]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * s390host.c -- hosting zSeries kernel virtual machines | |
3 | * | |
4 | * Copyright IBM Corp. 2008,2009 | |
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> | |
13 | * Christian Ehrhardt <ehrhardt@de.ibm.com> | |
14 | */ | |
15 | ||
16 | #include <linux/compiler.h> | |
17 | #include <linux/err.h> | |
18 | #include <linux/fs.h> | |
19 | #include <linux/hrtimer.h> | |
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> | |
25 | #include <linux/timer.h> | |
26 | #include <asm/asm-offsets.h> | |
27 | #include <asm/lowcore.h> | |
28 | #include <asm/pgtable.h> | |
29 | #include <asm/nmi.h> | |
30 | #include <asm/system.h> | |
31 | #include "kvm-s390.h" | |
32 | #include "gaccess.h" | |
33 | ||
34 | #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU | |
35 | ||
36 | struct kvm_stats_debugfs_item debugfs_entries[] = { | |
37 | { "userspace_handled", VCPU_STAT(exit_userspace) }, | |
38 | { "exit_null", VCPU_STAT(exit_null) }, | |
39 | { "exit_validity", VCPU_STAT(exit_validity) }, | |
40 | { "exit_stop_request", VCPU_STAT(exit_stop_request) }, | |
41 | { "exit_external_request", VCPU_STAT(exit_external_request) }, | |
42 | { "exit_external_interrupt", VCPU_STAT(exit_external_interrupt) }, | |
43 | { "exit_instruction", VCPU_STAT(exit_instruction) }, | |
44 | { "exit_program_interruption", VCPU_STAT(exit_program_interruption) }, | |
45 | { "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) }, | |
46 | { "instruction_lctlg", VCPU_STAT(instruction_lctlg) }, | |
47 | { "instruction_lctl", VCPU_STAT(instruction_lctl) }, | |
48 | { "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) }, | |
49 | { "deliver_external_call", VCPU_STAT(deliver_external_call) }, | |
50 | { "deliver_service_signal", VCPU_STAT(deliver_service_signal) }, | |
51 | { "deliver_virtio_interrupt", VCPU_STAT(deliver_virtio_interrupt) }, | |
52 | { "deliver_stop_signal", VCPU_STAT(deliver_stop_signal) }, | |
53 | { "deliver_prefix_signal", VCPU_STAT(deliver_prefix_signal) }, | |
54 | { "deliver_restart_signal", VCPU_STAT(deliver_restart_signal) }, | |
55 | { "deliver_program_interruption", VCPU_STAT(deliver_program_int) }, | |
56 | { "exit_wait_state", VCPU_STAT(exit_wait_state) }, | |
57 | { "instruction_stidp", VCPU_STAT(instruction_stidp) }, | |
58 | { "instruction_spx", VCPU_STAT(instruction_spx) }, | |
59 | { "instruction_stpx", VCPU_STAT(instruction_stpx) }, | |
60 | { "instruction_stap", VCPU_STAT(instruction_stap) }, | |
61 | { "instruction_storage_key", VCPU_STAT(instruction_storage_key) }, | |
62 | { "instruction_stsch", VCPU_STAT(instruction_stsch) }, | |
63 | { "instruction_chsc", VCPU_STAT(instruction_chsc) }, | |
64 | { "instruction_stsi", VCPU_STAT(instruction_stsi) }, | |
65 | { "instruction_stfl", VCPU_STAT(instruction_stfl) }, | |
66 | { "instruction_tprot", VCPU_STAT(instruction_tprot) }, | |
67 | { "instruction_sigp_sense", VCPU_STAT(instruction_sigp_sense) }, | |
68 | { "instruction_sigp_sense_running", VCPU_STAT(instruction_sigp_sense_running) }, | |
69 | { "instruction_sigp_external_call", VCPU_STAT(instruction_sigp_external_call) }, | |
70 | { "instruction_sigp_emergency", VCPU_STAT(instruction_sigp_emergency) }, | |
71 | { "instruction_sigp_stop", VCPU_STAT(instruction_sigp_stop) }, | |
72 | { "instruction_sigp_set_arch", VCPU_STAT(instruction_sigp_arch) }, | |
73 | { "instruction_sigp_set_prefix", VCPU_STAT(instruction_sigp_prefix) }, | |
74 | { "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart) }, | |
75 | { "diagnose_10", VCPU_STAT(diagnose_10) }, | |
76 | { "diagnose_44", VCPU_STAT(diagnose_44) }, | |
77 | { NULL } | |
78 | }; | |
79 | ||
