xref: /openbmc/linux/arch/s390/kvm/priv.c (revision 5927145e)
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 	int rc;
89 	u8 ar;
90 	u64 op2, val;
91 
92 	vcpu->stat.instruction_sck++;
93 
94 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
95 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
96 
97 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
98 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
99 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
100 	rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
101 	if (rc)
102 		return kvm_s390_inject_prog_cond(vcpu, rc);
103 
104 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
105 	kvm_s390_set_tod_clock(vcpu->kvm, val);
106 
107 	kvm_s390_set_psw_cc(vcpu, 0);
108 	return 0;
109 }
110 
111 static int handle_set_prefix(struct kvm_vcpu *vcpu)
112 {
113 	u64 operand2;
114 	u32 address;
115 	int rc;
116 	u8 ar;
117 
118 	vcpu->stat.instruction_spx++;
119 
120 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
121 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
122 
123 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
124 
125 	/* must be word boundary */
126 	if (operand2 & 3)
127 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
128 
129 	/* get the value */
130 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
131 	if (rc)
132 		return kvm_s390_inject_prog_cond(vcpu, rc);
133 
134 	address &= 0x7fffe000u;
135 
136 	/*
137 	 * Make sure the new value is valid memory. We only need to check the
138 	 * first page, since address is 8k aligned and memory pieces are always
139 	 * at least 1MB aligned and have at least a size of 1MB.
140 	 */
141 	if (kvm_is_error_gpa(vcpu->kvm, address))
142 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
143 
144 	kvm_s390_set_prefix(vcpu, address);
145 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
146 	return 0;
147 }
148 
149 static int handle_store_prefix(struct kvm_vcpu *vcpu)
150 {
151 	u64 operand2;
152 	u32 address;
153 	int rc;
154 	u8 ar;
155 
156 	vcpu->stat.instruction_stpx++;
157 
158 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
159 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
160 
161 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
162 
163 	/* must be word boundary */
164 	if (operand2 & 3)
165 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
166 
167 	address = kvm_s390_get_prefix(vcpu);
168 
169 	/* get the value */
170 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
171 	if (rc)
172 		return kvm_s390_inject_prog_cond(vcpu, rc);
173 
174 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
175 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
176 	return 0;
177 }
178 
179 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
180 {
181 	u16 vcpu_id = vcpu->vcpu_id;
182 	u64 ga;
183 	int rc;
184 	u8 ar;
185 
186 	vcpu->stat.instruction_stap++;
187 
188 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
189 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
190 
191 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
192 
193 	if (ga & 1)
194 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
195 
196 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
197 	if (rc)
198 		return kvm_s390_inject_prog_cond(vcpu, rc);
199 
200 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
201 	trace_kvm_s390_handle_stap(vcpu, ga);
202 	return 0;
203 }
204 
205 int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
206 {
207 	int rc = 0;
208 	struct kvm_s390_sie_block *sie_block = vcpu->arch.sie_block;
209 
210 	trace_kvm_s390_skey_related_inst(vcpu);
211 	if (!(sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)) &&
212 	    !kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
213 		return rc;
214 
215 	rc = s390_enable_skey();
216 	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
217 	if (!rc) {
218 		if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
219 			kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS);
220 		else
221 			sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE |
222 					     ICTL_RRBE);
223 	}
224 	return rc;
225 }
226 
227 static int try_handle_skey(struct kvm_vcpu *vcpu)
228 {
229 	int rc;
230 
231 	rc = kvm_s390_skey_check_enable(vcpu);
232 	if (rc)
233 		return rc;
234 	if (sclp.has_skey) {
235 		/* with storage-key facility, SIE interprets it for us */
236 		kvm_s390_retry_instr(vcpu);
237 		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
238 		return -EAGAIN;
239 	}
240 	return 0;
241 }
242 
243 static int handle_iske(struct kvm_vcpu *vcpu)
244 {
245 	unsigned long addr;
246 	unsigned char key;
247 	int reg1, reg2;
248 	int rc;
249 
250 	vcpu->stat.instruction_iske++;
251 
252 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
253 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
254 
255 	rc = try_handle_skey(vcpu);
256 	if (rc)
257 		return rc != -EAGAIN ? rc : 0;
258 
259 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
260 
261 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
262 	addr = kvm_s390_logical_to_effective(vcpu, addr);
263 	addr = kvm_s390_real_to_abs(vcpu, addr);
264 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
