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/io.h>
  27 #include <asm/ptrace.h>
  28 #include <asm/compat.h>
  29 #include "gaccess.h"
  30 #include "kvm-s390.h"
  31 #include "trace.h"
  32
  33 /* Handle SCK (SET CLOCK) interception */
  34 static int handle_set_clock(struct kvm_vcpu *vcpu)
  35 {
  36         struct kvm_vcpu *cpup;
  37         s64 hostclk, val;
  38         int i, rc;
  39         u64 op2;
  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);
  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, &val, sizeof(val));
  48         if (rc)
  49                 return kvm_s390_inject_prog_cond(vcpu, rc);
  50
  51         if (store_tod_clock(&hostclk)) {
  52                 kvm_s390_set_psw_cc(vcpu, 3);
  53                 return 0;
  54         }
  55         val = (val - hostclk) & ~0x3fUL;
  56
  57         mutex_lock(&vcpu->kvm->lock);
  58         kvm_for_each_vcpu(i, cpup, vcpu->kvm)
  59                 cpup->arch.sie_block->epoch = val;
  60         mutex_unlock(&vcpu->kvm->lock);
  61
  62         kvm_s390_set_psw_cc(vcpu, 0);
  63         return 0;
  64 }
  65
  66 static int handle_set_prefix(struct kvm_vcpu *vcpu)
  67 {
  68         u64 operand2;
  69         u32 address;
  70         int rc;
  71
  72         vcpu->stat.instruction_spx++;
  73
  74         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  75                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
  76
  77         operand2 = kvm_s390_get_base_disp_s(vcpu);
  78
  79         /* must be word boundary */
  80         if (operand2 & 3)
  81                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
  82
  83         /* get the value */
  84         rc = read_guest(vcpu, operand2, &address, sizeof(address));
  85         if (rc)
  86                 return kvm_s390_inject_prog_cond(vcpu, rc);
  87
  88         address &= 0x7fffe000u;
  89
  90         /*
  91          * Make sure the new value is valid memory. We only need to check the
  92          * first page, since address is 8k aligned and memory pieces are always
  93          * at least 1MB aligned and have at least a size of 1MB.
  94          */
  95         if (kvm_is_error_gpa(vcpu->kvm, address))
  96                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
  97
  98         kvm_s390_set_prefix(vcpu, address);
  99
 100         VCPU_EVENT(vcpu, 5, "setting prefix to %x", address);
 101         trace_kvm_s390_handle_prefix(vcpu, 1, address);
 102         return 0;
 103 }
 104
 105 static int handle_store_prefix(struct kvm_vcpu *vcpu)
 106 {
 107         u64 operand2;
 108         u32 address;
 109         int rc;
 110
 111         vcpu->stat.instruction_stpx++;
 112
 113         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 114                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 115
 116         operand2 = kvm_s390_get_base_disp_s(vcpu);
 117
 118         /* must be word boundary */
 119         if (operand2 & 3)
 120                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 121
 122         address = kvm_s390_get_prefix(vcpu);
 123
 124         /* get the value */
 125         rc = write_guest(vcpu, operand2, &address, sizeof(address));
 126         if (rc)
 127                 return kvm_s390_inject_prog_cond(vcpu, rc);
 128
 129         VCPU_EVENT(vcpu, 5, "storing prefix to %x", address);
 130         trace_kvm_s390_handle_prefix(vcpu, 0, address);
 131         return 0;
 132 }
 133
 134 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
 135 {
 136         u16 vcpu_id = vcpu->vcpu_id;
 137         u64 ga;
 138         int rc;
 139
 140         vcpu->stat.instruction_stap++;
 141
 142         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 143                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 144
 145         ga = kvm_s390_get_base_disp_s(vcpu);
 146
 147         if (ga & 1)
 148                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 149
 150         rc = write_guest(vcpu, ga, &vcpu_id, sizeof(vcpu_id));
 151         if (rc)
 152                 return kvm_s390_inject_prog_cond(vcpu, rc);
 153
 154         VCPU_EVENT(vcpu, 5, "storing cpu address to %llx", ga);
 155         trace_kvm_s390_handle_stap(vcpu, ga);
 156         return 0;
 157 }
 158
 159 static int __skey_check_enable(struct kvm_vcpu *vcpu)
 160 {
 161         int rc = 0;
 162         if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
 163                 return rc;
 164
 165         rc = s390_enable_skey();
 166         trace_kvm_s390_skey_related_inst(vcpu);
 167         vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
 168         return rc;
 169 }
 170
 171
 172 static int handle_skey(struct kvm_vcpu *vcpu)
 173 {
