xref: /linux/arch/s390/kvm/interrupt.c (revision fd639726bf15fca8ee1a00dce8e0096d0ad9bd18)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * handling kvm guest interrupts
4  *
5  * Copyright IBM Corp. 2008, 2015
6  *
7  *    Author(s): Carsten Otte <cotte@de.ibm.com>
8  */
9 
10 #include <linux/interrupt.h>
11 #include <linux/kvm_host.h>
12 #include <linux/hrtimer.h>
13 #include <linux/mmu_context.h>
14 #include <linux/signal.h>
15 #include <linux/slab.h>
16 #include <linux/bitmap.h>
17 #include <linux/vmalloc.h>
18 #include <asm/asm-offsets.h>
19 #include <asm/dis.h>
20 #include <linux/uaccess.h>
21 #include <asm/sclp.h>
22 #include <asm/isc.h>
23 #include <asm/gmap.h>
24 #include <asm/switch_to.h>
25 #include <asm/nmi.h>
26 #include "kvm-s390.h"
27 #include "gaccess.h"
28 #include "trace-s390.h"
29 
30 #define PFAULT_INIT 0x0600
31 #define PFAULT_DONE 0x0680
32 #define VIRTIO_PARAM 0x0d00
33 
34 /* handle external calls via sigp interpretation facility */
35 static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
36 {
37 	int c, scn;
38 
39 	if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
40 		return 0;
41 
42 	BUG_ON(!kvm_s390_use_sca_entries());
43 	read_lock(&vcpu->kvm->arch.sca_lock);
44 	if (vcpu->kvm->arch.use_esca) {
45 		struct esca_block *sca = vcpu->kvm->arch.sca;
46 		union esca_sigp_ctrl sigp_ctrl =
47 			sca->cpu[vcpu->vcpu_id].sigp_ctrl;
48 
49 		c = sigp_ctrl.c;
50 		scn = sigp_ctrl.scn;
51 	} else {
52 		struct bsca_block *sca = vcpu->kvm->arch.sca;
53 		union bsca_sigp_ctrl sigp_ctrl =
54 			sca->cpu[vcpu->vcpu_id].sigp_ctrl;
55 
56 		c = sigp_ctrl.c;
57 		scn = sigp_ctrl.scn;
58 	}
59 	read_unlock(&vcpu->kvm->arch.sca_lock);
60 
61 	if (src_id)
62 		*src_id = scn;
63 
64 	return c;
65 }
66 
67 static int sca_inject_ext_call(struct kvm_vcpu *vcpu, int src_id)
68 {
69 	int expect, rc;
70 
71 	BUG_ON(!kvm_s390_use_sca_entries());
72 	read_lock(&vcpu->kvm->arch.sca_lock);
73 	if (vcpu->kvm->arch.use_esca) {
74 		struct esca_block *sca = vcpu->kvm->arch.sca;
75 		union esca_sigp_ctrl *sigp_ctrl =
76 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
77 		union esca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
78 
79 		new_val.scn = src_id;
80 		new_val.c = 1;
81 		old_val.c = 0;
82 
83 		expect = old_val.value;
84 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
85 	} else {
86 		struct bsca_block *sca = vcpu->kvm->arch.sca;
87 		union bsca_sigp_ctrl *sigp_ctrl =
88 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
89 		union bsca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
90 
91 		new_val.scn = src_id;
92 		new_val.c = 1;
93 		old_val.c = 0;
94 
95 		expect = old_val.value;
96 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
97 	}
98 	read_unlock(&vcpu->kvm->arch.sca_lock);
99 
100 	if (rc != expect) {
101 		/* another external call is pending */
102 		return -EBUSY;
103 	}
104 	atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
105 	return 0;
106 }
107 
108 static void sca_clear_ext_call(struct kvm_vcpu *vcpu)
109 {
110 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
111 	int rc, expect;
112 
113 	if (!kvm_s390_use_sca_entries())
114 		return;
115 	atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
116 	read_lock(&vcpu->kvm->arch.sca_lock);
117 	if (vcpu->kvm->arch.use_esca) {
118 		struct esca_block *sca = vcpu->kvm->arch.sca;
119 		union esca_sigp_ctrl *sigp_ctrl =
120 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
121 		union esca_sigp_ctrl old = *sigp_ctrl;
122 
123 		expect = old.value;
124 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
125 	} else {
126 		struct bsca_block *sca = vcpu->kvm->arch.sca;
127 		union bsca_sigp_ctrl *sigp_ctrl =
128 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
129 		union bsca_sigp_ctrl old = *sigp_ctrl;
130 
131 		expect = old.value;
132 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
133 	}
134 	read_unlock(&vcpu->kvm->arch.sca_lock);
135 	WARN_ON(rc != expect); /* cannot clear? */
136 }
137 
138 int psw_extint_disabled(struct kvm_vcpu *vcpu)
139 {
140 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
141 }
142 
143 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
144 {
145 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
146 }
147 
148 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
149 {
150 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
151 }
152 
153 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
154 {
155 	return psw_extint_disabled(vcpu) &&
156 	       psw_ioint_disabled(vcpu) &&
157 	       psw_mchk_disabled(vcpu);
158 }
159 
160 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
161 {
162 	if (psw_extint_disabled(vcpu) ||
163 	    !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
164 		return 0;
165 	if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
166 		/* No timer interrupts when single stepping */
167 		return 0;
168 	return 1;
169 }
170 
171 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
172 {
173 	if (vcpu->arch.sie_block->ckc >= kvm_s390_get_tod_clock_fast(vcpu->kvm))
174 		return 0;
175 	return ckc_interrupts_enabled(vcpu);
176 }
177 
178 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
179 {
180 	return !psw_extint_disabled(vcpu) &&
181 	       (vcpu->arch.sie_block->gcr[0] & 0x400ul);
182 }
183 
184 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
185 {
186 	if (!cpu_timer_interrupts_enabled(vcpu))
187 		return 0;
188 	return kvm_s390_get_cpu_timer(vcpu) >> 63;
189 }
190 
191 static inline int is_ioirq(unsigned long irq_type)
192 {
193 	return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
194 		(irq_type <= IRQ_PEND_IO_ISC_7));
195 }
196 
197 static uint64_t isc_to_isc_bits(int isc)
198 {
199 	return (0x80 >> isc) << 24;
200 }
201 
202 static inline u8 int_word_to_isc(u32 int_word)
203 {
204 	return (int_word & 0x38000000) >> 27;
205 }
206 
207 static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
208 {
209 	return vcpu->kvm->arch.float_int.pending_irqs |
210 	       vcpu->arch.local_int.pending_irqs;
211 }
212 
213 static inline int isc_to_irq_type(unsigned long isc)
214 {
215 	return IRQ_PEND_IO_ISC_0 + isc;
216 }
217 
218 static inline int irq_type_to_isc(unsigned long irq_type)
219 {
220 	return irq_type - IRQ_PEND_IO_ISC_0;
221 }
222 
223 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
224 				   unsigned long active_mask)
225 {
226 	int i;
227 
228 	for (i = 0; i <= MAX_ISC; i++)
229 		if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
230 			active_mask &= ~(1UL << (isc_to_irq_type(i)));
231 
232 	return active_mask;
233 }
234 
235 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
236 {
237 	unsigned long active_mask;
238 
239 	active_mask = pending_irqs(vcpu);
240 	if (!active_mask)
241 		return 0;
242 
243 	if (psw_extint_disabled(vcpu))
244 		active_mask &= ~IRQ_PEND_EXT_MASK;
245 	if (psw_ioint_disabled(vcpu))
246 		active_mask &= ~IRQ_PEND_IO_MASK;
247 	else
248 		active_mask = disable_iscs(vcpu, active_mask);
249 	if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
250 		__clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
251 	if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
252 		__clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
253 	if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
254 		__clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
255 	if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
256 		__clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
257 	if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
258 		__clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
259 	if (psw_mchk_disabled(vcpu))
260 		active_mask &= ~IRQ_PEND_MCHK_MASK;
261 	/*
262 	 * Check both floating and local interrupt's cr14 because
263 	 * bit IRQ_PEND_MCHK_REP could be set in both cases.
264 	 */
265 	if (!(vcpu->arch.sie_block->gcr[14] &
266 	   (vcpu->kvm->arch.float_int.mchk.cr14 |
267 	   vcpu->arch.local_int.irq.mchk.cr14)))
268 		__clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
269 
270 	/*
271 	 * STOP irqs will never be actively delivered. They are triggered via
272 	 * intercept requests and cleared when the stop intercept is performed.
