xref: /qemu/target/s390x/kvm/kvm.c (revision 4be0fce498d0a08f18b3a9accdb9ded79484d30a)
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 "gdbstub/enums.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 "target/s390x/kvm/pv.h"
54 #include CONFIG_DEVICES
55 
56 #define kvm_vm_check_mem_attr(s, attr) \
57     kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
58 
59 #define IPA0_DIAG                       0x8300
60 #define IPA0_SIGP                       0xae00
61 #define IPA0_B2                         0xb200
62 #define IPA0_B9                         0xb900
63 #define IPA0_EB                         0xeb00
64 #define IPA0_E3                         0xe300
65 
66 #define PRIV_B2_SCLP_CALL               0x20
67 #define PRIV_B2_CSCH                    0x30
68 #define PRIV_B2_HSCH                    0x31
69 #define PRIV_B2_MSCH                    0x32
70 #define PRIV_B2_SSCH                    0x33
71 #define PRIV_B2_STSCH                   0x34
72 #define PRIV_B2_TSCH                    0x35
73 #define PRIV_B2_TPI                     0x36
74 #define PRIV_B2_SAL                     0x37
75 #define PRIV_B2_RSCH                    0x38
76 #define PRIV_B2_STCRW                   0x39
77 #define PRIV_B2_STCPS                   0x3a
78 #define PRIV_B2_RCHP                    0x3b
79 #define PRIV_B2_SCHM                    0x3c
80 #define PRIV_B2_CHSC                    0x5f
81 #define PRIV_B2_SIGA                    0x74
82 #define PRIV_B2_XSCH                    0x76
83 
84 #define PRIV_EB_SQBS                    0x8a
85 #define PRIV_EB_PCISTB                  0xd0
86 #define PRIV_EB_SIC                     0xd1
87 
88 #define PRIV_B9_EQBS                    0x9c
89 #define PRIV_B9_CLP                     0xa0
90 #define PRIV_B9_PTF                     0xa2
91 #define PRIV_B9_PCISTG                  0xd0
92 #define PRIV_B9_PCILG                   0xd2
93 #define PRIV_B9_RPCIT                   0xd3
94 
95 #define PRIV_E3_MPCIFC                  0xd0
96 #define PRIV_E3_STPCIFC                 0xd4
97 
98 #define DIAG_TIMEREVENT                 0x288
99 #define DIAG_IPL                        0x308
100 #define DIAG_SET_CONTROL_PROGRAM_CODES  0x318
101 #define DIAG_KVM_HYPERCALL              0x500
102 #define DIAG_KVM_BREAKPOINT             0x501
103 
104 #define ICPT_INSTRUCTION                0x04
105 #define ICPT_PROGRAM                    0x08
106 #define ICPT_EXT_INT                    0x14
107 #define ICPT_WAITPSW                    0x1c
108 #define ICPT_SOFT_INTERCEPT             0x24
109 #define ICPT_CPU_STOP                   0x28
110 #define ICPT_OPEREXC                    0x2c
111 #define ICPT_IO                         0x40
112 #define ICPT_PV_INSTR                   0x68
113 #define ICPT_PV_INSTR_NOTIFICATION      0x6c
114 
115 #define NR_LOCAL_IRQS 32
116 /*
117  * Needs to be big enough to contain max_cpus emergency signals
118  * and in addition NR_LOCAL_IRQS interrupts
119  */
120 #define VCPU_IRQ_BUF_SIZE(max_cpus) (sizeof(struct kvm_s390_irq) * \
121                                      (max_cpus + NR_LOCAL_IRQS))
122 /*
123  * KVM does only support memory slots up to KVM_MEM_MAX_NR_PAGES pages
124  * as the dirty bitmap must be managed by bitops that take an int as
125  * position indicator. This would end at an unaligned  address
126  * (0x7fffff00000). As future variants might provide larger pages
127  * and to make all addresses properly aligned, let us split at 4TB.
128  */
129 #define KVM_SLOT_MAX_BYTES (4UL * TiB)
130 
131 static CPUWatchpoint hw_watchpoint;
132 /*
133  * We don't use a list because this structure is also used to transmit the
134  * hardware breakpoints to the kernel.
135  */
136 static struct kvm_hw_breakpoint *hw_breakpoints;
137 static int nb_hw_breakpoints;
138 
139 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
140     KVM_CAP_LAST_INFO
141 };
142 
143 static int cap_async_pf;
144 static int cap_mem_op;
145 static int cap_mem_op_extension;
146 static int cap_s390_irq;
147 static int cap_ri;
148 static int cap_hpage_1m;
149 static int cap_vcpu_resets;
150 static int cap_protected;
151 static int cap_zpci_op;
152 static int cap_protected_dump;
153 
154 static bool mem_op_storage_key_support;
155 
156 static int active_cmma;
157 
158 static int kvm_s390_query_mem_limit(uint64_t *memory_limit)
159 {
160     struct kvm_device_attr attr = {
161         .group = KVM_S390_VM_MEM_CTRL,
162         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
163         .addr = (uint64_t) memory_limit,
164     };
165 
166     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
167 }
168 
169 int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit)
170 {
171     int rc;
172 
173     struct kvm_device_attr attr = {
174         .group = KVM_S390_VM_MEM_CTRL,
175         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
176         .addr = (uint64_t) &new_limit,
177     };
178 
179     if (!kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_LIMIT_SIZE)) {
180         return 0;
181     }
182 
183     rc = kvm_s390_query_mem_limit(hw_limit);
184     if (rc) {
185         return rc;
186     } else if (*hw_limit < new_limit) {
187         return -E2BIG;
188     }
189 
190     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
191 }
192 
193 int kvm_s390_cmma_active(void)
194 {
195     return active_cmma;
196 }
197 
198 static bool kvm_s390_cmma_available(void)
199 {
200     static bool initialized, value;
201 
202     if (!initialized) {
203         initialized = true;
204         value = kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_ENABLE_CMMA) &&
205                 kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_CLR_CMMA);
206     }
207     return value;
208 }
209 
210 void kvm_s390_cmma_reset(void)
211 {
212     int rc;
213     struct kvm_device_attr attr = {
214         .group = KVM_S390_VM_MEM_CTRL,
215         .attr = KVM_S390_VM_MEM_CLR_CMMA,
216     };
217 
218     if (!kvm_s390_cmma_active()) {
219         return;
220     }
221 
222     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
223     trace_kvm_clear_cmma(rc);
224 }
225 
226 static void kvm_s390_enable_cmma(void)
227 {
228     int rc;
229     struct kvm_device_attr attr = {
230         .group = KVM_S390_VM_MEM_CTRL,
231         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
232     };
233 
234     if (cap_hpage_1m) {
235         warn_report("CMM will not be enabled because it is not "
236                     "compatible with huge memory backings.");
237         return;
238     }
239     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
240     active_cmma = !rc;
241     trace_kvm_enable_cmma(rc);
242 }
243 
244 static void kvm_s390_set_crypto_attr(uint64_t attr)
245 {
246     struct kvm_device_attr attribute = {
247         .group = KVM_S390_VM_CRYPTO,
248         .attr  = attr,
249     };
250 
251     int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
252 
253     if (ret) {
254         error_report("Failed to set crypto device attribute %lu: %s",
255                      attr, strerror(-ret));
256     }
257 }
258 
259 static void kvm_s390_init_aes_kw(void)
260 {
261     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
262 
263     if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
264                                  NULL)) {
265             attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
266     }
267 
268     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
269             kvm_s390_set_crypto_attr(attr);
270     }
271 }
272 
273 static void kvm_s390_init_dea_kw(void)
274 {
275     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
276 
277     if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
278                                  NULL)) {
279             attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
280     }
281 
282     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
283             kvm_s390_set_crypto_attr(attr);
284     }
285 }
286 
287 void kvm_s390_crypto_reset(void)
288 {
289     if (s390_has_feat(S390_FEAT_MSA_EXT_3)) {
290         kvm_s390_init_aes_kw();
291         kvm_s390_init_dea_kw();
292     }
293 }
294 
295 void kvm_s390_set_max_pagesize(uint64_t pagesize, Error **errp)
296 {
297     if (pagesize == 4 * KiB) {
298         return;
299     }
300 
301     if (!hpage_1m_allowed()) {
302         error_setg(errp, "This QEMU machine does not support huge page "
303                    "mappings");
304         return;
305     }
306 
307     if (pagesize != 1 * MiB) {
308         error_setg(errp, "Memory backing with 2G pages was specified, "
309                    "but KVM does not support this memory backing");
310         return;
311     }
312 
313     if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_HPAGE_1M, 0)) {
314         error_setg(errp, "Memory backing with 1M pages was specified, "
315                    "but KVM does not support this memory backing");
316         return;
317     }
318 
319     cap_hpage_1m = 1;
320 }
321 
322 int kvm_s390_get_hpage_1m(void)
323 {
324     return cap_hpage_1m;
325 }
326 
327 static void ccw_machine_class_foreach(ObjectClass *oc, void *opaque)
328 {
329     MachineClass *mc = MACHINE_CLASS(oc);
330 
331     mc->default_cpu_type = S390_CPU_TYPE_NAME("host");
332 }
333 
334 int kvm_arch_get_default_type(MachineState *ms)
335 {
336     return 0;
337 }
338 
339 int kvm_arch_init(MachineState *ms, KVMState *s)
340 {
341     int required_caps[] = {
342         KVM_CAP_DEVICE_CTRL,
343         KVM_CAP_SYNC_REGS,
344     };
345 
346     for (int i = 0; i < ARRAY_SIZE(required_caps); i++) {
347         if (!kvm_check_extension(s, required_caps[i])) {
348             error_report("KVM is missing capability #%d - "
349                          "please use kernel 3.15 or newer", required_caps[i]);
350             return -1;
351         }
352     }
353 
354     object_class_foreach(ccw_machine_class_foreach, TYPE_S390_CCW_MACHINE,
355                          false, NULL);
356 
357     if (!kvm_check_extension(s, KVM_CAP_S390_COW)) {
358         error_report("KVM is missing capability KVM_CAP_S390_COW - "
359                      "unsupported environment");
360         return -1;
361     }
362 
363     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
364     cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
365     cap_mem_op_extension = kvm_check_extension(s, KVM_CAP_S390_MEM_OP_EXTENSION);
366     mem_op_storage_key_support = cap_mem_op_extension > 0;
367     cap_s390_irq = kvm_check_extension(s, KVM_CAP_S390_INJECT_IRQ);
368     cap_vcpu_resets = kvm_check_extension(s, KVM_CAP_S390_VCPU_RESETS);
369     cap_protected = kvm_check_extension(s, KVM_CAP_S390_PROTECTED);
370     cap_zpci_op = kvm_check_extension(s, KVM_CAP_S390_ZPCI_OP);
371     cap_protected_dump = kvm_check_extension(s, KVM_CAP_S390_PROTECTED_DUMP);
372 
373     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
374     kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
375     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
376     kvm_vm_enable_cap(s, KVM_CAP_S390_CPU_TOPOLOGY, 0);
377     if (ri_allowed()) {
378         if (kvm_vm_enable_cap(s, KVM_CAP_S390_RI, 0) == 0) {
379             cap_ri = 1;
380         }
381     }
382     if (cpu_model_allowed()) {
383         kvm_vm_enable_cap(s, KVM_CAP_S390_GS, 0);
384     }
385 
386     /*
387      * The migration interface for ais was introduced with kernel 4.13
388      * but the capability itself had been active since 4.12. As migration
389      * support is considered necessary, we only try to enable this for
390      * newer machine types if KVM_CAP_S390_AIS_MIGRATION is available.