265 	if (kvm_is_error_hva(addr))
266 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
267 
268 	down_read(&current->mm->mmap_sem);
269 	rc = get_guest_storage_key(current->mm, addr, &key);
270 	up_read(&current->mm->mmap_sem);
271 	if (rc)
272 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
273 	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
274 	vcpu->run->s.regs.gprs[reg1] |= key;
275 	return 0;
276 }
277 
278 static int handle_rrbe(struct kvm_vcpu *vcpu)
279 {
280 	unsigned long addr;
281 	int reg1, reg2;
282 	int rc;
283 
284 	vcpu->stat.instruction_rrbe++;
285 
286 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
287 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
288 
289 	rc = try_handle_skey(vcpu);
290 	if (rc)
291 		return rc != -EAGAIN ? rc : 0;
292 
293 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
294 
295 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
296 	addr = kvm_s390_logical_to_effective(vcpu, addr);
297 	addr = kvm_s390_real_to_abs(vcpu, addr);
298 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
299 	if (kvm_is_error_hva(addr))
300 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
301 
302 	down_read(&current->mm->mmap_sem);
303 	rc = reset_guest_reference_bit(current->mm, addr);
304 	up_read(&current->mm->mmap_sem);
305 	if (rc < 0)
306 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
307 
308 	kvm_s390_set_psw_cc(vcpu, rc);
309 	return 0;
310 }
311 
312 #define SSKE_NQ 0x8
313 #define SSKE_MR 0x4
314 #define SSKE_MC 0x2
315 #define SSKE_MB 0x1
316 static int handle_sske(struct kvm_vcpu *vcpu)
317 {
318 	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
319 	unsigned long start, end;
320 	unsigned char key, oldkey;
321 	int reg1, reg2;
322 	int rc;
323 
324 	vcpu->stat.instruction_sske++;
325 
326 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
327 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
328 
329 	rc = try_handle_skey(vcpu);
330 	if (rc)
331 		return rc != -EAGAIN ? rc : 0;
332 
333 	if (!test_kvm_facility(vcpu->kvm, 8))
334 		m3 &= ~SSKE_MB;
335 	if (!test_kvm_facility(vcpu->kvm, 10))
336 		m3 &= ~(SSKE_MC | SSKE_MR);
337 	if (!test_kvm_facility(vcpu->kvm, 14))
338 		m3 &= ~SSKE_NQ;
339 
340 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
341 
342 	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
343 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
344 	start = kvm_s390_logical_to_effective(vcpu, start);
345 	if (m3 & SSKE_MB) {
346 		/* start already designates an absolute address */
347 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
348 	} else {
349 		start = kvm_s390_real_to_abs(vcpu, start);
350 		end = start + PAGE_SIZE;
351 	}
352 
353 	while (start != end) {
354 		unsigned long addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
355 
356 		if (kvm_is_error_hva(addr))
357 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
358 
359 		down_read(&current->mm->mmap_sem);
360 		rc = cond_set_guest_storage_key(current->mm, addr, key, &oldkey,
361 						m3 & SSKE_NQ, m3 & SSKE_MR,
362 						m3 & SSKE_MC);
363 		up_read(&current->mm->mmap_sem);
364 		if (rc < 0)
365 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
366 		start += PAGE_SIZE;
367 	}
368 
369 	if (m3 & (SSKE_MC | SSKE_MR)) {
370 		if (m3 & SSKE_MB) {
371 			/* skey in reg1 is unpredictable */
372 			kvm_s390_set_psw_cc(vcpu, 3);
373 		} else {
374 			kvm_s390_set_psw_cc(vcpu, rc);
375 			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
376 			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
377 		}
378 	}
379 	if (m3 & SSKE_MB) {
380 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
381 			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
382 		else
383 			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
384 		end = kvm_s390_logical_to_effective(vcpu, end);
385 		vcpu->run->s.regs.gprs[reg2] |= end;
386 	}
387 	return 0;
388 }
389 
390 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
391 {
392 	vcpu->stat.instruction_ipte_interlock++;
393 	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
394 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
395 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
396 	kvm_s390_retry_instr(vcpu);
397 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
398 	return 0;
399 }
400 
401 static int handle_test_block(struct kvm_vcpu *vcpu)
402 {
403 	gpa_t addr;
404 	int reg2;
405 
406 	vcpu->stat.instruction_tb++;
407 
408 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
409 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
410 
411 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
412 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
413 	addr = kvm_s390_logical_to_effective(vcpu, addr);
414 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
415 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
416 	addr = kvm_s390_real_to_abs(vcpu, addr);
417 
418 	if (kvm_is_error_gpa(vcpu->kvm, addr))
419 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
420 	/*
421 	 * We don't expect errors on modern systems, and do not care
422 	 * about storage keys (yet), so let's just clear the page.
