xref: /qemu/target/s390x/kvm/kvm.c (revision e688df6bc4549f28534cdb001f168b8caae55b0c)
1 /*
2  * QEMU S390x KVM implementation
3  *
4  * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
5  * Copyright IBM Corp. 2012
6  *
7  * This library is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2 of the License, or (at your option) any later version.
11  *
12  * This library 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  * Lesser General Public License for more details.
16  *
17  * Contributions after 2012-10-29 are licensed under the terms of the
18  * GNU GPL, version 2 or (at your option) any later version.
19  *
20  * You should have received a copy of the GNU (Lesser) General Public
21  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
22  */
23 
24 #include "qemu/osdep.h"
25 #include <sys/ioctl.h>
26 
27 #include <linux/kvm.h>
28 #include <asm/ptrace.h>
29 
30 #include "qemu-common.h"
31 #include "cpu.h"
32 #include "internal.h"
33 #include "kvm_s390x.h"
34 #include "qapi/error.h"
35 #include "qemu/error-report.h"
36 #include "qemu/timer.h"
37 #include "sysemu/sysemu.h"
38 #include "sysemu/hw_accel.h"
39 #include "hw/hw.h"
40 #include "sysemu/device_tree.h"
41 #include "qapi/qmp/qjson.h"
42 #include "exec/gdbstub.h"
43 #include "exec/address-spaces.h"
44 #include "trace.h"
45 #include "qapi-event.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 
54 #ifndef DEBUG_KVM
55 #define DEBUG_KVM  0
56 #endif
57 
58 #define DPRINTF(fmt, ...) do {                \
59     if (DEBUG_KVM) {                          \
60         fprintf(stderr, fmt, ## __VA_ARGS__); \
61     }                                         \
62 } while (0)
63 
64 #define kvm_vm_check_mem_attr(s, attr) \
65     kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
66 
67 #define IPA0_DIAG                       0x8300
68 #define IPA0_SIGP                       0xae00
69 #define IPA0_B2                         0xb200
70 #define IPA0_B9                         0xb900
71 #define IPA0_EB                         0xeb00
72 #define IPA0_E3                         0xe300
73 
74 #define PRIV_B2_SCLP_CALL               0x20
75 #define PRIV_B2_CSCH                    0x30
76 #define PRIV_B2_HSCH                    0x31
77 #define PRIV_B2_MSCH                    0x32
78 #define PRIV_B2_SSCH                    0x33
79 #define PRIV_B2_STSCH                   0x34
80 #define PRIV_B2_TSCH                    0x35
81 #define PRIV_B2_TPI                     0x36
82 #define PRIV_B2_SAL                     0x37
83 #define PRIV_B2_RSCH                    0x38
84 #define PRIV_B2_STCRW                   0x39
85 #define PRIV_B2_STCPS                   0x3a
86 #define PRIV_B2_RCHP                    0x3b
87 #define PRIV_B2_SCHM                    0x3c
88 #define PRIV_B2_CHSC                    0x5f
89 #define PRIV_B2_SIGA                    0x74
90 #define PRIV_B2_XSCH                    0x76
91 
92 #define PRIV_EB_SQBS                    0x8a
93 #define PRIV_EB_PCISTB                  0xd0
94 #define PRIV_EB_SIC                     0xd1
95 
96 #define PRIV_B9_EQBS                    0x9c
97 #define PRIV_B9_CLP                     0xa0
98 #define PRIV_B9_PCISTG                  0xd0
99 #define PRIV_B9_PCILG                   0xd2
100 #define PRIV_B9_RPCIT                   0xd3
101 
102 #define PRIV_E3_MPCIFC                  0xd0
103 #define PRIV_E3_STPCIFC                 0xd4
104 
105 #define DIAG_TIMEREVENT                 0x288
106 #define DIAG_IPL                        0x308
107 #define DIAG_KVM_HYPERCALL              0x500
108 #define DIAG_KVM_BREAKPOINT             0x501
109 
110 #define ICPT_INSTRUCTION                0x04
111 #define ICPT_PROGRAM                    0x08
112 #define ICPT_EXT_INT                    0x14
113 #define ICPT_WAITPSW                    0x1c
114 #define ICPT_SOFT_INTERCEPT             0x24
115 #define ICPT_CPU_STOP                   0x28
116 #define ICPT_OPEREXC                    0x2c
117 #define ICPT_IO                         0x40
118 
119 #define NR_LOCAL_IRQS 32
120 /*
121  * Needs to be big enough to contain max_cpus emergency signals
122  * and in addition NR_LOCAL_IRQS interrupts
123  */
124 #define VCPU_IRQ_BUF_SIZE (sizeof(struct kvm_s390_irq) * \
125                            (max_cpus + NR_LOCAL_IRQS))
126 
127 static CPUWatchpoint hw_watchpoint;
128 /*
129  * We don't use a list because this structure is also used to transmit the
130  * hardware breakpoints to the kernel.
131  */
132 static struct kvm_hw_breakpoint *hw_breakpoints;
133 static int nb_hw_breakpoints;
134 
135 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
136     KVM_CAP_LAST_INFO
137 };
138 
139 static int cap_sync_regs;
140 static int cap_async_pf;
141 static int cap_mem_op;
142 static int cap_s390_irq;
143 static int cap_ri;
144 static int cap_gs;
145 
146 static int active_cmma;
147 
148 static void *legacy_s390_alloc(size_t size, uint64_t *align);
149 
150 static int kvm_s390_query_mem_limit(uint64_t *memory_limit)
151 {
152     struct kvm_device_attr attr = {
153         .group = KVM_S390_VM_MEM_CTRL,
154         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
155         .addr = (uint64_t) memory_limit,
156     };
157 
158     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
159 }
160 
161 int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit)
162 {
163     int rc;
164 
165     struct kvm_device_attr attr = {
166         .group = KVM_S390_VM_MEM_CTRL,
167         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
168         .addr = (uint64_t) &new_limit,
169     };
170 
171     if (!kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_LIMIT_SIZE)) {
172         return 0;
173     }
174 
175     rc = kvm_s390_query_mem_limit(hw_limit);
176     if (rc) {
177         return rc;
178     } else if (*hw_limit < new_limit) {
179         return -E2BIG;
180     }
181 
182     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
183 }
184 
185 int kvm_s390_cmma_active(void)
186 {
187     return active_cmma;
188 }
189 
190 static bool kvm_s390_cmma_available(void)
191 {
192     static bool initialized, value;
193 
194     if (!initialized) {
195         initialized = true;
196         value = kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_ENABLE_CMMA) &&
197                 kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_CLR_CMMA);
198     }
199     return value;
200 }
201 
202 void kvm_s390_cmma_reset(void)
203 {
204     int rc;
205     struct kvm_device_attr attr = {
206         .group = KVM_S390_VM_MEM_CTRL,
207         .attr = KVM_S390_VM_MEM_CLR_CMMA,
208     };
209 
210     if (!kvm_s390_cmma_active()) {
211         return;
212     }
213 
214     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
215     trace_kvm_clear_cmma(rc);
216 }
217 
218 static void kvm_s390_enable_cmma(void)
219 {
220     int rc;
221     struct kvm_device_attr attr = {
222         .group = KVM_S390_VM_MEM_CTRL,
223         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
224     };
225 
226     if (mem_path) {
227         warn_report("CMM will not be enabled because it is not "
228                     "compatible with hugetlbfs.");
229         return;
230     }
231     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
232     active_cmma = !rc;
233     trace_kvm_enable_cmma(rc);
234 }
235 
236 static void kvm_s390_set_attr(uint64_t attr)
237 {
238     struct kvm_device_attr attribute = {
239         .group = KVM_S390_VM_CRYPTO,
240         .attr  = attr,
241     };
242 
243     int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
244 
245     if (ret) {
246         error_report("Failed to set crypto device attribute %lu: %s",
247                      attr, strerror(-ret));
248     }
249 }
250 
251 static void kvm_s390_init_aes_kw(void)
252 {
253     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
254 
255     if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
256                                  NULL)) {
257             attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
258     }
259 
260     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
261             kvm_s390_set_attr(attr);
262     }
263 }
264 
265 static void kvm_s390_init_dea_kw(void)
266 {
267     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
268 
269     if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
270                                  NULL)) {
271             attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
272     }
273 
274     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
275             kvm_s390_set_attr(attr);
276     }
277 }
278 
279 void kvm_s390_crypto_reset(void)
280 {
281     if (s390_has_feat(S390_FEAT_MSA_EXT_3)) {
282         kvm_s390_init_aes_kw();
283         kvm_s390_init_dea_kw();
284     }
285 }
286 
287 int kvm_arch_init(MachineState *ms, KVMState *s)
288 {
289     MachineClass *mc = MACHINE_GET_CLASS(ms);
290 
291     mc->default_cpu_type = S390_CPU_TYPE_NAME("host");
292     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
293     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
294     cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
295     cap_s390_irq = kvm_check_extension(s, KVM_CAP_S390_INJECT_IRQ);
296 
297     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
298         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
299         phys_mem_set_alloc(legacy_s390_alloc);
300     }
301 
302     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
303     kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
304     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
305     if (ri_allowed()) {
306         if (kvm_vm_enable_cap(s, KVM_CAP_S390_RI, 0) == 0) {
307             cap_ri = 1;
308         }
309     }
310     if (cpu_model_allowed()) {
311         if (kvm_vm_enable_cap(s, KVM_CAP_S390_GS, 0) == 0) {
312             cap_gs = 1;
313         }
314     }
315 
316     /*
317      * The migration interface for ais was introduced with kernel 4.13
318      * but the capability itself had been active since 4.12. As migration
319      * support is considered necessary let's disable ais in the 2.10
320      * machine.
