xref: /linux/arch/s390/kvm/kvm-s390.c (revision 79d2e1919a2728ef49d938eb20ebd5903c14dfb0)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * hosting IBM Z kernel virtual machines (s390x)
4  *
5  * Copyright IBM Corp. 2008, 2020
6  *
7  *    Author(s): Carsten Otte <cotte@de.ibm.com>
8  *               Christian Borntraeger <borntraeger@de.ibm.com>
9  *               Christian Ehrhardt <ehrhardt@de.ibm.com>
10  *               Jason J. Herne <jjherne@us.ibm.com>
11  */
12 
13 #define KMSG_COMPONENT "kvm-s390"
14 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
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/mman.h>
24 #include <linux/module.h>
25 #include <linux/moduleparam.h>
26 #include <linux/random.h>
27 #include <linux/slab.h>
28 #include <linux/timer.h>
29 #include <linux/vmalloc.h>
30 #include <linux/bitmap.h>
31 #include <linux/sched/signal.h>
32 #include <linux/string.h>
33 #include <linux/pgtable.h>
34 #include <linux/mmu_notifier.h>
35 
36 #include <asm/access-regs.h>
37 #include <asm/asm-offsets.h>
38 #include <asm/lowcore.h>
39 #include <asm/stp.h>
40 #include <asm/gmap.h>
41 #include <asm/nmi.h>
42 #include <asm/isc.h>
43 #include <asm/sclp.h>
44 #include <asm/cpacf.h>
45 #include <asm/timex.h>
46 #include <asm/asm.h>
47 #include <asm/fpu.h>
48 #include <asm/ap.h>
49 #include <asm/uv.h>
50 #include "kvm-s390.h"
51 #include "gaccess.h"
52 #include "pci.h"
53 
54 #define CREATE_TRACE_POINTS
55 #include "trace.h"
56 #include "trace-s390.h"
57 
58 #define MEM_OP_MAX_SIZE 65536	/* Maximum transfer size for KVM_S390_MEM_OP */
59 #define LOCAL_IRQS 32
60 #define VCPU_IRQS_MAX_BUF (sizeof(struct kvm_s390_irq) * \
61 			   (KVM_MAX_VCPUS + LOCAL_IRQS))
62 
63 const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
64 	KVM_GENERIC_VM_STATS(),
65 	STATS_DESC_COUNTER(VM, inject_io),
66 	STATS_DESC_COUNTER(VM, inject_float_mchk),
67 	STATS_DESC_COUNTER(VM, inject_pfault_done),
68 	STATS_DESC_COUNTER(VM, inject_service_signal),
69 	STATS_DESC_COUNTER(VM, inject_virtio),
70 	STATS_DESC_COUNTER(VM, aen_forward),
71 	STATS_DESC_COUNTER(VM, gmap_shadow_reuse),
72 	STATS_DESC_COUNTER(VM, gmap_shadow_create),
73 	STATS_DESC_COUNTER(VM, gmap_shadow_r1_entry),
74 	STATS_DESC_COUNTER(VM, gmap_shadow_r2_entry),
75 	STATS_DESC_COUNTER(VM, gmap_shadow_r3_entry),
76 	STATS_DESC_COUNTER(VM, gmap_shadow_sg_entry),
77 	STATS_DESC_COUNTER(VM, gmap_shadow_pg_entry),
78 };
79 
80 const struct kvm_stats_header kvm_vm_stats_header = {
81 	.name_size = KVM_STATS_NAME_SIZE,
82 	.num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
83 	.id_offset = sizeof(struct kvm_stats_header),
84 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
85 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
86 		       sizeof(kvm_vm_stats_desc),
87 };
88 
89 const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
90 	KVM_GENERIC_VCPU_STATS(),
91 	STATS_DESC_COUNTER(VCPU, exit_userspace),
92 	STATS_DESC_COUNTER(VCPU, exit_null),
93 	STATS_DESC_COUNTER(VCPU, exit_external_request),
94 	STATS_DESC_COUNTER(VCPU, exit_io_request),
95 	STATS_DESC_COUNTER(VCPU, exit_external_interrupt),
96 	STATS_DESC_COUNTER(VCPU, exit_stop_request),
97 	STATS_DESC_COUNTER(VCPU, exit_validity),
98 	STATS_DESC_COUNTER(VCPU, exit_instruction),
99 	STATS_DESC_COUNTER(VCPU, exit_pei),
100 	STATS_DESC_COUNTER(VCPU, halt_no_poll_steal),
101 	STATS_DESC_COUNTER(VCPU, instruction_lctl),
102 	STATS_DESC_COUNTER(VCPU, instruction_lctlg),
103 	STATS_DESC_COUNTER(VCPU, instruction_stctl),
104 	STATS_DESC_COUNTER(VCPU, instruction_stctg),
105 	STATS_DESC_COUNTER(VCPU, exit_program_interruption),
106 	STATS_DESC_COUNTER(VCPU, exit_instr_and_program),
107 	STATS_DESC_COUNTER(VCPU, exit_operation_exception),
108 	STATS_DESC_COUNTER(VCPU, deliver_ckc),
109 	STATS_DESC_COUNTER(VCPU, deliver_cputm),
110 	STATS_DESC_COUNTER(VCPU, deliver_external_call),
111 	STATS_DESC_COUNTER(VCPU, deliver_emergency_signal),
112 	STATS_DESC_COUNTER(VCPU, deliver_service_signal),
113 	STATS_DESC_COUNTER(VCPU, deliver_virtio),
114 	STATS_DESC_COUNTER(VCPU, deliver_stop_signal),
115 	STATS_DESC_COUNTER(VCPU, deliver_prefix_signal),
116 	STATS_DESC_COUNTER(VCPU, deliver_restart_signal),
117 	STATS_DESC_COUNTER(VCPU, deliver_program),
118 	STATS_DESC_COUNTER(VCPU, deliver_io),
119 	STATS_DESC_COUNTER(VCPU, deliver_machine_check),
120 	STATS_DESC_COUNTER(VCPU, exit_wait_state),
121 	STATS_DESC_COUNTER(VCPU, inject_ckc),
122 	STATS_DESC_COUNTER(VCPU, inject_cputm),
123 	STATS_DESC_COUNTER(VCPU, inject_external_call),
124 	STATS_DESC_COUNTER(VCPU, inject_emergency_signal),
125 	STATS_DESC_COUNTER(VCPU, inject_mchk),
126 	STATS_DESC_COUNTER(VCPU, inject_pfault_init),
127 	STATS_DESC_COUNTER(VCPU, inject_program),
128 	STATS_DESC_COUNTER(VCPU, inject_restart),
129 	STATS_DESC_COUNTER(VCPU, inject_set_prefix),
130 	STATS_DESC_COUNTER(VCPU, inject_stop_signal),
131 	STATS_DESC_COUNTER(VCPU, instruction_epsw),
132 	STATS_DESC_COUNTER(VCPU, instruction_gs),
133 	STATS_DESC_COUNTER(VCPU, instruction_io_other),
134 	STATS_DESC_COUNTER(VCPU, instruction_lpsw),
135 	STATS_DESC_COUNTER(VCPU, instruction_lpswe),
136 	STATS_DESC_COUNTER(VCPU, instruction_lpswey),
137 	STATS_DESC_COUNTER(VCPU, instruction_pfmf),
138 	STATS_DESC_COUNTER(VCPU, instruction_ptff),
139 	STATS_DESC_COUNTER(VCPU, instruction_sck),
140 	STATS_DESC_COUNTER(VCPU, instruction_sckpf),
141 	STATS_DESC_COUNTER(VCPU, instruction_stidp),
142 	STATS_DESC_COUNTER(VCPU, instruction_spx),
143 	STATS_DESC_COUNTER(VCPU, instruction_stpx),
144 	STATS_DESC_COUNTER(VCPU, instruction_stap),
145 	STATS_DESC_COUNTER(VCPU, instruction_iske),
146 	STATS_DESC_COUNTER(VCPU, instruction_ri),
147 	STATS_DESC_COUNTER(VCPU, instruction_rrbe),
148 	STATS_DESC_COUNTER(VCPU, instruction_sske),
149 	STATS_DESC_COUNTER(VCPU, instruction_ipte_interlock),
150 	STATS_DESC_COUNTER(VCPU, instruction_stsi),
151 	STATS_DESC_COUNTER(VCPU, instruction_stfl),
152 	STATS_DESC_COUNTER(VCPU, instruction_tb),
153 	STATS_DESC_COUNTER(VCPU, instruction_tpi),
154 	STATS_DESC_COUNTER(VCPU, instruction_tprot),
155 	STATS_DESC_COUNTER(VCPU, instruction_tsch),
156 	STATS_DESC_COUNTER(VCPU, instruction_sie),
157 	STATS_DESC_COUNTER(VCPU, instruction_essa),
158 	STATS_DESC_COUNTER(VCPU, instruction_sthyi),
159 	STATS_DESC_COUNTER(VCPU, instruction_sigp_sense),
160 	STATS_DESC_COUNTER(VCPU, instruction_sigp_sense_running),
161 	STATS_DESC_COUNTER(VCPU, instruction_sigp_external_call),
162 	STATS_DESC_COUNTER(VCPU, instruction_sigp_emergency),
163 	STATS_DESC_COUNTER(VCPU, instruction_sigp_cond_emergency),
164 	STATS_DESC_COUNTER(VCPU, instruction_sigp_start),
165 	STATS_DESC_COUNTER(VCPU, instruction_sigp_stop),
166 	STATS_DESC_COUNTER(VCPU, instruction_sigp_stop_store_status),
167 	STATS_DESC_COUNTER(VCPU, instruction_sigp_store_status),
168 	STATS_DESC_COUNTER(VCPU, instruction_sigp_store_adtl_status),
169 	STATS_DESC_COUNTER(VCPU, instruction_sigp_arch),
170 	STATS_DESC_COUNTER(VCPU, instruction_sigp_prefix),
171 	STATS_DESC_COUNTER(VCPU, instruction_sigp_restart),
172 	STATS_DESC_COUNTER(VCPU, instruction_sigp_init_cpu_reset),
173 	STATS_DESC_COUNTER(VCPU, instruction_sigp_cpu_reset),
174 	STATS_DESC_COUNTER(VCPU, instruction_sigp_unknown),
175 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_10),
176 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_44),
177 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_9c),
178 	STATS_DESC_COUNTER(VCPU, diag_9c_ignored),
179 	STATS_DESC_COUNTER(VCPU, diag_9c_forward),
180 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_258),
181 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_308),
182 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_500),
183 	STATS_DESC_COUNTER(VCPU, instruction_diagnose_other),
184 	STATS_DESC_COUNTER(VCPU, pfault_sync)
185 };
186 
187 const struct kvm_stats_header kvm_vcpu_stats_header = {
188 	.name_size = KVM_STATS_NAME_SIZE,
189 	.num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
190 	.id_offset = sizeof(struct kvm_stats_header),
191 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
192 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
193 		       sizeof(kvm_vcpu_stats_desc),
194 };
195 
196 /* allow nested virtualization in KVM (if enabled by user space) */
197 static int nested;
198 module_param(nested, int, S_IRUGO);
199 MODULE_PARM_DESC(nested, "Nested virtualization support");
200 
201 /* allow 1m huge page guest backing, if !nested */
202 static int hpage;
203 module_param(hpage, int, 0444);
204 MODULE_PARM_DESC(hpage, "1m huge page backing support");
205 
206 /* maximum percentage of steal time for polling.  >100 is treated like 100 */
207 static u8 halt_poll_max_steal = 10;
208 module_param(halt_poll_max_steal, byte, 0644);
209 MODULE_PARM_DESC(halt_poll_max_steal, "Maximum percentage of steal time to allow polling");
210 
211 /* if set to true, the GISA will be initialized and used if available */
212 static bool use_gisa  = true;
213 module_param(use_gisa, bool, 0644);
214 MODULE_PARM_DESC(use_gisa, "Use the GISA if the host supports it.");
215 
216 /* maximum diag9c forwarding per second */
217 unsigned int diag9c_forwarding_hz;
218 module_param(diag9c_forwarding_hz, uint, 0644);
219 MODULE_PARM_DESC(diag9c_forwarding_hz, "Maximum diag9c forwarding per second, 0 to turn off");
220 
221 /*
222  * allow asynchronous deinit for protected guests; enable by default since
223  * the feature is opt-in anyway
224  */
225 static int async_destroy = 1;
226 module_param(async_destroy, int, 0444);
227 MODULE_PARM_DESC(async_destroy, "Asynchronous destroy for protected guests");
228 
229 /*
230  * For now we handle at most 16 double words as this is what the s390 base
231  * kernel handles and stores in the prefix page. If we ever need to go beyond
232  * this, this requires changes to code, but the external uapi can stay.
233  */
234 #define SIZE_INTERNAL 16
235 
236 /*
237  * Base feature mask that defines default mask for facilities. Consists of the
238  * defines in FACILITIES_KVM and the non-hypervisor managed bits.
239  */
240 static unsigned long kvm_s390_fac_base[SIZE_INTERNAL] = { FACILITIES_KVM };
241 /*
242  * Extended feature mask. Consists of the defines in FACILITIES_KVM_CPUMODEL
243  * and defines the facilities that can be enabled via a cpu model.
244  */
245 static unsigned long kvm_s390_fac_ext[SIZE_INTERNAL] = { FACILITIES_KVM_CPUMODEL };
246 
247 static unsigned long kvm_s390_fac_size(void)
248 {
249 	BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_MASK_SIZE_U64);
250 	BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_LIST_SIZE_U64);
251 	BUILD_BUG_ON(SIZE_INTERNAL * sizeof(unsigned long) >
252 		sizeof(stfle_fac_list));
253 
254 	return SIZE_INTERNAL;
255 }
256 
257 /* available cpu features supported by kvm */
258 static DECLARE_BITMAP(kvm_s390_available_cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
259 /* available subfunctions indicated via query / "test bit" */
260 static struct kvm_s390_vm_cpu_subfunc kvm_s390_available_subfunc;
261 
262 static struct gmap_notifier gmap_notifier;
263 static struct gmap_notifier vsie_gmap_notifier;
264 debug_info_t *kvm_s390_dbf;
265 debug_info_t *kvm_s390_dbf_uv;
266 
267 /* Section: not file related */
268 /* forward declarations */
269 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
270 			      unsigned long end);
271 static int sca_switch_to_extended(struct kvm *kvm);
272 
273 static void kvm_clock_sync_scb(struct kvm_s390_sie_block *scb, u64 delta)
274 {
275 	u8 delta_idx = 0;
276 
277 	/*
278 	 * The TOD jumps by delta, we have to compensate this by adding
279 	 * -delta to the epoch.
280 	 */
281 	delta = -delta;
282 
283 	/* sign-extension - we're adding to signed values below */
284 	if ((s64)delta < 0)
285 		delta_idx = -1;
286 
287 	scb->epoch += delta;
288 	if (scb->ecd & ECD_MEF) {
289 		scb->epdx += delta_idx;
290 		if (scb->epoch < delta)
291 			scb->epdx += 1;
292 	}
293 }
294 
295 /*
296  * This callback is executed during stop_machine(). All CPUs are therefore
297  * temporarily stopped. In order not to change guest behavior, we have to
298  * disable preemption whenever we touch the epoch of kvm and the VCPUs,
299  * so a CPU won't be stopped while calculating with the epoch.
300  */
301 static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
302 			  void *v)
303 {
304 	struct kvm *kvm;
305 	struct kvm_vcpu *vcpu;
306 	unsigned long i;
307 	unsigned long long *delta = v;
308 
309 	list_for_each_entry(kvm, &vm_list, vm_list) {
310 		kvm_for_each_vcpu(i, vcpu, kvm) {
311 			kvm_clock_sync_scb(vcpu->arch.sie_block, *delta);
312 			if (i == 0) {
313 				kvm->arch.epoch = vcpu->arch.sie_block->epoch;
314 				kvm->arch.epdx = vcpu->arch.sie_block->epdx;
315 			}
316 			if (vcpu->arch.cputm_enabled)
317 				vcpu->arch.cputm_start += *delta;
318 			if (vcpu->arch.vsie_block)
319 				kvm_clock_sync_scb(vcpu->arch.vsie_block,
320 						   *delta);
321 		}
322 	}
323 	return NOTIFY_OK;
324 }
325 
326 static struct notifier_block kvm_clock_notifier = {
327 	.notifier_call = kvm_clock_sync,
328 };
329 
330 static void allow_cpu_feat(unsigned long nr)
331 {
332 	set_bit_inv(nr, kvm_s390_available_cpu_feat);
333 }
334 
335 static inline int plo_test_bit(unsigned char nr)
336 {
337 	unsigned long function = (unsigned long)nr | 0x100;
338 	int cc;
339 
340 	asm volatile(
341 		"	lgr	0,%[function]\n"
342 		/* Parameter registers are ignored for "test bit" */
343 		"	plo	0,0,0,0(0)\n"
344 		CC_IPM(cc)
345 		: CC_OUT(cc, cc)
346 		: [function] "d" (function)
347 		: CC_CLOBBER_LIST("0"));
348 	return CC_TRANSFORM(cc) == 0;
349 }
350 
351 static __always_inline void pfcr_query(u8 (*query)[16])
352 {
353 	asm volatile(
354 		"	lghi	0,0\n"
355 		"	.insn   rsy,0xeb0000000016,0,0,%[query]\n"
356 		: [query] "=QS" (*query)
357 		:
358 		: "cc", "0");
359 }
360 
361 static __always_inline void __sortl_query(u8 (*query)[32])
362 {
363 	asm volatile(
364 		"	lghi	0,0\n"
365 		"	la	1,%[query]\n"
366 		/* Parameter registers are ignored */
367 		"	.insn	rre,0xb9380000,2,4\n"
368 		: [query] "=R" (*query)
369 		:
370 		: "cc", "0", "1");
371 }
372 
373 static __always_inline void __dfltcc_query(u8 (*query)[32])
374 {
375 	asm volatile(
376 		"	lghi	0,0\n"
377 		"	la	1,%[query]\n"
378 		/* Parameter registers are ignored */
379 		"	.insn	rrf,0xb9390000,2,4,6,0\n"
380 		: [query] "=R" (*query)
381 		:
382 		: "cc", "0", "1");
383 }
384 
385 static void __init kvm_s390_cpu_feat_init(void)
386 {
387 	int i;
388 
389 	for (i = 0; i < 256; ++i) {
390 		if (plo_test_bit(i))
391 			kvm_s390_available_subfunc.plo[i >> 3] |= 0x80 >> (i & 7);
392 	}
393 
394 	if (test_facility(28)) /* TOD-clock steering */
395 		ptff(kvm_s390_available_subfunc.ptff,
396 		     sizeof(kvm_s390_available_subfunc.ptff),
397 		     PTFF_QAF);
398 
399 	if (test_facility(17)) { /* MSA */
400 		__cpacf_query(CPACF_KMAC, (cpacf_mask_t *)
401 			      kvm_s390_available_subfunc.kmac);
402 		__cpacf_query(CPACF_KMC, (cpacf_mask_t *)
403 			      kvm_s390_available_subfunc.kmc);
404 		__cpacf_query(CPACF_KM, (cpacf_mask_t *)
405 			      kvm_s390_available_subfunc.km);
406 		__cpacf_query(CPACF_KIMD, (cpacf_mask_t *)
407 			      kvm_s390_available_subfunc.kimd);
408 		__cpacf_query(CPACF_KLMD, (cpacf_mask_t *)
409 			      kvm_s390_available_subfunc.klmd);
410 	}
411 	if (test_facility(76)) /* MSA3 */
412 		__cpacf_query(CPACF_PCKMO, (cpacf_mask_t *)
413 			      kvm_s390_available_subfunc.pckmo);
414 	if (test_facility(77)) { /* MSA4 */
415 		__cpacf_query(CPACF_KMCTR, (cpacf_mask_t *)
416 			      kvm_s390_available_subfunc.kmctr);
417 		__cpacf_query(CPACF_KMF, (cpacf_mask_t *)
418 			      kvm_s390_available_subfunc.kmf);
419 		__cpacf_query(CPACF_KMO, (cpacf_mask_t *)
420 			      kvm_s390_available_subfunc.kmo);
421 		__cpacf_query(CPACF_PCC, (cpacf_mask_t *)
422 			      kvm_s390_available_subfunc.pcc);
423 	}
424 	if (test_facility(57)) /* MSA5 */
425 		__cpacf_query(CPACF_PRNO, (cpacf_mask_t *)
426 			      kvm_s390_available_subfunc.ppno);
427 
428 	if (test_facility(146)) /* MSA8 */
429 		__cpacf_query(CPACF_KMA, (cpacf_mask_t *)
430 			      kvm_s390_available_subfunc.kma);
431 
432 	if (test_facility(155)) /* MSA9 */
433 		__cpacf_query(CPACF_KDSA, (cpacf_mask_t *)
434 			      kvm_s390_available_subfunc.kdsa);
435 
436 	if (test_facility(150)) /* SORTL */
437 		__sortl_query(&kvm_s390_available_subfunc.sortl);
438 
439 	if (test_facility(151)) /* DFLTCC */
440 		__dfltcc_query(&kvm_s390_available_subfunc.dfltcc);
441 
442 	if (test_facility(201))	/* PFCR */
443 		pfcr_query(&kvm_s390_available_subfunc.pfcr);
444 
445 	if (MACHINE_HAS_ESOP)
446 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP);
447 	/*
448 	 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
449 	 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
450 	 */
451 	if (!sclp.has_sief2 || !MACHINE_HAS_ESOP || !sclp.has_64bscao ||
452 	    !test_facility(3) || !nested)
453 		return;
454 	allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2);
455 	if (sclp.has_64bscao)
456 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO);
457 	if (sclp.has_siif)
458 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF);
459 	if (sclp.has_gpere)
460 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE);
461 	if (sclp.has_gsls)
462 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS);
463 	if (sclp.has_ib)
464 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB);
465 	if (sclp.has_cei)
466 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI);
467 	if (sclp.has_ibs)
468 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS);
469 	if (sclp.has_kss)
470 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_KSS);
471 	/*
472 	 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
473 	 * all skey handling functions read/set the skey from the PGSTE
474 	 * instead of the real storage key.
475 	 *
476 	 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
477 	 * pages being detected as preserved although they are resident.
478 	 *
479 	 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
480 	 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
481 	 *
482 	 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
483 	 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
484 	 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
485 	 *
486 	 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
487 	 * cannot easily shadow the SCA because of the ipte lock.
488 	 */
489 }
490 
491 static int __init __kvm_s390_init(void)
492 {
493 	int rc = -ENOMEM;
494 
495 	kvm_s390_dbf = debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
496 	if (!kvm_s390_dbf)
497 		return -ENOMEM;
498 
499 	kvm_s390_dbf_uv = debug_register("kvm-uv", 32, 1, 7 * sizeof(long));
500 	if (!kvm_s390_dbf_uv)
501 		goto err_kvm_uv;
502 
503 	if (debug_register_view(kvm_s390_dbf, &debug_sprintf_view) ||
504 	    debug_register_view(kvm_s390_dbf_uv, &debug_sprintf_view))
505 		goto err_debug_view;
506 
507 	kvm_s390_cpu_feat_init();
508 
509 	/* Register floating interrupt controller interface. */
510 	rc = kvm_register_device_ops(&kvm_flic_ops, KVM_DEV_TYPE_FLIC);
511 	if (rc) {
512 		pr_err("A FLIC registration call failed with rc=%d\n", rc);
513 		goto err_flic;
514 	}
515 
516 	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM)) {
517 		rc = kvm_s390_pci_init();
518 		if (rc) {
519 			pr_err("Unable to allocate AIFT for PCI\n");
520 			goto err_pci;
521 		}
522 	}
523 
524 	rc = kvm_s390_gib_init(GAL_ISC);
525 	if (rc)
526 		goto err_gib;
527 
528 	gmap_notifier.notifier_call = kvm_gmap_notifier;
529 	gmap_register_pte_notifier(&gmap_notifier);
530 	vsie_gmap_notifier.notifier_call = kvm_s390_vsie_gmap_notifier;
531 	gmap_register_pte_notifier(&vsie_gmap_notifier);
532 	atomic_notifier_chain_register(&s390_epoch_delta_notifier,
533 				       &kvm_clock_notifier);
534 
535 	return 0;
536 
537 err_gib:
538 	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM))
539 		kvm_s390_pci_exit();
540 err_pci:
541 err_flic:
542 err_debug_view:
543 	debug_unregister(kvm_s390_dbf_uv);
544 err_kvm_uv:
545 	debug_unregister(kvm_s390_dbf);
546 	return rc;
547 }
548 
549 static void __kvm_s390_exit(void)
550 {
551 	gmap_unregister_pte_notifier(&gmap_notifier);
552 	gmap_unregister_pte_notifier(&vsie_gmap_notifier);
553 	atomic_notifier_chain_unregister(&s390_epoch_delta_notifier,
554 					 &kvm_clock_notifier);
555 
556 	kvm_s390_gib_destroy();
557 	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM))
558 		kvm_s390_pci_exit();
559 	debug_unregister(kvm_s390_dbf);
560 	debug_unregister(kvm_s390_dbf_uv);
561 }
562 
563 /* Section: device related */
564 long kvm_arch_dev_ioctl(struct file *filp,
565 			unsigned int ioctl, unsigned long arg)
566 {
567 	if (ioctl == KVM_S390_ENABLE_SIE)
568 		return s390_enable_sie();
569 	return -EINVAL;
570 }
571 
572 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
573 {
574 	int r;
575 
576 	switch (ext) {
577 	case KVM_CAP_S390_PSW:
578 	case KVM_CAP_S390_GMAP:
579 	case KVM_CAP_SYNC_MMU:
580 #ifdef CONFIG_KVM_S390_UCONTROL
581 	case KVM_CAP_S390_UCONTROL:
582 #endif
583 	case KVM_CAP_ASYNC_PF:
584 	case KVM_CAP_SYNC_REGS:
585 	case KVM_CAP_ONE_REG:
586 	case KVM_CAP_ENABLE_CAP:
587 	case KVM_CAP_S390_CSS_SUPPORT:
588 	case KVM_CAP_IOEVENTFD:
589 	case KVM_CAP_S390_IRQCHIP:
590 	case KVM_CAP_VM_ATTRIBUTES:
591 	case KVM_CAP_MP_STATE:
592 	case KVM_CAP_IMMEDIATE_EXIT:
593 	case KVM_CAP_S390_INJECT_IRQ:
594 	case KVM_CAP_S390_USER_SIGP:
595 	case KVM_CAP_S390_USER_STSI:
596 	case KVM_CAP_S390_SKEYS:
597 	case KVM_CAP_S390_IRQ_STATE:
598 	case KVM_CAP_S390_USER_INSTR0:
599 	case KVM_CAP_S390_CMMA_MIGRATION:
600 	case KVM_CAP_S390_AIS:
601 	case KVM_CAP_S390_AIS_MIGRATION:
602 	case KVM_CAP_S390_VCPU_RESETS:
603 	case KVM_CAP_SET_GUEST_DEBUG:
604 	case KVM_CAP_S390_DIAG318:
605 	case KVM_CAP_IRQFD_RESAMPLE:
606 		r = 1;
607 		break;
608 	case KVM_CAP_SET_GUEST_DEBUG2:
609 		r = KVM_GUESTDBG_VALID_MASK;
610 		break;
611 	case KVM_CAP_S390_HPAGE_1M:
612 		r = 0;
613 		if (hpage && !(kvm && kvm_is_ucontrol(kvm)))
614 			r = 1;
615 		break;
616 	case KVM_CAP_S390_MEM_OP:
617 		r = MEM_OP_MAX_SIZE;
618 		break;
619 	case KVM_CAP_S390_MEM_OP_EXTENSION:
620 		/*
621 		 * Flag bits indicating which extensions are supported.
622 		 * If r > 0, the base extension must also be supported/indicated,
623 		 * in order to maintain backwards compatibility.
624 		 */
625 		r = KVM_S390_MEMOP_EXTENSION_CAP_BASE |
626 		    KVM_S390_MEMOP_EXTENSION_CAP_CMPXCHG;
627 		break;
628 	case KVM_CAP_NR_VCPUS:
629 	case KVM_CAP_MAX_VCPUS:
630 	case KVM_CAP_MAX_VCPU_ID:
631 		r = KVM_S390_BSCA_CPU_SLOTS;
632 		if (!kvm_s390_use_sca_entries())
633 			r = KVM_MAX_VCPUS;
634 		else if (sclp.has_esca && sclp.has_64bscao)
635 			r = KVM_S390_ESCA_CPU_SLOTS;
636 		if (ext == KVM_CAP_NR_VCPUS)
637 			r = min_t(unsigned int, num_online_cpus(), r);
638 		break;
639 	case KVM_CAP_S390_COW:
640 		r = MACHINE_HAS_ESOP;
641 		break;
642 	case KVM_CAP_S390_VECTOR_REGISTERS:
643 		r = test_facility(129);
644 		break;
645 	case KVM_CAP_S390_RI:
646 		r = test_facility(64);
647 		break;
648 	case KVM_CAP_S390_GS:
649 		r = test_facility(133);
650 		break;
651 	case KVM_CAP_S390_BPB:
652 		r = test_facility(82);
653 		break;
654 	case KVM_CAP_S390_PROTECTED_ASYNC_DISABLE:
655 		r = async_destroy && is_prot_virt_host();
656 		break;
657 	case KVM_CAP_S390_PROTECTED:
658 		r = is_prot_virt_host();
659 		break;
660 	case KVM_CAP_S390_PROTECTED_DUMP: {
661 		u64 pv_cmds_dump[] = {
662 			BIT_UVC_CMD_DUMP_INIT,
663 			BIT_UVC_CMD_DUMP_CONFIG_STOR_STATE,
664 			BIT_UVC_CMD_DUMP_CPU,
665 			BIT_UVC_CMD_DUMP_COMPLETE,
666 		};
667 		int i;
668 
669 		r = is_prot_virt_host();
670 
671 		for (i = 0; i < ARRAY_SIZE(pv_cmds_dump); i++) {
672 			if (!test_bit_inv(pv_cmds_dump[i],
673 					  (unsigned long *)&uv_info.inst_calls_list)) {
674 				r = 0;
675 				break;
676 			}
677 		}
678 		break;
679 	}
680 	case KVM_CAP_S390_ZPCI_OP:
681 		r = kvm_s390_pci_interp_allowed();
682 		break;
683 	case KVM_CAP_S390_CPU_TOPOLOGY:
684 		r = test_facility(11);
685 		break;
686 	default:
687 		r = 0;
688 	}
689 	return r;
690 }
691 
692 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
693 {
694 	int i;
695 	gfn_t cur_gfn, last_gfn;
696 	unsigned long gaddr, vmaddr;
697 	struct gmap *gmap = kvm->arch.gmap;
698 	DECLARE_BITMAP(bitmap, _PAGE_ENTRIES);
699 
700 	/* Loop over all guest segments */
701 	cur_gfn = memslot->base_gfn;
702 	last_gfn = memslot->base_gfn + memslot->npages;
703 	for (; cur_gfn <= last_gfn; cur_gfn += _PAGE_ENTRIES) {
704 		gaddr = gfn_to_gpa(cur_gfn);
705 		vmaddr = gfn_to_hva_memslot(memslot, cur_gfn);
706 		if (kvm_is_error_hva(vmaddr))
707 			continue;
708 
709 		bitmap_zero(bitmap, _PAGE_ENTRIES);
710 		gmap_sync_dirty_log_pmd(gmap, bitmap, gaddr, vmaddr);
711 		for (i = 0; i < _PAGE_ENTRIES; i++) {
712 			if (test_bit(i, bitmap))
713 				mark_page_dirty(kvm, cur_gfn + i);
714 		}
715 
716 		if (fatal_signal_pending(current))
717 			return;
718 		cond_resched();
719 	}
720 }
721 
722 /* Section: vm related */
723 static void sca_del_vcpu(struct kvm_vcpu *vcpu);
724 
725 /*
726  * Get (and clear) the dirty memory log for a memory slot.