 174         int rc = __skey_check_enable(vcpu);
 175
 176         if (rc)
 177                 return rc;
 178         vcpu->stat.instruction_storage_key++;
 179
 180         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 181                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 182
 183         kvm_s390_rewind_psw(vcpu, 4);
 184         VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
 185         return 0;
 186 }
 187
 188 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
 189 {
 190         vcpu->stat.instruction_ipte_interlock++;
 191         if (psw_bits(vcpu->arch.sie_block->gpsw).p)
 192                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 193         wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
 194         kvm_s390_rewind_psw(vcpu, 4);
 195         VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
 196         return 0;
 197 }
 198
 199 static int handle_test_block(struct kvm_vcpu *vcpu)
 200 {
 201         gpa_t addr;
 202         int reg2;
 203
 204         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 205                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 206
 207         kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
 208         addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 209         addr = kvm_s390_logical_to_effective(vcpu, addr);
 210         if (kvm_s390_check_low_addr_protection(vcpu, addr))
 211                 return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 212         addr = kvm_s390_real_to_abs(vcpu, addr);
 213
 214         if (kvm_is_error_gpa(vcpu->kvm, addr))
 215                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 216         /*
 217          * We don't expect errors on modern systems, and do not care
 218          * about storage keys (yet), so let's just clear the page.
 219          */
 220         if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
 221                 return -EFAULT;
 222         kvm_s390_set_psw_cc(vcpu, 0);
 223         vcpu->run->s.regs.gprs[0] = 0;
 224         return 0;
 225 }
 226
 227 static int handle_tpi(struct kvm_vcpu *vcpu)
 228 {
 229         struct kvm_s390_interrupt_info *inti;
 230         unsigned long len;
 231         u32 tpi_data[3];
 232         int cc, rc;
 233         u64 addr;
 234
 235         rc = 0;
 236         addr = kvm_s390_get_base_disp_s(vcpu);
 237         if (addr & 3)
 238                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 239         cc = 0;
 240         inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
 241         if (!inti)
 242                 goto no_interrupt;
 243         cc = 1;
 244         tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
 245         tpi_data[1] = inti->io.io_int_parm;
 246         tpi_data[2] = inti->io.io_int_word;
 247         if (addr) {
 248                 /*
 249                  * Store the two-word I/O interruption code into the
 250                  * provided area.
 251                  */
 252                 len = sizeof(tpi_data) - 4;
 253                 rc = write_guest(vcpu, addr, &tpi_data, len);
 254                 if (rc)
 255                         return kvm_s390_inject_prog_cond(vcpu, rc);
 256         } else {
 257                 /*
 258                  * Store the three-word I/O interruption code into
 259                  * the appropriate lowcore area.
 260                  */
 261                 len = sizeof(tpi_data);
 262                 if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len))
 263                         rc = -EFAULT;
 264         }
 265         /*
 266          * If we encounter a problem storing the interruption code, the
 267          * instruction is suppressed from the guest's view: reinject the
 268          * interrupt.
 269          */
 270         if (!rc)
 271                 kfree(inti);
 272         else
 273                 kvm_s390_reinject_io_int(vcpu->kvm, inti);
 274 no_interrupt:
 275         /* Set condition code and we're done. */
 276         if (!rc)
 277                 kvm_s390_set_psw_cc(vcpu, cc);
 278         return rc ? -EFAULT : 0;
 279 }
 280
 281 static int handle_tsch(struct kvm_vcpu *vcpu)
 282 {
 283         struct kvm_s390_interrupt_info *inti;
 284
 285         inti = kvm_s390_get_io_int(vcpu->kvm, 0,
 286                                    vcpu->run->s.regs.gprs[1]);
 287
 288         /*
 289          * Prepare exit to userspace.
 290          * We indicate whether we dequeued a pending I/O interrupt
 291          * so that userspace can re-inject it if the instruction gets
 292          * a program check. While this may re-order the pending I/O
 293          * interrupts, this is no problem since the priority is kept
 294          * intact.