391      */
392     if (cpu_model_allowed() && kvm_kernel_irqchip_allowed() &&
393         kvm_check_extension(s, KVM_CAP_S390_AIS_MIGRATION)) {
394         kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
395     }
396 
397     kvm_set_max_memslot_size(KVM_SLOT_MAX_BYTES);
398     return 0;
399 }
400 
401 int kvm_arch_irqchip_create(KVMState *s)
402 {
403     return 0;
404 }
405 
406 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
407 {
408     return cpu->cpu_index;
409 }
410 
411 int kvm_arch_init_vcpu(CPUState *cs)
412 {
413     unsigned int max_cpus = MACHINE(qdev_get_machine())->smp.max_cpus;
414     S390CPU *cpu = S390_CPU(cs);
415     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
416     cpu->irqstate = g_malloc0(VCPU_IRQ_BUF_SIZE(max_cpus));
417     return 0;
418 }
419 
420 int kvm_arch_destroy_vcpu(CPUState *cs)
421 {
422     S390CPU *cpu = S390_CPU(cs);
423 
424     g_free(cpu->irqstate);
425     cpu->irqstate = NULL;
426 
427     return 0;
428 }
429 
430 static void kvm_s390_reset_vcpu(S390CPU *cpu, unsigned long type)
431 {
432     CPUState *cs = CPU(cpu);
433 
434     /*
435      * The reset call is needed here to reset in-kernel vcpu data that
436      * we can't access directly from QEMU (i.e. with older kernels
437      * which don't support sync_regs/ONE_REG).  Before this ioctl
438      * cpu_synchronize_state() is called in common kvm code
439      * (kvm-all).
440      */
441     if (kvm_vcpu_ioctl(cs, type)) {
442         error_report("CPU reset failed on CPU %i type %lx",
443                      cs->cpu_index, type);
444     }
445 }
446 
447 void kvm_s390_reset_vcpu_initial(S390CPU *cpu)
448 {
449     kvm_s390_reset_vcpu(cpu, KVM_S390_INITIAL_RESET);
450 }
451 
452 void kvm_s390_reset_vcpu_clear(S390CPU *cpu)
453 {
454     if (cap_vcpu_resets) {
455         kvm_s390_reset_vcpu(cpu, KVM_S390_CLEAR_RESET);
456     } else {
457         kvm_s390_reset_vcpu(cpu, KVM_S390_INITIAL_RESET);
458     }
459 }
460 
461 void kvm_s390_reset_vcpu_normal(S390CPU *cpu)
462 {
463     if (cap_vcpu_resets) {
464         kvm_s390_reset_vcpu(cpu, KVM_S390_NORMAL_RESET);
465     }
466 }
467 
468 static int can_sync_regs(CPUState *cs, int regs)
469 {
470     return (cs->kvm_run->kvm_valid_regs & regs) == regs;
471 }
472 
473 #define KVM_SYNC_REQUIRED_REGS (KVM_SYNC_GPRS | KVM_SYNC_ACRS | \
474                                 KVM_SYNC_CRS | KVM_SYNC_PREFIX)
475 
476 int kvm_arch_put_registers(CPUState *cs, int level, Error **errp)
477 {
478     CPUS390XState *env = cpu_env(cs);
479     struct kvm_fpu fpu = {};
480     int r;
481     int i;
482 
483     g_assert(can_sync_regs(cs, KVM_SYNC_REQUIRED_REGS));
484 
485     /* always save the PSW  and the GPRS*/
486     cs->kvm_run->psw_addr = env->psw.addr;
487     cs->kvm_run->psw_mask = env->psw.mask;
488 
489     memcpy(cs->kvm_run->s.regs.gprs, env->regs, sizeof(cs->kvm_run->s.regs.gprs));
490     cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
491 
492     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
493         for (i = 0; i < 32; i++) {
494             cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0];
495             cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1];
496         }
497         cs->kvm_run->s.regs.fpc = env->fpc;
498         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
499     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
500         for (i = 0; i < 16; i++) {
501             cs->kvm_run->s.regs.fprs[i] = *get_freg(env, i);
502         }
503         cs->kvm_run->s.regs.fpc = env->fpc;
504         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_FPRS;
505     } else {
506         /* Floating point */
507         for (i = 0; i < 16; i++) {
508             fpu.fprs[i] = *get_freg(env, i);
509         }
510         fpu.fpc = env->fpc;
511 
512         r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
513         if (r < 0) {
514             return r;
515         }
516     }
517 
518     /* Do we need to save more than that? */
519     if (level == KVM_PUT_RUNTIME_STATE) {
520         return 0;
521     }
522 
523     /*
524      * Access registers, control registers and the prefix - these are
525      * always available via kvm_sync_regs in the kernels that we support
526      */
527     memcpy(cs->kvm_run->s.regs.acrs, env->aregs, sizeof(cs->kvm_run->s.regs.acrs));
528     memcpy(cs->kvm_run->s.regs.crs, env->cregs, sizeof(cs->kvm_run->s.regs.crs));
529     cs->kvm_run->s.regs.prefix = env->psa;
530     cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS | KVM_SYNC_CRS | KVM_SYNC_PREFIX;
531 
532     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
533         cs->kvm_run->s.regs.cputm = env->cputm;
534         cs->kvm_run->s.regs.ckc = env->ckc;
535         cs->kvm_run->s.regs.todpr = env->todpr;
536         cs->kvm_run->s.regs.gbea = env->gbea;
537         cs->kvm_run->s.regs.pp = env->pp;
538         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
539     } else {
540         /*
541          * These ONE_REGS are not protected by a capability. As they are only
542          * necessary for migration we just trace a possible error, but don't
543          * return with an error return code.
544          */
545         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
546         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
547         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
548         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
549         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
550     }
551 
552     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
553         memcpy(cs->kvm_run->s.regs.riccb, env->riccb, 64);
554         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_RICCB;
555     }
556 
557     /* pfault parameters */
558     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
559         cs->kvm_run->s.regs.pft = env->pfault_token;
560         cs->kvm_run->s.regs.pfs = env->pfault_select;
561         cs->kvm_run->s.regs.pfc = env->pfault_compare;
562         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
563     } else if (cap_async_pf) {
564         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
565         if (r < 0) {
566             return r;
567         }
568         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
569         if (r < 0) {
570             return r;
571         }
572         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
573         if (r < 0) {
574             return r;
575         }
576     }
577 
578     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
579         memcpy(cs->kvm_run->s.regs.gscb, env->gscb, 32);
580         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GSCB;
581     }
582 
583     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
584         cs->kvm_run->s.regs.bpbc = env->bpbc;
585         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_BPBC;
586     }
587 
588     if (can_sync_regs(cs, KVM_SYNC_ETOKEN)) {
589         cs->kvm_run->s.regs.etoken = env->etoken;
590         cs->kvm_run->s.regs.etoken_extension  = env->etoken_extension;
591         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ETOKEN;
592     }
593 
594     if (can_sync_regs(cs, KVM_SYNC_DIAG318)) {
595         cs->kvm_run->s.regs.diag318 = env->diag318_info;
596         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_DIAG318;
597     }
598 
599     return 0;
600 }
601 
602 int kvm_arch_get_registers(CPUState *cs, Error **errp)
603 {
604     CPUS390XState *env = cpu_env(cs);
605     struct kvm_fpu fpu;
606     int i, r;
607 
608     /* get the PSW */
609     env->psw.addr = cs->kvm_run->psw_addr;
610     env->psw.mask = cs->kvm_run->psw_mask;
611 
612     /* the GPRS, ACRS and CRS */
613     g_assert(can_sync_regs(cs, KVM_SYNC_REQUIRED_REGS));
614     memcpy(env->regs, cs->kvm_run->s.regs.gprs, sizeof(env->regs));
615     memcpy(env->aregs, cs->kvm_run->s.regs.acrs, sizeof(env->aregs));
616     memcpy(env->cregs, cs->kvm_run->s.regs.crs, sizeof(env->cregs));
617 
618     /* The prefix */
619     env->psa = cs->kvm_run->s.regs.prefix;
620 
621     /* Floating point and vector registers */
622     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
623         for (i = 0; i < 32; i++) {
624             env->vregs[i][0] = cs->kvm_run->s.regs.vrs[i][0];
625             env->vregs[i][1] = cs->kvm_run->s.regs.vrs[i][1];
626         }
627         env->fpc = cs->kvm_run->s.regs.fpc;
628     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
629         for (i = 0; i < 16; i++) {
630             *get_freg(env, i) = cs->kvm_run->s.regs.fprs[i];
631         }
632         env->fpc = cs->kvm_run->s.regs.fpc;
633     } else {
634         r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
635         if (r < 0) {
636             return r;
637         }
638         for (i = 0; i < 16; i++) {
639             *get_freg(env, i) = fpu.fprs[i];
640         }
641         env->fpc = fpu.fpc;
642     }
643 
644     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
645         env->cputm = cs->kvm_run->s.regs.cputm;
646         env->ckc = cs->kvm_run->s.regs.ckc;
647         env->todpr = cs->kvm_run->s.regs.todpr;
648         env->gbea = cs->kvm_run->s.regs.gbea;
649         env->pp = cs->kvm_run->s.regs.pp;
650     } else {
651         /*
652          * These ONE_REGS are not protected by a capability. As they are only
653          * necessary for migration we just trace a possible error, but don't
654          * return with an error return code.
