xref: /qemu/target/s390x/kvm/kvm.c (revision a976a99a)
1 /*
2  * QEMU S390x KVM implementation
3  *
4  * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
5  * Copyright IBM Corp. 2012
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License as published by
9  * the Free Software Foundation; either version 2 of the License, or
10  * (at your option) any later version.
11  *
12  * This program is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15  * General Public License for more details.
16  *
17  * You should have received a copy of the GNU General Public License
18  * along with this program; if not, see <http://www.gnu.org/licenses/>.
19  */
20 
21 #include "qemu/osdep.h"
22 #include <sys/ioctl.h>
23 
24 #include <linux/kvm.h>
25 #include <asm/ptrace.h>
26 
27 #include "cpu.h"
28 #include "s390x-internal.h"
29 #include "kvm_s390x.h"
30 #include "sysemu/kvm_int.h"
31 #include "qemu/cutils.h"
32 #include "qapi/error.h"
33 #include "qemu/error-report.h"
34 #include "qemu/timer.h"
35 #include "qemu/units.h"
36 #include "qemu/main-loop.h"
37 #include "qemu/mmap-alloc.h"
38 #include "qemu/log.h"
39 #include "sysemu/sysemu.h"
40 #include "sysemu/hw_accel.h"
41 #include "sysemu/runstate.h"
42 #include "sysemu/device_tree.h"
43 #include "exec/gdbstub.h"
44 #include "exec/ram_addr.h"
45 #include "trace.h"
46 #include "hw/s390x/s390-pci-inst.h"
47 #include "hw/s390x/s390-pci-bus.h"
48 #include "hw/s390x/ipl.h"
49 #include "hw/s390x/ebcdic.h"
50 #include "exec/memattrs.h"
51 #include "hw/s390x/s390-virtio-ccw.h"
52 #include "hw/s390x/s390-virtio-hcall.h"
53 #include "hw/s390x/pv.h"
54 
55 #ifndef DEBUG_KVM
56 #define DEBUG_KVM  0
57 #endif
58 
59 #define DPRINTF(fmt, ...) do {                \
60     if (DEBUG_KVM) {                          \
61         fprintf(stderr, fmt, ## __VA_ARGS__); \
62     }                                         \
63 } while (0)
64 
65 #define kvm_vm_check_mem_attr(s, attr) \
66     kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
67 
68 #define IPA0_DIAG                       0x8300
69 #define IPA0_SIGP                       0xae00
70 #define IPA0_B2                         0xb200
71 #define IPA0_B9                         0xb900
72 #define IPA0_EB                         0xeb00
73 #define IPA0_E3                         0xe300
74 
75 #define PRIV_B2_SCLP_CALL               0x20
76 #define PRIV_B2_CSCH                    0x30
77 #define PRIV_B2_HSCH                    0x31
78 #define PRIV_B2_MSCH                    0x32
79 #define PRIV_B2_SSCH                    0x33
80 #define PRIV_B2_STSCH                   0x34
81 #define PRIV_B2_TSCH                    0x35
82 #define PRIV_B2_TPI                     0x36
83 #define PRIV_B2_SAL                     0x37
84 #define PRIV_B2_RSCH                    0x38
85 #define PRIV_B2_STCRW                   0x39
86 #define PRIV_B2_STCPS                   0x3a
87 #define PRIV_B2_RCHP                    0x3b
88 #define PRIV_B2_SCHM                    0x3c
89 #define PRIV_B2_CHSC                    0x5f
90 #define PRIV_B2_SIGA                    0x74
91 #define PRIV_B2_XSCH                    0x76
92 
93 #define PRIV_EB_SQBS                    0x8a
94 #define PRIV_EB_PCISTB                  0xd0
95 #define PRIV_EB_SIC                     0xd1
96 
97 #define PRIV_B9_EQBS                    0x9c
98 #define PRIV_B9_CLP                     0xa0
99 #define PRIV_B9_PCISTG                  0xd0
100 #define PRIV_B9_PCILG                   0xd2
101 #define PRIV_B9_RPCIT                   0xd3
102 
103 #define PRIV_E3_MPCIFC                  0xd0
104 #define PRIV_E3_STPCIFC                 0xd4
105 
106 #define DIAG_TIMEREVENT                 0x288
107 #define DIAG_IPL                        0x308
108 #define DIAG_SET_CONTROL_PROGRAM_CODES  0x318
109 #define DIAG_KVM_HYPERCALL              0x500
110 #define DIAG_KVM_BREAKPOINT             0x501
111 
112 #define ICPT_INSTRUCTION                0x04
113 #define ICPT_PROGRAM                    0x08
114 #define ICPT_EXT_INT                    0x14
115 #define ICPT_WAITPSW                    0x1c
116 #define ICPT_SOFT_INTERCEPT             0x24
117 #define ICPT_CPU_STOP                   0x28
118 #define ICPT_OPEREXC                    0x2c
119 #define ICPT_IO                         0x40
120 #define ICPT_PV_INSTR                   0x68
121 #define ICPT_PV_INSTR_NOTIFICATION      0x6c
122 
123 #define NR_LOCAL_IRQS 32
124 /*
125  * Needs to be big enough to contain max_cpus emergency signals
126  * and in addition NR_LOCAL_IRQS interrupts
127  */
128 #define VCPU_IRQ_BUF_SIZE(max_cpus) (sizeof(struct kvm_s390_irq) * \
129                                      (max_cpus + NR_LOCAL_IRQS))
130 /*
131  * KVM does only support memory slots up to KVM_MEM_MAX_NR_PAGES pages
132  * as the dirty bitmap must be managed by bitops that take an int as
133  * position indicator. This would end at an unaligned  address
134  * (0x7fffff00000). As future variants might provide larger pages
135  * and to make all addresses properly aligned, let us split at 4TB.
136  */
137 #define KVM_SLOT_MAX_BYTES (4UL * TiB)
138 
139 static CPUWatchpoint hw_watchpoint;
140 /*
141  * We don't use a list because this structure is also used to transmit the
142  * hardware breakpoints to the kernel.
143  */
144 static struct kvm_hw_breakpoint *hw_breakpoints;
145 static int nb_hw_breakpoints;
146 
147 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
148     KVM_CAP_LAST_INFO
149 };
150 
151 static int cap_sync_regs;
152 static int cap_async_pf;
153 static int cap_mem_op;
154 static int cap_mem_op_extension;
155 static int cap_s390_irq;
156 static int cap_ri;
157 static int cap_hpage_1m;
158 static int cap_vcpu_resets;
159 static int cap_protected;
160 static int cap_zpci_op;
161 
162 static bool mem_op_storage_key_support;
163 
164 static int active_cmma;
165 
166 static int kvm_s390_query_mem_limit(uint64_t *memory_limit)
167 {
168     struct kvm_device_attr attr = {
169         .group = KVM_S390_VM_MEM_CTRL,
170         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
171         .addr = (uint64_t) memory_limit,
172     };
173 
174     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
175 }
176 
177 int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit)
178 {
179     int rc;
180 
181     struct kvm_device_attr attr = {
182         .group = KVM_S390_VM_MEM_CTRL,
183         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
184         .addr = (uint64_t) &new_limit,
185     };
186 
187     if (!kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_LIMIT_SIZE)) {
188         return 0;
189     }
190 
191     rc = kvm_s390_query_mem_limit(hw_limit);
192     if (rc) {
193         return rc;
194     } else if (*hw_limit < new_limit) {
195         return -E2BIG;
196     }
197 
198     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
199 }
200 
201 int kvm_s390_cmma_active(void)
202 {
203     return active_cmma;
204 }
205 
206 static bool kvm_s390_cmma_available(void)
207 {
208     static bool initialized, value;
209 
210     if (!initialized) {
211         initialized = true;
212         value = kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_ENABLE_CMMA) &&
213                 kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_CLR_CMMA);
214     }
215     return value;
216 }
217 
218 void kvm_s390_cmma_reset(void)
219 {
220     int rc;
221     struct kvm_device_attr attr = {
222         .group = KVM_S390_VM_MEM_CTRL,
223         .attr = KVM_S390_VM_MEM_CLR_CMMA,
224     };
225 
226     if (!kvm_s390_cmma_active()) {
227         return;
228     }
229 
230     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
231     trace_kvm_clear_cmma(rc);
232 }
233 
234 static void kvm_s390_enable_cmma(void)
235 {
236     int rc;
237     struct kvm_device_attr attr = {
238         .group = KVM_S390_VM_MEM_CTRL,
239         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
240     };
241 
242     if (cap_hpage_1m) {
243         warn_report("CMM will not be enabled because it is not "
244                     "compatible with huge memory backings.");
245         return;
246     }
247     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
248     active_cmma = !rc;
249     trace_kvm_enable_cmma(rc);
250 }
251 
252 static void kvm_s390_set_attr(uint64_t attr)
253 {
254     struct kvm_device_attr attribute = {
255         .group = KVM_S390_VM_CRYPTO,
256         .attr  = attr,
257     };
258 
259     int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
260 
261     if (ret) {
262         error_report("Failed to set crypto device attribute %lu: %s",
263                      attr, strerror(-ret));
264     }
265 }
266 
267 static void kvm_s390_init_aes_kw(void)
268 {
269     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
270 
271     if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
272                                  NULL)) {
273             attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
274     }
275 
276     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
277             kvm_s390_set_attr(attr);
278     }
279 }
280 
281 static void kvm_s390_init_dea_kw(void)
282 {
283     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
284 
285     if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
286                                  NULL)) {
287             attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
288     }
289 
290     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
291             kvm_s390_set_attr(attr);
292     }
293 }
294 
295 void kvm_s390_crypto_reset(void)
296 {
297     if (s390_has_feat(S390_FEAT_MSA_EXT_3)) {
298         kvm_s390_init_aes_kw();
299         kvm_s390_init_dea_kw();
300     }
301 }
302 
303 void kvm_s390_set_max_pagesize(uint64_t pagesize, Error **errp)
304 {
305     if (pagesize == 4 * KiB) {
306         return;
307     }
308 
309     if (!hpage_1m_allowed()) {
310         error_setg(errp, "This QEMU machine does not support huge page "
311                    "mappings");
312         return;
313     }
314 
315     if (pagesize != 1 * MiB) {
316         error_setg(errp, "Memory backing with 2G pages was specified, "
317                    "but KVM does not support this memory backing");
318         return;
319     }
320 
321     if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_HPAGE_1M, 0)) {
322         error_setg(errp, "Memory backing with 1M pages was specified, "
323                    "but KVM does not support this memory backing");
324         return;
325     }
326 
327     cap_hpage_1m = 1;
328 }
329 
330 int kvm_s390_get_hpage_1m(void)
331 {
332     return cap_hpage_1m;
333 }
334 
335 static void ccw_machine_class_foreach(ObjectClass *oc, void *opaque)
336 {
337     MachineClass *mc = MACHINE_CLASS(oc);
338 
339     mc->default_cpu_type = S390_CPU_TYPE_NAME("host");
340 }
341 
342 int kvm_arch_init(MachineState *ms, KVMState *s)
343 {
344     object_class_foreach(ccw_machine_class_foreach, TYPE_S390_CCW_MACHINE,
345                          false, NULL);
346 
347     if (!kvm_check_extension(kvm_state, KVM_CAP_DEVICE_CTRL)) {
348         error_report("KVM is missing capability KVM_CAP_DEVICE_CTRL - "
349                      "please use kernel 3.15 or newer");
350         return -1;
351     }
352     if (!kvm_check_extension(s, KVM_CAP_S390_COW)) {
353         error_report("KVM is missing capability KVM_CAP_S390_COW - "
354                      "unsupported environment");
355         return -1;
356     }
357 
358     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
359     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
360     cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
361     cap_mem_op_extension = kvm_check_extension(s, KVM_CAP_S390_MEM_OP_EXTENSION);
362     mem_op_storage_key_support = cap_mem_op_extension > 0;
363     cap_s390_irq = kvm_check_extension(s, KVM_CAP_S390_INJECT_IRQ);
364     cap_vcpu_resets = kvm_check_extension(s, KVM_CAP_S390_VCPU_RESETS);
365     cap_protected = kvm_check_extension(s, KVM_CAP_S390_PROTECTED);
366     cap_zpci_op = kvm_check_extension(s, KVM_CAP_S390_ZPCI_OP);
367 
368     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
369     kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
370     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
371     if (ri_allowed()) {
372         if (kvm_vm_enable_cap(s, KVM_CAP_S390_RI, 0) == 0) {
373             cap_ri = 1;
374         }
375     }
376     if (cpu_model_allowed()) {
377         kvm_vm_enable_cap(s, KVM_CAP_S390_GS, 0);
378     }
379 
380     /*
381      * The migration interface for ais was introduced with kernel 4.13
382      * but the capability itself had been active since 4.12. As migration
383      * support is considered necessary, we only try to enable this for
384      * newer machine types if KVM_CAP_S390_AIS_MIGRATION is available.
