xref: /linux/arch/s390/kvm/priv.c (revision 2fe05e1139a555ae91f00a812cb9520e7d3022ab)
1 /*
2  * handling privileged instructions
3  *
4  * Copyright IBM Corp. 2008, 2013
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License (version 2 only)
8  * as published by the Free Software Foundation.
9  *
10  *    Author(s): Carsten Otte <cotte@de.ibm.com>
11  *               Christian Borntraeger <borntraeger@de.ibm.com>
12  */
13 
14 #include <linux/kvm.h>
15 #include <linux/gfp.h>
16 #include <linux/errno.h>
17 #include <linux/compat.h>
18 #include <linux/mm_types.h>
19 
20 #include <asm/asm-offsets.h>
21 #include <asm/facility.h>
22 #include <asm/current.h>
23 #include <asm/debug.h>
24 #include <asm/ebcdic.h>
25 #include <asm/sysinfo.h>
26 #include <asm/pgtable.h>
27 #include <asm/page-states.h>
28 #include <asm/pgalloc.h>
29 #include <asm/gmap.h>
30 #include <asm/io.h>
31 #include <asm/ptrace.h>
32 #include <asm/compat.h>
33 #include <asm/sclp.h>
34 #include "gaccess.h"
35 #include "kvm-s390.h"
36 #include "trace.h"
37 
38 static int handle_ri(struct kvm_vcpu *vcpu)
39 {
40 	if (test_kvm_facility(vcpu->kvm, 64)) {
41 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
42 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
43 		kvm_s390_retry_instr(vcpu);
44 		return 0;
45 	} else
46 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
47 }
48 
49 int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
50 {
51 	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
52 		return handle_ri(vcpu);
53 	else
54 		return -EOPNOTSUPP;
55 }
56 
57 static int handle_gs(struct kvm_vcpu *vcpu)
58 {
59 	if (test_kvm_facility(vcpu->kvm, 133)) {
60 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (lazy)");
61 		preempt_disable();
62 		__ctl_set_bit(2, 4);
63 		current->thread.gs_cb = (struct gs_cb *)&vcpu->run->s.regs.gscb;
64 		restore_gs_cb(current->thread.gs_cb);
65 		preempt_enable();
66 		vcpu->arch.sie_block->ecb |= ECB_GS;
67 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
68 		vcpu->arch.gs_enabled = 1;
69 		kvm_s390_retry_instr(vcpu);
70 		return 0;
71 	} else
72 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
73 }
74 
75 int kvm_s390_handle_e3(struct kvm_vcpu *vcpu)
76 {
77 	int code = vcpu->arch.sie_block->ipb & 0xff;
78 
79 	if (code == 0x49 || code == 0x4d)
80 		return handle_gs(vcpu);
81 	else
82 		return -EOPNOTSUPP;
83 }
84 /* Handle SCK (SET CLOCK) interception */
85 static int handle_set_clock(struct kvm_vcpu *vcpu)
86 {
87 	int rc;
88 	u8 ar;
89 	u64 op2, val;
90 
91 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
92 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
93 
94 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
95 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
96 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
97 	rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
98 	if (rc)
99 		return kvm_s390_inject_prog_cond(vcpu, rc);
100 
101 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
102 	kvm_s390_set_tod_clock(vcpu->kvm, val);
103 
104 	kvm_s390_set_psw_cc(vcpu, 0);
105 	return 0;
106 }
107 
108 static int handle_set_prefix(struct kvm_vcpu *vcpu)
109 {
110 	u64 operand2;
111 	u32 address;
112 	int rc;
113 	u8 ar;
114 
115 	vcpu->stat.instruction_spx++;
116 
117 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
118 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
119 
120 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
121 
122 	/* must be word boundary */
123 	if (operand2 & 3)
124 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
125 
126 	/* get the value */
127 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
128 	if (rc)
129 		return kvm_s390_inject_prog_cond(vcpu, rc);
130 
131 	address &= 0x7fffe000u;
132 
133 	/*
134 	 * Make sure the new value is valid memory. We only need to check the
135 	 * first page, since address is 8k aligned and memory pieces are always
136 	 * at least 1MB aligned and have at least a size of 1MB.
