2 * handling kvm guest interrupts
4 * Copyright IBM Corp. 2008,2014
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 <asm/asm-offsets.h>
21 #include <asm/uaccess.h>
24 #include "trace-s390.h"
26 #define IOINT_SCHID_MASK 0x0000ffff
27 #define IOINT_SSID_MASK 0x00030000
28 #define IOINT_CSSID_MASK 0x03fc0000
29 #define IOINT_AI_MASK 0x04000000
30 #define PFAULT_INIT 0x0600
31 #define PFAULT_DONE 0x0680
32 #define VIRTIO_PARAM 0x0d00
34 static int is_ioint(u64 type)
36 return ((type & 0xfffe0000u) != 0xfffe0000u);
39 int psw_extint_disabled(struct kvm_vcpu *vcpu)
41 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
44 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
46 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
49 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
51 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
54 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
56 if ((vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PER) ||
57 (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO) ||
58 (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT))
63 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
65 if (psw_extint_disabled(vcpu) ||
66 !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
68 if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
69 /* No timer interrupts when single stepping */
74 static u64 int_word_to_isc_bits(u32 int_word)
76 u8 isc = (int_word & 0x38000000) >> 27;
78 return (0x80 >> isc) << 24;
81 static int __must_check __interrupt_is_deliverable(struct kvm_vcpu *vcpu,
82 struct kvm_s390_interrupt_info *inti)
85 case KVM_S390_INT_EXTERNAL_CALL:
86 if (psw_extint_disabled(vcpu))
88 if (vcpu->arch.sie_block->gcr[0] & 0x2000ul)
91 case KVM_S390_INT_EMERGENCY:
92 if (psw_extint_disabled(vcpu))
94 if (vcpu->arch.sie_block->gcr[0] & 0x4000ul)
97 case KVM_S390_INT_CLOCK_COMP:
98 return ckc_interrupts_enabled(vcpu);
99 case KVM_S390_INT_CPU_TIMER:
100 if (psw_extint_disabled(vcpu))
102 if (vcpu->arch.sie_block->gcr[0] & 0x400ul)
105 case KVM_S390_INT_SERVICE:
106 case KVM_S390_INT_PFAULT_INIT:
107 case KVM_S390_INT_PFAULT_DONE:
108 case KVM_S390_INT_VIRTIO:
109 if (psw_extint_disabled(vcpu))
111 if (vcpu->arch.sie_block->gcr[0] & 0x200ul)
114 case KVM_S390_PROGRAM_INT:
115 case KVM_S390_SIGP_STOP:
116 case KVM_S390_SIGP_SET_PREFIX:
117 case KVM_S390_RESTART:
120 if (psw_mchk_disabled(vcpu))
122 if (vcpu->arch.sie_block->gcr[14] & inti->mchk.cr14)
125 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
126 if (psw_ioint_disabled(vcpu))
128 if (vcpu->arch.sie_block->gcr[6] &
129 int_word_to_isc_bits(inti->io.io_int_word))
133 printk(KERN_WARNING "illegal interrupt type %llx\n",
140 static inline unsigned long pending_local_irqs(struct kvm_vcpu *vcpu)
142 return vcpu->arch.local_int.pending_irqs;
145 static unsigned long deliverable_local_irqs(struct kvm_vcpu *vcpu)
147 unsigned long active_mask = pending_local_irqs(vcpu);
149 if (psw_extint_disabled(vcpu))
150 active_mask &= ~IRQ_PEND_EXT_MASK;
151 if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
152 __clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
153 if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
154 __clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
155 if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
156 __clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
157 if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
158 __clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
159 if (psw_mchk_disabled(vcpu))
160 active_mask &= ~IRQ_PEND_MCHK_MASK;
165 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
167 atomic_set_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
168 set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
171 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
173 atomic_clear_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
174 clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
177 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
179 atomic_clear_mask(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
180 &vcpu->arch.sie_block->cpuflags);
181 vcpu->arch.sie_block->lctl = 0x0000;
182 vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
184 if (guestdbg_enabled(vcpu)) {
185 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
186 LCTL_CR10 | LCTL_CR11);
187 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
190 if (vcpu->arch.local_int.action_bits & ACTION_STOP_ON_STOP)
191 atomic_set_mask(CPUSTAT_STOP_INT, &vcpu->arch.sie_block->cpuflags);
194 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
196 atomic_set_mask(flag, &vcpu->arch.sie_block->cpuflags);
199 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
201 if (!(pending_local_irqs(vcpu) & IRQ_PEND_EXT_MASK))
203 if (psw_extint_disabled(vcpu))
204 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
206 vcpu->arch.sie_block->lctl |= LCTL_CR0;
209 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
211 if (!(pending_local_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
213 if (psw_mchk_disabled(vcpu))
214 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
216 vcpu->arch.sie_block->lctl |= LCTL_CR14;
219 /* Set interception request for non-deliverable local interrupts */
220 static void set_intercept_indicators_local(struct kvm_vcpu *vcpu)
222 set_intercept_indicators_ext(vcpu);
223 set_intercept_indicators_mchk(vcpu);
226 static void __set_intercept_indicator(struct kvm_vcpu *vcpu,
227 struct kvm_s390_interrupt_info *inti)
229 switch (inti->type) {
230 case KVM_S390_INT_SERVICE:
231 case KVM_S390_INT_PFAULT_DONE:
232 case KVM_S390_INT_VIRTIO:
233 if (psw_extint_disabled(vcpu))
234 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
236 vcpu->arch.sie_block->lctl |= LCTL_CR0;
239 if (psw_mchk_disabled(vcpu))
240 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
242 vcpu->arch.sie_block->lctl |= LCTL_CR14;
