xref: /openbmc/linux/arch/s390/kvm/priv.c (revision 6774def6)
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 void __skey_check_enable(struct kvm_vcpu *vcpu)
160 {
161 	if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
162 		return;
163 
164 	s390_enable_skey();
165 	trace_kvm_s390_skey_related_inst(vcpu);
166 	vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
167 }
168 
169 
170 static int handle_skey(struct kvm_vcpu *vcpu)
171 {
172 	__skey_check_enable(vcpu);
173 
174 	vcpu->stat.instruction_storage_key++;
175 
176 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
177 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
178 
179 	vcpu->arch.sie_block->gpsw.addr =
180 		__rewind_psw(vcpu->arch.sie_block->gpsw, 4);
181 	VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
182 	return 0;
183 }
184 
185 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
186 {
187 	psw_t *psw = &vcpu->arch.sie_block->gpsw;
188 
189 	vcpu->stat.instruction_ipte_interlock++;
190 	if (psw_bits(*psw).p)
191 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
192 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
193 	psw->addr = __rewind_psw(*psw, 4);
194 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
195 	return 0;
196 }
197 
198 static int handle_test_block(struct kvm_vcpu *vcpu)
199 {
200 	gpa_t addr;
201 	int reg2;
202 
203 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
204 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
205 
206 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
207 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
208 	addr = kvm_s390_logical_to_effective(vcpu, addr);
209 	if (kvm_s390_check_low_addr_protection(vcpu, addr))
210 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
211 	addr = kvm_s390_real_to_abs(vcpu, addr);
212 
213 	if (kvm_is_error_gpa(vcpu->kvm, addr))
214 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
215 	/*
216 	 * We don't expect errors on modern systems, and do not care
217 	 * about storage keys (yet), so let's just clear the page.
218 	 */
219 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
220 		return -EFAULT;
221 	kvm_s390_set_psw_cc(vcpu, 0);
222 	vcpu->run->s.regs.gprs[0] = 0;
223 	return 0;
224 }
225 
226 static int handle_tpi(struct kvm_vcpu *vcpu)
227 {
228 	struct kvm_s390_interrupt_info *inti;
229 	unsigned long len;
230 	u32 tpi_data[3];
231 	int cc, rc;
232 	u64 addr;
233 
234 	rc = 0;
235 	addr = kvm_s390_get_base_disp_s(vcpu);
236 	if (addr & 3)
237 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
238 	cc = 0;
239 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
240 	if (!inti)
241 		goto no_interrupt;
242 	cc = 1;
243 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
244 	tpi_data[1] = inti->io.io_int_parm;
245 	tpi_data[2] = inti->io.io_int_word;
246 	if (addr) {
247 		/*
248 		 * Store the two-word I/O interruption code into the
249 		 * provided area.
250 		 */
251 		len = sizeof(tpi_data) - 4;
252 		rc = write_guest(vcpu, addr, &tpi_data, len);
253 		if (rc)
254 			return kvm_s390_inject_prog_cond(vcpu, rc);
255 	} else {
256 		/*
257 		 * Store the three-word I/O interruption code into
258 		 * the appropriate lowcore area.
259 		 */
260 		len = sizeof(tpi_data);
261 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len))
262 			rc = -EFAULT;
263 	}
264 	/*
265 	 * If we encounter a problem storing the interruption code, the
266 	 * instruction is suppressed from the guest's view: reinject the
267 	 * interrupt.
268 	 */
269 	if (!rc)
270 		kfree(inti);
271 	else
272 		kvm_s390_reinject_io_int(vcpu->kvm, inti);
273 no_interrupt:
274 	/* Set condition code and we're done. */
275 	if (!rc)
276 		kvm_s390_set_psw_cc(vcpu, cc);
277 	return rc ? -EFAULT : 0;
278 }
279 
280 static int handle_tsch(struct kvm_vcpu *vcpu)
281 {
282 	struct kvm_s390_interrupt_info *inti;
283 
284 	inti = kvm_s390_get_io_int(vcpu->kvm, 0,
285 				   vcpu->run->s.regs.gprs[1]);
286 
287 	/*
288 	 * Prepare exit to userspace.
