KVM: PPC: booke: category E.HV (GS-mode) support
authorScott Wood <scottwood@freescale.com>
Tue, 20 Dec 2011 15:34:43 +0000 (15:34 +0000)
committerAvi Kivity <avi@redhat.com>
Sun, 8 Apr 2012 09:51:19 +0000 (12:51 +0300)
Chips such as e500mc that implement category E.HV in Power ISA 2.06
provide hardware virtualization features, including a new MSR mode for
guest state.  The guest OS can perform many operations without trapping
into the hypervisor, including transitions to and from guest userspace.

Since we can use SRR1[GS] to reliably tell whether an exception came from
guest state, instead of messing around with IVPR, we use DO_KVM similarly
to book3s.

Current issues include:
 - Machine checks from guest state are not routed to the host handler.
 - The guest can cause a host oops by executing an emulated instruction
   in a page that lacks read permission.  Existing e500/4xx support has
   the same problem.

Includes work by Ashish Kalra <Ashish.Kalra@freescale.com>,
Varun Sethi <Varun.Sethi@freescale.com>, and
Liu Yu <yu.liu@freescale.com>.

Signed-off-by: Scott Wood <scottwood@freescale.com>
[agraf: remove pt_regs usage]
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
18 files changed:
arch/powerpc/include/asm/dbell.h
arch/powerpc/include/asm/kvm_asm.h
arch/powerpc/include/asm/kvm_booke_hv_asm.h [new file with mode: 0644]
arch/powerpc/include/asm/kvm_host.h
arch/powerpc/include/asm/kvm_ppc.h
arch/powerpc/include/asm/mmu-book3e.h
arch/powerpc/include/asm/processor.h
arch/powerpc/include/asm/reg.h
arch/powerpc/include/asm/reg_booke.h
arch/powerpc/kernel/asm-offsets.c
arch/powerpc/kernel/head_booke.h
arch/powerpc/kvm/Kconfig
arch/powerpc/kvm/booke.c
arch/powerpc/kvm/booke.h
arch/powerpc/kvm/booke_emulate.c
arch/powerpc/kvm/bookehv_interrupts.S [new file with mode: 0644]
arch/powerpc/kvm/powerpc.c
arch/powerpc/kvm/timing.h

index efa74ac..d7365b0 100644 (file)
@@ -19,6 +19,7 @@
 
 #define PPC_DBELL_MSG_BRDCAST  (0x04000000)
 #define PPC_DBELL_TYPE(x)      (((x) & 0xf) << (63-36))
+#define PPC_DBELL_LPID(x)      ((x) << (63 - 49))
 enum ppc_dbell {
        PPC_DBELL = 0,          /* doorbell */
        PPC_DBELL_CRIT = 1,     /* critical doorbell */
index 7b1f0e0..0978152 100644 (file)
 #define BOOKE_INTERRUPT_SPE_FP_DATA 33
 #define BOOKE_INTERRUPT_SPE_FP_ROUND 34
 #define BOOKE_INTERRUPT_PERFORMANCE_MONITOR 35
+#define BOOKE_INTERRUPT_DOORBELL 36
+#define BOOKE_INTERRUPT_DOORBELL_CRITICAL 37
+
+/* booke_hv */
+#define BOOKE_INTERRUPT_GUEST_DBELL 38
+#define BOOKE_INTERRUPT_GUEST_DBELL_CRIT 39
+#define BOOKE_INTERRUPT_HV_SYSCALL 40
+#define BOOKE_INTERRUPT_HV_PRIV 41
 
 /* book3s */
 
diff --git a/arch/powerpc/include/asm/kvm_booke_hv_asm.h b/arch/powerpc/include/asm/kvm_booke_hv_asm.h
new file mode 100644 (file)
index 0000000..30a600f
--- /dev/null
@@ -0,0 +1,49 @@
+/*
+ * Copyright 2010-2011 Freescale Semiconductor, Inc.
+ *
+ * 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.
+ */
+
+#ifndef ASM_KVM_BOOKE_HV_ASM_H
+#define ASM_KVM_BOOKE_HV_ASM_H
+
+#ifdef __ASSEMBLY__
+
+/*
+ * All exceptions from guest state must go through KVM
+ * (except for those which are delivered directly to the guest) --
+ * there are no exceptions for which we fall through directly to
+ * the normal host handler.
+ *
+ * Expected inputs (normal exceptions):
+ *   SCRATCH0 = saved r10
+ *   r10 = thread struct
+ *   r11 = appropriate SRR1 variant (currently used as scratch)
+ *   r13 = saved CR
+ *   *(r10 + THREAD_NORMSAVE(0)) = saved r11
+ *   *(r10 + THREAD_NORMSAVE(2)) = saved r13
+ *
+ * Expected inputs (crit/mcheck/debug exceptions):
+ *   appropriate SCRATCH = saved r8
+ *   r8 = exception level stack frame
+ *   r9 = *(r8 + _CCR) = saved CR
+ *   r11 = appropriate SRR1 variant (currently used as scratch)
+ *   *(r8 + GPR9) = saved r9
+ *   *(r8 + GPR10) = saved r10 (r10 not yet clobbered)
+ *   *(r8 + GPR11) = saved r11
+ */
+.macro DO_KVM intno srr1
+#ifdef CONFIG_KVM_BOOKE_HV
+BEGIN_FTR_SECTION
+       mtocrf  0x80, r11       /* check MSR[GS] without clobbering reg */
+       bf      3, kvmppc_resume_\intno\()_\srr1
+       b       kvmppc_handler_\intno\()_\srr1
+kvmppc_resume_\intno\()_\srr1:
+END_FTR_SECTION_IFSET(CPU_FTR_EMB_HV)
+#endif
+.endm
+
+#endif /*__ASSEMBLY__ */
+#endif /* ASM_KVM_BOOKE_HV_ASM_H */
index 5b81cbc..e645623 100644 (file)
@@ -106,6 +106,8 @@ struct kvm_vcpu_stat {
        u32 dec_exits;
        u32 ext_intr_exits;
        u32 halt_wakeup;
+       u32 dbell_exits;
+       u32 gdbell_exits;
 #ifdef CONFIG_PPC_BOOK3S
        u32 pf_storage;
        u32 pf_instruc;
@@ -140,6 +142,7 @@ enum kvm_exit_types {
        EMULATED_TLBSX_EXITS,
        EMULATED_TLBWE_EXITS,
        EMULATED_RFI_EXITS,
+       EMULATED_RFCI_EXITS,
        DEC_EXITS,
        EXT_INTR_EXITS,
        HALT_WAKEUP,
@@ -147,6 +150,8 @@ enum kvm_exit_types {
        FP_UNAVAIL,
        DEBUG_EXITS,
        TIMEINGUEST,
+       DBELL_EXITS,
+       GDBELL_EXITS,
        __NUMBER_OF_KVM_EXIT_TYPES
 };
 
@@ -217,10 +222,10 @@ struct kvm_arch_memory_slot {
 };
 
 struct kvm_arch {
+       unsigned int lpid;
 #ifdef CONFIG_KVM_BOOK3S_64_HV
        unsigned long hpt_virt;
        struct revmap_entry *revmap;
-       unsigned int lpid;
        unsigned int host_lpid;
        unsigned long host_lpcr;
        unsigned long sdr1;
@@ -345,6 +350,17 @@ struct kvm_vcpu_arch {
        u64 vsr[64];
 #endif
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       u32 host_mas4;
+       u32 host_mas6;
+       u32 shadow_epcr;
+       u32 epcr;
+       u32 shadow_msrp;
+       u32 eplc;
+       u32 epsc;
+       u32 oldpir;
+#endif
+
 #ifdef CONFIG_PPC_BOOK3S
        /* For Gekko paired singles */
        u32 qpr[32];
@@ -428,6 +444,7 @@ struct kvm_vcpu_arch {
        ulong queued_esr;
        u32 tlbcfg[4];
        u32 mmucfg;
+       u32 epr;
 #endif
        gpa_t paddr_accessed;
 
index 731e920..e709975 100644 (file)
@@ -139,6 +139,9 @@ extern int kvmppc_core_prepare_memory_region(struct kvm *kvm,
 extern void kvmppc_core_commit_memory_region(struct kvm *kvm,
                                struct kvm_userspace_memory_region *mem);
 
+extern int kvmppc_bookehv_init(void);
+extern void kvmppc_bookehv_exit(void);
+
 /*
  * Cuts out inst bits with ordering according to spec.
  * That means the leftmost bit is zero. All given bits are included.
index cdb5421..eeabcdb 100644 (file)
 #define MAS4_TSIZED_MASK       0x00000f80      /* Default TSIZE */
 #define MAS4_TSIZED_SHIFT      7
 
+#define MAS5_SGS               0x80000000
+
 #define MAS6_SPID0             0x3FFF0000
 #define MAS6_SPID1             0x00007FFE
 #define MAS6_ISIZE(x)          MAS1_TSIZE(x)
 
