• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * handling privileged instructions
4  *
5  * Copyright IBM Corp. 2008, 2020
6  *
7  *    Author(s): Carsten Otte <cotte@de.ibm.com>
8  *               Christian Borntraeger <borntraeger@de.ibm.com>
9  */
10 
11 #include <linux/kvm.h>
12 #include <linux/gfp.h>
13 #include <linux/errno.h>
14 #include <linux/compat.h>
15 #include <linux/mm_types.h>
16 #include <linux/pgtable.h>
17 
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/page-states.h>
25 #include <asm/gmap.h>
26 #include <asm/io.h>
27 #include <asm/ptrace.h>
28 #include <asm/sclp.h>
29 #include <asm/ap.h>
30 #include "gaccess.h"
31 #include "kvm-s390.h"
32 #include "trace.h"
33 
handle_ri(struct kvm_vcpu * vcpu)34 static int handle_ri(struct kvm_vcpu *vcpu)
35 {
36 	vcpu->stat.instruction_ri++;
37 
38 	if (test_kvm_facility(vcpu->kvm, 64)) {
39 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (lazy)");
40 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
41 		kvm_s390_retry_instr(vcpu);
42 		return 0;
43 	} else
44 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
45 }
46 
kvm_s390_handle_aa(struct kvm_vcpu * vcpu)47 int kvm_s390_handle_aa(struct kvm_vcpu *vcpu)
48 {
49 	if ((vcpu->arch.sie_block->ipa & 0xf) <= 4)
50 		return handle_ri(vcpu);
51 	else
52 		return -EOPNOTSUPP;
53 }
54 
handle_gs(struct kvm_vcpu * vcpu)55 static int handle_gs(struct kvm_vcpu *vcpu)
56 {
57 	vcpu->stat.instruction_gs++;
58 
59 	if (test_kvm_facility(vcpu->kvm, 133)) {
60 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (lazy)");
61 		preempt_disable();
62 		__ctl_set_bit(2, 4);
63 		current->thread.gs_cb = (struct gs_cb *)&vcpu->run->s.regs.gscb;
64 		restore_gs_cb(current->thread.gs_cb);
65 		preempt_enable();
66 		vcpu->arch.sie_block->ecb |= ECB_GS;
67 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
68 		vcpu->arch.gs_enabled = 1;
69 		kvm_s390_retry_instr(vcpu);
70 		return 0;
71 	} else
72 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
73 }
74 
kvm_s390_handle_e3(struct kvm_vcpu * vcpu)75 int kvm_s390_handle_e3(struct kvm_vcpu *vcpu)
76 {
77 	int code = vcpu->arch.sie_block->ipb & 0xff;
78 
79 	if (code == 0x49 || code == 0x4d)
80 		return handle_gs(vcpu);
81 	else
82 		return -EOPNOTSUPP;
83 }
84 /* Handle SCK (SET CLOCK) interception */
handle_set_clock(struct kvm_vcpu * vcpu)85 static int handle_set_clock(struct kvm_vcpu *vcpu)
86 {
87 	struct kvm_s390_vm_tod_clock gtod = { 0 };
88 	int rc;
89 	u8 ar;
90 	u64 op2;
91 
92 	vcpu->stat.instruction_sck++;
93 
94 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
95 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
96 
97 	op2 = kvm_s390_get_base_disp_s(vcpu, &ar);
98 	if (op2 & 7)	/* Operand must be on a doubleword boundary */
99 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
100 	rc = read_guest(vcpu, op2, ar, &gtod.tod, sizeof(gtod.tod));
101 	if (rc)
102 		return kvm_s390_inject_prog_cond(vcpu, rc);
103 
104 	VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", gtod.tod);
105 	kvm_s390_set_tod_clock(vcpu->kvm, &gtod);
106 
107 	kvm_s390_set_psw_cc(vcpu, 0);
108 	return 0;
109 }
110 
handle_set_prefix(struct kvm_vcpu * vcpu)111 static int handle_set_prefix(struct kvm_vcpu *vcpu)
112 {
113 	u64 operand2;
114 	u32 address;
115 	int rc;
116 	u8 ar;
117 
118 	vcpu->stat.instruction_spx++;
119 
120 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
121 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
122 
123 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
124 
125 	/* must be word boundary */
126 	if (operand2 & 3)
127 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
128 
129 	/* get the value */
130 	rc = read_guest(vcpu, operand2, ar, &address, sizeof(address));
131 	if (rc)
132 		return kvm_s390_inject_prog_cond(vcpu, rc);
133 
134 	address &= 0x7fffe000u;
135 
136 	/*
137 	 * Make sure the new value is valid memory. We only need to check the
138 	 * first page, since address is 8k aligned and memory pieces are always
139 	 * at least 1MB aligned and have at least a size of 1MB.
140 	 */
141 	if (kvm_is_error_gpa(vcpu->kvm, address))
142 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
143 
144 	kvm_s390_set_prefix(vcpu, address);
145 	trace_kvm_s390_handle_prefix(vcpu, 1, address);
146 	return 0;
147 }
148 
handle_store_prefix(struct kvm_vcpu * vcpu)149 static int handle_store_prefix(struct kvm_vcpu *vcpu)
150 {
151 	u64 operand2;
152 	u32 address;
153 	int rc;
154 	u8 ar;
155 
156 	vcpu->stat.instruction_stpx++;
157 
158 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
159 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
160 
161 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
162 
163 	/* must be word boundary */
164 	if (operand2 & 3)
165 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
166 
167 	address = kvm_s390_get_prefix(vcpu);
168 
169 	/* get the value */
170 	rc = write_guest(vcpu, operand2, ar, &address, sizeof(address));
171 	if (rc)
172 		return kvm_s390_inject_prog_cond(vcpu, rc);
173 
174 	VCPU_EVENT(vcpu, 3, "STPX: storing prefix 0x%x into 0x%llx", address, operand2);
175 	trace_kvm_s390_handle_prefix(vcpu, 0, address);
176 	return 0;
177 }
178 
handle_store_cpu_address(struct kvm_vcpu * vcpu)179 static int handle_store_cpu_address(struct kvm_vcpu *vcpu)
180 {
181 	u16 vcpu_id = vcpu->vcpu_id;
182 	u64 ga;
183 	int rc;
184 	u8 ar;
185 
186 	vcpu->stat.instruction_stap++;
187 
188 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
189 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
190 
191 	ga = kvm_s390_get_base_disp_s(vcpu, &ar);
192 
193 	if (ga & 1)
194 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
195 
196 	rc = write_guest(vcpu, ga, ar, &vcpu_id, sizeof(vcpu_id));
197 	if (rc)
198 		return kvm_s390_inject_prog_cond(vcpu, rc);
199 
200 	VCPU_EVENT(vcpu, 3, "STAP: storing cpu address (%u) to 0x%llx", vcpu_id, ga);
201 	trace_kvm_s390_handle_stap(vcpu, ga);
202 	return 0;
203 }
204 
kvm_s390_skey_check_enable(struct kvm_vcpu * vcpu)205 int kvm_s390_skey_check_enable(struct kvm_vcpu *vcpu)
206 {
207 	int rc;
208 
209 	trace_kvm_s390_skey_related_inst(vcpu);
210 	/* Already enabled? */
211 	if (vcpu->arch.skey_enabled)
212 		return 0;
213 
214 	rc = s390_enable_skey();
215 	VCPU_EVENT(vcpu, 3, "enabling storage keys for guest: %d", rc);
216 	if (rc)
217 		return rc;
218 
219 	if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS))
220 		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS);
221 	if (!vcpu->kvm->arch.use_skf)
222 		vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
223 	else
224 		vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE);
225 	vcpu->arch.skey_enabled = true;
226 	return 0;
227 }
228 
try_handle_skey(struct kvm_vcpu * vcpu)229 static int try_handle_skey(struct kvm_vcpu *vcpu)
230 {
231 	int rc;
232 
233 	rc = kvm_s390_skey_check_enable(vcpu);
234 	if (rc)
235 		return rc;
236 	if (vcpu->kvm->arch.use_skf) {
237 		/* with storage-key facility, SIE interprets it for us */
238 		kvm_s390_retry_instr(vcpu);
