xref: /openbmc/linux/arch/s390/kvm/priv.c (revision cd4d09ec)
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 	int rc;
37 	ar_t ar;
38 	u64 op2, val;
39 
40 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
41 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
42 
43 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
44 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
45 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
46 	rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
47 	if (rc)
48 		return kvm_s390_inject_prog_cond(vcpu, rc);
49 
50 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
51 	kvm_s390_set_tod_clock(vcpu->kvm, val);
52 
53 	kvm_s390_set_psw_cc(vcpu, 0);
54 	return 0;
55 }
56 
57 static int handle_set_prefix(struct kvm_vcpu *vcpu)
58 {
59 	u64 operand2;
60 	u32 address;
61 	int rc;
62 	ar_t ar;
63 
64 	vcpu->stat.instruction_spx++;
65 
66 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
67 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
68 
69 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
70 
71 	/* must be word boundary */
72 	if (operand2 & 3)
73 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
74 
75 	/* get the value */
76 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
77 	if (rc)
78 		return kvm_s390_inject_prog_cond(vcpu, rc);
79 
80 	address &= 0x7fffe000u;
81 
82 	/*
83 	 * Make sure the new value is valid memory. We only need to check the
84 	 * first page, since address is 8k aligned and memory pieces are always
85 	 * at least 1MB aligned and have at least a size of 1MB.
86 	 */
87 	if (kvm_is_error_gpa(vcpu->kvm, address))
88 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
89 
90 	kvm_s390_set_prefix(vcpu, address);
91 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
92 	return 0;
93 }
94 
95 static int handle_store_prefix(struct kvm_vcpu *vcpu)
96 {
97 	u64 operand2;
98 	u32 address;
99 	int rc;
100 	ar_t ar;
101 
102 	vcpu->stat.instruction_stpx++;
103 
104 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
105 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
106 
107 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
108 
109 	/* must be word boundary */
110 	if (operand2 & 3)
111 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
112 
113 	address = kvm_s390_get_prefix(vcpu);
114 
115 	/* get the value */
116 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
117 	if (rc)
118 		return kvm_s390_inject_prog_cond(vcpu, rc);
119 
120 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
121 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
122 	return 0;
123 }
124 
125 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
126 {
127 	u16 vcpu_id = vcpu->vcpu_id;
128 	u64 ga;
129 	int rc;
130 	ar_t ar;
131 
132 	vcpu->stat.instruction_stap++;
133 
134 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
135 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
136 
137 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
138 
139 	if (ga & 1)
140 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
141 
142 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
143 	if (rc)
144 		return kvm_s390_inject_prog_cond(vcpu, rc);
145 
146 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
147 	trace_kvm_s390_handle_stap(vcpu, ga);
148 	return 0;
149 }
150 
151 static int __skey_check_enable(struct kvm_vcpu *vcpu)
152 {
153 	int rc = 0;
154 	if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
155 		return rc;
156 
157 	rc = s390_enable_skey();
158 	VCPU_EVENT(vcpu, 3, "%s", "enabling storage keys for guest");
159 	trace_kvm_s390_skey_related_inst(vcpu);
160 	vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
161 	return rc;
162 }
163 
164 
165 static int handle_skey(struct kvm_vcpu *vcpu)
166 {
167 	int rc = __skey_check_enable(vcpu);
168 
169 	if (rc)
170 		return rc;
171 	vcpu->stat.instruction_storage_key++;
172 
173 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
174 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
175 
176 	kvm_s390_rewind_psw(vcpu, 4);
177 	VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
178 	return 0;
179 }
180 
181 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
182 {
183 	vcpu->stat.instruction_ipte_interlock++;
184 	if (psw_bits(vcpu->arch.sie_block->gpsw).p)
185 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
186 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
187 	kvm_s390_rewind_psw(vcpu, 4);
188 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
189 	return 0;
190 }
191 
192 static int handle_test_block(struct kvm_vcpu *vcpu)
193 {
194 	gpa_t addr;
195 	int reg2;
196 
197 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
198 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
199 
200 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
201 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
202 	addr = kvm_s390_logical_to_effective(vcpu, addr);
203 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
204 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
205 	addr = kvm_s390_real_to_abs(vcpu, addr);
206 
207 	if (kvm_is_error_gpa(vcpu->kvm, addr))
208 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
209 	/*
210 	 * We don't expect errors on modern systems, and do not care
211 	 * about storage keys (yet), so let's just clear the page.
