arch/s390/kvm/priv.c

   1 /*
   2  * handling privileged instructions
   3  *
   4  * Copyright IBM Corp. 2008, 2013
   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  *               Christian Borntraeger <borntraeger@de.ibm.com>
  12  */
  13
  14 #include <linux/kvm.h>
  15 #include <linux/gfp.h>
  16 #include <linux/errno.h>
  17 #include <linux/compat.h>
  18 #include <asm/asm-offsets.h>
  19 #include <asm/facility.h>
  20 #include <asm/current.h>
  21 #include <asm/debug.h>
  22 #include <asm/ebcdic.h>
  23 #include <asm/sysinfo.h>
  24 #include <asm/pgtable.h>
  25 #include <asm/pgalloc.h>
  26 #include <asm/gmap.h>
  27 #include <asm/io.h>
  28 #include <asm/ptrace.h>
  29 #include <asm/compat.h>
  30 #include "gaccess.h"
  31 #include "kvm-s390.h"
  32 #include "trace.h"
  33
  34 /* Handle SCK (SET CLOCK) interception */
  35 static int handle_set_clock(struct kvm_vcpu *vcpu)
  36 {
  37         int rc;
  38         ar_t ar;
  39         u64 op2, val;
  40
  41         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  42                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
  43
  44         op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
  45         if (op2 & 7)    /* Operand must be on a doubleword boundary */
  46                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
  47         rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
  48         if (rc)
  49                 return kvm_s390_inject_prog_cond(vcpu, rc);
  50
  51         VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
  52         kvm_s390_set_tod_clock(vcpu->kvm, val);
  53
  54         kvm_s390_set_psw_cc(vcpu, 0);
  55         return 0;
  56 }
  57
  58 static int handle_set_prefix(struct kvm_vcpu *vcpu)
  59 {
  60         u64 operand2;
  61         u32 address;
  62         int rc;
  63         ar_t ar;
  64
  65         vcpu->stat.instruction_spx++;
  66
  67         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  68                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
  69
  70         operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
  71
  72         /* must be word boundary */
  73         if (operand2 & 3)
  74                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
  75
  76         /* get the value */
  77         rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
  78         if (rc)
  79                 return kvm_s390_inject_prog_cond(vcpu, rc);
  80
  81         address &= 0x7fffe000u;
  82
  83         /*
  84          * Make sure the new value is valid memory. We only need to check the
  85          * first page, since address is 8k aligned and memory pieces are always
  86          * at least 1MB aligned and have at least a size of 1MB.
  87          */
  88         if (kvm_is_error_gpa(vcpu->kvm, address))
  89                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
  90
  91         kvm_s390_set_prefix(vcpu, address);
  92         trace_kvm_s390_handle_prefix(vcpu, 1, address);
  93         return 0;
  94 }
  95
  96 static int handle_store_prefix(struct kvm_vcpu *vcpu)
  97 {
  98         u64 operand2;
  99         u32 address;
 100         int rc;
 101         ar_t ar;
 102
 103         vcpu->stat.instruction_stpx++;
 104
 105         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 106                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 107
 108         operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 109
 110         /* must be word boundary */
 111         if (operand2 & 3)
 112                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 113
 114         address = kvm_s390_get_prefix(vcpu);
 115
 116         /* get the value */
 117         rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
 118         if (rc)
 119                 return kvm_s390_inject_prog_cond(vcpu, rc);
 120
 121         VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
 122         trace_kvm_s390_handle_prefix(vcpu, 0, address);
 123         return 0;
 124 }
 125
 126 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
 127 {
 128         u16 vcpu_id = vcpu->vcpu_id;
 129         u64 ga;
 130         int rc;
 131         ar_t ar;
 132
 133         vcpu->stat.instruction_stap++;
 134
 135         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 136                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 137
 138         ga = kvm_s390_get_base_disp_s(vcpu, &ar);
 139
 140         if (ga & 1)
 141                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 142
 143         rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
 144         if (rc)
 145                 return kvm_s390_inject_prog_cond(vcpu, rc);
 146
 147         VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
 148         trace_kvm_s390_handle_stap(vcpu, ga);
 149         return 0;
 150 }
 151
 152 static int __skey_check_enable(struct kvm_vcpu *vcpu)
 153 {
 154         int rc = 0;
 155         if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
 156                 return rc;
 157
 158         rc = s390_enable_skey();
 159         VCPU_EVENT(vcpu, 3, "%s", "enabling storage keys for guest");
 160         trace_kvm_s390_skey_related_inst(vcpu);
