Linux Audio

Check our new training course

Loading...
v6.2
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * handling privileged instructions
   4 *
   5 * Copyright IBM Corp. 2008, 2020
   6 *
   7 *    Author(s): Carsten Otte <cotte@de.ibm.com>
   8 *               Christian Borntraeger <borntraeger@de.ibm.com>
   9 */
  10
  11#include <linux/kvm.h>
  12#include <linux/gfp.h>
  13#include <linux/errno.h>
  14#include <linux/mm_types.h>
  15#include <linux/pgtable.h>
  16
  17#include <asm/asm-offsets.h>
  18#include <asm/facility.h>
  19#include <asm/current.h>
  20#include <asm/debug.h>
  21#include <asm/ebcdic.h>
  22#include <asm/sysinfo.h>
  23#include <asm/page-states.h>
  24#include <asm/gmap.h>
  25#include <asm/io.h>
  26#include <asm/ptrace.h>
  27#include <asm/sclp.h>
  28#include <asm/ap.h>
  29#include "gaccess.h"
  30#include "kvm-s390.h"
  31#include "trace.h"
  32
  33static int handle_ri(struct kvm_vcpu *vcpu)
  34{
  35	vcpu->stat.instruction_ri++;
  36
  37	if (test_kvm_facility(vcpu->kvm, 64)) {
  38		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
  39		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
  40		kvm_s390_retry_instr(vcpu);
  41		return 0;
  42	} else
  43		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
  44}
  45
  46int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
  47{
  48	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
  49		return handle_ri(vcpu);
  50	else
  51		return -EOPNOTSUPP;
  52}
  53
  54static int handle_gs(struct kvm_vcpu *vcpu)
  55{
  56	vcpu->stat.instruction_gs++;
  57
  58	if (test_kvm_facility(vcpu->kvm, 133)) {
  59		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (lazy)");
  60		preempt_disable();
  61		__ctl_set_bit(2, 4);
  62		current->thread.gs_cb = (struct gs_cb *)&vcpu->run->s.regs.gscb;
  63		restore_gs_cb(current->thread.gs_cb);
  64		preempt_enable();
  65		vcpu->arch.sie_block->ecb |= ECB_GS;
  66		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
  67		vcpu->arch.gs_enabled = 1;
  68		kvm_s390_retry_instr(vcpu);
  69		return 0;
  70	} else
  71		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
  72}
  73
  74int kvm_s390_handle_e3(struct kvm_vcpu *vcpu)
  75{
  76	int code = vcpu->arch.sie_block->ipb & 0xff;
  77
  78	if (code == 0x49 || code == 0x4d)
  79		return handle_gs(vcpu);
  80	else
  81		return -EOPNOTSUPP;
  82}
  83/* Handle SCK (SET CLOCK) interception */
  84static int handle_set_clock(struct kvm_vcpu *vcpu)
  85{
  86	struct kvm_s390_vm_tod_clock gtod = { 0 };
  87	int rc;
  88	u8 ar;
  89	u64 op2;
  90
  91	vcpu->stat.instruction_sck++;
  92
  93	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  94		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
  95
  96	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
  97	if (op2 & 7)	/* Operand must be on a doubleword boundary */
  98		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
  99	rc = read_guest(vcpu, op2, ar, &gtod.tod, sizeof(gtod.tod));
 100	if (rc)
 101		return kvm_s390_inject_prog_cond(vcpu, rc);
 102
 103	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", gtod.tod);
 104	/*
 105	 * To set the TOD clock the kvm lock must be taken, but the vcpu lock
 106	 * is already held in handle_set_clock. The usual lock order is the
 107	 * opposite.  As SCK is deprecated and should not be used in several
 108	 * cases, for example when the multiple epoch facility or TOD clock
 109	 * steering facility is installed (see Principles of Operation),  a
 110	 * slow path can be used.  If the lock can not be taken via try_lock,
 111	 * the instruction will be retried via -EAGAIN at a later point in
 112	 * time.
 113	 */
 114	if (!kvm_s390_try_set_tod_clock(vcpu->kvm, &gtod)) {
 115		kvm_s390_retry_instr(vcpu);
 116		return -EAGAIN;
 117	}
 118
 119	kvm_s390_set_psw_cc(vcpu, 0);
 120	return 0;
 121}
 122
 123static int handle_set_prefix(struct kvm_vcpu *vcpu)
 124{
 125	u64 operand2;
 126	u32 address;
 127	int rc;
 128	u8 ar;
 129
 130	vcpu->stat.instruction_spx++;
 131
 132	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 133		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 134
 135	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 136
 137	/* must be word boundary */
 138	if (operand2 & 3)
 139		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 140
 141	/* get the value */
 142	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
 143	if (rc)
 144		return kvm_s390_inject_prog_cond(vcpu, rc);
 145
 146	address &= 0x7fffe000u;
 147
 148	/*
 149	 * Make sure the new value is valid memory. We only need to check the
 150	 * first page, since address is 8k aligned and memory pieces are always
 151	 * at least 1MB aligned and have at least a size of 1MB.
 152	 */
 153	if (kvm_is_error_gpa(vcpu->kvm, address))
 154		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 155
 156	kvm_s390_set_prefix(vcpu, address);
 157	trace_kvm_s390_handle_prefix(vcpu, 1, address);
 158	return 0;
 159}
 160
 161static int handle_store_prefix(struct kvm_vcpu *vcpu)
 162{
 163	u64 operand2;
 164	u32 address;
 165	int rc;
 166	u8 ar;
 167
 168	vcpu->stat.instruction_stpx++;
 169
 170	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 171		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 172
 173	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 174
 175	/* must be word boundary */
 176	if (operand2 & 3)
 177		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 178
 179	address = kvm_s390_get_prefix(vcpu);
 180
 181	/* get the value */
 182	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
 183	if (rc)
 184		return kvm_s390_inject_prog_cond(vcpu, rc);
 185
 186	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
 187	trace_kvm_s390_handle_prefix(vcpu, 0, address);
 188	return 0;
 189}
 190
 191static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
 192{
 193	u16 vcpu_id = vcpu->vcpu_id;
 194	u64 ga;
 195	int rc;
 196	u8 ar;
 197
 198	vcpu->stat.instruction_stap++;
 199
 200	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 201		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 202
 203	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
 204
 205	if (ga & 1)
 206		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 207
 208	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
 209	if (rc)
 210		return kvm_s390_inject_prog_cond(vcpu, rc);
 211
 212	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
 213	trace_kvm_s390_handle_stap(vcpu, ga);
 214	return 0;
 215}
 216
 217int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
 218{
 219	int rc;
 220
 221	trace_kvm_s390_skey_related_inst(vcpu);
 222	/* Already enabled? */
 223	if (vcpu->arch.skey_enabled)
 224		return 0;
 225
 226	rc = s390_enable_skey();
 227	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
 228	if (rc)
 229		return rc;
 230
 231	if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
 232		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS);
 233	if (!vcpu->kvm->arch.use_skf)
 234		vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
 235	else
 236		vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
 237	vcpu->arch.skey_enabled = true;
 238	return 0;
 239}
 240
 241static int try_handle_skey(struct kvm_vcpu *vcpu)
 242{
 243	int rc;
 244
 245	rc = kvm_s390_skey_check_enable(vcpu);
 246	if (rc)
 247		return rc;
 248	if (vcpu->kvm->arch.use_skf) {
 249		/* with storage-key facility, SIE interprets it for us */
 250		kvm_s390_retry_instr(vcpu);
 251		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
 252		return -EAGAIN;
 253	}
 254	return 0;
 255}
 256
 257static int handle_iske(struct kvm_vcpu *vcpu)
 258{
 259	unsigned long gaddr, vmaddr;
 260	unsigned char key;
 261	int reg1, reg2;
 262	bool unlocked;
 263	int rc;
 264
 265	vcpu->stat.instruction_iske++;
 266
 267	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 268		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 269
 270	rc = try_handle_skey(vcpu);
 271	if (rc)
 272		return rc != -EAGAIN ? rc : 0;
 273
 274	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 275
 276	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 277	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
 278	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
 279	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
 280	if (kvm_is_error_hva(vmaddr))
 281		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 282retry:
 283	unlocked = false;
 284	mmap_read_lock(current->mm);
 285	rc = get_guest_storage_key(current->mm, vmaddr, &key);
 286
 287	if (rc) {
 288		rc = fixup_user_fault(current->mm, vmaddr,
 289				      FAULT_FLAG_WRITE, &unlocked);
 290		if (!rc) {
 291			mmap_read_unlock(current->mm);
 292			goto retry;
 293		}
 294	}
 295	mmap_read_unlock(current->mm);
 296	if (rc == -EFAULT)
 297		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 298	if (rc < 0)
 299		return rc;
 300	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
 301	vcpu->run->s.regs.gprs[reg1] |= key;
 302	return 0;
 303}
 304
 305static int handle_rrbe(struct kvm_vcpu *vcpu)
 306{
 307	unsigned long vmaddr, gaddr;
 308	int reg1, reg2;
 309	bool unlocked;
 310	int rc;
 311
 312	vcpu->stat.instruction_rrbe++;
 313
 314	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 315		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 316
 317	rc = try_handle_skey(vcpu);
 318	if (rc)
 319		return rc != -EAGAIN ? rc : 0;
 320
 321	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 322
 323	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 324	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
 325	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
 326	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
 327	if (kvm_is_error_hva(vmaddr))
 328		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 329retry:
 330	unlocked = false;
 331	mmap_read_lock(current->mm);
 332	rc = reset_guest_reference_bit(current->mm, vmaddr);
 333	if (rc < 0) {
 334		rc = fixup_user_fault(current->mm, vmaddr,
 335				      FAULT_FLAG_WRITE, &unlocked);
 336		if (!rc) {
 337			mmap_read_unlock(current->mm);
 338			goto retry;
 339		}
 340	}
 341	mmap_read_unlock(current->mm);
 342	if (rc == -EFAULT)
 343		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 344	if (rc < 0)
 345		return rc;
 346	kvm_s390_set_psw_cc(vcpu, rc);
 347	return 0;
 348}
 349
 350#define SSKE_NQ 0x8
 351#define SSKE_MR 0x4
 352#define SSKE_MC 0x2
 353#define SSKE_MB 0x1
 354static int handle_sske(struct kvm_vcpu *vcpu)
 355{
 356	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
 357	unsigned long start, end;
 358	unsigned char key, oldkey;
 359	int reg1, reg2;
 360	bool unlocked;
 361	int rc;
 362
 363	vcpu->stat.instruction_sske++;
 364
 365	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 366		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 367
 368	rc = try_handle_skey(vcpu);
 369	if (rc)
 370		return rc != -EAGAIN ? rc : 0;
 371
 372	if (!test_kvm_facility(vcpu->kvm, 8))
 373		m3 &= ~SSKE_MB;
 374	if (!test_kvm_facility(vcpu->kvm, 10))
 375		m3 &= ~(SSKE_MC | SSKE_MR);
 376	if (!test_kvm_facility(vcpu->kvm, 14))
 377		m3 &= ~SSKE_NQ;
 378
 379	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 380
 381	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
 382	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 383	start = kvm_s390_logical_to_effective(vcpu, start);
 384	if (m3 & SSKE_MB) {
 385		/* start already designates an absolute address */
 386		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
 387	} else {
 388		start = kvm_s390_real_to_abs(vcpu, start);
 389		end = start + PAGE_SIZE;
 390	}
 391
 392	while (start != end) {
 393		unsigned long vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
 394		unlocked = false;
 395
 396		if (kvm_is_error_hva(vmaddr))
 397			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 398
 399		mmap_read_lock(current->mm);
 400		rc = cond_set_guest_storage_key(current->mm, vmaddr, key, &oldkey,
 401						m3 & SSKE_NQ, m3 & SSKE_MR,
 402						m3 & SSKE_MC);
 403
 404		if (rc < 0) {
 405			rc = fixup_user_fault(current->mm, vmaddr,
 406					      FAULT_FLAG_WRITE, &unlocked);
 407			rc = !rc ? -EAGAIN : rc;
 408		}
 409		mmap_read_unlock(current->mm);
 410		if (rc == -EFAULT)
 411			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 412		if (rc == -EAGAIN)
 413			continue;
 414		if (rc < 0)
 415			return rc;
 416		start += PAGE_SIZE;
 417	}
 418
 419	if (m3 & (SSKE_MC | SSKE_MR)) {
 420		if (m3 & SSKE_MB) {
 421			/* skey in reg1 is unpredictable */
 422			kvm_s390_set_psw_cc(vcpu, 3);
 423		} else {
 424			kvm_s390_set_psw_cc(vcpu, rc);
 425			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
 426			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
 427		}
 428	}
 429	if (m3 & SSKE_MB) {
 430		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
 431			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
 432		else
 433			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
 434		end = kvm_s390_logical_to_effective(vcpu, end);
 435		vcpu->run->s.regs.gprs[reg2] |= end;
 436	}
 437	return 0;
 438}
 439
 440static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
 441{
 442	vcpu->stat.instruction_ipte_interlock++;
 443	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
 444		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 445	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu->kvm));
 446	kvm_s390_retry_instr(vcpu);
 447	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
 448	return 0;
 449}
 450
 451static int handle_test_block(struct kvm_vcpu *vcpu)
 452{
 453	gpa_t addr;
 454	int reg2;
 455
 456	vcpu->stat.instruction_tb++;
 457
 458	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 459		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 460
 461	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
 462	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 463	addr = kvm_s390_logical_to_effective(vcpu, addr);
 464	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
 465		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 466	addr = kvm_s390_real_to_abs(vcpu, addr);
 467
 468	if (kvm_is_error_gpa(vcpu->kvm, addr))
 469		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 470	/*
 471	 * We don't expect errors on modern systems, and do not care
 472	 * about storage keys (yet), so let's just clear the page.
