2 * QEMU S390x KVM implementation
4 * Copyright (c) 2009 Alexander Graf <agraf@suse.de>
5 * Copyright IBM Corp. 2012
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.
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.
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.
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/>.
24 #include "qemu/osdep.h"
25 #include <sys/ioctl.h>
27 #include <linux/kvm.h>
28 #include <asm/ptrace.h>
30 #include "qemu-common.h"
32 #include "qemu/error-report.h"
33 #include "qemu/timer.h"
34 #include "sysemu/sysemu.h"
35 #include "sysemu/kvm.h"
37 #include "sysemu/device_tree.h"
38 #include "qapi/qmp/qjson.h"
39 #include "exec/gdbstub.h"
40 #include "exec/address-spaces.h"
42 #include "qapi-event.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"
50 /* #define DEBUG_KVM */
53 #define DPRINTF(fmt, ...) \
54 do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
56 #define DPRINTF(fmt, ...) \
60 #define kvm_vm_check_mem_attr(s, attr) \
61 kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
63 #define IPA0_DIAG 0x8300
64 #define IPA0_SIGP 0xae00
65 #define IPA0_B2 0xb200
66 #define IPA0_B9 0xb900
67 #define IPA0_EB 0xeb00
68 #define IPA0_E3 0xe300
70 #define PRIV_B2_SCLP_CALL 0x20
71 #define PRIV_B2_CSCH 0x30
72 #define PRIV_B2_HSCH 0x31
73 #define PRIV_B2_MSCH 0x32
74 #define PRIV_B2_SSCH 0x33
75 #define PRIV_B2_STSCH 0x34
76 #define PRIV_B2_TSCH 0x35
77 #define PRIV_B2_TPI 0x36
78 #define PRIV_B2_SAL 0x37
79 #define PRIV_B2_RSCH 0x38
80 #define PRIV_B2_STCRW 0x39
81 #define PRIV_B2_STCPS 0x3a
82 #define PRIV_B2_RCHP 0x3b
83 #define PRIV_B2_SCHM 0x3c
84 #define PRIV_B2_CHSC 0x5f
85 #define PRIV_B2_SIGA 0x74
86 #define PRIV_B2_XSCH 0x76
88 #define PRIV_EB_SQBS 0x8a
89 #define PRIV_EB_PCISTB 0xd0
90 #define PRIV_EB_SIC 0xd1
92 #define PRIV_B9_EQBS 0x9c
93 #define PRIV_B9_CLP 0xa0
94 #define PRIV_B9_PCISTG 0xd0
95 #define PRIV_B9_PCILG 0xd2
96 #define PRIV_B9_RPCIT 0xd3
98 #define PRIV_E3_MPCIFC 0xd0
99 #define PRIV_E3_STPCIFC 0xd4
101 #define DIAG_TIMEREVENT 0x288
102 #define DIAG_IPL 0x308
103 #define DIAG_KVM_HYPERCALL 0x500
104 #define DIAG_KVM_BREAKPOINT 0x501
106 #define ICPT_INSTRUCTION 0x04
107 #define ICPT_PROGRAM 0x08
108 #define ICPT_EXT_INT 0x14
109 #define ICPT_WAITPSW 0x1c
110 #define ICPT_SOFT_INTERCEPT 0x24
111 #define ICPT_CPU_STOP 0x28
112 #define ICPT_OPEREXC 0x2c
115 #define NR_LOCAL_IRQS 32
117 * Needs to be big enough to contain max_cpus emergency signals
118 * and in addition NR_LOCAL_IRQS interrupts
120 #define VCPU_IRQ_BUF_SIZE (sizeof(struct kvm_s390_irq) * \
121 (max_cpus + NR_LOCAL_IRQS))
123 static CPUWatchpoint hw_watchpoint
;
125 * We don't use a list because this structure is also used to transmit the
126 * hardware breakpoints to the kernel.
128 static struct kvm_hw_breakpoint
*hw_breakpoints
;
129 static int nb_hw_breakpoints
;
131 const KVMCapabilityInfo kvm_arch_required_capabilities
[] = {
135 static int cap_sync_regs
;
136 static int cap_async_pf
;
137 static int cap_mem_op
;
138 static int cap_s390_irq
;
141 static void *legacy_s390_alloc(size_t size
, uint64_t *align
);
143 static int kvm_s390_query_mem_limit(KVMState
*s
, uint64_t *memory_limit
)
145 struct kvm_device_attr attr
= {
146 .group
= KVM_S390_VM_MEM_CTRL
,
147 .attr
= KVM_S390_VM_MEM_LIMIT_SIZE
,
148 .addr
= (uint64_t) memory_limit
,
151 return kvm_vm_ioctl(s
, KVM_GET_DEVICE_ATTR
, &attr
);
154 int kvm_s390_set_mem_limit(KVMState
*s
, uint64_t new_limit
, uint64_t *hw_limit
)
158 struct kvm_device_attr attr
= {
159 .group
= KVM_S390_VM_MEM_CTRL
,
160 .attr
= KVM_S390_VM_MEM_LIMIT_SIZE
,
161 .addr
= (uint64_t) &new_limit
,
164 if (!kvm_vm_check_mem_attr(s
, KVM_S390_VM_MEM_LIMIT_SIZE
)) {
168 rc
= kvm_s390_query_mem_limit(s
, hw_limit
);
171 } else if (*hw_limit
< new_limit
) {
175 return kvm_vm_ioctl(s
, KVM_SET_DEVICE_ATTR
, &attr
);
178 static bool kvm_s390_cmma_available(void)
180 static bool initialized
, value
;
184 value
= kvm_vm_check_mem_attr(kvm_state
, KVM_S390_VM_MEM_ENABLE_CMMA
) &&
185 kvm_vm_check_mem_attr(kvm_state
, KVM_S390_VM_MEM_CLR_CMMA
);
190 void kvm_s390_cmma_reset(void)
193 struct kvm_device_attr attr
= {
194 .group
= KVM_S390_VM_MEM_CTRL
,
195 .attr
= KVM_S390_VM_MEM_CLR_CMMA
,
198 if (!mem_path
|| !kvm_s390_cmma_available()) {
202 rc
= kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
203 trace_kvm_clear_cmma(rc
);
206 static void kvm_s390_enable_cmma(void)
209 struct kvm_device_attr attr
= {
210 .group
= KVM_S390_VM_MEM_CTRL
,
211 .attr
= KVM_S390_VM_MEM_ENABLE_CMMA
,
214 rc
= kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
215 trace_kvm_enable_cmma(rc
);
218 static void kvm_s390_set_attr(uint64_t attr
)
220 struct kvm_device_attr attribute
= {
221 .group
= KVM_S390_VM_CRYPTO
,
225 int ret
= kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attribute
);
228 error_report("Failed to set crypto device attribute %lu: %s",
229 attr
, strerror(-ret
));
233 static void kvm_s390_init_aes_kw(void)
235 uint64_t attr
= KVM_S390_VM_CRYPTO_DISABLE_AES_KW
;
237 if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
239 attr
= KVM_S390_VM_CRYPTO_ENABLE_AES_KW
;
242 if (kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CRYPTO
, attr
)) {
243 kvm_s390_set_attr(attr
);
247 static void kvm_s390_init_dea_kw(void)
249 uint64_t attr
= KVM_S390_VM_CRYPTO_DISABLE_DEA_KW
;
251 if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
253 attr
= KVM_S390_VM_CRYPTO_ENABLE_DEA_KW
;
256 if (kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CRYPTO
, attr
)) {
257 kvm_s390_set_attr(attr
);
261 void kvm_s390_crypto_reset(void)
263 if (s390_has_feat(S390_FEAT_MSA_EXT_3
)) {
264 kvm_s390_init_aes_kw();
265 kvm_s390_init_dea_kw();
269 int kvm_arch_init(MachineState
*ms
, KVMState
*s
)
271 cap_sync_regs
= kvm_check_extension(s
, KVM_CAP_SYNC_REGS
);
272 cap_async_pf
= kvm_check_extension(s
, KVM_CAP_ASYNC_PF
);
273 cap_mem_op
= kvm_check_extension(s
, KVM_CAP_S390_MEM_OP
);
274 cap_s390_irq
= kvm_check_extension(s
, KVM_CAP_S390_INJECT_IRQ
);
276 if (!kvm_check_extension(s
, KVM_CAP_S390_GMAP
)
277 || !kvm_check_extension(s
, KVM_CAP_S390_COW
)) {
278 phys_mem_set_alloc(legacy_s390_alloc
);
281 kvm_vm_enable_cap(s
, KVM_CAP_S390_USER_SIGP
, 0);
282 kvm_vm_enable_cap(s
, KVM_CAP_S390_VECTOR_REGISTERS
, 0);
283 kvm_vm_enable_cap(s
, KVM_CAP_S390_USER_STSI
, 0);
285 if (kvm_vm_enable_cap(s
, KVM_CAP_S390_RI
, 0) == 0) {
293 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
)
295 return cpu
->cpu_index
;
298 int kvm_arch_init_vcpu(CPUState
*cs
)
300 S390CPU
*cpu
= S390_CPU(cs
);
301 kvm_s390_set_cpu_state(cpu
, cpu
->env
.cpu_state
);
302 cpu
->irqstate
= g_malloc0(VCPU_IRQ_BUF_SIZE
);
306 void kvm_s390_reset_vcpu(S390CPU
*cpu
)
308 CPUState
*cs
= CPU(cpu
);
310 /* The initial reset call is needed here to reset in-kernel
311 * vcpu data that we can't access directly from QEMU
312 * (i.e. with older kernels which don't support sync_regs/ONE_REG).