 161         vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
 162         return rc;
 163 }
 164
 165
 166 static int handle_skey(struct kvm_vcpu *vcpu)
 167 {
 168         int rc = __skey_check_enable(vcpu);
 169
 170         if (rc)
 171                 return rc;
 172         vcpu->stat.instruction_storage_key++;
 173
 174         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 175                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 176
 177         kvm_s390_retry_instr(vcpu);
 178         VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
 179         return 0;
 180 }
 181
 182 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
 183 {
 184         vcpu->stat.instruction_ipte_interlock++;
 185         if (psw_bits(vcpu->arch.sie_block->gpsw).p)
 186                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 187         wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
 188         kvm_s390_retry_instr(vcpu);
 189         VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
 190         return 0;
 191 }
 192
 193 static int handle_test_block(struct kvm_vcpu *vcpu)
 194 {
 195         gpa_t addr;
 196         int reg2;
 197
 198         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 199                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 200
 201         kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
 202         addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 203         addr = kvm_s390_logical_to_effective(vcpu, addr);
 204         if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
 205                 return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 206         addr = kvm_s390_real_to_abs(vcpu, addr);
 207
 208         if (kvm_is_error_gpa(vcpu->kvm, addr))
 209                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 210         /*
 211          * We don't expect errors on modern systems, and do not care
 212          * about storage keys (yet), so let's just clear the page.
 213          */
 214         if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
 215                 return -EFAULT;
 216         kvm_s390_set_psw_cc(vcpu, 0);
 217         vcpu->run->s.regs.gprs[0] = 0;
 218         return 0;
 219 }
 220
 221 static int handle_tpi(struct kvm_vcpu *vcpu)
 222 {
 223         struct kvm_s390_interrupt_info *inti;
 224         unsigned long len;
 225         u32 tpi_data[3];
 226         int rc;
 227         u64 addr;
 228         ar_t ar;
 229
 230         addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 231         if (addr & 3)
 232                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 233
 234         inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
 235         if (!inti) {
 236                 kvm_s390_set_psw_cc(vcpu, 0);
 237                 return 0;
 238         }
 239
 240         tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
 241         tpi_data[1] = inti->io.io_int_parm;
 242         tpi_data[2] = inti->io.io_int_word;
 243         if (addr) {
 244                 /*
 245                  * Store the two-word I/O interruption code into the
 246                  * provided area.
 247                  */
 248                 len = sizeof(tpi_data) - 4;
 249                 rc = write_guest(vcpu, addr, ar, &tpi_data, len);
 250                 if (rc) {
 251                         rc = kvm_s390_inject_prog_cond(vcpu, rc);
 252                         goto reinject_interrupt;
 253                 }
 254         } else {
 255                 /*
 256                  * Store the three-word I/O interruption code into
 257                  * the appropriate lowcore area.
 258                  */
 259                 len = sizeof(tpi_data);
 260                 if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
 261                         /* failed writes to the low core are not recoverable */
 262                         rc = -EFAULT;
 263                         goto reinject_interrupt;
 264                 }
 265         }
 266
 267         /* irq was successfully handed to the guest */
 268         kfree(inti);
 269         kvm_s390_set_psw_cc(vcpu, 1);
 270         return 0;
 271 reinject_interrupt:
 272         /*
 273          * If we encounter a problem storing the interruption code, the
 274          * instruction is suppressed from the guest's view: reinject the
 275          * interrupt.
 276          */
 277         if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
 278                 kfree(inti);
 279                 rc = -EFAULT;
 280         }
 281         /* don't set the cc, a pgm irq was injected or we drop to user space */
 282         return rc ? -EFAULT : 0;
 283 }
 284
 285 static int handle_tsch(struct kvm_vcpu *vcpu)
 286 {
 287         struct kvm_s390_interrupt_info *inti = NULL;
 288         const u64 isc_mask = 0xffUL << 24; /* all iscs set */
 289
 290         /* a valid schid has at least one bit set */
 291         if (vcpu->run->s.regs.gprs[1])
 292                 inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
 293                                            vcpu->run->s.regs.gprs[1]);
 294
 295         /*
 296          * Prepare exit to userspace.