727  */
728 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
729 			       struct kvm_dirty_log *log)
730 {
731 	int r;
732 	unsigned long n;
733 	struct kvm_memory_slot *memslot;
734 	int is_dirty;
735 
736 	if (kvm_is_ucontrol(kvm))
737 		return -EINVAL;
738 
739 	mutex_lock(&kvm->slots_lock);
740 
741 	r = -EINVAL;
742 	if (log->slot >= KVM_USER_MEM_SLOTS)
743 		goto out;
744 
745 	r = kvm_get_dirty_log(kvm, log, &is_dirty, &memslot);
746 	if (r)
747 		goto out;
748 
749 	/* Clear the dirty log */
750 	if (is_dirty) {
751 		n = kvm_dirty_bitmap_bytes(memslot);
752 		memset(memslot->dirty_bitmap, 0, n);
753 	}
754 	r = 0;
755 out:
756 	mutex_unlock(&kvm->slots_lock);
757 	return r;
758 }
759 
760 static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
761 {
762 	unsigned long i;
763 	struct kvm_vcpu *vcpu;
764 
765 	kvm_for_each_vcpu(i, vcpu, kvm) {
766 		kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
767 	}
768 }
769 
770 int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
771 {
772 	int r;
773 
774 	if (cap->flags)
775 		return -EINVAL;
776 
777 	switch (cap->cap) {
778 	case KVM_CAP_S390_IRQCHIP:
779 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
780 		kvm->arch.use_irqchip = 1;
781 		r = 0;
782 		break;
783 	case KVM_CAP_S390_USER_SIGP:
784 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
785 		kvm->arch.user_sigp = 1;
786 		r = 0;
787 		break;
788 	case KVM_CAP_S390_VECTOR_REGISTERS:
789 		mutex_lock(&kvm->lock);
790 		if (kvm->created_vcpus) {
791 			r = -EBUSY;
792 		} else if (cpu_has_vx()) {
793 			set_kvm_facility(kvm->arch.model.fac_mask, 129);
794 			set_kvm_facility(kvm->arch.model.fac_list, 129);
795 			if (test_facility(134)) {
796 				set_kvm_facility(kvm->arch.model.fac_mask, 134);
797 				set_kvm_facility(kvm->arch.model.fac_list, 134);
798 			}
799 			if (test_facility(135)) {
800 				set_kvm_facility(kvm->arch.model.fac_mask, 135);
801 				set_kvm_facility(kvm->arch.model.fac_list, 135);
802 			}
803 			if (test_facility(148)) {
804 				set_kvm_facility(kvm->arch.model.fac_mask, 148);
805 				set_kvm_facility(kvm->arch.model.fac_list, 148);
806 			}
807 			if (test_facility(152)) {
808 				set_kvm_facility(kvm->arch.model.fac_mask, 152);
809 				set_kvm_facility(kvm->arch.model.fac_list, 152);
810 			}
811 			if (test_facility(192)) {
812 				set_kvm_facility(kvm->arch.model.fac_mask, 192);
813 				set_kvm_facility(kvm->arch.model.fac_list, 192);
814 			}
815 			if (test_facility(198)) {
816 				set_kvm_facility(kvm->arch.model.fac_mask, 198);
817 				set_kvm_facility(kvm->arch.model.fac_list, 198);
818 			}
819 			if (test_facility(199)) {
820 				set_kvm_facility(kvm->arch.model.fac_mask, 199);
821 				set_kvm_facility(kvm->arch.model.fac_list, 199);
822 			}
823 			r = 0;
824 		} else
825 			r = -EINVAL;
826 		mutex_unlock(&kvm->lock);
827 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
828 			 r ? "(not available)" : "(success)");
829 		break;
830 	case KVM_CAP_S390_RI:
831 		r = -EINVAL;
832 		mutex_lock(&kvm->lock);
833 		if (kvm->created_vcpus) {
834 			r = -EBUSY;
835 		} else if (test_facility(64)) {
836 			set_kvm_facility(kvm->arch.model.fac_mask, 64);
837 			set_kvm_facility(kvm->arch.model.fac_list, 64);
838 			r = 0;
839 		}
840 		mutex_unlock(&kvm->lock);
841 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_RI %s",
842 			 r ? "(not available)" : "(success)");
843 		break;
844 	case KVM_CAP_S390_AIS:
845 		mutex_lock(&kvm->lock);
846 		if (kvm->created_vcpus) {
847 			r = -EBUSY;
848 		} else {
849 			set_kvm_facility(kvm->arch.model.fac_mask, 72);
850 			set_kvm_facility(kvm->arch.model.fac_list, 72);
851 			r = 0;
852 		}
853 		mutex_unlock(&kvm->lock);
854 		VM_EVENT(kvm, 3, "ENABLE: AIS %s",
855 			 r ? "(not available)" : "(success)");
856 		break;
857 	case KVM_CAP_S390_GS:
858 		r = -EINVAL;
859 		mutex_lock(&kvm->lock);
860 		if (kvm->created_vcpus) {
861 			r = -EBUSY;
862 		} else if (test_facility(133)) {
863 			set_kvm_facility(kvm->arch.model.fac_mask, 133);
864 			set_kvm_facility(kvm->arch.model.fac_list, 133);
865 			r = 0;
866 		}
867 		mutex_unlock(&kvm->lock);
868 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_GS %s",
869 			 r ? "(not available)" : "(success)");
870 		break;
871 	case KVM_CAP_S390_HPAGE_1M:
872 		mutex_lock(&kvm->lock);
873 		if (kvm->created_vcpus)
874 			r = -EBUSY;
875 		else if (!hpage || kvm->arch.use_cmma || kvm_is_ucontrol(kvm))
876 			r = -EINVAL;
877 		else {
878 			r = 0;
879 			mmap_write_lock(kvm->mm);
880 			kvm->mm->context.allow_gmap_hpage_1m = 1;
881 			mmap_write_unlock(kvm->mm);
882 			/*
883 			 * We might have to create fake 4k page
884 			 * tables. To avoid that the hardware works on
885 			 * stale PGSTEs, we emulate these instructions.
886 			 */
887 			kvm->arch.use_skf = 0;
888 			kvm->arch.use_pfmfi = 0;
889 		}
890 		mutex_unlock(&kvm->lock);
891 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_HPAGE %s",
892 			 r ? "(not available)" : "(success)");
893 		break;
894 	case KVM_CAP_S390_USER_STSI:
895 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
896 		kvm->arch.user_stsi = 1;
897 		r = 0;
898 		break;
899 	case KVM_CAP_S390_USER_INSTR0:
900 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
901 		kvm->arch.user_instr0 = 1;
902 		icpt_operexc_on_all_vcpus(kvm);
903 		r = 0;
904 		break;
905 	case KVM_CAP_S390_CPU_TOPOLOGY:
906 		r = -EINVAL;
907 		mutex_lock(&kvm->lock);
908 		if (kvm->created_vcpus) {
909 			r = -EBUSY;
910 		} else if (test_facility(11)) {
911 			set_kvm_facility(kvm->arch.model.fac_mask, 11);
912 			set_kvm_facility(kvm->arch.model.fac_list, 11);
913 			r = 0;
914 		}
915 		mutex_unlock(&kvm->lock);
916 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_CPU_TOPOLOGY %s",
917 			 r ? "(not available)" : "(success)");
918 		break;
919 	default:
920 		r = -EINVAL;
921 		break;
922 	}
923 	return r;
924 }
925 
926 static int kvm_s390_get_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
927 {
928 	int ret;
929 
930 	switch (attr->attr) {
931 	case KVM_S390_VM_MEM_LIMIT_SIZE:
932 		ret = 0;
933 		VM_EVENT(kvm, 3, "QUERY: max guest memory: %lu bytes",
934 			 kvm->arch.mem_limit);
935 		if (put_user(kvm->arch.mem_limit, (u64 __user *)attr->addr))
936 			ret = -EFAULT;
937 		break;
938 	default:
939 		ret = -ENXIO;
940 		break;
941 	}
942 	return ret;
943 }
944 
945 static int kvm_s390_set_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
946 {
947 	int ret;
948 	unsigned int idx;
949 	switch (attr->attr) {
950 	case KVM_S390_VM_MEM_ENABLE_CMMA:
951 		ret = -ENXIO;
952 		if (!sclp.has_cmma)
953 			break;
954 
955 		VM_EVENT(kvm, 3, "%s", "ENABLE: CMMA support");
956 		mutex_lock(&kvm->lock);
957 		if (kvm->created_vcpus)
958 			ret = -EBUSY;
959 		else if (kvm->mm->context.allow_gmap_hpage_1m)
960 			ret = -EINVAL;
961 		else {
962 			kvm->arch.use_cmma = 1;
963 			/* Not compatible with cmma. */
964 			kvm->arch.use_pfmfi = 0;
965 			ret = 0;
966 		}
967 		mutex_unlock(&kvm->lock);
968 		break;
969 	case KVM_S390_VM_MEM_CLR_CMMA:
970 		ret = -ENXIO;
971 		if (!sclp.has_cmma)
972 			break;
973 		ret = -EINVAL;
974 		if (!kvm->arch.use_cmma)
975 			break;
976 
977 		VM_EVENT(kvm, 3, "%s", "RESET: CMMA states");
978 		mutex_lock(&kvm->lock);
979 		idx = srcu_read_lock(&kvm->srcu);
980 		s390_reset_cmma(kvm->arch.gmap->mm);
981 		srcu_read_unlock(&kvm->srcu, idx);
982 		mutex_unlock(&kvm->lock);
983 		ret = 0;
984 		break;
985 	case KVM_S390_VM_MEM_LIMIT_SIZE: {
986 		unsigned long new_limit;
987 
988 		if (kvm_is_ucontrol(kvm))
989 			return -EINVAL;
990 
991 		if (get_user(new_limit, (u64 __user *)attr->addr))
992 			return -EFAULT;
993 
994 		if (kvm->arch.mem_limit != KVM_S390_NO_MEM_LIMIT &&
995 		    new_limit > kvm->arch.mem_limit)
996 			return -E2BIG;
997 
998 		if (!new_limit)
999 			return -EINVAL;
1000 
1001 		/* gmap_create takes last usable address */
1002 		if (new_limit != KVM_S390_NO_MEM_LIMIT)
1003 			new_limit -= 1;
1004 
1005 		ret = -EBUSY;
1006 		mutex_lock(&kvm->lock);
1007 		if (!kvm->created_vcpus) {
1008 			/* gmap_create will round the limit up */
1009 			struct gmap *new = gmap_create(current->mm, new_limit);
1010 
1011 			if (!new) {
1012 				ret = -ENOMEM;
1013 			} else {
1014 				gmap_remove(kvm->arch.gmap);
1015 				new->private = kvm;
1016 				kvm->arch.gmap = new;
1017 				ret = 0;
1018 			}
1019 		}
1020 		mutex_unlock(&kvm->lock);
1021 		VM_EVENT(kvm, 3, "SET: max guest address: %lu", new_limit);
1022 		VM_EVENT(kvm, 3, "New guest asce: 0x%pK",
1023 			 (void *) kvm->arch.gmap->asce);
1024 		break;
1025 	}
1026 	default:
1027 		ret = -ENXIO;
1028 		break;
1029 	}
1030 	return ret;
1031 }
1032 
1033 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
1034 
1035 void kvm_s390_vcpu_crypto_reset_all(struct kvm *kvm)
1036 {
1037 	struct kvm_vcpu *vcpu;
1038 	unsigned long i;
1039 
1040 	kvm_s390_vcpu_block_all(kvm);
1041 
1042 	kvm_for_each_vcpu(i, vcpu, kvm) {
1043 		kvm_s390_vcpu_crypto_setup(vcpu);
1044 		/* recreate the shadow crycb by leaving the VSIE handler */
1045 		kvm_s390_sync_request(KVM_REQ_VSIE_RESTART, vcpu);
1046 	}
1047 
1048 	kvm_s390_vcpu_unblock_all(kvm);
1049 }
1050 
1051 static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
1052 {
1053 	mutex_lock(&kvm->lock);
1054 	switch (attr->attr) {
1055 	case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
1056 		if (!test_kvm_facility(kvm, 76)) {
1057 			mutex_unlock(&kvm->lock);
1058 			return -EINVAL;
1059 		}
1060 		get_random_bytes(
1061 			kvm->arch.crypto.crycb->aes_wrapping_key_mask,
1062 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
1063 		kvm->arch.crypto.aes_kw = 1;
1064 		VM_EVENT(kvm, 3, "%s", "ENABLE: AES keywrapping support");
1065 		break;
1066 	case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
1067 		if (!test_kvm_facility(kvm, 76)) {
1068 			mutex_unlock(&kvm->lock);
1069 			return -EINVAL;
1070 		}
1071 		get_random_bytes(
1072 			kvm->arch.crypto.crycb->dea_wrapping_key_mask,
1073 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
1074 		kvm->arch.crypto.dea_kw = 1;
1075 		VM_EVENT(kvm, 3, "%s", "ENABLE: DEA keywrapping support");
1076 		break;
1077 	case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
1078 		if (!test_kvm_facility(kvm, 76)) {
1079 			mutex_unlock(&kvm->lock);
1080 			return -EINVAL;
1081 		}
1082 		kvm->arch.crypto.aes_kw = 0;
1083 		memset(kvm->arch.crypto.crycb->aes_wrapping_key_mask, 0,
1084 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
1085 		VM_EVENT(kvm, 3, "%s", "DISABLE: AES keywrapping support");
1086 		break;
1087 	case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
1088 		if (!test_kvm_facility(kvm, 76)) {
1089 			mutex_unlock(&kvm->lock);
1090 			return -EINVAL;
1091 		}
1092 		kvm->arch.crypto.dea_kw = 0;
1093 		memset(kvm->arch.crypto.crycb->dea_wrapping_key_mask, 0,
1094 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
1095 		VM_EVENT(kvm, 3, "%s", "DISABLE: DEA keywrapping support");
1096 		break;
1097 	case KVM_S390_VM_CRYPTO_ENABLE_APIE:
1098 		if (!ap_instructions_available()) {
1099 			mutex_unlock(&kvm->lock);
1100 			return -EOPNOTSUPP;
1101 		}
1102 		kvm->arch.crypto.apie = 1;
1103 		break;
1104 	case KVM_S390_VM_CRYPTO_DISABLE_APIE:
1105 		if (!ap_instructions_available()) {
1106 			mutex_unlock(&kvm->lock);
1107 			return -EOPNOTSUPP;
1108 		}
1109 		kvm->arch.crypto.apie = 0;
1110 		break;
1111 	default:
1112 		mutex_unlock(&kvm->lock);
1113 		return -ENXIO;
1114 	}
1115 
1116 	kvm_s390_vcpu_crypto_reset_all(kvm);
1117 	mutex_unlock(&kvm->lock);
1118 	return 0;
1119 }
1120 
1121 static void kvm_s390_vcpu_pci_setup(struct kvm_vcpu *vcpu)
1122 {
1123 	/* Only set the ECB bits after guest requests zPCI interpretation */
1124 	if (!vcpu->kvm->arch.use_zpci_interp)
1125 		return;
1126 
1127 	vcpu->arch.sie_block->ecb2 |= ECB2_ZPCI_LSI;
1128 	vcpu->arch.sie_block->ecb3 |= ECB3_AISII + ECB3_AISI;
1129 }
1130 
1131 void kvm_s390_vcpu_pci_enable_interp(struct kvm *kvm)
1132 {
1133 	struct kvm_vcpu *vcpu;
1134 	unsigned long i;
1135 
1136 	lockdep_assert_held(&kvm->lock);
1137 
1138 	if (!kvm_s390_pci_interp_allowed())
1139 		return;
1140 
1141 	/*
1142 	 * If host is configured for PCI and the necessary facilities are
1143 	 * available, turn on interpretation for the life of this guest
1144 	 */
1145 	kvm->arch.use_zpci_interp = 1;
1146 
1147 	kvm_s390_vcpu_block_all(kvm);
1148 
1149 	kvm_for_each_vcpu(i, vcpu, kvm) {
1150 		kvm_s390_vcpu_pci_setup(vcpu);
1151 		kvm_s390_sync_request(KVM_REQ_VSIE_RESTART, vcpu);
1152 	}
1153 
1154 	kvm_s390_vcpu_unblock_all(kvm);
1155 }
1156 
1157 static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
1158 {
1159 	unsigned long cx;
1160 	struct kvm_vcpu *vcpu;
1161 
1162 	kvm_for_each_vcpu(cx, vcpu, kvm)
1163 		kvm_s390_sync_request(req, vcpu);
1164 }
1165 
1166 /*
1167  * Must be called with kvm->srcu held to avoid races on memslots, and with
1168  * kvm->slots_lock to avoid races with ourselves and kvm_s390_vm_stop_migration.
1169  */
1170 static int kvm_s390_vm_start_migration(struct kvm *kvm)
1171 {
1172 	struct kvm_memory_slot *ms;
1173 	struct kvm_memslots *slots;
1174 	unsigned long ram_pages = 0;
1175 	int bkt;
1176 
1177 	/* migration mode already enabled */
1178 	if (kvm->arch.migration_mode)
1179 		return 0;
1180 	slots = kvm_memslots(kvm);
1181 	if (!slots || kvm_memslots_empty(slots))
1182 		return -EINVAL;
1183 
1184 	if (!kvm->arch.use_cmma) {
1185 		kvm->arch.migration_mode = 1;
1186 		return 0;
1187 	}
1188 	/* mark all the pages in active slots as dirty */
1189 	kvm_for_each_memslot(ms, bkt, slots) {
1190 		if (!ms->dirty_bitmap)
1191 			return -EINVAL;
1192 		/*
1193 		 * The second half of the bitmap is only used on x86,
1194 		 * and would be wasted otherwise, so we put it to good
1195 		 * use here to keep track of the state of the storage
1196 		 * attributes.
1197 		 */
1198 		memset(kvm_second_dirty_bitmap(ms), 0xff, kvm_dirty_bitmap_bytes(ms));
1199 		ram_pages += ms->npages;
1200 	}
1201 	atomic64_set(&kvm->arch.cmma_dirty_pages, ram_pages);
1202 	kvm->arch.migration_mode = 1;
1203 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_START_MIGRATION);
1204 	return 0;
1205 }
1206 
1207 /*
1208  * Must be called with kvm->slots_lock to avoid races with ourselves and
1209  * kvm_s390_vm_start_migration.
1210  */
1211 static int kvm_s390_vm_stop_migration(struct kvm *kvm)
1212 {
1213 	/* migration mode already disabled */
1214 	if (!kvm->arch.migration_mode)
1215 		return 0;
1216 	kvm->arch.migration_mode = 0;
1217 	if (kvm->arch.use_cmma)
1218 		kvm_s390_sync_request_broadcast(kvm, KVM_REQ_STOP_MIGRATION);
1219 	return 0;
1220 }
1221 
1222 static int kvm_s390_vm_set_migration(struct kvm *kvm,
1223 				     struct kvm_device_attr *attr)
1224 {
1225 	int res = -ENXIO;
1226 
1227 	mutex_lock(&kvm->slots_lock);
1228 	switch (attr->attr) {
1229 	case KVM_S390_VM_MIGRATION_START:
1230 		res = kvm_s390_vm_start_migration(kvm);
1231 		break;
1232 	case KVM_S390_VM_MIGRATION_STOP:
1233 		res = kvm_s390_vm_stop_migration(kvm);
1234 		break;
1235 	default:
1236 		break;
1237 	}
1238 	mutex_unlock(&kvm->slots_lock);
1239 
1240 	return res;
1241 }
1242 
1243 static int kvm_s390_vm_get_migration(struct kvm *kvm,
1244 				     struct kvm_device_attr *attr)
1245 {
1246 	u64 mig = kvm->arch.migration_mode;
1247 
1248 	if (attr->attr != KVM_S390_VM_MIGRATION_STATUS)
1249 		return -ENXIO;
1250 
1251 	if (copy_to_user((void __user *)attr->addr, &mig, sizeof(mig)))
1252 		return -EFAULT;
1253 	return 0;
1254 }
1255 
1256 static void __kvm_s390_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod);
1257 
1258 static int kvm_s390_set_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1259 {
1260 	struct kvm_s390_vm_tod_clock gtod;
1261 
1262 	if (copy_from_user(&gtod, (void __user *)attr->addr, sizeof(gtod)))
1263 		return -EFAULT;
1264 
1265 	if (!test_kvm_facility(kvm, 139) && gtod.epoch_idx)
1266 		return -EINVAL;
1267 	__kvm_s390_set_tod_clock(kvm, &gtod);
1268 
1269 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x, TOD base: 0x%llx",
1270 		gtod.epoch_idx, gtod.tod);
1271 
1272 	return 0;
1273 }
1274 
1275 static int kvm_s390_set_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1276 {
1277 	u8 gtod_high;
1278 
1279 	if (copy_from_user(&gtod_high, (void __user *)attr->addr,
1280 					   sizeof(gtod_high)))
1281 		return -EFAULT;
1282 
1283 	if (gtod_high != 0)
1284 		return -EINVAL;
1285 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x", gtod_high);
1286 
1287 	return 0;
1288 }
1289 
1290 static int kvm_s390_set_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1291 {
1292 	struct kvm_s390_vm_tod_clock gtod = { 0 };
1293 
1294 	if (copy_from_user(&gtod.tod, (void __user *)attr->addr,
1295 			   sizeof(gtod.tod)))
1296 		return -EFAULT;
1297 
1298 	__kvm_s390_set_tod_clock(kvm, &gtod);
1299 	VM_EVENT(kvm, 3, "SET: TOD base: 0x%llx", gtod.tod);
1300 	return 0;
1301 }
1302 
1303 static int kvm_s390_set_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1304 {
1305 	int ret;
1306 
1307 	if (attr->flags)
1308 		return -EINVAL;
1309 
1310 	mutex_lock(&kvm->lock);
1311 	/*
1312 	 * For protected guests, the TOD is managed by the ultravisor, so trying
1313 	 * to change it will never bring the expected results.
