Paul Mackerras | de56a94 | 2011-06-29 00:21:34 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright 2011 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com> |
| 3 | * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved. |
| 4 | * |
| 5 | * Authors: |
| 6 | * Paul Mackerras <paulus@au1.ibm.com> |
| 7 | * Alexander Graf <agraf@suse.de> |
| 8 | * Kevin Wolf <mail@kevin-wolf.de> |
| 9 | * |
| 10 | * Description: KVM functions specific to running on Book 3S |
| 11 | * processors in hypervisor mode (specifically POWER7 and later). |
| 12 | * |
| 13 | * This file is derived from arch/powerpc/kvm/book3s.c, |
| 14 | * by Alexander Graf <agraf@suse.de>. |
| 15 | * |
| 16 | * This program is free software; you can redistribute it and/or modify |
| 17 | * it under the terms of the GNU General Public License, version 2, as |
| 18 | * published by the Free Software Foundation. |
| 19 | */ |
| 20 | |
| 21 | #include <linux/kvm_host.h> |
| 22 | #include <linux/err.h> |
| 23 | #include <linux/slab.h> |
| 24 | #include <linux/preempt.h> |
| 25 | #include <linux/sched.h> |
| 26 | #include <linux/delay.h> |
| 27 | #include <linux/fs.h> |
| 28 | #include <linux/anon_inodes.h> |
| 29 | #include <linux/cpumask.h> |
| 30 | |
| 31 | #include <asm/reg.h> |
| 32 | #include <asm/cputable.h> |
| 33 | #include <asm/cacheflush.h> |
| 34 | #include <asm/tlbflush.h> |
| 35 | #include <asm/uaccess.h> |
| 36 | #include <asm/io.h> |
| 37 | #include <asm/kvm_ppc.h> |
| 38 | #include <asm/kvm_book3s.h> |
| 39 | #include <asm/mmu_context.h> |
| 40 | #include <asm/lppaca.h> |
| 41 | #include <asm/processor.h> |
| 42 | #include <linux/gfp.h> |
| 43 | #include <linux/sched.h> |
| 44 | #include <linux/vmalloc.h> |
| 45 | #include <linux/highmem.h> |
| 46 | |
| 47 | /* #define EXIT_DEBUG */ |
| 48 | /* #define EXIT_DEBUG_SIMPLE */ |
| 49 | /* #define EXIT_DEBUG_INT */ |
| 50 | |
| 51 | void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
| 52 | { |
| 53 | local_paca->kvm_hstate.kvm_vcpu = vcpu; |
| 54 | } |
| 55 | |
| 56 | void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu) |
| 57 | { |
| 58 | } |
| 59 | |
| 60 | void kvmppc_vcpu_block(struct kvm_vcpu *vcpu) |
| 61 | { |
| 62 | u64 now; |
| 63 | unsigned long dec_nsec; |
| 64 | |
| 65 | now = get_tb(); |
| 66 | if (now >= vcpu->arch.dec_expires && !kvmppc_core_pending_dec(vcpu)) |
| 67 | kvmppc_core_queue_dec(vcpu); |
| 68 | if (vcpu->arch.pending_exceptions) |
| 69 | return; |
| 70 | if (vcpu->arch.dec_expires != ~(u64)0) { |
| 71 | dec_nsec = (vcpu->arch.dec_expires - now) * NSEC_PER_SEC / |
| 72 | tb_ticks_per_sec; |
| 73 | hrtimer_start(&vcpu->arch.dec_timer, ktime_set(0, dec_nsec), |
| 74 | HRTIMER_MODE_REL); |
| 75 | } |
| 76 | |
| 77 | kvm_vcpu_block(vcpu); |
| 78 | vcpu->stat.halt_wakeup++; |
| 79 | |
| 80 | if (vcpu->arch.dec_expires != ~(u64)0) |
| 81 | hrtimer_try_to_cancel(&vcpu->arch.dec_timer); |
| 82 | } |
| 83 | |
| 84 | void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr) |
