xref: /linux/arch/s390/kvm/interrupt.c (revision c411ed854584a71b0e86ac3019b60e4789d88086)
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 	/*
255 	 * Check both floating and local interrupt's cr14 because
256 	 * bit IRQ_PEND_MCHK_REP could be set in both cases.
257 	 */
258 	if (!(vcpu->arch.sie_block->gcr[14] &
259 	   (vcpu->kvm->arch.float_int.mchk.cr14 |
260 	   vcpu->arch.local_int.irq.mchk.cr14)))
261 		__clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
262 
263 	/*
264 	 * STOP irqs will never be actively delivered. They are triggered via
265 	 * intercept requests and cleared when the stop intercept is performed.
266 	 */
267 	__clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
268 
269 	return active_mask;
270 }
271 
272 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
273 {
274 	atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
275 	set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
276 }
277 
278 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
279 {
280 	atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
281 	clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
282 }
283 
284 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
285 {
286 	atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
287 		    &vcpu->arch.sie_block->cpuflags);
288 	vcpu->arch.sie_block->lctl = 0x0000;
289 	vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
290 
291 	if (guestdbg_enabled(vcpu)) {
292 		vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
293 					       LCTL_CR10 | LCTL_CR11);
294 		vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
295 	}
296 }
297 
298 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
299 {
300 	atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
301 }
302 
303 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
304 {
305 	if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
306 		return;
307 	else if (psw_ioint_disabled(vcpu))
308 		__set_cpuflag(vcpu, CPUSTAT_IO_INT);
309 	else
310 		vcpu->arch.sie_block->lctl |= LCTL_CR6;
311 }
312 
313 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
314 {
315 	if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
316 		return;
317 	if (psw_extint_disabled(vcpu))
318 		__set_cpuflag(vcpu, CPUSTAT_EXT_INT);
319 	else
320 		vcpu->arch.sie_block->lctl |= LCTL_CR0;
321 }
322 
323 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
324 {
325 	if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
326 		return;
327 	if (psw_mchk_disabled(vcpu))
328 		vcpu->arch.sie_block->ictl |= ICTL_LPSW;
329 	else
330 		vcpu->arch.sie_block->lctl |= LCTL_CR14;
331 }
332 
333 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
334 {
335 	if (kvm_s390_is_stop_irq_pending(vcpu))
336 		__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
337 }
338 
339 /* Set interception request for non-deliverable interrupts */
340 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
341 {
342 	set_intercept_indicators_io(vcpu);
343 	set_intercept_indicators_ext(vcpu);
344 	set_intercept_indicators_mchk(vcpu);
345 	set_intercept_indicators_stop(vcpu);
346 }
347 
348 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
349 {
350 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
351 	int rc;
352 
353 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
354 					 0, 0);
355 
356 	rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
357 			   (u16 *)__LC_EXT_INT_CODE);
358 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
359 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
360 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
361 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
362 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
363 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
364 	return rc ? -EFAULT : 0;
365 }
366 
367 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
368 {
369 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
370 	int rc;
371 
372 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
373 					 0, 0);
374 
375 	rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
376 			   (u16 __user *)__LC_EXT_INT_CODE);
377 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
378 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
379 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
380 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
381 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
382 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
383 	return rc ? -EFAULT : 0;
384 }
385 
386 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
387 {
388 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
389 	struct kvm_s390_ext_info ext;
390 	int rc;
391 
392 	spin_lock(&li->lock);
393 	ext = li->irq.ext;
394 	clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
395 	li->irq.ext.ext_params2 = 0;
396 	spin_unlock(&li->lock);
397 
398 	VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
399 		   ext.ext_params2);
400 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
401 					 KVM_S390_INT_PFAULT_INIT,
402 					 0, ext.ext_params2);
403 
404 	rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
405 	rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
406 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
407 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
408 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
409 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
410 	rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
411 	return rc ? -EFAULT : 0;
412 }
413 
414 static int __write_machine_check(struct kvm_vcpu *vcpu,
415 				 struct kvm_s390_mchk_info *mchk)
416 {
417 	unsigned long ext_sa_addr;
418 	unsigned long lc;
419 	freg_t fprs[NUM_FPRS];
420 	union mci mci;
421 	int rc;
422 
423 	mci.val = mchk->mcic;
424 	/* take care of lazy register loading */
425 	save_fpu_regs();
426 	save_access_regs(vcpu->run->s.regs.acrs);
427 	if (MACHINE_HAS_GS && vcpu->arch.gs_enabled)
428 		save_gs_cb(current->thread.gs_cb);
429 
430 	/* Extended save area */
431 	rc = read_guest_lc(vcpu, __LC_MCESAD, &ext_sa_addr,
432 			   sizeof(unsigned long));
433 	/* Only bits 0 through 63-LC are used for address formation */
434 	lc = ext_sa_addr & MCESA_LC_MASK;
435 	if (test_kvm_facility(vcpu->kvm, 133)) {
436 		switch (lc) {
437 		case 0:
438 		case 10:
439 			ext_sa_addr &= ~0x3ffUL;
440 			break;
441 		case 11:
442 			ext_sa_addr &= ~0x7ffUL;
443 			break;
444 		case 12:
445 			ext_sa_addr &= ~0xfffUL;
446 			break;
447 		default:
448 			ext_sa_addr = 0;
449 			break;
450 		}
451 	} else {
452 		ext_sa_addr &= ~0x3ffUL;
453 	}
454 
455 	if (!rc && mci.vr && ext_sa_addr && test_kvm_facility(vcpu->kvm, 129)) {
456 		if (write_guest_abs(vcpu, ext_sa_addr, vcpu->run->s.regs.vrs,
457 				    512))
458 			mci.vr = 0;
459 	} else {
460 		mci.vr = 0;
461 	}
462 	if (!rc && mci.gs && ext_sa_addr && test_kvm_facility(vcpu->kvm, 133)
463 	    && (lc == 11 || lc == 12)) {
464 		if (write_guest_abs(vcpu, ext_sa_addr + 1024,
465 				    &vcpu->run->s.regs.gscb, 32))
466 			mci.gs = 0;
467 	} else {
468 		mci.gs = 0;
469 	}
470 
471 	/* General interruption information */
472 	rc |= put_guest_lc(vcpu, 1, (u8 __user *) __LC_AR_MODE_ID);
473 	rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
474 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
475 	rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
476 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
477 	rc |= put_guest_lc(vcpu, mci.val, (u64 __user *) __LC_MCCK_CODE);
478 
479 	/* Register-save areas */
480 	if (MACHINE_HAS_VX) {
481 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
482 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA, fprs, 128);
483 	} else {
484 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA,
485 				     vcpu->run->s.regs.fprs, 128);
486 	}
487 	rc |= write_guest_lc(vcpu, __LC_GPREGS_SAVE_AREA,
488 			     vcpu->run->s.regs.gprs, 128);
489 	rc |= put_guest_lc(vcpu, current->thread.fpu.fpc,
490 			   (u32 __user *) __LC_FP_CREG_SAVE_AREA);
491 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->todpr,
492 			   (u32 __user *) __LC_TOD_PROGREG_SAVE_AREA);
493 	rc |= put_guest_lc(vcpu, kvm_s390_get_cpu_timer(vcpu),
494 			   (u64 __user *) __LC_CPU_TIMER_SAVE_AREA);
495 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->ckc >> 8,
496 			   (u64 __user *) __LC_CLOCK_COMP_SAVE_AREA);
497 	rc |= write_guest_lc(vcpu, __LC_AREGS_SAVE_AREA,
498 			     &vcpu->run->s.regs.acrs, 64);
499 	rc |= write_guest_lc(vcpu, __LC_CREGS_SAVE_AREA,
500 			     &vcpu->arch.sie_block->gcr, 128);
501 
502 	/* Extended interruption information */
503 	rc |= put_guest_lc(vcpu, mchk->ext_damage_code,
504 			   (u32 __user *) __LC_EXT_DAMAGE_CODE);
505 	rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
506 			   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
507 	rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA, &mchk->fixed_logout,
508 			     sizeof(mchk->fixed_logout));
509 	return rc ? -EFAULT : 0;
510 }
511 
512 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
513 {
514 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
515 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
516 	struct kvm_s390_mchk_info mchk = {};
517 	int deliver = 0;
518 	int rc = 0;
519 
520 	spin_lock(&fi->lock);
521 	spin_lock(&li->lock);
522 	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
523 	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
524 		/*
525 		 * If there was an exigent machine check pending, then any
526 		 * repressible machine checks that might have been pending
527 		 * are indicated along with it, so always clear bits for
528 		 * repressible and exigent interrupts
529 		 */
530 		mchk = li->irq.mchk;
531 		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
532 		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
533 		memset(&li->irq.mchk, 0, sizeof(mchk));
534 		deliver = 1;
535 	}
536 	/*
537 	 * We indicate floating repressible conditions along with
538 	 * other pending conditions. Channel Report Pending and Channel
539 	 * Subsystem damage are the only two and and are indicated by
540 	 * bits in mcic and masked in cr14.
