]>
Commit | Line | Data |
---|---|---|
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> |
b0c632db HC |
14 | */ |
15 | ||
16 | #include <linux/compiler.h> | |
17 | #include <linux/err.h> | |
18 | #include <linux/fs.h> | |
ca872302 | 19 | #include <linux/hrtimer.h> |
b0c632db HC |
20 | #include <linux/init.h> |
21 | #include <linux/kvm.h> | |
22 | #include <linux/kvm_host.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/slab.h> | |
ba5c1e9b | 25 | #include <linux/timer.h> |
cbb870c8 | 26 | #include <asm/asm-offsets.h> |
b0c632db HC |
27 | #include <asm/lowcore.h> |
28 | #include <asm/pgtable.h> | |
f5daba1d | 29 | #include <asm/nmi.h> |
a0616cde | 30 | #include <asm/switch_to.h> |
78c4b59f | 31 | #include <asm/facility.h> |
1526bf9c | 32 | #include <asm/sclp.h> |
8f2abe6a | 33 | #include "kvm-s390.h" |
b0c632db HC |
34 | #include "gaccess.h" |
35 | ||
5786fffa CH |
36 | #define CREATE_TRACE_POINTS |
37 | #include "trace.h" | |
ade38c31 | 38 | #include "trace-s390.h" |
5786fffa | 39 | |
b0c632db HC |
40 | #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU |
41 | ||
42 | struct kvm_stats_debugfs_item debugfs_entries[] = { | |
43 | { "userspace_handled", VCPU_STAT(exit_userspace) }, | |
0eaeafa1 | 44 | { "exit_null", VCPU_STAT(exit_null) }, |
8f2abe6a CB |
45 | { "exit_validity", VCPU_STAT(exit_validity) }, |
46 | { "exit_stop_request", VCPU_STAT(exit_stop_request) }, | |
47 | { "exit_external_request", VCPU_STAT(exit_external_request) }, | |
48 | { "exit_external_interrupt", VCPU_STAT(exit_external_interrupt) }, | |
ba5c1e9b CO |
49 | { "exit_instruction", VCPU_STAT(exit_instruction) }, |
50 | { "exit_program_interruption", VCPU_STAT(exit_program_interruption) }, | |
51 | { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) }, | |
f5e10b09 | 52 | { "instruction_lctlg", VCPU_STAT(instruction_lctlg) }, |
ba5c1e9b CO |
53 | { "instruction_lctl", VCPU_STAT(instruction_lctl) }, |
54 | { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) }, | |
7697e71f | 55 | { "deliver_external_call", VCPU_STAT(deliver_external_call) }, |
ba5c1e9b CO |
56 | { "deliver_service_signal", VCPU_STAT(deliver_service_signal) }, |
57 | { "deliver_virtio_interrupt", VCPU_STAT(deliver_virtio_interrupt) }, | |
58 | { "deliver_stop_signal", VCPU_STAT(deliver_stop_signal) }, | |
59 | { "deliver_prefix_signal", VCPU_STAT(deliver_prefix_signal) }, | |
60 | { "deliver_restart_signal", VCPU_STAT(deliver_restart_signal) }, | |
61 | { "deliver_program_interruption", VCPU_STAT(deliver_program_int) }, | |
62 | { "exit_wait_state", VCPU_STAT(exit_wait_state) }, | |
69d0d3a3 | 63 | { "instruction_pfmf", VCPU_STAT(instruction_pfmf) }, |
453423dc CB |
64 | { "instruction_stidp", VCPU_STAT(instruction_stidp) }, |
65 | { "instruction_spx", VCPU_STAT(instruction_spx) }, | |
66 | { "instruction_stpx", VCPU_STAT(instruction_stpx) }, | |
67 | { "instruction_stap", VCPU_STAT(instruction_stap) }, | |
68 | { "instruction_storage_key", VCPU_STAT(instruction_storage_key) }, | |
69 | { "instruction_stsch", VCPU_STAT(instruction_stsch) }, | |
70 | { "instruction_chsc", VCPU_STAT(instruction_chsc) }, | |
71 | { "instruction_stsi", VCPU_STAT(instruction_stsi) }, | |
72 | { "instruction_stfl", VCPU_STAT(instruction_stfl) }, | |
bb25b9ba | 73 | { "instruction_tprot", VCPU_STAT(instruction_tprot) }, |
5288fbf0 | 74 | { "instruction_sigp_sense", VCPU_STAT(instruction_sigp_sense) }, |
bd59d3a4 | 75 | { "instruction_sigp_sense_running", VCPU_STAT(instruction_sigp_sense_running) }, |
7697e71f | 76 | { "instruction_sigp_external_call", VCPU_STAT(instruction_sigp_external_call) }, |
5288fbf0 CB |
77 | { "instruction_sigp_emergency", VCPU_STAT(instruction_sigp_emergency) }, |
78 | { "instruction_sigp_stop", VCPU_STAT(instruction_sigp_stop) }, | |
79 | { "instruction_sigp_set_arch", VCPU_STAT(instruction_sigp_arch) }, | |
80 | { "instruction_sigp_set_prefix", VCPU_STAT(instruction_sigp_prefix) }, | |
81 | { "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart) }, | |
388186bc | 82 | { "diagnose_10", VCPU_STAT(diagnose_10) }, |
e28acfea | 83 | { "diagnose_44", VCPU_STAT(diagnose_44) }, |
41628d33 | 84 | { "diagnose_9c", VCPU_STAT(diagnose_9c) }, |
b0c632db HC |
85 | { NULL } |
86 | }; | |
87 | ||
78c4b59f | 88 | unsigned long *vfacilities; |
2c70fe44 | 89 | static struct gmap_notifier gmap_notifier; |
b0c632db | 90 | |
78c4b59f MM |
91 | /* test availability of vfacility */ |
92 | static inline int test_vfacility(unsigned long nr) | |
93 | { | |
94 | return __test_facility(nr, (void *) vfacilities); | |
95 | } | |
96 | ||
b0c632db | 97 | /* Section: not file related */ |
10474ae8 | 98 | int kvm_arch_hardware_enable(void *garbage) |
b0c632db HC |
99 | { |
100 | /* every s390 is virtualization enabled ;-) */ | |
10474ae8 | 101 | return 0; |
b0c632db HC |
102 | } |
103 | ||
104 | void kvm_arch_hardware_disable(void *garbage) | |
105 | { | |
106 | } | |
107 | ||
2c70fe44 CB |
108 | static void kvm_gmap_notifier(struct gmap *gmap, unsigned long address); |
109 | ||
b0c632db HC |
110 | int kvm_arch_hardware_setup(void) |
111 | { | |
2c70fe44 CB |
112 | gmap_notifier.notifier_call = kvm_gmap_notifier; |
113 | gmap_register_ipte_notifier(&gmap_notifier); | |
b0c632db HC |
114 | return 0; |
115 | } | |
116 | ||
117 | void kvm_arch_hardware_unsetup(void) | |
118 | { | |
2c70fe44 | 119 | gmap_unregister_ipte_notifier(&gmap_notifier); |
b0c632db HC |
120 | } |
121 | ||
122 | void kvm_arch_check_processor_compat(void *rtn) | |
123 | { | |
124 | } | |
125 | ||
126 | int kvm_arch_init(void *opaque) | |
127 | { | |
128 | return 0; | |
129 | } | |
130 | ||
131 | void kvm_arch_exit(void) | |
132 | { | |
133 | } | |
134 | ||
135 | /* Section: device related */ | |
136 | long kvm_arch_dev_ioctl(struct file *filp, | |
137 | unsigned int ioctl, unsigned long arg) | |
138 | { | |
139 | if (ioctl == KVM_S390_ENABLE_SIE) | |
140 | return s390_enable_sie(); | |
141 | return -EINVAL; | |
142 | } | |
143 | ||
144 | int kvm_dev_ioctl_check_extension(long ext) | |
145 | { | |
d7b0b5eb CO |
146 | int r; |
147 | ||
2bd0ac4e | 148 | switch (ext) { |
d7b0b5eb | 149 | case KVM_CAP_S390_PSW: |
b6cf8788 | 150 | case KVM_CAP_S390_GMAP: |
52e16b18 | 151 | case KVM_CAP_SYNC_MMU: |
1efd0f59 CO |
152 | #ifdef CONFIG_KVM_S390_UCONTROL |
153 | case KVM_CAP_S390_UCONTROL: | |
154 | #endif | |
3c038e6b | 155 | case KVM_CAP_ASYNC_PF: |
