xref: /openbmc/linux/arch/s390/kvm/kvm-s390.c (revision 9b5db89e)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * hosting IBM Z kernel virtual machines (s390x)
4  *
5  * Copyright IBM Corp. 2008, 2018
6  *
7  *    Author(s): Carsten Otte <cotte@de.ibm.com>
8  *               Christian Borntraeger <borntraeger@de.ibm.com>
9  *               Heiko Carstens <heiko.carstens@de.ibm.com>
10  *               Christian Ehrhardt <ehrhardt@de.ibm.com>
11  *               Jason J. Herne <jjherne@us.ibm.com>
12  */
13 
14 #define KMSG_COMPONENT "kvm-s390"
15 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
16 
17 #include <linux/compiler.h>
18 #include <linux/err.h>
19 #include <linux/fs.h>
20 #include <linux/hrtimer.h>
21 #include <linux/init.h>
22 #include <linux/kvm.h>
23 #include <linux/kvm_host.h>
24 #include <linux/mman.h>
25 #include <linux/module.h>
26 #include <linux/moduleparam.h>
27 #include <linux/random.h>
28 #include <linux/slab.h>
29 #include <linux/timer.h>
30 #include <linux/vmalloc.h>
31 #include <linux/bitmap.h>
32 #include <linux/sched/signal.h>
33 #include <linux/string.h>
34 
35 #include <asm/asm-offsets.h>
36 #include <asm/lowcore.h>
37 #include <asm/stp.h>
38 #include <asm/pgtable.h>
39 #include <asm/gmap.h>
40 #include <asm/nmi.h>
41 #include <asm/switch_to.h>
42 #include <asm/isc.h>
43 #include <asm/sclp.h>
44 #include <asm/cpacf.h>
45 #include <asm/timex.h>
46 #include <asm/ap.h>
47 #include "kvm-s390.h"
48 #include "gaccess.h"
49 
50 #define CREATE_TRACE_POINTS
51 #include "trace.h"
52 #include "trace-s390.h"
53 
54 #define MEM_OP_MAX_SIZE 65536	/* Maximum transfer size for KVM_S390_MEM_OP */
55 #define LOCAL_IRQS 32
56 #define VCPU_IRQS_MAX_BUF (sizeof(struct kvm_s390_irq) * \
57 			   (KVM_MAX_VCPUS + LOCAL_IRQS))
58 
59 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
60 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
61 
62 struct kvm_stats_debugfs_item debugfs_entries[] = {
63 	{ "userspace_handled", VCPU_STAT(exit_userspace) },
64 	{ "exit_null", VCPU_STAT(exit_null) },
65 	{ "exit_validity", VCPU_STAT(exit_validity) },
66 	{ "exit_stop_request", VCPU_STAT(exit_stop_request) },
67 	{ "exit_external_request", VCPU_STAT(exit_external_request) },
68 	{ "exit_io_request", VCPU_STAT(exit_io_request) },
69 	{ "exit_external_interrupt", VCPU_STAT(exit_external_interrupt) },
70 	{ "exit_instruction", VCPU_STAT(exit_instruction) },
71 	{ "exit_pei", VCPU_STAT(exit_pei) },
72 	{ "exit_program_interruption", VCPU_STAT(exit_program_interruption) },
73 	{ "exit_instr_and_program_int", VCPU_STAT(exit_instr_and_program) },
74 	{ "exit_operation_exception", VCPU_STAT(exit_operation_exception) },
75 	{ "halt_successful_poll", VCPU_STAT(halt_successful_poll) },
76 	{ "halt_attempted_poll", VCPU_STAT(halt_attempted_poll) },
77 	{ "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
78 	{ "halt_no_poll_steal", VCPU_STAT(halt_no_poll_steal) },
79 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
80 	{ "instruction_lctlg", VCPU_STAT(instruction_lctlg) },
81 	{ "instruction_lctl", VCPU_STAT(instruction_lctl) },
82 	{ "instruction_stctl", VCPU_STAT(instruction_stctl) },
83 	{ "instruction_stctg", VCPU_STAT(instruction_stctg) },
84 	{ "deliver_ckc", VCPU_STAT(deliver_ckc) },
85 	{ "deliver_cputm", VCPU_STAT(deliver_cputm) },
86 	{ "deliver_emergency_signal", VCPU_STAT(deliver_emergency_signal) },
87 	{ "deliver_external_call", VCPU_STAT(deliver_external_call) },
88 	{ "deliver_service_signal", VCPU_STAT(deliver_service_signal) },
89 	{ "deliver_virtio", VCPU_STAT(deliver_virtio) },
90 	{ "deliver_stop_signal", VCPU_STAT(deliver_stop_signal) },
91 	{ "deliver_prefix_signal", VCPU_STAT(deliver_prefix_signal) },
92 	{ "deliver_restart_signal", VCPU_STAT(deliver_restart_signal) },
93 	{ "deliver_program", VCPU_STAT(deliver_program) },
94 	{ "deliver_io", VCPU_STAT(deliver_io) },
95 	{ "deliver_machine_check", VCPU_STAT(deliver_machine_check) },
96 	{ "exit_wait_state", VCPU_STAT(exit_wait_state) },
97 	{ "inject_ckc", VCPU_STAT(inject_ckc) },
98 	{ "inject_cputm", VCPU_STAT(inject_cputm) },
99 	{ "inject_external_call", VCPU_STAT(inject_external_call) },
100 	{ "inject_float_mchk", VM_STAT(inject_float_mchk) },
101 	{ "inject_emergency_signal", VCPU_STAT(inject_emergency_signal) },
102 	{ "inject_io", VM_STAT(inject_io) },
103 	{ "inject_mchk", VCPU_STAT(inject_mchk) },
104 	{ "inject_pfault_done", VM_STAT(inject_pfault_done) },
105 	{ "inject_program", VCPU_STAT(inject_program) },
106 	{ "inject_restart", VCPU_STAT(inject_restart) },
107 	{ "inject_service_signal", VM_STAT(inject_service_signal) },
108 	{ "inject_set_prefix", VCPU_STAT(inject_set_prefix) },
109 	{ "inject_stop_signal", VCPU_STAT(inject_stop_signal) },
110 	{ "inject_pfault_init", VCPU_STAT(inject_pfault_init) },
111 	{ "inject_virtio", VM_STAT(inject_virtio) },
112 	{ "instruction_epsw", VCPU_STAT(instruction_epsw) },
113 	{ "instruction_gs", VCPU_STAT(instruction_gs) },
114 	{ "instruction_io_other", VCPU_STAT(instruction_io_other) },
115 	{ "instruction_lpsw", VCPU_STAT(instruction_lpsw) },
116 	{ "instruction_lpswe", VCPU_STAT(instruction_lpswe) },
117 	{ "instruction_pfmf", VCPU_STAT(instruction_pfmf) },
118 	{ "instruction_ptff", VCPU_STAT(instruction_ptff) },
119 	{ "instruction_stidp", VCPU_STAT(instruction_stidp) },
120 	{ "instruction_sck", VCPU_STAT(instruction_sck) },
121 	{ "instruction_sckpf", VCPU_STAT(instruction_sckpf) },
122 	{ "instruction_spx", VCPU_STAT(instruction_spx) },
123 	{ "instruction_stpx", VCPU_STAT(instruction_stpx) },
124 	{ "instruction_stap", VCPU_STAT(instruction_stap) },
125 	{ "instruction_iske", VCPU_STAT(instruction_iske) },
126 	{ "instruction_ri", VCPU_STAT(instruction_ri) },
127 	{ "instruction_rrbe", VCPU_STAT(instruction_rrbe) },
128 	{ "instruction_sske", VCPU_STAT(instruction_sske) },
129 	{ "instruction_ipte_interlock", VCPU_STAT(instruction_ipte_interlock) },
130 	{ "instruction_essa", VCPU_STAT(instruction_essa) },
131 	{ "instruction_stsi", VCPU_STAT(instruction_stsi) },
132 	{ "instruction_stfl", VCPU_STAT(instruction_stfl) },
133 	{ "instruction_tb", VCPU_STAT(instruction_tb) },
134 	{ "instruction_tpi", VCPU_STAT(instruction_tpi) },
135 	{ "instruction_tprot", VCPU_STAT(instruction_tprot) },
136 	{ "instruction_tsch", VCPU_STAT(instruction_tsch) },
137 	{ "instruction_sthyi", VCPU_STAT(instruction_sthyi) },
138 	{ "instruction_sie", VCPU_STAT(instruction_sie) },
139 	{ "instruction_sigp_sense", VCPU_STAT(instruction_sigp_sense) },
140 	{ "instruction_sigp_sense_running", VCPU_STAT(instruction_sigp_sense_running) },
141 	{ "instruction_sigp_external_call", VCPU_STAT(instruction_sigp_external_call) },
142 	{ "instruction_sigp_emergency", VCPU_STAT(instruction_sigp_emergency) },
143 	{ "instruction_sigp_cond_emergency", VCPU_STAT(instruction_sigp_cond_emergency) },
144 	{ "instruction_sigp_start", VCPU_STAT(instruction_sigp_start) },
145 	{ "instruction_sigp_stop", VCPU_STAT(instruction_sigp_stop) },
146 	{ "instruction_sigp_stop_store_status", VCPU_STAT(instruction_sigp_stop_store_status) },
147 	{ "instruction_sigp_store_status", VCPU_STAT(instruction_sigp_store_status) },
148 	{ "instruction_sigp_store_adtl_status", VCPU_STAT(instruction_sigp_store_adtl_status) },
149 	{ "instruction_sigp_set_arch", VCPU_STAT(instruction_sigp_arch) },
150 	{ "instruction_sigp_set_prefix", VCPU_STAT(instruction_sigp_prefix) },
151 	{ "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart) },
152 	{ "instruction_sigp_cpu_reset", VCPU_STAT(instruction_sigp_cpu_reset) },
153 	{ "instruction_sigp_init_cpu_reset", VCPU_STAT(instruction_sigp_init_cpu_reset) },
154 	{ "instruction_sigp_unknown", VCPU_STAT(instruction_sigp_unknown) },
155 	{ "instruction_diag_10", VCPU_STAT(diagnose_10) },
156 	{ "instruction_diag_44", VCPU_STAT(diagnose_44) },
157 	{ "instruction_diag_9c", VCPU_STAT(diagnose_9c) },
158 	{ "instruction_diag_258", VCPU_STAT(diagnose_258) },
159 	{ "instruction_diag_308", VCPU_STAT(diagnose_308) },
160 	{ "instruction_diag_500", VCPU_STAT(diagnose_500) },
161 	{ "instruction_diag_other", VCPU_STAT(diagnose_other) },
162 	{ NULL }
163 };
164 
165 struct kvm_s390_tod_clock_ext {
166 	__u8 epoch_idx;
167 	__u64 tod;
168 	__u8 reserved[7];
169 } __packed;
170 
171 /* allow nested virtualization in KVM (if enabled by user space) */
172 static int nested;
173 module_param(nested, int, S_IRUGO);
174 MODULE_PARM_DESC(nested, "Nested virtualization support");
175 
176 /* allow 1m huge page guest backing, if !nested */
177 static int hpage;
178 module_param(hpage, int, 0444);
179 MODULE_PARM_DESC(hpage, "1m huge page backing support");
180 
181 /* maximum percentage of steal time for polling.  >100 is treated like 100 */
182 static u8 halt_poll_max_steal = 10;
183 module_param(halt_poll_max_steal, byte, 0644);
184 MODULE_PARM_DESC(halt_poll_max_steal, "Maximum percentage of steal time to allow polling");
185 
186 /*
187  * For now we handle at most 16 double words as this is what the s390 base
188  * kernel handles and stores in the prefix page. If we ever need to go beyond
189  * this, this requires changes to code, but the external uapi can stay.
190  */
191 #define SIZE_INTERNAL 16
192 
193 /*
194  * Base feature mask that defines default mask for facilities. Consists of the
195  * defines in FACILITIES_KVM and the non-hypervisor managed bits.
196  */
197 static unsigned long kvm_s390_fac_base[SIZE_INTERNAL] = { FACILITIES_KVM };
198 /*
199  * Extended feature mask. Consists of the defines in FACILITIES_KVM_CPUMODEL
200  * and defines the facilities that can be enabled via a cpu model.
201  */
202 static unsigned long kvm_s390_fac_ext[SIZE_INTERNAL] = { FACILITIES_KVM_CPUMODEL };
203 
204 static unsigned long kvm_s390_fac_size(void)
205 {
206 	BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_MASK_SIZE_U64);
207 	BUILD_BUG_ON(SIZE_INTERNAL > S390_ARCH_FAC_LIST_SIZE_U64);
208 	BUILD_BUG_ON(SIZE_INTERNAL * sizeof(unsigned long) >
209 		sizeof(S390_lowcore.stfle_fac_list));
210 
211 	return SIZE_INTERNAL;
212 }
213 
214 /* available cpu features supported by kvm */
215 static DECLARE_BITMAP(kvm_s390_available_cpu_feat, KVM_S390_VM_CPU_FEAT_NR_BITS);
216 /* available subfunctions indicated via query / "test bit" */
217 static struct kvm_s390_vm_cpu_subfunc kvm_s390_available_subfunc;
218 
219 static struct gmap_notifier gmap_notifier;
220 static struct gmap_notifier vsie_gmap_notifier;
221 debug_info_t *kvm_s390_dbf;
222 
223 /* Section: not file related */
224 int kvm_arch_hardware_enable(void)
225 {
226 	/* every s390 is virtualization enabled ;-) */
227 	return 0;
228 }
229 
230 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
231 			      unsigned long end);
232 
233 static void kvm_clock_sync_scb(struct kvm_s390_sie_block *scb, u64 delta)
234 {
235 	u8 delta_idx = 0;
236 
237 	/*
238 	 * The TOD jumps by delta, we have to compensate this by adding
239 	 * -delta to the epoch.
240 	 */
241 	delta = -delta;
242 
243 	/* sign-extension - we're adding to signed values below */
244 	if ((s64)delta < 0)
245 		delta_idx = -1;
246 
247 	scb->epoch += delta;
248 	if (scb->ecd & ECD_MEF) {
249 		scb->epdx += delta_idx;
250 		if (scb->epoch < delta)
251 			scb->epdx += 1;
252 	}
253 }
254 
255 /*
256  * This callback is executed during stop_machine(). All CPUs are therefore
257  * temporarily stopped. In order not to change guest behavior, we have to
258  * disable preemption whenever we touch the epoch of kvm and the VCPUs,
259  * so a CPU won't be stopped while calculating with the epoch.
260  */
261 static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
262 			  void *v)
263 {
264 	struct kvm *kvm;
265 	struct kvm_vcpu *vcpu;
266 	int i;
267 	unsigned long long *delta = v;
268 
269 	list_for_each_entry(kvm, &vm_list, vm_list) {
270 		kvm_for_each_vcpu(i, vcpu, kvm) {
271 			kvm_clock_sync_scb(vcpu->arch.sie_block, *delta);
272 			if (i == 0) {
273 				kvm->arch.epoch = vcpu->arch.sie_block->epoch;
274 				kvm->arch.epdx = vcpu->arch.sie_block->epdx;
275 			}
276 			if (vcpu->arch.cputm_enabled)
277 				vcpu->arch.cputm_start += *delta;
278 			if (vcpu->arch.vsie_block)
279 				kvm_clock_sync_scb(vcpu->arch.vsie_block,
280 						   *delta);
281 		}
282 	}
283 	return NOTIFY_OK;
284 }
285 
286 static struct notifier_block kvm_clock_notifier = {
287 	.notifier_call = kvm_clock_sync,
288 };
289 
290 int kvm_arch_hardware_setup(void)
291 {
292 	gmap_notifier.notifier_call = kvm_gmap_notifier;
293 	gmap_register_pte_notifier(&gmap_notifier);
294 	vsie_gmap_notifier.notifier_call = kvm_s390_vsie_gmap_notifier;
295 	gmap_register_pte_notifier(&vsie_gmap_notifier);
296 	atomic_notifier_chain_register(&s390_epoch_delta_notifier,
297 				       &kvm_clock_notifier);
298 	return 0;
299 }
300 
301 void kvm_arch_hardware_unsetup(void)
302 {
303 	gmap_unregister_pte_notifier(&gmap_notifier);
304 	gmap_unregister_pte_notifier(&vsie_gmap_notifier);
305 	atomic_notifier_chain_unregister(&s390_epoch_delta_notifier,
306 					 &kvm_clock_notifier);
307 }
308 
309 static void allow_cpu_feat(unsigned long nr)
310 {
311 	set_bit_inv(nr, kvm_s390_available_cpu_feat);
312 }
313 
314 static inline int plo_test_bit(unsigned char nr)
315 {
316 	register unsigned long r0 asm("0") = (unsigned long) nr | 0x100;
317 	int cc;
318 
319 	asm volatile(
320 		/* Parameter registers are ignored for "test bit" */
321 		"	plo	0,0,0,0(0)\n"
322 		"	ipm	%0\n"
323 		"	srl	%0,28\n"
324 		: "=d" (cc)
325 		: "d" (r0)
326 		: "cc");
327 	return cc == 0;
328 }
329 
330 static inline void __insn32_query(unsigned int opcode, u8 query[32])
331 {
332 	register unsigned long r0 asm("0") = 0;	/* query function */
333 	register unsigned long r1 asm("1") = (unsigned long) query;
334 
335 	asm volatile(
336 		/* Parameter regs are ignored */
337 		"	.insn	rrf,%[opc] << 16,2,4,6,0\n"
338 		: "=m" (*query)
339 		: "d" (r0), "a" (r1), [opc] "i" (opcode)
340 		: "cc");
341 }
342 
343 #define INSN_SORTL 0xb938
344 #define INSN_DFLTCC 0xb939
345 
346 static void kvm_s390_cpu_feat_init(void)
347 {
348 	int i;
349 
350 	for (i = 0; i < 256; ++i) {
351 		if (plo_test_bit(i))
352 			kvm_s390_available_subfunc.plo[i >> 3] |= 0x80 >> (i & 7);
353 	}
354 
355 	if (test_facility(28)) /* TOD-clock steering */
356 		ptff(kvm_s390_available_subfunc.ptff,
357 		     sizeof(kvm_s390_available_subfunc.ptff),
358 		     PTFF_QAF);
359 
360 	if (test_facility(17)) { /* MSA */
361 		__cpacf_query(CPACF_KMAC, (cpacf_mask_t *)
362 			      kvm_s390_available_subfunc.kmac);
363 		__cpacf_query(CPACF_KMC, (cpacf_mask_t *)
364 			      kvm_s390_available_subfunc.kmc);
365 		__cpacf_query(CPACF_KM, (cpacf_mask_t *)
366 			      kvm_s390_available_subfunc.km);
367 		__cpacf_query(CPACF_KIMD, (cpacf_mask_t *)
368 			      kvm_s390_available_subfunc.kimd);
369 		__cpacf_query(CPACF_KLMD, (cpacf_mask_t *)
370 			      kvm_s390_available_subfunc.klmd);
371 	}
372 	if (test_facility(76)) /* MSA3 */
373 		__cpacf_query(CPACF_PCKMO, (cpacf_mask_t *)
374 			      kvm_s390_available_subfunc.pckmo);
375 	if (test_facility(77)) { /* MSA4 */
376 		__cpacf_query(CPACF_KMCTR, (cpacf_mask_t *)
377 			      kvm_s390_available_subfunc.kmctr);
378 		__cpacf_query(CPACF_KMF, (cpacf_mask_t *)
379 			      kvm_s390_available_subfunc.kmf);
380 		__cpacf_query(CPACF_KMO, (cpacf_mask_t *)
381 			      kvm_s390_available_subfunc.kmo);
382 		__cpacf_query(CPACF_PCC, (cpacf_mask_t *)
383 			      kvm_s390_available_subfunc.pcc);
384 	}
385 	if (test_facility(57)) /* MSA5 */
386 		__cpacf_query(CPACF_PRNO, (cpacf_mask_t *)
387 			      kvm_s390_available_subfunc.ppno);
388 
389 	if (test_facility(146)) /* MSA8 */
390 		__cpacf_query(CPACF_KMA, (cpacf_mask_t *)
391 			      kvm_s390_available_subfunc.kma);
392 
393 	if (test_facility(155)) /* MSA9 */
394 		__cpacf_query(CPACF_KDSA, (cpacf_mask_t *)
395 			      kvm_s390_available_subfunc.kdsa);
396 
397 	if (test_facility(150)) /* SORTL */
398 		__insn32_query(INSN_SORTL, kvm_s390_available_subfunc.sortl);
399 
400 	if (test_facility(151)) /* DFLTCC */
401 		__insn32_query(INSN_DFLTCC, kvm_s390_available_subfunc.dfltcc);
402 
403 	if (MACHINE_HAS_ESOP)
404 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP);
405 	/*
406 	 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
407 	 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
408 	 */
409 	if (!sclp.has_sief2 || !MACHINE_HAS_ESOP || !sclp.has_64bscao ||
410 	    !test_facility(3) || !nested)
411 		return;
412 	allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2);
413 	if (sclp.has_64bscao)
414 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO);
415 	if (sclp.has_siif)
416 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF);
417 	if (sclp.has_gpere)
418 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE);
419 	if (sclp.has_gsls)
420 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS);
421 	if (sclp.has_ib)
422 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB);
423 	if (sclp.has_cei)
424 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI);
425 	if (sclp.has_ibs)
426 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS);
427 	if (sclp.has_kss)
428 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_KSS);
429 	/*
430 	 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
431 	 * all skey handling functions read/set the skey from the PGSTE
432 	 * instead of the real storage key.
