2 * QEMU S390x KVM implementation
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 <sys/types.h>
25 #include <sys/ioctl.h>
28 #include <linux/kvm.h>
29 #include <asm/ptrace.h>
31 #include "qemu-common.h"
32 #include "qemu/timer.h"
33 #include "sysemu/sysemu.h"
34 #include "sysemu/kvm.h"
37 #include "sysemu/device_tree.h"
38 #include "qapi/qmp/qjson.h"
39 #include "monitor/monitor.h"
40 #include "exec/gdbstub.h"
41 #include "exec/address-spaces.h"
43 #include "qapi-event.h"
44 #include "hw/s390x/s390-pci-inst.h"
45 #include "hw/s390x/s390-pci-bus.h"
46 #include "hw/s390x/ipl.h"
47 #include "hw/s390x/ebcdic.h"
49 /* #define DEBUG_KVM */
52 #define DPRINTF(fmt, ...) \
53 do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
55 #define DPRINTF(fmt, ...) \
59 #define kvm_vm_check_mem_attr(s, attr) \
60 kvm_vm_check_attr(s, KVM_S390_VM_MEM_CTRL, attr)
62 #define IPA0_DIAG 0x8300
63 #define IPA0_SIGP 0xae00
64 #define IPA0_B2 0xb200
65 #define IPA0_B9 0xb900
66 #define IPA0_EB 0xeb00
67 #define IPA0_E3 0xe300
69 #define PRIV_B2_SCLP_CALL 0x20
70 #define PRIV_B2_CSCH 0x30
71 #define PRIV_B2_HSCH 0x31
72 #define PRIV_B2_MSCH 0x32
73 #define PRIV_B2_SSCH 0x33
74 #define PRIV_B2_STSCH 0x34
75 #define PRIV_B2_TSCH 0x35
76 #define PRIV_B2_TPI 0x36
77 #define PRIV_B2_SAL 0x37
78 #define PRIV_B2_RSCH 0x38
79 #define PRIV_B2_STCRW 0x39
80 #define PRIV_B2_STCPS 0x3a
81 #define PRIV_B2_RCHP 0x3b
82 #define PRIV_B2_SCHM 0x3c
83 #define PRIV_B2_CHSC 0x5f
84 #define PRIV_B2_SIGA 0x74
85 #define PRIV_B2_XSCH 0x76
87 #define PRIV_EB_SQBS 0x8a
88 #define PRIV_EB_PCISTB 0xd0
89 #define PRIV_EB_SIC 0xd1
91 #define PRIV_B9_EQBS 0x9c
92 #define PRIV_B9_CLP 0xa0
93 #define PRIV_B9_PCISTG 0xd0
94 #define PRIV_B9_PCILG 0xd2
95 #define PRIV_B9_RPCIT 0xd3
97 #define PRIV_E3_MPCIFC 0xd0
98 #define PRIV_E3_STPCIFC 0xd4
100 #define DIAG_IPL 0x308
101 #define DIAG_KVM_HYPERCALL 0x500
102 #define DIAG_KVM_BREAKPOINT 0x501
104 #define ICPT_INSTRUCTION 0x04
105 #define ICPT_PROGRAM 0x08
106 #define ICPT_EXT_INT 0x14
107 #define ICPT_WAITPSW 0x1c
108 #define ICPT_SOFT_INTERCEPT 0x24
109 #define ICPT_CPU_STOP 0x28
112 static CPUWatchpoint hw_watchpoint;
114 * We don't use a list because this structure is also used to transmit the
115 * hardware breakpoints to the kernel.
117 static struct kvm_hw_breakpoint *hw_breakpoints;
118 static int nb_hw_breakpoints;
120 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
124 static int cap_sync_regs;
125 static int cap_async_pf;
126 static int cap_mem_op;
128 static void *legacy_s390_alloc(size_t size, uint64_t *align);
130 static int kvm_s390_query_mem_limit(KVMState *s, uint64_t *memory_limit)
132 struct kvm_device_attr attr = {
133 .group = KVM_S390_VM_MEM_CTRL,
134 .attr = KVM_S390_VM_MEM_LIMIT_SIZE,
135 .addr = (uint64_t) memory_limit,
138 return kvm_vm_ioctl(s, KVM_GET_DEVICE_ATTR, &attr);
141 int kvm_s390_set_mem_limit(KVMState *s, uint64_t new_limit, uint64_t *hw_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) &new_limit,
151 if (!kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_LIMIT_SIZE)) {
155 rc = kvm_s390_query_mem_limit(s, hw_limit);
158 } else if (*hw_limit < new_limit) {
162 return kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
165 void kvm_s390_clear_cmma_callback(void *opaque)
168 KVMState *s = opaque;
169 struct kvm_device_attr attr = {
170 .group = KVM_S390_VM_MEM_CTRL,
171 .attr = KVM_S390_VM_MEM_CLR_CMMA,
174 rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
175 trace_kvm_clear_cmma(rc);
178 static void kvm_s390_enable_cmma(KVMState *s)
181 struct kvm_device_attr attr = {
182 .group = KVM_S390_VM_MEM_CTRL,
183 .attr = KVM_S390_VM_MEM_ENABLE_CMMA,
186 if (!kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_ENABLE_CMMA) ||
187 !kvm_vm_check_mem_attr(s, KVM_S390_VM_MEM_CLR_CMMA)) {
191 rc = kvm_vm_ioctl(s, KVM_SET_DEVICE_ATTR, &attr);
193 qemu_register_reset(kvm_s390_clear_cmma_callback, s);
195 trace_kvm_enable_cmma(rc);
198 static void kvm_s390_set_attr(uint64_t attr)
200 struct kvm_device_attr attribute = {
201 .group = KVM_S390_VM_CRYPTO,
205 int ret = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attribute);
208 error_report("Failed to set crypto device attribute %lu: %s",
209 attr, strerror(-ret));
213 static void kvm_s390_init_aes_kw(void)
215 uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_AES_KW;
217 if (object_property_get_bool(OBJECT(qdev_get_machine()), "aes-key-wrap",
219 attr = KVM_S390_VM_CRYPTO_ENABLE_AES_KW;
222 if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
223 kvm_s390_set_attr(attr);
227 static void kvm_s390_init_dea_kw(void)
229 uint64_t attr = KVM_S390_VM_CRYPTO_DISABLE_DEA_KW;
231 if (object_property_get_bool(OBJECT(qdev_get_machine()), "dea-key-wrap",
233 attr = KVM_S390_VM_CRYPTO_ENABLE_DEA_KW;
236 if (kvm_vm_check_attr(kvm_state, KVM_S390_VM_CRYPTO, attr)) {
237 kvm_s390_set_attr(attr);
241 static void kvm_s390_init_crypto(void)
243 kvm_s390_init_aes_kw();
244 kvm_s390_init_dea_kw();
247 int kvm_arch_init(MachineState *ms, KVMState *s)
249 cap_sync_regs = kvm_check_extension(s, KVM_CAP_SYNC_REGS);
250 cap_async_pf = kvm_check_extension(s, KVM_CAP_ASYNC_PF);
251 cap_mem_op = kvm_check_extension(s, KVM_CAP_S390_MEM_OP);
253 kvm_s390_enable_cmma(s);
255 if (!kvm_check_extension(s, KVM_CAP_S390_GMAP)
256 || !kvm_check_extension(s, KVM_CAP_S390_COW)) {
257 phys_mem_set_alloc(legacy_s390_alloc);
260 kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
261 kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
266 unsigned long kvm_arch_vcpu_id(CPUState *cpu)
268 return cpu->cpu_index;
271 int kvm_arch_init_vcpu(CPUState *cs)
273 S390CPU *cpu = S390_CPU(cs);
274 kvm_s390_set_cpu_state(cpu, cpu->env.cpu_state);
278 void kvm_s390_reset_vcpu(S390CPU *cpu)
280 CPUState *cs = CPU(cpu);
282 /* The initial reset call is needed here to reset in-kernel
283 * vcpu data that we can't access directly from QEMU
284 * (i.e. with older kernels which don't support sync_regs/ONE_REG).
