1 // SPDX-License-Identifier: GPL-2.0
3 * hosting IBM Z kernel virtual machines (s390x)
5 * Copyright IBM Corp. 2008, 2020
7 * Author(s): Carsten Otte <cotte@de.ibm.com>
8 * Christian Borntraeger <borntraeger@de.ibm.com>
9 * Heiko Carstens <heiko.carstens@de.ibm.com>
10 * Christian Ehrhardt <ehrhardt@de.ibm.com>
11 * Jason J. Herne <jjherne@us.ibm.com>
14 #define KMSG_COMPONENT "kvm-s390"
15 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17 #include <linux/compiler.h>
18 #include <linux/err.h>
20 #include <linux/hrtimer.h>
21 #include <linux/init.h>
22 #include <linux/kvm.h>
23 #include <linux/kvm_host.h>
24 #include <linux/mman.h>
25 #include <linux/module.h>
26 #include <linux/moduleparam.h>
27 #include <linux/random.h>
28 #include <linux/slab.h>
29 #include <linux/timer.h>
30 #include <linux/vmalloc.h>
31 #include <linux/bitmap.h>
32 #include <linux/sched/signal.h>
33 #include <linux/string.h>
34 #include <linux/pgtable.h>
36 #include <asm/asm-offsets.h>
37 #include <asm/lowcore.h>
41 #include <asm/switch_to.h>
44 #include <asm/cpacf.h>
45 #include <asm/timex.h>
48 #include <asm/fpu/api.h>
52 #define CREATE_TRACE_POINTS
54 #include "trace-s390.h"
56 #define MEM_OP_MAX_SIZE 65536 /* Maximum transfer size for KVM_S390_MEM_OP */
58 #define VCPU_IRQS_MAX_BUF (sizeof(struct kvm_s390_irq) * \
59 (KVM_MAX_VCPUS + LOCAL_IRQS))
61 const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
62 KVM_GENERIC_VM_STATS(),
63 STATS_DESC_COUNTER(VM, inject_io),
64 STATS_DESC_COUNTER(VM, inject_float_mchk),
65 STATS_DESC_COUNTER(VM, inject_pfault_done),
66 STATS_DESC_COUNTER(VM, inject_service_signal),
67 STATS_DESC_COUNTER(VM, inject_virtio)
69 static_assert(ARRAY_SIZE(kvm_vm_stats_desc) ==
70 sizeof(struct kvm_vm_stat) / sizeof(u64));
72 const struct kvm_stats_header kvm_vm_stats_header = {
73 .name_size = KVM_STATS_NAME_SIZE,
74 .num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
75 .id_offset = sizeof(struct kvm_stats_header),
76 .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
77 .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
78 sizeof(kvm_vm_stats_desc),
81 const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
82 KVM_GENERIC_VCPU_STATS(),
83 STATS_DESC_COUNTER(VCPU, exit_userspace),
84 STATS_DESC_COUNTER(VCPU, exit_null),
85 STATS_DESC_COUNTER(VCPU, exit_external_request),
86 STATS_DESC_COUNTER(VCPU, exit_io_request),
87 STATS_DESC_COUNTER(VCPU, exit_external_interrupt),
88 STATS_DESC_COUNTER(VCPU, exit_stop_request),
89 STATS_DESC_COUNTER(VCPU, exit_validity),
90 STATS_DESC_COUNTER(VCPU, exit_instruction),
91 STATS_DESC_COUNTER(VCPU, exit_pei),
92 STATS_DESC_COUNTER(VCPU, halt_no_poll_steal),
93 STATS_DESC_COUNTER(VCPU, instruction_lctl),
94 STATS_DESC_COUNTER(VCPU, instruction_lctlg),
95 STATS_DESC_COUNTER(VCPU, instruction_stctl),
96 STATS_DESC_COUNTER(VCPU, instruction_stctg),
97 STATS_DESC_COUNTER(VCPU, exit_program_interruption),
98 STATS_DESC_COUNTER(VCPU, exit_instr_and_program),
99 STATS_DESC_COUNTER(VCPU, exit_operation_exception),
100 STATS_DESC_COUNTER(VCPU, deliver_ckc),
101 STATS_DESC_COUNTER(VCPU, deliver_cputm),
102 STATS_DESC_COUNTER(VCPU, deliver_external_call),
103 STATS_DESC_COUNTER(VCPU, deliver_emergency_signal),
104 STATS_DESC_COUNTER(VCPU, deliver_service_signal),
105 STATS_DESC_COUNTER(VCPU, deliver_virtio),
106 STATS_DESC_COUNTER(VCPU, deliver_stop_signal),
107 STATS_DESC_COUNTER(VCPU, deliver_prefix_signal),
108 STATS_DESC_COUNTER(VCPU, deliver_restart_signal),
109 STATS_DESC_COUNTER(VCPU, deliver_program),
110 STATS_DESC_COUNTER(VCPU, deliver_io),
111 STATS_DESC_COUNTER(VCPU, deliver_machine_check),
112 STATS_DESC_COUNTER(VCPU, exit_wait_state),
113 STATS_DESC_COUNTER(VCPU, inject_ckc),
114 STATS_DESC_COUNTER(VCPU, inject_cputm),
115 STATS_DESC_COUNTER(VCPU, inject_external_call),
116 STATS_DESC_COUNTER(VCPU, inject_emergency_signal),
117 STATS_DESC_COUNTER(VCPU, inject_mchk),
118 STATS_DESC_COUNTER(VCPU, inject_pfault_init),
119 STATS_DESC_COUNTER(VCPU, inject_program),
120 STATS_DESC_COUNTER(VCPU, inject_restart),
121 STATS_DESC_COUNTER(VCPU, inject_set_prefix),
122 STATS_DESC_COUNTER(VCPU, inject_stop_signal),
123 STATS_DESC_COUNTER(VCPU, instruction_epsw),
124 STATS_DESC_COUNTER(VCPU, instruction_gs),
125 STATS_DESC_COUNTER(VCPU, instruction_io_other),
126 STATS_DESC_COUNTER(VCPU, instruction_lpsw),
127 STATS_DESC_COUNTER(VCPU, instruction_lpswe),
128 STATS_DESC_COUNTER(VCPU, instruction_pfmf),
129 STATS_DESC_COUNTER(VCPU, instruction_ptff),
130 STATS_DESC_COUNTER(VCPU, instruction_sck),
131 STATS_DESC_COUNTER(VCPU, instruction_sckpf),
132 STATS_DESC_COUNTER(VCPU, instruction_stidp),
133 STATS_DESC_COUNTER(VCPU, instruction_spx),
134 STATS_DESC_COUNTER(VCPU, instruction_stpx),
135 STATS_DESC_COUNTER(VCPU, instruction_stap),
136 STATS_DESC_COUNTER(VCPU, instruction_iske),
137 STATS_DESC_COUNTER(VCPU, instruction_ri),
138 STATS_DESC_COUNTER(VCPU, instruction_rrbe),
139 STATS_DESC_COUNTER(VCPU, instruction_sske),
140 STATS_DESC_COUNTER(VCPU, instruction_ipte_interlock),
141 STATS_DESC_COUNTER(VCPU, instruction_stsi),
142 STATS_DESC_COUNTER(VCPU, instruction_stfl),
143 STATS_DESC_COUNTER(VCPU, instruction_tb),
144 STATS_DESC_COUNTER(VCPU, instruction_tpi),
145 STATS_DESC_COUNTER(VCPU, instruction_tprot),
146 STATS_DESC_COUNTER(VCPU, instruction_tsch),
147 STATS_DESC_COUNTER(VCPU, instruction_sie),
148 STATS_DESC_COUNTER(VCPU, instruction_essa),
149 STATS_DESC_COUNTER(VCPU, instruction_sthyi),
150 STATS_DESC_COUNTER(VCPU, instruction_sigp_sense),
151 STATS_DESC_COUNTER(VCPU, instruction_sigp_sense_running),
152 STATS_DESC_COUNTER(VCPU, instruction_sigp_external_call),
153 STATS_DESC_COUNTER(VCPU, instruction_sigp_emergency),
154 STATS_DESC_COUNTER(VCPU, instruction_sigp_cond_emergency),
155 STATS_DESC_COUNTER(VCPU, instruction_sigp_start),
156 STATS_DESC_COUNTER(VCPU, instruction_sigp_stop),
157 STATS_DESC_COUNTER(VCPU, instruction_sigp_stop_store_status),
158 STATS_DESC_COUNTER(VCPU, instruction_sigp_store_status),
159 STATS_DESC_COUNTER(VCPU, instruction_sigp_store_adtl_status),
160 STATS_DESC_COUNTER(VCPU, instruction_sigp_arch),
161 STATS_DESC_COUNTER(VCPU, instruction_sigp_prefix),
162 STATS_DESC_COUNTER(VCPU, instruction_sigp_restart),
163 STATS_DESC_COUNTER(VCPU, instruction_sigp_init_cpu_reset),
164 STATS_DESC_COUNTER(VCPU, instruction_sigp_cpu_reset),
165 STATS_DESC_COUNTER(VCPU, instruction_sigp_unknown),
166 STATS_DESC_COUNTER(VCPU, instruction_diagnose_10),
167 STATS_DESC_COUNTER(VCPU, instruction_diagnose_44),
168 STATS_DESC_COUNTER(VCPU, instruction_diagnose_9c),
169 STATS_DESC_COUNTER(VCPU, diag_9c_ignored),
170 STATS_DESC_COUNTER(VCPU, diag_9c_forward),
171 STATS_DESC_COUNTER(VCPU, instruction_diagnose_258),
172 STATS_DESC_COUNTER(VCPU, instruction_diagnose_308),
173 STATS_DESC_COUNTER(VCPU, instruction_diagnose_500),
174 STATS_DESC_COUNTER(VCPU, instruction_diagnose_other),
175 STATS_DESC_COUNTER(VCPU, pfault_sync)
177 static_assert(ARRAY_SIZE(kvm_vcpu_stats_desc) ==
178 sizeof(struct kvm_vcpu_stat) / sizeof(u64));
180 const struct kvm_stats_header kvm_vcpu_stats_header = {
181 .name_size = KVM_STATS_NAME_SIZE,
182 .num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
183 .id_offset = sizeof(struct kvm_stats_header),
184 .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
185 .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
186 sizeof(kvm_vcpu_stats_desc),
189 /* allow nested virtualization in KVM (if enabled by user space) */
191 module_param(nested, int, S_IRUGO);
192 MODULE_PARM_DESC(nested, "Nested virtualization support");
194 /* allow 1m huge page guest backing, if !nested */
196 module_param(hpage, int, 0444);
197 MODULE_PARM_DESC(hpage, "1m huge page backing support");
199 /* maximum percentage of steal time for polling. >100 is treated like 100 */
200 static u8 halt_poll_max_steal = 10;
201 module_param(halt_poll_max_steal, byte, 0644);
202 MODULE_PARM_DESC(halt_poll_max_steal, "Maximum percentage of steal time to allow polling");
204 /* if set to true, the GISA will be initialized and used if available */
205 static bool use_gisa = true;
206 module_param(use_gisa, bool, 0644);
207 MODULE_PARM_DESC(use_gisa, "Use the GISA if the host supports it.");
209 /* maximum diag9c forwarding per second */
210 unsigned int diag9c_forwarding_hz;
211 module_param(diag9c_forwarding_hz, uint, 0644);
212 MODULE_PARM_DESC(diag9c_forwarding_hz, "Maximum diag9c forwarding per second, 0 to turn off");
215 * For now we handle at most 16 double words as this is what the s390 base
216 * kernel handles and stores in the prefix page. If we ever need to go beyond
217 * this, this requires changes to code, but the external uapi can stay.
219 #define SIZE_INTERNAL 16
222 * Base feature mask that defines default mask for facilities. Consists of the
223 * defines in FACILITIES_KVM and the non-hypervisor managed bits.
225 static unsigned long kvm_s390_fac_base[SIZE_INTERNAL] = { FACILITIES_KVM };
227 * Extended feature mask. Consists of the defines in FACILITIES_KVM_CPUMODEL
228 * and defines the facilities that can be enabled via a cpu model.
230 static unsigned long kvm_s390_fac_ext[SIZE_INTERNAL] = { FACILITIES_KVM_CPUMODEL };
232 static unsigned long kvm_s390_fac_size(void)
234 BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_MASK_SIZE_U64);
235 BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_LIST_SIZE_U64);
236 BUILD_BUG_ON(SIZE_INTERNAL * sizeof(unsigned long) >
237 sizeof(stfle_fac_list));
239 return SIZE_INTERNAL;
242 /* available cpu features supported by kvm */
243 static DECLARE_BITMAP(kvm_s390_available_cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
244 /* available subfunctions indicated via query / "test bit" */
245 static struct kvm_s390_vm_cpu_subfunc kvm_s390_available_subfunc;
247 static struct gmap_notifier gmap_notifier;
248 static struct gmap_notifier vsie_gmap_notifier;
249 debug_info_t *kvm_s390_dbf;
250 debug_info_t *kvm_s390_dbf_uv;
252 /* Section: not file related */
253 int kvm_arch_hardware_enable(void)
255 /* every s390 is virtualization enabled ;-) */
259 int kvm_arch_check_processor_compat(void *opaque)
264 /* forward declarations */
265 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
267 static int sca_switch_to_extended(struct kvm *kvm);
269 static void kvm_clock_sync_scb(struct kvm_s390_sie_block *scb, u64 delta)
274 * The TOD jumps by delta, we have to compensate this by adding
275 * -delta to the epoch.
279 /* sign-extension - we're adding to signed values below */
284 if (scb->ecd & ECD_MEF) {
285 scb->epdx += delta_idx;
286 if (scb->epoch < delta)
292 * This callback is executed during stop_machine(). All CPUs are therefore
293 * temporarily stopped. In order not to change guest behavior, we have to
294 * disable preemption whenever we touch the epoch of kvm and the VCPUs,
295 * so a CPU won't be stopped while calculating with the epoch.
297 static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
301 struct kvm_vcpu *vcpu;
303 unsigned long long *delta = v;
305 list_for_each_entry(kvm, &vm_list, vm_list) {
306 kvm_for_each_vcpu(i, vcpu, kvm) {
307 kvm_clock_sync_scb(vcpu->arch.sie_block, *delta);
309 kvm->arch.epoch = vcpu->arch.sie_block->epoch;
310 kvm->arch.epdx = vcpu->arch.sie_block->epdx;
312 if (vcpu->arch.cputm_enabled)
313 vcpu->arch.cputm_start += *delta;
314 if (vcpu->arch.vsie_block)
315 kvm_clock_sync_scb(vcpu->arch.vsie_block,
322 static struct notifier_block kvm_clock_notifier = {
323 .notifier_call = kvm_clock_sync,
326 int kvm_arch_hardware_setup(void *opaque)
328 gmap_notifier.notifier_call = kvm_gmap_notifier;
329 gmap_register_pte_notifier(&gmap_notifier);
330 vsie_gmap_notifier.notifier_call = kvm_s390_vsie_gmap_notifier;
331 gmap_register_pte_notifier(&vsie_gmap_notifier);
332 atomic_notifier_chain_register(&s390_epoch_delta_notifier,
333 &kvm_clock_notifier);
337 void kvm_arch_hardware_unsetup(void)
339 gmap_unregister_pte_notifier(&gmap_notifier);
340 gmap_unregister_pte_notifier(&vsie_gmap_notifier);
341 atomic_notifier_chain_unregister(&s390_epoch_delta_notifier,
342 &kvm_clock_notifier);
345 static void allow_cpu_feat(unsigned long nr)
347 set_bit_inv(nr, kvm_s390_available_cpu_feat);
350 static inline int plo_test_bit(unsigned char nr)
352 unsigned long function = (unsigned long)nr | 0x100;
356 " lgr 0,%[function]\n"
357 /* Parameter registers are ignored for "test bit" */
362 : [function] "d" (function)
367 static __always_inline void __insn32_query(unsigned int opcode, u8 *query)
372 /* Parameter registers are ignored */
373 " .insn rrf,%[opc] << 16,2,4,6,0\n"
375 : [query] "d" ((unsigned long)query), [opc] "i" (opcode)
376 : "cc", "memory", "0", "1");
379 #define INSN_SORTL 0xb938
380 #define INSN_DFLTCC 0xb939
382 static void kvm_s390_cpu_feat_init(void)
386 for (i = 0; i < 256; ++i) {
388 kvm_s390_available_subfunc.plo[i >> 3] |= 0x80 >> (i & 7);
391 if (test_facility(28)) /* TOD-clock steering */
392 ptff(kvm_s390_available_subfunc.ptff,
393 sizeof(kvm_s390_available_subfunc.ptff),
396 if (test_facility(17)) { /* MSA */
397 __cpacf_query(CPACF_KMAC, (cpacf_mask_t *)
398 kvm_s390_available_subfunc.kmac);
399 __cpacf_query(CPACF_KMC, (cpacf_mask_t *)
400 kvm_s390_available_subfunc.kmc);
401 __cpacf_query(CPACF_KM, (cpacf_mask_t *)
402 kvm_s390_available_subfunc.km);
403 __cpacf_query(CPACF_KIMD, (cpacf_mask_t *)
404 kvm_s390_available_subfunc.kimd);
405 __cpacf_query(CPACF_KLMD, (cpacf_mask_t *)
406 kvm_s390_available_subfunc.klmd);
408 if (test_facility(76)) /* MSA3 */
409 __cpacf_query(CPACF_PCKMO, (cpacf_mask_t *)
410 kvm_s390_available_subfunc.pckmo);
411 if (test_facility(77)) { /* MSA4 */
412 __cpacf_query(CPACF_KMCTR, (cpacf_mask_t *)
413 kvm_s390_available_subfunc.kmctr);
414 __cpacf_query(CPACF_KMF, (cpacf_mask_t *)
415 kvm_s390_available_subfunc.kmf);
416 __cpacf_query(CPACF_KMO, (cpacf_mask_t *)
417 kvm_s390_available_subfunc.kmo);
418 __cpacf_query(CPACF_PCC, (cpacf_mask_t *)
419 kvm_s390_available_subfunc.pcc);
421 if (test_facility(57)) /* MSA5 */
422 __cpacf_query(CPACF_PRNO, (cpacf_mask_t *)
423 kvm_s390_available_subfunc.ppno);
425 if (test_facility(146)) /* MSA8 */
426 __cpacf_query(CPACF_KMA, (cpacf_mask_t *)
427 kvm_s390_available_subfunc.kma);
429 if (test_facility(155)) /* MSA9 */
430 __cpacf_query(CPACF_KDSA, (cpacf_mask_t *)
431 kvm_s390_available_subfunc.kdsa);
433 if (test_facility(150)) /* SORTL */
434 __insn32_query(INSN_SORTL, kvm_s390_available_subfunc.sortl);
436 if (test_facility(151)) /* DFLTCC */
437 __insn32_query(INSN_DFLTCC, kvm_s390_available_subfunc.dfltcc);
439 if (MACHINE_HAS_ESOP)
440 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP);
442 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
443 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
445 if (!sclp.has_sief2 || !MACHINE_HAS_ESOP || !sclp.has_64bscao ||
446 !test_facility(3) || !nested)
448 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2);
449 if (sclp.has_64bscao)
450 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO);
452 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF);
454 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE);
456 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS);
458 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB);
460 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI);
462 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS);
464 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_KSS);
466 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
467 * all skey handling functions read/set the skey from the PGSTE
468 * instead of the real storage key.
470 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
471 * pages being detected as preserved although they are resident.
473 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
474 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
476 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
477 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
478 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
480 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
481 * cannot easily shadow the SCA because of the ipte lock.
