mapping on local core,
> > +		 * and has valid ID mapping on other cores. We should
> > +		 * clear this ID mapping to disable related shadow TLBe.
> > +		 *
> > +		 * And we always clear corresponding ID mapping
> > +		 * if guest is invalidating a TLB1 entry.
> 
> Where?
> 
> > +		 */
> > +		if (tlbsel = 1 ||
> > +		    (pid = local_sid_lookup(&idt->id[ts][tid][pr]))<= 0) {
> > +			kvmppc_e500_id_table_reset_one(vcpu_e500, ts, tid, pr);
> > +			continue;
> > +		}

Here.

We only try to shoot down a single entry if it is TLB0 (and thus we know
it's backed by just one host TLB entry), and the shadow PID is active on
this host cpu (which means it is not active on others).  Otherwise we kill
the whole shadow PID.

I'll reword the comment to be clearer.

> > +		/* Guest is invalidating a TLB0 entry
> > +		 * which has a valid ID mapping.
> > +		 * We search host TLB0 to invalidate it's shadow TLBe */
> > +		val = (pid<<  16) | MAS6_SAS;
> 
> Is there no define for the shift? If not, please create one.

There is, not sure why it wasn't used.

> 
> > +		eaddr = get_tlb_eaddr(gtlbe);
> > +
> > +		local_irq_save(flags);
> > +
> > +		mtspr(SPRN_MAS6, val);
> > +		asm volatile("tlbsx 0, %[eaddr]" : : [eaddr] "r" (eaddr));
> > +		val = mfspr(SPRN_MAS1);
> > +		if (val&  MAS1_VALID) {
> > +			mtspr(SPRN_MAS1, val&  ~MAS1_VALID);
> > +			asm volatile("tlbwe");
> > +		}
> 
> Is all this faster than tlbivax? Sure, you might flush other guest's and 
> PID's addresses as well, but save here. Might be worth measuring.

tlbivax broadcasts, and requires tlbsync which broadcasts and needs a
global lock due to an erratum.

This is what Linux does for its own invalidations in the absense of tlbilx;
see __tlbil_va() in arch/powerpc/mm/tlb_nohash_low.S.

> > @@ -564,8 +780,8 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
> >
> >   	gtlbe =&vcpu_e500->gtlb_arch[tlbsel][esel];
> >
> > -	if (get_tlb_v(gtlbe)&&  tlbsel = 1)
> > -		kvmppc_e500_tlb1_invalidate(vcpu_e500, esel);
> > +	if (get_tlb_v(gtlbe))
> > +		kvmppc_e500_stlbe_invalidate(vcpu_e500, tlbsel, esel);
> >
> >   	gtlbe->mas1 = vcpu_e500->mas1;
> >   	gtlbe->mas2 = vcpu_e500->mas2;
> > @@ -582,6 +798,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
> >   		u64 eaddr;
> >   		u64 raddr;
> >
> > +		preempt_disable();
> 
> Are those preempt disables related or just happen to land in this patch? 
> I'd assume the actual tlb modification functions to care about preemption?

It's related.  We want to keep preemption off from the time we look up a
shadow ID to the time we write the TLB entry into hardware.  Atfer that, if
the shadow ID is invalidated, the TLB entry will be as well.

-Scott

[PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Scott Wood]

From: Liu Yu <yu.liu@freescale.com>

Dynamically assign host PIDs to guest PIDs, splitting each guest PID into
multiple host (shadow) PIDs based on kernel/user and MSR[IS/DS].  Use
both PID0 and PID1 so that the shadow PIDs for the right mode can be
selected, that correspond both to guest TID = zero and guest TID = guest
PID.

This allows us to significantly reduce the frequency of needing to
invalidate the entire TLB.  When the guest mode or PID changes, we just
update the host PID0/PID1.  And since the allocation of shadow PIDs is
global, multiple guests can share the TLB without conflict.

Note that KVM does not yet support the guest setting PID1 or PID2 to
a value other than zero.  This will need to be fixed for nested KVM
to work.  Until then, we enforce the requirement for guest PID1/PID2
to stay zero by failing the emulation if the guest tries to set them
to something else.

Signed-off-by: Liu Yu <yu.liu@freescale.com>
Signed-off-by: Scott Wood <scottwood@freescale.com>
---
 arch/powerpc/include/asm/kvm_e500.h |    8 +-
 arch/powerpc/include/asm/kvm_host.h |    1 +
 arch/powerpc/kernel/asm-offsets.c   |    1 +
 arch/powerpc/kvm/44x_tlb.c          |    4 +-
 arch/powerpc/kvm/booke.c            |   11 +-
 arch/powerpc/kvm/booke.h            |    1 +
 arch/powerpc/kvm/booke_interrupts.S |   11 ++
 arch/powerpc/kvm/e500_emulate.c     |    4 +
 arch/powerpc/kvm/e500_tlb.c         |  312 ++++++++++++++++++++++++++++++++---
 arch/powerpc/kvm/e500_tlb.h         |   13 ++-
 10 files changed, 334 insertions(+), 32 deletions(-)

