2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License, version 2, as
4 * published by the Free Software Foundation.
6 * This program is distributed in the hope that it will be useful,
7 * but WITHOUT ANY WARRANTY; without even the implied warranty of
8 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9 * GNU General Public License for more details.
11 * You should have received a copy of the GNU General Public License
12 * along with this program; if not, write to the Free Software
13 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
15 * Copyright IBM Corp. 2007
16 * Copyright 2010-2011 Freescale Semiconductor, Inc.
18 * Authors: Hollis Blanchard <hollisb@us.ibm.com>
19 * Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
20 * Scott Wood <scottwood@freescale.com>
21 * Varun Sethi <varun.sethi@freescale.com>
24 #include <linux/errno.h>
25 #include <linux/err.h>
26 #include <linux/kvm_host.h>
27 #include <linux/gfp.h>
28 #include <linux/module.h>
29 #include <linux/vmalloc.h>
32 #include <asm/cputable.h>
33 #include <asm/uaccess.h>
34 #include <asm/kvm_ppc.h>
35 #include <asm/cacheflush.h>
36 #include <asm/dbell.h>
37 #include <asm/hw_irq.h>
43 unsigned long kvmppc_booke_handlers;
45 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
46 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
48 struct kvm_stats_debugfs_item debugfs_entries[] = {
49 { "mmio", VCPU_STAT(mmio_exits) },
50 { "dcr", VCPU_STAT(dcr_exits) },
51 { "sig", VCPU_STAT(signal_exits) },
52 { "itlb_r", VCPU_STAT(itlb_real_miss_exits) },
53 { "itlb_v", VCPU_STAT(itlb_virt_miss_exits) },
54 { "dtlb_r", VCPU_STAT(dtlb_real_miss_exits) },
55 { "dtlb_v", VCPU_STAT(dtlb_virt_miss_exits) },
56 { "sysc", VCPU_STAT(syscall_exits) },
57 { "isi", VCPU_STAT(isi_exits) },
58 { "dsi", VCPU_STAT(dsi_exits) },
59 { "inst_emu", VCPU_STAT(emulated_inst_exits) },
60 { "dec", VCPU_STAT(dec_exits) },
61 { "ext_intr", VCPU_STAT(ext_intr_exits) },
62 { "halt_wakeup", VCPU_STAT(halt_wakeup) },
63 { "doorbell", VCPU_STAT(dbell_exits) },
64 { "guest doorbell", VCPU_STAT(gdbell_exits) },
68 /* TODO: use vcpu_printf() */
69 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
73 printk("pc: %08lx msr: %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
74 printk("lr: %08lx ctr: %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
75 printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
76 vcpu->arch.shared->srr1);
78 printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
80 for (i = 0; i < 32; i += 4) {
81 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
82 kvmppc_get_gpr(vcpu, i),
83 kvmppc_get_gpr(vcpu, i+1),
84 kvmppc_get_gpr(vcpu, i+2),
85 kvmppc_get_gpr(vcpu, i+3));
90 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
94 kvmppc_save_guest_spe(vcpu);
95 vcpu->arch.shadow_msr &= ~MSR_SPE;
99 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
103 kvmppc_load_guest_spe(vcpu);
104 vcpu->arch.shadow_msr |= MSR_SPE;
108 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
110 if (vcpu->arch.shared->msr & MSR_SPE) {
111 if (!(vcpu->arch.shadow_msr & MSR_SPE))
112 kvmppc_vcpu_enable_spe(vcpu);
113 } else if (vcpu->arch.shadow_msr & MSR_SPE) {
114 kvmppc_vcpu_disable_spe(vcpu);
118 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
124 * Helper function for "full" MSR writes. No need to call this if only
125 * EE/CE/ME/DE/RI are changing.
