Linux 4.2.2
[linux/fpc-iii.git] / arch / s390 / kvm / interrupt.c
blobc98d89708e99fb19ee2217c826f75f2bd0e3a41a
1 /*
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>
22 #include <asm/dis.h>
23 #include <asm/uaccess.h>
24 #include <asm/sclp.h>
25 #include <asm/isc.h>
26 #include "kvm-s390.h"
27 #include "gaccess.h"
28 #include "trace-s390.h"
30 #define IOINT_SCHID_MASK 0x0000ffff
31 #define IOINT_SSID_MASK 0x00030000
32 #define IOINT_CSSID_MASK 0x03fc0000
33 #define IOINT_AI_MASK 0x04000000
34 #define PFAULT_INIT 0x0600
35 #define PFAULT_DONE 0x0680
36 #define VIRTIO_PARAM 0x0d00
38 int psw_extint_disabled(struct kvm_vcpu *vcpu)
40 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
43 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
45 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
48 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
50 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
53 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
55 if ((vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PER) ||
56 (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO) ||
57 (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT))
58 return 0;
59 return 1;
62 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
64 if (psw_extint_disabled(vcpu) ||
65 !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
66 return 0;
67 if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
68 /* No timer interrupts when single stepping */
69 return 0;
70 return 1;
73 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
75 if (!(vcpu->arch.sie_block->ckc <
76 get_tod_clock_fast() + vcpu->arch.sie_block->epoch))
77 return 0;
78 return ckc_interrupts_enabled(vcpu);
81 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
83 return !psw_extint_disabled(vcpu) &&
84 (vcpu->arch.sie_block->gcr[0] & 0x400ul);
87 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
89 return (vcpu->arch.sie_block->cputm >> 63) &&
90 cpu_timer_interrupts_enabled(vcpu);
93 static inline int is_ioirq(unsigned long irq_type)
95 return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
96 (irq_type <= IRQ_PEND_IO_ISC_7));
99 static uint64_t isc_to_isc_bits(int isc)
101 return (0x80 >> isc) << 24;
104 static inline u8 int_word_to_isc(u32 int_word)
106 return (int_word & 0x38000000) >> 27;
109 static inline unsigned long pending_floating_irqs(struct kvm_vcpu *vcpu)
111 return vcpu->kvm->arch.float_int.pending_irqs;
114 static inline unsigned long pending_local_irqs(struct kvm_vcpu *vcpu)
116 return vcpu->arch.local_int.pending_irqs;
119 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
120 unsigned long active_mask)
122 int i;
124 for (i = 0; i <= MAX_ISC; i++)
125 if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
126 active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
128 return active_mask;
131 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
133 unsigned long active_mask;
135 active_mask = pending_local_irqs(vcpu);
136 active_mask |= pending_floating_irqs(vcpu);
137 if (!active_mask)
138 return 0;
140 if (psw_extint_disabled(vcpu))
141 active_mask &= ~IRQ_PEND_EXT_MASK;
142 if (psw_ioint_disabled(vcpu))
143 active_mask &= ~IRQ_PEND_IO_MASK;
144 else
145 active_mask = disable_iscs(vcpu, active_mask);
146 if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
147 __clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
148 if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
149 __clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
150 if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
151 __clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
152 if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
153 __clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
154 if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
155 __clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
156 if (psw_mchk_disabled(vcpu))
157 active_mask &= ~IRQ_PEND_MCHK_MASK;
158 if (!(vcpu->arch.sie_block->gcr[14] &
159 vcpu->kvm->arch.float_int.mchk.cr14))
160 __clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
163 * STOP irqs will never be actively delivered. They are triggered via
164 * intercept requests and cleared when the stop intercept is performed.
166 __clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
168 return active_mask;
171 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
173 atomic_set_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
174 set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
177 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
179 atomic_clear_mask(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
180 clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
183 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
185 atomic_clear_mask(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
186 &vcpu->arch.sie_block->cpuflags);
187 vcpu->arch.sie_block->lctl = 0x0000;
188 vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
190 if (guestdbg_enabled(vcpu)) {
191 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
192 LCTL_CR10 | LCTL_CR11);
193 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
197 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
199 atomic_set_mask(flag, &vcpu->arch.sie_block->cpuflags);
202 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
204 if (!(pending_floating_irqs(vcpu) & IRQ_PEND_IO_MASK))
205 return;
206 else if (psw_ioint_disabled(vcpu))
207 __set_cpuflag(vcpu, CPUSTAT_IO_INT);
208 else
209 vcpu->arch.sie_block->lctl |= LCTL_CR6;
212 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
214 if (!(pending_local_irqs(vcpu) & IRQ_PEND_EXT_MASK))
215 return;
216 if (psw_extint_disabled(vcpu))
217 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
218 else
219 vcpu->arch.sie_block->lctl |= LCTL_CR0;
222 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
224 if (!(pending_local_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
225 return;
226 if (psw_mchk_disabled(vcpu))
227 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
228 else
229 vcpu->arch.sie_block->lctl |= LCTL_CR14;
232 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
234 if (kvm_s390_is_stop_irq_pending(vcpu))
235 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
238 /* Set interception request for non-deliverable interrupts */
239 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
241 set_intercept_indicators_io(vcpu);
242 set_intercept_indicators_ext(vcpu);
243 set_intercept_indicators_mchk(vcpu);
244 set_intercept_indicators_stop(vcpu);
247 static u16 get_ilc(struct kvm_vcpu *vcpu)
249 switch (vcpu->arch.sie_block->icptcode) {
250 case ICPT_INST:
251 case ICPT_INSTPROGI:
252 case ICPT_OPEREXC:
253 case ICPT_PARTEXEC:
254 case ICPT_IOINST:
255 /* last instruction only stored for these icptcodes */
256 return insn_length(vcpu->arch.sie_block->ipa >> 8);
257 case ICPT_PROGI:
258 return vcpu->arch.sie_block->pgmilc;
259 default:
260 return 0;
264 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
266 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
267 int rc;
269 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
270 0, 0);
272 rc = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
273 (u16 *)__LC_EXT_INT_CODE);
274 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
275 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
276 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
277 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
278 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
279 clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
280 return rc ? -EFAULT : 0;
283 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
285 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
286 int rc;
288 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
289 0, 0);
291 rc = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
292 (u16 __user *)__LC_EXT_INT_CODE);
293 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
294 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
295 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
296 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
297 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
298 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
299 return rc ? -EFAULT : 0;
302 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
304 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
305 struct kvm_s390_ext_info ext;
306 int rc;
308 spin_lock(&li->lock);
309 ext = li->irq.ext;
310 clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
311 li->irq.ext.ext_params2 = 0;
312 spin_unlock(&li->lock);
314 VCPU_EVENT(vcpu, 4, "interrupt: pfault init parm:%x,parm64:%llx",
315 0, ext.ext_params2);
316 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
317 KVM_S390_INT_PFAULT_INIT,
318 0, ext.ext_params2);
320 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
321 rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
322 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
323 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
324 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
325 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
326 rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
327 return rc ? -EFAULT : 0;
330 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
332 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
333 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
334 struct kvm_s390_mchk_info mchk = {};
335 unsigned long adtl_status_addr;
336 int deliver = 0;
337 int rc = 0;
339 spin_lock(&fi->lock);
340 spin_lock(&li->lock);
341 if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
342 test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
344 * If there was an exigent machine check pending, then any
345 * repressible machine checks that might have been pending
346 * are indicated along with it, so always clear bits for
347 * repressible and exigent interrupts
349 mchk = li->irq.mchk;
350 clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
351 clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
352 memset(&li->irq.mchk, 0, sizeof(mchk));
353 deliver = 1;
356 * We indicate floating repressible conditions along with
357 * other pending conditions. Channel Report Pending and Channel
358 * Subsystem damage are the only two and and are indicated by
359 * bits in mcic and masked in cr14.