655          */
656         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
657         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
658         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
659         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
660         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
661     }
662 
663     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
664         memcpy(env->riccb, cs->kvm_run->s.regs.riccb, 64);
665     }
666 
667     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
668         memcpy(env->gscb, cs->kvm_run->s.regs.gscb, 32);
669     }
670 
671     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
672         env->bpbc = cs->kvm_run->s.regs.bpbc;
673     }
674 
675     if (can_sync_regs(cs, KVM_SYNC_ETOKEN)) {
676         env->etoken = cs->kvm_run->s.regs.etoken;
677         env->etoken_extension = cs->kvm_run->s.regs.etoken_extension;
678     }
679 
680     /* pfault parameters */
681     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
682         env->pfault_token = cs->kvm_run->s.regs.pft;
683         env->pfault_select = cs->kvm_run->s.regs.pfs;
684         env->pfault_compare = cs->kvm_run->s.regs.pfc;
685     } else if (cap_async_pf) {
686         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
687         if (r < 0) {
688             return r;
689         }
690         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
691         if (r < 0) {
692             return r;
693         }
694         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
695         if (r < 0) {
696             return r;
697         }
698     }
699 
700     if (can_sync_regs(cs, KVM_SYNC_DIAG318)) {
701         env->diag318_info = cs->kvm_run->s.regs.diag318;
702     }
703 
704     return 0;
705 }
706 
707 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
708 {
709     int r;
710     struct kvm_device_attr attr = {
711         .group = KVM_S390_VM_TOD,
712         .attr = KVM_S390_VM_TOD_LOW,
713         .addr = (uint64_t)tod_low,
714     };
715 
716     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
717     if (r) {
718         return r;
719     }
720 
721     attr.attr = KVM_S390_VM_TOD_HIGH;
722     attr.addr = (uint64_t)tod_high;
723     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
724 }
725 
726 int kvm_s390_get_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
727 {
728     int r;
729     struct kvm_s390_vm_tod_clock gtod;
730     struct kvm_device_attr attr = {
731         .group = KVM_S390_VM_TOD,
732         .attr = KVM_S390_VM_TOD_EXT,
733         .addr = (uint64_t)&gtod,
734     };
735 
736     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
737     *tod_high = gtod.epoch_idx;
738     *tod_low  = gtod.tod;
739 
740     return r;
741 }
742 
743 int kvm_s390_set_clock(uint8_t tod_high, uint64_t tod_low)
744 {
745     int r;
746     struct kvm_device_attr attr = {
747         .group = KVM_S390_VM_TOD,
748         .attr = KVM_S390_VM_TOD_LOW,
749         .addr = (uint64_t)&tod_low,
750     };
751 
752     r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
753     if (r) {
754         return r;
755     }
756 
757     attr.attr = KVM_S390_VM_TOD_HIGH;
758     attr.addr = (uint64_t)&tod_high;
759     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
760 }
761 
762 int kvm_s390_set_clock_ext(uint8_t tod_high, uint64_t tod_low)
763 {
764     struct kvm_s390_vm_tod_clock gtod = {
765         .epoch_idx = tod_high,
766         .tod  = tod_low,
767     };
768     struct kvm_device_attr attr = {
769         .group = KVM_S390_VM_TOD,
770         .attr = KVM_S390_VM_TOD_EXT,
771         .addr = (uint64_t)&gtod,
772     };
773 
774     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
775 }
776 
777 /**
778  * kvm_s390_mem_op:
779  * @addr:      the logical start address in guest memory
780  * @ar:        the access register number
781  * @hostbuf:   buffer in host memory. NULL = do only checks w/o copying
782  * @len:       length that should be transferred
783  * @is_write:  true = write, false = read
784  * Returns:    0 on success, non-zero if an exception or error occurred
785  *
786  * Use KVM ioctl to read/write from/to guest memory. An access exception
787  * is injected into the vCPU in case of translation errors.
788  */
789 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, uint8_t ar, void *hostbuf,
790                     int len, bool is_write)
791 {
792     struct kvm_s390_mem_op mem_op = {
793         .gaddr = addr,
794         .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
795         .size = len,
796         .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
797                        : KVM_S390_MEMOP_LOGICAL_READ,
798         .buf = (uint64_t)hostbuf,
799         .ar = ar,
800         .key = (cpu->env.psw.mask & PSW_MASK_KEY) >> PSW_SHIFT_KEY,
801     };
802     int ret;
803 
804     if (!cap_mem_op) {
805         return -ENOSYS;
806     }
807     if (!hostbuf) {
808         mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
809     }
810     if (mem_op_storage_key_support) {
811         mem_op.flags |= KVM_S390_MEMOP_F_SKEY_PROTECTION;
812     }
813 
814     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
815     if (ret < 0) {
816         warn_report("KVM_S390_MEM_OP failed: %s", strerror(-ret));
817     }
818     return ret;
819 }
820 
821 int kvm_s390_mem_op_pv(S390CPU *cpu, uint64_t offset, void *hostbuf,
822                        int len, bool is_write)
823 {
824     struct kvm_s390_mem_op mem_op = {
825         .sida_offset = offset,
826         .size = len,
827         .op = is_write ? KVM_S390_MEMOP_SIDA_WRITE
828                        : KVM_S390_MEMOP_SIDA_READ,
829         .buf = (uint64_t)hostbuf,
830     };
831     int ret;
832 
833     if (!cap_mem_op || !cap_protected) {
834         return -ENOSYS;
835     }
836 
837     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
838     if (ret < 0) {
839         error_report("KVM_S390_MEM_OP failed: %s", strerror(-ret));
840         abort();
841     }
842     return ret;
843 }
844 
845 static uint8_t const *sw_bp_inst;
846 static uint8_t sw_bp_ilen;
847 
848 static void determine_sw_breakpoint_instr(void)
849 {
850         /* DIAG 501 is used for sw breakpoints with old kernels */
851         static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
852         /* Instruction 0x0000 is used for sw breakpoints with recent kernels */
853         static const uint8_t instr_0x0000[] = {0x00, 0x00};
854 
855         if (sw_bp_inst) {
856             return;
857         }
858         if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_USER_INSTR0, 0)) {
859             sw_bp_inst = diag_501;
860             sw_bp_ilen = sizeof(diag_501);
861             trace_kvm_sw_breakpoint(4);
862         } else {
863             sw_bp_inst = instr_0x0000;
864             sw_bp_ilen = sizeof(instr_0x0000);
865             trace_kvm_sw_breakpoint(2);
866         }
867 }
868 
869 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
870 {
871     determine_sw_breakpoint_instr();
872 
873     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
874                             sw_bp_ilen, 0) ||
875         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)sw_bp_inst, sw_bp_ilen, 1)) {
876         return -EINVAL;
877     }
878     return 0;
879 }
880 
881 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
882 {
883     uint8_t t[MAX_ILEN];
884 
885     if (cpu_memory_rw_debug(cs, bp->pc, t, sw_bp_ilen, 0)) {
886         return -EINVAL;
887     } else if (memcmp(t, sw_bp_inst, sw_bp_ilen)) {
888         return -EINVAL;
889     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
890                                    sw_bp_ilen, 1)) {
891         return -EINVAL;
892     }
893 
894     return 0;
895 }
896 
897 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
898                                                     int len, int type)
899 {
900     int n;
901 
902     for (n = 0; n < nb_hw_breakpoints; n++) {
903         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
904             (hw_breakpoints[n].len == len || len == -1)) {
905             return &hw_breakpoints[n];
906         }
907     }
908 
909     return NULL;
910 }
911 
912 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
913 {
914     int size;
915 
916     if (find_hw_breakpoint(addr, len, type)) {
917         return -EEXIST;
918     }
919 
920     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
921 
922     if (!hw_breakpoints) {
923         nb_hw_breakpoints = 0;
924         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
925     } else {
926         hw_breakpoints =
927             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
928     }
929 
930     if (!hw_breakpoints) {
931         nb_hw_breakpoints = 0;
932         return -ENOMEM;
933     }
934 
935     hw_breakpoints[nb_hw_breakpoints].addr = addr;
936     hw_breakpoints[nb_hw_breakpoints].len = len;
937     hw_breakpoints[nb_hw_breakpoints].type = type;
938 
939     nb_hw_breakpoints++;
940 
941     return 0;
942 }
943 
944 int kvm_arch_insert_hw_breakpoint(vaddr addr, vaddr len, int type)
945 {
946     switch (type) {
947     case GDB_BREAKPOINT_HW:
948         type = KVM_HW_BP;
949         break;
950     case GDB_WATCHPOINT_WRITE:
951         if (len < 1) {
952             return -EINVAL;
953         }
954         type = KVM_HW_WP_WRITE;
955         break;
956     default:
957         return -ENOSYS;
958     }
959     return insert_hw_breakpoint(addr, len, type);
960 }
961 
962 int kvm_arch_remove_hw_breakpoint(vaddr addr, vaddr len, int type)
963 {
964     int size;
965     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
966 
967     if (bp == NULL) {
968         return -ENOENT;
969     }
970 
971     nb_hw_breakpoints--;
972     if (nb_hw_breakpoints > 0) {
973         /*
974          * In order to trim the array, move the last element to the position to
975          * be removed - if necessary.