127 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
129 u32 old_msr = vcpu->arch.shared->msr;
131 #ifdef CONFIG_KVM_BOOKE_HV
135 vcpu->arch.shared->msr = new_msr;
137 kvmppc_mmu_msr_notify(vcpu, old_msr);
138 kvmppc_vcpu_sync_spe(vcpu);
141 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
142 unsigned int priority)
144 set_bit(priority, &vcpu->arch.pending_exceptions);
147 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
148 ulong dear_flags, ulong esr_flags)
150 vcpu->arch.queued_dear = dear_flags;
151 vcpu->arch.queued_esr = esr_flags;
152 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
155 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
156 ulong dear_flags, ulong esr_flags)
158 vcpu->arch.queued_dear = dear_flags;
159 vcpu->arch.queued_esr = esr_flags;
160 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
163 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
166 vcpu->arch.queued_esr = esr_flags;
167 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
170 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
172 vcpu->arch.queued_esr = esr_flags;
173 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
176 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
178 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
181 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
183 return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
186 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
188 clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
191 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
192 struct kvm_interrupt *irq)
194 unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
196 if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
197 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
199 kvmppc_booke_queue_irqprio(vcpu, prio);
202 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
203 struct kvm_interrupt *irq)
205 clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
206 clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
209 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
211 #ifdef CONFIG_KVM_BOOKE_HV
212 mtspr(SPRN_GSRR0, srr0);
213 mtspr(SPRN_GSRR1, srr1);
215 vcpu->arch.shared->srr0 = srr0;
216 vcpu->arch.shared->srr1 = srr1;
220 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
222 vcpu->arch.csrr0 = srr0;
223 vcpu->arch.csrr1 = srr1;
226 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
228 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
229 vcpu->arch.dsrr0 = srr0;
230 vcpu->arch.dsrr1 = srr1;
232 set_guest_csrr(vcpu, srr0, srr1);
236 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
238 vcpu->arch.mcsrr0 = srr0;
239 vcpu->arch.mcsrr1 = srr1;
242 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
244 #ifdef CONFIG_KVM_BOOKE_HV
245 return mfspr(SPRN_GDEAR);
247 return vcpu->arch.shared->dar;
251 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
253 #ifdef CONFIG_KVM_BOOKE_HV
254 mtspr(SPRN_GDEAR, dear);
256 vcpu->arch.shared->dar = dear;
260 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
262 #ifdef CONFIG_KVM_BOOKE_HV
263 return mfspr(SPRN_GESR);
265 return vcpu->arch.shared->esr;
269 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
271 #ifdef CONFIG_KVM_BOOKE_HV
272 mtspr(SPRN_GESR, esr);
274 vcpu->arch.shared->esr = esr;
278 /* Deliver the interrupt of the corresponding priority, if possible. */
279 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
280 unsigned int priority)
284 bool update_esr = false, update_dear = false;
285 ulong crit_raw = vcpu->arch.shared->critical;
286 ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
288 bool keep_irq = false;
289 enum int_class int_class;
291 /* Truncate crit indicators in 32 bit mode */
292 if (!(vcpu->arch.shared->msr & MSR_SF)) {
293 crit_raw &= 0xffffffff;
294 crit_r1 &= 0xffffffff;
297 /* Critical section when crit == r1 */
298 crit = (crit_raw == crit_r1);
299 /* ... and we're in supervisor mode */
300 crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
302 if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
303 priority = BOOKE_IRQPRIO_EXTERNAL;
308 case BOOKE_IRQPRIO_DTLB_MISS:
309 case BOOKE_IRQPRIO_DATA_STORAGE:
312 case BOOKE_IRQPRIO_INST_STORAGE:
313 case BOOKE_IRQPRIO_PROGRAM:
316 case BOOKE_IRQPRIO_ITLB_MISS:
317 case BOOKE_IRQPRIO_SYSCALL:
318 case BOOKE_IRQPRIO_FP_UNAVAIL:
319 case BOOKE_IRQPRIO_SPE_UNAVAIL:
320 case BOOKE_IRQPRIO_SPE_FP_DATA:
321 case BOOKE_IRQPRIO_SPE_FP_ROUND:
322 case BOOKE_IRQPRIO_AP_UNAVAIL:
323 case BOOKE_IRQPRIO_ALIGNMENT:
325 msr_mask = MSR_CE | MSR_ME | MSR_DE;
326 int_class = INT_CLASS_NONCRIT;
328 case BOOKE_IRQPRIO_CRITICAL:
329 case BOOKE_IRQPRIO_DBELL_CRIT:
330 allowed = vcpu->arch.shared->msr & MSR_CE;
331 allowed = allowed && !crit;
333 int_class = INT_CLASS_CRIT;
335 case BOOKE_IRQPRIO_MACHINE_CHECK:
336 allowed = vcpu->arch.shared->msr & MSR_ME;
337 allowed = allowed && !crit;
338 int_class = INT_CLASS_MC;
340 case BOOKE_IRQPRIO_DECREMENTER:
341 case BOOKE_IRQPRIO_FIT:
344 case BOOKE_IRQPRIO_EXTERNAL:
345 case BOOKE_IRQPRIO_DBELL:
346 allowed = vcpu->arch.shared->msr & MSR_EE;
347 allowed = allowed && !crit;
348 msr_mask = MSR_CE | MSR_ME | MSR_DE;
349 int_class = INT_CLASS_NONCRIT;
351 case BOOKE_IRQPRIO_DEBUG:
352 allowed = vcpu->arch.shared->msr & MSR_DE;
353 allowed = allowed && !crit;
355 int_class = INT_CLASS_CRIT;
361 case INT_CLASS_NONCRIT:
362 set_guest_srr(vcpu, vcpu->arch.pc,
363 vcpu->arch.shared->msr);
366 set_guest_csrr(vcpu, vcpu->arch.pc,
367 vcpu->arch.shared->msr);
370 set_guest_dsrr(vcpu, vcpu->arch.pc,
371 vcpu->arch.shared->msr);
374 set_guest_mcsrr(vcpu, vcpu->arch.pc,
375 vcpu->arch.shared->msr);
379 vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
380 if (update_esr == true)
381 set_guest_esr(vcpu, vcpu->arch.queued_esr);
382 if (update_dear == true)
383 set_guest_dear(vcpu, vcpu->arch.queued_dear);
384 kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
387 clear_bit(priority, &vcpu->arch.pending_exceptions);
390 #ifdef CONFIG_KVM_BOOKE_HV
392 * If an interrupt is pending but masked, raise a guest doorbell
393 * so that we are notified when the guest enables the relevant
396 if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
397 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
398 if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
399 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
400 if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
401 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
407 static void update_timer_ints(struct kvm_vcpu *vcpu)
409 if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
410 kvmppc_core_queue_dec(vcpu);
412 kvmppc_core_dequeue_dec(vcpu);
415 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
417 unsigned long *pending = &vcpu->arch.pending_exceptions;
418 unsigned int priority;
420 if (vcpu->requests) {
421 if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu)) {
423 update_timer_ints(vcpu);
427 priority = __ffs(*pending);
428 while (priority < BOOKE_IRQPRIO_MAX) {
429 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
432 priority = find_next_bit(pending,
433 BITS_PER_BYTE * sizeof(*pending),
437 /* Tell the guest about our interrupt status */
438 vcpu->arch.shared->int_pending = !!*pending;
441 /* Check pending exceptions and deliver one, if possible. */
442 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
445 WARN_ON_ONCE(!irqs_disabled());
447 kvmppc_core_check_exceptions(vcpu);
449 if (vcpu->arch.shared->msr & MSR_WE) {
451 kvm_vcpu_block(vcpu);
454 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
462 * Common checks before entering the guest world. Call with interrupts
465 * returns !0 if a signal is pending and check_signal is true
467 static int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
471 WARN_ON_ONCE(!irqs_disabled());
473 if (need_resched()) {
480 if (signal_pending(current)) {
485 if (kvmppc_core_prepare_to_enter(vcpu)) {
486 /* interrupts got enabled in between, so we
487 are back at square 1 */
497 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
500 #ifdef CONFIG_PPC_FPU
506 if (!vcpu->arch.sane) {
507 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
512 if (kvmppc_prepare_to_enter(vcpu)) {
513 kvm_run->exit_reason = KVM_EXIT_INTR;
520 #ifdef CONFIG_PPC_FPU
521 /* Save userspace FPU state in stack */
523 memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
524 fpscr = current->thread.fpscr.val;
525 fpexc_mode = current->thread.fpexc_mode;
527 /* Restore guest FPU state to thread */
528 memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
529 current->thread.fpscr.val = vcpu->arch.fpscr;
532 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
533 * as always using the FPU. Kernel usage of FP (via
534 * enable_kernel_fp()) in this thread must not occur while
535 * vcpu->fpu_active is set.
