]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - arch/s390/kvm/kvm-s390.c
Merge remote-tracking branches 'regulator/topic/of', 'regulator/topic/pv88080', ...
[mirror_ubuntu-hirsute-kernel.git] / arch / s390 / kvm / kvm-s390.c
CommitLineData
b0c632db 1/*
a53c8fab 2 * hosting zSeries kernel virtual machines
b0c632db 3 *
a53c8fab 4 * Copyright IBM Corp. 2008, 2009
b0c632db
HC
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License (version 2 only)
8 * as published by the Free Software Foundation.
9 *
10 * Author(s): Carsten Otte <cotte@de.ibm.com>
11 * Christian Borntraeger <borntraeger@de.ibm.com>
12 * Heiko Carstens <heiko.carstens@de.ibm.com>
628eb9b8 13 * Christian Ehrhardt <ehrhardt@de.ibm.com>
15f36ebd 14 * Jason J. Herne <jjherne@us.ibm.com>
b0c632db
HC
15 */
16
17#include <linux/compiler.h>
18#include <linux/err.h>
19#include <linux/fs.h>
ca872302 20#include <linux/hrtimer.h>
b0c632db
HC
21#include <linux/init.h>
22#include <linux/kvm.h>
23#include <linux/kvm_host.h>
b2d73b2a 24#include <linux/mman.h>
b0c632db 25#include <linux/module.h>
a374e892 26#include <linux/random.h>
b0c632db 27#include <linux/slab.h>
ba5c1e9b 28#include <linux/timer.h>
41408c28 29#include <linux/vmalloc.h>
15c9705f 30#include <linux/bitmap.h>
cbb870c8 31#include <asm/asm-offsets.h>
b0c632db 32#include <asm/lowcore.h>
fd5ada04 33#include <asm/stp.h>
b0c632db 34#include <asm/pgtable.h>
1e133ab2 35#include <asm/gmap.h>
f5daba1d 36#include <asm/nmi.h>
a0616cde 37#include <asm/switch_to.h>
6d3da241 38#include <asm/isc.h>
1526bf9c 39#include <asm/sclp.h>
0a763c78 40#include <asm/cpacf.h>
221bb8a4 41#include <asm/timex.h>
8f2abe6a 42#include "kvm-s390.h"
b0c632db
HC
43#include "gaccess.h"
44
ea2cdd27
DH
45#define KMSG_COMPONENT "kvm-s390"
46#undef pr_fmt
47#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
48
5786fffa
CH
49#define CREATE_TRACE_POINTS
50#include "trace.h"
ade38c31 51#include "trace-s390.h"
5786fffa 52
41408c28 53#define MEM_OP_MAX_SIZE 65536 /* Maximum transfer size for KVM_S390_MEM_OP */
816c7667
JF
54#define LOCAL_IRQS 32
55#define VCPU_IRQS_MAX_BUF (sizeof(struct kvm_s390_irq) * \
56 (KVM_MAX_VCPUS + LOCAL_IRQS))
41408c28 57
b0c632db
HC
58#define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
59
60struct kvm_stats_debugfs_item debugfs_entries[] = {
61 { "userspace_handled", VCPU_STAT(exit_userspace) },
0eaeafa1 62 { "exit_null", VCPU_STAT(exit_null) },
8f2abe6a
CB
63 { "exit_validity", VCPU_STAT(exit_validity) },
64 { "exit_stop_request", VCPU_STAT(exit_stop_request) },
65 { "exit_external_request", VCPU_STAT(exit_external_request) },
66 { "exit_external_interrupt", VCPU_STAT(exit_external_interrupt) },
ba5c1e9b 67 { "exit_instruction", VCPU_STAT(exit_instruction) },
9ec6de19 68 { "exit_pei", VCPU_STAT(exit_pei) },
ba5c1e9b
CO
69 { "exit_program_interruption", VCPU_STAT(exit_program_interruption) },
70 { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) },
a011eeb2 71 { "exit_operation_exception", VCPU_STAT(exit_operation_exception) },
f7819512 72 { "halt_successful_poll", VCPU_STAT(halt_successful_poll) },
62bea5bf 73 { "halt_attempted_poll", VCPU_STAT(halt_attempted_poll) },
3491caf2 74 { "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
ce2e4f0b 75 { "halt_wakeup", VCPU_STAT(halt_wakeup) },
f5e10b09 76 { "instruction_lctlg", VCPU_STAT(instruction_lctlg) },
ba5c1e9b 77 { "instruction_lctl", VCPU_STAT(instruction_lctl) },
aba07508
DH
78 { "instruction_stctl", VCPU_STAT(instruction_stctl) },
79 { "instruction_stctg", VCPU_STAT(instruction_stctg) },
ba5c1e9b 80 { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) },
7697e71f 81 { "deliver_external_call", VCPU_STAT(deliver_external_call) },
ba5c1e9b
CO
82 { "deliver_service_signal", VCPU_STAT(deliver_service_signal) },
83 { "deliver_virtio_interrupt", VCPU_STAT(deliver_virtio_interrupt) },
84 { "deliver_stop_signal", VCPU_STAT(deliver_stop_signal) },
85 { "deliver_prefix_signal", VCPU_STAT(deliver_prefix_signal) },
86 { "deliver_restart_signal", VCPU_STAT(deliver_restart_signal) },
87 { "deliver_program_interruption", VCPU_STAT(deliver_program_int) },
88 { "exit_wait_state", VCPU_STAT(exit_wait_state) },
69d0d3a3 89 { "instruction_pfmf", VCPU_STAT(instruction_pfmf) },
453423dc
CB
90 { "instruction_stidp", VCPU_STAT(instruction_stidp) },
91 { "instruction_spx", VCPU_STAT(instruction_spx) },
92 { "instruction_stpx", VCPU_STAT(instruction_stpx) },
93 { "instruction_stap", VCPU_STAT(instruction_stap) },
94 { "instruction_storage_key", VCPU_STAT(instruction_storage_key) },
8a242234 95 { "instruction_ipte_interlock", VCPU_STAT(instruction_ipte_interlock) },
453423dc
CB
96 { "instruction_stsch", VCPU_STAT(instruction_stsch) },
97 { "instruction_chsc", VCPU_STAT(instruction_chsc) },
b31288fa 98 { "instruction_essa", VCPU_STAT(instruction_essa) },
453423dc
CB
99 { "instruction_stsi", VCPU_STAT(instruction_stsi) },
100 { "instruction_stfl", VCPU_STAT(instruction_stfl) },
bb25b9ba 101 { "instruction_tprot", VCPU_STAT(instruction_tprot) },
95ca2cb5 102 { "instruction_sthyi", VCPU_STAT(instruction_sthyi) },
a3508fbe 103 { "instruction_sie", VCPU_STAT(instruction_sie) },
5288fbf0 104 { "instruction_sigp_sense", VCPU_STAT(instruction_sigp_sense) },
bd59d3a4 105 { "instruction_sigp_sense_running", VCPU_STAT(instruction_sigp_sense_running) },
7697e71f 106 { "instruction_sigp_external_call", VCPU_STAT(instruction_sigp_external_call) },
5288fbf0 107 { "instruction_sigp_emergency", VCPU_STAT(instruction_sigp_emergency) },
42cb0c9f
DH
108 { "instruction_sigp_cond_emergency", VCPU_STAT(instruction_sigp_cond_emergency) },
109 { "instruction_sigp_start", VCPU_STAT(instruction_sigp_start) },
5288fbf0 110 { "instruction_sigp_stop", VCPU_STAT(instruction_sigp_stop) },
42cb0c9f
DH
111 { "instruction_sigp_stop_store_status", VCPU_STAT(instruction_sigp_stop_store_status) },
112 { "instruction_sigp_store_status", VCPU_STAT(instruction_sigp_store_status) },
cd7b4b61 113 { "instruction_sigp_store_adtl_status", VCPU_STAT(instruction_sigp_store_adtl_status) },
5288fbf0
CB
114 { "instruction_sigp_set_arch", VCPU_STAT(instruction_sigp_arch) },
115 { "instruction_sigp_set_prefix", VCPU_STAT(instruction_sigp_prefix) },
116 { "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart) },
42cb0c9f
DH
117 { "instruction_sigp_cpu_reset", VCPU_STAT(instruction_sigp_cpu_reset) },
118 { "instruction_sigp_init_cpu_reset", VCPU_STAT(instruction_sigp_init_cpu_reset) },
119 { "instruction_sigp_unknown", VCPU_STAT(instruction_sigp_unknown) },
388186bc 120 { "diagnose_10", VCPU_STAT(diagnose_10) },
e28acfea 121 { "diagnose_44", VCPU_STAT(diagnose_44) },
41628d33 122 { "diagnose_9c", VCPU_STAT(diagnose_9c) },
175a5c9e
CB
123 { "diagnose_258", VCPU_STAT(diagnose_258) },
124 { "diagnose_308", VCPU_STAT(diagnose_308) },
125 { "diagnose_500", VCPU_STAT(diagnose_500) },
b0c632db
HC
126 { NULL }
127};
128
a411edf1
DH
129/* allow nested virtualization in KVM (if enabled by user space) */
130static int nested;
131module_param(nested, int, S_IRUGO);
132MODULE_PARM_DESC(nested, "Nested virtualization support");
133
9d8d5786 134/* upper facilities limit for kvm */
60a37709
AY
135unsigned long kvm_s390_fac_list_mask[16] = {
136 0xffe6000000000000UL,
137 0x005e000000000000UL,
9d8d5786 138};
b0c632db 139
9d8d5786 140unsigned long kvm_s390_fac_list_mask_size(void)
78c4b59f 141{
9d8d5786
MM
142 BUILD_BUG_ON(ARRAY_SIZE(kvm_s390_fac_list_mask) > S390_ARCH_FAC_MASK_SIZE_U64);
143 return ARRAY_SIZE(kvm_s390_fac_list_mask);
78c4b59f
MM
144}
145
15c9705f
DH
146/* available cpu features supported by kvm */
147static DECLARE_BITMAP(kvm_s390_available_cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
0a763c78
DH
148/* available subfunctions indicated via query / "test bit" */
149static struct kvm_s390_vm_cpu_subfunc kvm_s390_available_subfunc;
15c9705f 150
9d8d5786 151static struct gmap_notifier gmap_notifier;
a3508fbe 152static struct gmap_notifier vsie_gmap_notifier;
78f26131 153debug_info_t *kvm_s390_dbf;
9d8d5786 154
b0c632db 155/* Section: not file related */
13a34e06 156int kvm_arch_hardware_enable(void)
b0c632db
HC
157{
158 /* every s390 is virtualization enabled ;-) */
10474ae8 159 return 0;
b0c632db
HC
160}
161
414d3b07
MS
162static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
163 unsigned long end);
2c70fe44 164
fdf03650
FZ
165/*
166 * This callback is executed during stop_machine(). All CPUs are therefore
167 * temporarily stopped. In order not to change guest behavior, we have to
168 * disable preemption whenever we touch the epoch of kvm and the VCPUs,
169 * so a CPU won't be stopped while calculating with the epoch.
170 */
171static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
172 void *v)
173{
174 struct kvm *kvm;
175 struct kvm_vcpu *vcpu;
176 int i;
177 unsigned long long *delta = v;
178
179 list_for_each_entry(kvm, &vm_list, vm_list) {
180 kvm->arch.epoch -= *delta;
181 kvm_for_each_vcpu(i, vcpu, kvm) {
182 vcpu->arch.sie_block->epoch -= *delta;
db0758b2
DH
183 if (vcpu->arch.cputm_enabled)
184 vcpu->arch.cputm_start += *delta;
91473b48
DH
185 if (vcpu->arch.vsie_block)
186 vcpu->arch.vsie_block->epoch -= *delta;
fdf03650
FZ
187 }
188 }
189 return NOTIFY_OK;
190}
191
192static struct notifier_block kvm_clock_notifier = {
193 .notifier_call = kvm_clock_sync,
194};
195
b0c632db
HC
196int kvm_arch_hardware_setup(void)
197{
2c70fe44 198 gmap_notifier.notifier_call = kvm_gmap_notifier;
b2d73b2a 199 gmap_register_pte_notifier(&gmap_notifier);
a3508fbe
DH
200 vsie_gmap_notifier.notifier_call = kvm_s390_vsie_gmap_notifier;
201 gmap_register_pte_notifier(&vsie_gmap_notifier);
fdf03650
FZ
202 atomic_notifier_chain_register(&s390_epoch_delta_notifier,
203 &kvm_clock_notifier);
b0c632db
HC
204 return 0;
205}
206
207void kvm_arch_hardware_unsetup(void)
208{
b2d73b2a 209 gmap_unregister_pte_notifier(&gmap_notifier);
a3508fbe 210 gmap_unregister_pte_notifier(&vsie_gmap_notifier);
fdf03650
FZ
211 atomic_notifier_chain_unregister(&s390_epoch_delta_notifier,
212 &kvm_clock_notifier);
b0c632db
HC
213}
214
22be5a13
DH
215static void allow_cpu_feat(unsigned long nr)
216{
217 set_bit_inv(nr, kvm_s390_available_cpu_feat);
218}
219
0a763c78
DH
220static inline int plo_test_bit(unsigned char nr)
221{
222 register unsigned long r0 asm("0") = (unsigned long) nr | 0x100;
223 int cc = 3; /* subfunction not available */
224
225 asm volatile(
226 /* Parameter registers are ignored for "test bit" */
227 " plo 0,0,0,0(0)\n"
228 " ipm %0\n"
229 " srl %0,28\n"
230 : "=d" (cc)
231 : "d" (r0)
232 : "cc");
233 return cc == 0;
234}
235
22be5a13
DH
236static void kvm_s390_cpu_feat_init(void)
237{
0a763c78
DH
238 int i;
239
240 for (i = 0; i < 256; ++i) {
241 if (plo_test_bit(i))
242 kvm_s390_available_subfunc.plo[i >> 3] |= 0x80 >> (i & 7);
243 }
244
245 if (test_facility(28)) /* TOD-clock steering */
221bb8a4
LT
246 ptff(kvm_s390_available_subfunc.ptff,
247 sizeof(kvm_s390_available_subfunc.ptff),
248 PTFF_QAF);
0a763c78
DH
249
250 if (test_facility(17)) { /* MSA */
251 __cpacf_query(CPACF_KMAC, kvm_s390_available_subfunc.kmac);
252 __cpacf_query(CPACF_KMC, kvm_s390_available_subfunc.kmc);
253 __cpacf_query(CPACF_KM, kvm_s390_available_subfunc.km);
254 __cpacf_query(CPACF_KIMD, kvm_s390_available_subfunc.kimd);
255 __cpacf_query(CPACF_KLMD, kvm_s390_available_subfunc.klmd);
256 }
257 if (test_facility(76)) /* MSA3 */
258 __cpacf_query(CPACF_PCKMO, kvm_s390_available_subfunc.pckmo);
259 if (test_facility(77)) { /* MSA4 */
260 __cpacf_query(CPACF_KMCTR, kvm_s390_available_subfunc.kmctr);
261 __cpacf_query(CPACF_KMF, kvm_s390_available_subfunc.kmf);
262 __cpacf_query(CPACF_KMO, kvm_s390_available_subfunc.kmo);
263 __cpacf_query(CPACF_PCC, kvm_s390_available_subfunc.pcc);
264 }
265 if (test_facility(57)) /* MSA5 */
266 __cpacf_query(CPACF_PPNO, kvm_s390_available_subfunc.ppno);
267
22be5a13
DH
268 if (MACHINE_HAS_ESOP)
269 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP);
a3508fbe
DH
270 /*
271 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
272 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
273 */
274 if (!sclp.has_sief2 || !MACHINE_HAS_ESOP || !sclp.has_64bscao ||
a411edf1 275 !test_facility(3) || !nested)
a3508fbe
DH
276 return;
277 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2);
19c439b5
DH
278 if (sclp.has_64bscao)
279 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO);
0615a326
DH
280 if (sclp.has_siif)
281 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF);
77d18f6d
DH
282 if (sclp.has_gpere)
283 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE);
a1b7b9b2
DH
284 if (sclp.has_gsls)
285 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS);
5630a8e8
DH
286 if (sclp.has_ib)
287 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB);
13ee3f67
DH
288 if (sclp.has_cei)
289 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI);
7fd7f39d
DH
290 if (sclp.has_ibs)
291 allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS);
5d3876a8
DH
292 /*
293 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
294 * all skey handling functions read/set the skey from the PGSTE
295 * instead of the real storage key.
296 *
297 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
298 * pages being detected as preserved although they are resident.
299 *
300 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
301 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
302 *
303 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
304 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
305 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
306 *
307 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
308 * cannot easily shadow the SCA because of the ipte lock.