1314 	 */
1315 	if (kvm_s390_pv_is_protected(kvm)) {
1316 		ret = -EOPNOTSUPP;
1317 		goto out_unlock;
1318 	}
1319 
1320 	switch (attr->attr) {
1321 	case KVM_S390_VM_TOD_EXT:
1322 		ret = kvm_s390_set_tod_ext(kvm, attr);
1323 		break;
1324 	case KVM_S390_VM_TOD_HIGH:
1325 		ret = kvm_s390_set_tod_high(kvm, attr);
1326 		break;
1327 	case KVM_S390_VM_TOD_LOW:
1328 		ret = kvm_s390_set_tod_low(kvm, attr);
1329 		break;
1330 	default:
1331 		ret = -ENXIO;
1332 		break;
1333 	}
1334 
1335 out_unlock:
1336 	mutex_unlock(&kvm->lock);
1337 	return ret;
1338 }
1339 
1340 static void kvm_s390_get_tod_clock(struct kvm *kvm,
1341 				   struct kvm_s390_vm_tod_clock *gtod)
1342 {
1343 	union tod_clock clk;
1344 
1345 	preempt_disable();
1346 
1347 	store_tod_clock_ext(&clk);
1348 
1349 	gtod->tod = clk.tod + kvm->arch.epoch;
1350 	gtod->epoch_idx = 0;
1351 	if (test_kvm_facility(kvm, 139)) {
1352 		gtod->epoch_idx = clk.ei + kvm->arch.epdx;
1353 		if (gtod->tod < clk.tod)
1354 			gtod->epoch_idx += 1;
1355 	}
1356 
1357 	preempt_enable();
1358 }
1359 
1360 static int kvm_s390_get_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1361 {
1362 	struct kvm_s390_vm_tod_clock gtod;
1363 
1364 	memset(&gtod, 0, sizeof(gtod));
1365 	kvm_s390_get_tod_clock(kvm, &gtod);
1366 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1367 		return -EFAULT;
1368 
1369 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x, TOD base: 0x%llx",
1370 		gtod.epoch_idx, gtod.tod);
1371 	return 0;
1372 }
1373 
1374 static int kvm_s390_get_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1375 {
1376 	u8 gtod_high = 0;
1377 
1378 	if (copy_to_user((void __user *)attr->addr, &gtod_high,
1379 					 sizeof(gtod_high)))
1380 		return -EFAULT;
1381 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x", gtod_high);
1382 
1383 	return 0;
1384 }
1385 
1386 static int kvm_s390_get_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1387 {
1388 	u64 gtod;
1389 
1390 	gtod = kvm_s390_get_tod_clock_fast(kvm);
1391 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1392 		return -EFAULT;
1393 	VM_EVENT(kvm, 3, "QUERY: TOD base: 0x%llx", gtod);
1394 
1395 	return 0;
1396 }
1397 
1398 static int kvm_s390_get_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1399 {
1400 	int ret;
1401 
1402 	if (attr->flags)
1403 		return -EINVAL;
1404 
1405 	switch (attr->attr) {
1406 	case KVM_S390_VM_TOD_EXT:
1407 		ret = kvm_s390_get_tod_ext(kvm, attr);
1408 		break;
1409 	case KVM_S390_VM_TOD_HIGH:
1410 		ret = kvm_s390_get_tod_high(kvm, attr);
1411 		break;
1412 	case KVM_S390_VM_TOD_LOW:
1413 		ret = kvm_s390_get_tod_low(kvm, attr);
1414 		break;
1415 	default:
1416 		ret = -ENXIO;
1417 		break;
1418 	}
1419 	return ret;
1420 }
1421 
1422 static int kvm_s390_set_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1423 {
1424 	struct kvm_s390_vm_cpu_processor *proc;
1425 	u16 lowest_ibc, unblocked_ibc;
1426 	int ret = 0;
1427 
1428 	mutex_lock(&kvm->lock);
1429 	if (kvm->created_vcpus) {
1430 		ret = -EBUSY;
1431 		goto out;
1432 	}
1433 	proc = kzalloc(sizeof(*proc), GFP_KERNEL_ACCOUNT);
1434 	if (!proc) {
1435 		ret = -ENOMEM;
1436 		goto out;
1437 	}
1438 	if (!copy_from_user(proc, (void __user *)attr->addr,
1439 			    sizeof(*proc))) {
1440 		kvm->arch.model.cpuid = proc->cpuid;
1441 		lowest_ibc = sclp.ibc >> 16 & 0xfff;
1442 		unblocked_ibc = sclp.ibc & 0xfff;
1443 		if (lowest_ibc && proc->ibc) {
1444 			if (proc->ibc > unblocked_ibc)
1445 				kvm->arch.model.ibc = unblocked_ibc;
1446 			else if (proc->ibc < lowest_ibc)
1447 				kvm->arch.model.ibc = lowest_ibc;
1448 			else
1449 				kvm->arch.model.ibc = proc->ibc;
1450 		}
1451 		memcpy(kvm->arch.model.fac_list, proc->fac_list,
1452 		       S390_ARCH_FAC_LIST_SIZE_BYTE);
1453 		VM_EVENT(kvm, 3, "SET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1454 			 kvm->arch.model.ibc,
1455 			 kvm->arch.model.cpuid);
1456 		VM_EVENT(kvm, 3, "SET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1457 			 kvm->arch.model.fac_list[0],
1458 			 kvm->arch.model.fac_list[1],
1459 			 kvm->arch.model.fac_list[2]);
1460 	} else
1461 		ret = -EFAULT;
1462 	kfree(proc);
1463 out:
1464 	mutex_unlock(&kvm->lock);
1465 	return ret;
1466 }
1467 
1468 static int kvm_s390_set_processor_feat(struct kvm *kvm,
1469 				       struct kvm_device_attr *attr)
1470 {
1471 	struct kvm_s390_vm_cpu_feat data;
1472 
1473 	if (copy_from_user(&data, (void __user *)attr->addr, sizeof(data)))
1474 		return -EFAULT;
1475 	if (!bitmap_subset((unsigned long *) data.feat,
1476 			   kvm_s390_available_cpu_feat,
1477 			   KVM_S390_VM_CPU_FEAT_NR_BITS))
1478 		return -EINVAL;
1479 
1480 	mutex_lock(&kvm->lock);
1481 	if (kvm->created_vcpus) {
1482 		mutex_unlock(&kvm->lock);
1483 		return -EBUSY;
1484 	}
1485 	bitmap_from_arr64(kvm->arch.cpu_feat, data.feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
1486 	mutex_unlock(&kvm->lock);
1487 	VM_EVENT(kvm, 3, "SET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1488 			 data.feat[0],
1489 			 data.feat[1],
1490 			 data.feat[2]);
1491 	return 0;
1492 }
1493 
1494 static int kvm_s390_set_processor_subfunc(struct kvm *kvm,
1495 					  struct kvm_device_attr *attr)
1496 {
1497 	mutex_lock(&kvm->lock);
1498 	if (kvm->created_vcpus) {
1499 		mutex_unlock(&kvm->lock);
1500 		return -EBUSY;
1501 	}
1502 
1503 	if (copy_from_user(&kvm->arch.model.subfuncs, (void __user *)attr->addr,
1504 			   sizeof(struct kvm_s390_vm_cpu_subfunc))) {
1505 		mutex_unlock(&kvm->lock);
1506 		return -EFAULT;
1507 	}
1508 	mutex_unlock(&kvm->lock);
1509 
1510 	VM_EVENT(kvm, 3, "SET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1511 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1512 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1513 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1514 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1515 	VM_EVENT(kvm, 3, "SET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1516 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1517 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1518 	VM_EVENT(kvm, 3, "SET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1519 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1520 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1521 	VM_EVENT(kvm, 3, "SET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1522 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1523 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1524 	VM_EVENT(kvm, 3, "SET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1525 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1526 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1527 	VM_EVENT(kvm, 3, "SET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1528 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1529 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1530 	VM_EVENT(kvm, 3, "SET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1531 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1532 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1533 	VM_EVENT(kvm, 3, "SET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1534 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1535 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1536 	VM_EVENT(kvm, 3, "SET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1537 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1538 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1539 	VM_EVENT(kvm, 3, "SET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1540 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1541 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1542 	VM_EVENT(kvm, 3, "SET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1543 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1544 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1545 	VM_EVENT(kvm, 3, "SET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1546 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1547 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1548 	VM_EVENT(kvm, 3, "SET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1549 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1550 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1551 	VM_EVENT(kvm, 3, "SET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1552 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1553 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1554 	VM_EVENT(kvm, 3, "SET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1555 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1556 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1557 	VM_EVENT(kvm, 3, "SET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1558 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1559 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1560 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1561 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1562 	VM_EVENT(kvm, 3, "SET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1563 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1564 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1565 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1566 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1567 	VM_EVENT(kvm, 3, "GET: guest PFCR   subfunc 0x%16.16lx.%16.16lx",
1568 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[0],
1569 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[1]);
1570 
1571 	return 0;
1572 }
1573 
1574 #define KVM_S390_VM_CPU_UV_FEAT_GUEST_MASK	\
1575 (						\
1576 	((struct kvm_s390_vm_cpu_uv_feat){	\
1577 		.ap = 1,			\
1578 		.ap_intr = 1,			\
1579 	})					\
1580 	.feat					\
1581 )
1582 
1583 static int kvm_s390_set_uv_feat(struct kvm *kvm, struct kvm_device_attr *attr)
1584 {
1585 	struct kvm_s390_vm_cpu_uv_feat __user *ptr = (void __user *)attr->addr;
1586 	unsigned long data, filter;
1587 
1588 	filter = uv_info.uv_feature_indications & KVM_S390_VM_CPU_UV_FEAT_GUEST_MASK;
1589 	if (get_user(data, &ptr->feat))
1590 		return -EFAULT;
1591 	if (!bitmap_subset(&data, &filter, KVM_S390_VM_CPU_UV_FEAT_NR_BITS))
1592 		return -EINVAL;
1593 
1594 	mutex_lock(&kvm->lock);
1595 	if (kvm->created_vcpus) {
1596 		mutex_unlock(&kvm->lock);
1597 		return -EBUSY;
1598 	}
1599 	kvm->arch.model.uv_feat_guest.feat = data;
1600 	mutex_unlock(&kvm->lock);
1601 
1602 	VM_EVENT(kvm, 3, "SET: guest UV-feat: 0x%16.16lx", data);
1603 
1604 	return 0;
1605 }
1606 
1607 static int kvm_s390_set_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1608 {
1609 	int ret = -ENXIO;
1610 
1611 	switch (attr->attr) {
1612 	case KVM_S390_VM_CPU_PROCESSOR:
1613 		ret = kvm_s390_set_processor(kvm, attr);
1614 		break;
1615 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1616 		ret = kvm_s390_set_processor_feat(kvm, attr);
1617 		break;
1618 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1619 		ret = kvm_s390_set_processor_subfunc(kvm, attr);
1620 		break;
1621 	case KVM_S390_VM_CPU_PROCESSOR_UV_FEAT_GUEST:
1622 		ret = kvm_s390_set_uv_feat(kvm, attr);
1623 		break;
1624 	}
1625 	return ret;
1626 }
1627 
1628 static int kvm_s390_get_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1629 {
1630 	struct kvm_s390_vm_cpu_processor *proc;
1631 	int ret = 0;
1632 
1633 	proc = kzalloc(sizeof(*proc), GFP_KERNEL_ACCOUNT);
1634 	if (!proc) {
1635 		ret = -ENOMEM;
1636 		goto out;
1637 	}
1638 	proc->cpuid = kvm->arch.model.cpuid;
1639 	proc->ibc = kvm->arch.model.ibc;
1640 	memcpy(&proc->fac_list, kvm->arch.model.fac_list,
1641 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1642 	VM_EVENT(kvm, 3, "GET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1643 		 kvm->arch.model.ibc,
1644 		 kvm->arch.model.cpuid);
1645 	VM_EVENT(kvm, 3, "GET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1646 		 kvm->arch.model.fac_list[0],
1647 		 kvm->arch.model.fac_list[1],
1648 		 kvm->arch.model.fac_list[2]);
1649 	if (copy_to_user((void __user *)attr->addr, proc, sizeof(*proc)))
1650 		ret = -EFAULT;
1651 	kfree(proc);
1652 out:
1653 	return ret;
1654 }
1655 
1656 static int kvm_s390_get_machine(struct kvm *kvm, struct kvm_device_attr *attr)
1657 {
1658 	struct kvm_s390_vm_cpu_machine *mach;
1659 	int ret = 0;
1660 
1661 	mach = kzalloc(sizeof(*mach), GFP_KERNEL_ACCOUNT);
1662 	if (!mach) {
1663 		ret = -ENOMEM;
1664 		goto out;
1665 	}
1666 	get_cpu_id((struct cpuid *) &mach->cpuid);
1667 	mach->ibc = sclp.ibc;
1668 	memcpy(&mach->fac_mask, kvm->arch.model.fac_mask,
1669 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1670 	memcpy((unsigned long *)&mach->fac_list, stfle_fac_list,
1671 	       sizeof(stfle_fac_list));
1672 	VM_EVENT(kvm, 3, "GET: host ibc:  0x%4.4x, host cpuid:  0x%16.16llx",
1673 		 kvm->arch.model.ibc,
1674 		 kvm->arch.model.cpuid);
1675 	VM_EVENT(kvm, 3, "GET: host facmask:  0x%16.16llx.%16.16llx.%16.16llx",
1676 		 mach->fac_mask[0],
1677 		 mach->fac_mask[1],
1678 		 mach->fac_mask[2]);
1679 	VM_EVENT(kvm, 3, "GET: host faclist:  0x%16.16llx.%16.16llx.%16.16llx",
1680 		 mach->fac_list[0],
1681 		 mach->fac_list[1],
1682 		 mach->fac_list[2]);
1683 	if (copy_to_user((void __user *)attr->addr, mach, sizeof(*mach)))
1684 		ret = -EFAULT;
1685 	kfree(mach);
1686 out:
1687 	return ret;
1688 }
1689 
1690 static int kvm_s390_get_processor_feat(struct kvm *kvm,
1691 				       struct kvm_device_attr *attr)
1692 {
1693 	struct kvm_s390_vm_cpu_feat data;
1694 
1695 	bitmap_to_arr64(data.feat, kvm->arch.cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
1696 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1697 		return -EFAULT;
1698 	VM_EVENT(kvm, 3, "GET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1699 			 data.feat[0],
1700 			 data.feat[1],
1701 			 data.feat[2]);
1702 	return 0;
1703 }
1704 
1705 static int kvm_s390_get_machine_feat(struct kvm *kvm,
1706 				     struct kvm_device_attr *attr)
1707 {
1708 	struct kvm_s390_vm_cpu_feat data;
1709 
1710 	bitmap_to_arr64(data.feat, kvm_s390_available_cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
1711 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1712 		return -EFAULT;
1713 	VM_EVENT(kvm, 3, "GET: host feat:  0x%16.16llx.0x%16.16llx.0x%16.16llx",
1714 			 data.feat[0],
1715 			 data.feat[1],
1716 			 data.feat[2]);
1717 	return 0;
1718 }
1719 
1720 static int kvm_s390_get_processor_subfunc(struct kvm *kvm,
1721 					  struct kvm_device_attr *attr)
1722 {
1723 	if (copy_to_user((void __user *)attr->addr, &kvm->arch.model.subfuncs,
1724 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1725 		return -EFAULT;
1726 
1727 	VM_EVENT(kvm, 3, "GET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1728 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1729 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1730 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1731 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1732 	VM_EVENT(kvm, 3, "GET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1733 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1734 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1735 	VM_EVENT(kvm, 3, "GET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1736 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1737 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1738 	VM_EVENT(kvm, 3, "GET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1739 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1740 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1741 	VM_EVENT(kvm, 3, "GET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1742 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1743 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1744 	VM_EVENT(kvm, 3, "GET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1745 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1746 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1747 	VM_EVENT(kvm, 3, "GET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1748 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1749 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1750 	VM_EVENT(kvm, 3, "GET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1751 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1752 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1753 	VM_EVENT(kvm, 3, "GET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1754 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1755 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1756 	VM_EVENT(kvm, 3, "GET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1757 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1758 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1759 	VM_EVENT(kvm, 3, "GET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1760 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1761 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1762 	VM_EVENT(kvm, 3, "GET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1763 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1764 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1765 	VM_EVENT(kvm, 3, "GET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1766 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1767 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1768 	VM_EVENT(kvm, 3, "GET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1769 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1770 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1771 	VM_EVENT(kvm, 3, "GET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1772 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1773 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1774 	VM_EVENT(kvm, 3, "GET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1775 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1776 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1777 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1778 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1779 	VM_EVENT(kvm, 3, "GET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1780 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1781 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1782 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1783 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1784 	VM_EVENT(kvm, 3, "GET: guest PFCR   subfunc 0x%16.16lx.%16.16lx",
1785 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[0],
1786 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[1]);
1787 
1788 	return 0;
1789 }
1790 
1791 static int kvm_s390_get_machine_subfunc(struct kvm *kvm,
1792 					struct kvm_device_attr *attr)
1793 {
1794 	if (copy_to_user((void __user *)attr->addr, &kvm_s390_available_subfunc,
1795 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1796 		return -EFAULT;
1797 
1798 	VM_EVENT(kvm, 3, "GET: host  PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1799 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[0],
1800 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[1],
1801 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[2],
1802 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[3]);
1803 	VM_EVENT(kvm, 3, "GET: host  PTFF   subfunc 0x%16.16lx.%16.16lx",
1804 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[0],
1805 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[1]);
1806 	VM_EVENT(kvm, 3, "GET: host  KMAC   subfunc 0x%16.16lx.%16.16lx",
1807 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[0],
1808 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[1]);
1809 	VM_EVENT(kvm, 3, "GET: host  KMC    subfunc 0x%16.16lx.%16.16lx",
1810 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[0],
1811 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[1]);
1812 	VM_EVENT(kvm, 3, "GET: host  KM     subfunc 0x%16.16lx.%16.16lx",
1813 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[0],
1814 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[1]);
1815 	VM_EVENT(kvm, 3, "GET: host  KIMD   subfunc 0x%16.16lx.%16.16lx",
1816 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[0],
1817 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[1]);
1818 	VM_EVENT(kvm, 3, "GET: host  KLMD   subfunc 0x%16.16lx.%16.16lx",
1819 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[0],
1820 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[1]);
1821 	VM_EVENT(kvm, 3, "GET: host  PCKMO  subfunc 0x%16.16lx.%16.16lx",
1822 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[0],
1823 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[1]);
1824 	VM_EVENT(kvm, 3, "GET: host  KMCTR  subfunc 0x%16.16lx.%16.16lx",
1825 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[0],
1826 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[1]);
1827 	VM_EVENT(kvm, 3, "GET: host  KMF    subfunc 0x%16.16lx.%16.16lx",
1828 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[0],
1829 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[1]);
1830 	VM_EVENT(kvm, 3, "GET: host  KMO    subfunc 0x%16.16lx.%16.16lx",
1831 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[0],
1832 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[1]);
1833 	VM_EVENT(kvm, 3, "GET: host  PCC    subfunc 0x%16.16lx.%16.16lx",
1834 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[0],
1835 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[1]);
1836 	VM_EVENT(kvm, 3, "GET: host  PPNO   subfunc 0x%16.16lx.%16.16lx",
1837 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[0],
1838 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[1]);
1839 	VM_EVENT(kvm, 3, "GET: host  KMA    subfunc 0x%16.16lx.%16.16lx",
1840 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[0],
1841 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[1]);
1842 	VM_EVENT(kvm, 3, "GET: host  KDSA   subfunc 0x%16.16lx.%16.16lx",
1843 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[0],
1844 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[1]);
1845 	VM_EVENT(kvm, 3, "GET: host  SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1846 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[0],
1847 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[1],
1848 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[2],
1849 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[3]);
1850 	VM_EVENT(kvm, 3, "GET: host  DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1851 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[0],
1852 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[1],
1853 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[2],
1854 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[3]);
1855 	VM_EVENT(kvm, 3, "GET: host  PFCR   subfunc 0x%16.16lx.%16.16lx",
1856 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[0],
1857 		 ((unsigned long *) &kvm_s390_available_subfunc.pfcr)[1]);
1858 
1859 	return 0;
1860 }
1861 
1862 static int kvm_s390_get_processor_uv_feat(struct kvm *kvm, struct kvm_device_attr *attr)
1863 {
1864 	struct kvm_s390_vm_cpu_uv_feat __user *dst = (void __user *)attr->addr;
1865 	unsigned long feat = kvm->arch.model.uv_feat_guest.feat;
1866 
1867 	if (put_user(feat, &dst->feat))
1868 		return -EFAULT;
1869 	VM_EVENT(kvm, 3, "GET: guest UV-feat: 0x%16.16lx", feat);
1870 
1871 	return 0;
1872 }
1873 
1874 static int kvm_s390_get_machine_uv_feat(struct kvm *kvm, struct kvm_device_attr *attr)
1875 {
1876 	struct kvm_s390_vm_cpu_uv_feat __user *dst = (void __user *)attr->addr;
1877 	unsigned long feat;
1878 
1879 	BUILD_BUG_ON(sizeof(*dst) != sizeof(uv_info.uv_feature_indications));
1880 
1881 	feat = uv_info.uv_feature_indications & KVM_S390_VM_CPU_UV_FEAT_GUEST_MASK;
1882 	if (put_user(feat, &dst->feat))
1883 		return -EFAULT;
1884 	VM_EVENT(kvm, 3, "GET: guest UV-feat: 0x%16.16lx", feat);
1885 
1886 	return 0;
1887 }
1888 
1889 static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1890 {
1891 	int ret = -ENXIO;
1892 
1893 	switch (attr->attr) {
1894 	case KVM_S390_VM_CPU_PROCESSOR:
1895 		ret = kvm_s390_get_processor(kvm, attr);
1896 		break;
1897 	case KVM_S390_VM_CPU_MACHINE:
1898 		ret = kvm_s390_get_machine(kvm, attr);
1899 		break;
1900 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1901 		ret = kvm_s390_get_processor_feat(kvm, attr);
1902 		break;
1903 	case KVM_S390_VM_CPU_MACHINE_FEAT:
1904 		ret = kvm_s390_get_machine_feat(kvm, attr);
1905 		break;
1906 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1907 		ret = kvm_s390_get_processor_subfunc(kvm, attr);
1908 		break;
1909 	case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1910 		ret = kvm_s390_get_machine_subfunc(kvm, attr);
1911 		break;
1912 	case KVM_S390_VM_CPU_PROCESSOR_UV_FEAT_GUEST:
1913 		ret = kvm_s390_get_processor_uv_feat(kvm, attr);
1914 		break;
1915 	case KVM_S390_VM_CPU_MACHINE_UV_FEAT_GUEST:
1916 		ret = kvm_s390_get_machine_uv_feat(kvm, attr);
1917 		break;
1918 	}
1919 	return ret;
1920 }
1921 
1922 /**
1923  * kvm_s390_update_topology_change_report - update CPU topology change report
1924  * @kvm: guest KVM description
1925  * @val: set or clear the MTCR bit
1926  *
1927  * Updates the Multiprocessor Topology-Change-Report bit to signal
1928  * the guest with a topology change.
1929  * This is only relevant if the topology facility is present.
1930  *
1931  * The SCA version, bsca or esca, doesn't matter as offset is the same.
1932  */
1933 static void kvm_s390_update_topology_change_report(struct kvm *kvm, bool val)
1934 {
1935 	union sca_utility new, old;
1936 	struct bsca_block *sca;
1937 
1938 	read_lock(&kvm->arch.sca_lock);
1939 	sca = kvm->arch.sca;
1940 	old = READ_ONCE(sca->utility);
1941 	do {
1942 		new = old;
1943 		new.mtcr = val;
1944 	} while (!try_cmpxchg(&sca->utility.val, &old.val, new.val));
1945 	read_unlock(&kvm->arch.sca_lock);
1946 }
1947 
1948 static int kvm_s390_set_topo_change_indication(struct kvm *kvm,
1949 					       struct kvm_device_attr *attr)
1950 {
1951 	if (!test_kvm_facility(kvm, 11))
1952 		return -ENXIO;
1953 
1954 	kvm_s390_update_topology_change_report(kvm, !!attr->attr);
1955 	return 0;
1956 }
1957 
1958 static int kvm_s390_get_topo_change_indication(struct kvm *kvm,
1959 					       struct kvm_device_attr *attr)
1960 {
1961 	u8 topo;
1962 
1963 	if (!test_kvm_facility(kvm, 11))
1964 		return -ENXIO;
1965 
1966 	read_lock(&kvm->arch.sca_lock);
1967 	topo = ((struct bsca_block *)kvm->arch.sca)->utility.mtcr;
1968 	read_unlock(&kvm->arch.sca_lock);
1969 
1970 	return put_user(topo, (u8 __user *)attr->addr);
1971 }
1972 
1973 static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1974 {
1975 	int ret;
1976 
1977 	switch (attr->group) {
1978 	case KVM_S390_VM_MEM_CTRL:
1979 		ret = kvm_s390_set_mem_control(kvm, attr);
1980 		break;
1981 	case KVM_S390_VM_TOD:
1982 		ret = kvm_s390_set_tod(kvm, attr);
1983 		break;
1984 	case KVM_S390_VM_CPU_MODEL:
1985 		ret = kvm_s390_set_cpu_model(kvm, attr);
1986 		break;
1987 	case KVM_S390_VM_CRYPTO:
1988 		ret = kvm_s390_vm_set_crypto(kvm, attr);
1989 		break;
1990 	case KVM_S390_VM_MIGRATION:
1991 		ret = kvm_s390_vm_set_migration(kvm, attr);
1992 		break;
1993 	case KVM_S390_VM_CPU_TOPOLOGY:
1994 		ret = kvm_s390_set_topo_change_indication(kvm, attr);
1995 		break;
1996 	default:
1997 		ret = -ENXIO;
1998 		break;
1999 	}
2000 
2001 	return ret;
2002 }
2003 
2004 static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
2005 {
2006 	int ret;
2007 
2008 	switch (attr->group) {
2009 	case KVM_S390_VM_MEM_CTRL:
2010 		ret = kvm_s390_get_mem_control(kvm, attr);
2011 		break;
2012 	case KVM_S390_VM_TOD:
2013 		ret = kvm_s390_get_tod(kvm, attr);
2014 		break;
2015 	case KVM_S390_VM_CPU_MODEL:
2016 		ret = kvm_s390_get_cpu_model(kvm, attr);
2017 		break;
2018 	case KVM_S390_VM_MIGRATION:
2019 		ret = kvm_s390_vm_get_migration(kvm, attr);
2020 		break;
2021 	case KVM_S390_VM_CPU_TOPOLOGY:
2022 		ret = kvm_s390_get_topo_change_indication(kvm, attr);
2023 		break;
2024 	default:
2025 		ret = -ENXIO;
2026 		break;
2027 	}
2028 
2029 	return ret;
2030 }
2031 
2032 static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
2033 {
2034 	int ret;
2035 
2036 	switch (attr->group) {
2037 	case KVM_S390_VM_MEM_CTRL:
2038 		switch (attr->attr) {
2039 		case KVM_S390_VM_MEM_ENABLE_CMMA:
2040 		case KVM_S390_VM_MEM_CLR_CMMA:
2041 			ret = sclp.has_cmma ? 0 : -ENXIO;
2042 			break;
2043 		case KVM_S390_VM_MEM_LIMIT_SIZE:
2044 			ret = 0;
2045 			break;
2046 		default:
2047 			ret = -ENXIO;
2048 			break;
2049 		}
2050 		break;
2051 	case KVM_S390_VM_TOD:
2052 		switch (attr->attr) {
2053 		case KVM_S390_VM_TOD_LOW:
2054 		case KVM_S390_VM_TOD_HIGH:
2055 			ret = 0;
2056 			break;
2057 		default:
2058 			ret = -ENXIO;
2059 			break;
2060 		}
2061 		break;
2062 	case KVM_S390_VM_CPU_MODEL:
2063 		switch (attr->attr) {
2064 		case KVM_S390_VM_CPU_PROCESSOR:
2065 		case KVM_S390_VM_CPU_MACHINE:
2066 		case KVM_S390_VM_CPU_PROCESSOR_FEAT:
2067 		case KVM_S390_VM_CPU_MACHINE_FEAT:
2068 		case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
2069 		case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
2070 		case KVM_S390_VM_CPU_MACHINE_UV_FEAT_GUEST:
2071 		case KVM_S390_VM_CPU_PROCESSOR_UV_FEAT_GUEST:
2072 			ret = 0;
2073 			break;
2074 		default:
2075 			ret = -ENXIO;
2076 			break;
2077 		}
2078 		break;
2079 	case KVM_S390_VM_CRYPTO:
2080 		switch (attr->attr) {
2081 		case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
2082 		case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
2083 		case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
2084 		case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
2085 			ret = 0;
2086 			break;
2087 		case KVM_S390_VM_CRYPTO_ENABLE_APIE:
2088 		case KVM_S390_VM_CRYPTO_DISABLE_APIE:
2089 			ret = ap_instructions_available() ? 0 : -ENXIO;
2090 			break;
2091 		default:
2092 			ret = -ENXIO;
2093 			break;
2094 		}
2095 		break;
2096 	case KVM_S390_VM_MIGRATION:
2097 		ret = 0;
2098 		break;
2099 	case KVM_S390_VM_CPU_TOPOLOGY:
2100 		ret = test_kvm_facility(kvm, 11) ? 0 : -ENXIO;
2101 		break;
2102 	default:
2103 		ret = -ENXIO;
2104 		break;
2105 	}
2106 
2107 	return ret;
2108 }
2109 
2110 static int kvm_s390_get_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
2111 {
2112 	uint8_t *keys;
2113 	uint64_t hva;
2114 	int srcu_idx, i, r = 0;
2115 
2116 	if (args->flags != 0)
2117 		return -EINVAL;
2118 
2119 	/* Is this guest using storage keys? */
2120 	if (!mm_uses_skeys(current->mm))
2121 		return KVM_S390_GET_SKEYS_NONE;
2122 
2123 	/* Enforce sane limit on memory allocation */
2124 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
2125 		return -EINVAL;
2126 
2127 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL_ACCOUNT);
2128 	if (!keys)
2129 		return -ENOMEM;
2130 
2131 	mmap_read_lock(current->mm);
2132 	srcu_idx = srcu_read_lock(&kvm->srcu);
2133 	for (i = 0; i < args->count; i++) {
2134 		hva = gfn_to_hva(kvm, args->start_gfn + i);
2135 		if (kvm_is_error_hva(hva)) {
2136 			r = -EFAULT;
2137 			break;
2138 		}
2139 
2140 		r = get_guest_storage_key(current->mm, hva, &keys[i]);
2141 		if (r)
2142 			break;
2143 	}
2144 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2145 	mmap_read_unlock(current->mm);
2146 
2147 	if (!r) {
2148 		r = copy_to_user((uint8_t __user *)args->skeydata_addr, keys,
2149 				 sizeof(uint8_t) * args->count);
2150 		if (r)
2151 			r = -EFAULT;
2152 	}
2153 
2154 	kvfree(keys);
2155 	return r;
2156 }
2157 
2158 static int kvm_s390_set_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
2159 {
2160 	uint8_t *keys;
2161 	uint64_t hva;
2162 	int srcu_idx, i, r = 0;
2163 	bool unlocked;
2164 
2165 	if (args->flags != 0)
2166 		return -EINVAL;
2167 
2168 	/* Enforce sane limit on memory allocation */
2169 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
2170 		return -EINVAL;
2171 
2172 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL_ACCOUNT);
2173 	if (!keys)
2174 		return -ENOMEM;
2175 
2176 	r = copy_from_user(keys, (uint8_t __user *)args->skeydata_addr,
2177 			   sizeof(uint8_t) * args->count);
2178 	if (r) {
2179 		r = -EFAULT;
2180 		goto out;
2181 	}
2182 
2183 	/* Enable storage key handling for the guest */
2184 	r = s390_enable_skey();
2185 	if (r)
2186 		goto out;
2187 
2188 	i = 0;
2189 	mmap_read_lock(current->mm);
2190 	srcu_idx = srcu_read_lock(&kvm->srcu);
2191         while (i < args->count) {
2192 		unlocked = false;
2193 		hva = gfn_to_hva(kvm, args->start_gfn + i);
2194 		if (kvm_is_error_hva(hva)) {
2195 			r = -EFAULT;
2196 			break;
2197 		}
2198 
2199 		/* Lowest order bit is reserved */
2200 		if (keys[i] & 0x01) {
2201 			r = -EINVAL;
2202 			break;
2203 		}
2204 
2205 		r = set_guest_storage_key(current->mm, hva, keys[i], 0);
2206 		if (r) {
2207 			r = fixup_user_fault(current->mm, hva,
2208 					     FAULT_FLAG_WRITE, &unlocked);
2209 			if (r)
2210 				break;
2211 		}
2212 		if (!r)
2213 			i++;
2214 	}
2215 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2216 	mmap_read_unlock(current->mm);
2217 out:
2218 	kvfree(keys);
2219 	return r;
2220 }
2221 
2222 /*
2223  * Base address and length must be sent at the start of each block, therefore
2224  * it's cheaper to send some clean data, as long as it's less than the size of
2225  * two longs.
2226  */
2227 #define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
2228 /* for consistency */
2229 #define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
2230 
2231 static int kvm_s390_peek_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
2232 			      u8 *res, unsigned long bufsize)
2233 {
2234 	unsigned long pgstev, hva, cur_gfn = args->start_gfn;
2235 
2236 	args->count = 0;
2237 	while (args->count < bufsize) {
2238 		hva = gfn_to_hva(kvm, cur_gfn);
2239 		/*
2240 		 * We return an error if the first value was invalid, but we
2241 		 * return successfully if at least one value was copied.
2242 		 */
2243 		if (kvm_is_error_hva(hva))
2244 			return args->count ? 0 : -EFAULT;
2245 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
2246 			pgstev = 0;
2247 		res[args->count++] = (pgstev >> 24) & 0x43;
2248 		cur_gfn++;
2249 	}
2250 
2251 	return 0;
2252 }
2253 
2254 static struct kvm_memory_slot *gfn_to_memslot_approx(struct kvm_memslots *slots,
2255 						     gfn_t gfn)
2256 {
2257 	return ____gfn_to_memslot(slots, gfn, true);
2258 }
2259 
2260 static unsigned long kvm_s390_next_dirty_cmma(struct kvm_memslots *slots,
2261 					      unsigned long cur_gfn)
2262 {
2263 	struct kvm_memory_slot *ms = gfn_to_memslot_approx(slots, cur_gfn);
2264 	unsigned long ofs = cur_gfn - ms->base_gfn;
2265 	struct rb_node *mnode = &ms->gfn_node[slots->node_idx];
2266 
2267 	if (ms->base_gfn + ms->npages <= cur_gfn) {
2268 		mnode = rb_next(mnode);
2269 		/* If we are above the highest slot, wrap around */
2270 		if (!mnode)
2271 			mnode = rb_first(&slots->gfn_tree);
2272 
2273 		ms = container_of(mnode, struct kvm_memory_slot, gfn_node[slots->node_idx]);
2274 		ofs = 0;
2275 	}
2276 
2277 	if (cur_gfn < ms->base_gfn)
2278 		ofs = 0;
2279 
2280 	ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, ofs);
2281 	while (ofs >= ms->npages && (mnode = rb_next(mnode))) {
2282 		ms = container_of(mnode, struct kvm_memory_slot, gfn_node[slots->node_idx]);
2283 		ofs = find_first_bit(kvm_second_dirty_bitmap(ms), ms->npages);
2284 	}
2285 	return ms->base_gfn + ofs;
2286 }
2287 
2288 static int kvm_s390_get_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
2289 			     u8 *res, unsigned long bufsize)
2290 {
2291 	unsigned long mem_end, cur_gfn, next_gfn, hva, pgstev;
2292 	struct kvm_memslots *slots = kvm_memslots(kvm);
2293 	struct kvm_memory_slot *ms;
2294 
2295 	if (unlikely(kvm_memslots_empty(slots)))
2296 		return 0;
2297 
2298 	cur_gfn = kvm_s390_next_dirty_cmma(slots, args->start_gfn);
2299 	ms = gfn_to_memslot(kvm, cur_gfn);
2300 	args->count = 0;
2301 	args->start_gfn = cur_gfn;
2302 	if (!ms)
2303 		return 0;
2304 	next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2305 	mem_end = kvm_s390_get_gfn_end(slots);
2306 
2307 	while (args->count < bufsize) {
2308 		hva = gfn_to_hva(kvm, cur_gfn);
2309 		if (kvm_is_error_hva(hva))
2310 			return 0;
2311 		/* Decrement only if we actually flipped the bit to 0 */
2312 		if (test_and_clear_bit(cur_gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
2313 			atomic64_dec(&kvm->arch.cmma_dirty_pages);
2314 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
2315 			pgstev = 0;
2316 		/* Save the value */
2317 		res[args->count++] = (pgstev >> 24) & 0x43;
2318 		/* If the next bit is too far away, stop. */
2319 		if (next_gfn > cur_gfn + KVM_S390_MAX_BIT_DISTANCE)
2320 			return 0;
2321 		/* If we reached the previous "next", find the next one */
2322 		if (cur_gfn == next_gfn)
2323 			next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2324 		/* Reached the end of memory or of the buffer, stop */
2325 		if ((next_gfn >= mem_end) ||
2326 		    (next_gfn - args->start_gfn >= bufsize))
2327 			return 0;
2328 		cur_gfn++;
2329 		/* Reached the end of the current memslot, take the next one. */
2330 		if (cur_gfn - ms->base_gfn >= ms->npages) {
2331 			ms = gfn_to_memslot(kvm, cur_gfn);
2332 			if (!ms)
2333 				return 0;
2334 		}
2335 	}
2336 	return 0;
2337 }
2338 
2339 /*
2340  * This function searches for the next page with dirty CMMA attributes, and
2341  * saves the attributes in the buffer up to either the end of the buffer or
2342  * until a block of at least KVM_S390_MAX_BIT_DISTANCE clean bits is found;
2343  * no trailing clean bytes are saved.
