1 // SPDX-License-Identifier: GPL-2.0
3 * hosting IBM Z kernel virtual machines (s390x)
5 * Copyright IBM Corp. 2008, 2017
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 #include <linux/compiler.h>
15 #include <linux/err.h>
17 #include <linux/hrtimer.h>
18 #include <linux/init.h>
19 #include <linux/kvm.h>
20 #include <linux/kvm_host.h>
21 #include <linux/mman.h>
22 #include <linux/module.h>
23 #include <linux/moduleparam.h>
24 #include <linux/random.h>
25 #include <linux/slab.h>
26 #include <linux/timer.h>
27 #include <linux/vmalloc.h>
28 #include <linux/bitmap.h>
29 #include <linux/sched/signal.h>
30 #include <linux/string.h>
32 #include <asm/asm-offsets.h>
33 #include <asm/lowcore.h>
35 #include <asm/pgtable.h>
38 #include <asm/switch_to.h>
41 #include <asm/cpacf.h>
42 #include <asm/timex.h>
46 #define KMSG_COMPONENT "kvm-s390"
48 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
50 #define CREATE_TRACE_POINTS
52 #include "trace-s390.h"
54 #define MEM_OP_MAX_SIZE 65536 /* Maximum transfer size for KVM_S390_MEM_OP */
56 #define VCPU_IRQS_MAX_BUF (sizeof(struct kvm_s390_irq) * \
57 (KVM_MAX_VCPUS + LOCAL_IRQS))
59 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
61 struct kvm_stats_debugfs_item debugfs_entries
[] = {
62 { "userspace_handled", VCPU_STAT(exit_userspace
) },
63 { "exit_null", VCPU_STAT(exit_null
) },
64 { "exit_validity", VCPU_STAT(exit_validity
) },
65 { "exit_stop_request", VCPU_STAT(exit_stop_request
) },
66 { "exit_external_request", VCPU_STAT(exit_external_request
) },
67 { "exit_external_interrupt", VCPU_STAT(exit_external_interrupt
) },
68 { "exit_instruction", VCPU_STAT(exit_instruction
) },
69 { "exit_pei", VCPU_STAT(exit_pei
) },
70 { "exit_program_interruption", VCPU_STAT(exit_program_interruption
) },
71 { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program
) },
72 { "exit_operation_exception", VCPU_STAT(exit_operation_exception
) },
73 { "halt_successful_poll", VCPU_STAT(halt_successful_poll
) },
74 { "halt_attempted_poll", VCPU_STAT(halt_attempted_poll
) },
75 { "halt_poll_invalid", VCPU_STAT(halt_poll_invalid
) },
76 { "halt_wakeup", VCPU_STAT(halt_wakeup
) },
77 { "instruction_lctlg", VCPU_STAT(instruction_lctlg
) },
78 { "instruction_lctl", VCPU_STAT(instruction_lctl
) },
79 { "instruction_stctl", VCPU_STAT(instruction_stctl
) },
80 { "instruction_stctg", VCPU_STAT(instruction_stctg
) },
81 { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal
) },
82 { "deliver_external_call", VCPU_STAT(deliver_external_call
) },
83 { "deliver_service_signal", VCPU_STAT(deliver_service_signal
) },
84 { "deliver_virtio_interrupt", VCPU_STAT(deliver_virtio_interrupt
) },
85 { "deliver_stop_signal", VCPU_STAT(deliver_stop_signal
) },
86 { "deliver_prefix_signal", VCPU_STAT(deliver_prefix_signal
) },
87 { "deliver_restart_signal", VCPU_STAT(deliver_restart_signal
) },
88 { "deliver_program_interruption", VCPU_STAT(deliver_program_int
) },
89 { "exit_wait_state", VCPU_STAT(exit_wait_state
) },
90 { "instruction_pfmf", VCPU_STAT(instruction_pfmf
) },
91 { "instruction_stidp", VCPU_STAT(instruction_stidp
) },
92 { "instruction_spx", VCPU_STAT(instruction_spx
) },
93 { "instruction_stpx", VCPU_STAT(instruction_stpx
) },
94 { "instruction_stap", VCPU_STAT(instruction_stap
) },
95 { "instruction_storage_key", VCPU_STAT(instruction_storage_key
) },
96 { "instruction_ipte_interlock", VCPU_STAT(instruction_ipte_interlock
) },
97 { "instruction_stsch", VCPU_STAT(instruction_stsch
) },
98 { "instruction_chsc", VCPU_STAT(instruction_chsc
) },
99 { "instruction_essa", VCPU_STAT(instruction_essa
) },
100 { "instruction_stsi", VCPU_STAT(instruction_stsi
) },
101 { "instruction_stfl", VCPU_STAT(instruction_stfl
) },
102 { "instruction_tprot", VCPU_STAT(instruction_tprot
) },
103 { "instruction_sthyi", VCPU_STAT(instruction_sthyi
) },
104 { "instruction_sie", VCPU_STAT(instruction_sie
) },
105 { "instruction_sigp_sense", VCPU_STAT(instruction_sigp_sense
) },
106 { "instruction_sigp_sense_running", VCPU_STAT(instruction_sigp_sense_running
) },
107 { "instruction_sigp_external_call", VCPU_STAT(instruction_sigp_external_call
) },
108 { "instruction_sigp_emergency", VCPU_STAT(instruction_sigp_emergency
) },
109 { "instruction_sigp_cond_emergency", VCPU_STAT(instruction_sigp_cond_emergency
) },
110 { "instruction_sigp_start", VCPU_STAT(instruction_sigp_start
) },
111 { "instruction_sigp_stop", VCPU_STAT(instruction_sigp_stop
) },
112 { "instruction_sigp_stop_store_status", VCPU_STAT(instruction_sigp_stop_store_status
) },
113 { "instruction_sigp_store_status", VCPU_STAT(instruction_sigp_store_status
) },
114 { "instruction_sigp_store_adtl_status", VCPU_STAT(instruction_sigp_store_adtl_status
) },
115 { "instruction_sigp_set_arch", VCPU_STAT(instruction_sigp_arch
) },
116 { "instruction_sigp_set_prefix", VCPU_STAT(instruction_sigp_prefix
) },
117 { "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart
) },
118 { "instruction_sigp_cpu_reset", VCPU_STAT(instruction_sigp_cpu_reset
) },
119 { "instruction_sigp_init_cpu_reset", VCPU_STAT(instruction_sigp_init_cpu_reset
) },
120 { "instruction_sigp_unknown", VCPU_STAT(instruction_sigp_unknown
) },
121 { "diagnose_10", VCPU_STAT(diagnose_10
) },
122 { "diagnose_44", VCPU_STAT(diagnose_44
) },
123 { "diagnose_9c", VCPU_STAT(diagnose_9c
) },
124 { "diagnose_258", VCPU_STAT(diagnose_258
) },
125 { "diagnose_308", VCPU_STAT(diagnose_308
) },
126 { "diagnose_500", VCPU_STAT(diagnose_500
) },
130 struct kvm_s390_tod_clock_ext
{
136 /* allow nested virtualization in KVM (if enabled by user space) */
138 module_param(nested
, int, S_IRUGO
);
139 MODULE_PARM_DESC(nested
, "Nested virtualization support");
141 /* upper facilities limit for kvm */
142 unsigned long kvm_s390_fac_list_mask
[16] = { FACILITIES_KVM
};
144 unsigned long kvm_s390_fac_list_mask_size(void)
146 BUILD_BUG_ON(ARRAY_SIZE(kvm_s390_fac_list_mask
) > S390_ARCH_FAC_MASK_SIZE_U64
);
147 return ARRAY_SIZE(kvm_s390_fac_list_mask
);
150 /* available cpu features supported by kvm */
151 static DECLARE_BITMAP(kvm_s390_available_cpu_feat
, KVM_S390_VM_CPU_FEAT_NR_BITS
);
152 /* available subfunctions indicated via query / "test bit" */
153 static struct kvm_s390_vm_cpu_subfunc kvm_s390_available_subfunc
;
155 static struct gmap_notifier gmap_notifier
;
156 static struct gmap_notifier vsie_gmap_notifier
;
157 debug_info_t
*kvm_s390_dbf
;
159 /* Section: not file related */
160 int kvm_arch_hardware_enable(void)
162 /* every s390 is virtualization enabled ;-) */
166 static void kvm_gmap_notifier(struct gmap
*gmap
, unsigned long start
,
170 * This callback is executed during stop_machine(). All CPUs are therefore
171 * temporarily stopped. In order not to change guest behavior, we have to
172 * disable preemption whenever we touch the epoch of kvm and the VCPUs,
173 * so a CPU won't be stopped while calculating with the epoch.
175 static int kvm_clock_sync(struct notifier_block
*notifier
, unsigned long val
,
179 struct kvm_vcpu
*vcpu
;
181 unsigned long long *delta
= v
;
183 list_for_each_entry(kvm
, &vm_list
, vm_list
) {
184 kvm
->arch
.epoch
-= *delta
;
185 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
186 vcpu
->arch
.sie_block
->epoch
-= *delta
;
187 if (vcpu
->arch
.cputm_enabled
)
188 vcpu
->arch
.cputm_start
+= *delta
;
189 if (vcpu
->arch
.vsie_block
)
190 vcpu
->arch
.vsie_block
->epoch
-= *delta
;
196 static struct notifier_block kvm_clock_notifier
= {
197 .notifier_call
= kvm_clock_sync
,
200 int kvm_arch_hardware_setup(void)
202 gmap_notifier
.notifier_call
= kvm_gmap_notifier
;
203 gmap_register_pte_notifier(&gmap_notifier
);
204 vsie_gmap_notifier
.notifier_call
= kvm_s390_vsie_gmap_notifier
;
205 gmap_register_pte_notifier(&vsie_gmap_notifier
);
206 atomic_notifier_chain_register(&s390_epoch_delta_notifier
,
207 &kvm_clock_notifier
);
211 void kvm_arch_hardware_unsetup(void)
213 gmap_unregister_pte_notifier(&gmap_notifier
);
214 gmap_unregister_pte_notifier(&vsie_gmap_notifier
);
215 atomic_notifier_chain_unregister(&s390_epoch_delta_notifier
,
216 &kvm_clock_notifier
);
219 static void allow_cpu_feat(unsigned long nr
)
221 set_bit_inv(nr
, kvm_s390_available_cpu_feat
);
224 static inline int plo_test_bit(unsigned char nr
)
226 register unsigned long r0
asm("0") = (unsigned long) nr
| 0x100;
230 /* Parameter registers are ignored for "test bit" */
240 static void kvm_s390_cpu_feat_init(void)
244 for (i
= 0; i
< 256; ++i
) {
246 kvm_s390_available_subfunc
.plo
[i
>> 3] |= 0x80 >> (i
& 7);
249 if (test_facility(28)) /* TOD-clock steering */
250 ptff(kvm_s390_available_subfunc
.ptff
,
251 sizeof(kvm_s390_available_subfunc
.ptff
),
254 if (test_facility(17)) { /* MSA */
255 __cpacf_query(CPACF_KMAC
, (cpacf_mask_t
*)
256 kvm_s390_available_subfunc
.kmac
);
257 __cpacf_query(CPACF_KMC
, (cpacf_mask_t
*)
258 kvm_s390_available_subfunc
.kmc
);
259 __cpacf_query(CPACF_KM
, (cpacf_mask_t
*)
260 kvm_s390_available_subfunc
.km
);
261 __cpacf_query(CPACF_KIMD
, (cpacf_mask_t
*)
262 kvm_s390_available_subfunc
.kimd
);
263 __cpacf_query(CPACF_KLMD
, (cpacf_mask_t
*)
264 kvm_s390_available_subfunc
.klmd
);
266 if (test_facility(76)) /* MSA3 */
267 __cpacf_query(CPACF_PCKMO
, (cpacf_mask_t
*)
268 kvm_s390_available_subfunc
.pckmo
);
269 if (test_facility(77)) { /* MSA4 */
270 __cpacf_query(CPACF_KMCTR
, (cpacf_mask_t
*)
271 kvm_s390_available_subfunc
.kmctr
);
272 __cpacf_query(CPACF_KMF
, (cpacf_mask_t
*)
273 kvm_s390_available_subfunc
.kmf
);
274 __cpacf_query(CPACF_KMO
, (cpacf_mask_t
*)
275 kvm_s390_available_subfunc
.kmo
);
276 __cpacf_query(CPACF_PCC
, (cpacf_mask_t
*)
277 kvm_s390_available_subfunc
.pcc
);
279 if (test_facility(57)) /* MSA5 */
280 __cpacf_query(CPACF_PRNO
, (cpacf_mask_t
*)
281 kvm_s390_available_subfunc
.ppno
);
283 if (test_facility(146)) /* MSA8 */
284 __cpacf_query(CPACF_KMA
, (cpacf_mask_t
*)
285 kvm_s390_available_subfunc
.kma
);
287 if (MACHINE_HAS_ESOP
)
288 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP
);
290 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
291 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
293 if (!sclp
.has_sief2
|| !MACHINE_HAS_ESOP
|| !sclp
.has_64bscao
||
294 !test_facility(3) || !nested
)
296 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2
);
297 if (sclp
.has_64bscao
)
298 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO
);
300 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF
);
302 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE
);
304 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS
);
306 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB
);
308 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI
);
310 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS
);
312 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_KSS
);
314 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
315 * all skey handling functions read/set the skey from the PGSTE
316 * instead of the real storage key.
318 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
319 * pages being detected as preserved although they are resident.
321 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
322 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
324 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
325 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
326 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
328 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
329 * cannot easily shadow the SCA because of the ipte lock.
333 int kvm_arch_init(void *opaque
)
335 kvm_s390_dbf
= debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
339 if (debug_register_view(kvm_s390_dbf
, &debug_sprintf_view
)) {
340 debug_unregister(kvm_s390_dbf
);
344 kvm_s390_cpu_feat_init();
346 /* Register floating interrupt controller interface. */
347 return kvm_register_device_ops(&kvm_flic_ops
, KVM_DEV_TYPE_FLIC
);
350 void kvm_arch_exit(void)
352 debug_unregister(kvm_s390_dbf
);
355 /* Section: device related */
356 long kvm_arch_dev_ioctl(struct file
*filp
,
357 unsigned int ioctl
, unsigned long arg
)
359 if (ioctl
== KVM_S390_ENABLE_SIE
)
360 return s390_enable_sie();
364 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
)
369 case KVM_CAP_S390_PSW
:
370 case KVM_CAP_S390_GMAP
:
371 case KVM_CAP_SYNC_MMU
:
372 #ifdef CONFIG_KVM_S390_UCONTROL
373 case KVM_CAP_S390_UCONTROL
:
375 case KVM_CAP_ASYNC_PF
:
376 case KVM_CAP_SYNC_REGS
:
377 case KVM_CAP_ONE_REG
:
378 case KVM_CAP_ENABLE_CAP
:
379 case KVM_CAP_S390_CSS_SUPPORT
:
380 case KVM_CAP_IOEVENTFD
:
381 case KVM_CAP_DEVICE_CTRL
:
382 case KVM_CAP_ENABLE_CAP_VM
:
383 case KVM_CAP_S390_IRQCHIP
:
384 case KVM_CAP_VM_ATTRIBUTES
:
385 case KVM_CAP_MP_STATE
:
386 case KVM_CAP_IMMEDIATE_EXIT
:
387 case KVM_CAP_S390_INJECT_IRQ
:
388 case KVM_CAP_S390_USER_SIGP
:
389 case KVM_CAP_S390_USER_STSI
:
390 case KVM_CAP_S390_SKEYS
:
391 case KVM_CAP_S390_IRQ_STATE
:
392 case KVM_CAP_S390_USER_INSTR0
:
393 case KVM_CAP_S390_CMMA_MIGRATION
:
394 case KVM_CAP_S390_AIS
:
395 case KVM_CAP_S390_AIS_MIGRATION
:
398 case KVM_CAP_S390_MEM_OP
:
401 case KVM_CAP_NR_VCPUS
:
402 case KVM_CAP_MAX_VCPUS
:
403 r
= KVM_S390_BSCA_CPU_SLOTS
;
404 if (!kvm_s390_use_sca_entries())
406 else if (sclp
.has_esca
&& sclp
.has_64bscao
)
407 r
= KVM_S390_ESCA_CPU_SLOTS
;
409 case KVM_CAP_NR_MEMSLOTS
:
410 r
= KVM_USER_MEM_SLOTS
;
412 case KVM_CAP_S390_COW
:
413 r
= MACHINE_HAS_ESOP
;
415 case KVM_CAP_S390_VECTOR_REGISTERS
:
418 case KVM_CAP_S390_RI
:
419 r
= test_facility(64);
421 case KVM_CAP_S390_GS
:
422 r
= test_facility(133);
430 static void kvm_s390_sync_dirty_log(struct kvm
*kvm
,
431 struct kvm_memory_slot
*memslot
)
433 gfn_t cur_gfn
, last_gfn
;
434 unsigned long address
;
435 struct gmap
*gmap
= kvm
->arch
.gmap
;
437 /* Loop over all guest pages */
438 last_gfn
= memslot
->base_gfn
+ memslot
->npages
;
439 for (cur_gfn
= memslot
->base_gfn
; cur_gfn
<= last_gfn
; cur_gfn
++) {
440 address
= gfn_to_hva_memslot(memslot
, cur_gfn
);
442 if (test_and_clear_guest_dirty(gmap
->mm
, address
))
443 mark_page_dirty(kvm
, cur_gfn
);
444 if (fatal_signal_pending(current
))
450 /* Section: vm related */
451 static void sca_del_vcpu(struct kvm_vcpu
*vcpu
);
454 * Get (and clear) the dirty memory log for a memory slot.
