arch/powerpc/kvm/book3s_hv.c

   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 struct kvm_vcpu *kvmppc_find_vcpu(struct kvm *kvm, int id)
 128 {
 129         int r;
 130         struct kvm_vcpu *v, *ret = NULL;
 131
 132         mutex_lock(&kvm->lock);
 133         kvm_for_each_vcpu(r, v, kvm) {
 134                 if (v->vcpu_id == id) {
 135                         ret = v;
 136                         break;
 137                 }
 138         }
 139         mutex_unlock(&kvm->lock);
 140         return ret;
 141 }
 142
 143 static void init_vpa(struct kvm_vcpu *vcpu, struct lppaca *vpa)
 144 {
 145         vpa->shared_proc = 1;
 146         vpa->yield_count = 1;
 147 }
 148
 149 static unsigned long do_h_register_vpa(struct kvm_vcpu *vcpu,
 150                                        unsigned long flags,
 151                                        unsigned long vcpuid, unsigned long vpa)
 152 {
 153         struct kvm *kvm = vcpu->kvm;
 154         unsigned long pg_index, ra, len;
 155         unsigned long pg_offset;
 156         void *va;
 157         struct kvm_vcpu *tvcpu;
 158
 159         tvcpu = kvmppc_find_vcpu(kvm, vcpuid);
 160         if (!tvcpu)
 161                 return H_PARAMETER;
 162
 163         flags >>= 63 - 18;
 164         flags &= 7;
 165         if (flags == 0 || flags == 4)
 166                 return H_PARAMETER;
 167         if (flags < 4) {
 168                 if (vpa & 0x7f)
 169                         return H_PARAMETER;
 170                 /* registering new area; convert logical addr to real */
 171                 pg_index = vpa >> kvm->arch.ram_porder;
 172                 pg_offset = vpa & (kvm->arch.ram_psize - 1);
 173                 if (pg_index >= kvm->arch.ram_npages)
 174                         return H_PARAMETER;
 175                 if (kvm->arch.ram_pginfo[pg_index].pfn == 0)
 176                         return H_PARAMETER;
 177                 ra = kvm->arch.ram_pginfo[pg_index].pfn << PAGE_SHIFT;
 178                 ra |= pg_offset;
 179                 va = __va(ra);
 180                 if (flags <= 1)
 181                         len = *(unsigned short *)(va + 4);
 182                 else
 183                         len = *(unsigned int *)(va + 4);
 184                 if (pg_offset + len > kvm->arch.ram_psize)
 185                         return H_PARAMETER;
 186                 switch (flags) {
 187                 case 1:         /* register VPA */
 188                         if (len < 640)
 189                                 return H_PARAMETER;
 190                         tvcpu->arch.vpa = va;
 191                         init_vpa(vcpu, va);
 192                         break;
 193                 case 2:         /* register DTL */
 194                         if (len < 48)
 195                                 return H_PARAMETER;
 196                         if (!tvcpu->arch.vpa)
 197                                 return H_RESOURCE;
 198                         len -= len % 48;
 199                         tvcpu->arch.dtl = va;
 200                         tvcpu->arch.dtl_end = va + len;
 201                         break;
 202                 case 3:         /* register SLB shadow buffer */
 203                         if (len < 8)
 204                                 return H_PARAMETER;
 205                         if (!tvcpu->arch.vpa)
 206                                 return H_RESOURCE;
 207                         tvcpu->arch.slb_shadow = va;
 208                         len = (len - 16) / 16;
 209                         tvcpu->arch.slb_shadow = va;
 210                         break;
 211                 }
 212         } else {
 213                 switch (flags) {
 214                 case 5:         /* unregister VPA */
 215                         if (tvcpu->arch.slb_shadow || tvcpu->arch.dtl)
 216                                 return H_RESOURCE;
 217                         tvcpu->arch.vpa = NULL;
 218                         break;
 219                 case 6:         /* unregister DTL */
 220                         tvcpu->arch.dtl = NULL;
 221                         break;
 222                 case 7:         /* unregister SLB shadow buffer */
 223                         tvcpu->arch.slb_shadow = NULL;
 224                         break;
 225                 }
 226         }
 227         return H_SUCCESS;
 228 }
 229
 230 int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu)
 231 {