| 85 | { |
| 86 | vcpu->arch.shregs.msr = msr; |
| 87 | } |
| 88 | |
| 89 | void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr) |
| 90 | { |
| 91 | vcpu->arch.pvr = pvr; |
| 92 | } |
| 93 | |
| 94 | void kvmppc_dump_regs(struct kvm_vcpu *vcpu) |
| 95 | { |
| 96 | int r; |
| 97 | |
| 98 | pr_err("vcpu %p (%d):\n", vcpu, vcpu->vcpu_id); |
| 99 | pr_err("pc = %.16lx msr = %.16llx trap = %x\n", |
| 100 | vcpu->arch.pc, vcpu->arch.shregs.msr, vcpu->arch.trap); |
| 101 | for (r = 0; r < 16; ++r) |
| 102 | pr_err("r%2d = %.16lx r%d = %.16lx\n", |
| 103 | r, kvmppc_get_gpr(vcpu, r), |
| 104 | r+16, kvmppc_get_gpr(vcpu, r+16)); |
| 105 | pr_err("ctr = %.16lx lr = %.16lx\n", |
| 106 | vcpu->arch.ctr, vcpu->arch.lr); |
| 107 | pr_err("srr0 = %.16llx srr1 = %.16llx\n", |
| 108 | vcpu->arch.shregs.srr0, vcpu->arch.shregs.srr1); |
| 109 | pr_err("sprg0 = %.16llx sprg1 = %.16llx\n", |
| 110 | vcpu->arch.shregs.sprg0, vcpu->arch.shregs.sprg1); |
| 111 | pr_err("sprg2 = %.16llx sprg3 = %.16llx\n", |
| 112 | vcpu->arch.shregs.sprg2, vcpu->arch.shregs.sprg3); |
| 113 | pr_err("cr = %.8x xer = %.16lx dsisr = %.8x\n", |
| 114 | vcpu->arch.cr, vcpu->arch.xer, vcpu->arch.shregs.dsisr); |
| 115 | pr_err("dar = %.16llx\n", vcpu->arch.shregs.dar); |
| 116 | pr_err("fault dar = %.16lx dsisr = %.8x\n", |
| 117 | vcpu->arch.fault_dar, vcpu->arch.fault_dsisr); |
| 118 | pr_err("SLB (%d entries):\n", vcpu->arch.slb_max); |
| 119 | for (r = 0; r < vcpu->arch.slb_max; ++r) |
| 120 | pr_err(" ESID = %.16llx VSID = %.16llx\n", |
| 121 | vcpu->arch.slb[r].orige, vcpu->arch.slb[r].origv); |
| 122 | pr_err("lpcr = %.16lx sdr1 = %.16lx last_inst = %.8x\n", |
| 123 | vcpu->arch.lpcr, vcpu->kvm->arch.sdr1, |
| 124 | vcpu->arch.last_inst); |
| 125 | } |
| 126 | |
| 127 | static int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu, |
| 128 | struct task_struct *tsk) |
| 129 | { |
| 130 | int r = RESUME_HOST; |
| 131 | |
| 132 | vcpu->stat.sum_exits++; |
| 133 | |
| 134 | run->exit_reason = KVM_EXIT_UNKNOWN; |
| 135 | run->ready_for_interrupt_injection = 1; |
| 136 | switch (vcpu->arch.trap) { |
| 137 | /* We're good on these - the host merely wanted to get our attention */ |
| 138 | case BOOK3S_INTERRUPT_HV_DECREMENTER: |
| 139 | vcpu->stat.dec_exits++; |
| 140 | r = RESUME_GUEST; |
| 141 | break; |
| 142 | case BOOK3S_INTERRUPT_EXTERNAL: |
| 143 | vcpu->stat.ext_intr_exits++; |
| 144 | r = RESUME_GUEST; |
| 145 | break; |
| 146 | case BOOK3S_INTERRUPT_PERFMON: |
| 147 | r = RESUME_GUEST; |
| 148 | break; |
| 149 | case BOOK3S_INTERRUPT_PROGRAM: |
| 150 | { |
| 151 | ulong flags; |
| 152 | /* |
| 153 | * Normally program interrupts are delivered directly |
| 154 | * to the guest by the hardware, but we can get here |
| 155 | * as a result of a hypervisor emulation interrupt |
| 156 | * (e40) getting turned into a 700 by BML RTAS. |
| 157 | */ |
| 158 | flags = vcpu->arch.shregs.msr & 0x1f0000ull; |
