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),
233 * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
235 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
236 unsigned int exit_nr)
238 enum emulation_result er;
243 run->exit_reason = KVM_EXIT_UNKNOWN;
244 run->ready_for_interrupt_injection = 1;
247 case BOOKE_INTERRUPT_MACHINE_CHECK:
248 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
249 kvmppc_dump_vcpu(vcpu);
253 case BOOKE_INTERRUPT_EXTERNAL:
254 case BOOKE_INTERRUPT_DECREMENTER:
255 /* Since we switched IVPR back to the host's value, the host
256 * handled this interrupt the moment we enabled interrupts.
257 * Now we just offer it a chance to reschedule the guest. */
259 /* XXX At this point the TLB still holds our shadow TLB, so if
260 * we do reschedule the host will fault over it. Perhaps we
261 * should politely restore the host's entries to minimize
262 * misses before ceding control. */
265 if (exit_nr == BOOKE_INTERRUPT_DECREMENTER)
266 vcpu->stat.dec_exits++;
268 vcpu->stat.ext_intr_exits++;
272 case BOOKE_INTERRUPT_PROGRAM:
273 if (vcpu->arch.msr & MSR_PR) {
274 /* Program traps generated by user-level software must be handled
275 * by the guest kernel. */
276 vcpu->arch.esr = vcpu->arch.fault_esr;
277 kvmppc_queue_exception(vcpu, BOOKE_INTERRUPT_PROGRAM);
282 er = kvmppc_emulate_instruction(run, vcpu);
285 /* Future optimization: only reload non-volatiles if
286 * they were actually modified by emulation. */
287 vcpu->stat.emulated_inst_exits++;
291 run->exit_reason = KVM_EXIT_DCR;
295 /* XXX Deliver Program interrupt to guest. */
296 printk(KERN_CRIT "%s: emulation at %x failed (%08x)\n",
297 __func__, vcpu->arch.pc, vcpu->arch.last_inst);
298 /* For debugging, encode the failing instruction and
299 * report it to userspace. */
300 run->hw.hardware_exit_reason = ~0ULL << 32;
301 run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
309 case BOOKE_INTERRUPT_FP_UNAVAIL:
310 kvmppc_queue_exception(vcpu, exit_nr);
314 case BOOKE_INTERRUPT_DATA_STORAGE:
315 vcpu->arch.dear = vcpu->arch.fault_dear;
316 vcpu->arch.esr = vcpu->arch.fault_esr;
317 kvmppc_queue_exception(vcpu, exit_nr);
318 vcpu->stat.dsi_exits++;
322 case BOOKE_INTERRUPT_INST_STORAGE:
323 vcpu->arch.esr = vcpu->arch.fault_esr;
324 kvmppc_queue_exception(vcpu, exit_nr);
325 vcpu->stat.isi_exits++;
329 case BOOKE_INTERRUPT_SYSCALL:
330 kvmppc_queue_exception(vcpu, exit_nr);
331 vcpu->stat.syscall_exits++;
335 case BOOKE_INTERRUPT_DTLB_MISS: {
337 unsigned long eaddr = vcpu->arch.fault_dear;
340 /* Check the guest TLB. */
341 gtlbe = kvmppc_44x_dtlb_search(vcpu, eaddr);
343 /* The guest didn't have a mapping for it. */
344 kvmppc_queue_exception(vcpu, exit_nr);
345 vcpu->arch.dear = vcpu->arch.fault_dear;
346 vcpu->arch.esr = vcpu->arch.fault_esr;
347 vcpu->stat.dtlb_real_miss_exits++;
352 vcpu->arch.paddr_accessed = tlb_xlate(gtlbe, eaddr);
353 gfn = vcpu->arch.paddr_accessed >> PAGE_SHIFT;
355 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
356 /* The guest TLB had a mapping, but the shadow TLB
357 * didn't, and it is RAM. This could be because:
358 * a) the entry is mapping the host kernel, or
359 * b) the guest used a large mapping which we're faking
360 * Either way, we need to satisfy the fault without
361 * invoking the guest. */
362 kvmppc_mmu_map(vcpu, eaddr, gfn, gtlbe->tid,
364 vcpu->stat.dtlb_virt_miss_exits++;
367 /* Guest has mapped and accessed a page which is not
369 r = kvmppc_emulate_mmio(run, vcpu);
375 case BOOKE_INTERRUPT_ITLB_MISS: {
377 unsigned long eaddr = vcpu->arch.pc;
382 /* Check the guest TLB. */
383 gtlbe = kvmppc_44x_itlb_search(vcpu, eaddr);
385 /* The guest didn't have a mapping for it. */
386 kvmppc_queue_exception(vcpu, exit_nr);
387 vcpu->stat.itlb_real_miss_exits++;
391 vcpu->stat.itlb_virt_miss_exits++;
393 gfn = tlb_xlate(gtlbe, eaddr) >> PAGE_SHIFT;
395 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
396 /* The guest TLB had a mapping, but the shadow TLB
397 * didn't. This could be because:
398 * a) the entry is mapping the host kernel, or
399 * b) the guest used a large mapping which we're faking
400 * Either way, we need to satisfy the fault without