289 	 * We indicate whether we dequeued a pending I/O interrupt
290 	 * so that userspace can re-inject it if the instruction gets
291 	 * a program check. While this may re-order the pending I/O
292 	 * interrupts, this is no problem since the priority is kept
293 	 * intact.
294 	 */
295 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
296 	vcpu->run->s390_tsch.dequeued = !!inti;
297 	if (inti) {
298 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
299 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
300 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
301 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
302 	}
303 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
304 	kfree(inti);
305 	return -EREMOTE;
306 }
307 
308 static int handle_io_inst(struct kvm_vcpu *vcpu)
309 {
310 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
311 
312 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
313 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
314 
315 	if (vcpu->kvm->arch.css_support) {
316 		/*
317 		 * Most I/O instructions will be handled by userspace.
318 		 * Exceptions are tpi and the interrupt portion of tsch.
319 		 */
320 		if (vcpu->arch.sie_block->ipa == 0xb236)
321 			return handle_tpi(vcpu);
322 		if (vcpu->arch.sie_block->ipa == 0xb235)
323 			return handle_tsch(vcpu);
324 		/* Handle in userspace. */
325 		return -EOPNOTSUPP;
326 	} else {
327 		/*
328 		 * Set condition code 3 to stop the guest from issuing channel
329 		 * I/O instructions.
330 		 */
331 		kvm_s390_set_psw_cc(vcpu, 3);
332 		return 0;
333 	}
334 }
335 
336 static int handle_stfl(struct kvm_vcpu *vcpu)
337 {
338 	int rc;
339 
340 	vcpu->stat.instruction_stfl++;
341 
342 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
343 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
344 
345 	rc = write_guest_lc(vcpu, offsetof(struct _lowcore, stfl_fac_list),
346 			    vfacilities, 4);
347 	if (rc)
348 		return rc;
349 	VCPU_EVENT(vcpu, 5, "store facility list value %x",
350 		   *(unsigned int *) vfacilities);
351 	trace_kvm_s390_handle_stfl(vcpu, *(unsigned int *) vfacilities);
352 	return 0;
353 }
354 
355 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
356 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
357 #define PSW_ADDR_24 0x0000000000ffffffUL
358 #define PSW_ADDR_31 0x000000007fffffffUL
359 
360 int is_valid_psw(psw_t *psw)
361 {
362 	if (psw->mask & PSW_MASK_UNASSIGNED)
363 		return 0;
364 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
365 		if (psw->addr & ~PSW_ADDR_31)
366 			return 0;
367 	}
368 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
369 		return 0;
370 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
371 		return 0;
372 	if (psw->addr & 1)
373 		return 0;
374 	return 1;
375 }
376 
377 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
378 {
379 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
380 	psw_compat_t new_psw;
381 	u64 addr;
382 	int rc;
383 
384 	if (gpsw->mask & PSW_MASK_PSTATE)
385 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
386 
387 	addr = kvm_s390_get_base_disp_s(vcpu);
388 	if (addr & 7)
389 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
390 
391 	rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
392 	if (rc)
393 		return kvm_s390_inject_prog_cond(vcpu, rc);
394 	if (!(new_psw.mask & PSW32_MASK_BASE))
395 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
396 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
397 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
398 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
399 	if (!is_valid_psw(gpsw))
400 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
401 	return 0;
402 }
403 
404 static int handle_lpswe(struct kvm_vcpu *vcpu)
405 {
406 	psw_t new_psw;
407 	u64 addr;
408 	int rc;
409 
410 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
411 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
412 
413 	addr = kvm_s390_get_base_disp_s(vcpu);
414 	if (addr & 7)
415 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
416 	rc = read_guest(vcpu, addr, &new_psw, sizeof(new_psw));
417 	if (rc)
418 		return kvm_s390_inject_prog_cond(vcpu, rc);
419 	vcpu->arch.sie_block->gpsw = new_psw;
420 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
421 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
422 	return 0;
423 }
424 
425 static int handle_stidp(struct kvm_vcpu *vcpu)
426 {
427 	u64 stidp_data = vcpu->arch.stidp_data;
428 	u64 operand2;