| 159 | kvmppc_core_queue_program(vcpu, flags); |
| 160 | r = RESUME_GUEST; |
| 161 | break; |
| 162 | } |
| 163 | case BOOK3S_INTERRUPT_SYSCALL: |
| 164 | { |
| 165 | /* hcall - punt to userspace */ |
| 166 | int i; |
| 167 | |
| 168 | if (vcpu->arch.shregs.msr & MSR_PR) { |
| 169 | /* sc 1 from userspace - reflect to guest syscall */ |
| 170 | kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_SYSCALL); |
| 171 | r = RESUME_GUEST; |
| 172 | break; |
| 173 | } |
| 174 | run->papr_hcall.nr = kvmppc_get_gpr(vcpu, 3); |
| 175 | for (i = 0; i < 9; ++i) |
| 176 | run->papr_hcall.args[i] = kvmppc_get_gpr(vcpu, 4 + i); |
| 177 | run->exit_reason = KVM_EXIT_PAPR_HCALL; |
| 178 | vcpu->arch.hcall_needed = 1; |
| 179 | r = RESUME_HOST; |
| 180 | break; |
| 181 | } |
| 182 | /* |
| 183 | * We get these next two if the guest does a bad real-mode access, |
| 184 | * as we have enabled VRMA (virtualized real mode area) mode in the |
| 185 | * LPCR. We just generate an appropriate DSI/ISI to the guest. |
| 186 | */ |
| 187 | case BOOK3S_INTERRUPT_H_DATA_STORAGE: |
| 188 | vcpu->arch.shregs.dsisr = vcpu->arch.fault_dsisr; |
| 189 | vcpu->arch.shregs.dar = vcpu->arch.fault_dar; |
| 190 | kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0); |
| 191 | r = RESUME_GUEST; |
| 192 | break; |
| 193 | case BOOK3S_INTERRUPT_H_INST_STORAGE: |
| 194 | kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, |
| 195 | 0x08000000); |
| 196 | r = RESUME_GUEST; |
| 197 | break; |
| 198 | /* |
| 199 | * This occurs if the guest executes an illegal instruction. |
| 200 | * We just generate a program interrupt to the guest, since |
| 201 | * we don't emulate any guest instructions at this stage. |
| 202 | */ |
| 203 | case BOOK3S_INTERRUPT_H_EMUL_ASSIST: |
| 204 | kvmppc_core_queue_program(vcpu, 0x80000); |
| 205 | r = RESUME_GUEST; |
| 206 | break; |
| 207 | default: |
| 208 | kvmppc_dump_regs(vcpu); |
| 209 | printk(KERN_EMERG "trap=0x%x | pc=0x%lx | msr=0x%llx\n", |
| 210 | vcpu->arch.trap, kvmppc_get_pc(vcpu), |
| 211 | vcpu->arch.shregs.msr); |
| 212 | r = RESUME_HOST; |
| 213 | BUG(); |
| 214 | break; |
| 215 | } |
| 216 | |
| 217 | |
| 218 | if (!(r & RESUME_HOST)) { |
| 219 | /* To avoid clobbering exit_reason, only check for signals if |
| 220 | * we aren't already exiting to userspace for some other |
| 221 | * reason. */ |
| 222 | if (signal_pending(tsk)) { |
| 223 | vcpu->stat.signal_exits++; |
| 224 | run->exit_reason = KVM_EXIT_INTR; |
| 225 | r = -EINTR; |
| 226 | } else { |
| 227 | kvmppc_core_deliver_interrupts(vcpu); |
| 228 | } |
| 229 | } |
| 230 | |
| 231 | return r; |
| 232 | } |
| 233 | |
| 234 | int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, |
| 235 | struct kvm_sregs *sregs) |
| 236 | { |
| 237 | int i; |
| 238 | |
| 239 | sregs->pvr = vcpu->arch.pvr; |
| 240 | |
| 241 | memset(sregs, 0, sizeof(struct kvm_sregs)); |
| 242 | for (i = 0; i < vcpu->arch.slb_max; i++) { |