313 * Before this ioctl cpu_synchronize_state() is called in common kvm
315 if (kvm_vcpu_ioctl(cs
, KVM_S390_INITIAL_RESET
, NULL
)) {
316 error_report("Initial CPU reset failed on CPU %i", cs
->cpu_index
);
320 static int can_sync_regs(CPUState
*cs
, int regs
)
322 return cap_sync_regs
&& (cs
->kvm_run
->kvm_valid_regs
& regs
) == regs
;
325 int kvm_arch_put_registers(CPUState
*cs
, int level
)
327 S390CPU
*cpu
= S390_CPU(cs
);
328 CPUS390XState
*env
= &cpu
->env
;
329 struct kvm_sregs sregs
;
330 struct kvm_regs regs
;
331 struct kvm_fpu fpu
= {};
335 /* always save the PSW and the GPRS*/
336 cs
->kvm_run
->psw_addr
= env
->psw
.addr
;
337 cs
->kvm_run
->psw_mask
= env
->psw
.mask
;
339 if (can_sync_regs(cs
, KVM_SYNC_GPRS
)) {
340 for (i
= 0; i
< 16; i
++) {
341 cs
->kvm_run
->s
.regs
.gprs
[i
] = env
->regs
[i
];
342 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_GPRS
;
345 for (i
= 0; i
< 16; i
++) {
346 regs
.gprs
[i
] = env
->regs
[i
];
348 r
= kvm_vcpu_ioctl(cs
, KVM_SET_REGS
, ®s
);
354 if (can_sync_regs(cs
, KVM_SYNC_VRS
)) {
355 for (i
= 0; i
< 32; i
++) {
356 cs
->kvm_run
->s
.regs
.vrs
[i
][0] = env
->vregs
[i
][0].ll
;
357 cs
->kvm_run
->s
.regs
.vrs
[i
][1] = env
->vregs
[i
][1].ll
;
359 cs
->kvm_run
->s
.regs
.fpc
= env
->fpc
;
360 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_VRS
;
361 } else if (can_sync_regs(cs
, KVM_SYNC_FPRS
)) {
362 for (i
= 0; i
< 16; i
++) {
363 cs
->kvm_run
->s
.regs
.fprs
[i
] = get_freg(env
, i
)->ll
;
365 cs
->kvm_run
->s
.regs
.fpc
= env
->fpc
;
366 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_FPRS
;
369 for (i
= 0; i
< 16; i
++) {
370 fpu
.fprs
[i
] = get_freg(env
, i
)->ll
;
374 r
= kvm_vcpu_ioctl(cs
, KVM_SET_FPU
, &fpu
);
380 /* Do we need to save more than that? */
381 if (level
== KVM_PUT_RUNTIME_STATE
) {
385 if (can_sync_regs(cs
, KVM_SYNC_ARCH0
)) {
386 cs
->kvm_run
->s
.regs
.cputm
= env
->cputm
;
387 cs
->kvm_run
->s
.regs
.ckc
= env
->ckc
;
388 cs
->kvm_run
->s
.regs
.todpr
= env
->todpr
;
389 cs
->kvm_run
->s
.regs
.gbea
= env
->gbea
;
390 cs
->kvm_run
->s
.regs
.pp
= env
->pp
;
391 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_ARCH0
;
394 * These ONE_REGS are not protected by a capability. As they are only
395 * necessary for migration we just trace a possible error, but don't
396 * return with an error return code.
398 kvm_set_one_reg(cs
, KVM_REG_S390_CPU_TIMER
, &env
->cputm
);
399 kvm_set_one_reg(cs
, KVM_REG_S390_CLOCK_COMP
, &env
->ckc
);
400 kvm_set_one_reg(cs
, KVM_REG_S390_TODPR
, &env
->todpr
);
401 kvm_set_one_reg(cs
, KVM_REG_S390_GBEA
, &env
->gbea
);
402 kvm_set_one_reg(cs
, KVM_REG_S390_PP
, &env
->pp
);
405 if (can_sync_regs(cs
, KVM_SYNC_RICCB
)) {
406 memcpy(cs
->kvm_run
->s
.regs
.riccb
, env
->riccb
, 64);
407 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_RICCB
;
410 /* pfault parameters */
411 if (can_sync_regs(cs
, KVM_SYNC_PFAULT
)) {
412 cs
->kvm_run
->s
.regs
.pft
= env
->pfault_token
;
413 cs
->kvm_run
->s
.regs
.pfs
= env
->pfault_select
;
414 cs
->kvm_run
->s
.regs
.pfc
= env
->pfault_compare
;
415 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_PFAULT
;
416 } else if (cap_async_pf
) {
417 r
= kvm_set_one_reg(cs
, KVM_REG_S390_PFTOKEN
, &env
->pfault_token
);
421 r
= kvm_set_one_reg(cs
, KVM_REG_S390_PFCOMPARE
, &env
->pfault_compare
);
425 r
= kvm_set_one_reg(cs
, KVM_REG_S390_PFSELECT
, &env
->pfault_select
);
431 /* access registers and control registers*/
432 if (can_sync_regs(cs
, KVM_SYNC_ACRS
| KVM_SYNC_CRS
)) {
433 for (i
= 0; i
< 16; i
++) {
434 cs
->kvm_run
->s
.regs
.acrs
[i
] = env
->aregs
[i
];
435 cs
->kvm_run
->s
.regs
.crs
[i
] = env
->cregs
[i
];
437 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_ACRS
;
438 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_CRS
;
440 for (i
= 0; i
< 16; i
++) {
441 sregs
.acrs
[i
] = env
->aregs
[i
];
442 sregs
.crs
[i
] = env
->cregs
[i
];
444 r
= kvm_vcpu_ioctl(cs
, KVM_SET_SREGS
, &sregs
);
450 /* Finally the prefix */
451 if (can_sync_regs(cs
, KVM_SYNC_PREFIX
)) {
452 cs
->kvm_run
->s
.regs
.prefix
= env
->psa
;
453 cs
->kvm_run
->kvm_dirty_regs
|= KVM_SYNC_PREFIX
;
455 /* prefix is only supported via sync regs */
460 int kvm_arch_get_registers(CPUState
*cs
)
462 S390CPU
*cpu
= S390_CPU(cs
);
463 CPUS390XState
*env
= &cpu
->env
;
464 struct kvm_sregs sregs
;
465 struct kvm_regs regs
;
470 env
->psw
.addr
= cs
->kvm_run
->psw_addr
;
471 env
->psw
.mask
= cs
->kvm_run
->psw_mask
;
474 if (can_sync_regs(cs
, KVM_SYNC_GPRS
)) {
475 for (i
= 0; i
< 16; i
++) {
476 env
->regs
[i
] = cs
->kvm_run
->s
.regs
.gprs
[i
];
479 r
= kvm_vcpu_ioctl(cs
, KVM_GET_REGS
, ®s
);
483 for (i
= 0; i
< 16; i
++) {
484 env
->regs
[i
] = regs
.gprs
[i
];
488 /* The ACRS and CRS */
489 if (can_sync_regs(cs
, KVM_SYNC_ACRS
| KVM_SYNC_CRS
)) {
490 for (i
= 0; i
< 16; i
++) {
491 env
->aregs
[i
] = cs
->kvm_run
->s
.regs
.acrs
[i
];
492 env
->cregs
[i
] = cs
->kvm_run
->s
.regs
.crs
[i
];
495 r
= kvm_vcpu_ioctl(cs
, KVM_GET_SREGS
, &sregs
);
499 for (i
= 0; i
< 16; i
++) {
500 env
->aregs
[i
] = sregs
.acrs
[i
];
501 env
->cregs
[i
] = sregs
.crs
[i
];
505 /* Floating point and vector registers */
506 if (can_sync_regs(cs
, KVM_SYNC_VRS
)) {
507 for (i
= 0; i
< 32; i
++) {
508 env
->vregs
[i
][0].ll
= cs
->kvm_run
->s
.regs
.vrs
[i
][0];
509 env
->vregs
[i
][1].ll
= cs
->kvm_run
->s
.regs
.vrs
[i
][1];
511 env
->fpc
= cs
->kvm_run
->s
.regs
.fpc
;
512 } else if (can_sync_regs(cs
, KVM_SYNC_FPRS
)) {
513 for (i
= 0; i
< 16; i
++) {
514 get_freg(env
, i
)->ll
= cs
->kvm_run
->s
.regs
.fprs
[i
];
516 env
->fpc
= cs
->kvm_run
->s
.regs
.fpc
;
518 r
= kvm_vcpu_ioctl(cs
, KVM_GET_FPU
, &fpu
);
522 for (i
= 0; i
< 16; i
++) {
523 get_freg(env
, i
)->ll
= fpu
.fprs
[i
];
529 if (can_sync_regs(cs
, KVM_SYNC_PREFIX
)) {
530 env
->psa
= cs
->kvm_run
->s
.regs
.prefix
;
533 if (can_sync_regs(cs
, KVM_SYNC_ARCH0
)) {
534 env
->cputm
= cs
->kvm_run
->s
.regs
.cputm
;
535 env
->ckc
= cs
->kvm_run
->s
.regs
.ckc
;
536 env
->todpr
= cs
->kvm_run
->s
.regs
.todpr
;
537 env
->gbea
= cs
->kvm_run
->s
.regs
.gbea
;
538 env
->pp
= cs
->kvm_run
->s
.regs
.pp
;
541 * These ONE_REGS are not protected by a capability. As they are only
542 * necessary for migration we just trace a possible error, but don't
543 * return with an error return code.
545 kvm_get_one_reg(cs
, KVM_REG_S390_CPU_TIMER
, &env
->cputm
);
546 kvm_get_one_reg(cs
, KVM_REG_S390_CLOCK_COMP
, &env
->ckc
);
547 kvm_get_one_reg(cs
, KVM_REG_S390_TODPR
, &env
->todpr
);
548 kvm_get_one_reg(cs
, KVM_REG_S390_GBEA
, &env
->gbea
);
549 kvm_get_one_reg(cs
, KVM_REG_S390_PP
, &env
->pp
);
552 if (can_sync_regs(cs
, KVM_SYNC_RICCB
)) {
553 memcpy(env
->riccb
, cs
->kvm_run
->s
.regs
.riccb
, 64);
556 /* pfault parameters */
557 if (can_sync_regs(cs
, KVM_SYNC_PFAULT
)) {
558 env
->pfault_token
= cs
->kvm_run
->s
.regs
.pft
;
559 env
->pfault_select
= cs
->kvm_run
->s
.regs
.pfs
;
560 env
->pfault_compare
= cs
->kvm_run
->s
.regs
.pfc
;
561 } else if (cap_async_pf
) {
562 r
= kvm_get_one_reg(cs
, KVM_REG_S390_PFTOKEN
, &env
->pfault_token
);
566 r
= kvm_get_one_reg(cs
, KVM_REG_S390_PFCOMPARE
, &env
->pfault_compare
);
570 r
= kvm_get_one_reg(cs
, KVM_REG_S390_PFSELECT
, &env
->pfault_select
);
579 int kvm_s390_get_clock(uint8_t *tod_high
, uint64_t *tod_low
)
582 struct kvm_device_attr attr
= {
583 .group
= KVM_S390_VM_TOD
,
584 .attr
= KVM_S390_VM_TOD_LOW
,
585 .addr
= (uint64_t)tod_low
,
588 r
= kvm_vm_ioctl(kvm_state
, KVM_GET_DEVICE_ATTR
, &attr
);
593 attr
.attr
= KVM_S390_VM_TOD_HIGH
;
594 attr
.addr
= (uint64_t)tod_high
;
595 return kvm_vm_ioctl(kvm_state
, KVM_GET_DEVICE_ATTR
, &attr
);
598 int kvm_s390_set_clock(uint8_t *tod_high
, uint64_t *tod_low
)
602 struct kvm_device_attr attr
= {
603 .group
= KVM_S390_VM_TOD
,
604 .attr
= KVM_S390_VM_TOD_LOW
,
605 .addr
= (uint64_t)tod_low
,
608 r
= kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
613 attr
.attr
= KVM_S390_VM_TOD_HIGH
;
614 attr
.addr
= (uint64_t)tod_high
;
615 return kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
620 * @addr: the logical start address in guest memory
621 * @ar: the access register number
622 * @hostbuf: buffer in host memory. NULL = do only checks w/o copying
623 * @len: length that should be transferred
624 * @is_write: true = write, false = read
625 * Returns: 0 on success, non-zero if an exception or error occurred
627 * Use KVM ioctl to read/write from/to guest memory. An access exception
628 * is injected into the vCPU in case of translation errors.
