xref: /linux/arch/s390/kvm/interrupt.c (revision 071bf69a0220253a44acb8b2a27f7a262b9a46bf)
1 /*
2  * handling kvm guest interrupts
3  *
4  * Copyright IBM Corp. 2008, 2015
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License (version 2 only)
8  * as published by the Free Software Foundation.
9  *
10  *    Author(s): Carsten Otte <cotte@de.ibm.com>
11  */
12 
13 #include <linux/interrupt.h>
14 #include <linux/kvm_host.h>
15 #include <linux/hrtimer.h>
16 #include <linux/mmu_context.h>
17 #include <linux/signal.h>
18 #include <linux/slab.h>
19 #include <linux/bitmap.h>
20 #include <linux/vmalloc.h>
21 #include <asm/asm-offsets.h>
22 #include <asm/dis.h>
23 #include <asm/uaccess.h>
24 #include <asm/sclp.h>
25 #include <asm/isc.h>
26 #include <asm/gmap.h>
27 #include "kvm-s390.h"
28 #include "gaccess.h"
29 #include "trace-s390.h"
30 
31 #define PFAULT_INIT 0x0600
32 #define PFAULT_DONE 0x0680
33 #define VIRTIO_PARAM 0x0d00
34 
35 /* handle external calls via sigp interpretation facility */
36 static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
37 {
38 	int c, scn;
39 
40 	if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
41 		return 0;
42 
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 	read_lock(&vcpu->kvm->arch.sca_lock);
72 	if (vcpu->kvm->arch.use_esca) {
73 		struct esca_block *sca = vcpu->kvm->arch.sca;
74 		union esca_sigp_ctrl *sigp_ctrl =
75 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
76 		union esca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
77 
78 		new_val.scn = src_id;
79 		new_val.c = 1;
80 		old_val.c = 0;
81 
82 		expect = old_val.value;
83 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
84 	} else {
85 		struct bsca_block *sca = vcpu->kvm->arch.sca;
86 		union bsca_sigp_ctrl *sigp_ctrl =
87 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
88 		union bsca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
89 
90 		new_val.scn = src_id;
91 		new_val.c = 1;
92 		old_val.c = 0;
93 
94 		expect = old_val.value;
95 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
96 	}
97 	read_unlock(&vcpu->kvm->arch.sca_lock);
98 
99 	if (rc != expect) {
100 		/* another external call is pending */
101 		return -EBUSY;
102 	}
103 	atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
104 	return 0;
105 }
106 
107 static void sca_clear_ext_call(struct kvm_vcpu *vcpu)
108 {
109 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
110 	int rc, expect;
111 
112 	atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
113 	read_lock(&vcpu->kvm->arch.sca_lock);
114 	if (vcpu->kvm->arch.use_esca) {
115 		struct esca_block *sca = vcpu->kvm->arch.sca;
116 		union esca_sigp_ctrl *sigp_ctrl =
117 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
118 		union esca_sigp_ctrl old = *sigp_ctrl;
119 
120 		expect = old.value;
121 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
122 	} else {
123 		struct bsca_block *sca = vcpu->kvm->arch.sca;
124 		union bsca_sigp_ctrl *sigp_ctrl =
125 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
126 		union bsca_sigp_ctrl old = *sigp_ctrl;
127 
128 		expect = old.value;
129 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
130 	}
131 	read_unlock(&vcpu->kvm->arch.sca_lock);
132 	WARN_ON(rc != expect); /* cannot clear? */
133 }
134 
135 int psw_extint_disabled(struct kvm_vcpu *vcpu)
136 {
137 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
138 }
139 
140 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
141 {
142 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
143 }
144 
145 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
146 {
147 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
148 }
149 
150 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
151 {
152 	return psw_extint_disabled(vcpu) &&
153 	       psw_ioint_disabled(vcpu) &&
154 	       psw_mchk_disabled(vcpu);
155 }
156 
157 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
158 {
159 	if (psw_extint_disabled(vcpu) ||
160 	    !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
161 		return 0;
162 	if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
163 		/* No timer interrupts when single stepping */
164 		return 0;
165 	return 1;
166 }
167 
168 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
169 {
170 	if (vcpu->arch.sie_block->ckc >= kvm_s390_get_tod_clock_fast(vcpu->kvm))
171 		return 0;
172 	return ckc_interrupts_enabled(vcpu);
173 }
174 
175 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
176 {
177 	return !psw_extint_disabled(vcpu) &&
178 	       (vcpu->arch.sie_block->gcr[0] & 0x400ul);
179 }
180 
181 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
182 {
183 	if (!cpu_timer_interrupts_enabled(vcpu))
184 		return 0;
185 	return kvm_s390_get_cpu_timer(vcpu) >> 63;
186 }
187 
188 static inline int is_ioirq(unsigned long irq_type)
189 {
190 	return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
191 		(irq_type <= IRQ_PEND_IO_ISC_7));
192 }
193 
194 static uint64_t isc_to_isc_bits(int isc)
195 {
196 	return (0x80 >> isc) << 24;
197 }
198 
199 static inline u8 int_word_to_isc(u32 int_word)
200 {
201 	return (int_word & 0x38000000) >> 27;
202 }
203 
204 static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
205 {
206 	return vcpu->kvm->arch.float_int.pending_irqs |
207 	       vcpu->arch.local_int.pending_irqs;
208 }
209 
210 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
211 				   unsigned long active_mask)
212 {
213 	int i;
214 
215 	for (i = 0; i <= MAX_ISC; i++)
216 		if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
217 			active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
218 
219 	return active_mask;
220 }
221 
222 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
223 {
224 	unsigned long active_mask;
225 
226 	active_mask = pending_irqs(vcpu);
227 	if (!active_mask)
228 		return 0;
229 
230 	if (psw_extint_disabled(vcpu))
231 		active_mask &= ~IRQ_PEND_EXT_MASK;
232 	if (psw_ioint_disabled(vcpu))
233 		active_mask &= ~IRQ_PEND_IO_MASK;
234 	else
235 		active_mask = disable_iscs(vcpu, active_mask);
236 	if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
237 		__clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
238 	if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
239 		__clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
240 	if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
241 		__clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
242 	if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
243 		__clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
244 	if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
245 		__clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
246 	if (psw_mchk_disabled(vcpu))
247 		active_mask &= ~IRQ_PEND_MCHK_MASK;
248 	if (!(vcpu->arch.sie_block->gcr[14] &
249 	      vcpu->kvm->arch.float_int.mchk.cr14))
250 		__clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
251 
252 	/*
253 	 * STOP irqs will never be actively delivered. They are triggered via
254 	 * intercept requests and cleared when the stop intercept is performed.
255 	 */
256 	__clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
257 
258 	return active_mask;
259 }
260 
261 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
262 {
263 	atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
264 	set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
265 }
266 
267 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
268 {
269 	atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
270 	clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
271 }
272 
273 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
274 {
275 	atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
276 		    &vcpu->arch.sie_block->cpuflags);
277 	vcpu->arch.sie_block->lctl = 0x0000;
278 	vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
279 
280 	if (guestdbg_enabled(vcpu)) {
281 		vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
282 					       LCTL_CR10 | LCTL_CR11);
283 		vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
284 	}
285 }
286 
287 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
288 {
289 	atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
290 }
291 
292 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
293 {
294 	if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
295 		return;
296 	else if (psw_ioint_disabled(vcpu))
297 		__set_cpuflag(vcpu, CPUSTAT_IO_INT);
298 	else
299 		vcpu->arch.sie_block->lctl |= LCTL_CR6;
300 }
301 
302 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
303 {
304 	if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
305 		return;
306 	if (psw_extint_disabled(vcpu))
307 		__set_cpuflag(vcpu, CPUSTAT_EXT_INT);
308 	else
309 		vcpu->arch.sie_block->lctl |= LCTL_CR0;
310 }
311 
312 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
313 {
314 	if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
315 		return;
316 	if (psw_mchk_disabled(vcpu))
317 		vcpu->arch.sie_block->ictl |= ICTL_LPSW;
318 	else
319 		vcpu->arch.sie_block->lctl |= LCTL_CR14;
320 }
321 
322 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
323 {
324 	if (kvm_s390_is_stop_irq_pending(vcpu))
325 		__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
326 }
327 
328 /* Set interception request for non-deliverable interrupts */
329 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
330 {
331 	set_intercept_indicators_io(vcpu);
332 	set_intercept_indicators_ext(vcpu);
333 	set_intercept_indicators_mchk(vcpu);
334 	set_intercept_indicators_stop(vcpu);
335 }
336 
337 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
338 {
339 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
340 	int rc;
341 
342 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
343 					 0, 0);
344 
345 	rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
346 			   (u16 *)__LC_EXT_INT_CODE);
347 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
348 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
349 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
350 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
351 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
352 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
353 	return rc ? -EFAULT : 0;
354 }
355 
356 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
357 {
358 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
359 	int rc;
360 
361 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
362 					 0, 0);
363 
364 	rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
365 			   (u16 __user *)__LC_EXT_INT_CODE);
366 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
367 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
368 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
369 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
370 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
371 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
372 	return rc ? -EFAULT : 0;
373 }
374 
375 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
376 {
377 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
378 	struct kvm_s390_ext_info ext;
379 	int rc;
380 
381 	spin_lock(&li->lock);
382 	ext = li->irq.ext;
383 	clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
384 	li->irq.ext.ext_params2 = 0;
385 	spin_unlock(&li->lock);
386 
387 	VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
388 		   ext.ext_params2);
389 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
390 					 KVM_S390_INT_PFAULT_INIT,
391 					 0, ext.ext_params2);
392 
393 	rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
394 	rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
395 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
396 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
397 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
398 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
399 	rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
400 	return rc ? -EFAULT : 0;
401 }
402 
403 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
404 {
405 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
406 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
407 	struct kvm_s390_mchk_info mchk = {};
408 	unsigned long adtl_status_addr;
409 	int deliver = 0;
410 	int rc = 0;
411 
412 	spin_lock(&fi->lock);
413 	spin_lock(&li->lock);
414 	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
415 	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
416 		/*
417 		 * If there was an exigent machine check pending, then any
418 		 * repressible machine checks that might have been pending
419 		 * are indicated along with it, so always clear bits for
420 		 * repressible and exigent interrupts
421 		 */
422 		mchk = li->irq.mchk;
423 		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
424 		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
425 		memset(&li->irq.mchk, 0, sizeof(mchk));
426 		deliver = 1;
427 	}
428 	/*
429 	 * We indicate floating repressible conditions along with
430 	 * other pending conditions. Channel Report Pending and Channel
431 	 * Subsystem damage are the only two and and are indicated by
432 	 * bits in mcic and masked in cr14.