 232         unsigned long req = kvmppc_get_gpr(vcpu, 3);
 233         unsigned long target, ret = H_SUCCESS;
 234         struct kvm_vcpu *tvcpu;
 235
 236         switch (req) {
 237         case H_CEDE:
 238                 vcpu->arch.shregs.msr |= MSR_EE;
 239                 vcpu->arch.ceded = 1;
 240                 smp_mb();
 241                 if (!vcpu->arch.prodded)
 242                         kvmppc_vcpu_block(vcpu);
 243                 else
 244                         vcpu->arch.prodded = 0;
 245                 smp_mb();
 246                 vcpu->arch.ceded = 0;
 247                 break;
 248         case H_PROD:
 249                 target = kvmppc_get_gpr(vcpu, 4);
 250                 tvcpu = kvmppc_find_vcpu(vcpu->kvm, target);
 251                 if (!tvcpu) {
 252                         ret = H_PARAMETER;
 253                         break;
 254                 }
 255                 tvcpu->arch.prodded = 1;
 256                 smp_mb();
 257                 if (vcpu->arch.ceded) {
 258                         if (waitqueue_active(&vcpu->wq)) {
 259                                 wake_up_interruptible(&vcpu->wq);
 260                                 vcpu->stat.halt_wakeup++;
 261                         }
 262                 }
 263                 break;
 264         case H_CONFER:
 265                 break;
 266         case H_REGISTER_VPA:
 267                 ret = do_h_register_vpa(vcpu, kvmppc_get_gpr(vcpu, 4),
 268                                         kvmppc_get_gpr(vcpu, 5),
 269                                         kvmppc_get_gpr(vcpu, 6));
 270                 break;
 271         default:
 272                 return RESUME_HOST;
 273         }
 274         kvmppc_set_gpr(vcpu, 3, ret);
 275         vcpu->arch.hcall_needed = 0;
 276         return RESUME_GUEST;
 277 }
 278
 279 static int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
 280                               struct task_struct *tsk)
 281 {
 282         int r = RESUME_HOST;
 283
 284         vcpu->stat.sum_exits++;
 285
 286         run->exit_reason = KVM_EXIT_UNKNOWN;
 287         run->ready_for_interrupt_injection = 1;
 288         switch (vcpu->arch.trap) {
 289         /* We're good on these - the host merely wanted to get our attention */
 290         case BOOK3S_INTERRUPT_HV_DECREMENTER:
 291                 vcpu->stat.dec_exits++;
 292                 r = RESUME_GUEST;
 293                 break;
 294         case BOOK3S_INTERRUPT_EXTERNAL:
 295                 vcpu->stat.ext_intr_exits++;
 296                 r = RESUME_GUEST;
 297                 break;
 298         case BOOK3S_INTERRUPT_PERFMON:
 299                 r = RESUME_GUEST;
 300                 break;
 301         case BOOK3S_INTERRUPT_PROGRAM:
 302         {
 303                 ulong flags;
 304                 /*
 305                  * Normally program interrupts are delivered directly
 306                  * to the guest by the hardware, but we can get here
 307                  * as a result of a hypervisor emulation interrupt
 308                  * (e40) getting turned into a 700 by BML RTAS.
 309                  */
 310                 flags = vcpu->arch.shregs.msr & 0x1f0000ull;
 311                 kvmppc_core_queue_program(vcpu, flags);
 312                 r = RESUME_GUEST;
 313                 break;
 314         }
 315         case BOOK3S_INTERRUPT_SYSCALL:
 316         {
 317                 /* hcall - punt to userspace */
 318                 int i;
 319
 320                 if (vcpu->arch.shregs.msr & MSR_PR) {
 321                         /* sc 1 from userspace - reflect to guest syscall */
 322                         kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_SYSCALL);
 323                         r = RESUME_GUEST;
 324                         break;
 325                 }
 326                 run->papr_hcall.nr = kvmppc_get_gpr(vcpu, 3);
 327                 for (i = 0; i < 9; ++i)
 328                         run->papr_hcall.args[i] = kvmppc_get_gpr(vcpu, 4 + i);
 329                 run->exit_reason = KVM_EXIT_PAPR_HCALL;
 330                 vcpu->arch.hcall_needed = 1;
 331                 r = RESUME_HOST;
 332                 break;
 333         }
 334         /*
 335          * We get these next two if the guest does a bad real-mode access,
 336          * as we have enabled VRMA (virtualized real mode area) mode in the
 337          * LPCR.  We just generate an appropriate DSI/ISI to the guest.