285 * Before this ioctl cpu_synchronize_state() is called in common kvm
287 if (kvm_vcpu_ioctl(cs, KVM_S390_INITIAL_RESET, NULL)) {
288 error_report("Initial CPU reset failed on CPU %i", cs->cpu_index);
291 kvm_s390_init_crypto();
294 static int can_sync_regs(CPUState *cs, int regs)
296 return cap_sync_regs && (cs->kvm_run->kvm_valid_regs & regs) == regs;
299 int kvm_arch_put_registers(CPUState *cs, int level)
301 S390CPU *cpu = S390_CPU(cs);
302 CPUS390XState *env = &cpu->env;
303 struct kvm_sregs sregs;
304 struct kvm_regs regs;
305 struct kvm_fpu fpu = {};
309 /* always save the PSW and the GPRS*/
310 cs->kvm_run->psw_addr = env->psw.addr;
311 cs->kvm_run->psw_mask = env->psw.mask;
313 if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
314 for (i = 0; i < 16; i++) {
315 cs->kvm_run->s.regs.gprs[i] = env->regs[i];
316 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_GPRS;
319 for (i = 0; i < 16; i++) {
320 regs.gprs[i] = env->regs[i];
322 r = kvm_vcpu_ioctl(cs, KVM_SET_REGS, ®s);
329 for (i = 0; i < 16; i++) {
330 fpu.fprs[i] = env->fregs[i].ll;
334 r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
339 /* Do we need to save more than that? */
340 if (level == KVM_PUT_RUNTIME_STATE) {
344 if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
345 cs->kvm_run->s.regs.cputm = env->cputm;
346 cs->kvm_run->s.regs.ckc = env->ckc;
347 cs->kvm_run->s.regs.todpr = env->todpr;
348 cs->kvm_run->s.regs.gbea = env->gbea;
349 cs->kvm_run->s.regs.pp = env->pp;
350 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ARCH0;
353 * These ONE_REGS are not protected by a capability. As they are only
354 * necessary for migration we just trace a possible error, but don't
355 * return with an error return code.
357 kvm_set_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
358 kvm_set_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
359 kvm_set_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
360 kvm_set_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
361 kvm_set_one_reg(cs, KVM_REG_S390_PP, &env->pp);
364 /* pfault parameters */
365 if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
366 cs->kvm_run->s.regs.pft = env->pfault_token;
367 cs->kvm_run->s.regs.pfs = env->pfault_select;
368 cs->kvm_run->s.regs.pfc = env->pfault_compare;
369 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PFAULT;
370 } else if (cap_async_pf) {
371 r = kvm_set_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
375 r = kvm_set_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
379 r = kvm_set_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
385 /* access registers and control registers*/
386 if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
387 for (i = 0; i < 16; i++) {
388 cs->kvm_run->s.regs.acrs[i] = env->aregs[i];
389 cs->kvm_run->s.regs.crs[i] = env->cregs[i];
391 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_ACRS;
392 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_CRS;
394 for (i = 0; i < 16; i++) {
395 sregs.acrs[i] = env->aregs[i];
396 sregs.crs[i] = env->cregs[i];
398 r = kvm_vcpu_ioctl(cs, KVM_SET_SREGS, &sregs);
404 /* Finally the prefix */
405 if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
406 cs->kvm_run->s.regs.prefix = env->psa;
407 cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_PREFIX;
409 /* prefix is only supported via sync regs */
414 int kvm_arch_get_registers(CPUState *cs)
416 S390CPU *cpu = S390_CPU(cs);
417 CPUS390XState *env = &cpu->env;
418 struct kvm_sregs sregs;
419 struct kvm_regs regs;
424 env->psw.addr = cs->kvm_run->psw_addr;
425 env->psw.mask = cs->kvm_run->psw_mask;
428 if (can_sync_regs(cs, KVM_SYNC_GPRS)) {
429 for (i = 0; i < 16; i++) {
430 env->regs[i] = cs->kvm_run->s.regs.gprs[i];
433 r = kvm_vcpu_ioctl(cs, KVM_GET_REGS, ®s);
437 for (i = 0; i < 16; i++) {
438 env->regs[i] = regs.gprs[i];
442 /* The ACRS and CRS */
443 if (can_sync_regs(cs, KVM_SYNC_ACRS | KVM_SYNC_CRS)) {
444 for (i = 0; i < 16; i++) {
445 env->aregs[i] = cs->kvm_run->s.regs.acrs[i];
446 env->cregs[i] = cs->kvm_run->s.regs.crs[i];
449 r = kvm_vcpu_ioctl(cs, KVM_GET_SREGS, &sregs);
453 for (i = 0; i < 16; i++) {
454 env->aregs[i] = sregs.acrs[i];
455 env->cregs[i] = sregs.crs[i];
460 r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
464 for (i = 0; i < 16; i++) {
465 env->fregs[i].ll = fpu.fprs[i];
470 if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
471 env->psa = cs->kvm_run->s.regs.prefix;
474 if (can_sync_regs(cs, KVM_SYNC_ARCH0)) {
475 env->cputm = cs->kvm_run->s.regs.cputm;
476 env->ckc = cs->kvm_run->s.regs.ckc;
477 env->todpr = cs->kvm_run->s.regs.todpr;
478 env->gbea = cs->kvm_run->s.regs.gbea;
479 env->pp = cs->kvm_run->s.regs.pp;
482 * These ONE_REGS are not protected by a capability. As they are only
483 * necessary for migration we just trace a possible error, but don't
484 * return with an error return code.