 #define MAS7_RPN               0xFFFFFFFF
 
+#define MAS8_TGS               0x80000000 /* Guest space */
+#define MAS8_VF                        0x40000000 /* Virtualization Fault */
+#define MAS8_TLPID             0x000000ff
+
 /* Bit definitions for MMUCFG */
 #define MMUCFG_MAVN    0x00000003      /* MMU Architecture Version Number */
 #define MMUCFG_MAVN_V1 0x00000000      /* v1.0 */
index 8e2d037..2a25ab0 100644 (file)
@@ -243,6 +243,9 @@ struct thread_struct {
 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
        void*           kvm_shadow_vcpu; /* KVM internal data */
 #endif /* CONFIG_KVM_BOOK3S_32_HANDLER */
+#if defined(CONFIG_KVM) && defined(CONFIG_BOOKE)
+       struct kvm_vcpu *kvm_vcpu;
+#endif
 #ifdef CONFIG_PPC64
        unsigned long   dscr;
        int             dscr_inherit;
index 9d7f0fb..f0cb7f4 100644 (file)
 #define   LPCR_LPES_SH 2
 #define   LPCR_RMI     0x00000002      /* real mode is cache inhibit */
 #define   LPCR_HDICE   0x00000001      /* Hyp Decr enable (HV,PR,EE) */
+#ifndef SPRN_LPID
 #define SPRN_LPID      0x13F   /* Logical Partition Identifier */
+#endif
 #define   LPID_RSVD    0x3ff           /* Reserved LPID for partn switching */
 #define        SPRN_HMER       0x150   /* Hardware m? error recovery */
 #define        SPRN_HMEER      0x151   /* Hardware m? enable error recovery */
index b86faa9..815e404 100644 (file)
@@ -61,18 +61,30 @@ extern u32 booke_wdt_period;
 #define SPRN_SPRG7W    0x117   /* Special Purpose Register General 7 Write */
 #define SPRN_EPCR      0x133   /* Embedded Processor Control Register */
 #define SPRN_DBCR2     0x136   /* Debug Control Register 2 */
+#define SPRN_MSRP      0x137   /* MSR Protect Register */
 #define SPRN_IAC3      0x13A   /* Instruction Address Compare 3 */
 #define SPRN_IAC4      0x13B   /* Instruction Address Compare 4 */
 #define SPRN_DVC1      0x13E   /* Data Value Compare Register 1 */
 #define SPRN_DVC2      0x13F   /* Data Value Compare Register 2 */
+#define SPRN_LPID      0x152   /* Logical Partition ID */
 #define SPRN_MAS8      0x155   /* MMU Assist Register 8 */
 #define SPRN_TLB0PS    0x158   /* TLB 0 Page Size Register */
 #define SPRN_TLB1PS    0x159   /* TLB 1 Page Size Register */
 #define SPRN_MAS5_MAS6 0x15c   /* MMU Assist Register 5 || 6 */
 #define SPRN_MAS8_MAS1 0x15d   /* MMU Assist Register 8 || 1 */
 #define SPRN_EPTCFG    0x15e   /* Embedded Page Table Config */
+#define SPRN_GSPRG0    0x170   /* Guest SPRG0 */
+#define SPRN_GSPRG1    0x171   /* Guest SPRG1 */
+#define SPRN_GSPRG2    0x172   /* Guest SPRG2 */
+#define SPRN_GSPRG3    0x173   /* Guest SPRG3 */
 #define SPRN_MAS7_MAS3 0x174   /* MMU Assist Register 7 || 3 */
 #define SPRN_MAS0_MAS1 0x175   /* MMU Assist Register 0 || 1 */
+#define SPRN_GSRR0     0x17A   /* Guest SRR0 */
+#define SPRN_GSRR1     0x17B   /* Guest SRR1 */
+#define SPRN_GEPR      0x17C   /* Guest EPR */
+#define SPRN_GDEAR     0x17D   /* Guest DEAR */
+#define SPRN_GPIR      0x17E   /* Guest PIR */
+#define SPRN_GESR      0x17F   /* Guest Exception Syndrome Register */
 #define SPRN_IVOR0     0x190   /* Interrupt Vector Offset Register 0 */
 #define SPRN_IVOR1     0x191   /* Interrupt Vector Offset Register 1 */
 #define SPRN_IVOR2     0x192   /* Interrupt Vector Offset Register 2 */
@@ -93,6 +105,13 @@ extern u32 booke_wdt_period;
 #define SPRN_IVOR39    0x1B1   /* Interrupt Vector Offset Register 39 */
 #define SPRN_IVOR40    0x1B2   /* Interrupt Vector Offset Register 40 */
 #define SPRN_IVOR41    0x1B3   /* Interrupt Vector Offset Register 41 */
+#define SPRN_GIVOR2    0x1B8   /* Guest IVOR2 */
+#define SPRN_GIVOR3    0x1B9   /* Guest IVOR3 */
+#define SPRN_GIVOR4    0x1BA   /* Guest IVOR4 */
+#define SPRN_GIVOR8    0x1BB   /* Guest IVOR8 */
+#define SPRN_GIVOR13   0x1BC   /* Guest IVOR13 */
+#define SPRN_GIVOR14   0x1BD   /* Guest IVOR14 */
+#define SPRN_GIVPR     0x1BF   /* Guest IVPR */
 #define SPRN_SPEFSCR   0x200   /* SPE & Embedded FP Status & Control */
 #define SPRN_BBEAR     0x201   /* Branch Buffer Entry Address Register */
 #define SPRN_BBTAR     0x202   /* Branch Buffer Target Address Register */
@@ -245,6 +264,10 @@ extern u32 booke_wdt_period;
 #define MCSR_LDG       0x00002000UL /* Guarded Load */
 #define MCSR_TLBSYNC   0x00000002UL /* Multiple tlbsyncs detected */
 #define MCSR_BSL2_ERR  0x00000001UL /* Backside L2 cache error */
+
+#define MSRP_UCLEP     0x04000000 /* Protect MSR[UCLE] */
+#define MSRP_DEP       0x00000200 /* Protect MSR[DE] */
+#define MSRP_PMMP      0x00000004 /* Protect MSR[PMM] */
 #endif
 
 #ifdef CONFIG_E200
@@ -599,6 +622,17 @@ extern u32 booke_wdt_period;
 #define SPRN_EPCR_DMIUH                0x00400000      /* Disable MAS Interrupt updates
                                                 * for hypervisor */
 
+/* Bit definitions for EPLC/EPSC */
+#define EPC_EPR                0x80000000 /* 1 = user, 0 = kernel */
+#define EPC_EPR_SHIFT  31
+#define EPC_EAS                0x40000000 /* Address Space */
+#define EPC_EAS_SHIFT  30
+#define EPC_EGS                0x20000000 /* 1 = guest, 0 = hypervisor */
+#define EPC_EGS_SHIFT  29
+#define EPC_ELPID      0x00ff0000
+#define EPC_ELPID_SHIFT        16
+#define EPC_EPID       0x00003fff
+#define EPC_EPID_SHIFT 0
 
 /*
  * The IBM-403 is an even more odd special case, as it is much
index 34b8afe..bbede58 100644 (file)
@@ -116,6 +116,9 @@ int main(void)
 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
        DEFINE(THREAD_KVM_SVCPU, offsetof(struct thread_struct, kvm_shadow_vcpu));
 #endif
+#ifdef CONFIG_KVM_BOOKE_HV
+       DEFINE(THREAD_KVM_VCPU, offsetof(struct thread_struct, kvm_vcpu));
+#endif
 
        DEFINE(TI_FLAGS, offsetof(struct thread_info, flags));
        DEFINE(TI_LOCAL_FLAGS, offsetof(struct thread_info, local_flags));
@@ -387,6 +390,7 @@ int main(void)
 #ifdef CONFIG_KVM
        DEFINE(VCPU_HOST_STACK, offsetof(struct kvm_vcpu, arch.host_stack));
        DEFINE(VCPU_HOST_PID, offsetof(struct kvm_vcpu, arch.host_pid));
+       DEFINE(VCPU_GUEST_PID, offsetof(struct kvm_vcpu, arch.pid));
        DEFINE(VCPU_GPRS, offsetof(struct kvm_vcpu, arch.gpr));
        DEFINE(VCPU_VRSAVE, offsetof(struct kvm_vcpu, arch.vrsave));
        DEFINE(VCPU_FPRS, offsetof(struct kvm_vcpu, arch.fpr));
@@ -429,9 +433,11 @@ int main(void)
        DEFINE(VCPU_SHARED_MAS4, offsetof(struct kvm_vcpu_arch_shared, mas4));
        DEFINE(VCPU_SHARED_MAS6, offsetof(struct kvm_vcpu_arch_shared, mas6));
 
+       DEFINE(VCPU_KVM, offsetof(struct kvm_vcpu, kvm));
+       DEFINE(KVM_LPID, offsetof(struct kvm, arch.lpid));
+
        /* book3s */
 #ifdef CONFIG_KVM_BOOK3S_64_HV
-       DEFINE(KVM_LPID, offsetof(struct kvm, arch.lpid));
        DEFINE(KVM_SDR1, offsetof(struct kvm, arch.sdr1));
        DEFINE(KVM_HOST_LPID, offsetof(struct kvm, arch.host_lpid));
        DEFINE(KVM_HOST_LPCR, offsetof(struct kvm, arch.host_lpcr));
@@ -446,7 +452,6 @@ int main(void)
        DEFINE(VCPU_DAR, offsetof(struct kvm_vcpu, arch.shregs.dar));
 #endif
 #ifdef CONFIG_PPC_BOOK3S
-       DEFINE(VCPU_KVM, offsetof(struct kvm_vcpu, kvm));
        DEFINE(VCPU_VCPUID, offsetof(struct kvm_vcpu, vcpu_id));
        DEFINE(VCPU_PURR, offsetof(struct kvm_vcpu, arch.purr));
        DEFINE(VCPU_SPURR, offsetof(struct kvm_vcpu, arch.spurr));
@@ -597,6 +602,12 @@ int main(void)
        DEFINE(VCPU_HOST_SPEFSCR, offsetof(struct kvm_vcpu, arch.host_spefscr));
 #endif
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       DEFINE(VCPU_HOST_MAS4, offsetof(struct kvm_vcpu, arch.host_mas4));
+       DEFINE(VCPU_HOST_MAS6, offsetof(struct kvm_vcpu, arch.host_mas6));
+       DEFINE(VCPU_EPLC, offsetof(struct kvm_vcpu, arch.eplc));
+#endif
+
 #ifdef CONFIG_KVM_EXIT_TIMING
        DEFINE(VCPU_TIMING_EXIT_TBU, offsetof(struct kvm_vcpu,
                                                arch.timing_exit.tv32.tbu));
index 51fd072..5f051ee 100644 (file)
@@ -3,6 +3,7 @@
 