137 	 */
138 	if (kvm_is_error_gpa(vcpu->kvm, address))
139 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
140 
141 	kvm_s390_set_prefix(vcpu, address);
142 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
143 	return 0;
144 }
145 
146 static int handle_store_prefix(struct kvm_vcpu *vcpu)
147 {
148 	u64 operand2;
149 	u32 address;
150 	int rc;
151 	u8 ar;
152 
153 	vcpu->stat.instruction_stpx++;
154 
155 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
156 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
157 
158 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
159 
160 	/* must be word boundary */
161 	if (operand2 & 3)
162 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
163 
164 	address = kvm_s390_get_prefix(vcpu);
165 
166 	/* get the value */
167 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
168 	if (rc)
169 		return kvm_s390_inject_prog_cond(vcpu, rc);
170 
171 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
172 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
173 	return 0;
174 }
175 
176 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
177 {
178 	u16 vcpu_id = vcpu->vcpu_id;
179 	u64 ga;
180 	int rc;
181 	u8 ar;
182 
183 	vcpu->stat.instruction_stap++;
184 
185 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
186 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
187 
188 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
189 
190 	if (ga & 1)
191 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
192 
193 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
194 	if (rc)
195 		return kvm_s390_inject_prog_cond(vcpu, rc);
196 
197 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
198 	trace_kvm_s390_handle_stap(vcpu, ga);
199 	return 0;
200 }
201 
202 int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
203 {
204 	int rc = 0;
205 	struct kvm_s390_sie_block *sie_block = vcpu->arch.sie_block;
206 
207 	trace_kvm_s390_skey_related_inst(vcpu);
208 	if (!(sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)) &&
209 	    !(atomic_read(&sie_block->cpuflags) & CPUSTAT_KSS))
210 		return rc;
211 
212 	rc = s390_enable_skey();
213 	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
214 	if (!rc) {
215 		if (atomic_read(&sie_block->cpuflags) & CPUSTAT_KSS)
216 			atomic_andnot(CPUSTAT_KSS, &sie_block->cpuflags);
217 		else
218 			sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE |
219 					     ICTL_RRBE);
220 	}
221 	return rc;
222 }
223 
224 static int try_handle_skey(struct kvm_vcpu *vcpu)
225 {
226 	int rc;
227 
228 	vcpu->stat.instruction_storage_key++;
229 	rc = kvm_s390_skey_check_enable(vcpu);
230 	if (rc)
231 		return rc;
232 	if (sclp.has_skey) {
233 		/* with storage-key facility, SIE interprets it for us */
234 		kvm_s390_retry_instr(vcpu);
235 		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
236 		return -EAGAIN;
237 	}
238 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
239 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
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 	rc = try_handle_skey(vcpu);
251 	if (rc)
252 		return rc != -EAGAIN ? rc : 0;
253 
254 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
255 
256 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
257 	addr = kvm_s390_logical_to_effective(vcpu, addr);
258 	addr = kvm_s390_real_to_abs(vcpu, addr);
259 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
260 	if (kvm_is_error_hva(addr))
261 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
262 
263 	down_read(&current->mm->mmap_sem);
264 	rc = get_guest_storage_key(current->mm, addr, &key);
265 	up_read(&current->mm->mmap_sem);
266 	if (rc)
267 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
268 	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
269 	vcpu->run->s.regs.gprs[reg1] |= key;
270 	return 0;
271 }
272 
273 static int handle_rrbe(struct kvm_vcpu *vcpu)
274 {
275 	unsigned long addr;
276 	int reg1, reg2;
277 	int rc;
278 
279 	rc = try_handle_skey(vcpu);
280 	if (rc)
281 		return rc != -EAGAIN ? rc : 0;
282 
283 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
284 
285 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
286 	addr = kvm_s390_logical_to_effective(vcpu, addr);
287 	addr = kvm_s390_real_to_abs(vcpu, addr);
288 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
289 	if (kvm_is_error_hva(addr))
290 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
291 
292 	down_read(&current->mm->mmap_sem);
293 	rc = reset_guest_reference_bit(current->mm, addr);
294 	up_read(&current->mm->mmap_sem);
295 	if (rc < 0)
296 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
297 
298 	kvm_s390_set_psw_cc(vcpu, rc);
299 	return 0;
300 }
301 
302 #define SSKE_NQ 0x8
303 #define SSKE_MR 0x4
304 #define SSKE_MC 0x2
305 #define SSKE_MB 0x1
306 static int handle_sske(struct kvm_vcpu *vcpu)
307 {
308 	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
309 	unsigned long start, end;
310 	unsigned char key, oldkey;
311 	int reg1, reg2;
312 	int rc;
313 
314 	rc = try_handle_skey(vcpu);
315 	if (rc)
316 		return rc != -EAGAIN ? rc : 0;
317 
318 	if (!test_kvm_facility(vcpu->kvm, 8))
319 		m3 &= ~SSKE_MB;
320 	if (!test_kvm_facility(vcpu->kvm, 10))
321 		m3 &= ~(SSKE_MC | SSKE_MR);
322 	if (!test_kvm_facility(vcpu->kvm, 14))
323 		m3 &= ~SSKE_NQ;
324 
325 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
326 
327 	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