244 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
245 if (psw_ioint_disabled(vcpu))
246 __set_cpuflag(vcpu, CPUSTAT_IO_INT);
248 vcpu->arch.sie_block->lctl |= LCTL_CR6;
255 static u16 get_ilc(struct kvm_vcpu *vcpu)
257 const unsigned short table[] = { 2, 4, 4, 6 };
259 switch (vcpu->arch.sie_block->icptcode) {
265 /* last instruction only stored for these icptcodes */
266 return table[vcpu->arch.sie_block->ipa >> 14];
268 return vcpu->arch.sie_block->pgmilc;
274 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
276 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
279 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
282 rc = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
283 (u16 *)__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_CPU_TIMER, &li->pending_irqs);
290 return rc ? -EFAULT : 0;
293 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
295 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
298 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
301 rc = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
302 (u16 __user *)__LC_EXT_INT_CODE);
303 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
304 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
305 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
306 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
307 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
308 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
309 return rc ? -EFAULT : 0;
312 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
314 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
315 struct kvm_s390_ext_info ext;
318 spin_lock(&li->lock);
320 clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
321 li->irq.ext.ext_params2 = 0;
322 spin_unlock(&li->lock);
324 VCPU_EVENT(vcpu, 4, "interrupt: pfault init parm:%x,parm64:%llx",
326 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
327 KVM_S390_INT_PFAULT_INIT,
330 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
331 rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
332 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
333 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
334 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
335 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
336 rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
337 return rc ? -EFAULT : 0;
340 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
342 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
343 struct kvm_s390_mchk_info mchk;
346 spin_lock(&li->lock);
349 * If there was an exigent machine check pending, then any repressible
350 * machine checks that might have been pending are indicated along
351 * with it, so always clear both bits
353 clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
354 clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
355 memset(&li->irq.mchk, 0, sizeof(mchk));
356 spin_unlock(&li->lock);
358 VCPU_EVENT(vcpu, 4, "interrupt: machine check mcic=%llx",
360 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_MCHK,
361 mchk.cr14, mchk.mcic);
363 rc = kvm_s390_vcpu_store_status(vcpu, KVM_S390_STORE_STATUS_PREFIXED);
364 rc |= put_guest_lc(vcpu, mchk.mcic,
365 (u64 __user *) __LC_MCCK_CODE);
366 rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
367 (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
368 rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
369 &mchk.fixed_logout, sizeof(mchk.fixed_logout));
370 rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
371 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
372 rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
373 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
374 return rc ? -EFAULT : 0;
377 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
379 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
382 VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu restart");
383 vcpu->stat.deliver_restart_signal++;
384 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
386 rc = write_guest_lc(vcpu,
387 offsetof(struct _lowcore, restart_old_psw),
388 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
389 rc |= read_guest_lc(vcpu, offsetof(struct _lowcore, restart_psw),
390 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
391 clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
392 return rc ? -EFAULT : 0;
395 static int __must_check __deliver_stop(struct kvm_vcpu *vcpu)
397 VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu stop");
398 vcpu->stat.deliver_stop_signal++;
399 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_SIGP_STOP,
402 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
403 clear_bit(IRQ_PEND_SIGP_STOP, &vcpu->arch.local_int.pending_irqs);
407 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
409 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
410 struct kvm_s390_prefix_info prefix;
412 spin_lock(&li->lock);
413 prefix = li->irq.prefix;
414 li->irq.prefix.address = 0;
415 clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
416 spin_unlock(&li->lock);
418 VCPU_EVENT(vcpu, 4, "interrupt: set prefix to %x", prefix.address);
419 vcpu->stat.deliver_prefix_signal++;
420 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
421 KVM_S390_SIGP_SET_PREFIX,
424 kvm_s390_set_prefix(vcpu, prefix.address);
428 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
430 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
434 spin_lock(&li->lock);
435 cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
436 clear_bit(cpu_addr, li->sigp_emerg_pending);
437 if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
438 clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
439 spin_unlock(&li->lock);
441 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp emerg");
442 vcpu->stat.deliver_emergency_signal++;
443 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