433 	 *
434 	 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
435 	 * pages being detected as preserved although they are resident.
436 	 *
437 	 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
438 	 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
439 	 *
440 	 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
441 	 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
442 	 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
443 	 *
444 	 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
445 	 * cannot easily shadow the SCA because of the ipte lock.
446 	 */
447 }
448 
449 int kvm_arch_init(void *opaque)
450 {
451 	int rc;
452 
453 	kvm_s390_dbf = debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
454 	if (!kvm_s390_dbf)
455 		return -ENOMEM;
456 
457 	if (debug_register_view(kvm_s390_dbf, &debug_sprintf_view)) {
458 		rc = -ENOMEM;
459 		goto out_debug_unreg;
460 	}
461 
462 	kvm_s390_cpu_feat_init();
463 
464 	/* Register floating interrupt controller interface. */
465 	rc = kvm_register_device_ops(&kvm_flic_ops, KVM_DEV_TYPE_FLIC);
466 	if (rc) {
467 		pr_err("A FLIC registration call failed with rc=%d\n", rc);
468 		goto out_debug_unreg;
469 	}
470 
471 	rc = kvm_s390_gib_init(GAL_ISC);
472 	if (rc)
473 		goto out_gib_destroy;
474 
475 	return 0;
476 
477 out_gib_destroy:
478 	kvm_s390_gib_destroy();
479 out_debug_unreg:
480 	debug_unregister(kvm_s390_dbf);
481 	return rc;
482 }
483 
484 void kvm_arch_exit(void)
485 {
486 	kvm_s390_gib_destroy();
487 	debug_unregister(kvm_s390_dbf);
488 }
489 
490 /* Section: device related */
491 long kvm_arch_dev_ioctl(struct file *filp,
492 			unsigned int ioctl, unsigned long arg)
493 {
494 	if (ioctl == KVM_S390_ENABLE_SIE)
495 		return s390_enable_sie();
496 	return -EINVAL;
497 }
498 
499 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
500 {
501 	int r;
502 
503 	switch (ext) {
504 	case KVM_CAP_S390_PSW:
505 	case KVM_CAP_S390_GMAP:
506 	case KVM_CAP_SYNC_MMU:
507 #ifdef CONFIG_KVM_S390_UCONTROL
508 	case KVM_CAP_S390_UCONTROL:
509 #endif
510 	case KVM_CAP_ASYNC_PF:
511 	case KVM_CAP_SYNC_REGS:
512 	case KVM_CAP_ONE_REG:
513 	case KVM_CAP_ENABLE_CAP:
514 	case KVM_CAP_S390_CSS_SUPPORT:
515 	case KVM_CAP_IOEVENTFD:
516 	case KVM_CAP_DEVICE_CTRL:
517 	case KVM_CAP_S390_IRQCHIP:
518 	case KVM_CAP_VM_ATTRIBUTES:
519 	case KVM_CAP_MP_STATE:
520 	case KVM_CAP_IMMEDIATE_EXIT:
521 	case KVM_CAP_S390_INJECT_IRQ:
522 	case KVM_CAP_S390_USER_SIGP:
523 	case KVM_CAP_S390_USER_STSI:
524 	case KVM_CAP_S390_SKEYS:
525 	case KVM_CAP_S390_IRQ_STATE:
526 	case KVM_CAP_S390_USER_INSTR0:
527 	case KVM_CAP_S390_CMMA_MIGRATION:
528 	case KVM_CAP_S390_AIS:
529 	case KVM_CAP_S390_AIS_MIGRATION:
530 		r = 1;
531 		break;
532 	case KVM_CAP_S390_HPAGE_1M:
533 		r = 0;
534 		if (hpage && !kvm_is_ucontrol(kvm))
535 			r = 1;
536 		break;
537 	case KVM_CAP_S390_MEM_OP:
538 		r = MEM_OP_MAX_SIZE;
539 		break;
540 	case KVM_CAP_NR_VCPUS:
541 	case KVM_CAP_MAX_VCPUS:
542 	case KVM_CAP_MAX_VCPU_ID:
543 		r = KVM_S390_BSCA_CPU_SLOTS;
544 		if (!kvm_s390_use_sca_entries())
545 			r = KVM_MAX_VCPUS;
546 		else if (sclp.has_esca && sclp.has_64bscao)
547 			r = KVM_S390_ESCA_CPU_SLOTS;
548 		break;
549 	case KVM_CAP_S390_COW:
550 		r = MACHINE_HAS_ESOP;
551 		break;
552 	case KVM_CAP_S390_VECTOR_REGISTERS:
553 		r = MACHINE_HAS_VX;
554 		break;
555 	case KVM_CAP_S390_RI:
556 		r = test_facility(64);
557 		break;
558 	case KVM_CAP_S390_GS:
559 		r = test_facility(133);
560 		break;
561 	case KVM_CAP_S390_BPB:
562 		r = test_facility(82);
563 		break;
564 	default:
565 		r = 0;
566 	}
567 	return r;
568 }
569 
570 static void kvm_s390_sync_dirty_log(struct kvm *kvm,
571 				    struct kvm_memory_slot *memslot)
572 {
573 	int i;
574 	gfn_t cur_gfn, last_gfn;
575 	unsigned long gaddr, vmaddr;
576 	struct gmap *gmap = kvm->arch.gmap;
577 	DECLARE_BITMAP(bitmap, _PAGE_ENTRIES);
578 
579 	/* Loop over all guest segments */
580 	cur_gfn = memslot->base_gfn;
581 	last_gfn = memslot->base_gfn + memslot->npages;
582 	for (; cur_gfn <= last_gfn; cur_gfn += _PAGE_ENTRIES) {
583 		gaddr = gfn_to_gpa(cur_gfn);
584 		vmaddr = gfn_to_hva_memslot(memslot, cur_gfn);
585 		if (kvm_is_error_hva(vmaddr))
586 			continue;
587 
588 		bitmap_zero(bitmap, _PAGE_ENTRIES);
589 		gmap_sync_dirty_log_pmd(gmap, bitmap, gaddr, vmaddr);
590 		for (i = 0; i < _PAGE_ENTRIES; i++) {
591 			if (test_bit(i, bitmap))
592 				mark_page_dirty(kvm, cur_gfn + i);
593 		}
594 
595 		if (fatal_signal_pending(current))
596 			return;
597 		cond_resched();
598 	}
599 }
600 
601 /* Section: vm related */
602 static void sca_del_vcpu(struct kvm_vcpu *vcpu);
603 
604 /*
605  * Get (and clear) the dirty memory log for a memory slot.
606  */
607 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
608 			       struct kvm_dirty_log *log)
609 {
610 	int r;
611 	unsigned long n;
612 	struct kvm_memslots *slots;
613 	struct kvm_memory_slot *memslot;
614 	int is_dirty = 0;
615 
616 	if (kvm_is_ucontrol(kvm))
617 		return -EINVAL;
618 
619 	mutex_lock(&kvm->slots_lock);
620 
621 	r = -EINVAL;
622 	if (log->slot >= KVM_USER_MEM_SLOTS)
623 		goto out;
624 
625 	slots = kvm_memslots(kvm);
626 	memslot = id_to_memslot(slots, log->slot);
627 	r = -ENOENT;
628 	if (!memslot->dirty_bitmap)
629 		goto out;
630 
631 	kvm_s390_sync_dirty_log(kvm, memslot);
632 	r = kvm_get_dirty_log(kvm, log, &is_dirty);
633 	if (r)
634 		goto out;
635 
636 	/* Clear the dirty log */
637 	if (is_dirty) {
638 		n = kvm_dirty_bitmap_bytes(memslot);
639 		memset(memslot->dirty_bitmap, 0, n);
640 	}
641 	r = 0;
642 out:
643 	mutex_unlock(&kvm->slots_lock);
644 	return r;
645 }
646 
647 static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
648 {
649 	unsigned int i;
650 	struct kvm_vcpu *vcpu;
651 
652 	kvm_for_each_vcpu(i, vcpu, kvm) {
653 		kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
654 	}
655 }
656 
657 int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
658 {
659 	int r;
660 
661 	if (cap->flags)
662 		return -EINVAL;
663 
664 	switch (cap->cap) {
665 	case KVM_CAP_S390_IRQCHIP:
666 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
667 		kvm->arch.use_irqchip = 1;
668 		r = 0;
669 		break;
670 	case KVM_CAP_S390_USER_SIGP:
671 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
672 		kvm->arch.user_sigp = 1;
673 		r = 0;
674 		break;
675 	case KVM_CAP_S390_VECTOR_REGISTERS:
676 		mutex_lock(&kvm->lock);
677 		if (kvm->created_vcpus) {
678 			r = -EBUSY;
679 		} else if (MACHINE_HAS_VX) {
680 			set_kvm_facility(kvm->arch.model.fac_mask, 129);
681 			set_kvm_facility(kvm->arch.model.fac_list, 129);
682 			if (test_facility(134)) {
683 				set_kvm_facility(kvm->arch.model.fac_mask, 134);
684 				set_kvm_facility(kvm->arch.model.fac_list, 134);
685 			}
686 			if (test_facility(135)) {
687 				set_kvm_facility(kvm->arch.model.fac_mask, 135);
688 				set_kvm_facility(kvm->arch.model.fac_list, 135);
689 			}
690 			if (test_facility(148)) {
691 				set_kvm_facility(kvm->arch.model.fac_mask, 148);
692 				set_kvm_facility(kvm->arch.model.fac_list, 148);
693 			}
694 			if (test_facility(152)) {
695 				set_kvm_facility(kvm->arch.model.fac_mask, 152);
696 				set_kvm_facility(kvm->arch.model.fac_list, 152);
697 			}
698 			r = 0;
699 		} else
700 			r = -EINVAL;
701 		mutex_unlock(&kvm->lock);
702 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
703 			 r ? "(not available)" : "(success)");
704 		break;
705 	case KVM_CAP_S390_RI:
706 		r = -EINVAL;
707 		mutex_lock(&kvm->lock);
708 		if (kvm->created_vcpus) {
709 			r = -EBUSY;
710 		} else if (test_facility(64)) {
711 			set_kvm_facility(kvm->arch.model.fac_mask, 64);
712 			set_kvm_facility(kvm->arch.model.fac_list, 64);
713 			r = 0;
714 		}
715 		mutex_unlock(&kvm->lock);
716 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_RI %s",
717 			 r ? "(not available)" : "(success)");
718 		break;
719 	case KVM_CAP_S390_AIS:
720 		mutex_lock(&kvm->lock);
721 		if (kvm->created_vcpus) {
722 			r = -EBUSY;
723 		} else {
724 			set_kvm_facility(kvm->arch.model.fac_mask, 72);
725 			set_kvm_facility(kvm->arch.model.fac_list, 72);
726 			r = 0;
727 		}
728 		mutex_unlock(&kvm->lock);
729 		VM_EVENT(kvm, 3, "ENABLE: AIS %s",
730 			 r ? "(not available)" : "(success)");
731 		break;
732 	case KVM_CAP_S390_GS:
733 		r = -EINVAL;
734 		mutex_lock(&kvm->lock);
735 		if (kvm->created_vcpus) {
736 			r = -EBUSY;
737 		} else if (test_facility(133)) {
738 			set_kvm_facility(kvm->arch.model.fac_mask, 133);
739 			set_kvm_facility(kvm->arch.model.fac_list, 133);
740 			r = 0;
741 		}
742 		mutex_unlock(&kvm->lock);
743 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_GS %s",
744 			 r ? "(not available)" : "(success)");
745 		break;
746 	case KVM_CAP_S390_HPAGE_1M:
747 		mutex_lock(&kvm->lock);
748 		if (kvm->created_vcpus)
749 			r = -EBUSY;
750 		else if (!hpage || kvm->arch.use_cmma || kvm_is_ucontrol(kvm))
751 			r = -EINVAL;
752 		else {
753 			r = 0;
754 			down_write(&kvm->mm->mmap_sem);
755 			kvm->mm->context.allow_gmap_hpage_1m = 1;
756 			up_write(&kvm->mm->mmap_sem);
757 			/*
758 			 * We might have to create fake 4k page
759 			 * tables. To avoid that the hardware works on
760 			 * stale PGSTEs, we emulate these instructions.