80 | static unsigned long long *facilities; | |
81 | ||
82 | /* Section: not file related */ | |
83 | int kvm_arch_hardware_enable(void *garbage) | |
84 | { | |
85 | /* every s390 is virtualization enabled ;-) */ | |
86 | return 0; | |
87 | } | |
88 | ||
89 | void kvm_arch_hardware_disable(void *garbage) | |
90 | { | |
91 | } | |
92 | ||
93 | int kvm_arch_hardware_setup(void) | |
94 | { | |
95 | return 0; | |
96 | } | |
97 | ||
98 | void kvm_arch_hardware_unsetup(void) | |
99 | { | |
100 | } | |
101 | ||
102 | void kvm_arch_check_processor_compat(void *rtn) | |
103 | { | |
104 | } | |
105 | ||
106 | int kvm_arch_init(void *opaque) | |
107 | { | |
108 | return 0; | |
109 | } | |
110 | ||
111 | void kvm_arch_exit(void) | |
112 | { | |
113 | } | |
114 | ||
115 | /* Section: device related */ | |
116 | long kvm_arch_dev_ioctl(struct file *filp, | |
117 | unsigned int ioctl, unsigned long arg) | |
118 | { | |
119 | if (ioctl == KVM_S390_ENABLE_SIE) | |
120 | return s390_enable_sie(); | |
121 | return -EINVAL; | |
122 | } | |
123 | ||
124 | int kvm_dev_ioctl_check_extension(long ext) | |
125 | { | |
126 | int r; | |
127 | ||
128 | switch (ext) { | |
129 | case KVM_CAP_S390_PSW: | |
130 | case KVM_CAP_S390_GMAP: | |
131 | case KVM_CAP_SYNC_MMU: | |
132 | #ifdef CONFIG_KVM_S390_UCONTROL | |
133 | case KVM_CAP_S390_UCONTROL: | |
134 | #endif | |
135 | r = 1; | |
136 | break; | |
137 | default: | |
138 | r = 0; | |
139 | } | |
140 | return r; | |
141 | } | |
142 | ||
143 | /* Section: vm related */ | |
144 | /* | |
145 | * Get (and clear) the dirty memory log for a memory slot. | |
146 | */ | |
147 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, | |
148 | struct kvm_dirty_log *log) | |
149 | { | |
150 | return 0; | |
151 | } | |
152 | ||
153 | long kvm_arch_vm_ioctl(struct file *filp, | |
154 | unsigned int ioctl, unsigned long arg) | |
155 | { | |
156 | struct kvm *kvm = filp->private_data; | |
157 | void __user *argp = (void __user *)arg; | |
158 | int r; | |
159 | ||
160 | switch (ioctl) { | |
161 | case KVM_S390_INTERRUPT: { | |
162 | struct kvm_s390_interrupt s390int; | |
163 | ||
164 | r = -EFAULT; | |
165 | if (copy_from_user(&s390int, argp, sizeof(s390int))) | |
166 | break; | |
167 | r = kvm_s390_inject_vm(kvm, &s390int); | |
168 | break; | |
169 | } | |
170 | default: | |
171 | r = -ENOTTY; | |
172 | } | |
173 | ||
174 | return r; | |
175 | } | |
176 | ||
177 | int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) | |
178 | { | |
179 | int rc; | |
180 | char debug_name[16]; | |
181 | ||
182 | rc = -EINVAL; | |
183 | #ifdef CONFIG_KVM_S390_UCONTROL | |
184 | if (type & ~KVM_VM_S390_UCONTROL) | |
185 | goto out_err; | |
186 | if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN))) | |
187 | goto out_err; | |
188 | #else | |
189 | if (type) | |
190 | goto out_err; | |
191 | #endif | |
192 | ||
193 | rc = s390_enable_sie(); | |
194 | if (rc) | |
195 | goto out_err; | |
196 | ||
197 | rc = -ENOMEM; | |
198 | ||
199 | kvm->arch.sca = (struct sca_block *) get_zeroed_page(GFP_KERNEL); | |
200 | if (!kvm->arch.sca) | |
201 | goto out_err; | |
202 | ||
203 | sprintf(debug_name, "kvm-%u", current->pid); | |
204 | ||
205 | kvm->arch.dbf = debug_register(debug_name, 8, 2, 8 * sizeof(long)); | |
206 | if (!kvm->arch.dbf) | |
207 | goto out_nodbf; | |
208 | ||
209 | spin_lock_init(&kvm->arch.float_int.lock); | |
210 | INIT_LIST_HEAD(&kvm->arch.float_int.list); | |
211 | ||
212 | debug_register_view(kvm->arch.dbf, &debug_sprintf_view); | |
213 | VM_EVENT(kvm, 3, "%s", "vm created"); | |
214 | ||
215 | if (type & KVM_VM_S390_UCONTROL) { | |
216 | kvm->arch.gmap = NULL; | |
217 | } else { | |
218 | kvm->arch.gmap = gmap_alloc(current->mm); | |
