aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2011-10-30 18:36:45 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2011-10-30 18:36:45 -0400
commit1bc87b00556e8f7ba30a1010471951c5b8f71114 (patch)
treee73c2d187e2dff0df97ed82e32b45e362b923117 /arch/powerpc
parentacff987d94cbdb4049f3706bed1f1792f8ef6837 (diff)
parentf1c1da2bde712812a3e0f9a7a7ebe7a916a4b5f4 (diff)
Merge branch 'kvm-updates/3.2' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm
* 'kvm-updates/3.2' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm: (75 commits) KVM: SVM: Keep intercepting task switching with NPT enabled KVM: s390: implement sigp external call KVM: s390: fix register setting KVM: s390: fix return value of kvm_arch_init_vm KVM: s390: check cpu_id prior to using it KVM: emulate lapic tsc deadline timer for guest x86: TSC deadline definitions KVM: Fix simultaneous NMIs KVM: x86 emulator: convert push %sreg/pop %sreg to direct decode KVM: x86 emulator: switch lds/les/lss/lfs/lgs to direct decode KVM: x86 emulator: streamline decode of segment registers KVM: x86 emulator: simplify OpMem64 decode KVM: x86 emulator: switch src decode to decode_operand() KVM: x86 emulator: qualify OpReg inhibit_byte_regs hack KVM: x86 emulator: switch OpImmUByte decode to decode_imm() KVM: x86 emulator: free up some flag bits near src, dst KVM: x86 emulator: switch src2 to generic decode_operand() KVM: x86 emulator: expand decode flags to 64 bits KVM: x86 emulator: split dst decode to a generic decode_operand() KVM: x86 emulator: move memop, memopp into emulation context ...
Diffstat (limited to 'arch/powerpc')
-rw-r--r--arch/powerpc/include/asm/kvm.h13
-rw-r--r--arch/powerpc/include/asm/kvm_book3s.h40
-rw-r--r--arch/powerpc/include/asm/kvm_book3s_asm.h2
-rw-r--r--arch/powerpc/include/asm/kvm_host.h30
-rw-r--r--arch/powerpc/include/asm/kvm_ppc.h1
-rw-r--r--arch/powerpc/kernel/asm-offsets.c13
-rw-r--r--arch/powerpc/kernel/exceptions-64s.S10
-rw-r--r--arch/powerpc/kvm/44x.c2
-rw-r--r--arch/powerpc/kvm/Makefile4
-rw-r--r--arch/powerpc/kvm/book3s_32_sr.S2
-rw-r--r--arch/powerpc/kvm/book3s_64_mmu.c8
-rw-r--r--arch/powerpc/kvm/book3s_64_slb.S2
-rw-r--r--arch/powerpc/kvm/book3s_emulate.c29
-rw-r--r--arch/powerpc/kvm/book3s_exports.c4
-rw-r--r--arch/powerpc/kvm/book3s_hv.c343
-rw-r--r--arch/powerpc/kvm/book3s_hv_rm_mmu.c33
-rw-r--r--arch/powerpc/kvm/book3s_hv_rmhandlers.S300
-rw-r--r--arch/powerpc/kvm/book3s_interrupts.S129
-rw-r--r--arch/powerpc/kvm/book3s_pr.c58
-rw-r--r--arch/powerpc/kvm/book3s_pr_papr.c158
-rw-r--r--arch/powerpc/kvm/book3s_rmhandlers.S54
-rw-r--r--arch/powerpc/kvm/book3s_segment.S117
-rw-r--r--arch/powerpc/kvm/booke.c10
-rw-r--r--arch/powerpc/kvm/e500.c2
-rw-r--r--arch/powerpc/kvm/powerpc.c55
25 files changed, 963 insertions, 456 deletions
diff --git a/arch/powerpc/include/asm/kvm.h b/arch/powerpc/include/asm/kvm.h
index a4f6c85431f8..08fe69edcd10 100644
--- a/arch/powerpc/include/asm/kvm.h
+++ b/arch/powerpc/include/asm/kvm.h
@@ -149,6 +149,12 @@ struct kvm_regs {
149#define KVM_SREGS_E_UPDATE_DBSR (1 << 3) 149#define KVM_SREGS_E_UPDATE_DBSR (1 << 3)
150 150
151/* 151/*
152 * Book3S special bits to indicate contents in the struct by maintaining
153 * backwards compatibility with older structs. If adding a new field,
154 * please make sure to add a flag for that new field */
155#define KVM_SREGS_S_HIOR (1 << 0)
156
157/*
152 * In KVM_SET_SREGS, reserved/pad fields must be left untouched from a 158 * In KVM_SET_SREGS, reserved/pad fields must be left untouched from a
153 * previous KVM_GET_REGS. 159 * previous KVM_GET_REGS.
154 * 160 *
@@ -173,6 +179,8 @@ struct kvm_sregs {
173 __u64 ibat[8]; 179 __u64 ibat[8];
174 __u64 dbat[8]; 180 __u64 dbat[8];
175 } ppc32; 181 } ppc32;
182 __u64 flags; /* KVM_SREGS_S_ */
183 __u64 hior;
176 } s; 184 } s;
177 struct { 185 struct {
178 union { 186 union {
@@ -276,6 +284,11 @@ struct kvm_guest_debug_arch {
276#define KVM_INTERRUPT_UNSET -2U 284#define KVM_INTERRUPT_UNSET -2U
277#define KVM_INTERRUPT_SET_LEVEL -3U 285#define KVM_INTERRUPT_SET_LEVEL -3U
278 286
287#define KVM_CPU_440 1
288#define KVM_CPU_E500V2 2
289#define KVM_CPU_3S_32 3
290#define KVM_CPU_3S_64 4
291
279/* for KVM_CAP_SPAPR_TCE */ 292/* for KVM_CAP_SPAPR_TCE */
280struct kvm_create_spapr_tce { 293struct kvm_create_spapr_tce {
281 __u64 liobn; 294 __u64 liobn;
diff --git a/arch/powerpc/include/asm/kvm_book3s.h b/arch/powerpc/include/asm/kvm_book3s.h
index 98da010252a3..a384ffdf33de 100644
--- a/arch/powerpc/include/asm/kvm_book3s.h
+++ b/arch/powerpc/include/asm/kvm_book3s.h
@@ -90,6 +90,8 @@ struct kvmppc_vcpu_book3s {
90#endif 90#endif
91 int context_id[SID_CONTEXTS]; 91 int context_id[SID_CONTEXTS];
92 92
93 bool hior_sregs; /* HIOR is set by SREGS, not PVR */
94
93 struct hlist_head hpte_hash_pte[HPTEG_HASH_NUM_PTE]; 95 struct hlist_head hpte_hash_pte[HPTEG_HASH_NUM_PTE];
94 struct hlist_head hpte_hash_pte_long[HPTEG_HASH_NUM_PTE_LONG]; 96 struct hlist_head hpte_hash_pte_long[HPTEG_HASH_NUM_PTE_LONG];
95 struct hlist_head hpte_hash_vpte[HPTEG_HASH_NUM_VPTE]; 97 struct hlist_head hpte_hash_vpte[HPTEG_HASH_NUM_VPTE];
@@ -139,15 +141,14 @@ extern void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr);
139extern int kvmppc_emulate_paired_single(struct kvm_run *run, struct kvm_vcpu *vcpu); 141extern int kvmppc_emulate_paired_single(struct kvm_run *run, struct kvm_vcpu *vcpu);
140extern pfn_t kvmppc_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn); 142extern pfn_t kvmppc_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
141 143
142extern void kvmppc_handler_lowmem_trampoline(void); 144extern void kvmppc_entry_trampoline(void);
143extern void kvmppc_handler_trampoline_enter(void);
144extern void kvmppc_rmcall(ulong srr0, ulong srr1);
145extern void kvmppc_hv_entry_trampoline(void); 145extern void kvmppc_hv_entry_trampoline(void);
146extern void kvmppc_load_up_fpu(void); 146extern void kvmppc_load_up_fpu(void);
147extern void kvmppc_load_up_altivec(void); 147extern void kvmppc_load_up_altivec(void);
148extern void kvmppc_load_up_vsx(void); 148extern void kvmppc_load_up_vsx(void);
149extern u32 kvmppc_alignment_dsisr(struct kvm_vcpu *vcpu, unsigned int inst); 149extern u32 kvmppc_alignment_dsisr(struct kvm_vcpu *vcpu, unsigned int inst);
150extern ulong kvmppc_alignment_dar(struct kvm_vcpu *vcpu, unsigned int inst); 150extern ulong kvmppc_alignment_dar(struct kvm_vcpu *vcpu, unsigned int inst);
151extern int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd);
151 152
152static inline struct kvmppc_vcpu_book3s *to_book3s(struct kvm_vcpu *vcpu) 153static inline struct kvmppc_vcpu_book3s *to_book3s(struct kvm_vcpu *vcpu)
153{ 154{
@@ -382,6 +383,39 @@ static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
382} 383}
383#endif 384#endif
384 385
386static inline unsigned long compute_tlbie_rb(unsigned long v, unsigned long r,
387 unsigned long pte_index)
388{
389 unsigned long rb, va_low;
390
391 rb = (v & ~0x7fUL) << 16; /* AVA field */
392 va_low = pte_index >> 3;
393 if (v & HPTE_V_SECONDARY)
394 va_low = ~va_low;
395 /* xor vsid from AVA */
396 if (!(v & HPTE_V_1TB_SEG))
397 va_low ^= v >> 12;
398 else
399 va_low ^= v >> 24;
400 va_low &= 0x7ff;
401 if (v & HPTE_V_LARGE) {
402 rb |= 1; /* L field */
403 if (cpu_has_feature(CPU_FTR_ARCH_206) &&
404 (r & 0xff000)) {
405 /* non-16MB large page, must be 64k */
406 /* (masks depend on page size) */
407 rb |= 0x1000; /* page encoding in LP field */
408 rb |= (va_low & 0x7f) << 16; /* 7b of VA in AVA/LP field */
409 rb |= (va_low & 0xfe); /* AVAL field (P7 doesn't seem to care) */
410 }
411 } else {
412 /* 4kB page */
413 rb |= (va_low & 0x7ff) << 12; /* remaining 11b of VA */
414 }
415 rb |= (v >> 54) & 0x300; /* B field */
416 return rb;
417}
418
385/* Magic register values loaded into r3 and r4 before the 'sc' assembly 419/* Magic register values loaded into r3 and r4 before the 'sc' assembly
386 * instruction for the OSI hypercalls */ 420 * instruction for the OSI hypercalls */
387#define OSI_SC_MAGIC_R3 0x113724FA 421#define OSI_SC_MAGIC_R3 0x113724FA
diff --git a/arch/powerpc/include/asm/kvm_book3s_asm.h b/arch/powerpc/include/asm/kvm_book3s_asm.h
index ef7b3688c3b6..1f2f5b6156bd 100644
--- a/arch/powerpc/include/asm/kvm_book3s_asm.h
+++ b/arch/powerpc/include/asm/kvm_book3s_asm.h
@@ -75,6 +75,8 @@ struct kvmppc_host_state {
75 ulong scratch0; 75 ulong scratch0;
76 ulong scratch1; 76 ulong scratch1;
77 u8 in_guest; 77 u8 in_guest;
78 u8 restore_hid5;
79 u8 napping;
78 80
79#ifdef CONFIG_KVM_BOOK3S_64_HV 81#ifdef CONFIG_KVM_BOOK3S_64_HV
80 struct kvm_vcpu *kvm_vcpu; 82 struct kvm_vcpu *kvm_vcpu;
diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
index cc22b282d755..bf8af5d5d5dc 100644
--- a/arch/powerpc/include/asm/kvm_host.h
+++ b/arch/powerpc/include/asm/kvm_host.h
@@ -198,21 +198,29 @@ struct kvm_arch {
198 */ 198 */
199struct kvmppc_vcore { 199struct kvmppc_vcore {
200 int n_runnable; 200 int n_runnable;
201 int n_blocked; 201 int n_busy;
202 int num_threads; 202 int num_threads;
203 int entry_exit_count; 203 int entry_exit_count;
204 int n_woken; 204 int n_woken;
205 int nap_count; 205 int nap_count;
206 int napping_threads;
206 u16 pcpu; 207 u16 pcpu;
207 u8 vcore_running; 208 u8 vcore_state;
208 u8 in_guest; 209 u8 in_guest;
209 struct list_head runnable_threads; 210 struct list_head runnable_threads;
210 spinlock_t lock; 211 spinlock_t lock;
212 wait_queue_head_t wq;
211}; 213};
212 214
213#define VCORE_ENTRY_COUNT(vc) ((vc)->entry_exit_count & 0xff) 215#define VCORE_ENTRY_COUNT(vc) ((vc)->entry_exit_count & 0xff)
214#define VCORE_EXIT_COUNT(vc) ((vc)->entry_exit_count >> 8) 216#define VCORE_EXIT_COUNT(vc) ((vc)->entry_exit_count >> 8)
215 217
218/* Values for vcore_state */
219#define VCORE_INACTIVE 0
220#define VCORE_RUNNING 1
221#define VCORE_EXITING 2
222#define VCORE_SLEEPING 3
223
216struct kvmppc_pte { 224struct kvmppc_pte {
217 ulong eaddr; 225 ulong eaddr;
218 u64 vpage; 226 u64 vpage;
@@ -258,14 +266,6 @@ struct kvm_vcpu_arch {
258 ulong host_stack; 266 ulong host_stack;
259 u32 host_pid; 267 u32 host_pid;
260#ifdef CONFIG_PPC_BOOK3S 268#ifdef CONFIG_PPC_BOOK3S
261 ulong host_msr;
262 ulong host_r2;
263 void *host_retip;
264 ulong trampoline_lowmem;
265 ulong trampoline_enter;
266 ulong highmem_handler;
267 ulong rmcall;
268 ulong host_paca_phys;
269 struct kvmppc_slb slb[64]; 269 struct kvmppc_slb slb[64];
270 int slb_max; /* 1 + index of last valid entry in slb[] */ 270 int slb_max; /* 1 + index of last valid entry in slb[] */
271 int slb_nr; /* total number of entries in SLB */ 271 int slb_nr; /* total number of entries in SLB */
@@ -389,6 +389,9 @@ struct kvm_vcpu_arch {
389 u8 dcr_is_write; 389 u8 dcr_is_write;
390 u8 osi_needed; 390 u8 osi_needed;
391 u8 osi_enabled; 391 u8 osi_enabled;
392 u8 papr_enabled;
393 u8 sane;
394 u8 cpu_type;
392 u8 hcall_needed; 395 u8 hcall_needed;
393 396
394 u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */ 397 u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */
@@ -408,11 +411,13 @@ struct kvm_vcpu_arch {
408 struct dtl *dtl; 411 struct dtl *dtl;
409 struct dtl *dtl_end; 412 struct dtl *dtl_end;
410 413
414 wait_queue_head_t *wqp;
411 struct kvmppc_vcore *vcore; 415 struct kvmppc_vcore *vcore;
412 int ret; 416 int ret;
413 int trap; 417 int trap;
414 int state; 418 int state;
415 int ptid; 419 int ptid;
420 bool timer_running;
416 wait_queue_head_t cpu_run; 421 wait_queue_head_t cpu_run;
417 422
418 struct kvm_vcpu_arch_shared *shared; 423 struct kvm_vcpu_arch_shared *shared;
@@ -428,8 +433,9 @@ struct kvm_vcpu_arch {
428#endif 433#endif
429}; 434};
430 435
431#define KVMPPC_VCPU_BUSY_IN_HOST 0 436/* Values for vcpu->arch.state */
432#define KVMPPC_VCPU_BLOCKED 1 437#define KVMPPC_VCPU_STOPPED 0