60b413c9 | 156 | case KVM_CAP_SYNC_REGS: |
14eebd91 | 157 | case KVM_CAP_ONE_REG: |
d6712df9 | 158 | case KVM_CAP_ENABLE_CAP: |
fa6b7fe9 | 159 | case KVM_CAP_S390_CSS_SUPPORT: |
10ccaa1e | 160 | case KVM_CAP_IOEVENTFD: |
c05c4186 | 161 | case KVM_CAP_DEVICE_CTRL: |
d7b0b5eb CO |
162 | r = 1; |
163 | break; | |
e726b1bd CB |
164 | case KVM_CAP_NR_VCPUS: |
165 | case KVM_CAP_MAX_VCPUS: | |
166 | r = KVM_MAX_VCPUS; | |
167 | break; | |
e1e2e605 NW |
168 | case KVM_CAP_NR_MEMSLOTS: |
169 | r = KVM_USER_MEM_SLOTS; | |
170 | break; | |
1526bf9c | 171 | case KVM_CAP_S390_COW: |
abf09bed | 172 | r = MACHINE_HAS_ESOP; |
1526bf9c | 173 | break; |
2bd0ac4e | 174 | default: |
d7b0b5eb | 175 | r = 0; |
2bd0ac4e | 176 | } |
d7b0b5eb | 177 | return r; |
b0c632db HC |
178 | } |
179 | ||
180 | /* Section: vm related */ | |
181 | /* | |
182 | * Get (and clear) the dirty memory log for a memory slot. | |
183 | */ | |
184 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, | |
185 | struct kvm_dirty_log *log) | |
186 | { | |
187 | return 0; | |
188 | } | |
189 | ||
190 | long kvm_arch_vm_ioctl(struct file *filp, | |
191 | unsigned int ioctl, unsigned long arg) | |
192 | { | |
193 | struct kvm *kvm = filp->private_data; | |
194 | void __user *argp = (void __user *)arg; | |
195 | int r; | |
196 | ||
197 | switch (ioctl) { | |
ba5c1e9b CO |
198 | case KVM_S390_INTERRUPT: { |
199 | struct kvm_s390_interrupt s390int; | |
200 | ||
201 | r = -EFAULT; | |
202 | if (copy_from_user(&s390int, argp, sizeof(s390int))) | |
203 | break; | |
204 | r = kvm_s390_inject_vm(kvm, &s390int); | |
205 | break; | |
206 | } | |
b0c632db | 207 | default: |
367e1319 | 208 | r = -ENOTTY; |
b0c632db HC |
209 | } |
210 | ||
211 | return r; | |
212 | } | |
213 | ||
e08b9637 | 214 | int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) |
b0c632db | 215 | { |
b0c632db HC |
216 | int rc; |
217 | char debug_name[16]; | |
218 | ||
e08b9637 CO |
219 | rc = -EINVAL; |
220 | #ifdef CONFIG_KVM_S390_UCONTROL | |
221 | if (type & ~KVM_VM_S390_UCONTROL) | |
222 | goto out_err; | |
223 | if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN))) | |
224 | goto out_err; | |
225 | #else | |
226 | if (type) | |
227 | goto out_err; | |
228 | #endif | |
229 | ||
b0c632db HC |
230 | rc = s390_enable_sie(); |
231 | if (rc) | |
d89f5eff | 232 | goto out_err; |
b0c632db | 233 | |
b290411a CO |
234 | rc = -ENOMEM; |
235 | ||
b0c632db HC |
236 | kvm->arch.sca = (struct sca_block *) get_zeroed_page(GFP_KERNEL); |
237 | if (!kvm->arch.sca) | |
d89f5eff | 238 | goto out_err; |
b0c632db HC |
239 | |
240 | sprintf(debug_name, "kvm-%u", current->pid); | |
241 | ||
242 | kvm->arch.dbf = debug_register(debug_name, 8, 2, 8 * sizeof(long)); | |
243 | if (!kvm->arch.dbf) | |
244 | goto out_nodbf; | |
245 | ||
ba5c1e9b CO |
246 | spin_lock_init(&kvm->arch.float_int.lock); |
247 | INIT_LIST_HEAD(&kvm->arch.float_int.list); | |
248 | ||
b0c632db HC |
249 | debug_register_view(kvm->arch.dbf, &debug_sprintf_view); |
250 | VM_EVENT(kvm, 3, "%s", "vm created"); | |
251 | ||
e08b9637 CO |
252 | if (type & KVM_VM_S390_UCONTROL) { |
253 | kvm->arch.gmap = NULL; | |
254 | } else { | |
255 | kvm->arch.gmap = gmap_alloc(current->mm); | |
256 | if (!kvm->arch.gmap) | |
257 | goto out_nogmap; | |
2c70fe44 | 258 | kvm->arch.gmap->private = kvm; |
24eb3a82 | 259 | kvm->arch.gmap->pfault_enabled = 0; |
e08b9637 | 260 | } |
fa6b7fe9 CH |
261 | |
262 | kvm->arch.css_support = 0; | |
263 | ||
d89f5eff | 264 | return 0; |
598841ca CO |
265 | out_nogmap: |
266 | debug_unregister(kvm->arch.dbf); | |
b0c632db HC |
267 | out_nodbf: |
268 | free_page((unsigned long)(kvm->arch.sca)); | |
d89f5eff JK |
269 | out_err: |
270 | return rc; | |
b0c632db HC |
271 | } |
272 | ||
d329c035 CB |
273 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) |
274 | { | |
275 | VCPU_EVENT(vcpu, 3, "%s", "free cpu"); | |
ade38c31 | 276 | trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id); |
3c038e6b | 277 | kvm_clear_async_pf_completion_queue(vcpu); |
58f9460b CO |
278 | if (!kvm_is_ucontrol(vcpu->kvm)) { |
279 | clear_bit(63 - vcpu->vcpu_id, | |
280 | (unsigned long *) &vcpu->kvm->arch.sca->mcn); | |
281 | if (vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sda == | |
282 | (__u64) vcpu->arch.sie_block) | |
283 | vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sda = 0; | |
284 | } | |
abf4a71e | 285 | smp_mb(); |
27e0393f CO |
286 | |
287 | if (kvm_is_ucontrol(vcpu->kvm)) | |
288 | gmap_free(vcpu->arch.gmap); | |
289 | ||
d329c035 | 290 | free_page((unsigned long)(vcpu->arch.sie_block)); |
6692cef3 | 291 | kvm_vcpu_uninit(vcpu); |
b110feaf | 292 | kmem_cache_free(kvm_vcpu_cache, vcpu); |
d329c035 CB |
293 | } |
294 | ||
295 | static void kvm_free_vcpus(struct kvm *kvm) | |
296 | { | |
297 | unsigned int i; | |
988a2cae | 298 | struct kvm_vcpu *vcpu; |
d329c035 | 299 | |
988a2cae GN |
300 | kvm_for_each_vcpu(i, vcpu, kvm) |
301 | kvm_arch_vcpu_destroy(vcpu); | |
302 | ||
303 | mutex_lock(&kvm->lock); | |
304 | for (i = 0; i < atomic_read(&kvm->online_vcpus); i++) | |
305 | kvm->vcpus[i] = NULL; | |
306 | ||
307 | atomic_set(&kvm->online_vcpus, 0); | |
308 | mutex_unlock(&kvm->lock); | |
d329c035 CB |
309 | } |
310 | ||
ad8ba2cd SY |
311 | void kvm_arch_sync_events(struct kvm *kvm) |
312 | { | |
313 | } | |
314 | ||
b0c632db HC |
315 | void kvm_arch_destroy_vm(struct kvm *kvm) |
316 | { | |
d329c035 | 317 | kvm_free_vcpus(kvm); |
b0c632db | 318 | free_page((unsigned long)(kvm->arch.sca)); |
d329c035 | 319 | debug_unregister(kvm->arch.dbf); |
27e0393f CO |
320 | if (!kvm_is_ucontrol(kvm)) |
321 | gmap_free(kvm->arch.gmap); | |
b0c632db HC |
322 | } |
323 | ||
324 | /* Section: vcpu related */ | |
325 | int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu) | |
326 | { | |
3c038e6b DD |
327 | vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID; |
328 | kvm_clear_async_pf_completion_queue(vcpu); | |
27e0393f CO |
329 | if (kvm_is_ucontrol(vcpu->kvm)) { |
330 | vcpu->arch.gmap = gmap_alloc(current->mm); | |
331 | if (!vcpu->arch.gmap) | |
332 | return -ENOMEM; | |
2c70fe44 | 333 | vcpu->arch.gmap->private = vcpu->kvm; |
27e0393f CO |
334 | return 0; |
335 | } | |
336 | ||
598841ca | 337 | vcpu->arch.gmap = vcpu->kvm->arch.gmap; |
59674c1a CB |
338 | vcpu->run->kvm_valid_regs = KVM_SYNC_PREFIX | |