456 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
457 struct kvm_dirty_log
*log
)
461 struct kvm_memslots
*slots
;
462 struct kvm_memory_slot
*memslot
;
465 if (kvm_is_ucontrol(kvm
))
468 mutex_lock(&kvm
->slots_lock
);
471 if (log
->slot
>= KVM_USER_MEM_SLOTS
)
474 slots
= kvm_memslots(kvm
);
475 memslot
= id_to_memslot(slots
, log
->slot
);
477 if (!memslot
->dirty_bitmap
)
480 kvm_s390_sync_dirty_log(kvm
, memslot
);
481 r
= kvm_get_dirty_log(kvm
, log
, &is_dirty
);
485 /* Clear the dirty log */
487 n
= kvm_dirty_bitmap_bytes(memslot
);
488 memset(memslot
->dirty_bitmap
, 0, n
);
492 mutex_unlock(&kvm
->slots_lock
);
496 static void icpt_operexc_on_all_vcpus(struct kvm
*kvm
)
499 struct kvm_vcpu
*vcpu
;
501 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
502 kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC
, vcpu
);
506 static int kvm_vm_ioctl_enable_cap(struct kvm
*kvm
, struct kvm_enable_cap
*cap
)
514 case KVM_CAP_S390_IRQCHIP
:
515 VM_EVENT(kvm
, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
516 kvm
->arch
.use_irqchip
= 1;
519 case KVM_CAP_S390_USER_SIGP
:
520 VM_EVENT(kvm
, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
521 kvm
->arch
.user_sigp
= 1;
524 case KVM_CAP_S390_VECTOR_REGISTERS
:
525 mutex_lock(&kvm
->lock
);
526 if (kvm
->created_vcpus
) {
528 } else if (MACHINE_HAS_VX
) {
529 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 129);
530 set_kvm_facility(kvm
->arch
.model
.fac_list
, 129);
531 if (test_facility(134)) {
532 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 134);
533 set_kvm_facility(kvm
->arch
.model
.fac_list
, 134);
535 if (test_facility(135)) {
536 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 135);
537 set_kvm_facility(kvm
->arch
.model
.fac_list
, 135);
542 mutex_unlock(&kvm
->lock
);
543 VM_EVENT(kvm
, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
544 r
? "(not available)" : "(success)");
546 case KVM_CAP_S390_RI
:
548 mutex_lock(&kvm
->lock
);
549 if (kvm
->created_vcpus
) {
551 } else if (test_facility(64)) {
552 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 64);
553 set_kvm_facility(kvm
->arch
.model
.fac_list
, 64);
556 mutex_unlock(&kvm
->lock
);
557 VM_EVENT(kvm
, 3, "ENABLE: CAP_S390_RI %s",
558 r
? "(not available)" : "(success)");
560 case KVM_CAP_S390_AIS
:
561 mutex_lock(&kvm
->lock
);
562 if (kvm
->created_vcpus
) {
565 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 72);
566 set_kvm_facility(kvm
->arch
.model
.fac_list
, 72);
569 mutex_unlock(&kvm
->lock
);
570 VM_EVENT(kvm
, 3, "ENABLE: AIS %s",
571 r
? "(not available)" : "(success)");
573 case KVM_CAP_S390_GS
:
575 mutex_lock(&kvm
->lock
);
576 if (atomic_read(&kvm
->online_vcpus
)) {
578 } else if (test_facility(133)) {
579 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 133);
580 set_kvm_facility(kvm
->arch
.model
.fac_list
, 133);
583 mutex_unlock(&kvm
->lock
);
584 VM_EVENT(kvm
, 3, "ENABLE: CAP_S390_GS %s",
585 r
? "(not available)" : "(success)");
587 case KVM_CAP_S390_USER_STSI
:
588 VM_EVENT(kvm
, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
589 kvm
->arch
.user_stsi
= 1;
592 case KVM_CAP_S390_USER_INSTR0
:
593 VM_EVENT(kvm
, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
594 kvm
->arch
.user_instr0
= 1;
595 icpt_operexc_on_all_vcpus(kvm
);
605 static int kvm_s390_get_mem_control(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
609 switch (attr
->attr
) {
610 case KVM_S390_VM_MEM_LIMIT_SIZE
:
612 VM_EVENT(kvm
, 3, "QUERY: max guest memory: %lu bytes",
613 kvm
->arch
.mem_limit
);
614 if (put_user(kvm
->arch
.mem_limit
, (u64 __user
*)attr
->addr
))
624 static int kvm_s390_set_mem_control(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
628 switch (attr
->attr
) {
629 case KVM_S390_VM_MEM_ENABLE_CMMA
:
635 VM_EVENT(kvm
, 3, "%s", "ENABLE: CMMA support");
636 mutex_lock(&kvm
->lock
);
637 if (!kvm
->created_vcpus
) {
638 kvm
->arch
.use_cmma
= 1;
641 mutex_unlock(&kvm
->lock
);
643 case KVM_S390_VM_MEM_CLR_CMMA
:
648 if (!kvm
->arch
.use_cmma
)
651 VM_EVENT(kvm
, 3, "%s", "RESET: CMMA states");
652 mutex_lock(&kvm
->lock
);
653 idx
= srcu_read_lock(&kvm
->srcu
);
654 s390_reset_cmma(kvm
->arch
.gmap
->mm
);
655 srcu_read_unlock(&kvm
->srcu
, idx
);
656 mutex_unlock(&kvm
->lock
);
659 case KVM_S390_VM_MEM_LIMIT_SIZE
: {
660 unsigned long new_limit
;
662 if (kvm_is_ucontrol(kvm
))
665 if (get_user(new_limit
, (u64 __user
*)attr
->addr
))
668 if (kvm
->arch
.mem_limit
!= KVM_S390_NO_MEM_LIMIT
&&
669 new_limit
> kvm
->arch
.mem_limit
)
675 /* gmap_create takes last usable address */
676 if (new_limit
!= KVM_S390_NO_MEM_LIMIT
)
680 mutex_lock(&kvm
->lock
);
681 if (!kvm
->created_vcpus
) {
682 /* gmap_create will round the limit up */
683 struct gmap
*new = gmap_create(current
->mm
, new_limit
);
688 gmap_remove(kvm
->arch
.gmap
);
690 kvm
->arch
.gmap
= new;
694 mutex_unlock(&kvm
->lock
);
695 VM_EVENT(kvm
, 3, "SET: max guest address: %lu", new_limit
);
696 VM_EVENT(kvm
, 3, "New guest asce: 0x%pK",
697 (void *) kvm
->arch
.gmap
->asce
);
707 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu
*vcpu
);
709 static int kvm_s390_vm_set_crypto(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
711 struct kvm_vcpu
*vcpu
;
714 if (!test_kvm_facility(kvm
, 76))
717 mutex_lock(&kvm
->lock
);
718 switch (attr
->attr
) {
719 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW
:
721 kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
,
722 sizeof(kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
));
723 kvm
->arch
.crypto
.aes_kw
= 1;
724 VM_EVENT(kvm
, 3, "%s", "ENABLE: AES keywrapping support");
726 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW
:
728 kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
,
729 sizeof(kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
));
730 kvm
->arch
.crypto
.dea_kw
= 1;
731 VM_EVENT(kvm
, 3, "%s", "ENABLE: DEA keywrapping support");
733 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW
:
734 kvm
->arch
.crypto
.aes_kw
= 0;
735 memset(kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
, 0,
736 sizeof(kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
));
737 VM_EVENT(kvm
, 3, "%s", "DISABLE: AES keywrapping support");
739 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW
:
740 kvm
->arch
.crypto
.dea_kw
= 0;
741 memset(kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
, 0,
742 sizeof(kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
));
743 VM_EVENT(kvm
, 3, "%s", "DISABLE: DEA keywrapping support");
746 mutex_unlock(&kvm
->lock
);
750 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
751 kvm_s390_vcpu_crypto_setup(vcpu
);
754 mutex_unlock(&kvm
->lock
);
758 static void kvm_s390_sync_request_broadcast(struct kvm
*kvm
, int req
)
761 struct kvm_vcpu
*vcpu
;
763 kvm_for_each_vcpu(cx
, vcpu
, kvm
)
764 kvm_s390_sync_request(req
, vcpu
);
768 * Must be called with kvm->srcu held to avoid races on memslots, and with
769 * kvm->lock to avoid races with ourselves and kvm_s390_vm_stop_migration.
771 static int kvm_s390_vm_start_migration(struct kvm
*kvm
)
773 struct kvm_s390_migration_state
*mgs
;
774 struct kvm_memory_slot
*ms
;
775 /* should be the only one */
776 struct kvm_memslots
*slots
;
777 unsigned long ram_pages
;
780 /* migration mode already enabled */
781 if (kvm
->arch
.migration_state
)
784 slots
= kvm_memslots(kvm
);
785 if (!slots
|| !slots
->used_slots
)
788 mgs
= kzalloc(sizeof(*mgs
), GFP_KERNEL
);
791 kvm
->arch
.migration_state
= mgs
;
793 if (kvm
->arch
.use_cmma
) {
795 * Get the first slot. They are reverse sorted by base_gfn, so
796 * the first slot is also the one at the end of the address
797 * space. We have verified above that at least one slot is
800 ms
= slots
->memslots
;
801 /* round up so we only use full longs */
802 ram_pages
= roundup(ms
->base_gfn
+ ms
->npages
, BITS_PER_LONG
);
803 /* allocate enough bytes to store all the bits */
804 mgs
->pgste_bitmap
= vmalloc(ram_pages
/ 8);
805 if (!mgs
->pgste_bitmap
) {
807 kvm
->arch
.migration_state
= NULL
;
811 mgs
->bitmap_size
= ram_pages
;
812 atomic64_set(&mgs
->dirty_pages
, ram_pages
);
813 /* mark all the pages in active slots as dirty */
814 for (slotnr
= 0; slotnr
< slots
->used_slots
; slotnr
++) {
815 ms
= slots
->memslots
+ slotnr
;
816 bitmap_set(mgs
->pgste_bitmap
, ms
->base_gfn
, ms
->npages
);
819 kvm_s390_sync_request_broadcast(kvm
, KVM_REQ_START_MIGRATION
);
825 * Must be called with kvm->lock to avoid races with ourselves and
826 * kvm_s390_vm_start_migration.
828 static int kvm_s390_vm_stop_migration(struct kvm
*kvm
)
830 struct kvm_s390_migration_state
*mgs
;
832 /* migration mode already disabled */
833 if (!kvm
->arch
.migration_state
)
835 mgs
= kvm
->arch
.migration_state
;
836 kvm
->arch
.migration_state
= NULL
;
838 if (kvm
->arch
.use_cmma
) {
839 kvm_s390_sync_request_broadcast(kvm
, KVM_REQ_STOP_MIGRATION
);
840 vfree(mgs
->pgste_bitmap
);
846 static int kvm_s390_vm_set_migration(struct kvm
*kvm
,
847 struct kvm_device_attr
*attr
)
849 int idx
, res
= -ENXIO
;
851 mutex_lock(&kvm
->lock
);
852 switch (attr
->attr
) {
853 case KVM_S390_VM_MIGRATION_START
:
854 idx
= srcu_read_lock(&kvm
->srcu
);
855 res
= kvm_s390_vm_start_migration(kvm
);
856 srcu_read_unlock(&kvm
->srcu
, idx
);
858 case KVM_S390_VM_MIGRATION_STOP
:
859 res
= kvm_s390_vm_stop_migration(kvm
);
864 mutex_unlock(&kvm
->lock
);
869 static int kvm_s390_vm_get_migration(struct kvm
*kvm
,
870 struct kvm_device_attr
*attr
)
872 u64 mig
= (kvm
->arch
.migration_state
!= NULL
);
874 if (attr
->attr
!= KVM_S390_VM_MIGRATION_STATUS
)
877 if (copy_to_user((void __user
*)attr
->addr
, &mig
, sizeof(mig
)))
882 static int kvm_s390_set_tod_ext(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
884 struct kvm_s390_vm_tod_clock gtod
;
886 if (copy_from_user(>od
, (void __user
*)attr
->addr
, sizeof(gtod
)))
889 if (test_kvm_facility(kvm
, 139))
890 kvm_s390_set_tod_clock_ext(kvm
, >od
);
891 else if (gtod
.epoch_idx
== 0)
892 kvm_s390_set_tod_clock(kvm
, gtod
.tod
);
896 VM_EVENT(kvm
, 3, "SET: TOD extension: 0x%x, TOD base: 0x%llx",
897 gtod
.epoch_idx
, gtod
.tod
);
902 static int kvm_s390_set_tod_high(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
906 if (copy_from_user(>od_high
, (void __user
*)attr
->addr
,
912 VM_EVENT(kvm
, 3, "SET: TOD extension: 0x%x", gtod_high
);
917 static int kvm_s390_set_tod_low(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
921 if (copy_from_user(>od
, (void __user
*)attr
->addr
, sizeof(gtod
)))
924 kvm_s390_set_tod_clock(kvm
, gtod
);
925 VM_EVENT(kvm
, 3, "SET: TOD base: 0x%llx", gtod
);
929 static int kvm_s390_set_tod(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
936 switch (attr
->attr
) {
937 case KVM_S390_VM_TOD_EXT
:
938 ret
= kvm_s390_set_tod_ext(kvm
, attr
);
940 case KVM_S390_VM_TOD_HIGH
:
941 ret
= kvm_s390_set_tod_high(kvm
, attr
);
943 case KVM_S390_VM_TOD_LOW
:
944 ret
= kvm_s390_set_tod_low(kvm
, attr
);
953 static void kvm_s390_get_tod_clock_ext(struct kvm
*kvm
,
954 struct kvm_s390_vm_tod_clock
*gtod
)
956 struct kvm_s390_tod_clock_ext htod
;
960 get_tod_clock_ext((char *)&htod
);
962 gtod
->tod
= htod
.tod
+ kvm
->arch
.epoch
;
963 gtod
->epoch_idx
= htod
.epoch_idx
+ kvm
->arch
.epdx
;
965 if (gtod
->tod
< htod
.tod
)
966 gtod
->epoch_idx
+= 1;
971 static int kvm_s390_get_tod_ext(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
973 struct kvm_s390_vm_tod_clock gtod
;
975 memset(>od
, 0, sizeof(gtod
));
977 if (test_kvm_facility(kvm
, 139))
978 kvm_s390_get_tod_clock_ext(kvm
, >od
);
980 gtod
.tod
= kvm_s390_get_tod_clock_fast(kvm
);
982 if (copy_to_user((void __user
*)attr
->addr
, >od
, sizeof(gtod
)))
985 VM_EVENT(kvm
, 3, "QUERY: TOD extension: 0x%x, TOD base: 0x%llx",
986 gtod
.epoch_idx
, gtod
.tod
);
990 static int kvm_s390_get_tod_high(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
994 if (copy_to_user((void __user
*)attr
->addr
, >od_high
,
997 VM_EVENT(kvm
, 3, "QUERY: TOD extension: 0x%x", gtod_high
);
1002 static int kvm_s390_get_tod_low(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1006 gtod
= kvm_s390_get_tod_clock_fast(kvm
);
1007 if (copy_to_user((void __user
*)attr
->addr
, >od
, sizeof(gtod
)))
1009 VM_EVENT(kvm
, 3, "QUERY: TOD base: 0x%llx", gtod
);
1014 static int kvm_s390_get_tod(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1021 switch (attr
->attr
) {
1022 case KVM_S390_VM_TOD_EXT
:
1023 ret
= kvm_s390_get_tod_ext(kvm
, attr
);
1025 case KVM_S390_VM_TOD_HIGH
:
1026 ret
= kvm_s390_get_tod_high(kvm
, attr
);
1028 case KVM_S390_VM_TOD_LOW
:
1029 ret
= kvm_s390_get_tod_low(kvm
, attr
);
1038 static int kvm_s390_set_processor(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1040 struct kvm_s390_vm_cpu_processor
*proc
;
1041 u16 lowest_ibc
, unblocked_ibc
;
1044 mutex_lock(&kvm
->lock
);
1045 if (kvm
->created_vcpus
) {
1049 proc
= kzalloc(sizeof(*proc
), GFP_KERNEL
);
1054 if (!copy_from_user(proc
, (void __user
*)attr
->addr
,
1056 kvm
->arch
.model
.cpuid
= proc
->cpuid
;
1057 lowest_ibc
= sclp
.ibc
>> 16 & 0xfff;
1058 unblocked_ibc
= sclp
.ibc
& 0xfff;
1059 if (lowest_ibc
&& proc
->ibc
) {
1060 if (proc
->ibc
> unblocked_ibc
)
1061 kvm
->arch
.model
.ibc
= unblocked_ibc
;
1062 else if (proc
->ibc
< lowest_ibc
)
1063 kvm
->arch
.model
.ibc
= lowest_ibc
;
1065 kvm
->arch
.model
.ibc
= proc
->ibc
;
1067 memcpy(kvm
->arch
.model
.fac_list
, proc
->fac_list
,
1068 S390_ARCH_FAC_LIST_SIZE_BYTE
);
1069 VM_EVENT(kvm
, 3, "SET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1070 kvm
->arch
.model
.ibc
,
1071 kvm
->arch
.model
.cpuid
);
1072 VM_EVENT(kvm
, 3, "SET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1073 kvm
->arch
.model
.fac_list
[0],
1074 kvm
->arch
.model
.fac_list
[1],
1075 kvm
->arch
.model
.fac_list
[2]);
1080 mutex_unlock(&kvm
->lock
);
1084 static int kvm_s390_set_processor_feat(struct kvm
*kvm
,
1085 struct kvm_device_attr
*attr
)
1087 struct kvm_s390_vm_cpu_feat data
;
1090 if (copy_from_user(&data
, (void __user
*)attr
->addr
, sizeof(data
)))
1092 if (!bitmap_subset((unsigned long *) data
.feat
,
1093 kvm_s390_available_cpu_feat
,
1094 KVM_S390_VM_CPU_FEAT_NR_BITS
))
1097 mutex_lock(&kvm
->lock
);
1098 if (!atomic_read(&kvm
->online_vcpus
)) {
1099 bitmap_copy(kvm
->arch
.cpu_feat
, (unsigned long *) data
.feat
,
1100 KVM_S390_VM_CPU_FEAT_NR_BITS
);
1103 mutex_unlock(&kvm
->lock
);
1107 static int kvm_s390_set_processor_subfunc(struct kvm
*kvm
,
1108 struct kvm_device_attr
*attr
)
1111 * Once supported by kernel + hw, we have to store the subfunctions
1112 * in kvm->arch and remember that user space configured them.