429 	int rc;
430 
431 	vcpu->stat.instruction_stidp++;
432 
433 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
434 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
435 
436 	operand2 = kvm_s390_get_base_disp_s(vcpu);
437 
438 	if (operand2 & 7)
439 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
440 
441 	rc = write_guest(vcpu, operand2, &stidp_data, sizeof(stidp_data));
442 	if (rc)
443 		return kvm_s390_inject_prog_cond(vcpu, rc);
444 
445 	VCPU_EVENT(vcpu, 5, "%s", "store cpu id");
446 	return 0;
447 }
448 
449 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
450 {
451 	int cpus = 0;
452 	int n;
453 
454 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
455 
456 	/* deal with other level 3 hypervisors */
457 	if (stsi(mem, 3, 2, 2))
458 		mem->count = 0;
459 	if (mem->count < 8)
460 		mem->count++;
461 	for (n = mem->count - 1; n > 0 ; n--)
462 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
463 
464 	mem->vm[0].cpus_total = cpus;
465 	mem->vm[0].cpus_configured = cpus;
466 	mem->vm[0].cpus_standby = 0;
467 	mem->vm[0].cpus_reserved = 0;
468 	mem->vm[0].caf = 1000;
469 	memcpy(mem->vm[0].name, "KVMguest", 8);
470 	ASCEBC(mem->vm[0].name, 8);
471 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
472 	ASCEBC(mem->vm[0].cpi, 16);
473 }
474 
475 static int handle_stsi(struct kvm_vcpu *vcpu)
476 {
477 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
478 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
479 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
480 	unsigned long mem = 0;
481 	u64 operand2;
482 	int rc = 0;
483 
484 	vcpu->stat.instruction_stsi++;
485 	VCPU_EVENT(vcpu, 4, "stsi: fc: %x sel1: %x sel2: %x", fc, sel1, sel2);
486 
487 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
488 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
489 
490 	if (fc > 3) {
491 		kvm_s390_set_psw_cc(vcpu, 3);
492 		return 0;
493 	}
494 
495 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
496 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
497 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
498 
499 	if (fc == 0) {
500 		vcpu->run->s.regs.gprs[0] = 3 << 28;
501 		kvm_s390_set_psw_cc(vcpu, 0);
502 		return 0;
503 	}
504 
505 	operand2 = kvm_s390_get_base_disp_s(vcpu);
506 
507 	if (operand2 & 0xfff)
508 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
509 
510 	switch (fc) {
511 	case 1: /* same handling for 1 and 2 */
512 	case 2:
513 		mem = get_zeroed_page(GFP_KERNEL);
514 		if (!mem)
515 			goto out_no_data;
516 		if (stsi((void *) mem, fc, sel1, sel2))
517 			goto out_no_data;
518 		break;
519 	case 3:
520 		if (sel1 != 2 || sel2 != 2)
521 			goto out_no_data;
522 		mem = get_zeroed_page(GFP_KERNEL);
523 		if (!mem)
524 			goto out_no_data;
525 		handle_stsi_3_2_2(vcpu, (void *) mem);
526 		break;
527 	}
528 
529 	rc = write_guest(vcpu, operand2, (void *)mem, PAGE_SIZE);
530 	if (rc) {
531 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
532 		goto out;
533 	}
534 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
535 	free_page(mem);
536 	kvm_s390_set_psw_cc(vcpu, 0);
537 	vcpu->run->s.regs.gprs[0] = 0;
538 	return 0;
539 out_no_data:
540 	kvm_s390_set_psw_cc(vcpu, 3);
541 out:
542 	free_page(mem);
543 	return rc;
544 }
545 
546 static const intercept_handler_t b2_handlers[256] = {
547 	[0x02] = handle_stidp,
548 	[0x04] = handle_set_clock,
549 	[0x10] = handle_set_prefix,
550 	[0x11] = handle_store_prefix,
551 	[0x12] = handle_store_cpu_address,
552 	[0x21] = handle_ipte_interlock,
553 	[0x29] = handle_skey,
554 	[0x2a] = handle_skey,
555 	[0x2b] = handle_skey,
556 	[0x2c] = handle_test_block,
557 	[0x30] = handle_io_inst,
558 	[0x31] = handle_io_inst,
559 	[0x32] = handle_io_inst,
560 	[0x33] = handle_io_inst,
561 	[0x34] = handle_io_inst,
562 	[0x35] = handle_io_inst,
563 	[0x36] = handle_io_inst,
564 	[0x37] = handle_io_inst,
565 	[0x38] = handle_io_inst,
566 	[0x39] = handle_io_inst,
567 	[0x3a] = handle_io_inst,
568 	[0x3b] = handle_io_inst,
569 	[0x3c] = handle_io_inst,
570 	[0x50] = handle_ipte_interlock,
571 	[0x5f] = handle_io_inst,
572 	[0x74] = handle_io_inst,
573 	[0x76] = handle_io_inst,
574 	[0x7d] = handle_stsi,
575 	[0xb1] = handle_stfl,
576 	[0xb2] = handle_lpswe,
577 };
578 
579 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
580 {
581 	intercept_handler_t handler;
582 
583 	/*
584 	 * A lot of B2 instructions are priviledged. Here we check for
585 	 * the privileged ones, that we can handle in the kernel.
