xref: /openbmc/linux/arch/s390/kvm/interrupt.c (revision 8730046c)
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 	freg_t fprs[NUM_FPRS];
414 	union mci mci;
415 	int rc;
416 
417 	mci.val = mchk->mcic;
418 	/* take care of lazy register loading */
419 	save_fpu_regs();
420 	save_access_regs(vcpu->run->s.regs.acrs);
421 
422 	/* Extended save area */
423 	rc = read_guest_lc(vcpu, __LC_VX_SAVE_AREA_ADDR, &ext_sa_addr,
424 			    sizeof(unsigned long));
425 	/* Only bits 0-53 are used for address formation */
426 	ext_sa_addr &= ~0x3ffUL;
427 	if (!rc && mci.vr && ext_sa_addr && test_kvm_facility(vcpu->kvm, 129)) {
428 		if (write_guest_abs(vcpu, ext_sa_addr, vcpu->run->s.regs.vrs,
429 				    512))
430 			mci.vr = 0;
431 	} else {
432 		mci.vr = 0;
433 	}
434 
435 	/* General interruption information */
436 	rc |= put_guest_lc(vcpu, 1, (u8 __user *) __LC_AR_MODE_ID);
437 	rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
438 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
439 	rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
440 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
441 	rc |= put_guest_lc(vcpu, mci.val, (u64 __user *) __LC_MCCK_CODE);
442 
443 	/* Register-save areas */
444 	if (MACHINE_HAS_VX) {
445 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
446 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA, fprs, 128);
447 	} else {
448 		rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA,
449 				     vcpu->run->s.regs.fprs, 128);
450 	}
451 	rc |= write_guest_lc(vcpu, __LC_GPREGS_SAVE_AREA,
452 			     vcpu->run->s.regs.gprs, 128);
453 	rc |= put_guest_lc(vcpu, current->thread.fpu.fpc,
454 			   (u32 __user *) __LC_FP_CREG_SAVE_AREA);
455 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->todpr,
456 			   (u32 __user *) __LC_TOD_PROGREG_SAVE_AREA);
457 	rc |= put_guest_lc(vcpu, kvm_s390_get_cpu_timer(vcpu),
458 			   (u64 __user *) __LC_CPU_TIMER_SAVE_AREA);
459 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->ckc >> 8,
460 			   (u64 __user *) __LC_CLOCK_COMP_SAVE_AREA);
461 	rc |= write_guest_lc(vcpu, __LC_AREGS_SAVE_AREA,
462 			     &vcpu->run->s.regs.acrs, 64);
463 	rc |= write_guest_lc(vcpu, __LC_CREGS_SAVE_AREA,
464 			     &vcpu->arch.sie_block->gcr, 128);
465 
466 	/* Extended interruption information */
467 	rc |= put_guest_lc(vcpu, mchk->ext_damage_code,
468 			   (u32 __user *) __LC_EXT_DAMAGE_CODE);
469 	rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
470 			   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
471 	rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA, &mchk->fixed_logout,
472 			     sizeof(mchk->fixed_logout));
473 	return rc ? -EFAULT : 0;
474 }
475 
476 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
477 {
478 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
479 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
480 	struct kvm_s390_mchk_info mchk = {};
481 	int deliver = 0;
482 	int rc = 0;
483 
484 	spin_lock(&fi->lock);
485 	spin_lock(&li->lock);
486 	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
487 	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
488 		/*
489 		 * If there was an exigent machine check pending, then any
490 		 * repressible machine checks that might have been pending
491 		 * are indicated along with it, so always clear bits for
492 		 * repressible and exigent interrupts
493 		 */
494 		mchk = li->irq.mchk;
495 		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
496 		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
497 		memset(&li->irq.mchk, 0, sizeof(mchk));
498 		deliver = 1;
499 	}
500 	/*
501 	 * We indicate floating repressible conditions along with
502 	 * other pending conditions. Channel Report Pending and Channel
503 	 * Subsystem damage are the only two and and are indicated by
504 	 * bits in mcic and masked in cr14.
505 	 */
506 	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
507 		mchk.mcic |= fi->mchk.mcic;
508 		mchk.cr14 |= fi->mchk.cr14;
509 		memset(&fi->mchk, 0, sizeof(mchk));
510 		deliver = 1;
511 	}
512 	spin_unlock(&li->lock);
513 	spin_unlock(&fi->lock);
514 
515 	if (deliver) {
516 		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
517 			   mchk.mcic);
518 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
519 						 KVM_S390_MCHK,
520 						 mchk.cr14, mchk.mcic);
521 		rc = __write_machine_check(vcpu, &mchk);
522 	}
523 	return rc;
524 }
525 
526 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
527 {
528 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
529 	int rc;
530 
531 	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
532 	vcpu->stat.deliver_restart_signal++;
533 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
534 
535 	rc  = write_guest_lc(vcpu,
536 			     offsetof(struct lowcore, restart_old_psw),
537 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
538 	rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
539 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
540 	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
541 	return rc ? -EFAULT : 0;
542 }
543 
544 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
545 {
546 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
547 	struct kvm_s390_prefix_info prefix;
548 
549 	spin_lock(&li->lock);
550 	prefix = li->irq.prefix;
551 	li->irq.prefix.address = 0;
552 	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
553 	spin_unlock(&li->lock);
554 
555 	vcpu->stat.deliver_prefix_signal++;
556 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
557 					 KVM_S390_SIGP_SET_PREFIX,
558 					 prefix.address, 0);
559 
560 	kvm_s390_set_prefix(vcpu, prefix.address);
561 	return 0;
562 }
563 
564 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
565 {
566 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
567 	int rc;
568 	int cpu_addr;
569 
570 	spin_lock(&li->lock);
571 	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
572 	clear_bit(cpu_addr, li->sigp_emerg_pending);
573 	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
574 		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
575 	spin_unlock(&li->lock);
576 
577 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
578 	vcpu->stat.deliver_emergency_signal++;
579 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
580 					 cpu_addr, 0);
581 
582 	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
583 			   (u16 *)__LC_EXT_INT_CODE);
584 	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
585 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
586 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
587 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
588 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
589 	return rc ? -EFAULT : 0;
590 }
591 
592 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
593 {
594 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
595 	struct kvm_s390_extcall_info extcall;
596 	int rc;
597 
598 	spin_lock(&li->lock);
599 	extcall = li->irq.extcall;
600 	li->irq.extcall.code = 0;
601 	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
602 	spin_unlock(&li->lock);
603 
604 	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
605 	vcpu->stat.deliver_external_call++;
606 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
607 					 KVM_S390_INT_EXTERNAL_CALL,
608 					 extcall.code, 0);
609 
610 	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
611 			   (u16 *)__LC_EXT_INT_CODE);
612 	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
613 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
614 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
615 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
616 			    sizeof(psw_t));
617 	return rc ? -EFAULT : 0;
618 }
619 
620 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
621 {
622 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
623 	struct kvm_s390_pgm_info pgm_info;
624 	int rc = 0, nullifying = false;
625 	u16 ilen;
626 
627 	spin_lock(&li->lock);
628 	pgm_info = li->irq.pgm;
629 	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
630 	memset(&li->irq.pgm, 0, sizeof(pgm_info));
631 	spin_unlock(&li->lock);
632 
633 	ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
634 	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
635 		   pgm_info.code, ilen);
636 	vcpu->stat.deliver_program_int++;
637 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
638 					 pgm_info.code, 0);
639 
640 	switch (pgm_info.code & ~PGM_PER) {
641 	case PGM_AFX_TRANSLATION:
642 	case PGM_ASX_TRANSLATION:
643 	case PGM_EX_TRANSLATION:
644 	case PGM_LFX_TRANSLATION:
645 	case PGM_LSTE_SEQUENCE:
646 	case PGM_LSX_TRANSLATION:
647 	case PGM_LX_TRANSLATION:
648 	case PGM_PRIMARY_AUTHORITY:
649 	case PGM_SECONDARY_AUTHORITY:
650 		nullifying = true;
651 		/* fall through */
652 	case PGM_SPACE_SWITCH:
653 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
654 				  (u64 *)__LC_TRANS_EXC_CODE);
655 		break;
656 	case PGM_ALEN_TRANSLATION:
657 	case PGM_ALE_SEQUENCE:
658 	case PGM_ASTE_INSTANCE:
659 	case PGM_ASTE_SEQUENCE:
660 	case PGM_ASTE_VALIDITY:
661 	case PGM_EXTENDED_AUTHORITY:
662 		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
663 				  (u8 *)__LC_EXC_ACCESS_ID);
664 		nullifying = true;
665 		break;
666 	case PGM_ASCE_TYPE:
667 	case PGM_PAGE_TRANSLATION:
668 	case PGM_REGION_FIRST_TRANS:
669 	case PGM_REGION_SECOND_TRANS:
670 	case PGM_REGION_THIRD_TRANS:
671 	case PGM_SEGMENT_TRANSLATION:
672 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
673 				  (u64 *)__LC_TRANS_EXC_CODE);
674 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
675 				   (u8 *)__LC_EXC_ACCESS_ID);
676 		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
677 				   (u8 *)__LC_OP_ACCESS_ID);
678 		nullifying = true;
679 		break;
680 	case PGM_MONITOR:
681 		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
682 				  (u16 *)__LC_MON_CLASS_NR);
683 		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
684 				   (u64 *)__LC_MON_CODE);
685 		break;
686 	case PGM_VECTOR_PROCESSING:
687 	case PGM_DATA:
688 		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
689 				  (u32 *)__LC_DATA_EXC_CODE);
690 		break;
691 	case PGM_PROTECTION:
692 		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
693 				  (u64 *)__LC_TRANS_EXC_CODE);
694 		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
695 				   (u8 *)__LC_EXC_ACCESS_ID);
696 		break;
697 	case PGM_STACK_FULL:
698 	case PGM_STACK_EMPTY:
699 	case PGM_STACK_SPECIFICATION:
700 	case PGM_STACK_TYPE:
701 	case PGM_STACK_OPERATION:
702 	case PGM_TRACE_TABEL:
703 	case PGM_CRYPTO_OPERATION:
704 		nullifying = true;
705 		break;
706 	}
707 
708 	if (pgm_info.code & PGM_PER) {
709 		rc |= put_guest_lc(vcpu, pgm_info.per_code,
710 				   (u8 *) __LC_PER_CODE);
711 		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
712 				   (u8 *)__LC_PER_ATMID);
713 		rc |= put_guest_lc(vcpu, pgm_info.per_address,
714 				   (u64 *) __LC_PER_ADDRESS);
715 		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
716 				   (u8 *) __LC_PER_ACCESS_ID);
717 	}
718 
719 	if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
720 		kvm_s390_rewind_psw(vcpu, ilen);
721 
722 	/* bit 1+2 of the target are the ilc, so we can directly use ilen */
723 	rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
724 	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
725 				 (u64 *) __LC_LAST_BREAK);
726 	rc |= put_guest_lc(vcpu, pgm_info.code,
727 			   (u16 *)__LC_PGM_INT_CODE);
728 	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
729 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
730 	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
731 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
732 	return rc ? -EFAULT : 0;
733 }
734 
735 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
736 {
737 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
738 	struct kvm_s390_ext_info ext;
739 	int rc = 0;
740 
741 	spin_lock(&fi->lock);
742 	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
743 		spin_unlock(&fi->lock);
744 		return 0;
745 	}
746 	ext = fi->srv_signal;
747 	memset(&fi->srv_signal, 0, sizeof(ext));
748 	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
749 	spin_unlock(&fi->lock);
750 
751 	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
752 		   ext.ext_params);
753 	vcpu->stat.deliver_service_signal++;
754 	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
755 					 ext.ext_params, 0);
756 
757 	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
758 	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
759 	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
760 			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
761 	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
762 			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
763 	rc |= put_guest_lc(vcpu, ext.ext_params,
764 			   (u32 *)__LC_EXT_PARAMS);
765 
766 	return rc ? -EFAULT : 0;
767 }
768 
769 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
770 {
771 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
772 	struct kvm_s390_interrupt_info *inti;
773 	int rc = 0;
774 
775 	spin_lock(&fi->lock);
776 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
777 					struct kvm_s390_interrupt_info,
778 					list);
779 	if (inti) {
780 		list_del(&inti->list);
781 		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
782 	}
783 	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
784 		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
785 	spin_unlock(&fi->lock);
786 
787 	if (inti) {
788 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
789 						 KVM_S390_INT_PFAULT_DONE, 0,
790 						 inti->ext.ext_params2);
791 		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
792 			   inti->ext.ext_params2);
793 
794 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
795 				(u16 *)__LC_EXT_INT_CODE);
796 		rc |= put_guest_lc(vcpu, PFAULT_DONE,
797 				(u16 *)__LC_EXT_CPU_ADDR);
798 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
799 				&vcpu->arch.sie_block->gpsw,
800 				sizeof(psw_t));
801 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
802 				&vcpu->arch.sie_block->gpsw,
803 				sizeof(psw_t));
804 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
805 				(u64 *)__LC_EXT_PARAMS2);
806 		kfree(inti);
807 	}
808 	return rc ? -EFAULT : 0;
809 }
810 
811 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
812 {
813 	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
814 	struct kvm_s390_interrupt_info *inti;
815 	int rc = 0;
816 
817 	spin_lock(&fi->lock);
818 	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
819 					struct kvm_s390_interrupt_info,
820 					list);
821 	if (inti) {
822 		VCPU_EVENT(vcpu, 4,
823 			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
824 			   inti->ext.ext_params, inti->ext.ext_params2);
825 		vcpu->stat.deliver_virtio_interrupt++;
826 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
827 				inti->type,
828 				inti->ext.ext_params,
829 				inti->ext.ext_params2);
830 		list_del(&inti->list);
831 		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
832 	}
833 	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
834 		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
835 	spin_unlock(&fi->lock);
836 
837 	if (inti) {
838 		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
839 				(u16 *)__LC_EXT_INT_CODE);
840 		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
841 				(u16 *)__LC_EXT_CPU_ADDR);
842 		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
843 				&vcpu->arch.sie_block->gpsw,
844 				sizeof(psw_t));
845 		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
846 				&vcpu->arch.sie_block->gpsw,
847 				sizeof(psw_t));
848 		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
849 				(u32 *)__LC_EXT_PARAMS);
850 		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
851 				(u64 *)__LC_EXT_PARAMS2);
852 		kfree(inti);
853 	}
854 	return rc ? -EFAULT : 0;
855 }
856 
857 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
858 				     unsigned long irq_type)
859 {
860 	struct list_head *isc_list;
861 	struct kvm_s390_float_interrupt *fi;
862 	struct kvm_s390_interrupt_info *inti = NULL;
863 	int rc = 0;
864 
865 	fi = &vcpu->kvm->arch.float_int;
866 
867 	spin_lock(&fi->lock);
868 	isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
869 	inti = list_first_entry_or_null(isc_list,
870 					struct kvm_s390_interrupt_info,
871 					list);
872 	if (inti) {
873 		if (inti->type & KVM_S390_INT_IO_AI_MASK)
874 			VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
875 		else
876 			VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
877 			inti->io.subchannel_id >> 8,
878 			inti->io.subchannel_id >> 1 & 0x3,
879 			inti->io.subchannel_nr);
880 
881 		vcpu->stat.deliver_io_int++;
882 		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
883 				inti->type,
884 				((__u32)inti->io.subchannel_id << 16) |
885 				inti->io.subchannel_nr,
886 				((__u64)inti->io.io_int_parm << 32) |
887 				inti->io.io_int_word);
888 		list_del(&inti->list);
889 		fi->counters[FIRQ_CNTR_IO] -= 1;
890 	}
891 	if (list_empty(isc_list))
892 		clear_bit(irq_type, &fi->pending_irqs);
893 	spin_unlock(&fi->lock);
894 
895 	if (inti) {
896 		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
897 				(u16 *)__LC_SUBCHANNEL_ID);
898 		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
899 				(u16 *)__LC_SUBCHANNEL_NR);
900 		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
901 				(u32 *)__LC_IO_INT_PARM);
902 		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
903 				(u32 *)__LC_IO_INT_WORD);
904 		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
905 				&vcpu->arch.sie_block->gpsw,
906 				sizeof(psw_t));
907 		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
908 				&vcpu->arch.sie_block->gpsw,
909 				sizeof(psw_t));
910 		kfree(inti);
911 	}
912 
913 	return rc ? -EFAULT : 0;
914 }
915 
916 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
917 
918 static const deliver_irq_t deliver_irq_funcs[] = {
919 	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
920 	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
921 	[IRQ_PEND_PROG]           = __deliver_prog,
922 	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
923 	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
924 	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
925 	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
926 	[IRQ_PEND_RESTART]        = __deliver_restart,
927 	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