2344  * In case no dirty bits were found, or if CMMA was not enabled or used, the
2345  * output buffer will indicate 0 as length.
2346  */
2347 static int kvm_s390_get_cmma_bits(struct kvm *kvm,
2348 				  struct kvm_s390_cmma_log *args)
2349 {
2350 	unsigned long bufsize;
2351 	int srcu_idx, peek, ret;
2352 	u8 *values;
2353 
2354 	if (!kvm->arch.use_cmma)
2355 		return -ENXIO;
2356 	/* Invalid/unsupported flags were specified */
2357 	if (args->flags & ~KVM_S390_CMMA_PEEK)
2358 		return -EINVAL;
2359 	/* Migration mode query, and we are not doing a migration */
2360 	peek = !!(args->flags & KVM_S390_CMMA_PEEK);
2361 	if (!peek && !kvm->arch.migration_mode)
2362 		return -EINVAL;
2363 	/* CMMA is disabled or was not used, or the buffer has length zero */
2364 	bufsize = min(args->count, KVM_S390_CMMA_SIZE_MAX);
2365 	if (!bufsize || !kvm->mm->context.uses_cmm) {
2366 		memset(args, 0, sizeof(*args));
2367 		return 0;
2368 	}
2369 	/* We are not peeking, and there are no dirty pages */
2370 	if (!peek && !atomic64_read(&kvm->arch.cmma_dirty_pages)) {
2371 		memset(args, 0, sizeof(*args));
2372 		return 0;
2373 	}
2374 
2375 	values = vmalloc(bufsize);
2376 	if (!values)
2377 		return -ENOMEM;
2378 
2379 	mmap_read_lock(kvm->mm);
2380 	srcu_idx = srcu_read_lock(&kvm->srcu);
2381 	if (peek)
2382 		ret = kvm_s390_peek_cmma(kvm, args, values, bufsize);
2383 	else
2384 		ret = kvm_s390_get_cmma(kvm, args, values, bufsize);
2385 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2386 	mmap_read_unlock(kvm->mm);
2387 
2388 	if (kvm->arch.migration_mode)
2389 		args->remaining = atomic64_read(&kvm->arch.cmma_dirty_pages);
2390 	else
2391 		args->remaining = 0;
2392 
2393 	if (copy_to_user((void __user *)args->values, values, args->count))
2394 		ret = -EFAULT;
2395 
2396 	vfree(values);
2397 	return ret;
2398 }
2399 
2400 /*
2401  * This function sets the CMMA attributes for the given pages. If the input
2402  * buffer has zero length, no action is taken, otherwise the attributes are
2403  * set and the mm->context.uses_cmm flag is set.
2404  */
2405 static int kvm_s390_set_cmma_bits(struct kvm *kvm,
2406 				  const struct kvm_s390_cmma_log *args)
2407 {
2408 	unsigned long hva, mask, pgstev, i;
2409 	uint8_t *bits;
2410 	int srcu_idx, r = 0;
2411 
2412 	mask = args->mask;
2413 
2414 	if (!kvm->arch.use_cmma)
2415 		return -ENXIO;
2416 	/* invalid/unsupported flags */
2417 	if (args->flags != 0)
2418 		return -EINVAL;
2419 	/* Enforce sane limit on memory allocation */
2420 	if (args->count > KVM_S390_CMMA_SIZE_MAX)
2421 		return -EINVAL;
2422 	/* Nothing to do */
2423 	if (args->count == 0)
2424 		return 0;
2425 
2426 	bits = vmalloc(array_size(sizeof(*bits), args->count));
2427 	if (!bits)
2428 		return -ENOMEM;
2429 
2430 	r = copy_from_user(bits, (void __user *)args->values, args->count);
2431 	if (r) {
2432 		r = -EFAULT;
2433 		goto out;
2434 	}
2435 
2436 	mmap_read_lock(kvm->mm);
2437 	srcu_idx = srcu_read_lock(&kvm->srcu);
2438 	for (i = 0; i < args->count; i++) {
2439 		hva = gfn_to_hva(kvm, args->start_gfn + i);
2440 		if (kvm_is_error_hva(hva)) {
2441 			r = -EFAULT;
2442 			break;
2443 		}
2444 
2445 		pgstev = bits[i];
2446 		pgstev = pgstev << 24;
2447 		mask &= _PGSTE_GPS_USAGE_MASK | _PGSTE_GPS_NODAT;
2448 		set_pgste_bits(kvm->mm, hva, mask, pgstev);
2449 	}
2450 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2451 	mmap_read_unlock(kvm->mm);
2452 
2453 	if (!kvm->mm->context.uses_cmm) {
2454 		mmap_write_lock(kvm->mm);
2455 		kvm->mm->context.uses_cmm = 1;
2456 		mmap_write_unlock(kvm->mm);
2457 	}
2458 out:
2459 	vfree(bits);
2460 	return r;
2461 }
2462 
2463 /**
2464  * kvm_s390_cpus_from_pv - Convert all protected vCPUs in a protected VM to
2465  * non protected.
2466  * @kvm: the VM whose protected vCPUs are to be converted
2467  * @rc: return value for the RC field of the UVC (in case of error)
2468  * @rrc: return value for the RRC field of the UVC (in case of error)
2469  *
2470  * Does not stop in case of error, tries to convert as many
2471  * CPUs as possible. In case of error, the RC and RRC of the last error are
2472  * returned.
2473  *
2474  * Return: 0 in case of success, otherwise -EIO
2475  */
2476 int kvm_s390_cpus_from_pv(struct kvm *kvm, u16 *rc, u16 *rrc)
2477 {
2478 	struct kvm_vcpu *vcpu;
2479 	unsigned long i;
2480 	u16 _rc, _rrc;
2481 	int ret = 0;
2482 
2483 	/*
2484 	 * We ignore failures and try to destroy as many CPUs as possible.
2485 	 * At the same time we must not free the assigned resources when
2486 	 * this fails, as the ultravisor has still access to that memory.
2487 	 * So kvm_s390_pv_destroy_cpu can leave a "wanted" memory leak
2488 	 * behind.
2489 	 * We want to return the first failure rc and rrc, though.
2490 	 */
2491 	kvm_for_each_vcpu(i, vcpu, kvm) {
2492 		mutex_lock(&vcpu->mutex);
2493 		if (kvm_s390_pv_destroy_cpu(vcpu, &_rc, &_rrc) && !ret) {
2494 			*rc = _rc;
2495 			*rrc = _rrc;
2496 			ret = -EIO;
2497 		}
2498 		mutex_unlock(&vcpu->mutex);
2499 	}
2500 	/* Ensure that we re-enable gisa if the non-PV guest used it but the PV guest did not. */
2501 	if (use_gisa)
2502 		kvm_s390_gisa_enable(kvm);
2503 	return ret;
2504 }
2505 
2506 /**
2507  * kvm_s390_cpus_to_pv - Convert all non-protected vCPUs in a protected VM
2508  * to protected.
2509  * @kvm: the VM whose protected vCPUs are to be converted
2510  * @rc: return value for the RC field of the UVC (in case of error)
2511  * @rrc: return value for the RRC field of the UVC (in case of error)
2512  *
2513  * Tries to undo the conversion in case of error.
2514  *
2515  * Return: 0 in case of success, otherwise -EIO
2516  */
2517 static int kvm_s390_cpus_to_pv(struct kvm *kvm, u16 *rc, u16 *rrc)
2518 {
2519 	unsigned long i;
2520 	int r = 0;
2521 	u16 dummy;
2522 
2523 	struct kvm_vcpu *vcpu;
2524 
2525 	/* Disable the GISA if the ultravisor does not support AIV. */
2526 	if (!uv_has_feature(BIT_UV_FEAT_AIV))
2527 		kvm_s390_gisa_disable(kvm);
2528 
2529 	kvm_for_each_vcpu(i, vcpu, kvm) {
2530 		mutex_lock(&vcpu->mutex);
2531 		r = kvm_s390_pv_create_cpu(vcpu, rc, rrc);
2532 		mutex_unlock(&vcpu->mutex);
2533 		if (r)
2534 			break;
2535 	}
2536 	if (r)
2537 		kvm_s390_cpus_from_pv(kvm, &dummy, &dummy);
2538 	return r;
2539 }
2540 
2541 /*
2542  * Here we provide user space with a direct interface to query UV
2543  * related data like UV maxima and available features as well as
2544  * feature specific data.
2545  *
2546  * To facilitate future extension of the data structures we'll try to
2547  * write data up to the maximum requested length.
2548  */
2549 static ssize_t kvm_s390_handle_pv_info(struct kvm_s390_pv_info *info)
2550 {
2551 	ssize_t len_min;
2552 
2553 	switch (info->header.id) {
2554 	case KVM_PV_INFO_VM: {
2555 		len_min =  sizeof(info->header) + sizeof(info->vm);
2556 
2557 		if (info->header.len_max < len_min)
2558 			return -EINVAL;
2559 
2560 		memcpy(info->vm.inst_calls_list,
2561 		       uv_info.inst_calls_list,
2562 		       sizeof(uv_info.inst_calls_list));
2563 
2564 		/* It's max cpuid not max cpus, so it's off by one */
2565 		info->vm.max_cpus = uv_info.max_guest_cpu_id + 1;
2566 		info->vm.max_guests = uv_info.max_num_sec_conf;
2567 		info->vm.max_guest_addr = uv_info.max_sec_stor_addr;
2568 		info->vm.feature_indication = uv_info.uv_feature_indications;
2569 
2570 		return len_min;
2571 	}
2572 	case KVM_PV_INFO_DUMP: {
2573 		len_min =  sizeof(info->header) + sizeof(info->dump);
2574 
2575 		if (info->header.len_max < len_min)
2576 			return -EINVAL;
2577 
2578 		info->dump.dump_cpu_buffer_len = uv_info.guest_cpu_stor_len;
2579 		info->dump.dump_config_mem_buffer_per_1m = uv_info.conf_dump_storage_state_len;
2580 		info->dump.dump_config_finalize_len = uv_info.conf_dump_finalize_len;
2581 		return len_min;
2582 	}
2583 	default:
2584 		return -EINVAL;
2585 	}
2586 }
2587 
2588 static int kvm_s390_pv_dmp(struct kvm *kvm, struct kvm_pv_cmd *cmd,
2589 			   struct kvm_s390_pv_dmp dmp)
2590 {
2591 	int r = -EINVAL;
2592 	void __user *result_buff = (void __user *)dmp.buff_addr;
2593 
2594 	switch (dmp.subcmd) {
2595 	case KVM_PV_DUMP_INIT: {
2596 		if (kvm->arch.pv.dumping)
2597 			break;
2598 
2599 		/*
2600 		 * Block SIE entry as concurrent dump UVCs could lead
2601 		 * to validities.
2602 		 */
2603 		kvm_s390_vcpu_block_all(kvm);
2604 
2605 		r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2606 				  UVC_CMD_DUMP_INIT, &cmd->rc, &cmd->rrc);
2607 		KVM_UV_EVENT(kvm, 3, "PROTVIRT DUMP INIT: rc %x rrc %x",
2608 			     cmd->rc, cmd->rrc);
2609 		if (!r) {
2610 			kvm->arch.pv.dumping = true;
2611 		} else {
2612 			kvm_s390_vcpu_unblock_all(kvm);
2613 			r = -EINVAL;
2614 		}
2615 		break;
2616 	}
2617 	case KVM_PV_DUMP_CONFIG_STOR_STATE: {
2618 		if (!kvm->arch.pv.dumping)
2619 			break;
2620 
2621 		/*
2622 		 * gaddr is an output parameter since we might stop
2623 		 * early. As dmp will be copied back in our caller, we
2624 		 * don't need to do it ourselves.
2625 		 */
2626 		r = kvm_s390_pv_dump_stor_state(kvm, result_buff, &dmp.gaddr, dmp.buff_len,
2627 						&cmd->rc, &cmd->rrc);
2628 		break;
2629 	}
2630 	case KVM_PV_DUMP_COMPLETE: {
2631 		if (!kvm->arch.pv.dumping)
2632 			break;
2633 
2634 		r = -EINVAL;
2635 		if (dmp.buff_len < uv_info.conf_dump_finalize_len)
2636 			break;
2637 
2638 		r = kvm_s390_pv_dump_complete(kvm, result_buff,
2639 					      &cmd->rc, &cmd->rrc);
2640 		break;
2641 	}
2642 	default:
2643 		r = -ENOTTY;
2644 		break;
2645 	}
2646 
2647 	return r;
2648 }
2649 
2650 static int kvm_s390_handle_pv(struct kvm *kvm, struct kvm_pv_cmd *cmd)
2651 {
2652 	const bool need_lock = (cmd->cmd != KVM_PV_ASYNC_CLEANUP_PERFORM);
2653 	void __user *argp = (void __user *)cmd->data;
2654 	int r = 0;
2655 	u16 dummy;
2656 
2657 	if (need_lock)
2658 		mutex_lock(&kvm->lock);
2659 
2660 	switch (cmd->cmd) {
2661 	case KVM_PV_ENABLE: {
2662 		r = -EINVAL;
2663 		if (kvm_s390_pv_is_protected(kvm))
2664 			break;
2665 
2666 		/*
2667 		 *  FMT 4 SIE needs esca. As we never switch back to bsca from
2668 		 *  esca, we need no cleanup in the error cases below
2669 		 */
2670 		r = sca_switch_to_extended(kvm);
2671 		if (r)
2672 			break;
2673 
2674 		r = s390_disable_cow_sharing();
2675 		if (r)
2676 			break;
2677 
2678 		r = kvm_s390_pv_init_vm(kvm, &cmd->rc, &cmd->rrc);
2679 		if (r)
2680 			break;
2681 
2682 		r = kvm_s390_cpus_to_pv(kvm, &cmd->rc, &cmd->rrc);
2683 		if (r)
2684 			kvm_s390_pv_deinit_vm(kvm, &dummy, &dummy);
2685 
2686 		/* we need to block service interrupts from now on */
2687 		set_bit(IRQ_PEND_EXT_SERVICE, &kvm->arch.float_int.masked_irqs);
2688 		break;
2689 	}
2690 	case KVM_PV_ASYNC_CLEANUP_PREPARE:
2691 		r = -EINVAL;
2692 		if (!kvm_s390_pv_is_protected(kvm) || !async_destroy)
2693 			break;
2694 
2695 		r = kvm_s390_cpus_from_pv(kvm, &cmd->rc, &cmd->rrc);
2696 		/*
2697 		 * If a CPU could not be destroyed, destroy VM will also fail.
2698 		 * There is no point in trying to destroy it. Instead return
2699 		 * the rc and rrc from the first CPU that failed destroying.
2700 		 */
2701 		if (r)
2702 			break;
2703 		r = kvm_s390_pv_set_aside(kvm, &cmd->rc, &cmd->rrc);
2704 
2705 		/* no need to block service interrupts any more */
2706 		clear_bit(IRQ_PEND_EXT_SERVICE, &kvm->arch.float_int.masked_irqs);
2707 		break;
2708 	case KVM_PV_ASYNC_CLEANUP_PERFORM:
2709 		r = -EINVAL;
2710 		if (!async_destroy)
2711 			break;
2712 		/* kvm->lock must not be held; this is asserted inside the function. */
2713 		r = kvm_s390_pv_deinit_aside_vm(kvm, &cmd->rc, &cmd->rrc);
2714 		break;
2715 	case KVM_PV_DISABLE: {
2716 		r = -EINVAL;
2717 		if (!kvm_s390_pv_is_protected(kvm))
2718 			break;
2719 
2720 		r = kvm_s390_cpus_from_pv(kvm, &cmd->rc, &cmd->rrc);
2721 		/*
2722 		 * If a CPU could not be destroyed, destroy VM will also fail.
2723 		 * There is no point in trying to destroy it. Instead return
2724 		 * the rc and rrc from the first CPU that failed destroying.
2725 		 */
2726 		if (r)
2727 			break;
2728 		r = kvm_s390_pv_deinit_cleanup_all(kvm, &cmd->rc, &cmd->rrc);
2729 
2730 		/* no need to block service interrupts any more */
2731 		clear_bit(IRQ_PEND_EXT_SERVICE, &kvm->arch.float_int.masked_irqs);
2732 		break;
2733 	}
2734 	case KVM_PV_SET_SEC_PARMS: {
2735 		struct kvm_s390_pv_sec_parm parms = {};
2736 		void *hdr;
2737 
2738 		r = -EINVAL;
2739 		if (!kvm_s390_pv_is_protected(kvm))
2740 			break;
2741 
2742 		r = -EFAULT;
2743 		if (copy_from_user(&parms, argp, sizeof(parms)))
2744 			break;
2745 
2746 		/* Currently restricted to 8KB */
2747 		r = -EINVAL;
2748 		if (parms.length > PAGE_SIZE * 2)
2749 			break;
2750 
2751 		r = -ENOMEM;
2752 		hdr = vmalloc(parms.length);
2753 		if (!hdr)
2754 			break;
2755 
2756 		r = -EFAULT;
2757 		if (!copy_from_user(hdr, (void __user *)parms.origin,
2758 				    parms.length))
2759 			r = kvm_s390_pv_set_sec_parms(kvm, hdr, parms.length,
2760 						      &cmd->rc, &cmd->rrc);
2761 
2762 		vfree(hdr);
2763 		break;
2764 	}
2765 	case KVM_PV_UNPACK: {
2766 		struct kvm_s390_pv_unp unp = {};
2767 
2768 		r = -EINVAL;
2769 		if (!kvm_s390_pv_is_protected(kvm) || !mm_is_protected(kvm->mm))
2770 			break;
2771 
2772 		r = -EFAULT;
2773 		if (copy_from_user(&unp, argp, sizeof(unp)))
2774 			break;
2775 
2776 		r = kvm_s390_pv_unpack(kvm, unp.addr, unp.size, unp.tweak,
2777 				       &cmd->rc, &cmd->rrc);
2778 		break;
2779 	}
2780 	case KVM_PV_VERIFY: {
2781 		r = -EINVAL;
2782 		if (!kvm_s390_pv_is_protected(kvm))
2783 			break;
2784 
2785 		r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2786 				  UVC_CMD_VERIFY_IMG, &cmd->rc, &cmd->rrc);
2787 		KVM_UV_EVENT(kvm, 3, "PROTVIRT VERIFY: rc %x rrc %x", cmd->rc,
2788 			     cmd->rrc);
2789 		break;
2790 	}
2791 	case KVM_PV_PREP_RESET: {
2792 		r = -EINVAL;
2793 		if (!kvm_s390_pv_is_protected(kvm))
2794 			break;
2795 
2796 		r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2797 				  UVC_CMD_PREPARE_RESET, &cmd->rc, &cmd->rrc);
2798 		KVM_UV_EVENT(kvm, 3, "PROTVIRT PREP RESET: rc %x rrc %x",
2799 			     cmd->rc, cmd->rrc);
2800 		break;
2801 	}
2802 	case KVM_PV_UNSHARE_ALL: {
2803 		r = -EINVAL;
2804 		if (!kvm_s390_pv_is_protected(kvm))
2805 			break;
2806 
2807 		r = uv_cmd_nodata(kvm_s390_pv_get_handle(kvm),
2808 				  UVC_CMD_SET_UNSHARE_ALL, &cmd->rc, &cmd->rrc);
2809 		KVM_UV_EVENT(kvm, 3, "PROTVIRT UNSHARE: rc %x rrc %x",
2810 			     cmd->rc, cmd->rrc);
2811 		break;
2812 	}
2813 	case KVM_PV_INFO: {
2814 		struct kvm_s390_pv_info info = {};
2815 		ssize_t data_len;
2816 
2817 		/*
2818 		 * No need to check the VM protection here.
2819 		 *
2820 		 * Maybe user space wants to query some of the data
2821 		 * when the VM is still unprotected. If we see the
2822 		 * need to fence a new data command we can still
2823 		 * return an error in the info handler.
2824 		 */
2825 
2826 		r = -EFAULT;
2827 		if (copy_from_user(&info, argp, sizeof(info.header)))
2828 			break;
2829 
2830 		r = -EINVAL;
2831 		if (info.header.len_max < sizeof(info.header))
2832 			break;
2833 
2834 		data_len = kvm_s390_handle_pv_info(&info);
2835 		if (data_len < 0) {
2836 			r = data_len;
2837 			break;
2838 		}
2839 		/*
2840 		 * If a data command struct is extended (multiple
2841 		 * times) this can be used to determine how much of it
2842 		 * is valid.
2843 		 */
2844 		info.header.len_written = data_len;
2845 
2846 		r = -EFAULT;
2847 		if (copy_to_user(argp, &info, data_len))
2848 			break;
2849 
2850 		r = 0;
2851 		break;
2852 	}
2853 	case KVM_PV_DUMP: {
2854 		struct kvm_s390_pv_dmp dmp;
2855 
2856 		r = -EINVAL;
2857 		if (!kvm_s390_pv_is_protected(kvm))
2858 			break;
2859 
2860 		r = -EFAULT;
2861 		if (copy_from_user(&dmp, argp, sizeof(dmp)))
2862 			break;
2863 
2864 		r = kvm_s390_pv_dmp(kvm, cmd, dmp);
2865 		if (r)
2866 			break;
2867 
2868 		if (copy_to_user(argp, &dmp, sizeof(dmp))) {
2869 			r = -EFAULT;
2870 			break;
2871 		}
2872 
2873 		break;
2874 	}
2875 	default:
2876 		r = -ENOTTY;
2877 	}
2878 	if (need_lock)
2879 		mutex_unlock(&kvm->lock);
2880 
2881 	return r;
2882 }
2883 
2884 static int mem_op_validate_common(struct kvm_s390_mem_op *mop, u64 supported_flags)
2885 {
2886 	if (mop->flags & ~supported_flags || !mop->size)
2887 		return -EINVAL;
2888 	if (mop->size > MEM_OP_MAX_SIZE)
2889 		return -E2BIG;
2890 	if (mop->flags & KVM_S390_MEMOP_F_SKEY_PROTECTION) {
2891 		if (mop->key > 0xf)
2892 			return -EINVAL;
2893 	} else {
2894 		mop->key = 0;
2895 	}
2896 	return 0;
2897 }
2898 
2899 static int kvm_s390_vm_mem_op_abs(struct kvm *kvm, struct kvm_s390_mem_op *mop)
2900 {
2901 	void __user *uaddr = (void __user *)mop->buf;
2902 	enum gacc_mode acc_mode;
2903 	void *tmpbuf = NULL;
2904 	int r, srcu_idx;
2905 
2906 	r = mem_op_validate_common(mop, KVM_S390_MEMOP_F_SKEY_PROTECTION |
2907 					KVM_S390_MEMOP_F_CHECK_ONLY);
2908 	if (r)
2909 		return r;
2910 
2911 	if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
2912 		tmpbuf = vmalloc(mop->size);
2913 		if (!tmpbuf)
2914 			return -ENOMEM;
2915 	}
2916 
2917 	srcu_idx = srcu_read_lock(&kvm->srcu);
2918 
2919 	if (!kvm_is_gpa_in_memslot(kvm, mop->gaddr)) {
2920 		r = PGM_ADDRESSING;
2921 		goto out_unlock;
2922 	}
2923 
2924 	acc_mode = mop->op == KVM_S390_MEMOP_ABSOLUTE_READ ? GACC_FETCH : GACC_STORE;
2925 	if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
2926 		r = check_gpa_range(kvm, mop->gaddr, mop->size, acc_mode, mop->key);
2927 		goto out_unlock;
2928 	}
2929 	if (acc_mode == GACC_FETCH) {
2930 		r = access_guest_abs_with_key(kvm, mop->gaddr, tmpbuf,
2931 					      mop->size, GACC_FETCH, mop->key);
2932 		if (r)
2933 			goto out_unlock;
2934 		if (copy_to_user(uaddr, tmpbuf, mop->size))
2935 			r = -EFAULT;
2936 	} else {
2937 		if (copy_from_user(tmpbuf, uaddr, mop->size)) {
2938 			r = -EFAULT;
2939 			goto out_unlock;
2940 		}
2941 		r = access_guest_abs_with_key(kvm, mop->gaddr, tmpbuf,
2942 					      mop->size, GACC_STORE, mop->key);
2943 	}
2944 
2945 out_unlock:
2946 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2947 
2948 	vfree(tmpbuf);
2949 	return r;
2950 }
2951 
2952 static int kvm_s390_vm_mem_op_cmpxchg(struct kvm *kvm, struct kvm_s390_mem_op *mop)
2953 {
2954 	void __user *uaddr = (void __user *)mop->buf;
2955 	void __user *old_addr = (void __user *)mop->old_addr;
2956 	union {
2957 		__uint128_t quad;
2958 		char raw[sizeof(__uint128_t)];
2959 	} old = { .quad = 0}, new = { .quad = 0 };
2960 	unsigned int off_in_quad = sizeof(new) - mop->size;
2961 	int r, srcu_idx;
2962 	bool success;
2963 
2964 	r = mem_op_validate_common(mop, KVM_S390_MEMOP_F_SKEY_PROTECTION);
2965 	if (r)
2966 		return r;
2967 	/*
2968 	 * This validates off_in_quad. Checking that size is a power
2969 	 * of two is not necessary, as cmpxchg_guest_abs_with_key
2970 	 * takes care of that
2971 	 */
2972 	if (mop->size > sizeof(new))
2973 		return -EINVAL;
2974 	if (copy_from_user(&new.raw[off_in_quad], uaddr, mop->size))
2975 		return -EFAULT;
2976 	if (copy_from_user(&old.raw[off_in_quad], old_addr, mop->size))
2977 		return -EFAULT;
2978 
2979 	srcu_idx = srcu_read_lock(&kvm->srcu);
2980 
2981 	if (!kvm_is_gpa_in_memslot(kvm, mop->gaddr)) {
2982 		r = PGM_ADDRESSING;
2983 		goto out_unlock;
2984 	}
2985 
2986 	r = cmpxchg_guest_abs_with_key(kvm, mop->gaddr, mop->size, &old.quad,
2987 				       new.quad, mop->key, &success);
2988 	if (!success && copy_to_user(old_addr, &old.raw[off_in_quad], mop->size))
2989 		r = -EFAULT;
2990 
2991 out_unlock:
2992 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2993 	return r;
2994 }
2995 
2996 static int kvm_s390_vm_mem_op(struct kvm *kvm, struct kvm_s390_mem_op *mop)
2997 {
2998 	/*
2999 	 * This is technically a heuristic only, if the kvm->lock is not
3000 	 * taken, it is not guaranteed that the vm is/remains non-protected.
3001 	 * This is ok from a kernel perspective, wrongdoing is detected
3002 	 * on the access, -EFAULT is returned and the vm may crash the
3003 	 * next time it accesses the memory in question.
3004 	 * There is no sane usecase to do switching and a memop on two
3005 	 * different CPUs at the same time.