541 	 */
542 	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
543 		mchk.mcic |= fi->mchk.mcic;
544 		mchk.cr14 |= fi->mchk.cr14;
545 		memset(&fi->mchk, 0, sizeof(mchk));
546 		deliver = 1;
547 	}
548 	spin_unlock(&li->lock);
549 	spin_unlock(&fi->lock);
550 
551 	if (deliver) {
552 		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
553 			   mchk.mcic);
554 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
555 						 KVM_S390_MCHK,
556 						 mchk.cr14, mchk.mcic);
557 		rc = __write_machine_check(vcpu, &mchk);
558 	}
559 	return rc;
560 }
561 
562 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
563 {
564 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
565 	int rc;
566 
567 	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
568 	vcpu->stat.deliver_restart_signal++;
569 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
570 
571 	rc  = write_guest_lc(vcpu,
572 			     offsetof(struct lowcore, restart_old_psw),
573 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
574 	rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
575 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
576 	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
577 	return rc ? -EFAULT : 0;
578 }
579 
580 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
581 {
582 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
583 	struct kvm_s390_prefix_info prefix;
584 
585 	spin_lock(&li->lock);
586 	prefix = li->irq.prefix;
587 	li->irq.prefix.address = 0;
588 	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
589 	spin_unlock(&li->lock);
590 
591 	vcpu->stat.deliver_prefix_signal++;
592 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
593 					 KVM_S390_SIGP_SET_PREFIX,
594 					 prefix.address, 0);
595 
596 	kvm_s390_set_prefix(vcpu, prefix.address);
597 	return 0;
598 }
599 
600 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
601 {
602 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
603 	int rc;
604 	int cpu_addr;
605 
606 	spin_lock(&li->lock);
607 	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
608 	clear_bit(cpu_addr, li->sigp_emerg_pending);
609 	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
610 		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
611 	spin_unlock(&li->lock);
612 
613 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
614 	vcpu->stat.deliver_emergency_signal++;
615 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
616 					 cpu_addr, 0);
617 
618 	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
619 			   (u16 *)__LC_EXT_INT_CODE);
620 	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
621 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
622 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
623 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
624 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
625 	return rc ? -EFAULT : 0;
626 }
627 
628 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
629 {
630 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
631 	struct kvm_s390_extcall_info extcall;
632 	int rc;
633 
634 	spin_lock(&li->lock);
635 	extcall = li->irq.extcall;
636 	li->irq.extcall.code = 0;
637 	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
638 	spin_unlock(&li->lock);
639 
640 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
641 	vcpu->stat.deliver_external_call++;
642 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
643 					 KVM_S390_INT_EXTERNAL_CALL,
644 					 extcall.code, 0);
645 
646 	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
647 			   (u16 *)__LC_EXT_INT_CODE);
648 	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
649 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
650 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
651 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
652 			    sizeof(psw_t));
653 	return rc ? -EFAULT : 0;
654 }
655 
656 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
657 {
658 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
659 	struct kvm_s390_pgm_info pgm_info;
660 	int rc = 0, nullifying = false;
661 	u16 ilen;
662 
663 	spin_lock(&li->lock);
664 	pgm_info = li->irq.pgm;
665 	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
666 	memset(&li->irq.pgm, 0, sizeof(pgm_info));
667 	spin_unlock(&li->lock);
668 
669 	ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
670 	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
671 		   pgm_info.code, ilen);
672 	vcpu->stat.deliver_program_int++;
673 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
674 					 pgm_info.code, 0);
675 
676 	switch (pgm_info.code & ~PGM_PER) {
677 	case PGM_AFX_TRANSLATION:
678 	case PGM_ASX_TRANSLATION:
679 	case PGM_EX_TRANSLATION:
680 	case PGM_LFX_TRANSLATION:
681 	case PGM_LSTE_SEQUENCE:
682 	case PGM_LSX_TRANSLATION:
683 	case PGM_LX_TRANSLATION:
684 	case PGM_PRIMARY_AUTHORITY:
685 	case PGM_SECONDARY_AUTHORITY:
686 		nullifying = true;
687 		/* fall through */
688 	case PGM_SPACE_SWITCH:
689 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
690 				  (u64 *)__LC_TRANS_EXC_CODE);
691 		break;
692 	case PGM_ALEN_TRANSLATION:
693 	case PGM_ALE_SEQUENCE:
694 	case PGM_ASTE_INSTANCE:
695 	case PGM_ASTE_SEQUENCE:
696 	case PGM_ASTE_VALIDITY:
697 	case PGM_EXTENDED_AUTHORITY:
698 		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
699 				  (u8 *)__LC_EXC_ACCESS_ID);
700 		nullifying = true;
701 		break;
702 	case PGM_ASCE_TYPE:
703 	case PGM_PAGE_TRANSLATION:
704 	case PGM_REGION_FIRST_TRANS:
705 	case PGM_REGION_SECOND_TRANS:
706 	case PGM_REGION_THIRD_TRANS:
707 	case PGM_SEGMENT_TRANSLATION:
708 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
709 				  (u64 *)__LC_TRANS_EXC_CODE);
710 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
711 				   (u8 *)__LC_EXC_ACCESS_ID);
712 		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
713 				   (u8 *)__LC_OP_ACCESS_ID);
714 		nullifying = true;
715 		break;
716 	case PGM_MONITOR:
717 		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
718 				  (u16 *)__LC_MON_CLASS_NR);
719 		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
720 				   (u64 *)__LC_MON_CODE);
721 		break;
722 	case PGM_VECTOR_PROCESSING:
723 	case PGM_DATA:
724 		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
725 				  (u32 *)__LC_DATA_EXC_CODE);
726 		break;
727 	case PGM_PROTECTION:
728 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
729 				  (u64 *)__LC_TRANS_EXC_CODE);
730 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
731 				   (u8 *)__LC_EXC_ACCESS_ID);
732 		break;
733 	case PGM_STACK_FULL:
734 	case PGM_STACK_EMPTY:
735 	case PGM_STACK_SPECIFICATION:
736 	case PGM_STACK_TYPE:
737 	case PGM_STACK_OPERATION:
738 	case PGM_TRACE_TABEL:
739 	case PGM_CRYPTO_OPERATION:
740 		nullifying = true;
741 		break;
742 	}
743 
744 	if (pgm_info.code & PGM_PER) {
745 		rc |= put_guest_lc(vcpu, pgm_info.per_code,
746 				   (u8 *) __LC_PER_CODE);
747 		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
748 				   (u8 *)__LC_PER_ATMID);
749 		rc |= put_guest_lc(vcpu, pgm_info.per_address,
750 				   (u64 *) __LC_PER_ADDRESS);
751 		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
752 				   (u8 *) __LC_PER_ACCESS_ID);
753 	}
754 
755 	if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
756 		kvm_s390_rewind_psw(vcpu, ilen);
757 
758 	/* bit 1+2 of the target are the ilc, so we can directly use ilen */
759 	rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
760 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
761 				 (u64 *) __LC_LAST_BREAK);
762 	rc |= put_guest_lc(vcpu, pgm_info.code,
763 			   (u16 *)__LC_PGM_INT_CODE);
764 	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
765 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
766 	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
767 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
768 	return rc ? -EFAULT : 0;
769 }
770 
771 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
772 {
773 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
774 	struct kvm_s390_ext_info ext;
775 	int rc = 0;
776 
777 	spin_lock(&fi->lock);
778 	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
779 		spin_unlock(&fi->lock);
780 		return 0;
781 	}
782 	ext = fi->srv_signal;
783 	memset(&fi->srv_signal, 0, sizeof(ext));
784 	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
785 	spin_unlock(&fi->lock);
786 
787 	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
788 		   ext.ext_params);
789 	vcpu->stat.deliver_service_signal++;
790 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
791 					 ext.ext_params, 0);
792 
793 	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
794 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
795 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
796 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
797 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
798 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
799 	rc |= put_guest_lc(vcpu, ext.ext_params,
800 			   (u32 *)__LC_EXT_PARAMS);
801 
802 	return rc ? -EFAULT : 0;
803 }
804 
805 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
806 {
807 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
808 	struct kvm_s390_interrupt_info *inti;
809 	int rc = 0;
810 
811 	spin_lock(&fi->lock);
812 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
813 					struct kvm_s390_interrupt_info,
814 					list);
815 	if (inti) {
816 		list_del(&inti->list);