537 vcpu->fpu_active = 1;
539 kvmppc_load_guest_fp(vcpu);
542 ret = __kvmppc_vcpu_run(kvm_run, vcpu);
544 #ifdef CONFIG_PPC_FPU
545 kvmppc_save_guest_fp(vcpu);
547 vcpu->fpu_active = 0;
549 /* Save guest FPU state from thread */
550 memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
551 vcpu->arch.fpscr = current->thread.fpscr.val;
553 /* Restore userspace FPU state from stack */
554 memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
555 current->thread.fpscr.val = fpscr;
556 current->thread.fpexc_mode = fpexc_mode;
566 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
568 enum emulation_result er;
570 er = kvmppc_emulate_instruction(run, vcpu);
573 /* don't overwrite subtypes, just account kvm_stats */
574 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
575 /* Future optimization: only reload non-volatiles if
576 * they were actually modified by emulation. */
577 return RESUME_GUEST_NV;
580 run->exit_reason = KVM_EXIT_DCR;
584 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
585 __func__, vcpu->arch.pc, vcpu->arch.last_inst);
586 /* For debugging, encode the failing instruction and
587 * report it to userspace. */
588 run->hw.hardware_exit_reason = ~0ULL << 32;
589 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
590 kvmppc_core_queue_program(vcpu, ESR_PIL);
598 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
600 ulong r1, ip, msr, lr;
602 asm("mr %0, 1" : "=r"(r1));
603 asm("mflr %0" : "=r"(lr));
604 asm("mfmsr %0" : "=r"(msr));
605 asm("bl 1f; 1: mflr %0" : "=r"(ip));
607 memset(regs, 0, sizeof(*regs));
614 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
615 unsigned int exit_nr)
620 case BOOKE_INTERRUPT_EXTERNAL:
621 kvmppc_fill_pt_regs(®s);
624 case BOOKE_INTERRUPT_DECREMENTER:
625 kvmppc_fill_pt_regs(®s);
626 timer_interrupt(®s);
628 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
629 case BOOKE_INTERRUPT_DOORBELL:
630 kvmppc_fill_pt_regs(®s);
631 doorbell_exception(®s);
634 case BOOKE_INTERRUPT_MACHINE_CHECK:
637 case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
638 kvmppc_fill_pt_regs(®s);
639 performance_monitor_exception(®s);
647 * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
649 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
650 unsigned int exit_nr)
654 /* update before a new last_exit_type is rewritten */
655 kvmppc_update_timing_stats(vcpu);
657 /* restart interrupts if they were meant for the host */
658 kvmppc_restart_interrupt(vcpu, exit_nr);
662 run->exit_reason = KVM_EXIT_UNKNOWN;
663 run->ready_for_interrupt_injection = 1;
666 case BOOKE_INTERRUPT_MACHINE_CHECK:
667 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
668 kvmppc_dump_vcpu(vcpu);
669 /* For debugging, send invalid exit reason to user space */
670 run->hw.hardware_exit_reason = ~1ULL << 32;
671 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
675 case BOOKE_INTERRUPT_EXTERNAL:
676 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
680 case BOOKE_INTERRUPT_DECREMENTER:
681 kvmppc_account_exit(vcpu, DEC_EXITS);
685 case BOOKE_INTERRUPT_DOORBELL:
686 kvmppc_account_exit(vcpu, DBELL_EXITS);
690 case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
691 kvmppc_account_exit(vcpu, GDBELL_EXITS);
694 * We are here because there is a pending guest interrupt
695 * which could not be delivered as MSR_CE or MSR_ME was not
696 * set. Once we break from here we will retry delivery.