239 		VCPU_EVENT(vcpu, 4, "%s", "retrying storage key operation");
240 		return -EAGAIN;
241 	}
242 	return 0;
243 }
244 
handle_iske(struct kvm_vcpu * vcpu)245 static int handle_iske(struct kvm_vcpu *vcpu)
246 {
247 	unsigned long gaddr, vmaddr;
248 	unsigned char key;
249 	int reg1, reg2;
250 	bool unlocked;
251 	int rc;
252 
253 	vcpu->stat.instruction_iske++;
254 
255 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
256 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
257 
258 	rc = try_handle_skey(vcpu);
259 	if (rc)
260 		return rc != -EAGAIN ? rc : 0;
261 
262 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
263 
264 	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
265 	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
266 	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
267 	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
268 	if (kvm_is_error_hva(vmaddr))
269 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
270 retry:
271 	unlocked = false;
272 	mmap_read_lock(current->mm);
273 	rc = get_guest_storage_key(current->mm, vmaddr, &key);
274 
275 	if (rc) {
276 		rc = fixup_user_fault(current->mm, vmaddr,
277 				      FAULT_FLAG_WRITE, &unlocked);
278 		if (!rc) {
279 			mmap_read_unlock(current->mm);
280 			goto retry;
281 		}
282 	}
283 	mmap_read_unlock(current->mm);
284 	if (rc == -EFAULT)
285 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
286 	if (rc < 0)
287 		return rc;
288 	vcpu->run->s.regs.gprs[reg1] &= ~0xff;
289 	vcpu->run->s.regs.gprs[reg1] |= key;
290 	return 0;
291 }
292 
handle_rrbe(struct kvm_vcpu * vcpu)293 static int handle_rrbe(struct kvm_vcpu *vcpu)
294 {
295 	unsigned long vmaddr, gaddr;
296 	int reg1, reg2;
297 	bool unlocked;
298 	int rc;
299 
300 	vcpu->stat.instruction_rrbe++;
301 
302 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
303 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
304 
305 	rc = try_handle_skey(vcpu);
306 	if (rc)
307 		return rc != -EAGAIN ? rc : 0;
308 
309 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
310 
311 	gaddr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
312 	gaddr = kvm_s390_logical_to_effective(vcpu, gaddr);
313 	gaddr = kvm_s390_real_to_abs(vcpu, gaddr);
314 	vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(gaddr));
315 	if (kvm_is_error_hva(vmaddr))
316 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
317 retry:
318 	unlocked = false;
319 	mmap_read_lock(current->mm);
320 	rc = reset_guest_reference_bit(current->mm, vmaddr);
321 	if (rc < 0) {
322 		rc = fixup_user_fault(current->mm, vmaddr,
323 				      FAULT_FLAG_WRITE, &unlocked);
324 		if (!rc) {
325 			mmap_read_unlock(current->mm);
326 			goto retry;
327 		}
328 	}
329 	mmap_read_unlock(current->mm);
330 	if (rc == -EFAULT)
331 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
332 	if (rc < 0)
333 		return rc;
334 	kvm_s390_set_psw_cc(vcpu, rc);
335 	return 0;
336 }
337 
338 #define SSKE_NQ 0x8
339 #define SSKE_MR 0x4
340 #define SSKE_MC 0x2
341 #define SSKE_MB 0x1
handle_sske(struct kvm_vcpu * vcpu)342 static int handle_sske(struct kvm_vcpu *vcpu)
343 {
344 	unsigned char m3 = vcpu->arch.sie_block->ipb >> 28;
345 	unsigned long start, end;
346 	unsigned char key, oldkey;
347 	int reg1, reg2;
348 	bool unlocked;
349 	int rc;
350 
351 	vcpu->stat.instruction_sske++;
352 
353 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
354 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
355 
356 	rc = try_handle_skey(vcpu);
357 	if (rc)
358 		return rc != -EAGAIN ? rc : 0;
359 
360 	if (!test_kvm_facility(vcpu->kvm, 8))
361 		m3 &= ~SSKE_MB;
362 	if (!test_kvm_facility(vcpu->kvm, 10))
363 		m3 &= ~(SSKE_MC | SSKE_MR);
364 	if (!test_kvm_facility(vcpu->kvm, 14))
365 		m3 &= ~SSKE_NQ;
366 
367 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
368 
369 	key = vcpu->run->s.regs.gprs[reg1] & 0xfe;
370 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
371 	start = kvm_s390_logical_to_effective(vcpu, start);
372 	if (m3 & SSKE_MB) {
373 		/* start already designates an absolute address */
374 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
375 	} else {
376 		start = kvm_s390_real_to_abs(vcpu, start);
377 		end = start + PAGE_SIZE;
378 	}
379 
380 	while (start != end) {
381 		unsigned long vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
382 		unlocked = false;
383 
384 		if (kvm_is_error_hva(vmaddr))
385 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
386 
387 		mmap_read_lock(current->mm);
388 		rc = cond_set_guest_storage_key(current->mm, vmaddr, key, &oldkey,
389 						m3 & SSKE_NQ, m3 & SSKE_MR,
390 						m3 & SSKE_MC);
391 
392 		if (rc < 0) {
393 			rc = fixup_user_fault(current->mm, vmaddr,
394 					      FAULT_FLAG_WRITE, &unlocked);
395 			rc = !rc ? -EAGAIN : rc;
396 		}
397 		mmap_read_unlock(current->mm);
398 		if (rc == -EFAULT)
399 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
400 		if (rc == -EAGAIN)
401 			continue;
402 		if (rc < 0)
403 			return rc;
404 		start += PAGE_SIZE;
405 	}
406 
407 	if (m3 & (SSKE_MC | SSKE_MR)) {
408 		if (m3 & SSKE_MB) {
409 			/* skey in reg1 is unpredictable */
410 			kvm_s390_set_psw_cc(vcpu, 3);
411 		} else {
412 			kvm_s390_set_psw_cc(vcpu, rc);
413 			vcpu->run->s.regs.gprs[reg1] &= ~0xff00UL;
414 			vcpu->run->s.regs.gprs[reg1] |= (u64) oldkey << 8;
415 		}
416 	}
417 	if (m3 & SSKE_MB) {
418 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT)
419 			vcpu->run->s.regs.gprs[reg2] &= ~PAGE_MASK;
420 		else
421 			vcpu->run->s.regs.gprs[reg2] &= ~0xfffff000UL;
422 		end = kvm_s390_logical_to_effective(vcpu, end);
423 		vcpu->run->s.regs.gprs[reg2] |= end;
424 	}
425 	return 0;
426 }
427 
handle_ipte_interlock(struct kvm_vcpu * vcpu)428 static int handle_ipte_interlock(struct kvm_vcpu *vcpu)
429 {
430 	vcpu->stat.instruction_ipte_interlock++;
431 	if (psw_bits(vcpu->arch.sie_block->gpsw).pstate)
432 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
433 	wait_event(vcpu->kvm->arch.ipte_wq, !ipte_lock_held(vcpu));
434 	kvm_s390_retry_instr(vcpu);
435 	VCPU_EVENT(vcpu, 4, "%s", "retrying ipte interlock operation");
436 	return 0;
437 }
438 
handle_test_block(struct kvm_vcpu * vcpu)439 static int handle_test_block(struct kvm_vcpu *vcpu)
440 {
441 	gpa_t addr;
442 	int reg2;
443 
444 	vcpu->stat.instruction_tb++;
445 
446 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
447 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
448 
449 	kvm_s390_get_regs_rre(vcpu, NULL, &reg2);
450 	addr = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
451 	addr = kvm_s390_logical_to_effective(vcpu, addr);
452 	if (kvm_s390_check_low_addr_prot_real(vcpu, addr))
453 		return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
454 	addr = kvm_s390_real_to_abs(vcpu, addr);
455 
456 	if (kvm_is_error_gpa(vcpu->kvm, addr))
457 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
458 	/*
459 	 * We don't expect errors on modern systems, and do not care
460 	 * about storage keys (yet), so let's just clear the page.