817 		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
818 	}
819 	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
820 		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
821 	spin_unlock(&fi->lock);
822 
823 	if (inti) {
824 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
825 						 KVM_S390_INT_PFAULT_DONE, 0,
826 						 inti->ext.ext_params2);
827 		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
828 			   inti->ext.ext_params2);
829 
830 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
831 				(u16 *)__LC_EXT_INT_CODE);
832 		rc |= put_guest_lc(vcpu, PFAULT_DONE,
833 				(u16 *)__LC_EXT_CPU_ADDR);
834 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
835 				&vcpu->arch.sie_block->gpsw,
836 				sizeof(psw_t));
837 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
838 				&vcpu->arch.sie_block->gpsw,
839 				sizeof(psw_t));
840 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
841 				(u64 *)__LC_EXT_PARAMS2);
842 		kfree(inti);
843 	}
844 	return rc ? -EFAULT : 0;
845 }
846 
847 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
848 {
849 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
850 	struct kvm_s390_interrupt_info *inti;
851 	int rc = 0;
852 
853 	spin_lock(&fi->lock);
854 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
855 					struct kvm_s390_interrupt_info,
856 					list);
857 	if (inti) {
858 		VCPU_EVENT(vcpu, 4,
859 			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
860 			   inti->ext.ext_params, inti->ext.ext_params2);
861 		vcpu->stat.deliver_virtio_interrupt++;
862 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
863 				inti->type,
864 				inti->ext.ext_params,
865 				inti->ext.ext_params2);
866 		list_del(&inti->list);
867 		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
868 	}
869 	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
870 		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
871 	spin_unlock(&fi->lock);
872 
873 	if (inti) {
874 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
875 				(u16 *)__LC_EXT_INT_CODE);
876 		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
877 				(u16 *)__LC_EXT_CPU_ADDR);
878 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
879 				&vcpu->arch.sie_block->gpsw,
880 				sizeof(psw_t));
881 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
882 				&vcpu->arch.sie_block->gpsw,
883 				sizeof(psw_t));
884 		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
885 				(u32 *)__LC_EXT_PARAMS);
886 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
887 				(u64 *)__LC_EXT_PARAMS2);
888 		kfree(inti);
889 	}
890 	return rc ? -EFAULT : 0;
891 }
892 
893 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
894 				     unsigned long irq_type)
895 {
896 	struct list_head *isc_list;
897 	struct kvm_s390_float_interrupt *fi;
898 	struct kvm_s390_interrupt_info *inti = NULL;
899 	int rc = 0;
900 
901 	fi = &vcpu->kvm->arch.float_int;
902 
903 	spin_lock(&fi->lock);
904 	isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
905 	inti = list_first_entry_or_null(isc_list,
906 					struct kvm_s390_interrupt_info,
907 					list);
908 	if (inti) {
909 		if (inti->type & KVM_S390_INT_IO_AI_MASK)
910 			VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
911 		else
912 			VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
913 			inti->io.subchannel_id >> 8,
914 			inti->io.subchannel_id >> 1 & 0x3,
915 			inti->io.subchannel_nr);
916 
917 		vcpu->stat.deliver_io_int++;
918 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
919 				inti->type,
920 				((__u32)inti->io.subchannel_id << 16) |
921 				inti->io.subchannel_nr,
922 				((__u64)inti->io.io_int_parm << 32) |
923 				inti->io.io_int_word);
924 		list_del(&inti->list);
925 		fi->counters[FIRQ_CNTR_IO] -= 1;
926 	}
927 	if (list_empty(isc_list))
928 		clear_bit(irq_type, &fi->pending_irqs);
929 	spin_unlock(&fi->lock);
930 
931 	if (inti) {
932 		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
933 				(u16 *)__LC_SUBCHANNEL_ID);
934 		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
935 				(u16 *)__LC_SUBCHANNEL_NR);
936 		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
937 				(u32 *)__LC_IO_INT_PARM);
938 		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
939 				(u32 *)__LC_IO_INT_WORD);
940 		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
941 				&vcpu->arch.sie_block->gpsw,
942 				sizeof(psw_t));
943 		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
944 				&vcpu->arch.sie_block->gpsw,
945 				sizeof(psw_t));
946 		kfree(inti);
947 	}
948 
949 	return rc ? -EFAULT : 0;
950 }
951 
952 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
953 
954 static const deliver_irq_t deliver_irq_funcs[] = {
955 	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
956 	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
957 	[IRQ_PEND_PROG]           = __deliver_prog,
958 	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
959 	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
960 	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
961 	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
962 	[IRQ_PEND_RESTART]        = __deliver_restart,
963 	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
964 	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
965 	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
966 	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
967 	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
968 };
969 
970 /* Check whether an external call is pending (deliverable or not) */
971 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
972 {
973 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
974 
975 	if (!sclp.has_sigpif)
976 		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
977 
978 	return sca_ext_call_pending(vcpu, NULL);
979 }
980 
981 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
982 {
983 	if (deliverable_irqs(vcpu))
984 		return 1;
985 
986 	if (kvm_cpu_has_pending_timer(vcpu))
987 		return 1;
988 
989 	/* external call pending and deliverable */
990 	if (kvm_s390_ext_call_pending(vcpu) &&
991 	    !psw_extint_disabled(vcpu) &&
992 	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
993 		return 1;
994 
995 	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
996 		return 1;
997 	return 0;
998 }
999 
1000 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
1001 {
1002 	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
1003 }
1004 
1005 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
1006 {
1007 	u64 now, cputm, sltime = 0;
1008 
1009 	if (ckc_interrupts_enabled(vcpu)) {
1010 		now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
1011 		sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
1012 		/* already expired or overflow? */
1013 		if (!sltime || vcpu->arch.sie_block->ckc <= now)
1014 			return 0;
1015 		if (cpu_timer_interrupts_enabled(vcpu)) {
1016 			cputm = kvm_s390_get_cpu_timer(vcpu);
1017 			/* already expired? */
1018 			if (cputm >> 63)
1019 				return 0;
1020 			return min(sltime, tod_to_ns(cputm));
1021 		}
1022 	} else if (cpu_timer_interrupts_enabled(vcpu)) {
1023 		sltime = kvm_s390_get_cpu_timer(vcpu);
1024 		/* already expired? */
1025 		if (sltime >> 63)
1026 			return 0;
1027 	}
1028 	return sltime;
1029 }
1030 
1031 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
1032 {
1033 	u64 sltime;
1034 
1035 	vcpu->stat.exit_wait_state++;
1036 
1037 	/* fast path */
1038 	if (kvm_arch_vcpu_runnable(vcpu))
1039 		return 0;
1040 
1041 	if (psw_interrupts_disabled(vcpu)) {
1042 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
1043 		return -EOPNOTSUPP; /* disabled wait */
1044 	}
1045 
1046 	if (!ckc_interrupts_enabled(vcpu) &&
1047 	    !cpu_timer_interrupts_enabled(vcpu)) {
1048 		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
1049 		__set_cpu_idle(vcpu);
1050 		goto no_timer;
1051 	}
1052 
1053 	sltime = __calculate_sltime(vcpu);
1054 	if (!sltime)
1055 		return 0;
1056 
1057 	__set_cpu_idle(vcpu);
1058 	hrtimer_start(&vcpu->arch.ckc_timer, sltime, HRTIMER_MODE_REL);
1059 	VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
1060 no_timer:
1061 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1062 	kvm_vcpu_block(vcpu);
1063 	__unset_cpu_idle(vcpu);
1064 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1065 
1066 	hrtimer_cancel(&vcpu->arch.ckc_timer);
1067 	return 0;
1068 }
1069 
1070 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
1071 {
1072 	/*
1073 	 * We cannot move this into the if, as the CPU might be already
1074 	 * in kvm_vcpu_block without having the waitqueue set (polling)
1075 	 */
1076 	vcpu->valid_wakeup = true;
1077 	if (swait_active(&vcpu->wq)) {
1078 		/*
1079 		 * The vcpu gave up the cpu voluntarily, mark it as a good
1080 		 * yield-candidate.
1081 		 */
1082 		vcpu->preempted = true;
1083 		swake_up(&vcpu->wq);
1084 		vcpu->stat.halt_wakeup++;
1085 	}
1086 	/*
1087 	 * The VCPU might not be sleeping but is executing the VSIE. Let's
1088 	 * kick it, so it leaves the SIE to process the request.
1089 	 */
1090 	kvm_s390_vsie_kick(vcpu);
1091 }
1092 
1093 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1094 {
1095 	struct kvm_vcpu *vcpu;
1096 	u64 sltime;
1097 
1098 	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1099 	sltime = __calculate_sltime(vcpu);
1100 
1101 	/*
1102 	 * If the monotonic clock runs faster than the tod clock we might be
1103 	 * woken up too early and have to go back to sleep to avoid deadlocks.