928 	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
929 	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
930 	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
931 	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
932 };
933 
934 /* Check whether an external call is pending (deliverable or not) */
935 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
936 {
937 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
938 
939 	if (!sclp.has_sigpif)
940 		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
941 
942 	return sca_ext_call_pending(vcpu, NULL);
943 }
944 
945 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
946 {
947 	if (deliverable_irqs(vcpu))
948 		return 1;
949 
950 	if (kvm_cpu_has_pending_timer(vcpu))
951 		return 1;
952 
953 	/* external call pending and deliverable */
954 	if (kvm_s390_ext_call_pending(vcpu) &&
955 	    !psw_extint_disabled(vcpu) &&
956 	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
957 		return 1;
958 
959 	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
960 		return 1;
961 	return 0;
962 }
963 
964 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
965 {
966 	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
967 }
968 
969 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
970 {
971 	u64 now, cputm, sltime = 0;
972 
973 	if (ckc_interrupts_enabled(vcpu)) {
974 		now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
975 		sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
976 		/* already expired or overflow? */
977 		if (!sltime || vcpu->arch.sie_block->ckc <= now)
978 			return 0;
979 		if (cpu_timer_interrupts_enabled(vcpu)) {
980 			cputm = kvm_s390_get_cpu_timer(vcpu);
981 			/* already expired? */
982 			if (cputm >> 63)
983 				return 0;
984 			return min(sltime, tod_to_ns(cputm));
985 		}
986 	} else if (cpu_timer_interrupts_enabled(vcpu)) {
987 		sltime = kvm_s390_get_cpu_timer(vcpu);
988 		/* already expired? */
989 		if (sltime >> 63)
990 			return 0;
991 	}
992 	return sltime;
993 }
994 
995 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
996 {
997 	u64 sltime;
998 
999 	vcpu->stat.exit_wait_state++;
1000 
1001 	/* fast path */
1002 	if (kvm_arch_vcpu_runnable(vcpu))
1003 		return 0;
1004 
1005 	if (psw_interrupts_disabled(vcpu)) {
1006 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
1007 		return -EOPNOTSUPP; /* disabled wait */
1008 	}
1009 
1010 	if (!ckc_interrupts_enabled(vcpu) &&
1011 	    !cpu_timer_interrupts_enabled(vcpu)) {
1012 		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
1013 		__set_cpu_idle(vcpu);
1014 		goto no_timer;
1015 	}
1016 
1017 	sltime = __calculate_sltime(vcpu);
1018 	if (!sltime)
1019 		return 0;
1020 
1021 	__set_cpu_idle(vcpu);
1022 	hrtimer_start(&vcpu->arch.ckc_timer, sltime, HRTIMER_MODE_REL);
1023 	VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
1024 no_timer:
1025 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1026 	kvm_vcpu_block(vcpu);
1027 	__unset_cpu_idle(vcpu);
1028 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1029 
1030 	hrtimer_cancel(&vcpu->arch.ckc_timer);
1031 	return 0;
1032 }
1033 
1034 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
1035 {
1036 	/*
1037 	 * We cannot move this into the if, as the CPU might be already
1038 	 * in kvm_vcpu_block without having the waitqueue set (polling)
1039 	 */
1040 	vcpu->valid_wakeup = true;
1041 	if (swait_active(&vcpu->wq)) {
1042 		/*
1043 		 * The vcpu gave up the cpu voluntarily, mark it as a good
1044 		 * yield-candidate.
1045 		 */
1046 		vcpu->preempted = true;
1047 		swake_up(&vcpu->wq);
1048 		vcpu->stat.halt_wakeup++;
1049 	}
1050 	/*
1051 	 * The VCPU might not be sleeping but is executing the VSIE. Let's
1052 	 * kick it, so it leaves the SIE to process the request.
1053 	 */
1054 	kvm_s390_vsie_kick(vcpu);
1055 }
1056 
1057 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1058 {
1059 	struct kvm_vcpu *vcpu;
1060 	u64 sltime;
1061 
1062 	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1063 	sltime = __calculate_sltime(vcpu);
1064 
1065 	/*
1066 	 * If the monotonic clock runs faster than the tod clock we might be
1067 	 * woken up too early and have to go back to sleep to avoid deadlocks.
1068 	 */
1069 	if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1070 		return HRTIMER_RESTART;
1071 	kvm_s390_vcpu_wakeup(vcpu);
1072 	return HRTIMER_NORESTART;
1073 }
1074 
1075 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1076 {
1077 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1078 
1079 	spin_lock(&li->lock);
1080 	li->pending_irqs = 0;
1081 	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1082 	memset(&li->irq, 0, sizeof(li->irq));
1083 	spin_unlock(&li->lock);
1084 
1085 	sca_clear_ext_call(vcpu);
1086 }
1087 
1088 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1089 {
1090 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1091 	deliver_irq_t func;
1092 	int rc = 0;
1093 	unsigned long irq_type;
1094 	unsigned long irqs;
1095 
1096 	__reset_intercept_indicators(vcpu);
1097 
1098 	/* pending ckc conditions might have been invalidated */
1099 	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1100 	if (ckc_irq_pending(vcpu))
1101 		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1102 
1103 	/* pending cpu timer conditions might have been invalidated */
1104 	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1105 	if (cpu_timer_irq_pending(vcpu))
1106 		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1107 
1108 	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1109 		/* bits are in the order of interrupt priority */
1110 		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1111 		if (is_ioirq(irq_type)) {
1112 			rc = __deliver_io(vcpu, irq_type);
1113 		} else {
1114 			func = deliver_irq_funcs[irq_type];
1115 			if (!func) {
1116 				WARN_ON_ONCE(func == NULL);
1117 				clear_bit(irq_type, &li->pending_irqs);
1118 				continue;
1119 			}
1120 			rc = func(vcpu);
1121 		}
1122 	}
1123 
1124 	set_intercept_indicators(vcpu);
1125 
1126 	return rc;
1127 }
1128 
1129 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1130 {
1131 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1132 
1133 	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1134 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1135 				   irq->u.pgm.code, 0);
1136 
1137 	if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1138 		/* auto detection if no valid ILC was given */
1139 		irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1140 		irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1141 		irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1142 	}
1143 
1144 	if (irq->u.pgm.code == PGM_PER) {
1145 		li->irq.pgm.code |= PGM_PER;
1146 		li->irq.pgm.flags = irq->u.pgm.flags;
1147 		/* only modify PER related information */
1148 		li->irq.pgm.per_address = irq->u.pgm.per_address;
1149 		li->irq.pgm.per_code = irq->u.pgm.per_code;
1150 		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1151 		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1152 	} else if (!(irq->u.pgm.code & PGM_PER)) {
1153 		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1154 				   irq->u.pgm.code;
1155 		li->irq.pgm.flags = irq->u.pgm.flags;
1156 		/* only modify non-PER information */
1157 		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1158 		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1159 		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1160 		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1161 		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1162 		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1163 	} else {
1164 		li->irq.pgm = irq->u.pgm;
1165 	}
1166 	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1167 	return 0;
1168 }
1169 
1170 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1171 {
1172 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1173 
1174 	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1175 		   irq->u.ext.ext_params2);
1176 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1177 				   irq->u.ext.ext_params,
1178 				   irq->u.ext.ext_params2);
1179 
1180 	li->irq.ext = irq->u.ext;
1181 	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1182 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1183 	return 0;
1184 }
1185 
1186 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1187 {
1188 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1189 	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1190 	uint16_t src_id = irq->u.extcall.code;
1191 
1192 	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1193 		   src_id);
1194 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1195 				   src_id, 0);
1196 
1197 	/* sending vcpu invalid */
1198 	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1199 		return -EINVAL;
1200 
1201 	if (sclp.has_sigpif)
1202 		return sca_inject_ext_call(vcpu, src_id);
1203 
1204 	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1205 		return -EBUSY;
1206 	*extcall = irq->u.extcall;
1207 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1208 	return 0;
1209 }
1210 
1211 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1212 {
1213 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1214 	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1215 
