xref: /linux/arch/s390/kvm/priv.c (revision d809aa238744ae5b7520b73ac5411862ccfdc1bc)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * handling privileged instructions
4  *
5  * Copyright IBM Corp. 2008, 2013
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License (version 2 only)
9  * as published by the Free Software Foundation.
10  *
11  *    Author(s): Carsten Otte <cotte@de.ibm.com>
12  *               Christian Borntraeger <borntraeger@de.ibm.com>
13  */
14 
15 #include <linux/kvm.h>
16 #include <linux/gfp.h>
17 #include <linux/errno.h>
18 #include <linux/compat.h>
19 #include <linux/mm_types.h>
20 
21 #include <asm/asm-offsets.h>
22 #include <asm/facility.h>
23 #include <asm/current.h>
24 #include <asm/debug.h>
25 #include <asm/ebcdic.h>
26 #include <asm/sysinfo.h>
27 #include <asm/pgtable.h>
28 #include <asm/page-states.h>
29 #include <asm/pgalloc.h>
30 #include <asm/gmap.h>
31 #include <asm/io.h>
32 #include <asm/ptrace.h>
33 #include <asm/compat.h>
34 #include <asm/sclp.h>
35 #include "gaccess.h"
36 #include "kvm-s390.h"
37 #include "trace.h"
38 
39 static int handle_ri(struct kvm_vcpu *vcpu)
40 {
41 	if (test_kvm_facility(vcpu->kvm, 64)) {
42 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
43 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
44 		kvm_s390_retry_instr(vcpu);
45 		return 0;
46 	} else
47 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
48 }
49 
50 int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
51 {
52 	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
53 		return handle_ri(vcpu);
54 	else
55 		return -EOPNOTSUPP;
56 }
57 
58 static int handle_gs(struct kvm_vcpu *vcpu)
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 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
93 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
94 
95 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
96 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
97 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
98 	rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
99 	if (rc)
100 		return kvm_s390_inject_prog_cond(vcpu, rc);
101 
102 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
103 	kvm_s390_set_tod_clock(vcpu->kvm, val);
104 
105 	kvm_s390_set_psw_cc(vcpu, 0);
106 	return 0;
107 }
108 
109 static int handle_set_prefix(struct kvm_vcpu *vcpu)
110 {
111 	u64 operand2;
112 	u32 address;
113 	int rc;
114 	u8 ar;
115 
116 	vcpu->stat.instruction_spx++;
117 
118 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
119 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
120 
121 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
122 
123 	/* must be word boundary */
124 	if (operand2 & 3)
125 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
126 
127 	/* get the value */
128 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
129 	if (rc)
130 		return kvm_s390_inject_prog_cond(vcpu, rc);
131 
132 	address &= 0x7fffe000u;
133 
134 	/*
135 	 * Make sure the new value is valid memory. We only need to check the
136 	 * first page, since address is 8k aligned and memory pieces are always
137 	 * at least 1MB aligned and have at least a size of 1MB.
138 	 */
139 	if (kvm_is_error_gpa(vcpu->kvm, address))
140 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
141 
142 	kvm_s390_set_prefix(vcpu, address);
143 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
144 	return 0;
145 }
146 
147 static int handle_store_prefix(struct kvm_vcpu *vcpu)
148 {
149 	u64 operand2;
150 	u32 address;
151 	int rc;
152 	u8 ar;
153 
154 	vcpu->stat.instruction_stpx++;
155 
156 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
157 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
158 
159 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
160 
161 	/* must be word boundary */
162 	if (operand2 & 3)
163 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
164 
165 	address = kvm_s390_get_prefix(vcpu);
166 
167 	/* get the value */
168 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
169 	if (rc)
170 		return kvm_s390_inject_prog_cond(vcpu, rc);
171 
172 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
173 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
174 	return 0;
175 }
176 
177 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
178 {
179 	u16 vcpu_id = vcpu->vcpu_id;
180 	u64 ga;
181 	int rc;
182 	u8 ar;
183 
184 	vcpu->stat.instruction_stap++;
185 
186 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
187 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
188 
189 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
190 
191 	if (ga & 1)
192 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
193 
194 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
195 	if (rc)
196 		return kvm_s390_inject_prog_cond(vcpu, rc);
197 
198 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
199 	trace_kvm_s390_handle_stap(vcpu, ga);
200 	return 0;
201 }
202 
203 int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
204 {
205 	int rc = 0;
206 	struct kvm_s390_sie_block *sie_block = vcpu->arch.sie_block;
207 
208 	trace_kvm_s390_skey_related_inst(vcpu);
209 	if (!(sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)) &&
210 	    !(atomic_read(&sie_block->cpuflags) & CPUSTAT_KSS))
211 		return rc;
212 
213 	rc = s390_enable_skey();
214 	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
215 	if (!rc) {
216 		if (atomic_read(&sie_block->cpuflags) & CPUSTAT_KSS)
217 			atomic_andnot(CPUSTAT_KSS, &sie_block->cpuflags);
218 		else
219 			sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE |
220 					     ICTL_RRBE);
221 	}
222 	return rc;
223 }
224 
225 static int try_handle_skey(struct kvm_vcpu *vcpu)
