1/*
2 * handling kvm guest interrupts
3 *
4 * Copyright IBM Corp. 2008, 2015
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License (version 2 only)
8 * as published by the Free Software Foundation.
9 *
10 *    Author(s): Carsten Otte <cotte@de.ibm.com>
11 */
12
13#include <linux/interrupt.h>
14#include <linux/kvm_host.h>
15#include <linux/hrtimer.h>
16#include <linux/mmu_context.h>
17#include <linux/signal.h>
18#include <linux/slab.h>
19#include <linux/bitmap.h>
20#include <linux/vmalloc.h>
21#include <asm/asm-offsets.h>
22#include <asm/dis.h>
23#include <asm/uaccess.h>
24#include <asm/sclp.h>
25#include <asm/isc.h>
26#include "kvm-s390.h"
27#include "gaccess.h"
28#include "trace-s390.h"
29
30#define IOINT_SCHID_MASK 0x0000ffff
31#define IOINT_SSID_MASK 0x00030000
32#define IOINT_CSSID_MASK 0x03fc0000
33#define PFAULT_INIT 0x0600
34#define PFAULT_DONE 0x0680
35#define VIRTIO_PARAM 0x0d00
36
37int psw_extint_disabled(struct kvm_vcpu *vcpu)
38{
39	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
40}
41
42static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
43{
44	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
45}
46
47static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
48{
49	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
50}
51
52static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
53{
54	return psw_extint_disabled(vcpu) &&
55	       psw_ioint_disabled(vcpu) &&
56	       psw_mchk_disabled(vcpu);
57}
58
59static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
60{
61	if (psw_extint_disabled(vcpu) ||
62	    !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
63		return 0;
64	if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
65		/* No timer interrupts when single stepping */
66		return 0;
67	return 1;
68}
69
70static int ckc_irq_pending(struct kvm_vcpu *vcpu)
71{
72	if (vcpu->arch.sie_block->ckc >= kvm_s390_get_tod_clock_fast(vcpu->kvm))
73		return 0;
74	return ckc_interrupts_enabled(vcpu);
75}
76
77static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
78{
79	return !psw_extint_disabled(vcpu) &&
80	       (vcpu->arch.sie_block->gcr[0] & 0x400ul);
81}
82
83static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
84{
85	return (vcpu->arch.sie_block->cputm >> 63) &&
86	       cpu_timer_interrupts_enabled(vcpu);
87}
88
89static inline int is_ioirq(unsigned long irq_type)
90{
91	return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
92		(irq_type <= IRQ_PEND_IO_ISC_7));
93}
94
95static uint64_t isc_to_isc_bits(int isc)
96{
97	return (0x80 >> isc) << 24;
98}
99
100static inline u8 int_word_to_isc(u32 int_word)
101{
102	return (int_word & 0x38000000) >> 27;
103}
104
105static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
106{
107	return vcpu->kvm->arch.float_int.pending_irqs |
108	       vcpu->arch.local_int.pending_irqs;
109}
110
111static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
112				   unsigned long active_mask)
113{
114	int i;
115
116	for (i = 0; i <= MAX_ISC; i++)
117		if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
118			active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
119
120	return active_mask;
121}
122
123static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
124{
125	unsigned long active_mask;
126
127	active_mask = pending_irqs(vcpu);
128	if (!active_mask)
129		return 0;
130
131	if (psw_extint_disabled(vcpu))
132		active_mask &= ~IRQ_PEND_EXT_MASK;
133	if (psw_ioint_disabled(vcpu))
134		active_mask &= ~IRQ_PEND_IO_MASK;
135	else
136		active_mask = disable_iscs(vcpu, active_mask);
137	if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
138		__clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
139	if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
140		__clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
141	if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
142		__clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
143	if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
144		__clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
145	if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
146		__clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
147	if (psw_mchk_disabled(vcpu))
148		active_mask &= ~IRQ_PEND_MCHK_MASK;
149	if (!(vcpu->arch.sie_block->gcr[14] &
150	      vcpu->kvm->arch.float_int.mchk.cr14))
151		__clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
152
153	/*
154	 * STOP irqs will never be actively delivered. They are triggered via
155	 * intercept requests and cleared when the stop intercept is performed.
156	 */
157	__clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
158
159	return active_mask;
160}
161
162static void __set_cpu_idle(struct kvm_vcpu *vcpu)
163{
164	atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
165	set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
166}
167
168static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
169{
170	atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
171	clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
172}
173
174static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
175{
176	atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
177		    &vcpu->arch.sie_block->cpuflags);
178	vcpu->arch.sie_block->lctl = 0x0000;
179	vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
180
181	if (guestdbg_enabled(vcpu)) {
182		vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
183					       LCTL_CR10 | LCTL_CR11);
184		vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
185	}
186}
187
188static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
189{
190	atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
191}
192
193static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
194{
195	if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
196		return;
197	else if (psw_ioint_disabled(vcpu))
198		__set_cpuflag(vcpu, CPUSTAT_IO_INT);
199	else
200		vcpu->arch.sie_block->lctl |= LCTL_CR6;
201}
202
203static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
204{
205	if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
206		return;
207	if (psw_extint_disabled(vcpu))
208		__set_cpuflag(vcpu, CPUSTAT_EXT_INT);
209	else
210		vcpu->arch.sie_block->lctl |= LCTL_CR0;
211}
212
213static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
214{
215	if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
216		return;
217	if (psw_mchk_disabled(vcpu))
218		vcpu->arch.sie_block->ictl |= ICTL_LPSW;
219	else
220		vcpu->arch.sie_block->lctl |= LCTL_CR14;
221}
222
223static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
224{
225	if (kvm_s390_is_stop_irq_pending(vcpu))
226		__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
227}
228
229/* Set interception request for non-deliverable interrupts */
230static void set_intercept_indicators(struct kvm_vcpu *vcpu)
231{
232	set_intercept_indicators_io(vcpu);
233	set_intercept_indicators_ext(vcpu);
234	set_intercept_indicators_mchk(vcpu);
235	set_intercept_indicators_stop(vcpu);
236}
237
238static u16 get_ilc(struct kvm_vcpu *vcpu)
239{
240	switch (vcpu->arch.sie_block->icptcode) {
241	case ICPT_INST:
242	case ICPT_INSTPROGI:
243	case ICPT_OPEREXC:
244	case ICPT_PARTEXEC:
245	case ICPT_IOINST:
246		/* last instruction only stored for these icptcodes */
247		return insn_length(vcpu->arch.sie_block->ipa >> 8);
248	case ICPT_PROGI:
249		return vcpu->arch.sie_block->pgmilc;
250	default:
251		return 0;
252	}
253}
254
255static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
256{
257	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
258	int rc;
259
260	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
261					 0, 0);
262
263	rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
264			   (u16 *)__LC_EXT_INT_CODE);
265	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
266	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
267			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
268	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
269			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
270	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
271	return rc ? -EFAULT : 0;
272}
273
274static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
275{
276	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
277	int rc;
278
279	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
280					 0, 0);
281
282	rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
283			   (u16 __user *)__LC_EXT_INT_CODE);
284	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
285	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
286			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
287	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
288			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
289	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
290	return rc ? -EFAULT : 0;
291}
292
293static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
294{
295	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
296	struct kvm_s390_ext_info ext;
297	int rc;
298
299	spin_lock(&li->lock);
300	ext = li->irq.ext;
301	clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
302	li->irq.ext.ext_params2 = 0;
303	spin_unlock(&li->lock);
304
305	VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
306		   ext.ext_params2);
307	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
308					 KVM_S390_INT_PFAULT_INIT,
309					 0, ext.ext_params2);
310
311	rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
312	rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
313	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
314			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
315	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
316			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
317	rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
318	return rc ? -EFAULT : 0;
319}
320
321static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
322{
323	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
324	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
325	struct kvm_s390_mchk_info mchk = {};
326	unsigned long adtl_status_addr;
327	int deliver = 0;
328	int rc = 0;
329
330	spin_lock(&fi->lock);
331	spin_lock(&li->lock);
332	if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
333	    test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
334		/*
335		 * If there was an exigent machine check pending, then any
336		 * repressible machine checks that might have been pending
337		 * are indicated along with it, so always clear bits for
338		 * repressible and exigent interrupts
339		 */
340		mchk = li->irq.mchk;
341		clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
342		clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
343		memset(&li->irq.mchk, 0, sizeof(mchk));
344		deliver = 1;
345	}
346	/*
347	 * We indicate floating repressible conditions along with
348	 * other pending conditions. Channel Report Pending and Channel
349	 * Subsystem damage are the only two and and are indicated by
350	 * bits in mcic and masked in cr14.