3006 	 */
3007 	if (kvm_s390_pv_get_handle(kvm))
3008 		return -EINVAL;
3009 
3010 	switch (mop->op) {
3011 	case KVM_S390_MEMOP_ABSOLUTE_READ:
3012 	case KVM_S390_MEMOP_ABSOLUTE_WRITE:
3013 		return kvm_s390_vm_mem_op_abs(kvm, mop);
3014 	case KVM_S390_MEMOP_ABSOLUTE_CMPXCHG:
3015 		return kvm_s390_vm_mem_op_cmpxchg(kvm, mop);
3016 	default:
3017 		return -EINVAL;
3018 	}
3019 }
3020 
3021 int kvm_arch_vm_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
3022 {
3023 	struct kvm *kvm = filp->private_data;
3024 	void __user *argp = (void __user *)arg;
3025 	struct kvm_device_attr attr;
3026 	int r;
3027 
3028 	switch (ioctl) {
3029 	case KVM_S390_INTERRUPT: {
3030 		struct kvm_s390_interrupt s390int;
3031 
3032 		r = -EFAULT;
3033 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
3034 			break;
3035 		r = kvm_s390_inject_vm(kvm, &s390int);
3036 		break;
3037 	}
3038 	case KVM_CREATE_IRQCHIP: {
3039 		r = -EINVAL;
3040 		if (kvm->arch.use_irqchip)
3041 			r = 0;
3042 		break;
3043 	}
3044 	case KVM_SET_DEVICE_ATTR: {
3045 		r = -EFAULT;
3046 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
3047 			break;
3048 		r = kvm_s390_vm_set_attr(kvm, &attr);
3049 		break;
3050 	}
3051 	case KVM_GET_DEVICE_ATTR: {
3052 		r = -EFAULT;
3053 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
3054 			break;
3055 		r = kvm_s390_vm_get_attr(kvm, &attr);
3056 		break;
3057 	}
3058 	case KVM_HAS_DEVICE_ATTR: {
3059 		r = -EFAULT;
3060 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
3061 			break;
3062 		r = kvm_s390_vm_has_attr(kvm, &attr);
3063 		break;
3064 	}
3065 	case KVM_S390_GET_SKEYS: {
3066 		struct kvm_s390_skeys args;
3067 
3068 		r = -EFAULT;
3069 		if (copy_from_user(&args, argp,
3070 				   sizeof(struct kvm_s390_skeys)))
3071 			break;
3072 		r = kvm_s390_get_skeys(kvm, &args);
3073 		break;
3074 	}
3075 	case KVM_S390_SET_SKEYS: {
3076 		struct kvm_s390_skeys args;
3077 
3078 		r = -EFAULT;
3079 		if (copy_from_user(&args, argp,
3080 				   sizeof(struct kvm_s390_skeys)))
3081 			break;
3082 		r = kvm_s390_set_skeys(kvm, &args);
3083 		break;
3084 	}
3085 	case KVM_S390_GET_CMMA_BITS: {
3086 		struct kvm_s390_cmma_log args;
3087 
3088 		r = -EFAULT;
3089 		if (copy_from_user(&args, argp, sizeof(args)))
3090 			break;
3091 		mutex_lock(&kvm->slots_lock);
3092 		r = kvm_s390_get_cmma_bits(kvm, &args);
3093 		mutex_unlock(&kvm->slots_lock);
3094 		if (!r) {
3095 			r = copy_to_user(argp, &args, sizeof(args));
3096 			if (r)
3097 				r = -EFAULT;
3098 		}
3099 		break;
3100 	}
3101 	case KVM_S390_SET_CMMA_BITS: {
3102 		struct kvm_s390_cmma_log args;
3103 
3104 		r = -EFAULT;
3105 		if (copy_from_user(&args, argp, sizeof(args)))
3106 			break;
3107 		mutex_lock(&kvm->slots_lock);
3108 		r = kvm_s390_set_cmma_bits(kvm, &args);
3109 		mutex_unlock(&kvm->slots_lock);
3110 		break;
3111 	}
3112 	case KVM_S390_PV_COMMAND: {
3113 		struct kvm_pv_cmd args;
3114 
3115 		/* protvirt means user cpu state */
3116 		kvm_s390_set_user_cpu_state_ctrl(kvm);
3117 		r = 0;
3118 		if (!is_prot_virt_host()) {
3119 			r = -EINVAL;
3120 			break;
3121 		}
3122 		if (copy_from_user(&args, argp, sizeof(args))) {
3123 			r = -EFAULT;
3124 			break;
3125 		}
3126 		if (args.flags) {
3127 			r = -EINVAL;
3128 			break;
3129 		}
3130 		/* must be called without kvm->lock */
3131 		r = kvm_s390_handle_pv(kvm, &args);
3132 		if (copy_to_user(argp, &args, sizeof(args))) {
3133 			r = -EFAULT;
3134 			break;
3135 		}
3136 		break;
3137 	}
3138 	case KVM_S390_MEM_OP: {
3139 		struct kvm_s390_mem_op mem_op;
3140 
3141 		if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
3142 			r = kvm_s390_vm_mem_op(kvm, &mem_op);
3143 		else
3144 			r = -EFAULT;
3145 		break;
3146 	}
3147 	case KVM_S390_ZPCI_OP: {
3148 		struct kvm_s390_zpci_op args;
3149 
3150 		r = -EINVAL;
3151 		if (!IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM))
3152 			break;
3153 		if (copy_from_user(&args, argp, sizeof(args))) {
3154 			r = -EFAULT;
3155 			break;
3156 		}
3157 		r = kvm_s390_pci_zpci_op(kvm, &args);
3158 		break;
3159 	}
3160 	default:
3161 		r = -ENOTTY;
3162 	}
3163 
3164 	return r;
3165 }
3166 
3167 static int kvm_s390_apxa_installed(void)
3168 {
3169 	struct ap_config_info info;
3170 
3171 	if (ap_instructions_available()) {
3172 		if (ap_qci(&info) == 0)
3173 			return info.apxa;
3174 	}
3175 
3176 	return 0;
3177 }
3178 
3179 /*
3180  * The format of the crypto control block (CRYCB) is specified in the 3 low
3181  * order bits of the CRYCB designation (CRYCBD) field as follows:
3182  * Format 0: Neither the message security assist extension 3 (MSAX3) nor the
3183  *	     AP extended addressing (APXA) facility are installed.
3184  * Format 1: The APXA facility is not installed but the MSAX3 facility is.
3185  * Format 2: Both the APXA and MSAX3 facilities are installed
3186  */
3187 static void kvm_s390_set_crycb_format(struct kvm *kvm)
3188 {
3189 	kvm->arch.crypto.crycbd = virt_to_phys(kvm->arch.crypto.crycb);
3190 
3191 	/* Clear the CRYCB format bits - i.e., set format 0 by default */
3192 	kvm->arch.crypto.crycbd &= ~(CRYCB_FORMAT_MASK);
3193 
3194 	/* Check whether MSAX3 is installed */
3195 	if (!test_kvm_facility(kvm, 76))
3196 		return;
3197 
3198 	if (kvm_s390_apxa_installed())
3199 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT2;
3200 	else
3201 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT1;
3202 }
3203 
3204 /*
3205  * kvm_arch_crypto_set_masks
3206  *
3207  * @kvm: pointer to the target guest's KVM struct containing the crypto masks
3208  *	 to be set.
3209  * @apm: the mask identifying the accessible AP adapters
3210  * @aqm: the mask identifying the accessible AP domains
3211  * @adm: the mask identifying the accessible AP control domains
3212  *
3213  * Set the masks that identify the adapters, domains and control domains to
3214  * which the KVM guest is granted access.
3215  *
3216  * Note: The kvm->lock mutex must be locked by the caller before invoking this
3217  *	 function.
3218  */
3219 void kvm_arch_crypto_set_masks(struct kvm *kvm, unsigned long *apm,
3220 			       unsigned long *aqm, unsigned long *adm)
3221 {
3222 	struct kvm_s390_crypto_cb *crycb = kvm->arch.crypto.crycb;
3223 
3224 	kvm_s390_vcpu_block_all(kvm);
3225 
3226 	switch (kvm->arch.crypto.crycbd & CRYCB_FORMAT_MASK) {
3227 	case CRYCB_FORMAT2: /* APCB1 use 256 bits */
3228 		memcpy(crycb->apcb1.apm, apm, 32);
3229 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx %016lx %016lx %016lx",
3230 			 apm[0], apm[1], apm[2], apm[3]);
3231 		memcpy(crycb->apcb1.aqm, aqm, 32);
3232 		VM_EVENT(kvm, 3, "SET CRYCB: aqm %016lx %016lx %016lx %016lx",
3233 			 aqm[0], aqm[1], aqm[2], aqm[3]);
3234 		memcpy(crycb->apcb1.adm, adm, 32);
3235 		VM_EVENT(kvm, 3, "SET CRYCB: adm %016lx %016lx %016lx %016lx",
3236 			 adm[0], adm[1], adm[2], adm[3]);
3237 		break;
3238 	case CRYCB_FORMAT1:
3239 	case CRYCB_FORMAT0: /* Fall through both use APCB0 */
3240 		memcpy(crycb->apcb0.apm, apm, 8);
3241 		memcpy(crycb->apcb0.aqm, aqm, 2);
3242 		memcpy(crycb->apcb0.adm, adm, 2);
3243 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx aqm %04x adm %04x",
3244 			 apm[0], *((unsigned short *)aqm),
3245 			 *((unsigned short *)adm));
3246 		break;
3247 	default:	/* Can not happen */
3248 		break;
3249 	}
3250 
3251 	/* recreate the shadow crycb for each vcpu */
3252 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
3253 	kvm_s390_vcpu_unblock_all(kvm);
3254 }
3255 EXPORT_SYMBOL_GPL(kvm_arch_crypto_set_masks);
3256 
3257 /*
3258  * kvm_arch_crypto_clear_masks
3259  *
3260  * @kvm: pointer to the target guest's KVM struct containing the crypto masks
3261  *	 to be cleared.
3262  *
3263  * Clear the masks that identify the adapters, domains and control domains to
3264  * which the KVM guest is granted access.
3265  *
3266  * Note: The kvm->lock mutex must be locked by the caller before invoking this
3267  *	 function.
3268  */
3269 void kvm_arch_crypto_clear_masks(struct kvm *kvm)
3270 {
3271 	kvm_s390_vcpu_block_all(kvm);
3272 
3273 	memset(&kvm->arch.crypto.crycb->apcb0, 0,
3274 	       sizeof(kvm->arch.crypto.crycb->apcb0));
3275 	memset(&kvm->arch.crypto.crycb->apcb1, 0,
3276 	       sizeof(kvm->arch.crypto.crycb->apcb1));
3277 
3278 	VM_EVENT(kvm, 3, "%s", "CLR CRYCB:");
3279 	/* recreate the shadow crycb for each vcpu */
3280 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
3281 	kvm_s390_vcpu_unblock_all(kvm);
3282 }
3283 EXPORT_SYMBOL_GPL(kvm_arch_crypto_clear_masks);
3284 
3285 static u64 kvm_s390_get_initial_cpuid(void)
3286 {
3287 	struct cpuid cpuid;
3288 
3289 	get_cpu_id(&cpuid);
3290 	cpuid.version = 0xff;
3291 	return *((u64 *) &cpuid);
3292 }
3293 
3294 static void kvm_s390_crypto_init(struct kvm *kvm)
3295 {
3296 	kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
3297 	kvm_s390_set_crycb_format(kvm);
3298 	init_rwsem(&kvm->arch.crypto.pqap_hook_rwsem);
3299 
3300 	if (!test_kvm_facility(kvm, 76))
3301 		return;
3302 
3303 	/* Enable AES/DEA protected key functions by default */
3304 	kvm->arch.crypto.aes_kw = 1;
3305 	kvm->arch.crypto.dea_kw = 1;
3306 	get_random_bytes(kvm->arch.crypto.crycb->aes_wrapping_key_mask,
3307 			 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
3308 	get_random_bytes(kvm->arch.crypto.crycb->dea_wrapping_key_mask,
3309 			 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
3310 }
3311 
3312 static void sca_dispose(struct kvm *kvm)
3313 {
3314 	if (kvm->arch.use_esca)
3315 		free_pages_exact(kvm->arch.sca, sizeof(struct esca_block));
3316 	else
3317 		free_page((unsigned long)(kvm->arch.sca));
3318 	kvm->arch.sca = NULL;
3319 }
3320 
3321 void kvm_arch_free_vm(struct kvm *kvm)
3322 {
3323 	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM))
3324 		kvm_s390_pci_clear_list(kvm);
3325 
3326 	__kvm_arch_free_vm(kvm);
3327 }
3328 
3329 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
3330 {
3331 	gfp_t alloc_flags = GFP_KERNEL_ACCOUNT;
3332 	int i, rc;
3333 	char debug_name[16];
3334 	static unsigned long sca_offset;
3335 
3336 	rc = -EINVAL;
3337 #ifdef CONFIG_KVM_S390_UCONTROL
3338 	if (type & ~KVM_VM_S390_UCONTROL)
3339 		goto out_err;
3340 	if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN)))
3341 		goto out_err;
3342 #else
3343 	if (type)
3344 		goto out_err;
3345 #endif
3346 
3347 	rc = s390_enable_sie();
3348 	if (rc)
3349 		goto out_err;
3350 
3351 	rc = -ENOMEM;
3352 
3353 	if (!sclp.has_64bscao)
3354 		alloc_flags |= GFP_DMA;
3355 	rwlock_init(&kvm->arch.sca_lock);
3356 	/* start with basic SCA */
3357 	kvm->arch.sca = (struct bsca_block *) get_zeroed_page(alloc_flags);
3358 	if (!kvm->arch.sca)
3359 		goto out_err;
3360 	mutex_lock(&kvm_lock);
3361 	sca_offset += 16;
3362 	if (sca_offset + sizeof(struct bsca_block) > PAGE_SIZE)
3363 		sca_offset = 0;
3364 	kvm->arch.sca = (struct bsca_block *)
3365 			((char *) kvm->arch.sca + sca_offset);
3366 	mutex_unlock(&kvm_lock);
3367 
3368 	sprintf(debug_name, "kvm-%u", current->pid);
3369 
3370 	kvm->arch.dbf = debug_register(debug_name, 32, 1, 7 * sizeof(long));
3371 	if (!kvm->arch.dbf)
3372 		goto out_err;
3373 
3374 	BUILD_BUG_ON(sizeof(struct sie_page2) != 4096);
3375 	kvm->arch.sie_page2 =
3376 	     (struct sie_page2 *) get_zeroed_page(GFP_KERNEL_ACCOUNT | GFP_DMA);
3377 	if (!kvm->arch.sie_page2)
3378 		goto out_err;
3379 
3380 	kvm->arch.sie_page2->kvm = kvm;
3381 	kvm->arch.model.fac_list = kvm->arch.sie_page2->fac_list;
3382 
3383 	for (i = 0; i < kvm_s390_fac_size(); i++) {
3384 		kvm->arch.model.fac_mask[i] = stfle_fac_list[i] &
3385 					      (kvm_s390_fac_base[i] |
3386 					       kvm_s390_fac_ext[i]);
3387 		kvm->arch.model.fac_list[i] = stfle_fac_list[i] &
3388 					      kvm_s390_fac_base[i];
3389 	}
3390 	kvm->arch.model.subfuncs = kvm_s390_available_subfunc;
3391 
3392 	/* we are always in czam mode - even on pre z14 machines */
3393 	set_kvm_facility(kvm->arch.model.fac_mask, 138);
3394 	set_kvm_facility(kvm->arch.model.fac_list, 138);
3395 	/* we emulate STHYI in kvm */
3396 	set_kvm_facility(kvm->arch.model.fac_mask, 74);
3397 	set_kvm_facility(kvm->arch.model.fac_list, 74);
3398 	if (MACHINE_HAS_TLB_GUEST) {
3399 		set_kvm_facility(kvm->arch.model.fac_mask, 147);
3400 		set_kvm_facility(kvm->arch.model.fac_list, 147);
3401 	}
3402 
3403 	if (css_general_characteristics.aiv && test_facility(65))
3404 		set_kvm_facility(kvm->arch.model.fac_mask, 65);
3405 
3406 	kvm->arch.model.cpuid = kvm_s390_get_initial_cpuid();
3407 	kvm->arch.model.ibc = sclp.ibc & 0x0fff;
3408 
3409 	kvm->arch.model.uv_feat_guest.feat = 0;
3410 
3411 	kvm_s390_crypto_init(kvm);
3412 
3413 	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM)) {
3414 		mutex_lock(&kvm->lock);
3415 		kvm_s390_pci_init_list(kvm);
3416 		kvm_s390_vcpu_pci_enable_interp(kvm);
3417 		mutex_unlock(&kvm->lock);
3418 	}
3419 
3420 	mutex_init(&kvm->arch.float_int.ais_lock);
3421 	spin_lock_init(&kvm->arch.float_int.lock);
3422 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
3423 		INIT_LIST_HEAD(&kvm->arch.float_int.lists[i]);
3424 	init_waitqueue_head(&kvm->arch.ipte_wq);
3425 	mutex_init(&kvm->arch.ipte_mutex);
3426 
3427 	debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
3428 	VM_EVENT(kvm, 3, "vm created with type %lu", type);
3429 
3430 	if (type & KVM_VM_S390_UCONTROL) {
3431 		kvm->arch.gmap = NULL;
3432 		kvm->arch.mem_limit = KVM_S390_NO_MEM_LIMIT;
3433 	} else {
3434 		if (sclp.hamax == U64_MAX)
3435 			kvm->arch.mem_limit = TASK_SIZE_MAX;
3436 		else
3437 			kvm->arch.mem_limit = min_t(unsigned long, TASK_SIZE_MAX,
3438 						    sclp.hamax + 1);
3439 		kvm->arch.gmap = gmap_create(current->mm, kvm->arch.mem_limit - 1);
3440 		if (!kvm->arch.gmap)
3441 			goto out_err;
3442 		kvm->arch.gmap->private = kvm;
3443 		kvm->arch.gmap->pfault_enabled = 0;
3444 	}
3445 
3446 	kvm->arch.use_pfmfi = sclp.has_pfmfi;
3447 	kvm->arch.use_skf = sclp.has_skey;
3448 	spin_lock_init(&kvm->arch.start_stop_lock);
3449 	kvm_s390_vsie_init(kvm);
3450 	if (use_gisa)
3451 		kvm_s390_gisa_init(kvm);
3452 	INIT_LIST_HEAD(&kvm->arch.pv.need_cleanup);
3453 	kvm->arch.pv.set_aside = NULL;
3454 	KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm, current->pid);
3455 
3456 	return 0;
3457 out_err:
3458 	free_page((unsigned long)kvm->arch.sie_page2);
3459 	debug_unregister(kvm->arch.dbf);
3460 	sca_dispose(kvm);
3461 	KVM_EVENT(3, "creation of vm failed: %d", rc);
3462 	return rc;
3463 }
3464 
3465 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
3466 {
3467 	u16 rc, rrc;
3468 
3469 	VCPU_EVENT(vcpu, 3, "%s", "free cpu");
3470 	trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id);
3471 	kvm_s390_clear_local_irqs(vcpu);
3472 	kvm_clear_async_pf_completion_queue(vcpu);
3473 	if (!kvm_is_ucontrol(vcpu->kvm))
3474 		sca_del_vcpu(vcpu);
3475 	kvm_s390_update_topology_change_report(vcpu->kvm, 1);
3476 
3477 	if (kvm_is_ucontrol(vcpu->kvm))
3478 		gmap_remove(vcpu->arch.gmap);
3479 
3480 	if (vcpu->kvm->arch.use_cmma)
3481 		kvm_s390_vcpu_unsetup_cmma(vcpu);
3482 	/* We can not hold the vcpu mutex here, we are already dying */
3483 	if (kvm_s390_pv_cpu_get_handle(vcpu))
3484 		kvm_s390_pv_destroy_cpu(vcpu, &rc, &rrc);
3485 	free_page((unsigned long)(vcpu->arch.sie_block));
3486 }
3487 
3488 void kvm_arch_destroy_vm(struct kvm *kvm)
3489 {
3490 	u16 rc, rrc;
3491 
3492 	kvm_destroy_vcpus(kvm);
3493 	sca_dispose(kvm);
3494 	kvm_s390_gisa_destroy(kvm);
3495 	/*
3496 	 * We are already at the end of life and kvm->lock is not taken.
3497 	 * This is ok as the file descriptor is closed by now and nobody
3498 	 * can mess with the pv state.
3499 	 */
3500 	kvm_s390_pv_deinit_cleanup_all(kvm, &rc, &rrc);
3501 	/*
3502 	 * Remove the mmu notifier only when the whole KVM VM is torn down,
3503 	 * and only if one was registered to begin with. If the VM is
3504 	 * currently not protected, but has been previously been protected,
3505 	 * then it's possible that the notifier is still registered.
3506 	 */
3507 	if (kvm->arch.pv.mmu_notifier.ops)
3508 		mmu_notifier_unregister(&kvm->arch.pv.mmu_notifier, kvm->mm);
3509 
3510 	debug_unregister(kvm->arch.dbf);
3511 	free_page((unsigned long)kvm->arch.sie_page2);
3512 	if (!kvm_is_ucontrol(kvm))
3513 		gmap_remove(kvm->arch.gmap);
3514 	kvm_s390_destroy_adapters(kvm);
3515 	kvm_s390_clear_float_irqs(kvm);
3516 	kvm_s390_vsie_destroy(kvm);
3517 	KVM_EVENT(3, "vm 0x%pK destroyed", kvm);
3518 }
3519 
3520 /* Section: vcpu related */
3521 static int __kvm_ucontrol_vcpu_init(struct kvm_vcpu *vcpu)
3522 {
3523 	vcpu->arch.gmap = gmap_create(current->mm, -1UL);
3524 	if (!vcpu->arch.gmap)
3525 		return -ENOMEM;
3526 	vcpu->arch.gmap->private = vcpu->kvm;
3527 
3528 	return 0;
3529 }
3530 
3531 static void sca_del_vcpu(struct kvm_vcpu *vcpu)
3532 {
3533 	if (!kvm_s390_use_sca_entries())
3534 		return;
3535 	read_lock(&vcpu->kvm->arch.sca_lock);
3536 	if (vcpu->kvm->arch.use_esca) {
3537 		struct esca_block *sca = vcpu->kvm->arch.sca;
3538 
3539 		clear_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
3540 		sca->cpu[vcpu->vcpu_id].sda = 0;
3541 	} else {
3542 		struct bsca_block *sca = vcpu->kvm->arch.sca;
3543 
3544 		clear_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
3545 		sca->cpu[vcpu->vcpu_id].sda = 0;
3546 	}
3547 	read_unlock(&vcpu->kvm->arch.sca_lock);
3548 }
3549 
3550 static void sca_add_vcpu(struct kvm_vcpu *vcpu)
3551 {
3552 	if (!kvm_s390_use_sca_entries()) {
3553 		phys_addr_t sca_phys = virt_to_phys(vcpu->kvm->arch.sca);
3554 
3555 		/* we still need the basic sca for the ipte control */
3556 		vcpu->arch.sie_block->scaoh = sca_phys >> 32;
3557 		vcpu->arch.sie_block->scaol = sca_phys;
3558 		return;
3559 	}
3560 	read_lock(&vcpu->kvm->arch.sca_lock);
3561 	if (vcpu->kvm->arch.use_esca) {
3562 		struct esca_block *sca = vcpu->kvm->arch.sca;
3563 		phys_addr_t sca_phys = virt_to_phys(sca);
3564 
3565 		sca->cpu[vcpu->vcpu_id].sda = virt_to_phys(vcpu->arch.sie_block);
3566 		vcpu->arch.sie_block->scaoh = sca_phys >> 32;
3567 		vcpu->arch.sie_block->scaol = sca_phys & ESCA_SCAOL_MASK;
3568 		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
3569 		set_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
3570 	} else {
3571 		struct bsca_block *sca = vcpu->kvm->arch.sca;
3572 		phys_addr_t sca_phys = virt_to_phys(sca);
3573 
3574 		sca->cpu[vcpu->vcpu_id].sda = virt_to_phys(vcpu->arch.sie_block);
3575 		vcpu->arch.sie_block->scaoh = sca_phys >> 32;
3576 		vcpu->arch.sie_block->scaol = sca_phys;
3577 		set_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
3578 	}
3579 	read_unlock(&vcpu->kvm->arch.sca_lock);
3580 }
3581 
3582 /* Basic SCA to Extended SCA data copy routines */
3583 static inline void sca_copy_entry(struct esca_entry *d, struct bsca_entry *s)
3584 {
3585 	d->sda = s->sda;
3586 	d->sigp_ctrl.c = s->sigp_ctrl.c;
3587 	d->sigp_ctrl.scn = s->sigp_ctrl.scn;
3588 }
3589 
3590 static void sca_copy_b_to_e(struct esca_block *d, struct bsca_block *s)
3591 {
3592 	int i;
3593 
3594 	d->ipte_control = s->ipte_control;
3595 	d->mcn[0] = s->mcn;
3596 	for (i = 0; i < KVM_S390_BSCA_CPU_SLOTS; i++)
3597 		sca_copy_entry(&d->cpu[i], &s->cpu[i]);
3598 }
3599 
3600 static int sca_switch_to_extended(struct kvm *kvm)
3601 {
3602 	struct bsca_block *old_sca = kvm->arch.sca;
3603 	struct esca_block *new_sca;
3604 	struct kvm_vcpu *vcpu;
3605 	unsigned long vcpu_idx;
3606 	u32 scaol, scaoh;
3607 	phys_addr_t new_sca_phys;
3608 
3609 	if (kvm->arch.use_esca)
3610 		return 0;
3611 
3612 	new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL_ACCOUNT | __GFP_ZERO);
3613 	if (!new_sca)
3614 		return -ENOMEM;
3615 
3616 	new_sca_phys = virt_to_phys(new_sca);
3617 	scaoh = new_sca_phys >> 32;
3618 	scaol = new_sca_phys & ESCA_SCAOL_MASK;
3619 
3620 	kvm_s390_vcpu_block_all(kvm);
3621 	write_lock(&kvm->arch.sca_lock);
3622 
3623 	sca_copy_b_to_e(new_sca, old_sca);
3624 
3625 	kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
3626 		vcpu->arch.sie_block->scaoh = scaoh;
3627 		vcpu->arch.sie_block->scaol = scaol;
3628 		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
3629 	}
3630 	kvm->arch.sca = new_sca;
3631 	kvm->arch.use_esca = 1;
3632 
3633 	write_unlock(&kvm->arch.sca_lock);
3634 	kvm_s390_vcpu_unblock_all(kvm);
3635 
3636 	free_page((unsigned long)old_sca);
3637 
3638 	VM_EVENT(kvm, 2, "Switched to ESCA (0x%pK -> 0x%pK)",
3639 		 old_sca, kvm->arch.sca);
3640 	return 0;
3641 }
3642 
3643 static int sca_can_add_vcpu(struct kvm *kvm, unsigned int id)
3644 {
3645 	int rc;
3646 
3647 	if (!kvm_s390_use_sca_entries()) {
3648 		if (id < KVM_MAX_VCPUS)
3649 			return true;
3650 		return false;
3651 	}
3652 	if (id < KVM_S390_BSCA_CPU_SLOTS)
3653 		return true;
3654 	if (!sclp.has_esca || !sclp.has_64bscao)
3655 		return false;
3656 
3657 	rc = kvm->arch.use_esca ? 0 : sca_switch_to_extended(kvm);
3658 
3659 	return rc == 0 && id < KVM_S390_ESCA_CPU_SLOTS;
3660 }
3661 
3662 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3663 static void __start_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3664 {
3665 	WARN_ON_ONCE(vcpu->arch.cputm_start != 0);
3666 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3667 	vcpu->arch.cputm_start = get_tod_clock_fast();
3668 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3669 }
3670 
3671 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3672 static void __stop_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3673 {
3674 	WARN_ON_ONCE(vcpu->arch.cputm_start == 0);
3675 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3676 	vcpu->arch.sie_block->cputm -= get_tod_clock_fast() - vcpu->arch.cputm_start;
3677 	vcpu->arch.cputm_start = 0;
3678 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3679 }
3680 
3681 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3682 static void __enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3683 {
3684 	WARN_ON_ONCE(vcpu->arch.cputm_enabled);
3685 	vcpu->arch.cputm_enabled = true;
3686 	__start_cpu_timer_accounting(vcpu);
3687 }
3688 
3689 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
3690 static void __disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3691 {
3692 	WARN_ON_ONCE(!vcpu->arch.cputm_enabled);
3693 	__stop_cpu_timer_accounting(vcpu);
3694 	vcpu->arch.cputm_enabled = false;
3695 }
3696 
3697 static void enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3698 {
3699 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3700 	__enable_cpu_timer_accounting(vcpu);
3701 	preempt_enable();
3702 }
3703 
3704 static void disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
3705 {
3706 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3707 	__disable_cpu_timer_accounting(vcpu);
3708 	preempt_enable();
3709 }
3710 
3711 /* set the cpu timer - may only be called from the VCPU thread itself */
3712 void kvm_s390_set_cpu_timer(struct kvm_vcpu *vcpu, __u64 cputm)
3713 {
3714 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3715 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
3716 	if (vcpu->arch.cputm_enabled)
3717 		vcpu->arch.cputm_start = get_tod_clock_fast();
3718 	vcpu->arch.sie_block->cputm = cputm;
3719 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
3720 	preempt_enable();
3721 }
3722 
3723 /* update and get the cpu timer - can also be called from other VCPU threads */
3724 __u64 kvm_s390_get_cpu_timer(struct kvm_vcpu *vcpu)
3725 {
3726 	unsigned int seq;
3727 	__u64 value;
3728 
3729 	if (unlikely(!vcpu->arch.cputm_enabled))
3730 		return vcpu->arch.sie_block->cputm;
3731 
3732 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
3733 	do {
3734 		seq = raw_read_seqcount(&vcpu->arch.cputm_seqcount);
3735 		/*
3736 		 * If the writer would ever execute a read in the critical
3737 		 * section, e.g. in irq context, we have a deadlock.
3738 		 */
3739 		WARN_ON_ONCE((seq & 1) && smp_processor_id() == vcpu->cpu);
3740 		value = vcpu->arch.sie_block->cputm;
3741 		/* if cputm_start is 0, accounting is being started/stopped */
3742 		if (likely(vcpu->arch.cputm_start))
3743 			value -= get_tod_clock_fast() - vcpu->arch.cputm_start;
3744 	} while (read_seqcount_retry(&vcpu->arch.cputm_seqcount, seq & ~1));
3745 	preempt_enable();
3746 	return value;
3747 }
3748 
3749 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
3750 {
3751 
3752 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_RUNNING);
3753 	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
3754 		__start_cpu_timer_accounting(vcpu);
3755 	vcpu->cpu = cpu;
3756 }
3757 
3758 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
3759 {
3760 	vcpu->cpu = -1;
3761 	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
3762 		__stop_cpu_timer_accounting(vcpu);
3763 	kvm_s390_clear_cpuflags(vcpu, CPUSTAT_RUNNING);
3764 
3765 }
3766 
3767 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
3768 {
3769 	mutex_lock(&vcpu->kvm->lock);
3770 	preempt_disable();
3771 	vcpu->arch.sie_block->epoch = vcpu->kvm->arch.epoch;
3772 	vcpu->arch.sie_block->epdx = vcpu->kvm->arch.epdx;
3773 	preempt_enable();
3774 	mutex_unlock(&vcpu->kvm->lock);
3775 	if (!kvm_is_ucontrol(vcpu->kvm)) {
3776 		vcpu->arch.gmap = vcpu->kvm->arch.gmap;
3777 		sca_add_vcpu(vcpu);
3778 	}
3779 	if (test_kvm_facility(vcpu->kvm, 74) || vcpu->kvm->arch.user_instr0)
3780 		vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
3781 }
3782 
3783 static bool kvm_has_pckmo_subfunc(struct kvm *kvm, unsigned long nr)
3784 {
3785 	if (test_bit_inv(nr, (unsigned long *)&kvm->arch.model.subfuncs.pckmo) &&
3786 	    test_bit_inv(nr, (unsigned long *)&kvm_s390_available_subfunc.pckmo))
3787 		return true;
3788 	return false;
3789 }
3790 
3791 static bool kvm_has_pckmo_ecc(struct kvm *kvm)
3792 {
3793 	/* At least one ECC subfunction must be present */
3794 	return kvm_has_pckmo_subfunc(kvm, 32) ||
3795 	       kvm_has_pckmo_subfunc(kvm, 33) ||
3796 	       kvm_has_pckmo_subfunc(kvm, 34) ||
3797 	       kvm_has_pckmo_subfunc(kvm, 40) ||
3798 	       kvm_has_pckmo_subfunc(kvm, 41);
3799 
3800 }
3801 
3802 static bool kvm_has_pckmo_hmac(struct kvm *kvm)
3803 {
3804 	/* At least one HMAC subfunction must be present */
3805 	return kvm_has_pckmo_subfunc(kvm, 118) ||
3806 	       kvm_has_pckmo_subfunc(kvm, 122);
3807 }
3808 
3809 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu)
3810 {
3811 	/*
3812 	 * If the AP instructions are not being interpreted and the MSAX3
3813 	 * facility is not configured for the guest, there is nothing to set up.