438#define KVMPPC_VCPU_BUSY_IN_HOST 1
433#define KVMPPC_VCPU_RUNNABLE 2 439#define KVMPPC_VCPU_RUNNABLE 2
434 440
435#endif /* __POWERPC_KVM_HOST_H__ */ 441#endif /* __POWERPC_KVM_HOST_H__ */
diff --git a/arch/powerpc/include/asm/kvm_ppc.h b/arch/powerpc/include/asm/kvm_ppc.h
index d121f49d62b8..46efd1a265c9 100644
--- a/arch/powerpc/include/asm/kvm_ppc.h
+++ b/arch/powerpc/include/asm/kvm_ppc.h
@@ -66,6 +66,7 @@ extern int kvmppc_emulate_instruction(struct kvm_run *run,
66extern int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu); 66extern int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu);
67extern void kvmppc_emulate_dec(struct kvm_vcpu *vcpu); 67extern void kvmppc_emulate_dec(struct kvm_vcpu *vcpu);
68extern u32 kvmppc_get_dec(struct kvm_vcpu *vcpu, u64 tb); 68extern u32 kvmppc_get_dec(struct kvm_vcpu *vcpu, u64 tb);
69extern int kvmppc_sanity_check(struct kvm_vcpu *vcpu);
69 70
70/* Core-specific hooks */ 71/* Core-specific hooks */
71 72
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 5f078bc2063e..69f7ffe7f674 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -44,6 +44,7 @@
44#include <asm/compat.h> 44#include <asm/compat.h>
45#include <asm/mmu.h> 45#include <asm/mmu.h>
46#include <asm/hvcall.h> 46#include <asm/hvcall.h>
47#include <asm/xics.h>
47#endif 48#endif
48#ifdef CONFIG_PPC_ISERIES 49#ifdef CONFIG_PPC_ISERIES
49#include <asm/iseries/alpaca.h> 50#include <asm/iseries/alpaca.h>
@@ -449,8 +450,6 @@ int main(void)
449#ifdef CONFIG_PPC_BOOK3S 450#ifdef CONFIG_PPC_BOOK3S
450 DEFINE(VCPU_KVM, offsetof(struct kvm_vcpu, kvm)); 451 DEFINE(VCPU_KVM, offsetof(struct kvm_vcpu, kvm));
451 DEFINE(VCPU_VCPUID, offsetof(struct kvm_vcpu, vcpu_id)); 452 DEFINE(VCPU_VCPUID, offsetof(struct kvm_vcpu, vcpu_id));
452 DEFINE(VCPU_HOST_RETIP, offsetof(struct kvm_vcpu, arch.host_retip));
453 DEFINE(VCPU_HOST_MSR, offsetof(struct kvm_vcpu, arch.host_msr));
454 DEFINE(VCPU_PURR, offsetof(struct kvm_vcpu, arch.purr)); 453 DEFINE(VCPU_PURR, offsetof(struct kvm_vcpu, arch.purr));
455 DEFINE(VCPU_SPURR, offsetof(struct kvm_vcpu, arch.spurr)); 454 DEFINE(VCPU_SPURR, offsetof(struct kvm_vcpu, arch.spurr));
456 DEFINE(VCPU_DSCR, offsetof(struct kvm_vcpu, arch.dscr)); 455 DEFINE(VCPU_DSCR, offsetof(struct kvm_vcpu, arch.dscr));
@@ -458,14 +457,12 @@ int main(void)
458 DEFINE(VCPU_UAMOR, offsetof(struct kvm_vcpu, arch.uamor)); 457 DEFINE(VCPU_UAMOR, offsetof(struct kvm_vcpu, arch.uamor));
459 DEFINE(VCPU_CTRL, offsetof(struct kvm_vcpu, arch.ctrl)); 458 DEFINE(VCPU_CTRL, offsetof(struct kvm_vcpu, arch.ctrl));
460 DEFINE(VCPU_DABR, offsetof(struct kvm_vcpu, arch.dabr)); 459 DEFINE(VCPU_DABR, offsetof(struct kvm_vcpu, arch.dabr));
461 DEFINE(VCPU_TRAMPOLINE_LOWMEM, offsetof(struct kvm_vcpu, arch.trampoline_lowmem));
462 DEFINE(VCPU_TRAMPOLINE_ENTER, offsetof(struct kvm_vcpu, arch.trampoline_enter));
463 DEFINE(VCPU_HIGHMEM_HANDLER, offsetof(struct kvm_vcpu, arch.highmem_handler));
464 DEFINE(VCPU_RMCALL, offsetof(struct kvm_vcpu, arch.rmcall));
465 DEFINE(VCPU_HFLAGS, offsetof(struct kvm_vcpu, arch.hflags)); 460 DEFINE(VCPU_HFLAGS, offsetof(struct kvm_vcpu, arch.hflags));
466 DEFINE(VCPU_DEC, offsetof(struct kvm_vcpu, arch.dec)); 461 DEFINE(VCPU_DEC, offsetof(struct kvm_vcpu, arch.dec));
467 DEFINE(VCPU_DEC_EXPIRES, offsetof(struct kvm_vcpu, arch.dec_expires)); 462 DEFINE(VCPU_DEC_EXPIRES, offsetof(struct kvm_vcpu, arch.dec_expires));
468 DEFINE(VCPU_PENDING_EXC, offsetof(struct kvm_vcpu, arch.pending_exceptions)); 463 DEFINE(VCPU_PENDING_EXC, offsetof(struct kvm_vcpu, arch.pending_exceptions));
464 DEFINE(VCPU_CEDED, offsetof(struct kvm_vcpu, arch.ceded));
465 DEFINE(VCPU_PRODDED, offsetof(struct kvm_vcpu, arch.prodded));
469 DEFINE(VCPU_VPA, offsetof(struct kvm_vcpu, arch.vpa)); 466 DEFINE(VCPU_VPA, offsetof(struct kvm_vcpu, arch.vpa));
470 DEFINE(VCPU_MMCR, offsetof(struct kvm_vcpu, arch.mmcr)); 467 DEFINE(VCPU_MMCR, offsetof(struct kvm_vcpu, arch.mmcr));
471 DEFINE(VCPU_PMC, offsetof(struct kvm_vcpu, arch.pmc)); 468 DEFINE(VCPU_PMC, offsetof(struct kvm_vcpu, arch.pmc));
@@ -481,6 +478,7 @@ int main(void)
481 DEFINE(VCORE_ENTRY_EXIT, offsetof(struct kvmppc_vcore, entry_exit_count)); 478 DEFINE(VCORE_ENTRY_EXIT, offsetof(struct kvmppc_vcore, entry_exit_count));
482 DEFINE(VCORE_NAP_COUNT, offsetof(struct kvmppc_vcore, nap_count)); 479 DEFINE(VCORE_NAP_COUNT, offsetof(struct kvmppc_vcore, nap_count));
483 DEFINE(VCORE_IN_GUEST, offsetof(struct kvmppc_vcore, in_guest)); 480 DEFINE(VCORE_IN_GUEST, offsetof(struct kvmppc_vcore, in_guest));
481 DEFINE(VCORE_NAPPING_THREADS, offsetof(struct kvmppc_vcore, napping_threads));
484 DEFINE(VCPU_SVCPU, offsetof(struct kvmppc_vcpu_book3s, shadow_vcpu) - 482 DEFINE(VCPU_SVCPU, offsetof(struct kvmppc_vcpu_book3s, shadow_vcpu) -
485 offsetof(struct kvmppc_vcpu_book3s, vcpu)); 483 offsetof(struct kvmppc_vcpu_book3s, vcpu));
486 DEFINE(VCPU_SLB_E, offsetof(struct kvmppc_slb, orige)); 484 DEFINE(VCPU_SLB_E, offsetof(struct kvmppc_slb, orige));
@@ -537,6 +535,8 @@ int main(void)
537 HSTATE_FIELD(HSTATE_SCRATCH0, scratch0); 535 HSTATE_FIELD(HSTATE_SCRATCH0, scratch0);
538 HSTATE_FIELD(HSTATE_SCRATCH1, scratch1); 536 HSTATE_FIELD(HSTATE_SCRATCH1, scratch1);
539 HSTATE_FIELD(HSTATE_IN_GUEST, in_guest); 537 HSTATE_FIELD(HSTATE_IN_GUEST, in_guest);
538 HSTATE_FIELD(HSTATE_RESTORE_HID5, restore_hid5);
539 HSTATE_FIELD(HSTATE_NAPPING, napping);
540 540
541#ifdef CONFIG_KVM_BOOK3S_64_HV 541#ifdef CONFIG_KVM_BOOK3S_64_HV
542 HSTATE_FIELD(HSTATE_KVM_VCPU, kvm_vcpu); 542 HSTATE_FIELD(HSTATE_KVM_VCPU, kvm_vcpu);
@@ -549,6 +549,7 @@ int main(void)
549 HSTATE_FIELD(HSTATE_DSCR, host_dscr); 549 HSTATE_FIELD(HSTATE_DSCR, host_dscr);
550 HSTATE_FIELD(HSTATE_DABR, dabr); 550 HSTATE_FIELD(HSTATE_DABR, dabr);
551 HSTATE_FIELD(HSTATE_DECEXP, dec_expires); 551 HSTATE_FIELD(HSTATE_DECEXP, dec_expires);
552 DEFINE(IPI_PRIORITY, IPI_PRIORITY);
552#endif /* CONFIG_KVM_BOOK3S_64_HV */ 553#endif /* CONFIG_KVM_BOOK3S_64_HV */
553 554
554#else /* CONFIG_PPC_BOOK3S */ 555#else /* CONFIG_PPC_BOOK3S */
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 41b02c792aa3..29ddd8b1c274 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -427,16 +427,6 @@ slb_miss_user_pseries:
427 b . /* prevent spec. execution */ 427 b . /* prevent spec. execution */
428#endif /* __DISABLED__ */ 428#endif /* __DISABLED__ */
429 429
430/* KVM's trampoline code needs to be close to the interrupt handlers */
431
432#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
433#ifdef CONFIG_KVM_BOOK3S_PR
434#include "../kvm/book3s_rmhandlers.S"
435#else
436#include "../kvm/book3s_hv_rmhandlers.S"
437#endif
438#endif
439
440 .align 7 430 .align 7
441 .globl __end_interrupts 431 .globl __end_interrupts
442__end_interrupts: 432__end_interrupts:
diff --git a/arch/powerpc/kvm/44x.c b/arch/powerpc/kvm/44x.c
index da3a1225c0ac..ca1f88b3dc59 100644
--- a/arch/powerpc/kvm/44x.c
+++ b/arch/powerpc/kvm/44x.c
@@ -78,6 +78,8 @@ int kvmppc_core_vcpu_setup(struct kvm_vcpu *vcpu)
78 for (i = 0; i < ARRAY_SIZE(vcpu_44x->shadow_refs); i++) 78 for (i = 0; i < ARRAY_SIZE(vcpu_44x->shadow_refs); i++)
79 vcpu_44x->shadow_refs[i].gtlb_index = -1; 79 vcpu_44x->shadow_refs[i].gtlb_index = -1;
80 80
81 vcpu->arch.cpu_type = KVM_CPU_440;
82
81 return 0; 83 return 0;
82} 84}
83 85
diff --git a/arch/powerpc/kvm/Makefile b/arch/powerpc/kvm/Makefile
index 08428e2c188d..3688aeecc4b2 100644
--- a/arch/powerpc/kvm/Makefile
+++ b/arch/powerpc/kvm/Makefile
@@ -43,18 +43,22 @@ kvm-book3s_64-objs-$(CONFIG_KVM_BOOK3S_64_PR) := \
43 fpu.o \ 43 fpu.o \
44 book3s_paired_singles.o \ 44 book3s_paired_singles.o \
45 book3s_pr.o \ 45 book3s_pr.o \
46 book3s_pr_papr.o \
46 book3s_emulate.o \ 47 book3s_emulate.o \
47 book3s_interrupts.o \ 48 book3s_interrupts.o \
48 book3s_mmu_hpte.o \ 49 book3s_mmu_hpte.o \
49 book3s_64_mmu_host.o \ 50 book3s_64_mmu_host.o \
50 book3s_64_mmu.o \ 51 book3s_64_mmu.o \
51 book3s_32_mmu.o 52 book3s_32_mmu.o
53kvm-book3s_64-builtin-objs-$(CONFIG_KVM_BOOK3S_64_PR) := \
54 book3s_rmhandlers.o
52 55
53kvm-book3s_64-objs-$(CONFIG_KVM_BOOK3S_64_HV) := \ 56kvm-book3s_64-objs-$(CONFIG_KVM_BOOK3S_64_HV) := \
54 book3s_hv.o \ 57 book3s_hv.o \
55 book3s_hv_interrupts.o \ 58 book3s_hv_interrupts.o \
56 book3s_64_mmu_hv.o 59 book3s_64_mmu_hv.o
57kvm-book3s_64-builtin-objs-$(CONFIG_KVM_BOOK3S_64_HV) := \ 60kvm-book3s_64-builtin-objs-$(CONFIG_KVM_BOOK3S_64_HV) := \
61 book3s_hv_rmhandlers.o \
58 book3s_hv_rm_mmu.o \ 62 book3s_hv_rm_mmu.o \
59 book3s_64_vio_hv.o \ 63 book3s_64_vio_hv.o \
60 book3s_hv_builtin.o 64 book3s_hv_builtin.o
diff --git a/arch/powerpc/kvm/book3s_32_sr.S b/arch/powerpc/kvm/book3s_32_sr.S
index 3608471ad2d8..7e06a6fc8d07 100644
--- a/arch/powerpc/kvm/book3s_32_sr.S
+++ b/arch/powerpc/kvm/book3s_32_sr.S
@@ -31,7 +31,7 @@
31 * R1 = host R1 31 * R1 = host R1
32 * R2 = host R2 32 * R2 = host R2
33 * R3 = shadow vcpu 33 * R3 = shadow vcpu
34 * all other volatile GPRS = free 34 * all other volatile GPRS = free except R4, R6
35 * SVCPU[CR] = guest CR 35 * SVCPU[CR] = guest CR
36 * SVCPU[XER] = guest XER 36 * SVCPU[XER] = guest XER
37 * SVCPU[CTR] = guest CTR 37 * SVCPU[CTR] = guest CTR
diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
index c6d3e194b6b4..b871721c0050 100644
--- a/arch/powerpc/kvm/book3s_64_mmu.c
+++ b/arch/powerpc/kvm/book3s_64_mmu.c
@@ -128,7 +128,13 @@ static hva_t kvmppc_mmu_book3s_64_get_pteg(
128 dprintk("MMU: page=0x%x sdr1=0x%llx pteg=0x%llx vsid=0x%llx\n", 128 dprintk("MMU: page=0x%x sdr1=0x%llx pteg=0x%llx vsid=0x%llx\n",
129 page, vcpu_book3s->sdr1, pteg, slbe->vsid); 129 page, vcpu_book3s->sdr1, pteg, slbe->vsid);
130 130
131 r = gfn_to_hva(vcpu_book3s->vcpu.kvm, pteg >> PAGE_SHIFT); 131 /* When running a PAPR guest, SDR1 contains a HVA address instead
132 of a GPA */
133 if (vcpu_book3s->vcpu.arch.papr_enabled)
134 r = pteg;
135 else
136 r = gfn_to_hva(vcpu_book3s->vcpu.kvm, pteg >> PAGE_SHIFT);
137
132 if (kvm_is_error_hva(r)) 138 if (kvm_is_error_hva(r))
133 return r; 139 return r;
134 return r | (pteg & ~PAGE_MASK); 140 return r | (pteg & ~PAGE_MASK);
diff --git a/arch/powerpc/kvm/book3s_64_slb.S b/arch/powerpc/kvm/book3s_64_slb.S
index 04e7d3bbfe8b..f2e6e48ea463 100644
--- a/arch/powerpc/kvm/book3s_64_slb.S
+++ b/arch/powerpc/kvm/book3s_64_slb.S
@@ -53,7 +53,7 @@ slb_exit_skip_ ## num:
53 * R1 = host R1 53 * R1 = host R1
54 * R2 = host R2 54 * R2 = host R2
55 * R3 = shadow vcpu 55 * R3 = shadow vcpu
56 * all other volatile GPRS = free 56 * all other volatile GPRS = free except R4, R6
57 * SVCPU[CR] = guest CR 57 * SVCPU[CR] = guest CR
58 * SVCPU[XER] = guest XER 58 * SVCPU[XER] = guest XER
59 * SVCPU[CTR] = guest CTR 59 * SVCPU[CTR] = guest CTR
diff --git a/arch/powerpc/kvm/book3s_emulate.c b/arch/powerpc/kvm/book3s_emulate.c
index 466846557089..0c9dc62532d0 100644
--- a/arch/powerpc/kvm/book3s_emulate.c
+++ b/arch/powerpc/kvm/book3s_emulate.c
@@ -63,6 +63,25 @@
63 * function pointers, so let's just disable the define. */ 63 * function pointers, so let's just disable the define. */
64#undef mfsrin 64#undef mfsrin
65 65
66enum priv_level {
67 PRIV_PROBLEM = 0,
68 PRIV_SUPER = 1,
69 PRIV_HYPER = 2,
70};
71
72static bool spr_allowed(struct kvm_vcpu *vcpu, enum priv_level level)
73{
74 /* PAPR VMs only access supervisor SPRs */
75 if (vcpu->arch.papr_enabled && (level > PRIV_SUPER))
76 return false;
77
78 /* Limit user space to its own small SPR set */
79 if ((vcpu->arch.shared->msr & MSR_PR) && level > PRIV_PROBLEM)
80 return false;
81
82 return true;
83}
84
66int kvmppc_core_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu, 85int kvmppc_core_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
67 unsigned int inst, int *advance) 86 unsigned int inst, int *advance)