446 rc = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
447 (u16 *)__LC_EXT_INT_CODE);
448 rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
449 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
450 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
451 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
452 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
453 return rc ? -EFAULT : 0;
456 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
458 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
459 struct kvm_s390_extcall_info extcall;
462 spin_lock(&li->lock);
463 extcall = li->irq.extcall;
464 li->irq.extcall.code = 0;
465 clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
466 spin_unlock(&li->lock);
468 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp ext call");
469 vcpu->stat.deliver_external_call++;
470 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
471 KVM_S390_INT_EXTERNAL_CALL,
474 rc = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
475 (u16 *)__LC_EXT_INT_CODE);
476 rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
477 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
478 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
479 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
481 return rc ? -EFAULT : 0;
484 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
486 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
487 struct kvm_s390_pgm_info pgm_info;
489 u16 ilc = get_ilc(vcpu);
491 spin_lock(&li->lock);
492 pgm_info = li->irq.pgm;
493 clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
494 memset(&li->irq.pgm, 0, sizeof(pgm_info));
495 spin_unlock(&li->lock);
497 VCPU_EVENT(vcpu, 4, "interrupt: pgm check code:%x, ilc:%x",
499 vcpu->stat.deliver_program_int++;
500 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
503 switch (pgm_info.code & ~PGM_PER) {
504 case PGM_AFX_TRANSLATION:
505 case PGM_ASX_TRANSLATION:
506 case PGM_EX_TRANSLATION:
507 case PGM_LFX_TRANSLATION:
508 case PGM_LSTE_SEQUENCE:
509 case PGM_LSX_TRANSLATION:
510 case PGM_LX_TRANSLATION:
511 case PGM_PRIMARY_AUTHORITY:
512 case PGM_SECONDARY_AUTHORITY:
513 case PGM_SPACE_SWITCH:
514 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
515 (u64 *)__LC_TRANS_EXC_CODE);
517 case PGM_ALEN_TRANSLATION:
518 case PGM_ALE_SEQUENCE:
519 case PGM_ASTE_INSTANCE:
520 case PGM_ASTE_SEQUENCE:
521 case PGM_ASTE_VALIDITY:
522 case PGM_EXTENDED_AUTHORITY:
523 rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
524 (u8 *)__LC_EXC_ACCESS_ID);
527 case PGM_PAGE_TRANSLATION:
528 case PGM_REGION_FIRST_TRANS:
529 case PGM_REGION_SECOND_TRANS:
530 case PGM_REGION_THIRD_TRANS:
531 case PGM_SEGMENT_TRANSLATION:
532 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
533 (u64 *)__LC_TRANS_EXC_CODE);
534 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
535 (u8 *)__LC_EXC_ACCESS_ID);
536 rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
537 (u8 *)__LC_OP_ACCESS_ID);
540 rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
541 (u16 *)__LC_MON_CLASS_NR);
542 rc |= put_guest_lc(vcpu, pgm_info.mon_code,
543 (u64 *)__LC_MON_CODE);
546 rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
547 (u32 *)__LC_DATA_EXC_CODE);
550 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
551 (u64 *)__LC_TRANS_EXC_CODE);
552 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
553 (u8 *)__LC_EXC_ACCESS_ID);
557 if (pgm_info.code & PGM_PER) {
558 rc |= put_guest_lc(vcpu, pgm_info.per_code,
559 (u8 *) __LC_PER_CODE);
560 rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
561 (u8 *)__LC_PER_ATMID);
562 rc |= put_guest_lc(vcpu, pgm_info.per_address,
563 (u64 *) __LC_PER_ADDRESS);
564 rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
565 (u8 *) __LC_PER_ACCESS_ID);
568 rc |= put_guest_lc(vcpu, ilc, (u16 *) __LC_PGM_ILC);
569 rc |= put_guest_lc(vcpu, pgm_info.code,
570 (u16 *)__LC_PGM_INT_CODE);
571 rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
572 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
573 rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
574 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
575 return rc ? -EFAULT : 0;
578 static int __must_check __deliver_service(struct kvm_vcpu *vcpu,
579 struct kvm_s390_interrupt_info *inti)
583 VCPU_EVENT(vcpu, 4, "interrupt: sclp parm:%x",
584 inti->ext.ext_params);
585 vcpu->stat.deliver_service_signal++;
586 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
587 inti->ext.ext_params, 0);
589 rc = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
590 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
591 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
592 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
593 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
594 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
595 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
596 (u32 *)__LC_EXT_PARAMS);
597 return rc ? -EFAULT : 0;
600 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu,
601 struct kvm_s390_interrupt_info *inti)
605 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
606 KVM_S390_INT_PFAULT_DONE, 0,
607 inti->ext.ext_params2);
609 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *)__LC_EXT_INT_CODE);
610 rc |= put_guest_lc(vcpu, PFAULT_DONE, (u16 *)__LC_EXT_CPU_ADDR);
611 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
612 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
613 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
614 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
615 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
616 (u64 *)__LC_EXT_PARAMS2);
617 return rc ? -EFAULT : 0;
620 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu,
621 struct kvm_s390_interrupt_info *inti)
625 VCPU_EVENT(vcpu, 4, "interrupt: virtio parm:%x,parm64:%llx",
626 inti->ext.ext_params, inti->ext.ext_params2);
627 vcpu->stat.deliver_virtio_interrupt++;