485 int kvm_arch_init(void *opaque)
489 kvm_s390_dbf = debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
493 kvm_s390_dbf_uv = debug_register("kvm-uv", 32, 1, 7 * sizeof(long));
494 if (!kvm_s390_dbf_uv)
497 if (debug_register_view(kvm_s390_dbf, &debug_sprintf_view) ||
498 debug_register_view(kvm_s390_dbf_uv, &debug_sprintf_view))
501 kvm_s390_cpu_feat_init();
503 /* Register floating interrupt controller interface. */
504 rc = kvm_register_device_ops(&kvm_flic_ops, KVM_DEV_TYPE_FLIC);
506 pr_err("A FLIC registration call failed with rc=%d\n", rc);
510 rc = kvm_s390_gib_init(GAL_ISC);
521 void kvm_arch_exit(void)
523 kvm_s390_gib_destroy();
524 debug_unregister(kvm_s390_dbf);
525 debug_unregister(kvm_s390_dbf_uv);
528 /* Section: device related */
529 long kvm_arch_dev_ioctl(struct file *filp,
530 unsigned int ioctl, unsigned long arg)
532 if (ioctl == KVM_S390_ENABLE_SIE)
533 return s390_enable_sie();
537 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
542 case KVM_CAP_S390_PSW:
543 case KVM_CAP_S390_GMAP:
544 case KVM_CAP_SYNC_MMU:
545 #ifdef CONFIG_KVM_S390_UCONTROL
546 case KVM_CAP_S390_UCONTROL:
548 case KVM_CAP_ASYNC_PF:
549 case KVM_CAP_SYNC_REGS:
550 case KVM_CAP_ONE_REG:
551 case KVM_CAP_ENABLE_CAP:
552 case KVM_CAP_S390_CSS_SUPPORT:
553 case KVM_CAP_IOEVENTFD:
554 case KVM_CAP_DEVICE_CTRL:
555 case KVM_CAP_S390_IRQCHIP:
556 case KVM_CAP_VM_ATTRIBUTES:
557 case KVM_CAP_MP_STATE:
558 case KVM_CAP_IMMEDIATE_EXIT:
559 case KVM_CAP_S390_INJECT_IRQ:
560 case KVM_CAP_S390_USER_SIGP:
561 case KVM_CAP_S390_USER_STSI:
562 case KVM_CAP_S390_SKEYS:
563 case KVM_CAP_S390_IRQ_STATE:
564 case KVM_CAP_S390_USER_INSTR0:
565 case KVM_CAP_S390_CMMA_MIGRATION:
566 case KVM_CAP_S390_AIS:
567 case KVM_CAP_S390_AIS_MIGRATION:
568 case KVM_CAP_S390_VCPU_RESETS:
569 case KVM_CAP_SET_GUEST_DEBUG:
570 case KVM_CAP_S390_DIAG318:
573 case KVM_CAP_SET_GUEST_DEBUG2:
574 r = KVM_GUESTDBG_VALID_MASK;
576 case KVM_CAP_S390_HPAGE_1M:
578 if (hpage && !kvm_is_ucontrol(kvm))
581 case KVM_CAP_S390_MEM_OP:
584 case KVM_CAP_NR_VCPUS:
585 case KVM_CAP_MAX_VCPUS:
586 case KVM_CAP_MAX_VCPU_ID:
587 r = KVM_S390_BSCA_CPU_SLOTS;
588 if (!kvm_s390_use_sca_entries())
590 else if (sclp.has_esca && sclp.has_64bscao)
591 r = KVM_S390_ESCA_CPU_SLOTS;
593 case KVM_CAP_S390_COW:
594 r = MACHINE_HAS_ESOP;
596 case KVM_CAP_S390_VECTOR_REGISTERS:
599 case KVM_CAP_S390_RI:
600 r = test_facility(64);
602 case KVM_CAP_S390_GS:
603 r = test_facility(133);
605 case KVM_CAP_S390_BPB:
606 r = test_facility(82);
608 case KVM_CAP_S390_PROTECTED:
609 r = is_prot_virt_host();
617 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
620 gfn_t cur_gfn, last_gfn;
621 unsigned long gaddr, vmaddr;
622 struct gmap *gmap = kvm->arch.gmap;
623 DECLARE_BITMAP(bitmap, _PAGE_ENTRIES);
625 /* Loop over all guest segments */
626 cur_gfn = memslot->base_gfn;
627 last_gfn = memslot->base_gfn + memslot->npages;
628 for (; cur_gfn <= last_gfn; cur_gfn += _PAGE_ENTRIES) {
629 gaddr = gfn_to_gpa(cur_gfn);
630 vmaddr = gfn_to_hva_memslot(memslot, cur_gfn);
631 if (kvm_is_error_hva(vmaddr))
634 bitmap_zero(bitmap, _PAGE_ENTRIES);
635 gmap_sync_dirty_log_pmd(gmap, bitmap, gaddr, vmaddr);
636 for (i = 0; i < _PAGE_ENTRIES; i++) {
637 if (test_bit(i, bitmap))
638 mark_page_dirty(kvm, cur_gfn + i);
641 if (fatal_signal_pending(current))
647 /* Section: vm related */
648 static void sca_del_vcpu(struct kvm_vcpu *vcpu);
651 * Get (and clear) the dirty memory log for a memory slot.
653 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
654 struct kvm_dirty_log *log)
658 struct kvm_memory_slot *memslot;
661 if (kvm_is_ucontrol(kvm))
664 mutex_lock(&kvm->slots_lock);
667 if (log->slot >= KVM_USER_MEM_SLOTS)
670 r = kvm_get_dirty_log(kvm, log, &is_dirty, &memslot);
674 /* Clear the dirty log */
676 n = kvm_dirty_bitmap_bytes(memslot);
677 memset(memslot->dirty_bitmap, 0, n);
681 mutex_unlock(&kvm->slots_lock);
685 static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
688 struct kvm_vcpu *vcpu;
690 kvm_for_each_vcpu(i, vcpu, kvm) {
691 kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
695 int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
703 case KVM_CAP_S390_IRQCHIP:
704 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
705 kvm->arch.use_irqchip = 1;
708 case KVM_CAP_S390_USER_SIGP:
709 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
710 kvm->arch.user_sigp = 1;
713 case KVM_CAP_S390_VECTOR_REGISTERS:
714 mutex_lock(&kvm->lock);
715 if (kvm->created_vcpus) {
717 } else if (MACHINE_HAS_VX) {
718 set_kvm_facility(kvm->arch.model.fac_mask, 129);
719 set_kvm_facility(kvm->arch.model.fac_list, 129);
720 if (test_facility(134)) {
721 set_kvm_facility(kvm->arch.model.fac_mask, 134);
722 set_kvm_facility(kvm->arch.model.fac_list, 134);
724 if (test_facility(135)) {
725 set_kvm_facility(kvm->arch.model.fac_mask, 135);
726 set_kvm_facility(kvm->arch.model.fac_list, 135);
728 if (test_facility(148)) {
729 set_kvm_facility(kvm->arch.model.fac_mask, 148);
730 set_kvm_facility(kvm->arch.model.fac_list, 148);
732 if (test_facility(152)) {
733 set_kvm_facility(kvm->arch.model.fac_mask, 152);
734 set_kvm_facility(kvm->arch.model.fac_list, 152);
736 if (test_facility(192)) {
737 set_kvm_facility(kvm->arch.model.fac_mask, 192);
738 set_kvm_facility(kvm->arch.model.fac_list, 192);
743 mutex_unlock(&kvm->lock);
744 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
745 r ? "(not available)" : "(success)");
747 case KVM_CAP_S390_RI:
749 mutex_lock(&kvm->lock);
750 if (kvm->created_vcpus) {
752 } else if (test_facility(64)) {
753 set_kvm_facility(kvm->arch.model.fac_mask, 64);
754 set_kvm_facility(kvm->arch.model.fac_list, 64);
757 mutex_unlock(&kvm->lock);
758 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_RI %s",
759 r ? "(not available)" : "(success)");
761 case KVM_CAP_S390_AIS:
762 mutex_lock(&kvm->lock);
763 if (kvm->created_vcpus) {
766 set_kvm_facility(kvm->arch.model.fac_mask, 72);
767 set_kvm_facility(kvm->arch.model.fac_list, 72);
770 mutex_unlock(&kvm->lock);
771 VM_EVENT(kvm, 3, "ENABLE: AIS %s",
772 r ? "(not available)" : "(success)");
774 case KVM_CAP_S390_GS:
776 mutex_lock(&kvm->lock);
777 if (kvm->created_vcpus) {
779 } else if (test_facility(133)) {
780 set_kvm_facility(kvm->arch.model.fac_mask, 133);
781 set_kvm_facility(kvm->arch.model.fac_list, 133);
784 mutex_unlock(&kvm->lock);
785 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_GS %s",
786 r ? "(not available)" : "(success)");
788 case KVM_CAP_S390_HPAGE_1M:
789 mutex_lock(&kvm->lock);
790 if (kvm->created_vcpus)
792 else if (!hpage || kvm->arch.use_cmma || kvm_is_ucontrol(kvm))
796 mmap_write_lock(kvm->mm);
797 kvm->mm->context.allow_gmap_hpage_1m = 1;
798 mmap_write_unlock(kvm->mm);
800 * We might have to create fake 4k page
801 * tables. To avoid that the hardware works on
802 * stale PGSTEs, we emulate these instructions.
804 kvm->arch.use_skf = 0;
805 kvm->arch.use_pfmfi = 0;
807 mutex_unlock(&kvm->lock);
808 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_HPAGE %s",
809 r ? "(not available)" : "(success)");
811 case KVM_CAP_S390_USER_STSI:
812 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
813 kvm->arch.user_stsi = 1;
816 case KVM_CAP_S390_USER_INSTR0:
817 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
818 kvm->arch.user_instr0 = 1;
819 icpt_operexc_on_all_vcpus(kvm);
829 static int kvm_s390_get_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
833 switch (attr->attr) {
834 case KVM_S390_VM_MEM_LIMIT_SIZE:
836 VM_EVENT(kvm, 3, "QUERY: max guest memory: %lu bytes",
837 kvm->arch.mem_limit);
838 if (put_user(kvm->arch.mem_limit, (u64 __user *)attr->addr))
848 static int kvm_s390_set_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
852 switch (attr->attr) {
853 case KVM_S390_VM_MEM_ENABLE_CMMA:
858 VM_EVENT(kvm, 3, "%s", "ENABLE: CMMA support");
859 mutex_lock(&kvm->lock);
860 if (kvm->created_vcpus)
862 else if (kvm->mm->context.allow_gmap_hpage_1m)
865 kvm->arch.use_cmma = 1;
866 /* Not compatible with cmma. */
867 kvm->arch.use_pfmfi = 0;
870 mutex_unlock(&kvm->lock);
872 case KVM_S390_VM_MEM_CLR_CMMA:
877 if (!kvm->arch.use_cmma)
880 VM_EVENT(kvm, 3, "%s", "RESET: CMMA states");
881 mutex_lock(&kvm->lock);
882 idx = srcu_read_lock(&kvm->srcu);
883 s390_reset_cmma(kvm->arch.gmap->mm);
884 srcu_read_unlock(&kvm->srcu, idx);
885 mutex_unlock(&kvm->lock);
888 case KVM_S390_VM_MEM_LIMIT_SIZE: {
889 unsigned long new_limit;
891 if (kvm_is_ucontrol(kvm))
894 if (get_user(new_limit, (u64 __user *)attr->addr))
897 if (kvm->arch.mem_limit != KVM_S390_NO_MEM_LIMIT &&
898 new_limit > kvm->arch.mem_limit)
904 /* gmap_create takes last usable address */
905 if (new_limit != KVM_S390_NO_MEM_LIMIT)
909 mutex_lock(&kvm->lock);
910 if (!kvm->created_vcpus) {
911 /* gmap_create will round the limit up */
912 struct gmap *new = gmap_create(current->mm, new_limit);
917 gmap_remove(kvm->arch.gmap);
919 kvm->arch.gmap = new;
923 mutex_unlock(&kvm->lock);
924 VM_EVENT(kvm, 3, "SET: max guest address: %lu", new_limit);
925 VM_EVENT(kvm, 3, "New guest asce: 0x%pK",
926 (void *) kvm->arch.gmap->asce);
936 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
938 void kvm_s390_vcpu_crypto_reset_all(struct kvm *kvm)
940 struct kvm_vcpu *vcpu;
943 kvm_s390_vcpu_block_all(kvm);
945 kvm_for_each_vcpu(i, vcpu, kvm) {
946 kvm_s390_vcpu_crypto_setup(vcpu);
947 /* recreate the shadow crycb by leaving the VSIE handler */
948 kvm_s390_sync_request(KVM_REQ_VSIE_RESTART, vcpu);
951 kvm_s390_vcpu_unblock_all(kvm);
954 static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
956 mutex_lock(&kvm->lock);
957 switch (attr->attr) {
958 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
959 if (!test_kvm_facility(kvm, 76)) {
960 mutex_unlock(&kvm->lock);
964 kvm->arch.crypto.crycb->aes_wrapping_key_mask,
965 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
966 kvm->arch.crypto.aes_kw = 1;
967 VM_EVENT(kvm, 3, "%s", "ENABLE: AES keywrapping support");
969 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
970 if (!test_kvm_facility(kvm, 76)) {
971 mutex_unlock(&kvm->lock);
975 kvm->arch.crypto.crycb->dea_wrapping_key_mask,
976 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
977 kvm->arch.crypto.dea_kw = 1;
978 VM_EVENT(kvm, 3, "%s", "ENABLE: DEA keywrapping support");
980 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
981 if (!test_kvm_facility(kvm, 76)) {
982 mutex_unlock(&kvm->lock);
985 kvm->arch.crypto.aes_kw = 0;
986 memset(kvm->arch.crypto.crycb->aes_wrapping_key_mask, 0,
987 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
988 VM_EVENT(kvm, 3, "%s", "DISABLE: AES keywrapping support");
990 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
991 if (!test_kvm_facility(kvm, 76)) {
992 mutex_unlock(&kvm->lock);
995 kvm->arch.crypto.dea_kw = 0;
996 memset(kvm->arch.crypto.crycb->dea_wrapping_key_mask, 0,
997 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
998 VM_EVENT(kvm, 3, "%s", "DISABLE: DEA keywrapping support");
1000 case KVM_S390_VM_CRYPTO_ENABLE_APIE:
1001 if (!ap_instructions_available()) {
1002 mutex_unlock(&kvm->lock);
1005 kvm->arch.crypto.apie = 1;
1007 case KVM_S390_VM_CRYPTO_DISABLE_APIE:
1008 if (!ap_instructions_available()) {
1009 mutex_unlock(&kvm->lock);
1012 kvm->arch.crypto.apie = 0;
1015 mutex_unlock(&kvm->lock);
1019 kvm_s390_vcpu_crypto_reset_all(kvm);
1020 mutex_unlock(&kvm->lock);
1024 static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
1027 struct kvm_vcpu *vcpu;
1029 kvm_for_each_vcpu(cx, vcpu, kvm)
1030 kvm_s390_sync_request(req, vcpu);
1034 * Must be called with kvm->srcu held to avoid races on memslots, and with
1035 * kvm->slots_lock to avoid races with ourselves and kvm_s390_vm_stop_migration.
1037 static int kvm_s390_vm_start_migration(struct kvm *kvm)
1039 struct kvm_memory_slot *ms;
1040 struct kvm_memslots *slots;
1041 unsigned long ram_pages = 0;
1044 /* migration mode already enabled */
1045 if (kvm->arch.migration_mode)
1047 slots = kvm_memslots(kvm);
1048 if (!slots || !slots->used_slots)
1051 if (!kvm->arch.use_cmma) {
1052 kvm->arch.migration_mode = 1;
1055 /* mark all the pages in active slots as dirty */
1056 for (slotnr = 0; slotnr < slots->used_slots; slotnr++) {
1057 ms = slots->memslots + slotnr;
1058 if (!ms->dirty_bitmap)
1061 * The second half of the bitmap is only used on x86,
1062 * and would be wasted otherwise, so we put it to good
1063 * use here to keep track of the state of the storage
1066 memset(kvm_second_dirty_bitmap(ms), 0xff, kvm_dirty_bitmap_bytes(ms));
1067 ram_pages += ms->npages;
1069 atomic64_set(&kvm->arch.cmma_dirty_pages, ram_pages);
1070 kvm->arch.migration_mode = 1;
1071 kvm_s390_sync_request_broadcast(kvm, KVM_REQ_START_MIGRATION);
1076 * Must be called with kvm->slots_lock to avoid races with ourselves and
1077 * kvm_s390_vm_start_migration.