 #include <asm/ptrace.h>        /* for STACK_FRAME_REGS_MARKER */
 #include <asm/kvm_asm.h>
+#include <asm/kvm_booke_hv_asm.h>
 
 /*
  * Macros used for common Book-e exception handling
@@ -36,8 +37,9 @@
        stw     r11, THREAD_NORMSAVE(0)(r10);                                \
        stw     r13, THREAD_NORMSAVE(2)(r10);                                \
        mfcr    r13;                    /* save CR in r13 for now          */\
-       mfspr   r11,SPRN_SRR1;          /* check whether user or kernel    */\
-       andi.   r11,r11,MSR_PR;                                              \
+       mfspr   r11, SPRN_SRR1;                                              \
+       DO_KVM  BOOKE_INTERRUPT_##intno SPRN_SRR1;                           \
+       andi.   r11, r11, MSR_PR;       /* check whether user or kernel    */\
        mr      r11, r1;                                                     \
        beq     1f;                                                          \
        /* if from user, start at top of this thread's kernel stack */       \
        stw     r10,GPR10(r8);                                               \
        stw     r11,GPR11(r8);                                               \
        stw     r9,_CCR(r8);            /* save CR on stack                */\
-       mfspr   r10,exc_level_srr1;     /* check whether user or kernel    */\
-       andi.   r10,r10,MSR_PR;                                              \
+       mfspr   r11,exc_level_srr1;     /* check whether user or kernel    */\
+       DO_KVM  BOOKE_INTERRUPT_##intno exc_level_srr1;                      \
+       andi.   r11,r11,MSR_PR;                                              \
        mfspr   r11,SPRN_SPRG_THREAD;   /* if from user, start at top of   */\
        lwz     r11,THREAD_INFO-THREAD(r11); /* this thread's kernel stack */\
        addi    r11,r11,EXC_LVL_FRAME_OVERHEAD; /* allocate stack frame    */\
                        SPRN_MCSRR0, SPRN_MCSRR1)
 
 /*
+ * Guest Doorbell -- this is a bit odd in that uses GSRR0/1 despite
+ * being delivered to the host.  This exception can only happen
+ * inside a KVM guest -- so we just handle up to the DO_KVM rather
+ * than try to fit this into one of the existing prolog macros.
+ */
+#define GUEST_DOORBELL_EXCEPTION \
+       START_EXCEPTION(GuestDoorbell);                                      \
+       mtspr   SPRN_SPRG_WSCRATCH0, r10;       /* save one register */      \
+       mfspr   r10, SPRN_SPRG_THREAD;                                       \
+       stw     r11, THREAD_NORMSAVE(0)(r10);                                \
+       mfspr   r11, SPRN_SRR1;                                              \
+       stw     r13, THREAD_NORMSAVE(2)(r10);                                \
+       mfcr    r13;                    /* save CR in r13 for now          */\
+       DO_KVM  BOOKE_INTERRUPT_GUEST_DBELL SPRN_GSRR1;                      \
+       trap
+
+/*
  * Exception vectors.
  */
 #define        START_EXCEPTION(label)                                               \
index 8f64709..2c33cd3 100644 (file)
@@ -90,6 +90,9 @@ config KVM_BOOK3S_64_PR
        depends on KVM_BOOK3S_64 && !KVM_BOOK3S_64_HV
        select KVM_BOOK3S_PR
 
+config KVM_BOOKE_HV
+       bool
+
 config KVM_440
        bool "KVM support for PowerPC 440 processors"
        depends on EXPERIMENTAL && 44x
index 2ee9bae..75dbaeb 100644 (file)
@@ -17,6 +17,8 @@
  *
  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
+ *          Scott Wood <scottwood@freescale.com>
+ *          Varun Sethi <varun.sethi@freescale.com>
  */
 
 #include <linux/errno.h>
 #include <asm/cputable.h>
 #include <asm/uaccess.h>
 #include <asm/kvm_ppc.h>
-#include "timing.h"
 #include <asm/cacheflush.h>
+#include <asm/dbell.h>
+#include <asm/hw_irq.h>
+#include <asm/irq.h>
 
+#include "timing.h"
 #include "booke.h"
 
 unsigned long kvmppc_booke_handlers;
@@ -55,6 +60,8 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
        { "dec",        VCPU_STAT(dec_exits) },
        { "ext_intr",   VCPU_STAT(ext_intr_exits) },
        { "halt_wakeup", VCPU_STAT(halt_wakeup) },
+       { "doorbell", VCPU_STAT(dbell_exits) },
+       { "guest doorbell", VCPU_STAT(gdbell_exits) },
        { NULL }
 };
 
@@ -121,6 +128,10 @@ void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
 {
        u32 old_msr = vcpu->arch.shared->msr;
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       new_msr |= MSR_GS;
+#endif
+
        vcpu->arch.shared->msr = new_msr;
 
        kvmppc_mmu_msr_notify(vcpu, old_msr);
@@ -195,6 +206,75 @@ void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
        clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
 }
 
+static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
+{
+#ifdef CONFIG_KVM_BOOKE_HV
+       mtspr(SPRN_GSRR0, srr0);
+       mtspr(SPRN_GSRR1, srr1);
+#else
+       vcpu->arch.shared->srr0 = srr0;
+       vcpu->arch.shared->srr1 = srr1;
+#endif
+}
+
+static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
+{
+       vcpu->arch.csrr0 = srr0;
+       vcpu->arch.csrr1 = srr1;
+}
+
+static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
+{
+       if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
+               vcpu->arch.dsrr0 = srr0;
+               vcpu->arch.dsrr1 = srr1;
+       } else {
+               set_guest_csrr(vcpu, srr0, srr1);
+       }
+}
+
+static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
+{
+       vcpu->arch.mcsrr0 = srr0;
+       vcpu->arch.mcsrr1 = srr1;
+}
+
+static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
+{
+#ifdef CONFIG_KVM_BOOKE_HV
+       return mfspr(SPRN_GDEAR);
+#else
+       return vcpu->arch.shared->dar;
+#endif
+}
+
+static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
+{
+#ifdef CONFIG_KVM_BOOKE_HV
+       mtspr(SPRN_GDEAR, dear);
+#else
+       vcpu->arch.shared->dar = dear;
+#endif
+}
+
+static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
+{
+#ifdef CONFIG_KVM_BOOKE_HV
+       return mfspr(SPRN_GESR);
+#else
+       return vcpu->arch.shared->esr;
+#endif
+}
+
+static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
+{
+#ifdef CONFIG_KVM_BOOKE_HV
+       mtspr(SPRN_GESR, esr);
+#else
+       vcpu->arch.shared->esr = esr;
+#endif
+}
+
 /* Deliver the interrupt of the corresponding priority, if possible. */
 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
                                         unsigned int priority)
@@ -206,6 +286,7 @@ static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
        ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
        bool crit;
        bool keep_irq = false;
+       enum int_class int_class;
 
        /* Truncate crit indicators in 32 bit mode */
        if (!(vcpu->arch.shared->msr & MSR_SF)) {
@@ -241,16 +322,20 @@ static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
        case BOOKE_IRQPRIO_AP_UNAVAIL:
        case BOOKE_IRQPRIO_ALIGNMENT:
                allowed = 1;
-               msr_mask = MSR_CE|MSR_ME|MSR_DE;
+               msr_mask = MSR_GS | MSR_CE | MSR_ME | MSR_DE;
+               int_class = INT_CLASS_NONCRIT;
                break;
        case BOOKE_IRQPRIO_CRITICAL:
-       case BOOKE_IRQPRIO_WATCHDOG:
                allowed = vcpu->arch.shared->msr & MSR_CE;
-               msr_mask = MSR_ME;
+               allowed = allowed && !crit;
+               msr_mask = MSR_GS | MSR_ME;
+               int_class = INT_CLASS_CRIT;
                break;
        case BOOKE_IRQPRIO_MACHINE_CHECK:
                allowed = vcpu->arch.shared->msr & MSR_ME;
-               msr_mask = 0;
+               allowed = allowed && !crit;
+               msr_mask = MSR_GS;
+               int_class = INT_CLASS_MC;
                break;
        case BOOKE_IRQPRIO_DECREMENTER:
        case BOOKE_IRQPRIO_FIT:
@@ -259,28 +344,62 @@ static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
        case BOOKE_IRQPRIO_EXTERNAL:
                allowed = vcpu->arch.shared->msr & MSR_EE;
                allowed = allowed && !crit;
-               msr_mask = MSR_CE|MSR_ME|MSR_DE;
+               msr_mask = MSR_GS | MSR_CE | MSR_ME | MSR_DE;
+               int_class = INT_CLASS_NONCRIT;
                break;
        case BOOKE_IRQPRIO_DEBUG:
                allowed = vcpu->arch.shared->msr & MSR_DE;
-               msr_mask = MSR_ME;
+               allowed = allowed && !crit;
+               msr_mask = MSR_GS | MSR_ME;
+               int_class = INT_CLASS_CRIT;
                break;
        }
 