 473	 */
 474	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
 475		return -EFAULT;
 476	kvm_s390_set_psw_cc(vcpu, 0);
 477	vcpu->run->s.regs.gprs[0] = 0;
 478	return 0;
 479}
 480
 481static int handle_tpi(struct kvm_vcpu *vcpu)
 482{
 483	struct kvm_s390_interrupt_info *inti;
 484	unsigned long len;
 485	u32 tpi_data[3];
 486	int rc;
 487	u64 addr;
 488	u8 ar;
 489
 490	vcpu->stat.instruction_tpi++;
 491
 492	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 493	if (addr & 3)
 494		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 495
 496	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
 497	if (!inti) {
 498		kvm_s390_set_psw_cc(vcpu, 0);
 499		return 0;
 500	}
 501
 502	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
 503	tpi_data[1] = inti->io.io_int_parm;
 504	tpi_data[2] = inti->io.io_int_word;
 505	if (addr) {
 506		/*
 507		 * Store the two-word I/O interruption code into the
 508		 * provided area.
 509		 */
 510		len = sizeof(tpi_data) - 4;
 511		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
 512		if (rc) {
 513			rc = kvm_s390_inject_prog_cond(vcpu, rc);
 514			goto reinject_interrupt;
 515		}
 516	} else {
 517		/*
 518		 * Store the three-word I/O interruption code into
 519		 * the appropriate lowcore area.
 520		 */
 521		len = sizeof(tpi_data);
 522		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
 523			/* failed writes to the low core are not recoverable */
 524			rc = -EFAULT;
 525			goto reinject_interrupt;
 526		}
 527	}
 528
 529	/* irq was successfully handed to the guest */
 530	kfree(inti);
 531	kvm_s390_set_psw_cc(vcpu, 1);
 532	return 0;
 533reinject_interrupt:
 534	/*
 535	 * If we encounter a problem storing the interruption code, the
 536	 * instruction is suppressed from the guest's view: reinject the
 537	 * interrupt.
 538	 */
 539	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
 540		kfree(inti);
 541		rc = -EFAULT;
 542	}
 543	/* don't set the cc, a pgm irq was injected or we drop to user space */
 544	return rc ? -EFAULT : 0;
 545}
 546
 547static int handle_tsch(struct kvm_vcpu *vcpu)
 548{
 549	struct kvm_s390_interrupt_info *inti = NULL;
 550	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
 551
 552	vcpu->stat.instruction_tsch++;
 553
 554	/* a valid schid has at least one bit set */
 555	if (vcpu->run->s.regs.gprs[1])
 556		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
 557					   vcpu->run->s.regs.gprs[1]);
 558
 559	/*
 560	 * Prepare exit to userspace.
 561	 * We indicate whether we dequeued a pending I/O interrupt
 562	 * so that userspace can re-inject it if the instruction gets
 563	 * a program check. While this may re-order the pending I/O
 564	 * interrupts, this is no problem since the priority is kept
 565	 * intact.
 566	 */
 567	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
 568	vcpu->run->s390_tsch.dequeued = !!inti;
 569	if (inti) {
 570		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
 571		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
 572		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
 573		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
 574	}
 575	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
 576	kfree(inti);
 577	return -EREMOTE;
 578}
 579
 580static int handle_io_inst(struct kvm_vcpu *vcpu)
 581{
 582	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
 583
 584	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 585		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 586
 587	if (vcpu->kvm->arch.css_support) {
 588		/*
 589		 * Most I/O instructions will be handled by userspace.
 590		 * Exceptions are tpi and the interrupt portion of tsch.
 591		 */
 592		if (vcpu->arch.sie_block->ipa == 0xb236)
 593			return handle_tpi(vcpu);
 594		if (vcpu->arch.sie_block->ipa == 0xb235)
 595			return handle_tsch(vcpu);
 596		/* Handle in userspace. */
 597		vcpu->stat.instruction_io_other++;
 598		return -EOPNOTSUPP;
 599	} else {
 600		/*
 601		 * Set condition code 3 to stop the guest from issuing channel
 602		 * I/O instructions.
 603		 */
 604		kvm_s390_set_psw_cc(vcpu, 3);
 605		return 0;
 606	}
 607}
 608
 609/*
 610 * handle_pqap: Handling pqap interception
 611 * @vcpu: the vcpu having issue the pqap instruction
 612 *
 613 * We now support PQAP/AQIC instructions and we need to correctly
 614 * answer the guest even if no dedicated driver's hook is available.
 615 *
 616 * The intercepting code calls a dedicated callback for this instruction
 617 * if a driver did register one in the CRYPTO satellite of the
 618 * SIE block.
 619 *
 620 * If no callback is available, the queues are not available, return this
 621 * response code to the caller and set CC to 3.
 622 * Else return the response code returned by the callback.
 623 */
 624static int handle_pqap(struct kvm_vcpu *vcpu)
 625{
 626	struct ap_queue_status status = {};
 627	crypto_hook pqap_hook;
 628	unsigned long reg0;
 629	int ret;
 630	uint8_t fc;
 631
 632	/* Verify that the AP instruction are available */
 633	if (!ap_instructions_available())
 634		return -EOPNOTSUPP;
 635	/* Verify that the guest is allowed to use AP instructions */
 636	if (!(vcpu->arch.sie_block->eca & ECA_APIE))
 637		return -EOPNOTSUPP;
 638	/*
 639	 * The only possibly intercepted functions when AP instructions are
 640	 * available for the guest are AQIC and TAPQ with the t bit set
 641	 * since we do not set IC.3 (FIII) we currently will only intercept
 642	 * the AQIC function code.
 643	 * Note: running nested under z/VM can result in intercepts for other
 644	 * function codes, e.g. PQAP(QCI). We do not support this and bail out.
 645	 */
 646	reg0 = vcpu->run->s.regs.gprs[0];
 647	fc = (reg0 >> 24) & 0xff;
 648	if (fc != 0x03)
 649		return -EOPNOTSUPP;
 650
 651	/* PQAP instruction is allowed for guest kernel only */
 652	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 653		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 654
 655	/* Common PQAP instruction specification exceptions */
 656	/* bits 41-47 must all be zeros */
 657	if (reg0 & 0x007f0000UL)
 658		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 659	/* APFT not install and T bit set */
 660	if (!test_kvm_facility(vcpu->kvm, 15) && (reg0 & 0x00800000UL))
 661		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 662	/* APXA not installed and APID greater 64 or APQI greater 16 */
 663	if (!(vcpu->kvm->arch.crypto.crycbd & 0x02) && (reg0 & 0x0000c0f0UL))
 664		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 665
 666	/* AQIC function code specific exception */
 667	/* facility 65 not present for AQIC function code */
 668	if (!test_kvm_facility(vcpu->kvm, 65))
 669		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 670
 671	/*
 672	 * If the hook callback is registered, there will be a pointer to the
 673	 * hook function pointer in the kvm_s390_crypto structure. Lock the
 674	 * owner, retrieve the hook function pointer and call the hook.
 675	 */
 676	down_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 677	if (vcpu->kvm->arch.crypto.pqap_hook) {
 678		pqap_hook = *vcpu->kvm->arch.crypto.pqap_hook;
 679		ret = pqap_hook(vcpu);
 680		if (!ret && vcpu->run->s.regs.gprs[1] & 0x00ff0000)
 681			kvm_s390_set_psw_cc(vcpu, 3);
 
 
 
 
 682		up_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 683		return ret;
 684	}
 685	up_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 686	/*
 687	 * A vfio_driver must register a hook.
 688	 * No hook means no driver to enable the SIE CRYCB and no queues.
 689	 * We send this response to the guest.
 690	 */
 691	status.response_code = 0x01;
 692	memcpy(&vcpu->run->s.regs.gprs[1], &status, sizeof(status));
 693	kvm_s390_set_psw_cc(vcpu, 3);
 694	return 0;
 695}
 696
 697static int handle_stfl(struct kvm_vcpu *vcpu)
 698{
 699	int rc;
 700	unsigned int fac;
 701
 702	vcpu->stat.instruction_stfl++;
 703
 704	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 705		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 706
 707	/*
 708	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
 709	 * into a u32 memory representation. They will remain bits 0-31.