630 int kvm_s390_mem_op(S390CPU
*cpu
, vaddr addr
, uint8_t ar
, void *hostbuf
,
631 int len
, bool is_write
)
633 struct kvm_s390_mem_op mem_op
= {
635 .flags
= KVM_S390_MEMOP_F_INJECT_EXCEPTION
,
637 .op
= is_write
? KVM_S390_MEMOP_LOGICAL_WRITE
638 : KVM_S390_MEMOP_LOGICAL_READ
,
639 .buf
= (uint64_t)hostbuf
,
648 mem_op
.flags
|= KVM_S390_MEMOP_F_CHECK_ONLY
;
651 ret
= kvm_vcpu_ioctl(CPU(cpu
), KVM_S390_MEM_OP
, &mem_op
);
653 error_printf("KVM_S390_MEM_OP failed: %s\n", strerror(-ret
));
659 * Legacy layout for s390:
660 * Older S390 KVM requires the topmost vma of the RAM to be
661 * smaller than an system defined value, which is at least 256GB.
662 * Larger systems have larger values. We put the guest between
663 * the end of data segment (system break) and this value. We
664 * use 32GB as a base to have enough room for the system break
665 * to grow. We also have to use MAP parameters that avoid
666 * read-only mapping of guest pages.
668 static void *legacy_s390_alloc(size_t size
, uint64_t *align
)
672 mem
= mmap((void *) 0x800000000ULL
, size
,
673 PROT_EXEC
|PROT_READ
|PROT_WRITE
,
674 MAP_SHARED
| MAP_ANONYMOUS
| MAP_FIXED
, -1, 0);
675 return mem
== MAP_FAILED
? NULL
: mem
;
678 static uint8_t const *sw_bp_inst
;
679 static uint8_t sw_bp_ilen
;
681 static void determine_sw_breakpoint_instr(void)
683 /* DIAG 501 is used for sw breakpoints with old kernels */
684 static const uint8_t diag_501
[] = {0x83, 0x24, 0x05, 0x01};
685 /* Instruction 0x0000 is used for sw breakpoints with recent kernels */
686 static const uint8_t instr_0x0000
[] = {0x00, 0x00};
691 if (kvm_vm_enable_cap(kvm_state
, KVM_CAP_S390_USER_INSTR0
, 0)) {
692 sw_bp_inst
= diag_501
;
693 sw_bp_ilen
= sizeof(diag_501
);
694 DPRINTF("KVM: will use 4-byte sw breakpoints.\n");
696 sw_bp_inst
= instr_0x0000
;
697 sw_bp_ilen
= sizeof(instr_0x0000
);
698 DPRINTF("KVM: will use 2-byte sw breakpoints.\n");
702 int kvm_arch_insert_sw_breakpoint(CPUState
*cs
, struct kvm_sw_breakpoint
*bp
)
704 determine_sw_breakpoint_instr();
706 if (cpu_memory_rw_debug(cs
, bp
->pc
, (uint8_t *)&bp
->saved_insn
,
708 cpu_memory_rw_debug(cs
, bp
->pc
, (uint8_t *)sw_bp_inst
, sw_bp_ilen
, 1)) {
714 int kvm_arch_remove_sw_breakpoint(CPUState
*cs
, struct kvm_sw_breakpoint
*bp
)
718 if (cpu_memory_rw_debug(cs
, bp
->pc
, t
, sw_bp_ilen
, 0)) {
720 } else if (memcmp(t
, sw_bp_inst
, sw_bp_ilen
)) {
722 } else if (cpu_memory_rw_debug(cs
, bp
->pc
, (uint8_t *)&bp
->saved_insn
,
730 static struct kvm_hw_breakpoint
*find_hw_breakpoint(target_ulong addr
,
735 for (n
= 0; n
< nb_hw_breakpoints
; n
++) {
736 if (hw_breakpoints
[n
].addr
== addr
&& hw_breakpoints
[n
].type
== type
&&
737 (hw_breakpoints
[n
].len
== len
|| len
== -1)) {
738 return &hw_breakpoints
[n
];
745 static int insert_hw_breakpoint(target_ulong addr
, int len
, int type
)
749 if (find_hw_breakpoint(addr
, len
, type
)) {
753 size
= (nb_hw_breakpoints
+ 1) * sizeof(struct kvm_hw_breakpoint
);
755 if (!hw_breakpoints
) {
756 nb_hw_breakpoints
= 0;
757 hw_breakpoints
= (struct kvm_hw_breakpoint
*)g_try_malloc(size
);
760 (struct kvm_hw_breakpoint
*)g_try_realloc(hw_breakpoints
, size
);
763 if (!hw_breakpoints
) {
764 nb_hw_breakpoints
= 0;
768 hw_breakpoints
[nb_hw_breakpoints
].addr
= addr
;
769 hw_breakpoints
[nb_hw_breakpoints
].len
= len
;
770 hw_breakpoints
[nb_hw_breakpoints
].type
= type
;
777 int kvm_arch_insert_hw_breakpoint(target_ulong addr
,
778 target_ulong len
, int type
)
781 case GDB_BREAKPOINT_HW
:
784 case GDB_WATCHPOINT_WRITE
:
788 type
= KVM_HW_WP_WRITE
;
793 return insert_hw_breakpoint(addr
, len
, type
);
796 int kvm_arch_remove_hw_breakpoint(target_ulong addr
,
797 target_ulong len
, int type
)
800 struct kvm_hw_breakpoint
*bp
= find_hw_breakpoint(addr
, len
, type
);
807 if (nb_hw_breakpoints
> 0) {
809 * In order to trim the array, move the last element to the position to
810 * be removed - if necessary.
812 if (bp
!= &hw_breakpoints
[nb_hw_breakpoints
]) {
813 *bp
= hw_breakpoints
[nb_hw_breakpoints
];
815 size
= nb_hw_breakpoints
* sizeof(struct kvm_hw_breakpoint
);
817 (struct kvm_hw_breakpoint
*)g_realloc(hw_breakpoints
, size
);
819 g_free(hw_breakpoints
);
820 hw_breakpoints
= NULL
;
826 void kvm_arch_remove_all_hw_breakpoints(void)
828 nb_hw_breakpoints
= 0;
829 g_free(hw_breakpoints
);
830 hw_breakpoints
= NULL
;
833 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
)
837 if (nb_hw_breakpoints
> 0) {
838 dbg
->arch
.nr_hw_bp
= nb_hw_breakpoints
;
839 dbg
->arch
.hw_bp
= hw_breakpoints
;
841 for (i
= 0; i
< nb_hw_breakpoints
; ++i
) {
842 hw_breakpoints
[i
].phys_addr
= s390_cpu_get_phys_addr_debug(cpu
,
843 hw_breakpoints
[i
].addr
);
845 dbg
->control
|= KVM_GUESTDBG_ENABLE
| KVM_GUESTDBG_USE_HW_BP
;
847 dbg
->arch
.nr_hw_bp
= 0;
848 dbg
->arch
.hw_bp
= NULL
;
852 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
)
856 MemTxAttrs
kvm_arch_post_run(CPUState
*cs
, struct kvm_run
*run
)
858 return MEMTXATTRS_UNSPECIFIED
;
861 int kvm_arch_process_async_events(CPUState
*cs
)
866 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq
*irq
,
867 struct kvm_s390_interrupt
*interrupt
)
871 interrupt
->type
= irq
->type
;
873 case KVM_S390_INT_VIRTIO
:
874 interrupt
->parm
= irq
->u
.ext
.ext_params
;
876 case KVM_S390_INT_PFAULT_INIT
:
877 case KVM_S390_INT_PFAULT_DONE
:
878 interrupt
->parm64
= irq
->u
.ext
.ext_params2
;
880 case KVM_S390_PROGRAM_INT
:
881 interrupt
->parm
= irq
->u
.pgm
.code
;
883 case KVM_S390_SIGP_SET_PREFIX
:
884 interrupt
->parm
= irq
->u
.prefix
.address
;
886 case KVM_S390_INT_SERVICE
:
887 interrupt
->parm
= irq
->u
.ext
.ext_params
;
890 interrupt
->parm
= irq
->u
.mchk
.cr14
;
891 interrupt
->parm64
= irq
->u
.mchk
.mcic
;
893 case KVM_S390_INT_EXTERNAL_CALL
:
894 interrupt
->parm
= irq
->u
.extcall
.code
;
896 case KVM_S390_INT_EMERGENCY
:
897 interrupt
->parm
= irq
->u
.emerg
.code
;
899 case KVM_S390_SIGP_STOP
:
900 case KVM_S390_RESTART
:
901 break; /* These types have no parameters */
902 case KVM_S390_INT_IO_MIN
...KVM_S390_INT_IO_MAX
:
903 interrupt
->parm
= irq
->u
.io
.subchannel_id
<< 16;
904 interrupt
->parm
|= irq
->u
.io
.subchannel_nr
;
905 interrupt
->parm64
= (uint64_t)irq
->u
.io
.io_int_parm
<< 32;
906 interrupt
->parm64
|= irq
->u
.io
.io_int_word
;
915 static void inject_vcpu_irq_legacy(CPUState
*cs
, struct kvm_s390_irq
*irq
)
917 struct kvm_s390_interrupt kvmint
= {};
920 r
= s390_kvm_irq_to_interrupt(irq
, &kvmint
);
922 fprintf(stderr
, "%s called with bogus interrupt\n", __func__
);
926 r
= kvm_vcpu_ioctl(cs
, KVM_S390_INTERRUPT
, &kvmint
);
928 fprintf(stderr
, "KVM failed to inject interrupt\n");
933 void kvm_s390_vcpu_interrupt(S390CPU
*cpu
, struct kvm_s390_irq
*irq
)
935 CPUState
*cs
= CPU(cpu
);
939 r
= kvm_vcpu_ioctl(cs
, KVM_S390_IRQ
, irq
);
943 error_report("KVM failed to inject interrupt %llx", irq
->type
);
947 inject_vcpu_irq_legacy(cs
, irq
);
950 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq
*irq
)
952 struct kvm_s390_interrupt kvmint
= {};
955 r
= s390_kvm_irq_to_interrupt(irq
, &kvmint
);
957 fprintf(stderr
, "%s called with bogus interrupt\n", __func__
);
961 r
= kvm_vm_ioctl(kvm_state
, KVM_S390_INTERRUPT
, &kvmint
);
963 fprintf(stderr
, "KVM failed to inject interrupt\n");
968 void kvm_s390_floating_interrupt(struct kvm_s390_irq
*irq
)
970 static bool use_flic
= true;
974 r
= kvm_s390_inject_flic(irq
);
982 __kvm_s390_floating_interrupt(irq
);
985 void kvm_s390_service_interrupt(uint32_t parm