361 if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
362 mchk.mcic |= fi->mchk.mcic;
363 mchk.cr14 |= fi->mchk.cr14;
364 memset(&fi->mchk, 0, sizeof(mchk));
365 deliver = 1;
367 spin_unlock(&li->lock);
368 spin_unlock(&fi->lock);
370 if (deliver) {
371 VCPU_EVENT(vcpu, 4, "interrupt: machine check mcic=%llx",
372 mchk.mcic);
373 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
374 KVM_S390_MCHK,
375 mchk.cr14, mchk.mcic);
377 rc = kvm_s390_vcpu_store_status(vcpu,
378 KVM_S390_STORE_STATUS_PREFIXED);
379 rc |= read_guest_lc(vcpu, __LC_VX_SAVE_AREA_ADDR,
380 &adtl_status_addr,
381 sizeof(unsigned long));
382 rc |= kvm_s390_vcpu_store_adtl_status(vcpu,
383 adtl_status_addr);
384 rc |= put_guest_lc(vcpu, mchk.mcic,
385 (u64 __user *) __LC_MCCK_CODE);
386 rc |= put_guest_lc(vcpu, mchk.failing_storage_address,
387 (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
388 rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA,
389 &mchk.fixed_logout,
390 sizeof(mchk.fixed_logout));
391 rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
392 &vcpu->arch.sie_block->gpsw,
393 sizeof(psw_t));
394 rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
395 &vcpu->arch.sie_block->gpsw,
396 sizeof(psw_t));
398 return rc ? -EFAULT : 0;
401 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
403 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
404 int rc;
406 VCPU_EVENT(vcpu, 4, "%s", "interrupt: cpu restart");
407 vcpu->stat.deliver_restart_signal++;
408 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
410 rc = write_guest_lc(vcpu,
411 offsetof(struct _lowcore, restart_old_psw),
412 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
413 rc |= read_guest_lc(vcpu, offsetof(struct _lowcore, restart_psw),
414 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
415 clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
416 return rc ? -EFAULT : 0;
419 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
421 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
422 struct kvm_s390_prefix_info prefix;
424 spin_lock(&li->lock);
425 prefix = li->irq.prefix;
426 li->irq.prefix.address = 0;
427 clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
428 spin_unlock(&li->lock);
430 VCPU_EVENT(vcpu, 4, "interrupt: set prefix to %x", prefix.address);
431 vcpu->stat.deliver_prefix_signal++;
432 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
433 KVM_S390_SIGP_SET_PREFIX,
434 prefix.address, 0);
436 kvm_s390_set_prefix(vcpu, prefix.address);
437 return 0;
440 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
442 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
443 int rc;
444 int cpu_addr;
446 spin_lock(&li->lock);
447 cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
448 clear_bit(cpu_addr, li->sigp_emerg_pending);
449 if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
450 clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
451 spin_unlock(&li->lock);
453 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp emerg");
454 vcpu->stat.deliver_emergency_signal++;
455 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
456 cpu_addr, 0);
458 rc = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
459 (u16 *)__LC_EXT_INT_CODE);
460 rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
461 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
462 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
463 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
464 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
465 return rc ? -EFAULT : 0;
468 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
470 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
471 struct kvm_s390_extcall_info extcall;
472 int rc;
474 spin_lock(&li->lock);
475 extcall = li->irq.extcall;
476 li->irq.extcall.code = 0;
477 clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
478 spin_unlock(&li->lock);
480 VCPU_EVENT(vcpu, 4, "%s", "interrupt: sigp ext call");
481 vcpu->stat.deliver_external_call++;
482 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
483 KVM_S390_INT_EXTERNAL_CALL,
484 extcall.code, 0);
486 rc = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
487 (u16 *)__LC_EXT_INT_CODE);
488 rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
489 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
490 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
491 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
492 sizeof(psw_t));
493 return rc ? -EFAULT : 0;
496 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
498 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
499 struct kvm_s390_pgm_info pgm_info;
500 int rc = 0, nullifying = false;
501 u16 ilc = get_ilc(vcpu);
503 spin_lock(&li->lock);
504 pgm_info = li->irq.pgm;
505 clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
506 memset(&li->irq.pgm, 0, sizeof(pgm_info));
507 spin_unlock(&li->lock);
509 VCPU_EVENT(vcpu, 4, "interrupt: pgm check code:%x, ilc:%x",
510 pgm_info.code, ilc);
511 vcpu->stat.deliver_program_int++;
512 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
513 pgm_info.code, 0);
515 switch (pgm_info.code & ~PGM_PER) {
516 case PGM_AFX_TRANSLATION:
517 case PGM_ASX_TRANSLATION:
518 case PGM_EX_TRANSLATION:
519 case PGM_LFX_TRANSLATION:
520 case PGM_LSTE_SEQUENCE:
521 case PGM_LSX_TRANSLATION:
522 case PGM_LX_TRANSLATION:
523 case PGM_PRIMARY_AUTHORITY:
524 case PGM_SECONDARY_AUTHORITY:
525 nullifying = true;
526 /* fall through */
527 case PGM_SPACE_SWITCH:
528 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
529 (u64 *)__LC_TRANS_EXC_CODE);
530 break;
531 case PGM_ALEN_TRANSLATION:
532 case PGM_ALE_SEQUENCE:
533 case PGM_ASTE_INSTANCE:
534 case PGM_ASTE_SEQUENCE:
535 case PGM_ASTE_VALIDITY:
536 case PGM_EXTENDED_AUTHORITY:
537 rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
538 (u8 *)__LC_EXC_ACCESS_ID);
539 nullifying = true;
540 break;
541 case PGM_ASCE_TYPE:
542 case PGM_PAGE_TRANSLATION:
543 case PGM_REGION_FIRST_TRANS:
544 case PGM_REGION_SECOND_TRANS:
545 case PGM_REGION_THIRD_TRANS:
546 case PGM_SEGMENT_TRANSLATION:
547 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
548 (u64 *)__LC_TRANS_EXC_CODE);
549 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
550 (u8 *)__LC_EXC_ACCESS_ID);
551 rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
552 (u8 *)__LC_OP_ACCESS_ID);
553 nullifying = true;
554 break;
555 case PGM_MONITOR:
556 rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
557 (u16 *)__LC_MON_CLASS_NR);
558 rc |= put_guest_lc(vcpu, pgm_info.mon_code,
559 (u64 *)__LC_MON_CODE);
560 break;
561 case PGM_VECTOR_PROCESSING:
562 case PGM_DATA:
563 rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
564 (u32 *)__LC_DATA_EXC_CODE);
565 break;
566 case PGM_PROTECTION:
567 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
568 (u64 *)__LC_TRANS_EXC_CODE);
569 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
570 (u8 *)__LC_EXC_ACCESS_ID);
571 break;
572 case PGM_STACK_FULL:
573 case PGM_STACK_EMPTY:
574 case PGM_STACK_SPECIFICATION:
575 case PGM_STACK_TYPE:
576 case PGM_STACK_OPERATION:
577 case PGM_TRACE_TABEL:
578 case PGM_CRYPTO_OPERATION:
579 nullifying = true;
580 break;
583 if (pgm_info.code & PGM_PER) {
584 rc |= put_guest_lc(vcpu, pgm_info.per_code,
585 (u8 *) __LC_PER_CODE);
586 rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
587 (u8 *)__LC_PER_ATMID);
588 rc |= put_guest_lc(vcpu, pgm_info.per_address,
589 (u64 *) __LC_PER_ADDRESS);
590 rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
591 (u8 *) __LC_PER_ACCESS_ID);
594 if (nullifying && vcpu->arch.sie_block->icptcode == ICPT_INST)
595 kvm_s390_rewind_psw(vcpu, ilc);
597 rc |= put_guest_lc(vcpu, ilc, (u16 *) __LC_PGM_ILC);
598 rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
599 (u64 *) __LC_LAST_BREAK);
600 rc |= put_guest_lc(vcpu, pgm_info.code,
601 (u16 *)__LC_PGM_INT_CODE);
602 rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
603 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
604 rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
605 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
606 return rc ? -EFAULT : 0;
609 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
611 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
612 struct kvm_s390_ext_info ext;
613 int rc = 0;
615 spin_lock(&fi->lock);
