xref: /openbmc/linux/arch/s390/kvm/priv.c (revision 4f205687)
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/gmap.h>
27 #include <asm/io.h>
28 #include <asm/ptrace.h>
29 #include <asm/compat.h>
30 #include "gaccess.h"
31 #include "kvm-s390.h"
32 #include "trace.h"
33 
34 /* Handle SCK (SET CLOCK) interception */
35 static int handle_set_clock(struct kvm_vcpu *vcpu)
36 {
37 	int rc;
38 	ar_t ar;
39 	u64 op2, val;
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, &ar);
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, ar, &val, sizeof(val));
48 	if (rc)
49 		return kvm_s390_inject_prog_cond(vcpu, rc);
50 
51 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
52 	kvm_s390_set_tod_clock(vcpu->kvm, val);
53 
54 	kvm_s390_set_psw_cc(vcpu, 0);
55 	return 0;
56 }
57 
58 static int handle_set_prefix(struct kvm_vcpu *vcpu)
59 {
60 	u64 operand2;
61 	u32 address;
62 	int rc;
63 	ar_t ar;
64 
65 	vcpu->stat.instruction_spx++;
66 
67 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
68 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
69 
70 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
71 
72 	/* must be word boundary */
73 	if (operand2 & 3)
74 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
75 
76 	/* get the value */
77 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
78 	if (rc)
79 		return kvm_s390_inject_prog_cond(vcpu, rc);
80 
81 	address &= 0x7fffe000u;
82 
83 	/*
84 	 * Make sure the new value is valid memory. We only need to check the
85 	 * first page, since address is 8k aligned and memory pieces are always
86 	 * at least 1MB aligned and have at least a size of 1MB.
87 	 */
88 	if (kvm_is_error_gpa(vcpu->kvm, address))
89 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
90 
91 	kvm_s390_set_prefix(vcpu, address);
92 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
93 	return 0;
94 }
95 
96 static int handle_store_prefix(struct kvm_vcpu *vcpu)
97 {
98 	u64 operand2;
99 	u32 address;
100 	int rc;
101 	ar_t ar;
102 
103 	vcpu->stat.instruction_stpx++;
104 
105 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
106 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
107 
108 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
109 
110 	/* must be word boundary */
111 	if (operand2 & 3)
112 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
113 
114 	address = kvm_s390_get_prefix(vcpu);
115 
116 	/* get the value */
117 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
118 	if (rc)
119 		return kvm_s390_inject_prog_cond(vcpu, rc);
120 
121 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
122 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
123 	return 0;
124 }
125 
126 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
127 {
128 	u16 vcpu_id = vcpu->vcpu_id;
129 	u64 ga;
130 	int rc;
131 	ar_t ar;
132 
133 	vcpu->stat.instruction_stap++;
134 
135 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
136 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
137 
138 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
139 
140 	if (ga & 1)
141 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
142 
143 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
144 	if (rc)
145 		return kvm_s390_inject_prog_cond(vcpu, rc);
146 
147 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
148 	trace_kvm_s390_handle_stap(vcpu, ga);
149 	return 0;
150 }
151 
152 static int __skey_check_enable(struct kvm_vcpu *vcpu)
153 {
154 	int rc = 0;
155 	if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
156 		return rc;
157 
158 	rc = s390_enable_skey();
159 	VCPU_EVENT(vcpu, 3, "%s", "enabling storage keys for guest");
160 	trace_kvm_s390_skey_related_inst(vcpu);
161 	vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
162 	return rc;
163 }
164 
165 
166 static int handle_skey(struct kvm_vcpu *vcpu)
167 {
168 	int rc = __skey_check_enable(vcpu);
169 
170 	if (rc)
171 		return rc;
172 	vcpu->stat.instruction_storage_key++;
173 
174 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
175 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
176 
177 	kvm_s390_retry_instr(vcpu);
178 	VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
179 	return 0;
180 }
181 
182 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
183 {
184 	vcpu->stat.instruction_ipte_interlock++;
185 	if (psw_bits(vcpu->arch.sie_block->gpsw).p)
186 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
187 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
188 	kvm_s390_retry_instr(vcpu);
189 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
190 	return 0;
191 }
192 
193 static int handle_test_block(struct kvm_vcpu *vcpu)
194 {
195 	gpa_t addr;
196 	int reg2;
197 
198 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
199 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
200 
201 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
202 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
203 	addr = kvm_s390_logical_to_effective(vcpu, addr);
204 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
205 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
206 	addr = kvm_s390_real_to_abs(vcpu, addr);
207 
208 	if (kvm_is_error_gpa(vcpu->kvm, addr))
209 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
210 	/*
211 	 * We don't expect errors on modern systems, and do not care
212 	 * about storage keys (yet), so let's just clear the page.