diff --git a/arch/powerpc/include/asm/kvm_e500.h b/arch/powerpc/include/asm/kvm_e500.h
index 4a6d77a..adbfca9 100644
--- a/arch/powerpc/include/asm/kvm_e500.h
+++ b/arch/powerpc/include/asm/kvm_e500.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 Freescale Semiconductor, Inc. All rights reserved.
+ * Copyright (C) 2008-2011 Freescale Semiconductor, Inc. All rights reserved.
  *
  * Author: Yu Liu, <yu.liu@freescale.com>
  *
@@ -37,6 +37,8 @@ struct tlbe_priv {
 	unsigned int flags; /* E500_TLB_* */
 };
 
+struct vcpu_id_table;
+
 struct kvmppc_vcpu_e500 {
 	/* Unmodified copy of the guest's TLB. */
 	struct tlbe *gtlb_arch[E500_TLB_NUM];
@@ -59,6 +61,10 @@ struct kvmppc_vcpu_e500 {
 	u32 mas5;
 	u32 mas6;
 	u32 mas7;
+
+	/* vcpu id table */
+	struct vcpu_id_table *idt;
+
 	u32 l1csr0;
 	u32 l1csr1;
 	u32 hid0;
diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
index c4ce105..6e05b2d 100644
--- a/arch/powerpc/include/asm/kvm_host.h
+++ b/arch/powerpc/include/asm/kvm_host.h
@@ -255,6 +255,7 @@ struct kvm_vcpu_arch {
 	u32 pvr;
 
 	u32 shadow_pid;
+	u32 shadow_pid1;
 	u32 pid;
 	u32 swap_pid;
 
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index ecd2b3a..faf8461 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -402,6 +402,7 @@ int main(void)
 	DEFINE(VCPU_SPRG6, offsetof(struct kvm_vcpu, arch.sprg6));
 	DEFINE(VCPU_SPRG7, offsetof(struct kvm_vcpu, arch.sprg7));
 	DEFINE(VCPU_SHADOW_PID, offsetof(struct kvm_vcpu, arch.shadow_pid));
+	DEFINE(VCPU_SHADOW_PID1, offsetof(struct kvm_vcpu, arch.shadow_pid1));
 	DEFINE(VCPU_SHARED, offsetof(struct kvm_vcpu, arch.shared));
 	DEFINE(VCPU_SHARED_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
 	DEFINE(VCPU_SHADOW_MSR, offsetof(struct kvm_vcpu, arch.shadow_msr));
diff --git a/arch/powerpc/kvm/44x_tlb.c b/arch/powerpc/kvm/44x_tlb.c
index 5f3cff8..33aa715 100644
--- a/arch/powerpc/kvm/44x_tlb.c
+++ b/arch/powerpc/kvm/44x_tlb.c
@@ -387,8 +387,10 @@ static void kvmppc_44x_invalidate(struct kvm_vcpu *vcpu,
 	}
 }
 
-void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
 {
+	int usermode = vcpu->arch.shared->msr & MSR_PR;
+
 	vcpu->arch.shadow_pid = !usermode;
 }
 
diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
index 45ad454..d2288af 100644
--- a/arch/powerpc/kvm/booke.c
+++ b/arch/powerpc/kvm/booke.c
@@ -113,15 +113,18 @@ static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
 }
 #endif
 
-/* Helper function for "full" MSR writes. No need to call this if only EE is
- * changing. */
+/*
+ * Helper function for "full" MSR writes.  No need to call this if only
+ * EE/CE/ME/DE/RI are changing.
+ */
 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 {
-	if ((new_msr & MSR_PR) != (vcpu->arch.shared->msr & MSR_PR))
-		kvmppc_mmu_priv_switch(vcpu, new_msr & MSR_PR);
+	u32 old_msr = vcpu->arch.shared->msr;
 
 	vcpu->arch.shared->msr = new_msr;
 
+	kvmppc_mmu_msr_notify(vcpu, old_msr);
+
 	if (vcpu->arch.shared->msr & MSR_WE) {
 		kvm_vcpu_block(vcpu);
 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
diff --git a/arch/powerpc/kvm/booke.h b/arch/powerpc/kvm/booke.h
index 0fa1732..8e1fe33 100644
--- a/arch/powerpc/kvm/booke.h
+++ b/arch/powerpc/kvm/booke.h
@@ -53,6 +53,7 @@
 extern unsigned long kvmppc_booke_handlers;
 
 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr);
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr);
 
 int kvmppc_booke_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
                             unsigned int inst, int *advance);
diff --git a/arch/powerpc/kvm/booke_interrupts.S b/arch/powerpc/kvm/booke_interrupts.S
index 8cb3dfe..42f2fb1 100644
--- a/arch/powerpc/kvm/booke_interrupts.S
+++ b/arch/powerpc/kvm/booke_interrupts.S
@@ -191,6 +191,12 @@ _GLOBAL(kvmppc_resume_host)
 	lwz	r3, VCPU_HOST_PID(r4)
 	mtspr	SPRN_PID, r3
 