1216 	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1217 		   irq->u.prefix.address);
1218 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1219 				   irq->u.prefix.address, 0);
1220 
1221 	if (!is_vcpu_stopped(vcpu))
1222 		return -EBUSY;
1223 
1224 	*prefix = irq->u.prefix;
1225 	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1226 	return 0;
1227 }
1228 
1229 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1230 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1231 {
1232 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1233 	struct kvm_s390_stop_info *stop = &li->irq.stop;
1234 	int rc = 0;
1235 
1236 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1237 
1238 	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1239 		return -EINVAL;
1240 
1241 	if (is_vcpu_stopped(vcpu)) {
1242 		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1243 			rc = kvm_s390_store_status_unloaded(vcpu,
1244 						KVM_S390_STORE_STATUS_NOADDR);
1245 		return rc;
1246 	}
1247 
1248 	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1249 		return -EBUSY;
1250 	stop->flags = irq->u.stop.flags;
1251 	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1252 	return 0;
1253 }
1254 
1255 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1256 				 struct kvm_s390_irq *irq)
1257 {
1258 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1259 
1260 	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1261 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1262 
1263 	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1264 	return 0;
1265 }
1266 
1267 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1268 				   struct kvm_s390_irq *irq)
1269 {
1270 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1271 
1272 	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1273 		   irq->u.emerg.code);
1274 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1275 				   irq->u.emerg.code, 0);
1276 
1277 	/* sending vcpu invalid */
1278 	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1279 		return -EINVAL;
1280 
1281 	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1282 	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1283 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1284 	return 0;
1285 }
1286 
1287 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1288 {
1289 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1290 	struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1291 
1292 	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1293 		   irq->u.mchk.mcic);
1294 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1295 				   irq->u.mchk.mcic);
1296 
1297 	/*
1298 	 * Because repressible machine checks can be indicated along with
1299 	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1300 	 * we need to combine cr14, mcic and external damage code.
1301 	 * Failing storage address and the logout area should not be or'ed
1302 	 * together, we just indicate the last occurrence of the corresponding
1303 	 * machine check
1304 	 */
1305 	mchk->cr14 |= irq->u.mchk.cr14;
1306 	mchk->mcic |= irq->u.mchk.mcic;
1307 	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1308 	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1309 	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1310 	       sizeof(mchk->fixed_logout));
1311 	if (mchk->mcic & MCHK_EX_MASK)
1312 		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1313 	else if (mchk->mcic & MCHK_REP_MASK)
1314 		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1315 	return 0;
1316 }
1317 
1318 static int __inject_ckc(struct kvm_vcpu *vcpu)
1319 {
1320 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1321 
1322 	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1323 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1324 				   0, 0);
1325 
1326 	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1327 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1328 	return 0;
1329 }
1330 
1331 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1332 {
1333 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1334 
1335 	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1336 	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1337 				   0, 0);
1338 
1339 	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1340 	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1341 	return 0;
1342 }
1343 
1344 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1345 						  int isc, u32 schid)
1346 {
1347 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1348 	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1349 	struct kvm_s390_interrupt_info *iter;
1350 	u16 id = (schid & 0xffff0000U) >> 16;
1351 	u16 nr = schid & 0x0000ffffU;
1352 
1353 	spin_lock(&fi->lock);
1354 	list_for_each_entry(iter, isc_list, list) {
1355 		if (schid && (id != iter->io.subchannel_id ||
1356 			      nr != iter->io.subchannel_nr))
1357 			continue;
1358 		/* found an appropriate entry */
1359 		list_del_init(&iter->list);
1360 		fi->counters[FIRQ_CNTR_IO] -= 1;
1361 		if (list_empty(isc_list))
1362 			clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1363 		spin_unlock(&fi->lock);
1364 		return iter;
1365 	}
1366 	spin_unlock(&fi->lock);
1367 	return NULL;
1368 }
1369 
1370 /*
1371  * Dequeue and return an I/O interrupt matching any of the interruption
1372  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1373  */
1374 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1375 						    u64 isc_mask, u32 schid)
1376 {
1377 	struct kvm_s390_interrupt_info *inti = NULL;
1378 	int isc;
1379 
1380 	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1381 		if (isc_mask & isc_to_isc_bits(isc))
1382 			inti = get_io_int(kvm, isc, schid);
1383 	}
1384 	return inti;
1385 }
1386 
1387 #define SCCB_MASK 0xFFFFFFF8
1388 #define SCCB_EVENT_PENDING 0x3
1389 
1390 static int __inject_service(struct kvm *kvm,
1391 			     struct kvm_s390_interrupt_info *inti)
1392 {
1393 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1394 
1395 	spin_lock(&fi->lock);
1396 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1397 	/*
1398 	 * Early versions of the QEMU s390 bios will inject several
1399 	 * service interrupts after another without handling a
1400 	 * condition code indicating busy.
1401 	 * We will silently ignore those superfluous sccb values.
1402 	 * A future version of QEMU will take care of serialization
1403 	 * of servc requests
1404 	 */
1405 	if (fi->srv_signal.ext_params & SCCB_MASK)
1406 		goto out;
1407 	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1408 	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1409 out:
1410 	spin_unlock(&fi->lock);
1411 	kfree(inti);
1412 	return 0;
1413 }
1414 
1415 static int __inject_virtio(struct kvm *kvm,
1416 			    struct kvm_s390_interrupt_info *inti)
1417 {
1418 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1419 
1420 	spin_lock(&fi->lock);
1421 	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1422 		spin_unlock(&fi->lock);
1423 		return -EBUSY;
1424 	}
1425 	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1426 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1427 	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1428 	spin_unlock(&fi->lock);
1429 	return 0;
1430 }
1431 
1432 static int __inject_pfault_done(struct kvm *kvm,
1433 				 struct kvm_s390_interrupt_info *inti)
1434 {
1435 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1436 
1437 	spin_lock(&fi->lock);
1438 	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1439 		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1440 		spin_unlock(&fi->lock);
1441 		return -EBUSY;
1442 	}
1443 	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1444 	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1445 	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1446 	spin_unlock(&fi->lock);
1447 	return 0;
1448 }
1449 
1450 #define CR_PENDING_SUBCLASS 28
1451 static int __inject_float_mchk(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 	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1458 	fi->mchk.mcic |= inti->mchk.mcic;
1459 	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1460 	spin_unlock(&fi->lock);
1461 	kfree(inti);
1462 	return 0;
1463 }
1464 
1465 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1466 {
1467 	struct kvm_s390_float_interrupt *fi;
1468 	struct list_head *list;
1469 	int isc;
1470 
1471 	fi = &kvm->arch.float_int;
1472 	spin_lock(&fi->lock);
1473 	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1474 		spin_unlock(&fi->lock);
1475 		return -EBUSY;
1476 	}
1477 	fi->counters[FIRQ_CNTR_IO] += 1;
1478 
1479 	if (inti->type & KVM_S390_INT_IO_AI_MASK)
1480 		VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1481 	else
1482 		VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1483 			inti->io.subchannel_id >> 8,
1484 			inti->io.subchannel_id >> 1 & 0x3,
1485 			inti->io.subchannel_nr);
1486 	isc = int_word_to_isc(inti->io.io_int_word);
1487 	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1488 	list_add_tail(&inti->list, list);
1489 	set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1490 	spin_unlock(&fi->lock);
1491 	return 0;
1492 }
1493 
1494 /*
1495  * Find a destination VCPU for a floating irq and kick it.