| 243 | sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige; |
| 244 | sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv; |
| 245 | } |
| 246 | |
| 247 | return 0; |
| 248 | } |
| 249 | |
| 250 | int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, |
| 251 | struct kvm_sregs *sregs) |
| 252 | { |
| 253 | int i, j; |
| 254 | |
| 255 | kvmppc_set_pvr(vcpu, sregs->pvr); |
| 256 | |
| 257 | j = 0; |
| 258 | for (i = 0; i < vcpu->arch.slb_nr; i++) { |
| 259 | if (sregs->u.s.ppc64.slb[i].slbe & SLB_ESID_V) { |
| 260 | vcpu->arch.slb[j].orige = sregs->u.s.ppc64.slb[i].slbe; |
| 261 | vcpu->arch.slb[j].origv = sregs->u.s.ppc64.slb[i].slbv; |
| 262 | ++j; |
| 263 | } |
| 264 | } |
| 265 | vcpu->arch.slb_max = j; |
| 266 | |
| 267 | return 0; |
| 268 | } |
| 269 | |
| 270 | int kvmppc_core_check_processor_compat(void) |
| 271 | { |
| 272 | if (cpu_has_feature(CPU_FTR_HVMODE_206)) |
| 273 | return 0; |
| 274 | return -EIO; |
| 275 | } |
| 276 | |
| 277 | struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id) |
| 278 | { |
| 279 | struct kvm_vcpu *vcpu; |
| 280 | int err = -ENOMEM; |
| 281 | unsigned long lpcr; |
| 282 | |
| 283 | vcpu = kzalloc(sizeof(struct kvm_vcpu), GFP_KERNEL); |
| 284 | if (!vcpu) |
| 285 | goto out; |
| 286 | |
| 287 | err = kvm_vcpu_init(vcpu, kvm, id); |
| 288 | if (err) |
| 289 | goto free_vcpu; |
| 290 | |
| 291 | vcpu->arch.shared = &vcpu->arch.shregs; |
| 292 | vcpu->arch.last_cpu = -1; |
| 293 | vcpu->arch.mmcr[0] = MMCR0_FC; |
| 294 | vcpu->arch.ctrl = CTRL_RUNLATCH; |
| 295 | /* default to host PVR, since we can't spoof it */ |
| 296 | vcpu->arch.pvr = mfspr(SPRN_PVR); |
| 297 | kvmppc_set_pvr(vcpu, vcpu->arch.pvr); |
| 298 | |
| 299 | lpcr = kvm->arch.host_lpcr & (LPCR_PECE | LPCR_LPES); |
| 300 | lpcr |= LPCR_VPM0 | LPCR_VRMA_L | (4UL << LPCR_DPFD_SH) | LPCR_HDICE; |
| 301 | vcpu->arch.lpcr = lpcr; |
| 302 | |
| 303 | kvmppc_mmu_book3s_hv_init(vcpu); |
| 304 | |
| 305 | return vcpu; |
| 306 | |
| 307 | free_vcpu: |
| 308 | kfree(vcpu); |
| 309 | out: |
| 310 | return ERR_PTR(err); |
| 311 | } |
| 312 | |
| 313 | void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu) |
| 314 | { |
| 315 | kvm_vcpu_uninit(vcpu); |
| 316 | kfree(vcpu); |
| 317 | } |
| 318 | |
| 319 | extern int __kvmppc_vcore_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu); |
| 320 | |
| 321 | int kvmppc_vcpu_run(struct kvm_run *run, struct kvm_vcpu *vcpu) |
| 322 | { |
| 323 | u64 now; |
| 324 | |
| 325 | if (signal_pending(current)) { |
| 326 | run->exit_reason = KVM_EXIT_INTR; |
| 327 | return -EINTR; |
| 328 | } |
| 329 | |
| 330 | flush_fp_to_thread(current); |
| 331 | flush_altivec_to_thread(current); |
| 332 | flush_vsx_to_thread(current); |
| 333 | preempt_disable(); |
| 334 | |
| 335 | /* |
| 336 | * Make sure we are running on thread 0, and that |
| 337 | * secondary threads are offline. |
| 338 | * XXX we should also block attempts to bring any |
| 339 | * secondary threads online. |
| 340 | */ |