213 	 */
214 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
215 		return -EFAULT;
216 	kvm_s390_set_psw_cc(vcpu, 0);
217 	vcpu->run->s.regs.gprs[0] = 0;
218 	return 0;
219 }
220 
221 static int handle_tpi(struct kvm_vcpu *vcpu)
222 {
223 	struct kvm_s390_interrupt_info *inti;
224 	unsigned long len;
225 	u32 tpi_data[3];
226 	int rc;
227 	u64 addr;
228 	ar_t ar;
229 
230 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
231 	if (addr & 3)
232 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
233 
234 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
235 	if (!inti) {
236 		kvm_s390_set_psw_cc(vcpu, 0);
237 		return 0;
238 	}
239 
240 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
241 	tpi_data[1] = inti->io.io_int_parm;
242 	tpi_data[2] = inti->io.io_int_word;
243 	if (addr) {
244 		/*
245 		 * Store the two-word I/O interruption code into the
246 		 * provided area.
247 		 */
248 		len = sizeof(tpi_data) - 4;
249 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
250 		if (rc) {
251 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
252 			goto reinject_interrupt;
253 		}
254 	} else {
255 		/*
256 		 * Store the three-word I/O interruption code into
257 		 * the appropriate lowcore area.
258 		 */
259 		len = sizeof(tpi_data);
260 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
261 			/* failed writes to the low core are not recoverable */
262 			rc = -EFAULT;
263 			goto reinject_interrupt;
264 		}
265 	}
266 
267 	/* irq was successfully handed to the guest */
268 	kfree(inti);
269 	kvm_s390_set_psw_cc(vcpu, 1);
270 	return 0;
271 reinject_interrupt:
272 	/*
273 	 * If we encounter a problem storing the interruption code, the
274 	 * instruction is suppressed from the guest's view: reinject the
275 	 * interrupt.
276 	 */
277 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
278 		kfree(inti);
279 		rc = -EFAULT;
280 	}
281 	/* don't set the cc, a pgm irq was injected or we drop to user space */
282 	return rc ? -EFAULT : 0;
283 }
284 
285 static int handle_tsch(struct kvm_vcpu *vcpu)
286 {
287 	struct kvm_s390_interrupt_info *inti = NULL;
288 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
289 
290 	/* a valid schid has at least one bit set */
291 	if (vcpu->run->s.regs.gprs[1])
292 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
293 					   vcpu->run->s.regs.gprs[1]);
294 
295 	/*
296 	 * Prepare exit to userspace.
297 	 * We indicate whether we dequeued a pending I/O interrupt
298 	 * so that userspace can re-inject it if the instruction gets
299 	 * a program check. While this may re-order the pending I/O
300 	 * interrupts, this is no problem since the priority is kept
301 	 * intact.
302 	 */
303 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
304 	vcpu->run->s390_tsch.dequeued = !!inti;
305 	if (inti) {
306 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
307 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
308 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
309 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
310 	}
311 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
312 	kfree(inti);
313 	return -EREMOTE;
314 }
315 
316 static int handle_io_inst(struct kvm_vcpu *vcpu)
317 {
318 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
319 
320 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
321 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
322 
323 	if (vcpu->kvm->arch.css_support) {
324 		/*
325 		 * Most I/O instructions will be handled by userspace.
326 		 * Exceptions are tpi and the interrupt portion of tsch.
327 		 */
328 		if (vcpu->arch.sie_block->ipa == 0xb236)
329 			return handle_tpi(vcpu);
330 		if (vcpu->arch.sie_block->ipa == 0xb235)
331 			return handle_tsch(vcpu);
332 		/* Handle in userspace. */
333 		return -EOPNOTSUPP;
334 	} else {
335 		/*
336 		 * Set condition code 3 to stop the guest from issuing channel
337 		 * I/O instructions.
338 		 */
339 		kvm_s390_set_psw_cc(vcpu, 3);
340 		return 0;
341 	}
342 }
343 
344 static int handle_stfl(struct kvm_vcpu *vcpu)
345 {
346 	int rc;
347 	unsigned int fac;
348 
349 	vcpu->stat.instruction_stfl++;
350 
351 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
352 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
353 
354 	/*
355 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
356 	 * into a u32 memory representation. They will remain bits 0-31.