226 {
227 	int rc;
228 
229 	vcpu->stat.instruction_storage_key++;
230 	rc = kvm_s390_skey_check_enable(vcpu);
231 	if (rc)
232 		return rc;
233 	if (sclp.has_skey) {
234 		/* with storage-key facility, SIE interprets it for us */
235 		kvm_s390_retry_instr(vcpu);
236 		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
237 		return -EAGAIN;
238 	}
239 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
240 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
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 	rc = try_handle_skey(vcpu);
252 	if (rc)
253 		return rc != -EAGAIN ? rc : 0;
254 
255 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
256 
257 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
258 	addr = kvm_s390_logical_to_effective(vcpu, addr);
259 	addr = kvm_s390_real_to_abs(vcpu, addr);
260 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
261 	if (kvm_is_error_hva(addr))
262 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
263 
264 	down_read(&current->mm->mmap_sem);
265 	rc = get_guest_storage_key(current->mm, addr, &key);
266 	up_read(&current->mm->mmap_sem);
267 	if (rc)
268 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
269 	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
270 	vcpu->run->s.regs.gprs[reg1] |= key;
271 	return 0;
272 }
273 
274 static int handle_rrbe(struct kvm_vcpu *vcpu)
275 {
276 	unsigned long addr;
277 	int reg1, reg2;
278 	int rc;
279 
280 	rc = try_handle_skey(vcpu);
281 	if (rc)
282 		return rc != -EAGAIN ? rc : 0;
283 
284 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
285 
286 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
287 	addr = kvm_s390_logical_to_effective(vcpu, addr);
288 	addr = kvm_s390_real_to_abs(vcpu, addr);
289 	addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(addr));
290 	if (kvm_is_error_hva(addr))
291 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
292 
293 	down_read(&current->mm->mmap_sem);
294 	rc = reset_guest_reference_bit(current->mm, addr);
295 	up_read(&current->mm->mmap_sem);
296 	if (rc < 0)
297 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
298 
299 	kvm_s390_set_psw_cc(vcpu, rc);
300 	return 0;
301 }
302 
303 #define SSKE_NQ 0x8
304 #define SSKE_MR 0x4
305 #define SSKE_MC 0x2
306 #define SSKE_MB 0x1
307 static int handle_sske(struct kvm_vcpu *vcpu)
308 {
309 	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
310 	unsigned long start, end;
311 	unsigned char key, oldkey;
312 	int reg1, reg2;
313 	int rc;
314 
315 	rc = try_handle_skey(vcpu);
316 	if (rc)
317 		return rc != -EAGAIN ? rc : 0;
318 
319 	if (!test_kvm_facility(vcpu->kvm, 8))
320 		m3 &= ~SSKE_MB;
321 	if (!test_kvm_facility(vcpu->kvm, 10))
322 		m3 &= ~(SSKE_MC | SSKE_MR);
323 	if (!test_kvm_facility(vcpu->kvm, 14))
324 		m3 &= ~SSKE_NQ;
325 
326 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
327 
328 	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
329 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
330 	start = kvm_s390_logical_to_effective(vcpu, start);
331 	if (m3 & SSKE_MB) {
332 		/* start already designates an absolute address */
333 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
334 	} else {
335 		start = kvm_s390_real_to_abs(vcpu, start);
336 		end = start + PAGE_SIZE;
337 	}
338 
339 	while (start != end) {
340 		unsigned long addr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
341 
342 		if (kvm_is_error_hva(addr))
343 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
344 
345 		down_read(&current->mm->mmap_sem);
346 		rc = cond_set_guest_storage_key(current->mm, addr, key, &oldkey,
347 						m3 & SSKE_NQ, m3 & SSKE_MR,
348 						m3 & SSKE_MC);
349 		up_read(&current->mm->mmap_sem);
350 		if (rc < 0)
351 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
352 		start += PAGE_SIZE;
353 	}
354 
355 	if (m3 & (SSKE_MC | SSKE_MR)) {
356 		if (m3 & SSKE_MB) {
357 			/* skey in reg1 is unpredictable */
358 			kvm_s390_set_psw_cc(vcpu, 3);
359 		} else {
360 			kvm_s390_set_psw_cc(vcpu, rc);
361 			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
362 			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
363 		}
364 	}
365 	if (m3 & SSKE_MB) {
366 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
367 			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
368 		else
369 			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
370 		end = kvm_s390_logical_to_effective(vcpu, end);
371 		vcpu->run->s.regs.gprs[reg2] |= end;
372 	}
373 	return 0;
374 }
375 
376 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
377 {
378 	vcpu->stat.instruction_ipte_interlock++;
379 	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
380 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
381 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
382 	kvm_s390_retry_instr(vcpu);
383 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
384 	return 0;
385 }
386 
387 static int handle_test_block(struct kvm_vcpu *vcpu)
388 {
389 	gpa_t addr;
390 	int reg2;
391 
392 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
393 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
394 
395 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
396 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
397 	addr = kvm_s390_logical_to_effective(vcpu, addr);
398 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
399 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
400 	addr = kvm_s390_real_to_abs(vcpu, addr);
401 
402 	if (kvm_is_error_gpa(vcpu->kvm, addr))
403 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
404 	/*
405 	 * We don't expect errors on modern systems, and do not care
406 	 * about storage keys (yet), so let's just clear the page.