 297          * We indicate whether we dequeued a pending I/O interrupt
 298          * so that userspace can re-inject it if the instruction gets
 299          * a program check. While this may re-order the pending I/O
 300          * interrupts, this is no problem since the priority is kept
 301          * intact.
 302          */
 303         vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
 304         vcpu->run->s390_tsch.dequeued = !!inti;
 305         if (inti) {
 306                 vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
 307                 vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
 308                 vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
 309                 vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
 310         }
 311         vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
 312         kfree(inti);
 313         return -EREMOTE;
 314 }
 315
 316 static int handle_io_inst(struct kvm_vcpu *vcpu)
 317 {
 318         VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
 319
 320         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 321                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 322
 323         if (vcpu->kvm->arch.css_support) {
 324                 /*
 325                  * Most I/O instructions will be handled by userspace.
 326                  * Exceptions are tpi and the interrupt portion of tsch.
 327                  */
 328                 if (vcpu->arch.sie_block->ipa == 0xb236)
 329                         return handle_tpi(vcpu);
 330                 if (vcpu->arch.sie_block->ipa == 0xb235)
 331                         return handle_tsch(vcpu);
 332                 /* Handle in userspace. */
 333                 return -EOPNOTSUPP;
 334         } else {
 335                 /*
 336                  * Set condition code 3 to stop the guest from issuing channel
 337                  * I/O instructions.
 338                  */
 339                 kvm_s390_set_psw_cc(vcpu, 3);
 340                 return 0;
 341         }
 342 }
 343
 344 static int handle_stfl(struct kvm_vcpu *vcpu)
 345 {
 346         int rc;
 347         unsigned int fac;
 348
 349         vcpu->stat.instruction_stfl++;
 350
 351         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 352                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 353
 354         /*
 355          * We need to shift the lower 32 facility bits (bit 0-31) from a u64
 356          * into a u32 memory representation. They will remain bits 0-31.
 357          */
 358         fac = *vcpu->kvm->arch.model.fac_list >> 32;
 359         rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
 360                             &fac, sizeof(fac));
 361         if (rc)
 362                 return rc;
 363         VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
 364         trace_kvm_s390_handle_stfl(vcpu, fac);
 365         return 0;
 366 }
 367
 368 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
 369 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
 370 #define PSW_ADDR_24 0x0000000000ffffffUL
 371 #define PSW_ADDR_31 0x000000007fffffffUL
 372
 373 int is_valid_psw(psw_t *psw)
 374 {
 375         if (psw->mask & PSW_MASK_UNASSIGNED)
 376                 return 0;
 377         if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
 378                 if (psw->addr & ~PSW_ADDR_31)
 379                         return 0;
 380         }
 381         if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
 382                 return 0;
 383         if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
 384                 return 0;
 385         if (psw->addr & 1)
 386                 return 0;
 387         return 1;
 388 }
 389
 390 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
 391 {
 392         psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
 393         psw_compat_t new_psw;
 394         u64 addr;
 395         int rc;
 396         ar_t ar;
 397
 398         if (gpsw->mask & PSW_MASK_PSTATE)
 399                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 400
 401         addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 402         if (addr & 7)
 403                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 404
 405         rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 406         if (rc)
 407                 return kvm_s390_inject_prog_cond(vcpu, rc);
 408         if (!(new_psw.mask & PSW32_MASK_BASE))
 409                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 410         gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
 411         gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
 412         gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
 413         if (!is_valid_psw(gpsw))
 414                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 415         return 0;
 416 }
 417
 418 static int handle_lpswe(struct kvm_vcpu *vcpu)
 419 {
 420         psw_t new_psw;
 421         u64 addr;
 422         int rc;
 423         ar_t ar;
 424
 425         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 426                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 427
 428         addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 429         if (addr & 7)
 430                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 431         rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 432         if (rc)
 433                 return kvm_s390_inject_prog_cond(vcpu, rc);
 434         vcpu->arch.sie_block->gpsw = new_psw;
 435         if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
 436                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 437         return 0;
 438 }
 439
 440 static int handle_stidp(struct kvm_vcpu *vcpu)
 441 {
 442         u64 stidp_data = vcpu->arch.stidp_data;