616 if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
617 spin_unlock(&fi->lock);
618 return 0;
620 ext = fi->srv_signal;
621 memset(&fi->srv_signal, 0, sizeof(ext));
622 clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
623 spin_unlock(&fi->lock);
625 VCPU_EVENT(vcpu, 4, "interrupt: sclp parm:%x",
626 ext.ext_params);
627 vcpu->stat.deliver_service_signal++;
628 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
629 ext.ext_params, 0);
631 rc = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
632 rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
633 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
634 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
635 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
636 &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
637 rc |= put_guest_lc(vcpu, ext.ext_params,
638 (u32 *)__LC_EXT_PARAMS);
640 return rc ? -EFAULT : 0;
643 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
645 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
646 struct kvm_s390_interrupt_info *inti;
647 int rc = 0;
649 spin_lock(&fi->lock);
650 inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
651 struct kvm_s390_interrupt_info,
652 list);
653 if (inti) {
654 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
655 KVM_S390_INT_PFAULT_DONE, 0,
656 inti->ext.ext_params2);
657 list_del(&inti->list);
658 fi->counters[FIRQ_CNTR_PFAULT] -= 1;
660 if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
661 clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
662 spin_unlock(&fi->lock);
664 if (inti) {
665 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
666 (u16 *)__LC_EXT_INT_CODE);
667 rc |= put_guest_lc(vcpu, PFAULT_DONE,
668 (u16 *)__LC_EXT_CPU_ADDR);
669 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
670 &vcpu->arch.sie_block->gpsw,
671 sizeof(psw_t));
672 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
673 &vcpu->arch.sie_block->gpsw,
674 sizeof(psw_t));
675 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
676 (u64 *)__LC_EXT_PARAMS2);
677 kfree(inti);
679 return rc ? -EFAULT : 0;
682 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
684 struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
685 struct kvm_s390_interrupt_info *inti;
686 int rc = 0;
688 spin_lock(&fi->lock);
689 inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
690 struct kvm_s390_interrupt_info,
691 list);
692 if (inti) {
693 VCPU_EVENT(vcpu, 4,
694 "interrupt: virtio parm:%x,parm64:%llx",
695 inti->ext.ext_params, inti->ext.ext_params2);
696 vcpu->stat.deliver_virtio_interrupt++;
697 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
698 inti->type,
699 inti->ext.ext_params,
700 inti->ext.ext_params2);
701 list_del(&inti->list);
702 fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
704 if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
705 clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
706 spin_unlock(&fi->lock);
708 if (inti) {
709 rc = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
710 (u16 *)__LC_EXT_INT_CODE);
711 rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
712 (u16 *)__LC_EXT_CPU_ADDR);
713 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
714 &vcpu->arch.sie_block->gpsw,
715 sizeof(psw_t));
716 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
717 &vcpu->arch.sie_block->gpsw,
718 sizeof(psw_t));
719 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
720 (u32 *)__LC_EXT_PARAMS);
721 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
722 (u64 *)__LC_EXT_PARAMS2);
723 kfree(inti);
725 return rc ? -EFAULT : 0;
728 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
729 unsigned long irq_type)
731 struct list_head *isc_list;
732 struct kvm_s390_float_interrupt *fi;
733 struct kvm_s390_interrupt_info *inti = NULL;
734 int rc = 0;
736 fi = &vcpu->kvm->arch.float_int;
738 spin_lock(&fi->lock);
739 isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
740 inti = list_first_entry_or_null(isc_list,
741 struct kvm_s390_interrupt_info,
742 list);
743 if (inti) {
744 VCPU_EVENT(vcpu, 4, "interrupt: I/O %llx", inti->type);
745 vcpu->stat.deliver_io_int++;
746 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
747 inti->type,
748 ((__u32)inti->io.subchannel_id << 16) |
749 inti->io.subchannel_nr,
750 ((__u64)inti->io.io_int_parm << 32) |
751 inti->io.io_int_word);
752 list_del(&inti->list);
753 fi->counters[FIRQ_CNTR_IO] -= 1;
755 if (list_empty(isc_list))
756 clear_bit(irq_type, &fi->pending_irqs);
757 spin_unlock(&fi->lock);
759 if (inti) {
760 rc = put_guest_lc(vcpu, inti->io.subchannel_id,
761 (u16 *)__LC_SUBCHANNEL_ID);
762 rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
763 (u16 *)__LC_SUBCHANNEL_NR);
764 rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
765 (u32 *)__LC_IO_INT_PARM);
766 rc |= put_guest_lc(vcpu, inti->io.io_int_word,
767 (u32 *)__LC_IO_INT_WORD);
768 rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
769 &vcpu->arch.sie_block->gpsw,
770 sizeof(psw_t));
771 rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
772 &vcpu->arch.sie_block->gpsw,
773 sizeof(psw_t));
774 kfree(inti);
777 return rc ? -EFAULT : 0;
780 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
782 static const deliver_irq_t deliver_irq_funcs[] = {
783 [IRQ_PEND_MCHK_EX] = __deliver_machine_check,
784 [IRQ_PEND_MCHK_REP] = __deliver_machine_check,
785 [IRQ_PEND_PROG] = __deliver_prog,
786 [IRQ_PEND_EXT_EMERGENCY] = __deliver_emergency_signal,
787 [IRQ_PEND_EXT_EXTERNAL] = __deliver_external_call,
788 [IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
789 [IRQ_PEND_EXT_CPU_TIMER] = __deliver_cpu_timer,
790 [IRQ_PEND_RESTART] = __deliver_restart,
791 [IRQ_PEND_SET_PREFIX] = __deliver_set_prefix,
792 [IRQ_PEND_PFAULT_INIT] = __deliver_pfault_init,
793 [IRQ_PEND_EXT_SERVICE] = __deliver_service,
794 [IRQ_PEND_PFAULT_DONE] = __deliver_pfault_done,
795 [IRQ_PEND_VIRTIO] = __deliver_virtio,
798 /* Check whether an external call is pending (deliverable or not) */
799 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
801 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
802 uint8_t sigp_ctrl = vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
804 if (!sclp.has_sigpif)
805 return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
807 return (sigp_ctrl & SIGP_CTRL_C) &&
808 (atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND);
811 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
813 int rc;
815 rc = !!deliverable_irqs(vcpu);
817 if (!rc && kvm_cpu_has_pending_timer(vcpu))
818 rc = 1;
820 /* external call pending and deliverable */
821 if (!rc && kvm_s390_ext_call_pending(vcpu) &&
822 !psw_extint_disabled(vcpu) &&
823 (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
824 rc = 1;
826 if (!rc && !exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
827 rc = 1;
829 return rc;
832 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
834 return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
837 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
839 u64 now, sltime;
841 vcpu->stat.exit_wait_state++;
843 /* fast path */
844 if (kvm_cpu_has_pending_timer(vcpu) || kvm_arch_vcpu_runnable(vcpu))
845 return 0;
847 if (psw_interrupts_disabled(vcpu)) {
848 VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
849 return -EOPNOTSUPP; /* disabled wait */
852 if (!ckc_interrupts_enabled(vcpu)) {
853 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
854 __set_cpu_idle(vcpu);
855 goto no_timer;
858 now = get_tod_clock_fast() + vcpu->arch.sie_block->epoch;
859 sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
861 /* underflow */
862 if (vcpu->arch.sie_block->ckc < now)
863 return 0;
865 __set_cpu_idle(vcpu);
866 hrtimer_start(&vcpu->arch.ckc_timer, ktime_set (0, sltime) , HRTIMER_MODE_REL);
867 VCPU_EVENT(vcpu, 5, "enabled wait via clock comparator: %llx ns", sltime);
868 no_timer:
869 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
870 kvm_vcpu_block(vcpu);
871 __unset_cpu_idle(vcpu);
872 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
874 hrtimer_cancel(&vcpu->arch.ckc_timer);
875 return 0;
878 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
880 if (waitqueue_active(&vcpu->wq)) {
882 * The vcpu gave up the cpu voluntarily, mark it as a good
883 * yield-candidate.