357 	 */
358 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
359 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
360 			    &fac, sizeof(fac));
361 	if (rc)
362 		return rc;
363 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
364 	trace_kvm_s390_handle_stfl(vcpu, fac);
365 	return 0;
366 }
367 
368 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
369 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
370 #define PSW_ADDR_24 0x0000000000ffffffUL
371 #define PSW_ADDR_31 0x000000007fffffffUL
372 
373 int is_valid_psw(psw_t *psw)
374 {
375 	if (psw->mask & PSW_MASK_UNASSIGNED)
376 		return 0;
377 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
378 		if (psw->addr & ~PSW_ADDR_31)
379 			return 0;
380 	}
381 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
382 		return 0;
383 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
384 		return 0;
385 	if (psw->addr & 1)
386 		return 0;
387 	return 1;
388 }
389 
390 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
391 {
392 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
393 	psw_compat_t new_psw;
394 	u64 addr;
395 	int rc;
396 	ar_t ar;
397 
398 	if (gpsw->mask & PSW_MASK_PSTATE)
399 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
400 
401 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
402 	if (addr & 7)
403 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
404 
405 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
406 	if (rc)
407 		return kvm_s390_inject_prog_cond(vcpu, rc);
408 	if (!(new_psw.mask & PSW32_MASK_BASE))
409 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
410 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
411 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
412 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
413 	if (!is_valid_psw(gpsw))
414 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
415 	return 0;
416 }
417 
418 static int handle_lpswe(struct kvm_vcpu *vcpu)
419 {
420 	psw_t new_psw;
421 	u64 addr;
422 	int rc;
423 	ar_t ar;
424 
425 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
426 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
427 
428 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
429 	if (addr & 7)
430 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
431 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
432 	if (rc)
433 		return kvm_s390_inject_prog_cond(vcpu, rc);
434 	vcpu->arch.sie_block->gpsw = new_psw;
435 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
436 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
437 	return 0;
438 }
439 
440 static int handle_stidp(struct kvm_vcpu *vcpu)
441 {
442 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
443 	u64 operand2;
444 	int rc;
445 	ar_t ar;
446 
447 	vcpu->stat.instruction_stidp++;
448 
449 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
450 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
451 
452 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
453 
454 	if (operand2 & 7)
455 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
456 
457 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
458 	if (rc)
459 		return kvm_s390_inject_prog_cond(vcpu, rc);
460 
461 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
462 	return 0;
463 }
464 
465 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
466 {
467 	int cpus = 0;
468 	int n;
469 
470 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
471 
472 	/* deal with other level 3 hypervisors */
473 	if (stsi(mem, 3, 2, 2))
474 		mem->count = 0;
475 	if (mem->count < 8)
476 		mem->count++;
477 	for (n = mem->count - 1; n > 0 ; n--)
478 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
479 
480 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
481 	mem->vm[0].cpus_total = cpus;
482 	mem->vm[0].cpus_configured = cpus;
483 	mem->vm[0].cpus_standby = 0;
484 	mem->vm[0].cpus_reserved = 0;
485 	mem->vm[0].caf = 1000;