321      */
322     /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
323 
324     return 0;
325 }
326 
327 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s)
328 {
329     return 0;
330 }
331 
332 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
333 {
334     return cpu->cpu_index;
335 }
336 
337 int kvm_arch_init_vcpu(CPUState *cs)
338 {
339     S390CPU *cpu = S390_CPU(cs);
340     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
341     cpu->irqstate = g_malloc0(VCPU_IRQ_BUF_SIZE);
342     return 0;
343 }
344 
345 void kvm_s390_reset_vcpu(S390CPU *cpu)
346 {
347     CPUState *cs = CPU(cpu);
348 
349     /* The initial reset call is needed here to reset in-kernel
350      * vcpu data that we can't access directly from QEMU
351      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
352      * Before this ioctl cpu_synchronize_state() is called in common kvm
353      * code (kvm-all) */
354     if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
355         error_report("Initial CPU reset failed on CPU %i", cs->cpu_index);
356     }
357 }
358 
359 static int can_sync_regs(CPUState *cs, int regs)
360 {
361     return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
362 }
363 
364 int kvm_arch_put_registers(CPUState *cs, int level)
365 {
366     S390CPU *cpu = S390_CPU(cs);
367     CPUS390XState *env = &cpu->env;
368     struct kvm_sregs sregs;
369     struct kvm_regs regs;
370     struct kvm_fpu fpu = {};
371     int r;
372     int i;
373 
374     /* always save the PSW  and the GPRS*/
375     cs->kvm_run->psw_addr = env->psw.addr;
376     cs->kvm_run->psw_mask = env->psw.mask;
377 
378     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
379         for (i = 0; i < 16; i++) {
380             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
381             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
382         }
383     } else {
384         for (i = 0; i < 16; i++) {
385             regs.gprs[i] = env->regs[i];
386         }
387         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
388         if (r < 0) {
389             return r;
390         }
391     }
392 
393     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
394         for (i = 0; i < 32; i++) {
395             cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0].ll;
396             cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1].ll;
397         }
398         cs->kvm_run->s.regs.fpc = env->fpc;
399         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
400     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
401         for (i = 0; i < 16; i++) {
402             cs->kvm_run->s.regs.fprs[i] = get_freg(env, i)->ll;
403         }
404         cs->kvm_run->s.regs.fpc = env->fpc;
405         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_FPRS;
406     } else {
407         /* Floating point */
408         for (i = 0; i < 16; i++) {
409             fpu.fprs[i] = get_freg(env, i)->ll;
410         }
411         fpu.fpc = env->fpc;
412 
413         r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
414         if (r < 0) {
415             return r;
416         }
417     }
418 
419     /* Do we need to save more than that? */
420     if (level == KVM_PUT_RUNTIME_STATE) {
421         return 0;
422     }
423 
424     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
425         cs->kvm_run->s.regs.cputm = env->cputm;
426         cs->kvm_run->s.regs.ckc = env->ckc;
427         cs->kvm_run->s.regs.todpr = env->todpr;
428         cs->kvm_run->s.regs.gbea = env->gbea;
429         cs->kvm_run->s.regs.pp = env->pp;
430         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
431     } else {
432         /*
433          * These ONE_REGS are not protected by a capability. As they are only
434          * necessary for migration we just trace a possible error, but don't
435          * return with an error return code.
436          */
437         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
438         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
439         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
440         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
441         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
442     }
443 
444     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
445         memcpy(cs->kvm_run->s.regs.riccb, env->riccb, 64);
446         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_RICCB;
447     }
448 
449     /* pfault parameters */
450     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
451         cs->kvm_run->s.regs.pft = env->pfault_token;
452         cs->kvm_run->s.regs.pfs = env->pfault_select;
453         cs->kvm_run->s.regs.pfc = env->pfault_compare;
454         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
455     } else if (cap_async_pf) {
456         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
457         if (r < 0) {
458             return r;
459         }
460         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
461         if (r < 0) {
462             return r;
463         }
464         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
465         if (r < 0) {
466             return r;
467         }
468     }
469 
470     /* access registers and control registers*/
471     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
472         for (i = 0; i < 16; i++) {
473             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
474             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
475         }
476         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
477         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
478     } else {
479         for (i = 0; i < 16; i++) {
480             sregs.acrs[i] = env->aregs[i];
481             sregs.crs[i] = env->cregs[i];
482         }
483         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
484         if (r < 0) {
485             return r;
486         }
487     }
488 
489     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
490         memcpy(cs->kvm_run->s.regs.gscb, env->gscb, 32);
491         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GSCB;
492     }
493 
494     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
495         cs->kvm_run->s.regs.bpbc = env->bpbc;
496         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_BPBC;
497     }
498 
499     /* Finally the prefix */
500     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
501         cs->kvm_run->s.regs.prefix = env->psa;
502         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
503     } else {
504         /* prefix is only supported via sync regs */
505     }
506     return 0;
507 }
508 
509 int kvm_arch_get_registers(CPUState *cs)
510 {
511     S390CPU *cpu = S390_CPU(cs);
512     CPUS390XState *env = &cpu->env;
513     struct kvm_sregs sregs;
514     struct kvm_regs regs;
515     struct kvm_fpu fpu;
516     int i, r;
517 
518     /* get the PSW */
519     env->psw.addr = cs->kvm_run->psw_addr;
520     env->psw.mask = cs->kvm_run->psw_mask;
521 
522     /* the GPRS */
523     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
524         for (i = 0; i < 16; i++) {
525             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
526         }
527     } else {
528         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
529         if (r < 0) {
530             return r;
531         }
532          for (i = 0; i < 16; i++) {
533             env->regs[i] = regs.gprs[i];
534         }
535     }
536 
537     /* The ACRS and CRS */
538     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
539         for (i = 0; i < 16; i++) {
540             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
541             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
542         }
543     } else {
544         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
545         if (r < 0) {
546             return r;
547         }
548          for (i = 0; i < 16; i++) {
549             env->aregs[i] = sregs.acrs[i];
550             env->cregs[i] = sregs.crs[i];
551         }
552     }
553 
554     /* Floating point and vector registers */
555     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
556         for (i = 0; i < 32; i++) {
557             env->vregs[i][0].ll = cs->kvm_run->s.regs.vrs[i][0];
558             env->vregs[i][1].ll = cs->kvm_run->s.regs.vrs[i][1];
559         }
560         env->fpc = cs->kvm_run->s.regs.fpc;
561     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
562         for (i = 0; i < 16; i++) {
563             get_freg(env, i)->ll = cs->kvm_run->s.regs.fprs[i];
564         }
565         env->fpc = cs->kvm_run->s.regs.fpc;
566     } else {
567         r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
568         if (r < 0) {
569             return r;
570         }
571         for (i = 0; i < 16; i++) {
572             get_freg(env, i)->ll = fpu.fprs[i];
573         }
574         env->fpc = fpu.fpc;
575     }
576 
577     /* The prefix */
578     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
579         env->psa = cs->kvm_run->s.regs.prefix;
580     }
581 
582     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
583         env->cputm = cs->kvm_run->s.regs.cputm;
584         env->ckc = cs->kvm_run->s.regs.ckc;
585         env->todpr = cs->kvm_run->s.regs.todpr;
586         env->gbea = cs->kvm_run->s.regs.gbea;
587         env->pp = cs->kvm_run->s.regs.pp;
588     } else {
589         /*
590          * These ONE_REGS are not protected by a capability. As they are only
591          * necessary for migration we just trace a possible error, but don't
592          * return with an error return code.
