xref: /linux/arch/s390/kvm/priv.c (revision 957e3facd147510f2cf8780e38606f1d707f0e33)
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 <asm/asm-offsets.h>
19 #include <asm/facility.h>
20 #include <asm/current.h>
21 #include <asm/debug.h>
22 #include <asm/ebcdic.h>
23 #include <asm/sysinfo.h>
24 #include <asm/pgtable.h>
25 #include <asm/pgalloc.h>
26 #include <asm/io.h>
27 #include <asm/ptrace.h>
28 #include <asm/compat.h>
29 #include "gaccess.h"
30 #include "kvm-s390.h"
31 #include "trace.h"
32 
33 /* Handle SCK (SET CLOCK) interception */
34 static int handle_set_clock(struct kvm_vcpu *vcpu)
35 {
36 	struct kvm_vcpu *cpup;
37 	s64 hostclk, val;
38 	int i, rc;
39 	u64 op2;
40 
41 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
42 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
43 
44 	op2 = kvm_s390_get_base_disp_s(vcpu);
45 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
46 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
47 	rc = read_guest(vcpu, op2, &val, sizeof(val));
48 	if (rc)
49 		return kvm_s390_inject_prog_cond(vcpu, rc);
50 
51 	if (store_tod_clock(&hostclk)) {
52 		kvm_s390_set_psw_cc(vcpu, 3);
53 		return 0;
54 	}
55 	val = (val - hostclk) & ~0x3fUL;
56 
57 	mutex_lock(&vcpu->kvm->lock);
58 	kvm_for_each_vcpu(i, cpup, vcpu->kvm)
59 		cpup->arch.sie_block->epoch = val;
60 	mutex_unlock(&vcpu->kvm->lock);
61 
62 	kvm_s390_set_psw_cc(vcpu, 0);
63 	return 0;
64 }
65 
66 static int handle_set_prefix(struct kvm_vcpu *vcpu)
67 {
68 	u64 operand2;
69 	u32 address;
70 	int rc;
71 
72 	vcpu->stat.instruction_spx++;
73 
74 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
75 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
76 
77 	operand2 = kvm_s390_get_base_disp_s(vcpu);
78 
79 	/* must be word boundary */
80 	if (operand2 & 3)
81 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
82 
83 	/* get the value */
84 	rc = read_guest(vcpu, operand2, &address, sizeof(address));
85 	if (rc)
86 		return kvm_s390_inject_prog_cond(vcpu, rc);
87 
88 	address &= 0x7fffe000u;
89 
90 	/*
91 	 * Make sure the new value is valid memory. We only need to check the
92 	 * first page, since address is 8k aligned and memory pieces are always
93 	 * at least 1MB aligned and have at least a size of 1MB.
94 	 */
95 	if (kvm_is_error_gpa(vcpu->kvm, address))
96 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
97 
98 	kvm_s390_set_prefix(vcpu, address);
99 
100 	VCPU_EVENT(vcpu, 5, "setting prefix to %x", address);
101 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
102 	return 0;
103 }
104 
105 static int handle_store_prefix(struct kvm_vcpu *vcpu)
106 {
107 	u64 operand2;
108 	u32 address;
109 	int rc;
110 
111 	vcpu->stat.instruction_stpx++;
112 
113 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
114 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
115 
116 	operand2 = kvm_s390_get_base_disp_s(vcpu);
117 
118 	/* must be word boundary */
119 	if (operand2 & 3)
120 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
121 
122 	address = kvm_s390_get_prefix(vcpu);
123 
124 	/* get the value */
125 	rc = write_guest(vcpu, operand2, &address, sizeof(address));
126 	if (rc)
127 		return kvm_s390_inject_prog_cond(vcpu, rc);
128 
129 	VCPU_EVENT(vcpu, 5, "storing prefix to %x", address);
130 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
131 	return 0;
132 }
133 
134 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
135 {
136 	u16 vcpu_id = vcpu->vcpu_id;
137 	u64 ga;
138 	int rc;
139 
140 	vcpu->stat.instruction_stap++;
141 
142 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
143 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
144 
145 	ga = kvm_s390_get_base_disp_s(vcpu);
146 
147 	if (ga & 1)
148 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
149 
150 	rc = write_guest(vcpu, ga, &vcpu_id, sizeof(vcpu_id));
151 	if (rc)
152 		return kvm_s390_inject_prog_cond(vcpu, rc);
153 
154 	VCPU_EVENT(vcpu, 5, "storing cpu address to %llx", ga);
155 	trace_kvm_s390_handle_stap(vcpu, ga);