273 	 */
274 	__clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
275 
276 	return active_mask;
277 }
278 
279 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
280 {
281 	atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
282 	set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
283 }
284 
285 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
286 {
287 	atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
288 	clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
289 }
290 
291 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
292 {
293 	atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
294 		    &vcpu->arch.sie_block->cpuflags);
295 	vcpu->arch.sie_block->lctl = 0x0000;
296 	vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
297 
298 	if (guestdbg_enabled(vcpu)) {
299 		vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
300 					       LCTL_CR10 | LCTL_CR11);
301 		vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
302 	}
303 }
304 
305 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
306 {
307 	atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
308 }
309 
310 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
311 {
312 	if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
313 		return;
314 	else if (psw_ioint_disabled(vcpu))
315 		__set_cpuflag(vcpu, CPUSTAT_IO_INT);
316 	else
317 		vcpu->arch.sie_block->lctl |= LCTL_CR6;
318 }
319 
320 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
321 {
322 	if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
323 		return;
324 	if (psw_extint_disabled(vcpu))
325 		__set_cpuflag(vcpu, CPUSTAT_EXT_INT);
326 	else
327 		vcpu->arch.sie_block->lctl |= LCTL_CR0;
328 }
329 
330 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
331 {
332 	if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
333 		return;
334 	if (psw_mchk_disabled(vcpu))
335 		vcpu->arch.sie_block->ictl |= ICTL_LPSW;
336 	else
337 		vcpu->arch.sie_block->lctl |= LCTL_CR14;
338 }
339 
340 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
341 {
342 	if (kvm_s390_is_stop_irq_pending(vcpu))
343 		__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
344 }
345 
346 /* Set interception request for non-deliverable interrupts */
347 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
348 {
349 	set_intercept_indicators_io(vcpu);
350 	set_intercept_indicators_ext(vcpu);
351 	set_intercept_indicators_mchk(vcpu);
352 	set_intercept_indicators_stop(vcpu);
353 }
354 
355 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
356 {
357 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
358 	int rc;
359 
360 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
361 					 0, 0);
362 
363 	rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
364 			   (u16 *)__LC_EXT_INT_CODE);
365 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
366 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
367 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
368 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
369 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
370 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
371 	return rc ? -EFAULT : 0;
372 }
373 
374 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
375 {
376 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
377 	int rc;
378 
379 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
380 					 0, 0);
381 
382 	rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
383 			   (u16 __user *)__LC_EXT_INT_CODE);
384 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
385 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
386 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
387 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
388 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
389 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
390 	return rc ? -EFAULT : 0;
391 }
392 
393 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
394 {
395 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
396 	struct kvm_s390_ext_info ext;
397 	int rc;
398 
399 	spin_lock(&li->lock);
400 	ext = li->irq.ext;
401 	clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
402 	li->irq.ext.ext_params2 = 0;
403 	spin_unlock(&li->lock);
404 
405 	VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
406 		   ext.ext_params2);
407 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
408 					 KVM_S390_INT_PFAULT_INIT,
409 					 0, ext.ext_params2);
410 
411 	rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
412 	rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
413 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
414 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
415 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
416 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
417 	rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
418 	return rc ? -EFAULT : 0;
419 }
420 
421 static int __write_machine_check(struct kvm_vcpu *vcpu,
422 				 struct kvm_s390_mchk_info *mchk)
423 {
424 	unsigned long ext_sa_addr;
425 	unsigned long lc;
426 	freg_t fprs[NUM_FPRS];
427 	union mci mci;
428 	int rc;
429 
430 	mci.val = mchk->mcic;
431 	/* take care of lazy register loading */
432 	save_fpu_regs();
433 	save_access_regs(vcpu->run->s.regs.acrs);
434 	if (MACHINE_HAS_GS && vcpu->arch.gs_enabled)
435 		save_gs_cb(current->thread.gs_cb);
436 
437 	/* Extended save area */
438 	rc = read_guest_lc(vcpu, __LC_MCESAD, &ext_sa_addr,
439 			   sizeof(unsigned long));
440 	/* Only bits 0 through 63-LC are used for address formation */
441 	lc = ext_sa_addr & MCESA_LC_MASK;
442 	if (test_kvm_facility(vcpu->kvm, 133)) {
443 		switch (lc) {
444 		case 0:
445 		case 10:
446 			ext_sa_addr &= ~0x3ffUL;
447 			break;
448 		case 11:
449 			ext_sa_addr &= ~0x7ffUL;
450 			break;
451 		case 12:
452 			ext_sa_addr &= ~0xfffUL;
453 			break;
454 		default:
455 			ext_sa_addr = 0;
456 			break;
457 		}
458 	} else {
459 		ext_sa_addr &= ~0x3ffUL;
460 	}
461 
462 	if (!rc && mci.vr && ext_sa_addr && test_kvm_facility(vcpu->kvm, 129)) {
463 		if (write_guest_abs(vcpu, ext_sa_addr, vcpu->run->s.regs.vrs,
464 				    512))
465 			mci.vr = 0;
466 	} else {
467 		mci.vr = 0;
468 	}
469 	if (!rc && mci.gs && ext_sa_addr && test_kvm_facility(vcpu->kvm, 133)
470 	    && (lc == 11 || lc == 12)) {
471 		if (write_guest_abs(vcpu, ext_sa_addr + 1024,
472 				    &vcpu->run->s.regs.gscb, 32))
473 			mci.gs = 0;
474 	} else {
475 		mci.gs = 0;
476 	}
477 
478 	/* General interruption information */
479 	rc |= put_guest_lc(vcpu, 1, (u8 __user *) __LC_AR_MODE_ID);
480 	rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
481 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
482 	rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
483 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
484 	rc |= put_guest_lc(vcpu, mci.val, (u64 __user *) __LC_MCCK_CODE);
485 
486 	/* Register-save areas */
487 	if (MACHINE_HAS_VX) {
488 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
489 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA, fprs, 128);
490 	} else {
491 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA,
492 				     vcpu->run->s.regs.fprs, 128);
493 	}
494 	rc |= write_guest_lc(vcpu, __LC_GPREGS_SAVE_AREA,
495 			     vcpu->run->s.regs.gprs, 128);
496 	rc |= put_guest_lc(vcpu, current->thread.fpu.fpc,
497 			   (u32 __user *) __LC_FP_CREG_SAVE_AREA);
498 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->todpr,
499 			   (u32 __user *) __LC_TOD_PROGREG_SAVE_AREA);
500 	rc |= put_guest_lc(vcpu, kvm_s390_get_cpu_timer(vcpu),
501 			   (u64 __user *) __LC_CPU_TIMER_SAVE_AREA);
502 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->ckc >> 8,
503 			   (u64 __user *) __LC_CLOCK_COMP_SAVE_AREA);
504 	rc |= write_guest_lc(vcpu, __LC_AREGS_SAVE_AREA,
505 			     &vcpu->run->s.regs.acrs, 64);
506 	rc |= write_guest_lc(vcpu, __LC_CREGS_SAVE_AREA,
507 			     &vcpu->arch.sie_block->gcr, 128);
508 
509 	/* Extended interruption information */
510 	rc |= put_guest_lc(vcpu, mchk->ext_damage_code,
511 			   (u32 __user *) __LC_EXT_DAMAGE_CODE);
512 	rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
513 			   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
514 	rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA, &mchk->fixed_logout,
515 			     sizeof(mchk->fixed_logout));
516 	return rc ? -EFAULT : 0;
517 }
518 
519 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
520 {
521 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
522 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
523 	struct kvm_s390_mchk_info mchk = {};
524 	int deliver = 0;
525 	int rc = 0;
526 
527 	spin_lock(&fi->lock);
528 	spin_lock(&li->lock);
529 	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
530 	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
531 		/*
532 		 * If there was an exigent machine check pending, then any
533 		 * repressible machine checks that might have been pending
534 		 * are indicated along with it, so always clear bits for
535 		 * repressible and exigent interrupts
536 		 */
537 		mchk = li->irq.mchk;
538 		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
539 		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
540 		memset(&li->irq.mchk, 0, sizeof(mchk));
541 		deliver = 1;
542 	}
543 	/*
544 	 * We indicate floating repressible conditions along with
545 	 * other pending conditions. Channel Report Pending and Channel
546 	 * Subsystem damage are the only two and and are indicated by
547 	 * bits in mcic and masked in cr14.
548 	 */
549 	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
550 		mchk.mcic |= fi->mchk.mcic;
551 		mchk.cr14 |= fi->mchk.cr14;
552 		memset(&fi->mchk, 0, sizeof(mchk));
553 		deliver = 1;
554 	}
555 	spin_unlock(&li->lock);
556 	spin_unlock(&fi->lock);
557 
558 	if (deliver) {
559 		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
560 			   mchk.mcic);
561 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
562 						 KVM_S390_MCHK,
563 						 mchk.cr14, mchk.mcic);
564 		rc = __write_machine_check(vcpu, &mchk);
565 	}
566 	return rc;
567 }
568 
569 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
570 {
571 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
572 	int rc;
573 
574 	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
575 	vcpu->stat.deliver_restart_signal++;
576 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
577 
578 	rc  = write_guest_lc(vcpu,
579 			     offsetof(struct lowcore, restart_old_psw),
580 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
581 	rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
582 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
583 	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
584 	return rc ? -EFAULT : 0;
585 }
586 
587 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
588 {
589 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
590 	struct kvm_s390_prefix_info prefix;
591 
592 	spin_lock(&li->lock);
593 	prefix = li->irq.prefix;
594 	li->irq.prefix.address = 0;
595 	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
596 	spin_unlock(&li->lock);
597 
598 	vcpu->stat.deliver_prefix_signal++;
599 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
600 					 KVM_S390_SIGP_SET_PREFIX,
601 					 prefix.address, 0);
602 
603 	kvm_s390_set_prefix(vcpu, prefix.address);
604 	return 0;
605 }
606 
607 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
608 {
609 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
610 	int rc;
611 	int cpu_addr;
612 
613 	spin_lock(&li->lock);
614 	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
615 	clear_bit(cpu_addr, li->sigp_emerg_pending);
616 	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
617 		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
618 	spin_unlock(&li->lock);
619 
620 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
621 	vcpu->stat.deliver_emergency_signal++;
622 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
623 					 cpu_addr, 0);
624 
625 	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
626 			   (u16 *)__LC_EXT_INT_CODE);
627 	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
628 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
629 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
630 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
631 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
632 	return rc ? -EFAULT : 0;
633 }
634 
635 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
636 {
637 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
638 	struct kvm_s390_extcall_info extcall;
639 	int rc;
640 
641 	spin_lock(&li->lock);
642 	extcall = li->irq.extcall;
643 	li->irq.extcall.code = 0;
644 	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
645 	spin_unlock(&li->lock);
646 
647 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
648 	vcpu->stat.deliver_external_call++;
649 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
650 					 KVM_S390_INT_EXTERNAL_CALL,
651 					 extcall.code, 0);
652 
653 	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
654 			   (u16 *)__LC_EXT_INT_CODE);
655 	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
656 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
657 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
658 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
659 			    sizeof(psw_t));
660 	return rc ? -EFAULT : 0;
661 }
662 
663 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
664 {
665 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
666 	struct kvm_s390_pgm_info pgm_info;
667 	int rc = 0, nullifying = false;
668 	u16 ilen;
669 
670 	spin_lock(&li->lock);
671 	pgm_info = li->irq.pgm;
672 	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
673 	memset(&li->irq.pgm, 0, sizeof(pgm_info));
674 	spin_unlock(&li->lock);
675 
676 	ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