68{ 87{
@@ -296,6 +315,8 @@ int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
296 315
297 switch (sprn) { 316 switch (sprn) {
298 case SPRN_SDR1: 317 case SPRN_SDR1:
318 if (!spr_allowed(vcpu, PRIV_HYPER))
319 goto unprivileged;
299 to_book3s(vcpu)->sdr1 = spr_val; 320 to_book3s(vcpu)->sdr1 = spr_val;
300 break; 321 break;
301 case SPRN_DSISR: 322 case SPRN_DSISR:
@@ -390,6 +411,7 @@ int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
390 case SPRN_PMC4_GEKKO: 411 case SPRN_PMC4_GEKKO:
391 case SPRN_WPAR_GEKKO: 412 case SPRN_WPAR_GEKKO:
392 break; 413 break;
414unprivileged:
393 default: 415 default:
394 printk(KERN_INFO "KVM: invalid SPR write: %d\n", sprn); 416 printk(KERN_INFO "KVM: invalid SPR write: %d\n", sprn);
395#ifndef DEBUG_SPR 417#ifndef DEBUG_SPR
@@ -421,6 +443,8 @@ int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, int sprn, int rt)
421 break; 443 break;
422 } 444 }
423 case SPRN_SDR1: 445 case SPRN_SDR1:
446 if (!spr_allowed(vcpu, PRIV_HYPER))
447 goto unprivileged;
424 kvmppc_set_gpr(vcpu, rt, to_book3s(vcpu)->sdr1); 448 kvmppc_set_gpr(vcpu, rt, to_book3s(vcpu)->sdr1);
425 break; 449 break;
426 case SPRN_DSISR: 450 case SPRN_DSISR:
@@ -449,6 +473,10 @@ int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, int sprn, int rt)
449 case SPRN_HID5: 473 case SPRN_HID5:
450 kvmppc_set_gpr(vcpu, rt, to_book3s(vcpu)->hid[5]); 474 kvmppc_set_gpr(vcpu, rt, to_book3s(vcpu)->hid[5]);
451 break; 475 break;
476 case SPRN_CFAR:
477 case SPRN_PURR:
478 kvmppc_set_gpr(vcpu, rt, 0);
479 break;
452 case SPRN_GQR0: 480 case SPRN_GQR0:
453 case SPRN_GQR1: 481 case SPRN_GQR1:
454 case SPRN_GQR2: 482 case SPRN_GQR2:
@@ -476,6 +504,7 @@ int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, int sprn, int rt)
476 kvmppc_set_gpr(vcpu, rt, 0); 504 kvmppc_set_gpr(vcpu, rt, 0);
477 break; 505 break;
478 default: 506 default:
507unprivileged:
479 printk(KERN_INFO "KVM: invalid SPR read: %d\n", sprn); 508 printk(KERN_INFO "KVM: invalid SPR read: %d\n", sprn);
480#ifndef DEBUG_SPR 509#ifndef DEBUG_SPR
481 emulated = EMULATE_FAIL; 510 emulated = EMULATE_FAIL;
diff --git a/arch/powerpc/kvm/book3s_exports.c b/arch/powerpc/kvm/book3s_exports.c
index 88c8f26add02..f7f63a00ab1f 100644
--- a/arch/powerpc/kvm/book3s_exports.c
+++ b/arch/powerpc/kvm/book3s_exports.c
@@ -23,9 +23,7 @@
23#ifdef CONFIG_KVM_BOOK3S_64_HV 23#ifdef CONFIG_KVM_BOOK3S_64_HV
24EXPORT_SYMBOL_GPL(kvmppc_hv_entry_trampoline); 24EXPORT_SYMBOL_GPL(kvmppc_hv_entry_trampoline);
25#else 25#else
26EXPORT_SYMBOL_GPL(kvmppc_handler_trampoline_enter); 26EXPORT_SYMBOL_GPL(kvmppc_entry_trampoline);
27EXPORT_SYMBOL_GPL(kvmppc_handler_lowmem_trampoline);
28EXPORT_SYMBOL_GPL(kvmppc_rmcall);
29EXPORT_SYMBOL_GPL(kvmppc_load_up_fpu); 27EXPORT_SYMBOL_GPL(kvmppc_load_up_fpu);
30#ifdef CONFIG_ALTIVEC 28#ifdef CONFIG_ALTIVEC
31EXPORT_SYMBOL_GPL(kvmppc_load_up_altivec); 29EXPORT_SYMBOL_GPL(kvmppc_load_up_altivec);
diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
index cc0d7f1b19ab..4644c7986d80 100644
--- a/arch/powerpc/kvm/book3s_hv.c
+++ b/arch/powerpc/kvm/book3s_hv.c
@@ -62,6 +62,8 @@
62/* #define EXIT_DEBUG_SIMPLE */ 62/* #define EXIT_DEBUG_SIMPLE */
63/* #define EXIT_DEBUG_INT */ 63/* #define EXIT_DEBUG_INT */
64 64
65static void kvmppc_end_cede(struct kvm_vcpu *vcpu);
66
65void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu) 67void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
66{ 68{
67 local_paca->kvm_hstate.kvm_vcpu = vcpu; 69 local_paca->kvm_hstate.kvm_vcpu = vcpu;
@@ -72,40 +74,10 @@ void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
72{ 74{
73} 75}
74 76
75static void kvmppc_vcpu_blocked(struct kvm_vcpu *vcpu);
76static void kvmppc_vcpu_unblocked(struct kvm_vcpu *vcpu);
77
78void kvmppc_vcpu_block(struct kvm_vcpu *vcpu)
79{
80 u64 now;
81 unsigned long dec_nsec;
82
83 now = get_tb();
84 if (now >= vcpu->arch.dec_expires && !kvmppc_core_pending_dec(vcpu))
85 kvmppc_core_queue_dec(vcpu);
86 if (vcpu->arch.pending_exceptions)
87 return;
88 if (vcpu->arch.dec_expires != ~(u64)0) {
89 dec_nsec = (vcpu->arch.dec_expires - now) * NSEC_PER_SEC /
90 tb_ticks_per_sec;
91 hrtimer_start(&vcpu->arch.dec_timer, ktime_set(0, dec_nsec),
92 HRTIMER_MODE_REL);
93 }
94
95 kvmppc_vcpu_blocked(vcpu);
96
97 kvm_vcpu_block(vcpu);
98 vcpu->stat.halt_wakeup++;
99
100 if (vcpu->arch.dec_expires != ~(u64)0)
101 hrtimer_try_to_cancel(&vcpu->arch.dec_timer);
102
103 kvmppc_vcpu_unblocked(vcpu);
104}
105
106void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr) 77void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
107{ 78{
108 vcpu->arch.shregs.msr = msr; 79 vcpu->arch.shregs.msr = msr;
80 kvmppc_end_cede(vcpu);
109} 81}
110 82
111void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr) 83void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
@@ -257,15 +229,6 @@ int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu)
257 229
258 switch (req) { 230 switch (req) {
259 case H_CEDE: 231 case H_CEDE:
260 vcpu->arch.shregs.msr |= MSR_EE;
261 vcpu->arch.ceded = 1;
262 smp_mb();
263 if (!vcpu->arch.prodded)
264 kvmppc_vcpu_block(vcpu);
265 else
266 vcpu->arch.prodded = 0;
267 smp_mb();
268 vcpu->arch.ceded = 0;
269 break; 232 break;
270 case H_PROD: 233 case H_PROD:
271 target = kvmppc_get_gpr(vcpu, 4); 234 target = kvmppc_get_gpr(vcpu, 4);
@@ -388,20 +351,6 @@ static int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
388 break; 351 break;
389 } 352 }
390 353
391
392 if (!(r & RESUME_HOST)) {
393 /* To avoid clobbering exit_reason, only check for signals if
394 * we aren't already exiting to userspace for some other
395 * reason. */
396 if (signal_pending(tsk)) {
397 vcpu->stat.signal_exits++;
398 run->exit_reason = KVM_EXIT_INTR;
399 r = -EINTR;
400 } else {
401 kvmppc_core_deliver_interrupts(vcpu);
402 }
403 }
404
405 return r; 354 return r;
406} 355}
407 356
@@ -479,13 +428,9 @@ struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
479 kvmppc_mmu_book3s_hv_init(vcpu); 428 kvmppc_mmu_book3s_hv_init(vcpu);
480 429
481 /* 430 /*
482 * Some vcpus may start out in stopped state. If we initialize 431 * We consider the vcpu stopped until we see the first run ioctl for it.
483 * them to busy-in-host state they will stop other vcpus in the
484 * vcore from running. Instead we initialize them to blocked
485 * state, effectively considering them to be stopped until we
486 * see the first run ioctl for them.
487 */ 432 */
488 vcpu->arch.state = KVMPPC_VCPU_BLOCKED; 433 vcpu->arch.state = KVMPPC_VCPU_STOPPED;
489 434
490 init_waitqueue_head(&vcpu->arch.cpu_run); 435 init_waitqueue_head(&vcpu->arch.cpu_run);
491 436
@@ -496,6 +441,7 @@ struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
496 if (vcore) { 441 if (vcore) {
497 INIT_LIST_HEAD(&vcore->runnable_threads); 442 INIT_LIST_HEAD(&vcore->runnable_threads);
498 spin_lock_init(&vcore->lock); 443 spin_lock_init(&vcore->lock);
444 init_waitqueue_head(&vcore->wq);
499 } 445 }
500 kvm->arch.vcores[core] = vcore; 446 kvm->arch.vcores[core] = vcore;
501 } 447 }
@@ -506,10 +452,12 @@ struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
506 452
507 spin_lock(&vcore->lock); 453 spin_lock(&vcore->lock);
508 ++vcore->num_threads; 454 ++vcore->num_threads;
509 ++vcore->n_blocked;
510 spin_unlock(&vcore->lock); 455 spin_unlock(&vcore->lock);
511 vcpu->arch.vcore = vcore; 456 vcpu->arch.vcore = vcore;
512 457
458 vcpu->arch.cpu_type = KVM_CPU_3S_64;
459 kvmppc_sanity_check(vcpu);
460
513 return vcpu; 461 return vcpu;
514 462
515free_vcpu: 463free_vcpu:
@@ -524,30 +472,31 @@ void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
524 kfree(vcpu); 472 kfree(vcpu);
525} 473}
526 474
527static void kvmppc_vcpu_blocked(struct kvm_vcpu *vcpu) 475static void kvmppc_set_timer(struct kvm_vcpu *vcpu)
528{ 476{
529 struct kvmppc_vcore *vc = vcpu->arch.vcore; 477 unsigned long dec_nsec, now;
530 478
531 spin_lock(&vc->lock); 479 now = get_tb();
532 vcpu->arch.state = KVMPPC_VCPU_BLOCKED; 480 if (now > vcpu->arch.dec_expires) {
533 ++vc->n_blocked; 481 /* decrementer has already gone negative */
534 if (vc->n_runnable > 0 && 482 kvmppc_core_queue_dec(vcpu);
535 vc->n_runnable + vc->n_blocked == vc->num_threads) { 483 kvmppc_core_deliver_interrupts(vcpu);
536 vcpu = list_first_entry(&vc->runnable_threads, struct kvm_vcpu, 484 return;
537 arch.run_list);
538 wake_up(&vcpu->arch.cpu_run);
539 } 485 }
540 spin_unlock(&vc->lock); 486 dec_nsec = (vcpu->arch.dec_expires - now) * NSEC_PER_SEC
487 / tb_ticks_per_sec;
488 hrtimer_start(&vcpu->arch.dec_timer, ktime_set(0, dec_nsec),
489 HRTIMER_MODE_REL);
490 vcpu->arch.timer_running = 1;
541} 491}
542 492
543static void kvmppc_vcpu_unblocked(struct kvm_vcpu *vcpu) 493static void kvmppc_end_cede(struct kvm_vcpu *vcpu)
544{ 494{
545 struct kvmppc_vcore *vc = vcpu->arch.vcore; 495 vcpu->arch.ceded = 0;
546 496 if (vcpu->arch.timer_running) {
547 spin_lock(&vc->lock); 497 hrtimer_try_to_cancel(&vcpu->arch.dec_timer);
548 vcpu->arch.state = KVMPPC_VCPU_BUSY_IN_HOST; 498 vcpu->arch.timer_running = 0;
549 --vc->n_blocked; 499 }
550 spin_unlock(&vc->lock);
551} 500}
552 501
553extern int __kvmppc_vcore_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu); 502extern int __kvmppc_vcore_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
@@ -562,6 +511,7 @@ static void kvmppc_remove_runnable(struct kvmppc_vcore *vc,
562 return; 511 return;
563 vcpu->arch.state = KVMPPC_VCPU_BUSY_IN_HOST; 512 vcpu->arch.state = KVMPPC_VCPU_BUSY_IN_HOST;
564 --vc->n_runnable; 513 --vc->n_runnable;
514 ++vc->n_busy;
565 /* decrement the physical thread id of each following vcpu */ 515 /* decrement the physical thread id of each following vcpu */
566 v = vcpu; 516 v = vcpu;
567 list_for_each_entry_continue(v, &vc->runnable_threads, arch.run_list) 517 list_for_each_entry_continue(v, &vc->runnable_threads, arch.run_list)
@@ -575,15 +525,20 @@ static void kvmppc_start_thread(struct kvm_vcpu *vcpu)
575 struct paca_struct *tpaca; 525 struct paca_struct *tpaca;
576 struct kvmppc_vcore *vc = vcpu->arch.vcore; 526 struct kvmppc_vcore *vc = vcpu->arch.vcore;
577 527
528 if (vcpu->arch.timer_running) {
529 hrtimer_try_to_cancel(&vcpu->arch.dec_timer);
530 vcpu->arch.timer_running = 0;
531 }
578 cpu = vc->pcpu + vcpu->arch.ptid; 532 cpu = vc->pcpu + vcpu->arch.ptid;
579 tpaca = &paca[cpu]; 533 tpaca = &paca[cpu];
580 tpaca->kvm_hstate.kvm_vcpu = vcpu; 534 tpaca->kvm_hstate.kvm_vcpu = vcpu;
581 tpaca->kvm_hstate.kvm_vcore = vc; 535 tpaca->kvm_hstate.kvm_vcore = vc;
536 tpaca->kvm_hstate.napping = 0;
537 vcpu->cpu = vc->pcpu;
582 smp_wmb(); 538 smp_wmb();
583#ifdef CONFIG_PPC_ICP_NATIVE 539#ifdef CONFIG_PPC_ICP_NATIVE
584 if (vcpu->arch.ptid) { 540 if (vcpu->arch.ptid) {
585 tpaca->cpu_start = 0x80; 541 tpaca->cpu_start = 0x80;
586 tpaca->kvm_hstate.in_guest = KVM_GUEST_MODE_GUEST;
587 wmb(); 542 wmb();
588 xics_wake_cpu(cpu); 543 xics_wake_cpu(cpu);
589 ++vc->n_woken; 544 ++vc->n_woken;
@@ -631,9 +586,10 @@ static int on_primary_thread(void)
631 */ 586 */
632static int kvmppc_run_core(struct kvmppc_vcore *vc) 587static int kvmppc_run_core(struct kvmppc_vcore *vc)
633{ 588{
634 struct kvm_vcpu *vcpu, *vnext; 589 struct kvm_vcpu *vcpu, *vcpu0, *vnext;
635 long ret; 590 long ret;
636 u64 now; 591 u64 now;
592 int ptid;
637 593
638 /* don't start if any threads have a signal pending */ 594 /* don't start if any threads have a signal pending */
639 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list) 595 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list)
@@ -652,29 +608,50 @@ static int kvmppc_run_core(struct kvmppc_vcore *vc)
652 goto out; 608 goto out;
653 } 609 }
654 610
611 /*
612 * Assign physical thread IDs, first to non-ceded vcpus
613 * and then to ceded ones.
