xref: /qemu/target/s390x/kvm/kvm.c (revision 25a3173a0cd318445f274d82543d9a8133b4c384)
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 "exec/gdbstub.h"
42 #include "trace.h"
43 #include "hw/s390x/s390-pci-inst.h"
44 #include "hw/s390x/s390-pci-bus.h"
45 #include "hw/s390x/ipl.h"
46 #include "hw/s390x/ebcdic.h"
47 #include "exec/memattrs.h"
48 #include "hw/s390x/s390-virtio-ccw.h"
49 #include "hw/s390x/s390-virtio-hcall.h"
50 
51 #ifndef DEBUG_KVM
52 #define DEBUG_KVM  0
53 #endif
54 
55 #define DPRINTF(fmt, ...) do {                \
56     if (DEBUG_KVM) {                          \
57         fprintf(stderr, fmt, ## __VA_ARGS__); \
58     }                                         \
59 } while (0)
60 
61 #define kvm_vm_check_mem_attr(s, attr) \
62     kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
63 
64 #define IPA0_DIAG                       0x8300
65 #define IPA0_SIGP                       0xae00
66 #define IPA0_B2                         0xb200
67 #define IPA0_B9                         0xb900
68 #define IPA0_EB                         0xeb00
69 #define IPA0_E3                         0xe300
70 
71 #define PRIV_B2_SCLP_CALL               0x20
72 #define PRIV_B2_CSCH                    0x30
73 #define PRIV_B2_HSCH                    0x31
74 #define PRIV_B2_MSCH                    0x32
75 #define PRIV_B2_SSCH                    0x33
76 #define PRIV_B2_STSCH                   0x34
77 #define PRIV_B2_TSCH                    0x35
78 #define PRIV_B2_TPI                     0x36
79 #define PRIV_B2_SAL                     0x37
80 #define PRIV_B2_RSCH                    0x38
81 #define PRIV_B2_STCRW                   0x39
82 #define PRIV_B2_STCPS                   0x3a
83 #define PRIV_B2_RCHP                    0x3b
84 #define PRIV_B2_SCHM                    0x3c
85 #define PRIV_B2_CHSC                    0x5f
86 #define PRIV_B2_SIGA                    0x74
87 #define PRIV_B2_XSCH                    0x76
88 
89 #define PRIV_EB_SQBS                    0x8a
90 #define PRIV_EB_PCISTB                  0xd0
91 #define PRIV_EB_SIC                     0xd1
92 
93 #define PRIV_B9_EQBS                    0x9c
94 #define PRIV_B9_CLP                     0xa0
95 #define PRIV_B9_PCISTG                  0xd0
96 #define PRIV_B9_PCILG                   0xd2
97 #define PRIV_B9_RPCIT                   0xd3
98 
99 #define PRIV_E3_MPCIFC                  0xd0
100 #define PRIV_E3_STPCIFC                 0xd4
101 
102 #define DIAG_TIMEREVENT                 0x288
103 #define DIAG_IPL                        0x308
104 #define DIAG_KVM_HYPERCALL              0x500
105 #define DIAG_KVM_BREAKPOINT             0x501
106 
107 #define ICPT_INSTRUCTION                0x04
108 #define ICPT_PROGRAM                    0x08
109 #define ICPT_EXT_INT                    0x14
110 #define ICPT_WAITPSW                    0x1c
111 #define ICPT_SOFT_INTERCEPT             0x24
112 #define ICPT_CPU_STOP                   0x28
113 #define ICPT_OPEREXC                    0x2c
114 #define ICPT_IO                         0x40
115 
116 #define NR_LOCAL_IRQS 32
117 /*
118  * Needs to be big enough to contain max_cpus emergency signals
119  * and in addition NR_LOCAL_IRQS interrupts
120  */
121 #define VCPU_IRQ_BUF_SIZE (sizeof(struct kvm_s390_irq) * \
122                            (max_cpus + NR_LOCAL_IRQS))
123 
124 static CPUWatchpoint hw_watchpoint;
125 /*
126  * We don't use a list because this structure is also used to transmit the
127  * hardware breakpoints to the kernel.
128  */
129 static struct kvm_hw_breakpoint *hw_breakpoints;
130 static int nb_hw_breakpoints;
131 
132 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
133     KVM_CAP_LAST_INFO
134 };
135 
136 static int cap_sync_regs;
137 static int cap_async_pf;
138 static int cap_mem_op;
139 static int cap_s390_irq;
140 static int cap_ri;
141 static int cap_gs;
142 
143 static int active_cmma;
144 
145 static void *legacy_s390_alloc(size_t size, uint64_t *align, bool shared);
146 
147 static int kvm_s390_query_mem_limit(uint64_t *memory_limit)
148 {
149     struct kvm_device_attr attr = {
150         .group = KVM_S390_VM_MEM_CTRL,
151         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
152         .addr = (uint64_t) memory_limit,
153     };
154 
155     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
156 }
157 
158 int kvm_s390_set_mem_limit(uint64_t new_limit, uint64_t *hw_limit)
159 {
160     int rc;
161 
162     struct kvm_device_attr attr = {
163         .group = KVM_S390_VM_MEM_CTRL,
164         .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
165         .addr = (uint64_t) &new_limit,
166     };
167 
168     if (!kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_LIMIT_SIZE)) {
169         return 0;
170     }
171 
172     rc = kvm_s390_query_mem_limit(hw_limit);
173     if (rc) {
174         return rc;
175     } else if (*hw_limit < new_limit) {
176         return -E2BIG;
177     }
178 
179     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
180 }
181 
182 int kvm_s390_cmma_active(void)
183 {
184     return active_cmma;
185 }
186 
187 static bool kvm_s390_cmma_available(void)
188 {
189     static bool initialized, value;
190 
191     if (!initialized) {
192         initialized = true;
193         value = kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_ENABLE_CMMA) &&
194                 kvm_vm_check_mem_attr(kvm_state, KVM_S390_VM_MEM_CLR_CMMA);
195     }
196     return value;
197 }
198 
199 void kvm_s390_cmma_reset(void)
200 {
201     int rc;
202     struct kvm_device_attr attr = {
203         .group = KVM_S390_VM_MEM_CTRL,
204         .attr = KVM_S390_VM_MEM_CLR_CMMA,
205     };
206 
207     if (!kvm_s390_cmma_active()) {
208         return;
209     }
210 
211     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
212     trace_kvm_clear_cmma(rc);
213 }
214 
215 static void kvm_s390_enable_cmma(void)
216 {
217     int rc;
218     struct kvm_device_attr attr = {
219         .group = KVM_S390_VM_MEM_CTRL,
220         .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
221     };
222 
223     if (mem_path) {
224         warn_report("CMM will not be enabled because it is not "
225                     "compatible with hugetlbfs.");
226         return;
227     }
228     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
229     active_cmma = !rc;
230     trace_kvm_enable_cmma(rc);
231 }
232 
233 static void kvm_s390_set_attr(uint64_t attr)
234 {
235     struct kvm_device_attr attribute = {
236         .group = KVM_S390_VM_CRYPTO,
237         .attr  = attr,
238     };
239 
240     int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
241 
242     if (ret) {
243         error_report("Failed to set crypto device attribute %lu: %s",
244                      attr, strerror(-ret));
245     }
246 }
247 
248 static void kvm_s390_init_aes_kw(void)
249 {
250     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
251 
252     if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
253                                  NULL)) {
254             attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
255     }
256 
257     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
258             kvm_s390_set_attr(attr);
259     }
260 }
261 
262 static void kvm_s390_init_dea_kw(void)
263 {
264     uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
265 
266     if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
267                                  NULL)) {
268             attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
269     }
270 
271     if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
272             kvm_s390_set_attr(attr);
273     }
274 }
275 
276 void kvm_s390_crypto_reset(void)
277 {
278     if (s390_has_feat(S390_FEAT_MSA_EXT_3)) {
279         kvm_s390_init_aes_kw();
280         kvm_s390_init_dea_kw();
281     }
282 }
283 
284 int kvm_arch_init(MachineState *ms, KVMState *s)
285 {
286     MachineClass *mc = MACHINE_GET_CLASS(ms);
287 
288     mc->default_cpu_type = S390_CPU_TYPE_NAME("host");
289     cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
290     cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
291     cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
292     cap_s390_irq = kvm_check_extension(s, KVM_CAP_S390_INJECT_IRQ);
293 
294     if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
295         || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
296         phys_mem_set_alloc(legacy_s390_alloc);
297     }
298 
299     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
300     kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
301     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
302     if (ri_allowed()) {
303         if (kvm_vm_enable_cap(s, KVM_CAP_S390_RI, 0) == 0) {
304             cap_ri = 1;
305         }
306     }
307     if (cpu_model_allowed()) {
308         if (kvm_vm_enable_cap(s, KVM_CAP_S390_GS, 0) == 0) {
309             cap_gs = 1;
310         }
311     }
312 
313     /*
314      * The migration interface for ais was introduced with kernel 4.13
315      * but the capability itself had been active since 4.12. As migration
316      * support is considered necessary let's disable ais in the 2.10
317      * machine.