701 case BOOKE_INTERRUPT_GUEST_DBELL:
702 kvmppc_account_exit(vcpu, GDBELL_EXITS);
705 * We are here because there is a pending guest interrupt
706 * which could not be delivered as MSR_EE was not set. Once
707 * we break from here we will retry delivery.
712 case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
716 case BOOKE_INTERRUPT_HV_PRIV:
717 r = emulation_exit(run, vcpu);
720 case BOOKE_INTERRUPT_PROGRAM:
721 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
723 * Program traps generated by user-level software must
724 * be handled by the guest kernel.
726 * In GS mode, hypervisor privileged instructions trap
727 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
728 * actual program interrupts, handled by the guest.
730 kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
732 kvmppc_account_exit(vcpu, USR_PR_INST);
736 r = emulation_exit(run, vcpu);
739 case BOOKE_INTERRUPT_FP_UNAVAIL:
740 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
741 kvmppc_account_exit(vcpu, FP_UNAVAIL);
746 case BOOKE_INTERRUPT_SPE_UNAVAIL: {
747 if (vcpu->arch.shared->msr & MSR_SPE)
748 kvmppc_vcpu_enable_spe(vcpu);
750 kvmppc_booke_queue_irqprio(vcpu,
751 BOOKE_IRQPRIO_SPE_UNAVAIL);
756 case BOOKE_INTERRUPT_SPE_FP_DATA:
757 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
761 case BOOKE_INTERRUPT_SPE_FP_ROUND:
762 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
766 case BOOKE_INTERRUPT_SPE_UNAVAIL:
768 * Guest wants SPE, but host kernel doesn't support it. Send
769 * an "unimplemented operation" program check to the guest.
771 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
776 * These really should never happen without CONFIG_SPE,
777 * as we should never enable the real MSR[SPE] in the guest.
779 case BOOKE_INTERRUPT_SPE_FP_DATA:
780 case BOOKE_INTERRUPT_SPE_FP_ROUND:
781 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
782 __func__, exit_nr, vcpu->arch.pc);
783 run->hw.hardware_exit_reason = exit_nr;
788 case BOOKE_INTERRUPT_DATA_STORAGE:
789 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
790 vcpu->arch.fault_esr);
791 kvmppc_account_exit(vcpu, DSI_EXITS);
795 case BOOKE_INTERRUPT_INST_STORAGE:
796 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
797 kvmppc_account_exit(vcpu, ISI_EXITS);
801 #ifdef CONFIG_KVM_BOOKE_HV
802 case BOOKE_INTERRUPT_HV_SYSCALL:
803 if (!(vcpu->arch.shared->msr & MSR_PR)) {
804 kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
807 * hcall from guest userspace -- send privileged
808 * instruction program check.