212 	 */
213 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
214 		return -EFAULT;
215 	kvm_s390_set_psw_cc(vcpu, 0);
216 	vcpu->run->s.regs.gprs[0] = 0;
217 	return 0;
218 }
219 
220 static int handle_tpi(struct kvm_vcpu *vcpu)
221 {
222 	struct kvm_s390_interrupt_info *inti;
223 	unsigned long len;
224 	u32 tpi_data[3];
225 	int rc;
226 	u64 addr;
227 	ar_t ar;
228 
229 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
230 	if (addr & 3)
231 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
232 
233 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
234 	if (!inti) {
235 		kvm_s390_set_psw_cc(vcpu, 0);
236 		return 0;
237 	}
238 
239 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
240 	tpi_data[1] = inti->io.io_int_parm;
241 	tpi_data[2] = inti->io.io_int_word;
242 	if (addr) {
243 		/*
244 		 * Store the two-word I/O interruption code into the
245 		 * provided area.
246 		 */
247 		len = sizeof(tpi_data) - 4;
248 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
249 		if (rc) {
250 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
251 			goto reinject_interrupt;
252 		}
253 	} else {
254 		/*
255 		 * Store the three-word I/O interruption code into
256 		 * the appropriate lowcore area.
257 		 */
258 		len = sizeof(tpi_data);
259 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
260 			/* failed writes to the low core are not recoverable */
261 			rc = -EFAULT;
262 			goto reinject_interrupt;
263 		}
264 	}
265 
266 	/* irq was successfully handed to the guest */
267 	kfree(inti);
268 	kvm_s390_set_psw_cc(vcpu, 1);
269 	return 0;
270 reinject_interrupt:
271 	/*
272 	 * If we encounter a problem storing the interruption code, the
273 	 * instruction is suppressed from the guest's view: reinject the
274 	 * interrupt.
275 	 */
276 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
277 		kfree(inti);
278 		rc = -EFAULT;
279 	}
280 	/* don't set the cc, a pgm irq was injected or we drop to user space */
281 	return rc ? -EFAULT : 0;
282 }
283 
284 static int handle_tsch(struct kvm_vcpu *vcpu)
285 {
286 	struct kvm_s390_interrupt_info *inti = NULL;
287 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
288 
289 	/* a valid schid has at least one bit set */
290 	if (vcpu->run->s.regs.gprs[1])
291 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
292 					   vcpu->run->s.regs.gprs[1]);
293 
294 	/*
295 	 * Prepare exit to userspace.
296 	 * We indicate whether we dequeued a pending I/O interrupt
297 	 * so that userspace can re-inject it if the instruction gets
298 	 * a program check. While this may re-order the pending I/O
299 	 * interrupts, this is no problem since the priority is kept
300 	 * intact.
301 	 */
302 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
303 	vcpu->run->s390_tsch.dequeued = !!inti;
304 	if (inti) {
305 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
306 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
307 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
308 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
309 	}
310 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
311 	kfree(inti);
312 	return -EREMOTE;
313 }
314 
315 static int handle_io_inst(struct kvm_vcpu *vcpu)
316 {
317 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
318 
319 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
320 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
321 
322 	if (vcpu->kvm->arch.css_support) {
323 		/*
324 		 * Most I/O instructions will be handled by userspace.
325 		 * Exceptions are tpi and the interrupt portion of tsch.
326 		 */
327 		if (vcpu->arch.sie_block->ipa == 0xb236)
328 			return handle_tpi(vcpu);
329 		if (vcpu->arch.sie_block->ipa == 0xb235)
330 			return handle_tsch(vcpu);
331 		/* Handle in userspace. */
332 		return -EOPNOTSUPP;
333 	} else {
334 		/*
335 		 * Set condition code 3 to stop the guest from issuing channel
336 		 * I/O instructions.
337 		 */
338 		kvm_s390_set_psw_cc(vcpu, 3);
339 		return 0;
340 	}
341 }
342 
343 static int handle_stfl(struct kvm_vcpu *vcpu)
344 {
345 	int rc;
346 	unsigned int fac;
347 
348 	vcpu->stat.instruction_stfl++;
349 
350 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
351 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
352 
353 	/*
354 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
355 	 * into a u32 memory representation. They will remain bits 0-31.