)
987 struct kvm_s390_irq irq
= {
988 .type
= KVM_S390_INT_SERVICE
,
989 .u
.ext
.ext_params
= parm
,
992 kvm_s390_floating_interrupt(&irq
);
995 static void enter_pgmcheck(S390CPU
*cpu
, uint16_t code
)
997 struct kvm_s390_irq irq
= {
998 .type
= KVM_S390_PROGRAM_INT
,
1002 kvm_s390_vcpu_interrupt(cpu
, &irq
);
1005 void kvm_s390_access_exception(S390CPU
*cpu
, uint16_t code
, uint64_t te_code
)
1007 struct kvm_s390_irq irq
= {
1008 .type
= KVM_S390_PROGRAM_INT
,
1010 .u
.pgm
.trans_exc_code
= te_code
,
1011 .u
.pgm
.exc_access_id
= te_code
& 3,
1014 kvm_s390_vcpu_interrupt(cpu
, &irq
);
1017 static int kvm_sclp_service_call(S390CPU
*cpu
, struct kvm_run
*run
,
1020 CPUS390XState
*env
= &cpu
->env
;
1025 cpu_synchronize_state(CPU(cpu
));
1026 sccb
= env
->regs
[ipbh0
& 0xf];
1027 code
= env
->regs
[(ipbh0
& 0xf0) >> 4];
1029 r
= sclp_service_call(env
, sccb
, code
);
1031 enter_pgmcheck(cpu
, -r
);
1039 static int handle_b2(S390CPU
*cpu
, struct kvm_run
*run
, uint8_t ipa1
)
1041 CPUS390XState
*env
= &cpu
->env
;
1043 uint16_t ipbh0
= (run
->s390_sieic
.ipb
& 0xffff0000) >> 16;
1045 cpu_synchronize_state(CPU(cpu
));
1049 ioinst_handle_xsch(cpu
, env
->regs
[1]);
1052 ioinst_handle_csch(cpu
, env
->regs
[1]);
1055 ioinst_handle_hsch(cpu
, env
->regs
[1]);
1058 ioinst_handle_msch(cpu
, env
->regs
[1], run
->s390_sieic
.ipb
);
1061 ioinst_handle_ssch(cpu
, env
->regs
[1], run
->s390_sieic
.ipb
);
1064 ioinst_handle_stcrw(cpu
, run
->s390_sieic
.ipb
);
1067 ioinst_handle_stsch(cpu
, env
->regs
[1], run
->s390_sieic
.ipb
);
1070 /* We should only get tsch via KVM_EXIT_S390_TSCH. */
1071 fprintf(stderr
, "Spurious tsch intercept\n");
1074 ioinst_handle_chsc(cpu
, run
->s390_sieic
.ipb
);
1077 /* This should have been handled by kvm already. */
1078 fprintf(stderr
, "Spurious tpi intercept\n");
1081 ioinst_handle_schm(cpu
, env
->regs
[1], env
->regs
[2],
1082 run
->s390_sieic
.ipb
);
1085 ioinst_handle_rsch(cpu
, env
->regs
[1]);
1088 ioinst_handle_rchp(cpu
, env
->regs
[1]);
1091 /* We do not provide this instruction, it is suppressed. */
1094 ioinst_handle_sal(cpu
, env
->regs
[1]);
1097 /* Not provided, set CC = 3 for subchannel not operational */
1100 case PRIV_B2_SCLP_CALL
:
1101 rc
= kvm_sclp_service_call(cpu
, run
, ipbh0
);
1105 DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1
);
1112 static uint64_t get_base_disp_rxy(S390CPU
*cpu
, struct kvm_run
*run
,
1115 CPUS390XState
*env
= &cpu
->env
;
1116 uint32_t x2
= (run
->s390_sieic
.ipa
& 0x000f);
1117 uint32_t base2
= run
->s390_sieic
.ipb
>> 28;
1118 uint32_t disp2
= ((run
->s390_sieic
.ipb
& 0x0fff0000) >> 16) +
1119 ((run
->s390_sieic
.ipb
& 0xff00) << 4);
1121 if (disp2
& 0x80000) {
1122 disp2
+= 0xfff00000;
1128 return (base2
? env
->regs
[base2
] : 0) +
1129 (x2
? env
->regs
[x2
] : 0) + (long)(int)disp2
;
1132 static uint64_t get_base_disp_rsy(S390CPU
*cpu
, struct kvm_run
*run
,
1135 CPUS390XState
*env
= &cpu
->env
;
1136 uint32_t base2
= run
->s390_sieic
.ipb
>> 28;
1137 uint32_t disp2
= ((run
->s390_sieic
.ipb
& 0x0fff0000) >> 16) +
1138 ((run
->s390_sieic
.ipb
& 0xff00) << 4);
1140 if (disp2
& 0x80000) {
1141 disp2
+= 0xfff00000;
1147 return (base2
? env
->regs
[base2
] : 0) + (long)(int)disp2
;
1150 static int kvm_clp_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1152 uint8_t r2
= (run
->s390_sieic
.ipb
& 0x000f0000) >> 16;
1154 return clp_service_call(cpu
, r2
);
1157 static int kvm_pcilg_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1159 uint8_t r1
= (run
->s390_sieic
.ipb
& 0x00f00000) >> 20;
1160 uint8_t r2
= (run
->s390_sieic
.ipb
& 0x000f0000) >> 16;
1162 return pcilg_service_call(cpu
, r1
, r2
);
1165 static int kvm_pcistg_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1167 uint8_t r1
= (run
->s390_sieic
.ipb
& 0x00f00000) >> 20;
1168 uint8_t r2
= (run
->s390_sieic
.ipb
& 0x000f0000) >> 16;
1170 return pcistg_service_call(cpu
, r1
, r2
);
1173 static int kvm_stpcifc_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1175 uint8_t r1
= (run
->s390_sieic
.ipa
& 0x00f0) >> 4;
1179 cpu_synchronize_state(CPU(cpu
));
1180 fiba
= get_base_disp_rxy(cpu
, run
, &ar
);
1182 return stpcifc_service_call(cpu
, r1
, fiba
, ar
);
1185 static int kvm_sic_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1191 static int kvm_rpcit_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1193 uint8_t r1
= (run
->s390_sieic
.ipb
& 0x00f00000) >> 20;
1194 uint8_t r2
= (run
->s390_sieic
.ipb
& 0x000f0000) >> 16;
1196 return rpcit_service_call(cpu
, r1
, r2
);
1199 static int kvm_pcistb_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1201 uint8_t r1
= (run
->s390_sieic
.ipa
& 0x00f0) >> 4;
1202 uint8_t r3
= run
->s390_sieic
.ipa
& 0x000f;
1206 cpu_synchronize_state(CPU(cpu
));
1207 gaddr
= get_base_disp_rsy(cpu
, run
, &ar
);
1209 return pcistb_service_call(cpu
, r1
, r3
, gaddr
, ar
);
1212 static int kvm_mpcifc_service_call(S390CPU
*cpu
, struct kvm_run
*run
)
1214 uint8_t r1
= (run
->s390_sieic
.ipa
& 0x00f0) >> 4;
1218 cpu_synchronize_state(CPU(cpu
));
1219 fiba
= get_base_disp_rxy(cpu
, run
, &ar
);
1221 return mpcifc_service_call(cpu
, r1
, fiba
, ar
);
1224 static int handle_b9(S390CPU
*cpu
, struct kvm_run
*run
, uint8_t ipa1
)
1230 r
= kvm_clp_service_call(cpu
, run
);
1232 case PRIV_B9_PCISTG
:
1233 r
= kvm_pcistg_service_call(cpu
, run
);
1236 r
= kvm_pcilg_service_call(cpu
, run
);
1239 r
= kvm_rpcit_service_call(cpu
, run
);
1242 /* just inject exception */
1247 DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1
);
1254 static int handle_eb(S390CPU
*cpu
, struct kvm_run
*run
, uint8_t ipbl
)
1259 case PRIV_EB_PCISTB
:
1260 r
= kvm_pcistb_service_call(cpu
, run
);
1263 r
= kvm_sic_service_call(cpu
, run
);
1266 /* just inject exception */
1271 DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl
);
1278 static int handle_e3(S390CPU
*cpu
, struct kvm_run
*run
, uint8_t ipbl
)
1283 case PRIV_E3_MPCIFC
:
1284 r
= kvm_mpcifc_service_call(cpu
, run
);
1286 case PRIV_E3_STPCIFC
:
1287 r
= kvm_stpcifc_service_call(cpu
, run
);
1291 DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl
);
1298 static int handle_hypercall(S390CPU
*cpu
, struct kvm_run
*run
)
1300 CPUS390XState
*env
= &cpu
->env
;
1303 cpu_synchronize_state(CPU(cpu
));
1304 ret
= s390_virtio_hypercall(env
);
1305 if (ret
== -EINVAL
) {
1306 enter_pgmcheck(cpu
, PGM_SPECIFICATION
);
1313 static void kvm_handle_diag_288(S390CPU
*cpu
, struct kvm_run
*run
)
1318 cpu_synchronize_state(CPU(cpu
));
1319 r1
= (run
->s390_sieic
.ipa
& 0x00f0) >> 4;
1320 r3
= run
->s390_sieic
.ipa
& 0x000f;
1321 rc
= handle_diag_288(&cpu
->env
, r1
, r3
);
1323 enter_pgmcheck(cpu
, PGM_SPECIFICATION
);
1327 static void kvm_handle_diag_308(S390CPU
*cpu
, struct kvm_run
*run
)
1331 cpu_synchronize_state(CPU(cpu
));
1332 r1
= (run
->s390_sieic
.ipa
& 0x00f0) >> 4;
1333 r3
= run
->s390_sieic
.ipa
& 0x000f;
1334 handle_diag_308(&cpu
->env
, r1
, r3
);
1337 static int handle_sw_breakpoint(S390CPU
*cpu
, struct kvm_run
*run
)
1339 CPUS390XState
*env
= &cpu
->env
;
1342 cpu_synchronize_state(CPU(cpu
));
1344 pc
= env
->psw
.addr
- sw_bp_ilen
;
1345 if (kvm_find_sw_breakpoint(CPU(cpu
), pc
)) {
1353 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1355 static int handle_diag(S390CPU
*cpu
, struct kvm_run
*run
, uint32_t ipb
)
1361 * For any diagnose call we support, bits 48-63 of the resulting
1362 * address specify the function code; the remainder is ignored.