1104 	 */
1105 	if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1106 		return HRTIMER_RESTART;
1107 	kvm_s390_vcpu_wakeup(vcpu);
1108 	return HRTIMER_NORESTART;
1109 }
1110 
1111 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1112 {
1113 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1114 
1115 	spin_lock(&li->lock);
1116 	li->pending_irqs = 0;
1117 	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1118 	memset(&li->irq, 0, sizeof(li->irq));
1119 	spin_unlock(&li->lock);
1120 
1121 	sca_clear_ext_call(vcpu);
1122 }
1123 
1124 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1125 {
1126 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1127 	deliver_irq_t func;
1128 	int rc = 0;
1129 	unsigned long irq_type;
1130 	unsigned long irqs;
1131 
1132 	__reset_intercept_indicators(vcpu);
1133 
1134 	/* pending ckc conditions might have been invalidated */
1135 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1136 	if (ckc_irq_pending(vcpu))
1137 		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1138 
1139 	/* pending cpu timer conditions might have been invalidated */
1140 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1141 	if (cpu_timer_irq_pending(vcpu))
1142 		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1143 
1144 	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1145 		/* bits are in the order of interrupt priority */
1146 		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1147 		if (is_ioirq(irq_type)) {
1148 			rc = __deliver_io(vcpu, irq_type);
1149 		} else {
1150 			func = deliver_irq_funcs[irq_type];
1151 			if (!func) {
1152 				WARN_ON_ONCE(func == NULL);
1153 				clear_bit(irq_type, &li->pending_irqs);
1154 				continue;
1155 			}
1156 			rc = func(vcpu);
1157 		}
1158 	}
1159 
1160 	set_intercept_indicators(vcpu);
1161 
1162 	return rc;
1163 }
1164 
1165 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1166 {
1167 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1168 
1169 	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1170 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1171 				   irq->u.pgm.code, 0);
1172 
1173 	if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1174 		/* auto detection if no valid ILC was given */
1175 		irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1176 		irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1177 		irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1178 	}
1179 
1180 	if (irq->u.pgm.code == PGM_PER) {
1181 		li->irq.pgm.code |= PGM_PER;
1182 		li->irq.pgm.flags = irq->u.pgm.flags;
1183 		/* only modify PER related information */
1184 		li->irq.pgm.per_address = irq->u.pgm.per_address;
1185 		li->irq.pgm.per_code = irq->u.pgm.per_code;
1186 		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1187 		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1188 	} else if (!(irq->u.pgm.code & PGM_PER)) {
1189 		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1190 				   irq->u.pgm.code;
1191 		li->irq.pgm.flags = irq->u.pgm.flags;
1192 		/* only modify non-PER information */
1193 		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1194 		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1195 		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1196 		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1197 		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1198 		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1199 	} else {
1200 		li->irq.pgm = irq->u.pgm;
1201 	}
1202 	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1203 	return 0;
1204 }
1205 
1206 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1207 {
1208 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1209 
1210 	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1211 		   irq->u.ext.ext_params2);
1212 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1213 				   irq->u.ext.ext_params,
1214 				   irq->u.ext.ext_params2);
1215 
1216 	li->irq.ext = irq->u.ext;
1217 	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1218 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1219 	return 0;
1220 }
1221 
1222 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1223 {
1224 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1225 	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1226 	uint16_t src_id = irq->u.extcall.code;
1227 
1228 	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1229 		   src_id);
1230 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1231 				   src_id, 0);
1232 
1233 	/* sending vcpu invalid */
1234 	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1235 		return -EINVAL;
1236 
1237 	if (sclp.has_sigpif)
1238 		return sca_inject_ext_call(vcpu, src_id);
1239 
1240 	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1241 		return -EBUSY;
1242 	*extcall = irq->u.extcall;
1243 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1244 	return 0;
1245 }
1246 
1247 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1248 {
1249 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1250 	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1251 
1252 	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1253 		   irq->u.prefix.address);
1254 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1255 				   irq->u.prefix.address, 0);
1256 
1257 	if (!is_vcpu_stopped(vcpu))
1258 		return -EBUSY;
1259 
1260 	*prefix = irq->u.prefix;
1261 	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1262 	return 0;
1263 }
1264 
1265 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1266 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1267 {
1268 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1269 	struct kvm_s390_stop_info *stop = &li->irq.stop;
1270 	int rc = 0;
1271 
1272 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1273 
1274 	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1275 		return -EINVAL;
1276 
1277 	if (is_vcpu_stopped(vcpu)) {
1278 		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1279 			rc = kvm_s390_store_status_unloaded(vcpu,
1280 						KVM_S390_STORE_STATUS_NOADDR);
1281 		return rc;
1282 	}
1283 
1284 	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1285 		return -EBUSY;
1286 	stop->flags = irq->u.stop.flags;
1287 	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1288 	return 0;
1289 }
1290 
1291 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1292 				 struct kvm_s390_irq *irq)
1293 {
1294 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1295 
1296 	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1297 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1298 
1299 	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1300 	return 0;
1301 }
1302 
1303 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1304 				   struct kvm_s390_irq *irq)
1305 {
1306 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1307 
1308 	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1309 		   irq->u.emerg.code);
1310 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1311 				   irq->u.emerg.code, 0);
1312 
1313 	/* sending vcpu invalid */
1314 	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1315 		return -EINVAL;
1316 
1317 	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1318 	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1319 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1320 	return 0;
1321 }
1322 
1323 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1324 {
1325 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1326 	struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1327 
1328 	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1329 		   irq->u.mchk.mcic);
1330 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1331 				   irq->u.mchk.mcic);
1332 
1333 	/*
1334 	 * Because repressible machine checks can be indicated along with
1335 	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1336 	 * we need to combine cr14, mcic and external damage code.
1337 	 * Failing storage address and the logout area should not be or'ed
1338 	 * together, we just indicate the last occurrence of the corresponding
1339 	 * machine check
1340 	 */
1341 	mchk->cr14 |= irq->u.mchk.cr14;
1342 	mchk->mcic |= irq->u.mchk.mcic;
1343 	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1344 	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1345 	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1346 	       sizeof(mchk->fixed_logout));
1347 	if (mchk->mcic & MCHK_EX_MASK)
1348 		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1349 	else if (mchk->mcic & MCHK_REP_MASK)
1350 		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1351 	return 0;
1352 }
1353 
1354 static int __inject_ckc(struct kvm_vcpu *vcpu)
1355 {
1356 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1357 
1358 	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1359 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1360 				   0, 0);
1361 
1362 	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1363 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1364 	return 0;
1365 }
1366 
1367 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1368 {
1369 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1370 
1371 	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1372 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1373 				   0, 0);
1374 
1375 	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1376 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1377 	return 0;
1378 }
1379 
1380 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1381 						  int isc, u32 schid)
1382 {
1383 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1384 	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1385 	struct kvm_s390_interrupt_info *iter;
1386 	u16 id = (schid & 0xffff0000U) >> 16;
1387 	u16 nr = schid & 0x0000ffffU;
1388 
1389 	spin_lock(&fi->lock);
1390 	list_for_each_entry(iter, isc_list, list) {
1391 		if (schid && (id != iter->io.subchannel_id ||
1392 			      nr != iter->io.subchannel_nr))
1393 			continue;
1394 		/* found an appropriate entry */
1395 		list_del_init(&iter->list);
1396 		fi->counters[FIRQ_CNTR_IO] -= 1;
1397 		if (list_empty(isc_list))
1398 			clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1399 		spin_unlock(&fi->lock);
1400 		return iter;
1401 	}
1402 	spin_unlock(&fi->lock);
1403 	return NULL;
1404 }
1405 
1406 /*
1407  * Dequeue and return an I/O interrupt matching any of the interruption
1408  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1409  */
1410 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1411 						    u64 isc_mask, u32 schid)
1412 {
1413 	struct kvm_s390_interrupt_info *inti = NULL;
1414 	int isc;
1415 
1416 	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1417 		if (isc_mask & isc_to_isc_bits(isc))
1418 			inti = get_io_int(kvm, isc, schid);
1419 	}
1420 	return inti;
1421 }
1422 
1423 #define SCCB_MASK 0xFFFFFFF8
1424 #define SCCB_EVENT_PENDING 0x3
1425 
1426 static int __inject_service(struct kvm *kvm,
1427 			     struct kvm_s390_interrupt_info *inti)
1428 {
1429 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1430 
1431 	spin_lock(&fi->lock);
1432 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1433 	/*
1434 	 * Early versions of the QEMU s390 bios will inject several
1435 	 * service interrupts after another without handling a
1436 	 * condition code indicating busy.
1437 	 * We will silently ignore those superfluous sccb values.
1438 	 * A future version of QEMU will take care of serialization
1439 	 * of servc requests
1440 	 */
1441 	if (fi->srv_signal.ext_params & SCCB_MASK)
1442 		goto out;
1443 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1444 	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1445 out:
1446 	spin_unlock(&fi->lock);
1447 	kfree(inti);
1448 	return 0;
1449 }
1450 
1451 static int __inject_virtio(struct kvm *kvm,
1452 			    struct kvm_s390_interrupt_info *inti)
1453 {
1454 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1455 
1456 	spin_lock(&fi->lock);
1457 	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1458 		spin_unlock(&fi->lock);
1459 		return -EBUSY;
1460 	}
1461 	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1462 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1463 	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1464 	spin_unlock(&fi->lock);
1465 	return 0;
1466 }
1467 
1468 static int __inject_pfault_done(struct kvm *kvm,
1469 				 struct kvm_s390_interrupt_info *inti)
1470 {
1471 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1472 
1473 	spin_lock(&fi->lock);
1474 	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1475 		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1476 		spin_unlock(&fi->lock);
1477 		return -EBUSY;
1478 	}
1479 	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1480 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1481 	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1482 	spin_unlock(&fi->lock);
1483 	return 0;
1484 }
1485 
1486 #define CR_PENDING_SUBCLASS 28
1487 static int __inject_float_mchk(struct kvm *kvm,
1488 				struct kvm_s390_interrupt_info *inti)
1489 {
1490 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1491 
1492 	spin_lock(&fi->lock);
1493 	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1494 	fi->mchk.mcic |= inti->mchk.mcic;
1495 	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1496 	spin_unlock(&fi->lock);
1497 	kfree(inti);
1498 	return 0;
1499 }
1500 
1501 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1502 {
1503 	struct kvm_s390_float_interrupt *fi;
1504 	struct list_head *list;
1505 	int isc;
1506 
1507 	fi = &kvm->arch.float_int;
1508 	spin_lock(&fi->lock);
1509 	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1510 		spin_unlock(&fi->lock);
1511 		return -EBUSY;
1512 	}
1513 	fi->counters[FIRQ_CNTR_IO] += 1;
1514 
1515 	if (inti->type & KVM_S390_INT_IO_AI_MASK)
1516 		VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1517 	else
1518 		VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1519 			inti->io.subchannel_id >> 8,
1520 			inti->io.subchannel_id >> 1 & 0x3,
1521 			inti->io.subchannel_nr);
1522 	isc = int_word_to_isc(inti->io.io_int_word);
1523 	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1524 	list_add_tail(&inti->list, list);
1525 	set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1526 	spin_unlock(&fi->lock);
1527 	return 0;
1528 }
1529 
1530 /*
1531  * Find a destination VCPU for a floating irq and kick it.
