arch/s390/kvm/interrupt.c

   1 /*
   2  * handling kvm guest interrupts
   3  *
   4  * Copyright IBM Corp. 2008,2014
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License (version 2 only)
   8  * as published by the Free Software Foundation.
   9  *
  10  *    Author(s): Carsten Otte <cotte@de.ibm.com>
  11  */
  12
  13 #include <linux/interrupt.h>
  14 #include <linux/kvm_host.h>
  15 #include <linux/hrtimer.h>
  16 #include <linux/mmu_context.h>
  17 #include <linux/signal.h>
  18 #include <linux/slab.h>
  19 #include <linux/bitmap.h>
  20 #include <asm/asm-offsets.h>
  21 #include <asm/uaccess.h>
  22 #include <asm/sclp.h>
  23 #include "kvm-s390.h"
  24 #include "gaccess.h"
  25 #include "trace-s390.h"
  26
  27 #define IOINT_SCHID_MASK 0x0000ffff
  28 #define IOINT_SSID_MASK 0x00030000
  29 #define IOINT_CSSID_MASK 0x03fc0000
  30 #define IOINT_AI_MASK 0x04000000
  31 #define PFAULT_INIT 0x0600
  32 #define PFAULT_DONE 0x0680
  33 #define VIRTIO_PARAM 0x0d00
  34
  35 static int is_ioint(u64 type)
  36 {
  37         return ((type & 0xfffe0000u) != 0xfffe0000u);
  38 }
  39
  40 int psw_extint_disabled(struct kvm_vcpu *vcpu)
  41 {
  42         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
  43 }
  44
  45 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
  46 {
  47         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
  48 }
  49
  50 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
  51 {
  52         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
  53 }
  54
  55 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
  56 {
  57         if ((vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PER) ||
  58             (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO) ||
  59             (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT))
  60                 return 0;
  61         return 1;
  62 }
  63
  64 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
  65 {
  66         if (psw_extint_disabled(vcpu) ||
  67             !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
  68                 return 0;
  69         if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
  70                 /* No timer interrupts when single stepping */
  71                 return 0;
  72         return 1;
  73 }
  74
  75 static u64 int_word_to_isc_bits(u32 int_word)
  76 {
  77         u8 isc = (int_word & 0x38000000) >> 27;
  78
  79         return (0x80 >> isc) << 24;
  80 }
  81
  82 static int __must_check __interrupt_is_deliverable(struct kvm_vcpu *vcpu,
  83                                       struct kvm_s390_interrupt_info *inti)
  84 {
  85         switch (inti->type) {
  86         case KVM_S390_INT_EXTERNAL_CALL:
  87                 if (psw_extint_disabled(vcpu))
  88                         return 0;
  89                 if (vcpu->arch.sie_block->gcr[0] & 0x2000ul)
  90                         return 1;
  91                 return 0;
  92         case KVM_S390_INT_EMERGENCY:
  93                 if (psw_extint_disabled(vcpu))
  94                         return 0;
  95                 if (vcpu->arch.sie_block->gcr[0] & 0x4000ul)
  96                         return 1;
  97                 return 0;
  98         case KVM_S390_INT_CLOCK_COMP:
  99                 return ckc_interrupts_enabled(vcpu);
 100         case KVM_S390_INT_CPU_TIMER:
 101                 if (psw_extint_disabled(vcpu))
 102                         return 0;
 103                 if (vcpu->arch.sie_block->gcr[0] & 0x400ul)
 104                         return 1;
 105                 return 0;
 106         case KVM_S390_INT_SERVICE:
 107         case KVM_S390_INT_PFAULT_INIT:
 108         case KVM_S390_INT_PFAULT_DONE:
 109         case KVM_S390_INT_VIRTIO:
 110                 if (psw_extint_disabled(vcpu))
 111                         return 0;
 112                 if (vcpu->arch.sie_block->gcr[0] & 0x200ul)
 113                         return 1;
 114                 return 0;
 115         case KVM_S390_PROGRAM_INT:
 116         case KVM_S390_SIGP_STOP:
 117         case KVM_S390_SIGP_SET_PREFIX:
 118         case KVM_S390_RESTART:
 119                 return 1;
 120         case KVM_S390_MCHK:
 121                 if (psw_mchk_disabled(vcpu))
 122                         return 0;
 123                 if (vcpu->arch.sie_block->gcr[14] & inti->mchk.cr14)
 124                         return 1;
 125                 return 0;
 126         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 127                 if (psw_ioint_disabled(vcpu))
 128                         return 0;
 129                 if (vcpu->arch.sie_block->gcr[6] &
 130                     int_word_to_isc_bits(inti->io.io_int_word))
 131                         return 1;
 132                 return 0;
 133         default:
 134                 printk(KERN_WARNING "illegal interrupt type %llx\n",
 135                        inti->type);
 136                 BUG();
 137         }
 138         return 0;
 139 }
 140
 141 static inline unsigned long pending_local_irqs(struct kvm_vcpu *vcpu)
 142 {
 143         return vcpu->arch.local_int.pending_irqs;
 144 }
 145
 146 static unsigned long deliverable_local_irqs(struct kvm_vcpu *vcpu)
 147 {
 148         unsigned long active_mask = pending_local_irqs(vcpu);
 149
 150         if (psw_extint_disabled(vcpu))
 151                 active_mask &= ~IRQ_PEND_EXT_MASK;
 152         if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
 153                 __clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
 154         if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
 155                 __clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
 156         if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
 157                 __clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
 158         if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
 159                 __clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
 160         if (psw_mchk_disabled(vcpu))
 161                 active_mask &= ~IRQ_PEND_MCHK_MASK;
 162
 163         /*
 164          * STOP irqs will never be actively delivered. They are triggered via
 165          * intercept requests and cleared when the stop intercept is performed.