593          */
594         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
595         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
596         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
597         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
598         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
599     }
600 
601     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
602         memcpy(env->riccb, cs->kvm_run->s.regs.riccb, 64);
603     }
604 
605     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
606         memcpy(env->gscb, cs->kvm_run->s.regs.gscb, 32);
607     }
608 
609     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
610         env->bpbc = cs->kvm_run->s.regs.bpbc;
611     }
612 
613     /* pfault parameters */
614     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
615         env->pfault_token = cs->kvm_run->s.regs.pft;
616         env->pfault_select = cs->kvm_run->s.regs.pfs;
617         env->pfault_compare = cs->kvm_run->s.regs.pfc;
618     } else if (cap_async_pf) {
619         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
620         if (r < 0) {
621             return r;
622         }
623         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
624         if (r < 0) {
625             return r;
626         }
627         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
628         if (r < 0) {
629             return r;
630         }
631     }
632 
633     return 0;
634 }
635 
636 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
637 {
638     int r;
639     struct kvm_device_attr attr = {
640         .group = KVM_S390_VM_TOD,
641         .attr = KVM_S390_VM_TOD_LOW,
642         .addr = (uint64_t)tod_low,
643     };
644 
645     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
646     if (r) {
647         return r;
648     }
649 
650     attr.attr = KVM_S390_VM_TOD_HIGH;
651     attr.addr = (uint64_t)tod_high;
652     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
653 }
654 
655 int kvm_s390_get_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
656 {
657     int r;
658     struct kvm_s390_vm_tod_clock gtod;
659     struct kvm_device_attr attr = {
660         .group = KVM_S390_VM_TOD,
661         .attr = KVM_S390_VM_TOD_EXT,
662         .addr = (uint64_t)&gtod,
663     };
664 
665     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
666     *tod_high = gtod.epoch_idx;
667     *tod_low  = gtod.tod;
668 
669     return r;
670 }
671 
672 int kvm_s390_set_clock(uint8_t *tod_high, uint64_t *tod_low)
673 {
674     int r;
675     struct kvm_device_attr attr = {
676         .group = KVM_S390_VM_TOD,
677         .attr = KVM_S390_VM_TOD_LOW,
678         .addr = (uint64_t)tod_low,
679     };
680 
681     r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
682     if (r) {
683         return r;
684     }
685 
686     attr.attr = KVM_S390_VM_TOD_HIGH;
687     attr.addr = (uint64_t)tod_high;
688     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
689 }
690 
691 int kvm_s390_set_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
692 {
693     struct kvm_s390_vm_tod_clock gtod = {
694         .epoch_idx = *tod_high,
695         .tod  = *tod_low,
696     };
697     struct kvm_device_attr attr = {
698         .group = KVM_S390_VM_TOD,
699         .attr = KVM_S390_VM_TOD_EXT,
700         .addr = (uint64_t)&gtod,
701     };
702 
703     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
704 }
705 
706 /**
707  * kvm_s390_mem_op:
708  * @addr:      the logical start address in guest memory
709  * @ar:        the access register number
710  * @hostbuf:   buffer in host memory. NULL = do only checks w/o copying
711  * @len:       length that should be transferred
712  * @is_write:  true = write, false = read
713  * Returns:    0 on success, non-zero if an exception or error occurred
714  *
715  * Use KVM ioctl to read/write from/to guest memory. An access exception
716  * is injected into the vCPU in case of translation errors.
717  */
718 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, uint8_t ar, void *hostbuf,
719                     int len, bool is_write)
720 {
721     struct kvm_s390_mem_op mem_op = {
722         .gaddr = addr,
723         .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
724         .size = len,
725         .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
726                        : KVM_S390_MEMOP_LOGICAL_READ,
727         .buf = (uint64_t)hostbuf,
728         .ar = ar,
729     };
730     int ret;
731 
732     if (!cap_mem_op) {
733         return -ENOSYS;
734     }
735     if (!hostbuf) {
736         mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
737     }
738 
739     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
740     if (ret < 0) {
741         error_printf("KVM_S390_MEM_OP failed: %s\n", strerror(-ret));
742     }
743     return ret;
744 }
745 
746 /*
747  * Legacy layout for s390:
748  * Older S390 KVM requires the topmost vma of the RAM to be
749  * smaller than an system defined value, which is at least 256GB.
750  * Larger systems have larger values. We put the guest between
751  * the end of data segment (system break) and this value. We
752  * use 32GB as a base to have enough room for the system break
753  * to grow. We also have to use MAP parameters that avoid
754  * read-only mapping of guest pages.
755  */
756 static void *legacy_s390_alloc(size_t size, uint64_t *align)
757 {
758     void *mem;
759 
760     mem = mmap((void *) 0x800000000ULL, size,
761                PROT_EXEC|PROT_READ|PROT_WRITE,
762                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
763     return mem == MAP_FAILED ? NULL : mem;
764 }
765 
766 static uint8_t const *sw_bp_inst;
767 static uint8_t sw_bp_ilen;
768 
769 static void determine_sw_breakpoint_instr(void)
770 {
771         /* DIAG 501 is used for sw breakpoints with old kernels */
772         static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
773         /* Instruction 0x0000 is used for sw breakpoints with recent kernels */
774         static const uint8_t instr_0x0000[] = {0x00, 0x00};
775 
776         if (sw_bp_inst) {
777             return;
778         }
779         if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_USER_INSTR0, 0)) {
780             sw_bp_inst = diag_501;
781             sw_bp_ilen = sizeof(diag_501);
782             DPRINTF("KVM: will use 4-byte sw breakpoints.\n");
783         } else {
784             sw_bp_inst = instr_0x0000;
785             sw_bp_ilen = sizeof(instr_0x0000);
786             DPRINTF("KVM: will use 2-byte sw breakpoints.\n");
787         }
788 }
789 
790 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
791 {
792     determine_sw_breakpoint_instr();
793 
794     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
795                             sw_bp_ilen, 0) ||
796         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)sw_bp_inst, sw_bp_ilen, 1)) {
797         return -EINVAL;
798     }
799     return 0;
800 }
801 
802 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
803 {
804     uint8_t t[MAX_ILEN];
805 
806     if (cpu_memory_rw_debug(cs, bp->pc, t, sw_bp_ilen, 0)) {
807         return -EINVAL;
808     } else if (memcmp(t, sw_bp_inst, sw_bp_ilen)) {
809         return -EINVAL;
810     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
811                                    sw_bp_ilen, 1)) {
812         return -EINVAL;
813     }
814 
815     return 0;
816 }
817 
818 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
819                                                     int len, int type)
820 {
821     int n;
822 
823     for (n = 0; n < nb_hw_breakpoints; n++) {
824         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
825             (hw_breakpoints[n].len == len || len == -1)) {
826             return &hw_breakpoints[n];
827         }
828     }
829 
830     return NULL;
831 }
832 
833 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
834 {
835     int size;
836 
837     if (find_hw_breakpoint(addr, len, type)) {
838         return -EEXIST;
839     }
840 
841     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
842 
843     if (!hw_breakpoints) {
844         nb_hw_breakpoints = 0;
845         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
846     } else {
847         hw_breakpoints =
848             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
849     }
850 
851     if (!hw_breakpoints) {
852         nb_hw_breakpoints = 0;
853         return -ENOMEM;
854     }
855 
856     hw_breakpoints[nb_hw_breakpoints].addr = addr;
857     hw_breakpoints[nb_hw_breakpoints].len = len;
858     hw_breakpoints[nb_hw_breakpoints].type = type;
859 
860     nb_hw_breakpoints++;
861 
862     return 0;
863 }
864 
865 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
866                                   target_ulong len, int type)
867 {
868     switch (type) {
869     case GDB_BREAKPOINT_HW:
870         type = KVM_HW_BP;
871         break;
872     case GDB_WATCHPOINT_WRITE:
873         if (len < 1) {
874             return -EINVAL;
875         }
876         type = KVM_HW_WP_WRITE;
877         break;
878     default:
879         return -ENOSYS;
880     }
881     return insert_hw_breakpoint(addr, len, type);
882 }
883 
884 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
885                                   target_ulong len, int type)
886 {
887     int size;
888     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
889 
890     if (bp == NULL) {
891         return -ENOENT;
892     }
893 
894     nb_hw_breakpoints--;
895     if (nb_hw_breakpoints > 0) {
896         /*
897          * In order to trim the array, move the last element to the position to
898          * be removed - if necessary.