433 	 */
434 	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
435 		mchk.mcic |= fi->mchk.mcic;
436 		mchk.cr14 |= fi->mchk.cr14;
437 		memset(&fi->mchk, 0, sizeof(mchk));
438 		deliver = 1;
439 	}
440 	spin_unlock(&li->lock);
441 	spin_unlock(&fi->lock);
442 
443 	if (deliver) {
444 		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
445 			   mchk.mcic);
446 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
447 						 KVM_S390_MCHK,
448 						 mchk.cr14, mchk.mcic);
449 
450 		rc  = kvm_s390_vcpu_store_status(vcpu,
451 						 KVM_S390_STORE_STATUS_PREFIXED);
452 		rc |= read_guest_lc(vcpu, __LC_VX_SAVE_AREA_ADDR,
453 				    &adtl_status_addr,
454 				    sizeof(unsigned long));
455 		rc |= kvm_s390_vcpu_store_adtl_status(vcpu,
456 						      adtl_status_addr);
457 		rc |= put_guest_lc(vcpu, mchk.mcic,
458 				   (u64 __user *) __LC_MCCK_CODE);
459 		rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
460 				   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
461 		rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
462 				     &mchk.fixed_logout,
463 				     sizeof(mchk.fixed_logout));
464 		rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
465 				     &vcpu->arch.sie_block->gpsw,
466 				     sizeof(psw_t));
467 		rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
468 				    &vcpu->arch.sie_block->gpsw,
469 				    sizeof(psw_t));
470 	}
471 	return rc ? -EFAULT : 0;
472 }
473 
474 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
475 {
476 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
477 	int rc;
478 
479 	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
480 	vcpu->stat.deliver_restart_signal++;
481 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
482 
483 	rc  = write_guest_lc(vcpu,
484 			     offsetof(struct lowcore, restart_old_psw),
485 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
486 	rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
487 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
488 	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
489 	return rc ? -EFAULT : 0;
490 }
491 
492 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
493 {
494 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
495 	struct kvm_s390_prefix_info prefix;
496 
497 	spin_lock(&li->lock);
498 	prefix = li->irq.prefix;
499 	li->irq.prefix.address = 0;
500 	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
501 	spin_unlock(&li->lock);
502 
503 	vcpu->stat.deliver_prefix_signal++;
504 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
505 					 KVM_S390_SIGP_SET_PREFIX,
506 					 prefix.address, 0);
507 
508 	kvm_s390_set_prefix(vcpu, prefix.address);
509 	return 0;
510 }
511 
512 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
513 {
514 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
515 	int rc;
516 	int cpu_addr;
517 
518 	spin_lock(&li->lock);
519 	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
520 	clear_bit(cpu_addr, li->sigp_emerg_pending);
521 	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
522 		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
523 	spin_unlock(&li->lock);
524 
525 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
526 	vcpu->stat.deliver_emergency_signal++;
527 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
528 					 cpu_addr, 0);
529 
530 	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
531 			   (u16 *)__LC_EXT_INT_CODE);
532 	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
533 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
534 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
535 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
536 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
537 	return rc ? -EFAULT : 0;
538 }
539 
540 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
541 {
542 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
543 	struct kvm_s390_extcall_info extcall;
544 	int rc;
545 
546 	spin_lock(&li->lock);
547 	extcall = li->irq.extcall;
548 	li->irq.extcall.code = 0;
549 	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
550 	spin_unlock(&li->lock);
551 
552 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
553 	vcpu->stat.deliver_external_call++;
554 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
555 					 KVM_S390_INT_EXTERNAL_CALL,
556 					 extcall.code, 0);
557 
558 	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
559 			   (u16 *)__LC_EXT_INT_CODE);
560 	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
561 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
562 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
563 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
564 			    sizeof(psw_t));
565 	return rc ? -EFAULT : 0;
566 }
567 
568 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
569 {
570 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
571 	struct kvm_s390_pgm_info pgm_info;
572 	int rc = 0, nullifying = false;
573 	u16 ilen;
574 
575 	spin_lock(&li->lock);
576 	pgm_info = li->irq.pgm;
577 	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
578 	memset(&li->irq.pgm, 0, sizeof(pgm_info));
579 	spin_unlock(&li->lock);
580 
581 	ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
582 	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
583 		   pgm_info.code, ilen);
584 	vcpu->stat.deliver_program_int++;
585 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
586 					 pgm_info.code, 0);
587 
588 	switch (pgm_info.code & ~PGM_PER) {
589 	case PGM_AFX_TRANSLATION:
590 	case PGM_ASX_TRANSLATION:
591 	case PGM_EX_TRANSLATION:
592 	case PGM_LFX_TRANSLATION:
593 	case PGM_LSTE_SEQUENCE:
594 	case PGM_LSX_TRANSLATION:
595 	case PGM_LX_TRANSLATION:
596 	case PGM_PRIMARY_AUTHORITY:
597 	case PGM_SECONDARY_AUTHORITY:
598 		nullifying = true;
599 		/* fall through */
600 	case PGM_SPACE_SWITCH:
601 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
602 				  (u64 *)__LC_TRANS_EXC_CODE);
603 		break;
604 	case PGM_ALEN_TRANSLATION:
605 	case PGM_ALE_SEQUENCE:
606 	case PGM_ASTE_INSTANCE:
607 	case PGM_ASTE_SEQUENCE:
608 	case PGM_ASTE_VALIDITY:
609 	case PGM_EXTENDED_AUTHORITY:
610 		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
611 				  (u8 *)__LC_EXC_ACCESS_ID);
612 		nullifying = true;
613 		break;
614 	case PGM_ASCE_TYPE:
615 	case PGM_PAGE_TRANSLATION:
616 	case PGM_REGION_FIRST_TRANS:
617 	case PGM_REGION_SECOND_TRANS:
618 	case PGM_REGION_THIRD_TRANS:
619 	case PGM_SEGMENT_TRANSLATION:
620 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
621 				  (u64 *)__LC_TRANS_EXC_CODE);
622 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
623 				   (u8 *)__LC_EXC_ACCESS_ID);
624 		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
625 				   (u8 *)__LC_OP_ACCESS_ID);
626 		nullifying = true;
627 		break;
628 	case PGM_MONITOR:
629 		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
630 				  (u16 *)__LC_MON_CLASS_NR);
631 		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
632 				   (u64 *)__LC_MON_CODE);
633 		break;
634 	case PGM_VECTOR_PROCESSING:
635 	case PGM_DATA:
636 		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
637 				  (u32 *)__LC_DATA_EXC_CODE);
638 		break;
639 	case PGM_PROTECTION:
640 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
641 				  (u64 *)__LC_TRANS_EXC_CODE);
642 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
643 				   (u8 *)__LC_EXC_ACCESS_ID);
644 		break;
645 	case PGM_STACK_FULL:
646 	case PGM_STACK_EMPTY:
647 	case PGM_STACK_SPECIFICATION:
648 	case PGM_STACK_TYPE:
649 	case PGM_STACK_OPERATION:
650 	case PGM_TRACE_TABEL:
651 	case PGM_CRYPTO_OPERATION:
652 		nullifying = true;
653 		break;
654 	}
655 
656 	if (pgm_info.code & PGM_PER) {
657 		rc |= put_guest_lc(vcpu, pgm_info.per_code,
658 				   (u8 *) __LC_PER_CODE);
659 		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
660 				   (u8 *)__LC_PER_ATMID);
661 		rc |= put_guest_lc(vcpu, pgm_info.per_address,
662 				   (u64 *) __LC_PER_ADDRESS);
663 		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
664 				   (u8 *) __LC_PER_ACCESS_ID);
665 	}
666 
667 	if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
668 		kvm_s390_rewind_psw(vcpu, ilen);
669 
670 	/* bit 1+2 of the target are the ilc, so we can directly use ilen */
671 	rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
672 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
673 				 (u64 *) __LC_LAST_BREAK);
674 	rc |= put_guest_lc(vcpu, pgm_info.code,
675 			   (u16 *)__LC_PGM_INT_CODE);
676 	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
677 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
678 	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
679 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
680 	return rc ? -EFAULT : 0;
681 }
682 
683 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
684 {
685 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
686 	struct kvm_s390_ext_info ext;
687 	int rc = 0;
688 
689 	spin_lock(&fi->lock);
690 	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
691 		spin_unlock(&fi->lock);
692 		return 0;
693 	}
694 	ext = fi->srv_signal;
695 	memset(&fi->srv_signal, 0, sizeof(ext));
696 	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
697 	spin_unlock(&fi->lock);
698 
699 	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
700 		   ext.ext_params);
701 	vcpu->stat.deliver_service_signal++;
702 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
703 					 ext.ext_params, 0);
704 
705 	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