1532  */
1533 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1534 {
1535 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1536 	struct kvm_s390_local_interrupt *li;
1537 	struct kvm_vcpu *dst_vcpu;
1538 	int sigcpu, online_vcpus, nr_tries = 0;
1539 
1540 	online_vcpus = atomic_read(&kvm->online_vcpus);
1541 	if (!online_vcpus)
1542 		return;
1543 
1544 	/* find idle VCPUs first, then round robin */
1545 	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1546 	if (sigcpu == online_vcpus) {
1547 		do {
1548 			sigcpu = fi->next_rr_cpu;
1549 			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1550 			/* avoid endless loops if all vcpus are stopped */
1551 			if (nr_tries++ >= online_vcpus)
1552 				return;
1553 		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1554 	}
1555 	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1556 
1557 	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1558 	li = &dst_vcpu->arch.local_int;
1559 	spin_lock(&li->lock);
1560 	switch (type) {
1561 	case KVM_S390_MCHK:
1562 		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1563 		break;
1564 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1565 		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1566 		break;
1567 	default:
1568 		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1569 		break;
1570 	}
1571 	spin_unlock(&li->lock);
1572 	kvm_s390_vcpu_wakeup(dst_vcpu);
1573 }
1574 
1575 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1576 {
1577 	u64 type = READ_ONCE(inti->type);
1578 	int rc;
1579 
1580 	switch (type) {
1581 	case KVM_S390_MCHK:
1582 		rc = __inject_float_mchk(kvm, inti);
1583 		break;
1584 	case KVM_S390_INT_VIRTIO:
1585 		rc = __inject_virtio(kvm, inti);
1586 		break;
1587 	case KVM_S390_INT_SERVICE:
1588 		rc = __inject_service(kvm, inti);
1589 		break;
1590 	case KVM_S390_INT_PFAULT_DONE:
1591 		rc = __inject_pfault_done(kvm, inti);
1592 		break;
1593 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1594 		rc = __inject_io(kvm, inti);
1595 		break;
1596 	default:
1597 		rc = -EINVAL;
1598 	}
1599 	if (rc)
1600 		return rc;
1601 
1602 	__floating_irq_kick(kvm, type);
1603 	return 0;
1604 }
1605 
1606 int kvm_s390_inject_vm(struct kvm *kvm,
1607 		       struct kvm_s390_interrupt *s390int)
1608 {
1609 	struct kvm_s390_interrupt_info *inti;
1610 	int rc;
1611 
1612 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1613 	if (!inti)
1614 		return -ENOMEM;
1615 
1616 	inti->type = s390int->type;
1617 	switch (inti->type) {
1618 	case KVM_S390_INT_VIRTIO:
1619 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1620 			 s390int->parm, s390int->parm64);
1621 		inti->ext.ext_params = s390int->parm;
1622 		inti->ext.ext_params2 = s390int->parm64;
1623 		break;
1624 	case KVM_S390_INT_SERVICE:
1625 		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1626 		inti->ext.ext_params = s390int->parm;
1627 		break;
1628 	case KVM_S390_INT_PFAULT_DONE:
1629 		inti->ext.ext_params2 = s390int->parm64;
1630 		break;
1631 	case KVM_S390_MCHK:
1632 		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1633 			 s390int->parm64);
1634 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1635 		inti->mchk.mcic = s390int->parm64;
1636 		break;
1637 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1638 		inti->io.subchannel_id = s390int->parm >> 16;
1639 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1640 		inti->io.io_int_parm = s390int->parm64 >> 32;
1641 		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1642 		break;
1643 	default:
1644 		kfree(inti);
1645 		return -EINVAL;
1646 	}
1647 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1648 				 2);
1649 
1650 	rc = __inject_vm(kvm, inti);
1651 	if (rc)
1652 		kfree(inti);
1653 	return rc;
1654 }
1655 
1656 int kvm_s390_reinject_io_int(struct kvm *kvm,
1657 			      struct kvm_s390_interrupt_info *inti)
1658 {
1659 	return __inject_vm(kvm, inti);
1660 }
1661 
1662 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1663 		       struct kvm_s390_irq *irq)
1664 {
1665 	irq->type = s390int->type;
1666 	switch (irq->type) {
1667 	case KVM_S390_PROGRAM_INT:
1668 		if (s390int->parm & 0xffff0000)
1669 			return -EINVAL;
1670 		irq->u.pgm.code = s390int->parm;
1671 		break;
1672 	case KVM_S390_SIGP_SET_PREFIX:
1673 		irq->u.prefix.address = s390int->parm;
1674 		break;
1675 	case KVM_S390_SIGP_STOP:
1676 		irq->u.stop.flags = s390int->parm;
1677 		break;
1678 	case KVM_S390_INT_EXTERNAL_CALL:
1679 		if (s390int->parm & 0xffff0000)
1680 			return -EINVAL;
1681 		irq->u.extcall.code = s390int->parm;
1682 		break;
1683 	case KVM_S390_INT_EMERGENCY:
1684 		if (s390int->parm & 0xffff0000)
1685 			return -EINVAL;
1686 		irq->u.emerg.code = s390int->parm;
1687 		break;
1688 	case KVM_S390_MCHK:
1689 		irq->u.mchk.mcic = s390int->parm64;
1690 		break;
1691 	}
1692 	return 0;
1693 }
1694 
1695 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1696 {
1697 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1698 
1699 	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1700 }
1701 
1702 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1703 {
1704 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1705 
1706 	spin_lock(&li->lock);
1707 	li->irq.stop.flags = 0;
1708 	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1709 	spin_unlock(&li->lock);
1710 }
1711 
1712 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1713 {
1714 	int rc;
1715 
1716 	switch (irq->type) {
1717 	case KVM_S390_PROGRAM_INT:
1718 		rc = __inject_prog(vcpu, irq);
1719 		break;
1720 	case KVM_S390_SIGP_SET_PREFIX:
1721 		rc = __inject_set_prefix(vcpu, irq);
1722 		break;
1723 	case KVM_S390_SIGP_STOP:
1724 		rc = __inject_sigp_stop(vcpu, irq);
1725 		break;
1726 	case KVM_S390_RESTART:
1727 		rc = __inject_sigp_restart(vcpu, irq);
1728 		break;
1729 	case KVM_S390_INT_CLOCK_COMP:
1730 		rc = __inject_ckc(vcpu);
1731 		break;
1732 	case KVM_S390_INT_CPU_TIMER:
1733 		rc = __inject_cpu_timer(vcpu);
1734 		break;
1735 	case KVM_S390_INT_EXTERNAL_CALL:
1736 		rc = __inject_extcall(vcpu, irq);
1737 		break;
1738 	case KVM_S390_INT_EMERGENCY:
1739 		rc = __inject_sigp_emergency(vcpu, irq);
1740 		break;
1741 	case KVM_S390_MCHK:
1742 		rc = __inject_mchk(vcpu, irq);
1743 		break;
1744 	case KVM_S390_INT_PFAULT_INIT:
1745 		rc = __inject_pfault_init(vcpu, irq);
1746 		break;
1747 	case KVM_S390_INT_VIRTIO:
1748 	case KVM_S390_INT_SERVICE:
1749 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1750 	default:
1751 		rc = -EINVAL;
1752 	}
1753 
1754 	return rc;
1755 }
1756 
1757 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1758 {
1759 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1760 	int rc;
1761 
1762 	spin_lock(&li->lock);
1763 	rc = do_inject_vcpu(vcpu, irq);
1764 	spin_unlock(&li->lock);
1765 	if (!rc)
1766 		kvm_s390_vcpu_wakeup(vcpu);
1767 	return rc;
1768 }
1769 
1770 static inline void clear_irq_list(struct list_head *_list)
1771 {
1772 	struct kvm_s390_interrupt_info *inti, *n;
1773 
1774 	list_for_each_entry_safe(inti, n, _list, list) {
1775 		list_del(&inti->list);
1776 		kfree(inti);
1777 	}
1778 }
1779 
1780 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1781 		       struct kvm_s390_irq *irq)
1782 {
1783 	irq->type = inti->type;
1784 	switch (inti->type) {
1785 	case KVM_S390_INT_PFAULT_INIT:
1786 	case KVM_S390_INT_PFAULT_DONE:
1787 	case KVM_S390_INT_VIRTIO:
1788 		irq->u.ext = inti->ext;
1789 		break;
1790 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1791 		irq->u.io = inti->io;
1792 		break;
1793 	}
1794 }
1795 
1796 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1797 {
1798 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1799 	int i;
1800 
1801 	spin_lock(&fi->lock);
1802 	fi->pending_irqs = 0;
1803 	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1804 	memset(&fi->mchk, 0, sizeof(fi->mchk));
1805 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1806 		clear_irq_list(&fi->lists[i]);
1807 	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1808 		fi->counters[i] = 0;
1809 	spin_unlock(&fi->lock);
1810 };
1811 
1812 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1813 {
1814 	struct kvm_s390_interrupt_info *inti;
1815 	struct kvm_s390_float_interrupt *fi;
1816 	struct kvm_s390_irq *buf;
1817 	struct kvm_s390_irq *irq;
1818 	int max_irqs;
1819 	int ret = 0;
1820 	int n = 0;
1821 	int i;
1822 
1823 	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1824 		return -EINVAL;
1825 
1826 	/*
1827 	 * We are already using -ENOMEM to signal
1828 	 * userspace it may retry with a bigger buffer,
1829 	 * so we need to use something else for this case
1830 	 */
1831 	buf = vzalloc(len);
1832 	if (!buf)
1833 		return -ENOBUFS;
1834 
1835 	max_irqs = len / sizeof(struct kvm_s390_irq);
1836 
1837 	fi = &kvm->arch.float_int;
1838 	spin_lock(&fi->lock);
1839 	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1840 		list_for_each_entry(inti, &fi->lists[i], list) {
1841 			if (n == max_irqs) {
1842 				/* signal userspace to try again */
1843 				ret = -ENOMEM;
