2 * handling kvm guest interrupts
4 * Copyright IBM Corp. 2008, 2015
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.
10 * Author(s): Carsten Otte <cotte@de.ibm.com>
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>
23 #include <asm/uaccess.h>
29 #include "trace-s390.h"
31 #define IOINT_SCHID_MASK 0x0000ffff
32 #define IOINT_SSID_MASK 0x00030000
33 #define IOINT_CSSID_MASK 0x03fc0000
34 #define PFAULT_INIT 0x0600
35 #define PFAULT_DONE 0x0680
36 #define VIRTIO_PARAM 0x0d00
38 /* handle external calls via sigp interpretation facility */
39 static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
43 if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
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;
55 struct bsca_block *sca = vcpu->kvm->arch.sca;
56 union bsca_sigp_ctrl sigp_ctrl =
57 sca->cpu[vcpu->vcpu_id].sigp_ctrl;
62 read_unlock(&vcpu->kvm->arch.sca_lock);
70 static int sca_inject_ext_call(struct kvm_vcpu *vcpu, int src_id)
74 read_lock(&vcpu->kvm->arch.sca_lock);
75 if (vcpu->kvm->arch.use_esca) {
76 struct esca_block *sca = vcpu->kvm->arch.sca;
77 union esca_sigp_ctrl *sigp_ctrl =
78 &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
79 union esca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
85 expect = old_val.value;
86 rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
88 struct bsca_block *sca = vcpu->kvm->arch.sca;
89 union bsca_sigp_ctrl *sigp_ctrl =
90 &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
91 union bsca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
97 expect = old_val.value;
98 rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
100 read_unlock(&vcpu->kvm->arch.sca_lock);
103 /* another external call is pending */
106 atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
110 static void sca_clear_ext_call(struct kvm_vcpu *vcpu)
112 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
115 atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
116 read_lock(&vcpu->kvm->arch.sca_lock);
117 if (vcpu->kvm->arch.use_esca) {
118 struct esca_block *sca = vcpu->kvm->arch.sca;
119 union esca_sigp_ctrl *sigp_ctrl =
120 &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
121 union esca_sigp_ctrl old = *sigp_ctrl;
124 rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
126 struct bsca_block *sca = vcpu->kvm->arch.sca;
127 union bsca_sigp_ctrl *sigp_ctrl =
128 &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
129 union bsca_sigp_ctrl old = *sigp_ctrl;
132 rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
134 read_unlock(&vcpu->kvm->arch.sca_lock);
135 WARN_ON(rc != expect); /* cannot clear? */
138 int psw_extint_disabled(struct kvm_vcpu *vcpu)
140 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
143 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
145 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
148 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
150 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
153 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
155 return psw_extint_disabled(vcpu) &&
156 psw_ioint_disabled(vcpu) &&
157 psw_mchk_disabled(vcpu);
160 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
162 if (psw_extint_disabled(vcpu) ||
163 !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
165 if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
166 /* No timer interrupts when single stepping */
171 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
173 if (vcpu->arch.sie_block->ckc >= kvm_s390_get_tod_clock_fast(vcpu->kvm))
175 return ckc_interrupts_enabled(vcpu);
178 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
180 return !psw_extint_disabled(vcpu) &&
181 (vcpu->arch.sie_block->gcr[0] & 0x400ul);
184 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
186 if (!cpu_timer_interrupts_enabled(vcpu))
188 return kvm_s390_get_cpu_timer(vcpu) >> 63;
191 static inline int is_ioirq(unsigned long irq_type)
193 return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
194 (irq_type <= IRQ_PEND_IO_ISC_7));
197 static uint64_t isc_to_isc_bits(int isc)
199 return (0x80 >> isc) << 24;
202 static inline u8 int_word_to_isc(u32 int_word)
204 return (int_word & 0x38000000) >> 27;
207 static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
209 return vcpu->kvm->arch.float_int.pending_irqs |
210 vcpu->arch.local_int.pending_irqs;
213 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
214 unsigned long active_mask)
218 for (i = 0; i <= MAX_ISC; i++)
219 if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
220 active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
225 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
227 unsigned long active_mask;
229 active_mask = pending_irqs(vcpu);
233 if (psw_extint_disabled(vcpu))
234 active_mask &= ~IRQ_PEND_EXT_MASK;
235 if (psw_ioint_disabled(vcpu))
236 active_mask &= ~IRQ_PEND_IO_MASK;
238 active_mask = disable_iscs(vcpu, active_mask);
239 if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
240 __clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
241 if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
242 __clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
243 if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
244 __clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
245 if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
246 __clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
247 if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
248 __clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
249 if (psw_mchk_disabled(vcpu))
250 active_mask &= ~IRQ_PEND_MCHK_MASK;
251 if (!(vcpu->arch.sie_block->gcr[14] &
252 vcpu->kvm->arch.float_int.mchk.cr14))
253 __clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
256 * STOP irqs will never be actively delivered. They are triggered via
257 * intercept requests and cleared when the stop intercept is performed.
259 __clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
264 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
266 atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
267 set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
270 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
272 atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
273 clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
276 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
278 atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
279 &vcpu->arch.sie_block->cpuflags);
280 vcpu->arch.sie_block->lctl = 0x0000;
281 vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
283 if (guestdbg_enabled(vcpu)) {
284 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
285 LCTL_CR10 | LCTL_CR11);
286 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
290 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
292 atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
295 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
297 if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
299 else if (psw_ioint_disabled(vcpu))
300 __set_cpuflag(vcpu, CPUSTAT_IO_INT);
302 vcpu->arch.sie_block->lctl |= LCTL_CR6;
305 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
307 if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
309 if (psw_extint_disabled(vcpu))
310 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
312 vcpu->arch.sie_block->lctl |= LCTL_CR0;
315 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
317 if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
319 if (psw_mchk_disabled(vcpu))
320 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
322 vcpu->arch.sie_block->lctl |= LCTL_CR14;
325 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
327 if (kvm_s390_is_stop_irq_pending(vcpu))
328 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
331 /* Set interception request for non-deliverable interrupts */
332 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
334 set_intercept_indicators_io(vcpu);
335 set_intercept_indicators_ext(vcpu);
336 set_intercept_indicators_mchk(vcpu);
337 set_intercept_indicators_stop(vcpu);
340 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
342 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
345 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
348 rc = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
349 (u16 *)__LC_EXT_INT_CODE);
350 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
351 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
352 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
353 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
354 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
355 clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
356 return rc ? -EFAULT : 0;
359 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
361 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
364 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
367 rc = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
368 (u16 __user *)__LC_EXT_INT_CODE);
369 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
370 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
371 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
372 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
373 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
374 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
375 return rc ? -EFAULT : 0;
378 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
380 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
381 struct kvm_s390_ext_info ext;
384 spin_lock(&li->lock);
386 clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
387 li->irq.ext.ext_params2 = 0;
388 spin_unlock(&li->lock);
390 VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
392 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
393 KVM_S390_INT_PFAULT_INIT,
396 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
397 rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
398 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
399 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
400 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
401 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
402 rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
403 return rc ? -EFAULT : 0;
406 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
408 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
409 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
410 struct kvm_s390_mchk_info mchk = {};
411 unsigned long adtl_status_addr;
415 spin_lock(&fi->lock);
416 spin_lock(&li->lock);
417 if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
418 test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
420 * If there was an exigent machine check pending, then any
421 * repressible machine checks that might have been pending
422 * are indicated along with it, so always clear bits for
423 * repressible and exigent interrupts
426 clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
427 clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
428 memset(&li->irq.mchk, 0, sizeof(mchk));
432 * We indicate floating repressible conditions along with
433 * other pending conditions. Channel Report Pending and Channel
434 * Subsystem damage are the only two and and are indicated by
435 * bits in mcic and masked in cr14.