356 	 */
357 	fac = *vcpu->kvm->arch.model.fac->list >> 32;
358 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
359 			    &fac, sizeof(fac));
360 	if (rc)
361 		return rc;
362 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
363 	trace_kvm_s390_handle_stfl(vcpu, fac);
364 	return 0;
365 }
366 
367 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
368 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
369 #define PSW_ADDR_24 0x0000000000ffffffUL
370 #define PSW_ADDR_31 0x000000007fffffffUL
371 
372 int is_valid_psw(psw_t *psw)
373 {
374 	if (psw->mask & PSW_MASK_UNASSIGNED)
375 		return 0;
376 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
377 		if (psw->addr & ~PSW_ADDR_31)
378 			return 0;
379 	}
380 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
381 		return 0;
382 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
383 		return 0;
384 	if (psw->addr & 1)
385 		return 0;
386 	return 1;
387 }
388 
389 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
390 {
391 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
392 	psw_compat_t new_psw;
393 	u64 addr;
394 	int rc;
395 	ar_t ar;
396 
397 	if (gpsw->mask & PSW_MASK_PSTATE)
398 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
399 
400 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
401 	if (addr & 7)
402 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
403 
404 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
405 	if (rc)
406 		return kvm_s390_inject_prog_cond(vcpu, rc);
407 	if (!(new_psw.mask & PSW32_MASK_BASE))
408 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
409 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
410 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
411 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
412 	if (!is_valid_psw(gpsw))
413 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
414 	return 0;
415 }
416 
417 static int handle_lpswe(struct kvm_vcpu *vcpu)
418 {
419 	psw_t new_psw;
420 	u64 addr;
421 	int rc;
422 	ar_t ar;
423 
424 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
425 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
426 
427 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
428 	if (addr & 7)
429 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
430 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
431 	if (rc)
432 		return kvm_s390_inject_prog_cond(vcpu, rc);
433 	vcpu->arch.sie_block->gpsw = new_psw;
434 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
435 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
436 	return 0;
437 }
438 
439 static int handle_stidp(struct kvm_vcpu *vcpu)
440 {
441 	u64 stidp_data = vcpu->arch.stidp_data;
442 	u64 operand2;
443 	int rc;
444 	ar_t ar;
445 
446 	vcpu->stat.instruction_stidp++;
447 
448 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
449 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
450 
451 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
452 
453 	if (operand2 & 7)
454 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
455 
456 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
457 	if (rc)
458 		return kvm_s390_inject_prog_cond(vcpu, rc);
459 
460 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
461 	return 0;
462 }
463 
464 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
465 {
466 	int cpus = 0;
467 	int n;
468 
469 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
470 
471 	/* deal with other level 3 hypervisors */
472 	if (stsi(mem, 3, 2, 2))
473 		mem->count = 0;
474 	if (mem->count < 8)
475 		mem->count++;
476 	for (n = mem->count - 1; n > 0 ; n--)
477 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
478 
479 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
480 	mem->vm[0].cpus_total = cpus;
481 	mem->vm[0].cpus_configured = cpus;
482 	mem->vm[0].cpus_standby = 0;
483 	mem->vm[0].cpus_reserved = 0;
484 	mem->vm[0].caf = 1000;
485 	memcpy(mem->vm[0].name, "KVMguest", 8);
486 	ASCEBC(mem->vm[0].name, 8);
487 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
488 	ASCEBC(mem->vm[0].cpi, 16);
489 }
490 
491 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, ar_t ar,
492 				 u8 fc, u8 sel1, u16 sel2)
493 {
494 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
495 	vcpu->run->s390_stsi.addr = addr;
496 	vcpu->run->s390_stsi.ar = ar;
497 	vcpu->run->s390_stsi.fc = fc;
498 	vcpu->run->s390_stsi.sel1 = sel1;
499 	vcpu->run->s390_stsi.sel2 = sel2;
500 }
501 
502 static int handle_stsi(struct kvm_vcpu *vcpu)
503 {
504 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
505 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
506 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
507 	unsigned long mem = 0;
508 	u64 operand2;
509 	int rc = 0;
510 	ar_t ar;
511 
512 	vcpu->stat.instruction_stsi++;
513 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
514 
515 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
516 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
517 
518 	if (fc > 3) {
519 		kvm_s390_set_psw_cc(vcpu, 3);
520 		return 0;
521 	}
522 
523 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
524 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
525 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
526 
527 	if (fc == 0) {