486 kvm_get_one_reg(cs, KVM_REG_S390_CPU_TIMER, &env->cputm);
487 kvm_get_one_reg(cs, KVM_REG_S390_CLOCK_COMP, &env->ckc);
488 kvm_get_one_reg(cs, KVM_REG_S390_TODPR, &env->todpr);
489 kvm_get_one_reg(cs, KVM_REG_S390_GBEA, &env->gbea);
490 kvm_get_one_reg(cs, KVM_REG_S390_PP, &env->pp);
493 /* pfault parameters */
494 if (can_sync_regs(cs, KVM_SYNC_PFAULT)) {
495 env->pfault_token = cs->kvm_run->s.regs.pft;
496 env->pfault_select = cs->kvm_run->s.regs.pfs;
497 env->pfault_compare = cs->kvm_run->s.regs.pfc;
498 } else if (cap_async_pf) {
499 r = kvm_get_one_reg(cs, KVM_REG_S390_PFTOKEN, &env->pfault_token);
503 r = kvm_get_one_reg(cs, KVM_REG_S390_PFCOMPARE, &env->pfault_compare);
507 r = kvm_get_one_reg(cs, KVM_REG_S390_PFSELECT, &env->pfault_select);
516 int kvm_s390_get_clock(uint8_t *tod_high, uint64_t *tod_low)
519 struct kvm_device_attr attr = {
520 .group = KVM_S390_VM_TOD,
521 .attr = KVM_S390_VM_TOD_LOW,
522 .addr = (uint64_t)tod_low,
525 r = kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
530 attr.attr = KVM_S390_VM_TOD_HIGH;
531 attr.addr = (uint64_t)tod_high;
532 return kvm_vm_ioctl(kvm_state, KVM_GET_DEVICE_ATTR, &attr);
535 int kvm_s390_set_clock(uint8_t *tod_high, uint64_t *tod_low)
539 struct kvm_device_attr attr = {
540 .group = KVM_S390_VM_TOD,
541 .attr = KVM_S390_VM_TOD_LOW,
542 .addr = (uint64_t)tod_low,
545 r = kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
550 attr.attr = KVM_S390_VM_TOD_HIGH;
551 attr.addr = (uint64_t)tod_high;
552 return kvm_vm_ioctl(kvm_state, KVM_SET_DEVICE_ATTR, &attr);
557 * @addr: the logical start address in guest memory
558 * @hostbuf: buffer in host memory. NULL = do only checks w/o copying
559 * @len: length that should be transfered
560 * @is_write: true = write, false = read
561 * Returns: 0 on success, non-zero if an exception or error occured
563 * Use KVM ioctl to read/write from/to guest memory. An access exception
564 * is injected into the vCPU in case of translation errors.
566 int kvm_s390_mem_op(S390CPU *cpu, vaddr addr, void *hostbuf, int len,
569 struct kvm_s390_mem_op mem_op = {
571 .flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION,
573 .op = is_write ? KVM_S390_MEMOP_LOGICAL_WRITE
574 : KVM_S390_MEMOP_LOGICAL_READ,
575 .buf = (uint64_t)hostbuf,
583 mem_op.flags |= KVM_S390_MEMOP_F_CHECK_ONLY;
586 ret = kvm_vcpu_ioctl(CPU(cpu), KVM_S390_MEM_OP, &mem_op);
588 error_printf("KVM_S390_MEM_OP failed: %s\n", strerror(-ret));
594 * Legacy layout for s390:
595 * Older S390 KVM requires the topmost vma of the RAM to be
596 * smaller than an system defined value, which is at least 256GB.
597 * Larger systems have larger values. We put the guest between
598 * the end of data segment (system break) and this value. We
599 * use 32GB as a base to have enough room for the system break
600 * to grow. We also have to use MAP parameters that avoid
601 * read-only mapping of guest pages.
603 static void *legacy_s390_alloc(size_t size, uint64_t *align)
607 mem = mmap((void *) 0x800000000ULL, size,
608 PROT_EXEC|PROT_READ|PROT_WRITE,
609 MAP_SHARED | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
610 return mem == MAP_FAILED ? NULL : mem;
613 /* DIAG 501 is used for sw breakpoints */
614 static const uint8_t diag_501[] = {0x83, 0x24, 0x05, 0x01};
616 int kvm_arch_insert_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
619 if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
620 sizeof(diag_501), 0) ||
621 cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)diag_501,
622 sizeof(diag_501), 1)) {
628 int kvm_arch_remove_sw_breakpoint(CPUState *cs, struct kvm_sw_breakpoint *bp)
630 uint8_t t[sizeof(diag_501)];
632 if (cpu_memory_rw_debug(cs, bp->pc, t, sizeof(diag_501), 0)) {
634 } else if (memcmp(t, diag_501, sizeof(diag_501))) {
636 } else if (cpu_memory_rw_debug(cs, bp->pc, (uint8_t *)&bp->saved_insn,
637 sizeof(diag_501), 1)) {
644 static struct kvm_hw_breakpoint *find_hw_breakpoint(target_ulong addr,
649 for (n = 0; n < nb_hw_breakpoints; n++) {
650 if (hw_breakpoints[n].addr == addr && hw_breakpoints[n].type == type &&
651 (hw_breakpoints[n].len == len || len == -1)) {
652 return &hw_breakpoints[n];
659 static int insert_hw_breakpoint(target_ulong addr, int len, int type)
663 if (find_hw_breakpoint(addr, len, type)) {
667 size = (nb_hw_breakpoints + 1) * sizeof(struct kvm_hw_breakpoint);
669 if (!hw_breakpoints) {
670 nb_hw_breakpoints = 0;
671 hw_breakpoints = (struct kvm_hw_breakpoint *)g_try_malloc(size);
674 (struct kvm_hw_breakpoint *)g_try_realloc(hw_breakpoints, size);
677 if (!hw_breakpoints) {
678 nb_hw_breakpoints = 0;
682 hw_breakpoints[nb_hw_breakpoints].addr = addr;
683 hw_breakpoints[nb_hw_breakpoints].len = len;
684 hw_breakpoints[nb_hw_breakpoints].type = type;
691 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
692 target_ulong len, int type)
695 case GDB_BREAKPOINT_HW:
698 case GDB_WATCHPOINT_WRITE:
702 type = KVM_HW_WP_WRITE;
707 return insert_hw_breakpoint(addr, len, type);
710 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
711 target_ulong len, int type)
714 struct kvm_hw_breakpoint *bp = find_hw_breakpoint(addr, len, type);
721 if (nb_hw_breakpoints > 0) {
723 * In order to trim the array, move the last element to the position to
724 * be removed - if necessary.