407 	 */
408 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
409 		return -EFAULT;
410 	kvm_s390_set_psw_cc(vcpu, 0);
411 	vcpu->run->s.regs.gprs[0] = 0;
412 	return 0;
413 }
414 
415 static int handle_tpi(struct kvm_vcpu *vcpu)
416 {
417 	struct kvm_s390_interrupt_info *inti;
418 	unsigned long len;
419 	u32 tpi_data[3];
420 	int rc;
421 	u64 addr;
422 	u8 ar;
423 
424 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
425 	if (addr & 3)
426 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
427 
428 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
429 	if (!inti) {
430 		kvm_s390_set_psw_cc(vcpu, 0);
431 		return 0;
432 	}
433 
434 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
435 	tpi_data[1] = inti->io.io_int_parm;
436 	tpi_data[2] = inti->io.io_int_word;
437 	if (addr) {
438 		/*
439 		 * Store the two-word I/O interruption code into the
440 		 * provided area.
441 		 */
442 		len = sizeof(tpi_data) - 4;
443 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
444 		if (rc) {
445 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
446 			goto reinject_interrupt;
447 		}
448 	} else {
449 		/*
450 		 * Store the three-word I/O interruption code into
451 		 * the appropriate lowcore area.
452 		 */
453 		len = sizeof(tpi_data);
454 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
455 			/* failed writes to the low core are not recoverable */
456 			rc = -EFAULT;
457 			goto reinject_interrupt;
458 		}
459 	}
460 
461 	/* irq was successfully handed to the guest */
462 	kfree(inti);
463 	kvm_s390_set_psw_cc(vcpu, 1);
464 	return 0;
465 reinject_interrupt:
466 	/*
467 	 * If we encounter a problem storing the interruption code, the
468 	 * instruction is suppressed from the guest's view: reinject the
469 	 * interrupt.
470 	 */
471 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
472 		kfree(inti);
473 		rc = -EFAULT;
474 	}
475 	/* don't set the cc, a pgm irq was injected or we drop to user space */
476 	return rc ? -EFAULT : 0;
477 }
478 
479 static int handle_tsch(struct kvm_vcpu *vcpu)
480 {
481 	struct kvm_s390_interrupt_info *inti = NULL;
482 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
483 
484 	/* a valid schid has at least one bit set */
485 	if (vcpu->run->s.regs.gprs[1])
486 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
487 					   vcpu->run->s.regs.gprs[1]);
488 
489 	/*
490 	 * Prepare exit to userspace.
491 	 * We indicate whether we dequeued a pending I/O interrupt
492 	 * so that userspace can re-inject it if the instruction gets
493 	 * a program check. While this may re-order the pending I/O
494 	 * interrupts, this is no problem since the priority is kept
495 	 * intact.
496 	 */
497 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
498 	vcpu->run->s390_tsch.dequeued = !!inti;
499 	if (inti) {
500 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
501 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
502 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
503 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
504 	}
505 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
506 	kfree(inti);
507 	return -EREMOTE;
508 }
509 
510 static int handle_io_inst(struct kvm_vcpu *vcpu)
511 {
512 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
513 
514 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
515 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
516 
517 	if (vcpu->kvm->arch.css_support) {
518 		/*
519 		 * Most I/O instructions will be handled by userspace.
520 		 * Exceptions are tpi and the interrupt portion of tsch.
521 		 */
522 		if (vcpu->arch.sie_block->ipa == 0xb236)
523 			return handle_tpi(vcpu);
524 		if (vcpu->arch.sie_block->ipa == 0xb235)
525 			return handle_tsch(vcpu);
526 		/* Handle in userspace. */
527 		return -EOPNOTSUPP;
528 	} else {
529 		/*
530 		 * Set condition code 3 to stop the guest from issuing channel
531 		 * I/O instructions.