461 	 */
462 	if (kvm_clear_guest(vcpu->kvm, addr, PAGE_SIZE))
463 		return -EFAULT;
464 	kvm_s390_set_psw_cc(vcpu, 0);
465 	vcpu->run->s.regs.gprs[0] = 0;
466 	return 0;
467 }
468 
handle_tpi(struct kvm_vcpu * vcpu)469 static int handle_tpi(struct kvm_vcpu *vcpu)
470 {
471 	struct kvm_s390_interrupt_info *inti;
472 	unsigned long len;
473 	u32 tpi_data[3];
474 	int rc;
475 	u64 addr;
476 	u8 ar;
477 
478 	vcpu->stat.instruction_tpi++;
479 
480 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
481 	if (addr & 3)
482 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
483 
484 	inti = kvm_s390_get_io_int(vcpu->kvm, vcpu->arch.sie_block->gcr[6], 0);
485 	if (!inti) {
486 		kvm_s390_set_psw_cc(vcpu, 0);
487 		return 0;
488 	}
489 
490 	tpi_data[0] = inti->io.subchannel_id << 16 | inti->io.subchannel_nr;
491 	tpi_data[1] = inti->io.io_int_parm;
492 	tpi_data[2] = inti->io.io_int_word;
493 	if (addr) {
494 		/*
495 		 * Store the two-word I/O interruption code into the
496 		 * provided area.
497 		 */
498 		len = sizeof(tpi_data) - 4;
499 		rc = write_guest(vcpu, addr, ar, &tpi_data, len);
500 		if (rc) {
501 			rc = kvm_s390_inject_prog_cond(vcpu, rc);
502 			goto reinject_interrupt;
503 		}
504 	} else {
505 		/*
506 		 * Store the three-word I/O interruption code into
507 		 * the appropriate lowcore area.
508 		 */
509 		len = sizeof(tpi_data);
510 		if (write_guest_lc(vcpu, __LC_SUBCHANNEL_ID, &tpi_data, len)) {
511 			/* failed writes to the low core are not recoverable */
512 			rc = -EFAULT;
513 			goto reinject_interrupt;
514 		}
515 	}
516 
517 	/* irq was successfully handed to the guest */
518 	kfree(inti);
519 	kvm_s390_set_psw_cc(vcpu, 1);
520 	return 0;
521 reinject_interrupt:
522 	/*
523 	 * If we encounter a problem storing the interruption code, the
524 	 * instruction is suppressed from the guest's view: reinject the
525 	 * interrupt.
526 	 */
527 	if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
528 		kfree(inti);
529 		rc = -EFAULT;
530 	}
531 	/* don't set the cc, a pgm irq was injected or we drop to user space */
532 	return rc ? -EFAULT : 0;
533 }
534 
handle_tsch(struct kvm_vcpu * vcpu)535 static int handle_tsch(struct kvm_vcpu *vcpu)
536 {
537 	struct kvm_s390_interrupt_info *inti = NULL;
538 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
539 
540 	vcpu->stat.instruction_tsch++;
541 
542 	/* a valid schid has at least one bit set */
543 	if (vcpu->run->s.regs.gprs[1])
544 		inti = kvm_s390_get_io_int(vcpu->kvm, isc_mask,
545 					   vcpu->run->s.regs.gprs[1]);
546 
547 	/*
548 	 * Prepare exit to userspace.
549 	 * We indicate whether we dequeued a pending I/O interrupt
550 	 * so that userspace can re-inject it if the instruction gets
551 	 * a program check. While this may re-order the pending I/O
552 	 * interrupts, this is no problem since the priority is kept
553 	 * intact.
554 	 */
555 	vcpu->run->exit_reason = KVM_EXIT_S390_TSCH;
556 	vcpu->run->s390_tsch.dequeued = !!inti;
557 	if (inti) {
558 		vcpu->run->s390_tsch.subchannel_id = inti->io.subchannel_id;
559 		vcpu->run->s390_tsch.subchannel_nr = inti->io.subchannel_nr;
560 		vcpu->run->s390_tsch.io_int_parm = inti->io.io_int_parm;
561 		vcpu->run->s390_tsch.io_int_word = inti->io.io_int_word;
562 	}
563 	vcpu->run->s390_tsch.ipb = vcpu->arch.sie_block->ipb;
564 	kfree(inti);
565 	return -EREMOTE;
566 }
567 
handle_io_inst(struct kvm_vcpu * vcpu)568 static int handle_io_inst(struct kvm_vcpu *vcpu)
569 {
570 	VCPU_EVENT(vcpu, 4, "%s", "I/O instruction");
571 
572 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
573 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
574 
575 	if (vcpu->kvm->arch.css_support) {
576 		/*
577 		 * Most I/O instructions will be handled by userspace.
578 		 * Exceptions are tpi and the interrupt portion of tsch.
579 		 */
580 		if (vcpu->arch.sie_block->ipa == 0xb236)
581 			return handle_tpi(vcpu);
582 		if (vcpu->arch.sie_block->ipa == 0xb235)
583 			return handle_tsch(vcpu);
584 		/* Handle in userspace. */
585 		vcpu->stat.instruction_io_other++;
586 		return -EOPNOTSUPP;
587 	} else {
588 		/*
589 		 * Set condition code 3 to stop the guest from issuing channel
590 		 * I/O instructions.
591 		 */
592 		kvm_s390_set_psw_cc(vcpu, 3);
593 		return 0;
594 	}
595 }
596 
597 /*
598  * handle_pqap: Handling pqap interception
599  * @vcpu: the vcpu having issue the pqap instruction
600  *
601  * We now support PQAP/AQIC instructions and we need to correctly
602  * answer the guest even if no dedicated driver's hook is available.
603  *
604  * The intercepting code calls a dedicated callback for this instruction
605  * if a driver did register one in the CRYPTO satellite of the
606  * SIE block.
607  *
608  * If no callback is available, the queues are not available, return this
609  * response code to the caller and set CC to 3.
610  * Else return the response code returned by the callback.