339 | KVM_SYNC_GPRS | | |
9eed0735 CB |
340 | KVM_SYNC_ACRS | |
341 | KVM_SYNC_CRS; | |
b0c632db HC |
342 | return 0; |
343 | } | |
344 | ||
345 | void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) | |
346 | { | |
6692cef3 | 347 | /* Nothing todo */ |
b0c632db HC |
348 | } |
349 | ||
350 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) | |
351 | { | |
4725c860 MS |
352 | save_fp_ctl(&vcpu->arch.host_fpregs.fpc); |
353 | save_fp_regs(vcpu->arch.host_fpregs.fprs); | |
b0c632db | 354 | save_access_regs(vcpu->arch.host_acrs); |
4725c860 MS |
355 | restore_fp_ctl(&vcpu->arch.guest_fpregs.fpc); |
356 | restore_fp_regs(vcpu->arch.guest_fpregs.fprs); | |
59674c1a | 357 | restore_access_regs(vcpu->run->s.regs.acrs); |
480e5926 | 358 | gmap_enable(vcpu->arch.gmap); |
9e6dabef | 359 | atomic_set_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags); |
b0c632db HC |
360 | } |
361 | ||
362 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) | |
363 | { | |
9e6dabef | 364 | atomic_clear_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags); |
480e5926 | 365 | gmap_disable(vcpu->arch.gmap); |
4725c860 MS |
366 | save_fp_ctl(&vcpu->arch.guest_fpregs.fpc); |
367 | save_fp_regs(vcpu->arch.guest_fpregs.fprs); | |
59674c1a | 368 | save_access_regs(vcpu->run->s.regs.acrs); |
4725c860 MS |
369 | restore_fp_ctl(&vcpu->arch.host_fpregs.fpc); |
370 | restore_fp_regs(vcpu->arch.host_fpregs.fprs); | |
b0c632db HC |
371 | restore_access_regs(vcpu->arch.host_acrs); |
372 | } | |
373 | ||
374 | static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu) | |
375 | { | |
376 | /* this equals initial cpu reset in pop, but we don't switch to ESA */ | |
377 | vcpu->arch.sie_block->gpsw.mask = 0UL; | |
378 | vcpu->arch.sie_block->gpsw.addr = 0UL; | |
8d26cf7b | 379 | kvm_s390_set_prefix(vcpu, 0); |
b0c632db HC |
380 | vcpu->arch.sie_block->cputm = 0UL; |
381 | vcpu->arch.sie_block->ckc = 0UL; | |
382 | vcpu->arch.sie_block->todpr = 0; | |
383 | memset(vcpu->arch.sie_block->gcr, 0, 16 * sizeof(__u64)); | |
384 | vcpu->arch.sie_block->gcr[0] = 0xE0UL; | |
385 | vcpu->arch.sie_block->gcr[14] = 0xC2000000UL; | |
386 | vcpu->arch.guest_fpregs.fpc = 0; | |
387 | asm volatile("lfpc %0" : : "Q" (vcpu->arch.guest_fpregs.fpc)); | |
388 | vcpu->arch.sie_block->gbea = 1; | |
3c038e6b DD |
389 | vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID; |
390 | kvm_clear_async_pf_completion_queue(vcpu); | |
61bde82c | 391 | atomic_set_mask(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags); |
b0c632db HC |
392 | } |
393 | ||
42897d86 MT |
394 | int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu) |
395 | { | |
396 | return 0; | |
397 | } | |
398 | ||
b0c632db HC |
399 | int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu) |
400 | { | |
9e6dabef CH |
401 | atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH | |
402 | CPUSTAT_SM | | |
69d0d3a3 CB |
403 | CPUSTAT_STOPPED | |
404 | CPUSTAT_GED); | |
fc34531d | 405 | vcpu->arch.sie_block->ecb = 6; |
7feb6bb8 MM |
406 | if (test_vfacility(50) && test_vfacility(73)) |
407 | vcpu->arch.sie_block->ecb |= 0x10; | |
408 | ||
69d0d3a3 | 409 | vcpu->arch.sie_block->ecb2 = 8; |
b0c632db | 410 | vcpu->arch.sie_block->eca = 0xC1002001U; |
78c4b59f | 411 | vcpu->arch.sie_block->fac = (int) (long) vfacilities; |
ca872302 CB |
412 | hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS); |
413 | tasklet_init(&vcpu->arch.tasklet, kvm_s390_tasklet, | |
414 | (unsigned long) vcpu); | |
415 | vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup; | |
453423dc | 416 | get_cpu_id(&vcpu->arch.cpu_id); |
92e6ecf3 | 417 | vcpu->arch.cpu_id.version = 0xff; |
b0c632db HC |
418 | return 0; |
419 | } | |
420 | ||
421 | struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, | |
422 | unsigned int id) | |
423 | { | |
4d47555a | 424 | struct kvm_vcpu *vcpu; |
7feb6bb8 | 425 | struct sie_page *sie_page; |
4d47555a CO |
426 | int rc = -EINVAL; |
427 | ||
428 | if (id >= KVM_MAX_VCPUS) | |
429 | goto out; | |
430 | ||
431 | rc = -ENOMEM; | |
b0c632db | 432 | |
b110feaf | 433 | vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL); |
b0c632db | 434 | if (!vcpu) |
4d47555a | 435 | goto out; |
b0c632db | 436 | |
7feb6bb8 MM |
437 | sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL); |
438 | if (!sie_page) | |
b0c632db HC |
439 | goto out_free_cpu; |
440 | ||
7feb6bb8 MM |
441 | vcpu->arch.sie_block = &sie_page->sie_block; |
442 | vcpu->arch.sie_block->itdba = (unsigned long) &sie_page->itdb; | |
443 | ||
b0c632db | 444 | vcpu->arch.sie_block->icpua = id; |
58f9460b CO |
445 | if (!kvm_is_ucontrol(kvm)) { |
446 | if (!kvm->arch.sca) { | |
447 | WARN_ON_ONCE(1); | |
448 | goto out_free_cpu; | |
449 | } | |
450 | if (!kvm->arch.sca->cpu[id].sda) | |
451 | kvm->arch.sca->cpu[id].sda = | |
452 | (__u64) vcpu->arch.sie_block; | |
453 | vcpu->arch.sie_block->scaoh = | |
454 | (__u32)(((__u64)kvm->arch.sca) >> 32); | |
455 | vcpu->arch.sie_block->scaol = (__u32)(__u64)kvm->arch.sca; | |
456 | set_bit(63 - id, (unsigned long *) &kvm->arch.sca->mcn); | |
457 | } | |
b0c632db | 458 | |
ba5c1e9b CO |
459 | spin_lock_init(&vcpu->arch.local_int.lock); |
460 | INIT_LIST_HEAD(&vcpu->arch.local_int.list); | |
461 | vcpu->arch.local_int.float_int = &kvm->arch.float_int; | |
b037a4f3 | 462 | spin_lock(&kvm->arch.float_int.lock); |
ba5c1e9b | 463 | kvm->arch.float_int.local_int[id] = &vcpu->arch.local_int; |
d0321a24 | 464 | vcpu->arch.local_int.wq = &vcpu->wq; |
5288fbf0 | 465 | vcpu->arch.local_int.cpuflags = &vcpu->arch.sie_block->cpuflags; |
b037a4f3 | 466 | spin_unlock(&kvm->arch.float_int.lock); |
ba5c1e9b | 467 | |
b0c632db HC |
468 | rc = kvm_vcpu_init(vcpu, kvm, id); |
469 | if (rc) | |
7b06bf2f | 470 | goto out_free_sie_block; |
b0c632db HC |
471 | VM_EVENT(kvm, 3, "create cpu %d at %p, sie block at %p", id, vcpu, |
472 | vcpu->arch.sie_block); | |
ade38c31 | 473 | trace_kvm_s390_create_vcpu(id, vcpu, vcpu->arch.sie_block); |
b0c632db | 474 | |
b0c632db | 475 | return vcpu; |
7b06bf2f WY |
476 | out_free_sie_block: |
477 | free_page((unsigned long)(vcpu->arch.sie_block)); | |
b0c632db | 478 | out_free_cpu: |
b110feaf | 479 | kmem_cache_free(kvm_vcpu_cache, vcpu); |
4d47555a | 480 | out: |
b0c632db HC |
481 | return ERR_PTR(rc); |
482 | } | |
483 | ||
b0c632db HC |
484 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu) |