532 		 */
533 		kvm_s390_set_psw_cc(vcpu, 3);
534 		return 0;
535 	}
536 }
537 
538 static int handle_stfl(struct kvm_vcpu *vcpu)
539 {
540 	int rc;
541 	unsigned int fac;
542 
543 	vcpu->stat.instruction_stfl++;
544 
545 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
546 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
547 
548 	/*
549 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
550 	 * into a u32 memory representation. They will remain bits 0-31.
551 	 */
552 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
553 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
554 			    &fac, sizeof(fac));
555 	if (rc)
556 		return rc;
557 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
558 	trace_kvm_s390_handle_stfl(vcpu, fac);
559 	return 0;
560 }
561 
562 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
563 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
564 #define PSW_ADDR_24 0x0000000000ffffffUL
565 #define PSW_ADDR_31 0x000000007fffffffUL
566 
567 int is_valid_psw(psw_t *psw)
568 {
569 	if (psw->mask & PSW_MASK_UNASSIGNED)
570 		return 0;
571 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
572 		if (psw->addr & ~PSW_ADDR_31)
573 			return 0;
574 	}
575 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
576 		return 0;
577 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
578 		return 0;
579 	if (psw->addr & 1)
580 		return 0;
581 	return 1;
582 }
583 
584 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
585 {
586 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
587 	psw_compat_t new_psw;
588 	u64 addr;
589 	int rc;
590 	u8 ar;
591 
592 	if (gpsw->mask & PSW_MASK_PSTATE)
593 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
594 
595 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
596 	if (addr & 7)
597 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
598 
599 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
600 	if (rc)
601 		return kvm_s390_inject_prog_cond(vcpu, rc);
602 	if (!(new_psw.mask & PSW32_MASK_BASE))
603 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
604 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
605 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
606 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
607 	if (!is_valid_psw(gpsw))
608 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
609 	return 0;
610 }
611 
612 static int handle_lpswe(struct kvm_vcpu *vcpu)
613 {
614 	psw_t new_psw;
615 	u64 addr;
616 	int rc;
617 	u8 ar;
618 
619 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
620 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
621 
622 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
623 	if (addr & 7)
624 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
625 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
626 	if (rc)
627 		return kvm_s390_inject_prog_cond(vcpu, rc);
628 	vcpu->arch.sie_block->gpsw = new_psw;
629 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
630 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
631 	return 0;
632 }
633 
634 static int handle_stidp(struct kvm_vcpu *vcpu)
635 {
636 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
637 	u64 operand2;
638 	int rc;
639 	u8 ar;
640 
641 	vcpu->stat.instruction_stidp++;
642 
643 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
644 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
645 
646 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
647 
648 	if (operand2 & 7)
649 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
650 
651 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
652 	if (rc)
653 		return kvm_s390_inject_prog_cond(vcpu, rc);
654 
655 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
656 	return 0;
657 }
658 
659 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
660 {
661 	int cpus = 0;
662 	int n;
663 
664 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
665 
666 	/* deal with other level 3 hypervisors */
667 	if (stsi(mem, 3, 2, 2))
668 		mem->count = 0;
669 	if (mem->count < 8)
670 		mem->count++;
671 	for (n = mem->count - 1; n > 0 ; n--)
672 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
673 