 166          */
 167         __clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
 168
 169         return active_mask;
 170 }
 171
 172 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
 173 {
 174         atomic_set_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
 175         set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
 176 }
 177
 178 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
 179 {
 180         atomic_clear_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
 181         clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
 182 }
 183
 184 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
 185 {
 186         atomic_clear_mask(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
 187                           &vcpu->arch.sie_block->cpuflags);
 188         vcpu->arch.sie_block->lctl = 0x0000;
 189         vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
 190
 191         if (guestdbg_enabled(vcpu)) {
 192                 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
 193                                                LCTL_CR10 | LCTL_CR11);
 194                 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
 195         }
 196 }
 197
 198 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
 199 {
 200         atomic_set_mask(flag, &vcpu->arch.sie_block->cpuflags);
 201 }
 202
 203 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
 204 {
 205         if (!(pending_local_irqs(vcpu) & IRQ_PEND_EXT_MASK))
 206                 return;
 207         if (psw_extint_disabled(vcpu))
 208                 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
 209         else
 210                 vcpu->arch.sie_block->lctl |= LCTL_CR0;
 211 }
 212
 213 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
 214 {
 215         if (!(pending_local_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
 216                 return;
 217         if (psw_mchk_disabled(vcpu))
 218                 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
 219         else
 220                 vcpu->arch.sie_block->lctl |= LCTL_CR14;
 221 }
 222
 223 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
 224 {
 225         if (kvm_s390_is_stop_irq_pending(vcpu))
 226                 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
 227 }
 228
 229 /* Set interception request for non-deliverable local interrupts */
 230 static void set_intercept_indicators_local(struct kvm_vcpu *vcpu)
 231 {
 232         set_intercept_indicators_ext(vcpu);
 233         set_intercept_indicators_mchk(vcpu);
 234         set_intercept_indicators_stop(vcpu);
 235 }
 236
 237 static void __set_intercept_indicator(struct kvm_vcpu *vcpu,
 238                                       struct kvm_s390_interrupt_info *inti)
 239 {
 240         switch (inti->type) {
 241         case KVM_S390_INT_SERVICE:
 242         case KVM_S390_INT_PFAULT_DONE:
 243         case KVM_S390_INT_VIRTIO:
 244                 if (psw_extint_disabled(vcpu))
 245                         __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
 246                 else
 247                         vcpu->arch.sie_block->lctl |= LCTL_CR0;
 248                 break;
 249         case KVM_S390_MCHK:
 250                 if (psw_mchk_disabled(vcpu))
 251                         vcpu->arch.sie_block->ictl |= ICTL_LPSW;
 252                 else
 253                         vcpu->arch.sie_block->lctl |= LCTL_CR14;
 254                 break;
 255         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 256                 if (psw_ioint_disabled(vcpu))
 257                         __set_cpuflag(vcpu, CPUSTAT_IO_INT);
 258                 else
 259                         vcpu->arch.sie_block->lctl |= LCTL_CR6;
 260                 break;
 261         default:
 262                 BUG();
 263         }
 264 }
 265
 266 static u16 get_ilc(struct kvm_vcpu *vcpu)
 267 {
 268         const unsigned short table[] = { 2, 4, 4, 6 };
 269
 270         switch (vcpu->arch.sie_block->icptcode) {
 271         case ICPT_INST:
 272         case ICPT_INSTPROGI:
 273         case ICPT_OPEREXC:
 274         case ICPT_PARTEXEC:
 275         case ICPT_IOINST:
 276                 /* last instruction only stored for these icptcodes */
 277                 return table[vcpu->arch.sie_block->ipa >> 14];
 278         case ICPT_PROGI:
 279                 return vcpu->arch.sie_block->pgmilc;
 280         default:
 281                 return 0;
 282         }
 283 }
 284
 285 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
 286 {
 287         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 288         int rc;
 289
 290         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
 291                                          0, 0);
 292
 293         rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
 294                            (u16 *)__LC_EXT_INT_CODE);
 295         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 296         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 297                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 298         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 299                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 300         clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
 301         return rc ? -EFAULT : 0;
 302 }
 303
 304 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
 305 {
 306         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 307         int rc;
 308
 309         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
 310                                          0, 0);
 311
 312         rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
 313                            (u16 __user *)__LC_EXT_INT_CODE);
 314         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 315         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 316                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 317         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 318                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 319         clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
 320         return rc ? -EFAULT : 0;
 321 }
 322
 323 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
 324 {
 325         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 326         struct kvm_s390_ext_info ext;
 327         int rc;
 328
 329         spin_lock(&li->lock);
 330         ext = li->irq.ext;
 331         clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
 332         li->irq.ext.ext_params2 = 0;
 333         spin_unlock(&li->lock);
 334
 335         VCPU_EVENT(vcpu, 4, "interrupt: pfault init parm:%x,parm64:%llx",
 336                    0, ext.ext_params2);
 337         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 338                                          KVM_S390_INT_PFAULT_INIT,
 339                                          0, ext.ext_params2);
 340
 341         rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
 342         rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
 343         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 344                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 345         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 346                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 347         rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
 348         return rc ? -EFAULT : 0;
 349 }
 350
 351 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
 352 {
 353         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 354         struct kvm_s390_mchk_info mchk;
 355         int rc;
 356
 357         spin_lock(&li->lock);
 358         mchk = li->irq.mchk;
 359         /*
 360          * If there was an exigent machine check pending, then any repressible
 361          * machine checks that might have been pending are indicated along
 362          * with it, so always clear both bits
 363          */
 364         clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
 365         clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
 366         memset(&li->irq.mchk, 0, sizeof(mchk));
 367         spin_unlock(&li->lock);
 368
 369         VCPU_EVENT(vcpu, 4, "interrupt: machine check mcic=%llx",
 370                    mchk.mcic);
 371         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_MCHK,
 372                                          mchk.cr14, mchk.mcic);
 373
 374         rc  = kvm_s390_vcpu_store_status(vcpu, KVM_S390_STORE_STATUS_PREFIXED);
 375         rc |= put_guest_lc(vcpu, mchk.mcic,
 376                            (u64 __user *) __LC_MCCK_CODE);
 377         rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
 378                            (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
 379         rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
 380                              &mchk.fixed_logout, sizeof(mchk.fixed_logout));
 381         rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
 382                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 383         rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
 384                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 385         return rc ? -EFAULT : 0;
 386 }
 387
 388 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
 389 {
 390         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 391         int rc;
 392
 393         VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu restart");
 394         vcpu->stat.deliver_restart_signal++;
 395         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
 396
 397         rc  = write_guest_lc(vcpu,
 398                              offsetof(struct _lowcore, restart_old_psw),
 399                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 400         rc |= read_guest_lc(vcpu, offsetof(struct _lowcore, restart_psw),
 401                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 402         clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
 403         return rc ? -EFAULT : 0;
 404 }
 405
 406 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
 407 {
 408         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 409         struct kvm_s390_prefix_info prefix;