309 */
22be5a13
DH
310}
311
b0c632db
HC
312int kvm_arch_init(void *opaque)
313{
78f26131
CB
314 kvm_s390_dbf = debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
315 if (!kvm_s390_dbf)
316 return -ENOMEM;
317
318 if (debug_register_view(kvm_s390_dbf, &debug_sprintf_view)) {
319 debug_unregister(kvm_s390_dbf);
320 return -ENOMEM;
321 }
322
22be5a13
DH
323 kvm_s390_cpu_feat_init();
324
84877d93
CH
325 /* Register floating interrupt controller interface. */
326 return kvm_register_device_ops(&kvm_flic_ops, KVM_DEV_TYPE_FLIC);
b0c632db
HC
327}
328
78f26131
CB
329void kvm_arch_exit(void)
330{
331 debug_unregister(kvm_s390_dbf);
332}
333
b0c632db
HC
334/* Section: device related */
335long kvm_arch_dev_ioctl(struct file *filp,
336 unsigned int ioctl, unsigned long arg)
337{
338 if (ioctl == KVM_S390_ENABLE_SIE)
339 return s390_enable_sie();
340 return -EINVAL;
341}
342
784aa3d7 343int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
b0c632db 344{
d7b0b5eb
CO
345 int r;
346
2bd0ac4e 347 switch (ext) {
d7b0b5eb 348 case KVM_CAP_S390_PSW:
b6cf8788 349 case KVM_CAP_S390_GMAP:
52e16b18 350 case KVM_CAP_SYNC_MMU:
1efd0f59
CO
351#ifdef CONFIG_KVM_S390_UCONTROL
352 case KVM_CAP_S390_UCONTROL:
353#endif
3c038e6b 354 case KVM_CAP_ASYNC_PF:
60b413c9 355 case KVM_CAP_SYNC_REGS:
14eebd91 356 case KVM_CAP_ONE_REG:
d6712df9 357 case KVM_CAP_ENABLE_CAP:
fa6b7fe9 358 case KVM_CAP_S390_CSS_SUPPORT:
10ccaa1e 359 case KVM_CAP_IOEVENTFD:
c05c4186 360 case KVM_CAP_DEVICE_CTRL:
d938dc55 361 case KVM_CAP_ENABLE_CAP_VM:
78599d90 362 case KVM_CAP_S390_IRQCHIP:
f2061656 363 case KVM_CAP_VM_ATTRIBUTES:
6352e4d2 364 case KVM_CAP_MP_STATE:
47b43c52 365 case KVM_CAP_S390_INJECT_IRQ:
2444b352 366 case KVM_CAP_S390_USER_SIGP:
e44fc8c9 367 case KVM_CAP_S390_USER_STSI:
30ee2a98 368 case KVM_CAP_S390_SKEYS:
816c7667 369 case KVM_CAP_S390_IRQ_STATE:
6502a34c 370 case KVM_CAP_S390_USER_INSTR0:
d7b0b5eb
CO
371 r = 1;
372 break;
41408c28
TH
373 case KVM_CAP_S390_MEM_OP:
374 r = MEM_OP_MAX_SIZE;
375 break;
e726b1bd
CB
376 case KVM_CAP_NR_VCPUS:
377 case KVM_CAP_MAX_VCPUS:
76a6dd72
DH
378 r = KVM_S390_BSCA_CPU_SLOTS;
379 if (sclp.has_esca && sclp.has_64bscao)
380 r = KVM_S390_ESCA_CPU_SLOTS;
e726b1bd 381 break;
e1e2e605
NW
382 case KVM_CAP_NR_MEMSLOTS:
383 r = KVM_USER_MEM_SLOTS;
384 break;
1526bf9c 385 case KVM_CAP_S390_COW:
abf09bed 386 r = MACHINE_HAS_ESOP;
1526bf9c 387 break;
68c55750
EF
388 case KVM_CAP_S390_VECTOR_REGISTERS:
389 r = MACHINE_HAS_VX;
390 break;
c6e5f166
FZ
391 case KVM_CAP_S390_RI:
392 r = test_facility(64);
393 break;
2bd0ac4e 394 default:
d7b0b5eb 395 r = 0;
2bd0ac4e 396 }
d7b0b5eb 397 return r;
b0c632db
HC
398}
399
15f36ebd
JH
400static void kvm_s390_sync_dirty_log(struct kvm *kvm,
401 struct kvm_memory_slot *memslot)
402{
403 gfn_t cur_gfn, last_gfn;
404 unsigned long address;
405 struct gmap *gmap = kvm->arch.gmap;
406
15f36ebd
JH
407 /* Loop over all guest pages */
408 last_gfn = memslot->base_gfn + memslot->npages;
409 for (cur_gfn = memslot->base_gfn; cur_gfn <= last_gfn; cur_gfn++) {
410 address = gfn_to_hva_memslot(memslot, cur_gfn);
411
1e133ab2 412 if (test_and_clear_guest_dirty(gmap->mm, address))
15f36ebd 413 mark_page_dirty(kvm, cur_gfn);
1763f8d0
CB
414 if (fatal_signal_pending(current))
415 return;
70c88a00 416 cond_resched();
15f36ebd 417 }
15f36ebd
JH
418}
419
b0c632db 420/* Section: vm related */
a6e2f683
ED
421static void sca_del_vcpu(struct kvm_vcpu *vcpu);
422
b0c632db
HC
423/*
424 * Get (and clear) the dirty memory log for a memory slot.
425 */
426int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
427 struct kvm_dirty_log *log)
428{
15f36ebd
JH
429 int r;
430 unsigned long n;
9f6b8029 431 struct kvm_memslots *slots;
15f36ebd
JH
432 struct kvm_memory_slot *memslot;
433 int is_dirty = 0;
434
435 mutex_lock(&kvm->slots_lock);
436
437 r = -EINVAL;
438 if (log->slot >= KVM_USER_MEM_SLOTS)
439 goto out;
440
9f6b8029
PB
441 slots = kvm_memslots(kvm);
442 memslot = id_to_memslot(slots, log->slot);
15f36ebd
JH
443 r = -ENOENT;
444 if (!memslot->dirty_bitmap)
445 goto out;
446
447 kvm_s390_sync_dirty_log(kvm, memslot);
448 r = kvm_get_dirty_log(kvm, log, &is_dirty);
449 if (r)
450 goto out;
451
452 /* Clear the dirty log */
453 if (is_dirty) {
454 n = kvm_dirty_bitmap_bytes(memslot);
455 memset(memslot->dirty_bitmap, 0, n);
456 }
457 r = 0;
458out:
459 mutex_unlock(&kvm->slots_lock);
460 return r;
b0c632db
HC
461}
462
6502a34c
DH
463static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
464{
465 unsigned int i;
466 struct kvm_vcpu *vcpu;
467
468 kvm_for_each_vcpu(i, vcpu, kvm) {
469 kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
470 }
471}
472
d938dc55
CH
473static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
474{
475 int r;
476
477 if (cap->flags)
478 return -EINVAL;
479
480 switch (cap->cap) {
84223598 481 case KVM_CAP_S390_IRQCHIP:
c92ea7b9 482 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
84223598
CH
483 kvm->arch.use_irqchip = 1;
484 r = 0;
485 break;
2444b352 486 case KVM_CAP_S390_USER_SIGP:
c92ea7b9 487 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
2444b352
DH
488 kvm->arch.user_sigp = 1;
489 r = 0;
490 break;
68c55750 491 case KVM_CAP_S390_VECTOR_REGISTERS:
5967c17b 492 mutex_lock(&kvm->lock);
a03825bb 493 if (kvm->created_vcpus) {
5967c17b
DH
494 r = -EBUSY;
495 } else if (MACHINE_HAS_VX) {
c54f0d6a
DH
496 set_kvm_facility(kvm->arch.model.fac_mask, 129);
497 set_kvm_facility(kvm->arch.model.fac_list, 129);
18280d8b
MM
498 r = 0;
499 } else
500 r = -EINVAL;
5967c17b 501 mutex_unlock(&kvm->lock);
c92ea7b9
CB
502 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
503 r ? "(not available)" : "(success)");
68c55750 504 break;
c6e5f166
FZ
505 case KVM_CAP_S390_RI:
506 r = -EINVAL;
507 mutex_lock(&kvm->lock);
a03825bb 508 if (kvm->created_vcpus) {
c6e5f166
FZ
509 r = -EBUSY;
510 } else if (test_facility(64)) {
c54f0d6a
DH
511 set_kvm_facility(kvm->arch.model.fac_mask, 64);
512 set_kvm_facility(kvm->arch.model.fac_list, 64);
c6e5f166
FZ
513 r = 0;
514 }
515 mutex_unlock(&kvm->lock);
516 VM_EVENT(kvm, 3, "ENABLE: CAP_S390_RI %s",
517 r ? "(not available)" : "(success)");
518 break;
e44fc8c9 519 case KVM_CAP_S390_USER_STSI:
c92ea7b9 520 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
e44fc8c9
ET
521 kvm->arch.user_stsi = 1;
522 r = 0;
523 break;
6502a34c
DH
524 case KVM_CAP_S390_USER_INSTR0:
525 VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
526 kvm->arch.user_instr0 = 1;
527 icpt_operexc_on_all_vcpus(kvm);
528 r = 0;
529 break;
d938dc55
CH
530 default:
531 r = -EINVAL;
532 break;
533 }
534 return r;
535}
536
8c0a7ce6
DD
537static int kvm_s390_get_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
538{
539 int ret;
540
541 switch (attr->attr) {
542 case KVM_S390_VM_MEM_LIMIT_SIZE:
543 ret = 0;
c92ea7b9 544 VM_EVENT(kvm, 3, "QUERY: max guest memory: %lu bytes",
a3a92c31
DD
545 kvm->arch.mem_limit);
546 if (put_user(kvm->arch.mem_limit, (u64 __user *)attr->addr))
8c0a7ce6
DD
547 ret = -EFAULT;
548 break;
549 default:
550 ret = -ENXIO;
551 break;
552 }
553 return ret;
554}
555
556static int kvm_s390_set_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
4f718eab
DD
557{
558 int ret;
559 unsigned int idx;
560 switch (attr->attr) {
561 case KVM_S390_VM_MEM_ENABLE_CMMA:
f9cbd9b0 562 ret = -ENXIO;
c24cc9c8 563 if (!sclp.has_cmma)
e6db1d61
DD
564 break;
565
4f718eab 566 ret = -EBUSY;
c92ea7b9 567 VM_EVENT(kvm, 3, "%s", "ENABLE: CMMA support");
4f718eab 568 mutex_lock(&kvm->lock);
a03825bb 569 if (!kvm->created_vcpus) {
4f718eab
DD
570 kvm->arch.use_cmma = 1;
571 ret = 0;
572 }
573 mutex_unlock(&kvm->lock);
574 break;
575 case KVM_S390_VM_MEM_CLR_CMMA:
f9cbd9b0
DH
576 ret = -ENXIO;
577 if (!sclp.has_cmma)
578 break;
c3489155
DD
579 ret = -EINVAL;
580 if (!kvm->arch.use_cmma)
581 break;
582
c92ea7b9 583 VM_EVENT(kvm, 3, "%s", "RESET: CMMA states");
4f718eab
DD
584 mutex_lock(&kvm->lock);
585 idx = srcu_read_lock(&kvm->srcu);
a13cff31 586 s390_reset_cmma(kvm->arch.gmap->mm);
4f718eab
DD
587 srcu_read_unlock(&kvm->srcu, idx);
588 mutex_unlock(&kvm->lock);
589 ret = 0;
590 break;
8c0a7ce6
DD
591 case KVM_S390_VM_MEM_LIMIT_SIZE: {
592 unsigned long new_limit;
593
594 if (kvm_is_ucontrol(kvm))
595 return -EINVAL;
596
597 if (get_user(new_limit, (u64 __user *)attr->addr))
598 return -EFAULT;
599
a3a92c31
DD
600 if (kvm->arch.mem_limit != KVM_S390_NO_MEM_LIMIT &&
601 new_limit > kvm->arch.mem_limit)
8c0a7ce6
DD
602 return -E2BIG;
603
a3a92c31
DD
604 if (!new_limit)
605 return -EINVAL;
606
6ea427bb 607 /* gmap_create takes last usable address */
a3a92c31
DD
608 if (new_limit != KVM_S390_NO_MEM_LIMIT)
609 new_limit -= 1;
610
8c0a7ce6
DD
611 ret = -EBUSY;
612 mutex_lock(&kvm->lock);
a03825bb 613 if (!kvm->created_vcpus) {
6ea427bb
MS
614 /* gmap_create will round the limit up */
615 struct gmap *new = gmap_create(current->mm, new_limit);
8c0a7ce6
DD
616
617 if (!new) {
618 ret = -ENOMEM;
619 } else {
6ea427bb 620 gmap_remove(kvm->arch.gmap);
8c0a7ce6
DD
621 new->private = kvm;
622 kvm->arch.gmap = new;
623 ret = 0;
624 }
625 }
626 mutex_unlock(&kvm->lock);
a3a92c31
DD
627 VM_EVENT(kvm, 3, "SET: max guest address: %lu", new_limit);
628 VM_EVENT(kvm, 3, "New guest asce: 0x%pK",
629 (void *) kvm->arch.gmap->asce);
8c0a7ce6
DD
630 break;
631 }
4f718eab
DD
632 default:
633 ret = -ENXIO;
634 break;
635 }
636 return ret;
637}
638
a374e892
TK
639static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
640
641static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
642{
643 struct kvm_vcpu *vcpu;
644 int i;
645
9d8d5786 646 if (!test_kvm_facility(kvm, 76))
a374e892
TK
647 return -EINVAL;
648
649 mutex_lock(&kvm->lock);
650 switch (attr->attr) {
651 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
652 get_random_bytes(
653 kvm->arch.crypto.crycb->aes_wrapping_key_mask,
654 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
655 kvm->arch.crypto.aes_kw = 1;
c92ea7b9 656 VM_EVENT(kvm, 3, "%s", "ENABLE: AES keywrapping support");
a374e892
TK
657 break;
658 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
659 get_random_bytes(
660 kvm->arch.crypto.crycb->dea_wrapping_key_mask,
661 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
662 kvm->arch.crypto.dea_kw = 1;
c92ea7b9 663 VM_EVENT(kvm, 3, "%s", "ENABLE: DEA keywrapping support");
a374e892
TK
664 break;
665 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
666 kvm->arch.crypto.aes_kw = 0;
667 memset(kvm->arch.crypto.crycb->aes_wrapping_key_mask, 0,
668 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
c92ea7b9 669 VM_EVENT(kvm, 3, "%s", "DISABLE: AES keywrapping support");
a374e892
TK
670 break;
671 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
672 kvm->arch.crypto.dea_kw = 0;
673 memset(kvm->arch.crypto.crycb->dea_wrapping_key_mask, 0,
674 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
c92ea7b9 675 VM_EVENT(kvm, 3, "%s", "DISABLE: DEA keywrapping support");
a374e892
TK
676 break;
677 default:
678 mutex_unlock(&kvm->lock);
679 return -ENXIO;
680 }
681
682 kvm_for_each_vcpu(i, vcpu, kvm) {
683 kvm_s390_vcpu_crypto_setup(vcpu);
684 exit_sie(vcpu);
685 }
686 mutex_unlock(&kvm->lock);
687 return 0;
688}
689
72f25020
JH
690static int kvm_s390_set_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
691{
692 u8 gtod_high;
693
694 if (copy_from_user(&gtod_high, (void __user *)attr->addr,
695 sizeof(gtod_high)))
696 return -EFAULT;
697
698 if (gtod_high != 0)
699 return -EINVAL;
58c383c6 700 VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x", gtod_high);
72f25020
JH
701
702 return 0;
703}
704
705static int kvm_s390_set_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
706{
5a3d883a 707 u64 gtod;
72f25020
JH
708
709 if (copy_from_user(&gtod, (void __user *)attr->addr, sizeof(gtod)))
710 return -EFAULT;
711
25ed1675 712 kvm_s390_set_tod_clock(kvm, gtod);
58c383c6 713 VM_EVENT(kvm, 3, "SET: TOD base: 0x%llx", gtod);
72f25020
JH
714 return 0;
715}
716
717static int kvm_s390_set_tod(struct kvm *kvm, struct kvm_device_attr *attr)
718{
719 int ret;
720
721 if (attr->flags)
722 return -EINVAL;
723
724 switch (attr->attr) {
725 case KVM_S390_VM_TOD_HIGH:
726 ret = kvm_s390_set_tod_high(kvm, attr);
727 break;
728 case KVM_S390_VM_TOD_LOW:
729 ret = kvm_s390_set_tod_low(kvm, attr);
730 break;
731 default:
732 ret = -ENXIO;
733 break;
734 }
735 return ret;
736}
737
738static int kvm_s390_get_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
739{
740 u8 gtod_high = 0;
741
742 if (copy_to_user((void __user *)attr->addr, &gtod_high,
743 sizeof(gtod_high)))
744 return -EFAULT;
58c383c6 745 VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x", gtod_high);
72f25020
JH
746
747 return 0;
748}
749
750static int kvm_s390_get_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
751{
5a3d883a 752 u64 gtod;
72f25020 753
60417fcc 754 gtod = kvm_s390_get_tod_clock_fast(kvm);
72f25020
JH
755 if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
756 return -EFAULT;
58c383c6 757 VM_EVENT(kvm, 3, "QUERY: TOD base: 0x%llx", gtod);
72f25020
JH
758
759 return 0;
760}
761
762static int kvm_s390_get_tod(struct kvm *kvm, struct kvm_device_attr *attr)
763{
764 int ret;
765
766 if (attr->flags)
767 return -EINVAL;
768
769 switch (attr->attr) {
770 case KVM_S390_VM_TOD_HIGH:
771 ret = kvm_s390_get_tod_high(kvm, attr);
772 break;
773 case KVM_S390_VM_TOD_LOW:
774 ret = kvm_s390_get_tod_low(kvm, attr);
775 break;
776 default:
777 ret = -ENXIO;
778 break;
779 }
780 return ret;
781}
782
658b6eda
MM
783static int kvm_s390_set_processor(struct kvm *kvm, struct kvm_device_attr *attr)
784{
785 struct kvm_s390_vm_cpu_processor *proc;
053dd230 786 u16 lowest_ibc, unblocked_ibc;
658b6eda
MM
787 int ret = 0;
788
789 mutex_lock(&kvm->lock);
a03825bb 790 if (kvm->created_vcpus) {
658b6eda
MM
791 ret = -EBUSY;
792 goto out;
793 }
794 proc = kzalloc(sizeof(*proc), GFP_KERNEL);
795 if (!proc) {
796 ret = -ENOMEM;
797 goto out;
798 }
799 if (!copy_from_user(proc, (void __user *)attr->addr,
800 sizeof(*proc))) {
9bb0ec09 801 kvm->arch.model.cpuid = proc->cpuid;
053dd230
DH
802 lowest_ibc = sclp.ibc >> 16 & 0xfff;
803 unblocked_ibc = sclp.ibc & 0xfff;
0487c44d 804 if (lowest_ibc && proc->ibc) {
053dd230
DH
805 if (proc->ibc > unblocked_ibc)
806 kvm->arch.model.ibc = unblocked_ibc;
807 else if (proc->ibc < lowest_ibc)
808 kvm->arch.model.ibc = lowest_ibc;
809 else
810 kvm->arch.model.ibc = proc->ibc;
811 }
c54f0d6a 812 memcpy(kvm->arch.model.fac_list, proc->fac_list,
658b6eda
MM
813 S390_ARCH_FAC_LIST_SIZE_BYTE);
814 } else
815 ret = -EFAULT;
816 kfree(proc);
817out:
818 mutex_unlock(&kvm->lock);
819 return ret;
820}
821
15c9705f
DH
822static int kvm_s390_set_processor_feat(struct kvm *kvm,
823 struct kvm_device_attr *attr)
824{
825 struct kvm_s390_vm_cpu_feat data;
826 int ret = -EBUSY;
827
828 if (copy_from_user(&data, (void __user *)attr->addr, sizeof(data)))
829 return -EFAULT;
830 if (!bitmap_subset((unsigned long *) data.feat,
831 kvm_s390_available_cpu_feat,
832 KVM_S390_VM_CPU_FEAT_NR_BITS))
833 return -EINVAL;
834
835 mutex_lock(&kvm->lock);
836 if (!atomic_read(&kvm->online_vcpus)) {
837 bitmap_copy(kvm->arch.cpu_feat, (unsigned long *) data.feat,
838 KVM_S390_VM_CPU_FEAT_NR_BITS);
839 ret = 0;
840 }
841 mutex_unlock(&kvm->lock);
842 return ret;
843}
844
0a763c78
DH
845static int kvm_s390_set_processor_subfunc(struct kvm *kvm,
846 struct kvm_device_attr *attr)
847{
848 /*
849 * Once supported by kernel + hw, we have to store the subfunctions
850 * in kvm->arch and remember that user space configured them.