810 kvmppc_core_queue_program(vcpu, ESR_PPR);
816 case BOOKE_INTERRUPT_SYSCALL:
817 if (!(vcpu->arch.shared->msr & MSR_PR) &&
818 (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
819 /* KVM PV hypercalls */
820 kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
824 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
826 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
831 case BOOKE_INTERRUPT_DTLB_MISS: {
832 unsigned long eaddr = vcpu->arch.fault_dear;
837 #ifdef CONFIG_KVM_E500V2
838 if (!(vcpu->arch.shared->msr & MSR_PR) &&
839 (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
840 kvmppc_map_magic(vcpu);
841 kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
848 /* Check the guest TLB. */
849 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
850 if (gtlb_index < 0) {
851 /* The guest didn't have a mapping for it. */
852 kvmppc_core_queue_dtlb_miss(vcpu,
853 vcpu->arch.fault_dear,
854 vcpu->arch.fault_esr);
855 kvmppc_mmu_dtlb_miss(vcpu);
856 kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
861 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
862 gfn = gpaddr >> PAGE_SHIFT;
864 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
865 /* The guest TLB had a mapping, but the shadow TLB
866 * didn't, and it is RAM. This could be because:
867 * a) the entry is mapping the host kernel, or
868 * b) the guest used a large mapping which we're faking
869 * Either way, we need to satisfy the fault without
870 * invoking the guest. */
871 kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
872 kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
875 /* Guest has mapped and accessed a page which is not
877 vcpu->arch.paddr_accessed = gpaddr;
878 r = kvmppc_emulate_mmio(run, vcpu);
879 kvmppc_account_exit(vcpu, MMIO_EXITS);
885 case BOOKE_INTERRUPT_ITLB_MISS: {
886 unsigned long eaddr = vcpu->arch.pc;
893 /* Check the guest TLB. */
894 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
895 if (gtlb_index < 0) {
896 /* The guest didn't have a mapping for it. */
897 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
898 kvmppc_mmu_itlb_miss(vcpu);
899 kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
903 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
905 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
906 gfn = gpaddr >> PAGE_SHIFT;
908 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
909 /* The guest TLB had a mapping, but the shadow TLB
910 * didn't. This could be because:
911 * a) the entry is mapping the host kernel, or
912 * b) the guest used a large mapping which we're faking
913 * Either way, we need to satisfy the fault without
914 * invoking the guest. */
915 kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
917 /* Guest mapped and leaped at non-RAM! */
918 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
924 case BOOKE_INTERRUPT_DEBUG: {
927 vcpu->arch.pc = mfspr(SPRN_CSRR0);
929 /* clear IAC events in DBSR register */
930 dbsr = mfspr(SPRN_DBSR);
931 dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
932 mtspr(SPRN_DBSR, dbsr);
934 run->exit_reason = KVM_EXIT_DEBUG;
935 kvmppc_account_exit(vcpu, DEBUG_EXITS);
941 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
946 * To avoid clobbering exit_reason, only check for signals if we
947 * aren't already exiting to userspace for some other reason.
949 if (!(r & RESUME_HOST)) {
951 if (kvmppc_prepare_to_enter(vcpu)) {
952 run->exit_reason = KVM_EXIT_INTR;
953 r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
954 kvmppc_account_exit(vcpu, SIGNAL_EXITS);
961 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
962 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
968 vcpu->arch.shared->pir = vcpu->vcpu_id;
969 kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
970 kvmppc_set_msr(vcpu, 0);
972 #ifndef CONFIG_KVM_BOOKE_HV
973 vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
974 vcpu->arch.shadow_pid = 1;
975 vcpu->arch.shared->msr = 0;
978 /* Eye-catching numbers so we know if the guest takes an interrupt
979 * before it's programmed its own IVPR/IVORs. */
980 vcpu->arch.ivpr = 0x55550000;
981 for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
982 vcpu->arch.ivor[i] = 0x7700 | i * 4;
984 kvmppc_init_timing_stats(vcpu);
986 r = kvmppc_core_vcpu_setup(vcpu);
987 kvmppc_sanity_check(vcpu);
991 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
995 regs->pc = vcpu->arch.pc;
996 regs->cr = kvmppc_get_cr(vcpu);
997 regs->ctr = vcpu->arch.ctr;
998 regs->lr = vcpu->arch.lr;
999 regs->xer = kvmppc_get_xer(vcpu);