 410
 411         spin_lock(&li->lock);
 412         prefix = li->irq.prefix;
 413         li->irq.prefix.address = 0;
 414         clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
 415         spin_unlock(&li->lock);
 416
 417         VCPU_EVENT(vcpu, 4, "interrupt: set prefix to %x", prefix.address);
 418         vcpu->stat.deliver_prefix_signal++;
 419         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 420                                          KVM_S390_SIGP_SET_PREFIX,
 421                                          prefix.address, 0);
 422
 423         kvm_s390_set_prefix(vcpu, prefix.address);
 424         return 0;
 425 }
 426
 427 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
 428 {
 429         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 430         int rc;
 431         int cpu_addr;
 432
 433         spin_lock(&li->lock);
 434         cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
 435         clear_bit(cpu_addr, li->sigp_emerg_pending);
 436         if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
 437                 clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
 438         spin_unlock(&li->lock);
 439
 440         VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp emerg");
 441         vcpu->stat.deliver_emergency_signal++;
 442         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
 443                                          cpu_addr, 0);
 444
 445         rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
 446                            (u16 *)__LC_EXT_INT_CODE);
 447         rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
 448         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 449                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 450         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 451                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 452         return rc ? -EFAULT : 0;
 453 }
 454
 455 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
 456 {
 457         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 458         struct kvm_s390_extcall_info extcall;
 459         int rc;
 460
 461         spin_lock(&li->lock);
 462         extcall = li->irq.extcall;
 463         li->irq.extcall.code = 0;
 464         clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
 465         spin_unlock(&li->lock);
 466
 467         VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp ext call");
 468         vcpu->stat.deliver_external_call++;
 469         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 470                                          KVM_S390_INT_EXTERNAL_CALL,
 471                                          extcall.code, 0);
 472
 473         rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
 474                            (u16 *)__LC_EXT_INT_CODE);
 475         rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
 476         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 477                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 478         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
 479                             sizeof(psw_t));
 480         return rc ? -EFAULT : 0;
 481 }
 482
 483 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
 484 {
 485         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 486         struct kvm_s390_pgm_info pgm_info;
 487         int rc = 0;
 488         u16 ilc = get_ilc(vcpu);
 489
 490         spin_lock(&li->lock);
 491         pgm_info = li->irq.pgm;
 492         clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
 493         memset(&li->irq.pgm, 0, sizeof(pgm_info));
 494         spin_unlock(&li->lock);
 495
 496         VCPU_EVENT(vcpu, 4, "interrupt: pgm check code:%x, ilc:%x",
 497                    pgm_info.code, ilc);
 498         vcpu->stat.deliver_program_int++;
 499         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
 500                                          pgm_info.code, 0);
 501
 502         switch (pgm_info.code & ~PGM_PER) {
 503         case PGM_AFX_TRANSLATION:
 504         case PGM_ASX_TRANSLATION:
 505         case PGM_EX_TRANSLATION:
 506         case PGM_LFX_TRANSLATION:
 507         case PGM_LSTE_SEQUENCE:
 508         case PGM_LSX_TRANSLATION:
 509         case PGM_LX_TRANSLATION:
 510         case PGM_PRIMARY_AUTHORITY:
 511         case PGM_SECONDARY_AUTHORITY:
 512         case PGM_SPACE_SWITCH:
 513                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 514                                   (u64 *)__LC_TRANS_EXC_CODE);
 515                 break;
 516         case PGM_ALEN_TRANSLATION:
 517         case PGM_ALE_SEQUENCE:
 518         case PGM_ASTE_INSTANCE:
 519         case PGM_ASTE_SEQUENCE:
 520         case PGM_ASTE_VALIDITY:
 521         case PGM_EXTENDED_AUTHORITY:
 522                 rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
 523                                   (u8 *)__LC_EXC_ACCESS_ID);
 524                 break;
 525         case PGM_ASCE_TYPE:
 526         case PGM_PAGE_TRANSLATION:
 527         case PGM_REGION_FIRST_TRANS:
 528         case PGM_REGION_SECOND_TRANS:
 529         case PGM_REGION_THIRD_TRANS:
 530         case PGM_SEGMENT_TRANSLATION:
 531                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 532                                   (u64 *)__LC_TRANS_EXC_CODE);
 533                 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
 534                                    (u8 *)__LC_EXC_ACCESS_ID);
 535                 rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
 536                                    (u8 *)__LC_OP_ACCESS_ID);
 537                 break;
 538         case PGM_MONITOR:
 539                 rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
 540                                   (u16 *)__LC_MON_CLASS_NR);
 541                 rc |= put_guest_lc(vcpu, pgm_info.mon_code,
 542                                    (u64 *)__LC_MON_CODE);
 543                 break;
 544         case PGM_DATA:
 545                 rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
 546                                   (u32 *)__LC_DATA_EXC_CODE);
 547                 break;
 548         case PGM_PROTECTION:
 549                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 550                                   (u64 *)__LC_TRANS_EXC_CODE);
 551                 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
 552                                    (u8 *)__LC_EXC_ACCESS_ID);
 553                 break;
 554         }
 555
 556         if (pgm_info.code & PGM_PER) {
 557                 rc |= put_guest_lc(vcpu, pgm_info.per_code,
 558                                    (u8 *) __LC_PER_CODE);
 559                 rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
 560                                    (u8 *)__LC_PER_ATMID);
 561                 rc |= put_guest_lc(vcpu, pgm_info.per_address,
 562                                    (u64 *) __LC_PER_ADDRESS);
 563                 rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
 564                                    (u8 *) __LC_PER_ACCESS_ID);
 565         }
 566
 567         rc |= put_guest_lc(vcpu, ilc, (u16 *) __LC_PGM_ILC);
 568         rc |= put_guest_lc(vcpu, pgm_info.code,
 569                            (u16 *)__LC_PGM_INT_CODE);
 570         rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
 571                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 572         rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
 573                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 574         return rc ? -EFAULT : 0;
 575 }
 576
 577 static int __must_check __deliver_service(struct kvm_vcpu *vcpu,
 578                                           struct kvm_s390_interrupt_info *inti)
 579 {
 580         int rc;
 581
 582         VCPU_EVENT(vcpu, 4, "interrupt: sclp parm:%x",
 583                    inti->ext.ext_params);
 584         vcpu->stat.deliver_service_signal++;
 585         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
 586                                          inti->ext.ext_params, 0);
 587
 588         rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
 589         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 590         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 591                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 592         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 593                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 594         rc |= put_guest_lc(vcpu, inti->ext.ext_params,
 595                            (u32 *)__LC_EXT_PARAMS);
 596         return rc ? -EFAULT : 0;
 597 }
 598
 599 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu,
 600                                            struct kvm_s390_interrupt_info *inti)
 601 {
 602         int rc;
 603
 604         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 605                                          KVM_S390_INT_PFAULT_DONE, 0,
 606                                          inti->ext.ext_params2);
 607
 608         rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *)__LC_EXT_INT_CODE);
 609         rc |= put_guest_lc(vcpu, PFAULT_DONE, (u16 *)__LC_EXT_CPU_ADDR);
 610         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 611                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 612         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 613                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 614         rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
 615                            (u64 *)__LC_EXT_PARAMS2);
 616         return rc ? -EFAULT : 0;
 617 }
 618
 619 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu,
 620                                          struct kvm_s390_interrupt_info *inti)
 621 {
 622         int rc;
 623
 624         VCPU_EVENT(vcpu, 4, "interrupt: virtio parm:%x,parm64:%llx",
 625                    inti->ext.ext_params, inti->ext.ext_params2);
 626         vcpu->stat.deliver_virtio_interrupt++;
 627         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
 628                                          inti->ext.ext_params,