156 	return 0;
157 }
158 
159 static int __skey_check_enable(struct kvm_vcpu *vcpu)
160 {
161 	int rc = 0;
162 	if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
163 		return rc;
164 
165 	rc = s390_enable_skey();
166 	trace_kvm_s390_skey_related_inst(vcpu);
167 	vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
168 	return rc;
169 }
170 
171 
172 static int handle_skey(struct kvm_vcpu *vcpu)
173 {
174 	int rc = __skey_check_enable(vcpu);
175 
176 	if (rc)
177 		return rc;
178 	vcpu->stat.instruction_storage_key++;
179 
180 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
181 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
182 
183 	vcpu->arch.sie_block->gpsw.addr =
184 		__rewind_psw(vcpu->arch.sie_block->gpsw, 4);
185 	VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
186 	return 0;
187 }
188 
189 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
190 {
191 	psw_t *psw = &vcpu->arch.sie_block->gpsw;
192 
193 	vcpu->stat.instruction_ipte_interlock++;
194 	if (psw_bits(*psw).p)
195 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
196 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
197 	psw->addr = __rewind_psw(*psw, 4);
198 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
199 	return 0;
200 }
201 
202 static int handle_test_block(struct kvm_vcpu *vcpu)
203 {
204 	gpa_t addr;
205 	int reg2;
206 
207 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
208 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
209 
210 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
211 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
212 	addr = kvm_s390_logical_to_effective(vcpu, addr);
213 	if (kvm_s390_check_low_addr_protection(vcpu, addr))
214 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
215 	addr = kvm_s390_real_to_abs(vcpu, addr);
216 
217 	if (kvm_is_error_gpa(vcpu->kvm, addr))
218 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
219 	/*
220 	 * We don't expect errors on modern systems, and do not care
221 	 * about storage keys (yet), so let's just clear the page.
222 	 */
223 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
224 		return -EFAULT;
225 	kvm_s390_set_psw_cc(vcpu, 0);
226 	vcpu->run->s.regs.gprs[0] = 0;
227 	return 0;
228 }
229 
230 static int handle_tpi(struct kvm_vcpu *vcpu)
231 {
232 	struct kvm_s390_interrupt_info *inti;
233 	unsigned long len;
234 	u32 tpi_data[3];
235 	int cc, rc;
236 	u64 addr;
237 
238 	rc = 0;
239 	addr = kvm_s390_get_base_disp_s(vcpu);
240 	if (addr & 3)
241 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
242 	cc = 0;
243 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
244 	if (!inti)
245 		goto no_interrupt;
246 	cc = 1;
247 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
248 	tpi_data[1] = inti->io.io_int_parm;
249 	tpi_data[2] = inti->io.io_int_word;
250 	if (addr) {
251 		/*
252 		 * Store the two-word I/O interruption code into the
253 		 * provided area.
254 		 */
255 		len = sizeof(tpi_data) - 4;
256 		rc = write_guest(vcpu, addr, &tpi_data, len);
257 		if (rc)
258 			return kvm_s390_inject_prog_cond(vcpu, rc);
259 	} else {
260 		/*
261 		 * Store the three-word I/O interruption code into
262 		 * the appropriate lowcore area.
263 		 */
264 		len = sizeof(tpi_data);
265 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len))
266 			rc = -EFAULT;
267 	}
268 	/*
269 	 * If we encounter a problem storing the interruption code, the
270 	 * instruction is suppressed from the guest's view: reinject the
271 	 * interrupt.
272 	 */
273 	if (!rc)
274 		kfree(inti);
275 	else
276 		kvm_s390_reinject_io_int(vcpu->kvm, inti);
277 no_interrupt:
278 	/* Set condition code and we're done. */
279 	if (!rc)
280 		kvm_s390_set_psw_cc(vcpu, cc);
281 	return rc ? -EFAULT : 0;
282 }
283 
284 static int handle_tsch(struct kvm_vcpu *vcpu)
285 {
286 	struct kvm_s390_interrupt_info *inti;
287 
288 	inti = kvm_s390_get_io_int(vcpu->kvm, 0,
289 				   vcpu->run->s.regs.gprs[1]);
290 
291 	/*
292 	 * Prepare exit to userspace.