586 	 * Anything else goes to userspace.
587 	 */
588 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
589 	if (handler)
590 		return handler(vcpu);
591 
592 	return -EOPNOTSUPP;
593 }
594 
595 static int handle_epsw(struct kvm_vcpu *vcpu)
596 {
597 	int reg1, reg2;
598 
599 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
600 
601 	/* This basically extracts the mask half of the psw. */
602 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
603 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
604 	if (reg2) {
605 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
606 		vcpu->run->s.regs.gprs[reg2] |=
607 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
608 	}
609 	return 0;
610 }
611 
612 #define PFMF_RESERVED   0xfffc0101UL
613 #define PFMF_SK         0x00020000UL
614 #define PFMF_CF         0x00010000UL
615 #define PFMF_UI         0x00008000UL
616 #define PFMF_FSC        0x00007000UL
617 #define PFMF_NQ         0x00000800UL
618 #define PFMF_MR         0x00000400UL
619 #define PFMF_MC         0x00000200UL
620 #define PFMF_KEY        0x000000feUL
621 
622 static int handle_pfmf(struct kvm_vcpu *vcpu)
623 {
624 	int reg1, reg2;
625 	unsigned long start, end;
626 
627 	vcpu->stat.instruction_pfmf++;
628 
629 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
630 
631 	if (!MACHINE_HAS_PFMF)
632 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
633 
634 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
635 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
636 
637 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
638 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
639 
640 	/* Only provide non-quiescing support if the host supports it */
641 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
642 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
643 
644 	/* No support for conditional-SSKE */
645 	if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
646 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
647 
648 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
649 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
650 		if (kvm_s390_check_low_addr_protection(vcpu, start))
651 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
652 	}
653 
654 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
655 	case 0x00000000:
656 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
657 		break;
658 	case 0x00001000:
659 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
660 		break;
661 	/* We dont support EDAT2
662 	case 0x00002000:
663 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
664 		break;*/
665 	default:
666 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
667 	}
668 	while (start < end) {
669 		unsigned long useraddr, abs_addr;
670 
671 		/* Translate guest address to host address */
672 		if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
673 			abs_addr = kvm_s390_real_to_abs(vcpu, start);
674 		else
675 			abs_addr = start;
676 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
677 		if (kvm_is_error_hva(useraddr))
678 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
679 
680 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
681 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
682 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
683 		}
684 
685 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
686 			__skey_check_enable(vcpu);
687 			if (set_guest_storage_key(current->mm, useraddr,
688 					vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
689 					vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
690 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
691 		}
692 
693 		start += PAGE_SIZE;
694 	}
695 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
696 		vcpu->run->s.regs.gprs[reg2] = end;
697 	return 0;
698 }
699 
700 static int handle_essa(struct kvm_vcpu *vcpu)
701 {
702 	/* entries expected to be 1FF */
703 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
704 	unsigned long *cbrlo, cbrle;
705 	struct gmap *gmap;
706 	int i;
707 
708 	VCPU_EVENT(vcpu, 5, "cmma release %d pages", entries);
709 	gmap = vcpu->arch.gmap;
710 	vcpu->stat.instruction_essa++;
711 	if (!kvm_s390_cmma_enabled(vcpu->kvm))
712 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
713 