 629                                          inti->ext.ext_params2);
 630
 631         rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *)__LC_EXT_INT_CODE);
 632         rc |= put_guest_lc(vcpu, VIRTIO_PARAM, (u16 *)__LC_EXT_CPU_ADDR);
 633         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 634                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 635         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 636                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 637         rc |= put_guest_lc(vcpu, inti->ext.ext_params,
 638                            (u32 *)__LC_EXT_PARAMS);
 639         rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
 640                            (u64 *)__LC_EXT_PARAMS2);
 641         return rc ? -EFAULT : 0;
 642 }
 643
 644 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
 645                                      struct kvm_s390_interrupt_info *inti)
 646 {
 647         int rc;
 648
 649         VCPU_EVENT(vcpu, 4, "interrupt: I/O %llx", inti->type);
 650         vcpu->stat.deliver_io_int++;
 651         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
 652                                          ((__u32)inti->io.subchannel_id << 16) |
 653                                                 inti->io.subchannel_nr,
 654                                          ((__u64)inti->io.io_int_parm << 32) |
 655                                                 inti->io.io_int_word);
 656
 657         rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
 658                            (u16 *)__LC_SUBCHANNEL_ID);
 659         rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
 660                            (u16 *)__LC_SUBCHANNEL_NR);
 661         rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
 662                            (u32 *)__LC_IO_INT_PARM);
 663         rc |= put_guest_lc(vcpu, inti->io.io_int_word,
 664                            (u32 *)__LC_IO_INT_WORD);
 665         rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
 666                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 667         rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
 668                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 669         return rc ? -EFAULT : 0;
 670 }
 671
 672 static int __must_check __deliver_mchk_floating(struct kvm_vcpu *vcpu,
 673                                            struct kvm_s390_interrupt_info *inti)
 674 {
 675         struct kvm_s390_mchk_info *mchk = &inti->mchk;
 676         int rc;
 677
 678         VCPU_EVENT(vcpu, 4, "interrupt: machine check mcic=%llx",
 679                    mchk->mcic);
 680         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_MCHK,
 681                                          mchk->cr14, mchk->mcic);
 682
 683         rc  = kvm_s390_vcpu_store_status(vcpu, KVM_S390_STORE_STATUS_PREFIXED);
 684         rc |= put_guest_lc(vcpu, mchk->mcic,
 685                         (u64 __user *) __LC_MCCK_CODE);
 686         rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
 687                         (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
 688         rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
 689                              &mchk->fixed_logout, sizeof(mchk->fixed_logout));
 690         rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
 691                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 692         rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
 693                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 694         return rc ? -EFAULT : 0;
 695 }
 696
 697 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
 698
 699 static const deliver_irq_t deliver_irq_funcs[] = {
 700         [IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
 701         [IRQ_PEND_PROG]           = __deliver_prog,
 702         [IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
 703         [IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
 704         [IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
 705         [IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
 706         [IRQ_PEND_RESTART]        = __deliver_restart,
 707         [IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
 708         [IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
 709 };
 710
 711 static int __must_check __deliver_floating_interrupt(struct kvm_vcpu *vcpu,
 712                                            struct kvm_s390_interrupt_info *inti)
 713 {
 714         int rc;
 715
 716         switch (inti->type) {
 717         case KVM_S390_INT_SERVICE:
 718                 rc = __deliver_service(vcpu, inti);
 719                 break;
 720         case KVM_S390_INT_PFAULT_DONE:
 721                 rc = __deliver_pfault_done(vcpu, inti);
 722                 break;
 723         case KVM_S390_INT_VIRTIO:
 724                 rc = __deliver_virtio(vcpu, inti);
 725                 break;
 726         case KVM_S390_MCHK:
 727                 rc = __deliver_mchk_floating(vcpu, inti);
 728                 break;
 729         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
 730                 rc = __deliver_io(vcpu, inti);
 731                 break;
 732         default:
 733                 BUG();
 734         }
 735
 736         return rc;
 737 }
 738
 739 /* Check whether an external call is pending (deliverable or not) */
 740 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
 741 {
 742         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 743         uint8_t sigp_ctrl = vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
 744
 745         if (!sclp_has_sigpif())
 746                 return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
 747
 748         return (sigp_ctrl & SIGP_CTRL_C) &&
 749                (atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND);
 750 }
 751
 752 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
 753 {
 754         struct kvm_s390_float_interrupt *fi = vcpu->arch.local_int.float_int;
 755         struct kvm_s390_interrupt_info  *inti;
 756         int rc;
 757
 758         rc = !!deliverable_local_irqs(vcpu);
 759
 760         if ((!rc) && atomic_read(&fi->active)) {
 761                 spin_lock(&fi->lock);
 762                 list_for_each_entry(inti, &fi->list, list)
 763                         if (__interrupt_is_deliverable(vcpu, inti)) {
 764                                 rc = 1;
 765                                 break;
 766                         }
 767                 spin_unlock(&fi->lock);
 768         }
 769
 770         if (!rc && kvm_cpu_has_pending_timer(vcpu))
 771                 rc = 1;
 772
 773         /* external call pending and deliverable */
 774         if (!rc && kvm_s390_ext_call_pending(vcpu) &&
 775             !psw_extint_disabled(vcpu) &&
 776             (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
 777                 rc = 1;
 778
 779         if (!rc && !exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
 780                 rc = 1;
 781
 782         return rc;
 783 }
 784
 785 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 786 {
 787         if (!(vcpu->arch.sie_block->ckc <
 788               get_tod_clock_fast() + vcpu->arch.sie_block->epoch))
 789                 return 0;
 790         if (!ckc_interrupts_enabled(vcpu))
 791                 return 0;
 792         return 1;
 793 }
 794
 795 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
 796 {
 797         u64 now, sltime;
 798
 799         vcpu->stat.exit_wait_state++;
 800
 801         /* fast path */
 802         if (kvm_cpu_has_pending_timer(vcpu) || kvm_arch_vcpu_runnable(vcpu))
 803                 return 0;
 804
 805         if (psw_interrupts_disabled(vcpu)) {
 806                 VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
 807                 return -EOPNOTSUPP; /* disabled wait */
 808         }
 809
 810         if (!ckc_interrupts_enabled(vcpu)) {
 811                 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
 812                 __set_cpu_idle(vcpu);
 813                 goto no_timer;
 814         }
 815
 816         now = get_tod_clock_fast() + vcpu->arch.sie_block->epoch;
 817         sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
 818
 819         /* underflow */
 820         if (vcpu->arch.sie_block->ckc < now)
 821                 return 0;
 822
 823         __set_cpu_idle(vcpu);
 824         hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
 825         VCPU_EVENT(vcpu, 5, "enabled wait via clock comparator: %llx ns", sltime);
 826 no_timer:
 827         srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 828         kvm_vcpu_block(vcpu);
 829         __unset_cpu_idle(vcpu);
 830         vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 831
 832         hrtimer_cancel(&vcpu->arch.ckc_timer);
 833         return 0;
 834 }
 835
 836 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
 837 {
 838         if (waitqueue_active(&vcpu->wq)) {
 839                 /*
 840                  * The vcpu gave up the cpu voluntarily, mark it as a good
 841                  * yield-candidate.
 842                  */
 843                 vcpu->preempted = true;
 844                 wake_up_interruptible(&vcpu->wq);
 845                 vcpu->stat.halt_wakeup++;
 846         }
 847 }
 848
 849 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
 850 {
 851         struct kvm_vcpu *vcpu;
 852         u64 now, sltime;
 853
 854         vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
 855         now = get_tod_clock_fast() + vcpu->arch.sie_block->epoch;
 856         sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
 857
 858         /*
 859          * If the monotonic clock runs faster than the tod clock we might be
 860          * woken up too early and have to go back to sleep to avoid deadlocks.
 861          */
 862         if (vcpu->arch.sie_block->ckc > now &&
 863             hrtimer_forward_now(timer, ns_to_ktime(sltime)))
 864                 return HRTIMER_RESTART;
 865         kvm_s390_vcpu_wakeup(vcpu);
 866         return HRTIMER_NORESTART;
 867 }
 868
 869 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
 870 {