1364 func_code
= decode_basedisp_rs(&cpu
->env
, ipb
, NULL
) & DIAG_KVM_CODE_MASK
;
1365 switch (func_code
) {
1366 case DIAG_TIMEREVENT
:
1367 kvm_handle_diag_288(cpu
, run
);
1370 kvm_handle_diag_308(cpu
, run
);
1372 case DIAG_KVM_HYPERCALL
:
1373 r
= handle_hypercall(cpu
, run
);
1375 case DIAG_KVM_BREAKPOINT
:
1376 r
= handle_sw_breakpoint(cpu
, run
);
1379 DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code
);
1380 enter_pgmcheck(cpu
, PGM_SPECIFICATION
);
1387 typedef struct SigpInfo
{
1391 uint64_t *status_reg
;
1394 static void set_sigp_status(SigpInfo
*si
, uint64_t status
)
1396 *si
->status_reg
&= 0xffffffff00000000ULL
;
1397 *si
->status_reg
|= status
;
1398 si
->cc
= SIGP_CC_STATUS_STORED
;
1401 static void sigp_start(void *arg
)
1405 if (s390_cpu_get_state(si
->cpu
) != CPU_STATE_STOPPED
) {
1406 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1410 s390_cpu_set_state(CPU_STATE_OPERATING
, si
->cpu
);
1411 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1414 static void sigp_stop(void *arg
)
1417 struct kvm_s390_irq irq
= {
1418 .type
= KVM_S390_SIGP_STOP
,
1421 if (s390_cpu_get_state(si
->cpu
) != CPU_STATE_OPERATING
) {
1422 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1426 /* disabled wait - sleeping in user space */
1427 if (CPU(si
->cpu
)->halted
) {
1428 s390_cpu_set_state(CPU_STATE_STOPPED
, si
->cpu
);
1430 /* execute the stop function */
1431 si
->cpu
->env
.sigp_order
= SIGP_STOP
;
1432 kvm_s390_vcpu_interrupt(si
->cpu
, &irq
);
1434 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1437 #define ADTL_SAVE_AREA_SIZE 1024
1438 static int kvm_s390_store_adtl_status(S390CPU
*cpu
, hwaddr addr
)
1441 hwaddr len
= ADTL_SAVE_AREA_SIZE
;
1443 mem
= cpu_physical_memory_map(addr
, &len
, 1);
1447 if (len
!= ADTL_SAVE_AREA_SIZE
) {
1448 cpu_physical_memory_unmap(mem
, len
, 1, 0);
1452 memcpy(mem
, &cpu
->env
.vregs
, 512);
1454 cpu_physical_memory_unmap(mem
, len
, 1, len
);
1459 #define KVM_S390_STORE_STATUS_DEF_ADDR offsetof(LowCore, floating_pt_save_area)
1460 #define SAVE_AREA_SIZE 512
1461 static int kvm_s390_store_status(S390CPU
*cpu
, hwaddr addr
, bool store_arch
)
1463 static const uint8_t ar_id
= 1;
1464 uint64_t ckc
= cpu
->env
.ckc
>> 8;
1467 hwaddr len
= SAVE_AREA_SIZE
;
1469 mem
= cpu_physical_memory_map(addr
, &len
, 1);
1473 if (len
!= SAVE_AREA_SIZE
) {
1474 cpu_physical_memory_unmap(mem
, len
, 1, 0);
1479 cpu_physical_memory_write(offsetof(LowCore
, ar_access_id
), &ar_id
, 1);
1481 for (i
= 0; i
< 16; ++i
) {
1482 *((uint64_t *)mem
+ i
) = get_freg(&cpu
->env
, i
)->ll
;
1484 memcpy(mem
+ 128, &cpu
->env
.regs
, 128);
1485 memcpy(mem
+ 256, &cpu
->env
.psw
, 16);
1486 memcpy(mem
+ 280, &cpu
->env
.psa
, 4);
1487 memcpy(mem
+ 284, &cpu
->env
.fpc
, 4);
1488 memcpy(mem
+ 292, &cpu
->env
.todpr
, 4);
1489 memcpy(mem
+ 296, &cpu
->env
.cputm
, 8);
1490 memcpy(mem
+ 304, &ckc
, 8);
1491 memcpy(mem
+ 320, &cpu
->env
.aregs
, 64);
1492 memcpy(mem
+ 384, &cpu
->env
.cregs
, 128);
1494 cpu_physical_memory_unmap(mem
, len
, 1, len
);
1499 static void sigp_stop_and_store_status(void *arg
)
1502 struct kvm_s390_irq irq
= {
1503 .type
= KVM_S390_SIGP_STOP
,
1506 /* disabled wait - sleeping in user space */
1507 if (s390_cpu_get_state(si
->cpu
) == CPU_STATE_OPERATING
&&
1508 CPU(si
->cpu
)->halted
) {
1509 s390_cpu_set_state(CPU_STATE_STOPPED
, si
->cpu
);
1512 switch (s390_cpu_get_state(si
->cpu
)) {
1513 case CPU_STATE_OPERATING
:
1514 si
->cpu
->env
.sigp_order
= SIGP_STOP_STORE_STATUS
;
1515 kvm_s390_vcpu_interrupt(si
->cpu
, &irq
);
1516 /* store will be performed when handling the stop intercept */
1518 case CPU_STATE_STOPPED
:
1519 /* already stopped, just store the status */
1520 cpu_synchronize_state(CPU(si
->cpu
));
1521 kvm_s390_store_status(si
->cpu
, KVM_S390_STORE_STATUS_DEF_ADDR
, true);
1524 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1527 static void sigp_store_status_at_address(void *arg
)
1530 uint32_t address
= si
->param
& 0x7ffffe00u
;
1532 /* cpu has to be stopped */
1533 if (s390_cpu_get_state(si
->cpu
) != CPU_STATE_STOPPED
) {
1534 set_sigp_status(si
, SIGP_STAT_INCORRECT_STATE
);
1538 cpu_synchronize_state(CPU(si
->cpu
));
1540 if (kvm_s390_store_status(si
->cpu
, address
, false)) {
1541 set_sigp_status(si
, SIGP_STAT_INVALID_PARAMETER
);
1544 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1547 static void sigp_store_adtl_status(void *arg
)
1551 if (!s390_has_feat(S390_FEAT_VECTOR
)) {
1552 set_sigp_status(si
, SIGP_STAT_INVALID_ORDER
);
1556 /* cpu has to be stopped */
1557 if (s390_cpu_get_state(si
->cpu
) != CPU_STATE_STOPPED
) {
1558 set_sigp_status(si
, SIGP_STAT_INCORRECT_STATE
);
1562 /* parameter must be aligned to 1024-byte boundary */
1563 if (si
->param
& 0x3ff) {
1564 set_sigp_status(si
, SIGP_STAT_INVALID_PARAMETER
);
1568 cpu_synchronize_state(CPU(si
->cpu
));
1570 if (kvm_s390_store_adtl_status(si
->cpu
, si
->param
)) {
1571 set_sigp_status(si
, SIGP_STAT_INVALID_PARAMETER
);
1574 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1577 static void sigp_restart(void *arg
)
1580 struct kvm_s390_irq irq
= {
1581 .type
= KVM_S390_RESTART
,
1584 switch (s390_cpu_get_state(si
->cpu
)) {
1585 case CPU_STATE_STOPPED
:
1586 /* the restart irq has to be delivered prior to any other pending irq */
1587 cpu_synchronize_state(CPU(si
->cpu
));
1588 do_restart_interrupt(&si
->cpu
->env
);
1589 s390_cpu_set_state(CPU_STATE_OPERATING
, si
->cpu
);
1591 case CPU_STATE_OPERATING
:
1592 kvm_s390_vcpu_interrupt(si
->cpu
, &irq
);
1595 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1598 int kvm_s390_cpu_restart(S390CPU
*cpu
)
1604 run_on_cpu(CPU(cpu
), sigp_restart
, &si
);
1605 DPRINTF("DONE: KVM cpu restart: %p\n", &cpu
->env
);
1609 static void sigp_initial_cpu_reset(void *arg
)
1612 CPUState
*cs
= CPU(si
->cpu
);
1613 S390CPUClass
*scc
= S390_CPU_GET_CLASS(si
->cpu
);
1615 cpu_synchronize_state(cs
);
1616 scc
->initial_cpu_reset(cs
);
1617 cpu_synchronize_post_reset(cs
);
1618 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1621 static void sigp_cpu_reset(void *arg
)
1624 CPUState
*cs
= CPU(si
->cpu
);
1625 S390CPUClass
*scc
= S390_CPU_GET_CLASS(si
->cpu
);
1627 cpu_synchronize_state(cs
);
1629 cpu_synchronize_post_reset(cs
);
1630 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1633 static void sigp_set_prefix(void *arg
)
1636 uint32_t addr
= si
->param
& 0x7fffe000u
;
1638 cpu_synchronize_state(CPU(si
->cpu
));
1640 if (!address_space_access_valid(&address_space_memory
, addr
,
1641 sizeof(struct LowCore
), false)) {
1642 set_sigp_status(si
, SIGP_STAT_INVALID_PARAMETER
);
1646 /* cpu has to be stopped */
1647 if (s390_cpu_get_state(si
->cpu
) != CPU_STATE_STOPPED
) {
1648 set_sigp_status(si
, SIGP_STAT_INCORRECT_STATE
);
1652 si
->cpu
->env
.psa
= addr
;
1653 cpu_synchronize_post_init(CPU(si
->cpu
));
1654 si
->cc
= SIGP_CC_ORDER_CODE_ACCEPTED
;
1657 static int handle_sigp_single_dst(S390CPU
*dst_cpu
, uint8_t order
,
1658 uint64_t param
, uint64_t *status_reg
)
1663 .status_reg
= status_reg
,
1666 /* cpu available? */
1667 if (dst_cpu
== NULL
) {
1668 return SIGP_CC_NOT_OPERATIONAL
;
1671 /* only resets can break pending orders */
1672 if (dst_cpu
->env
.sigp_order
!= 0 &&
1673 order
!= SIGP_CPU_RESET
&&
1674 order
!= SIGP_INITIAL_CPU_RESET
) {
1675 return SIGP_CC_BUSY
;
1680 run_on_cpu(CPU(dst_cpu
), sigp_start
, &si
);
1683 run_on_cpu(CPU(dst_cpu
), sigp_stop
, &si
);
1686 run_on_cpu(CPU(dst_cpu
), sigp_restart
, &si
);
1688 case SIGP_STOP_STORE_STATUS
:
1689 run_on_cpu(CPU(dst_cpu
), sigp_stop_and_store_status
, &si
);
1691 case SIGP_STORE_STATUS_ADDR
:
1692 run_on_cpu(CPU(dst_cpu
), sigp_store_status_at_address
, &si
);
1694 case SIGP_STORE_ADTL_STATUS
:
1695 run_on_cpu(CPU(dst_cpu
), sigp_store_adtl_status
, &si
);
1697 case SIGP_SET_PREFIX
:
1698 run_on_cpu(CPU(dst_cpu
), sigp_set_prefix
, &si
);
1700 case SIGP_INITIAL_CPU_RESET
:
1701 run_on_cpu(CPU(dst_cpu
), sigp_initial_cpu_reset
, &si
);
1703 case SIGP_CPU_RESET
:
1704 run_on_cpu(CPU(dst_cpu
), sigp_cpu_reset
, &si
);
1707 DPRINTF("KVM: unknown SIGP: 0x%x\n", order
);
1708 set_sigp_status(&si
, SIGP_STAT_INVALID_ORDER
);
1714 static int sigp_set_architecture(S390CPU
*cpu
, uint32_t param
,
1715 uint64_t *status_reg
)
1720 /* due to the BQL, we are the only active cpu */
1721 CPU_FOREACH(cur_cs
) {
1722 cur_cpu
= S390_CPU(cur_cs
);
1723 if (cur_cpu
->env
.sigp_order
!= 0) {
1724 return SIGP_CC_BUSY
;
1726 cpu_synchronize_state(cur_cs
);
1727 /* all but the current one have to be stopped */
1728 if (cur_cpu
!= cpu
&&
1729 s390_cpu_get_state(cur_cpu
) != CPU_STATE_STOPPED
) {
1730 *status_reg
&= 0xffffffff00000000ULL
;
1731 *status_reg
|= SIGP_STAT_INCORRECT_STATE
;
1732 return SIGP_CC_STATUS_STORED
;
1736 switch (param
& 0xff) {
1737 case SIGP_MODE_ESA_S390
:
1739 return SIGP_CC_NOT_OPERATIONAL
;
1740 case SIGP_MODE_Z_ARCH_TRANS_ALL_PSW
:
1741 case SIGP_MODE_Z_ARCH_TRANS_CUR_PSW
:
1742 CPU_FOREACH(cur_cs
) {
1743 cur_cpu
= S390_CPU(cur_cs
);
1744 cur_cpu
->env
.pfault_token
= -1UL;
1748 *status_reg
&= 0xffffffff00000000ULL
;
1749 *status_reg
|= SIGP_STAT_INVALID_PARAMETER
;
1750 return SIGP_CC_STATUS_STORED
;