351	 */
352	if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
353		mchk.mcic |= fi->mchk.mcic;
354		mchk.cr14 |= fi->mchk.cr14;
355		memset(&fi->mchk, 0, sizeof(mchk));
356		deliver = 1;
357	}
358	spin_unlock(&li->lock);
359	spin_unlock(&fi->lock);
360
361	if (deliver) {
362		VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
363			   mchk.mcic);
364		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
365						 KVM_S390_MCHK,
366						 mchk.cr14, mchk.mcic);
367
368		rc  = kvm_s390_vcpu_store_status(vcpu,
369						 KVM_S390_STORE_STATUS_PREFIXED);
370		rc |= read_guest_lc(vcpu, __LC_VX_SAVE_AREA_ADDR,
371				    &adtl_status_addr,
372				    sizeof(unsigned long));
373		rc |= kvm_s390_vcpu_store_adtl_status(vcpu,
374						      adtl_status_addr);
375		rc |= put_guest_lc(vcpu, mchk.mcic,
376				   (u64 __user *) __LC_MCCK_CODE);
377		rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
378				   (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
379		rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
380				     &mchk.fixed_logout,
381				     sizeof(mchk.fixed_logout));
382		rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
383				     &vcpu->arch.sie_block->gpsw,
384				     sizeof(psw_t));
385		rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
386				    &vcpu->arch.sie_block->gpsw,
387				    sizeof(psw_t));
388	}
389	return rc ? -EFAULT : 0;
390}
391
392static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
393{
394	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
395	int rc;
396
397	VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
398	vcpu->stat.deliver_restart_signal++;
399	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
400
401	rc  = write_guest_lc(vcpu,
402			     offsetof(struct _lowcore, restart_old_psw),
403			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
404	rc |= read_guest_lc(vcpu, offsetof(struct _lowcore, restart_psw),
405			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
406	clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
407	return rc ? -EFAULT : 0;
408}
409
410static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
411{
412	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
413	struct kvm_s390_prefix_info prefix;
414
415	spin_lock(&li->lock);
416	prefix = li->irq.prefix;
417	li->irq.prefix.address = 0;
418	clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
419	spin_unlock(&li->lock);
420
421	vcpu->stat.deliver_prefix_signal++;
422	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
423					 KVM_S390_SIGP_SET_PREFIX,
424					 prefix.address, 0);
425
426	kvm_s390_set_prefix(vcpu, prefix.address);
427	return 0;
428}
429
430static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
431{
432	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
433	int rc;
434	int cpu_addr;
435
436	spin_lock(&li->lock);
437	cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
438	clear_bit(cpu_addr, li->sigp_emerg_pending);
439	if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
440		clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
441	spin_unlock(&li->lock);
442
443	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
444	vcpu->stat.deliver_emergency_signal++;
445	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
446					 cpu_addr, 0);
447
448	rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
449			   (u16 *)__LC_EXT_INT_CODE);
450	rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
451	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
452			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
453	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
454			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
455	return rc ? -EFAULT : 0;
456}
457
458static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
459{
460	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
461	struct kvm_s390_extcall_info extcall;
462	int rc;
463
464	spin_lock(&li->lock);
465	extcall = li->irq.extcall;
466	li->irq.extcall.code = 0;
467	clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
468	spin_unlock(&li->lock);
469
470	VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
471	vcpu->stat.deliver_external_call++;
472	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
473					 KVM_S390_INT_EXTERNAL_CALL,
474					 extcall.code, 0);
475
476	rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
477			   (u16 *)__LC_EXT_INT_CODE);
478	rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
479	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
480			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
481	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
482			    sizeof(psw_t));
483	return rc ? -EFAULT : 0;
484}
485
486static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
487{
488	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
489	struct kvm_s390_pgm_info pgm_info;
490	int rc = 0, nullifying = false;
491	u16 ilc = get_ilc(vcpu);
492
493	spin_lock(&li->lock);
494	pgm_info = li->irq.pgm;
495	clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
496	memset(&li->irq.pgm, 0, sizeof(pgm_info));
497	spin_unlock(&li->lock);
498
499	VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilc:%d",
500		   pgm_info.code, ilc);
501	vcpu->stat.deliver_program_int++;
502	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
503					 pgm_info.code, 0);
504
505	switch (pgm_info.code & ~PGM_PER) {
506	case PGM_AFX_TRANSLATION:
507	case PGM_ASX_TRANSLATION:
508	case PGM_EX_TRANSLATION:
509	case PGM_LFX_TRANSLATION:
510	case PGM_LSTE_SEQUENCE:
511	case PGM_LSX_TRANSLATION:
512	case PGM_LX_TRANSLATION:
513	case PGM_PRIMARY_AUTHORITY:
514	case PGM_SECONDARY_AUTHORITY:
515		nullifying = true;
516		/* fall through */
517	case PGM_SPACE_SWITCH:
518		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
519				  (u64 *)__LC_TRANS_EXC_CODE);
520		break;
521	case PGM_ALEN_TRANSLATION:
522	case PGM_ALE_SEQUENCE:
523	case PGM_ASTE_INSTANCE:
524	case PGM_ASTE_SEQUENCE:
525	case PGM_ASTE_VALIDITY:
526	case PGM_EXTENDED_AUTHORITY:
527		rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
528				  (u8 *)__LC_EXC_ACCESS_ID);
529		nullifying = true;
530		break;
531	case PGM_ASCE_TYPE:
532	case PGM_PAGE_TRANSLATION:
533	case PGM_REGION_FIRST_TRANS:
534	case PGM_REGION_SECOND_TRANS:
535	case PGM_REGION_THIRD_TRANS:
536	case PGM_SEGMENT_TRANSLATION:
537		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
538				  (u64 *)__LC_TRANS_EXC_CODE);
539		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
540				   (u8 *)__LC_EXC_ACCESS_ID);
541		rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
542				   (u8 *)__LC_OP_ACCESS_ID);
543		nullifying = true;
544		break;
545	case PGM_MONITOR:
546		rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
547				  (u16 *)__LC_MON_CLASS_NR);
548		rc |= put_guest_lc(vcpu, pgm_info.mon_code,
549				   (u64 *)__LC_MON_CODE);
550		break;
551	case PGM_VECTOR_PROCESSING:
552	case PGM_DATA:
553		rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
554				  (u32 *)__LC_DATA_EXC_CODE);
555		break;
556	case PGM_PROTECTION:
557		rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
558				  (u64 *)__LC_TRANS_EXC_CODE);
559		rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
560				   (u8 *)__LC_EXC_ACCESS_ID);
561		break;
562	case PGM_STACK_FULL:
563	case PGM_STACK_EMPTY:
564	case PGM_STACK_SPECIFICATION:
565	case PGM_STACK_TYPE:
566	case PGM_STACK_OPERATION:
567	case PGM_TRACE_TABEL:
568	case PGM_CRYPTO_OPERATION:
569		nullifying = true;
570		break;
571	}
572
573	if (pgm_info.code & PGM_PER) {
574		rc |= put_guest_lc(vcpu, pgm_info.per_code,
575				   (u8 *) __LC_PER_CODE);
576		rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
577				   (u8 *)__LC_PER_ATMID);
578		rc |= put_guest_lc(vcpu, pgm_info.per_address,
579				   (u64 *) __LC_PER_ADDRESS);
580		rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
581				   (u8 *) __LC_PER_ACCESS_ID);
582	}
583
584	if (nullifying && vcpu->arch.sie_block->icptcode == ICPT_INST)
585		kvm_s390_rewind_psw(vcpu, ilc);
586
587	rc |= put_guest_lc(vcpu, ilc, (u16 *) __LC_PGM_ILC);
588	rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
589				 (u64 *) __LC_LAST_BREAK);
590	rc |= put_guest_lc(vcpu, pgm_info.code,
591			   (u16 *)__LC_PGM_INT_CODE);
592	rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
593			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
594	rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
595			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
596	return rc ? -EFAULT : 0;
597}
598
599static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
600{
601	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
602	struct kvm_s390_ext_info ext;
603	int rc = 0;
604
605	spin_lock(&fi->lock);
606	if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
607		spin_unlock(&fi->lock);
608		return 0;