485 | { | |
486 | /* kvm common code refers to this, but never calls it */ | |
487 | BUG(); | |
488 | return 0; | |
489 | } | |
490 | ||
49b99e1e CB |
491 | void s390_vcpu_block(struct kvm_vcpu *vcpu) |
492 | { | |
493 | atomic_set_mask(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20); | |
494 | } | |
495 | ||
496 | void s390_vcpu_unblock(struct kvm_vcpu *vcpu) | |
497 | { | |
498 | atomic_clear_mask(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20); | |
499 | } | |
500 | ||
501 | /* | |
502 | * Kick a guest cpu out of SIE and wait until SIE is not running. | |
503 | * If the CPU is not running (e.g. waiting as idle) the function will | |
504 | * return immediately. */ | |
505 | void exit_sie(struct kvm_vcpu *vcpu) | |
506 | { | |
507 | atomic_set_mask(CPUSTAT_STOP_INT, &vcpu->arch.sie_block->cpuflags); | |
508 | while (vcpu->arch.sie_block->prog0c & PROG_IN_SIE) | |
509 | cpu_relax(); | |
510 | } | |
511 | ||
512 | /* Kick a guest cpu out of SIE and prevent SIE-reentry */ | |
513 | void exit_sie_sync(struct kvm_vcpu *vcpu) | |
514 | { | |
515 | s390_vcpu_block(vcpu); | |
516 | exit_sie(vcpu); | |
517 | } | |
518 | ||
2c70fe44 CB |
519 | static void kvm_gmap_notifier(struct gmap *gmap, unsigned long address) |
520 | { | |
521 | int i; | |
522 | struct kvm *kvm = gmap->private; | |
523 | struct kvm_vcpu *vcpu; | |
524 | ||
525 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
526 | /* match against both prefix pages */ | |
527 | if (vcpu->arch.sie_block->prefix == (address & ~0x1000UL)) { | |
528 | VCPU_EVENT(vcpu, 2, "gmap notifier for %lx", address); | |
529 | kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu); | |
530 | exit_sie_sync(vcpu); | |
531 | } | |
532 | } | |
533 | } | |
534 | ||
b6d33834 CD |
535 | int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu) |
536 | { | |
537 | /* kvm common code refers to this, but never calls it */ | |
538 | BUG(); | |
539 | return 0; | |
540 | } | |
541 | ||
14eebd91 CO |
542 | static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, |
543 | struct kvm_one_reg *reg) | |
544 | { | |
545 | int r = -EINVAL; | |
546 | ||
547 | switch (reg->id) { | |
29b7c71b CO |
548 | case KVM_REG_S390_TODPR: |
549 | r = put_user(vcpu->arch.sie_block->todpr, | |
550 | (u32 __user *)reg->addr); | |
551 | break; | |
552 | case KVM_REG_S390_EPOCHDIFF: | |
553 | r = put_user(vcpu->arch.sie_block->epoch, | |
554 | (u64 __user *)reg->addr); | |
555 | break; | |
46a6dd1c J |
556 | case KVM_REG_S390_CPU_TIMER: |
557 | r = put_user(vcpu->arch.sie_block->cputm, | |
558 | (u64 __user *)reg->addr); | |
559 | break; | |
560 | case KVM_REG_S390_CLOCK_COMP: | |
561 | r = put_user(vcpu->arch.sie_block->ckc, | |
562 | (u64 __user *)reg->addr); | |
563 | break; | |
536336c2 DD |
564 | case KVM_REG_S390_PFTOKEN: |
565 | r = put_user(vcpu->arch.pfault_token, | |
566 | (u64 __user *)reg->addr); | |
567 | break; | |
568 | case KVM_REG_S390_PFCOMPARE: | |
569 | r = put_user(vcpu->arch.pfault_compare, | |
570 | (u64 __user *)reg->addr); | |
571 | break; | |
572 | case KVM_REG_S390_PFSELECT: | |
573 | r = put_user(vcpu->arch.pfault_select, | |
574 | (u64 __user *)reg->addr); | |
575 | break; | |
14eebd91 CO |
576 | default: |
577 | break; | |
578 | } | |
579 | ||
580 | return r; | |
581 | } | |
582 | ||
583 | static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, | |
584 | struct kvm_one_reg *reg) | |
585 | { | |
586 | int r = -EINVAL; | |
587 | ||
588 | switch (reg->id) { | |
29b7c71b CO |
589 | case KVM_REG_S390_TODPR: |
590 | r = get_user(vcpu->arch.sie_block->todpr, | |
591 | (u32 __user *)reg->addr); | |
592 | break; | |
593 | case KVM_REG_S390_EPOCHDIFF: | |
594 | r = get_user(vcpu->arch.sie_block->epoch, | |
595 | (u64 __user *)reg->addr); | |
596 | break; | |
46a6dd1c J |
597 | case KVM_REG_S390_CPU_TIMER: |
598 | r = get_user(vcpu->arch.sie_block->cputm, | |
599 | (u64 __user *)reg->addr); | |
600 | break; | |
601 | case KVM_REG_S390_CLOCK_COMP: | |
602 | r = get_user(vcpu->arch.sie_block->ckc, | |
603 | (u64 __user *)reg->addr); | |
604 | break; | |
536336c2 DD |
605 | case KVM_REG_S390_PFTOKEN: |
606 | r = get_user(vcpu->arch.pfault_token, | |
607 | (u64 __user *)reg->addr); | |
608 | break; | |
609 | case KVM_REG_S390_PFCOMPARE: | |
610 | r = get_user(vcpu->arch.pfault_compare, | |
611 | (u64 __user *)reg->addr); | |
612 | break; | |
613 | case KVM_REG_S390_PFSELECT: | |
614 | r = get_user(vcpu->arch.pfault_select, | |
615 | (u64 __user *)reg->addr); | |
616 | break; | |
14eebd91 CO |
617 | default: |
618 | break; | |
619 | } | |
620 | ||
621 | return r; | |
622 | } | |
b6d33834 | 623 | |
b0c632db HC |
624 | static int kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu) |
625 | { | |
b0c632db | 626 | kvm_s390_vcpu_initial_reset(vcpu); |
b0c632db HC |
627 | return 0; |
628 | } | |
629 | ||
630 | int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
631 | { | |
5a32c1af | 632 | memcpy(&vcpu->run->s.regs.gprs, ®s->gprs, sizeof(regs->gprs)); |
b0c632db HC |
633 | return 0; |
634 | } | |
635 | ||
636 | int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
637 | { | |
5a32c1af | 638 | memcpy(®s->gprs, &vcpu->run->s.regs.gprs, sizeof(regs->gprs)); |
b0c632db HC |
639 | return 0; |
640 | } | |
641 | ||
642 | int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, | |
643 | struct kvm_sregs *sregs) | |
644 | { | |
59674c1a | 645 | memcpy(&vcpu->run->s.regs.acrs, &sregs->acrs, sizeof(sregs->acrs)); |
b0c632db | 646 | memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs)); |
59674c1a | 647 | restore_access_regs(vcpu->run->s.regs.acrs); |
b0c632db HC |
648 | return 0; |
649 | } | |
650 | ||
651 | int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, | |
652 | struct kvm_sregs *sregs) | |
653 | { | |
59674c1a | 654 | memcpy(&sregs->acrs, &vcpu->run->s.regs.acrs, sizeof(sregs->acrs)); |
b0c632db | 655 | memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs)); |
b0c632db HC |
656 | return 0; |
657 | } | |
658 | ||
659 | int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
660 | { | |
4725c860 MS |
661 | if (test_fp_ctl(fpu->fpc)) |
662 | return -EINVAL; | |
b0c632db | 663 | memcpy(&vcpu->arch.guest_fpregs.fprs, &fpu->fprs, sizeof(fpu->fprs)); |
4725c860 MS |
664 | vcpu->arch.guest_fpregs.fpc = fpu->fpc; |
665 | restore_fp_ctl(&vcpu->arch.guest_fpregs.fpc); | |