628 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
629 inti->ext.ext_params,
630 inti->ext.ext_params2);
632 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *)__LC_EXT_INT_CODE);
633 rc |= put_guest_lc(vcpu, VIRTIO_PARAM, (u16 *)__LC_EXT_CPU_ADDR);
634 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
635 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
636 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
637 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
638 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
639 (u32 *)__LC_EXT_PARAMS);
640 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
641 (u64 *)__LC_EXT_PARAMS2);
642 return rc ? -EFAULT : 0;
645 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
646 struct kvm_s390_interrupt_info *inti)
650 VCPU_EVENT(vcpu, 4, "interrupt: I/O %llx", inti->type);
651 vcpu->stat.deliver_io_int++;
652 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, inti->type,
653 ((__u32)inti->io.subchannel_id << 16) |
654 inti->io.subchannel_nr,
655 ((__u64)inti->io.io_int_parm << 32) |
656 inti->io.io_int_word);
658 rc = put_guest_lc(vcpu, inti->io.subchannel_id,
659 (u16 *)__LC_SUBCHANNEL_ID);
660 rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
661 (u16 *)__LC_SUBCHANNEL_NR);
662 rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
663 (u32 *)__LC_IO_INT_PARM);
664 rc |= put_guest_lc(vcpu, inti->io.io_int_word,
665 (u32 *)__LC_IO_INT_WORD);
666 rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
667 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
668 rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
669 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
670 return rc ? -EFAULT : 0;
673 static int __must_check __deliver_mchk_floating(struct kvm_vcpu *vcpu,
674 struct kvm_s390_interrupt_info *inti)
676 struct kvm_s390_mchk_info *mchk = &inti->mchk;
679 VCPU_EVENT(vcpu, 4, "interrupt: machine check mcic=%llx",
681 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_MCHK,
682 mchk->cr14, mchk->mcic);
684 rc = kvm_s390_vcpu_store_status(vcpu, KVM_S390_STORE_STATUS_PREFIXED);
685 rc |= put_guest_lc(vcpu, mchk->mcic,
686 (u64 __user *) __LC_MCCK_CODE);
687 rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
688 (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
689 rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
690 &mchk->fixed_logout, sizeof(mchk->fixed_logout));
691 rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
692 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
693 rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
694 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
695 return rc ? -EFAULT : 0;
698 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
700 static const deliver_irq_t deliver_irq_funcs[] = {
701 [IRQ_PEND_MCHK_EX] = __deliver_machine_check,
702 [IRQ_PEND_PROG] = __deliver_prog,
703 [IRQ_PEND_EXT_EMERGENCY] = __deliver_emergency_signal,
704 [IRQ_PEND_EXT_EXTERNAL] = __deliver_external_call,
705 [IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
706 [IRQ_PEND_EXT_CPU_TIMER] = __deliver_cpu_timer,
707 [IRQ_PEND_RESTART] = __deliver_restart,
708 [IRQ_PEND_SIGP_STOP] = __deliver_stop,
709 [IRQ_PEND_SET_PREFIX] = __deliver_set_prefix,
710 [IRQ_PEND_PFAULT_INIT] = __deliver_pfault_init,
713 static int __must_check __deliver_floating_interrupt(struct kvm_vcpu *vcpu,
714 struct kvm_s390_interrupt_info *inti)
718 switch (inti->type) {
719 case KVM_S390_INT_SERVICE:
720 rc = __deliver_service(vcpu, inti);
722 case KVM_S390_INT_PFAULT_DONE:
723 rc = __deliver_pfault_done(vcpu, inti);
725 case KVM_S390_INT_VIRTIO:
726 rc = __deliver_virtio(vcpu, inti);
729 rc = __deliver_mchk_floating(vcpu, inti);
731 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
732 rc = __deliver_io(vcpu, inti);
741 /* Check whether SIGP interpretation facility has an external call pending */
742 int kvm_s390_si_ext_call_pending(struct kvm_vcpu *vcpu)
744 atomic_t *sigp_ctrl = &vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].ctrl;
746 if (!psw_extint_disabled(vcpu) &&
747 (vcpu->arch.sie_block->gcr[0] & 0x2000ul) &&
748 (atomic_read(sigp_ctrl) & SIGP_CTRL_C) &&
749 (atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
755 int kvm_cpu_has_interrupt(struct kvm_vcpu *vcpu)
757 struct kvm_s390_float_interrupt *fi = vcpu->arch.local_int.float_int;
758 struct kvm_s390_interrupt_info *inti;
761 rc = !!deliverable_local_irqs(vcpu);
763 if ((!rc) && atomic_read(&fi->active)) {
764 spin_lock(&fi->lock);
765 list_for_each_entry(inti, &fi->list, list)
766 if (__interrupt_is_deliverable(vcpu, inti)) {
770 spin_unlock(&fi->lock);
773 if (!rc && kvm_cpu_has_pending_timer(vcpu))
776 if (!rc && kvm_s390_si_ext_call_pending(vcpu))
782 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
784 if (!(vcpu->arch.sie_block->ckc <
785 get_tod_clock_fast() + vcpu->arch.sie_block->epoch))
787 if (!ckc_interrupts_enabled(vcpu))
792 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
796 vcpu->stat.exit_wait_state++;
799 if (kvm_cpu_has_pending_timer(vcpu) || kvm_arch_vcpu_runnable(vcpu))
802 if (psw_interrupts_disabled(vcpu)) {
803 VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
804 return -EOPNOTSUPP; /* disabled wait */
807 __set_cpu_idle(vcpu);
808 if (!ckc_interrupts_enabled(vcpu)) {
809 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
813 now = get_tod_clock_fast() + vcpu->arch.sie_block->epoch;
814 sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
815 hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
816 VCPU_EVENT(vcpu, 5, "enabled wait via clock comparator: %llx ns", sltime);
818 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
819 kvm_vcpu_block(vcpu);
820 __unset_cpu_idle(vcpu);
821 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
823 hrtimer_try_to_cancel(&vcpu->arch.ckc_timer);
827 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