761 			 */
762 			kvm->arch.use_skf = 0;
763 			kvm->arch.use_pfmfi = 0;
764 		}
765 		mutex_unlock(&kvm->lock);
766 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_HPAGE %s",
767 			 r ? "(not available)" : "(success)");
768 		break;
769 	case KVM_CAP_S390_USER_STSI:
770 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
771 		kvm->arch.user_stsi = 1;
772 		r = 0;
773 		break;
774 	case KVM_CAP_S390_USER_INSTR0:
775 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
776 		kvm->arch.user_instr0 = 1;
777 		icpt_operexc_on_all_vcpus(kvm);
778 		r = 0;
779 		break;
780 	default:
781 		r = -EINVAL;
782 		break;
783 	}
784 	return r;
785 }
786 
787 static int kvm_s390_get_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
788 {
789 	int ret;
790 
791 	switch (attr->attr) {
792 	case KVM_S390_VM_MEM_LIMIT_SIZE:
793 		ret = 0;
794 		VM_EVENT(kvm, 3, "QUERY: max guest memory: %lu bytes",
795 			 kvm->arch.mem_limit);
796 		if (put_user(kvm->arch.mem_limit, (u64 __user *)attr->addr))
797 			ret = -EFAULT;
798 		break;
799 	default:
800 		ret = -ENXIO;
801 		break;
802 	}
803 	return ret;
804 }
805 
806 static int kvm_s390_set_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
807 {
808 	int ret;
809 	unsigned int idx;
810 	switch (attr->attr) {
811 	case KVM_S390_VM_MEM_ENABLE_CMMA:
812 		ret = -ENXIO;
813 		if (!sclp.has_cmma)
814 			break;
815 
816 		VM_EVENT(kvm, 3, "%s", "ENABLE: CMMA support");
817 		mutex_lock(&kvm->lock);
818 		if (kvm->created_vcpus)
819 			ret = -EBUSY;
820 		else if (kvm->mm->context.allow_gmap_hpage_1m)
821 			ret = -EINVAL;
822 		else {
823 			kvm->arch.use_cmma = 1;
824 			/* Not compatible with cmma. */
825 			kvm->arch.use_pfmfi = 0;
826 			ret = 0;
827 		}
828 		mutex_unlock(&kvm->lock);
829 		break;
830 	case KVM_S390_VM_MEM_CLR_CMMA:
831 		ret = -ENXIO;
832 		if (!sclp.has_cmma)
833 			break;
834 		ret = -EINVAL;
835 		if (!kvm->arch.use_cmma)
836 			break;
837 
838 		VM_EVENT(kvm, 3, "%s", "RESET: CMMA states");
839 		mutex_lock(&kvm->lock);
840 		idx = srcu_read_lock(&kvm->srcu);
841 		s390_reset_cmma(kvm->arch.gmap->mm);
842 		srcu_read_unlock(&kvm->srcu, idx);
843 		mutex_unlock(&kvm->lock);
844 		ret = 0;
845 		break;
846 	case KVM_S390_VM_MEM_LIMIT_SIZE: {
847 		unsigned long new_limit;
848 
849 		if (kvm_is_ucontrol(kvm))
850 			return -EINVAL;
851 
852 		if (get_user(new_limit, (u64 __user *)attr->addr))
853 			return -EFAULT;
854 
855 		if (kvm->arch.mem_limit != KVM_S390_NO_MEM_LIMIT &&
856 		    new_limit > kvm->arch.mem_limit)
857 			return -E2BIG;
858 
859 		if (!new_limit)
860 			return -EINVAL;
861 
862 		/* gmap_create takes last usable address */
863 		if (new_limit != KVM_S390_NO_MEM_LIMIT)
864 			new_limit -= 1;
865 
866 		ret = -EBUSY;
867 		mutex_lock(&kvm->lock);
868 		if (!kvm->created_vcpus) {
869 			/* gmap_create will round the limit up */
870 			struct gmap *new = gmap_create(current->mm, new_limit);
871 
872 			if (!new) {
873 				ret = -ENOMEM;
874 			} else {
875 				gmap_remove(kvm->arch.gmap);
876 				new->private = kvm;
877 				kvm->arch.gmap = new;
878 				ret = 0;
879 			}
880 		}
881 		mutex_unlock(&kvm->lock);
882 		VM_EVENT(kvm, 3, "SET: max guest address: %lu", new_limit);
883 		VM_EVENT(kvm, 3, "New guest asce: 0x%pK",
884 			 (void *) kvm->arch.gmap->asce);
885 		break;
886 	}
887 	default:
888 		ret = -ENXIO;
889 		break;
890 	}
891 	return ret;
892 }
893 
894 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
895 
896 void kvm_s390_vcpu_crypto_reset_all(struct kvm *kvm)
897 {
898 	struct kvm_vcpu *vcpu;
899 	int i;
900 
901 	kvm_s390_vcpu_block_all(kvm);
902 
903 	kvm_for_each_vcpu(i, vcpu, kvm) {
904 		kvm_s390_vcpu_crypto_setup(vcpu);
905 		/* recreate the shadow crycb by leaving the VSIE handler */
906 		kvm_s390_sync_request(KVM_REQ_VSIE_RESTART, vcpu);
907 	}
908 
909 	kvm_s390_vcpu_unblock_all(kvm);
910 }
911 
912 static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
913 {
914 	mutex_lock(&kvm->lock);
915 	switch (attr->attr) {
916 	case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
917 		if (!test_kvm_facility(kvm, 76)) {
918 			mutex_unlock(&kvm->lock);
919 			return -EINVAL;
920 		}
921 		get_random_bytes(
922 			kvm->arch.crypto.crycb->aes_wrapping_key_mask,
923 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
924 		kvm->arch.crypto.aes_kw = 1;
925 		VM_EVENT(kvm, 3, "%s", "ENABLE: AES keywrapping support");
926 		break;
927 	case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
928 		if (!test_kvm_facility(kvm, 76)) {
929 			mutex_unlock(&kvm->lock);
930 			return -EINVAL;
931 		}
932 		get_random_bytes(
933 			kvm->arch.crypto.crycb->dea_wrapping_key_mask,
934 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
935 		kvm->arch.crypto.dea_kw = 1;
936 		VM_EVENT(kvm, 3, "%s", "ENABLE: DEA keywrapping support");
937 		break;
938 	case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
939 		if (!test_kvm_facility(kvm, 76)) {
940 			mutex_unlock(&kvm->lock);
941 			return -EINVAL;
942 		}
943 		kvm->arch.crypto.aes_kw = 0;
944 		memset(kvm->arch.crypto.crycb->aes_wrapping_key_mask, 0,
945 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
946 		VM_EVENT(kvm, 3, "%s", "DISABLE: AES keywrapping support");
947 		break;
948 	case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
949 		if (!test_kvm_facility(kvm, 76)) {
950 			mutex_unlock(&kvm->lock);
951 			return -EINVAL;
952 		}
953 		kvm->arch.crypto.dea_kw = 0;
954 		memset(kvm->arch.crypto.crycb->dea_wrapping_key_mask, 0,
955 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
956 		VM_EVENT(kvm, 3, "%s", "DISABLE: DEA keywrapping support");
957 		break;
958 	case KVM_S390_VM_CRYPTO_ENABLE_APIE:
959 		if (!ap_instructions_available()) {
960 			mutex_unlock(&kvm->lock);
961 			return -EOPNOTSUPP;
962 		}
963 		kvm->arch.crypto.apie = 1;
964 		break;
965 	case KVM_S390_VM_CRYPTO_DISABLE_APIE:
966 		if (!ap_instructions_available()) {
967 			mutex_unlock(&kvm->lock);
968 			return -EOPNOTSUPP;
969 		}
970 		kvm->arch.crypto.apie = 0;
971 		break;
972 	default:
973 		mutex_unlock(&kvm->lock);
974 		return -ENXIO;
975 	}
976 
977 	kvm_s390_vcpu_crypto_reset_all(kvm);
978 	mutex_unlock(&kvm->lock);
979 	return 0;
980 }
981 
982 static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
983 {
984 	int cx;
985 	struct kvm_vcpu *vcpu;
986 
987 	kvm_for_each_vcpu(cx, vcpu, kvm)
988 		kvm_s390_sync_request(req, vcpu);
989 }
990 
991 /*
992  * Must be called with kvm->srcu held to avoid races on memslots, and with
993  * kvm->slots_lock to avoid races with ourselves and kvm_s390_vm_stop_migration.
994  */
995 static int kvm_s390_vm_start_migration(struct kvm *kvm)
996 {
997 	struct kvm_memory_slot *ms;
998 	struct kvm_memslots *slots;
999 	unsigned long ram_pages = 0;
1000 	int slotnr;
1001 
1002 	/* migration mode already enabled */
1003 	if (kvm->arch.migration_mode)
1004 		return 0;
1005 	slots = kvm_memslots(kvm);
1006 	if (!slots || !slots->used_slots)
1007 		return -EINVAL;
1008 
1009 	if (!kvm->arch.use_cmma) {
1010 		kvm->arch.migration_mode = 1;
1011 		return 0;
1012 	}
1013 	/* mark all the pages in active slots as dirty */
1014 	for (slotnr = 0; slotnr < slots->used_slots; slotnr++) {
1015 		ms = slots->memslots + slotnr;
1016 		/*
1017 		 * The second half of the bitmap is only used on x86,
1018 		 * and would be wasted otherwise, so we put it to good
1019 		 * use here to keep track of the state of the storage
1020 		 * attributes.
1021 		 */
1022 		memset(kvm_second_dirty_bitmap(ms), 0xff, kvm_dirty_bitmap_bytes(ms));
1023 		ram_pages += ms->npages;
1024 	}
1025 	atomic64_set(&kvm->arch.cmma_dirty_pages, ram_pages);
1026 	kvm->arch.migration_mode = 1;
1027 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_START_MIGRATION);
1028 	return 0;
1029 }
1030 
1031 /*
1032  * Must be called with kvm->slots_lock to avoid races with ourselves and
1033  * kvm_s390_vm_start_migration.
1034  */
1035 static int kvm_s390_vm_stop_migration(struct kvm *kvm)
1036 {
1037 	/* migration mode already disabled */
1038 	if (!kvm->arch.migration_mode)
1039 		return 0;
1040 	kvm->arch.migration_mode = 0;
1041 	if (kvm->arch.use_cmma)
1042 		kvm_s390_sync_request_broadcast(kvm, KVM_REQ_STOP_MIGRATION);
1043 	return 0;
1044 }
1045 
1046 static int kvm_s390_vm_set_migration(struct kvm *kvm,
1047 				     struct kvm_device_attr *attr)
1048 {
1049 	int res = -ENXIO;
1050 
1051 	mutex_lock(&kvm->slots_lock);
1052 	switch (attr->attr) {
1053 	case KVM_S390_VM_MIGRATION_START:
1054 		res = kvm_s390_vm_start_migration(kvm);
1055 		break;
1056 	case KVM_S390_VM_MIGRATION_STOP:
1057 		res = kvm_s390_vm_stop_migration(kvm);
1058 		break;
1059 	default:
1060 		break;
1061 	}
1062 	mutex_unlock(&kvm->slots_lock);
1063 
1064 	return res;
1065 }
1066 
1067 static int kvm_s390_vm_get_migration(struct kvm *kvm,
1068 				     struct kvm_device_attr *attr)
1069 {
1070 	u64 mig = kvm->arch.migration_mode;
1071 
1072 	if (attr->attr != KVM_S390_VM_MIGRATION_STATUS)
1073 		return -ENXIO;
1074 
1075 	if (copy_to_user((void __user *)attr->addr, &mig, sizeof(mig)))
1076 		return -EFAULT;
1077 	return 0;
1078 }
1079 
1080 static int kvm_s390_set_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1081 {
1082 	struct kvm_s390_vm_tod_clock gtod;
1083 
1084 	if (copy_from_user(&gtod, (void __user *)attr->addr, sizeof(gtod)))
1085 		return -EFAULT;
1086 
1087 	if (!test_kvm_facility(kvm, 139) && gtod.epoch_idx)
1088 		return -EINVAL;
1089 	kvm_s390_set_tod_clock(kvm, &gtod);
1090 
1091 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x, TOD base: 0x%llx",
1092 		gtod.epoch_idx, gtod.tod);
1093 
1094 	return 0;
1095 }
1096 
1097 static int kvm_s390_set_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1098 {
1099 	u8 gtod_high;
1100 
1101 	if (copy_from_user(&gtod_high, (void __user *)attr->addr,
1102 					   sizeof(gtod_high)))
1103 		return -EFAULT;
1104 
1105 	if (gtod_high != 0)
1106 		return -EINVAL;
1107 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x", gtod_high);
1108 
1109 	return 0;
1110 }
1111 
1112 static int kvm_s390_set_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1113 {
1114 	struct kvm_s390_vm_tod_clock gtod = { 0 };
1115 
1116 	if (copy_from_user(&gtod.tod, (void __user *)attr->addr,
1117 			   sizeof(gtod.tod)))
1118 		return -EFAULT;
1119 
1120 	kvm_s390_set_tod_clock(kvm, &gtod);
1121 	VM_EVENT(kvm, 3, "SET: TOD base: 0x%llx", gtod.tod);
1122 	return 0;
1123 }
1124 
1125 static int kvm_s390_set_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1126 {
1127 	int ret;
1128 
1129 	if (attr->flags)
1130 		return -EINVAL;
1131 
1132 	switch (attr->attr) {
1133 	case KVM_S390_VM_TOD_EXT:
1134 		ret = kvm_s390_set_tod_ext(kvm, attr);
1135 		break;
1136 	case KVM_S390_VM_TOD_HIGH:
1137 		ret = kvm_s390_set_tod_high(kvm, attr);
1138 		break;
1139 	case KVM_S390_VM_TOD_LOW:
1140 		ret = kvm_s390_set_tod_low(kvm, attr);
1141 		break;
1142 	default:
1143 		ret = -ENXIO;
1144 		break;
1145 	}
1146 	return ret;
1147 }
1148 
1149 static void kvm_s390_get_tod_clock(struct kvm *kvm,
1150 				   struct kvm_s390_vm_tod_clock *gtod)
1151 {
1152 	struct kvm_s390_tod_clock_ext htod;
1153 
1154 	preempt_disable();
1155 
1156 	get_tod_clock_ext((char *)&htod);
1157 
1158 	gtod->tod = htod.tod + kvm->arch.epoch;
1159 	gtod->epoch_idx = 0;
1160 	if (test_kvm_facility(kvm, 139)) {
1161 		gtod->epoch_idx = htod.epoch_idx + kvm->arch.epdx;
1162 		if (gtod->tod < htod.tod)
1163 			gtod->epoch_idx += 1;
1164 	}
1165 
1166 	preempt_enable();
1167 }
1168 
1169 static int kvm_s390_get_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1170 {
1171 	struct kvm_s390_vm_tod_clock gtod;
1172 
1173 	memset(&gtod, 0, sizeof(gtod));
1174 	kvm_s390_get_tod_clock(kvm, &gtod);
1175 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1176 		return -EFAULT;
1177 
1178 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x, TOD base: 0x%llx",
1179 		gtod.epoch_idx, gtod.tod);
1180 	return 0;
1181 }
1182 
1183 static int kvm_s390_get_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1184 {
1185 	u8 gtod_high = 0;
1186 
1187 	if (copy_to_user((void __user *)attr->addr, &gtod_high,
1188 					 sizeof(gtod_high)))
1189 		return -EFAULT;
1190 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x", gtod_high);
1191 
1192 	return 0;
1193 }
1194 
1195 static int kvm_s390_get_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1196 {
1197 	u64 gtod;
1198 
1199 	gtod = kvm_s390_get_tod_clock_fast(kvm);
1200 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1201 		return -EFAULT;
1202 	VM_EVENT(kvm, 3, "QUERY: TOD base: 0x%llx", gtod);
1203 
1204 	return 0;
1205 }
1206 
1207 static int kvm_s390_get_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1208 {
1209 	int ret;
1210 
1211 	if (attr->flags)
1212 		return -EINVAL;
1213 
1214 	switch (attr->attr) {
1215 	case KVM_S390_VM_TOD_EXT:
1216 		ret = kvm_s390_get_tod_ext(kvm, attr);
1217 		break;
1218 	case KVM_S390_VM_TOD_HIGH:
1219 		ret = kvm_s390_get_tod_high(kvm, attr);
1220 		break;
1221 	case KVM_S390_VM_TOD_LOW:
1222 		ret = kvm_s390_get_tod_low(kvm, attr);
1223 		break;
1224 	default:
1225 		ret = -ENXIO;
1226 		break;
1227 	}
1228 	return ret;
1229 }
1230 
1231 static int kvm_s390_set_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1232 {
1233 	struct kvm_s390_vm_cpu_processor *proc;
1234 	u16 lowest_ibc, unblocked_ibc;
1235 	int ret = 0;
1236 
1237 	mutex_lock(&kvm->lock);
1238 	if (kvm->created_vcpus) {
1239 		ret = -EBUSY;
1240 		goto out;
1241 	}
1242 	proc = kzalloc(sizeof(*proc), GFP_KERNEL);
1243 	if (!proc) {
1244 		ret = -ENOMEM;
1245 		goto out;
1246 	}
1247 	if (!copy_from_user(proc, (void __user *)attr->addr,
1248 			    sizeof(*proc))) {
1249 		kvm->arch.model.cpuid = proc->cpuid;
1250 		lowest_ibc = sclp.ibc >> 16 & 0xfff;
1251 		unblocked_ibc = sclp.ibc & 0xfff;
1252 		if (lowest_ibc && proc->ibc) {
1253 			if (proc->ibc > unblocked_ibc)
1254 				kvm->arch.model.ibc = unblocked_ibc;
1255 			else if (proc->ibc < lowest_ibc)
1256 				kvm->arch.model.ibc = lowest_ibc;
1257 			else
1258 				kvm->arch.model.ibc = proc->ibc;
1259 		}
1260 		memcpy(kvm->arch.model.fac_list, proc->fac_list,
1261 		       S390_ARCH_FAC_LIST_SIZE_BYTE);
1262 		VM_EVENT(kvm, 3, "SET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1263 			 kvm->arch.model.ibc,
1264 			 kvm->arch.model.cpuid);
1265 		VM_EVENT(kvm, 3, "SET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1266 			 kvm->arch.model.fac_list[0],
1267 			 kvm->arch.model.fac_list[1],
1268 			 kvm->arch.model.fac_list[2]);
1269 	} else
1270 		ret = -EFAULT;
1271 	kfree(proc);
1272 out:
1273 	mutex_unlock(&kvm->lock);
1274 	return ret;
1275 }
1276 
1277 static int kvm_s390_set_processor_feat(struct kvm *kvm,
1278 				       struct kvm_device_attr *attr)
1279 {
1280 	struct kvm_s390_vm_cpu_feat data;
1281 
1282 	if (copy_from_user(&data, (void __user *)attr->addr, sizeof(data)))
1283 		return -EFAULT;
1284 	if (!bitmap_subset((unsigned long *) data.feat,
1285 			   kvm_s390_available_cpu_feat,
1286 			   KVM_S390_VM_CPU_FEAT_NR_BITS))
1287 		return -EINVAL;
1288 
1289 	mutex_lock(&kvm->lock);
1290 	if (kvm->created_vcpus) {
1291 		mutex_unlock(&kvm->lock);
1292 		return -EBUSY;
1293 	}
1294 	bitmap_copy(kvm->arch.cpu_feat, (unsigned long *) data.feat,
1295 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1296 	mutex_unlock(&kvm->lock);
1297 	VM_EVENT(kvm, 3, "SET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1298 			 data.feat[0],
1299 			 data.feat[1],
1300 			 data.feat[2]);
1301 	return 0;
1302 }
1303 
1304 static int kvm_s390_set_processor_subfunc(struct kvm *kvm,
1305 					  struct kvm_device_attr *attr)
1306 {
1307 	mutex_lock(&kvm->lock);
1308 	if (kvm->created_vcpus) {
1309 		mutex_unlock(&kvm->lock);
1310 		return -EBUSY;
1311 	}
1312 
1313 	if (copy_from_user(&kvm->arch.model.subfuncs, (void __user *)attr->addr,
1314 			   sizeof(struct kvm_s390_vm_cpu_subfunc))) {
1315 		mutex_unlock(&kvm->lock);
1316 		return -EFAULT;
1317 	}
1318 	mutex_unlock(&kvm->lock);
1319 
1320 	VM_EVENT(kvm, 3, "SET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1321 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1322 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1323 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1324 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1325 	VM_EVENT(kvm, 3, "SET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1326 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1327 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1328 	VM_EVENT(kvm, 3, "SET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1329 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1330 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1331 	VM_EVENT(kvm, 3, "SET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1332 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1333 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1334 	VM_EVENT(kvm, 3, "SET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1335 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1336 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1337 	VM_EVENT(kvm, 3, "SET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1338 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1339 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1340 	VM_EVENT(kvm, 3, "SET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1341 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1342 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1343 	VM_EVENT(kvm, 3, "SET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1344 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1345 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1346 	VM_EVENT(kvm, 3, "SET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1347 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1348 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1349 	VM_EVENT(kvm, 3, "SET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1350 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1351 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1352 	VM_EVENT(kvm, 3, "SET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1353 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1354 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1355 	VM_EVENT(kvm, 3, "SET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1356 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1357 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1358 	VM_EVENT(kvm, 3, "SET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1359 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1360 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1361 	VM_EVENT(kvm, 3, "SET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1362 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1363 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1364 	VM_EVENT(kvm, 3, "SET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1365 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1366 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1367 	VM_EVENT(kvm, 3, "SET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1368 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1369 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1370 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1371 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1372 	VM_EVENT(kvm, 3, "SET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1373 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1374 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1375 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1376 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1377 