| 341 | if (threads_per_core > 1) { |
| 342 | int cpu = smp_processor_id(); |
| 343 | int thr = cpu_thread_in_core(cpu); |
| 344 | |
| 345 | if (thr) |
| 346 | goto out; |
| 347 | while (++thr < threads_per_core) |
| 348 | if (cpu_online(cpu + thr)) |
| 349 | goto out; |
| 350 | } |
| 351 | |
| 352 | kvm_guest_enter(); |
| 353 | |
| 354 | __kvmppc_vcore_entry(NULL, vcpu); |
| 355 | |
| 356 | kvm_guest_exit(); |
| 357 | |
| 358 | preempt_enable(); |
| 359 | kvm_resched(vcpu); |
| 360 | |
| 361 | now = get_tb(); |
| 362 | /* cancel pending dec exception if dec is positive */ |
| 363 | if (now < vcpu->arch.dec_expires && kvmppc_core_pending_dec(vcpu)) |
| 364 | kvmppc_core_dequeue_dec(vcpu); |
| 365 | |
| 366 | return kvmppc_handle_exit(run, vcpu, current); |
| 367 | |
| 368 | out: |
| 369 | preempt_enable(); |
| 370 | return -EBUSY; |
| 371 | } |
| 372 | |
| 373 | int kvmppc_core_prepare_memory_region(struct kvm *kvm, |
| 374 | struct kvm_userspace_memory_region *mem) |
| 375 | { |
| 376 | if (mem->guest_phys_addr == 0 && mem->memory_size != 0) |
| 377 | return kvmppc_prepare_vrma(kvm, mem); |
| 378 | return 0; |
| 379 | } |
| 380 | |
| 381 | void kvmppc_core_commit_memory_region(struct kvm *kvm, |
| 382 | struct kvm_userspace_memory_region *mem) |
| 383 | { |
| 384 | if (mem->guest_phys_addr == 0 && mem->memory_size != 0) |
| 385 | kvmppc_map_vrma(kvm, mem); |
| 386 | } |
| 387 | |
| 388 | int kvmppc_core_init_vm(struct kvm *kvm) |
| 389 | { |
| 390 | long r; |
| 391 | |
| 392 | /* Allocate hashed page table */ |
| 393 | r = kvmppc_alloc_hpt(kvm); |
| 394 | |
| 395 | return r; |
| 396 | } |
| 397 | |
| 398 | void kvmppc_core_destroy_vm(struct kvm *kvm) |
| 399 | { |
| 400 | kvmppc_free_hpt(kvm); |
| 401 | } |
| 402 | |
| 403 | /* These are stubs for now */ |
| 404 | void kvmppc_mmu_pte_pflush(struct kvm_vcpu *vcpu, ulong pa_start, ulong pa_end) |
| 405 | { |
| 406 | } |
| 407 | |
| 408 | /* We don't need to emulate any privileged instructions or dcbz */ |
| 409 | int kvmppc_core_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu, |
| 410 | unsigned int inst, int *advance) |
| 411 | { |
| 412 | return EMULATE_FAIL; |
| 413 | } |
| 414 | |
| 415 | int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs) |
| 416 | { |
| 417 | return EMULATE_FAIL; |
| 418 | } |
| 419 | |
| 420 | int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, int sprn, int rt) |
| 421 | { |
| 422 | return EMULATE_FAIL; |
| 423 | } |
| 424 | |
| 425 | static int kvmppc_book3s_hv_init(void) |
| 426 | { |
| 427 | int r; |
| 428 | |
| 429 | r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE); |
| 430 | |
| 431 | if (r) |
| 432 | return r; |
| 433 | |
| 434 | r = kvmppc_mmu_hv_init(); |
| 435 | |
| 436 | return r; |
| 437 | } |
| 438 | |
| 439 | static void kvmppc_book3s_hv_exit(void) |
| 440 | { |
| 441 | kvm_exit(); |
| 442 | } |
| 443 | |
| 444 | module_init(kvmppc_book3s_hv_init); |
| 445 | module_exit(kvmppc_book3s_hv_exit); |