437 if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
438 mchk.mcic |= fi->mchk.mcic;
439 mchk.cr14 |= fi->mchk.cr14;
440 memset(&fi->mchk, 0, sizeof(mchk));
443 spin_unlock(&li->lock);
444 spin_unlock(&fi->lock);
447 VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
449 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
451 mchk.cr14, mchk.mcic);
453 rc = kvm_s390_vcpu_store_status(vcpu,
454 KVM_S390_STORE_STATUS_PREFIXED);
455 rc |= read_guest_lc(vcpu, __LC_VX_SAVE_AREA_ADDR,
457 sizeof(unsigned long));
458 rc |= kvm_s390_vcpu_store_adtl_status(vcpu,
460 rc |= put_guest_lc(vcpu, mchk.mcic,
461 (u64 __user *) __LC_MCCK_CODE);
462 rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
463 (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
464 rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
466 sizeof(mchk.fixed_logout));
467 rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
468 &vcpu->arch.sie_block->gpsw,
470 rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
471 &vcpu->arch.sie_block->gpsw,
474 return rc ? -EFAULT : 0;
477 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
479 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
482 VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
483 vcpu->stat.deliver_restart_signal++;
484 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
486 rc = write_guest_lc(vcpu,
487 offsetof(struct lowcore, restart_old_psw),
488 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
489 rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
490 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
491 clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
492 return rc ? -EFAULT : 0;
495 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
497 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
498 struct kvm_s390_prefix_info prefix;
500 spin_lock(&li->lock);
501 prefix = li->irq.prefix;
502 li->irq.prefix.address = 0;
503 clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
504 spin_unlock(&li->lock);
506 vcpu->stat.deliver_prefix_signal++;
507 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
508 KVM_S390_SIGP_SET_PREFIX,
511 kvm_s390_set_prefix(vcpu, prefix.address);
515 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
517 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
521 spin_lock(&li->lock);
522 cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
523 clear_bit(cpu_addr, li->sigp_emerg_pending);
524 if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
525 clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
526 spin_unlock(&li->lock);
528 VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
529 vcpu->stat.deliver_emergency_signal++;
530 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
533 rc = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
534 (u16 *)__LC_EXT_INT_CODE);
535 rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
536 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
537 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
538 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
539 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
540 return rc ? -EFAULT : 0;
543 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
545 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
546 struct kvm_s390_extcall_info extcall;
549 spin_lock(&li->lock);
550 extcall = li->irq.extcall;
551 li->irq.extcall.code = 0;
552 clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
553 spin_unlock(&li->lock);
555 VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
556 vcpu->stat.deliver_external_call++;
557 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
558 KVM_S390_INT_EXTERNAL_CALL,
561 rc = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
562 (u16 *)__LC_EXT_INT_CODE);
563 rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
564 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
565 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
566 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
568 return rc ? -EFAULT : 0;
571 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
573 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
574 struct kvm_s390_pgm_info pgm_info;
575 int rc = 0, nullifying = false;
578 spin_lock(&li->lock);
579 pgm_info = li->irq.pgm;
580 clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
581 memset(&li->irq.pgm, 0, sizeof(pgm_info));
582 spin_unlock(&li->lock);
584 ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
585 VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
586 pgm_info.code, ilen);
587 vcpu->stat.deliver_program_int++;
588 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
591 switch (pgm_info.code & ~PGM_PER) {
592 case PGM_AFX_TRANSLATION:
593 case PGM_ASX_TRANSLATION:
594 case PGM_EX_TRANSLATION:
595 case PGM_LFX_TRANSLATION:
596 case PGM_LSTE_SEQUENCE:
597 case PGM_LSX_TRANSLATION:
598 case PGM_LX_TRANSLATION:
599 case PGM_PRIMARY_AUTHORITY:
600 case PGM_SECONDARY_AUTHORITY:
603 case PGM_SPACE_SWITCH:
604 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
605 (u64 *)__LC_TRANS_EXC_CODE);
607 case PGM_ALEN_TRANSLATION:
608 case PGM_ALE_SEQUENCE:
609 case PGM_ASTE_INSTANCE:
610 case PGM_ASTE_SEQUENCE:
611 case PGM_ASTE_VALIDITY:
612 case PGM_EXTENDED_AUTHORITY:
613 rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
614 (u8 *)__LC_EXC_ACCESS_ID);
618 case PGM_PAGE_TRANSLATION:
619 case PGM_REGION_FIRST_TRANS:
620 case PGM_REGION_SECOND_TRANS:
621 case PGM_REGION_THIRD_TRANS:
622 case PGM_SEGMENT_TRANSLATION:
623 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
624 (u64 *)__LC_TRANS_EXC_CODE);
625 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
626 (u8 *)__LC_EXC_ACCESS_ID);
627 rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
628 (u8 *)__LC_OP_ACCESS_ID);
632 rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
633 (u16 *)__LC_MON_CLASS_NR);
634 rc |= put_guest_lc(vcpu, pgm_info.mon_code,
635 (u64 *)__LC_MON_CODE);
637 case PGM_VECTOR_PROCESSING:
639 rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
640 (u32 *)__LC_DATA_EXC_CODE);
643 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
644 (u64 *)__LC_TRANS_EXC_CODE);
645 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
646 (u8 *)__LC_EXC_ACCESS_ID);
649 case PGM_STACK_EMPTY:
650 case PGM_STACK_SPECIFICATION:
652 case PGM_STACK_OPERATION:
653 case PGM_TRACE_TABEL:
654 case PGM_CRYPTO_OPERATION:
659 if (pgm_info.code & PGM_PER) {
660 rc |= put_guest_lc(vcpu, pgm_info.per_code,
661 (u8 *) __LC_PER_CODE);
662 rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
663 (u8 *)__LC_PER_ATMID);
664 rc |= put_guest_lc(vcpu, pgm_info.per_address,
665 (u64 *) __LC_PER_ADDRESS);
666 rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
667 (u8 *) __LC_PER_ACCESS_ID);
670 if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
671 kvm_s390_rewind_psw(vcpu, ilen);
673 /* bit 1+2 of the target are the ilc, so we can directly use ilen */
674 rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
675 rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
676 (u64 *) __LC_LAST_BREAK);
677 rc |= put_guest_lc(vcpu, pgm_info.code,
678 (u16 *)__LC_PGM_INT_CODE);
679 rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
680 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
681 rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
682 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
683 return rc ? -EFAULT : 0;
686 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
688 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
689 struct kvm_s390_ext_info ext;
692 spin_lock(&fi->lock);
693 if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
