xref: /openbmc/linux/arch/s390/kvm/priv.c (revision 82e6fdd6)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * handling privileged instructions
4  *
5  * Copyright IBM Corp. 2008, 2018
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/compat.h>
15 #include <linux/mm_types.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/pgtable.h>
24 #include <asm/page-states.h>
25 #include <asm/pgalloc.h>
26 #include <asm/gmap.h>
27 #include <asm/io.h>
28 #include <asm/ptrace.h>
29 #include <asm/compat.h>
30 #include <asm/sclp.h>
31 #include "gaccess.h"
32 #include "kvm-s390.h"
33 #include "trace.h"
34 
35 static int handle_ri(struct kvm_vcpu *vcpu)
36 {
37 	vcpu->stat.instruction_ri++;
38 
39 	if (test_kvm_facility(vcpu->kvm, 64)) {
40 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
41 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
42 		kvm_s390_retry_instr(vcpu);
43 		return 0;
44 	} else
45 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
46 }
47 
48 int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
49 {
50 	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
51 		return handle_ri(vcpu);
52 	else
53 		return -EOPNOTSUPP;
54 }
55 
56 static int handle_gs(struct kvm_vcpu *vcpu)
57 {
58 	vcpu->stat.instruction_gs++;
59 
60 	if (test_kvm_facility(vcpu->kvm, 133)) {
61 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (lazy)");
62 		preempt_disable();
63 		__ctl_set_bit(2, 4);
64 		current->thread.gs_cb = (struct gs_cb *)&vcpu->run->s.regs.gscb;
65 		restore_gs_cb(current->thread.gs_cb);
66 		preempt_enable();
67 		vcpu->arch.sie_block->ecb |= ECB_GS;
68 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
69 		vcpu->arch.gs_enabled = 1;
70 		kvm_s390_retry_instr(vcpu);
71 		return 0;
72 	} else
73 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
74 }
75 
76 int kvm_s390_handle_e3(struct kvm_vcpu *vcpu)
77 {
78 	int code = vcpu->arch.sie_block->ipb & 0xff;
79 
80 	if (code == 0x49 || code == 0x4d)
81 		return handle_gs(vcpu);
82 	else
83 		return -EOPNOTSUPP;
84 }
85 /* Handle SCK (SET CLOCK) interception */
86 static int handle_set_clock(struct kvm_vcpu *vcpu)
87 {
88 	struct kvm_s390_vm_tod_clock gtod = { 0 };
89 	int rc;
90 	u8 ar;
91 	u64 op2;
92 
93 	vcpu->stat.instruction_sck++;
94 
95 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
96 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
97 
98 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
99 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
100 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
101 	rc = read_guest(vcpu, op2, ar, &gtod.tod, sizeof(gtod.tod));
102 	if (rc)
103 		return kvm_s390_inject_prog_cond(vcpu, rc);
104 
105 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", gtod.tod);
106 	kvm_s390_set_tod_clock(vcpu->kvm, &gtod);
107 
108 	kvm_s390_set_psw_cc(vcpu, 0);
109 	return 0;
110 }
111 
112 static int handle_set_prefix(struct kvm_vcpu *vcpu)
113 {
114 	u64 operand2;
115 	u32 address;
116 	int rc;
117 	u8 ar;
118 
119 	vcpu->stat.instruction_spx++;
120 
121 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
122 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
123 
124 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
125 
126 	/* must be word boundary */
127 	if (operand2 & 3)
128 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
129 
130 	/* get the value */
131 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
132 	if (rc)
133 		return kvm_s390_inject_prog_cond(vcpu, rc);
134 
135 	address &= 0x7fffe000u;
136 
137 	/*
138 	 * Make sure the new value is valid memory. We only need to check the
139 	 * first page, since address is 8k aligned and memory pieces are always
140 	 * at least 1MB aligned and have at least a size of 1MB.