1079 static int kvm_s390_vm_stop_migration(struct kvm *kvm)
1081 /* migration mode already disabled */
1082 if (!kvm->arch.migration_mode)
1084 kvm->arch.migration_mode = 0;
1085 if (kvm->arch.use_cmma)
1086 kvm_s390_sync_request_broadcast(kvm, KVM_REQ_STOP_MIGRATION);
1090 static int kvm_s390_vm_set_migration(struct kvm *kvm,
1091 struct kvm_device_attr *attr)
1095 mutex_lock(&kvm->slots_lock);
1096 switch (attr->attr) {
1097 case KVM_S390_VM_MIGRATION_START:
1098 res = kvm_s390_vm_start_migration(kvm);
1100 case KVM_S390_VM_MIGRATION_STOP:
1101 res = kvm_s390_vm_stop_migration(kvm);
1106 mutex_unlock(&kvm->slots_lock);
1111 static int kvm_s390_vm_get_migration(struct kvm *kvm,
1112 struct kvm_device_attr *attr)
1114 u64 mig = kvm->arch.migration_mode;
1116 if (attr->attr != KVM_S390_VM_MIGRATION_STATUS)
1119 if (copy_to_user((void __user *)attr->addr, &mig, sizeof(mig)))
1124 static int kvm_s390_set_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1126 struct kvm_s390_vm_tod_clock gtod;
1128 if (copy_from_user(>od, (void __user *)attr->addr, sizeof(gtod)))
1131 if (!test_kvm_facility(kvm, 139) && gtod.epoch_idx)
1133 kvm_s390_set_tod_clock(kvm, >od);
1135 VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x, TOD base: 0x%llx",
1136 gtod.epoch_idx, gtod.tod);
1141 static int kvm_s390_set_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1145 if (copy_from_user(>od_high, (void __user *)attr->addr,
1151 VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x", gtod_high);
1156 static int kvm_s390_set_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1158 struct kvm_s390_vm_tod_clock gtod = { 0 };
1160 if (copy_from_user(>od.tod, (void __user *)attr->addr,
1164 kvm_s390_set_tod_clock(kvm, >od);
1165 VM_EVENT(kvm, 3, "SET: TOD base: 0x%llx", gtod.tod);
1169 static int kvm_s390_set_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1176 switch (attr->attr) {
1177 case KVM_S390_VM_TOD_EXT:
1178 ret = kvm_s390_set_tod_ext(kvm, attr);
1180 case KVM_S390_VM_TOD_HIGH:
1181 ret = kvm_s390_set_tod_high(kvm, attr);
1183 case KVM_S390_VM_TOD_LOW:
1184 ret = kvm_s390_set_tod_low(kvm, attr);
1193 static void kvm_s390_get_tod_clock(struct kvm *kvm,
1194 struct kvm_s390_vm_tod_clock *gtod)
1196 union tod_clock clk;
1200 store_tod_clock_ext(&clk);
1202 gtod->tod = clk.tod + kvm->arch.epoch;
1203 gtod->epoch_idx = 0;
1204 if (test_kvm_facility(kvm, 139)) {
1205 gtod->epoch_idx = clk.ei + kvm->arch.epdx;
1206 if (gtod->tod < clk.tod)
1207 gtod->epoch_idx += 1;
1213 static int kvm_s390_get_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1215 struct kvm_s390_vm_tod_clock gtod;
1217 memset(>od, 0, sizeof(gtod));
1218 kvm_s390_get_tod_clock(kvm, >od);
1219 if (copy_to_user((void __user *)attr->addr, >od, sizeof(gtod)))
1222 VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x, TOD base: 0x%llx",
1223 gtod.epoch_idx, gtod.tod);
1227 static int kvm_s390_get_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1231 if (copy_to_user((void __user *)attr->addr, >od_high,
1234 VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x", gtod_high);
1239 static int kvm_s390_get_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1243 gtod = kvm_s390_get_tod_clock_fast(kvm);
1244 if (copy_to_user((void __user *)attr->addr, >od, sizeof(gtod)))
1246 VM_EVENT(kvm, 3, "QUERY: TOD base: 0x%llx", gtod);
1251 static int kvm_s390_get_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1258 switch (attr->attr) {
1259 case KVM_S390_VM_TOD_EXT:
1260 ret = kvm_s390_get_tod_ext(kvm, attr);
1262 case KVM_S390_VM_TOD_HIGH:
1263 ret = kvm_s390_get_tod_high(kvm, attr);
1265 case KVM_S390_VM_TOD_LOW:
1266 ret = kvm_s390_get_tod_low(kvm, attr);
1275 static int kvm_s390_set_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1277 struct kvm_s390_vm_cpu_processor *proc;
1278 u16 lowest_ibc, unblocked_ibc;
1281 mutex_lock(&kvm->lock);
1282 if (kvm->created_vcpus) {
1286 proc = kzalloc(sizeof(*proc), GFP_KERNEL_ACCOUNT);
1291 if (!copy_from_user(proc, (void __user *)attr->addr,
1293 kvm->arch.model.cpuid = proc->cpuid;
1294 lowest_ibc = sclp.ibc >> 16 & 0xfff;
1295 unblocked_ibc = sclp.ibc & 0xfff;
1296 if (lowest_ibc && proc->ibc) {
1297 if (proc->ibc > unblocked_ibc)
1298 kvm->arch.model.ibc = unblocked_ibc;
1299 else if (proc->ibc < lowest_ibc)
1300 kvm->arch.model.ibc = lowest_ibc;
1302 kvm->arch.model.ibc = proc->ibc;
1304 memcpy(kvm->arch.model.fac_list, proc->fac_list,
1305 S390_ARCH_FAC_LIST_SIZE_BYTE);
1306 VM_EVENT(kvm, 3, "SET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1307 kvm->arch.model.ibc,
1308 kvm->arch.model.cpuid);
1309 VM_EVENT(kvm, 3, "SET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1310 kvm->arch.model.fac_list[0],
1311 kvm->arch.model.fac_list[1],
1312 kvm->arch.model.fac_list[2]);
1317 mutex_unlock(&kvm->lock);
1321 static int kvm_s390_set_processor_feat(struct kvm *kvm,
1322 struct kvm_device_attr *attr)
1324 struct kvm_s390_vm_cpu_feat data;
1326 if (copy_from_user(&data, (void __user *)attr->addr, sizeof(data)))
1328 if (!bitmap_subset((unsigned long *) data.feat,
1329 kvm_s390_available_cpu_feat,
1330 KVM_S390_VM_CPU_FEAT_NR_BITS))
1333 mutex_lock(&kvm->lock);
1334 if (kvm->created_vcpus) {
1335 mutex_unlock(&kvm->lock);
1338 bitmap_copy(kvm->arch.cpu_feat, (unsigned long *) data.feat,
1339 KVM_S390_VM_CPU_FEAT_NR_BITS);
1340 mutex_unlock(&kvm->lock);
1341 VM_EVENT(kvm, 3, "SET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1348 static int kvm_s390_set_processor_subfunc(struct kvm *kvm,
1349 struct kvm_device_attr *attr)
1351 mutex_lock(&kvm->lock);
1352 if (kvm->created_vcpus) {
1353 mutex_unlock(&kvm->lock);
1357 if (copy_from_user(&kvm->arch.model.subfuncs, (void __user *)attr->addr,
1358 sizeof(struct kvm_s390_vm_cpu_subfunc))) {
1359 mutex_unlock(&kvm->lock);
1362 mutex_unlock(&kvm->lock);
1364 VM_EVENT(kvm, 3, "SET: guest PLO subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1365 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1366 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1367 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1368 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1369 VM_EVENT(kvm, 3, "SET: guest PTFF subfunc 0x%16.16lx.%16.16lx",
1370 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1371 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1372 VM_EVENT(kvm, 3, "SET: guest KMAC subfunc 0x%16.16lx.%16.16lx",
1373 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1374 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1375 VM_EVENT(kvm, 3, "SET: guest KMC subfunc 0x%16.16lx.%16.16lx",
1376 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1377 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1378 VM_EVENT(kvm, 3, "SET: guest KM subfunc 0x%16.16lx.%16.16lx",
1379 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1380 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1381 VM_EVENT(kvm, 3, "SET: guest KIMD subfunc 0x%16.16lx.%16.16lx",
1382 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1383 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1384 VM_EVENT(kvm, 3, "SET: guest KLMD subfunc 0x%16.16lx.%16.16lx",
1385 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1386 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1387 VM_EVENT(kvm, 3, "SET: guest PCKMO subfunc 0x%16.16lx.%16.16lx",
1388 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1389 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1390 VM_EVENT(kvm, 3, "SET: guest KMCTR subfunc 0x%16.16lx.%16.16lx",
1391 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1392 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1393 VM_EVENT(kvm, 3, "SET: guest KMF subfunc 0x%16.16lx.%16.16lx",
1394 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1395 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1396 VM_EVENT(kvm, 3, "SET: guest KMO subfunc 0x%16.16lx.%16.16lx",
1397 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1398 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1399 VM_EVENT(kvm, 3, "SET: guest PCC subfunc 0x%16.16lx.%16.16lx",
1400 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1401 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1402 VM_EVENT(kvm, 3, "SET: guest PPNO subfunc 0x%16.16lx.%16.16lx",
1403 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1404 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1405 VM_EVENT(kvm, 3, "SET: guest KMA subfunc 0x%16.16lx.%16.16lx",
1406 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1407 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1408 VM_EVENT(kvm, 3, "SET: guest KDSA subfunc 0x%16.16lx.%16.16lx",
1409 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1410 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1411 VM_EVENT(kvm, 3, "SET: guest SORTL subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1412 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1413 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1414 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1415 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1416 VM_EVENT(kvm, 3, "SET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1417 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1418 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1419 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1420 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1425 static int kvm_s390_set_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1429 switch (attr->attr) {
1430 case KVM_S390_VM_CPU_PROCESSOR:
1431 ret = kvm_s390_set_processor(kvm, attr);
1433 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1434 ret = kvm_s390_set_processor_feat(kvm, attr);
1436 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1437 ret = kvm_s390_set_processor_subfunc(kvm, attr);
1443 static int kvm_s390_get_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1445 struct kvm_s390_vm_cpu_processor *proc;
1448 proc = kzalloc(sizeof(*proc), GFP_KERNEL_ACCOUNT);
1453 proc->cpuid = kvm->arch.model.cpuid;
1454 proc->ibc = kvm->arch.model.ibc;
1455 memcpy(&proc->fac_list, kvm->arch.model.fac_list,
1456 S390_ARCH_FAC_LIST_SIZE_BYTE);
1457 VM_EVENT(kvm, 3, "GET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1458 kvm->arch.model.ibc,
1459 kvm->arch.model.cpuid);
1460 VM_EVENT(kvm, 3, "GET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1461 kvm->arch.model.fac_list[0],
1462 kvm->arch.model.fac_list[1],
1463 kvm->arch.model.fac_list[2]);
1464 if (copy_to_user((void __user *)attr->addr, proc, sizeof(*proc)))
1471 static int kvm_s390_get_machine(struct kvm *kvm, struct kvm_device_attr *attr)
1473 struct kvm_s390_vm_cpu_machine *mach;
1476 mach = kzalloc(sizeof(*mach), GFP_KERNEL_ACCOUNT);
1481 get_cpu_id((struct cpuid *) &mach->cpuid);
1482 mach->ibc = sclp.ibc;
1483 memcpy(&mach->fac_mask, kvm->arch.model.fac_mask,
1484 S390_ARCH_FAC_LIST_SIZE_BYTE);
1485 memcpy((unsigned long *)&mach->fac_list, stfle_fac_list,
1486 sizeof(stfle_fac_list));
1487 VM_EVENT(kvm, 3, "GET: host ibc: 0x%4.4x, host cpuid: 0x%16.16llx",
1488 kvm->arch.model.ibc,
1489 kvm->arch.model.cpuid);
1490 VM_EVENT(kvm, 3, "GET: host facmask: 0x%16.16llx.%16.16llx.%16.16llx",
1494 VM_EVENT(kvm, 3, "GET: host faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1498 if (copy_to_user((void __user *)attr->addr, mach, sizeof(*mach)))
1505 static int kvm_s390_get_processor_feat(struct kvm *kvm,
1506 struct kvm_device_attr *attr)
1508 struct kvm_s390_vm_cpu_feat data;
1510 bitmap_copy((unsigned long *) data.feat, kvm->arch.cpu_feat,
1511 KVM_S390_VM_CPU_FEAT_NR_BITS);
1512 if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1514 VM_EVENT(kvm, 3, "GET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1521 static int kvm_s390_get_machine_feat(struct kvm *kvm,
1522 struct kvm_device_attr *attr)
1524 struct kvm_s390_vm_cpu_feat data;
1526 bitmap_copy((unsigned long *) data.feat,
1527 kvm_s390_available_cpu_feat,
1528 KVM_S390_VM_CPU_FEAT_NR_BITS);
1529 if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1531 VM_EVENT(kvm, 3, "GET: host feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1538 static int kvm_s390_get_processor_subfunc(struct kvm *kvm,
1539 struct kvm_device_attr *attr)
1541 if (copy_to_user((void __user *)attr->addr, &kvm->arch.model.subfuncs,
1542 sizeof(struct kvm_s390_vm_cpu_subfunc)))
1545 VM_EVENT(kvm, 3, "GET: guest PLO subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1546 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1547 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1548 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1549 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1550 VM_EVENT(kvm, 3, "GET: guest PTFF subfunc 0x%16.16lx.%16.16lx",
1551 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1552 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1553 VM_EVENT(kvm, 3, "GET: guest KMAC subfunc 0x%16.16lx.%16.16lx",
1554 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1555 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1556 VM_EVENT(kvm, 3, "GET: guest KMC subfunc 0x%16.16lx.%16.16lx",
1557 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1558 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1559 VM_EVENT(kvm, 3, "GET: guest KM subfunc 0x%16.16lx.%16.16lx",
1560 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1561 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1562 VM_EVENT(kvm, 3, "GET: guest KIMD subfunc 0x%16.16lx.%16.16lx",
1563 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1564 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1565 VM_EVENT(kvm, 3, "GET: guest KLMD subfunc 0x%16.16lx.%16.16lx",
1566 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1567 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1568 VM_EVENT(kvm, 3, "GET: guest PCKMO subfunc 0x%16.16lx.%16.16lx",
1569 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1570 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1571 VM_EVENT(kvm, 3, "GET: guest KMCTR subfunc 0x%16.16lx.%16.16lx",
1572 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1573 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1574 VM_EVENT(kvm, 3, "GET: guest KMF subfunc 0x%16.16lx.%16.16lx",
1575 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1576 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1577 VM_EVENT(kvm, 3, "GET: guest KMO subfunc 0x%16.16lx.%16.16lx",
1578 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1579 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1580 VM_EVENT(kvm, 3, "GET: guest PCC subfunc 0x%16.16lx.%16.16lx",
1581 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1582 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1583 VM_EVENT(kvm, 3, "GET: guest PPNO subfunc 0x%16.16lx.%16.16lx",
1584 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1585 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1586 VM_EVENT(kvm, 3, "GET: guest KMA subfunc 0x%16.16lx.%16.16lx",
1587 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1588 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1589 VM_EVENT(kvm, 3, "GET: guest KDSA subfunc 0x%16.16lx.%16.16lx",
1590 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1591 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1592 VM_EVENT(kvm, 3, "GET: guest SORTL subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1593 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1594 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1595 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1596 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1597 VM_EVENT(kvm, 3, "GET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1598 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1599 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1600 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1601 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1606 static int kvm_s390_get_machine_subfunc(struct kvm *kvm,
1607 struct kvm_device_attr *attr)
1609 if (copy_to_user((void __user *)attr->addr, &kvm_s390_available_subfunc,
1610 sizeof(struct kvm_s390_vm_cpu_subfunc)))
1613 VM_EVENT(kvm, 3, "GET: host PLO subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1614 ((unsigned long *) &kvm_s390_available_subfunc.plo)[0],
1615 ((unsigned long *) &kvm_s390_available_subfunc.plo)[1],
1616 ((unsigned long *) &kvm_s390_available_subfunc.plo)[2],
1617 ((unsigned long *) &kvm_s390_available_subfunc.plo)[3]);
1618 VM_EVENT(kvm, 3, "GET: host PTFF subfunc 0x%16.16lx.%16.16lx",
1619 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[0],
1620 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[1]);
1621 VM_EVENT(kvm, 3, "GET: host KMAC subfunc 0x%16.16lx.%16.16lx",
1622 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[0],
1623 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[1]);
1624 VM_EVENT(kvm, 3, "GET: host KMC subfunc 0x%16.16lx.%16.16lx",
1625 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[0],
1626 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[1]);
1627 VM_EVENT(kvm, 3, "GET: host KM subfunc 0x%16.16lx.%16.16lx",
1628 ((unsigned long *) &kvm_s390_available_subfunc.km)[0],
1629 ((unsigned long *) &kvm_s390_available_subfunc.km)[1]);
1630 VM_EVENT(kvm, 3, "GET: host KIMD subfunc 0x%16.16lx.%16.16lx",
1631 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[0],
1632 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[1]);
1633 VM_EVENT(kvm, 3, "GET: host KLMD subfunc 0x%16.16lx.%16.16lx",
1634 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[0],
1635 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[1]);
1636 VM_EVENT(kvm, 3, "GET: host PCKMO subfunc 0x%16.16lx.%16.16lx",
1637 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[0],
1638 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[1]);
1639 VM_EVENT(kvm, 3, "GET: host KMCTR subfunc 0x%16.16lx.%16.16lx",
1640 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[0],
1641 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[1]);
1642 VM_EVENT(kvm, 3, "GET: host KMF subfunc 0x%16.16lx.%16.16lx",
1643 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[0],
1644 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[1]);
1645 VM_EVENT(kvm, 3, "GET: host KMO subfunc 0x%16.16lx.%16.16lx",
1646 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[0],
1647 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[1]);
1648 VM_EVENT(kvm, 3, "GET: host PCC subfunc 0x%16.16lx.%16.16lx",
1649 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[0],
1650 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[1]);
1651 VM_EVENT(kvm, 3, "GET: host PPNO subfunc 0x%16.16lx.%16.16lx",
1652 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[0],
1653 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[1]);
1654 VM_EVENT(kvm, 3, "GET: host KMA subfunc 0x%16.16lx.%16.16lx",
1655 ((unsigned long *) &kvm_s390_available_subfunc.kma)[0],
1656 ((unsigned long *) &kvm_s390_available_subfunc.kma)[1]);
1657 VM_EVENT(kvm, 3, "GET: host KDSA subfunc 0x%16.16lx.%16.16lx",
1658 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[0],
1659 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[1]);
1660 VM_EVENT(kvm, 3, "GET: host SORTL subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1661 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[0],
1662 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[1],
1663 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[2],
1664 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[3]);
1665 VM_EVENT(kvm, 3, "GET: host DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1666 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[0],
1667 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[1],
1668 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[2],
1669 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[3]);
1674 static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1678 switch (attr->attr) {
1679 case KVM_S390_VM_CPU_PROCESSOR:
1680 ret = kvm_s390_get_processor(kvm, attr);
1682 case KVM_S390_VM_CPU_MACHINE:
1683 ret = kvm_s390_get_machine(kvm, attr);
1685 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1686 ret = kvm_s390_get_processor_feat(kvm, attr);
1688 case KVM_S390_VM_CPU_MACHINE_FEAT:
1689 ret = kvm_s390_get_machine_feat(kvm, attr);
1691 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1692 ret = kvm_s390_get_processor_subfunc(kvm, attr);
1694 case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1695 ret = kvm_s390_get_machine_subfunc(kvm, attr);
1701 static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1705 switch (attr->group) {
1706 case KVM_S390_VM_MEM_CTRL:
1707 ret = kvm_s390_set_mem_control(kvm, attr);
1709 case KVM_S390_VM_TOD:
1710 ret = kvm_s390_set_tod(kvm, attr);
1712 case KVM_S390_VM_CPU_MODEL:
1713 ret = kvm_s390_set_cpu_model(kvm, attr);
1715 case KVM_S390_VM_CRYPTO:
1716 ret = kvm_s390_vm_set_crypto(kvm, attr);
1718 case KVM_S390_VM_MIGRATION:
1719 ret = kvm_s390_vm_set_migration(kvm, attr);
1729 static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1733 switch (attr->group) {
1734 case KVM_S390_VM_MEM_CTRL:
1735 ret = kvm_s390_get_mem_control(kvm, attr);
1737 case KVM_S390_VM_TOD:
1738 ret = kvm_s390_get_tod(kvm, attr);
1740 case KVM_S390_VM_CPU_MODEL:
1741 ret = kvm_s390_get_cpu_model(kvm, attr);
1743 case KVM_S390_VM_MIGRATION:
1744 ret = kvm_s390_vm_get_migration(kvm, attr);
1754 static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1758 switch (attr->group) {
1759 case KVM_S390_VM_MEM_CTRL:
1760 switch (attr->attr) {
1761 case KVM_S390_VM_MEM_ENABLE_CMMA:
1762 case KVM_S390_VM_MEM_CLR_CMMA:
1763 ret = sclp.has_cmma ? 0 : -ENXIO;
1765 case KVM_S390_VM_MEM_LIMIT_SIZE:
1773 case KVM_S390_VM_TOD:
1774 switch (attr->attr) {
1775 case KVM_S390_VM_TOD_LOW:
1776 case KVM_S390_VM_TOD_HIGH:
1784 case KVM_S390_VM_CPU_MODEL:
1785 switch (attr->attr) {
1786 case KVM_S390_VM_CPU_PROCESSOR:
1787 case KVM_S390_VM_CPU_MACHINE:
1788 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1789 case KVM_S390_VM_CPU_MACHINE_FEAT:
1790 case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1791 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1799 case KVM_S390_VM_CRYPTO:
1800 switch (attr->attr) {
1801 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
1802 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
1803 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
1804 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
1807 case KVM_S390_VM_CRYPTO_ENABLE_APIE:
1808 case KVM_S390_VM_CRYPTO_DISABLE_APIE:
1809 ret = ap_instructions_available() ? 0 : -ENXIO;
1816 case KVM_S390_VM_MIGRATION:
1827 static long kvm_s390_get_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1831 int srcu_idx, i, r = 0;
1833 if (args->flags != 0)
1836 /* Is this guest using storage keys? */
1837 if (!mm_uses_skeys(current->mm))
1838 return KVM_S390_GET_SKEYS_NONE;
1840 /* Enforce sane limit on memory allocation */
1841 if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1844 keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL_ACCOUNT);
1848 mmap_read_lock(current->mm);
1849 srcu_idx = srcu_read_lock(&kvm->srcu);
1850 for (i = 0; i < args->count; i++) {
1851 hva = gfn_to_hva(kvm, args->start_gfn + i);
1852 if (kvm_is_error_hva(hva)) {
1857 r = get_guest_storage_key(current->mm, hva, &keys[i]);
1861 srcu_read_unlock(&kvm->srcu, srcu_idx);
1862 mmap_read_unlock(current->mm);
1865 r = copy_to_user((uint8_t __user *)args->skeydata_addr, keys,
1866 sizeof(uint8_t) * args->count);
1875 static long kvm_s390_set_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1879 int srcu_idx, i, r = 0;
1882 if (args->flags != 0)
1885 /* Enforce sane limit on memory allocation */
1886 if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1889 keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL_ACCOUNT);
1893 r = copy_from_user(keys, (uint8_t __user *)args->skeydata_addr,
1894 sizeof(uint8_t) * args->count);
1900 /* Enable storage key handling for the guest */
1901 r = s390_enable_skey();
1906 mmap_read_lock(current->mm);
1907 srcu_idx = srcu_read_lock(&kvm->srcu);
1908 while (i < args->count) {
1910 hva = gfn_to_hva(kvm, args->start_gfn + i);
1911 if (kvm_is_error_hva(hva)) {
1916 /* Lowest order bit is reserved */
1917 if (keys[i] & 0x01) {
1922 r = set_guest_storage_key(current->mm, hva, keys[i], 0);
1924 r = fixup_user_fault(current->mm, hva,
1925 FAULT_FLAG_WRITE, &unlocked);
1932 srcu_read_unlock(&kvm->srcu, srcu_idx);
1933 mmap_read_unlock(current->mm);
1940 * Base address and length must be sent at the start of each block, therefore
1941 * it's cheaper to send some clean data, as long as it's less than the size of
1944 #define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
1945 /* for consistency */
1946 #define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
1949 * Similar to gfn_to_memslot, but returns the index of a memslot also when the
1950 * address falls in a hole. In that case the index of one of the memslots
1951 * bordering the hole is returned.