1378 	return 0;
1379 }
1380 
1381 static int kvm_s390_set_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1382 {
1383 	int ret = -ENXIO;
1384 
1385 	switch (attr->attr) {
1386 	case KVM_S390_VM_CPU_PROCESSOR:
1387 		ret = kvm_s390_set_processor(kvm, attr);
1388 		break;
1389 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1390 		ret = kvm_s390_set_processor_feat(kvm, attr);
1391 		break;
1392 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1393 		ret = kvm_s390_set_processor_subfunc(kvm, attr);
1394 		break;
1395 	}
1396 	return ret;
1397 }
1398 
1399 static int kvm_s390_get_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1400 {
1401 	struct kvm_s390_vm_cpu_processor *proc;
1402 	int ret = 0;
1403 
1404 	proc = kzalloc(sizeof(*proc), GFP_KERNEL);
1405 	if (!proc) {
1406 		ret = -ENOMEM;
1407 		goto out;
1408 	}
1409 	proc->cpuid = kvm->arch.model.cpuid;
1410 	proc->ibc = kvm->arch.model.ibc;
1411 	memcpy(&proc->fac_list, kvm->arch.model.fac_list,
1412 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1413 	VM_EVENT(kvm, 3, "GET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1414 		 kvm->arch.model.ibc,
1415 		 kvm->arch.model.cpuid);
1416 	VM_EVENT(kvm, 3, "GET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1417 		 kvm->arch.model.fac_list[0],
1418 		 kvm->arch.model.fac_list[1],
1419 		 kvm->arch.model.fac_list[2]);
1420 	if (copy_to_user((void __user *)attr->addr, proc, sizeof(*proc)))
1421 		ret = -EFAULT;
1422 	kfree(proc);
1423 out:
1424 	return ret;
1425 }
1426 
1427 static int kvm_s390_get_machine(struct kvm *kvm, struct kvm_device_attr *attr)
1428 {
1429 	struct kvm_s390_vm_cpu_machine *mach;
1430 	int ret = 0;
1431 
1432 	mach = kzalloc(sizeof(*mach), GFP_KERNEL);
1433 	if (!mach) {
1434 		ret = -ENOMEM;
1435 		goto out;
1436 	}
1437 	get_cpu_id((struct cpuid *) &mach->cpuid);
1438 	mach->ibc = sclp.ibc;
1439 	memcpy(&mach->fac_mask, kvm->arch.model.fac_mask,
1440 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1441 	memcpy((unsigned long *)&mach->fac_list, S390_lowcore.stfle_fac_list,
1442 	       sizeof(S390_lowcore.stfle_fac_list));
1443 	VM_EVENT(kvm, 3, "GET: host ibc:  0x%4.4x, host cpuid:  0x%16.16llx",
1444 		 kvm->arch.model.ibc,
1445 		 kvm->arch.model.cpuid);
1446 	VM_EVENT(kvm, 3, "GET: host facmask:  0x%16.16llx.%16.16llx.%16.16llx",
1447 		 mach->fac_mask[0],
1448 		 mach->fac_mask[1],
1449 		 mach->fac_mask[2]);
1450 	VM_EVENT(kvm, 3, "GET: host faclist:  0x%16.16llx.%16.16llx.%16.16llx",
1451 		 mach->fac_list[0],
1452 		 mach->fac_list[1],
1453 		 mach->fac_list[2]);
1454 	if (copy_to_user((void __user *)attr->addr, mach, sizeof(*mach)))
1455 		ret = -EFAULT;
1456 	kfree(mach);
1457 out:
1458 	return ret;
1459 }
1460 
1461 static int kvm_s390_get_processor_feat(struct kvm *kvm,
1462 				       struct kvm_device_attr *attr)
1463 {
1464 	struct kvm_s390_vm_cpu_feat data;
1465 
1466 	bitmap_copy((unsigned long *) data.feat, kvm->arch.cpu_feat,
1467 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1468 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1469 		return -EFAULT;
1470 	VM_EVENT(kvm, 3, "GET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1471 			 data.feat[0],
1472 			 data.feat[1],
1473 			 data.feat[2]);
1474 	return 0;
1475 }
1476 
1477 static int kvm_s390_get_machine_feat(struct kvm *kvm,
1478 				     struct kvm_device_attr *attr)
1479 {
1480 	struct kvm_s390_vm_cpu_feat data;
1481 
1482 	bitmap_copy((unsigned long *) data.feat,
1483 		    kvm_s390_available_cpu_feat,
1484 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1485 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1486 		return -EFAULT;
1487 	VM_EVENT(kvm, 3, "GET: host 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_get_processor_subfunc(struct kvm *kvm,
1495 					  struct kvm_device_attr *attr)
1496 {
1497 	if (copy_to_user((void __user *)attr->addr, &kvm->arch.model.subfuncs,
1498 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1499 		return -EFAULT;
1500 
1501 	VM_EVENT(kvm, 3, "GET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1502 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1503 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1504 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1505 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1506 	VM_EVENT(kvm, 3, "GET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1507 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1508 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1509 	VM_EVENT(kvm, 3, "GET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1510 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1511 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1512 	VM_EVENT(kvm, 3, "GET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1513 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1514 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1515 	VM_EVENT(kvm, 3, "GET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1516 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1517 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1518 	VM_EVENT(kvm, 3, "GET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1519 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1520 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1521 	VM_EVENT(kvm, 3, "GET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1522 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1523 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1524 	VM_EVENT(kvm, 3, "GET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1525 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1526 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1527 	VM_EVENT(kvm, 3, "GET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1528 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1529 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1530 	VM_EVENT(kvm, 3, "GET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1531 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1532 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1533 	VM_EVENT(kvm, 3, "GET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1534 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1535 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1536 	VM_EVENT(kvm, 3, "GET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1537 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1538 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1539 	VM_EVENT(kvm, 3, "GET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1540 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1541 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1542 	VM_EVENT(kvm, 3, "GET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1543 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1544 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1545 	VM_EVENT(kvm, 3, "GET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1546 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1547 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1548 	VM_EVENT(kvm, 3, "GET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1549 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1550 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1551 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1552 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1553 	VM_EVENT(kvm, 3, "GET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1554 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1555 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1556 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1557 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1558 
1559 	return 0;
1560 }
1561 
1562 static int kvm_s390_get_machine_subfunc(struct kvm *kvm,
1563 					struct kvm_device_attr *attr)
1564 {
1565 	if (copy_to_user((void __user *)attr->addr, &kvm_s390_available_subfunc,
1566 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1567 		return -EFAULT;
1568 
1569 	VM_EVENT(kvm, 3, "GET: host  PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1570 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[0],
1571 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[1],
1572 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[2],
1573 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[3]);
1574 	VM_EVENT(kvm, 3, "GET: host  PTFF   subfunc 0x%16.16lx.%16.16lx",
1575 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[0],
1576 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[1]);
1577 	VM_EVENT(kvm, 3, "GET: host  KMAC   subfunc 0x%16.16lx.%16.16lx",
1578 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[0],
1579 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[1]);
1580 	VM_EVENT(kvm, 3, "GET: host  KMC    subfunc 0x%16.16lx.%16.16lx",
1581 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[0],
1582 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[1]);
1583 	VM_EVENT(kvm, 3, "GET: host  KM     subfunc 0x%16.16lx.%16.16lx",
1584 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[0],
1585 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[1]);
1586 	VM_EVENT(kvm, 3, "GET: host  KIMD   subfunc 0x%16.16lx.%16.16lx",
1587 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[0],
1588 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[1]);
1589 	VM_EVENT(kvm, 3, "GET: host  KLMD   subfunc 0x%16.16lx.%16.16lx",
1590 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[0],
1591 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[1]);
1592 	VM_EVENT(kvm, 3, "GET: host  PCKMO  subfunc 0x%16.16lx.%16.16lx",
1593 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[0],
1594 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[1]);
1595 	VM_EVENT(kvm, 3, "GET: host  KMCTR  subfunc 0x%16.16lx.%16.16lx",
1596 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[0],
1597 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[1]);
1598 	VM_EVENT(kvm, 3, "GET: host  KMF    subfunc 0x%16.16lx.%16.16lx",
1599 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[0],
1600 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[1]);
1601 	VM_EVENT(kvm, 3, "GET: host  KMO    subfunc 0x%16.16lx.%16.16lx",
1602 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[0],
1603 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[1]);
1604 	VM_EVENT(kvm, 3, "GET: host  PCC    subfunc 0x%16.16lx.%16.16lx",
1605 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[0],
1606 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[1]);
1607 	VM_EVENT(kvm, 3, "GET: host  PPNO   subfunc 0x%16.16lx.%16.16lx",
1608 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[0],
1609 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[1]);
1610 	VM_EVENT(kvm, 3, "GET: host  KMA    subfunc 0x%16.16lx.%16.16lx",
1611 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[0],
1612 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[1]);
1613 	VM_EVENT(kvm, 3, "GET: host  KDSA   subfunc 0x%16.16lx.%16.16lx",
1614 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[0],
1615 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[1]);
1616 	VM_EVENT(kvm, 3, "GET: host  SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1617 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[0],
1618 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[1],
1619 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[2],
1620 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[3]);
1621 	VM_EVENT(kvm, 3, "GET: host  DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1622 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[0],
1623 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[1],
1624 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[2],
1625 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[3]);
1626 
1627 	return 0;
1628 }
1629 
1630 static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1631 {
1632 	int ret = -ENXIO;
1633 
1634 	switch (attr->attr) {
1635 	case KVM_S390_VM_CPU_PROCESSOR:
1636 		ret = kvm_s390_get_processor(kvm, attr);
1637 		break;
1638 	case KVM_S390_VM_CPU_MACHINE:
1639 		ret = kvm_s390_get_machine(kvm, attr);
1640 		break;
1641 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1642 		ret = kvm_s390_get_processor_feat(kvm, attr);
1643 		break;
1644 	case KVM_S390_VM_CPU_MACHINE_FEAT:
1645 		ret = kvm_s390_get_machine_feat(kvm, attr);
1646 		break;
1647 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1648 		ret = kvm_s390_get_processor_subfunc(kvm, attr);
1649 		break;
1650 	case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1651 		ret = kvm_s390_get_machine_subfunc(kvm, attr);
1652 		break;
1653 	}
1654 	return ret;
1655 }
1656 
1657 static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1658 {
1659 	int ret;
1660 
1661 	switch (attr->group) {
1662 	case KVM_S390_VM_MEM_CTRL:
1663 		ret = kvm_s390_set_mem_control(kvm, attr);
1664 		break;
1665 	case KVM_S390_VM_TOD:
1666 		ret = kvm_s390_set_tod(kvm, attr);
1667 		break;
1668 	case KVM_S390_VM_CPU_MODEL:
1669 		ret = kvm_s390_set_cpu_model(kvm, attr);
1670 		break;
1671 	case KVM_S390_VM_CRYPTO:
1672 		ret = kvm_s390_vm_set_crypto(kvm, attr);
1673 		break;
1674 	case KVM_S390_VM_MIGRATION:
1675 		ret = kvm_s390_vm_set_migration(kvm, attr);
1676 		break;
1677 	default:
1678 		ret = -ENXIO;
1679 		break;
1680 	}
1681 
1682 	return ret;
1683 }
1684 
1685 static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1686 {
1687 	int ret;
1688 
1689 	switch (attr->group) {
1690 	case KVM_S390_VM_MEM_CTRL:
1691 		ret = kvm_s390_get_mem_control(kvm, attr);
1692 		break;
1693 	case KVM_S390_VM_TOD:
1694 		ret = kvm_s390_get_tod(kvm, attr);
1695 		break;
1696 	case KVM_S390_VM_CPU_MODEL:
1697 		ret = kvm_s390_get_cpu_model(kvm, attr);
1698 		break;
1699 	case KVM_S390_VM_MIGRATION:
1700 		ret = kvm_s390_vm_get_migration(kvm, attr);
1701 		break;
1702 	default:
1703 		ret = -ENXIO;
1704 		break;
1705 	}
1706 
1707 	return ret;
1708 }
1709 
1710 static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1711 {
1712 	int ret;
1713 
1714 	switch (attr->group) {
1715 	case KVM_S390_VM_MEM_CTRL:
1716 		switch (attr->attr) {
1717 		case KVM_S390_VM_MEM_ENABLE_CMMA:
1718 		case KVM_S390_VM_MEM_CLR_CMMA:
1719 			ret = sclp.has_cmma ? 0 : -ENXIO;
1720 			break;
1721 		case KVM_S390_VM_MEM_LIMIT_SIZE:
1722 			ret = 0;
1723 			break;
1724 		default:
1725 			ret = -ENXIO;
1726 			break;
1727 		}
1728 		break;
1729 	case KVM_S390_VM_TOD:
1730 		switch (attr->attr) {
1731 		case KVM_S390_VM_TOD_LOW:
1732 		case KVM_S390_VM_TOD_HIGH:
1733 			ret = 0;
1734 			break;
1735 		default:
1736 			ret = -ENXIO;
1737 			break;
1738 		}
1739 		break;
1740 	case KVM_S390_VM_CPU_MODEL:
1741 		switch (attr->attr) {
1742 		case KVM_S390_VM_CPU_PROCESSOR:
1743 		case KVM_S390_VM_CPU_MACHINE:
1744 		case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1745 		case KVM_S390_VM_CPU_MACHINE_FEAT:
1746 		case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1747 		case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1748 			ret = 0;
1749 			break;
1750 		default:
1751 			ret = -ENXIO;
1752 			break;
1753 		}
1754 		break;
1755 	case KVM_S390_VM_CRYPTO:
1756 		switch (attr->attr) {
1757 		case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
1758 		case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
1759 		case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
1760 		case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
1761 			ret = 0;
1762 			break;
1763 		case KVM_S390_VM_CRYPTO_ENABLE_APIE:
1764 		case KVM_S390_VM_CRYPTO_DISABLE_APIE:
1765 			ret = ap_instructions_available() ? 0 : -ENXIO;
1766 			break;
1767 		default:
1768 			ret = -ENXIO;
1769 			break;
1770 		}
1771 		break;
1772 	case KVM_S390_VM_MIGRATION:
1773 		ret = 0;
1774 		break;
1775 	default:
1776 		ret = -ENXIO;
1777 		break;
1778 	}
1779 
1780 	return ret;
1781 }
1782 
1783 static long kvm_s390_get_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1784 {
1785 	uint8_t *keys;
1786 	uint64_t hva;
1787 	int srcu_idx, i, r = 0;
1788 
1789 	if (args->flags != 0)
1790 		return -EINVAL;
1791 
1792 	/* Is this guest using storage keys? */
1793 	if (!mm_uses_skeys(current->mm))
1794 		return KVM_S390_GET_SKEYS_NONE;
1795 
1796 	/* Enforce sane limit on memory allocation */
1797 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1798 		return -EINVAL;
1799 
1800 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL);
1801 	if (!keys)
1802 		return -ENOMEM;
1803 
1804 	down_read(&current->mm->mmap_sem);
1805 	srcu_idx = srcu_read_lock(&kvm->srcu);
1806 	for (i = 0; i < args->count; i++) {
1807 		hva = gfn_to_hva(kvm, args->start_gfn + i);
1808 		if (kvm_is_error_hva(hva)) {
1809 			r = -EFAULT;
1810 			break;
1811 		}
1812 
1813 		r = get_guest_storage_key(current->mm, hva, &keys[i]);
1814 		if (r)
1815 			break;
1816 	}
1817 	srcu_read_unlock(&kvm->srcu, srcu_idx);
1818 	up_read(&current->mm->mmap_sem);
1819 
1820 	if (!r) {
1821 		r = copy_to_user((uint8_t __user *)args->skeydata_addr, keys,
1822 				 sizeof(uint8_t) * args->count);
1823 		if (r)
1824 			r = -EFAULT;
1825 	}
1826 
1827 	kvfree(keys);
1828 	return r;
1829 }
1830 
1831 static long kvm_s390_set_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1832 {
1833 	uint8_t *keys;
1834 	uint64_t hva;
1835 	int srcu_idx, i, r = 0;
1836 	bool unlocked;
1837 
1838 	if (args->flags != 0)
1839 		return -EINVAL;
1840 
1841 	/* Enforce sane limit on memory allocation */
1842 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1843 		return -EINVAL;
1844 
1845 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL);
1846 	if (!keys)
1847 		return -ENOMEM;
1848 
1849 	r = copy_from_user(keys, (uint8_t __user *)args->skeydata_addr,
1850 			   sizeof(uint8_t) * args->count);
1851 	if (r) {
1852 		r = -EFAULT;
1853 		goto out;
1854 	}
1855 
1856 	/* Enable storage key handling for the guest */
1857 	r = s390_enable_skey();
1858 	if (r)
1859 		goto out;
1860 
1861 	i = 0;
1862 	down_read(&current->mm->mmap_sem);
1863 	srcu_idx = srcu_read_lock(&kvm->srcu);
1864         while (i < args->count) {
1865 		unlocked = false;
1866 		hva = gfn_to_hva(kvm, args->start_gfn + i);
1867 		if (kvm_is_error_hva(hva)) {
1868 			r = -EFAULT;
1869 			break;
1870 		}
1871 
1872 		/* Lowest order bit is reserved */
1873 		if (keys[i] & 0x01) {
1874 			r = -EINVAL;
1875 			break;
1876 		}
1877 
1878 		r = set_guest_storage_key(current->mm, hva, keys[i], 0);
1879 		if (r) {
1880 			r = fixup_user_fault(current, current->mm, hva,
1881 					     FAULT_FLAG_WRITE, &unlocked);
1882 			if (r)
1883 				break;
1884 		}
1885 		if (!r)
1886 			i++;
1887 	}
1888 	srcu_read_unlock(&kvm->srcu, srcu_idx);
1889 	up_read(&current->mm->mmap_sem);
1890 out:
1891 	kvfree(keys);
1892 	return r;
1893 }
1894 
1895 /*
1896  * Base address and length must be sent at the start of each block, therefore
1897  * it's cheaper to send some clean data, as long as it's less than the size of
1898  * two longs.