609	}
610	ext = fi->srv_signal;
611	memset(&fi->srv_signal, 0, sizeof(ext));
612	clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
613	spin_unlock(&fi->lock);
614
615	VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
616		   ext.ext_params);
617	vcpu->stat.deliver_service_signal++;
618	trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
619					 ext.ext_params, 0);
620
621	rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
622	rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
623	rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
624			     &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
625	rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
626			    &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
627	rc |= put_guest_lc(vcpu, ext.ext_params,
628			   (u32 *)__LC_EXT_PARAMS);
629
630	return rc ? -EFAULT : 0;
631}
632
633static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
634{
635	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
636	struct kvm_s390_interrupt_info *inti;
637	int rc = 0;
638
639	spin_lock(&fi->lock);
640	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
641					struct kvm_s390_interrupt_info,
642					list);
643	if (inti) {
644		list_del(&inti->list);
645		fi->counters[FIRQ_CNTR_PFAULT] -= 1;
646	}
647	if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
648		clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
649	spin_unlock(&fi->lock);
650
651	if (inti) {
652		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
653						 KVM_S390_INT_PFAULT_DONE, 0,
654						 inti->ext.ext_params2);
655		VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
656			   inti->ext.ext_params2);
657
658		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
659				(u16 *)__LC_EXT_INT_CODE);
660		rc |= put_guest_lc(vcpu, PFAULT_DONE,
661				(u16 *)__LC_EXT_CPU_ADDR);
662		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
663				&vcpu->arch.sie_block->gpsw,
664				sizeof(psw_t));
665		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
666				&vcpu->arch.sie_block->gpsw,
667				sizeof(psw_t));
668		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
669				(u64 *)__LC_EXT_PARAMS2);
670		kfree(inti);
671	}
672	return rc ? -EFAULT : 0;
673}
674
675static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
676{
677	struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
678	struct kvm_s390_interrupt_info *inti;
679	int rc = 0;
680
681	spin_lock(&fi->lock);
682	inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
683					struct kvm_s390_interrupt_info,
684					list);
685	if (inti) {
686		VCPU_EVENT(vcpu, 4,
687			   "deliver: virtio parm: 0x%x,parm64: 0x%llx",
688			   inti->ext.ext_params, inti->ext.ext_params2);
689		vcpu->stat.deliver_virtio_interrupt++;
690		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
691				inti->type,
692				inti->ext.ext_params,
693				inti->ext.ext_params2);
694		list_del(&inti->list);
695		fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
696	}
697	if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
698		clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
699	spin_unlock(&fi->lock);
700
701	if (inti) {
702		rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
703				(u16 *)__LC_EXT_INT_CODE);
704		rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
705				(u16 *)__LC_EXT_CPU_ADDR);
706		rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
707				&vcpu->arch.sie_block->gpsw,
708				sizeof(psw_t));
709		rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
710				&vcpu->arch.sie_block->gpsw,
711				sizeof(psw_t));
712		rc |= put_guest_lc(vcpu, inti->ext.ext_params,
713				(u32 *)__LC_EXT_PARAMS);
714		rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
715				(u64 *)__LC_EXT_PARAMS2);
716		kfree(inti);
717	}
718	return rc ? -EFAULT : 0;
719}
720
721static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
722				     unsigned long irq_type)
723{
724	struct list_head *isc_list;
725	struct kvm_s390_float_interrupt *fi;
726	struct kvm_s390_interrupt_info *inti = NULL;
727	int rc = 0;
728
729	fi = &vcpu->kvm->arch.float_int;
730
731	spin_lock(&fi->lock);
732	isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
733	inti = list_first_entry_or_null(isc_list,
734					struct kvm_s390_interrupt_info,
735					list);
736	if (inti) {
737		VCPU_EVENT(vcpu, 4, "deliver: I/O 0x%llx", inti->type);
738		vcpu->stat.deliver_io_int++;
739		trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
740				inti->type,
741				((__u32)inti->io.subchannel_id << 16) |
742				inti->io.subchannel_nr,
743				((__u64)inti->io.io_int_parm << 32) |
744				inti->io.io_int_word);
745		list_del(&inti->list);
746		fi->counters[FIRQ_CNTR_IO] -= 1;
747	}
748	if (list_empty(isc_list))
749		clear_bit(irq_type, &fi->pending_irqs);
750	spin_unlock(&fi->lock);
751
752	if (inti) {
753		rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
754				(u16 *)__LC_SUBCHANNEL_ID);
755		rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
756				(u16 *)__LC_SUBCHANNEL_NR);
757		rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
758				(u32 *)__LC_IO_INT_PARM);
759		rc |= put_guest_lc(vcpu, inti->io.io_int_word,
760				(u32 *)__LC_IO_INT_WORD);
761		rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
762				&vcpu->arch.sie_block->gpsw,
763				sizeof(psw_t));
764		rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
765				&vcpu->arch.sie_block->gpsw,
766				sizeof(psw_t));
767		kfree(inti);
768	}
769
770	return rc ? -EFAULT : 0;
771}
772
773typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
774
775static const deliver_irq_t deliver_irq_funcs[] = {
776	[IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
777	[IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
778	[IRQ_PEND_PROG]           = __deliver_prog,
779	[IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
780	[IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
781	[IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
782	[IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
783	[IRQ_PEND_RESTART]        = __deliver_restart,
784	[IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
785	[IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
786	[IRQ_PEND_EXT_SERVICE]    = __deliver_service,
787	[IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
788	[IRQ_PEND_VIRTIO]         = __deliver_virtio,
789};
790
791/* Check whether an external call is pending (deliverable or not) */
792int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
793{
794	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
795	uint8_t sigp_ctrl = vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
796
797	if (!sclp.has_sigpif)
798		return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
799
800	return (sigp_ctrl & SIGP_CTRL_C) &&
801	       (atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND);
802}
803
804int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
805{
806	if (deliverable_irqs(vcpu))
807		return 1;
808
809	if (kvm_cpu_has_pending_timer(vcpu))
810		return 1;
811
812	/* external call pending and deliverable */
813	if (kvm_s390_ext_call_pending(vcpu) &&
814	    !psw_extint_disabled(vcpu) &&
815	    (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
816		return 1;
817
818	if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
819		return 1;
820	return 0;
821}
822
823int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
824{
825	return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
826}
827
828int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
829{
830	u64 now, sltime;
831
832	vcpu->stat.exit_wait_state++;
833
834	/* fast path */
835	if (kvm_arch_vcpu_runnable(vcpu))
836		return 0;
837
838	if (psw_interrupts_disabled(vcpu)) {
839		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
840		return -EOPNOTSUPP; /* disabled wait */
841	}
842
843	if (!ckc_interrupts_enabled(vcpu)) {
844		VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
845		__set_cpu_idle(vcpu);
846		goto no_timer;
847	}
848
849	now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
850	sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
851
852	/* underflow */
853	if (vcpu->arch.sie_block->ckc < now)
854		return 0;
855
856	__set_cpu_idle(vcpu);
857	hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
858	VCPU_EVENT(vcpu, 4, "enabled wait via clock comparator: %llu ns", sltime);
859no_timer:
860	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
861	kvm_vcpu_block(vcpu);
862	__unset_cpu_idle(vcpu);
863	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
864
865	hrtimer_cancel(&vcpu->arch.ckc_timer);
866	return 0;
867}
868
869void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
870{
871	if (waitqueue_active(&vcpu->wq)) {
872		/*
873		 * The vcpu gave up the cpu voluntarily, mark it as a good
874		 * yield-candidate.