219 | if (!kvm->arch.gmap) | |
220 | goto out_nogmap; | |
221 | } | |
222 | return 0; | |
223 | out_nogmap: | |
224 | debug_unregister(kvm->arch.dbf); | |
225 | out_nodbf: | |
226 | free_page((unsigned long)(kvm->arch.sca)); | |
227 | out_err: | |
228 | return rc; | |
229 | } | |
230 | ||
231 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) | |
232 | { | |
233 | VCPU_EVENT(vcpu, 3, "%s", "free cpu"); | |
234 | if (!kvm_is_ucontrol(vcpu->kvm)) { | |
235 | clear_bit(63 - vcpu->vcpu_id, | |
236 | (unsigned long *) &vcpu->kvm->arch.sca->mcn); | |
237 | if (vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sda == | |
238 | (__u64) vcpu->arch.sie_block) | |
239 | vcpu->kvm->arch.sca->cpu[vcpu->vcpu_id].sda = 0; | |
240 | } | |
241 | smp_mb(); | |
242 | ||
243 | if (kvm_is_ucontrol(vcpu->kvm)) | |
244 | gmap_free(vcpu->arch.gmap); | |
245 | ||
246 | free_page((unsigned long)(vcpu->arch.sie_block)); | |
247 | kvm_vcpu_uninit(vcpu); | |
248 | kfree(vcpu); | |
249 | } | |
250 | ||
251 | static void kvm_free_vcpus(struct kvm *kvm) | |
252 | { | |
253 | unsigned int i; | |
254 | struct kvm_vcpu *vcpu; | |
255 | ||
256 | kvm_for_each_vcpu(i, vcpu, kvm) | |
257 | kvm_arch_vcpu_destroy(vcpu); | |
258 | ||
259 | mutex_lock(&kvm->lock); | |
260 | for (i = 0; i < atomic_read(&kvm->online_vcpus); i++) | |
261 | kvm->vcpus[i] = NULL; | |
262 | ||
263 | atomic_set(&kvm->online_vcpus, 0); | |
264 | mutex_unlock(&kvm->lock); | |
265 | } | |
266 | ||
267 | void kvm_arch_sync_events(struct kvm *kvm) | |
268 | { | |
269 | } | |
270 | ||
271 | void kvm_arch_destroy_vm(struct kvm *kvm) | |
272 | { | |
273 | kvm_free_vcpus(kvm); | |
274 | free_page((unsigned long)(kvm->arch.sca)); | |
275 | debug_unregister(kvm->arch.dbf); | |
276 | if (!kvm_is_ucontrol(kvm)) | |
277 | gmap_free(kvm->arch.gmap); | |
278 | } | |
279 | ||
280 | /* Section: vcpu related */ | |
281 | int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu) | |
282 | { | |
283 | if (kvm_is_ucontrol(vcpu->kvm)) { | |
284 | vcpu->arch.gmap = gmap_alloc(current->mm); | |
285 | if (!vcpu->arch.gmap) | |
286 | return -ENOMEM; | |
287 | return 0; | |
288 | } | |
289 | ||
290 | vcpu->arch.gmap = vcpu->kvm->arch.gmap; | |
291 | return 0; | |
292 | } | |
293 | ||
294 | void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) | |
295 | { | |
296 | /* Nothing todo */ | |
297 | } | |
298 | ||
299 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) | |
300 | { | |
301 | save_fp_regs(&vcpu->arch.host_fpregs); | |
302 | save_access_regs(vcpu->arch.host_acrs); | |
303 | vcpu->arch.guest_fpregs.fpc &= FPC_VALID_MASK; | |
304 | restore_fp_regs(&vcpu->arch.guest_fpregs); | |
305 | restore_access_regs(vcpu->arch.guest_acrs); | |
306 | gmap_enable(vcpu->arch.gmap); | |
307 | atomic_set_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags); | |
308 | } | |
309 | ||
310 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) | |
311 | { | |
312 | atomic_clear_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags); | |
313 | gmap_disable(vcpu->arch.gmap); | |
314 | save_fp_regs(&vcpu->arch.guest_fpregs); | |
315 | save_access_regs(vcpu->arch.guest_acrs); | |
316 | restore_fp_regs(&vcpu->arch.host_fpregs); | |
317 | restore_access_regs(vcpu->arch.host_acrs); | |
318 | } | |
319 | ||
320 | static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu) | |
321 | { | |
322 | /* this equals initial cpu reset in pop, but we don't switch to ESA */ | |
323 | vcpu->arch.sie_block->gpsw.mask = 0UL; | |
324 | vcpu->arch.sie_block->gpsw.addr = 0UL; | |
325 | kvm_s390_set_prefix(vcpu, 0); | |
326 | vcpu->arch.sie_block->cputm = 0UL; | |
327 | vcpu->arch.sie_block->ckc = 0UL; | |
328 | vcpu->arch.sie_block->todpr = 0; | |