 338          */
 339         case BOOK3S_INTERRUPT_H_DATA_STORAGE:
 340                 vcpu->arch.shregs.dsisr = vcpu->arch.fault_dsisr;
 341                 vcpu->arch.shregs.dar = vcpu->arch.fault_dar;
 342                 kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
 343                 r = RESUME_GUEST;
 344                 break;
 345         case BOOK3S_INTERRUPT_H_INST_STORAGE:
 346                 kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE,
 347                                         0x08000000);
 348                 r = RESUME_GUEST;
 349                 break;
 350         /*
 351          * This occurs if the guest executes an illegal instruction.
 352          * We just generate a program interrupt to the guest, since
 353          * we don't emulate any guest instructions at this stage.
 354          */
 355         case BOOK3S_INTERRUPT_H_EMUL_ASSIST:
 356                 kvmppc_core_queue_program(vcpu, 0x80000);
 357                 r = RESUME_GUEST;
 358                 break;
 359         default:
 360                 kvmppc_dump_regs(vcpu);
 361                 printk(KERN_EMERG "trap=0x%x | pc=0x%lx | msr=0x%llx\n",
 362                         vcpu->arch.trap, kvmppc_get_pc(vcpu),
 363                         vcpu->arch.shregs.msr);
 364                 r = RESUME_HOST;
 365                 BUG();
 366                 break;
 367         }
 368
 369
 370         if (!(r & RESUME_HOST)) {
 371                 /* To avoid clobbering exit_reason, only check for signals if
 372                  * we aren't already exiting to userspace for some other
 373                  * reason. */
 374                 if (signal_pending(tsk)) {
 375                         vcpu->stat.signal_exits++;
 376                         run->exit_reason = KVM_EXIT_INTR;
 377                         r = -EINTR;
 378                 } else {
 379                         kvmppc_core_deliver_interrupts(vcpu);
 380                 }
 381         }
 382
 383         return r;
 384 }
 385
 386 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 387                                   struct kvm_sregs *sregs)
 388 {
 389         int i;
 390
 391         sregs->pvr = vcpu->arch.pvr;
 392
 393         memset(sregs, 0, sizeof(struct kvm_sregs));
 394         for (i = 0; i < vcpu->arch.slb_max; i++) {
 395                 sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige;
 396                 sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
 397         }
 398
 399         return 0;
 400 }
 401
 402 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 403                                   struct kvm_sregs *sregs)
 404 {
 405         int i, j;
 406
 407         kvmppc_set_pvr(vcpu, sregs->pvr);
 408
 409         j = 0;
 410         for (i = 0; i < vcpu->arch.slb_nr; i++) {
 411                 if (sregs->u.s.ppc64.slb[i].slbe & SLB_ESID_V) {
 412                         vcpu->arch.slb[j].orige = sregs->u.s.ppc64.slb[i].slbe;
 413                         vcpu->arch.slb[j].origv = sregs->u.s.ppc64.slb[i].slbv;
 414                         ++j;
 415                 }
 416         }
 417         vcpu->arch.slb_max = j;
 418
 419         return 0;
 420 }
 421
 422 int kvmppc_core_check_processor_compat(void)
 423 {
 424         if (cpu_has_feature(CPU_FTR_HVMODE_206))
 425                 return 0;
 426         return -EIO;
 427 }
 428
 429 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
 430 {
 431         struct kvm_vcpu *vcpu;
 432         int err = -ENOMEM;
 433         unsigned long lpcr;
 434
 435         vcpu = kzalloc(sizeof(struct kvm_vcpu), GFP_KERNEL);
 436         if (!vcpu)
 437                 goto out;
 438
 439         err = kvm_vcpu_init(vcpu, kvm, id);
 440         if (err)
 441                 goto free_vcpu;
 442
 443         vcpu->arch.shared = &vcpu->arch.shregs;
 444         vcpu->arch.last_cpu = -1;
 445         vcpu->arch.mmcr[0] = MMCR0_FC;
 446         vcpu->arch.ctrl = CTRL_RUNLATCH;
 447         /* default to host PVR, since we can't spoof it */
 448         vcpu->arch.pvr = mfspr(SPRN_PVR);
 449         kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
 450
 451         lpcr = kvm->arch.host_lpcr & (LPCR_PECE | LPCR_LPES);
 452         lpcr |= LPCR_VPM0 | LPCR_VRMA_L | (4UL << LPCR_DPFD_SH) | LPCR_HDICE;
 453         vcpu->arch.lpcr = lpcr;
 454
 455         kvmppc_mmu_book3s_hv_init(vcpu);
 456
 457         return vcpu;
 458
 459 free_vcpu:
 460         kfree(vcpu);
 461 out:
 462         return ERR_PTR(err);
 463 }
 464
 465 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
 466 {
 467         kvm_vcpu_uninit(vcpu);
 468         kfree(vcpu);
 469 }
 470
 471 extern int __kvmppc_vcore_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
 472
 473 static int kvmppc_run_vcpu(struct kvm_run *run, struct kvm_vcpu *vcpu)
 474 {
 475         u64 now;
 476
 477         if (signal_pending(current)) {
 478                 run->exit_reason = KVM_EXIT_INTR;
 479                 return -EINTR;
 480         }
 481
 482         flush_fp_to_thread(current);
 483         flush_altivec_to_thread(current);
 484         flush_vsx_to_thread(current);
 485         preempt_disable();
 486
 487         /*
 488          * Make sure we are running on thread 0, and that
 489          * secondary threads are offline.