875		 */
876		vcpu->preempted = true;
877		wake_up_interruptible(&vcpu->wq);
878		vcpu->stat.halt_wakeup++;
879	}
880}
881
882enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
883{
884	struct kvm_vcpu *vcpu;
885	u64 now, sltime;
886
887	vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
888	now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
889	sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
890
891	/*
892	 * If the monotonic clock runs faster than the tod clock we might be
893	 * woken up too early and have to go back to sleep to avoid deadlocks.
894	 */
895	if (vcpu->arch.sie_block->ckc > now &&
896	    hrtimer_forward_now(timer, ns_to_ktime(sltime)))
897		return HRTIMER_RESTART;
898	kvm_s390_vcpu_wakeup(vcpu);
899	return HRTIMER_NORESTART;
900}
901
902void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
903{
904	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
905
906	spin_lock(&li->lock);
907	li->pending_irqs = 0;
908	bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
909	memset(&li->irq, 0, sizeof(li->irq));
910	spin_unlock(&li->lock);
911
912	/* clear pending external calls set by sigp interpretation facility */
913	atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
914	vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl = 0;
915}
916
917int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
918{
919	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
920	deliver_irq_t func;
921	int rc = 0;
922	unsigned long irq_type;
923	unsigned long irqs;
924
925	__reset_intercept_indicators(vcpu);
926
927	/* pending ckc conditions might have been invalidated */
928	clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
929	if (ckc_irq_pending(vcpu))
930		set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
931
932	/* pending cpu timer conditions might have been invalidated */
933	clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
934	if (cpu_timer_irq_pending(vcpu))
935		set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
936
937	while ((irqs = deliverable_irqs(vcpu)) && !rc) {
938		/* bits are in the order of interrupt priority */
939		irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
940		if (is_ioirq(irq_type)) {
941			rc = __deliver_io(vcpu, irq_type);
942		} else {
943			func = deliver_irq_funcs[irq_type];
944			if (!func) {
945				WARN_ON_ONCE(func == NULL);
946				clear_bit(irq_type, &li->pending_irqs);
947				continue;
948			}
949			rc = func(vcpu);
950		}
951	}
952
953	set_intercept_indicators(vcpu);
954
955	return rc;
956}
957
958static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
959{
960	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
961
962	VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
963	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
964				   irq->u.pgm.code, 0);
965
966	if (irq->u.pgm.code == PGM_PER) {
967		li->irq.pgm.code |= PGM_PER;
968		/* only modify PER related information */
969		li->irq.pgm.per_address = irq->u.pgm.per_address;
970		li->irq.pgm.per_code = irq->u.pgm.per_code;
971		li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
972		li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
973	} else if (!(irq->u.pgm.code & PGM_PER)) {
974		li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
975				   irq->u.pgm.code;
976		/* only modify non-PER information */
977		li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
978		li->irq.pgm.mon_code = irq->u.pgm.mon_code;
979		li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
980		li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
981		li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
982		li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
983	} else {
984		li->irq.pgm = irq->u.pgm;
985	}
986	set_bit(IRQ_PEND_PROG, &li->pending_irqs);
987	return 0;
988}
989
990static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
991{
992	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
993
994	VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
995		   irq->u.ext.ext_params2);
996	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
997				   irq->u.ext.ext_params,
998				   irq->u.ext.ext_params2);
999
1000	li->irq.ext = irq->u.ext;
1001	set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1002	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1003	return 0;
1004}
1005
1006static int __inject_extcall_sigpif(struct kvm_vcpu *vcpu, uint16_t src_id)
1007{
1008	unsigned char new_val, old_val;
1009	uint8_t *sigp_ctrl = &vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
1010
1011	new_val = SIGP_CTRL_C | (src_id & SIGP_CTRL_SCN_MASK);
1012	old_val = *sigp_ctrl & ~SIGP_CTRL_C;
1013	if (cmpxchg(sigp_ctrl, old_val, new_val) != old_val) {
1014		/* another external call is pending */
1015		return -EBUSY;
1016	}
1017	atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
1018	return 0;
1019}
1020
1021static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1022{
1023	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1024	struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1025	uint16_t src_id = irq->u.extcall.code;
1026
1027	VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1028		   src_id);
1029	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1030				   src_id, 0);
1031
1032	/* sending vcpu invalid */
1033	if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1034		return -EINVAL;
1035
1036	if (sclp.has_sigpif)
1037		return __inject_extcall_sigpif(vcpu, src_id);
1038
1039	if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1040		return -EBUSY;
1041	*extcall = irq->u.extcall;
1042	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1043	return 0;
1044}
1045
1046static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1047{
1048	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1049	struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1050
1051	VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1052		   irq->u.prefix.address);
1053	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1054				   irq->u.prefix.address, 0);
1055
1056	if (!is_vcpu_stopped(vcpu))
1057		return -EBUSY;
1058
1059	*prefix = irq->u.prefix;
1060	set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1061	return 0;
1062}
1063
1064#define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1065static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1066{
1067	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1068	struct kvm_s390_stop_info *stop = &li->irq.stop;
1069	int rc = 0;
1070
1071	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1072
1073	if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1074		return -EINVAL;
1075
1076	if (is_vcpu_stopped(vcpu)) {
1077		if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1078			rc = kvm_s390_store_status_unloaded(vcpu,
1079						KVM_S390_STORE_STATUS_NOADDR);
1080		return rc;
1081	}
1082
1083	if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1084		return -EBUSY;
1085	stop->flags = irq->u.stop.flags;
1086	__set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1087	return 0;
1088}
1089
1090static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1091				 struct kvm_s390_irq *irq)
1092{
1093	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1094
1095	VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1096	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1097
1098	set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1099	return 0;
1100}
1101
1102static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1103				   struct kvm_s390_irq *irq)
1104{
1105	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1106
1107	VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1108		   irq->u.emerg.code);
1109	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1110				   irq->u.emerg.code, 0);
1111
1112	/* sending vcpu invalid */
1113	if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1114		return -EINVAL;
1115
1116	set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1117	set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1118	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1119	return 0;
1120}
1121
1122static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1123{
1124	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1125	struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1126
1127	VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1128		   irq->u.mchk.mcic);
1129	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1130				   irq->u.mchk.mcic);
1131
1132	/*
1133	 * Because repressible machine checks can be indicated along with
1134	 * exigent machine checks (PoP, Chapter 11, Interruption action)
1135	 * we need to combine cr14, mcic and external damage code.
1136	 * Failing storage address and the logout area should not be or'ed
1137	 * together, we just indicate the last occurrence of the corresponding
1138	 * machine check
1139	 */
1140	mchk->cr14 |= irq->u.mchk.cr14;
1141	mchk->mcic |= irq->u.mchk.mcic;
1142	mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1143	mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1144	memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1145	       sizeof(mchk->fixed_logout));
1146	if (mchk->mcic & MCHK_EX_MASK)
1147		set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1148	else if (mchk->mcic & MCHK_REP_MASK)
1149		set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1150	return 0;
1151}
1152
1153static int __inject_ckc(struct kvm_vcpu *vcpu)
1154{
1155	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1156
1157	VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1158	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1159				   0, 0);
1160
1161	set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1162	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1163	return 0;
1164}
1165
1166static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1167{
1168	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1169
1170	VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1171	trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1172				   0, 0);
1173
1174	set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1175	atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1176	return 0;
1177}
1178
1179static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1180						  int isc, u32 schid)
1181{
1182	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1183	struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1184	struct kvm_s390_interrupt_info *iter;
1185	u16 id = (schid & 0xffff0000U) >> 16;
1186	u16 nr = schid & 0x0000ffffU;
1187
1188	spin_lock(&fi->lock);
1189	list_for_each_entry(iter, isc_list, list) {
1190		if (schid && (id != iter->io.subchannel_id ||
1191			      nr != iter->io.subchannel_nr))
1192			continue;
1193		/* found an appropriate entry */
1194		list_del_init(&iter->list);
1195		fi->counters[FIRQ_CNTR_IO] -= 1;
1196		if (list_empty(isc_list))
1197			clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1198		spin_unlock(&fi->lock);
1199		return iter;
1200	}
1201	spin_unlock(&fi->lock);
1202	return NULL;
1203}
1204
1205/*
1206 * Dequeue and return an I/O interrupt matching any of the interruption
1207 * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1208 */
1209struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1210						    u64 isc_mask, u32 schid)
1211{
1212	struct kvm_s390_interrupt_info *inti = NULL;
1213	int isc;
1214
1215	for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1216		if (isc_mask & isc_to_isc_bits(isc))
1217			inti = get_io_int(kvm, isc, schid);
1218	}
1219	return inti;
1220}
1221
1222#define SCCB_MASK 0xFFFFFFF8
1223#define SCCB_EVENT_PENDING 0x3
1224
1225static int __inject_service(struct kvm *kvm,
1226			     struct kvm_s390_interrupt_info *inti)
1227{
1228	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1229
1230	spin_lock(&fi->lock);
1231	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1232	/*
1233	 * Early versions of the QEMU s390 bios will inject several
1234	 * service interrupts after another without handling a
1235	 * condition code indicating busy.
