The shadow id mappings is based on Hollis's idea.
We can benifit a lot from this trick:
1. Support AS=1 in guest.
So that OSes other than Linux can be expected to run in the guest.
2. Minimize the frequency of TLB flushes.
Signed-off-by: Liu Yu <yu.liu@xxxxxxxxxxxxx>
---
arch/powerpc/include/asm/kvm_e500.h | 3 +
arch/powerpc/include/asm/kvm_ppc.h | 1 +
arch/powerpc/kernel/asm-offsets.c | 1 +
arch/powerpc/kvm/booke.h | 4 +
arch/powerpc/kvm/booke_interrupts.S | 11 +++
arch/powerpc/kvm/e500_emulate.c | 10 ++-
arch/powerpc/kvm/e500_tlb.c | 146 +++++++++++++++++++++++++++++++++--
arch/powerpc/kvm/e500_tlb.h | 2 +
8 files changed, 167 insertions(+), 11 deletions(-)
diff --git a/arch/powerpc/include/asm/kvm_e500.h b/arch/powerpc/include/asm/kvm_e500.h
index ffa111b..e5ca811 100644
--- a/arch/powerpc/include/asm/kvm_e500.h
+++ b/arch/powerpc/include/asm/kvm_e500.h
@@ -43,6 +43,9 @@ struct kvmppc_vcpu_e500 {
/* Pages which are referenced in the shadow TLB. */
struct kvmppc_e500_shadow_ref *shadow_refs[E500_TLB_NUM];
+ /* MMU id mapping */
+ void *id_mapping;
+
unsigned int guest_tlb_size[E500_TLB_NUM];
unsigned int shadow_tlb_size[E500_TLB_NUM];
unsigned int guest_tlb_nv[E500_TLB_NUM];
diff --git a/arch/powerpc/include/asm/kvm_ppc.h b/arch/powerpc/include/asm/kvm_ppc.h
index 2c6ee34..40823c4 100644
--- a/arch/powerpc/include/asm/kvm_ppc.h
+++ b/arch/powerpc/include/asm/kvm_ppc.h
@@ -57,6 +57,7 @@ extern void kvmppc_emulate_dec(struct kvm_vcpu *vcpu);
extern void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 gvaddr, gpa_t gpaddr,
unsigned int gtlb_idx);
extern void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode);
+extern void kvmppc_mmu_as_switch(struct kvm_vcpu *vcpu, unsigned int as);
extern void kvmppc_mmu_switch_pid(struct kvm_vcpu *vcpu, u32 pid);
extern void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu);
extern int kvmppc_mmu_dtlb_index(struct kvm_vcpu *vcpu, gva_t eaddr);
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 42fe4da..89d28df 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -379,6 +379,7 @@ int main(void)
DEFINE(VCPU_LAST_INST, offsetof(struct kvm_vcpu, arch.last_inst));
DEFINE(VCPU_FAULT_DEAR, offsetof(struct kvm_vcpu, arch.fault_dear));
DEFINE(VCPU_FAULT_ESR, offsetof(struct kvm_vcpu, arch.fault_esr));
+ DEFINE(VCPU_SWAP_PID, offsetof(struct kvm_vcpu, arch.swap_pid));
#endif
#ifdef CONFIG_44x
DEFINE(PGD_T_LOG2, PGD_T_LOG2);
diff --git a/arch/powerpc/kvm/booke.h b/arch/powerpc/kvm/booke.h
index d59bcca..96e6cc0 100644
--- a/arch/powerpc/kvm/booke.h
+++ b/arch/powerpc/kvm/booke.h
@@ -57,6 +57,10 @@ static inline void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
if ((new_msr & MSR_PR) != (vcpu->arch.msr & MSR_PR))
kvmppc_mmu_priv_switch(vcpu, new_msr & MSR_PR);
+ if ((new_msr & (MSR_IS | MSR_DS)) !=
+ (vcpu->arch.msr & (MSR_IS | MSR_DS)))
+ kvmppc_mmu_as_switch(vcpu, new_msr & (MSR_IS | MSR_DS));
+
vcpu->arch.msr = new_msr;
if (vcpu->arch.msr & MSR_WE) {
diff --git a/arch/powerpc/kvm/booke_interrupts.S b/arch/powerpc/kvm/booke_interrupts.S
index d0c6f84..12383fe 100644
--- a/arch/powerpc/kvm/booke_interrupts.S