694 spin_unlock(&fi->lock);
697 ext = fi->srv_signal;
698 memset(&fi->srv_signal, 0, sizeof(ext));
699 clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
700 spin_unlock(&fi->lock);
702 VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
704 vcpu->stat.deliver_service_signal++;
705 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
708 rc = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
709 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
710 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
711 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
712 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
713 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
714 rc |= put_guest_lc(vcpu, ext.ext_params,
715 (u32 *)__LC_EXT_PARAMS);
717 return rc ? -EFAULT : 0;
720 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
722 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
723 struct kvm_s390_interrupt_info *inti;
726 spin_lock(&fi->lock);
727 inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
728 struct kvm_s390_interrupt_info,
731 list_del(&inti->list);
732 fi->counters[FIRQ_CNTR_PFAULT] -= 1;
734 if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
735 clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
736 spin_unlock(&fi->lock);
739 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
740 KVM_S390_INT_PFAULT_DONE, 0,
741 inti->ext.ext_params2);
742 VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
743 inti->ext.ext_params2);
745 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
746 (u16 *)__LC_EXT_INT_CODE);
747 rc |= put_guest_lc(vcpu, PFAULT_DONE,
748 (u16 *)__LC_EXT_CPU_ADDR);
749 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
750 &vcpu->arch.sie_block->gpsw,
752 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
753 &vcpu->arch.sie_block->gpsw,
755 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
756 (u64 *)__LC_EXT_PARAMS2);
759 return rc ? -EFAULT : 0;
762 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
764 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
765 struct kvm_s390_interrupt_info *inti;
768 spin_lock(&fi->lock);
769 inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
770 struct kvm_s390_interrupt_info,
774 "deliver: virtio parm: 0x%x,parm64: 0x%llx",
775 inti->ext.ext_params, inti->ext.ext_params2);
776 vcpu->stat.deliver_virtio_interrupt++;
777 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
779 inti->ext.ext_params,
780 inti->ext.ext_params2);
781 list_del(&inti->list);
782 fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
784 if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
785 clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
786 spin_unlock(&fi->lock);
789 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
790 (u16 *)__LC_EXT_INT_CODE);
791 rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
792 (u16 *)__LC_EXT_CPU_ADDR);
793 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
794 &vcpu->arch.sie_block->gpsw,
796 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
797 &vcpu->arch.sie_block->gpsw,
799 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
800 (u32 *)__LC_EXT_PARAMS);
801 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
802 (u64 *)__LC_EXT_PARAMS2);
805 return rc ? -EFAULT : 0;
808 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
809 unsigned long irq_type)
811 struct list_head *isc_list;
812 struct kvm_s390_float_interrupt *fi;
813 struct kvm_s390_interrupt_info *inti = NULL;
816 fi = &vcpu->kvm->arch.float_int;
818 spin_lock(&fi->lock);
819 isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
820 inti = list_first_entry_or_null(isc_list,
821 struct kvm_s390_interrupt_info,
824 VCPU_EVENT(vcpu, 4, "deliver: I/O 0x%llx", inti->type);
825 vcpu->stat.deliver_io_int++;
826 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
828 ((__u32)inti->io.subchannel_id << 16) |
829 inti->io.subchannel_nr,
830 ((__u64)inti->io.io_int_parm << 32) |
831 inti->io.io_int_word);
832 list_del(&inti->list);
833 fi->counters[FIRQ_CNTR_IO] -= 1;
835 if (list_empty(isc_list))
836 clear_bit(irq_type, &fi->pending_irqs);
837 spin_unlock(&fi->lock);
840 rc = put_guest_lc(vcpu, inti->io.subchannel_id,
841 (u16 *)__LC_SUBCHANNEL_ID);
842 rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
843 (u16 *)__LC_SUBCHANNEL_NR);
844 rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
845 (u32 *)__LC_IO_INT_PARM);
846 rc |= put_guest_lc(vcpu, inti->io.io_int_word,
847 (u32 *)__LC_IO_INT_WORD);
848 rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
849 &vcpu->arch.sie_block->gpsw,
851 rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
852 &vcpu->arch.sie_block->gpsw,
857 return rc ? -EFAULT : 0;
860 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
862 static const deliver_irq_t deliver_irq_funcs[] = {
863 [IRQ_PEND_MCHK_EX] = __deliver_machine_check,
864 [IRQ_PEND_MCHK_REP] = __deliver_machine_check,
865 [IRQ_PEND_PROG] = __deliver_prog,
866 [IRQ_PEND_EXT_EMERGENCY] = __deliver_emergency_signal,
867 [IRQ_PEND_EXT_EXTERNAL] = __deliver_external_call,
868 [IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
869 [IRQ_PEND_EXT_CPU_TIMER] = __deliver_cpu_timer,
870 [IRQ_PEND_RESTART] = __deliver_restart,
871 [IRQ_PEND_SET_PREFIX] = __deliver_set_prefix,
872 [IRQ_PEND_PFAULT_INIT] = __deliver_pfault_init,
873 [IRQ_PEND_EXT_SERVICE] = __deliver_service,
874 [IRQ_PEND_PFAULT_DONE] = __deliver_pfault_done,
875 [IRQ_PEND_VIRTIO] = __deliver_virtio,
878 /* Check whether an external call is pending (deliverable or not) */
879 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
881 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
883 if (!sclp.has_sigpif)
884 return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
886 return sca_ext_call_pending(vcpu, NULL);
889 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
891 if (deliverable_irqs(vcpu))
894 if (kvm_cpu_has_pending_timer(vcpu))
897 /* external call pending and deliverable */
898 if (kvm_s390_ext_call_pending(vcpu) &&
899 !psw_extint_disabled(vcpu) &&
900 (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
903 if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
908 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
910 return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
913 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
915 u64 now, cputm, sltime = 0;
917 if (ckc_interrupts_enabled(vcpu)) {
918 now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
919 sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
920 /* already expired or overflow? */
921 if (!sltime || vcpu->arch.sie_block->ckc <= now)
923 if (cpu_timer_interrupts_enabled(vcpu)) {
924 cputm = kvm_s390_get_cpu_timer(vcpu);
925 /* already expired? */
928 return min(sltime, tod_to_ns(cputm));
930 } else if (cpu_timer_interrupts_enabled(vcpu)) {
931 sltime = kvm_s390_get_cpu_timer(vcpu);
932 /* already expired? */
939 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
943 vcpu->stat.exit_wait_state++;
946 if (kvm_arch_vcpu_runnable(vcpu))
949 if (psw_interrupts_disabled(vcpu)) {
950 VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
951 return -EOPNOTSUPP; /* disabled wait */
954 if (!ckc_interrupts_enabled(vcpu) &&
955 !cpu_timer_interrupts_enabled(vcpu)) {
956 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
957 __set_cpu_idle(vcpu);
961 sltime = __calculate_sltime(vcpu);
965 __set_cpu_idle(vcpu);
966 hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
967 VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
969 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
