xref: /openbmc/linux/arch/s390/kvm/interrupt.c (revision 020c5260)
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 <linux/uaccess.h>
24 #include <asm/sclp.h>
25 #include <asm/isc.h>
26 #include <asm/gmap.h>
27 #include <asm/switch_to.h>
28 #include <asm/nmi.h>
29 #include "kvm-s390.h"
30 #include "gaccess.h"
31 #include "trace-s390.h"
32 
33 #define PFAULT_INIT 0x0600
34 #define PFAULT_DONE 0x0680
35 #define VIRTIO_PARAM 0x0d00
36 
37 /* handle external calls via sigp interpretation facility */
38 static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
39 {
40 	int c, scn;
41 
42 	if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
43 		return 0;
44 
45 	BUG_ON(!kvm_s390_use_sca_entries());
46 	read_lock(&vcpu->kvm->arch.sca_lock);
47 	if (vcpu->kvm->arch.use_esca) {
48 		struct esca_block *sca = vcpu->kvm->arch.sca;
49 		union esca_sigp_ctrl sigp_ctrl =
50 			sca->cpu[vcpu->vcpu_id].sigp_ctrl;
51 
52 		c = sigp_ctrl.c;
53 		scn = sigp_ctrl.scn;
54 	} else {
55 		struct bsca_block *sca = vcpu->kvm->arch.sca;
56 		union bsca_sigp_ctrl sigp_ctrl =
57 			sca->cpu[vcpu->vcpu_id].sigp_ctrl;
58 
59 		c = sigp_ctrl.c;
60 		scn = sigp_ctrl.scn;
61 	}
62 	read_unlock(&vcpu->kvm->arch.sca_lock);
63 
64 	if (src_id)
65 		*src_id = scn;
66 
67 	return c;
68 }
69 
70 static int sca_inject_ext_call(struct kvm_vcpu *vcpu, int src_id)
71 {
72 	int expect, rc;
73 
74 	BUG_ON(!kvm_s390_use_sca_entries());
75 	read_lock(&vcpu->kvm->arch.sca_lock);
76 	if (vcpu->kvm->arch.use_esca) {
77 		struct esca_block *sca = vcpu->kvm->arch.sca;
78 		union esca_sigp_ctrl *sigp_ctrl =
79 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
80 		union esca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
81 
82 		new_val.scn = src_id;
83 		new_val.c = 1;
84 		old_val.c = 0;
85 
86 		expect = old_val.value;
87 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
88 	} else {
89 		struct bsca_block *sca = vcpu->kvm->arch.sca;
90 		union bsca_sigp_ctrl *sigp_ctrl =
91 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
92 		union bsca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
93 
94 		new_val.scn = src_id;
95 		new_val.c = 1;
96 		old_val.c = 0;
97 
98 		expect = old_val.value;
99 		rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
100 	}
101 	read_unlock(&vcpu->kvm->arch.sca_lock);
102 
103 	if (rc != expect) {
104 		/* another external call is pending */
105 		return -EBUSY;
106 	}
107 	atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
108 	return 0;
109 }
110 
111 static void sca_clear_ext_call(struct kvm_vcpu *vcpu)
112 {
113 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
114 	int rc, expect;
115 
116 	if (!kvm_s390_use_sca_entries())
117 		return;
118 	atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
119 	read_lock(&vcpu->kvm->arch.sca_lock);
120 	if (vcpu->kvm->arch.use_esca) {
121 		struct esca_block *sca = vcpu->kvm->arch.sca;
122 		union esca_sigp_ctrl *sigp_ctrl =
123 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
124 		union esca_sigp_ctrl old = *sigp_ctrl;
125 
126 		expect = old.value;
127 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
128 	} else {
129 		struct bsca_block *sca = vcpu->kvm->arch.sca;
130 		union bsca_sigp_ctrl *sigp_ctrl =
131 			&(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
132 		union bsca_sigp_ctrl old = *sigp_ctrl;
133 
134 		expect = old.value;
135 		rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
136 	}
137 	read_unlock(&vcpu->kvm->arch.sca_lock);
138 	WARN_ON(rc != expect); /* cannot clear? */
139 }
140 
141 int psw_extint_disabled(struct kvm_vcpu *vcpu)
142 {
143 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
144 }
145 
146 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
147 {
148 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
149 }
150 
151 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
152 {
153 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
154 }
155 
156 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
157 {
158 	return psw_extint_disabled(vcpu) &&
159 	       psw_ioint_disabled(vcpu) &&
160 	       psw_mchk_disabled(vcpu);
161 }
162 
163 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
164 {
165 	if (psw_extint_disabled(vcpu) ||
166 	    !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
167 		return 0;
168 	if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
169 		/* No timer interrupts when single stepping */
170 		return 0;
171 	return 1;
172 }
173 
174 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
175 {
176 	if (vcpu->arch.sie_block->ckc >= kvm_s390_get_tod_clock_fast(vcpu->kvm))
177 		return 0;
178 	return ckc_interrupts_enabled(vcpu);
179 }
180 
181 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
182 {
183 	return !psw_extint_disabled(vcpu) &&
184 	       (vcpu->arch.sie_block->gcr[0] & 0x400ul);
185 }
186 
187 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
188 {
189 	if (!cpu_timer_interrupts_enabled(vcpu))
190 		return 0;
191 	return kvm_s390_get_cpu_timer(vcpu) >> 63;
192 }
193 
194 static inline int is_ioirq(unsigned long irq_type)
195 {
196 	return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
197 		(irq_type <= IRQ_PEND_IO_ISC_7));
198 }
199 
200 static uint64_t isc_to_isc_bits(int isc)
201 {
202 	return (0x80 >> isc) << 24;
203 }
204 
205 static inline u8 int_word_to_isc(u32 int_word)
206 {
207 	return (int_word & 0x38000000) >> 27;
208 }
209 
210 static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
211 {
212 	return vcpu->kvm->arch.float_int.pending_irqs |
213 	       vcpu->arch.local_int.pending_irqs;
214 }
215 
216 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
217 				   unsigned long active_mask)
218 {
219 	int i;
220 
221 	for (i = 0; i <= MAX_ISC; i++)
222 		if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
223 			active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
224 
225 	return active_mask;
226 }
227 
228 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
229 {
230 	unsigned long active_mask;
231 
232 	active_mask = pending_irqs(vcpu);
233 	if (!active_mask)
234 		return 0;
235 
236 	if (psw_extint_disabled(vcpu))
237 		active_mask &= ~IRQ_PEND_EXT_MASK;
238 	if (psw_ioint_disabled(vcpu))
239 		active_mask &= ~IRQ_PEND_IO_MASK;
240 	else
241 		active_mask = disable_iscs(vcpu, active_mask);
242 	if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
243 		__clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
244 	if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
245 		__clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
246 	if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
247 		__clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
248 	if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
249 		__clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
250 	if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
251 		__clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
252 	if (psw_mchk_disabled(vcpu))
253 		active_mask &= ~IRQ_PEND_MCHK_MASK;
254 	if (!(vcpu->arch.sie_block->gcr[14] &
255 	      vcpu->kvm->arch.float_int.mchk.cr14))
256 		__clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
257 
258 	/*
259 	 * STOP irqs will never be actively delivered. They are triggered via
260 	 * intercept requests and cleared when the stop intercept is performed.