328 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
329 	start = kvm_s390_logical_to_effective(vcpu, start);
330 	if (m3 & SSKE_MB) {
331 		/* start already designates an absolute address */
332 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
333 	} else {
334 		start = kvm_s390_real_to_abs(vcpu, start);
335 		end = start + PAGE_SIZE;
336 	}
337 
338 	while (start != end) {
339 		unsigned long addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
340 
341 		if (kvm_is_error_hva(addr))
342 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
343 
344 		down_read(&current->mm->mmap_sem);
345 		rc = cond_set_guest_storage_key(current->mm, addr, key, &oldkey,
346 						m3 & SSKE_NQ, m3 & SSKE_MR,
347 						m3 & SSKE_MC);
348 		up_read(&current->mm->mmap_sem);
349 		if (rc < 0)
350 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
351 		start += PAGE_SIZE;
352 	}
353 
354 	if (m3 & (SSKE_MC | SSKE_MR)) {
355 		if (m3 & SSKE_MB) {
356 			/* skey in reg1 is unpredictable */
357 			kvm_s390_set_psw_cc(vcpu, 3);
358 		} else {
359 			kvm_s390_set_psw_cc(vcpu, rc);
360 			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
361 			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
362 		}
363 	}
364 	if (m3 & SSKE_MB) {
365 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
366 			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
367 		else
368 			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
369 		end = kvm_s390_logical_to_effective(vcpu, end);
370 		vcpu->run->s.regs.gprs[reg2] |= end;
371 	}
372 	return 0;
373 }
374 
375 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
376 {
377 	vcpu->stat.instruction_ipte_interlock++;
378 	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
379 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
380 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
381 	kvm_s390_retry_instr(vcpu);
382 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
383 	return 0;
384 }
385 
386 static int handle_test_block(struct kvm_vcpu *vcpu)
387 {
388 	gpa_t addr;
389 	int reg2;
390 
391 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
392 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
393 
394 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
395 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
396 	addr = kvm_s390_logical_to_effective(vcpu, addr);
397 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
398 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
399 	addr = kvm_s390_real_to_abs(vcpu, addr);
400 
401 	if (kvm_is_error_gpa(vcpu->kvm, addr))
402 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
403 	/*
404 	 * We don't expect errors on modern systems, and do not care
405 	 * about storage keys (yet), so let's just clear the page.
406 	 */
407 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
408 		return -EFAULT;
409 	kvm_s390_set_psw_cc(vcpu, 0);
410 	vcpu->run->s.regs.gprs[0] = 0;
411 	return 0;
412 }
413 
414 static int handle_tpi(struct kvm_vcpu *vcpu)
415 {
416 	struct kvm_s390_interrupt_info *inti;
417 	unsigned long len;
418 	u32 tpi_data[3];
419 	int rc;
420 	u64 addr;
421 	u8 ar;
422 
423 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
424 	if (addr & 3)
425 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
426 
427 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
428 	if (!inti) {
429 		kvm_s390_set_psw_cc(vcpu, 0);
430 		return 0;
431 	}
432 
433 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
434 	tpi_data[1] = inti->io.io_int_parm;
435 	tpi_data[2] = inti->io.io_int_word;
436 	if (addr) {
437 		/*
438 		 * Store the two-word I/O interruption code into the
439 		 * provided area.
440 		 */
441 		len = sizeof(tpi_data) - 4;
442 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
443 		if (rc) {
444 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
445 			goto reinject_interrupt;
446 		}
447 	} else {
448 		/*
449 		 * Store the three-word I/O interruption code into
450 		 * the appropriate lowcore area.
451 		 */
452 		len = sizeof(tpi_data);
453 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
454 			/* failed writes to the low core are not recoverable */
455 			rc = -EFAULT;
456 			goto reinject_interrupt;
457 		}
458 	}
459 
460 	/* irq was successfully handed to the guest */
461 	kfree(inti);
462 	kvm_s390_set_psw_cc(vcpu, 1);
463 	return 0;
464 reinject_interrupt:
465 	/*
466 	 * If we encounter a problem storing the interruption code, the
467 	 * instruction is suppressed from the guest's view: reinject the
468 	 * interrupt.
469 	 */
470 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
471 		kfree(inti);
472 		rc = -EFAULT;
473 	}
474 	/* don't set the cc, a pgm irq was injected or we drop to user space */
475 	return rc ? -EFAULT : 0;
476 }
477 
478 static int handle_tsch(struct kvm_vcpu *vcpu)
479 {
480 	struct kvm_s390_interrupt_info *inti = NULL;
481 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
482 
483 	/* a valid schid has at least one bit set */
484 	if (vcpu->run->s.regs.gprs[1])
485 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
486 					   vcpu->run->s.regs.gprs[1]);
487 
488 	/*
489 	 * Prepare exit to userspace.
490 	 * We indicate whether we dequeued a pending I/O interrupt
491 	 * so that userspace can re-inject it if the instruction gets
492 	 * a program check. While this may re-order the pending I/O
493 	 * interrupts, this is no problem since the priority is kept
494 	 * intact.