677 	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
678 		   pgm_info.code, ilen);
679 	vcpu->stat.deliver_program_int++;
680 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
681 					 pgm_info.code, 0);
682 
683 	switch (pgm_info.code & ~PGM_PER) {
684 	case PGM_AFX_TRANSLATION:
685 	case PGM_ASX_TRANSLATION:
686 	case PGM_EX_TRANSLATION:
687 	case PGM_LFX_TRANSLATION:
688 	case PGM_LSTE_SEQUENCE:
689 	case PGM_LSX_TRANSLATION:
690 	case PGM_LX_TRANSLATION:
691 	case PGM_PRIMARY_AUTHORITY:
692 	case PGM_SECONDARY_AUTHORITY:
693 		nullifying = true;
694 		/* fall through */
695 	case PGM_SPACE_SWITCH:
696 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
697 				  (u64 *)__LC_TRANS_EXC_CODE);
698 		break;
699 	case PGM_ALEN_TRANSLATION:
700 	case PGM_ALE_SEQUENCE:
701 	case PGM_ASTE_INSTANCE:
702 	case PGM_ASTE_SEQUENCE:
703 	case PGM_ASTE_VALIDITY:
704 	case PGM_EXTENDED_AUTHORITY:
705 		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
706 				  (u8 *)__LC_EXC_ACCESS_ID);
707 		nullifying = true;
708 		break;
709 	case PGM_ASCE_TYPE:
710 	case PGM_PAGE_TRANSLATION:
711 	case PGM_REGION_FIRST_TRANS:
712 	case PGM_REGION_SECOND_TRANS:
713 	case PGM_REGION_THIRD_TRANS:
714 	case PGM_SEGMENT_TRANSLATION:
715 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
716 				  (u64 *)__LC_TRANS_EXC_CODE);
717 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
718 				   (u8 *)__LC_EXC_ACCESS_ID);
719 		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
720 				   (u8 *)__LC_OP_ACCESS_ID);
721 		nullifying = true;
722 		break;
723 	case PGM_MONITOR:
724 		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
725 				  (u16 *)__LC_MON_CLASS_NR);
726 		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
727 				   (u64 *)__LC_MON_CODE);
728 		break;
729 	case PGM_VECTOR_PROCESSING:
730 	case PGM_DATA:
731 		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
732 				  (u32 *)__LC_DATA_EXC_CODE);
733 		break;
734 	case PGM_PROTECTION:
735 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
736 				  (u64 *)__LC_TRANS_EXC_CODE);
737 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
738 				   (u8 *)__LC_EXC_ACCESS_ID);
739 		break;
740 	case PGM_STACK_FULL:
741 	case PGM_STACK_EMPTY:
742 	case PGM_STACK_SPECIFICATION:
743 	case PGM_STACK_TYPE:
744 	case PGM_STACK_OPERATION:
745 	case PGM_TRACE_TABEL:
746 	case PGM_CRYPTO_OPERATION:
747 		nullifying = true;
748 		break;
749 	}
750 
751 	if (pgm_info.code & PGM_PER) {
752 		rc |= put_guest_lc(vcpu, pgm_info.per_code,
753 				   (u8 *) __LC_PER_CODE);
754 		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
755 				   (u8 *)__LC_PER_ATMID);
756 		rc |= put_guest_lc(vcpu, pgm_info.per_address,
757 				   (u64 *) __LC_PER_ADDRESS);
758 		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
759 				   (u8 *) __LC_PER_ACCESS_ID);
760 	}
761 
762 	if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
763 		kvm_s390_rewind_psw(vcpu, ilen);
764 
765 	/* bit 1+2 of the target are the ilc, so we can directly use ilen */
766 	rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
767 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
768 				 (u64 *) __LC_LAST_BREAK);
769 	rc |= put_guest_lc(vcpu, pgm_info.code,
770 			   (u16 *)__LC_PGM_INT_CODE);
771 	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
772 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
773 	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
774 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
775 	return rc ? -EFAULT : 0;
776 }
777 
778 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
779 {
780 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
781 	struct kvm_s390_ext_info ext;
782 	int rc = 0;
783 
784 	spin_lock(&fi->lock);
785 	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
786 		spin_unlock(&fi->lock);
787 		return 0;
788 	}
789 	ext = fi->srv_signal;
790 	memset(&fi->srv_signal, 0, sizeof(ext));
791 	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
792 	spin_unlock(&fi->lock);
793 
794 	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
795 		   ext.ext_params);
796 	vcpu->stat.deliver_service_signal++;
797 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
798 					 ext.ext_params, 0);
799 
800 	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
801 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
802 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
803 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
804 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
805 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
806 	rc |= put_guest_lc(vcpu, ext.ext_params,
807 			   (u32 *)__LC_EXT_PARAMS);
808 
809 	return rc ? -EFAULT : 0;
810 }
811 
812 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
813 {
814 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
815 	struct kvm_s390_interrupt_info *inti;
816 	int rc = 0;
817 
818 	spin_lock(&fi->lock);
819 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
820 					struct kvm_s390_interrupt_info,
821 					list);
822 	if (inti) {
823 		list_del(&inti->list);
824 		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
825 	}
826 	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
827 		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
828 	spin_unlock(&fi->lock);
829 
830 	if (inti) {
831 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
832 						 KVM_S390_INT_PFAULT_DONE, 0,
833 						 inti->ext.ext_params2);
834 		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
835 			   inti->ext.ext_params2);
836 
837 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
838 				(u16 *)__LC_EXT_INT_CODE);
839 		rc |= put_guest_lc(vcpu, PFAULT_DONE,
840 				(u16 *)__LC_EXT_CPU_ADDR);
841 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
842 				&vcpu->arch.sie_block->gpsw,
843 				sizeof(psw_t));
844 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
845 				&vcpu->arch.sie_block->gpsw,
846 				sizeof(psw_t));
847 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
848 				(u64 *)__LC_EXT_PARAMS2);
849 		kfree(inti);
850 	}
851 	return rc ? -EFAULT : 0;
852 }
853 
854 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
855 {
856 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
857 	struct kvm_s390_interrupt_info *inti;
858 	int rc = 0;
859 
860 	spin_lock(&fi->lock);
861 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
862 					struct kvm_s390_interrupt_info,
863 					list);
864 	if (inti) {
865 		VCPU_EVENT(vcpu, 4,
866 			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
867 			   inti->ext.ext_params, inti->ext.ext_params2);
868 		vcpu->stat.deliver_virtio_interrupt++;
869 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
870 				inti->type,
871 				inti->ext.ext_params,
872 				inti->ext.ext_params2);
873 		list_del(&inti->list);
874 		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
875 	}
876 	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
877 		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
878 	spin_unlock(&fi->lock);
879 
880 	if (inti) {
881 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
882 				(u16 *)__LC_EXT_INT_CODE);
883 		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
884 				(u16 *)__LC_EXT_CPU_ADDR);
885 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
886 				&vcpu->arch.sie_block->gpsw,
887 				sizeof(psw_t));
888 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
889 				&vcpu->arch.sie_block->gpsw,
890 				sizeof(psw_t));
891 		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
892 				(u32 *)__LC_EXT_PARAMS);
893 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
894 				(u64 *)__LC_EXT_PARAMS2);
895 		kfree(inti);
896 	}
897 	return rc ? -EFAULT : 0;
898 }
899 
900 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
901 				     unsigned long irq_type)
902 {
903 	struct list_head *isc_list;
904 	struct kvm_s390_float_interrupt *fi;
905 	struct kvm_s390_interrupt_info *inti = NULL;
906 	int rc = 0;
907 
908 	fi = &vcpu->kvm->arch.float_int;
909 
910 	spin_lock(&fi->lock);
911 	isc_list = &fi->lists[irq_type_to_isc(irq_type)];
912 	inti = list_first_entry_or_null(isc_list,
913 					struct kvm_s390_interrupt_info,
914 					list);
915 	if (inti) {
916 		if (inti->type & KVM_S390_INT_IO_AI_MASK)
917 			VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
918 		else
919 			VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
920 			inti->io.subchannel_id >> 8,
921 			inti->io.subchannel_id >> 1 & 0x3,
922 			inti->io.subchannel_nr);
923 
924 		vcpu->stat.deliver_io_int++;
925 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
926 				inti->type,
927 				((__u32)inti->io.subchannel_id << 16) |
928 				inti->io.subchannel_nr,
929 				((__u64)inti->io.io_int_parm << 32) |
930 				inti->io.io_int_word);
931 		list_del(&inti->list);
932 		fi->counters[FIRQ_CNTR_IO] -= 1;
933 	}
934 	if (list_empty(isc_list))
935 		clear_bit(irq_type, &fi->pending_irqs);
936 	spin_unlock(&fi->lock);
937 
938 	if (inti) {
939 		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
940 				(u16 *)__LC_SUBCHANNEL_ID);
941 		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
942 				(u16 *)__LC_SUBCHANNEL_NR);
943 		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
944 				(u32 *)__LC_IO_INT_PARM);
945 		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
946 				(u32 *)__LC_IO_INT_WORD);
947 		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
948 				&vcpu->arch.sie_block->gpsw,
949 				sizeof(psw_t));
950 		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
951 				&vcpu->arch.sie_block->gpsw,
952 				sizeof(psw_t));
953 		kfree(inti);
954 	}
955 
956 	return rc ? -EFAULT : 0;
957 }
958 
959 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
960 
961 static const deliver_irq_t deliver_irq_funcs[] = {
962 	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
963 	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
964 	[IRQ_PEND_PROG]           = __deliver_prog,
965 	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
966 	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
967 	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
968 	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
969 	[IRQ_PEND_RESTART]        = __deliver_restart,
970 	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
971 	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
972 	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
973 	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
974 	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
975 };
976 
977 /* Check whether an external call is pending (deliverable or not) */
978 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
979 {
980 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
981 
982 	if (!sclp.has_sigpif)
983 		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
984 
985 	return sca_ext_call_pending(vcpu, NULL);
986 }
987 
988 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
989 {
990 	if (deliverable_irqs(vcpu))
991 		return 1;
992 
993 	if (kvm_cpu_has_pending_timer(vcpu))
994 		return 1;
995 
996 	/* external call pending and deliverable */
997 	if (kvm_s390_ext_call_pending(vcpu) &&
998 	    !psw_extint_disabled(vcpu) &&
999 	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
1000 		return 1;
1001 
1002 	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
1003 		return 1;
1004 	return 0;
1005 }
1006 
1007 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
1008 {
1009 	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
1010 }
1011 
1012 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
1013 {
1014 	u64 now, cputm, sltime = 0;
1015 
1016 	if (ckc_interrupts_enabled(vcpu)) {
1017 		now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
1018 		sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
1019 		/* already expired or overflow? */
1020 		if (!sltime || vcpu->arch.sie_block->ckc <= now)
1021 			return 0;
1022 		if (cpu_timer_interrupts_enabled(vcpu)) {
1023 			cputm = kvm_s390_get_cpu_timer(vcpu);
1024 			/* already expired? */
1025 			if (cputm >> 63)
1026 				return 0;
1027 			return min(sltime, tod_to_ns(cputm));
1028 		}
1029 	} else if (cpu_timer_interrupts_enabled(vcpu)) {
1030 		sltime = kvm_s390_get_cpu_timer(vcpu);
1031 		/* already expired? */
1032 		if (sltime >> 63)
1033 			return 0;
1034 	}
1035 	return sltime;
1036 }
1037 
1038 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
1039 {
1040 	u64 sltime;
1041 
1042 	vcpu->stat.exit_wait_state++;
1043 
1044 	/* fast path */
1045 	if (kvm_arch_vcpu_runnable(vcpu))
1046 		return 0;
1047 
1048 	if (psw_interrupts_disabled(vcpu)) {
1049 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
1050 		return -EOPNOTSUPP; /* disabled wait */
1051 	}
1052 
1053 	if (!ckc_interrupts_enabled(vcpu) &&
1054 	    !cpu_timer_interrupts_enabled(vcpu)) {
1055 		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
1056 		__set_cpu_idle(vcpu);
1057 		goto no_timer;
1058 	}
1059 
1060 	sltime = __calculate_sltime(vcpu);
1061 	if (!sltime)
1062 		return 0;
1063 
1064 	__set_cpu_idle(vcpu);
1065 	hrtimer_start(&vcpu->arch.ckc_timer, sltime, HRTIMER_MODE_REL);
1066 	VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
1067 no_timer:
1068 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1069 	kvm_vcpu_block(vcpu);
1070 	__unset_cpu_idle(vcpu);
1071 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1072 
1073 	hrtimer_cancel(&vcpu->arch.ckc_timer);
1074 	return 0;
1075 }
1076 
1077 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
1078 {
1079 	/*
1080 	 * We cannot move this into the if, as the CPU might be already
1081 	 * in kvm_vcpu_block without having the waitqueue set (polling)
1082 	 */
1083 	vcpu->valid_wakeup = true;
1084 	/*
1085 	 * This is mostly to document, that the read in swait_active could
1086 	 * be moved before other stores, leading to subtle races.