385      */
386     if (cpu_model_allowed() && kvm_kernel_irqchip_allowed() &&
387         kvm_check_extension(s, KVM_CAP_S390_AIS_MIGRATION)) {
388         kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
389     }
390 
391     kvm_set_max_memslot_size(KVM_SLOT_MAX_BYTES);
392     return 0;
393 }
394 
395 int kvm_arch_irqchip_create(KVMState *s)
396 {
397     return 0;
398 }
399 
400 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
401 {
402     return cpu->cpu_index;
403 }
404 
405 int kvm_arch_init_vcpu(CPUState *cs)
406 {
407     unsigned int max_cpus = MACHINE(qdev_get_machine())->smp.max_cpus;
408     S390CPU *cpu = S390_CPU(cs);
409     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
410     cpu->irqstate = g_malloc0(VCPU_IRQ_BUF_SIZE(max_cpus));
411     return 0;
412 }
413 
414 int kvm_arch_destroy_vcpu(CPUState *cs)
415 {
416     S390CPU *cpu = S390_CPU(cs);
417 
418     g_free(cpu->irqstate);
419     cpu->irqstate = NULL;
420 
421     return 0;
422 }
423 
424 static void kvm_s390_reset_vcpu(S390CPU *cpu, unsigned long type)
425 {
426     CPUState *cs = CPU(cpu);
427 
428     /*
429      * The reset call is needed here to reset in-kernel vcpu data that
430      * we can't access directly from QEMU (i.e. with older kernels
431      * which don't support sync_regs/ONE_REG).  Before this ioctl
432      * cpu_synchronize_state() is called in common kvm code
433      * (kvm-all).
434      */
435     if (kvm_vcpu_ioctl(cs, type)) {
436         error_report("CPU reset failed on CPU %i type %lx",
437                      cs->cpu_index, type);
438     }
439 }
440 
441 void kvm_s390_reset_vcpu_initial(S390CPU *cpu)
442 {
443     kvm_s390_reset_vcpu(cpu, KVM_S390_INITIAL_RESET);
444 }
445 
446 void kvm_s390_reset_vcpu_clear(S390CPU *cpu)
447 {
448     if (cap_vcpu_resets) {
449         kvm_s390_reset_vcpu(cpu, KVM_S390_CLEAR_RESET);
450     } else {
451         kvm_s390_reset_vcpu(cpu, KVM_S390_INITIAL_RESET);
452     }
453 }
454 
455 void kvm_s390_reset_vcpu_normal(S390CPU *cpu)
456 {
457     if (cap_vcpu_resets) {
458         kvm_s390_reset_vcpu(cpu, KVM_S390_NORMAL_RESET);
459     }
460 }
461 
462 static int can_sync_regs(CPUState *cs, int regs)
463 {
464     return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
465 }
466 
467 int kvm_arch_put_registers(CPUState *cs, int level)
468 {
469     S390CPU *cpu = S390_CPU(cs);
470     CPUS390XState *env = &cpu->env;
471     struct kvm_sregs sregs;
472     struct kvm_regs regs;
473     struct kvm_fpu fpu = {};
474     int r;
475     int i;
476 
477     /* always save the PSW  and the GPRS*/
478     cs->kvm_run->psw_addr = env->psw.addr;
479     cs->kvm_run->psw_mask = env->psw.mask;
480 
481     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
482         for (i = 0; i < 16; i++) {
483             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
484             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
485         }
486     } else {
487         for (i = 0; i < 16; i++) {
488             regs.gprs[i] = env->regs[i];
489         }
490         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
491         if (r < 0) {
492             return r;
493         }
494     }
495 
496     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
497         for (i = 0; i < 32; i++) {
498             cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0];
499             cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1];
500         }
501         cs->kvm_run->s.regs.fpc = env->fpc;
502         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
503     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
504         for (i = 0; i < 16; i++) {
505             cs->kvm_run->s.regs.fprs[i] = *get_freg(env, i);
506         }
507         cs->kvm_run->s.regs.fpc = env->fpc;
508         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_FPRS;
509     } else {
510         /* Floating point */
511         for (i = 0; i < 16; i++) {
512             fpu.fprs[i] = *get_freg(env, i);
513         }
514         fpu.fpc = env->fpc;
515 
516         r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
517         if (r < 0) {
518             return r;
519         }
520     }
521 
522     /* Do we need to save more than that? */
523     if (level == KVM_PUT_RUNTIME_STATE) {
524         return 0;
525     }
526 
527     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
528         cs->kvm_run->s.regs.cputm = env->cputm;
529         cs->kvm_run->s.regs.ckc = env->ckc;
530         cs->kvm_run->s.regs.todpr = env->todpr;
531         cs->kvm_run->s.regs.gbea = env->gbea;
532         cs->kvm_run->s.regs.pp = env->pp;
533         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
534     } else {
535         /*
536          * These ONE_REGS are not protected by a capability. As they are only
537          * necessary for migration we just trace a possible error, but don't
538          * return with an error return code.
539          */
540         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
541         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
542         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
543         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
544         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
545     }
546 
547     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
548         memcpy(cs->kvm_run->s.regs.riccb, env->riccb, 64);
549         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_RICCB;
550     }
551 
552     /* pfault parameters */
553     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
554         cs->kvm_run->s.regs.pft = env->pfault_token;
555         cs->kvm_run->s.regs.pfs = env->pfault_select;
556         cs->kvm_run->s.regs.pfc = env->pfault_compare;
557         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
558     } else if (cap_async_pf) {
559         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
560         if (r < 0) {
561             return r;
562         }
563         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
564         if (r < 0) {
565             return r;
566         }
567         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
568         if (r < 0) {
569             return r;
570         }
571     }
572 
573     /* access registers and control registers*/
574     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
575         for (i = 0; i < 16; i++) {
576             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
577             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
578         }
579         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
580         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
581     } else {
582         for (i = 0; i < 16; i++) {
583             sregs.acrs[i] = env->aregs[i];
584             sregs.crs[i] = env->cregs[i];
585         }
586         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
587         if (r < 0) {
588             return r;
589         }
590     }
591 
592     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
593         memcpy(cs->kvm_run->s.regs.gscb, env->gscb, 32);
594         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GSCB;
595     }
596 
597     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
598         cs->kvm_run->s.regs.bpbc = env->bpbc;
599         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_BPBC;
600     }
601 
602     if (can_sync_regs(cs, KVM_SYNC_ETOKEN)) {
603         cs->kvm_run->s.regs.etoken = env->etoken;
604         cs->kvm_run->s.regs.etoken_extension  = env->etoken_extension;
605         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ETOKEN;
606     }
607 
608     if (can_sync_regs(cs, KVM_SYNC_DIAG318)) {
609         cs->kvm_run->s.regs.diag318 = env->diag318_info;
610         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_DIAG318;
611     }
612 
613     /* Finally the prefix */
614     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
615         cs->kvm_run->s.regs.prefix = env->psa;
616         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
617     } else {
618         /* prefix is only supported via sync regs */
619     }
620     return 0;
621 }
622 
623 int kvm_arch_get_registers(CPUState *cs)
624 {
625     S390CPU *cpu = S390_CPU(cs);
626     CPUS390XState *env = &cpu->env;
627     struct kvm_sregs sregs;
628     struct kvm_regs regs;
629     struct kvm_fpu fpu;
630     int i, r;
631 
632     /* get the PSW */
633     env->psw.addr = cs->kvm_run->psw_addr;
634     env->psw.mask = cs->kvm_run->psw_mask;
635 
636     /* the GPRS */
637     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
638         for (i = 0; i < 16; i++) {
639             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
640         }
641     } else {
642         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
643         if (r < 0) {
644             return r;
645         }
646          for (i = 0; i < 16; i++) {
647             env->regs[i] = regs.gprs[i];
648         }
649     }
650 
651     /* The ACRS and CRS */
652     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
653         for (i = 0; i < 16; i++) {
654             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
655             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
656         }
657     } else {
658         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
659         if (r < 0) {
660             return r;
661         }
662          for (i = 0; i < 16; i++) {
663             env->aregs[i] = sregs.acrs[i];
664             env->cregs[i] = sregs.crs[i];
665         }
666     }
667 
668     /* Floating point and vector registers */
669     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
670         for (i = 0; i < 32; i++) {
671             env->vregs[i][0] = cs->kvm_run->s.regs.vrs[i][0];
672             env->vregs[i][1] = cs->kvm_run->s.regs.vrs[i][1];
673         }
674         env->fpc = cs->kvm_run->s.regs.fpc;
675     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
676         for (i = 0; i < 16; i++) {
677             *get_freg(env, i) = cs->kvm_run->s.regs.fprs[i];
678         }
679         env->fpc = cs->kvm_run->s.regs.fpc;
680     } else {
681         r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
682         if (r < 0) {
683             return r;
684         }
685         for (i = 0; i < 16; i++) {
686             *get_freg(env, i) = fpu.fprs[i];
687         }
688         env->fpc = fpu.fpc;
689     }
690 
691     /* The prefix */
692     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
693         env->psa = cs->kvm_run->s.regs.prefix;
694     }
695 
696     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
697         env->cputm = cs->kvm_run->s.regs.cputm;
698         env->ckc = cs->kvm_run->s.regs.ckc;
699         env->todpr = cs->kvm_run->s.regs.todpr;
700         env->gbea = cs->kvm_run->s.regs.gbea;
701         env->pp = cs->kvm_run->s.regs.pp;
702     } else {
703         /*
704          * These ONE_REGS are not protected by a capability. As they are only
705          * necessary for migration we just trace a possible error, but don't
706          * return with an error return code.