423 	 */
424 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
425 		return -EFAULT;
426 	kvm_s390_set_psw_cc(vcpu, 0);
427 	vcpu->run->s.regs.gprs[0] = 0;
428 	return 0;
429 }
430 
431 static int handle_tpi(struct kvm_vcpu *vcpu)
432 {
433 	struct kvm_s390_interrupt_info *inti;
434 	unsigned long len;
435 	u32 tpi_data[3];
436 	int rc;
437 	u64 addr;
438 	u8 ar;
439 
440 	vcpu->stat.instruction_tpi++;
441 
442 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
443 	if (addr & 3)
444 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
445 
446 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
447 	if (!inti) {
448 		kvm_s390_set_psw_cc(vcpu, 0);
449 		return 0;
450 	}
451 
452 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
453 	tpi_data[1] = inti->io.io_int_parm;
454 	tpi_data[2] = inti->io.io_int_word;
455 	if (addr) {
456 		/*
457 		 * Store the two-word I/O interruption code into the
458 		 * provided area.
459 		 */
460 		len = sizeof(tpi_data) - 4;
461 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
462 		if (rc) {
463 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
464 			goto reinject_interrupt;
465 		}
466 	} else {
467 		/*
468 		 * Store the three-word I/O interruption code into
469 		 * the appropriate lowcore area.
470 		 */
471 		len = sizeof(tpi_data);
472 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
473 			/* failed writes to the low core are not recoverable */
474 			rc = -EFAULT;
475 			goto reinject_interrupt;
476 		}
477 	}
478 
479 	/* irq was successfully handed to the guest */
480 	kfree(inti);
481 	kvm_s390_set_psw_cc(vcpu, 1);
482 	return 0;
483 reinject_interrupt:
484 	/*
485 	 * If we encounter a problem storing the interruption code, the
486 	 * instruction is suppressed from the guest's view: reinject the
487 	 * interrupt.
488 	 */
489 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
490 		kfree(inti);
491 		rc = -EFAULT;
492 	}
493 	/* don't set the cc, a pgm irq was injected or we drop to user space */
494 	return rc ? -EFAULT : 0;
495 }
496 
497 static int handle_tsch(struct kvm_vcpu *vcpu)
498 {
499 	struct kvm_s390_interrupt_info *inti = NULL;
500 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
501 
502 	vcpu->stat.instruction_tsch++;
503 
504 	/* a valid schid has at least one bit set */
505 	if (vcpu->run->s.regs.gprs[1])
506 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
507 					   vcpu->run->s.regs.gprs[1]);
508 
509 	/*
510 	 * Prepare exit to userspace.
511 	 * We indicate whether we dequeued a pending I/O interrupt
512 	 * so that userspace can re-inject it if the instruction gets
513 	 * a program check. While this may re-order the pending I/O
514 	 * interrupts, this is no problem since the priority is kept
515 	 * intact.
516 	 */
517 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
518 	vcpu->run->s390_tsch.dequeued = !!inti;
519 	if (inti) {
520 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
521 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
522 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
523 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
524 	}
525 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
526 	kfree(inti);
527 	return -EREMOTE;
528 }
529 
530 static int handle_io_inst(struct kvm_vcpu *vcpu)
531 {
532 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
533 
534 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
535 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
536 
537 	if (vcpu->kvm->arch.css_support) {
538 		/*
539 		 * Most I/O instructions will be handled by userspace.
540 		 * Exceptions are tpi and the interrupt portion of tsch.
541 		 */
542 		if (vcpu->arch.sie_block->ipa == 0xb236)
543 			return handle_tpi(vcpu);
544 		if (vcpu->arch.sie_block->ipa == 0xb235)
545 			return handle_tsch(vcpu);
546 		/* Handle in userspace. */
547 		vcpu->stat.instruction_io_other++;
548 		return -EOPNOTSUPP;
549 	} else {
550 		/*
551 		 * Set condition code 3 to stop the guest from issuing channel
552 		 * I/O instructions.
553 		 */
554 		kvm_s390_set_psw_cc(vcpu, 3);
555 		return 0;
556 	}
557 }
558 
559 static int handle_stfl(struct kvm_vcpu *vcpu)
560 {
561 	int rc;
562 	unsigned int fac;
563 
564 	vcpu->stat.instruction_stfl++;
565 
566 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
567 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
568 
569 	/*
570 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
571 	 * into a u32 memory representation. They will remain bits 0-31.