674 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
675 	mem->vm[0].cpus_total = cpus;
676 	mem->vm[0].cpus_configured = cpus;
677 	mem->vm[0].cpus_standby = 0;
678 	mem->vm[0].cpus_reserved = 0;
679 	mem->vm[0].caf = 1000;
680 	memcpy(mem->vm[0].name, "KVMguest", 8);
681 	ASCEBC(mem->vm[0].name, 8);
682 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
683 	ASCEBC(mem->vm[0].cpi, 16);
684 }
685 
686 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
687 				 u8 fc, u8 sel1, u16 sel2)
688 {
689 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
690 	vcpu->run->s390_stsi.addr = addr;
691 	vcpu->run->s390_stsi.ar = ar;
692 	vcpu->run->s390_stsi.fc = fc;
693 	vcpu->run->s390_stsi.sel1 = sel1;
694 	vcpu->run->s390_stsi.sel2 = sel2;
695 }
696 
697 static int handle_stsi(struct kvm_vcpu *vcpu)
698 {
699 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
700 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
701 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
702 	unsigned long mem = 0;
703 	u64 operand2;
704 	int rc = 0;
705 	u8 ar;
706 
707 	vcpu->stat.instruction_stsi++;
708 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
709 
710 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
711 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
712 
713 	if (fc > 3) {
714 		kvm_s390_set_psw_cc(vcpu, 3);
715 		return 0;
716 	}
717 
718 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
719 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
720 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
721 
722 	if (fc == 0) {
723 		vcpu->run->s.regs.gprs[0] = 3 << 28;
724 		kvm_s390_set_psw_cc(vcpu, 0);
725 		return 0;
726 	}
727 
728 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
729 
730 	if (operand2 & 0xfff)
731 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
732 
733 	switch (fc) {
734 	case 1: /* same handling for 1 and 2 */
735 	case 2:
736 		mem = get_zeroed_page(GFP_KERNEL);
737 		if (!mem)
738 			goto out_no_data;
739 		if (stsi((void *) mem, fc, sel1, sel2))
740 			goto out_no_data;
741 		break;
742 	case 3:
743 		if (sel1 != 2 || sel2 != 2)
744 			goto out_no_data;
745 		mem = get_zeroed_page(GFP_KERNEL);
746 		if (!mem)
747 			goto out_no_data;
748 		handle_stsi_3_2_2(vcpu, (void *) mem);
749 		break;
750 	}
751 
752 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
753 	if (rc) {
754 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
755 		goto out;
756 	}
757 	if (vcpu->kvm->arch.user_stsi) {
758 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
759 		rc = -EREMOTE;
760 	}
761 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
762 	free_page(mem);
763 	kvm_s390_set_psw_cc(vcpu, 0);
764 	vcpu->run->s.regs.gprs[0] = 0;
765 	return rc;
766 out_no_data:
767 	kvm_s390_set_psw_cc(vcpu, 3);
768 out:
769 	free_page(mem);
770 	return rc;
771 }
772 
773 static const intercept_handler_t b2_handlers[256] = {
774 	[0x02] = handle_stidp,
775 	[0x04] = handle_set_clock,
776 	[0x10] = handle_set_prefix,
777 	[0x11] = handle_store_prefix,
778 	[0x12] = handle_store_cpu_address,
779 	[0x14] = kvm_s390_handle_vsie,
780 	[0x21] = handle_ipte_interlock,
781 	[0x29] = handle_iske,
782 	[0x2a] = handle_rrbe,
783 	[0x2b] = handle_sske,
784 	[0x2c] = handle_test_block,
785 	[0x30] = handle_io_inst,
786 	[0x31] = handle_io_inst,
787 	[0x32] = handle_io_inst,
788 	[0x33] = handle_io_inst,
789 	[0x34] = handle_io_inst,
790 	[0x35] = handle_io_inst,
791 	[0x36] = handle_io_inst,
792 	[0x37] = handle_io_inst,
793 	[0x38] = handle_io_inst,
794 	[0x39] = handle_io_inst,
795 	[0x3a] = handle_io_inst,
796 	[0x3b] = handle_io_inst,
797 	[0x3c] = handle_io_inst,
798 	[0x50] = handle_ipte_interlock,
799 	[0x56] = handle_sthyi,
800 	[0x5f] = handle_io_inst,
801 	[0x74] = handle_io_inst,
802 	[0x76] = handle_io_inst,
803 	[0x7d] = handle_stsi,
804 	[0xb1] = handle_stfl,
805 	[0xb2] = handle_lpswe,
806 };
807 
808 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
809 {
810 	intercept_handler_t handler;
811 
812 	/*
813 	 * A lot of B2 instructions are priviledged. Here we check for
814 	 * the privileged ones, that we can handle in the kernel.
815 	 * Anything else goes to userspace.