1000 regs->msr = vcpu->arch.shared->msr;
1001 regs->srr0 = vcpu->arch.shared->srr0;
1002 regs->srr1 = vcpu->arch.shared->srr1;
1003 regs->pid = vcpu->arch.pid;
1004 regs->sprg0 = vcpu->arch.shared->sprg0;
1005 regs->sprg1 = vcpu->arch.shared->sprg1;
1006 regs->sprg2 = vcpu->arch.shared->sprg2;
1007 regs->sprg3 = vcpu->arch.shared->sprg3;
1008 regs->sprg4 = vcpu->arch.shared->sprg4;
1009 regs->sprg5 = vcpu->arch.shared->sprg5;
1010 regs->sprg6 = vcpu->arch.shared->sprg6;
1011 regs->sprg7 = vcpu->arch.shared->sprg7;
1013 for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1014 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1019 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1023 vcpu->arch.pc = regs->pc;
1024 kvmppc_set_cr(vcpu, regs->cr);
1025 vcpu->arch.ctr = regs->ctr;
1026 vcpu->arch.lr = regs->lr;
1027 kvmppc_set_xer(vcpu, regs->xer);
1028 kvmppc_set_msr(vcpu, regs->msr);
1029 vcpu->arch.shared->srr0 = regs->srr0;
1030 vcpu->arch.shared->srr1 = regs->srr1;
1031 kvmppc_set_pid(vcpu, regs->pid);
1032 vcpu->arch.shared->sprg0 = regs->sprg0;
1033 vcpu->arch.shared->sprg1 = regs->sprg1;
1034 vcpu->arch.shared->sprg2 = regs->sprg2;
1035 vcpu->arch.shared->sprg3 = regs->sprg3;
1036 vcpu->arch.shared->sprg4 = regs->sprg4;
1037 vcpu->arch.shared->sprg5 = regs->sprg5;
1038 vcpu->arch.shared->sprg6 = regs->sprg6;
1039 vcpu->arch.shared->sprg7 = regs->sprg7;
1041 for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1042 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1047 static void get_sregs_base(struct kvm_vcpu *vcpu,
1048 struct kvm_sregs *sregs)
1052 sregs->u.e.features |= KVM_SREGS_E_BASE;
1054 sregs->u.e.csrr0 = vcpu->arch.csrr0;
1055 sregs->u.e.csrr1 = vcpu->arch.csrr1;
1056 sregs->u.e.mcsr = vcpu->arch.mcsr;
1057 sregs->u.e.esr = get_guest_esr(vcpu);
1058 sregs->u.e.dear = get_guest_dear(vcpu);
1059 sregs->u.e.tsr = vcpu->arch.tsr;
1060 sregs->u.e.tcr = vcpu->arch.tcr;
1061 sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1063 sregs->u.e.vrsave = vcpu->arch.vrsave;
1066 static int set_sregs_base(struct kvm_vcpu *vcpu,
1067 struct kvm_sregs *sregs)
1069 if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1072 vcpu->arch.csrr0 = sregs->u.e.csrr0;
1073 vcpu->arch.csrr1 = sregs->u.e.csrr1;
1074 vcpu->arch.mcsr = sregs->u.e.mcsr;
1075 set_guest_esr(vcpu, sregs->u.e.esr);
1076 set_guest_dear(vcpu, sregs->u.e.dear);
1077 vcpu->arch.vrsave = sregs->u.e.vrsave;
1078 kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1080 if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1081 vcpu->arch.dec = sregs->u.e.dec;
1082 kvmppc_emulate_dec(vcpu);
1085 if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1086 vcpu->arch.tsr = sregs->u.e.tsr;
1087 update_timer_ints(vcpu);
1093 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1094 struct kvm_sregs *sregs)
1096 sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1098 sregs->u.e.pir = vcpu->vcpu_id;
1099 sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1100 sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1101 sregs->u.e.decar = vcpu->arch.decar;
1102 sregs->u.e.ivpr = vcpu->arch.ivpr;
1105 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1106 struct kvm_sregs *sregs)
1108 if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1111 if (sregs->u.e.pir != vcpu->vcpu_id)
1114 vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1115 vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1116 vcpu->arch.decar = sregs->u.e.decar;
1117 vcpu->arch.ivpr = sregs->u.e.ivpr;
1122 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1124 sregs->u.e.features |= KVM_SREGS_E_IVOR;
1126 sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1127 sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1128 sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1129 sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1130 sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1131 sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1132 sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1133 sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1134 sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1135 sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1136 sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1137 sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1138 sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1139 sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1140 sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1141 sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1144 