3814 	 */
3815 	if (!vcpu->kvm->arch.crypto.apie && !test_kvm_facility(vcpu->kvm, 76))
3816 		return;
3817 
3818 	vcpu->arch.sie_block->crycbd = vcpu->kvm->arch.crypto.crycbd;
3819 	vcpu->arch.sie_block->ecb3 &= ~(ECB3_AES | ECB3_DEA);
3820 	vcpu->arch.sie_block->eca &= ~ECA_APIE;
3821 	vcpu->arch.sie_block->ecd &= ~(ECD_ECC | ECD_HMAC);
3822 
3823 	if (vcpu->kvm->arch.crypto.apie)
3824 		vcpu->arch.sie_block->eca |= ECA_APIE;
3825 
3826 	/* Set up protected key support */
3827 	if (vcpu->kvm->arch.crypto.aes_kw) {
3828 		vcpu->arch.sie_block->ecb3 |= ECB3_AES;
3829 		/* ecc/hmac is also wrapped with AES key */
3830 		if (kvm_has_pckmo_ecc(vcpu->kvm))
3831 			vcpu->arch.sie_block->ecd |= ECD_ECC;
3832 		if (kvm_has_pckmo_hmac(vcpu->kvm))
3833 			vcpu->arch.sie_block->ecd |= ECD_HMAC;
3834 	}
3835 
3836 	if (vcpu->kvm->arch.crypto.dea_kw)
3837 		vcpu->arch.sie_block->ecb3 |= ECB3_DEA;
3838 }
3839 
3840 void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu)
3841 {
3842 	free_page((unsigned long)phys_to_virt(vcpu->arch.sie_block->cbrlo));
3843 	vcpu->arch.sie_block->cbrlo = 0;
3844 }
3845 
3846 int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu)
3847 {
3848 	void *cbrlo_page = (void *)get_zeroed_page(GFP_KERNEL_ACCOUNT);
3849 
3850 	if (!cbrlo_page)
3851 		return -ENOMEM;
3852 
3853 	vcpu->arch.sie_block->cbrlo = virt_to_phys(cbrlo_page);
3854 	return 0;
3855 }
3856 
3857 static void kvm_s390_vcpu_setup_model(struct kvm_vcpu *vcpu)
3858 {
3859 	struct kvm_s390_cpu_model *model = &vcpu->kvm->arch.model;
3860 
3861 	vcpu->arch.sie_block->ibc = model->ibc;
3862 	if (test_kvm_facility(vcpu->kvm, 7))
3863 		vcpu->arch.sie_block->fac = virt_to_phys(model->fac_list);
3864 }
3865 
3866 static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
3867 {
3868 	int rc = 0;
3869 	u16 uvrc, uvrrc;
3870 
3871 	atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH |
3872 						    CPUSTAT_SM |
3873 						    CPUSTAT_STOPPED);
3874 
3875 	if (test_kvm_facility(vcpu->kvm, 78))
3876 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED2);
3877 	else if (test_kvm_facility(vcpu->kvm, 8))
3878 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED);
3879 
3880 	kvm_s390_vcpu_setup_model(vcpu);
3881 
3882 	/* pgste_set_pte has special handling for !MACHINE_HAS_ESOP */
3883 	if (MACHINE_HAS_ESOP)
3884 		vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
3885 	if (test_kvm_facility(vcpu->kvm, 9))
3886 		vcpu->arch.sie_block->ecb |= ECB_SRSI;
3887 	if (test_kvm_facility(vcpu->kvm, 11))
3888 		vcpu->arch.sie_block->ecb |= ECB_PTF;
3889 	if (test_kvm_facility(vcpu->kvm, 73))
3890 		vcpu->arch.sie_block->ecb |= ECB_TE;
3891 	if (!kvm_is_ucontrol(vcpu->kvm))
3892 		vcpu->arch.sie_block->ecb |= ECB_SPECI;
3893 
3894 	if (test_kvm_facility(vcpu->kvm, 8) && vcpu->kvm->arch.use_pfmfi)
3895 		vcpu->arch.sie_block->ecb2 |= ECB2_PFMFI;
3896 	if (test_kvm_facility(vcpu->kvm, 130))
3897 		vcpu->arch.sie_block->ecb2 |= ECB2_IEP;
3898 	vcpu->arch.sie_block->eca = ECA_MVPGI | ECA_PROTEXCI;
3899 	if (sclp.has_cei)
3900 		vcpu->arch.sie_block->eca |= ECA_CEI;
3901 	if (sclp.has_ib)
3902 		vcpu->arch.sie_block->eca |= ECA_IB;
3903 	if (sclp.has_siif)
3904 		vcpu->arch.sie_block->eca |= ECA_SII;
3905 	if (sclp.has_sigpif)
3906 		vcpu->arch.sie_block->eca |= ECA_SIGPI;
3907 	if (test_kvm_facility(vcpu->kvm, 129)) {
3908 		vcpu->arch.sie_block->eca |= ECA_VX;
3909 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
3910 	}
3911 	if (test_kvm_facility(vcpu->kvm, 139))
3912 		vcpu->arch.sie_block->ecd |= ECD_MEF;
3913 	if (test_kvm_facility(vcpu->kvm, 156))
3914 		vcpu->arch.sie_block->ecd |= ECD_ETOKENF;
3915 	if (vcpu->arch.sie_block->gd) {
3916 		vcpu->arch.sie_block->eca |= ECA_AIV;
3917 		VCPU_EVENT(vcpu, 3, "AIV gisa format-%u enabled for cpu %03u",
3918 			   vcpu->arch.sie_block->gd & 0x3, vcpu->vcpu_id);
3919 	}
3920 	vcpu->arch.sie_block->sdnxo = virt_to_phys(&vcpu->run->s.regs.sdnx) | SDNXC;
3921 	vcpu->arch.sie_block->riccbd = virt_to_phys(&vcpu->run->s.regs.riccb);
3922 
3923 	if (sclp.has_kss)
3924 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_KSS);
3925 	else
3926 		vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
3927 
3928 	if (vcpu->kvm->arch.use_cmma) {
3929 		rc = kvm_s390_vcpu_setup_cmma(vcpu);
3930 		if (rc)
3931 			return rc;
3932 	}
3933 	hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
3934 	vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup;
3935 
3936 	vcpu->arch.sie_block->hpid = HPID_KVM;
3937 
3938 	kvm_s390_vcpu_crypto_setup(vcpu);
3939 
3940 	kvm_s390_vcpu_pci_setup(vcpu);
3941 
3942 	mutex_lock(&vcpu->kvm->lock);
3943 	if (kvm_s390_pv_is_protected(vcpu->kvm)) {
3944 		rc = kvm_s390_pv_create_cpu(vcpu, &uvrc, &uvrrc);
3945 		if (rc)
3946 			kvm_s390_vcpu_unsetup_cmma(vcpu);
3947 	}
3948 	mutex_unlock(&vcpu->kvm->lock);
3949 
3950 	return rc;
3951 }
3952 
3953 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
3954 {
3955 	if (!kvm_is_ucontrol(kvm) && !sca_can_add_vcpu(kvm, id))
3956 		return -EINVAL;
3957 	return 0;
3958 }
3959 
3960 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
3961 {
3962 	struct sie_page *sie_page;
3963 	int rc;
3964 
3965 	BUILD_BUG_ON(sizeof(struct sie_page) != 4096);
3966 	sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL_ACCOUNT);
3967 	if (!sie_page)
3968 		return -ENOMEM;
3969 
3970 	vcpu->arch.sie_block = &sie_page->sie_block;
3971 	vcpu->arch.sie_block->itdba = virt_to_phys(&sie_page->itdb);
3972 
3973 	/* the real guest size will always be smaller than msl */
3974 	vcpu->arch.sie_block->mso = 0;
3975 	vcpu->arch.sie_block->msl = sclp.hamax;
3976 
3977 	vcpu->arch.sie_block->icpua = vcpu->vcpu_id;
3978 	spin_lock_init(&vcpu->arch.local_int.lock);
3979 	vcpu->arch.sie_block->gd = kvm_s390_get_gisa_desc(vcpu->kvm);
3980 	seqcount_init(&vcpu->arch.cputm_seqcount);
3981 
3982 	vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
3983 	kvm_clear_async_pf_completion_queue(vcpu);
3984 	vcpu->run->kvm_valid_regs = KVM_SYNC_PREFIX |
3985 				    KVM_SYNC_GPRS |
3986 				    KVM_SYNC_ACRS |
3987 				    KVM_SYNC_CRS |
3988 				    KVM_SYNC_ARCH0 |
3989 				    KVM_SYNC_PFAULT |
3990 				    KVM_SYNC_DIAG318;
3991 	vcpu->arch.acrs_loaded = false;
3992 	kvm_s390_set_prefix(vcpu, 0);
3993 	if (test_kvm_facility(vcpu->kvm, 64))
3994 		vcpu->run->kvm_valid_regs |= KVM_SYNC_RICCB;
3995 	if (test_kvm_facility(vcpu->kvm, 82))
3996 		vcpu->run->kvm_valid_regs |= KVM_SYNC_BPBC;
3997 	if (test_kvm_facility(vcpu->kvm, 133))
3998 		vcpu->run->kvm_valid_regs |= KVM_SYNC_GSCB;
3999 	if (test_kvm_facility(vcpu->kvm, 156))
4000 		vcpu->run->kvm_valid_regs |= KVM_SYNC_ETOKEN;
4001 	/* fprs can be synchronized via vrs, even if the guest has no vx. With
4002 	 * cpu_has_vx(), (load|store)_fpu_regs() will work with vrs format.
4003 	 */
4004 	if (cpu_has_vx())
4005 		vcpu->run->kvm_valid_regs |= KVM_SYNC_VRS;
4006 	else
4007 		vcpu->run->kvm_valid_regs |= KVM_SYNC_FPRS;
4008 
4009 	if (kvm_is_ucontrol(vcpu->kvm)) {
4010 		rc = __kvm_ucontrol_vcpu_init(vcpu);
4011 		if (rc)
4012 			goto out_free_sie_block;
4013 	}
4014 
4015 	VM_EVENT(vcpu->kvm, 3, "create cpu %d at 0x%pK, sie block at 0x%pK",
4016 		 vcpu->vcpu_id, vcpu, vcpu->arch.sie_block);
4017 	trace_kvm_s390_create_vcpu(vcpu->vcpu_id, vcpu, vcpu->arch.sie_block);
4018 
4019 	rc = kvm_s390_vcpu_setup(vcpu);
4020 	if (rc)
4021 		goto out_ucontrol_uninit;
4022 
4023 	kvm_s390_update_topology_change_report(vcpu->kvm, 1);
4024 	return 0;
4025 
4026 out_ucontrol_uninit:
4027 	if (kvm_is_ucontrol(vcpu->kvm))
4028 		gmap_remove(vcpu->arch.gmap);
4029 out_free_sie_block:
4030 	free_page((unsigned long)(vcpu->arch.sie_block));
4031 	return rc;
4032 }
4033 
4034 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
4035 {
4036 	clear_bit(vcpu->vcpu_idx, vcpu->kvm->arch.gisa_int.kicked_mask);
4037 	return kvm_s390_vcpu_has_irq(vcpu, 0);
4038 }
4039 
4040 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu)
4041 {
4042 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE);
4043 }
4044 
4045 void kvm_s390_vcpu_block(struct kvm_vcpu *vcpu)
4046 {
4047 	atomic_or(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
4048 	exit_sie(vcpu);
4049 }
4050 
4051 void kvm_s390_vcpu_unblock(struct kvm_vcpu *vcpu)
4052 {
4053 	atomic_andnot(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
4054 }
4055 
4056 static void kvm_s390_vcpu_request(struct kvm_vcpu *vcpu)
4057 {
4058 	atomic_or(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
4059 	exit_sie(vcpu);
4060 }
4061 
4062 bool kvm_s390_vcpu_sie_inhibited(struct kvm_vcpu *vcpu)
4063 {
4064 	return atomic_read(&vcpu->arch.sie_block->prog20) &
4065 	       (PROG_BLOCK_SIE | PROG_REQUEST);
4066 }
4067 
4068 static void kvm_s390_vcpu_request_handled(struct kvm_vcpu *vcpu)
4069 {
4070 	atomic_andnot(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
4071 }
4072 
4073 /*
4074  * Kick a guest cpu out of (v)SIE and wait until (v)SIE is not running.
4075  * If the CPU is not running (e.g. waiting as idle) the function will
4076  * return immediately. */
4077 void exit_sie(struct kvm_vcpu *vcpu)
4078 {
4079 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOP_INT);
4080 	kvm_s390_vsie_kick(vcpu);
4081 	while (vcpu->arch.sie_block->prog0c & PROG_IN_SIE)
4082 		cpu_relax();
4083 }
4084 
4085 /* Kick a guest cpu out of SIE to process a request synchronously */
4086 void kvm_s390_sync_request(int req, struct kvm_vcpu *vcpu)
4087 {
4088 	__kvm_make_request(req, vcpu);
4089 	kvm_s390_vcpu_request(vcpu);
4090 }
4091 
4092 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
4093 			      unsigned long end)
4094 {
4095 	struct kvm *kvm = gmap->private;
4096 	struct kvm_vcpu *vcpu;
4097 	unsigned long prefix;
4098 	unsigned long i;
4099 
4100 	trace_kvm_s390_gmap_notifier(start, end, gmap_is_shadow(gmap));
4101 
4102 	if (gmap_is_shadow(gmap))
4103 		return;
4104 	if (start >= 1UL << 31)
4105 		/* We are only interested in prefix pages */
4106 		return;
4107 	kvm_for_each_vcpu(i, vcpu, kvm) {
4108 		/* match against both prefix pages */
4109 		prefix = kvm_s390_get_prefix(vcpu);
4110 		if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
4111 			VCPU_EVENT(vcpu, 2, "gmap notifier for %lx-%lx",
4112 				   start, end);
4113 			kvm_s390_sync_request(KVM_REQ_REFRESH_GUEST_PREFIX, vcpu);
4114 		}
4115 	}
4116 }
4117 
4118 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
4119 {
4120 	/* do not poll with more than halt_poll_max_steal percent of steal time */
4121 	if (get_lowcore()->avg_steal_timer * 100 / (TICK_USEC << 12) >=
4122 	    READ_ONCE(halt_poll_max_steal)) {
4123 		vcpu->stat.halt_no_poll_steal++;
4124 		return true;
4125 	}
4126 	return false;
4127 }
4128 
4129 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
4130 {
4131 	/* kvm common code refers to this, but never calls it */
4132 	BUG();
4133 	return 0;
4134 }
4135 
4136 static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu,
4137 					   struct kvm_one_reg *reg)
4138 {
4139 	int r = -EINVAL;
4140 
4141 	switch (reg->id) {
4142 	case KVM_REG_S390_TODPR:
4143 		r = put_user(vcpu->arch.sie_block->todpr,
4144 			     (u32 __user *)reg->addr);
4145 		break;
4146 	case KVM_REG_S390_EPOCHDIFF:
4147 		r = put_user(vcpu->arch.sie_block->epoch,
4148 			     (u64 __user *)reg->addr);
4149 		break;
4150 	case KVM_REG_S390_CPU_TIMER:
4151 		r = put_user(kvm_s390_get_cpu_timer(vcpu),
4152 			     (u64 __user *)reg->addr);
4153 		break;
4154 	case KVM_REG_S390_CLOCK_COMP:
4155 		r = put_user(vcpu->arch.sie_block->ckc,
4156 			     (u64 __user *)reg->addr);
4157 		break;
4158 	case KVM_REG_S390_PFTOKEN:
4159 		r = put_user(vcpu->arch.pfault_token,
4160 			     (u64 __user *)reg->addr);
4161 		break;
4162 	case KVM_REG_S390_PFCOMPARE:
4163 		r = put_user(vcpu->arch.pfault_compare,
4164 			     (u64 __user *)reg->addr);
4165 		break;
4166 	case KVM_REG_S390_PFSELECT:
4167 		r = put_user(vcpu->arch.pfault_select,
4168 			     (u64 __user *)reg->addr);
4169 		break;
4170 	case KVM_REG_S390_PP:
4171 		r = put_user(vcpu->arch.sie_block->pp,
4172 			     (u64 __user *)reg->addr);
4173 		break;
4174 	case KVM_REG_S390_GBEA:
4175 		r = put_user(vcpu->arch.sie_block->gbea,
4176 			     (u64 __user *)reg->addr);
4177 		break;
4178 	default:
4179 		break;
4180 	}
4181 
4182 	return r;
4183 }
4184 
4185 static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu,
4186 					   struct kvm_one_reg *reg)
4187 {
4188 	int r = -EINVAL;
4189 	__u64 val;
4190 
4191 	switch (reg->id) {
4192 	case KVM_REG_S390_TODPR:
4193 		r = get_user(vcpu->arch.sie_block->todpr,
4194 			     (u32 __user *)reg->addr);
4195 		break;
4196 	case KVM_REG_S390_EPOCHDIFF:
4197 		r = get_user(vcpu->arch.sie_block->epoch,
4198 			     (u64 __user *)reg->addr);
4199 		break;
4200 	case KVM_REG_S390_CPU_TIMER:
4201 		r = get_user(val, (u64 __user *)reg->addr);
4202 		if (!r)
4203 			kvm_s390_set_cpu_timer(vcpu, val);
4204 		break;
4205 	case KVM_REG_S390_CLOCK_COMP:
4206 		r = get_user(vcpu->arch.sie_block->ckc,
4207 			     (u64 __user *)reg->addr);
4208 		break;
4209 	case KVM_REG_S390_PFTOKEN:
4210 		r = get_user(vcpu->arch.pfault_token,
4211 			     (u64 __user *)reg->addr);
4212 		if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
4213 			kvm_clear_async_pf_completion_queue(vcpu);
4214 		break;
4215 	case KVM_REG_S390_PFCOMPARE:
4216 		r = get_user(vcpu->arch.pfault_compare,
4217 			     (u64 __user *)reg->addr);
4218 		break;
4219 	case KVM_REG_S390_PFSELECT:
4220 		r = get_user(vcpu->arch.pfault_select,
4221 			     (u64 __user *)reg->addr);
4222 		break;
4223 	case KVM_REG_S390_PP:
4224 		r = get_user(vcpu->arch.sie_block->pp,
4225 			     (u64 __user *)reg->addr);
4226 		break;
4227 	case KVM_REG_S390_GBEA:
4228 		r = get_user(vcpu->arch.sie_block->gbea,
4229 			     (u64 __user *)reg->addr);
4230 		break;
4231 	default:
4232 		break;
4233 	}
4234 
4235 	return r;
4236 }
4237 
4238 static void kvm_arch_vcpu_ioctl_normal_reset(struct kvm_vcpu *vcpu)
4239 {
4240 	vcpu->arch.sie_block->gpsw.mask &= ~PSW_MASK_RI;
4241 	vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
4242 	memset(vcpu->run->s.regs.riccb, 0, sizeof(vcpu->run->s.regs.riccb));
4243 
4244 	kvm_clear_async_pf_completion_queue(vcpu);
4245 	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
4246 		kvm_s390_vcpu_stop(vcpu);
4247 	kvm_s390_clear_local_irqs(vcpu);
4248 }
4249 
4250 static void kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu)
4251 {
4252 	/* Initial reset is a superset of the normal reset */
4253 	kvm_arch_vcpu_ioctl_normal_reset(vcpu);
4254 
4255 	/*
4256 	 * This equals initial cpu reset in pop, but we don't switch to ESA.
4257 	 * We do not only reset the internal data, but also ...
4258 	 */
4259 	vcpu->arch.sie_block->gpsw.mask = 0;
4260 	vcpu->arch.sie_block->gpsw.addr = 0;
4261 	kvm_s390_set_prefix(vcpu, 0);
4262 	kvm_s390_set_cpu_timer(vcpu, 0);
4263 	vcpu->arch.sie_block->ckc = 0;
4264 	memset(vcpu->arch.sie_block->gcr, 0, sizeof(vcpu->arch.sie_block->gcr));
4265 	vcpu->arch.sie_block->gcr[0] = CR0_INITIAL_MASK;
4266 	vcpu->arch.sie_block->gcr[14] = CR14_INITIAL_MASK;
4267 
4268 	/* ... the data in sync regs */
4269 	memset(vcpu->run->s.regs.crs, 0, sizeof(vcpu->run->s.regs.crs));
4270 	vcpu->run->s.regs.ckc = 0;
4271 	vcpu->run->s.regs.crs[0] = CR0_INITIAL_MASK;
4272 	vcpu->run->s.regs.crs[14] = CR14_INITIAL_MASK;
4273 	vcpu->run->psw_addr = 0;
4274 	vcpu->run->psw_mask = 0;
4275 	vcpu->run->s.regs.todpr = 0;
4276 	vcpu->run->s.regs.cputm = 0;
4277 	vcpu->run->s.regs.ckc = 0;
4278 	vcpu->run->s.regs.pp = 0;
4279 	vcpu->run->s.regs.gbea = 1;
4280 	vcpu->run->s.regs.fpc = 0;
4281 	/*
4282 	 * Do not reset these registers in the protected case, as some of
4283 	 * them are overlaid and they are not accessible in this case
4284 	 * anyway.
4285 	 */
4286 	if (!kvm_s390_pv_cpu_is_protected(vcpu)) {
4287 		vcpu->arch.sie_block->gbea = 1;
4288 		vcpu->arch.sie_block->pp = 0;
4289 		vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
4290 		vcpu->arch.sie_block->todpr = 0;
4291 	}
4292 }
4293 
4294 static void kvm_arch_vcpu_ioctl_clear_reset(struct kvm_vcpu *vcpu)
4295 {
4296 	struct kvm_sync_regs *regs = &vcpu->run->s.regs;
4297 
4298 	/* Clear reset is a superset of the initial reset */
4299 	kvm_arch_vcpu_ioctl_initial_reset(vcpu);
4300 
4301 	memset(&regs->gprs, 0, sizeof(regs->gprs));
4302 	memset(&regs->vrs, 0, sizeof(regs->vrs));
4303 	memset(&regs->acrs, 0, sizeof(regs->acrs));
4304 	memset(&regs->gscb, 0, sizeof(regs->gscb));
4305 
4306 	regs->etoken = 0;
4307 	regs->etoken_extension = 0;
4308 }
4309 
4310 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
4311 {
4312 	vcpu_load(vcpu);
4313 	memcpy(&vcpu->run->s.regs.gprs, &regs->gprs, sizeof(regs->gprs));
4314 	vcpu_put(vcpu);
4315 	return 0;
4316 }
4317 
4318 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
4319 {
4320 	vcpu_load(vcpu);
4321 	memcpy(&regs->gprs, &vcpu->run->s.regs.gprs, sizeof(regs->gprs));
4322 	vcpu_put(vcpu);
4323 	return 0;
4324 }
4325 
4326 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
4327 				  struct kvm_sregs *sregs)
4328 {
4329 	vcpu_load(vcpu);
4330 
4331 	memcpy(&vcpu->run->s.regs.acrs, &sregs->acrs, sizeof(sregs->acrs));
4332 	memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs));
4333 
4334 	vcpu_put(vcpu);
4335 	return 0;
4336 }
4337 
4338 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
4339 				  struct kvm_sregs *sregs)
4340 {
4341 	vcpu_load(vcpu);
4342 
4343 	memcpy(&sregs->acrs, &vcpu->run->s.regs.acrs, sizeof(sregs->acrs));
4344 	memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs));
4345 
4346 	vcpu_put(vcpu);
4347 	return 0;
4348 }
4349 
4350 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
4351 {
4352 	int ret = 0;
4353 
4354 	vcpu_load(vcpu);
4355 
4356 	vcpu->run->s.regs.fpc = fpu->fpc;
4357 	if (cpu_has_vx())
4358 		convert_fp_to_vx((__vector128 *) vcpu->run->s.regs.vrs,
4359 				 (freg_t *) fpu->fprs);
4360 	else
4361 		memcpy(vcpu->run->s.regs.fprs, &fpu->fprs, sizeof(fpu->fprs));
4362 
4363 	vcpu_put(vcpu);
4364 	return ret;
4365 }
4366 
4367 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
4368 {
4369 	vcpu_load(vcpu);
4370 
4371 	if (cpu_has_vx())
4372 		convert_vx_to_fp((freg_t *) fpu->fprs,
4373 				 (__vector128 *) vcpu->run->s.regs.vrs);
4374 	else
4375 		memcpy(fpu->fprs, vcpu->run->s.regs.fprs, sizeof(fpu->fprs));
4376 	fpu->fpc = vcpu->run->s.regs.fpc;
4377 
4378 	vcpu_put(vcpu);
4379 	return 0;
4380 }
4381 
4382 static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw)
4383 {
4384 	int rc = 0;
4385 
4386 	if (!is_vcpu_stopped(vcpu))
4387 		rc = -EBUSY;
4388 	else {
4389 		vcpu->run->psw_mask = psw.mask;
4390 		vcpu->run->psw_addr = psw.addr;
4391 	}
4392 	return rc;
4393 }
4394 
4395 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
4396 				  struct kvm_translation *tr)
4397 {
4398 	return -EINVAL; /* not implemented yet */
4399 }
4400 
4401 #define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
4402 			      KVM_GUESTDBG_USE_HW_BP | \
4403 			      KVM_GUESTDBG_ENABLE)
4404 
4405 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
4406 					struct kvm_guest_debug *dbg)
4407 {
4408 	int rc = 0;
4409 
4410 	vcpu_load(vcpu);
4411 
4412 	vcpu->guest_debug = 0;
4413 	kvm_s390_clear_bp_data(vcpu);
4414 
4415 	if (dbg->control & ~VALID_GUESTDBG_FLAGS) {
4416 		rc = -EINVAL;
4417 		goto out;
4418 	}
4419 	if (!sclp.has_gpere) {
4420 		rc = -EINVAL;
4421 		goto out;
4422 	}
4423 
4424 	if (dbg->control & KVM_GUESTDBG_ENABLE) {
4425 		vcpu->guest_debug = dbg->control;
4426 		/* enforce guest PER */
4427 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_P);
4428 
4429 		if (dbg->control & KVM_GUESTDBG_USE_HW_BP)
4430 			rc = kvm_s390_import_bp_data(vcpu, dbg);
4431 	} else {
4432 		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
4433 		vcpu->arch.guestdbg.last_bp = 0;
4434 	}
4435 
4436 	if (rc) {
4437 		vcpu->guest_debug = 0;
4438 		kvm_s390_clear_bp_data(vcpu);
4439 		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
4440 	}
4441 
4442 out:
4443 	vcpu_put(vcpu);
4444 	return rc;
4445 }
4446 
4447 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
4448 				    struct kvm_mp_state *mp_state)
4449 {
4450 	int ret;
4451 
4452 	vcpu_load(vcpu);
4453 
4454 	/* CHECK_STOP and LOAD are not supported yet */
4455 	ret = is_vcpu_stopped(vcpu) ? KVM_MP_STATE_STOPPED :
4456 				      KVM_MP_STATE_OPERATING;
4457 
4458 	vcpu_put(vcpu);
4459 	return ret;
4460 }
4461 
4462 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
4463 				    struct kvm_mp_state *mp_state)
4464 {
4465 	int rc = 0;
4466 
4467 	vcpu_load(vcpu);
4468 
4469 	/* user space knows about this interface - let it control the state */
4470 	kvm_s390_set_user_cpu_state_ctrl(vcpu->kvm);
4471 
4472 	switch (mp_state->mp_state) {
4473 	case KVM_MP_STATE_STOPPED:
4474 		rc = kvm_s390_vcpu_stop(vcpu);
4475 		break;
4476 	case KVM_MP_STATE_OPERATING:
4477 		rc = kvm_s390_vcpu_start(vcpu);
4478 		break;
4479 	case KVM_MP_STATE_LOAD:
4480 		if (!kvm_s390_pv_cpu_is_protected(vcpu)) {
4481 			rc = -ENXIO;
4482 			break;
4483 		}
4484 		rc = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_OPR_LOAD);
4485 		break;
4486 	case KVM_MP_STATE_CHECK_STOP:
4487 		fallthrough;	/* CHECK_STOP and LOAD are not supported yet */
4488 	default:
4489 		rc = -ENXIO;
4490 	}
4491 
4492 	vcpu_put(vcpu);
4493 	return rc;
4494 }
4495 
4496 static bool ibs_enabled(struct kvm_vcpu *vcpu)
4497 {
4498 	return kvm_s390_test_cpuflags(vcpu, CPUSTAT_IBS);
4499 }
4500 
4501 static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
4502 {
4503 retry:
4504 	kvm_s390_vcpu_request_handled(vcpu);
4505 	if (!kvm_request_pending(vcpu))
4506 		return 0;
4507 	/*
4508 	 * If the guest prefix changed, re-arm the ipte notifier for the
4509 	 * guest prefix page. gmap_mprotect_notify will wait on the ptl lock.