1496  */
1497 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1498 {
1499 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1500 	struct kvm_s390_local_interrupt *li;
1501 	struct kvm_vcpu *dst_vcpu;
1502 	int sigcpu, online_vcpus, nr_tries = 0;
1503 
1504 	online_vcpus = atomic_read(&kvm->online_vcpus);
1505 	if (!online_vcpus)
1506 		return;
1507 
1508 	/* find idle VCPUs first, then round robin */
1509 	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1510 	if (sigcpu == online_vcpus) {
1511 		do {
1512 			sigcpu = fi->next_rr_cpu;
1513 			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1514 			/* avoid endless loops if all vcpus are stopped */
1515 			if (nr_tries++ >= online_vcpus)
1516 				return;
1517 		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1518 	}
1519 	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1520 
1521 	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1522 	li = &dst_vcpu->arch.local_int;
1523 	spin_lock(&li->lock);
1524 	switch (type) {
1525 	case KVM_S390_MCHK:
1526 		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1527 		break;
1528 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1529 		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1530 		break;
1531 	default:
1532 		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1533 		break;
1534 	}
1535 	spin_unlock(&li->lock);
1536 	kvm_s390_vcpu_wakeup(dst_vcpu);
1537 }
1538 
1539 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1540 {
1541 	u64 type = READ_ONCE(inti->type);
1542 	int rc;
1543 
1544 	switch (type) {
1545 	case KVM_S390_MCHK:
1546 		rc = __inject_float_mchk(kvm, inti);
1547 		break;
1548 	case KVM_S390_INT_VIRTIO:
1549 		rc = __inject_virtio(kvm, inti);
1550 		break;
1551 	case KVM_S390_INT_SERVICE:
1552 		rc = __inject_service(kvm, inti);
1553 		break;
1554 	case KVM_S390_INT_PFAULT_DONE:
1555 		rc = __inject_pfault_done(kvm, inti);
1556 		break;
1557 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1558 		rc = __inject_io(kvm, inti);
1559 		break;
1560 	default:
1561 		rc = -EINVAL;
1562 	}
1563 	if (rc)
1564 		return rc;
1565 
1566 	__floating_irq_kick(kvm, type);
1567 	return 0;
1568 }
1569 
1570 int kvm_s390_inject_vm(struct kvm *kvm,
1571 		       struct kvm_s390_interrupt *s390int)
1572 {
1573 	struct kvm_s390_interrupt_info *inti;
1574 	int rc;
1575 
1576 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1577 	if (!inti)
1578 		return -ENOMEM;
1579 
1580 	inti->type = s390int->type;
1581 	switch (inti->type) {
1582 	case KVM_S390_INT_VIRTIO:
1583 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1584 			 s390int->parm, s390int->parm64);
1585 		inti->ext.ext_params = s390int->parm;
1586 		inti->ext.ext_params2 = s390int->parm64;
1587 		break;
1588 	case KVM_S390_INT_SERVICE:
1589 		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1590 		inti->ext.ext_params = s390int->parm;
1591 		break;
1592 	case KVM_S390_INT_PFAULT_DONE:
1593 		inti->ext.ext_params2 = s390int->parm64;
1594 		break;
1595 	case KVM_S390_MCHK:
1596 		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1597 			 s390int->parm64);
1598 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1599 		inti->mchk.mcic = s390int->parm64;
1600 		break;
1601 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1602 		inti->io.subchannel_id = s390int->parm >> 16;
1603 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1604 		inti->io.io_int_parm = s390int->parm64 >> 32;
1605 		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1606 		break;
1607 	default:
1608 		kfree(inti);
1609 		return -EINVAL;
1610 	}
1611 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1612 				 2);
1613 
1614 	rc = __inject_vm(kvm, inti);
1615 	if (rc)
1616 		kfree(inti);
1617 	return rc;
1618 }
1619 
1620 int kvm_s390_reinject_io_int(struct kvm *kvm,
1621 			      struct kvm_s390_interrupt_info *inti)
1622 {
1623 	return __inject_vm(kvm, inti);
1624 }
1625 
1626 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1627 		       struct kvm_s390_irq *irq)
1628 {
1629 	irq->type = s390int->type;
1630 	switch (irq->type) {
1631 	case KVM_S390_PROGRAM_INT:
1632 		if (s390int->parm & 0xffff0000)
1633 			return -EINVAL;
1634 		irq->u.pgm.code = s390int->parm;
1635 		break;
1636 	case KVM_S390_SIGP_SET_PREFIX:
1637 		irq->u.prefix.address = s390int->parm;
1638 		break;
1639 	case KVM_S390_SIGP_STOP:
1640 		irq->u.stop.flags = s390int->parm;
1641 		break;
1642 	case KVM_S390_INT_EXTERNAL_CALL:
1643 		if (s390int->parm & 0xffff0000)
1644 			return -EINVAL;
1645 		irq->u.extcall.code = s390int->parm;
1646 		break;
1647 	case KVM_S390_INT_EMERGENCY:
1648 		if (s390int->parm & 0xffff0000)
1649 			return -EINVAL;
1650 		irq->u.emerg.code = s390int->parm;
1651 		break;
1652 	case KVM_S390_MCHK:
1653 		irq->u.mchk.mcic = s390int->parm64;
1654 		break;
1655 	}
1656 	return 0;
1657 }
1658 
1659 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1660 {
1661 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1662 
1663 	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1664 }
1665 
1666 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1667 {
1668 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1669 
1670 	spin_lock(&li->lock);
1671 	li->irq.stop.flags = 0;
1672 	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1673 	spin_unlock(&li->lock);
1674 }
1675 
1676 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1677 {
1678 	int rc;
1679 
1680 	switch (irq->type) {
1681 	case KVM_S390_PROGRAM_INT:
1682 		rc = __inject_prog(vcpu, irq);
1683 		break;
1684 	case KVM_S390_SIGP_SET_PREFIX:
1685 		rc = __inject_set_prefix(vcpu, irq);
1686 		break;
1687 	case KVM_S390_SIGP_STOP:
1688 		rc = __inject_sigp_stop(vcpu, irq);
1689 		break;
1690 	case KVM_S390_RESTART:
1691 		rc = __inject_sigp_restart(vcpu, irq);
1692 		break;
1693 	case KVM_S390_INT_CLOCK_COMP:
1694 		rc = __inject_ckc(vcpu);
1695 		break;
1696 	case KVM_S390_INT_CPU_TIMER:
1697 		rc = __inject_cpu_timer(vcpu);
1698 		break;
1699 	case KVM_S390_INT_EXTERNAL_CALL:
1700 		rc = __inject_extcall(vcpu, irq);
1701 		break;
1702 	case KVM_S390_INT_EMERGENCY:
1703 		rc = __inject_sigp_emergency(vcpu, irq);
1704 		break;
1705 	case KVM_S390_MCHK:
1706 		rc = __inject_mchk(vcpu, irq);
1707 		break;
1708 	case KVM_S390_INT_PFAULT_INIT:
1709 		rc = __inject_pfault_init(vcpu, irq);
1710 		break;
1711 	case KVM_S390_INT_VIRTIO:
1712 	case KVM_S390_INT_SERVICE:
1713 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1714 	default:
1715 		rc = -EINVAL;
1716 	}
1717 
1718 	return rc;
1719 }
1720 
1721 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1722 {
1723 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1724 	int rc;
1725 
1726 	spin_lock(&li->lock);
1727 	rc = do_inject_vcpu(vcpu, irq);
1728 	spin_unlock(&li->lock);
1729 	if (!rc)