706 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
707 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
708 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
709 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
710 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
711 	rc |= put_guest_lc(vcpu, ext.ext_params,
712 			   (u32 *)__LC_EXT_PARAMS);
713 
714 	return rc ? -EFAULT : 0;
715 }
716 
717 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
718 {
719 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
720 	struct kvm_s390_interrupt_info *inti;
721 	int rc = 0;
722 
723 	spin_lock(&fi->lock);
724 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
725 					struct kvm_s390_interrupt_info,
726 					list);
727 	if (inti) {
728 		list_del(&inti->list);
729 		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
730 	}
731 	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
732 		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
733 	spin_unlock(&fi->lock);
734 
735 	if (inti) {
736 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
737 						 KVM_S390_INT_PFAULT_DONE, 0,
738 						 inti->ext.ext_params2);
739 		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
740 			   inti->ext.ext_params2);
741 
742 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
743 				(u16 *)__LC_EXT_INT_CODE);
744 		rc |= put_guest_lc(vcpu, PFAULT_DONE,
745 				(u16 *)__LC_EXT_CPU_ADDR);
746 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
747 				&vcpu->arch.sie_block->gpsw,
748 				sizeof(psw_t));
749 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
750 				&vcpu->arch.sie_block->gpsw,
751 				sizeof(psw_t));
752 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
753 				(u64 *)__LC_EXT_PARAMS2);
754 		kfree(inti);
755 	}
756 	return rc ? -EFAULT : 0;
757 }
758 
759 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
760 {
761 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
762 	struct kvm_s390_interrupt_info *inti;
763 	int rc = 0;
764 
765 	spin_lock(&fi->lock);
766 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
767 					struct kvm_s390_interrupt_info,
768 					list);
769 	if (inti) {
770 		VCPU_EVENT(vcpu, 4,
771 			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
772 			   inti->ext.ext_params, inti->ext.ext_params2);
773 		vcpu->stat.deliver_virtio_interrupt++;
774 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
775 				inti->type,
776 				inti->ext.ext_params,
777 				inti->ext.ext_params2);
778 		list_del(&inti->list);
779 		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
780 	}
781 	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
782 		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
783 	spin_unlock(&fi->lock);
784 
785 	if (inti) {
786 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
787 				(u16 *)__LC_EXT_INT_CODE);
788 		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
789 				(u16 *)__LC_EXT_CPU_ADDR);
790 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
791 				&vcpu->arch.sie_block->gpsw,
792 				sizeof(psw_t));
793 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
794 				&vcpu->arch.sie_block->gpsw,
795 				sizeof(psw_t));
796 		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
797 				(u32 *)__LC_EXT_PARAMS);
798 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
799 				(u64 *)__LC_EXT_PARAMS2);
800 		kfree(inti);
801 	}
802 	return rc ? -EFAULT : 0;
803 }
804 
805 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
806 				     unsigned long irq_type)
807 {
808 	struct list_head *isc_list;
809 	struct kvm_s390_float_interrupt *fi;
810 	struct kvm_s390_interrupt_info *inti = NULL;
811 	int rc = 0;
812 
813 	fi = &vcpu->kvm->arch.float_int;
814 
815 	spin_lock(&fi->lock);
816 	isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
817 	inti = list_first_entry_or_null(isc_list,
818 					struct kvm_s390_interrupt_info,
819 					list);
820 	if (inti) {
821 		if (inti->type & KVM_S390_INT_IO_AI_MASK)
822 			VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
823 		else
824 			VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
825 			inti->io.subchannel_id >> 8,
826 			inti->io.subchannel_id >> 1 & 0x3,
827 			inti->io.subchannel_nr);
828 
829 		vcpu->stat.deliver_io_int++;
830 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
831 				inti->type,
832 				((__u32)inti->io.subchannel_id << 16) |
833 				inti->io.subchannel_nr,
834 				((__u64)inti->io.io_int_parm << 32) |
835 				inti->io.io_int_word);
836 		list_del(&inti->list);
837 		fi->counters[FIRQ_CNTR_IO] -= 1;
838 	}
839 	if (list_empty(isc_list))
840 		clear_bit(irq_type, &fi->pending_irqs);
841 	spin_unlock(&fi->lock);
842 
843 	if (inti) {
844 		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
845 				(u16 *)__LC_SUBCHANNEL_ID);
846 		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
847 				(u16 *)__LC_SUBCHANNEL_NR);
848 		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
849 				(u32 *)__LC_IO_INT_PARM);
850 		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
851 				(u32 *)__LC_IO_INT_WORD);
852 		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
853 				&vcpu->arch.sie_block->gpsw,
854 				sizeof(psw_t));
855 		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
856 				&vcpu->arch.sie_block->gpsw,
857 				sizeof(psw_t));
858 		kfree(inti);
859 	}
860 
861 	return rc ? -EFAULT : 0;
862 }
863 
864 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
865 
866 static const deliver_irq_t deliver_irq_funcs[] = {
867 	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
868 	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
869 	[IRQ_PEND_PROG]           = __deliver_prog,
870 	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
871 	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
872 	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
873 	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
874 	[IRQ_PEND_RESTART]        = __deliver_restart,
875 	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
876 	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
877 	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
878 	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
879 	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
880 };
881 
882 /* Check whether an external call is pending (deliverable or not) */
883 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
884 {
885 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
886 
887 	if (!sclp.has_sigpif)
888 		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
889 
890 	return sca_ext_call_pending(vcpu, NULL);
891 }
892 
893 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
894 {
895 	if (deliverable_irqs(vcpu))
896 		return 1;
897 
898 	if (kvm_cpu_has_pending_timer(vcpu))
899 		return 1;
900 
901 	/* external call pending and deliverable */
902 	if (kvm_s390_ext_call_pending(vcpu) &&
903 	    !psw_extint_disabled(vcpu) &&
904 	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
905 		return 1;
906 
907 	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
908 		return 1;
909 	return 0;
910 }
911 
912 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
913 {
914 	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
915 }
916 
917 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
918 {
919 	u64 now, cputm, sltime = 0;
920 
921 	if (ckc_interrupts_enabled(vcpu)) {
922 		now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
923 		sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
924 		/* already expired or overflow? */
925 		if (!sltime || vcpu->arch.sie_block->ckc <= now)
926 			return 0;
927 		if (cpu_timer_interrupts_enabled(vcpu)) {
928 			cputm = kvm_s390_get_cpu_timer(vcpu);
929 			/* already expired? */
930 			if (cputm >> 63)
931 				return 0;
932 			return min(sltime, tod_to_ns(cputm));
933 		}
934 	} else if (cpu_timer_interrupts_enabled(vcpu)) {
935 		sltime = kvm_s390_get_cpu_timer(vcpu);
936 		/* already expired? */
937 		if (sltime >> 63)
938 			return 0;
939 	}
940 	return sltime;
941 }
942 
943 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
944 {
945 	u64 sltime;
946 
947 	vcpu->stat.exit_wait_state++;
948 
949 	/* fast path */
950 	if (kvm_arch_vcpu_runnable(vcpu))
951 		return 0;
952 
953 	if (psw_interrupts_disabled(vcpu)) {
954 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
955 		return -EOPNOTSUPP; /* disabled wait */
956 	}
957 
958 	if (!ckc_interrupts_enabled(vcpu) &&
959 	    !cpu_timer_interrupts_enabled(vcpu)) {
960 		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
961 		__set_cpu_idle(vcpu);
962 		goto no_timer;
963 	}
964 
965 	sltime = __calculate_sltime(vcpu);
966 	if (!sltime)
967 		return 0;
968 
969 	__set_cpu_idle(vcpu);
970 	hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
971 	VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
972 no_timer:
973 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
974 	kvm_vcpu_block(vcpu);
975 	__unset_cpu_idle(vcpu);
976 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
977 
978 	hrtimer_cancel(&vcpu->arch.ckc_timer);
979 	return 0;
980 }
981 
982 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
983 {
984 	/*
985 	 * We cannot move this into the if, as the CPU might be already
986 	 * in kvm_vcpu_block without having the waitqueue set (polling)
987 	 */
988 	vcpu->valid_wakeup = true;
989 	if (swait_active(&vcpu->wq)) {
990 		/*
991 		 * The vcpu gave up the cpu voluntarily, mark it as a good
992 		 * yield-candidate.