816 	 */
817 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
818 	if (handler)
819 		return handler(vcpu);
820 
821 	return -EOPNOTSUPP;
822 }
823 
824 static int handle_epsw(struct kvm_vcpu *vcpu)
825 {
826 	int reg1, reg2;
827 
828 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
829 
830 	/* This basically extracts the mask half of the psw. */
831 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
832 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
833 	if (reg2) {
834 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
835 		vcpu->run->s.regs.gprs[reg2] |=
836 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
837 	}
838 	return 0;
839 }
840 
841 #define PFMF_RESERVED   0xfffc0101UL
842 #define PFMF_SK         0x00020000UL
843 #define PFMF_CF         0x00010000UL
844 #define PFMF_UI         0x00008000UL
845 #define PFMF_FSC        0x00007000UL
846 #define PFMF_NQ         0x00000800UL
847 #define PFMF_MR         0x00000400UL
848 #define PFMF_MC         0x00000200UL
849 #define PFMF_KEY        0x000000feUL
850 
851 static int handle_pfmf(struct kvm_vcpu *vcpu)
852 {
853 	bool mr = false, mc = false, nq;
854 	int reg1, reg2;
855 	unsigned long start, end;
856 	unsigned char key;
857 
858 	vcpu->stat.instruction_pfmf++;
859 
860 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
861 
862 	if (!test_kvm_facility(vcpu->kvm, 8))
863 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
864 
865 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
866 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
867 
868 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
869 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
870 
871 	/* Only provide non-quiescing support if enabled for the guest */
872 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
873 	    !test_kvm_facility(vcpu->kvm, 14))
874 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
875 
876 	/* Only provide conditional-SSKE support if enabled for the guest */
877 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
878 	    test_kvm_facility(vcpu->kvm, 10)) {
879 		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
880 		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
881 	}
882 
883 	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
884 	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
885 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
886 	start = kvm_s390_logical_to_effective(vcpu, start);
887 
888 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
889 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
890 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
891 	}
892 
893 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
894 	case 0x00000000:
895 		/* only 4k frames specify a real address */
896 		start = kvm_s390_real_to_abs(vcpu, start);
897 		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
898 		break;
899 	case 0x00001000:
900 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
901 		break;
902 	case 0x00002000:
903 		/* only support 2G frame size if EDAT2 is available and we are
904 		   not in 24-bit addressing mode */
905 		if (!test_kvm_facility(vcpu->kvm, 78) ||
906 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
907 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
908 		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
909 		break;
910 	default:
911 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
912 	}
913 
914 	while (start != end) {
915 		unsigned long useraddr;
916 
917 		/* Translate guest address to host address */
918 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
919 		if (kvm_is_error_hva(useraddr))
920 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
921 
922 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
923 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
924 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
925 		}
926 
927 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
928 			int rc = kvm_s390_skey_check_enable(vcpu);
929 
930 			if (rc)
931 				return rc;
932 			down_read(&current->mm->mmap_sem);
933 			rc = cond_set_guest_storage_key(current->mm, useraddr,
934 							key, NULL, nq, mr, mc);
935 			up_read(&current->mm->mmap_sem);
936 			if (rc < 0)
937 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
938 		}
939 
940 		start += PAGE_SIZE;
941 	}
942 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
943 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
944 			vcpu->run->s.regs.gprs[reg2] = end;
945 		} else {
946 			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
947 			end = kvm_s390_logical_to_effective(vcpu, end);
948 			vcpu->run->s.regs.gprs[reg2] |= end;
949 		}
950 	}
951 	return 0;
952 }
953 
954 static inline int do_essa(struct kvm_vcpu *vcpu, const int orc)
955 {
956 	struct kvm_s390_migration_state *ms = vcpu->kvm->arch.migration_state;
957 	int r1, r2, nappended, entries;
958 	unsigned long gfn, hva, res, pgstev, ptev;
959 	unsigned long *cbrlo;
960 
961 	/*
962 	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
963 	 * machine check here we either handle it or crash
964 	 */
965 
966 	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
967 	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
968 	hva = gfn_to_hva(vcpu->kvm, gfn);
969 	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
970 
971 	if (kvm_is_error_hva(hva))
972 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
973 
974 	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
975 	if (nappended < 0) {
976 		res = orc ? 0x10 : 0;
977 		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
978 		return 0;
979 	}
980 	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
981 	/*
982 	 * Set the block-content state part of the result. 0 means resident, so
983 	 * nothing to do if the page is valid. 2 is for preserved pages
984 	 * (non-present and non-zero), and 3 for zero pages (non-present and
985 	 * zero).
986 	 */
987 	if (ptev & _PAGE_INVALID) {
988 		res |= 2;
989 		if (pgstev & _PGSTE_GPS_ZERO)
990 			res |= 1;
991 	}
992 	if (pgstev & _PGSTE_GPS_NODAT)
993 		res |= 0x20;
994 	vcpu->run->s.regs.gprs[r1] = res;
995 	/*
996 	 * It is possible that all the normal 511 slots were full, in which case
997 	 * we will now write in the 512th slot, which is reserved for host use.
998 	 * In both cases we let the normal essa handling code process all the
999 	 * slots, including the reserved one, if needed.