899          */
900         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
901             *bp = hw_breakpoints[nb_hw_breakpoints];
902         }
903         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
904         hw_breakpoints =
905              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
906     } else {
907         g_free(hw_breakpoints);
908         hw_breakpoints = NULL;
909     }
910 
911     return 0;
912 }
913 
914 void kvm_arch_remove_all_hw_breakpoints(void)
915 {
916     nb_hw_breakpoints = 0;
917     g_free(hw_breakpoints);
918     hw_breakpoints = NULL;
919 }
920 
921 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
922 {
923     int i;
924 
925     if (nb_hw_breakpoints > 0) {
926         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
927         dbg->arch.hw_bp = hw_breakpoints;
928 
929         for (i = 0; i < nb_hw_breakpoints; ++i) {
930             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
931                                                        hw_breakpoints[i].addr);
932         }
933         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
934     } else {
935         dbg->arch.nr_hw_bp = 0;
936         dbg->arch.hw_bp = NULL;
937     }
938 }
939 
940 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
941 {
942 }
943 
944 MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
945 {
946     return MEMTXATTRS_UNSPECIFIED;
947 }
948 
949 int kvm_arch_process_async_events(CPUState *cs)
950 {
951     return cs->halted;
952 }
953 
954 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
955                                      struct kvm_s390_interrupt *interrupt)
956 {
957     int r = 0;
958 
959     interrupt->type = irq->type;
960     switch (irq->type) {
961     case KVM_S390_INT_VIRTIO:
962         interrupt->parm = irq->u.ext.ext_params;
963         /* fall through */
964     case KVM_S390_INT_PFAULT_INIT:
965     case KVM_S390_INT_PFAULT_DONE:
966         interrupt->parm64 = irq->u.ext.ext_params2;
967         break;
968     case KVM_S390_PROGRAM_INT:
969         interrupt->parm = irq->u.pgm.code;
970         break;
971     case KVM_S390_SIGP_SET_PREFIX:
972         interrupt->parm = irq->u.prefix.address;
973         break;
974     case KVM_S390_INT_SERVICE:
975         interrupt->parm = irq->u.ext.ext_params;
976         break;
977     case KVM_S390_MCHK:
978         interrupt->parm = irq->u.mchk.cr14;
979         interrupt->parm64 = irq->u.mchk.mcic;
980         break;
981     case KVM_S390_INT_EXTERNAL_CALL:
982         interrupt->parm = irq->u.extcall.code;
983         break;
984     case KVM_S390_INT_EMERGENCY:
985         interrupt->parm = irq->u.emerg.code;
986         break;
987     case KVM_S390_SIGP_STOP:
988     case KVM_S390_RESTART:
989         break; /* These types have no parameters */
990     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
991         interrupt->parm = irq->u.io.subchannel_id << 16;
992         interrupt->parm |= irq->u.io.subchannel_nr;
993         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
994         interrupt->parm64 |= irq->u.io.io_int_word;
995         break;
996     default:
997         r = -EINVAL;
998         break;
999     }
1000     return r;
1001 }
1002 
1003 static void inject_vcpu_irq_legacy(CPUState *cs, struct kvm_s390_irq *irq)
1004 {
1005     struct kvm_s390_interrupt kvmint = {};
1006     int r;
1007 
1008     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1009     if (r < 0) {
1010         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1011         exit(1);
1012     }
1013 
1014     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
1015     if (r < 0) {
1016         fprintf(stderr, "KVM failed to inject interrupt\n");
1017         exit(1);
1018     }
1019 }
1020 
1021 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
1022 {
1023     CPUState *cs = CPU(cpu);
1024     int r;
1025 
1026     if (cap_s390_irq) {
1027         r = kvm_vcpu_ioctl(cs, KVM_S390_IRQ, irq);
1028         if (!r) {
1029             return;
1030         }
1031         error_report("KVM failed to inject interrupt %llx", irq->type);
1032         exit(1);
1033     }
1034 
1035     inject_vcpu_irq_legacy(cs, irq);
1036 }
1037 
1038 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
1039 {
1040     struct kvm_s390_interrupt kvmint = {};
1041     int r;
1042 
1043     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1044     if (r < 0) {
1045         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1046         exit(1);
1047     }
1048 
1049     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
1050     if (r < 0) {
1051         fprintf(stderr, "KVM failed to inject interrupt\n");
1052         exit(1);
1053     }
1054 }
1055 
1056 void kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
1057 {
1058     static bool use_flic = true;
1059     int r;
1060 
1061     if (use_flic) {
1062         r = kvm_s390_inject_flic(irq);
1063         if (r == -ENOSYS) {
1064             use_flic = false;
1065         }
1066         if (!r) {
1067             return;
1068         }
1069     }
1070     __kvm_s390_floating_interrupt(irq);
1071 }
1072 
1073 void kvm_s390_service_interrupt(uint32_t parm)
1074 {
1075     struct kvm_s390_irq irq = {
1076         .type = KVM_S390_INT_SERVICE,
1077         .u.ext.ext_params = parm,
1078     };
1079 
1080     kvm_s390_floating_interrupt(&irq);
1081 }
1082 
1083 void kvm_s390_program_interrupt(S390CPU *cpu, uint16_t code)
1084 {
1085     struct kvm_s390_irq irq = {
1086         .type = KVM_S390_PROGRAM_INT,
1087         .u.pgm.code = code,
1088     };
1089 
1090     kvm_s390_vcpu_interrupt(cpu, &irq);
1091 }
1092 
1093 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
1094 {
1095     struct kvm_s390_irq irq = {
1096         .type = KVM_S390_PROGRAM_INT,
1097         .u.pgm.code = code,
1098         .u.pgm.trans_exc_code = te_code,
1099         .u.pgm.exc_access_id = te_code & 3,
1100     };
1101 
1102     kvm_s390_vcpu_interrupt(cpu, &irq);
1103 }
1104 
1105 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
1106                                  uint16_t ipbh0)
1107 {
1108     CPUS390XState *env = &cpu->env;
1109     uint64_t sccb;
1110     uint32_t code;
1111     int r = 0;
1112 
1113     cpu_synchronize_state(CPU(cpu));
1114     sccb = env->regs[ipbh0 & 0xf];
1115     code = env->regs[(ipbh0 & 0xf0) >> 4];
1116 
1117     r = sclp_service_call(env, sccb, code);
1118     if (r < 0) {
1119         kvm_s390_program_interrupt(cpu, -r);
1120     } else {
1121         setcc(cpu, r);
1122     }
1123 
1124     return 0;
1125 }
1126 
1127 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1128 {
1129     CPUS390XState *env = &cpu->env;
1130     int rc = 0;
1131     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
1132 
1133     cpu_synchronize_state(CPU(cpu));
1134 
1135     switch (ipa1) {
1136     case PRIV_B2_XSCH:
1137         ioinst_handle_xsch(cpu, env->regs[1], RA_IGNORED);
1138         break;
1139     case PRIV_B2_CSCH:
1140         ioinst_handle_csch(cpu, env->regs[1], RA_IGNORED);
1141         break;
1142     case PRIV_B2_HSCH:
1143         ioinst_handle_hsch(cpu, env->regs[1], RA_IGNORED);
1144         break;
1145     case PRIV_B2_MSCH:
1146         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1147         break;
1148     case PRIV_B2_SSCH:
1149         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1150         break;
1151     case PRIV_B2_STCRW:
1152         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb, RA_IGNORED);
1153         break;
1154     case PRIV_B2_STSCH:
1155         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1156         break;
1157     case PRIV_B2_TSCH:
1158         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1159         fprintf(stderr, "Spurious tsch intercept\n");
1160         break;
1161     case PRIV_B2_CHSC:
1162         ioinst_handle_chsc(cpu, run->s390_sieic.ipb, RA_IGNORED);
1163         break;
1164     case PRIV_B2_TPI:
1165         /* This should have been handled by kvm already. */
1166         fprintf(stderr, "Spurious tpi intercept\n");
1167         break;
1168     case PRIV_B2_SCHM:
1169         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
1170                            run->s390_sieic.ipb, RA_IGNORED);
1171         break;
1172     case PRIV_B2_RSCH:
1173         ioinst_handle_rsch(cpu, env->regs[1], RA_IGNORED);
1174         break;
1175     case PRIV_B2_RCHP:
1176         ioinst_handle_rchp(cpu, env->regs[1], RA_IGNORED);
1177         break;
1178     case PRIV_B2_STCPS:
1179         /* We do not provide this instruction, it is suppressed. */
1180         break;
1181     case PRIV_B2_SAL:
1182         ioinst_handle_sal(cpu, env->regs[1], RA_IGNORED);
1183         break;
1184     case PRIV_B2_SIGA:
1185         /* Not provided, set CC = 3 for subchannel not operational */
1186         setcc(cpu, 3);
1187         break;
1188     case PRIV_B2_SCLP_CALL:
1189         rc = kvm_sclp_service_call(cpu, run, ipbh0);
1190         break;
1191     default:
1192         rc = -1;
1193         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
1194         break;
1195     }
1196 
1197     return rc;
1198 }
1199 
1200 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run,
1201                                   uint8_t *ar)
1202 {
1203     CPUS390XState *env = &cpu->env;
1204     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1205     uint32_t base2 = run->s390_sieic.ipb >> 28;
1206     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1207                      ((run->s390_sieic.ipb & 0xff00) << 4);
1208 
1209     if (disp2 & 0x80000) {
1210         disp2 += 0xfff00000;
1211     }
1212     if (ar) {
1213         *ar = base2;
1214     }
1215 
1216     return (base2 ? env->regs[base2] : 0) +
1217            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1218 }
1219 
1220 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run,
1221                                   uint8_t *ar)
1222 {
1223     CPUS390XState *env = &cpu->env;
1224     uint32_t base2 = run->s390_sieic.ipb >> 28;
1225     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1226                      ((run->s390_sieic.ipb & 0xff00) << 4);
1227 
1228     if (disp2 & 0x80000) {
1229         disp2 += 0xfff00000;
1230     }
1231     if (ar) {
1232         *ar = base2;
1233     }
1234 
1235     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1236 }
1237 
1238 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1239 {
1240     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1241 