726 if (bp != &hw_breakpoints[nb_hw_breakpoints]) {
727 *bp = hw_breakpoints[nb_hw_breakpoints];
729 size = nb_hw_breakpoints * sizeof(struct kvm_hw_breakpoint);
731 (struct kvm_hw_breakpoint *)g_realloc(hw_breakpoints, size);
733 g_free(hw_breakpoints);
734 hw_breakpoints = NULL;
740 void kvm_arch_remove_all_hw_breakpoints(void)
742 nb_hw_breakpoints = 0;
743 g_free(hw_breakpoints);
744 hw_breakpoints = NULL;
747 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
751 if (nb_hw_breakpoints > 0) {
752 dbg->arch.nr_hw_bp = nb_hw_breakpoints;
753 dbg->arch.hw_bp = hw_breakpoints;
755 for (i = 0; i < nb_hw_breakpoints; ++i) {
756 hw_breakpoints[i].phys_addr = s390_cpu_get_phys_addr_debug(cpu,
757 hw_breakpoints[i].addr);
759 dbg->control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_HW_BP;
761 dbg->arch.nr_hw_bp = 0;
762 dbg->arch.hw_bp = NULL;
766 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
770 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run)
774 int kvm_arch_process_async_events(CPUState *cs)
779 static int s390_kvm_irq_to_interrupt(struct kvm_s390_irq *irq,
780 struct kvm_s390_interrupt *interrupt)
784 interrupt->type = irq->type;
786 case KVM_S390_INT_VIRTIO:
787 interrupt->parm = irq->u.ext.ext_params;
789 case KVM_S390_INT_PFAULT_INIT:
790 case KVM_S390_INT_PFAULT_DONE:
791 interrupt->parm64 = irq->u.ext.ext_params2;
793 case KVM_S390_PROGRAM_INT:
794 interrupt->parm = irq->u.pgm.code;
796 case KVM_S390_SIGP_SET_PREFIX:
797 interrupt->parm = irq->u.prefix.address;
799 case KVM_S390_INT_SERVICE:
800 interrupt->parm = irq->u.ext.ext_params;
803 interrupt->parm = irq->u.mchk.cr14;
804 interrupt->parm64 = irq->u.mchk.mcic;
806 case KVM_S390_INT_EXTERNAL_CALL:
807 interrupt->parm = irq->u.extcall.code;
809 case KVM_S390_INT_EMERGENCY:
810 interrupt->parm = irq->u.emerg.code;
812 case KVM_S390_SIGP_STOP:
813 case KVM_S390_RESTART:
814 break; /* These types have no parameters */
815 case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
816 interrupt->parm = irq->u.io.subchannel_id << 16;
817 interrupt->parm |= irq->u.io.subchannel_nr;
818 interrupt->parm64 = (uint64_t)irq->u.io.io_int_parm << 32;
819 interrupt->parm64 |= irq->u.io.io_int_word;
828 void kvm_s390_vcpu_interrupt(S390CPU *cpu, struct kvm_s390_irq *irq)
830 struct kvm_s390_interrupt kvmint = {};
831 CPUState *cs = CPU(cpu);
834 r = s390_kvm_irq_to_interrupt(irq, &kvmint);
836 fprintf(stderr, "%s called with bogus interrupt\n", __func__);
840 r = kvm_vcpu_ioctl(cs, KVM_S390_INTERRUPT, &kvmint);
842 fprintf(stderr, "KVM failed to inject interrupt\n");
847 static void __kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
849 struct kvm_s390_interrupt kvmint = {};
852 r = s390_kvm_irq_to_interrupt(irq, &kvmint);
854 fprintf(stderr, "%s called with bogus interrupt\n", __func__);
858 r = kvm_vm_ioctl(kvm_state, KVM_S390_INTERRUPT, &kvmint);
860 fprintf(stderr, "KVM failed to inject interrupt\n");
865 void kvm_s390_floating_interrupt(struct kvm_s390_irq *irq)
867 static bool use_flic = true;
871 r = kvm_s390_inject_flic(irq);
879 __kvm_s390_floating_interrupt(irq);
882 void kvm_s390_virtio_irq(int config_change, uint64_t token)
884 struct kvm_s390_irq irq = {
885 .type = KVM_S390_INT_VIRTIO,
886 .u.ext.ext_params = config_change,
887 .u.ext.ext_params2 = token,
890 kvm_s390_floating_interrupt(&irq);
893 void kvm_s390_service_interrupt(uint32_t parm)
895 struct kvm_s390_irq irq = {
896 .type = KVM_S390_INT_SERVICE,
897 .u.ext.ext_params = parm,
900 kvm_s390_floating_interrupt(&irq);
903 static void enter_pgmcheck(S390CPU *cpu, uint16_t code)
905 struct kvm_s390_irq irq = {
906 .type = KVM_S390_PROGRAM_INT,
910 kvm_s390_vcpu_interrupt(cpu, &irq);
913 void kvm_s390_access_exception(S390CPU *cpu, uint16_t code, uint64_t te_code)
915 struct kvm_s390_irq irq = {
916 .type = KVM_S390_PROGRAM_INT,
918 .u.pgm.trans_exc_code = te_code,
919 .u.pgm.exc_access_id = te_code & 3,
922 kvm_s390_vcpu_interrupt(cpu, &irq);
925 static int kvm_sclp_service_call(S390CPU *cpu, struct kvm_run *run,
928 CPUS390XState *env = &cpu->env;
933 cpu_synchronize_state(CPU(cpu));
934 sccb = env->regs[ipbh0 & 0xf];
935 code = env->regs[(ipbh0 & 0xf0) >> 4];
937 r = sclp_service_call(env, sccb, code);
939 enter_pgmcheck(cpu, -r);
947 static int handle_b2(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
949 CPUS390XState *env = &cpu->env;
951 uint16_t ipbh0 = (run->s390_sieic.ipb & 0xffff0000) >> 16;
953 cpu_synchronize_state(CPU(cpu));
957 ioinst_handle_xsch(cpu, env->regs[1]);
960 ioinst_handle_csch(cpu, env->regs[1]);
963 ioinst_handle_hsch(cpu, env->regs[1]);
966 ioinst_handle_msch(cpu, env->regs[1], run->s390_sieic.ipb);
969 ioinst_handle_ssch(cpu, env->regs[1], run->s390_sieic.ipb);
972 ioinst_handle_stcrw(cpu, run->s390_sieic.ipb);
975 ioinst_handle_stsch(cpu, env->regs[1], run->s390_sieic.ipb);
978 /* We should only get tsch via KVM_EXIT_S390_TSCH. */
979 fprintf(stderr, "Spurious tsch intercept\n");
982 ioinst_handle_chsc(cpu, run->s390_sieic.ipb);
985 /* This should have been handled by kvm already. */
986 fprintf(stderr, "Spurious tpi intercept\n");
989 ioinst_handle_schm(cpu, env->regs[1], env->regs[2],
990 run->s390_sieic.ipb);
993 ioinst_handle_rsch(cpu, env->regs[1]);
996 ioinst_handle_rchp(cpu, env->regs[1]);
999 /* We do not provide this instruction, it is suppressed. */
1002 ioinst_handle_sal(cpu, env->regs[1]);
1005 /* Not provided, set CC = 3 for subchannel not operational */
1008 case PRIV_B2_SCLP_CALL:
1009 rc = kvm_sclp_service_call(cpu, run, ipbh0);
1013 DPRINTF("KVM: unhandled PRIV: 0xb2%x\n", ipa1);
1020 static uint64_t get_base_disp_rxy(S390CPU *cpu, struct kvm_run *run)
1022 CPUS390XState *env = &cpu->env;