572 	 */
573 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
574 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
575 			    &fac, sizeof(fac));
576 	if (rc)
577 		return rc;
578 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
579 	trace_kvm_s390_handle_stfl(vcpu, fac);
580 	return 0;
581 }
582 
583 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
584 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
585 #define PSW_ADDR_24 0x0000000000ffffffUL
586 #define PSW_ADDR_31 0x000000007fffffffUL
587 
588 int is_valid_psw(psw_t *psw)
589 {
590 	if (psw->mask & PSW_MASK_UNASSIGNED)
591 		return 0;
592 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
593 		if (psw->addr & ~PSW_ADDR_31)
594 			return 0;
595 	}
596 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
597 		return 0;
598 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
599 		return 0;
600 	if (psw->addr & 1)
601 		return 0;
602 	return 1;
603 }
604 
605 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
606 {
607 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
608 	psw_compat_t new_psw;
609 	u64 addr;
610 	int rc;
611 	u8 ar;
612 
613 	vcpu->stat.instruction_lpsw++;
614 
615 	if (gpsw->mask & PSW_MASK_PSTATE)
616 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
617 
618 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
619 	if (addr & 7)
620 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
621 
622 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
623 	if (rc)
624 		return kvm_s390_inject_prog_cond(vcpu, rc);
625 	if (!(new_psw.mask & PSW32_MASK_BASE))
626 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
627 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
628 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
629 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
630 	if (!is_valid_psw(gpsw))
631 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
632 	return 0;
633 }
634 
635 static int handle_lpswe(struct kvm_vcpu *vcpu)
636 {
637 	psw_t new_psw;
638 	u64 addr;
639 	int rc;
640 	u8 ar;
641 
642 	vcpu->stat.instruction_lpswe++;
643 
644 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
645 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
646 
647 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
648 	if (addr & 7)
649 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
650 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
651 	if (rc)
652 		return kvm_s390_inject_prog_cond(vcpu, rc);
653 	vcpu->arch.sie_block->gpsw = new_psw;
654 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
655 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
656 	return 0;
657 }
658 
659 static int handle_stidp(struct kvm_vcpu *vcpu)
660 {
661 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
662 	u64 operand2;
663 	int rc;
664 	u8 ar;
665 
666 	vcpu->stat.instruction_stidp++;
667 
668 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
669 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
670 
671 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
672 
673 	if (operand2 & 7)
674 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
675 
676 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
677 	if (rc)
678 		return kvm_s390_inject_prog_cond(vcpu, rc);
679 
680 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
681 	return 0;
682 }
683 
684 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
685 {
686 	int cpus = 0;
687 	int n;
688 
689 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
690 
691 	/* deal with other level 3 hypervisors */
692 	if (stsi(mem, 3, 2, 2))
693 		mem->count = 0;
694 	if (mem->count < 8)
695 		mem->count++;
696 	for (n = mem->count - 1; n > 0 ; n--)
697 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
698 
699 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
700 	mem->vm[0].cpus_total = cpus;
701 	mem->vm[0].cpus_configured = cpus;
702 	mem->vm[0].cpus_standby = 0;
703 	mem->vm[0].cpus_reserved = 0;
704 	mem->vm[0].caf = 1000;
705 	memcpy(mem->vm[0].name, "KVMguest", 8);
706 	ASCEBC(mem->vm[0].name, 8);
707 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
708 	ASCEBC(mem->vm[0].cpi, 16);
709 }
710 
711 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
712 				 u8 fc, u8 sel1, u16 sel2)
713 {
714 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
715 	vcpu->run->s390_stsi.addr = addr;
716 	vcpu->run->s390_stsi.ar = ar;
717 	vcpu->run->s390_stsi.fc = fc;
718 	vcpu->run->s390_stsi.sel1 = sel1;
719 	vcpu->run->s390_stsi.sel2 = sel2;
720 }
721 
722 static int handle_stsi(struct kvm_vcpu *vcpu)
723 {
724 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
725 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
726 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
727 	unsigned long mem = 0;
728 	u64 operand2;
729 	int rc = 0;
730 	u8 ar;
731 
732 	vcpu->stat.instruction_stsi++;
733 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
734 
735 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
736 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
737 
738 	if (fc > 3) {
739 		kvm_s390_set_psw_cc(vcpu, 3);
740 		return 0;
741 	}
742 
743 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
744 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
745 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
746 
747 	if (fc == 0) {