611  */
handle_pqap(struct kvm_vcpu * vcpu)612 static int handle_pqap(struct kvm_vcpu *vcpu)
613 {
614 	struct ap_queue_status status = {};
615 	unsigned long reg0;
616 	int ret;
617 	uint8_t fc;
618 
619 	/* Verify that the AP instruction are available */
620 	if (!ap_instructions_available())
621 		return -EOPNOTSUPP;
622 	/* Verify that the guest is allowed to use AP instructions */
623 	if (!(vcpu->arch.sie_block->eca & ECA_APIE))
624 		return -EOPNOTSUPP;
625 	/*
626 	 * The only possibly intercepted functions when AP instructions are
627 	 * available for the guest are AQIC and TAPQ with the t bit set
628 	 * since we do not set IC.3 (FIII) we currently will only intercept
629 	 * the AQIC function code.
630 	 * Note: running nested under z/VM can result in intercepts for other
631 	 * function codes, e.g. PQAP(QCI). We do not support this and bail out.
632 	 */
633 	reg0 = vcpu->run->s.regs.gprs[0];
634 	fc = (reg0 >> 24) & 0xff;
635 	if (fc != 0x03)
636 		return -EOPNOTSUPP;
637 
638 	/* PQAP instruction is allowed for guest kernel only */
639 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
640 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
641 
642 	/* Common PQAP instruction specification exceptions */
643 	/* bits 41-47 must all be zeros */
644 	if (reg0 & 0x007f0000UL)
645 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
646 	/* APFT not install and T bit set */
647 	if (!test_kvm_facility(vcpu->kvm, 15) && (reg0 & 0x00800000UL))
648 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
649 	/* APXA not installed and APID greater 64 or APQI greater 16 */
650 	if (!(vcpu->kvm->arch.crypto.crycbd & 0x02) && (reg0 & 0x0000c0f0UL))
651 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
652 
653 	/* AQIC function code specific exception */
654 	/* facility 65 not present for AQIC function code */
655 	if (!test_kvm_facility(vcpu->kvm, 65))
656 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
657 
658 	/*
659 	 * Verify that the hook callback is registered, lock the owner
660 	 * and call the hook.
661 	 */
662 	if (vcpu->kvm->arch.crypto.pqap_hook) {
663 		if (!try_module_get(vcpu->kvm->arch.crypto.pqap_hook->owner))
664 			return -EOPNOTSUPP;
665 		ret = vcpu->kvm->arch.crypto.pqap_hook->hook(vcpu);
666 		module_put(vcpu->kvm->arch.crypto.pqap_hook->owner);
667 		if (!ret && vcpu->run->s.regs.gprs[1] & 0x00ff0000)
668 			kvm_s390_set_psw_cc(vcpu, 3);
669 		return ret;
670 	}
671 	/*
672 	 * A vfio_driver must register a hook.
673 	 * No hook means no driver to enable the SIE CRYCB and no queues.
674 	 * We send this response to the guest.
675 	 */
676 	status.response_code = 0x01;
677 	memcpy(&vcpu->run->s.regs.gprs[1], &status, sizeof(status));
678 	kvm_s390_set_psw_cc(vcpu, 3);
679 	return 0;
680 }
681 
handle_stfl(struct kvm_vcpu * vcpu)682 static int handle_stfl(struct kvm_vcpu *vcpu)
683 {
684 	int rc;
685 	unsigned int fac;
686 
687 	vcpu->stat.instruction_stfl++;
688 
689 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
690 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
691 
692 	/*
693 	 * We need to shift the lower 32 facility bits (bit 0-31) from a u64
694 	 * into a u32 memory representation. They will remain bits 0-31.
695 	 */
696 	fac = *vcpu->kvm->arch.model.fac_list >> 32;
697 	rc = write_guest_lc(vcpu, offsetof(struct lowcore, stfl_fac_list),
698 			    &fac, sizeof(fac));
699 	if (rc)
700 		return rc;
701 	VCPU_EVENT(vcpu, 3, "STFL: store facility list 0x%x", fac);
702 	trace_kvm_s390_handle_stfl(vcpu, fac);
703 	return 0;
704 }
705 
706 #define PSW_MASK_ADDR_MODE (PSW_MASK_EA | PSW_MASK_BA)
707 #define PSW_MASK_UNASSIGNED 0xb80800fe7fffffffUL
708 #define PSW_ADDR_24 0x0000000000ffffffUL
709 #define PSW_ADDR_31 0x000000007fffffffUL
710 
is_valid_psw(psw_t * psw)711 int is_valid_psw(psw_t *psw)
712 {
713 	if (psw->mask & PSW_MASK_UNASSIGNED)
714 		return 0;
715 	if ((psw->mask & PSW_MASK_ADDR_MODE) == PSW_MASK_BA) {
716 		if (psw->addr & ~PSW_ADDR_31)
717 			return 0;
718 	}
719 	if (!(psw->mask & PSW_MASK_ADDR_MODE) && (psw->addr & ~PSW_ADDR_24))
720 		return 0;
721 	if ((psw->mask & PSW_MASK_ADDR_MODE) ==  PSW_MASK_EA)
722 		return 0;
723 	if (psw->addr & 1)
724 		return 0;
725 	return 1;
726 }
727 
kvm_s390_handle_lpsw(struct kvm_vcpu * vcpu)728 int kvm_s390_handle_lpsw(struct kvm_vcpu *vcpu)
729 {
730 	psw_t *gpsw = &vcpu->arch.sie_block->gpsw;
731 	psw_compat_t new_psw;
732 	u64 addr;
733 	int rc;
734 	u8 ar;
735 
736 	vcpu->stat.instruction_lpsw++;
737 
738 	if (gpsw->mask & PSW_MASK_PSTATE)
739 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
740 
741 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
742 	if (addr & 7)
743 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
744 
745 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
746 	if (rc)
747 		return kvm_s390_inject_prog_cond(vcpu, rc);
748 	if (!(new_psw.mask & PSW32_MASK_BASE))
749 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
750 	gpsw->mask = (new_psw.mask & ~PSW32_MASK_BASE) << 32;
751 	gpsw->mask |= new_psw.addr & PSW32_ADDR_AMODE;
752 	gpsw->addr = new_psw.addr & ~PSW32_ADDR_AMODE;
753 	if (!is_valid_psw(gpsw))
754 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
755 	return 0;
756 }
757 
handle_lpswe(struct kvm_vcpu * vcpu)758 static int handle_lpswe(struct kvm_vcpu *vcpu)
759 {
760 	psw_t new_psw;
761 	u64 addr;
762 	int rc;
763 	u8 ar;
764 
765 	vcpu->stat.instruction_lpswe++;
766 
767 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
768 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
769 
770 	addr = kvm_s390_get_base_disp_s(vcpu, &ar);
771 	if (addr & 7)
772 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
773 	rc = read_guest(vcpu, addr, ar, &new_psw, sizeof(new_psw));
774 	if (rc)
775 		return kvm_s390_inject_prog_cond(vcpu, rc);
776 	vcpu->arch.sie_block->gpsw = new_psw;
777 	if (!is_valid_psw(&vcpu->arch.sie_block->gpsw))
778 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
779 	return 0;
780 }
781 
handle_stidp(struct kvm_vcpu * vcpu)782 static int handle_stidp(struct kvm_vcpu *vcpu)