 710	 */
 711	fac = *vcpu->kvm->arch.model.fac_list >> 32;
 712	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
 713			    &fac, sizeof(fac));
 714	if (rc)
 715		return rc;
 716	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
 717	trace_kvm_s390_handle_stfl(vcpu, fac);
 718	return 0;
 719}
 720
 721#define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
 722#define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
 723#define PSW_ADDR_24 0x0000000000ffffffUL
 724#define PSW_ADDR_31 0x000000007fffffffUL
 725
 726int is_valid_psw(psw_t *psw)
 727{
 728	if (psw->mask & PSW_MASK_UNASSIGNED)
 729		return 0;
 730	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
 731		if (psw->addr & ~PSW_ADDR_31)
 732			return 0;
 733	}
 734	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
 735		return 0;
 736	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
 737		return 0;
 738	if (psw->addr & 1)
 739		return 0;
 740	return 1;
 741}
 742
 743int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
 744{
 745	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
 746	psw_compat_t new_psw;
 747	u64 addr;
 748	int rc;
 749	u8 ar;
 750
 751	vcpu->stat.instruction_lpsw++;
 752
 753	if (gpsw->mask & PSW_MASK_PSTATE)
 754		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 755
 756	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 757	if (addr & 7)
 758		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 759
 760	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 761	if (rc)
 762		return kvm_s390_inject_prog_cond(vcpu, rc);
 763	if (!(new_psw.mask & PSW32_MASK_BASE))
 764		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 765	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
 766	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
 767	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
 768	if (!is_valid_psw(gpsw))
 769		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 770	return 0;
 771}
 772
 773static int handle_lpswe(struct kvm_vcpu *vcpu)
 774{
 775	psw_t new_psw;
 776	u64 addr;
 777	int rc;
 778	u8 ar;
 779
 780	vcpu->stat.instruction_lpswe++;
 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	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 786	if (addr & 7)
 787		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 788	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 789	if (rc)
 790		return kvm_s390_inject_prog_cond(vcpu, rc);
 791	vcpu->arch.sie_block->gpsw = new_psw;
 792	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
 793		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 794	return 0;
 795}
 796
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 797static int handle_stidp(struct kvm_vcpu *vcpu)
 798{
 799	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
 800	u64 operand2;
 801	int rc;
 802	u8 ar;
 803
 804	vcpu->stat.instruction_stidp++;
 805
 806	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 807		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 808
 809	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 810
 811	if (operand2 & 7)
 812		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 813
 814	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
 815	if (rc)
 816		return kvm_s390_inject_prog_cond(vcpu, rc);
 817
 818	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
 819	return 0;
 820}
 821
 822static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
 823{
 824	int cpus = 0;
 825	int n;
 826
 827	cpus = atomic_read(&vcpu->kvm->online_vcpus);
 828
 829	/* deal with other level 3 hypervisors */
 830	if (stsi(mem, 3, 2, 2))
 831		mem->count = 0;
 832	if (mem->count < 8)
 833		mem->count++;
 834	for (n = mem->count - 1; n > 0 ; n--)
 835		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
 836
 837	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
 838	mem->vm[0].cpus_total = cpus;
 839	mem->vm[0].cpus_configured = cpus;
 840	mem->vm[0].cpus_standby = 0;
 841	mem->vm[0].cpus_reserved = 0;
 842	mem->vm[0].caf = 1000;
 843	memcpy(mem->vm[0].name, "KVMguest", 8);
 844	ASCEBC(mem->vm[0].name, 8);
 845	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
 846	ASCEBC(mem->vm[0].cpi, 16);
 847}
 848
 849static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
 850				 u8 fc, u8 sel1, u16 sel2)
 851{
 852	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
 853	vcpu->run->s390_stsi.addr = addr;
 854	vcpu->run->s390_stsi.ar = ar;
 855	vcpu->run->s390_stsi.fc = fc;
 856	vcpu->run->s390_stsi.sel1 = sel1;
 857	vcpu->run->s390_stsi.sel2 = sel2;
 858}
 859
 860static int handle_stsi(struct kvm_vcpu *vcpu)
 861{
 862	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
 863	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
 864	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
 865	unsigned long mem = 0;
 866	u64 operand2;
 867	int rc = 0;
 868	u8 ar;
 869
 870	vcpu->stat.instruction_stsi++;
 871	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
 872
 873	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 874		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 875
 876	/* Bailout forbidden function codes */
 877	if (fc > 3 && fc != 15)
 878		goto out_no_data;
 879
 880	/*
 881	 * fc 15 is provided only with
 882	 *   - PTF/CPU topology support through facility 15
 883	 *   - KVM_CAP_S390_USER_STSI
 884	 */
 885	if (fc == 15 && (!test_kvm_facility(vcpu->kvm, 11) ||
 886			 !vcpu->kvm->arch.user_stsi))
 887		goto out_no_data;
 888
 889	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
 890	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
 891		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 892
 893	if (fc == 0) {
 894		vcpu->run->s.regs.gprs[0] = 3 << 28;
 895		kvm_s390_set_psw_cc(vcpu, 0);
 896		return 0;
 897	}
 898
 899	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 900
 901	if (!kvm_s390_pv_cpu_is_protected(vcpu) && (operand2 & 0xfff))
 902		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 903
 904	switch (fc) {
 905	case 1: /* same handling for 1 and 2 */
 906	case 2:
 907		mem = get_zeroed_page(GFP_KERNEL_ACCOUNT);
 908		if (!mem)
 909			goto out_no_data;
 910		if (stsi((void *) mem, fc, sel1, sel2))
 911			goto out_no_data;
 912		break;
 913	case 3:
 914		if (sel1 != 2 || sel2 != 2)
 915			goto out_no_data;
 916		mem = get_zeroed_page(GFP_KERNEL_ACCOUNT);
 917		if (!mem)
 918			goto out_no_data;
 919		handle_stsi_3_2_2(vcpu, (void *) mem);
 920		break;
 921	case 15: /* fc 15 is fully handled in userspace */
 922		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
 923		trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 924		return -EREMOTE;
 925	}
 926	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
 927		memcpy(sida_addr(vcpu->arch.sie_block), (void *)mem, PAGE_SIZE);
 928		rc = 0;
 929	} else {
 930		rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
 931	}
 932	if (rc) {
 933		rc = kvm_s390_inject_prog_cond(vcpu, rc);
 934		goto out;
 935	}
 936	if (vcpu->kvm->arch.user_stsi) {
 937		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
 938		rc = -EREMOTE;
 939	}
 940	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 941	free_page(mem);
 942	kvm_s390_set_psw_cc(vcpu, 0);
 943	vcpu->run->s.regs.gprs[0] = 0;
 944	return rc;
 945out_no_data:
 946	kvm_s390_set_psw_cc(vcpu, 3);
 947out:
 948	free_page(mem);
 949	return rc;
 950}
 951
 952int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
 953{
 954	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
 955	case 0x02:
 956		return handle_stidp(vcpu);
 957	case 0x04:
 958		return handle_set_clock(vcpu);
 959	case 0x10:
 960		return handle_set_prefix(vcpu);
 961	case 0x11:
 962		return handle_store_prefix(vcpu);
 963	case 0x12:
 964		return handle_store_cpu_address(vcpu);
 965	case 0x14:
 966		return kvm_s390_handle_vsie(vcpu);
 967	case 0x21:
 968	case 0x50:
 969		return handle_ipte_interlock(vcpu);
 970	case 0x29:
 971		return handle_iske(vcpu);
 972	case 0x2a:
 973		return handle_rrbe(vcpu);
 974	case 0x2b:
 975		return handle_sske(vcpu);
 976	case 0x2c:
 977		return handle_test_block(vcpu);
 978	case 0x30:
 979	case 0x31:
 980	case 0x32:
 981	case 0x33:
 982	case 0x34:
 983	case 0x35:
 984	case 0x36:
 985	case 0x37:
 986	case 0x38:
 987	case 0x39:
 988	case 0x3a:
 989	case 0x3b:
 990	case 0x3c:
 991	case 0x5f:
 992	case 0x74:
 993	case 0x76:
 994		return handle_io_inst(vcpu);
 995	case 0x56:
 996		return handle_sthyi(vcpu);
 997	case 0x7d:
 998		return handle_stsi(vcpu);
 999	case 0xaf:
1000		return handle_pqap(vcpu);
1001	case 0xb1:
1002		return handle_stfl(vcpu);
1003	case 0xb2:
1004		return handle_lpswe(vcpu);
1005	default:
1006		return -EOPNOTSUPP;
1007	}
1008}
1009
1010static int handle_epsw(struct kvm_vcpu *vcpu)
1011{
1012	int reg1, reg2;
1013
1014	vcpu->stat.instruction_epsw++;
1015
1016	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
1017
1018	/* This basically extracts the mask half of the psw. */
1019	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
1020	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
1021	if (reg2) {
1022		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
1023		vcpu->run->s.regs.gprs[reg2] |=
1024			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
1025	}
1026	return 0;
1027}
1028
1029#define PFMF_RESERVED   0xfffc0101UL
1030#define PFMF_SK         0x00020000UL
1031#define PFMF_CF         0x00010000UL
1032#define PFMF_UI         0x00008000UL
1033#define PFMF_FSC        0x00007000UL
1034#define PFMF_NQ         0x00000800UL
1035#define PFMF_MR         0x00000400UL
1036#define PFMF_MC         0x00000200UL
1037#define PFMF_KEY        0x000000feUL
1038
1039static int handle_pfmf(struct kvm_vcpu *vcpu)
1040{
1041	bool mr = false, mc = false, nq;
1042	int reg1, reg2;
1043	unsigned long start, end;
1044	unsigned char key;
1045
1046	vcpu->stat.instruction_pfmf++;
1047
1048	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
1049
1050	if (!test_kvm_facility(vcpu->kvm, 8))
1051		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1052
1053	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1054		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1055
1056	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
1057		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1058
1059	/* Only provide non-quiescing support if enabled for the guest */
1060	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
1061	    !test_kvm_facility(vcpu->kvm, 14))
1062		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1063
1064	/* Only provide conditional-SSKE support if enabled for the guest */
1065	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
1066	    test_kvm_facility(vcpu->kvm, 10)) {
1067		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
1068		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
1069	}
1070
1071	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
1072	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
1073	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
1074	start = kvm_s390_logical_to_effective(vcpu, start);
1075
1076	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1077		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
1078			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
1079	}
1080
1081	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1082	case 0x00000000:
1083		/* only 4k frames specify a real address */
1084		start = kvm_s390_real_to_abs(vcpu, start);
1085		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
1086		break;
1087	case 0x00001000:
1088		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
1089		break;
1090	case 0x00002000:
1091		/* only support 2G frame size if EDAT2 is available and we are
1092		   not in 24-bit addressing mode */
1093		if (!test_kvm_facility(vcpu->kvm, 78) ||
1094		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
1095			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1096		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
1097		break;
1098	default:
1099		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1100	}
1101
1102	while (start != end) {
1103		unsigned long vmaddr;
1104		bool unlocked = false;
1105
1106		/* Translate guest address to host address */
1107		vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
1108		if (kvm_is_error_hva(vmaddr))
1109			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1110
1111		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1112			if (kvm_clear_guest(vcpu->kvm, start, PAGE_SIZE))
1113				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1114		}
1115
1116		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
1117			int rc = kvm_s390_skey_check_enable(vcpu);
1118
1119			if (rc)
1120				return rc;
1121			mmap_read_lock(current->mm);
1122			rc = cond_set_guest_storage_key(current->mm, vmaddr,
1123							key, NULL, nq, mr, mc);
1124			if (rc < 0) {
1125				rc = fixup_user_fault(current->mm, vmaddr,
1126						      FAULT_FLAG_WRITE, &unlocked);
1127				rc = !rc ? -EAGAIN : rc;
1128			}
1129			mmap_read_unlock(current->mm);
1130			if (rc == -EFAULT)
1131				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1132			if (rc == -EAGAIN)
1133				continue;
1134			if (rc < 0)
1135				return rc;
1136		}
1137		start += PAGE_SIZE;
1138	}
1139	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1140		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
1141			vcpu->run->s.regs.gprs[reg2] = end;
1142		} else {
1143			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
1144			end = kvm_s390_logical_to_effective(vcpu, end);
1145			vcpu->run->s.regs.gprs[reg2] |= end;
1146		}
1147	}
1148	return 0;
1149}
1150
1151/*
1152 * Must be called with relevant read locks held (kvm->mm->mmap_lock, kvm->srcu)
1153 */
1154static inline int __do_essa(struct kvm_vcpu *vcpu, const int orc)
1155{
1156	int r1, r2, nappended, entries;
1157	unsigned long gfn, hva, res, pgstev, ptev;
1158	unsigned long *cbrlo;
1159
1160	/*
1161	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
1162	 * machine check here we either handle it or crash
1163	 */
1164
1165	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
1166	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
1167	hva = gfn_to_hva(vcpu->kvm, gfn);
1168	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1169
1170	if (kvm_is_error_hva(hva))
1171		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1172
1173	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
1174	if (nappended < 0) {
1175		res = orc ? 0x10 : 0;
1176		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
1177		return 0;
1178	}
1179	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
1180	/*
1181	 * Set the block-content state part of the result. 0 means resident, so
1182	 * nothing to do if the page is valid. 2 is for preserved pages
1183	 * (non-present and non-zero), and 3 for zero pages (non-present and
1184	 * zero).