748 		vcpu->run->s.regs.gprs[0] = 3 << 28;
749 		kvm_s390_set_psw_cc(vcpu, 0);
750 		return 0;
751 	}
752 
753 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
754 
755 	if (operand2 & 0xfff)
756 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
757 
758 	switch (fc) {
759 	case 1: /* same handling for 1 and 2 */
760 	case 2:
761 		mem = get_zeroed_page(GFP_KERNEL);
762 		if (!mem)
763 			goto out_no_data;
764 		if (stsi((void *) mem, fc, sel1, sel2))
765 			goto out_no_data;
766 		break;
767 	case 3:
768 		if (sel1 != 2 || sel2 != 2)
769 			goto out_no_data;
770 		mem = get_zeroed_page(GFP_KERNEL);
771 		if (!mem)
772 			goto out_no_data;
773 		handle_stsi_3_2_2(vcpu, (void *) mem);
774 		break;
775 	}
776 
777 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
778 	if (rc) {
779 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
780 		goto out;
781 	}
782 	if (vcpu->kvm->arch.user_stsi) {
783 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
784 		rc = -EREMOTE;
785 	}
786 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
787 	free_page(mem);
788 	kvm_s390_set_psw_cc(vcpu, 0);
789 	vcpu->run->s.regs.gprs[0] = 0;
790 	return rc;
791 out_no_data:
792 	kvm_s390_set_psw_cc(vcpu, 3);
793 out:
794 	free_page(mem);
795 	return rc;
796 }
797 
798 static const intercept_handler_t b2_handlers[256] = {
799 	[0x02] = handle_stidp,
800 	[0x04] = handle_set_clock,
801 	[0x10] = handle_set_prefix,
802 	[0x11] = handle_store_prefix,
803 	[0x12] = handle_store_cpu_address,
804 	[0x14] = kvm_s390_handle_vsie,
805 	[0x21] = handle_ipte_interlock,
806 	[0x29] = handle_iske,
807 	[0x2a] = handle_rrbe,
808 	[0x2b] = handle_sske,
809 	[0x2c] = handle_test_block,
810 	[0x30] = handle_io_inst,
811 	[0x31] = handle_io_inst,
812 	[0x32] = handle_io_inst,
813 	[0x33] = handle_io_inst,
814 	[0x34] = handle_io_inst,
815 	[0x35] = handle_io_inst,
816 	[0x36] = handle_io_inst,
817 	[0x37] = handle_io_inst,
818 	[0x38] = handle_io_inst,
819 	[0x39] = handle_io_inst,
820 	[0x3a] = handle_io_inst,
821 	[0x3b] = handle_io_inst,
822 	[0x3c] = handle_io_inst,
823 	[0x50] = handle_ipte_interlock,
824 	[0x56] = handle_sthyi,
825 	[0x5f] = handle_io_inst,
826 	[0x74] = handle_io_inst,
827 	[0x76] = handle_io_inst,
828 	[0x7d] = handle_stsi,
829 	[0xb1] = handle_stfl,
830 	[0xb2] = handle_lpswe,
831 };
832 
833 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
834 {
835 	intercept_handler_t handler;
836 
837 	/*
838 	 * A lot of B2 instructions are priviledged. Here we check for
839 	 * the privileged ones, that we can handle in the kernel.
840 	 * Anything else goes to userspace.
841 	 */
842 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
843 	if (handler)
844 		return handler(vcpu);
845 
846 	return -EOPNOTSUPP;
847 }
848 
849 static int handle_epsw(struct kvm_vcpu *vcpu)
850 {
851 	int reg1, reg2;
852 
853 	vcpu->stat.instruction_epsw++;
854 
855 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
856 
857 	/* This basically extracts the mask half of the psw. */
858 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
859 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
860 	if (reg2) {
861 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
862 		vcpu->run->s.regs.gprs[reg2] |=
863 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
864 	}
865 	return 0;
866 }
867 
868 #define PFMF_RESERVED   0xfffc0101UL
869 #define PFMF_SK         0x00020000UL
870 #define PFMF_CF         0x00010000UL
871 #define PFMF_UI         0x00008000UL
872 #define PFMF_FSC        0x00007000UL
873 #define PFMF_NQ         0x00000800UL
874 #define PFMF_MR         0x00000400UL
875 #define PFMF_MC         0x00000200UL
876 #define PFMF_KEY        0x000000feUL
877 
878 static int handle_pfmf(struct kvm_vcpu *vcpu)
879 {
880 	bool mr = false, mc = false, nq;
881 	int reg1, reg2;
882 	unsigned long start, end;
883 	unsigned char key;
884 
885 	vcpu->stat.instruction_pfmf++;
886 
887 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
888 
889 	if (!test_kvm_facility(vcpu->kvm, 8))
890 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
891 
892 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
893 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
894 
895 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
896 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
897 
898 	/* Only provide non-quiescing support if enabled for the guest */
899 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
900 	    !test_kvm_facility(vcpu->kvm, 14))
901 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
902 
903 	/* Only provide conditional-SSKE support if enabled for the guest */
904 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
905 	    test_kvm_facility(vcpu->kvm, 10)) {
906 		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
907 		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
908 	}
909 
910 	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
911 	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
912 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
913 	start = kvm_s390_logical_to_effective(vcpu, start);
914 
915 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
916 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
917 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
918 	}
919 
920 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