+++ b/arch/powerpc/kvm/booke_interrupts.S
@@ -192,6 +192,12 @@ _GLOBAL(kvmppc_resume_host)
lwz r3, VCPU_HOST_PID(r4)
mtspr SPRN_PID, r3
+#ifdef CONFIG_E500
+ /* we cheat and know 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
@@ -350,6 +356,11 @@ lightweight_exit:
lwz r3, VCPU_SHADOW_PID(r4)
mtspr SPRN_PID, r3
+#ifdef CONFIG_E500
+ lwz r3, VCPU_SWAP_PID(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 3f76041..8d85655 100644
--- a/arch/powerpc/kvm/e500_emulate.c
+++ b/arch/powerpc/kvm/e500_emulate.c
@@ -76,10 +76,14 @@ int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
int emulated = EMULATE_DONE;
switch (sprn) {
- case SPRN_PID:
- vcpu_e500->pid[0] = vcpu->arch.shadow_pid =
- vcpu->arch.pid = vcpu->arch.gpr[rs];
+ case SPRN_PID: {
+ unsigned int as = !!(vcpu->arch.msr & (MSR_IS | MSR_DS));
+
+ vcpu_e500->pid[0] = vcpu->arch.pid = vcpu->arch.gpr[rs];
+ vcpu->arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500, as,
+ get_cur_pid(vcpu));
break;
+ }
case SPRN_PID1:
vcpu_e500->pid[1] = vcpu->arch.gpr[rs]; break;
case SPRN_PID2:
diff --git a/arch/powerpc/kvm/e500_tlb.c b/arch/powerpc/kvm/e500_tlb.c
index 847dfec..7b614a9 100644
--- a/arch/powerpc/kvm/e500_tlb.c
+++ b/arch/powerpc/kvm/e500_tlb.c
@@ -27,6 +27,101 @@
static unsigned int tlb1_entry_num;
+struct id_mapping {
+ unsigned char id[2][256];
+};
+
+struct shadow_id {
+ void *ref;
+};
+
+static DEFINE_PER_CPU(struct shadow_id[256], host_sid);
+
+static inline int e500_id_create_mapping(unsigned char *entry)
+{
+ unsigned long sid;
+
+ preempt_disable();
+ sid = (unsigned long)++(__get_cpu_var(host_sid)[0].ref);
+ if (sid < 256) {
+ *entry = (unsigned char)sid;
+ __get_cpu_var(host_sid)[sid].ref = entry;
+ return sid;
+ }
+ preempt_enable();
+
+ return -1;
+}
+
+static inline void e500_id_destroy_all(void)
+{
+ preempt_disable();
+ memset(__get_cpu_var(host_sid), 0, sizeof(__get_cpu_var(host_sid)));
+ preempt_enable();
+}
+
+static inline int e500_id_find_mapping(unsigned char *entry)
+{
+ if (*entry && __get_cpu_var(host_sid)[*entry].ref == entry)
+ return *entry;
+ return -1;
+}
+
+static int kvmppc_e500_alloc_idm(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+ vcpu_e500->id_mapping =
+ (struct id_mapping *)kzalloc(sizeof(struct id_mapping), GFP_KERNEL);
+ if (vcpu_e500->id_mapping != NULL)
+ return 0;
+ return -1;
+}
+
+static void kvmppc_e500_free_idm(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+ kfree(vcpu_e500->id_mapping);
+ vcpu_e500->id_mapping = NULL;
+ return;
+}
+
+static inline void kvmppc_e500_reset_idm(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+ memset(vcpu_e500->id_mapping, 0, sizeof(struct id_mapping));
+}
+
+static void inline kvmppc_e500_update_spid(struct kvmppc_vcpu_e500 *vcpu_e500)
+{
+ unsigned int as = !!(vcpu_e500->vcpu.arch.msr & (MSR_IS | MSR_DS));
+
+ vcpu_e500->vcpu.arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500, as,
+ get_cur_pid(&vcpu_e500->vcpu));
+ vcpu_e500->vcpu.arch.swap_pid = kvmppc_e500_get_sid(vcpu_e500, as, 0);
+}
+
+/*
+ * Map guest (vcpu,as,id) to individual shadow id.
