xref: /linux/arch/s390/kvm/priv.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
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 	ar_t ar;
40 	u64 op2;
41 
42 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
43 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
44 
45 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
46 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
47 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
48 	rc = read_guest(vcpu, op2, ar, &val, sizeof(val));
49 	if (rc)
50 		return kvm_s390_inject_prog_cond(vcpu, rc);
51 
52 	if (store_tod_clock(&hostclk)) {
53 		kvm_s390_set_psw_cc(vcpu, 3);
54 		return 0;
55 	}
56 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", val);
57 	val = (val - hostclk) & ~0x3fUL;
58 
59 	mutex_lock(&vcpu->kvm->lock);
60 	preempt_disable();
61 	kvm_for_each_vcpu(i, cpup, vcpu->kvm)
62 		cpup->arch.sie_block->epoch = val;
63 	preempt_enable();
64 	mutex_unlock(&vcpu->kvm->lock);
65 
66 	kvm_s390_set_psw_cc(vcpu, 0);
67 	return 0;
68 }
69 
70 static int handle_set_prefix(struct kvm_vcpu *vcpu)
71 {
72 	u64 operand2;
73 	u32 address;
74 	int rc;
75 	ar_t ar;
76 
77 	vcpu->stat.instruction_spx++;
78 
79 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
80 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
81 
82 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
83 
84 	/* must be word boundary */
85 	if (operand2 & 3)
86 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
87 
88 	/* get the value */
89 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
90 	if (rc)
91 		return kvm_s390_inject_prog_cond(vcpu, rc);
92 
93 	address &= 0x7fffe000u;
94 
95 	/*
96 	 * Make sure the new value is valid memory. We only need to check the
97 	 * first page, since address is 8k aligned and memory pieces are always
98 	 * at least 1MB aligned and have at least a size of 1MB.
99 	 */
100 	if (kvm_is_error_gpa(vcpu->kvm, address))
101 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
102 
103 	kvm_s390_set_prefix(vcpu, address);
104 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
105 	return 0;
106 }
107 
108 static int handle_store_prefix(struct kvm_vcpu *vcpu)
109 {
110 	u64 operand2;
111 	u32 address;
112 	int rc;
113 	ar_t ar;
114 
115 	vcpu->stat.instruction_stpx++;
116 
117 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
118 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
119 
120 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
121 
122 	/* must be word boundary */
123 	if (operand2 & 3)
124 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
125 
126 	address = kvm_s390_get_prefix(vcpu);
127 
128 	/* get the value */
129 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
130 	if (rc)
131 		return kvm_s390_inject_prog_cond(vcpu, rc);
132 
133 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
134 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
135 	return 0;
136 }
137 
138 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
139 {
140 	u16 vcpu_id = vcpu->vcpu_id;
141 	u64 ga;
142 	int rc;
143 	ar_t ar;
144 
145 	vcpu->stat.instruction_stap++;
146 
147 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
148 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
149 
150 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
151 
152 	if (ga & 1)
153 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
154 
155 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
156 	if (rc)
157 		return kvm_s390_inject_prog_cond(vcpu, rc);
158 
159 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
160 	trace_kvm_s390_handle_stap(vcpu, ga);
161 	return 0;
162 }
163 
164 static int __skey_check_enable(struct kvm_vcpu *vcpu)
165 {
166 	int rc = 0;
167 	if (!(vcpu->arch.sie_block->ictl & (ICTL_ISKE | ICTL_SSKE | ICTL_RRBE)))
168 		return rc;
169 
170 	rc = s390_enable_skey();
171 	VCPU_EVENT(vcpu, 3, "%s", "enabling storage keys for guest");
172 	trace_kvm_s390_skey_related_inst(vcpu);
173 	vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
174 	return rc;
175 }
176 
177 
178 static int handle_skey(struct kvm_vcpu *vcpu)
179 {
180 	int rc = __skey_check_enable(vcpu);
181 
182 	if (rc)
183 		return rc;
184 	vcpu->stat.instruction_storage_key++;
185 
186 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
187 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
188 
189 	kvm_s390_rewind_psw(vcpu, 4);
190 	VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
191 	return 0;
192 }
193 
194 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
195 {
196 	vcpu->stat.instruction_ipte_interlock++;
197 	if (psw_bits(vcpu->arch.sie_block->gpsw).p)
198 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
199 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
200 	kvm_s390_rewind_psw(vcpu, 4);
201 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
202 	return 0;
203 }
204 
205 static int handle_test_block(struct kvm_vcpu *vcpu)
206 {
207 	gpa_t addr;
208 	int reg2;
209 
210 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
211 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
212 
213 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
214 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
215 	addr = kvm_s390_logical_to_effective(vcpu, addr);
216 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
217 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
218 	addr = kvm_s390_real_to_abs(vcpu, addr);
219 
220 	if (kvm_is_error_gpa(vcpu->kvm, addr))
221 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
222 	/*
223 	 * We don't expect errors on modern systems, and do not care
224 	 * about storage keys (yet), so let's just clear the page.