707          */
708         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
709         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
710         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
711         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
712         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
713     }
714 
715     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
716         memcpy(env->riccb, cs->kvm_run->s.regs.riccb, 64);
717     }
718 
719     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
720         memcpy(env->gscb, cs->kvm_run->s.regs.gscb, 32);
721     }
722 
723     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
724         env->bpbc = cs->kvm_run->s.regs.bpbc;
725     }
726 
727     if (can_sync_regs(cs, KVM_SYNC_ETOKEN)) {
728         env->etoken = cs->kvm_run->s.regs.etoken;
729         env->etoken_extension = cs->kvm_run->s.regs.etoken_extension;
730     }
731 
732     /* pfault parameters */
733     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
734         env->pfault_token = cs->kvm_run->s.regs.pft;
735         env->pfault_select = cs->kvm_run->s.regs.pfs;
736         env->pfault_compare = cs->kvm_run->s.regs.pfc;
737     } else if (cap_async_pf) {
738         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
739         if (r < 0) {
740             return r;
741         }
742         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
743         if (r < 0) {
744             return r;
745         }
746         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
747         if (r < 0) {
748             return r;
749         }
750     }
751 
752     if (can_sync_regs(cs, KVM_SYNC_DIAG318)) {
753         env->diag318_info = cs->kvm_run->s.regs.diag318;
754     }
755 
756     return 0;
757 }
758 
759 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
760 {
761     int r;
762     struct kvm_device_attr attr = {
763         .group = KVM_S390_VM_TOD,
764         .attr = KVM_S390_VM_TOD_LOW,
765         .addr = (uint64_t)tod_low,
766     };
767 
768     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
769     if (r) {
770         return r;
771     }
772 
773     attr.attr = KVM_S390_VM_TOD_HIGH;
774     attr.addr = (uint64_t)tod_high;
775     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
776 }
777 
778 int kvm_s390_get_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
779 {
780     int r;
781     struct kvm_s390_vm_tod_clock gtod;
782     struct kvm_device_attr attr = {
783         .group = KVM_S390_VM_TOD,
784         .attr = KVM_S390_VM_TOD_EXT,
785         .addr = (uint64_t)&gtod,
786     };
787 
788     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
789     *tod_high = gtod.epoch_idx;
790     *tod_low  = gtod.tod;
791 
792     return r;
793 }
794 
795 int kvm_s390_set_clock(uint8_t tod_high, uint64_t tod_low)
796 {
797     int r;
798     struct kvm_device_attr attr = {
799         .group = KVM_S390_VM_TOD,
800         .attr = KVM_S390_VM_TOD_LOW,
801         .addr = (uint64_t)&tod_low,
802     };
803 
804     r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
805     if (r) {
806         return r;
807     }
808 
809     attr.attr = KVM_S390_VM_TOD_HIGH;
810     attr.addr = (uint64_t)&tod_high;
811     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
812 }
813 
814 int kvm_s390_set_clock_ext(uint8_t tod_high, uint64_t tod_low)
815 {
816     struct kvm_s390_vm_tod_clock gtod = {
817         .epoch_idx = tod_high,
818         .tod  = tod_low,
819     };
820     struct kvm_device_attr attr = {
821         .group = KVM_S390_VM_TOD,
822         .attr = KVM_S390_VM_TOD_EXT,
823         .addr = (uint64_t)&gtod,
824     };
825 
826     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
827 }
828 
829 /**
830  * kvm_s390_mem_op:
831  * @addr:      the logical start address in guest memory
832  * @ar:        the access register number
833  * @hostbuf:   buffer in host memory. NULL = do only checks w/o copying
834  * @len:       length that should be transferred
835  * @is_write:  true = write, false = read
836  * Returns:    0 on success, non-zero if an exception or error occurred
837  *
838  * Use KVM ioctl to read/write from/to guest memory. An access exception
839  * is injected into the vCPU in case of translation errors.
840  */
841 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, uint8_t ar, void *hostbuf,
842                     int len, bool is_write)
843 {
844     struct kvm_s390_mem_op mem_op = {
845         .gaddr = addr,
846         .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
847         .size = len,
848         .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
849                        : KVM_S390_MEMOP_LOGICAL_READ,
850         .buf = (uint64_t)hostbuf,
851         .ar = ar,
852         .key = (cpu->env.psw.mask & PSW_MASK_KEY) >> PSW_SHIFT_KEY,
853     };
854     int ret;
855 
856     if (!cap_mem_op) {
857         return -ENOSYS;
858     }
859     if (!hostbuf) {
860         mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
861     }
862     if (mem_op_storage_key_support) {
863         mem_op.flags |= KVM_S390_MEMOP_F_SKEY_PROTECTION;
864     }
865 
866     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
867     if (ret < 0) {
868         warn_report("KVM_S390_MEM_OP failed: %s", strerror(-ret));
869     }
870     return ret;
871 }
872 
873 int kvm_s390_mem_op_pv(S390CPU *cpu, uint64_t offset, void *hostbuf,
874                        int len, bool is_write)
875 {
876     struct kvm_s390_mem_op mem_op = {
877         .sida_offset = offset,
878         .size = len,
879         .op = is_write ? KVM_S390_MEMOP_SIDA_WRITE
880                        : KVM_S390_MEMOP_SIDA_READ,
881         .buf = (uint64_t)hostbuf,
882     };
883     int ret;
884 
885     if (!cap_mem_op || !cap_protected) {
886         return -ENOSYS;
887     }
888 
889     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
890     if (ret < 0) {
891         error_report("KVM_S390_MEM_OP failed: %s", strerror(-ret));
892         abort();
893     }
894     return ret;
895 }
896 
897 static uint8_t const *sw_bp_inst;
898 static uint8_t sw_bp_ilen;
899 
900 static void determine_sw_breakpoint_instr(void)
901 {
902         /* DIAG 501 is used for sw breakpoints with old kernels */
903         static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
904         /* Instruction 0x0000 is used for sw breakpoints with recent kernels */
905         static const uint8_t instr_0x0000[] = {0x00, 0x00};
906 
907         if (sw_bp_inst) {
908             return;
909         }
910         if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_USER_INSTR0, 0)) {
911             sw_bp_inst = diag_501;
912             sw_bp_ilen = sizeof(diag_501);
913             DPRINTF("KVM: will use 4-byte sw breakpoints.\n");
914         } else {
915             sw_bp_inst = instr_0x0000;
916             sw_bp_ilen = sizeof(instr_0x0000);
917             DPRINTF("KVM: will use 2-byte sw breakpoints.\n");
918         }
919 }
920 
921 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
922 {
923     determine_sw_breakpoint_instr();
924 
925     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
926                             sw_bp_ilen, 0) ||
927         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)sw_bp_inst, sw_bp_ilen, 1)) {
928         return -EINVAL;
929     }
930     return 0;
931 }
932 
933 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
934 {
935     uint8_t t[MAX_ILEN];
936 
937     if (cpu_memory_rw_debug(cs, bp->pc, t, sw_bp_ilen, 0)) {
938         return -EINVAL;
939     } else if (memcmp(t, sw_bp_inst, sw_bp_ilen)) {
940         return -EINVAL;
941     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
942                                    sw_bp_ilen, 1)) {
943         return -EINVAL;
944     }
945 
946     return 0;
947 }
948 
949 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
950                                                     int len, int type)
951 {
952     int n;
953 
954     for (n = 0; n < nb_hw_breakpoints; n++) {
955         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
956             (hw_breakpoints[n].len == len || len == -1)) {
957             return &hw_breakpoints[n];
958         }
959     }
960 
961     return NULL;
962 }
963 
964 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
965 {
966     int size;
967 
968     if (find_hw_breakpoint(addr, len, type)) {
969         return -EEXIST;
970     }
971 
972     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
973 
974     if (!hw_breakpoints) {
975         nb_hw_breakpoints = 0;
976         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
977     } else {
978         hw_breakpoints =
979             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
980     }
981 
982     if (!hw_breakpoints) {
983         nb_hw_breakpoints = 0;
984         return -ENOMEM;
985     }
986 
987     hw_breakpoints[nb_hw_breakpoints].addr = addr;
988     hw_breakpoints[nb_hw_breakpoints].len = len;
989     hw_breakpoints[nb_hw_breakpoints].type = type;
990 
991     nb_hw_breakpoints++;
992 
993     return 0;
994 }
995 
996 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
997                                   target_ulong len, int type)
998 {
999     switch (type) {
1000     case GDB_BREAKPOINT_HW:
1001         type = KVM_HW_BP;
1002         break;
1003     case GDB_WATCHPOINT_WRITE:
1004         if (len < 1) {
1005             return -EINVAL;
1006         }
1007         type = KVM_HW_WP_WRITE;
1008         break;
1009     default:
1010         return -ENOSYS;
1011     }
1012     return insert_hw_breakpoint(addr, len, type);
1013 }
1014 
1015 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
1016                                   target_ulong len, int type)
1017 {
1018     int size;
1019     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
1020 
1021     if (bp == NULL) {
1022         return -ENOENT;
1023     }
1024 
1025     nb_hw_breakpoints--;
1026     if (nb_hw_breakpoints > 0) {
1027         /*
1028          * In order to trim the array, move the last element to the position to
1029          * be removed - if necessary.