+#ifdef CONFIG_FSL_BOOKE
+	/* we cheat and know that Linux doesn't use PID1 which is always 0 */
+	lis	r3, 0
+	mtspr	SPRN_PID1, r3
+#endif
+
 	/* Restore host IVPR before re-enabling interrupts. We cheat and know
 	 * that Linux IVPR is always 0xc0000000. */
 	lis	r3, 0xc000
@@ -365,6 +371,11 @@ lightweight_exit:
 	lwz	r3, VCPU_SHADOW_PID(r4)
 	mtspr	SPRN_PID, r3
 
+#ifdef CONFIG_FSL_BOOKE
+	lwz	r3, VCPU_SHADOW_PID1(r4)
+	mtspr	SPRN_PID1, r3
+#endif
+
 #ifdef CONFIG_44x
 	iccci	0, 0 /* XXX hack */
 #endif
diff --git a/arch/powerpc/kvm/e500_emulate.c b/arch/powerpc/kvm/e500_emulate.c
index 69cd665..d48ae39 100644
--- a/arch/powerpc/kvm/e500_emulate.c
+++ b/arch/powerpc/kvm/e500_emulate.c
@@ -81,8 +81,12 @@ int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
 		kvmppc_set_pid(vcpu, spr_val);
 		break;
 	case SPRN_PID1:
+		if (spr_val != 0)
+			return EMULATE_FAIL;
 		vcpu_e500->pid[1] = spr_val; break;
 	case SPRN_PID2:
+		if (spr_val != 0)
+			return EMULATE_FAIL;
 		vcpu_e500->pid[2] = spr_val; break;
 	case SPRN_MAS0:
 		vcpu_e500->mas0 = spr_val; break;
diff --git a/arch/powerpc/kvm/e500_tlb.c b/arch/powerpc/kvm/e500_tlb.c
index 9d1e28d..1474f0d 100644
--- a/arch/powerpc/kvm/e500_tlb.c
+++ b/arch/powerpc/kvm/e500_tlb.c
@@ -28,8 +28,196 @@
 
 #define to_htlb1_esel(esel) (tlb1_entry_num - (esel) - 1)
 
+struct id {
+	unsigned long val;
+	struct id **pentry;
+};
+
+#define NUM_TIDS 256
+
+/*
+ * This table provide mappings from:
+ * (guestAS,guestTID,guestPR) --> ID of physical cpu
+ * guestAS	[0..1]
+ * guestTID	[0..255]
+ * guestPR	[0..1]
+ * ID		[1..255]
+ * Each vcpu keeps one vcpu_id_table.
+ */
+struct vcpu_id_table {
+	struct id id[2][NUM_TIDS][2];
+};
+
+/*
+ * This table provide reversed mappings of vcpu_id_table:
+ * ID --> address of vcpu_id_table item.
+ * Each physical core has one pcpu_id_table.
+ */
+struct pcpu_id_table {
+	struct id *entry[NUM_TIDS];
+};
+
+static DEFINE_PER_CPU(struct pcpu_id_table, pcpu_sids);
+
+/* This variable keeps last used shadow ID on local core.
+ * The valid range of shadow ID is [1..255] */
+static DEFINE_PER_CPU(unsigned long, pcpu_last_used_sid);
+
 static unsigned int tlb1_entry_num;
 