921 	case 0x00000000:
922 		/* only 4k frames specify a real address */
923 		start = kvm_s390_real_to_abs(vcpu, start);
924 		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
925 		break;
926 	case 0x00001000:
927 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
928 		break;
929 	case 0x00002000:
930 		/* only support 2G frame size if EDAT2 is available and we are
931 		   not in 24-bit addressing mode */
932 		if (!test_kvm_facility(vcpu->kvm, 78) ||
933 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
934 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
935 		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
936 		break;
937 	default:
938 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
939 	}
940 
941 	while (start != end) {
942 		unsigned long useraddr;
943 
944 		/* Translate guest address to host address */
945 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
946 		if (kvm_is_error_hva(useraddr))
947 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
948 
949 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
950 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
951 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
952 		}
953 
954 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
955 			int rc = kvm_s390_skey_check_enable(vcpu);
956 
957 			if (rc)
958 				return rc;
959 			down_read(&current->mm->mmap_sem);
960 			rc = cond_set_guest_storage_key(current->mm, useraddr,
961 							key, NULL, nq, mr, mc);
962 			up_read(&current->mm->mmap_sem);
963 			if (rc < 0)
964 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
965 		}
966 
967 		start += PAGE_SIZE;
968 	}
969 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
970 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
971 			vcpu->run->s.regs.gprs[reg2] = end;
972 		} else {
973 			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
974 			end = kvm_s390_logical_to_effective(vcpu, end);
975 			vcpu->run->s.regs.gprs[reg2] |= end;
976 		}
977 	}
978 	return 0;
979 }
980 
981 static inline int do_essa(struct kvm_vcpu *vcpu, const int orc)
982 {
983 	struct kvm_s390_migration_state *ms = vcpu->kvm->arch.migration_state;
984 	int r1, r2, nappended, entries;
985 	unsigned long gfn, hva, res, pgstev, ptev;
986 	unsigned long *cbrlo;
987 
988 	/*
989 	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
990 	 * machine check here we either handle it or crash
991 	 */
992 
993 	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
994 	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
995 	hva = gfn_to_hva(vcpu->kvm, gfn);
996 	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
997 
998 	if (kvm_is_error_hva(hva))
999 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1000 
1001 	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
1002 	if (nappended < 0) {
1003 		res = orc ? 0x10 : 0;
1004 		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
1005 		return 0;
1006 	}
1007 	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
1008 	/*
1009 	 * Set the block-content state part of the result. 0 means resident, so
1010 	 * nothing to do if the page is valid. 2 is for preserved pages
1011 	 * (non-present and non-zero), and 3 for zero pages (non-present and
1012 	 * zero).
1013 	 */
1014 	if (ptev & _PAGE_INVALID) {
1015 		res |= 2;
1016 		if (pgstev & _PGSTE_GPS_ZERO)
1017 			res |= 1;
1018 	}
1019 	if (pgstev & _PGSTE_GPS_NODAT)
1020 		res |= 0x20;
1021 	vcpu->run->s.regs.gprs[r1] = res;
1022 	/*
1023 	 * It is possible that all the normal 511 slots were full, in which case
1024 	 * we will now write in the 512th slot, which is reserved for host use.
1025 	 * In both cases we let the normal essa handling code process all the
1026 	 * slots, including the reserved one, if needed.
1027 	 */
1028 	if (nappended > 0) {
1029 		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1030 		cbrlo[entries] = gfn << PAGE_SHIFT;
1031 	}
1032 
1033 	if (orc && gfn < ms->bitmap_size) {
1034 		/* increment only if we are really flipping the bit to 1 */
1035 		if (!test_and_set_bit(gfn, ms->pgste_bitmap))
1036 			atomic64_inc(&ms->dirty_pages);
1037 	}
1038 
1039 	return nappended;
1040 }
1041 
1042 static int handle_essa(struct kvm_vcpu *vcpu)
1043 {
1044 	/* entries expected to be 1FF */
1045 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1046 	unsigned long *cbrlo;
1047 	struct gmap *gmap;
1048 	int i, orc;
1049 
1050 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1051 	gmap = vcpu->arch.gmap;
1052 	vcpu->stat.instruction_essa++;
1053 	if (!vcpu->kvm->arch.use_cmma)
1054 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1055 
1056 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1057 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1058 	/* Check for invalid operation request code */
1059 	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1060 	/* ORCs 0-6 are always valid */
1061 	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1062 						: ESSA_SET_STABLE_IF_RESIDENT))
1063 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1064 
1065 	if (likely(!vcpu->kvm->arch.migration_state)) {
1066 		/*
1067 		 * CMMA is enabled in the KVM settings, but is disabled in
1068 		 * the SIE block and in the mm_context, and we are not doing
1069 		 * a migration. Enable CMMA in the mm_context.