        if (allowed) {
-               vcpu->arch.shared->srr0 = vcpu->arch.pc;
-               vcpu->arch.shared->srr1 = vcpu->arch.shared->msr;
+               switch (int_class) {
+               case INT_CLASS_NONCRIT:
+                       set_guest_srr(vcpu, vcpu->arch.pc,
+                                     vcpu->arch.shared->msr);
+                       break;
+               case INT_CLASS_CRIT:
+                       set_guest_csrr(vcpu, vcpu->arch.pc,
+                                      vcpu->arch.shared->msr);
+                       break;
+               case INT_CLASS_DBG:
+                       set_guest_dsrr(vcpu, vcpu->arch.pc,
+                                      vcpu->arch.shared->msr);
+                       break;
+               case INT_CLASS_MC:
+                       set_guest_mcsrr(vcpu, vcpu->arch.pc,
+                                       vcpu->arch.shared->msr);
+                       break;
+               }
+
                vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
                if (update_esr == true)
-                       vcpu->arch.shared->esr = vcpu->arch.queued_esr;
+                       set_guest_esr(vcpu, vcpu->arch.queued_esr);
                if (update_dear == true)
-                       vcpu->arch.shared->dar = vcpu->arch.queued_dear;
+                       set_guest_dear(vcpu, vcpu->arch.queued_dear);
                kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
 
                if (!keep_irq)
                        clear_bit(priority, &vcpu->arch.pending_exceptions);
        }
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       /*
+        * If an interrupt is pending but masked, raise a guest doorbell
+        * so that we are notified when the guest enables the relevant
+        * MSR bit.
+        */
+       if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
+               kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
+       if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
+               kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
+       if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
+               kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
+#endif
+
        return allowed;
 }
 
@@ -344,6 +463,11 @@ int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
                return -EINVAL;
        }
 
+       if (!current->thread.kvm_vcpu) {
+               WARN(1, "no vcpu\n");
+               return -EPERM;
+       }
+
        local_irq_disable();
 
        kvmppc_core_prepare_to_enter(vcpu);
@@ -363,6 +487,38 @@ out:
        return ret;
 }
 
+static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
+{
+       enum emulation_result er;
+
+       er = kvmppc_emulate_instruction(run, vcpu);
+       switch (er) {
+       case EMULATE_DONE:
+               /* don't overwrite subtypes, just account kvm_stats */
+               kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
+               /* Future optimization: only reload non-volatiles if
+                * they were actually modified by emulation. */
+               return RESUME_GUEST_NV;
+
+       case EMULATE_DO_DCR:
+               run->exit_reason = KVM_EXIT_DCR;
+               return RESUME_HOST;
+
+       case EMULATE_FAIL:
+               /* XXX Deliver Program interrupt to guest. */
+               printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
+                      __func__, vcpu->arch.pc, vcpu->arch.last_inst);
+               /* For debugging, encode the failing instruction and
+                * report it to userspace. */
+               run->hw.hardware_exit_reason = ~0ULL << 32;
+               run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
+               return RESUME_HOST;
+
+       default:
+               BUG();
+       }
+}
+
 /**
  * kvmppc_handle_exit
  *
@@ -371,12 +527,30 @@ out:
 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
                        unsigned int exit_nr)
 {
-       enum emulation_result er;
        int r = RESUME_HOST;
 
        /* update before a new last_exit_type is rewritten */
        kvmppc_update_timing_stats(vcpu);
 
+       switch (exit_nr) {
+       case BOOKE_INTERRUPT_EXTERNAL:
+               do_IRQ(current->thread.regs);
+               break;
+
+       case BOOKE_INTERRUPT_DECREMENTER:
+               timer_interrupt(current->thread.regs);
+               break;
+
+#if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
+       case BOOKE_INTERRUPT_DOORBELL:
+               doorbell_exception(current->thread.regs);
+               break;
+#endif
+       case BOOKE_INTERRUPT_MACHINE_CHECK:
+               /* FIXME */
+               break;
+       }
+
        local_irq_enable();
 
        run->exit_reason = KVM_EXIT_UNKNOWN;
@@ -384,30 +558,56 @@ int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 
        switch (exit_nr) {
        case BOOKE_INTERRUPT_MACHINE_CHECK:
-               printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
-               kvmppc_dump_vcpu(vcpu);
-               r = RESUME_HOST;
+               kvm_resched(vcpu);
+               r = RESUME_GUEST;
                break;
 
        case BOOKE_INTERRUPT_EXTERNAL:
                kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
-               if (need_resched())
-                       cond_resched();
+               kvm_resched(vcpu);
                r = RESUME_GUEST;
                break;
 
        case BOOKE_INTERRUPT_DECREMENTER:
-               /* Since we switched IVPR back to the host's value, the host
-                * handled this interrupt the moment we enabled interrupts.
-                * Now we just offer it a chance to reschedule the guest. */
                kvmppc_account_exit(vcpu, DEC_EXITS);
-               if (need_resched())
-                       cond_resched();
+               kvm_resched(vcpu);
                r = RESUME_GUEST;
                break;
 
+       case BOOKE_INTERRUPT_DOORBELL:
+               kvmppc_account_exit(vcpu, DBELL_EXITS);
+               kvm_resched(vcpu);
+               r = RESUME_GUEST;
+               break;
+
+       case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
+               kvmppc_account_exit(vcpu, GDBELL_EXITS);
+
+               /*
+                * We are here because there is a pending guest interrupt
+                * which could not be delivered as MSR_CE or MSR_ME was not
+                * set.  Once we break from here we will retry delivery.
+                */
+               r = RESUME_GUEST;
+               break;
+
+       case BOOKE_INTERRUPT_GUEST_DBELL:
+               kvmppc_account_exit(vcpu, GDBELL_EXITS);
+
+               /*
+                * We are here because there is a pending guest interrupt
+                * which could not be delivered as MSR_EE was not set.  Once
+                * we break from here we will retry delivery.
+                */
+               r = RESUME_GUEST;
+               break;
+
+       case BOOKE_INTERRUPT_HV_PRIV:
+               r = emulation_exit(run, vcpu);
+               break;
+
        case BOOKE_INTERRUPT_PROGRAM:
-               if (vcpu->arch.shared->msr & MSR_PR) {
+               if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
                        /* Program traps generated by user-level software must be handled
                         * by the guest kernel. */
                        kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
@@ -416,32 +616,7 @@ int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
                        break;
                }
 
-               er = kvmppc_emulate_instruction(run, vcpu);
-               switch (er) {
-               case EMULATE_DONE:
-                       /* don't overwrite subtypes, just account kvm_stats */
-                       kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
-                       /* Future optimization: only reload non-volatiles if
-                        * they were actually modified by emulation. */
-                       r = RESUME_GUEST_NV;
-                       break;
-               case EMULATE_DO_DCR:
-                       run->exit_reason = KVM_EXIT_DCR;
-                       r = RESUME_HOST;
-                       break;
-               case EMULATE_FAIL:
-                       /* XXX Deliver Program interrupt to guest. */
-                       printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
-                              __func__, vcpu->arch.pc, vcpu->arch.last_inst);
-                       /* For debugging, encode the failing instruction and
-                        * report it to userspace. */
-                       run->hw.hardware_exit_reason = ~0ULL << 32;
-                       run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
-                       r = RESUME_HOST;
-                       break;
-               default:
-                       BUG();
-               }
+               r = emulation_exit(run, vcpu);
                break;
 
        case BOOKE_INTERRUPT_FP_UNAVAIL:
@@ -506,6 +681,21 @@ int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
                r = RESUME_GUEST;
                break;
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       case BOOKE_INTERRUPT_HV_SYSCALL:
+               if (!(vcpu->arch.shared->msr & MSR_PR)) {
+                       kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
+               } else {
+                       /*
+                        * hcall from guest userspace -- send privileged
+                        * instruction program check.
+                        */
+                       kvmppc_core_queue_program(vcpu, ESR_PPR);
+               }
+
+               r = RESUME_GUEST;
+               break;
+#else
        case BOOKE_INTERRUPT_SYSCALL:
                if (!(vcpu->arch.shared->msr & MSR_PR) &&
                    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
@@ -519,6 +709,7 @@ int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
                kvmppc_account_exit(vcpu, SYSCALL_EXITS);
                r = RESUME_GUEST;
                break;
+#endif
 
        case BOOKE_INTERRUPT_DTLB_MISS: {
                unsigned long eaddr = vcpu->arch.fault_dear;
@@ -659,12 +850,15 @@ int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
        int r;
 
        vcpu->arch.pc = 0;
-       vcpu->arch.shared->msr = 0;
-       vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
        vcpu->arch.shared->pir = vcpu->vcpu_id;
        kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
+       kvmppc_set_msr(vcpu, 0);
 
+#ifndef CONFIG_KVM_BOOKE_HV
+       vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
        vcpu->arch.shadow_pid = 1;
+       vcpu->arch.shared->msr = 0;
+#endif
 