976          */
977         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
978             *bp = hw_breakpoints[nb_hw_breakpoints];
979         }
980         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
981         hw_breakpoints =
982              g_realloc(hw_breakpoints, size);
983     } else {
984         g_free(hw_breakpoints);
985         hw_breakpoints = NULL;
986     }
987 
988     return 0;
989 }
990 
991 void kvm_arch_remove_all_hw_breakpoints(void)
992 {
993     nb_hw_breakpoints = 0;
994     g_free(hw_breakpoints);
995     hw_breakpoints = NULL;
996 }
997 
998 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
999 {
1000     int i;
1001 
1002     if (nb_hw_breakpoints > 0) {
1003         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
1004         dbg->arch.hw_bp = hw_breakpoints;
1005 
1006         for (i = 0; i < nb_hw_breakpoints; ++i) {
1007             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
1008                                                        hw_breakpoints[i].addr);
1009         }
1010         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
1011     } else {
1012         dbg->arch.nr_hw_bp = 0;
1013         dbg->arch.hw_bp = NULL;
1014     }
1015 }
1016 
1017 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
1018 {
1019 }
1020 
1021 MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
1022 {
1023     return MEMTXATTRS_UNSPECIFIED;
1024 }
1025 
1026 int kvm_arch_process_async_events(CPUState *cs)
1027 {
1028     return cs->halted;
1029 }
1030 
1031 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
1032                                      struct kvm_s390_interrupt *interrupt)
1033 {
1034     int r = 0;
1035 
1036     interrupt->type = irq->type;
1037     switch (irq->type) {
1038     case KVM_S390_INT_VIRTIO:
1039         interrupt->parm = irq->u.ext.ext_params;
1040         /* fall through */
1041     case KVM_S390_INT_PFAULT_INIT:
1042     case KVM_S390_INT_PFAULT_DONE:
1043         interrupt->parm64 = irq->u.ext.ext_params2;
1044         break;
1045     case KVM_S390_PROGRAM_INT:
1046         interrupt->parm = irq->u.pgm.code;
1047         break;
1048     case KVM_S390_SIGP_SET_PREFIX:
1049         interrupt->parm = irq->u.prefix.address;
1050         break;
1051     case KVM_S390_INT_SERVICE:
1052         interrupt->parm = irq->u.ext.ext_params;
1053         break;
1054     case KVM_S390_MCHK:
1055         interrupt->parm = irq->u.mchk.cr14;
1056         interrupt->parm64 = irq->u.mchk.mcic;
1057         break;
1058     case KVM_S390_INT_EXTERNAL_CALL:
1059         interrupt->parm = irq->u.extcall.code;
1060         break;
1061     case KVM_S390_INT_EMERGENCY:
1062         interrupt->parm = irq->u.emerg.code;
1063         break;
1064     case KVM_S390_SIGP_STOP:
1065     case KVM_S390_RESTART:
1066         break; /* These types have no parameters */
1067     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1068         interrupt->parm = irq->u.io.subchannel_id << 16;
1069         interrupt->parm |= irq->u.io.subchannel_nr;
1070         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
1071         interrupt->parm64 |= irq->u.io.io_int_word;
1072         break;
1073     default:
1074         r = -EINVAL;
1075         break;
1076     }
1077     return r;
1078 }
1079 
1080 static void inject_vcpu_irq_legacy(CPUState *cs, struct kvm_s390_irq *irq)
1081 {
1082     struct kvm_s390_interrupt kvmint = {};
1083     int r;
1084 
1085     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1086     if (r < 0) {
1087         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1088         exit(1);
1089     }
1090 
1091     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
1092     if (r < 0) {
1093         fprintf(stderr, "KVM failed to inject interrupt\n");
1094         exit(1);
1095     }
1096 }
1097 
1098 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
1099 {
1100     CPUState *cs = CPU(cpu);
1101     int r;
1102 
1103     if (cap_s390_irq) {
1104         r = kvm_vcpu_ioctl(cs, KVM_S390_IRQ, irq);
1105         if (!r) {
1106             return;
1107         }
1108         error_report("KVM failed to inject interrupt %llx", irq->type);
1109         exit(1);
1110     }
1111 
1112     inject_vcpu_irq_legacy(cs, irq);
1113 }
1114 
1115 void kvm_s390_floating_interrupt_legacy(struct kvm_s390_irq *irq)
1116 {
1117     struct kvm_s390_interrupt kvmint = {};
1118     int r;
1119 
1120     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1121     if (r < 0) {
1122         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1123         exit(1);
1124     }
1125 
1126     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
1127     if (r < 0) {
1128         fprintf(stderr, "KVM failed to inject interrupt\n");
1129         exit(1);
1130     }
1131 }
1132 
1133 void kvm_s390_program_interrupt(S390CPU *cpu, uint16_t code)
1134 {
1135     struct kvm_s390_irq irq = {
1136         .type = KVM_S390_PROGRAM_INT,
1137         .u.pgm.code = code,
1138     };
1139     qemu_log_mask(CPU_LOG_INT, "program interrupt at %#" PRIx64 "\n",
1140                   cpu->env.psw.addr);
1141     kvm_s390_vcpu_interrupt(cpu, &irq);
1142 }
1143 
1144 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
1145 {
1146     struct kvm_s390_irq irq = {
1147         .type = KVM_S390_PROGRAM_INT,
1148         .u.pgm.code = code,
1149         .u.pgm.trans_exc_code = te_code,
1150         .u.pgm.exc_access_id = te_code & 3,
1151     };
1152 
1153     kvm_s390_vcpu_interrupt(cpu, &irq);
1154 }
1155 
1156 static void kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
1157                                  uint16_t ipbh0)
1158 {
1159     CPUS390XState *env = &cpu->env;
1160     uint64_t sccb;
1161     uint32_t code;
1162     int r;
1163 
1164     sccb = env->regs[ipbh0 & 0xf];
1165     code = env->regs[(ipbh0 & 0xf0) >> 4];
1166 
1167     switch (run->s390_sieic.icptcode) {
1168     case ICPT_PV_INSTR_NOTIFICATION:
1169         g_assert(s390_is_pv());
1170         /* The notification intercepts are currently handled by KVM */
1171         error_report("unexpected SCLP PV notification");
1172         exit(1);
1173         break;
1174     case ICPT_PV_INSTR:
1175         g_assert(s390_is_pv());
1176         sclp_service_call_protected(cpu, sccb, code);
1177         /* Setting the CC is done by the Ultravisor. */
1178         break;
1179     case ICPT_INSTRUCTION:
1180         g_assert(!s390_is_pv());
1181         r = sclp_service_call(cpu, sccb, code);
1182         if (r < 0) {
1183             kvm_s390_program_interrupt(cpu, -r);
1184             return;
1185         }
1186         setcc(cpu, r);
1187     }
1188 }
1189 
1190 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1191 {
1192     CPUS390XState *env = &cpu->env;
1193     int rc = 0;
1194     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
1195 
1196     switch (ipa1) {
1197     case PRIV_B2_XSCH:
1198         ioinst_handle_xsch(cpu, env->regs[1], RA_IGNORED);
1199         break;
1200     case PRIV_B2_CSCH:
1201         ioinst_handle_csch(cpu, env->regs[1], RA_IGNORED);
1202         break;
1203     case PRIV_B2_HSCH:
1204         ioinst_handle_hsch(cpu, env->regs[1], RA_IGNORED);
1205         break;
1206     case PRIV_B2_MSCH:
1207         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1208         break;
1209     case PRIV_B2_SSCH:
1210         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1211         break;
1212     case PRIV_B2_STCRW:
1213         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb, RA_IGNORED);
1214         break;
1215     case PRIV_B2_STSCH:
1216         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1217         break;
1218     case PRIV_B2_TSCH:
1219         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1220         fprintf(stderr, "Spurious tsch intercept\n");
1221         break;
1222     case PRIV_B2_CHSC:
1223         ioinst_handle_chsc(cpu, run->s390_sieic.ipb, RA_IGNORED);
1224         break;
1225     case PRIV_B2_TPI:
1226         /* This should have been handled by kvm already. */
1227         fprintf(stderr, "Spurious tpi intercept\n");
1228         break;
1229     case PRIV_B2_SCHM:
1230         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
1231                            run->s390_sieic.ipb, RA_IGNORED);
1232         break;
1233     case PRIV_B2_RSCH:
1234         ioinst_handle_rsch(cpu, env->regs[1], RA_IGNORED);
1235         break;
1236     case PRIV_B2_RCHP:
1237         ioinst_handle_rchp(cpu, env->regs[1], RA_IGNORED);
1238         break;
1239     case PRIV_B2_STCPS:
1240         /* We do not provide this instruction, it is suppressed. */
1241         break;
1242     case PRIV_B2_SAL:
1243         ioinst_handle_sal(cpu, env->regs[1], RA_IGNORED);
1244         break;
1245     case PRIV_B2_SIGA:
1246         /* Not provided, set CC = 3 for subchannel not operational */
1247         setcc(cpu, 3);
1248         break;
1249     case PRIV_B2_SCLP_CALL:
1250         kvm_sclp_service_call(cpu, run, ipbh0);
1251         break;
1252     default:
1253         rc = -1;
1254         trace_kvm_insn_unhandled_priv(ipa1);
1255         break;
1256     }
1257 
1258     return rc;
1259 }
1260 
1261 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run,
1262                                   uint8_t *ar)
1263 {
1264     CPUS390XState *env = &cpu->env;
1265     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1266     uint32_t base2 = run->s390_sieic.ipb >> 28;
1267     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1268                      ((run->s390_sieic.ipb & 0xff00) << 4);
1269 
1270     if (disp2 & 0x80000) {
1271         disp2 += 0xfff00000;
1272     }
1273     if (ar) {
1274         *ar = base2;
1275     }
1276 
1277     return (base2 ? env->regs[base2] : 0) +
1278            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1279 }
1280 
1281 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run,
1282                                   uint8_t *ar)
1283 {
1284     CPUS390XState *env = &cpu->env;
1285     uint32_t base2 = run->s390_sieic.ipb >> 28;
1286     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1287                      ((run->s390_sieic.ipb & 0xff00) << 4);
1288 
1289     if (disp2 & 0x80000) {
1290         disp2 += 0xfff00000;
1291     }
1292     if (ar) {
1293         *ar = base2;
1294     }
1295 
1296     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1297 }
1298 
1299 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1300 {
1301     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1302 
1303     if (s390_has_feat(S390_FEAT_ZPCI)) {
1304         return clp_service_call(cpu, r2, RA_IGNORED);
1305     } else {
1306         return -1;
1307     }
1308 }
1309 
1310 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1311 {
1312     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1313     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1314 
1315     if (s390_has_feat(S390_FEAT_ZPCI)) {
1316         return pcilg_service_call(cpu, r1, r2, RA_IGNORED);
1317     } else {
1318         return -1;
1319     }
1320 }
1321 
1322 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1323 {
1324     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1325     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1326 
1327     if (s390_has_feat(S390_FEAT_ZPCI)) {
1328         return pcistg_service_call(cpu, r1, r2, RA_IGNORED);
1329     } else {
1330         return -1;
1331     }
1332 }
1333 
1334 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1335 {
1336     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1337     uint64_t fiba;
1338     uint8_t ar;
1339 
1340     if (s390_has_feat(S390_FEAT_ZPCI)) {
1341         fiba = get_base_disp_rxy(cpu, run, &ar);
1342 
1343         return stpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1344     } else {
1345         return -1;
1346     }
1347 }
1348 
1349 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1350 {
1351     CPUS390XState *env = &cpu->env;
1352     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1353     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1354     uint8_t isc;
1355     uint16_t mode;
1356     int r;
1357 
1358     mode = env->regs[r1] & 0xffff;
1359     isc = (env->regs[r3] >> 27) & 0x7;
1360     r = css_do_sic(cpu, isc, mode);
1361     if (r) {
1362         kvm_s390_program_interrupt(cpu, -r);
1363     }
1364 
1365     return 0;
1366 }
1367 
1368 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1369 {
1370     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1371     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1372 
1373     if (s390_has_feat(S390_FEAT_ZPCI)) {
1374         return rpcit_service_call(cpu, r1, r2, RA_IGNORED);
1375     } else {
1376         return -1;
1377     }
1378 }
1379 