1242     if (s390_has_feat(S390_FEAT_ZPCI)) {
1243         return clp_service_call(cpu, r2, RA_IGNORED);
1244     } else {
1245         return -1;
1246     }
1247 }
1248 
1249 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1250 {
1251     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1252     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1253 
1254     if (s390_has_feat(S390_FEAT_ZPCI)) {
1255         return pcilg_service_call(cpu, r1, r2, RA_IGNORED);
1256     } else {
1257         return -1;
1258     }
1259 }
1260 
1261 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1262 {
1263     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1264     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1265 
1266     if (s390_has_feat(S390_FEAT_ZPCI)) {
1267         return pcistg_service_call(cpu, r1, r2, RA_IGNORED);
1268     } else {
1269         return -1;
1270     }
1271 }
1272 
1273 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1274 {
1275     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1276     uint64_t fiba;
1277     uint8_t ar;
1278 
1279     if (s390_has_feat(S390_FEAT_ZPCI)) {
1280         cpu_synchronize_state(CPU(cpu));
1281         fiba = get_base_disp_rxy(cpu, run, &ar);
1282 
1283         return stpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1284     } else {
1285         return -1;
1286     }
1287 }
1288 
1289 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1290 {
1291     CPUS390XState *env = &cpu->env;
1292     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1293     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1294     uint8_t isc;
1295     uint16_t mode;
1296     int r;
1297 
1298     cpu_synchronize_state(CPU(cpu));
1299     mode = env->regs[r1] & 0xffff;
1300     isc = (env->regs[r3] >> 27) & 0x7;
1301     r = css_do_sic(env, isc, mode);
1302     if (r) {
1303         kvm_s390_program_interrupt(cpu, -r);
1304     }
1305 
1306     return 0;
1307 }
1308 
1309 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1310 {
1311     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1312     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1313 
1314     if (s390_has_feat(S390_FEAT_ZPCI)) {
1315         return rpcit_service_call(cpu, r1, r2, RA_IGNORED);
1316     } else {
1317         return -1;
1318     }
1319 }
1320 
1321 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1322 {
1323     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1324     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1325     uint64_t gaddr;
1326     uint8_t ar;
1327 
1328     if (s390_has_feat(S390_FEAT_ZPCI)) {
1329         cpu_synchronize_state(CPU(cpu));
1330         gaddr = get_base_disp_rsy(cpu, run, &ar);
1331 
1332         return pcistb_service_call(cpu, r1, r3, gaddr, ar, RA_IGNORED);
1333     } else {
1334         return -1;
1335     }
1336 }
1337 
1338 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1339 {
1340     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1341     uint64_t fiba;
1342     uint8_t ar;
1343 
1344     if (s390_has_feat(S390_FEAT_ZPCI)) {
1345         cpu_synchronize_state(CPU(cpu));
1346         fiba = get_base_disp_rxy(cpu, run, &ar);
1347 
1348         return mpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1349     } else {
1350         return -1;
1351     }
1352 }
1353 
1354 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1355 {
1356     int r = 0;
1357 
1358     switch (ipa1) {
1359     case PRIV_B9_CLP:
1360         r = kvm_clp_service_call(cpu, run);
1361         break;
1362     case PRIV_B9_PCISTG:
1363         r = kvm_pcistg_service_call(cpu, run);
1364         break;
1365     case PRIV_B9_PCILG:
1366         r = kvm_pcilg_service_call(cpu, run);
1367         break;
1368     case PRIV_B9_RPCIT:
1369         r = kvm_rpcit_service_call(cpu, run);
1370         break;
1371     case PRIV_B9_EQBS:
1372         /* just inject exception */
1373         r = -1;
1374         break;
1375     default:
1376         r = -1;
1377         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
1378         break;
1379     }
1380 
1381     return r;
1382 }
1383 
1384 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1385 {
1386     int r = 0;
1387 
1388     switch (ipbl) {
1389     case PRIV_EB_PCISTB:
1390         r = kvm_pcistb_service_call(cpu, run);
1391         break;
1392     case PRIV_EB_SIC:
1393         r = kvm_sic_service_call(cpu, run);
1394         break;
1395     case PRIV_EB_SQBS:
1396         /* just inject exception */
1397         r = -1;
1398         break;
1399     default:
1400         r = -1;
1401         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1402         break;
1403     }
1404 
1405     return r;
1406 }
1407 
1408 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1409 {
1410     int r = 0;
1411 
1412     switch (ipbl) {
1413     case PRIV_E3_MPCIFC:
1414         r = kvm_mpcifc_service_call(cpu, run);
1415         break;
1416     case PRIV_E3_STPCIFC:
1417         r = kvm_stpcifc_service_call(cpu, run);
1418         break;
1419     default:
1420         r = -1;
1421         DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1422         break;
1423     }
1424 
1425     return r;
1426 }
1427 
1428 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1429 {
1430     CPUS390XState *env = &cpu->env;
1431     int ret;
1432 
1433     cpu_synchronize_state(CPU(cpu));
1434     ret = s390_virtio_hypercall(env);
1435     if (ret == -EINVAL) {
1436         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1437         return 0;
1438     }
1439 
1440     return ret;
1441 }
1442 
1443 static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
1444 {
1445     uint64_t r1, r3;
1446     int rc;
1447 
1448     cpu_synchronize_state(CPU(cpu));
1449     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1450     r3 = run->s390_sieic.ipa & 0x000f;
1451     rc = handle_diag_288(&cpu->env, r1, r3);
1452     if (rc) {
1453         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1454     }
1455 }
1456 
1457 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1458 {
1459     uint64_t r1, r3;
1460 
1461     cpu_synchronize_state(CPU(cpu));
1462     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1463     r3 = run->s390_sieic.ipa & 0x000f;
1464     handle_diag_308(&cpu->env, r1, r3, RA_IGNORED);
1465 }
1466 
1467 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1468 {
1469     CPUS390XState *env = &cpu->env;
1470     unsigned long pc;
1471 
1472     cpu_synchronize_state(CPU(cpu));
1473 
1474     pc = env->psw.addr - sw_bp_ilen;
1475     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1476         env->psw.addr = pc;
1477         return EXCP_DEBUG;
1478     }
1479 
1480     return -ENOENT;
1481 }
1482 
1483 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1484 
1485 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1486 {
1487     int r = 0;
1488     uint16_t func_code;
1489 
1490     /*
1491      * For any diagnose call we support, bits 48-63 of the resulting
1492      * address specify the function code; the remainder is ignored.
1493      */
1494     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
1495     switch (func_code) {
1496     case DIAG_TIMEREVENT:
1497         kvm_handle_diag_288(cpu, run);
1498         break;
1499     case DIAG_IPL:
1500         kvm_handle_diag_308(cpu, run);
1501         break;
1502     case DIAG_KVM_HYPERCALL:
1503         r = handle_hypercall(cpu, run);
1504         break;
1505     case DIAG_KVM_BREAKPOINT:
1506         r = handle_sw_breakpoint(cpu, run);
1507         break;
1508     default:
1509         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1510         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1511         break;
1512     }
1513 
1514     return r;
1515 }
1516 
1517 static int kvm_s390_handle_sigp(S390CPU *cpu, uint8_t ipa1, uint32_t ipb)
1518 {
1519     CPUS390XState *env = &cpu->env;
1520     const uint8_t r1 = ipa1 >> 4;
1521     const uint8_t r3 = ipa1 & 0x0f;
1522     int ret;
1523     uint8_t order;
1524 
1525     cpu_synchronize_state(CPU(cpu));
1526 
1527     /* get order code */
1528     order = decode_basedisp_rs(env, ipb, NULL) & SIGP_ORDER_MASK;
1529 
1530     ret = handle_sigp(env, order, r1, r3);
1531     setcc(cpu, ret);
1532     return 0;
1533 }
1534 
1535 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1536 {
1537     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1538     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1539     int r = -1;
1540 
1541     DPRINTF("handle_instruction 0x%x 0x%x\n",
1542             run->s390_sieic.ipa, run->s390_sieic.ipb);
1543     switch (ipa0) {
1544     case IPA0_B2:
1545         r = handle_b2(cpu, run, ipa1);
1546         break;
1547     case IPA0_B9:
1548         r = handle_b9(cpu, run, ipa1);
1549         break;
1550     case IPA0_EB:
1551         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1552         break;
1553     case IPA0_E3:
1554         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1555         break;
1556     case IPA0_DIAG:
1557         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1558         break;
1559     case IPA0_SIGP:
1560         r = kvm_s390_handle_sigp(cpu, ipa1, run->s390_sieic.ipb);
1561         break;
1562     }
1563 
1564     if (r < 0) {
1565         r = 0;
1566         kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1567     }
1568 
1569     return r;
1570 }
1571 
1572 static void unmanageable_intercept(S390CPU *cpu, const char *str, int pswoffset)
1573 {
1574     CPUState *cs = CPU(cpu);
1575 
1576     error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1577                  str, cs->cpu_index, ldq_phys(cs->as, cpu->env.psa + pswoffset),
1578                  ldq_phys(cs->as, cpu->env.psa + pswoffset + 8));
1579     s390_cpu_halt(cpu);
1580     qemu_system_guest_panicked(NULL);
1581 }
1582 
1583 /* try to detect pgm check loops */
1584 static int handle_oper_loop(S390CPU *cpu, struct kvm_run *run)
1585 {
1586     CPUState *cs = CPU(cpu);
1587     PSW oldpsw, newpsw;
1588 
1589     cpu_synchronize_state(cs);
1590     newpsw.mask = ldq_phys(cs->as, cpu->env.psa +
1591                            offsetof(LowCore, program_new_psw));
1592     newpsw.addr = ldq_phys(cs->as, cpu->env.psa +
1593                            offsetof(LowCore, program_new_psw) + 8);
1594     oldpsw.mask  = run->psw_mask;
1595     oldpsw.addr  = run->psw_addr;
1596     /*
1597      * Avoid endless loops of operation exceptions, if the pgm new
1598      * PSW will cause a new operation exception.