851 */
852 return -ENXIO;
853}
854
658b6eda
MM
855static int kvm_s390_set_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
856{
857 int ret = -ENXIO;
858
859 switch (attr->attr) {
860 case KVM_S390_VM_CPU_PROCESSOR:
861 ret = kvm_s390_set_processor(kvm, attr);
862 break;
15c9705f
DH
863 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
864 ret = kvm_s390_set_processor_feat(kvm, attr);
865 break;
0a763c78
DH
866 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
867 ret = kvm_s390_set_processor_subfunc(kvm, attr);
868 break;
658b6eda
MM
869 }
870 return ret;
871}
872
873static int kvm_s390_get_processor(struct kvm *kvm, struct kvm_device_attr *attr)
874{
875 struct kvm_s390_vm_cpu_processor *proc;
876 int ret = 0;
877
878 proc = kzalloc(sizeof(*proc), GFP_KERNEL);
879 if (!proc) {
880 ret = -ENOMEM;
881 goto out;
882 }
9bb0ec09 883 proc->cpuid = kvm->arch.model.cpuid;
658b6eda 884 proc->ibc = kvm->arch.model.ibc;
c54f0d6a
DH
885 memcpy(&proc->fac_list, kvm->arch.model.fac_list,
886 S390_ARCH_FAC_LIST_SIZE_BYTE);
658b6eda
MM
887 if (copy_to_user((void __user *)attr->addr, proc, sizeof(*proc)))
888 ret = -EFAULT;
889 kfree(proc);
890out:
891 return ret;
892}
893
894static int kvm_s390_get_machine(struct kvm *kvm, struct kvm_device_attr *attr)
895{
896 struct kvm_s390_vm_cpu_machine *mach;
897 int ret = 0;
898
899 mach = kzalloc(sizeof(*mach), GFP_KERNEL);
900 if (!mach) {
901 ret = -ENOMEM;
902 goto out;
903 }
904 get_cpu_id((struct cpuid *) &mach->cpuid);
37c5f6c8 905 mach->ibc = sclp.ibc;
c54f0d6a 906 memcpy(&mach->fac_mask, kvm->arch.model.fac_mask,
981467c9 907 S390_ARCH_FAC_LIST_SIZE_BYTE);
658b6eda 908 memcpy((unsigned long *)&mach->fac_list, S390_lowcore.stfle_fac_list,
94422ee8 909 S390_ARCH_FAC_LIST_SIZE_BYTE);
658b6eda
MM
910 if (copy_to_user((void __user *)attr->addr, mach, sizeof(*mach)))
911 ret = -EFAULT;
912 kfree(mach);
913out:
914 return ret;
915}
916
15c9705f
DH
917static int kvm_s390_get_processor_feat(struct kvm *kvm,
918 struct kvm_device_attr *attr)
919{
920 struct kvm_s390_vm_cpu_feat data;
921
922 bitmap_copy((unsigned long *) data.feat, kvm->arch.cpu_feat,
923 KVM_S390_VM_CPU_FEAT_NR_BITS);
924 if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
925 return -EFAULT;
926 return 0;
927}
928
929static int kvm_s390_get_machine_feat(struct kvm *kvm,
930 struct kvm_device_attr *attr)
931{
932 struct kvm_s390_vm_cpu_feat data;
933
934 bitmap_copy((unsigned long *) data.feat,
935 kvm_s390_available_cpu_feat,
936 KVM_S390_VM_CPU_FEAT_NR_BITS);
937 if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
938 return -EFAULT;
939 return 0;
940}
941
0a763c78
DH
942static int kvm_s390_get_processor_subfunc(struct kvm *kvm,
943 struct kvm_device_attr *attr)
944{
945 /*
946 * Once we can actually configure subfunctions (kernel + hw support),
947 * we have to check if they were already set by user space, if so copy
948 * them from kvm->arch.
949 */
950 return -ENXIO;
951}
952
953static int kvm_s390_get_machine_subfunc(struct kvm *kvm,
954 struct kvm_device_attr *attr)
955{
956 if (copy_to_user((void __user *)attr->addr, &kvm_s390_available_subfunc,
957 sizeof(struct kvm_s390_vm_cpu_subfunc)))
958 return -EFAULT;
959 return 0;
960}
658b6eda
MM
961static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
962{
963 int ret = -ENXIO;
964
965 switch (attr->attr) {
966 case KVM_S390_VM_CPU_PROCESSOR:
967 ret = kvm_s390_get_processor(kvm, attr);
968 break;
969 case KVM_S390_VM_CPU_MACHINE:
970 ret = kvm_s390_get_machine(kvm, attr);
971 break;
15c9705f
DH
972 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
973 ret = kvm_s390_get_processor_feat(kvm, attr);
974 break;
975 case KVM_S390_VM_CPU_MACHINE_FEAT:
976 ret = kvm_s390_get_machine_feat(kvm, attr);
977 break;
0a763c78
DH
978 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
979 ret = kvm_s390_get_processor_subfunc(kvm, attr);
980 break;
981 case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
982 ret = kvm_s390_get_machine_subfunc(kvm, attr);
983 break;
658b6eda
MM
984 }
985 return ret;
986}
987
f2061656
DD
988static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
989{
990 int ret;
991
992 switch (attr->group) {
4f718eab 993 case KVM_S390_VM_MEM_CTRL:
8c0a7ce6 994 ret = kvm_s390_set_mem_control(kvm, attr);
4f718eab 995 break;
72f25020
JH
996 case KVM_S390_VM_TOD:
997 ret = kvm_s390_set_tod(kvm, attr);
998 break;
658b6eda
MM
999 case KVM_S390_VM_CPU_MODEL:
1000 ret = kvm_s390_set_cpu_model(kvm, attr);
1001 break;
a374e892
TK
1002 case KVM_S390_VM_CRYPTO:
1003 ret = kvm_s390_vm_set_crypto(kvm, attr);
1004 break;
f2061656
DD
1005 default:
1006 ret = -ENXIO;
1007 break;
1008 }
1009
1010 return ret;
1011}
1012
1013static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1014{
8c0a7ce6
DD
1015 int ret;
1016
1017 switch (attr->group) {
1018 case KVM_S390_VM_MEM_CTRL:
1019 ret = kvm_s390_get_mem_control(kvm, attr);
1020 break;
72f25020
JH
1021 case KVM_S390_VM_TOD:
1022 ret = kvm_s390_get_tod(kvm, attr);
1023 break;
658b6eda
MM
1024 case KVM_S390_VM_CPU_MODEL:
1025 ret = kvm_s390_get_cpu_model(kvm, attr);
1026 break;
8c0a7ce6
DD
1027 default:
1028 ret = -ENXIO;
1029 break;
1030 }
1031
1032 return ret;
f2061656
DD
1033}
1034
1035static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1036{
1037 int ret;
1038
1039 switch (attr->group) {
4f718eab
DD
1040 case KVM_S390_VM_MEM_CTRL:
1041 switch (attr->attr) {
1042 case KVM_S390_VM_MEM_ENABLE_CMMA:
1043 case KVM_S390_VM_MEM_CLR_CMMA:
f9cbd9b0
DH
1044 ret = sclp.has_cmma ? 0 : -ENXIO;
1045 break;
8c0a7ce6 1046 case KVM_S390_VM_MEM_LIMIT_SIZE:
4f718eab
DD
1047 ret = 0;
1048 break;
1049 default:
1050 ret = -ENXIO;
1051 break;
1052 }
1053 break;
72f25020
JH
1054 case KVM_S390_VM_TOD:
1055 switch (attr->attr) {
1056 case KVM_S390_VM_TOD_LOW:
1057 case KVM_S390_VM_TOD_HIGH:
1058 ret = 0;
1059 break;
1060 default:
1061 ret = -ENXIO;
1062 break;
1063 }
1064 break;
658b6eda
MM
1065 case KVM_S390_VM_CPU_MODEL:
1066 switch (attr->attr) {
1067 case KVM_S390_VM_CPU_PROCESSOR:
1068 case KVM_S390_VM_CPU_MACHINE:
15c9705f
DH
1069 case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1070 case KVM_S390_VM_CPU_MACHINE_FEAT:
0a763c78 1071 case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
658b6eda
MM
1072 ret = 0;
1073 break;
0a763c78
DH
1074 /* configuring subfunctions is not supported yet */
1075 case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
658b6eda
MM
1076 default:
1077 ret = -ENXIO;
1078 break;
1079 }
1080 break;
a374e892
TK
1081 case KVM_S390_VM_CRYPTO:
1082 switch (attr->attr) {
1083 case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
1084 case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
1085 case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
1086 case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
1087 ret = 0;
1088 break;
1089 default:
1090 ret = -ENXIO;
1091 break;
1092 }
1093 break;
f2061656
DD
1094 default:
1095 ret = -ENXIO;
1096 break;
1097 }
1098
1099 return ret;
1100}
1101
30ee2a98
JH
1102static long kvm_s390_get_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1103{
1104 uint8_t *keys;
1105 uint64_t hva;
30ee2a98
JH
1106 int i, r = 0;
1107
1108 if (args->flags != 0)
1109 return -EINVAL;
1110
1111 /* Is this guest using storage keys? */
1112 if (!mm_use_skey(current->mm))
1113 return KVM_S390_GET_SKEYS_NONE;
1114
1115 /* Enforce sane limit on memory allocation */
1116 if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1117 return -EINVAL;
1118
1119 keys = kmalloc_array(args->count, sizeof(uint8_t),
1120 GFP_KERNEL | __GFP_NOWARN);
1121 if (!keys)
1122 keys = vmalloc(sizeof(uint8_t) * args->count);
1123 if (!keys)
1124 return -ENOMEM;
1125
d3ed1cee 1126 down_read(&current->mm->mmap_sem);
30ee2a98
JH
1127 for (i = 0; i < args->count; i++) {
1128 hva = gfn_to_hva(kvm, args->start_gfn + i);
1129 if (kvm_is_error_hva(hva)) {
1130 r = -EFAULT;
d3ed1cee 1131 break;
30ee2a98
JH
1132 }
1133
154c8c19
DH
1134 r = get_guest_storage_key(current->mm, hva, &keys[i]);
1135 if (r)
d3ed1cee 1136 break;
30ee2a98 1137 }
d3ed1cee
MS
1138 up_read(&current->mm->mmap_sem);
1139
1140 if (!r) {
1141 r = copy_to_user((uint8_t __user *)args->skeydata_addr, keys,
1142 sizeof(uint8_t) * args->count);
1143 if (r)
1144 r = -EFAULT;
30ee2a98
JH
1145 }
1146
30ee2a98
JH
1147 kvfree(keys);
1148 return r;
1149}
1150
1151static long kvm_s390_set_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1152{
1153 uint8_t *keys;
1154 uint64_t hva;
1155 int i, r = 0;
1156
1157 if (args->flags != 0)
1158 return -EINVAL;
1159
1160 /* Enforce sane limit on memory allocation */
1161 if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1162 return -EINVAL;
1163
1164 keys = kmalloc_array(args->count, sizeof(uint8_t),
1165 GFP_KERNEL | __GFP_NOWARN);
1166 if (!keys)
1167 keys = vmalloc(sizeof(uint8_t) * args->count);
1168 if (!keys)
1169 return -ENOMEM;
1170
1171 r = copy_from_user(keys, (uint8_t __user *)args->skeydata_addr,
1172 sizeof(uint8_t) * args->count);
1173 if (r) {
1174 r = -EFAULT;
1175 goto out;
1176 }
1177
1178 /* Enable storage key handling for the guest */
14d4a425
DD
1179 r = s390_enable_skey();
1180 if (r)
1181 goto out;
30ee2a98 1182
d3ed1cee 1183 down_read(&current->mm->mmap_sem);
30ee2a98
JH
1184 for (i = 0; i < args->count; i++) {
1185 hva = gfn_to_hva(kvm, args->start_gfn + i);
1186 if (kvm_is_error_hva(hva)) {
1187 r = -EFAULT;
d3ed1cee 1188 break;
30ee2a98
JH
1189 }
1190
1191 /* Lowest order bit is reserved */
1192 if (keys[i] & 0x01) {
1193 r = -EINVAL;
d3ed1cee 1194 break;
30ee2a98
JH
1195 }
1196
fe69eabf 1197 r = set_guest_storage_key(current->mm, hva, keys[i], 0);
30ee2a98 1198 if (r)
d3ed1cee 1199 break;
30ee2a98 1200 }
d3ed1cee 1201 up_read(&current->mm->mmap_sem);
30ee2a98
JH
1202out:
1203 kvfree(keys);
1204 return r;
1205}
1206
b0c632db
HC
1207long kvm_arch_vm_ioctl(struct file *filp,
1208 unsigned int ioctl, unsigned long arg)
1209{
1210 struct kvm *kvm = filp->private_data;
1211 void __user *argp = (void __user *)arg;
f2061656 1212 struct kvm_device_attr attr;
b0c632db
HC
1213 int r;
1214
1215 switch (ioctl) {
ba5c1e9b
CO
1216 case KVM_S390_INTERRUPT: {
1217 struct kvm_s390_interrupt s390int;
1218
1219 r = -EFAULT;
1220 if (copy_from_user(&s390int, argp, sizeof(s390int)))
1221 break;
1222 r = kvm_s390_inject_vm(kvm, &s390int);
1223 break;
1224 }
d938dc55
CH
1225 case KVM_ENABLE_CAP: {
1226 struct kvm_enable_cap cap;
1227 r = -EFAULT;
1228 if (copy_from_user(&cap, argp, sizeof(cap)))
1229 break;
1230 r = kvm_vm_ioctl_enable_cap(kvm, &cap);
1231 break;
1232 }
84223598
CH
1233 case KVM_CREATE_IRQCHIP: {
1234 struct kvm_irq_routing_entry routing;
1235
1236 r = -EINVAL;
1237 if (kvm->arch.use_irqchip) {
1238 /* Set up dummy routing. */
1239 memset(&routing, 0, sizeof(routing));
152b2839 1240 r = kvm_set_irq_routing(kvm, &routing, 0, 0);
84223598
CH
1241 }
1242 break;
1243 }
f2061656
DD
1244 case KVM_SET_DEVICE_ATTR: {
1245 r = -EFAULT;
1246 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
1247 break;
1248 r = kvm_s390_vm_set_attr(kvm, &attr);
1249 break;
1250 }
1251 case KVM_GET_DEVICE_ATTR: {
1252 r = -EFAULT;
1253 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
1254 break;
1255 r = kvm_s390_vm_get_attr(kvm, &attr);
1256 break;
1257 }
1258 case KVM_HAS_DEVICE_ATTR: {
1259 r = -EFAULT;
1260 if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
1261 break;
1262 r = kvm_s390_vm_has_attr(kvm, &attr);
1263 break;
1264 }
30ee2a98
JH
1265 case KVM_S390_GET_SKEYS: {
1266 struct kvm_s390_skeys args;
1267
1268 r = -EFAULT;
1269 if (copy_from_user(&args, argp,
1270 sizeof(struct kvm_s390_skeys)))
1271 break;
1272 r = kvm_s390_get_skeys(kvm, &args);
1273 break;
1274 }
1275 case KVM_S390_SET_SKEYS: {
1276 struct kvm_s390_skeys args;
1277
1278 r = -EFAULT;
1279 if (copy_from_user(&args, argp,
1280 sizeof(struct kvm_s390_skeys)))
1281 break;
1282 r = kvm_s390_set_skeys(kvm, &args);
1283 break;
1284 }
b0c632db 1285 default:
367e1319 1286 r = -ENOTTY;
b0c632db
HC
1287 }
1288
1289 return r;
1290}
1291
45c9b47c
TK
1292static int kvm_s390_query_ap_config(u8 *config)
1293{
1294 u32 fcn_code = 0x04000000UL;
86044c8c 1295 u32 cc = 0;
45c9b47c 1296
86044c8c 1297 memset(config, 0, 128);
45c9b47c
TK
1298 asm volatile(
1299 "lgr 0,%1\n"
1300 "lgr 2,%2\n"
1301 ".long 0xb2af0000\n" /* PQAP(QCI) */
86044c8c 1302 "0: ipm %0\n"
45c9b47c 1303 "srl %0,28\n"
86044c8c
CB
1304 "1:\n"
1305 EX_TABLE(0b, 1b)
1306 : "+r" (cc)
45c9b47c
TK
1307 : "r" (fcn_code), "r" (config)
1308 : "cc", "0", "2", "memory"
1309 );
1310
1311 return cc;
1312}
1313
1314static int kvm_s390_apxa_installed(void)
1315{
1316 u8 config[128];
1317 int cc;
1318
a6aacc3f 1319 if (test_facility(12)) {