        /* Eye-catching numbers so we know if the guest takes an interrupt
         * before it's programmed its own IVPR/IVORs. */
@@ -745,8 +939,8 @@ static void get_sregs_base(struct kvm_vcpu *vcpu,
        sregs->u.e.csrr0 = vcpu->arch.csrr0;
        sregs->u.e.csrr1 = vcpu->arch.csrr1;
        sregs->u.e.mcsr = vcpu->arch.mcsr;
-       sregs->u.e.esr = vcpu->arch.shared->esr;
-       sregs->u.e.dear = vcpu->arch.shared->dar;
+       sregs->u.e.esr = get_guest_esr(vcpu);
+       sregs->u.e.dear = get_guest_dear(vcpu);
        sregs->u.e.tsr = vcpu->arch.tsr;
        sregs->u.e.tcr = vcpu->arch.tcr;
        sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
@@ -763,8 +957,8 @@ static int set_sregs_base(struct kvm_vcpu *vcpu,
        vcpu->arch.csrr0 = sregs->u.e.csrr0;
        vcpu->arch.csrr1 = sregs->u.e.csrr1;
        vcpu->arch.mcsr = sregs->u.e.mcsr;
-       vcpu->arch.shared->esr = sregs->u.e.esr;
-       vcpu->arch.shared->dar = sregs->u.e.dear;
+       set_guest_esr(vcpu, sregs->u.e.esr);
+       set_guest_dear(vcpu, sregs->u.e.dear);
        vcpu->arch.vrsave = sregs->u.e.vrsave;
        kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
 
@@ -961,14 +1155,17 @@ void kvmppc_decrementer_func(unsigned long data)
 
 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 {
+       current->thread.kvm_vcpu = vcpu;
 }
 
 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
 {
+       current->thread.kvm_vcpu = NULL;
 }
 
 int __init kvmppc_booke_init(void)
 {
+#ifndef CONFIG_KVM_BOOKE_HV
        unsigned long ivor[16];
        unsigned long max_ivor = 0;
        int i;
@@ -1011,7 +1208,7 @@ int __init kvmppc_booke_init(void)
        }
        flush_icache_range(kvmppc_booke_handlers,
                           kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
-
+#endif /* !BOOKE_HV */
        return 0;
 }
 
index 05d1d99..d53bcf2 100644 (file)
 #define BOOKE_IRQPRIO_PERFORMANCE_MONITOR 19
 /* Internal pseudo-irqprio for level triggered externals */
 #define BOOKE_IRQPRIO_EXTERNAL_LEVEL 20
-#define BOOKE_IRQPRIO_MAX 20
+#define BOOKE_IRQPRIO_DBELL 21
+#define BOOKE_IRQPRIO_DBELL_CRIT 22
+#define BOOKE_IRQPRIO_MAX 23
+
+#define BOOKE_IRQMASK_EE ((1 << BOOKE_IRQPRIO_EXTERNAL_LEVEL) | \
+                         (1 << BOOKE_IRQPRIO_PERFORMANCE_MONITOR) | \
+                         (1 << BOOKE_IRQPRIO_DBELL) | \
+                         (1 << BOOKE_IRQPRIO_DECREMENTER) | \
+                         (1 << BOOKE_IRQPRIO_FIT) | \
+                         (1 << BOOKE_IRQPRIO_EXTERNAL))
+
+#define BOOKE_IRQMASK_CE ((1 << BOOKE_IRQPRIO_DBELL_CRIT) | \
+                         (1 << BOOKE_IRQPRIO_WATCHDOG) | \
+                         (1 << BOOKE_IRQPRIO_CRITICAL))
 
 extern unsigned long kvmppc_booke_handlers;
 
@@ -74,4 +87,13 @@ void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu);
 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu);
 
+enum int_class {
+       INT_CLASS_NONCRIT,
+       INT_CLASS_CRIT,
+       INT_CLASS_MC,
+       INT_CLASS_DBG,
+};
+
+void kvmppc_set_pending_interrupt(struct kvm_vcpu *vcpu, enum int_class type);
+
 #endif /* __KVM_BOOKE_H__ */
index 3e652da..904412b 100644 (file)
@@ -99,6 +99,12 @@ int kvmppc_booke_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
        return emulated;
 }
 
+/*
+ * NOTE: some of these registers are not emulated on BOOKE_HV (GS-mode).
+ * Their backing store is in real registers, and these functions
+ * will return the wrong result if called for them in another context
+ * (such as debugging).
+ */
 int kvmppc_booke_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
 {
        int emulated = EMULATE_DONE;
@@ -122,9 +128,11 @@ int kvmppc_booke_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
                kvmppc_set_tcr(vcpu, spr_val);
                break;
 
-       /* Note: SPRG4-7 are user-readable. These values are
-        * loaded into the real SPRGs when resuming the
-        * guest. */
+       /*
+        * Note: SPRG4-7 are user-readable.
+        * These values are loaded into the real SPRGs when resuming the
+        * guest (PR-mode only).
+        */
        case SPRN_SPRG4:
                vcpu->arch.shared->sprg4 = spr_val; break;
        case SPRN_SPRG5:
@@ -136,6 +144,9 @@ int kvmppc_booke_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
 