993 		 */
994 		vcpu->preempted = true;
995 		swake_up(&vcpu->wq);
996 		vcpu->stat.halt_wakeup++;
997 	}
998 	/*
999 	 * The VCPU might not be sleeping but is executing the VSIE. Let's
1000 	 * kick it, so it leaves the SIE to process the request.
1001 	 */
1002 	kvm_s390_vsie_kick(vcpu);
1003 }
1004 
1005 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1006 {
1007 	struct kvm_vcpu *vcpu;
1008 	u64 sltime;
1009 
1010 	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1011 	sltime = __calculate_sltime(vcpu);
1012 
1013 	/*
1014 	 * If the monotonic clock runs faster than the tod clock we might be
1015 	 * woken up too early and have to go back to sleep to avoid deadlocks.
1016 	 */
1017 	if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1018 		return HRTIMER_RESTART;
1019 	kvm_s390_vcpu_wakeup(vcpu);
1020 	return HRTIMER_NORESTART;
1021 }
1022 
1023 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1024 {
1025 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1026 
1027 	spin_lock(&li->lock);
1028 	li->pending_irqs = 0;
1029 	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1030 	memset(&li->irq, 0, sizeof(li->irq));
1031 	spin_unlock(&li->lock);
1032 
1033 	sca_clear_ext_call(vcpu);
1034 }
1035 
1036 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1037 {
1038 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1039 	deliver_irq_t func;
1040 	int rc = 0;
1041 	unsigned long irq_type;
1042 	unsigned long irqs;
1043 
1044 	__reset_intercept_indicators(vcpu);
1045 
1046 	/* pending ckc conditions might have been invalidated */
1047 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1048 	if (ckc_irq_pending(vcpu))
1049 		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1050 
1051 	/* pending cpu timer conditions might have been invalidated */
1052 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1053 	if (cpu_timer_irq_pending(vcpu))
1054 		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1055 
1056 	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1057 		/* bits are in the order of interrupt priority */
1058 		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1059 		if (is_ioirq(irq_type)) {
1060 			rc = __deliver_io(vcpu, irq_type);
1061 		} else {
1062 			func = deliver_irq_funcs[irq_type];
1063 			if (!func) {
1064 				WARN_ON_ONCE(func == NULL);
1065 				clear_bit(irq_type, &li->pending_irqs);
1066 				continue;
1067 			}
1068 			rc = func(vcpu);
1069 		}
1070 	}
1071 
1072 	set_intercept_indicators(vcpu);
1073 
1074 	return rc;
1075 }
1076 
1077 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1078 {
1079 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1080 
1081 	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1082 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1083 				   irq->u.pgm.code, 0);
1084 
1085 	if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1086 		/* auto detection if no valid ILC was given */
1087 		irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1088 		irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1089 		irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1090 	}
1091 
1092 	if (irq->u.pgm.code == PGM_PER) {
1093 		li->irq.pgm.code |= PGM_PER;
1094 		li->irq.pgm.flags = irq->u.pgm.flags;
1095 		/* only modify PER related information */
1096 		li->irq.pgm.per_address = irq->u.pgm.per_address;
1097 		li->irq.pgm.per_code = irq->u.pgm.per_code;
1098 		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1099 		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1100 	} else if (!(irq->u.pgm.code & PGM_PER)) {
1101 		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1102 				   irq->u.pgm.code;
1103 		li->irq.pgm.flags = irq->u.pgm.flags;
1104 		/* only modify non-PER information */
1105 		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1106 		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1107 		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1108 		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1109 		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1110 		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1111 	} else {
1112 		li->irq.pgm = irq->u.pgm;
1113 	}
1114 	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1115 	return 0;
1116 }
1117 
1118 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1119 {
1120 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1121 
1122 	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1123 		   irq->u.ext.ext_params2);
1124 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1125 				   irq->u.ext.ext_params,
1126 				   irq->u.ext.ext_params2);
1127 
1128 	li->irq.ext = irq->u.ext;
1129 	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1130 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1131 	return 0;
1132 }
1133 
1134 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1135 {
1136 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1137 	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1138 	uint16_t src_id = irq->u.extcall.code;
1139 
1140 	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1141 		   src_id);
1142 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1143 				   src_id, 0);
1144 
1145 	/* sending vcpu invalid */
1146 	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1147 		return -EINVAL;
1148 
1149 	if (sclp.has_sigpif)
1150 		return sca_inject_ext_call(vcpu, src_id);
1151 
1152 	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1153 		return -EBUSY;
1154 	*extcall = irq->u.extcall;
1155 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1156 	return 0;
1157 }
1158 
1159 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1160 {
1161 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1162 	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1163 
1164 	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1165 		   irq->u.prefix.address);
1166 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1167 				   irq->u.prefix.address, 0);
1168 
1169 	if (!is_vcpu_stopped(vcpu))
1170 		return -EBUSY;
1171 
1172 	*prefix = irq->u.prefix;
1173 	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1174 	return 0;
1175 }
1176 
1177 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1178 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1179 {
1180 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1181 	struct kvm_s390_stop_info *stop = &li->irq.stop;
1182 	int rc = 0;
1183 
1184 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1185 
1186 	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1187 		return -EINVAL;
1188 
1189 	if (is_vcpu_stopped(vcpu)) {
1190 		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1191 			rc = kvm_s390_store_status_unloaded(vcpu,
1192 						KVM_S390_STORE_STATUS_NOADDR);
1193 		return rc;
1194 	}
1195 
1196 	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1197 		return -EBUSY;
1198 	stop->flags = irq->u.stop.flags;
1199 	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1200 	return 0;
1201 }
1202 
1203 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1204 				 struct kvm_s390_irq *irq)
1205 {
1206 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1207 
1208 	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1209 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1210 
1211 	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1212 	return 0;
1213 }
1214 
1215 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1216 				   struct kvm_s390_irq *irq)
1217 {
1218 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1219 
1220 	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1221 		   irq->u.emerg.code);
1222 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1223 				   irq->u.emerg.code, 0);
1224 
1225 	/* sending vcpu invalid */
1226 	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1227 		return -EINVAL;
1228 
1229 	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1230 	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1231 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1232 	return 0;
1233 }
1234 
1235 static int __inject_mchk(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_mchk_info *mchk = &li->irq.mchk;
1239 
1240 	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1241 		   irq->u.mchk.mcic);
1242 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1243 				   irq->u.mchk.mcic);
1244 
1245 	/*
1246 	 * Because repressible machine checks can be indicated along with
1247 	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1248 	 * we need to combine cr14, mcic and external damage code.