1380 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1381 {
1382     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1383     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1384     uint64_t gaddr;
1385     uint8_t ar;
1386 
1387     if (s390_has_feat(S390_FEAT_ZPCI)) {
1388         gaddr = get_base_disp_rsy(cpu, run, &ar);
1389 
1390         return pcistb_service_call(cpu, r1, r3, gaddr, ar, RA_IGNORED);
1391     } else {
1392         return -1;
1393     }
1394 }
1395 
1396 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1397 {
1398     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1399     uint64_t fiba;
1400     uint8_t ar;
1401 
1402     if (s390_has_feat(S390_FEAT_ZPCI)) {
1403         fiba = get_base_disp_rxy(cpu, run, &ar);
1404 
1405         return mpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1406     } else {
1407         return -1;
1408     }
1409 }
1410 
1411 static void kvm_handle_ptf(S390CPU *cpu, struct kvm_run *run)
1412 {
1413     uint8_t r1 = (run->s390_sieic.ipb >> 20) & 0x0f;
1414 
1415     s390_handle_ptf(cpu, r1, RA_IGNORED);
1416 }
1417 
1418 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1419 {
1420     int r = 0;
1421 
1422     switch (ipa1) {
1423     case PRIV_B9_CLP:
1424         r = kvm_clp_service_call(cpu, run);
1425         break;
1426     case PRIV_B9_PCISTG:
1427         r = kvm_pcistg_service_call(cpu, run);
1428         break;
1429     case PRIV_B9_PCILG:
1430         r = kvm_pcilg_service_call(cpu, run);
1431         break;
1432     case PRIV_B9_RPCIT:
1433         r = kvm_rpcit_service_call(cpu, run);
1434         break;
1435     case PRIV_B9_PTF:
1436         kvm_handle_ptf(cpu, run);
1437         break;
1438     case PRIV_B9_EQBS:
1439         /* just inject exception */
1440         r = -1;
1441         break;
1442     default:
1443         r = -1;
1444         trace_kvm_insn_unhandled_priv(ipa1);
1445         break;
1446     }
1447 
1448     return r;
1449 }
1450 
1451 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1452 {
1453     int r = 0;
1454 
1455     switch (ipbl) {
1456     case PRIV_EB_PCISTB:
1457         r = kvm_pcistb_service_call(cpu, run);
1458         break;
1459     case PRIV_EB_SIC:
1460         r = kvm_sic_service_call(cpu, run);
1461         break;
1462     case PRIV_EB_SQBS:
1463         /* just inject exception */
1464         r = -1;
1465         break;
1466     default:
1467         r = -1;
1468         trace_kvm_insn_unhandled_priv(ipbl);
1469         break;
1470     }
1471 
1472     return r;
1473 }
1474 
1475 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1476 {
1477     int r = 0;
1478 
1479     switch (ipbl) {
1480     case PRIV_E3_MPCIFC:
1481         r = kvm_mpcifc_service_call(cpu, run);
1482         break;
1483     case PRIV_E3_STPCIFC:
1484         r = kvm_stpcifc_service_call(cpu, run);
1485         break;
1486     default:
1487         r = -1;
1488         trace_kvm_insn_unhandled_priv(ipbl);
1489         break;
1490     }
1491 
1492     return r;
1493 }
1494 
1495 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1496 {
1497     CPUS390XState *env = &cpu->env;
1498     int ret = -EINVAL;
1499 
1500 #ifdef CONFIG_S390_CCW_VIRTIO
1501     ret = s390_virtio_hypercall(env);
1502 #endif /* CONFIG_S390_CCW_VIRTIO */
1503     if (ret == -EINVAL) {
1504         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1505         return 0;
1506     }
1507 
1508     return ret;
1509 }
1510 
1511 static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
1512 {
1513     uint64_t r1, r3;
1514     int rc;
1515 
1516     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1517     r3 = run->s390_sieic.ipa & 0x000f;
1518     rc = handle_diag_288(&cpu->env, r1, r3);
1519     if (rc) {
1520         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1521     }
1522 }
1523 
1524 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1525 {
1526     uint64_t r1, r3;
1527 
1528     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1529     r3 = run->s390_sieic.ipa & 0x000f;
1530     handle_diag_308(&cpu->env, r1, r3, RA_IGNORED);
1531 }
1532 
1533 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1534 {
1535     CPUS390XState *env = &cpu->env;
1536     unsigned long pc;
1537 
1538     pc = env->psw.addr - sw_bp_ilen;
1539     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1540         env->psw.addr = pc;
1541         return EXCP_DEBUG;
1542     }
1543 
1544     return -ENOENT;
1545 }
1546 
1547 void kvm_s390_set_diag318(CPUState *cs, uint64_t diag318_info)
1548 {
1549     CPUS390XState *env = &S390_CPU(cs)->env;
1550 
1551     /* Feat bit is set only if KVM supports sync for diag318 */
1552     if (s390_has_feat(S390_FEAT_DIAG_318)) {
1553         env->diag318_info = diag318_info;
1554         cs->kvm_run->s.regs.diag318 = diag318_info;
1555         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_DIAG318;
1556         /*
1557          * diag 318 info is zeroed during a clear reset and
1558          * diag 308 IPL subcodes.
1559          */
1560     }
1561 }
1562 
1563 static void handle_diag_318(S390CPU *cpu, struct kvm_run *run)
1564 {
1565     uint64_t reg = (run->s390_sieic.ipa & 0x00f0) >> 4;
1566     uint64_t diag318_info = run->s.regs.gprs[reg];
1567     CPUState *t;
1568 
1569     /*
1570      * DIAG 318 can only be enabled with KVM support. As such, let's
1571      * ensure a guest cannot execute this instruction erroneously.
1572      */
1573     if (!s390_has_feat(S390_FEAT_DIAG_318)) {
1574         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1575         return;
1576     }
1577 
1578     CPU_FOREACH(t) {
1579         run_on_cpu(t, s390_do_cpu_set_diag318,
1580                    RUN_ON_CPU_HOST_ULONG(diag318_info));
1581     }
1582 }
1583 
1584 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1585 
1586 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1587 {
1588     int r = 0;
1589     uint16_t func_code;
1590 
1591     /*
1592      * For any diagnose call we support, bits 48-63 of the resulting
1593      * address specify the function code; the remainder is ignored.
1594      */
1595     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
1596     switch (func_code) {
1597     case DIAG_TIMEREVENT:
1598         kvm_handle_diag_288(cpu, run);
1599         break;
1600     case DIAG_IPL:
1601         kvm_handle_diag_308(cpu, run);
1602         break;
1603     case DIAG_SET_CONTROL_PROGRAM_CODES:
1604         handle_diag_318(cpu, run);
1605         break;
1606     case DIAG_KVM_HYPERCALL:
1607         r = handle_hypercall(cpu, run);
1608         break;
1609     case DIAG_KVM_BREAKPOINT:
1610         r = handle_sw_breakpoint(cpu, run);
1611         break;
1612     default:
1613         trace_kvm_insn_diag(func_code);
1614         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1615         break;
1616     }
1617 
1618     return r;
1619 }
1620 
1621 static int kvm_s390_handle_sigp(S390CPU *cpu, uint8_t ipa1, uint32_t ipb)
1622 {
1623     CPUS390XState *env = &cpu->env;
1624     const uint8_t r1 = ipa1 >> 4;
1625     const uint8_t r3 = ipa1 & 0x0f;
1626     int ret;
1627     uint8_t order;
1628 
1629     /* get order code */
1630     order = decode_basedisp_rs(env, ipb, NULL) & SIGP_ORDER_MASK;
1631 
1632     ret = handle_sigp(env, order, r1, r3);
1633     setcc(cpu, ret);
1634     return 0;
1635 }
1636 
1637 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1638 {
1639     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1640     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1641     int r = -1;
1642 
1643     trace_kvm_insn(run->s390_sieic.ipa, run->s390_sieic.ipb);
1644     switch (ipa0) {
1645     case IPA0_B2:
1646         r = handle_b2(cpu, run, ipa1);
1647         break;
1648     case IPA0_B9:
1649         r = handle_b9(cpu, run, ipa1);
1650         break;
1651     case IPA0_EB:
1652         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1653         break;
1654     case IPA0_E3:
1655         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1656         break;
1657     case IPA0_DIAG:
1658         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1659         break;
1660     case IPA0_SIGP:
1661         r = kvm_s390_handle_sigp(cpu, ipa1, run->s390_sieic.ipb);
1662         break;
1663     }
1664 
1665     if (r < 0) {
1666         r = 0;
1667         kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1668     }
1669 
1670     return r;
1671 }
1672 
1673 static void unmanageable_intercept(S390CPU *cpu, S390CrashReason reason,
1674                                    int pswoffset)
1675 {
1676     CPUState *cs = CPU(cpu);
1677 
1678     s390_cpu_halt(cpu);
1679     cpu->env.crash_reason = reason;
1680     qemu_system_guest_panicked(cpu_get_crash_info(cs));
1681 }
1682 
1683 /* try to detect pgm check loops */
1684 static int handle_oper_loop(S390CPU *cpu, struct kvm_run *run)
1685 {
1686     CPUState *cs = CPU(cpu);
1687     PSW oldpsw, newpsw;
1688 
1689     newpsw.mask = ldq_phys(cs->as, cpu->env.psa +
1690                            offsetof(LowCore, program_new_psw));
1691     newpsw.addr = ldq_phys(cs->as, cpu->env.psa +
1692                            offsetof(LowCore, program_new_psw) + 8);
1693     oldpsw.mask  = run->psw_mask;
1694     oldpsw.addr  = run->psw_addr;
1695     /*
1696      * Avoid endless loops of operation exceptions, if the pgm new
1697      * PSW will cause a new operation exception.
1698      * The heuristic checks if the pgm new psw is within 6 bytes before
1699      * the faulting psw address (with same DAT, AS settings) and the
1700      * new psw is not a wait psw and the fault was not triggered by
1701      * problem state. In that case go into crashed state.
1702      */
1703 
1704     if (oldpsw.addr - newpsw.addr <= 6 &&
1705         !(newpsw.mask & PSW_MASK_WAIT) &&
1706         !(oldpsw.mask & PSW_MASK_PSTATE) &&
1707         (newpsw.mask & PSW_MASK_ASC) == (oldpsw.mask & PSW_MASK_ASC) &&
1708         (newpsw.mask & PSW_MASK_DAT) == (oldpsw.mask & PSW_MASK_DAT)) {
1709         unmanageable_intercept(cpu, S390_CRASH_REASON_OPINT_LOOP,
1710                                offsetof(LowCore, program_new_psw));
1711         return EXCP_HALTED;
1712     }
1713     return 0;
1714 }
1715 
1716 static int handle_intercept(S390CPU *cpu)
1717 {
1718     CPUState *cs = CPU(cpu);
1719     struct kvm_run *run = cs->kvm_run;
1720     int icpt_code = run->s390_sieic.icptcode;
1721     int r = 0;
1722 
1723     trace_kvm_intercept(icpt_code, (long)run->psw_addr);
1724     switch (icpt_code) {
1725         case ICPT_INSTRUCTION:
1726         case ICPT_PV_INSTR:
1727         case ICPT_PV_INSTR_NOTIFICATION:
1728             r = handle_instruction(cpu, run);
1729             break;
1730         case ICPT_PROGRAM:
1731             unmanageable_intercept(cpu, S390_CRASH_REASON_PGMINT_LOOP,
1732                                    offsetof(LowCore, program_new_psw));
1733             r = EXCP_HALTED;
1734             break;
1735         case ICPT_EXT_INT:
1736             unmanageable_intercept(cpu, S390_CRASH_REASON_EXTINT_LOOP,
1737                                    offsetof(LowCore, external_new_psw));
1738             r = EXCP_HALTED;
1739             break;
1740         case ICPT_WAITPSW:
1741             /* disabled wait, since enabled wait is handled in kernel */
1742             s390_handle_wait(cpu);
1743             r = EXCP_HALTED;
1744             break;
1745         case ICPT_CPU_STOP:
1746             do_stop_interrupt(&cpu->env);
1747             r = EXCP_HALTED;
1748             break;
1749         case ICPT_OPEREXC:
1750             /* check for break points */
1751             r = handle_sw_breakpoint(cpu, run);
1752             if (r == -ENOENT) {
1753                 /* Then check for potential pgm check loops */
1754                 r = handle_oper_loop(cpu, run);
1755                 if (r == 0) {
1756                     kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1757                 }
1758             }
1759             break;
1760         case ICPT_SOFT_INTERCEPT:
1761             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1762             exit(1);
1763             break;
1764         case ICPT_IO:
1765             fprintf(stderr, "KVM unimplemented icpt IO\n");
1766             exit(1);
1767             break;
1768         default:
1769             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1770             exit(1);
1771             break;
1772     }
1773 
1774     return r;
1775 }
1776 
1777 static int handle_tsch(S390CPU *cpu)
1778 {
1779     CPUState *cs = CPU(cpu);
1780     struct kvm_run *run = cs->kvm_run;
1781     int ret;
1782 
1783     ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb,
1784                              RA_IGNORED);
1785     if (ret < 0) {
1786         /*
1787          * Failure.