        case SPRN_IVPR:
                vcpu->arch.ivpr = spr_val;
+#ifdef CONFIG_KVM_BOOKE_HV
+               mtspr(SPRN_GIVPR, spr_val);
+#endif
                break;
        case SPRN_IVOR0:
                vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = spr_val;
@@ -145,6 +156,9 @@ int kvmppc_booke_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
                break;
        case SPRN_IVOR2:
                vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = spr_val;
+#ifdef CONFIG_KVM_BOOKE_HV
+               mtspr(SPRN_GIVOR2, spr_val);
+#endif
                break;
        case SPRN_IVOR3:
                vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = spr_val;
@@ -163,6 +177,9 @@ int kvmppc_booke_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
                break;
        case SPRN_IVOR8:
                vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = spr_val;
+#ifdef CONFIG_KVM_BOOKE_HV
+               mtspr(SPRN_GIVOR8, spr_val);
+#endif
                break;
        case SPRN_IVOR9:
                vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = spr_val;
diff --git a/arch/powerpc/kvm/bookehv_interrupts.S b/arch/powerpc/kvm/bookehv_interrupts.S
new file mode 100644 (file)
index 0000000..9eaeebd
--- /dev/null
@@ -0,0 +1,587 @@
+/*
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright (C) 2010-2011 Freescale Semiconductor, Inc.
+ *
+ * Author: Varun Sethi <varun.sethi@freescale.com>
+ * Author: Scott Wood <scotwood@freescale.com>
+ *
+ * This file is derived from arch/powerpc/kvm/booke_interrupts.S
+ */
+
+#include <asm/ppc_asm.h>
+#include <asm/kvm_asm.h>
+#include <asm/reg.h>
+#include <asm/mmu-44x.h>
+#include <asm/page.h>
+#include <asm/asm-compat.h>
+#include <asm/asm-offsets.h>
+#include <asm/bitsperlong.h>
+
+#include "../kernel/head_booke.h" /* for THREAD_NORMSAVE() */
+
+#define GET_VCPU(vcpu, thread) \
+       PPC_LL  vcpu, THREAD_KVM_VCPU(thread)
+
+#define SET_VCPU(vcpu)         \
+        PPC_STL        vcpu, (THREAD + THREAD_KVM_VCPU)(r2)
+
+#define LONGBYTES              (BITS_PER_LONG / 8)
+
+#define VCPU_GPR(n)            (VCPU_GPRS + (n * LONGBYTES))
+#define VCPU_GUEST_SPRG(n)     (VCPU_GUEST_SPRGS + (n * LONGBYTES))
+
+/* The host stack layout: */
+#define HOST_R1         (0 * LONGBYTES) /* Implied by stwu. */
+#define HOST_CALLEE_LR  (1 * LONGBYTES)
+#define HOST_RUN        (2 * LONGBYTES) /* struct kvm_run */
+/*
+ * r2 is special: it holds 'current', and it made nonvolatile in the
+ * kernel with the -ffixed-r2 gcc option.
+ */
+#define HOST_R2         (3 * LONGBYTES)
+#define HOST_NV_GPRS    (4 * LONGBYTES)
+#define HOST_NV_GPR(n)  (HOST_NV_GPRS + ((n - 14) * LONGBYTES))
+#define HOST_MIN_STACK_SIZE (HOST_NV_GPR(31) + LONGBYTES)
+#define HOST_STACK_SIZE ((HOST_MIN_STACK_SIZE + 15) & ~15) /* Align. */
+#define HOST_STACK_LR   (HOST_STACK_SIZE + LONGBYTES) /* In caller stack frame. */
+
+#define NEED_EMU               0x00000001 /* emulation -- save nv regs */
+#define NEED_DEAR              0x00000002 /* save faulting DEAR */
+#define NEED_ESR               0x00000004 /* save faulting ESR */
+
+/*
+ * On entry:
+ * r4 = vcpu, r5 = srr0, r6 = srr1
+ * saved in vcpu: cr, ctr, r3-r13
+ */
+.macro kvm_handler_common intno, srr0, flags
+       mfspr   r10, SPRN_PID
+       lwz     r8, VCPU_HOST_PID(r4)
+       PPC_LL  r11, VCPU_SHARED(r4)
+       PPC_STL r14, VCPU_GPR(r14)(r4) /* We need a non-volatile GPR. */
+       li      r14, \intno
+
+       stw     r10, VCPU_GUEST_PID(r4)
+       mtspr   SPRN_PID, r8
+
+       .if     \flags & NEED_EMU
+       lwz     r9, VCPU_KVM(r4)
+       .endif
+
+#ifdef CONFIG_KVM_EXIT_TIMING
+       /* save exit time */
+1:     mfspr   r7, SPRN_TBRU
+       mfspr   r8, SPRN_TBRL
+       mfspr   r9, SPRN_TBRU
+       cmpw    r9, r7
+       PPC_STL r8, VCPU_TIMING_EXIT_TBL(r4)
+       bne-    1b
+       PPC_STL r9, VCPU_TIMING_EXIT_TBU(r4)
+#endif
+
+       oris    r8, r6, MSR_CE@h
+#ifndef CONFIG_64BIT
+       stw     r6, (VCPU_SHARED_MSR + 4)(r11)
+#else
+       std     r6, (VCPU_SHARED_MSR)(r11)
+#endif
+       ori     r8, r8, MSR_ME | MSR_RI
+       PPC_STL r5, VCPU_PC(r4)
+
+       /*
+        * Make sure CE/ME/RI are set (if appropriate for exception type)
+        * whether or not the guest had it set.  Since mfmsr/mtmsr are
+        * somewhat expensive, skip in the common case where the guest
+        * had all these bits set (and thus they're still set if
+        * appropriate for the exception type).
+        */
+       cmpw    r6, r8
+       .if     \flags & NEED_EMU
+       lwz     r9, KVM_LPID(r9)
+       .endif
+       beq     1f
+       mfmsr   r7
+       .if     \srr0 != SPRN_MCSRR0 && \srr0 != SPRN_CSRR0
+       oris    r7, r7, MSR_CE@h
+       .endif
+       .if     \srr0 != SPRN_MCSRR0
+       ori     r7, r7, MSR_ME | MSR_RI
+       .endif
+       mtmsr   r7
+1:
+
+       .if     \flags & NEED_EMU
+       /*
+        * This assumes you have external PID support.
+        * To support a bookehv CPU without external PID, you'll
+        * need to look up the TLB entry and create a temporary mapping.
+        *
+        * FIXME: we don't currently handle if the lwepx faults.  PR-mode
+        * booke doesn't handle it either.  Since Linux doesn't use
+        * broadcast tlbivax anymore, the only way this should happen is
+        * if the guest maps its memory execute-but-not-read, or if we
+        * somehow take a TLB miss in the middle of this entry code and
+        * evict the relevant entry.  On e500mc, all kernel lowmem is
+        * bolted into TLB1 large page mappings, and we don't use
+        * broadcast invalidates, so we should not take a TLB miss here.
+        *
+        * Later we'll need to deal with faults here.  Disallowing guest
+        * mappings that are execute-but-not-read could be an option on
+        * e500mc, but not on chips with an LRAT if it is used.
+        */
+
+       mfspr   r3, SPRN_EPLC   /* will already have correct ELPID and EGS */
+       PPC_STL r15, VCPU_GPR(r15)(r4)
+       PPC_STL r16, VCPU_GPR(r16)(r4)
+       PPC_STL r17, VCPU_GPR(r17)(r4)
+       PPC_STL r18, VCPU_GPR(r18)(r4)
+       PPC_STL r19, VCPU_GPR(r19)(r4)
+       mr      r8, r3
+       PPC_STL r20, VCPU_GPR(r20)(r4)
+       rlwimi  r8, r6, EPC_EAS_SHIFT - MSR_IR_LG, EPC_EAS
+       PPC_STL r21, VCPU_GPR(r21)(r4)
+       rlwimi  r8, r6, EPC_EPR_SHIFT - MSR_PR_LG, EPC_EPR
+       PPC_STL r22, VCPU_GPR(r22)(r4)
+       rlwimi  r8, r10, EPC_EPID_SHIFT, EPC_EPID
+       PPC_STL r23, VCPU_GPR(r23)(r4)
+       PPC_STL r24, VCPU_GPR(r24)(r4)
+       PPC_STL r25, VCPU_GPR(r25)(r4)
+       PPC_STL r26, VCPU_GPR(r26)(r4)
+       PPC_STL r27, VCPU_GPR(r27)(r4)
+       PPC_STL r28, VCPU_GPR(r28)(r4)
+       PPC_STL r29, VCPU_GPR(r29)(r4)
+       PPC_STL r30, VCPU_GPR(r30)(r4)
+       PPC_STL r31, VCPU_GPR(r31)(r4)
+       mtspr   SPRN_EPLC, r8
+       isync
+       lwepx   r9, 0, r5
+       mtspr   SPRN_EPLC, r3
+       stw     r9, VCPU_LAST_INST(r4)
+       .endif
+
+       .if     \flags & NEED_ESR
+       mfspr   r8, SPRN_ESR
+       PPC_STL r8, VCPU_FAULT_ESR(r4)
+       .endif
+
+       .if     \flags & NEED_DEAR
+       mfspr   r9, SPRN_DEAR
+       PPC_STL r9, VCPU_FAULT_DEAR(r4)
+       .endif
+
+       b       kvmppc_resume_host
+.endm
+
+/*
+ * For input register values, see arch/powerpc/include/asm/kvm_booke_hv_asm.h
+ */
+.macro kvm_handler intno srr0, srr1, flags