318      */
319     /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
320 
321     return 0;
322 }
323 
324 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s)
325 {
326     return 0;
327 }
328 
329 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
330 {
331     return cpu->cpu_index;
332 }
333 
334 int kvm_arch_init_vcpu(CPUState *cs)
335 {
336     S390CPU *cpu = S390_CPU(cs);
337     kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
338     cpu->irqstate = g_malloc0(VCPU_IRQ_BUF_SIZE);
339     return 0;
340 }
341 
342 void kvm_s390_reset_vcpu(S390CPU *cpu)
343 {
344     CPUState *cs = CPU(cpu);
345 
346     /* The initial reset call is needed here to reset in-kernel
347      * vcpu data that we can't access directly from QEMU
348      * (i.e. with older kernels which don't support sync_regs/ONE_REG).
349      * Before this ioctl cpu_synchronize_state() is called in common kvm
350      * code (kvm-all) */
351     if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
352         error_report("Initial CPU reset failed on CPU %i", cs->cpu_index);
353     }
354 }
355 
356 static int can_sync_regs(CPUState *cs, int regs)
357 {
358     return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
359 }
360 
361 int kvm_arch_put_registers(CPUState *cs, int level)
362 {
363     S390CPU *cpu = S390_CPU(cs);
364     CPUS390XState *env = &cpu->env;
365     struct kvm_sregs sregs;
366     struct kvm_regs regs;
367     struct kvm_fpu fpu = {};
368     int r;
369     int i;
370 
371     /* always save the PSW  and the GPRS*/
372     cs->kvm_run->psw_addr = env->psw.addr;
373     cs->kvm_run->psw_mask = env->psw.mask;
374 
375     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
376         for (i = 0; i < 16; i++) {
377             cs->kvm_run->s.regs.gprs[i] = env->regs[i];
378             cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
379         }
380     } else {
381         for (i = 0; i < 16; i++) {
382             regs.gprs[i] = env->regs[i];
383         }
384         r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, &regs);
385         if (r < 0) {
386             return r;
387         }
388     }
389 
390     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
391         for (i = 0; i < 32; i++) {
392             cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0].ll;
393             cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1].ll;
394         }
395         cs->kvm_run->s.regs.fpc = env->fpc;
396         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
397     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
398         for (i = 0; i < 16; i++) {
399             cs->kvm_run->s.regs.fprs[i] = get_freg(env, i)->ll;
400         }
401         cs->kvm_run->s.regs.fpc = env->fpc;
402         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_FPRS;
403     } else {
404         /* Floating point */
405         for (i = 0; i < 16; i++) {
406             fpu.fprs[i] = get_freg(env, i)->ll;
407         }
408         fpu.fpc = env->fpc;
409 
410         r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
411         if (r < 0) {
412             return r;
413         }
414     }
415 
416     /* Do we need to save more than that? */
417     if (level == KVM_PUT_RUNTIME_STATE) {
418         return 0;
419     }
420 
421     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
422         cs->kvm_run->s.regs.cputm = env->cputm;
423         cs->kvm_run->s.regs.ckc = env->ckc;
424         cs->kvm_run->s.regs.todpr = env->todpr;
425         cs->kvm_run->s.regs.gbea = env->gbea;
426         cs->kvm_run->s.regs.pp = env->pp;
427         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
428     } else {
429         /*
430          * These ONE_REGS are not protected by a capability. As they are only
431          * necessary for migration we just trace a possible error, but don't
432          * return with an error return code.
433          */
434         kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
435         kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
436         kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
437         kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
438         kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
439     }
440 
441     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
442         memcpy(cs->kvm_run->s.regs.riccb, env->riccb, 64);
443         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_RICCB;
444     }
445 
446     /* pfault parameters */
447     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
448         cs->kvm_run->s.regs.pft = env->pfault_token;
449         cs->kvm_run->s.regs.pfs = env->pfault_select;
450         cs->kvm_run->s.regs.pfc = env->pfault_compare;
451         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
452     } else if (cap_async_pf) {
453         r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
454         if (r < 0) {
455             return r;
456         }
457         r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
458         if (r < 0) {
459             return r;
460         }
461         r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
462         if (r < 0) {
463             return r;
464         }
465     }
466 
467     /* access registers and control registers*/
468     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
469         for (i = 0; i < 16; i++) {
470             cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
471             cs->kvm_run->s.regs.crs[i] = env->cregs[i];
472         }
473         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
474         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
475     } else {
476         for (i = 0; i < 16; i++) {
477             sregs.acrs[i] = env->aregs[i];
478             sregs.crs[i] = env->cregs[i];
479         }
480         r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
481         if (r < 0) {
482             return r;
483         }
484     }
485 
486     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
487         memcpy(cs->kvm_run->s.regs.gscb, env->gscb, 32);
488         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GSCB;
489     }
490 
491     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
492         cs->kvm_run->s.regs.bpbc = env->bpbc;
493         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_BPBC;
494     }
495 
496     /* Finally the prefix */
497     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
498         cs->kvm_run->s.regs.prefix = env->psa;
499         cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
500     } else {
501         /* prefix is only supported via sync regs */
502     }
503     return 0;
504 }
505 
506 int kvm_arch_get_registers(CPUState *cs)
507 {
508     S390CPU *cpu = S390_CPU(cs);
509     CPUS390XState *env = &cpu->env;
510     struct kvm_sregs sregs;
511     struct kvm_regs regs;
512     struct kvm_fpu fpu;
513     int i, r;
514 
515     /* get the PSW */
516     env->psw.addr = cs->kvm_run->psw_addr;
517     env->psw.mask = cs->kvm_run->psw_mask;
518 
519     /* the GPRS */
520     if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
521         for (i = 0; i < 16; i++) {
522             env->regs[i] = cs->kvm_run->s.regs.gprs[i];
523         }
524     } else {
525         r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, &regs);
526         if (r < 0) {
527             return r;
528         }
529          for (i = 0; i < 16; i++) {
530             env->regs[i] = regs.gprs[i];
531         }
532     }
533 
534     /* The ACRS and CRS */
535     if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
536         for (i = 0; i < 16; i++) {
537             env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
538             env->cregs[i] = cs->kvm_run->s.regs.crs[i];
539         }
540     } else {
541         r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
542         if (r < 0) {
543             return r;
544         }
545          for (i = 0; i < 16; i++) {
546             env->aregs[i] = sregs.acrs[i];
547             env->cregs[i] = sregs.crs[i];
548         }
549     }
550 
551     /* Floating point and vector registers */
552     if (can_sync_regs(cs, KVM_SYNC_VRS)) {
553         for (i = 0; i < 32; i++) {
554             env->vregs[i][0].ll = cs->kvm_run->s.regs.vrs[i][0];
555             env->vregs[i][1].ll = cs->kvm_run->s.regs.vrs[i][1];
556         }
557         env->fpc = cs->kvm_run->s.regs.fpc;
558     } else if (can_sync_regs(cs, KVM_SYNC_FPRS)) {
559         for (i = 0; i < 16; i++) {
560             get_freg(env, i)->ll = cs->kvm_run->s.regs.fprs[i];
561         }
562         env->fpc = cs->kvm_run->s.regs.fpc;
563     } else {
564         r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
565         if (r < 0) {
566             return r;
567         }
568         for (i = 0; i < 16; i++) {
569             get_freg(env, i)->ll = fpu.fprs[i];
570         }
571         env->fpc = fpu.fpc;
572     }
573 
574     /* The prefix */
575     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
576         env->psa = cs->kvm_run->s.regs.prefix;
577     }
578 
579     if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
580         env->cputm = cs->kvm_run->s.regs.cputm;
581         env->ckc = cs->kvm_run->s.regs.ckc;
582         env->todpr = cs->kvm_run->s.regs.todpr;
583         env->gbea = cs->kvm_run->s.regs.gbea;
584         env->pp = cs->kvm_run->s.regs.pp;
585     } else {
586         /*
587          * These ONE_REGS are not protected by a capability. As they are only
588          * necessary for migration we just trace a possible error, but don't
589          * return with an error return code.