885 vcpu->preempted = true;
886 wake_up_interruptible(&vcpu->wq);
887 vcpu->stat.halt_wakeup++;
891 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
893 struct kvm_vcpu *vcpu;
894 u64 now, sltime;
896 vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
897 now = get_tod_clock_fast() + vcpu->arch.sie_block->epoch;
898 sltime = tod_to_ns(vcpu->arch.sie_block->ckc - now);
901 * If the monotonic clock runs faster than the tod clock we might be
902 * woken up too early and have to go back to sleep to avoid deadlocks.
904 if (vcpu->arch.sie_block->ckc > now &&
905 hrtimer_forward_now(timer, ns_to_ktime(sltime)))
906 return HRTIMER_RESTART;
907 kvm_s390_vcpu_wakeup(vcpu);
908 return HRTIMER_NORESTART;
911 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
913 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
915 spin_lock(&li->lock);
916 li->pending_irqs = 0;
917 bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
918 memset(&li->irq, 0, sizeof(li->irq));
919 spin_unlock(&li->lock);
921 /* clear pending external calls set by sigp interpretation facility */
922 atomic_clear_mask(CPUSTAT_ECALL_PEND, li->cpuflags);
923 vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl = 0;
926 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
928 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
929 deliver_irq_t func;
930 int rc = 0;
931 unsigned long irq_type;
932 unsigned long irqs;
934 __reset_intercept_indicators(vcpu);
936 /* pending ckc conditions might have been invalidated */
937 clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
938 if (ckc_irq_pending(vcpu))
939 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
941 /* pending cpu timer conditions might have been invalidated */
942 clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
943 if (cpu_timer_irq_pending(vcpu))
944 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
946 while ((irqs = deliverable_irqs(vcpu)) && !rc) {
947 /* bits are in the order of interrupt priority */
948 irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
949 if (is_ioirq(irq_type)) {
950 rc = __deliver_io(vcpu, irq_type);
951 } else {
952 func = deliver_irq_funcs[irq_type];
953 if (!func) {
954 WARN_ON_ONCE(func == NULL);
955 clear_bit(irq_type, &li->pending_irqs);
956 continue;
958 rc = func(vcpu);
962 set_intercept_indicators(vcpu);
964 return rc;
967 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
969 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
971 li->irq.pgm = irq->u.pgm;
972 set_bit(IRQ_PEND_PROG, &li->pending_irqs);
973 return 0;
976 int kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code)
978 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
979 struct kvm_s390_irq irq;
981 VCPU_EVENT(vcpu, 3, "inject: program check %d (from kernel)", code);
982 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT, code,
983 0, 1);
984 spin_lock(&li->lock);
985 irq.u.pgm.code = code;
986 __inject_prog(vcpu, &irq);
987 BUG_ON(waitqueue_active(li->wq));
988 spin_unlock(&li->lock);
989 return 0;
992 int kvm_s390_inject_prog_irq(struct kvm_vcpu *vcpu,
993 struct kvm_s390_pgm_info *pgm_info)
995 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
996 struct kvm_s390_irq irq;
997 int rc;
999 VCPU_EVENT(vcpu, 3, "inject: prog irq %d (from kernel)",
1000 pgm_info->code);
1001 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1002 pgm_info->code, 0, 1);
1003 spin_lock(&li->lock);
1004 irq.u.pgm = *pgm_info;
1005 rc = __inject_prog(vcpu, &irq);
1006 BUG_ON(waitqueue_active(li->wq));
1007 spin_unlock(&li->lock);
1008 return rc;
1011 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1013 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1015 VCPU_EVENT(vcpu, 3, "inject: external irq params:%x, params2:%llx",
1016 irq->u.ext.ext_params, irq->u.ext.ext_params2);
1017 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1018 irq->u.ext.ext_params,
1019 irq->u.ext.ext_params2, 2);
1021 li->irq.ext = irq->u.ext;
1022 set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1023 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1024 return 0;
1027 static int __inject_extcall_sigpif(struct kvm_vcpu *vcpu, uint16_t src_id)
1029 unsigned char new_val, old_val;
1030 uint8_t *sigp_ctrl = &vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
1032 new_val = SIGP_CTRL_C | (src_id & SIGP_CTRL_SCN_MASK);
1033 old_val = *sigp_ctrl & ~SIGP_CTRL_C;
1034 if (cmpxchg(sigp_ctrl, old_val, new_val) != old_val) {
1035 /* another external call is pending */
1036 return -EBUSY;
1038 atomic_set_mask(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
1039 return 0;
1042 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1044 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1045 struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1046 uint16_t src_id = irq->u.extcall.code;
1048 VCPU_EVENT(vcpu, 3, "inject: external call source-cpu:%u",
1049 src_id);
1050 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1051 src_id, 0, 2);
1053 /* sending vcpu invalid */
1054 if (src_id >= KVM_MAX_VCPUS ||
1055 kvm_get_vcpu(vcpu->kvm, src_id) == NULL)
1056 return -EINVAL;
1058 if (sclp.has_sigpif)
1059 return __inject_extcall_sigpif(vcpu, src_id);
1061 if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1062 return -EBUSY;
1063 *extcall = irq->u.extcall;
1064 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1065 return 0;
1068 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1070 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1071 struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1073 VCPU_EVENT(vcpu, 3, "inject: set prefix to %x (from user)",
1074 irq->u.prefix.address);
1075 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1076 irq->u.prefix.address, 0, 2);
1078 if (!is_vcpu_stopped(vcpu))
1079 return -EBUSY;
1081 *prefix = irq->u.prefix;
1082 set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1083 return 0;
1086 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1087 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1089 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1090 struct kvm_s390_stop_info *stop = &li->irq.stop;
1091 int rc = 0;
1093 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0, 2);
1095 if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1096 return -EINVAL;
1098 if (is_vcpu_stopped(vcpu)) {
1099 if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1100 rc = kvm_s390_store_status_unloaded(vcpu,
1101 KVM_S390_STORE_STATUS_NOADDR);
1102 return rc;
1105 if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1106 return -EBUSY;
1107 stop->flags = irq->u.stop.flags;
1108 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1109 return 0;
1112 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1113 struct kvm_s390_irq *irq)
1115 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1117 VCPU_EVENT(vcpu, 3, "inject: restart type %llx", irq->type);
1118 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0, 2);
1120 set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1121 return 0;
1124 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1125 struct kvm_s390_irq *irq)
1127 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1129 VCPU_EVENT(vcpu, 3, "inject: emergency %u\n",
1130 irq->u.emerg.code);
1131 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1132 irq->u.emerg.code, 0, 2);
1134 set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1135 set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1136 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1137 return 0;
1140 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1142 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1143 struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1145 VCPU_EVENT(vcpu, 5, "inject: machine check parm64:%llx",
1146 irq->u.mchk.mcic);
1147 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1148 irq->u.mchk.mcic, 2);
1151 * Because repressible machine checks can be indicated along with
1152 * exigent machine checks (PoP, Chapter 11, Interruption action)
1153 * we need to combine cr14, mcic and external damage code.