1087 	 * All current users do not store or use an atomic like update
1088 	 */
1089 	smp_mb__after_atomic();
1090 	if (swait_active(&vcpu->wq)) {
1091 		/*
1092 		 * The vcpu gave up the cpu voluntarily, mark it as a good
1093 		 * yield-candidate.
1094 		 */
1095 		vcpu->preempted = true;
1096 		swake_up(&vcpu->wq);
1097 		vcpu->stat.halt_wakeup++;
1098 	}
1099 	/*
1100 	 * The VCPU might not be sleeping but is executing the VSIE. Let's
1101 	 * kick it, so it leaves the SIE to process the request.
1102 	 */
1103 	kvm_s390_vsie_kick(vcpu);
1104 }
1105 
1106 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1107 {
1108 	struct kvm_vcpu *vcpu;
1109 	u64 sltime;
1110 
1111 	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1112 	sltime = __calculate_sltime(vcpu);
1113 
1114 	/*
1115 	 * If the monotonic clock runs faster than the tod clock we might be
1116 	 * woken up too early and have to go back to sleep to avoid deadlocks.
1117 	 */
1118 	if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1119 		return HRTIMER_RESTART;
1120 	kvm_s390_vcpu_wakeup(vcpu);
1121 	return HRTIMER_NORESTART;
1122 }
1123 
1124 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1125 {
1126 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1127 
1128 	spin_lock(&li->lock);
1129 	li->pending_irqs = 0;
1130 	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1131 	memset(&li->irq, 0, sizeof(li->irq));
1132 	spin_unlock(&li->lock);
1133 
1134 	sca_clear_ext_call(vcpu);
1135 }
1136 
1137 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1138 {
1139 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1140 	deliver_irq_t func;
1141 	int rc = 0;
1142 	unsigned long irq_type;
1143 	unsigned long irqs;
1144 
1145 	__reset_intercept_indicators(vcpu);
1146 
1147 	/* pending ckc conditions might have been invalidated */
1148 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1149 	if (ckc_irq_pending(vcpu))
1150 		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1151 
1152 	/* pending cpu timer conditions might have been invalidated */
1153 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1154 	if (cpu_timer_irq_pending(vcpu))
1155 		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1156 
1157 	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1158 		/* bits are in the order of interrupt priority */
1159 		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1160 		if (is_ioirq(irq_type)) {
1161 			rc = __deliver_io(vcpu, irq_type);
1162 		} else {
1163 			func = deliver_irq_funcs[irq_type];
1164 			if (!func) {
1165 				WARN_ON_ONCE(func == NULL);
1166 				clear_bit(irq_type, &li->pending_irqs);
1167 				continue;
1168 			}
1169 			rc = func(vcpu);
1170 		}
1171 	}
1172 
1173 	set_intercept_indicators(vcpu);
1174 
1175 	return rc;
1176 }
1177 
1178 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1179 {
1180 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1181 
1182 	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1183 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1184 				   irq->u.pgm.code, 0);
1185 
1186 	if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1187 		/* auto detection if no valid ILC was given */
1188 		irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1189 		irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1190 		irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1191 	}
1192 
1193 	if (irq->u.pgm.code == PGM_PER) {
1194 		li->irq.pgm.code |= PGM_PER;
1195 		li->irq.pgm.flags = irq->u.pgm.flags;
1196 		/* only modify PER related information */
1197 		li->irq.pgm.per_address = irq->u.pgm.per_address;
1198 		li->irq.pgm.per_code = irq->u.pgm.per_code;
1199 		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1200 		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1201 	} else if (!(irq->u.pgm.code & PGM_PER)) {
1202 		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1203 				   irq->u.pgm.code;
1204 		li->irq.pgm.flags = irq->u.pgm.flags;
1205 		/* only modify non-PER information */
1206 		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1207 		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1208 		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1209 		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1210 		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1211 		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1212 	} else {
1213 		li->irq.pgm = irq->u.pgm;
1214 	}
1215 	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1216 	return 0;
1217 }
1218 
1219 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1220 {
1221 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1222 
1223 	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1224 		   irq->u.ext.ext_params2);
1225 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1226 				   irq->u.ext.ext_params,
1227 				   irq->u.ext.ext_params2);
1228 
1229 	li->irq.ext = irq->u.ext;
1230 	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1231 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1232 	return 0;
1233 }
1234 
1235 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1236 {
1237 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1238 	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1239 	uint16_t src_id = irq->u.extcall.code;
1240 
1241 	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1242 		   src_id);
1243 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1244 				   src_id, 0);
1245 
1246 	/* sending vcpu invalid */
1247 	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1248 		return -EINVAL;
1249 
1250 	if (sclp.has_sigpif)
1251 		return sca_inject_ext_call(vcpu, src_id);
1252 
1253 	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1254 		return -EBUSY;
1255 	*extcall = irq->u.extcall;
1256 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1257 	return 0;
1258 }
1259 
1260 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1261 {
1262 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1263 	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1264 
1265 	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1266 		   irq->u.prefix.address);
1267 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1268 				   irq->u.prefix.address, 0);
1269 
1270 	if (!is_vcpu_stopped(vcpu))
1271 		return -EBUSY;
1272 
1273 	*prefix = irq->u.prefix;
1274 	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1275 	return 0;
1276 }
1277 
1278 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1279 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1280 {
1281 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1282 	struct kvm_s390_stop_info *stop = &li->irq.stop;
1283 	int rc = 0;
1284 
1285 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1286 
1287 	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1288 		return -EINVAL;
1289 
1290 	if (is_vcpu_stopped(vcpu)) {
1291 		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1292 			rc = kvm_s390_store_status_unloaded(vcpu,
1293 						KVM_S390_STORE_STATUS_NOADDR);
1294 		return rc;
1295 	}
1296 
1297 	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1298 		return -EBUSY;
1299 	stop->flags = irq->u.stop.flags;
1300 	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1301 	return 0;
1302 }
1303 
1304 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1305 				 struct kvm_s390_irq *irq)
1306 {
1307 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1308 
1309 	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1310 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1311 
1312 	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1313 	return 0;
1314 }
1315 
1316 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1317 				   struct kvm_s390_irq *irq)
1318 {
1319 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1320 
1321 	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1322 		   irq->u.emerg.code);
1323 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1324 				   irq->u.emerg.code, 0);
1325 
1326 	/* sending vcpu invalid */
1327 	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1328 		return -EINVAL;
1329 
1330 	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1331 	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1332 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1333 	return 0;
1334 }
1335 
1336 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1337 {
1338 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1339 	struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1340 
1341 	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1342 		   irq->u.mchk.mcic);
1343 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1344 				   irq->u.mchk.mcic);
1345 
1346 	/*
1347 	 * Because repressible machine checks can be indicated along with
1348 	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1349 	 * we need to combine cr14, mcic and external damage code.
1350 	 * Failing storage address and the logout area should not be or'ed
1351 	 * together, we just indicate the last occurrence of the corresponding
1352 	 * machine check
1353 	 */
1354 	mchk->cr14 |= irq->u.mchk.cr14;
1355 	mchk->mcic |= irq->u.mchk.mcic;
1356 	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1357 	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1358 	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1359 	       sizeof(mchk->fixed_logout));
1360 	if (mchk->mcic & MCHK_EX_MASK)
1361 		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1362 	else if (mchk->mcic & MCHK_REP_MASK)
1363 		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1364 	return 0;
1365 }
1366 
1367 static int __inject_ckc(struct kvm_vcpu *vcpu)
1368 {
1369 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1370 
1371 	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1372 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1373 				   0, 0);
1374 
1375 	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1376 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1377 	return 0;
1378 }
1379 
1380 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1381 {
1382 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1383 
1384 	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1385 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1386 				   0, 0);
1387 
1388 	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1389 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1390 	return 0;
1391 }
1392 
1393 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1394 						  int isc, u32 schid)
1395 {
1396 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1397 	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1398 	struct kvm_s390_interrupt_info *iter;
1399 	u16 id = (schid & 0xffff0000U) >> 16;
1400 	u16 nr = schid & 0x0000ffffU;
1401 
1402 	spin_lock(&fi->lock);
1403 	list_for_each_entry(iter, isc_list, list) {
1404 		if (schid && (id != iter->io.subchannel_id ||
1405 			      nr != iter->io.subchannel_nr))
1406 			continue;
1407 		/* found an appropriate entry */
1408 		list_del_init(&iter->list);
1409 		fi->counters[FIRQ_CNTR_IO] -= 1;
1410 		if (list_empty(isc_list))
1411 			clear_bit(isc_to_irq_type(isc), &fi->pending_irqs);
1412 		spin_unlock(&fi->lock);
1413 		return iter;
1414 	}
1415 	spin_unlock(&fi->lock);
1416 	return NULL;
1417 }
1418 
1419 /*
1420  * Dequeue and return an I/O interrupt matching any of the interruption
1421  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1422  */
1423 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1424 						    u64 isc_mask, u32 schid)
1425 {
1426 	struct kvm_s390_interrupt_info *inti = NULL;
1427 	int isc;
1428 
1429 	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1430 		if (isc_mask & isc_to_isc_bits(isc))
1431 			inti = get_io_int(kvm, isc, schid);
1432 	}
1433 	return inti;
1434 }
1435 
1436 #define SCCB_MASK 0xFFFFFFF8
1437 #define SCCB_EVENT_PENDING 0x3
1438 
1439 static int __inject_service(struct kvm *kvm,
1440 			     struct kvm_s390_interrupt_info *inti)
1441 {
1442 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1443 
1444 	spin_lock(&fi->lock);
1445 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1446 	/*
1447 	 * Early versions of the QEMU s390 bios will inject several
1448 	 * service interrupts after another without handling a
1449 	 * condition code indicating busy.