141 	 */
142 	if (kvm_is_error_gpa(vcpu->kvm, address))
143 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
144 
145 	kvm_s390_set_prefix(vcpu, address);
146 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
147 	return 0;
148 }
149 
150 static int handle_store_prefix(struct kvm_vcpu *vcpu)
151 {
152 	u64 operand2;
153 	u32 address;
154 	int rc;
155 	u8 ar;
156 
157 	vcpu->stat.instruction_stpx++;
158 
159 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
160 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
161 
162 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
163 
164 	/* must be word boundary */
165 	if (operand2 & 3)
166 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
167 
168 	address = kvm_s390_get_prefix(vcpu);
169 
170 	/* get the value */
171 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
172 	if (rc)
173 		return kvm_s390_inject_prog_cond(vcpu, rc);
174 
175 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
176 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
177 	return 0;
178 }
179 
180 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
181 {
182 	u16 vcpu_id = vcpu->vcpu_id;
183 	u64 ga;
184 	int rc;
185 	u8 ar;
186 
187 	vcpu->stat.instruction_stap++;
188 
189 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
190 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
191 
192 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
193 
194 	if (ga & 1)
195 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
196 
197 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
198 	if (rc)
199 		return kvm_s390_inject_prog_cond(vcpu, rc);
200 
201 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
202 	trace_kvm_s390_handle_stap(vcpu, ga);
203 	return 0;
204 }
205 
206 int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
207 {
208 	int rc = 0;
209 	struct kvm_s390_sie_block *sie_block = vcpu->arch.sie_block;
210 
211 	trace_kvm_s390_skey_related_inst(vcpu);
212 	if (!(sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)) &&
213 	    !kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
214 		return rc;
215 
216 	rc = s390_enable_skey();
217 	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
218 	if (!rc) {
219 		if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
220 			kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS);
221 		else
222 			sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE |
223 					     ICTL_RRBE);
224 	}
225 	return rc;
226 }
227 
228 static int try_handle_skey(struct kvm_vcpu *vcpu)
229 {
230 	int rc;
231 
232 	rc = kvm_s390_skey_check_enable(vcpu);
233 	if (rc)
234 		return rc;
235 	if (sclp.has_skey) {
236 		/* with storage-key facility, SIE interprets it for us */
237 		kvm_s390_retry_instr(vcpu);
238 		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
239 		return -EAGAIN;
240 	}
241 	return 0;
242 }
243 
244 static int handle_iske(struct kvm_vcpu *vcpu)
245 {
246 	unsigned long addr;
247 	unsigned char key;
248 	int reg1, reg2;
249 	int rc;
250 
251 	vcpu->stat.instruction_iske++;
252 
253 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
254 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
255 
256 	rc = try_handle_skey(vcpu);
257 	if (rc)
258 		return rc != -EAGAIN ? rc : 0;
259 
260 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
261 
262 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
263 	addr = kvm_s390_logical_to_effective(vcpu, addr);
264 	addr = kvm_s390_real_to_abs(vcpu, addr);
265 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
266 	if (kvm_is_error_hva(addr))
267 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
268 
269 	down_read(&current->mm->mmap_sem);
270 	rc = get_guest_storage_key(current->mm, addr, &key);
271 	up_read(&current->mm->mmap_sem);
272 	if (rc)
273 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
274 	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
275 	vcpu->run->s.regs.gprs[reg1] |= key;
276 	return 0;
277 }
278 
279 static int handle_rrbe(struct kvm_vcpu *vcpu)
280 {
281 	unsigned long addr;
282 	int reg1, reg2;
283 	int rc;
284 
285 	vcpu->stat.instruction_rrbe++;
286 
287 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
288 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
289 
290 	rc = try_handle_skey(vcpu);
291 	if (rc)
292 		return rc != -EAGAIN ? rc : 0;
293 
294 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
295 
296 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
297 	addr = kvm_s390_logical_to_effective(vcpu, addr);
298 	addr = kvm_s390_real_to_abs(vcpu, addr);
299 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
300 	if (kvm_is_error_hva(addr))
301 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
302 
303 	down_read(&current->mm->mmap_sem);
304 	rc = reset_guest_reference_bit(current->mm, addr);
305 	up_read(&current->mm->mmap_sem);
306 	if (rc < 0)
307 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
308 
309 	kvm_s390_set_psw_cc(vcpu, rc);
310 	return 0;
311 }
312 
313 #define SSKE_NQ 0x8
314 #define SSKE_MR 0x4
315 #define SSKE_MC 0x2
316 #define SSKE_MB 0x1
317 static int handle_sske(struct kvm_vcpu *vcpu)
318 {
319 	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
320 	unsigned long start, end;
321 	unsigned char key, oldkey;
322 	int reg1, reg2;
323 	int rc;
324 
325 	vcpu->stat.instruction_sske++;
326 
327 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
328 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
329 
330 	rc = try_handle_skey(vcpu);
331 	if (rc)
332 		return rc != -EAGAIN ? rc : 0;
333 
334 	if (!test_kvm_facility(vcpu->kvm, 8))
335 		m3 &= ~SSKE_MB;
336 	if (!test_kvm_facility(vcpu->kvm, 10))
337 		m3 &= ~(SSKE_MC | SSKE_MR);
338 	if (!test_kvm_facility(vcpu->kvm, 14))