829 if (waitqueue_active(&vcpu->wq)) {
831 * The vcpu gave up the cpu voluntarily, mark it as a good
834 vcpu->preempted = true;
835 wake_up_interruptible(&vcpu->wq);
836 vcpu->stat.halt_wakeup++;
840 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
842 struct kvm_vcpu *vcpu;
844 vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
845 kvm_s390_vcpu_wakeup(vcpu);
847 return HRTIMER_NORESTART;
850 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
852 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
854 spin_lock(&li->lock);
855 li->pending_irqs = 0;
856 bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
857 memset(&li->irq, 0, sizeof(li->irq));
858 spin_unlock(&li->lock);
860 /* clear pending external calls set by sigp interpretation facility */
861 atomic_clear_mask(CPUSTAT_ECALL_PEND, li->cpuflags);
862 atomic_clear_mask(SIGP_CTRL_C,
863 &vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].ctrl);
866 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
868 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
869 struct kvm_s390_float_interrupt *fi = vcpu->arch.local_int.float_int;
870 struct kvm_s390_interrupt_info *n, *inti = NULL;
874 unsigned long irq_type;
875 unsigned long deliverable_irqs;
877 __reset_intercept_indicators(vcpu);
879 /* pending ckc conditions might have been invalidated */
880 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
881 if (kvm_cpu_has_pending_timer(vcpu))
882 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
885 deliverable_irqs = deliverable_local_irqs(vcpu);
886 /* bits are in the order of interrupt priority */
887 irq_type = find_first_bit(&deliverable_irqs, IRQ_PEND_COUNT);
888 if (irq_type == IRQ_PEND_COUNT)
890 func = deliver_irq_funcs[irq_type];
892 WARN_ON_ONCE(func == NULL);
893 clear_bit(irq_type, &li->pending_irqs);
897 } while (!rc && irq_type != IRQ_PEND_COUNT);
899 set_intercept_indicators_local(vcpu);
901 if (!rc && atomic_read(&fi->active)) {
904 spin_lock(&fi->lock);
905 list_for_each_entry_safe(inti, n, &fi->list, list) {
906 if (__interrupt_is_deliverable(vcpu, inti)) {
907 list_del(&inti->list);
912 __set_intercept_indicator(vcpu, inti);
914 if (list_empty(&fi->list))
915 atomic_set(&fi->active, 0);
916 spin_unlock(&fi->lock);
918 rc = __deliver_floating_interrupt(vcpu, inti);
921 } while (!rc && deliver);
927 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
929 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
931 li->irq.pgm = irq->u.pgm;
932 set_bit(IRQ_PEND_PROG, &li->pending_irqs);
936 int kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code)
938 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
939 struct kvm_s390_irq irq;
941 VCPU_EVENT(vcpu, 3, "inject: program check %d (from kernel)", code);
942 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT, code,
944 spin_lock(&li->lock);
945 irq.u.pgm.code = code;
946 __inject_prog(vcpu, &irq);
947 BUG_ON(waitqueue_active(li->wq));
948 spin_unlock(&li->lock);
952 int kvm_s390_inject_prog_irq(struct kvm_vcpu *vcpu,
953 struct kvm_s390_pgm_info *pgm_info)
955 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
956 struct kvm_s390_irq irq;
959 VCPU_EVENT(vcpu, 3, "inject: prog irq %d (from kernel)",
961 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
962 pgm_info->code, 0, 1);
963 spin_lock(&li->lock);
964 irq.u.pgm = *pgm_info;
965 rc = __inject_prog(vcpu, &irq);
966 BUG_ON(waitqueue_active(li->wq));
967 spin_unlock(&li->lock);
971 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
973 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
975 VCPU_EVENT(vcpu, 3, "inject: external irq params:%x, params2:%llx",
976 irq->u.ext.ext_params, irq->u.ext.ext_params2);
977 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
978 irq->u.ext.ext_params,
979 irq->u.ext.ext_params2, 2);
981 li->irq.ext = irq->u.ext;
982 set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
983 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
987 int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
989 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
990 struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
992 VCPU_EVENT(vcpu, 3, "inject: external call source-cpu:%u",
993 irq->u.extcall.code);
994 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
995 irq->u.extcall.code, 0, 2);
997 *extcall = irq->u.extcall;
998 set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
999 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1003 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1005 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1006 struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1008 VCPU_EVENT(vcpu, 3, "inject: set prefix to %x (from user)",
1010 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1011 prefix->address, 0, 2);
1013 *prefix = irq->u.prefix;
1014 set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1018 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1020 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1022 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0, 2);
1024 li->action_bits |= ACTION_STOP_ON_STOP;
1025 set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1029 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1030 struct kvm_s390_irq *irq)
1032 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1034 VCPU_EVENT(vcpu, 3, "inject: restart type %llx", irq->type);
1035 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0, 2);
1037 set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1041 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1042 struct kvm_s390_irq *irq)
1044 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1045 struct kvm_s390_emerg_info *emerg = &li->irq.emerg;
1047 VCPU_EVENT(vcpu, 3, "inject: emergency %u\n",
1049 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1052 set_bit(emerg->code, li->sigp_emerg_pending);