1899  */
1900 #define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
1901 /* for consistency */
1902 #define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
1903 
1904 /*
1905  * Similar to gfn_to_memslot, but returns the index of a memslot also when the
1906  * address falls in a hole. In that case the index of one of the memslots
1907  * bordering the hole is returned.
1908  */
1909 static int gfn_to_memslot_approx(struct kvm_memslots *slots, gfn_t gfn)
1910 {
1911 	int start = 0, end = slots->used_slots;
1912 	int slot = atomic_read(&slots->lru_slot);
1913 	struct kvm_memory_slot *memslots = slots->memslots;
1914 
1915 	if (gfn >= memslots[slot].base_gfn &&
1916 	    gfn < memslots[slot].base_gfn + memslots[slot].npages)
1917 		return slot;
1918 
1919 	while (start < end) {
1920 		slot = start + (end - start) / 2;
1921 
1922 		if (gfn >= memslots[slot].base_gfn)
1923 			end = slot;
1924 		else
1925 			start = slot + 1;
1926 	}
1927 
1928 	if (gfn >= memslots[start].base_gfn &&
1929 	    gfn < memslots[start].base_gfn + memslots[start].npages) {
1930 		atomic_set(&slots->lru_slot, start);
1931 	}
1932 
1933 	return start;
1934 }
1935 
1936 static int kvm_s390_peek_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
1937 			      u8 *res, unsigned long bufsize)
1938 {
1939 	unsigned long pgstev, hva, cur_gfn = args->start_gfn;
1940 
1941 	args->count = 0;
1942 	while (args->count < bufsize) {
1943 		hva = gfn_to_hva(kvm, cur_gfn);
1944 		/*
1945 		 * We return an error if the first value was invalid, but we
1946 		 * return successfully if at least one value was copied.
1947 		 */
1948 		if (kvm_is_error_hva(hva))
1949 			return args->count ? 0 : -EFAULT;
1950 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
1951 			pgstev = 0;
1952 		res[args->count++] = (pgstev >> 24) & 0x43;
1953 		cur_gfn++;
1954 	}
1955 
1956 	return 0;
1957 }
1958 
1959 static unsigned long kvm_s390_next_dirty_cmma(struct kvm_memslots *slots,
1960 					      unsigned long cur_gfn)
1961 {
1962 	int slotidx = gfn_to_memslot_approx(slots, cur_gfn);
1963 	struct kvm_memory_slot *ms = slots->memslots + slotidx;
1964 	unsigned long ofs = cur_gfn - ms->base_gfn;
1965 
1966 	if (ms->base_gfn + ms->npages <= cur_gfn) {
1967 		slotidx--;
1968 		/* If we are above the highest slot, wrap around */
1969 		if (slotidx < 0)
1970 			slotidx = slots->used_slots - 1;
1971 
1972 		ms = slots->memslots + slotidx;
1973 		ofs = 0;
1974 	}
1975 	ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, ofs);
1976 	while ((slotidx > 0) && (ofs >= ms->npages)) {
1977 		slotidx--;
1978 		ms = slots->memslots + slotidx;
1979 		ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, 0);
1980 	}
1981 	return ms->base_gfn + ofs;
1982 }
1983 
1984 static int kvm_s390_get_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
1985 			     u8 *res, unsigned long bufsize)
1986 {
1987 	unsigned long mem_end, cur_gfn, next_gfn, hva, pgstev;
1988 	struct kvm_memslots *slots = kvm_memslots(kvm);
1989 	struct kvm_memory_slot *ms;
1990 
1991 	cur_gfn = kvm_s390_next_dirty_cmma(slots, args->start_gfn);
1992 	ms = gfn_to_memslot(kvm, cur_gfn);
1993 	args->count = 0;
1994 	args->start_gfn = cur_gfn;
1995 	if (!ms)
1996 		return 0;
1997 	next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
1998 	mem_end = slots->memslots[0].base_gfn + slots->memslots[0].npages;
1999 
2000 	while (args->count < bufsize) {
2001 		hva = gfn_to_hva(kvm, cur_gfn);
2002 		if (kvm_is_error_hva(hva))
2003 			return 0;
2004 		/* Decrement only if we actually flipped the bit to 0 */
2005 		if (test_and_clear_bit(cur_gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
2006 			atomic64_dec(&kvm->arch.cmma_dirty_pages);
2007 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
2008 			pgstev = 0;
2009 		/* Save the value */
2010 		res[args->count++] = (pgstev >> 24) & 0x43;
2011 		/* If the next bit is too far away, stop. */
2012 		if (next_gfn > cur_gfn + KVM_S390_MAX_BIT_DISTANCE)
2013 			return 0;
2014 		/* If we reached the previous "next", find the next one */
2015 		if (cur_gfn == next_gfn)
2016 			next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2017 		/* Reached the end of memory or of the buffer, stop */
2018 		if ((next_gfn >= mem_end) ||
2019 		    (next_gfn - args->start_gfn >= bufsize))
2020 			return 0;
2021 		cur_gfn++;
2022 		/* Reached the end of the current memslot, take the next one. */
2023 		if (cur_gfn - ms->base_gfn >= ms->npages) {
2024 			ms = gfn_to_memslot(kvm, cur_gfn);
2025 			if (!ms)
2026 				return 0;
2027 		}
2028 	}
2029 	return 0;
2030 }
2031 
2032 /*
2033  * This function searches for the next page with dirty CMMA attributes, and
2034  * saves the attributes in the buffer up to either the end of the buffer or
2035  * until a block of at least KVM_S390_MAX_BIT_DISTANCE clean bits is found;
2036  * no trailing clean bytes are saved.
2037  * In case no dirty bits were found, or if CMMA was not enabled or used, the
2038  * output buffer will indicate 0 as length.
2039  */
2040 static int kvm_s390_get_cmma_bits(struct kvm *kvm,
2041 				  struct kvm_s390_cmma_log *args)
2042 {
2043 	unsigned long bufsize;
2044 	int srcu_idx, peek, ret;
2045 	u8 *values;
2046 
2047 	if (!kvm->arch.use_cmma)
2048 		return -ENXIO;
2049 	/* Invalid/unsupported flags were specified */
2050 	if (args->flags & ~KVM_S390_CMMA_PEEK)
2051 		return -EINVAL;
2052 	/* Migration mode query, and we are not doing a migration */
2053 	peek = !!(args->flags & KVM_S390_CMMA_PEEK);
2054 	if (!peek && !kvm->arch.migration_mode)
2055 		return -EINVAL;
2056 	/* CMMA is disabled or was not used, or the buffer has length zero */
2057 	bufsize = min(args->count, KVM_S390_CMMA_SIZE_MAX);
2058 	if (!bufsize || !kvm->mm->context.uses_cmm) {
2059 		memset(args, 0, sizeof(*args));
2060 		return 0;
2061 	}
2062 	/* We are not peeking, and there are no dirty pages */
2063 	if (!peek && !atomic64_read(&kvm->arch.cmma_dirty_pages)) {
2064 		memset(args, 0, sizeof(*args));
2065 		return 0;
2066 	}
2067 
2068 	values = vmalloc(bufsize);
2069 	if (!values)
2070 		return -ENOMEM;
2071 
2072 	down_read(&kvm->mm->mmap_sem);
2073 	srcu_idx = srcu_read_lock(&kvm->srcu);
2074 	if (peek)
2075 		ret = kvm_s390_peek_cmma(kvm, args, values, bufsize);
2076 	else
2077 		ret = kvm_s390_get_cmma(kvm, args, values, bufsize);
2078 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2079 	up_read(&kvm->mm->mmap_sem);
2080 
2081 	if (kvm->arch.migration_mode)
2082 		args->remaining = atomic64_read(&kvm->arch.cmma_dirty_pages);
2083 	else
2084 		args->remaining = 0;
2085 
2086 	if (copy_to_user((void __user *)args->values, values, args->count))
2087 		ret = -EFAULT;
2088 
2089 	vfree(values);
2090 	return ret;
2091 }
2092 
2093 /*
2094  * This function sets the CMMA attributes for the given pages. If the input
2095  * buffer has zero length, no action is taken, otherwise the attributes are
2096  * set and the mm->context.uses_cmm flag is set.
2097  */
2098 static int kvm_s390_set_cmma_bits(struct kvm *kvm,
2099 				  const struct kvm_s390_cmma_log *args)
2100 {
2101 	unsigned long hva, mask, pgstev, i;
2102 	uint8_t *bits;
2103 	int srcu_idx, r = 0;
2104 
2105 	mask = args->mask;
2106 
2107 	if (!kvm->arch.use_cmma)
2108 		return -ENXIO;
2109 	/* invalid/unsupported flags */
2110 	if (args->flags != 0)
2111 		return -EINVAL;
2112 	/* Enforce sane limit on memory allocation */
2113 	if (args->count > KVM_S390_CMMA_SIZE_MAX)
2114 		return -EINVAL;
2115 	/* Nothing to do */
2116 	if (args->count == 0)
2117 		return 0;
2118 
2119 	bits = vmalloc(array_size(sizeof(*bits), args->count));
2120 	if (!bits)
2121 		return -ENOMEM;
2122 
2123 	r = copy_from_user(bits, (void __user *)args->values, args->count);
2124 	if (r) {
2125 		r = -EFAULT;
2126 		goto out;
2127 	}
2128 
2129 	down_read(&kvm->mm->mmap_sem);
2130 	srcu_idx = srcu_read_lock(&kvm->srcu);
2131 	for (i = 0; i < args->count; i++) {
2132 		hva = gfn_to_hva(kvm, args->start_gfn + i);
2133 		if (kvm_is_error_hva(hva)) {
2134 			r = -EFAULT;
2135 			break;
2136 		}
2137 
2138 		pgstev = bits[i];
2139 		pgstev = pgstev << 24;
2140 		mask &= _PGSTE_GPS_USAGE_MASK | _PGSTE_GPS_NODAT;
2141 		set_pgste_bits(kvm->mm, hva, mask, pgstev);
2142 	}
2143 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2144 	up_read(&kvm->mm->mmap_sem);
2145 
2146 	if (!kvm->mm->context.uses_cmm) {
2147 		down_write(&kvm->mm->mmap_sem);
2148 		kvm->mm->context.uses_cmm = 1;
2149 		up_write(&kvm->mm->mmap_sem);
2150 	}
2151 out:
2152 	vfree(bits);
2153 	return r;
2154 }
2155 
2156 long kvm_arch_vm_ioctl(struct file *filp,
2157 		       unsigned int ioctl, unsigned long arg)
2158 {
2159 	struct kvm *kvm = filp->private_data;
2160 	void __user *argp = (void __user *)arg;
2161 	struct kvm_device_attr attr;
2162 	int r;
2163 
2164 	switch (ioctl) {
2165 	case KVM_S390_INTERRUPT: {
2166 		struct kvm_s390_interrupt s390int;
2167 
2168 		r = -EFAULT;
2169 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
2170 			break;
2171 		r = kvm_s390_inject_vm(kvm, &s390int);
2172 		break;
2173 	}
2174 	case KVM_CREATE_IRQCHIP: {
2175 		struct kvm_irq_routing_entry routing;
2176 
2177 		r = -EINVAL;
2178 		if (kvm->arch.use_irqchip) {
2179 			/* Set up dummy routing. */
2180 			memset(&routing, 0, sizeof(routing));
2181 			r = kvm_set_irq_routing(kvm, &routing, 0, 0);
2182 		}
2183 		break;
2184 	}
2185 	case KVM_SET_DEVICE_ATTR: {
2186 		r = -EFAULT;
2187 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2188 			break;
2189 		r = kvm_s390_vm_set_attr(kvm, &attr);
2190 		break;
2191 	}
2192 	case KVM_GET_DEVICE_ATTR: {
2193 		r = -EFAULT;
2194 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2195 			break;
2196 		r = kvm_s390_vm_get_attr(kvm, &attr);
2197 		break;
2198 	}
2199 	case KVM_HAS_DEVICE_ATTR: {
2200 		r = -EFAULT;
2201 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2202 			break;
2203 		r = kvm_s390_vm_has_attr(kvm, &attr);
2204 		break;
2205 	}
2206 	case KVM_S390_GET_SKEYS: {
2207 		struct kvm_s390_skeys args;
2208 
2209 		r = -EFAULT;
2210 		if (copy_from_user(&args, argp,
2211 				   sizeof(struct kvm_s390_skeys)))
2212 			break;
2213 		r = kvm_s390_get_skeys(kvm, &args);
2214 		break;
2215 	}
2216 	case KVM_S390_SET_SKEYS: {
2217 		struct kvm_s390_skeys args;
2218 
2219 		r = -EFAULT;
2220 		if (copy_from_user(&args, argp,
2221 				   sizeof(struct kvm_s390_skeys)))
2222 			break;
2223 		r = kvm_s390_set_skeys(kvm, &args);
2224 		break;
2225 	}
2226 	case KVM_S390_GET_CMMA_BITS: {
2227 		struct kvm_s390_cmma_log args;
2228 
2229 		r = -EFAULT;
2230 		if (copy_from_user(&args, argp, sizeof(args)))
2231 			break;
2232 		mutex_lock(&kvm->slots_lock);
2233 		r = kvm_s390_get_cmma_bits(kvm, &args);
2234 		mutex_unlock(&kvm->slots_lock);
2235 		if (!r) {
2236 			r = copy_to_user(argp, &args, sizeof(args));
2237 			if (r)
2238 				r = -EFAULT;
2239 		}
2240 		break;
2241 	}
2242 	case KVM_S390_SET_CMMA_BITS: {
2243 		struct kvm_s390_cmma_log args;
2244 
2245 		r = -EFAULT;
2246 		if (copy_from_user(&args, argp, sizeof(args)))
2247 			break;
2248 		mutex_lock(&kvm->slots_lock);
2249 		r = kvm_s390_set_cmma_bits(kvm, &args);
2250 		mutex_unlock(&kvm->slots_lock);
2251 		break;
2252 	}
2253 	default:
2254 		r = -ENOTTY;
2255 	}
2256 
2257 	return r;
2258 }
2259 
2260 static int kvm_s390_apxa_installed(void)
2261 {
2262 	struct ap_config_info info;
2263 
2264 	if (ap_instructions_available()) {
2265 		if (ap_qci(&info) == 0)
2266 			return info.apxa;
2267 	}
2268 
2269 	return 0;
2270 }
2271 
2272 /*
2273  * The format of the crypto control block (CRYCB) is specified in the 3 low
2274  * order bits of the CRYCB designation (CRYCBD) field as follows:
2275  * Format 0: Neither the message security assist extension 3 (MSAX3) nor the
2276  *	     AP extended addressing (APXA) facility are installed.
2277  * Format 1: The APXA facility is not installed but the MSAX3 facility is.