1185	 */
1186	if (ptev & _PAGE_INVALID) {
1187		res |= 2;
1188		if (pgstev & _PGSTE_GPS_ZERO)
1189			res |= 1;
1190	}
1191	if (pgstev & _PGSTE_GPS_NODAT)
1192		res |= 0x20;
1193	vcpu->run->s.regs.gprs[r1] = res;
1194	/*
1195	 * It is possible that all the normal 511 slots were full, in which case
1196	 * we will now write in the 512th slot, which is reserved for host use.
1197	 * In both cases we let the normal essa handling code process all the
1198	 * slots, including the reserved one, if needed.
1199	 */
1200	if (nappended > 0) {
1201		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1202		cbrlo[entries] = gfn << PAGE_SHIFT;
1203	}
1204
1205	if (orc) {
1206		struct kvm_memory_slot *ms = gfn_to_memslot(vcpu->kvm, gfn);
1207
1208		/* Increment only if we are really flipping the bit */
1209		if (ms && !test_and_set_bit(gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
1210			atomic64_inc(&vcpu->kvm->arch.cmma_dirty_pages);
1211	}
1212
1213	return nappended;
1214}
1215
1216static int handle_essa(struct kvm_vcpu *vcpu)
1217{
1218	/* entries expected to be 1FF */
1219	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1220	unsigned long *cbrlo;
1221	struct gmap *gmap;
1222	int i, orc;
1223
1224	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1225	gmap = vcpu->arch.gmap;
1226	vcpu->stat.instruction_essa++;
1227	if (!vcpu->kvm->arch.use_cmma)
1228		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1229
1230	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1231		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1232	/* Check for invalid operation request code */
1233	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1234	/* ORCs 0-6 are always valid */
1235	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1236						: ESSA_SET_STABLE_IF_RESIDENT))
1237		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1238
1239	if (!vcpu->kvm->arch.migration_mode) {
1240		/*
1241		 * CMMA is enabled in the KVM settings, but is disabled in
1242		 * the SIE block and in the mm_context, and we are not doing
1243		 * a migration. Enable CMMA in the mm_context.
1244		 * Since we need to take a write lock to write to the context
1245		 * to avoid races with storage keys handling, we check if the
1246		 * value really needs to be written to; if the value is
1247		 * already correct, we do nothing and avoid the lock.
1248		 */
1249		if (vcpu->kvm->mm->context.uses_cmm == 0) {
1250			mmap_write_lock(vcpu->kvm->mm);
1251			vcpu->kvm->mm->context.uses_cmm = 1;
1252			mmap_write_unlock(vcpu->kvm->mm);
1253		}
1254		/*
1255		 * If we are here, we are supposed to have CMMA enabled in
1256		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1257		 * while the context use_cmma flag is per process.
1258		 * It's possible that the context flag is enabled and the
1259		 * SIE flag is not, so we set the flag always; if it was
1260		 * already set, nothing changes, otherwise we enable it
1261		 * on this CPU too.
1262		 */
1263		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1264		/* Retry the ESSA instruction */
1265		kvm_s390_retry_instr(vcpu);
1266	} else {
1267		int srcu_idx;
1268
1269		mmap_read_lock(vcpu->kvm->mm);
1270		srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1271		i = __do_essa(vcpu, orc);
1272		srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
1273		mmap_read_unlock(vcpu->kvm->mm);
1274		if (i < 0)
1275			return i;
1276		/* Account for the possible extra cbrl entry */
1277		entries += i;
1278	}
1279	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1280	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1281	mmap_read_lock(gmap->mm);
1282	for (i = 0; i < entries; ++i)
1283		__gmap_zap(gmap, cbrlo[i]);
1284	mmap_read_unlock(gmap->mm);
1285	return 0;
1286}
1287
1288int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1289{
1290	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1291	case 0x8a:
1292	case 0x8e:
1293	case 0x8f:
1294		return handle_ipte_interlock(vcpu);
1295	case 0x8d:
1296		return handle_epsw(vcpu);
1297	case 0xab:
1298		return handle_essa(vcpu);
1299	case 0xaf:
1300		return handle_pfmf(vcpu);
1301	default:
1302		return -EOPNOTSUPP;
1303	}
1304}
1305
1306int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1307{
1308	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1309	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1310	int reg, rc, nr_regs;
1311	u32 ctl_array[16];
1312	u64 ga;
1313	u8 ar;
1314
1315	vcpu->stat.instruction_lctl++;
1316
1317	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1318		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1319
1320	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1321
1322	if (ga & 3)
1323		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1324
1325	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1326	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1327
1328	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1329	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1330	if (rc)
1331		return kvm_s390_inject_prog_cond(vcpu, rc);
1332	reg = reg1;
1333	nr_regs = 0;
1334	do {
1335		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1336		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1337		if (reg == reg3)
1338			break;
1339		reg = (reg + 1) % 16;
1340	} while (1);
1341	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1342	return 0;
1343}
1344
1345int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1346{
1347	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1348	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1349	int reg, rc, nr_regs;
1350	u32 ctl_array[16];
1351	u64 ga;
1352	u8 ar;
1353
1354	vcpu->stat.instruction_stctl++;
1355
1356	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1357		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1358
1359	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1360
1361	if (ga & 3)
1362		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1363
1364	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1365	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1366
1367	reg = reg1;
1368	nr_regs = 0;
1369	do {
1370		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1371		if (reg == reg3)
1372			break;
1373		reg = (reg + 1) % 16;
1374	} while (1);
1375	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1376	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1377}
1378
1379static int handle_lctlg(struct kvm_vcpu *vcpu)
1380{
1381	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1382	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1383	int reg, rc, nr_regs;
1384	u64 ctl_array[16];
1385	u64 ga;
1386	u8 ar;
1387
1388	vcpu->stat.instruction_lctlg++;
1389
1390	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1391		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1392
1393	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1394
1395	if (ga & 7)
1396		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1397
1398	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1399	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1400
1401	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1402	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1403	if (rc)
1404		return kvm_s390_inject_prog_cond(vcpu, rc);
1405	reg = reg1;
1406	nr_regs = 0;
1407	do {
1408		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1409		if (reg == reg3)
1410			break;
1411		reg = (reg + 1) % 16;
1412	} while (1);
1413	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1414	return 0;
1415}
1416
1417static int handle_stctg(struct kvm_vcpu *vcpu)
1418{
1419	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1420	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1421	int reg, rc, nr_regs;
1422	u64 ctl_array[16];
1423	u64 ga;
1424	u8 ar;
1425
1426	vcpu->stat.instruction_stctg++;
1427
1428	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1429		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1430
1431	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1432
1433	if (ga & 7)
1434		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1435
1436	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1437	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1438
1439	reg = reg1;
1440	nr_regs = 0;
1441	do {
1442		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1443		if (reg == reg3)
1444			break;
1445		reg = (reg + 1) % 16;
1446	} while (1);
1447	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1448	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1449}
1450
1451int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1452{
1453	switch (vcpu->arch.sie_block->ipb & 0x000000ff) {
1454	case 0x25:
1455		return handle_stctg(vcpu);
1456	case 0x2f:
1457		return handle_lctlg(vcpu);
1458	case 0x60:
1459	case 0x61:
1460	case 0x62:
1461		return handle_ri(vcpu);
 
 
1462	default:
1463		return -EOPNOTSUPP;
1464	}
1465}
1466
1467static int handle_tprot(struct kvm_vcpu *vcpu)
1468{
1469	u64 address, operand2;
1470	unsigned long gpa;
1471	u8 access_key;
1472	bool writable;
1473	int ret, cc;
1474	u8 ar;
1475
1476	vcpu->stat.instruction_tprot++;
1477
1478	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1479		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1480
1481	kvm_s390_get_base_disp_sse(vcpu, &address, &operand2, &ar, NULL);
1482	access_key = (operand2 & 0xf0) >> 4;
1483
1484	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1485		ipte_lock(vcpu->kvm);
1486
1487	ret = guest_translate_address_with_key(vcpu, address, ar, &gpa,
1488					       GACC_STORE, access_key);
1489	if (ret == 0) {
1490		gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1491	} else if (ret == PGM_PROTECTION) {
1492		writable = false;
1493		/* Write protected? Try again with read-only... */
1494		ret = guest_translate_address_with_key(vcpu, address, ar, &gpa,
1495						       GACC_FETCH, access_key);
1496	}
1497	if (ret >= 0) {
1498		cc = -1;
1499
1500		/* Fetching permitted; storing permitted */
1501		if (ret == 0 && writable)
1502			cc = 0;
1503		/* Fetching permitted; storing not permitted */
1504		else if (ret == 0 && !writable)
1505			cc = 1;
1506		/* Fetching not permitted; storing not permitted */
1507		else if (ret == PGM_PROTECTION)
1508			cc = 2;
1509		/* Translation not available */
1510		else if (ret != PGM_ADDRESSING && ret != PGM_TRANSLATION_SPEC)
1511			cc = 3;
1512
1513		if (cc != -1) {
1514			kvm_s390_set_psw_cc(vcpu, cc);
1515			ret = 0;
1516		} else {
1517			ret = kvm_s390_inject_program_int(vcpu, ret);
1518		}
1519	}
1520
1521	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1522		ipte_unlock(vcpu->kvm);
1523	return ret;
1524}
1525
1526int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1527{
1528	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1529	case 0x01:
1530		return handle_tprot(vcpu);
1531	default:
1532		return -EOPNOTSUPP;
1533	}
1534}
1535
1536static int handle_sckpf(struct kvm_vcpu *vcpu)
1537{
1538	u32 value;
1539
1540	vcpu->stat.instruction_sckpf++;
1541
1542	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1543		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1544
1545	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1546		return kvm_s390_inject_program_int(vcpu,
1547						   PGM_SPECIFICATION);
1548
1549	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1550	vcpu->arch.sie_block->todpr = value;
1551
1552	return 0;
1553}
1554
1555static int handle_ptff(struct kvm_vcpu *vcpu)
1556{
1557	vcpu->stat.instruction_ptff++;
1558
1559	/* we don't emulate any control instructions yet */
1560	kvm_s390_set_psw_cc(vcpu, 3);
1561	return 0;
1562}
1563
1564int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1565{
1566	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1567	case 0x04:
1568		return handle_ptff(vcpu);
1569	case 0x07:
1570		return handle_sckpf(vcpu);
1571	default:
1572		return -EOPNOTSUPP;
1573	}
1574}
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * handling privileged instructions
   4 *
   5 * Copyright IBM Corp. 2008, 2020
   6 *
   7 *    Author(s): Carsten Otte <cotte@de.ibm.com>
   8 *               Christian Borntraeger <borntraeger@de.ibm.com>
   9 */
  10
  11#include <linux/kvm.h>
  12#include <linux/gfp.h>
  13#include <linux/errno.h>
  14#include <linux/mm_types.h>
  15#include <linux/pgtable.h>
  16#include <linux/io.h>
  17#include <asm/asm-offsets.h>
  18#include <asm/facility.h>
  19#include <asm/current.h>
  20#include <asm/debug.h>
  21#include <asm/ebcdic.h>
  22#include <asm/sysinfo.h>
  23#include <asm/page-states.h>
  24#include <asm/gmap.h>
 
  25#include <asm/ptrace.h>
  26#include <asm/sclp.h>
  27#include <asm/ap.h>
  28#include "gaccess.h"
  29#include "kvm-s390.h"
  30#include "trace.h"
  31
  32static int handle_ri(struct kvm_vcpu *vcpu)
  33{
  34	vcpu->stat.instruction_ri++;
  35
  36	if (test_kvm_facility(vcpu->kvm, 64)) {
  37		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
  38		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
  39		kvm_s390_retry_instr(vcpu);
  40		return 0;
  41	} else
  42		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
  43}
  44
  45int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
  46{
  47	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
  48		return handle_ri(vcpu);
  49	else
  50		return -EOPNOTSUPP;
  51}
  52
  53static int handle_gs(struct kvm_vcpu *vcpu)
  54{
  55	vcpu->stat.instruction_gs++;
  56
  57	if (test_kvm_facility(vcpu->kvm, 133)) {
  58		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (lazy)");
  59		preempt_disable();
  60		local_ctl_set_bit(2, CR2_GUARDED_STORAGE_BIT);
  61		current->thread.gs_cb = (struct gs_cb *)&vcpu->run->s.regs.gscb;
  62		restore_gs_cb(current->thread.gs_cb);
  63		preempt_enable();
  64		vcpu->arch.sie_block->ecb |= ECB_GS;
  65		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
  66		vcpu->arch.gs_enabled = 1;
  67		kvm_s390_retry_instr(vcpu);
  68		return 0;
  69	} else
  70		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
  71}
  72
  73int kvm_s390_handle_e3(struct kvm_vcpu *vcpu)
  74{
  75	int code = vcpu->arch.sie_block->ipb & 0xff;
  76
  77	if (code == 0x49 || code == 0x4d)
  78		return handle_gs(vcpu);
  79	else
  80		return -EOPNOTSUPP;
  81}
  82/* Handle SCK (SET CLOCK) interception */
  83static int handle_set_clock(struct kvm_vcpu *vcpu)
  84{
  85	struct kvm_s390_vm_tod_clock gtod = { 0 };
  86	int rc;
  87	u8 ar;
  88	u64 op2;
  89
  90	vcpu->stat.instruction_sck++;
  91
  92	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  93		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
  94
  95	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
  96	if (op2 & 7)	/* Operand must be on a doubleword boundary */
  97		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
  98	rc = read_guest(vcpu, op2, ar, &gtod.tod, sizeof(gtod.tod));
  99	if (rc)
 100		return kvm_s390_inject_prog_cond(vcpu, rc);
 101
 102	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", gtod.tod);
 103	/*
 104	 * To set the TOD clock the kvm lock must be taken, but the vcpu lock
 105	 * is already held in handle_set_clock. The usual lock order is the
 106	 * opposite.  As SCK is deprecated and should not be used in several
 107	 * cases, for example when the multiple epoch facility or TOD clock
 108	 * steering facility is installed (see Principles of Operation),  a
 109	 * slow path can be used.  If the lock can not be taken via try_lock,
 110	 * the instruction will be retried via -EAGAIN at a later point in
 111	 * time.