1117 static int kvm_s390_set_cpu_model(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1121 switch (attr
->attr
) {
1122 case KVM_S390_VM_CPU_PROCESSOR
:
1123 ret
= kvm_s390_set_processor(kvm
, attr
);
1125 case KVM_S390_VM_CPU_PROCESSOR_FEAT
:
1126 ret
= kvm_s390_set_processor_feat(kvm
, attr
);
1128 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC
:
1129 ret
= kvm_s390_set_processor_subfunc(kvm
, attr
);
1135 static int kvm_s390_get_processor(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1137 struct kvm_s390_vm_cpu_processor
*proc
;
1140 proc
= kzalloc(sizeof(*proc
), GFP_KERNEL
);
1145 proc
->cpuid
= kvm
->arch
.model
.cpuid
;
1146 proc
->ibc
= kvm
->arch
.model
.ibc
;
1147 memcpy(&proc
->fac_list
, kvm
->arch
.model
.fac_list
,
1148 S390_ARCH_FAC_LIST_SIZE_BYTE
);
1149 VM_EVENT(kvm
, 3, "GET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1150 kvm
->arch
.model
.ibc
,
1151 kvm
->arch
.model
.cpuid
);
1152 VM_EVENT(kvm
, 3, "GET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1153 kvm
->arch
.model
.fac_list
[0],
1154 kvm
->arch
.model
.fac_list
[1],
1155 kvm
->arch
.model
.fac_list
[2]);
1156 if (copy_to_user((void __user
*)attr
->addr
, proc
, sizeof(*proc
)))
1163 static int kvm_s390_get_machine(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1165 struct kvm_s390_vm_cpu_machine
*mach
;
1168 mach
= kzalloc(sizeof(*mach
), GFP_KERNEL
);
1173 get_cpu_id((struct cpuid
*) &mach
->cpuid
);
1174 mach
->ibc
= sclp
.ibc
;
1175 memcpy(&mach
->fac_mask
, kvm
->arch
.model
.fac_mask
,
1176 S390_ARCH_FAC_LIST_SIZE_BYTE
);
1177 memcpy((unsigned long *)&mach
->fac_list
, S390_lowcore
.stfle_fac_list
,
1178 sizeof(S390_lowcore
.stfle_fac_list
));
1179 VM_EVENT(kvm
, 3, "GET: host ibc: 0x%4.4x, host cpuid: 0x%16.16llx",
1180 kvm
->arch
.model
.ibc
,
1181 kvm
->arch
.model
.cpuid
);
1182 VM_EVENT(kvm
, 3, "GET: host facmask: 0x%16.16llx.%16.16llx.%16.16llx",
1186 VM_EVENT(kvm
, 3, "GET: host faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1190 if (copy_to_user((void __user
*)attr
->addr
, mach
, sizeof(*mach
)))
1197 static int kvm_s390_get_processor_feat(struct kvm
*kvm
,
1198 struct kvm_device_attr
*attr
)
1200 struct kvm_s390_vm_cpu_feat data
;
1202 bitmap_copy((unsigned long *) data
.feat
, kvm
->arch
.cpu_feat
,
1203 KVM_S390_VM_CPU_FEAT_NR_BITS
);
1204 if (copy_to_user((void __user
*)attr
->addr
, &data
, sizeof(data
)))
1209 static int kvm_s390_get_machine_feat(struct kvm
*kvm
,
1210 struct kvm_device_attr
*attr
)
1212 struct kvm_s390_vm_cpu_feat data
;
1214 bitmap_copy((unsigned long *) data
.feat
,
1215 kvm_s390_available_cpu_feat
,
1216 KVM_S390_VM_CPU_FEAT_NR_BITS
);
1217 if (copy_to_user((void __user
*)attr
->addr
, &data
, sizeof(data
)))
1222 static int kvm_s390_get_processor_subfunc(struct kvm
*kvm
,
1223 struct kvm_device_attr
*attr
)
1226 * Once we can actually configure subfunctions (kernel + hw support),
1227 * we have to check if they were already set by user space, if so copy
1228 * them from kvm->arch.
1233 static int kvm_s390_get_machine_subfunc(struct kvm
*kvm
,
1234 struct kvm_device_attr
*attr
)
1236 if (copy_to_user((void __user
*)attr
->addr
, &kvm_s390_available_subfunc
,
1237 sizeof(struct kvm_s390_vm_cpu_subfunc
)))
1241 static int kvm_s390_get_cpu_model(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1245 switch (attr
->attr
) {
1246 case KVM_S390_VM_CPU_PROCESSOR
:
1247 ret
= kvm_s390_get_processor(kvm
, attr
);
1249 case KVM_S390_VM_CPU_MACHINE
:
1250 ret
= kvm_s390_get_machine(kvm
, attr
);
1252 case KVM_S390_VM_CPU_PROCESSOR_FEAT
:
1253 ret
= kvm_s390_get_processor_feat(kvm
, attr
);
1255 case KVM_S390_VM_CPU_MACHINE_FEAT
:
1256 ret
= kvm_s390_get_machine_feat(kvm
, attr
);
1258 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC
:
1259 ret
= kvm_s390_get_processor_subfunc(kvm
, attr
);
1261 case KVM_S390_VM_CPU_MACHINE_SUBFUNC
:
1262 ret
= kvm_s390_get_machine_subfunc(kvm
, attr
);
1268 static int kvm_s390_vm_set_attr(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1272 switch (attr
->group
) {
1273 case KVM_S390_VM_MEM_CTRL
:
1274 ret
= kvm_s390_set_mem_control(kvm
, attr
);
1276 case KVM_S390_VM_TOD
:
1277 ret
= kvm_s390_set_tod(kvm
, attr
);
1279 case KVM_S390_VM_CPU_MODEL
:
1280 ret
= kvm_s390_set_cpu_model(kvm
, attr
);
1282 case KVM_S390_VM_CRYPTO
:
1283 ret
= kvm_s390_vm_set_crypto(kvm
, attr
);
1285 case KVM_S390_VM_MIGRATION
:
1286 ret
= kvm_s390_vm_set_migration(kvm
, attr
);
1296 static int kvm_s390_vm_get_attr(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1300 switch (attr
->group
) {
1301 case KVM_S390_VM_MEM_CTRL
:
1302 ret
= kvm_s390_get_mem_control(kvm
, attr
);
1304 case KVM_S390_VM_TOD
:
1305 ret
= kvm_s390_get_tod(kvm
, attr
);
1307 case KVM_S390_VM_CPU_MODEL
:
1308 ret
= kvm_s390_get_cpu_model(kvm
, attr
);
1310 case KVM_S390_VM_MIGRATION
:
1311 ret
= kvm_s390_vm_get_migration(kvm
, attr
);
1321 static int kvm_s390_vm_has_attr(struct kvm
*kvm
, struct kvm_device_attr
*attr
)
1325 switch (attr
->group
) {
1326 case KVM_S390_VM_MEM_CTRL
:
1327 switch (attr
->attr
) {
1328 case KVM_S390_VM_MEM_ENABLE_CMMA
:
1329 case KVM_S390_VM_MEM_CLR_CMMA
:
1330 ret
= sclp
.has_cmma
? 0 : -ENXIO
;
1332 case KVM_S390_VM_MEM_LIMIT_SIZE
:
1340 case KVM_S390_VM_TOD
:
1341 switch (attr
->attr
) {
1342 case KVM_S390_VM_TOD_LOW
:
1343 case KVM_S390_VM_TOD_HIGH
:
1351 case KVM_S390_VM_CPU_MODEL
:
1352 switch (attr
->attr
) {
1353 case KVM_S390_VM_CPU_PROCESSOR
:
1354 case KVM_S390_VM_CPU_MACHINE
:
1355 case KVM_S390_VM_CPU_PROCESSOR_FEAT
:
1356 case KVM_S390_VM_CPU_MACHINE_FEAT
:
1357 case KVM_S390_VM_CPU_MACHINE_SUBFUNC
:
1360 /* configuring subfunctions is not supported yet */
1361 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC
:
1367 case KVM_S390_VM_CRYPTO
:
1368 switch (attr
->attr
) {
1369 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW
:
1370 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW
:
1371 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW
:
1372 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW
:
1380 case KVM_S390_VM_MIGRATION
:
1391 static long kvm_s390_get_skeys(struct kvm
*kvm
, struct kvm_s390_skeys
*args
)
1395 int srcu_idx
, i
, r
= 0;
1397 if (args
->flags
!= 0)
1400 /* Is this guest using storage keys? */
1401 if (!mm_use_skey(current
->mm
))
1402 return KVM_S390_GET_SKEYS_NONE
;
1404 /* Enforce sane limit on memory allocation */
1405 if (args
->count
< 1 || args
->count
> KVM_S390_SKEYS_MAX
)
1408 keys
= kvmalloc_array(args
->count
, sizeof(uint8_t), GFP_KERNEL
);
1412 down_read(¤t
->mm
->mmap_sem
);
1413 srcu_idx
= srcu_read_lock(&kvm
->srcu
);
1414 for (i
= 0; i
< args
->count
; i
++) {
1415 hva
= gfn_to_hva(kvm
, args
->start_gfn
+ i
);
1416 if (kvm_is_error_hva(hva
)) {
1421 r
= get_guest_storage_key(current
->mm
, hva
, &keys
[i
]);
1425 srcu_read_unlock(&kvm
->srcu
, srcu_idx
);
1426 up_read(¤t
->mm
->mmap_sem
);
1429 r
= copy_to_user((uint8_t __user
*)args
->skeydata_addr
, keys
,
1430 sizeof(uint8_t) * args
->count
);
1439 static long kvm_s390_set_skeys(struct kvm
*kvm
, struct kvm_s390_skeys
*args
)
1443 int srcu_idx
, i
, r
= 0;
1445 if (args
->flags
!= 0)
1448 /* Enforce sane limit on memory allocation */
1449 if (args
->count
< 1 || args
->count
> KVM_S390_SKEYS_MAX
)
1452 keys
= kvmalloc_array(args
->count
, sizeof(uint8_t), GFP_KERNEL
);
1456 r
= copy_from_user(keys
, (uint8_t __user
*)args
->skeydata_addr
,
1457 sizeof(uint8_t) * args
->count
);
1463 /* Enable storage key handling for the guest */
1464 r
= s390_enable_skey();
1468 down_read(¤t
->mm
->mmap_sem
);
1469 srcu_idx
= srcu_read_lock(&kvm
->srcu
);
1470 for (i
= 0; i
< args
->count
; i
++) {
1471 hva
= gfn_to_hva(kvm
, args
->start_gfn
+ i
);
1472 if (kvm_is_error_hva(hva
)) {
1477 /* Lowest order bit is reserved */
1478 if (keys
[i
] & 0x01) {
1483 r
= set_guest_storage_key(current
->mm
, hva
, keys
[i
], 0);
1487 srcu_read_unlock(&kvm
->srcu
, srcu_idx
);
1488 up_read(¤t
->mm
->mmap_sem
);
1495 * Base address and length must be sent at the start of each block, therefore
1496 * it's cheaper to send some clean data, as long as it's less than the size of
1499 #define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
1500 /* for consistency */
1501 #define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
1504 * This function searches for the next page with dirty CMMA attributes, and
1505 * saves the attributes in the buffer up to either the end of the buffer or
1506 * until a block of at least KVM_S390_MAX_BIT_DISTANCE clean bits is found;
1507 * no trailing clean bytes are saved.
1508 * In case no dirty bits were found, or if CMMA was not enabled or used, the
1509 * output buffer will indicate 0 as length.
1511 static int kvm_s390_get_cmma_bits(struct kvm
*kvm
,
1512 struct kvm_s390_cmma_log
*args
)
1514 struct kvm_s390_migration_state
*s
= kvm
->arch
.migration_state
;
1515 unsigned long bufsize
, hva
, pgstev
, i
, next
, cur
;
1516 int srcu_idx
, peek
, r
= 0, rr
;
1519 cur
= args
->start_gfn
;
1520 i
= next
= pgstev
= 0;
1522 if (unlikely(!kvm
->arch
.use_cmma
))
1524 /* Invalid/unsupported flags were specified */
1525 if (args
->flags
& ~KVM_S390_CMMA_PEEK
)
1527 /* Migration mode query, and we are not doing a migration */
1528 peek
= !!(args
->flags
& KVM_S390_CMMA_PEEK
);
1531 /* CMMA is disabled or was not used, or the buffer has length zero */
1532 bufsize
= min(args
->count
, KVM_S390_CMMA_SIZE_MAX
);
1533 if (!bufsize
|| !kvm
->mm
->context
.use_cmma
) {
1534 memset(args
, 0, sizeof(*args
));
1539 /* We are not peeking, and there are no dirty pages */
1540 if (!atomic64_read(&s
->dirty_pages
)) {
1541 memset(args
, 0, sizeof(*args
));
1544 cur
= find_next_bit(s
->pgste_bitmap
, s
->bitmap_size
,
1546 if (cur
>= s
->bitmap_size
) /* nothing found, loop back */
1547 cur
= find_next_bit(s
->pgste_bitmap
, s
->bitmap_size
, 0);
1548 if (cur
>= s
->bitmap_size
) { /* again! (very unlikely) */
1549 memset(args
, 0, sizeof(*args
));
1552 next
= find_next_bit(s
->pgste_bitmap
, s
->bitmap_size
, cur
+ 1);
1555 res
= vmalloc(bufsize
);
1559 args
->start_gfn
= cur
;
1561 down_read(&kvm
->mm
->mmap_sem
);
1562 srcu_idx
= srcu_read_lock(&kvm
->srcu
);
1563 while (i
< bufsize
) {
1564 hva
= gfn_to_hva(kvm
, cur
);
1565 if (kvm_is_error_hva(hva
)) {
1569 /* decrement only if we actually flipped the bit to 0 */
1570 if (!peek
&& test_and_clear_bit(cur
, s
->pgste_bitmap
))
1571 atomic64_dec(&s
->dirty_pages
);
1572 r
= get_pgste(kvm
->mm
, hva
, &pgstev
);
1575 /* save the value */
1576 res
[i
++] = (pgstev
>> 24) & 0x43;
1578 * if the next bit is too far away, stop.
1579 * if we reached the previous "next", find the next one
1582 if (next
> cur
+ KVM_S390_MAX_BIT_DISTANCE
)
1585 next
= find_next_bit(s
->pgste_bitmap
,
1586 s
->bitmap_size
, cur
+ 1);
1587 /* reached the end of the bitmap or of the buffer, stop */
1588 if ((next
>= s
->bitmap_size
) ||
1589 (next
>= args
->start_gfn
+ bufsize
))
1594 srcu_read_unlock(&kvm
->srcu
, srcu_idx
);
1595 up_read(&kvm
->mm
->mmap_sem
);
1597 args
->remaining
= s
? atomic64_read(&s
->dirty_pages
) : 0;
1599 rr
= copy_to_user((void __user
*)args
->values
, res
, args
->count
);
1608 * This function sets the CMMA attributes for the given pages. If the input
1609 * buffer has zero length, no action is taken, otherwise the attributes are
1610 * set and the mm->context.use_cmma flag is set.