 443         u64 operand2;
 444         int rc;
 445         ar_t ar;
 446
 447         vcpu->stat.instruction_stidp++;
 448
 449         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 450                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 451
 452         operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 453
 454         if (operand2 & 7)
 455                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 456
 457         rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
 458         if (rc)
 459                 return kvm_s390_inject_prog_cond(vcpu, rc);
 460
 461         VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
 462         return 0;
 463 }
 464
 465 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
 466 {
 467         int cpus = 0;
 468         int n;
 469
 470         cpus = atomic_read(&vcpu->kvm->online_vcpus);
 471
 472         /* deal with other level 3 hypervisors */
 473         if (stsi(mem, 3, 2, 2))
 474                 mem->count = 0;
 475         if (mem->count < 8)
 476                 mem->count++;
 477         for (n = mem->count - 1; n > 0 ; n--)
 478                 memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
 479
 480         memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
 481         mem->vm[0].cpus_total = cpus;
 482         mem->vm[0].cpus_configured = cpus;
 483         mem->vm[0].cpus_standby = 0;
 484         mem->vm[0].cpus_reserved = 0;
 485         mem->vm[0].caf = 1000;
 486         memcpy(mem->vm[0].name, "KVMguest", 8);
 487         ASCEBC(mem->vm[0].name, 8);
 488         memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
 489         ASCEBC(mem->vm[0].cpi, 16);
 490 }
 491
 492 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, ar_t ar,
 493                                  u8 fc, u8 sel1, u16 sel2)
 494 {
 495         vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
 496         vcpu->run->s390_stsi.addr = addr;
 497         vcpu->run->s390_stsi.ar = ar;
 498         vcpu->run->s390_stsi.fc = fc;
 499         vcpu->run->s390_stsi.sel1 = sel1;
 500         vcpu->run->s390_stsi.sel2 = sel2;
 501 }
 502
 503 static int handle_stsi(struct kvm_vcpu *vcpu)
 504 {
 505         int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
 506         int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
 507         int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
 508         unsigned long mem = 0;
 509         u64 operand2;
 510         int rc = 0;
 511         ar_t ar;
 512
 513         vcpu->stat.instruction_stsi++;
 514         VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
 515
 516         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 517                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 518
 519         if (fc > 3) {
 520                 kvm_s390_set_psw_cc(vcpu, 3);
 521                 return 0;
 522         }
 523
 524         if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
 525             || vcpu->run->s.regs.gprs[1] & 0xffff0000)
 526                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 527
 528         if (fc == 0) {
 529                 vcpu->run->s.regs.gprs[0] = 3 << 28;
 530                 kvm_s390_set_psw_cc(vcpu, 0);
 531                 return 0;
 532         }
 533
 534         operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 535
 536         if (operand2 & 0xfff)
 537                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 538
 539         switch (fc) {
 540         case 1: /* same handling for 1 and 2 */
 541         case 2:
 542                 mem = get_zeroed_page(GFP_KERNEL);
 543                 if (!mem)
 544                         goto out_no_data;
 545                 if (stsi((void *) mem, fc, sel1, sel2))
 546                         goto out_no_data;
 547                 break;
 548         case 3:
 549                 if (sel1 != 2 || sel2 != 2)
 550                         goto out_no_data;
 551                 mem = get_zeroed_page(GFP_KERNEL);
 552                 if (!mem)
 553                         goto out_no_data;
 554                 handle_stsi_3_2_2(vcpu, (void *) mem);
 555                 break;
 556         }
 557
 558         rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
 559         if (rc) {
 560                 rc = kvm_s390_inject_prog_cond(vcpu, rc);
 561                 goto out;
 562         }
 563         if (vcpu->kvm->arch.user_stsi) {
 564                 insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
 565                 rc = -EREMOTE;
 566         }
 567         trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 568         free_page(mem);
 569         kvm_s390_set_psw_cc(vcpu, 0);
 570         vcpu->run->s.regs.gprs[0] = 0;
 571         return rc;
 572 out_no_data:
 573         kvm_s390_set_psw_cc(vcpu, 3);
 574 out:
 575         free_page(mem);
 576         return rc;
 577 }
 578
 579 static const intercept_handler_t b2_handlers[256] = {
 580         [0x02] = handle_stidp,
 581         [0x04] = handle_set_clock,
 582         [0x10] = handle_set_prefix,
 583         [0x11] = handle_store_prefix,
 584         [0x12] = handle_store_cpu_address,
 585         [0x21] = handle_ipte_interlock,
 586         [0x29] = handle_skey,
 587         [0x2a] = handle_skey,
 588         [0x2b] = handle_skey,
 589         [0x2c] = handle_test_block,
 590         [0x30] = handle_io_inst,
 591         [0x31] = handle_io_inst,
 592         [0x32] = handle_io_inst,
 593         [0x33] = handle_io_inst,
 594         [0x34] = handle_io_inst,
 595         [0x35] = handle_io_inst,
 596         [0x36] = handle_io_inst,
 597         [0x37] = handle_io_inst,
 598         [0x38] = handle_io_inst,
 599         [0x39] = handle_io_inst,
 600         [0x3a] = handle_io_inst,
 601         [0x3b] = handle_io_inst,
 602         [0x3c] = handle_io_inst,