339 		m3 &= ~SSKE_NQ;
340 
341 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
342 
343 	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
344 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
345 	start = kvm_s390_logical_to_effective(vcpu, start);
346 	if (m3 & SSKE_MB) {
347 		/* start already designates an absolute address */
348 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
349 	} else {
350 		start = kvm_s390_real_to_abs(vcpu, start);
351 		end = start + PAGE_SIZE;
352 	}
353 
354 	while (start != end) {
355 		unsigned long addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
356 
357 		if (kvm_is_error_hva(addr))
358 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
359 
360 		down_read(&current->mm->mmap_sem);
361 		rc = cond_set_guest_storage_key(current->mm, addr, key, &oldkey,
362 						m3 & SSKE_NQ, m3 & SSKE_MR,
363 						m3 & SSKE_MC);
364 		up_read(&current->mm->mmap_sem);
365 		if (rc < 0)
366 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
367 		start += PAGE_SIZE;
368 	}
369 
370 	if (m3 & (SSKE_MC | SSKE_MR)) {
371 		if (m3 & SSKE_MB) {
372 			/* skey in reg1 is unpredictable */
373 			kvm_s390_set_psw_cc(vcpu, 3);
374 		} else {
375 			kvm_s390_set_psw_cc(vcpu, rc);
376 			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
377 			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
378 		}
379 	}
380 	if (m3 & SSKE_MB) {
381 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
382 			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
383 		else
384 			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
385 		end = kvm_s390_logical_to_effective(vcpu, end);
386 		vcpu->run->s.regs.gprs[reg2] |= end;
387 	}
388 	return 0;
389 }
390 
391 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
392 {
393 	vcpu->stat.instruction_ipte_interlock++;
394 	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
395 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
396 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
397 	kvm_s390_retry_instr(vcpu);
398 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
399 	return 0;
400 }
401 
402 static int handle_test_block(struct kvm_vcpu *vcpu)
403 {
404 	gpa_t addr;
405 	int reg2;
406 
407 	vcpu->stat.instruction_tb++;
408 
409 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
410 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
411 
412 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
413 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
414 	addr = kvm_s390_logical_to_effective(vcpu, addr);
415 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
416 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
417 	addr = kvm_s390_real_to_abs(vcpu, addr);
418 
419 	if (kvm_is_error_gpa(vcpu->kvm, addr))
420 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
421 	/*
422 	 * We don't expect errors on modern systems, and do not care
423 	 * about storage keys (yet), so let's just clear the page.
424 	 */
425 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
426 		return -EFAULT;
427 	kvm_s390_set_psw_cc(vcpu, 0);
428 	vcpu->run->s.regs.gprs[0] = 0;
429 	return 0;
430 }
431 
432 static int handle_tpi(struct kvm_vcpu *vcpu)
433 {
434 	struct kvm_s390_interrupt_info *inti;
435 	unsigned long len;
436 	u32 tpi_data[3];
437 	int rc;
438 	u64 addr;
439 	u8 ar;
440 
441 	vcpu->stat.instruction_tpi++;
442 
443 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
444 	if (addr & 3)
445 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
446 
447 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
448 	if (!inti) {
449 		kvm_s390_set_psw_cc(vcpu, 0);
450 		return 0;
451 	}
452 
453 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
454 	tpi_data[1] = inti->io.io_int_parm;
455 	tpi_data[2] = inti->io.io_int_word;
456 	if (addr) {
457 		/*
458 		 * Store the two-word I/O interruption code into the
459 		 * provided area.
460 		 */
461 		len = sizeof(tpi_data) - 4;
462 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
463 		if (rc) {
464 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
465 			goto reinject_interrupt;
466 		}
467 	} else {
468 		/*
469 		 * Store the three-word I/O interruption code into
470 		 * the appropriate lowcore area.
471 		 */
472 		len = sizeof(tpi_data);
473 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
474 			/* failed writes to the low core are not recoverable */
475 			rc = -EFAULT;
476 			goto reinject_interrupt;
477 		}
478 	}
479 
480 	/* irq was successfully handed to the guest */
481 	kfree(inti);
482 	kvm_s390_set_psw_cc(vcpu, 1);
483 	return 0;
484 reinject_interrupt:
485 	/*
486 	 * If we encounter a problem storing the interruption code, the
487 	 * instruction is suppressed from the guest's view: reinject the
488 	 * interrupt.
489 	 */
490 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
491 		kfree(inti);
492 		rc = -EFAULT;
493 	}
494 	/* don't set the cc, a pgm irq was injected or we drop to user space */
495 	return rc ? -EFAULT : 0;
496 }
497 
498 static int handle_tsch(struct kvm_vcpu *vcpu)
499 {
500 	struct kvm_s390_interrupt_info *inti = NULL;
501 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
502 
503 	vcpu->stat.instruction_tsch++;
504 
505 	/* a valid schid has at least one bit set */
506 	if (vcpu->run->s.regs.gprs[1])
507 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
508 					   vcpu->run->s.regs.gprs[1]);
509 
510 	/*
511 	 * Prepare exit to userspace.
512 	 * We indicate whether we dequeued a pending I/O interrupt
513 	 * so that userspace can re-inject it if the instruction gets
514 	 * a program check. While this may re-order the pending I/O
515 	 * interrupts, this is no problem since the priority is kept
516 	 * intact.