1030          */
1031         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
1032             *bp = hw_breakpoints[nb_hw_breakpoints];
1033         }
1034         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
1035         hw_breakpoints =
1036              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
1037     } else {
1038         g_free(hw_breakpoints);
1039         hw_breakpoints = NULL;
1040     }
1041 
1042     return 0;
1043 }
1044 
1045 void kvm_arch_remove_all_hw_breakpoints(void)
1046 {
1047     nb_hw_breakpoints = 0;
1048     g_free(hw_breakpoints);
1049     hw_breakpoints = NULL;
1050 }
1051 
1052 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
1053 {
1054     int i;
1055 
1056     if (nb_hw_breakpoints > 0) {
1057         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
1058         dbg->arch.hw_bp = hw_breakpoints;
1059 
1060         for (i = 0; i < nb_hw_breakpoints; ++i) {
1061             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
1062                                                        hw_breakpoints[i].addr);
1063         }
1064         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
1065     } else {
1066         dbg->arch.nr_hw_bp = 0;
1067         dbg->arch.hw_bp = NULL;
1068     }
1069 }
1070 
1071 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
1072 {
1073 }
1074 
1075 MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
1076 {
1077     return MEMTXATTRS_UNSPECIFIED;
1078 }
1079 
1080 int kvm_arch_process_async_events(CPUState *cs)
1081 {
1082     return cs->halted;
1083 }
1084 
1085 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
1086                                      struct kvm_s390_interrupt *interrupt)
1087 {
1088     int r = 0;
1089 
1090     interrupt->type = irq->type;
1091     switch (irq->type) {
1092     case KVM_S390_INT_VIRTIO:
1093         interrupt->parm = irq->u.ext.ext_params;
1094         /* fall through */
1095     case KVM_S390_INT_PFAULT_INIT:
1096     case KVM_S390_INT_PFAULT_DONE:
1097         interrupt->parm64 = irq->u.ext.ext_params2;
1098         break;
1099     case KVM_S390_PROGRAM_INT:
1100         interrupt->parm = irq->u.pgm.code;
1101         break;
1102     case KVM_S390_SIGP_SET_PREFIX:
1103         interrupt->parm = irq->u.prefix.address;
1104         break;
1105     case KVM_S390_INT_SERVICE:
1106         interrupt->parm = irq->u.ext.ext_params;
1107         break;
1108     case KVM_S390_MCHK:
1109         interrupt->parm = irq->u.mchk.cr14;
1110         interrupt->parm64 = irq->u.mchk.mcic;
1111         break;
1112     case KVM_S390_INT_EXTERNAL_CALL:
1113         interrupt->parm = irq->u.extcall.code;
1114         break;
1115     case KVM_S390_INT_EMERGENCY:
1116         interrupt->parm = irq->u.emerg.code;
1117         break;
1118     case KVM_S390_SIGP_STOP:
1119     case KVM_S390_RESTART:
1120         break; /* These types have no parameters */
1121     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1122         interrupt->parm = irq->u.io.subchannel_id << 16;
1123         interrupt->parm |= irq->u.io.subchannel_nr;
1124         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
1125         interrupt->parm64 |= irq->u.io.io_int_word;
1126         break;
1127     default:
1128         r = -EINVAL;
1129         break;
1130     }
1131     return r;
1132 }
1133 
1134 static void inject_vcpu_irq_legacy(CPUState *cs, struct kvm_s390_irq *irq)
1135 {
1136     struct kvm_s390_interrupt kvmint = {};
1137     int r;
1138 
1139     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1140     if (r < 0) {
1141         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1142         exit(1);
1143     }
1144 
1145     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
1146     if (r < 0) {
1147         fprintf(stderr, "KVM failed to inject interrupt\n");
1148         exit(1);
1149     }
1150 }
1151 
1152 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
1153 {
1154     CPUState *cs = CPU(cpu);
1155     int r;
1156 
1157     if (cap_s390_irq) {
1158         r = kvm_vcpu_ioctl(cs, KVM_S390_IRQ, irq);
1159         if (!r) {
1160             return;
1161         }
1162         error_report("KVM failed to inject interrupt %llx", irq->type);
1163         exit(1);
1164     }
1165 
1166     inject_vcpu_irq_legacy(cs, irq);
1167 }
1168 
1169 void kvm_s390_floating_interrupt_legacy(struct kvm_s390_irq *irq)
1170 {
1171     struct kvm_s390_interrupt kvmint = {};
1172     int r;
1173 
1174     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1175     if (r < 0) {
1176         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1177         exit(1);
1178     }
1179 
1180     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
1181     if (r < 0) {
1182         fprintf(stderr, "KVM failed to inject interrupt\n");
1183         exit(1);
1184     }
1185 }
1186 
1187 void kvm_s390_program_interrupt(S390CPU *cpu, uint16_t code)
1188 {
1189     struct kvm_s390_irq irq = {
1190         .type = KVM_S390_PROGRAM_INT,
1191         .u.pgm.code = code,
1192     };
1193     qemu_log_mask(CPU_LOG_INT, "program interrupt at %#" PRIx64 "\n",
1194                   cpu->env.psw.addr);
1195     kvm_s390_vcpu_interrupt(cpu, &irq);
1196 }
1197 
1198 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
1199 {
1200     struct kvm_s390_irq irq = {
1201         .type = KVM_S390_PROGRAM_INT,
1202         .u.pgm.code = code,
1203         .u.pgm.trans_exc_code = te_code,
1204         .u.pgm.exc_access_id = te_code & 3,
1205     };
1206 
1207     kvm_s390_vcpu_interrupt(cpu, &irq);
1208 }
1209 
1210 static void kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
1211                                  uint16_t ipbh0)
1212 {
1213     CPUS390XState *env = &cpu->env;
1214     uint64_t sccb;
1215     uint32_t code;
1216     int r;
1217 
1218     sccb = env->regs[ipbh0 & 0xf];
1219     code = env->regs[(ipbh0 & 0xf0) >> 4];
1220 
1221     switch (run->s390_sieic.icptcode) {
1222     case ICPT_PV_INSTR_NOTIFICATION:
1223         g_assert(s390_is_pv());
1224         /* The notification intercepts are currently handled by KVM */
1225         error_report("unexpected SCLP PV notification");
1226         exit(1);
1227         break;
1228     case ICPT_PV_INSTR:
1229         g_assert(s390_is_pv());
1230         sclp_service_call_protected(env, sccb, code);
1231         /* Setting the CC is done by the Ultravisor. */
1232         break;
1233     case ICPT_INSTRUCTION:
1234         g_assert(!s390_is_pv());
1235         r = sclp_service_call(env, sccb, code);
1236         if (r < 0) {
1237             kvm_s390_program_interrupt(cpu, -r);
1238             return;
1239         }
1240         setcc(cpu, r);
1241     }
1242 }
1243 
1244 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1245 {
1246     CPUS390XState *env = &cpu->env;
1247     int rc = 0;
1248     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
1249 
1250     switch (ipa1) {
1251     case PRIV_B2_XSCH:
1252         ioinst_handle_xsch(cpu, env->regs[1], RA_IGNORED);
1253         break;
1254     case PRIV_B2_CSCH:
1255         ioinst_handle_csch(cpu, env->regs[1], RA_IGNORED);
1256         break;
1257     case PRIV_B2_HSCH:
1258         ioinst_handle_hsch(cpu, env->regs[1], RA_IGNORED);
1259         break;
1260     case PRIV_B2_MSCH:
1261         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1262         break;
1263     case PRIV_B2_SSCH:
1264         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1265         break;
1266     case PRIV_B2_STCRW:
1267         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb, RA_IGNORED);
1268         break;
1269     case PRIV_B2_STSCH:
1270         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1271         break;
1272     case PRIV_B2_TSCH:
1273         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1274         fprintf(stderr, "Spurious tsch intercept\n");
1275         break;
1276     case PRIV_B2_CHSC:
1277         ioinst_handle_chsc(cpu, run->s390_sieic.ipb, RA_IGNORED);
1278         break;
1279     case PRIV_B2_TPI:
1280         /* This should have been handled by kvm already. */
1281         fprintf(stderr, "Spurious tpi intercept\n");
1282         break;
1283     case PRIV_B2_SCHM:
1284         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
1285                            run->s390_sieic.ipb, RA_IGNORED);
1286         break;
1287     case PRIV_B2_RSCH:
1288         ioinst_handle_rsch(cpu, env->regs[1], RA_IGNORED);
1289         break;
1290     case PRIV_B2_RCHP:
1291         ioinst_handle_rchp(cpu, env->regs[1], RA_IGNORED);
1292         break;
1293     case PRIV_B2_STCPS:
1294         /* We do not provide this instruction, it is suppressed. */
1295         break;
1296     case PRIV_B2_SAL:
1297         ioinst_handle_sal(cpu, env->regs[1], RA_IGNORED);
1298         break;
1299     case PRIV_B2_SIGA:
1300         /* Not provided, set CC = 3 for subchannel not operational */
1301         setcc(cpu, 3);
1302         break;
1303     case PRIV_B2_SCLP_CALL:
1304         kvm_sclp_service_call(cpu, run, ipbh0);
1305         break;
1306     default:
1307         rc = -1;
1308         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
1309         break;
1310     }
1311 
1312     return rc;
1313 }
1314 
1315 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run,
1316                                   uint8_t *ar)
1317 {
1318     CPUS390XState *env = &cpu->env;
1319     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1320     uint32_t base2 = run->s390_sieic.ipb >> 28;
1321     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1322                      ((run->s390_sieic.ipb & 0xff00) << 4);
1323 
1324     if (disp2 & 0x80000) {
1325         disp2 += 0xfff00000;
1326     }
1327     if (ar) {
1328         *ar = base2;
1329     }
1330 
1331     return (base2 ? env->regs[base2] : 0) +
1332            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1333 }
1334 
1335 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run,
1336                                   uint8_t *ar)
1337 {
1338     CPUS390XState *env = &cpu->env;
1339     uint32_t base2 = run->s390_sieic.ipb >> 28;
1340     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1341                      ((run->s390_sieic.ipb & 0xff00) << 4);
1342 
1343     if (disp2 & 0x80000) {
1344         disp2 += 0xfff00000;
1345     }
1346     if (ar) {
1347         *ar = base2;
1348     }
1349 
1350     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1351 }
1352 
1353 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1354 {
1355     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1356 
1357     if (s390_has_feat(S390_FEAT_ZPCI)) {
1358         return clp_service_call(cpu, r2, RA_IGNORED);
1359     } else {
1360         return -1;
1361     }
1362 }
1363 
1364 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1365 {
1366     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1367     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1368 
1369     if (s390_has_feat(S390_FEAT_ZPCI)) {