293 	 * We indicate whether we dequeued a pending I/O interrupt
294 	 * so that userspace can re-inject it if the instruction gets
295 	 * a program check. While this may re-order the pending I/O
296 	 * interrupts, this is no problem since the priority is kept
297 	 * intact.
298 	 */
299 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
300 	vcpu->run->s390_tsch.dequeued = !!inti;
301 	if (inti) {
302 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
303 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
304 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
305 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
306 	}
307 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
308 	kfree(inti);
309 	return -EREMOTE;
310 }
311 
312 static int handle_io_inst(struct kvm_vcpu *vcpu)
313 {
314 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
315 
316 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
317 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
318 
319 	if (vcpu->kvm->arch.css_support) {
320 		/*
321 		 * Most I/O instructions will be handled by userspace.
322 		 * Exceptions are tpi and the interrupt portion of tsch.
323 		 */
324 		if (vcpu->arch.sie_block->ipa == 0xb236)
325 			return handle_tpi(vcpu);
326 		if (vcpu->arch.sie_block->ipa == 0xb235)
327 			return handle_tsch(vcpu);
328 		/* Handle in userspace. */
329 		return -EOPNOTSUPP;
330 	} else {
331 		/*
332 		 * Set condition code 3 to stop the guest from issuing channel
333 		 * I/O instructions.
334 		 */
335 		kvm_s390_set_psw_cc(vcpu, 3);
336 		return 0;
337 	}
338 }
339 
340 static int handle_stfl(struct kvm_vcpu *vcpu)
341 {
342 	int rc;
343 
344 	vcpu->stat.instruction_stfl++;
345 
346 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
347 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
348 
349 	rc = write_guest_lc(vcpu, offsetof(struct _lowcore, stfl_fac_list),
350 			    vfacilities, 4);
351 	if (rc)
352 		return rc;
353 	VCPU_EVENT(vcpu, 5, "store facility list value %x",
354 		   *(unsigned int *) vfacilities);
355 	trace_kvm_s390_handle_stfl(vcpu, *(unsigned int *) vfacilities);
356 	return 0;
357 }
358 
359 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
360 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
361 #define PSW_ADDR_24 0x0000000000ffffffUL
362 #define PSW_ADDR_31 0x000000007fffffffUL
363 
364 int is_valid_psw(psw_t *psw)
365 {
366 	if (psw->mask & PSW_MASK_UNASSIGNED)
367 		return 0;
368 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
369 		if (psw->addr & ~PSW_ADDR_31)
370 			return 0;
371 	}
372 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
373 		return 0;
374 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
375 		return 0;
376 	if (psw->addr & 1)
377 		return 0;
378 	return 1;
379 }
380 
381 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
382 {
383 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
384 	psw_compat_t new_psw;
385 	u64 addr;
386 	int rc;
387 
388 	if (gpsw->mask & PSW_MASK_PSTATE)
389 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
390 
391 	addr = kvm_s390_get_base_disp_s(vcpu);
392 	if (addr & 7)
393 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
394 
395 	rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
396 	if (rc)
397 		return kvm_s390_inject_prog_cond(vcpu, rc);
398 	if (!(new_psw.mask & PSW32_MASK_BASE))
399 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
400 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
401 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
402 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
403 	if (!is_valid_psw(gpsw))
404 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
405 	return 0;
406 }
407 
408 static int handle_lpswe(struct kvm_vcpu *vcpu)
409 {
410 	psw_t new_psw;
411 	u64 addr;
412 	int rc;
413 
414 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
415 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
416 
417 	addr = kvm_s390_get_base_disp_s(vcpu);
418 	if (addr & 7)
419 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
420 	rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
421 	if (rc)
422 		return kvm_s390_inject_prog_cond(vcpu, rc);
423 	vcpu->arch.sie_block->gpsw = new_psw;
424 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
425 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
426 	return 0;
427 }
428 