1599      * The heuristic checks if the pgm new psw is within 6 bytes before
1600      * the faulting psw address (with same DAT, AS settings) and the
1601      * new psw is not a wait psw and the fault was not triggered by
1602      * problem state. In that case go into crashed state.
1603      */
1604 
1605     if (oldpsw.addr - newpsw.addr <= 6 &&
1606         !(newpsw.mask & PSW_MASK_WAIT) &&
1607         !(oldpsw.mask & PSW_MASK_PSTATE) &&
1608         (newpsw.mask & PSW_MASK_ASC) == (oldpsw.mask & PSW_MASK_ASC) &&
1609         (newpsw.mask & PSW_MASK_DAT) == (oldpsw.mask & PSW_MASK_DAT)) {
1610         unmanageable_intercept(cpu, "operation exception loop",
1611                                offsetof(LowCore, program_new_psw));
1612         return EXCP_HALTED;
1613     }
1614     return 0;
1615 }
1616 
1617 static int handle_intercept(S390CPU *cpu)
1618 {
1619     CPUState *cs = CPU(cpu);
1620     struct kvm_run *run = cs->kvm_run;
1621     int icpt_code = run->s390_sieic.icptcode;
1622     int r = 0;
1623 
1624     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1625             (long)cs->kvm_run->psw_addr);
1626     switch (icpt_code) {
1627         case ICPT_INSTRUCTION:
1628             r = handle_instruction(cpu, run);
1629             break;
1630         case ICPT_PROGRAM:
1631             unmanageable_intercept(cpu, "program interrupt",
1632                                    offsetof(LowCore, program_new_psw));
1633             r = EXCP_HALTED;
1634             break;
1635         case ICPT_EXT_INT:
1636             unmanageable_intercept(cpu, "external interrupt",
1637                                    offsetof(LowCore, external_new_psw));
1638             r = EXCP_HALTED;
1639             break;
1640         case ICPT_WAITPSW:
1641             /* disabled wait, since enabled wait is handled in kernel */
1642             cpu_synchronize_state(cs);
1643             s390_handle_wait(cpu);
1644             r = EXCP_HALTED;
1645             break;
1646         case ICPT_CPU_STOP:
1647             do_stop_interrupt(&cpu->env);
1648             r = EXCP_HALTED;
1649             break;
1650         case ICPT_OPEREXC:
1651             /* check for break points */
1652             r = handle_sw_breakpoint(cpu, run);
1653             if (r == -ENOENT) {
1654                 /* Then check for potential pgm check loops */
1655                 r = handle_oper_loop(cpu, run);
1656                 if (r == 0) {
1657                     kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1658                 }
1659             }
1660             break;
1661         case ICPT_SOFT_INTERCEPT:
1662             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1663             exit(1);
1664             break;
1665         case ICPT_IO:
1666             fprintf(stderr, "KVM unimplemented icpt IO\n");
1667             exit(1);
1668             break;
1669         default:
1670             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1671             exit(1);
1672             break;
1673     }
1674 
1675     return r;
1676 }
1677 
1678 static int handle_tsch(S390CPU *cpu)
1679 {
1680     CPUState *cs = CPU(cpu);
1681     struct kvm_run *run = cs->kvm_run;
1682     int ret;
1683 
1684     cpu_synchronize_state(cs);
1685 
1686     ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb,
1687                              RA_IGNORED);
1688     if (ret < 0) {
1689         /*
1690          * Failure.
1691          * If an I/O interrupt had been dequeued, we have to reinject it.
1692          */
1693         if (run->s390_tsch.dequeued) {
1694             kvm_s390_io_interrupt(run->s390_tsch.subchannel_id,
1695                                   run->s390_tsch.subchannel_nr,
1696                                   run->s390_tsch.io_int_parm,
1697                                   run->s390_tsch.io_int_word);
1698         }
1699         ret = 0;
1700     }
1701     return ret;
1702 }
1703 
1704 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr, uint8_t ar)
1705 {
1706     struct sysib_322 sysib;
1707     int del;
1708 
1709     if (s390_cpu_virt_mem_read(cpu, addr, ar, &sysib, sizeof(sysib))) {
1710         return;
1711     }
1712     /* Shift the stack of Extended Names to prepare for our own data */
1713     memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1714             sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1715     /* First virt level, that doesn't provide Ext Names delimits stack. It is
1716      * assumed it's not capable of managing Extended Names for lower levels.
1717      */
1718     for (del = 1; del < sysib.count; del++) {
1719         if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1720             break;
1721         }
1722     }
1723     if (del < sysib.count) {
1724         memset(sysib.ext_names[del], 0,
1725                sizeof(sysib.ext_names[0]) * (sysib.count - del));
1726     }
1727     /* Insert short machine name in EBCDIC, padded with blanks */
1728     if (qemu_name) {
1729         memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1730         ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1731                                                     strlen(qemu_name)));
1732     }
1733     sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1734     memset(sysib.ext_names[0], 0, sizeof(sysib.ext_names[0]));
1735     /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1736      * considered by s390 as not capable of providing any Extended Name.
1737      * Therefore if no name was specified on qemu invocation, we go with the
1738      * same "KVMguest" default, which KVM has filled into short name field.
1739      */
1740     if (qemu_name) {
1741         strncpy((char *)sysib.ext_names[0], qemu_name,
1742                 sizeof(sysib.ext_names[0]));
1743     } else {
1744         strcpy((char *)sysib.ext_names[0], "KVMguest");
1745     }
1746     /* Insert UUID */
1747     memcpy(sysib.vm[0].uuid, &qemu_uuid, sizeof(sysib.vm[0].uuid));
1748 
1749     s390_cpu_virt_mem_write(cpu, addr, ar, &sysib, sizeof(sysib));
1750 }
1751 
1752 static int handle_stsi(S390CPU *cpu)
1753 {
1754     CPUState *cs = CPU(cpu);
1755     struct kvm_run *run = cs->kvm_run;
1756 
1757     switch (run->s390_stsi.fc) {
1758     case 3:
1759         if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1760             return 0;
1761         }
1762         /* Only sysib 3.2.2 needs post-handling for now. */
1763         insert_stsi_3_2_2(cpu, run->s390_stsi.addr, run->s390_stsi.ar);
1764         return 0;
1765     default:
1766         return 0;
1767     }
1768 }
1769 
1770 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1771 {
1772     CPUState *cs = CPU(cpu);
1773     struct kvm_run *run = cs->kvm_run;
1774 
1775     int ret = 0;
1776     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1777 
1778     switch (arch_info->type) {
1779     case KVM_HW_WP_WRITE:
1780         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1781             cs->watchpoint_hit = &hw_watchpoint;
1782             hw_watchpoint.vaddr = arch_info->addr;
1783             hw_watchpoint.flags = BP_MEM_WRITE;
1784             ret = EXCP_DEBUG;
1785         }
1786         break;
1787     case KVM_HW_BP:
1788         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1789             ret = EXCP_DEBUG;
1790         }
1791         break;
1792     case KVM_SINGLESTEP:
1793         if (cs->singlestep_enabled) {
1794             ret = EXCP_DEBUG;
1795         }
1796         break;
1797     default:
1798         ret = -ENOSYS;
1799     }
1800 
1801     return ret;
1802 }
1803 
1804 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1805 {
1806     S390CPU *cpu = S390_CPU(cs);
1807     int ret = 0;
1808 
1809     qemu_mutex_lock_iothread();
1810 
1811     switch (run->exit_reason) {
1812         case KVM_EXIT_S390_SIEIC:
1813             ret = handle_intercept(cpu);
1814             break;
1815         case KVM_EXIT_S390_RESET:
1816             s390_reipl_request();
1817             break;
1818         case KVM_EXIT_S390_TSCH:
1819             ret = handle_tsch(cpu);
1820             break;
1821         case KVM_EXIT_S390_STSI:
1822             ret = handle_stsi(cpu);
1823             break;
1824         case KVM_EXIT_DEBUG:
1825             ret = kvm_arch_handle_debug_exit(cpu);
1826             break;
1827         default:
1828             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1829             break;
1830     }
1831     qemu_mutex_unlock_iothread();
1832 
1833     if (ret == 0) {
1834         ret = EXCP_INTERRUPT;
1835     }
1836     return ret;
1837 }
1838 
1839 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1840 {
1841     return true;
1842 }
1843 
1844 void kvm_s390_io_interrupt(uint16_t subchannel_id,
1845                            uint16_t subchannel_nr, uint32_t io_int_parm,
1846                            uint32_t io_int_word)
1847 {
1848     struct kvm_s390_irq irq = {
1849         .u.io.subchannel_id = subchannel_id,
1850         .u.io.subchannel_nr = subchannel_nr,
1851         .u.io.io_int_parm = io_int_parm,
1852         .u.io.io_int_word = io_int_word,
1853     };
1854 
1855     if (io_int_word & IO_INT_WORD_AI) {
1856         irq.type = KVM_S390_INT_IO(1, 0, 0, 0);
1857     } else {
1858         irq.type = KVM_S390_INT_IO(0, (subchannel_id & 0xff00) >> 8,
1859                                       (subchannel_id & 0x0006),
1860                                       subchannel_nr);
1861     }
1862     kvm_s390_floating_interrupt(&irq);
1863 }
1864 
1865 void kvm_s390_crw_mchk(void)
1866 {
1867     struct kvm_s390_irq irq = {
1868         .type = KVM_S390_MCHK,
1869         .u.mchk.cr14 = CR14_CHANNEL_REPORT_SC,
1870         .u.mchk.mcic = s390_build_validity_mcic() | MCIC_SC_CP,
1871     };
1872     kvm_s390_floating_interrupt(&irq);
1873 }
1874 
1875 void kvm_s390_enable_css_support(S390CPU *cpu)
1876 {
1877     int r;
1878 
1879     /* Activate host kernel channel subsystem support. */
1880     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1881     assert(r == 0);
1882 }
1883 
1884 void kvm_arch_init_irq_routing(KVMState *s)
1885 {
1886     /*
1887      * Note that while irqchip capabilities generally imply that cpustates
1888      * are handled in-kernel, it is not true for s390 (yet); therefore, we
1889      * have to override the common code kvm_halt_in_kernel_allowed setting.