666 | restore_fp_regs(vcpu->arch.guest_fpregs.fprs); | |
b0c632db HC |
667 | return 0; |
668 | } | |
669 | ||
670 | int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
671 | { | |
b0c632db HC |
672 | memcpy(&fpu->fprs, &vcpu->arch.guest_fpregs.fprs, sizeof(fpu->fprs)); |
673 | fpu->fpc = vcpu->arch.guest_fpregs.fpc; | |
b0c632db HC |
674 | return 0; |
675 | } | |
676 | ||
677 | static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw) | |
678 | { | |
679 | int rc = 0; | |
680 | ||
9e6dabef | 681 | if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_STOPPED)) |
b0c632db | 682 | rc = -EBUSY; |
d7b0b5eb CO |
683 | else { |
684 | vcpu->run->psw_mask = psw.mask; | |
685 | vcpu->run->psw_addr = psw.addr; | |
686 | } | |
b0c632db HC |
687 | return rc; |
688 | } | |
689 | ||
690 | int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, | |
691 | struct kvm_translation *tr) | |
692 | { | |
693 | return -EINVAL; /* not implemented yet */ | |
694 | } | |
695 | ||
d0bfb940 JK |
696 | int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, |
697 | struct kvm_guest_debug *dbg) | |
b0c632db HC |
698 | { |
699 | return -EINVAL; /* not implemented yet */ | |
700 | } | |
701 | ||
62d9f0db MT |
702 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
703 | struct kvm_mp_state *mp_state) | |
704 | { | |
705 | return -EINVAL; /* not implemented yet */ | |
706 | } | |
707 | ||
708 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
709 | struct kvm_mp_state *mp_state) | |
710 | { | |
711 | return -EINVAL; /* not implemented yet */ | |
712 | } | |
713 | ||
2c70fe44 CB |
714 | static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu) |
715 | { | |
716 | /* | |
717 | * We use MMU_RELOAD just to re-arm the ipte notifier for the | |
718 | * guest prefix page. gmap_ipte_notify will wait on the ptl lock. | |
719 | * This ensures that the ipte instruction for this request has | |
720 | * already finished. We might race against a second unmapper that | |
721 | * wants to set the blocking bit. Lets just retry the request loop. | |
722 | */ | |
723 | while (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu)) { | |
724 | int rc; | |
725 | rc = gmap_ipte_notify(vcpu->arch.gmap, | |
726 | vcpu->arch.sie_block->prefix, | |
727 | PAGE_SIZE * 2); | |
728 | if (rc) | |
729 | return rc; | |
730 | s390_vcpu_unblock(vcpu); | |
731 | } | |
732 | return 0; | |
733 | } | |
734 | ||
24eb3a82 DD |
735 | static long kvm_arch_fault_in_sync(struct kvm_vcpu *vcpu) |
736 | { | |
737 | long rc; | |
738 | hva_t fault = gmap_fault(current->thread.gmap_addr, vcpu->arch.gmap); | |
739 | struct mm_struct *mm = current->mm; | |
740 | down_read(&mm->mmap_sem); | |
741 | rc = get_user_pages(current, mm, fault, 1, 1, 0, NULL, NULL); | |
742 | up_read(&mm->mmap_sem); | |
743 | return rc; | |
744 | } | |
745 | ||
3c038e6b DD |
746 | static void __kvm_inject_pfault_token(struct kvm_vcpu *vcpu, bool start_token, |
747 | unsigned long token) | |
748 | { | |
749 | struct kvm_s390_interrupt inti; | |
750 | inti.parm64 = token; | |
751 | ||
752 | if (start_token) { | |
753 | inti.type = KVM_S390_INT_PFAULT_INIT; | |
754 | WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &inti)); | |
755 | } else { | |
756 | inti.type = KVM_S390_INT_PFAULT_DONE; | |
757 | WARN_ON_ONCE(kvm_s390_inject_vm(vcpu->kvm, &inti)); | |
758 | } | |
759 | } | |
760 | ||
761 | void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, | |
762 | struct kvm_async_pf *work) | |
763 | { | |
764 | trace_kvm_s390_pfault_init(vcpu, work->arch.pfault_token); | |
765 | __kvm_inject_pfault_token(vcpu, true, work->arch.pfault_token); | |
766 | } | |
767 | ||
768 | void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, | |
769 | struct kvm_async_pf *work) | |
770 | { | |
771 | trace_kvm_s390_pfault_done(vcpu, work->arch.pfault_token); | |
772 | __kvm_inject_pfault_token(vcpu, false, work->arch.pfault_token); | |
773 | } | |
774 | ||
775 | void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, | |
776 | struct kvm_async_pf *work) | |
777 | { | |
778 | /* s390 will always inject the page directly */ | |
779 | } | |
780 | ||
781 | bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu) | |
782 | { | |
783 | /* | |
784 | * s390 will always inject the page directly, | |
785 | * but we still want check_async_completion to cleanup | |
786 | */ | |
787 | return true; | |
788 | } | |
789 | ||
790 | static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu) | |
791 | { | |
792 | hva_t hva; | |
793 | struct kvm_arch_async_pf arch; | |
794 | int rc; | |
795 | ||
796 | if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID) | |
797 | return 0; | |
798 | if ((vcpu->arch.sie_block->gpsw.mask & vcpu->arch.pfault_select) != | |
799 | vcpu->arch.pfault_compare) | |
800 | return 0; | |
801 | if (psw_extint_disabled(vcpu)) | |
802 | return 0; | |
803 | if (kvm_cpu_has_interrupt(vcpu)) | |
804 | return 0; | |
805 | if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul)) | |
806 | return 0; | |
807 | if (!vcpu->arch.gmap->pfault_enabled) | |
808 | return 0; | |
809 | ||
810 | hva = gmap_fault(current->thread.gmap_addr, vcpu->arch.gmap); | |
811 | if (copy_from_guest(vcpu, &arch.pfault_token, vcpu->arch.pfault_token, 8)) | |
812 | return 0; | |
813 | ||
814 | rc = kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch); | |
815 | return rc; | |
816 | } | |
817 | ||
3fb4c40f | 818 | static int vcpu_pre_run(struct kvm_vcpu *vcpu) |
b0c632db | 819 | { |
3fb4c40f | 820 | int rc, cpuflags; |
e168bf8d | 821 | |
3c038e6b DD |
822 | /* |
823 | * On s390 notifications for arriving pages will be delivered directly | |
824 | * to the guest but the house keeping for completed pfaults is | |
825 | * handled outside the worker. | |
826 | */ | |
827 | kvm_check_async_pf_completion(vcpu); | |
828 | ||
5a32c1af | 829 | memcpy(&vcpu->arch.sie_block->gg14, &vcpu->run->s.regs.gprs[14], 16); |
b0c632db HC |
830 | |
831 | if (need_resched()) | |
832 | schedule(); | |
833 | ||
71cde587 CB |
834 | if (test_thread_flag(TIF_MCCK_PENDING)) |
835 | s390_handle_mcck(); | |
836 | ||
d6b6d166 CO |
837 | if (!kvm_is_ucontrol(vcpu->kvm)) |
838 | kvm_s390_deliver_pending_interrupts(vcpu); | |
0ff31867 | 839 | |
2c70fe44 CB |
840 | rc = kvm_s390_handle_requests(vcpu); |
841 | if (rc) | |
842 | return rc; | |
843 | ||