486 	memcpy(mem->vm[0].name, "KVMguest", 8);
487 	ASCEBC(mem->vm[0].name, 8);
488 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
489 	ASCEBC(mem->vm[0].cpi, 16);
490 }
491 
492 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, ar_t ar,
493 				 u8 fc, u8 sel1, u16 sel2)
494 {
495 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
496 	vcpu->run->s390_stsi.addr = addr;
497 	vcpu->run->s390_stsi.ar = ar;
498 	vcpu->run->s390_stsi.fc = fc;
499 	vcpu->run->s390_stsi.sel1 = sel1;
500 	vcpu->run->s390_stsi.sel2 = sel2;
501 }
502 
503 static int handle_stsi(struct kvm_vcpu *vcpu)
504 {
505 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
506 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
507 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
508 	unsigned long mem = 0;
509 	u64 operand2;
510 	int rc = 0;
511 	ar_t ar;
512 
513 	vcpu->stat.instruction_stsi++;
514 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
515 
516 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
517 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
518 
519 	if (fc > 3) {
520 		kvm_s390_set_psw_cc(vcpu, 3);
521 		return 0;
522 	}
523 
524 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
525 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
526 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
527 
528 	if (fc == 0) {
529 		vcpu->run->s.regs.gprs[0] = 3 << 28;
530 		kvm_s390_set_psw_cc(vcpu, 0);
531 		return 0;
532 	}
533 
534 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
535 
536 	if (operand2 & 0xfff)
537 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
538 
539 	switch (fc) {
540 	case 1: /* same handling for 1 and 2 */
541 	case 2:
542 		mem = get_zeroed_page(GFP_KERNEL);
543 		if (!mem)
544 			goto out_no_data;
545 		if (stsi((void *) mem, fc, sel1, sel2))
546 			goto out_no_data;
547 		break;
548 	case 3:
549 		if (sel1 != 2 || sel2 != 2)
550 			goto out_no_data;
551 		mem = get_zeroed_page(GFP_KERNEL);
552 		if (!mem)
553 			goto out_no_data;
554 		handle_stsi_3_2_2(vcpu, (void *) mem);
555 		break;
556 	}
557 
558 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
559 	if (rc) {
560 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
561 		goto out;
562 	}
563 	if (vcpu->kvm->arch.user_stsi) {
564 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
565 		rc = -EREMOTE;
566 	}
567 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
568 	free_page(mem);
569 	kvm_s390_set_psw_cc(vcpu, 0);
570 	vcpu->run->s.regs.gprs[0] = 0;
571 	return rc;
572 out_no_data:
573 	kvm_s390_set_psw_cc(vcpu, 3);
574 out:
575 	free_page(mem);
576 	return rc;
577 }
578 
579 static const intercept_handler_t b2_handlers[256] = {
580 	[0x02] = handle_stidp,
581 	[0x04] = handle_set_clock,
582 	[0x10] = handle_set_prefix,
583 	[0x11] = handle_store_prefix,
584 	[0x12] = handle_store_cpu_address,
585 	[0x21] = handle_ipte_interlock,
586 	[0x29] = handle_skey,
587 	[0x2a] = handle_skey,
588 	[0x2b] = handle_skey,
589 	[0x2c] = handle_test_block,
590 	[0x30] = handle_io_inst,
591 	[0x31] = handle_io_inst,
592 	[0x32] = handle_io_inst,
593 	[0x33] = handle_io_inst,
594 	[0x34] = handle_io_inst,
595 	[0x35] = handle_io_inst,
596 	[0x36] = handle_io_inst,
597 	[0x37] = handle_io_inst,
598 	[0x38] = handle_io_inst,
599 	[0x39] = handle_io_inst,
600 	[0x3a] = handle_io_inst,
601 	[0x3b] = handle_io_inst,
602 	[0x3c] = handle_io_inst,
603 	[0x50] = handle_ipte_interlock,
604 	[0x5f] = handle_io_inst,
605 	[0x74] = handle_io_inst,
606 	[0x76] = handle_io_inst,
607 	[0x7d] = handle_stsi,
608 	[0xb1] = handle_stfl,
609 	[0xb2] = handle_lpswe,
610 };
611 
612 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
613 {
614 	intercept_handler_t handler;
615 
616 	/*
617 	 * A lot of B2 instructions are priviledged. Here we check for
618 	 * the privileged ones, that we can handle in the kernel.
619 	 * Anything else goes to userspace.