1730 		kvm_s390_vcpu_wakeup(vcpu);
1731 	return rc;
1732 }
1733 
1734 static inline void clear_irq_list(struct list_head *_list)
1735 {
1736 	struct kvm_s390_interrupt_info *inti, *n;
1737 
1738 	list_for_each_entry_safe(inti, n, _list, list) {
1739 		list_del(&inti->list);
1740 		kfree(inti);
1741 	}
1742 }
1743 
1744 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1745 		       struct kvm_s390_irq *irq)
1746 {
1747 	irq->type = inti->type;
1748 	switch (inti->type) {
1749 	case KVM_S390_INT_PFAULT_INIT:
1750 	case KVM_S390_INT_PFAULT_DONE:
1751 	case KVM_S390_INT_VIRTIO:
1752 		irq->u.ext = inti->ext;
1753 		break;
1754 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1755 		irq->u.io = inti->io;
1756 		break;
1757 	}
1758 }
1759 
1760 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1761 {
1762 	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1763 	int i;
1764 
1765 	spin_lock(&fi->lock);
1766 	fi->pending_irqs = 0;
1767 	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1768 	memset(&fi->mchk, 0, sizeof(fi->mchk));
1769 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1770 		clear_irq_list(&fi->lists[i]);
1771 	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1772 		fi->counters[i] = 0;
1773 	spin_unlock(&fi->lock);
1774 };
1775 
1776 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1777 {
1778 	struct kvm_s390_interrupt_info *inti;
1779 	struct kvm_s390_float_interrupt *fi;
1780 	struct kvm_s390_irq *buf;
1781 	struct kvm_s390_irq *irq;
1782 	int max_irqs;
1783 	int ret = 0;
1784 	int n = 0;
1785 	int i;
1786 
1787 	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1788 		return -EINVAL;
1789 
1790 	/*
1791 	 * We are already using -ENOMEM to signal
1792 	 * userspace it may retry with a bigger buffer,
1793 	 * so we need to use something else for this case
1794 	 */
1795 	buf = vzalloc(len);
1796 	if (!buf)
1797 		return -ENOBUFS;
1798 
1799 	max_irqs = len / sizeof(struct kvm_s390_irq);
1800 
1801 	fi = &kvm->arch.float_int;
1802 	spin_lock(&fi->lock);
1803 	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1804 		list_for_each_entry(inti, &fi->lists[i], list) {
1805 			if (n == max_irqs) {
1806 				/* signal userspace to try again */
1807 				ret = -ENOMEM;
1808 				goto out;
1809 			}
1810 			inti_to_irq(inti, &buf[n]);
1811 			n++;
1812 		}
1813 	}
1814 	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1815 		if (n == max_irqs) {
1816 			/* signal userspace to try again */
1817 			ret = -ENOMEM;
1818 			goto out;
1819 		}
1820 		irq = (struct kvm_s390_irq *) &buf[n];
1821 		irq->type = KVM_S390_INT_SERVICE;
1822 		irq->u.ext = fi->srv_signal;
1823 		n++;
1824 	}
1825 	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1826 		if (n == max_irqs) {
1827 				/* signal userspace to try again */
1828 				ret = -ENOMEM;
1829 				goto out;
1830 		}
1831 		irq = (struct kvm_s390_irq *) &buf[n];
1832 		irq->type = KVM_S390_MCHK;
1833 		irq->u.mchk = fi->mchk;
1834 		n++;
1835 }
1836 
1837 out:
1838 	spin_unlock(&fi->lock);
1839 	if (!ret && n > 0) {
1840 		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1841 			ret = -EFAULT;
1842 	}
1843 	vfree(buf);
1844 
1845 	return ret < 0 ? ret : n;
1846 }
1847 
1848 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1849 {
1850 	int r;
1851 
1852 	switch (attr->group) {
1853 	case KVM_DEV_FLIC_GET_ALL_IRQS:
1854 		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1855 					  attr->attr);
1856 		break;
1857 	default:
1858 		r = -EINVAL;
1859 	}
1860 
1861 	return r;
1862 }
1863 
1864 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1865 				     u64 addr)
1866 {
1867 	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1868 	void *target = NULL;
1869 	void __user *source;
1870 	u64 size;
1871 
1872 	if (get_user(inti->type, (u64 __user *)addr))
1873 		return -EFAULT;
1874 
1875 	switch (inti->type) {
1876 	case KVM_S390_INT_PFAULT_INIT:
1877 	case KVM_S390_INT_PFAULT_DONE:
1878 	case KVM_S390_INT_VIRTIO:
1879 	case KVM_S390_INT_SERVICE:
1880 		target = (void *) &inti->ext;
1881 		source = &uptr->u.ext;
1882 		size = sizeof(inti->ext);
1883 		break;
1884 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1885 		target = (void *) &inti->io;
1886 		source = &uptr->u.io;
1887 		size = sizeof(inti->io);
1888 		break;
1889 	case KVM_S390_MCHK:
1890 		target = (void *) &inti->mchk;
1891 		source = &uptr->u.mchk;
1892 		size = sizeof(inti->mchk);
1893 		break;
1894 	default:
1895 		return -EINVAL;
1896 	}
1897 
1898 	if (copy_from_user(target, source, size))
1899 		return -EFAULT;
1900 
1901 	return 0;
1902 }
1903 
1904 static int enqueue_floating_irq(struct kvm_device *dev,
1905 				struct kvm_device_attr *attr)
1906 {
1907 	struct kvm_s390_interrupt_info *inti = NULL;
1908 	int r = 0;
1909 	int len = attr->attr;
1910 
1911 	if (len % sizeof(struct kvm_s390_irq) != 0)
1912 		return -EINVAL;
1913 	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1914 		return -EINVAL;
1915 
1916 	while (len >= sizeof(struct kvm_s390_irq)) {
1917 		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1918 		if (!inti)
1919 			return -ENOMEM;
1920 
1921 		r = copy_irq_from_user(inti, attr->addr);
1922 		if (r) {
1923 			kfree(inti);
1924 			return r;
1925 		}
1926 		r = __inject_vm(dev->kvm, inti);
1927 		if (r) {
1928 			kfree(inti);
1929 			return r;
1930 		}
1931 		len -= sizeof(struct kvm_s390_irq);
1932 		attr->addr += sizeof(struct kvm_s390_irq);
1933 	}
1934 
1935 	return r;
1936 }
1937 
1938 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1939 {
1940 	if (id >= MAX_S390_IO_ADAPTERS)
1941 		return NULL;
1942 	return kvm->arch.adapters[id];
1943 }
1944 
1945 static int register_io_adapter(struct kvm_device *dev,
1946 			       struct kvm_device_attr *attr)
1947 {
1948 	struct s390_io_adapter *adapter;
1949 	struct kvm_s390_io_adapter adapter_info;
1950 
1951 	if (copy_from_user(&adapter_info,
1952 			   (void __user *)attr->addr, sizeof(adapter_info)))
1953 		return -EFAULT;
1954 
1955 	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1956 	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1957 		return -EINVAL;
1958 
1959 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1960 	if (!adapter)
1961 		return -ENOMEM;
1962 
1963 	INIT_LIST_HEAD(&adapter->maps);
1964 	init_rwsem(&adapter->maps_lock);
1965 	atomic_set(&adapter->nr_maps, 0);
1966 	adapter->id = adapter_info.id;
1967 	adapter->isc = adapter_info.isc;
1968 	adapter->maskable = adapter_info.maskable;
1969 	adapter->masked = false;
1970 	adapter->swap = adapter_info.swap;
1971 	dev->kvm->arch.adapters[adapter->id] = adapter;
1972 
1973 	return 0;
1974 }
1975 
1976 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1977 {
1978 	int ret;
1979 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1980 
1981 	if (!adapter || !adapter->maskable)