329 | memset(vcpu->arch.sie_block->gcr, 0, 16 * sizeof(__u64)); | |
330 | vcpu->arch.sie_block->gcr[0] = 0xE0UL; | |
331 | vcpu->arch.sie_block->gcr[14] = 0xC2000000UL; | |
332 | vcpu->arch.guest_fpregs.fpc = 0; | |
333 | asm volatile("lfpc %0" : : "Q" (vcpu->arch.guest_fpregs.fpc)); | |
334 | vcpu->arch.sie_block->gbea = 1; | |
335 | } | |
336 | ||
337 | int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu) | |
338 | { | |
339 | atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH | | |
340 | CPUSTAT_SM | | |
341 | CPUSTAT_STOPPED); | |
342 | vcpu->arch.sie_block->ecb = 6; | |
343 | vcpu->arch.sie_block->eca = 0xC1002001U; | |
344 | vcpu->arch.sie_block->fac = (int) (long) facilities; | |
345 | hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS); | |
346 | tasklet_init(&vcpu->arch.tasklet, kvm_s390_tasklet, | |
347 | (unsigned long) vcpu); | |
348 | vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup; | |
349 | get_cpu_id(&vcpu->arch.cpu_id); | |
350 | vcpu->arch.cpu_id.version = 0xff; | |
351 | return 0; | |
352 | } | |
353 | ||
354 | struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, | |
355 | unsigned int id) | |
356 | { | |
357 | struct kvm_vcpu *vcpu; | |
358 | int rc = -EINVAL; | |
359 | ||
360 | if (id >= KVM_MAX_VCPUS) | |
361 | goto out; | |
362 | ||
363 | rc = -ENOMEM; | |
364 | ||
365 | vcpu = kzalloc(sizeof(struct kvm_vcpu), GFP_KERNEL); | |
366 | if (!vcpu) | |
367 | goto out; | |
368 | ||
369 | vcpu->arch.sie_block = (struct kvm_s390_sie_block *) | |
370 | get_zeroed_page(GFP_KERNEL); | |
371 | ||
372 | if (!vcpu->arch.sie_block) | |
373 | goto out_free_cpu; | |
374 | ||
375 | vcpu->arch.sie_block->icpua = id; | |
376 | if (!kvm_is_ucontrol(kvm)) { | |
377 | if (!kvm->arch.sca) { | |
378 | WARN_ON_ONCE(1); | |
379 | goto out_free_cpu; | |
380 | } | |
381 | if (!kvm->arch.sca->cpu[id].sda) | |
382 | kvm->arch.sca->cpu[id].sda = | |
383 | (__u64) vcpu->arch.sie_block; | |
384 | vcpu->arch.sie_block->scaoh = | |
385 | (__u32)(((__u64)kvm->arch.sca) >> 32); | |
386 | vcpu->arch.sie_block->scaol = (__u32)(__u64)kvm->arch.sca; | |
387 | set_bit(63 - id, (unsigned long *) &kvm->arch.sca->mcn); | |
388 | } | |
389 | ||
390 | spin_lock_init(&vcpu->arch.local_int.lock); | |
391 | INIT_LIST_HEAD(&vcpu->arch.local_int.list); | |
392 | vcpu->arch.local_int.float_int = &kvm->arch.float_int; | |
393 | spin_lock(&kvm->arch.float_int.lock); | |
394 | kvm->arch.float_int.local_int[id] = &vcpu->arch.local_int; | |
395 | init_waitqueue_head(&vcpu->arch.local_int.wq); | |
396 | vcpu->arch.local_int.cpuflags = &vcpu->arch.sie_block->cpuflags; | |
397 | spin_unlock(&kvm->arch.float_int.lock); | |
398 | ||
399 | rc = kvm_vcpu_init(vcpu, kvm, id); | |
400 | if (rc) | |
401 | goto out_free_sie_block; | |
402 | VM_EVENT(kvm, 3, "create cpu %d at %p, sie block at %p", id, vcpu, | |
403 | vcpu->arch.sie_block); | |
404 | ||
405 | return vcpu; | |
406 | out_free_sie_block: | |
407 | free_page((unsigned long)(vcpu->arch.sie_block)); | |
408 | out_free_cpu: | |
409 | kfree(vcpu); | |
410 | out: | |
411 | return ERR_PTR(rc); | |
412 | } | |
413 | ||
414 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu) | |
415 | { | |
416 | /* kvm common code refers to this, but never calls it */ | |
417 | BUG(); | |
418 | return 0; | |
419 | } | |
420 | ||
421 | static int kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu) | |
422 | { | |
423 | kvm_s390_vcpu_initial_reset(vcpu); | |
424 | return 0; | |
425 | } | |
426 | ||
427 | int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
428 | { | |
429 | memcpy(&vcpu->arch.guest_gprs, ®s->gprs, sizeof(regs->gprs)); | |
430 | return 0; | |
431 | } | |