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1146 if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1149 vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1150 vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1151 vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1152 vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1153 vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1154 vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1155 vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1156 vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1157 vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1158 vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1159 vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1160 vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1161 vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1162 vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1163 vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1164 vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1169 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1170 struct kvm_sregs *sregs)
1172 sregs->pvr = vcpu->arch.pvr;
1174 get_sregs_base(vcpu, sregs);
1175 get_sregs_arch206(vcpu, sregs);
1176 kvmppc_core_get_sregs(vcpu, sregs);
1180 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1181 struct kvm_sregs *sregs)
1185 if (vcpu->arch.pvr != sregs->pvr)
1188 ret = set_sregs_base(vcpu, sregs);
1192 ret = set_sregs_arch206(vcpu, sregs);
1196 return kvmppc_core_set_sregs(vcpu, sregs);
1199 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1204 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1209 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1214 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1219 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1220 struct kvm_translation *tr)
1224 r = kvmppc_core_vcpu_translate(vcpu, tr);
1228 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1233 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1234 struct kvm_userspace_memory_region *mem)
1239 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1240 struct kvm_userspace_memory_region *mem)
1244 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1246 vcpu->arch.tcr = new_tcr;
1247 update_timer_ints(vcpu);
1250 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1252 set_bits(tsr_bits, &vcpu->arch.tsr);
1254 kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1255 kvm_vcpu_kick(vcpu);
1258 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1260 clear_bits(tsr_bits, &vcpu->arch.tsr);
1261 update_timer_ints(vcpu);
1264 void kvmppc_decrementer_func(unsigned long data)
1266 struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1268 kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1271 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1273 current->thread.kvm_vcpu = vcpu;
1276 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1278 current->thread.kvm_vcpu = NULL;
1281 int __init kvmppc_booke_init(void)
1283 #ifndef CONFIG_KVM_BOOKE_HV
1284 unsigned long ivor[16];
1285 unsigned long max_ivor = 0;
1288 /* We install our own exception handlers by hijacking IVPR. IVPR must
1289 * be 16-bit aligned, so we need a 64KB allocation. */
1290 kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1292 if (!kvmppc_booke_handlers)
1295 /* XXX make sure our handlers are smaller than Linux's */
1297 /* Copy our interrupt handlers to match host IVORs. That way we don't
1298 * have to swap the IVORs on every guest/host transition. */
1299 ivor[0] = mfspr(SPRN_IVOR0);
1300 ivor[1] = mfspr(SPRN_IVOR1);
1301 ivor[2] = mfspr(SPRN_IVOR2);
1302 ivor[3] = mfspr(SPRN_IVOR3);
1303 ivor[4] = mfspr(SPRN_IVOR4);
1304 ivor[5] = mfspr(SPRN_IVOR5);
1305 ivor[6] = mfspr(SPRN_IVOR6);
1306 ivor[7] = mfspr(SPRN_IVOR7);
1307 ivor[8] = mfspr(SPRN_IVOR8);
1308 ivor[9] = mfspr(SPRN_IVOR9);
1309 ivor[10] = mfspr(SPRN_IVOR10);
1310 ivor[11] = mfspr(SPRN_IVOR11);
1311 ivor[12] = mfspr(SPRN_IVOR12);
1312 ivor[13] = mfspr(SPRN_IVOR13);
1313 ivor[14] = mfspr(SPRN_IVOR14);
1314 ivor[15] = mfspr(SPRN_IVOR15);
1316 for (i = 0; i < 16; i++) {
1317 if (ivor[i] > max_ivor)
1320 memcpy((void *)kvmppc_booke_handlers + ivor[i],
1321 kvmppc_handlers_start + i * kvmppc_handler_len,
1322 kvmppc_handler_len);
1324 flush_icache_range(kvmppc_booke_handlers,
1325 kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1326 #endif /* !BOOKE_HV */
1330 void __exit kvmppc_booke_exit(void)
1332 free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);