+/*
+ * Allocate a free shadow id and setup a valid sid mapping in given entry.
+ * A mapping is only valid when vcpu_id_table and pcpu_id_table are match.
+ *
+ * The caller must have preemption disabled, and keep it that way until
+ * it has finished with the returned shadow id (either written into the
+ * TLB or arch.shadow_pid, or discarded).
+ */
+static inline int local_sid_setup_one(struct id *entry)
+{
+	unsigned long sid;
+	int ret = -1;
+
+	sid = ++(__get_cpu_var(pcpu_last_used_sid));
+	if (sid < NUM_TIDS) {
+		__get_cpu_var(pcpu_sids).entry[sid] = entry;
+		entry->val = sid;
+		entry->pentry = &__get_cpu_var(pcpu_sids).entry[sid];
+		ret = sid;
+	}
+
+	/*
+	 * If sid = NUM_TIDS, we've run out of sids.  We return -1, and
+	 * the caller will invalidate everything and start over.
+	 *
+	 * sid > NUM_TIDS indicates a race, which we disable preemption to
+	 * avoid.
+	 */
+	WARN_ON(sid > NUM_TIDS);
+
+	return ret;
+}
+
+/*
+ * Check if given entry contain a valid shadow id mapping.
+ * An ID mapping is considered valid only if
+ * both vcpu and pcpu know this mapping.
+ *
+ * The caller must have preemption disabled, and keep it that way until
+ * it has finished with the returned shadow id (either written into the
+ * TLB or arch.shadow_pid, or discarded).
+ */
+static inline int local_sid_lookup(struct id *entry)
+{
+	if (entry && entry->val != 0 &&
+	    __get_cpu_var(pcpu_sids).entry[entry->val] = entry &&
+	    entry->pentry = &__get_cpu_var(pcpu_sids).entry[entry->val])
+		return entry->val;
+	return -1;
+}
+
+/* Invalidate all id mappings on local core */
+static inline void local_sid_destroy_all(void)
+{
+	preempt_disable();
+	__get_cpu_var(pcpu_last_used_sid) = 0;
+	memset(&__get_cpu_var(pcpu_sids), 0, sizeof(__get_cpu_var(pcpu_sids)));
+	preempt_enable();
+}
+
+static void *kvmppc_e500_id_table_alloc(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	vcpu_e500->idt = kzalloc(sizeof(struct vcpu_id_table), GFP_KERNEL);
+	return vcpu_e500->idt;
+}
+
+static void kvmppc_e500_id_table_free(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	kfree(vcpu_e500->idt);
+}
+
+/* Invalidate all mappings on vcpu */
+static void kvmppc_e500_id_table_reset_all(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	memset(vcpu_e500->idt, 0, sizeof(struct vcpu_id_table));
+
+	/* Update shadow pid when mappings are changed */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+}
+
+/* Invalidate one ID mapping on vcpu */
+static inline void kvmppc_e500_id_table_reset_one(
+			       struct kvmppc_vcpu_e500 *vcpu_e500,
+			       int as, int pid, int pr)
+{
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+
+	BUG_ON(as >= 2);
+	BUG_ON(pid >= NUM_TIDS);
+	BUG_ON(pr >= 2);
+
+	idt->id[as][pid][pr].val = 0;
+	idt->id[as][pid][pr].pentry = NULL;
+
+	/* Update shadow pid when mappings are changed */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+}
+
+/*
+ * Map guest (vcpu,AS,ID,PR) to physical core shadow id.
+ * This function first lookup if a valid mapping exists,
+ * if not, then creates a new one.
+ *
+ * The caller must have preemption disabled, and keep it that way until
+ * it has finished with the returned shadow id (either written into the
+ * TLB or arch.shadow_pid, or discarded).
+ */
+static unsigned int kvmppc_e500_get_sid(struct kvmppc_vcpu_e500 *vcpu_e500,
+					unsigned int as, unsigned int gid,
+					unsigned int pr, int avoid_recursion)
+{
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+	int sid;
+
+	BUG_ON(as >= 2);
+	BUG_ON(gid >= NUM_TIDS);
+	BUG_ON(pr >= 2);
+
+	sid = local_sid_lookup(&idt->id[as][gid][pr]);
+
+	while (sid <= 0) {
+		/* No mapping yet */
+		sid = local_sid_setup_one(&idt->id[as][gid][pr]);
+		if (sid <= 0) {
+			_tlbil_all();
+			local_sid_destroy_all();
+		}
+
+		/* Update shadow pid when mappings are changed */
+		if (!avoid_recursion)
+			kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+	}
+
+	return sid;
+}
+
+/* Map guest pid to shadow.
+ * We use PID to keep shadow of current guest non-zero PID,
+ * and use PID1 to keep shadow of guest zero PID.
+ * So that guest tlbe with TID=0 can be accessed at any time */
+void kvmppc_e500_recalc_shadow_pid(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+	preempt_disable();
+	vcpu_e500->vcpu.arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500,
+			get_cur_as(&vcpu_e500->vcpu),
+			get_cur_pid(&vcpu_e500->vcpu),
+			get_cur_pr(&vcpu_e500->vcpu), 1);
+	vcpu_e500->vcpu.arch.shadow_pid1 = kvmppc_e500_get_sid(vcpu_e500,
+			get_cur_as(&vcpu_e500->vcpu), 0,
+			get_cur_pr(&vcpu_e500->vcpu), 1);
+	preempt_enable();
+}
+
 void kvmppc_dump_tlbs(struct kvm_vcpu *vcpu)
 {
 	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
@@ -134,14 +322,19 @@ static inline void write_host_tlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
 
 void kvmppc_map_magic(struct kvm_vcpu *vcpu)
 {
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
 	struct tlbe magic;
 	ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
+	unsigned int stid;
 	pfn_t pfn;
 
 	pfn = (pfn_t)virt_to_phys((void *)shared_page) >> PAGE_SHIFT;
 	get_page(pfn_to_page(pfn));
 
-	magic.mas1 = MAS1_VALID | MAS1_TS |
+	preempt_disable();
+	stid = kvmppc_e500_get_sid(vcpu_e500, 0, 0, 0, 0);
+
+	magic.mas1 = MAS1_VALID | MAS1_TS | MAS1_TID(stid) |
 		     MAS1_TSIZE(BOOK3E_PAGESZ_4K);
 	magic.mas2 = vcpu->arch.magic_page_ea | MAS2_M;
 	magic.mas3 = (pfn << PAGE_SHIFT) |
@@ -149,15 +342,76 @@ void kvmppc_map_magic(struct kvm_vcpu *vcpu)
 	magic.mas7 = pfn >> (32 - PAGE_SHIFT);
 