1890      */
1891     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1892         kvm_gsi_routing_allowed = true;
1893         kvm_halt_in_kernel_allowed = false;
1894     }
1895 }
1896 
1897 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1898                                     int vq, bool assign)
1899 {
1900     struct kvm_ioeventfd kick = {
1901         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1902         KVM_IOEVENTFD_FLAG_DATAMATCH,
1903         .fd = event_notifier_get_fd(notifier),
1904         .datamatch = vq,
1905         .addr = sch,
1906         .len = 8,
1907     };
1908     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
1909         return -ENOSYS;
1910     }
1911     if (!assign) {
1912         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
1913     }
1914     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
1915 }
1916 
1917 int kvm_s390_get_memslot_count(void)
1918 {
1919     return kvm_check_extension(kvm_state, KVM_CAP_NR_MEMSLOTS);
1920 }
1921 
1922 int kvm_s390_get_ri(void)
1923 {
1924     return cap_ri;
1925 }
1926 
1927 int kvm_s390_get_gs(void)
1928 {
1929     return cap_gs;
1930 }
1931 
1932 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
1933 {
1934     struct kvm_mp_state mp_state = {};
1935     int ret;
1936 
1937     /* the kvm part might not have been initialized yet */
1938     if (CPU(cpu)->kvm_state == NULL) {
1939         return 0;
1940     }
1941 
1942     switch (cpu_state) {
1943     case CPU_STATE_STOPPED:
1944         mp_state.mp_state = KVM_MP_STATE_STOPPED;
1945         break;
1946     case CPU_STATE_CHECK_STOP:
1947         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
1948         break;
1949     case CPU_STATE_OPERATING:
1950         mp_state.mp_state = KVM_MP_STATE_OPERATING;
1951         break;
1952     case CPU_STATE_LOAD:
1953         mp_state.mp_state = KVM_MP_STATE_LOAD;
1954         break;
1955     default:
1956         error_report("Requested CPU state is not a valid S390 CPU state: %u",
1957                      cpu_state);
1958         exit(1);
1959     }
1960 
1961     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
1962     if (ret) {
1963         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
1964                                        strerror(-ret));
1965     }
1966 
1967     return ret;
1968 }
1969 
1970 void kvm_s390_vcpu_interrupt_pre_save(S390CPU *cpu)
1971 {
1972     struct kvm_s390_irq_state irq_state = {
1973         .buf = (uint64_t) cpu->irqstate,
1974         .len = VCPU_IRQ_BUF_SIZE,
1975     };
1976     CPUState *cs = CPU(cpu);
1977     int32_t bytes;
1978 
1979     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
1980         return;
1981     }
1982 
1983     bytes = kvm_vcpu_ioctl(cs, KVM_S390_GET_IRQ_STATE, &irq_state);
1984     if (bytes < 0) {
1985         cpu->irqstate_saved_size = 0;
1986         error_report("Migration of interrupt state failed");
1987         return;
1988     }
1989 
1990     cpu->irqstate_saved_size = bytes;
1991 }
1992 
1993 int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
1994 {
1995     CPUState *cs = CPU(cpu);
1996     struct kvm_s390_irq_state irq_state = {
1997         .buf = (uint64_t) cpu->irqstate,
1998         .len = cpu->irqstate_saved_size,
1999     };
2000     int r;
2001 
2002     if (cpu->irqstate_saved_size == 0) {
2003         return 0;
2004     }
2005 
2006     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
2007         return -ENOSYS;
2008     }
2009 
2010     r = kvm_vcpu_ioctl(cs, KVM_S390_SET_IRQ_STATE, &irq_state);
2011     if (r) {
2012         error_report("Setting interrupt state failed %d", r);
2013     }
2014     return r;
2015 }
2016 
2017 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
2018                              uint64_t address, uint32_t data, PCIDevice *dev)
2019 {
2020     S390PCIBusDevice *pbdev;
2021     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
2022 
2023     if (!dev) {
2024         DPRINTF("add_msi_route no pci device\n");
2025         return -ENODEV;
2026     }
2027 
2028     pbdev = s390_pci_find_dev_by_target(s390_get_phb(), DEVICE(dev)->id);
2029     if (!pbdev) {
2030         DPRINTF("add_msi_route no zpci device\n");
2031         return -ENODEV;
2032     }
2033 
2034     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
2035     route->flags = 0;
2036     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
2037     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
2038     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
2039     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset + vec;
2040     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
2041     return 0;
2042 }
2043 
2044 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
2045                                 int vector, PCIDevice *dev)
2046 {
2047     return 0;
2048 }
2049 
2050 int kvm_arch_release_virq_post(int virq)
2051 {
2052     return 0;
2053 }
2054 
2055 int kvm_arch_msi_data_to_gsi(uint32_t data)
2056 {
2057     abort();
2058 }
2059 
2060 static int query_cpu_subfunc(S390FeatBitmap features)
2061 {
2062     struct kvm_s390_vm_cpu_subfunc prop;
2063     struct kvm_device_attr attr = {
2064         .group = KVM_S390_VM_CPU_MODEL,
2065         .attr = KVM_S390_VM_CPU_MACHINE_SUBFUNC,
2066         .addr = (uint64_t) &prop,
2067     };
2068     int rc;
2069 
2070     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2071     if (rc) {
2072         return  rc;
2073     }
2074 
2075     /*
2076      * We're going to add all subfunctions now, if the corresponding feature
2077      * is available that unlocks the query functions.
2078      */
2079     s390_add_from_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2080     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2081         s390_add_from_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2082     }
2083     if (test_bit(S390_FEAT_MSA, features)) {
2084         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2085         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2086         s390_add_from_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2087         s390_add_from_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2088         s390_add_from_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2089     }
2090     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2091         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2092     }
2093     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2094         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2095         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2096         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2097         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2098     }
2099     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2100         s390_add_from_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2101     }
2102     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2103         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2104     }
2105     return 0;
2106 }
2107 
2108 static int configure_cpu_subfunc(const S390FeatBitmap features)
2109 {
2110     struct kvm_s390_vm_cpu_subfunc prop = {};
2111     struct kvm_device_attr attr = {
2112         .group = KVM_S390_VM_CPU_MODEL,
2113         .attr = KVM_S390_VM_CPU_PROCESSOR_SUBFUNC,
2114         .addr = (uint64_t) &prop,
2115     };
2116 
2117     if (!kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2118                            KVM_S390_VM_CPU_PROCESSOR_SUBFUNC)) {
2119         /* hardware support might be missing, IBC will handle most of this */
2120         return 0;
2121     }
2122 
2123     s390_fill_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2124     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2125         s390_fill_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2126     }
2127     if (test_bit(S390_FEAT_MSA, features)) {
2128         s390_fill_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2129         s390_fill_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2130         s390_fill_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2131         s390_fill_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2132         s390_fill_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2133     }
2134     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2135         s390_fill_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2136     }
2137     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2138         s390_fill_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2139         s390_fill_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2140         s390_fill_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2141         s390_fill_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2142     }
2143     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2144         s390_fill_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2145     }
2146     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2147         s390_fill_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2148     }