614 */
615 ptid = 0;
616 vcpu0 = NULL;
617 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list) {
618 if (!vcpu->arch.ceded) {
619 if (!ptid)
620 vcpu0 = vcpu;
621 vcpu->arch.ptid = ptid++;
622 }
623 }
624 if (!vcpu0)
625 return 0; /* nothing to run */
626 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list)
627 if (vcpu->arch.ceded)
628 vcpu->arch.ptid = ptid++;
629
655 vc->n_woken = 0; 630 vc->n_woken = 0;
656 vc->nap_count = 0; 631 vc->nap_count = 0;
657 vc->entry_exit_count = 0; 632 vc->entry_exit_count = 0;
658 vc->vcore_running = 1; 633 vc->vcore_state = VCORE_RUNNING;
659 vc->in_guest = 0; 634 vc->in_guest = 0;
660 vc->pcpu = smp_processor_id(); 635 vc->pcpu = smp_processor_id();
636 vc->napping_threads = 0;
661 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list) 637 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list)
662 kvmppc_start_thread(vcpu); 638 kvmppc_start_thread(vcpu);
663 vcpu = list_first_entry(&vc->runnable_threads, struct kvm_vcpu,
664 arch.run_list);
665 639
640 preempt_disable();
666 spin_unlock(&vc->lock); 641 spin_unlock(&vc->lock);
667 642
668 preempt_disable();
669 kvm_guest_enter(); 643 kvm_guest_enter();
670 __kvmppc_vcore_entry(NULL, vcpu); 644 __kvmppc_vcore_entry(NULL, vcpu0);
671 645
672 /* wait for secondary threads to finish writing their state to memory */
673 spin_lock(&vc->lock); 646 spin_lock(&vc->lock);
647 /* disable sending of IPIs on virtual external irqs */
648 list_for_each_entry(vcpu, &vc->runnable_threads, arch.run_list)
649 vcpu->cpu = -1;
650 /* wait for secondary threads to finish writing their state to memory */
674 if (vc->nap_count < vc->n_woken) 651 if (vc->nap_count < vc->n_woken)
675 kvmppc_wait_for_nap(vc); 652 kvmppc_wait_for_nap(vc);
676 /* prevent other vcpu threads from doing kvmppc_start_thread() now */ 653 /* prevent other vcpu threads from doing kvmppc_start_thread() now */
677 vc->vcore_running = 2; 654 vc->vcore_state = VCORE_EXITING;
678 spin_unlock(&vc->lock); 655 spin_unlock(&vc->lock);
679 656
680 /* make sure updates to secondary vcpu structs are visible now */ 657 /* make sure updates to secondary vcpu structs are visible now */
@@ -690,22 +667,26 @@ static int kvmppc_run_core(struct kvmppc_vcore *vc)
690 if (now < vcpu->arch.dec_expires && 667 if (now < vcpu->arch.dec_expires &&
691 kvmppc_core_pending_dec(vcpu)) 668 kvmppc_core_pending_dec(vcpu))
692 kvmppc_core_dequeue_dec(vcpu); 669 kvmppc_core_dequeue_dec(vcpu);
693 if (!vcpu->arch.trap) { 670
694 if (signal_pending(vcpu->arch.run_task)) { 671 ret = RESUME_GUEST;
695 vcpu->arch.kvm_run->exit_reason = KVM_EXIT_INTR; 672 if (vcpu->arch.trap)
696 vcpu->arch.ret = -EINTR; 673 ret = kvmppc_handle_exit(vcpu->arch.kvm_run, vcpu,
697 } 674 vcpu->arch.run_task);
698 continue; /* didn't get to run */ 675
699 }
700 ret = kvmppc_handle_exit(vcpu->arch.kvm_run, vcpu,
701 vcpu->arch.run_task);
702 vcpu->arch.ret = ret; 676 vcpu->arch.ret = ret;
703 vcpu->arch.trap = 0; 677 vcpu->arch.trap = 0;
678
679 if (vcpu->arch.ceded) {
680 if (ret != RESUME_GUEST)
681 kvmppc_end_cede(vcpu);
682 else
683 kvmppc_set_timer(vcpu);
684 }
704 } 685 }
705 686
706 spin_lock(&vc->lock); 687 spin_lock(&vc->lock);
707 out: 688 out:
708 vc->vcore_running = 0; 689 vc->vcore_state = VCORE_INACTIVE;
709 list_for_each_entry_safe(vcpu, vnext, &vc->runnable_threads, 690 list_for_each_entry_safe(vcpu, vnext, &vc->runnable_threads,
710 arch.run_list) { 691 arch.run_list) {
711 if (vcpu->arch.ret != RESUME_GUEST) { 692 if (vcpu->arch.ret != RESUME_GUEST) {
@@ -717,82 +698,130 @@ static int kvmppc_run_core(struct kvmppc_vcore *vc)
717 return 1; 698 return 1;
718} 699}
719 700
720static int kvmppc_run_vcpu(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu) 701/*
702 * Wait for some other vcpu thread to execute us, and
703 * wake us up when we need to handle something in the host.
704 */
705static void kvmppc_wait_for_exec(struct kvm_vcpu *vcpu, int wait_state)
721{ 706{
722 int ptid;
723 int wait_state;
724 struct kvmppc_vcore *vc;
725 DEFINE_WAIT(wait); 707 DEFINE_WAIT(wait);
726 708
727 /* No need to go into the guest when all we do is going out */ 709 prepare_to_wait(&vcpu->arch.cpu_run, &wait, wait_state);
728 if (signal_pending(current)) { 710 if (vcpu->arch.state == KVMPPC_VCPU_RUNNABLE)
729 kvm_run->exit_reason = KVM_EXIT_INTR; 711 schedule();
730 return -EINTR; 712 finish_wait(&vcpu->arch.cpu_run, &wait);
713}
714
715/*
716 * All the vcpus in this vcore are idle, so wait for a decrementer
717 * or external interrupt to one of the vcpus. vc->lock is held.
718 */
719static void kvmppc_vcore_blocked(struct kvmppc_vcore *vc)
720{
721 DEFINE_WAIT(wait);
722 struct kvm_vcpu *v;
723 int all_idle = 1;
724
725 prepare_to_wait(&vc->wq, &wait, TASK_INTERRUPTIBLE);
726 vc->vcore_state = VCORE_SLEEPING;
727 spin_unlock(&vc->lock);
728 list_for_each_entry(v, &vc->runnable_threads, arch.run_list) {
729 if (!v->arch.ceded || v->arch.pending_exceptions) {
730 all_idle = 0;
731 break;
732 }
731 } 733 }
734 if (all_idle)
735 schedule();
736 finish_wait(&vc->wq, &wait);
737 spin_lock(&vc->lock);
738 vc->vcore_state = VCORE_INACTIVE;
739}
732 740
733 /* On PPC970, check that we have an RMA region */ 741static int kvmppc_run_vcpu(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
734 if (!vcpu->kvm->arch.rma && cpu_has_feature(CPU_FTR_ARCH_201)) 742{
735 return -EPERM; 743 int n_ceded;
744 int prev_state;
745 struct kvmppc_vcore *vc;
746 struct kvm_vcpu *v, *vn;
736 747
737 kvm_run->exit_reason = 0; 748 kvm_run->exit_reason = 0;
738 vcpu->arch.ret = RESUME_GUEST; 749 vcpu->arch.ret = RESUME_GUEST;
739 vcpu->arch.trap = 0; 750 vcpu->arch.trap = 0;
740 751
741 flush_fp_to_thread(current);
742 flush_altivec_to_thread(current);
743 flush_vsx_to_thread(current);
744
745 /* 752 /*
746 * Synchronize with other threads in this virtual core 753 * Synchronize with other threads in this virtual core
747 */ 754 */
748 vc = vcpu->arch.vcore; 755 vc = vcpu->arch.vcore;
749 spin_lock(&vc->lock); 756 spin_lock(&vc->lock);
750 /* This happens the first time this is called for a vcpu */ 757 vcpu->arch.ceded = 0;
751 if (vcpu->arch.state == KVMPPC_VCPU_BLOCKED)
752 --vc->n_blocked;
753 vcpu->arch.state = KVMPPC_VCPU_RUNNABLE;
754 ptid = vc->n_runnable;
755 vcpu->arch.run_task = current; 758 vcpu->arch.run_task = current;
756 vcpu->arch.kvm_run = kvm_run; 759 vcpu->arch.kvm_run = kvm_run;
757 vcpu->arch.ptid = ptid; 760 prev_state = vcpu->arch.state;
761 vcpu->arch.state = KVMPPC_VCPU_RUNNABLE;
758 list_add_tail(&vcpu->arch.run_list, &vc->runnable_threads); 762 list_add_tail(&vcpu->arch.run_list, &vc->runnable_threads);
759 ++vc->n_runnable; 763 ++vc->n_runnable;
760 764
761 wait_state = TASK_INTERRUPTIBLE; 765 /*
762 while (vcpu->arch.state == KVMPPC_VCPU_RUNNABLE) { 766 * This happens the first time this is called for a vcpu.
763 if (signal_pending(current)) { 767 * If the vcore is already running, we may be able to start
764 if (!vc->vcore_running) { 768 * this thread straight away and have it join in.
765 kvm_run->exit_reason = KVM_EXIT_INTR; 769 */
766 vcpu->arch.ret = -EINTR; 770 if (prev_state == KVMPPC_VCPU_STOPPED) {
767 break; 771 if (vc->vcore_state == VCORE_RUNNING &&
768 } 772 VCORE_EXIT_COUNT(vc) == 0) {
769 /* have to wait for vcore to stop executing guest */ 773 vcpu->arch.ptid = vc->n_runnable - 1;
770 wait_state = TASK_UNINTERRUPTIBLE; 774 kvmppc_start_thread(vcpu);
771 smp_send_reschedule(vc->pcpu);
772 } 775 }
773 776
774 if (!vc->vcore_running && 777 } else if (prev_state == KVMPPC_VCPU_BUSY_IN_HOST)
775 vc->n_runnable + vc->n_blocked == vc->num_threads) { 778 --vc->n_busy;
776 /* we can run now */
777 if (kvmppc_run_core(vc))
778 continue;
779 }
780 779
781 if (vc->vcore_running == 1 && VCORE_EXIT_COUNT(vc) == 0) 780 while (vcpu->arch.state == KVMPPC_VCPU_RUNNABLE &&
782 kvmppc_start_thread(vcpu); 781 !signal_pending(current)) {
782 if (vc->n_busy || vc->vcore_state != VCORE_INACTIVE) {
783 spin_unlock(&vc->lock);
784 kvmppc_wait_for_exec(vcpu, TASK_INTERRUPTIBLE);
785 spin_lock(&vc->lock);
786 continue;
787 }
788 n_ceded = 0;
789 list_for_each_entry(v, &vc->runnable_threads, arch.run_list)
790 n_ceded += v->arch.ceded;
791 if (n_ceded == vc->n_runnable)
792 kvmppc_vcore_blocked(vc);
793 else
794 kvmppc_run_core(vc);
795
796 list_for_each_entry_safe(v, vn, &vc->runnable_threads,
797 arch.run_list) {
798 kvmppc_core_deliver_interrupts(v);
799 if (signal_pending(v->arch.run_task)) {
800 kvmppc_remove_runnable(vc, v);
801 v->stat.signal_exits++;
802 v->arch.kvm_run->exit_reason = KVM_EXIT_INTR;
803 v->arch.ret = -EINTR;
804 wake_up(&v->arch.cpu_run);
805 }
806 }
807 }
783 808
784 /* wait for other threads to come in, or wait for vcore */ 809 if (signal_pending(current)) {
785 prepare_to_wait(&vcpu->arch.cpu_run, &wait, wait_state); 810 if (vc->vcore_state == VCORE_RUNNING ||
786 spin_unlock(&vc->lock); 811 vc->vcore_state == VCORE_EXITING) {
787 schedule(); 812 spin_unlock(&vc->lock);
788 finish_wait(&vcpu->arch.cpu_run, &wait); 813 kvmppc_wait_for_exec(vcpu, TASK_UNINTERRUPTIBLE);
789 spin_lock(&vc->lock); 814 spin_lock(&vc->lock);
815 }
816 if (vcpu->arch.state == KVMPPC_VCPU_RUNNABLE) {
817 kvmppc_remove_runnable(vc, vcpu);
818 vcpu->stat.signal_exits++;
819 kvm_run->exit_reason = KVM_EXIT_INTR;
820 vcpu->arch.ret = -EINTR;
821 }
790 } 822 }
791 823
792 if (vcpu->arch.state == KVMPPC_VCPU_RUNNABLE)
793 kvmppc_remove_runnable(vc, vcpu);
794 spin_unlock(&vc->lock); 824 spin_unlock(&vc->lock);
795
796 return vcpu->arch.ret; 825 return vcpu->arch.ret;
797} 826}
798 827
@@ -800,6 +829,26 @@ int kvmppc_vcpu_run(struct kvm_run *run, struct kvm_vcpu *vcpu)
800{ 829{
801 int r; 830 int r;
802 831
832 if (!vcpu->arch.sane) {
833 run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
834 return -EINVAL;
835 }
836
837 /* No need to go into the guest when all we'll do is come back out */
838 if (signal_pending(current)) {
839 run->exit_reason = KVM_EXIT_INTR;
840 return -EINTR;
841 }
842
843 /* On PPC970, check that we have an RMA region */
844 if (!vcpu->kvm->arch.rma && cpu_has_feature(CPU_FTR_ARCH_201))
845 return -EPERM;
846
847 flush_fp_to_thread(current);
848 flush_altivec_to_thread(current);
849 flush_vsx_to_thread(current);
850 vcpu->arch.wqp = &vcpu->arch.vcore->wq;
851
803 do { 852 do {
804 r = kvmppc_run_vcpu(run, vcpu); 853 r = kvmppc_run_vcpu(run, vcpu);
805 854
diff --git a/arch/powerpc/kvm/book3s_hv_rm_mmu.c b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
index fcfe6b055558..bacb0cfa3602 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_mmu.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
@@ -110,39 +110,6 @@ long kvmppc_h_enter(struct kvm_vcpu *vcpu, unsigned long flags,
110 return H_SUCCESS; 110 return H_SUCCESS;
111} 111}
112 112
113static unsigned long compute_tlbie_rb(unsigned long v, unsigned long r,
114 unsigned long pte_index)
115{
116 unsigned long rb, va_low;
117
118 rb = (v & ~0x7fUL) << 16; /* AVA field */
119 va_low = pte_index >> 3;
120 if (v & HPTE_V_SECONDARY)
121 va_low = ~va_low;
122 /* xor vsid from AVA */
123 if (!(v & HPTE_V_1TB_SEG))
124 va_low ^= v >> 12;
125 else
126 va_low ^= v >> 24;
127 va_low &= 0x7ff;
128 if (v & HPTE_V_LARGE) {
129 rb |= 1; /* L field */
130 if (cpu_has_feature(CPU_FTR_ARCH_206) &&
131 (r & 0xff000)) {
132 /* non-16MB large page, must be 64k */
133 /* (masks depend on page size) */
134 rb |= 0x1000; /* page encoding in LP field */
135 rb |= (va_low & 0x7f) << 16; /* 7b of VA in AVA/LP field */
136 rb |= (va_low & 0xfe); /* AVAL field (P7 doesn't seem to care) */
137 }
138 } else {
139 /* 4kB page */
140 rb |= (va_low & 0x7ff) << 12; /* remaining 11b of VA */
141 }
142 rb |= (v >> 54) & 0x300; /* B field */
143 return rb;
144}
145
146#define LOCK_TOKEN (*(u32 *)(&get_paca()->lock_token)) 113#define LOCK_TOKEN (*(u32 *)(&get_paca()->lock_token))
147 114
148static inline int try_lock_tlbie(unsigned int *lock) 115static inline int try_lock_tlbie(unsigned int *lock)
diff --git a/arch/powerpc/kvm/book3s_hv_rmhandlers.S b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
index de2950135e6e..f422231d9235 100644
--- a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
+++ b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
@@ -20,7 +20,10 @@
20#include <asm/ppc_asm.h> 20#include <asm/ppc_asm.h>
21#include <asm/kvm_asm.h> 21#include <asm/kvm_asm.h>
22#include <asm/reg.h> 22#include <asm/reg.h>
23#include <asm/mmu.h>
23#include <asm/page.h> 24#include <asm/page.h>
25#include <asm/ptrace.h>
26#include <asm/hvcall.h>
24#include <asm/asm-offsets.h> 27#include <asm/asm-offsets.h>
25#include <asm/exception-64s.h> 28#include <asm/exception-64s.h>
26 29
@@ -49,7 +52,7 @@ kvmppc_skip_Hinterrupt:
49 b . 52 b .