225 	 */
226 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
227 		return -EFAULT;
228 	kvm_s390_set_psw_cc(vcpu, 0);
229 	vcpu->run->s.regs.gprs[0] = 0;
230 	return 0;
231 }
232 
233 static int handle_tpi(struct kvm_vcpu *vcpu)
234 {
235 	struct kvm_s390_interrupt_info *inti;
236 	unsigned long len;
237 	u32 tpi_data[3];
238 	int rc;
239 	u64 addr;
240 	ar_t ar;
241 
242 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
243 	if (addr & 3)
244 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
245 
246 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
247 	if (!inti) {
248 		kvm_s390_set_psw_cc(vcpu, 0);
249 		return 0;
250 	}
251 
252 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
253 	tpi_data[1] = inti->io.io_int_parm;
254 	tpi_data[2] = inti->io.io_int_word;
255 	if (addr) {
256 		/*
257 		 * Store the two-word I/O interruption code into the
258 		 * provided area.
259 		 */
260 		len = sizeof(tpi_data) - 4;
261 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
262 		if (rc) {
263 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
264 			goto reinject_interrupt;
265 		}
266 	} else {
267 		/*
268 		 * Store the three-word I/O interruption code into
269 		 * the appropriate lowcore area.
270 		 */
271 		len = sizeof(tpi_data);
272 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
273 			/* failed writes to the low core are not recoverable */
274 			rc = -EFAULT;
275 			goto reinject_interrupt;
276 		}
277 	}
278 
279 	/* irq was successfully handed to the guest */
280 	kfree(inti);
281 	kvm_s390_set_psw_cc(vcpu, 1);
282 	return 0;
283 reinject_interrupt:
284 	/*
285 	 * If we encounter a problem storing the interruption code, the
286 	 * instruction is suppressed from the guest's view: reinject the
287 	 * interrupt.
288 	 */
289 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
290 		kfree(inti);
291 		rc = -EFAULT;
292 	}
293 	/* don't set the cc, a pgm irq was injected or we drop to user space */
294 	return rc ? -EFAULT : 0;
295 }
296 
297 static int handle_tsch(struct kvm_vcpu *vcpu)
298 {
299 	struct kvm_s390_interrupt_info *inti = NULL;
300 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
301 
302 	/* a valid schid has at least one bit set */
303 	if (vcpu->run->s.regs.gprs[1])
304 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
305 					   vcpu->run->s.regs.gprs[1]);
306 
307 	/*
308 	 * Prepare exit to userspace.
309 	 * We indicate whether we dequeued a pending I/O interrupt
310 	 * so that userspace can re-inject it if the instruction gets
311 	 * a program check. While this may re-order the pending I/O
312 	 * interrupts, this is no problem since the priority is kept
313 	 * intact.
314 	 */
315 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
316 	vcpu->run->s390_tsch.dequeued = !!inti;
317 	if (inti) {
318 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
319 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
320 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
321 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
322 	}
323 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
324 	kfree(inti);
325 	return -EREMOTE;
326 }
327 
328 static int handle_io_inst(struct kvm_vcpu *vcpu)
329 {
330 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
331 
332 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
333 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
334 
335 	if (vcpu->kvm->arch.css_support) {
336 		/*
337 		 * Most I/O instructions will be handled by userspace.