1953 static int gfn_to_memslot_approx(struct kvm_memslots *slots, gfn_t gfn)
1955 int start = 0, end = slots->used_slots;
1956 int slot = atomic_read(&slots->lru_slot);
1957 struct kvm_memory_slot *memslots = slots->memslots;
1959 if (gfn >= memslots[slot].base_gfn &&
1960 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1963 while (start < end) {
1964 slot = start + (end - start) / 2;
1966 if (gfn >= memslots[slot].base_gfn)
1972 if (start >= slots->used_slots)
1973 return slots->used_slots - 1;
1975 if (gfn >= memslots[start].base_gfn &&
1976 gfn < memslots[start].base_gfn + memslots[start].npages) {
1977 atomic_set(&slots->lru_slot, start);
1983 static int kvm_s390_peek_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
1984 u8 *res, unsigned long bufsize)
1986 unsigned long pgstev, hva, cur_gfn = args->start_gfn;
1989 while (args->count < bufsize) {
1990 hva = gfn_to_hva(kvm, cur_gfn);
1992 * We return an error if the first value was invalid, but we
1993 * return successfully if at least one value was copied.
1995 if (kvm_is_error_hva(hva))
1996 return args->count ? 0 : -EFAULT;
1997 if (get_pgste(kvm->mm, hva, &pgstev) < 0)
1999 res[args->count++] = (pgstev >> 24) & 0x43;
2006 static unsigned long kvm_s390_next_dirty_cmma(struct kvm_memslots *slots,
2007 unsigned long cur_gfn)
2009 int slotidx = gfn_to_memslot_approx(slots, cur_gfn);
2010 struct kvm_memory_slot *ms = slots->memslots + slotidx;
2011 unsigned long ofs = cur_gfn - ms->base_gfn;
2013 if (ms->base_gfn + ms->npages <= cur_gfn) {
2015 /* If we are above the highest slot, wrap around */
2017 slotidx = slots->used_slots - 1;
2019 ms = slots->memslots + slotidx;
2022 ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, ofs);
2023 while ((slotidx > 0) && (ofs >= ms->npages)) {
2025 ms = slots->memslots + slotidx;
2026 ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, 0);
2028 return ms->base_gfn + ofs;
2031 static int kvm_s390_get_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
2032 u8 *res, unsigned long bufsize)
2034 unsigned long mem_end, cur_gfn, next_gfn, hva, pgstev;
2035 struct kvm_memslots *slots = kvm_memslots(kvm);
2036 struct kvm_memory_slot *ms;
2038 if (unlikely(!slots->used_slots))
2041 cur_gfn = kvm_s390_next_dirty_cmma(slots, args->start_gfn);
2042 ms = gfn_to_memslot(kvm, cur_gfn);
2044 args->start_gfn = cur_gfn;
2047 next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2048 mem_end = slots->memslots[0].base_gfn + slots->memslots[0].npages;
2050 while (args->count < bufsize) {
2051 hva = gfn_to_hva(kvm, cur_gfn);
2052 if (kvm_is_error_hva(hva))
2054 /* Decrement only if we actually flipped the bit to 0 */
2055 if (test_and_clear_bit(cur_gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
2056 atomic64_dec(&kvm->arch.cmma_dirty_pages);
2057 if (get_pgste(kvm->mm, hva, &pgstev) < 0)
2059 /* Save the value */
2060 res[args->count++] = (pgstev >> 24) & 0x43;
2061 /* If the next bit is too far away, stop. */
2062 if (next_gfn > cur_gfn + KVM_S390_MAX_BIT_DISTANCE)
2064 /* If we reached the previous "next", find the next one */
2065 if (cur_gfn == next_gfn)
2066 next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2067 /* Reached the end of memory or of the buffer, stop */
2068 if ((next_gfn >= mem_end) ||
2069 (next_gfn - args->start_gfn >= bufsize))
2072 /* Reached the end of the current memslot, take the next one. */
2073 if (cur_gfn - ms->base_gfn >= ms->npages) {
2074 ms = gfn_to_memslot(kvm, cur_gfn);
2083 * This function searches for the next page with dirty CMMA attributes, and
2084 * saves the attributes in the buffer up to either the end of the buffer or
2085 * until a block of at least KVM_S390_MAX_BIT_DISTANCE clean bits is found;
2086 * no trailing clean bytes are saved.
2087 * In case no dirty bits were found, or if CMMA was not enabled or used, the
2088 * output buffer will indicate 0 as length.
2090 static int kvm_s390_get_cmma_bits(struct kvm *kvm,
2091 struct kvm_s390_cmma_log *args)
2093 unsigned long bufsize;
2094 int srcu_idx, peek, ret;
2097 if (!kvm->arch.use_cmma)
2099 /* Invalid/unsupported flags were specified */
2100 if (args->flags & ~KVM_S390_CMMA_PEEK)
2102 /* Migration mode query, and we are not doing a migration */
2103 peek = !!(args->flags & KVM_S390_CMMA_PEEK);
2104 if (!peek && !kvm->arch.migration_mode)
2106 /* CMMA is disabled or was not used, or the buffer has length zero */
2107 bufsize = min(args->count, KVM_S390_CMMA_SIZE_MAX);
2108 if (!bufsize || !kvm->mm->context.uses_cmm) {
2109 memset(args, 0, sizeof(*args));
2112 /* We are not peeking, and there are no dirty pages */
2113 if (!peek && !atomic64_read(&kvm->arch.cmma_dirty_pages)) {
2114 memset(args, 0, sizeof(*args));
2118 values = vmalloc(bufsize);
2122 mmap_read_lock(kvm->mm);
2123 srcu_idx = srcu_read_lock(&kvm->srcu);
2125 ret = kvm_s390_peek_cmma(kvm, args, values, bufsize);
2127 ret = kvm_s390_get_cmma(kvm, args, values, bufsize);
2128 srcu_read_unlock(&kvm->srcu, srcu_idx);
2129 mmap_read_unlock(kvm->mm);
2131 if (kvm->arch.migration_mode)
2132 args->remaining = atomic64_read(&kvm->arch.cmma_dirty_pages);
2134 args->remaining = 0;
2136 if (copy_to_user((void __user *)args->values, values, args->count))
2144 * This function sets the CMMA attributes for the given pages. If the input
2145 * buffer has zero length, no action is taken, otherwise the attributes are
2146 * set and the mm->context.uses_cmm flag is set.
2148 static int kvm_s390_set_cmma_bits(struct kvm *kvm,
2149 const struct kvm_s390_cmma_log *args)
2151 unsigned long hva, mask, pgstev, i;
2153 int srcu_idx, r = 0;
2157 if (!kvm->arch.use_cmma)
2159 /* invalid/unsupported flags */
2160 if (args->flags != 0)
2162 /* Enforce sane limit on memory allocation */
2163 if (args->count > KVM_S390_CMMA_SIZE_MAX)
2166 if (args->count == 0)
2169 bits = vmalloc(array_size(sizeof(*bits), args->count));
2173 r = copy_from_user(bits, (void __user *)args->values, args->count);
2179 mmap_read_lock(kvm->mm);
2180 srcu_idx = srcu_read_lock(&kvm->srcu);
2181 for (i = 0; i < args->count; i++) {
2182 hva = gfn_to_hva(kvm, args->start_gfn + i);
2183 if (kvm_is_error_hva(hva)) {
2189 pgstev = pgstev << 24;
2190 mask &= _PGSTE_GPS_USAGE_MASK | _PGSTE_GPS_NODAT;
2191 set_pgste_bits(kvm->mm, hva, mask, pgstev);
2193 srcu_read_unlock(&kvm->srcu, srcu_idx);
2194 mmap_read_unlock(kvm->mm);
2196 if (!kvm->mm->context.uses_cmm) {
2197 mmap_write_lock(kvm->mm);
2198 kvm->mm->context.uses_cmm = 1;
2199 mmap_write_unlock(kvm->mm);
2206 static int kvm_s390_cpus_from_pv(struct kvm *kvm, u16 *rcp, u16 *rrcp)
2208 struct kvm_vcpu *vcpu;
2214 * We ignore failures and try to destroy as many CPUs as possible.
2215 * At the same time we must not free the assigned resources when
2216 * this fails, as the ultravisor has still access to that memory.
2217 * So kvm_s390_pv_destroy_cpu can leave a "wanted" memory leak
2219 * We want to return the first failure rc and rrc, though.
2221 kvm_for_each_vcpu(i, vcpu, kvm) {
2222 mutex_lock(&vcpu->mutex);
2223 if (kvm_s390_pv_destroy_cpu(vcpu, &rc, &rrc) && !ret) {
2228 mutex_unlock(&vcpu->mutex);
2233 static int kvm_s390_cpus_to_pv(struct kvm *kvm, u16 *rc, u16 *rrc)
2238 struct kvm_vcpu *vcpu;
2240 kvm_for_each_vcpu(i, vcpu, kvm) {
2241 mutex_lock(&vcpu->mutex);
2242 r = kvm_s390_pv_create_cpu(vcpu, rc, rrc);
2243 mutex_unlock(&vcpu->mutex);
2248 kvm_s390_cpus_from_pv(kvm, &dummy, &dummy);
2252 static int kvm_s390_handle_pv(struct kvm *kvm, struct kvm_pv_cmd *cmd)
2256 void __user *argp = (void __user *)cmd->data;
2259 case KVM_PV_ENABLE: {
2261 if (kvm_s390_pv_is_protected(kvm))
2265 * FMT 4 SIE needs esca. As we never switch back to bsca from
2266 * esca, we need no cleanup in the error cases below
2268 r = sca_switch_to_extended(kvm);
2272 mmap_write_lock(current->mm);
2273 r = gmap_mark_unmergeable();
2274 mmap_write_unlock(current->mm);
2278 r = kvm_s390_pv_init_vm(kvm, &cmd->rc, &cmd->rrc);
2282 r = kvm_s390_cpus_to_pv(kvm, &cmd->rc, &cmd->rrc);
2284 kvm_s390_pv_deinit_vm(kvm, &dummy, &dummy);
2286 /* we need to block service interrupts from now on */
2287 set_bit(IRQ_PEND_EXT_SERVICE, &kvm->arch.float_int.masked_irqs);
2290 case KVM_PV_DISABLE: {
2292 if (!kvm_s390_pv_is_protected(kvm))
2295 r = kvm_s390_cpus_from_pv(kvm, &cmd->rc, &cmd->rrc);
2297 * If a CPU could not be destroyed, destroy VM will also fail.
2298 * There is no point in trying to destroy it. Instead return
2299 * the rc and rrc from the first CPU that failed destroying.
2303 r = kvm_s390_pv_deinit_vm(kvm, &cmd->rc, &cmd->rrc);
2305 /* no need to block service interrupts any more */
2306 clear_bit(IRQ_PEND_EXT_SERVICE, &kvm->arch.float_int.masked_irqs);
2309 case KVM_PV_SET_SEC_PARMS: {
2310 struct kvm_s390_pv_sec_parm parms = {};
2314 if (!kvm_s390_pv_is_protected(kvm))
2318 if (copy_from_user(&parms, argp, sizeof(parms)))
2321 /* Currently restricted to 8KB */
2323 if (parms.length > PAGE_SIZE * 2)
2327 hdr = vmalloc(parms.length);
2332 if (!copy_from_user(hdr, (void __user *)parms.origin,
2334 r = kvm_s390_pv_set_sec_parms(kvm, hdr, parms.length,
2335 &cmd->rc, &cmd->rrc);
2340 case KVM_PV_UNPACK: {
2341 struct kvm_s390_pv_unp unp = {};
2344 if (!kvm_s390_pv_is_protected(kvm) || !mm_is_protected(kvm->mm))
2348 if (copy_from_user(&unp, argp, sizeof(unp)))
2351 r = kvm_s390_pv_unpack(kvm, unp.addr, unp.size, unp.tweak,
2352 &cmd->rc, &cmd->rrc);
2355 case KVM_PV_VERIFY: {
2357 if (!kvm_s390_pv_is_protected(kvm))
2360 r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2361 UVC_CMD_VERIFY_IMG, &cmd->rc, &cmd->rrc);
2362 KVM_UV_EVENT(kvm, 3, "PROTVIRT VERIFY: rc %x rrc %x", cmd->rc,
2366 case KVM_PV_PREP_RESET: {
2368 if (!kvm_s390_pv_is_protected(kvm))
2371 r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2372 UVC_CMD_PREPARE_RESET, &cmd->rc, &cmd->rrc);
2373 KVM_UV_EVENT(kvm, 3, "PROTVIRT PREP RESET: rc %x rrc %x",
2377 case KVM_PV_UNSHARE_ALL: {
2379 if (!kvm_s390_pv_is_protected(kvm))
2382 r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2383 UVC_CMD_SET_UNSHARE_ALL, &cmd->rc, &cmd->rrc);
2384 KVM_UV_EVENT(kvm, 3, "PROTVIRT UNSHARE: rc %x rrc %x",
2394 long kvm_arch_vm_ioctl(struct file *filp,
2395 unsigned int ioctl, unsigned long arg)
2397 struct kvm *kvm = filp->private_data;
2398 void __user *argp = (void __user *)arg;
2399 struct kvm_device_attr attr;
2403 case KVM_S390_INTERRUPT: {
2404 struct kvm_s390_interrupt s390int;
2407 if (copy_from_user(&s390int, argp, sizeof(s390int)))
2409 r = kvm_s390_inject_vm(kvm, &s390int);
2412 case KVM_CREATE_IRQCHIP: {
2413 struct kvm_irq_routing_entry routing;
2416 if (kvm->arch.use_irqchip) {
2417 /* Set up dummy routing. */
2418 memset(&routing, 0, sizeof(routing));
2419 r = kvm_set_irq_routing(kvm, &routing, 0, 0);
2423 case KVM_SET_DEVICE_ATTR: {
2425 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2427 r = kvm_s390_vm_set_attr(kvm, &attr);
2430 case KVM_GET_DEVICE_ATTR: {
2432 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2434 r = kvm_s390_vm_get_attr(kvm, &attr);
2437 case KVM_HAS_DEVICE_ATTR: {
2439 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2441 r = kvm_s390_vm_has_attr(kvm, &attr);
2444 case KVM_S390_GET_SKEYS: {
2445 struct kvm_s390_skeys args;
2448 if (copy_from_user(&args, argp,
2449 sizeof(struct kvm_s390_skeys)))
2451 r = kvm_s390_get_skeys(kvm, &args);
2454 case KVM_S390_SET_SKEYS: {
2455 struct kvm_s390_skeys args;
2458 if (copy_from_user(&args, argp,
2459 sizeof(struct kvm_s390_skeys)))
2461 r = kvm_s390_set_skeys(kvm, &args);
2464 case KVM_S390_GET_CMMA_BITS: {
2465 struct kvm_s390_cmma_log args;
2468 if (copy_from_user(&args, argp, sizeof(args)))
2470 mutex_lock(&kvm->slots_lock);
2471 r = kvm_s390_get_cmma_bits(kvm, &args);
2472 mutex_unlock(&kvm->slots_lock);
2474 r = copy_to_user(argp, &args, sizeof(args));
2480 case KVM_S390_SET_CMMA_BITS: {
2481 struct kvm_s390_cmma_log args;
2484 if (copy_from_user(&args, argp, sizeof(args)))
2486 mutex_lock(&kvm->slots_lock);
2487 r = kvm_s390_set_cmma_bits(kvm, &args);
2488 mutex_unlock(&kvm->slots_lock);
2491 case KVM_S390_PV_COMMAND: {
2492 struct kvm_pv_cmd args;
2494 /* protvirt means user sigp */
2495 kvm->arch.user_cpu_state_ctrl = 1;
2497 if (!is_prot_virt_host()) {
2501 if (copy_from_user(&args, argp, sizeof(args))) {
2509 mutex_lock(&kvm->lock);
2510 r = kvm_s390_handle_pv(kvm, &args);
2511 mutex_unlock(&kvm->lock);
2512 if (copy_to_user(argp, &args, sizeof(args))) {
2525 static int kvm_s390_apxa_installed(void)
2527 struct ap_config_info info;
2529 if (ap_instructions_available()) {
2530 if (ap_qci(&info) == 0)
2538 * The format of the crypto control block (CRYCB) is specified in the 3 low
2539 * order bits of the CRYCB designation (CRYCBD) field as follows:
2540 * Format 0: Neither the message security assist extension 3 (MSAX3) nor the
2541 * AP extended addressing (APXA) facility are installed.
2542 * Format 1: The APXA facility is not installed but the MSAX3 facility is.
2543 * Format 2: Both the APXA and MSAX3 facilities are installed
2545 static void kvm_s390_set_crycb_format(struct kvm *kvm)
2547 kvm->arch.crypto.crycbd = (__u32)(unsigned long) kvm->arch.crypto.crycb;
2549 /* Clear the CRYCB format bits - i.e., set format 0 by default */
2550 kvm->arch.crypto.crycbd &= ~(CRYCB_FORMAT_MASK);
2552 /* Check whether MSAX3 is installed */
2553 if (!test_kvm_facility(kvm, 76))
2556 if (kvm_s390_apxa_installed())
2557 kvm->arch.crypto.crycbd |= CRYCB_FORMAT2;
2559 kvm->arch.crypto.crycbd |= CRYCB_FORMAT1;
2563 * kvm_arch_crypto_set_masks
2565 * @kvm: pointer to the target guest's KVM struct containing the crypto masks
2567 * @apm: the mask identifying the accessible AP adapters
2568 * @aqm: the mask identifying the accessible AP domains
2569 * @adm: the mask identifying the accessible AP control domains
2571 * Set the masks that identify the adapters, domains and control domains to
2572 * which the KVM guest is granted access.