1450 	 * We will silently ignore those superfluous sccb values.
1451 	 * A future version of QEMU will take care of serialization
1452 	 * of servc requests
1453 	 */
1454 	if (fi->srv_signal.ext_params & SCCB_MASK)
1455 		goto out;
1456 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1457 	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1458 out:
1459 	spin_unlock(&fi->lock);
1460 	kfree(inti);
1461 	return 0;
1462 }
1463 
1464 static int __inject_virtio(struct kvm *kvm,
1465 			    struct kvm_s390_interrupt_info *inti)
1466 {
1467 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1468 
1469 	spin_lock(&fi->lock);
1470 	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1471 		spin_unlock(&fi->lock);
1472 		return -EBUSY;
1473 	}
1474 	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1475 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1476 	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1477 	spin_unlock(&fi->lock);
1478 	return 0;
1479 }
1480 
1481 static int __inject_pfault_done(struct kvm *kvm,
1482 				 struct kvm_s390_interrupt_info *inti)
1483 {
1484 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1485 
1486 	spin_lock(&fi->lock);
1487 	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1488 		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1489 		spin_unlock(&fi->lock);
1490 		return -EBUSY;
1491 	}
1492 	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1493 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1494 	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1495 	spin_unlock(&fi->lock);
1496 	return 0;
1497 }
1498 
1499 #define CR_PENDING_SUBCLASS 28
1500 static int __inject_float_mchk(struct kvm *kvm,
1501 				struct kvm_s390_interrupt_info *inti)
1502 {
1503 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1504 
1505 	spin_lock(&fi->lock);
1506 	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1507 	fi->mchk.mcic |= inti->mchk.mcic;
1508 	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1509 	spin_unlock(&fi->lock);
1510 	kfree(inti);
1511 	return 0;
1512 }
1513 
1514 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1515 {
1516 	struct kvm_s390_float_interrupt *fi;
1517 	struct list_head *list;
1518 	int isc;
1519 
1520 	fi = &kvm->arch.float_int;
1521 	spin_lock(&fi->lock);
1522 	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1523 		spin_unlock(&fi->lock);
1524 		return -EBUSY;
1525 	}
1526 	fi->counters[FIRQ_CNTR_IO] += 1;
1527 
1528 	if (inti->type & KVM_S390_INT_IO_AI_MASK)
1529 		VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1530 	else
1531 		VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1532 			inti->io.subchannel_id >> 8,
1533 			inti->io.subchannel_id >> 1 & 0x3,
1534 			inti->io.subchannel_nr);
1535 	isc = int_word_to_isc(inti->io.io_int_word);
1536 	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1537 	list_add_tail(&inti->list, list);
1538 	set_bit(isc_to_irq_type(isc), &fi->pending_irqs);
1539 	spin_unlock(&fi->lock);
1540 	return 0;
1541 }
1542 
1543 /*
1544  * Find a destination VCPU for a floating irq and kick it.
1545  */
1546 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1547 {
1548 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1549 	struct kvm_s390_local_interrupt *li;
1550 	struct kvm_vcpu *dst_vcpu;
1551 	int sigcpu, online_vcpus, nr_tries = 0;
1552 
1553 	online_vcpus = atomic_read(&kvm->online_vcpus);
1554 	if (!online_vcpus)
1555 		return;
1556 
1557 	/* find idle VCPUs first, then round robin */
1558 	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1559 	if (sigcpu == online_vcpus) {
1560 		do {
1561 			sigcpu = fi->next_rr_cpu;
1562 			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1563 			/* avoid endless loops if all vcpus are stopped */
1564 			if (nr_tries++ >= online_vcpus)
1565 				return;
1566 		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1567 	}
1568 	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1569 
1570 	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1571 	li = &dst_vcpu->arch.local_int;
1572 	spin_lock(&li->lock);
1573 	switch (type) {
1574 	case KVM_S390_MCHK:
1575 		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1576 		break;
1577 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1578 		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1579 		break;
1580 	default:
1581 		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1582 		break;
1583 	}
1584 	spin_unlock(&li->lock);
1585 	kvm_s390_vcpu_wakeup(dst_vcpu);
1586 }
1587 
1588 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1589 {
1590 	u64 type = READ_ONCE(inti->type);
1591 	int rc;
1592 
1593 	switch (type) {
1594 	case KVM_S390_MCHK:
1595 		rc = __inject_float_mchk(kvm, inti);
1596 		break;
1597 	case KVM_S390_INT_VIRTIO:
1598 		rc = __inject_virtio(kvm, inti);
1599 		break;
1600 	case KVM_S390_INT_SERVICE:
1601 		rc = __inject_service(kvm, inti);
1602 		break;
1603 	case KVM_S390_INT_PFAULT_DONE:
1604 		rc = __inject_pfault_done(kvm, inti);
1605 		break;
1606 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1607 		rc = __inject_io(kvm, inti);
1608 		break;
1609 	default:
1610 		rc = -EINVAL;
1611 	}
1612 	if (rc)
1613 		return rc;
1614 
1615 	__floating_irq_kick(kvm, type);
1616 	return 0;
1617 }
1618 
1619 int kvm_s390_inject_vm(struct kvm *kvm,
1620 		       struct kvm_s390_interrupt *s390int)
1621 {
1622 	struct kvm_s390_interrupt_info *inti;
1623 	int rc;
1624 
1625 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1626 	if (!inti)
1627 		return -ENOMEM;
1628 
1629 	inti->type = s390int->type;
1630 	switch (inti->type) {
1631 	case KVM_S390_INT_VIRTIO:
1632 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1633 			 s390int->parm, s390int->parm64);
1634 		inti->ext.ext_params = s390int->parm;
1635 		inti->ext.ext_params2 = s390int->parm64;
1636 		break;
1637 	case KVM_S390_INT_SERVICE:
1638 		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1639 		inti->ext.ext_params = s390int->parm;
1640 		break;
1641 	case KVM_S390_INT_PFAULT_DONE:
1642 		inti->ext.ext_params2 = s390int->parm64;
1643 		break;
1644 	case KVM_S390_MCHK:
1645 		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1646 			 s390int->parm64);
1647 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1648 		inti->mchk.mcic = s390int->parm64;
1649 		break;
1650 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1651 		inti->io.subchannel_id = s390int->parm >> 16;
1652 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1653 		inti->io.io_int_parm = s390int->parm64 >> 32;
1654 		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1655 		break;
1656 	default:
1657 		kfree(inti);
1658 		return -EINVAL;
1659 	}
1660 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1661 				 2);
1662 
1663 	rc = __inject_vm(kvm, inti);
1664 	if (rc)
1665 		kfree(inti);
1666 	return rc;
1667 }
1668 
1669 int kvm_s390_reinject_io_int(struct kvm *kvm,
1670 			      struct kvm_s390_interrupt_info *inti)
1671 {
1672 	return __inject_vm(kvm, inti);
1673 }
1674 
1675 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1676 		       struct kvm_s390_irq *irq)
1677 {
1678 	irq->type = s390int->type;
1679 	switch (irq->type) {
1680 	case KVM_S390_PROGRAM_INT:
1681 		if (s390int->parm & 0xffff0000)
1682 			return -EINVAL;
1683 		irq->u.pgm.code = s390int->parm;
1684 		break;
1685 	case KVM_S390_SIGP_SET_PREFIX:
1686 		irq->u.prefix.address = s390int->parm;
1687 		break;
1688 	case KVM_S390_SIGP_STOP:
1689 		irq->u.stop.flags = s390int->parm;
1690 		break;
1691 	case KVM_S390_INT_EXTERNAL_CALL:
1692 		if (s390int->parm & 0xffff0000)
1693 			return -EINVAL;
1694 		irq->u.extcall.code = s390int->parm;
1695 		break;
1696 	case KVM_S390_INT_EMERGENCY:
1697 		if (s390int->parm & 0xffff0000)
1698 			return -EINVAL;
1699 		irq->u.emerg.code = s390int->parm;
1700 		break;
1701 	case KVM_S390_MCHK:
1702 		irq->u.mchk.mcic = s390int->parm64;
1703 		break;
1704 	}
1705 	return 0;
1706 }
1707 
1708 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1709 {
1710 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1711 
1712 	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1713 }
1714 
1715 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1716 {
1717 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1718 
1719 	spin_lock(&li->lock);