+ */
+unsigned int kvmppc_e500_get_sid(struct kvmppc_vcpu_e500 *vcpu_e500,
+ int as, int gid)
+{
+ struct id_mapping *idm = vcpu_e500->id_mapping;
+ int sid;
+
+ sid = e500_id_find_mapping(&idm->id[as][gid]);
+
+ while (sid <= 0) {
+ /* None mapping yet */
+ sid = e500_id_create_mapping(&idm->id[as][gid]);
+ if(sid <= 0) {
+ BUG_ON(sid == 0);
+ _tlbil_all();
+ e500_id_destroy_all();
+ kvmppc_e500_update_spid(vcpu_e500);
+ }
+ }
+
+ return sid;
+}
+
void kvmppc_dump_tlbs(struct kvm_vcpu *vcpu)
{
struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
@@ -131,11 +226,14 @@ static inline void write_host_tlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
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 */
+ kvmppc_e500_update_spid(vcpu_e500);
}
void kvmppc_e500_tlb_put(struct kvm_vcpu *vcpu)
{
- _tlbil_all();
}
/* Search the guest TLB for a matching entry. */
@@ -245,10 +343,14 @@ static inline void kvmppc_e500_setup_stlbe(struct kvmppc_vcpu_e500 *vcpu_e500,
u64 gvaddr, struct tlbe *stlbe)
{
hpa_t hpaddr = page_to_phys(ref->page);
+ unsigned int stid;
+
+ stid = kvmppc_e500_get_sid(vcpu_e500, get_tlb_ts(gtlbe),
+ get_tlb_tid(gtlbe));
/* Force TS=1 IPROT=0 TSIZE=4KB for all guest mappings. */
stlbe->mas1 = MAS1_TSIZE(BOOKE_PAGESZ_4K)
- | 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.msr & MSR_PR);
@@ -315,11 +417,29 @@ static int kvmppc_e500_tlb1_map(struct kvmppc_vcpu_e500 *vcpu_e500,
* proper permission bits. */
void kvmppc_mmu_priv_switch(struct kvm_vcpu *vcpu, int usermode)
{
+ struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+
if (usermode) {
- _tlbil_all();
+ /* clear PID for guest kernel mapping */
+ vcpu->arch.swap_pid = 0;
+ } else {
+ /* set PID for guest kernel mapping
+ * We assume:
+ * 1. AS = 0 when enter supervise mode
+ * 2. TID = 0 for supervise code/data mappings */
+ vcpu->arch.swap_pid = kvmppc_e500_get_sid(vcpu_e500, 0, 0);
}
}
+void kvmppc_mmu_as_switch(struct kvm_vcpu *vcpu, u32 new_as)
+{
+ struct kvmppc_vcpu_e500 *vcpu_e500 = to_e500(vcpu);
+ u32 pid = get_cur_pid(vcpu);
+
+ vcpu->arch.shadow_pid = kvmppc_e500_get_sid(vcpu_e500, !!new_as, pid);
+ vcpu->arch.swap_pid = kvmppc_e500_get_sid(vcpu_e500, !!new_as, 0);
+}
+
static inline int kvmppc_e500_gtlbe_invalidate(
struct kvmppc_vcpu_e500 *vcpu_e500,
int tlbsel, int esel)
@@ -345,7 +465,9 @@ int kvmppc_e500_emul_mt_mmucsr0(struct kvmppc_vcpu_e500 *vcpu_e500, ulong value)