 295          */
 296         vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
 297         vcpu->run->s390_tsch.dequeued = !!inti;
 298         if (inti) {
 299                 vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
 300                 vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
 301                 vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
 302                 vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
 303         }
 304         vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
 305         kfree(inti);
 306         return -EREMOTE;
 307 }
 308
 309 static int handle_io_inst(struct kvm_vcpu *vcpu)
 310 {
 311         VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
 312
 313         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 314                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 315
 316         if (vcpu->kvm->arch.css_support) {
 317                 /*
 318                  * Most I/O instructions will be handled by userspace.
 319                  * Exceptions are tpi and the interrupt portion of tsch.
 320                  */
 321                 if (vcpu->arch.sie_block->ipa == 0xb236)
 322                         return handle_tpi(vcpu);
 323                 if (vcpu->arch.sie_block->ipa == 0xb235)
 324                         return handle_tsch(vcpu);
 325                 /* Handle in userspace. */
 326                 return -EOPNOTSUPP;
 327         } else {
 328                 /*
 329                  * Set condition code 3 to stop the guest from issuing channel
 330                  * I/O instructions.
 331                  */
 332                 kvm_s390_set_psw_cc(vcpu, 3);
 333                 return 0;
 334         }
 335 }
 336
 337 static int handle_stfl(struct kvm_vcpu *vcpu)
 338 {
 339         int rc;
 340         unsigned int fac;
 341
 342         vcpu->stat.instruction_stfl++;
 343
 344         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 345                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 346
 347         /*
 348          * We need to shift the lower 32 facility bits (bit 0-31) from a u64
 349          * into a u32 memory representation. They will remain bits 0-31.
 350          */
 351         fac = *vcpu->kvm->arch.model.fac->list >> 32;
 352         rc = write_guest_lc(vcpu, offsetof(struct _lowcore, stfl_fac_list),
 353                             &fac, sizeof(fac));
 354         if (rc)
 355                 return rc;
 356         VCPU_EVENT(vcpu, 5, "store facility list value %x", fac);
 357         trace_kvm_s390_handle_stfl(vcpu, fac);
 358         return 0;
 359 }
 360
 361 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
 362 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
 363 #define PSW_ADDR_24 0x0000000000ffffffUL
 364 #define PSW_ADDR_31 0x000000007fffffffUL
 365
 366 int is_valid_psw(psw_t *psw)
 367 {
 368         if (psw->mask & PSW_MASK_UNASSIGNED)
 369                 return 0;
 370         if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
 371                 if (psw->addr & ~PSW_ADDR_31)
 372                         return 0;
 373         }
 374         if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
 375                 return 0;
 376         if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
 377                 return 0;
 378         if (psw->addr & 1)
 379                 return 0;
 380         return 1;
 381 }
 382
 383 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
 384 {
 385         psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
 386         psw_compat_t new_psw;
 387         u64 addr;
 388         int rc;
 389
 390         if (gpsw->mask & PSW_MASK_PSTATE)
 391                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 392
 393         addr = kvm_s390_get_base_disp_s(vcpu);
 394         if (addr & 7)
 395                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 396
 397         rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
 398         if (rc)
 399                 return kvm_s390_inject_prog_cond(vcpu, rc);
 400         if (!(new_psw.mask & PSW32_MASK_BASE))
 401                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 402         gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
 403         gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
 404         gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
 405         if (!is_valid_psw(gpsw))
 406                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 407         return 0;
 408 }
 409
 410 static int handle_lpswe(struct kvm_vcpu *vcpu)
 411 {
 412         psw_t new_psw;
 413         u64 addr;
 414         int rc;
 415
 416         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 417                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 418
 419         addr = kvm_s390_get_base_disp_s(vcpu);
 420         if (addr & 7)
 421                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 422         rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
 423         if (rc)
 424                 return kvm_s390_inject_prog_cond(vcpu, rc);
 425         vcpu->arch.sie_block->gpsw = new_psw;
 426         if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
 427                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 428         return 0;
 429 }
 430
 431 static int handle_stidp(struct kvm_vcpu *vcpu)
 432 {
 433         u64 stidp_data = vcpu->arch.stidp_data;
 434         u64 operand2;
 435         int rc;
 436
 437         vcpu->stat.instruction_stidp++;