1236	 * We will silently ignore those superfluous sccb values.
1237	 * A future version of QEMU will take care of serialization
1238	 * of servc requests
1239	 */
1240	if (fi->srv_signal.ext_params & SCCB_MASK)
1241		goto out;
1242	fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1243	set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1244out:
1245	spin_unlock(&fi->lock);
1246	kfree(inti);
1247	return 0;
1248}
1249
1250static int __inject_virtio(struct kvm *kvm,
1251			    struct kvm_s390_interrupt_info *inti)
1252{
1253	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1254
1255	spin_lock(&fi->lock);
1256	if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1257		spin_unlock(&fi->lock);
1258		return -EBUSY;
1259	}
1260	fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1261	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1262	set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1263	spin_unlock(&fi->lock);
1264	return 0;
1265}
1266
1267static int __inject_pfault_done(struct kvm *kvm,
1268				 struct kvm_s390_interrupt_info *inti)
1269{
1270	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1271
1272	spin_lock(&fi->lock);
1273	if (fi->counters[FIRQ_CNTR_PFAULT] >=
1274		(ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1275		spin_unlock(&fi->lock);
1276		return -EBUSY;
1277	}
1278	fi->counters[FIRQ_CNTR_PFAULT] += 1;
1279	list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1280	set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1281	spin_unlock(&fi->lock);
1282	return 0;
1283}
1284
1285#define CR_PENDING_SUBCLASS 28
1286static int __inject_float_mchk(struct kvm *kvm,
1287				struct kvm_s390_interrupt_info *inti)
1288{
1289	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1290
1291	spin_lock(&fi->lock);
1292	fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1293	fi->mchk.mcic |= inti->mchk.mcic;
1294	set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1295	spin_unlock(&fi->lock);
1296	kfree(inti);
1297	return 0;
1298}
1299
1300static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1301{
1302	struct kvm_s390_float_interrupt *fi;
1303	struct list_head *list;
1304	int isc;
1305
1306	fi = &kvm->arch.float_int;
1307	spin_lock(&fi->lock);
1308	if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1309		spin_unlock(&fi->lock);
1310		return -EBUSY;
1311	}
1312	fi->counters[FIRQ_CNTR_IO] += 1;
1313
1314	isc = int_word_to_isc(inti->io.io_int_word);
1315	list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1316	list_add_tail(&inti->list, list);
1317	set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1318	spin_unlock(&fi->lock);
1319	return 0;
1320}
1321
1322/*
1323 * Find a destination VCPU for a floating irq and kick it.
1324 */
1325static void __floating_irq_kick(struct kvm *kvm, u64 type)
1326{
1327	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1328	struct kvm_s390_local_interrupt *li;
1329	struct kvm_vcpu *dst_vcpu;
1330	int sigcpu, online_vcpus, nr_tries = 0;
1331
1332	online_vcpus = atomic_read(&kvm->online_vcpus);
1333	if (!online_vcpus)
1334		return;
1335
1336	/* find idle VCPUs first, then round robin */
1337	sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1338	if (sigcpu == online_vcpus) {
1339		do {
1340			sigcpu = fi->next_rr_cpu;
1341			fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1342			/* avoid endless loops if all vcpus are stopped */
1343			if (nr_tries++ >= online_vcpus)
1344				return;
1345		} while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1346	}
1347	dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1348
1349	/* make the VCPU drop out of the SIE, or wake it up if sleeping */
1350	li = &dst_vcpu->arch.local_int;
1351	spin_lock(&li->lock);
1352	switch (type) {
1353	case KVM_S390_MCHK:
1354		atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1355		break;
1356	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1357		atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1358		break;
1359	default:
1360		atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1361		break;
1362	}
1363	spin_unlock(&li->lock);
1364	kvm_s390_vcpu_wakeup(dst_vcpu);
1365}
1366
1367static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1368{
1369	u64 type = READ_ONCE(inti->type);
1370	int rc;
1371
1372	switch (type) {
1373	case KVM_S390_MCHK:
1374		rc = __inject_float_mchk(kvm, inti);
1375		break;
1376	case KVM_S390_INT_VIRTIO:
1377		rc = __inject_virtio(kvm, inti);
1378		break;
1379	case KVM_S390_INT_SERVICE:
1380		rc = __inject_service(kvm, inti);
1381		break;
1382	case KVM_S390_INT_PFAULT_DONE:
1383		rc = __inject_pfault_done(kvm, inti);
1384		break;
1385	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1386		rc = __inject_io(kvm, inti);
1387		break;
1388	default:
1389		rc = -EINVAL;
1390	}
1391	if (rc)
1392		return rc;
1393
1394	__floating_irq_kick(kvm, type);
1395	return 0;
1396}
1397
1398int kvm_s390_inject_vm(struct kvm *kvm,
1399		       struct kvm_s390_interrupt *s390int)
1400{
1401	struct kvm_s390_interrupt_info *inti;
1402	int rc;
1403
1404	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1405	if (!inti)
1406		return -ENOMEM;
1407
1408	inti->type = s390int->type;
1409	switch (inti->type) {
1410	case KVM_S390_INT_VIRTIO:
1411		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1412			 s390int->parm, s390int->parm64);
1413		inti->ext.ext_params = s390int->parm;
1414		inti->ext.ext_params2 = s390int->parm64;
1415		break;
1416	case KVM_S390_INT_SERVICE:
1417		VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1418		inti->ext.ext_params = s390int->parm;
1419		break;
1420	case KVM_S390_INT_PFAULT_DONE:
1421		inti->ext.ext_params2 = s390int->parm64;
1422		break;
1423	case KVM_S390_MCHK:
1424		VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1425			 s390int->parm64);
1426		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1427		inti->mchk.mcic = s390int->parm64;
1428		break;
1429	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1430		if (inti->type & KVM_S390_INT_IO_AI_MASK)
1431			VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
1432		else
1433			VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
1434				 s390int->type & IOINT_CSSID_MASK,
1435				 s390int->type & IOINT_SSID_MASK,
1436				 s390int->type & IOINT_SCHID_MASK);
1437		inti->io.subchannel_id = s390int->parm >> 16;
1438		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1439		inti->io.io_int_parm = s390int->parm64 >> 32;
1440		inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1441		break;
1442	default:
1443		kfree(inti);
1444		return -EINVAL;
1445	}
1446	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1447				 2);
1448
1449	rc = __inject_vm(kvm, inti);