261 	 */
262 	__clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
263 
264 	return active_mask;
265 }
266 
267 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
268 {
269 	atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
270 	set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
271 }
272 
273 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
274 {
275 	atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
276 	clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
277 }
278 
279 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
280 {
281 	atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
282 		    &vcpu->arch.sie_block->cpuflags);
283 	vcpu->arch.sie_block->lctl = 0x0000;
284 	vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
285 
286 	if (guestdbg_enabled(vcpu)) {
287 		vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
288 					       LCTL_CR10 | LCTL_CR11);
289 		vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
290 	}
291 }
292 
293 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
294 {
295 	atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
296 }
297 
298 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
299 {
300 	if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
301 		return;
302 	else if (psw_ioint_disabled(vcpu))
303 		__set_cpuflag(vcpu, CPUSTAT_IO_INT);
304 	else
305 		vcpu->arch.sie_block->lctl |= LCTL_CR6;
306 }
307 
308 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
309 {
310 	if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
311 		return;
312 	if (psw_extint_disabled(vcpu))
313 		__set_cpuflag(vcpu, CPUSTAT_EXT_INT);
314 	else
315 		vcpu->arch.sie_block->lctl |= LCTL_CR0;
316 }
317 
318 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
319 {
320 	if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
321 		return;
322 	if (psw_mchk_disabled(vcpu))
323 		vcpu->arch.sie_block->ictl |= ICTL_LPSW;
324 	else
325 		vcpu->arch.sie_block->lctl |= LCTL_CR14;
326 }
327 
328 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
329 {
330 	if (kvm_s390_is_stop_irq_pending(vcpu))
331 		__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
332 }
333 
334 /* Set interception request for non-deliverable interrupts */
335 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
336 {
337 	set_intercept_indicators_io(vcpu);
338 	set_intercept_indicators_ext(vcpu);
339 	set_intercept_indicators_mchk(vcpu);
340 	set_intercept_indicators_stop(vcpu);
341 }
342 
343 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
344 {
345 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
346 	int rc;
347 
348 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
349 					 0, 0);
350 
351 	rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
352 			   (u16 *)__LC_EXT_INT_CODE);
353 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
354 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
355 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
356 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
357 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
358 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
359 	return rc ? -EFAULT : 0;
360 }
361 
362 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
363 {
364 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
365 	int rc;
366 
367 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
368 					 0, 0);
369 
370 	rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
371 			   (u16 __user *)__LC_EXT_INT_CODE);
372 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
373 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
374 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
375 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
376 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
377 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
378 	return rc ? -EFAULT : 0;
379 }
380 
381 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
382 {
383 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
384 	struct kvm_s390_ext_info ext;
385 	int rc;
386 
387 	spin_lock(&li->lock);
388 	ext = li->irq.ext;
389 	clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
390 	li->irq.ext.ext_params2 = 0;
391 	spin_unlock(&li->lock);
392 
393 	VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
394 		   ext.ext_params2);
395 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
396 					 KVM_S390_INT_PFAULT_INIT,
397 					 0, ext.ext_params2);
398 
399 	rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
400 	rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
401 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
402 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
403 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
404 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
405 	rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
406 	return rc ? -EFAULT : 0;
407 }
408 
409 static int __write_machine_check(struct kvm_vcpu *vcpu,
410 				 struct kvm_s390_mchk_info *mchk)
411 {
412 	unsigned long ext_sa_addr;
413 	unsigned long lc;
414 	freg_t fprs[NUM_FPRS];
415 	union mci mci;
416 	int rc;
417 
418 	mci.val = mchk->mcic;
419 	/* take care of lazy register loading */
420 	save_fpu_regs();
421 	save_access_regs(vcpu->run->s.regs.acrs);
422 	if (MACHINE_HAS_GS && vcpu->arch.gs_enabled)
423 		save_gs_cb(current->thread.gs_cb);
424 
425 	/* Extended save area */
426 	rc = read_guest_lc(vcpu, __LC_MCESAD, &ext_sa_addr,
427 			   sizeof(unsigned long));
428 	/* Only bits 0 through 63-LC are used for address formation */
429 	lc = ext_sa_addr & MCESA_LC_MASK;
430 	if (test_kvm_facility(vcpu->kvm, 133)) {
431 		switch (lc) {
432 		case 0:
433 		case 10:
434 			ext_sa_addr &= ~0x3ffUL;
435 			break;
436 		case 11:
437 			ext_sa_addr &= ~0x7ffUL;
438 			break;
439 		case 12:
440 			ext_sa_addr &= ~0xfffUL;
441 			break;
442 		default:
443 			ext_sa_addr = 0;
444 			break;
445 		}
446 	} else {
447 		ext_sa_addr &= ~0x3ffUL;
448 	}
449 
450 	if (!rc && mci.vr && ext_sa_addr && test_kvm_facility(vcpu->kvm, 129)) {
451 		if (write_guest_abs(vcpu, ext_sa_addr, vcpu->run->s.regs.vrs,
452 				    512))
453 			mci.vr = 0;
454 	} else {
455 		mci.vr = 0;
456 	}
457 	if (!rc && mci.gs && ext_sa_addr && test_kvm_facility(vcpu->kvm, 133)
458 	    && (lc == 11 || lc == 12)) {
459 		if (write_guest_abs(vcpu, ext_sa_addr + 1024,
460 				    &vcpu->run->s.regs.gscb, 32))
461 			mci.gs = 0;
462 	} else {
463 		mci.gs = 0;
464 	}
465 
466 	/* General interruption information */
467 	rc |= put_guest_lc(vcpu, 1, (u8 __user *) __LC_AR_MODE_ID);
468 	rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
469 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
470 	rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
471 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
472 	rc |= put_guest_lc(vcpu, mci.val, (u64 __user *) __LC_MCCK_CODE);
473 
474 	/* Register-save areas */
475 	if (MACHINE_HAS_VX) {
476 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
477 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA, fprs, 128);
478 	} else {
479 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA,
480 				     vcpu->run->s.regs.fprs, 128);
481 	}
482 	rc |= write_guest_lc(vcpu, __LC_GPREGS_SAVE_AREA,
483 			     vcpu->run->s.regs.gprs, 128);
484 	rc |= put_guest_lc(vcpu, current->thread.fpu.fpc,
485 			   (u32 __user *) __LC_FP_CREG_SAVE_AREA);
486 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->todpr,
487 			   (u32 __user *) __LC_TOD_PROGREG_SAVE_AREA);
488 	rc |= put_guest_lc(vcpu, kvm_s390_get_cpu_timer(vcpu),
489 			   (u64 __user *) __LC_CPU_TIMER_SAVE_AREA);
490 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->ckc >> 8,
491 			   (u64 __user *) __LC_CLOCK_COMP_SAVE_AREA);
492 	rc |= write_guest_lc(vcpu, __LC_AREGS_SAVE_AREA,
493 			     &vcpu->run->s.regs.acrs, 64);
494 	rc |= write_guest_lc(vcpu, __LC_CREGS_SAVE_AREA,
495 			     &vcpu->arch.sie_block->gcr, 128);
496 
497 	/* Extended interruption information */
498 	rc |= put_guest_lc(vcpu, mchk->ext_damage_code,
499 			   (u32 __user *) __LC_EXT_DAMAGE_CODE);
500 	rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
501 			   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
502 	rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA, &mchk->fixed_logout,
503 			     sizeof(mchk->fixed_logout));
504 	return rc ? -EFAULT : 0;
505 }
506 
507 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
508 {
509 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
510 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
511 	struct kvm_s390_mchk_info mchk = {};
512 	int deliver = 0;
513 	int rc = 0;
514 
515 	spin_lock(&fi->lock);
516 	spin_lock(&li->lock);
517 	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
518 	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
519 		/*
520 		 * If there was an exigent machine check pending, then any
521 		 * repressible machine checks that might have been pending
522 		 * are indicated along with it, so always clear bits for
523 		 * repressible and exigent interrupts
524 		 */
525 		mchk = li->irq.mchk;
526 		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
527 		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
528 		memset(&li->irq.mchk, 0, sizeof(mchk));
529 		deliver = 1;
530 	}
531 	/*
532 	 * We indicate floating repressible conditions along with
533 	 * other pending conditions. Channel Report Pending and Channel
534 	 * Subsystem damage are the only two and and are indicated by
535 	 * bits in mcic and masked in cr14.
536 	 */
537 	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
538 		mchk.mcic |= fi->mchk.mcic;
539 		mchk.cr14 |= fi->mchk.cr14;
540 		memset(&fi->mchk, 0, sizeof(mchk));
541 		deliver = 1;
542 	}
543 	spin_unlock(&li->lock);
544 	spin_unlock(&fi->lock);
545 
546 	if (deliver) {
547 		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
548 			   mchk.mcic);
549 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
550 						 KVM_S390_MCHK,
551 						 mchk.cr14, mchk.mcic);
552 		rc = __write_machine_check(vcpu, &mchk);
553 	}
554 	return rc;
555 }
556 
557 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
558 {
559 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
560 	int rc;
561 
562 	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
563 	vcpu->stat.deliver_restart_signal++;
564 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
565 
566 	rc  = write_guest_lc(vcpu,
567 			     offsetof(struct lowcore, restart_old_psw),
568 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
569 	rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
570 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
571 	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
572 	return rc ? -EFAULT : 0;
573 }
574 
575 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
576 {
577 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
578 	struct kvm_s390_prefix_info prefix;
579 
580 	spin_lock(&li->lock);
581 	prefix = li->irq.prefix;
582 	li->irq.prefix.address = 0;
583 	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
584 	spin_unlock(&li->lock);
585 
586 	vcpu->stat.deliver_prefix_signal++;
587 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
588 					 KVM_S390_SIGP_SET_PREFIX,
589 					 prefix.address, 0);
590 
591 	kvm_s390_set_prefix(vcpu, prefix.address);
592 	return 0;
593 }
594 
595 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
596 {
597 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
598 	int rc;
599 	int cpu_addr;
600 
601 	spin_lock(&li->lock);
602 	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
603 	clear_bit(cpu_addr, li->sigp_emerg_pending);
604 	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
605 		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
606 	spin_unlock(&li->lock);
607 
608 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
609 	vcpu->stat.deliver_emergency_signal++;
610 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
611 					 cpu_addr, 0);
612 
613 	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
614 			   (u16 *)__LC_EXT_INT_CODE);
615 	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
616 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
617 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
618 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
619 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
620 	return rc ? -EFAULT : 0;
621 }
622 
623 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
624 {
625 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
626 	struct kvm_s390_extcall_info extcall;
627 	int rc;
628 
629 	spin_lock(&li->lock);
630 	extcall = li->irq.extcall;
631 	li->irq.extcall.code = 0;
632 	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
633 	spin_unlock(&li->lock);
634 
635 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
636 	vcpu->stat.deliver_external_call++;
637 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
638 					 KVM_S390_INT_EXTERNAL_CALL,
639 					 extcall.code, 0);
640 
641 	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
642 			   (u16 *)__LC_EXT_INT_CODE);
643 	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
644 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
645 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
646 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
647 			    sizeof(psw_t));
648 	return rc ? -EFAULT : 0;
649 }
650 
651 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
652 {
653 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
654 	struct kvm_s390_pgm_info pgm_info;
655 	int rc = 0, nullifying = false;
656 	u16 ilen;
657 
658 	spin_lock(&li->lock);
659 	pgm_info = li->irq.pgm;
660 	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
661 	memset(&li->irq.pgm, 0, sizeof(pgm_info));
662 	spin_unlock(&li->lock);
663 
664 	ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