338 		 * Exceptions are tpi and the interrupt portion of tsch.
339 		 */
340 		if (vcpu->arch.sie_block->ipa == 0xb236)
341 			return handle_tpi(vcpu);
342 		if (vcpu->arch.sie_block->ipa == 0xb235)
343 			return handle_tsch(vcpu);
344 		/* Handle in userspace. */
345 		return -EOPNOTSUPP;
346 	} else {
347 		/*
348 		 * Set condition code 3 to stop the guest from issuing channel
349 		 * I/O instructions.
350 		 */
351 		kvm_s390_set_psw_cc(vcpu, 3);
352 		return 0;
353 	}
354 }
355 
356 static int handle_stfl(struct kvm_vcpu *vcpu)
357 {
358 	int rc;
359 	unsigned int fac;
360 
361 	vcpu->stat.instruction_stfl++;
362 
363 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
364 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
365 
366 	/*
367 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
368 	 * into a u32 memory representation. They will remain bits 0-31.
369 	 */
370 	fac = *vcpu->kvm->arch.model.fac->list >> 32;
371 	rc = write_guest_lc(vcpu, offsetof(struct _lowcore, stfl_fac_list),
372 			    &fac, sizeof(fac));
373 	if (rc)
374 		return rc;
375 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
376 	trace_kvm_s390_handle_stfl(vcpu, fac);
377 	return 0;
378 }
379 
380 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
381 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
382 #define PSW_ADDR_24 0x0000000000ffffffUL
383 #define PSW_ADDR_31 0x000000007fffffffUL
384 
385 int is_valid_psw(psw_t *psw)
386 {
387 	if (psw->mask & PSW_MASK_UNASSIGNED)
388 		return 0;
389 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
390 		if (psw->addr & ~PSW_ADDR_31)
391 			return 0;
392 	}
393 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
394 		return 0;
395 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
396 		return 0;
397 	if (psw->addr & 1)
398 		return 0;
399 	return 1;
400 }
401 
402 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
403 {
404 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
405 	psw_compat_t new_psw;
406 	u64 addr;
407 	int rc;
408 	ar_t ar;
409 
410 	if (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, &ar);
414 	if (addr & 7)
415 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
416 
417 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
418 	if (rc)
419 		return kvm_s390_inject_prog_cond(vcpu, rc);
420 	if (!(new_psw.mask & PSW32_MASK_BASE))
421 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
422 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
423 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
424 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
425 	if (!is_valid_psw(gpsw))
426 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
427 	return 0;
428 }
429 
430 static int handle_lpswe(struct kvm_vcpu *vcpu)
431 {
432 	psw_t new_psw;
433 	u64 addr;
434 	int rc;
435 	ar_t ar;
436 
437 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
438 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
439 
440 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
441 	if (addr & 7)
442 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
443 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
444 	if (rc)
445 		return kvm_s390_inject_prog_cond(vcpu, rc);
446 	vcpu->arch.sie_block->gpsw = new_psw;
447 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
448 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
449 	return 0;
450 }
451 
452 static int handle_stidp(struct kvm_vcpu *vcpu)
453 {
454 	u64 stidp_data = vcpu->arch.stidp_data;
455 	u64 operand2;
456 	int rc;
457 	ar_t ar;
458 
459 	vcpu->stat.instruction_stidp++;
460 
461 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
462 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
463 
464 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
465 
466 	if (operand2 & 7)
467 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
468 
469 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
470 	if (rc)
471 		return kvm_s390_inject_prog_cond(vcpu, rc);
472 
473 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
474 	return 0;
475 }
476 
477 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
478 {
479 	int cpus = 0;
480 	int n;
481 
482 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
483 
484 	/* deal with other level 3 hypervisors */