1023 uint32_t x2 = (run->s390_sieic.ipa & 0x000f);
1024 uint32_t base2 = run->s390_sieic.ipb >> 28;
1025 uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1026 ((run->s390_sieic.ipb & 0xff00) << 4);
1028 if (disp2 & 0x80000) {
1029 disp2 += 0xfff00000;
1032 return (base2 ? env->regs[base2] : 0) +
1033 (x2 ? env->regs[x2] : 0) + (long)(int)disp2;
1036 static uint64_t get_base_disp_rsy(S390CPU *cpu, struct kvm_run *run)
1038 CPUS390XState *env = &cpu->env;
1039 uint32_t base2 = run->s390_sieic.ipb >> 28;
1040 uint32_t disp2 = ((run->s390_sieic.ipb & 0x0fff0000) >> 16) +
1041 ((run->s390_sieic.ipb & 0xff00) << 4);
1043 if (disp2 & 0x80000) {
1044 disp2 += 0xfff00000;
1047 return (base2 ? env->regs[base2] : 0) + (long)(int)disp2;
1050 static int kvm_clp_service_call(S390CPU *cpu, struct kvm_run *run)
1052 uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1054 return clp_service_call(cpu, r2);
1057 static int kvm_pcilg_service_call(S390CPU *cpu, struct kvm_run *run)
1059 uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1060 uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1062 return pcilg_service_call(cpu, r1, r2);
1065 static int kvm_pcistg_service_call(S390CPU *cpu, struct kvm_run *run)
1067 uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1068 uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1070 return pcistg_service_call(cpu, r1, r2);
1073 static int kvm_stpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1075 uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1078 cpu_synchronize_state(CPU(cpu));
1079 fiba = get_base_disp_rxy(cpu, run);
1081 return stpcifc_service_call(cpu, r1, fiba);
1084 static int kvm_sic_service_call(S390CPU *cpu, struct kvm_run *run)
1090 static int kvm_rpcit_service_call(S390CPU *cpu, struct kvm_run *run)
1092 uint8_t r1 = (run->s390_sieic.ipb & 0x00f00000) >> 20;
1093 uint8_t r2 = (run->s390_sieic.ipb & 0x000f0000) >> 16;
1095 return rpcit_service_call(cpu, r1, r2);
1098 static int kvm_pcistb_service_call(S390CPU *cpu, struct kvm_run *run)
1100 uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1101 uint8_t r3 = run->s390_sieic.ipa & 0x000f;
1104 cpu_synchronize_state(CPU(cpu));
1105 gaddr = get_base_disp_rsy(cpu, run);
1107 return pcistb_service_call(cpu, r1, r3, gaddr);
1110 static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run)
1112 uint8_t r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1115 cpu_synchronize_state(CPU(cpu));
1116 fiba = get_base_disp_rxy(cpu, run);
1118 return mpcifc_service_call(cpu, r1, fiba);
1121 static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1127 r = kvm_clp_service_call(cpu, run);
1129 case PRIV_B9_PCISTG:
1130 r = kvm_pcistg_service_call(cpu, run);
1133 r = kvm_pcilg_service_call(cpu, run);
1136 r = kvm_rpcit_service_call(cpu, run);
1139 /* just inject exception */
1144 DPRINTF("KVM: unhandled PRIV: 0xb9%x\n", ipa1);
1151 static int handle_eb(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1156 case PRIV_EB_PCISTB:
1157 r = kvm_pcistb_service_call(cpu, run);
1160 r = kvm_sic_service_call(cpu, run);
1163 /* just inject exception */
1168 DPRINTF("KVM: unhandled PRIV: 0xeb%x\n", ipbl);
1175 static int handle_e3(S390CPU *cpu, struct kvm_run *run, uint8_t ipbl)
1180 case PRIV_E3_MPCIFC:
1181 r = kvm_mpcifc_service_call(cpu, run);
1183 case PRIV_E3_STPCIFC:
1184 r = kvm_stpcifc_service_call(cpu, run);
1188 DPRINTF("KVM: unhandled PRIV: 0xe3%x\n", ipbl);
1195 static int handle_hypercall(S390CPU *cpu, struct kvm_run *run)
1197 CPUS390XState *env = &cpu->env;
1200 cpu_synchronize_state(CPU(cpu));
1201 ret = s390_virtio_hypercall(env);
1202 if (ret == -EINVAL) {
1203 enter_pgmcheck(cpu, PGM_SPECIFICATION);
1210 static void kvm_handle_diag_308(S390CPU *cpu, struct kvm_run *run)
1214 cpu_synchronize_state(CPU(cpu));
1215 r1 = (run->s390_sieic.ipa & 0x00f0) >> 4;
1216 r3 = run->s390_sieic.ipa & 0x000f;
1217 handle_diag_308(&cpu->env, r1, r3);
1220 static int handle_sw_breakpoint(S390CPU *cpu, struct kvm_run *run)
1222 CPUS390XState *env = &cpu->env;
1225 cpu_synchronize_state(CPU(cpu));
1227 pc = env->psw.addr - 4;
1228 if (kvm_find_sw_breakpoint(CPU(cpu), pc)) {
1236 #define DIAG_KVM_CODE_MASK 0x000000000000ffff
1238 static int handle_diag(S390CPU *cpu, struct kvm_run *run, uint32_t ipb)
1244 * For any diagnose call we support, bits 48-63 of the resulting
1245 * address specify the function code; the remainder is ignored.
1247 func_code = decode_basedisp_rs(&cpu->env, ipb) & DIAG_KVM_CODE_MASK;
1248 switch (func_code) {
1250 kvm_handle_diag_308(cpu, run);
1252 case DIAG_KVM_HYPERCALL:
1253 r = handle_hypercall(cpu, run);
1255 case DIAG_KVM_BREAKPOINT:
1256 r = handle_sw_breakpoint(cpu, run);
1259 DPRINTF("KVM: unknown DIAG: 0x%x\n", func_code);
1260 enter_pgmcheck(cpu, PGM_SPECIFICATION);
1267 typedef struct SigpInfo {
1271 uint64_t *status_reg;
1274 static void set_sigp_status(SigpInfo *si, uint64_t status)
1276 *si->status_reg &= 0xffffffff00000000ULL;
1277 *si->status_reg |= status;
1278 si->cc = SIGP_CC_STATUS_STORED;
1281 static void sigp_start(void *arg)
1285 if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1286 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1290 s390_cpu_set_state(CPU_STATE_OPERATING, si->cpu);
1291 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1294 static void sigp_stop(void *arg)
1297 struct kvm_s390_irq irq = {
1298 .type = KVM_S390_SIGP_STOP,
1301 if (s390_cpu_get_state(si->cpu) != CPU_STATE_OPERATING) {
1302 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1306 /* disabled wait - sleeping in user space */
1307 if (CPU(si->cpu)->halted) {
1308 s390_cpu_set_state(CPU_STATE_STOPPED, si->cpu);