1612 static int kvm_s390_set_cmma_bits(struct kvm
*kvm
,
1613 const struct kvm_s390_cmma_log
*args
)
1615 unsigned long hva
, mask
, pgstev
, i
;
1617 int srcu_idx
, r
= 0;
1621 if (!kvm
->arch
.use_cmma
)
1623 /* invalid/unsupported flags */
1624 if (args
->flags
!= 0)
1626 /* Enforce sane limit on memory allocation */
1627 if (args
->count
> KVM_S390_CMMA_SIZE_MAX
)
1630 if (args
->count
== 0)
1633 bits
= vmalloc(sizeof(*bits
) * args
->count
);
1637 r
= copy_from_user(bits
, (void __user
*)args
->values
, args
->count
);
1643 down_read(&kvm
->mm
->mmap_sem
);
1644 srcu_idx
= srcu_read_lock(&kvm
->srcu
);
1645 for (i
= 0; i
< args
->count
; i
++) {
1646 hva
= gfn_to_hva(kvm
, args
->start_gfn
+ i
);
1647 if (kvm_is_error_hva(hva
)) {
1653 pgstev
= pgstev
<< 24;
1654 mask
&= _PGSTE_GPS_USAGE_MASK
| _PGSTE_GPS_NODAT
;
1655 set_pgste_bits(kvm
->mm
, hva
, mask
, pgstev
);
1657 srcu_read_unlock(&kvm
->srcu
, srcu_idx
);
1658 up_read(&kvm
->mm
->mmap_sem
);
1660 if (!kvm
->mm
->context
.use_cmma
) {
1661 down_write(&kvm
->mm
->mmap_sem
);
1662 kvm
->mm
->context
.use_cmma
= 1;
1663 up_write(&kvm
->mm
->mmap_sem
);
1670 long kvm_arch_vm_ioctl(struct file
*filp
,
1671 unsigned int ioctl
, unsigned long arg
)
1673 struct kvm
*kvm
= filp
->private_data
;
1674 void __user
*argp
= (void __user
*)arg
;
1675 struct kvm_device_attr attr
;
1679 case KVM_S390_INTERRUPT
: {
1680 struct kvm_s390_interrupt s390int
;
1683 if (copy_from_user(&s390int
, argp
, sizeof(s390int
)))
1685 r
= kvm_s390_inject_vm(kvm
, &s390int
);
1688 case KVM_ENABLE_CAP
: {
1689 struct kvm_enable_cap cap
;
1691 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
1693 r
= kvm_vm_ioctl_enable_cap(kvm
, &cap
);
1696 case KVM_CREATE_IRQCHIP
: {
1697 struct kvm_irq_routing_entry routing
;
1700 if (kvm
->arch
.use_irqchip
) {
1701 /* Set up dummy routing. */
1702 memset(&routing
, 0, sizeof(routing
));
1703 r
= kvm_set_irq_routing(kvm
, &routing
, 0, 0);
1707 case KVM_SET_DEVICE_ATTR
: {
1709 if (copy_from_user(&attr
, (void __user
*)arg
, sizeof(attr
)))
1711 r
= kvm_s390_vm_set_attr(kvm
, &attr
);
1714 case KVM_GET_DEVICE_ATTR
: {
1716 if (copy_from_user(&attr
, (void __user
*)arg
, sizeof(attr
)))
1718 r
= kvm_s390_vm_get_attr(kvm
, &attr
);
1721 case KVM_HAS_DEVICE_ATTR
: {
1723 if (copy_from_user(&attr
, (void __user
*)arg
, sizeof(attr
)))
1725 r
= kvm_s390_vm_has_attr(kvm
, &attr
);
1728 case KVM_S390_GET_SKEYS
: {
1729 struct kvm_s390_skeys args
;
1732 if (copy_from_user(&args
, argp
,
1733 sizeof(struct kvm_s390_skeys
)))
1735 r
= kvm_s390_get_skeys(kvm
, &args
);
1738 case KVM_S390_SET_SKEYS
: {
1739 struct kvm_s390_skeys args
;
1742 if (copy_from_user(&args
, argp
,
1743 sizeof(struct kvm_s390_skeys
)))
1745 r
= kvm_s390_set_skeys(kvm
, &args
);
1748 case KVM_S390_GET_CMMA_BITS
: {
1749 struct kvm_s390_cmma_log args
;
1752 if (copy_from_user(&args
, argp
, sizeof(args
)))
1754 r
= kvm_s390_get_cmma_bits(kvm
, &args
);
1756 r
= copy_to_user(argp
, &args
, sizeof(args
));
1762 case KVM_S390_SET_CMMA_BITS
: {
1763 struct kvm_s390_cmma_log args
;
1766 if (copy_from_user(&args
, argp
, sizeof(args
)))
1768 r
= kvm_s390_set_cmma_bits(kvm
, &args
);
1778 static int kvm_s390_query_ap_config(u8
*config
)
1780 u32 fcn_code
= 0x04000000UL
;
1783 memset(config
, 0, 128);
1787 ".long 0xb2af0000\n" /* PQAP(QCI) */
1793 : "r" (fcn_code
), "r" (config
)
1794 : "cc", "0", "2", "memory"
1800 static int kvm_s390_apxa_installed(void)
1805 if (test_facility(12)) {
1806 cc
= kvm_s390_query_ap_config(config
);
1809 pr_err("PQAP(QCI) failed with cc=%d", cc
);
1811 return config
[0] & 0x40;
1817 static void kvm_s390_set_crycb_format(struct kvm
*kvm
)
1819 kvm
->arch
.crypto
.crycbd
= (__u32
)(unsigned long) kvm
->arch
.crypto
.crycb
;
1821 if (kvm_s390_apxa_installed())
1822 kvm
->arch
.crypto
.crycbd
|= CRYCB_FORMAT2
;
1824 kvm
->arch
.crypto
.crycbd
|= CRYCB_FORMAT1
;
1827 static u64
kvm_s390_get_initial_cpuid(void)
1832 cpuid
.version
= 0xff;
1833 return *((u64
*) &cpuid
);
1836 static void kvm_s390_crypto_init(struct kvm
*kvm
)
1838 if (!test_kvm_facility(kvm
, 76))
1841 kvm
->arch
.crypto
.crycb
= &kvm
->arch
.sie_page2
->crycb
;
1842 kvm_s390_set_crycb_format(kvm
);
1844 /* Enable AES/DEA protected key functions by default */
1845 kvm
->arch
.crypto
.aes_kw
= 1;
1846 kvm
->arch
.crypto
.dea_kw
= 1;
1847 get_random_bytes(kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
,
1848 sizeof(kvm
->arch
.crypto
.crycb
->aes_wrapping_key_mask
));
1849 get_random_bytes(kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
,
1850 sizeof(kvm
->arch
.crypto
.crycb
->dea_wrapping_key_mask
));
1853 static void sca_dispose(struct kvm
*kvm
)
1855 if (kvm
->arch
.use_esca
)
1856 free_pages_exact(kvm
->arch
.sca
, sizeof(struct esca_block
));
1858 free_page((unsigned long)(kvm
->arch
.sca
));
1859 kvm
->arch
.sca
= NULL
;
1862 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
)
1864 gfp_t alloc_flags
= GFP_KERNEL
;
1866 char debug_name
[16];
1867 static unsigned long sca_offset
;
1870 #ifdef CONFIG_KVM_S390_UCONTROL
1871 if (type
& ~KVM_VM_S390_UCONTROL
)
1873 if ((type
& KVM_VM_S390_UCONTROL
) && (!capable(CAP_SYS_ADMIN
)))
1880 rc
= s390_enable_sie();
1886 kvm
->arch
.use_esca
= 0; /* start with basic SCA */
1887 if (!sclp
.has_64bscao
)
1888 alloc_flags
|= GFP_DMA
;
1889 rwlock_init(&kvm
->arch
.sca_lock
);
1890 kvm
->arch
.sca
= (struct bsca_block
*) get_zeroed_page(alloc_flags
);
1893 spin_lock(&kvm_lock
);
1895 if (sca_offset
+ sizeof(struct bsca_block
) > PAGE_SIZE
)
1897 kvm
->arch
.sca
= (struct bsca_block
*)
1898 ((char *) kvm
->arch
.sca
+ sca_offset
);
1899 spin_unlock(&kvm_lock
);
1901 sprintf(debug_name
, "kvm-%u", current
->pid
);
1903 kvm
->arch
.dbf
= debug_register(debug_name
, 32, 1, 7 * sizeof(long));
1907 kvm
->arch
.sie_page2
=
1908 (struct sie_page2
*) get_zeroed_page(GFP_KERNEL
| GFP_DMA
);
1909 if (!kvm
->arch
.sie_page2
)
1912 /* Populate the facility mask initially. */
1913 memcpy(kvm
->arch
.model
.fac_mask
, S390_lowcore
.stfle_fac_list
,
1914 sizeof(S390_lowcore
.stfle_fac_list
));
1915 for (i
= 0; i
< S390_ARCH_FAC_LIST_SIZE_U64
; i
++) {
1916 if (i
< kvm_s390_fac_list_mask_size())
1917 kvm
->arch
.model
.fac_mask
[i
] &= kvm_s390_fac_list_mask
[i
];
1919 kvm
->arch
.model
.fac_mask
[i
] = 0UL;
1922 /* Populate the facility list initially. */
1923 kvm
->arch
.model
.fac_list
= kvm
->arch
.sie_page2
->fac_list
;
1924 memcpy(kvm
->arch
.model
.fac_list
, kvm
->arch
.model
.fac_mask
,
1925 S390_ARCH_FAC_LIST_SIZE_BYTE
);
1927 /* we are always in czam mode - even on pre z14 machines */
1928 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 138);
1929 set_kvm_facility(kvm
->arch
.model
.fac_list
, 138);
1930 /* we emulate STHYI in kvm */
1931 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 74);
1932 set_kvm_facility(kvm
->arch
.model
.fac_list
, 74);
1933 if (MACHINE_HAS_TLB_GUEST
) {
1934 set_kvm_facility(kvm
->arch
.model
.fac_mask
, 147);
1935 set_kvm_facility(kvm
->arch
.model
.fac_list
, 147);
1938 kvm
->arch
.model
.cpuid
= kvm_s390_get_initial_cpuid();
1939 kvm
->arch
.model
.ibc
= sclp
.ibc
& 0x0fff;
1941 kvm_s390_crypto_init(kvm
);
1943 mutex_init(&kvm
->arch
.float_int
.ais_lock
);
1944 kvm
->arch
.float_int
.simm
= 0;
1945 kvm
->arch
.float_int
.nimm
= 0;
1946 spin_lock_init(&kvm
->arch
.float_int
.lock
);
1947 for (i
= 0; i
< FIRQ_LIST_COUNT
; i
++)
1948 INIT_LIST_HEAD(&kvm
->arch
.float_int
.lists
[i
]);
1949 init_waitqueue_head(&kvm
->arch
.ipte_wq
);
1950 mutex_init(&kvm
->arch
.ipte_mutex
);
1952 debug_register_view(kvm
->arch
.dbf
, &debug_sprintf_view
);
1953 VM_EVENT(kvm
, 3, "vm created with type %lu", type
);
1955 if (type
& KVM_VM_S390_UCONTROL
) {
1956 kvm
->arch
.gmap
= NULL
;
1957 kvm
->arch
.mem_limit
= KVM_S390_NO_MEM_LIMIT
;
1959 if (sclp
.hamax
== U64_MAX
)
1960 kvm
->arch
.mem_limit
= TASK_SIZE_MAX
;
1962 kvm
->arch
.mem_limit
= min_t(unsigned long, TASK_SIZE_MAX
,
1964 kvm
->arch
.gmap
= gmap_create(current
->mm
, kvm
->arch
.mem_limit
- 1);
1965 if (!kvm
->arch
.gmap
)
1967 kvm
->arch
.gmap
->private = kvm
;
1968 kvm
->arch
.gmap
->pfault_enabled
= 0;
1971 kvm
->arch
.css_support
= 0;
1972 kvm
->arch
.use_irqchip
= 0;
1973 kvm
->arch
.epoch
= 0;
1975 spin_lock_init(&kvm
->arch
.start_stop_lock
);
1976 kvm_s390_vsie_init(kvm
);
1977 KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm
, current
->pid
);
1981 free_page((unsigned long)kvm
->arch
.sie_page2
);
1982 debug_unregister(kvm
->arch
.dbf
);
1984 KVM_EVENT(3, "creation of vm failed: %d", rc
);
1988 bool kvm_arch_has_vcpu_debugfs(void)
1993 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
)
1998 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
2000 VCPU_EVENT(vcpu
, 3, "%s", "free cpu");
2001 trace_kvm_s390_destroy_vcpu(vcpu
->vcpu_id
);
2002 kvm_s390_clear_local_irqs(vcpu
);
2003 kvm_clear_async_pf_completion_queue(vcpu
);
2004 if (!kvm_is_ucontrol(vcpu
->kvm
))
2007 if (kvm_is_ucontrol(vcpu
->kvm
))
2008 gmap_remove(vcpu
->arch
.gmap
);
2010 if (vcpu
->kvm
->arch
.use_cmma
)
2011 kvm_s390_vcpu_unsetup_cmma(vcpu
);
2012 free_page((unsigned long)(vcpu
->arch
.sie_block
));
2014 kvm_vcpu_uninit(vcpu
);
2015 kmem_cache_free(kvm_vcpu_cache
, vcpu
);
2018 static void kvm_free_vcpus(struct kvm
*kvm
)
2021 struct kvm_vcpu
*vcpu
;
2023 kvm_for_each_vcpu(i
, vcpu
, kvm
)
2024 kvm_arch_vcpu_destroy(vcpu
);
2026 mutex_lock(&kvm
->lock
);
2027 for (i
= 0; i
< atomic_read(&kvm
->online_vcpus
); i
++)
2028 kvm
->vcpus
[i
] = NULL
;
2030 atomic_set(&kvm
->online_vcpus
, 0);
2031 mutex_unlock(&kvm
->lock
);
2034 void kvm_arch_destroy_vm(struct kvm
*kvm
)
2036 kvm_free_vcpus(kvm
);
2038 debug_unregister(kvm
->arch
.dbf
);
2039 free_page((unsigned long)kvm
->arch
.sie_page2
);
2040 if (!kvm_is_ucontrol(kvm
))
2041 gmap_remove(kvm
->arch
.gmap
);
2042 kvm_s390_destroy_adapters(kvm
);
2043 kvm_s390_clear_float_irqs(kvm
);
2044 kvm_s390_vsie_destroy(kvm
);
2045 if (kvm
->arch
.migration_state
) {
2046 vfree(kvm
->arch
.migration_state
->pgste_bitmap
);
2047 kfree(kvm
->arch
.migration_state
);
2049 KVM_EVENT(3, "vm 0x%pK destroyed", kvm
);
2052 /* Section: vcpu related */
2053 static int __kvm_ucontrol_vcpu_init(struct kvm_vcpu
*vcpu
)
2055 vcpu
->arch
.gmap
= gmap_create(current
->mm
, -1UL);
2056 if (!vcpu
->arch
.gmap
)
2058 vcpu
->arch
.gmap
->private = vcpu
->kvm
;
2063 static void sca_del_vcpu(struct kvm_vcpu
*vcpu
)
2065 if (!kvm_s390_use_sca_entries())
2067 read_lock(&vcpu
->kvm
->arch
.sca_lock
);
2068 if (vcpu
->kvm
->arch
.use_esca
) {
2069 struct esca_block
*sca
= vcpu
->kvm
->arch
.sca
;
2071 clear_bit_inv(vcpu
->vcpu_id
, (unsigned long *) sca
->mcn
);
2072 sca
->cpu
[vcpu
->vcpu_id
].sda
= 0;
2074 struct bsca_block
*sca
= vcpu
->kvm
->arch
.sca
;
2076 clear_bit_inv(vcpu
->vcpu_id
, (unsigned long *) &sca
->mcn
);
2077 sca
->cpu
[vcpu
->vcpu_id
].sda
= 0;
2079 read_unlock(&vcpu
->kvm
->arch
.sca_lock
);
2082 static void sca_add_vcpu(struct kvm_vcpu
*vcpu
)
2084 if (!kvm_s390_use_sca_entries()) {
2085 struct bsca_block
*sca
= vcpu
->kvm
->arch
.sca
;
2087 /* we still need the basic sca for the ipte control */
2088 vcpu
->arch
.sie_block
->scaoh
= (__u32
)(((__u64
)sca
) >> 32);
2089 vcpu
->arch
.sie_block
->scaol
= (__u32
)(__u64
)sca
;
2091 read_lock(&vcpu
->kvm
->arch
.sca_lock
);
2092 if (vcpu
->kvm
->arch
.use_esca
) {
2093 struct esca_block
*sca
= vcpu
->kvm
->arch
.sca
;
2095 sca
->cpu
[vcpu
->vcpu_id
].sda
= (__u64
) vcpu
->arch
.sie_block
;
2096 vcpu
->arch
.sie_block
->scaoh
= (__u32
)(((__u64
)sca
) >> 32);
2097 vcpu
->arch
.sie_block
->scaol
= (__u32
)(__u64
)sca
& ~0x3fU
;
2098 vcpu
->arch
.sie_block
->ecb2
|= ECB2_ESCA
;
2099 set_bit_inv(vcpu
->vcpu_id
, (unsigned long *) sca
->mcn
);
2101 struct bsca_block
*sca
= vcpu
->kvm
->arch
.sca
;
2103 sca
->cpu
[vcpu
->vcpu_id
].sda
= (__u64
) vcpu
->arch
.sie_block
;
2104 vcpu
->arch
.sie_block
->scaoh
= (__u32
)(((__u64
)sca
) >> 32);
2105 vcpu
->arch
.sie_block
->scaol
= (__u32
)(__u64
)sca
;
2106 set_bit_inv(vcpu
->vcpu_id
, (unsigned long *) &sca
->mcn
);
2108 read_unlock(&vcpu
->kvm
->arch
.sca_lock
);
2111 /* Basic SCA to Extended SCA data copy routines */
2112 static inline void sca_copy_entry(struct esca_entry
*d
, struct bsca_entry
*s
)
2115 d
->sigp_ctrl
.c
= s
->sigp_ctrl
.c
;
2116 d
->sigp_ctrl
.scn
= s
->sigp_ctrl
.scn
;
2119 static void sca_copy_b_to_e(struct esca_block
*d
, struct bsca_block
*s
)
2123 d
->ipte_control
= s
->ipte_control
;
2125 for (i
= 0; i
< KVM_S390_BSCA_CPU_SLOTS
; i
++)
2126 sca_copy_entry(&d
->cpu
[i
], &s
->cpu
[i
]);
2129 static int sca_switch_to_extended(struct kvm
*kvm
)
2131 struct bsca_block
*old_sca
= kvm
->arch
.sca
;
2132 struct esca_block
*new_sca
;
2133 struct kvm_vcpu
*vcpu
;
2134 unsigned int vcpu_idx
;
2137 new_sca
= alloc_pages_exact(sizeof(*new_sca
), GFP_KERNEL
|__GFP_ZERO
);
2141 scaoh
= (u32
)((u64
)(new_sca
) >> 32);
2142 scaol
= (u32
)(u64
)(new_sca
) & ~0x3fU
;
2144 kvm_s390_vcpu_block_all(kvm
);
2145 write_lock(&kvm
->arch
.sca_lock
);
2147 sca_copy_b_to_e(new_sca
, old_sca
);
2149 kvm_for_each_vcpu(vcpu_idx
, vcpu
, kvm
) {
2150 vcpu
->arch
.sie_block
->scaoh
= scaoh
;
2151 vcpu
->arch
.sie_block
->scaol
= scaol
;
2152 vcpu
->arch
.sie_block
->ecb2
|= ECB2_ESCA
;
2154 kvm
->arch
.sca
= new_sca
;
2155 kvm
->arch
.use_esca
= 1;
2157 write_unlock(&kvm
->arch
.sca_lock
);
2158 kvm_s390_vcpu_unblock_all(kvm
);
2160 free_page((unsigned long)old_sca
);
2162 VM_EVENT(kvm
, 2, "Switched to ESCA (0x%pK -> 0x%pK)",
2163 old_sca
, kvm
->arch
.sca
);
2167 static int sca_can_add_vcpu(struct kvm
*kvm
, unsigned int id
)
2171 if (!kvm_s390_use_sca_entries()) {
2172 if (id
< KVM_MAX_VCPUS
)
2176 if (id
< KVM_S390_BSCA_CPU_SLOTS
)
2178 if (!sclp
.has_esca
|| !sclp
.has_64bscao
)
2181 mutex_lock(&kvm
->lock
);
2182 rc
= kvm
->arch
.use_esca
? 0 : sca_switch_to_extended(kvm
);
2183 mutex_unlock(&kvm
->lock
);
2185 return rc
== 0 && id
< KVM_S390_ESCA_CPU_SLOTS
;
2188 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
2190 vcpu
->arch
.pfault_token
= KVM_S390_PFAULT_TOKEN_INVALID
;
2191 kvm_clear_async_pf_completion_queue(vcpu
);
2192 vcpu
->run
->kvm_valid_regs
= KVM_SYNC_PREFIX
|
2198 kvm_s390_set_prefix(vcpu
, 0);
2199 if (test_kvm_facility(vcpu
->kvm
, 64))
2200 vcpu
->run
->kvm_valid_regs
|= KVM_SYNC_RICCB
;
2201 if (test_kvm_facility(vcpu
->kvm
, 133))
2202 vcpu
->run
->kvm_valid_regs
|= KVM_SYNC_GSCB
;
2203 /* fprs can be synchronized via vrs, even if the guest has no vx. With
2204 * MACHINE_HAS_VX, (load|store)_fpu_regs() will work with vrs format.