485 	if (stsi(mem, 3, 2, 2))
486 		mem->count = 0;
487 	if (mem->count < 8)
488 		mem->count++;
489 	for (n = mem->count - 1; n > 0 ; n--)
490 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
491 
492 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
493 	mem->vm[0].cpus_total = cpus;
494 	mem->vm[0].cpus_configured = cpus;
495 	mem->vm[0].cpus_standby = 0;
496 	mem->vm[0].cpus_reserved = 0;
497 	mem->vm[0].caf = 1000;
498 	memcpy(mem->vm[0].name, "KVMguest", 8);
499 	ASCEBC(mem->vm[0].name, 8);
500 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
501 	ASCEBC(mem->vm[0].cpi, 16);
502 }
503 
504 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, ar_t ar,
505 				 u8 fc, u8 sel1, u16 sel2)
506 {
507 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
508 	vcpu->run->s390_stsi.addr = addr;
509 	vcpu->run->s390_stsi.ar = ar;
510 	vcpu->run->s390_stsi.fc = fc;
511 	vcpu->run->s390_stsi.sel1 = sel1;
512 	vcpu->run->s390_stsi.sel2 = sel2;
513 }
514 
515 static int handle_stsi(struct kvm_vcpu *vcpu)
516 {
517 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
518 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
519 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
520 	unsigned long mem = 0;
521 	u64 operand2;
522 	int rc = 0;
523 	ar_t ar;
524 
525 	vcpu->stat.instruction_stsi++;
526 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
527 
528 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
529 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
530 
531 	if (fc > 3) {
532 		kvm_s390_set_psw_cc(vcpu, 3);
533 		return 0;
534 	}
535 
536 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
537 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
538 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
539 
540 	if (fc == 0) {
541 		vcpu->run->s.regs.gprs[0] = 3 << 28;
542 		kvm_s390_set_psw_cc(vcpu, 0);
543 		return 0;
544 	}
545 
546 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
547 
548 	if (operand2 & 0xfff)
549 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
550 
551 	switch (fc) {
552 	case 1: /* same handling for 1 and 2 */
553 	case 2:
554 		mem = get_zeroed_page(GFP_KERNEL);
555 		if (!mem)
556 			goto out_no_data;
557 		if (stsi((void *) mem, fc, sel1, sel2))
558 			goto out_no_data;
559 		break;
560 	case 3:
561 		if (sel1 != 2 || sel2 != 2)
562 			goto out_no_data;
563 		mem = get_zeroed_page(GFP_KERNEL);
564 		if (!mem)
565 			goto out_no_data;
566 		handle_stsi_3_2_2(vcpu, (void *) mem);
567 		break;
568 	}
569 
570 	rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
571 	if (rc) {
572 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
573 		goto out;
574 	}
575 	if (vcpu->kvm->arch.user_stsi) {
576 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
577 		rc = -EREMOTE;
578 	}
579 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
580 	free_page(mem);
581 	kvm_s390_set_psw_cc(vcpu, 0);
582 	vcpu->run->s.regs.gprs[0] = 0;
583 	return rc;
584 out_no_data:
585 	kvm_s390_set_psw_cc(vcpu, 3);
586 out:
587 	free_page(mem);
588 	return rc;
589 }
590 
591 static const intercept_handler_t b2_handlers[256] = {
592 	[0x02] = handle_stidp,
593 	[0x04] = handle_set_clock,
594 	[0x10] = handle_set_prefix,
595 	[0x11] = handle_store_prefix,
596 	[0x12] = handle_store_cpu_address,
597 	[0x21] = handle_ipte_interlock,
598 	[0x29] = handle_skey,
599 	[0x2a] = handle_skey,
600 	[0x2b] = handle_skey,
601 	[0x2c] = handle_test_block,
602 	[0x30] = handle_io_inst,
603 	[0x31] = handle_io_inst,
604 	[0x32] = handle_io_inst,
605 	[0x33] = handle_io_inst,
606 	[0x34] = handle_io_inst,
607 	[0x35] = handle_io_inst,
608 	[0x36] = handle_io_inst,
609 	[0x37] = handle_io_inst,
610 	[0x38] = handle_io_inst,
611 	[0x39] = handle_io_inst,
612 	[0x3a] = handle_io_inst,
613 	[0x3b] = handle_io_inst,
614 	[0x3c] = handle_io_inst,
615 	[0x50] = handle_ipte_interlock,
616 	[0x5f] = handle_io_inst,
617 	[0x74] = handle_io_inst,
618 	[0x76] = handle_io_inst,
619 	[0x7d] = handle_stsi,
620 	[0xb1] = handle_stfl,
621 	[0xb2] = handle_lpswe,
622 };
623 
624 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
625 {
626 	intercept_handler_t handler;
627 
628 	/*
629 	 * A lot of B2 instructions are priviledged. Here we check for
630 	 * the privileged ones, that we can handle in the kernel.