1450	if (rc)
1451		kfree(inti);
1452	return rc;
1453}
1454
1455int kvm_s390_reinject_io_int(struct kvm *kvm,
1456			      struct kvm_s390_interrupt_info *inti)
1457{
1458	return __inject_vm(kvm, inti);
1459}
1460
1461int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1462		       struct kvm_s390_irq *irq)
1463{
1464	irq->type = s390int->type;
1465	switch (irq->type) {
1466	case KVM_S390_PROGRAM_INT:
1467		if (s390int->parm & 0xffff0000)
1468			return -EINVAL;
1469		irq->u.pgm.code = s390int->parm;
1470		break;
1471	case KVM_S390_SIGP_SET_PREFIX:
1472		irq->u.prefix.address = s390int->parm;
1473		break;
1474	case KVM_S390_SIGP_STOP:
1475		irq->u.stop.flags = s390int->parm;
1476		break;
1477	case KVM_S390_INT_EXTERNAL_CALL:
1478		if (s390int->parm & 0xffff0000)
1479			return -EINVAL;
1480		irq->u.extcall.code = s390int->parm;
1481		break;
1482	case KVM_S390_INT_EMERGENCY:
1483		if (s390int->parm & 0xffff0000)
1484			return -EINVAL;
1485		irq->u.emerg.code = s390int->parm;
1486		break;
1487	case KVM_S390_MCHK:
1488		irq->u.mchk.mcic = s390int->parm64;
1489		break;
1490	}
1491	return 0;
1492}
1493
1494int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1495{
1496	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1497
1498	return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1499}
1500
1501void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1502{
1503	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1504
1505	spin_lock(&li->lock);
1506	li->irq.stop.flags = 0;
1507	clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1508	spin_unlock(&li->lock);
1509}
1510
1511static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1512{
1513	int rc;
1514
1515	switch (irq->type) {
1516	case KVM_S390_PROGRAM_INT:
1517		rc = __inject_prog(vcpu, irq);
1518		break;
1519	case KVM_S390_SIGP_SET_PREFIX:
1520		rc = __inject_set_prefix(vcpu, irq);
1521		break;
1522	case KVM_S390_SIGP_STOP:
1523		rc = __inject_sigp_stop(vcpu, irq);
1524		break;
1525	case KVM_S390_RESTART:
1526		rc = __inject_sigp_restart(vcpu, irq);
1527		break;
1528	case KVM_S390_INT_CLOCK_COMP:
1529		rc = __inject_ckc(vcpu);
1530		break;
1531	case KVM_S390_INT_CPU_TIMER:
1532		rc = __inject_cpu_timer(vcpu);
1533		break;
1534	case KVM_S390_INT_EXTERNAL_CALL:
1535		rc = __inject_extcall(vcpu, irq);
1536		break;
1537	case KVM_S390_INT_EMERGENCY:
1538		rc = __inject_sigp_emergency(vcpu, irq);
1539		break;
1540	case KVM_S390_MCHK:
1541		rc = __inject_mchk(vcpu, irq);
1542		break;
1543	case KVM_S390_INT_PFAULT_INIT:
1544		rc = __inject_pfault_init(vcpu, irq);
1545		break;
1546	case KVM_S390_INT_VIRTIO:
1547	case KVM_S390_INT_SERVICE:
1548	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1549	default:
1550		rc = -EINVAL;
1551	}
1552
1553	return rc;
1554}
1555
1556int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1557{
1558	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1559	int rc;
1560
1561	spin_lock(&li->lock);
1562	rc = do_inject_vcpu(vcpu, irq);
1563	spin_unlock(&li->lock);
1564	if (!rc)
1565		kvm_s390_vcpu_wakeup(vcpu);
1566	return rc;
1567}
1568
1569static inline void clear_irq_list(struct list_head *_list)
1570{
1571	struct kvm_s390_interrupt_info *inti, *n;
1572
1573	list_for_each_entry_safe(inti, n, _list, list) {
1574		list_del(&inti->list);
1575		kfree(inti);
1576	}
1577}
1578
1579static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1580		       struct kvm_s390_irq *irq)
1581{
1582	irq->type = inti->type;
1583	switch (inti->type) {
1584	case KVM_S390_INT_PFAULT_INIT:
1585	case KVM_S390_INT_PFAULT_DONE:
1586	case KVM_S390_INT_VIRTIO:
1587		irq->u.ext = inti->ext;
1588		break;
1589	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1590		irq->u.io = inti->io;
1591		break;
1592	}
1593}
1594
1595void kvm_s390_clear_float_irqs(struct kvm *kvm)
1596{
1597	struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1598	int i;
1599
1600	spin_lock(&fi->lock);
1601	fi->pending_irqs = 0;
1602	memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1603	memset(&fi->mchk, 0, sizeof(fi->mchk));
1604	for (i = 0; i < FIRQ_LIST_COUNT; i++)
1605		clear_irq_list(&fi->lists[i]);
1606	for (i = 0; i < FIRQ_MAX_COUNT; i++)
1607		fi->counters[i] = 0;
1608	spin_unlock(&fi->lock);
1609};
1610
1611static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1612{
1613	struct kvm_s390_interrupt_info *inti;
1614	struct kvm_s390_float_interrupt *fi;
1615	struct kvm_s390_irq *buf;
1616	struct kvm_s390_irq *irq;
1617	int max_irqs;
1618	int ret = 0;
1619	int n = 0;
1620	int i;
1621
1622	if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1623		return -EINVAL;
1624
1625	/*
1626	 * We are already using -ENOMEM to signal
1627	 * userspace it may retry with a bigger buffer,
1628	 * so we need to use something else for this case
1629	 */
1630	buf = vzalloc(len);
1631	if (!buf)
1632		return -ENOBUFS;
1633
1634	max_irqs = len / sizeof(struct kvm_s390_irq);
1635
1636	fi = &kvm->arch.float_int;
1637	spin_lock(&fi->lock);
1638	for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1639		list_for_each_entry(inti, &fi->lists[i], list) {
1640			if (n == max_irqs) {
1641				/* signal userspace to try again */
1642				ret = -ENOMEM;
1643				goto out;
1644			}
1645			inti_to_irq(inti, &buf[n]);
1646			n++;
1647		}
1648	}
1649	if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1650		if (n == max_irqs) {
1651			/* signal userspace to try again */
1652			ret = -ENOMEM;
1653			goto out;
1654		}
1655		irq = (struct kvm_s390_irq *) &buf[n];
1656		irq->type = KVM_S390_INT_SERVICE;
1657		irq->u.ext = fi->srv_signal;
1658		n++;
1659	}
1660	if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1661		if (n == max_irqs) {
1662				/* signal userspace to try again */
1663				ret = -ENOMEM;
1664				goto out;
1665		}
1666		irq = (struct kvm_s390_irq *) &buf[n];
1667		irq->type = KVM_S390_MCHK;
1668		irq->u.mchk = fi->mchk;
1669		n++;
1670}
1671
1672out:
1673	spin_unlock(&fi->lock);
1674	if (!ret && n > 0) {
1675		if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1676			ret = -EFAULT;
1677	}
1678	vfree(buf);
1679
1680	return ret < 0 ? ret : n;
1681}
1682
1683static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1684{
1685	int r;
1686
1687	switch (attr->group) {
1688	case KVM_DEV_FLIC_GET_ALL_IRQS:
1689		r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1690					  attr->attr);
1691		break;
1692	default:
1693		r = -EINVAL;
1694	}
1695
1696	return r;
1697}
1698
1699static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1700				     u64 addr)
1701{
1702	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1703	void *target = NULL;
1704	void __user *source;
1705	u64 size;
1706