1753 return SIGP_CC_ORDER_CODE_ACCEPTED
;
1756 #define SIGP_ORDER_MASK 0x000000ff
1758 static int handle_sigp(S390CPU
*cpu
, struct kvm_run
*run
, uint8_t ipa1
)
1760 CPUS390XState
*env
= &cpu
->env
;
1761 const uint8_t r1
= ipa1
>> 4;
1762 const uint8_t r3
= ipa1
& 0x0f;
1765 uint64_t *status_reg
;
1767 S390CPU
*dst_cpu
= NULL
;
1769 cpu_synchronize_state(CPU(cpu
));
1771 /* get order code */
1772 order
= decode_basedisp_rs(env
, run
->s390_sieic
.ipb
, NULL
)
1774 status_reg
= &env
->regs
[r1
];
1775 param
= (r1
% 2) ? env
->regs
[r1
] : env
->regs
[r1
+ 1];
1779 ret
= sigp_set_architecture(cpu
, param
, status_reg
);
1782 /* all other sigp orders target a single vcpu */
1783 dst_cpu
= s390_cpu_addr2state(env
->regs
[r3
]);
1784 ret
= handle_sigp_single_dst(dst_cpu
, order
, param
, status_reg
);
1787 trace_kvm_sigp_finished(order
, CPU(cpu
)->cpu_index
,
1788 dst_cpu
? CPU(dst_cpu
)->cpu_index
: -1, ret
);
1798 static int handle_instruction(S390CPU
*cpu
, struct kvm_run
*run
)
1800 unsigned int ipa0
= (run
->s390_sieic
.ipa
& 0xff00);
1801 uint8_t ipa1
= run
->s390_sieic
.ipa
& 0x00ff;
1804 DPRINTF("handle_instruction 0x%x 0x%x\n",
1805 run
->s390_sieic
.ipa
, run
->s390_sieic
.ipb
);
1808 r
= handle_b2(cpu
, run
, ipa1
);
1811 r
= handle_b9(cpu
, run
, ipa1
);
1814 r
= handle_eb(cpu
, run
, run
->s390_sieic
.ipb
& 0xff);
1817 r
= handle_e3(cpu
, run
, run
->s390_sieic
.ipb
& 0xff);
1820 r
= handle_diag(cpu
, run
, run
->s390_sieic
.ipb
);
1823 r
= handle_sigp(cpu
, run
, ipa1
);
1829 enter_pgmcheck(cpu
, 0x0001);
1835 static bool is_special_wait_psw(CPUState
*cs
)
1837 /* signal quiesce */
1838 return cs
->kvm_run
->psw_addr
== 0xfffUL
;
1841 static void unmanageable_intercept(S390CPU
*cpu
, const char *str
, int pswoffset
)
1843 CPUState
*cs
= CPU(cpu
);
1845 error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1846 str
, cs
->cpu_index
, ldq_phys(cs
->as
, cpu
->env
.psa
+ pswoffset
),
1847 ldq_phys(cs
->as
, cpu
->env
.psa
+ pswoffset
+ 8));
1849 qemu_system_guest_panicked();
1852 static int handle_intercept(S390CPU
*cpu
)
1854 CPUState
*cs
= CPU(cpu
);
1855 struct kvm_run
*run
= cs
->kvm_run
;
1856 int icpt_code
= run
->s390_sieic
.icptcode
;
1859 DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code
,
1860 (long)cs
->kvm_run
->psw_addr
);
1861 switch (icpt_code
) {
1862 case ICPT_INSTRUCTION
:
1863 r
= handle_instruction(cpu
, run
);
1866 unmanageable_intercept(cpu
, "program interrupt",
1867 offsetof(LowCore
, program_new_psw
));
1871 unmanageable_intercept(cpu
, "external interrupt",
1872 offsetof(LowCore
, external_new_psw
));
1876 /* disabled wait, since enabled wait is handled in kernel */
1877 cpu_synchronize_state(cs
);
1878 if (s390_cpu_halt(cpu
) == 0) {
1879 if (is_special_wait_psw(cs
)) {
1880 qemu_system_shutdown_request();
1882 qemu_system_guest_panicked();
1888 if (s390_cpu_set_state(CPU_STATE_STOPPED
, cpu
) == 0) {
1889 qemu_system_shutdown_request();
1891 if (cpu
->env
.sigp_order
== SIGP_STOP_STORE_STATUS
) {
1892 kvm_s390_store_status(cpu
, KVM_S390_STORE_STATUS_DEF_ADDR
,
1895 cpu
->env
.sigp_order
= 0;
1899 /* currently only instr 0x0000 after enabled via capability */
1900 r
= handle_sw_breakpoint(cpu
, run
);
1902 enter_pgmcheck(cpu
, PGM_OPERATION
);
1906 case ICPT_SOFT_INTERCEPT
:
1907 fprintf(stderr
, "KVM unimplemented icpt SOFT\n");
1911 fprintf(stderr
, "KVM unimplemented icpt IO\n");
1915 fprintf(stderr
, "Unknown intercept code: %d\n", icpt_code
);
1923 static int handle_tsch(S390CPU
*cpu
)
1925 CPUState
*cs
= CPU(cpu
);
1926 struct kvm_run
*run
= cs
->kvm_run
;
1929 cpu_synchronize_state(cs
);
1931 ret
= ioinst_handle_tsch(cpu
, cpu
->env
.regs
[1], run
->s390_tsch
.ipb
);
1935 * If an I/O interrupt had been dequeued, we have to reinject it.
1937 if (run
->s390_tsch
.dequeued
) {
1938 kvm_s390_io_interrupt(run
->s390_tsch
.subchannel_id
,
1939 run
->s390_tsch
.subchannel_nr
,
1940 run
->s390_tsch
.io_int_parm
,
1941 run
->s390_tsch
.io_int_word
);
1948 static void insert_stsi_3_2_2(S390CPU
*cpu
, __u64 addr
, uint8_t ar
)
1950 struct sysib_322 sysib
;
1953 if (s390_cpu_virt_mem_read(cpu
, addr
, ar
, &sysib
, sizeof(sysib
))) {
1956 /* Shift the stack of Extended Names to prepare for our own data */
1957 memmove(&sysib
.ext_names
[1], &sysib
.ext_names
[0],
1958 sizeof(sysib
.ext_names
[0]) * (sysib
.count
- 1));
1959 /* First virt level, that doesn't provide Ext Names delimits stack. It is
1960 * assumed it's not capable of managing Extended Names for lower levels.
1962 for (del
= 1; del
< sysib
.count
; del
++) {
1963 if (!sysib
.vm
[del
].ext_name_encoding
|| !sysib
.ext_names
[del
][0]) {
1967 if (del
< sysib
.count
) {
1968 memset(sysib
.ext_names
[del
], 0,
1969 sizeof(sysib
.ext_names
[0]) * (sysib
.count
- del
));
1971 /* Insert short machine name in EBCDIC, padded with blanks */
1973 memset(sysib
.vm
[0].name
, 0x40, sizeof(sysib
.vm
[0].name
));
1974 ebcdic_put(sysib
.vm
[0].name
, qemu_name
, MIN(sizeof(sysib
.vm
[0].name
),
1975 strlen(qemu_name
)));
1977 sysib
.vm
[0].ext_name_encoding
= 2; /* 2 = UTF-8 */
1978 memset(sysib
.ext_names
[0], 0, sizeof(sysib
.ext_names
[0]));
1979 /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1980 * considered by s390 as not capable of providing any Extended Name.
1981 * Therefore if no name was specified on qemu invocation, we go with the
1982 * same "KVMguest" default, which KVM has filled into short name field.
1985 strncpy((char *)sysib
.ext_names
[0], qemu_name
,
1986 sizeof(sysib
.ext_names
[0]));
1988 strcpy((char *)sysib
.ext_names
[0], "KVMguest");
1991 memcpy(sysib
.vm
[0].uuid
, qemu_uuid
, sizeof(sysib
.vm
[0].uuid
));
1993 s390_cpu_virt_mem_write(cpu
, addr
, ar
, &sysib
, sizeof(sysib
));
1996 static int handle_stsi(S390CPU
*cpu
)
1998 CPUState
*cs
= CPU(cpu
);
1999 struct kvm_run
*run
= cs
->kvm_run
;
2001 switch (run
->s390_stsi
.fc
) {
2003 if (run
->s390_stsi
.sel1
!= 2 || run
->s390_stsi
.sel2
!= 2) {
2006 /* Only sysib 3.2.2 needs post-handling for now. */
2007 insert_stsi_3_2_2(cpu
, run
->s390_stsi
.addr
, run
->s390_stsi
.ar
);
2014 static int kvm_arch_handle_debug_exit(S390CPU
*cpu
)
2016 CPUState
*cs
= CPU(cpu
);
2017 struct kvm_run
*run
= cs
->kvm_run
;
2020 struct kvm_debug_exit_arch
*arch_info
= &run
->debug
.arch
;
2022 switch (arch_info
->type
) {
2023 case KVM_HW_WP_WRITE
:
2024 if (find_hw_breakpoint(arch_info
->addr
, -1, arch_info
->type
)) {
2025 cs
->watchpoint_hit
= &hw_watchpoint
;
2026 hw_watchpoint
.vaddr
= arch_info
->addr
;
2027 hw_watchpoint
.flags
= BP_MEM_WRITE
;
2032 if (find_hw_breakpoint(arch_info
->addr
, -1, arch_info
->type
)) {
2036 case KVM_SINGLESTEP
:
2037 if (cs
->singlestep_enabled
) {
2048 int kvm_arch_handle_exit(CPUState
*cs
, struct kvm_run
*run
)
2050 S390CPU
*cpu
= S390_CPU(cs
);
2053 qemu_mutex_lock_iothread();
2055 switch (run
->exit_reason
) {
2056 case KVM_EXIT_S390_SIEIC
:
2057 ret
= handle_intercept(cpu
);
2059 case KVM_EXIT_S390_RESET
:
2060 s390_reipl_request();
2062 case KVM_EXIT_S390_TSCH
:
2063 ret
= handle_tsch(cpu
);
2065 case KVM_EXIT_S390_STSI
:
2066 ret
= handle_stsi(cpu
);
2068 case KVM_EXIT_DEBUG
:
2069 ret
= kvm_arch_handle_debug_exit(cpu
);
2072 fprintf(stderr
, "Unknown KVM exit: %d\n", run
->exit_reason
);
2075 qemu_mutex_unlock_iothread();
2078 ret
= EXCP_INTERRUPT
;
2083 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
)
2088 int kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
)
2093 int kvm_arch_on_sigbus(int code
, void *addr
)
2098 void kvm_s390_io_interrupt(uint16_t subchannel_id
,
2099 uint16_t subchannel_nr
, uint32_t io_int_parm
,
2100 uint32_t io_int_word
)
2102 struct kvm_s390_irq irq
= {
2103 .u
.io
.subchannel_id
= subchannel_id
,
2104 .u
.io
.subchannel_nr
= subchannel_nr
,
2105 .u
.io
.io_int_parm
= io_int_parm
,
2106 .u
.io
.io_int_word
= io_int_word
,
2109 if (io_int_word
& IO_INT_WORD_AI
) {
2110 irq
.type
= KVM_S390_INT_IO(1, 0, 0, 0);
2112 irq
.type
= KVM_S390_INT_IO(0, (subchannel_id
& 0xff00) >> 8,
2113 (subchannel_id
& 0x0006),
2116 kvm_s390_floating_interrupt(&irq
);
2119 static uint64_t build_channel_report_mcic(void)
2123 /* subclass: indicate channel report pending */
2125 /* subclass modifiers: none */
2126 /* storage errors: none */
2127 /* validity bits: no damage */
2128 MCIC_VB_WP
| MCIC_VB_MS
| MCIC_VB_PM
| MCIC_VB_IA
| MCIC_VB_FP
|
2129 MCIC_VB_GR
| MCIC_VB_CR
| MCIC_VB_ST
| MCIC_VB_AR
| MCIC_VB_PR
|
2130 MCIC_VB_FC
| MCIC_VB_CT
| MCIC_VB_CC
;
2131 if (s390_has_feat(S390_FEAT_VECTOR
)) {
2137 void kvm_s390_crw_mchk(void)
2139 struct kvm_s390_irq irq
= {
2140 .type
= KVM_S390_MCHK
,
2141 .u
.mchk
.cr14
= 1 << 28,
2142 .u
.mchk
.mcic
= build_channel_report_mcic(),
2144 kvm_s390_floating_interrupt(&irq
);
2147 void kvm_s390_enable_css_support(S390CPU
*cpu
)
2151 /* Activate host kernel channel subsystem support. */
2152 r
= kvm_vcpu_enable_cap(CPU(cpu
), KVM_CAP_S390_CSS_SUPPORT
, 0);
2156 void kvm_arch_init_irq_routing(KVMState
*s
)
2159 * Note that while irqchip capabilities generally imply that cpustates
2160 * are handled in-kernel, it is not true for s390 (yet); therefore, we
2161 * have to override the common code kvm_halt_in_kernel_allowed setting.