495 	 */
496 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
497 	vcpu->run->s390_tsch.dequeued = !!inti;
498 	if (inti) {
499 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
500 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
501 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
502 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
503 	}
504 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
505 	kfree(inti);
506 	return -EREMOTE;
507 }
508 
509 static int handle_io_inst(struct kvm_vcpu *vcpu)
510 {
511 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
512 
513 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
514 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
515 
516 	if (vcpu->kvm->arch.css_support) {
517 		/*
518 		 * Most I/O instructions will be handled by userspace.
519 		 * Exceptions are tpi and the interrupt portion of tsch.
520 		 */
521 		if (vcpu->arch.sie_block->ipa == 0xb236)
522 			return handle_tpi(vcpu);
523 		if (vcpu->arch.sie_block->ipa == 0xb235)
524 			return handle_tsch(vcpu);
525 		/* Handle in userspace. */
526 		return -EOPNOTSUPP;
527 	} else {
528 		/*
529 		 * Set condition code 3 to stop the guest from issuing channel
530 		 * I/O instructions.
531 		 */
532 		kvm_s390_set_psw_cc(vcpu, 3);
533 		return 0;
534 	}
535 }
536 
537 static int handle_stfl(struct kvm_vcpu *vcpu)
538 {
539 	int rc;
540 	unsigned int fac;
541 
542 	vcpu->stat.instruction_stfl++;
543 
544 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
545 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
546 
547 	/*
548 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
549 	 * into a u32 memory representation. They will remain bits 0-31.
550 	 */
551 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
552 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
553 			    &fac, sizeof(fac));
554 	if (rc)
555 		return rc;
556 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
557 	trace_kvm_s390_handle_stfl(vcpu, fac);
558 	return 0;
559 }
560 
561 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
562 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
563 #define PSW_ADDR_24 0x0000000000ffffffUL
564 #define PSW_ADDR_31 0x000000007fffffffUL
565 
566 int is_valid_psw(psw_t *psw)
567 {
568 	if (psw->mask & PSW_MASK_UNASSIGNED)
569 		return 0;
570 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
571 		if (psw->addr & ~PSW_ADDR_31)
572 			return 0;
573 	}
574 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
575 		return 0;
576 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
577 		return 0;
578 	if (psw->addr & 1)
579 		return 0;
580 	return 1;
581 }
582 
583 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
584 {
585 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
586 	psw_compat_t new_psw;
587 	u64 addr;
588 	int rc;
589 	u8 ar;
590 
591 	if (gpsw->mask & PSW_MASK_PSTATE)
592 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
593 
594 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
595 	if (addr & 7)
596 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
597 
598 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
599 	if (rc)
600 		return kvm_s390_inject_prog_cond(vcpu, rc);
601 	if (!(new_psw.mask & PSW32_MASK_BASE))
602 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
603 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
604 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
605 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
606 	if (!is_valid_psw(gpsw))
607 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
608 	return 0;
609 }
610 
611 static int handle_lpswe(struct kvm_vcpu *vcpu)
612 {
613 	psw_t new_psw;
614 	u64 addr;
615 	int rc;
616 	u8 ar;
617 
618 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
619 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
620 
621 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
622 	if (addr & 7)
623 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
624 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
625 	if (rc)
626 		return kvm_s390_inject_prog_cond(vcpu, rc);
627 	vcpu->arch.sie_block->gpsw = new_psw;
628 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
629 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
630 	return 0;
631 }
632 
633 static int handle_stidp(struct kvm_vcpu *vcpu)
634 {
635 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
636 	u64 operand2;
637 	int rc;
638 	u8 ar;
639 
640 	vcpu->stat.instruction_stidp++;
641 
642 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
643 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
644 
645 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
646 
647 	if (operand2 & 7)
648 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
649 