45c9b47c
TK
1320 cc = kvm_s390_query_ap_config(config);
1321
1322 if (cc)
1323 pr_err("PQAP(QCI) failed with cc=%d", cc);
1324 else
1325 return config[0] & 0x40;
1326 }
1327
1328 return 0;
1329}
1330
1331static void kvm_s390_set_crycb_format(struct kvm *kvm)
1332{
1333 kvm->arch.crypto.crycbd = (__u32)(unsigned long) kvm->arch.crypto.crycb;
1334
1335 if (kvm_s390_apxa_installed())
1336 kvm->arch.crypto.crycbd |= CRYCB_FORMAT2;
1337 else
1338 kvm->arch.crypto.crycbd |= CRYCB_FORMAT1;
1339}
1340
9bb0ec09 1341static u64 kvm_s390_get_initial_cpuid(void)
9d8d5786 1342{
9bb0ec09
DH
1343 struct cpuid cpuid;
1344
1345 get_cpu_id(&cpuid);
1346 cpuid.version = 0xff;
1347 return *((u64 *) &cpuid);
9d8d5786
MM
1348}
1349
c54f0d6a 1350static void kvm_s390_crypto_init(struct kvm *kvm)
5102ee87 1351{
9d8d5786 1352 if (!test_kvm_facility(kvm, 76))
c54f0d6a 1353 return;
5102ee87 1354
c54f0d6a 1355 kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
45c9b47c 1356 kvm_s390_set_crycb_format(kvm);
5102ee87 1357
ed6f76b4
TK
1358 /* Enable AES/DEA protected key functions by default */
1359 kvm->arch.crypto.aes_kw = 1;
1360 kvm->arch.crypto.dea_kw = 1;
1361 get_random_bytes(kvm->arch.crypto.crycb->aes_wrapping_key_mask,
1362 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
1363 get_random_bytes(kvm->arch.crypto.crycb->dea_wrapping_key_mask,
1364 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
5102ee87
TK
1365}
1366
7d43bafc
ED
1367static void sca_dispose(struct kvm *kvm)
1368{
1369 if (kvm->arch.use_esca)
5e044315 1370 free_pages_exact(kvm->arch.sca, sizeof(struct esca_block));
7d43bafc
ED
1371 else
1372 free_page((unsigned long)(kvm->arch.sca));
1373 kvm->arch.sca = NULL;
1374}
1375
e08b9637 1376int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
b0c632db 1377{
76a6dd72 1378 gfp_t alloc_flags = GFP_KERNEL;
9d8d5786 1379 int i, rc;
b0c632db 1380 char debug_name[16];
f6c137ff 1381 static unsigned long sca_offset;
b0c632db 1382
e08b9637
CO
1383 rc = -EINVAL;
1384#ifdef CONFIG_KVM_S390_UCONTROL
1385 if (type & ~KVM_VM_S390_UCONTROL)
1386 goto out_err;
1387 if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN)))
1388 goto out_err;
1389#else
1390 if (type)
1391 goto out_err;
1392#endif
1393
b0c632db
HC
1394 rc = s390_enable_sie();
1395 if (rc)
d89f5eff 1396 goto out_err;
b0c632db 1397
b290411a
CO
1398 rc = -ENOMEM;
1399
7d0a5e62
JF
1400 ratelimit_state_init(&kvm->arch.sthyi_limit, 5 * HZ, 500);
1401
7d43bafc 1402 kvm->arch.use_esca = 0; /* start with basic SCA */
76a6dd72
DH
1403 if (!sclp.has_64bscao)
1404 alloc_flags |= GFP_DMA;
5e044315 1405 rwlock_init(&kvm->arch.sca_lock);
76a6dd72 1406 kvm->arch.sca = (struct bsca_block *) get_zeroed_page(alloc_flags);
b0c632db 1407 if (!kvm->arch.sca)
d89f5eff 1408 goto out_err;
f6c137ff 1409 spin_lock(&kvm_lock);
c5c2c393 1410 sca_offset += 16;
bc784cce 1411 if (sca_offset + sizeof(struct bsca_block) > PAGE_SIZE)
c5c2c393 1412 sca_offset = 0;
bc784cce
ED
1413 kvm->arch.sca = (struct bsca_block *)
1414 ((char *) kvm->arch.sca + sca_offset);
f6c137ff 1415 spin_unlock(&kvm_lock);
b0c632db
HC
1416
1417 sprintf(debug_name, "kvm-%u", current->pid);
1418
1cb9cf72 1419 kvm->arch.dbf = debug_register(debug_name, 32, 1, 7 * sizeof(long));
b0c632db 1420 if (!kvm->arch.dbf)
40f5b735 1421 goto out_err;
b0c632db 1422
c54f0d6a
DH
1423 kvm->arch.sie_page2 =
1424 (struct sie_page2 *) get_zeroed_page(GFP_KERNEL | GFP_DMA);
1425 if (!kvm->arch.sie_page2)
40f5b735 1426 goto out_err;
9d8d5786 1427
fb5bf93f 1428 /* Populate the facility mask initially. */
c54f0d6a 1429 memcpy(kvm->arch.model.fac_mask, S390_lowcore.stfle_fac_list,
94422ee8 1430 S390_ARCH_FAC_LIST_SIZE_BYTE);
9d8d5786
MM
1431 for (i = 0; i < S390_ARCH_FAC_LIST_SIZE_U64; i++) {
1432 if (i < kvm_s390_fac_list_mask_size())
c54f0d6a 1433 kvm->arch.model.fac_mask[i] &= kvm_s390_fac_list_mask[i];
9d8d5786 1434 else
c54f0d6a 1435 kvm->arch.model.fac_mask[i] = 0UL;
9d8d5786
MM
1436 }
1437
981467c9 1438 /* Populate the facility list initially. */
c54f0d6a
DH
1439 kvm->arch.model.fac_list = kvm->arch.sie_page2->fac_list;
1440 memcpy(kvm->arch.model.fac_list, kvm->arch.model.fac_mask,
981467c9
MM
1441 S390_ARCH_FAC_LIST_SIZE_BYTE);
1442
95ca2cb5
JF
1443 set_kvm_facility(kvm->arch.model.fac_mask, 74);
1444 set_kvm_facility(kvm->arch.model.fac_list, 74);
1445
9bb0ec09 1446 kvm->arch.model.cpuid = kvm_s390_get_initial_cpuid();
37c5f6c8 1447 kvm->arch.model.ibc = sclp.ibc & 0x0fff;
9d8d5786 1448
c54f0d6a 1449 kvm_s390_crypto_init(kvm);
5102ee87 1450
ba5c1e9b 1451 spin_lock_init(&kvm->arch.float_int.lock);
6d3da241
JF
1452 for (i = 0; i < FIRQ_LIST_COUNT; i++)
1453 INIT_LIST_HEAD(&kvm->arch.float_int.lists[i]);
8a242234 1454 init_waitqueue_head(&kvm->arch.ipte_wq);
a6b7e459 1455 mutex_init(&kvm->arch.ipte_mutex);
ba5c1e9b 1456
b0c632db 1457 debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
78f26131 1458 VM_EVENT(kvm, 3, "vm created with type %lu", type);
b0c632db 1459
e08b9637
CO
1460 if (type & KVM_VM_S390_UCONTROL) {
1461 kvm->arch.gmap = NULL;
a3a92c31 1462 kvm->arch.mem_limit = KVM_S390_NO_MEM_LIMIT;
e08b9637 1463 } else {
32e6b236
GH
1464 if (sclp.hamax == U64_MAX)
1465 kvm->arch.mem_limit = TASK_MAX_SIZE;
1466 else
1467 kvm->arch.mem_limit = min_t(unsigned long, TASK_MAX_SIZE,
1468 sclp.hamax + 1);
6ea427bb 1469 kvm->arch.gmap = gmap_create(current->mm, kvm->arch.mem_limit - 1);
e08b9637 1470 if (!kvm->arch.gmap)
40f5b735 1471 goto out_err;
2c70fe44 1472 kvm->arch.gmap->private = kvm;
24eb3a82 1473 kvm->arch.gmap->pfault_enabled = 0;
e08b9637 1474 }
fa6b7fe9
CH
1475
1476 kvm->arch.css_support = 0;
84223598 1477 kvm->arch.use_irqchip = 0;
72f25020 1478 kvm->arch.epoch = 0;
fa6b7fe9 1479
8ad35755 1480 spin_lock_init(&kvm->arch.start_stop_lock);
a3508fbe 1481 kvm_s390_vsie_init(kvm);
8335713a 1482 KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm, current->pid);
8ad35755 1483
d89f5eff 1484 return 0;
40f5b735 1485out_err:
c54f0d6a 1486 free_page((unsigned long)kvm->arch.sie_page2);
598841ca 1487 debug_unregister(kvm->arch.dbf);
7d43bafc 1488 sca_dispose(kvm);
78f26131 1489 KVM_EVENT(3, "creation of vm failed: %d", rc);
d89f5eff 1490 return rc;
b0c632db
HC
1491}
1492
d329c035
CB
1493void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
1494{
1495 VCPU_EVENT(vcpu, 3, "%s", "free cpu");
ade38c31 1496 trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id);
67335e63 1497 kvm_s390_clear_local_irqs(vcpu);
3c038e6b 1498 kvm_clear_async_pf_completion_queue(vcpu);
bc784cce 1499 if (!kvm_is_ucontrol(vcpu->kvm))
a6e2f683 1500 sca_del_vcpu(vcpu);
27e0393f
CO
1501
1502 if (kvm_is_ucontrol(vcpu->kvm))
6ea427bb 1503 gmap_remove(vcpu->arch.gmap);
27e0393f 1504
e6db1d61 1505 if (vcpu->kvm->arch.use_cmma)
b31605c1 1506 kvm_s390_vcpu_unsetup_cmma(vcpu);
d329c035 1507 free_page((unsigned long)(vcpu->arch.sie_block));
b31288fa 1508
6692cef3 1509 kvm_vcpu_uninit(vcpu);
b110feaf 1510 kmem_cache_free(kvm_vcpu_cache, vcpu);
d329c035
CB
1511}
1512
1513static void kvm_free_vcpus(struct kvm *kvm)
1514{
1515 unsigned int i;
988a2cae 1516 struct kvm_vcpu *vcpu;
d329c035 1517
988a2cae
GN
1518 kvm_for_each_vcpu(i, vcpu, kvm)
1519 kvm_arch_vcpu_destroy(vcpu);
1520
1521 mutex_lock(&kvm->lock);
1522 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
1523 kvm->vcpus[i] = NULL;
1524
1525 atomic_set(&kvm->online_vcpus, 0);
1526 mutex_unlock(&kvm->lock);
d329c035
CB
1527}
1528
b0c632db
HC
1529void kvm_arch_destroy_vm(struct kvm *kvm)
1530{
d329c035 1531 kvm_free_vcpus(kvm);
7d43bafc 1532 sca_dispose(kvm);
d329c035 1533 debug_unregister(kvm->arch.dbf);
c54f0d6a 1534 free_page((unsigned long)kvm->arch.sie_page2);
27e0393f 1535 if (!kvm_is_ucontrol(kvm))
6ea427bb 1536 gmap_remove(kvm->arch.gmap);
841b91c5 1537 kvm_s390_destroy_adapters(kvm);
67335e63 1538 kvm_s390_clear_float_irqs(kvm);
a3508fbe 1539 kvm_s390_vsie_destroy(kvm);
8335713a 1540 KVM_EVENT(3, "vm 0x%pK destroyed", kvm);
b0c632db
HC
1541}
1542
1543/* Section: vcpu related */
dafd032a
DD
1544static int __kvm_ucontrol_vcpu_init(struct kvm_vcpu *vcpu)
1545{
6ea427bb 1546 vcpu->arch.gmap = gmap_create(current->mm, -1UL);
dafd032a
DD
1547 if (!vcpu->arch.gmap)
1548 return -ENOMEM;
1549 vcpu->arch.gmap->private = vcpu->kvm;
1550
1551 return 0;
1552}
1553
a6e2f683
ED
1554static void sca_del_vcpu(struct kvm_vcpu *vcpu)
1555{
5e044315 1556 read_lock(&vcpu->kvm->arch.sca_lock);
7d43bafc
ED
1557 if (vcpu->kvm->arch.use_esca) {
1558 struct esca_block *sca = vcpu->kvm->arch.sca;
a6e2f683 1559
7d43bafc 1560 clear_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
10ce32d5 1561 sca->cpu[vcpu->vcpu_id].sda = 0;
7d43bafc
ED
1562 } else {
1563 struct bsca_block *sca = vcpu->kvm->arch.sca;
1564
1565 clear_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
10ce32d5 1566 sca->cpu[vcpu->vcpu_id].sda = 0;
7d43bafc 1567 }
5e044315 1568 read_unlock(&vcpu->kvm->arch.sca_lock);
a6e2f683
ED
1569}
1570
eaa78f34 1571static void sca_add_vcpu(struct kvm_vcpu *vcpu)
a6e2f683 1572{
eaa78f34
DH
1573 read_lock(&vcpu->kvm->arch.sca_lock);
1574 if (vcpu->kvm->arch.use_esca) {
1575 struct esca_block *sca = vcpu->kvm->arch.sca;
7d43bafc 1576
eaa78f34 1577 sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
7d43bafc
ED
1578 vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
1579 vcpu->arch.sie_block->scaol = (__u32)(__u64)sca & ~0x3fU;
25508824 1580 vcpu->arch.sie_block->ecb2 |= 0x04U;
eaa78f34 1581 set_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
7d43bafc 1582 } else {
eaa78f34 1583 struct bsca_block *sca = vcpu->kvm->arch.sca;
a6e2f683 1584
eaa78f34 1585 sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
7d43bafc
ED
1586 vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
1587 vcpu->arch.sie_block->scaol = (__u32)(__u64)sca;
eaa78f34 1588 set_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
7d43bafc 1589 }
eaa78f34 1590 read_unlock(&vcpu->kvm->arch.sca_lock);
5e044315
ED
1591}
1592
1593/* Basic SCA to Extended SCA data copy routines */
1594static inline void sca_copy_entry(struct esca_entry *d, struct bsca_entry *s)
1595{
1596 d->sda = s->sda;
1597 d->sigp_ctrl.c = s->sigp_ctrl.c;
1598 d->sigp_ctrl.scn = s->sigp_ctrl.scn;
1599}
1600
1601static void sca_copy_b_to_e(struct esca_block *d, struct bsca_block *s)
1602{
1603 int i;
1604
1605 d->ipte_control = s->ipte_control;
1606 d->mcn[0] = s->mcn;
1607 for (i = 0; i < KVM_S390_BSCA_CPU_SLOTS; i++)
1608 sca_copy_entry(&d->cpu[i], &s->cpu[i]);
1609}
1610
1611static int sca_switch_to_extended(struct kvm *kvm)
1612{
1613 struct bsca_block *old_sca = kvm->arch.sca;
1614 struct esca_block *new_sca;
1615 struct kvm_vcpu *vcpu;
1616 unsigned int vcpu_idx;
1617 u32 scaol, scaoh;
1618
1619 new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL|__GFP_ZERO);
1620 if (!new_sca)
1621 return -ENOMEM;
1622
1623 scaoh = (u32)((u64)(new_sca) >> 32);
1624 scaol = (u32)(u64)(new_sca) & ~0x3fU;
1625
1626 kvm_s390_vcpu_block_all(kvm);
1627 write_lock(&kvm->arch.sca_lock);
1628
1629 sca_copy_b_to_e(new_sca, old_sca);
1630
1631 kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
1632 vcpu->arch.sie_block->scaoh = scaoh;
1633 vcpu->arch.sie_block->scaol = scaol;
1634 vcpu->arch.sie_block->ecb2 |= 0x04U;
1635 }
1636 kvm->arch.sca = new_sca;
1637 kvm->arch.use_esca = 1;
1638
1639 write_unlock(&kvm->arch.sca_lock);
1640 kvm_s390_vcpu_unblock_all(kvm);
1641
1642 free_page((unsigned long)old_sca);
1643
8335713a
CB
1644 VM_EVENT(kvm, 2, "Switched to ESCA (0x%pK -> 0x%pK)",
1645 old_sca, kvm->arch.sca);
5e044315 1646 return 0;
a6e2f683
ED
1647}
1648
1649static int sca_can_add_vcpu(struct kvm *kvm, unsigned int id)
1650{
5e044315
ED
1651 int rc;
1652
1653 if (id < KVM_S390_BSCA_CPU_SLOTS)
1654 return true;
76a6dd72 1655 if (!sclp.has_esca || !sclp.has_64bscao)
5e044315
ED
1656 return false;
1657
1658 mutex_lock(&kvm->lock);
1659 rc = kvm->arch.use_esca ? 0 : sca_switch_to_extended(kvm);
1660 mutex_unlock(&kvm->lock);
1661
1662 return rc == 0 && id < KVM_S390_ESCA_CPU_SLOTS;
a6e2f683
ED
1663}
1664
b0c632db
HC
1665int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
1666{
3c038e6b
DD
1667 vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
1668 kvm_clear_async_pf_completion_queue(vcpu);
59674c1a
CB
1669 vcpu->run->kvm_valid_regs = KVM_SYNC_PREFIX |
1670 KVM_SYNC_GPRS |
9eed0735 1671 KVM_SYNC_ACRS |
b028ee3e
DH
1672 KVM_SYNC_CRS |
1673 KVM_SYNC_ARCH0 |
1674 KVM_SYNC_PFAULT;
75a4615c 1675 kvm_s390_set_prefix(vcpu, 0);
c6e5f166
FZ
1676 if (test_kvm_facility(vcpu->kvm, 64))
1677 vcpu->run->kvm_valid_regs |= KVM_SYNC_RICCB;
f6aa6dc4
DH
1678 /* fprs can be synchronized via vrs, even if the guest has no vx. With
1679 * MACHINE_HAS_VX, (load|store)_fpu_regs() will work with vrs format.