+_GLOBAL(kvmppc_handler_\intno\()_\srr1)
+       GET_VCPU(r11, r10)
+       PPC_STL r3, VCPU_GPR(r3)(r11)
+       mfspr   r3, SPRN_SPRG_RSCRATCH0
+       PPC_STL r4, VCPU_GPR(r4)(r11)
+       PPC_LL  r4, THREAD_NORMSAVE(0)(r10)
+       PPC_STL r5, VCPU_GPR(r5)(r11)
+       PPC_STL r13, VCPU_CR(r11)
+       mfspr   r5, \srr0
+       PPC_STL r3, VCPU_GPR(r10)(r11)
+       PPC_LL  r3, THREAD_NORMSAVE(2)(r10)
+       PPC_STL r6, VCPU_GPR(r6)(r11)
+       PPC_STL r4, VCPU_GPR(r11)(r11)
+       mfspr   r6, \srr1
+       PPC_STL r7, VCPU_GPR(r7)(r11)
+       PPC_STL r8, VCPU_GPR(r8)(r11)
+       PPC_STL r9, VCPU_GPR(r9)(r11)
+       PPC_STL r3, VCPU_GPR(r13)(r11)
+       mfctr   r7
+       PPC_STL r12, VCPU_GPR(r12)(r11)
+       PPC_STL r7, VCPU_CTR(r11)
+       mr      r4, r11
+       kvm_handler_common \intno, \srr0, \flags
+.endm
+
+.macro kvm_lvl_handler intno scratch srr0, srr1, flags
+_GLOBAL(kvmppc_handler_\intno\()_\srr1)
+       mfspr   r10, SPRN_SPRG_THREAD
+       GET_VCPU(r11, r10)
+       PPC_STL r3, VCPU_GPR(r3)(r11)
+       mfspr   r3, \scratch
+       PPC_STL r4, VCPU_GPR(r4)(r11)
+       PPC_LL  r4, GPR9(r8)
+       PPC_STL r5, VCPU_GPR(r5)(r11)
+       PPC_STL r9, VCPU_CR(r11)
+       mfspr   r5, \srr0
+       PPC_STL r3, VCPU_GPR(r8)(r11)
+       PPC_LL  r3, GPR10(r8)
+       PPC_STL r6, VCPU_GPR(r6)(r11)
+       PPC_STL r4, VCPU_GPR(r9)(r11)
+       mfspr   r6, \srr1
+       PPC_LL  r4, GPR11(r8)
+       PPC_STL r7, VCPU_GPR(r7)(r11)
+       PPC_STL r8, VCPU_GPR(r8)(r11)
+       PPC_STL r3, VCPU_GPR(r10)(r11)
+       mfctr   r7
+       PPC_STL r12, VCPU_GPR(r12)(r11)
+       PPC_STL r4, VCPU_GPR(r11)(r11)
+       PPC_STL r7, VCPU_CTR(r11)
+       mr      r4, r11
+       kvm_handler_common \intno, \srr0, \flags
+.endm
+
+kvm_lvl_handler BOOKE_INTERRUPT_CRITICAL, \
+       SPRN_SPRG_RSCRATCH_CRIT, SPRN_CSRR0, SPRN_CSRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_MACHINE_CHECK, \
+       SPRN_SPRG_RSCRATCH_MC, SPRN_MCSRR0, SPRN_MCSRR1, 0
+kvm_handler BOOKE_INTERRUPT_DATA_STORAGE, \
+       SPRN_SRR0, SPRN_SRR1, (NEED_EMU | NEED_DEAR)
+kvm_handler BOOKE_INTERRUPT_INST_STORAGE, SPRN_SRR0, SPRN_SRR1, NEED_ESR
+kvm_handler BOOKE_INTERRUPT_EXTERNAL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_ALIGNMENT, \
+       SPRN_SRR0, SPRN_SRR1, (NEED_DEAR | NEED_ESR)
+kvm_handler BOOKE_INTERRUPT_PROGRAM, SPRN_SRR0, SPRN_SRR1, NEED_ESR
+kvm_handler BOOKE_INTERRUPT_FP_UNAVAIL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_SYSCALL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_AP_UNAVAIL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_DECREMENTER, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_FIT, SPRN_SRR0, SPRN_SRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_WATCHDOG, \
+       SPRN_SPRG_RSCRATCH_CRIT, SPRN_CSRR0, SPRN_CSRR1, 0
+kvm_handler BOOKE_INTERRUPT_DTLB_MISS, \
+       SPRN_SRR0, SPRN_SRR1, (NEED_EMU | NEED_DEAR | NEED_ESR)
+kvm_handler BOOKE_INTERRUPT_ITLB_MISS, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_SPE_UNAVAIL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_SPE_FP_DATA, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_SPE_FP_ROUND, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_PERFORMANCE_MONITOR, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_DOORBELL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_DOORBELL_CRITICAL, \
+       SPRN_SPRG_RSCRATCH_CRIT, SPRN_CSRR0, SPRN_CSRR1, 0
+kvm_handler BOOKE_INTERRUPT_HV_PRIV, SPRN_SRR0, SPRN_SRR1, NEED_EMU
+kvm_handler BOOKE_INTERRUPT_HV_SYSCALL, SPRN_SRR0, SPRN_SRR1, 0
+kvm_handler BOOKE_INTERRUPT_GUEST_DBELL, SPRN_GSRR0, SPRN_GSRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_GUEST_DBELL_CRIT, \
+       SPRN_SPRG_RSCRATCH_CRIT, SPRN_CSRR0, SPRN_CSRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_DEBUG, \
+       SPRN_SPRG_RSCRATCH_CRIT, SPRN_CSRR0, SPRN_CSRR1, 0
+kvm_lvl_handler BOOKE_INTERRUPT_DEBUG, \
+       SPRN_SPRG_RSCRATCH_DBG, SPRN_DSRR0, SPRN_DSRR1, 0
+
+
+/* Registers:
+ *  SPRG_SCRATCH0: guest r10
+ *  r4: vcpu pointer
+ *  r11: vcpu->arch.shared
+ *  r14: KVM exit number
+ */
+_GLOBAL(kvmppc_resume_host)
+       /* Save remaining volatile guest register state to vcpu. */
+       mfspr   r3, SPRN_VRSAVE
+       PPC_STL r0, VCPU_GPR(r0)(r4)
+       PPC_STL r1, VCPU_GPR(r1)(r4)
+       mflr    r5
+       mfspr   r6, SPRN_SPRG4
+       PPC_STL r2, VCPU_GPR(r2)(r4)
+       PPC_STL r5, VCPU_LR(r4)
+       mfspr   r7, SPRN_SPRG5
+       PPC_STL r3, VCPU_VRSAVE(r4)
+       PPC_STL r6, VCPU_SHARED_SPRG4(r11)
+       mfspr   r8, SPRN_SPRG6
+       PPC_STL r7, VCPU_SHARED_SPRG5(r11)
+       mfspr   r9, SPRN_SPRG7
+       PPC_STL r8, VCPU_SHARED_SPRG6(r11)
+       mfxer   r3
+       PPC_STL r9, VCPU_SHARED_SPRG7(r11)
+
+       /* save guest MAS registers and restore host mas4 & mas6 */
+       mfspr   r5, SPRN_MAS0
+       PPC_STL r3, VCPU_XER(r4)
+       mfspr   r6, SPRN_MAS1
+       stw     r5, VCPU_SHARED_MAS0(r11)
+       mfspr   r7, SPRN_MAS2
+       stw     r6, VCPU_SHARED_MAS1(r11)
+#ifndef CONFIG_64BIT
+       stw     r7, (VCPU_SHARED_MAS2 + 4)(r11)
+#else
+       std     r7, (VCPU_SHARED_MAS2)(r11)
+#endif
+       mfspr   r5, SPRN_MAS3
+       mfspr   r6, SPRN_MAS4
+       stw     r5, VCPU_SHARED_MAS7_3+4(r11)
+       mfspr   r7, SPRN_MAS6
+       stw     r6, VCPU_SHARED_MAS4(r11)
+       mfspr   r5, SPRN_MAS7
+       lwz     r6, VCPU_HOST_MAS4(r4)
+       stw     r7, VCPU_SHARED_MAS6(r11)
+       lwz     r8, VCPU_HOST_MAS6(r4)
+       mtspr   SPRN_MAS4, r6
+       stw     r5, VCPU_SHARED_MAS7_3+0(r11)
+       mtspr   SPRN_MAS6, r8
+       mfspr   r3, SPRN_EPCR
+       rlwinm  r3, r3, 0, ~SPRN_EPCR_DMIUH
+       mtspr   SPRN_EPCR, r3
+       isync
+
+       /* Restore host stack pointer */
+       PPC_LL  r1, VCPU_HOST_STACK(r4)
+       PPC_LL  r2, HOST_R2(r1)
+
+       /* Switch to kernel stack and jump to handler. */
+       PPC_LL  r3, HOST_RUN(r1)
+       mr      r5, r14 /* intno */
+       mr      r14, r4 /* Save vcpu pointer. */
+       bl      kvmppc_handle_exit
+
+       /* Restore vcpu pointer and the nonvolatiles we used. */
+       mr      r4, r14
+       PPC_LL  r14, VCPU_GPR(r14)(r4)
+
+       andi.   r5, r3, RESUME_FLAG_NV
+       beq     skip_nv_load
+       PPC_LL  r15, VCPU_GPR(r15)(r4)
+       PPC_LL  r16, VCPU_GPR(r16)(r4)
+       PPC_LL  r17, VCPU_GPR(r17)(r4)
+       PPC_LL  r18, VCPU_GPR(r18)(r4)
+       PPC_LL  r19, VCPU_GPR(r19)(r4)
+       PPC_LL  r20, VCPU_GPR(r20)(r4)
+       PPC_LL  r21, VCPU_GPR(r21)(r4)
+       PPC_LL  r22, VCPU_GPR(r22)(r4)
+       PPC_LL  r23, VCPU_GPR(r23)(r4)
+       PPC_LL  r24, VCPU_GPR(r24)(r4)
+       PPC_LL  r25, VCPU_GPR(r25)(r4)
+       PPC_LL  r26, VCPU_GPR(r26)(r4)
+       PPC_LL  r27, VCPU_GPR(r27)(r4)
+       PPC_LL  r28, VCPU_GPR(r28)(r4)
+       PPC_LL  r29, VCPU_GPR(r29)(r4)
+       PPC_LL  r30, VCPU_GPR(r30)(r4)
+       PPC_LL  r31, VCPU_GPR(r31)(r4)
+skip_nv_load:
+       /* Should we return to the guest? */
+       andi.   r5, r3, RESUME_FLAG_HOST
+       beq     lightweight_exit
+
+       srawi   r3, r3, 2 /* Shift -ERR back down. */
+
+heavyweight_exit:
+       /* Not returning to guest. */
+       PPC_LL  r5, HOST_STACK_LR(r1)
+
+       /*
+        * We already saved guest volatile register state; now save the
+        * non-volatiles.
+        */
+
+       PPC_STL r15, VCPU_GPR(r15)(r4)
+       PPC_STL r16, VCPU_GPR(r16)(r4)
+       PPC_STL r17, VCPU_GPR(r17)(r4)