1982 		return -EINVAL;
1983 	ret = adapter->masked;
1984 	adapter->masked = masked;
1985 	return ret;
1986 }
1987 
1988 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1989 {
1990 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1991 	struct s390_map_info *map;
1992 	int ret;
1993 
1994 	if (!adapter || !addr)
1995 		return -EINVAL;
1996 
1997 	map = kzalloc(sizeof(*map), GFP_KERNEL);
1998 	if (!map) {
1999 		ret = -ENOMEM;
2000 		goto out;
2001 	}
2002 	INIT_LIST_HEAD(&map->list);
2003 	map->guest_addr = addr;
2004 	map->addr = gmap_translate(kvm->arch.gmap, addr);
2005 	if (map->addr == -EFAULT) {
2006 		ret = -EFAULT;
2007 		goto out;
2008 	}
2009 	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
2010 	if (ret < 0)
2011 		goto out;
2012 	BUG_ON(ret != 1);
2013 	down_write(&adapter->maps_lock);
2014 	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
2015 		list_add_tail(&map->list, &adapter->maps);
2016 		ret = 0;
2017 	} else {
2018 		put_page(map->page);
2019 		ret = -EINVAL;
2020 	}
2021 	up_write(&adapter->maps_lock);
2022 out:
2023 	if (ret)
2024 		kfree(map);
2025 	return ret;
2026 }
2027 
2028 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
2029 {
2030 	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2031 	struct s390_map_info *map, *tmp;
2032 	int found = 0;
2033 
2034 	if (!adapter || !addr)
2035 		return -EINVAL;
2036 
2037 	down_write(&adapter->maps_lock);
2038 	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
2039 		if (map->guest_addr == addr) {
2040 			found = 1;
2041 			atomic_dec(&adapter->nr_maps);
2042 			list_del(&map->list);
2043 			put_page(map->page);
2044 			kfree(map);
2045 			break;
2046 		}
2047 	}
2048 	up_write(&adapter->maps_lock);
2049 
2050 	return found ? 0 : -EINVAL;
2051 }
2052 
2053 void kvm_s390_destroy_adapters(struct kvm *kvm)
2054 {
2055 	int i;
2056 	struct s390_map_info *map, *tmp;
2057 
2058 	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2059 		if (!kvm->arch.adapters[i])
2060 			continue;
2061 		list_for_each_entry_safe(map, tmp,
2062 					 &kvm->arch.adapters[i]->maps, list) {
2063 			list_del(&map->list);
2064 			put_page(map->page);
2065 			kfree(map);
2066 		}
2067 		kfree(kvm->arch.adapters[i]);
2068 	}
2069 }
2070 
2071 static int modify_io_adapter(struct kvm_device *dev,
2072 			     struct kvm_device_attr *attr)
2073 {
2074 	struct kvm_s390_io_adapter_req req;
2075 	struct s390_io_adapter *adapter;
2076 	int ret;
2077 
2078 	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2079 		return -EFAULT;
2080 
2081 	adapter = get_io_adapter(dev->kvm, req.id);
2082 	if (!adapter)
2083 		return -EINVAL;
2084 	switch (req.type) {
2085 	case KVM_S390_IO_ADAPTER_MASK:
2086 		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2087 		if (ret > 0)
2088 			ret = 0;
2089 		break;
2090 	case KVM_S390_IO_ADAPTER_MAP:
2091 		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2092 		break;
2093 	case KVM_S390_IO_ADAPTER_UNMAP:
2094 		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2095 		break;
2096 	default:
2097 		ret = -EINVAL;
2098 	}
2099 
2100 	return ret;
2101 }
2102 
2103 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2104 
2105 {
2106 	const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2107 	u32 schid;
2108 
2109 	if (attr->flags)
2110 		return -EINVAL;
2111 	if (attr->attr != sizeof(schid))
2112 		return -EINVAL;
2113 	if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2114 		return -EFAULT;
2115 	kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2116 	/*
2117 	 * If userspace is conforming to the architecture, we can have at most
2118 	 * one pending I/O interrupt per subchannel, so this is effectively a
2119 	 * clear all.
2120 	 */
2121 	return 0;
2122 }
2123 
2124 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2125 {
2126 	int r = 0;
2127 	unsigned int i;
2128 	struct kvm_vcpu *vcpu;
2129 
2130 	switch (attr->group) {
2131 	case KVM_DEV_FLIC_ENQUEUE:
2132 		r = enqueue_floating_irq(dev, attr);
2133 		break;
2134 	case KVM_DEV_FLIC_CLEAR_IRQS:
2135 		kvm_s390_clear_float_irqs(dev->kvm);
2136 		break;
2137 	case KVM_DEV_FLIC_APF_ENABLE:
2138 		dev->kvm->arch.gmap->pfault_enabled = 1;
2139 		break;
2140 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2141 		dev->kvm->arch.gmap->pfault_enabled = 0;
2142 		/*
2143 		 * Make sure no async faults are in transition when
2144 		 * clearing the queues. So we don't need to worry
2145 		 * about late coming workers.
2146 		 */
2147 		synchronize_srcu(&dev->kvm->srcu);
2148 		kvm_for_each_vcpu(i, vcpu, dev->kvm)
2149 			kvm_clear_async_pf_completion_queue(vcpu);
2150 		break;
2151 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2152 		r = register_io_adapter(dev, attr);
2153 		break;
2154 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2155 		r = modify_io_adapter(dev, attr);
2156 		break;
2157 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2158 		r = clear_io_irq(dev->kvm, attr);
2159 		break;
2160 	default:
2161 		r = -EINVAL;
2162 	}
2163 
2164 	return r;
2165 }
2166 
2167 static int flic_has_attr(struct kvm_device *dev,
2168 			     struct kvm_device_attr *attr)
2169 {
2170 	switch (attr->group) {
2171 	case KVM_DEV_FLIC_GET_ALL_IRQS:
2172 	case KVM_DEV_FLIC_ENQUEUE:
2173 	case KVM_DEV_FLIC_CLEAR_IRQS:
2174 	case KVM_DEV_FLIC_APF_ENABLE:
2175 	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2176 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
2177 	case KVM_DEV_FLIC_ADAPTER_MODIFY:
2178 	case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2179 		return 0;
2180 	}
2181 	return -ENXIO;
2182 }
2183 
2184 static int flic_create(struct kvm_device *dev, u32 type)
2185 {
2186 	if (!dev)
2187 		return -EINVAL;
2188 	if (dev->kvm->arch.flic)
2189 		return -EINVAL;
2190 	dev->kvm->arch.flic = dev;
2191 	return 0;
2192 }
2193 
2194 static void flic_destroy(struct kvm_device *dev)
2195 {
2196 	dev->kvm->arch.flic = NULL;
2197 	kfree(dev);
2198 }
2199 
2200 /* s390 floating irq controller (flic) */
2201 struct kvm_device_ops kvm_flic_ops = {
2202 	.name = "kvm-flic",
2203 	.get_attr = flic_get_attr,
2204 	.set_attr = flic_set_attr,
2205 	.has_attr = flic_has_attr,
2206 	.create = flic_create,
2207 	.destroy = flic_destroy,
2208 };
2209 
2210 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2211 {
2212 	unsigned long bit;
2213 
2214 	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2215 
2216 	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2217 }
2218 
2219 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2220 					  u64 addr)
2221 {
2222 	struct s390_map_info *map;
2223 
2224 	if (!adapter)
2225 		return NULL;
2226 
2227 	list_for_each_entry(map, &adapter->maps, list) {
2228 		if (map->guest_addr == addr)
2229 			return map;
2230 	}
2231 	return NULL;
2232 }
2233 
2234 static int adapter_indicators_set(struct kvm *kvm,