 112	 */
 113	if (!kvm_s390_try_set_tod_clock(vcpu->kvm, &gtod)) {
 114		kvm_s390_retry_instr(vcpu);
 115		return -EAGAIN;
 116	}
 117
 118	kvm_s390_set_psw_cc(vcpu, 0);
 119	return 0;
 120}
 121
 122static int handle_set_prefix(struct kvm_vcpu *vcpu)
 123{
 124	u64 operand2;
 125	u32 address;
 126	int rc;
 127	u8 ar;
 128
 129	vcpu->stat.instruction_spx++;
 130
 131	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 132		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 133
 134	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 135
 136	/* must be word boundary */
 137	if (operand2 & 3)
 138		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 139
 140	/* get the value */
 141	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
 142	if (rc)
 143		return kvm_s390_inject_prog_cond(vcpu, rc);
 144
 145	address &= 0x7fffe000u;
 146
 147	/*
 148	 * Make sure the new value is valid memory. We only need to check the
 149	 * first page, since address is 8k aligned and memory pieces are always
 150	 * at least 1MB aligned and have at least a size of 1MB.
 151	 */
 152	if (!kvm_is_gpa_in_memslot(vcpu->kvm, address))
 153		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 154
 155	kvm_s390_set_prefix(vcpu, address);
 156	trace_kvm_s390_handle_prefix(vcpu, 1, address);
 157	return 0;
 158}
 159
 160static int handle_store_prefix(struct kvm_vcpu *vcpu)
 161{
 162	u64 operand2;
 163	u32 address;
 164	int rc;
 165	u8 ar;
 166
 167	vcpu->stat.instruction_stpx++;
 168
 169	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 170		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 171
 172	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 173
 174	/* must be word boundary */
 175	if (operand2 & 3)
 176		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 177
 178	address = kvm_s390_get_prefix(vcpu);
 179
 180	/* get the value */
 181	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
 182	if (rc)
 183		return kvm_s390_inject_prog_cond(vcpu, rc);
 184
 185	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
 186	trace_kvm_s390_handle_prefix(vcpu, 0, address);
 187	return 0;
 188}
 189
 190static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
 191{
 192	u16 vcpu_id = vcpu->vcpu_id;
 193	u64 ga;
 194	int rc;
 195	u8 ar;
 196
 197	vcpu->stat.instruction_stap++;
 198
 199	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 200		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 201
 202	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
 203
 204	if (ga & 1)
 205		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 206
 207	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
 208	if (rc)
 209		return kvm_s390_inject_prog_cond(vcpu, rc);
 210
 211	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
 212	trace_kvm_s390_handle_stap(vcpu, ga);
 213	return 0;
 214}
 215
 216int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
 217{
 218	int rc;
 219
 220	trace_kvm_s390_skey_related_inst(vcpu);
 221	/* Already enabled? */
 222	if (vcpu->arch.skey_enabled)
 223		return 0;
 224
 225	rc = s390_enable_skey();
 226	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
 227	if (rc)
 228		return rc;
 229
 230	if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
 231		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS);
 232	if (!vcpu->kvm->arch.use_skf)
 233		vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
 234	else
 235		vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
 236	vcpu->arch.skey_enabled = true;
 237	return 0;
 238}
 239
 240static int try_handle_skey(struct kvm_vcpu *vcpu)
 241{
 242	int rc;
 243
 244	rc = kvm_s390_skey_check_enable(vcpu);
 245	if (rc)
 246		return rc;
 247	if (vcpu->kvm->arch.use_skf) {
 248		/* with storage-key facility, SIE interprets it for us */
 249		kvm_s390_retry_instr(vcpu);
 250		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
 251		return -EAGAIN;
 252	}
 253	return 0;
 254}
 255
 256static int handle_iske(struct kvm_vcpu *vcpu)
 257{
 258	unsigned long gaddr, vmaddr;
 259	unsigned char key;
 260	int reg1, reg2;
 261	bool unlocked;
 262	int rc;
 263
 264	vcpu->stat.instruction_iske++;
 265
 266	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 267		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 268
 269	rc = try_handle_skey(vcpu);
 270	if (rc)
 271		return rc != -EAGAIN ? rc : 0;
 272
 273	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 274
 275	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 276	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
 277	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
 278	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
 279	if (kvm_is_error_hva(vmaddr))
 280		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 281retry:
 282	unlocked = false;
 283	mmap_read_lock(current->mm);
 284	rc = get_guest_storage_key(current->mm, vmaddr, &key);
 285
 286	if (rc) {
 287		rc = fixup_user_fault(current->mm, vmaddr,
 288				      FAULT_FLAG_WRITE, &unlocked);
 289		if (!rc) {
 290			mmap_read_unlock(current->mm);
 291			goto retry;
 292		}
 293	}
 294	mmap_read_unlock(current->mm);
 295	if (rc == -EFAULT)
 296		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 297	if (rc < 0)
 298		return rc;
 299	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
 300	vcpu->run->s.regs.gprs[reg1] |= key;
 301	return 0;
 302}
 303
 304static int handle_rrbe(struct kvm_vcpu *vcpu)
 305{
 306	unsigned long vmaddr, gaddr;
 307	int reg1, reg2;
 308	bool unlocked;
 309	int rc;
 310
 311	vcpu->stat.instruction_rrbe++;
 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	rc = try_handle_skey(vcpu);
 317	if (rc)
 318		return rc != -EAGAIN ? rc : 0;
 319
 320	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 321
 322	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 323	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
 324	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
 325	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
 326	if (kvm_is_error_hva(vmaddr))
 327		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 328retry:
 329	unlocked = false;
 330	mmap_read_lock(current->mm);
 331	rc = reset_guest_reference_bit(current->mm, vmaddr);
 332	if (rc < 0) {
 333		rc = fixup_user_fault(current->mm, vmaddr,
 334				      FAULT_FLAG_WRITE, &unlocked);
 335		if (!rc) {
 336			mmap_read_unlock(current->mm);
 337			goto retry;
 338		}
 339	}
 340	mmap_read_unlock(current->mm);
 341	if (rc == -EFAULT)
 342		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 343	if (rc < 0)
 344		return rc;
 345	kvm_s390_set_psw_cc(vcpu, rc);
 346	return 0;
 347}
 348
 349#define SSKE_NQ 0x8
 350#define SSKE_MR 0x4
 351#define SSKE_MC 0x2
 352#define SSKE_MB 0x1
 353static int handle_sske(struct kvm_vcpu *vcpu)
 354{
 355	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
 356	unsigned long start, end;
 357	unsigned char key, oldkey;
 358	int reg1, reg2;
 359	bool unlocked;
 360	int rc;
 361
 362	vcpu->stat.instruction_sske++;
 363
 364	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 365		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 366
 367	rc = try_handle_skey(vcpu);
 368	if (rc)
 369		return rc != -EAGAIN ? rc : 0;
 370
 371	if (!test_kvm_facility(vcpu->kvm, 8))
 372		m3 &= ~SSKE_MB;
 373	if (!test_kvm_facility(vcpu->kvm, 10))
 374		m3 &= ~(SSKE_MC | SSKE_MR);
 375	if (!test_kvm_facility(vcpu->kvm, 14))
 376		m3 &= ~SSKE_NQ;
 377
 378	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
 379
 380	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
 381	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 382	start = kvm_s390_logical_to_effective(vcpu, start);
 383	if (m3 & SSKE_MB) {
 384		/* start already designates an absolute address */
 385		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
 386	} else {
 387		start = kvm_s390_real_to_abs(vcpu, start);
 388		end = start + PAGE_SIZE;
 389	}
 390
 391	while (start != end) {
 392		unsigned long vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
 393		unlocked = false;
 394
 395		if (kvm_is_error_hva(vmaddr))
 396			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 397
 398		mmap_read_lock(current->mm);
 399		rc = cond_set_guest_storage_key(current->mm, vmaddr, key, &oldkey,
 400						m3 & SSKE_NQ, m3 & SSKE_MR,
 401						m3 & SSKE_MC);
 402
 403		if (rc < 0) {
 404			rc = fixup_user_fault(current->mm, vmaddr,
 405					      FAULT_FLAG_WRITE, &unlocked);
 406			rc = !rc ? -EAGAIN : rc;
 407		}
 408		mmap_read_unlock(current->mm);
 409		if (rc == -EFAULT)
 410			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 411		if (rc == -EAGAIN)
 412			continue;
 413		if (rc < 0)
 414			return rc;
 415		start += PAGE_SIZE;
 416	}
 417
 418	if (m3 & (SSKE_MC | SSKE_MR)) {
 419		if (m3 & SSKE_MB) {
 420			/* skey in reg1 is unpredictable */
 421			kvm_s390_set_psw_cc(vcpu, 3);
 422		} else {
 423			kvm_s390_set_psw_cc(vcpu, rc);
 424			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
 425			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
 426		}
 427	}
 428	if (m3 & SSKE_MB) {
 429		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
 430			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
 431		else
 432			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
 433		end = kvm_s390_logical_to_effective(vcpu, end);
 434		vcpu->run->s.regs.gprs[reg2] |= end;
 435	}
 436	return 0;
 437}
 438
 439static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
 440{
 441	vcpu->stat.instruction_ipte_interlock++;
 442	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
 443		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 444	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu->kvm));
 445	kvm_s390_retry_instr(vcpu);
 446	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
 447	return 0;
 448}
 449
 450static int handle_test_block(struct kvm_vcpu *vcpu)
 451{
 452	gpa_t addr;
 453	int reg2;
 454
 455	vcpu->stat.instruction_tb++;
 456
 457	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 458		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 459
 460	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
 461	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
 462	addr = kvm_s390_logical_to_effective(vcpu, addr);
 463	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
 464		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
 465	addr = kvm_s390_real_to_abs(vcpu, addr);
 466
 467	if (!kvm_is_gpa_in_memslot(vcpu->kvm, addr))
 468		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
 469	/*
 470	 * We don't expect errors on modern systems, and do not care
 471	 * about storage keys (yet), so let's just clear the page.