970 kvm_vcpu_block(vcpu);
971 __unset_cpu_idle(vcpu);
972 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
974 hrtimer_cancel(&vcpu->arch.ckc_timer);
978 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
981 * We cannot move this into the if, as the CPU might be already
982 * in kvm_vcpu_block without having the waitqueue set (polling)
984 vcpu->valid_wakeup = true;
985 if (swait_active(&vcpu->wq)) {
987 * The vcpu gave up the cpu voluntarily, mark it as a good
990 vcpu->preempted = true;
992 vcpu->stat.halt_wakeup++;
996 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
998 struct kvm_vcpu *vcpu;
1001 vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1002 sltime = __calculate_sltime(vcpu);
1005 * If the monotonic clock runs faster than the tod clock we might be
1006 * woken up too early and have to go back to sleep to avoid deadlocks.
1008 if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1009 return HRTIMER_RESTART;
1010 kvm_s390_vcpu_wakeup(vcpu);
1011 return HRTIMER_NORESTART;
1014 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1016 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1018 spin_lock(&li->lock);
1019 li->pending_irqs = 0;
1020 bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1021 memset(&li->irq, 0, sizeof(li->irq));
1022 spin_unlock(&li->lock);
1024 sca_clear_ext_call(vcpu);
1027 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1029 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1032 unsigned long irq_type;
1035 __reset_intercept_indicators(vcpu);
1037 /* pending ckc conditions might have been invalidated */
1038 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1039 if (ckc_irq_pending(vcpu))
1040 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1042 /* pending cpu timer conditions might have been invalidated */
1043 clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1044 if (cpu_timer_irq_pending(vcpu))
1045 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1047 while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1048 /* bits are in the order of interrupt priority */
1049 irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1050 if (is_ioirq(irq_type)) {
1051 rc = __deliver_io(vcpu, irq_type);
1053 func = deliver_irq_funcs[irq_type];
1055 WARN_ON_ONCE(func == NULL);
1056 clear_bit(irq_type, &li->pending_irqs);
1063 set_intercept_indicators(vcpu);
1068 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1070 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1072 VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1073 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1074 irq->u.pgm.code, 0);
1076 if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1077 /* auto detection if no valid ILC was given */
1078 irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1079 irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1080 irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1083 if (irq->u.pgm.code == PGM_PER) {
1084 li->irq.pgm.code |= PGM_PER;
1085 li->irq.pgm.flags = irq->u.pgm.flags;
1086 /* only modify PER related information */
1087 li->irq.pgm.per_address = irq->u.pgm.per_address;
1088 li->irq.pgm.per_code = irq->u.pgm.per_code;
1089 li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1090 li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1091 } else if (!(irq->u.pgm.code & PGM_PER)) {
1092 li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1094 li->irq.pgm.flags = irq->u.pgm.flags;
1095 /* only modify non-PER information */
1096 li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1097 li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1098 li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1099 li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1100 li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1101 li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1103 li->irq.pgm = irq->u.pgm;
1105 set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1109 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1111 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1113 VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1114 irq->u.ext.ext_params2);
1115 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1116 irq->u.ext.ext_params,
1117 irq->u.ext.ext_params2);
1119 li->irq.ext = irq->u.ext;
1120 set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1121 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1125 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1127 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1128 struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1129 uint16_t src_id = irq->u.extcall.code;
1131 VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1133 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1136 /* sending vcpu invalid */
1137 if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1140 if (sclp.has_sigpif)
1141 return sca_inject_ext_call(vcpu, src_id);
1143 if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1145 *extcall = irq->u.extcall;
1146 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1150 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1152 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1153 struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1155 VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1156 irq->u.prefix.address);
1157 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1158 irq->u.prefix.address, 0);
1160 if (!is_vcpu_stopped(vcpu))
1163 *prefix = irq->u.prefix;
1164 set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1168 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1169 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1171 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1172 struct kvm_s390_stop_info *stop = &li->irq.stop;
1175 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1177 if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1180 if (is_vcpu_stopped(vcpu)) {
1181 if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1182 rc = kvm_s390_store_status_unloaded(vcpu,
1183 KVM_S390_STORE_STATUS_NOADDR);
1187 if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1189 stop->flags = irq->u.stop.flags;
1190 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1194 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1195 struct kvm_s390_irq *irq)
1197 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1199 VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1200 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1202 set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1206 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1207 struct kvm_s390_irq *irq)
1209 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1211 VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1213 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1214 irq->u.emerg.code, 0);
1216 /* sending vcpu invalid */
1217 if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1220 set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1221 set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1222 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1226 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1228 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1229 struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1231 VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1233 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1237 * Because repressible machine checks can be indicated along with
1238 * exigent machine checks (PoP, Chapter 11, Interruption action)
1239 * we need to combine cr14, mcic and external damage code.