 603         [0x50] = handle_ipte_interlock,
 604         [0x5f] = handle_io_inst,
 605         [0x74] = handle_io_inst,
 606         [0x76] = handle_io_inst,
 607         [0x7d] = handle_stsi,
 608         [0xb1] = handle_stfl,
 609         [0xb2] = handle_lpswe,
 610 };
 611
 612 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
 613 {
 614         intercept_handler_t handler;
 615
 616         /*
 617          * A lot of B2 instructions are priviledged. Here we check for
 618          * the privileged ones, that we can handle in the kernel.
 619          * Anything else goes to userspace.
 620          */
 621         handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
 622         if (handler)
 623                 return handler(vcpu);
 624
 625         return -EOPNOTSUPP;
 626 }
 627
 628 static int handle_epsw(struct kvm_vcpu *vcpu)
 629 {
 630         int reg1, reg2;
 631
 632         kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 633
 634         /* This basically extracts the mask half of the psw. */
 635         vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
 636         vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
 637         if (reg2) {
 638                 vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
 639                 vcpu->run->s.regs.gprs[reg2] |=
 640                         vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
 641         }
 642         return 0;
 643 }
 644
 645 #define PFMF_RESERVED   0xfffc0101UL
 646 #define PFMF_SK         0x00020000UL
 647 #define PFMF_CF         0x00010000UL
 648 #define PFMF_UI         0x00008000UL
 649 #define PFMF_FSC        0x00007000UL
 650 #define PFMF_NQ         0x00000800UL
 651 #define PFMF_MR         0x00000400UL
 652 #define PFMF_MC         0x00000200UL
 653 #define PFMF_KEY        0x000000feUL
 654
 655 static int handle_pfmf(struct kvm_vcpu *vcpu)
 656 {
 657         int reg1, reg2;
 658         unsigned long start, end;
 659
 660         vcpu->stat.instruction_pfmf++;
 661
 662         kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 663
 664         if (!test_kvm_facility(vcpu->kvm, 8))
 665                 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
 666
 667         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 668                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 669
 670         if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
 671                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 672
 673         /* Only provide non-quiescing support if the host supports it */
 674         if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
 675                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 676
 677         /* No support for conditional-SSKE */
 678         if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
 679                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 680
 681         start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 682         start = kvm_s390_logical_to_effective(vcpu, start);
 683
 684         switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
 685         case 0x00000000:
 686                 end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
 687                 break;
 688         case 0x00001000:
 689                 end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
 690                 break;
 691         case 0x00002000:
 692                 /* only support 2G frame size if EDAT2 is available and we are
 693                    not in 24-bit addressing mode */
 694                 if (!test_kvm_facility(vcpu->kvm, 78) ||
 695                     psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_24BIT)
 696                         return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 697                 end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
 698                 break;
 699         default:
 700                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 701         }
 702
 703         if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
 704                 if (kvm_s390_check_low_addr_prot_real(vcpu, start))
 705                         return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 706         }
 707
 708         while (start < end) {
 709                 unsigned long useraddr, abs_addr;
 710
 711                 /* Translate guest address to host address */
 712                 if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
 713                         abs_addr = kvm_s390_real_to_abs(vcpu, start);
 714                 else
 715                         abs_addr = start;
 716                 useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
 717                 if (kvm_is_error_hva(useraddr))
 718                         return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 719
 720                 if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
 721                         if (clear_user((void __user *)useraddr, PAGE_SIZE))
 722                                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 723                 }
 724
 725                 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
 726                         int rc = __skey_check_enable(vcpu);
 727
 728                         if (rc)
 729                                 return rc;
 730                         if (set_guest_storage_key(current->mm, useraddr,
 731                                         vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
 732                                         vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
 733                                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 734                 }
 735
 736                 start += PAGE_SIZE;
 737         }
 738         if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