b0c632db | 844 | vcpu->arch.sie_block->icptcode = 0; |
3fb4c40f TH |
845 | cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags); |
846 | VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags); | |
847 | trace_kvm_s390_sie_enter(vcpu, cpuflags); | |
2b29a9fd | 848 | |
3fb4c40f TH |
849 | return 0; |
850 | } | |
851 | ||
852 | static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason) | |
853 | { | |
24eb3a82 | 854 | int rc = -1; |
2b29a9fd DD |
855 | |
856 | VCPU_EVENT(vcpu, 6, "exit sie icptcode %d", | |
857 | vcpu->arch.sie_block->icptcode); | |
858 | trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode); | |
859 | ||
3fb4c40f | 860 | if (exit_reason >= 0) { |
7c470539 | 861 | rc = 0; |
210b1607 TH |
862 | } else if (kvm_is_ucontrol(vcpu->kvm)) { |
863 | vcpu->run->exit_reason = KVM_EXIT_S390_UCONTROL; | |
864 | vcpu->run->s390_ucontrol.trans_exc_code = | |
865 | current->thread.gmap_addr; | |
866 | vcpu->run->s390_ucontrol.pgm_code = 0x10; | |
867 | rc = -EREMOTE; | |
24eb3a82 DD |
868 | |
869 | } else if (current->thread.gmap_pfault) { | |
3c038e6b | 870 | trace_kvm_s390_major_guest_pfault(vcpu); |
24eb3a82 | 871 | current->thread.gmap_pfault = 0; |
3c038e6b DD |
872 | if (kvm_arch_setup_async_pf(vcpu) || |
873 | (kvm_arch_fault_in_sync(vcpu) >= 0)) | |
24eb3a82 DD |
874 | rc = 0; |
875 | } | |
876 | ||
877 | if (rc == -1) { | |
699bde3b CB |
878 | VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction"); |
879 | trace_kvm_s390_sie_fault(vcpu); | |
880 | rc = kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); | |
1f0d0f09 | 881 | } |
b0c632db | 882 | |
5a32c1af | 883 | memcpy(&vcpu->run->s.regs.gprs[14], &vcpu->arch.sie_block->gg14, 16); |
3fb4c40f | 884 | |
a76ccff6 TH |
885 | if (rc == 0) { |
886 | if (kvm_is_ucontrol(vcpu->kvm)) | |
887 | rc = -EOPNOTSUPP; | |
888 | else | |
889 | rc = kvm_handle_sie_intercept(vcpu); | |
890 | } | |
891 | ||
3fb4c40f TH |
892 | return rc; |
893 | } | |
894 | ||
895 | static int __vcpu_run(struct kvm_vcpu *vcpu) | |
896 | { | |
897 | int rc, exit_reason; | |
898 | ||
800c1065 TH |
899 | /* |
900 | * We try to hold kvm->srcu during most of vcpu_run (except when run- | |
901 | * ning the guest), so that memslots (and other stuff) are protected | |
902 | */ | |
903 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); | |
904 | ||
a76ccff6 TH |
905 | do { |
906 | rc = vcpu_pre_run(vcpu); | |
907 | if (rc) | |
908 | break; | |
3fb4c40f | 909 | |
800c1065 | 910 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); |
a76ccff6 TH |
911 | /* |
912 | * As PF_VCPU will be used in fault handler, between | |
913 | * guest_enter and guest_exit should be no uaccess. | |
914 | */ | |
915 | preempt_disable(); | |
916 | kvm_guest_enter(); | |
917 | preempt_enable(); | |
918 | exit_reason = sie64a(vcpu->arch.sie_block, | |
919 | vcpu->run->s.regs.gprs); | |
920 | kvm_guest_exit(); | |
800c1065 | 921 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); |
a76ccff6 TH |
922 | |
923 | rc = vcpu_post_run(vcpu, exit_reason); | |
924 | } while (!signal_pending(current) && !rc); | |
3fb4c40f | 925 | |
800c1065 | 926 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); |
e168bf8d | 927 | return rc; |
b0c632db HC |
928 | } |
929 | ||
930 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |
931 | { | |
8f2abe6a | 932 | int rc; |
b0c632db HC |
933 | sigset_t sigsaved; |
934 | ||
b0c632db HC |
935 | if (vcpu->sigset_active) |
936 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
937 | ||
9e6dabef | 938 | atomic_clear_mask(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags); |
b0c632db | 939 | |
ba5c1e9b CO |
940 | BUG_ON(vcpu->kvm->arch.float_int.local_int[vcpu->vcpu_id] == NULL); |
941 | ||
8f2abe6a CB |
942 | switch (kvm_run->exit_reason) { |
943 | case KVM_EXIT_S390_SIEIC: | |
8f2abe6a | 944 | case KVM_EXIT_UNKNOWN: |
9ace903d | 945 | case KVM_EXIT_INTR: |
8f2abe6a | 946 | case KVM_EXIT_S390_RESET: |
e168bf8d | 947 | case KVM_EXIT_S390_UCONTROL: |
fa6b7fe9 | 948 | case KVM_EXIT_S390_TSCH: |
8f2abe6a CB |
949 | break; |
950 | default: | |
951 | BUG(); | |
952 | } | |
953 | ||
d7b0b5eb CO |
954 | vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask; |
955 | vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr; | |
60b413c9 CB |
956 | if (kvm_run->kvm_dirty_regs & KVM_SYNC_PREFIX) { |
957 | kvm_run->kvm_dirty_regs &= ~KVM_SYNC_PREFIX; | |
958 | kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix); | |
959 | } | |
9eed0735 CB |
960 | if (kvm_run->kvm_dirty_regs & KVM_SYNC_CRS) { |
961 | kvm_run->kvm_dirty_regs &= ~KVM_SYNC_CRS; | |
962 | memcpy(&vcpu->arch.sie_block->gcr, &kvm_run->s.regs.crs, 128); | |
963 | kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix); | |
964 | } | |
d7b0b5eb | 965 | |
dab4079d | 966 | might_fault(); |
a76ccff6 | 967 | rc = __vcpu_run(vcpu); |
9ace903d | 968 | |
b1d16c49 CE |
969 | if (signal_pending(current) && !rc) { |
970 | kvm_run->exit_reason = KVM_EXIT_INTR; | |
8f2abe6a | 971 | rc = -EINTR; |
b1d16c49 | 972 | } |
8f2abe6a | 973 | |
b8e660b8 | 974 | if (rc == -EOPNOTSUPP) { |
8f2abe6a CB |
975 | /* intercept cannot be handled in-kernel, prepare kvm-run */ |
976 | kvm_run->exit_reason = KVM_EXIT_S390_SIEIC; | |
977 | kvm_run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode; | |
8f2abe6a CB |
978 | kvm_run->s390_sieic.ipa = vcpu->arch.sie_block->ipa; |
979 | kvm_run->s390_sieic.ipb = vcpu->arch.sie_block->ipb; | |
980 | rc = 0; | |
981 | } | |
982 | ||
983 | if (rc == -EREMOTE) { | |
984 | /* intercept was handled, but userspace support is needed | |
985 | * kvm_run has been prepared by the handler */ | |
986 | rc = 0; | |
987 | } | |
b0c632db | 988 | |
d7b0b5eb CO |
989 | kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask; |
990 | kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr; | |
60b413c9 | 991 | kvm_run->s.regs.prefix = vcpu->arch.sie_block->prefix; |
9eed0735 | 992 | memcpy(&kvm_run->s.regs.crs, &vcpu->arch.sie_block->gcr, 128); |
d7b0b5eb | 993 | |
b0c632db HC |
994 | if (vcpu->sigset_active) |
995 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
996 | ||
b0c632db | 997 | vcpu->stat.exit_userspace++; |
7e8e6ab4 | 998 | return rc; |
b0c632db HC |
999 | } |
1000 | ||
092670cd | 1001 | static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, void *from, |
b0c632db HC |
1002 | unsigned long n, int prefix) |