665 	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
666 		   pgm_info.code, ilen);
667 	vcpu->stat.deliver_program_int++;
668 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
669 					 pgm_info.code, 0);
670 
671 	switch (pgm_info.code & ~PGM_PER) {
672 	case PGM_AFX_TRANSLATION:
673 	case PGM_ASX_TRANSLATION:
674 	case PGM_EX_TRANSLATION:
675 	case PGM_LFX_TRANSLATION:
676 	case PGM_LSTE_SEQUENCE:
677 	case PGM_LSX_TRANSLATION:
678 	case PGM_LX_TRANSLATION:
679 	case PGM_PRIMARY_AUTHORITY:
680 	case PGM_SECONDARY_AUTHORITY:
681 		nullifying = true;
682 		/* fall through */
683 	case PGM_SPACE_SWITCH:
684 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
685 				  (u64 *)__LC_TRANS_EXC_CODE);
686 		break;
687 	case PGM_ALEN_TRANSLATION:
688 	case PGM_ALE_SEQUENCE:
689 	case PGM_ASTE_INSTANCE:
690 	case PGM_ASTE_SEQUENCE:
691 	case PGM_ASTE_VALIDITY:
692 	case PGM_EXTENDED_AUTHORITY:
693 		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
694 				  (u8 *)__LC_EXC_ACCESS_ID);
695 		nullifying = true;
696 		break;
697 	case PGM_ASCE_TYPE:
698 	case PGM_PAGE_TRANSLATION:
699 	case PGM_REGION_FIRST_TRANS:
700 	case PGM_REGION_SECOND_TRANS:
701 	case PGM_REGION_THIRD_TRANS:
702 	case PGM_SEGMENT_TRANSLATION:
703 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
704 				  (u64 *)__LC_TRANS_EXC_CODE);
705 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
706 				   (u8 *)__LC_EXC_ACCESS_ID);
707 		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
708 				   (u8 *)__LC_OP_ACCESS_ID);
709 		nullifying = true;
710 		break;
711 	case PGM_MONITOR:
712 		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
713 				  (u16 *)__LC_MON_CLASS_NR);
714 		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
715 				   (u64 *)__LC_MON_CODE);
716 		break;
717 	case PGM_VECTOR_PROCESSING:
718 	case PGM_DATA:
719 		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
720 				  (u32 *)__LC_DATA_EXC_CODE);
721 		break;
722 	case PGM_PROTECTION:
723 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
724 				  (u64 *)__LC_TRANS_EXC_CODE);
725 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
726 				   (u8 *)__LC_EXC_ACCESS_ID);
727 		break;
728 	case PGM_STACK_FULL:
729 	case PGM_STACK_EMPTY:
730 	case PGM_STACK_SPECIFICATION:
731 	case PGM_STACK_TYPE:
732 	case PGM_STACK_OPERATION:
733 	case PGM_TRACE_TABEL:
734 	case PGM_CRYPTO_OPERATION:
735 		nullifying = true;
736 		break;
737 	}
738 
739 	if (pgm_info.code & PGM_PER) {
740 		rc |= put_guest_lc(vcpu, pgm_info.per_code,
741 				   (u8 *) __LC_PER_CODE);
742 		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
743 				   (u8 *)__LC_PER_ATMID);
744 		rc |= put_guest_lc(vcpu, pgm_info.per_address,
745 				   (u64 *) __LC_PER_ADDRESS);
746 		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
747 				   (u8 *) __LC_PER_ACCESS_ID);
748 	}
749 
750 	if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
751 		kvm_s390_rewind_psw(vcpu, ilen);
752 
753 	/* bit 1+2 of the target are the ilc, so we can directly use ilen */
754 	rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
755 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
756 				 (u64 *) __LC_LAST_BREAK);
757 	rc |= put_guest_lc(vcpu, pgm_info.code,
758 			   (u16 *)__LC_PGM_INT_CODE);
759 	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
760 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
761 	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
762 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
763 	return rc ? -EFAULT : 0;
764 }
765 
766 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
767 {
768 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
769 	struct kvm_s390_ext_info ext;
770 	int rc = 0;
771 
772 	spin_lock(&fi->lock);
773 	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
774 		spin_unlock(&fi->lock);
775 		return 0;
776 	}
777 	ext = fi->srv_signal;
778 	memset(&fi->srv_signal, 0, sizeof(ext));
779 	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
780 	spin_unlock(&fi->lock);
781 
782 	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
783 		   ext.ext_params);
784 	vcpu->stat.deliver_service_signal++;
785 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
786 					 ext.ext_params, 0);
787 
788 	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
789 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
790 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
791 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
792 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
793 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
794 	rc |= put_guest_lc(vcpu, ext.ext_params,
795 			   (u32 *)__LC_EXT_PARAMS);
796 
797 	return rc ? -EFAULT : 0;
798 }
799 
800 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
801 {
802 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
803 	struct kvm_s390_interrupt_info *inti;
804 	int rc = 0;
805 
806 	spin_lock(&fi->lock);
807 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
808 					struct kvm_s390_interrupt_info,
809 					list);
810 	if (inti) {
811 		list_del(&inti->list);
812 		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
813 	}
814 	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
815 		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
816 	spin_unlock(&fi->lock);
817 
818 	if (inti) {
819 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
820 						 KVM_S390_INT_PFAULT_DONE, 0,
821 						 inti->ext.ext_params2);
822 		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
823 			   inti->ext.ext_params2);
824 
825 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
826 				(u16 *)__LC_EXT_INT_CODE);
827 		rc |= put_guest_lc(vcpu, PFAULT_DONE,
828 				(u16 *)__LC_EXT_CPU_ADDR);
829 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
830 				&vcpu->arch.sie_block->gpsw,
831 				sizeof(psw_t));
832 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
833 				&vcpu->arch.sie_block->gpsw,
834 				sizeof(psw_t));
835 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
836 				(u64 *)__LC_EXT_PARAMS2);
837 		kfree(inti);
838 	}
839 	return rc ? -EFAULT : 0;
840 }
841 
842 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
843 {
844 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
845 	struct kvm_s390_interrupt_info *inti;
846 	int rc = 0;
847 
848 	spin_lock(&fi->lock);
849 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
850 					struct kvm_s390_interrupt_info,
851 					list);
852 	if (inti) {
853 		VCPU_EVENT(vcpu, 4,
854 			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
855 			   inti->ext.ext_params, inti->ext.ext_params2);
856 		vcpu->stat.deliver_virtio_interrupt++;
857 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
858 				inti->type,
859 				inti->ext.ext_params,
860 				inti->ext.ext_params2);
861 		list_del(&inti->list);
862 		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
863 	}
864 	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
865 		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
866 	spin_unlock(&fi->lock);
867 
868 	if (inti) {
869 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
870 				(u16 *)__LC_EXT_INT_CODE);
871 		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
872 				(u16 *)__LC_EXT_CPU_ADDR);
873 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
874 				&vcpu->arch.sie_block->gpsw,
875 				sizeof(psw_t));
876 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
877 				&vcpu->arch.sie_block->gpsw,
878 				sizeof(psw_t));
879 		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
880 				(u32 *)__LC_EXT_PARAMS);
881 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
882 				(u64 *)__LC_EXT_PARAMS2);
883 		kfree(inti);
884 	}
885 	return rc ? -EFAULT : 0;
886 }
887 
888 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
889 				     unsigned long irq_type)
890 {
891 	struct list_head *isc_list;
892 	struct kvm_s390_float_interrupt *fi;
893 	struct kvm_s390_interrupt_info *inti = NULL;
894 	int rc = 0;
895 
896 	fi = &vcpu->kvm->arch.float_int;
897 
898 	spin_lock(&fi->lock);
899 	isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
900 	inti = list_first_entry_or_null(isc_list,
901 					struct kvm_s390_interrupt_info,
902 					list);
903 	if (inti) {
904 		if (inti->type & KVM_S390_INT_IO_AI_MASK)
905 			VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
906 		else
907 			VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
908 			inti->io.subchannel_id >> 8,
909 			inti->io.subchannel_id >> 1 & 0x3,
910 			inti->io.subchannel_nr);
911 
912 		vcpu->stat.deliver_io_int++;
913 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
914 				inti->type,
915 				((__u32)inti->io.subchannel_id << 16) |
916 				inti->io.subchannel_nr,
917 				((__u64)inti->io.io_int_parm << 32) |
918 				inti->io.io_int_word);
919 		list_del(&inti->list);
920 		fi->counters[FIRQ_CNTR_IO] -= 1;
921 	}
922 	if (list_empty(isc_list))
923 		clear_bit(irq_type, &fi->pending_irqs);
924 	spin_unlock(&fi->lock);
925 
926 	if (inti) {
927 		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
928 				(u16 *)__LC_SUBCHANNEL_ID);
929 		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
930 				(u16 *)__LC_SUBCHANNEL_NR);
931 		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
932 				(u32 *)__LC_IO_INT_PARM);
933 		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
934 				(u32 *)__LC_IO_INT_WORD);
935 		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
936 				&vcpu->arch.sie_block->gpsw,
937 				sizeof(psw_t));
938 		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
939 				&vcpu->arch.sie_block->gpsw,
940 				sizeof(psw_t));
941 		kfree(inti);
942 	}
943 
944 	return rc ? -EFAULT : 0;
945 }
946 
947 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
948 
949 static const deliver_irq_t deliver_irq_funcs[] = {
950 	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
951 	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
952 	[IRQ_PEND_PROG]           = __deliver_prog,
953 	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
954 	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
955 	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
956 	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
957 	[IRQ_PEND_RESTART]        = __deliver_restart,
958 	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
959 	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
960 	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
961 	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
962 	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
963 };
964 
965 /* Check whether an external call is pending (deliverable or not) */
966 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
967 {
968 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
969 
970 	if (!sclp.has_sigpif)
971 		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
972 
973 	return sca_ext_call_pending(vcpu, NULL);
974 }
975 
976 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
977 {
978 	if (deliverable_irqs(vcpu))
979 		return 1;
980 
981 	if (kvm_cpu_has_pending_timer(vcpu))
982 		return 1;
983 
984 	/* external call pending and deliverable */
985 	if (kvm_s390_ext_call_pending(vcpu) &&
986 	    !psw_extint_disabled(vcpu) &&
987 	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
988 		return 1;
989 
990 	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
991 		return 1;
992 	return 0;
993 }
994 
995 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
996 {
997 	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
998 }
999 
1000 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
1001 {
1002 	u64 now, cputm, sltime = 0;
1003 
1004 	if (ckc_interrupts_enabled(vcpu)) {
1005 		now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
1006 		sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
1007 		/* already expired or overflow? */
1008 		if (!sltime || vcpu->arch.sie_block->ckc <= now)
1009 			return 0;
1010 		if (cpu_timer_interrupts_enabled(vcpu)) {
1011 			cputm = kvm_s390_get_cpu_timer(vcpu);
1012 			/* already expired? */
1013 			if (cputm >> 63)
1014 				return 0;
1015 			return min(sltime, tod_to_ns(cputm));
1016 		}
1017 	} else if (cpu_timer_interrupts_enabled(vcpu)) {
1018 		sltime = kvm_s390_get_cpu_timer(vcpu);
1019 		/* already expired? */
1020 		if (sltime >> 63)
1021 			return 0;
1022 	}
1023 	return sltime;
1024 }
1025 
1026 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
1027 {
1028 	u64 sltime;
1029 
1030 	vcpu->stat.exit_wait_state++;
1031 
1032 	/* fast path */
1033 	if (kvm_arch_vcpu_runnable(vcpu))
1034 		return 0;
1035 
1036 	if (psw_interrupts_disabled(vcpu)) {
1037 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
1038 		return -EOPNOTSUPP; /* disabled wait */
1039 	}
1040 
1041 	if (!ckc_interrupts_enabled(vcpu) &&
1042 	    !cpu_timer_interrupts_enabled(vcpu)) {
1043 		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
1044 		__set_cpu_idle(vcpu);
1045 		goto no_timer;
1046 	}
1047 
1048 	sltime = __calculate_sltime(vcpu);
1049 	if (!sltime)
1050 		return 0;
1051 
1052 	__set_cpu_idle(vcpu);
1053 	hrtimer_start(&vcpu->arch.ckc_timer, sltime, HRTIMER_MODE_REL);
1054 	VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
1055 no_timer:
1056 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1057 	kvm_vcpu_block(vcpu);
1058 	__unset_cpu_idle(vcpu);
1059 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1060 
1061 	hrtimer_cancel(&vcpu->arch.ckc_timer);
1062 	return 0;
1063 }
1064 
1065 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
1066 {
1067 	/*
1068 	 * We cannot move this into the if, as the CPU might be already
1069 	 * in kvm_vcpu_block without having the waitqueue set (polling)
1070 	 */
1071 	vcpu->valid_wakeup = true;
1072 	if (swait_active(&vcpu->wq)) {
1073 		/*
1074 		 * The vcpu gave up the cpu voluntarily, mark it as a good
1075 		 * yield-candidate.