2163 if (kvm_check_extension(s
, KVM_CAP_IRQ_ROUTING
)) {
2164 kvm_gsi_routing_allowed
= true;
2165 kvm_halt_in_kernel_allowed
= false;
2169 int kvm_s390_assign_subch_ioeventfd(EventNotifier
*notifier
, uint32_t sch
,
2170 int vq
, bool assign
)
2172 struct kvm_ioeventfd kick
= {
2173 .flags
= KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY
|
2174 KVM_IOEVENTFD_FLAG_DATAMATCH
,
2175 .fd
= event_notifier_get_fd(notifier
),
2180 if (!kvm_check_extension(kvm_state
, KVM_CAP_IOEVENTFD
)) {
2184 kick
.flags
|= KVM_IOEVENTFD_FLAG_DEASSIGN
;
2186 return kvm_vm_ioctl(kvm_state
, KVM_IOEVENTFD
, &kick
);
2189 int kvm_s390_get_memslot_count(KVMState
*s
)
2191 return kvm_check_extension(s
, KVM_CAP_NR_MEMSLOTS
);
2194 int kvm_s390_get_ri(void)
2199 int kvm_s390_set_cpu_state(S390CPU
*cpu
, uint8_t cpu_state
)
2201 struct kvm_mp_state mp_state
= {};
2204 /* the kvm part might not have been initialized yet */
2205 if (CPU(cpu
)->kvm_state
== NULL
) {
2209 switch (cpu_state
) {
2210 case CPU_STATE_STOPPED
:
2211 mp_state
.mp_state
= KVM_MP_STATE_STOPPED
;
2213 case CPU_STATE_CHECK_STOP
:
2214 mp_state
.mp_state
= KVM_MP_STATE_CHECK_STOP
;
2216 case CPU_STATE_OPERATING
:
2217 mp_state
.mp_state
= KVM_MP_STATE_OPERATING
;
2219 case CPU_STATE_LOAD
:
2220 mp_state
.mp_state
= KVM_MP_STATE_LOAD
;
2223 error_report("Requested CPU state is not a valid S390 CPU state: %u",
2228 ret
= kvm_vcpu_ioctl(CPU(cpu
), KVM_SET_MP_STATE
, &mp_state
);
2230 trace_kvm_failed_cpu_state_set(CPU(cpu
)->cpu_index
, cpu_state
,
2237 void kvm_s390_vcpu_interrupt_pre_save(S390CPU
*cpu
)
2239 struct kvm_s390_irq_state irq_state
;
2240 CPUState
*cs
= CPU(cpu
);
2243 if (!kvm_check_extension(kvm_state
, KVM_CAP_S390_IRQ_STATE
)) {
2247 irq_state
.buf
= (uint64_t) cpu
->irqstate
;
2248 irq_state
.len
= VCPU_IRQ_BUF_SIZE
;
2250 bytes
= kvm_vcpu_ioctl(cs
, KVM_S390_GET_IRQ_STATE
, &irq_state
);
2252 cpu
->irqstate_saved_size
= 0;
2253 error_report("Migration of interrupt state failed");
2257 cpu
->irqstate_saved_size
= bytes
;
2260 int kvm_s390_vcpu_interrupt_post_load(S390CPU
*cpu
)
2262 CPUState
*cs
= CPU(cpu
);
2263 struct kvm_s390_irq_state irq_state
;
2266 if (cpu
->irqstate_saved_size
== 0) {
2270 if (!kvm_check_extension(kvm_state
, KVM_CAP_S390_IRQ_STATE
)) {
2274 irq_state
.buf
= (uint64_t) cpu
->irqstate
;
2275 irq_state
.len
= cpu
->irqstate_saved_size
;
2277 r
= kvm_vcpu_ioctl(cs
, KVM_S390_SET_IRQ_STATE
, &irq_state
);
2279 error_report("Setting interrupt state failed %d", r
);
2284 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
2285 uint64_t address
, uint32_t data
, PCIDevice
*dev
)
2287 S390PCIBusDevice
*pbdev
;
2288 uint32_t idx
= data
>> ZPCI_MSI_VEC_BITS
;
2289 uint32_t vec
= data
& ZPCI_MSI_VEC_MASK
;
2291 pbdev
= s390_pci_find_dev_by_idx(idx
);
2293 DPRINTF("add_msi_route no dev\n");
2297 pbdev
->routes
.adapter
.ind_offset
= vec
;
2299 route
->type
= KVM_IRQ_ROUTING_S390_ADAPTER
;
2301 route
->u
.adapter
.summary_addr
= pbdev
->routes
.adapter
.summary_addr
;
2302 route
->u
.adapter
.ind_addr
= pbdev
->routes
.adapter
.ind_addr
;
2303 route
->u
.adapter
.summary_offset
= pbdev
->routes
.adapter
.summary_offset
;
2304 route
->u
.adapter
.ind_offset
= pbdev
->routes
.adapter
.ind_offset
;
2305 route
->u
.adapter
.adapter_id
= pbdev
->routes
.adapter
.adapter_id
;
2309 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry
*route
,
2310 int vector
, PCIDevice
*dev
)
2315 int kvm_arch_release_virq_post(int virq
)
2320 int kvm_arch_msi_data_to_gsi(uint32_t data
)
2325 static inline int test_bit_inv(long nr
, const unsigned long *addr
)
2327 return test_bit(BE_BIT_NR(nr
), addr
);
2330 static inline void set_bit_inv(long nr
, unsigned long *addr
)
2332 set_bit(BE_BIT_NR(nr
), addr
);
2335 static int query_cpu_subfunc(S390FeatBitmap features
)
2337 struct kvm_s390_vm_cpu_subfunc prop
;
2338 struct kvm_device_attr attr
= {
2339 .group
= KVM_S390_VM_CPU_MODEL
,
2340 .attr
= KVM_S390_VM_CPU_MACHINE_SUBFUNC
,
2341 .addr
= (uint64_t) &prop
,
2345 rc
= kvm_vm_ioctl(kvm_state
, KVM_GET_DEVICE_ATTR
, &attr
);
2351 * We're going to add all subfunctions now, if the corresponding feature
2352 * is available that unlocks the query functions.