2278  * Format 2: Both the APXA and MSAX3 facilities are installed
2279  */
2280 static void kvm_s390_set_crycb_format(struct kvm *kvm)
2281 {
2282 	kvm->arch.crypto.crycbd = (__u32)(unsigned long) kvm->arch.crypto.crycb;
2283 
2284 	/* Clear the CRYCB format bits - i.e., set format 0 by default */
2285 	kvm->arch.crypto.crycbd &= ~(CRYCB_FORMAT_MASK);
2286 
2287 	/* Check whether MSAX3 is installed */
2288 	if (!test_kvm_facility(kvm, 76))
2289 		return;
2290 
2291 	if (kvm_s390_apxa_installed())
2292 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT2;
2293 	else
2294 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT1;
2295 }
2296 
2297 void kvm_arch_crypto_set_masks(struct kvm *kvm, unsigned long *apm,
2298 			       unsigned long *aqm, unsigned long *adm)
2299 {
2300 	struct kvm_s390_crypto_cb *crycb = kvm->arch.crypto.crycb;
2301 
2302 	mutex_lock(&kvm->lock);
2303 	kvm_s390_vcpu_block_all(kvm);
2304 
2305 	switch (kvm->arch.crypto.crycbd & CRYCB_FORMAT_MASK) {
2306 	case CRYCB_FORMAT2: /* APCB1 use 256 bits */
2307 		memcpy(crycb->apcb1.apm, apm, 32);
2308 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx %016lx %016lx %016lx",
2309 			 apm[0], apm[1], apm[2], apm[3]);
2310 		memcpy(crycb->apcb1.aqm, aqm, 32);
2311 		VM_EVENT(kvm, 3, "SET CRYCB: aqm %016lx %016lx %016lx %016lx",
2312 			 aqm[0], aqm[1], aqm[2], aqm[3]);
2313 		memcpy(crycb->apcb1.adm, adm, 32);
2314 		VM_EVENT(kvm, 3, "SET CRYCB: adm %016lx %016lx %016lx %016lx",
2315 			 adm[0], adm[1], adm[2], adm[3]);
2316 		break;
2317 	case CRYCB_FORMAT1:
2318 	case CRYCB_FORMAT0: /* Fall through both use APCB0 */
2319 		memcpy(crycb->apcb0.apm, apm, 8);
2320 		memcpy(crycb->apcb0.aqm, aqm, 2);
2321 		memcpy(crycb->apcb0.adm, adm, 2);
2322 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx aqm %04x adm %04x",
2323 			 apm[0], *((unsigned short *)aqm),
2324 			 *((unsigned short *)adm));
2325 		break;
2326 	default:	/* Can not happen */
2327 		break;
2328 	}
2329 
2330 	/* recreate the shadow crycb for each vcpu */
2331 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2332 	kvm_s390_vcpu_unblock_all(kvm);
2333 	mutex_unlock(&kvm->lock);
2334 }
2335 EXPORT_SYMBOL_GPL(kvm_arch_crypto_set_masks);
2336 
2337 void kvm_arch_crypto_clear_masks(struct kvm *kvm)
2338 {
2339 	mutex_lock(&kvm->lock);
2340 	kvm_s390_vcpu_block_all(kvm);
2341 
2342 	memset(&kvm->arch.crypto.crycb->apcb0, 0,
2343 	       sizeof(kvm->arch.crypto.crycb->apcb0));
2344 	memset(&kvm->arch.crypto.crycb->apcb1, 0,
2345 	       sizeof(kvm->arch.crypto.crycb->apcb1));
2346 
2347 	VM_EVENT(kvm, 3, "%s", "CLR CRYCB:");
2348 	/* recreate the shadow crycb for each vcpu */
2349 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2350 	kvm_s390_vcpu_unblock_all(kvm);
2351 	mutex_unlock(&kvm->lock);
2352 }
2353 EXPORT_SYMBOL_GPL(kvm_arch_crypto_clear_masks);
2354 
2355 static u64 kvm_s390_get_initial_cpuid(void)
2356 {
2357 	struct cpuid cpuid;
2358 
2359 	get_cpu_id(&cpuid);
2360 	cpuid.version = 0xff;
2361 	return *((u64 *) &cpuid);
2362 }
2363 
2364 static void kvm_s390_crypto_init(struct kvm *kvm)
2365 {
2366 	kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
2367 	kvm_s390_set_crycb_format(kvm);
2368 
2369 	if (!test_kvm_facility(kvm, 76))
2370 		return;
2371 
2372 	/* Enable AES/DEA protected key functions by default */
2373 	kvm->arch.crypto.aes_kw = 1;
2374 	kvm->arch.crypto.dea_kw = 1;
2375 	get_random_bytes(kvm->arch.crypto.crycb->aes_wrapping_key_mask,
2376 			 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
2377 	get_random_bytes(kvm->arch.crypto.crycb->dea_wrapping_key_mask,
2378 			 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
2379 }
2380 
2381 static void sca_dispose(struct kvm *kvm)
2382 {
2383 	if (kvm->arch.use_esca)
2384 		free_pages_exact(kvm->arch.sca, sizeof(struct esca_block));
2385 	else
2386 		free_page((unsigned long)(kvm->arch.sca));
2387 	kvm->arch.sca = NULL;
2388 }
2389 
2390 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
2391 {
2392 	gfp_t alloc_flags = GFP_KERNEL;
2393 	int i, rc;
2394 	char debug_name[16];
2395 	static unsigned long sca_offset;
2396 
2397 	rc = -EINVAL;
2398 #ifdef CONFIG_KVM_S390_UCONTROL
2399 	if (type & ~KVM_VM_S390_UCONTROL)
2400 		goto out_err;
2401 	if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN)))
2402 		goto out_err;
2403 #else
2404 	if (type)
2405 		goto out_err;
2406 #endif
2407 
2408 	rc = s390_enable_sie();
2409 	if (rc)
2410 		goto out_err;
2411 
2412 	rc = -ENOMEM;
2413 
2414 	if (!sclp.has_64bscao)
2415 		alloc_flags |= GFP_DMA;
2416 	rwlock_init(&kvm->arch.sca_lock);
2417 	/* start with basic SCA */
2418 	kvm->arch.sca = (struct bsca_block *) get_zeroed_page(alloc_flags);
2419 	if (!kvm->arch.sca)
2420 		goto out_err;
2421 	spin_lock(&kvm_lock);
2422 	sca_offset += 16;
2423 	if (sca_offset + sizeof(struct bsca_block) > PAGE_SIZE)
2424 		sca_offset = 0;
2425 	kvm->arch.sca = (struct bsca_block *)
2426 			((char *) kvm->arch.sca + sca_offset);
2427 	spin_unlock(&kvm_lock);
2428 
2429 	sprintf(debug_name, "kvm-%u", current->pid);
2430 
2431 	kvm->arch.dbf = debug_register(debug_name, 32, 1, 7 * sizeof(long));
2432 	if (!kvm->arch.dbf)
2433 		goto out_err;
2434 
2435 	BUILD_BUG_ON(sizeof(struct sie_page2) != 4096);
2436 	kvm->arch.sie_page2 =
2437 	     (struct sie_page2 *) get_zeroed_page(GFP_KERNEL | GFP_DMA);
2438 	if (!kvm->arch.sie_page2)
2439 		goto out_err;
2440 
2441 	kvm->arch.sie_page2->kvm = kvm;
2442 	kvm->arch.model.fac_list = kvm->arch.sie_page2->fac_list;
2443 
2444 	for (i = 0; i < kvm_s390_fac_size(); i++) {
2445 		kvm->arch.model.fac_mask[i] = S390_lowcore.stfle_fac_list[i] &
2446 					      (kvm_s390_fac_base[i] |
2447 					       kvm_s390_fac_ext[i]);
2448 		kvm->arch.model.fac_list[i] = S390_lowcore.stfle_fac_list[i] &
2449 					      kvm_s390_fac_base[i];
2450 	}
2451 	kvm->arch.model.subfuncs = kvm_s390_available_subfunc;
2452 
2453 	/* we are always in czam mode - even on pre z14 machines */
2454 	set_kvm_facility(kvm->arch.model.fac_mask, 138);
2455 	set_kvm_facility(kvm->arch.model.fac_list, 138);
2456 	/* we emulate STHYI in kvm */
2457 	set_kvm_facility(kvm->arch.model.fac_mask, 74);
2458 	set_kvm_facility(kvm->arch.model.fac_list, 74);
2459 	if (MACHINE_HAS_TLB_GUEST) {
2460 		set_kvm_facility(kvm->arch.model.fac_mask, 147);
2461 		set_kvm_facility(kvm->arch.model.fac_list, 147);
2462 	}
2463 
2464 	kvm->arch.model.cpuid = kvm_s390_get_initial_cpuid();
2465 	kvm->arch.model.ibc = sclp.ibc & 0x0fff;
2466 
2467 	kvm_s390_crypto_init(kvm);
2468 
2469 	mutex_init(&kvm->arch.float_int.ais_lock);
2470 	spin_lock_init(&kvm->arch.float_int.lock);
2471 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
2472 		INIT_LIST_HEAD(&kvm->arch.float_int.lists[i]);
2473 	init_waitqueue_head(&kvm->arch.ipte_wq);
2474 	mutex_init(&kvm->arch.ipte_mutex);
2475 
2476 	debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
2477 	VM_EVENT(kvm, 3, "vm created with type %lu", type);
2478 
2479 	if (type & KVM_VM_S390_UCONTROL) {
2480 		kvm->arch.gmap = NULL;
2481 		kvm->arch.mem_limit = KVM_S390_NO_MEM_LIMIT;
2482 	} else {
2483 		if (sclp.hamax == U64_MAX)
2484 			kvm->arch.mem_limit = TASK_SIZE_MAX;
2485 		else
2486 			kvm->arch.mem_limit = min_t(unsigned long, TASK_SIZE_MAX,
2487 						    sclp.hamax + 1);
2488 		kvm->arch.gmap = gmap_create(current->mm, kvm->arch.mem_limit - 1);
2489 		if (!kvm->arch.gmap)
2490 			goto out_err;
2491 		kvm->arch.gmap->private = kvm;
2492 		kvm->arch.gmap->pfault_enabled = 0;
2493 	}
2494 
2495 	kvm->arch.use_pfmfi = sclp.has_pfmfi;
2496 	kvm->arch.use_skf = sclp.has_skey;
2497 	spin_lock_init(&kvm->arch.start_stop_lock);
2498 	kvm_s390_vsie_init(kvm);
2499 	kvm_s390_gisa_init(kvm);
2500 	KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm, current->pid);
2501 
2502 	return 0;
2503 out_err:
2504 	free_page((unsigned long)kvm->arch.sie_page2);
2505 	debug_unregister(kvm->arch.dbf);
2506 	sca_dispose(kvm);
2507 	KVM_EVENT(3, "creation of vm failed: %d", rc);
2508 	return rc;
2509 }
2510 
2511 bool kvm_arch_has_vcpu_debugfs(void)
2512 {
2513 	return false;
2514 }
2515 
2516 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
2517 {
2518 	return 0;
2519 }
2520 
2521 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
2522 {
2523 	VCPU_EVENT(vcpu, 3, "%s", "free cpu");
2524 	trace_kvm_s390_destroy_vcpu(vcpu->vcpu_id);
2525 	kvm_s390_clear_local_irqs(vcpu);
2526 	kvm_clear_async_pf_completion_queue(vcpu);
2527 	if (!kvm_is_ucontrol(vcpu->kvm))
2528 		sca_del_vcpu(vcpu);
2529 
2530 	if (kvm_is_ucontrol(vcpu->kvm))
2531 		gmap_remove(vcpu->arch.gmap);
2532 
2533 	if (vcpu->kvm->arch.use_cmma)
2534 		kvm_s390_vcpu_unsetup_cmma(vcpu);
2535 	free_page((unsigned long)(vcpu->arch.sie_block));
2536 
2537 	kvm_vcpu_uninit(vcpu);
2538 	kmem_cache_free(kvm_vcpu_cache, vcpu);
2539 }
2540 
2541 static void kvm_free_vcpus(struct kvm *kvm)
2542 {
2543 	unsigned int i;
2544 	struct kvm_vcpu *vcpu;
2545 
2546 	kvm_for_each_vcpu(i, vcpu, kvm)
2547 		kvm_arch_vcpu_destroy(vcpu);
2548 
2549 	mutex_lock(&kvm->lock);
2550 	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
2551 		kvm->vcpus[i] = NULL;
2552 
2553 	atomic_set(&kvm->online_vcpus, 0);
2554 	mutex_unlock(&kvm->lock);
2555 }
2556 
2557 void kvm_arch_destroy_vm(struct kvm *kvm)
2558 {
2559 	kvm_free_vcpus(kvm);
2560 	sca_dispose(kvm);
2561 	debug_unregister(kvm->arch.dbf);
2562 	kvm_s390_gisa_destroy(kvm);
2563 	free_page((unsigned long)kvm->arch.sie_page2);
2564 	if (!kvm_is_ucontrol(kvm))
2565 		gmap_remove(kvm->arch.gmap);
2566 	kvm_s390_destroy_adapters(kvm);
2567 	kvm_s390_clear_float_irqs(kvm);
2568 	kvm_s390_vsie_destroy(kvm);
2569 	KVM_EVENT(3, "vm 0x%pK destroyed", kvm);
2570 }
2571 
2572 /* Section: vcpu related */
2573 static int __kvm_ucontrol_vcpu_init(struct kvm_vcpu *vcpu)
2574 {
2575 	vcpu->arch.gmap = gmap_create(current->mm, -1UL);
2576 	if (!vcpu->arch.gmap)
2577 		return -ENOMEM;
2578 	vcpu->arch.gmap->private = vcpu->kvm;
2579 
2580 	return 0;
2581 }
2582 
2583 static void sca_del_vcpu(struct kvm_vcpu *vcpu)
2584 {
2585 	if (!kvm_s390_use_sca_entries())
2586 		return;
2587 	read_lock(&vcpu->kvm->arch.sca_lock);
2588 	if (vcpu->kvm->arch.use_esca) {
2589 		struct esca_block *sca = vcpu->kvm->arch.sca;
2590 
2591 		clear_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
2592 		sca->cpu[vcpu->vcpu_id].sda = 0;
2593 	} else {
2594 		struct bsca_block *sca = vcpu->kvm->arch.sca;
2595 
2596 		clear_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
2597 		sca->cpu[vcpu->vcpu_id].sda = 0;
2598 	}
2599 	read_unlock(&vcpu->kvm->arch.sca_lock);
2600 }
2601 
2602 static void sca_add_vcpu(struct kvm_vcpu *vcpu)
2603 {
2604 	if (!kvm_s390_use_sca_entries()) {
2605 		struct bsca_block *sca = vcpu->kvm->arch.sca;
2606 
2607 		/* we still need the basic sca for the ipte control */
2608 		vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2609 		vcpu->arch.sie_block->scaol = (__u32)(__u64)sca;
2610 		return;
2611 	}
2612 	read_lock(&vcpu->kvm->arch.sca_lock);
2613 	if (vcpu->kvm->arch.use_esca) {
2614 		struct esca_block *sca = vcpu->kvm->arch.sca;
2615 
2616 		sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
2617 		vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2618 		vcpu->arch.sie_block->scaol = (__u32)(__u64)sca & ~0x3fU;
2619 		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
2620 		set_bit_inv(vcpu->vcpu_id, (unsigned long *) sca->mcn);
2621 	} else {
2622 		struct bsca_block *sca = vcpu->kvm->arch.sca;
2623 
2624 		sca->cpu[vcpu->vcpu_id].sda = (__u64) vcpu->arch.sie_block;
2625 		vcpu->arch.sie_block->scaoh = (__u32)(((__u64)sca) >> 32);
2626 		vcpu->arch.sie_block->scaol = (__u32)(__u64)sca;
2627 		set_bit_inv(vcpu->vcpu_id, (unsigned long *) &sca->mcn);
2628 	}
2629 	read_unlock(&vcpu->kvm->arch.sca_lock);
2630 }
2631 
2632 /* Basic SCA to Extended SCA data copy routines */
2633 static inline void sca_copy_entry(struct esca_entry *d, struct bsca_entry *s)
2634 {
2635 	d->sda = s->sda;
2636 	d->sigp_ctrl.c = s->sigp_ctrl.c;
2637 	d->sigp_ctrl.scn = s->sigp_ctrl.scn;
2638 }
2639 
2640 static void sca_copy_b_to_e(struct esca_block *d, struct bsca_block *s)
2641 {
2642 	int i;
2643 
2644 	d->ipte_control = s->ipte_control;
2645 	d->mcn[0] = s->mcn;
2646 	for (i = 0; i < KVM_S390_BSCA_CPU_SLOTS; i++)
2647 		sca_copy_entry(&d->cpu[i], &s->cpu[i]);
2648 }
2649 
2650 static int sca_switch_to_extended(struct kvm *kvm)
2651 {
2652 	struct bsca_block *old_sca = kvm->arch.sca;
2653 	struct esca_block *new_sca;
2654 	struct kvm_vcpu *vcpu;
2655 	unsigned int vcpu_idx;
2656 	u32 scaol, scaoh;
2657 
2658 	new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL|__GFP_ZERO);
2659 	if (!new_sca)
2660 		return -ENOMEM;
2661 
2662 	scaoh = (u32)((u64)(new_sca) >> 32);
2663 	scaol = (u32)(u64)(new_sca) & ~0x3fU;
2664 
2665 	kvm_s390_vcpu_block_all(kvm);
2666 	write_lock(&kvm->arch.sca_lock);
2667 
2668 	sca_copy_b_to_e(new_sca, old_sca);
2669 
2670 	kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
2671 		vcpu->arch.sie_block->scaoh = scaoh;
2672 		vcpu->arch.sie_block->scaol = scaol;
2673 		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
2674 	}
2675 	kvm->arch.sca = new_sca;
2676 	kvm->arch.use_esca = 1;
2677 
2678 	write_unlock(&kvm->arch.sca_lock);
2679 	kvm_s390_vcpu_unblock_all(kvm);
2680 
2681 	free_page((unsigned long)old_sca);
2682 
2683 	VM_EVENT(kvm, 2, "Switched to ESCA (0x%pK -> 0x%pK)",
2684 		 old_sca, kvm->arch.sca);
2685 	return 0;
2686 }
2687 
2688 static int sca_can_add_vcpu(struct kvm *kvm, unsigned int id)
2689 {
2690 	int rc;
2691 
2692 	if (!kvm_s390_use_sca_entries()) {
2693 		if (id < KVM_MAX_VCPUS)
2694 			return true;
2695 		return false;
2696 	}
2697 	if (id < KVM_S390_BSCA_CPU_SLOTS)
2698 		return true;
2699 	if (!sclp.has_esca || !sclp.has_64bscao)
2700 		return false;
2701 
2702 	mutex_lock(&kvm->lock);
2703 	rc = kvm->arch.use_esca ? 0 : sca_switch_to_extended(kvm);
2704 	mutex_unlock(&kvm->lock);
2705 
2706 	return rc == 0 && id < KVM_S390_ESCA_CPU_SLOTS;
2707 }
2708 
2709 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
2710 {
2711 	vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
2712 	kvm_clear_async_pf_completion_queue(vcpu);
2713 	vcpu->run->kvm_valid_regs = KVM_SYNC_PREFIX |
2714 				    KVM_SYNC_GPRS |
2715 				    KVM_SYNC_ACRS |
2716 				    KVM_SYNC_CRS |
2717 				    KVM_SYNC_ARCH0 |
2718 				    KVM_SYNC_PFAULT;
2719 	kvm_s390_set_prefix(vcpu, 0);
2720 	if (test_kvm_facility(vcpu->kvm, 64))
2721 		vcpu->run->kvm_valid_regs |= KVM_SYNC_RICCB;
2722 	if (test_kvm_facility(vcpu->kvm, 82))
2723 		vcpu->run->kvm_valid_regs |= KVM_SYNC_BPBC;
2724 	if (test_kvm_facility(vcpu->kvm, 133))
2725 		vcpu->run->kvm_valid_regs |= KVM_SYNC_GSCB;
2726 	if (test_kvm_facility(vcpu->kvm, 156))
2727 		vcpu->run->kvm_valid_regs |= KVM_SYNC_ETOKEN;
2728 	/* fprs can be synchronized via vrs, even if the guest has no vx. With
2729 	 * MACHINE_HAS_VX, (load|store)_fpu_regs() will work with vrs format.