650 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
651 	if (rc)
652 		return kvm_s390_inject_prog_cond(vcpu, rc);
653 
654 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
655 	return 0;
656 }
657 
658 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
659 {
660 	int cpus = 0;
661 	int n;
662 
663 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
664 
665 	/* deal with other level 3 hypervisors */
666 	if (stsi(mem, 3, 2, 2))
667 		mem->count = 0;
668 	if (mem->count < 8)
669 		mem->count++;
670 	for (n = mem->count - 1; n > 0 ; n--)
671 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
672 
673 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
674 	mem->vm[0].cpus_total = cpus;
675 	mem->vm[0].cpus_configured = cpus;
676 	mem->vm[0].cpus_standby = 0;
677 	mem->vm[0].cpus_reserved = 0;
678 	mem->vm[0].caf = 1000;
679 	memcpy(mem->vm[0].name, "KVMguest", 8);
680 	ASCEBC(mem->vm[0].name, 8);
681 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
682 	ASCEBC(mem->vm[0].cpi, 16);
683 }
684 
685 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
686 				 u8 fc, u8 sel1, u16 sel2)
687 {
688 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
689 	vcpu->run->s390_stsi.addr = addr;
690 	vcpu->run->s390_stsi.ar = ar;
691 	vcpu->run->s390_stsi.fc = fc;
692 	vcpu->run->s390_stsi.sel1 = sel1;
693 	vcpu->run->s390_stsi.sel2 = sel2;
694 }
695 
696 static int handle_stsi(struct kvm_vcpu *vcpu)
697 {
698 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
699 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
700 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
701 	unsigned long mem = 0;
702 	u64 operand2;
703 	int rc = 0;
704 	u8 ar;
705 
706 	vcpu->stat.instruction_stsi++;
707 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
708 
709 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
710 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
711 
712 	if (fc > 3) {
713 		kvm_s390_set_psw_cc(vcpu, 3);
714 		return 0;
715 	}
716 
717 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
718 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
719 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
720 
721 	if (fc == 0) {
722 		vcpu->run->s.regs.gprs[0] = 3 << 28;
723 		kvm_s390_set_psw_cc(vcpu, 0);
724 		return 0;
725 	}
726 
727 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
728 
729 	if (operand2 & 0xfff)
730 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
731 
732 	switch (fc) {
733 	case 1: /* same handling for 1 and 2 */
734 	case 2:
735 		mem = get_zeroed_page(GFP_KERNEL);
736 		if (!mem)
737 			goto out_no_data;
738 		if (stsi((void *) mem, fc, sel1, sel2))
739 			goto out_no_data;
740 		break;
741 	case 3:
742 		if (sel1 != 2 || sel2 != 2)
743 			goto out_no_data;
744 		mem = get_zeroed_page(GFP_KERNEL);
745 		if (!mem)
746 			goto out_no_data;
747 		handle_stsi_3_2_2(vcpu, (void *) mem);
748 		break;
749 	}
750 
751 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
752 	if (rc) {
753 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
754 		goto out;
755 	}
756 	if (vcpu->kvm->arch.user_stsi) {
757 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
758 		rc = -EREMOTE;
759 	}
760 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
761 	free_page(mem);
762 	kvm_s390_set_psw_cc(vcpu, 0);
763 	vcpu->run->s.regs.gprs[0] = 0;
764 	return rc;
765 out_no_data:
766 	kvm_s390_set_psw_cc(vcpu, 3);
767 out:
768 	free_page(mem);
769 	return rc;
770 }
771 
772 static const intercept_handler_t b2_handlers[256] = {
773 	[0x02] = handle_stidp,
774 	[0x04] = handle_set_clock,
775 	[0x10] = handle_set_prefix,
776 	[0x11] = handle_store_prefix,
777 	[0x12] = handle_store_cpu_address,
778 	[0x14] = kvm_s390_handle_vsie,
779 	[0x21] = handle_ipte_interlock,
780 	[0x29] = handle_iske,
781 	[0x2a] = handle_rrbe,
782 	[0x2b] = handle_sske,
783 	[0x2c] = handle_test_block,
784 	[0x30] = handle_io_inst,
785 	[0x31] = handle_io_inst,
786 	[0x32] = handle_io_inst,
787 	[0x33] = handle_io_inst,
788 	[0x34] = handle_io_inst,
789 	[0x35] = handle_io_inst,
790 	[0x36] = handle_io_inst,
791 	[0x37] = handle_io_inst,
792 	[0x38] = handle_io_inst,
793 	[0x39] = handle_io_inst,
794 	[0x3a] = handle_io_inst,
795 	[0x3b] = handle_io_inst,
796 	[0x3c] = handle_io_inst,
797 	[0x50] = handle_ipte_interlock,
798 	[0x56] = handle_sthyi,
799 	[0x5f] = handle_io_inst,
800 	[0x74] = handle_io_inst,
801 	[0x76] = handle_io_inst,
802 	[0x7d] = handle_stsi,
803 	[0xb1] = handle_stfl,
804 	[0xb2] = handle_lpswe,
805 };
806 
807 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
808 {
809 	intercept_handler_t handler;
810 
811 	/*
812 	 * A lot of B2 instructions are priviledged. Here we check for
813 	 * the privileged ones, that we can handle in the kernel.
814 	 * Anything else goes to userspace.