590          */
591         kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
592         kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
593         kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
594         kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
595         kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
596     }
597 
598     if (can_sync_regs(cs, KVM_SYNC_RICCB)) {
599         memcpy(env->riccb, cs->kvm_run->s.regs.riccb, 64);
600     }
601 
602     if (can_sync_regs(cs, KVM_SYNC_GSCB)) {
603         memcpy(env->gscb, cs->kvm_run->s.regs.gscb, 32);
604     }
605 
606     if (can_sync_regs(cs, KVM_SYNC_BPBC)) {
607         env->bpbc = cs->kvm_run->s.regs.bpbc;
608     }
609 
610     /* pfault parameters */
611     if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
612         env->pfault_token = cs->kvm_run->s.regs.pft;
613         env->pfault_select = cs->kvm_run->s.regs.pfs;
614         env->pfault_compare = cs->kvm_run->s.regs.pfc;
615     } else if (cap_async_pf) {
616         r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
617         if (r < 0) {
618             return r;
619         }
620         r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
621         if (r < 0) {
622             return r;
623         }
624         r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
625         if (r < 0) {
626             return r;
627         }
628     }
629 
630     return 0;
631 }
632 
633 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
634 {
635     int r;
636     struct kvm_device_attr attr = {
637         .group = KVM_S390_VM_TOD,
638         .attr = KVM_S390_VM_TOD_LOW,
639         .addr = (uint64_t)tod_low,
640     };
641 
642     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
643     if (r) {
644         return r;
645     }
646 
647     attr.attr = KVM_S390_VM_TOD_HIGH;
648     attr.addr = (uint64_t)tod_high;
649     return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
650 }
651 
652 int kvm_s390_get_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
653 {
654     int r;
655     struct kvm_s390_vm_tod_clock gtod;
656     struct kvm_device_attr attr = {
657         .group = KVM_S390_VM_TOD,
658         .attr = KVM_S390_VM_TOD_EXT,
659         .addr = (uint64_t)&gtod,
660     };
661 
662     r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
663     *tod_high = gtod.epoch_idx;
664     *tod_low  = gtod.tod;
665 
666     return r;
667 }
668 
669 int kvm_s390_set_clock(uint8_t *tod_high, uint64_t *tod_low)
670 {
671     int r;
672     struct kvm_device_attr attr = {
673         .group = KVM_S390_VM_TOD,
674         .attr = KVM_S390_VM_TOD_LOW,
675         .addr = (uint64_t)tod_low,
676     };
677 
678     r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
679     if (r) {
680         return r;
681     }
682 
683     attr.attr = KVM_S390_VM_TOD_HIGH;
684     attr.addr = (uint64_t)tod_high;
685     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
686 }
687 
688 int kvm_s390_set_clock_ext(uint8_t *tod_high, uint64_t *tod_low)
689 {
690     struct kvm_s390_vm_tod_clock gtod = {
691         .epoch_idx = *tod_high,
692         .tod  = *tod_low,
693     };
694     struct kvm_device_attr attr = {
695         .group = KVM_S390_VM_TOD,
696         .attr = KVM_S390_VM_TOD_EXT,
697         .addr = (uint64_t)&gtod,
698     };
699 
700     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
701 }
702 
703 /**
704  * kvm_s390_mem_op:
705  * @addr:      the logical start address in guest memory
706  * @ar:        the access register number
707  * @hostbuf:   buffer in host memory. NULL = do only checks w/o copying
708  * @len:       length that should be transferred
709  * @is_write:  true = write, false = read
710  * Returns:    0 on success, non-zero if an exception or error occurred
711  *
712  * Use KVM ioctl to read/write from/to guest memory. An access exception
713  * is injected into the vCPU in case of translation errors.
714  */
715 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, uint8_t ar, void *hostbuf,
716                     int len, bool is_write)
717 {
718     struct kvm_s390_mem_op mem_op = {
719         .gaddr = addr,
720         .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
721         .size = len,
722         .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
723                        : KVM_S390_MEMOP_LOGICAL_READ,
724         .buf = (uint64_t)hostbuf,
725         .ar = ar,
726     };
727     int ret;
728 
729     if (!cap_mem_op) {
730         return -ENOSYS;
731     }
732     if (!hostbuf) {
733         mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
734     }
735 
736     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
737     if (ret < 0) {
738         error_printf("KVM_S390_MEM_OP failed: %s\n", strerror(-ret));
739     }
740     return ret;
741 }
742 
743 /*
744  * Legacy layout for s390:
745  * Older S390 KVM requires the topmost vma of the RAM to be
746  * smaller than an system defined value, which is at least 256GB.
747  * Larger systems have larger values. We put the guest between
748  * the end of data segment (system break) and this value. We
749  * use 32GB as a base to have enough room for the system break
750  * to grow. We also have to use MAP parameters that avoid
751  * read-only mapping of guest pages.
752  */
753 static void *legacy_s390_alloc(size_t size, uint64_t *align, bool shared)
754 {
755     void *mem;
756 
757     mem = mmap((void *) 0x800000000ULL, size,
758                PROT_EXEC|PROT_READ|PROT_WRITE,
759                MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
760     return mem == MAP_FAILED ? NULL : mem;
761 }
762 
763 static uint8_t const *sw_bp_inst;
764 static uint8_t sw_bp_ilen;
765 
766 static void determine_sw_breakpoint_instr(void)
767 {
768         /* DIAG 501 is used for sw breakpoints with old kernels */
769         static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
770         /* Instruction 0x0000 is used for sw breakpoints with recent kernels */
771         static const uint8_t instr_0x0000[] = {0x00, 0x00};
772 
773         if (sw_bp_inst) {
774             return;
775         }
776         if (kvm_vm_enable_cap(kvm_state, KVM_CAP_S390_USER_INSTR0, 0)) {
777             sw_bp_inst = diag_501;
778             sw_bp_ilen = sizeof(diag_501);
779             DPRINTF("KVM: will use 4-byte sw breakpoints.\n");
780         } else {
781             sw_bp_inst = instr_0x0000;
782             sw_bp_ilen = sizeof(instr_0x0000);
783             DPRINTF("KVM: will use 2-byte sw breakpoints.\n");
784         }
785 }
786 
787 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
788 {
789     determine_sw_breakpoint_instr();
790 
791     if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
792                             sw_bp_ilen, 0) ||
793         cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)sw_bp_inst, sw_bp_ilen, 1)) {
794         return -EINVAL;
795     }
796     return 0;
797 }
798 
799 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
800 {
801     uint8_t t[MAX_ILEN];
802 
803     if (cpu_memory_rw_debug(cs, bp->pc, t, sw_bp_ilen, 0)) {
804         return -EINVAL;
805     } else if (memcmp(t, sw_bp_inst, sw_bp_ilen)) {
806         return -EINVAL;
807     } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
808                                    sw_bp_ilen, 1)) {
809         return -EINVAL;
810     }
811 
812     return 0;
813 }
814 
815 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
816                                                     int len, int type)
817 {
818     int n;
819 
820     for (n = 0; n < nb_hw_breakpoints; n++) {
821         if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
822             (hw_breakpoints[n].len == len || len == -1)) {
823             return &hw_breakpoints[n];
824         }
825     }
826 
827     return NULL;
828 }
829 
830 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
831 {
832     int size;
833 
834     if (find_hw_breakpoint(addr, len, type)) {
835         return -EEXIST;
836     }
837 
838     size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
839 
840     if (!hw_breakpoints) {
841         nb_hw_breakpoints = 0;
842         hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
843     } else {
844         hw_breakpoints =
845             (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
846     }
847 
848     if (!hw_breakpoints) {
849         nb_hw_breakpoints = 0;
850         return -ENOMEM;
851     }
852 
853     hw_breakpoints[nb_hw_breakpoints].addr = addr;
854     hw_breakpoints[nb_hw_breakpoints].len = len;
855     hw_breakpoints[nb_hw_breakpoints].type = type;
856 
857     nb_hw_breakpoints++;
858 
859     return 0;
860 }
861 
862 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
863                                   target_ulong len, int type)
864 {
865     switch (type) {
866     case GDB_BREAKPOINT_HW:
867         type = KVM_HW_BP;
868         break;
869     case GDB_WATCHPOINT_WRITE:
870         if (len < 1) {
871             return -EINVAL;
872         }
873         type = KVM_HW_WP_WRITE;
874         break;
875     default:
876         return -ENOSYS;
877     }
878     return insert_hw_breakpoint(addr, len, type);
879 }
880 
881 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
882                                   target_ulong len, int type)
883 {
884     int size;
885     struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
886 
887     if (bp == NULL) {
888         return -ENOENT;
889     }
890 
891     nb_hw_breakpoints--;
892     if (nb_hw_breakpoints > 0) {
893         /*
894          * In order to trim the array, move the last element to the position to
895          * be removed - if necessary.