1076 		 */
1077 		vcpu->preempted = true;
1078 		swake_up(&vcpu->wq);
1079 		vcpu->stat.halt_wakeup++;
1080 	}
1081 	/*
1082 	 * The VCPU might not be sleeping but is executing the VSIE. Let's
1083 	 * kick it, so it leaves the SIE to process the request.
1084 	 */
1085 	kvm_s390_vsie_kick(vcpu);
1086 }
1087 
1088 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1089 {
1090 	struct kvm_vcpu *vcpu;
1091 	u64 sltime;
1092 
1093 	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1094 	sltime = __calculate_sltime(vcpu);
1095 
1096 	/*
1097 	 * If the monotonic clock runs faster than the tod clock we might be
1098 	 * woken up too early and have to go back to sleep to avoid deadlocks.
1099 	 */
1100 	if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1101 		return HRTIMER_RESTART;
1102 	kvm_s390_vcpu_wakeup(vcpu);
1103 	return HRTIMER_NORESTART;
1104 }
1105 
1106 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1107 {
1108 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1109 
1110 	spin_lock(&li->lock);
1111 	li->pending_irqs = 0;
1112 	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1113 	memset(&li->irq, 0, sizeof(li->irq));
1114 	spin_unlock(&li->lock);
1115 
1116 	sca_clear_ext_call(vcpu);
1117 }
1118 
1119 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1120 {
1121 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1122 	deliver_irq_t func;
1123 	int rc = 0;
1124 	unsigned long irq_type;
1125 	unsigned long irqs;
1126 
1127 	__reset_intercept_indicators(vcpu);
1128 
1129 	/* pending ckc conditions might have been invalidated */
1130 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1131 	if (ckc_irq_pending(vcpu))
1132 		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1133 
1134 	/* pending cpu timer conditions might have been invalidated */
1135 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1136 	if (cpu_timer_irq_pending(vcpu))
1137 		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1138 
1139 	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1140 		/* bits are in the order of interrupt priority */
1141 		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1142 		if (is_ioirq(irq_type)) {
1143 			rc = __deliver_io(vcpu, irq_type);
1144 		} else {
1145 			func = deliver_irq_funcs[irq_type];
1146 			if (!func) {
1147 				WARN_ON_ONCE(func == NULL);
1148 				clear_bit(irq_type, &li->pending_irqs);
1149 				continue;
1150 			}
1151 			rc = func(vcpu);
1152 		}
1153 	}
1154 
1155 	set_intercept_indicators(vcpu);
1156 
1157 	return rc;
1158 }
1159 
1160 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1161 {
1162 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1163 
1164 	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1165 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1166 				   irq->u.pgm.code, 0);
1167 
1168 	if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1169 		/* auto detection if no valid ILC was given */
1170 		irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1171 		irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1172 		irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1173 	}
1174 
1175 	if (irq->u.pgm.code == PGM_PER) {
1176 		li->irq.pgm.code |= PGM_PER;
1177 		li->irq.pgm.flags = irq->u.pgm.flags;
1178 		/* only modify PER related information */
1179 		li->irq.pgm.per_address = irq->u.pgm.per_address;
1180 		li->irq.pgm.per_code = irq->u.pgm.per_code;
1181 		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1182 		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1183 	} else if (!(irq->u.pgm.code & PGM_PER)) {
1184 		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1185 				   irq->u.pgm.code;
1186 		li->irq.pgm.flags = irq->u.pgm.flags;
1187 		/* only modify non-PER information */
1188 		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1189 		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1190 		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1191 		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1192 		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1193 		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1194 	} else {
1195 		li->irq.pgm = irq->u.pgm;
1196 	}
1197 	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1198 	return 0;
1199 }
1200 
1201 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1202 {
1203 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1204 
1205 	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1206 		   irq->u.ext.ext_params2);
1207 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1208 				   irq->u.ext.ext_params,
1209 				   irq->u.ext.ext_params2);
1210 
1211 	li->irq.ext = irq->u.ext;
1212 	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1213 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1214 	return 0;
1215 }
1216 
1217 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1218 {
1219 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1220 	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1221 	uint16_t src_id = irq->u.extcall.code;
1222 
1223 	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1224 		   src_id);
1225 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1226 				   src_id, 0);
1227 
1228 	/* sending vcpu invalid */
1229 	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1230 		return -EINVAL;
1231 
1232 	if (sclp.has_sigpif)
1233 		return sca_inject_ext_call(vcpu, src_id);
1234 
1235 	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1236 		return -EBUSY;
1237 	*extcall = irq->u.extcall;
1238 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1239 	return 0;
1240 }
1241 
1242 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1243 {
1244 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1245 	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1246 
1247 	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1248 		   irq->u.prefix.address);
1249 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1250 				   irq->u.prefix.address, 0);
1251 
1252 	if (!is_vcpu_stopped(vcpu))
1253 		return -EBUSY;
1254 
1255 	*prefix = irq->u.prefix;
1256 	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1257 	return 0;
1258 }
1259 
1260 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1261 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1262 {
1263 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1264 	struct kvm_s390_stop_info *stop = &li->irq.stop;
1265 	int rc = 0;
1266 
1267 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1268 
1269 	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1270 		return -EINVAL;
1271 
1272 	if (is_vcpu_stopped(vcpu)) {
1273 		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1274 			rc = kvm_s390_store_status_unloaded(vcpu,
1275 						KVM_S390_STORE_STATUS_NOADDR);
1276 		return rc;
1277 	}
1278 
1279 	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1280 		return -EBUSY;
1281 	stop->flags = irq->u.stop.flags;
1282 	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1283 	return 0;
1284 }
1285 
1286 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1287 				 struct kvm_s390_irq *irq)
1288 {
1289 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1290 
1291 	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1292 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1293 
1294 	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1295 	return 0;
1296 }
1297 
1298 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1299 				   struct kvm_s390_irq *irq)
1300 {
1301 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1302 
1303 	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1304 		   irq->u.emerg.code);
1305 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1306 				   irq->u.emerg.code, 0);
1307 
1308 	/* sending vcpu invalid */
1309 	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1310 		return -EINVAL;
1311 
1312 	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1313 	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1314 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1315 	return 0;
1316 }
1317 
1318 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1319 {
1320 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1321 	struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1322 
1323 	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1324 		   irq->u.mchk.mcic);
1325 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1326 				   irq->u.mchk.mcic);
1327 
1328 	/*
1329 	 * Because repressible machine checks can be indicated along with
1330 	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1331 	 * we need to combine cr14, mcic and external damage code.