 871         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 872
 873         spin_lock(&li->lock);
 874         li->pending_irqs = 0;
 875         bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
 876         memset(&li->irq, 0, sizeof(li->irq));
 877         spin_unlock(&li->lock);
 878
 879         /* clear pending external calls set by sigp interpretation facility */
 880         atomic_clear_mask(CPUSTAT_ECALL_PEND, li->cpuflags);
 881         vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl = 0;
 882 }
 883
 884 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
 885 {
 886         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 887         struct kvm_s390_float_interrupt *fi = vcpu->arch.local_int.float_int;
 888         struct kvm_s390_interrupt_info  *n, *inti = NULL;
 889         deliver_irq_t func;
 890         int deliver;
 891         int rc = 0;
 892         unsigned long irq_type;
 893         unsigned long deliverable_irqs;
 894
 895         __reset_intercept_indicators(vcpu);
 896
 897         /* pending ckc conditions might have been invalidated */
 898         clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
 899         if (kvm_cpu_has_pending_timer(vcpu))
 900                 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
 901
 902         do {
 903                 deliverable_irqs = deliverable_local_irqs(vcpu);
 904                 /* bits are in the order of interrupt priority */
 905                 irq_type = find_first_bit(&deliverable_irqs, IRQ_PEND_COUNT);
 906                 if (irq_type == IRQ_PEND_COUNT)
 907                         break;
 908                 func = deliver_irq_funcs[irq_type];
 909                 if (!func) {
 910                         WARN_ON_ONCE(func == NULL);
 911                         clear_bit(irq_type, &li->pending_irqs);
 912                         continue;
 913                 }
 914                 rc = func(vcpu);
 915         } while (!rc && irq_type != IRQ_PEND_COUNT);
 916
 917         set_intercept_indicators_local(vcpu);
 918
 919         if (!rc && atomic_read(&fi->active)) {
 920                 do {
 921                         deliver = 0;
 922                         spin_lock(&fi->lock);
 923                         list_for_each_entry_safe(inti, n, &fi->list, list) {
 924                                 if (__interrupt_is_deliverable(vcpu, inti)) {
 925                                         list_del(&inti->list);
 926                                         fi->irq_count--;
 927                                         deliver = 1;
 928                                         break;
 929                                 }
 930                                 __set_intercept_indicator(vcpu, inti);
 931                         }
 932                         if (list_empty(&fi->list))
 933                                 atomic_set(&fi->active, 0);
 934                         spin_unlock(&fi->lock);
 935                         if (deliver) {
 936                                 rc = __deliver_floating_interrupt(vcpu, inti);
 937                                 kfree(inti);
 938                         }
 939                 } while (!rc && deliver);
 940         }
 941
 942         return rc;
 943 }
 944
 945 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
 946 {
 947         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 948
 949         li->irq.pgm = irq->u.pgm;
 950         set_bit(IRQ_PEND_PROG, &li->pending_irqs);
 951         return 0;
 952 }
 953
 954 int kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code)
 955 {
 956         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 957         struct kvm_s390_irq irq;
 958
 959         VCPU_EVENT(vcpu, 3, "inject: program check %d (from kernel)", code);
 960         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT, code,
 961                                    0, 1);
 962         spin_lock(&li->lock);
 963         irq.u.pgm.code = code;
 964         __inject_prog(vcpu, &irq);
 965         BUG_ON(waitqueue_active(li->wq));
 966         spin_unlock(&li->lock);
 967         return 0;
 968 }
 969
 970 int kvm_s390_inject_prog_irq(struct kvm_vcpu *vcpu,
 971                              struct kvm_s390_pgm_info *pgm_info)
 972 {
 973         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 974         struct kvm_s390_irq irq;
 975         int rc;
 976
 977         VCPU_EVENT(vcpu, 3, "inject: prog irq %d (from kernel)",
 978                    pgm_info->code);
 979         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
 980                                    pgm_info->code, 0, 1);
 981         spin_lock(&li->lock);
 982         irq.u.pgm = *pgm_info;
 983         rc = __inject_prog(vcpu, &irq);
 984         BUG_ON(waitqueue_active(li->wq));
 985         spin_unlock(&li->lock);
 986         return rc;
 987 }
 988
 989 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
 990 {
 991         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 992
 993         VCPU_EVENT(vcpu, 3, "inject: external irq params:%x, params2:%llx",
 994                    irq->u.ext.ext_params, irq->u.ext.ext_params2);
 995         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
 996                                    irq->u.ext.ext_params,
 997                                    irq->u.ext.ext_params2, 2);
 998
 999         li->irq.ext = irq->u.ext;
1000         set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1001         atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1002         return 0;
1003 }
1004
1005 static int __inject_extcall_sigpif(struct kvm_vcpu *vcpu, uint16_t src_id)
1006 {
1007         unsigned char new_val, old_val;
1008         uint8_t *sigp_ctrl = &vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
1009
1010         new_val = SIGP_CTRL_C | (src_id & SIGP_CTRL_SCN_MASK);
1011         old_val = *sigp_ctrl & ~SIGP_CTRL_C;
1012         if (cmpxchg(sigp_ctrl, old_val, new_val) != old_val) {
1013                 /* another external call is pending */
1014                 return -EBUSY;
1015         }
1016         atomic_set_mask(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
1017         return 0;
1018 }
1019
1020 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1021 {
1022         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1023         struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1024         uint16_t src_id = irq->u.extcall.code;
1025
1026         VCPU_EVENT(vcpu, 3, "inject: external call source-cpu:%u",
1027                    src_id);
1028         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1029                                    src_id, 0, 2);
1030
1031         /* sending vcpu invalid */
1032         if (src_id >= KVM_MAX_VCPUS ||
1033             kvm_get_vcpu(vcpu->kvm, src_id) == NULL)
1034                 return -EINVAL;
1035
1036         if (sclp_has_sigpif())
1037                 return __inject_extcall_sigpif(vcpu, src_id);
1038
1039         if (!test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1040                 return -EBUSY;
1041         *extcall = irq->u.extcall;
1042         atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1043         return 0;
1044 }
1045
1046 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1047 {
1048         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1049         struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1050
1051         VCPU_EVENT(vcpu, 3, "inject: set prefix to %x (from user)",
1052                    irq->u.prefix.address);
1053         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1054                                    irq->u.prefix.address, 0, 2);
1055
1056         if (!is_vcpu_stopped(vcpu))
1057                 return -EBUSY;
1058
1059         *prefix = irq->u.prefix;
1060         set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1061         return 0;
1062 }
1063
1064 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1065 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1066 {
1067         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1068         struct kvm_s390_stop_info *stop = &li->irq.stop;
1069         int rc = 0;
1070
1071         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0, 2);
1072
1073         if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1074                 return -EINVAL;
1075
1076         if (is_vcpu_stopped(vcpu)) {
1077                 if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1078                         rc = kvm_s390_store_status_unloaded(vcpu,
1079                                                 KVM_S390_STORE_STATUS_NOADDR);
1080                 return rc;
1081         }
1082
1083         if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1084                 return -EBUSY;
1085         stop->flags = irq->u.stop.flags;
1086         __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1087         return 0;
1088 }
1089
1090 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1091                                  struct kvm_s390_irq *irq)
1092 {
1093         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1094
1095         VCPU_EVENT(vcpu, 3, "inject: restart type %llx", irq->type);
1096         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0, 2);
1097
1098         set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1099         return 0;
1100 }
1101
1102 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1103                                    struct kvm_s390_irq *irq)
1104 {
1105         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1106
1107         VCPU_EVENT(vcpu, 3, "inject: emergency %u\n",
1108                    irq->u.emerg.code);
1109         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1110                                    irq->u.emerg.code, 0, 2);
1111
1112         set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1113         set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1114         atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1115         return 0;
1116 }
1117
1118 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1119 {
1120         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1121         struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1122