1310 /* execute the stop function */
1311 si->cpu->env.sigp_order = SIGP_STOP;
1312 kvm_s390_vcpu_interrupt(si->cpu, &irq);
1314 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1317 #define KVM_S390_STORE_STATUS_DEF_ADDR offsetof(LowCore, floating_pt_save_area)
1318 #define SAVE_AREA_SIZE 512
1319 static int kvm_s390_store_status(S390CPU *cpu, hwaddr addr, bool store_arch)
1321 static const uint8_t ar_id = 1;
1322 uint64_t ckc = cpu->env.ckc >> 8;
1324 hwaddr len = SAVE_AREA_SIZE;
1326 mem = cpu_physical_memory_map(addr, &len, 1);
1330 if (len != SAVE_AREA_SIZE) {
1331 cpu_physical_memory_unmap(mem, len, 1, 0);
1336 cpu_physical_memory_write(offsetof(LowCore, ar_access_id), &ar_id, 1);
1338 memcpy(mem, &cpu->env.fregs, 128);
1339 memcpy(mem + 128, &cpu->env.regs, 128);
1340 memcpy(mem + 256, &cpu->env.psw, 16);
1341 memcpy(mem + 280, &cpu->env.psa, 4);
1342 memcpy(mem + 284, &cpu->env.fpc, 4);
1343 memcpy(mem + 292, &cpu->env.todpr, 4);
1344 memcpy(mem + 296, &cpu->env.cputm, 8);
1345 memcpy(mem + 304, &ckc, 8);
1346 memcpy(mem + 320, &cpu->env.aregs, 64);
1347 memcpy(mem + 384, &cpu->env.cregs, 128);
1349 cpu_physical_memory_unmap(mem, len, 1, len);
1354 static void sigp_stop_and_store_status(void *arg)
1357 struct kvm_s390_irq irq = {
1358 .type = KVM_S390_SIGP_STOP,
1361 /* disabled wait - sleeping in user space */
1362 if (s390_cpu_get_state(si->cpu) == CPU_STATE_OPERATING &&
1363 CPU(si->cpu)->halted) {
1364 s390_cpu_set_state(CPU_STATE_STOPPED, si->cpu);
1367 switch (s390_cpu_get_state(si->cpu)) {
1368 case CPU_STATE_OPERATING:
1369 si->cpu->env.sigp_order = SIGP_STOP_STORE_STATUS;
1370 kvm_s390_vcpu_interrupt(si->cpu, &irq);
1371 /* store will be performed when handling the stop intercept */
1373 case CPU_STATE_STOPPED:
1374 /* already stopped, just store the status */
1375 cpu_synchronize_state(CPU(si->cpu));
1376 kvm_s390_store_status(si->cpu, KVM_S390_STORE_STATUS_DEF_ADDR, true);
1379 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1382 static void sigp_store_status_at_address(void *arg)
1385 uint32_t address = si->param & 0x7ffffe00u;
1387 /* cpu has to be stopped */
1388 if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1389 set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
1393 cpu_synchronize_state(CPU(si->cpu));
1395 if (kvm_s390_store_status(si->cpu, address, false)) {
1396 set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1399 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1402 static void sigp_restart(void *arg)
1405 struct kvm_s390_irq irq = {
1406 .type = KVM_S390_RESTART,
1409 switch (s390_cpu_get_state(si->cpu)) {
1410 case CPU_STATE_STOPPED:
1411 /* the restart irq has to be delivered prior to any other pending irq */
1412 cpu_synchronize_state(CPU(si->cpu));
1413 do_restart_interrupt(&si->cpu->env);
1414 s390_cpu_set_state(CPU_STATE_OPERATING, si->cpu);
1416 case CPU_STATE_OPERATING:
1417 kvm_s390_vcpu_interrupt(si->cpu, &irq);
1420 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1423 int kvm_s390_cpu_restart(S390CPU *cpu)
1429 run_on_cpu(CPU(cpu), sigp_restart, &si);
1430 DPRINTF("DONE: KVM cpu restart: %p\n", &cpu->env);
1434 static void sigp_initial_cpu_reset(void *arg)
1437 CPUState *cs = CPU(si->cpu);
1438 S390CPUClass *scc = S390_CPU_GET_CLASS(si->cpu);
1440 cpu_synchronize_state(cs);
1441 scc->initial_cpu_reset(cs);
1442 cpu_synchronize_post_reset(cs);
1443 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1446 static void sigp_cpu_reset(void *arg)
1449 CPUState *cs = CPU(si->cpu);
1450 S390CPUClass *scc = S390_CPU_GET_CLASS(si->cpu);
1452 cpu_synchronize_state(cs);
1454 cpu_synchronize_post_reset(cs);
1455 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1458 static void sigp_set_prefix(void *arg)
1461 uint32_t addr = si->param & 0x7fffe000u;
1463 cpu_synchronize_state(CPU(si->cpu));
1465 if (!address_space_access_valid(&address_space_memory, addr,
1466 sizeof(struct LowCore), false)) {
1467 set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
1471 /* cpu has to be stopped */
1472 if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
1473 set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
1477 si->cpu->env.psa = addr;
1478 cpu_synchronize_post_init(CPU(si->cpu));
1479 si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
1482 static int handle_sigp_single_dst(S390CPU *dst_cpu, uint8_t order,
1483 uint64_t param, uint64_t *status_reg)
1488 .status_reg = status_reg,
1491 /* cpu available? */
1492 if (dst_cpu == NULL) {
1493 return SIGP_CC_NOT_OPERATIONAL;
1496 /* only resets can break pending orders */
1497 if (dst_cpu->env.sigp_order != 0 &&
1498 order != SIGP_CPU_RESET &&
1499 order != SIGP_INITIAL_CPU_RESET) {
1500 return SIGP_CC_BUSY;
1505 run_on_cpu(CPU(dst_cpu), sigp_start, &si);
1508 run_on_cpu(CPU(dst_cpu), sigp_stop, &si);
1511 run_on_cpu(CPU(dst_cpu), sigp_restart, &si);
1513 case SIGP_STOP_STORE_STATUS:
1514 run_on_cpu(CPU(dst_cpu), sigp_stop_and_store_status, &si);
1516 case SIGP_STORE_STATUS_ADDR:
1517 run_on_cpu(CPU(dst_cpu), sigp_store_status_at_address, &si);
1519 case SIGP_SET_PREFIX:
1520 run_on_cpu(CPU(dst_cpu), sigp_set_prefix, &si);
1522 case SIGP_INITIAL_CPU_RESET:
1523 run_on_cpu(CPU(dst_cpu), sigp_initial_cpu_reset, &si);
1525 case SIGP_CPU_RESET:
1526 run_on_cpu(CPU(dst_cpu), sigp_cpu_reset, &si);
1529 DPRINTF("KVM: unknown SIGP: 0x%x\n", order);
1530 set_sigp_status(&si, SIGP_STAT_INVALID_ORDER);
1536 static int sigp_set_architecture(S390CPU *cpu, uint32_t param,
1537 uint64_t *status_reg)
1542 /* due to the BQL, we are the only active cpu */
1543 CPU_FOREACH(cur_cs) {
1544 cur_cpu = S390_CPU(cur_cs);
1545 if (cur_cpu->env.sigp_order != 0) {
1546 return SIGP_CC_BUSY;