 438
 439         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 440                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 441
 442         operand2 = kvm_s390_get_base_disp_s(vcpu);
 443
 444         if (operand2 & 7)
 445                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 446
 447         rc = write_guest(vcpu, operand2, &stidp_data, sizeof(stidp_data));
 448         if (rc)
 449                 return kvm_s390_inject_prog_cond(vcpu, rc);
 450
 451         VCPU_EVENT(vcpu, 5, "%s", "store cpu id");
 452         return 0;
 453 }
 454
 455 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
 456 {
 457         int cpus = 0;
 458         int n;
 459
 460         cpus = atomic_read(&vcpu->kvm->online_vcpus);
 461
 462         /* deal with other level 3 hypervisors */
 463         if (stsi(mem, 3, 2, 2))
 464                 mem->count = 0;
 465         if (mem->count < 8)
 466                 mem->count++;
 467         for (n = mem->count - 1; n > 0 ; n--)
 468                 memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
 469
 470         mem->vm[0].cpus_total = cpus;
 471         mem->vm[0].cpus_configured = cpus;
 472         mem->vm[0].cpus_standby = 0;
 473         mem->vm[0].cpus_reserved = 0;
 474         mem->vm[0].caf = 1000;
 475         memcpy(mem->vm[0].name, "KVMguest", 8);
 476         ASCEBC(mem->vm[0].name, 8);
 477         memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
 478         ASCEBC(mem->vm[0].cpi, 16);
 479 }
 480
 481 static int handle_stsi(struct kvm_vcpu *vcpu)
 482 {
 483         int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
 484         int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
 485         int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
 486         unsigned long mem = 0;
 487         u64 operand2;
 488         int rc = 0;
 489
 490         vcpu->stat.instruction_stsi++;
 491         VCPU_EVENT(vcpu, 4, "stsi: fc: %x sel1: %x sel2: %x", fc, sel1, sel2);
 492
 493         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 494                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 495
 496         if (fc > 3) {
 497                 kvm_s390_set_psw_cc(vcpu, 3);
 498                 return 0;
 499         }
 500
 501         if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
 502             || vcpu->run->s.regs.gprs[1] & 0xffff0000)
 503                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 504
 505         if (fc == 0) {
 506                 vcpu->run->s.regs.gprs[0] = 3 << 28;
 507                 kvm_s390_set_psw_cc(vcpu, 0);
 508                 return 0;
 509         }
 510
 511         operand2 = kvm_s390_get_base_disp_s(vcpu);
 512
 513         if (operand2 & 0xfff)
 514                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 515
 516         switch (fc) {
 517         case 1: /* same handling for 1 and 2 */
 518         case 2:
 519                 mem = get_zeroed_page(GFP_KERNEL);
 520                 if (!mem)
 521                         goto out_no_data;
 522                 if (stsi((void *) mem, fc, sel1, sel2))
 523                         goto out_no_data;
 524                 break;
 525         case 3:
 526                 if (sel1 != 2 || sel2 != 2)
 527                         goto out_no_data;
 528                 mem = get_zeroed_page(GFP_KERNEL);
 529                 if (!mem)
 530                         goto out_no_data;
 531                 handle_stsi_3_2_2(vcpu, (void *) mem);
 532                 break;
 533         }
 534
 535         rc = write_guest(vcpu, operand2, (void *)mem, PAGE_SIZE);
 536         if (rc) {
 537                 rc = kvm_s390_inject_prog_cond(vcpu, rc);
 538                 goto out;
 539         }
 540         trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 541         free_page(mem);
 542         kvm_s390_set_psw_cc(vcpu, 0);
 543         vcpu->run->s.regs.gprs[0] = 0;
 544         return 0;
 545 out_no_data:
 546         kvm_s390_set_psw_cc(vcpu, 3);
 547 out:
 548         free_page(mem);
 549         return rc;
 550 }
 551
 552 static const intercept_handler_t b2_handlers[256] = {
 553         [0x02] = handle_stidp,
 554         [0x04] = handle_set_clock,
 555         [0x10] = handle_set_prefix,
 556         [0x11] = handle_store_prefix,
 557         [0x12] = handle_store_cpu_address,
 558         [0x21] = handle_ipte_interlock,
 559         [0x29] = handle_skey,
 560         [0x2a] = handle_skey,
 561         [0x2b] = handle_skey,
 562         [0x2c] = handle_test_block,
 563         [0x30] = handle_io_inst,
 564         [0x31] = handle_io_inst,
 565         [0x32] = handle_io_inst,
 566         [0x33] = handle_io_inst,
 567         [0x34] = handle_io_inst,
 568         [0x35] = handle_io_inst,
 569         [0x36] = handle_io_inst,
 570         [0x37] = handle_io_inst,
 571         [0x38] = handle_io_inst,
 572         [0x39] = handle_io_inst,
 573         [0x3a] = handle_io_inst,
 574         [0x3b] = handle_io_inst,
 575         [0x3c] = handle_io_inst,
 576         [0x50] = handle_ipte_interlock,
 577         [0x5f] = handle_io_inst,
 578         [0x74] = handle_io_inst,
 579         [0x76] = handle_io_inst,
 580         [0x7d] = handle_stsi,
 581         [0xb1] = handle_stfl,
 582         [0xb2] = handle_lpswe,
 583 };
 584
 585 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
 586 {
 587         intercept_handler_t handler;
 588
 589         /*
 590          * A lot of B2 instructions are priviledged. Here we check for
 591          * the privileged ones, that we can handle in the kernel.