1123         VCPU_EVENT(vcpu, 5, "inject: machine check parm64:%llx",
1124                    irq->u.mchk.mcic);
1125         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1126                                    irq->u.mchk.mcic, 2);
1127
1128         /*
1129          * Because repressible machine checks can be indicated along with
1130          * exigent machine checks (PoP, Chapter 11, Interruption action)
1131          * we need to combine cr14, mcic and external damage code.
1132          * Failing storage address and the logout area should not be or'ed
1133          * together, we just indicate the last occurrence of the corresponding
1134          * machine check
1135          */
1136         mchk->cr14 |= irq->u.mchk.cr14;
1137         mchk->mcic |= irq->u.mchk.mcic;
1138         mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1139         mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1140         memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1141                sizeof(mchk->fixed_logout));
1142         if (mchk->mcic & MCHK_EX_MASK)
1143                 set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1144         else if (mchk->mcic & MCHK_REP_MASK)
1145                 set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1146         return 0;
1147 }
1148
1149 static int __inject_ckc(struct kvm_vcpu *vcpu)
1150 {
1151         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1152
1153         VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CLOCK_COMP);
1154         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1155                                    0, 0, 2);
1156
1157         set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1158         atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1159         return 0;
1160 }
1161
1162 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1163 {
1164         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1165
1166         VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CPU_TIMER);
1167         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1168                                    0, 0, 2);
1169
1170         set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1171         atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1172         return 0;
1173 }
1174
1175
1176 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1177                                                     u64 cr6, u64 schid)
1178 {
1179         struct kvm_s390_float_interrupt *fi;
1180         struct kvm_s390_interrupt_info *inti, *iter;
1181
1182         if ((!schid && !cr6) || (schid && cr6))
1183                 return NULL;
1184         fi = &kvm->arch.float_int;
1185         spin_lock(&fi->lock);
1186         inti = NULL;
1187         list_for_each_entry(iter, &fi->list, list) {
1188                 if (!is_ioint(iter->type))
1189                         continue;
1190                 if (cr6 &&
1191                     ((cr6 & int_word_to_isc_bits(iter->io.io_int_word)) == 0))
1192                         continue;
1193                 if (schid) {
1194                         if (((schid & 0x00000000ffff0000) >> 16) !=
1195                             iter->io.subchannel_id)
1196                                 continue;
1197                         if ((schid & 0x000000000000ffff) !=
1198                             iter->io.subchannel_nr)
1199                                 continue;
1200                 }
1201                 inti = iter;
1202                 break;
1203         }
1204         if (inti) {
1205                 list_del_init(&inti->list);
1206                 fi->irq_count--;
1207         }
1208         if (list_empty(&fi->list))
1209                 atomic_set(&fi->active, 0);
1210         spin_unlock(&fi->lock);
1211         return inti;
1212 }
1213
1214 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1215 {
1216         struct kvm_s390_local_interrupt *li;
1217         struct kvm_s390_float_interrupt *fi;
1218         struct kvm_s390_interrupt_info *iter;
1219         struct kvm_vcpu *dst_vcpu = NULL;
1220         int sigcpu;
1221         int rc = 0;
1222
1223         fi = &kvm->arch.float_int;
1224         spin_lock(&fi->lock);
1225         if (fi->irq_count >= KVM_S390_MAX_FLOAT_IRQS) {
1226                 rc = -EINVAL;
1227                 goto unlock_fi;
1228         }
1229         fi->irq_count++;
1230         if (!is_ioint(inti->type)) {
1231                 list_add_tail(&inti->list, &fi->list);
1232         } else {
1233                 u64 isc_bits = int_word_to_isc_bits(inti->io.io_int_word);
1234
1235                 /* Keep I/O interrupts sorted in isc order. */
1236                 list_for_each_entry(iter, &fi->list, list) {
1237                         if (!is_ioint(iter->type))
1238                                 continue;
1239                         if (int_word_to_isc_bits(iter->io.io_int_word)
1240                             <= isc_bits)
1241                                 continue;
1242                         break;
1243                 }
1244                 list_add_tail(&inti->list, &iter->list);
1245         }
1246         atomic_set(&fi->active, 1);
1247         if (atomic_read(&kvm->online_vcpus) == 0)
1248                 goto unlock_fi;
1249         sigcpu = find_first_bit(fi->idle_mask, KVM_MAX_VCPUS);
1250         if (sigcpu == KVM_MAX_VCPUS) {
1251                 do {
1252                         sigcpu = fi->next_rr_cpu++;
1253                         if (sigcpu == KVM_MAX_VCPUS)
1254                                 sigcpu = fi->next_rr_cpu = 0;
1255                 } while (kvm_get_vcpu(kvm, sigcpu) == NULL);
1256         }
1257         dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1258         li = &dst_vcpu->arch.local_int;
1259         spin_lock(&li->lock);
1260         switch (inti->type) {
1261         case KVM_S390_MCHK:
1262                 atomic_set_mask(CPUSTAT_STOP_INT, li->cpuflags);
1263                 break;
1264         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1265                 atomic_set_mask(CPUSTAT_IO_INT, li->cpuflags);
1266                 break;
1267         default:
1268                 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1269                 break;
1270         }
1271         spin_unlock(&li->lock);
1272         kvm_s390_vcpu_wakeup(kvm_get_vcpu(kvm, sigcpu));
1273 unlock_fi:
1274         spin_unlock(&fi->lock);
1275         return rc;
1276 }
1277
1278 int kvm_s390_inject_vm(struct kvm *kvm,
1279                        struct kvm_s390_interrupt *s390int)
1280 {
1281         struct kvm_s390_interrupt_info *inti;
1282         int rc;
1283
1284         inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1285         if (!inti)
1286                 return -ENOMEM;
1287
1288         inti->type = s390int->type;
1289         switch (inti->type) {
1290         case KVM_S390_INT_VIRTIO:
1291                 VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1292                          s390int->parm, s390int->parm64);
1293                 inti->ext.ext_params = s390int->parm;
1294                 inti->ext.ext_params2 = s390int->parm64;
1295                 break;
1296         case KVM_S390_INT_SERVICE:
1297                 VM_EVENT(kvm, 5, "inject: sclp parm:%x", s390int->parm);
1298                 inti->ext.ext_params = s390int->parm;
1299                 break;
1300         case KVM_S390_INT_PFAULT_DONE:
1301                 inti->ext.ext_params2 = s390int->parm64;
1302                 break;
1303         case KVM_S390_MCHK:
1304                 VM_EVENT(kvm, 5, "inject: machine check parm64:%llx",
1305                          s390int->parm64);
1306                 inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1307                 inti->mchk.mcic = s390int->parm64;
1308                 break;
1309         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1310                 if (inti->type & IOINT_AI_MASK)
1311                         VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
1312                 else
1313                         VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
1314                                  s390int->type & IOINT_CSSID_MASK,
1315                                  s390int->type & IOINT_SSID_MASK,
1316                                  s390int->type & IOINT_SCHID_MASK);
1317                 inti->io.subchannel_id = s390int->parm >> 16;
1318                 inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1319                 inti->io.io_int_parm = s390int->parm64 >> 32;
1320                 inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1321                 break;
1322         default:
1323                 kfree(inti);
1324                 return -EINVAL;
1325         }
1326         trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1327                                  2);
1328
1329         rc = __inject_vm(kvm, inti);
1330         if (rc)
1331                 kfree(inti);
1332         return rc;
1333 }
1334
1335 void kvm_s390_reinject_io_int(struct kvm *kvm,
1336                               struct kvm_s390_interrupt_info *inti)
1337 {
1338         __inject_vm(kvm, inti);
1339 }
1340
1341 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1342                        struct kvm_s390_irq *irq)
1343 {
1344         irq->type = s390int->type;
1345         switch (irq->type) {
1346         case KVM_S390_PROGRAM_INT:
1347                 if (s390int->parm & 0xffff0000)
1348                         return -EINVAL;
1349                 irq->u.pgm.code = s390int->parm;
1350                 break;
1351         case KVM_S390_SIGP_SET_PREFIX:
1352                 irq->u.prefix.address = s390int->parm;
1353                 break;
1354         case KVM_S390_SIGP_STOP:
1355                 irq->u.stop.flags = s390int->parm;
1356                 break;
1357         case KVM_S390_INT_EXTERNAL_CALL:
1358                 if (s390int->parm & 0xffff0000)
1359                         return -EINVAL;
1360                 irq->u.extcall.code = s390int->parm;
1361                 break;
1362         case KVM_S390_INT_EMERGENCY:
1363                 if (s390int->parm & 0xffff0000)
1364                         return -EINVAL;
1365                 irq->u.emerg.code = s390int->parm;