631 	 * Anything else goes to userspace.
632 	 */
633 	handler = b2_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
634 	if (handler)
635 		return handler(vcpu);
636 
637 	return -EOPNOTSUPP;
638 }
639 
640 static int handle_epsw(struct kvm_vcpu *vcpu)
641 {
642 	int reg1, reg2;
643 
644 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
645 
646 	/* This basically extracts the mask half of the psw. */
647 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
648 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
649 	if (reg2) {
650 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
651 		vcpu->run->s.regs.gprs[reg2] |=
652 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
653 	}
654 	return 0;
655 }
656 
657 #define PFMF_RESERVED   0xfffc0101UL
658 #define PFMF_SK         0x00020000UL
659 #define PFMF_CF         0x00010000UL
660 #define PFMF_UI         0x00008000UL
661 #define PFMF_FSC        0x00007000UL
662 #define PFMF_NQ         0x00000800UL
663 #define PFMF_MR         0x00000400UL
664 #define PFMF_MC         0x00000200UL
665 #define PFMF_KEY        0x000000feUL
666 
667 static int handle_pfmf(struct kvm_vcpu *vcpu)
668 {
669 	int reg1, reg2;
670 	unsigned long start, end;
671 
672 	vcpu->stat.instruction_pfmf++;
673 
674 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
675 
676 	if (!MACHINE_HAS_PFMF)
677 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
678 
679 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
680 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
681 
682 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
683 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
684 
685 	/* Only provide non-quiescing support if the host supports it */
686 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ && !test_facility(14))
687 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
688 
689 	/* No support for conditional-SSKE */
690 	if (vcpu->run->s.regs.gprs[reg1] & (PFMF_MR | PFMF_MC))
691 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
692 
693 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
694 	start = kvm_s390_logical_to_effective(vcpu, start);
695 
696 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
697 	case 0x00000000:
698 		end = (start + (1UL << 12)) & ~((1UL << 12) - 1);
699 		break;
700 	case 0x00001000:
701 		end = (start + (1UL << 20)) & ~((1UL << 20) - 1);
702 		break;
703 	case 0x00002000:
704 		/* only support 2G frame size if EDAT2 is available and we are
705 		   not in 24-bit addressing mode */
706 		if (!test_kvm_facility(vcpu->kvm, 78) ||
707 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_AMODE_24BIT)
708 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
709 		end = (start + (1UL << 31)) & ~((1UL << 31) - 1);
710 		break;
711 	default:
712 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
713 	}
714 
715 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
716 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
717 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
718 	}
719 
720 	while (start < end) {
721 		unsigned long useraddr, abs_addr;
722 
723 		/* Translate guest address to host address */
724 		if ((vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) == 0)
725 			abs_addr = kvm_s390_real_to_abs(vcpu, start);
726 		else
727 			abs_addr = start;
728 		useraddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(abs_addr));
729 		if (kvm_is_error_hva(useraddr))
730 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
731 
732 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
733 			if (clear_user((void __user *)useraddr, PAGE_SIZE))
734 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
735 		}
736 
737 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
738 			int rc = __skey_check_enable(vcpu);
739 
740 			if (rc)
741 				return rc;
742 			if (set_guest_storage_key(current->mm, useraddr,
743 					vcpu->run->s.regs.gprs[reg1] & PFMF_KEY,
744 					vcpu->run->s.regs.gprs[reg1] & PFMF_NQ))
745 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
746 		}
747 
748 		start += PAGE_SIZE;
749 	}
750 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC)
751 		vcpu->run->s.regs.gprs[reg2] = end;
752 	return 0;
753 }
754 
755 static int handle_essa(struct kvm_vcpu *vcpu)
756 {
757 	/* entries expected to be 1FF */
758 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
759 	unsigned long *cbrlo, cbrle;
760 	struct gmap *gmap;
761 	int i;
762 
763 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
764 	gmap = vcpu->arch.gmap;