1680 */
1681 if (MACHINE_HAS_VX)
68c55750 1682 vcpu->run->kvm_valid_regs |= KVM_SYNC_VRS;
6fd8e67d
DH
1683 else
1684 vcpu->run->kvm_valid_regs |= KVM_SYNC_FPRS;
dafd032a
DD
1685
1686 if (kvm_is_ucontrol(vcpu->kvm))
1687 return __kvm_ucontrol_vcpu_init(vcpu);
1688
b0c632db
HC
1689 return 0;
1690}
1691
db0758b2
DH
1692/* needs disabled preemption to protect from TOD sync and vcpu_load/put */
1693static void __start_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1694{
1695 WARN_ON_ONCE(vcpu->arch.cputm_start != 0);
9c23a131 1696 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
db0758b2 1697 vcpu->arch.cputm_start = get_tod_clock_fast();
9c23a131 1698 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
db0758b2
DH
1699}
1700
1701/* needs disabled preemption to protect from TOD sync and vcpu_load/put */
1702static void __stop_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1703{
1704 WARN_ON_ONCE(vcpu->arch.cputm_start == 0);
9c23a131 1705 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
db0758b2
DH
1706 vcpu->arch.sie_block->cputm -= get_tod_clock_fast() - vcpu->arch.cputm_start;
1707 vcpu->arch.cputm_start = 0;
9c23a131 1708 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
db0758b2
DH
1709}
1710
1711/* needs disabled preemption to protect from TOD sync and vcpu_load/put */
1712static void __enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1713{
1714 WARN_ON_ONCE(vcpu->arch.cputm_enabled);
1715 vcpu->arch.cputm_enabled = true;
1716 __start_cpu_timer_accounting(vcpu);
1717}
1718
1719/* needs disabled preemption to protect from TOD sync and vcpu_load/put */
1720static void __disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1721{
1722 WARN_ON_ONCE(!vcpu->arch.cputm_enabled);
1723 __stop_cpu_timer_accounting(vcpu);
1724 vcpu->arch.cputm_enabled = false;
1725}
1726
1727static void enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1728{
1729 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
1730 __enable_cpu_timer_accounting(vcpu);
1731 preempt_enable();
1732}
1733
1734static void disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
1735{
1736 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
1737 __disable_cpu_timer_accounting(vcpu);
1738 preempt_enable();
1739}
1740
4287f247
DH
1741/* set the cpu timer - may only be called from the VCPU thread itself */
1742void kvm_s390_set_cpu_timer(struct kvm_vcpu *vcpu, __u64 cputm)
1743{
db0758b2 1744 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
9c23a131 1745 raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
db0758b2
DH
1746 if (vcpu->arch.cputm_enabled)
1747 vcpu->arch.cputm_start = get_tod_clock_fast();
4287f247 1748 vcpu->arch.sie_block->cputm = cputm;
9c23a131 1749 raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
db0758b2 1750 preempt_enable();
4287f247
DH
1751}
1752
db0758b2 1753/* update and get the cpu timer - can also be called from other VCPU threads */
4287f247
DH
1754__u64 kvm_s390_get_cpu_timer(struct kvm_vcpu *vcpu)
1755{
9c23a131 1756 unsigned int seq;
db0758b2 1757 __u64 value;
db0758b2
DH
1758
1759 if (unlikely(!vcpu->arch.cputm_enabled))
1760 return vcpu->arch.sie_block->cputm;
1761
9c23a131
DH
1762 preempt_disable(); /* protect from TOD sync and vcpu_load/put */
1763 do {
1764 seq = raw_read_seqcount(&vcpu->arch.cputm_seqcount);
1765 /*
1766 * If the writer would ever execute a read in the critical
1767 * section, e.g. in irq context, we have a deadlock.
1768 */
1769 WARN_ON_ONCE((seq & 1) && smp_processor_id() == vcpu->cpu);
1770 value = vcpu->arch.sie_block->cputm;
1771 /* if cputm_start is 0, accounting is being started/stopped */
1772 if (likely(vcpu->arch.cputm_start))
1773 value -= get_tod_clock_fast() - vcpu->arch.cputm_start;
1774 } while (read_seqcount_retry(&vcpu->arch.cputm_seqcount, seq & ~1));
1775 preempt_enable();
db0758b2 1776 return value;
4287f247
DH
1777}
1778
b0c632db
HC
1779void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1780{
9977e886 1781 /* Save host register state */
d0164ee2 1782 save_fpu_regs();
9abc2a08
DH
1783 vcpu->arch.host_fpregs.fpc = current->thread.fpu.fpc;
1784 vcpu->arch.host_fpregs.regs = current->thread.fpu.regs;
9977e886 1785
6fd8e67d
DH
1786 if (MACHINE_HAS_VX)
1787 current->thread.fpu.regs = vcpu->run->s.regs.vrs;
1788 else
1789 current->thread.fpu.regs = vcpu->run->s.regs.fprs;
9abc2a08 1790 current->thread.fpu.fpc = vcpu->run->s.regs.fpc;
9977e886 1791 if (test_fp_ctl(current->thread.fpu.fpc))
96b2d7a8 1792 /* User space provided an invalid FPC, let's clear it */
9977e886
HB
1793 current->thread.fpu.fpc = 0;
1794
1795 save_access_regs(vcpu->arch.host_acrs);
59674c1a 1796 restore_access_regs(vcpu->run->s.regs.acrs);
37d9df98 1797 gmap_enable(vcpu->arch.enabled_gmap);
805de8f4 1798 atomic_or(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags);
5ebda316 1799 if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
db0758b2 1800 __start_cpu_timer_accounting(vcpu);
01a745ac 1801 vcpu->cpu = cpu;
b0c632db
HC
1802}
1803
1804void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
1805{
01a745ac 1806 vcpu->cpu = -1;
5ebda316 1807 if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
db0758b2 1808 __stop_cpu_timer_accounting(vcpu);
805de8f4 1809 atomic_andnot(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags);
37d9df98
DH
1810 vcpu->arch.enabled_gmap = gmap_get_enabled();
1811 gmap_disable(vcpu->arch.enabled_gmap);
9977e886 1812
9abc2a08 1813 /* Save guest register state */
d0164ee2 1814 save_fpu_regs();
9abc2a08 1815 vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
9977e886 1816
9abc2a08
DH
1817 /* Restore host register state */
1818 current->thread.fpu.fpc = vcpu->arch.host_fpregs.fpc;
1819 current->thread.fpu.regs = vcpu->arch.host_fpregs.regs;
9977e886
HB
1820
1821 save_access_regs(vcpu->run->s.regs.acrs);
b0c632db
HC
1822 restore_access_regs(vcpu->arch.host_acrs);
1823}
1824
1825static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu)
1826{
1827 /* this equals initial cpu reset in pop, but we don't switch to ESA */
1828 vcpu->arch.sie_block->gpsw.mask = 0UL;
1829 vcpu->arch.sie_block->gpsw.addr = 0UL;
8d26cf7b 1830 kvm_s390_set_prefix(vcpu, 0);
4287f247 1831 kvm_s390_set_cpu_timer(vcpu, 0);
b0c632db
HC
1832 vcpu->arch.sie_block->ckc = 0UL;
1833 vcpu->arch.sie_block->todpr = 0;
1834 memset(vcpu->arch.sie_block->gcr, 0, 16 * sizeof(__u64));
1835 vcpu->arch.sie_block->gcr[0] = 0xE0UL;
1836 vcpu->arch.sie_block->gcr[14] = 0xC2000000UL;
9abc2a08
DH
1837 /* make sure the new fpc will be lazily loaded */
1838 save_fpu_regs();
1839 current->thread.fpu.fpc = 0;
b0c632db 1840 vcpu->arch.sie_block->gbea = 1;
672550fb 1841 vcpu->arch.sie_block->pp = 0;
3c038e6b
DD
1842 vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
1843 kvm_clear_async_pf_completion_queue(vcpu);
6352e4d2
DH
1844 if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
1845 kvm_s390_vcpu_stop(vcpu);
2ed10cc1 1846 kvm_s390_clear_local_irqs(vcpu);
b0c632db
HC
1847}
1848
31928aa5 1849void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
42897d86 1850{
72f25020 1851 mutex_lock(&vcpu->kvm->lock);
fdf03650 1852 preempt_disable();
72f25020 1853 vcpu->arch.sie_block->epoch = vcpu->kvm->arch.epoch;
fdf03650 1854 preempt_enable();
72f25020 1855 mutex_unlock(&vcpu->kvm->lock);
25508824 1856 if (!kvm_is_ucontrol(vcpu->kvm)) {
dafd032a 1857 vcpu->arch.gmap = vcpu->kvm->arch.gmap;
eaa78f34 1858 sca_add_vcpu(vcpu);
25508824 1859 }
6502a34c
DH
1860 if (test_kvm_facility(vcpu->kvm, 74) || vcpu->kvm->arch.user_instr0)
1861 vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
37d9df98
DH
1862 /* make vcpu_load load the right gmap on the first trigger */
1863 vcpu->arch.enabled_gmap = vcpu->arch.gmap;
42897d86
MT
1864}
1865
5102ee87
TK
1866static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu)
1867{
9d8d5786 1868 if (!test_kvm_facility(vcpu->kvm, 76))
5102ee87
TK
1869 return;
1870
a374e892
TK
1871 vcpu->arch.sie_block->ecb3 &= ~(ECB3_AES | ECB3_DEA);
1872
1873 if (vcpu->kvm->arch.crypto.aes_kw)
1874 vcpu->arch.sie_block->ecb3 |= ECB3_AES;
1875 if (vcpu->kvm->arch.crypto.dea_kw)
1876 vcpu->arch.sie_block->ecb3 |= ECB3_DEA;
1877
5102ee87
TK
1878 vcpu->arch.sie_block->crycbd = vcpu->kvm->arch.crypto.crycbd;
1879}
1880
b31605c1
DD
1881void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu)
1882{
1883 free_page(vcpu->arch.sie_block->cbrlo);
1884 vcpu->arch.sie_block->cbrlo = 0;
1885}
1886
1887int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu)
1888{
1889 vcpu->arch.sie_block->cbrlo = get_zeroed_page(GFP_KERNEL);
1890 if (!vcpu->arch.sie_block->cbrlo)
1891 return -ENOMEM;
1892
1893 vcpu->arch.sie_block->ecb2 |= 0x80;
1894 vcpu->arch.sie_block->ecb2 &= ~0x08;
1895 return 0;
1896}
1897
91520f1a
MM
1898static void kvm_s390_vcpu_setup_model(struct kvm_vcpu *vcpu)
1899{
1900 struct kvm_s390_cpu_model *model = &vcpu->kvm->arch.model;
1901
91520f1a 1902 vcpu->arch.sie_block->ibc = model->ibc;
80bc79dc 1903 if (test_kvm_facility(vcpu->kvm, 7))
c54f0d6a 1904 vcpu->arch.sie_block->fac = (u32)(u64) model->fac_list;
91520f1a
MM
1905}
1906
b0c632db
HC
1907int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1908{
b31605c1 1909 int rc = 0;
b31288fa 1910
9e6dabef
CH
1911 atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH |
1912 CPUSTAT_SM |
a4a4f191
GH
1913 CPUSTAT_STOPPED);
1914
53df84f8 1915 if (test_kvm_facility(vcpu->kvm, 78))
805de8f4 1916 atomic_or(CPUSTAT_GED2, &vcpu->arch.sie_block->cpuflags);
53df84f8 1917 else if (test_kvm_facility(vcpu->kvm, 8))
805de8f4 1918 atomic_or(CPUSTAT_GED, &vcpu->arch.sie_block->cpuflags);
a4a4f191 1919
91520f1a
MM
1920 kvm_s390_vcpu_setup_model(vcpu);
1921
bdab09f3
DH
1922 /* pgste_set_pte has special handling for !MACHINE_HAS_ESOP */
1923 if (MACHINE_HAS_ESOP)
1924 vcpu->arch.sie_block->ecb |= 0x02;
bd50e8ec
DH
1925 if (test_kvm_facility(vcpu->kvm, 9))
1926 vcpu->arch.sie_block->ecb |= 0x04;
f597d24e 1927 if (test_kvm_facility(vcpu->kvm, 73))
7feb6bb8
MM
1928 vcpu->arch.sie_block->ecb |= 0x10;
1929
873b425e 1930 if (test_kvm_facility(vcpu->kvm, 8) && sclp.has_pfmfi)
d6af0b49 1931 vcpu->arch.sie_block->ecb2 |= 0x08;
48ee7d3a
DH
1932 vcpu->arch.sie_block->eca = 0x1002000U;
1933 if (sclp.has_cei)
1934 vcpu->arch.sie_block->eca |= 0x80000000U;
11ad65b7
DH
1935 if (sclp.has_ib)
1936 vcpu->arch.sie_block->eca |= 0x40000000U;
37c5f6c8 1937 if (sclp.has_siif)
217a4406 1938 vcpu->arch.sie_block->eca |= 1;
37c5f6c8 1939 if (sclp.has_sigpif)
ea5f4969 1940 vcpu->arch.sie_block->eca |= 0x10000000U;
c6e5f166
FZ
1941 if (test_kvm_facility(vcpu->kvm, 64))
1942 vcpu->arch.sie_block->ecb3 |= 0x01;
18280d8b 1943 if (test_kvm_facility(vcpu->kvm, 129)) {
13211ea7
EF
1944 vcpu->arch.sie_block->eca |= 0x00020000;
1945 vcpu->arch.sie_block->ecd |= 0x20000000;
1946 }
c6e5f166 1947 vcpu->arch.sie_block->riccbd = (unsigned long) &vcpu->run->s.regs.riccb;
492d8642 1948 vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
5a5e6536 1949
e6db1d61 1950 if (vcpu->kvm->arch.use_cmma) {
b31605c1
DD
1951 rc = kvm_s390_vcpu_setup_cmma(vcpu);
1952 if (rc)
1953 return rc;
b31288fa 1954 }
0ac96caf 1955 hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
ca872302 1956 vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup;
9d8d5786 1957
5102ee87
TK
1958 kvm_s390_vcpu_crypto_setup(vcpu);
1959
b31605c1 1960 return rc;
b0c632db
HC
1961}
1962
1963struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
1964 unsigned int id)
1965{
4d47555a 1966 struct kvm_vcpu *vcpu;
7feb6bb8 1967 struct sie_page *sie_page;
4d47555a
CO
1968 int rc = -EINVAL;
1969
4215825e 1970 if (!kvm_is_ucontrol(kvm) && !sca_can_add_vcpu(kvm, id))
4d47555a
CO
1971 goto out;
1972
1973 rc = -ENOMEM;
b0c632db 1974
b110feaf 1975 vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
b0c632db 1976 if (!vcpu)
4d47555a 1977 goto out;
b0c632db 1978
7feb6bb8
MM
1979 sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL);
1980 if (!sie_page)
b0c632db
HC
1981 goto out_free_cpu;
1982
7feb6bb8
MM
1983 vcpu->arch.sie_block = &sie_page->sie_block;
1984 vcpu->arch.sie_block->itdba = (unsigned long) &sie_page->itdb;
1985
efed1104
DH
1986 /* the real guest size will always be smaller than msl */
1987 vcpu->arch.sie_block->mso = 0;
1988 vcpu->arch.sie_block->msl = sclp.hamax;
1989
b0c632db 1990 vcpu->arch.sie_block->icpua = id;
ba5c1e9b 1991 spin_lock_init(&vcpu->arch.local_int.lock);
ba5c1e9b 1992 vcpu->arch.local_int.float_int = &kvm->arch.float_int;
d0321a24 1993 vcpu->arch.local_int.wq = &vcpu->wq;
5288fbf0 1994 vcpu->arch.local_int.cpuflags = &vcpu->arch.sie_block->cpuflags;
9c23a131 1995 seqcount_init(&vcpu->arch.cputm_seqcount);
ba5c1e9b 1996
b0c632db
HC
1997 rc = kvm_vcpu_init(vcpu, kvm, id);
1998 if (rc)
9abc2a08 1999 goto out_free_sie_block;
8335713a 2000 VM_EVENT(kvm, 3, "create cpu %d at 0x%pK, sie block at 0x%pK", id, vcpu,
b0c632db 2001 vcpu->arch.sie_block);
ade38c31 2002 trace_kvm_s390_create_vcpu(id, vcpu, vcpu->arch.sie_block);
b0c632db 2003
b0c632db 2004 return vcpu;
7b06bf2f
WY
2005out_free_sie_block:
2006 free_page((unsigned long)(vcpu->arch.sie_block));
b0c632db 2007out_free_cpu:
b110feaf 2008 kmem_cache_free(kvm_vcpu_cache, vcpu);
4d47555a 2009out:
b0c632db
HC
2010 return ERR_PTR(rc);
2011}
2012
b0c632db
HC
2013int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
2014{
9a022067 2015 return kvm_s390_vcpu_has_irq(vcpu, 0);
b0c632db
HC
2016}
2017
27406cd5 2018void kvm_s390_vcpu_block(struct kvm_vcpu *vcpu)
49b99e1e 2019{
805de8f4 2020 atomic_or(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
61a6df54 2021 exit_sie(vcpu);
49b99e1e
CB
2022}
2023
27406cd5 2024void kvm_s390_vcpu_unblock(struct kvm_vcpu *vcpu)
49b99e1e 2025{
805de8f4 2026 atomic_andnot(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
49b99e1e
CB
2027}
2028
8e236546
CB
2029static void kvm_s390_vcpu_request(struct kvm_vcpu *vcpu)
2030{
805de8f4 2031 atomic_or(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
61a6df54 2032 exit_sie(vcpu);
8e236546
CB
2033}
2034
2035static void kvm_s390_vcpu_request_handled(struct kvm_vcpu *vcpu)
2036{
9bf9fde2 2037 atomic_andnot(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
8e236546
CB
2038}
2039
49b99e1e
CB
2040/*
2041 * Kick a guest cpu out of SIE and wait until SIE is not running.