896          */
897         if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
898             *bp = hw_breakpoints[nb_hw_breakpoints];
899         }
900         size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
901         hw_breakpoints =
902              (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
903     } else {
904         g_free(hw_breakpoints);
905         hw_breakpoints = NULL;
906     }
907 
908     return 0;
909 }
910 
911 void kvm_arch_remove_all_hw_breakpoints(void)
912 {
913     nb_hw_breakpoints = 0;
914     g_free(hw_breakpoints);
915     hw_breakpoints = NULL;
916 }
917 
918 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
919 {
920     int i;
921 
922     if (nb_hw_breakpoints > 0) {
923         dbg->arch.nr_hw_bp = nb_hw_breakpoints;
924         dbg->arch.hw_bp = hw_breakpoints;
925 
926         for (i = 0; i < nb_hw_breakpoints; ++i) {
927             hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
928                                                        hw_breakpoints[i].addr);
929         }
930         dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
931     } else {
932         dbg->arch.nr_hw_bp = 0;
933         dbg->arch.hw_bp = NULL;
934     }
935 }
936 
937 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
938 {
939 }
940 
941 MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
942 {
943     return MEMTXATTRS_UNSPECIFIED;
944 }
945 
946 int kvm_arch_process_async_events(CPUState *cs)
947 {
948     return cs->halted;
949 }
950 
951 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
952                                      struct kvm_s390_interrupt *interrupt)
953 {
954     int r = 0;
955 
956     interrupt->type = irq->type;
957     switch (irq->type) {
958     case KVM_S390_INT_VIRTIO:
959         interrupt->parm = irq->u.ext.ext_params;
960         /* fall through */
961     case KVM_S390_INT_PFAULT_INIT:
962     case KVM_S390_INT_PFAULT_DONE:
963         interrupt->parm64 = irq->u.ext.ext_params2;
964         break;
965     case KVM_S390_PROGRAM_INT:
966         interrupt->parm = irq->u.pgm.code;
967         break;
968     case KVM_S390_SIGP_SET_PREFIX:
969         interrupt->parm = irq->u.prefix.address;
970         break;
971     case KVM_S390_INT_SERVICE:
972         interrupt->parm = irq->u.ext.ext_params;
973         break;
974     case KVM_S390_MCHK:
975         interrupt->parm = irq->u.mchk.cr14;
976         interrupt->parm64 = irq->u.mchk.mcic;
977         break;
978     case KVM_S390_INT_EXTERNAL_CALL:
979         interrupt->parm = irq->u.extcall.code;
980         break;
981     case KVM_S390_INT_EMERGENCY:
982         interrupt->parm = irq->u.emerg.code;
983         break;
984     case KVM_S390_SIGP_STOP:
985     case KVM_S390_RESTART:
986         break; /* These types have no parameters */
987     case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
988         interrupt->parm = irq->u.io.subchannel_id << 16;
989         interrupt->parm |= irq->u.io.subchannel_nr;
990         interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
991         interrupt->parm64 |= irq->u.io.io_int_word;
992         break;
993     default:
994         r = -EINVAL;
995         break;
996     }
997     return r;
998 }
999 
1000 static void inject_vcpu_irq_legacy(CPUState *cs, struct kvm_s390_irq *irq)
1001 {
1002     struct kvm_s390_interrupt kvmint = {};
1003     int r;
1004 
1005     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1006     if (r < 0) {
1007         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1008         exit(1);
1009     }
1010 
1011     r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
1012     if (r < 0) {
1013         fprintf(stderr, "KVM failed to inject interrupt\n");
1014         exit(1);
1015     }
1016 }
1017 
1018 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
1019 {
1020     CPUState *cs = CPU(cpu);
1021     int r;
1022 
1023     if (cap_s390_irq) {
1024         r = kvm_vcpu_ioctl(cs, KVM_S390_IRQ, irq);
1025         if (!r) {
1026             return;
1027         }
1028         error_report("KVM failed to inject interrupt %llx", irq->type);
1029         exit(1);
1030     }
1031 
1032     inject_vcpu_irq_legacy(cs, irq);
1033 }
1034 
1035 void kvm_s390_floating_interrupt_legacy(struct kvm_s390_irq *irq)
1036 {
1037     struct kvm_s390_interrupt kvmint = {};
1038     int r;
1039 
1040     r = s390_kvm_irq_to_interrupt(irq, &kvmint);
1041     if (r < 0) {
1042         fprintf(stderr, "%s called with bogus interrupt\n", __func__);
1043         exit(1);
1044     }
1045 
1046     r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
1047     if (r < 0) {
1048         fprintf(stderr, "KVM failed to inject interrupt\n");
1049         exit(1);
1050     }
1051 }
1052 
1053 void kvm_s390_program_interrupt(S390CPU *cpu, uint16_t code)
1054 {
1055     struct kvm_s390_irq irq = {
1056         .type = KVM_S390_PROGRAM_INT,
1057         .u.pgm.code = code,
1058     };
1059 
1060     kvm_s390_vcpu_interrupt(cpu, &irq);
1061 }
1062 
1063 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
1064 {
1065     struct kvm_s390_irq irq = {
1066         .type = KVM_S390_PROGRAM_INT,
1067         .u.pgm.code = code,
1068         .u.pgm.trans_exc_code = te_code,
1069         .u.pgm.exc_access_id = te_code & 3,
1070     };
1071 
1072     kvm_s390_vcpu_interrupt(cpu, &irq);
1073 }
1074 
1075 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
1076                                  uint16_t ipbh0)
1077 {
1078     CPUS390XState *env = &cpu->env;
1079     uint64_t sccb;
1080     uint32_t code;
1081     int r = 0;
1082 
1083     sccb = env->regs[ipbh0 & 0xf];
1084     code = env->regs[(ipbh0 & 0xf0) >> 4];
1085 
1086     r = sclp_service_call(env, sccb, code);
1087     if (r < 0) {
1088         kvm_s390_program_interrupt(cpu, -r);
1089     } else {
1090         setcc(cpu, r);
1091     }
1092 
1093     return 0;
1094 }
1095 
1096 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1097 {
1098     CPUS390XState *env = &cpu->env;
1099     int rc = 0;
1100     uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
1101 
1102     switch (ipa1) {
1103     case PRIV_B2_XSCH:
1104         ioinst_handle_xsch(cpu, env->regs[1], RA_IGNORED);
1105         break;
1106     case PRIV_B2_CSCH:
1107         ioinst_handle_csch(cpu, env->regs[1], RA_IGNORED);
1108         break;
1109     case PRIV_B2_HSCH:
1110         ioinst_handle_hsch(cpu, env->regs[1], RA_IGNORED);
1111         break;
1112     case PRIV_B2_MSCH:
1113         ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1114         break;
1115     case PRIV_B2_SSCH:
1116         ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1117         break;
1118     case PRIV_B2_STCRW:
1119         ioinst_handle_stcrw(cpu, run->s390_sieic.ipb, RA_IGNORED);
1120         break;
1121     case PRIV_B2_STSCH:
1122         ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb, RA_IGNORED);
1123         break;
1124     case PRIV_B2_TSCH:
1125         /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1126         fprintf(stderr, "Spurious tsch intercept\n");
1127         break;
1128     case PRIV_B2_CHSC:
1129         ioinst_handle_chsc(cpu, run->s390_sieic.ipb, RA_IGNORED);
1130         break;
1131     case PRIV_B2_TPI:
1132         /* This should have been handled by kvm already. */
1133         fprintf(stderr, "Spurious tpi intercept\n");
1134         break;
1135     case PRIV_B2_SCHM:
1136         ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
1137                            run->s390_sieic.ipb, RA_IGNORED);
1138         break;
1139     case PRIV_B2_RSCH:
1140         ioinst_handle_rsch(cpu, env->regs[1], RA_IGNORED);
1141         break;
1142     case PRIV_B2_RCHP:
1143         ioinst_handle_rchp(cpu, env->regs[1], RA_IGNORED);
1144         break;
1145     case PRIV_B2_STCPS:
1146         /* We do not provide this instruction, it is suppressed. */
1147         break;
1148     case PRIV_B2_SAL:
1149         ioinst_handle_sal(cpu, env->regs[1], RA_IGNORED);
1150         break;
1151     case PRIV_B2_SIGA:
1152         /* Not provided, set CC = 3 for subchannel not operational */
1153         setcc(cpu, 3);
1154         break;
1155     case PRIV_B2_SCLP_CALL:
1156         rc = kvm_sclp_service_call(cpu, run, ipbh0);
1157         break;
1158     default:
1159         rc = -1;
1160         DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
1161         break;
1162     }
1163 
1164     return rc;
1165 }
1166 
1167 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run,
1168                                   uint8_t *ar)
1169 {
1170     CPUS390XState *env = &cpu->env;
1171     uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1172     uint32_t base2 = run->s390_sieic.ipb >> 28;
1173     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1174                      ((run->s390_sieic.ipb & 0xff00) << 4);
1175 
1176     if (disp2 & 0x80000) {
1177         disp2 += 0xfff00000;
1178     }
1179     if (ar) {
1180         *ar = base2;
1181     }
1182 
1183     return (base2 ? env->regs[base2] : 0) +
1184            (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1185 }
1186 
1187 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run,
1188                                   uint8_t *ar)
1189 {
1190     CPUS390XState *env = &cpu->env;
1191     uint32_t base2 = run->s390_sieic.ipb >> 28;
1192     uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1193                      ((run->s390_sieic.ipb & 0xff00) << 4);
1194 
1195     if (disp2 & 0x80000) {
1196         disp2 += 0xfff00000;
1197     }
1198     if (ar) {
1199         *ar = base2;
1200     }
1201 
1202     return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1203 }
1204 
1205 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1206 {
1207     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1208 
1209     if (s390_has_feat(S390_FEAT_ZPCI)) {
1210         return clp_service_call(cpu, r2, RA_IGNORED);
1211     } else {
1212         return -1;
1213     }
1214 }
1215 
1216 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1217 {