783 {
784 	u64 stidp_data = vcpu->kvm->arch.model.cpuid;
785 	u64 operand2;
786 	int rc;
787 	u8 ar;
788 
789 	vcpu->stat.instruction_stidp++;
790 
791 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
792 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
793 
794 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
795 
796 	if (operand2 & 7)
797 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
798 
799 	rc = write_guest(vcpu, operand2, ar, &stidp_data, sizeof(stidp_data));
800 	if (rc)
801 		return kvm_s390_inject_prog_cond(vcpu, rc);
802 
803 	VCPU_EVENT(vcpu, 3, "STIDP: store cpu id 0x%llx", stidp_data);
804 	return 0;
805 }
806 
handle_stsi_3_2_2(struct kvm_vcpu * vcpu,struct sysinfo_3_2_2 * mem)807 static void handle_stsi_3_2_2(struct kvm_vcpu *vcpu, struct sysinfo_3_2_2 *mem)
808 {
809 	int cpus = 0;
810 	int n;
811 
812 	cpus = atomic_read(&vcpu->kvm->online_vcpus);
813 
814 	/* deal with other level 3 hypervisors */
815 	if (stsi(mem, 3, 2, 2))
816 		mem->count = 0;
817 	if (mem->count < 8)
818 		mem->count++;
819 	for (n = mem->count - 1; n > 0 ; n--)
820 		memcpy(&mem->vm[n], &mem->vm[n - 1], sizeof(mem->vm[0]));
821 
822 	memset(&mem->vm[0], 0, sizeof(mem->vm[0]));
823 	mem->vm[0].cpus_total = cpus;
824 	mem->vm[0].cpus_configured = cpus;
825 	mem->vm[0].cpus_standby = 0;
826 	mem->vm[0].cpus_reserved = 0;
827 	mem->vm[0].caf = 1000;
828 	memcpy(mem->vm[0].name, "KVMguest", 8);
829 	ASCEBC(mem->vm[0].name, 8);
830 	memcpy(mem->vm[0].cpi, "KVM/Linux       ", 16);
831 	ASCEBC(mem->vm[0].cpi, 16);
832 }
833 
insert_stsi_usr_data(struct kvm_vcpu * vcpu,u64 addr,u8 ar,u8 fc,u8 sel1,u16 sel2)834 static void insert_stsi_usr_data(struct kvm_vcpu *vcpu, u64 addr, u8 ar,
835 				 u8 fc, u8 sel1, u16 sel2)
836 {
837 	vcpu->run->exit_reason = KVM_EXIT_S390_STSI;
838 	vcpu->run->s390_stsi.addr = addr;
839 	vcpu->run->s390_stsi.ar = ar;
840 	vcpu->run->s390_stsi.fc = fc;
841 	vcpu->run->s390_stsi.sel1 = sel1;
842 	vcpu->run->s390_stsi.sel2 = sel2;
843 }
844 
handle_stsi(struct kvm_vcpu * vcpu)845 static int handle_stsi(struct kvm_vcpu *vcpu)
846 {
847 	int fc = (vcpu->run->s.regs.gprs[0] & 0xf0000000) >> 28;
848 	int sel1 = vcpu->run->s.regs.gprs[0] & 0xff;
849 	int sel2 = vcpu->run->s.regs.gprs[1] & 0xffff;
850 	unsigned long mem = 0;
851 	u64 operand2;
852 	int rc = 0;
853 	u8 ar;
854 
855 	vcpu->stat.instruction_stsi++;
856 	VCPU_EVENT(vcpu, 3, "STSI: fc: %u sel1: %u sel2: %u", fc, sel1, sel2);
857 
858 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
859 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
860 
861 	if (fc > 3) {
862 		kvm_s390_set_psw_cc(vcpu, 3);
863 		return 0;
864 	}
865 
866 	if (vcpu->run->s.regs.gprs[0] & 0x0fffff00
867 	    || vcpu->run->s.regs.gprs[1] & 0xffff0000)
868 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
869 
870 	if (fc == 0) {
871 		vcpu->run->s.regs.gprs[0] = 3 << 28;
872 		kvm_s390_set_psw_cc(vcpu, 0);
873 		return 0;
874 	}
875 
876 	operand2 = kvm_s390_get_base_disp_s(vcpu, &ar);
877 
878 	if (!kvm_s390_pv_cpu_is_protected(vcpu) && (operand2 & 0xfff))
879 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
880 
881 	switch (fc) {
882 	case 1: /* same handling for 1 and 2 */
883 	case 2:
884 		mem = get_zeroed_page(GFP_KERNEL);
885 		if (!mem)
886 			goto out_no_data;
887 		if (stsi((void *) mem, fc, sel1, sel2))
888 			goto out_no_data;
889 		break;
890 	case 3:
891 		if (sel1 != 2 || sel2 != 2)
892 			goto out_no_data;
893 		mem = get_zeroed_page(GFP_KERNEL);
894 		if (!mem)
895 			goto out_no_data;
896 		handle_stsi_3_2_2(vcpu, (void *) mem);
897 		break;
898 	}
899 	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
900 		memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,
901 		       PAGE_SIZE);
902 		rc = 0;
903 	} else {
904 		rc = write_guest(vcpu, operand2, ar, (void *)mem, PAGE_SIZE);
905 	}
906 	if (rc) {
907 		rc = kvm_s390_inject_prog_cond(vcpu, rc);
908 		goto out;
909 	}
910 	if (vcpu->kvm->arch.user_stsi) {
911 		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
912 		rc = -EREMOTE;
913 	}
914 	trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
915 	free_page(mem);
916 	kvm_s390_set_psw_cc(vcpu, 0);
917 	vcpu->run->s.regs.gprs[0] = 0;
918 	return rc;
919 out_no_data:
920 	kvm_s390_set_psw_cc(vcpu, 3);
921 out:
922 	free_page(mem);
923 	return rc;
924 }
925 
kvm_s390_handle_b2(struct kvm_vcpu * vcpu)926 int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
927 {
928 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
929 	case 0x02:
930 		return handle_stidp(vcpu);
931 	case 0x04:
932 		return handle_set_clock(vcpu);
933 	case 0x10:
934 		return handle_set_prefix(vcpu);
935 	case 0x11:
936 		return handle_store_prefix(vcpu);
937 	case 0x12:
938 		return handle_store_cpu_address(vcpu);
939 	case 0x14:
940 		return kvm_s390_handle_vsie(vcpu);
941 	case 0x21:
942 	case 0x50:
943 		return handle_ipte_interlock(vcpu);
944 	case 0x29:
945 		return handle_iske(vcpu);
946 	case 0x2a:
947 		return handle_rrbe(vcpu);
948 	case 0x2b:
949 		return handle_sske(vcpu);
950 	case 0x2c:
951 		return handle_test_block(vcpu);
952 	case 0x30:
953 	case 0x31:
954 	case 0x32:
955 	case 0x33:
956 	case 0x34:
957 	case 0x35:
958 	case 0x36:
959 	case 0x37:
960 	case 0x38:
961 	case 0x39:
962 	case 0x3a:
963 	case 0x3b:
964 	case 0x3c:
965 	case 0x5f:
966 	case 0x74:
967 	case 0x76:
968 		return handle_io_inst(vcpu);
969 	case 0x56:
970 		return handle_sthyi(vcpu);
971 	case 0x7d:
972 		return handle_stsi(vcpu);
973 	case 0xaf:
974 		return handle_pqap(vcpu);
975 	case 0xb1:
976 		return handle_stfl(vcpu);
977 	case 0xb2:
978 		return handle_lpswe(vcpu);
979 	default:
980 		return -EOPNOTSUPP;
981 	}
982 }
983 
handle_epsw(struct kvm_vcpu * vcpu)984 static int handle_epsw(struct kvm_vcpu *vcpu)
985 {
986 	int reg1, reg2;
987 
988 	vcpu->stat.instruction_epsw++;
989 
990 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
991 
992 	/* This basically extracts the mask half of the psw. */
993 	vcpu->run->s.regs.gprs[reg1] &= 0xffffffff00000000UL;
994 	vcpu->run->s.regs.gprs[reg1] |= vcpu->arch.sie_block->gpsw.mask >> 32;