1370         return pcilg_service_call(cpu, r1, r2, RA_IGNORED);
1371     } else {
1372         return -1;
1373     }
1374 }
1375 
1376 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1377 {
1378     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1379     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1380 
1381     if (s390_has_feat(S390_FEAT_ZPCI)) {
1382         return pcistg_service_call(cpu, r1, r2, RA_IGNORED);
1383     } else {
1384         return -1;
1385     }
1386 }
1387 
1388 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1389 {
1390     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1391     uint64_t fiba;
1392     uint8_t ar;
1393 
1394     if (s390_has_feat(S390_FEAT_ZPCI)) {
1395         fiba = get_base_disp_rxy(cpu, run, &ar);
1396 
1397         return stpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1398     } else {
1399         return -1;
1400     }
1401 }
1402 
1403 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1404 {
1405     CPUS390XState *env = &cpu->env;
1406     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1407     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1408     uint8_t isc;
1409     uint16_t mode;
1410     int r;
1411 
1412     mode = env->regs[r1] & 0xffff;
1413     isc = (env->regs[r3] >> 27) & 0x7;
1414     r = css_do_sic(env, isc, mode);
1415     if (r) {
1416         kvm_s390_program_interrupt(cpu, -r);
1417     }
1418 
1419     return 0;
1420 }
1421 
1422 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1423 {
1424     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1425     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1426 
1427     if (s390_has_feat(S390_FEAT_ZPCI)) {
1428         return rpcit_service_call(cpu, r1, r2, RA_IGNORED);
1429     } else {
1430         return -1;
1431     }
1432 }
1433 
1434 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1435 {
1436     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1437     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1438     uint64_t gaddr;
1439     uint8_t ar;
1440 
1441     if (s390_has_feat(S390_FEAT_ZPCI)) {
1442         gaddr = get_base_disp_rsy(cpu, run, &ar);
1443 
1444         return pcistb_service_call(cpu, r1, r3, gaddr, ar, RA_IGNORED);
1445     } else {
1446         return -1;
1447     }
1448 }
1449 
1450 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1451 {
1452     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1453     uint64_t fiba;
1454     uint8_t ar;
1455 
1456     if (s390_has_feat(S390_FEAT_ZPCI)) {
1457         fiba = get_base_disp_rxy(cpu, run, &ar);
1458 
1459         return mpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1460     } else {
1461         return -1;
1462     }
1463 }
1464 
1465 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1466 {
1467     int r = 0;
1468 
1469     switch (ipa1) {
1470     case PRIV_B9_CLP:
1471         r = kvm_clp_service_call(cpu, run);
1472         break;
1473     case PRIV_B9_PCISTG:
1474         r = kvm_pcistg_service_call(cpu, run);
1475         break;
1476     case PRIV_B9_PCILG:
1477         r = kvm_pcilg_service_call(cpu, run);
1478         break;
1479     case PRIV_B9_RPCIT:
1480         r = kvm_rpcit_service_call(cpu, run);
1481         break;
1482     case PRIV_B9_EQBS:
1483         /* just inject exception */
1484         r = -1;
1485         break;
1486     default:
1487         r = -1;
1488         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
1489         break;
1490     }
1491 
1492     return r;
1493 }
1494 
1495 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1496 {
1497     int r = 0;
1498 
1499     switch (ipbl) {
1500     case PRIV_EB_PCISTB:
1501         r = kvm_pcistb_service_call(cpu, run);
1502         break;
1503     case PRIV_EB_SIC:
1504         r = kvm_sic_service_call(cpu, run);
1505         break;
1506     case PRIV_EB_SQBS:
1507         /* just inject exception */
1508         r = -1;
1509         break;
1510     default:
1511         r = -1;
1512         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1513         break;
1514     }
1515 
1516     return r;
1517 }
1518 
1519 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1520 {
1521     int r = 0;
1522 
1523     switch (ipbl) {
1524     case PRIV_E3_MPCIFC:
1525         r = kvm_mpcifc_service_call(cpu, run);
1526         break;
1527     case PRIV_E3_STPCIFC:
1528         r = kvm_stpcifc_service_call(cpu, run);
1529         break;
1530     default:
1531         r = -1;
1532         DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1533         break;
1534     }
1535 
1536     return r;
1537 }
1538 
1539 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1540 {
1541     CPUS390XState *env = &cpu->env;
1542     int ret;
1543 
1544     ret = s390_virtio_hypercall(env);
1545     if (ret == -EINVAL) {
1546         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1547         return 0;
1548     }
1549 
1550     return ret;
1551 }
1552 
1553 static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
1554 {
1555     uint64_t r1, r3;
1556     int rc;
1557 
1558     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1559     r3 = run->s390_sieic.ipa & 0x000f;
1560     rc = handle_diag_288(&cpu->env, r1, r3);
1561     if (rc) {
1562         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1563     }
1564 }
1565 
1566 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1567 {
1568     uint64_t r1, r3;
1569 
1570     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1571     r3 = run->s390_sieic.ipa & 0x000f;
1572     handle_diag_308(&cpu->env, r1, r3, RA_IGNORED);
1573 }
1574 
1575 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1576 {
1577     CPUS390XState *env = &cpu->env;
1578     unsigned long pc;
1579 
1580     pc = env->psw.addr - sw_bp_ilen;
1581     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1582         env->psw.addr = pc;
1583         return EXCP_DEBUG;
1584     }
1585 
1586     return -ENOENT;
1587 }
1588 
1589 void kvm_s390_set_diag318(CPUState *cs, uint64_t diag318_info)
1590 {
1591     CPUS390XState *env = &S390_CPU(cs)->env;
1592 
1593     /* Feat bit is set only if KVM supports sync for diag318 */
1594     if (s390_has_feat(S390_FEAT_DIAG_318)) {
1595         env->diag318_info = diag318_info;
1596         cs->kvm_run->s.regs.diag318 = diag318_info;
1597         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_DIAG318;
1598         /*
1599          * diag 318 info is zeroed during a clear reset and
1600          * diag 308 IPL subcodes.
1601          */
1602     }
1603 }
1604 
1605 static void handle_diag_318(S390CPU *cpu, struct kvm_run *run)
1606 {
1607     uint64_t reg = (run->s390_sieic.ipa & 0x00f0) >> 4;
1608     uint64_t diag318_info = run->s.regs.gprs[reg];
1609     CPUState *t;
1610 
1611     /*
1612      * DIAG 318 can only be enabled with KVM support. As such, let's
1613      * ensure a guest cannot execute this instruction erroneously.
1614      */
1615     if (!s390_has_feat(S390_FEAT_DIAG_318)) {
1616         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1617         return;
1618     }
1619 
1620     CPU_FOREACH(t) {
1621         run_on_cpu(t, s390_do_cpu_set_diag318,
1622                    RUN_ON_CPU_HOST_ULONG(diag318_info));
1623     }
1624 }
1625 
1626 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1627 
1628 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1629 {
1630     int r = 0;
1631     uint16_t func_code;
1632 
1633     /*
1634      * For any diagnose call we support, bits 48-63 of the resulting
1635      * address specify the function code; the remainder is ignored.
1636      */
1637     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
1638     switch (func_code) {
1639     case DIAG_TIMEREVENT:
1640         kvm_handle_diag_288(cpu, run);
1641         break;
1642     case DIAG_IPL:
1643         kvm_handle_diag_308(cpu, run);
1644         break;
1645     case DIAG_SET_CONTROL_PROGRAM_CODES:
1646         handle_diag_318(cpu, run);
1647         break;
1648     case DIAG_KVM_HYPERCALL:
1649         r = handle_hypercall(cpu, run);
1650         break;
1651     case DIAG_KVM_BREAKPOINT:
1652         r = handle_sw_breakpoint(cpu, run);
1653         break;
1654     default:
1655         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1656         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1657         break;
1658     }
1659 
1660     return r;
1661 }
1662 
1663 static int kvm_s390_handle_sigp(S390CPU *cpu, uint8_t ipa1, uint32_t ipb)
1664 {
1665     CPUS390XState *env = &cpu->env;
1666     const uint8_t r1 = ipa1 >> 4;
1667     const uint8_t r3 = ipa1 & 0x0f;
1668     int ret;
1669     uint8_t order;
1670 
1671     /* get order code */
1672     order = decode_basedisp_rs(env, ipb, NULL) & SIGP_ORDER_MASK;
1673 
1674     ret = handle_sigp(env, order, r1, r3);
1675     setcc(cpu, ret);
1676     return 0;
1677 }
1678 
1679 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1680 {
1681     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1682     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1683     int r = -1;
1684 
1685     DPRINTF("handle_instruction 0x%x 0x%x\n",
1686             run->s390_sieic.ipa, run->s390_sieic.ipb);
1687     switch (ipa0) {
1688     case IPA0_B2:
1689         r = handle_b2(cpu, run, ipa1);
1690         break;
1691     case IPA0_B9:
1692         r = handle_b9(cpu, run, ipa1);
1693         break;
1694     case IPA0_EB:
1695         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1696         break;
1697     case IPA0_E3:
1698         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1699         break;
1700     case IPA0_DIAG:
1701         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1702         break;
1703     case IPA0_SIGP:
1704         r = kvm_s390_handle_sigp(cpu, ipa1, run->s390_sieic.ipb);
1705         break;
1706     }
1707 
1708     if (r < 0) {
1709         r = 0;
1710         kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1711     }
1712 
1713     return r;
1714 }
1715 
1716 static void unmanageable_intercept(S390CPU *cpu, S390CrashReason reason,
1717                                    int pswoffset)
1718 {
1719     CPUState *cs = CPU(cpu);
1720 
1721     s390_cpu_halt(cpu);
1722     cpu->env.crash_reason = reason;
1723     qemu_system_guest_panicked(cpu_get_crash_info(cs));
1724 }
1725 
1726 /* try to detect pgm check loops */
1727 static int handle_oper_loop(S390CPU *cpu, struct kvm_run *run)
1728 {
1729     CPUState *cs = CPU(cpu);
1730     PSW oldpsw, newpsw;
1731 
1732     newpsw.mask = ldq_phys(cs->as, cpu->env.psa +
1733                            offsetof(LowCore, program_new_psw));
1734     newpsw.addr = ldq_phys(cs->as, cpu->env.psa +
1735                            offsetof(LowCore, program_new_psw) + 8);
1736     oldpsw.mask  = run->psw_mask;
1737     oldpsw.addr  = run->psw_addr;
1738     /*
1739      * Avoid endless loops of operation exceptions, if the pgm new
1740      * PSW will cause a new operation exception.