1240 * Failing storage address and the logout area should not be or'ed
1241 * together, we just indicate the last occurrence of the corresponding
1244 mchk->cr14 |= irq->u.mchk.cr14;
1245 mchk->mcic |= irq->u.mchk.mcic;
1246 mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1247 mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1248 memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1249 sizeof(mchk->fixed_logout));
1250 if (mchk->mcic & MCHK_EX_MASK)
1251 set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1252 else if (mchk->mcic & MCHK_REP_MASK)
1253 set_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
1257 static int __inject_ckc(struct kvm_vcpu *vcpu)
1259 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1261 VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1262 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1265 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1266 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1270 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1272 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1274 VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1275 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1278 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1279 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1283 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1286 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1287 struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1288 struct kvm_s390_interrupt_info *iter;
1289 u16 id = (schid & 0xffff0000U) >> 16;
1290 u16 nr = schid & 0x0000ffffU;
1292 spin_lock(&fi->lock);
1293 list_for_each_entry(iter, isc_list, list) {
1294 if (schid && (id != iter->io.subchannel_id ||
1295 nr != iter->io.subchannel_nr))
1297 /* found an appropriate entry */
1298 list_del_init(&iter->list);
1299 fi->counters[FIRQ_CNTR_IO] -= 1;
1300 if (list_empty(isc_list))
1301 clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1302 spin_unlock(&fi->lock);
1305 spin_unlock(&fi->lock);
1310 * Dequeue and return an I/O interrupt matching any of the interruption
1311 * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1313 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1314 u64 isc_mask, u32 schid)
1316 struct kvm_s390_interrupt_info *inti = NULL;
1319 for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1320 if (isc_mask & isc_to_isc_bits(isc))
1321 inti = get_io_int(kvm, isc, schid);
1326 #define SCCB_MASK 0xFFFFFFF8
1327 #define SCCB_EVENT_PENDING 0x3
1329 static int __inject_service(struct kvm *kvm,
1330 struct kvm_s390_interrupt_info *inti)
1332 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1334 spin_lock(&fi->lock);
1335 fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1337 * Early versions of the QEMU s390 bios will inject several
1338 * service interrupts after another without handling a
1339 * condition code indicating busy.
1340 * We will silently ignore those superfluous sccb values.
1341 * A future version of QEMU will take care of serialization
1344 if (fi->srv_signal.ext_params & SCCB_MASK)
1346 fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1347 set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1349 spin_unlock(&fi->lock);
1354 static int __inject_virtio(struct kvm *kvm,
1355 struct kvm_s390_interrupt_info *inti)
1357 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1359 spin_lock(&fi->lock);
1360 if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1361 spin_unlock(&fi->lock);
1364 fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1365 list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1366 set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1367 spin_unlock(&fi->lock);
1371 static int __inject_pfault_done(struct kvm *kvm,
1372 struct kvm_s390_interrupt_info *inti)
1374 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1376 spin_lock(&fi->lock);
1377 if (fi->counters[FIRQ_CNTR_PFAULT] >=
1378 (ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1379 spin_unlock(&fi->lock);
1382 fi->counters[FIRQ_CNTR_PFAULT] += 1;
1383 list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1384 set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1385 spin_unlock(&fi->lock);
1389 #define CR_PENDING_SUBCLASS 28
1390 static int __inject_float_mchk(struct kvm *kvm,
1391 struct kvm_s390_interrupt_info *inti)
1393 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1395 spin_lock(&fi->lock);
1396 fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1397 fi->mchk.mcic |= inti->mchk.mcic;
1398 set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1399 spin_unlock(&fi->lock);
1404 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1406 struct kvm_s390_float_interrupt *fi;
1407 struct list_head *list;
1410 fi = &kvm->arch.float_int;
1411 spin_lock(&fi->lock);
1412 if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1413 spin_unlock(&fi->lock);
1416 fi->counters[FIRQ_CNTR_IO] += 1;
1418 isc = int_word_to_isc(inti->io.io_int_word);
1419 list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1420 list_add_tail(&inti->list, list);
1421 set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1422 spin_unlock(&fi->lock);
1427 * Find a destination VCPU for a floating irq and kick it.