1548 cpu_synchronize_state(cur_cs);
1549 /* all but the current one have to be stopped */
1550 if (cur_cpu != cpu &&
1551 s390_cpu_get_state(cur_cpu) != CPU_STATE_STOPPED) {
1552 *status_reg &= 0xffffffff00000000ULL;
1553 *status_reg |= SIGP_STAT_INCORRECT_STATE;
1554 return SIGP_CC_STATUS_STORED;
1558 switch (param & 0xff) {
1559 case SIGP_MODE_ESA_S390:
1561 return SIGP_CC_NOT_OPERATIONAL;
1562 case SIGP_MODE_Z_ARCH_TRANS_ALL_PSW:
1563 case SIGP_MODE_Z_ARCH_TRANS_CUR_PSW:
1564 CPU_FOREACH(cur_cs) {
1565 cur_cpu = S390_CPU(cur_cs);
1566 cur_cpu->env.pfault_token = -1UL;
1570 *status_reg &= 0xffffffff00000000ULL;
1571 *status_reg |= SIGP_STAT_INVALID_PARAMETER;
1572 return SIGP_CC_STATUS_STORED;
1575 return SIGP_CC_ORDER_CODE_ACCEPTED;
1578 #define SIGP_ORDER_MASK 0x000000ff
1580 static int handle_sigp(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)
1582 CPUS390XState *env = &cpu->env;
1583 const uint8_t r1 = ipa1 >> 4;
1584 const uint8_t r3 = ipa1 & 0x0f;
1587 uint64_t *status_reg;
1589 S390CPU *dst_cpu = NULL;
1591 cpu_synchronize_state(CPU(cpu));
1593 /* get order code */
1594 order = decode_basedisp_rs(env, run->s390_sieic.ipb) & SIGP_ORDER_MASK;
1595 status_reg = &env->regs[r1];
1596 param = (r1 % 2) ? env->regs[r1] : env->regs[r1 + 1];
1600 ret = sigp_set_architecture(cpu, param, status_reg);
1603 /* all other sigp orders target a single vcpu */
1604 dst_cpu = s390_cpu_addr2state(env->regs[r3]);
1605 ret = handle_sigp_single_dst(dst_cpu, order, param, status_reg);
1608 trace_kvm_sigp_finished(order, CPU(cpu)->cpu_index,
1609 dst_cpu ? CPU(dst_cpu)->cpu_index : -1, ret);
1619 static int handle_instruction(S390CPU *cpu, struct kvm_run *run)
1621 unsigned int ipa0 = (run->s390_sieic.ipa & 0xff00);
1622 uint8_t ipa1 = run->s390_sieic.ipa & 0x00ff;
1625 DPRINTF("handle_instruction 0x%x 0x%x\n",
1626 run->s390_sieic.ipa, run->s390_sieic.ipb);
1629 r = handle_b2(cpu, run, ipa1);
1632 r = handle_b9(cpu, run, ipa1);
1635 r = handle_eb(cpu, run, run->s390_sieic.ipb & 0xff);
1638 r = handle_e3(cpu, run, run->s390_sieic.ipb & 0xff);
1641 r = handle_diag(cpu, run, run->s390_sieic.ipb);
1644 r = handle_sigp(cpu, run, ipa1);
1650 enter_pgmcheck(cpu, 0x0001);
1656 static bool is_special_wait_psw(CPUState *cs)
1658 /* signal quiesce */
1659 return cs->kvm_run->psw_addr == 0xfffUL;
1662 static void guest_panicked(void)
1664 qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_PAUSE,
1666 vm_stop(RUN_STATE_GUEST_PANICKED);
1669 static void unmanageable_intercept(S390CPU *cpu, const char *str, int pswoffset)
1671 CPUState *cs = CPU(cpu);
1673 error_report("Unmanageable %s! CPU%i new PSW: 0x%016lx:%016lx",
1674 str, cs->cpu_index, ldq_phys(cs->as, cpu->env.psa + pswoffset),
1675 ldq_phys(cs->as, cpu->env.psa + pswoffset + 8));
1680 static int handle_intercept(S390CPU *cpu)
1682 CPUState *cs = CPU(cpu);
1683 struct kvm_run *run = cs->kvm_run;
1684 int icpt_code = run->s390_sieic.icptcode;
1687 DPRINTF("intercept: 0x%x (at 0x%lx)\n", icpt_code,
1688 (long)cs->kvm_run->psw_addr);
1689 switch (icpt_code) {
1690 case ICPT_INSTRUCTION:
1691 r = handle_instruction(cpu, run);
1694 unmanageable_intercept(cpu, "program interrupt",
1695 offsetof(LowCore, program_new_psw));
1699 unmanageable_intercept(cpu, "external interrupt",
1700 offsetof(LowCore, external_new_psw));
1704 /* disabled wait, since enabled wait is handled in kernel */
1705 cpu_synchronize_state(cs);
1706 if (s390_cpu_halt(cpu) == 0) {
1707 if (is_special_wait_psw(cs)) {
1708 qemu_system_shutdown_request();
1716 if (s390_cpu_set_state(CPU_STATE_STOPPED, cpu) == 0) {
1717 qemu_system_shutdown_request();
1719 if (cpu->env.sigp_order == SIGP_STOP_STORE_STATUS) {
1720 kvm_s390_store_status(cpu, KVM_S390_STORE_STATUS_DEF_ADDR,
1723 cpu->env.sigp_order = 0;
1726 case ICPT_SOFT_INTERCEPT:
1727 fprintf(stderr, "KVM unimplemented icpt SOFT\n");
1731 fprintf(stderr, "KVM unimplemented icpt IO\n");
1735 fprintf(stderr, "Unknown intercept code: %d\n", icpt_code);
1743 static int handle_tsch(S390CPU *cpu)
1745 CPUState *cs = CPU(cpu);
1746 struct kvm_run *run = cs->kvm_run;
1749 cpu_synchronize_state(cs);
1751 ret = ioinst_handle_tsch(cpu, cpu->env.regs[1], run->s390_tsch.ipb);
1755 * If an I/O interrupt had been dequeued, we have to reinject it.
1757 if (run->s390_tsch.dequeued) {
1758 kvm_s390_io_interrupt(run->s390_tsch.subchannel_id,
1759 run->s390_tsch.subchannel_nr,
1760 run->s390_tsch.io_int_parm,
1761 run->s390_tsch.io_int_word);
1768 static void insert_stsi_3_2_2(S390CPU *cpu, __u64 addr)
1770 struct sysib_322 sysib;
1773 if (s390_cpu_virt_mem_read(cpu, addr, &sysib, sizeof(sysib))) {
1776 /* Shift the stack of Extended Names to prepare for our own data */
1777 memmove(&sysib.ext_names[1], &sysib.ext_names[0],
1778 sizeof(sysib.ext_names[0]) * (sysib.count - 1));
1779 /* First virt level, that doesn't provide Ext Names delimits stack. It is
1780 * assumed it's not capable of managing Extended Names for lower levels.
1782 for (del = 1; del < sysib.count; del++) {
1783 if (!sysib.vm[del].ext_name_encoding || !sysib.ext_names[del][0]) {
1787 if (del < sysib.count) {
1788 memset(sysib.ext_names[del], 0,
1789 sizeof(sysib.ext_names[0]) * (sysib.count - del));
1791 /* Insert short machine name in EBCDIC, padded with blanks */
1793 memset(sysib.vm[0].name, 0x40, sizeof(sysib.vm[0].name));
1794 ebcdic_put(sysib.vm[0].name, qemu_name, MIN(sizeof(sysib.vm[0].name),
1795 strlen(qemu_name)));
1797 sysib.vm[0].ext_name_encoding = 2; /* 2 = UTF-8 */
1798 memset(sysib.ext_names[0], 0, sizeof(sysib.ext_names[0]));
1799 /* If hypervisor specifies zero Extended Name in STSI322 SYSIB, it's
1800 * considered by s390 as not capable of providing any Extended Name.