995 	if (reg2) {
996 		vcpu->run->s.regs.gprs[reg2] &= 0xffffffff00000000UL;
997 		vcpu->run->s.regs.gprs[reg2] |=
998 			vcpu->arch.sie_block->gpsw.mask & 0x00000000ffffffffUL;
999 	}
1000 	return 0;
1001 }
1002 
1003 #define PFMF_RESERVED   0xfffc0101UL
1004 #define PFMF_SK         0x00020000UL
1005 #define PFMF_CF         0x00010000UL
1006 #define PFMF_UI         0x00008000UL
1007 #define PFMF_FSC        0x00007000UL
1008 #define PFMF_NQ         0x00000800UL
1009 #define PFMF_MR         0x00000400UL
1010 #define PFMF_MC         0x00000200UL
1011 #define PFMF_KEY        0x000000feUL
1012 
handle_pfmf(struct kvm_vcpu * vcpu)1013 static int handle_pfmf(struct kvm_vcpu *vcpu)
1014 {
1015 	bool mr = false, mc = false, nq;
1016 	int reg1, reg2;
1017 	unsigned long start, end;
1018 	unsigned char key;
1019 
1020 	vcpu->stat.instruction_pfmf++;
1021 
1022 	kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
1023 
1024 	if (!test_kvm_facility(vcpu->kvm, 8))
1025 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1026 
1027 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1028 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1029 
1030 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_RESERVED)
1031 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1032 
1033 	/* Only provide non-quiescing support if enabled for the guest */
1034 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_NQ &&
1035 	    !test_kvm_facility(vcpu->kvm, 14))
1036 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1037 
1038 	/* Only provide conditional-SSKE support if enabled for the guest */
1039 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK &&
1040 	    test_kvm_facility(vcpu->kvm, 10)) {
1041 		mr = vcpu->run->s.regs.gprs[reg1] & PFMF_MR;
1042 		mc = vcpu->run->s.regs.gprs[reg1] & PFMF_MC;
1043 	}
1044 
1045 	nq = vcpu->run->s.regs.gprs[reg1] & PFMF_NQ;
1046 	key = vcpu->run->s.regs.gprs[reg1] & PFMF_KEY;
1047 	start = vcpu->run->s.regs.gprs[reg2] & PAGE_MASK;
1048 	start = kvm_s390_logical_to_effective(vcpu, start);
1049 
1050 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1051 		if (kvm_s390_check_low_addr_prot_real(vcpu, start))
1052 			return kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
1053 	}
1054 
1055 	switch (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1056 	case 0x00000000:
1057 		/* only 4k frames specify a real address */
1058 		start = kvm_s390_real_to_abs(vcpu, start);
1059 		end = (start + PAGE_SIZE) & ~(PAGE_SIZE - 1);
1060 		break;
1061 	case 0x00001000:
1062 		end = (start + _SEGMENT_SIZE) & ~(_SEGMENT_SIZE - 1);
1063 		break;
1064 	case 0x00002000:
1065 		/* only support 2G frame size if EDAT2 is available and we are
1066 		   not in 24-bit addressing mode */
1067 		if (!test_kvm_facility(vcpu->kvm, 78) ||
1068 		    psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_24BIT)
1069 			return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1070 		end = (start + _REGION3_SIZE) & ~(_REGION3_SIZE - 1);
1071 		break;
1072 	default:
1073 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1074 	}
1075 
1076 	while (start != end) {
1077 		unsigned long vmaddr;
1078 		bool unlocked = false;
1079 
1080 		/* Translate guest address to host address */
1081 		vmaddr = gfn_to_hva(vcpu->kvm, gpa_to_gfn(start));
1082 		if (kvm_is_error_hva(vmaddr))
1083 			return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1084 
1085 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_CF) {
1086 			if (kvm_clear_guest(vcpu->kvm, start, PAGE_SIZE))
1087 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1088 		}
1089 
1090 		if (vcpu->run->s.regs.gprs[reg1] & PFMF_SK) {
1091 			int rc = kvm_s390_skey_check_enable(vcpu);
1092 
1093 			if (rc)
1094 				return rc;
1095 			mmap_read_lock(current->mm);
1096 			rc = cond_set_guest_storage_key(current->mm, vmaddr,
1097 							key, NULL, nq, mr, mc);
1098 			if (rc < 0) {
1099 				rc = fixup_user_fault(current->mm, vmaddr,
1100 						      FAULT_FLAG_WRITE, &unlocked);
1101 				rc = !rc ? -EAGAIN : rc;
1102 			}
1103 			mmap_read_unlock(current->mm);
1104 			if (rc == -EFAULT)
1105 				return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1106 			if (rc == -EAGAIN)
1107 				continue;
1108 			if (rc < 0)
1109 				return rc;
1110 		}
1111 		start += PAGE_SIZE;
1112 	}
1113 	if (vcpu->run->s.regs.gprs[reg1] & PFMF_FSC) {
1114 		if (psw_bits(vcpu->arch.sie_block->gpsw).eaba == PSW_BITS_AMODE_64BIT) {
1115 			vcpu->run->s.regs.gprs[reg2] = end;
1116 		} else {
1117 			vcpu->run->s.regs.gprs[reg2] &= ~0xffffffffUL;
1118 			end = kvm_s390_logical_to_effective(vcpu, end);
1119 			vcpu->run->s.regs.gprs[reg2] |= end;
1120 		}
1121 	}
1122 	return 0;
1123 }
1124 
1125 /*
1126  * Must be called with relevant read locks held (kvm->mm->mmap_lock, kvm->srcu)
1127  */
__do_essa(struct kvm_vcpu * vcpu,const int orc)1128 static inline int __do_essa(struct kvm_vcpu *vcpu, const int orc)
1129 {
1130 	int r1, r2, nappended, entries;
1131 	unsigned long gfn, hva, res, pgstev, ptev;
1132 	unsigned long *cbrlo;
1133 
1134 	/*
1135 	 * We don't need to set SD.FPF.SK to 1 here, because if we have a
1136 	 * machine check here we either handle it or crash
1137 	 */
1138 
1139 	kvm_s390_get_regs_rre(vcpu, &r1, &r2);
1140 	gfn = vcpu->run->s.regs.gprs[r2] >> PAGE_SHIFT;
1141 	hva = gfn_to_hva(vcpu->kvm, gfn);
1142 	entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1143 
1144 	if (kvm_is_error_hva(hva))
1145 		return kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1146 
1147 	nappended = pgste_perform_essa(vcpu->kvm->mm, hva, orc, &ptev, &pgstev);
1148 	if (nappended < 0) {
1149 		res = orc ? 0x10 : 0;
1150 		vcpu->run->s.regs.gprs[r1] = res; /* Exception Indication */
1151 		return 0;
1152 	}
1153 	res = (pgstev & _PGSTE_GPS_USAGE_MASK) >> 22;
1154 	/*
1155 	 * Set the block-content state part of the result. 0 means resident, so
1156 	 * nothing to do if the page is valid. 2 is for preserved pages
1157 	 * (non-present and non-zero), and 3 for zero pages (non-present and
1158 	 * zero).