1003 | { | |
1004 | if (prefix) | |
1005 | return copy_to_guest(vcpu, guestdest, from, n); | |
1006 | else | |
1007 | return copy_to_guest_absolute(vcpu, guestdest, from, n); | |
1008 | } | |
1009 | ||
1010 | /* | |
1011 | * store status at address | |
1012 | * we use have two special cases: | |
1013 | * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit | |
1014 | * KVM_S390_STORE_STATUS_PREFIXED: -> prefix | |
1015 | */ | |
e879892c | 1016 | int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long addr) |
b0c632db | 1017 | { |
092670cd | 1018 | unsigned char archmode = 1; |
b0c632db | 1019 | int prefix; |
178bd789 | 1020 | u64 clkcomp; |
b0c632db HC |
1021 | |
1022 | if (addr == KVM_S390_STORE_STATUS_NOADDR) { | |
1023 | if (copy_to_guest_absolute(vcpu, 163ul, &archmode, 1)) | |
1024 | return -EFAULT; | |
1025 | addr = SAVE_AREA_BASE; | |
1026 | prefix = 0; | |
1027 | } else if (addr == KVM_S390_STORE_STATUS_PREFIXED) { | |
1028 | if (copy_to_guest(vcpu, 163ul, &archmode, 1)) | |
1029 | return -EFAULT; | |
1030 | addr = SAVE_AREA_BASE; | |
1031 | prefix = 1; | |
1032 | } else | |
1033 | prefix = 0; | |
1034 | ||
f64ca217 | 1035 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, fp_regs), |
b0c632db HC |
1036 | vcpu->arch.guest_fpregs.fprs, 128, prefix)) |
1037 | return -EFAULT; | |
1038 | ||
f64ca217 | 1039 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, gp_regs), |
5a32c1af | 1040 | vcpu->run->s.regs.gprs, 128, prefix)) |
b0c632db HC |
1041 | return -EFAULT; |
1042 | ||
f64ca217 | 1043 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, psw), |
b0c632db HC |
1044 | &vcpu->arch.sie_block->gpsw, 16, prefix)) |
1045 | return -EFAULT; | |
1046 | ||
f64ca217 | 1047 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, pref_reg), |
b0c632db HC |
1048 | &vcpu->arch.sie_block->prefix, 4, prefix)) |
1049 | return -EFAULT; | |
1050 | ||
1051 | if (__guestcopy(vcpu, | |
f64ca217 | 1052 | addr + offsetof(struct save_area, fp_ctrl_reg), |
b0c632db HC |
1053 | &vcpu->arch.guest_fpregs.fpc, 4, prefix)) |
1054 | return -EFAULT; | |
1055 | ||
f64ca217 | 1056 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, tod_reg), |
b0c632db HC |
1057 | &vcpu->arch.sie_block->todpr, 4, prefix)) |
1058 | return -EFAULT; | |
1059 | ||
f64ca217 | 1060 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, timer), |
b0c632db HC |
1061 | &vcpu->arch.sie_block->cputm, 8, prefix)) |
1062 | return -EFAULT; | |
1063 | ||
178bd789 | 1064 | clkcomp = vcpu->arch.sie_block->ckc >> 8; |
f64ca217 | 1065 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, clk_cmp), |
178bd789 | 1066 | &clkcomp, 8, prefix)) |
b0c632db HC |
1067 | return -EFAULT; |
1068 | ||
f64ca217 | 1069 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, acc_regs), |
59674c1a | 1070 | &vcpu->run->s.regs.acrs, 64, prefix)) |
b0c632db HC |
1071 | return -EFAULT; |
1072 | ||
1073 | if (__guestcopy(vcpu, | |
f64ca217 | 1074 | addr + offsetof(struct save_area, ctrl_regs), |
b0c632db HC |
1075 | &vcpu->arch.sie_block->gcr, 128, prefix)) |
1076 | return -EFAULT; | |
1077 | return 0; | |
1078 | } | |
1079 | ||
e879892c TH |
1080 | int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr) |
1081 | { | |
1082 | /* | |
1083 | * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy | |
1084 | * copying in vcpu load/put. Lets update our copies before we save | |
1085 | * it into the save area | |
1086 | */ | |
1087 | save_fp_ctl(&vcpu->arch.guest_fpregs.fpc); | |
1088 | save_fp_regs(vcpu->arch.guest_fpregs.fprs); | |
1089 | save_access_regs(vcpu->run->s.regs.acrs); | |
1090 | ||
1091 | return kvm_s390_store_status_unloaded(vcpu, addr); | |
1092 | } | |
1093 | ||
d6712df9 CH |
1094 | static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu, |
1095 | struct kvm_enable_cap *cap) | |
1096 | { | |
1097 | int r; | |
1098 | ||
1099 | if (cap->flags) | |
1100 | return -EINVAL; | |
1101 | ||
1102 | switch (cap->cap) { | |
fa6b7fe9 CH |
1103 | case KVM_CAP_S390_CSS_SUPPORT: |
1104 | if (!vcpu->kvm->arch.css_support) { | |
1105 | vcpu->kvm->arch.css_support = 1; | |
1106 | trace_kvm_s390_enable_css(vcpu->kvm); | |
1107 | } | |
1108 | r = 0; | |
1109 | break; | |
d6712df9 CH |
1110 | default: |
1111 | r = -EINVAL; | |
1112 | break; | |
1113 | } | |
1114 | return r; | |
1115 | } | |
1116 | ||
b0c632db HC |
1117 | long kvm_arch_vcpu_ioctl(struct file *filp, |
1118 | unsigned int ioctl, unsigned long arg) | |
1119 | { | |
1120 | struct kvm_vcpu *vcpu = filp->private_data; | |
1121 | void __user *argp = (void __user *)arg; | |
800c1065 | 1122 | int idx; |
bc923cc9 | 1123 | long r; |
b0c632db | 1124 | |
93736624 AK |
1125 | switch (ioctl) { |
1126 | case KVM_S390_INTERRUPT: { | |
ba5c1e9b CO |
1127 | struct kvm_s390_interrupt s390int; |
1128 | ||
93736624 | 1129 | r = -EFAULT; |
ba5c1e9b | 1130 | if (copy_from_user(&s390int, argp, sizeof(s390int))) |
93736624 AK |
1131 | break; |
1132 | r = kvm_s390_inject_vcpu(vcpu, &s390int); | |
1133 | break; | |
ba5c1e9b | 1134 | } |
b0c632db | 1135 | case KVM_S390_STORE_STATUS: |
800c1065 | 1136 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
bc923cc9 | 1137 | r = kvm_s390_vcpu_store_status(vcpu, arg); |
800c1065 | 1138 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
bc923cc9 | 1139 | break; |
b0c632db HC |
1140 | case KVM_S390_SET_INITIAL_PSW: { |
1141 | psw_t psw; | |
1142 | ||
bc923cc9 | 1143 | r = -EFAULT; |
b0c632db | 1144 | if (copy_from_user(&psw, argp, sizeof(psw))) |
bc923cc9 AK |
1145 | break; |
1146 | r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw); | |
1147 | break; | |
b0c632db HC |
1148 | } |
1149 | case KVM_S390_INITIAL_RESET: | |
bc923cc9 AK |
1150 | r = kvm_arch_vcpu_ioctl_initial_reset(vcpu); |
1151 | break; | |
14eebd91 CO |
1152 | case KVM_SET_ONE_REG: |
1153 | case KVM_GET_ONE_REG: { | |
1154 | struct kvm_one_reg reg; | |
1155 | r = -EFAULT; | |
1156 | if (copy_from_user(®, argp, sizeof(reg))) | |
1157 | break; | |
1158 | if (ioctl == KVM_SET_ONE_REG) | |
1159 | r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, ®); | |
1160 | else | |
1161 | r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, ®); | |
1162 | break; | |
1163 | } | |
27e0393f CO |
1164 | #ifdef CONFIG_KVM_S390_UCONTROL |
1165 | case KVM_S390_UCAS_MAP: { | |
1166 | struct kvm_s390_ucas_mapping ucasmap; | |
1167 | ||