815 	 */
816 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
817 	if (handler)
818 		return handler(vcpu);
819 
820 	return -EOPNOTSUPP;
821 }
822 
823 static int handle_epsw(struct kvm_vcpu *vcpu)
824 {
825 	int reg1, reg2;
826 
827 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
828 
829 	/* This basically extracts the mask half of the psw. */
830 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
831 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
832 	if (reg2) {
833 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
834 		vcpu->run->s.regs.gprs[reg2] |=
835 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
836 	}
837 	return 0;
838 }
839 
840 #define PFMF_RESERVED   0xfffc0101UL
841 #define PFMF_SK         0x00020000UL
842 #define PFMF_CF         0x00010000UL
843 #define PFMF_UI         0x00008000UL
844 #define PFMF_FSC        0x00007000UL
845 #define PFMF_NQ         0x00000800UL
846 #define PFMF_MR         0x00000400UL
847 #define PFMF_MC         0x00000200UL
848 #define PFMF_KEY        0x000000feUL
849 
850 static int handle_pfmf(struct kvm_vcpu *vcpu)
851 {
852 	bool mr = false, mc = false, nq;
853 	int reg1, reg2;
854 	unsigned long start, end;
855 	unsigned char key;
856 
857 	vcpu->stat.instruction_pfmf++;
858 
859 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
860 
861 	if (!test_kvm_facility(vcpu->kvm, 8))
862 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
863 
864 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
865 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
866 
867 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
868 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
869 
870 	/* Only provide non-quiescing support if enabled for the guest */
871 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
872 	    !test_kvm_facility(vcpu->kvm, 14))
873 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
874 
875 	/* Only provide conditional-SSKE support if enabled for the guest */
876 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
877 	    test_kvm_facility(vcpu->kvm, 10)) {
878 		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
879 		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
880 	}
881 
882 	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
883 	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
884 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
885 	start = kvm_s390_logical_to_effective(vcpu, start);
886 
887 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
888 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
889 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
890 	}
891 
892 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
893 	case 0x00000000:
894 		/* only 4k frames specify a real address */
895 		start = kvm_s390_real_to_abs(vcpu, start);
896 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
897 		break;
898 	case 0x00001000:
899 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
900 		break;
901 	case 0x00002000:
902 		/* only support 2G frame size if EDAT2 is available and we are
903 		   not in 24-bit addressing mode */
904 		if (!test_kvm_facility(vcpu->kvm, 78) ||
905 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
906 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
907 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
908 		break;
909 	default:
910 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
911 	}
912 
913 	while (start != end) {
914 		unsigned long useraddr;
915 
916 		/* Translate guest address to host address */
917 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
918 		if (kvm_is_error_hva(useraddr))
919 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
920 
921 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
922 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
923 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
924 		}
925 
926 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
927 			int rc = kvm_s390_skey_check_enable(vcpu);
928 
929 			if (rc)
930 				return rc;
931 			down_read(&current->mm->mmap_sem);
932 			rc = cond_set_guest_storage_key(current->mm, useraddr,
933 							key, NULL, nq, mr, mc);
934 			up_read(&current->mm->mmap_sem);
935 			if (rc < 0)
936 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
937 		}
938 
939 		start += PAGE_SIZE;
940 	}
941 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
942 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
943 			vcpu->run->s.regs.gprs[reg2] = end;
944 		} else {
945 			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
946 			end = kvm_s390_logical_to_effective(vcpu, end);
947 			vcpu->run->s.regs.gprs[reg2] |= end;
948 		}
949 	}
950 	return 0;
951 }
952 
953 static inline int do_essa(struct kvm_vcpu *vcpu, const int orc)
954 {
955 	struct kvm_s390_migration_state *ms = vcpu->kvm->arch.migration_state;
956 	int r1, r2, nappended, entries;
957 	unsigned long gfn, hva, res, pgstev, ptev;
958 	unsigned long *cbrlo;
959 
960 	/*
961 	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
962 	 * machine check here we either handle it or crash
963 	 */
964 
965 	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
966 	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
967 	hva = gfn_to_hva(vcpu->kvm, gfn);
968 	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
969 
970 	if (kvm_is_error_hva(hva))
971 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
972 
973 	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
974 	if (nappended < 0) {
975 		res = orc ? 0x10 : 0;
976 		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
977 		return 0;
978 	}
979 	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
980 	/*
981 	 * Set the block-content state part of the result. 0 means resident, so
982 	 * nothing to do if the page is valid. 2 is for preserved pages
983 	 * (non-present and non-zero), and 3 for zero pages (non-present and
984 	 * zero).
985 	 */
986 	if (ptev & _PAGE_INVALID) {
987 		res |= 2;
988 		if (pgstev & _PGSTE_GPS_ZERO)
989 			res |= 1;
990 	}
991 	vcpu->run->s.regs.gprs[r1] = res;
992 	/*
993 	 * It is possible that all the normal 511 slots were full, in which case
994 	 * we will now write in the 512th slot, which is reserved for host use.
995 	 * In both cases we let the normal essa handling code process all the
996 	 * slots, including the reserved one, if needed.