1366                 break;
1367         case KVM_S390_MCHK:
1368                 irq->u.mchk.mcic = s390int->parm64;
1369                 break;
1370         }
1371         return 0;
1372 }
1373
1374 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1375 {
1376         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1377
1378         return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1379 }
1380
1381 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1382 {
1383         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1384
1385         spin_lock(&li->lock);
1386         li->irq.stop.flags = 0;
1387         clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1388         spin_unlock(&li->lock);
1389 }
1390
1391 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1392 {
1393         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1394         int rc;
1395
1396         spin_lock(&li->lock);
1397         switch (irq->type) {
1398         case KVM_S390_PROGRAM_INT:
1399                 VCPU_EVENT(vcpu, 3, "inject: program check %d (from user)",
1400                            irq->u.pgm.code);
1401                 rc = __inject_prog(vcpu, irq);
1402                 break;
1403         case KVM_S390_SIGP_SET_PREFIX:
1404                 rc = __inject_set_prefix(vcpu, irq);
1405                 break;
1406         case KVM_S390_SIGP_STOP:
1407                 rc = __inject_sigp_stop(vcpu, irq);
1408                 break;
1409         case KVM_S390_RESTART:
1410                 rc = __inject_sigp_restart(vcpu, irq);
1411                 break;
1412         case KVM_S390_INT_CLOCK_COMP:
1413                 rc = __inject_ckc(vcpu);
1414                 break;
1415         case KVM_S390_INT_CPU_TIMER:
1416                 rc = __inject_cpu_timer(vcpu);
1417                 break;
1418         case KVM_S390_INT_EXTERNAL_CALL:
1419                 rc = __inject_extcall(vcpu, irq);
1420                 break;
1421         case KVM_S390_INT_EMERGENCY:
1422                 rc = __inject_sigp_emergency(vcpu, irq);
1423                 break;
1424         case KVM_S390_MCHK:
1425                 rc = __inject_mchk(vcpu, irq);
1426                 break;
1427         case KVM_S390_INT_PFAULT_INIT:
1428                 rc = __inject_pfault_init(vcpu, irq);
1429                 break;
1430         case KVM_S390_INT_VIRTIO:
1431         case KVM_S390_INT_SERVICE:
1432         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1433         default:
1434                 rc = -EINVAL;
1435         }
1436         spin_unlock(&li->lock);
1437         if (!rc)
1438                 kvm_s390_vcpu_wakeup(vcpu);
1439         return rc;
1440 }
1441
1442 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1443 {
1444         struct kvm_s390_float_interrupt *fi;
1445         struct kvm_s390_interrupt_info  *n, *inti = NULL;
1446
1447         fi = &kvm->arch.float_int;
1448         spin_lock(&fi->lock);
1449         list_for_each_entry_safe(inti, n, &fi->list, list) {
1450                 list_del(&inti->list);
1451                 kfree(inti);
1452         }
1453         fi->irq_count = 0;
1454         atomic_set(&fi->active, 0);
1455         spin_unlock(&fi->lock);
1456 }
1457
1458 static inline int copy_irq_to_user(struct kvm_s390_interrupt_info *inti,
1459                                    u8 *addr)
1460 {
1461         struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1462         struct kvm_s390_irq irq = {0};
1463
1464         irq.type = inti->type;
1465         switch (inti->type) {
1466         case KVM_S390_INT_PFAULT_INIT:
1467         case KVM_S390_INT_PFAULT_DONE:
1468         case KVM_S390_INT_VIRTIO:
1469         case KVM_S390_INT_SERVICE:
1470                 irq.u.ext = inti->ext;
1471                 break;
1472         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1473                 irq.u.io = inti->io;
1474                 break;
1475         case KVM_S390_MCHK:
1476                 irq.u.mchk = inti->mchk;
1477                 break;
1478         default:
1479                 return -EINVAL;
1480         }
1481
1482         if (copy_to_user(uptr, &irq, sizeof(irq)))
1483                 return -EFAULT;
1484
1485         return 0;
1486 }
1487
1488 static int get_all_floating_irqs(struct kvm *kvm, __u8 *buf, __u64 len)
1489 {
1490         struct kvm_s390_interrupt_info *inti;
1491         struct kvm_s390_float_interrupt *fi;
1492         int ret = 0;
1493         int n = 0;
1494
1495         fi = &kvm->arch.float_int;
1496         spin_lock(&fi->lock);
1497
1498         list_for_each_entry(inti, &fi->list, list) {
1499                 if (len < sizeof(struct kvm_s390_irq)) {
1500                         /* signal userspace to try again */
1501                         ret = -ENOMEM;
1502                         break;
1503                 }
1504                 ret = copy_irq_to_user(inti, buf);
1505                 if (ret)
1506                         break;
1507                 buf += sizeof(struct kvm_s390_irq);
1508                 len -= sizeof(struct kvm_s390_irq);
1509                 n++;
1510         }
1511
1512         spin_unlock(&fi->lock);
1513
1514         return ret < 0 ? ret : n;
1515 }
1516
1517 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1518 {
1519         int r;
1520
1521         switch (attr->group) {
1522         case KVM_DEV_FLIC_GET_ALL_IRQS:
1523                 r = get_all_floating_irqs(dev->kvm, (u8 *) attr->addr,
1524                                           attr->attr);
1525                 break;
1526         default:
1527                 r = -EINVAL;
1528         }
1529
1530         return r;
1531 }
1532
1533 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1534                                      u64 addr)
1535 {
1536         struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1537         void *target = NULL;
1538         void __user *source;
1539         u64 size;
1540
1541         if (get_user(inti->type, (u64 __user *)addr))
1542                 return -EFAULT;
1543
1544         switch (inti->type) {
1545         case KVM_S390_INT_PFAULT_INIT:
1546         case KVM_S390_INT_PFAULT_DONE:
1547         case KVM_S390_INT_VIRTIO:
1548         case KVM_S390_INT_SERVICE:
1549                 target = (void *) &inti->ext;
1550                 source = &uptr->u.ext;
1551                 size = sizeof(inti->ext);
1552                 break;
1553         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1554                 target = (void *) &inti->io;
1555                 source = &uptr->u.io;
1556                 size = sizeof(inti->io);
1557                 break;
1558         case KVM_S390_MCHK:
1559                 target = (void *) &inti->mchk;
1560                 source = &uptr->u.mchk;
1561                 size = sizeof(inti->mchk);
1562                 break;
1563         default:
1564                 return -EINVAL;
1565         }
1566
1567         if (copy_from_user(target, source, size))
1568                 return -EFAULT;
1569
1570         return 0;
1571 }
1572
1573 static int enqueue_floating_irq(struct kvm_device *dev,
1574                                 struct kvm_device_attr *attr)
1575 {
1576         struct kvm_s390_interrupt_info *inti = NULL;
1577         int r = 0;
1578         int len = attr->attr;
1579
1580         if (len % sizeof(struct kvm_s390_irq) != 0)
1581                 return -EINVAL;
1582         else if (len > KVM_S390_FLIC_MAX_BUFFER)
1583                 return -EINVAL;
1584
1585         while (len >= sizeof(struct kvm_s390_irq)) {
1586                 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1587                 if (!inti)
1588                         return -ENOMEM;
1589
1590                 r = copy_irq_from_user(inti, attr->addr);
1591                 if (r) {
1592                         kfree(inti);
1593                         return r;
1594                 }
1595                 r = __inject_vm(dev->kvm, inti);
1596                 if (r) {
1597                         kfree(inti);
1598                         return r;
1599                 }
1600                 len -= sizeof(struct kvm_s390_irq);
1601                 attr->addr += sizeof(struct kvm_s390_irq);
1602         }
1603
1604         return r;
1605 }
1606
1607 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1608 {
1609         if (id >= MAX_S390_IO_ADAPTERS)
1610                 return NULL;
1611         return kvm->arch.adapters[id];
1612 }
1613
1614 static int register_io_adapter(struct kvm_device *dev,
1615                                struct kvm_device_attr *attr)
1616 {
1617         struct s390_io_adapter *adapter;
1618         struct kvm_s390_io_adapter adapter_info;
1619
1620         if (copy_from_user(&adapter_info,
1621                            (void __user *)attr->addr, sizeof(adapter_info)))
1622                 return -EFAULT;
1623
1624         if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1625             (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1626                 return -EINVAL;
1627
1628         adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1629         if (!adapter)
1630                 return -ENOMEM;
1631
1632         INIT_LIST_HEAD(&adapter->maps);
1633         init_rwsem(&adapter->maps_lock);
1634         atomic_set(&adapter->nr_maps, 0);
1635         adapter->id = adapter_info.id;
1636         adapter->isc = adapter_info.isc;
1637         adapter->maskable = adapter_info.maskable;
1638         adapter->masked = false;
1639         adapter->swap = adapter_info.swap;
1640         dev->kvm->arch.adapters[adapter->id] = adapter;
1641
1642         return 0;
1643 }
1644
1645 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1646 {
1647         int ret;
1648         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1649
1650         if (!adapter || !adapter->maskable)
1651                 return -EINVAL;
1652         ret = adapter->masked;
1653         adapter->masked = masked;
1654         return ret;
1655 }
1656
1657 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1658 {
1659         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1660         struct s390_map_info *map;
1661         int ret;
1662
1663         if (!adapter || !addr)
1664                 return -EINVAL;
1665
