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
17 * Authors: Hollis Blanchard <hollisb@us.ibm.com>
18 * Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
21 #include <linux/errno.h>
22 #include <linux/err.h>
23 #include <linux/kvm_host.h>
24 #include <linux/module.h>
25 #include <linux/vmalloc.h>
27 #include <asm/cputable.h>
28 #include <asm/uaccess.h>
29 #include <asm/kvm_ppc.h>
33 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
34 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
36 struct kvm_stats_debugfs_item debugfs_entries
[] = {
37 { "exits", VCPU_STAT(sum_exits
) },
38 { "mmio", VCPU_STAT(mmio_exits
) },
39 { "dcr", VCPU_STAT(dcr_exits
) },
40 { "sig", VCPU_STAT(signal_exits
) },
41 { "light", VCPU_STAT(light_exits
) },
42 { "itlb_r", VCPU_STAT(itlb_real_miss_exits
) },
43 { "itlb_v", VCPU_STAT(itlb_virt_miss_exits
) },
44 { "dtlb_r", VCPU_STAT(dtlb_real_miss_exits
) },
45 { "dtlb_v", VCPU_STAT(dtlb_virt_miss_exits
) },
46 { "sysc", VCPU_STAT(syscall_exits
) },
47 { "isi", VCPU_STAT(isi_exits
) },
48 { "dsi", VCPU_STAT(dsi_exits
) },
49 { "inst_emu", VCPU_STAT(emulated_inst_exits
) },
50 { "dec", VCPU_STAT(dec_exits
) },
51 { "ext_intr", VCPU_STAT(ext_intr_exits
) },
52 { "halt_wakeup", VCPU_STAT(halt_wakeup
) },
56 static const u32 interrupt_msr_mask
[16] = {
57 [BOOKE_INTERRUPT_CRITICAL
] = MSR_ME
,
58 [BOOKE_INTERRUPT_MACHINE_CHECK
] = 0,
59 [BOOKE_INTERRUPT_DATA_STORAGE
] = MSR_CE
|MSR_ME
|MSR_DE
,
60 [BOOKE_INTERRUPT_INST_STORAGE
] = MSR_CE
|MSR_ME
|MSR_DE
,
61 [BOOKE_INTERRUPT_EXTERNAL
] = MSR_CE
|MSR_ME
|MSR_DE
,
62 [BOOKE_INTERRUPT_ALIGNMENT
] = MSR_CE
|MSR_ME
|MSR_DE
,
63 [BOOKE_INTERRUPT_PROGRAM
] = MSR_CE
|MSR_ME
|MSR_DE
,
64 [BOOKE_INTERRUPT_FP_UNAVAIL
] = MSR_CE
|MSR_ME
|MSR_DE
,
65 [BOOKE_INTERRUPT_SYSCALL
] = MSR_CE
|MSR_ME
|MSR_DE
,
66 [BOOKE_INTERRUPT_AP_UNAVAIL
] = MSR_CE
|MSR_ME
|MSR_DE
,
67 [BOOKE_INTERRUPT_DECREMENTER
] = MSR_CE
|MSR_ME
|MSR_DE
,
68 [BOOKE_INTERRUPT_FIT
] = MSR_CE
|MSR_ME
|MSR_DE
,
69 [BOOKE_INTERRUPT_WATCHDOG
] = MSR_ME
,
70 [BOOKE_INTERRUPT_DTLB_MISS
] = MSR_CE
|MSR_ME
|MSR_DE
,
71 [BOOKE_INTERRUPT_ITLB_MISS
] = MSR_CE
|MSR_ME
|MSR_DE
,
72 [BOOKE_INTERRUPT_DEBUG
] = MSR_ME
,
75 const unsigned char exception_priority
[] = {
76 [BOOKE_INTERRUPT_DATA_STORAGE
] = 0,
77 [BOOKE_INTERRUPT_INST_STORAGE
] = 1,
78 [BOOKE_INTERRUPT_ALIGNMENT
] = 2,
79 [BOOKE_INTERRUPT_PROGRAM
] = 3,
80 [BOOKE_INTERRUPT_FP_UNAVAIL
] = 4,
81 [BOOKE_INTERRUPT_SYSCALL
] = 5,
82 [BOOKE_INTERRUPT_AP_UNAVAIL
] = 6,
83 [BOOKE_INTERRUPT_DTLB_MISS
] = 7,
84 [BOOKE_INTERRUPT_ITLB_MISS
] = 8,
85 [BOOKE_INTERRUPT_MACHINE_CHECK
] = 9,
86 [BOOKE_INTERRUPT_DEBUG
] = 10,
87 [BOOKE_INTERRUPT_CRITICAL
] = 11,
88 [BOOKE_INTERRUPT_WATCHDOG
] = 12,
89 [BOOKE_INTERRUPT_EXTERNAL
] = 13,
90 [BOOKE_INTERRUPT_FIT
] = 14,
91 [BOOKE_INTERRUPT_DECREMENTER
] = 15,
94 const unsigned char priority_exception
[] = {
95 BOOKE_INTERRUPT_DATA_STORAGE
,
96 BOOKE_INTERRUPT_INST_STORAGE
,
97 BOOKE_INTERRUPT_ALIGNMENT
,
98 BOOKE_INTERRUPT_PROGRAM
,
99 BOOKE_INTERRUPT_FP_UNAVAIL
,