2574 * Note: The kvm->lock mutex must be locked by the caller before invoking this
2577 void kvm_arch_crypto_set_masks(struct kvm *kvm, unsigned long *apm,
2578 unsigned long *aqm, unsigned long *adm)
2580 struct kvm_s390_crypto_cb *crycb = kvm->arch.crypto.crycb;
2582 kvm_s390_vcpu_block_all(kvm);
2584 switch (kvm->arch.crypto.crycbd & CRYCB_FORMAT_MASK) {
2585 case CRYCB_FORMAT2: /* APCB1 use 256 bits */
2586 memcpy(crycb->apcb1.apm, apm, 32);
2587 VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx %016lx %016lx %016lx",
2588 apm[0], apm[1], apm[2], apm[3]);
2589 memcpy(crycb->apcb1.aqm, aqm, 32);
2590 VM_EVENT(kvm, 3, "SET CRYCB: aqm %016lx %016lx %016lx %016lx",
2591 aqm[0], aqm[1], aqm[2], aqm[3]);
2592 memcpy(crycb->apcb1.adm, adm, 32);
2593 VM_EVENT(kvm, 3, "SET CRYCB: adm %016lx %016lx %016lx %016lx",
2594 adm[0], adm[1], adm[2], adm[3]);
2597 case CRYCB_FORMAT0: /* Fall through both use APCB0 */
2598 memcpy(crycb->apcb0.apm, apm, 8);
2599 memcpy(crycb->apcb0.aqm, aqm, 2);
2600 memcpy(crycb->apcb0.adm, adm, 2);
2601 VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx aqm %04x adm %04x",
2602 apm[0], *((unsigned short *)aqm),
2603 *((unsigned short *)adm));
2605 default: /* Can not happen */
2609 /* recreate the shadow crycb for each vcpu */
2610 kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2611 kvm_s390_vcpu_unblock_all(kvm);
2613 EXPORT_SYMBOL_GPL(kvm_arch_crypto_set_masks);
2616 * kvm_arch_crypto_clear_masks
2618 * @kvm: pointer to the target guest's KVM struct containing the crypto masks
2621 * Clear the masks that identify the adapters, domains and control domains to
2622 * which the KVM guest is granted access.
2624 * Note: The kvm->lock mutex must be locked by the caller before invoking this
2627 void kvm_arch_crypto_clear_masks(struct kvm *kvm)
2629 kvm_s390_vcpu_block_all(kvm);
2631 memset(&kvm->arch.crypto.crycb->apcb0, 0,
2632 sizeof(kvm->arch.crypto.crycb->apcb0));
2633 memset(&kvm->arch.crypto.crycb->apcb1, 0,
2634 sizeof(kvm->arch.crypto.crycb->apcb1));
2636 VM_EVENT(kvm, 3, "%s", "CLR CRYCB:");
2637 /* recreate the shadow crycb for each vcpu */
2638 kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2639 kvm_s390_vcpu_unblock_all(kvm);
2641 EXPORT_SYMBOL_GPL(kvm_arch_crypto_clear_masks);
2643 static u64 kvm_s390_get_initial_cpuid(void)
2648 cpuid.version = 0xff;
2649 return *((u64 *) &cpuid);
2652 static void kvm_s390_crypto_init(struct kvm *kvm)
2654 kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
2655 kvm_s390_set_crycb_format(kvm);
2656 init_rwsem(&kvm->arch.crypto.pqap_hook_rwsem);
2658 if (!test_kvm_facility(kvm, 76))
2661 /* Enable AES/DEA protected key functions by default */
2662 kvm->arch.crypto.aes_kw = 1;
2663 kvm->arch.crypto.dea_kw = 1;
2664 get_random_bytes(kvm->arch.crypto.crycb->aes_wrapping_key_mask,
2665 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
2666 get_random_bytes(kvm->arch.crypto.crycb->dea_wrapping_key_mask,
2667 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
2670 static void sca_dispose(struct kvm *kvm)
2672 if (kvm->arch.use_esca)
2673 free_pages_exact(kvm->arch.sca, sizeof(struct esca_block));
2675 free_page((unsigned long)(kvm->arch.sca));
2676 kvm->arch.sca = NULL;
2679 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
2681 gfp_t alloc_flags = GFP_KERNEL_ACCOUNT;
2683 char debug_name[16];
2684 static unsigned long sca_offset;
2687 #ifdef CONFIG_KVM_S390_UCONTROL
2688 if (type & ~KVM_VM_S390_UCONTROL)
2690 if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN)))
2697 rc = s390_enable_sie();
2703 if (!sclp.has_64bscao)
2704 alloc_flags |= GFP_DMA;
2705 rwlock_init(&kvm->arch.sca_lock);
2706 /* start with basic SCA */
2707 kvm->arch.sca = (struct bsca_block *) get_zeroed_page(alloc_flags);
2710 mutex_lock(&kvm_lock);
2712 if (sca_offset + sizeof(struct bsca_block) > PAGE_SIZE)
2714 kvm->arch.sca = (struct bsca_block *)
2715 ((char *) kvm->arch.sca + sca_offset);
2716 mutex_unlock(&kvm_lock);
2718 sprintf(debug_name, "kvm-%u", current->pid);
2720 kvm->arch.dbf = debug_register(debug_name, 32, 1, 7 * sizeof(long));
2724 BUILD_BUG_ON(sizeof(struct sie_page2) != 4096);
2725 kvm->arch.sie_page2 =
2726 (struct sie_page2 *) get_zeroed_page(GFP_KERNEL_ACCOUNT | GFP_DMA);
2727 if (!kvm->arch.sie_page2)
2730 kvm->arch.sie_page2->kvm = kvm;
2731 kvm->arch.model.fac_list = kvm->arch.sie_page2->fac_list;
2733 for (i = 0; i < kvm_s390_fac_size(); i++) {
2734 kvm->arch.model.fac_mask[i] = stfle_fac_list[i] &
2735 (kvm_s390_fac_base[i] |
2736 kvm_s390_fac_ext[i]);
2737 kvm->arch.model.fac_list[i] = stfle_fac_list[i] &
2738 kvm_s390_fac_base[i];
2740 kvm->arch.model.subfuncs = kvm_s390_available_subfunc;
2742 /* we are always in czam mode - even on pre z14 machines */
2743 set_kvm_facility(kvm->arch.model.fac_mask, 138);
2744 set_kvm_facility(kvm->arch.model.fac_list, 138);
2745 /* we emulate STHYI in kvm */
2746 set_kvm_facility(kvm->arch.model.fac_mask, 74);
2747 set_kvm_facility(kvm->arch.model.fac_list, 74);
2748 if (MACHINE_HAS_TLB_GUEST) {
2749 set_kvm_facility(kvm->arch.model.fac_mask, 147);
2750 set_kvm_facility(kvm->arch.model.fac_list, 147);
2753 if (css_general_characteristics.aiv && test_facility(65))
2754 set_kvm_facility(kvm->arch.model.fac_mask, 65);
2756 kvm->arch.model.cpuid = kvm_s390_get_initial_cpuid();
2757 kvm->arch.model.ibc = sclp.ibc & 0x0fff;
2759 kvm_s390_crypto_init(kvm);
2761 mutex_init(&kvm->arch.float_int.ais_lock);
2762 spin_lock_init(&kvm->arch.float_int.lock);
2763 for (i = 0; i < FIRQ_LIST_COUNT; i++)
2764 INIT_LIST_HEAD(&kvm->arch.float_int.lists[i]);
2765 init_waitqueue_head(&kvm->arch.ipte_wq);
2766 mutex_init(&kvm->arch.ipte_mutex);
2768 debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
2769 VM_EVENT(kvm, 3, "vm created with type %lu", type);
2771 if (type & KVM_VM_S390_UCONTROL) {
2772 kvm->arch.gmap = NULL;
2773 kvm->arch.mem_limit = KVM_S390_NO_MEM_LIMIT;
2775 if (sclp.hamax == U64_MAX)
2776 kvm->arch.mem_limit = TASK_SIZE_MAX;
2778 kvm->arch.mem_limit = min_t(unsigned long, TASK_SIZE_MAX,
2780 kvm->arch.gmap = gmap_create(current->mm, kvm->arch.mem_limit - 1);
2781 if (!kvm->arch.gmap)
2783 kvm->arch.gmap->private = kvm;
2784 kvm->arch.gmap->pfault_enabled = 0;
2787 kvm->arch.use_pfmfi = sclp.has_pfmfi;
2788 kvm->arch.use_skf = sclp.has_skey;
2789 spin_lock_init(&kvm->arch.start_stop_lock);
2790 kvm_s390_vsie_init(kvm);
2792 kvm_s390_gisa_init(kvm);
2793 KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm, current->pid);
2797 free_page((unsigned long)kvm->arch.sie_page2);
2798 debug_unregister(kvm->arch.dbf);
2800 KVM_EVENT(3, "creation of vm failed: %d", rc);
2804 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
2808 VCPU_EVENT(vcpu, 3, "%s", "free cpu");
2809 trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id);
2810 kvm_s390_clear_local_irqs(vcpu);
2811 kvm_clear_async_pf_completion_queue(vcpu);
2812 if (!kvm_is_ucontrol(vcpu->kvm))
2815 if (kvm_is_ucontrol(vcpu->kvm))
2816 gmap_remove(vcpu->arch.gmap);
2818 if (vcpu->kvm->arch.use_cmma)
2819 kvm_s390_vcpu_unsetup_cmma(vcpu);
2820 /* We can not hold the vcpu mutex here, we are already dying */
2821 if (kvm_s390_pv_cpu_get_handle(vcpu))
2822 kvm_s390_pv_destroy_cpu(vcpu, &rc, &rrc);
2823 free_page((unsigned long)(vcpu->arch.sie_block));
2826 static void kvm_free_vcpus(struct kvm *kvm)
2829 struct kvm_vcpu *vcpu;
2831 kvm_for_each_vcpu(i, vcpu, kvm)
2832 kvm_vcpu_destroy(vcpu);
2834 mutex_lock(&kvm->lock);
2835 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
2836 kvm->vcpus[i] = NULL;
2838 atomic_set(&kvm->online_vcpus, 0);
2839 mutex_unlock(&kvm->lock);
2842 void kvm_arch_destroy_vm(struct kvm *kvm)
2846 kvm_free_vcpus(kvm);
2848 kvm_s390_gisa_destroy(kvm);
2850 * We are already at the end of life and kvm->lock is not taken.
2851 * This is ok as the file descriptor is closed by now and nobody
2852 * can mess with the pv state. To avoid lockdep_assert_held from
2853 * complaining we do not use kvm_s390_pv_is_protected.
2855 if (kvm_s390_pv_get_handle(kvm))
2856 kvm_s390_pv_deinit_vm(kvm, &rc, &rrc);
2857 debug_unregister(kvm->arch.dbf);
2858 free_page((unsigned long)kvm->arch.sie_page2);
2859 if (!kvm_is_ucontrol(kvm))
2860 gmap_remove(kvm->arch.gmap);
2861 kvm_s390_destroy_adapters(kvm);
2862 kvm_s390_clear_float_irqs(kvm);
2863 kvm_s390_vsie_destroy(kvm);
2864 KVM_EVENT(3, "vm 0x%pK destroyed", kvm);
2867 /* Section: vcpu related */
2868 static int __kvm_ucontrol_vcpu_init(struct kvm_vcpu *vcpu)
2870 vcpu->arch.gmap = gmap_create(current->mm, -1UL);
2871 if (!vcpu->arch.gmap)
2873 vcpu->arch.gmap->private = vcpu->kvm;
2878 static void sca_del_vcpu(struct kvm_vcpu *vcpu)
2880 if (!kvm_s390_use_sca_entries())
2882 read_lock(&vcpu->kvm->arch.sca_lock);
2883 if (vcpu->kvm->arch.use_esca) {
2884 struct esca_block *sca = vcpu->kvm->arch.sca;
2886 clear_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
2887 sca->cpu[vcpu->vcpu_id].sda = 0;
2889 struct bsca_block *sca = vcpu->kvm->arch.sca;
2891 clear_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
2892 sca->cpu[vcpu->vcpu_id].sda = 0;
2894 read_unlock(&vcpu->kvm->arch.sca_lock);
2897 static void sca_add_vcpu(struct kvm_vcpu *vcpu)
2899 if (!kvm_s390_use_sca_entries()) {
2900 struct bsca_block *sca = vcpu->kvm->arch.sca;
2902 /* we still need the basic sca for the ipte control */
2903 vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2904 vcpu->arch.sie_block->scaol = (__u32)(__u64)sca;
2907 read_lock(&vcpu->kvm->arch.sca_lock);
2908 if (vcpu->kvm->arch.use_esca) {
2909 struct esca_block *sca = vcpu->kvm->arch.sca;
2911 sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
2912 vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2913 vcpu->arch.sie_block->scaol = (__u32)(__u64)sca & ~0x3fU;
2914 vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
2915 set_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
2917 struct bsca_block *sca = vcpu->kvm->arch.sca;
2919 sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
2920 vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2921 vcpu->arch.sie_block->scaol = (__u32)(__u64)sca;
2922 set_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
2924 read_unlock(&vcpu->kvm->arch.sca_lock);
2927 /* Basic SCA to Extended SCA data copy routines */
2928 static inline void sca_copy_entry(struct esca_entry *d, struct bsca_entry *s)
2931 d->sigp_ctrl.c = s->sigp_ctrl.c;
2932 d->sigp_ctrl.scn = s->sigp_ctrl.scn;
2935 static void sca_copy_b_to_e(struct esca_block *d, struct bsca_block *s)
2939 d->ipte_control = s->ipte_control;
2941 for (i = 0; i < KVM_S390_BSCA_CPU_SLOTS; i++)
2942 sca_copy_entry(&d->cpu[i], &s->cpu[i]);
2945 static int sca_switch_to_extended(struct kvm *kvm)
2947 struct bsca_block *old_sca = kvm->arch.sca;
2948 struct esca_block *new_sca;
2949 struct kvm_vcpu *vcpu;
2950 unsigned int vcpu_idx;
2953 if (kvm->arch.use_esca)
2956 new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL_ACCOUNT | __GFP_ZERO);
2960 scaoh = (u32)((u64)(new_sca) >> 32);
2961 scaol = (u32)(u64)(new_sca) & ~0x3fU;
2963 kvm_s390_vcpu_block_all(kvm);
2964 write_lock(&kvm->arch.sca_lock);
2966 sca_copy_b_to_e(new_sca, old_sca);
2968 kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
2969 vcpu->arch.sie_block->scaoh = scaoh;
2970 vcpu->arch.sie_block->scaol = scaol;
2971 vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
2973 kvm->arch.sca = new_sca;
2974 kvm->arch.use_esca = 1;
2976 write_unlock(&kvm->arch.sca_lock);
2977 kvm_s390_vcpu_unblock_all(kvm);
2979 free_page((unsigned long)old_sca);
2981 VM_EVENT(kvm, 2, "Switched to ESCA (0x%pK -> 0x%pK)",
2982 old_sca, kvm->arch.sca);
2986 static int sca_can_add_vcpu(struct kvm *kvm, unsigned int id)
2990 if (!kvm_s390_use_sca_entries()) {
2991 if (id < KVM_MAX_VCPUS)
2995 if (id < KVM_S390_BSCA_CPU_SLOTS)
2997 if (!sclp.has_esca || !sclp.has_64bscao)
3000 mutex_lock(&kvm->lock);
3001 rc = kvm->arch.use_esca ? 0 : sca_switch_to_extended(kvm);
3002 mutex_unlock(&kvm->lock);
3004 return rc == 0 && id < KVM_S390_ESCA_CPU_SLOTS;
3007 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3008 static void __start_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3010 WARN_ON_ONCE(vcpu->arch.cputm_start != 0);
3011 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3012 vcpu->arch.cputm_start = get_tod_clock_fast();
3013 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3016 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3017 static void __stop_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3019 WARN_ON_ONCE(vcpu->arch.cputm_start == 0);
3020 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3021 vcpu->arch.sie_block->cputm -= get_tod_clock_fast() - vcpu->arch.cputm_start;
3022 vcpu->arch.cputm_start = 0;
3023 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3026 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3027 static void __enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3029 WARN_ON_ONCE(vcpu->arch.cputm_enabled);
3030 vcpu->arch.cputm_enabled = true;
3031 __start_cpu_timer_accounting(vcpu);
3034 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3035 static void __disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3037 WARN_ON_ONCE(!vcpu->arch.cputm_enabled);
3038 __stop_cpu_timer_accounting(vcpu);
3039 vcpu->arch.cputm_enabled = false;
3042 static void enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3044 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3045 __enable_cpu_timer_accounting(vcpu);
3049 static void disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3051 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3052 __disable_cpu_timer_accounting(vcpu);
3056 /* set the cpu timer - may only be called from the VCPU thread itself */
3057 void kvm_s390_set_cpu_timer(struct kvm_vcpu *vcpu, __u64 cputm)
3059 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3060 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3061 if (vcpu->arch.cputm_enabled)
3062 vcpu->arch.cputm_start = get_tod_clock_fast();
3063 vcpu->arch.sie_block->cputm = cputm;
3064 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3068 /* update and get the cpu timer - can also be called from other VCPU threads */
3069 __u64 kvm_s390_get_cpu_timer(struct kvm_vcpu *vcpu)
3074 if (unlikely(!vcpu->arch.cputm_enabled))
3075 return vcpu->arch.sie_block->cputm;
3077 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3079 seq = raw_read_seqcount(&vcpu->arch.cputm_seqcount);
3081 * If the writer would ever execute a read in the critical
3082 * section, e.g. in irq context, we have a deadlock.
3084 WARN_ON_ONCE((seq & 1) && smp_processor_id() == vcpu->cpu);
3085 value = vcpu->arch.sie_block->cputm;
3086 /* if cputm_start is 0, accounting is being started/stopped */
3087 if (likely(vcpu->arch.cputm_start))
3088 value -= get_tod_clock_fast() - vcpu->arch.cputm_start;
3089 } while (read_seqcount_retry(&vcpu->arch.cputm_seqcount, seq & ~1));
3094 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
3097 gmap_enable(vcpu->arch.enabled_gmap);
3098 kvm_s390_set_cpuflags(vcpu, CPUSTAT_RUNNING);
3099 if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
3100 __start_cpu_timer_accounting(vcpu);
3104 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
3107 if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
3108 __stop_cpu_timer_accounting(vcpu);
3109 kvm_s390_clear_cpuflags(vcpu, CPUSTAT_RUNNING);
3110 vcpu->arch.enabled_gmap = gmap_get_enabled();
3111 gmap_disable(vcpu->arch.enabled_gmap);
3115 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
3117 mutex_lock(&vcpu->kvm->lock);
3119 vcpu->arch.sie_block->epoch = vcpu->kvm->arch.epoch;
3120 vcpu->arch.sie_block->epdx = vcpu->kvm->arch.epdx;
3122 mutex_unlock(&vcpu->kvm->lock);
3123 if (!kvm_is_ucontrol(vcpu->kvm)) {
3124 vcpu->arch.gmap = vcpu->kvm->arch.gmap;
3127 if (test_kvm_facility(vcpu->kvm, 74) || vcpu->kvm->arch.user_instr0)
3128 vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
3129 /* make vcpu_load load the right gmap on the first trigger */
3130 vcpu->arch.enabled_gmap = vcpu->arch.gmap;
3133 static bool kvm_has_pckmo_subfunc(struct kvm *kvm, unsigned long nr)
3135 if (test_bit_inv(nr, (unsigned long *)&kvm->arch.model.subfuncs.pckmo) &&
3136 test_bit_inv(nr, (unsigned long *)&kvm_s390_available_subfunc.pckmo))
3141 static bool kvm_has_pckmo_ecc(struct kvm *kvm)
3143 /* At least one ECC subfunction must be present */
3144 return kvm_has_pckmo_subfunc(kvm, 32) ||
3145 kvm_has_pckmo_subfunc(kvm, 33) ||
3146 kvm_has_pckmo_subfunc(kvm, 34) ||
3147 kvm_has_pckmo_subfunc(kvm, 40) ||
3148 kvm_has_pckmo_subfunc(kvm, 41);
3152 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu)
3155 * If the AP instructions are not being interpreted and the MSAX3
3156 * facility is not configured for the guest, there is nothing to set up.