1159 	 */
1160 	if (ptev & _PAGE_INVALID) {
1161 		res |= 2;
1162 		if (pgstev & _PGSTE_GPS_ZERO)
1163 			res |= 1;
1164 	}
1165 	if (pgstev & _PGSTE_GPS_NODAT)
1166 		res |= 0x20;
1167 	vcpu->run->s.regs.gprs[r1] = res;
1168 	/*
1169 	 * It is possible that all the normal 511 slots were full, in which case
1170 	 * we will now write in the 512th slot, which is reserved for host use.
1171 	 * In both cases we let the normal essa handling code process all the
1172 	 * slots, including the reserved one, if needed.
1173 	 */
1174 	if (nappended > 0) {
1175 		cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo & PAGE_MASK);
1176 		cbrlo[entries] = gfn << PAGE_SHIFT;
1177 	}
1178 
1179 	if (orc) {
1180 		struct kvm_memory_slot *ms = gfn_to_memslot(vcpu->kvm, gfn);
1181 
1182 		/* Increment only if we are really flipping the bit */
1183 		if (ms && !test_and_set_bit(gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
1184 			atomic64_inc(&vcpu->kvm->arch.cmma_dirty_pages);
1185 	}
1186 
1187 	return nappended;
1188 }
1189 
handle_essa(struct kvm_vcpu * vcpu)1190 static int handle_essa(struct kvm_vcpu *vcpu)
1191 {
1192 	/* entries expected to be 1FF */
1193 	int entries = (vcpu->arch.sie_block->cbrlo & ~PAGE_MASK) >> 3;
1194 	unsigned long *cbrlo;
1195 	struct gmap *gmap;
1196 	int i, orc;
1197 
1198 	VCPU_EVENT(vcpu, 4, "ESSA: release %d pages", entries);
1199 	gmap = vcpu->arch.gmap;
1200 	vcpu->stat.instruction_essa++;
1201 	if (!vcpu->kvm->arch.use_cmma)
1202 		return kvm_s390_inject_program_int(vcpu, PGM_OPERATION);
1203 
1204 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1205 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1206 	/* Check for invalid operation request code */
1207 	orc = (vcpu->arch.sie_block->ipb & 0xf0000000) >> 28;
1208 	/* ORCs 0-6 are always valid */
1209 	if (orc > (test_kvm_facility(vcpu->kvm, 147) ? ESSA_SET_STABLE_NODAT
1210 						: ESSA_SET_STABLE_IF_RESIDENT))
1211 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1212 
1213 	if (!vcpu->kvm->arch.migration_mode) {
1214 		/*
1215 		 * CMMA is enabled in the KVM settings, but is disabled in
1216 		 * the SIE block and in the mm_context, and we are not doing
1217 		 * a migration. Enable CMMA in the mm_context.
1218 		 * Since we need to take a write lock to write to the context
1219 		 * to avoid races with storage keys handling, we check if the
1220 		 * value really needs to be written to; if the value is
1221 		 * already correct, we do nothing and avoid the lock.
1222 		 */
1223 		if (vcpu->kvm->mm->context.uses_cmm == 0) {
1224 			mmap_write_lock(vcpu->kvm->mm);
1225 			vcpu->kvm->mm->context.uses_cmm = 1;
1226 			mmap_write_unlock(vcpu->kvm->mm);
1227 		}
1228 		/*
1229 		 * If we are here, we are supposed to have CMMA enabled in
1230 		 * the SIE block. Enabling CMMA works on a per-CPU basis,
1231 		 * while the context use_cmma flag is per process.
1232 		 * It's possible that the context flag is enabled and the
1233 		 * SIE flag is not, so we set the flag always; if it was
1234 		 * already set, nothing changes, otherwise we enable it
1235 		 * on this CPU too.
1236 		 */
1237 		vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
1238 		/* Retry the ESSA instruction */
1239 		kvm_s390_retry_instr(vcpu);
1240 	} else {
1241 		int srcu_idx;
1242 
1243 		mmap_read_lock(vcpu->kvm->mm);
1244 		srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1245 		i = __do_essa(vcpu, orc);
1246 		srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
1247 		mmap_read_unlock(vcpu->kvm->mm);
1248 		if (i < 0)
1249 			return i;
1250 		/* Account for the possible extra cbrl entry */
1251 		entries += i;
1252 	}
1253 	vcpu->arch.sie_block->cbrlo &= PAGE_MASK;	/* reset nceo */
1254 	cbrlo = phys_to_virt(vcpu->arch.sie_block->cbrlo);
1255 	mmap_read_lock(gmap->mm);
1256 	for (i = 0; i < entries; ++i)
1257 		__gmap_zap(gmap, cbrlo[i]);
1258 	mmap_read_unlock(gmap->mm);
1259 	return 0;
1260 }
1261 
kvm_s390_handle_b9(struct kvm_vcpu * vcpu)1262 int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
1263 {
1264 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1265 	case 0x8a:
1266 	case 0x8e:
1267 	case 0x8f:
1268 		return handle_ipte_interlock(vcpu);
1269 	case 0x8d:
1270 		return handle_epsw(vcpu);
1271 	case 0xab:
1272 		return handle_essa(vcpu);
1273 	case 0xaf:
1274 		return handle_pfmf(vcpu);
1275 	default:
1276 		return -EOPNOTSUPP;
1277 	}
1278 }
1279 
kvm_s390_handle_lctl(struct kvm_vcpu * vcpu)1280 int kvm_s390_handle_lctl(struct kvm_vcpu *vcpu)
1281 {
1282 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1283 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1284 	int reg, rc, nr_regs;
1285 	u32 ctl_array[16];
1286 	u64 ga;
1287 	u8 ar;
1288 
1289 	vcpu->stat.instruction_lctl++;
1290 
1291 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1292 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1293 
1294 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1295 
1296 	if (ga & 3)
1297 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1298 
1299 	VCPU_EVENT(vcpu, 4, "LCTL: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1300 	trace_kvm_s390_handle_lctl(vcpu, 0, reg1, reg3, ga);
1301 
1302 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1303 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1304 	if (rc)
1305 		return kvm_s390_inject_prog_cond(vcpu, rc);
1306 	reg = reg1;
1307 	nr_regs = 0;
1308 	do {
1309 		vcpu->arch.sie_block->gcr[reg] &= 0xffffffff00000000ul;
1310 		vcpu->arch.sie_block->gcr[reg] |= ctl_array[nr_regs++];
1311 		if (reg == reg3)
1312 			break;
1313 		reg = (reg + 1) % 16;
1314 	} while (1);
1315 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1316 	return 0;
1317 }
1318 
kvm_s390_handle_stctl(struct kvm_vcpu * vcpu)1319 int kvm_s390_handle_stctl(struct kvm_vcpu *vcpu)
1320 {
1321 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1322 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1323 	int reg, rc, nr_regs;
1324 	u32 ctl_array[16];
1325 	u64 ga;
1326 	u8 ar;
1327 
1328 	vcpu->stat.instruction_stctl++;
1329 
1330 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1331 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1332 
1333 	ga = kvm_s390_get_base_disp_rs(vcpu, &ar);
1334 
1335 	if (ga & 3)
1336 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1337 