517 	 */
518 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
519 	vcpu->run->s390_tsch.dequeued = !!inti;
520 	if (inti) {
521 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
522 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
523 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
524 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
525 	}
526 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
527 	kfree(inti);
528 	return -EREMOTE;
529 }
530 
531 static int handle_io_inst(struct kvm_vcpu *vcpu)
532 {
533 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
534 
535 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
536 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
537 
538 	if (vcpu->kvm->arch.css_support) {
539 		/*
540 		 * Most I/O instructions will be handled by userspace.
541 		 * Exceptions are tpi and the interrupt portion of tsch.
542 		 */
543 		if (vcpu->arch.sie_block->ipa == 0xb236)
544 			return handle_tpi(vcpu);
545 		if (vcpu->arch.sie_block->ipa == 0xb235)
546 			return handle_tsch(vcpu);
547 		/* Handle in userspace. */
548 		vcpu->stat.instruction_io_other++;
549 		return -EOPNOTSUPP;
550 	} else {
551 		/*
552 		 * Set condition code 3 to stop the guest from issuing channel
553 		 * I/O instructions.
554 		 */
555 		kvm_s390_set_psw_cc(vcpu, 3);
556 		return 0;
557 	}
558 }
559 
560 static int handle_stfl(struct kvm_vcpu *vcpu)
561 {
562 	int rc;
563 	unsigned int fac;
564 
565 	vcpu->stat.instruction_stfl++;
566 
567 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
568 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
569 
570 	/*
571 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
572 	 * into a u32 memory representation. They will remain bits 0-31.
573 	 */
574 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
575 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
576 			    &fac, sizeof(fac));
577 	if (rc)
578 		return rc;
579 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
580 	trace_kvm_s390_handle_stfl(vcpu, fac);
581 	return 0;
582 }
583 
584 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
585 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
586 #define PSW_ADDR_24 0x0000000000ffffffUL
587 #define PSW_ADDR_31 0x000000007fffffffUL
588 
589 int is_valid_psw(psw_t *psw)
590 {
591 	if (psw->mask & PSW_MASK_UNASSIGNED)
592 		return 0;
593 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
594 		if (psw->addr & ~PSW_ADDR_31)
595 			return 0;
596 	}
597 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
598 		return 0;
599 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
600 		return 0;
601 	if (psw->addr & 1)
602 		return 0;
603 	return 1;
604 }
605 
606 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
607 {
608 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
609 	psw_compat_t new_psw;
610 	u64 addr;
611 	int rc;
612 	u8 ar;
613 
614 	vcpu->stat.instruction_lpsw++;
615 
616 	if (gpsw->mask & PSW_MASK_PSTATE)
617 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
618 
619 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
620 	if (addr & 7)
621 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
622 
623 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
624 	if (rc)
625 		return kvm_s390_inject_prog_cond(vcpu, rc);
626 	if (!(new_psw.mask & PSW32_MASK_BASE))
627 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
628 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
629 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
630 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
631 	if (!is_valid_psw(gpsw))
632 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
633 	return 0;
634 }
635 
636 static int handle_lpswe(struct kvm_vcpu *vcpu)
637 {
638 	psw_t new_psw;
639 	u64 addr;
640 	int rc;
641 	u8 ar;
642 
643 	vcpu->stat.instruction_lpswe++;
644 
645 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
646 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
647 
648 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
649 	if (addr & 7)
650 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
651 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
652 	if (rc)
653 		return kvm_s390_inject_prog_cond(vcpu, rc);
654 	vcpu->arch.sie_block->gpsw = new_psw;
655 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
656 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
657 	return 0;
658 }
659 
660 static int handle_stidp(struct kvm_vcpu *vcpu)
661 {
662 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
663 	u64 operand2;
664 	int rc;
665 	u8 ar;
666 
667 	vcpu->stat.instruction_stidp++;
668 
669 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
670 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
671 
672 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
673 
674 	if (operand2 & 7)
675 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
676 
677 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
678 	if (rc)
679 		return kvm_s390_inject_prog_cond(vcpu, rc);
680 
681 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
682 	return 0;
683 }
684 
685 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
686 {
687 	int cpus = 0;
688 	int n;
689 
690 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
691 
692 	/* deal with other level 3 hypervisors */
693 	if (stsi(mem, 3, 2, 2))