 472	 */
 473	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
 474		return -EFAULT;
 475	kvm_s390_set_psw_cc(vcpu, 0);
 476	vcpu->run->s.regs.gprs[0] = 0;
 477	return 0;
 478}
 479
 480static int handle_tpi(struct kvm_vcpu *vcpu)
 481{
 482	struct kvm_s390_interrupt_info *inti;
 483	unsigned long len;
 484	u32 tpi_data[3];
 485	int rc;
 486	u64 addr;
 487	u8 ar;
 488
 489	vcpu->stat.instruction_tpi++;
 490
 491	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 492	if (addr & 3)
 493		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 494
 495	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
 496	if (!inti) {
 497		kvm_s390_set_psw_cc(vcpu, 0);
 498		return 0;
 499	}
 500
 501	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
 502	tpi_data[1] = inti->io.io_int_parm;
 503	tpi_data[2] = inti->io.io_int_word;
 504	if (addr) {
 505		/*
 506		 * Store the two-word I/O interruption code into the
 507		 * provided area.
 508		 */
 509		len = sizeof(tpi_data) - 4;
 510		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
 511		if (rc) {
 512			rc = kvm_s390_inject_prog_cond(vcpu, rc);
 513			goto reinject_interrupt;
 514		}
 515	} else {
 516		/*
 517		 * Store the three-word I/O interruption code into
 518		 * the appropriate lowcore area.
 519		 */
 520		len = sizeof(tpi_data);
 521		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
 522			/* failed writes to the low core are not recoverable */
 523			rc = -EFAULT;
 524			goto reinject_interrupt;
 525		}
 526	}
 527
 528	/* irq was successfully handed to the guest */
 529	kfree(inti);
 530	kvm_s390_set_psw_cc(vcpu, 1);
 531	return 0;
 532reinject_interrupt:
 533	/*
 534	 * If we encounter a problem storing the interruption code, the
 535	 * instruction is suppressed from the guest's view: reinject the
 536	 * interrupt.
 537	 */
 538	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
 539		kfree(inti);
 540		rc = -EFAULT;
 541	}
 542	/* don't set the cc, a pgm irq was injected or we drop to user space */
 543	return rc ? -EFAULT : 0;
 544}
 545
 546static int handle_tsch(struct kvm_vcpu *vcpu)
 547{
 548	struct kvm_s390_interrupt_info *inti = NULL;
 549	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
 550
 551	vcpu->stat.instruction_tsch++;
 552
 553	/* a valid schid has at least one bit set */
 554	if (vcpu->run->s.regs.gprs[1])
 555		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
 556					   vcpu->run->s.regs.gprs[1]);
 557
 558	/*
 559	 * Prepare exit to userspace.
 560	 * We indicate whether we dequeued a pending I/O interrupt
 561	 * so that userspace can re-inject it if the instruction gets
 562	 * a program check. While this may re-order the pending I/O
 563	 * interrupts, this is no problem since the priority is kept
 564	 * intact.
 565	 */
 566	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
 567	vcpu->run->s390_tsch.dequeued = !!inti;
 568	if (inti) {
 569		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
 570		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
 571		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
 572		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
 573	}
 574	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
 575	kfree(inti);
 576	return -EREMOTE;
 577}
 578
 579static int handle_io_inst(struct kvm_vcpu *vcpu)
 580{
 581	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
 582
 583	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 584		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 585
 586	if (vcpu->kvm->arch.css_support) {
 587		/*
 588		 * Most I/O instructions will be handled by userspace.
 589		 * Exceptions are tpi and the interrupt portion of tsch.
 590		 */
 591		if (vcpu->arch.sie_block->ipa == 0xb236)
 592			return handle_tpi(vcpu);
 593		if (vcpu->arch.sie_block->ipa == 0xb235)
 594			return handle_tsch(vcpu);
 595		/* Handle in userspace. */
 596		vcpu->stat.instruction_io_other++;
 597		return -EOPNOTSUPP;
 598	} else {
 599		/*
 600		 * Set condition code 3 to stop the guest from issuing channel
 601		 * I/O instructions.
 602		 */
 603		kvm_s390_set_psw_cc(vcpu, 3);
 604		return 0;
 605	}
 606}
 607
 608/*
 609 * handle_pqap: Handling pqap interception
 610 * @vcpu: the vcpu having issue the pqap instruction
 611 *
 612 * We now support PQAP/AQIC instructions and we need to correctly
 613 * answer the guest even if no dedicated driver's hook is available.
 614 *
 615 * The intercepting code calls a dedicated callback for this instruction
 616 * if a driver did register one in the CRYPTO satellite of the
 617 * SIE block.
 618 *
 619 * If no callback is available, the queues are not available, return this
 620 * response code to the caller and set CC to 3.
 621 * Else return the response code returned by the callback.
 622 */
 623static int handle_pqap(struct kvm_vcpu *vcpu)
 624{
 625	struct ap_queue_status status = {};
 626	crypto_hook pqap_hook;
 627	unsigned long reg0;
 628	int ret;
 629	uint8_t fc;
 630
 631	/* Verify that the AP instruction are available */
 632	if (!ap_instructions_available())
 633		return -EOPNOTSUPP;
 634	/* Verify that the guest is allowed to use AP instructions */
 635	if (!(vcpu->arch.sie_block->eca & ECA_APIE))
 636		return -EOPNOTSUPP;
 637	/*
 638	 * The only possibly intercepted functions when AP instructions are
 639	 * available for the guest are AQIC and TAPQ with the t bit set
 640	 * since we do not set IC.3 (FIII) we currently will only intercept
 641	 * the AQIC function code.
 642	 * Note: running nested under z/VM can result in intercepts for other
 643	 * function codes, e.g. PQAP(QCI). We do not support this and bail out.
 644	 */
 645	reg0 = vcpu->run->s.regs.gprs[0];
 646	fc = (reg0 >> 24) & 0xff;
 647	if (fc != 0x03)
 648		return -EOPNOTSUPP;
 649
 650	/* PQAP instruction is allowed for guest kernel only */
 651	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 652		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 653
 654	/* Common PQAP instruction specification exceptions */
 655	/* bits 41-47 must all be zeros */
 656	if (reg0 & 0x007f0000UL)
 657		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 658	/* APFT not install and T bit set */
 659	if (!test_kvm_facility(vcpu->kvm, 15) && (reg0 & 0x00800000UL))
 660		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 661	/* APXA not installed and APID greater 64 or APQI greater 16 */
 662	if (!(vcpu->kvm->arch.crypto.crycbd & 0x02) && (reg0 & 0x0000c0f0UL))
 663		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 664
 665	/* AQIC function code specific exception */
 666	/* facility 65 not present for AQIC function code */
 667	if (!test_kvm_facility(vcpu->kvm, 65))
 668		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 669
 670	/*
 671	 * If the hook callback is registered, there will be a pointer to the
 672	 * hook function pointer in the kvm_s390_crypto structure. Lock the
 673	 * owner, retrieve the hook function pointer and call the hook.
 674	 */
 675	down_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 676	if (vcpu->kvm->arch.crypto.pqap_hook) {
 677		pqap_hook = *vcpu->kvm->arch.crypto.pqap_hook;
 678		ret = pqap_hook(vcpu);
 679		if (!ret) {
 680			if (vcpu->run->s.regs.gprs[1] & 0x00ff0000)
 681				kvm_s390_set_psw_cc(vcpu, 3);
 682			else
 683				kvm_s390_set_psw_cc(vcpu, 0);
 684		}
 685		up_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 686		return ret;
 687	}
 688	up_read(&vcpu->kvm->arch.crypto.pqap_hook_rwsem);
 689	/*
 690	 * A vfio_driver must register a hook.
 691	 * No hook means no driver to enable the SIE CRYCB and no queues.
 692	 * We send this response to the guest.
 693	 */
 694	status.response_code = 0x01;
 695	memcpy(&vcpu->run->s.regs.gprs[1], &status, sizeof(status));
 696	kvm_s390_set_psw_cc(vcpu, 3);
 697	return 0;
 698}
 699
 700static int handle_stfl(struct kvm_vcpu *vcpu)
 701{
 702	int rc;
 703	unsigned int fac;
 704
 705	vcpu->stat.instruction_stfl++;
 706
 707	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 708		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 709
 710	/*
 711	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
 712	 * into a u32 memory representation. They will remain bits 0-31.