2207 vcpu
->run
->kvm_valid_regs
|= KVM_SYNC_VRS
;
2209 vcpu
->run
->kvm_valid_regs
|= KVM_SYNC_FPRS
;
2211 if (kvm_is_ucontrol(vcpu
->kvm
))
2212 return __kvm_ucontrol_vcpu_init(vcpu
);
2217 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2218 static void __start_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2220 WARN_ON_ONCE(vcpu
->arch
.cputm_start
!= 0);
2221 raw_write_seqcount_begin(&vcpu
->arch
.cputm_seqcount
);
2222 vcpu
->arch
.cputm_start
= get_tod_clock_fast();
2223 raw_write_seqcount_end(&vcpu
->arch
.cputm_seqcount
);
2226 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2227 static void __stop_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2229 WARN_ON_ONCE(vcpu
->arch
.cputm_start
== 0);
2230 raw_write_seqcount_begin(&vcpu
->arch
.cputm_seqcount
);
2231 vcpu
->arch
.sie_block
->cputm
-= get_tod_clock_fast() - vcpu
->arch
.cputm_start
;
2232 vcpu
->arch
.cputm_start
= 0;
2233 raw_write_seqcount_end(&vcpu
->arch
.cputm_seqcount
);
2236 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2237 static void __enable_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2239 WARN_ON_ONCE(vcpu
->arch
.cputm_enabled
);
2240 vcpu
->arch
.cputm_enabled
= true;
2241 __start_cpu_timer_accounting(vcpu
);
2244 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2245 static void __disable_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2247 WARN_ON_ONCE(!vcpu
->arch
.cputm_enabled
);
2248 __stop_cpu_timer_accounting(vcpu
);
2249 vcpu
->arch
.cputm_enabled
= false;
2252 static void enable_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2254 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2255 __enable_cpu_timer_accounting(vcpu
);
2259 static void disable_cpu_timer_accounting(struct kvm_vcpu
*vcpu
)
2261 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2262 __disable_cpu_timer_accounting(vcpu
);
2266 /* set the cpu timer - may only be called from the VCPU thread itself */
2267 void kvm_s390_set_cpu_timer(struct kvm_vcpu
*vcpu
, __u64 cputm
)
2269 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2270 raw_write_seqcount_begin(&vcpu
->arch
.cputm_seqcount
);
2271 if (vcpu
->arch
.cputm_enabled
)
2272 vcpu
->arch
.cputm_start
= get_tod_clock_fast();
2273 vcpu
->arch
.sie_block
->cputm
= cputm
;
2274 raw_write_seqcount_end(&vcpu
->arch
.cputm_seqcount
);
2278 /* update and get the cpu timer - can also be called from other VCPU threads */
2279 __u64
kvm_s390_get_cpu_timer(struct kvm_vcpu
*vcpu
)
2284 if (unlikely(!vcpu
->arch
.cputm_enabled
))
2285 return vcpu
->arch
.sie_block
->cputm
;
2287 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2289 seq
= raw_read_seqcount(&vcpu
->arch
.cputm_seqcount
);
2291 * If the writer would ever execute a read in the critical
2292 * section, e.g. in irq context, we have a deadlock.
2294 WARN_ON_ONCE((seq
& 1) && smp_processor_id() == vcpu
->cpu
);
2295 value
= vcpu
->arch
.sie_block
->cputm
;
2296 /* if cputm_start is 0, accounting is being started/stopped */
2297 if (likely(vcpu
->arch
.cputm_start
))
2298 value
-= get_tod_clock_fast() - vcpu
->arch
.cputm_start
;
2299 } while (read_seqcount_retry(&vcpu
->arch
.cputm_seqcount
, seq
& ~1));
2304 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
2307 gmap_enable(vcpu
->arch
.enabled_gmap
);
2308 atomic_or(CPUSTAT_RUNNING
, &vcpu
->arch
.sie_block
->cpuflags
);
2309 if (vcpu
->arch
.cputm_enabled
&& !is_vcpu_idle(vcpu
))
2310 __start_cpu_timer_accounting(vcpu
);
2314 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
2317 if (vcpu
->arch
.cputm_enabled
&& !is_vcpu_idle(vcpu
))
2318 __stop_cpu_timer_accounting(vcpu
);
2319 atomic_andnot(CPUSTAT_RUNNING
, &vcpu
->arch
.sie_block
->cpuflags
);
2320 vcpu
->arch
.enabled_gmap
= gmap_get_enabled();
2321 gmap_disable(vcpu
->arch
.enabled_gmap
);
2325 static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu
*vcpu
)
2327 /* this equals initial cpu reset in pop, but we don't switch to ESA */
2328 vcpu
->arch
.sie_block
->gpsw
.mask
= 0UL;
2329 vcpu
->arch
.sie_block
->gpsw
.addr
= 0UL;
2330 kvm_s390_set_prefix(vcpu
, 0);
2331 kvm_s390_set_cpu_timer(vcpu
, 0);
2332 vcpu
->arch
.sie_block
->ckc
= 0UL;
2333 vcpu
->arch
.sie_block
->todpr
= 0;
2334 memset(vcpu
->arch
.sie_block
->gcr
, 0, 16 * sizeof(__u64
));
2335 vcpu
->arch
.sie_block
->gcr
[0] = 0xE0UL
;
2336 vcpu
->arch
.sie_block
->gcr
[14] = 0xC2000000UL
;
2337 /* make sure the new fpc will be lazily loaded */
2339 current
->thread
.fpu
.fpc
= 0;
2340 vcpu
->arch
.sie_block
->gbea
= 1;
2341 vcpu
->arch
.sie_block
->pp
= 0;
2342 vcpu
->arch
.pfault_token
= KVM_S390_PFAULT_TOKEN_INVALID
;
2343 kvm_clear_async_pf_completion_queue(vcpu
);
2344 if (!kvm_s390_user_cpu_state_ctrl(vcpu
->kvm
))
2345 kvm_s390_vcpu_stop(vcpu
);
2346 kvm_s390_clear_local_irqs(vcpu
);
2349 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
)
2351 mutex_lock(&vcpu
->kvm
->lock
);
2353 vcpu
->arch
.sie_block
->epoch
= vcpu
->kvm
->arch
.epoch
;
2355 mutex_unlock(&vcpu
->kvm
->lock
);
2356 if (!kvm_is_ucontrol(vcpu
->kvm
)) {
2357 vcpu
->arch
.gmap
= vcpu
->kvm
->arch
.gmap
;
2360 if (test_kvm_facility(vcpu
->kvm
, 74) || vcpu
->kvm
->arch
.user_instr0
)
2361 vcpu
->arch
.sie_block
->ictl
|= ICTL_OPEREXC
;
2362 /* make vcpu_load load the right gmap on the first trigger */
2363 vcpu
->arch
.enabled_gmap
= vcpu
->arch
.gmap
;
2366 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu
*vcpu
)
2368 if (!test_kvm_facility(vcpu
->kvm
, 76))
2371 vcpu
->arch
.sie_block
->ecb3
&= ~(ECB3_AES
| ECB3_DEA
);
2373 if (vcpu
->kvm
->arch
.crypto
.aes_kw
)
2374 vcpu
->arch
.sie_block
->ecb3
|= ECB3_AES
;
2375 if (vcpu
->kvm
->arch
.crypto
.dea_kw
)
2376 vcpu
->arch
.sie_block
->ecb3
|= ECB3_DEA
;
2378 vcpu
->arch
.sie_block
->crycbd
= vcpu
->kvm
->arch
.crypto
.crycbd
;
2381 void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu
*vcpu
)
2383 free_page(vcpu
->arch
.sie_block
->cbrlo
);
2384 vcpu
->arch
.sie_block
->cbrlo
= 0;
2387 int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu
*vcpu
)
2389 vcpu
->arch
.sie_block
->cbrlo
= get_zeroed_page(GFP_KERNEL
);
2390 if (!vcpu
->arch
.sie_block
->cbrlo
)
2393 vcpu
->arch
.sie_block
->ecb2
&= ~ECB2_PFMFI
;
2397 static void kvm_s390_vcpu_setup_model(struct kvm_vcpu
*vcpu
)
2399 struct kvm_s390_cpu_model
*model
= &vcpu
->kvm
->arch
.model
;
2401 vcpu
->arch
.sie_block
->ibc
= model
->ibc
;
2402 if (test_kvm_facility(vcpu
->kvm
, 7))
2403 vcpu
->arch
.sie_block
->fac
= (u32
)(u64
) model
->fac_list
;
2406 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
2410 atomic_set(&vcpu
->arch
.sie_block
->cpuflags
, CPUSTAT_ZARCH
|
2414 if (test_kvm_facility(vcpu
->kvm
, 78))
2415 atomic_or(CPUSTAT_GED2
, &vcpu
->arch
.sie_block
->cpuflags
);
2416 else if (test_kvm_facility(vcpu
->kvm
, 8))
2417 atomic_or(CPUSTAT_GED
, &vcpu
->arch
.sie_block
->cpuflags
);
2419 kvm_s390_vcpu_setup_model(vcpu
);
2421 /* pgste_set_pte has special handling for !MACHINE_HAS_ESOP */
2422 if (MACHINE_HAS_ESOP
)
2423 vcpu
->arch
.sie_block
->ecb
|= ECB_HOSTPROTINT
;
2424 if (test_kvm_facility(vcpu
->kvm
, 9))
2425 vcpu
->arch
.sie_block
->ecb
|= ECB_SRSI
;
2426 if (test_kvm_facility(vcpu
->kvm
, 73))
2427 vcpu
->arch
.sie_block
->ecb
|= ECB_TE
;
2429 if (test_kvm_facility(vcpu
->kvm
, 8) && sclp
.has_pfmfi
)
2430 vcpu
->arch
.sie_block
->ecb2
|= ECB2_PFMFI
;
2431 if (test_kvm_facility(vcpu
->kvm
, 130))
2432 vcpu
->arch
.sie_block
->ecb2
|= ECB2_IEP
;
2433 vcpu
->arch
.sie_block
->eca
= ECA_MVPGI
| ECA_PROTEXCI
;
2435 vcpu
->arch
.sie_block
->eca
|= ECA_CEI
;
2437 vcpu
->arch
.sie_block
->eca
|= ECA_IB
;
2439 vcpu
->arch
.sie_block
->eca
|= ECA_SII
;
2440 if (sclp
.has_sigpif
)
2441 vcpu
->arch
.sie_block
->eca
|= ECA_SIGPI
;
2442 if (test_kvm_facility(vcpu
->kvm
, 129)) {
2443 vcpu
->arch
.sie_block
->eca
|= ECA_VX
;
2444 vcpu
->arch
.sie_block
->ecd
|= ECD_HOSTREGMGMT
;
2446 if (test_kvm_facility(vcpu
->kvm
, 139))
2447 vcpu
->arch
.sie_block
->ecd
|= ECD_MEF
;
2449 vcpu
->arch
.sie_block
->sdnxo
= ((unsigned long) &vcpu
->run
->s
.regs
.sdnx
)
2451 vcpu
->arch
.sie_block
->riccbd
= (unsigned long) &vcpu
->run
->s
.regs
.riccb
;
2454 atomic_or(CPUSTAT_KSS
, &vcpu
->arch
.sie_block
->cpuflags
);
2456 vcpu
->arch
.sie_block
->ictl
|= ICTL_ISKE
| ICTL_SSKE
| ICTL_RRBE
;
2458 if (vcpu
->kvm
->arch
.use_cmma
) {
2459 rc
= kvm_s390_vcpu_setup_cmma(vcpu
);
2463 hrtimer_init(&vcpu
->arch
.ckc_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
2464 vcpu
->arch
.ckc_timer
.function
= kvm_s390_idle_wakeup
;
2466 kvm_s390_vcpu_crypto_setup(vcpu
);
2471 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
,
2474 struct kvm_vcpu
*vcpu
;
2475 struct sie_page
*sie_page
;
2478 if (!kvm_is_ucontrol(kvm
) && !sca_can_add_vcpu(kvm
, id
))
2483 vcpu
= kmem_cache_zalloc(kvm_vcpu_cache
, GFP_KERNEL
);
2487 BUILD_BUG_ON(sizeof(struct sie_page
) != 4096);
2488 sie_page
= (struct sie_page
*) get_zeroed_page(GFP_KERNEL
);
2492 vcpu
->arch
.sie_block
= &sie_page
->sie_block
;
2493 vcpu
->arch
.sie_block
->itdba
= (unsigned long) &sie_page
->itdb
;
2495 /* the real guest size will always be smaller than msl */
2496 vcpu
->arch
.sie_block
->mso
= 0;
2497 vcpu
->arch
.sie_block
->msl
= sclp
.hamax
;
2499 vcpu
->arch
.sie_block
->icpua
= id
;
2500 spin_lock_init(&vcpu
->arch
.local_int
.lock
);
2501 vcpu
->arch
.local_int
.float_int
= &kvm
->arch
.float_int
;
2502 vcpu
->arch
.local_int
.wq
= &vcpu
->wq
;
2503 vcpu
->arch
.local_int
.cpuflags
= &vcpu
->arch
.sie_block
->cpuflags
;
2504 seqcount_init(&vcpu
->arch
.cputm_seqcount
);
2506 rc
= kvm_vcpu_init(vcpu
, kvm
, id
);
2508 goto out_free_sie_block
;
2509 VM_EVENT(kvm
, 3, "create cpu %d at 0x%pK, sie block at 0x%pK", id
, vcpu
,
2510 vcpu
->arch
.sie_block
);
2511 trace_kvm_s390_create_vcpu(id
, vcpu
, vcpu
->arch
.sie_block
);
2515 free_page((unsigned long)(vcpu
->arch
.sie_block
));
2517 kmem_cache_free(kvm_vcpu_cache
, vcpu
);
2522 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
)
2524 return kvm_s390_vcpu_has_irq(vcpu
, 0);
2527 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu
*vcpu
)
2529 return !(vcpu
->arch
.sie_block
->gpsw
.mask
& PSW_MASK_PSTATE
);
2532 void kvm_s390_vcpu_block(struct kvm_vcpu
*vcpu
)
2534 atomic_or(PROG_BLOCK_SIE
, &vcpu
->arch
.sie_block
->prog20
);
2538 void kvm_s390_vcpu_unblock(struct kvm_vcpu
*vcpu
)
2540 atomic_andnot(PROG_BLOCK_SIE
, &vcpu
->arch
.sie_block
->prog20
);
2543 static void kvm_s390_vcpu_request(struct kvm_vcpu
*vcpu
)
2545 atomic_or(PROG_REQUEST
, &vcpu
->arch
.sie_block
->prog20
);
2549 static void kvm_s390_vcpu_request_handled(struct kvm_vcpu
*vcpu
)
2551 atomic_andnot(PROG_REQUEST
, &vcpu
->arch
.sie_block
->prog20
);
2555 * Kick a guest cpu out of SIE and wait until SIE is not running.