1332 	 * Failing storage address and the logout area should not be or'ed
1333 	 * together, we just indicate the last occurrence of the corresponding
1334 	 * machine check
1335 	 */
1336 	mchk->cr14 |= irq->u.mchk.cr14;
1337 	mchk->mcic |= irq->u.mchk.mcic;
1338 	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1339 	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1340 	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1341 	       sizeof(mchk->fixed_logout));
1342 	if (mchk->mcic & MCHK_EX_MASK)
1343 		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1344 	else if (mchk->mcic & MCHK_REP_MASK)
1345 		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1346 	return 0;
1347 }
1348 
1349 static int __inject_ckc(struct kvm_vcpu *vcpu)
1350 {
1351 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1352 
1353 	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1354 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1355 				   0, 0);
1356 
1357 	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1358 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1359 	return 0;
1360 }
1361 
1362 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1363 {
1364 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1365 
1366 	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1367 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1368 				   0, 0);
1369 
1370 	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1371 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1372 	return 0;
1373 }
1374 
1375 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1376 						  int isc, u32 schid)
1377 {
1378 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1379 	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1380 	struct kvm_s390_interrupt_info *iter;
1381 	u16 id = (schid & 0xffff0000U) >> 16;
1382 	u16 nr = schid & 0x0000ffffU;
1383 
1384 	spin_lock(&fi->lock);
1385 	list_for_each_entry(iter, isc_list, list) {
1386 		if (schid && (id != iter->io.subchannel_id ||
1387 			      nr != iter->io.subchannel_nr))
1388 			continue;
1389 		/* found an appropriate entry */
1390 		list_del_init(&iter->list);
1391 		fi->counters[FIRQ_CNTR_IO] -= 1;
1392 		if (list_empty(isc_list))
1393 			clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1394 		spin_unlock(&fi->lock);
1395 		return iter;
1396 	}
1397 	spin_unlock(&fi->lock);
1398 	return NULL;
1399 }
1400 
1401 /*
1402  * Dequeue and return an I/O interrupt matching any of the interruption
1403  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1404  */
1405 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1406 						    u64 isc_mask, u32 schid)
1407 {
1408 	struct kvm_s390_interrupt_info *inti = NULL;
1409 	int isc;
1410 
1411 	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1412 		if (isc_mask & isc_to_isc_bits(isc))
1413 			inti = get_io_int(kvm, isc, schid);
1414 	}
1415 	return inti;
1416 }
1417 
1418 #define SCCB_MASK 0xFFFFFFF8
1419 #define SCCB_EVENT_PENDING 0x3
1420 
1421 static int __inject_service(struct kvm *kvm,
1422 			     struct kvm_s390_interrupt_info *inti)
1423 {
1424 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1425 
1426 	spin_lock(&fi->lock);
1427 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1428 	/*
1429 	 * Early versions of the QEMU s390 bios will inject several
1430 	 * service interrupts after another without handling a
1431 	 * condition code indicating busy.
1432 	 * We will silently ignore those superfluous sccb values.
1433 	 * A future version of QEMU will take care of serialization
1434 	 * of servc requests
1435 	 */
1436 	if (fi->srv_signal.ext_params & SCCB_MASK)
1437 		goto out;
1438 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1439 	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1440 out:
1441 	spin_unlock(&fi->lock);
1442 	kfree(inti);
1443 	return 0;
1444 }
1445 
1446 static int __inject_virtio(struct kvm *kvm,
1447 			    struct kvm_s390_interrupt_info *inti)
1448 {
1449 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1450 
1451 	spin_lock(&fi->lock);
1452 	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1453 		spin_unlock(&fi->lock);
1454 		return -EBUSY;
1455 	}
1456 	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1457 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1458 	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1459 	spin_unlock(&fi->lock);
1460 	return 0;
1461 }
1462 
1463 static int __inject_pfault_done(struct kvm *kvm,
1464 				 struct kvm_s390_interrupt_info *inti)
1465 {
1466 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1467 
1468 	spin_lock(&fi->lock);
1469 	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1470 		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1471 		spin_unlock(&fi->lock);
1472 		return -EBUSY;
1473 	}
1474 	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1475 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1476 	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1477 	spin_unlock(&fi->lock);
1478 	return 0;
1479 }
1480 
1481 #define CR_PENDING_SUBCLASS 28
1482 static int __inject_float_mchk(struct kvm *kvm,
1483 				struct kvm_s390_interrupt_info *inti)
1484 {
1485 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1486 
1487 	spin_lock(&fi->lock);
1488 	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1489 	fi->mchk.mcic |= inti->mchk.mcic;
1490 	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1491 	spin_unlock(&fi->lock);
1492 	kfree(inti);
1493 	return 0;
1494 }
1495 
1496 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1497 {
1498 	struct kvm_s390_float_interrupt *fi;
1499 	struct list_head *list;
1500 	int isc;
1501 
1502 	fi = &kvm->arch.float_int;
1503 	spin_lock(&fi->lock);
1504 	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1505 		spin_unlock(&fi->lock);
1506 		return -EBUSY;
1507 	}
1508 	fi->counters[FIRQ_CNTR_IO] += 1;
1509 
1510 	if (inti->type & KVM_S390_INT_IO_AI_MASK)
1511 		VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1512 	else
1513 		VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1514 			inti->io.subchannel_id >> 8,
1515 			inti->io.subchannel_id >> 1 & 0x3,
1516 			inti->io.subchannel_nr);
1517 	isc = int_word_to_isc(inti->io.io_int_word);
1518 	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1519 	list_add_tail(&inti->list, list);
1520 	set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1521 	spin_unlock(&fi->lock);
1522 	return 0;
1523 }
1524 
1525 /*
1526  * Find a destination VCPU for a floating irq and kick it.
1527  */
1528 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1529 {
1530 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1531 	struct kvm_s390_local_interrupt *li;
1532 	struct kvm_vcpu *dst_vcpu;
1533 	int sigcpu, online_vcpus, nr_tries = 0;
1534 
1535 	online_vcpus = atomic_read(&kvm->online_vcpus);
1536 	if (!online_vcpus)
1537 		return;
1538 
1539 	/* find idle VCPUs first, then round robin */
1540 	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1541 	if (sigcpu == online_vcpus) {
1542 		do {
1543 			sigcpu = fi->next_rr_cpu;
1544 			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1545 			/* avoid endless loops if all vcpus are stopped */
1546 			if (nr_tries++ >= online_vcpus)
1547 				return;
1548 		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1549 	}
1550 	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1551 
1552 	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1553 	li = &dst_vcpu->arch.local_int;
1554 	spin_lock(&li->lock);
1555 	switch (type) {
1556 	case KVM_S390_MCHK:
1557 		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1558 		break;
1559 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1560 		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1561 		break;
1562 	default:
1563 		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1564 		break;
1565 	}
1566 	spin_unlock(&li->lock);
1567 	kvm_s390_vcpu_wakeup(dst_vcpu);
1568 }
1569 
1570 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1571 {
1572 	u64 type = READ_ONCE(inti->type);
1573 	int rc;
1574 
1575 	switch (type) {
1576 	case KVM_S390_MCHK:
1577 		rc = __inject_float_mchk(kvm, inti);
1578 		break;
1579 	case KVM_S390_INT_VIRTIO:
1580 		rc = __inject_virtio(kvm, inti);
1581 		break;
1582 	case KVM_S390_INT_SERVICE:
1583 		rc = __inject_service(kvm, inti);
1584 		break;
1585 	case KVM_S390_INT_PFAULT_DONE:
1586 		rc = __inject_pfault_done(kvm, inti);
1587 		break;
1588 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1589 		rc = __inject_io(kvm, inti);
1590 		break;
1591 	default:
1592 		rc = -EINVAL;
1593 	}
1594 	if (rc)
1595 		return rc;
1596 
1597 	__floating_irq_kick(kvm, type);
1598 	return 0;
1599 }
1600 
1601 int kvm_s390_inject_vm(struct kvm *kvm,
1602 		       struct kvm_s390_interrupt *s390int)
1603 {
1604 	struct kvm_s390_interrupt_info *inti;
1605 	int rc;
1606 
1607 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1608 	if (!inti)
1609 		return -ENOMEM;
1610 
1611 	inti->type = s390int->type;
1612 	switch (inti->type) {