1788          * If an I/O interrupt had been dequeued, we have to reinject it.
1789          */
1790         if (run->s390_tsch.dequeued) {
1791             s390_io_interrupt(run->s390_tsch.subchannel_id,
1792                               run->s390_tsch.subchannel_nr,
1793                               run->s390_tsch.io_int_parm,
1794                               run->s390_tsch.io_int_word);
1795         }
1796         ret = 0;
1797     }
1798     return ret;
1799 }
1800 
1801 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr, uint8_t ar)
1802 {
1803     const MachineState *ms = MACHINE(qdev_get_machine());
1804     uint16_t conf_cpus = 0, reserved_cpus = 0;
1805     SysIB_322 sysib;
1806     int del, i;
1807 
1808     if (s390_is_pv()) {
1809         s390_cpu_pv_mem_read(cpu, 0, &sysib, sizeof(sysib));
1810     } else if (s390_cpu_virt_mem_read(cpu, addr, ar, &sysib, sizeof(sysib))) {
1811         return;
1812     }
1813     /* Shift the stack of Extended Names to prepare for our own data */
1814     memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1815             sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1816     /* First virt level, that doesn't provide Ext Names delimits stack. It is
1817      * assumed it's not capable of managing Extended Names for lower levels.
1818      */
1819     for (del = 1; del < sysib.count; del++) {
1820         if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1821             break;
1822         }
1823     }
1824     if (del < sysib.count) {
1825         memset(sysib.ext_names[del], 0,
1826                sizeof(sysib.ext_names[0]) * (sysib.count - del));
1827     }
1828 
1829     /* count the cpus and split them into configured and reserved ones */
1830     for (i = 0; i < ms->possible_cpus->len; i++) {
1831         if (ms->possible_cpus->cpus[i].cpu) {
1832             conf_cpus++;
1833         } else {
1834             reserved_cpus++;
1835         }
1836     }
1837     sysib.vm[0].total_cpus = conf_cpus + reserved_cpus;
1838     sysib.vm[0].conf_cpus = conf_cpus;
1839     sysib.vm[0].reserved_cpus = reserved_cpus;
1840 
1841     /* Insert short machine name in EBCDIC, padded with blanks */
1842     if (qemu_name) {
1843         memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1844         ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1845                                                     strlen(qemu_name)));
1846     }
1847     sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1848     /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1849      * considered by s390 as not capable of providing any Extended Name.
1850      * Therefore if no name was specified on qemu invocation, we go with the
1851      * same "KVMguest" default, which KVM has filled into short name field.
1852      */
1853     strpadcpy((char *)sysib.ext_names[0],
1854               sizeof(sysib.ext_names[0]),
1855               qemu_name ?: "KVMguest", '\0');
1856 
1857     /* Insert UUID */
1858     memcpy(sysib.vm[0].uuid, &qemu_uuid, sizeof(sysib.vm[0].uuid));
1859 
1860     if (s390_is_pv()) {
1861         s390_cpu_pv_mem_write(cpu, 0, &sysib, sizeof(sysib));
1862     } else {
1863         s390_cpu_virt_mem_write(cpu, addr, ar, &sysib, sizeof(sysib));
1864     }
1865 }
1866 
1867 static int handle_stsi(S390CPU *cpu)
1868 {
1869     CPUState *cs = CPU(cpu);
1870     struct kvm_run *run = cs->kvm_run;
1871 
1872     switch (run->s390_stsi.fc) {
1873     case 3:
1874         if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1875             return 0;
1876         }
1877         insert_stsi_3_2_2(cpu, run->s390_stsi.addr, run->s390_stsi.ar);
1878         return 0;
1879     case 15:
1880         insert_stsi_15_1_x(cpu, run->s390_stsi.sel2, run->s390_stsi.addr,
1881                            run->s390_stsi.ar, RA_IGNORED);
1882         return 0;
1883     default:
1884         return 0;
1885     }
1886 }
1887 
1888 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1889 {
1890     CPUState *cs = CPU(cpu);
1891     struct kvm_run *run = cs->kvm_run;
1892 
1893     int ret = 0;
1894     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1895 
1896     switch (arch_info->type) {
1897     case KVM_HW_WP_WRITE:
1898         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1899             cs->watchpoint_hit = &hw_watchpoint;
1900             hw_watchpoint.vaddr = arch_info->addr;
1901             hw_watchpoint.flags = BP_MEM_WRITE;
1902             ret = EXCP_DEBUG;
1903         }
1904         break;
1905     case KVM_HW_BP:
1906         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1907             ret = EXCP_DEBUG;
1908         }
1909         break;
1910     case KVM_SINGLESTEP:
1911         if (cs->singlestep_enabled) {
1912             ret = EXCP_DEBUG;
1913         }
1914         break;
1915     default:
1916         ret = -ENOSYS;
1917     }
1918 
1919     return ret;
1920 }
1921 
1922 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1923 {
1924     S390CPU *cpu = S390_CPU(cs);
1925     int ret = 0;
1926 
1927     bql_lock();
1928 
1929     kvm_cpu_synchronize_state(cs);
1930 
1931     switch (run->exit_reason) {
1932         case KVM_EXIT_S390_SIEIC:
1933             ret = handle_intercept(cpu);
1934             break;
1935         case KVM_EXIT_S390_RESET:
1936             s390_ipl_reset_request(cs, S390_RESET_REIPL);
1937             break;
1938         case KVM_EXIT_S390_TSCH:
1939             ret = handle_tsch(cpu);
1940             break;
1941         case KVM_EXIT_S390_STSI:
1942             ret = handle_stsi(cpu);
1943             break;
1944         case KVM_EXIT_DEBUG:
1945             ret = kvm_arch_handle_debug_exit(cpu);
1946             break;
1947         default:
1948             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1949             break;
1950     }
1951     bql_unlock();
1952 
1953     if (ret == 0) {
1954         ret = EXCP_INTERRUPT;
1955     }
1956     return ret;
1957 }
1958 
1959 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1960 {
1961     return true;
1962 }
1963 
1964 void kvm_s390_enable_css_support(S390CPU *cpu)
1965 {
1966     int r;
1967 
1968     /* Activate host kernel channel subsystem support. */
1969     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1970     assert(r == 0);
1971 }
1972 
1973 void kvm_arch_init_irq_routing(KVMState *s)
1974 {
1975     /*
1976      * Note that while irqchip capabilities generally imply that cpustates
1977      * are handled in-kernel, it is not true for s390 (yet); therefore, we
1978      * have to override the common code kvm_halt_in_kernel_allowed setting.
1979      */
1980     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1981         kvm_gsi_routing_allowed = true;
1982         kvm_halt_in_kernel_allowed = false;
1983     }
1984 }
1985 
1986 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1987                                     int vq, bool assign)
1988 {
1989     struct kvm_ioeventfd kick = {
1990         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1991         KVM_IOEVENTFD_FLAG_DATAMATCH,
1992         .fd = event_notifier_get_fd(notifier),
1993         .datamatch = vq,
1994         .addr = sch,
1995         .len = 8,
1996     };
1997     trace_kvm_assign_subch_ioeventfd(kick.fd, kick.addr, assign,
1998                                      kick.datamatch);
1999     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
2000         return -ENOSYS;
2001     }
2002     if (!assign) {
2003         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
2004     }
2005     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
2006 }
2007 
2008 int kvm_s390_get_protected_dump(void)
2009 {
2010     return cap_protected_dump;
2011 }
2012 
2013 int kvm_s390_get_ri(void)
2014 {
2015     return cap_ri;
2016 }
2017 
2018 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
2019 {
2020     struct kvm_mp_state mp_state = {};
2021     int ret;
2022 
2023     /* the kvm part might not have been initialized yet */
2024     if (CPU(cpu)->kvm_state == NULL) {
2025         return 0;
2026     }
2027 
2028     switch (cpu_state) {
2029     case S390_CPU_STATE_STOPPED:
2030         mp_state.mp_state = KVM_MP_STATE_STOPPED;
2031         break;
2032     case S390_CPU_STATE_CHECK_STOP:
2033         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
2034         break;
2035     case S390_CPU_STATE_OPERATING:
2036         mp_state.mp_state = KVM_MP_STATE_OPERATING;
2037         break;
2038     case S390_CPU_STATE_LOAD:
2039         mp_state.mp_state = KVM_MP_STATE_LOAD;
2040         break;
2041     default:
2042         error_report("Requested CPU state is not a valid S390 CPU state: %u",
2043                      cpu_state);
2044         exit(1);
2045     }
2046 
2047     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
2048     if (ret) {
2049         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
2050                                        strerror(-ret));
2051     }
2052 
2053     return ret;
2054 }
2055 
2056 void kvm_s390_vcpu_interrupt_pre_save(S390CPU *cpu)
2057 {
2058     unsigned int max_cpus = MACHINE(qdev_get_machine())->smp.max_cpus;
2059     struct kvm_s390_irq_state irq_state = {
2060         .buf = (uint64_t) cpu->irqstate,
2061         .len = VCPU_IRQ_BUF_SIZE(max_cpus),
2062     };
2063     CPUState *cs = CPU(cpu);
2064     int32_t bytes;
2065 
2066     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2067         return;
2068     }
2069 
2070     bytes = kvm_vcpu_ioctl(cs, KVM_S390_GET_IRQ_STATE, &irq_state);
2071     if (bytes < 0) {
2072         cpu->irqstate_saved_size = 0;
2073         error_report("Migration of interrupt state failed");
2074         return;
2075     }
2076 
2077     cpu->irqstate_saved_size = bytes;
2078 }
2079 
2080 int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
2081 {
2082     CPUState *cs = CPU(cpu);
2083     struct kvm_s390_irq_state irq_state = {
2084         .buf = (uint64_t) cpu->irqstate,
2085         .len = cpu->irqstate_saved_size,
2086     };
2087     int r;
2088 
2089     if (cpu->irqstate_saved_size == 0) {
2090         return 0;
2091     }
2092 
2093     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2094         return -ENOSYS;
2095     }
2096 
2097     r = kvm_vcpu_ioctl(cs, KVM_S390_SET_IRQ_STATE, &irq_state);
2098     if (r) {
2099         error_report("Setting interrupt state failed %d", r);
2100     }
2101     return r;
2102 }
2103 
2104 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
2105                              uint64_t address, uint32_t data, PCIDevice *dev)
2106 {
2107     S390PCIBusDevice *pbdev;
2108     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
2109 
2110     if (!dev) {
2111         trace_kvm_msi_route_fixup("no pci device");
2112         return -ENODEV;
2113     }
2114 
2115     pbdev = s390_pci_find_dev_by_target(s390_get_phb(), DEVICE(dev)->id);
2116     if (!pbdev) {
2117         trace_kvm_msi_route_fixup("no zpci device");
2118         return -ENODEV;
2119     }
2120 
2121     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
2122     route->flags = 0;
2123     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
2124     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
2125     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
2126     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset + vec;
2127     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
2128     return 0;
2129 }
2130 
2131 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
2132                                 int vector, PCIDevice *dev)
2133 {
2134     return 0;
2135 }
2136 
2137 int kvm_arch_release_virq_post(int virq)
2138 {
2139     return 0;
2140 }
2141 
2142 int kvm_arch_msi_data_to_gsi(uint32_t data)
2143 {
2144     abort();
2145 }
2146 
2147 static int query_cpu_subfunc(S390FeatBitmap features)
2148 {
2149     struct kvm_s390_vm_cpu_subfunc prop = {};
2150     struct kvm_device_attr attr = {
2151         .group = KVM_S390_VM_CPU_MODEL,
2152         .attr = KVM_S390_VM_CPU_MACHINE_SUBFUNC,
2153         .addr = (uint64_t) &prop,
2154     };
2155     int rc;
2156 
2157     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2158     if (rc) {
2159         return  rc;
2160     }
2161 
2162     /*
2163      * We're going to add all subfunctions now, if the corresponding feature
2164      * is available that unlocks the query functions.