2042 * If the CPU is not running (e.g. waiting as idle) the function will
2043 * return immediately. */
2044void exit_sie(struct kvm_vcpu *vcpu)
2045{
805de8f4 2046 atomic_or(CPUSTAT_STOP_INT, &vcpu->arch.sie_block->cpuflags);
49b99e1e
CB
2047 while (vcpu->arch.sie_block->prog0c & PROG_IN_SIE)
2048 cpu_relax();
2049}
2050
8e236546
CB
2051/* Kick a guest cpu out of SIE to process a request synchronously */
2052void kvm_s390_sync_request(int req, struct kvm_vcpu *vcpu)
49b99e1e 2053{
8e236546
CB
2054 kvm_make_request(req, vcpu);
2055 kvm_s390_vcpu_request(vcpu);
49b99e1e
CB
2056}
2057
414d3b07
MS
2058static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
2059 unsigned long end)
2c70fe44 2060{
2c70fe44
CB
2061 struct kvm *kvm = gmap->private;
2062 struct kvm_vcpu *vcpu;
414d3b07
MS
2063 unsigned long prefix;
2064 int i;
2c70fe44 2065
65d0b0d4
DH
2066 if (gmap_is_shadow(gmap))
2067 return;
414d3b07
MS
2068 if (start >= 1UL << 31)
2069 /* We are only interested in prefix pages */
2070 return;
2c70fe44
CB
2071 kvm_for_each_vcpu(i, vcpu, kvm) {
2072 /* match against both prefix pages */
414d3b07
MS
2073 prefix = kvm_s390_get_prefix(vcpu);
2074 if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
2075 VCPU_EVENT(vcpu, 2, "gmap notifier for %lx-%lx",
2076 start, end);
8e236546 2077 kvm_s390_sync_request(KVM_REQ_MMU_RELOAD, vcpu);
2c70fe44
CB
2078 }
2079 }
2080}
2081
b6d33834
CD
2082int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
2083{
2084 /* kvm common code refers to this, but never calls it */
2085 BUG();
2086 return 0;
2087}
2088
14eebd91
CO
2089static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu,
2090 struct kvm_one_reg *reg)
2091{
2092 int r = -EINVAL;
2093
2094 switch (reg->id) {
29b7c71b
CO
2095 case KVM_REG_S390_TODPR:
2096 r = put_user(vcpu->arch.sie_block->todpr,
2097 (u32 __user *)reg->addr);
2098 break;
2099 case KVM_REG_S390_EPOCHDIFF:
2100 r = put_user(vcpu->arch.sie_block->epoch,
2101 (u64 __user *)reg->addr);
2102 break;
46a6dd1c 2103 case KVM_REG_S390_CPU_TIMER:
4287f247 2104 r = put_user(kvm_s390_get_cpu_timer(vcpu),
46a6dd1c
J
2105 (u64 __user *)reg->addr);
2106 break;
2107 case KVM_REG_S390_CLOCK_COMP:
2108 r = put_user(vcpu->arch.sie_block->ckc,
2109 (u64 __user *)reg->addr);
2110 break;
536336c2
DD
2111 case KVM_REG_S390_PFTOKEN:
2112 r = put_user(vcpu->arch.pfault_token,
2113 (u64 __user *)reg->addr);
2114 break;
2115 case KVM_REG_S390_PFCOMPARE:
2116 r = put_user(vcpu->arch.pfault_compare,
2117 (u64 __user *)reg->addr);
2118 break;
2119 case KVM_REG_S390_PFSELECT:
2120 r = put_user(vcpu->arch.pfault_select,
2121 (u64 __user *)reg->addr);
2122 break;
672550fb
CB
2123 case KVM_REG_S390_PP:
2124 r = put_user(vcpu->arch.sie_block->pp,
2125 (u64 __user *)reg->addr);
2126 break;
afa45ff5
CB
2127 case KVM_REG_S390_GBEA:
2128 r = put_user(vcpu->arch.sie_block->gbea,
2129 (u64 __user *)reg->addr);
2130 break;
14eebd91
CO
2131 default:
2132 break;
2133 }
2134
2135 return r;
2136}
2137
2138static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu,
2139 struct kvm_one_reg *reg)
2140{
2141 int r = -EINVAL;
4287f247 2142 __u64 val;
14eebd91
CO
2143
2144 switch (reg->id) {
29b7c71b
CO
2145 case KVM_REG_S390_TODPR:
2146 r = get_user(vcpu->arch.sie_block->todpr,
2147 (u32 __user *)reg->addr);
2148 break;
2149 case KVM_REG_S390_EPOCHDIFF:
2150 r = get_user(vcpu->arch.sie_block->epoch,
2151 (u64 __user *)reg->addr);
2152 break;
46a6dd1c 2153 case KVM_REG_S390_CPU_TIMER:
4287f247
DH
2154 r = get_user(val, (u64 __user *)reg->addr);
2155 if (!r)
2156 kvm_s390_set_cpu_timer(vcpu, val);
46a6dd1c
J
2157 break;
2158 case KVM_REG_S390_CLOCK_COMP:
2159 r = get_user(vcpu->arch.sie_block->ckc,
2160 (u64 __user *)reg->addr);
2161 break;
536336c2
DD
2162 case KVM_REG_S390_PFTOKEN:
2163 r = get_user(vcpu->arch.pfault_token,
2164 (u64 __user *)reg->addr);
9fbd8082
DH
2165 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
2166 kvm_clear_async_pf_completion_queue(vcpu);
536336c2
DD
2167 break;
2168 case KVM_REG_S390_PFCOMPARE:
2169 r = get_user(vcpu->arch.pfault_compare,
2170 (u64 __user *)reg->addr);
2171 break;
2172 case KVM_REG_S390_PFSELECT:
2173 r = get_user(vcpu->arch.pfault_select,
2174 (u64 __user *)reg->addr);
2175 break;
672550fb
CB
2176 case KVM_REG_S390_PP:
2177 r = get_user(vcpu->arch.sie_block->pp,
2178 (u64 __user *)reg->addr);
2179 break;
afa45ff5
CB
2180 case KVM_REG_S390_GBEA:
2181 r = get_user(vcpu->arch.sie_block->gbea,
2182 (u64 __user *)reg->addr);
2183 break;
14eebd91
CO
2184 default:
2185 break;
2186 }
2187
2188 return r;
2189}
b6d33834 2190
b0c632db
HC
2191static int kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu)
2192{
b0c632db 2193 kvm_s390_vcpu_initial_reset(vcpu);
b0c632db
HC
2194 return 0;
2195}
2196
2197int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
2198{
5a32c1af 2199 memcpy(&vcpu->run->s.regs.gprs, &regs->gprs, sizeof(regs->gprs));
b0c632db
HC
2200 return 0;
2201}
2202
2203int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
2204{
5a32c1af 2205 memcpy(&regs->gprs, &vcpu->run->s.regs.gprs, sizeof(regs->gprs));
b0c632db
HC
2206 return 0;
2207}
2208
2209int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
2210 struct kvm_sregs *sregs)
2211{
59674c1a 2212 memcpy(&vcpu->run->s.regs.acrs, &sregs->acrs, sizeof(sregs->acrs));
b0c632db 2213 memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs));
59674c1a 2214 restore_access_regs(vcpu->run->s.regs.acrs);
b0c632db
HC
2215 return 0;
2216}
2217
2218int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
2219 struct kvm_sregs *sregs)
2220{
59674c1a 2221 memcpy(&sregs->acrs, &vcpu->run->s.regs.acrs, sizeof(sregs->acrs));
b0c632db 2222 memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs));
b0c632db
HC
2223 return 0;
2224}
2225
2226int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
2227{
9abc2a08
DH
2228 /* make sure the new values will be lazily loaded */
2229 save_fpu_regs();
4725c860
MS
2230 if (test_fp_ctl(fpu->fpc))
2231 return -EINVAL;
9abc2a08
DH
2232 current->thread.fpu.fpc = fpu->fpc;
2233 if (MACHINE_HAS_VX)
a7d4b8f2
DH
2234 convert_fp_to_vx((__vector128 *) vcpu->run->s.regs.vrs,
2235 (freg_t *) fpu->fprs);
9abc2a08 2236 else
a7d4b8f2 2237 memcpy(vcpu->run->s.regs.fprs, &fpu->fprs, sizeof(fpu->fprs));
b0c632db
HC
2238 return 0;
2239}
2240
2241int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
2242{
9abc2a08
DH
2243 /* make sure we have the latest values */
2244 save_fpu_regs();
2245 if (MACHINE_HAS_VX)
a7d4b8f2
DH
2246 convert_vx_to_fp((freg_t *) fpu->fprs,
2247 (__vector128 *) vcpu->run->s.regs.vrs);
9abc2a08 2248 else
a7d4b8f2 2249 memcpy(fpu->fprs, vcpu->run->s.regs.fprs, sizeof(fpu->fprs));
9abc2a08 2250 fpu->fpc = current->thread.fpu.fpc;
b0c632db
HC
2251 return 0;
2252}
2253
2254static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw)
2255{
2256 int rc = 0;
2257
7a42fdc2 2258 if (!is_vcpu_stopped(vcpu))
b0c632db 2259 rc = -EBUSY;
d7b0b5eb
CO
2260 else {
2261 vcpu->run->psw_mask = psw.mask;
2262 vcpu->run->psw_addr = psw.addr;
2263 }
b0c632db
HC
2264 return rc;
2265}
2266
2267int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
2268 struct kvm_translation *tr)
2269{
2270 return -EINVAL; /* not implemented yet */
2271}
2272
27291e21
DH
2273#define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
2274 KVM_GUESTDBG_USE_HW_BP | \
2275 KVM_GUESTDBG_ENABLE)
2276
d0bfb940
JK
2277int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2278 struct kvm_guest_debug *dbg)
b0c632db 2279{
27291e21
DH
2280 int rc = 0;
2281
2282 vcpu->guest_debug = 0;
2283 kvm_s390_clear_bp_data(vcpu);
2284
2de3bfc2 2285 if (dbg->control & ~VALID_GUESTDBG_FLAGS)
27291e21 2286 return -EINVAL;
89b5b4de
DH
2287 if (!sclp.has_gpere)
2288 return -EINVAL;
27291e21
DH
2289
2290 if (dbg->control & KVM_GUESTDBG_ENABLE) {
2291 vcpu->guest_debug = dbg->control;
2292 /* enforce guest PER */
805de8f4 2293 atomic_or(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
27291e21
DH
2294
2295 if (dbg->control & KVM_GUESTDBG_USE_HW_BP)
2296 rc = kvm_s390_import_bp_data(vcpu, dbg);
2297 } else {
805de8f4 2298 atomic_andnot(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
27291e21
DH
2299 vcpu->arch.guestdbg.last_bp = 0;
2300 }
2301
2302 if (rc) {
2303 vcpu->guest_debug = 0;
2304 kvm_s390_clear_bp_data(vcpu);
805de8f4 2305 atomic_andnot(CPUSTAT_P, &vcpu->arch.sie_block->cpuflags);
27291e21
DH
2306 }
2307
2308 return rc;
b0c632db
HC
2309}
2310
62d9f0db
MT
2311int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
2312 struct kvm_mp_state *mp_state)
2313{
6352e4d2
DH
2314 /* CHECK_STOP and LOAD are not supported yet */
2315 return is_vcpu_stopped(vcpu) ? KVM_MP_STATE_STOPPED :
2316 KVM_MP_STATE_OPERATING;
62d9f0db
MT
2317}
2318
2319int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
2320 struct kvm_mp_state *mp_state)
2321{
6352e4d2
DH
2322 int rc = 0;
2323
2324 /* user space knows about this interface - let it control the state */
2325 vcpu->kvm->arch.user_cpu_state_ctrl = 1;
2326
2327 switch (mp_state->mp_state) {
2328 case KVM_MP_STATE_STOPPED:
2329 kvm_s390_vcpu_stop(vcpu);
2330 break;
2331 case KVM_MP_STATE_OPERATING:
2332 kvm_s390_vcpu_start(vcpu);
2333 break;
2334 case KVM_MP_STATE_LOAD:
2335 case KVM_MP_STATE_CHECK_STOP:
2336 /* fall through - CHECK_STOP and LOAD are not supported yet */
2337 default:
2338 rc = -ENXIO;
2339 }
2340
2341 return rc;
62d9f0db
MT
2342}
2343
8ad35755
DH
2344static bool ibs_enabled(struct kvm_vcpu *vcpu)
2345{
2346 return atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_IBS;
2347}
2348
2c70fe44
CB
2349static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
2350{
8ad35755 2351retry:
8e236546 2352 kvm_s390_vcpu_request_handled(vcpu);
586b7ccd
CB
2353 if (!vcpu->requests)
2354 return 0;
2c70fe44
CB
2355 /*
2356 * We use MMU_RELOAD just to re-arm the ipte notifier for the
b2d73b2a 2357 * guest prefix page. gmap_mprotect_notify will wait on the ptl lock.
2c70fe44
CB
2358 * This ensures that the ipte instruction for this request has
2359 * already finished. We might race against a second unmapper that
2360 * wants to set the blocking bit. Lets just retry the request loop.
2361 */
8ad35755 2362 if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu)) {
2c70fe44 2363 int rc;
b2d73b2a
MS
2364 rc = gmap_mprotect_notify(vcpu->arch.gmap,
2365 kvm_s390_get_prefix(vcpu),
2366 PAGE_SIZE * 2, PROT_WRITE);
aca411a4
JN
2367 if (rc) {
2368 kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu);
2c70fe44 2369 return rc;
aca411a4 2370 }
8ad35755 2371 goto retry;
2c70fe44 2372 }
8ad35755 2373
d3d692c8
DH
2374 if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
2375 vcpu->arch.sie_block->ihcpu = 0xffff;
2376 goto retry;
2377 }
2378
8ad35755
DH
2379 if (kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu)) {
2380 if (!ibs_enabled(vcpu)) {
2381 trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 1);
805de8f4 2382 atomic_or(CPUSTAT_IBS,
8ad35755
DH
2383 &vcpu->arch.sie_block->cpuflags);
2384 }
2385 goto retry;
2c70fe44 2386 }
8ad35755
DH
2387
2388 if (kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu)) {
2389 if (ibs_enabled(vcpu)) {
2390 trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 0);
805de8f4 2391 atomic_andnot(CPUSTAT_IBS,
8ad35755
DH
2392 &vcpu->arch.sie_block->cpuflags);
2393 }
2394 goto retry;
2395 }
2396
6502a34c
DH
2397 if (kvm_check_request(KVM_REQ_ICPT_OPEREXC, vcpu)) {
2398 vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
2399 goto retry;
2400 }
2401
0759d068
DH
2402 /* nothing to do, just clear the request */
2403 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
2404
2c70fe44
CB
2405 return 0;
2406}
2407
25ed1675
DH
2408void kvm_s390_set_tod_clock(struct kvm *kvm, u64 tod)
2409{
2410 struct kvm_vcpu *vcpu;
2411 int i;
2412
2413 mutex_lock(&kvm->lock);
2414 preempt_disable();
2415 kvm->arch.epoch = tod - get_tod_clock();
2416 kvm_s390_vcpu_block_all(kvm);
2417 kvm_for_each_vcpu(i, vcpu, kvm)
2418 vcpu->arch.sie_block->epoch = kvm->arch.epoch;
2419 kvm_s390_vcpu_unblock_all(kvm);
2420 preempt_enable();
2421 mutex_unlock(&kvm->lock);
2422}
2423
fa576c58
TH
2424/**
2425 * kvm_arch_fault_in_page - fault-in guest page if necessary
2426 * @vcpu: The corresponding virtual cpu
2427 * @gpa: Guest physical address
2428 * @writable: Whether the page should be writable or not
2429 *
2430 * Make sure that a guest page has been faulted-in on the host.
2431 *
2432 * Return: Zero on success, negative error code otherwise.