1741      * The heuristic checks if the pgm new psw is within 6 bytes before
1742      * the faulting psw address (with same DAT, AS settings) and the
1743      * new psw is not a wait psw and the fault was not triggered by
1744      * problem state. In that case go into crashed state.
1745      */
1746 
1747     if (oldpsw.addr - newpsw.addr <= 6 &&
1748         !(newpsw.mask & PSW_MASK_WAIT) &&
1749         !(oldpsw.mask & PSW_MASK_PSTATE) &&
1750         (newpsw.mask & PSW_MASK_ASC) == (oldpsw.mask & PSW_MASK_ASC) &&
1751         (newpsw.mask & PSW_MASK_DAT) == (oldpsw.mask & PSW_MASK_DAT)) {
1752         unmanageable_intercept(cpu, S390_CRASH_REASON_OPINT_LOOP,
1753                                offsetof(LowCore, program_new_psw));
1754         return EXCP_HALTED;
1755     }
1756     return 0;
1757 }
1758 
1759 static int handle_intercept(S390CPU *cpu)
1760 {
1761     CPUState *cs = CPU(cpu);
1762     struct kvm_run *run = cs->kvm_run;
1763     int icpt_code = run->s390_sieic.icptcode;
1764     int r = 0;
1765 
1766     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code, (long)run->psw_addr);
1767     switch (icpt_code) {
1768         case ICPT_INSTRUCTION:
1769         case ICPT_PV_INSTR:
1770         case ICPT_PV_INSTR_NOTIFICATION:
1771             r = handle_instruction(cpu, run);
1772             break;
1773         case ICPT_PROGRAM:
1774             unmanageable_intercept(cpu, S390_CRASH_REASON_PGMINT_LOOP,
1775                                    offsetof(LowCore, program_new_psw));
1776             r = EXCP_HALTED;
1777             break;
1778         case ICPT_EXT_INT:
1779             unmanageable_intercept(cpu, S390_CRASH_REASON_EXTINT_LOOP,
1780                                    offsetof(LowCore, external_new_psw));
1781             r = EXCP_HALTED;
1782             break;
1783         case ICPT_WAITPSW:
1784             /* disabled wait, since enabled wait is handled in kernel */
1785             s390_handle_wait(cpu);
1786             r = EXCP_HALTED;
1787             break;
1788         case ICPT_CPU_STOP:
1789             do_stop_interrupt(&cpu->env);
1790             r = EXCP_HALTED;
1791             break;
1792         case ICPT_OPEREXC:
1793             /* check for break points */
1794             r = handle_sw_breakpoint(cpu, run);
1795             if (r == -ENOENT) {
1796                 /* Then check for potential pgm check loops */
1797                 r = handle_oper_loop(cpu, run);
1798                 if (r == 0) {
1799                     kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1800                 }
1801             }
1802             break;
1803         case ICPT_SOFT_INTERCEPT:
1804             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1805             exit(1);
1806             break;
1807         case ICPT_IO:
1808             fprintf(stderr, "KVM unimplemented icpt IO\n");
1809             exit(1);
1810             break;
1811         default:
1812             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1813             exit(1);
1814             break;
1815     }
1816 
1817     return r;
1818 }
1819 
1820 static int handle_tsch(S390CPU *cpu)
1821 {
1822     CPUState *cs = CPU(cpu);
1823     struct kvm_run *run = cs->kvm_run;
1824     int ret;
1825 
1826     ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb,
1827                              RA_IGNORED);
1828     if (ret < 0) {
1829         /*
1830          * Failure.
1831          * If an I/O interrupt had been dequeued, we have to reinject it.
1832          */
1833         if (run->s390_tsch.dequeued) {
1834             s390_io_interrupt(run->s390_tsch.subchannel_id,
1835                               run->s390_tsch.subchannel_nr,
1836                               run->s390_tsch.io_int_parm,
1837                               run->s390_tsch.io_int_word);
1838         }
1839         ret = 0;
1840     }
1841     return ret;
1842 }
1843 
1844 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr, uint8_t ar)
1845 {
1846     const MachineState *ms = MACHINE(qdev_get_machine());
1847     uint16_t conf_cpus = 0, reserved_cpus = 0;
1848     SysIB_322 sysib;
1849     int del, i;
1850 
1851     if (s390_is_pv()) {
1852         s390_cpu_pv_mem_read(cpu, 0, &sysib, sizeof(sysib));
1853     } else if (s390_cpu_virt_mem_read(cpu, addr, ar, &sysib, sizeof(sysib))) {
1854         return;
1855     }
1856     /* Shift the stack of Extended Names to prepare for our own data */
1857     memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1858             sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1859     /* First virt level, that doesn't provide Ext Names delimits stack. It is
1860      * assumed it's not capable of managing Extended Names for lower levels.
1861      */
1862     for (del = 1; del < sysib.count; del++) {
1863         if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1864             break;
1865         }
1866     }
1867     if (del < sysib.count) {
1868         memset(sysib.ext_names[del], 0,
1869                sizeof(sysib.ext_names[0]) * (sysib.count - del));
1870     }
1871 
1872     /* count the cpus and split them into configured and reserved ones */
1873     for (i = 0; i < ms->possible_cpus->len; i++) {
1874         if (ms->possible_cpus->cpus[i].cpu) {
1875             conf_cpus++;
1876         } else {
1877             reserved_cpus++;
1878         }
1879     }
1880     sysib.vm[0].total_cpus = conf_cpus + reserved_cpus;
1881     sysib.vm[0].conf_cpus = conf_cpus;
1882     sysib.vm[0].reserved_cpus = reserved_cpus;
1883 
1884     /* Insert short machine name in EBCDIC, padded with blanks */
1885     if (qemu_name) {
1886         memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1887         ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1888                                                     strlen(qemu_name)));
1889     }
1890     sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1891     /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1892      * considered by s390 as not capable of providing any Extended Name.
1893      * Therefore if no name was specified on qemu invocation, we go with the
1894      * same "KVMguest" default, which KVM has filled into short name field.
1895      */
1896     strpadcpy((char *)sysib.ext_names[0],
1897               sizeof(sysib.ext_names[0]),
1898               qemu_name ?: "KVMguest", '\0');
1899 
1900     /* Insert UUID */
1901     memcpy(sysib.vm[0].uuid, &qemu_uuid, sizeof(sysib.vm[0].uuid));
1902 
1903     if (s390_is_pv()) {
1904         s390_cpu_pv_mem_write(cpu, 0, &sysib, sizeof(sysib));
1905     } else {
1906         s390_cpu_virt_mem_write(cpu, addr, ar, &sysib, sizeof(sysib));
1907     }
1908 }
1909 
1910 static int handle_stsi(S390CPU *cpu)
1911 {
1912     CPUState *cs = CPU(cpu);
1913     struct kvm_run *run = cs->kvm_run;
1914 
1915     switch (run->s390_stsi.fc) {
1916     case 3:
1917         if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1918             return 0;
1919         }
1920         /* Only sysib 3.2.2 needs post-handling for now. */
1921         insert_stsi_3_2_2(cpu, run->s390_stsi.addr, run->s390_stsi.ar);
1922         return 0;
1923     default:
1924         return 0;
1925     }
1926 }
1927 
1928 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1929 {
1930     CPUState *cs = CPU(cpu);
1931     struct kvm_run *run = cs->kvm_run;
1932 
1933     int ret = 0;
1934     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1935 
1936     switch (arch_info->type) {
1937     case KVM_HW_WP_WRITE:
1938         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1939             cs->watchpoint_hit = &hw_watchpoint;
1940             hw_watchpoint.vaddr = arch_info->addr;
1941             hw_watchpoint.flags = BP_MEM_WRITE;
1942             ret = EXCP_DEBUG;
1943         }
1944         break;
1945     case KVM_HW_BP:
1946         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1947             ret = EXCP_DEBUG;
1948         }
1949         break;
1950     case KVM_SINGLESTEP:
1951         if (cs->singlestep_enabled) {
1952             ret = EXCP_DEBUG;
1953         }
1954         break;
1955     default:
1956         ret = -ENOSYS;
1957     }
1958 
1959     return ret;
1960 }
1961 
1962 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1963 {
1964     S390CPU *cpu = S390_CPU(cs);
1965     int ret = 0;
1966 
1967     qemu_mutex_lock_iothread();
1968 
1969     kvm_cpu_synchronize_state(cs);
1970 
1971     switch (run->exit_reason) {
1972         case KVM_EXIT_S390_SIEIC:
1973             ret = handle_intercept(cpu);
1974             break;
1975         case KVM_EXIT_S390_RESET:
1976             s390_ipl_reset_request(cs, S390_RESET_REIPL);
1977             break;
1978         case KVM_EXIT_S390_TSCH:
1979             ret = handle_tsch(cpu);
1980             break;
1981         case KVM_EXIT_S390_STSI:
1982             ret = handle_stsi(cpu);
1983             break;
1984         case KVM_EXIT_DEBUG:
1985             ret = kvm_arch_handle_debug_exit(cpu);
1986             break;
1987         default:
1988             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1989             break;
1990     }
1991     qemu_mutex_unlock_iothread();
1992 
1993     if (ret == 0) {
1994         ret = EXCP_INTERRUPT;
1995     }
1996     return ret;
1997 }
1998 
1999 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
2000 {
2001     return true;
2002 }
2003 
2004 void kvm_s390_enable_css_support(S390CPU *cpu)
2005 {
2006     int r;
2007 
2008     /* Activate host kernel channel subsystem support. */
2009     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
2010     assert(r == 0);
2011 }
2012 
2013 void kvm_arch_init_irq_routing(KVMState *s)
2014 {
2015     /*
2016      * Note that while irqchip capabilities generally imply that cpustates
2017      * are handled in-kernel, it is not true for s390 (yet); therefore, we
2018      * have to override the common code kvm_halt_in_kernel_allowed setting.