401 * invoking the guest. */
402 kvmppc_mmu_map(vcpu, eaddr, gfn, gtlbe->tid,
405 /* Guest mapped and leaped at non-RAM! */
406 kvmppc_queue_exception(vcpu,
407 BOOKE_INTERRUPT_MACHINE_CHECK);
413 case BOOKE_INTERRUPT_DEBUG: {
416 vcpu->arch.pc = mfspr(SPRN_CSRR0);
418 /* clear IAC events in DBSR register */
419 dbsr = mfspr(SPRN_DBSR);
420 dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
421 mtspr(SPRN_DBSR, dbsr);
423 run->exit_reason = KVM_EXIT_DEBUG;
429 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
435 kvmppc_check_and_deliver_interrupts(vcpu);
437 /* Do some exit accounting. */
438 vcpu->stat.sum_exits++;
439 if (!(r & RESUME_HOST)) {
440 /* To avoid clobbering exit_reason, only check for signals if
441 * we aren't already exiting to userspace for some other
443 if (signal_pending(current)) {
444 run->exit_reason = KVM_EXIT_INTR;
445 r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
447 vcpu->stat.signal_exits++;
449 vcpu->stat.light_exits++;
452 switch (run->exit_reason) {
454 vcpu->stat.mmio_exits++;
457 vcpu->stat.dcr_exits++;
460 vcpu->stat.signal_exits++;
468 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
469 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
471 struct tlbe *tlbe = &vcpu->arch.guest_tlb[0];
474 tlbe->word0 = PPC44x_TLB_16M | PPC44x_TLB_VALID;
476 tlbe->word2 = PPC44x_TLB_SX | PPC44x_TLB_SW | PPC44x_TLB_SR;
480 tlbe->word0 = 0xef600000 | PPC44x_TLB_4K | PPC44x_TLB_VALID;
481 tlbe->word1 = 0xef600000;
482 tlbe->word2 = PPC44x_TLB_SX | PPC44x_TLB_SW | PPC44x_TLB_SR
483 | PPC44x_TLB_I | PPC44x_TLB_G;
487 vcpu->arch.gpr[1] = (16<<20) - 8; /* -8 for the callee-save LR slot */
489 vcpu->arch.shadow_pid = 1;
491 /* Eye-catching number so we know if the guest takes an interrupt
492 * before it's programmed its own IVPR. */
493 vcpu->arch.ivpr = 0x55550000;
495 /* Since the guest can directly access the timebase, it must know the
496 * real timebase frequency. Accordingly, it must see the state of
498 vcpu->arch.ccr1 = mfspr(SPRN_CCR1);
503 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
507 regs->pc = vcpu->arch.pc;
508 regs->cr = vcpu->arch.cr;
509 regs->ctr = vcpu->arch.ctr;
510 regs->lr = vcpu->arch.lr;
511 regs->xer = vcpu->arch.xer;
512 regs->msr = vcpu->arch.msr;
513 regs->srr0 = vcpu->arch.srr0;
514 regs->srr1 = vcpu->arch.srr1;
515 regs->pid = vcpu->arch.pid;
516 regs->sprg0 = vcpu->arch.sprg0;
517 regs->sprg1 = vcpu->arch.sprg1;
518 regs->sprg2 = vcpu->arch.sprg2;
519 regs->sprg3 = vcpu->arch.sprg3;
520 regs->sprg5 = vcpu->arch.sprg4;
521 regs->sprg6 = vcpu->arch.sprg5;
522 regs->sprg7 = vcpu->arch.sprg6;
524 for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
525 regs->gpr[i] = vcpu->arch.gpr[i];
530 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
534 vcpu->arch.pc = regs->pc;
535 vcpu->arch.cr = regs->cr;
536 vcpu->arch.ctr = regs->ctr;
537 vcpu->arch.lr = regs->lr;
538 vcpu->arch.xer = regs->xer;
539 vcpu->arch.msr = regs->msr;
540 vcpu->arch.srr0 = regs->srr0;
541 vcpu->arch.srr1 = regs->srr1;
542 vcpu->arch.sprg0 = regs->sprg0;
543 vcpu->arch.sprg1 = regs->sprg1;
544 vcpu->arch.sprg2 = regs->sprg2;
545 vcpu->arch.sprg3 = regs->sprg3;
546 vcpu->arch.sprg5 = regs->sprg4;
547 vcpu->arch.sprg6 = regs->sprg5;
548 vcpu->arch.sprg7 = regs->sprg6;
550 for (i = 0; i < ARRAY_SIZE(vcpu->arch.gpr); i++)
551 vcpu->arch.gpr[i] = regs->gpr[i];
556 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
557 struct kvm_sregs *sregs)
562 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
563 struct kvm_sregs *sregs)
568 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
573 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
578 /* 'linear_address' is actually an encoding of AS|PID|EADDR . */
579 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
580 struct kvm_translation *tr)
588 eaddr = tr->linear_address;
589 pid = (tr->linear_address >> 32) & 0xff;
590 as = (tr->linear_address >> 40) & 0x1;
592 index = kvmppc_44x_tlb_index(vcpu, eaddr, pid, as);
598 gtlbe = &vcpu->arch.guest_tlb[index];
600 tr->physical_address = tlb_xlate(gtlbe, eaddr);
601 /* XXX what does "writeable" and "usermode" even mean? */