2433 */
2434long kvm_arch_fault_in_page(struct kvm_vcpu *vcpu, gpa_t gpa, int writable)
24eb3a82 2435{
527e30b4
MS
2436 return gmap_fault(vcpu->arch.gmap, gpa,
2437 writable ? FAULT_FLAG_WRITE : 0);
24eb3a82
DD
2438}
2439
3c038e6b
DD
2440static void __kvm_inject_pfault_token(struct kvm_vcpu *vcpu, bool start_token,
2441 unsigned long token)
2442{
2443 struct kvm_s390_interrupt inti;
383d0b05 2444 struct kvm_s390_irq irq;
3c038e6b
DD
2445
2446 if (start_token) {
383d0b05
JF
2447 irq.u.ext.ext_params2 = token;
2448 irq.type = KVM_S390_INT_PFAULT_INIT;
2449 WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &irq));
3c038e6b
DD
2450 } else {
2451 inti.type = KVM_S390_INT_PFAULT_DONE;
383d0b05 2452 inti.parm64 = token;
3c038e6b
DD
2453 WARN_ON_ONCE(kvm_s390_inject_vm(vcpu->kvm, &inti));
2454 }
2455}
2456
2457void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
2458 struct kvm_async_pf *work)
2459{
2460 trace_kvm_s390_pfault_init(vcpu, work->arch.pfault_token);
2461 __kvm_inject_pfault_token(vcpu, true, work->arch.pfault_token);
2462}
2463
2464void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
2465 struct kvm_async_pf *work)
2466{
2467 trace_kvm_s390_pfault_done(vcpu, work->arch.pfault_token);
2468 __kvm_inject_pfault_token(vcpu, false, work->arch.pfault_token);
2469}
2470
2471void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
2472 struct kvm_async_pf *work)
2473{
2474 /* s390 will always inject the page directly */
2475}
2476
2477bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu)
2478{
2479 /*
2480 * s390 will always inject the page directly,
2481 * but we still want check_async_completion to cleanup
2482 */
2483 return true;
2484}
2485
2486static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu)
2487{
2488 hva_t hva;
2489 struct kvm_arch_async_pf arch;
2490 int rc;
2491
2492 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
2493 return 0;
2494 if ((vcpu->arch.sie_block->gpsw.mask & vcpu->arch.pfault_select) !=
2495 vcpu->arch.pfault_compare)
2496 return 0;
2497 if (psw_extint_disabled(vcpu))
2498 return 0;
9a022067 2499 if (kvm_s390_vcpu_has_irq(vcpu, 0))
3c038e6b
DD
2500 return 0;
2501 if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
2502 return 0;
2503 if (!vcpu->arch.gmap->pfault_enabled)
2504 return 0;
2505
81480cc1
HC
2506 hva = gfn_to_hva(vcpu->kvm, gpa_to_gfn(current->thread.gmap_addr));
2507 hva += current->thread.gmap_addr & ~PAGE_MASK;
2508 if (read_guest_real(vcpu, vcpu->arch.pfault_token, &arch.pfault_token, 8))
3c038e6b
DD
2509 return 0;
2510
2511 rc = kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch);
2512 return rc;
2513}
2514
3fb4c40f 2515static int vcpu_pre_run(struct kvm_vcpu *vcpu)
b0c632db 2516{
3fb4c40f 2517 int rc, cpuflags;
e168bf8d 2518
3c038e6b
DD
2519 /*
2520 * On s390 notifications for arriving pages will be delivered directly
2521 * to the guest but the house keeping for completed pfaults is
2522 * handled outside the worker.
2523 */
2524 kvm_check_async_pf_completion(vcpu);
2525
7ec7c8c7
CB
2526 vcpu->arch.sie_block->gg14 = vcpu->run->s.regs.gprs[14];
2527 vcpu->arch.sie_block->gg15 = vcpu->run->s.regs.gprs[15];
b0c632db
HC
2528
2529 if (need_resched())
2530 schedule();
2531
d3a73acb 2532 if (test_cpu_flag(CIF_MCCK_PENDING))
71cde587
CB
2533 s390_handle_mcck();
2534
79395031
JF
2535 if (!kvm_is_ucontrol(vcpu->kvm)) {
2536 rc = kvm_s390_deliver_pending_interrupts(vcpu);
2537 if (rc)
2538 return rc;
2539 }
0ff31867 2540
2c70fe44
CB
2541 rc = kvm_s390_handle_requests(vcpu);
2542 if (rc)
2543 return rc;
2544
27291e21
DH
2545 if (guestdbg_enabled(vcpu)) {
2546 kvm_s390_backup_guest_per_regs(vcpu);
2547 kvm_s390_patch_guest_per_regs(vcpu);
2548 }
2549
b0c632db 2550 vcpu->arch.sie_block->icptcode = 0;
3fb4c40f
TH
2551 cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags);
2552 VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags);
2553 trace_kvm_s390_sie_enter(vcpu, cpuflags);
2b29a9fd 2554
3fb4c40f
TH
2555 return 0;
2556}
2557
492d8642
TH
2558static int vcpu_post_run_fault_in_sie(struct kvm_vcpu *vcpu)
2559{
56317920
DH
2560 struct kvm_s390_pgm_info pgm_info = {
2561 .code = PGM_ADDRESSING,
2562 };
2563 u8 opcode, ilen;
492d8642
TH
2564 int rc;
2565
2566 VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction");
2567 trace_kvm_s390_sie_fault(vcpu);
2568
2569 /*
2570 * We want to inject an addressing exception, which is defined as a
2571 * suppressing or terminating exception. However, since we came here
2572 * by a DAT access exception, the PSW still points to the faulting
2573 * instruction since DAT exceptions are nullifying. So we've got
2574 * to look up the current opcode to get the length of the instruction
2575 * to be able to forward the PSW.
2576 */
65977322 2577 rc = read_guest_instr(vcpu, &opcode, 1);
56317920 2578 ilen = insn_length(opcode);
9b0d721a
DH
2579 if (rc < 0) {
2580 return rc;
2581 } else if (rc) {
2582 /* Instruction-Fetching Exceptions - we can't detect the ilen.
2583 * Forward by arbitrary ilc, injection will take care of
2584 * nullification if necessary.
2585 */
2586 pgm_info = vcpu->arch.pgm;
2587 ilen = 4;
2588 }
56317920
DH
2589 pgm_info.flags = ilen | KVM_S390_PGM_FLAGS_ILC_VALID;
2590 kvm_s390_forward_psw(vcpu, ilen);
2591 return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
492d8642
TH
2592}
2593
3fb4c40f
TH
2594static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
2595{
2b29a9fd
DD
2596 VCPU_EVENT(vcpu, 6, "exit sie icptcode %d",
2597 vcpu->arch.sie_block->icptcode);
2598 trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode);
2599
27291e21
DH
2600 if (guestdbg_enabled(vcpu))
2601 kvm_s390_restore_guest_per_regs(vcpu);
2602
7ec7c8c7
CB
2603 vcpu->run->s.regs.gprs[14] = vcpu->arch.sie_block->gg14;
2604 vcpu->run->s.regs.gprs[15] = vcpu->arch.sie_block->gg15;
71f116bf
DH
2605
2606 if (vcpu->arch.sie_block->icptcode > 0) {
2607 int rc = kvm_handle_sie_intercept(vcpu);
2608
2609 if (rc != -EOPNOTSUPP)
2610 return rc;
2611 vcpu->run->exit_reason = KVM_EXIT_S390_SIEIC;
2612 vcpu->run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode;
2613 vcpu->run->s390_sieic.ipa = vcpu->arch.sie_block->ipa;
2614 vcpu->run->s390_sieic.ipb = vcpu->arch.sie_block->ipb;
2615 return -EREMOTE;
2616 } else if (exit_reason != -EFAULT) {
2617 vcpu->stat.exit_null++;
2618 return 0;
210b1607
TH
2619 } else if (kvm_is_ucontrol(vcpu->kvm)) {
2620 vcpu->run->exit_reason = KVM_EXIT_S390_UCONTROL;
2621 vcpu->run->s390_ucontrol.trans_exc_code =
2622 current->thread.gmap_addr;
2623 vcpu->run->s390_ucontrol.pgm_code = 0x10;
71f116bf 2624 return -EREMOTE;
24eb3a82 2625 } else if (current->thread.gmap_pfault) {
3c038e6b 2626 trace_kvm_s390_major_guest_pfault(vcpu);
24eb3a82 2627 current->thread.gmap_pfault = 0;
71f116bf
DH
2628 if (kvm_arch_setup_async_pf(vcpu))
2629 return 0;
2630 return kvm_arch_fault_in_page(vcpu, current->thread.gmap_addr, 1);
a76ccff6 2631 }
71f116bf 2632 return vcpu_post_run_fault_in_sie(vcpu);
3fb4c40f
TH
2633}
2634
2635static int __vcpu_run(struct kvm_vcpu *vcpu)
2636{
2637 int rc, exit_reason;
2638
800c1065
TH
2639 /*
2640 * We try to hold kvm->srcu during most of vcpu_run (except when run-
2641 * ning the guest), so that memslots (and other stuff) are protected
2642 */
2643 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
2644
a76ccff6
TH
2645 do {
2646 rc = vcpu_pre_run(vcpu);
2647 if (rc)
2648 break;
3fb4c40f 2649
800c1065 2650 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
a76ccff6
TH
2651 /*
2652 * As PF_VCPU will be used in fault handler, between
2653 * guest_enter and guest_exit should be no uaccess.
2654 */
0097d12e 2655 local_irq_disable();
6edaa530 2656 guest_enter_irqoff();
db0758b2 2657 __disable_cpu_timer_accounting(vcpu);
0097d12e 2658 local_irq_enable();
a76ccff6
TH
2659 exit_reason = sie64a(vcpu->arch.sie_block,
2660 vcpu->run->s.regs.gprs);
0097d12e 2661 local_irq_disable();
db0758b2 2662 __enable_cpu_timer_accounting(vcpu);
6edaa530 2663 guest_exit_irqoff();
0097d12e 2664 local_irq_enable();
800c1065 2665 vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
a76ccff6
TH
2666
2667 rc = vcpu_post_run(vcpu, exit_reason);
27291e21 2668 } while (!signal_pending(current) && !guestdbg_exit_pending(vcpu) && !rc);
3fb4c40f 2669
800c1065 2670 srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
e168bf8d 2671 return rc;
b0c632db
HC
2672}
2673
b028ee3e
DH
2674static void sync_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
2675{
2676 vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask;
2677 vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr;
2678 if (kvm_run->kvm_dirty_regs & KVM_SYNC_PREFIX)
2679 kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix);
2680 if (kvm_run->kvm_dirty_regs & KVM_SYNC_CRS) {
2681 memcpy(&vcpu->arch.sie_block->gcr, &kvm_run->s.regs.crs, 128);
d3d692c8
DH
2682 /* some control register changes require a tlb flush */
2683 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
b028ee3e
DH
2684 }
2685 if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
4287f247 2686 kvm_s390_set_cpu_timer(vcpu, kvm_run->s.regs.cputm);
b028ee3e
DH
2687 vcpu->arch.sie_block->ckc = kvm_run->s.regs.ckc;
2688 vcpu->arch.sie_block->todpr = kvm_run->s.regs.todpr;
2689 vcpu->arch.sie_block->pp = kvm_run->s.regs.pp;
2690 vcpu->arch.sie_block->gbea = kvm_run->s.regs.gbea;
2691 }
2692 if (kvm_run->kvm_dirty_regs & KVM_SYNC_PFAULT) {
2693 vcpu->arch.pfault_token = kvm_run->s.regs.pft;
2694 vcpu->arch.pfault_select = kvm_run->s.regs.pfs;
2695 vcpu->arch.pfault_compare = kvm_run->s.regs.pfc;
9fbd8082
DH
2696 if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
2697 kvm_clear_async_pf_completion_queue(vcpu);
b028ee3e
DH
2698 }
2699 kvm_run->kvm_dirty_regs = 0;
2700}
2701
2702static void store_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
2703{
2704 kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask;
2705 kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr;
2706 kvm_run->s.regs.prefix = kvm_s390_get_prefix(vcpu);
2707 memcpy(&kvm_run->s.regs.crs, &vcpu->arch.sie_block->gcr, 128);
4287f247 2708 kvm_run->s.regs.cputm = kvm_s390_get_cpu_timer(vcpu);
b028ee3e
DH
2709 kvm_run->s.regs.ckc = vcpu->arch.sie_block->ckc;
2710 kvm_run->s.regs.todpr = vcpu->arch.sie_block->todpr;
2711 kvm_run->s.regs.pp = vcpu->arch.sie_block->pp;
2712 kvm_run->s.regs.gbea = vcpu->arch.sie_block->gbea;
2713 kvm_run->s.regs.pft = vcpu->arch.pfault_token;
2714 kvm_run->s.regs.pfs = vcpu->arch.pfault_select;
2715 kvm_run->s.regs.pfc = vcpu->arch.pfault_compare;
2716}
2717
b0c632db
HC
2718int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
2719{
8f2abe6a 2720 int rc;
b0c632db
HC
2721 sigset_t sigsaved;
2722
27291e21
DH
2723 if (guestdbg_exit_pending(vcpu)) {
2724 kvm_s390_prepare_debug_exit(vcpu);
2725 return 0;
2726 }
2727
b0c632db
HC
2728 if (vcpu->sigset_active)
2729 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
2730
6352e4d2
DH
2731 if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm)) {
2732 kvm_s390_vcpu_start(vcpu);
2733 } else if (is_vcpu_stopped(vcpu)) {
ea2cdd27 2734 pr_err_ratelimited("can't run stopped vcpu %d\n",
6352e4d2
DH
2735 vcpu->vcpu_id);
2736 return -EINVAL;
2737 }
b0c632db 2738
b028ee3e 2739 sync_regs(vcpu, kvm_run);
db0758b2 2740 enable_cpu_timer_accounting(vcpu);
d7b0b5eb 2741
dab4079d 2742 might_fault();
a76ccff6 2743 rc = __vcpu_run(vcpu);
9ace903d 2744
b1d16c49
CE
2745 if (signal_pending(current) && !rc) {
2746 kvm_run->exit_reason = KVM_EXIT_INTR;
8f2abe6a 2747 rc = -EINTR;
b1d16c49 2748 }
8f2abe6a 2749
27291e21
DH
2750 if (guestdbg_exit_pending(vcpu) && !rc) {
2751 kvm_s390_prepare_debug_exit(vcpu);
2752 rc = 0;
2753 }
2754
8f2abe6a 2755 if (rc == -EREMOTE) {
71f116bf 2756 /* userspace support is needed, kvm_run has been prepared */
8f2abe6a
CB
2757 rc = 0;
2758 }
b0c632db 2759
db0758b2 2760 disable_cpu_timer_accounting(vcpu);
b028ee3e 2761 store_regs(vcpu, kvm_run);
d7b0b5eb 2762
b0c632db
HC
2763 if (vcpu->sigset_active)
2764 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
2765
b0c632db 2766 vcpu->stat.exit_userspace++;
7e8e6ab4 2767 return rc;
b0c632db
HC
2768}
2769
b0c632db
HC
2770/*
2771 * store status at address
2772 * we use have two special cases:
2773 * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
2774 * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
2775 */
d0bce605 2776int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
b0c632db 2777{
092670cd 2778 unsigned char archmode = 1;
9abc2a08 2779 freg_t fprs[NUM_FPRS];
fda902cb 2780 unsigned int px;
4287f247 2781 u64 clkcomp, cputm;
d0bce605 2782 int rc;
b0c632db 2783
d9a3a09a 2784 px = kvm_s390_get_prefix(vcpu);
d0bce605
HC
2785 if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
2786 if (write_guest_abs(vcpu, 163, &archmode, 1))
b0c632db 2787 return -EFAULT;
d9a3a09a 2788 gpa = 0;
d0bce605
HC
2789 } else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
2790 if (write_guest_real(vcpu, 163, &archmode, 1))
b0c632db 2791 return -EFAULT;
d9a3a09a
MS
2792 gpa = px;
2793 } else
2794 gpa -= __LC_FPREGS_SAVE_AREA;
9abc2a08
DH
2795
2796 /* manually convert vector registers if necessary */
2797 if (MACHINE_HAS_VX) {
9522b37f 2798 convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
9abc2a08
DH
2799 rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
2800 fprs, 128);
2801 } else {
2802 rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
6fd8e67d 2803 vcpu->run->s.regs.fprs, 128);
9abc2a08 2804 }
d9a3a09a 2805 rc |= write_guest_abs(vcpu, gpa + __LC_GPREGS_SAVE_AREA,
d0bce605 2806 vcpu->run->s.regs.gprs, 128);
d9a3a09a 2807 rc |= write_guest_abs(vcpu, gpa + __LC_PSW_SAVE_AREA,
d0bce605 2808 &vcpu->arch.sie_block->gpsw, 16);
d9a3a09a 2809 rc |= write_guest_abs(vcpu, gpa + __LC_PREFIX_SAVE_AREA,
fda902cb 2810 &px, 4);
d9a3a09a 2811 rc |= write_guest_abs(vcpu, gpa + __LC_FP_CREG_SAVE_AREA,
9abc2a08 2812 &vcpu->run->s.regs.fpc, 4);
d9a3a09a 2813 rc |= write_guest_abs(vcpu, gpa + __LC_TOD_PROGREG_SAVE_AREA,
d0bce605 2814 &vcpu->arch.sie_block->todpr, 4);
4287f247 2815 cputm = kvm_s390_get_cpu_timer(vcpu);
d9a3a09a 2816 rc |= write_guest_abs(vcpu, gpa + __LC_CPU_TIMER_SAVE_AREA,
4287f247 2817 &cputm, 8);
178bd789 2818 clkcomp = vcpu->arch.sie_block->ckc >> 8;
d9a3a09a 2819 rc |= write_guest_abs(vcpu, gpa + __LC_CLOCK_COMP_SAVE_AREA,
d0bce605 2820 &clkcomp, 8);
d9a3a09a 2821 rc |= write_guest_abs(vcpu, gpa + __LC_AREGS_SAVE_AREA,
d0bce605 2822 &vcpu->run->s.regs.acrs, 64);
d9a3a09a 2823 rc |= write_guest_abs(vcpu, gpa + __LC_CREGS_SAVE_AREA,
d0bce605
HC
2824 &vcpu->arch.sie_block->gcr, 128);
2825 return rc ? -EFAULT : 0;
b0c632db
HC
2826}
2827
e879892c
TH
2828int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
2829{
2830 /*
2831 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
2832 * copying in vcpu load/put. Lets update our copies before we save
2833 * it into the save area
2834 */
d0164ee2 2835 save_fpu_regs();
9abc2a08 2836 vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
e879892c
TH
2837 save_access_regs(vcpu->run->s.regs.acrs);
2838
2839 return kvm_s390_store_status_unloaded(vcpu, addr);
2840}
2841
bc17de7c
EF
2842/*
2843 * store additional status at address
2844 */
2845int kvm_s390_store_adtl_status_unloaded(struct kvm_vcpu *vcpu,
2846 unsigned long gpa)
2847{
2848 /* Only bits 0-53 are used for address formation */
2849 if (!(gpa & ~0x3ff))
2850 return 0;
2851
2852 return write_guest_abs(vcpu, gpa & ~0x3ff,
2853 (void *)&vcpu->run->s.regs.vrs, 512);
2854}
2855
2856int kvm_s390_vcpu_store_adtl_status(struct kvm_vcpu *vcpu, unsigned long addr)
2857{
2858 if (!test_kvm_facility(vcpu->kvm, 129))
2859 return 0;
2860
2861 /*
2862 * The guest VXRS are in the host VXRs due to the lazy
9977e886
HB
2863 * copying in vcpu load/put. We can simply call save_fpu_regs()
2864 * to save the current register state because we are in the
2865 * middle of a load/put cycle.