 739                 vcpu->run->s.regs.gprs[reg2] = end;
 740         return 0;
 741 }
 742
 743 static int handle_essa(struct kvm_vcpu *vcpu)
 744 {
 745         /* entries expected to be 1FF */
 746         int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
 747         unsigned long *cbrlo, cbrle;
 748         struct gmap *gmap;
 749         int i;
 750
 751         VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
 752         gmap = vcpu->arch.gmap;
 753         vcpu->stat.instruction_essa++;
 754         if (!vcpu->kvm->arch.use_cmma)
 755                 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
 756
 757         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 758                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 759
 760         if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
 761                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 762
 763         /* Retry the ESSA instruction */
 764         kvm_s390_retry_instr(vcpu);
 765         vcpu->arch.sie_block->cbrlo &= PAGE_MASK;       /* reset nceo */
 766         cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
 767         down_read(&gmap->mm->mmap_sem);
 768         for (i = 0; i < entries; ++i) {
 769                 cbrle = cbrlo[i];
 770                 if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
 771                         /* invalid entry */
 772                         break;
 773                 /* try to free backing */
 774                 __gmap_zap(gmap, cbrle);
 775         }
 776         up_read(&gmap->mm->mmap_sem);
 777         if (i < entries)
 778                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 779         return 0;
 780 }
 781
 782 static const intercept_handler_t b9_handlers[256] = {
 783         [0x8a] = handle_ipte_interlock,
 784         [0x8d] = handle_epsw,
 785         [0x8e] = handle_ipte_interlock,
 786         [0x8f] = handle_ipte_interlock,
 787         [0xab] = handle_essa,
 788         [0xaf] = handle_pfmf,
 789 };
 790
 791 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
 792 {
 793         intercept_handler_t handler;
 794
 795         /* This is handled just as for the B2 instructions. */
 796         handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
 797         if (handler)
 798                 return handler(vcpu);
 799
 800         return -EOPNOTSUPP;
 801 }
 802
 803 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
 804 {
 805         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 806         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 807         int reg, rc, nr_regs;
 808         u32 ctl_array[16];
 809         u64 ga;
 810         ar_t ar;
 811
 812         vcpu->stat.instruction_lctl++;
 813
 814         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 815                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 816
 817         ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
 818
 819         if (ga & 3)
 820                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 821
 822         VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
 823         trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
 824
 825         nr_regs = ((reg3 - reg1) & 0xf) + 1;
 826         rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
 827         if (rc)
 828                 return kvm_s390_inject_prog_cond(vcpu, rc);
 829         reg = reg1;
 830         nr_regs = 0;
 831         do {
 832                 vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
 833                 vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
 834                 if (reg == reg3)
 835                         break;
 836                 reg = (reg + 1) % 16;
 837         } while (1);
 838         kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 839         return 0;
 840 }
 841
 842 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
 843 {
 844         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 845         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 846         int reg, rc, nr_regs;
 847         u32 ctl_array[16];
 848         u64 ga;
 849         ar_t ar;
 850
 851         vcpu->stat.instruction_stctl++;
 852
 853         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 854                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 855
 856         ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
 857
 858         if (ga & 3)
 859                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 860
 861         VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
 862         trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
 863
 864         reg = reg1;
 865         nr_regs = 0;
 866         do {
 867                 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
 868                 if (reg == reg3)
 869                         break;
 870                 reg = (reg + 1) % 16;
 871         } while (1);
 872         rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
 873         return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
 874 }
 875
 876 static int handle_lctlg(struct kvm_vcpu *vcpu)
 877 {
 878         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 879         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 880         int reg, rc, nr_regs;
 881         u64 ctl_array[16];
 882         u64 ga;
 883         ar_t ar;
 884
 885         vcpu->stat.instruction_lctlg++;
 886
 887         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 888                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 889
 890         ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
 891
 892         if (ga & 7)
 893                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 894
 895         VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