765 	vcpu->stat.instruction_essa++;
766 	if (!vcpu->kvm->arch.use_cmma)
767 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
768 
769 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
770 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
771 
772 	if (((vcpu->arch.sie_block->ipb & 0xf0000000) >> 28) > 6)
773 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
774 
775 	/* Rewind PSW to repeat the ESSA instruction */
776 	kvm_s390_rewind_psw(vcpu, 4);
777 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
778 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
779 	down_read(&gmap->mm->mmap_sem);
780 	for (i = 0; i < entries; ++i) {
781 		cbrle = cbrlo[i];
782 		if (unlikely(cbrle & ~PAGE_MASK || cbrle < 2 * PAGE_SIZE))
783 			/* invalid entry */
784 			break;
785 		/* try to free backing */
786 		__gmap_zap(gmap, cbrle);
787 	}
788 	up_read(&gmap->mm->mmap_sem);
789 	if (i < entries)
790 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
791 	return 0;
792 }
793 
794 static const intercept_handler_t b9_handlers[256] = {
795 	[0x8a] = handle_ipte_interlock,
796 	[0x8d] = handle_epsw,
797 	[0x8e] = handle_ipte_interlock,
798 	[0x8f] = handle_ipte_interlock,
799 	[0xab] = handle_essa,
800 	[0xaf] = handle_pfmf,
801 };
802 
803 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
804 {
805 	intercept_handler_t handler;
806 
807 	/* This is handled just as for the B2 instructions. */
808 	handler = b9_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
809 	if (handler)
810 		return handler(vcpu);
811 
812 	return -EOPNOTSUPP;
813 }
814 
815 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
816 {
817 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
818 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
819 	int reg, rc, nr_regs;
820 	u32 ctl_array[16];
821 	u64 ga;
822 	ar_t ar;
823 
824 	vcpu->stat.instruction_lctl++;
825 
826 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
827 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
828 
829 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
830 
831 	if (ga & 3)
832 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
833 
834 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
835 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
836 
837 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
838 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
839 	if (rc)
840 		return kvm_s390_inject_prog_cond(vcpu, rc);
841 	reg = reg1;
842 	nr_regs = 0;
843 	do {
844 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
845 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
846 		if (reg == reg3)
847 			break;
848 		reg = (reg + 1) % 16;
849 	} while (1);
850 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
851 	return 0;
852 }
853 
854 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
855 {
856 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
857 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
858 	int reg, rc, nr_regs;
859 	u32 ctl_array[16];
860 	u64 ga;
861 	ar_t ar;
862 
863 	vcpu->stat.instruction_stctl++;
864 
865 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
866 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
867 
868 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
869 
870 	if (ga & 3)
871 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
872 
873 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
874 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
875 
876 	reg = reg1;
877 	nr_regs = 0;
878 	do {
879 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
880 		if (reg == reg3)
881 			break;
882 		reg = (reg + 1) % 16;
883 	} while (1);
884 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
885 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
886 }
887 
888 static int handle_lctlg(struct kvm_vcpu *vcpu)
889 {
890 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
891 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
892 	int reg, rc, nr_regs;
893 	u64 ctl_array[16];
894 	u64 ga;
895 	ar_t ar;
896 
897 	vcpu->stat.instruction_lctlg++;
898 
899 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
900 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
901 
902 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
903 
904 	if (ga & 7)
905 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
906 
907 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
908 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
909 
910 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