50 53
51/* 54/*
52 * Call kvmppc_handler_trampoline_enter in real mode. 55 * Call kvmppc_hv_entry in real mode.
53 * Must be called with interrupts hard-disabled. 56 * Must be called with interrupts hard-disabled.
54 * 57 *
55 * Input Registers: 58 * Input Registers:
@@ -89,6 +92,12 @@ _GLOBAL(kvmppc_hv_entry_trampoline)
89kvm_start_guest: 92kvm_start_guest:
90 ld r1,PACAEMERGSP(r13) 93 ld r1,PACAEMERGSP(r13)
91 subi r1,r1,STACK_FRAME_OVERHEAD 94 subi r1,r1,STACK_FRAME_OVERHEAD
95 ld r2,PACATOC(r13)
96
97 /* were we napping due to cede? */
98 lbz r0,HSTATE_NAPPING(r13)
99 cmpwi r0,0
100 bne kvm_end_cede
92 101
93 /* get vcpu pointer */ 102 /* get vcpu pointer */
94 ld r4, HSTATE_KVM_VCPU(r13) 103 ld r4, HSTATE_KVM_VCPU(r13)
@@ -276,15 +285,9 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
276 cmpwi r0,0 285 cmpwi r0,0
277 beq 20b 286 beq 20b
278 287
279 /* Set LPCR. Set the MER bit if there is a pending external irq. */ 288 /* Set LPCR and RMOR. */
28010: ld r8,KVM_LPCR(r9) 28910: ld r8,KVM_LPCR(r9)
281 ld r0,VCPU_PENDING_EXC(r4) 290 mtspr SPRN_LPCR,r8
282 li r7,(1 << BOOK3S_IRQPRIO_EXTERNAL)
283 oris r7,r7,(1 << BOOK3S_IRQPRIO_EXTERNAL_LEVEL)@h
284 and. r0,r0,r7
285 beq 11f
286 ori r8,r8,LPCR_MER
28711: mtspr SPRN_LPCR,r8
288 ld r8,KVM_RMOR(r9) 291 ld r8,KVM_RMOR(r9)
289 mtspr SPRN_RMOR,r8 292 mtspr SPRN_RMOR,r8
290 isync 293 isync
@@ -448,19 +451,50 @@ toc_tlbie_lock:
448 mtctr r6 451 mtctr r6
449 mtxer r7 452 mtxer r7
450 453
451 /* Move SRR0 and SRR1 into the respective regs */ 454kvmppc_cede_reentry: /* r4 = vcpu, r13 = paca */
452 ld r6, VCPU_SRR0(r4) 455 ld r6, VCPU_SRR0(r4)
453 ld r7, VCPU_SRR1(r4) 456 ld r7, VCPU_SRR1(r4)
454 mtspr SPRN_SRR0, r6
455 mtspr SPRN_SRR1, r7
456
457 ld r10, VCPU_PC(r4) 457 ld r10, VCPU_PC(r4)
458 ld r11, VCPU_MSR(r4) /* r11 = vcpu->arch.msr & ~MSR_HV */
458 459
459 ld r11, VCPU_MSR(r4) /* r10 = vcpu->arch.msr & ~MSR_HV */
460 rldicl r11, r11, 63 - MSR_HV_LG, 1 460 rldicl r11, r11, 63 - MSR_HV_LG, 1
461 rotldi r11, r11, 1 + MSR_HV_LG 461 rotldi r11, r11, 1 + MSR_HV_LG
462 ori r11, r11, MSR_ME 462 ori r11, r11, MSR_ME
463 463
464 /* Check if we can deliver an external or decrementer interrupt now */
465 ld r0,VCPU_PENDING_EXC(r4)
466 li r8,(1 << BOOK3S_IRQPRIO_EXTERNAL)
467 oris r8,r8,(1 << BOOK3S_IRQPRIO_EXTERNAL_LEVEL)@h
468 and r0,r0,r8
469 cmpdi cr1,r0,0
470 andi. r0,r11,MSR_EE
471 beq cr1,11f
472BEGIN_FTR_SECTION
473 mfspr r8,SPRN_LPCR
474 ori r8,r8,LPCR_MER
475 mtspr SPRN_LPCR,r8
476 isync
477END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
478 beq 5f
479 li r0,BOOK3S_INTERRUPT_EXTERNAL
48012: mr r6,r10
481 mr r10,r0
482 mr r7,r11
483 li r11,(MSR_ME << 1) | 1 /* synthesize MSR_SF | MSR_ME */
484 rotldi r11,r11,63
485 b 5f
48611: beq 5f
487 mfspr r0,SPRN_DEC
488 cmpwi r0,0
489 li r0,BOOK3S_INTERRUPT_DECREMENTER
490 blt 12b
491
492 /* Move SRR0 and SRR1 into the respective regs */
4935: mtspr SPRN_SRR0, r6
494 mtspr SPRN_SRR1, r7
495 li r0,0
496 stb r0,VCPU_CEDED(r4) /* cancel cede */
497
464fast_guest_return: 498fast_guest_return:
465 mtspr SPRN_HSRR0,r10 499 mtspr SPRN_HSRR0,r10
466 mtspr SPRN_HSRR1,r11 500 mtspr SPRN_HSRR1,r11
@@ -574,21 +608,20 @@ kvmppc_interrupt:
574 /* See if this is something we can handle in real mode */ 608 /* See if this is something we can handle in real mode */
575 cmpwi r12,BOOK3S_INTERRUPT_SYSCALL 609 cmpwi r12,BOOK3S_INTERRUPT_SYSCALL
576 beq hcall_try_real_mode 610 beq hcall_try_real_mode
577hcall_real_cont:
578 611
579 /* Check for mediated interrupts (could be done earlier really ...) */ 612 /* Check for mediated interrupts (could be done earlier really ...) */
580BEGIN_FTR_SECTION 613BEGIN_FTR_SECTION
581 cmpwi r12,BOOK3S_INTERRUPT_EXTERNAL 614 cmpwi r12,BOOK3S_INTERRUPT_EXTERNAL
582 bne+ 1f 615 bne+ 1f
583 ld r5,VCPU_KVM(r9)
584 ld r5,KVM_LPCR(r5)
585 andi. r0,r11,MSR_EE 616 andi. r0,r11,MSR_EE
586 beq 1f 617 beq 1f
618 mfspr r5,SPRN_LPCR
587 andi. r0,r5,LPCR_MER 619 andi. r0,r5,LPCR_MER
588 bne bounce_ext_interrupt 620 bne bounce_ext_interrupt
5891: 6211:
590END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206) 622END_FTR_SECTION_IFSET(CPU_FTR_ARCH_206)
591 623
624hcall_real_cont: /* r9 = vcpu, r12 = trap, r13 = paca */
592 /* Save DEC */ 625 /* Save DEC */
593 mfspr r5,SPRN_DEC 626 mfspr r5,SPRN_DEC
594 mftb r6 627 mftb r6
@@ -682,7 +715,7 @@ END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_201)
682 slbia 715 slbia
683 ptesync 716 ptesync
684 717
685hdec_soon: 718hdec_soon: /* r9 = vcpu, r12 = trap, r13 = paca */
686BEGIN_FTR_SECTION 719BEGIN_FTR_SECTION
687 b 32f 720 b 32f
688END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201) 721END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
@@ -700,6 +733,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
700 addi r0,r3,0x100 733 addi r0,r3,0x100
701 stwcx. r0,0,r6 734 stwcx. r0,0,r6
702 bne 41b 735 bne 41b
736 lwsync
703 737
704 /* 738 /*
705 * At this point we have an interrupt that we have to pass 739 * At this point we have an interrupt that we have to pass
@@ -713,18 +747,39 @@ END_FTR_SECTION_IFSET(CPU_FTR_ARCH_201)
713 * interrupt, since the other threads will already be on their 747 * interrupt, since the other threads will already be on their
714 * way here in that case. 748 * way here in that case.
715 */ 749 */
750 cmpwi r3,0x100 /* Are we the first here? */
751 bge 43f
752 cmpwi r3,1 /* Are any other threads in the guest? */
753 ble 43f
716 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER 754 cmpwi r12,BOOK3S_INTERRUPT_HV_DECREMENTER
717 beq 40f 755 beq 40f
718 cmpwi r3,0x100 /* Are we the first here? */
719 bge 40f
720 cmpwi r3,1
721 ble 40f
722 li r0,0 756 li r0,0
723 mtspr SPRN_HDEC,r0 757 mtspr SPRN_HDEC,r0
72440: 75840:
759 /*
760 * Send an IPI to any napping threads, since an HDEC interrupt
761 * doesn't wake CPUs up from nap.
762 */
763 lwz r3,VCORE_NAPPING_THREADS(r5)
764 lwz r4,VCPU_PTID(r9)
765 li r0,1
766 sldi r0,r0,r4
767 andc. r3,r3,r0 /* no sense IPI'ing ourselves */
768 beq 43f
769 mulli r4,r4,PACA_SIZE /* get paca for thread 0 */
770 subf r6,r4,r13
77142: andi. r0,r3,1
772 beq 44f
773 ld r8,HSTATE_XICS_PHYS(r6) /* get thread's XICS reg addr */
774 li r0,IPI_PRIORITY
775 li r7,XICS_QIRR
776 stbcix r0,r7,r8 /* trigger the IPI */
77744: srdi. r3,r3,1
778 addi r6,r6,PACA_SIZE
779 bne 42b
725 780
726 /* Secondary threads wait for primary to do partition switch */ 781 /* Secondary threads wait for primary to do partition switch */
727 ld r4,VCPU_KVM(r9) /* pointer to struct kvm */ 78243: ld r4,VCPU_KVM(r9) /* pointer to struct kvm */
728 ld r5,HSTATE_KVM_VCORE(r13) 783 ld r5,HSTATE_KVM_VCORE(r13)
729 lwz r3,VCPU_PTID(r9) 784 lwz r3,VCPU_PTID(r9)
730 cmpwi r3,0 785 cmpwi r3,0
@@ -1077,7 +1132,6 @@ hcall_try_real_mode:
1077hcall_real_fallback: 1132hcall_real_fallback:
1078 li r12,BOOK3S_INTERRUPT_SYSCALL 1133 li r12,BOOK3S_INTERRUPT_SYSCALL
1079 ld r9, HSTATE_KVM_VCPU(r13) 1134 ld r9, HSTATE_KVM_VCPU(r13)
1080 ld r11, VCPU_MSR(r9)
1081 1135
1082 b hcall_real_cont 1136 b hcall_real_cont
1083 1137
@@ -1139,7 +1193,7 @@ hcall_real_table:
1139 .long 0 /* 0xd4 */ 1193 .long 0 /* 0xd4 */
1140 .long 0 /* 0xd8 */ 1194 .long 0 /* 0xd8 */
1141 .long 0 /* 0xdc */ 1195 .long 0 /* 0xdc */
1142 .long 0 /* 0xe0 */ 1196 .long .kvmppc_h_cede - hcall_real_table
1143 .long 0 /* 0xe4 */ 1197 .long 0 /* 0xe4 */
1144 .long 0 /* 0xe8 */ 1198 .long 0 /* 0xe8 */
1145 .long 0 /* 0xec */ 1199 .long 0 /* 0xec */
@@ -1168,7 +1222,8 @@ bounce_ext_interrupt:
1168 mtspr SPRN_SRR0,r10 1222 mtspr SPRN_SRR0,r10
1169 mtspr SPRN_SRR1,r11 1223 mtspr SPRN_SRR1,r11
1170 li r10,BOOK3S_INTERRUPT_EXTERNAL 1224 li r10,BOOK3S_INTERRUPT_EXTERNAL
1171 LOAD_REG_IMMEDIATE(r11,MSR_SF | MSR_ME); 1225 li r11,(MSR_ME << 1) | 1 /* synthesize MSR_SF | MSR_ME */
1226 rotldi r11,r11,63
1172 b fast_guest_return 1227 b fast_guest_return
1173 1228
1174_GLOBAL(kvmppc_h_set_dabr) 1229_GLOBAL(kvmppc_h_set_dabr)
@@ -1177,6 +1232,178 @@ _GLOBAL(kvmppc_h_set_dabr)
1177 li r3,0 1232 li r3,0
1178 blr 1233 blr
1179 1234
1235_GLOBAL(kvmppc_h_cede)
1236 ori r11,r11,MSR_EE
1237 std r11,VCPU_MSR(r3)
1238 li r0,1
1239 stb r0,VCPU_CEDED(r3)
1240 sync /* order setting ceded vs. testing prodded */
1241 lbz r5,VCPU_PRODDED(r3)
1242 cmpwi r5,0
1243 bne 1f
1244 li r0,0 /* set trap to 0 to say hcall is handled */
1245 stw r0,VCPU_TRAP(r3)
1246 li r0,H_SUCCESS
1247 std r0,VCPU_GPR(r3)(r3)
1248BEGIN_FTR_SECTION
1249 b 2f /* just send it up to host on 970 */
1250END_FTR_SECTION_IFCLR(CPU_FTR_ARCH_206)
1251
1252 /*
1253 * Set our bit in the bitmask of napping threads unless all the
1254 * other threads are already napping, in which case we send this
1255 * up to the host.
1256 */
1257 ld r5,HSTATE_KVM_VCORE(r13)
1258 lwz r6,VCPU_PTID(r3)
1259 lwz r8,VCORE_ENTRY_EXIT(r5)
1260 clrldi r8,r8,56
1261 li r0,1
1262 sld r0,r0,r6
1263 addi r6,r5,VCORE_NAPPING_THREADS
126431: lwarx r4,0,r6
1265 or r4,r4,r0
1266 popcntw r7,r4
1267 cmpw r7,r8
1268 bge 2f
1269 stwcx. r4,0,r6
1270 bne 31b
1271 li r0,1
1272 stb r0,HSTATE_NAPPING(r13)
1273 /* order napping_threads update vs testing entry_exit_count */
1274 lwsync
1275 mr r4,r3
1276 lwz r7,VCORE_ENTRY_EXIT(r5)
1277 cmpwi r7,0x100
1278 bge 33f /* another thread already exiting */
1279
1280/*
1281 * Although not specifically required by the architecture, POWER7
1282 * preserves the following registers in nap mode, even if an SMT mode
1283 * switch occurs: SLB entries, PURR, SPURR, AMOR, UAMOR, AMR, SPRG0-3,
1284 * DAR, DSISR, DABR, DABRX, DSCR, PMCx, MMCRx, SIAR, SDAR.
1285 */
1286 /* Save non-volatile GPRs */
1287 std r14, VCPU_GPR(r14)(r3)
1288 std r15, VCPU_GPR(r15)(r3)
1289 std r16, VCPU_GPR(r16)(r3)
1290 std r17, VCPU_GPR(r17)(r3)
1291 std r18, VCPU_GPR(r18)(r3)
1292 std r19, VCPU_GPR(r19)(r3)
1293 std r20, VCPU_GPR(r20)(r3)
1294 std r21, VCPU_GPR(r21)(r3)
1295 std r22, VCPU_GPR(r22)(r3)
1296 std r23, VCPU_GPR(r23)(r3)
1297 std r24, VCPU_GPR(r24)(r3)
1298 std r25, VCPU_GPR(r25)(r3)
1299 std r26, VCPU_GPR(r26)(r3)
1300 std r27, VCPU_GPR(r27)(r3)
1301 std r28, VCPU_GPR(r28)(r3)
1302 std r29, VCPU_GPR(r29)(r3)
1303 std r30, VCPU_GPR(r30)(r3)
1304 std r31, VCPU_GPR(r31)(r3)
1305
1306 /* save FP state */
1307 bl .kvmppc_save_fp
1308
1309 /*
1310 * Take a nap until a decrementer or external interrupt occurs,
1311 * with PECE1 (wake on decr) and PECE0 (wake on external) set in LPCR
1312 */
1313 li r0,0x80
1314 stb r0,PACAPROCSTART(r13)
1315 mfspr r5,SPRN_LPCR
1316 ori r5,r5,LPCR_PECE0 | LPCR_PECE1
1317 mtspr SPRN_LPCR,r5
1318 isync
1319 li r0, 0
1320 std r0, HSTATE_SCRATCH0(r13)
1321 ptesync
1322 ld r0, HSTATE_SCRATCH0(r13)
13231: cmpd r0, r0
1324 bne 1b
1325 nap
1326 b .