1218     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1219     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1220 
1221     if (s390_has_feat(S390_FEAT_ZPCI)) {
1222         return pcilg_service_call(cpu, r1, r2, RA_IGNORED);
1223     } else {
1224         return -1;
1225     }
1226 }
1227 
1228 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1229 {
1230     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1231     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1232 
1233     if (s390_has_feat(S390_FEAT_ZPCI)) {
1234         return pcistg_service_call(cpu, r1, r2, RA_IGNORED);
1235     } else {
1236         return -1;
1237     }
1238 }
1239 
1240 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1241 {
1242     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1243     uint64_t fiba;
1244     uint8_t ar;
1245 
1246     if (s390_has_feat(S390_FEAT_ZPCI)) {
1247         fiba = get_base_disp_rxy(cpu, run, &ar);
1248 
1249         return stpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1250     } else {
1251         return -1;
1252     }
1253 }
1254 
1255 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1256 {
1257     CPUS390XState *env = &cpu->env;
1258     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1259     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1260     uint8_t isc;
1261     uint16_t mode;
1262     int r;
1263 
1264     mode = env->regs[r1] & 0xffff;
1265     isc = (env->regs[r3] >> 27) & 0x7;
1266     r = css_do_sic(env, isc, mode);
1267     if (r) {
1268         kvm_s390_program_interrupt(cpu, -r);
1269     }
1270 
1271     return 0;
1272 }
1273 
1274 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1275 {
1276     uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1277     uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1278 
1279     if (s390_has_feat(S390_FEAT_ZPCI)) {
1280         return rpcit_service_call(cpu, r1, r2, RA_IGNORED);
1281     } else {
1282         return -1;
1283     }
1284 }
1285 
1286 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1287 {
1288     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1289     uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1290     uint64_t gaddr;
1291     uint8_t ar;
1292 
1293     if (s390_has_feat(S390_FEAT_ZPCI)) {
1294         gaddr = get_base_disp_rsy(cpu, run, &ar);
1295 
1296         return pcistb_service_call(cpu, r1, r3, gaddr, ar, RA_IGNORED);
1297     } else {
1298         return -1;
1299     }
1300 }
1301 
1302 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1303 {
1304     uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1305     uint64_t fiba;
1306     uint8_t ar;
1307 
1308     if (s390_has_feat(S390_FEAT_ZPCI)) {
1309         fiba = get_base_disp_rxy(cpu, run, &ar);
1310 
1311         return mpcifc_service_call(cpu, r1, fiba, ar, RA_IGNORED);
1312     } else {
1313         return -1;
1314     }
1315 }
1316 
1317 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1318 {
1319     int r = 0;
1320 
1321     switch (ipa1) {
1322     case PRIV_B9_CLP:
1323         r = kvm_clp_service_call(cpu, run);
1324         break;
1325     case PRIV_B9_PCISTG:
1326         r = kvm_pcistg_service_call(cpu, run);
1327         break;
1328     case PRIV_B9_PCILG:
1329         r = kvm_pcilg_service_call(cpu, run);
1330         break;
1331     case PRIV_B9_RPCIT:
1332         r = kvm_rpcit_service_call(cpu, run);
1333         break;
1334     case PRIV_B9_EQBS:
1335         /* just inject exception */
1336         r = -1;
1337         break;
1338     default:
1339         r = -1;
1340         DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
1341         break;
1342     }
1343 
1344     return r;
1345 }
1346 
1347 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1348 {
1349     int r = 0;
1350 
1351     switch (ipbl) {
1352     case PRIV_EB_PCISTB:
1353         r = kvm_pcistb_service_call(cpu, run);
1354         break;
1355     case PRIV_EB_SIC:
1356         r = kvm_sic_service_call(cpu, run);
1357         break;
1358     case PRIV_EB_SQBS:
1359         /* just inject exception */
1360         r = -1;
1361         break;
1362     default:
1363         r = -1;
1364         DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1365         break;
1366     }
1367 
1368     return r;
1369 }
1370 
1371 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1372 {
1373     int r = 0;
1374 
1375     switch (ipbl) {
1376     case PRIV_E3_MPCIFC:
1377         r = kvm_mpcifc_service_call(cpu, run);
1378         break;
1379     case PRIV_E3_STPCIFC:
1380         r = kvm_stpcifc_service_call(cpu, run);
1381         break;
1382     default:
1383         r = -1;
1384         DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1385         break;
1386     }
1387 
1388     return r;
1389 }
1390 
1391 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1392 {
1393     CPUS390XState *env = &cpu->env;
1394     int ret;
1395 
1396     ret = s390_virtio_hypercall(env);
1397     if (ret == -EINVAL) {
1398         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1399         return 0;
1400     }
1401 
1402     return ret;
1403 }
1404 
1405 static void kvm_handle_diag_288(S390CPU *cpu, struct kvm_run *run)
1406 {
1407     uint64_t r1, r3;
1408     int rc;
1409 
1410     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1411     r3 = run->s390_sieic.ipa & 0x000f;
1412     rc = handle_diag_288(&cpu->env, r1, r3);
1413     if (rc) {
1414         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1415     }
1416 }
1417 
1418 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1419 {
1420     uint64_t r1, r3;
1421 
1422     r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1423     r3 = run->s390_sieic.ipa & 0x000f;
1424     handle_diag_308(&cpu->env, r1, r3, RA_IGNORED);
1425 }
1426 
1427 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1428 {
1429     CPUS390XState *env = &cpu->env;
1430     unsigned long pc;
1431 
1432     pc = env->psw.addr - sw_bp_ilen;
1433     if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1434         env->psw.addr = pc;
1435         return EXCP_DEBUG;
1436     }
1437 
1438     return -ENOENT;
1439 }
1440 
1441 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1442 
1443 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1444 {
1445     int r = 0;
1446     uint16_t func_code;
1447 
1448     /*
1449      * For any diagnose call we support, bits 48-63 of the resulting
1450      * address specify the function code; the remainder is ignored.
1451      */
1452     func_code = decode_basedisp_rs(&cpu->env, ipb, NULL) & DIAG_KVM_CODE_MASK;
1453     switch (func_code) {
1454     case DIAG_TIMEREVENT:
1455         kvm_handle_diag_288(cpu, run);
1456         break;
1457     case DIAG_IPL:
1458         kvm_handle_diag_308(cpu, run);
1459         break;
1460     case DIAG_KVM_HYPERCALL:
1461         r = handle_hypercall(cpu, run);
1462         break;
1463     case DIAG_KVM_BREAKPOINT:
1464         r = handle_sw_breakpoint(cpu, run);
1465         break;
1466     default:
1467         DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1468         kvm_s390_program_interrupt(cpu, PGM_SPECIFICATION);
1469         break;
1470     }
1471 
1472     return r;
1473 }
1474 
1475 static int kvm_s390_handle_sigp(S390CPU *cpu, uint8_t ipa1, uint32_t ipb)
1476 {
1477     CPUS390XState *env = &cpu->env;
1478     const uint8_t r1 = ipa1 >> 4;
1479     const uint8_t r3 = ipa1 & 0x0f;
1480     int ret;
1481     uint8_t order;
1482 
1483     /* get order code */
1484     order = decode_basedisp_rs(env, ipb, NULL) & SIGP_ORDER_MASK;
1485 
1486     ret = handle_sigp(env, order, r1, r3);
1487     setcc(cpu, ret);
1488     return 0;
1489 }
1490 
1491 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1492 {
1493     unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1494     uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1495     int r = -1;
1496 
1497     DPRINTF("handle_instruction 0x%x 0x%x\n",
1498             run->s390_sieic.ipa, run->s390_sieic.ipb);
1499     switch (ipa0) {
1500     case IPA0_B2:
1501         r = handle_b2(cpu, run, ipa1);
1502         break;
1503     case IPA0_B9:
1504         r = handle_b9(cpu, run, ipa1);
1505         break;
1506     case IPA0_EB:
1507         r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1508         break;
1509     case IPA0_E3:
1510         r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1511         break;
1512     case IPA0_DIAG:
1513         r = handle_diag(cpu, run, run->s390_sieic.ipb);
1514         break;
1515     case IPA0_SIGP:
1516         r = kvm_s390_handle_sigp(cpu, ipa1, run->s390_sieic.ipb);
1517         break;
1518     }
1519 
1520     if (r < 0) {
1521         r = 0;
1522         kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1523     }
1524 
1525     return r;
1526 }
1527 
1528 static void unmanageable_intercept(S390CPU *cpu, S390CrashReason reason,
1529                                    int pswoffset)
1530 {
1531     CPUState *cs = CPU(cpu);
1532 
1533     s390_cpu_halt(cpu);
1534     cpu->env.crash_reason = reason;
1535     qemu_system_guest_panicked(cpu_get_crash_info(cs));
1536 }
1537 
1538 /* try to detect pgm check loops */
1539 static int handle_oper_loop(S390CPU *cpu, struct kvm_run *run)
1540 {
1541     CPUState *cs = CPU(cpu);
1542     PSW oldpsw, newpsw;
1543 
1544     newpsw.mask = ldq_phys(cs->as, cpu->env.psa +
1545                            offsetof(LowCore, program_new_psw));
1546     newpsw.addr = ldq_phys(cs->as, cpu->env.psa +
1547                            offsetof(LowCore, program_new_psw) + 8);
1548     oldpsw.mask  = run->psw_mask;
1549     oldpsw.addr  = run->psw_addr;
1550     /*
1551      * Avoid endless loops of operation exceptions, if the pgm new
1552      * PSW will cause a new operation exception.
1553      * The heuristic checks if the pgm new psw is within 6 bytes before
1554      * the faulting psw address (with same DAT, AS settings) and the
1555      * new psw is not a wait psw and the fault was not triggered by
1556      * problem state. In that case go into crashed state.