4510 	 * This ensures that the ipte instruction for this request has
4511 	 * already finished. We might race against a second unmapper that
4512 	 * wants to set the blocking bit. Lets just retry the request loop.
4513 	 */
4514 	if (kvm_check_request(KVM_REQ_REFRESH_GUEST_PREFIX, vcpu)) {
4515 		int rc;
4516 		rc = gmap_mprotect_notify(vcpu->arch.gmap,
4517 					  kvm_s390_get_prefix(vcpu),
4518 					  PAGE_SIZE * 2, PROT_WRITE);
4519 		if (rc) {
4520 			kvm_make_request(KVM_REQ_REFRESH_GUEST_PREFIX, vcpu);
4521 			return rc;
4522 		}
4523 		goto retry;
4524 	}
4525 
4526 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
4527 		vcpu->arch.sie_block->ihcpu = 0xffff;
4528 		goto retry;
4529 	}
4530 
4531 	if (kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu)) {
4532 		if (!ibs_enabled(vcpu)) {
4533 			trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 1);
4534 			kvm_s390_set_cpuflags(vcpu, CPUSTAT_IBS);
4535 		}
4536 		goto retry;
4537 	}
4538 
4539 	if (kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu)) {
4540 		if (ibs_enabled(vcpu)) {
4541 			trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 0);
4542 			kvm_s390_clear_cpuflags(vcpu, CPUSTAT_IBS);
4543 		}
4544 		goto retry;
4545 	}
4546 
4547 	if (kvm_check_request(KVM_REQ_ICPT_OPEREXC, vcpu)) {
4548 		vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
4549 		goto retry;
4550 	}
4551 
4552 	if (kvm_check_request(KVM_REQ_START_MIGRATION, vcpu)) {
4553 		/*
4554 		 * Disable CMM virtualization; we will emulate the ESSA
4555 		 * instruction manually, in order to provide additional
4556 		 * functionalities needed for live migration.
4557 		 */
4558 		vcpu->arch.sie_block->ecb2 &= ~ECB2_CMMA;
4559 		goto retry;
4560 	}
4561 
4562 	if (kvm_check_request(KVM_REQ_STOP_MIGRATION, vcpu)) {
4563 		/*
4564 		 * Re-enable CMM virtualization if CMMA is available and
4565 		 * CMM has been used.
4566 		 */
4567 		if ((vcpu->kvm->arch.use_cmma) &&
4568 		    (vcpu->kvm->mm->context.uses_cmm))
4569 			vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
4570 		goto retry;
4571 	}
4572 
4573 	/* we left the vsie handler, nothing to do, just clear the request */
4574 	kvm_clear_request(KVM_REQ_VSIE_RESTART, vcpu);
4575 
4576 	return 0;
4577 }
4578 
4579 static void __kvm_s390_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod)
4580 {
4581 	struct kvm_vcpu *vcpu;
4582 	union tod_clock clk;
4583 	unsigned long i;
4584 
4585 	preempt_disable();
4586 
4587 	store_tod_clock_ext(&clk);
4588 
4589 	kvm->arch.epoch = gtod->tod - clk.tod;
4590 	kvm->arch.epdx = 0;
4591 	if (test_kvm_facility(kvm, 139)) {
4592 		kvm->arch.epdx = gtod->epoch_idx - clk.ei;
4593 		if (kvm->arch.epoch > gtod->tod)
4594 			kvm->arch.epdx -= 1;
4595 	}
4596 
4597 	kvm_s390_vcpu_block_all(kvm);
4598 	kvm_for_each_vcpu(i, vcpu, kvm) {
4599 		vcpu->arch.sie_block->epoch = kvm->arch.epoch;
4600 		vcpu->arch.sie_block->epdx  = kvm->arch.epdx;
4601 	}
4602 
4603 	kvm_s390_vcpu_unblock_all(kvm);
4604 	preempt_enable();
4605 }
4606 
4607 int kvm_s390_try_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod)
4608 {
4609 	if (!mutex_trylock(&kvm->lock))
4610 		return 0;
4611 	__kvm_s390_set_tod_clock(kvm, gtod);
4612 	mutex_unlock(&kvm->lock);
4613 	return 1;
4614 }
4615 
4616 static void __kvm_inject_pfault_token(struct kvm_vcpu *vcpu, bool start_token,
4617 				      unsigned long token)
4618 {
4619 	struct kvm_s390_interrupt inti;
4620 	struct kvm_s390_irq irq;
4621 
4622 	if (start_token) {
4623 		irq.u.ext.ext_params2 = token;
4624 		irq.type = KVM_S390_INT_PFAULT_INIT;
4625 		WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &irq));
4626 	} else {
4627 		inti.type = KVM_S390_INT_PFAULT_DONE;
4628 		inti.parm64 = token;
4629 		WARN_ON_ONCE(kvm_s390_inject_vm(vcpu->kvm, &inti));
4630 	}
4631 }
4632 
4633 bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
4634 				     struct kvm_async_pf *work)
4635 {
4636 	trace_kvm_s390_pfault_init(vcpu, work->arch.pfault_token);
4637 	__kvm_inject_pfault_token(vcpu, true, work->arch.pfault_token);
4638 
4639 	return true;
4640 }
4641 
4642 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
4643 				 struct kvm_async_pf *work)
4644 {
4645 	trace_kvm_s390_pfault_done(vcpu, work->arch.pfault_token);
4646 	__kvm_inject_pfault_token(vcpu, false, work->arch.pfault_token);
4647 }
4648 
4649 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
4650 			       struct kvm_async_pf *work)
4651 {
4652 	/* s390 will always inject the page directly */
4653 }
4654 
4655 bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu)
4656 {
4657 	/*
4658 	 * s390 will always inject the page directly,
4659 	 * but we still want check_async_completion to cleanup
4660 	 */
4661 	return true;
4662 }
4663 
4664 static bool kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu)
4665 {
4666 	hva_t hva;
4667 	struct kvm_arch_async_pf arch;
4668 
4669 	if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
4670 		return false;
4671 	if ((vcpu->arch.sie_block->gpsw.mask & vcpu->arch.pfault_select) !=
4672 	    vcpu->arch.pfault_compare)
4673 		return false;
4674 	if (psw_extint_disabled(vcpu))
4675 		return false;
4676 	if (kvm_s390_vcpu_has_irq(vcpu, 0))
4677 		return false;
4678 	if (!(vcpu->arch.sie_block->gcr[0] & CR0_SERVICE_SIGNAL_SUBMASK))
4679 		return false;
4680 	if (!vcpu->arch.gmap->pfault_enabled)
4681 		return false;
4682 
4683 	hva = gfn_to_hva(vcpu->kvm, current->thread.gmap_teid.addr);
4684 	if (read_guest_real(vcpu, vcpu->arch.pfault_token, &arch.pfault_token, 8))
4685 		return false;
4686 
4687 	return kvm_setup_async_pf(vcpu, current->thread.gmap_teid.addr * PAGE_SIZE, hva, &arch);
4688 }
4689 
4690 static int vcpu_pre_run(struct kvm_vcpu *vcpu)
4691 {
4692 	int rc, cpuflags;
4693 
4694 	/*
4695 	 * On s390 notifications for arriving pages will be delivered directly
4696 	 * to the guest but the house keeping for completed pfaults is
4697 	 * handled outside the worker.
4698 	 */
4699 	kvm_check_async_pf_completion(vcpu);
4700 
4701 	vcpu->arch.sie_block->gg14 = vcpu->run->s.regs.gprs[14];
4702 	vcpu->arch.sie_block->gg15 = vcpu->run->s.regs.gprs[15];
4703 
4704 	if (need_resched())
4705 		schedule();
4706 
4707 	if (!kvm_is_ucontrol(vcpu->kvm)) {
4708 		rc = kvm_s390_deliver_pending_interrupts(vcpu);
4709 		if (rc || guestdbg_exit_pending(vcpu))
4710 			return rc;
4711 	}
4712 
4713 	rc = kvm_s390_handle_requests(vcpu);
4714 	if (rc)
4715 		return rc;
4716 
4717 	if (guestdbg_enabled(vcpu)) {
4718 		kvm_s390_backup_guest_per_regs(vcpu);
4719 		kvm_s390_patch_guest_per_regs(vcpu);
4720 	}
4721 
4722 	clear_bit(vcpu->vcpu_idx, vcpu->kvm->arch.gisa_int.kicked_mask);
4723 
4724 	vcpu->arch.sie_block->icptcode = 0;
4725 	current->thread.gmap_int_code = 0;
4726 	cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags);
4727 	VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags);
4728 	trace_kvm_s390_sie_enter(vcpu, cpuflags);
4729 
4730 	return 0;
4731 }
4732 
4733 static int vcpu_post_run_addressing_exception(struct kvm_vcpu *vcpu)
4734 {
4735 	struct kvm_s390_pgm_info pgm_info = {
4736 		.code = PGM_ADDRESSING,
4737 	};
4738 	u8 opcode, ilen;
4739 	int rc;
4740 
4741 	VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction");
4742 	trace_kvm_s390_sie_fault(vcpu);
4743 
4744 	/*
4745 	 * We want to inject an addressing exception, which is defined as a
4746 	 * suppressing or terminating exception. However, since we came here
4747 	 * by a DAT access exception, the PSW still points to the faulting
4748 	 * instruction since DAT exceptions are nullifying. So we've got
4749 	 * to look up the current opcode to get the length of the instruction
4750 	 * to be able to forward the PSW.
4751 	 */
4752 	rc = read_guest_instr(vcpu, vcpu->arch.sie_block->gpsw.addr, &opcode, 1);
4753 	ilen = insn_length(opcode);
4754 	if (rc < 0) {
4755 		return rc;
4756 	} else if (rc) {
4757 		/* Instruction-Fetching Exceptions - we can't detect the ilen.
4758 		 * Forward by arbitrary ilc, injection will take care of
4759 		 * nullification if necessary.
4760 		 */
4761 		pgm_info = vcpu->arch.pgm;
4762 		ilen = 4;
4763 	}
4764 	pgm_info.flags = ilen | KVM_S390_PGM_FLAGS_ILC_VALID;
4765 	kvm_s390_forward_psw(vcpu, ilen);
4766 	return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
4767 }
4768 
4769 static int vcpu_post_run_handle_fault(struct kvm_vcpu *vcpu)
4770 {
4771 	unsigned int flags = 0;
4772 	unsigned long gaddr;
4773 	int rc = 0;
4774 
4775 	gaddr = current->thread.gmap_teid.addr * PAGE_SIZE;
4776 	if (kvm_s390_cur_gmap_fault_is_write())
4777 		flags = FAULT_FLAG_WRITE;
4778 
4779 	switch (current->thread.gmap_int_code & PGM_INT_CODE_MASK) {
4780 	case 0:
4781 		vcpu->stat.exit_null++;
4782 		break;
4783 	case PGM_NON_SECURE_STORAGE_ACCESS:
4784 		KVM_BUG(current->thread.gmap_teid.as != PSW_BITS_AS_PRIMARY, vcpu->kvm,
4785 			"Unexpected program interrupt 0x%x, TEID 0x%016lx",
4786 			current->thread.gmap_int_code, current->thread.gmap_teid.val);
4787 		/*
4788 		 * This is normal operation; a page belonging to a protected
4789 		 * guest has not been imported yet. Try to import the page into
4790 		 * the protected guest.
4791 		 */
4792 		if (gmap_convert_to_secure(vcpu->arch.gmap, gaddr) == -EINVAL)
4793 			send_sig(SIGSEGV, current, 0);
4794 		break;
4795 	case PGM_SECURE_STORAGE_ACCESS:
4796 	case PGM_SECURE_STORAGE_VIOLATION:
4797 		KVM_BUG(current->thread.gmap_teid.as != PSW_BITS_AS_PRIMARY, vcpu->kvm,
4798 			"Unexpected program interrupt 0x%x, TEID 0x%016lx",
4799 			current->thread.gmap_int_code, current->thread.gmap_teid.val);
4800 		/*
4801 		 * This can happen after a reboot with asynchronous teardown;
4802 		 * the new guest (normal or protected) will run on top of the
4803 		 * previous protected guest. The old pages need to be destroyed
4804 		 * so the new guest can use them.
4805 		 */
4806 		if (gmap_destroy_page(vcpu->arch.gmap, gaddr)) {
4807 			/*
4808 			 * Either KVM messed up the secure guest mapping or the
4809 			 * same page is mapped into multiple secure guests.
4810 			 *
4811 			 * This exception is only triggered when a guest 2 is
4812 			 * running and can therefore never occur in kernel
4813 			 * context.
4814 			 */
4815 			pr_warn_ratelimited("Secure storage violation (%x) in task: %s, pid %d\n",
4816 					    current->thread.gmap_int_code, current->comm,
4817 					    current->pid);
4818 			send_sig(SIGSEGV, current, 0);
4819 		}
4820 		break;
4821 	case PGM_PROTECTION:
4822 	case PGM_SEGMENT_TRANSLATION:
4823 	case PGM_PAGE_TRANSLATION:
4824 	case PGM_ASCE_TYPE:
4825 	case PGM_REGION_FIRST_TRANS:
4826 	case PGM_REGION_SECOND_TRANS:
4827 	case PGM_REGION_THIRD_TRANS:
4828 		KVM_BUG(current->thread.gmap_teid.as != PSW_BITS_AS_PRIMARY, vcpu->kvm,
4829 			"Unexpected program interrupt 0x%x, TEID 0x%016lx",
4830 			current->thread.gmap_int_code, current->thread.gmap_teid.val);
4831 		if (vcpu->arch.gmap->pfault_enabled) {
4832 			rc = gmap_fault(vcpu->arch.gmap, gaddr, flags | FAULT_FLAG_RETRY_NOWAIT);
4833 			if (rc == -EFAULT)
4834 				return vcpu_post_run_addressing_exception(vcpu);
4835 			if (rc == -EAGAIN) {
4836 				trace_kvm_s390_major_guest_pfault(vcpu);
4837 				if (kvm_arch_setup_async_pf(vcpu))
4838 					return 0;
4839 				vcpu->stat.pfault_sync++;
4840 			} else {
4841 				return rc;
4842 			}
4843 		}
4844 		rc = gmap_fault(vcpu->arch.gmap, gaddr, flags);
4845 		if (rc == -EFAULT) {
4846 			if (kvm_is_ucontrol(vcpu->kvm)) {
4847 				vcpu->run->exit_reason = KVM_EXIT_S390_UCONTROL;
4848 				vcpu->run->s390_ucontrol.trans_exc_code = gaddr;
4849 				vcpu->run->s390_ucontrol.pgm_code = 0x10;
4850 				return -EREMOTE;
4851 			}
4852 			return vcpu_post_run_addressing_exception(vcpu);
4853 		}
4854 		break;
4855 	default:
4856 		KVM_BUG(1, vcpu->kvm, "Unexpected program interrupt 0x%x, TEID 0x%016lx",
4857 			current->thread.gmap_int_code, current->thread.gmap_teid.val);
4858 		send_sig(SIGSEGV, current, 0);
4859 		break;
4860 	}
4861 	return rc;
4862 }
4863 
4864 static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
4865 {
4866 	struct mcck_volatile_info *mcck_info;
4867 	struct sie_page *sie_page;
4868 	int rc;
4869 
4870 	VCPU_EVENT(vcpu, 6, "exit sie icptcode %d",
4871 		   vcpu->arch.sie_block->icptcode);
4872 	trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode);
4873 
4874 	if (guestdbg_enabled(vcpu))
4875 		kvm_s390_restore_guest_per_regs(vcpu);
4876 
4877 	vcpu->run->s.regs.gprs[14] = vcpu->arch.sie_block->gg14;
4878 	vcpu->run->s.regs.gprs[15] = vcpu->arch.sie_block->gg15;
4879 
4880 	if (exit_reason == -EINTR) {
4881 		VCPU_EVENT(vcpu, 3, "%s", "machine check");
4882 		sie_page = container_of(vcpu->arch.sie_block,
4883 					struct sie_page, sie_block);
4884 		mcck_info = &sie_page->mcck_info;
4885 		kvm_s390_reinject_machine_check(vcpu, mcck_info);
4886 		return 0;
4887 	}
4888 
4889 	if (vcpu->arch.sie_block->icptcode > 0) {
4890 		rc = kvm_handle_sie_intercept(vcpu);
4891 
4892 		if (rc != -EOPNOTSUPP)
4893 			return rc;
4894 		vcpu->run->exit_reason = KVM_EXIT_S390_SIEIC;
4895 		vcpu->run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode;
4896 		vcpu->run->s390_sieic.ipa = vcpu->arch.sie_block->ipa;
4897 		vcpu->run->s390_sieic.ipb = vcpu->arch.sie_block->ipb;
4898 		return -EREMOTE;
4899 	}
4900 
4901 	return vcpu_post_run_handle_fault(vcpu);
4902 }
4903 
4904 #define PSW_INT_MASK (PSW_MASK_EXT | PSW_MASK_IO | PSW_MASK_MCHECK)
4905 static int __vcpu_run(struct kvm_vcpu *vcpu)
4906 {
4907 	int rc, exit_reason;
4908 	struct sie_page *sie_page = (struct sie_page *)vcpu->arch.sie_block;
4909 
4910 	/*
4911 	 * We try to hold kvm->srcu during most of vcpu_run (except when run-
4912 	 * ning the guest), so that memslots (and other stuff) are protected
4913 	 */
4914 	kvm_vcpu_srcu_read_lock(vcpu);
4915 
4916 	do {
4917 		rc = vcpu_pre_run(vcpu);
4918 		if (rc || guestdbg_exit_pending(vcpu))
4919 			break;
4920 
4921 		kvm_vcpu_srcu_read_unlock(vcpu);
4922 		/*
4923 		 * As PF_VCPU will be used in fault handler, between
4924 		 * guest_enter and guest_exit should be no uaccess.
4925 		 */
4926 		local_irq_disable();
4927 		guest_enter_irqoff();
4928 		__disable_cpu_timer_accounting(vcpu);
4929 		local_irq_enable();
4930 		if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4931 			memcpy(sie_page->pv_grregs,
4932 			       vcpu->run->s.regs.gprs,
4933 			       sizeof(sie_page->pv_grregs));
4934 		}
4935 		exit_reason = sie64a(vcpu->arch.sie_block,
4936 				     vcpu->run->s.regs.gprs,
4937 				     vcpu->arch.gmap->asce);
4938 		if (kvm_s390_pv_cpu_is_protected(vcpu)) {
4939 			memcpy(vcpu->run->s.regs.gprs,
4940 			       sie_page->pv_grregs,
4941 			       sizeof(sie_page->pv_grregs));
4942 			/*
4943 			 * We're not allowed to inject interrupts on intercepts
4944 			 * that leave the guest state in an "in-between" state
4945 			 * where the next SIE entry will do a continuation.
4946 			 * Fence interrupts in our "internal" PSW.
4947 			 */
4948 			if (vcpu->arch.sie_block->icptcode == ICPT_PV_INSTR ||
4949 			    vcpu->arch.sie_block->icptcode == ICPT_PV_PREF) {
4950 				vcpu->arch.sie_block->gpsw.mask &= ~PSW_INT_MASK;
4951 			}
4952 		}
4953 		local_irq_disable();
4954 		__enable_cpu_timer_accounting(vcpu);
4955 		guest_exit_irqoff();
4956 		local_irq_enable();
4957 		kvm_vcpu_srcu_read_lock(vcpu);
4958 
4959 		rc = vcpu_post_run(vcpu, exit_reason);
4960 	} while (!signal_pending(current) && !guestdbg_exit_pending(vcpu) && !rc);
4961 
4962 	kvm_vcpu_srcu_read_unlock(vcpu);
4963 	return rc;
4964 }
4965 
4966 static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
4967 {
4968 	struct kvm_run *kvm_run = vcpu->run;
4969 	struct runtime_instr_cb *riccb;
4970 	struct gs_cb *gscb;
4971 
4972 	riccb = (struct runtime_instr_cb *) &kvm_run->s.regs.riccb;
4973 	gscb = (struct gs_cb *) &kvm_run->s.regs.gscb;
4974 	vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask;
4975 	vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr;
4976 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
4977 		vcpu->arch.sie_block->todpr = kvm_run->s.regs.todpr;
4978 		vcpu->arch.sie_block->pp = kvm_run->s.regs.pp;
4979 		vcpu->arch.sie_block->gbea = kvm_run->s.regs.gbea;
4980 	}
4981 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_PFAULT) {
4982 		vcpu->arch.pfault_token = kvm_run->s.regs.pft;
4983 		vcpu->arch.pfault_select = kvm_run->s.regs.pfs;
4984 		vcpu->arch.pfault_compare = kvm_run->s.regs.pfc;
4985 		if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
4986 			kvm_clear_async_pf_completion_queue(vcpu);
4987 	}
4988 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
4989 		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
4990 		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
4991 		VCPU_EVENT(vcpu, 3, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
4992 	}
4993 	/*
4994 	 * If userspace sets the riccb (e.g. after migration) to a valid state,
4995 	 * we should enable RI here instead of doing the lazy enablement.
4996 	 */
4997 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_RICCB) &&
4998 	    test_kvm_facility(vcpu->kvm, 64) &&
4999 	    riccb->v &&
5000 	    !(vcpu->arch.sie_block->ecb3 & ECB3_RI)) {
5001 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (sync_regs)");
5002 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
5003 	}
5004 	/*
5005 	 * If userspace sets the gscb (e.g. after migration) to non-zero,
5006 	 * we should enable GS here instead of doing the lazy enablement.
5007 	 */
5008 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_GSCB) &&
5009 	    test_kvm_facility(vcpu->kvm, 133) &&
5010 	    gscb->gssm &&
5011 	    !vcpu->arch.gs_enabled) {
5012 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (sync_regs)");
5013 		vcpu->arch.sie_block->ecb |= ECB_GS;
5014 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
5015 		vcpu->arch.gs_enabled = 1;
5016 	}
5017 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_BPBC) &&
5018 	    test_kvm_facility(vcpu->kvm, 82)) {
5019 		vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
5020 		vcpu->arch.sie_block->fpf |= kvm_run->s.regs.bpbc ? FPF_BPBC : 0;
5021 	}
5022 	if (MACHINE_HAS_GS) {
5023 		preempt_disable();
5024 		local_ctl_set_bit(2, CR2_GUARDED_STORAGE_BIT);
5025 		if (current->thread.gs_cb) {
5026 			vcpu->arch.host_gscb = current->thread.gs_cb;
5027 			save_gs_cb(vcpu->arch.host_gscb);
5028 		}
5029 		if (vcpu->arch.gs_enabled) {
5030 			current->thread.gs_cb = (struct gs_cb *)
5031 						&vcpu->run->s.regs.gscb;
5032 			restore_gs_cb(current->thread.gs_cb);
5033 		}
5034 		preempt_enable();
5035 	}
5036 	/* SIE will load etoken directly from SDNX and therefore kvm_run */
5037 }
5038 
5039 static void sync_regs(struct kvm_vcpu *vcpu)
5040 {
5041 	struct kvm_run *kvm_run = vcpu->run;
5042 
5043 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_PREFIX)
5044 		kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix);
5045 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_CRS) {
5046 		memcpy(&vcpu->arch.sie_block->gcr, &kvm_run->s.regs.crs, 128);
5047 		/* some control register changes require a tlb flush */
5048 		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
5049 	}
5050 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
5051 		kvm_s390_set_cpu_timer(vcpu, kvm_run->s.regs.cputm);
5052 		vcpu->arch.sie_block->ckc = kvm_run->s.regs.ckc;
5053 	}
5054 	save_access_regs(vcpu->arch.host_acrs);
5055 	restore_access_regs(vcpu->run->s.regs.acrs);
5056 	vcpu->arch.acrs_loaded = true;
5057 	kvm_s390_fpu_load(vcpu->run);
5058 	/* Sync fmt2 only data */
5059 	if (likely(!kvm_s390_pv_cpu_is_protected(vcpu))) {
5060 		sync_regs_fmt2(vcpu);
5061 	} else {
5062 		/*
5063 		 * In several places we have to modify our internal view to
5064 		 * not do things that are disallowed by the ultravisor. For
5065 		 * example we must not inject interrupts after specific exits
5066 		 * (e.g. 112 prefix page not secure). We do this by turning
5067 		 * off the machine check, external and I/O interrupt bits
5068 		 * of our PSW copy. To avoid getting validity intercepts, we
5069 		 * do only accept the condition code from userspace.
5070 		 */
5071 		vcpu->arch.sie_block->gpsw.mask &= ~PSW_MASK_CC;
5072 		vcpu->arch.sie_block->gpsw.mask |= kvm_run->psw_mask &
5073 						   PSW_MASK_CC;
5074 	}
5075 
5076 	kvm_run->kvm_dirty_regs = 0;
5077 }
5078 
5079 static void store_regs_fmt2(struct kvm_vcpu *vcpu)
5080 {
5081 	struct kvm_run *kvm_run = vcpu->run;
5082 
5083 	kvm_run->s.regs.todpr = vcpu->arch.sie_block->todpr;
5084 	kvm_run->s.regs.pp = vcpu->arch.sie_block->pp;
5085 	kvm_run->s.regs.gbea = vcpu->arch.sie_block->gbea;
5086 	kvm_run->s.regs.bpbc = (vcpu->arch.sie_block->fpf & FPF_BPBC) == FPF_BPBC;
5087 	kvm_run->s.regs.diag318 = vcpu->arch.diag318_info.val;
5088 	if (MACHINE_HAS_GS) {
5089 		preempt_disable();
5090 		local_ctl_set_bit(2, CR2_GUARDED_STORAGE_BIT);
5091 		if (vcpu->arch.gs_enabled)
5092 			save_gs_cb(current->thread.gs_cb);
5093 		current->thread.gs_cb = vcpu->arch.host_gscb;
5094 		restore_gs_cb(vcpu->arch.host_gscb);
5095 		if (!vcpu->arch.host_gscb)
5096 			local_ctl_clear_bit(2, CR2_GUARDED_STORAGE_BIT);
5097 		vcpu->arch.host_gscb = NULL;
5098 		preempt_enable();
5099 	}
5100 	/* SIE will save etoken directly into SDNX and therefore kvm_run */
5101 }
5102 
5103 static void store_regs(struct kvm_vcpu *vcpu)
5104 {
5105 	struct kvm_run *kvm_run = vcpu->run;
5106 
5107 	kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask;
5108 	kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr;
5109 	kvm_run->s.regs.prefix = kvm_s390_get_prefix(vcpu);
5110 	memcpy(&kvm_run->s.regs.crs, &vcpu->arch.sie_block->gcr, 128);
5111 	kvm_run->s.regs.cputm = kvm_s390_get_cpu_timer(vcpu);
5112 	kvm_run->s.regs.ckc = vcpu->arch.sie_block->ckc;
5113 	kvm_run->s.regs.pft = vcpu->arch.pfault_token;
5114 	kvm_run->s.regs.pfs = vcpu->arch.pfault_select;
5115 	kvm_run->s.regs.pfc = vcpu->arch.pfault_compare;
5116 	save_access_regs(vcpu->run->s.regs.acrs);
5117 	restore_access_regs(vcpu->arch.host_acrs);
5118 	vcpu->arch.acrs_loaded = false;
5119 	kvm_s390_fpu_store(vcpu->run);
5120 	if (likely(!kvm_s390_pv_cpu_is_protected(vcpu)))
5121 		store_regs_fmt2(vcpu);
5122 }
5123 
5124 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
5125 {
5126 	struct kvm_run *kvm_run = vcpu->run;
5127 	DECLARE_KERNEL_FPU_ONSTACK32(fpu);
5128 	int rc;
5129 
5130 	/*
5131 	 * Running a VM while dumping always has the potential to
5132 	 * produce inconsistent dump data. But for PV vcpus a SIE
5133 	 * entry while dumping could also lead to a fatal validity
5134 	 * intercept which we absolutely want to avoid.