1720 	li->irq.stop.flags = 0;
1721 	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1722 	spin_unlock(&li->lock);
1723 }
1724 
1725 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1726 {
1727 	int rc;
1728 
1729 	switch (irq->type) {
1730 	case KVM_S390_PROGRAM_INT:
1731 		rc = __inject_prog(vcpu, irq);
1732 		break;
1733 	case KVM_S390_SIGP_SET_PREFIX:
1734 		rc = __inject_set_prefix(vcpu, irq);
1735 		break;
1736 	case KVM_S390_SIGP_STOP:
1737 		rc = __inject_sigp_stop(vcpu, irq);
1738 		break;
1739 	case KVM_S390_RESTART:
1740 		rc = __inject_sigp_restart(vcpu, irq);
1741 		break;
1742 	case KVM_S390_INT_CLOCK_COMP:
1743 		rc = __inject_ckc(vcpu);
1744 		break;
1745 	case KVM_S390_INT_CPU_TIMER:
1746 		rc = __inject_cpu_timer(vcpu);
1747 		break;
1748 	case KVM_S390_INT_EXTERNAL_CALL:
1749 		rc = __inject_extcall(vcpu, irq);
1750 		break;
1751 	case KVM_S390_INT_EMERGENCY:
1752 		rc = __inject_sigp_emergency(vcpu, irq);
1753 		break;
1754 	case KVM_S390_MCHK:
1755 		rc = __inject_mchk(vcpu, irq);
1756 		break;
1757 	case KVM_S390_INT_PFAULT_INIT:
1758 		rc = __inject_pfault_init(vcpu, irq);
1759 		break;
1760 	case KVM_S390_INT_VIRTIO:
1761 	case KVM_S390_INT_SERVICE:
1762 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1763 	default:
1764 		rc = -EINVAL;
1765 	}
1766 
1767 	return rc;
1768 }
1769 
1770 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1771 {
1772 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1773 	int rc;
1774 
1775 	spin_lock(&li->lock);
1776 	rc = do_inject_vcpu(vcpu, irq);
1777 	spin_unlock(&li->lock);
1778 	if (!rc)
1779 		kvm_s390_vcpu_wakeup(vcpu);
1780 	return rc;
1781 }
1782 
1783 static inline void clear_irq_list(struct list_head *_list)
1784 {
1785 	struct kvm_s390_interrupt_info *inti, *n;
1786 
1787 	list_for_each_entry_safe(inti, n, _list, list) {
1788 		list_del(&inti->list);
1789 		kfree(inti);
1790 	}
1791 }
1792 
1793 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1794 		       struct kvm_s390_irq *irq)
1795 {
1796 	irq->type = inti->type;
1797 	switch (inti->type) {
1798 	case KVM_S390_INT_PFAULT_INIT:
1799 	case KVM_S390_INT_PFAULT_DONE:
1800 	case KVM_S390_INT_VIRTIO:
1801 		irq->u.ext = inti->ext;
1802 		break;
1803 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1804 		irq->u.io = inti->io;
1805 		break;
1806 	}
1807 }
1808 
1809 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1810 {
1811 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1812 	int i;
1813 
1814 	spin_lock(&fi->lock);
1815 	fi->pending_irqs = 0;
1816 	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1817 	memset(&fi->mchk, 0, sizeof(fi->mchk));
1818 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1819 		clear_irq_list(&fi->lists[i]);
1820 	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1821 		fi->counters[i] = 0;
1822 	spin_unlock(&fi->lock);
1823 };
1824 
1825 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1826 {
1827 	struct kvm_s390_interrupt_info *inti;
1828 	struct kvm_s390_float_interrupt *fi;
1829 	struct kvm_s390_irq *buf;
1830 	struct kvm_s390_irq *irq;
1831 	int max_irqs;
1832 	int ret = 0;
1833 	int n = 0;
1834 	int i;
1835 
1836 	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1837 		return -EINVAL;
1838 
1839 	/*
1840 	 * We are already using -ENOMEM to signal
1841 	 * userspace it may retry with a bigger buffer,
1842 	 * so we need to use something else for this case
1843 	 */
1844 	buf = vzalloc(len);
1845 	if (!buf)
1846 		return -ENOBUFS;
1847 
1848 	max_irqs = len / sizeof(struct kvm_s390_irq);
1849 
1850 	fi = &kvm->arch.float_int;
1851 	spin_lock(&fi->lock);
1852 	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1853 		list_for_each_entry(inti, &fi->lists[i], list) {
1854 			if (n == max_irqs) {
1855 				/* signal userspace to try again */
1856 				ret = -ENOMEM;
1857 				goto out;
1858 			}
1859 			inti_to_irq(inti, &buf[n]);
1860 			n++;
1861 		}
1862 	}
1863 	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1864 		if (n == max_irqs) {
1865 			/* signal userspace to try again */
1866 			ret = -ENOMEM;
1867 			goto out;
1868 		}
1869 		irq = (struct kvm_s390_irq *) &buf[n];
1870 		irq->type = KVM_S390_INT_SERVICE;
1871 		irq->u.ext = fi->srv_signal;
1872 		n++;
1873 	}
1874 	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1875 		if (n == max_irqs) {
1876 				/* signal userspace to try again */
1877 				ret = -ENOMEM;
1878 				goto out;
1879 		}
1880 		irq = (struct kvm_s390_irq *) &buf[n];
1881 		irq->type = KVM_S390_MCHK;
1882 		irq->u.mchk = fi->mchk;
1883 		n++;
1884 }
1885 
1886 out:
1887 	spin_unlock(&fi->lock);
1888 	if (!ret && n > 0) {
1889 		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1890 			ret = -EFAULT;
1891 	}
1892 	vfree(buf);
1893 
1894 	return ret < 0 ? ret : n;
1895 }
1896 
1897 static int flic_ais_mode_get_all(struct kvm *kvm, struct kvm_device_attr *attr)
1898 {
1899 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1900 	struct kvm_s390_ais_all ais;
1901 
1902 	if (attr->attr < sizeof(ais))
1903 		return -EINVAL;
1904 
1905 	if (!test_kvm_facility(kvm, 72))
1906 		return -ENOTSUPP;
1907 
1908 	mutex_lock(&fi->ais_lock);
1909 	ais.simm = fi->simm;
1910 	ais.nimm = fi->nimm;
1911 	mutex_unlock(&fi->ais_lock);
1912 
1913 	if (copy_to_user((void __user *)attr->addr, &ais, sizeof(ais)))
1914 		return -EFAULT;
1915 
1916 	return 0;
1917 }
1918 
1919 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1920 {
1921 	int r;
1922 
1923 	switch (attr->group) {
1924 	case KVM_DEV_FLIC_GET_ALL_IRQS:
1925 		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1926 					  attr->attr);
1927 		break;
1928 	case KVM_DEV_FLIC_AISM_ALL:
1929 		r = flic_ais_mode_get_all(dev->kvm, attr);
1930 		break;
1931 	default:
1932 		r = -EINVAL;
1933 	}
1934 
1935 	return r;
1936 }
1937 
1938 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1939 				     u64 addr)
1940 {
1941 	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1942 	void *target = NULL;
1943 	void __user *source;
1944 	u64 size;
1945 
1946 	if (get_user(inti->type, (u64 __user *)addr))
1947 		return -EFAULT;
1948 
1949 	switch (inti->type) {
1950 	case KVM_S390_INT_PFAULT_INIT:
1951 	case KVM_S390_INT_PFAULT_DONE:
1952 	case KVM_S390_INT_VIRTIO:
1953 	case KVM_S390_INT_SERVICE:
1954 		target = (void *) &inti->ext;
1955 		source = &uptr->u.ext;
1956 		size = sizeof(inti->ext);
1957 		break;
1958 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1959 		target = (void *) &inti->io;
1960 		source = &uptr->u.io;
1961 		size = sizeof(inti->io);
1962 		break;
1963 	case KVM_S390_MCHK:
1964 		target = (void *) &inti->mchk;
1965 		source = &uptr->u.mchk;
1966 		size = sizeof(inti->mchk);
1967 		break;
1968 	default:
1969 		return -EINVAL;
1970 	}
1971 
1972 	if (copy_from_user(target, source, size))
1973 		return -EFAULT;
1974 
1975 	return 0;
1976 }
1977 
1978 static int enqueue_floating_irq(struct kvm_device *dev,
1979 				struct kvm_device_attr *attr)
1980 {
1981 	struct kvm_s390_interrupt_info *inti = NULL;
1982 	int r = 0;
1983 	int len = attr->attr;
1984 
1985 	if (len % sizeof(struct kvm_s390_irq) != 0)
1986 		return -EINVAL;
1987 	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1988 		return -EINVAL;
1989 
1990 	while (len >= sizeof(struct kvm_s390_irq)) {
1991 		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1992 		if (!inti)
1993 			return -ENOMEM;
1994 
1995 		r = copy_irq_from_user(inti, attr->addr);
1996 		if (r) {
1997 			kfree(inti);
1998 			return r;
1999 		}
2000 		r = __inject_vm(dev->kvm, inti);
2001 		if (r) {
2002 			kfree(inti);
2003 			return r;
2004 		}
2005 		len -= sizeof(struct kvm_s390_irq);
2006 		attr->addr += sizeof(struct kvm_s390_irq);
2007 	}
2008 
2009 	return r;
2010 }
2011 
2012 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
2013 {
2014 	if (id >= MAX_S390_IO_ADAPTERS)
2015 		return NULL;
2016 	return kvm->arch.adapters[id];
2017 }
2018 
2019 static int register_io_adapter(struct kvm_device *dev,
2020 			       struct kvm_device_attr *attr)
2021 {
2022 	struct s390_io_adapter *adapter;
2023 	struct kvm_s390_io_adapter adapter_info;
2024 
2025 	if (copy_from_user(&adapter_info,
2026 			   (void __user *)attr->addr, sizeof(adapter_info)))
2027 		return -EFAULT;
2028 
2029 	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