432 | ||
433 | int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
434 | { | |
435 | memcpy(®s->gprs, &vcpu->arch.guest_gprs, sizeof(regs->gprs)); | |
436 | return 0; | |
437 | } | |
438 | ||
439 | int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, | |
440 | struct kvm_sregs *sregs) | |
441 | { | |
442 | memcpy(&vcpu->arch.guest_acrs, &sregs->acrs, sizeof(sregs->acrs)); | |
443 | memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs)); | |
444 | restore_access_regs(vcpu->arch.guest_acrs); | |
445 | return 0; | |
446 | } | |
447 | ||
448 | int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, | |
449 | struct kvm_sregs *sregs) | |
450 | { | |
451 | memcpy(&sregs->acrs, &vcpu->arch.guest_acrs, sizeof(sregs->acrs)); | |
452 | memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs)); | |
453 | return 0; | |
454 | } | |
455 | ||
456 | int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
457 | { | |
458 | memcpy(&vcpu->arch.guest_fpregs.fprs, &fpu->fprs, sizeof(fpu->fprs)); | |
459 | vcpu->arch.guest_fpregs.fpc = fpu->fpc; | |
460 | restore_fp_regs(&vcpu->arch.guest_fpregs); | |
461 | return 0; | |
462 | } | |
463 | ||
464 | int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
465 | { | |
466 | memcpy(&fpu->fprs, &vcpu->arch.guest_fpregs.fprs, sizeof(fpu->fprs)); | |
467 | fpu->fpc = vcpu->arch.guest_fpregs.fpc; | |
468 | return 0; | |
469 | } | |
470 | ||
471 | static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw) | |
472 | { | |
473 | int rc = 0; | |
474 | ||
475 | if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_STOPPED)) | |
476 | rc = -EBUSY; | |
477 | else { | |
478 | vcpu->run->psw_mask = psw.mask; | |
479 | vcpu->run->psw_addr = psw.addr; | |
480 | } | |
481 | return rc; | |
482 | } | |
483 | ||
484 | int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, | |
485 | struct kvm_translation *tr) | |
486 | { | |
487 | return -EINVAL; /* not implemented yet */ | |
488 | } | |
489 | ||
490 | int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, | |
491 | struct kvm_guest_debug *dbg) | |
492 | { | |
493 | return -EINVAL; /* not implemented yet */ | |
494 | } | |
495 | ||
496 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, | |
497 | struct kvm_mp_state *mp_state) | |
498 | { | |
499 | return -EINVAL; /* not implemented yet */ | |
500 | } | |
501 | ||
502 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
503 | struct kvm_mp_state *mp_state) | |
504 | { | |
505 | return -EINVAL; /* not implemented yet */ | |
506 | } | |
507 | ||
508 | static int __vcpu_run(struct kvm_vcpu *vcpu) | |
509 | { | |
510 | int rc; | |
511 | ||
512 | memcpy(&vcpu->arch.sie_block->gg14, &vcpu->arch.guest_gprs[14], 16); | |
513 | ||
514 | if (need_resched()) | |
515 | schedule(); | |
516 | ||
517 | if (test_thread_flag(TIF_MCCK_PENDING)) | |
518 | s390_handle_mcck(); | |
519 | ||
520 | if (!kvm_is_ucontrol(vcpu->kvm)) | |
521 | kvm_s390_deliver_pending_interrupts(vcpu); | |
522 | ||
523 | vcpu->arch.sie_block->icptcode = 0; | |
524 | local_irq_disable(); | |
525 | kvm_guest_enter(); | |
526 | local_irq_enable(); | |
527 | VCPU_EVENT(vcpu, 6, "entering sie flags %x", | |
528 | atomic_read(&vcpu->arch.sie_block->cpuflags)); | |
529 | rc = sie64a(vcpu->arch.sie_block, vcpu->arch.guest_gprs); | |
530 | if (rc) { | |
531 | if (kvm_is_ucontrol(vcpu->kvm)) { | |
532 | rc = SIE_INTERCEPT_UCONTROL; | |
533 | } else { | |
534 | VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction"); | |
535 | kvm_s390_inject_program_int(vcpu, PGM_ADDRESSING); | |
536 | rc = 0; | |
537 | } | |
538 | } | |
539 | VCPU_EVENT(vcpu, 6, "exit sie icptcode %d", | |
540 | vcpu->arch.sie_block->icptcode); | |
541 | local_irq_disable(); | |