1053 set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1054 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1058 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1060 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1061 struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1063 VCPU_EVENT(vcpu, 5, "inject: machine check parm64:%llx",
1065 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1069 * Because repressible machine checks can be indicated along with
1070 * exigent machine checks (PoP, Chapter 11, Interruption action)
1071 * we need to combine cr14, mcic and external damage code.
1072 * Failing storage address and the logout area should not be or'ed
1073 * together, we just indicate the last occurrence of the corresponding
1076 mchk->cr14 |= irq->u.mchk.cr14;
1077 mchk->mcic |= irq->u.mchk.mcic;
1078 mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1079 mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1080 memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1081 sizeof(mchk->fixed_logout));
1082 if (mchk->mcic & MCHK_EX_MASK)
1083 set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1084 else if (mchk->mcic & MCHK_REP_MASK)
1085 set_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
1089 static int __inject_ckc(struct kvm_vcpu *vcpu)
1091 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1093 VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CLOCK_COMP);
1094 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1097 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1098 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1102 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1104 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1106 VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CPU_TIMER);
1107 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1110 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1111 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1116 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1119 struct kvm_s390_float_interrupt *fi;
1120 struct kvm_s390_interrupt_info *inti, *iter;
1122 if ((!schid && !cr6) || (schid && cr6))
1124 mutex_lock(&kvm->lock);
1125 fi = &kvm->arch.float_int;
1126 spin_lock(&fi->lock);
1128 list_for_each_entry(iter, &fi->list, list) {
1129 if (!is_ioint(iter->type))
1132 ((cr6 & int_word_to_isc_bits(iter->io.io_int_word)) == 0))
1135 if (((schid & 0x00000000ffff0000) >> 16) !=
1136 iter->io.subchannel_id)
1138 if ((schid & 0x000000000000ffff) !=
1139 iter->io.subchannel_nr)
1146 list_del_init(&inti->list);
1149 if (list_empty(&fi->list))
1150 atomic_set(&fi->active, 0);
1151 spin_unlock(&fi->lock);
1152 mutex_unlock(&kvm->lock);
1156 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1158 struct kvm_s390_local_interrupt *li;
1159 struct kvm_s390_float_interrupt *fi;
1160 struct kvm_s390_interrupt_info *iter;
1161 struct kvm_vcpu *dst_vcpu = NULL;
1165 mutex_lock(&kvm->lock);
1166 fi = &kvm->arch.float_int;
1167 spin_lock(&fi->lock);
1168 if (fi->irq_count >= KVM_S390_MAX_FLOAT_IRQS) {
1173 if (!is_ioint(inti->type)) {
1174 list_add_tail(&inti->list, &fi->list);
1176 u64 isc_bits = int_word_to_isc_bits(inti->io.io_int_word);
1178 /* Keep I/O interrupts sorted in isc order. */
1179 list_for_each_entry(iter, &fi->list, list) {
1180 if (!is_ioint(iter->type))
1182 if (int_word_to_isc_bits(iter->io.io_int_word)
1187 list_add_tail(&inti->list, &iter->list);
1189 atomic_set(&fi->active, 1);
1190 sigcpu = find_first_bit(fi->idle_mask, KVM_MAX_VCPUS);
1191 if (sigcpu == KVM_MAX_VCPUS) {
1193 sigcpu = fi->next_rr_cpu++;
1194 if (sigcpu == KVM_MAX_VCPUS)
1195 sigcpu = fi->next_rr_cpu = 0;
1196 } while (kvm_get_vcpu(kvm, sigcpu) == NULL);
1198 dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1199 li = &dst_vcpu->arch.local_int;
1200 spin_lock(&li->lock);
1201 switch (inti->type) {
1203 atomic_set_mask(CPUSTAT_STOP_INT, li->cpuflags);
1205 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1206 atomic_set_mask(CPUSTAT_IO_INT, li->cpuflags);
1209 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1212 spin_unlock(&li->lock);
1213 kvm_s390_vcpu_wakeup(kvm_get_vcpu(kvm, sigcpu));
1215 spin_unlock(&fi->lock);
1216 mutex_unlock(&kvm->lock);
1220 int kvm_s390_inject_vm(struct kvm *kvm,
1221 struct kvm_s390_interrupt *s390int)
1223 struct kvm_s390_interrupt_info *inti;
1225 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1229 inti->type = s390int->type;
1230 switch (inti->type) {
1231 case KVM_S390_INT_VIRTIO:
1232 VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1233 s390int->parm, s390int->parm64);
1234 inti->ext.ext_params = s390int->parm;
1235 inti->ext.ext_params2 = s390int->parm64;
1237 case KVM_S390_INT_SERVICE:
1238 VM_EVENT(kvm, 5, "inject: sclp parm:%x", s390int->parm);
1239 inti->ext.ext_params = s390int->parm;
1241 case KVM_S390_INT_PFAULT_DONE:
1242 inti->type = s390int->type;
1243 inti->ext.ext_params2 = s390int->parm64;
1246 VM_EVENT(kvm, 5, "inject: machine check parm64:%llx",
1248 inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1249 inti->mchk.mcic = s390int->parm64;
1251 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1252 if (inti->type & IOINT_AI_MASK)
1253 VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
1255 VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
1256 s390int->type & IOINT_CSSID_MASK,
1257 s390int->type & IOINT_SSID_MASK,
1258 s390int->type & IOINT_SCHID_MASK);
1259 inti->io.subchannel_id = s390int->parm >> 16;
1260 inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1261 inti->io.io_int_parm = s390int->parm64 >> 32;
1262 inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1268 trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1271 return __inject_vm(kvm, inti);
1274 void kvm_s390_reinject_io_int(struct kvm *kvm,
1275 struct kvm_s390_interrupt_info *inti)
1277 __inject_vm(kvm, inti);
1280 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1281 struct kvm_s390_irq *irq)