429 static int handle_stidp(struct kvm_vcpu *vcpu)
430 {
431 	u64 stidp_data = vcpu->arch.stidp_data;
432 	u64 operand2;
433 	int rc;
434 
435 	vcpu->stat.instruction_stidp++;
436 
437 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
438 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
439 
440 	operand2 = kvm_s390_get_base_disp_s(vcpu);
441 
442 	if (operand2 & 7)
443 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
444 
445 	rc = write_guest(vcpu, operand2, &stidp_data, sizeof(stidp_data));
446 	if (rc)
447 		return kvm_s390_inject_prog_cond(vcpu, rc);
448 
449 	VCPU_EVENT(vcpu, 5, "%s", "store cpu id");
450 	return 0;
451 }
452 
453 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
454 {
455 	int cpus = 0;
456 	int n;
457 
458 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
459 
460 	/* deal with other level 3 hypervisors */
461 	if (stsi(mem, 3, 2, 2))
462 		mem->count = 0;
463 	if (mem->count < 8)
464 		mem->count++;
465 	for (n = mem->count - 1; n > 0 ; n--)
466 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
467 
468 	mem->vm[0].cpus_total = cpus;
469 	mem->vm[0].cpus_configured = cpus;
470 	mem->vm[0].cpus_standby = 0;
471 	mem->vm[0].cpus_reserved = 0;
472 	mem->vm[0].caf = 1000;
473 	memcpy(mem->vm[0].name, "KVMguest", 8);
474 	ASCEBC(mem->vm[0].name, 8);
475 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
476 	ASCEBC(mem->vm[0].cpi, 16);
477 }
478 
479 static int handle_stsi(struct kvm_vcpu *vcpu)
480 {
481 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
482 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
483 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
484 	unsigned long mem = 0;
485 	u64 operand2;
486 	int rc = 0;
487 
488 	vcpu->stat.instruction_stsi++;
489 	VCPU_EVENT(vcpu, 4, "stsi: fc: %x sel1: %x sel2: %x", fc, sel1, sel2);
490 
491 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
492 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
493 
494 	if (fc > 3) {
495 		kvm_s390_set_psw_cc(vcpu, 3);
496 		return 0;
497 	}
498 
499 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
500 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
501 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
502 
503 	if (fc == 0) {
504 		vcpu->run->s.regs.gprs[0] = 3 << 28;
505 		kvm_s390_set_psw_cc(vcpu, 0);
506 		return 0;
507 	}
508 
509 	operand2 = kvm_s390_get_base_disp_s(vcpu);
510 
511 	if (operand2 & 0xfff)
512 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
513 
514 	switch (fc) {
515 	case 1: /* same handling for 1 and 2 */
516 	case 2:
517 		mem = get_zeroed_page(GFP_KERNEL);
518 		if (!mem)
519 			goto out_no_data;
520 		if (stsi((void *) mem, fc, sel1, sel2))
521 			goto out_no_data;
522 		break;
523 	case 3:
524 		if (sel1 != 2 || sel2 != 2)
525 			goto out_no_data;
526 		mem = get_zeroed_page(GFP_KERNEL);
527 		if (!mem)
528 			goto out_no_data;
529 		handle_stsi_3_2_2(vcpu, (void *) mem);
530 		break;
531 	}
532 
533 	rc = write_guest(vcpu, operand2, (void *)mem, PAGE_SIZE);
534 	if (rc) {
535 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
536 		goto out;
537 	}
538 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
539 	free_page(mem);
540 	kvm_s390_set_psw_cc(vcpu, 0);
541 	vcpu->run->s.regs.gprs[0] = 0;
542 	return 0;
543 out_no_data:
544 	kvm_s390_set_psw_cc(vcpu, 3);
545 out:
546 	free_page(mem);
547 	return rc;
548 }
549 
550 static const intercept_handler_t b2_handlers[256] = {
551 	[0x02] = handle_stidp,
552 	[0x04] = handle_set_clock,
553 	[0x10] = handle_set_prefix,
554 	[0x11] = handle_store_prefix,
555 	[0x12] = handle_store_cpu_address,
556 	[0x21] = handle_ipte_interlock,
557 	[0x29] = handle_skey,
558 	[0x2a] = handle_skey,
559 	[0x2b] = handle_skey,
560 	[0x2c] = handle_test_block,
561 	[0x30] = handle_io_inst,
562 	[0x31] = handle_io_inst,
563 	[0x32] = handle_io_inst,
564 	[0x33] = handle_io_inst,
565 	[0x34] = handle_io_inst,
566 	[0x35] = handle_io_inst,
567 	[0x36] = handle_io_inst,
568 	[0x37] = handle_io_inst,
569 	[0x38] = handle_io_inst,
570 	[0x39] = handle_io_inst,
571 	[0x3a] = handle_io_inst,
572 	[0x3b] = handle_io_inst,
573 	[0x3c] = handle_io_inst,
574 	[0x50] = handle_ipte_interlock,
575 	[0x5f] = handle_io_inst,
576 	[0x74] = handle_io_inst,
577 	[0x76] = handle_io_inst,
578 	[0x7d] = handle_stsi,
579 	[0xb1] = handle_stfl,
580 	[0xb2] = handle_lpswe,
581 };
582 