620 	 */
621 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
622 	if (handler)
623 		return handler(vcpu);
624 
625 	return -EOPNOTSUPP;
626 }
627 
628 static int handle_epsw(struct kvm_vcpu *vcpu)
629 {
630 	int reg1, reg2;
631 
632 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
633 
634 	/* This basically extracts the mask half of the psw. */
635 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
636 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
637 	if (reg2) {
638 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
639 		vcpu->run->s.regs.gprs[reg2] |=
640 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
641 	}
642 	return 0;
643 }
644 
645 #define PFMF_RESERVED   0xfffc0101UL
646 #define PFMF_SK         0x00020000UL
647 #define PFMF_CF         0x00010000UL
648 #define PFMF_UI         0x00008000UL
649 #define PFMF_FSC        0x00007000UL
650 #define PFMF_NQ         0x00000800UL
651 #define PFMF_MR         0x00000400UL
652 #define PFMF_MC         0x00000200UL
653 #define PFMF_KEY        0x000000feUL
654 
655 static int handle_pfmf(struct kvm_vcpu *vcpu)
656 {
657 	int reg1, reg2;
658 	unsigned long start, end;
659 
660 	vcpu->stat.instruction_pfmf++;
661 
662 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
663 
664 	if (!test_kvm_facility(vcpu->kvm, 8))
665 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
666 
667 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
668 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
669 
670 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
671 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
672 
673 	/* Only provide non-quiescing support if enabled for the guest */
674 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
675 	    !test_kvm_facility(vcpu->kvm, 14))
676 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
677 
678 	/* No support for conditional-SSKE */
679 	if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
680 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
681 
682 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
683 	start = kvm_s390_logical_to_effective(vcpu, start);
684 
685 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
686 	case 0x00000000:
687 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
688 		break;
689 	case 0x00001000:
690 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
691 		break;
692 	case 0x00002000:
693 		/* only support 2G frame size if EDAT2 is available and we are
694 		   not in 24-bit addressing mode */
695 		if (!test_kvm_facility(vcpu->kvm, 78) ||
696 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_24BIT)
697 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
698 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
699 		break;
700 	default:
701 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
702 	}
703 
704 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
705 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
706 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
707 	}
708 
709 	while (start < end) {
710 		unsigned long useraddr, abs_addr;
711 
712 		/* Translate guest address to host address */
713 		if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
714 			abs_addr = kvm_s390_real_to_abs(vcpu, start);
715 		else
716 			abs_addr = start;
717 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
718 		if (kvm_is_error_hva(useraddr))
719 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
720 
721 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
722 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
723 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
724 		}
725 
726 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
727 			int rc = __skey_check_enable(vcpu);
728 
729 			if (rc)
730 				return rc;
731 			if (set_guest_storage_key(current->mm, useraddr,
732 					vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
733 					vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
734 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
735 		}
736 
737 		start += PAGE_SIZE;
738 	}
739 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
740 		vcpu->run->s.regs.gprs[reg2] = end;
741 	return 0;
742 }
743 
744 static int handle_essa(struct kvm_vcpu *vcpu)
745 {
746 	/* entries expected to be 1FF */
747 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
748 	unsigned long *cbrlo;
749 	struct gmap *gmap;
750 	int i;
751 
752 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
753 	gmap = vcpu->arch.gmap;
754 	vcpu->stat.instruction_essa++;
755 	if (!vcpu->kvm->arch.use_cmma)
756 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
757 
758 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
759 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
760 
761 	if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
762 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
763 
764 	/* Retry the ESSA instruction */
765 	kvm_s390_retry_instr(vcpu);
766 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
767 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
768 	down_read(&gmap->mm->mmap_sem);
769 	for (i = 0; i < entries; ++i)
770 		__gmap_zap(gmap, cbrlo[i]);
771 	up_read(&gmap->mm->mmap_sem);
772 	return 0;
773 }
774 
775 static const intercept_handler_t b9_handlers[256] = {
776 	[0x8a] = handle_ipte_interlock,
777 	[0x8d] = handle_epsw,
778 	[0x8e] = handle_ipte_interlock,
779 	[0x8f] = handle_ipte_interlock,
780 	[0xab] = handle_essa,
781 	[0xaf] = handle_pfmf,
782 };
783 
784 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
785 {
786 	intercept_handler_t handler;
787 
788 	/* This is handled just as for the B2 instructions. */
789 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
790 	if (handler)
791 		return handler(vcpu);
792 
793 	return -EOPNOTSUPP;
794 }
795 
796 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
797 {
798 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
799 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
800 	int reg, rc, nr_regs;
801 	u32 ctl_array[16];
802 	u64 ga;
803 	ar_t ar;
804 
805 	vcpu->stat.instruction_lctl++;
806 
807 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
808 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
809 
810 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
811 
812 	if (ga & 3)
813 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
814 
815 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
816 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
817 
818 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
819 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
820 	if (rc)