100 BOOKE_INTERRUPT_SYSCALL
,
101 BOOKE_INTERRUPT_AP_UNAVAIL
,
102 BOOKE_INTERRUPT_DTLB_MISS
,
103 BOOKE_INTERRUPT_ITLB_MISS
,
104 BOOKE_INTERRUPT_MACHINE_CHECK
,
105 BOOKE_INTERRUPT_DEBUG
,
106 BOOKE_INTERRUPT_CRITICAL
,
107 BOOKE_INTERRUPT_WATCHDOG
,
108 BOOKE_INTERRUPT_EXTERNAL
,
110 BOOKE_INTERRUPT_DECREMENTER
,
114 void kvmppc_dump_tlbs(struct kvm_vcpu
*vcpu
)
119 printk("vcpu %d TLB dump:\n", vcpu
->vcpu_id
);
120 printk("| %2s | %3s | %8s | %8s | %8s |\n",
121 "nr", "tid", "word0", "word1", "word2");
123 for (i
= 0; i
< PPC44x_TLB_SIZE
; i
++) {
124 tlbe
= &vcpu
->arch
.guest_tlb
[i
];
125 if (tlbe
->word0
& PPC44x_TLB_VALID
)
126 printk(" G%2d | %02X | %08X | %08X | %08X |\n",
127 i
, tlbe
->tid
, tlbe
->word0
, tlbe
->word1
,
131 for (i
= 0; i
< PPC44x_TLB_SIZE
; i
++) {
132 tlbe
= &vcpu
->arch
.shadow_tlb
[i
];
133 if (tlbe
->word0
& PPC44x_TLB_VALID
)
134 printk(" S%2d | %02X | %08X | %08X | %08X |\n",
135 i
, tlbe
->tid
, tlbe
->word0
, tlbe
->word1
,
140 /* TODO: use vcpu_printf() */
141 void kvmppc_dump_vcpu(struct kvm_vcpu
*vcpu
)
145 printk("pc: %08x msr: %08x\n", vcpu
->arch
.pc
, vcpu
->arch
.msr
);
146 printk("lr: %08x ctr: %08x\n", vcpu
->arch
.lr
, vcpu
->arch
.ctr
);
147 printk("srr0: %08x srr1: %08x\n", vcpu
->arch
.srr0
, vcpu
->arch
.srr1
);
149 printk("exceptions: %08lx\n", vcpu
->arch
.pending_exceptions
);
151 for (i
= 0; i
< 32; i
+= 4) {
152 printk("gpr%02d: %08x %08x %08x %08x\n", i
,
156 vcpu
->arch
.gpr
[i
+3]);
160 /* Check if we are ready to deliver the interrupt */
161 static int kvmppc_can_deliver_interrupt(struct kvm_vcpu
*vcpu
, int interrupt
)
166 case BOOKE_INTERRUPT_CRITICAL
:
167 r
= vcpu
->arch
.msr
& MSR_CE
;
169 case BOOKE_INTERRUPT_MACHINE_CHECK
:
170 r
= vcpu
->arch
.msr
& MSR_ME
;
172 case BOOKE_INTERRUPT_EXTERNAL
:
173 r
= vcpu
->arch
.msr
& MSR_EE
;
175 case BOOKE_INTERRUPT_DECREMENTER
:
176 r
= vcpu
->arch
.msr
& MSR_EE
;
178 case BOOKE_INTERRUPT_FIT
:
179 r
= vcpu
->arch
.msr
& MSR_EE
;
181 case BOOKE_INTERRUPT_WATCHDOG
:
182 r
= vcpu
->arch
.msr
& MSR_CE
;
184 case BOOKE_INTERRUPT_DEBUG
:
185 r
= vcpu
->arch
.msr
& MSR_DE
;
194 static void kvmppc_deliver_interrupt(struct kvm_vcpu
*vcpu
, int interrupt
)
197 case BOOKE_INTERRUPT_DECREMENTER
:
198 vcpu
->arch
.tsr
|= TSR_DIS
;
202 vcpu
->arch
.srr0
= vcpu
->arch
.pc
;
203 vcpu
->arch
.srr1
= vcpu
->arch
.msr
;
204 vcpu
->arch
.pc
= vcpu
->arch
.ivpr
| vcpu
->arch
.ivor
[interrupt
];
205 kvmppc_set_msr(vcpu
, vcpu
->arch
.msr
& interrupt_msr_mask
[interrupt
]);
208 /* Check pending exceptions and deliver one, if possible. */
209 void kvmppc_check_and_deliver_interrupts(struct kvm_vcpu
*vcpu
)
211 unsigned long *pending
= &vcpu
->arch
.pending_exceptions
;
212 unsigned int exception
;
213 unsigned int priority
;
215 priority
= find_first_bit(pending
, BITS_PER_BYTE
* sizeof(*pending
));
216 while (priority
<= BOOKE_MAX_INTERRUPT
) {
217 exception
= priority_exception
[priority
];
218 if (kvmppc_can_deliver_interrupt(vcpu
, exception
)) {
219 kvmppc_clear_exception(vcpu
, exception
);
220 kvmppc_deliver_interrupt(vcpu