997 	 */
998 	if (nappended > 0) {
999 		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1000 		cbrlo[entries] = gfn << PAGE_SHIFT;
1001 	}
1002 
1003 	if (orc) {
1004 		/* increment only if we are really flipping the bit to 1 */
1005 		if (!test_and_set_bit(gfn, ms->pgste_bitmap))
1006 			atomic64_inc(&ms->dirty_pages);
1007 	}
1008 
1009 	return nappended;
1010 }
1011 
1012 static int handle_essa(struct kvm_vcpu *vcpu)
1013 {
1014 	/* entries expected to be 1FF */
1015 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1016 	unsigned long *cbrlo;
1017 	struct gmap *gmap;
1018 	int i, orc;
1019 
1020 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1021 	gmap = vcpu->arch.gmap;
1022 	vcpu->stat.instruction_essa++;
1023 	if (!vcpu->kvm->arch.use_cmma)
1024 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1025 
1026 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1027 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1028 	/* Check for invalid operation request code */
1029 	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1030 	if (orc > ESSA_MAX)
1031 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1032 
1033 	if (likely(!vcpu->kvm->arch.migration_state)) {
1034 		/*
1035 		 * CMMA is enabled in the KVM settings, but is disabled in
1036 		 * the SIE block and in the mm_context, and we are not doing
1037 		 * a migration. Enable CMMA in the mm_context.
1038 		 * Since we need to take a write lock to write to the context
1039 		 * to avoid races with storage keys handling, we check if the
1040 		 * value really needs to be written to; if the value is
1041 		 * already correct, we do nothing and avoid the lock.
1042 		 */
1043 		if (vcpu->kvm->mm->context.use_cmma == 0) {
1044 			down_write(&vcpu->kvm->mm->mmap_sem);
1045 			vcpu->kvm->mm->context.use_cmma = 1;
1046 			up_write(&vcpu->kvm->mm->mmap_sem);
1047 		}
1048 		/*
1049 		 * If we are here, we are supposed to have CMMA enabled in
1050 		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1051 		 * while the context use_cmma flag is per process.
1052 		 * It's possible that the context flag is enabled and the
1053 		 * SIE flag is not, so we set the flag always; if it was
1054 		 * already set, nothing changes, otherwise we enable it
1055 		 * on this CPU too.
1056 		 */
1057 		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1058 		/* Retry the ESSA instruction */
1059 		kvm_s390_retry_instr(vcpu);
1060 	} else {
1061 		/* Account for the possible extra cbrl entry */
1062 		i = do_essa(vcpu, orc);
1063 		if (i < 0)
1064 			return i;
1065 		entries += i;
1066 	}
1067 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1068 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1069 	down_read(&gmap->mm->mmap_sem);
1070 	for (i = 0; i < entries; ++i)
1071 		__gmap_zap(gmap, cbrlo[i]);
1072 	up_read(&gmap->mm->mmap_sem);
1073 	return 0;
1074 }
1075 
1076 static const intercept_handler_t b9_handlers[256] = {
1077 	[0x8a] = handle_ipte_interlock,
1078 	[0x8d] = handle_epsw,
1079 	[0x8e] = handle_ipte_interlock,
1080 	[0x8f] = handle_ipte_interlock,
1081 	[0xab] = handle_essa,
1082 	[0xaf] = handle_pfmf,
1083 };
1084 
1085 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1086 {
1087 	intercept_handler_t handler;
1088 
1089 	/* This is handled just as for the B2 instructions. */
1090 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1091 	if (handler)
1092 		return handler(vcpu);
1093 
1094 	return -EOPNOTSUPP;
1095 }
1096 
1097 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1098 {
1099 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1100 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1101 	int reg, rc, nr_regs;
1102 	u32 ctl_array[16];
1103 	u64 ga;
1104 	u8 ar;
1105 
1106 	vcpu->stat.instruction_lctl++;
1107 
1108 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1109 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1110 
1111 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1112 
1113 	if (ga & 3)
1114 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1115 
1116 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1117 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1118 
1119 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1120 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1121 	if (rc)
1122 		return kvm_s390_inject_prog_cond(vcpu, rc);
1123 	reg = reg1;
1124 	nr_regs = 0;
1125 	do {
1126 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1127 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1128 		if (reg == reg3)
1129 			break;
1130 		reg = (reg + 1) % 16;
1131 	} while (1);
1132 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1133 	return 0;
1134 }
1135 
1136 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1137 {
1138 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1139 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1140 	int reg, rc, nr_regs;
1141 	u32 ctl_array[16];
1142 	u64 ga;
1143 	u8 ar;
1144 
1145 	vcpu->stat.instruction_stctl++;
1146 
1147 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1148 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1149 
1150 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1151 
1152 	if (ga & 3)
1153 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1154 
1155 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1156 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1157 
1158 	reg = reg1;
1159 	nr_regs = 0;
1160 	do {
1161 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1162 		if (reg == reg3)
1163 			break;
1164 		reg = (reg + 1) % 16;
1165 	} while (1);
1166 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1167 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1168 }