583 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
584 {
585 	intercept_handler_t handler;
586 
587 	/*
588 	 * A lot of B2 instructions are priviledged. Here we check for
589 	 * the privileged ones, that we can handle in the kernel.
590 	 * Anything else goes to userspace.
591 	 */
592 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
593 	if (handler)
594 		return handler(vcpu);
595 
596 	return -EOPNOTSUPP;
597 }
598 
599 static int handle_epsw(struct kvm_vcpu *vcpu)
600 {
601 	int reg1, reg2;
602 
603 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
604 
605 	/* This basically extracts the mask half of the psw. */
606 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
607 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
608 	if (reg2) {
609 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
610 		vcpu->run->s.regs.gprs[reg2] |=
611 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
612 	}
613 	return 0;
614 }
615 
616 #define PFMF_RESERVED   0xfffc0101UL
617 #define PFMF_SK         0x00020000UL
618 #define PFMF_CF         0x00010000UL
619 #define PFMF_UI         0x00008000UL
620 #define PFMF_FSC        0x00007000UL
621 #define PFMF_NQ         0x00000800UL
622 #define PFMF_MR         0x00000400UL
623 #define PFMF_MC         0x00000200UL
624 #define PFMF_KEY        0x000000feUL
625 
626 static int handle_pfmf(struct kvm_vcpu *vcpu)
627 {
628 	int reg1, reg2;
629 	unsigned long start, end;
630 
631 	vcpu->stat.instruction_pfmf++;
632 
633 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
634 
635 	if (!MACHINE_HAS_PFMF)
636 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
637 
638 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
639 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
640 
641 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
642 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
643 
644 	/* Only provide non-quiescing support if the host supports it */
645 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
646 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
647 
648 	/* No support for conditional-SSKE */
649 	if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
650 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
651 
652 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
653 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
654 		if (kvm_s390_check_low_addr_protection(vcpu, start))
655 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
656 	}
657 
658 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
659 	case 0x00000000:
660 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
661 		break;
662 	case 0x00001000:
663 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
664 		break;
665 	/* We dont support EDAT2
666 	case 0x00002000:
667 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
668 		break;*/
669 	default:
670 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
671 	}
672 	while (start < end) {
673 		unsigned long useraddr, abs_addr;
674 
675 		/* Translate guest address to host address */
676 		if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
677 			abs_addr = kvm_s390_real_to_abs(vcpu, start);
678 		else
679 			abs_addr = start;
680 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
681 		if (kvm_is_error_hva(useraddr))
682 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
683 
684 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
685 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
686 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
687 		}
688 
689 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
690 			int rc = __skey_check_enable(vcpu);
691 
692 			if (rc)
693 				return rc;
694 			if (set_guest_storage_key(current->mm, useraddr,
695 					vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
696 					vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
697 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
698 		}
699 
700 		start += PAGE_SIZE;
701 	}
702 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
703 		vcpu->run->s.regs.gprs[reg2] = end;
704 	return 0;
705 }
706 
707 static int handle_essa(struct kvm_vcpu *vcpu)
708 {
709 	/* entries expected to be 1FF */
710 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
711 	unsigned long *cbrlo, cbrle;
712 	struct gmap *gmap;
713 	int i;
714 
715 	VCPU_EVENT(vcpu, 5, "cmma release %d pages", entries);