542 | kvm_guest_exit(); | |
543 | local_irq_enable(); | |
544 | ||
545 | memcpy(&vcpu->arch.guest_gprs[14], &vcpu->arch.sie_block->gg14, 16); | |
546 | return rc; | |
547 | } | |
548 | ||
549 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |
550 | { | |
551 | int rc; | |
552 | sigset_t sigsaved; | |
553 | ||
554 | rerun_vcpu: | |
555 | if (vcpu->sigset_active) | |
556 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
557 | ||
558 | atomic_clear_mask(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags); | |
559 | ||
560 | BUG_ON(vcpu->kvm->arch.float_int.local_int[vcpu->vcpu_id] == NULL); | |
561 | ||
562 | switch (kvm_run->exit_reason) { | |
563 | case KVM_EXIT_S390_SIEIC: | |
564 | case KVM_EXIT_UNKNOWN: | |
565 | case KVM_EXIT_INTR: | |
566 | case KVM_EXIT_S390_RESET: | |
567 | case KVM_EXIT_S390_UCONTROL: | |
568 | break; | |
569 | default: | |
570 | BUG(); | |
571 | } | |
572 | ||
573 | vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask; | |
574 | vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr; | |
575 | ||
576 | might_fault(); | |
577 | ||
578 | do { | |
579 | rc = __vcpu_run(vcpu); | |
580 | if (rc) | |
581 | break; | |
582 | if (kvm_is_ucontrol(vcpu->kvm)) | |
583 | rc = -EOPNOTSUPP; | |
584 | else | |
585 | rc = kvm_handle_sie_intercept(vcpu); | |
586 | } while (!signal_pending(current) && !rc); | |
587 | ||
588 | if (rc == SIE_INTERCEPT_RERUNVCPU) | |
589 | goto rerun_vcpu; | |
590 | ||
591 | if (signal_pending(current) && !rc) { | |
592 | kvm_run->exit_reason = KVM_EXIT_INTR; | |
593 | rc = -EINTR; | |
594 | } | |
595 | ||
596 | #ifdef CONFIG_KVM_S390_UCONTROL | |
597 | if (rc == SIE_INTERCEPT_UCONTROL) { | |
598 | kvm_run->exit_reason = KVM_EXIT_S390_UCONTROL; | |
599 | kvm_run->s390_ucontrol.trans_exc_code = | |
600 | current->thread.gmap_addr; | |
601 | kvm_run->s390_ucontrol.pgm_code = 0x10; | |
602 | rc = 0; | |
603 | } | |
604 | #endif | |
605 | ||
606 | if (rc == -EOPNOTSUPP) { | |
607 | /* intercept cannot be handled in-kernel, prepare kvm-run */ | |
608 | kvm_run->exit_reason = KVM_EXIT_S390_SIEIC; | |
609 | kvm_run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode; | |
610 | kvm_run->s390_sieic.ipa = vcpu->arch.sie_block->ipa; | |
611 | kvm_run->s390_sieic.ipb = vcpu->arch.sie_block->ipb; | |
612 | rc = 0; | |
613 | } | |
614 | ||
615 | if (rc == -EREMOTE) { | |
616 | /* intercept was handled, but userspace support is needed | |
617 | * kvm_run has been prepared by the handler */ | |
618 | rc = 0; | |
619 | } | |
620 | ||
621 | kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask; | |
622 | kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr; | |
623 | ||
624 | if (vcpu->sigset_active) | |
625 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
626 | ||
627 | vcpu->stat.exit_userspace++; | |
628 | return rc; | |
629 | } | |
630 | ||
631 | static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, void *from, | |
632 | unsigned long n, int prefix) | |
633 | { | |
634 | if (prefix) | |
635 | return copy_to_guest(vcpu, guestdest, from, n); | |
636 | else | |
637 | return copy_to_guest_absolute(vcpu, guestdest, from, n); | |
638 | } | |
639 | ||
640 | /* | |
641 | * store status at address | |
642 | * we use have two special cases: | |
643 | * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit | |
644 | * KVM_S390_STORE_STATUS_PREFIXED: -> prefix | |
645 | */ | |
646 | int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr) | |
647 | { | |
648 | unsigned char archmode = 1; | |
649 | int prefix; | |
650 | ||
651 | if (addr == KVM_S390_STORE_STATUS_NOADDR) { | |
652 | if (copy_to_guest_absolute(vcpu, 163ul, &archmode, 1)) | |
653 | return -EFAULT; | |