1707	if (get_user(inti->type, (u64 __user *)addr))
1708		return -EFAULT;
1709
1710	switch (inti->type) {
1711	case KVM_S390_INT_PFAULT_INIT:
1712	case KVM_S390_INT_PFAULT_DONE:
1713	case KVM_S390_INT_VIRTIO:
1714	case KVM_S390_INT_SERVICE:
1715		target = (void *) &inti->ext;
1716		source = &uptr->u.ext;
1717		size = sizeof(inti->ext);
1718		break;
1719	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1720		target = (void *) &inti->io;
1721		source = &uptr->u.io;
1722		size = sizeof(inti->io);
1723		break;
1724	case KVM_S390_MCHK:
1725		target = (void *) &inti->mchk;
1726		source = &uptr->u.mchk;
1727		size = sizeof(inti->mchk);
1728		break;
1729	default:
1730		return -EINVAL;
1731	}
1732
1733	if (copy_from_user(target, source, size))
1734		return -EFAULT;
1735
1736	return 0;
1737}
1738
1739static int enqueue_floating_irq(struct kvm_device *dev,
1740				struct kvm_device_attr *attr)
1741{
1742	struct kvm_s390_interrupt_info *inti = NULL;
1743	int r = 0;
1744	int len = attr->attr;
1745
1746	if (len % sizeof(struct kvm_s390_irq) != 0)
1747		return -EINVAL;
1748	else if (len > KVM_S390_FLIC_MAX_BUFFER)
1749		return -EINVAL;
1750
1751	while (len >= sizeof(struct kvm_s390_irq)) {
1752		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1753		if (!inti)
1754			return -ENOMEM;
1755
1756		r = copy_irq_from_user(inti, attr->addr);
1757		if (r) {
1758			kfree(inti);
1759			return r;
1760		}
1761		r = __inject_vm(dev->kvm, inti);
1762		if (r) {
1763			kfree(inti);
1764			return r;
1765		}
1766		len -= sizeof(struct kvm_s390_irq);
1767		attr->addr += sizeof(struct kvm_s390_irq);
1768	}
1769
1770	return r;
1771}
1772
1773static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1774{
1775	if (id >= MAX_S390_IO_ADAPTERS)
1776		return NULL;
1777	return kvm->arch.adapters[id];
1778}
1779
1780static int register_io_adapter(struct kvm_device *dev,
1781			       struct kvm_device_attr *attr)
1782{
1783	struct s390_io_adapter *adapter;
1784	struct kvm_s390_io_adapter adapter_info;
1785
1786	if (copy_from_user(&adapter_info,
1787			   (void __user *)attr->addr, sizeof(adapter_info)))
1788		return -EFAULT;
1789
1790	if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1791	    (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1792		return -EINVAL;
1793
1794	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1795	if (!adapter)
1796		return -ENOMEM;
1797
1798	INIT_LIST_HEAD(&adapter->maps);
1799	init_rwsem(&adapter->maps_lock);
1800	atomic_set(&adapter->nr_maps, 0);
1801	adapter->id = adapter_info.id;
1802	adapter->isc = adapter_info.isc;
1803	adapter->maskable = adapter_info.maskable;
1804	adapter->masked = false;
1805	adapter->swap = adapter_info.swap;
1806	dev->kvm->arch.adapters[adapter->id] = adapter;
1807
1808	return 0;
1809}
1810
1811int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1812{
1813	int ret;
1814	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1815
1816	if (!adapter || !adapter->maskable)
1817		return -EINVAL;
1818	ret = adapter->masked;
1819	adapter->masked = masked;
1820	return ret;
1821}
1822
1823static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1824{
1825	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1826	struct s390_map_info *map;
1827	int ret;
1828
1829	if (!adapter || !addr)
1830		return -EINVAL;
1831
1832	map = kzalloc(sizeof(*map), GFP_KERNEL);
1833	if (!map) {
1834		ret = -ENOMEM;
1835		goto out;
1836	}
1837	INIT_LIST_HEAD(&map->list);
1838	map->guest_addr = addr;
1839	map->addr = gmap_translate(kvm->arch.gmap, addr);
1840	if (map->addr == -EFAULT) {
1841		ret = -EFAULT;
1842		goto out;
1843	}
1844	ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1845	if (ret < 0)
1846		goto out;
1847	BUG_ON(ret != 1);
1848	down_write(&adapter->maps_lock);
1849	if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1850		list_add_tail(&map->list, &adapter->maps);
1851		ret = 0;
1852	} else {
1853		put_page(map->page);
1854		ret = -EINVAL;
1855	}
1856	up_write(&adapter->maps_lock);
1857out:
1858	if (ret)
1859		kfree(map);
1860	return ret;
1861}
1862
1863static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1864{
1865	struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1866	struct s390_map_info *map, *tmp;
1867	int found = 0;
1868
1869	if (!adapter || !addr)
1870		return -EINVAL;
1871
1872	down_write(&adapter->maps_lock);
1873	list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1874		if (map->guest_addr == addr) {
1875			found = 1;
1876			atomic_dec(&adapter->nr_maps);
1877			list_del(&map->list);
1878			put_page(map->page);
1879			kfree(map);
1880			break;
1881		}
1882	}
1883	up_write(&adapter->maps_lock);
1884
1885	return found ? 0 : -EINVAL;
1886}
1887
1888void kvm_s390_destroy_adapters(struct kvm *kvm)
1889{
1890	int i;
1891	struct s390_map_info *map, *tmp;
1892
1893	for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
1894		if (!kvm->arch.adapters[i])
1895			continue;
1896		list_for_each_entry_safe(map, tmp,
1897					 &kvm->arch.adapters[i]->maps, list) {
1898			list_del(&map->list);
1899			put_page(map->page);
1900			kfree(map);
1901		}
1902		kfree(kvm->arch.adapters[i]);
1903	}
1904}
1905
1906static int modify_io_adapter(struct kvm_device *dev,
1907			     struct kvm_device_attr *attr)
1908{
1909	struct kvm_s390_io_adapter_req req;
1910	struct s390_io_adapter *adapter;
1911	int ret;
1912
1913	if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
1914		return -EFAULT;
1915
1916	adapter = get_io_adapter(dev->kvm, req.id);
1917	if (!adapter)
1918		return -EINVAL;
1919	switch (req.type) {
1920	case KVM_S390_IO_ADAPTER_MASK:
1921		ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
1922		if (ret > 0)
1923			ret = 0;
1924		break;
1925	case KVM_S390_IO_ADAPTER_MAP:
1926		ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
1927		break;
1928	case KVM_S390_IO_ADAPTER_UNMAP:
1929		ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
1930		break;
1931	default:
1932		ret = -EINVAL;
1933	}
1934
1935	return ret;
1936}
1937
1938static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1939{
1940	int r = 0;
1941	unsigned int i;
1942	struct kvm_vcpu *vcpu;
1943
1944	switch (attr->group) {
1945	case KVM_DEV_FLIC_ENQUEUE:
1946		r = enqueue_floating_irq(dev, attr);
1947		break;
1948	case KVM_DEV_FLIC_CLEAR_IRQS:
1949		kvm_s390_clear_float_irqs(dev->kvm);
1950		break;
1951	case KVM_DEV_FLIC_APF_ENABLE:
1952		dev->kvm->arch.gmap->pfault_enabled = 1;
1953		break;
1954	case KVM_DEV_FLIC_APF_DISABLE_WAIT:
1955		dev->kvm->arch.gmap->pfault_enabled = 0;
1956		/*
1957		 * Make sure no async faults are in transition when
1958		 * clearing the queues. So we don't need to worry
1959		 * about late coming workers.