1154 * Failing storage address and the logout area should not be or'ed
1155 * together, we just indicate the last occurrence of the corresponding
1156 * machine check
1158 mchk->cr14 |= irq->u.mchk.cr14;
1159 mchk->mcic |= irq->u.mchk.mcic;
1160 mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1161 mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1162 memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1163 sizeof(mchk->fixed_logout));
1164 if (mchk->mcic & MCHK_EX_MASK)
1165 set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1166 else if (mchk->mcic & MCHK_REP_MASK)
1167 set_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
1168 return 0;
1171 static int __inject_ckc(struct kvm_vcpu *vcpu)
1173 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1175 VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CLOCK_COMP);
1176 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1177 0, 0, 2);
1179 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1180 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1181 return 0;
1184 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1186 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1188 VCPU_EVENT(vcpu, 3, "inject: type %x", KVM_S390_INT_CPU_TIMER);
1189 trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1190 0, 0, 2);
1192 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1193 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1194 return 0;
1197 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1198 int isc, u32 schid)
1200 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1201 struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1202 struct kvm_s390_interrupt_info *iter;
1203 u16 id = (schid & 0xffff0000U) >> 16;
1204 u16 nr = schid & 0x0000ffffU;
1206 spin_lock(&fi->lock);
1207 list_for_each_entry(iter, isc_list, list) {
1208 if (schid && (id != iter->io.subchannel_id ||
1209 nr != iter->io.subchannel_nr))
1210 continue;
1211 /* found an appropriate entry */
1212 list_del_init(&iter->list);
1213 fi->counters[FIRQ_CNTR_IO] -= 1;
1214 if (list_empty(isc_list))
1215 clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1216 spin_unlock(&fi->lock);
1217 return iter;
1219 spin_unlock(&fi->lock);
1220 return NULL;
1224 * Dequeue and return an I/O interrupt matching any of the interruption
1225 * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1227 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1228 u64 isc_mask, u32 schid)
1230 struct kvm_s390_interrupt_info *inti = NULL;
1231 int isc;
1233 for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1234 if (isc_mask & isc_to_isc_bits(isc))
1235 inti = get_io_int(kvm, isc, schid);
1237 return inti;
1240 #define SCCB_MASK 0xFFFFFFF8
1241 #define SCCB_EVENT_PENDING 0x3
1243 static int __inject_service(struct kvm *kvm,
1244 struct kvm_s390_interrupt_info *inti)
1246 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1248 spin_lock(&fi->lock);
1249 fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1251 * Early versions of the QEMU s390 bios will inject several
1252 * service interrupts after another without handling a
1253 * condition code indicating busy.
1254 * We will silently ignore those superfluous sccb values.
1255 * A future version of QEMU will take care of serialization
1256 * of servc requests
1258 if (fi->srv_signal.ext_params & SCCB_MASK)
1259 goto out;
1260 fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1261 set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1262 out:
1263 spin_unlock(&fi->lock);
1264 kfree(inti);
1265 return 0;
1268 static int __inject_virtio(struct kvm *kvm,
1269 struct kvm_s390_interrupt_info *inti)
1271 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1273 spin_lock(&fi->lock);
1274 if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1275 spin_unlock(&fi->lock);
1276 return -EBUSY;
1278 fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1279 list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1280 set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1281 spin_unlock(&fi->lock);
1282 return 0;
1285 static int __inject_pfault_done(struct kvm *kvm,
1286 struct kvm_s390_interrupt_info *inti)
1288 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1290 spin_lock(&fi->lock);
1291 if (fi->counters[FIRQ_CNTR_PFAULT] >=
1292 (ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1293 spin_unlock(&fi->lock);
1294 return -EBUSY;
1296 fi->counters[FIRQ_CNTR_PFAULT] += 1;
1297 list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1298 set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1299 spin_unlock(&fi->lock);
1300 return 0;
1303 #define CR_PENDING_SUBCLASS 28
1304 static int __inject_float_mchk(struct kvm *kvm,
1305 struct kvm_s390_interrupt_info *inti)
1307 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1309 spin_lock(&fi->lock);
1310 fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1311 fi->mchk.mcic |= inti->mchk.mcic;
1312 set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1313 spin_unlock(&fi->lock);
1314 kfree(inti);
1315 return 0;
1318 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1320 struct kvm_s390_float_interrupt *fi;
1321 struct list_head *list;
1322 int isc;
1324 fi = &kvm->arch.float_int;
1325 spin_lock(&fi->lock);
1326 if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1327 spin_unlock(&fi->lock);
1328 return -EBUSY;
1330 fi->counters[FIRQ_CNTR_IO] += 1;
1332 isc = int_word_to_isc(inti->io.io_int_word);
1333 list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1334 list_add_tail(&inti->list, list);
1335 set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1336 spin_unlock(&fi->lock);
1337 return 0;
1341 * Find a destination VCPU for a floating irq and kick it.
1343 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1345 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1346 struct kvm_s390_local_interrupt *li;
1347 struct kvm_vcpu *dst_vcpu;
1348 int sigcpu, online_vcpus, nr_tries = 0;
1350 online_vcpus = atomic_read(&kvm->online_vcpus);
1351 if (!online_vcpus)
1352 return;
1354 /* find idle VCPUs first, then round robin */
1355 sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1356 if (sigcpu == online_vcpus) {
1357 do {
1358 sigcpu = fi->next_rr_cpu;
1359 fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1360 /* avoid endless loops if all vcpus are stopped */
1361 if (nr_tries++ >= online_vcpus)
1362 return;
1363 } while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1365 dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1367 /* make the VCPU drop out of the SIE, or wake it up if sleeping */
1368 li = &dst_vcpu->arch.local_int;
1369 spin_lock(&li->lock);
1370 switch (type) {
1371 case KVM_S390_MCHK:
1372 atomic_set_mask(CPUSTAT_STOP_INT, li->cpuflags);
1373 break;
1374 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1375 atomic_set_mask(CPUSTAT_IO_INT, li->cpuflags);
1376 break;
1377 default:
1378 atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
1379 break;
1381 spin_unlock(&li->lock);
1382 kvm_s390_vcpu_wakeup(dst_vcpu);
1385 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1387 struct kvm_s390_float_interrupt *fi;
1388 u64 type = READ_ONCE(inti->type);
1389 int rc;
1391 fi = &kvm->arch.float_int;
1393 switch (type) {
1394 case KVM_S390_MCHK:
1395 rc = __inject_float_mchk(kvm, inti);
1396 break;
1397 case KVM_S390_INT_VIRTIO:
1398 rc = __inject_virtio(kvm, inti);
1399 break;
1400 case KVM_S390_INT_SERVICE:
1401 rc = __inject_service(kvm, inti);
1402 break;
1403 case KVM_S390_INT_PFAULT_DONE:
1404 rc = __inject_pfault_done(kvm, inti);
1405 break;
1406 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1407 rc = __inject_io(kvm, inti);
1408 break;
1409 default:
1410 rc = -EINVAL;
1412 if (rc)
1413 return rc;
1415 __floating_irq_kick(kvm, type);
1416 return 0;
1419 int kvm_s390_inject_vm(struct kvm *kvm,
1420 struct kvm_s390_interrupt *s390int)
1422 struct kvm_s390_interrupt_info *inti;
1423 int rc;
1425 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1426 if (!inti)
1427 return -ENOMEM;
1429 inti->type = s390int->type;
1430 switch (inti->type) {
1431 case KVM_S390_INT_VIRTIO:
1432 VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1433 s390int->parm, s390int->parm64);
1434 inti->ext.ext_params = s390int->parm;
1435 inti->ext.ext_params2 = s390int->parm64;
1436 break;
1437 case KVM_S390_INT_SERVICE:
1438 VM_EVENT(kvm, 5, "inject: sclp parm:%x", s390int->parm);
1439 inti->ext.ext_params = s390int->parm;
1440 break;
1441 case KVM_S390_INT_PFAULT_DONE:
1442 inti->ext.ext_params2 = s390int->parm64;
1443 break;
1444 case KVM_S390_MCHK:
1445 VM_EVENT(kvm, 5, "inject: machine check parm64:%llx",
1446 s390int->parm64);
1447 inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1448 inti->mchk.mcic = s390int->parm64;
1449 break;
1450 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1451 if (inti->type & IOINT_AI_MASK)
1452 VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
1453 else
1454 VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
1455 s390int->type & IOINT_CSSID_MASK,
1456 s390int->type & IOINT_SSID_MASK,
1457 s390int->type & IOINT_SCHID_MASK);
1458 inti->io.subchannel_id = s390int->parm >> 16;
1459 inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1460 inti->io.io_int_parm = s390int->parm64 >> 32;
1461 inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1462 break;
1463 default:
1464 kfree(inti);
1465 return -EINVAL;
1467 trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1470 rc = __inject_vm(kvm, inti);
1471 if (rc)
1472 kfree(inti);
1473 return rc;
1476 int kvm_s390_reinject_io_int(struct kvm *kvm,
1477 struct kvm_s390_interrupt_info *inti)
1479 return __inject_vm(kvm, inti);
1482 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1483 struct kvm_s390_irq *irq)
1485 irq->type = s390int->type;
1486 switch (irq->type) {
1487 case KVM_S390_PROGRAM_INT:
1488 if (s390int->parm & 0xffff0000)
1489 return -EINVAL;
1490 irq->u.pgm.code = s390int->parm;
1491 break;
1492 case KVM_S390_SIGP_SET_PREFIX:
1493 irq->u.prefix.address = s390int->parm;
1494 break;
1495 case KVM_S390_SIGP_STOP:
1496 irq->u.stop.flags = s390int->parm;
1497 break;
1498 case KVM_S390_INT_EXTERNAL_CALL:
1499 if (s390int->parm & 0xffff0000)
1500 return -EINVAL;
1501 irq->u.extcall.code = s390int->parm;
1502 break;
1503 case KVM_S390_INT_EMERGENCY:
1504 if (s390int->parm & 0xffff0000)
1505 return -EINVAL;
1506 irq->u.emerg.code = s390int->parm;
1507 break;
1508 case KVM_S390_MCHK:
1509 irq->u.mchk.mcic = s390int->parm64;
1510 break;
1512 return 0;
1515 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1517 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1519 return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1522 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1524 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1526 spin_lock(&li->lock);
1527 li->irq.stop.flags = 0;
1528 clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1529 spin_unlock(&li->lock);
1532 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1534 int rc;
1536 switch (irq->type) {
1537 case KVM_S390_PROGRAM_INT:
1538 VCPU_EVENT(vcpu, 3, "inject: program check %d (from user)",
1539 irq->u.pgm.code);
1540 rc = __inject_prog(vcpu, irq);
1541 break;
1542 case KVM_S390_SIGP_SET_PREFIX:
1543 rc = __inject_set_prefix(vcpu, irq);
1544 break;
1545 case KVM_S390_SIGP_STOP:
1546 rc = __inject_sigp_stop(vcpu, irq);
1547 break;
1548 case KVM_S390_RESTART:
1549 rc = __inject_sigp_restart(vcpu, irq);
1550 break;
1551 case KVM_S390_INT_CLOCK_COMP:
1552 rc = __inject_ckc(vcpu);
1553 break;
1554 case KVM_S390_INT_CPU_TIMER:
1555 rc = __inject_cpu_timer(vcpu);
1556 break;
1557 case KVM_S390_INT_EXTERNAL_CALL:
1558 rc = __inject_extcall(vcpu, irq);
1559 break;
1560 case KVM_S390_INT_EMERGENCY:
1561 rc = __inject_sigp_emergency(vcpu, irq);
1562 break;
1563 case KVM_S390_MCHK:
1564 rc = __inject_mchk(vcpu, irq);
1565 break;
1566 case KVM_S390_INT_PFAULT_INIT:
1567 rc = __inject_pfault_init(vcpu, irq);
1568 break;
1569 case KVM_S390_INT_VIRTIO:
1570 case KVM_S390_INT_SERVICE:
1571 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1572 default:
1573 rc = -EINVAL;
1576 return rc;
1579 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1581 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1582 int rc;
1584 spin_lock(&li->lock);
1585 rc = do_inject_vcpu(vcpu, irq);
1586 spin_unlock(&li->lock);
1587 if (!rc)
1588 kvm_s390_vcpu_wakeup(vcpu);
1589 return rc;
1592 static inline void clear_irq_list(struct list_head *_list)
1594 struct kvm_s390_interrupt_info *inti, *n;
1596 list_for_each_entry_safe(inti, n, _list, list) {
1597 list_del(&inti->list);
1598 kfree(inti);
1602 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1603 struct kvm_s390_irq *irq)
1605 irq->type = inti->type;
1606 switch (inti->type) {
1607 case KVM_S390_INT_PFAULT_INIT:
1608 case KVM_S390_INT_PFAULT_DONE:
1609 case KVM_S390_INT_VIRTIO:
1610 irq->u.ext = inti->ext;
1611 break;
1612 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1613 irq->u.io = inti->io;
1614 break;
1618 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1620 struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1621 int i;
1623 spin_lock(&fi->lock);
1624 fi->pending_irqs = 0;
1625 memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1626 memset(&fi->mchk, 0, sizeof(fi->mchk));
1627 for (i = 0; i < FIRQ_LIST_COUNT; i++)
1628 clear_irq_list(&fi->lists[i]);
1629 for (i = 0; i < FIRQ_MAX_COUNT; i++)
1630 fi->counters[i] = 0;
1631 spin_unlock(&fi->lock);
1634 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1636 struct kvm_s390_interrupt_info *inti;
1637 struct kvm_s390_float_interrupt *fi;
1638 struct kvm_s390_irq *buf;
1639 struct kvm_s390_irq *irq;
1640 int max_irqs;
1641 int ret = 0;
1642 int n = 0;
1643 int i;
1645 if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1646 return -EINVAL;
1649 * We are already using -ENOMEM to signal
1650 * userspace it may retry with a bigger buffer,
1651 * so we need to use something else for this case
1653 buf = vzalloc(len);
1654 if (!buf)
1655 return -ENOBUFS;
1657 max_irqs = len / sizeof(struct kvm_s390_irq);
1659 fi = &kvm->arch.float_int;
1660 spin_lock(&fi->lock);
1661 for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1662 list_for_each_entry(inti, &fi->lists[i], list) {
1663 if (n == max_irqs) {
1664 /* signal userspace to try again */
1665 ret = -ENOMEM;
1666 goto out;
1668 inti_to_irq(inti, &buf[n]);
1669 n++;
1672 if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1673 if (n == max_irqs) {
1674 /* signal userspace to try again */