1613 	case KVM_S390_INT_VIRTIO:
1614 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1615 			 s390int->parm, s390int->parm64);
1616 		inti->ext.ext_params = s390int->parm;
1617 		inti->ext.ext_params2 = s390int->parm64;
1618 		break;
1619 	case KVM_S390_INT_SERVICE:
1620 		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1621 		inti->ext.ext_params = s390int->parm;
1622 		break;
1623 	case KVM_S390_INT_PFAULT_DONE:
1624 		inti->ext.ext_params2 = s390int->parm64;
1625 		break;
1626 	case KVM_S390_MCHK:
1627 		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1628 			 s390int->parm64);
1629 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1630 		inti->mchk.mcic = s390int->parm64;
1631 		break;
1632 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1633 		inti->io.subchannel_id = s390int->parm >> 16;
1634 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1635 		inti->io.io_int_parm = s390int->parm64 >> 32;
1636 		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1637 		break;
1638 	default:
1639 		kfree(inti);
1640 		return -EINVAL;
1641 	}
1642 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1643 				 2);
1644 
1645 	rc = __inject_vm(kvm, inti);
1646 	if (rc)
1647 		kfree(inti);
1648 	return rc;
1649 }
1650 
1651 int kvm_s390_reinject_io_int(struct kvm *kvm,
1652 			      struct kvm_s390_interrupt_info *inti)
1653 {
1654 	return __inject_vm(kvm, inti);
1655 }
1656 
1657 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1658 		       struct kvm_s390_irq *irq)
1659 {
1660 	irq->type = s390int->type;
1661 	switch (irq->type) {
1662 	case KVM_S390_PROGRAM_INT:
1663 		if (s390int->parm & 0xffff0000)
1664 			return -EINVAL;
1665 		irq->u.pgm.code = s390int->parm;
1666 		break;
1667 	case KVM_S390_SIGP_SET_PREFIX:
1668 		irq->u.prefix.address = s390int->parm;
1669 		break;
1670 	case KVM_S390_SIGP_STOP:
1671 		irq->u.stop.flags = s390int->parm;
1672 		break;
1673 	case KVM_S390_INT_EXTERNAL_CALL:
1674 		if (s390int->parm & 0xffff0000)
1675 			return -EINVAL;
1676 		irq->u.extcall.code = s390int->parm;
1677 		break;
1678 	case KVM_S390_INT_EMERGENCY:
1679 		if (s390int->parm & 0xffff0000)
1680 			return -EINVAL;
1681 		irq->u.emerg.code = s390int->parm;
1682 		break;
1683 	case KVM_S390_MCHK:
1684 		irq->u.mchk.mcic = s390int->parm64;
1685 		break;
1686 	}
1687 	return 0;
1688 }
1689 
1690 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1691 {
1692 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1693 
1694 	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1695 }
1696 
1697 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1698 {
1699 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1700 
1701 	spin_lock(&li->lock);
1702 	li->irq.stop.flags = 0;
1703 	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1704 	spin_unlock(&li->lock);
1705 }
1706 
1707 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1708 {
1709 	int rc;
1710 
1711 	switch (irq->type) {
1712 	case KVM_S390_PROGRAM_INT:
1713 		rc = __inject_prog(vcpu, irq);
1714 		break;
1715 	case KVM_S390_SIGP_SET_PREFIX:
1716 		rc = __inject_set_prefix(vcpu, irq);
1717 		break;
1718 	case KVM_S390_SIGP_STOP:
1719 		rc = __inject_sigp_stop(vcpu, irq);
1720 		break;
1721 	case KVM_S390_RESTART:
1722 		rc = __inject_sigp_restart(vcpu, irq);
1723 		break;
1724 	case KVM_S390_INT_CLOCK_COMP:
1725 		rc = __inject_ckc(vcpu);
1726 		break;
1727 	case KVM_S390_INT_CPU_TIMER:
1728 		rc = __inject_cpu_timer(vcpu);
1729 		break;
1730 	case KVM_S390_INT_EXTERNAL_CALL:
1731 		rc = __inject_extcall(vcpu, irq);
1732 		break;
1733 	case KVM_S390_INT_EMERGENCY:
1734 		rc = __inject_sigp_emergency(vcpu, irq);
1735 		break;
1736 	case KVM_S390_MCHK:
1737 		rc = __inject_mchk(vcpu, irq);
1738 		break;
1739 	case KVM_S390_INT_PFAULT_INIT:
1740 		rc = __inject_pfault_init(vcpu, irq);
1741 		break;
1742 	case KVM_S390_INT_VIRTIO:
1743 	case KVM_S390_INT_SERVICE:
1744 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1745 	default:
1746 		rc = -EINVAL;
1747 	}
1748 
1749 	return rc;
1750 }
1751 
1752 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1753 {
1754 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1755 	int rc;
1756 
1757 	spin_lock(&li->lock);
1758 	rc = do_inject_vcpu(vcpu, irq);
1759 	spin_unlock(&li->lock);
1760 	if (!rc)
1761 		kvm_s390_vcpu_wakeup(vcpu);
1762 	return rc;
1763 }
1764 
1765 static inline void clear_irq_list(struct list_head *_list)
1766 {
1767 	struct kvm_s390_interrupt_info *inti, *n;
1768 
1769 	list_for_each_entry_safe(inti, n, _list, list) {
1770 		list_del(&inti->list);
1771 		kfree(inti);
1772 	}
1773 }
1774 
1775 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1776 		       struct kvm_s390_irq *irq)
1777 {
1778 	irq->type = inti->type;
1779 	switch (inti->type) {
1780 	case KVM_S390_INT_PFAULT_INIT:
1781 	case KVM_S390_INT_PFAULT_DONE:
1782 	case KVM_S390_INT_VIRTIO:
1783 		irq->u.ext = inti->ext;
1784 		break;
1785 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1786 		irq->u.io = inti->io;
1787 		break;
1788 	}
1789 }
1790 
1791 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1792 {
1793 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1794 	int i;
1795 
1796 	spin_lock(&fi->lock);
1797 	fi->pending_irqs = 0;
1798 	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1799 	memset(&fi->mchk, 0, sizeof(fi->mchk));
1800 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1801 		clear_irq_list(&fi->lists[i]);
1802 	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1803 		fi->counters[i] = 0;
1804 	spin_unlock(&fi->lock);
1805 };
1806 
1807 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1808 {
1809 	struct kvm_s390_interrupt_info *inti;
1810 	struct kvm_s390_float_interrupt *fi;
1811 	struct kvm_s390_irq *buf;
1812 	struct kvm_s390_irq *irq;
1813 	int max_irqs;
1814 	int ret = 0;
1815 	int n = 0;
1816 	int i;
1817 
1818 	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1819 		return -EINVAL;
1820 
1821 	/*
1822 	 * We are already using -ENOMEM to signal
1823 	 * userspace it may retry with a bigger buffer,
1824 	 * so we need to use something else for this case
1825 	 */
1826 	buf = vzalloc(len);
1827 	if (!buf)
1828 		return -ENOBUFS;
1829 
1830 	max_irqs = len / sizeof(struct kvm_s390_irq);
1831 
1832 	fi = &kvm->arch.float_int;
1833 	spin_lock(&fi->lock);
1834 	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1835 		list_for_each_entry(inti, &fi->lists[i], list) {
1836 			if (n == max_irqs) {
1837 				/* signal userspace to try again */
1838 				ret = -ENOMEM;
1839 				goto out;
1840 			}
1841 			inti_to_irq(inti, &buf[n]);
1842 			n++;
1843 		}
1844 	}
1845 	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1846 		if (n == max_irqs) {
1847 			/* signal userspace to try again */
1848 			ret = -ENOMEM;
1849 			goto out;
1850 		}
1851 		irq = (struct kvm_s390_irq *) &buf[n];
1852 		irq->type = KVM_S390_INT_SERVICE;
1853 		irq->u.ext = fi->srv_signal;
1854 		n++;
1855 	}
1856 	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1857 		if (n == max_irqs) {
1858 				/* signal userspace to try again */
1859 				ret = -ENOMEM;
1860 				goto out;
1861 		}
1862 		irq = (struct kvm_s390_irq *) &buf[n];
1863 		irq->type = KVM_S390_MCHK;
1864 		irq->u.mchk = fi->mchk;
1865 		n++;
1866 }
1867 
1868 out:
1869 	spin_unlock(&fi->lock);
1870 	if (!ret && n > 0) {
1871 		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1872 			ret = -EFAULT;
1873 	}
1874 	vfree(buf);
1875 
1876 	return ret < 0 ? ret : n;
1877 }
1878 
1879 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1880 {
1881 	int r;
1882 
1883 	switch (attr->group) {
1884 	case KVM_DEV_FLIC_GET_ALL_IRQS:
1885 		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1886 					  attr->attr);
1887 		break;
1888 	default:
1889 		r = -EINVAL;
1890 	}
1891 
1892 	return r;
1893 }
1894 
1895 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1896 				     u64 addr)
1897 {
1898 	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1899 	void *target = NULL;
1900 	void __user *source;
1901 	u64 size;
1902 
1903 	if (get_user(inti->type, (u64 __user *)addr))
1904 		return -EFAULT;
1905 
1906 	switch (inti->type) {
1907 	case KVM_S390_INT_PFAULT_INIT:
1908 	case KVM_S390_INT_PFAULT_DONE:
1909 	case KVM_S390_INT_VIRTIO:
1910 	case KVM_S390_INT_SERVICE:
1911 		target = (void *) &inti->ext;
1912 		source = &uptr->u.ext;
1913 		size = sizeof(inti->ext);
1914 		break;
1915 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1916 		target = (void *) &inti->io;
1917 		source = &uptr->u.io;
1918 		size = sizeof(inti->io);
1919 		break;
1920 	case KVM_S390_MCHK:
1921 		target = (void *) &inti->mchk;
1922 		source = &uptr->u.mchk;
1923 		size = sizeof(inti->mchk);
1924 		break;
1925 	default:
1926 		return -EINVAL;
1927 	}
1928 
1929 	if (copy_from_user(target, source, size))
1930 		return -EFAULT;
1931 
1932 	return 0;
1933 }
1934 
1935 static int enqueue_floating_irq(struct kvm_device *dev,
1936 				struct kvm_device_attr *attr)
1937 {
1938 	struct kvm_s390_interrupt_info *inti = NULL;