1429 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1431 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1432 struct kvm_s390_local_interrupt *li;
1433 struct kvm_vcpu *dst_vcpu;
1434 int sigcpu, online_vcpus, nr_tries = 0;
1436 online_vcpus = atomic_read(&kvm->online_vcpus);
1440 /* find idle VCPUs first, then round robin */
1441 sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1442 if (sigcpu == online_vcpus) {
1444 sigcpu = fi->next_rr_cpu;
1445 fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1446 /* avoid endless loops if all vcpus are stopped */
1447 if (nr_tries++ >= online_vcpus)
1449 } while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1451 dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1453 /* make the VCPU drop out of the SIE, or wake it up if sleeping */
1454 li = &dst_vcpu->arch.local_int;
1455 spin_lock(&li->lock);
1458 atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1460 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1461 atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1464 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1467 spin_unlock(&li->lock);
1468 kvm_s390_vcpu_wakeup(dst_vcpu);
1471 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1473 u64 type = READ_ONCE(inti->type);
1478 rc = __inject_float_mchk(kvm, inti);
1480 case KVM_S390_INT_VIRTIO:
1481 rc = __inject_virtio(kvm, inti);
1483 case KVM_S390_INT_SERVICE:
1484 rc = __inject_service(kvm, inti);
1486 case KVM_S390_INT_PFAULT_DONE:
1487 rc = __inject_pfault_done(kvm, inti);
1489 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1490 rc = __inject_io(kvm, inti);
1498 __floating_irq_kick(kvm, type);
1502 int kvm_s390_inject_vm(struct kvm *kvm,
1503 struct kvm_s390_interrupt *s390int)
1505 struct kvm_s390_interrupt_info *inti;
1508 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1512 inti->type = s390int->type;
1513 switch (inti->type) {
1514 case KVM_S390_INT_VIRTIO:
1515 VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1516 s390int->parm, s390int->parm64);
1517 inti->ext.ext_params = s390int->parm;
1518 inti->ext.ext_params2 = s390int->parm64;
1520 case KVM_S390_INT_SERVICE:
1521 VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1522 inti->ext.ext_params = s390int->parm;
1524 case KVM_S390_INT_PFAULT_DONE:
1525 inti->ext.ext_params2 = s390int->parm64;
1528 VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1530 inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1531 inti->mchk.mcic = s390int->parm64;
1533 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1534 if (inti->type & KVM_S390_INT_IO_AI_MASK)
1535 VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
1537 VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
1538 s390int->type & IOINT_CSSID_MASK,
1539 s390int->type & IOINT_SSID_MASK,
1540 s390int->type & IOINT_SCHID_MASK);
1541 inti->io.subchannel_id = s390int->parm >> 16;
1542 inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1543 inti->io.io_int_parm = s390int->parm64 >> 32;
1544 inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1550 trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1553 rc = __inject_vm(kvm, inti);
1559 int kvm_s390_reinject_io_int(struct kvm *kvm,
1560 struct kvm_s390_interrupt_info *inti)
1562 return __inject_vm(kvm, inti);
1565 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1566 struct kvm_s390_irq *irq)
1568 irq->type = s390int->type;
1569 switch (irq->type) {
1570 case KVM_S390_PROGRAM_INT:
1571 if (s390int->parm & 0xffff0000)
1573 irq->u.pgm.code = s390int->parm;
1575 case KVM_S390_SIGP_SET_PREFIX:
1576 irq->u.prefix.address = s390int->parm;
1578 case KVM_S390_SIGP_STOP:
1579 irq->u.stop.flags = s390int->parm;
1581 case KVM_S390_INT_EXTERNAL_CALL:
1582 if (s390int->parm & 0xffff0000)
1584 irq->u.extcall.code = s390int->parm;
1586 case KVM_S390_INT_EMERGENCY:
1587 if (s390int->parm & 0xffff0000)
1589 irq->u.emerg.code = s390int->parm;
1592 irq->u.mchk.mcic = s390int->parm64;
1598 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1600 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1602 return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1605 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1607 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1609 spin_lock(&li->lock);
1610 li->irq.stop.flags = 0;
1611 clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1612 spin_unlock(&li->lock);
1615 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1619 switch (irq->type) {
1620 case KVM_S390_PROGRAM_INT:
1621 rc = __inject_prog(vcpu, irq);
1623 case KVM_S390_SIGP_SET_PREFIX:
1624 rc = __inject_set_prefix(vcpu, irq);
1626 case KVM_S390_SIGP_STOP:
1627 rc = __inject_sigp_stop(vcpu, irq);
1629 case KVM_S390_RESTART:
1630 rc = __inject_sigp_restart(vcpu, irq);
1632 case KVM_S390_INT_CLOCK_COMP:
1633 rc = __inject_ckc(vcpu);
1635 case KVM_S390_INT_CPU_TIMER:
1636 rc = __inject_cpu_timer(vcpu);
1638 case KVM_S390_INT_EXTERNAL_CALL:
1639 rc = __inject_extcall(vcpu, irq);
1641 case KVM_S390_INT_EMERGENCY:
1642 rc = __inject_sigp_emergency(vcpu, irq);
1645 rc = __inject_mchk(vcpu, irq);
1647 case KVM_S390_INT_PFAULT_INIT:
1648 rc = __inject_pfault_init(vcpu, irq);
1650 case KVM_S390_INT_VIRTIO:
1651 case KVM_S390_INT_SERVICE:
1652 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1660 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1662 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1665 spin_lock(&li->lock);
1666 rc = do_inject_vcpu(vcpu, irq);
1667 spin_unlock(&li->lock);
1669 kvm_s390_vcpu_wakeup(vcpu);
1673 static inline void clear_irq_list(struct list_head *_list)
1675 struct kvm_s390_interrupt_info *inti, *n;
1677 list_for_each_entry_safe(inti, n, _list, list) {
1678 list_del(&inti->list);
1683 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1684 struct kvm_s390_irq *irq)
1686 irq->type = inti->type;
1687 switch (inti->type) {
1688 case KVM_S390_INT_PFAULT_INIT:
1689 case KVM_S390_INT_PFAULT_DONE:
1690 case KVM_S390_INT_VIRTIO:
1691 irq->u.ext = inti->ext;
1693 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1694 irq->u.io = inti->io;
1699 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1701 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1704 spin_lock(&fi->lock);
1705 fi->pending_irqs = 0;
1706 memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1707 memset(&fi->mchk, 0, sizeof(fi->mchk));
1708 for (i = 0; i < FIRQ_LIST_COUNT; i++)
1709 clear_irq_list(&fi->lists[i]);
1710 for (i = 0; i < FIRQ_MAX_COUNT; i++)
1711 fi->counters[i] = 0;
1712 spin_unlock(&fi->lock);
1715 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1717 struct kvm_s390_interrupt_info *inti;
1718 struct kvm_s390_float_interrupt *fi;
1719 struct kvm_s390_irq *buf;
1720 struct kvm_s390_irq *irq;
1726 if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1730 * We are already using -ENOMEM to signal