, exception
);
224 priority
= find_next_bit(pending
,
225 BITS_PER_BYTE
* sizeof(*pending
),
230 static int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
232 enum emulation_result er
;
235 er
= kvmppc_emulate_instruction(run
, vcpu
);
238 /* Future optimization: only reload non-volatiles if they were
239 * actually modified. */
242 case EMULATE_DO_MMIO
:
243 run
->exit_reason
= KVM_EXIT_MMIO
;
244 /* We must reload nonvolatiles because "update" load/store
245 * instructions modify register state. */
246 /* Future optimization: only reload non-volatiles if they were
247 * actually modified. */
251 /* XXX Deliver Program interrupt to guest. */
252 printk(KERN_EMERG
"%s: emulation failed (%08x)\n", __func__
,
253 vcpu
->arch
.last_inst
);
266 * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
268 int kvmppc_handle_exit(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
269 unsigned int exit_nr
)
271 enum emulation_result er
;
276 run
->exit_reason
= KVM_EXIT_UNKNOWN
;
277 run
->ready_for_interrupt_injection
= 1;
280 case BOOKE_INTERRUPT_MACHINE_CHECK
:
281 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR
));
282 kvmppc_dump_vcpu(vcpu
);
286 case BOOKE_INTERRUPT_EXTERNAL
:
287 case BOOKE_INTERRUPT_DECREMENTER
:
288 /* Since we switched IVPR back to the host's value, the host
289 * handled this interrupt the moment we enabled interrupts.
290 * Now we just offer it a chance to reschedule the guest. */
292 /* XXX At this point the TLB still holds our shadow TLB, so if
293 * we do reschedule the host will fault over it. Perhaps we
294 * should politely restore the host's entries to minimize
295 * misses before ceding control. */
298 if (exit_nr
== BOOKE_INTERRUPT_DECREMENTER
)
299 vcpu
->stat
.dec_exits
++;
301 vcpu
->stat
.ext_intr_exits
++;
305 case BOOKE_INTERRUPT_PROGRAM
:
306 if (vcpu
->arch
.msr
& MSR_PR
) {
307 /* Program traps generated by user-level software must be handled
308 * by the guest kernel. */
309 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
310 kvmppc_queue_exception(vcpu
, BOOKE_INTERRUPT_PROGRAM
);
315 er
= kvmppc_emulate_instruction(run
, vcpu
);
318 /* Future optimization: only reload non-volatiles if
319 * they were actually modified by emulation. */
320 vcpu
->stat
.emulated_inst_exits
++;
324 run
->exit_reason
= KVM_EXIT_DCR
;
328 /* XXX Deliver Program interrupt to guest. */
329 printk(KERN_CRIT
"%s: emulation at %x failed (%08x)\n",
330 __func__
, vcpu
->arch
.pc
, vcpu
->arch
.last_inst
);
331 /* For debugging, encode the failing instruction and
332 * report it to userspace. */
333 run
->hw
.hardware_exit_reason
= ~0ULL << 32;
334 run
->hw
.hardware_exit_reason
|= vcpu
->arch
.last_inst
;
342 case BOOKE_INTERRUPT_FP_UNAVAIL
:
343 kvmppc_queue_exception(vcpu
, exit_nr
);
347 case BOOKE_INTERRUPT_DATA_STORAGE
:
348 vcpu
->arch
.dear
= vcpu
->arch
.fault_dear
;
349 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
350 kvmppc_queue_exception(vcpu
, exit_nr
);
351 vcpu
->stat
.dsi_exits
++;
355 case BOOKE_INTERRUPT_INST_STORAGE