5135 	 */
5136 	if (vcpu->kvm->arch.pv.dumping)
5137 		return -EINVAL;
5138 
5139 	if (!vcpu->wants_to_run)
5140 		return -EINTR;
5141 
5142 	if (kvm_run->kvm_valid_regs & ~KVM_SYNC_S390_VALID_FIELDS ||
5143 	    kvm_run->kvm_dirty_regs & ~KVM_SYNC_S390_VALID_FIELDS)
5144 		return -EINVAL;
5145 
5146 	vcpu_load(vcpu);
5147 
5148 	if (guestdbg_exit_pending(vcpu)) {
5149 		kvm_s390_prepare_debug_exit(vcpu);
5150 		rc = 0;
5151 		goto out;
5152 	}
5153 
5154 	kvm_sigset_activate(vcpu);
5155 
5156 	/*
5157 	 * no need to check the return value of vcpu_start as it can only have
5158 	 * an error for protvirt, but protvirt means user cpu state
5159 	 */
5160 	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm)) {
5161 		kvm_s390_vcpu_start(vcpu);
5162 	} else if (is_vcpu_stopped(vcpu)) {
5163 		pr_err_ratelimited("can't run stopped vcpu %d\n",
5164 				   vcpu->vcpu_id);
5165 		rc = -EINVAL;
5166 		goto out;
5167 	}
5168 
5169 	kernel_fpu_begin(&fpu, KERNEL_FPC | KERNEL_VXR);
5170 	sync_regs(vcpu);
5171 	enable_cpu_timer_accounting(vcpu);
5172 
5173 	might_fault();
5174 	rc = __vcpu_run(vcpu);
5175 
5176 	if (signal_pending(current) && !rc) {
5177 		kvm_run->exit_reason = KVM_EXIT_INTR;
5178 		rc = -EINTR;
5179 	}
5180 
5181 	if (guestdbg_exit_pending(vcpu) && !rc)  {
5182 		kvm_s390_prepare_debug_exit(vcpu);
5183 		rc = 0;
5184 	}
5185 
5186 	if (rc == -EREMOTE) {
5187 		/* userspace support is needed, kvm_run has been prepared */
5188 		rc = 0;
5189 	}
5190 
5191 	disable_cpu_timer_accounting(vcpu);
5192 	store_regs(vcpu);
5193 	kernel_fpu_end(&fpu, KERNEL_FPC | KERNEL_VXR);
5194 
5195 	kvm_sigset_deactivate(vcpu);
5196 
5197 	vcpu->stat.exit_userspace++;
5198 out:
5199 	vcpu_put(vcpu);
5200 	return rc;
5201 }
5202 
5203 /*
5204  * store status at address
5205  * we use have two special cases:
5206  * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
5207  * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
5208  */
5209 int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
5210 {
5211 	unsigned char archmode = 1;
5212 	freg_t fprs[NUM_FPRS];
5213 	unsigned int px;
5214 	u64 clkcomp, cputm;
5215 	int rc;
5216 
5217 	px = kvm_s390_get_prefix(vcpu);
5218 	if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
5219 		if (write_guest_abs(vcpu, 163, &archmode, 1))
5220 			return -EFAULT;
5221 		gpa = 0;
5222 	} else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
5223 		if (write_guest_real(vcpu, 163, &archmode, 1))
5224 			return -EFAULT;
5225 		gpa = px;
5226 	} else
5227 		gpa -= __LC_FPREGS_SAVE_AREA;
5228 
5229 	/* manually convert vector registers if necessary */
5230 	if (cpu_has_vx()) {
5231 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
5232 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
5233 				     fprs, 128);
5234 	} else {
5235 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
5236 				     vcpu->run->s.regs.fprs, 128);
5237 	}
5238 	rc |= write_guest_abs(vcpu, gpa + __LC_GPREGS_SAVE_AREA,
5239 			      vcpu->run->s.regs.gprs, 128);
5240 	rc |= write_guest_abs(vcpu, gpa + __LC_PSW_SAVE_AREA,
5241 			      &vcpu->arch.sie_block->gpsw, 16);
5242 	rc |= write_guest_abs(vcpu, gpa + __LC_PREFIX_SAVE_AREA,
5243 			      &px, 4);
5244 	rc |= write_guest_abs(vcpu, gpa + __LC_FP_CREG_SAVE_AREA,
5245 			      &vcpu->run->s.regs.fpc, 4);
5246 	rc |= write_guest_abs(vcpu, gpa + __LC_TOD_PROGREG_SAVE_AREA,
5247 			      &vcpu->arch.sie_block->todpr, 4);
5248 	cputm = kvm_s390_get_cpu_timer(vcpu);
5249 	rc |= write_guest_abs(vcpu, gpa + __LC_CPU_TIMER_SAVE_AREA,
5250 			      &cputm, 8);
5251 	clkcomp = vcpu->arch.sie_block->ckc >> 8;
5252 	rc |= write_guest_abs(vcpu, gpa + __LC_CLOCK_COMP_SAVE_AREA,
5253 			      &clkcomp, 8);
5254 	rc |= write_guest_abs(vcpu, gpa + __LC_AREGS_SAVE_AREA,
5255 			      &vcpu->run->s.regs.acrs, 64);
5256 	rc |= write_guest_abs(vcpu, gpa + __LC_CREGS_SAVE_AREA,
5257 			      &vcpu->arch.sie_block->gcr, 128);
5258 	return rc ? -EFAULT : 0;
5259 }
5260 
5261 int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
5262 {
5263 	/*
5264 	 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
5265 	 * switch in the run ioctl. Let's update our copies before we save
5266 	 * it into the save area
5267 	 */
5268 	kvm_s390_fpu_store(vcpu->run);
5269 	save_access_regs(vcpu->run->s.regs.acrs);
5270 
5271 	return kvm_s390_store_status_unloaded(vcpu, addr);
5272 }
5273 
5274 static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
5275 {
5276 	kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu);
5277 	kvm_s390_sync_request(KVM_REQ_DISABLE_IBS, vcpu);
5278 }
5279 
5280 static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
5281 {
5282 	unsigned long i;
5283 	struct kvm_vcpu *vcpu;
5284 
5285 	kvm_for_each_vcpu(i, vcpu, kvm) {
5286 		__disable_ibs_on_vcpu(vcpu);
5287 	}
5288 }
5289 
5290 static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
5291 {
5292 	if (!sclp.has_ibs)
5293 		return;
5294 	kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu);
5295 	kvm_s390_sync_request(KVM_REQ_ENABLE_IBS, vcpu);
5296 }
5297 
5298 int kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
5299 {
5300 	int i, online_vcpus, r = 0, started_vcpus = 0;
5301 
5302 	if (!is_vcpu_stopped(vcpu))
5303 		return 0;
5304 
5305 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
5306 	/* Only one cpu at a time may enter/leave the STOPPED state. */
5307 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
5308 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
5309 
5310 	/* Let's tell the UV that we want to change into the operating state */
5311 	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
5312 		r = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_OPR);
5313 		if (r) {
5314 			spin_unlock(&vcpu->kvm->arch.start_stop_lock);
5315 			return r;
5316 		}
5317 	}
5318 
5319 	for (i = 0; i < online_vcpus; i++) {
5320 		if (!is_vcpu_stopped(kvm_get_vcpu(vcpu->kvm, i)))
5321 			started_vcpus++;
5322 	}
5323 
5324 	if (started_vcpus == 0) {
5325 		/* we're the only active VCPU -> speed it up */
5326 		__enable_ibs_on_vcpu(vcpu);
5327 	} else if (started_vcpus == 1) {
5328 		/*
5329 		 * As we are starting a second VCPU, we have to disable
5330 		 * the IBS facility on all VCPUs to remove potentially
5331 		 * outstanding ENABLE requests.
5332 		 */
5333 		__disable_ibs_on_all_vcpus(vcpu->kvm);
5334 	}
5335 
5336 	kvm_s390_clear_cpuflags(vcpu, CPUSTAT_STOPPED);
5337 	/*
5338 	 * The real PSW might have changed due to a RESTART interpreted by the
5339 	 * ultravisor. We block all interrupts and let the next sie exit
5340 	 * refresh our view.
5341 	 */
5342 	if (kvm_s390_pv_cpu_is_protected(vcpu))
5343 		vcpu->arch.sie_block->gpsw.mask &= ~PSW_INT_MASK;
5344 	/*
5345 	 * Another VCPU might have used IBS while we were offline.
5346 	 * Let's play safe and flush the VCPU at startup.
5347 	 */
5348 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
5349 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
5350 	return 0;
5351 }
5352 
5353 int kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
5354 {
5355 	int i, online_vcpus, r = 0, started_vcpus = 0;
5356 	struct kvm_vcpu *started_vcpu = NULL;
5357 
5358 	if (is_vcpu_stopped(vcpu))
5359 		return 0;
5360 
5361 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
5362 	/* Only one cpu at a time may enter/leave the STOPPED state. */
5363 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
5364 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
5365 
5366 	/* Let's tell the UV that we want to change into the stopped state */
5367 	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
5368 		r = kvm_s390_pv_set_cpu_state(vcpu, PV_CPU_STATE_STP);
5369 		if (r) {
5370 			spin_unlock(&vcpu->kvm->arch.start_stop_lock);
5371 			return r;
5372 		}
5373 	}
5374 
5375 	/*
5376 	 * Set the VCPU to STOPPED and THEN clear the interrupt flag,
5377 	 * now that the SIGP STOP and SIGP STOP AND STORE STATUS orders
5378 	 * have been fully processed. This will ensure that the VCPU
5379 	 * is kept BUSY if another VCPU is inquiring with SIGP SENSE.
5380 	 */
5381 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOPPED);
5382 	kvm_s390_clear_stop_irq(vcpu);
5383 
5384 	__disable_ibs_on_vcpu(vcpu);
5385 
5386 	for (i = 0; i < online_vcpus; i++) {
5387 		struct kvm_vcpu *tmp = kvm_get_vcpu(vcpu->kvm, i);
5388 
5389 		if (!is_vcpu_stopped(tmp)) {
5390 			started_vcpus++;
5391 			started_vcpu = tmp;
5392 		}
5393 	}
5394 
5395 	if (started_vcpus == 1) {
5396 		/*
5397 		 * As we only have one VCPU left, we want to enable the
5398 		 * IBS facility for that VCPU to speed it up.
5399 		 */
5400 		__enable_ibs_on_vcpu(started_vcpu);
5401 	}
5402 
5403 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
5404 	return 0;
5405 }
5406 
5407 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
5408 				     struct kvm_enable_cap *cap)
5409 {
5410 	int r;
5411 
5412 	if (cap->flags)
5413 		return -EINVAL;
5414 
5415 	switch (cap->cap) {
5416 	case KVM_CAP_S390_CSS_SUPPORT:
5417 		if (!vcpu->kvm->arch.css_support) {
5418 			vcpu->kvm->arch.css_support = 1;
5419 			VM_EVENT(vcpu->kvm, 3, "%s", "ENABLE: CSS support");
5420 			trace_kvm_s390_enable_css(vcpu->kvm);
5421 		}
5422 		r = 0;
5423 		break;
5424 	default:
5425 		r = -EINVAL;
5426 		break;
5427 	}
5428 	return r;
5429 }
5430 
5431 static long kvm_s390_vcpu_sida_op(struct kvm_vcpu *vcpu,
5432 				  struct kvm_s390_mem_op *mop)
5433 {
5434 	void __user *uaddr = (void __user *)mop->buf;
5435 	void *sida_addr;
5436 	int r = 0;
5437 
5438 	if (mop->flags || !mop->size)
5439 		return -EINVAL;
5440 	if (mop->size + mop->sida_offset < mop->size)
5441 		return -EINVAL;
5442 	if (mop->size + mop->sida_offset > sida_size(vcpu->arch.sie_block))
5443 		return -E2BIG;
5444 	if (!kvm_s390_pv_cpu_is_protected(vcpu))
5445 		return -EINVAL;
5446 
5447 	sida_addr = (char *)sida_addr(vcpu->arch.sie_block) + mop->sida_offset;
5448 
5449 	switch (mop->op) {
5450 	case KVM_S390_MEMOP_SIDA_READ:
5451 		if (copy_to_user(uaddr, sida_addr, mop->size))
5452 			r = -EFAULT;
5453 
5454 		break;
5455 	case KVM_S390_MEMOP_SIDA_WRITE:
5456 		if (copy_from_user(sida_addr, uaddr, mop->size))
5457 			r = -EFAULT;
5458 		break;
5459 	}
5460 	return r;
5461 }
5462 
5463 static long kvm_s390_vcpu_mem_op(struct kvm_vcpu *vcpu,
5464 				 struct kvm_s390_mem_op *mop)
5465 {
5466 	void __user *uaddr = (void __user *)mop->buf;
5467 	enum gacc_mode acc_mode;
5468 	void *tmpbuf = NULL;
5469 	int r;
5470 
5471 	r = mem_op_validate_common(mop, KVM_S390_MEMOP_F_INJECT_EXCEPTION |
5472 					KVM_S390_MEMOP_F_CHECK_ONLY |
5473 					KVM_S390_MEMOP_F_SKEY_PROTECTION);
5474 	if (r)
5475 		return r;
5476 	if (mop->ar >= NUM_ACRS)
5477 		return -EINVAL;
5478 	if (kvm_s390_pv_cpu_is_protected(vcpu))
5479 		return -EINVAL;
5480 	if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
5481 		tmpbuf = vmalloc(mop->size);
5482 		if (!tmpbuf)
5483 			return -ENOMEM;
5484 	}
5485 
5486 	acc_mode = mop->op == KVM_S390_MEMOP_LOGICAL_READ ? GACC_FETCH : GACC_STORE;
5487 	if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
5488 		r = check_gva_range(vcpu, mop->gaddr, mop->ar, mop->size,
5489 				    acc_mode, mop->key);
5490 		goto out_inject;
5491 	}
5492 	if (acc_mode == GACC_FETCH) {
5493 		r = read_guest_with_key(vcpu, mop->gaddr, mop->ar, tmpbuf,
5494 					mop->size, mop->key);
5495 		if (r)
5496 			goto out_inject;
5497 		if (copy_to_user(uaddr, tmpbuf, mop->size)) {
5498 			r = -EFAULT;
5499 			goto out_free;
5500 		}
5501 	} else {
5502 		if (copy_from_user(tmpbuf, uaddr, mop->size)) {
5503 			r = -EFAULT;
5504 			goto out_free;
5505 		}
5506 		r = write_guest_with_key(vcpu, mop->gaddr, mop->ar, tmpbuf,
5507 					 mop->size, mop->key);
5508 	}
5509 
5510 out_inject:
5511 	if (r > 0 && (mop->flags & KVM_S390_MEMOP_F_INJECT_EXCEPTION) != 0)
5512 		kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
5513 
5514 out_free:
5515 	vfree(tmpbuf);
5516 	return r;
5517 }
5518 
5519 static long kvm_s390_vcpu_memsida_op(struct kvm_vcpu *vcpu,
5520 				     struct kvm_s390_mem_op *mop)
5521 {
5522 	int r, srcu_idx;
5523 
5524 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
5525 
5526 	switch (mop->op) {
5527 	case KVM_S390_MEMOP_LOGICAL_READ:
5528 	case KVM_S390_MEMOP_LOGICAL_WRITE:
5529 		r = kvm_s390_vcpu_mem_op(vcpu, mop);
5530 		break;
5531 	case KVM_S390_MEMOP_SIDA_READ:
5532 	case KVM_S390_MEMOP_SIDA_WRITE:
5533 		/* we are locked against sida going away by the vcpu->mutex */
5534 		r = kvm_s390_vcpu_sida_op(vcpu, mop);
5535 		break;
5536 	default:
5537 		r = -EINVAL;
5538 	}
5539 
5540 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
5541 	return r;
5542 }
5543 
5544 long kvm_arch_vcpu_async_ioctl(struct file *filp,
5545 			       unsigned int ioctl, unsigned long arg)
5546 {
5547 	struct kvm_vcpu *vcpu = filp->private_data;
5548 	void __user *argp = (void __user *)arg;
5549 	int rc;
5550 
5551 	switch (ioctl) {
5552 	case KVM_S390_IRQ: {
5553 		struct kvm_s390_irq s390irq;
5554 
5555 		if (copy_from_user(&s390irq, argp, sizeof(s390irq)))
5556 			return -EFAULT;
5557 		rc = kvm_s390_inject_vcpu(vcpu, &s390irq);
5558 		break;
5559 	}
5560 	case KVM_S390_INTERRUPT: {
5561 		struct kvm_s390_interrupt s390int;
5562 		struct kvm_s390_irq s390irq = {};
5563 
5564 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
5565 			return -EFAULT;
5566 		if (s390int_to_s390irq(&s390int, &s390irq))
5567 			return -EINVAL;
5568 		rc = kvm_s390_inject_vcpu(vcpu, &s390irq);
5569 		break;
5570 	}
5571 	default:
5572 		rc = -ENOIOCTLCMD;
5573 		break;
5574 	}
5575 
5576 	/*
5577 	 * To simplify single stepping of userspace-emulated instructions,
5578 	 * KVM_EXIT_S390_SIEIC exit sets KVM_GUESTDBG_EXIT_PENDING (see
5579 	 * should_handle_per_ifetch()). However, if userspace emulation injects
5580 	 * an interrupt, it needs to be cleared, so that KVM_EXIT_DEBUG happens
5581 	 * after (and not before) the interrupt delivery.
5582 	 */
5583 	if (!rc)
5584 		vcpu->guest_debug &= ~KVM_GUESTDBG_EXIT_PENDING;
5585 
5586 	return rc;
5587 }
5588 
5589 static int kvm_s390_handle_pv_vcpu_dump(struct kvm_vcpu *vcpu,
5590 					struct kvm_pv_cmd *cmd)
5591 {
5592 	struct kvm_s390_pv_dmp dmp;
5593 	void *data;
5594 	int ret;
5595 
5596 	/* Dump initialization is a prerequisite */
5597 	if (!vcpu->kvm->arch.pv.dumping)
5598 		return -EINVAL;
5599 
5600 	if (copy_from_user(&dmp, (__u8 __user *)cmd->data, sizeof(dmp)))
5601 		return -EFAULT;
5602 
5603 	/* We only handle this subcmd right now */
5604 	if (dmp.subcmd != KVM_PV_DUMP_CPU)
5605 		return -EINVAL;
5606 
5607 	/* CPU dump length is the same as create cpu storage donation. */
5608 	if (dmp.buff_len != uv_info.guest_cpu_stor_len)
5609 		return -EINVAL;
5610 
5611 	data = kvzalloc(uv_info.guest_cpu_stor_len, GFP_KERNEL);
5612 	if (!data)
5613 		return -ENOMEM;
5614 
5615 	ret = kvm_s390_pv_dump_cpu(vcpu, data, &cmd->rc, &cmd->rrc);
5616 
5617 	VCPU_EVENT(vcpu, 3, "PROTVIRT DUMP CPU %d rc %x rrc %x",
5618 		   vcpu->vcpu_id, cmd->rc, cmd->rrc);
5619 
5620 	if (ret)
5621 		ret = -EINVAL;
5622 
5623 	/* On success copy over the dump data */
5624 	if (!ret && copy_to_user((__u8 __user *)dmp.buff_addr, data, uv_info.guest_cpu_stor_len))
5625 		ret = -EFAULT;
5626 
5627 	kvfree(data);
5628 	return ret;
5629 }
5630 
5631 long kvm_arch_vcpu_ioctl(struct file *filp,
5632 			 unsigned int ioctl, unsigned long arg)
5633 {
5634 	struct kvm_vcpu *vcpu = filp->private_data;
5635 	void __user *argp = (void __user *)arg;
5636 	int idx;
5637 	long r;
5638 	u16 rc, rrc;
5639 
5640 	vcpu_load(vcpu);
5641 
5642 	switch (ioctl) {
5643 	case KVM_S390_STORE_STATUS:
5644 		idx = srcu_read_lock(&vcpu->kvm->srcu);
5645 		r = kvm_s390_store_status_unloaded(vcpu, arg);
5646 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
5647 		break;
5648 	case KVM_S390_SET_INITIAL_PSW: {
5649 		psw_t psw;
5650 
5651 		r = -EFAULT;
5652 		if (copy_from_user(&psw, argp, sizeof(psw)))
5653 			break;
5654 		r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw);
5655 		break;
5656 	}
5657 	case KVM_S390_CLEAR_RESET:
5658 		r = 0;
5659 		kvm_arch_vcpu_ioctl_clear_reset(vcpu);
5660 		if (kvm_s390_pv_cpu_is_protected(vcpu)) {
5661 			r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
5662 					  UVC_CMD_CPU_RESET_CLEAR, &rc, &rrc);
5663 			VCPU_EVENT(vcpu, 3, "PROTVIRT RESET CLEAR VCPU: rc %x rrc %x",
5664 				   rc, rrc);
5665 		}
5666 		break;
5667 	case KVM_S390_INITIAL_RESET:
5668 		r = 0;
5669 		kvm_arch_vcpu_ioctl_initial_reset(vcpu);
5670 		if (kvm_s390_pv_cpu_is_protected(vcpu)) {
5671 			r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
5672 					  UVC_CMD_CPU_RESET_INITIAL,
5673 					  &rc, &rrc);
5674 			VCPU_EVENT(vcpu, 3, "PROTVIRT RESET INITIAL VCPU: rc %x rrc %x",
5675 				   rc, rrc);
5676 		}
5677 		break;
5678 	case KVM_S390_NORMAL_RESET:
5679 		r = 0;
5680 		kvm_arch_vcpu_ioctl_normal_reset(vcpu);
5681 		if (kvm_s390_pv_cpu_is_protected(vcpu)) {
5682 			r = uv_cmd_nodata(kvm_s390_pv_cpu_get_handle(vcpu),
5683 					  UVC_CMD_CPU_RESET, &rc, &rrc);
5684 			VCPU_EVENT(vcpu, 3, "PROTVIRT RESET NORMAL VCPU: rc %x rrc %x",
5685 				   rc, rrc);
5686 		}
5687 		break;
5688 	case KVM_SET_ONE_REG:
5689 	case KVM_GET_ONE_REG: {
5690 		struct kvm_one_reg reg;
5691 		r = -EINVAL;
5692 		if (kvm_s390_pv_cpu_is_protected(vcpu))
5693 			break;
5694 		r = -EFAULT;
5695 		if (copy_from_user(&reg, argp, sizeof(reg)))
5696 			break;
5697 		if (ioctl == KVM_SET_ONE_REG)
5698 			r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, &reg);
5699 		else
5700 			r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, &reg);
5701 		break;
5702 	}
5703 #ifdef CONFIG_KVM_S390_UCONTROL
5704 	case KVM_S390_UCAS_MAP: {
5705 		struct kvm_s390_ucas_mapping ucasmap;
5706 
5707 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
5708 			r = -EFAULT;
5709 			break;
5710 		}
5711 
5712 		if (!kvm_is_ucontrol(vcpu->kvm)) {
5713 			r = -EINVAL;
5714 			break;
5715 		}
5716 
5717 		r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr,
5718 				     ucasmap.vcpu_addr, ucasmap.length);
5719 		break;
5720 	}
5721 	case KVM_S390_UCAS_UNMAP: {
5722 		struct kvm_s390_ucas_mapping ucasmap;
5723 
5724 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
5725 			r = -EFAULT;
5726 			break;
5727 		}
5728 
5729 		if (!kvm_is_ucontrol(vcpu->kvm)) {
5730 			r = -EINVAL;
5731 			break;
5732 		}
5733 
5734 		r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr,
5735 			ucasmap.length);
5736 		break;
5737 	}
5738 #endif
5739 	case KVM_S390_VCPU_FAULT: {
5740 		r = gmap_fault(vcpu->arch.gmap, arg, 0);
5741 		break;
5742 	}
5743 	case KVM_ENABLE_CAP:
5744 	{
5745 		struct kvm_enable_cap cap;
5746 		r = -EFAULT;
5747 		if (copy_from_user(&cap, argp, sizeof(cap)))
5748 			break;
5749 		r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
5750 		break;
5751 	}
5752 	case KVM_S390_MEM_OP: {
5753 		struct kvm_s390_mem_op mem_op;
5754 
5755 		if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
5756 			r = kvm_s390_vcpu_memsida_op(vcpu, &mem_op);
5757 		else
5758 			r = -EFAULT;
5759 		break;
5760 	}
5761 	case KVM_S390_SET_IRQ_STATE: {
5762 		struct kvm_s390_irq_state irq_state;
5763 
5764 		r = -EFAULT;
5765 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
5766 			break;
5767 		if (irq_state.len > VCPU_IRQS_MAX_BUF ||
5768 		    irq_state.len == 0 ||
5769 		    irq_state.len % sizeof(struct kvm_s390_irq) > 0) {
5770 			r = -EINVAL;
5771 			break;
5772 		}
5773 		/* do not use irq_state.flags, it will break old QEMUs */
5774 		r = kvm_s390_set_irq_state(vcpu,
5775 					   (void __user *) irq_state.buf,
5776 					   irq_state.len);
5777 		break;
5778 	}
5779 	case KVM_S390_GET_IRQ_STATE: {
5780 		struct kvm_s390_irq_state irq_state;
5781 
5782 		r = -EFAULT;
5783 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
5784 			break;
5785 		if (irq_state.len == 0) {
5786 			r = -EINVAL;
5787 			break;
5788 		}
5789 		/* do not use irq_state.flags, it will break old QEMUs */
5790 		r = kvm_s390_get_irq_state(vcpu,
5791 					   (__u8 __user *)  irq_state.buf,
5792 					   irq_state.len);
5793 		break;
5794 	}
5795 	case KVM_S390_PV_CPU_COMMAND: {
5796 		struct kvm_pv_cmd cmd;
5797 
5798 		r = -EINVAL;
5799 		if (!is_prot_virt_host())
5800 			break;
5801 
5802 		r = -EFAULT;
5803 		if (copy_from_user(&cmd, argp, sizeof(cmd)))
5804 			break;
5805 
5806 		r = -EINVAL;
5807 		if (cmd.flags)
5808 			break;
5809 
5810 		/* We only handle this cmd right now */
5811 		if (cmd.cmd != KVM_PV_DUMP)
5812 			break;
5813 
5814 		r = kvm_s390_handle_pv_vcpu_dump(vcpu, &cmd);
5815 
5816 		/* Always copy over UV rc / rrc data */
5817 		if (copy_to_user((__u8 __user *)argp, &cmd.rc,
5818 				 sizeof(cmd.rc) + sizeof(cmd.rrc)))
5819 			r = -EFAULT;
5820 		break;
5821 	}
5822 	default:
5823 		r = -ENOTTY;
5824 	}
5825 
5826 	vcpu_put(vcpu);
5827 	return r;
5828 }
5829 
5830 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
5831 {
5832 #ifdef CONFIG_KVM_S390_UCONTROL
5833 	if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET)
5834 		 && (kvm_is_ucontrol(vcpu->kvm))) {
5835 		vmf->page = virt_to_page(vcpu->arch.sie_block);
5836 		get_page(vmf->page);
5837 		return 0;
5838 	}
5839 #endif
5840 	return VM_FAULT_SIGBUS;
5841 }
5842 
5843 bool kvm_arch_irqchip_in_kernel(struct kvm *kvm)
5844 {
5845 	return true;
5846 }
5847 
5848 /* Section: memory related */
5849 int kvm_arch_prepare_memory_region(struct kvm *kvm,
5850 				   const struct kvm_memory_slot *old,
5851 				   struct kvm_memory_slot *new,
5852 				   enum kvm_mr_change change)
5853 {
5854 	gpa_t size;
5855 
5856 	if (kvm_is_ucontrol(kvm))
5857 		return -EINVAL;
5858 
5859 	/* When we are protected, we should not change the memory slots */
5860 	if (kvm_s390_pv_get_handle(kvm))
5861 		return -EINVAL;
5862 
5863 	if (change != KVM_MR_DELETE && change != KVM_MR_FLAGS_ONLY) {
5864 		/*
5865 		 * A few sanity checks. We can have memory slots which have to be
5866 		 * located/ended at a segment boundary (1MB). The memory in userland is
5867 		 * ok to be fragmented into various different vmas. It is okay to mmap()
5868 		 * and munmap() stuff in this slot after doing this call at any time
5869 		 */
5870 
5871 		if (new->userspace_addr & 0xffffful)
5872 			return -EINVAL;
5873 
5874 		size = new->npages * PAGE_SIZE;
5875 		if (size & 0xffffful)
5876 			return -EINVAL;
5877 
5878 		if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
5879 			return -EINVAL;
5880 	}
5881 
5882 	if (!kvm->arch.migration_mode)
5883 		return 0;
5884 
5885 	/*
5886 	 * Turn off migration mode when:
5887 	 * - userspace creates a new memslot with dirty logging off,
5888 	 * - userspace modifies an existing memslot (MOVE or FLAGS_ONLY) and
5889 	 *   dirty logging is turned off.
5890 	 * Migration mode expects dirty page logging being enabled to store
5891 	 * its dirty bitmap.
5892 	 */
5893 	if (change != KVM_MR_DELETE &&
5894 	    !(new->flags & KVM_MEM_LOG_DIRTY_PAGES))
5895 		WARN(kvm_s390_vm_stop_migration(kvm),
5896 		     "Failed to stop migration mode");
5897 
5898 	return 0;
5899 }
5900 
5901 void kvm_arch_commit_memory_region(struct kvm *kvm,
5902 				struct kvm_memory_slot *old,
5903 				const struct kvm_memory_slot *new,
5904 				enum kvm_mr_change change)
5905 {
5906 	int rc = 0;
5907 
5908 	switch (change) {
5909 	case KVM_MR_DELETE:
5910 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
5911 					old->npages * PAGE_SIZE);
5912 		break;
5913 	case KVM_MR_MOVE:
5914 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
5915 					old->npages * PAGE_SIZE);
5916 		if (rc)
5917 			break;
5918 		fallthrough;
5919 	case KVM_MR_CREATE:
5920 		rc = gmap_map_segment(kvm->arch.gmap, new->userspace_addr,
5921 				      new->base_gfn * PAGE_SIZE,
5922 				      new->npages * PAGE_SIZE);
5923 		break;
5924 	case KVM_MR_FLAGS_ONLY:
5925 		break;
5926 	default:
5927 		WARN(1, "Unknown KVM MR CHANGE: %d\n", change);
5928 	}
5929 	if (rc)
5930 		pr_warn("failed to commit memory region\n");
5931 	return;
5932 }
5933 
5934 static inline unsigned long nonhyp_mask(int i)
5935 {
5936 	unsigned int nonhyp_fai = (sclp.hmfai << i * 2) >> 30;
5937 
5938 	return 0x0000ffffffffffffUL >> (nonhyp_fai << 4);
5939 }
5940 
5941 static int __init kvm_s390_init(void)
5942 {
5943 	int i, r;
5944 
5945 	if (!sclp.has_sief2) {
5946 		pr_info("SIE is not available\n");
5947 		return -ENODEV;
5948 	}
5949 
5950 	if (nested && hpage) {
5951 		pr_info("A KVM host that supports nesting cannot back its KVM guests with huge pages\n");
5952 		return -EINVAL;
5953 	}
5954 
5955 	for (i = 0; i < 16; i++)
5956 		kvm_s390_fac_base[i] |=
5957 			stfle_fac_list[i] & nonhyp_mask(i);
5958 
5959 	r = __kvm_s390_init();
5960 	if (r)
5961 		return r;
5962 
5963 	r = kvm_init(sizeof(struct kvm_vcpu), 0, THIS_MODULE);
5964 	if (r) {
5965 		__kvm_s390_exit();
5966 		return r;
5967 	}
5968 	return 0;
5969 }
5970 
5971 static void __exit kvm_s390_exit(void)
5972 {
5973 	kvm_exit();
5974 
5975 	__kvm_s390_exit();
5976 }
5977 
5978 module_init(kvm_s390_init);
5979 module_exit(kvm_s390_exit);
5980 
5981 /*
5982  * Enable autoloading of the kvm module.
5983  * Note that we add the module alias here instead of virt/kvm/kvm_main.c
5984  * since x86 takes a different approach.
5985  */
5986 #include <linux/miscdevice.h>
5987 MODULE_ALIAS_MISCDEV(KVM_MINOR);
5988 MODULE_ALIAS("devname:kvm");
5989