1168 | if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) { | |
1169 | r = -EFAULT; | |
1170 | break; | |
1171 | } | |
1172 | ||
1173 | if (!kvm_is_ucontrol(vcpu->kvm)) { | |
1174 | r = -EINVAL; | |
1175 | break; | |
1176 | } | |
1177 | ||
1178 | r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr, | |
1179 | ucasmap.vcpu_addr, ucasmap.length); | |
1180 | break; | |
1181 | } | |
1182 | case KVM_S390_UCAS_UNMAP: { | |
1183 | struct kvm_s390_ucas_mapping ucasmap; | |
1184 | ||
1185 | if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) { | |
1186 | r = -EFAULT; | |
1187 | break; | |
1188 | } | |
1189 | ||
1190 | if (!kvm_is_ucontrol(vcpu->kvm)) { | |
1191 | r = -EINVAL; | |
1192 | break; | |
1193 | } | |
1194 | ||
1195 | r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr, | |
1196 | ucasmap.length); | |
1197 | break; | |
1198 | } | |
1199 | #endif | |
ccc7910f CO |
1200 | case KVM_S390_VCPU_FAULT: { |
1201 | r = gmap_fault(arg, vcpu->arch.gmap); | |
1202 | if (!IS_ERR_VALUE(r)) | |
1203 | r = 0; | |
1204 | break; | |
1205 | } | |
d6712df9 CH |
1206 | case KVM_ENABLE_CAP: |
1207 | { | |
1208 | struct kvm_enable_cap cap; | |
1209 | r = -EFAULT; | |
1210 | if (copy_from_user(&cap, argp, sizeof(cap))) | |
1211 | break; | |
1212 | r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap); | |
1213 | break; | |
1214 | } | |
b0c632db | 1215 | default: |
3e6afcf1 | 1216 | r = -ENOTTY; |
b0c632db | 1217 | } |
bc923cc9 | 1218 | return r; |
b0c632db HC |
1219 | } |
1220 | ||
5b1c1493 CO |
1221 | int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf) |
1222 | { | |
1223 | #ifdef CONFIG_KVM_S390_UCONTROL | |
1224 | if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET) | |
1225 | && (kvm_is_ucontrol(vcpu->kvm))) { | |
1226 | vmf->page = virt_to_page(vcpu->arch.sie_block); | |
1227 | get_page(vmf->page); | |
1228 | return 0; | |
1229 | } | |
1230 | #endif | |
1231 | return VM_FAULT_SIGBUS; | |
1232 | } | |
1233 | ||
5587027c | 1234 | void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free, |
db3fe4eb TY |
1235 | struct kvm_memory_slot *dont) |
1236 | { | |
1237 | } | |
1238 | ||
5587027c AK |
1239 | int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot, |
1240 | unsigned long npages) | |
db3fe4eb TY |
1241 | { |
1242 | return 0; | |
1243 | } | |
1244 | ||
e59dbe09 TY |
1245 | void kvm_arch_memslots_updated(struct kvm *kvm) |
1246 | { | |
1247 | } | |
1248 | ||
b0c632db | 1249 | /* Section: memory related */ |
f7784b8e MT |
1250 | int kvm_arch_prepare_memory_region(struct kvm *kvm, |
1251 | struct kvm_memory_slot *memslot, | |
7b6195a9 TY |
1252 | struct kvm_userspace_memory_region *mem, |
1253 | enum kvm_mr_change change) | |
b0c632db | 1254 | { |
dd2887e7 NW |
1255 | /* A few sanity checks. We can have memory slots which have to be |
1256 | located/ended at a segment boundary (1MB). The memory in userland is | |
1257 | ok to be fragmented into various different vmas. It is okay to mmap() | |
1258 | and munmap() stuff in this slot after doing this call at any time */ | |
b0c632db | 1259 | |
598841ca | 1260 | if (mem->userspace_addr & 0xffffful) |
b0c632db HC |
1261 | return -EINVAL; |
1262 | ||
598841ca | 1263 | if (mem->memory_size & 0xffffful) |
b0c632db HC |
1264 | return -EINVAL; |
1265 | ||
f7784b8e MT |
1266 | return 0; |
1267 | } | |
1268 | ||
1269 | void kvm_arch_commit_memory_region(struct kvm *kvm, | |
1270 | struct kvm_userspace_memory_region *mem, | |
8482644a TY |
1271 | const struct kvm_memory_slot *old, |
1272 | enum kvm_mr_change change) | |
f7784b8e | 1273 | { |
f7850c92 | 1274 | int rc; |
f7784b8e | 1275 | |
2cef4deb CB |
1276 | /* If the basics of the memslot do not change, we do not want |
1277 | * to update the gmap. Every update causes several unnecessary | |
1278 | * segment translation exceptions. This is usually handled just | |
1279 | * fine by the normal fault handler + gmap, but it will also | |
1280 | * cause faults on the prefix page of running guest CPUs. | |
1281 | */ | |
1282 | if (old->userspace_addr == mem->userspace_addr && | |
1283 | old->base_gfn * PAGE_SIZE == mem->guest_phys_addr && | |
1284 | old->npages * PAGE_SIZE == mem->memory_size) | |
1285 | return; | |
598841ca CO |
1286 | |
1287 | rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr, | |
1288 | mem->guest_phys_addr, mem->memory_size); | |
1289 | if (rc) | |
f7850c92 | 1290 | printk(KERN_WARNING "kvm-s390: failed to commit memory region\n"); |
598841ca | 1291 | return; |
b0c632db HC |
1292 | } |
1293 | ||
2df72e9b MT |
1294 | void kvm_arch_flush_shadow_all(struct kvm *kvm) |
1295 | { | |
1296 | } | |
1297 | ||
1298 | void kvm_arch_flush_shadow_memslot(struct kvm *kvm, | |
1299 | struct kvm_memory_slot *slot) | |
34d4cb8f MT |
1300 | { |
1301 | } | |
1302 | ||
b0c632db HC |
1303 | static int __init kvm_s390_init(void) |
1304 | { | |
ef50f7ac | 1305 | int ret; |
0ee75bea | 1306 | ret = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE); |
ef50f7ac CB |
1307 | if (ret) |
1308 | return ret; | |
1309 | ||
1310 | /* | |
1311 | * guests can ask for up to 255+1 double words, we need a full page | |
25985edc | 1312 | * to hold the maximum amount of facilities. On the other hand, we |
ef50f7ac CB |
1313 | * only set facilities that are known to work in KVM. |
1314 | */ | |
78c4b59f MM |
1315 | vfacilities = (unsigned long *) get_zeroed_page(GFP_KERNEL|GFP_DMA); |
1316 | if (!vfacilities) { | |
ef50f7ac CB |
1317 | kvm_exit(); |
1318 | return -ENOMEM; | |
1319 | } | |
78c4b59f | 1320 | memcpy(vfacilities, S390_lowcore.stfle_fac_list, 16); |
d208c79d | 1321 | vfacilities[0] &= 0xff82fff3f4fc2000UL; |
7feb6bb8 | 1322 | vfacilities[1] &= 0x005c000000000000UL; |
ef50f7ac | 1323 | return 0; |
b0c632db HC |
1324 | } |
1325 | ||
1326 | static void __exit kvm_s390_exit(void) | |
1327 | { | |
78c4b59f | 1328 | free_page((unsigned long) vfacilities); |
b0c632db HC |
1329 | kvm_exit(); |
1330 | } | |
1331 | ||
1332 | module_init(kvm_s390_init); | |
1333 | module_exit(kvm_s390_exit); | |
566af940 CH |
1334 | |
1335 | /* | |
1336 | * Enable autoloading of the kvm module. | |
1337 | * Note that we add the module alias here instead of virt/kvm/kvm_main.c | |
1338 | * since x86 takes a different approach. | |
1339 | */ | |
1340 | #include <linux/miscdevice.h> | |
1341 | MODULE_ALIAS_MISCDEV(KVM_MINOR); | |
1342 | MODULE_ALIAS("devname:kvm"); |