654 | addr = SAVE_AREA_BASE; | |
655 | prefix = 0; | |
656 | } else if (addr == KVM_S390_STORE_STATUS_PREFIXED) { | |
657 | if (copy_to_guest(vcpu, 163ul, &archmode, 1)) | |
658 | return -EFAULT; | |
659 | addr = SAVE_AREA_BASE; | |
660 | prefix = 1; | |
661 | } else | |
662 | prefix = 0; | |
663 | ||
664 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, fp_regs), | |
665 | vcpu->arch.guest_fpregs.fprs, 128, prefix)) | |
666 | return -EFAULT; | |
667 | ||
668 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, gp_regs), | |
669 | vcpu->arch.guest_gprs, 128, prefix)) | |
670 | return -EFAULT; | |
671 | ||
672 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, psw), | |
673 | &vcpu->arch.sie_block->gpsw, 16, prefix)) | |
674 | return -EFAULT; | |
675 | ||
676 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, pref_reg), | |
677 | &vcpu->arch.sie_block->prefix, 4, prefix)) | |
678 | return -EFAULT; | |
679 | ||
680 | if (__guestcopy(vcpu, | |
681 | addr + offsetof(struct save_area, fp_ctrl_reg), | |
682 | &vcpu->arch.guest_fpregs.fpc, 4, prefix)) | |
683 | return -EFAULT; | |
684 | ||
685 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, tod_reg), | |
686 | &vcpu->arch.sie_block->todpr, 4, prefix)) | |
687 | return -EFAULT; | |
688 | ||
689 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, timer), | |
690 | &vcpu->arch.sie_block->cputm, 8, prefix)) | |
691 | return -EFAULT; | |
692 | ||
693 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, clk_cmp), | |
694 | &vcpu->arch.sie_block->ckc, 8, prefix)) | |
695 | return -EFAULT; | |
696 | ||
697 | if (__guestcopy(vcpu, addr + offsetof(struct save_area, acc_regs), | |
698 | &vcpu->arch.guest_acrs, 64, prefix)) | |
699 | return -EFAULT; | |
700 | ||
701 | if (__guestcopy(vcpu, | |
702 | addr + offsetof(struct save_area, ctrl_regs), | |
703 | &vcpu->arch.sie_block->gcr, 128, prefix)) | |
704 | return -EFAULT; | |
705 | return 0; | |
706 | } | |
707 | ||
708 | long kvm_arch_vcpu_ioctl(struct file *filp, | |
709 | unsigned int ioctl, unsigned long arg) | |
710 | { | |
711 | struct kvm_vcpu *vcpu = filp->private_data; | |
712 | void __user *argp = (void __user *)arg; | |
713 | long r; | |
714 | ||
715 | switch (ioctl) { | |
716 | case KVM_S390_INTERRUPT: { | |
717 | struct kvm_s390_interrupt s390int; | |
718 | ||
719 | r = -EFAULT; | |
720 | if (copy_from_user(&s390int, argp, sizeof(s390int))) | |
721 | break; | |
722 | r = kvm_s390_inject_vcpu(vcpu, &s390int); | |
723 | break; | |
724 | } | |
725 | case KVM_S390_STORE_STATUS: | |
726 | r = kvm_s390_vcpu_store_status(vcpu, arg); | |
727 | break; | |
728 | case KVM_S390_SET_INITIAL_PSW: { | |
729 | psw_t psw; | |
730 | ||
731 | r = -EFAULT; | |
732 | if (copy_from_user(&psw, argp, sizeof(psw))) | |
733 | break; | |
734 | r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw); | |
735 | break; | |
736 | } | |
737 | case KVM_S390_INITIAL_RESET: | |
738 | r = kvm_arch_vcpu_ioctl_initial_reset(vcpu); | |
739 | break; | |
740 | #ifdef CONFIG_KVM_S390_UCONTROL | |
741 | case KVM_S390_UCAS_MAP: { | |
742 | struct kvm_s390_ucas_mapping ucasmap; | |
743 | ||
744 | if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) { | |
745 | r = -EFAULT; | |
746 | break; | |
747 | } | |
748 | ||
749 | if (!kvm_is_ucontrol(vcpu->kvm)) { | |
750 | r = -EINVAL; | |
751 | break; | |
752 | } | |
753 | ||
754 | r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr, | |
755 | ucasmap.vcpu_addr, ucasmap.length); | |
756 | break; | |
757 | } | |
758 | case KVM_S390_UCAS_UNMAP: { | |
759 | struct kvm_s390_ucas_mapping ucasmap; | |
760 | ||
761 | if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) { | |
762 | r = -EFAULT; | |