1000 	 */
1001 	if (nappended > 0) {
1002 		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1003 		cbrlo[entries] = gfn << PAGE_SHIFT;
1004 	}
1005 
1006 	if (orc) {
1007 		/* increment only if we are really flipping the bit to 1 */
1008 		if (!test_and_set_bit(gfn, ms->pgste_bitmap))
1009 			atomic64_inc(&ms->dirty_pages);
1010 	}
1011 
1012 	return nappended;
1013 }
1014 
1015 static int handle_essa(struct kvm_vcpu *vcpu)
1016 {
1017 	/* entries expected to be 1FF */
1018 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1019 	unsigned long *cbrlo;
1020 	struct gmap *gmap;
1021 	int i, orc;
1022 
1023 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1024 	gmap = vcpu->arch.gmap;
1025 	vcpu->stat.instruction_essa++;
1026 	if (!vcpu->kvm->arch.use_cmma)
1027 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1028 
1029 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1030 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1031 	/* Check for invalid operation request code */
1032 	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1033 	/* ORCs 0-6 are always valid */
1034 	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1035 						: ESSA_SET_STABLE_IF_RESIDENT))
1036 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1037 
1038 	if (likely(!vcpu->kvm->arch.migration_state)) {
1039 		/*
1040 		 * CMMA is enabled in the KVM settings, but is disabled in
1041 		 * the SIE block and in the mm_context, and we are not doing
1042 		 * a migration. Enable CMMA in the mm_context.
1043 		 * Since we need to take a write lock to write to the context
1044 		 * to avoid races with storage keys handling, we check if the
1045 		 * value really needs to be written to; if the value is
1046 		 * already correct, we do nothing and avoid the lock.
1047 		 */
1048 		if (vcpu->kvm->mm->context.use_cmma == 0) {
1049 			down_write(&vcpu->kvm->mm->mmap_sem);
1050 			vcpu->kvm->mm->context.use_cmma = 1;
1051 			up_write(&vcpu->kvm->mm->mmap_sem);
1052 		}
1053 		/*
1054 		 * If we are here, we are supposed to have CMMA enabled in
1055 		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1056 		 * while the context use_cmma flag is per process.
1057 		 * It's possible that the context flag is enabled and the
1058 		 * SIE flag is not, so we set the flag always; if it was
1059 		 * already set, nothing changes, otherwise we enable it
1060 		 * on this CPU too.
1061 		 */
1062 		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1063 		/* Retry the ESSA instruction */
1064 		kvm_s390_retry_instr(vcpu);
1065 	} else {
1066 		/* Account for the possible extra cbrl entry */
1067 		i = do_essa(vcpu, orc);
1068 		if (i < 0)
1069 			return i;
1070 		entries += i;
1071 	}
1072 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1073 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1074 	down_read(&gmap->mm->mmap_sem);
1075 	for (i = 0; i < entries; ++i)
1076 		__gmap_zap(gmap, cbrlo[i]);
1077 	up_read(&gmap->mm->mmap_sem);
1078 	return 0;
1079 }
1080 
1081 static const intercept_handler_t b9_handlers[256] = {
1082 	[0x8a] = handle_ipte_interlock,
1083 	[0x8d] = handle_epsw,
1084 	[0x8e] = handle_ipte_interlock,
1085 	[0x8f] = handle_ipte_interlock,
1086 	[0xab] = handle_essa,
1087 	[0xaf] = handle_pfmf,
1088 };
1089 
1090 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1091 {
1092 	intercept_handler_t handler;
1093 
1094 	/* This is handled just as for the B2 instructions. */
1095 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1096 	if (handler)
1097 		return handler(vcpu);
1098 
1099 	return -EOPNOTSUPP;
1100 }
1101 
1102 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1103 {
1104 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1105 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1106 	int reg, rc, nr_regs;
1107 	u32 ctl_array[16];
1108 	u64 ga;
1109 	u8 ar;
1110 
1111 	vcpu->stat.instruction_lctl++;
1112 
1113 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1114 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1115 
1116 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1117 
1118 	if (ga & 3)
1119 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1120 
1121 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1122 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1123 
1124 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1125 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1126 	if (rc)
1127 		return kvm_s390_inject_prog_cond(vcpu, rc);
1128 	reg = reg1;
1129 	nr_regs = 0;
1130 	do {
1131 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1132 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1133 		if (reg == reg3)
1134 			break;
1135 		reg = (reg + 1) % 16;
1136 	} while (1);
1137 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1138 	return 0;
1139 }
1140 
1141 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1142 {
1143 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1144 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1145 	int reg, rc, nr_regs;
1146 	u32 ctl_array[16];
1147 	u64 ga;
1148 	u8 ar;
1149 
1150 	vcpu->stat.instruction_stctl++;
1151 
1152 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1153 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1154 
1155 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1156 
1157 	if (ga & 3)
1158 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1159 
1160 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1161 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1162 
1163 	reg = reg1;
1164 	nr_regs = 0;
1165 	do {
1166 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1167 		if (reg == reg3)
1168 			break;
1169 		reg = (reg + 1) % 16;
1170 	} while (1);