1249 	 * Failing storage address and the logout area should not be or'ed
1250 	 * together, we just indicate the last occurrence of the corresponding
1251 	 * machine check
1252 	 */
1253 	mchk->cr14 |= irq->u.mchk.cr14;
1254 	mchk->mcic |= irq->u.mchk.mcic;
1255 	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1256 	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1257 	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1258 	       sizeof(mchk->fixed_logout));
1259 	if (mchk->mcic & MCHK_EX_MASK)
1260 		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1261 	else if (mchk->mcic & MCHK_REP_MASK)
1262 		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1263 	return 0;
1264 }
1265 
1266 static int __inject_ckc(struct kvm_vcpu *vcpu)
1267 {
1268 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1269 
1270 	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1271 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1272 				   0, 0);
1273 
1274 	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1275 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1276 	return 0;
1277 }
1278 
1279 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1280 {
1281 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1282 
1283 	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1284 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1285 				   0, 0);
1286 
1287 	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1288 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1289 	return 0;
1290 }
1291 
1292 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1293 						  int isc, u32 schid)
1294 {
1295 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1296 	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1297 	struct kvm_s390_interrupt_info *iter;
1298 	u16 id = (schid & 0xffff0000U) >> 16;
1299 	u16 nr = schid & 0x0000ffffU;
1300 
1301 	spin_lock(&fi->lock);
1302 	list_for_each_entry(iter, isc_list, list) {
1303 		if (schid && (id != iter->io.subchannel_id ||
1304 			      nr != iter->io.subchannel_nr))
1305 			continue;
1306 		/* found an appropriate entry */
1307 		list_del_init(&iter->list);
1308 		fi->counters[FIRQ_CNTR_IO] -= 1;
1309 		if (list_empty(isc_list))
1310 			clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1311 		spin_unlock(&fi->lock);
1312 		return iter;
1313 	}
1314 	spin_unlock(&fi->lock);
1315 	return NULL;
1316 }
1317 
1318 /*
1319  * Dequeue and return an I/O interrupt matching any of the interruption
1320  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1321  */
1322 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1323 						    u64 isc_mask, u32 schid)
1324 {
1325 	struct kvm_s390_interrupt_info *inti = NULL;
1326 	int isc;
1327 
1328 	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1329 		if (isc_mask & isc_to_isc_bits(isc))
1330 			inti = get_io_int(kvm, isc, schid);
1331 	}
1332 	return inti;
1333 }
1334 
1335 #define SCCB_MASK 0xFFFFFFF8
1336 #define SCCB_EVENT_PENDING 0x3
1337 
1338 static int __inject_service(struct kvm *kvm,
1339 			     struct kvm_s390_interrupt_info *inti)
1340 {
1341 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1342 
1343 	spin_lock(&fi->lock);
1344 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1345 	/*
1346 	 * Early versions of the QEMU s390 bios will inject several
1347 	 * service interrupts after another without handling a
1348 	 * condition code indicating busy.
1349 	 * We will silently ignore those superfluous sccb values.
1350 	 * A future version of QEMU will take care of serialization
1351 	 * of servc requests
1352 	 */
1353 	if (fi->srv_signal.ext_params & SCCB_MASK)
1354 		goto out;
1355 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1356 	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1357 out:
1358 	spin_unlock(&fi->lock);
1359 	kfree(inti);
1360 	return 0;
1361 }
1362 
1363 static int __inject_virtio(struct kvm *kvm,
1364 			    struct kvm_s390_interrupt_info *inti)
1365 {
1366 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1367 
1368 	spin_lock(&fi->lock);
1369 	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1370 		spin_unlock(&fi->lock);
1371 		return -EBUSY;
1372 	}
1373 	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1374 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1375 	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1376 	spin_unlock(&fi->lock);
1377 	return 0;
1378 }
1379 
1380 static int __inject_pfault_done(struct kvm *kvm,
1381 				 struct kvm_s390_interrupt_info *inti)
1382 {
1383 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1384 
1385 	spin_lock(&fi->lock);
1386 	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1387 		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1388 		spin_unlock(&fi->lock);
1389 		return -EBUSY;
1390 	}
1391 	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1392 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1393 	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1394 	spin_unlock(&fi->lock);
1395 	return 0;
1396 }
1397 
1398 #define CR_PENDING_SUBCLASS 28
1399 static int __inject_float_mchk(struct kvm *kvm,
1400 				struct kvm_s390_interrupt_info *inti)
1401 {
1402 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1403 
1404 	spin_lock(&fi->lock);
1405 	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1406 	fi->mchk.mcic |= inti->mchk.mcic;
1407 	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1408 	spin_unlock(&fi->lock);
1409 	kfree(inti);
1410 	return 0;
1411 }
1412 
1413 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1414 {
1415 	struct kvm_s390_float_interrupt *fi;
1416 	struct list_head *list;
1417 	int isc;
1418 
1419 	fi = &kvm->arch.float_int;
1420 	spin_lock(&fi->lock);
1421 	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1422 		spin_unlock(&fi->lock);
1423 		return -EBUSY;
1424 	}
1425 	fi->counters[FIRQ_CNTR_IO] += 1;
1426 
1427 	if (inti->type & KVM_S390_INT_IO_AI_MASK)
1428 		VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1429 	else
1430 		VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1431 			inti->io.subchannel_id >> 8,
1432 			inti->io.subchannel_id >> 1 & 0x3,
1433 			inti->io.subchannel_nr);
1434 	isc = int_word_to_isc(inti->io.io_int_word);
1435 	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1436 	list_add_tail(&inti->list, list);
1437 	set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1438 	spin_unlock(&fi->lock);
1439 	return 0;
1440 }
1441 
1442 /*
1443  * Find a destination VCPU for a floating irq and kick it.
1444  */
1445 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1446 {
1447 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1448 	struct kvm_s390_local_interrupt *li;
1449 	struct kvm_vcpu *dst_vcpu;
1450 	int sigcpu, online_vcpus, nr_tries = 0;
1451 
1452 	online_vcpus = atomic_read(&kvm->online_vcpus);
1453 	if (!online_vcpus)
1454 		return;
1455 
1456 	/* find idle VCPUs first, then round robin */
1457 	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1458 	if (sigcpu == online_vcpus) {
1459 		do {
1460 			sigcpu = fi->next_rr_cpu;
1461 			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1462 			/* avoid endless loops if all vcpus are stopped */
1463 			if (nr_tries++ >= online_vcpus)
1464 				return;
1465 		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1466 	}
1467 	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1468 
1469 	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1470 	li = &dst_vcpu->arch.local_int;
1471 	spin_lock(&li->lock);
1472 	switch (type) {
1473 	case KVM_S390_MCHK:
1474 		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1475 		break;
1476 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1477 		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1478 		break;
1479 	default:
1480 		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1481 		break;
1482 	}
1483 	spin_unlock(&li->lock);
1484 	kvm_s390_vcpu_wakeup(dst_vcpu);
1485 }
1486 
1487 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1488 {
1489 	u64 type = READ_ONCE(inti->type);
1490 	int rc;
1491 
1492 	switch (type) {
1493 	case KVM_S390_MCHK:
1494 		rc = __inject_float_mchk(kvm, inti);
1495 		break;
1496 	case KVM_S390_INT_VIRTIO:
1497 		rc = __inject_virtio(kvm, inti);
1498 		break;
1499 	case KVM_S390_INT_SERVICE:
1500 		rc = __inject_service(kvm, inti);
1501 		break;
1502 	case KVM_S390_INT_PFAULT_DONE:
1503 		rc = __inject_pfault_done(kvm, inti);
1504 		break;
1505 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1506 		rc = __inject_io(kvm, inti);
1507 		break;
1508 	default:
1509 		rc = -EINVAL;
1510 	}
1511 	if (rc)
1512 		return rc;
1513 
1514 	__floating_irq_kick(kvm, type);
1515 	return 0;
1516 }
1517 
1518 int kvm_s390_inject_vm(struct kvm *kvm,
1519 		       struct kvm_s390_interrupt *s390int)
1520 {
1521 	struct kvm_s390_interrupt_info *inti;
1522 	int rc;
1523 
1524 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1525 	if (!inti)
1526 		return -ENOMEM;
1527 
1528 	inti->type = s390int->type;
1529 	switch (inti->type) {
1530 	case KVM_S390_INT_VIRTIO:
1531 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1532 			 s390int->parm, s390int->parm64);
1533 		inti->ext.ext_params = s390int->parm;
1534 		inti->ext.ext_params2 = s390int->parm64;
1535 		break;
1536 	case KVM_S390_INT_SERVICE:
1537 		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1538 		inti->ext.ext_params = s390int->parm;
1539 		break;
1540 	case KVM_S390_INT_PFAULT_DONE:
1541 		inti->ext.ext_params2 = s390int->parm64;
1542 		break;
1543 	case KVM_S390_MCHK:
1544 		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1545 			 s390int->parm64);
1546 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1547 		inti->mchk.mcic = s390int->parm64;
1548 		break;
1549 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1550 		inti->io.subchannel_id = s390int->parm >> 16;
1551 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1552 		inti->io.io_int_parm = s390int->parm64 >> 32;
1553 		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1554 		break;
1555 	default:
1556 		kfree(inti);
1557 		return -EINVAL;
1558 	}
1559 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1560 				 2);
1561 
1562 	rc = __inject_vm(kvm, inti);
1563 	if (rc)
1564 		kfree(inti);
1565 	return rc;
1566 }
1567 
1568 int kvm_s390_reinject_io_int(struct kvm *kvm,
1569 			      struct kvm_s390_interrupt_info *inti)
1570 {
1571 	return __inject_vm(kvm, inti);
1572 }
1573 
1574 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1575 		       struct kvm_s390_irq *irq)
1576 {
1577 	irq->type = s390int->type;
1578 	switch (irq->type) {
1579 	case KVM_S390_PROGRAM_INT:
1580 		if (s390int->parm & 0xffff0000)
1581 			return -EINVAL;
1582 		irq->u.pgm.code = s390int->parm;
1583 		break;
1584 	case KVM_S390_SIGP_SET_PREFIX:
1585 		irq->u.prefix.address = s390int->parm;
1586 		break;
1587 	case KVM_S390_SIGP_STOP:
1588 		irq->u.stop.flags = s390int->parm;
1589 		break;
1590 	case KVM_S390_INT_EXTERNAL_CALL:
1591 		if (s390int->parm & 0xffff0000)
1592 			return -EINVAL;
1593 		irq->u.extcall.code = s390int->parm;
1594 		break;
1595 	case KVM_S390_INT_EMERGENCY:
1596 		if (s390int->parm & 0xffff0000)
1597 			return -EINVAL;
1598 		irq->u.emerg.code = s390int->parm;
1599 		break;
1600 	case KVM_S390_MCHK:
1601 		irq->u.mchk.mcic = s390int->parm64;
1602 		break;
1603 	}
1604 	return 0;
1605 }
1606 
1607 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1608 {
1609 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1610 
1611 	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1612 }
1613 
1614 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1615 {
1616 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1617 
1618 	spin_lock(&li->lock);
1619 	li->irq.stop.flags = 0;
1620 	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1621 	spin_unlock(&li->lock);
1622 }
1623 
1624 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1625 {
1626 	int rc;
1627 
1628 	switch (irq->type) {
1629 	case KVM_S390_PROGRAM_INT:
1630 		rc = __inject_prog(vcpu, irq);
1631 		break;
1632 	case KVM_S390_SIGP_SET_PREFIX:
1633 		rc = __inject_set_prefix(vcpu, irq);
1634 		break;
1635 	case KVM_S390_SIGP_STOP:
1636 		rc = __inject_sigp_stop(vcpu, irq);
1637 		break;
1638 	case KVM_S390_RESTART:
1639 		rc = __inject_sigp_restart(vcpu, irq);
1640 		break;
1641 	case KVM_S390_INT_CLOCK_COMP:
1642 		rc = __inject_ckc(vcpu);
1643 		break;
1644 	case KVM_S390_INT_CPU_TIMER:
1645 		rc = __inject_cpu_timer(vcpu);
1646 		break;
1647 	case KVM_S390_INT_EXTERNAL_CALL:
1648 		rc = __inject_extcall(vcpu, irq);
1649 		break;
1650 	case KVM_S390_INT_EMERGENCY:
1651 		rc = __inject_sigp_emergency(vcpu, irq);
1652 		break;
1653 	case KVM_S390_MCHK:
1654 		rc = __inject_mchk(vcpu, irq);
1655 		break;
1656 	case KVM_S390_INT_PFAULT_INIT:
1657 		rc = __inject_pfault_init(vcpu, irq);
1658 		break;
1659 	case KVM_S390_INT_VIRTIO:
1660 	case KVM_S390_INT_SERVICE:
1661 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1662 	default:
1663 		rc = -EINVAL;
1664 	}
1665 
1666 	return rc;
1667 }
1668 
1669 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1670 {
1671 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1672 	int rc;
1673 
1674 	spin_lock(&li->lock);
1675 	rc = do_inject_vcpu(vcpu, irq);
1676 	spin_unlock(&li->lock);
1677 	if (!rc)
1678 		kvm_s390_vcpu_wakeup(vcpu);
1679 	return rc;
1680 }
1681 
1682 static inline void clear_irq_list(struct list_head *_list)
1683 {
1684 	struct kvm_s390_interrupt_info *inti, *n;
1685 
1686 	list_for_each_entry_safe(inti, n, _list, list) {
1687 		list_del(&inti->list);
1688 		kfree(inti);
1689 	}
1690 }
1691 
1692 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1693 		       struct kvm_s390_irq *irq)
1694 {
1695 	irq->type = inti->type;
1696 	switch (inti->type) {
1697 	case KVM_S390_INT_PFAULT_INIT:
1698 	case KVM_S390_INT_PFAULT_DONE:
1699 	case KVM_S390_INT_VIRTIO:
1700 		irq->u.ext = inti->ext;
1701 		break;
1702 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1703 		irq->u.io = inti->io;
1704 		break;
1705 	}
1706 }
1707 
1708 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1709 {
1710 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1711 	int i;
1712 
1713 	spin_lock(&fi->lock);
1714 	fi->pending_irqs = 0;
1715 	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1716 	memset(&fi->mchk, 0, sizeof(fi->mchk));
1717 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1718 		clear_irq_list(&fi->lists[i]);
1719 	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1720 		fi->counters[i] = 0;
1721 	spin_unlock(&fi->lock);
1722 };
1723 
1724 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1725 {
1726 	struct kvm_s390_interrupt_info *inti;
1727 	struct kvm_s390_float_interrupt *fi;
1728 	struct kvm_s390_irq *buf;
1729 	struct kvm_s390_irq *irq;
1730 	int max_irqs;
1731 	int ret = 0;
1732 	int n = 0;
1733 	int i;
1734 
1735 	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1736 		return -EINVAL;
1737 
1738 	/*
1739 	 * We are already using -ENOMEM to signal
1740 	 * userspace it may retry with a bigger buffer,
1741 	 * so we need to use something else for this case
1742 	 */
1743 	buf = vzalloc(len);
1744 	if (!buf)
1745 		return -ENOBUFS;
1746 
1747 	max_irqs = len / sizeof(struct kvm_s390_irq);
1748 
1749 	fi = &kvm->arch.float_int;
1750 	spin_lock(&fi->lock);
1751 	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1752 		list_for_each_entry(inti, &fi->lists[i], list) {
1753 			if (n == max_irqs) {
1754 				/* signal userspace to try again */
1755 				ret = -ENOMEM;
1756 				goto out;
1757 			}
1758 			inti_to_irq(inti, &buf[n]);
1759 			n++;
1760 		}
1761 	}
1762 	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1763 		if (n == max_irqs) {
1764 			/* signal userspace to try again */
1765 			ret = -ENOMEM;
1766 			goto out;
1767 		}
1768 		irq = (struct kvm_s390_irq *) &buf[n];
1769 		irq->type = KVM_S390_INT_SERVICE;
1770 		irq->u.ext = fi->srv_signal;
1771 		n++;
1772 	}
1773 	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1774 		if (n == max_irqs) {
1775 				/* signal userspace to try again */
1776 				ret = -ENOMEM;
1777 				goto out;
1778 		}
1779 		irq = (struct kvm_s390_irq *) &buf[n];
1780 		irq->type = KVM_S390_MCHK;
1781 		irq->u.mchk = fi->mchk;
1782 		n++;
1783 }
1784 
1785 out:
1786 	spin_unlock(&fi->lock);
1787 	if (!ret && n > 0) {