714 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
715 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
716 
717 	if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
718 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
719 
720 	/* Rewind PSW to repeat the ESSA instruction */
721 	vcpu->arch.sie_block->gpsw.addr =
722 		__rewind_psw(vcpu->arch.sie_block->gpsw, 4);
723 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
724 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
725 	down_read(&gmap->mm->mmap_sem);
726 	for (i = 0; i < entries; ++i) {
727 		cbrle = cbrlo[i];
728 		if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
729 			/* invalid entry */
730 			break;
731 		/* try to free backing */
732 		__gmap_zap(gmap, cbrle);
733 	}
734 	up_read(&gmap->mm->mmap_sem);
735 	if (i < entries)
736 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
737 	return 0;
738 }
739 
740 static const intercept_handler_t b9_handlers[256] = {
741 	[0x8a] = handle_ipte_interlock,
742 	[0x8d] = handle_epsw,
743 	[0x8e] = handle_ipte_interlock,
744 	[0x8f] = handle_ipte_interlock,
745 	[0xab] = handle_essa,
746 	[0xaf] = handle_pfmf,
747 };
748 
749 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
750 {
751 	intercept_handler_t handler;
752 
753 	/* This is handled just as for the B2 instructions. */
754 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
755 	if (handler)
756 		return handler(vcpu);
757 
758 	return -EOPNOTSUPP;
759 }
760 
761 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
762 {
763 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
764 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
765 	u32 val = 0;
766 	int reg, rc;
767 	u64 ga;
768 
769 	vcpu->stat.instruction_lctl++;
770 
771 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
772 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
773 
774 	ga = kvm_s390_get_base_disp_rs(vcpu);
775 
776 	if (ga & 3)
777 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
778 
779 	VCPU_EVENT(vcpu, 5, "lctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
780 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
781 
782 	reg = reg1;
783 	do {
784 		rc = read_guest(vcpu, ga, &val, sizeof(val));
785 		if (rc)
786 			return kvm_s390_inject_prog_cond(vcpu, rc);
787 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
788 		vcpu->arch.sie_block->gcr[reg] |= val;
789 		ga += 4;
790 		if (reg == reg3)
791 			break;
792 		reg = (reg + 1) % 16;
793 	} while (1);
794 
795 	return 0;
796 }
797 
798 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
799 {
800 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
801 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
802 	u64 ga;
803 	u32 val;
804 	int reg, rc;
805 
806 	vcpu->stat.instruction_stctl++;
807 
808 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
809 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
810 
811 	ga = kvm_s390_get_base_disp_rs(vcpu);
812 
813 	if (ga & 3)
814 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
815 
816 	VCPU_EVENT(vcpu, 5, "stctl r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
817 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
818 
819 	reg = reg1;
820 	do {
821 		val = vcpu->arch.sie_block->gcr[reg] &  0x00000000fffffffful;
822 		rc = write_guest(vcpu, ga, &val, sizeof(val));
823 		if (rc)
824 			return kvm_s390_inject_prog_cond(vcpu, rc);
825 		ga += 4;
826 		if (reg == reg3)
827 			break;
828 		reg = (reg + 1) % 16;
829 	} while (1);
830 
831 	return 0;
832 }
833 
834 static int handle_lctlg(struct kvm_vcpu *vcpu)
835 {
836 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
837 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
838 	u64 ga, val;
839 	int reg, rc;
840 
841 	vcpu->stat.instruction_lctlg++;
842 
843 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
844 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
845 
846 	ga = kvm_s390_get_base_disp_rsy(vcpu);
847 
848 	if (ga & 7)
849 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
850 
851 	reg = reg1;
852 
853 	VCPU_EVENT(vcpu, 5, "lctlg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
854 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
855 
856 	do {
857 		rc = read_guest(vcpu, ga, &val, sizeof(val));
858 		if (rc)