528 		vcpu->run->s.regs.gprs[0] = 3 << 28;
529 		kvm_s390_set_psw_cc(vcpu, 0);
530 		return 0;
531 	}
532 
533 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
534 
535 	if (operand2 & 0xfff)
536 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
537 
538 	switch (fc) {
539 	case 1: /* same handling for 1 and 2 */
540 	case 2:
541 		mem = get_zeroed_page(GFP_KERNEL);
542 		if (!mem)
543 			goto out_no_data;
544 		if (stsi((void *) mem, fc, sel1, sel2))
545 			goto out_no_data;
546 		break;
547 	case 3:
548 		if (sel1 != 2 || sel2 != 2)
549 			goto out_no_data;
550 		mem = get_zeroed_page(GFP_KERNEL);
551 		if (!mem)
552 			goto out_no_data;
553 		handle_stsi_3_2_2(vcpu, (void *) mem);
554 		break;
555 	}
556 
557 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
558 	if (rc) {
559 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
560 		goto out;
561 	}
562 	if (vcpu->kvm->arch.user_stsi) {
563 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
564 		rc = -EREMOTE;
565 	}
566 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
567 	free_page(mem);
568 	kvm_s390_set_psw_cc(vcpu, 0);
569 	vcpu->run->s.regs.gprs[0] = 0;
570 	return rc;
571 out_no_data:
572 	kvm_s390_set_psw_cc(vcpu, 3);
573 out:
574 	free_page(mem);
575 	return rc;
576 }
577 
578 static const intercept_handler_t b2_handlers[256] = {
579 	[0x02] = handle_stidp,
580 	[0x04] = handle_set_clock,
581 	[0x10] = handle_set_prefix,
582 	[0x11] = handle_store_prefix,
583 	[0x12] = handle_store_cpu_address,
584 	[0x21] = handle_ipte_interlock,
585 	[0x29] = handle_skey,
586 	[0x2a] = handle_skey,
587 	[0x2b] = handle_skey,
588 	[0x2c] = handle_test_block,
589 	[0x30] = handle_io_inst,
590 	[0x31] = handle_io_inst,
591 	[0x32] = handle_io_inst,
592 	[0x33] = handle_io_inst,
593 	[0x34] = handle_io_inst,
594 	[0x35] = handle_io_inst,
595 	[0x36] = handle_io_inst,
596 	[0x37] = handle_io_inst,
597 	[0x38] = handle_io_inst,
598 	[0x39] = handle_io_inst,
599 	[0x3a] = handle_io_inst,
600 	[0x3b] = handle_io_inst,
601 	[0x3c] = handle_io_inst,
602 	[0x50] = handle_ipte_interlock,
603 	[0x5f] = handle_io_inst,
604 	[0x74] = handle_io_inst,
605 	[0x76] = handle_io_inst,
606 	[0x7d] = handle_stsi,
607 	[0xb1] = handle_stfl,
608 	[0xb2] = handle_lpswe,
609 };
610 
611 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
612 {
613 	intercept_handler_t handler;
614 
615 	/*
616 	 * A lot of B2 instructions are priviledged. Here we check for
617 	 * the privileged ones, that we can handle in the kernel.
618 	 * Anything else goes to userspace.
619 	 */
620 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
621 	if (handler)
622 		return handler(vcpu);
623 
624 	return -EOPNOTSUPP;
625 }
626 
627 static int handle_epsw(struct kvm_vcpu *vcpu)
628 {
629 	int reg1, reg2;
630 
631 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
632 
633 	/* This basically extracts the mask half of the psw. */
634 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
635 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
636 	if (reg2) {
637 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
638 		vcpu->run->s.regs.gprs[reg2] |=
639 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
640 	}
641 	return 0;
642 }
643 
644 #define PFMF_RESERVED   0xfffc0101UL
645 #define PFMF_SK         0x00020000UL
646 #define PFMF_CF         0x00010000UL
647 #define PFMF_UI         0x00008000UL
648 #define PFMF_FSC        0x00007000UL
649 #define PFMF_NQ         0x00000800UL
650 #define PFMF_MR         0x00000400UL
651 #define PFMF_MC         0x00000200UL
652 #define PFMF_KEY        0x000000feUL
653 
654 static int handle_pfmf(struct kvm_vcpu *vcpu)
655 {
656 	int reg1, reg2;
657 	unsigned long start, end;
658 
659 	vcpu->stat.instruction_pfmf++;
660 
661 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
662 
663 	if (!test_kvm_facility(vcpu->kvm, 8))
664 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
665 
666 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
667 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
668 
669 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
670 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
671 
672 	/* Only provide non-quiescing support if the host supports it */
673 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
674 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
675 
676 	/* No support for conditional-SSKE */
677 	if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
678 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
679 
680 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
681 	start = kvm_s390_logical_to_effective(vcpu, start);
682 
683 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
684 	case 0x00000000:
685 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
686 		break;
687 	case 0x00001000:
688 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
689 		break;
690 	case 0x00002000:
691 		/* only support 2G frame size if EDAT2 is available and we are
692 		   not in 24-bit addressing mode */
693 		if (!test_kvm_facility(vcpu->kvm, 78) ||
694 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_24BIT)
695 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
696 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
697 		break;
698 	default:
699 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
700 	}
701 
702 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
703 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
704 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
705 	}
706 
707 	while (start < end) {
708 		unsigned long useraddr, abs_addr;
709 
710 		/* Translate guest address to host address */
711 		if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
712 			abs_addr = kvm_s390_real_to_abs(vcpu, start);
713 		else
714 			abs_addr = start;
715 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
716 		if (kvm_is_error_hva(useraddr))
717 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
718 
719 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
720 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
721 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
722 		}
723 
724 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
725 			int rc = __skey_check_enable(vcpu);
726 
727 			if (rc)
728 				return rc;
729 			if (set_guest_storage_key(current->mm, useraddr,
730 					vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
731 					vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
732 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
733 		}
734 
735 		start += PAGE_SIZE;
736 	}
737 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
738 		vcpu->run->s.regs.gprs[reg2] = end;
739 	return 0;
740 }
741 
742 static int handle_essa(struct kvm_vcpu *vcpu)
743 {
744 	/* entries expected to be 1FF */
745 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
746 	unsigned long *cbrlo, cbrle;
747 	struct gmap *gmap;
748 	int i;
749 
750 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
751 	gmap = vcpu->arch.gmap;
752 	vcpu->stat.instruction_essa++;
753 	if (!vcpu->kvm->arch.use_cmma)
754 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
755 
756 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
757 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
758 
759 	if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
760 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
761 
762 	/* Rewind PSW to repeat the ESSA instruction */
763 	kvm_s390_rewind_psw(vcpu, 4);
764 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
765 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
766 	down_read(&gmap->mm->mmap_sem);
767 	for (i = 0; i < entries; ++i) {
768 		cbrle = cbrlo[i];
769 		if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
770 			/* invalid entry */
771 			break;
772 		/* try to free backing */
773 		__gmap_zap(gmap, cbrle);
774 	}
775 	up_read(&gmap->mm->mmap_sem);
776 	if (i < entries)
777 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
778 	return 0;
779 }
780 
781 static const intercept_handler_t b9_handlers[256] = {
782 	[0x8a] = handle_ipte_interlock,
783 	[0x8d] = handle_epsw,
784 	[0x8e] = handle_ipte_interlock,
785 	[0x8f] = handle_ipte_interlock,
786 	[0xab] = handle_essa,
787 	[0xaf] = handle_pfmf,
788 };
789 
790 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
791 {
792 	intercept_handler_t handler;
793 
794 	/* This is handled just as for the B2 instructions. */
795 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
796 	if (handler)
797 		return handler(vcpu);
798 
799 	return -EOPNOTSUPP;
800 }
801 
802 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
803 {
804 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
805 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
806 	int reg, rc, nr_regs;
807 	u32 ctl_array[16];
808 	u64 ga;
809 	ar_t ar;
810 
811 	vcpu->stat.instruction_lctl++;
812 
813 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
814 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
815 
816 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
817 
818 	if (ga & 3)
819 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
820 
821 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
822 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
823 
824 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
825 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
826 	if (rc)
827 		return kvm_s390_inject_prog_cond(vcpu, rc);
828 	reg = reg1;
829 	nr_regs = 0;
830 	do {
831 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
832 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
833 		if (reg == reg3)
834 			break;
835 		reg = (reg + 1) % 16;
836 	} while (1);
837 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
838 	return 0;
839 }
840 
841 int kvm_s390_handle_stctl(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 	int reg, rc, nr_regs;
846 	u32 ctl_array[16];
847 	u64 ga;
848 	ar_t ar;
849 
850 	vcpu->stat.instruction_stctl++;
851 
852 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
853 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
854 
855 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
856 
857 	if (ga & 3)
858 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
859 
860 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
861 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
862 
863 	reg = reg1;
864 	nr_regs = 0;
865 	do {
866 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
867 		if (reg == reg3)
868 			break;
869 		reg = (reg + 1) % 16;
870 	} while (1);