+       PPC_STL r18, VCPU_GPR(r18)(r4)
+       PPC_STL r19, VCPU_GPR(r19)(r4)
+       PPC_STL r20, VCPU_GPR(r20)(r4)
+       PPC_STL r21, VCPU_GPR(r21)(r4)
+       PPC_STL r22, VCPU_GPR(r22)(r4)
+       PPC_STL r23, VCPU_GPR(r23)(r4)
+       PPC_STL r24, VCPU_GPR(r24)(r4)
+       PPC_STL r25, VCPU_GPR(r25)(r4)
+       PPC_STL r26, VCPU_GPR(r26)(r4)
+       PPC_STL r27, VCPU_GPR(r27)(r4)
+       PPC_STL r28, VCPU_GPR(r28)(r4)
+       PPC_STL r29, VCPU_GPR(r29)(r4)
+       PPC_STL r30, VCPU_GPR(r30)(r4)
+       PPC_STL r31, VCPU_GPR(r31)(r4)
+
+       /* Load host non-volatile register state from host stack. */
+       PPC_LL  r14, HOST_NV_GPR(r14)(r1)
+       PPC_LL  r15, HOST_NV_GPR(r15)(r1)
+       PPC_LL  r16, HOST_NV_GPR(r16)(r1)
+       PPC_LL  r17, HOST_NV_GPR(r17)(r1)
+       PPC_LL  r18, HOST_NV_GPR(r18)(r1)
+       PPC_LL  r19, HOST_NV_GPR(r19)(r1)
+       PPC_LL  r20, HOST_NV_GPR(r20)(r1)
+       PPC_LL  r21, HOST_NV_GPR(r21)(r1)
+       PPC_LL  r22, HOST_NV_GPR(r22)(r1)
+       PPC_LL  r23, HOST_NV_GPR(r23)(r1)
+       PPC_LL  r24, HOST_NV_GPR(r24)(r1)
+       PPC_LL  r25, HOST_NV_GPR(r25)(r1)
+       PPC_LL  r26, HOST_NV_GPR(r26)(r1)
+       PPC_LL  r27, HOST_NV_GPR(r27)(r1)
+       PPC_LL  r28, HOST_NV_GPR(r28)(r1)
+       PPC_LL  r29, HOST_NV_GPR(r29)(r1)
+       PPC_LL  r30, HOST_NV_GPR(r30)(r1)
+       PPC_LL  r31, HOST_NV_GPR(r31)(r1)
+
+       /* Return to kvm_vcpu_run(). */
+       mtlr    r5
+       addi    r1, r1, HOST_STACK_SIZE
+       /* r3 still contains the return code from kvmppc_handle_exit(). */
+       blr
+
+/* Registers:
+ *  r3: kvm_run pointer
+ *  r4: vcpu pointer
+ */
+_GLOBAL(__kvmppc_vcpu_run)
+       stwu    r1, -HOST_STACK_SIZE(r1)
+       PPC_STL r1, VCPU_HOST_STACK(r4) /* Save stack pointer to vcpu. */
+
+       /* Save host state to stack. */
+       PPC_STL r3, HOST_RUN(r1)
+       mflr    r3
+       PPC_STL r3, HOST_STACK_LR(r1)
+
+       /* Save host non-volatile register state to stack. */
+       PPC_STL r14, HOST_NV_GPR(r14)(r1)
+       PPC_STL r15, HOST_NV_GPR(r15)(r1)
+       PPC_STL r16, HOST_NV_GPR(r16)(r1)
+       PPC_STL r17, HOST_NV_GPR(r17)(r1)
+       PPC_STL r18, HOST_NV_GPR(r18)(r1)
+       PPC_STL r19, HOST_NV_GPR(r19)(r1)
+       PPC_STL r20, HOST_NV_GPR(r20)(r1)
+       PPC_STL r21, HOST_NV_GPR(r21)(r1)
+       PPC_STL r22, HOST_NV_GPR(r22)(r1)
+       PPC_STL r23, HOST_NV_GPR(r23)(r1)
+       PPC_STL r24, HOST_NV_GPR(r24)(r1)
+       PPC_STL r25, HOST_NV_GPR(r25)(r1)
+       PPC_STL r26, HOST_NV_GPR(r26)(r1)
+       PPC_STL r27, HOST_NV_GPR(r27)(r1)
+       PPC_STL r28, HOST_NV_GPR(r28)(r1)
+       PPC_STL r29, HOST_NV_GPR(r29)(r1)
+       PPC_STL r30, HOST_NV_GPR(r30)(r1)
+       PPC_STL r31, HOST_NV_GPR(r31)(r1)
+
+       /* Load guest non-volatiles. */
+       PPC_LL  r14, VCPU_GPR(r14)(r4)
+       PPC_LL  r15, VCPU_GPR(r15)(r4)
+       PPC_LL  r16, VCPU_GPR(r16)(r4)
+       PPC_LL  r17, VCPU_GPR(r17)(r4)
+       PPC_LL  r18, VCPU_GPR(r18)(r4)
+       PPC_LL  r19, VCPU_GPR(r19)(r4)
+       PPC_LL  r20, VCPU_GPR(r20)(r4)
+       PPC_LL  r21, VCPU_GPR(r21)(r4)
+       PPC_LL  r22, VCPU_GPR(r22)(r4)
+       PPC_LL  r23, VCPU_GPR(r23)(r4)
+       PPC_LL  r24, VCPU_GPR(r24)(r4)
+       PPC_LL  r25, VCPU_GPR(r25)(r4)
+       PPC_LL  r26, VCPU_GPR(r26)(r4)
+       PPC_LL  r27, VCPU_GPR(r27)(r4)
+       PPC_LL  r28, VCPU_GPR(r28)(r4)
+       PPC_LL  r29, VCPU_GPR(r29)(r4)
+       PPC_LL  r30, VCPU_GPR(r30)(r4)
+       PPC_LL  r31, VCPU_GPR(r31)(r4)
+
+
+lightweight_exit:
+       PPC_STL r2, HOST_R2(r1)
+
+       mfspr   r3, SPRN_PID
+       stw     r3, VCPU_HOST_PID(r4)
+       lwz     r3, VCPU_GUEST_PID(r4)
+       mtspr   SPRN_PID, r3
+
+       /* Save vcpu pointer for the exception handlers
+        * must be done before loading guest r2.
+        */
+//     SET_VCPU(r4)
+
+       PPC_LL  r11, VCPU_SHARED(r4)
+       /* Save host mas4 and mas6 and load guest MAS registers */
+       mfspr   r3, SPRN_MAS4
+       stw     r3, VCPU_HOST_MAS4(r4)
+       mfspr   r3, SPRN_MAS6
+       stw     r3, VCPU_HOST_MAS6(r4)
+       lwz     r3, VCPU_SHARED_MAS0(r11)
+       lwz     r5, VCPU_SHARED_MAS1(r11)
+#ifndef CONFIG_64BIT
+       lwz     r6, (VCPU_SHARED_MAS2 + 4)(r11)
+#else
+       ld      r6, (VCPU_SHARED_MAS2)(r11)
+#endif
+       lwz     r7, VCPU_SHARED_MAS7_3+4(r11)
+       lwz     r8, VCPU_SHARED_MAS4(r11)
+       mtspr   SPRN_MAS0, r3
+       mtspr   SPRN_MAS1, r5
+       mtspr   SPRN_MAS2, r6
+       mtspr   SPRN_MAS3, r7
+       mtspr   SPRN_MAS4, r8
+       lwz     r3, VCPU_SHARED_MAS6(r11)
+       lwz     r5, VCPU_SHARED_MAS7_3+0(r11)
+       mtspr   SPRN_MAS6, r3
+       mtspr   SPRN_MAS7, r5
+       /* Disable MAS register updates via exception */
+       mfspr   r3, SPRN_EPCR
+       oris    r3, r3, SPRN_EPCR_DMIUH@h
+       mtspr   SPRN_EPCR, r3
+
+       /*
+        * Host interrupt handlers may have clobbered these guest-readable
+        * SPRGs, so we need to reload them here with the guest's values.
+        */
+       lwz     r3, VCPU_VRSAVE(r4)
+       lwz     r5, VCPU_SHARED_SPRG4(r11)
+       mtspr   SPRN_VRSAVE, r3
+       lwz     r6, VCPU_SHARED_SPRG5(r11)
+       mtspr   SPRN_SPRG4W, r5
+       lwz     r7, VCPU_SHARED_SPRG6(r11)
+       mtspr   SPRN_SPRG5W, r6
+       lwz     r8, VCPU_SHARED_SPRG7(r11)
+       mtspr   SPRN_SPRG6W, r7
+       mtspr   SPRN_SPRG7W, r8
+
+       /* Load some guest volatiles. */
+       PPC_LL  r3, VCPU_LR(r4)
+       PPC_LL  r5, VCPU_XER(r4)
+       PPC_LL  r6, VCPU_CTR(r4)
+       PPC_LL  r7, VCPU_CR(r4)
+       PPC_LL  r8, VCPU_PC(r4)
+#ifndef CONFIG_64BIT
+       lwz     r9, (VCPU_SHARED_MSR + 4)(r11)
+#else
+       ld      r9, (VCPU_SHARED_MSR)(r11)
+#endif
+       PPC_LL  r0, VCPU_GPR(r0)(r4)
+       PPC_LL  r1, VCPU_GPR(r1)(r4)
+       PPC_LL  r2, VCPU_GPR(r2)(r4)
+       PPC_LL  r10, VCPU_GPR(r10)(r4)
+       PPC_LL  r11, VCPU_GPR(r11)(r4)
+       PPC_LL  r12, VCPU_GPR(r12)(r4)
+       PPC_LL  r13, VCPU_GPR(r13)(r4)
+       mtlr    r3
+       mtxer   r5
+       mtctr   r6
+       mtcr    r7
+       mtsrr0  r8
+       mtsrr1  r9
+
+#ifdef CONFIG_KVM_EXIT_TIMING
+       /* save enter time */
+1:
+       mfspr   r6, SPRN_TBRU
+       mfspr   r7, SPRN_TBRL
+       mfspr   r8, SPRN_TBRU
+       cmpw    r8, r6
+       PPC_STL r7, VCPU_TIMING_LAST_ENTER_TBL(r4)
+       bne     1b
+       PPC_STL r8, VCPU_TIMING_LAST_ENTER_TBU(r4)
+#endif
+
+       /* Finish loading guest volatiles and jump to guest. */
+       PPC_LL  r5, VCPU_GPR(r5)(r4)
+       PPC_LL  r6, VCPU_GPR(r6)(r4)
+       PPC_LL  r7, VCPU_GPR(r7)(r4)
+       PPC_LL  r8, VCPU_GPR(r8)(r4)
+       PPC_LL  r9, VCPU_GPR(r9)(r4)
+
+       PPC_LL  r3, VCPU_GPR(r3)(r4)
+       PPC_LL  r4, VCPU_GPR(r4)(r4)
+       rfi
index cd53e08..6a530e4 100644 (file)
@@ -114,6 +114,11 @@ int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
                goto out;
 #endif
 
+#ifdef CONFIG_KVM_BOOKE_HV
+       if (!cpu_has_feature(CPU_FTR_EMB_HV))
+               goto out;
+#endif
+
        r = true;
 
 out:
index 8167d42..bf191e7 100644 (file)
@@ -93,6 +93,12 @@ static inline void kvmppc_account_exit_stat(struct kvm_vcpu *vcpu, int type)
        case SIGNAL_EXITS:
                vcpu->stat.signal_exits++;
                break;
+       case DBELL_EXITS:
+               vcpu->stat.dbell_exits++;
+               break;
+       case GDBELL_EXITS:
+               vcpu->stat.gdbell_exits++;
+               break;
        }
 }