1327
1328kvm_end_cede:
1329 /* Woken by external or decrementer interrupt */
1330 ld r1, HSTATE_HOST_R1(r13)
1331 ld r2, PACATOC(r13)
1332
1333 /* If we're a secondary thread and we got here by an IPI, ack it */
1334 ld r4,HSTATE_KVM_VCPU(r13)
1335 lwz r3,VCPU_PTID(r4)
1336 cmpwi r3,0
1337 beq 27f
1338 mfspr r3,SPRN_SRR1
1339 rlwinm r3,r3,44-31,0x7 /* extract wake reason field */
1340 cmpwi r3,4 /* was it an external interrupt? */
1341 bne 27f
1342 ld r5, HSTATE_XICS_PHYS(r13)
1343 li r0,0xff
1344 li r6,XICS_QIRR
1345 li r7,XICS_XIRR
1346 lwzcix r8,r5,r7 /* ack the interrupt */
1347 sync
1348 stbcix r0,r5,r6 /* clear it */
1349 stwcix r8,r5,r7 /* EOI it */
135027:
1351 /* load up FP state */
1352 bl kvmppc_load_fp
1353
1354 /* Load NV GPRS */
1355 ld r14, VCPU_GPR(r14)(r4)
1356 ld r15, VCPU_GPR(r15)(r4)
1357 ld r16, VCPU_GPR(r16)(r4)
1358 ld r17, VCPU_GPR(r17)(r4)
1359 ld r18, VCPU_GPR(r18)(r4)
1360 ld r19, VCPU_GPR(r19)(r4)
1361 ld r20, VCPU_GPR(r20)(r4)
1362 ld r21, VCPU_GPR(r21)(r4)
1363 ld r22, VCPU_GPR(r22)(r4)
1364 ld r23, VCPU_GPR(r23)(r4)
1365 ld r24, VCPU_GPR(r24)(r4)
1366 ld r25, VCPU_GPR(r25)(r4)
1367 ld r26, VCPU_GPR(r26)(r4)
1368 ld r27, VCPU_GPR(r27)(r4)
1369 ld r28, VCPU_GPR(r28)(r4)
1370 ld r29, VCPU_GPR(r29)(r4)
1371 ld r30, VCPU_GPR(r30)(r4)
1372 ld r31, VCPU_GPR(r31)(r4)
1373
1374 /* clear our bit in vcore->napping_threads */
137533: ld r5,HSTATE_KVM_VCORE(r13)
1376 lwz r3,VCPU_PTID(r4)
1377 li r0,1
1378 sld r0,r0,r3
1379 addi r6,r5,VCORE_NAPPING_THREADS
138032: lwarx r7,0,r6
1381 andc r7,r7,r0
1382 stwcx. r7,0,r6
1383 bne 32b
1384 li r0,0
1385 stb r0,HSTATE_NAPPING(r13)
1386
1387 /* see if any other thread is already exiting */
1388 lwz r0,VCORE_ENTRY_EXIT(r5)
1389 cmpwi r0,0x100
1390 blt kvmppc_cede_reentry /* if not go back to guest */
1391
1392 /* some threads are exiting, so go to the guest exit path */
1393 b hcall_real_fallback
1394
1395 /* cede when already previously prodded case */
13961: li r0,0
1397 stb r0,VCPU_PRODDED(r3)
1398 sync /* order testing prodded vs. clearing ceded */
1399 stb r0,VCPU_CEDED(r3)
1400 li r3,H_SUCCESS
1401 blr
1402
1403 /* we've ceded but we want to give control to the host */
14042: li r3,H_TOO_HARD
1405 blr
1406
1180secondary_too_late: 1407secondary_too_late:
1181 ld r5,HSTATE_KVM_VCORE(r13) 1408 ld r5,HSTATE_KVM_VCORE(r13)
1182 HMT_LOW 1409 HMT_LOW
@@ -1194,14 +1421,20 @@ secondary_too_late:
1194 slbmte r6,r5 1421 slbmte r6,r5
11951: addi r11,r11,16 14221: addi r11,r11,16
1196 .endr 1423 .endr
1197 b 50f
1198 1424
1199secondary_nap: 1425secondary_nap:
1200 /* Clear any pending IPI */ 1426 /* Clear any pending IPI - assume we're a secondary thread */
120150: ld r5, HSTATE_XICS_PHYS(r13) 1427 ld r5, HSTATE_XICS_PHYS(r13)
1428 li r7, XICS_XIRR
1429 lwzcix r3, r5, r7 /* ack any pending interrupt */
1430 rlwinm. r0, r3, 0, 0xffffff /* any pending? */
1431 beq 37f
1432 sync
1202 li r0, 0xff 1433 li r0, 0xff
1203 li r6, XICS_QIRR 1434 li r6, XICS_QIRR
1204 stbcix r0, r5, r6 1435 stbcix r0, r5, r6 /* clear the IPI */
1436 stwcix r3, r5, r7 /* EOI it */
143737: sync
1205 1438
1206 /* increment the nap count and then go to nap mode */ 1439 /* increment the nap count and then go to nap mode */
1207 ld r4, HSTATE_KVM_VCORE(r13) 1440 ld r4, HSTATE_KVM_VCORE(r13)
@@ -1211,13 +1444,12 @@ secondary_nap:
1211 addi r3, r3, 1 1444 addi r3, r3, 1
1212 stwcx. r3, 0, r4 1445 stwcx. r3, 0, r4
1213 bne 51b 1446 bne 51b
1214 isync
1215 1447
1448 li r3, LPCR_PECE0
1216 mfspr r4, SPRN_LPCR 1449 mfspr r4, SPRN_LPCR
1217 li r0, LPCR_PECE 1450 rlwimi r4, r3, 0, LPCR_PECE0 | LPCR_PECE1
1218 andc r4, r4, r0
1219 ori r4, r4, LPCR_PECE0 /* exit nap on interrupt */
1220 mtspr SPRN_LPCR, r4 1451 mtspr SPRN_LPCR, r4
1452 isync
1221 li r0, 0 1453 li r0, 0
1222 std r0, HSTATE_SCRATCH0(r13) 1454 std r0, HSTATE_SCRATCH0(r13)
1223 ptesync 1455 ptesync
diff --git a/arch/powerpc/kvm/book3s_interrupts.S b/arch/powerpc/kvm/book3s_interrupts.S
index c54b0e30cf3f..0a8515a5c042 100644
--- a/arch/powerpc/kvm/book3s_interrupts.S
+++ b/arch/powerpc/kvm/book3s_interrupts.S
@@ -29,27 +29,11 @@
29#define ULONG_SIZE 8 29#define ULONG_SIZE 8
30#define FUNC(name) GLUE(.,name) 30#define FUNC(name) GLUE(.,name)
31 31
32#define GET_SHADOW_VCPU_R13
33
34#define DISABLE_INTERRUPTS \
35 mfmsr r0; \
36 rldicl r0,r0,48,1; \
37 rotldi r0,r0,16; \
38 mtmsrd r0,1; \
39
40#elif defined(CONFIG_PPC_BOOK3S_32) 32#elif defined(CONFIG_PPC_BOOK3S_32)
41 33
42#define ULONG_SIZE 4 34#define ULONG_SIZE 4
43#define FUNC(name) name 35#define FUNC(name) name
44 36
45#define GET_SHADOW_VCPU_R13 \
46 lwz r13, (THREAD + THREAD_KVM_SVCPU)(r2)
47
48#define DISABLE_INTERRUPTS \
49 mfmsr r0; \
50 rlwinm r0,r0,0,17,15; \
51 mtmsr r0; \
52
53#endif /* CONFIG_PPC_BOOK3S_XX */ 37#endif /* CONFIG_PPC_BOOK3S_XX */
54 38
55 39
@@ -108,44 +92,17 @@ kvm_start_entry:
108 92
109kvm_start_lightweight: 93kvm_start_lightweight:
110 94
111 GET_SHADOW_VCPU_R13
112 PPC_LL r3, VCPU_HIGHMEM_HANDLER(r4)
113 PPC_STL r3, HSTATE_VMHANDLER(r13)
114
115 PPC_LL r10, VCPU_SHADOW_MSR(r4) /* r10 = vcpu->arch.shadow_msr */
116
117 DISABLE_INTERRUPTS
118
119#ifdef CONFIG_PPC_BOOK3S_64 95#ifdef CONFIG_PPC_BOOK3S_64
120 /* Some guests may need to have dcbz set to 32 byte length.
121 *
122 * Usually we ensure that by patching the guest's instructions
123 * to trap on dcbz and emulate it in the hypervisor.
124 *
125 * If we can, we should tell the CPU to use 32 byte dcbz though,
126 * because that's a lot faster.
127 */
128
129 PPC_LL r3, VCPU_HFLAGS(r4) 96 PPC_LL r3, VCPU_HFLAGS(r4)
130 rldicl. r3, r3, 0, 63 /* CR = ((r3 & 1) == 0) */ 97 rldicl r3, r3, 0, 63 /* r3 &= 1 */
131 beq no_dcbz32_on 98 stb r3, HSTATE_RESTORE_HID5(r13)
132
133 mfspr r3,SPRN_HID5
134 ori r3, r3, 0x80 /* XXX HID5_dcbz32 = 0x80 */
135 mtspr SPRN_HID5,r3
136
137no_dcbz32_on:
138
139#endif /* CONFIG_PPC_BOOK3S_64 */ 99#endif /* CONFIG_PPC_BOOK3S_64 */
140 100
141 PPC_LL r6, VCPU_RMCALL(r4) 101 PPC_LL r4, VCPU_SHADOW_MSR(r4) /* get shadow_msr */
142 mtctr r6
143
144 PPC_LL r3, VCPU_TRAMPOLINE_ENTER(r4)
145 LOAD_REG_IMMEDIATE(r4, MSR_KERNEL & ~(MSR_IR | MSR_DR))
146 102
147 /* Jump to segment patching handler and into our guest */ 103 /* Jump to segment patching handler and into our guest */
148 bctr 104 bl FUNC(kvmppc_entry_trampoline)
105 nop
149 106
150/* 107/*
151 * This is the handler in module memory. It gets jumped at from the 108 * This is the handler in module memory. It gets jumped at from the
@@ -170,21 +127,6 @@ kvmppc_handler_highmem:
170 /* R7 = vcpu */ 127 /* R7 = vcpu */
171 PPC_LL r7, GPR4(r1) 128 PPC_LL r7, GPR4(r1)
172 129
173#ifdef CONFIG_PPC_BOOK3S_64
174
175 PPC_LL r5, VCPU_HFLAGS(r7)
176 rldicl. r5, r5, 0, 63 /* CR = ((r5 & 1) == 0) */
177 beq no_dcbz32_off
178
179 li r4, 0
180 mfspr r5,SPRN_HID5
181 rldimi r5,r4,6,56
182 mtspr SPRN_HID5,r5
183
184no_dcbz32_off:
185
186#endif /* CONFIG_PPC_BOOK3S_64 */
187
188 PPC_STL r14, VCPU_GPR(r14)(r7) 130 PPC_STL r14, VCPU_GPR(r14)(r7)
189 PPC_STL r15, VCPU_GPR(r15)(r7) 131 PPC_STL r15, VCPU_GPR(r15)(r7)
190 PPC_STL r16, VCPU_GPR(r16)(r7) 132 PPC_STL r16, VCPU_GPR(r16)(r7)
@@ -204,67 +146,6 @@ no_dcbz32_off:
204 PPC_STL r30, VCPU_GPR(r30)(r7) 146 PPC_STL r30, VCPU_GPR(r30)(r7)
205 PPC_STL r31, VCPU_GPR(r31)(r7) 147 PPC_STL r31, VCPU_GPR(r31)(r7)
206 148
207 /* Restore host msr -> SRR1 */
208 PPC_LL r6, VCPU_HOST_MSR(r7)
209
210 /*
211 * For some interrupts, we need to call the real Linux
212 * handler, so it can do work for us. This has to happen
213 * as if the interrupt arrived from the kernel though,
214 * so let's fake it here where most state is restored.
215 *
216 * Call Linux for hardware interrupts/decrementer
217 * r3 = address of interrupt handler (exit reason)
218 */
219
220 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
221 beq call_linux_handler
222 cmpwi r12, BOOK3S_INTERRUPT_DECREMENTER
223 beq call_linux_handler
224 cmpwi r12, BOOK3S_INTERRUPT_PERFMON
225 beq call_linux_handler
226
227 /* Back to EE=1 */
228 mtmsr r6
229 sync
230 b kvm_return_point
231
232call_linux_handler:
233
234 /*
235 * If we land here we need to jump back to the handler we
236 * came from.
237 *
238 * We have a page that we can access from real mode, so let's
239 * jump back to that and use it as a trampoline to get back into the
240 * interrupt handler!
241 *
242 * R3 still contains the exit code,
243 * R5 VCPU_HOST_RETIP and
244 * R6 VCPU_HOST_MSR
245 */
246
247 /* Restore host IP -> SRR0 */
248 PPC_LL r5, VCPU_HOST_RETIP(r7)
249
250 /* XXX Better move to a safe function?