2019      */
2020     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
2021         kvm_gsi_routing_allowed = true;
2022         kvm_halt_in_kernel_allowed = false;
2023     }
2024 }
2025 
2026 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
2027                                     int vq, bool assign)
2028 {
2029     struct kvm_ioeventfd kick = {
2030         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
2031         KVM_IOEVENTFD_FLAG_DATAMATCH,
2032         .fd = event_notifier_get_fd(notifier),
2033         .datamatch = vq,
2034         .addr = sch,
2035         .len = 8,
2036     };
2037     trace_kvm_assign_subch_ioeventfd(kick.fd, kick.addr, assign,
2038                                      kick.datamatch);
2039     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
2040         return -ENOSYS;
2041     }
2042     if (!assign) {
2043         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
2044     }
2045     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
2046 }
2047 
2048 int kvm_s390_get_ri(void)
2049 {
2050     return cap_ri;
2051 }
2052 
2053 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
2054 {
2055     struct kvm_mp_state mp_state = {};
2056     int ret;
2057 
2058     /* the kvm part might not have been initialized yet */
2059     if (CPU(cpu)->kvm_state == NULL) {
2060         return 0;
2061     }
2062 
2063     switch (cpu_state) {
2064     case S390_CPU_STATE_STOPPED:
2065         mp_state.mp_state = KVM_MP_STATE_STOPPED;
2066         break;
2067     case S390_CPU_STATE_CHECK_STOP:
2068         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
2069         break;
2070     case S390_CPU_STATE_OPERATING:
2071         mp_state.mp_state = KVM_MP_STATE_OPERATING;
2072         break;
2073     case S390_CPU_STATE_LOAD:
2074         mp_state.mp_state = KVM_MP_STATE_LOAD;
2075         break;
2076     default:
2077         error_report("Requested CPU state is not a valid S390 CPU state: %u",
2078                      cpu_state);
2079         exit(1);
2080     }
2081 
2082     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
2083     if (ret) {
2084         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
2085                                        strerror(-ret));
2086     }
2087 
2088     return ret;
2089 }
2090 
2091 void kvm_s390_vcpu_interrupt_pre_save(S390CPU *cpu)
2092 {
2093     unsigned int max_cpus = MACHINE(qdev_get_machine())->smp.max_cpus;
2094     struct kvm_s390_irq_state irq_state = {
2095         .buf = (uint64_t) cpu->irqstate,
2096         .len = VCPU_IRQ_BUF_SIZE(max_cpus),
2097     };
2098     CPUState *cs = CPU(cpu);
2099     int32_t bytes;
2100 
2101     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2102         return;
2103     }
2104 
2105     bytes = kvm_vcpu_ioctl(cs, KVM_S390_GET_IRQ_STATE, &irq_state);
2106     if (bytes < 0) {
2107         cpu->irqstate_saved_size = 0;
2108         error_report("Migration of interrupt state failed");
2109         return;
2110     }
2111 
2112     cpu->irqstate_saved_size = bytes;
2113 }
2114 
2115 int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
2116 {
2117     CPUState *cs = CPU(cpu);
2118     struct kvm_s390_irq_state irq_state = {
2119         .buf = (uint64_t) cpu->irqstate,
2120         .len = cpu->irqstate_saved_size,
2121     };
2122     int r;
2123 
2124     if (cpu->irqstate_saved_size == 0) {
2125         return 0;
2126     }
2127 
2128     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2129         return -ENOSYS;
2130     }
2131 
2132     r = kvm_vcpu_ioctl(cs, KVM_S390_SET_IRQ_STATE, &irq_state);
2133     if (r) {
2134         error_report("Setting interrupt state failed %d", r);
2135     }
2136     return r;
2137 }
2138 
2139 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
2140                              uint64_t address, uint32_t data, PCIDevice *dev)
2141 {
2142     S390PCIBusDevice *pbdev;
2143     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
2144 
2145     if (!dev) {
2146         DPRINTF("add_msi_route no pci device\n");
2147         return -ENODEV;
2148     }
2149 
2150     pbdev = s390_pci_find_dev_by_target(s390_get_phb(), DEVICE(dev)->id);
2151     if (!pbdev) {
2152         DPRINTF("add_msi_route no zpci device\n");
2153         return -ENODEV;
2154     }
2155 
2156     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
2157     route->flags = 0;
2158     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
2159     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
2160     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
2161     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset + vec;
2162     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
2163     return 0;
2164 }
2165 
2166 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
2167                                 int vector, PCIDevice *dev)
2168 {
2169     return 0;
2170 }
2171 
2172 int kvm_arch_release_virq_post(int virq)
2173 {
2174     return 0;
2175 }
2176 
2177 int kvm_arch_msi_data_to_gsi(uint32_t data)
2178 {
2179     abort();
2180 }
2181 
2182 static int query_cpu_subfunc(S390FeatBitmap features)
2183 {
2184     struct kvm_s390_vm_cpu_subfunc prop = {};
2185     struct kvm_device_attr attr = {
2186         .group = KVM_S390_VM_CPU_MODEL,
2187         .attr = KVM_S390_VM_CPU_MACHINE_SUBFUNC,
2188         .addr = (uint64_t) &prop,
2189     };
2190     int rc;
2191 
2192     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2193     if (rc) {
2194         return  rc;
2195     }
2196 
2197     /*
2198      * We're going to add all subfunctions now, if the corresponding feature
2199      * is available that unlocks the query functions.
2200      */
2201     s390_add_from_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2202     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2203         s390_add_from_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2204     }
2205     if (test_bit(S390_FEAT_MSA, features)) {
2206         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2207         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2208         s390_add_from_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2209         s390_add_from_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2210         s390_add_from_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2211     }
2212     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2213         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2214     }
2215     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2216         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2217         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2218         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2219         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2220     }
2221     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2222         s390_add_from_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2223     }
2224     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2225         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2226     }
2227     if (test_bit(S390_FEAT_MSA_EXT_9, features)) {
2228         s390_add_from_feat_block(features, S390_FEAT_TYPE_KDSA, prop.kdsa);
2229     }
2230     if (test_bit(S390_FEAT_ESORT_BASE, features)) {
2231         s390_add_from_feat_block(features, S390_FEAT_TYPE_SORTL, prop.sortl);
2232     }
2233     if (test_bit(S390_FEAT_DEFLATE_BASE, features)) {
2234         s390_add_from_feat_block(features, S390_FEAT_TYPE_DFLTCC, prop.dfltcc);
2235     }
2236     return 0;
2237 }
2238 
2239 static int configure_cpu_subfunc(const S390FeatBitmap features)
2240 {
2241     struct kvm_s390_vm_cpu_subfunc prop = {};
2242     struct kvm_device_attr attr = {
2243         .group = KVM_S390_VM_CPU_MODEL,
2244         .attr = KVM_S390_VM_CPU_PROCESSOR_SUBFUNC,
2245         .addr = (uint64_t) &prop,
2246     };
2247 
2248     if (!kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2249                            KVM_S390_VM_CPU_PROCESSOR_SUBFUNC)) {
2250         /* hardware support might be missing, IBC will handle most of this */
2251         return 0;
2252     }
2253 
2254     s390_fill_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2255     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2256         s390_fill_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2257     }
2258     if (test_bit(S390_FEAT_MSA, features)) {
2259         s390_fill_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2260         s390_fill_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2261         s390_fill_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2262         s390_fill_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2263         s390_fill_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2264     }
2265     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2266         s390_fill_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2267     }
2268     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2269         s390_fill_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2270         s390_fill_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2271         s390_fill_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2272         s390_fill_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2273     }
2274     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2275         s390_fill_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2276     }
2277     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2278         s390_fill_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2279     }
2280     if (test_bit(S390_FEAT_MSA_EXT_9, features)) {
2281         s390_fill_feat_block(features, S390_FEAT_TYPE_KDSA, prop.kdsa);
2282     }
2283     if (test_bit(S390_FEAT_ESORT_BASE, features)) {
2284         s390_fill_feat_block(features, S390_FEAT_TYPE_SORTL, prop.sortl);
2285     }
2286     if (test_bit(S390_FEAT_DEFLATE_BASE, features)) {
2287         s390_fill_feat_block(features, S390_FEAT_TYPE_DFLTCC, prop.dfltcc);
2288     }
2289     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2290 }
2291 
2292 static int kvm_to_feat[][2] = {
2293     { KVM_S390_VM_CPU_FEAT_ESOP, S390_FEAT_ESOP },
2294     { KVM_S390_VM_CPU_FEAT_SIEF2, S390_FEAT_SIE_F2 },
2295     { KVM_S390_VM_CPU_FEAT_64BSCAO , S390_FEAT_SIE_64BSCAO },
2296     { KVM_S390_VM_CPU_FEAT_SIIF, S390_FEAT_SIE_SIIF },
2297     { KVM_S390_VM_CPU_FEAT_GPERE, S390_FEAT_SIE_GPERE },
2298     { KVM_S390_VM_CPU_FEAT_GSLS, S390_FEAT_SIE_GSLS },
2299     { KVM_S390_VM_CPU_FEAT_IB, S390_FEAT_SIE_IB },
2300     { KVM_S390_VM_CPU_FEAT_CEI, S390_FEAT_SIE_CEI },