2030 	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
2031 		return -EINVAL;
2032 
2033 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
2034 	if (!adapter)
2035 		return -ENOMEM;
2036 
2037 	INIT_LIST_HEAD(&adapter->maps);
2038 	init_rwsem(&adapter->maps_lock);
2039 	atomic_set(&adapter->nr_maps, 0);
2040 	adapter->id = adapter_info.id;
2041 	adapter->isc = adapter_info.isc;
2042 	adapter->maskable = adapter_info.maskable;
2043 	adapter->masked = false;
2044 	adapter->swap = adapter_info.swap;
2045 	adapter->suppressible = (adapter_info.flags) &
2046 				KVM_S390_ADAPTER_SUPPRESSIBLE;
2047 	dev->kvm->arch.adapters[adapter->id] = adapter;
2048 
2049 	return 0;
2050 }
2051 
2052 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
2053 {
2054 	int ret;
2055 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2056 
2057 	if (!adapter || !adapter->maskable)
2058 		return -EINVAL;
2059 	ret = adapter->masked;
2060 	adapter->masked = masked;
2061 	return ret;
2062 }
2063 
2064 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
2065 {
2066 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2067 	struct s390_map_info *map;
2068 	int ret;
2069 
2070 	if (!adapter || !addr)
2071 		return -EINVAL;
2072 
2073 	map = kzalloc(sizeof(*map), GFP_KERNEL);
2074 	if (!map) {
2075 		ret = -ENOMEM;
2076 		goto out;
2077 	}
2078 	INIT_LIST_HEAD(&map->list);
2079 	map->guest_addr = addr;
2080 	map->addr = gmap_translate(kvm->arch.gmap, addr);
2081 	if (map->addr == -EFAULT) {
2082 		ret = -EFAULT;
2083 		goto out;
2084 	}
2085 	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
2086 	if (ret < 0)
2087 		goto out;
2088 	BUG_ON(ret != 1);
2089 	down_write(&adapter->maps_lock);
2090 	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
2091 		list_add_tail(&map->list, &adapter->maps);
2092 		ret = 0;
2093 	} else {
2094 		put_page(map->page);
2095 		ret = -EINVAL;
2096 	}
2097 	up_write(&adapter->maps_lock);
2098 out:
2099 	if (ret)
2100 		kfree(map);
2101 	return ret;
2102 }
2103 
2104 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
2105 {
2106 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2107 	struct s390_map_info *map, *tmp;
2108 	int found = 0;
2109 
2110 	if (!adapter || !addr)
2111 		return -EINVAL;
2112 
2113 	down_write(&adapter->maps_lock);
2114 	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
2115 		if (map->guest_addr == addr) {
2116 			found = 1;
2117 			atomic_dec(&adapter->nr_maps);
2118 			list_del(&map->list);
2119 			put_page(map->page);
2120 			kfree(map);
2121 			break;
2122 		}
2123 	}
2124 	up_write(&adapter->maps_lock);
2125 
2126 	return found ? 0 : -EINVAL;
2127 }
2128 
2129 void kvm_s390_destroy_adapters(struct kvm *kvm)
2130 {
2131 	int i;
2132 	struct s390_map_info *map, *tmp;
2133 
2134 	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2135 		if (!kvm->arch.adapters[i])
2136 			continue;
2137 		list_for_each_entry_safe(map, tmp,
2138 					 &kvm->arch.adapters[i]->maps, list) {
2139 			list_del(&map->list);
2140 			put_page(map->page);
2141 			kfree(map);
2142 		}
2143 		kfree(kvm->arch.adapters[i]);
2144 	}
2145 }
2146 
2147 static int modify_io_adapter(struct kvm_device *dev,
2148 			     struct kvm_device_attr *attr)
2149 {
2150 	struct kvm_s390_io_adapter_req req;
2151 	struct s390_io_adapter *adapter;
2152 	int ret;
2153 
2154 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2155 		return -EFAULT;
2156 
2157 	adapter = get_io_adapter(dev->kvm, req.id);
2158 	if (!adapter)
2159 		return -EINVAL;
2160 	switch (req.type) {
2161 	case KVM_S390_IO_ADAPTER_MASK:
2162 		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2163 		if (ret > 0)
2164 			ret = 0;
2165 		break;
2166 	case KVM_S390_IO_ADAPTER_MAP:
2167 		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2168 		break;
2169 	case KVM_S390_IO_ADAPTER_UNMAP:
2170 		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2171 		break;
2172 	default:
2173 		ret = -EINVAL;
2174 	}
2175 
2176 	return ret;
2177 }
2178 
2179 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2180 
2181 {
2182 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2183 	u32 schid;
2184 
2185 	if (attr->flags)
2186 		return -EINVAL;
2187 	if (attr->attr != sizeof(schid))
2188 		return -EINVAL;
2189 	if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2190 		return -EFAULT;
2191 	if (!schid)
2192 		return -EINVAL;
2193 	kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2194 	/*
2195 	 * If userspace is conforming to the architecture, we can have at most
2196 	 * one pending I/O interrupt per subchannel, so this is effectively a
2197 	 * clear all.
2198 	 */
2199 	return 0;
2200 }
2201 
2202 static int modify_ais_mode(struct kvm *kvm, struct kvm_device_attr *attr)
2203 {
2204 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2205 	struct kvm_s390_ais_req req;
2206 	int ret = 0;
2207 
2208 	if (!test_kvm_facility(kvm, 72))
2209 		return -ENOTSUPP;
2210 
2211 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2212 		return -EFAULT;
2213 
2214 	if (req.isc > MAX_ISC)
2215 		return -EINVAL;
2216 
2217 	trace_kvm_s390_modify_ais_mode(req.isc,
2218 				       (fi->simm & AIS_MODE_MASK(req.isc)) ?
2219 				       (fi->nimm & AIS_MODE_MASK(req.isc)) ?
2220 				       2 : KVM_S390_AIS_MODE_SINGLE :
2221 				       KVM_S390_AIS_MODE_ALL, req.mode);
2222 
2223 	mutex_lock(&fi->ais_lock);
2224 	switch (req.mode) {
2225 	case KVM_S390_AIS_MODE_ALL:
2226 		fi->simm &= ~AIS_MODE_MASK(req.isc);
2227 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2228 		break;
2229 	case KVM_S390_AIS_MODE_SINGLE:
2230 		fi->simm |= AIS_MODE_MASK(req.isc);
2231 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2232 		break;
2233 	default:
2234 		ret = -EINVAL;
2235 	}
2236 	mutex_unlock(&fi->ais_lock);
2237 
2238 	return ret;
2239 }
2240 
2241 static int kvm_s390_inject_airq(struct kvm *kvm,
2242 				struct s390_io_adapter *adapter)
2243 {
2244 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2245 	struct kvm_s390_interrupt s390int = {
2246 		.type = KVM_S390_INT_IO(1, 0, 0, 0),
2247 		.parm = 0,
2248 		.parm64 = (adapter->isc << 27) | 0x80000000,
2249 	};
2250 	int ret = 0;
2251 
2252 	if (!test_kvm_facility(kvm, 72) || !adapter->suppressible)
2253 		return kvm_s390_inject_vm(kvm, &s390int);
2254 
2255 	mutex_lock(&fi->ais_lock);
2256 	if (fi->nimm & AIS_MODE_MASK(adapter->isc)) {
2257 		trace_kvm_s390_airq_suppressed(adapter->id, adapter->isc);
2258 		goto out;
2259 	}
2260 
2261 	ret = kvm_s390_inject_vm(kvm, &s390int);
2262 	if (!ret && (fi->simm & AIS_MODE_MASK(adapter->isc))) {
2263 		fi->nimm |= AIS_MODE_MASK(adapter->isc);
2264 		trace_kvm_s390_modify_ais_mode(adapter->isc,
2265 					       KVM_S390_AIS_MODE_SINGLE, 2);
2266 	}
2267 out:
2268 	mutex_unlock(&fi->ais_lock);
2269 	return ret;
2270 }
2271 
2272 static int flic_inject_airq(struct kvm *kvm, struct kvm_device_attr *attr)
2273 {
2274 	unsigned int id = attr->attr;
2275 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2276 
2277 	if (!adapter)
2278 		return -EINVAL;
2279 
2280 	return kvm_s390_inject_airq(kvm, adapter);
2281 }
2282 
2283 static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr)
2284 {
2285 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2286 	struct kvm_s390_ais_all ais;
2287 
2288 	if (!test_kvm_facility(kvm, 72))
2289 		return -ENOTSUPP;
2290 
2291 	if (copy_from_user(&ais, (void __user *)attr->addr, sizeof(ais)))
2292 		return -EFAULT;
2293 
2294 	mutex_lock(&fi->ais_lock);
2295 	fi->simm = ais.simm;
2296 	fi->nimm = ais.nimm;
2297 	mutex_unlock(&fi->ais_lock);
2298 
2299 	return 0;
2300 }
2301 
2302 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2303 {
2304 	int r = 0;
2305 	unsigned int i;
2306 	struct kvm_vcpu *vcpu;
2307 
2308 	switch (attr->group) {
2309 	case KVM_DEV_FLIC_ENQUEUE:
2310 		r = enqueue_floating_irq(dev, attr);
2311 		break;
2312 	case KVM_DEV_FLIC_CLEAR_IRQS:
2313 		kvm_s390_clear_float_irqs(dev->kvm);
2314 		break;
2315 	case KVM_DEV_FLIC_APF_ENABLE:
2316 		dev->kvm->arch.gmap->pfault_enabled = 1;
2317 		break;
2318 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2319 		dev->kvm->arch.gmap->pfault_enabled = 0;
2320 		/*
2321 		 * Make sure no async faults are in transition when
2322 		 * clearing the queues. So we don't need to worry
2323 		 * about late coming workers.