1844 				goto out;
1845 			}
1846 			inti_to_irq(inti, &buf[n]);
1847 			n++;
1848 		}
1849 	}
1850 	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1851 		if (n == max_irqs) {
1852 			/* signal userspace to try again */
1853 			ret = -ENOMEM;
1854 			goto out;
1855 		}
1856 		irq = (struct kvm_s390_irq *) &buf[n];
1857 		irq->type = KVM_S390_INT_SERVICE;
1858 		irq->u.ext = fi->srv_signal;
1859 		n++;
1860 	}
1861 	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1862 		if (n == max_irqs) {
1863 				/* signal userspace to try again */
1864 				ret = -ENOMEM;
1865 				goto out;
1866 		}
1867 		irq = (struct kvm_s390_irq *) &buf[n];
1868 		irq->type = KVM_S390_MCHK;
1869 		irq->u.mchk = fi->mchk;
1870 		n++;
1871 }
1872 
1873 out:
1874 	spin_unlock(&fi->lock);
1875 	if (!ret && n > 0) {
1876 		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1877 			ret = -EFAULT;
1878 	}
1879 	vfree(buf);
1880 
1881 	return ret < 0 ? ret : n;
1882 }
1883 
1884 static int flic_ais_mode_get_all(struct kvm *kvm, struct kvm_device_attr *attr)
1885 {
1886 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1887 	struct kvm_s390_ais_all ais;
1888 
1889 	if (attr->attr < sizeof(ais))
1890 		return -EINVAL;
1891 
1892 	if (!test_kvm_facility(kvm, 72))
1893 		return -ENOTSUPP;
1894 
1895 	mutex_lock(&fi->ais_lock);
1896 	ais.simm = fi->simm;
1897 	ais.nimm = fi->nimm;
1898 	mutex_unlock(&fi->ais_lock);
1899 
1900 	if (copy_to_user((void __user *)attr->addr, &ais, sizeof(ais)))
1901 		return -EFAULT;
1902 
1903 	return 0;
1904 }
1905 
1906 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1907 {
1908 	int r;
1909 
1910 	switch (attr->group) {
1911 	case KVM_DEV_FLIC_GET_ALL_IRQS:
1912 		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1913 					  attr->attr);
1914 		break;
1915 	case KVM_DEV_FLIC_AISM_ALL:
1916 		r = flic_ais_mode_get_all(dev->kvm, attr);
1917 		break;
1918 	default:
1919 		r = -EINVAL;
1920 	}
1921 
1922 	return r;
1923 }
1924 
1925 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1926 				     u64 addr)
1927 {
1928 	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1929 	void *target = NULL;
1930 	void __user *source;
1931 	u64 size;
1932 
1933 	if (get_user(inti->type, (u64 __user *)addr))
1934 		return -EFAULT;
1935 
1936 	switch (inti->type) {
1937 	case KVM_S390_INT_PFAULT_INIT:
1938 	case KVM_S390_INT_PFAULT_DONE:
1939 	case KVM_S390_INT_VIRTIO:
1940 	case KVM_S390_INT_SERVICE:
1941 		target = (void *) &inti->ext;
1942 		source = &uptr->u.ext;
1943 		size = sizeof(inti->ext);
1944 		break;
1945 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1946 		target = (void *) &inti->io;
1947 		source = &uptr->u.io;
1948 		size = sizeof(inti->io);
1949 		break;
1950 	case KVM_S390_MCHK:
1951 		target = (void *) &inti->mchk;
1952 		source = &uptr->u.mchk;
1953 		size = sizeof(inti->mchk);
1954 		break;
1955 	default:
1956 		return -EINVAL;
1957 	}
1958 
1959 	if (copy_from_user(target, source, size))
1960 		return -EFAULT;
1961 
1962 	return 0;
1963 }
1964 
1965 static int enqueue_floating_irq(struct kvm_device *dev,
1966 				struct kvm_device_attr *attr)
1967 {
1968 	struct kvm_s390_interrupt_info *inti = NULL;
1969 	int r = 0;
1970 	int len = attr->attr;
1971 
1972 	if (len % sizeof(struct kvm_s390_irq) != 0)
1973 		return -EINVAL;
1974 	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1975 		return -EINVAL;
1976 
1977 	while (len >= sizeof(struct kvm_s390_irq)) {
1978 		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1979 		if (!inti)
1980 			return -ENOMEM;
1981 
1982 		r = copy_irq_from_user(inti, attr->addr);
1983 		if (r) {
1984 			kfree(inti);
1985 			return r;
1986 		}
1987 		r = __inject_vm(dev->kvm, inti);
1988 		if (r) {
1989 			kfree(inti);
1990 			return r;
1991 		}
1992 		len -= sizeof(struct kvm_s390_irq);
1993 		attr->addr += sizeof(struct kvm_s390_irq);
1994 	}
1995 
1996 	return r;
1997 }
1998 
1999 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
2000 {
2001 	if (id >= MAX_S390_IO_ADAPTERS)
2002 		return NULL;
2003 	return kvm->arch.adapters[id];
2004 }
2005 
2006 static int register_io_adapter(struct kvm_device *dev,
2007 			       struct kvm_device_attr *attr)
2008 {
2009 	struct s390_io_adapter *adapter;
2010 	struct kvm_s390_io_adapter adapter_info;
2011 
2012 	if (copy_from_user(&adapter_info,
2013 			   (void __user *)attr->addr, sizeof(adapter_info)))
2014 		return -EFAULT;
2015 
2016 	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
2017 	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
2018 		return -EINVAL;
2019 
2020 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
2021 	if (!adapter)
2022 		return -ENOMEM;
2023 
2024 	INIT_LIST_HEAD(&adapter->maps);
2025 	init_rwsem(&adapter->maps_lock);
2026 	atomic_set(&adapter->nr_maps, 0);
2027 	adapter->id = adapter_info.id;
2028 	adapter->isc = adapter_info.isc;
2029 	adapter->maskable = adapter_info.maskable;
2030 	adapter->masked = false;
2031 	adapter->swap = adapter_info.swap;
2032 	adapter->suppressible = (adapter_info.flags) &
2033 				KVM_S390_ADAPTER_SUPPRESSIBLE;
2034 	dev->kvm->arch.adapters[adapter->id] = adapter;
2035 
2036 	return 0;
2037 }
2038 
2039 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
2040 {
2041 	int ret;
2042 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2043 
2044 	if (!adapter || !adapter->maskable)
2045 		return -EINVAL;
2046 	ret = adapter->masked;
2047 	adapter->masked = masked;
2048 	return ret;
2049 }
2050 
2051 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
2052 {
2053 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2054 	struct s390_map_info *map;
2055 	int ret;
2056 
2057 	if (!adapter || !addr)
2058 		return -EINVAL;
2059 
2060 	map = kzalloc(sizeof(*map), GFP_KERNEL);
2061 	if (!map) {
2062 		ret = -ENOMEM;
2063 		goto out;
2064 	}
2065 	INIT_LIST_HEAD(&map->list);
2066 	map->guest_addr = addr;
2067 	map->addr = gmap_translate(kvm->arch.gmap, addr);
2068 	if (map->addr == -EFAULT) {
2069 		ret = -EFAULT;
2070 		goto out;
2071 	}
2072 	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
2073 	if (ret < 0)
2074 		goto out;
2075 	BUG_ON(ret != 1);
2076 	down_write(&adapter->maps_lock);
2077 	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
2078 		list_add_tail(&map->list, &adapter->maps);
2079 		ret = 0;
2080 	} else {
2081 		put_page(map->page);
2082 		ret = -EINVAL;
2083 	}
2084 	up_write(&adapter->maps_lock);
2085 out:
2086 	if (ret)
2087 		kfree(map);
2088 	return ret;
2089 }
2090 
2091 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
2092 {
2093 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2094 	struct s390_map_info *map, *tmp;
2095 	int found = 0;
2096 
2097 	if (!adapter || !addr)
2098 		return -EINVAL;
2099 
2100 	down_write(&adapter->maps_lock);
2101 	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
2102 		if (map->guest_addr == addr) {
2103 			found = 1;
2104 			atomic_dec(&adapter->nr_maps);
2105 			list_del(&map->list);
2106 			put_page(map->page);
2107 			kfree(map);
2108 			break;
2109 		}
2110 	}
2111 	up_write(&adapter->maps_lock);
2112 
2113 	return found ? 0 : -EINVAL;
2114 }
2115 
2116 void kvm_s390_destroy_adapters(struct kvm *kvm)
2117 {
2118 	int i;
2119 	struct s390_map_info *map, *tmp;
2120 
2121 	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2122 		if (!kvm->arch.adapters[i])
2123 			continue;
2124 		list_for_each_entry_safe(map, tmp,
2125 					 &kvm->arch.adapters[i]->maps, list) {
2126 			list_del(&map->list);
2127 			put_page(map->page);
2128 			kfree(map);
2129 		}
2130 		kfree(kvm->arch.adapters[i]);
2131 	}
2132 }
2133 
2134 static int modify_io_adapter(struct kvm_device *dev,
2135 			     struct kvm_device_attr *attr)
2136 {
2137 	struct kvm_s390_io_adapter_req req;
2138 	struct s390_io_adapter *adapter;
2139 	int ret;
2140 
2141 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2142 		return -EFAULT;
2143 
2144 	adapter = get_io_adapter(dev->kvm, req.id);
2145 	if (!adapter)
2146 		return -EINVAL;
2147 	switch (req.type) {
2148 	case KVM_S390_IO_ADAPTER_MASK:
2149 		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2150 		if (ret > 0)
2151 			ret = 0;
2152 		break;
2153 	case KVM_S390_IO_ADAPTER_MAP:
2154 		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2155 		break;
2156 	case KVM_S390_IO_ADAPTER_UNMAP:
2157 		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2158 		break;
2159 	default:
2160 		ret = -EINVAL;
2161 	}
2162 
2163 	return ret;
2164 }
2165 
2166 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2167 
2168 {
2169 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2170 	u32 schid;
2171 
2172 	if (attr->flags)
2173 		return -EINVAL;
2174 	if (attr->attr != sizeof(schid))
2175 		return -EINVAL;
2176 	if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2177 		return -EFAULT;
2178 	kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2179 	/*
2180 	 * If userspace is conforming to the architecture, we can have at most
2181 	 * one pending I/O interrupt per subchannel, so this is effectively a
2182 	 * clear all.