1675 ret = -ENOMEM;
1676 goto out;
1678 irq = (struct kvm_s390_irq *) &buf[n];
1679 irq->type = KVM_S390_INT_SERVICE;
1680 irq->u.ext = fi->srv_signal;
1681 n++;
1683 if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1684 if (n == max_irqs) {
1685 /* signal userspace to try again */
1686 ret = -ENOMEM;
1687 goto out;
1689 irq = (struct kvm_s390_irq *) &buf[n];
1690 irq->type = KVM_S390_MCHK;
1691 irq->u.mchk = fi->mchk;
1692 n++;
1695 out:
1696 spin_unlock(&fi->lock);
1697 if (!ret && n > 0) {
1698 if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1699 ret = -EFAULT;
1701 vfree(buf);
1703 return ret < 0 ? ret : n;
1706 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1708 int r;
1710 switch (attr->group) {
1711 case KVM_DEV_FLIC_GET_ALL_IRQS:
1712 r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1713 attr->attr);
1714 break;
1715 default:
1716 r = -EINVAL;
1719 return r;
1722 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1723 u64 addr)
1725 struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1726 void *target = NULL;
1727 void __user *source;
1728 u64 size;
1730 if (get_user(inti->type, (u64 __user *)addr))
1731 return -EFAULT;
1733 switch (inti->type) {
1734 case KVM_S390_INT_PFAULT_INIT:
1735 case KVM_S390_INT_PFAULT_DONE:
1736 case KVM_S390_INT_VIRTIO:
1737 case KVM_S390_INT_SERVICE:
1738 target = (void *) &inti->ext;
1739 source = &uptr->u.ext;
1740 size = sizeof(inti->ext);
1741 break;
1742 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1743 target = (void *) &inti->io;
1744 source = &uptr->u.io;
1745 size = sizeof(inti->io);
1746 break;
1747 case KVM_S390_MCHK:
1748 target = (void *) &inti->mchk;
1749 source = &uptr->u.mchk;
1750 size = sizeof(inti->mchk);
1751 break;
1752 default:
1753 return -EINVAL;
1756 if (copy_from_user(target, source, size))
1757 return -EFAULT;
1759 return 0;
1762 static int enqueue_floating_irq(struct kvm_device *dev,
1763 struct kvm_device_attr *attr)
1765 struct kvm_s390_interrupt_info *inti = NULL;
1766 int r = 0;
1767 int len = attr->attr;
1769 if (len % sizeof(struct kvm_s390_irq) != 0)
1770 return -EINVAL;
1771 else if (len > KVM_S390_FLIC_MAX_BUFFER)
1772 return -EINVAL;
1774 while (len >= sizeof(struct kvm_s390_irq)) {
1775 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1776 if (!inti)
1777 return -ENOMEM;
1779 r = copy_irq_from_user(inti, attr->addr);
1780 if (r) {
1781 kfree(inti);
1782 return r;
1784 r = __inject_vm(dev->kvm, inti);
1785 if (r) {
1786 kfree(inti);
1787 return r;
1789 len -= sizeof(struct kvm_s390_irq);
1790 attr->addr += sizeof(struct kvm_s390_irq);
1793 return r;
1796 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
1798 if (id >= MAX_S390_IO_ADAPTERS)
1799 return NULL;
1800 return kvm->arch.adapters[id];
1803 static int register_io_adapter(struct kvm_device *dev,
1804 struct kvm_device_attr *attr)
1806 struct s390_io_adapter *adapter;
1807 struct kvm_s390_io_adapter adapter_info;
1809 if (copy_from_user(&adapter_info,
1810 (void __user *)attr->addr, sizeof(adapter_info)))
1811 return -EFAULT;
1813 if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
1814 (dev->kvm->arch.adapters[adapter_info.id] != NULL))
1815 return -EINVAL;
1817 adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
1818 if (!adapter)
1819 return -ENOMEM;
1821 INIT_LIST_HEAD(&adapter->maps);
1822 init_rwsem(&adapter->maps_lock);
1823 atomic_set(&adapter->nr_maps, 0);
1824 adapter->id = adapter_info.id;
1825 adapter->isc = adapter_info.isc;
1826 adapter->maskable = adapter_info.maskable;
1827 adapter->masked = false;
1828 adapter->swap = adapter_info.swap;
1829 dev->kvm->arch.adapters[adapter->id] = adapter;
1831 return 0;
1834 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
1836 int ret;
1837 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1839 if (!adapter || !adapter->maskable)
1840 return -EINVAL;
1841 ret = adapter->masked;
1842 adapter->masked = masked;
1843 return ret;
1846 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
1848 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1849 struct s390_map_info *map;
1850 int ret;
1852 if (!adapter || !addr)
1853 return -EINVAL;
1855 map = kzalloc(sizeof(*map), GFP_KERNEL);
1856 if (!map) {
1857 ret = -ENOMEM;
1858 goto out;
1860 INIT_LIST_HEAD(&map->list);
1861 map->guest_addr = addr;
1862 map->addr = gmap_translate(kvm->arch.gmap, addr);
1863 if (map->addr == -EFAULT) {
1864 ret = -EFAULT;
1865 goto out;
1867 ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
1868 if (ret < 0)
1869 goto out;
1870 BUG_ON(ret != 1);
1871 down_write(&adapter->maps_lock);
1872 if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
1873 list_add_tail(&map->list, &adapter->maps);
1874 ret = 0;
1875 } else {
1876 put_page(map->page);
1877 ret = -EINVAL;
1879 up_write(&adapter->maps_lock);
1880 out:
1881 if (ret)
1882 kfree(map);
1883 return ret;
1886 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
1888 struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
1889 struct s390_map_info *map, *tmp;
1890 int found = 0;
1892 if (!adapter || !addr)
1893 return -EINVAL;
1895 down_write(&adapter->maps_lock);
1896 list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
1897 if (map->guest_addr == addr) {
1898 found = 1;
1899 atomic_dec(&adapter->nr_maps);
1900 list_del(&map->list);
1901 put_page(map->page);
1902 kfree(map);
1903 break;
1906 up_write(&adapter->maps_lock);
1908 return found ? 0 : -EINVAL;
1911 void kvm_s390_destroy_adapters(struct kvm *kvm)
1913 int i;
1914 struct s390_map_info *map, *tmp;
1916 for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
1917 if (!kvm->arch.adapters[i])
1918 continue;
1919 list_for_each_entry_safe(map, tmp,
1920 &kvm->arch.adapters[i]->maps, list) {
1921 list_del(&map->list);
1922 put_page(map->page);
1923 kfree(map);
1925 kfree(kvm->arch.adapters[i]);
1929 static int modify_io_adapter(struct kvm_device *dev,
1930 struct kvm_device_attr *attr)
1932 struct kvm_s390_io_adapter_req req;
1933 struct s390_io_adapter *adapter;
1934 int ret;
1936 if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
1937 return -EFAULT;
1939 adapter = get_io_adapter(dev->kvm, req.id);
1940 if (!adapter)
1941 return -EINVAL;
1942 switch (req.type) {
1943 case KVM_S390_IO_ADAPTER_MASK:
1944 ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
1945 if (ret > 0)
1946 ret = 0;
1947 break;
1948 case KVM_S390_IO_ADAPTER_MAP:
1949 ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
1950 break;
1951 case KVM_S390_IO_ADAPTER_UNMAP:
1952 ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
1953 break;
1954 default:
1955 ret = -EINVAL;
1958 return ret;
1961 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1963 int r = 0;
1964 unsigned int i;
1965 struct kvm_vcpu *vcpu;
1967 switch (attr->group) {
1968 case KVM_DEV_FLIC_ENQUEUE:
1969 r = enqueue_floating_irq(dev, attr);
1970 break;
1971 case KVM_DEV_FLIC_CLEAR_IRQS:
1972 kvm_s390_clear_float_irqs(dev->kvm);
1973 break;
1974 case KVM_DEV_FLIC_APF_ENABLE:
1975 dev->kvm->arch.gmap->pfault_enabled = 1;
1976 break;
1977 case KVM_DEV_FLIC_APF_DISABLE_WAIT:
1978 dev->kvm->arch.gmap->pfault_enabled = 0;
1980 * Make sure no async faults are in transition when
1981 * clearing the queues. So we don't need to worry
1982 * about late coming workers.