for (esel = 0; esel < vcpu_e500->guest_tlb_size[1]; esel++)
kvmppc_e500_gtlbe_invalidate(vcpu_e500, 1, esel);
- _tlbil_all();
+ /* Reset vcpu shadow id mapping */
+ kvmppc_e500_reset_idm(vcpu_e500);
+ kvmppc_e500_update_spid(vcpu_e500);
return EMULATE_DONE;
}
@@ -376,7 +498,9 @@ int kvmppc_e500_emul_tlbivax(struct kvm_vcpu *vcpu, int ra, int rb)
kvmppc_e500_gtlbe_invalidate(vcpu_e500, tlbsel, esel);
}
- _tlbil_all();
+ /* Reset vcpu shadow id mapping */
+ kvmppc_e500_reset_idm(vcpu_e500);
+ kvmppc_e500_update_spid(vcpu_e500);
return EMULATE_DONE;
}
@@ -555,9 +679,6 @@ void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
for (stlbsel = 0; stlbsel < 2; stlbsel++)
for (i = 0; i < vcpu_e500->guest_tlb_size[stlbsel]; i++)
kvmppc_e500_shadow_release(vcpu_e500, stlbsel, i);
-
- /* discard all guest mapping */
- _tlbil_all();
}
void kvmppc_mmu_map(struct kvm_vcpu *vcpu, u64 eaddr, gpa_t gpaddr,
@@ -629,6 +750,9 @@ void kvmppc_e500_tlb_setup(struct kvmppc_vcpu_e500 *vcpu_e500)
tlbe->mas2 = (0xe0004500 & 0xFFFFF000) | MAS2_I | MAS2_G;
tlbe->mas3 = (0xe0004500 & 0xFFFFF000) | E500_TLB_SUPER_PERM_MASK;
tlbe->mas7 = 0;
+
+ /* Setup shadow PID before start guest */
+ kvmppc_e500_update_spid(vcpu_e500);
}
int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
@@ -657,8 +781,13 @@ int kvmppc_e500_tlb_init(struct kvmppc_vcpu_e500 *vcpu_e500)
if (vcpu_e500->shadow_refs[1] == NULL)
goto err_out_ref0;
+ if(kvmppc_e500_alloc_idm(vcpu_e500) < 0)
+ goto err_out_ref1;
+
return 0;
+err_out_ref1:
+ kfree(vcpu_e500->shadow_refs[1]);
err_out_ref0:
kfree(vcpu_e500->shadow_refs[0]);
err_out_guest1:
@@ -671,6 +800,7 @@ err_out:
void kvmppc_e500_tlb_uninit(struct kvmppc_vcpu_e500 *vcpu_e500)
{
+ kvmppc_e500_free_idm(vcpu_e500);
kfree(vcpu_e500->shadow_refs[1]);
kfree(vcpu_e500->shadow_refs[0]);
kfree(vcpu_e500->guest_tlb[1]);
diff --git a/arch/powerpc/kvm/e500_tlb.h b/arch/powerpc/kvm/e500_tlb.h
index 45b064b..5c6e56c 100644
--- a/arch/powerpc/kvm/e500_tlb.h
+++ b/arch/powerpc/kvm/e500_tlb.h
@@ -55,6 +55,8 @@ 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 unsigned int kvmppc_e500_get_sid(struct kvmppc_vcpu_e500 *vcpu_e500,
+ int as, int gid);
/* TLB helper functions */
static inline unsigned int get_tlb_size(const struct tlbe *tlbe)
--
1.5.4
--
To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