1283 irq->type = s390int->type;
1284 switch (irq->type) {
1285 case KVM_S390_PROGRAM_INT:
1286 if (s390int->parm & 0xffff0000)
1288 irq->u.pgm.code = s390int->parm;
1290 case KVM_S390_SIGP_SET_PREFIX:
1291 irq->u.prefix.address = s390int->parm;
1293 case KVM_S390_INT_EXTERNAL_CALL:
1294 if (irq->u.extcall.code & 0xffff0000)
1296 irq->u.extcall.code = s390int->parm;
1298 case KVM_S390_INT_EMERGENCY:
1299 if (irq->u.emerg.code & 0xffff0000)
1301 irq->u.emerg.code = s390int->parm;
1304 irq->u.mchk.mcic = s390int->parm64;
1310 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1312 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1315 spin_lock(&li->lock);
1316 switch (irq->type) {
1317 case KVM_S390_PROGRAM_INT:
1318 VCPU_EVENT(vcpu, 3, "inject: program check %d (from user)",
1320 rc = __inject_prog(vcpu, irq);
1322 case KVM_S390_SIGP_SET_PREFIX:
1323 rc = __inject_set_prefix(vcpu, irq);
1325 case KVM_S390_SIGP_STOP:
1326 rc = __inject_sigp_stop(vcpu, irq);
1328 case KVM_S390_RESTART:
1329 rc = __inject_sigp_restart(vcpu, irq);
1331 case KVM_S390_INT_CLOCK_COMP:
1332 rc = __inject_ckc(vcpu);
1334 case KVM_S390_INT_CPU_TIMER:
1335 rc = __inject_cpu_timer(vcpu);
1337 case KVM_S390_INT_EXTERNAL_CALL:
1338 rc = __inject_extcall(vcpu, irq);
1340 case KVM_S390_INT_EMERGENCY:
1341 rc = __inject_sigp_emergency(vcpu, irq);
1344 rc = __inject_mchk(vcpu, irq);
1346 case KVM_S390_INT_PFAULT_INIT:
1347 rc = __inject_pfault_init(vcpu, irq);
1349 case KVM_S390_INT_VIRTIO:
1350 case KVM_S390_INT_SERVICE:
1351 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1355 spin_unlock(&li->lock);
1357 kvm_s390_vcpu_wakeup(vcpu);
1361 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1363 struct kvm_s390_float_interrupt *fi;
1364 struct kvm_s390_interrupt_info *n, *inti = NULL;
1366 mutex_lock(&kvm->lock);
1367 fi = &kvm->arch.float_int;
1368 spin_lock(&fi->lock);
1369 list_for_each_entry_safe(inti, n, &fi->list, list) {
1370 list_del(&inti->list);
1374 atomic_set(&fi->active, 0);
1375 spin_unlock(&fi->lock);
1376 mutex_unlock(&kvm->lock);
1379 static inline int copy_irq_to_user(struct kvm_s390_interrupt_info *inti,
1382 struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1383 struct kvm_s390_irq irq = {0};
1385 irq.type = inti->type;
1386 switch (inti->type) {
1387 case KVM_S390_INT_PFAULT_INIT:
1388 case KVM_S390_INT_PFAULT_DONE:
1389 case KVM_S390_INT_VIRTIO:
1390 case KVM_S390_INT_SERVICE:
1391 irq.u.ext = inti->ext;
1393 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1394 irq.u.io = inti->io;
1397 irq.u.mchk = inti->mchk;
1403 if (copy_to_user(uptr, &irq, sizeof(irq)))
1409 static int get_all_floating_irqs(struct kvm *kvm, __u8 *buf, __u64 len)
1411 struct kvm_s390_interrupt_info *inti;
1412 struct kvm_s390_float_interrupt *fi;
1416 mutex_lock(&kvm->lock);
1417 fi = &kvm->arch.float_int;
1418 spin_lock(&fi->lock);
1420 list_for_each_entry(inti, &fi->list, list) {
1421 if (len < sizeof(struct kvm_s390_irq)) {
1422 /* signal userspace to try again */
1426 ret = copy_irq_to_user(inti, buf);
1429 buf += sizeof(struct kvm_s390_irq);
1430 len -= sizeof(struct kvm_s390_irq);
1434 spin_unlock(&fi->lock);
1435 mutex_unlock(&kvm->lock);
1437 return ret < 0 ? ret : n;
1440 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1444 switch (attr->group) {
1445 case KVM_DEV_FLIC_GET_ALL_IRQS:
1446 r = get_all_floating_irqs(dev->kvm, (u8 *) attr->addr,
1456 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1459 struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1460 void *target = NULL;
1461 void __user *source;
1464 if (get_user(inti->type, (u64 __user *)addr))
1467 switch (inti->type) {
1468 case KVM_S390_INT_PFAULT_INIT:
1469 case KVM_S390_INT_PFAULT_DONE:
1470 case KVM_S390_INT_VIRTIO:
1471 case KVM_S390_INT_SERVICE:
1472 target = (void *) &inti->ext;
1473 source = &uptr->u.ext;
1474 size = sizeof(inti->ext);
1476 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1477 target = (void *) &inti->io;
1478 source = &uptr->u.io;
1479 size = sizeof(inti->io);
1482 target = (void *) &inti->mchk;
1483 source = &uptr->u.mchk;
1484 size = sizeof(inti->mchk);
1490 if (copy_from_user(target, source, size))
1496 static int enqueue_floating_irq(struct kvm_device *dev,
1497 struct kvm_device_attr *attr)
1499 struct kvm_s390_interrupt_info *inti = NULL;
1501 int len = attr->attr;
1503 if (len % sizeof(struct kvm_s390_irq) != 0)
1505 else if (len > KVM_S390_FLIC_MAX_BUFFER)
1508 while (len >= sizeof(struct kvm_s390_irq)) {
1509 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1513 r = copy_irq_from_user(inti, attr->addr);
1518 r = __inject_vm(dev->kvm, inti);
1523 len -= sizeof(struct kvm_s390_irq);
1524 attr->addr += sizeof(struct kvm_s390_irq);
1530 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1532 if (id >= MAX_S390_IO_ADAPTERS)
1534 return kvm->arch.adapters[id];
1537 static int register_io_adapter(struct kvm_device *dev,
1538 struct kvm_device_attr *attr)
1540 struct s390_io_adapter *adapter;
1541 struct kvm_s390_io_adapter adapter_info;
1543 if (copy_from_user(&adapter_info,
1544 (void __user *)attr->addr, sizeof(adapter_info)))
1547 if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1548 (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1551 adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1555 INIT_LIST_HEAD(&adapter->maps);
1556 init_rwsem(&adapter->maps_lock);
1557 atomic_set(&adapter->nr_maps, 0);
1558 adapter->id = adapter_info.id;
1559 adapter->isc = adapter_info.isc;
1560 adapter->maskable = adapter_info.maskable;
1561 adapter->masked = false;
1562 adapter->swap = adapter_info.swap;
1563 dev->kvm->arch.adapters[adapter->id] = adapter;
1568 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1571 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1573 if (!adapter || !adapter->maskable)
1575 ret = adapter->masked;
1576 adapter->masked = masked;