1939 	int r = 0;
1940 	int len = attr->attr;
1941 
1942 	if (len % sizeof(struct kvm_s390_irq) != 0)
1943 		return -EINVAL;
1944 	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1945 		return -EINVAL;
1946 
1947 	while (len >= sizeof(struct kvm_s390_irq)) {
1948 		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1949 		if (!inti)
1950 			return -ENOMEM;
1951 
1952 		r = copy_irq_from_user(inti, attr->addr);
1953 		if (r) {
1954 			kfree(inti);
1955 			return r;
1956 		}
1957 		r = __inject_vm(dev->kvm, inti);
1958 		if (r) {
1959 			kfree(inti);
1960 			return r;
1961 		}
1962 		len -= sizeof(struct kvm_s390_irq);
1963 		attr->addr += sizeof(struct kvm_s390_irq);
1964 	}
1965 
1966 	return r;
1967 }
1968 
1969 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1970 {
1971 	if (id >= MAX_S390_IO_ADAPTERS)
1972 		return NULL;
1973 	return kvm->arch.adapters[id];
1974 }
1975 
1976 static int register_io_adapter(struct kvm_device *dev,
1977 			       struct kvm_device_attr *attr)
1978 {
1979 	struct s390_io_adapter *adapter;
1980 	struct kvm_s390_io_adapter adapter_info;
1981 
1982 	if (copy_from_user(&adapter_info,
1983 			   (void __user *)attr->addr, sizeof(adapter_info)))
1984 		return -EFAULT;
1985 
1986 	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1987 	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1988 		return -EINVAL;
1989 
1990 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1991 	if (!adapter)
1992 		return -ENOMEM;
1993 
1994 	INIT_LIST_HEAD(&adapter->maps);
1995 	init_rwsem(&adapter->maps_lock);
1996 	atomic_set(&adapter->nr_maps, 0);
1997 	adapter->id = adapter_info.id;
1998 	adapter->isc = adapter_info.isc;
1999 	adapter->maskable = adapter_info.maskable;
2000 	adapter->masked = false;
2001 	adapter->swap = adapter_info.swap;
2002 	adapter->suppressible = (adapter_info.flags) &
2003 				KVM_S390_ADAPTER_SUPPRESSIBLE;
2004 	dev->kvm->arch.adapters[adapter->id] = adapter;
2005 
2006 	return 0;
2007 }
2008 
2009 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
2010 {
2011 	int ret;
2012 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2013 
2014 	if (!adapter || !adapter->maskable)
2015 		return -EINVAL;
2016 	ret = adapter->masked;
2017 	adapter->masked = masked;
2018 	return ret;
2019 }
2020 
2021 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
2022 {
2023 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2024 	struct s390_map_info *map;
2025 	int ret;
2026 
2027 	if (!adapter || !addr)
2028 		return -EINVAL;
2029 
2030 	map = kzalloc(sizeof(*map), GFP_KERNEL);
2031 	if (!map) {
2032 		ret = -ENOMEM;
2033 		goto out;
2034 	}
2035 	INIT_LIST_HEAD(&map->list);
2036 	map->guest_addr = addr;
2037 	map->addr = gmap_translate(kvm->arch.gmap, addr);
2038 	if (map->addr == -EFAULT) {
2039 		ret = -EFAULT;
2040 		goto out;
2041 	}
2042 	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
2043 	if (ret < 0)
2044 		goto out;
2045 	BUG_ON(ret != 1);
2046 	down_write(&adapter->maps_lock);
2047 	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
2048 		list_add_tail(&map->list, &adapter->maps);
2049 		ret = 0;
2050 	} else {
2051 		put_page(map->page);
2052 		ret = -EINVAL;
2053 	}
2054 	up_write(&adapter->maps_lock);
2055 out:
2056 	if (ret)
2057 		kfree(map);
2058 	return ret;
2059 }
2060 
2061 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
2062 {
2063 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2064 	struct s390_map_info *map, *tmp;
2065 	int found = 0;
2066 
2067 	if (!adapter || !addr)
2068 		return -EINVAL;
2069 
2070 	down_write(&adapter->maps_lock);
2071 	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
2072 		if (map->guest_addr == addr) {
2073 			found = 1;
2074 			atomic_dec(&adapter->nr_maps);
2075 			list_del(&map->list);
2076 			put_page(map->page);
2077 			kfree(map);
2078 			break;
2079 		}
2080 	}
2081 	up_write(&adapter->maps_lock);
2082 
2083 	return found ? 0 : -EINVAL;
2084 }
2085 
2086 void kvm_s390_destroy_adapters(struct kvm *kvm)
2087 {
2088 	int i;
2089 	struct s390_map_info *map, *tmp;
2090 
2091 	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2092 		if (!kvm->arch.adapters[i])
2093 			continue;
2094 		list_for_each_entry_safe(map, tmp,
2095 					 &kvm->arch.adapters[i]->maps, list) {
2096 			list_del(&map->list);
2097 			put_page(map->page);
2098 			kfree(map);
2099 		}
2100 		kfree(kvm->arch.adapters[i]);
2101 	}
2102 }
2103 
2104 static int modify_io_adapter(struct kvm_device *dev,
2105 			     struct kvm_device_attr *attr)
2106 {
2107 	struct kvm_s390_io_adapter_req req;
2108 	struct s390_io_adapter *adapter;
2109 	int ret;
2110 
2111 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2112 		return -EFAULT;
2113 
2114 	adapter = get_io_adapter(dev->kvm, req.id);
2115 	if (!adapter)
2116 		return -EINVAL;
2117 	switch (req.type) {
2118 	case KVM_S390_IO_ADAPTER_MASK:
2119 		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2120 		if (ret > 0)
2121 			ret = 0;
2122 		break;
2123 	case KVM_S390_IO_ADAPTER_MAP:
2124 		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2125 		break;
2126 	case KVM_S390_IO_ADAPTER_UNMAP:
2127 		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2128 		break;
2129 	default:
2130 		ret = -EINVAL;
2131 	}
2132 
2133 	return ret;
2134 }
2135 
2136 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2137 
2138 {
2139 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2140 	u32 schid;
2141 
2142 	if (attr->flags)
2143 		return -EINVAL;
2144 	if (attr->attr != sizeof(schid))
2145 		return -EINVAL;
2146 	if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2147 		return -EFAULT;
2148 	kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2149 	/*
2150 	 * If userspace is conforming to the architecture, we can have at most
2151 	 * one pending I/O interrupt per subchannel, so this is effectively a
2152 	 * clear all.
2153 	 */
2154 	return 0;
2155 }
2156 
2157 static int modify_ais_mode(struct kvm *kvm, struct kvm_device_attr *attr)
2158 {
2159 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2160 	struct kvm_s390_ais_req req;
2161 	int ret = 0;
2162 
2163 	if (!fi->ais_enabled)
2164 		return -ENOTSUPP;
2165 
2166 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2167 		return -EFAULT;
2168 
2169 	if (req.isc > MAX_ISC)
2170 		return -EINVAL;
2171 
2172 	trace_kvm_s390_modify_ais_mode(req.isc,
2173 				       (fi->simm & AIS_MODE_MASK(req.isc)) ?
2174 				       (fi->nimm & AIS_MODE_MASK(req.isc)) ?
2175 				       2 : KVM_S390_AIS_MODE_SINGLE :
2176 				       KVM_S390_AIS_MODE_ALL, req.mode);
2177 
2178 	mutex_lock(&fi->ais_lock);
2179 	switch (req.mode) {
2180 	case KVM_S390_AIS_MODE_ALL:
2181 		fi->simm &= ~AIS_MODE_MASK(req.isc);
2182 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2183 		break;
2184 	case KVM_S390_AIS_MODE_SINGLE:
2185 		fi->simm |= AIS_MODE_MASK(req.isc);
2186 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2187 		break;
2188 	default:
2189 		ret = -EINVAL;
2190 	}
2191 	mutex_unlock(&fi->ais_lock);
2192 
2193 	return ret;
2194 }
2195 
2196 static int kvm_s390_inject_airq(struct kvm *kvm,
2197 				struct s390_io_adapter *adapter)
2198 {
2199 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2200 	struct kvm_s390_interrupt s390int = {
2201 		.type = KVM_S390_INT_IO(1, 0, 0, 0),
2202 		.parm = 0,
2203 		.parm64 = (adapter->isc << 27) | 0x80000000,
2204 	};
2205 	int ret = 0;
2206 
2207 	if (!fi->ais_enabled || !adapter->suppressible)
2208 		return kvm_s390_inject_vm(kvm, &s390int);
2209 
2210 	mutex_lock(&fi->ais_lock);
2211 	if (fi->nimm & AIS_MODE_MASK(adapter->isc)) {
2212 		trace_kvm_s390_airq_suppressed(adapter->id, adapter->isc);
2213 		goto out;
2214 	}
2215 
2216 	ret = kvm_s390_inject_vm(kvm, &s390int);
2217 	if (!ret && (fi->simm & AIS_MODE_MASK(adapter->isc))) {
2218 		fi->nimm |= AIS_MODE_MASK(adapter->isc);
2219 		trace_kvm_s390_modify_ais_mode(adapter->isc,
2220 					       KVM_S390_AIS_MODE_SINGLE, 2);
2221 	}
2222 out:
2223 	mutex_unlock(&fi->ais_lock);
2224 	return ret;
2225 }
2226 
2227 static int flic_inject_airq(struct kvm *kvm, struct kvm_device_attr *attr)
2228 {
2229 	unsigned int id = attr->attr;
2230 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2231 
2232 	if (!adapter)
2233 		return -EINVAL;
2234 
2235 	return kvm_s390_inject_airq(kvm, adapter);
2236 }
2237 
2238 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2239 {
2240 	int r = 0;
2241 	unsigned int i;
2242 	struct kvm_vcpu *vcpu;
2243 
2244 	switch (attr->group) {
2245 	case KVM_DEV_FLIC_ENQUEUE:
2246 		r = enqueue_floating_irq(dev, attr);
2247 		break;
2248 	case KVM_DEV_FLIC_CLEAR_IRQS:
2249 		kvm_s390_clear_float_irqs(dev->kvm);
2250 		break;
2251 	case KVM_DEV_FLIC_APF_ENABLE:
2252 		dev->kvm->arch.gmap->pfault_enabled = 1;
2253 		break;
2254 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2255 		dev->kvm->arch.gmap->pfault_enabled = 0;
2256 		/*
2257 		 * Make sure no async faults are in transition when
2258 		 * clearing the queues. So we don't need to worry
2259 		 * about late coming workers.