 	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index));
+	preempt_enable();
 }
 
 void kvmppc_e500_tlb_load(struct kvm_vcpu *vcpu, int cpu)
 {
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+
+	/* Shadow PID may be expired on local core */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
 }
 
 void kvmppc_e500_tlb_put(struct kvm_vcpu *vcpu)
 {
-	_tlbil_all();
+}
+
+static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
+					 int tlbsel, int esel)
+{
+	struct tlbe *gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
+	struct vcpu_id_table *idt = vcpu_e500->idt;
+	unsigned int pr, tid, ts, pid;
+	u32 val, eaddr;
+	unsigned long flags;
+
+	ts = get_tlb_ts(gtlbe);
+	tid = get_tlb_tid(gtlbe);
+
+	preempt_disable();
+
+	/* One guest ID may be mapped to two shadow IDs */
+	for (pr = 0; pr < 2; pr++) {
+		/*
+		 * The shadow PID can have a valid mapping on at most one
+		 * host CPU.  In the common case, it will be valid on this
+		 * CPU, in which case (for TLB0) we do a local invalidation
+		 * of the specific address.
+		 *
+		 * If the shadow PID is not valid on the current host CPU, or
+		 * if we're invalidating a TLB1 entry, we invalidate the
+		 * entire shadow PID.
+		 */
+		if (tlbsel = 1 ||
+		    (pid = local_sid_lookup(&idt->id[ts][tid][pr])) <= 0) {
+			kvmppc_e500_id_table_reset_one(vcpu_e500, ts, tid, pr);
+			continue;
+		}
+
+		/*
+		 * The guest is invalidating a TLB0 entry which in in a PID
+		 * that has a valid shadow mapping on this host CPU.  We
+		 * search host TLB0 to invalidate it's shadow TLB entry,
+		 * similar to __tlbil_va except that we need to look in AS1.
+		 */
+		val = (pid << MAS6_SPID_SHIFT) | MAS6_SAS;
+		eaddr = get_tlb_eaddr(gtlbe);
+
+		local_irq_save(flags);
+
+		mtspr(SPRN_MAS6, val);
+		asm volatile("tlbsx 0, %[eaddr]" : : [eaddr] "r" (eaddr));
+		val = mfspr(SPRN_MAS1);
+		if (val & MAS1_VALID) {
+			mtspr(SPRN_MAS1, val & ~MAS1_VALID);
+			asm volatile("tlbwe");
+		}
+
+		local_irq_restore(flags);
+	}
+
+	preempt_enable();
 }
 
 /* Search the guest TLB for a matching entry. */
@@ -216,12 +470,6 @@ static inline void kvmppc_e500_priv_release(struct tlbe_priv *priv)
 	}
 }
 
-static void kvmppc_e500_tlb1_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
-					int esel)
-{
-	mtspr(SPRN_MMUCSR0, MMUCSR0_TLB1FI);
-}
-
 static inline void kvmppc_e500_deliver_tlb_miss(struct kvm_vcpu *vcpu,
 		unsigned int eaddr, int as)
 {
@@ -255,10 +503,15 @@ static inline void kvmppc_e500_setup_stlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
 					   u64 gvaddr, struct tlbe *stlbe)
 {
 	pfn_t pfn = priv->pfn;
+	unsigned int stid;
+
+	stid = kvmppc_e500_get_sid(vcpu_e500, get_tlb_ts(gtlbe),
+				   get_tlb_tid(gtlbe),
+				   get_cur_pr(&vcpu_e500->vcpu), 0);
 
 	/* Force TS=1 IPROT=0 for all guest mappings. */
 	stlbe->mas1 = MAS1_TSIZE(tsize)
-		| MAS1_TID(get_tlb_tid(gtlbe)) | MAS1_TS | MAS1_VALID;
+		| MAS1_TID(stid) | MAS1_TS | MAS1_VALID;
 	stlbe->mas2 = (gvaddr & MAS2_EPN)
 		| e500_shadow_mas2_attrib(gtlbe->mas2,
 				vcpu_e500->vcpu.arch.shared->msr & MSR_PR);
@@ -414,14 +667,12 @@ static int kvmppc_e500_tlb1_map(struct kvmppc_vcpu_e500 *vcpu_e500,
 	return victim;
 }
 