2324 		 */
2325 		synchronize_srcu(&dev->kvm->srcu);
2326 		kvm_for_each_vcpu(i, vcpu, dev->kvm)
2327 			kvm_clear_async_pf_completion_queue(vcpu);
2328 		break;
2329 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2330 		r = register_io_adapter(dev, attr);
2331 		break;
2332 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2333 		r = modify_io_adapter(dev, attr);
2334 		break;
2335 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2336 		r = clear_io_irq(dev->kvm, attr);
2337 		break;
2338 	case KVM_DEV_FLIC_AISM:
2339 		r = modify_ais_mode(dev->kvm, attr);
2340 		break;
2341 	case KVM_DEV_FLIC_AIRQ_INJECT:
2342 		r = flic_inject_airq(dev->kvm, attr);
2343 		break;
2344 	case KVM_DEV_FLIC_AISM_ALL:
2345 		r = flic_ais_mode_set_all(dev->kvm, attr);
2346 		break;
2347 	default:
2348 		r = -EINVAL;
2349 	}
2350 
2351 	return r;
2352 }
2353 
2354 static int flic_has_attr(struct kvm_device *dev,
2355 			     struct kvm_device_attr *attr)
2356 {
2357 	switch (attr->group) {
2358 	case KVM_DEV_FLIC_GET_ALL_IRQS:
2359 	case KVM_DEV_FLIC_ENQUEUE:
2360 	case KVM_DEV_FLIC_CLEAR_IRQS:
2361 	case KVM_DEV_FLIC_APF_ENABLE:
2362 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2363 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2364 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2365 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2366 	case KVM_DEV_FLIC_AISM:
2367 	case KVM_DEV_FLIC_AIRQ_INJECT:
2368 	case KVM_DEV_FLIC_AISM_ALL:
2369 		return 0;
2370 	}
2371 	return -ENXIO;
2372 }
2373 
2374 static int flic_create(struct kvm_device *dev, u32 type)
2375 {
2376 	if (!dev)
2377 		return -EINVAL;
2378 	if (dev->kvm->arch.flic)
2379 		return -EINVAL;
2380 	dev->kvm->arch.flic = dev;
2381 	return 0;
2382 }
2383 
2384 static void flic_destroy(struct kvm_device *dev)
2385 {
2386 	dev->kvm->arch.flic = NULL;
2387 	kfree(dev);
2388 }
2389 
2390 /* s390 floating irq controller (flic) */
2391 struct kvm_device_ops kvm_flic_ops = {
2392 	.name = "kvm-flic",
2393 	.get_attr = flic_get_attr,
2394 	.set_attr = flic_set_attr,
2395 	.has_attr = flic_has_attr,
2396 	.create = flic_create,
2397 	.destroy = flic_destroy,
2398 };
2399 
2400 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2401 {
2402 	unsigned long bit;
2403 
2404 	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2405 
2406 	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2407 }
2408 
2409 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2410 					  u64 addr)
2411 {
2412 	struct s390_map_info *map;
2413 
2414 	if (!adapter)
2415 		return NULL;
2416 
2417 	list_for_each_entry(map, &adapter->maps, list) {
2418 		if (map->guest_addr == addr)
2419 			return map;
2420 	}
2421 	return NULL;
2422 }
2423 
2424 static int adapter_indicators_set(struct kvm *kvm,
2425 				  struct s390_io_adapter *adapter,
2426 				  struct kvm_s390_adapter_int *adapter_int)
2427 {
2428 	unsigned long bit;
2429 	int summary_set, idx;
2430 	struct s390_map_info *info;
2431 	void *map;
2432 
2433 	info = get_map_info(adapter, adapter_int->ind_addr);
2434 	if (!info)
2435 		return -1;
2436 	map = page_address(info->page);
2437 	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2438 	set_bit(bit, map);
2439 	idx = srcu_read_lock(&kvm->srcu);
2440 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2441 	set_page_dirty_lock(info->page);
2442 	info = get_map_info(adapter, adapter_int->summary_addr);
2443 	if (!info) {
2444 		srcu_read_unlock(&kvm->srcu, idx);
2445 		return -1;
2446 	}
2447 	map = page_address(info->page);
2448 	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2449 			  adapter->swap);
2450 	summary_set = test_and_set_bit(bit, map);
2451 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2452 	set_page_dirty_lock(info->page);
2453 	srcu_read_unlock(&kvm->srcu, idx);
2454 	return summary_set ? 0 : 1;
2455 }
2456 
2457 /*
2458  * < 0 - not injected due to error
2459  * = 0 - coalesced, summary indicator already active
2460  * > 0 - injected interrupt
2461  */
2462 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2463 			   struct kvm *kvm, int irq_source_id, int level,
2464 			   bool line_status)
2465 {
2466 	int ret;
2467 	struct s390_io_adapter *adapter;
2468 
2469 	/* We're only interested in the 0->1 transition. */
2470 	if (!level)
2471 		return 0;
2472 	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2473 	if (!adapter)
2474 		return -1;
2475 	down_read(&adapter->maps_lock);
2476 	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2477 	up_read(&adapter->maps_lock);
2478 	if ((ret > 0) && !adapter->masked) {
2479 		ret = kvm_s390_inject_airq(kvm, adapter);
2480 		if (ret == 0)
2481 			ret = 1;
2482 	}
2483 	return ret;
2484 }
2485 
2486 /*
2487  * Inject the machine check to the guest.
2488  */
2489 void kvm_s390_reinject_machine_check(struct kvm_vcpu *vcpu,
2490 				     struct mcck_volatile_info *mcck_info)
2491 {
2492 	struct kvm_s390_interrupt_info inti;
2493 	struct kvm_s390_irq irq;
2494 	struct kvm_s390_mchk_info *mchk;
2495 	union mci mci;
2496 	__u64 cr14 = 0;         /* upper bits are not used */
2497 	int rc;
2498 
2499 	mci.val = mcck_info->mcic;
2500 	if (mci.sr)
2501 		cr14 |= CR14_RECOVERY_SUBMASK;
2502 	if (mci.dg)
2503 		cr14 |= CR14_DEGRADATION_SUBMASK;
2504 	if (mci.w)
2505 		cr14 |= CR14_WARNING_SUBMASK;
2506 
2507 	mchk = mci.ck ? &inti.mchk : &irq.u.mchk;
2508 	mchk->cr14 = cr14;
2509 	mchk->mcic = mcck_info->mcic;
2510 	mchk->ext_damage_code = mcck_info->ext_damage_code;
2511 	mchk->failing_storage_address = mcck_info->failing_storage_address;
2512 	if (mci.ck) {
2513 		/* Inject the floating machine check */
2514 		inti.type = KVM_S390_MCHK;
2515 		rc = __inject_vm(vcpu->kvm, &inti);
2516 	} else {
2517 		/* Inject the machine check to specified vcpu */
2518 		irq.type = KVM_S390_MCHK;
2519 		rc = kvm_s390_inject_vcpu(vcpu, &irq);
2520 	}
2521 	WARN_ON_ONCE(rc);
2522 }
2523 
2524 int kvm_set_routing_entry(struct kvm *kvm,
2525 			  struct kvm_kernel_irq_routing_entry *e,
2526 			  const struct kvm_irq_routing_entry *ue)
2527 {
2528 	int ret;
2529 
2530 	switch (ue->type) {
2531 	case KVM_IRQ_ROUTING_S390_ADAPTER:
2532 		e->set = set_adapter_int;
2533 		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2534 		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2535 		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2536 		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2537 		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2538 		ret = 0;
2539 		break;
2540 	default:
2541 		ret = -EINVAL;
2542 	}
2543 
2544 	return ret;
2545 }
2546 
2547 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2548 		int irq_source_id, int level, bool line_status)
2549 {
2550 	return -EINVAL;
2551 }
2552 
2553 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2554 {
2555 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2556 	struct kvm_s390_irq *buf;
2557 	int r = 0;
2558 	int n;
2559 
2560 	buf = vmalloc(len);
2561 	if (!buf)
2562 		return -ENOMEM;
2563 
2564 	if (copy_from_user((void *) buf, irqstate, len)) {
2565 		r = -EFAULT;
2566 		goto out_free;
2567 	}
2568 
2569 	/*
2570 	 * Don't allow setting the interrupt state
2571 	 * when there are already interrupts pending
2572 	 */
2573 	spin_lock(&li->lock);
2574 	if (li->pending_irqs) {
2575 		r = -EBUSY;
2576 		goto out_unlock;
2577 	}
2578 
2579 	for (n = 0; n < len / sizeof(*buf); n++) {
2580 		r = do_inject_vcpu(vcpu, &buf[n]);
2581 		if (r)
2582 			break;
2583 	}
2584 
2585 out_unlock:
2586 	spin_unlock(&li->lock);
2587 out_free:
2588 	vfree(buf);
2589 
2590 	return r;
2591 }
2592 
2593 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2594 			    struct kvm_s390_irq *irq,
2595 			    unsigned long irq_type)
2596 {
2597 	switch (irq_type) {
2598 	case IRQ_PEND_MCHK_EX:
2599 	case IRQ_PEND_MCHK_REP:
2600 		irq->type = KVM_S390_MCHK;
2601 		irq->u.mchk = li->irq.mchk;
2602 		break;
2603 	case IRQ_PEND_PROG:
2604 		irq->type = KVM_S390_PROGRAM_INT;
2605 		irq->u.pgm = li->irq.pgm;
2606 		break;
2607 	case IRQ_PEND_PFAULT_INIT:
2608 		irq->type = KVM_S390_INT_PFAULT_INIT;
2609 		irq->u.ext = li->irq.ext;
2610 		break;
2611 	case IRQ_PEND_EXT_EXTERNAL:
2612 		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2613 		irq->u.extcall = li->irq.extcall;
2614 		break;
2615 	case IRQ_PEND_EXT_CLOCK_COMP:
2616 		irq->type = KVM_S390_INT_CLOCK_COMP;
2617 		break;
2618 	case IRQ_PEND_EXT_CPU_TIMER:
2619 		irq->type = KVM_S390_INT_CPU_TIMER;
2620 		break;
2621 	case IRQ_PEND_SIGP_STOP:
2622 		irq->type = KVM_S390_SIGP_STOP;
2623 		irq->u.stop = li->irq.stop;
2624 		break;
2625 	case IRQ_PEND_RESTART:
2626 		irq->type = KVM_S390_RESTART;
2627 		break;
2628 	case IRQ_PEND_SET_PREFIX:
2629 		irq->type = KVM_S390_SIGP_SET_PREFIX;
2630 		irq->u.prefix = li->irq.prefix;
2631 		break;
2632 	}
2633 }
2634 
2635 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2636 {
2637 	int scn;
2638 	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2639 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2640 	unsigned long pending_irqs;
2641 	struct kvm_s390_irq irq;
2642 	unsigned long irq_type;
2643 	int cpuaddr;
2644 	int n = 0;
2645 
2646 	spin_lock(&li->lock);
2647 	pending_irqs = li->pending_irqs;
2648 	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2649 	       sizeof(sigp_emerg_pending));
2650 	spin_unlock(&li->lock);
2651 
2652 	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2653 		memset(&irq, 0, sizeof(irq));
2654 		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2655 			continue;
2656 		if (n + sizeof(irq) > len)
2657 			return -ENOBUFS;
2658 		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2659 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2660 			return -EFAULT;
2661 		n += sizeof(irq);
2662 	}
2663 
2664 	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2665 		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2666 			memset(&irq, 0, sizeof(irq));
2667 			if (n + sizeof(irq) > len)
2668 				return -ENOBUFS;
2669 			irq.type = KVM_S390_INT_EMERGENCY;
2670 			irq.u.emerg.code = cpuaddr;
2671 			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2672 				return -EFAULT;
2673 			n += sizeof(irq);
2674 		}
2675 	}
2676 
2677 	if (sca_ext_call_pending(vcpu, &scn)) {
2678 		if (n + sizeof(irq) > len)
2679 			return -ENOBUFS;
2680 		memset(&irq, 0, sizeof(irq));
2681 		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2682 		irq.u.extcall.code = scn;
2683 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2684 			return -EFAULT;
2685 		n += sizeof(irq);
2686 	}
2687 
2688 	return n;
2689 }
2690