251 * What if we get an HTAB flush in between mtsrr0 and mtsrr1? */
252
253 mtlr r12
254
255 PPC_LL r4, VCPU_TRAMPOLINE_LOWMEM(r7)
256 mtsrr0 r4
257 LOAD_REG_IMMEDIATE(r3, MSR_KERNEL & ~(MSR_IR | MSR_DR))
258 mtsrr1 r3
259
260 RFI
261
262.global kvm_return_point
263kvm_return_point:
264
265 /* Jump back to lightweight entry if we're supposed to */
266 /* go back into the guest */
267
268 /* Pass the exit number as 3rd argument to kvmppc_handle_exit */ 149 /* Pass the exit number as 3rd argument to kvmppc_handle_exit */
269 mr r5, r12 150 mr r5, r12
270 151
diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
index 0c0d3f274437..d417511abfb1 100644
--- a/arch/powerpc/kvm/book3s_pr.c
+++ b/arch/powerpc/kvm/book3s_pr.c
@@ -150,16 +150,22 @@ void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
150#ifdef CONFIG_PPC_BOOK3S_64 150#ifdef CONFIG_PPC_BOOK3S_64
151 if ((pvr >= 0x330000) && (pvr < 0x70330000)) { 151 if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
152 kvmppc_mmu_book3s_64_init(vcpu); 152 kvmppc_mmu_book3s_64_init(vcpu);
153 to_book3s(vcpu)->hior = 0xfff00000; 153 if (!to_book3s(vcpu)->hior_sregs)
154 to_book3s(vcpu)->hior = 0xfff00000;
154 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL; 155 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
156 vcpu->arch.cpu_type = KVM_CPU_3S_64;
155 } else 157 } else
156#endif 158#endif
157 { 159 {
158 kvmppc_mmu_book3s_32_init(vcpu); 160 kvmppc_mmu_book3s_32_init(vcpu);
159 to_book3s(vcpu)->hior = 0; 161 if (!to_book3s(vcpu)->hior_sregs)
162 to_book3s(vcpu)->hior = 0;
160 to_book3s(vcpu)->msr_mask = 0xffffffffULL; 163 to_book3s(vcpu)->msr_mask = 0xffffffffULL;
164 vcpu->arch.cpu_type = KVM_CPU_3S_32;
161 } 165 }
162 166
167 kvmppc_sanity_check(vcpu);
168
163 /* If we are in hypervisor level on 970, we can tell the CPU to 169 /* If we are in hypervisor level on 970, we can tell the CPU to
164 * treat DCBZ as 32 bytes store */ 170 * treat DCBZ as 32 bytes store */
165 vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32; 171 vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
@@ -646,7 +652,27 @@ program_interrupt:
646 break; 652 break;
647 } 653 }
648 case BOOK3S_INTERRUPT_SYSCALL: 654 case BOOK3S_INTERRUPT_SYSCALL:
649 if (vcpu->arch.osi_enabled && 655 if (vcpu->arch.papr_enabled &&
656 (kvmppc_get_last_inst(vcpu) == 0x44000022) &&
657 !(vcpu->arch.shared->msr & MSR_PR)) {
658 /* SC 1 papr hypercalls */
659 ulong cmd = kvmppc_get_gpr(vcpu, 3);
660 int i;
661
662 if (kvmppc_h_pr(vcpu, cmd) == EMULATE_DONE) {
663 r = RESUME_GUEST;
664 break;
665 }
666
667 run->papr_hcall.nr = cmd;
668 for (i = 0; i < 9; ++i) {
669 ulong gpr = kvmppc_get_gpr(vcpu, 4 + i);
670 run->papr_hcall.args[i] = gpr;
671 }
672 run->exit_reason = KVM_EXIT_PAPR_HCALL;
673 vcpu->arch.hcall_needed = 1;
674 r = RESUME_HOST;
675 } else if (vcpu->arch.osi_enabled &&
650 (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) && 676 (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
651 (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) { 677 (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
652 /* MOL hypercalls */ 678 /* MOL hypercalls */
@@ -770,6 +796,9 @@ int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
770 } 796 }
771 } 797 }
772 798
799 if (sregs->u.s.flags & KVM_SREGS_S_HIOR)
800 sregs->u.s.hior = to_book3s(vcpu)->hior;
801
773 return 0; 802 return 0;
774} 803}
775 804
@@ -806,6 +835,11 @@ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
806 /* Flush the MMU after messing with the segments */ 835 /* Flush the MMU after messing with the segments */
807 kvmppc_mmu_pte_flush(vcpu, 0, 0); 836 kvmppc_mmu_pte_flush(vcpu, 0, 0);
808 837
838 if (sregs->u.s.flags & KVM_SREGS_S_HIOR) {
839 to_book3s(vcpu)->hior_sregs = true;
840 to_book3s(vcpu)->hior = sregs->u.s.hior;
841 }
842
809 return 0; 843 return 0;
810} 844}
811 845
@@ -841,8 +875,6 @@ struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
841 if (!p) 875 if (!p)
842 goto uninit_vcpu; 876 goto uninit_vcpu;
843 877
844 vcpu->arch.host_retip = kvm_return_point;
845 vcpu->arch.host_msr = mfmsr();
846#ifdef CONFIG_PPC_BOOK3S_64 878#ifdef CONFIG_PPC_BOOK3S_64
847 /* default to book3s_64 (970fx) */ 879 /* default to book3s_64 (970fx) */
848 vcpu->arch.pvr = 0x3C0301; 880 vcpu->arch.pvr = 0x3C0301;
@@ -853,16 +885,6 @@ struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
853 kvmppc_set_pvr(vcpu, vcpu->arch.pvr); 885 kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
854 vcpu->arch.slb_nr = 64; 886 vcpu->arch.slb_nr = 64;
855 887
856 /* remember where some real-mode handlers are */
857 vcpu->arch.trampoline_lowmem = __pa(kvmppc_handler_lowmem_trampoline);
858 vcpu->arch.trampoline_enter = __pa(kvmppc_handler_trampoline_enter);
859 vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
860#ifdef CONFIG_PPC_BOOK3S_64
861 vcpu->arch.rmcall = *(ulong*)kvmppc_rmcall;
862#else
863 vcpu->arch.rmcall = (ulong)kvmppc_rmcall;
864#endif
865
866 vcpu->arch.shadow_msr = MSR_USER64; 888 vcpu->arch.shadow_msr = MSR_USER64;
867 889
868 err = kvmppc_mmu_init(vcpu); 890 err = kvmppc_mmu_init(vcpu);
@@ -908,6 +930,12 @@ int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
908#endif 930#endif
909 ulong ext_msr; 931 ulong ext_msr;
910 932
933 /* Check if we can run the vcpu at all */
934 if (!vcpu->arch.sane) {
935 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
936 return -EINVAL;
937 }
938
911 /* No need to go into the guest when all we do is going out */ 939 /* No need to go into the guest when all we do is going out */
912 if (signal_pending(current)) { 940 if (signal_pending(current)) {
913 kvm_run->exit_reason = KVM_EXIT_INTR; 941 kvm_run->exit_reason = KVM_EXIT_INTR;
diff --git a/arch/powerpc/kvm/book3s_pr_papr.c b/arch/powerpc/kvm/book3s_pr_papr.c
new file mode 100644
index 000000000000..b9589324797b
--- /dev/null
+++ b/arch/powerpc/kvm/book3s_pr_papr.c
@@ -0,0 +1,158 @@
1/*
2 * Copyright (C) 2011. Freescale Inc. All rights reserved.
3 *
4 * Authors:
5 * Alexander Graf <agraf@suse.de>
6 * Paul Mackerras <paulus@samba.org>
7 *
8 * Description:
9 *
10 * Hypercall handling for running PAPR guests in PR KVM on Book 3S
11 * processors.
12 *
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License, version 2, as
15 * published by the Free Software Foundation.
16 */
17
18#include <asm/uaccess.h>
19#include <asm/kvm_ppc.h>
20#include <asm/kvm_book3s.h>
21
22static unsigned long get_pteg_addr(struct kvm_vcpu *vcpu, long pte_index)
23{
24 struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
25 unsigned long pteg_addr;
26
27 pte_index <<= 4;
28 pte_index &= ((1 << ((vcpu_book3s->sdr1 & 0x1f) + 11)) - 1) << 7 | 0x70;
29 pteg_addr = vcpu_book3s->sdr1 & 0xfffffffffffc0000ULL;
30 pteg_addr |= pte_index;
31
32 return pteg_addr;
33}
34
35static int kvmppc_h_pr_enter(struct kvm_vcpu *vcpu)
36{
37 long flags = kvmppc_get_gpr(vcpu, 4);
38 long pte_index = kvmppc_get_gpr(vcpu, 5);
39 unsigned long pteg[2 * 8];
40 unsigned long pteg_addr, i, *hpte;
41
42 pte_index &= ~7UL;
43 pteg_addr = get_pteg_addr(vcpu, pte_index);
44
45 copy_from_user(pteg, (void __user *)pteg_addr, sizeof(pteg));
46 hpte = pteg;
47
48 if (likely((flags & H_EXACT) == 0)) {
49 pte_index &= ~7UL;
50 for (i = 0; ; ++i) {
51 if (i == 8)
52 return H_PTEG_FULL;
53 if ((*hpte & HPTE_V_VALID) == 0)
54 break;
55 hpte += 2;
56 }
57 } else {
58 i = kvmppc_get_gpr(vcpu, 5) & 7UL;
59 hpte += i * 2;
60 }
61
62 hpte[0] = kvmppc_get_gpr(vcpu, 6);
63 hpte[1] = kvmppc_get_gpr(vcpu, 7);
64 copy_to_user((void __user *)pteg_addr, pteg, sizeof(pteg));
65 kvmppc_set_gpr(vcpu, 3, H_SUCCESS);
66 kvmppc_set_gpr(vcpu, 4, pte_index | i);
67
68 return EMULATE_DONE;
69}
70
71static int kvmppc_h_pr_remove(struct kvm_vcpu *vcpu)
72{
73 unsigned long flags= kvmppc_get_gpr(vcpu, 4);
74 unsigned long pte_index = kvmppc_get_gpr(vcpu, 5);
75 unsigned long avpn = kvmppc_get_gpr(vcpu, 6);
76 unsigned long v = 0, pteg, rb;
77 unsigned long pte[2];
78
79 pteg = get_pteg_addr(vcpu, pte_index);
80 copy_from_user(pte, (void __user *)pteg, sizeof(pte));
81
82 if ((pte[0] & HPTE_V_VALID) == 0 ||
83 ((flags & H_AVPN) && (pte[0] & ~0x7fUL) != avpn) ||
84 ((flags & H_ANDCOND) && (pte[0] & avpn) != 0)) {
85 kvmppc_set_gpr(vcpu, 3, H_NOT_FOUND);
86 return EMULATE_DONE;
87 }
88
89 copy_to_user((void __user *)pteg, &v, sizeof(v));
90
91 rb = compute_tlbie_rb(pte[0], pte[1], pte_index);
92 vcpu->arch.mmu.tlbie(vcpu, rb, rb & 1 ? true : false);
93
94 kvmppc_set_gpr(vcpu, 3, H_SUCCESS);
95 kvmppc_set_gpr(vcpu, 4, pte[0]);
96 kvmppc_set_gpr(vcpu, 5, pte[1]);
97
98 return EMULATE_DONE;
99}
100
101static int kvmppc_h_pr_protect(struct kvm_vcpu *vcpu)
102{
103 unsigned long flags = kvmppc_get_gpr(vcpu, 4);
104 unsigned long pte_index = kvmppc_get_gpr(vcpu, 5);
105 unsigned long avpn = kvmppc_get_gpr(vcpu, 6);
106 unsigned long rb, pteg, r, v;
107 unsigned long pte[2];
108
109 pteg = get_pteg_addr(vcpu, pte_index);
110 copy_from_user(pte, (void __user *)pteg, sizeof(pte));
111
112 if ((pte[0] & HPTE_V_VALID) == 0 ||
113 ((flags & H_AVPN) && (pte[0] & ~0x7fUL) != avpn)) {
114 kvmppc_set_gpr(vcpu, 3, H_NOT_FOUND);
115 return EMULATE_DONE;
116 }
117
118 v = pte[0];
119 r = pte[1];
120 r &= ~(HPTE_R_PP0 | HPTE_R_PP | HPTE_R_N | HPTE_R_KEY_HI |
121 HPTE_R_KEY_LO);
122 r |= (flags << 55) & HPTE_R_PP0;
123 r |= (flags << 48) & HPTE_R_KEY_HI;
124 r |= flags & (HPTE_R_PP | HPTE_R_N | HPTE_R_KEY_LO);
125
126 pte[1] = r;
127
128 rb = compute_tlbie_rb(v, r, pte_index);
129 vcpu->arch.mmu.tlbie(vcpu, rb, rb & 1 ? true : false);
130 copy_to_user((void __user *)pteg, pte, sizeof(pte));
131
132 kvmppc_set_gpr(vcpu, 3, H_SUCCESS);
133
134 return EMULATE_DONE;
135}
136
137int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd)
138{
139 switch (cmd) {
140 case H_ENTER:
141 return kvmppc_h_pr_enter(vcpu);
142 case H_REMOVE:
143 return kvmppc_h_pr_remove(vcpu);
144 case H_PROTECT:
145 return kvmppc_h_pr_protect(vcpu);
146 case H_BULK_REMOVE:
147 /* We just flush all PTEs, so user space can
148 handle the HPT modifications */
149 kvmppc_mmu_pte_flush(vcpu, 0, 0);
150 break;
151 case H_CEDE:
152 kvm_vcpu_block(vcpu);
153 vcpu->stat.halt_wakeup++;
154 return EMULATE_DONE;
155 }
156
157 return EMULATE_FAIL;
158}
diff --git a/arch/powerpc/kvm/book3s_rmhandlers.S b/arch/powerpc/kvm/book3s_rmhandlers.S
index c1f877c4a884..34187585c507 100644
--- a/arch/powerpc/kvm/book3s_rmhandlers.S
+++ b/arch/powerpc/kvm/book3s_rmhandlers.S
@@ -20,6 +20,7 @@
20#include <asm/ppc_asm.h> 20#include <asm/ppc_asm.h>
21#include <asm/kvm_asm.h> 21#include <asm/kvm_asm.h>
22#include <asm/reg.h> 22#include <asm/reg.h>
23#include <asm/mmu.h>
23#include <asm/page.h> 24#include <asm/page.h>
24#include <asm/asm-offsets.h> 25#include <asm/asm-offsets.h>
25 26
@@ -35,10 +36,10 @@
35 36
36#if defined(CONFIG_PPC_BOOK3S_64) 37#if defined(CONFIG_PPC_BOOK3S_64)
37 38
38#define LOAD_SHADOW_VCPU(reg) GET_PACA(reg)
39#define MSR_NOIRQ MSR_KERNEL & ~(MSR_IR | MSR_DR)
40#define FUNC(name) GLUE(.,name) 39#define FUNC(name) GLUE(.,name)
40#define MTMSR_EERI(reg) mtmsrd (reg),1
41 41
42 .globl kvmppc_skip_interrupt
42kvmppc_skip_interrupt: 43kvmppc_skip_interrupt:
43 /* 44 /*
44 * Here all GPRs are unchanged from when the interrupt happened 45 * Here all GPRs are unchanged from when the interrupt happened
@@ -51,6 +52,7 @@ kvmppc_skip_interrupt:
51 rfid 52 rfid
52 b . 53 b .
53 54
55 .globl kvmppc_skip_Hinterrupt
54kvmppc_skip_Hinterrupt: 56kvmppc_skip_Hinterrupt:
55 /* 57 /*
56 * Here all GPRs are unchanged from when the interrupt happened 58 * Here all GPRs are unchanged from when the interrupt happened
@@ -65,8 +67,8 @@ kvmppc_skip_Hinterrupt:
65 67
66#elif defined(CONFIG_PPC_BOOK3S_32) 68#elif defined(CONFIG_PPC_BOOK3S_32)
67 69
68#define MSR_NOIRQ MSR_KERNEL
69#define FUNC(name) name 70#define FUNC(name) name
71#define MTMSR_EERI(reg) mtmsr (reg)
70 72
71.macro INTERRUPT_TRAMPOLINE intno 73.macro INTERRUPT_TRAMPOLINE intno
72 74
@@ -167,40 +169,24 @@ kvmppc_handler_skip_ins:
167#endif 169#endif
168 170
169/* 171/*
170 * This trampoline brings us back to a real mode handler 172 * Call kvmppc_handler_trampoline_enter in real mode
171 *
172 * Input Registers:
173 *
174 * R5 = SRR0
175 * R6 = SRR1
176 * LR = real-mode IP
177 * 173 *
174 * On entry, r4 contains the guest shadow MSR
178 */ 175 */
179.global kvmppc_handler_lowmem_trampoline 176_GLOBAL(kvmppc_entry_trampoline)
180kvmppc_handler_lowmem_trampoline: 177 mfmsr r5
181 178 LOAD_REG_ADDR(r7, kvmppc_handler_trampoline_enter)
182 mtsrr0 r5 179 toreal(r7)
180
181 li r9, MSR_RI
182 ori r9, r9, MSR_EE
183 andc r9, r5, r9 /* Clear EE and RI in MSR value */
184 li r6, MSR_IR | MSR_DR
185 ori r6, r6, MSR_EE
186 andc r6, r5, r6 /* Clear EE, DR and IR in MSR value */
187 MTMSR_EERI(r9) /* Clear EE and RI in MSR */
188 mtsrr0 r7 /* before we set srr0/1 */
183 mtsrr1 r6 189 mtsrr1 r6
184 blr
185kvmppc_handler_lowmem_trampoline_end:
186
187/*
188 * Call a function in real mode
189 *
190 * Input Registers:
191 *
192 * R3 = function
193 * R4 = MSR
194 * R5 = scratch register
195 *
196 */
197_GLOBAL(kvmppc_rmcall)
198 LOAD_REG_IMMEDIATE(r5, MSR_NOIRQ)
199 mtmsr r5 /* Disable relocation and interrupts, so mtsrr
200 doesn't get interrupted */
201 sync
202 mtsrr0 r3
203 mtsrr1 r4
204 RFI 190 RFI
205 191
206#if defined(CONFIG_PPC_BOOK3S_32) 192#if defined(CONFIG_PPC_BOOK3S_32)
diff --git a/arch/powerpc/kvm/book3s_segment.S b/arch/powerpc/kvm/book3s_segment.S
index aed32e517212..0676ae249b9f 100644
--- a/arch/powerpc/kvm/book3s_segment.S
+++ b/arch/powerpc/kvm/book3s_segment.S
@@ -23,6 +23,7 @@
23 23
24#define GET_SHADOW_VCPU(reg) \ 24#define GET_SHADOW_VCPU(reg) \
25 mr reg, r13 25 mr reg, r13
26#define MTMSR_EERI(reg) mtmsrd (reg),1
26 27
27#elif defined(CONFIG_PPC_BOOK3S_32) 28#elif defined(CONFIG_PPC_BOOK3S_32)
28 29
@@ -30,6 +31,7 @@
30 tophys(reg, r2); \ 31 tophys(reg, r2); \
31 lwz reg, (THREAD + THREAD_KVM_SVCPU)(reg); \ 32 lwz reg, (THREAD + THREAD_KVM_SVCPU)(reg); \
32 tophys(reg, reg) 33 tophys(reg, reg)
34#define MTMSR_EERI(reg) mtmsr (reg)
33 35
34#endif 36#endif
35 37
@@ -57,10 +59,12 @@ kvmppc_handler_trampoline_enter:
57 /* Required state: 59 /* Required state:
58 * 60 *
59 * MSR = ~IR|DR 61 * MSR = ~IR|DR
60 * R13 = PACA
61 * R1 = host R1 62 * R1 = host R1
62 * R2 = host R2 63 * R2 = host R2
63 * R10 = guest MSR 64 * R4 = guest shadow MSR
65 * R5 = normal host MSR
66 * R6 = current host MSR (EE, IR, DR off)
67 * LR = highmem guest exit code
64 * all other volatile GPRS = free 68 * all other volatile GPRS = free
65 * SVCPU[CR] = guest CR 69 * SVCPU[CR] = guest CR
66 * SVCPU[XER] = guest XER 70 * SVCPU[XER] = guest XER
@@ -71,15 +75,15 @@ kvmppc_handler_trampoline_enter:
71 /* r3 = shadow vcpu */ 75 /* r3 = shadow vcpu */
72 GET_SHADOW_VCPU(r3) 76 GET_SHADOW_VCPU(r3)
73 77
78 /* Save guest exit handler address and MSR */
79 mflr r0
80 PPC_STL r0, HSTATE_VMHANDLER(r3)
81 PPC_STL r5, HSTATE_HOST_MSR(r3)
82
74 /* Save R1/R2 in the PACA (64-bit) or shadow_vcpu (32-bit) */ 83 /* Save R1/R2 in the PACA (64-bit) or shadow_vcpu (32-bit) */
75 PPC_STL r1, HSTATE_HOST_R1(r3) 84 PPC_STL r1, HSTATE_HOST_R1(r3)
76 PPC_STL r2, HSTATE_HOST_R2(r3) 85 PPC_STL r2, HSTATE_HOST_R2(r3)
77 86
78 /* Move SRR0 and SRR1 into the respective regs */
79 PPC_LL r9, SVCPU_PC(r3)
80 mtsrr0 r9
81 mtsrr1 r10
82
83 /* Activate guest mode, so faults get handled by KVM */ 87 /* Activate guest mode, so faults get handled by KVM */
84 li r11, KVM_GUEST_MODE_GUEST 88 li r11, KVM_GUEST_MODE_GUEST
85 stb r11, HSTATE_IN_GUEST(r3) 89 stb r11, HSTATE_IN_GUEST(r3)
@@ -87,17 +91,46 @@ kvmppc_handler_trampoline_enter:
87 /* Switch to guest segment. This is subarch specific. */ 91 /* Switch to guest segment. This is subarch specific. */
88 LOAD_GUEST_SEGMENTS 92 LOAD_GUEST_SEGMENTS
89 93
94#ifdef CONFIG_PPC_BOOK3S_64
95 /* Some guests may need to have dcbz set to 32 byte length.