1731 * userspace it may retry with a bigger buffer,
1732 * so we need to use something else for this case
1738 max_irqs = len / sizeof(struct kvm_s390_irq);
1740 fi = &kvm->arch.float_int;
1741 spin_lock(&fi->lock);
1742 for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1743 list_for_each_entry(inti, &fi->lists[i], list) {
1744 if (n == max_irqs) {
1745 /* signal userspace to try again */
1749 inti_to_irq(inti, &buf[n]);
1753 if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1754 if (n == max_irqs) {
1755 /* signal userspace to try again */
1759 irq = (struct kvm_s390_irq *) &buf[n];
1760 irq->type = KVM_S390_INT_SERVICE;
1761 irq->u.ext = fi->srv_signal;
1764 if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1765 if (n == max_irqs) {
1766 /* signal userspace to try again */
1770 irq = (struct kvm_s390_irq *) &buf[n];
1771 irq->type = KVM_S390_MCHK;
1772 irq->u.mchk = fi->mchk;
1777 spin_unlock(&fi->lock);
1778 if (!ret && n > 0) {
1779 if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1784 return ret < 0 ? ret : n;
1787 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1791 switch (attr->group) {
1792 case KVM_DEV_FLIC_GET_ALL_IRQS:
1793 r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1803 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1806 struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1807 void *target = NULL;
1808 void __user *source;
1811 if (get_user(inti->type, (u64 __user *)addr))
1814 switch (inti->type) {
1815 case KVM_S390_INT_PFAULT_INIT:
1816 case KVM_S390_INT_PFAULT_DONE:
1817 case KVM_S390_INT_VIRTIO:
1818 case KVM_S390_INT_SERVICE:
1819 target = (void *) &inti->ext;
1820 source = &uptr->u.ext;
1821 size = sizeof(inti->ext);
1823 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1824 target = (void *) &inti->io;
1825 source = &uptr->u.io;
1826 size = sizeof(inti->io);
1829 target = (void *) &inti->mchk;
1830 source = &uptr->u.mchk;
1831 size = sizeof(inti->mchk);
1837 if (copy_from_user(target, source, size))
1843 static int enqueue_floating_irq(struct kvm_device *dev,
1844 struct kvm_device_attr *attr)
1846 struct kvm_s390_interrupt_info *inti = NULL;
1848 int len = attr->attr;
1850 if (len % sizeof(struct kvm_s390_irq) != 0)
1852 else if (len > KVM_S390_FLIC_MAX_BUFFER)
1855 while (len >= sizeof(struct kvm_s390_irq)) {
1856 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1860 r = copy_irq_from_user(inti, attr->addr);
1865 r = __inject_vm(dev->kvm, inti);
1870 len -= sizeof(struct kvm_s390_irq);
1871 attr->addr += sizeof(struct kvm_s390_irq);
1877 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1879 if (id >= MAX_S390_IO_ADAPTERS)
1881 return kvm->arch.adapters[id];
1884 static int register_io_adapter(struct kvm_device *dev,
1885 struct kvm_device_attr *attr)
1887 struct s390_io_adapter *adapter;
1888 struct kvm_s390_io_adapter adapter_info;
1890 if (copy_from_user(&adapter_info,
1891 (void __user *)attr->addr, sizeof(adapter_info)))
1894 if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1895 (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1898 adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1902 INIT_LIST_HEAD(&adapter->maps);
1903 init_rwsem(&adapter->maps_lock);
1904 atomic_set(&adapter->nr_maps, 0);
1905 adapter->id = adapter_info.id;
1906 adapter->isc = adapter_info.isc;
1907 adapter->maskable = adapter_info.maskable;
1908 adapter->masked = false;
1909 adapter->swap = adapter_info.swap;
1910 dev->kvm->arch.adapters[adapter->id] = adapter;
1915 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1918 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1920 if (!adapter || !adapter->maskable)
1922 ret = adapter->masked;
1923 adapter->masked = masked;
1927 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1929 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1930 struct s390_map_info *map;
1933 if (!adapter || !addr)
1936 map = kzalloc(sizeof(*map), GFP_KERNEL);
1941 INIT_LIST_HEAD(&map->list);
1942 map->guest_addr = addr;
1943 map->addr = gmap_translate(kvm->arch.gmap, addr);
1944 if (map->addr == -EFAULT) {
1948 ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1952 down_write(&adapter->maps_lock);
1953 if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1954 list_add_tail(&map->list, &adapter->maps);
1957 put_page(map->page);
1960 up_write(&adapter->maps_lock);
1967 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1969 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1970 struct s390_map_info *map, *tmp;
1973 if (!adapter || !addr)
1976 down_write(&adapter->maps_lock);
1977 list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1978 if (map->guest_addr == addr) {
1980 atomic_dec(&adapter->nr_maps);
1981 list_del(&map->list);
1982 put_page(map->page);
1987 up_write(&adapter->maps_lock);
1989 return found ? 0 : -EINVAL;
1992 void kvm_s390_destroy_adapters(struct kvm *kvm)
1995 struct s390_map_info *map, *tmp;
1997 for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
1998 if (!kvm->arch.adapters[i])
2000 list_for_each_entry_safe(map, tmp,
2001 &kvm->arch.adapters[i]->maps, list) {
2002 list_del(&map->list);
2003 put_page(map->page);
2006 kfree(kvm->arch.adapters[i]);
2010 static int modify_io_adapter(struct kvm_device *dev,
2011 struct kvm_device_attr *attr)
2013 struct kvm_s390_io_adapter_req req;
2014 struct s390_io_adapter *adapter;
2017 if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2020 adapter = get_io_adapter(dev->kvm, req.id);
2024 case KVM_S390_IO_ADAPTER_MASK:
2025 ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2029 case KVM_S390_IO_ADAPTER_MAP:
2030 ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2032 case KVM_S390_IO_ADAPTER_UNMAP:
2033 ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2042 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2045 const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2050 if (attr->attr != sizeof(schid))
2052 if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2054 kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2056 * If userspace is conforming to the architecture, we can have at most
2057 * one pending I/O interrupt per subchannel, so this is effectively a
2063 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2067 struct kvm_vcpu *vcpu;
2069 switch (attr->group) {
2070 case KVM_DEV_FLIC_ENQUEUE:
2071 r = enqueue_floating_irq(dev, attr);
2073 case KVM_DEV_FLIC_CLEAR_IRQS:
2074 kvm_s390_clear_float_irqs(dev->kvm);
2076 case KVM_DEV_FLIC_APF_ENABLE:
2077 dev->kvm->arch.gmap->pfault_enabled = 1;
2079 case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2080 dev->kvm->arch.gmap->pfault_enabled = 0;
2082 * Make sure no async faults are in transition when
2083 * clearing the queues. So we don't need to worry
2084 * about late coming workers.