871 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
872 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
873 }
874 
875 static int handle_lctlg(struct kvm_vcpu *vcpu)
876 {
877 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
878 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
879 	int reg, rc, nr_regs;
880 	u64 ctl_array[16];
881 	u64 ga;
882 	ar_t ar;
883 
884 	vcpu->stat.instruction_lctlg++;
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, &ar);
890 
891 	if (ga & 7)
892 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
893 
894 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
895 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
896 
897 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
898 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
899 	if (rc)
900 		return kvm_s390_inject_prog_cond(vcpu, rc);
901 	reg = reg1;
902 	nr_regs = 0;
903 	do {
904 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
905 		if (reg == reg3)
906 			break;
907 		reg = (reg + 1) % 16;
908 	} while (1);
909 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
910 	return 0;
911 }
912 
913 static int handle_stctg(struct kvm_vcpu *vcpu)
914 {
915 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
916 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
917 	int reg, rc, nr_regs;
918 	u64 ctl_array[16];
919 	u64 ga;
920 	ar_t ar;
921 
922 	vcpu->stat.instruction_stctg++;
923 
924 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
925 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
926 
927 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
928 
929 	if (ga & 7)
930 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
931 
932 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
933 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
934 
935 	reg = reg1;
936 	nr_regs = 0;
937 	do {
938 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
939 		if (reg == reg3)
940 			break;
941 		reg = (reg + 1) % 16;
942 	} while (1);
943 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
944 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
945 }
946 
947 static const intercept_handler_t eb_handlers[256] = {
948 	[0x2f] = handle_lctlg,
949 	[0x25] = handle_stctg,
950 };
951 
952 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
953 {
954 	intercept_handler_t handler;
955 
956 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
957 	if (handler)
958 		return handler(vcpu);
959 	return -EOPNOTSUPP;
960 }
961 
962 static int handle_tprot(struct kvm_vcpu *vcpu)
963 {
964 	u64 address1, address2;
965 	unsigned long hva, gpa;
966 	int ret = 0, cc = 0;
967 	bool writable;
968 	ar_t ar;
969 
970 	vcpu->stat.instruction_tprot++;
971 
972 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
973 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
974 
975 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
976 
977 	/* we only handle the Linux memory detection case:
978 	 * access key == 0
979 	 * everything else goes to userspace. */
980 	if (address2 & 0xf0)
981 		return -EOPNOTSUPP;
982 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
983 		ipte_lock(vcpu);
984 	ret = guest_translate_address(vcpu, address1, ar, &gpa, 1);
985 	if (ret == PGM_PROTECTION) {
986 		/* Write protected? Try again with read-only... */
987 		cc = 1;
988 		ret = guest_translate_address(vcpu, address1, ar, &gpa, 0);
989 	}
990 	if (ret) {
991 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
992 			ret = kvm_s390_inject_program_int(vcpu, ret);
993 		} else if (ret > 0) {
994 			/* Translation not available */
995 			kvm_s390_set_psw_cc(vcpu, 3);
996 			ret = 0;
997 		}
998 		goto out_unlock;
999 	}
1000 
1001 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1002 	if (kvm_is_error_hva(hva)) {
1003 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1004 	} else {
1005 		if (!writable)
1006 			cc = 1;		/* Write not permitted ==> read-only */
1007 		kvm_s390_set_psw_cc(vcpu, cc);
1008 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1009 	}
1010 out_unlock:
1011 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1012 		ipte_unlock(vcpu);
1013 	return ret;
1014 }
1015 
1016 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1017 {
1018 	/* For e5xx... instructions we only handle TPROT */
1019 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1020 		return handle_tprot(vcpu);
1021 	return -EOPNOTSUPP;
1022 }
1023 
1024 static int handle_sckpf(struct kvm_vcpu *vcpu)
1025 {
1026 	u32 value;
1027 
1028 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1029 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1030 
1031 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1032 		return kvm_s390_inject_program_int(vcpu,
1033 						   PGM_SPECIFICATION);
1034 
1035 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1036 	vcpu->arch.sie_block->todpr = value;
1037 
1038 	return 0;
1039 }
1040 
1041 static const intercept_handler_t x01_handlers[256] = {
1042 	[0x07] = handle_sckpf,
1043 };
1044 
1045 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1046 {
1047 	intercept_handler_t handler;
1048 
1049 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1050 	if (handler)
1051 		return handler(vcpu);
1052 	return -EOPNOTSUPP;
1053 }
1054