2866 *
2867 * Let's update our copies before we save it into the save area.
bc17de7c 2868 */
d0164ee2 2869 save_fpu_regs();
bc17de7c
EF
2870
2871 return kvm_s390_store_adtl_status_unloaded(vcpu, addr);
2872}
2873
8ad35755
DH
2874static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
2875{
2876 kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu);
8e236546 2877 kvm_s390_sync_request(KVM_REQ_DISABLE_IBS, vcpu);
8ad35755
DH
2878}
2879
2880static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
2881{
2882 unsigned int i;
2883 struct kvm_vcpu *vcpu;
2884
2885 kvm_for_each_vcpu(i, vcpu, kvm) {
2886 __disable_ibs_on_vcpu(vcpu);
2887 }
2888}
2889
2890static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
2891{
09a400e7
DH
2892 if (!sclp.has_ibs)
2893 return;
8ad35755 2894 kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu);
8e236546 2895 kvm_s390_sync_request(KVM_REQ_ENABLE_IBS, vcpu);
8ad35755
DH
2896}
2897
6852d7b6
DH
2898void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
2899{
8ad35755
DH
2900 int i, online_vcpus, started_vcpus = 0;
2901
2902 if (!is_vcpu_stopped(vcpu))
2903 return;
2904
6852d7b6 2905 trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
8ad35755 2906 /* Only one cpu at a time may enter/leave the STOPPED state. */
433b9ee4 2907 spin_lock(&vcpu->kvm->arch.start_stop_lock);
8ad35755
DH
2908 online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
2909
2910 for (i = 0; i < online_vcpus; i++) {
2911 if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
2912 started_vcpus++;
2913 }
2914
2915 if (started_vcpus == 0) {
2916 /* we're the only active VCPU -> speed it up */
2917 __enable_ibs_on_vcpu(vcpu);
2918 } else if (started_vcpus == 1) {
2919 /*
2920 * As we are starting a second VCPU, we have to disable
2921 * the IBS facility on all VCPUs to remove potentially
2922 * oustanding ENABLE requests.
2923 */
2924 __disable_ibs_on_all_vcpus(vcpu->kvm);
2925 }
2926
805de8f4 2927 atomic_andnot(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
8ad35755
DH
2928 /*
2929 * Another VCPU might have used IBS while we were offline.
2930 * Let's play safe and flush the VCPU at startup.
2931 */
d3d692c8 2932 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
433b9ee4 2933 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
8ad35755 2934 return;
6852d7b6
DH
2935}
2936
2937void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
2938{
8ad35755
DH
2939 int i, online_vcpus, started_vcpus = 0;
2940 struct kvm_vcpu *started_vcpu = NULL;
2941
2942 if (is_vcpu_stopped(vcpu))
2943 return;
2944
6852d7b6 2945 trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
8ad35755 2946 /* Only one cpu at a time may enter/leave the STOPPED state. */
433b9ee4 2947 spin_lock(&vcpu->kvm->arch.start_stop_lock);
8ad35755
DH
2948 online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
2949
32f5ff63 2950 /* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
6cddd432 2951 kvm_s390_clear_stop_irq(vcpu);
32f5ff63 2952
805de8f4 2953 atomic_or(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
8ad35755
DH
2954 __disable_ibs_on_vcpu(vcpu);
2955
2956 for (i = 0; i < online_vcpus; i++) {
2957 if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
2958 started_vcpus++;
2959 started_vcpu = vcpu->kvm->vcpus[i];
2960 }
2961 }
2962
2963 if (started_vcpus == 1) {
2964 /*
2965 * As we only have one VCPU left, we want to enable the
2966 * IBS facility for that VCPU to speed it up.
2967 */
2968 __enable_ibs_on_vcpu(started_vcpu);
2969 }
2970
433b9ee4 2971 spin_unlock(&vcpu->kvm->arch.start_stop_lock);
8ad35755 2972 return;
6852d7b6
DH
2973}
2974
d6712df9
CH
2975static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
2976 struct kvm_enable_cap *cap)
2977{
2978 int r;
2979
2980 if (cap->flags)
2981 return -EINVAL;
2982
2983 switch (cap->cap) {
fa6b7fe9
CH
2984 case KVM_CAP_S390_CSS_SUPPORT:
2985 if (!vcpu->kvm->arch.css_support) {
2986 vcpu->kvm->arch.css_support = 1;
c92ea7b9 2987 VM_EVENT(vcpu->kvm, 3, "%s", "ENABLE: CSS support");
fa6b7fe9
CH
2988 trace_kvm_s390_enable_css(vcpu->kvm);
2989 }
2990 r = 0;
2991 break;
d6712df9
CH
2992 default:
2993 r = -EINVAL;
2994 break;
2995 }
2996 return r;
2997}
2998
41408c28
TH
2999static long kvm_s390_guest_mem_op(struct kvm_vcpu *vcpu,
3000 struct kvm_s390_mem_op *mop)
3001{
3002 void __user *uaddr = (void __user *)mop->buf;
3003 void *tmpbuf = NULL;
3004 int r, srcu_idx;
3005 const u64 supported_flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION
3006 | KVM_S390_MEMOP_F_CHECK_ONLY;
3007
3008 if (mop->flags & ~supported_flags)
3009 return -EINVAL;
3010
3011 if (mop->size > MEM_OP_MAX_SIZE)
3012 return -E2BIG;
3013
3014 if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
3015 tmpbuf = vmalloc(mop->size);
3016 if (!tmpbuf)
3017 return -ENOMEM;
3018 }
3019
3020 srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
3021
3022 switch (mop->op) {
3023 case KVM_S390_MEMOP_LOGICAL_READ:
3024 if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
92c96321
DH
3025 r = check_gva_range(vcpu, mop->gaddr, mop->ar,
3026 mop->size, GACC_FETCH);
41408c28
TH
3027 break;
3028 }
3029 r = read_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
3030 if (r == 0) {
3031 if (copy_to_user(uaddr, tmpbuf, mop->size))
3032 r = -EFAULT;
3033 }
3034 break;
3035 case KVM_S390_MEMOP_LOGICAL_WRITE:
3036 if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
92c96321
DH
3037 r = check_gva_range(vcpu, mop->gaddr, mop->ar,
3038 mop->size, GACC_STORE);
41408c28
TH
3039 break;
3040 }
3041 if (copy_from_user(tmpbuf, uaddr, mop->size)) {
3042 r = -EFAULT;
3043 break;
3044 }
3045 r = write_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
3046 break;
3047 default:
3048 r = -EINVAL;
3049 }
3050
3051 srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
3052
3053 if (r > 0 && (mop->flags & KVM_S390_MEMOP_F_INJECT_EXCEPTION) != 0)
3054 kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
3055
3056 vfree(tmpbuf);
3057 return r;
3058}
3059
b0c632db
HC
3060long kvm_arch_vcpu_ioctl(struct file *filp,
3061 unsigned int ioctl, unsigned long arg)
3062{
3063 struct kvm_vcpu *vcpu = filp->private_data;
3064 void __user *argp = (void __user *)arg;
800c1065 3065 int idx;
bc923cc9 3066 long r;
b0c632db 3067
93736624 3068 switch (ioctl) {
47b43c52
JF
3069 case KVM_S390_IRQ: {
3070 struct kvm_s390_irq s390irq;
3071
3072 r = -EFAULT;
3073 if (copy_from_user(&s390irq, argp, sizeof(s390irq)))
3074 break;
3075 r = kvm_s390_inject_vcpu(vcpu, &s390irq);
3076 break;
3077 }
93736624 3078 case KVM_S390_INTERRUPT: {
ba5c1e9b 3079 struct kvm_s390_interrupt s390int;
383d0b05 3080 struct kvm_s390_irq s390irq;
ba5c1e9b 3081
93736624 3082 r = -EFAULT;
ba5c1e9b 3083 if (copy_from_user(&s390int, argp, sizeof(s390int)))
93736624 3084 break;
383d0b05
JF
3085 if (s390int_to_s390irq(&s390int, &s390irq))
3086 return -EINVAL;
3087 r = kvm_s390_inject_vcpu(vcpu, &s390irq);
93736624 3088 break;
ba5c1e9b 3089 }
b0c632db 3090 case KVM_S390_STORE_STATUS:
800c1065 3091 idx = srcu_read_lock(&vcpu->kvm->srcu);
bc923cc9 3092 r = kvm_s390_vcpu_store_status(vcpu, arg);
800c1065 3093 srcu_read_unlock(&vcpu->kvm->srcu, idx);
bc923cc9 3094 break;
b0c632db
HC
3095 case KVM_S390_SET_INITIAL_PSW: {
3096 psw_t psw;
3097
bc923cc9 3098 r = -EFAULT;
b0c632db 3099 if (copy_from_user(&psw, argp, sizeof(psw)))
bc923cc9
AK
3100 break;
3101 r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw);
3102 break;
b0c632db
HC
3103 }
3104 case KVM_S390_INITIAL_RESET:
bc923cc9
AK
3105 r = kvm_arch_vcpu_ioctl_initial_reset(vcpu);
3106 break;
14eebd91
CO
3107 case KVM_SET_ONE_REG:
3108 case KVM_GET_ONE_REG: {
3109 struct kvm_one_reg reg;
3110 r = -EFAULT;
3111 if (copy_from_user(&reg, argp, sizeof(reg)))
3112 break;
3113 if (ioctl == KVM_SET_ONE_REG)
3114 r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, &reg);
3115 else
3116 r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, &reg);
3117 break;
3118 }
27e0393f
CO
3119#ifdef CONFIG_KVM_S390_UCONTROL
3120 case KVM_S390_UCAS_MAP: {
3121 struct kvm_s390_ucas_mapping ucasmap;
3122
3123 if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
3124 r = -EFAULT;
3125 break;
3126 }
3127
3128 if (!kvm_is_ucontrol(vcpu->kvm)) {
3129 r = -EINVAL;
3130 break;
3131 }
3132
3133 r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr,
3134 ucasmap.vcpu_addr, ucasmap.length);
3135 break;
3136 }
3137 case KVM_S390_UCAS_UNMAP: {
3138 struct kvm_s390_ucas_mapping ucasmap;
3139
3140 if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
3141 r = -EFAULT;
3142 break;
3143 }
3144
3145 if (!kvm_is_ucontrol(vcpu->kvm)) {
3146 r = -EINVAL;
3147 break;
3148 }
3149
3150 r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr,
3151 ucasmap.length);
3152 break;
3153 }
3154#endif
ccc7910f 3155 case KVM_S390_VCPU_FAULT: {
527e30b4 3156 r = gmap_fault(vcpu->arch.gmap, arg, 0);
ccc7910f
CO
3157 break;
3158 }
d6712df9
CH
3159 case KVM_ENABLE_CAP:
3160 {
3161 struct kvm_enable_cap cap;
3162 r = -EFAULT;
3163 if (copy_from_user(&cap, argp, sizeof(cap)))
3164 break;
3165 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
3166 break;
3167 }
41408c28
TH
3168 case KVM_S390_MEM_OP: {
3169 struct kvm_s390_mem_op mem_op;
3170
3171 if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
3172 r = kvm_s390_guest_mem_op(vcpu, &mem_op);
3173 else
3174 r = -EFAULT;
3175 break;
3176 }
816c7667
JF
3177 case KVM_S390_SET_IRQ_STATE: {
3178 struct kvm_s390_irq_state irq_state;
3179
3180 r = -EFAULT;
3181 if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
3182 break;
3183 if (irq_state.len > VCPU_IRQS_MAX_BUF ||
3184 irq_state.len == 0 ||
3185 irq_state.len % sizeof(struct kvm_s390_irq) > 0) {
3186 r = -EINVAL;
3187 break;
3188 }
3189 r = kvm_s390_set_irq_state(vcpu,
3190 (void __user *) irq_state.buf,
3191 irq_state.len);
3192 break;
3193 }
3194 case KVM_S390_GET_IRQ_STATE: {
3195 struct kvm_s390_irq_state irq_state;
3196
3197 r = -EFAULT;
3198 if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
3199 break;
3200 if (irq_state.len == 0) {
3201 r = -EINVAL;
3202 break;
3203 }
3204 r = kvm_s390_get_irq_state(vcpu,
3205 (__u8 __user *) irq_state.buf,
3206 irq_state.len);
3207 break;
3208 }
b0c632db 3209 default:
3e6afcf1 3210 r = -ENOTTY;
b0c632db 3211 }
bc923cc9 3212 return r;
b0c632db
HC
3213}
3214
5b1c1493
CO
3215int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
3216{
3217#ifdef CONFIG_KVM_S390_UCONTROL
3218 if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET)
3219 && (kvm_is_ucontrol(vcpu->kvm))) {
3220 vmf->page = virt_to_page(vcpu->arch.sie_block);
3221 get_page(vmf->page);
3222 return 0;
3223 }
3224#endif
3225 return VM_FAULT_SIGBUS;
3226}
3227
5587027c
AK
3228int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
3229 unsigned long npages)
db3fe4eb
TY
3230{
3231 return 0;
3232}
3233
b0c632db 3234/* Section: memory related */
f7784b8e
MT
3235int kvm_arch_prepare_memory_region(struct kvm *kvm,
3236 struct kvm_memory_slot *memslot,
09170a49 3237 const struct kvm_userspace_memory_region *mem,
7b6195a9 3238 enum kvm_mr_change change)
b0c632db 3239{
dd2887e7
NW
3240 /* A few sanity checks. We can have memory slots which have to be
3241 located/ended at a segment boundary (1MB). The memory in userland is
3242 ok to be fragmented into various different vmas. It is okay to mmap()
3243 and munmap() stuff in this slot after doing this call at any time */
b0c632db 3244
598841ca 3245 if (mem->userspace_addr & 0xffffful)
b0c632db
HC
3246 return -EINVAL;
3247
598841ca 3248 if (mem->memory_size & 0xffffful)
b0c632db
HC
3249 return -EINVAL;
3250
a3a92c31
DD
3251 if (mem->guest_phys_addr + mem->memory_size > kvm->arch.mem_limit)
3252 return -EINVAL;
3253
f7784b8e
MT
3254 return 0;
3255}
3256
3257void kvm_arch_commit_memory_region(struct kvm *kvm,
09170a49 3258 const struct kvm_userspace_memory_region *mem,
8482644a 3259 const struct kvm_memory_slot *old,
f36f3f28 3260 const struct kvm_memory_slot *new,
8482644a 3261 enum kvm_mr_change change)
f7784b8e 3262{
f7850c92 3263 int rc;
f7784b8e 3264
2cef4deb
CB
3265 /* If the basics of the memslot do not change, we do not want
3266 * to update the gmap. Every update causes several unnecessary
3267 * segment translation exceptions. This is usually handled just
3268 * fine by the normal fault handler + gmap, but it will also
3269 * cause faults on the prefix page of running guest CPUs.
3270 */
3271 if (old->userspace_addr == mem->userspace_addr &&
3272 old->base_gfn * PAGE_SIZE == mem->guest_phys_addr &&
3273 old->npages * PAGE_SIZE == mem->memory_size)
3274 return;
598841ca
CO
3275
3276 rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr,
3277 mem->guest_phys_addr, mem->memory_size);
3278 if (rc)
ea2cdd27 3279 pr_warn("failed to commit memory region\n");
598841ca 3280 return;
b0c632db
HC
3281}
3282
60a37709
AY
3283static inline unsigned long nonhyp_mask(int i)
3284{
3285 unsigned int nonhyp_fai = (sclp.hmfai << i * 2) >> 30;
3286
3287 return 0x0000ffffffffffffUL >> (nonhyp_fai << 4);
3288}
3289
3491caf2
CB
3290void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu)
3291{
3292 vcpu->valid_wakeup = false;
3293}
3294
b0c632db
HC
3295static int __init kvm_s390_init(void)
3296{
60a37709
AY
3297 int i;
3298
07197fd0
DH
3299 if (!sclp.has_sief2) {
3300 pr_info("SIE not available\n");
3301 return -ENODEV;
3302 }
3303
60a37709
AY
3304 for (i = 0; i < 16; i++)
3305 kvm_s390_fac_list_mask[i] |=
3306 S390_lowcore.stfle_fac_list[i] & nonhyp_mask(i);
3307
9d8d5786 3308 return kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
b0c632db
HC
3309}
3310
3311static void __exit kvm_s390_exit(void)
3312{
3313 kvm_exit();
3314}
3315
3316module_init(kvm_s390_init);
3317module_exit(kvm_s390_exit);
566af940
CH
3318
3319/*
3320 * Enable autoloading of the kvm module.
3321 * Note that we add the module alias here instead of virt/kvm/kvm_main.c
3322 * since x86 takes a different approach.
3323 */
3324#include <linux/miscdevice.h>
3325MODULE_ALIAS_MISCDEV(KVM_MINOR);
3326MODULE_ALIAS("devname:kvm");