3158 if (!vcpu->kvm->arch.crypto.apie && !test_kvm_facility(vcpu->kvm, 76))
3161 vcpu->arch.sie_block->crycbd = vcpu->kvm->arch.crypto.crycbd;
3162 vcpu->arch.sie_block->ecb3 &= ~(ECB3_AES | ECB3_DEA);
3163 vcpu->arch.sie_block->eca &= ~ECA_APIE;
3164 vcpu->arch.sie_block->ecd &= ~ECD_ECC;
3166 if (vcpu->kvm->arch.crypto.apie)
3167 vcpu->arch.sie_block->eca |= ECA_APIE;
3169 /* Set up protected key support */
3170 if (vcpu->kvm->arch.crypto.aes_kw) {
3171 vcpu->arch.sie_block->ecb3 |= ECB3_AES;
3172 /* ecc is also wrapped with AES key */
3173 if (kvm_has_pckmo_ecc(vcpu->kvm))
3174 vcpu->arch.sie_block->ecd |= ECD_ECC;
3177 if (vcpu->kvm->arch.crypto.dea_kw)
3178 vcpu->arch.sie_block->ecb3 |= ECB3_DEA;
3181 void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu)
3183 free_page(vcpu->arch.sie_block->cbrlo);
3184 vcpu->arch.sie_block->cbrlo = 0;
3187 int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu)
3189 vcpu->arch.sie_block->cbrlo = get_zeroed_page(GFP_KERNEL_ACCOUNT);
3190 if (!vcpu->arch.sie_block->cbrlo)
3195 static void kvm_s390_vcpu_setup_model(struct kvm_vcpu *vcpu)
3197 struct kvm_s390_cpu_model *model = &vcpu->kvm->arch.model;
3199 vcpu->arch.sie_block->ibc = model->ibc;
3200 if (test_kvm_facility(vcpu->kvm, 7))
3201 vcpu->arch.sie_block->fac = (u32)(u64) model->fac_list;
3204 static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
3209 atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH |
3213 if (test_kvm_facility(vcpu->kvm, 78))
3214 kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED2);
3215 else if (test_kvm_facility(vcpu->kvm, 8))
3216 kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED);
3218 kvm_s390_vcpu_setup_model(vcpu);
3220 /* pgste_set_pte has special handling for !MACHINE_HAS_ESOP */
3221 if (MACHINE_HAS_ESOP)
3222 vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
3223 if (test_kvm_facility(vcpu->kvm, 9))
3224 vcpu->arch.sie_block->ecb |= ECB_SRSI;
3225 if (test_kvm_facility(vcpu->kvm, 73))
3226 vcpu->arch.sie_block->ecb |= ECB_TE;
3228 if (test_kvm_facility(vcpu->kvm, 8) && vcpu->kvm->arch.use_pfmfi)
3229 vcpu->arch.sie_block->ecb2 |= ECB2_PFMFI;
3230 if (test_kvm_facility(vcpu->kvm, 130))
3231 vcpu->arch.sie_block->ecb2 |= ECB2_IEP;
3232 vcpu->arch.sie_block->eca = ECA_MVPGI | ECA_PROTEXCI;
3234 vcpu->arch.sie_block->eca |= ECA_CEI;
3236 vcpu->arch.sie_block->eca |= ECA_IB;
3238 vcpu->arch.sie_block->eca |= ECA_SII;
3239 if (sclp.has_sigpif)
3240 vcpu->arch.sie_block->eca |= ECA_SIGPI;
3241 if (test_kvm_facility(vcpu->kvm, 129)) {
3242 vcpu->arch.sie_block->eca |= ECA_VX;
3243 vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
3245 if (test_kvm_facility(vcpu->kvm, 139))
3246 vcpu->arch.sie_block->ecd |= ECD_MEF;
3247 if (test_kvm_facility(vcpu->kvm, 156))
3248 vcpu->arch.sie_block->ecd |= ECD_ETOKENF;
3249 if (vcpu->arch.sie_block->gd) {
3250 vcpu->arch.sie_block->eca |= ECA_AIV;
3251 VCPU_EVENT(vcpu, 3, "AIV gisa format-%u enabled for cpu %03u",
3252 vcpu->arch.sie_block->gd & 0x3, vcpu->vcpu_id);
3254 vcpu->arch.sie_block->sdnxo = ((unsigned long) &vcpu->run->s.regs.sdnx)
3256 vcpu->arch.sie_block->riccbd = (unsigned long) &vcpu->run->s.regs.riccb;
3259 kvm_s390_set_cpuflags(vcpu, CPUSTAT_KSS);
3261 vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
3263 if (vcpu->kvm->arch.use_cmma) {
3264 rc = kvm_s390_vcpu_setup_cmma(vcpu);
3268 hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
3269 vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup;
3271 vcpu->arch.sie_block->hpid = HPID_KVM;
3273 kvm_s390_vcpu_crypto_setup(vcpu);
3275 mutex_lock(&vcpu->kvm->lock);
3276 if (kvm_s390_pv_is_protected(vcpu->kvm)) {
3277 rc = kvm_s390_pv_create_cpu(vcpu, &uvrc, &uvrrc);
3279 kvm_s390_vcpu_unsetup_cmma(vcpu);
3281 mutex_unlock(&vcpu->kvm->lock);
3286 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
3288 if (!kvm_is_ucontrol(kvm) && !sca_can_add_vcpu(kvm, id))
3293 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
3295 struct sie_page *sie_page;
3298 BUILD_BUG_ON(sizeof(struct sie_page) != 4096);
3299 sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL_ACCOUNT);
3303 vcpu->arch.sie_block = &sie_page->sie_block;
3304 vcpu->arch.sie_block->itdba = (unsigned long) &sie_page->itdb;
3306 /* the real guest size will always be smaller than msl */
3307 vcpu->arch.sie_block->mso = 0;
3308 vcpu->arch.sie_block->msl = sclp.hamax;
3310 vcpu->arch.sie_block->icpua = vcpu->vcpu_id;
3311 spin_lock_init(&vcpu->arch.local_int.lock);
3312 vcpu->arch.sie_block->gd = (u32)(u64)vcpu->kvm->arch.gisa_int.origin;
3313 if (vcpu->arch.sie_block->gd && sclp.has_gisaf)
3314 vcpu->arch.sie_block->gd |= GISA_FORMAT1;
3315 seqcount_init(&vcpu->arch.cputm_seqcount);
3317 vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
3318 kvm_clear_async_pf_completion_queue(vcpu);
3319 vcpu->run->kvm_valid_regs = KVM_SYNC_PREFIX |
3326 kvm_s390_set_prefix(vcpu, 0);
3327 if (test_kvm_facility(vcpu->kvm, 64))
3328 vcpu->run->kvm_valid_regs |= KVM_SYNC_RICCB;
3329 if (test_kvm_facility(vcpu->kvm, 82))
3330 vcpu->run->kvm_valid_regs |= KVM_SYNC_BPBC;
3331 if (test_kvm_facility(vcpu->kvm, 133))
3332 vcpu->run->kvm_valid_regs |= KVM_SYNC_GSCB;
3333 if (test_kvm_facility(vcpu->kvm, 156))
3334 vcpu->run->kvm_valid_regs |= KVM_SYNC_ETOKEN;
3335 /* fprs can be synchronized via vrs, even if the guest has no vx. With
3336 * MACHINE_HAS_VX, (load|store)_fpu_regs() will work with vrs format.
3339 vcpu->run->kvm_valid_regs |= KVM_SYNC_VRS;
3341 vcpu->run->kvm_valid_regs |= KVM_SYNC_FPRS;
3343 if (kvm_is_ucontrol(vcpu->kvm)) {
3344 rc = __kvm_ucontrol_vcpu_init(vcpu);
3346 goto out_free_sie_block;
3349 VM_EVENT(vcpu->kvm, 3, "create cpu %d at 0x%pK, sie block at 0x%pK",
3350 vcpu->vcpu_id, vcpu, vcpu->arch.sie_block);
3351 trace_kvm_s390_create_vcpu(vcpu->vcpu_id, vcpu, vcpu->arch.sie_block);
3353 rc = kvm_s390_vcpu_setup(vcpu);
3355 goto out_ucontrol_uninit;
3358 out_ucontrol_uninit:
3359 if (kvm_is_ucontrol(vcpu->kvm))
3360 gmap_remove(vcpu->arch.gmap);
3362 free_page((unsigned long)(vcpu->arch.sie_block));
3366 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
3368 return kvm_s390_vcpu_has_irq(vcpu, 0);
3371 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu)
3373 return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE);
3376 void kvm_s390_vcpu_block(struct kvm_vcpu *vcpu)
3378 atomic_or(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
3382 void kvm_s390_vcpu_unblock(struct kvm_vcpu *vcpu)
3384 atomic_andnot(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
3387 static void kvm_s390_vcpu_request(struct kvm_vcpu *vcpu)
3389 atomic_or(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
3393 bool kvm_s390_vcpu_sie_inhibited(struct kvm_vcpu *vcpu)
3395 return atomic_read(&vcpu->arch.sie_block->prog20) &
3396 (PROG_BLOCK_SIE | PROG_REQUEST);
3399 static void kvm_s390_vcpu_request_handled(struct kvm_vcpu *vcpu)
3401 atomic_andnot(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
3405 * Kick a guest cpu out of (v)SIE and wait until (v)SIE is not running.
3406 * If the CPU is not running (e.g. waiting as idle) the function will
3407 * return immediately. */
3408 void exit_sie(struct kvm_vcpu *vcpu)
3410 kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOP_INT);
3411 kvm_s390_vsie_kick(vcpu);
3412 while (vcpu->arch.sie_block->prog0c & PROG_IN_SIE)
3416 /* Kick a guest cpu out of SIE to process a request synchronously */
3417 void kvm_s390_sync_request(int req, struct kvm_vcpu *vcpu)
3419 kvm_make_request(req, vcpu);
3420 kvm_s390_vcpu_request(vcpu);
3423 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
3426 struct kvm *kvm = gmap->private;
3427 struct kvm_vcpu *vcpu;
3428 unsigned long prefix;
3431 if (gmap_is_shadow(gmap))
3433 if (start >= 1UL << 31)
3434 /* We are only interested in prefix pages */
3436 kvm_for_each_vcpu(i, vcpu, kvm) {
3437 /* match against both prefix pages */
3438 prefix = kvm_s390_get_prefix(vcpu);
3439 if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
3440 VCPU_EVENT(vcpu, 2, "gmap notifier for %lx-%lx",
3442 kvm_s390_sync_request(KVM_REQ_MMU_RELOAD, vcpu);
3447 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
3449 /* do not poll with more than halt_poll_max_steal percent of steal time */
3450 if (S390_lowcore.avg_steal_timer * 100 / (TICK_USEC << 12) >=
3451 halt_poll_max_steal) {
3452 vcpu->stat.halt_no_poll_steal++;
3458 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
3460 /* kvm common code refers to this, but never calls it */
3465 static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu,
3466 struct kvm_one_reg *reg)
3471 case KVM_REG_S390_TODPR:
3472 r = put_user(vcpu->arch.sie_block->todpr,
3473 (u32 __user *)reg->addr);
3475 case KVM_REG_S390_EPOCHDIFF:
3476 r = put_user(vcpu->arch.sie_block->epoch,
3477 (u64 __user *)reg->addr);
3479 case KVM_REG_S390_CPU_TIMER:
3480 r = put_user(kvm_s390_get_cpu_timer(vcpu),
3481 (u64 __user *)reg->addr);
3483 case KVM_REG_S390_CLOCK_COMP:
3484 r = put_user(vcpu->arch.sie_block->ckc,
3485 (u64 __user *)reg->addr);
3487 case KVM_REG_S390_PFTOKEN:
3488 r = put_user(vcpu->arch.pfault_token,
3489 (u64 __user *)reg->addr);
3491 case KVM_REG_S390_PFCOMPARE:
3492 r = put_user(vcpu->arch.pfault_compare,
3493 (u64 __user *)reg->addr);
3495 case KVM_REG_S390_PFSELECT:
3496 r = put_user(vcpu->arch.pfault_select,
3497 (u64 __user *)reg->addr);
3499 case KVM_REG_S390_PP:
3500 r = put_user(vcpu->arch.sie_block->pp,
3501 (u64 __user *)reg->addr);
3503 case KVM_REG_S390_GBEA:
3504 r = put_user(vcpu->arch.sie_block->gbea,
3505 (u64 __user *)reg->addr);
3514 static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu,
3515 struct kvm_one_reg *reg)
3521 case KVM_REG_S390_TODPR:
3522 r = get_user(vcpu->arch.sie_block->todpr,
3523 (u32 __user *)reg->addr);
3525 case KVM_REG_S390_EPOCHDIFF:
3526 r = get_user(vcpu->arch.sie_block->epoch,
3527 (u64 __user *)reg->addr);
3529 case KVM_REG_S390_CPU_TIMER:
3530 r = get_user(val, (u64 __user *)reg->addr);
3532 kvm_s390_set_cpu_timer(vcpu, val);
3534 case KVM_REG_S390_CLOCK_COMP:
3535 r = get_user(vcpu->arch.sie_block->ckc,
3536 (u64 __user *)reg->addr);
3538 case KVM_REG_S390_PFTOKEN:
3539 r = get_user(vcpu->arch.pfault_token,
3540 (u64 __user *)reg->addr);
3541 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
3542 kvm_clear_async_pf_completion_queue(vcpu);
3544 case KVM_REG_S390_PFCOMPARE:
3545 r = get_user(vcpu->arch.pfault_compare,
3546 (u64 __user *)reg->addr);
3548 case KVM_REG_S390_PFSELECT:
3549 r = get_user(vcpu->arch.pfault_select,
3550 (u64 __user *)reg->addr);
3552 case KVM_REG_S390_PP:
3553 r = get_user(vcpu->arch.sie_block->pp,
3554 (u64 __user *)reg->addr);
3556 case KVM_REG_S390_GBEA:
3557 r = get_user(vcpu->arch.sie_block->gbea,
3558 (u64 __user *)reg->addr);
3567 static void kvm_arch_vcpu_ioctl_normal_reset(struct kvm_vcpu *vcpu)
3569 vcpu->arch.sie_block->gpsw.mask &= ~PSW_MASK_RI;
3570 vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
3571 memset(vcpu->run->s.regs.riccb, 0, sizeof(vcpu->run->s.regs.riccb));
3573 kvm_clear_async_pf_completion_queue(vcpu);
3574 if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
3575 kvm_s390_vcpu_stop(vcpu);
3576 kvm_s390_clear_local_irqs(vcpu);
3579 static void kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu)
3581 /* Initial reset is a superset of the normal reset */
3582 kvm_arch_vcpu_ioctl_normal_reset(vcpu);
3585 * This equals initial cpu reset in pop, but we don't switch to ESA.
3586 * We do not only reset the internal data, but also ...
3588 vcpu->arch.sie_block->gpsw.mask = 0;
3589 vcpu->arch.sie_block->gpsw.addr = 0;
3590 kvm_s390_set_prefix(vcpu, 0);
3591 kvm_s390_set_cpu_timer(vcpu, 0);
3592 vcpu->arch.sie_block->ckc = 0;
3593 memset(vcpu->arch.sie_block->gcr, 0, sizeof(vcpu->arch.sie_block->gcr));
3594 vcpu->arch.sie_block->gcr[0] = CR0_INITIAL_MASK;
3595 vcpu->arch.sie_block->gcr[14] = CR14_INITIAL_MASK;
3597 /* ... the data in sync regs */
3598 memset(vcpu->run->s.regs.crs, 0, sizeof(vcpu->run->s.regs.crs));
3599 vcpu->run->s.regs.ckc = 0;
3600 vcpu->run->s.regs.crs[0] = CR0_INITIAL_MASK;
3601 vcpu->run->s.regs.crs[14] = CR14_INITIAL_MASK;
3602 vcpu->run->psw_addr = 0;
3603 vcpu->run->psw_mask = 0;
3604 vcpu->run->s.regs.todpr = 0;
3605 vcpu->run->s.regs.cputm = 0;
3606 vcpu->run->s.regs.ckc = 0;
3607 vcpu->run->s.regs.pp = 0;
3608 vcpu->run->s.regs.gbea = 1;
3609 vcpu->run->s.regs.fpc = 0;
3611 * Do not reset these registers in the protected case, as some of
3612 * them are overlayed and they are not accessible in this case
3615 if (!kvm_s390_pv_cpu_is_protected(vcpu)) {
3616 vcpu->arch.sie_block->gbea = 1;
3617 vcpu->arch.sie_block->pp = 0;
3618 vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
3619 vcpu->arch.sie_block->todpr = 0;
3623 static void kvm_arch_vcpu_ioctl_clear_reset(struct kvm_vcpu *vcpu)
3625 struct kvm_sync_regs *regs = &vcpu->run->s.regs;
3627 /* Clear reset is a superset of the initial reset */
3628 kvm_arch_vcpu_ioctl_initial_reset(vcpu);
3630 memset(®s->gprs, 0, sizeof(regs->gprs));
3631 memset(®s->vrs, 0, sizeof(regs->vrs));
3632 memset(®s->acrs, 0, sizeof(regs->acrs));
3633 memset(®s->gscb, 0, sizeof(regs->gscb));
3636 regs->etoken_extension = 0;
3639 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
3642 memcpy(&vcpu->run->s.regs.gprs, ®s->gprs, sizeof(regs->gprs));
3647 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
3650 memcpy(®s->gprs, &vcpu->run->s.regs.gprs, sizeof(regs->gprs));
3655 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
3656 struct kvm_sregs *sregs)
3660 memcpy(&vcpu->run->s.regs.acrs, &sregs->acrs, sizeof(sregs->acrs));
3661 memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs));
3667 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
3668 struct kvm_sregs *sregs)
3672 memcpy(&sregs->acrs, &vcpu->run->s.regs.acrs, sizeof(sregs->acrs));
3673 memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs));
3679 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
3685 if (test_fp_ctl(fpu->fpc)) {
3689 vcpu->run->s.regs.fpc = fpu->fpc;
3691 convert_fp_to_vx((__vector128 *) vcpu->run->s.regs.vrs,
3692 (freg_t *) fpu->fprs);
3694 memcpy(vcpu->run->s.regs.fprs, &fpu->fprs, sizeof(fpu->fprs));
3701 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
3705 /* make sure we have the latest values */
3708 convert_vx_to_fp((freg_t *) fpu->fprs,
3709 (__vector128 *) vcpu->run->s.regs.vrs);
3711 memcpy(fpu->fprs, vcpu->run->s.regs.fprs, sizeof(fpu->fprs));
3712 fpu->fpc = vcpu->run->s.regs.fpc;
3718 static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw)
3722 if (!is_vcpu_stopped(vcpu))
3725 vcpu->run->psw_mask = psw.mask;
3726 vcpu->run->psw_addr = psw.addr;
3731 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
3732 struct kvm_translation *tr)
3734 return -EINVAL; /* not implemented yet */
3737 #define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
3738 KVM_GUESTDBG_USE_HW_BP | \
3739 KVM_GUESTDBG_ENABLE)
3741 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
3742 struct kvm_guest_debug *dbg)
3748 vcpu->guest_debug = 0;
3749 kvm_s390_clear_bp_data(vcpu);
3751 if (dbg->control & ~VALID_GUESTDBG_FLAGS) {
3755 if (!sclp.has_gpere) {
3760 if (dbg->control & KVM_GUESTDBG_ENABLE) {
3761 vcpu->guest_debug = dbg->control;
3762 /* enforce guest PER */
3763 kvm_s390_set_cpuflags(vcpu, CPUSTAT_P);
3765 if (dbg->control & KVM_GUESTDBG_USE_HW_BP)
3766 rc = kvm_s390_import_bp_data(vcpu, dbg);
3768 kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
3769 vcpu->arch.guestdbg.last_bp = 0;
3773 vcpu->guest_debug = 0;
3774 kvm_s390_clear_bp_data(vcpu);
3775 kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
3783 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
3784 struct kvm_mp_state *mp_state)
3790 /* CHECK_STOP and LOAD are not supported yet */
3791 ret = is_vcpu_stopped(vcpu) ? KVM_MP_STATE_STOPPED :
3792 KVM_MP_STATE_OPERATING;
3798 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
3799 struct kvm_mp_state *mp_state)
3805 /* user space knows about this interface - let it control the state */
3806 vcpu->kvm->arch.user_cpu_state_ctrl = 1;
3808 switch (mp_state->mp_state) {
3809 case KVM_MP_STATE_STOPPED:
3810 rc = kvm_s390_vcpu_stop(vcpu);
3812 case KVM_MP_STATE_OPERATING:
3813 rc = kvm_s390_vcpu_start(vcpu);
3815 case KVM_MP_STATE_LOAD:
3816 if (!kvm_s390_pv_cpu_is_protected(vcpu)) {
3820 rc = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_OPR_LOAD);
3822 case KVM_MP_STATE_CHECK_STOP:
3823 fallthrough; /* CHECK_STOP and LOAD are not supported yet */
3832 static bool ibs_enabled(struct kvm_vcpu *vcpu)
3834 return kvm_s390_test_cpuflags(vcpu, CPUSTAT_IBS);
3837 static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
3840 kvm_s390_vcpu_request_handled(vcpu);
3841 if (!kvm_request_pending(vcpu))
3844 * We use MMU_RELOAD just to re-arm the ipte notifier for the
3845 * guest prefix page. gmap_mprotect_notify will wait on the ptl lock.