859 			return kvm_s390_inject_prog_cond(vcpu, rc);
860 		vcpu->arch.sie_block->gcr[reg] = val;
861 		ga += 8;
862 		if (reg == reg3)
863 			break;
864 		reg = (reg + 1) % 16;
865 	} while (1);
866 
867 	return 0;
868 }
869 
870 static int handle_stctg(struct kvm_vcpu *vcpu)
871 {
872 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
873 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
874 	u64 ga, val;
875 	int reg, rc;
876 
877 	vcpu->stat.instruction_stctg++;
878 
879 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
880 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
881 
882 	ga = kvm_s390_get_base_disp_rsy(vcpu);
883 
884 	if (ga & 7)
885 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
886 
887 	reg = reg1;
888 
889 	VCPU_EVENT(vcpu, 5, "stctg r1:%x, r3:%x, addr:%llx", reg1, reg3, ga);
890 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
891 
892 	do {
893 		val = vcpu->arch.sie_block->gcr[reg];
894 		rc = write_guest(vcpu, ga, &val, sizeof(val));
895 		if (rc)
896 			return kvm_s390_inject_prog_cond(vcpu, rc);
897 		ga += 8;
898 		if (reg == reg3)
899 			break;
900 		reg = (reg + 1) % 16;
901 	} while (1);
902 
903 	return 0;
904 }
905 
906 static const intercept_handler_t eb_handlers[256] = {
907 	[0x2f] = handle_lctlg,
908 	[0x25] = handle_stctg,
909 };
910 
911 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
912 {
913 	intercept_handler_t handler;
914 
915 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
916 	if (handler)
917 		return handler(vcpu);
918 	return -EOPNOTSUPP;
919 }
920 
921 static int handle_tprot(struct kvm_vcpu *vcpu)
922 {
923 	u64 address1, address2;
924 	unsigned long hva, gpa;
925 	int ret = 0, cc = 0;
926 	bool writable;
927 
928 	vcpu->stat.instruction_tprot++;
929 
930 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
931 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
932 
933 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2);
934 
935 	/* we only handle the Linux memory detection case:
936 	 * access key == 0
937 	 * everything else goes to userspace. */
938 	if (address2 & 0xf0)
939 		return -EOPNOTSUPP;
940 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
941 		ipte_lock(vcpu);
942 	ret = guest_translate_address(vcpu, address1, &gpa, 1);
943 	if (ret == PGM_PROTECTION) {
944 		/* Write protected? Try again with read-only... */
945 		cc = 1;
946 		ret = guest_translate_address(vcpu, address1, &gpa, 0);
947 	}
948 	if (ret) {
949 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
950 			ret = kvm_s390_inject_program_int(vcpu, ret);
951 		} else if (ret > 0) {
952 			/* Translation not available */
953 			kvm_s390_set_psw_cc(vcpu, 3);
954 			ret = 0;
955 		}
956 		goto out_unlock;
957 	}
958 
959 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
960 	if (kvm_is_error_hva(hva)) {
961 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
962 	} else {
963 		if (!writable)
964 			cc = 1;		/* Write not permitted ==> read-only */
965 		kvm_s390_set_psw_cc(vcpu, cc);
966 		/* Note: CC2 only occurs for storage keys (not supported yet) */
967 	}
968 out_unlock:
969 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
970 		ipte_unlock(vcpu);
971 	return ret;
972 }
973 
974 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
975 {
976 	/* For e5xx... instructions we only handle TPROT */
977 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
978 		return handle_tprot(vcpu);
979 	return -EOPNOTSUPP;
980 }
981 
982 static int handle_sckpf(struct kvm_vcpu *vcpu)
983 {
984 	u32 value;
985 
986 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
987 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
988 
989 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
990 		return kvm_s390_inject_program_int(vcpu,
991 						   PGM_SPECIFICATION);
992 
993 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
994 	vcpu->arch.sie_block->todpr = value;
995 
996 	return 0;
997 }
998 
999 static const intercept_handler_t x01_handlers[256] = {
1000 	[0x07] = handle_sckpf,
1001 };
1002 
1003 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1004 {
1005 	intercept_handler_t handler;
1006 
1007 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1008 	if (handler)
1009 		return handler(vcpu);
1010 	return -EOPNOTSUPP;
1011 }
1012