1801 * Therefore if no name was specified on qemu invocation, we go with the
1802 * same "KVMguest" default, which KVM has filled into short name field.
1805 strncpy((char *)sysib.ext_names[0], qemu_name,
1806 sizeof(sysib.ext_names[0]));
1808 strcpy((char *)sysib.ext_names[0], "KVMguest");
1811 memcpy(sysib.vm[0].uuid, qemu_uuid, sizeof(sysib.vm[0].uuid));
1813 s390_cpu_virt_mem_write(cpu, addr, &sysib, sizeof(sysib));
1816 static int handle_stsi(S390CPU *cpu)
1818 CPUState *cs = CPU(cpu);
1819 struct kvm_run *run = cs->kvm_run;
1821 switch (run->s390_stsi.fc) {
1823 if (run->s390_stsi.sel1 != 2 || run->s390_stsi.sel2 != 2) {
1826 /* Only sysib 3.2.2 needs post-handling for now. */
1827 insert_stsi_3_2_2(cpu, run->s390_stsi.addr);
1834 static int kvm_arch_handle_debug_exit(S390CPU *cpu)
1836 CPUState *cs = CPU(cpu);
1837 struct kvm_run *run = cs->kvm_run;
1840 struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
1842 switch (arch_info->type) {
1843 case KVM_HW_WP_WRITE:
1844 if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1845 cs->watchpoint_hit = &hw_watchpoint;
1846 hw_watchpoint.vaddr = arch_info->addr;
1847 hw_watchpoint.flags = BP_MEM_WRITE;
1852 if (find_hw_breakpoint(arch_info->addr, -1, arch_info->type)) {
1856 case KVM_SINGLESTEP:
1857 if (cs->singlestep_enabled) {
1868 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
1870 S390CPU *cpu = S390_CPU(cs);
1873 switch (run->exit_reason) {
1874 case KVM_EXIT_S390_SIEIC:
1875 ret = handle_intercept(cpu);
1877 case KVM_EXIT_S390_RESET:
1878 s390_reipl_request();
1880 case KVM_EXIT_S390_TSCH:
1881 ret = handle_tsch(cpu);
1883 case KVM_EXIT_S390_STSI:
1884 ret = handle_stsi(cpu);
1886 case KVM_EXIT_DEBUG:
1887 ret = kvm_arch_handle_debug_exit(cpu);
1890 fprintf(stderr, "Unknown KVM exit: %d\n", run->exit_reason);
1895 ret = EXCP_INTERRUPT;
1900 bool kvm_arch_stop_on_emulation_error(CPUState *cpu)
1905 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr)
1910 int kvm_arch_on_sigbus(int code, void *addr)
1915 void kvm_s390_io_interrupt(uint16_t subchannel_id,
1916 uint16_t subchannel_nr, uint32_t io_int_parm,
1917 uint32_t io_int_word)
1919 struct kvm_s390_irq irq = {
1920 .u.io.subchannel_id = subchannel_id,
1921 .u.io.subchannel_nr = subchannel_nr,
1922 .u.io.io_int_parm = io_int_parm,
1923 .u.io.io_int_word = io_int_word,
1926 if (io_int_word & IO_INT_WORD_AI) {
1927 irq.type = KVM_S390_INT_IO(1, 0, 0, 0);
1929 irq.type = ((subchannel_id & 0xff00) << 24) |
1930 ((subchannel_id & 0x00060) << 22) | (subchannel_nr << 16);
1932 kvm_s390_floating_interrupt(&irq);
1935 void kvm_s390_crw_mchk(void)
1937 struct kvm_s390_irq irq = {
1938 .type = KVM_S390_MCHK,
1939 .u.mchk.cr14 = 1 << 28,
1940 .u.mchk.mcic = 0x00400f1d40330000ULL,
1942 kvm_s390_floating_interrupt(&irq);
1945 void kvm_s390_enable_css_support(S390CPU *cpu)
1949 /* Activate host kernel channel subsystem support. */
1950 r = kvm_vcpu_enable_cap(CPU(cpu), KVM_CAP_S390_CSS_SUPPORT, 0);
1954 void kvm_arch_init_irq_routing(KVMState *s)
1957 * Note that while irqchip capabilities generally imply that cpustates
1958 * are handled in-kernel, it is not true for s390 (yet); therefore, we
1959 * have to override the common code kvm_halt_in_kernel_allowed setting.
1961 if (kvm_check_extension(s, KVM_CAP_IRQ_ROUTING)) {
1962 kvm_gsi_routing_allowed = true;
1963 kvm_halt_in_kernel_allowed = false;
1967 int kvm_s390_assign_subch_ioeventfd(EventNotifier *notifier, uint32_t sch,
1968 int vq, bool assign)
1970 struct kvm_ioeventfd kick = {
1971 .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
1972 KVM_IOEVENTFD_FLAG_DATAMATCH,
1973 .fd = event_notifier_get_fd(notifier),
1978 if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
1982 kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
1984 return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
1987 int kvm_s390_get_memslot_count(KVMState *s)
1989 return kvm_check_extension(s, KVM_CAP_NR_MEMSLOTS);
1992 int kvm_s390_set_cpu_state(S390CPU *cpu, uint8_t cpu_state)
1994 struct kvm_mp_state mp_state = {};
1997 /* the kvm part might not have been initialized yet */
1998 if (CPU(cpu)->kvm_state == NULL) {
2002 switch (cpu_state) {
2003 case CPU_STATE_STOPPED:
2004 mp_state.mp_state = KVM_MP_STATE_STOPPED;
2006 case CPU_STATE_CHECK_STOP:
2007 mp_state.mp_state = KVM_MP_STATE_CHECK_STOP;
2009 case CPU_STATE_OPERATING:
2010 mp_state.mp_state = KVM_MP_STATE_OPERATING;
2012 case CPU_STATE_LOAD:
2013 mp_state.mp_state = KVM_MP_STATE_LOAD;
2016 error_report("Requested CPU state is not a valid S390 CPU state: %u",
2021 ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_MP_STATE, &mp_state);
2023 trace_kvm_failed_cpu_state_set(CPU(cpu)->cpu_index, cpu_state,
2030 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
2031 uint64_t address, uint32_t data)
2033 S390PCIBusDevice *pbdev;
2034 uint32_t fid = data >> ZPCI_MSI_VEC_BITS;
2035 uint32_t vec = data & ZPCI_MSI_VEC_MASK;
2037 pbdev = s390_pci_find_dev_by_fid(fid);
2039 DPRINTF("add_msi_route no dev\n");
2043 pbdev->routes.adapter.ind_offset = vec;
2045 route->type = KVM_IRQ_ROUTING_S390_ADAPTER;
2047 route->u.adapter.summary_addr = pbdev->routes.adapter.summary_addr;
2048 route->u.adapter.ind_addr = pbdev->routes.adapter.ind_addr;
2049 route->u.adapter.summary_offset = pbdev->routes.adapter.summary_offset;
2050 route->u.adapter.ind_offset = pbdev->routes.adapter.ind_offset;
2051 route->u.adapter.adapter_id = pbdev->routes.adapter.adapter_id;