763 | break; | |
764 | } | |
765 | ||
766 | if (!kvm_is_ucontrol(vcpu->kvm)) { | |
767 | r = -EINVAL; | |
768 | break; | |
769 | } | |
770 | ||
771 | r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr, | |
772 | ucasmap.length); | |
773 | break; | |
774 | } | |
775 | #endif | |
776 | case KVM_S390_VCPU_FAULT: { | |
777 | r = gmap_fault(arg, vcpu->arch.gmap); | |
778 | if (!IS_ERR_VALUE(r)) | |
779 | r = 0; | |
780 | break; | |
781 | } | |
782 | default: | |
783 | r = -ENOTTY; | |
784 | } | |
785 | return r; | |
786 | } | |
787 | ||
788 | int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf) | |
789 | { | |
790 | #ifdef CONFIG_KVM_S390_UCONTROL | |
791 | if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET) | |
792 | && (kvm_is_ucontrol(vcpu->kvm))) { | |
793 | vmf->page = virt_to_page(vcpu->arch.sie_block); | |
794 | get_page(vmf->page); | |
795 | return 0; | |
796 | } | |
797 | #endif | |
798 | return VM_FAULT_SIGBUS; | |
799 | } | |
800 | ||
801 | /* Section: memory related */ | |
802 | int kvm_arch_prepare_memory_region(struct kvm *kvm, | |
803 | struct kvm_memory_slot *memslot, | |
804 | struct kvm_memory_slot old, | |
805 | struct kvm_userspace_memory_region *mem, | |
806 | int user_alloc) | |
807 | { | |
808 | /* A few sanity checks. We can have exactly one memory slot which has | |
809 | to start at guest virtual zero and which has to be located at a | |
810 | page boundary in userland and which has to end at a page boundary. | |
811 | The memory in userland is ok to be fragmented into various different | |
812 | vmas. It is okay to mmap() and munmap() stuff in this slot after | |
813 | doing this call at any time */ | |
814 | ||
815 | if (mem->slot) | |
816 | return -EINVAL; | |
817 | ||
818 | if (mem->guest_phys_addr) | |
819 | return -EINVAL; | |
820 | ||
821 | if (mem->userspace_addr & 0xffffful) | |
822 | return -EINVAL; | |
823 | ||
824 | if (mem->memory_size & 0xffffful) | |
825 | return -EINVAL; | |
826 | ||
827 | if (!user_alloc) | |
828 | return -EINVAL; | |
829 | ||
830 | return 0; | |
831 | } | |
832 | ||
833 | void kvm_arch_commit_memory_region(struct kvm *kvm, | |
834 | struct kvm_userspace_memory_region *mem, | |
835 | struct kvm_memory_slot old, | |
836 | int user_alloc) | |
837 | { | |
838 | int rc; | |
839 | ||
840 | ||
841 | rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr, | |
842 | mem->guest_phys_addr, mem->memory_size); | |
843 | if (rc) | |
844 | printk(KERN_WARNING "kvm-s390: failed to commit memory region\n"); | |
845 | return; | |
846 | } | |
847 | ||
848 | void kvm_arch_flush_shadow(struct kvm *kvm) | |
849 | { | |
850 | } | |
851 | ||
852 | static int __init kvm_s390_init(void) | |
853 | { | |
854 | int ret; | |
855 | ret = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE); | |
856 | if (ret) | |
857 | return ret; | |
858 | ||
859 | /* | |
860 | * guests can ask for up to 255+1 double words, we need a full page | |
861 | * to hold the maximum amount of facilities. On the other hand, we | |
862 | * only set facilities that are known to work in KVM. | |
863 | */ | |
864 | facilities = (unsigned long long *) get_zeroed_page(GFP_KERNEL|GFP_DMA); | |
865 | if (!facilities) { | |
866 | kvm_exit(); | |
867 | return -ENOMEM; | |
868 | } | |
869 | memcpy(facilities, S390_lowcore.stfle_fac_list, 16); | |
870 | facilities[0] &= 0xff00fff3f47c0000ULL; | |
871 | facilities[1] &= 0x201c000000000000ULL; | |
872 | return 0; | |
873 | } | |
874 | ||
875 | static void __exit kvm_s390_exit(void) | |
876 | { | |
877 | free_page((unsigned long) facilities); | |
878 | kvm_exit(); | |
879 | } | |
880 | ||
881 | module_init(kvm_s390_init); | |
882 | module_exit(kvm_s390_exit); |