1070 		 * Since we need to take a write lock to write to the context
1071 		 * to avoid races with storage keys handling, we check if the
1072 		 * value really needs to be written to; if the value is
1073 		 * already correct, we do nothing and avoid the lock.
1074 		 */
1075 		if (vcpu->kvm->mm->context.use_cmma == 0) {
1076 			down_write(&vcpu->kvm->mm->mmap_sem);
1077 			vcpu->kvm->mm->context.use_cmma = 1;
1078 			up_write(&vcpu->kvm->mm->mmap_sem);
1079 		}
1080 		/*
1081 		 * If we are here, we are supposed to have CMMA enabled in
1082 		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1083 		 * while the context use_cmma flag is per process.
1084 		 * It's possible that the context flag is enabled and the
1085 		 * SIE flag is not, so we set the flag always; if it was
1086 		 * already set, nothing changes, otherwise we enable it
1087 		 * on this CPU too.
1088 		 */
1089 		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1090 		/* Retry the ESSA instruction */
1091 		kvm_s390_retry_instr(vcpu);
1092 	} else {
1093 		/* Account for the possible extra cbrl entry */
1094 		i = do_essa(vcpu, orc);
1095 		if (i < 0)
1096 			return i;
1097 		entries += i;
1098 	}
1099 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1100 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1101 	down_read(&gmap->mm->mmap_sem);
1102 	for (i = 0; i < entries; ++i)
1103 		__gmap_zap(gmap, cbrlo[i]);
1104 	up_read(&gmap->mm->mmap_sem);
1105 	return 0;
1106 }
1107 
1108 static const intercept_handler_t b9_handlers[256] = {
1109 	[0x8a] = handle_ipte_interlock,
1110 	[0x8d] = handle_epsw,
1111 	[0x8e] = handle_ipte_interlock,
1112 	[0x8f] = handle_ipte_interlock,
1113 	[0xab] = handle_essa,
1114 	[0xaf] = handle_pfmf,
1115 };
1116 
1117 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1118 {
1119 	intercept_handler_t handler;
1120 
1121 	/* This is handled just as for the B2 instructions. */
1122 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1123 	if (handler)
1124 		return handler(vcpu);
1125 
1126 	return -EOPNOTSUPP;
1127 }
1128 
1129 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1130 {
1131 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1132 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1133 	int reg, rc, nr_regs;
1134 	u32 ctl_array[16];
1135 	u64 ga;
1136 	u8 ar;
1137 
1138 	vcpu->stat.instruction_lctl++;
1139 
1140 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1141 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1142 
1143 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1144 
1145 	if (ga & 3)
1146 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1147 
1148 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1149 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1150 
1151 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1152 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1153 	if (rc)
1154 		return kvm_s390_inject_prog_cond(vcpu, rc);
1155 	reg = reg1;
1156 	nr_regs = 0;
1157 	do {
1158 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1159 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1160 		if (reg == reg3)
1161 			break;
1162 		reg = (reg + 1) % 16;
1163 	} while (1);
1164 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1165 	return 0;
1166 }
1167 
1168 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1169 {
1170 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1171 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1172 	int reg, rc, nr_regs;
1173 	u32 ctl_array[16];
1174 	u64 ga;
1175 	u8 ar;
1176 
1177 	vcpu->stat.instruction_stctl++;
1178 
1179 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1180 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1181 
1182 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1183 
1184 	if (ga & 3)
1185 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1186 
1187 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1188 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1189 
1190 	reg = reg1;
1191 	nr_regs = 0;
1192 	do {
1193 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1194 		if (reg == reg3)
1195 			break;
1196 		reg = (reg + 1) % 16;
1197 	} while (1);
1198 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1199 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1200 }
1201 
1202 static int handle_lctlg(struct kvm_vcpu *vcpu)
1203 {
1204 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1205 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1206 	int reg, rc, nr_regs;
1207 	u64 ctl_array[16];
1208 	u64 ga;
1209 	u8 ar;
1210 
1211 	vcpu->stat.instruction_lctlg++;
1212 
1213 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1214 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1215 
1216 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1217 
1218 	if (ga & 7)
1219 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1220 
1221 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1222 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1223 