2354 s390_add_from_feat_block(features
, S390_FEAT_TYPE_PLO
, prop
.plo
);
2355 if (test_bit(S390_FEAT_TOD_CLOCK_STEERING
, features
)) {
2356 s390_add_from_feat_block(features
, S390_FEAT_TYPE_PTFF
, prop
.ptff
);
2358 if (test_bit(S390_FEAT_MSA
, features
)) {
2359 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KMAC
, prop
.kmac
);
2360 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KMC
, prop
.kmc
);
2361 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KM
, prop
.km
);
2362 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KIMD
, prop
.kimd
);
2363 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KLMD
, prop
.klmd
);
2365 if (test_bit(S390_FEAT_MSA_EXT_3
, features
)) {
2366 s390_add_from_feat_block(features
, S390_FEAT_TYPE_PCKMO
, prop
.pckmo
);
2368 if (test_bit(S390_FEAT_MSA_EXT_4
, features
)) {
2369 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KMCTR
, prop
.kmctr
);
2370 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KMF
, prop
.kmf
);
2371 s390_add_from_feat_block(features
, S390_FEAT_TYPE_KMO
, prop
.kmo
);
2372 s390_add_from_feat_block(features
, S390_FEAT_TYPE_PCC
, prop
.pcc
);
2374 if (test_bit(S390_FEAT_MSA_EXT_5
, features
)) {
2375 s390_add_from_feat_block(features
, S390_FEAT_TYPE_PPNO
, prop
.ppno
);
2380 static int configure_cpu_subfunc(const S390FeatBitmap features
)
2382 struct kvm_s390_vm_cpu_subfunc prop
= {};
2383 struct kvm_device_attr attr
= {
2384 .group
= KVM_S390_VM_CPU_MODEL
,
2385 .attr
= KVM_S390_VM_CPU_PROCESSOR_SUBFUNC
,
2386 .addr
= (uint64_t) &prop
,
2389 if (!kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2390 KVM_S390_VM_CPU_PROCESSOR_SUBFUNC
)) {
2391 /* hardware support might be missing, IBC will handle most of this */
2395 s390_fill_feat_block(features
, S390_FEAT_TYPE_PLO
, prop
.plo
);
2396 if (test_bit(S390_FEAT_TOD_CLOCK_STEERING
, features
)) {
2397 s390_fill_feat_block(features
, S390_FEAT_TYPE_PTFF
, prop
.ptff
);
2398 prop
.ptff
[0] |= 0x80; /* query is always available */
2400 if (test_bit(S390_FEAT_MSA
, features
)) {
2401 s390_fill_feat_block(features
, S390_FEAT_TYPE_KMAC
, prop
.kmac
);
2402 prop
.kmac
[0] |= 0x80; /* query is always available */
2403 s390_fill_feat_block(features
, S390_FEAT_TYPE_KMC
, prop
.kmc
);
2404 prop
.kmc
[0] |= 0x80; /* query is always available */
2405 s390_fill_feat_block(features
, S390_FEAT_TYPE_KM
, prop
.km
);
2406 prop
.km
[0] |= 0x80; /* query is always available */
2407 s390_fill_feat_block(features
, S390_FEAT_TYPE_KIMD
, prop
.kimd
);
2408 prop
.kimd
[0] |= 0x80; /* query is always available */
2409 s390_fill_feat_block(features
, S390_FEAT_TYPE_KLMD
, prop
.klmd
);
2410 prop
.klmd
[0] |= 0x80; /* query is always available */
2412 if (test_bit(S390_FEAT_MSA_EXT_3
, features
)) {
2413 s390_fill_feat_block(features
, S390_FEAT_TYPE_PCKMO
, prop
.pckmo
);
2414 prop
.pckmo
[0] |= 0x80; /* query is always available */
2416 if (test_bit(S390_FEAT_MSA_EXT_4
, features
)) {
2417 s390_fill_feat_block(features
, S390_FEAT_TYPE_KMCTR
, prop
.kmctr
);
2418 prop
.kmctr
[0] |= 0x80; /* query is always available */
2419 s390_fill_feat_block(features
, S390_FEAT_TYPE_KMF
, prop
.kmf
);
2420 prop
.kmf
[0] |= 0x80; /* query is always available */
2421 s390_fill_feat_block(features
, S390_FEAT_TYPE_KMO
, prop
.kmo
);
2422 prop
.kmo
[0] |= 0x80; /* query is always available */
2423 s390_fill_feat_block(features
, S390_FEAT_TYPE_PCC
, prop
.pcc
);
2424 prop
.pcc
[0] |= 0x80; /* query is always available */
2426 if (test_bit(S390_FEAT_MSA_EXT_5
, features
)) {
2427 s390_fill_feat_block(features
, S390_FEAT_TYPE_PPNO
, prop
.ppno
);
2428 prop
.ppno
[0] |= 0x80; /* query is always available */
2430 return kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
2433 static int kvm_to_feat
[][2] = {
2434 { KVM_S390_VM_CPU_FEAT_ESOP
, S390_FEAT_ESOP
},
2435 { KVM_S390_VM_CPU_FEAT_SIEF2
, S390_FEAT_SIE_F2
},
2436 { KVM_S390_VM_CPU_FEAT_64BSCAO
, S390_FEAT_SIE_64BSCAO
},
2437 { KVM_S390_VM_CPU_FEAT_SIIF
, S390_FEAT_SIE_SIIF
},
2438 { KVM_S390_VM_CPU_FEAT_GPERE
, S390_FEAT_SIE_GPERE
},
2439 { KVM_S390_VM_CPU_FEAT_GSLS
, S390_FEAT_SIE_GSLS
},
2440 { KVM_S390_VM_CPU_FEAT_IB
, S390_FEAT_SIE_IB
},
2441 { KVM_S390_VM_CPU_FEAT_CEI
, S390_FEAT_SIE_CEI
},
2442 { KVM_S390_VM_CPU_FEAT_IBS
, S390_FEAT_SIE_IBS
},
2443 { KVM_S390_VM_CPU_FEAT_SKEY
, S390_FEAT_SIE_SKEY
},
2444 { KVM_S390_VM_CPU_FEAT_CMMA
, S390_FEAT_SIE_CMMA
},
2445 { KVM_S390_VM_CPU_FEAT_PFMFI
, S390_FEAT_SIE_PFMFI
},
2446 { KVM_S390_VM_CPU_FEAT_SIGPIF
, S390_FEAT_SIE_SIGPIF
},
2449 static int query_cpu_feat(S390FeatBitmap features
)
2451 struct kvm_s390_vm_cpu_feat prop
;
2452 struct kvm_device_attr attr
= {
2453 .group
= KVM_S390_VM_CPU_MODEL
,
2454 .attr
= KVM_S390_VM_CPU_MACHINE_FEAT
,
2455 .addr
= (uint64_t) &prop
,
2460 rc
= kvm_vm_ioctl(kvm_state
, KVM_GET_DEVICE_ATTR
, &attr
);
2465 for (i
= 0; i
< ARRAY_SIZE(kvm_to_feat
); i
++) {
2466 if (test_bit_inv(kvm_to_feat
[i
][0], (unsigned long *)prop
.feat
)) {
2467 set_bit(kvm_to_feat
[i
][1], features
);
2473 static int configure_cpu_feat(const S390FeatBitmap features
)
2475 struct kvm_s390_vm_cpu_feat prop
= {};
2476 struct kvm_device_attr attr
= {
2477 .group
= KVM_S390_VM_CPU_MODEL
,
2478 .attr
= KVM_S390_VM_CPU_PROCESSOR_FEAT
,
2479 .addr
= (uint64_t) &prop
,
2483 for (i
= 0; i
< ARRAY_SIZE(kvm_to_feat
); i
++) {
2484 if (test_bit(kvm_to_feat
[i
][1], features
)) {
2485 set_bit_inv(kvm_to_feat
[i
][0], (unsigned long *)prop
.feat
);
2488 return kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
2491 bool kvm_s390_cpu_models_supported(void)
2493 if (!cpu_model_allowed()) {
2494 /* compatibility machines interfere with the cpu model */
2497 return kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2498 KVM_S390_VM_CPU_MACHINE
) &&
2499 kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2500 KVM_S390_VM_CPU_PROCESSOR
) &&
2501 kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2502 KVM_S390_VM_CPU_MACHINE_FEAT
) &&
2503 kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2504 KVM_S390_VM_CPU_PROCESSOR_FEAT
) &&
2505 kvm_vm_check_attr(kvm_state
, KVM_S390_VM_CPU_MODEL
,
2506 KVM_S390_VM_CPU_MACHINE_SUBFUNC
);
2509 void kvm_s390_get_host_cpu_model(S390CPUModel
*model
, Error
**errp
)
2511 struct kvm_s390_vm_cpu_machine prop
= {};
2512 struct kvm_device_attr attr
= {
2513 .group
= KVM_S390_VM_CPU_MODEL
,
2514 .attr
= KVM_S390_VM_CPU_MACHINE
,
2515 .addr
= (uint64_t) &prop
,
2517 uint16_t unblocked_ibc
= 0, cpu_type
= 0;
2520 memset(model
, 0, sizeof(*model
));
2522 if (!kvm_s390_cpu_models_supported()) {
2523 error_setg(errp
, "KVM doesn't support CPU models");
2527 /* query the basic cpu model properties */
2528 rc
= kvm_vm_ioctl(kvm_state
, KVM_GET_DEVICE_ATTR
, &attr
);
2530 error_setg(errp
, "KVM: Error querying host CPU model: %d", rc
);
2534 cpu_type
= cpuid_type(prop
.cpuid
);
2535 if (has_ibc(prop
.ibc
)) {
2536 model
->lowest_ibc
= lowest_ibc(prop
.ibc
);
2537 unblocked_ibc
= unblocked_ibc(prop
.ibc
);
2539 model
->cpu_id
= cpuid_id(prop
.cpuid
);
2540 model
->cpu_ver
= 0xff;
2542 /* get supported cpu features indicated via STFL(E) */
2543 s390_add_from_feat_block(model
->features
, S390_FEAT_TYPE_STFL
,
2544 (uint8_t *) prop
.fac_mask
);
2545 /* dat-enhancement facility 2 has no bit but was introduced with stfle */
2546 if (test_bit(S390_FEAT_STFLE
, model
->features
)) {
2547 set_bit(S390_FEAT_DAT_ENH_2
, model
->features
);
2549 /* get supported cpu features indicated e.g. via SCLP */
2550 rc
= query_cpu_feat(model
->features
);
2552 error_setg(errp
, "KVM: Error querying CPU features: %d", rc
);
2555 /* get supported cpu subfunctions indicated via query / test bit */
2556 rc
= query_cpu_subfunc(model
->features
);
2558 error_setg(errp
, "KVM: Error querying CPU subfunctions: %d", rc
);
2562 /* with cpu model support, CMM is only indicated if really available */
2563 if (kvm_s390_cmma_available()) {
2564 set_bit(S390_FEAT_CMM
, model
->features
);
2567 if (s390_known_cpu_type(cpu_type
)) {
2568 /* we want the exact model, even if some features are missing */
2569 model
->def
= s390_find_cpu_def(cpu_type
, ibc_gen(unblocked_ibc
),
2570 ibc_ec_ga(unblocked_ibc
), NULL
);
2572 /* model unknown, e.g. too new - search using features */
2573 model
->def
= s390_find_cpu_def(0, ibc_gen(unblocked_ibc
),
2574 ibc_ec_ga(unblocked_ibc
),
2578 error_setg(errp
, "KVM: host CPU model could not be identified");
2581 /* strip of features that are not part of the maximum model */
2582 bitmap_and(model
->features
, model
->features
, model
->def
->full_feat
,
2586 void kvm_s390_apply_cpu_model(const S390CPUModel
*model
, Error
**errp
)
2588 struct kvm_s390_vm_cpu_processor prop
= {
2591 struct kvm_device_attr attr
= {
2592 .group
= KVM_S390_VM_CPU_MODEL
,
2593 .attr
= KVM_S390_VM_CPU_PROCESSOR
,
2594 .addr
= (uint64_t) &prop
,
2599 /* compatibility handling if cpu models are disabled */
2600 if (kvm_s390_cmma_available() && !mem_path
) {
2601 kvm_s390_enable_cmma();
2605 if (!kvm_s390_cpu_models_supported()) {
2606 error_setg(errp
, "KVM doesn't support CPU models");
2609 prop
.cpuid
= s390_cpuid_from_cpu_model(model
);
2610 prop
.ibc
= s390_ibc_from_cpu_model(model
);
2611 /* configure cpu features indicated via STFL(e) */
2612 s390_fill_feat_block(model
->features
, S390_FEAT_TYPE_STFL
,
2613 (uint8_t *) prop
.fac_list
);
2614 rc
= kvm_vm_ioctl(kvm_state
, KVM_SET_DEVICE_ATTR
, &attr
);
2616 error_setg(errp
, "KVM: Error configuring the CPU model: %d", rc
);
2619 /* configure cpu features indicated e.g. via SCLP */
2620 rc
= configure_cpu_feat(model
->features
);
2622 error_setg(errp
, "KVM: Error configuring CPU features: %d", rc
);
2625 /* configure cpu subfunctions indicated via query / test bit */
2626 rc
= configure_cpu_subfunc(model
->features
);
2628 error_setg(errp
, "KVM: Error configuring CPU subfunctions: %d", rc
);
2631 /* enable CMM via CMMA - disable on hugetlbfs */
2632 if (test_bit(S390_FEAT_CMM
, model
->features
)) {
2634 error_report("Warning: CMM will not be enabled because it is not "
2635 "compatible to hugetlbfs.");
2637 kvm_s390_enable_cmma();