 896         trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
 897
 898         nr_regs = ((reg3 - reg1) & 0xf) + 1;
 899         rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
 900         if (rc)
 901                 return kvm_s390_inject_prog_cond(vcpu, rc);
 902         reg = reg1;
 903         nr_regs = 0;
 904         do {
 905                 vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
 906                 if (reg == reg3)
 907                         break;
 908                 reg = (reg + 1) % 16;
 909         } while (1);
 910         kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 911         return 0;
 912 }
 913
 914 static int handle_stctg(struct kvm_vcpu *vcpu)
 915 {
 916         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 917         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 918         int reg, rc, nr_regs;
 919         u64 ctl_array[16];
 920         u64 ga;
 921         ar_t ar;
 922
 923         vcpu->stat.instruction_stctg++;
 924
 925         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 926                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 927
 928         ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
 929
 930         if (ga & 7)
 931                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 932
 933         VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
 934         trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
 935
 936         reg = reg1;
 937         nr_regs = 0;
 938         do {
 939                 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
 940                 if (reg == reg3)
 941                         break;
 942                 reg = (reg + 1) % 16;
 943         } while (1);
 944         rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
 945         return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
 946 }
 947
 948 static const intercept_handler_t eb_handlers[256] = {
 949         [0x2f] = handle_lctlg,
 950         [0x25] = handle_stctg,
 951 };
 952
 953 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
 954 {
 955         intercept_handler_t handler;
 956
 957         handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
 958         if (handler)
 959                 return handler(vcpu);
 960         return -EOPNOTSUPP;
 961 }
 962
 963 static int handle_tprot(struct kvm_vcpu *vcpu)
 964 {
 965         u64 address1, address2;
 966         unsigned long hva, gpa;
 967         int ret = 0, cc = 0;
 968         bool writable;
 969         ar_t ar;
 970
 971         vcpu->stat.instruction_tprot++;
 972
 973         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 974                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 975
 976         kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
 977
 978         /* we only handle the Linux memory detection case:
 979          * access key == 0
 980          * everything else goes to userspace. */
 981         if (address2 & 0xf0)
 982                 return -EOPNOTSUPP;
 983         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
 984                 ipte_lock(vcpu);
 985         ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
 986         if (ret == PGM_PROTECTION) {
 987                 /* Write protected? Try again with read-only... */
 988                 cc = 1;
 989                 ret = guest_translate_address(vcpu, address1, ar, &gpa,
 990                                               GACC_FETCH);
 991         }
 992         if (ret) {
 993                 if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
 994                         ret = kvm_s390_inject_program_int(vcpu, ret);
 995                 } else if (ret > 0) {
 996                         /* Translation not available */
 997                         kvm_s390_set_psw_cc(vcpu, 3);
 998                         ret = 0;
 999                 }
1000                 goto out_unlock;
1001         }
1002
1003         hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1004         if (kvm_is_error_hva(hva)) {
1005                 ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1006         } else {
1007                 if (!writable)
1008                         cc = 1;         /* Write not permitted ==> read-only */
1009                 kvm_s390_set_psw_cc(vcpu, cc);
1010                 /* Note: CC2 only occurs for storage keys (not supported yet) */
1011         }
1012 out_unlock:
1013         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1014                 ipte_unlock(vcpu);
1015         return ret;
1016 }
1017
1018 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1019 {
1020         /* For e5xx... instructions we only handle TPROT */
1021         if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1022                 return handle_tprot(vcpu);
1023         return -EOPNOTSUPP;
1024 }
1025
1026 static int handle_sckpf(struct kvm_vcpu *vcpu)
1027 {
1028         u32 value;
1029
1030         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1031                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1032
1033         if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1034                 return kvm_s390_inject_program_int(vcpu,
1035                                                    PGM_SPECIFICATION);
1036
1037         value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1038         vcpu->arch.sie_block->todpr = value;
1039
1040         return 0;
1041 }
1042
1043 static const intercept_handler_t x01_handlers[256] = {
1044         [0x07] = handle_sckpf,
1045 };
1046
1047 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1048 {
1049         intercept_handler_t handler;
1050
1051         handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1052         if (handler)
1053                 return handler(vcpu);
1054         return -EOPNOTSUPP;
1055 }