694 		mem->count = 0;
695 	if (mem->count < 8)
696 		mem->count++;
697 	for (n = mem->count - 1; n > 0 ; n--)
698 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
699 
700 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
701 	mem->vm[0].cpus_total = cpus;
702 	mem->vm[0].cpus_configured = cpus;
703 	mem->vm[0].cpus_standby = 0;
704 	mem->vm[0].cpus_reserved = 0;
705 	mem->vm[0].caf = 1000;
706 	memcpy(mem->vm[0].name, "KVMguest", 8);
707 	ASCEBC(mem->vm[0].name, 8);
708 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
709 	ASCEBC(mem->vm[0].cpi, 16);
710 }
711 
712 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
713 				 u8 fc, u8 sel1, u16 sel2)
714 {
715 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
716 	vcpu->run->s390_stsi.addr = addr;
717 	vcpu->run->s390_stsi.ar = ar;
718 	vcpu->run->s390_stsi.fc = fc;
719 	vcpu->run->s390_stsi.sel1 = sel1;
720 	vcpu->run->s390_stsi.sel2 = sel2;
721 }
722 
723 static int handle_stsi(struct kvm_vcpu *vcpu)
724 {
725 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
726 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
727 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
728 	unsigned long mem = 0;
729 	u64 operand2;
730 	int rc = 0;
731 	u8 ar;
732 
733 	vcpu->stat.instruction_stsi++;
734 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
735 
736 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
737 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
738 
739 	if (fc > 3) {
740 		kvm_s390_set_psw_cc(vcpu, 3);
741 		return 0;
742 	}
743 
744 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
745 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
746 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
747 
748 	if (fc == 0) {
749 		vcpu->run->s.regs.gprs[0] = 3 << 28;
750 		kvm_s390_set_psw_cc(vcpu, 0);
751 		return 0;
752 	}
753 
754 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
755 
756 	if (operand2 & 0xfff)
757 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
758 
759 	switch (fc) {
760 	case 1: /* same handling for 1 and 2 */
761 	case 2:
762 		mem = get_zeroed_page(GFP_KERNEL);
763 		if (!mem)
764 			goto out_no_data;
765 		if (stsi((void *) mem, fc, sel1, sel2))
766 			goto out_no_data;
767 		break;
768 	case 3:
769 		if (sel1 != 2 || sel2 != 2)
770 			goto out_no_data;
771 		mem = get_zeroed_page(GFP_KERNEL);
772 		if (!mem)
773 			goto out_no_data;
774 		handle_stsi_3_2_2(vcpu, (void *) mem);
775 		break;
776 	}
777 
778 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
779 	if (rc) {
780 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
781 		goto out;
782 	}
783 	if (vcpu->kvm->arch.user_stsi) {
784 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
785 		rc = -EREMOTE;
786 	}
787 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
788 	free_page(mem);
789 	kvm_s390_set_psw_cc(vcpu, 0);
790 	vcpu->run->s.regs.gprs[0] = 0;
791 	return rc;
792 out_no_data:
793 	kvm_s390_set_psw_cc(vcpu, 3);
794 out:
795 	free_page(mem);
796 	return rc;
797 }
798 
799 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
800 {
801 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
802 	case 0x02:
803 		return handle_stidp(vcpu);
804 	case 0x04:
805 		return handle_set_clock(vcpu);
806 	case 0x10:
807 		return handle_set_prefix(vcpu);
808 	case 0x11:
809 		return handle_store_prefix(vcpu);
810 	case 0x12:
811 		return handle_store_cpu_address(vcpu);
812 	case 0x14:
813 		return kvm_s390_handle_vsie(vcpu);
814 	case 0x21:
815 	case 0x50:
816 		return handle_ipte_interlock(vcpu);
817 	case 0x29:
818 		return handle_iske(vcpu);
819 	case 0x2a:
820 		return handle_rrbe(vcpu);
821 	case 0x2b:
822 		return handle_sske(vcpu);
823 	case 0x2c:
824 		return handle_test_block(vcpu);
825 	case 0x30:
826 	case 0x31:
827 	case 0x32:
828 	case 0x33:
829 	case 0x34:
830 	case 0x35:
831 	case 0x36:
832 	case 0x37:
833 	case 0x38:
834 	case 0x39:
835 	case 0x3a:
836 	case 0x3b:
837 	case 0x3c:
838 	case 0x5f:
839 	case 0x74:
840 	case 0x76:
841 		return handle_io_inst(vcpu);
842 	case 0x56:
843 		return handle_sthyi(vcpu);
844 	case 0x7d:
845 		return handle_stsi(vcpu);
846 	case 0xb1:
847 		return handle_stfl(vcpu);
848 	case 0xb2:
849 		return handle_lpswe(vcpu);
850 	default:
851 		return -EOPNOTSUPP;
852 	}
853 }
854 
855 static int handle_epsw(struct kvm_vcpu *vcpu)
856 {
857 	int reg1, reg2;
858 
859 	vcpu->stat.instruction_epsw++;
860 
861 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
862 
863 	/* This basically extracts the mask half of the psw. */
864 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
865 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
866 	if (reg2) {
867 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
868 		vcpu->run->s.regs.gprs[reg2] |=
869 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
870 	}
871 	return 0;
872 }
873 
874 #define PFMF_RESERVED   0xfffc0101UL
875 #define PFMF_SK         0x00020000UL
876 #define PFMF_CF         0x00010000UL
877 #define PFMF_UI         0x00008000UL
878 #define PFMF_FSC        0x00007000UL
879 #define PFMF_NQ         0x00000800UL
880 #define PFMF_MR         0x00000400UL
881 #define PFMF_MC         0x00000200UL
882 #define PFMF_KEY        0x000000feUL
883 
884 static int handle_pfmf(struct kvm_vcpu *vcpu)
885 {
886 	bool mr = false, mc = false, nq;