 592          * Anything else goes to userspace.
 593          */
 594         handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
 595         if (handler)
 596                 return handler(vcpu);
 597
 598         return -EOPNOTSUPP;
 599 }
 600
 601 static int handle_epsw(struct kvm_vcpu *vcpu)
 602 {
 603         int reg1, reg2;
 604
 605         kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 606
 607         /* This basically extracts the mask half of the psw. */
 608         vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
 609         vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
 610         if (reg2) {
 611                 vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
 612                 vcpu->run->s.regs.gprs[reg2] |=
 613                         vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
 614         }
 615         return 0;
 616 }
 617
 618 #define PFMF_RESERVED   0xfffc0101UL
 619 #define PFMF_SK         0x00020000UL
 620 #define PFMF_CF         0x00010000UL
 621 #define PFMF_UI         0x00008000UL
 622 #define PFMF_FSC        0x00007000UL
 623 #define PFMF_NQ         0x00000800UL
 624 #define PFMF_MR         0x00000400UL
 625 #define PFMF_MC         0x00000200UL
 626 #define PFMF_KEY        0x000000feUL
 627
 628 static int handle_pfmf(struct kvm_vcpu *vcpu)
 629 {
 630         int reg1, reg2;
 631         unsigned long start, end;
 632
 633         vcpu->stat.instruction_pfmf++;
 634
 635         kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 636
 637         if (!MACHINE_HAS_PFMF)
 638                 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
 639
 640         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 641                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 642
 643         if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
 644                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 645
 646         /* Only provide non-quiescing support if the host supports it */
 647         if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
 648                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 649
 650         /* No support for conditional-SSKE */
 651         if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
 652                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 653
 654         start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 655         start = kvm_s390_logical_to_effective(vcpu, start);
 656
 657         switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
 658         case 0x00000000:
 659                 end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
 660                 break;
 661         case 0x00001000:
 662                 end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
 663                 break;
 664         /* We dont support EDAT2
 665         case 0x00002000:
 666                 end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
 667                 break;*/
 668         default:
 669                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 670         }
 671
 672         if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
 673                 if (kvm_s390_check_low_addr_protection(vcpu, start))
 674                         return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 675         }
 676
 677         while (start < end) {
 678                 unsigned long useraddr, abs_addr;
 679
 680                 /* Translate guest address to host address */
 681                 if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
 682                         abs_addr = kvm_s390_real_to_abs(vcpu, start);
 683                 else
 684                         abs_addr = start;
 685                 useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
 686                 if (kvm_is_error_hva(useraddr))
 687                         return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 688
 689                 if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
 690                         if (clear_user((void __user *)useraddr, PAGE_SIZE))
 691                                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 692                 }
 693
 694                 if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
 695                         int rc = __skey_check_enable(vcpu);
 696
 697                         if (rc)
 698                                 return rc;
 699                         if (set_guest_storage_key(current->mm, useraddr,
 700                                         vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
 701                                         vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
 702                                 return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 703                 }
 704
 705                 start += PAGE_SIZE;
 706         }
 707         if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
 708                 vcpu->run->s.regs.gprs[reg2] = end;
 709         return 0;
 710 }
 711
 712 static int handle_essa(struct kvm_vcpu *vcpu)
 713 {
 714         /* entries expected to be 1FF */
 715         int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
 716         unsigned long *cbrlo, cbrle;
 717         struct gmap *gmap;
 718         int i;
 719
 720         VCPU_EVENT(vcpu, 5, "cmma release %d pages", entries);
 721         gmap = vcpu->arch.gmap;
 722         vcpu->stat.instruction_essa++;