1171 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1172 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1173 }
1174 
1175 static int handle_lctlg(struct kvm_vcpu *vcpu)
1176 {
1177 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1178 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1179 	int reg, rc, nr_regs;
1180 	u64 ctl_array[16];
1181 	u64 ga;
1182 	u8 ar;
1183 
1184 	vcpu->stat.instruction_lctlg++;
1185 
1186 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1187 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1188 
1189 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1190 
1191 	if (ga & 7)
1192 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1193 
1194 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1195 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1196 
1197 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1198 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1199 	if (rc)
1200 		return kvm_s390_inject_prog_cond(vcpu, rc);
1201 	reg = reg1;
1202 	nr_regs = 0;
1203 	do {
1204 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1205 		if (reg == reg3)
1206 			break;
1207 		reg = (reg + 1) % 16;
1208 	} while (1);
1209 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1210 	return 0;
1211 }
1212 
1213 static int handle_stctg(struct kvm_vcpu *vcpu)
1214 {
1215 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1216 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1217 	int reg, rc, nr_regs;
1218 	u64 ctl_array[16];
1219 	u64 ga;
1220 	u8 ar;
1221 
1222 	vcpu->stat.instruction_stctg++;
1223 
1224 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1225 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1226 
1227 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1228 
1229 	if (ga & 7)
1230 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1231 
1232 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1233 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1234 
1235 	reg = reg1;
1236 	nr_regs = 0;
1237 	do {
1238 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1239 		if (reg == reg3)
1240 			break;
1241 		reg = (reg + 1) % 16;
1242 	} while (1);
1243 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1244 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1245 }
1246 
1247 static const intercept_handler_t eb_handlers[256] = {
1248 	[0x2f] = handle_lctlg,
1249 	[0x25] = handle_stctg,
1250 	[0x60] = handle_ri,
1251 	[0x61] = handle_ri,
1252 	[0x62] = handle_ri,
1253 };
1254 
1255 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1256 {
1257 	intercept_handler_t handler;
1258 
1259 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
1260 	if (handler)
1261 		return handler(vcpu);
1262 	return -EOPNOTSUPP;
1263 }
1264 
1265 static int handle_tprot(struct kvm_vcpu *vcpu)
1266 {
1267 	u64 address1, address2;
1268 	unsigned long hva, gpa;
1269 	int ret = 0, cc = 0;
1270 	bool writable;
1271 	u8 ar;
1272 
1273 	vcpu->stat.instruction_tprot++;
1274 
1275 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1276 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1277 
1278 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1279 
1280 	/* we only handle the Linux memory detection case:
1281 	 * access key == 0
1282 	 * everything else goes to userspace. */
1283 	if (address2 & 0xf0)
1284 		return -EOPNOTSUPP;
1285 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1286 		ipte_lock(vcpu);
1287 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1288 	if (ret == PGM_PROTECTION) {
1289 		/* Write protected? Try again with read-only... */
1290 		cc = 1;
1291 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
1292 					      GACC_FETCH);
1293 	}
1294 	if (ret) {
1295 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1296 			ret = kvm_s390_inject_program_int(vcpu, ret);
1297 		} else if (ret > 0) {
1298 			/* Translation not available */
1299 			kvm_s390_set_psw_cc(vcpu, 3);
1300 			ret = 0;
1301 		}
1302 		goto out_unlock;
1303 	}
1304 
1305 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1306 	if (kvm_is_error_hva(hva)) {
1307 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1308 	} else {
1309 		if (!writable)
1310 			cc = 1;		/* Write not permitted ==> read-only */
1311 		kvm_s390_set_psw_cc(vcpu, cc);
1312 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1313 	}
1314 out_unlock:
1315 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1316 		ipte_unlock(vcpu);
1317 	return ret;
1318 }
1319 
1320 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1321 {
1322 	/* For e5xx... instructions we only handle TPROT */
1323 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1324 		return handle_tprot(vcpu);
1325 	return -EOPNOTSUPP;
1326 }
1327 
1328 static int handle_sckpf(struct kvm_vcpu *vcpu)
1329 {
1330 	u32 value;
1331 
1332 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1333 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1334 
1335 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1336 		return kvm_s390_inject_program_int(vcpu,
1337 						   PGM_SPECIFICATION);
1338 
1339 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1340 	vcpu->arch.sie_block->todpr = value;
1341 
1342 	return 0;
1343 }
1344 
1345 static int handle_ptff(struct kvm_vcpu *vcpu)
1346 {
1347 	/* we don't emulate any control instructions yet */
1348 	kvm_s390_set_psw_cc(vcpu, 3);
1349 	return 0;
1350 }
1351 
1352 static const intercept_handler_t x01_handlers[256] = {
1353 	[0x04] = handle_ptff,
1354 	[0x07] = handle_sckpf,
1355 };
1356 
1357 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1358 {
1359 	intercept_handler_t handler;
1360 
1361 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1362 	if (handler)
1363 		return handler(vcpu);
1364 	return -EOPNOTSUPP;
1365 }
1366