2556 * If the CPU is not running (e.g. waiting as idle) the function will
2557 * return immediately. */
2558 void exit_sie(struct kvm_vcpu
*vcpu
)
2560 atomic_or(CPUSTAT_STOP_INT
, &vcpu
->arch
.sie_block
->cpuflags
);
2561 while (vcpu
->arch
.sie_block
->prog0c
& PROG_IN_SIE
)
2565 /* Kick a guest cpu out of SIE to process a request synchronously */
2566 void kvm_s390_sync_request(int req
, struct kvm_vcpu
*vcpu
)
2568 kvm_make_request(req
, vcpu
);
2569 kvm_s390_vcpu_request(vcpu
);
2572 static void kvm_gmap_notifier(struct gmap
*gmap
, unsigned long start
,
2575 struct kvm
*kvm
= gmap
->private;
2576 struct kvm_vcpu
*vcpu
;
2577 unsigned long prefix
;
2580 if (gmap_is_shadow(gmap
))
2582 if (start
>= 1UL << 31)
2583 /* We are only interested in prefix pages */
2585 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
2586 /* match against both prefix pages */
2587 prefix
= kvm_s390_get_prefix(vcpu
);
2588 if (prefix
<= end
&& start
<= prefix
+ 2*PAGE_SIZE
- 1) {
2589 VCPU_EVENT(vcpu
, 2, "gmap notifier for %lx-%lx",
2591 kvm_s390_sync_request(KVM_REQ_MMU_RELOAD
, vcpu
);
2596 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
)
2598 /* kvm common code refers to this, but never calls it */
2603 static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu
*vcpu
,
2604 struct kvm_one_reg
*reg
)
2609 case KVM_REG_S390_TODPR
:
2610 r
= put_user(vcpu
->arch
.sie_block
->todpr
,
2611 (u32 __user
*)reg
->addr
);
2613 case KVM_REG_S390_EPOCHDIFF
:
2614 r
= put_user(vcpu
->arch
.sie_block
->epoch
,
2615 (u64 __user
*)reg
->addr
);
2617 case KVM_REG_S390_CPU_TIMER
:
2618 r
= put_user(kvm_s390_get_cpu_timer(vcpu
),
2619 (u64 __user
*)reg
->addr
);
2621 case KVM_REG_S390_CLOCK_COMP
:
2622 r
= put_user(vcpu
->arch
.sie_block
->ckc
,
2623 (u64 __user
*)reg
->addr
);
2625 case KVM_REG_S390_PFTOKEN
:
2626 r
= put_user(vcpu
->arch
.pfault_token
,
2627 (u64 __user
*)reg
->addr
);
2629 case KVM_REG_S390_PFCOMPARE
:
2630 r
= put_user(vcpu
->arch
.pfault_compare
,
2631 (u64 __user
*)reg
->addr
);
2633 case KVM_REG_S390_PFSELECT
:
2634 r
= put_user(vcpu
->arch
.pfault_select
,
2635 (u64 __user
*)reg
->addr
);
2637 case KVM_REG_S390_PP
:
2638 r
= put_user(vcpu
->arch
.sie_block
->pp
,
2639 (u64 __user
*)reg
->addr
);
2641 case KVM_REG_S390_GBEA
:
2642 r
= put_user(vcpu
->arch
.sie_block
->gbea
,
2643 (u64 __user
*)reg
->addr
);
2652 static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu
*vcpu
,
2653 struct kvm_one_reg
*reg
)
2659 case KVM_REG_S390_TODPR
:
2660 r
= get_user(vcpu
->arch
.sie_block
->todpr
,
2661 (u32 __user
*)reg
->addr
);
2663 case KVM_REG_S390_EPOCHDIFF
:
2664 r
= get_user(vcpu
->arch
.sie_block
->epoch
,
2665 (u64 __user
*)reg
->addr
);
2667 case KVM_REG_S390_CPU_TIMER
:
2668 r
= get_user(val
, (u64 __user
*)reg
->addr
);
2670 kvm_s390_set_cpu_timer(vcpu
, val
);
2672 case KVM_REG_S390_CLOCK_COMP
:
2673 r
= get_user(vcpu
->arch
.sie_block
->ckc
,
2674 (u64 __user
*)reg
->addr
);
2676 case KVM_REG_S390_PFTOKEN
:
2677 r
= get_user(vcpu
->arch
.pfault_token
,
2678 (u64 __user
*)reg
->addr
);
2679 if (vcpu
->arch
.pfault_token
== KVM_S390_PFAULT_TOKEN_INVALID
)
2680 kvm_clear_async_pf_completion_queue(vcpu
);
2682 case KVM_REG_S390_PFCOMPARE
:
2683 r
= get_user(vcpu
->arch
.pfault_compare
,
2684 (u64 __user
*)reg
->addr
);
2686 case KVM_REG_S390_PFSELECT
:
2687 r
= get_user(vcpu
->arch
.pfault_select
,
2688 (u64 __user
*)reg
->addr
);
2690 case KVM_REG_S390_PP
:
2691 r
= get_user(vcpu
->arch
.sie_block
->pp
,
2692 (u64 __user
*)reg
->addr
);
2694 case KVM_REG_S390_GBEA
:
2695 r
= get_user(vcpu
->arch
.sie_block
->gbea
,
2696 (u64 __user
*)reg
->addr
);
2705 static int kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu
*vcpu
)
2707 kvm_s390_vcpu_initial_reset(vcpu
);
2711 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
2713 memcpy(&vcpu
->run
->s
.regs
.gprs
, ®s
->gprs
, sizeof(regs
->gprs
));
2717 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
2719 memcpy(®s
->gprs
, &vcpu
->run
->s
.regs
.gprs
, sizeof(regs
->gprs
));
2723 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
2724 struct kvm_sregs
*sregs
)
2726 memcpy(&vcpu
->run
->s
.regs
.acrs
, &sregs
->acrs
, sizeof(sregs
->acrs
));
2727 memcpy(&vcpu
->arch
.sie_block
->gcr
, &sregs
->crs
, sizeof(sregs
->crs
));
2731 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
2732 struct kvm_sregs
*sregs
)
2734 memcpy(&sregs
->acrs
, &vcpu
->run
->s
.regs
.acrs
, sizeof(sregs
->acrs
));
2735 memcpy(&sregs
->crs
, &vcpu
->arch
.sie_block
->gcr
, sizeof(sregs
->crs
));
2739 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
2741 if (test_fp_ctl(fpu
->fpc
))
2743 vcpu
->run
->s
.regs
.fpc
= fpu
->fpc
;
2745 convert_fp_to_vx((__vector128
*) vcpu
->run
->s
.regs
.vrs
,
2746 (freg_t
*) fpu
->fprs
);
2748 memcpy(vcpu
->run
->s
.regs
.fprs
, &fpu
->fprs
, sizeof(fpu
->fprs
));
2752 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
2754 /* make sure we have the latest values */
2757 convert_vx_to_fp((freg_t
*) fpu
->fprs
,
2758 (__vector128
*) vcpu
->run
->s
.regs
.vrs
);
2760 memcpy(fpu
->fprs
, vcpu
->run
->s
.regs
.fprs
, sizeof(fpu
->fprs
));
2761 fpu
->fpc
= vcpu
->run
->s
.regs
.fpc
;
2765 static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu
*vcpu
, psw_t psw
)
2769 if (!is_vcpu_stopped(vcpu
))
2772 vcpu
->run
->psw_mask
= psw
.mask
;
2773 vcpu
->run
->psw_addr
= psw
.addr
;
2778 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
2779 struct kvm_translation
*tr
)
2781 return -EINVAL
; /* not implemented yet */
2784 #define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
2785 KVM_GUESTDBG_USE_HW_BP | \
2786 KVM_GUESTDBG_ENABLE)
2788 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
2789 struct kvm_guest_debug
*dbg
)
2793 vcpu
->guest_debug
= 0;
2794 kvm_s390_clear_bp_data(vcpu
);
2796 if (dbg
->control
& ~VALID_GUESTDBG_FLAGS
)
2798 if (!sclp
.has_gpere
)
2801 if (dbg
->control
& KVM_GUESTDBG_ENABLE
) {
2802 vcpu
->guest_debug
= dbg
->control
;
2803 /* enforce guest PER */
2804 atomic_or(CPUSTAT_P
, &vcpu
->arch
.sie_block
->cpuflags
);
2806 if (dbg
->control
& KVM_GUESTDBG_USE_HW_BP
)
2807 rc
= kvm_s390_import_bp_data(vcpu
, dbg
);
2809 atomic_andnot(CPUSTAT_P
, &vcpu
->arch
.sie_block
->cpuflags
);
2810 vcpu
->arch
.guestdbg
.last_bp
= 0;
2814 vcpu
->guest_debug
= 0;
2815 kvm_s390_clear_bp_data(vcpu
);
2816 atomic_andnot(CPUSTAT_P
, &vcpu
->arch
.sie_block
->cpuflags
);
2822 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
2823 struct kvm_mp_state
*mp_state
)
2825 /* CHECK_STOP and LOAD are not supported yet */
2826 return is_vcpu_stopped(vcpu
) ? KVM_MP_STATE_STOPPED
:
2827 KVM_MP_STATE_OPERATING
;
2830 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
2831 struct kvm_mp_state
*mp_state
)
2835 /* user space knows about this interface - let it control the state */
2836 vcpu
->kvm
->arch
.user_cpu_state_ctrl
= 1;
2838 switch (mp_state
->mp_state
) {
2839 case KVM_MP_STATE_STOPPED
:
2840 kvm_s390_vcpu_stop(vcpu
);
2842 case KVM_MP_STATE_OPERATING
:
2843 kvm_s390_vcpu_start(vcpu
);
2845 case KVM_MP_STATE_LOAD
:
2846 case KVM_MP_STATE_CHECK_STOP
:
2847 /* fall through - CHECK_STOP and LOAD are not supported yet */
2855 static bool ibs_enabled(struct kvm_vcpu
*vcpu
)
2857 return atomic_read(&vcpu
->arch
.sie_block
->cpuflags
) & CPUSTAT_IBS
;
2860 static int kvm_s390_handle_requests(struct kvm_vcpu
*vcpu
)
2863 kvm_s390_vcpu_request_handled(vcpu
);
2864 if (!kvm_request_pending(vcpu
))
2867 * We use MMU_RELOAD just to re-arm the ipte notifier for the
2868 * guest prefix page. gmap_mprotect_notify will wait on the ptl lock.
2869 * This ensures that the ipte instruction for this request has
2870 * already finished. We might race against a second unmapper that
2871 * wants to set the blocking bit. Lets just retry the request loop.
2873 if (kvm_check_request(KVM_REQ_MMU_RELOAD
, vcpu
)) {
2875 rc
= gmap_mprotect_notify(vcpu
->arch
.gmap
,
2876 kvm_s390_get_prefix(vcpu
),
2877 PAGE_SIZE
* 2, PROT_WRITE
);
2879 kvm_make_request(KVM_REQ_MMU_RELOAD
, vcpu
);
2885 if (kvm_check_request(KVM_REQ_TLB_FLUSH
, vcpu
)) {
2886 vcpu
->arch
.sie_block
->ihcpu
= 0xffff;
2890 if (kvm_check_request(KVM_REQ_ENABLE_IBS
, vcpu
)) {
2891 if (!ibs_enabled(vcpu
)) {
2892 trace_kvm_s390_enable_disable_ibs(vcpu
->vcpu_id
, 1);
2893 atomic_or(CPUSTAT_IBS
,
2894 &vcpu
->arch
.sie_block
->cpuflags
);
2899 if (kvm_check_request(KVM_REQ_DISABLE_IBS
, vcpu
)) {
2900 if (ibs_enabled(vcpu
)) {
2901 trace_kvm_s390_enable_disable_ibs(vcpu
->vcpu_id
, 0);
2902 atomic_andnot(CPUSTAT_IBS
,
2903 &vcpu
->arch
.sie_block
->cpuflags
);
2908 if (kvm_check_request(KVM_REQ_ICPT_OPEREXC
, vcpu
)) {
2909 vcpu
->arch
.sie_block
->ictl
|= ICTL_OPEREXC
;
2913 if (kvm_check_request(KVM_REQ_START_MIGRATION
, vcpu
)) {
2915 * Disable CMMA virtualization; we will emulate the ESSA
2916 * instruction manually, in order to provide additional
2917 * functionalities needed for live migration.
2919 vcpu
->arch
.sie_block
->ecb2
&= ~ECB2_CMMA
;
2923 if (kvm_check_request(KVM_REQ_STOP_MIGRATION
, vcpu
)) {
2925 * Re-enable CMMA virtualization if CMMA is available and
2928 if ((vcpu
->kvm
->arch
.use_cmma
) &&
2929 (vcpu
->kvm
->mm
->context
.use_cmma
))
2930 vcpu
->arch
.sie_block
->ecb2
|= ECB2_CMMA
;
2934 /* nothing to do, just clear the request */
2935 kvm_clear_request(KVM_REQ_UNHALT
, vcpu
);
2940 void kvm_s390_set_tod_clock_ext(struct kvm
*kvm
,
2941 const struct kvm_s390_vm_tod_clock
*gtod
)
2943 struct kvm_vcpu
*vcpu
;
2944 struct kvm_s390_tod_clock_ext htod
;
2947 mutex_lock(&kvm
->lock
);
2950 get_tod_clock_ext((char *)&htod
);
2952 kvm
->arch
.epoch
= gtod
->tod
- htod
.tod
;
2953 kvm
->arch
.epdx
= gtod
->epoch_idx
- htod
.epoch_idx
;
2955 if (kvm
->arch
.epoch
> gtod
->tod
)
2956 kvm
->arch
.epdx
-= 1;
2958 kvm_s390_vcpu_block_all(kvm
);
2959 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
2960 vcpu
->arch
.sie_block
->epoch
= kvm
->arch
.epoch
;
2961 vcpu
->arch
.sie_block
->epdx
= kvm
->arch
.epdx
;
2964 kvm_s390_vcpu_unblock_all(kvm
);
2966 mutex_unlock(&kvm
->lock
);
2969 void kvm_s390_set_tod_clock(struct kvm
*kvm
, u64 tod
)
2971 struct kvm_vcpu
*vcpu
;
2974 mutex_lock(&kvm
->lock
);
2976 kvm
->arch
.epoch
= tod
- get_tod_clock();
2977 kvm_s390_vcpu_block_all(kvm
);
2978 kvm_for_each_vcpu(i
, vcpu
, kvm
)
2979 vcpu
->arch
.sie_block
->epoch
= kvm
->arch
.epoch
;
2980 kvm_s390_vcpu_unblock_all(kvm
);
2982 mutex_unlock(&kvm
->lock
);
2986 * kvm_arch_fault_in_page - fault-in guest page if necessary
2987 * @vcpu: The corresponding virtual cpu
2988 * @gpa: Guest physical address
2989 * @writable: Whether the page should be writable or not
2991 * Make sure that a guest page has been faulted-in on the host.
2993 * Return: Zero on success, negative error code otherwise.