2730 	 */
2731 	if (MACHINE_HAS_VX)
2732 		vcpu->run->kvm_valid_regs |= KVM_SYNC_VRS;
2733 	else
2734 		vcpu->run->kvm_valid_regs |= KVM_SYNC_FPRS;
2735 
2736 	if (kvm_is_ucontrol(vcpu->kvm))
2737 		return __kvm_ucontrol_vcpu_init(vcpu);
2738 
2739 	return 0;
2740 }
2741 
2742 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2743 static void __start_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2744 {
2745 	WARN_ON_ONCE(vcpu->arch.cputm_start != 0);
2746 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
2747 	vcpu->arch.cputm_start = get_tod_clock_fast();
2748 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
2749 }
2750 
2751 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2752 static void __stop_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2753 {
2754 	WARN_ON_ONCE(vcpu->arch.cputm_start == 0);
2755 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
2756 	vcpu->arch.sie_block->cputm -= get_tod_clock_fast() - vcpu->arch.cputm_start;
2757 	vcpu->arch.cputm_start = 0;
2758 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
2759 }
2760 
2761 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2762 static void __enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2763 {
2764 	WARN_ON_ONCE(vcpu->arch.cputm_enabled);
2765 	vcpu->arch.cputm_enabled = true;
2766 	__start_cpu_timer_accounting(vcpu);
2767 }
2768 
2769 /* needs disabled preemption to protect from TOD sync and vcpu_load/put */
2770 static void __disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2771 {
2772 	WARN_ON_ONCE(!vcpu->arch.cputm_enabled);
2773 	__stop_cpu_timer_accounting(vcpu);
2774 	vcpu->arch.cputm_enabled = false;
2775 }
2776 
2777 static void enable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2778 {
2779 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2780 	__enable_cpu_timer_accounting(vcpu);
2781 	preempt_enable();
2782 }
2783 
2784 static void disable_cpu_timer_accounting(struct kvm_vcpu *vcpu)
2785 {
2786 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2787 	__disable_cpu_timer_accounting(vcpu);
2788 	preempt_enable();
2789 }
2790 
2791 /* set the cpu timer - may only be called from the VCPU thread itself */
2792 void kvm_s390_set_cpu_timer(struct kvm_vcpu *vcpu, __u64 cputm)
2793 {
2794 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2795 	raw_write_seqcount_begin(&vcpu->arch.cputm_seqcount);
2796 	if (vcpu->arch.cputm_enabled)
2797 		vcpu->arch.cputm_start = get_tod_clock_fast();
2798 	vcpu->arch.sie_block->cputm = cputm;
2799 	raw_write_seqcount_end(&vcpu->arch.cputm_seqcount);
2800 	preempt_enable();
2801 }
2802 
2803 /* update and get the cpu timer - can also be called from other VCPU threads */
2804 __u64 kvm_s390_get_cpu_timer(struct kvm_vcpu *vcpu)
2805 {
2806 	unsigned int seq;
2807 	__u64 value;
2808 
2809 	if (unlikely(!vcpu->arch.cputm_enabled))
2810 		return vcpu->arch.sie_block->cputm;
2811 
2812 	preempt_disable(); /* protect from TOD sync and vcpu_load/put */
2813 	do {
2814 		seq = raw_read_seqcount(&vcpu->arch.cputm_seqcount);
2815 		/*
2816 		 * If the writer would ever execute a read in the critical
2817 		 * section, e.g. in irq context, we have a deadlock.
2818 		 */
2819 		WARN_ON_ONCE((seq & 1) && smp_processor_id() == vcpu->cpu);
2820 		value = vcpu->arch.sie_block->cputm;
2821 		/* if cputm_start is 0, accounting is being started/stopped */
2822 		if (likely(vcpu->arch.cputm_start))
2823 			value -= get_tod_clock_fast() - vcpu->arch.cputm_start;
2824 	} while (read_seqcount_retry(&vcpu->arch.cputm_seqcount, seq & ~1));
2825 	preempt_enable();
2826 	return value;
2827 }
2828 
2829 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2830 {
2831 
2832 	gmap_enable(vcpu->arch.enabled_gmap);
2833 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_RUNNING);
2834 	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
2835 		__start_cpu_timer_accounting(vcpu);
2836 	vcpu->cpu = cpu;
2837 }
2838 
2839 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
2840 {
2841 	vcpu->cpu = -1;
2842 	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
2843 		__stop_cpu_timer_accounting(vcpu);
2844 	kvm_s390_clear_cpuflags(vcpu, CPUSTAT_RUNNING);
2845 	vcpu->arch.enabled_gmap = gmap_get_enabled();
2846 	gmap_disable(vcpu->arch.enabled_gmap);
2847 
2848 }
2849 
2850 static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu)
2851 {
2852 	/* this equals initial cpu reset in pop, but we don't switch to ESA */
2853 	vcpu->arch.sie_block->gpsw.mask = 0UL;
2854 	vcpu->arch.sie_block->gpsw.addr = 0UL;
2855 	kvm_s390_set_prefix(vcpu, 0);
2856 	kvm_s390_set_cpu_timer(vcpu, 0);
2857 	vcpu->arch.sie_block->ckc       = 0UL;
2858 	vcpu->arch.sie_block->todpr     = 0;
2859 	memset(vcpu->arch.sie_block->gcr, 0, 16 * sizeof(__u64));
2860 	vcpu->arch.sie_block->gcr[0]  = CR0_UNUSED_56 |
2861 					CR0_INTERRUPT_KEY_SUBMASK |
2862 					CR0_MEASUREMENT_ALERT_SUBMASK;
2863 	vcpu->arch.sie_block->gcr[14] = CR14_UNUSED_32 |
2864 					CR14_UNUSED_33 |
2865 					CR14_EXTERNAL_DAMAGE_SUBMASK;
2866 	/* make sure the new fpc will be lazily loaded */
2867 	save_fpu_regs();
2868 	current->thread.fpu.fpc = 0;
2869 	vcpu->arch.sie_block->gbea = 1;
2870 	vcpu->arch.sie_block->pp = 0;
2871 	vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
2872 	vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
2873 	kvm_clear_async_pf_completion_queue(vcpu);
2874 	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
2875 		kvm_s390_vcpu_stop(vcpu);
2876 	kvm_s390_clear_local_irqs(vcpu);
2877 }
2878 
2879 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
2880 {
2881 	mutex_lock(&vcpu->kvm->lock);
2882 	preempt_disable();
2883 	vcpu->arch.sie_block->epoch = vcpu->kvm->arch.epoch;
2884 	vcpu->arch.sie_block->epdx = vcpu->kvm->arch.epdx;
2885 	preempt_enable();
2886 	mutex_unlock(&vcpu->kvm->lock);
2887 	if (!kvm_is_ucontrol(vcpu->kvm)) {
2888 		vcpu->arch.gmap = vcpu->kvm->arch.gmap;
2889 		sca_add_vcpu(vcpu);
2890 	}
2891 	if (test_kvm_facility(vcpu->kvm, 74) || vcpu->kvm->arch.user_instr0)
2892 		vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
2893 	/* make vcpu_load load the right gmap on the first trigger */
2894 	vcpu->arch.enabled_gmap = vcpu->arch.gmap;
2895 }
2896 
2897 static bool kvm_has_pckmo_subfunc(struct kvm *kvm, unsigned long nr)
2898 {
2899 	if (test_bit_inv(nr, (unsigned long *)&kvm->arch.model.subfuncs.pckmo) &&
2900 	    test_bit_inv(nr, (unsigned long *)&kvm_s390_available_subfunc.pckmo))
2901 		return true;
2902 	return false;
2903 }
2904 
2905 static bool kvm_has_pckmo_ecc(struct kvm *kvm)
2906 {
2907 	/* At least one ECC subfunction must be present */
2908 	return kvm_has_pckmo_subfunc(kvm, 32) ||
2909 	       kvm_has_pckmo_subfunc(kvm, 33) ||
2910 	       kvm_has_pckmo_subfunc(kvm, 34) ||
2911 	       kvm_has_pckmo_subfunc(kvm, 40) ||
2912 	       kvm_has_pckmo_subfunc(kvm, 41);
2913 
2914 }
2915 
2916 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu)
2917 {
2918 	/*
2919 	 * If the AP instructions are not being interpreted and the MSAX3
2920 	 * facility is not configured for the guest, there is nothing to set up.
2921 	 */
2922 	if (!vcpu->kvm->arch.crypto.apie && !test_kvm_facility(vcpu->kvm, 76))
2923 		return;
2924 
2925 	vcpu->arch.sie_block->crycbd = vcpu->kvm->arch.crypto.crycbd;
2926 	vcpu->arch.sie_block->ecb3 &= ~(ECB3_AES | ECB3_DEA);
2927 	vcpu->arch.sie_block->eca &= ~ECA_APIE;
2928 	vcpu->arch.sie_block->ecd &= ~ECD_ECC;
2929 
2930 	if (vcpu->kvm->arch.crypto.apie)
2931 		vcpu->arch.sie_block->eca |= ECA_APIE;
2932 
2933 	/* Set up protected key support */
2934 	if (vcpu->kvm->arch.crypto.aes_kw) {
2935 		vcpu->arch.sie_block->ecb3 |= ECB3_AES;
2936 		/* ecc is also wrapped with AES key */
2937 		if (kvm_has_pckmo_ecc(vcpu->kvm))
2938 			vcpu->arch.sie_block->ecd |= ECD_ECC;
2939 	}
2940 
2941 	if (vcpu->kvm->arch.crypto.dea_kw)
2942 		vcpu->arch.sie_block->ecb3 |= ECB3_DEA;
2943 }
2944 
2945 void kvm_s390_vcpu_unsetup_cmma(struct kvm_vcpu *vcpu)
2946 {
2947 	free_page(vcpu->arch.sie_block->cbrlo);
2948 	vcpu->arch.sie_block->cbrlo = 0;
2949 }
2950 
2951 int kvm_s390_vcpu_setup_cmma(struct kvm_vcpu *vcpu)
2952 {
2953 	vcpu->arch.sie_block->cbrlo = get_zeroed_page(GFP_KERNEL);
2954 	if (!vcpu->arch.sie_block->cbrlo)
2955 		return -ENOMEM;
2956 	return 0;
2957 }
2958 
2959 static void kvm_s390_vcpu_setup_model(struct kvm_vcpu *vcpu)
2960 {
2961 	struct kvm_s390_cpu_model *model = &vcpu->kvm->arch.model;
2962 
2963 	vcpu->arch.sie_block->ibc = model->ibc;
2964 	if (test_kvm_facility(vcpu->kvm, 7))
2965 		vcpu->arch.sie_block->fac = (u32)(u64) model->fac_list;
2966 }
2967 
2968 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
2969 {
2970 	int rc = 0;
2971 
2972 	atomic_set(&vcpu->arch.sie_block->cpuflags, CPUSTAT_ZARCH |
2973 						    CPUSTAT_SM |
2974 						    CPUSTAT_STOPPED);
2975 
2976 	if (test_kvm_facility(vcpu->kvm, 78))
2977 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED2);
2978 	else if (test_kvm_facility(vcpu->kvm, 8))
2979 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_GED);
2980 
2981 	kvm_s390_vcpu_setup_model(vcpu);
2982 
2983 	/* pgste_set_pte has special handling for !MACHINE_HAS_ESOP */
2984 	if (MACHINE_HAS_ESOP)
2985 		vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
2986 	if (test_kvm_facility(vcpu->kvm, 9))
2987 		vcpu->arch.sie_block->ecb |= ECB_SRSI;
2988 	if (test_kvm_facility(vcpu->kvm, 73))
2989 		vcpu->arch.sie_block->ecb |= ECB_TE;
2990 
2991 	if (test_kvm_facility(vcpu->kvm, 8) && vcpu->kvm->arch.use_pfmfi)
2992 		vcpu->arch.sie_block->ecb2 |= ECB2_PFMFI;
2993 	if (test_kvm_facility(vcpu->kvm, 130))
2994 		vcpu->arch.sie_block->ecb2 |= ECB2_IEP;
2995 	vcpu->arch.sie_block->eca = ECA_MVPGI | ECA_PROTEXCI;
2996 	if (sclp.has_cei)
2997 		vcpu->arch.sie_block->eca |= ECA_CEI;
2998 	if (sclp.has_ib)
2999 		vcpu->arch.sie_block->eca |= ECA_IB;
3000 	if (sclp.has_siif)
3001 		vcpu->arch.sie_block->eca |= ECA_SII;
3002 	if (sclp.has_sigpif)
3003 		vcpu->arch.sie_block->eca |= ECA_SIGPI;
3004 	if (test_kvm_facility(vcpu->kvm, 129)) {
3005 		vcpu->arch.sie_block->eca |= ECA_VX;
3006 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
3007 	}
3008 	if (test_kvm_facility(vcpu->kvm, 139))
3009 		vcpu->arch.sie_block->ecd |= ECD_MEF;
3010 	if (test_kvm_facility(vcpu->kvm, 156))
3011 		vcpu->arch.sie_block->ecd |= ECD_ETOKENF;
3012 	if (vcpu->arch.sie_block->gd) {
3013 		vcpu->arch.sie_block->eca |= ECA_AIV;
3014 		VCPU_EVENT(vcpu, 3, "AIV gisa format-%u enabled for cpu %03u",
3015 			   vcpu->arch.sie_block->gd & 0x3, vcpu->vcpu_id);
3016 	}
3017 	vcpu->arch.sie_block->sdnxo = ((unsigned long) &vcpu->run->s.regs.sdnx)
3018 					| SDNXC;
3019 	vcpu->arch.sie_block->riccbd = (unsigned long) &vcpu->run->s.regs.riccb;
3020 
3021 	if (sclp.has_kss)
3022 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_KSS);
3023 	else
3024 		vcpu->arch.sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE;
3025 
3026 	if (vcpu->kvm->arch.use_cmma) {
3027 		rc = kvm_s390_vcpu_setup_cmma(vcpu);
3028 		if (rc)
3029 			return rc;
3030 	}
3031 	hrtimer_init(&vcpu->arch.ckc_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
3032 	vcpu->arch.ckc_timer.function = kvm_s390_idle_wakeup;
3033 
3034 	vcpu->arch.sie_block->hpid = HPID_KVM;
3035 
3036 	kvm_s390_vcpu_crypto_setup(vcpu);
3037 
3038 	return rc;
3039 }
3040 
3041 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
3042 				      unsigned int id)
3043 {
3044 	struct kvm_vcpu *vcpu;
3045 	struct sie_page *sie_page;
3046 	int rc = -EINVAL;
3047 
3048 	if (!kvm_is_ucontrol(kvm) && !sca_can_add_vcpu(kvm, id))
3049 		goto out;
3050 
3051 	rc = -ENOMEM;
3052 
3053 	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
3054 	if (!vcpu)
3055 		goto out;
3056 
3057 	BUILD_BUG_ON(sizeof(struct sie_page) != 4096);
3058 	sie_page = (struct sie_page *) get_zeroed_page(GFP_KERNEL);
3059 	if (!sie_page)
3060 		goto out_free_cpu;
3061 
3062 	vcpu->arch.sie_block = &sie_page->sie_block;
3063 	vcpu->arch.sie_block->itdba = (unsigned long) &sie_page->itdb;
3064 
3065 	/* the real guest size will always be smaller than msl */
3066 	vcpu->arch.sie_block->mso = 0;
3067 	vcpu->arch.sie_block->msl = sclp.hamax;
3068 
3069 	vcpu->arch.sie_block->icpua = id;
3070 	spin_lock_init(&vcpu->arch.local_int.lock);
3071 	vcpu->arch.sie_block->gd = (u32)(u64)kvm->arch.gisa_int.origin;
3072 	if (vcpu->arch.sie_block->gd && sclp.has_gisaf)
3073 		vcpu->arch.sie_block->gd |= GISA_FORMAT1;
3074 	seqcount_init(&vcpu->arch.cputm_seqcount);
3075 
3076 	rc = kvm_vcpu_init(vcpu, kvm, id);
3077 	if (rc)
3078 		goto out_free_sie_block;
3079 	VM_EVENT(kvm, 3, "create cpu %d at 0x%pK, sie block at 0x%pK", id, vcpu,
3080 		 vcpu->arch.sie_block);
3081 	trace_kvm_s390_create_vcpu(id, vcpu, vcpu->arch.sie_block);
3082 
3083 	return vcpu;
3084 out_free_sie_block:
3085 	free_page((unsigned long)(vcpu->arch.sie_block));
3086 out_free_cpu:
3087 	kmem_cache_free(kvm_vcpu_cache, vcpu);
3088 out:
3089 	return ERR_PTR(rc);
3090 }
3091 
3092 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
3093 {
3094 	return kvm_s390_vcpu_has_irq(vcpu, 0);
3095 }
3096 
3097 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu)
3098 {
3099 	return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE);
3100 }
3101 
3102 void kvm_s390_vcpu_block(struct kvm_vcpu *vcpu)
3103 {
3104 	atomic_or(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
3105 	exit_sie(vcpu);
3106 }
3107 
3108 void kvm_s390_vcpu_unblock(struct kvm_vcpu *vcpu)
3109 {
3110 	atomic_andnot(PROG_BLOCK_SIE, &vcpu->arch.sie_block->prog20);
3111 }
3112 
3113 static void kvm_s390_vcpu_request(struct kvm_vcpu *vcpu)
3114 {
3115 	atomic_or(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
3116 	exit_sie(vcpu);
3117 }
3118 
3119 bool kvm_s390_vcpu_sie_inhibited(struct kvm_vcpu *vcpu)
3120 {
3121 	return atomic_read(&vcpu->arch.sie_block->prog20) &
3122 	       (PROG_BLOCK_SIE | PROG_REQUEST);
3123 }
3124 
3125 static void kvm_s390_vcpu_request_handled(struct kvm_vcpu *vcpu)
3126 {
3127 	atomic_andnot(PROG_REQUEST, &vcpu->arch.sie_block->prog20);
3128 }
3129 
3130 /*
3131  * Kick a guest cpu out of (v)SIE and wait until (v)SIE is not running.