1788 		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1789 			ret = -EFAULT;
1790 	}
1791 	vfree(buf);
1792 
1793 	return ret < 0 ? ret : n;
1794 }
1795 
1796 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1797 {
1798 	int r;
1799 
1800 	switch (attr->group) {
1801 	case KVM_DEV_FLIC_GET_ALL_IRQS:
1802 		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1803 					  attr->attr);
1804 		break;
1805 	default:
1806 		r = -EINVAL;
1807 	}
1808 
1809 	return r;
1810 }
1811 
1812 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1813 				     u64 addr)
1814 {
1815 	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1816 	void *target = NULL;
1817 	void __user *source;
1818 	u64 size;
1819 
1820 	if (get_user(inti->type, (u64 __user *)addr))
1821 		return -EFAULT;
1822 
1823 	switch (inti->type) {
1824 	case KVM_S390_INT_PFAULT_INIT:
1825 	case KVM_S390_INT_PFAULT_DONE:
1826 	case KVM_S390_INT_VIRTIO:
1827 	case KVM_S390_INT_SERVICE:
1828 		target = (void *) &inti->ext;
1829 		source = &uptr->u.ext;
1830 		size = sizeof(inti->ext);
1831 		break;
1832 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1833 		target = (void *) &inti->io;
1834 		source = &uptr->u.io;
1835 		size = sizeof(inti->io);
1836 		break;
1837 	case KVM_S390_MCHK:
1838 		target = (void *) &inti->mchk;
1839 		source = &uptr->u.mchk;
1840 		size = sizeof(inti->mchk);
1841 		break;
1842 	default:
1843 		return -EINVAL;
1844 	}
1845 
1846 	if (copy_from_user(target, source, size))
1847 		return -EFAULT;
1848 
1849 	return 0;
1850 }
1851 
1852 static int enqueue_floating_irq(struct kvm_device *dev,
1853 				struct kvm_device_attr *attr)
1854 {
1855 	struct kvm_s390_interrupt_info *inti = NULL;
1856 	int r = 0;
1857 	int len = attr->attr;
1858 
1859 	if (len % sizeof(struct kvm_s390_irq) != 0)
1860 		return -EINVAL;
1861 	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1862 		return -EINVAL;
1863 
1864 	while (len >= sizeof(struct kvm_s390_irq)) {
1865 		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1866 		if (!inti)
1867 			return -ENOMEM;
1868 
1869 		r = copy_irq_from_user(inti, attr->addr);
1870 		if (r) {
1871 			kfree(inti);
1872 			return r;
1873 		}
1874 		r = __inject_vm(dev->kvm, inti);
1875 		if (r) {
1876 			kfree(inti);
1877 			return r;
1878 		}
1879 		len -= sizeof(struct kvm_s390_irq);
1880 		attr->addr += sizeof(struct kvm_s390_irq);
1881 	}
1882 
1883 	return r;
1884 }
1885 
1886 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1887 {
1888 	if (id >= MAX_S390_IO_ADAPTERS)
1889 		return NULL;
1890 	return kvm->arch.adapters[id];
1891 }
1892 
1893 static int register_io_adapter(struct kvm_device *dev,
1894 			       struct kvm_device_attr *attr)
1895 {
1896 	struct s390_io_adapter *adapter;
1897 	struct kvm_s390_io_adapter adapter_info;
1898 
1899 	if (copy_from_user(&adapter_info,
1900 			   (void __user *)attr->addr, sizeof(adapter_info)))
1901 		return -EFAULT;
1902 
1903 	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1904 	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1905 		return -EINVAL;
1906 
1907 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1908 	if (!adapter)
1909 		return -ENOMEM;
1910 
1911 	INIT_LIST_HEAD(&adapter->maps);
1912 	init_rwsem(&adapter->maps_lock);
1913 	atomic_set(&adapter->nr_maps, 0);
1914 	adapter->id = adapter_info.id;
1915 	adapter->isc = adapter_info.isc;
1916 	adapter->maskable = adapter_info.maskable;
1917 	adapter->masked = false;
1918 	adapter->swap = adapter_info.swap;
1919 	dev->kvm->arch.adapters[adapter->id] = adapter;
1920 
1921 	return 0;
1922 }
1923 
1924 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1925 {
1926 	int ret;
1927 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1928 
1929 	if (!adapter || !adapter->maskable)
1930 		return -EINVAL;
1931 	ret = adapter->masked;
1932 	adapter->masked = masked;
1933 	return ret;
1934 }
1935 
1936 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1937 {
1938 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1939 	struct s390_map_info *map;
1940 	int ret;
1941 
1942 	if (!adapter || !addr)
1943 		return -EINVAL;
1944 
1945 	map = kzalloc(sizeof(*map), GFP_KERNEL);
1946 	if (!map) {
1947 		ret = -ENOMEM;
1948 		goto out;
1949 	}
1950 	INIT_LIST_HEAD(&map->list);
1951 	map->guest_addr = addr;
1952 	map->addr = gmap_translate(kvm->arch.gmap, addr);
1953 	if (map->addr == -EFAULT) {
1954 		ret = -EFAULT;
1955 		goto out;
1956 	}
1957 	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1958 	if (ret < 0)
1959 		goto out;
1960 	BUG_ON(ret != 1);
1961 	down_write(&adapter->maps_lock);
1962 	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1963 		list_add_tail(&map->list, &adapter->maps);
1964 		ret = 0;
1965 	} else {
1966 		put_page(map->page);
1967 		ret = -EINVAL;
1968 	}
1969 	up_write(&adapter->maps_lock);
1970 out:
1971 	if (ret)
1972 		kfree(map);
1973 	return ret;
1974 }
1975 
1976 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1977 {
1978 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1979 	struct s390_map_info *map, *tmp;
1980 	int found = 0;
1981 
1982 	if (!adapter || !addr)
1983 		return -EINVAL;
1984 
1985 	down_write(&adapter->maps_lock);
1986 	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1987 		if (map->guest_addr == addr) {
1988 			found = 1;
1989 			atomic_dec(&adapter->nr_maps);
1990 			list_del(&map->list);
1991 			put_page(map->page);
1992 			kfree(map);
1993 			break;
1994 		}
1995 	}
1996 	up_write(&adapter->maps_lock);
1997 
1998 	return found ? 0 : -EINVAL;
1999 }
2000 
2001 void kvm_s390_destroy_adapters(struct kvm *kvm)
2002 {
2003 	int i;
2004 	struct s390_map_info *map, *tmp;
2005 
2006 	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2007 		if (!kvm->arch.adapters[i])
2008 			continue;
2009 		list_for_each_entry_safe(map, tmp,
2010 					 &kvm->arch.adapters[i]->maps, list) {
2011 			list_del(&map->list);
2012 			put_page(map->page);
2013 			kfree(map);
2014 		}
2015 		kfree(kvm->arch.adapters[i]);
2016 	}
2017 }
2018 
2019 static int modify_io_adapter(struct kvm_device *dev,
2020 			     struct kvm_device_attr *attr)
2021 {
2022 	struct kvm_s390_io_adapter_req req;
2023 	struct s390_io_adapter *adapter;
2024 	int ret;
2025 
2026 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2027 		return -EFAULT;
2028 
2029 	adapter = get_io_adapter(dev->kvm, req.id);
2030 	if (!adapter)
2031 		return -EINVAL;
2032 	switch (req.type) {
2033 	case KVM_S390_IO_ADAPTER_MASK:
2034 		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2035 		if (ret > 0)
2036 			ret = 0;
2037 		break;
2038 	case KVM_S390_IO_ADAPTER_MAP:
2039 		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2040 		break;
2041 	case KVM_S390_IO_ADAPTER_UNMAP:
2042 		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2043 		break;
2044 	default:
2045 		ret = -EINVAL;
2046 	}
2047 
2048 	return ret;
2049 }
2050 
2051 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2052 
2053 {
2054 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2055 	u32 schid;
2056 
2057 	if (attr->flags)
2058 		return -EINVAL;
2059 	if (attr->attr != sizeof(schid))
2060 		return -EINVAL;
2061 	if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2062 		return -EFAULT;
2063 	kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2064 	/*
2065 	 * If userspace is conforming to the architecture, we can have at most
2066 	 * one pending I/O interrupt per subchannel, so this is effectively a
2067 	 * clear all.
2068 	 */
2069 	return 0;
2070 }
2071 
2072 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2073 {
2074 	int r = 0;
2075 	unsigned int i;
2076 	struct kvm_vcpu *vcpu;
2077 
2078 	switch (attr->group) {
2079 	case KVM_DEV_FLIC_ENQUEUE:
2080 		r = enqueue_floating_irq(dev, attr);
2081 		break;
2082 	case KVM_DEV_FLIC_CLEAR_IRQS:
2083 		kvm_s390_clear_float_irqs(dev->kvm);
2084 		break;
2085 	case KVM_DEV_FLIC_APF_ENABLE:
2086 		dev->kvm->arch.gmap->pfault_enabled = 1;
2087 		break;
2088 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2089 		dev->kvm->arch.gmap->pfault_enabled = 0;
2090 		/*
2091 		 * Make sure no async faults are in transition when
2092 		 * clearing the queues. So we don't need to worry
2093 		 * about late coming workers.