1557      */
1558 
1559     if (oldpsw.addr - newpsw.addr <= 6 &&
1560         !(newpsw.mask & PSW_MASK_WAIT) &&
1561         !(oldpsw.mask & PSW_MASK_PSTATE) &&
1562         (newpsw.mask & PSW_MASK_ASC) == (oldpsw.mask & PSW_MASK_ASC) &&
1563         (newpsw.mask & PSW_MASK_DAT) == (oldpsw.mask & PSW_MASK_DAT)) {
1564         unmanageable_intercept(cpu, S390_CRASH_REASON_OPINT_LOOP,
1565                                offsetof(LowCore, program_new_psw));
1566         return EXCP_HALTED;
1567     }
1568     return 0;
1569 }
1570 
1571 static int handle_intercept(S390CPU *cpu)
1572 {
1573     CPUState *cs = CPU(cpu);
1574     struct kvm_run *run = cs->kvm_run;
1575     int icpt_code = run->s390_sieic.icptcode;
1576     int r = 0;
1577 
1578     DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1579             (long)cs->kvm_run->psw_addr);
1580     switch (icpt_code) {
1581         case ICPT_INSTRUCTION:
1582             r = handle_instruction(cpu, run);
1583             break;
1584         case ICPT_PROGRAM:
1585             unmanageable_intercept(cpu, S390_CRASH_REASON_PGMINT_LOOP,
1586                                    offsetof(LowCore, program_new_psw));
1587             r = EXCP_HALTED;
1588             break;
1589         case ICPT_EXT_INT:
1590             unmanageable_intercept(cpu, S390_CRASH_REASON_EXTINT_LOOP,
1591                                    offsetof(LowCore, external_new_psw));
1592             r = EXCP_HALTED;
1593             break;
1594         case ICPT_WAITPSW:
1595             /* disabled wait, since enabled wait is handled in kernel */
1596             s390_handle_wait(cpu);
1597             r = EXCP_HALTED;
1598             break;
1599         case ICPT_CPU_STOP:
1600             do_stop_interrupt(&cpu->env);
1601             r = EXCP_HALTED;
1602             break;
1603         case ICPT_OPEREXC:
1604             /* check for break points */
1605             r = handle_sw_breakpoint(cpu, run);
1606             if (r == -ENOENT) {
1607                 /* Then check for potential pgm check loops */
1608                 r = handle_oper_loop(cpu, run);
1609                 if (r == 0) {
1610                     kvm_s390_program_interrupt(cpu, PGM_OPERATION);
1611                 }
1612             }
1613             break;
1614         case ICPT_SOFT_INTERCEPT:
1615             fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1616             exit(1);
1617             break;
1618         case ICPT_IO:
1619             fprintf(stderr, "KVM unimplemented icpt IO\n");
1620             exit(1);
1621             break;
1622         default:
1623             fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1624             exit(1);
1625             break;
1626     }
1627 
1628     return r;
1629 }
1630 
1631 static int handle_tsch(S390CPU *cpu)
1632 {
1633     CPUState *cs = CPU(cpu);
1634     struct kvm_run *run = cs->kvm_run;
1635     int ret;
1636 
1637     ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb,
1638                              RA_IGNORED);
1639     if (ret < 0) {
1640         /*
1641          * Failure.
1642          * If an I/O interrupt had been dequeued, we have to reinject it.
1643          */
1644         if (run->s390_tsch.dequeued) {
1645             s390_io_interrupt(run->s390_tsch.subchannel_id,
1646                               run->s390_tsch.subchannel_nr,
1647                               run->s390_tsch.io_int_parm,
1648                               run->s390_tsch.io_int_word);
1649         }
1650         ret = 0;
1651     }
1652     return ret;
1653 }
1654 
1655 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr, uint8_t ar)
1656 {
1657     SysIB_322 sysib;
1658     int del;
1659 
1660     if (s390_cpu_virt_mem_read(cpu, addr, ar, &sysib, sizeof(sysib))) {
1661         return;
1662     }
1663     /* Shift the stack of Extended Names to prepare for our own data */
1664     memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1665             sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1666     /* First virt level, that doesn't provide Ext Names delimits stack. It is
1667      * assumed it's not capable of managing Extended Names for lower levels.
1668      */
1669     for (del = 1; del < sysib.count; del++) {
1670         if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1671             break;
1672         }
1673     }
1674     if (del < sysib.count) {
1675         memset(sysib.ext_names[del], 0,
1676                sizeof(sysib.ext_names[0]) * (sysib.count - del));
1677     }
1678     /* Insert short machine name in EBCDIC, padded with blanks */
1679     if (qemu_name) {
1680         memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1681         ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1682                                                     strlen(qemu_name)));
1683     }
1684     sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1685     memset(sysib.ext_names[0], 0, sizeof(sysib.ext_names[0]));
1686     /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1687      * considered by s390 as not capable of providing any Extended Name.
1688      * Therefore if no name was specified on qemu invocation, we go with the
1689      * same "KVMguest" default, which KVM has filled into short name field.
1690      */
1691     if (qemu_name) {
1692         strncpy((char *)sysib.ext_names[0], qemu_name,
1693                 sizeof(sysib.ext_names[0]));
1694     } else {
1695         strcpy((char *)sysib.ext_names[0], "KVMguest");
1696     }
1697     /* Insert UUID */
1698     memcpy(sysib.vm[0].uuid, &qemu_uuid, sizeof(sysib.vm[0].uuid));
1699 
1700     s390_cpu_virt_mem_write(cpu, addr, ar, &sysib, sizeof(sysib));
1701 }
1702 
1703 static int handle_stsi(S390CPU *cpu)
1704 {
1705     CPUState *cs = CPU(cpu);
1706     struct kvm_run *run = cs->kvm_run;
1707 
1708     switch (run->s390_stsi.fc) {
1709     case 3:
1710         if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1711             return 0;
1712         }
1713         /* Only sysib 3.2.2 needs post-handling for now. */
1714         insert_stsi_3_2_2(cpu, run->s390_stsi.addr, run->s390_stsi.ar);
1715         return 0;
1716     default:
1717         return 0;
1718     }
1719 }
1720 
1721 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1722 {
1723     CPUState *cs = CPU(cpu);
1724     struct kvm_run *run = cs->kvm_run;
1725 
1726     int ret = 0;
1727     struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1728 
1729     switch (arch_info->type) {
1730     case KVM_HW_WP_WRITE:
1731         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1732             cs->watchpoint_hit = &hw_watchpoint;
1733             hw_watchpoint.vaddr = arch_info->addr;
1734             hw_watchpoint.flags = BP_MEM_WRITE;
1735             ret = EXCP_DEBUG;
1736         }
1737         break;
1738     case KVM_HW_BP:
1739         if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1740             ret = EXCP_DEBUG;
1741         }
1742         break;
1743     case KVM_SINGLESTEP:
1744         if (cs->singlestep_enabled) {
1745             ret = EXCP_DEBUG;
1746         }
1747         break;
1748     default:
1749         ret = -ENOSYS;
1750     }
1751 
1752     return ret;
1753 }
1754 
1755 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1756 {
1757     S390CPU *cpu = S390_CPU(cs);
1758     int ret = 0;
1759 
1760     qemu_mutex_lock_iothread();
1761 
1762     kvm_cpu_synchronize_state(cs);
1763 
1764     switch (run->exit_reason) {
1765         case KVM_EXIT_S390_SIEIC:
1766             ret = handle_intercept(cpu);
1767             break;
1768         case KVM_EXIT_S390_RESET:
1769             s390_ipl_reset_request(cs, S390_RESET_REIPL);
1770             break;
1771         case KVM_EXIT_S390_TSCH:
1772             ret = handle_tsch(cpu);
1773             break;
1774         case KVM_EXIT_S390_STSI:
1775             ret = handle_stsi(cpu);
1776             break;
1777         case KVM_EXIT_DEBUG:
1778             ret = kvm_arch_handle_debug_exit(cpu);
1779             break;
1780         default:
1781             fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1782             break;
1783     }
1784     qemu_mutex_unlock_iothread();
1785 
1786     if (ret == 0) {
1787         ret = EXCP_INTERRUPT;
1788     }
1789     return ret;
1790 }
1791 
1792 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1793 {
1794     return true;
1795 }
1796 
1797 void kvm_s390_enable_css_support(S390CPU *cpu)
1798 {
1799     int r;
1800 
1801     /* Activate host kernel channel subsystem support. */
1802     r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1803     assert(r == 0);
1804 }
1805 
1806 void kvm_arch_init_irq_routing(KVMState *s)
1807 {
1808     /*
1809      * Note that while irqchip capabilities generally imply that cpustates
1810      * are handled in-kernel, it is not true for s390 (yet); therefore, we
1811      * have to override the common code kvm_halt_in_kernel_allowed setting.