2183 	 */
2184 	return 0;
2185 }
2186 
2187 static int modify_ais_mode(struct kvm *kvm, struct kvm_device_attr *attr)
2188 {
2189 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2190 	struct kvm_s390_ais_req req;
2191 	int ret = 0;
2192 
2193 	if (!test_kvm_facility(kvm, 72))
2194 		return -ENOTSUPP;
2195 
2196 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2197 		return -EFAULT;
2198 
2199 	if (req.isc > MAX_ISC)
2200 		return -EINVAL;
2201 
2202 	trace_kvm_s390_modify_ais_mode(req.isc,
2203 				       (fi->simm & AIS_MODE_MASK(req.isc)) ?
2204 				       (fi->nimm & AIS_MODE_MASK(req.isc)) ?
2205 				       2 : KVM_S390_AIS_MODE_SINGLE :
2206 				       KVM_S390_AIS_MODE_ALL, req.mode);
2207 
2208 	mutex_lock(&fi->ais_lock);
2209 	switch (req.mode) {
2210 	case KVM_S390_AIS_MODE_ALL:
2211 		fi->simm &= ~AIS_MODE_MASK(req.isc);
2212 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2213 		break;
2214 	case KVM_S390_AIS_MODE_SINGLE:
2215 		fi->simm |= AIS_MODE_MASK(req.isc);
2216 		fi->nimm &= ~AIS_MODE_MASK(req.isc);
2217 		break;
2218 	default:
2219 		ret = -EINVAL;
2220 	}
2221 	mutex_unlock(&fi->ais_lock);
2222 
2223 	return ret;
2224 }
2225 
2226 static int kvm_s390_inject_airq(struct kvm *kvm,
2227 				struct s390_io_adapter *adapter)
2228 {
2229 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2230 	struct kvm_s390_interrupt s390int = {
2231 		.type = KVM_S390_INT_IO(1, 0, 0, 0),
2232 		.parm = 0,
2233 		.parm64 = (adapter->isc << 27) | 0x80000000,
2234 	};
2235 	int ret = 0;
2236 
2237 	if (!test_kvm_facility(kvm, 72) || !adapter->suppressible)
2238 		return kvm_s390_inject_vm(kvm, &s390int);
2239 
2240 	mutex_lock(&fi->ais_lock);
2241 	if (fi->nimm & AIS_MODE_MASK(adapter->isc)) {
2242 		trace_kvm_s390_airq_suppressed(adapter->id, adapter->isc);
2243 		goto out;
2244 	}
2245 
2246 	ret = kvm_s390_inject_vm(kvm, &s390int);
2247 	if (!ret && (fi->simm & AIS_MODE_MASK(adapter->isc))) {
2248 		fi->nimm |= AIS_MODE_MASK(adapter->isc);
2249 		trace_kvm_s390_modify_ais_mode(adapter->isc,
2250 					       KVM_S390_AIS_MODE_SINGLE, 2);
2251 	}
2252 out:
2253 	mutex_unlock(&fi->ais_lock);
2254 	return ret;
2255 }
2256 
2257 static int flic_inject_airq(struct kvm *kvm, struct kvm_device_attr *attr)
2258 {
2259 	unsigned int id = attr->attr;
2260 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2261 
2262 	if (!adapter)
2263 		return -EINVAL;
2264 
2265 	return kvm_s390_inject_airq(kvm, adapter);
2266 }
2267 
2268 static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr)
2269 {
2270 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2271 	struct kvm_s390_ais_all ais;
2272 
2273 	if (!test_kvm_facility(kvm, 72))
2274 		return -ENOTSUPP;
2275 
2276 	if (copy_from_user(&ais, (void __user *)attr->addr, sizeof(ais)))
2277 		return -EFAULT;
2278 
2279 	mutex_lock(&fi->ais_lock);
2280 	fi->simm = ais.simm;
2281 	fi->nimm = ais.nimm;
2282 	mutex_unlock(&fi->ais_lock);
2283 
2284 	return 0;
2285 }
2286 
2287 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2288 {
2289 	int r = 0;
2290 	unsigned int i;
2291 	struct kvm_vcpu *vcpu;
2292 
2293 	switch (attr->group) {
2294 	case KVM_DEV_FLIC_ENQUEUE:
2295 		r = enqueue_floating_irq(dev, attr);
2296 		break;
2297 	case KVM_DEV_FLIC_CLEAR_IRQS:
2298 		kvm_s390_clear_float_irqs(dev->kvm);
2299 		break;
2300 	case KVM_DEV_FLIC_APF_ENABLE:
2301 		dev->kvm->arch.gmap->pfault_enabled = 1;
2302 		break;
2303 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2304 		dev->kvm->arch.gmap->pfault_enabled = 0;
2305 		/*
2306 		 * Make sure no async faults are in transition when
2307 		 * clearing the queues. So we don't need to worry
2308 		 * about late coming workers.
2309 		 */
2310 		synchronize_srcu(&dev->kvm->srcu);
2311 		kvm_for_each_vcpu(i, vcpu, dev->kvm)
2312 			kvm_clear_async_pf_completion_queue(vcpu);
2313 		break;
2314 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2315 		r = register_io_adapter(dev, attr);
2316 		break;
2317 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2318 		r = modify_io_adapter(dev, attr);
2319 		break;
2320 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2321 		r = clear_io_irq(dev->kvm, attr);
2322 		break;
2323 	case KVM_DEV_FLIC_AISM:
2324 		r = modify_ais_mode(dev->kvm, attr);
2325 		break;
2326 	case KVM_DEV_FLIC_AIRQ_INJECT:
2327 		r = flic_inject_airq(dev->kvm, attr);
2328 		break;
2329 	case KVM_DEV_FLIC_AISM_ALL:
2330 		r = flic_ais_mode_set_all(dev->kvm, attr);
2331 		break;
2332 	default:
2333 		r = -EINVAL;
2334 	}
2335 
2336 	return r;
2337 }
2338 
2339 static int flic_has_attr(struct kvm_device *dev,
2340 			     struct kvm_device_attr *attr)
2341 {
2342 	switch (attr->group) {
2343 	case KVM_DEV_FLIC_GET_ALL_IRQS:
2344 	case KVM_DEV_FLIC_ENQUEUE:
2345 	case KVM_DEV_FLIC_CLEAR_IRQS:
2346 	case KVM_DEV_FLIC_APF_ENABLE:
2347 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2348 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2349 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2350 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2351 	case KVM_DEV_FLIC_AISM:
2352 	case KVM_DEV_FLIC_AIRQ_INJECT:
2353 	case KVM_DEV_FLIC_AISM_ALL:
2354 		return 0;
2355 	}
2356 	return -ENXIO;
2357 }
2358 
2359 static int flic_create(struct kvm_device *dev, u32 type)
2360 {
2361 	if (!dev)
2362 		return -EINVAL;
2363 	if (dev->kvm->arch.flic)
2364 		return -EINVAL;
2365 	dev->kvm->arch.flic = dev;
2366 	return 0;
2367 }
2368 
2369 static void flic_destroy(struct kvm_device *dev)
2370 {
2371 	dev->kvm->arch.flic = NULL;
2372 	kfree(dev);
2373 }
2374 
2375 /* s390 floating irq controller (flic) */
2376 struct kvm_device_ops kvm_flic_ops = {
2377 	.name = "kvm-flic",
2378 	.get_attr = flic_get_attr,
2379 	.set_attr = flic_set_attr,
2380 	.has_attr = flic_has_attr,
2381 	.create = flic_create,
2382 	.destroy = flic_destroy,
2383 };
2384 
2385 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2386 {
2387 	unsigned long bit;
2388 
2389 	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2390 
2391 	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2392 }
2393 
2394 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2395 					  u64 addr)
2396 {
2397 	struct s390_map_info *map;
2398 
2399 	if (!adapter)
2400 		return NULL;
2401 
2402 	list_for_each_entry(map, &adapter->maps, list) {
2403 		if (map->guest_addr == addr)
2404 			return map;
2405 	}
2406 	return NULL;
2407 }
2408 
2409 static int adapter_indicators_set(struct kvm *kvm,
2410 				  struct s390_io_adapter *adapter,
2411 				  struct kvm_s390_adapter_int *adapter_int)
2412 {
2413 	unsigned long bit;
2414 	int summary_set, idx;
2415 	struct s390_map_info *info;
2416 	void *map;
2417 
2418 	info = get_map_info(adapter, adapter_int->ind_addr);
2419 	if (!info)
2420 		return -1;
2421 	map = page_address(info->page);
2422 	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2423 	set_bit(bit, map);
2424 	idx = srcu_read_lock(&kvm->srcu);
2425 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2426 	set_page_dirty_lock(info->page);
2427 	info = get_map_info(adapter, adapter_int->summary_addr);
2428 	if (!info) {
2429 		srcu_read_unlock(&kvm->srcu, idx);
2430 		return -1;
2431 	}
2432 	map = page_address(info->page);
2433 	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2434 			  adapter->swap);
2435 	summary_set = test_and_set_bit(bit, map);
2436 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2437 	set_page_dirty_lock(info->page);
2438 	srcu_read_unlock(&kvm->srcu, idx);
2439 	return summary_set ? 0 : 1;
2440 }
2441 
2442 /*
2443  * < 0 - not injected due to error
2444  * = 0 - coalesced, summary indicator already active
2445  * > 0 - injected interrupt
2446  */
2447 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2448 			   struct kvm *kvm, int irq_source_id, int level,
2449 			   bool line_status)
2450 {
2451 	int ret;
2452 	struct s390_io_adapter *adapter;
2453 
2454 	/* We're only interested in the 0->1 transition. */
2455 	if (!level)
2456 		return 0;
2457 	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2458 	if (!adapter)
2459 		return -1;
2460 	down_read(&adapter->maps_lock);
2461 	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2462 	up_read(&adapter->maps_lock);
2463 	if ((ret > 0) && !adapter->masked) {
2464 		ret = kvm_s390_inject_airq(kvm, adapter);
2465 		if (ret == 0)
2466 			ret = 1;
2467 	}
2468 	return ret;
2469 }
2470 
2471 /*
2472  * Inject the machine check to the guest.