2094 		 */
2095 		synchronize_srcu(&dev->kvm->srcu);
2096 		kvm_for_each_vcpu(i, vcpu, dev->kvm)
2097 			kvm_clear_async_pf_completion_queue(vcpu);
2098 		break;
2099 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2100 		r = register_io_adapter(dev, attr);
2101 		break;
2102 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2103 		r = modify_io_adapter(dev, attr);
2104 		break;
2105 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2106 		r = clear_io_irq(dev->kvm, attr);
2107 		break;
2108 	default:
2109 		r = -EINVAL;
2110 	}
2111 
2112 	return r;
2113 }
2114 
2115 static int flic_has_attr(struct kvm_device *dev,
2116 			     struct kvm_device_attr *attr)
2117 {
2118 	switch (attr->group) {
2119 	case KVM_DEV_FLIC_GET_ALL_IRQS:
2120 	case KVM_DEV_FLIC_ENQUEUE:
2121 	case KVM_DEV_FLIC_CLEAR_IRQS:
2122 	case KVM_DEV_FLIC_APF_ENABLE:
2123 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2124 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2125 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2126 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2127 		return 0;
2128 	}
2129 	return -ENXIO;
2130 }
2131 
2132 static int flic_create(struct kvm_device *dev, u32 type)
2133 {
2134 	if (!dev)
2135 		return -EINVAL;
2136 	if (dev->kvm->arch.flic)
2137 		return -EINVAL;
2138 	dev->kvm->arch.flic = dev;
2139 	return 0;
2140 }
2141 
2142 static void flic_destroy(struct kvm_device *dev)
2143 {
2144 	dev->kvm->arch.flic = NULL;
2145 	kfree(dev);
2146 }
2147 
2148 /* s390 floating irq controller (flic) */
2149 struct kvm_device_ops kvm_flic_ops = {
2150 	.name = "kvm-flic",
2151 	.get_attr = flic_get_attr,
2152 	.set_attr = flic_set_attr,
2153 	.has_attr = flic_has_attr,
2154 	.create = flic_create,
2155 	.destroy = flic_destroy,
2156 };
2157 
2158 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2159 {
2160 	unsigned long bit;
2161 
2162 	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2163 
2164 	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2165 }
2166 
2167 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2168 					  u64 addr)
2169 {
2170 	struct s390_map_info *map;
2171 
2172 	if (!adapter)
2173 		return NULL;
2174 
2175 	list_for_each_entry(map, &adapter->maps, list) {
2176 		if (map->guest_addr == addr)
2177 			return map;
2178 	}
2179 	return NULL;
2180 }
2181 
2182 static int adapter_indicators_set(struct kvm *kvm,
2183 				  struct s390_io_adapter *adapter,
2184 				  struct kvm_s390_adapter_int *adapter_int)
2185 {
2186 	unsigned long bit;
2187 	int summary_set, idx;
2188 	struct s390_map_info *info;
2189 	void *map;
2190 
2191 	info = get_map_info(adapter, adapter_int->ind_addr);
2192 	if (!info)
2193 		return -1;
2194 	map = page_address(info->page);
2195 	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2196 	set_bit(bit, map);
2197 	idx = srcu_read_lock(&kvm->srcu);
2198 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2199 	set_page_dirty_lock(info->page);
2200 	info = get_map_info(adapter, adapter_int->summary_addr);
2201 	if (!info) {
2202 		srcu_read_unlock(&kvm->srcu, idx);
2203 		return -1;
2204 	}
2205 	map = page_address(info->page);
2206 	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2207 			  adapter->swap);
2208 	summary_set = test_and_set_bit(bit, map);
2209 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2210 	set_page_dirty_lock(info->page);
2211 	srcu_read_unlock(&kvm->srcu, idx);
2212 	return summary_set ? 0 : 1;
2213 }
2214 
2215 /*
2216  * < 0 - not injected due to error
2217  * = 0 - coalesced, summary indicator already active
2218  * > 0 - injected interrupt
2219  */
2220 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2221 			   struct kvm *kvm, int irq_source_id, int level,
2222 			   bool line_status)
2223 {
2224 	int ret;
2225 	struct s390_io_adapter *adapter;
2226 
2227 	/* We're only interested in the 0->1 transition. */
2228 	if (!level)
2229 		return 0;
2230 	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2231 	if (!adapter)
2232 		return -1;
2233 	down_read(&adapter->maps_lock);
2234 	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2235 	up_read(&adapter->maps_lock);
2236 	if ((ret > 0) && !adapter->masked) {
2237 		struct kvm_s390_interrupt s390int = {
2238 			.type = KVM_S390_INT_IO(1, 0, 0, 0),
2239 			.parm = 0,
2240 			.parm64 = (adapter->isc << 27) | 0x80000000,
2241 		};
2242 		ret = kvm_s390_inject_vm(kvm, &s390int);
2243 		if (ret == 0)
2244 			ret = 1;
2245 	}
2246 	return ret;
2247 }
2248 
2249 int kvm_set_routing_entry(struct kvm *kvm,
2250 			  struct kvm_kernel_irq_routing_entry *e,
2251 			  const struct kvm_irq_routing_entry *ue)
2252 {
2253 	int ret;
2254 
2255 	switch (ue->type) {
2256 	case KVM_IRQ_ROUTING_S390_ADAPTER:
2257 		e->set = set_adapter_int;
2258 		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2259 		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2260 		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2261 		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2262 		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2263 		ret = 0;
2264 		break;
2265 	default:
2266 		ret = -EINVAL;
2267 	}
2268 
2269 	return ret;
2270 }
2271 
2272 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2273 		int irq_source_id, int level, bool line_status)
2274 {
2275 	return -EINVAL;
2276 }
2277 
2278 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2279 {
2280 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2281 	struct kvm_s390_irq *buf;
2282 	int r = 0;
2283 	int n;
2284 
2285 	buf = vmalloc(len);
2286 	if (!buf)
2287 		return -ENOMEM;
2288 
2289 	if (copy_from_user((void *) buf, irqstate, len)) {
2290 		r = -EFAULT;
2291 		goto out_free;
2292 	}
2293 
2294 	/*
2295 	 * Don't allow setting the interrupt state
2296 	 * when there are already interrupts pending
2297 	 */
2298 	spin_lock(&li->lock);
2299 	if (li->pending_irqs) {
2300 		r = -EBUSY;
2301 		goto out_unlock;
2302 	}
2303 
2304 	for (n = 0; n < len / sizeof(*buf); n++) {
2305 		r = do_inject_vcpu(vcpu, &buf[n]);
2306 		if (r)
2307 			break;
2308 	}
2309 
2310 out_unlock:
2311 	spin_unlock(&li->lock);
2312 out_free:
2313 	vfree(buf);
2314 
2315 	return r;
2316 }
2317 
2318 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2319 			    struct kvm_s390_irq *irq,
2320 			    unsigned long irq_type)
2321 {
2322 	switch (irq_type) {
2323 	case IRQ_PEND_MCHK_EX:
2324 	case IRQ_PEND_MCHK_REP:
2325 		irq->type = KVM_S390_MCHK;
2326 		irq->u.mchk = li->irq.mchk;
2327 		break;
2328 	case IRQ_PEND_PROG:
2329 		irq->type = KVM_S390_PROGRAM_INT;
2330 		irq->u.pgm = li->irq.pgm;
2331 		break;
2332 	case IRQ_PEND_PFAULT_INIT:
2333 		irq->type = KVM_S390_INT_PFAULT_INIT;
2334 		irq->u.ext = li->irq.ext;
2335 		break;
2336 	case IRQ_PEND_EXT_EXTERNAL:
2337 		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2338 		irq->u.extcall = li->irq.extcall;
2339 		break;
2340 	case IRQ_PEND_EXT_CLOCK_COMP:
2341 		irq->type = KVM_S390_INT_CLOCK_COMP;
2342 		break;
2343 	case IRQ_PEND_EXT_CPU_TIMER:
2344 		irq->type = KVM_S390_INT_CPU_TIMER;
2345 		break;
2346 	case IRQ_PEND_SIGP_STOP:
2347 		irq->type = KVM_S390_SIGP_STOP;
2348 		irq->u.stop = li->irq.stop;
2349 		break;
2350 	case IRQ_PEND_RESTART:
2351 		irq->type = KVM_S390_RESTART;
2352 		break;
2353 	case IRQ_PEND_SET_PREFIX:
2354 		irq->type = KVM_S390_SIGP_SET_PREFIX;
2355 		irq->u.prefix = li->irq.prefix;
2356 		break;
2357 	}
2358 }
2359 
2360 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2361 {
2362 	int scn;
2363 	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2364 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2365 	unsigned long pending_irqs;
2366 	struct kvm_s390_irq irq;
2367 	unsigned long irq_type;
2368 	int cpuaddr;
2369 	int n = 0;
2370 
2371 	spin_lock(&li->lock);
2372 	pending_irqs = li->pending_irqs;
2373 	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2374 	       sizeof(sigp_emerg_pending));
2375 	spin_unlock(&li->lock);
2376 
2377 	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2378 		memset(&irq, 0, sizeof(irq));
2379 		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2380 			continue;
2381 		if (n + sizeof(irq) > len)
2382 			return -ENOBUFS;
2383 		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2384 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2385 			return -EFAULT;
2386 		n += sizeof(irq);
2387 	}
2388 
2389 	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2390 		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2391 			memset(&irq, 0, sizeof(irq));
2392 			if (n + sizeof(irq) > len)
2393 				return -ENOBUFS;
2394 			irq.type = KVM_S390_INT_EMERGENCY;
2395 			irq.u.emerg.code = cpuaddr;
2396 			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2397 				return -EFAULT;
2398 			n += sizeof(irq);
2399 		}
2400 	}
2401 
2402 	if (sca_ext_call_pending(vcpu, &scn)) {
2403 		if (n + sizeof(irq) > len)
2404 			return -ENOBUFS;
2405 		memset(&irq, 0, sizeof(irq));
2406 		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2407 		irq.u.extcall.code = scn;
2408 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2409 			return -EFAULT;
2410 		n += sizeof(irq);
2411 	}
2412 
2413 	return n;
2414 }
2415