1224 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1225 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1226 	if (rc)
1227 		return kvm_s390_inject_prog_cond(vcpu, rc);
1228 	reg = reg1;
1229 	nr_regs = 0;
1230 	do {
1231 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1232 		if (reg == reg3)
1233 			break;
1234 		reg = (reg + 1) % 16;
1235 	} while (1);
1236 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1237 	return 0;
1238 }
1239 
1240 static int handle_stctg(struct kvm_vcpu *vcpu)
1241 {
1242 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1243 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1244 	int reg, rc, nr_regs;
1245 	u64 ctl_array[16];
1246 	u64 ga;
1247 	u8 ar;
1248 
1249 	vcpu->stat.instruction_stctg++;
1250 
1251 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1252 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1253 
1254 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1255 
1256 	if (ga & 7)
1257 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1258 
1259 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1260 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1261 
1262 	reg = reg1;
1263 	nr_regs = 0;
1264 	do {
1265 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1266 		if (reg == reg3)
1267 			break;
1268 		reg = (reg + 1) % 16;
1269 	} while (1);
1270 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1271 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1272 }
1273 
1274 static const intercept_handler_t eb_handlers[256] = {
1275 	[0x2f] = handle_lctlg,
1276 	[0x25] = handle_stctg,
1277 	[0x60] = handle_ri,
1278 	[0x61] = handle_ri,
1279 	[0x62] = handle_ri,
1280 };
1281 
1282 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1283 {
1284 	intercept_handler_t handler;
1285 
1286 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
1287 	if (handler)
1288 		return handler(vcpu);
1289 	return -EOPNOTSUPP;
1290 }
1291 
1292 static int handle_tprot(struct kvm_vcpu *vcpu)
1293 {
1294 	u64 address1, address2;
1295 	unsigned long hva, gpa;
1296 	int ret = 0, cc = 0;
1297 	bool writable;
1298 	u8 ar;
1299 
1300 	vcpu->stat.instruction_tprot++;
1301 
1302 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1303 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1304 
1305 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1306 
1307 	/* we only handle the Linux memory detection case:
1308 	 * access key == 0
1309 	 * everything else goes to userspace. */
1310 	if (address2 & 0xf0)
1311 		return -EOPNOTSUPP;
1312 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1313 		ipte_lock(vcpu);
1314 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1315 	if (ret == PGM_PROTECTION) {
1316 		/* Write protected? Try again with read-only... */
1317 		cc = 1;
1318 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
1319 					      GACC_FETCH);
1320 	}
1321 	if (ret) {
1322 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1323 			ret = kvm_s390_inject_program_int(vcpu, ret);
1324 		} else if (ret > 0) {
1325 			/* Translation not available */
1326 			kvm_s390_set_psw_cc(vcpu, 3);
1327 			ret = 0;
1328 		}
1329 		goto out_unlock;
1330 	}
1331 
1332 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1333 	if (kvm_is_error_hva(hva)) {
1334 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1335 	} else {
1336 		if (!writable)
1337 			cc = 1;		/* Write not permitted ==> read-only */
1338 		kvm_s390_set_psw_cc(vcpu, cc);
1339 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1340 	}
1341 out_unlock:
1342 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1343 		ipte_unlock(vcpu);
1344 	return ret;
1345 }
1346 
1347 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1348 {
1349 	/* For e5xx... instructions we only handle TPROT */
1350 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1351 		return handle_tprot(vcpu);
1352 	return -EOPNOTSUPP;
1353 }
1354 
1355 static int handle_sckpf(struct kvm_vcpu *vcpu)
1356 {
1357 	u32 value;
1358 
1359 	vcpu->stat.instruction_sckpf++;
1360 
1361 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1362 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1363 
1364 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1365 		return kvm_s390_inject_program_int(vcpu,
1366 						   PGM_SPECIFICATION);
1367 
1368 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1369 	vcpu->arch.sie_block->todpr = value;
1370 
1371 	return 0;
1372 }
1373 
1374 static int handle_ptff(struct kvm_vcpu *vcpu)
1375 {
1376 	vcpu->stat.instruction_ptff++;
1377 
1378 	/* we don't emulate any control instructions yet */
1379 	kvm_s390_set_psw_cc(vcpu, 3);
1380 	return 0;
1381 }
1382 
1383 static const intercept_handler_t x01_handlers[256] = {
1384 	[0x04] = handle_ptff,
1385 	[0x07] = handle_sckpf,
1386 };
1387 
1388 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1389 {
1390 	intercept_handler_t handler;
1391 
1392 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1393 	if (handler)
1394 		return handler(vcpu);
1395 	return -EOPNOTSUPP;
1396 }
1397