:
356 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
357 kvmppc_queue_exception(vcpu
, exit_nr
);
358 vcpu
->stat
.isi_exits
++;
362 case BOOKE_INTERRUPT_SYSCALL
:
363 kvmppc_queue_exception(vcpu
, exit_nr
);
364 vcpu
->stat
.syscall_exits
++;
368 case BOOKE_INTERRUPT_DTLB_MISS
: {
370 unsigned long eaddr
= vcpu
->arch
.fault_dear
;
373 /* Check the guest TLB. */
374 gtlbe
= kvmppc_44x_dtlb_search(vcpu
, eaddr
);
376 /* The guest didn't have a mapping for it. */
377 kvmppc_queue_exception(vcpu
, exit_nr
);
378 vcpu
->arch
.dear
= vcpu
->arch
.fault_dear
;
379 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
380 vcpu
->stat
.dtlb_real_miss_exits
++;
385 vcpu
->arch
.paddr_accessed
= tlb_xlate(gtlbe
, eaddr
);
386 gfn
= vcpu
->arch
.paddr_accessed
>> PAGE_SHIFT
;
388 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
389 /* The guest TLB had a mapping, but the shadow TLB
390 * didn't, and it is RAM. This could be because:
391 * a) the entry is mapping the host kernel, or
392 * b) the guest used a large mapping which we're faking
393 * Either way, we need to satisfy the fault without
394 * invoking the guest. */
395 kvmppc_mmu_map(vcpu
, eaddr
, gfn
, gtlbe
->tid
,
397 vcpu
->stat
.dtlb_virt_miss_exits
++;
400 /* Guest has mapped and accessed a page which is not
402 r
= kvmppc_emulate_mmio(run
, vcpu
);
408 case BOOKE_INTERRUPT_ITLB_MISS
: {
410 unsigned long eaddr
= vcpu
->arch
.pc
;
415 /* Check the guest TLB. */
416 gtlbe
= kvmppc_44x_itlb_search(vcpu
, eaddr
);
418 /* The guest didn't have a mapping for it. */
419 kvmppc_queue_exception(vcpu
, exit_nr
);
420 vcpu
->stat
.itlb_real_miss_exits
++;
424 vcpu
->stat
.itlb_virt_miss_exits
++;
426 gfn
= tlb_xlate(gtlbe
, eaddr
) >> PAGE_SHIFT
;
428 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
429 /* The guest TLB had a mapping, but the shadow TLB
430 * didn't. This could be because:
431 * a) the entry is mapping the host kernel, or
432 * b) the guest used a large mapping which we're faking
433 * Either way, we need to satisfy the fault without
434 * invoking the guest. */
435 kvmppc_mmu_map(vcpu
, eaddr
, gfn
, gtlbe
->tid
,
438 /* Guest mapped and leaped at non-RAM! */
439 kvmppc_queue_exception(vcpu
,
440 BOOKE_INTERRUPT_MACHINE_CHECK
);
447 printk(KERN_EMERG
"exit_nr %d\n", exit_nr
);
453 kvmppc_check_and_deliver_interrupts(vcpu
);
455 /* Do some exit accounting. */
456 vcpu
->stat
.sum_exits
++;
457 if (!(r
& RESUME_HOST
)) {
458 /* To avoid clobbering exit_reason, only check for signals if
459 * we aren't already exiting to userspace for some other
461 if (signal_pending(current
)) {
462 run
->exit_reason
= KVM_EXIT_INTR
;
463 r
= (-EINTR
<< 2) | RESUME_HOST
| (r
& RESUME_FLAG_NV
);
465 vcpu
->stat
.signal_exits
++;
467 vcpu
->stat
.light_exits
++;
470 switch (run
->exit_reason
) {
472 vcpu
->stat
.mmio_exits
++;
475 vcpu
->stat
.dcr_exits
++;
478 vcpu
->stat
.signal_exits
++;
486 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
487 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
489 struct tlbe
*tlbe