2473  */
2474 void kvm_s390_reinject_machine_check(struct kvm_vcpu *vcpu,
2475 				     struct mcck_volatile_info *mcck_info)
2476 {
2477 	struct kvm_s390_interrupt_info inti;
2478 	struct kvm_s390_irq irq;
2479 	struct kvm_s390_mchk_info *mchk;
2480 	union mci mci;
2481 	__u64 cr14 = 0;         /* upper bits are not used */
2482 
2483 	mci.val = mcck_info->mcic;
2484 	if (mci.sr)
2485 		cr14 |= MCCK_CR14_RECOVERY_SUB_MASK;
2486 	if (mci.dg)
2487 		cr14 |= MCCK_CR14_DEGRAD_SUB_MASK;
2488 	if (mci.w)
2489 		cr14 |= MCCK_CR14_WARN_SUB_MASK;
2490 
2491 	mchk = mci.ck ? &inti.mchk : &irq.u.mchk;
2492 	mchk->cr14 = cr14;
2493 	mchk->mcic = mcck_info->mcic;
2494 	mchk->ext_damage_code = mcck_info->ext_damage_code;
2495 	mchk->failing_storage_address = mcck_info->failing_storage_address;
2496 	if (mci.ck) {
2497 		/* Inject the floating machine check */
2498 		inti.type = KVM_S390_MCHK;
2499 		WARN_ON_ONCE(__inject_vm(vcpu->kvm, &inti));
2500 	} else {
2501 		/* Inject the machine check to specified vcpu */
2502 		irq.type = KVM_S390_MCHK;
2503 		WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &irq));
2504 	}
2505 }
2506 
2507 int kvm_set_routing_entry(struct kvm *kvm,
2508 			  struct kvm_kernel_irq_routing_entry *e,
2509 			  const struct kvm_irq_routing_entry *ue)
2510 {
2511 	int ret;
2512 
2513 	switch (ue->type) {
2514 	case KVM_IRQ_ROUTING_S390_ADAPTER:
2515 		e->set = set_adapter_int;
2516 		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2517 		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2518 		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2519 		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2520 		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2521 		ret = 0;
2522 		break;
2523 	default:
2524 		ret = -EINVAL;
2525 	}
2526 
2527 	return ret;
2528 }
2529 
2530 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2531 		int irq_source_id, int level, bool line_status)
2532 {
2533 	return -EINVAL;
2534 }
2535 
2536 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2537 {
2538 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2539 	struct kvm_s390_irq *buf;
2540 	int r = 0;
2541 	int n;
2542 
2543 	buf = vmalloc(len);
2544 	if (!buf)
2545 		return -ENOMEM;
2546 
2547 	if (copy_from_user((void *) buf, irqstate, len)) {
2548 		r = -EFAULT;
2549 		goto out_free;
2550 	}
2551 
2552 	/*
2553 	 * Don't allow setting the interrupt state
2554 	 * when there are already interrupts pending
2555 	 */
2556 	spin_lock(&li->lock);
2557 	if (li->pending_irqs) {
2558 		r = -EBUSY;
2559 		goto out_unlock;
2560 	}
2561 
2562 	for (n = 0; n < len / sizeof(*buf); n++) {
2563 		r = do_inject_vcpu(vcpu, &buf[n]);
2564 		if (r)
2565 			break;
2566 	}
2567 
2568 out_unlock:
2569 	spin_unlock(&li->lock);
2570 out_free:
2571 	vfree(buf);
2572 
2573 	return r;
2574 }
2575 
2576 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2577 			    struct kvm_s390_irq *irq,
2578 			    unsigned long irq_type)
2579 {
2580 	switch (irq_type) {
2581 	case IRQ_PEND_MCHK_EX:
2582 	case IRQ_PEND_MCHK_REP:
2583 		irq->type = KVM_S390_MCHK;
2584 		irq->u.mchk = li->irq.mchk;
2585 		break;
2586 	case IRQ_PEND_PROG:
2587 		irq->type = KVM_S390_PROGRAM_INT;
2588 		irq->u.pgm = li->irq.pgm;
2589 		break;
2590 	case IRQ_PEND_PFAULT_INIT:
2591 		irq->type = KVM_S390_INT_PFAULT_INIT;
2592 		irq->u.ext = li->irq.ext;
2593 		break;
2594 	case IRQ_PEND_EXT_EXTERNAL:
2595 		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2596 		irq->u.extcall = li->irq.extcall;
2597 		break;
2598 	case IRQ_PEND_EXT_CLOCK_COMP:
2599 		irq->type = KVM_S390_INT_CLOCK_COMP;
2600 		break;
2601 	case IRQ_PEND_EXT_CPU_TIMER:
2602 		irq->type = KVM_S390_INT_CPU_TIMER;
2603 		break;
2604 	case IRQ_PEND_SIGP_STOP:
2605 		irq->type = KVM_S390_SIGP_STOP;
2606 		irq->u.stop = li->irq.stop;
2607 		break;
2608 	case IRQ_PEND_RESTART:
2609 		irq->type = KVM_S390_RESTART;
2610 		break;
2611 	case IRQ_PEND_SET_PREFIX:
2612 		irq->type = KVM_S390_SIGP_SET_PREFIX;
2613 		irq->u.prefix = li->irq.prefix;
2614 		break;
2615 	}
2616 }
2617 
2618 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2619 {
2620 	int scn;
2621 	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2622 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2623 	unsigned long pending_irqs;
2624 	struct kvm_s390_irq irq;
2625 	unsigned long irq_type;
2626 	int cpuaddr;
2627 	int n = 0;
2628 
2629 	spin_lock(&li->lock);
2630 	pending_irqs = li->pending_irqs;
2631 	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2632 	       sizeof(sigp_emerg_pending));
2633 	spin_unlock(&li->lock);
2634 
2635 	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2636 		memset(&irq, 0, sizeof(irq));
2637 		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2638 			continue;
2639 		if (n + sizeof(irq) > len)
2640 			return -ENOBUFS;
2641 		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2642 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2643 			return -EFAULT;
2644 		n += sizeof(irq);
2645 	}
2646 
2647 	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2648 		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2649 			memset(&irq, 0, sizeof(irq));
2650 			if (n + sizeof(irq) > len)
2651 				return -ENOBUFS;
2652 			irq.type = KVM_S390_INT_EMERGENCY;
2653 			irq.u.emerg.code = cpuaddr;
2654 			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2655 				return -EFAULT;
2656 			n += sizeof(irq);
2657 		}
2658 	}
2659 
2660 	if (sca_ext_call_pending(vcpu, &scn)) {
2661 		if (n + sizeof(irq) > len)
2662 			return -ENOBUFS;
2663 		memset(&irq, 0, sizeof(irq));
2664 		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2665 		irq.u.extcall.code = scn;
2666 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2667 			return -EFAULT;
2668 		n += sizeof(irq);
2669 	}
2670 
2671 	return n;
2672 }
2673