2165      */
2166     s390_add_from_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2167     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2168         s390_add_from_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2169     }
2170     if (test_bit(S390_FEAT_MSA, features)) {
2171         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2172         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2173         s390_add_from_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2174         s390_add_from_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2175         s390_add_from_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2176     }
2177     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2178         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2179     }
2180     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2181         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2182         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2183         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2184         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2185     }
2186     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2187         s390_add_from_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2188     }
2189     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2190         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2191     }
2192     if (test_bit(S390_FEAT_MSA_EXT_9, features)) {
2193         s390_add_from_feat_block(features, S390_FEAT_TYPE_KDSA, prop.kdsa);
2194     }
2195     if (test_bit(S390_FEAT_ESORT_BASE, features)) {
2196         s390_add_from_feat_block(features, S390_FEAT_TYPE_SORTL, prop.sortl);
2197     }
2198     if (test_bit(S390_FEAT_DEFLATE_BASE, features)) {
2199         s390_add_from_feat_block(features, S390_FEAT_TYPE_DFLTCC, prop.dfltcc);
2200     }
2201     if (test_bit(S390_FEAT_CCF_BASE, features)) {
2202         s390_add_from_feat_block(features, S390_FEAT_TYPE_PFCR, prop.pfcr);
2203     }
2204     return 0;
2205 }
2206 
2207 static int configure_cpu_subfunc(const S390FeatBitmap features)
2208 {
2209     struct kvm_s390_vm_cpu_subfunc prop = {};
2210     struct kvm_device_attr attr = {
2211         .group = KVM_S390_VM_CPU_MODEL,
2212         .attr = KVM_S390_VM_CPU_PROCESSOR_SUBFUNC,
2213         .addr = (uint64_t) &prop,
2214     };
2215 
2216     if (!kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2217                            KVM_S390_VM_CPU_PROCESSOR_SUBFUNC)) {
2218         /* hardware support might be missing, IBC will handle most of this */
2219         return 0;
2220     }
2221 
2222     s390_fill_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2223     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2224         s390_fill_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2225     }
2226     if (test_bit(S390_FEAT_MSA, features)) {
2227         s390_fill_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2228         s390_fill_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2229         s390_fill_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2230         s390_fill_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2231         s390_fill_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2232     }
2233     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2234         s390_fill_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2235     }
2236     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2237         s390_fill_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2238         s390_fill_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2239         s390_fill_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2240         s390_fill_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2241     }
2242     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2243         s390_fill_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2244     }
2245     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2246         s390_fill_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2247     }
2248     if (test_bit(S390_FEAT_MSA_EXT_9, features)) {
2249         s390_fill_feat_block(features, S390_FEAT_TYPE_KDSA, prop.kdsa);
2250     }
2251     if (test_bit(S390_FEAT_ESORT_BASE, features)) {
2252         s390_fill_feat_block(features, S390_FEAT_TYPE_SORTL, prop.sortl);
2253     }
2254     if (test_bit(S390_FEAT_DEFLATE_BASE, features)) {
2255         s390_fill_feat_block(features, S390_FEAT_TYPE_DFLTCC, prop.dfltcc);
2256     }
2257     if (test_bit(S390_FEAT_CCF_BASE, features)) {
2258         s390_fill_feat_block(features, S390_FEAT_TYPE_PFCR, prop.pfcr);
2259     }
2260     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2261 }
2262 
2263 static bool ap_available(void)
2264 {
2265     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO,
2266                              KVM_S390_VM_CRYPTO_ENABLE_APIE);
2267 }
2268 
2269 static bool ap_enabled(const S390FeatBitmap features)
2270 {
2271     return test_bit(S390_FEAT_AP, features);
2272 }
2273 
2274 static bool uv_feat_supported(void)
2275 {
2276     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2277                              KVM_S390_VM_CPU_PROCESSOR_UV_FEAT_GUEST);
2278 }
2279 
2280 static int query_uv_feat_guest(S390FeatBitmap features)
2281 {
2282     struct kvm_s390_vm_cpu_uv_feat prop = {};
2283     struct kvm_device_attr attr = {
2284         .group = KVM_S390_VM_CPU_MODEL,
2285         .attr = KVM_S390_VM_CPU_MACHINE_UV_FEAT_GUEST,
2286         .addr = (uint64_t) &prop,
2287     };
2288     int rc;
2289 
2290     /* AP support check is currently the only user of the UV feature test */
2291     if (!(uv_feat_supported() && ap_available())) {
2292         return 0;
2293     }
2294 
2295     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2296     if (rc) {
2297         return  rc;
2298     }
2299 
2300     if (prop.ap) {
2301         set_bit(S390_FEAT_UV_FEAT_AP, features);
2302     }
2303     if (prop.ap_intr) {
2304         set_bit(S390_FEAT_UV_FEAT_AP_INTR, features);
2305     }
2306 
2307     return 0;
2308 }
2309 
2310 static int kvm_to_feat[][2] = {
2311     { KVM_S390_VM_CPU_FEAT_ESOP, S390_FEAT_ESOP },
2312     { KVM_S390_VM_CPU_FEAT_SIEF2, S390_FEAT_SIE_F2 },
2313     { KVM_S390_VM_CPU_FEAT_64BSCAO , S390_FEAT_SIE_64BSCAO },
2314     { KVM_S390_VM_CPU_FEAT_SIIF, S390_FEAT_SIE_SIIF },
2315     { KVM_S390_VM_CPU_FEAT_GPERE, S390_FEAT_SIE_GPERE },
2316     { KVM_S390_VM_CPU_FEAT_GSLS, S390_FEAT_SIE_GSLS },
2317     { KVM_S390_VM_CPU_FEAT_IB, S390_FEAT_SIE_IB },
2318     { KVM_S390_VM_CPU_FEAT_CEI, S390_FEAT_SIE_CEI },
2319     { KVM_S390_VM_CPU_FEAT_IBS, S390_FEAT_SIE_IBS },
2320     { KVM_S390_VM_CPU_FEAT_SKEY, S390_FEAT_SIE_SKEY },
2321     { KVM_S390_VM_CPU_FEAT_CMMA, S390_FEAT_SIE_CMMA },
2322     { KVM_S390_VM_CPU_FEAT_PFMFI, S390_FEAT_SIE_PFMFI},
2323     { KVM_S390_VM_CPU_FEAT_SIGPIF, S390_FEAT_SIE_SIGPIF},
2324     { KVM_S390_VM_CPU_FEAT_KSS, S390_FEAT_SIE_KSS},
2325 };
2326 
2327 static int query_cpu_feat(S390FeatBitmap features)
2328 {
2329     struct kvm_s390_vm_cpu_feat prop = {};
2330     struct kvm_device_attr attr = {
2331         .group = KVM_S390_VM_CPU_MODEL,
2332         .attr = KVM_S390_VM_CPU_MACHINE_FEAT,
2333         .addr = (uint64_t) &prop,
2334     };
2335     int rc;
2336     int i;
2337 
2338     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2339     if (rc) {
2340         return  rc;
2341     }
2342 
2343     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2344         if (test_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat)) {
2345             set_bit(kvm_to_feat[i][1], features);
2346         }
2347     }
2348     return 0;
2349 }
2350 
2351 static int configure_cpu_feat(const S390FeatBitmap features)
2352 {
2353     struct kvm_s390_vm_cpu_feat prop = {};
2354     struct kvm_device_attr attr = {
2355         .group = KVM_S390_VM_CPU_MODEL,
2356         .attr = KVM_S390_VM_CPU_PROCESSOR_FEAT,
2357         .addr = (uint64_t) &prop,
2358     };
2359     int i;
2360 
2361     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2362         if (test_bit(kvm_to_feat[i][1], features)) {
2363             set_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat);
2364         }
2365     }
2366     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2367 }
2368 
2369 bool kvm_s390_cpu_models_supported(void)
2370 {
2371     if (!cpu_model_allowed()) {
2372         /* compatibility machines interfere with the cpu model */
2373         return false;
2374     }
2375     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2376                              KVM_S390_VM_CPU_MACHINE) &&
2377            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2378                              KVM_S390_VM_CPU_PROCESSOR) &&
2379            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2380                              KVM_S390_VM_CPU_MACHINE_FEAT) &&
2381            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2382                              KVM_S390_VM_CPU_PROCESSOR_FEAT) &&
2383            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2384                              KVM_S390_VM_CPU_MACHINE_SUBFUNC);
2385 }
2386 
2387 bool kvm_s390_get_host_cpu_model(S390CPUModel *model, Error **errp)
2388 {
2389     struct kvm_s390_vm_cpu_machine prop = {};
2390     struct kvm_device_attr attr = {
2391         .group = KVM_S390_VM_CPU_MODEL,
2392         .attr = KVM_S390_VM_CPU_MACHINE,
2393         .addr = (uint64_t) &prop,
2394     };
2395     uint16_t unblocked_ibc = 0, cpu_type = 0;
2396     int rc;
2397 
2398     memset(model, 0, sizeof(*model));