 490          * XXX we should also block attempts to bring any
 491          * secondary threads online.
 492          */
 493         if (threads_per_core > 1) {
 494                 int cpu = smp_processor_id();
 495                 int thr = cpu_thread_in_core(cpu);
 496
 497                 if (thr)
 498                         goto out;
 499                 while (++thr < threads_per_core)
 500                         if (cpu_online(cpu + thr))
 501                                 goto out;
 502         }
 503
 504         kvm_guest_enter();
 505
 506         __kvmppc_vcore_entry(NULL, vcpu);
 507
 508         kvm_guest_exit();
 509
 510         preempt_enable();
 511         kvm_resched(vcpu);
 512
 513         now = get_tb();
 514         /* cancel pending dec exception if dec is positive */
 515         if (now < vcpu->arch.dec_expires && kvmppc_core_pending_dec(vcpu))
 516                 kvmppc_core_dequeue_dec(vcpu);
 517
 518         return kvmppc_handle_exit(run, vcpu, current);
 519
 520  out:
 521         preempt_enable();
 522         return -EBUSY;
 523 }
 524
 525 int kvmppc_vcpu_run(struct kvm_run *run, struct kvm_vcpu *vcpu)
 526 {
 527         int r;
 528
 529         do {
 530                 r = kvmppc_run_vcpu(run, vcpu);
 531
 532                 if (run->exit_reason == KVM_EXIT_PAPR_HCALL &&
 533                     !(vcpu->arch.shregs.msr & MSR_PR)) {
 534                         r = kvmppc_pseries_do_hcall(vcpu);
 535                         kvmppc_core_deliver_interrupts(vcpu);
 536                 }
 537         } while (r == RESUME_GUEST);
 538         return r;
 539 }
 540
 541 static long kvmppc_stt_npages(unsigned long window_size)
 542 {
 543         return ALIGN((window_size >> SPAPR_TCE_SHIFT)
 544                      * sizeof(u64), PAGE_SIZE) / PAGE_SIZE;
 545 }
 546
 547 static void release_spapr_tce_table(struct kvmppc_spapr_tce_table *stt)
 548 {
 549         struct kvm *kvm = stt->kvm;
 550         int i;
 551
 552         mutex_lock(&kvm->lock);
 553         list_del(&stt->list);
 554         for (i = 0; i < kvmppc_stt_npages(stt->window_size); i++)
 555                 __free_page(stt->pages[i]);
 556         kfree(stt);
 557         mutex_unlock(&kvm->lock);
 558
 559         kvm_put_kvm(kvm);
 560 }
 561
 562 static int kvm_spapr_tce_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 563 {
 564         struct kvmppc_spapr_tce_table *stt = vma->vm_file->private_data;
 565         struct page *page;
 566
 567         if (vmf->pgoff >= kvmppc_stt_npages(stt->window_size))
 568                 return VM_FAULT_SIGBUS;
 569
 570         page = stt->pages[vmf->pgoff];
 571         get_page(page);
 572         vmf->page = page;
 573         return 0;
 574 }
 575
 576 static const struct vm_operations_struct kvm_spapr_tce_vm_ops = {
 577         .fault = kvm_spapr_tce_fault,
 578 };
 579
 580 static int kvm_spapr_tce_mmap(struct file *file, struct vm_area_struct *vma)
 581 {
 582         vma->vm_ops = &kvm_spapr_tce_vm_ops;
 583         return 0;
 584 }
 585
 586 static int kvm_spapr_tce_release(struct inode *inode, struct file *filp)
 587 {
 588         struct kvmppc_spapr_tce_table *stt = filp->private_data;
 589
 590         release_spapr_tce_table(stt);
 591         return 0;
 592 }
 593
 594 static struct file_operations kvm_spapr_tce_fops = {
 595         .mmap           = kvm_spapr_tce_mmap,
 596         .release        = kvm_spapr_tce_release,
 597 };
 598
 599 long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 600                                    struct kvm_create_spapr_tce *args)
 601 {
 602         struct kvmppc_spapr_tce_table *stt = NULL;
 603         long npages;