1338 	VCPU_EVENT(vcpu, 4, "STCTL r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1339 	trace_kvm_s390_handle_stctl(vcpu, 0, reg1, reg3, ga);
1340 
1341 	reg = reg1;
1342 	nr_regs = 0;
1343 	do {
1344 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1345 		if (reg == reg3)
1346 			break;
1347 		reg = (reg + 1) % 16;
1348 	} while (1);
1349 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u32));
1350 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1351 }
1352 
handle_lctlg(struct kvm_vcpu * vcpu)1353 static int handle_lctlg(struct kvm_vcpu *vcpu)
1354 {
1355 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1356 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1357 	int reg, rc, nr_regs;
1358 	u64 ctl_array[16];
1359 	u64 ga;
1360 	u8 ar;
1361 
1362 	vcpu->stat.instruction_lctlg++;
1363 
1364 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1365 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1366 
1367 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1368 
1369 	if (ga & 7)
1370 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1371 
1372 	VCPU_EVENT(vcpu, 4, "LCTLG: r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1373 	trace_kvm_s390_handle_lctl(vcpu, 1, reg1, reg3, ga);
1374 
1375 	nr_regs = ((reg3 - reg1) & 0xf) + 1;
1376 	rc = read_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1377 	if (rc)
1378 		return kvm_s390_inject_prog_cond(vcpu, rc);
1379 	reg = reg1;
1380 	nr_regs = 0;
1381 	do {
1382 		vcpu->arch.sie_block->gcr[reg] = ctl_array[nr_regs++];
1383 		if (reg == reg3)
1384 			break;
1385 		reg = (reg + 1) % 16;
1386 	} while (1);
1387 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
1388 	return 0;
1389 }
1390 
handle_stctg(struct kvm_vcpu * vcpu)1391 static int handle_stctg(struct kvm_vcpu *vcpu)
1392 {
1393 	int reg1 = (vcpu->arch.sie_block->ipa & 0x00f0) >> 4;
1394 	int reg3 = vcpu->arch.sie_block->ipa & 0x000f;
1395 	int reg, rc, nr_regs;
1396 	u64 ctl_array[16];
1397 	u64 ga;
1398 	u8 ar;
1399 
1400 	vcpu->stat.instruction_stctg++;
1401 
1402 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1403 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1404 
1405 	ga = kvm_s390_get_base_disp_rsy(vcpu, &ar);
1406 
1407 	if (ga & 7)
1408 		return kvm_s390_inject_program_int(vcpu, PGM_SPECIFICATION);
1409 
1410 	VCPU_EVENT(vcpu, 4, "STCTG r1:%d, r3:%d, addr: 0x%llx", reg1, reg3, ga);
1411 	trace_kvm_s390_handle_stctl(vcpu, 1, reg1, reg3, ga);
1412 
1413 	reg = reg1;
1414 	nr_regs = 0;
1415 	do {
1416 		ctl_array[nr_regs++] = vcpu->arch.sie_block->gcr[reg];
1417 		if (reg == reg3)
1418 			break;
1419 		reg = (reg + 1) % 16;
1420 	} while (1);
1421 	rc = write_guest(vcpu, ga, ar, ctl_array, nr_regs * sizeof(u64));
1422 	return rc ? kvm_s390_inject_prog_cond(vcpu, rc) : 0;
1423 }
1424 
kvm_s390_handle_eb(struct kvm_vcpu * vcpu)1425 int kvm_s390_handle_eb(struct kvm_vcpu *vcpu)
1426 {
1427 	switch (vcpu->arch.sie_block->ipb & 0x000000ff) {
1428 	case 0x25:
1429 		return handle_stctg(vcpu);
1430 	case 0x2f:
1431 		return handle_lctlg(vcpu);
1432 	case 0x60:
1433 	case 0x61:
1434 	case 0x62:
1435 		return handle_ri(vcpu);
1436 	default:
1437 		return -EOPNOTSUPP;
1438 	}
1439 }
1440 
handle_tprot(struct kvm_vcpu * vcpu)1441 static int handle_tprot(struct kvm_vcpu *vcpu)
1442 {
1443 	u64 address1, address2;
1444 	unsigned long hva, gpa;
1445 	int ret = 0, cc = 0;
1446 	bool writable;
1447 	u8 ar;
1448 
1449 	vcpu->stat.instruction_tprot++;
1450 
1451 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1452 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1453 
1454 	kvm_s390_get_base_disp_sse(vcpu, &address1, &address2, &ar, NULL);
1455 
1456 	/* we only handle the Linux memory detection case:
1457 	 * access key == 0
1458 	 * everything else goes to userspace. */
1459 	if (address2 & 0xf0)
1460 		return -EOPNOTSUPP;
1461 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1462 		ipte_lock(vcpu);
1463 	ret = guest_translate_address(vcpu, address1, ar, &gpa, GACC_STORE);
1464 	if (ret == PGM_PROTECTION) {
1465 		/* Write protected? Try again with read-only... */
1466 		cc = 1;
1467 		ret = guest_translate_address(vcpu, address1, ar, &gpa,
1468 					      GACC_FETCH);
1469 	}
1470 	if (ret) {
1471 		if (ret == PGM_ADDRESSING || ret == PGM_TRANSLATION_SPEC) {
1472 			ret = kvm_s390_inject_program_int(vcpu, ret);
1473 		} else if (ret > 0) {
1474 			/* Translation not available */
1475 			kvm_s390_set_psw_cc(vcpu, 3);
1476 			ret = 0;
1477 		}
1478 		goto out_unlock;
1479 	}
1480 
1481 	hva = gfn_to_hva_prot(vcpu->kvm, gpa_to_gfn(gpa), &writable);
1482 	if (kvm_is_error_hva(hva)) {
1483 		ret = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING);
1484 	} else {
1485 		if (!writable)
1486 			cc = 1;		/* Write not permitted ==> read-only */
1487 		kvm_s390_set_psw_cc(vcpu, cc);
1488 		/* Note: CC2 only occurs for storage keys (not supported yet) */
1489 	}
1490 out_unlock:
1491 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_DAT)
1492 		ipte_unlock(vcpu);
1493 	return ret;
1494 }
1495 
kvm_s390_handle_e5(struct kvm_vcpu * vcpu)1496 int kvm_s390_handle_e5(struct kvm_vcpu *vcpu)
1497 {
1498 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1499 	case 0x01:
1500 		return handle_tprot(vcpu);
1501 	default:
1502 		return -EOPNOTSUPP;
1503 	}
1504 }
1505 
handle_sckpf(struct kvm_vcpu * vcpu)1506 static int handle_sckpf(struct kvm_vcpu *vcpu)
1507 {
1508 	u32 value;
1509 
1510 	vcpu->stat.instruction_sckpf++;
1511 
1512 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
1513 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
1514 
1515 	if (vcpu->run->s.regs.gprs[0] & 0x00000000ffff0000)
1516 		return kvm_s390_inject_program_int(vcpu,
1517 						   PGM_SPECIFICATION);
1518 
1519 	value = vcpu->run->s.regs.gprs[0] & 0x000000000000ffff;
1520 	vcpu->arch.sie_block->todpr = value;
1521 
1522 	return 0;
1523 }
1524 
handle_ptff(struct kvm_vcpu * vcpu)1525 static int handle_ptff(struct kvm_vcpu *vcpu)
1526 {
1527 	vcpu->stat.instruction_ptff++;
1528 
1529 	/* we don't emulate any control instructions yet */
1530 	kvm_s390_set_psw_cc(vcpu, 3);
1531 	return 0;
1532 }
1533 
kvm_s390_handle_01(struct kvm_vcpu * vcpu)1534 int kvm_s390_handle_01(struct kvm_vcpu *vcpu)
1535 {
1536 	switch (vcpu->arch.sie_block->ipa & 0x00ff) {
1537 	case 0x04:
1538 		return handle_ptff(vcpu);
1539 	case 0x07:
1540 		return handle_sckpf(vcpu);
1541 	default:
1542 		return -EOPNOTSUPP;
1543 	}
1544 }
1545