2995 long kvm_arch_fault_in_page(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int writable
)
2997 return gmap_fault(vcpu
->arch
.gmap
, gpa
,
2998 writable
? FAULT_FLAG_WRITE
: 0);
3001 static void __kvm_inject_pfault_token(struct kvm_vcpu
*vcpu
, bool start_token
,
3002 unsigned long token
)
3004 struct kvm_s390_interrupt inti
;
3005 struct kvm_s390_irq irq
;
3008 irq
.u
.ext
.ext_params2
= token
;
3009 irq
.type
= KVM_S390_INT_PFAULT_INIT
;
3010 WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu
, &irq
));
3012 inti
.type
= KVM_S390_INT_PFAULT_DONE
;
3013 inti
.parm64
= token
;
3014 WARN_ON_ONCE(kvm_s390_inject_vm(vcpu
->kvm
, &inti
));
3018 void kvm_arch_async_page_not_present(struct kvm_vcpu
*vcpu
,
3019 struct kvm_async_pf
*work
)
3021 trace_kvm_s390_pfault_init(vcpu
, work
->arch
.pfault_token
);
3022 __kvm_inject_pfault_token(vcpu
, true, work
->arch
.pfault_token
);
3025 void kvm_arch_async_page_present(struct kvm_vcpu
*vcpu
,
3026 struct kvm_async_pf
*work
)
3028 trace_kvm_s390_pfault_done(vcpu
, work
->arch
.pfault_token
);
3029 __kvm_inject_pfault_token(vcpu
, false, work
->arch
.pfault_token
);
3032 void kvm_arch_async_page_ready(struct kvm_vcpu
*vcpu
,
3033 struct kvm_async_pf
*work
)
3035 /* s390 will always inject the page directly */
3038 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu
*vcpu
)
3041 * s390 will always inject the page directly,
3042 * but we still want check_async_completion to cleanup
3047 static int kvm_arch_setup_async_pf(struct kvm_vcpu
*vcpu
)
3050 struct kvm_arch_async_pf arch
;
3053 if (vcpu
->arch
.pfault_token
== KVM_S390_PFAULT_TOKEN_INVALID
)
3055 if ((vcpu
->arch
.sie_block
->gpsw
.mask
& vcpu
->arch
.pfault_select
) !=
3056 vcpu
->arch
.pfault_compare
)
3058 if (psw_extint_disabled(vcpu
))
3060 if (kvm_s390_vcpu_has_irq(vcpu
, 0))
3062 if (!(vcpu
->arch
.sie_block
->gcr
[0] & 0x200ul
))
3064 if (!vcpu
->arch
.gmap
->pfault_enabled
)
3067 hva
= gfn_to_hva(vcpu
->kvm
, gpa_to_gfn(current
->thread
.gmap_addr
));
3068 hva
+= current
->thread
.gmap_addr
& ~PAGE_MASK
;
3069 if (read_guest_real(vcpu
, vcpu
->arch
.pfault_token
, &arch
.pfault_token
, 8))
3072 rc
= kvm_setup_async_pf(vcpu
, current
->thread
.gmap_addr
, hva
, &arch
);
3076 static int vcpu_pre_run(struct kvm_vcpu
*vcpu
)
3081 * On s390 notifications for arriving pages will be delivered directly
3082 * to the guest but the house keeping for completed pfaults is
3083 * handled outside the worker.
3085 kvm_check_async_pf_completion(vcpu
);
3087 vcpu
->arch
.sie_block
->gg14
= vcpu
->run
->s
.regs
.gprs
[14];
3088 vcpu
->arch
.sie_block
->gg15
= vcpu
->run
->s
.regs
.gprs
[15];
3093 if (test_cpu_flag(CIF_MCCK_PENDING
))
3096 if (!kvm_is_ucontrol(vcpu
->kvm
)) {
3097 rc
= kvm_s390_deliver_pending_interrupts(vcpu
);
3102 rc
= kvm_s390_handle_requests(vcpu
);
3106 if (guestdbg_enabled(vcpu
)) {
3107 kvm_s390_backup_guest_per_regs(vcpu
);
3108 kvm_s390_patch_guest_per_regs(vcpu
);
3111 vcpu
->arch
.sie_block
->icptcode
= 0;
3112 cpuflags
= atomic_read(&vcpu
->arch
.sie_block
->cpuflags
);
3113 VCPU_EVENT(vcpu
, 6, "entering sie flags %x", cpuflags
);
3114 trace_kvm_s390_sie_enter(vcpu
, cpuflags
);
3119 static int vcpu_post_run_fault_in_sie(struct kvm_vcpu
*vcpu
)
3121 struct kvm_s390_pgm_info pgm_info
= {
3122 .code
= PGM_ADDRESSING
,
3127 VCPU_EVENT(vcpu
, 3, "%s", "fault in sie instruction");
3128 trace_kvm_s390_sie_fault(vcpu
);
3131 * We want to inject an addressing exception, which is defined as a
3132 * suppressing or terminating exception. However, since we came here
3133 * by a DAT access exception, the PSW still points to the faulting
3134 * instruction since DAT exceptions are nullifying. So we've got
3135 * to look up the current opcode to get the length of the instruction
3136 * to be able to forward the PSW.
3138 rc
= read_guest_instr(vcpu
, vcpu
->arch
.sie_block
->gpsw
.addr
, &opcode
, 1);
3139 ilen
= insn_length(opcode
);
3143 /* Instruction-Fetching Exceptions - we can't detect the ilen.
3144 * Forward by arbitrary ilc, injection will take care of
3145 * nullification if necessary.
3147 pgm_info
= vcpu
->arch
.pgm
;
3150 pgm_info
.flags
= ilen
| KVM_S390_PGM_FLAGS_ILC_VALID
;
3151 kvm_s390_forward_psw(vcpu
, ilen
);
3152 return kvm_s390_inject_prog_irq(vcpu
, &pgm_info
);
3155 static int vcpu_post_run(struct kvm_vcpu
*vcpu
, int exit_reason
)
3157 struct mcck_volatile_info
*mcck_info
;
3158 struct sie_page
*sie_page
;
3160 VCPU_EVENT(vcpu
, 6, "exit sie icptcode %d",
3161 vcpu
->arch
.sie_block
->icptcode
);
3162 trace_kvm_s390_sie_exit(vcpu
, vcpu
->arch
.sie_block
->icptcode
);
3164 if (guestdbg_enabled(vcpu
))
3165 kvm_s390_restore_guest_per_regs(vcpu
);
3167 vcpu
->run
->s
.regs
.gprs
[14] = vcpu
->arch
.sie_block
->gg14
;
3168 vcpu
->run
->s
.regs
.gprs
[15] = vcpu
->arch
.sie_block
->gg15
;
3170 if (exit_reason
== -EINTR
) {
3171 VCPU_EVENT(vcpu
, 3, "%s", "machine check");
3172 sie_page
= container_of(vcpu
->arch
.sie_block
,
3173 struct sie_page
, sie_block
);
3174 mcck_info
= &sie_page
->mcck_info
;
3175 kvm_s390_reinject_machine_check(vcpu
, mcck_info
);
3179 if (vcpu
->arch
.sie_block
->icptcode
> 0) {
3180 int rc
= kvm_handle_sie_intercept(vcpu
);
3182 if (rc
!= -EOPNOTSUPP
)
3184 vcpu
->run
->exit_reason
= KVM_EXIT_S390_SIEIC
;
3185 vcpu
->run
->s390_sieic
.icptcode
= vcpu
->arch
.sie_block
->icptcode
;
3186 vcpu
->run
->s390_sieic
.ipa
= vcpu
->arch
.sie_block
->ipa
;
3187 vcpu
->run
->s390_sieic
.ipb
= vcpu
->arch
.sie_block
->ipb
;
3189 } else if (exit_reason
!= -EFAULT
) {
3190 vcpu
->stat
.exit_null
++;
3192 } else if (kvm_is_ucontrol(vcpu
->kvm
)) {
3193 vcpu
->run
->exit_reason
= KVM_EXIT_S390_UCONTROL
;
3194 vcpu
->run
->s390_ucontrol
.trans_exc_code
=
3195 current
->thread
.gmap_addr
;
3196 vcpu
->run
->s390_ucontrol
.pgm_code
= 0x10;
3198 } else if (current
->thread
.gmap_pfault
) {
3199 trace_kvm_s390_major_guest_pfault(vcpu
);
3200 current
->thread
.gmap_pfault
= 0;
3201 if (kvm_arch_setup_async_pf(vcpu
))
3203 return kvm_arch_fault_in_page(vcpu
, current
->thread
.gmap_addr
, 1);
3205 return vcpu_post_run_fault_in_sie(vcpu
);
3208 static int __vcpu_run(struct kvm_vcpu
*vcpu
)
3210 int rc
, exit_reason
;
3213 * We try to hold kvm->srcu during most of vcpu_run (except when run-
3214 * ning the guest), so that memslots (and other stuff) are protected
3216 vcpu
->srcu_idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
3219 rc
= vcpu_pre_run(vcpu
);
3223 srcu_read_unlock(&vcpu
->kvm
->srcu
, vcpu
->srcu_idx
);
3225 * As PF_VCPU will be used in fault handler, between
3226 * guest_enter and guest_exit should be no uaccess.
3228 local_irq_disable();
3229 guest_enter_irqoff();
3230 __disable_cpu_timer_accounting(vcpu
);
3232 exit_reason
= sie64a(vcpu
->arch
.sie_block
,
3233 vcpu
->run
->s
.regs
.gprs
);
3234 local_irq_disable();
3235 __enable_cpu_timer_accounting(vcpu
);
3236 guest_exit_irqoff();
3238 vcpu
->srcu_idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
3240 rc
= vcpu_post_run(vcpu
, exit_reason
);
3241 } while (!signal_pending(current
) && !guestdbg_exit_pending(vcpu
) && !rc
);
3243 srcu_read_unlock(&vcpu
->kvm
->srcu
, vcpu
->srcu_idx
);
3247 static void sync_regs(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
3249 struct runtime_instr_cb
*riccb
;
3252 riccb
= (struct runtime_instr_cb
*) &kvm_run
->s
.regs
.riccb
;
3253 gscb
= (struct gs_cb
*) &kvm_run
->s
.regs
.gscb
;
3254 vcpu
->arch
.sie_block
->gpsw
.mask
= kvm_run
->psw_mask
;
3255 vcpu
->arch
.sie_block
->gpsw
.addr
= kvm_run
->psw_addr
;
3256 if (kvm_run
->kvm_dirty_regs
& KVM_SYNC_PREFIX
)
3257 kvm_s390_set_prefix(vcpu
, kvm_run
->s
.regs
.prefix
);
3258 if (kvm_run
->kvm_dirty_regs
& KVM_SYNC_CRS
) {
3259 memcpy(&vcpu
->arch
.sie_block
->gcr
, &kvm_run
->s
.regs
.crs
, 128);
3260 /* some control register changes require a tlb flush */
3261 kvm_make_request(KVM_REQ_TLB_FLUSH
, vcpu
);
3263 if (kvm_run
->kvm_dirty_regs
& KVM_SYNC_ARCH0
) {
3264 kvm_s390_set_cpu_timer(vcpu
, kvm_run
->s
.regs
.cputm
);
3265 vcpu
->arch
.sie_block
->ckc
= kvm_run
->s
.regs
.ckc
;
3266 vcpu
->arch
.sie_block
->todpr
= kvm_run
->s
.regs
.todpr
;
3267 vcpu
->arch
.sie_block
->pp
= kvm_run
->s
.regs
.pp
;
3268 vcpu
->arch
.sie_block
->gbea
= kvm_run
->s
.regs
.gbea
;
3270 if (kvm_run
->kvm_dirty_regs
& KVM_SYNC_PFAULT
) {
3271 vcpu
->arch
.pfault_token
= kvm_run
->s
.regs
.pft
;
3272 vcpu
->arch
.pfault_select
= kvm_run
->s
.regs
.pfs
;
3273 vcpu
->arch
.pfault_compare
= kvm_run
->s
.regs
.pfc
;
3274 if (vcpu
->arch
.pfault_token
== KVM_S390_PFAULT_TOKEN_INVALID
)
3275 kvm_clear_async_pf_completion_queue(vcpu
);
3278 * If userspace sets the riccb (e.g. after migration) to a valid state,
3279 * we should enable RI here instead of doing the lazy enablement.
3281 if ((kvm_run
->kvm_dirty_regs
& KVM_SYNC_RICCB
) &&
3282 test_kvm_facility(vcpu
->kvm
, 64) &&
3284 !(vcpu
->arch
.sie_block
->ecb3
& ECB3_RI
)) {
3285 VCPU_EVENT(vcpu
, 3, "%s", "ENABLE: RI (sync_regs)");
3286 vcpu
->arch
.sie_block
->ecb3
|= ECB3_RI
;
3289 * If userspace sets the gscb (e.g. after migration) to non-zero,
3290 * we should enable GS here instead of doing the lazy enablement.
3292 if ((kvm_run
->kvm_dirty_regs
& KVM_SYNC_GSCB
) &&
3293 test_kvm_facility(vcpu
->kvm
, 133) &&
3295 !vcpu
->arch
.gs_enabled
) {
3296 VCPU_EVENT(vcpu
, 3, "%s", "ENABLE: GS (sync_regs)");
3297 vcpu
->arch
.sie_block
->ecb
|= ECB_GS
;
3298 vcpu
->arch
.sie_block
->ecd
|= ECD_HOSTREGMGMT
;
3299 vcpu
->arch
.gs_enabled
= 1;
3301 save_access_regs(vcpu
->arch
.host_acrs
);
3302 restore_access_regs(vcpu
->run
->s
.regs
.acrs
);
3303 /* save host (userspace) fprs/vrs */
3305 vcpu
->arch
.host_fpregs
.fpc
= current
->thread
.fpu
.fpc
;
3306 vcpu
->arch
.host_fpregs
.regs
= current
->thread
.fpu
.regs
;
3308 current
->thread
.fpu
.regs
= vcpu
->run
->s
.regs
.vrs
;
3310 current
->thread
.fpu
.regs
= vcpu
->run
->s
.regs
.fprs
;
3311 current
->thread
.fpu
.fpc
= vcpu
->run
->s
.regs
.fpc
;
3312 if (test_fp_ctl(current
->thread
.fpu
.fpc
))
3313 /* User space provided an invalid FPC, let's clear it */
3314 current
->thread
.fpu
.fpc
= 0;
3315 if (MACHINE_HAS_GS
) {
3317 __ctl_set_bit(2, 4);
3318 if (current
->thread
.gs_cb
) {
3319 vcpu
->arch
.host_gscb
= current
->thread
.gs_cb
;
3320 save_gs_cb(vcpu
->arch
.host_gscb
);
3322 if (vcpu
->arch
.gs_enabled
) {
3323 current
->thread
.gs_cb
= (struct gs_cb
*)
3324 &vcpu
->run
->s
.regs
.gscb
;
3325 restore_gs_cb(current
->thread
.gs_cb
);
3330 kvm_run
->kvm_dirty_regs
= 0;
3333 static void store_regs(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
3335 kvm_run
->psw_mask
= vcpu
->arch
.sie_block
->gpsw
.mask
;
3336 kvm_run
->psw_addr
= vcpu
->arch
.sie_block
->gpsw
.addr
;
3337 kvm_run
->s
.regs
.prefix
= kvm_s390_get_prefix(vcpu
);
3338 memcpy(&kvm_run
->s
.regs
.crs
, &vcpu
->arch
.sie_block
->gcr
, 128);
3339 kvm_run
->s
.regs
.cputm
= kvm_s390_get_cpu_timer(vcpu
);
3340 kvm_run
->s
.regs
.ckc
= vcpu
->arch
.sie_block
->ckc
;
3341 kvm_run
->s
.regs
.todpr
= vcpu
->arch
.sie_block
->todpr
;
3342 kvm_run
->s
.regs
.pp
= vcpu
->arch
.sie_block
->pp
;
3343 kvm_run
->s
.regs
.gbea
= vcpu
->arch
.sie_block
->gbea
;
3344 kvm_run
->s
.regs
.pft
= vcpu
->arch
.pfault_token
;
3345 kvm_run
->s
.regs
.pfs
= vcpu
->arch
.pfault_select
;
3346 kvm_run
->s
.regs
.pfc
= vcpu
->arch
.pfault_compare
;
3347 save_access_regs(vcpu
->run
->s
.regs
.acrs
);
3348 restore_access_regs(vcpu
->arch
.host_acrs
);
3349 /* Save guest register state */
3351 vcpu
->run
->s
.regs
.fpc
= current
->thread
.fpu
.fpc
;
3352 /* Restore will be done lazily at return */
3353 current
->thread
.fpu
.fpc
= vcpu
->arch
.host_fpregs
.fpc
;
3354 current
->thread
.fpu
.regs
= vcpu
->arch
.host_fpregs
.regs
;
3355 if (MACHINE_HAS_GS
) {
3356 __ctl_set_bit(2, 4);
3357 if (vcpu
->arch
.gs_enabled
)
3358 save_gs_cb(current
->thread
.gs_cb
);
3360 current
->thread
.gs_cb
= vcpu
->arch
.host_gscb
;
3361 restore_gs_cb(vcpu
->arch
.host_gscb
);
3363 if (!vcpu
->arch
.host_gscb
)
3364 __ctl_clear_bit(2, 4);
3365 vcpu
->arch
.host_gscb
= NULL
;
3370 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
3374 if (kvm_run
->immediate_exit
)
3377 if (guestdbg_exit_pending(vcpu
)) {
3378 kvm_s390_prepare_debug_exit(vcpu
);
3382 kvm_sigset_activate(vcpu
);
3384 if (!kvm_s390_user_cpu_state_ctrl(vcpu
->kvm
)) {
3385 kvm_s390_vcpu_start(vcpu
);
3386 } else if (is_vcpu_stopped(vcpu
)) {
3387 pr_err_ratelimited("can't run stopped vcpu %d\n",
3392 sync_regs(vcpu
, kvm_run
);
3393 enable_cpu_timer_accounting(vcpu
);
3396 rc
= __vcpu_run(vcpu
);
3398 if (signal_pending(current
) && !rc
) {
3399 kvm_run
->exit_reason
= KVM_EXIT_INTR
;
3403 if (guestdbg_exit_pending(vcpu
) && !rc
) {
3404 kvm_s390_prepare_debug_exit(vcpu
);
3408 if (rc
== -EREMOTE
) {
3409 /* userspace support is needed, kvm_run has been prepared */
3413 disable_cpu_timer_accounting(vcpu
);
3414 store_regs(vcpu
, kvm_run
);
3416 kvm_sigset_deactivate(vcpu
);
3418 vcpu
->stat
.exit_userspace
++;
3423 * store status at address
3424 * we use have two special cases:
3425 * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
3426 * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
3428 int kvm_s390_store_status_unloaded(struct kvm_vcpu
*vcpu
, unsigned long gpa
)
3430 unsigned char archmode
= 1;
3431 freg_t fprs
[NUM_FPRS
];
3436 px
= kvm_s390_get_prefix(vcpu
);
3437 if (gpa
== KVM_S390_STORE_STATUS_NOADDR
) {
3438 if (write_guest_abs(vcpu
, 163, &archmode
, 1))
3441 } else if (gpa
== KVM_S390_STORE_STATUS_PREFIXED
) {
3442 if (write_guest_real(vcpu
, 163, &archmode
, 1))
3446 gpa
-= __LC_FPREGS_SAVE_AREA
;
3448 /* manually convert vector registers if necessary */
3449 if (MACHINE_HAS_VX
) {
3450 convert_vx_to_fp(fprs
, (__vector128
*) vcpu
->run
->s
.regs
.vrs
);
3451 rc
= write_guest_abs(vcpu
, gpa
+ __LC_FPREGS_SAVE_AREA
,
3454 rc
= write_guest_abs(vcpu
, gpa
+ __LC_FPREGS_SAVE_AREA
,
3455 vcpu
->run
->s
.regs
.fprs
, 128);
3457 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_GPREGS_SAVE_AREA
,
3458 vcpu
->run
->s
.regs
.gprs
, 128);
3459 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_PSW_SAVE_AREA
,
3460 &vcpu
->arch
.sie_block
->gpsw
, 16);
3461 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_PREFIX_SAVE_AREA
,
3463 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_FP_CREG_SAVE_AREA
,
3464 &vcpu
->run
->s
.regs
.fpc
, 4);
3465 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_TOD_PROGREG_SAVE_AREA
,
3466 &vcpu
->arch
.sie_block
->todpr
, 4);
3467 cputm
= kvm_s390_get_cpu_timer(vcpu
);
3468 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_CPU_TIMER_SAVE_AREA
,
3470 clkcomp
= vcpu
->arch
.sie_block
->ckc
>> 8;
3471 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_CLOCK_COMP_SAVE_AREA
,
3473 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_AREGS_SAVE_AREA
,
3474 &vcpu
->run
->s
.regs
.acrs
, 64);
3475 rc
|= write_guest_abs(vcpu
, gpa
+ __LC_CREGS_SAVE_AREA
,
3476 &vcpu
->arch
.sie_block
->gcr
, 128);
3477 return rc
? -EFAULT
: 0;
3480 int kvm_s390_vcpu_store_status(struct kvm_vcpu
*vcpu
, unsigned long addr
)
3483 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
3484 * switch in the run ioctl. Let's update our copies before we save
3485 * it into the save area
3488 vcpu
->run
->s
.regs
.fpc
= current
->thread
.fpu
.fpc
;
3489 save_access_regs(vcpu
->run
->s
.regs
.acrs
);
3491 return kvm_s390_store_status_unloaded(vcpu
, addr
);
3494 static void __disable_ibs_on_vcpu(struct kvm_vcpu
*vcpu
)
3496 kvm_check_request(KVM_REQ_ENABLE_IBS
, vcpu
);
3497 kvm_s390_sync_request(KVM_REQ_DISABLE_IBS
, vcpu
);
3500 static void __disable_ibs_on_all_vcpus(struct kvm
*kvm
)
3503 struct kvm_vcpu
*vcpu
;
3505 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
3506 __disable_ibs_on_vcpu(vcpu
);
3510 static void __enable_ibs_on_vcpu(struct kvm_vcpu
*vcpu
)
3514 kvm_check_request(KVM_REQ_DISABLE_IBS
, vcpu
);
3515 kvm_s390_sync_request(KVM_REQ_ENABLE_IBS
, vcpu
);
3518 void kvm_s390_vcpu_start(struct kvm_vcpu
*vcpu
)
3520 int i
, online_vcpus
, started_vcpus
= 0;
3522 if (!is_vcpu_stopped(vcpu
))
3525 trace_kvm_s390_vcpu_start_stop(vcpu
->vcpu_id
, 1);
3526 /* Only one cpu at a time may enter/leave the STOPPED state. */
3527 spin_lock(&vcpu
->kvm
->arch
.start_stop_lock
);
3528 online_vcpus
= atomic_read(&vcpu
->kvm
->online_vcpus
);
3530 for (i
= 0; i
< online_vcpus
; i
++) {
3531 if (!is_vcpu_stopped(vcpu
->kvm
->vcpus
[i
]))
3535 if (started_vcpus
== 0) {
3536 /* we're the only active VCPU -> speed it up */
3537 __enable_ibs_on_vcpu(vcpu
);
3538 } else if (started_vcpus
== 1) {
3540 * As we are starting a second VCPU, we have to disable
3541 * the IBS facility on all VCPUs to remove potentially
3542 * oustanding ENABLE requests.