2086 synchronize_srcu(&dev->kvm->srcu);
2087 kvm_for_each_vcpu(i, vcpu, dev->kvm)
2088 kvm_clear_async_pf_completion_queue(vcpu);
2090 case KVM_DEV_FLIC_ADAPTER_REGISTER:
2091 r = register_io_adapter(dev, attr);
2093 case KVM_DEV_FLIC_ADAPTER_MODIFY:
2094 r = modify_io_adapter(dev, attr);
2096 case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2097 r = clear_io_irq(dev->kvm, attr);
2106 static int flic_has_attr(struct kvm_device *dev,
2107 struct kvm_device_attr *attr)
2109 switch (attr->group) {
2110 case KVM_DEV_FLIC_GET_ALL_IRQS:
2111 case KVM_DEV_FLIC_ENQUEUE:
2112 case KVM_DEV_FLIC_CLEAR_IRQS:
2113 case KVM_DEV_FLIC_APF_ENABLE:
2114 case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2115 case KVM_DEV_FLIC_ADAPTER_REGISTER:
2116 case KVM_DEV_FLIC_ADAPTER_MODIFY:
2117 case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2123 static int flic_create(struct kvm_device *dev, u32 type)
2127 if (dev->kvm->arch.flic)
2129 dev->kvm->arch.flic = dev;
2133 static void flic_destroy(struct kvm_device *dev)
2135 dev->kvm->arch.flic = NULL;
2139 /* s390 floating irq controller (flic) */
2140 struct kvm_device_ops kvm_flic_ops = {
2142 .get_attr = flic_get_attr,
2143 .set_attr = flic_set_attr,
2144 .has_attr = flic_has_attr,
2145 .create = flic_create,
2146 .destroy = flic_destroy,
2149 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2153 bit = bit_nr + (addr % PAGE_SIZE) * 8;
2155 return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2158 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2161 struct s390_map_info *map;
2166 list_for_each_entry(map, &adapter->maps, list) {
2167 if (map->guest_addr == addr)
2173 static int adapter_indicators_set(struct kvm *kvm,
2174 struct s390_io_adapter *adapter,
2175 struct kvm_s390_adapter_int *adapter_int)
2178 int summary_set, idx;
2179 struct s390_map_info *info;
2182 info = get_map_info(adapter, adapter_int->ind_addr);
2185 map = page_address(info->page);
2186 bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2188 idx = srcu_read_lock(&kvm->srcu);
2189 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2190 set_page_dirty_lock(info->page);
2191 info = get_map_info(adapter, adapter_int->summary_addr);
2193 srcu_read_unlock(&kvm->srcu, idx);
2196 map = page_address(info->page);
2197 bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2199 summary_set = test_and_set_bit(bit, map);
2200 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2201 set_page_dirty_lock(info->page);
2202 srcu_read_unlock(&kvm->srcu, idx);
2203 return summary_set ? 0 : 1;
2207 * < 0 - not injected due to error
2208 * = 0 - coalesced, summary indicator already active
2209 * > 0 - injected interrupt
2211 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2212 struct kvm *kvm, int irq_source_id, int level,
2216 struct s390_io_adapter *adapter;
2218 /* We're only interested in the 0->1 transition. */
2221 adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2224 down_read(&adapter->maps_lock);
2225 ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2226 up_read(&adapter->maps_lock);
2227 if ((ret > 0) && !adapter->masked) {
2228 struct kvm_s390_interrupt s390int = {
2229 .type = KVM_S390_INT_IO(1, 0, 0, 0),
2231 .parm64 = (adapter->isc << 27) | 0x80000000,
2233 ret = kvm_s390_inject_vm(kvm, &s390int);
2240 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
2241 const struct kvm_irq_routing_entry *ue)
2246 case KVM_IRQ_ROUTING_S390_ADAPTER:
2247 e->set = set_adapter_int;
2248 e->adapter.summary_addr = ue->u.adapter.summary_addr;
2249 e->adapter.ind_addr = ue->u.adapter.ind_addr;
2250 e->adapter.summary_offset = ue->u.adapter.summary_offset;
2251 e->adapter.ind_offset = ue->u.adapter.ind_offset;
2252 e->adapter.adapter_id = ue->u.adapter.adapter_id;
2262 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2263 int irq_source_id, int level, bool line_status)
2268 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2270 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2271 struct kvm_s390_irq *buf;
2279 if (copy_from_user((void *) buf, irqstate, len)) {
2285 * Don't allow setting the interrupt state
2286 * when there are already interrupts pending
2288 spin_lock(&li->lock);
2289 if (li->pending_irqs) {
2294 for (n = 0; n < len / sizeof(*buf); n++) {
2295 r = do_inject_vcpu(vcpu, &buf[n]);
2301 spin_unlock(&li->lock);
2308 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2309 struct kvm_s390_irq *irq,
2310 unsigned long irq_type)
2313 case IRQ_PEND_MCHK_EX:
2314 case IRQ_PEND_MCHK_REP:
2315 irq->type = KVM_S390_MCHK;
2316 irq->u.mchk = li->irq.mchk;
2319 irq->type = KVM_S390_PROGRAM_INT;
2320 irq->u.pgm = li->irq.pgm;
2322 case IRQ_PEND_PFAULT_INIT:
2323 irq->type = KVM_S390_INT_PFAULT_INIT;
2324 irq->u.ext = li->irq.ext;
2326 case IRQ_PEND_EXT_EXTERNAL:
2327 irq->type = KVM_S390_INT_EXTERNAL_CALL;
2328 irq->u.extcall = li->irq.extcall;
2330 case IRQ_PEND_EXT_CLOCK_COMP:
2331 irq->type = KVM_S390_INT_CLOCK_COMP;
2333 case IRQ_PEND_EXT_CPU_TIMER:
2334 irq->type = KVM_S390_INT_CPU_TIMER;
2336 case IRQ_PEND_SIGP_STOP:
2337 irq->type = KVM_S390_SIGP_STOP;
2338 irq->u.stop = li->irq.stop;
2340 case IRQ_PEND_RESTART:
2341 irq->type = KVM_S390_RESTART;
2343 case IRQ_PEND_SET_PREFIX:
2344 irq->type = KVM_S390_SIGP_SET_PREFIX;
2345 irq->u.prefix = li->irq.prefix;
2350 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2353 unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2354 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2355 unsigned long pending_irqs;
2356 struct kvm_s390_irq irq;
2357 unsigned long irq_type;
2361 spin_lock(&li->lock);
2362 pending_irqs = li->pending_irqs;
2363 memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2364 sizeof(sigp_emerg_pending));
2365 spin_unlock(&li->lock);
2367 for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2368 memset(&irq, 0, sizeof(irq));
2369 if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2371 if (n + sizeof(irq) > len)
2373 store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2374 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2379 if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2380 for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2381 memset(&irq, 0, sizeof(irq));
2382 if (n + sizeof(irq) > len)
2384 irq.type = KVM_S390_INT_EMERGENCY;
2385 irq.u.emerg.code = cpuaddr;
2386 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2392 if (sca_ext_call_pending(vcpu, &scn)) {
2393 if (n + sizeof(irq) > len)
2395 memset(&irq, 0, sizeof(irq));
2396 irq.type = KVM_S390_INT_EXTERNAL_CALL;
2397 irq.u.extcall.code = scn;
2398 if (copy_to_user(&buf[n], &irq, sizeof(irq)))