1812      */
1813     if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1814         kvm_gsi_routing_allowed = true;
1815         kvm_halt_in_kernel_allowed = false;
1816     }
1817 }
1818 
1819 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1820                                     int vq, bool assign)
1821 {
1822     struct kvm_ioeventfd kick = {
1823         .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1824         KVM_IOEVENTFD_FLAG_DATAMATCH,
1825         .fd = event_notifier_get_fd(notifier),
1826         .datamatch = vq,
1827         .addr = sch,
1828         .len = 8,
1829     };
1830     if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
1831         return -ENOSYS;
1832     }
1833     if (!assign) {
1834         kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
1835     }
1836     return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
1837 }
1838 
1839 int kvm_s390_get_ri(void)
1840 {
1841     return cap_ri;
1842 }
1843 
1844 int kvm_s390_get_gs(void)
1845 {
1846     return cap_gs;
1847 }
1848 
1849 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
1850 {
1851     struct kvm_mp_state mp_state = {};
1852     int ret;
1853 
1854     /* the kvm part might not have been initialized yet */
1855     if (CPU(cpu)->kvm_state == NULL) {
1856         return 0;
1857     }
1858 
1859     switch (cpu_state) {
1860     case S390_CPU_STATE_STOPPED:
1861         mp_state.mp_state = KVM_MP_STATE_STOPPED;
1862         break;
1863     case S390_CPU_STATE_CHECK_STOP:
1864         mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
1865         break;
1866     case S390_CPU_STATE_OPERATING:
1867         mp_state.mp_state = KVM_MP_STATE_OPERATING;
1868         break;
1869     case S390_CPU_STATE_LOAD:
1870         mp_state.mp_state = KVM_MP_STATE_LOAD;
1871         break;
1872     default:
1873         error_report("Requested CPU state is not a valid S390 CPU state: %u",
1874                      cpu_state);
1875         exit(1);
1876     }
1877 
1878     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
1879     if (ret) {
1880         trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
1881                                        strerror(-ret));
1882     }
1883 
1884     return ret;
1885 }
1886 
1887 void kvm_s390_vcpu_interrupt_pre_save(S390CPU *cpu)
1888 {
1889     struct kvm_s390_irq_state irq_state = {
1890         .buf = (uint64_t) cpu->irqstate,
1891         .len = VCPU_IRQ_BUF_SIZE,
1892     };
1893     CPUState *cs = CPU(cpu);
1894     int32_t bytes;
1895 
1896     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
1897         return;
1898     }
1899 
1900     bytes = kvm_vcpu_ioctl(cs, KVM_S390_GET_IRQ_STATE, &irq_state);
1901     if (bytes < 0) {
1902         cpu->irqstate_saved_size = 0;
1903         error_report("Migration of interrupt state failed");
1904         return;
1905     }
1906 
1907     cpu->irqstate_saved_size = bytes;
1908 }
1909 
1910 int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
1911 {
1912     CPUState *cs = CPU(cpu);
1913     struct kvm_s390_irq_state irq_state = {
1914         .buf = (uint64_t) cpu->irqstate,
1915         .len = cpu->irqstate_saved_size,
1916     };
1917     int r;
1918 
1919     if (cpu->irqstate_saved_size == 0) {
1920         return 0;
1921     }
1922 
1923     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
1924         return -ENOSYS;
1925     }
1926 
1927     r = kvm_vcpu_ioctl(cs, KVM_S390_SET_IRQ_STATE, &irq_state);
1928     if (r) {
1929         error_report("Setting interrupt state failed %d", r);
1930     }
1931     return r;
1932 }
1933 
1934 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
1935                              uint64_t address, uint32_t data, PCIDevice *dev)
1936 {
1937     S390PCIBusDevice *pbdev;
1938     uint32_t vec = data & ZPCI_MSI_VEC_MASK;
1939 
1940     if (!dev) {
1941         DPRINTF("add_msi_route no pci device\n");
1942         return -ENODEV;
1943     }
1944 
1945     pbdev = s390_pci_find_dev_by_target(s390_get_phb(), DEVICE(dev)->id);
1946     if (!pbdev) {
1947         DPRINTF("add_msi_route no zpci device\n");
1948         return -ENODEV;
1949     }
1950 
1951     route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
1952     route->flags = 0;
1953     route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
1954     route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
1955     route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
1956     route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset + vec;
1957     route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;
1958     return 0;
1959 }
1960 
1961 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
1962                                 int vector, PCIDevice *dev)
1963 {
1964     return 0;
1965 }
1966 
1967 int kvm_arch_release_virq_post(int virq)
1968 {
1969     return 0;
1970 }
1971 
1972 int kvm_arch_msi_data_to_gsi(uint32_t data)
1973 {
1974     abort();
1975 }
1976 
1977 static int query_cpu_subfunc(S390FeatBitmap features)
1978 {
1979     struct kvm_s390_vm_cpu_subfunc prop;
1980     struct kvm_device_attr attr = {
1981         .group = KVM_S390_VM_CPU_MODEL,
1982         .attr = KVM_S390_VM_CPU_MACHINE_SUBFUNC,
1983         .addr = (uint64_t) &prop,
1984     };
1985     int rc;
1986 
1987     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
1988     if (rc) {
1989         return  rc;
1990     }
1991 
1992     /*
1993      * We're going to add all subfunctions now, if the corresponding feature
1994      * is available that unlocks the query functions.
1995      */
1996     s390_add_from_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
1997     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
1998         s390_add_from_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
1999     }
2000     if (test_bit(S390_FEAT_MSA, features)) {
2001         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2002         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2003         s390_add_from_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2004         s390_add_from_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2005         s390_add_from_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2006     }
2007     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2008         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2009     }
2010     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2011         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2012         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2013         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2014         s390_add_from_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2015     }
2016     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2017         s390_add_from_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2018     }
2019     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2020         s390_add_from_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2021     }
2022     return 0;
2023 }
2024 
2025 static int configure_cpu_subfunc(const S390FeatBitmap features)
2026 {
2027     struct kvm_s390_vm_cpu_subfunc prop = {};
2028     struct kvm_device_attr attr = {
2029         .group = KVM_S390_VM_CPU_MODEL,
2030         .attr = KVM_S390_VM_CPU_PROCESSOR_SUBFUNC,
2031         .addr = (uint64_t) &prop,
2032     };
2033 
2034     if (!kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2035                            KVM_S390_VM_CPU_PROCESSOR_SUBFUNC)) {
2036         /* hardware support might be missing, IBC will handle most of this */
2037         return 0;
2038     }
2039 
2040     s390_fill_feat_block(features, S390_FEAT_TYPE_PLO, prop.plo);
2041     if (test_bit(S390_FEAT_TOD_CLOCK_STEERING, features)) {
2042         s390_fill_feat_block(features, S390_FEAT_TYPE_PTFF, prop.ptff);
2043     }
2044     if (test_bit(S390_FEAT_MSA, features)) {
2045         s390_fill_feat_block(features, S390_FEAT_TYPE_KMAC, prop.kmac);
2046         s390_fill_feat_block(features, S390_FEAT_TYPE_KMC, prop.kmc);
2047         s390_fill_feat_block(features, S390_FEAT_TYPE_KM, prop.km);
2048         s390_fill_feat_block(features, S390_FEAT_TYPE_KIMD, prop.kimd);
2049         s390_fill_feat_block(features, S390_FEAT_TYPE_KLMD, prop.klmd);
2050     }
2051     if (test_bit(S390_FEAT_MSA_EXT_3, features)) {
2052         s390_fill_feat_block(features, S390_FEAT_TYPE_PCKMO, prop.pckmo);
2053     }
2054     if (test_bit(S390_FEAT_MSA_EXT_4, features)) {
2055         s390_fill_feat_block(features, S390_FEAT_TYPE_KMCTR, prop.kmctr);
2056         s390_fill_feat_block(features, S390_FEAT_TYPE_KMF, prop.kmf);
2057         s390_fill_feat_block(features, S390_FEAT_TYPE_KMO, prop.kmo);
2058         s390_fill_feat_block(features, S390_FEAT_TYPE_PCC, prop.pcc);
2059     }
2060     if (test_bit(S390_FEAT_MSA_EXT_5, features)) {
2061         s390_fill_feat_block(features, S390_FEAT_TYPE_PPNO, prop.ppno);
2062     }
2063     if (test_bit(S390_FEAT_MSA_EXT_8, features)) {
2064         s390_fill_feat_block(features, S390_FEAT_TYPE_KMA, prop.kma);
2065     }
2066     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2067 }
2068 
2069 static int kvm_to_feat[][2] = {
2070     { KVM_S390_VM_CPU_FEAT_ESOP, S390_FEAT_ESOP },
2071     { KVM_S390_VM_CPU_FEAT_SIEF2, S390_FEAT_SIE_F2 },
2072     { KVM_S390_VM_CPU_FEAT_64BSCAO , S390_FEAT_SIE_64BSCAO },