3544 __disable_ibs_on_all_vcpus(vcpu
->kvm
);
3547 atomic_andnot(CPUSTAT_STOPPED
, &vcpu
->arch
.sie_block
->cpuflags
);
3549 * Another VCPU might have used IBS while we were offline.
3550 * Let's play safe and flush the VCPU at startup.
3552 kvm_make_request(KVM_REQ_TLB_FLUSH
, vcpu
);
3553 spin_unlock(&vcpu
->kvm
->arch
.start_stop_lock
);
3557 void kvm_s390_vcpu_stop(struct kvm_vcpu
*vcpu
)
3559 int i
, online_vcpus
, started_vcpus
= 0;
3560 struct kvm_vcpu
*started_vcpu
= NULL
;
3562 if (is_vcpu_stopped(vcpu
))
3565 trace_kvm_s390_vcpu_start_stop(vcpu
->vcpu_id
, 0);
3566 /* Only one cpu at a time may enter/leave the STOPPED state. */
3567 spin_lock(&vcpu
->kvm
->arch
.start_stop_lock
);
3568 online_vcpus
= atomic_read(&vcpu
->kvm
->online_vcpus
);
3570 /* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
3571 kvm_s390_clear_stop_irq(vcpu
);
3573 atomic_or(CPUSTAT_STOPPED
, &vcpu
->arch
.sie_block
->cpuflags
);
3574 __disable_ibs_on_vcpu(vcpu
);
3576 for (i
= 0; i
< online_vcpus
; i
++) {
3577 if (!is_vcpu_stopped(vcpu
->kvm
->vcpus
[i
])) {
3579 started_vcpu
= vcpu
->kvm
->vcpus
[i
];
3583 if (started_vcpus
== 1) {
3585 * As we only have one VCPU left, we want to enable the
3586 * IBS facility for that VCPU to speed it up.
3588 __enable_ibs_on_vcpu(started_vcpu
);
3591 spin_unlock(&vcpu
->kvm
->arch
.start_stop_lock
);
3595 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
3596 struct kvm_enable_cap
*cap
)
3604 case KVM_CAP_S390_CSS_SUPPORT
:
3605 if (!vcpu
->kvm
->arch
.css_support
) {
3606 vcpu
->kvm
->arch
.css_support
= 1;
3607 VM_EVENT(vcpu
->kvm
, 3, "%s", "ENABLE: CSS support");
3608 trace_kvm_s390_enable_css(vcpu
->kvm
);
3619 static long kvm_s390_guest_mem_op(struct kvm_vcpu
*vcpu
,
3620 struct kvm_s390_mem_op
*mop
)
3622 void __user
*uaddr
= (void __user
*)mop
->buf
;
3623 void *tmpbuf
= NULL
;
3625 const u64 supported_flags
= KVM_S390_MEMOP_F_INJECT_EXCEPTION
3626 | KVM_S390_MEMOP_F_CHECK_ONLY
;
3628 if (mop
->flags
& ~supported_flags
)
3631 if (mop
->size
> MEM_OP_MAX_SIZE
)
3634 if (!(mop
->flags
& KVM_S390_MEMOP_F_CHECK_ONLY
)) {
3635 tmpbuf
= vmalloc(mop
->size
);
3640 srcu_idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
3643 case KVM_S390_MEMOP_LOGICAL_READ
:
3644 if (mop
->flags
& KVM_S390_MEMOP_F_CHECK_ONLY
) {
3645 r
= check_gva_range(vcpu
, mop
->gaddr
, mop
->ar
,
3646 mop
->size
, GACC_FETCH
);
3649 r
= read_guest(vcpu
, mop
->gaddr
, mop
->ar
, tmpbuf
, mop
->size
);
3651 if (copy_to_user(uaddr
, tmpbuf
, mop
->size
))
3655 case KVM_S390_MEMOP_LOGICAL_WRITE
:
3656 if (mop
->flags
& KVM_S390_MEMOP_F_CHECK_ONLY
) {
3657 r
= check_gva_range(vcpu
, mop
->gaddr
, mop
->ar
,
3658 mop
->size
, GACC_STORE
);
3661 if (copy_from_user(tmpbuf
, uaddr
, mop
->size
)) {
3665 r
= write_guest(vcpu
, mop
->gaddr
, mop
->ar
, tmpbuf
, mop
->size
);
3671 srcu_read_unlock(&vcpu
->kvm
->srcu
, srcu_idx
);
3673 if (r
> 0 && (mop
->flags
& KVM_S390_MEMOP_F_INJECT_EXCEPTION
) != 0)
3674 kvm_s390_inject_prog_irq(vcpu
, &vcpu
->arch
.pgm
);
3680 long kvm_arch_vcpu_ioctl(struct file
*filp
,
3681 unsigned int ioctl
, unsigned long arg
)
3683 struct kvm_vcpu
*vcpu
= filp
->private_data
;
3684 void __user
*argp
= (void __user
*)arg
;
3689 case KVM_S390_IRQ
: {
3690 struct kvm_s390_irq s390irq
;
3693 if (copy_from_user(&s390irq
, argp
, sizeof(s390irq
)))
3695 r
= kvm_s390_inject_vcpu(vcpu
, &s390irq
);
3698 case KVM_S390_INTERRUPT
: {
3699 struct kvm_s390_interrupt s390int
;
3700 struct kvm_s390_irq s390irq
;
3703 if (copy_from_user(&s390int
, argp
, sizeof(s390int
)))
3705 if (s390int_to_s390irq(&s390int
, &s390irq
))
3707 r
= kvm_s390_inject_vcpu(vcpu
, &s390irq
);
3710 case KVM_S390_STORE_STATUS
:
3711 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
3712 r
= kvm_s390_vcpu_store_status(vcpu
, arg
);
3713 srcu_read_unlock(&vcpu
->kvm
->srcu
, idx
);
3715 case KVM_S390_SET_INITIAL_PSW
: {
3719 if (copy_from_user(&psw
, argp
, sizeof(psw
)))
3721 r
= kvm_arch_vcpu_ioctl_set_initial_psw(vcpu
, psw
);
3724 case KVM_S390_INITIAL_RESET
:
3725 r
= kvm_arch_vcpu_ioctl_initial_reset(vcpu
);
3727 case KVM_SET_ONE_REG
:
3728 case KVM_GET_ONE_REG
: {
3729 struct kvm_one_reg reg
;
3731 if (copy_from_user(®
, argp
, sizeof(reg
)))
3733 if (ioctl
== KVM_SET_ONE_REG
)
3734 r
= kvm_arch_vcpu_ioctl_set_one_reg(vcpu
, ®
);
3736 r
= kvm_arch_vcpu_ioctl_get_one_reg(vcpu
, ®
);
3739 #ifdef CONFIG_KVM_S390_UCONTROL
3740 case KVM_S390_UCAS_MAP
: {
3741 struct kvm_s390_ucas_mapping ucasmap
;
3743 if (copy_from_user(&ucasmap
, argp
, sizeof(ucasmap
))) {
3748 if (!kvm_is_ucontrol(vcpu
->kvm
)) {
3753 r
= gmap_map_segment(vcpu
->arch
.gmap
, ucasmap
.user_addr
,
3754 ucasmap
.vcpu_addr
, ucasmap
.length
);
3757 case KVM_S390_UCAS_UNMAP
: {
3758 struct kvm_s390_ucas_mapping ucasmap
;
3760 if (copy_from_user(&ucasmap
, argp
, sizeof(ucasmap
))) {
3765 if (!kvm_is_ucontrol(vcpu
->kvm
)) {
3770 r
= gmap_unmap_segment(vcpu
->arch
.gmap
, ucasmap
.vcpu_addr
,
3775 case KVM_S390_VCPU_FAULT
: {
3776 r
= gmap_fault(vcpu
->arch
.gmap
, arg
, 0);
3779 case KVM_ENABLE_CAP
:
3781 struct kvm_enable_cap cap
;
3783 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
3785 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
3788 case KVM_S390_MEM_OP
: {
3789 struct kvm_s390_mem_op mem_op
;
3791 if (copy_from_user(&mem_op
, argp
, sizeof(mem_op
)) == 0)
3792 r
= kvm_s390_guest_mem_op(vcpu
, &mem_op
);
3797 case KVM_S390_SET_IRQ_STATE
: {
3798 struct kvm_s390_irq_state irq_state
;
3801 if (copy_from_user(&irq_state
, argp
, sizeof(irq_state
)))
3803 if (irq_state
.len
> VCPU_IRQS_MAX_BUF
||
3804 irq_state
.len
== 0 ||
3805 irq_state
.len
% sizeof(struct kvm_s390_irq
) > 0) {
3809 /* do not use irq_state.flags, it will break old QEMUs */
3810 r
= kvm_s390_set_irq_state(vcpu
,
3811 (void __user
*) irq_state
.buf
,
3815 case KVM_S390_GET_IRQ_STATE
: {
3816 struct kvm_s390_irq_state irq_state
;
3819 if (copy_from_user(&irq_state
, argp
, sizeof(irq_state
)))
3821 if (irq_state
.len
== 0) {
3825 /* do not use irq_state.flags, it will break old QEMUs */
3826 r
= kvm_s390_get_irq_state(vcpu
,
3827 (__u8 __user
*) irq_state
.buf
,
3837 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
)
3839 #ifdef CONFIG_KVM_S390_UCONTROL
3840 if ((vmf
->pgoff
== KVM_S390_SIE_PAGE_OFFSET
)
3841 && (kvm_is_ucontrol(vcpu
->kvm
))) {
3842 vmf
->page
= virt_to_page(vcpu
->arch
.sie_block
);
3843 get_page(vmf
->page
);
3847 return VM_FAULT_SIGBUS
;
3850 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
3851 unsigned long npages
)
3856 /* Section: memory related */
3857 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
3858 struct kvm_memory_slot
*memslot
,
3859 const struct kvm_userspace_memory_region
*mem
,
3860 enum kvm_mr_change change
)
3862 /* A few sanity checks. We can have memory slots which have to be
3863 located/ended at a segment boundary (1MB). The memory in userland is
3864 ok to be fragmented into various different vmas. It is okay to mmap()
3865 and munmap() stuff in this slot after doing this call at any time */
3867 if (mem
->userspace_addr
& 0xffffful
)
3870 if (mem
->memory_size
& 0xffffful
)
3873 if (mem
->guest_phys_addr
+ mem
->memory_size
> kvm
->arch
.mem_limit
)
3879 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
3880 const struct kvm_userspace_memory_region
*mem
,
3881 const struct kvm_memory_slot
*old
,
3882 const struct kvm_memory_slot
*new,
3883 enum kvm_mr_change change
)
3887 /* If the basics of the memslot do not change, we do not want
3888 * to update the gmap. Every update causes several unnecessary
3889 * segment translation exceptions. This is usually handled just
3890 * fine by the normal fault handler + gmap, but it will also
3891 * cause faults on the prefix page of running guest CPUs.
3893 if (old
->userspace_addr
== mem
->userspace_addr
&&
3894 old
->base_gfn
* PAGE_SIZE
== mem
->guest_phys_addr
&&
3895 old
->npages
* PAGE_SIZE
== mem
->memory_size
)
3898 rc
= gmap_map_segment(kvm
->arch
.gmap
, mem
->userspace_addr
,
3899 mem
->guest_phys_addr
, mem
->memory_size
);
3901 pr_warn("failed to commit memory region\n");
3905 static inline unsigned long nonhyp_mask(int i
)
3907 unsigned int nonhyp_fai
= (sclp
.hmfai
<< i
* 2) >> 30;
3909 return 0x0000ffffffffffffUL
>> (nonhyp_fai
<< 4);
3912 void kvm_arch_vcpu_block_finish(struct kvm_vcpu
*vcpu
)
3914 vcpu
->valid_wakeup
= false;
3917 static int __init
kvm_s390_init(void)
3921 if (!sclp
.has_sief2
) {
3922 pr_info("SIE not available\n");
3926 for (i
= 0; i
< 16; i
++)
3927 kvm_s390_fac_list_mask
[i
] |=
3928 S390_lowcore
.stfle_fac_list
[i
] & nonhyp_mask(i
);
3930 return kvm_init(NULL
, sizeof(struct kvm_vcpu
), 0, THIS_MODULE
);
3933 static void __exit
kvm_s390_exit(void)
3938 module_init(kvm_s390_init
);
3939 module_exit(kvm_s390_exit
);
3942 * Enable autoloading of the kvm module.
3943 * Note that we add the module alias here instead of virt/kvm/kvm_main.c
3944 * since x86 takes a different approach.
3946 #include <linux/miscdevice.h>
3947 MODULE_ALIAS_MISCDEV(KVM_MINOR
);
3948 MODULE_ALIAS("devname:kvm");