2235 				  struct s390_io_adapter *adapter,
2236 				  struct kvm_s390_adapter_int *adapter_int)
2237 {
2238 	unsigned long bit;
2239 	int summary_set, idx;
2240 	struct s390_map_info *info;
2241 	void *map;
2242 
2243 	info = get_map_info(adapter, adapter_int->ind_addr);
2244 	if (!info)
2245 		return -1;
2246 	map = page_address(info->page);
2247 	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2248 	set_bit(bit, map);
2249 	idx = srcu_read_lock(&kvm->srcu);
2250 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2251 	set_page_dirty_lock(info->page);
2252 	info = get_map_info(adapter, adapter_int->summary_addr);
2253 	if (!info) {
2254 		srcu_read_unlock(&kvm->srcu, idx);
2255 		return -1;
2256 	}
2257 	map = page_address(info->page);
2258 	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2259 			  adapter->swap);
2260 	summary_set = test_and_set_bit(bit, map);
2261 	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2262 	set_page_dirty_lock(info->page);
2263 	srcu_read_unlock(&kvm->srcu, idx);
2264 	return summary_set ? 0 : 1;
2265 }
2266 
2267 /*
2268  * < 0 - not injected due to error
2269  * = 0 - coalesced, summary indicator already active
2270  * > 0 - injected interrupt
2271  */
2272 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2273 			   struct kvm *kvm, int irq_source_id, int level,
2274 			   bool line_status)
2275 {
2276 	int ret;
2277 	struct s390_io_adapter *adapter;
2278 
2279 	/* We're only interested in the 0->1 transition. */
2280 	if (!level)
2281 		return 0;
2282 	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2283 	if (!adapter)
2284 		return -1;
2285 	down_read(&adapter->maps_lock);
2286 	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2287 	up_read(&adapter->maps_lock);
2288 	if ((ret > 0) && !adapter->masked) {
2289 		struct kvm_s390_interrupt s390int = {
2290 			.type = KVM_S390_INT_IO(1, 0, 0, 0),
2291 			.parm = 0,
2292 			.parm64 = (adapter->isc << 27) | 0x80000000,
2293 		};
2294 		ret = kvm_s390_inject_vm(kvm, &s390int);
2295 		if (ret == 0)
2296 			ret = 1;
2297 	}
2298 	return ret;
2299 }
2300 
2301 int kvm_set_routing_entry(struct kvm *kvm,
2302 			  struct kvm_kernel_irq_routing_entry *e,
2303 			  const struct kvm_irq_routing_entry *ue)
2304 {
2305 	int ret;
2306 
2307 	switch (ue->type) {
2308 	case KVM_IRQ_ROUTING_S390_ADAPTER:
2309 		e->set = set_adapter_int;
2310 		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2311 		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2312 		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2313 		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2314 		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2315 		ret = 0;
2316 		break;
2317 	default:
2318 		ret = -EINVAL;
2319 	}
2320 
2321 	return ret;
2322 }
2323 
2324 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2325 		int irq_source_id, int level, bool line_status)
2326 {
2327 	return -EINVAL;
2328 }
2329 
2330 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2331 {
2332 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2333 	struct kvm_s390_irq *buf;
2334 	int r = 0;
2335 	int n;
2336 
2337 	buf = vmalloc(len);
2338 	if (!buf)
2339 		return -ENOMEM;
2340 
2341 	if (copy_from_user((void *) buf, irqstate, len)) {
2342 		r = -EFAULT;
2343 		goto out_free;
2344 	}
2345 
2346 	/*
2347 	 * Don't allow setting the interrupt state
2348 	 * when there are already interrupts pending
2349 	 */
2350 	spin_lock(&li->lock);
2351 	if (li->pending_irqs) {
2352 		r = -EBUSY;
2353 		goto out_unlock;
2354 	}
2355 
2356 	for (n = 0; n < len / sizeof(*buf); n++) {
2357 		r = do_inject_vcpu(vcpu, &buf[n]);
2358 		if (r)
2359 			break;
2360 	}
2361 
2362 out_unlock:
2363 	spin_unlock(&li->lock);
2364 out_free:
2365 	vfree(buf);
2366 
2367 	return r;
2368 }
2369 
2370 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2371 			    struct kvm_s390_irq *irq,
2372 			    unsigned long irq_type)
2373 {
2374 	switch (irq_type) {
2375 	case IRQ_PEND_MCHK_EX:
2376 	case IRQ_PEND_MCHK_REP:
2377 		irq->type = KVM_S390_MCHK;
2378 		irq->u.mchk = li->irq.mchk;
2379 		break;
2380 	case IRQ_PEND_PROG:
2381 		irq->type = KVM_S390_PROGRAM_INT;
2382 		irq->u.pgm = li->irq.pgm;
2383 		break;
2384 	case IRQ_PEND_PFAULT_INIT:
2385 		irq->type = KVM_S390_INT_PFAULT_INIT;
2386 		irq->u.ext = li->irq.ext;
2387 		break;
2388 	case IRQ_PEND_EXT_EXTERNAL:
2389 		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2390 		irq->u.extcall = li->irq.extcall;
2391 		break;
2392 	case IRQ_PEND_EXT_CLOCK_COMP:
2393 		irq->type = KVM_S390_INT_CLOCK_COMP;
2394 		break;
2395 	case IRQ_PEND_EXT_CPU_TIMER:
2396 		irq->type = KVM_S390_INT_CPU_TIMER;
2397 		break;
2398 	case IRQ_PEND_SIGP_STOP:
2399 		irq->type = KVM_S390_SIGP_STOP;
2400 		irq->u.stop = li->irq.stop;
2401 		break;
2402 	case IRQ_PEND_RESTART:
2403 		irq->type = KVM_S390_RESTART;
2404 		break;
2405 	case IRQ_PEND_SET_PREFIX:
2406 		irq->type = KVM_S390_SIGP_SET_PREFIX;
2407 		irq->u.prefix = li->irq.prefix;
2408 		break;
2409 	}
2410 }
2411 
2412 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2413 {
2414 	int scn;
2415 	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2416 	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2417 	unsigned long pending_irqs;
2418 	struct kvm_s390_irq irq;
2419 	unsigned long irq_type;
2420 	int cpuaddr;
2421 	int n = 0;
2422 
2423 	spin_lock(&li->lock);
2424 	pending_irqs = li->pending_irqs;
2425 	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2426 	       sizeof(sigp_emerg_pending));
2427 	spin_unlock(&li->lock);
2428 
2429 	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2430 		memset(&irq, 0, sizeof(irq));
2431 		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2432 			continue;
2433 		if (n + sizeof(irq) > len)
2434 			return -ENOBUFS;
2435 		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2436 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2437 			return -EFAULT;
2438 		n += sizeof(irq);
2439 	}
2440 
2441 	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2442 		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2443 			memset(&irq, 0, sizeof(irq));
2444 			if (n + sizeof(irq) > len)
2445 				return -ENOBUFS;
2446 			irq.type = KVM_S390_INT_EMERGENCY;
2447 			irq.u.emerg.code = cpuaddr;
2448 			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2449 				return -EFAULT;
2450 			n += sizeof(irq);
2451 		}
2452 	}
2453 
2454 	if (sca_ext_call_pending(vcpu, &scn)) {
2455 		if (n + sizeof(irq) > len)
2456 			return -ENOBUFS;
2457 		memset(&irq, 0, sizeof(irq));
2458 		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2459 		irq.u.extcall.code = scn;
2460 		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2461 			return -EFAULT;
2462 		n += sizeof(irq);
2463 	}
2464 
2465 	return n;
2466 }
2467