 604         int ret = -ENOMEM;
 605         int i;
 606
 607         /* Check this LIOBN hasn't been previously allocated */
 608         list_for_each_entry(stt, &kvm->arch.spapr_tce_tables, list) {
 609                 if (stt->liobn == args->liobn)
 610                         return -EBUSY;
 611         }
 612
 613         npages = kvmppc_stt_npages(args->window_size);
 614
 615         stt = kzalloc(sizeof(*stt) + npages* sizeof(struct page *),
 616                       GFP_KERNEL);
 617         if (!stt)
 618                 goto fail;
 619
 620         stt->liobn = args->liobn;
 621         stt->window_size = args->window_size;
 622         stt->kvm = kvm;
 623
 624         for (i = 0; i < npages; i++) {
 625                 stt->pages[i] = alloc_page(GFP_KERNEL | __GFP_ZERO);
 626                 if (!stt->pages[i])
 627                         goto fail;
 628         }
 629
 630         kvm_get_kvm(kvm);
 631
 632         mutex_lock(&kvm->lock);
 633         list_add(&stt->list, &kvm->arch.spapr_tce_tables);
 634
 635         mutex_unlock(&kvm->lock);
 636
 637         return anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
 638                                 stt, O_RDWR);
 639
 640 fail:
 641         if (stt) {
 642                 for (i = 0; i < npages; i++)
 643                         if (stt->pages[i])
 644                                 __free_page(stt->pages[i]);
 645
 646                 kfree(stt);
 647         }
 648         return ret;
 649 }
 650
 651 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
 652                                 struct kvm_userspace_memory_region *mem)
 653 {
 654         if (mem->guest_phys_addr == 0 && mem->memory_size != 0)
 655                 return kvmppc_prepare_vrma(kvm, mem);
 656         return 0;
 657 }
 658
 659 void kvmppc_core_commit_memory_region(struct kvm *kvm,
 660                                 struct kvm_userspace_memory_region *mem)
 661 {
 662         if (mem->guest_phys_addr == 0 && mem->memory_size != 0)
 663                 kvmppc_map_vrma(kvm, mem);
 664 }
 665
 666 int kvmppc_core_init_vm(struct kvm *kvm)
 667 {
 668         long r;
 669
 670         /* Allocate hashed page table */
 671         r = kvmppc_alloc_hpt(kvm);
 672         if (r)
 673                 return r;
 674
 675         INIT_LIST_HEAD(&kvm->arch.spapr_tce_tables);
 676         return 0;
 677 }
 678
 679 void kvmppc_core_destroy_vm(struct kvm *kvm)
 680 {
 681         kvmppc_free_hpt(kvm);
 682         WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
 683 }
 684
 685 /* These are stubs for now */
 686 void kvmppc_mmu_pte_pflush(struct kvm_vcpu *vcpu, ulong pa_start, ulong pa_end)
 687 {
 688 }
 689
 690 /* We don't need to emulate any privileged instructions or dcbz */
 691 int kvmppc_core_emulate_op(struct kvm_run *run, struct kvm_vcpu *vcpu,
 692                            unsigned int inst, int *advance)
 693 {
 694         return EMULATE_FAIL;
 695 }
 696
 697 int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, int sprn, int rs)
 698 {
 699         return EMULATE_FAIL;
 700 }
 701
 702 int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, int sprn, int rt)
 703 {
 704         return EMULATE_FAIL;
 705 }
 706
 707 static int kvmppc_book3s_hv_init(void)
 708 {
 709         int r;
 710
 711         r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
 712
 713         if (r)
 714                 return r;
 715
 716         r = kvmppc_mmu_hv_init();
 717
 718         return r;
 719 }
 720
 721 static void kvmppc_book3s_hv_exit(void)
 722 {
 723         kvm_exit();
 724 }
 725
 726 module_init(kvmppc_book3s_hv_init);
 727 module_exit(kvmppc_book3s_hv_exit);