1960		 */
1961		synchronize_srcu(&dev->kvm->srcu);
1962		kvm_for_each_vcpu(i, vcpu, dev->kvm)
1963			kvm_clear_async_pf_completion_queue(vcpu);
1964		break;
1965	case KVM_DEV_FLIC_ADAPTER_REGISTER:
1966		r = register_io_adapter(dev, attr);
1967		break;
1968	case KVM_DEV_FLIC_ADAPTER_MODIFY:
1969		r = modify_io_adapter(dev, attr);
1970		break;
1971	default:
1972		r = -EINVAL;
1973	}
1974
1975	return r;
1976}
1977
1978static int flic_create(struct kvm_device *dev, u32 type)
1979{
1980	if (!dev)
1981		return -EINVAL;
1982	if (dev->kvm->arch.flic)
1983		return -EINVAL;
1984	dev->kvm->arch.flic = dev;
1985	return 0;
1986}
1987
1988static void flic_destroy(struct kvm_device *dev)
1989{
1990	dev->kvm->arch.flic = NULL;
1991	kfree(dev);
1992}
1993
1994/* s390 floating irq controller (flic) */
1995struct kvm_device_ops kvm_flic_ops = {
1996	.name = "kvm-flic",
1997	.get_attr = flic_get_attr,
1998	.set_attr = flic_set_attr,
1999	.create = flic_create,
2000	.destroy = flic_destroy,
2001};
2002
2003static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2004{
2005	unsigned long bit;
2006
2007	bit = bit_nr + (addr % PAGE_SIZE) * 8;
2008
2009	return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2010}
2011
2012static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2013					  u64 addr)
2014{
2015	struct s390_map_info *map;
2016
2017	if (!adapter)
2018		return NULL;
2019
2020	list_for_each_entry(map, &adapter->maps, list) {
2021		if (map->guest_addr == addr)
2022			return map;
2023	}
2024	return NULL;
2025}
2026
2027static int adapter_indicators_set(struct kvm *kvm,
2028				  struct s390_io_adapter *adapter,
2029				  struct kvm_s390_adapter_int *adapter_int)
2030{
2031	unsigned long bit;
2032	int summary_set, idx;
2033	struct s390_map_info *info;
2034	void *map;
2035
2036	info = get_map_info(adapter, adapter_int->ind_addr);
2037	if (!info)
2038		return -1;
2039	map = page_address(info->page);
2040	bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2041	set_bit(bit, map);
2042	idx = srcu_read_lock(&kvm->srcu);
2043	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2044	set_page_dirty_lock(info->page);
2045	info = get_map_info(adapter, adapter_int->summary_addr);
2046	if (!info) {
2047		srcu_read_unlock(&kvm->srcu, idx);
2048		return -1;
2049	}
2050	map = page_address(info->page);
2051	bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2052			  adapter->swap);
2053	summary_set = test_and_set_bit(bit, map);
2054	mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2055	set_page_dirty_lock(info->page);
2056	srcu_read_unlock(&kvm->srcu, idx);
2057	return summary_set ? 0 : 1;
2058}
2059
2060/*
2061 * < 0 - not injected due to error
2062 * = 0 - coalesced, summary indicator already active
2063 * > 0 - injected interrupt
2064 */
2065static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2066			   struct kvm *kvm, int irq_source_id, int level,
2067			   bool line_status)
2068{
2069	int ret;
2070	struct s390_io_adapter *adapter;
2071
2072	/* We're only interested in the 0->1 transition. */
2073	if (!level)
2074		return 0;
2075	adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2076	if (!adapter)
2077		return -1;
2078	down_read(&adapter->maps_lock);
2079	ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2080	up_read(&adapter->maps_lock);
2081	if ((ret > 0) && !adapter->masked) {
2082		struct kvm_s390_interrupt s390int = {
2083			.type = KVM_S390_INT_IO(1, 0, 0, 0),
2084			.parm = 0,
2085			.parm64 = (adapter->isc << 27) | 0x80000000,
2086		};
2087		ret = kvm_s390_inject_vm(kvm, &s390int);
2088		if (ret == 0)
2089			ret = 1;
2090	}
2091	return ret;
2092}
2093
2094int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
2095			  const struct kvm_irq_routing_entry *ue)
2096{
2097	int ret;
2098
2099	switch (ue->type) {
2100	case KVM_IRQ_ROUTING_S390_ADAPTER:
2101		e->set = set_adapter_int;
2102		e->adapter.summary_addr = ue->u.adapter.summary_addr;
2103		e->adapter.ind_addr = ue->u.adapter.ind_addr;
2104		e->adapter.summary_offset = ue->u.adapter.summary_offset;
2105		e->adapter.ind_offset = ue->u.adapter.ind_offset;
2106		e->adapter.adapter_id = ue->u.adapter.adapter_id;
2107		ret = 0;
2108		break;
2109	default:
2110		ret = -EINVAL;
2111	}
2112
2113	return ret;
2114}
2115
2116int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2117		int irq_source_id, int level, bool line_status)
2118{
2119	return -EINVAL;
2120}
2121
2122int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2123{
2124	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2125	struct kvm_s390_irq *buf;
2126	int r = 0;
2127	int n;
2128
2129	buf = vmalloc(len);
2130	if (!buf)
2131		return -ENOMEM;
2132
2133	if (copy_from_user((void *) buf, irqstate, len)) {
2134		r = -EFAULT;
2135		goto out_free;
2136	}
2137
2138	/*
2139	 * Don't allow setting the interrupt state
2140	 * when there are already interrupts pending
2141	 */
2142	spin_lock(&li->lock);
2143	if (li->pending_irqs) {
2144		r = -EBUSY;
2145		goto out_unlock;
2146	}
2147
2148	for (n = 0; n < len / sizeof(*buf); n++) {
2149		r = do_inject_vcpu(vcpu, &buf[n]);
2150		if (r)
2151			break;
2152	}
2153
2154out_unlock:
2155	spin_unlock(&li->lock);
2156out_free:
2157	vfree(buf);
2158
2159	return r;
2160}
2161
2162static void store_local_irq(struct kvm_s390_local_interrupt *li,
2163			    struct kvm_s390_irq *irq,
2164			    unsigned long irq_type)
2165{
2166	switch (irq_type) {
2167	case IRQ_PEND_MCHK_EX:
2168	case IRQ_PEND_MCHK_REP:
2169		irq->type = KVM_S390_MCHK;
2170		irq->u.mchk = li->irq.mchk;
2171		break;
2172	case IRQ_PEND_PROG:
2173		irq->type = KVM_S390_PROGRAM_INT;
2174		irq->u.pgm = li->irq.pgm;
2175		break;
2176	case IRQ_PEND_PFAULT_INIT:
2177		irq->type = KVM_S390_INT_PFAULT_INIT;
2178		irq->u.ext = li->irq.ext;
2179		break;
2180	case IRQ_PEND_EXT_EXTERNAL:
2181		irq->type = KVM_S390_INT_EXTERNAL_CALL;
2182		irq->u.extcall = li->irq.extcall;
2183		break;
2184	case IRQ_PEND_EXT_CLOCK_COMP:
2185		irq->type = KVM_S390_INT_CLOCK_COMP;
2186		break;
2187	case IRQ_PEND_EXT_CPU_TIMER:
2188		irq->type = KVM_S390_INT_CPU_TIMER;
2189		break;
2190	case IRQ_PEND_SIGP_STOP:
2191		irq->type = KVM_S390_SIGP_STOP;
2192		irq->u.stop = li->irq.stop;
2193		break;
2194	case IRQ_PEND_RESTART:
2195		irq->type = KVM_S390_RESTART;
2196		break;
2197	case IRQ_PEND_SET_PREFIX:
2198		irq->type = KVM_S390_SIGP_SET_PREFIX;
2199		irq->u.prefix = li->irq.prefix;
2200		break;
2201	}
2202}
2203
2204int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2205{
2206	uint8_t sigp_ctrl = vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
2207	unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2208	struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2209	unsigned long pending_irqs;
2210	struct kvm_s390_irq irq;
2211	unsigned long irq_type;
2212	int cpuaddr;
2213	int n = 0;
2214
2215	spin_lock(&li->lock);
2216	pending_irqs = li->pending_irqs;
2217	memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2218	       sizeof(sigp_emerg_pending));
2219	spin_unlock(&li->lock);
2220
2221	for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2222		memset(&irq, 0, sizeof(irq));
2223		if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2224			continue;
2225		if (n + sizeof(irq) > len)
2226			return -ENOBUFS;
2227		store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2228		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2229			return -EFAULT;
2230		n += sizeof(irq);
2231	}
2232
2233	if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2234		for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2235			memset(&irq, 0, sizeof(irq));
2236			if (n + sizeof(irq) > len)
2237				return -ENOBUFS;
2238			irq.type = KVM_S390_INT_EMERGENCY;
2239			irq.u.emerg.code = cpuaddr;
2240			if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2241				return -EFAULT;
2242			n += sizeof(irq);
2243		}
2244	}
2245
2246	if ((sigp_ctrl & SIGP_CTRL_C) &&
2247	    (atomic_read(&vcpu->arch.sie_block->cpuflags) &
2248	     CPUSTAT_ECALL_PEND)) {
2249		if (n + sizeof(irq) > len)
2250			return -ENOBUFS;
2251		memset(&irq, 0, sizeof(irq));
2252		irq.type = KVM_S390_INT_EXTERNAL_CALL;
2253		irq.u.extcall.code = sigp_ctrl & SIGP_CTRL_SCN_MASK;
2254		if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2255			return -EFAULT;
2256		n += sizeof(irq);
2257	}
2258
2259	return n;
2260}
2261