2149     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2150 }
2151 
2152 static int kvm_to_feat[][2] = {
2153     { KVM_S390_VM_CPU_FEAT_ESOP, S390_FEAT_ESOP },
2154     { KVM_S390_VM_CPU_FEAT_SIEF2, S390_FEAT_SIE_F2 },
2155     { KVM_S390_VM_CPU_FEAT_64BSCAO , S390_FEAT_SIE_64BSCAO },
2156     { KVM_S390_VM_CPU_FEAT_SIIF, S390_FEAT_SIE_SIIF },
2157     { KVM_S390_VM_CPU_FEAT_GPERE, S390_FEAT_SIE_GPERE },
2158     { KVM_S390_VM_CPU_FEAT_GSLS, S390_FEAT_SIE_GSLS },
2159     { KVM_S390_VM_CPU_FEAT_IB, S390_FEAT_SIE_IB },
2160     { KVM_S390_VM_CPU_FEAT_CEI, S390_FEAT_SIE_CEI },
2161     { KVM_S390_VM_CPU_FEAT_IBS, S390_FEAT_SIE_IBS },
2162     { KVM_S390_VM_CPU_FEAT_SKEY, S390_FEAT_SIE_SKEY },
2163     { KVM_S390_VM_CPU_FEAT_CMMA, S390_FEAT_SIE_CMMA },
2164     { KVM_S390_VM_CPU_FEAT_PFMFI, S390_FEAT_SIE_PFMFI},
2165     { KVM_S390_VM_CPU_FEAT_SIGPIF, S390_FEAT_SIE_SIGPIF},
2166     { KVM_S390_VM_CPU_FEAT_KSS, S390_FEAT_SIE_KSS},
2167 };
2168 
2169 static int query_cpu_feat(S390FeatBitmap features)
2170 {
2171     struct kvm_s390_vm_cpu_feat prop;
2172     struct kvm_device_attr attr = {
2173         .group = KVM_S390_VM_CPU_MODEL,
2174         .attr = KVM_S390_VM_CPU_MACHINE_FEAT,
2175         .addr = (uint64_t) &prop,
2176     };
2177     int rc;
2178     int i;
2179 
2180     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2181     if (rc) {
2182         return  rc;
2183     }
2184 
2185     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2186         if (test_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat)) {
2187             set_bit(kvm_to_feat[i][1], features);
2188         }
2189     }
2190     return 0;
2191 }
2192 
2193 static int configure_cpu_feat(const S390FeatBitmap features)
2194 {
2195     struct kvm_s390_vm_cpu_feat prop = {};
2196     struct kvm_device_attr attr = {
2197         .group = KVM_S390_VM_CPU_MODEL,
2198         .attr = KVM_S390_VM_CPU_PROCESSOR_FEAT,
2199         .addr = (uint64_t) &prop,
2200     };
2201     int i;
2202 
2203     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2204         if (test_bit(kvm_to_feat[i][1], features)) {
2205             set_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat);
2206         }
2207     }
2208     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2209 }
2210 
2211 bool kvm_s390_cpu_models_supported(void)
2212 {
2213     if (!cpu_model_allowed()) {
2214         /* compatibility machines interfere with the cpu model */
2215         return false;
2216     }
2217     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2218                              KVM_S390_VM_CPU_MACHINE) &&
2219            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2220                              KVM_S390_VM_CPU_PROCESSOR) &&
2221            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2222                              KVM_S390_VM_CPU_MACHINE_FEAT) &&
2223            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2224                              KVM_S390_VM_CPU_PROCESSOR_FEAT) &&
2225            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2226                              KVM_S390_VM_CPU_MACHINE_SUBFUNC);
2227 }
2228 
2229 void kvm_s390_get_host_cpu_model(S390CPUModel *model, Error **errp)
2230 {
2231     struct kvm_s390_vm_cpu_machine prop = {};
2232     struct kvm_device_attr attr = {
2233         .group = KVM_S390_VM_CPU_MODEL,
2234         .attr = KVM_S390_VM_CPU_MACHINE,
2235         .addr = (uint64_t) &prop,
2236     };
2237     uint16_t unblocked_ibc = 0, cpu_type = 0;
2238     int rc;
2239 
2240     memset(model, 0, sizeof(*model));
2241 
2242     if (!kvm_s390_cpu_models_supported()) {
2243         error_setg(errp, "KVM doesn't support CPU models");
2244         return;
2245     }
2246 
2247     /* query the basic cpu model properties */
2248     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2249     if (rc) {
2250         error_setg(errp, "KVM: Error querying host CPU model: %d", rc);
2251         return;
2252     }
2253 
2254     cpu_type = cpuid_type(prop.cpuid);
2255     if (has_ibc(prop.ibc)) {
2256         model->lowest_ibc = lowest_ibc(prop.ibc);
2257         unblocked_ibc = unblocked_ibc(prop.ibc);
2258     }
2259     model->cpu_id = cpuid_id(prop.cpuid);
2260     model->cpu_id_format = cpuid_format(prop.cpuid);
2261     model->cpu_ver = 0xff;
2262 
2263     /* get supported cpu features indicated via STFL(E) */
2264     s390_add_from_feat_block(model->features, S390_FEAT_TYPE_STFL,
2265                              (uint8_t *) prop.fac_mask);
2266     /* dat-enhancement facility 2 has no bit but was introduced with stfle */
2267     if (test_bit(S390_FEAT_STFLE, model->features)) {
2268         set_bit(S390_FEAT_DAT_ENH_2, model->features);
2269     }
2270     /* get supported cpu features indicated e.g. via SCLP */
2271     rc = query_cpu_feat(model->features);
2272     if (rc) {
2273         error_setg(errp, "KVM: Error querying CPU features: %d", rc);
2274         return;
2275     }
2276     /* get supported cpu subfunctions indicated via query / test bit */
2277     rc = query_cpu_subfunc(model->features);
2278     if (rc) {
2279         error_setg(errp, "KVM: Error querying CPU subfunctions: %d", rc);
2280         return;
2281     }
2282 
2283     /* with cpu model support, CMM is only indicated if really available */
2284     if (kvm_s390_cmma_available()) {
2285         set_bit(S390_FEAT_CMM, model->features);
2286     } else {
2287         /* no cmm -> no cmm nt */
2288         clear_bit(S390_FEAT_CMM_NT, model->features);
2289     }
2290 
2291     /* bpb needs kernel support for migration, VSIE and reset */
2292     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_BPB)) {
2293         clear_bit(S390_FEAT_BPB, model->features);
2294     }
2295 
2296     /* We emulate a zPCI bus and AEN, therefore we don't need HW support */
2297     if (pci_available) {
2298         set_bit(S390_FEAT_ZPCI, model->features);
2299     }
2300     set_bit(S390_FEAT_ADAPTER_EVENT_NOTIFICATION, model->features);
2301 
2302     if (s390_known_cpu_type(cpu_type)) {
2303         /* we want the exact model, even if some features are missing */
2304         model->def = s390_find_cpu_def(cpu_type, ibc_gen(unblocked_ibc),
2305                                        ibc_ec_ga(unblocked_ibc), NULL);
2306     } else {
2307         /* model unknown, e.g. too new - search using features */
2308         model->def = s390_find_cpu_def(0, ibc_gen(unblocked_ibc),
2309                                        ibc_ec_ga(unblocked_ibc),
2310                                        model->features);
2311     }
2312     if (!model->def) {
2313         error_setg(errp, "KVM: host CPU model could not be identified");
2314         return;
2315     }
2316     /* strip of features that are not part of the maximum model */
2317     bitmap_and(model->features, model->features, model->def->full_feat,
2318                S390_FEAT_MAX);
2319 }
2320 
2321 void kvm_s390_apply_cpu_model(const S390CPUModel *model, Error **errp)
2322 {
2323     struct kvm_s390_vm_cpu_processor prop  = {
2324         .fac_list = { 0 },
2325     };
2326     struct kvm_device_attr attr = {
2327         .group = KVM_S390_VM_CPU_MODEL,
2328         .attr = KVM_S390_VM_CPU_PROCESSOR,
2329         .addr = (uint64_t) &prop,
2330     };
2331     int rc;
2332 
2333     if (!model) {
2334         /* compatibility handling if cpu models are disabled */
2335         if (kvm_s390_cmma_available()) {
2336             kvm_s390_enable_cmma();
2337         }
2338         return;
2339     }
2340     if (!kvm_s390_cpu_models_supported()) {
2341         error_setg(errp, "KVM doesn't support CPU models");
2342         return;
2343     }
2344     prop.cpuid = s390_cpuid_from_cpu_model(model);
2345     prop.ibc = s390_ibc_from_cpu_model(model);
2346     /* configure cpu features indicated via STFL(e) */
2347     s390_fill_feat_block(model->features, S390_FEAT_TYPE_STFL,
2348                          (uint8_t *) prop.fac_list);
2349     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2350     if (rc) {
2351         error_setg(errp, "KVM: Error configuring the CPU model: %d", rc);
2352         return;
2353     }
2354     /* configure cpu features indicated e.g. via SCLP */
2355     rc = configure_cpu_feat(model->features);
2356     if (rc) {
2357         error_setg(errp, "KVM: Error configuring CPU features: %d", rc);
2358         return;
2359     }
2360     /* configure cpu subfunctions indicated via query / test bit */
2361     rc = configure_cpu_subfunc(model->features);
2362     if (rc) {
2363         error_setg(errp, "KVM: Error configuring CPU subfunctions: %d", rc);
2364         return;
2365     }
2366     /* enable CMM via CMMA */
2367     if (test_bit(S390_FEAT_CMM, model->features)) {
2368         kvm_s390_enable_cmma();
2369     }
2370 }
2371 
2372 void kvm_s390_restart_interrupt(S390CPU *cpu)
2373 {
2374     struct kvm_s390_irq irq = {
2375         .type = KVM_S390_RESTART,
2376     };
2377 
2378     kvm_s390_vcpu_interrupt(cpu, &irq);
2379 }
2380 
2381 void kvm_s390_stop_interrupt(S390CPU *cpu)
2382 {
2383     struct kvm_s390_irq irq = {
2384         .type = KVM_S390_SIGP_STOP,
2385     };
2386 
2387     kvm_s390_vcpu_interrupt(cpu, &irq);
2388 }
2389