887 	int reg1, reg2;
888 	unsigned long start, end;
889 	unsigned char key;
890 
891 	vcpu->stat.instruction_pfmf++;
892 
893 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
894 
895 	if (!test_kvm_facility(vcpu->kvm, 8))
896 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
897 
898 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
899 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
900 
901 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
902 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
903 
904 	/* Only provide non-quiescing support if enabled for the guest */
905 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
906 	    !test_kvm_facility(vcpu->kvm, 14))
907 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
908 
909 	/* Only provide conditional-SSKE support if enabled for the guest */
910 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
911 	    test_kvm_facility(vcpu->kvm, 10)) {
912 		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
913 		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
914 	}
915 
916 	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
917 	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
918 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
919 	start = kvm_s390_logical_to_effective(vcpu, start);
920 
921 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
922 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
923 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
924 	}
925 
926 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
927 	case 0x00000000:
928 		/* only 4k frames specify a real address */
929 		start = kvm_s390_real_to_abs(vcpu, start);
930 		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
931 		break;
932 	case 0x00001000:
933 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
934 		break;
935 	case 0x00002000:
936 		/* only support 2G frame size if EDAT2 is available and we are
937 		   not in 24-bit addressing mode */
938 		if (!test_kvm_facility(vcpu->kvm, 78) ||
939 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
940 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
941 		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
942 		break;
943 	default:
944 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
945 	}
946 
947 	while (start != end) {
948 		unsigned long useraddr;
949 
950 		/* Translate guest address to host address */
951 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
952 		if (kvm_is_error_hva(useraddr))
953 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
954 
955 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
956 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
957 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
958 		}
959 
960 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
961 			int rc = kvm_s390_skey_check_enable(vcpu);
962 
963 			if (rc)
964 				return rc;
965 			down_read(&current->mm->mmap_sem);
966 			rc = cond_set_guest_storage_key(current->mm, useraddr,
967 							key, NULL, nq, mr, mc);
968 			up_read(&current->mm->mmap_sem);
969 			if (rc < 0)
970 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
971 		}
972 
973 		start += PAGE_SIZE;
974 	}
975 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
976 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
977 			vcpu->run->s.regs.gprs[reg2] = end;
978 		} else {
979 			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
980 			end = kvm_s390_logical_to_effective(vcpu, end);
981 			vcpu->run->s.regs.gprs[reg2] |= end;
982 		}
983 	}
984 	return 0;
985 }
986 
987 static inline int do_essa(struct kvm_vcpu *vcpu, const int orc)
988 {
989 	struct kvm_s390_migration_state *ms = vcpu->kvm->arch.migration_state;
990 	int r1, r2, nappended, entries;
991 	unsigned long gfn, hva, res, pgstev, ptev;
992 	unsigned long *cbrlo;
993 
994 	/*
995 	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
996 	 * machine check here we either handle it or crash
997 	 */
998 
999 	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
1000 	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
1001 	hva = gfn_to_hva(vcpu->kvm, gfn);
1002 	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1003 
1004 	if (kvm_is_error_hva(hva))
1005 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1006 
1007 	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
1008 	if (nappended < 0) {
1009 		res = orc ? 0x10 : 0;
1010 		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
1011 		return 0;
1012 	}
1013 	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
1014 	/*
1015 	 * Set the block-content state part of the result. 0 means resident, so
1016 	 * nothing to do if the page is valid. 2 is for preserved pages
1017 	 * (non-present and non-zero), and 3 for zero pages (non-present and
1018 	 * zero).
1019 	 */
1020 	if (ptev & _PAGE_INVALID) {
1021 		res |= 2;
1022 		if (pgstev & _PGSTE_GPS_ZERO)
1023 			res |= 1;
1024 	}
1025 	if (pgstev & _PGSTE_GPS_NODAT)
1026 		res |= 0x20;
1027 	vcpu->run->s.regs.gprs[r1] = res;
1028 	/*
1029 	 * It is possible that all the normal 511 slots were full, in which case
1030 	 * we will now write in the 512th slot, which is reserved for host use.
1031 	 * In both cases we let the normal essa handling code process all the
1032 	 * slots, including the reserved one, if needed.