1169 
1170 static int handle_lctlg(struct kvm_vcpu *vcpu)
1171 {
1172 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1173 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1174 	int reg, rc, nr_regs;
1175 	u64 ctl_array[16];
1176 	u64 ga;
1177 	u8 ar;
1178 
1179 	vcpu->stat.instruction_lctlg++;
1180 
1181 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1182 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1183 
1184 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1185 
1186 	if (ga & 7)
1187 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1188 
1189 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1190 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1191 
1192 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1193 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1194 	if (rc)
1195 		return kvm_s390_inject_prog_cond(vcpu, rc);
1196 	reg = reg1;
1197 	nr_regs = 0;
1198 	do {
1199 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1200 		if (reg == reg3)
1201 			break;
1202 		reg = (reg + 1) % 16;
1203 	} while (1);
1204 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1205 	return 0;
1206 }
1207 
1208 static int handle_stctg(struct kvm_vcpu *vcpu)
1209 {
1210 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1211 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1212 	int reg, rc, nr_regs;
1213 	u64 ctl_array[16];
1214 	u64 ga;
1215 	u8 ar;
1216 
1217 	vcpu->stat.instruction_stctg++;
1218 
1219 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1220 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1221 
1222 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1223 
1224 	if (ga & 7)
1225 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1226 
1227 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1228 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1229 
1230 	reg = reg1;
1231 	nr_regs = 0;
1232 	do {
1233 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1234 		if (reg == reg3)
1235 			break;
1236 		reg = (reg + 1) % 16;
1237 	} while (1);
1238 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1239 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1240 }
1241 
1242 static const intercept_handler_t eb_handlers[256] = {
1243 	[0x2f] = handle_lctlg,
1244 	[0x25] = handle_stctg,
1245 	[0x60] = handle_ri,
1246 	[0x61] = handle_ri,
1247 	[0x62] = handle_ri,
1248 };
1249 
1250 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1251 {
1252 	intercept_handler_t handler;
1253 
1254 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
1255 	if (handler)
1256 		return handler(vcpu);
1257 	return -EOPNOTSUPP;
1258 }
1259 
1260 static int handle_tprot(struct kvm_vcpu *vcpu)
1261 {
1262 	u64 address1, address2;
1263 	unsigned long hva, gpa;
1264 	int ret = 0, cc = 0;
1265 	bool writable;
1266 	u8 ar;
1267 
1268 	vcpu->stat.instruction_tprot++;
1269 
1270 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1271 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1272 
1273 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1274 
1275 	/* we only handle the Linux memory detection case:
1276 	 * access key == 0
1277 	 * everything else goes to userspace. */
1278 	if (address2 & 0xf0)
1279 		return -EOPNOTSUPP;
1280 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1281 		ipte_lock(vcpu);
1282 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1283 	if (ret == PGM_PROTECTION) {
1284 		/* Write protected? Try again with read-only... */
1285 		cc = 1;
1286 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
1287 					      GACC_FETCH);
1288 	}
1289 	if (ret) {
1290 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1291 			ret = kvm_s390_inject_program_int(vcpu, ret);
1292 		} else if (ret > 0) {
1293 			/* Translation not available */
1294 			kvm_s390_set_psw_cc(vcpu, 3);
1295 			ret = 0;
1296 		}
1297 		goto out_unlock;
1298 	}
1299 
1300 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1301 	if (kvm_is_error_hva(hva)) {
1302 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1303 	} else {
1304 		if (!writable)
1305 			cc = 1;		/* Write not permitted ==> read-only */
1306 		kvm_s390_set_psw_cc(vcpu, cc);
1307 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1308 	}
1309 out_unlock:
1310 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1311 		ipte_unlock(vcpu);
1312 	return ret;
1313 }
1314 
1315 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1316 {
1317 	/* For e5xx... instructions we only handle TPROT */
1318 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1319 		return handle_tprot(vcpu);
1320 	return -EOPNOTSUPP;
1321 }
1322 
1323 static int handle_sckpf(struct kvm_vcpu *vcpu)
1324 {
1325 	u32 value;
1326 
1327 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1328 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1329 
1330 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1331 		return kvm_s390_inject_program_int(vcpu,
1332 						   PGM_SPECIFICATION);
1333 
1334 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1335 	vcpu->arch.sie_block->todpr = value;
1336 
1337 	return 0;
1338 }
1339 
1340 static int handle_ptff(struct kvm_vcpu *vcpu)
1341 {
1342 	/* we don't emulate any control instructions yet */
1343 	kvm_s390_set_psw_cc(vcpu, 3);
1344 	return 0;
1345 }
1346 
1347 static const intercept_handler_t x01_handlers[256] = {
1348 	[0x04] = handle_ptff,
1349 	[0x07] = handle_sckpf,
1350 };
1351 
1352 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1353 {
1354 	intercept_handler_t handler;
1355 
1356 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1357 	if (handler)
1358 		return handler(vcpu);
1359 	return -EOPNOTSUPP;
1360 }
1361