3846 * This ensures that the ipte instruction for this request has
3847 * already finished. We might race against a second unmapper that
3848 * wants to set the blocking bit. Lets just retry the request loop.
3850 if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu)) {
3852 rc = gmap_mprotect_notify(vcpu->arch.gmap,
3853 kvm_s390_get_prefix(vcpu),
3854 PAGE_SIZE * 2, PROT_WRITE);
3856 kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu);
3862 if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
3863 vcpu->arch.sie_block->ihcpu = 0xffff;
3867 if (kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu)) {
3868 if (!ibs_enabled(vcpu)) {
3869 trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 1);
3870 kvm_s390_set_cpuflags(vcpu, CPUSTAT_IBS);
3875 if (kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu)) {
3876 if (ibs_enabled(vcpu)) {
3877 trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 0);
3878 kvm_s390_clear_cpuflags(vcpu, CPUSTAT_IBS);
3883 if (kvm_check_request(KVM_REQ_ICPT_OPEREXC, vcpu)) {
3884 vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
3888 if (kvm_check_request(KVM_REQ_START_MIGRATION, vcpu)) {
3890 * Disable CMM virtualization; we will emulate the ESSA
3891 * instruction manually, in order to provide additional
3892 * functionalities needed for live migration.
3894 vcpu->arch.sie_block->ecb2 &= ~ECB2_CMMA;
3898 if (kvm_check_request(KVM_REQ_STOP_MIGRATION, vcpu)) {
3900 * Re-enable CMM virtualization if CMMA is available and
3901 * CMM has been used.
3903 if ((vcpu->kvm->arch.use_cmma) &&
3904 (vcpu->kvm->mm->context.uses_cmm))
3905 vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
3909 /* nothing to do, just clear the request */
3910 kvm_clear_request(KVM_REQ_UNHALT, vcpu);
3911 /* we left the vsie handler, nothing to do, just clear the request */
3912 kvm_clear_request(KVM_REQ_VSIE_RESTART, vcpu);
3917 void kvm_s390_set_tod_clock(struct kvm *kvm,
3918 const struct kvm_s390_vm_tod_clock *gtod)
3920 struct kvm_vcpu *vcpu;
3921 union tod_clock clk;
3924 mutex_lock(&kvm->lock);
3927 store_tod_clock_ext(&clk);
3929 kvm->arch.epoch = gtod->tod - clk.tod;
3931 if (test_kvm_facility(kvm, 139)) {
3932 kvm->arch.epdx = gtod->epoch_idx - clk.ei;
3933 if (kvm->arch.epoch > gtod->tod)
3934 kvm->arch.epdx -= 1;
3937 kvm_s390_vcpu_block_all(kvm);
3938 kvm_for_each_vcpu(i, vcpu, kvm) {
3939 vcpu->arch.sie_block->epoch = kvm->arch.epoch;
3940 vcpu->arch.sie_block->epdx = kvm->arch.epdx;
3943 kvm_s390_vcpu_unblock_all(kvm);
3945 mutex_unlock(&kvm->lock);
3949 * kvm_arch_fault_in_page - fault-in guest page if necessary
3950 * @vcpu: The corresponding virtual cpu
3951 * @gpa: Guest physical address
3952 * @writable: Whether the page should be writable or not
3954 * Make sure that a guest page has been faulted-in on the host.
3956 * Return: Zero on success, negative error code otherwise.
3958 long kvm_arch_fault_in_page(struct kvm_vcpu *vcpu, gpa_t gpa, int writable)
3960 return gmap_fault(vcpu->arch.gmap, gpa,
3961 writable ? FAULT_FLAG_WRITE : 0);
3964 static void __kvm_inject_pfault_token(struct kvm_vcpu *vcpu, bool start_token,
3965 unsigned long token)
3967 struct kvm_s390_interrupt inti;
3968 struct kvm_s390_irq irq;
3971 irq.u.ext.ext_params2 = token;
3972 irq.type = KVM_S390_INT_PFAULT_INIT;
3973 WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &irq));
3975 inti.type = KVM_S390_INT_PFAULT_DONE;
3976 inti.parm64 = token;
3977 WARN_ON_ONCE(kvm_s390_inject_vm(vcpu->kvm, &inti));
3981 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
3982 struct kvm_async_pf *work)
3984 trace_kvm_s390_pfault_init(vcpu, work->arch.pfault_token);
3985 __kvm_inject_pfault_token(vcpu, true, work->arch.pfault_token);
3990 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
3991 struct kvm_async_pf *work)
3993 trace_kvm_s390_pfault_done(vcpu, work->arch.pfault_token);
3994 __kvm_inject_pfault_token(vcpu, false, work->arch.pfault_token);
3997 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
3998 struct kvm_async_pf *work)
4000 /* s390 will always inject the page directly */
4003 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu)
4006 * s390 will always inject the page directly,
4007 * but we still want check_async_completion to cleanup
4012 static bool kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu)
4015 struct kvm_arch_async_pf arch;
4017 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
4019 if ((vcpu->arch.sie_block->gpsw.mask & vcpu->arch.pfault_select) !=
4020 vcpu->arch.pfault_compare)
4022 if (psw_extint_disabled(vcpu))
4024 if (kvm_s390_vcpu_has_irq(vcpu, 0))
4026 if (!(vcpu->arch.sie_block->gcr[0] & CR0_SERVICE_SIGNAL_SUBMASK))
4028 if (!vcpu->arch.gmap->pfault_enabled)
4031 hva = gfn_to_hva(vcpu->kvm, gpa_to_gfn(current->thread.gmap_addr));
4032 hva += current->thread.gmap_addr & ~PAGE_MASK;
4033 if (read_guest_real(vcpu, vcpu->arch.pfault_token, &arch.pfault_token, 8))
4036 return kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch);
4039 static int vcpu_pre_run(struct kvm_vcpu *vcpu)
4044 * On s390 notifications for arriving pages will be delivered directly
4045 * to the guest but the house keeping for completed pfaults is
4046 * handled outside the worker.
4048 kvm_check_async_pf_completion(vcpu);
4050 vcpu->arch.sie_block->gg14 = vcpu->run->s.regs.gprs[14];
4051 vcpu->arch.sie_block->gg15 = vcpu->run->s.regs.gprs[15];
4056 if (!kvm_is_ucontrol(vcpu->kvm)) {
4057 rc = kvm_s390_deliver_pending_interrupts(vcpu);
4062 rc = kvm_s390_handle_requests(vcpu);
4066 if (guestdbg_enabled(vcpu)) {
4067 kvm_s390_backup_guest_per_regs(vcpu);
4068 kvm_s390_patch_guest_per_regs(vcpu);
4071 clear_bit(vcpu->vcpu_id, vcpu->kvm->arch.gisa_int.kicked_mask);
4073 vcpu->arch.sie_block->icptcode = 0;
4074 cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags);
4075 VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags);
4076 trace_kvm_s390_sie_enter(vcpu, cpuflags);
4081 static int vcpu_post_run_fault_in_sie(struct kvm_vcpu *vcpu)
4083 struct kvm_s390_pgm_info pgm_info = {
4084 .code = PGM_ADDRESSING,
4089 VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction");
4090 trace_kvm_s390_sie_fault(vcpu);
4093 * We want to inject an addressing exception, which is defined as a
4094 * suppressing or terminating exception. However, since we came here
4095 * by a DAT access exception, the PSW still points to the faulting
4096 * instruction since DAT exceptions are nullifying. So we've got
4097 * to look up the current opcode to get the length of the instruction
4098 * to be able to forward the PSW.
4100 rc = read_guest_instr(vcpu, vcpu->arch.sie_block->gpsw.addr, &opcode, 1);
4101 ilen = insn_length(opcode);
4105 /* Instruction-Fetching Exceptions - we can't detect the ilen.
4106 * Forward by arbitrary ilc, injection will take care of
4107 * nullification if necessary.
4109 pgm_info = vcpu->arch.pgm;
4112 pgm_info.flags = ilen | KVM_S390_PGM_FLAGS_ILC_VALID;
4113 kvm_s390_forward_psw(vcpu, ilen);
4114 return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
4117 static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
4119 struct mcck_volatile_info *mcck_info;
4120 struct sie_page *sie_page;
4122 VCPU_EVENT(vcpu, 6, "exit sie icptcode %d",
4123 vcpu->arch.sie_block->icptcode);
4124 trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode);
4126 if (guestdbg_enabled(vcpu))
4127 kvm_s390_restore_guest_per_regs(vcpu);
4129 vcpu->run->s.regs.gprs[14] = vcpu->arch.sie_block->gg14;
4130 vcpu->run->s.regs.gprs[15] = vcpu->arch.sie_block->gg15;
4132 if (exit_reason == -EINTR) {
4133 VCPU_EVENT(vcpu, 3, "%s", "machine check");
4134 sie_page = container_of(vcpu->arch.sie_block,
4135 struct sie_page, sie_block);
4136 mcck_info = &sie_page->mcck_info;
4137 kvm_s390_reinject_machine_check(vcpu, mcck_info);
4141 if (vcpu->arch.sie_block->icptcode > 0) {
4142 int rc = kvm_handle_sie_intercept(vcpu);
4144 if (rc != -EOPNOTSUPP)
4146 vcpu->run->exit_reason = KVM_EXIT_S390_SIEIC;
4147 vcpu->run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode;
4148 vcpu->run->s390_sieic.ipa = vcpu->arch.sie_block->ipa;
4149 vcpu->run->s390_sieic.ipb = vcpu->arch.sie_block->ipb;
4151 } else if (exit_reason != -EFAULT) {
4152 vcpu->stat.exit_null++;
4154 } else if (kvm_is_ucontrol(vcpu->kvm)) {
4155 vcpu->run->exit_reason = KVM_EXIT_S390_UCONTROL;
4156 vcpu->run->s390_ucontrol.trans_exc_code =
4157 current->thread.gmap_addr;
4158 vcpu->run->s390_ucontrol.pgm_code = 0x10;
4160 } else if (current->thread.gmap_pfault) {
4161 trace_kvm_s390_major_guest_pfault(vcpu);
4162 current->thread.gmap_pfault = 0;
4163 if (kvm_arch_setup_async_pf(vcpu))
4165 vcpu->stat.pfault_sync++;
4166 return kvm_arch_fault_in_page(vcpu, current->thread.gmap_addr, 1);
4168 return vcpu_post_run_fault_in_sie(vcpu);
4171 #define PSW_INT_MASK (PSW_MASK_EXT | PSW_MASK_IO | PSW_MASK_MCHECK)
4172 static int __vcpu_run(struct kvm_vcpu *vcpu)
4174 int rc, exit_reason;
4175 struct sie_page *sie_page = (struct sie_page *)vcpu->arch.sie_block;
4178 * We try to hold kvm->srcu during most of vcpu_run (except when run-
4179 * ning the guest), so that memslots (and other stuff) are protected
4181 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
4184 rc = vcpu_pre_run(vcpu);
4188 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
4190 * As PF_VCPU will be used in fault handler, between
4191 * guest_enter and guest_exit should be no uaccess.
4193 local_irq_disable();
4194 guest_enter_irqoff();
4195 __disable_cpu_timer_accounting(vcpu);
4197 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4198 memcpy(sie_page->pv_grregs,
4199 vcpu->run->s.regs.gprs,
4200 sizeof(sie_page->pv_grregs));
4202 if (test_cpu_flag(CIF_FPU))
4204 exit_reason = sie64a(vcpu->arch.sie_block,
4205 vcpu->run->s.regs.gprs);
4206 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4207 memcpy(vcpu->run->s.regs.gprs,
4208 sie_page->pv_grregs,
4209 sizeof(sie_page->pv_grregs));
4211 * We're not allowed to inject interrupts on intercepts
4212 * that leave the guest state in an "in-between" state
4213 * where the next SIE entry will do a continuation.
4214 * Fence interrupts in our "internal" PSW.
4216 if (vcpu->arch.sie_block->icptcode == ICPT_PV_INSTR ||
4217 vcpu->arch.sie_block->icptcode == ICPT_PV_PREF) {
4218 vcpu->arch.sie_block->gpsw.mask &= ~PSW_INT_MASK;
4221 local_irq_disable();
4222 __enable_cpu_timer_accounting(vcpu);
4223 guest_exit_irqoff();
4225 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
4227 rc = vcpu_post_run(vcpu, exit_reason);
4228 } while (!signal_pending(current) && !guestdbg_exit_pending(vcpu) && !rc);
4230 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
4234 static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
4236 struct kvm_run *kvm_run = vcpu->run;
4237 struct runtime_instr_cb *riccb;
4240 riccb = (struct runtime_instr_cb *) &kvm_run->s.regs.riccb;
4241 gscb = (struct gs_cb *) &kvm_run->s.regs.gscb;
4242 vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask;
4243 vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr;
4244 if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
4245 vcpu->arch.sie_block->todpr = kvm_run->s.regs.todpr;
4246 vcpu->arch.sie_block->pp = kvm_run->s.regs.pp;
4247 vcpu->arch.sie_block->gbea = kvm_run->s.regs.gbea;
4249 if (kvm_run->kvm_dirty_regs & KVM_SYNC_PFAULT) {
4250 vcpu->arch.pfault_token = kvm_run->s.regs.pft;
4251 vcpu->arch.pfault_select = kvm_run->s.regs.pfs;
4252 vcpu->arch.pfault_compare = kvm_run->s.regs.pfc;
4253 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
4254 kvm_clear_async_pf_completion_queue(vcpu);
4256 if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
4257 vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
4258 vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
4261 * If userspace sets the riccb (e.g. after migration) to a valid state,
4262 * we should enable RI here instead of doing the lazy enablement.
4264 if ((kvm_run->kvm_dirty_regs & KVM_SYNC_RICCB) &&
4265 test_kvm_facility(vcpu->kvm, 64) &&
4267 !(vcpu->arch.sie_block->ecb3 & ECB3_RI)) {
4268 VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (sync_regs)");
4269 vcpu->arch.sie_block->ecb3 |= ECB3_RI;
4272 * If userspace sets the gscb (e.g. after migration) to non-zero,
4273 * we should enable GS here instead of doing the lazy enablement.
4275 if ((kvm_run->kvm_dirty_regs & KVM_SYNC_GSCB) &&
4276 test_kvm_facility(vcpu->kvm, 133) &&
4278 !vcpu->arch.gs_enabled) {
4279 VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (sync_regs)");
4280 vcpu->arch.sie_block->ecb |= ECB_GS;
4281 vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
4282 vcpu->arch.gs_enabled = 1;
4284 if ((kvm_run->kvm_dirty_regs & KVM_SYNC_BPBC) &&
4285 test_kvm_facility(vcpu->kvm, 82)) {
4286 vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
4287 vcpu->arch.sie_block->fpf |= kvm_run->s.regs.bpbc ? FPF_BPBC : 0;
4289 if (MACHINE_HAS_GS) {
4291 __ctl_set_bit(2, 4);
4292 if (current->thread.gs_cb) {
4293 vcpu->arch.host_gscb = current->thread.gs_cb;
4294 save_gs_cb(vcpu->arch.host_gscb);
4296 if (vcpu->arch.gs_enabled) {
4297 current->thread.gs_cb = (struct gs_cb *)
4298 &vcpu->run->s.regs.gscb;
4299 restore_gs_cb(current->thread.gs_cb);
4303 /* SIE will load etoken directly from SDNX and therefore kvm_run */
4306 static void sync_regs(struct kvm_vcpu *vcpu)
4308 struct kvm_run *kvm_run = vcpu->run;
4310 if (kvm_run->kvm_dirty_regs & KVM_SYNC_PREFIX)
4311 kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix);
4312 if (kvm_run->kvm_dirty_regs & KVM_SYNC_CRS) {
4313 memcpy(&vcpu->arch.sie_block->gcr, &kvm_run->s.regs.crs, 128);
4314 /* some control register changes require a tlb flush */
4315 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4317 if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
4318 kvm_s390_set_cpu_timer(vcpu, kvm_run->s.regs.cputm);
4319 vcpu->arch.sie_block->ckc = kvm_run->s.regs.ckc;
4321 save_access_regs(vcpu->arch.host_acrs);
4322 restore_access_regs(vcpu->run->s.regs.acrs);
4323 /* save host (userspace) fprs/vrs */
4325 vcpu->arch.host_fpregs.fpc = current->thread.fpu.fpc;
4326 vcpu->arch.host_fpregs.regs = current->thread.fpu.regs;
4328 current->thread.fpu.regs = vcpu->run->s.regs.vrs;
4330 current->thread.fpu.regs = vcpu->run->s.regs.fprs;
4331 current->thread.fpu.fpc = vcpu->run->s.regs.fpc;
4332 if (test_fp_ctl(current->thread.fpu.fpc))
4333 /* User space provided an invalid FPC, let's clear it */
4334 current->thread.fpu.fpc = 0;
4336 /* Sync fmt2 only data */
4337 if (likely(!kvm_s390_pv_cpu_is_protected(vcpu))) {
4338 sync_regs_fmt2(vcpu);
4341 * In several places we have to modify our internal view to
4342 * not do things that are disallowed by the ultravisor. For
4343 * example we must not inject interrupts after specific exits
4344 * (e.g. 112 prefix page not secure). We do this by turning
4345 * off the machine check, external and I/O interrupt bits
4346 * of our PSW copy. To avoid getting validity intercepts, we
4347 * do only accept the condition code from userspace.