1580 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1582 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1583 struct s390_map_info *map;
1586 if (!adapter || !addr)
1589 map = kzalloc(sizeof(*map), GFP_KERNEL);
1594 INIT_LIST_HEAD(&map->list);
1595 map->guest_addr = addr;
1596 map->addr = gmap_translate(kvm->arch.gmap, addr);
1597 if (map->addr == -EFAULT) {
1601 ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1605 down_write(&adapter->maps_lock);
1606 if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1607 list_add_tail(&map->list, &adapter->maps);
1610 put_page(map->page);
1613 up_write(&adapter->maps_lock);
1620 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1622 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1623 struct s390_map_info *map, *tmp;
1626 if (!adapter || !addr)
1629 down_write(&adapter->maps_lock);
1630 list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1631 if (map->guest_addr == addr) {
1633 atomic_dec(&adapter->nr_maps);
1634 list_del(&map->list);
1635 put_page(map->page);
1640 up_write(&adapter->maps_lock);
1642 return found ? 0 : -EINVAL;
1645 void kvm_s390_destroy_adapters(struct kvm *kvm)
1648 struct s390_map_info *map, *tmp;
1650 for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
1651 if (!kvm->arch.adapters[i])
1653 list_for_each_entry_safe(map, tmp,
1654 &kvm->arch.adapters[i]->maps, list) {
1655 list_del(&map->list);
1656 put_page(map->page);
1659 kfree(kvm->arch.adapters[i]);
1663 static int modify_io_adapter(struct kvm_device *dev,
1664 struct kvm_device_attr *attr)
1666 struct kvm_s390_io_adapter_req req;
1667 struct s390_io_adapter *adapter;
1670 if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
1673 adapter = get_io_adapter(dev->kvm, req.id);
1677 case KVM_S390_IO_ADAPTER_MASK:
1678 ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
1682 case KVM_S390_IO_ADAPTER_MAP:
1683 ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
1685 case KVM_S390_IO_ADAPTER_UNMAP:
1686 ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
1695 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1699 struct kvm_vcpu *vcpu;
1701 switch (attr->group) {
1702 case KVM_DEV_FLIC_ENQUEUE:
1703 r = enqueue_floating_irq(dev, attr);
1705 case KVM_DEV_FLIC_CLEAR_IRQS:
1706 kvm_s390_clear_float_irqs(dev->kvm);
1708 case KVM_DEV_FLIC_APF_ENABLE:
1709 dev->kvm->arch.gmap->pfault_enabled = 1;
1711 case KVM_DEV_FLIC_APF_DISABLE_WAIT:
1712 dev->kvm->arch.gmap->pfault_enabled = 0;
1714 * Make sure no async faults are in transition when
1715 * clearing the queues. So we don't need to worry
1716 * about late coming workers.
1718 synchronize_srcu(&dev->kvm->srcu);
1719 kvm_for_each_vcpu(i, vcpu, dev->kvm)
1720 kvm_clear_async_pf_completion_queue(vcpu);
1722 case KVM_DEV_FLIC_ADAPTER_REGISTER:
1723 r = register_io_adapter(dev, attr);
1725 case KVM_DEV_FLIC_ADAPTER_MODIFY:
1726 r = modify_io_adapter(dev, attr);
1735 static int flic_create(struct kvm_device *dev, u32 type)
1739 if (dev->kvm->arch.flic)
1741 dev->kvm->arch.flic = dev;
1745 static void flic_destroy(struct kvm_device *dev)
1747 dev->kvm->arch.flic = NULL;
1751 /* s390 floating irq controller (flic) */
1752 struct kvm_device_ops kvm_flic_ops = {
1754 .get_attr = flic_get_attr,
1755 .set_attr = flic_set_attr,
1756 .create = flic_create,
1757 .destroy = flic_destroy,
1760 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
1764 bit = bit_nr + (addr % PAGE_SIZE) * 8;
1766 return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
1769 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
1772 struct s390_map_info *map;
1777 list_for_each_entry(map, &adapter->maps, list) {
1778 if (map->guest_addr == addr)
1784 static int adapter_indicators_set(struct kvm *kvm,
1785 struct s390_io_adapter *adapter,
1786 struct kvm_s390_adapter_int *adapter_int)
1789 int summary_set, idx;
1790 struct s390_map_info *info;
1793 info = get_map_info(adapter, adapter_int->ind_addr);
1796 map = page_address(info->page);
1797 bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
1799 idx = srcu_read_lock(&kvm->srcu);
1800 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
1801 set_page_dirty_lock(info->page);
1802 info = get_map_info(adapter, adapter_int->summary_addr);
1804 srcu_read_unlock(&kvm->srcu, idx);
1807 map = page_address(info->page);
1808 bit = get_ind_bit(info->addr, adapter_int->summary_offset,
1810 summary_set = test_and_set_bit(bit, map);
1811 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
1812 set_page_dirty_lock(info->page);
1813 srcu_read_unlock(&kvm->srcu, idx);
1814 return summary_set ? 0 : 1;
1818 * < 0 - not injected due to error
1819 * = 0 - coalesced, summary indicator already active
1820 * > 0 - injected interrupt
1822 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
1823 struct kvm *kvm, int irq_source_id, int level,
1827 struct s390_io_adapter *adapter;
1829 /* We're only interested in the 0->1 transition. */
1832 adapter = get_io_adapter(kvm, e->adapter.adapter_id);
1835 down_read(&adapter->maps_lock);
1836 ret = adapter_indicators_set(kvm, adapter, &e->adapter);
1837 up_read(&adapter->maps_lock);
1838 if ((ret > 0) && !adapter->masked) {
1839 struct kvm_s390_interrupt s390int = {
1840 .type = KVM_S390_INT_IO(1, 0, 0, 0),
1842 .parm64 = (adapter->isc << 27) | 0x80000000,
1844 ret = kvm_s390_inject_vm(kvm, &s390int);
1851 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1852 const struct kvm_irq_routing_entry *ue)
1857 case KVM_IRQ_ROUTING_S390_ADAPTER:
1858 e->set = set_adapter_int;
1859 e->adapter.summary_addr = ue->u.adapter.summary_addr;
1860 e->adapter.ind_addr = ue->u.adapter.ind_addr;
1861 e->adapter.summary_offset = ue->u.adapter.summary_offset;
1862 e->adapter.ind_offset = ue->u.adapter.ind_offset;
1863 e->adapter.adapter_id = ue->u.adapter.adapter_id;
1873 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
1874 int irq_source_id, int level, bool line_status)