-/* Invalidate all guest kernel mappings when enter usermode,
- * so that when they fault back in they will get the
- * proper permission bits. */
-void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
+void kvmppc_mmu_msr_notify(struct kvm_vcpu *vcpu, u32 old_msr)
 {
-	if (usermode) {
-		_tlbil_all();
-	}
+	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+
+	/* Recalc shadow pid since MSR changes */
+	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
 }
 
 static inline int kvmppc_e500_gtlbe_invalidate(
@@ -449,7 +700,8 @@ int kvmppc_e500_emul_mt_mmucsr0(struct kvmppc_vcpu_e500 *vcpu_e500, ulong value)
 		for (esel = 0; esel < vcpu_e500->gtlb_size[1]; esel++)
 			kvmppc_e500_gtlbe_invalidate(vcpu_e500, 1, esel);
 
-	_tlbil_all();
+	/* Invalidate all vcpu id mappings */
+	kvmppc_e500_id_table_reset_all(vcpu_e500);
 
 	return EMULATE_DONE;
 }
@@ -480,7 +732,8 @@ int kvmppc_e500_emul_tlbivax(struct kvm_vcpu *vcpu, int ra, int rb)
 			kvmppc_e500_gtlbe_invalidate(vcpu_e500, tlbsel, esel);
 	}
 
-	_tlbil_all();
+	/* Invalidate all vcpu id mappings */
+	kvmppc_e500_id_table_reset_all(vcpu_e500);
 
 	return EMULATE_DONE;
 }
@@ -564,8 +817,8 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 
 	gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
 
-	if (get_tlb_v(gtlbe) && tlbsel = 1)
-		kvmppc_e500_tlb1_invalidate(vcpu_e500, esel);
+	if (get_tlb_v(gtlbe))
+		kvmppc_e500_stlbe_invalidate(vcpu_e500, tlbsel, esel);
 
 	gtlbe->mas1 = vcpu_e500->mas1;
 	gtlbe->mas2 = vcpu_e500->mas2;
@@ -582,6 +835,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 		u64 eaddr;
 		u64 raddr;
 
+		preempt_disable();
 		switch (tlbsel) {
 		case 0:
 			/* TLB0 */
@@ -611,6 +865,7 @@ int kvmppc_e500_emul_tlbwe(struct kvm_vcpu *vcpu)
 			BUG();
 		}
 		write_host_tlbe(vcpu_e500, stlbsel, sesel, &stlbe);
+		preempt_enable();
 	}
 
 	kvmppc_set_exit_type(vcpu, EMULATED_TLBWE_EXITS);
@@ -672,6 +927,7 @@ void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 eaddr, gpa_t gpaddr,
 
 	gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
 
+	preempt_disable();
 	switch (tlbsel) {
 	case 0:
 		stlbsel = 0;
@@ -697,6 +953,7 @@ void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 eaddr, gpa_t gpaddr,
 	}
 
 	write_host_tlbe(vcpu_e500, stlbsel, sesel, &stlbe);
+	preempt_enable();
 }
 
 int kvmppc_e500_tlb_search(struct kvm_vcpu *vcpu,
@@ -718,8 +975,10 @@ void kvmppc_set_pid(struct kvm_vcpu *vcpu, u32 pid)
 {
 	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
 
-	vcpu_e500->pid[0] = vcpu->arch.shadow_pid -		vcpu->arch.pid = pid;
+	if (vcpu->arch.pid != pid) {
+		vcpu_e500->pid[0] = vcpu->arch.pid = pid;
+		kvmppc_e500_recalc_shadow_pid(vcpu_e500);
+	}
 }
 
 void kvmppc_e500_tlb_setup(struct kvmppc_vcpu_e500 *vcpu_e500)
@@ -767,6 +1026,9 @@ int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
 	if (vcpu_e500->gtlb_priv[1] = NULL)
 		goto err_out_priv0;
 
+	if (kvmppc_e500_id_table_alloc(vcpu_e500) = NULL)
+		goto err_out_priv1;
+
 	/* Init TLB configuration register */
 	vcpu_e500->tlb0cfg = mfspr(SPRN_TLB0CFG) & ~0xfffUL;
 	vcpu_e500->tlb0cfg |= vcpu_e500->gtlb_size[0];
@@ -775,6 +1037,8 @@ int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
 
 	return 0;
 
+err_out_priv1:
+	kfree(vcpu_e500->gtlb_priv[1]);
 err_out_priv0:
 	kfree(vcpu_e500->gtlb_priv[0]);
 err_out_guest1:
@@ -797,9 +1061,7 @@ void kvmppc_e500_tlb_uninit(struct kvmppc_vcpu_e500 *vcpu_e500)
 			kvmppc_e500_priv_release(priv);
 		}
 