 713	 */
 714	fac = *vcpu->kvm->arch.model.fac_list >> 32;
 715	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
 716			    &fac, sizeof(fac));
 717	if (rc)
 718		return rc;
 719	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
 720	trace_kvm_s390_handle_stfl(vcpu, fac);
 721	return 0;
 722}
 723
 724#define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
 725#define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
 726#define PSW_ADDR_24 0x0000000000ffffffUL
 727#define PSW_ADDR_31 0x000000007fffffffUL
 728
 729int is_valid_psw(psw_t *psw)
 730{
 731	if (psw->mask & PSW_MASK_UNASSIGNED)
 732		return 0;
 733	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
 734		if (psw->addr & ~PSW_ADDR_31)
 735			return 0;
 736	}
 737	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
 738		return 0;
 739	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
 740		return 0;
 741	if (psw->addr & 1)
 742		return 0;
 743	return 1;
 744}
 745
 746int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
 747{
 748	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
 749	psw_compat_t new_psw;
 750	u64 addr;
 751	int rc;
 752	u8 ar;
 753
 754	vcpu->stat.instruction_lpsw++;
 755
 756	if (gpsw->mask & PSW_MASK_PSTATE)
 757		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 758
 759	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 760	if (addr & 7)
 761		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 762
 763	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 764	if (rc)
 765		return kvm_s390_inject_prog_cond(vcpu, rc);
 766	if (!(new_psw.mask & PSW32_MASK_BASE))
 767		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 768	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
 769	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
 770	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
 771	if (!is_valid_psw(gpsw))
 772		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 773	return 0;
 774}
 775
 776static int handle_lpswe(struct kvm_vcpu *vcpu)
 777{
 778	psw_t new_psw;
 779	u64 addr;
 780	int rc;
 781	u8 ar;
 782
 783	vcpu->stat.instruction_lpswe++;
 784
 785	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 786		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 787
 788	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
 789	if (addr & 7)
 790		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 791	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 792	if (rc)
 793		return kvm_s390_inject_prog_cond(vcpu, rc);
 794	vcpu->arch.sie_block->gpsw = new_psw;
 795	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
 796		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 797	return 0;
 798}
 799
 800static int handle_lpswey(struct kvm_vcpu *vcpu)
 801{
 802	psw_t new_psw;
 803	u64 addr;
 804	int rc;
 805	u8 ar;
 806
 807	vcpu->stat.instruction_lpswey++;
 808
 809	if (!test_kvm_facility(vcpu->kvm, 193))
 810		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
 811
 812	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 813		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 814
 815	addr = kvm_s390_get_base_disp_siy(vcpu, &ar);
 816	if (addr & 7)
 817		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 818
 819	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
 820	if (rc)
 821		return kvm_s390_inject_prog_cond(vcpu, rc);
 822
 823	vcpu->arch.sie_block->gpsw = new_psw;
 824	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
 825		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 826
 827	return 0;
 828}
 829
 830static int handle_stidp(struct kvm_vcpu *vcpu)
 831{
 832	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
 833	u64 operand2;
 834	int rc;
 835	u8 ar;
 836
 837	vcpu->stat.instruction_stidp++;
 838
 839	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 840		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 841
 842	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 843
 844	if (operand2 & 7)
 845		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 846
 847	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
 848	if (rc)
 849		return kvm_s390_inject_prog_cond(vcpu, rc);
 850
 851	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
 852	return 0;
 853}
 854
 855static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
 856{
 857	int cpus = 0;
 858	int n;
 859
 860	cpus = atomic_read(&vcpu->kvm->online_vcpus);
 861
 862	/* deal with other level 3 hypervisors */
 863	if (stsi(mem, 3, 2, 2))
 864		mem->count = 0;
 865	if (mem->count < 8)
 866		mem->count++;
 867	for (n = mem->count - 1; n > 0 ; n--)
 868		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
 869
 870	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
 871	mem->vm[0].cpus_total = cpus;
 872	mem->vm[0].cpus_configured = cpus;
 873	mem->vm[0].cpus_standby = 0;
 874	mem->vm[0].cpus_reserved = 0;
 875	mem->vm[0].caf = 1000;
 876	memcpy(mem->vm[0].name, "KVMguest", 8);
 877	ASCEBC(mem->vm[0].name, 8);
 878	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
 879	ASCEBC(mem->vm[0].cpi, 16);
 880}
 881
 882static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
 883				 u8 fc, u8 sel1, u16 sel2)
 884{
 885	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
 886	vcpu->run->s390_stsi.addr = addr;
 887	vcpu->run->s390_stsi.ar = ar;
 888	vcpu->run->s390_stsi.fc = fc;
 889	vcpu->run->s390_stsi.sel1 = sel1;
 890	vcpu->run->s390_stsi.sel2 = sel2;
 891}
 892
 893static int handle_stsi(struct kvm_vcpu *vcpu)
 894{
 895	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
 896	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
 897	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
 898	unsigned long mem = 0;
 899	u64 operand2;
 900	int rc = 0;
 901	u8 ar;
 902
 903	vcpu->stat.instruction_stsi++;
 904	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
 905
 906	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 907		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 908
 909	/* Bailout forbidden function codes */
 910	if (fc > 3 && fc != 15)
 911		goto out_no_data;
 912
 913	/*
 914	 * fc 15 is provided only with
 915	 *   - PTF/CPU topology support through facility 15
 916	 *   - KVM_CAP_S390_USER_STSI
 917	 */
 918	if (fc == 15 && (!test_kvm_facility(vcpu->kvm, 11) ||
 919			 !vcpu->kvm->arch.user_stsi))
 920		goto out_no_data;
 921
 922	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
 923	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
 924		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 925
 926	if (fc == 0) {
 927		vcpu->run->s.regs.gprs[0] = 3 << 28;
 928		kvm_s390_set_psw_cc(vcpu, 0);
 929		return 0;
 930	}
 931
 932	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
 933
 934	if (!kvm_s390_pv_cpu_is_protected(vcpu) && (operand2 & 0xfff))
 935		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
 936
 937	switch (fc) {
 938	case 1: /* same handling for 1 and 2 */
 939	case 2:
 940		mem = get_zeroed_page(GFP_KERNEL_ACCOUNT);
 941		if (!mem)
 942			goto out_no_data;
 943		if (stsi((void *) mem, fc, sel1, sel2))
 944			goto out_no_data;
 945		break;
 946	case 3:
 947		if (sel1 != 2 || sel2 != 2)
 948			goto out_no_data;
 949		mem = get_zeroed_page(GFP_KERNEL_ACCOUNT);
 950		if (!mem)
 951			goto out_no_data;
 952		handle_stsi_3_2_2(vcpu, (void *) mem);
 953		break;
 954	case 15: /* fc 15 is fully handled in userspace */
 955		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
 956		trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 957		return -EREMOTE;
 958	}
 959	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
 960		memcpy(sida_addr(vcpu->arch.sie_block), (void *)mem, PAGE_SIZE);
 961		rc = 0;
 962	} else {
 963		rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
 964	}
 965	if (rc) {
 966		rc = kvm_s390_inject_prog_cond(vcpu, rc);
 967		goto out;
 968	}
 969	if (vcpu->kvm->arch.user_stsi) {
 970		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
 971		rc = -EREMOTE;
 972	}
 973	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
 974	free_page(mem);
 975	kvm_s390_set_psw_cc(vcpu, 0);
 976	vcpu->run->s.regs.gprs[0] = 0;
 977	return rc;
 978out_no_data:
 979	kvm_s390_set_psw_cc(vcpu, 3);
 980out:
 981	free_page(mem);
 982	return rc;
 983}
 984
 985int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
 986{
 987	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
 988	case 0x02:
 989		return handle_stidp(vcpu);
 990	case 0x04:
 991		return handle_set_clock(vcpu);
 992	case 0x10:
 993		return handle_set_prefix(vcpu);
 994	case 0x11:
 995		return handle_store_prefix(vcpu);
 996	case 0x12:
 997		return handle_store_cpu_address(vcpu);
 998	case 0x14:
 999		return kvm_s390_handle_vsie(vcpu);
1000	case 0x21:
1001	case 0x50:
1002		return handle_ipte_interlock(vcpu);
1003	case 0x29:
1004		return handle_iske(vcpu);
1005	case 0x2a:
1006		return handle_rrbe(vcpu);
1007	case 0x2b:
1008		return handle_sske(vcpu);
1009	case 0x2c:
1010		return handle_test_block(vcpu);
1011	case 0x30:
1012	case 0x31:
1013	case 0x32:
1014	case 0x33:
1015	case 0x34:
1016	case 0x35:
1017	case 0x36:
1018	case 0x37:
1019	case 0x38:
1020	case 0x39:
1021	case 0x3a:
1022	case 0x3b:
1023	case 0x3c:
1024	case 0x5f:
1025	case 0x74:
1026	case 0x76:
1027		return handle_io_inst(vcpu);
1028	case 0x56:
1029		return handle_sthyi(vcpu);
1030	case 0x7d:
1031		return handle_stsi(vcpu);
1032	case 0xaf:
1033		return handle_pqap(vcpu);
1034	case 0xb1:
1035		return handle_stfl(vcpu);
1036	case 0xb2:
1037		return handle_lpswe(vcpu);
1038	default:
1039		return -EOPNOTSUPP;
1040	}
1041}
1042
1043static int handle_epsw(struct kvm_vcpu *vcpu)
1044{
1045	int reg1, reg2;
1046
1047	vcpu->stat.instruction_epsw++;
1048
1049	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
1050
1051	/* This basically extracts the mask half of the psw. */
1052	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
1053	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
1054	if (reg2) {
1055		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
1056		vcpu->run->s.regs.gprs[reg2] |=
1057			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
1058	}
1059	return 0;
1060}
1061
1062#define PFMF_RESERVED   0xfffc0101UL
1063#define PFMF_SK         0x00020000UL
1064#define PFMF_CF         0x00010000UL
1065#define PFMF_UI         0x00008000UL
1066#define PFMF_FSC        0x00007000UL
1067#define PFMF_NQ         0x00000800UL
1068#define PFMF_MR         0x00000400UL
1069#define PFMF_MC         0x00000200UL
1070#define PFMF_KEY        0x000000feUL
1071
1072static int handle_pfmf(struct kvm_vcpu *vcpu)
1073{
1074	bool mr = false, mc = false, nq;
1075	int reg1, reg2;
1076	unsigned long start, end;
1077	unsigned char key;
1078
1079	vcpu->stat.instruction_pfmf++;
1080
1081	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
1082
1083	if (!test_kvm_facility(vcpu->kvm, 8))
1084		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1085
1086	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1087		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1088
1089	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
1090		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1091
1092	/* Only provide non-quiescing support if enabled for the guest */
1093	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
1094	    !test_kvm_facility(vcpu->kvm, 14))
1095		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1096
1097	/* Only provide conditional-SSKE support if enabled for the guest */
1098	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
1099	    test_kvm_facility(vcpu->kvm, 10)) {
1100		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
1101		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
1102	}
1103
1104	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
1105	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
1106	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
1107	start = kvm_s390_logical_to_effective(vcpu, start);
1108
1109	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1110		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
1111			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
1112	}
1113
1114	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1115	case 0x00000000:
1116		/* only 4k frames specify a real address */
1117		start = kvm_s390_real_to_abs(vcpu, start);
1118		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
1119		break;
1120	case 0x00001000:
1121		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
1122		break;
1123	case 0x00002000:
1124		/* only support 2G frame size if EDAT2 is available and we are
1125		   not in 24-bit addressing mode */
1126		if (!test_kvm_facility(vcpu->kvm, 78) ||
1127		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
1128			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1129		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
1130		break;
1131	default:
1132		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1133	}
1134
1135	while (start != end) {
1136		unsigned long vmaddr;
1137		bool unlocked = false;
1138
1139		/* Translate guest address to host address */
1140		vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
1141		if (kvm_is_error_hva(vmaddr))
1142			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1143
1144		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1145			if (kvm_clear_guest(vcpu->kvm, start, PAGE_SIZE))
1146				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1147		}
1148
1149		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
1150			int rc = kvm_s390_skey_check_enable(vcpu);
1151
1152			if (rc)
1153				return rc;
1154			mmap_read_lock(current->mm);
1155			rc = cond_set_guest_storage_key(current->mm, vmaddr,
1156							key, NULL, nq, mr, mc);
1157			if (rc < 0) {
1158				rc = fixup_user_fault(current->mm, vmaddr,
1159						      FAULT_FLAG_WRITE, &unlocked);
1160				rc = !rc ? -EAGAIN : rc;
1161			}
1162			mmap_read_unlock(current->mm);
1163			if (rc == -EFAULT)
1164				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1165			if (rc == -EAGAIN)
1166				continue;
1167			if (rc < 0)
1168				return rc;
1169		}
1170		start += PAGE_SIZE;
1171	}
1172	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1173		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
1174			vcpu->run->s.regs.gprs[reg2] = end;
1175		} else {
1176			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
1177			end = kvm_s390_logical_to_effective(vcpu, end);
1178			vcpu->run->s.regs.gprs[reg2] |= end;
1179		}
1180	}
1181	return 0;
1182}
1183
1184/*
1185 * Must be called with relevant read locks held (kvm->mm->mmap_lock, kvm->srcu)
1186 */
1187static inline int __do_essa(struct kvm_vcpu *vcpu, const int orc)
1188{
1189	int r1, r2, nappended, entries;
1190	unsigned long gfn, hva, res, pgstev, ptev;
1191	unsigned long *cbrlo;
1192
1193	/*
1194	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
1195	 * machine check here we either handle it or crash
1196	 */
1197
1198	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
1199	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
1200	hva = gfn_to_hva(vcpu->kvm, gfn);
1201	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1202
1203	if (kvm_is_error_hva(hva))
1204		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1205
1206	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
1207	if (nappended < 0) {
1208		res = orc ? 0x10 : 0;
1209		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
1210		return 0;
1211	}
1212	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
1213	/*
1214	 * Set the block-content state part of the result. 0 means resident, so
1215	 * nothing to do if the page is valid. 2 is for preserved pages
1216	 * (non-present and non-zero), and 3 for zero pages (non-present and
1217	 * zero).