= &vcpu
->arch
.guest_tlb
[0];
492 tlbe
->word0
= PPC44x_TLB_16M
| PPC44x_TLB_VALID
;
494 tlbe
->word2
= PPC44x_TLB_SX
| PPC44x_TLB_SW
| PPC44x_TLB_SR
;
498 tlbe
->word0
= 0xef600000 | PPC44x_TLB_4K
| PPC44x_TLB_VALID
;
499 tlbe
->word1
= 0xef600000;
500 tlbe
->word2
= PPC44x_TLB_SX
| PPC44x_TLB_SW
| PPC44x_TLB_SR
501 | PPC44x_TLB_I
| PPC44x_TLB_G
;
505 vcpu
->arch
.gpr
[1] = (16<<20) - 8; /* -8 for the callee-save LR slot */
507 /* Eye-catching number so we know if the guest takes an interrupt
508 * before it's programmed its own IVPR. */
509 vcpu
->arch
.ivpr
= 0x55550000;
511 /* Since the guest can directly access the timebase, it must know the
512 * real timebase frequency. Accordingly, it must see the state of
514 vcpu
->arch
.ccr1
= mfspr(SPRN_CCR1
);
519 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
523 regs
->pc
= vcpu
->arch
.pc
;
524 regs
->cr
= vcpu
->arch
.cr
;
525 regs
->ctr
= vcpu
->arch
.ctr
;
526 regs
->lr
= vcpu
->arch
.lr
;
527 regs
->xer
= vcpu
->arch
.xer
;
528 regs
->msr
= vcpu
->arch
.msr
;
529 regs
->srr0
= vcpu
->arch
.srr0
;
530 regs
->srr1
= vcpu
->arch
.srr1
;
531 regs
->pid
= vcpu
->arch
.pid
;
532 regs
->sprg0
= vcpu
->arch
.sprg0
;
533 regs
->sprg1
= vcpu
->arch
.sprg1
;
534 regs
->sprg2
= vcpu
->arch
.sprg2
;
535 regs
->sprg3
= vcpu
->arch
.sprg3
;
536 regs
->sprg5
= vcpu
->arch
.sprg4
;
537 regs
->sprg6
= vcpu
->arch
.sprg5
;
538 regs
->sprg7
= vcpu
->arch
.sprg6
;
540 for (i
= 0; i
< ARRAY_SIZE(regs
->gpr
); i
++)
541 regs
->gpr
[i
] = vcpu
->arch
.gpr
[i
];
546 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
550 vcpu
->arch
.pc
= regs
->pc
;
551 vcpu
->arch
.cr
= regs
->cr
;
552 vcpu
->arch
.ctr
= regs
->ctr
;
553 vcpu
->arch
.lr
= regs
->lr
;
554 vcpu
->arch
.xer
= regs
->xer
;
555 vcpu
->arch
.msr
= regs
->msr
;
556 vcpu
->arch
.srr0
= regs
->srr0
;
557 vcpu
->arch
.srr1
= regs
->srr1
;
558 vcpu
->arch
.sprg0
= regs
->sprg0
;
559 vcpu
->arch
.sprg1
= regs
->sprg1
;
560 vcpu
->arch
.sprg2
= regs
->sprg2
;
561 vcpu
->arch
.sprg3
= regs
->sprg3
;
562 vcpu
->arch
.sprg5
= regs
->sprg4
;
563 vcpu
->arch
.sprg6
= regs
->sprg5
;
564 vcpu
->arch
.sprg7
= regs
->sprg6
;
566 for (i
= 0; i
< ARRAY_SIZE(vcpu
->arch
.gpr
); i
++)
567 vcpu
->arch
.gpr
[i
] = regs
->gpr
[i
];
572 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
573 struct kvm_sregs
*sregs
)
578 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
579 struct kvm_sregs
*sregs
)
584 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
589 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
594 /* 'linear_address' is actually an encoding of AS|PID|EADDR . */
595 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
596 struct kvm_translation
*tr
)
604 eaddr
= tr
->linear_address
;
605 pid
= (tr
->linear_address
>> 32) & 0xff;
606 as
= (tr
->linear_address
>> 40) & 0x1;
608 index
= kvmppc_44x_tlb_index(vcpu
, eaddr
, pid
, as
);
614 gtlbe
= &vcpu
->arch
.guest_tlb
[index
];
616 tr
->physical_address
= tlb_xlate(gtlbe
, eaddr
);
617 /* XXX what does "writeable" and "usermode" even mean? */