2301     { KVM_S390_VM_CPU_FEAT_IBS, S390_FEAT_SIE_IBS },
2302     { KVM_S390_VM_CPU_FEAT_SKEY, S390_FEAT_SIE_SKEY },
2303     { KVM_S390_VM_CPU_FEAT_CMMA, S390_FEAT_SIE_CMMA },
2304     { KVM_S390_VM_CPU_FEAT_PFMFI, S390_FEAT_SIE_PFMFI},
2305     { KVM_S390_VM_CPU_FEAT_SIGPIF, S390_FEAT_SIE_SIGPIF},
2306     { KVM_S390_VM_CPU_FEAT_KSS, S390_FEAT_SIE_KSS},
2307 };
2308 
2309 static int query_cpu_feat(S390FeatBitmap features)
2310 {
2311     struct kvm_s390_vm_cpu_feat prop = {};
2312     struct kvm_device_attr attr = {
2313         .group = KVM_S390_VM_CPU_MODEL,
2314         .attr = KVM_S390_VM_CPU_MACHINE_FEAT,
2315         .addr = (uint64_t) &prop,
2316     };
2317     int rc;
2318     int i;
2319 
2320     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2321     if (rc) {
2322         return  rc;
2323     }
2324 
2325     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2326         if (test_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat)) {
2327             set_bit(kvm_to_feat[i][1], features);
2328         }
2329     }
2330     return 0;
2331 }
2332 
2333 static int configure_cpu_feat(const S390FeatBitmap features)
2334 {
2335     struct kvm_s390_vm_cpu_feat prop = {};
2336     struct kvm_device_attr attr = {
2337         .group = KVM_S390_VM_CPU_MODEL,
2338         .attr = KVM_S390_VM_CPU_PROCESSOR_FEAT,
2339         .addr = (uint64_t) &prop,
2340     };
2341     int i;
2342 
2343     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2344         if (test_bit(kvm_to_feat[i][1], features)) {
2345             set_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat);
2346         }
2347     }
2348     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2349 }
2350 
2351 bool kvm_s390_cpu_models_supported(void)
2352 {
2353     if (!cpu_model_allowed()) {
2354         /* compatibility machines interfere with the cpu model */
2355         return false;
2356     }
2357     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2358                              KVM_S390_VM_CPU_MACHINE) &&
2359            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2360                              KVM_S390_VM_CPU_PROCESSOR) &&
2361            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2362                              KVM_S390_VM_CPU_MACHINE_FEAT) &&
2363            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2364                              KVM_S390_VM_CPU_PROCESSOR_FEAT) &&
2365            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2366                              KVM_S390_VM_CPU_MACHINE_SUBFUNC);
2367 }
2368 
2369 void kvm_s390_get_host_cpu_model(S390CPUModel *model, Error **errp)
2370 {
2371     struct kvm_s390_vm_cpu_machine prop = {};
2372     struct kvm_device_attr attr = {
2373         .group = KVM_S390_VM_CPU_MODEL,
2374         .attr = KVM_S390_VM_CPU_MACHINE,
2375         .addr = (uint64_t) &prop,
2376     };
2377     uint16_t unblocked_ibc = 0, cpu_type = 0;
2378     int rc;
2379 
2380     memset(model, 0, sizeof(*model));
2381 
2382     if (!kvm_s390_cpu_models_supported()) {
2383         error_setg(errp, "KVM doesn't support CPU models");
2384         return;
2385     }
2386 
2387     /* query the basic cpu model properties */
2388     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2389     if (rc) {
2390         error_setg(errp, "KVM: Error querying host CPU model: %d", rc);
2391         return;
2392     }
2393 
2394     cpu_type = cpuid_type(prop.cpuid);
2395     if (has_ibc(prop.ibc)) {
2396         model->lowest_ibc = lowest_ibc(prop.ibc);
2397         unblocked_ibc = unblocked_ibc(prop.ibc);
2398     }
2399     model->cpu_id = cpuid_id(prop.cpuid);
2400     model->cpu_id_format = cpuid_format(prop.cpuid);
2401     model->cpu_ver = 0xff;
2402 
2403     /* get supported cpu features indicated via STFL(E) */
2404     s390_add_from_feat_block(model->features, S390_FEAT_TYPE_STFL,
2405                              (uint8_t *) prop.fac_mask);
2406     /* dat-enhancement facility 2 has no bit but was introduced with stfle */
2407     if (test_bit(S390_FEAT_STFLE, model->features)) {
2408         set_bit(S390_FEAT_DAT_ENH_2, model->features);
2409     }
2410     /* get supported cpu features indicated e.g. via SCLP */
2411     rc = query_cpu_feat(model->features);
2412     if (rc) {
2413         error_setg(errp, "KVM: Error querying CPU features: %d", rc);
2414         return;
2415     }
2416     /* get supported cpu subfunctions indicated via query / test bit */
2417     rc = query_cpu_subfunc(model->features);
2418     if (rc) {
2419         error_setg(errp, "KVM: Error querying CPU subfunctions: %d", rc);
2420         return;
2421     }
2422 
2423     /* PTFF subfunctions might be indicated although kernel support missing */
2424     if (!test_bit(S390_FEAT_MULTIPLE_EPOCH, model->features)) {
2425         clear_bit(S390_FEAT_PTFF_QSIE, model->features);
2426         clear_bit(S390_FEAT_PTFF_QTOUE, model->features);
2427         clear_bit(S390_FEAT_PTFF_STOE, model->features);
2428         clear_bit(S390_FEAT_PTFF_STOUE, model->features);
2429     }
2430 
2431     /* with cpu model support, CMM is only indicated if really available */
2432     if (kvm_s390_cmma_available()) {
2433         set_bit(S390_FEAT_CMM, model->features);
2434     } else {
2435         /* no cmm -> no cmm nt */
2436         clear_bit(S390_FEAT_CMM_NT, model->features);
2437     }
2438 
2439     /* bpb needs kernel support for migration, VSIE and reset */
2440     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_BPB)) {
2441         clear_bit(S390_FEAT_BPB, model->features);
2442     }
2443 
2444     /*
2445      * If we have support for protected virtualization, indicate
2446      * the protected virtualization IPL unpack facility.
2447      */
2448     if (cap_protected) {
2449         set_bit(S390_FEAT_UNPACK, model->features);
2450     }
2451 
2452     /* We emulate a zPCI bus and AEN, therefore we don't need HW support */
2453     set_bit(S390_FEAT_ZPCI, model->features);
2454     set_bit(S390_FEAT_ADAPTER_EVENT_NOTIFICATION, model->features);
2455 
2456     if (s390_known_cpu_type(cpu_type)) {
2457         /* we want the exact model, even if some features are missing */
2458         model->def = s390_find_cpu_def(cpu_type, ibc_gen(unblocked_ibc),
2459                                        ibc_ec_ga(unblocked_ibc), NULL);
2460     } else {
2461         /* model unknown, e.g. too new - search using features */
2462         model->def = s390_find_cpu_def(0, ibc_gen(unblocked_ibc),
2463                                        ibc_ec_ga(unblocked_ibc),
2464                                        model->features);
2465     }
2466     if (!model->def) {
2467         error_setg(errp, "KVM: host CPU model could not be identified");
2468         return;
2469     }
2470     /* for now, we can only provide the AP feature with HW support */
2471     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO,
2472         KVM_S390_VM_CRYPTO_ENABLE_APIE)) {
2473         set_bit(S390_FEAT_AP, model->features);
2474     }
2475 
2476     /*
2477      * Extended-Length SCCB is handled entirely within QEMU.
2478      * For PV guests this is completely fenced by the Ultravisor, as Service
2479      * Call error checking and STFLE interpretation are handled via SIE.
2480      */
2481     set_bit(S390_FEAT_EXTENDED_LENGTH_SCCB, model->features);
2482 
2483     if (kvm_check_extension(kvm_state, KVM_CAP_S390_DIAG318)) {
2484         set_bit(S390_FEAT_DIAG_318, model->features);
2485     }
2486 
2487     /* strip of features that are not part of the maximum model */
2488     bitmap_and(model->features, model->features, model->def->full_feat,
2489                S390_FEAT_MAX);
2490 }
2491 
2492 static void kvm_s390_configure_apie(bool interpret)
2493 {
2494     uint64_t attr = interpret ? KVM_S390_VM_CRYPTO_ENABLE_APIE :
2495                                 KVM_S390_VM_CRYPTO_DISABLE_APIE;
2496 
2497     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
2498         kvm_s390_set_attr(attr);
2499     }
2500 }
2501 
2502 void kvm_s390_apply_cpu_model(const S390CPUModel *model, Error **errp)
2503 {
2504     struct kvm_s390_vm_cpu_processor prop  = {
2505         .fac_list = { 0 },
2506     };
2507     struct kvm_device_attr attr = {
2508         .group = KVM_S390_VM_CPU_MODEL,
2509         .attr = KVM_S390_VM_CPU_PROCESSOR,
2510         .addr = (uint64_t) &prop,
2511     };
2512     int rc;
2513 
2514     if (!model) {
2515         /* compatibility handling if cpu models are disabled */
2516         if (kvm_s390_cmma_available()) {
2517             kvm_s390_enable_cmma();
2518         }
2519         return;
2520     }
2521     if (!kvm_s390_cpu_models_supported()) {
2522         error_setg(errp, "KVM doesn't support CPU models");
2523         return;
2524     }
2525     prop.cpuid = s390_cpuid_from_cpu_model(model);
2526     prop.ibc = s390_ibc_from_cpu_model(model);
2527     /* configure cpu features indicated via STFL(e) */
2528     s390_fill_feat_block(model->features, S390_FEAT_TYPE_STFL,
2529                          (uint8_t *) prop.fac_list);
2530     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2531     if (rc) {
2532         error_setg(errp, "KVM: Error configuring the CPU model: %d", rc);
2533         return;
2534     }
2535     /* configure cpu features indicated e.g. via SCLP */
2536     rc = configure_cpu_feat(model->features);
2537     if (rc) {
2538         error_setg(errp, "KVM: Error configuring CPU features: %d", rc);
2539         return;
2540     }
2541     /* configure cpu subfunctions indicated via query / test bit */
2542     rc = configure_cpu_subfunc(model->features);
2543     if (rc) {
2544         error_setg(errp, "KVM: Error configuring CPU subfunctions: %d", rc);
2545         return;
2546     }
2547     /* enable CMM via CMMA */
2548     if (test_bit(S390_FEAT_CMM, model->features)) {
2549         kvm_s390_enable_cmma();
2550     }
2551 
2552     if (test_bit(S390_FEAT_AP, model->features)) {
2553         kvm_s390_configure_apie(true);
2554     }
2555 }
2556 
2557 void kvm_s390_restart_interrupt(S390CPU *cpu)
2558 {
2559     struct kvm_s390_irq irq = {
2560         .type = KVM_S390_RESTART,
2561     };
2562 
2563     kvm_s390_vcpu_interrupt(cpu, &irq);
2564 }
2565 
2566 void kvm_s390_stop_interrupt(S390CPU *cpu)
2567 {
2568     struct kvm_s390_irq irq = {
2569         .type = KVM_S390_SIGP_STOP,
2570     };
2571 
2572     kvm_s390_vcpu_interrupt(cpu, &irq);
2573 }
2574 
2575 bool kvm_arch_cpu_check_are_resettable(void)
2576 {
2577     return true;
2578 }
2579 
2580 int kvm_s390_get_zpci_op(void)
2581 {
2582     return cap_zpci_op;
2583 }
2584