 723         if (!kvm_s390_cmma_enabled(vcpu->kvm))
 724                 return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
 725
 726         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 727                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 728
 729         if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
 730                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 731
 732         /* Rewind PSW to repeat the ESSA instruction */
 733         kvm_s390_rewind_psw(vcpu, 4);
 734         vcpu->arch.sie_block->cbrlo &= PAGE_MASK;       /* reset nceo */
 735         cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
 736         down_read(&gmap->mm->mmap_sem);
 737         for (i = 0; i < entries; ++i) {
 738                 cbrle = cbrlo[i];
 739                 if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
 740                         /* invalid entry */
 741                         break;
 742                 /* try to free backing */
 743                 __gmap_zap(gmap, cbrle);
 744         }
 745         up_read(&gmap->mm->mmap_sem);
 746         if (i < entries)
 747                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 748         return 0;
 749 }
 750
 751 static const intercept_handler_t b9_handlers[256] = {
 752         [0x8a] = handle_ipte_interlock,
 753         [0x8d] = handle_epsw,
 754         [0x8e] = handle_ipte_interlock,
 755         [0x8f] = handle_ipte_interlock,
 756         [0xab] = handle_essa,
 757         [0xaf] = handle_pfmf,
 758 };
 759
 760 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
 761 {
 762         intercept_handler_t handler;
 763
 764         /* This is handled just as for the B2 instructions. */
 765         handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
 766         if (handler)
 767                 return handler(vcpu);
 768
 769         return -EOPNOTSUPP;
 770 }
 771
 772 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
 773 {
 774         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 775         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 776         int reg, rc, nr_regs;
 777         u32 ctl_array[16];
 778         u64 ga;
 779
 780         vcpu->stat.instruction_lctl++;
 781
 782         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 783                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 784
 785         ga = kvm_s390_get_base_disp_rs(vcpu);
 786
 787         if (ga & 3)
 788                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 789
 790         VCPU_EVENT(vcpu, 5, "lctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
 791         trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
 792
 793         nr_regs = ((reg3 - reg1) & 0xf) + 1;
 794         rc = read_guest(vcpu, ga, ctl_array, nr_regs * sizeof(u32));
 795         if (rc)
 796                 return kvm_s390_inject_prog_cond(vcpu, rc);
 797         reg = reg1;
 798         nr_regs = 0;
 799         do {
 800                 vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
 801                 vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
 802                 if (reg == reg3)
 803                         break;
 804                 reg = (reg + 1) % 16;
 805         } while (1);
 806         kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 807         return 0;
 808 }
 809
 810 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
 811 {
 812         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 813         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 814         int reg, rc, nr_regs;
 815         u32 ctl_array[16];
 816         u64 ga;
 817
 818         vcpu->stat.instruction_stctl++;
 819
 820         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 821                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 822
 823         ga = kvm_s390_get_base_disp_rs(vcpu);
 824
 825         if (ga & 3)
 826                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 827
 828         VCPU_EVENT(vcpu, 5, "stctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
 829         trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
 830
 831         reg = reg1;
 832         nr_regs = 0;
 833         do {
 834                 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
 835                 if (reg == reg3)
 836                         break;
 837                 reg = (reg + 1) % 16;
 838         } while (1);
 839         rc = write_guest(vcpu, ga, ctl_array, nr_regs * sizeof(u32));
 840         return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
 841 }
 842
 843 static int handle_lctlg(struct kvm_vcpu *vcpu)
 844 {
 845         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 846         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 847         int reg, rc, nr_regs;
 848         u64 ctl_array[16];
 849         u64 ga;
 850
 851         vcpu->stat.instruction_lctlg++;
 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_rsy(vcpu);
 857
 858         if (ga & 7)
 859                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 860
 861         VCPU_EVENT(vcpu, 5, "lctlg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
 862         trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
 863
 864         nr_regs = ((reg3 - reg1) & 0xf) + 1;
 865         rc = read_guest(vcpu, ga, ctl_array, nr_regs * sizeof(u64));
 866         if (rc)
 867                 return kvm_s390_inject_prog_cond(vcpu, rc);
 868         reg = reg1;