1984 synchronize_srcu(&dev->kvm->srcu);
1985 kvm_for_each_vcpu(i, vcpu, dev->kvm)
1986 kvm_clear_async_pf_completion_queue(vcpu);
1987 break;
1988 case KVM_DEV_FLIC_ADAPTER_REGISTER:
1989 r = register_io_adapter(dev, attr);
1990 break;
1991 case KVM_DEV_FLIC_ADAPTER_MODIFY:
1992 r = modify_io_adapter(dev, attr);
1993 break;
1994 default:
1995 r = -EINVAL;
1998 return r;
2001 static int flic_create(struct kvm_device *dev, u32 type)
2003 if (!dev)
2004 return -EINVAL;
2005 if (dev->kvm->arch.flic)
2006 return -EINVAL;
2007 dev->kvm->arch.flic = dev;
2008 return 0;
2011 static void flic_destroy(struct kvm_device *dev)
2013 dev->kvm->arch.flic = NULL;
2014 kfree(dev);
2017 /* s390 floating irq controller (flic) */
2018 struct kvm_device_ops kvm_flic_ops = {
2019 .name = "kvm-flic",
2020 .get_attr = flic_get_attr,
2021 .set_attr = flic_set_attr,
2022 .create = flic_create,
2023 .destroy = flic_destroy,
2026 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2028 unsigned long bit;
2030 bit = bit_nr + (addr % PAGE_SIZE) * 8;
2032 return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2035 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2036 u64 addr)
2038 struct s390_map_info *map;
2040 if (!adapter)
2041 return NULL;
2043 list_for_each_entry(map, &adapter->maps, list) {
2044 if (map->guest_addr == addr)
2045 return map;
2047 return NULL;
2050 static int adapter_indicators_set(struct kvm *kvm,
2051 struct s390_io_adapter *adapter,
2052 struct kvm_s390_adapter_int *adapter_int)
2054 unsigned long bit;
2055 int summary_set, idx;
2056 struct s390_map_info *info;
2057 void *map;
2059 info = get_map_info(adapter, adapter_int->ind_addr);
2060 if (!info)
2061 return -1;
2062 map = page_address(info->page);
2063 bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2064 set_bit(bit, map);
2065 idx = srcu_read_lock(&kvm->srcu);
2066 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2067 set_page_dirty_lock(info->page);
2068 info = get_map_info(adapter, adapter_int->summary_addr);
2069 if (!info) {
2070 srcu_read_unlock(&kvm->srcu, idx);
2071 return -1;
2073 map = page_address(info->page);
2074 bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2075 adapter->swap);
2076 summary_set = test_and_set_bit(bit, map);
2077 mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2078 set_page_dirty_lock(info->page);
2079 srcu_read_unlock(&kvm->srcu, idx);
2080 return summary_set ? 0 : 1;
2084 * < 0 - not injected due to error
2085 * = 0 - coalesced, summary indicator already active
2086 * > 0 - injected interrupt
2088 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2089 struct kvm *kvm, int irq_source_id, int level,
2090 bool line_status)
2092 int ret;
2093 struct s390_io_adapter *adapter;
2095 /* We're only interested in the 0->1 transition. */
2096 if (!level)
2097 return 0;
2098 adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2099 if (!adapter)
2100 return -1;
2101 down_read(&adapter->maps_lock);
2102 ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2103 up_read(&adapter->maps_lock);
2104 if ((ret > 0) && !adapter->masked) {
2105 struct kvm_s390_interrupt s390int = {
2106 .type = KVM_S390_INT_IO(1, 0, 0, 0),
2107 .parm = 0,
2108 .parm64 = (adapter->isc << 27) | 0x80000000,
2110 ret = kvm_s390_inject_vm(kvm, &s390int);
2111 if (ret == 0)
2112 ret = 1;
2114 return ret;
2117 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
2118 const struct kvm_irq_routing_entry *ue)
2120 int ret;
2122 switch (ue->type) {
2123 case KVM_IRQ_ROUTING_S390_ADAPTER:
2124 e->set = set_adapter_int;
2125 e->adapter.summary_addr = ue->u.adapter.summary_addr;
2126 e->adapter.ind_addr = ue->u.adapter.ind_addr;
2127 e->adapter.summary_offset = ue->u.adapter.summary_offset;
2128 e->adapter.ind_offset = ue->u.adapter.ind_offset;
2129 e->adapter.adapter_id = ue->u.adapter.adapter_id;
2130 ret = 0;
2131 break;
2132 default:
2133 ret = -EINVAL;
2136 return ret;
2139 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2140 int irq_source_id, int level, bool line_status)
2142 return -EINVAL;
2145 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2147 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2148 struct kvm_s390_irq *buf;
2149 int r = 0;
2150 int n;
2152 buf = vmalloc(len);
2153 if (!buf)
2154 return -ENOMEM;
2156 if (copy_from_user((void *) buf, irqstate, len)) {
2157 r = -EFAULT;
2158 goto out_free;
2162 * Don't allow setting the interrupt state
2163 * when there are already interrupts pending
2165 spin_lock(&li->lock);
2166 if (li->pending_irqs) {
2167 r = -EBUSY;
2168 goto out_unlock;
2171 for (n = 0; n < len / sizeof(*buf); n++) {
2172 r = do_inject_vcpu(vcpu, &buf[n]);
2173 if (r)
2174 break;
2177 out_unlock:
2178 spin_unlock(&li->lock);
2179 out_free:
2180 vfree(buf);
2182 return r;
2185 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2186 struct kvm_s390_irq *irq,
2187 unsigned long irq_type)
2189 switch (irq_type) {
2190 case IRQ_PEND_MCHK_EX:
2191 case IRQ_PEND_MCHK_REP:
2192 irq->type = KVM_S390_MCHK;
2193 irq->u.mchk = li->irq.mchk;
2194 break;
2195 case IRQ_PEND_PROG:
2196 irq->type = KVM_S390_PROGRAM_INT;
2197 irq->u.pgm = li->irq.pgm;
2198 break;
2199 case IRQ_PEND_PFAULT_INIT:
2200 irq->type = KVM_S390_INT_PFAULT_INIT;
2201 irq->u.ext = li->irq.ext;
2202 break;
2203 case IRQ_PEND_EXT_EXTERNAL:
2204 irq->type = KVM_S390_INT_EXTERNAL_CALL;
2205 irq->u.extcall = li->irq.extcall;
2206 break;
2207 case IRQ_PEND_EXT_CLOCK_COMP:
2208 irq->type = KVM_S390_INT_CLOCK_COMP;
2209 break;
2210 case IRQ_PEND_EXT_CPU_TIMER:
2211 irq->type = KVM_S390_INT_CPU_TIMER;
2212 break;
2213 case IRQ_PEND_SIGP_STOP:
2214 irq->type = KVM_S390_SIGP_STOP;
2215 irq->u.stop = li->irq.stop;
2216 break;
2217 case IRQ_PEND_RESTART:
2218 irq->type = KVM_S390_RESTART;
2219 break;
2220 case IRQ_PEND_SET_PREFIX:
2221 irq->type = KVM_S390_SIGP_SET_PREFIX;
2222 irq->u.prefix = li->irq.prefix;
2223 break;
2227 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2229 uint8_t sigp_ctrl = vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sigp_ctrl;
2230 unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2231 struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2232 unsigned long pending_irqs;
2233 struct kvm_s390_irq irq;
2234 unsigned long irq_type;
2235 int cpuaddr;
2236 int n = 0;
2238 spin_lock(&li->lock);
2239 pending_irqs = li->pending_irqs;
2240 memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2241 sizeof(sigp_emerg_pending));
2242 spin_unlock(&li->lock);
2244 for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2245 memset(&irq, 0, sizeof(irq));
2246 if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2247 continue;
2248 if (n + sizeof(irq) > len)
2249 return -ENOBUFS;
2250 store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2251 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2252 return -EFAULT;
2253 n += sizeof(irq);
2256 if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2257 for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2258 memset(&irq, 0, sizeof(irq));
2259 if (n + sizeof(irq) > len)
2260 return -ENOBUFS;
2261 irq.type = KVM_S390_INT_EMERGENCY;
2262 irq.u.emerg.code = cpuaddr;
2263 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2264 return -EFAULT;
2265 n += sizeof(irq);
2269 if ((sigp_ctrl & SIGP_CTRL_C) &&
2270 (atomic_read(&vcpu->arch.sie_block->cpuflags) &
2271 CPUSTAT_ECALL_PEND)) {
2272 if (n + sizeof(irq) > len)
2273 return -ENOBUFS;
2274 memset(&irq, 0, sizeof(irq));
2275 irq.type = KVM_S390_INT_EXTERNAL_CALL;
2276 irq.u.extcall.code = sigp_ctrl & SIGP_CTRL_SCN_MASK;
2277 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2278 return -EFAULT;
2279 n += sizeof(irq);
2282 return n;