96 *
97 * Usually we ensure that by patching the guest's instructions
98 * to trap on dcbz and emulate it in the hypervisor.
99 *
100 * If we can, we should tell the CPU to use 32 byte dcbz though,
101 * because that's a lot faster.
102 */
103 lbz r0, HSTATE_RESTORE_HID5(r3)
104 cmpwi r0, 0
105 beq no_dcbz32_on
106
107 mfspr r0,SPRN_HID5
108 ori r0, r0, 0x80 /* XXX HID5_dcbz32 = 0x80 */
109 mtspr SPRN_HID5,r0
110no_dcbz32_on:
111
112#endif /* CONFIG_PPC_BOOK3S_64 */
113
90 /* Enter guest */ 114 /* Enter guest */
91 115
92 PPC_LL r4, SVCPU_CTR(r3) 116 PPC_LL r8, SVCPU_CTR(r3)
93 PPC_LL r5, SVCPU_LR(r3) 117 PPC_LL r9, SVCPU_LR(r3)
94 lwz r6, SVCPU_CR(r3) 118 lwz r10, SVCPU_CR(r3)
95 lwz r7, SVCPU_XER(r3) 119 lwz r11, SVCPU_XER(r3)
120
121 mtctr r8
122 mtlr r9
123 mtcr r10
124 mtxer r11
96 125
97 mtctr r4 126 /* Move SRR0 and SRR1 into the respective regs */
98 mtlr r5 127 PPC_LL r9, SVCPU_PC(r3)
99 mtcr r6 128 /* First clear RI in our current MSR value */
100 mtxer r7 129 li r0, MSR_RI
130 andc r6, r6, r0
131 MTMSR_EERI(r6)
132 mtsrr0 r9
133 mtsrr1 r4
101 134
102 PPC_LL r0, SVCPU_R0(r3) 135 PPC_LL r0, SVCPU_R0(r3)
103 PPC_LL r1, SVCPU_R1(r3) 136 PPC_LL r1, SVCPU_R1(r3)
@@ -213,11 +246,16 @@ END_FTR_SECTION_IFSET(CPU_FTR_HVMODE)
213 beq ld_last_inst 246 beq ld_last_inst
214 cmpwi r12, BOOK3S_INTERRUPT_PROGRAM 247 cmpwi r12, BOOK3S_INTERRUPT_PROGRAM
215 beq ld_last_inst 248 beq ld_last_inst
249 cmpwi r12, BOOK3S_INTERRUPT_SYSCALL
250 beq ld_last_prev_inst
216 cmpwi r12, BOOK3S_INTERRUPT_ALIGNMENT 251 cmpwi r12, BOOK3S_INTERRUPT_ALIGNMENT
217 beq- ld_last_inst 252 beq- ld_last_inst
218 253
219 b no_ld_last_inst 254 b no_ld_last_inst
220 255
256ld_last_prev_inst:
257 addi r3, r3, -4
258
221ld_last_inst: 259ld_last_inst:
222 /* Save off the guest instruction we're at */ 260 /* Save off the guest instruction we're at */
223 261
@@ -254,6 +292,43 @@ no_ld_last_inst:
254 /* Switch back to host MMU */ 292 /* Switch back to host MMU */
255 LOAD_HOST_SEGMENTS 293 LOAD_HOST_SEGMENTS
256 294
295#ifdef CONFIG_PPC_BOOK3S_64
296
297 lbz r5, HSTATE_RESTORE_HID5(r13)
298 cmpwi r5, 0
299 beq no_dcbz32_off
300
301 li r4, 0
302 mfspr r5,SPRN_HID5
303 rldimi r5,r4,6,56
304 mtspr SPRN_HID5,r5
305
306no_dcbz32_off:
307
308#endif /* CONFIG_PPC_BOOK3S_64 */
309
310 /*
311 * For some interrupts, we need to call the real Linux
312 * handler, so it can do work for us. This has to happen
313 * as if the interrupt arrived from the kernel though,
314 * so let's fake it here where most state is restored.
315 *
316 * Having set up SRR0/1 with the address where we want
317 * to continue with relocation on (potentially in module
318 * space), we either just go straight there with rfi[d],
319 * or we jump to an interrupt handler with bctr if there
320 * is an interrupt to be handled first. In the latter
321 * case, the rfi[d] at the end of the interrupt handler
322 * will get us back to where we want to continue.
323 */
324
325 cmpwi r12, BOOK3S_INTERRUPT_EXTERNAL
326 beq 1f
327 cmpwi r12, BOOK3S_INTERRUPT_DECREMENTER
328 beq 1f
329 cmpwi r12, BOOK3S_INTERRUPT_PERFMON
3301: mtctr r12
331
257 /* Register usage at this point: 332 /* Register usage at this point:
258 * 333 *
259 * R1 = host R1 334 * R1 = host R1
@@ -264,13 +339,15 @@ no_ld_last_inst:
264 * 339 *
265 */ 340 */
266 341
267 /* RFI into the highmem handler */ 342 PPC_LL r6, HSTATE_HOST_MSR(r13)
268 mfmsr r7
269 ori r7, r7, MSR_IR|MSR_DR|MSR_RI|MSR_ME /* Enable paging */
270 mtsrr1 r7
271 /* Load highmem handler address */
272 PPC_LL r8, HSTATE_VMHANDLER(r13) 343 PPC_LL r8, HSTATE_VMHANDLER(r13)
344
345 /* Restore host msr -> SRR1 */
346 mtsrr1 r6
347 /* Load highmem handler address */
273 mtsrr0 r8 348 mtsrr0 r8
274 349
350 /* RFI into the highmem handler, or jump to interrupt handler */
351 beqctr
275 RFI 352 RFI
276kvmppc_handler_trampoline_exit_end: 353kvmppc_handler_trampoline_exit_end:
diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
index ee45fa01220e..bb6c988f010a 100644
--- a/arch/powerpc/kvm/booke.c
+++ b/arch/powerpc/kvm/booke.c
@@ -316,6 +316,11 @@ int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
316{ 316{
317 int ret; 317 int ret;
318 318
319 if (!vcpu->arch.sane) {
320 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
321 return -EINVAL;
322 }
323
319 local_irq_disable(); 324 local_irq_disable();
320 kvm_guest_enter(); 325 kvm_guest_enter();
321 ret = __kvmppc_vcpu_run(kvm_run, vcpu); 326 ret = __kvmppc_vcpu_run(kvm_run, vcpu);
@@ -618,6 +623,7 @@ int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
618int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu) 623int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
619{ 624{
620 int i; 625 int i;
626 int r;
621 627
622 vcpu->arch.pc = 0; 628 vcpu->arch.pc = 0;
623 vcpu->arch.shared->msr = 0; 629 vcpu->arch.shared->msr = 0;
@@ -634,7 +640,9 @@ int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
634 640
635 kvmppc_init_timing_stats(vcpu); 641 kvmppc_init_timing_stats(vcpu);
636 642
637 return kvmppc_core_vcpu_setup(vcpu); 643 r = kvmppc_core_vcpu_setup(vcpu);
644 kvmppc_sanity_check(vcpu);
645 return r;
638} 646}
639 647
640int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) 648int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
diff --git a/arch/powerpc/kvm/e500.c b/arch/powerpc/kvm/e500.c
index 797a7447c268..26d20903f2bc 100644
--- a/arch/powerpc/kvm/e500.c
+++ b/arch/powerpc/kvm/e500.c
@@ -73,6 +73,8 @@ int kvmppc_core_vcpu_setup(struct kvm_vcpu *vcpu)
73 /* Since booke kvm only support one core, update all vcpus' PIR to 0 */ 73 /* Since booke kvm only support one core, update all vcpus' PIR to 0 */
74 vcpu->vcpu_id = 0; 74 vcpu->vcpu_id = 0;
75 75
76 vcpu->arch.cpu_type = KVM_CPU_E500V2;
77
76 return 0; 78 return 0;
77} 79}
78 80
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index a107c9be0fb1..0d843c6ba315 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -39,12 +39,8 @@
39 39
40int kvm_arch_vcpu_runnable(struct kvm_vcpu *v) 40int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
41{ 41{
42#ifndef CONFIG_KVM_BOOK3S_64_HV
43 return !(v->arch.shared->msr & MSR_WE) || 42 return !(v->arch.shared->msr & MSR_WE) ||
44 !!(v->arch.pending_exceptions); 43 !!(v->arch.pending_exceptions);
45#else
46 return !(v->arch.ceded) || !!(v->arch.pending_exceptions);
47#endif
48} 44}
49 45
50int kvmppc_kvm_pv(struct kvm_vcpu *vcpu) 46int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
@@ -95,6 +91,31 @@ int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
95 return r; 91 return r;
96} 92}
97 93
94int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
95{
96 int r = false;
97
98 /* We have to know what CPU to virtualize */
99 if (!vcpu->arch.pvr)
100 goto out;
101
102 /* PAPR only works with book3s_64 */
103 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
104 goto out;
105
106#ifdef CONFIG_KVM_BOOK3S_64_HV
107 /* HV KVM can only do PAPR mode for now */
108 if (!vcpu->arch.papr_enabled)
109 goto out;
110#endif
111
112 r = true;
113
114out:
115 vcpu->arch.sane = r;
116 return r ? 0 : -EINVAL;
117}
118
98int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu) 119int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
99{ 120{
100 enum emulation_result er; 121 enum emulation_result er;
@@ -188,6 +209,8 @@ int kvm_dev_ioctl_check_extension(long ext)
188 case KVM_CAP_PPC_BOOKE_SREGS: 209 case KVM_CAP_PPC_BOOKE_SREGS:
189#else 210#else
190 case KVM_CAP_PPC_SEGSTATE: 211 case KVM_CAP_PPC_SEGSTATE:
212 case KVM_CAP_PPC_HIOR:
213 case KVM_CAP_PPC_PAPR:
191#endif 214#endif
192 case KVM_CAP_PPC_UNSET_IRQ: 215 case KVM_CAP_PPC_UNSET_IRQ:
193 case KVM_CAP_PPC_IRQ_LEVEL: 216 case KVM_CAP_PPC_IRQ_LEVEL:
@@ -258,6 +281,7 @@ struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
258{ 281{
259 struct kvm_vcpu *vcpu; 282 struct kvm_vcpu *vcpu;
260 vcpu = kvmppc_core_vcpu_create(kvm, id); 283 vcpu = kvmppc_core_vcpu_create(kvm, id);
284 vcpu->arch.wqp = &vcpu->wq;
261 if (!IS_ERR(vcpu)) 285 if (!IS_ERR(vcpu))
262 kvmppc_create_vcpu_debugfs(vcpu, id); 286 kvmppc_create_vcpu_debugfs(vcpu, id);
263 return vcpu; 287 return vcpu;
@@ -289,8 +313,8 @@ static void kvmppc_decrementer_func(unsigned long data)
289 313
290 kvmppc_core_queue_dec(vcpu); 314 kvmppc_core_queue_dec(vcpu);
291 315
292 if (waitqueue_active(&vcpu->wq)) { 316 if (waitqueue_active(vcpu->arch.wqp)) {
293 wake_up_interruptible(&vcpu->wq); 317 wake_up_interruptible(vcpu->arch.wqp);
294 vcpu->stat.halt_wakeup++; 318 vcpu->stat.halt_wakeup++;
295 } 319 }
296} 320}
@@ -543,13 +567,15 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
543 567
544int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq) 568int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
545{ 569{
546 if (irq->irq == KVM_INTERRUPT_UNSET) 570 if (irq->irq == KVM_INTERRUPT_UNSET) {
547 kvmppc_core_dequeue_external(vcpu, irq); 571 kvmppc_core_dequeue_external(vcpu, irq);
548 else 572 return 0;
549 kvmppc_core_queue_external(vcpu, irq); 573 }
574
575 kvmppc_core_queue_external(vcpu, irq);
550 576
551 if (waitqueue_active(&vcpu->wq)) { 577 if (waitqueue_active(vcpu->arch.wqp)) {
552 wake_up_interruptible(&vcpu->wq); 578 wake_up_interruptible(vcpu->arch.wqp);
553 vcpu->stat.halt_wakeup++; 579 vcpu->stat.halt_wakeup++;
554 } else if (vcpu->cpu != -1) { 580 } else if (vcpu->cpu != -1) {
555 smp_send_reschedule(vcpu->cpu); 581 smp_send_reschedule(vcpu->cpu);
@@ -571,11 +597,18 @@ static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
571 r = 0; 597 r = 0;
572 vcpu->arch.osi_enabled = true; 598 vcpu->arch.osi_enabled = true;
573 break; 599 break;
600 case KVM_CAP_PPC_PAPR:
601 r = 0;
602 vcpu->arch.papr_enabled = true;
603 break;
574 default: 604 default:
575 r = -EINVAL; 605 r = -EINVAL;
576 break; 606 break;
577 } 607 }
578 608
609 if (!r)
610 r = kvmppc_sanity_check(vcpu);
611
579 return r; 612 return r;
580} 613}
581 614
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-26 23:32:56 -0500