-	/* discard all guest mapping */
-	_tlbil_all();
-
+	kvmppc_e500_id_table_free(vcpu_e500);
 	kfree(vcpu_e500->gtlb_arch[1]);
 	kfree(vcpu_e500->gtlb_arch[0]);
 }
diff --git a/arch/powerpc/kvm/e500_tlb.h b/arch/powerpc/kvm/e500_tlb.h
index 458946b..59b88e9 100644
--- a/arch/powerpc/kvm/e500_tlb.h
+++ b/arch/powerpc/kvm/e500_tlb.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 Freescale Semiconductor, Inc. All rights reserved.
+ * Copyright (C) 2008-2011 Freescale Semiconductor, Inc. All rights reserved.
  *
  * Author: Yu Liu, yu.liu@freescale.com
  *
@@ -55,6 +55,7 @@ extern void kvmppc_e500_tlb_load(struct kvm_vcpu *, int);
 extern int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *);
 extern void kvmppc_e500_tlb_uninit(struct kvmppc_vcpu_e500 *);
 extern void kvmppc_e500_tlb_setup(struct kvmppc_vcpu_e500 *);
+extern void kvmppc_e500_recalc_shadow_pid(struct kvmppc_vcpu_e500 *);
 
 /* TLB helper functions */
 static inline unsigned int get_tlb_size(const struct tlbe *tlbe)
@@ -110,6 +111,16 @@ static inline unsigned int get_cur_pid(struct kvm_vcpu *vcpu)
 	return vcpu->arch.pid & 0xff;
 }
 
+static inline unsigned int get_cur_as(struct kvm_vcpu *vcpu)
+{
+	return !!(vcpu->arch.shared->msr & (MSR_IS | MSR_DS));
+}
+
+static inline unsigned int get_cur_pr(struct kvm_vcpu *vcpu)
+{
+	return !!(vcpu->arch.shared->msr & MSR_PR);
+}
+
 static inline unsigned int get_cur_spid(
 		const struct kvmppc_vcpu_e500 *vcpu_e500)
 {
-- 
1.7.4.1

Re: [PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Alexander Graf]

On 03.06.2011, at 01:17, Scott Wood wrote:

> From: Liu Yu <yu.liu@freescale.com>
> 
> Dynamically assign host PIDs to guest PIDs, splitting each guest PID into
> multiple host (shadow) PIDs based on kernel/user and MSR[IS/DS].  Use
> both PID0 and PID1 so that the shadow PIDs for the right mode can be
> selected, that correspond both to guest TID = zero and guest TID = guest
> PID.
> 
> This allows us to significantly reduce the frequency of needing to
> invalidate the entire TLB.  When the guest mode or PID changes, we just
> update the host PID0/PID1.  And since the allocation of shadow PIDs is
> global, multiple guests can share the TLB without conflict.
> 
> Note that KVM does not yet support the guest setting PID1 or PID2 to
> a value other than zero.  This will need to be fixed for nested KVM
> to work.  Until then, we enforce the requirement for guest PID1/PID2
> to stay zero by failing the emulation if the guest tries to set them
> to something else.
> 
> Signed-off-by: Liu Yu <yu.liu@freescale.com>
> Signed-off-by: Scott Wood <scottwood@freescale.com>
> ---
> arch/powerpc/include/asm/kvm_e500.h |    8 +-
> arch/powerpc/include/asm/kvm_host.h |    1 +
> arch/powerpc/kernel/asm-offsets.c   |    1 +
> arch/powerpc/kvm/44x_tlb.c          |    4 +-
> arch/powerpc/kvm/booke.c            |   11 +-
> arch/powerpc/kvm/booke.h            |    1 +
> arch/powerpc/kvm/booke_interrupts.S |   11 ++
> arch/powerpc/kvm/e500_emulate.c     |    4 +
> arch/powerpc/kvm/e500_tlb.c         |  312 ++++++++++++++++++++++++++++++++---
> arch/powerpc/kvm/e500_tlb.h         |   13 ++-
> 10 files changed, 334 insertions(+), 32 deletions(-)
> 
> 

[...]

> @@ -149,15 +342,76 @@ void kvmppc_map_magic(struct kvm_vcpu *vcpu)
> 	magic.mas7 = pfn >> (32 - PAGE_SHIFT);
> 
> 	__write_host_tlbe(&magic, MAS0_TLBSEL(1) | MAS0_ESEL(tlbcam_index));
> +	preempt_enable();
> }
> 
> void kvmppc_e500_tlb_load(struct kvm_vcpu *vcpu, int cpu)
> {
> +	struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
> +
> +	/* Shadow PID may be expired on local core */
> +	kvmppc_e500_recalc_shadow_pid(vcpu_e500);
> }
> 
> void kvmppc_e500_tlb_put(struct kvm_vcpu *vcpu)
> {
> -	_tlbil_all();
> +}
> +
> +static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
> +					 int tlbsel, int esel)
> +{
> +	struct tlbe *gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
> +	struct vcpu_id_table *idt = vcpu_e500->idt;
> +	unsigned int pr, tid, ts, pid;
> +	u32 val, eaddr;
> +	unsigned long flags;
> +
> +	ts = get_tlb_ts(gtlbe);
> +	tid = get_tlb_tid(gtlbe);
> +
> +	preempt_disable();
> +
> +	/* One guest ID may be mapped to two shadow IDs */
> +	for (pr = 0; pr < 2; pr++) {
> +		/*
> +		 * The shadow PID can have a valid mapping on at most one
> +		 * host CPU.  In the common case, it will be valid on this

Not sure I understand this part. Who ensures that a shadow pid is only valid on a single CPU? From what I see, the only guarantee is that there's at most one shadow pid for each AS-PID-PR combination per CPU. But that means the invalidate could very well reach out to a different CPU as well, no?