 869         nr_regs = 0;
 870         do {
 871                 vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
 872                 if (reg == reg3)
 873                         break;
 874                 reg = (reg + 1) % 16;
 875         } while (1);
 876         kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 877         return 0;
 878 }
 879
 880 static int handle_stctg(struct kvm_vcpu *vcpu)
 881 {
 882         int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
 883         int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
 884         int reg, rc, nr_regs;
 885         u64 ctl_array[16];
 886         u64 ga;
 887
 888         vcpu->stat.instruction_stctg++;
 889
 890         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 891                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 892
 893         ga = kvm_s390_get_base_disp_rsy(vcpu);
 894
 895         if (ga & 7)
 896                 return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 897
 898         VCPU_EVENT(vcpu, 5, "stctg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
 899         trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
 900
 901         reg = reg1;
 902         nr_regs = 0;
 903         do {
 904                 ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
 905                 if (reg == reg3)
 906                         break;
 907                 reg = (reg + 1) % 16;
 908         } while (1);
 909         rc = write_guest(vcpu, ga, ctl_array, nr_regs * sizeof(u64));
 910         return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
 911 }
 912
 913 static const intercept_handler_t eb_handlers[256] = {
 914         [0x2f] = handle_lctlg,
 915         [0x25] = handle_stctg,
 916 };
 917
 918 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
 919 {
 920         intercept_handler_t handler;
 921
 922         handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
 923         if (handler)
 924                 return handler(vcpu);
 925         return -EOPNOTSUPP;
 926 }
 927
 928 static int handle_tprot(struct kvm_vcpu *vcpu)
 929 {
 930         u64 address1, address2;
 931         unsigned long hva, gpa;
 932         int ret = 0, cc = 0;
 933         bool writable;
 934
 935         vcpu->stat.instruction_tprot++;
 936
 937         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 938                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 939
 940         kvm_s390_get_base_disp_sse(vcpu, &address1, &address2);
 941
 942         /* we only handle the Linux memory detection case:
 943          * access key == 0
 944          * everything else goes to userspace. */
 945         if (address2 & 0xf0)
 946                 return -EOPNOTSUPP;
 947         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
 948                 ipte_lock(vcpu);
 949         ret = guest_translate_address(vcpu, address1, &gpa, 1);
 950         if (ret == PGM_PROTECTION) {
 951                 /* Write protected? Try again with read-only... */
 952                 cc = 1;
 953                 ret = guest_translate_address(vcpu, address1, &gpa, 0);
 954         }
 955         if (ret) {
 956                 if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
 957                         ret = kvm_s390_inject_program_int(vcpu, ret);
 958                 } else if (ret > 0) {
 959                         /* Translation not available */
 960                         kvm_s390_set_psw_cc(vcpu, 3);
 961                         ret = 0;
 962                 }
 963                 goto out_unlock;
 964         }
 965
 966         hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
 967         if (kvm_is_error_hva(hva)) {
 968                 ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 969         } else {
 970                 if (!writable)
 971                         cc = 1;         /* Write not permitted ==> read-only */
 972                 kvm_s390_set_psw_cc(vcpu, cc);
 973                 /* Note: CC2 only occurs for storage keys (not supported yet) */
 974         }
 975 out_unlock:
 976         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
 977                 ipte_unlock(vcpu);
 978         return ret;
 979 }
 980
 981 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
 982 {
 983         /* For e5xx... instructions we only handle TPROT */
 984         if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
 985                 return handle_tprot(vcpu);
 986         return -EOPNOTSUPP;
 987 }
 988
 989 static int handle_sckpf(struct kvm_vcpu *vcpu)
 990 {
 991         u32 value;
 992
 993         if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 994                 return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 995
 996         if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
 997                 return kvm_s390_inject_program_int(vcpu,
 998                                                    PGM_SPECIFICATION);
 999
1000         value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1001         vcpu->arch.sie_block->todpr = value;
1002
1003         return 0;
1004 }
1005
1006 static const intercept_handler_t x01_handlers[256] = {
1007         [0x07] = handle_sckpf,
1008 };
1009
1010 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1011 {
1012         intercept_handler_t handler;
1013
1014         handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1015         if (handler)
1016                 return handler(vcpu);
1017         return -EOPNOTSUPP;
1018 }