716 	gmap = vcpu->arch.gmap;
717 	vcpu->stat.instruction_essa++;
718 	if (!kvm_s390_cmma_enabled(vcpu->kvm))
719 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
720 
721 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
722 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
723 
724 	if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
725 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
726 
727 	/* Rewind PSW to repeat the ESSA instruction */
728 	vcpu->arch.sie_block->gpsw.addr =
729 		__rewind_psw(vcpu->arch.sie_block->gpsw, 4);
730 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
731 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
732 	down_read(&gmap->mm->mmap_sem);
733 	for (i = 0; i < entries; ++i) {
734 		cbrle = cbrlo[i];
735 		if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
736 			/* invalid entry */
737 			break;
738 		/* try to free backing */
739 		__gmap_zap(gmap, cbrle);
740 	}
741 	up_read(&gmap->mm->mmap_sem);
742 	if (i < entries)
743 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
744 	return 0;
745 }
746 
747 static const intercept_handler_t b9_handlers[256] = {
748 	[0x8a] = handle_ipte_interlock,
749 	[0x8d] = handle_epsw,
750 	[0x8e] = handle_ipte_interlock,
751 	[0x8f] = handle_ipte_interlock,
752 	[0xab] = handle_essa,
753 	[0xaf] = handle_pfmf,
754 };
755 
756 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
757 {
758 	intercept_handler_t handler;
759 
760 	/* This is handled just as for the B2 instructions. */
761 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
762 	if (handler)
763 		return handler(vcpu);
764 
765 	return -EOPNOTSUPP;
766 }
767 
768 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
769 {
770 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
771 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
772 	u32 val = 0;
773 	int reg, rc;
774 	u64 ga;
775 
776 	vcpu->stat.instruction_lctl++;
777 
778 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
779 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
780 
781 	ga = kvm_s390_get_base_disp_rs(vcpu);
782 
783 	if (ga & 3)
784 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
785 
786 	VCPU_EVENT(vcpu, 5, "lctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
787 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
788 
789 	reg = reg1;
790 	do {
791 		rc = read_guest(vcpu, ga, &val, sizeof(val));
792 		if (rc)
793 			return kvm_s390_inject_prog_cond(vcpu, rc);
794 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
795 		vcpu->arch.sie_block->gcr[reg] |= val;
796 		ga += 4;
797 		if (reg == reg3)
798 			break;
799 		reg = (reg + 1) % 16;
800 	} while (1);
801 
802 	return 0;
803 }
804 
805 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
806 {
807 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
808 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
809 	u64 ga;
810 	u32 val;
811 	int reg, rc;
812 
813 	vcpu->stat.instruction_stctl++;
814 
815 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
816 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
817 
818 	ga = kvm_s390_get_base_disp_rs(vcpu);
819 
820 	if (ga & 3)
821 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
822 
823 	VCPU_EVENT(vcpu, 5, "stctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
824 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
825 
826 	reg = reg1;
827 	do {
828 		val = vcpu->arch.sie_block->gcr[reg] &  0x00000000fffffffful;
829 		rc = write_guest(vcpu, ga, &val, sizeof(val));
830 		if (rc)
831 			return kvm_s390_inject_prog_cond(vcpu, rc);
832 		ga += 4;
833 		if (reg == reg3)
834 			break;
835 		reg = (reg + 1) % 16;
836 	} while (1);
837 
838 	return 0;
839 }
840 
841 static int handle_lctlg(struct kvm_vcpu *vcpu)
842 {
843 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
844 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
845 	u64 ga, val;
846 	int reg, rc;
847 
848 	vcpu->stat.instruction_lctlg++;
849 
850 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
851 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
852 
853 	ga = kvm_s390_get_base_disp_rsy(vcpu);
854 
855 	if (ga & 7)
856 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
857 
858 	reg = reg1;
859 
860 	VCPU_EVENT(vcpu, 5, "lctlg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
861 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
862 