1218	 */
1219	if (ptev & _PAGE_INVALID) {
1220		res |= 2;
1221		if (pgstev & _PGSTE_GPS_ZERO)
1222			res |= 1;
1223	}
1224	if (pgstev & _PGSTE_GPS_NODAT)
1225		res |= 0x20;
1226	vcpu->run->s.regs.gprs[r1] = res;
1227	/*
1228	 * It is possible that all the normal 511 slots were full, in which case
1229	 * we will now write in the 512th slot, which is reserved for host use.
1230	 * In both cases we let the normal essa handling code process all the
1231	 * slots, including the reserved one, if needed.
1232	 */
1233	if (nappended > 0) {
1234		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1235		cbrlo[entries] = gfn << PAGE_SHIFT;
1236	}
1237
1238	if (orc) {
1239		struct kvm_memory_slot *ms = gfn_to_memslot(vcpu->kvm, gfn);
1240
1241		/* Increment only if we are really flipping the bit */
1242		if (ms && !test_and_set_bit(gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
1243			atomic64_inc(&vcpu->kvm->arch.cmma_dirty_pages);
1244	}
1245
1246	return nappended;
1247}
1248
1249static int handle_essa(struct kvm_vcpu *vcpu)
1250{
1251	/* entries expected to be 1FF */
1252	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1253	unsigned long *cbrlo;
1254	struct gmap *gmap;
1255	int i, orc;
1256
1257	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1258	gmap = vcpu->arch.gmap;
1259	vcpu->stat.instruction_essa++;
1260	if (!vcpu->kvm->arch.use_cmma)
1261		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1262
1263	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1264		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1265	/* Check for invalid operation request code */
1266	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1267	/* ORCs 0-6 are always valid */
1268	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1269						: ESSA_SET_STABLE_IF_RESIDENT))
1270		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1271
1272	if (!vcpu->kvm->arch.migration_mode) {
1273		/*
1274		 * CMMA is enabled in the KVM settings, but is disabled in
1275		 * the SIE block and in the mm_context, and we are not doing
1276		 * a migration. Enable CMMA in the mm_context.
1277		 * Since we need to take a write lock to write to the context
1278		 * to avoid races with storage keys handling, we check if the
1279		 * value really needs to be written to; if the value is
1280		 * already correct, we do nothing and avoid the lock.
1281		 */
1282		if (vcpu->kvm->mm->context.uses_cmm == 0) {
1283			mmap_write_lock(vcpu->kvm->mm);
1284			vcpu->kvm->mm->context.uses_cmm = 1;
1285			mmap_write_unlock(vcpu->kvm->mm);
1286		}
1287		/*
1288		 * If we are here, we are supposed to have CMMA enabled in
1289		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1290		 * while the context use_cmma flag is per process.
1291		 * It's possible that the context flag is enabled and the
1292		 * SIE flag is not, so we set the flag always; if it was
1293		 * already set, nothing changes, otherwise we enable it
1294		 * on this CPU too.
1295		 */
1296		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1297		/* Retry the ESSA instruction */
1298		kvm_s390_retry_instr(vcpu);
1299	} else {
1300		int srcu_idx;
1301
1302		mmap_read_lock(vcpu->kvm->mm);
1303		srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1304		i = __do_essa(vcpu, orc);
1305		srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
1306		mmap_read_unlock(vcpu->kvm->mm);
1307		if (i < 0)
1308			return i;
1309		/* Account for the possible extra cbrl entry */
1310		entries += i;
1311	}
1312	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1313	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1314	mmap_read_lock(gmap->mm);
1315	for (i = 0; i < entries; ++i)
1316		__gmap_zap(gmap, cbrlo[i]);
1317	mmap_read_unlock(gmap->mm);
1318	return 0;
1319}
1320
1321int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1322{
1323	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1324	case 0x8a:
1325	case 0x8e:
1326	case 0x8f:
1327		return handle_ipte_interlock(vcpu);
1328	case 0x8d:
1329		return handle_epsw(vcpu);
1330	case 0xab:
1331		return handle_essa(vcpu);
1332	case 0xaf:
1333		return handle_pfmf(vcpu);
1334	default:
1335		return -EOPNOTSUPP;
1336	}
1337}
1338
1339int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1340{
1341	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1342	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1343	int reg, rc, nr_regs;
1344	u32 ctl_array[16];
1345	u64 ga;
1346	u8 ar;
1347
1348	vcpu->stat.instruction_lctl++;
1349
1350	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1351		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1352
1353	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1354
1355	if (ga & 3)
1356		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1357
1358	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1359	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1360
1361	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1362	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1363	if (rc)
1364		return kvm_s390_inject_prog_cond(vcpu, rc);
1365	reg = reg1;
1366	nr_regs = 0;
1367	do {
1368		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1369		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1370		if (reg == reg3)
1371			break;
1372		reg = (reg + 1) % 16;
1373	} while (1);
1374	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1375	return 0;
1376}
1377
1378int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1379{
1380	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1381	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1382	int reg, rc, nr_regs;
1383	u32 ctl_array[16];
1384	u64 ga;
1385	u8 ar;
1386
1387	vcpu->stat.instruction_stctl++;
1388
1389	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1390		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1391
1392	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1393
1394	if (ga & 3)
1395		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1396
1397	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1398	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1399
1400	reg = reg1;
1401	nr_regs = 0;
1402	do {
1403		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1404		if (reg == reg3)
1405			break;
1406		reg = (reg + 1) % 16;
1407	} while (1);
1408	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1409	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1410}
1411
1412static int handle_lctlg(struct kvm_vcpu *vcpu)
1413{
1414	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1415	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1416	int reg, rc, nr_regs;
1417	u64 ctl_array[16];
1418	u64 ga;
1419	u8 ar;
1420
1421	vcpu->stat.instruction_lctlg++;
1422
1423	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1424		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1425
1426	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1427
1428	if (ga & 7)
1429		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1430
1431	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1432	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1433
1434	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1435	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1436	if (rc)
1437		return kvm_s390_inject_prog_cond(vcpu, rc);
1438	reg = reg1;
1439	nr_regs = 0;
1440	do {
1441		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1442		if (reg == reg3)
1443			break;
1444		reg = (reg + 1) % 16;
1445	} while (1);
1446	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1447	return 0;
1448}
1449
1450static int handle_stctg(struct kvm_vcpu *vcpu)
1451{
1452	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1453	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1454	int reg, rc, nr_regs;
1455	u64 ctl_array[16];
1456	u64 ga;
1457	u8 ar;
1458
1459	vcpu->stat.instruction_stctg++;
1460
1461	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1462		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1463
1464	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1465
1466	if (ga & 7)
1467		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1468
1469	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1470	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1471
1472	reg = reg1;
1473	nr_regs = 0;
1474	do {
1475		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1476		if (reg == reg3)
1477			break;
1478		reg = (reg + 1) % 16;
1479	} while (1);
1480	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1481	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1482}
1483
1484int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1485{
1486	switch (vcpu->arch.sie_block->ipb & 0x000000ff) {
1487	case 0x25:
1488		return handle_stctg(vcpu);
1489	case 0x2f:
1490		return handle_lctlg(vcpu);
1491	case 0x60:
1492	case 0x61:
1493	case 0x62:
1494		return handle_ri(vcpu);
1495	case 0x71:
1496		return handle_lpswey(vcpu);
1497	default:
1498		return -EOPNOTSUPP;
1499	}
1500}
1501
1502static int handle_tprot(struct kvm_vcpu *vcpu)
1503{
1504	u64 address, operand2;
1505	unsigned long gpa;
1506	u8 access_key;
1507	bool writable;
1508	int ret, cc;
1509	u8 ar;
1510
1511	vcpu->stat.instruction_tprot++;
1512
1513	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1514		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1515
1516	kvm_s390_get_base_disp_sse(vcpu, &address, &operand2, &ar, NULL);
1517	access_key = (operand2 & 0xf0) >> 4;
1518
1519	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1520		ipte_lock(vcpu->kvm);
1521
1522	ret = guest_translate_address_with_key(vcpu, address, ar, &gpa,
1523					       GACC_STORE, access_key);
1524	if (ret == 0) {
1525		gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1526	} else if (ret == PGM_PROTECTION) {
1527		writable = false;
1528		/* Write protected? Try again with read-only... */
1529		ret = guest_translate_address_with_key(vcpu, address, ar, &gpa,
1530						       GACC_FETCH, access_key);
1531	}
1532	if (ret >= 0) {
1533		cc = -1;
1534
1535		/* Fetching permitted; storing permitted */
1536		if (ret == 0 && writable)
1537			cc = 0;
1538		/* Fetching permitted; storing not permitted */
1539		else if (ret == 0 && !writable)
1540			cc = 1;
1541		/* Fetching not permitted; storing not permitted */
1542		else if (ret == PGM_PROTECTION)
1543			cc = 2;
1544		/* Translation not available */
1545		else if (ret != PGM_ADDRESSING && ret != PGM_TRANSLATION_SPEC)
1546			cc = 3;
1547
1548		if (cc != -1) {
1549			kvm_s390_set_psw_cc(vcpu, cc);
1550			ret = 0;
1551		} else {
1552			ret = kvm_s390_inject_program_int(vcpu, ret);
1553		}
1554	}
1555
1556	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1557		ipte_unlock(vcpu->kvm);
1558	return ret;
1559}
1560
1561int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1562{
1563	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1564	case 0x01:
1565		return handle_tprot(vcpu);
1566	default:
1567		return -EOPNOTSUPP;
1568	}
1569}
1570
1571static int handle_sckpf(struct kvm_vcpu *vcpu)
1572{
1573	u32 value;
1574
1575	vcpu->stat.instruction_sckpf++;
1576
1577	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1578		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1579
1580	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1581		return kvm_s390_inject_program_int(vcpu,
1582						   PGM_SPECIFICATION);
1583
1584	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1585	vcpu->arch.sie_block->todpr = value;
1586
1587	return 0;
1588}
1589
1590static int handle_ptff(struct kvm_vcpu *vcpu)
1591{
1592	vcpu->stat.instruction_ptff++;
1593
1594	/* we don't emulate any control instructions yet */
1595	kvm_s390_set_psw_cc(vcpu, 3);
1596	return 0;
1597}
1598
1599int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1600{
1601	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1602	case 0x04:
1603		return handle_ptff(vcpu);
1604	case 0x07:
1605		return handle_sckpf(vcpu);
1606	default:
1607		return -EOPNOTSUPP;
1608	}
1609}