4349 vcpu->arch.sie_block->gpsw.mask &= ~PSW_MASK_CC;
4350 vcpu->arch.sie_block->gpsw.mask |= kvm_run->psw_mask &
4354 kvm_run->kvm_dirty_regs = 0;
4357 static void store_regs_fmt2(struct kvm_vcpu *vcpu)
4359 struct kvm_run *kvm_run = vcpu->run;
4361 kvm_run->s.regs.todpr = vcpu->arch.sie_block->todpr;
4362 kvm_run->s.regs.pp = vcpu->arch.sie_block->pp;
4363 kvm_run->s.regs.gbea = vcpu->arch.sie_block->gbea;
4364 kvm_run->s.regs.bpbc = (vcpu->arch.sie_block->fpf & FPF_BPBC) == FPF_BPBC;
4365 kvm_run->s.regs.diag318 = vcpu->arch.diag318_info.val;
4366 if (MACHINE_HAS_GS) {
4368 __ctl_set_bit(2, 4);
4369 if (vcpu->arch.gs_enabled)
4370 save_gs_cb(current->thread.gs_cb);
4371 current->thread.gs_cb = vcpu->arch.host_gscb;
4372 restore_gs_cb(vcpu->arch.host_gscb);
4373 if (!vcpu->arch.host_gscb)
4374 __ctl_clear_bit(2, 4);
4375 vcpu->arch.host_gscb = NULL;
4378 /* SIE will save etoken directly into SDNX and therefore kvm_run */
4381 static void store_regs(struct kvm_vcpu *vcpu)
4383 struct kvm_run *kvm_run = vcpu->run;
4385 kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask;
4386 kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr;
4387 kvm_run->s.regs.prefix = kvm_s390_get_prefix(vcpu);
4388 memcpy(&kvm_run->s.regs.crs, &vcpu->arch.sie_block->gcr, 128);
4389 kvm_run->s.regs.cputm = kvm_s390_get_cpu_timer(vcpu);
4390 kvm_run->s.regs.ckc = vcpu->arch.sie_block->ckc;
4391 kvm_run->s.regs.pft = vcpu->arch.pfault_token;
4392 kvm_run->s.regs.pfs = vcpu->arch.pfault_select;
4393 kvm_run->s.regs.pfc = vcpu->arch.pfault_compare;
4394 save_access_regs(vcpu->run->s.regs.acrs);
4395 restore_access_regs(vcpu->arch.host_acrs);
4396 /* Save guest register state */
4398 vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
4399 /* Restore will be done lazily at return */
4400 current->thread.fpu.fpc = vcpu->arch.host_fpregs.fpc;
4401 current->thread.fpu.regs = vcpu->arch.host_fpregs.regs;
4402 if (likely(!kvm_s390_pv_cpu_is_protected(vcpu)))
4403 store_regs_fmt2(vcpu);
4406 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
4408 struct kvm_run *kvm_run = vcpu->run;
4411 if (kvm_run->immediate_exit)
4414 if (kvm_run->kvm_valid_regs & ~KVM_SYNC_S390_VALID_FIELDS ||
4415 kvm_run->kvm_dirty_regs & ~KVM_SYNC_S390_VALID_FIELDS)
4420 if (guestdbg_exit_pending(vcpu)) {
4421 kvm_s390_prepare_debug_exit(vcpu);
4426 kvm_sigset_activate(vcpu);
4429 * no need to check the return value of vcpu_start as it can only have
4430 * an error for protvirt, but protvirt means user cpu state
4432 if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm)) {
4433 kvm_s390_vcpu_start(vcpu);
4434 } else if (is_vcpu_stopped(vcpu)) {
4435 pr_err_ratelimited("can't run stopped vcpu %d\n",
4442 enable_cpu_timer_accounting(vcpu);
4445 rc = __vcpu_run(vcpu);
4447 if (signal_pending(current) && !rc) {
4448 kvm_run->exit_reason = KVM_EXIT_INTR;
4452 if (guestdbg_exit_pending(vcpu) && !rc) {
4453 kvm_s390_prepare_debug_exit(vcpu);
4457 if (rc == -EREMOTE) {
4458 /* userspace support is needed, kvm_run has been prepared */
4462 disable_cpu_timer_accounting(vcpu);
4465 kvm_sigset_deactivate(vcpu);
4467 vcpu->stat.exit_userspace++;
4474 * store status at address
4475 * we use have two special cases:
4476 * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
4477 * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
4479 int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
4481 unsigned char archmode = 1;
4482 freg_t fprs[NUM_FPRS];
4487 px = kvm_s390_get_prefix(vcpu);
4488 if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
4489 if (write_guest_abs(vcpu, 163, &archmode, 1))
4492 } else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
4493 if (write_guest_real(vcpu, 163, &archmode, 1))
4497 gpa -= __LC_FPREGS_SAVE_AREA;
4499 /* manually convert vector registers if necessary */
4500 if (MACHINE_HAS_VX) {
4501 convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
4502 rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4505 rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4506 vcpu->run->s.regs.fprs, 128);
4508 rc |= write_guest_abs(vcpu, gpa + __LC_GPREGS_SAVE_AREA,
4509 vcpu->run->s.regs.gprs, 128);
4510 rc |= write_guest_abs(vcpu, gpa + __LC_PSW_SAVE_AREA,
4511 &vcpu->arch.sie_block->gpsw, 16);
4512 rc |= write_guest_abs(vcpu, gpa + __LC_PREFIX_SAVE_AREA,
4514 rc |= write_guest_abs(vcpu, gpa + __LC_FP_CREG_SAVE_AREA,
4515 &vcpu->run->s.regs.fpc, 4);
4516 rc |= write_guest_abs(vcpu, gpa + __LC_TOD_PROGREG_SAVE_AREA,
4517 &vcpu->arch.sie_block->todpr, 4);
4518 cputm = kvm_s390_get_cpu_timer(vcpu);
4519 rc |= write_guest_abs(vcpu, gpa + __LC_CPU_TIMER_SAVE_AREA,
4521 clkcomp = vcpu->arch.sie_block->ckc >> 8;
4522 rc |= write_guest_abs(vcpu, gpa + __LC_CLOCK_COMP_SAVE_AREA,
4524 rc |= write_guest_abs(vcpu, gpa + __LC_AREGS_SAVE_AREA,
4525 &vcpu->run->s.regs.acrs, 64);
4526 rc |= write_guest_abs(vcpu, gpa + __LC_CREGS_SAVE_AREA,
4527 &vcpu->arch.sie_block->gcr, 128);
4528 return rc ? -EFAULT : 0;
4531 int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
4534 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
4535 * switch in the run ioctl. Let's update our copies before we save
4536 * it into the save area
4539 vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
4540 save_access_regs(vcpu->run->s.regs.acrs);
4542 return kvm_s390_store_status_unloaded(vcpu, addr);
4545 static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4547 kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu);
4548 kvm_s390_sync_request(KVM_REQ_DISABLE_IBS, vcpu);
4551 static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
4554 struct kvm_vcpu *vcpu;
4556 kvm_for_each_vcpu(i, vcpu, kvm) {
4557 __disable_ibs_on_vcpu(vcpu);
4561 static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4565 kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu);
4566 kvm_s390_sync_request(KVM_REQ_ENABLE_IBS, vcpu);
4569 int kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
4571 int i, online_vcpus, r = 0, started_vcpus = 0;
4573 if (!is_vcpu_stopped(vcpu))
4576 trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
4577 /* Only one cpu at a time may enter/leave the STOPPED state. */
4578 spin_lock(&vcpu->kvm->arch.start_stop_lock);
4579 online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4581 /* Let's tell the UV that we want to change into the operating state */
4582 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4583 r = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_OPR);
4585 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4590 for (i = 0; i < online_vcpus; i++) {
4591 if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
4595 if (started_vcpus == 0) {
4596 /* we're the only active VCPU -> speed it up */
4597 __enable_ibs_on_vcpu(vcpu);
4598 } else if (started_vcpus == 1) {
4600 * As we are starting a second VCPU, we have to disable
4601 * the IBS facility on all VCPUs to remove potentially
4602 * outstanding ENABLE requests.
4604 __disable_ibs_on_all_vcpus(vcpu->kvm);
4607 kvm_s390_clear_cpuflags(vcpu, CPUSTAT_STOPPED);
4609 * The real PSW might have changed due to a RESTART interpreted by the
4610 * ultravisor. We block all interrupts and let the next sie exit
4613 if (kvm_s390_pv_cpu_is_protected(vcpu))
4614 vcpu->arch.sie_block->gpsw.mask &= ~PSW_INT_MASK;
4616 * Another VCPU might have used IBS while we were offline.
4617 * Let's play safe and flush the VCPU at startup.
4619 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4620 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4624 int kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
4626 int i, online_vcpus, r = 0, started_vcpus = 0;
4627 struct kvm_vcpu *started_vcpu = NULL;
4629 if (is_vcpu_stopped(vcpu))
4632 trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
4633 /* Only one cpu at a time may enter/leave the STOPPED state. */
4634 spin_lock(&vcpu->kvm->arch.start_stop_lock);
4635 online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4637 /* Let's tell the UV that we want to change into the stopped state */
4638 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4639 r = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_STP);
4641 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4646 /* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
4647 kvm_s390_clear_stop_irq(vcpu);
4649 kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOPPED);
4650 __disable_ibs_on_vcpu(vcpu);
4652 for (i = 0; i < online_vcpus; i++) {
4653 if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
4655 started_vcpu = vcpu->kvm->vcpus[i];
4659 if (started_vcpus == 1) {
4661 * As we only have one VCPU left, we want to enable the
4662 * IBS facility for that VCPU to speed it up.
4664 __enable_ibs_on_vcpu(started_vcpu);
4667 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4671 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
4672 struct kvm_enable_cap *cap)
4680 case KVM_CAP_S390_CSS_SUPPORT:
4681 if (!vcpu->kvm->arch.css_support) {
4682 vcpu->kvm->arch.css_support = 1;
4683 VM_EVENT(vcpu->kvm, 3, "%s", "ENABLE: CSS support");
4684 trace_kvm_s390_enable_css(vcpu->kvm);
4695 static long kvm_s390_guest_sida_op(struct kvm_vcpu *vcpu,
4696 struct kvm_s390_mem_op *mop)
4698 void __user *uaddr = (void __user *)mop->buf;
4701 if (mop->flags || !mop->size)
4703 if (mop->size + mop->sida_offset < mop->size)
4705 if (mop->size + mop->sida_offset > sida_size(vcpu->arch.sie_block))
4709 case KVM_S390_MEMOP_SIDA_READ:
4710 if (copy_to_user(uaddr, (void *)(sida_origin(vcpu->arch.sie_block) +
4711 mop->sida_offset), mop->size))
4715 case KVM_S390_MEMOP_SIDA_WRITE:
4716 if (copy_from_user((void *)(sida_origin(vcpu->arch.sie_block) +
4717 mop->sida_offset), uaddr, mop->size))
4723 static long kvm_s390_guest_mem_op(struct kvm_vcpu *vcpu,
4724 struct kvm_s390_mem_op *mop)
4726 void __user *uaddr = (void __user *)mop->buf;
4727 void *tmpbuf = NULL;
4729 const u64 supported_flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION
4730 | KVM_S390_MEMOP_F_CHECK_ONLY;
4732 if (mop->flags & ~supported_flags || mop->ar >= NUM_ACRS || !mop->size)
4735 if (mop->size > MEM_OP_MAX_SIZE)
4738 if (kvm_s390_pv_cpu_is_protected(vcpu))
4741 if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
4742 tmpbuf = vmalloc(mop->size);
4748 case KVM_S390_MEMOP_LOGICAL_READ:
4749 if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4750 r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4751 mop->size, GACC_FETCH);
4754 r = read_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4756 if (copy_to_user(uaddr, tmpbuf, mop->size))
4760 case KVM_S390_MEMOP_LOGICAL_WRITE:
4761 if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4762 r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4763 mop->size, GACC_STORE);
4766 if (copy_from_user(tmpbuf, uaddr, mop->size)) {
4770 r = write_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4774 if (r > 0 && (mop->flags & KVM_S390_MEMOP_F_INJECT_EXCEPTION) != 0)
4775 kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
4781 static long kvm_s390_guest_memsida_op(struct kvm_vcpu *vcpu,
4782 struct kvm_s390_mem_op *mop)
4786 srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
4789 case KVM_S390_MEMOP_LOGICAL_READ:
4790 case KVM_S390_MEMOP_LOGICAL_WRITE:
4791 r = kvm_s390_guest_mem_op(vcpu, mop);
4793 case KVM_S390_MEMOP_SIDA_READ:
4794 case KVM_S390_MEMOP_SIDA_WRITE:
4795 /* we are locked against sida going away by the vcpu->mutex */
4796 r = kvm_s390_guest_sida_op(vcpu, mop);
4802 srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
4806 long kvm_arch_vcpu_async_ioctl(struct file *filp,
4807 unsigned int ioctl, unsigned long arg)
4809 struct kvm_vcpu *vcpu = filp->private_data;
4810 void __user *argp = (void __user *)arg;
4813 case KVM_S390_IRQ: {
4814 struct kvm_s390_irq s390irq;
4816 if (copy_from_user(&s390irq, argp, sizeof(s390irq)))
4818 return kvm_s390_inject_vcpu(vcpu, &s390irq);
4820 case KVM_S390_INTERRUPT: {
4821 struct kvm_s390_interrupt s390int;
4822 struct kvm_s390_irq s390irq = {};
4824 if (copy_from_user(&s390int, argp, sizeof(s390int)))
4826 if (s390int_to_s390irq(&s390int, &s390irq))
4828 return kvm_s390_inject_vcpu(vcpu, &s390irq);
4831 return -ENOIOCTLCMD;
4834 long kvm_arch_vcpu_ioctl(struct file *filp,
4835 unsigned int ioctl, unsigned long arg)
4837 struct kvm_vcpu *vcpu = filp->private_data;
4838 void __user *argp = (void __user *)arg;
4846 case KVM_S390_STORE_STATUS:
4847 idx = srcu_read_lock(&vcpu->kvm->srcu);
4848 r = kvm_s390_store_status_unloaded(vcpu, arg);
4849 srcu_read_unlock(&vcpu->kvm->srcu, idx);
4851 case KVM_S390_SET_INITIAL_PSW: {
4855 if (copy_from_user(&psw, argp, sizeof(psw)))
4857 r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw);
4860 case KVM_S390_CLEAR_RESET:
4862 kvm_arch_vcpu_ioctl_clear_reset(vcpu);
4863 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4864 r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
4865 UVC_CMD_CPU_RESET_CLEAR, &rc, &rrc);
4866 VCPU_EVENT(vcpu, 3, "PROTVIRT RESET CLEAR VCPU: rc %x rrc %x",
4870 case KVM_S390_INITIAL_RESET:
4872 kvm_arch_vcpu_ioctl_initial_reset(vcpu);
4873 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4874 r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
4875 UVC_CMD_CPU_RESET_INITIAL,
4877 VCPU_EVENT(vcpu, 3, "PROTVIRT RESET INITIAL VCPU: rc %x rrc %x",
4881 case KVM_S390_NORMAL_RESET:
4883 kvm_arch_vcpu_ioctl_normal_reset(vcpu);
4884 if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4885 r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
4886 UVC_CMD_CPU_RESET, &rc, &rrc);
4887 VCPU_EVENT(vcpu, 3, "PROTVIRT RESET NORMAL VCPU: rc %x rrc %x",
4891 case KVM_SET_ONE_REG:
4892 case KVM_GET_ONE_REG: {
4893 struct kvm_one_reg reg;
4895 if (kvm_s390_pv_cpu_is_protected(vcpu))
4898 if (copy_from_user(®, argp, sizeof(reg)))
4900 if (ioctl == KVM_SET_ONE_REG)
4901 r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, ®);
4903 r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, ®);
4906 #ifdef CONFIG_KVM_S390_UCONTROL
4907 case KVM_S390_UCAS_MAP: {
4908 struct kvm_s390_ucas_mapping ucasmap;
4910 if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4915 if (!kvm_is_ucontrol(vcpu->kvm)) {
4920 r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr,
4921 ucasmap.vcpu_addr, ucasmap.length);
4924 case KVM_S390_UCAS_UNMAP: {
4925 struct kvm_s390_ucas_mapping ucasmap;
4927 if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4932 if (!kvm_is_ucontrol(vcpu->kvm)) {
4937 r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr,
4942 case KVM_S390_VCPU_FAULT: {
4943 r = gmap_fault(vcpu->arch.gmap, arg, 0);
4946 case KVM_ENABLE_CAP:
4948 struct kvm_enable_cap cap;
4950 if (copy_from_user(&cap, argp, sizeof(cap)))
4952 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
4955 case KVM_S390_MEM_OP: {
4956 struct kvm_s390_mem_op mem_op;
4958 if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
4959 r = kvm_s390_guest_memsida_op(vcpu, &mem_op);
4964 case KVM_S390_SET_IRQ_STATE: {
4965 struct kvm_s390_irq_state irq_state;
4968 if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4970 if (irq_state.len > VCPU_IRQS_MAX_BUF ||
4971 irq_state.len == 0 ||
4972 irq_state.len % sizeof(struct kvm_s390_irq) > 0) {
4976 /* do not use irq_state.flags, it will break old QEMUs */
4977 r = kvm_s390_set_irq_state(vcpu,
4978 (void __user *) irq_state.buf,
4982 case KVM_S390_GET_IRQ_STATE: {
4983 struct kvm_s390_irq_state irq_state;
4986 if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4988 if (irq_state.len == 0) {
4992 /* do not use irq_state.flags, it will break old QEMUs */
4993 r = kvm_s390_get_irq_state(vcpu,
4994 (__u8 __user *) irq_state.buf,
5006 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
5008 #ifdef CONFIG_KVM_S390_UCONTROL
5009 if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET)
5010 && (kvm_is_ucontrol(vcpu->kvm))) {
5011 vmf->page = virt_to_page(vcpu->arch.sie_block);
5012 get_page(vmf->page);
5016 return VM_FAULT_SIGBUS;
5019 /* Section: memory related */
5020 int kvm_arch_prepare_memory_region(struct kvm *kvm,
5021 struct kvm_memory_slot *memslot,
5022 const struct kvm_userspace_memory_region *mem,
5023 enum kvm_mr_change change)
5025 /* A few sanity checks. We can have memory slots which have to be
5026 located/ended at a segment boundary (1MB). The memory in userland is
5027 ok to be fragmented into various different vmas. It is okay to mmap()
5028 and munmap() stuff in this slot after doing this call at any time */
5030 if (mem->userspace_addr & 0xffffful)
5033 if (mem->memory_size & 0xffffful)
5036 if (mem->guest_phys_addr + mem->memory_size > kvm->arch.mem_limit)
5039 /* When we are protected, we should not change the memory slots */
5040 if (kvm_s390_pv_get_handle(kvm))
5045 void kvm_arch_commit_memory_region(struct kvm *kvm,
5046 const struct kvm_userspace_memory_region *mem,
5047 struct kvm_memory_slot *old,
5048 const struct kvm_memory_slot *new,
5049 enum kvm_mr_change change)
5055 rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
5056 old->npages * PAGE_SIZE);
5059 rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
5060 old->npages * PAGE_SIZE);
5065 rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr,
5066 mem->guest_phys_addr, mem->memory_size);
5068 case KVM_MR_FLAGS_ONLY:
5071 WARN(1, "Unknown KVM MR CHANGE: %d\n", change);
5074 pr_warn("failed to commit memory region\n");
5078 static inline unsigned long nonhyp_mask(int i)
5080 unsigned int nonhyp_fai = (sclp.hmfai << i * 2) >> 30;
5082 return 0x0000ffffffffffffUL >> (nonhyp_fai << 4);
5085 void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu)
5087 vcpu->valid_wakeup = false;
5090 static int __init kvm_s390_init(void)
5094 if (!sclp.has_sief2) {
5095 pr_info("SIE is not available\n");
5099 if (nested && hpage) {
5100 pr_info("A KVM host that supports nesting cannot back its KVM guests with huge pages\n");
5104 for (i = 0; i < 16; i++)
5105 kvm_s390_fac_base[i] |=
5106 stfle_fac_list[i] & nonhyp_mask(i);
5108 return kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
5111 static void __exit kvm_s390_exit(void)
5116 module_init(kvm_s390_init);
5117 module_exit(kvm_s390_exit);
5120 * Enable autoloading of the kvm module.
5121 * Note that we add the module alias here instead of virt/kvm/kvm_main.c
5122 * since x86 takes a different approach.
5124 #include <linux/miscdevice.h>
5125 MODULE_ALIAS_MISCDEV(KVM_MINOR);
5126 MODULE_ALIAS("devname:kvm");