2073     { KVM_S390_VM_CPU_FEAT_SIIF, S390_FEAT_SIE_SIIF },
2074     { KVM_S390_VM_CPU_FEAT_GPERE, S390_FEAT_SIE_GPERE },
2075     { KVM_S390_VM_CPU_FEAT_GSLS, S390_FEAT_SIE_GSLS },
2076     { KVM_S390_VM_CPU_FEAT_IB, S390_FEAT_SIE_IB },
2077     { KVM_S390_VM_CPU_FEAT_CEI, S390_FEAT_SIE_CEI },
2078     { KVM_S390_VM_CPU_FEAT_IBS, S390_FEAT_SIE_IBS },
2079     { KVM_S390_VM_CPU_FEAT_SKEY, S390_FEAT_SIE_SKEY },
2080     { KVM_S390_VM_CPU_FEAT_CMMA, S390_FEAT_SIE_CMMA },
2081     { KVM_S390_VM_CPU_FEAT_PFMFI, S390_FEAT_SIE_PFMFI},
2082     { KVM_S390_VM_CPU_FEAT_SIGPIF, S390_FEAT_SIE_SIGPIF},
2083     { KVM_S390_VM_CPU_FEAT_KSS, S390_FEAT_SIE_KSS},
2084 };
2085 
2086 static int query_cpu_feat(S390FeatBitmap features)
2087 {
2088     struct kvm_s390_vm_cpu_feat prop;
2089     struct kvm_device_attr attr = {
2090         .group = KVM_S390_VM_CPU_MODEL,
2091         .attr = KVM_S390_VM_CPU_MACHINE_FEAT,
2092         .addr = (uint64_t) &prop,
2093     };
2094     int rc;
2095     int i;
2096 
2097     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2098     if (rc) {
2099         return  rc;
2100     }
2101 
2102     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2103         if (test_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat)) {
2104             set_bit(kvm_to_feat[i][1], features);
2105         }
2106     }
2107     return 0;
2108 }
2109 
2110 static int configure_cpu_feat(const S390FeatBitmap features)
2111 {
2112     struct kvm_s390_vm_cpu_feat prop = {};
2113     struct kvm_device_attr attr = {
2114         .group = KVM_S390_VM_CPU_MODEL,
2115         .attr = KVM_S390_VM_CPU_PROCESSOR_FEAT,
2116         .addr = (uint64_t) &prop,
2117     };
2118     int i;
2119 
2120     for (i = 0; i < ARRAY_SIZE(kvm_to_feat); i++) {
2121         if (test_bit(kvm_to_feat[i][1], features)) {
2122             set_be_bit(kvm_to_feat[i][0], (uint8_t *) prop.feat);
2123         }
2124     }
2125     return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2126 }
2127 
2128 bool kvm_s390_cpu_models_supported(void)
2129 {
2130     if (!cpu_model_allowed()) {
2131         /* compatibility machines interfere with the cpu model */
2132         return false;
2133     }
2134     return kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2135                              KVM_S390_VM_CPU_MACHINE) &&
2136            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2137                              KVM_S390_VM_CPU_PROCESSOR) &&
2138            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2139                              KVM_S390_VM_CPU_MACHINE_FEAT) &&
2140            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2141                              KVM_S390_VM_CPU_PROCESSOR_FEAT) &&
2142            kvm_vm_check_attr(kvm_state, KVM_S390_VM_CPU_MODEL,
2143                              KVM_S390_VM_CPU_MACHINE_SUBFUNC);
2144 }
2145 
2146 void kvm_s390_get_host_cpu_model(S390CPUModel *model, Error **errp)
2147 {
2148     struct kvm_s390_vm_cpu_machine prop = {};
2149     struct kvm_device_attr attr = {
2150         .group = KVM_S390_VM_CPU_MODEL,
2151         .attr = KVM_S390_VM_CPU_MACHINE,
2152         .addr = (uint64_t) &prop,
2153     };
2154     uint16_t unblocked_ibc = 0, cpu_type = 0;
2155     int rc;
2156 
2157     memset(model, 0, sizeof(*model));
2158 
2159     if (!kvm_s390_cpu_models_supported()) {
2160         error_setg(errp, "KVM doesn't support CPU models");
2161         return;
2162     }
2163 
2164     /* query the basic cpu model properties */
2165     rc = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
2166     if (rc) {
2167         error_setg(errp, "KVM: Error querying host CPU model: %d", rc);
2168         return;
2169     }
2170 
2171     cpu_type = cpuid_type(prop.cpuid);
2172     if (has_ibc(prop.ibc)) {
2173         model->lowest_ibc = lowest_ibc(prop.ibc);
2174         unblocked_ibc = unblocked_ibc(prop.ibc);
2175     }
2176     model->cpu_id = cpuid_id(prop.cpuid);
2177     model->cpu_id_format = cpuid_format(prop.cpuid);
2178     model->cpu_ver = 0xff;
2179 
2180     /* get supported cpu features indicated via STFL(E) */
2181     s390_add_from_feat_block(model->features, S390_FEAT_TYPE_STFL,
2182                              (uint8_t *) prop.fac_mask);
2183     /* dat-enhancement facility 2 has no bit but was introduced with stfle */
2184     if (test_bit(S390_FEAT_STFLE, model->features)) {
2185         set_bit(S390_FEAT_DAT_ENH_2, model->features);
2186     }
2187     /* get supported cpu features indicated e.g. via SCLP */
2188     rc = query_cpu_feat(model->features);
2189     if (rc) {
2190         error_setg(errp, "KVM: Error querying CPU features: %d", rc);
2191         return;
2192     }
2193     /* get supported cpu subfunctions indicated via query / test bit */
2194     rc = query_cpu_subfunc(model->features);
2195     if (rc) {
2196         error_setg(errp, "KVM: Error querying CPU subfunctions: %d", rc);
2197         return;
2198     }
2199 
2200     /* PTFF subfunctions might be indicated although kernel support missing */
2201     if (!test_bit(S390_FEAT_MULTIPLE_EPOCH, model->features)) {
2202         clear_bit(S390_FEAT_PTFF_QSIE, model->features);
2203         clear_bit(S390_FEAT_PTFF_QTOUE, model->features);
2204         clear_bit(S390_FEAT_PTFF_STOE, model->features);
2205         clear_bit(S390_FEAT_PTFF_STOUE, model->features);
2206     }
2207 
2208     /* with cpu model support, CMM is only indicated if really available */
2209     if (kvm_s390_cmma_available()) {
2210         set_bit(S390_FEAT_CMM, model->features);
2211     } else {
2212         /* no cmm -> no cmm nt */
2213         clear_bit(S390_FEAT_CMM_NT, model->features);
2214     }
2215 
2216     /* bpb needs kernel support for migration, VSIE and reset */
2217     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_BPB)) {
2218         clear_bit(S390_FEAT_BPB, model->features);
2219     }
2220 
2221     /* We emulate a zPCI bus and AEN, therefore we don't need HW support */
2222     if (pci_available) {
2223         set_bit(S390_FEAT_ZPCI, model->features);
2224     }
2225     set_bit(S390_FEAT_ADAPTER_EVENT_NOTIFICATION, model->features);
2226 
2227     if (s390_known_cpu_type(cpu_type)) {
2228         /* we want the exact model, even if some features are missing */
2229         model->def = s390_find_cpu_def(cpu_type, ibc_gen(unblocked_ibc),
2230                                        ibc_ec_ga(unblocked_ibc), NULL);
2231     } else {
2232         /* model unknown, e.g. too new - search using features */
2233         model->def = s390_find_cpu_def(0, ibc_gen(unblocked_ibc),
2234                                        ibc_ec_ga(unblocked_ibc),
2235                                        model->features);
2236     }
2237     if (!model->def) {
2238         error_setg(errp, "KVM: host CPU model could not be identified");
2239         return;
2240     }
2241     /* strip of features that are not part of the maximum model */
2242     bitmap_and(model->features, model->features, model->def->full_feat,
2243                S390_FEAT_MAX);
2244 }
2245 
2246 void kvm_s390_apply_cpu_model(const S390CPUModel *model, Error **errp)
2247 {
2248     struct kvm_s390_vm_cpu_processor prop  = {
2249         .fac_list = { 0 },
2250     };
2251     struct kvm_device_attr attr = {
2252         .group = KVM_S390_VM_CPU_MODEL,
2253         .attr = KVM_S390_VM_CPU_PROCESSOR,
2254         .addr = (uint64_t) &prop,
2255     };
2256     int rc;
2257 
2258     if (!model) {
2259         /* compatibility handling if cpu models are disabled */
2260         if (kvm_s390_cmma_available()) {
2261             kvm_s390_enable_cmma();
2262         }
2263         return;
2264     }
2265     if (!kvm_s390_cpu_models_supported()) {
2266         error_setg(errp, "KVM doesn't support CPU models");
2267         return;
2268     }
2269     prop.cpuid = s390_cpuid_from_cpu_model(model);
2270     prop.ibc = s390_ibc_from_cpu_model(model);
2271     /* configure cpu features indicated via STFL(e) */
2272     s390_fill_feat_block(model->features, S390_FEAT_TYPE_STFL,
2273                          (uint8_t *) prop.fac_list);
2274     rc = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
2275     if (rc) {
2276         error_setg(errp, "KVM: Error configuring the CPU model: %d", rc);
2277         return;
2278     }
2279     /* configure cpu features indicated e.g. via SCLP */
2280     rc = configure_cpu_feat(model->features);
2281     if (rc) {
2282         error_setg(errp, "KVM: Error configuring CPU features: %d", rc);
2283         return;
2284     }
2285     /* configure cpu subfunctions indicated via query / test bit */
2286     rc = configure_cpu_subfunc(model->features);
2287     if (rc) {
2288         error_setg(errp, "KVM: Error configuring CPU subfunctions: %d", rc);
2289         return;
2290     }
2291     /* enable CMM via CMMA */
2292     if (test_bit(S390_FEAT_CMM, model->features)) {
2293         kvm_s390_enable_cmma();
2294     }
2295 }
2296 
2297 void kvm_s390_restart_interrupt(S390CPU *cpu)
2298 {
2299     struct kvm_s390_irq irq = {
2300         .type = KVM_S390_RESTART,
2301     };
2302 
2303     kvm_s390_vcpu_interrupt(cpu, &irq);
2304 }
2305 
2306 void kvm_s390_stop_interrupt(S390CPU *cpu)
2307 {
2308     struct kvm_s390_irq irq = {
2309         .type = KVM_S390_SIGP_STOP,
2310     };
2311 
2312     kvm_s390_vcpu_interrupt(cpu, &irq);
2313 }
2314