1033 	 */
1034 	if (nappended > 0) {
1035 		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1036 		cbrlo[entries] = gfn << PAGE_SHIFT;
1037 	}
1038 
1039 	if (orc && gfn < ms->bitmap_size) {
1040 		/* increment only if we are really flipping the bit to 1 */
1041 		if (!test_and_set_bit(gfn, ms->pgste_bitmap))
1042 			atomic64_inc(&ms->dirty_pages);
1043 	}
1044 
1045 	return nappended;
1046 }
1047 
1048 static int handle_essa(struct kvm_vcpu *vcpu)
1049 {
1050 	/* entries expected to be 1FF */
1051 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1052 	unsigned long *cbrlo;
1053 	struct gmap *gmap;
1054 	int i, orc;
1055 
1056 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1057 	gmap = vcpu->arch.gmap;
1058 	vcpu->stat.instruction_essa++;
1059 	if (!vcpu->kvm->arch.use_cmma)
1060 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1061 
1062 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1063 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1064 	/* Check for invalid operation request code */
1065 	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1066 	/* ORCs 0-6 are always valid */
1067 	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1068 						: ESSA_SET_STABLE_IF_RESIDENT))
1069 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1070 
1071 	if (likely(!vcpu->kvm->arch.migration_state)) {
1072 		/*
1073 		 * CMMA is enabled in the KVM settings, but is disabled in
1074 		 * the SIE block and in the mm_context, and we are not doing
1075 		 * a migration. Enable CMMA in the mm_context.
1076 		 * Since we need to take a write lock to write to the context
1077 		 * to avoid races with storage keys handling, we check if the
1078 		 * value really needs to be written to; if the value is
1079 		 * already correct, we do nothing and avoid the lock.
1080 		 */
1081 		if (vcpu->kvm->mm->context.use_cmma == 0) {
1082 			down_write(&vcpu->kvm->mm->mmap_sem);
1083 			vcpu->kvm->mm->context.use_cmma = 1;
1084 			up_write(&vcpu->kvm->mm->mmap_sem);
1085 		}
1086 		/*
1087 		 * If we are here, we are supposed to have CMMA enabled in
1088 		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1089 		 * while the context use_cmma flag is per process.
1090 		 * It's possible that the context flag is enabled and the
1091 		 * SIE flag is not, so we set the flag always; if it was
1092 		 * already set, nothing changes, otherwise we enable it
1093 		 * on this CPU too.
1094 		 */
1095 		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1096 		/* Retry the ESSA instruction */
1097 		kvm_s390_retry_instr(vcpu);
1098 	} else {
1099 		/* Account for the possible extra cbrl entry */
1100 		i = do_essa(vcpu, orc);
1101 		if (i < 0)
1102 			return i;
1103 		entries += i;
1104 	}
1105 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1106 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1107 	down_read(&gmap->mm->mmap_sem);
1108 	for (i = 0; i < entries; ++i)
1109 		__gmap_zap(gmap, cbrlo[i]);
1110 	up_read(&gmap->mm->mmap_sem);
1111 	return 0;
1112 }
1113 
1114 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1115 {
1116 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1117 	case 0x8a:
1118 	case 0x8e:
1119 	case 0x8f:
1120 		return handle_ipte_interlock(vcpu);
1121 	case 0x8d:
1122 		return handle_epsw(vcpu);
1123 	case 0xab:
1124 		return handle_essa(vcpu);
1125 	case 0xaf:
1126 		return handle_pfmf(vcpu);
1127 	default:
1128 		return -EOPNOTSUPP;
1129 	}
1130 }
1131 
1132 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1133 {
1134 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1135 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1136 	int reg, rc, nr_regs;
1137 	u32 ctl_array[16];
1138 	u64 ga;
1139 	u8 ar;
1140 
1141 	vcpu->stat.instruction_lctl++;
1142 
1143 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1144 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1145 
1146 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1147 
1148 	if (ga & 3)
1149 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1150 
1151 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1152 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1153 
1154 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1155 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1156 	if (rc)
1157 		return kvm_s390_inject_prog_cond(vcpu, rc);
1158 	reg = reg1;
1159 	nr_regs = 0;
1160 	do {
1161 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1162 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1163 		if (reg == reg3)
1164 			break;
1165 		reg = (reg + 1) % 16;
1166 	} while (1);
1167 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1168 	return 0;
1169 }
1170 
1171 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1172 {
1173 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1174 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1175 	int reg, rc, nr_regs;
1176 	u32 ctl_array[16];
1177 	u64 ga;
1178 	u8 ar;
1179 
1180 	vcpu->stat.instruction_stctl++;
1181 
1182 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1183 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1184 
1185 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1186 
1187 	if (ga & 3)
1188 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1189 
1190 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1191 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1192 
1193 	reg = reg1;
1194 	nr_regs = 0;
1195 	do {
1196 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1197 		if (reg == reg3)
1198 			break;
1199 		reg = (reg + 1) % 16;
1200 	} while (1);