863 	do {
864 		rc = read_guest(vcpu, ga, &val, sizeof(val));
865 		if (rc)
866 			return kvm_s390_inject_prog_cond(vcpu, rc);
867 		vcpu->arch.sie_block->gcr[reg] = val;
868 		ga += 8;
869 		if (reg == reg3)
870 			break;
871 		reg = (reg + 1) % 16;
872 	} while (1);
873 
874 	return 0;
875 }
876 
877 static int handle_stctg(struct kvm_vcpu *vcpu)
878 {
879 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
880 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
881 	u64 ga, val;
882 	int reg, rc;
883 
884 	vcpu->stat.instruction_stctg++;
885 
886 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
887 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
888 
889 	ga = kvm_s390_get_base_disp_rsy(vcpu);
890 
891 	if (ga & 7)
892 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
893 
894 	reg = reg1;
895 
896 	VCPU_EVENT(vcpu, 5, "stctg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
897 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
898 
899 	do {
900 		val = vcpu->arch.sie_block->gcr[reg];
901 		rc = write_guest(vcpu, ga, &val, sizeof(val));
902 		if (rc)
903 			return kvm_s390_inject_prog_cond(vcpu, rc);
904 		ga += 8;
905 		if (reg == reg3)
906 			break;
907 		reg = (reg + 1) % 16;
908 	} while (1);
909 
910 	return 0;
911 }
912 
913 static const intercept_handler_t eb_handlers[256] = {
914 	[0x2f] = handle_lctlg,
915 	[0x25] = handle_stctg,
916 };
917 
918 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
919 {
920 	intercept_handler_t handler;
921 
922 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
923 	if (handler)
924 		return handler(vcpu);
925 	return -EOPNOTSUPP;
926 }
927 
928 static int handle_tprot(struct kvm_vcpu *vcpu)
929 {
930 	u64 address1, address2;
931 	unsigned long hva, gpa;
932 	int ret = 0, cc = 0;
933 	bool writable;
934 
935 	vcpu->stat.instruction_tprot++;
936 
937 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
938 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
939 
940 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2);
941 
942 	/* we only handle the Linux memory detection case:
943 	 * access key == 0
944 	 * everything else goes to userspace. */
945 	if (address2 & 0xf0)
946 		return -EOPNOTSUPP;
947 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
948 		ipte_lock(vcpu);
949 	ret = guest_translate_address(vcpu, address1, &gpa, 1);
950 	if (ret == PGM_PROTECTION) {
951 		/* Write protected? Try again with read-only... */
952 		cc = 1;
953 		ret = guest_translate_address(vcpu, address1, &gpa, 0);
954 	}
955 	if (ret) {
956 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
957 			ret = kvm_s390_inject_program_int(vcpu, ret);
958 		} else if (ret > 0) {
959 			/* Translation not available */
960 			kvm_s390_set_psw_cc(vcpu, 3);
961 			ret = 0;
962 		}
963 		goto out_unlock;
964 	}
965 
966 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
967 	if (kvm_is_error_hva(hva)) {
968 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
969 	} else {
970 		if (!writable)
971 			cc = 1;		/* Write not permitted ==> read-only */
972 		kvm_s390_set_psw_cc(vcpu, cc);
973 		/* Note: CC2 only occurs for storage keys (not supported yet) */
974 	}
975 out_unlock:
976 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
977 		ipte_unlock(vcpu);
978 	return ret;
979 }
980 
981 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
982 {
983 	/* For e5xx... instructions we only handle TPROT */
984 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
985 		return handle_tprot(vcpu);
986 	return -EOPNOTSUPP;
987 }
988 
989 static int handle_sckpf(struct kvm_vcpu *vcpu)
990 {
991 	u32 value;
992 
993 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
994 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
995 
996 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
997 		return kvm_s390_inject_program_int(vcpu,
998 						   PGM_SPECIFICATION);
999 
1000 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1001 	vcpu->arch.sie_block->todpr = value;
1002 
1003 	return 0;
1004 }
1005 
1006 static const intercept_handler_t x01_handlers[256] = {
1007 	[0x07] = handle_sckpf,
1008 };
1009 
1010 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1011 {
1012 	intercept_handler_t handler;
1013 
1014 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1015 	if (handler)
1016 		return handler(vcpu);
1017 	return -EOPNOTSUPP;
1018 }
1019