911 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
912 	if (rc)
913 		return kvm_s390_inject_prog_cond(vcpu, rc);
914 	reg = reg1;
915 	nr_regs = 0;
916 	do {
917 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
918 		if (reg == reg3)
919 			break;
920 		reg = (reg + 1) % 16;
921 	} while (1);
922 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
923 	return 0;
924 }
925 
926 static int handle_stctg(struct kvm_vcpu *vcpu)
927 {
928 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
929 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
930 	int reg, rc, nr_regs;
931 	u64 ctl_array[16];
932 	u64 ga;
933 	ar_t ar;
934 
935 	vcpu->stat.instruction_stctg++;
936 
937 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
938 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
939 
940 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
941 
942 	if (ga & 7)
943 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
944 
945 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
946 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
947 
948 	reg = reg1;
949 	nr_regs = 0;
950 	do {
951 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
952 		if (reg == reg3)
953 			break;
954 		reg = (reg + 1) % 16;
955 	} while (1);
956 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
957 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
958 }
959 
960 static const intercept_handler_t eb_handlers[256] = {
961 	[0x2f] = handle_lctlg,
962 	[0x25] = handle_stctg,
963 };
964 
965 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
966 {
967 	intercept_handler_t handler;
968 
969 	handler = eb_handlers[vcpu->arch.sie_block->ipb & 0xff];
970 	if (handler)
971 		return handler(vcpu);
972 	return -EOPNOTSUPP;
973 }
974 
975 static int handle_tprot(struct kvm_vcpu *vcpu)
976 {
977 	u64 address1, address2;
978 	unsigned long hva, gpa;
979 	int ret = 0, cc = 0;
980 	bool writable;
981 	ar_t ar;
982 
983 	vcpu->stat.instruction_tprot++;
984 
985 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
986 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
987 
988 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
989 
990 	/* we only handle the Linux memory detection case:
991 	 * access key == 0
992 	 * everything else goes to userspace. */
993 	if (address2 & 0xf0)
994 		return -EOPNOTSUPP;
995 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
996 		ipte_lock(vcpu);
997 	ret = guest_translate_address(vcpu, address1, ar, &gpa, 1);
998 	if (ret == PGM_PROTECTION) {
999 		/* Write protected? Try again with read-only... */
1000 		cc = 1;
1001 		ret = guest_translate_address(vcpu, address1, ar, &gpa, 0);
1002 	}
1003 	if (ret) {
1004 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1005 			ret = kvm_s390_inject_program_int(vcpu, ret);
1006 		} else if (ret > 0) {
1007 			/* Translation not available */
1008 			kvm_s390_set_psw_cc(vcpu, 3);
1009 			ret = 0;
1010 		}
1011 		goto out_unlock;
1012 	}
1013 
1014 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1015 	if (kvm_is_error_hva(hva)) {
1016 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1017 	} else {
1018 		if (!writable)
1019 			cc = 1;		/* Write not permitted ==> read-only */
1020 		kvm_s390_set_psw_cc(vcpu, cc);
1021 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1022 	}
1023 out_unlock:
1024 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1025 		ipte_unlock(vcpu);
1026 	return ret;
1027 }
1028 
1029 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1030 {
1031 	/* For e5xx... instructions we only handle TPROT */
1032 	if ((vcpu->arch.sie_block->ipa & 0x00ff) == 0x01)
1033 		return handle_tprot(vcpu);
1034 	return -EOPNOTSUPP;
1035 }
1036 
1037 static int handle_sckpf(struct kvm_vcpu *vcpu)
1038 {
1039 	u32 value;
1040 
1041 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1042 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1043 
1044 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1045 		return kvm_s390_inject_program_int(vcpu,
1046 						   PGM_SPECIFICATION);
1047 
1048 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1049 	vcpu->arch.sie_block->todpr = value;
1050 
1051 	return 0;
1052 }
1053 
1054 static const intercept_handler_t x01_handlers[256] = {
1055 	[0x07] = handle_sckpf,
1056 };
1057 
1058 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1059 {
1060 	intercept_handler_t handler;
1061 
1062 	handler = x01_handlers[vcpu->arch.sie_block->ipa & 0x00ff];
1063 	if (handler)
1064 		return handler(vcpu);
1065 	return -EOPNOTSUPP;
1066 }
1067