1201 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1202 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1203 }
1204 
1205 static int handle_lctlg(struct kvm_vcpu *vcpu)
1206 {
1207 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1208 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1209 	int reg, rc, nr_regs;
1210 	u64 ctl_array[16];
1211 	u64 ga;
1212 	u8 ar;
1213 
1214 	vcpu->stat.instruction_lctlg++;
1215 
1216 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1217 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1218 
1219 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1220 
1221 	if (ga & 7)
1222 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1223 
1224 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1225 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1226 
1227 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1228 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1229 	if (rc)
1230 		return kvm_s390_inject_prog_cond(vcpu, rc);
1231 	reg = reg1;
1232 	nr_regs = 0;
1233 	do {
1234 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1235 		if (reg == reg3)
1236 			break;
1237 		reg = (reg + 1) % 16;
1238 	} while (1);
1239 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1240 	return 0;
1241 }
1242 
1243 static int handle_stctg(struct kvm_vcpu *vcpu)
1244 {
1245 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1246 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1247 	int reg, rc, nr_regs;
1248 	u64 ctl_array[16];
1249 	u64 ga;
1250 	u8 ar;
1251 
1252 	vcpu->stat.instruction_stctg++;
1253 
1254 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1255 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1256 
1257 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1258 
1259 	if (ga & 7)
1260 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1261 
1262 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1263 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1264 
1265 	reg = reg1;
1266 	nr_regs = 0;
1267 	do {
1268 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1269 		if (reg == reg3)
1270 			break;
1271 		reg = (reg + 1) % 16;
1272 	} while (1);
1273 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1274 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1275 }
1276 
1277 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1278 {
1279 	switch (vcpu->arch.sie_block->ipb & 0x000000ff) {
1280 	case 0x25:
1281 		return handle_stctg(vcpu);
1282 	case 0x2f:
1283 		return handle_lctlg(vcpu);
1284 	case 0x60:
1285 	case 0x61:
1286 	case 0x62:
1287 		return handle_ri(vcpu);
1288 	default:
1289 		return -EOPNOTSUPP;
1290 	}
1291 }
1292 
1293 static int handle_tprot(struct kvm_vcpu *vcpu)
1294 {
1295 	u64 address1, address2;
1296 	unsigned long hva, gpa;
1297 	int ret = 0, cc = 0;
1298 	bool writable;
1299 	u8 ar;
1300 
1301 	vcpu->stat.instruction_tprot++;
1302 
1303 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1304 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1305 
1306 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1307 
1308 	/* we only handle the Linux memory detection case:
1309 	 * access key == 0
1310 	 * everything else goes to userspace. */
1311 	if (address2 & 0xf0)
1312 		return -EOPNOTSUPP;
1313 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1314 		ipte_lock(vcpu);
1315 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1316 	if (ret == PGM_PROTECTION) {
1317 		/* Write protected? Try again with read-only... */
1318 		cc = 1;
1319 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
1320 					      GACC_FETCH);
1321 	}
1322 	if (ret) {
1323 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1324 			ret = kvm_s390_inject_program_int(vcpu, ret);
1325 		} else if (ret > 0) {
1326 			/* Translation not available */
1327 			kvm_s390_set_psw_cc(vcpu, 3);
1328 			ret = 0;
1329 		}
1330 		goto out_unlock;
1331 	}
1332 
1333 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1334 	if (kvm_is_error_hva(hva)) {
1335 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1336 	} else {
1337 		if (!writable)
1338 			cc = 1;		/* Write not permitted ==> read-only */
1339 		kvm_s390_set_psw_cc(vcpu, cc);
1340 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1341 	}
1342 out_unlock:
1343 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1344 		ipte_unlock(vcpu);
1345 	return ret;
1346 }
1347 
1348 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1349 {
1350 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1351 	case 0x01:
1352 		return handle_tprot(vcpu);
1353 	default:
1354 		return -EOPNOTSUPP;
1355 	}
1356 }
1357 
1358 static int handle_sckpf(struct kvm_vcpu *vcpu)
1359 {
1360 	u32 value;
1361 
1362 	vcpu->stat.instruction_sckpf++;
1363 
1364 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1365 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1366 
1367 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1368 		return kvm_s390_inject_program_int(vcpu,
1369 						   PGM_SPECIFICATION);
1370 
1371 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1372 	vcpu->arch.sie_block->todpr = value;
1373 
1374 	return 0;
1375 }
1376 
1377 static int handle_ptff(struct kvm_vcpu *vcpu)
1378 {
1379 	vcpu->stat.instruction_ptff++;
1380 
1381 	/* we don't emulate any control instructions yet */
1382 	kvm_s390_set_psw_cc(vcpu, 3);
1383 	return 0;
1384 }
1385 
1386 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1387 {
1388 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1389 	case 0x04:
1390 		return handle_ptff(vcpu);
1391 	case 0x07:
1392 		return handle_sckpf(vcpu);
1393 	default:
1394 		return -EOPNOTSUPP;
1395 	}
1396 }
1397