Also, the shadow PID could be valid on multiple cores even. So if the guest is on CPU A, maps a shadow pid, then migrates to CPU B, uses another shadow PID, we have 2 mappings active, right?

> +		 * CPU, in which case (for TLB0) we do a local invalidation
> +		 * of the specific address.
> +		 *
> +		 * If the shadow PID is not valid on the current host CPU, or
> +		 * if we're invalidating a TLB1 entry, we invalidate the
> +		 * entire shadow PID.
> +		 */
> +		if (tlbsel = 1 ||
> +		    (pid = local_sid_lookup(&idt->id[ts][tid][pr])) <= 0) {
> +			kvmppc_e500_id_table_reset_one(vcpu_e500, ts, tid, pr);
> +			continue;
> +		}
> +
> +		/*
> +		 * The guest is invalidating a TLB0 entry which in in a PID

in in?



Alex

Re: [PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Scott Wood]

On Tue, 14 Jun 2011 12:41:03 +0200
Alexander Graf <agraf@suse.de> wrote:

> 
> On 03.06.2011, at 01:17, Scott Wood wrote:
> 
> > +static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
> > +					 int tlbsel, int esel)
> > +{
> > +	struct tlbe *gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
> > +	struct vcpu_id_table *idt = vcpu_e500->idt;
> > +	unsigned int pr, tid, ts, pid;
> > +	u32 val, eaddr;
> > +	unsigned long flags;
> > +
> > +	ts = get_tlb_ts(gtlbe);
> > +	tid = get_tlb_tid(gtlbe);
> > +
> > +	preempt_disable();
> > +
> > +	/* One guest ID may be mapped to two shadow IDs */
> > +	for (pr = 0; pr < 2; pr++) {
> > +		/*
> > +		 * The shadow PID can have a valid mapping on at most one
> > +		 * host CPU.  In the common case, it will be valid on this
> 
> Not sure I understand this part. Who ensures that a shadow pid is only valid on a single CPU? 

vcpu_e500->idt->id[...]->pentry can only point to one place at a time.  Any
other shadow PIDs (e.g. on other host CPUs) that it used to point to are
now invalid, and will not be re-used if the vcpu returns to that host CPU
(other than by having that host CPU reset its shadow PIDs once
they're exhausted, invalidating everything).

The linkage must match in both directions in local_sid_lookup(), or a new
shadow ID will be allocated (and the old one put out of use when pentry is
updated).

> > +		/*
> > +		 * The guest is invalidating a TLB0 entry which in in a PID
> 
> in in?
> 

"is in", sorry.

-Scott

Re: [PATCH 11/13] KVM: PPC: e500: Add shadow PID support

[Alexander Graf]

On 14.06.2011, at 21:37, Scott Wood wrote:

> On Tue, 14 Jun 2011 12:41:03 +0200
> Alexander Graf <agraf@suse.de> wrote:
> 
>> 
>> On 03.06.2011, at 01:17, Scott Wood wrote:
>> 
>>> +static void kvmppc_e500_stlbe_invalidate(struct kvmppc_vcpu_e500 *vcpu_e500,
>>> +					 int tlbsel, int esel)
>>> +{
>>> +	struct tlbe *gtlbe = &vcpu_e500->gtlb_arch[tlbsel][esel];
>>> +	struct vcpu_id_table *idt = vcpu_e500->idt;
>>> +	unsigned int pr, tid, ts, pid;
>>> +	u32 val, eaddr;
>>> +	unsigned long flags;
>>> +
>>> +	ts = get_tlb_ts(gtlbe);
>>> +	tid = get_tlb_tid(gtlbe);
>>> +
>>> +	preempt_disable();
>>> +
>>> +	/* One guest ID may be mapped to two shadow IDs */
>>> +	for (pr = 0; pr < 2; pr++) {
>>> +		/*
>>> +		 * The shadow PID can have a valid mapping on at most one
>>> +		 * host CPU.  In the common case, it will be valid on this
>> 
>> Not sure I understand this part. Who ensures that a shadow pid is only valid on a single CPU? 
> 
> vcpu_e500->idt->id[...]->pentry can only point to one place at a time.  Any
> other shadow PIDs (e.g. on other host CPUs) that it used to point to are
> now invalid, and will not be re-used if the vcpu returns to that host CPU
> (other than by having that host CPU reset its shadow PIDs once
> they're exhausted, invalidating everything).
> 
> The linkage must match in both directions in local_sid_lookup(), or a new
> shadow ID will be allocated (and the old one put out of use when pentry is
> updated).

Ah, and since the other direction is vcpu local, there's no need to do anything atomically. Very nice indeed.


Alex