1666         map = kzalloc(sizeof(*map), GFP_KERNEL);
1667         if (!map) {
1668                 ret = -ENOMEM;
1669                 goto out;
1670         }
1671         INIT_LIST_HEAD(&map->list);
1672         map->guest_addr = addr;
1673         map->addr = gmap_translate(kvm->arch.gmap, addr);
1674         if (map->addr == -EFAULT) {
1675                 ret = -EFAULT;
1676                 goto out;
1677         }
1678         ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1679         if (ret < 0)
1680                 goto out;
1681         BUG_ON(ret != 1);
1682         down_write(&adapter->maps_lock);
1683         if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1684                 list_add_tail(&map->list, &adapter->maps);
1685                 ret = 0;
1686         } else {
1687                 put_page(map->page);
1688                 ret = -EINVAL;
1689         }
1690         up_write(&adapter->maps_lock);
1691 out:
1692         if (ret)
1693                 kfree(map);
1694         return ret;
1695 }
1696
1697 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1698 {
1699         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1700         struct s390_map_info *map, *tmp;
1701         int found = 0;
1702
1703         if (!adapter || !addr)
1704                 return -EINVAL;
1705
1706         down_write(&adapter->maps_lock);
1707         list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1708                 if (map->guest_addr == addr) {
1709                         found = 1;
1710                         atomic_dec(&adapter->nr_maps);
1711                         list_del(&map->list);
1712                         put_page(map->page);
1713                         kfree(map);
1714                         break;
1715                 }
1716         }
1717         up_write(&adapter->maps_lock);
1718
1719         return found ? 0 : -EINVAL;
1720 }
1721
1722 void kvm_s390_destroy_adapters(struct kvm *kvm)
1723 {
1724         int i;
1725         struct s390_map_info *map, *tmp;
1726
1727         for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
1728                 if (!kvm->arch.adapters[i])
1729                         continue;
1730                 list_for_each_entry_safe(map, tmp,
1731                                          &kvm->arch.adapters[i]->maps, list) {
1732                         list_del(&map->list);
1733                         put_page(map->page);
1734                         kfree(map);
1735                 }
1736                 kfree(kvm->arch.adapters[i]);
1737         }
1738 }
1739
1740 static int modify_io_adapter(struct kvm_device *dev,
1741                              struct kvm_device_attr *attr)
1742 {
1743         struct kvm_s390_io_adapter_req req;
1744         struct s390_io_adapter *adapter;
1745         int ret;
1746
1747         if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
1748                 return -EFAULT;
1749
1750         adapter = get_io_adapter(dev->kvm, req.id);
1751         if (!adapter)
1752                 return -EINVAL;
1753         switch (req.type) {
1754         case KVM_S390_IO_ADAPTER_MASK:
1755                 ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
1756                 if (ret > 0)
1757                         ret = 0;
1758                 break;
1759         case KVM_S390_IO_ADAPTER_MAP:
1760                 ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
1761                 break;
1762         case KVM_S390_IO_ADAPTER_UNMAP:
1763                 ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
1764                 break;
1765         default:
1766                 ret = -EINVAL;
1767         }
1768
1769         return ret;
1770 }
1771
1772 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1773 {
1774         int r = 0;
1775         unsigned int i;
1776         struct kvm_vcpu *vcpu;
1777
1778         switch (attr->group) {
1779         case KVM_DEV_FLIC_ENQUEUE:
1780                 r = enqueue_floating_irq(dev, attr);
1781                 break;
1782         case KVM_DEV_FLIC_CLEAR_IRQS:
1783                 kvm_s390_clear_float_irqs(dev->kvm);
1784                 break;
1785         case KVM_DEV_FLIC_APF_ENABLE:
1786                 dev->kvm->arch.gmap->pfault_enabled = 1;
1787                 break;
1788         case KVM_DEV_FLIC_APF_DISABLE_WAIT:
1789                 dev->kvm->arch.gmap->pfault_enabled = 0;
1790                 /*
1791                  * Make sure no async faults are in transition when
1792                  * clearing the queues. So we don't need to worry
1793                  * about late coming workers.
1794                  */
1795                 synchronize_srcu(&dev->kvm->srcu);
1796                 kvm_for_each_vcpu(i, vcpu, dev->kvm)
1797                         kvm_clear_async_pf_completion_queue(vcpu);
1798                 break;
1799         case KVM_DEV_FLIC_ADAPTER_REGISTER:
1800                 r = register_io_adapter(dev, attr);
1801                 break;
1802         case KVM_DEV_FLIC_ADAPTER_MODIFY:
1803                 r = modify_io_adapter(dev, attr);
1804                 break;
1805         default:
1806                 r = -EINVAL;
1807         }
1808
1809         return r;
1810 }
1811
1812 static int flic_create(struct kvm_device *dev, u32 type)
1813 {
1814         if (!dev)
1815                 return -EINVAL;
1816         if (dev->kvm->arch.flic)
1817                 return -EINVAL;
1818         dev->kvm->arch.flic = dev;
1819         return 0;
1820 }
1821
1822 static void flic_destroy(struct kvm_device *dev)
1823 {
1824         dev->kvm->arch.flic = NULL;
1825         kfree(dev);
1826 }
1827
1828 /* s390 floating irq controller (flic) */
1829 struct kvm_device_ops kvm_flic_ops = {
1830         .name = "kvm-flic",
1831         .get_attr = flic_get_attr,
1832         .set_attr = flic_set_attr,
1833         .create = flic_create,
1834         .destroy = flic_destroy,
1835 };
1836
1837 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
1838 {
1839         unsigned long bit;
1840
1841         bit = bit_nr + (addr % PAGE_SIZE) * 8;
1842
1843         return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
1844 }
1845
1846 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
1847                                           u64 addr)
1848 {
1849         struct s390_map_info *map;
1850
1851         if (!adapter)
1852                 return NULL;
1853
1854         list_for_each_entry(map, &adapter->maps, list) {
1855                 if (map->guest_addr == addr)
1856                         return map;
1857         }
1858         return NULL;
1859 }
1860
1861 static int adapter_indicators_set(struct kvm *kvm,
1862                                   struct s390_io_adapter *adapter,
1863                                   struct kvm_s390_adapter_int *adapter_int)
1864 {
1865         unsigned long bit;
1866         int summary_set, idx;
1867         struct s390_map_info *info;
1868         void *map;
1869
1870         info = get_map_info(adapter, adapter_int->ind_addr);
1871         if (!info)
1872                 return -1;
1873         map = page_address(info->page);
1874         bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
1875         set_bit(bit, map);
1876         idx = srcu_read_lock(&kvm->srcu);
1877         mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
1878         set_page_dirty_lock(info->page);
1879         info = get_map_info(adapter, adapter_int->summary_addr);
1880         if (!info) {
1881                 srcu_read_unlock(&kvm->srcu, idx);
1882                 return -1;
1883         }
1884         map = page_address(info->page);
1885         bit = get_ind_bit(info->addr, adapter_int->summary_offset,
1886                           adapter->swap);
1887         summary_set = test_and_set_bit(bit, map);
1888         mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
1889         set_page_dirty_lock(info->page);
1890         srcu_read_unlock(&kvm->srcu, idx);
1891         return summary_set ? 0 : 1;
1892 }
1893
1894 /*
1895  * < 0 - not injected due to error
1896  * = 0 - coalesced, summary indicator already active
1897  * > 0 - injected interrupt
1898  */
1899 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
1900                            struct kvm *kvm, int irq_source_id, int level,
1901                            bool line_status)
1902 {
1903         int ret;
1904         struct s390_io_adapter *adapter;
1905
1906         /* We're only interested in the 0->1 transition. */
1907         if (!level)
1908                 return 0;
1909         adapter = get_io_adapter(kvm, e->adapter.adapter_id);
1910         if (!adapter)
1911                 return -1;
1912         down_read(&adapter->maps_lock);
1913         ret = adapter_indicators_set(kvm, adapter, &e->adapter);
1914         up_read(&adapter->maps_lock);
1915         if ((ret > 0) && !adapter->masked) {
1916                 struct kvm_s390_interrupt s390int = {
1917                         .type = KVM_S390_INT_IO(1, 0, 0, 0),
1918                         .parm = 0,
1919                         .parm64 = (adapter->isc << 27) | 0x80000000,
1920                 };
1921                 ret = kvm_s390_inject_vm(kvm, &s390int);
1922                 if (ret == 0)
1923                         ret = 1;
1924         }
1925         return ret;
1926 }
1927
1928 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1929                           const struct kvm_irq_routing_entry *ue)
1930 {
1931         int ret;
1932
1933         switch (ue->type) {
1934         case KVM_IRQ_ROUTING_S390_ADAPTER:
1935                 e->set = set_adapter_int;
1936                 e->adapter.summary_addr = ue->u.adapter.summary_addr;
1937                 e->adapter.ind_addr = ue->u.adapter.ind_addr;
1938                 e->adapter.summary_offset = ue->u.adapter.summary_offset;
1939                 e->adapter.ind_offset = ue->u.adapter.ind_offset;
1940                 e->adapter.adapter_id = ue->u.adapter.adapter_id;
1941                 ret = 0;
1942                 break;
1943         default:
1944                 ret = -EINVAL;
1945         }
1946
1947         return ret;
1948 }
1949
1950 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
1951                 int irq_source_id, int level, bool line_status)
1952 {
1953         return -EINVAL;
1954 }