821 		return kvm_s390_inject_prog_cond(vcpu, rc);
822 	reg = reg1;
823 	nr_regs = 0;
824 	do {
825 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
826 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
827 		if (reg == reg3)
828 			break;
829 		reg = (reg + 1) % 16;
830 	} while (1);
831 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
832 	return 0;
833 }
834 
835 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
836 {
837 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
838 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
839 	int reg, rc, nr_regs;
840 	u32 ctl_array[16];
841 	u64 ga;
842 	ar_t ar;
843 
844 	vcpu->stat.instruction_stctl++;
845 
846 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
847 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
848 
849 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
850 
851 	if (ga & 3)
852 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
853 
854 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
855 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
856 
857 	reg = reg1;
858 	nr_regs = 0;
859 	do {
860 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
861 		if (reg == reg3)
862 			break;
863 		reg = (reg + 1) % 16;
864 	} while (1);
865 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
866 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
867 }
868 
869 static int handle_lctlg(struct kvm_vcpu *vcpu)
870 {
871 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
872 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
873 	int reg, rc, nr_regs;
874 	u64 ctl_array[16];
875 	u64 ga;
876 	ar_t ar;
877 
878 	vcpu->stat.instruction_lctlg++;
879 
880 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
881 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
882 
883 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
884 
885 	if (ga & 7)
886 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
887 
888 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
889 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
890 
891 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
892 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
893 	if (rc)
894 		return kvm_s390_inject_prog_cond(vcpu, rc);
895 	reg = reg1;
896 	nr_regs = 0;
897 	do {
898 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
899 		if (reg == reg3)
900 			break;
901 		reg = (reg + 1) % 16;
902 	} while (1);
903 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
904 	return 0;
905 }
906 
907 static int handle_stctg(struct kvm_vcpu *vcpu)
908 {
909 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
910 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
911 	int reg, rc, nr_regs;
912 	u64 ctl_array[16];
913 	u64 ga;
914 	ar_t ar;
915 
916 	vcpu->stat.instruction_stctg++;
917 
918 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
919 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
920 
921 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
922 
923 	if (ga & 7)
924 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
925 
926 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
927 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
928 
929 	reg = reg1;
930 	nr_regs = 0;
931 	do {
932 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
933 		if (reg == reg3)
934 			break;
935 		reg = (reg + 1) % 16;
936 	} while (1);
937 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
938 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
939 }
940 
941 static const intercept_handler_t eb_handlers[256] = {
942 	[0x2f] = handle_lctlg,
943 	[0x25] = handle_stctg,
944 };
945 
946 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
947 {
948 	intercept_handler_t handler;
949 
950 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
951 	if (handler)
952 		return handler(vcpu);
953 	return -EOPNOTSUPP;
954 }
955 
956 static int handle_tprot(struct kvm_vcpu *vcpu)
957 {
958 	u64 address1, address2;
959 	unsigned long hva, gpa;
960 	int ret = 0, cc = 0;
961 	bool writable;
962 	ar_t ar;
963 
964 	vcpu->stat.instruction_tprot++;
965 
966 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
967 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
968 
969 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
970 
971 	/* we only handle the Linux memory detection case:
972 	 * access key == 0
973 	 * everything else goes to userspace. */
974 	if (address2 & 0xf0)
975 		return -EOPNOTSUPP;
976 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
977 		ipte_lock(vcpu);
978 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
979 	if (ret == PGM_PROTECTION) {
980 		/* Write protected? Try again with read-only... */
981 		cc = 1;
982 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
983 					      GACC_FETCH);
984 	}
985 	if (ret) {
986 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
987 			ret = kvm_s390_inject_program_int(vcpu, ret);
988 		} else if (ret > 0) {
989 			/* Translation not available */
990 			kvm_s390_set_psw_cc(vcpu, 3);
991 			ret = 0;
992 		}
993 		goto out_unlock;
994 	}
995 
996 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
997 	if (kvm_is_error_hva(hva)) {
998 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
999 	} else {
1000 		if (!writable)
1001 			cc = 1;		/* Write not permitted ==> read-only */
1002 		kvm_s390_set_psw_cc(vcpu, cc);
1003 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1004 	}
1005 out_unlock:
1006 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1007 		ipte_unlock(vcpu);
1008 	return ret;
1009 }
1010 
1011 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1012 {
1013 	/* For e5xx... instructions we only handle TPROT */
1014 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1015 		return handle_tprot(vcpu);
1016 	return -EOPNOTSUPP;
1017 }
1018 
1019 static int handle_sckpf(struct kvm_vcpu *vcpu)
1020 {
1021 	u32 value;
1022 
1023 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1024 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1025 
1026 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1027 		return kvm_s390_inject_program_int(vcpu,
1028 						   PGM_SPECIFICATION);
1029 
1030 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1031 	vcpu->arch.sie_block->todpr = value;
1032 
1033 	return 0;
1034 }
1035 
1036 static const intercept_handler_t x01_handlers[256] = {
1037 	[0x07] = handle_sckpf,
1038 };
1039 
1040 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1041 {
1042 	intercept_handler_t handler;
1043 
1044 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1045 	if (handler)
1046 		return handler(vcpu);
1047 	return -EOPNOTSUPP;
1048 }
1049