2399 
2400     if (!kvm_s390_cpu_models_supported()) {
2401         error_setg(errp, "KVM doesn't support CPU models");
2402         return false;
2403     }
2404 
2405     /* query the basic cpu model properties */
2406     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2407     if (rc) {
2408         error_setg(errp, "KVM: Error querying host CPU model: %d", rc);
2409         return false;
2410     }
2411 
2412     cpu_type = cpuid_type(prop.cpuid);
2413     if (has_ibc(prop.ibc)) {
2414         model->lowest_ibc = lowest_ibc(prop.ibc);
2415         unblocked_ibc = unblocked_ibc(prop.ibc);
2416     }
2417     model->cpu_id = cpuid_id(prop.cpuid);
2418     model->cpu_id_format = cpuid_format(prop.cpuid);
2419     model->cpu_ver = 0xff;
2420 
2421     /* get supported cpu features indicated via STFL(E) */
2422     s390_add_from_feat_block(model->features, S390_FEAT_TYPE_STFL,
2423                              (uint8_t *) prop.fac_mask);
2424     /* dat-enhancement facility 2 has no bit but was introduced with stfle */
2425     if (test_bit(S390_FEAT_STFLE, model->features)) {
2426         set_bit(S390_FEAT_DAT_ENH_2, model->features);
2427     }
2428     /* get supported cpu features indicated e.g. via SCLP */
2429     rc = query_cpu_feat(model->features);
2430     if (rc) {
2431         error_setg(errp, "KVM: Error querying CPU features: %d", rc);
2432         return false;
2433     }
2434     /* get supported cpu subfunctions indicated via query / test bit */
2435     rc = query_cpu_subfunc(model->features);
2436     if (rc) {
2437         error_setg(errp, "KVM: Error querying CPU subfunctions: %d", rc);
2438         return false;
2439     }
2440 
2441     /* PTFF subfunctions might be indicated although kernel support missing */
2442     if (!test_bit(S390_FEAT_MULTIPLE_EPOCH, model->features)) {
2443         clear_bit(S390_FEAT_PTFF_QSIE, model->features);
2444         clear_bit(S390_FEAT_PTFF_QTOUE, model->features);
2445         clear_bit(S390_FEAT_PTFF_STOE, model->features);
2446         clear_bit(S390_FEAT_PTFF_STOUE, model->features);
2447     }
2448 
2449     /* with cpu model support, CMM is only indicated if really available */
2450     if (kvm_s390_cmma_available()) {
2451         set_bit(S390_FEAT_CMM, model->features);
2452     } else {
2453         /* no cmm -> no cmm nt */
2454         clear_bit(S390_FEAT_CMM_NT, model->features);
2455     }
2456 
2457     /* bpb needs kernel support for migration, VSIE and reset */
2458     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_BPB)) {
2459         clear_bit(S390_FEAT_BPB, model->features);
2460     }
2461 
2462     /*
2463      * If we have support for protected virtualization, indicate
2464      * the protected virtualization IPL unpack facility.
2465      */
2466     if (cap_protected) {
2467         set_bit(S390_FEAT_UNPACK, model->features);
2468     }
2469 
2470     /*
2471      * If we have kernel support for CPU Topology indicate the
2472      * configuration-topology facility.
2473      */
2474     if (kvm_check_extension(kvm_state, KVM_CAP_S390_CPU_TOPOLOGY)) {
2475         set_bit(S390_FEAT_CONFIGURATION_TOPOLOGY, model->features);
2476     }
2477 
2478     /* We emulate a zPCI bus and AEN, therefore we don't need HW support */
2479     set_bit(S390_FEAT_ZPCI, model->features);
2480     set_bit(S390_FEAT_ADAPTER_EVENT_NOTIFICATION, model->features);
2481 
2482     if (s390_known_cpu_type(cpu_type)) {
2483         /* we want the exact model, even if some features are missing */
2484         model->def = s390_find_cpu_def(cpu_type, ibc_gen(unblocked_ibc),
2485                                        ibc_ec_ga(unblocked_ibc), NULL);
2486     } else {
2487         /* model unknown, e.g. too new - search using features */
2488         model->def = s390_find_cpu_def(0, ibc_gen(unblocked_ibc),
2489                                        ibc_ec_ga(unblocked_ibc),
2490                                        model->features);
2491     }
2492     if (!model->def) {
2493         error_setg(errp, "KVM: host CPU model could not be identified");
2494         return false;
2495     }
2496     /* for now, we can only provide the AP feature with HW support */
2497     if (ap_available()) {
2498         set_bit(S390_FEAT_AP, model->features);
2499     }
2500 
2501     /*
2502      * Extended-Length SCCB is handled entirely within QEMU.
2503      * For PV guests this is completely fenced by the Ultravisor, as Service
2504      * Call error checking and STFLE interpretation are handled via SIE.
2505      */
2506     set_bit(S390_FEAT_EXTENDED_LENGTH_SCCB, model->features);
2507 
2508     if (kvm_check_extension(kvm_state, KVM_CAP_S390_DIAG318)) {
2509         set_bit(S390_FEAT_DIAG_318, model->features);
2510     }
2511 
2512     /* Test for Ultravisor features that influence secure guest behavior */
2513     query_uv_feat_guest(model->features);
2514 
2515     /* strip of features that are not part of the maximum model */
2516     bitmap_and(model->features, model->features, model->def->full_feat,
2517                S390_FEAT_MAX);
2518     return true;
2519 }
2520 
2521 static int configure_uv_feat_guest(const S390FeatBitmap features)
2522 {
2523     struct kvm_s390_vm_cpu_uv_feat uv_feat = {};
2524     struct kvm_device_attr attribute = {
2525         .group = KVM_S390_VM_CPU_MODEL,
2526         .attr = KVM_S390_VM_CPU_PROCESSOR_UV_FEAT_GUEST,
2527         .addr = (__u64) &uv_feat,
2528     };
2529 
2530     /* AP support check is currently the only user of the UV feature test */
2531     if (!(uv_feat_supported() && ap_enabled(features))) {
2532         return 0;
2533     }
2534 
2535     if (test_bit(S390_FEAT_UV_FEAT_AP, features)) {
2536         uv_feat.ap = 1;
2537     }
2538     if (test_bit(S390_FEAT_UV_FEAT_AP_INTR, features)) {
2539         uv_feat.ap_intr = 1;
2540     }
2541 
2542     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
2543 }
2544 
2545 static void kvm_s390_configure_apie(bool interpret)
2546 {
2547     uint64_t attr = interpret ? KVM_S390_VM_CRYPTO_ENABLE_APIE :
2548                                 KVM_S390_VM_CRYPTO_DISABLE_APIE;
2549 
2550     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
2551         kvm_s390_set_crypto_attr(attr);
2552     }
2553 }
2554 
2555 bool kvm_s390_apply_cpu_model(const S390CPUModel *model, Error **errp)
2556 {
2557     struct kvm_s390_vm_cpu_processor prop  = {
2558         .fac_list = { 0 },
2559     };
2560     struct kvm_device_attr attr = {
2561         .group = KVM_S390_VM_CPU_MODEL,
2562         .attr = KVM_S390_VM_CPU_PROCESSOR,
2563         .addr = (uint64_t) &prop,
2564     };
2565     int rc;
2566 
2567     if (!model) {
2568         /* compatibility handling if cpu models are disabled */
2569         if (kvm_s390_cmma_available()) {
2570             kvm_s390_enable_cmma();
2571         }
2572         return true;
2573     }
2574     if (!kvm_s390_cpu_models_supported()) {
2575         error_setg(errp, "KVM doesn't support CPU models");
2576         return false;
2577     }
2578     prop.cpuid = s390_cpuid_from_cpu_model(model);
2579     prop.ibc = s390_ibc_from_cpu_model(model);
2580     /* configure cpu features indicated via STFL(e) */
2581     s390_fill_feat_block(model->features, S390_FEAT_TYPE_STFL,
2582                          (uint8_t *) prop.fac_list);
2583     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2584     if (rc) {
2585         error_setg(errp, "KVM: Error configuring the CPU model: %d", rc);
2586         return false;
2587     }
2588     /* configure cpu features indicated e.g. via SCLP */
2589     rc = configure_cpu_feat(model->features);
2590     if (rc) {
2591         error_setg(errp, "KVM: Error configuring CPU features: %d", rc);
2592         return false;
2593     }
2594     /* configure cpu subfunctions indicated via query / test bit */
2595     rc = configure_cpu_subfunc(model->features);
2596     if (rc) {
2597         error_setg(errp, "KVM: Error configuring CPU subfunctions: %d", rc);
2598         return false;
2599     }
2600     /* enable CMM via CMMA */
2601     if (test_bit(S390_FEAT_CMM, model->features)) {
2602         kvm_s390_enable_cmma();
2603     }
2604 
2605     if (ap_enabled(model->features)) {
2606         kvm_s390_configure_apie(true);
2607     }
2608 
2609     /* configure UV-features for the guest indicated via query / test_bit */
2610     rc = configure_uv_feat_guest(model->features);
2611     if (rc) {
2612         error_setg(errp, "KVM: Error configuring CPU UV features %d", rc);
2613         return false;
2614     }
2615     return true;
2616 }
2617 
2618 void kvm_s390_restart_interrupt(S390CPU *cpu)
2619 {
2620     struct kvm_s390_irq irq = {
2621         .type = KVM_S390_RESTART,
2622     };
2623 
2624     kvm_s390_vcpu_interrupt(cpu, &irq);
2625 }
2626 
2627 void kvm_s390_stop_interrupt(S390CPU *cpu)
2628 {
2629     struct kvm_s390_irq irq = {
2630         .type = KVM_S390_SIGP_STOP,
2631     };
2632 
2633     kvm_s390_vcpu_interrupt(cpu, &irq);
2634 }
2635 
2636 int kvm_s390_get_zpci_op(void)
2637 {
2638     return cap_zpci_op;
2639 }
2640 
2641 int kvm_s390_topology_set_mtcr(uint64_t attr)
2642 {
2643     struct kvm_device_attr attribute = {
2644         .group = KVM_S390_VM_CPU_TOPOLOGY,
2645         .attr  = attr,
2646     };
2647 
2648     if (!s390_has_feat(S390_FEAT_CONFIGURATION_TOPOLOGY)) {
2649         return 0;
2650     }
2651     if (!kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_TOPOLOGY, attr)) {
2652         return -ENOTSUP;
2653     }
2654 
2655     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
2656 }
2657 
2658 void kvm_arch_accel_class_init(ObjectClass *oc)
2659 {
2660 }
2661