2260 		 */
2261 		synchronize_srcu(&dev->kvm->srcu);
2262 		kvm_for_each_vcpu(i, vcpu, dev->kvm)
2263 			kvm_clear_async_pf_completion_queue(vcpu);
2264 		break;
2265 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2266 		r = register_io_adapter(dev, attr);
2267 		break;
2268 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2269 		r = modify_io_adapter(dev, attr);
2270 		break;
2271 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2272 		r = clear_io_irq(dev->kvm, attr);
2273 		break;
2274 	case KVM_DEV_FLIC_AISM:
2275 		r = modify_ais_mode(dev->kvm, attr);
2276 		break;
2277 	case KVM_DEV_FLIC_AIRQ_INJECT:
2278 		r = flic_inject_airq(dev->kvm, attr);
2279 		break;
2280 	default:
2281 		r = -EINVAL;
2282 	}
2283 
2284 	return r;
2285 }
2286 
2287 static int flic_has_attr(struct kvm_device *dev,
2288 			     struct kvm_device_attr *attr)
2289 {
2290 	switch (attr->group) {
2291 	case KVM_DEV_FLIC_GET_ALL_IRQS:
2292 	case KVM_DEV_FLIC_ENQUEUE:
2293 	case KVM_DEV_FLIC_CLEAR_IRQS:
2294 	case KVM_DEV_FLIC_APF_ENABLE:
2295 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2296 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2297 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2298 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2299 	case KVM_DEV_FLIC_AISM:
2300 	case KVM_DEV_FLIC_AIRQ_INJECT:
2301 		return 0;
2302 	}
2303 	return -ENXIO;
2304 }
2305 
2306 static int flic_create(struct kvm_device *dev, u32 type)
2307 {
2308 	if (!dev)
2309 		return -EINVAL;
2310 	if (dev->kvm->arch.flic)
2311 		return -EINVAL;
2312 	dev->kvm->arch.flic = dev;
2313 	return 0;
2314 }
2315 
2316 static void flic_destroy(struct kvm_device *dev)
2317 {
2318 	dev->kvm->arch.flic = NULL;
2319 	kfree(dev);
2320 }
2321 
2322 /* s390 floating irq controller (flic) */
2323 struct kvm_device_ops kvm_flic_ops = {
2324 	.name = "kvm-flic",
2325 	.get_attr = flic_get_attr,
2326 	.set_attr = flic_set_attr,
2327 	.has_attr = flic_has_attr,
2328 	.create = flic_create,
2329 	.destroy = flic_destroy,
2330 };
2331 
2332 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2333 {
2334 	unsigned long bit;
2335 
2336 	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2337 
2338 	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2339 }
2340 
2341 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2342 					  u64 addr)
2343 {
2344 	struct s390_map_info *map;
2345 
2346 	if (!adapter)
2347 		return NULL;
2348 
2349 	list_for_each_entry(map, &adapter->maps, list) {
2350 		if (map->guest_addr == addr)
2351 			return map;
2352 	}
2353 	return NULL;
2354 }
2355 
2356 static int adapter_indicators_set(struct kvm *kvm,
2357 				  struct s390_io_adapter *adapter,
2358 				  struct kvm_s390_adapter_int *adapter_int)
2359 {
2360 	unsigned long bit;
2361 	int summary_set, idx;
2362 	struct s390_map_info *info;
2363 	void *map;
2364 
2365 	info = get_map_info(adapter, adapter_int->ind_addr);
2366 	if (!info)
2367 		return -1;
2368 	map = page_address(info->page);
2369 	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2370 	set_bit(bit, map);
2371 	idx = srcu_read_lock(&kvm->srcu);
2372 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2373 	set_page_dirty_lock(info->page);
2374 	info = get_map_info(adapter, adapter_int->summary_addr);
2375 	if (!info) {
2376 		srcu_read_unlock(&kvm->srcu, idx);
2377 		return -1;
2378 	}
2379 	map = page_address(info->page);
2380 	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2381 			  adapter->swap);
2382 	summary_set = test_and_set_bit(bit, map);
2383 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2384 	set_page_dirty_lock(info->page);
2385 	srcu_read_unlock(&kvm->srcu, idx);
2386 	return summary_set ? 0 : 1;
2387 }
2388 
2389 /*
2390  * < 0 - not injected due to error
2391  * = 0 - coalesced, summary indicator already active
2392  * > 0 - injected interrupt
2393  */
2394 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2395 			   struct kvm *kvm, int irq_source_id, int level,
2396 			   bool line_status)
2397 {
2398 	int ret;
2399 	struct s390_io_adapter *adapter;
2400 
2401 	/* We're only interested in the 0->1 transition. */
2402 	if (!level)
2403 		return 0;
2404 	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2405 	if (!adapter)
2406 		return -1;
2407 	down_read(&adapter->maps_lock);
2408 	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2409 	up_read(&adapter->maps_lock);
2410 	if ((ret > 0) && !adapter->masked) {
2411 		ret = kvm_s390_inject_airq(kvm, adapter);
2412 		if (ret == 0)
2413 			ret = 1;
2414 	}
2415 	return ret;
2416 }
2417 
2418 int kvm_set_routing_entry(struct kvm *kvm,
2419 			  struct kvm_kernel_irq_routing_entry *e,
2420 			  const struct kvm_irq_routing_entry *ue)
2421 {
2422 	int ret;
2423 
2424 	switch (ue->type) {
2425 	case KVM_IRQ_ROUTING_S390_ADAPTER:
2426 		e->set = set_adapter_int;
2427 		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2428 		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2429 		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2430 		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2431 		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2432 		ret = 0;
2433 		break;
2434 	default:
2435 		ret = -EINVAL;
2436 	}
2437 
2438 	return ret;
2439 }
2440 
2441 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2442 		int irq_source_id, int level, bool line_status)
2443 {
2444 	return -EINVAL;
2445 }
2446 
2447 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2448 {
2449 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2450 	struct kvm_s390_irq *buf;
2451 	int r = 0;
2452 	int n;
2453 
2454 	buf = vmalloc(len);
2455 	if (!buf)
2456 		return -ENOMEM;
2457 
2458 	if (copy_from_user((void *) buf, irqstate, len)) {
2459 		r = -EFAULT;
2460 		goto out_free;
2461 	}
2462 
2463 	/*
2464 	 * Don't allow setting the interrupt state
2465 	 * when there are already interrupts pending
2466 	 */
2467 	spin_lock(&li->lock);
2468 	if (li->pending_irqs) {
2469 		r = -EBUSY;
2470 		goto out_unlock;
2471 	}
2472 
2473 	for (n = 0; n < len / sizeof(*buf); n++) {
2474 		r = do_inject_vcpu(vcpu, &buf[n]);
2475 		if (r)
2476 			break;
2477 	}
2478 
2479 out_unlock:
2480 	spin_unlock(&li->lock);
2481 out_free:
2482 	vfree(buf);
2483 
2484 	return r;
2485 }
2486 
2487 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2488 			    struct kvm_s390_irq *irq,
2489 			    unsigned long irq_type)
2490 {
2491 	switch (irq_type) {
2492 	case IRQ_PEND_MCHK_EX:
2493 	case IRQ_PEND_MCHK_REP:
2494 		irq->type = KVM_S390_MCHK;
2495 		irq->u.mchk = li->irq.mchk;
2496 		break;
2497 	case IRQ_PEND_PROG:
2498 		irq->type = KVM_S390_PROGRAM_INT;
2499 		irq->u.pgm = li->irq.pgm;
2500 		break;
2501 	case IRQ_PEND_PFAULT_INIT:
2502 		irq->type = KVM_S390_INT_PFAULT_INIT;
2503 		irq->u.ext = li->irq.ext;
2504 		break;
2505 	case IRQ_PEND_EXT_EXTERNAL:
2506 		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2507 		irq->u.extcall = li->irq.extcall;
2508 		break;
2509 	case IRQ_PEND_EXT_CLOCK_COMP:
2510 		irq->type = KVM_S390_INT_CLOCK_COMP;
2511 		break;
2512 	case IRQ_PEND_EXT_CPU_TIMER:
2513 		irq->type = KVM_S390_INT_CPU_TIMER;
2514 		break;
2515 	case IRQ_PEND_SIGP_STOP:
2516 		irq->type = KVM_S390_SIGP_STOP;
2517 		irq->u.stop = li->irq.stop;
2518 		break;
2519 	case IRQ_PEND_RESTART:
2520 		irq->type = KVM_S390_RESTART;
2521 		break;
2522 	case IRQ_PEND_SET_PREFIX:
2523 		irq->type = KVM_S390_SIGP_SET_PREFIX;
2524 		irq->u.prefix = li->irq.prefix;
2525 		break;
2526 	}
2527 }
2528 
2529 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2530 {
2531 	int scn;
2532 	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2533 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2534 	unsigned long pending_irqs;
2535 	struct kvm_s390_irq irq;
2536 	unsigned long irq_type;
2537 	int cpuaddr;
2538 	int n = 0;
2539 
2540 	spin_lock(&li->lock);
2541 	pending_irqs = li->pending_irqs;
2542 	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2543 	       sizeof(sigp_emerg_pending));
2544 	spin_unlock(&li->lock);
2545 
2546 	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2547 		memset(&irq, 0, sizeof(irq));
2548 		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2549 			continue;
2550 		if (n + sizeof(irq) > len)
2551 			return -ENOBUFS;
2552 		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2553 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2554 			return -EFAULT;
2555 		n += sizeof(irq);
2556 	}
2557 
2558 	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2559 		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2560 			memset(&irq, 0, sizeof(irq));
2561 			if (n + sizeof(irq) > len)
2562 				return -ENOBUFS;
2563 			irq.type = KVM_S390_INT_EMERGENCY;
2564 			irq.u.emerg.code = cpuaddr;
2565 			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2566 				return -EFAULT;
2567 			n += sizeof(irq);
2568 		}
2569 	}
2570 
2571 	if (sca_ext_call_pending(vcpu, &scn)) {
2572 		if (n + sizeof(irq) > len)
2573 			return -ENOBUFS;
2574 		memset(&irq, 0, sizeof(irq));
2575 		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2576 		irq.u.extcall.code = scn;
2577 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2578 			return -EFAULT;
2579 		n += sizeof(irq);
2580 	}
2581 
2582 	return n;
2583 }
2584