3132  * If the CPU is not running (e.g. waiting as idle) the function will
3133  * return immediately. */
3134 void exit_sie(struct kvm_vcpu *vcpu)
3135 {
3136 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOP_INT);
3137 	kvm_s390_vsie_kick(vcpu);
3138 	while (vcpu->arch.sie_block->prog0c & PROG_IN_SIE)
3139 		cpu_relax();
3140 }
3141 
3142 /* Kick a guest cpu out of SIE to process a request synchronously */
3143 void kvm_s390_sync_request(int req, struct kvm_vcpu *vcpu)
3144 {
3145 	kvm_make_request(req, vcpu);
3146 	kvm_s390_vcpu_request(vcpu);
3147 }
3148 
3149 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
3150 			      unsigned long end)
3151 {
3152 	struct kvm *kvm = gmap->private;
3153 	struct kvm_vcpu *vcpu;
3154 	unsigned long prefix;
3155 	int i;
3156 
3157 	if (gmap_is_shadow(gmap))
3158 		return;
3159 	if (start >= 1UL << 31)
3160 		/* We are only interested in prefix pages */
3161 		return;
3162 	kvm_for_each_vcpu(i, vcpu, kvm) {
3163 		/* match against both prefix pages */
3164 		prefix = kvm_s390_get_prefix(vcpu);
3165 		if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
3166 			VCPU_EVENT(vcpu, 2, "gmap notifier for %lx-%lx",
3167 				   start, end);
3168 			kvm_s390_sync_request(KVM_REQ_MMU_RELOAD, vcpu);
3169 		}
3170 	}
3171 }
3172 
3173 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
3174 {
3175 	/* do not poll with more than halt_poll_max_steal percent of steal time */
3176 	if (S390_lowcore.avg_steal_timer * 100 / (TICK_USEC << 12) >=
3177 	    halt_poll_max_steal) {
3178 		vcpu->stat.halt_no_poll_steal++;
3179 		return true;
3180 	}
3181 	return false;
3182 }
3183 
3184 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
3185 {
3186 	/* kvm common code refers to this, but never calls it */
3187 	BUG();
3188 	return 0;
3189 }
3190 
3191 static int kvm_arch_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu,
3192 					   struct kvm_one_reg *reg)
3193 {
3194 	int r = -EINVAL;
3195 
3196 	switch (reg->id) {
3197 	case KVM_REG_S390_TODPR:
3198 		r = put_user(vcpu->arch.sie_block->todpr,
3199 			     (u32 __user *)reg->addr);
3200 		break;
3201 	case KVM_REG_S390_EPOCHDIFF:
3202 		r = put_user(vcpu->arch.sie_block->epoch,
3203 			     (u64 __user *)reg->addr);
3204 		break;
3205 	case KVM_REG_S390_CPU_TIMER:
3206 		r = put_user(kvm_s390_get_cpu_timer(vcpu),
3207 			     (u64 __user *)reg->addr);
3208 		break;
3209 	case KVM_REG_S390_CLOCK_COMP:
3210 		r = put_user(vcpu->arch.sie_block->ckc,
3211 			     (u64 __user *)reg->addr);
3212 		break;
3213 	case KVM_REG_S390_PFTOKEN:
3214 		r = put_user(vcpu->arch.pfault_token,
3215 			     (u64 __user *)reg->addr);
3216 		break;
3217 	case KVM_REG_S390_PFCOMPARE:
3218 		r = put_user(vcpu->arch.pfault_compare,
3219 			     (u64 __user *)reg->addr);
3220 		break;
3221 	case KVM_REG_S390_PFSELECT:
3222 		r = put_user(vcpu->arch.pfault_select,
3223 			     (u64 __user *)reg->addr);
3224 		break;
3225 	case KVM_REG_S390_PP:
3226 		r = put_user(vcpu->arch.sie_block->pp,
3227 			     (u64 __user *)reg->addr);
3228 		break;
3229 	case KVM_REG_S390_GBEA:
3230 		r = put_user(vcpu->arch.sie_block->gbea,
3231 			     (u64 __user *)reg->addr);
3232 		break;
3233 	default:
3234 		break;
3235 	}
3236 
3237 	return r;
3238 }
3239 
3240 static int kvm_arch_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu,
3241 					   struct kvm_one_reg *reg)
3242 {
3243 	int r = -EINVAL;
3244 	__u64 val;
3245 
3246 	switch (reg->id) {
3247 	case KVM_REG_S390_TODPR:
3248 		r = get_user(vcpu->arch.sie_block->todpr,
3249 			     (u32 __user *)reg->addr);
3250 		break;
3251 	case KVM_REG_S390_EPOCHDIFF:
3252 		r = get_user(vcpu->arch.sie_block->epoch,
3253 			     (u64 __user *)reg->addr);
3254 		break;
3255 	case KVM_REG_S390_CPU_TIMER:
3256 		r = get_user(val, (u64 __user *)reg->addr);
3257 		if (!r)
3258 			kvm_s390_set_cpu_timer(vcpu, val);
3259 		break;
3260 	case KVM_REG_S390_CLOCK_COMP:
3261 		r = get_user(vcpu->arch.sie_block->ckc,
3262 			     (u64 __user *)reg->addr);
3263 		break;
3264 	case KVM_REG_S390_PFTOKEN:
3265 		r = get_user(vcpu->arch.pfault_token,
3266 			     (u64 __user *)reg->addr);
3267 		if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
3268 			kvm_clear_async_pf_completion_queue(vcpu);
3269 		break;
3270 	case KVM_REG_S390_PFCOMPARE:
3271 		r = get_user(vcpu->arch.pfault_compare,
3272 			     (u64 __user *)reg->addr);
3273 		break;
3274 	case KVM_REG_S390_PFSELECT:
3275 		r = get_user(vcpu->arch.pfault_select,
3276 			     (u64 __user *)reg->addr);
3277 		break;
3278 	case KVM_REG_S390_PP:
3279 		r = get_user(vcpu->arch.sie_block->pp,
3280 			     (u64 __user *)reg->addr);
3281 		break;
3282 	case KVM_REG_S390_GBEA:
3283 		r = get_user(vcpu->arch.sie_block->gbea,
3284 			     (u64 __user *)reg->addr);
3285 		break;
3286 	default:
3287 		break;
3288 	}
3289 
3290 	return r;
3291 }
3292 
3293 static int kvm_arch_vcpu_ioctl_initial_reset(struct kvm_vcpu *vcpu)
3294 {
3295 	kvm_s390_vcpu_initial_reset(vcpu);
3296 	return 0;
3297 }
3298 
3299 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
3300 {
3301 	vcpu_load(vcpu);
3302 	memcpy(&vcpu->run->s.regs.gprs, &regs->gprs, sizeof(regs->gprs));
3303 	vcpu_put(vcpu);
3304 	return 0;
3305 }
3306 
3307 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
3308 {
3309 	vcpu_load(vcpu);
3310 	memcpy(&regs->gprs, &vcpu->run->s.regs.gprs, sizeof(regs->gprs));
3311 	vcpu_put(vcpu);
3312 	return 0;
3313 }
3314 
3315 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
3316 				  struct kvm_sregs *sregs)
3317 {
3318 	vcpu_load(vcpu);
3319 
3320 	memcpy(&vcpu->run->s.regs.acrs, &sregs->acrs, sizeof(sregs->acrs));
3321 	memcpy(&vcpu->arch.sie_block->gcr, &sregs->crs, sizeof(sregs->crs));
3322 
3323 	vcpu_put(vcpu);
3324 	return 0;
3325 }
3326 
3327 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
3328 				  struct kvm_sregs *sregs)
3329 {
3330 	vcpu_load(vcpu);
3331 
3332 	memcpy(&sregs->acrs, &vcpu->run->s.regs.acrs, sizeof(sregs->acrs));
3333 	memcpy(&sregs->crs, &vcpu->arch.sie_block->gcr, sizeof(sregs->crs));
3334 
3335 	vcpu_put(vcpu);
3336 	return 0;
3337 }
3338 
3339 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
3340 {
3341 	int ret = 0;
3342 
3343 	vcpu_load(vcpu);
3344 
3345 	if (test_fp_ctl(fpu->fpc)) {
3346 		ret = -EINVAL;
3347 		goto out;
3348 	}
3349 	vcpu->run->s.regs.fpc = fpu->fpc;
3350 	if (MACHINE_HAS_VX)
3351 		convert_fp_to_vx((__vector128 *) vcpu->run->s.regs.vrs,
3352 				 (freg_t *) fpu->fprs);
3353 	else
3354 		memcpy(vcpu->run->s.regs.fprs, &fpu->fprs, sizeof(fpu->fprs));
3355 
3356 out:
3357 	vcpu_put(vcpu);
3358 	return ret;
3359 }
3360 
3361 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
3362 {
3363 	vcpu_load(vcpu);
3364 
3365 	/* make sure we have the latest values */
3366 	save_fpu_regs();
3367 	if (MACHINE_HAS_VX)
3368 		convert_vx_to_fp((freg_t *) fpu->fprs,
3369 				 (__vector128 *) vcpu->run->s.regs.vrs);
3370 	else
3371 		memcpy(fpu->fprs, vcpu->run->s.regs.fprs, sizeof(fpu->fprs));
3372 	fpu->fpc = vcpu->run->s.regs.fpc;
3373 
3374 	vcpu_put(vcpu);
3375 	return 0;
3376 }
3377 
3378 static int kvm_arch_vcpu_ioctl_set_initial_psw(struct kvm_vcpu *vcpu, psw_t psw)
3379 {
3380 	int rc = 0;
3381 
3382 	if (!is_vcpu_stopped(vcpu))
3383 		rc = -EBUSY;
3384 	else {
3385 		vcpu->run->psw_mask = psw.mask;
3386 		vcpu->run->psw_addr = psw.addr;
3387 	}
3388 	return rc;
3389 }
3390 
3391 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
3392 				  struct kvm_translation *tr)
3393 {
3394 	return -EINVAL; /* not implemented yet */
3395 }
3396 
3397 #define VALID_GUESTDBG_FLAGS (KVM_GUESTDBG_SINGLESTEP | \
3398 			      KVM_GUESTDBG_USE_HW_BP | \
3399 			      KVM_GUESTDBG_ENABLE)
3400 
3401 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
3402 					struct kvm_guest_debug *dbg)
3403 {
3404 	int rc = 0;
3405 
3406 	vcpu_load(vcpu);
3407 
3408 	vcpu->guest_debug = 0;
3409 	kvm_s390_clear_bp_data(vcpu);
3410 
3411 	if (dbg->control & ~VALID_GUESTDBG_FLAGS) {
3412 		rc = -EINVAL;
3413 		goto out;
3414 	}
3415 	if (!sclp.has_gpere) {
3416 		rc = -EINVAL;
3417 		goto out;
3418 	}
3419 
3420 	if (dbg->control & KVM_GUESTDBG_ENABLE) {
3421 		vcpu->guest_debug = dbg->control;
3422 		/* enforce guest PER */
3423 		kvm_s390_set_cpuflags(vcpu, CPUSTAT_P);
3424 
3425 		if (dbg->control & KVM_GUESTDBG_USE_HW_BP)
3426 			rc = kvm_s390_import_bp_data(vcpu, dbg);
3427 	} else {
3428 		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
3429 		vcpu->arch.guestdbg.last_bp = 0;
3430 	}
3431 
3432 	if (rc) {
3433 		vcpu->guest_debug = 0;
3434 		kvm_s390_clear_bp_data(vcpu);
3435 		kvm_s390_clear_cpuflags(vcpu, CPUSTAT_P);
3436 	}
3437 
3438 out:
3439 	vcpu_put(vcpu);
3440 	return rc;
3441 }
3442 
3443 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
3444 				    struct kvm_mp_state *mp_state)
3445 {
3446 	int ret;
3447 
3448 	vcpu_load(vcpu);
3449 
3450 	/* CHECK_STOP and LOAD are not supported yet */
3451 	ret = is_vcpu_stopped(vcpu) ? KVM_MP_STATE_STOPPED :
3452 				      KVM_MP_STATE_OPERATING;
3453 
3454 	vcpu_put(vcpu);
3455 	return ret;
3456 }
3457 
3458 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
3459 				    struct kvm_mp_state *mp_state)
3460 {
3461 	int rc = 0;
3462 
3463 	vcpu_load(vcpu);
3464 
3465 	/* user space knows about this interface - let it control the state */
3466 	vcpu->kvm->arch.user_cpu_state_ctrl = 1;
3467 
3468 	switch (mp_state->mp_state) {
3469 	case KVM_MP_STATE_STOPPED:
3470 		kvm_s390_vcpu_stop(vcpu);
3471 		break;
3472 	case KVM_MP_STATE_OPERATING:
3473 		kvm_s390_vcpu_start(vcpu);
3474 		break;
3475 	case KVM_MP_STATE_LOAD:
3476 	case KVM_MP_STATE_CHECK_STOP:
3477 		/* fall through - CHECK_STOP and LOAD are not supported yet */
3478 	default:
3479 		rc = -ENXIO;
3480 	}
3481 
3482 	vcpu_put(vcpu);
3483 	return rc;
3484 }
3485 
3486 static bool ibs_enabled(struct kvm_vcpu *vcpu)
3487 {
3488 	return kvm_s390_test_cpuflags(vcpu, CPUSTAT_IBS);
3489 }
3490 
3491 static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
3492 {
3493 retry:
3494 	kvm_s390_vcpu_request_handled(vcpu);
3495 	if (!kvm_request_pending(vcpu))
3496 		return 0;
3497 	/*
3498 	 * We use MMU_RELOAD just to re-arm the ipte notifier for the
3499 	 * guest prefix page. gmap_mprotect_notify will wait on the ptl lock.
3500 	 * This ensures that the ipte instruction for this request has
3501 	 * already finished. We might race against a second unmapper that
3502 	 * wants to set the blocking bit. Lets just retry the request loop.
3503 	 */
3504 	if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu)) {
3505 		int rc;
3506 		rc = gmap_mprotect_notify(vcpu->arch.gmap,
3507 					  kvm_s390_get_prefix(vcpu),
3508 					  PAGE_SIZE * 2, PROT_WRITE);
3509 		if (rc) {
3510 			kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu);
3511 			return rc;
3512 		}
3513 		goto retry;
3514 	}
3515 
3516 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
3517 		vcpu->arch.sie_block->ihcpu = 0xffff;
3518 		goto retry;
3519 	}
3520 
3521 	if (kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu)) {
3522 		if (!ibs_enabled(vcpu)) {
3523 			trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 1);
3524 			kvm_s390_set_cpuflags(vcpu, CPUSTAT_IBS);
3525 		}
3526 		goto retry;
3527 	}
3528 
3529 	if (kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu)) {
3530 		if (ibs_enabled(vcpu)) {
3531 			trace_kvm_s390_enable_disable_ibs(vcpu->vcpu_id, 0);
3532 			kvm_s390_clear_cpuflags(vcpu, CPUSTAT_IBS);
3533 		}
3534 		goto retry;
3535 	}
3536 
3537 	if (kvm_check_request(KVM_REQ_ICPT_OPEREXC, vcpu)) {
3538 		vcpu->arch.sie_block->ictl |= ICTL_OPEREXC;
3539 		goto retry;
3540 	}
3541 
3542 	if (kvm_check_request(KVM_REQ_START_MIGRATION, vcpu)) {
3543 		/*
3544 		 * Disable CMM virtualization; we will emulate the ESSA
3545 		 * instruction manually, in order to provide additional
3546 		 * functionalities needed for live migration.
3547 		 */
3548 		vcpu->arch.sie_block->ecb2 &= ~ECB2_CMMA;
3549 		goto retry;
3550 	}
3551 
3552 	if (kvm_check_request(KVM_REQ_STOP_MIGRATION, vcpu)) {
3553 		/*
3554 		 * Re-enable CMM virtualization if CMMA is available and
3555 		 * CMM has been used.
3556 		 */
3557 		if ((vcpu->kvm->arch.use_cmma) &&
3558 		    (vcpu->kvm->mm->context.uses_cmm))
3559 			vcpu->arch.sie_block->ecb2 |= ECB2_CMMA;
3560 		goto retry;
3561 	}
3562 
3563 	/* nothing to do, just clear the request */
3564 	kvm_clear_request(KVM_REQ_UNHALT, vcpu);
3565 	/* we left the vsie handler, nothing to do, just clear the request */
3566 	kvm_clear_request(KVM_REQ_VSIE_RESTART, vcpu);
3567 
3568 	return 0;
3569 }
3570 
3571 void kvm_s390_set_tod_clock(struct kvm *kvm,
3572 			    const struct kvm_s390_vm_tod_clock *gtod)
3573 {
3574 	struct kvm_vcpu *vcpu;
3575 	struct kvm_s390_tod_clock_ext htod;
3576 	int i;
3577 
3578 	mutex_lock(&kvm->lock);
3579 	preempt_disable();
3580 
3581 	get_tod_clock_ext((char *)&htod);
3582 
3583 	kvm->arch.epoch = gtod->tod - htod.tod;
3584 	kvm->arch.epdx = 0;
3585 	if (test_kvm_facility(kvm, 139)) {
3586 		kvm->arch.epdx = gtod->epoch_idx - htod.epoch_idx;
3587 		if (kvm->arch.epoch > gtod->tod)
3588 			kvm->arch.epdx -= 1;
3589 	}
3590 
3591 	kvm_s390_vcpu_block_all(kvm);
3592 	kvm_for_each_vcpu(i, vcpu, kvm) {
3593 		vcpu->arch.sie_block->epoch = kvm->arch.epoch;
3594 		vcpu->arch.sie_block->epdx  = kvm->arch.epdx;
3595 	}
3596 
3597 	kvm_s390_vcpu_unblock_all(kvm);
3598 	preempt_enable();
3599 	mutex_unlock(&kvm->lock);
3600 }
3601 
3602 /**
3603  * kvm_arch_fault_in_page - fault-in guest page if necessary
3604  * @vcpu: The corresponding virtual cpu
3605  * @gpa: Guest physical address
3606  * @writable: Whether the page should be writable or not
3607  *
3608  * Make sure that a guest page has been faulted-in on the host.
3609  *
3610  * Return: Zero on success, negative error code otherwise.
3611  */
3612 long kvm_arch_fault_in_page(struct kvm_vcpu *vcpu, gpa_t gpa, int writable)
3613 {
3614 	return gmap_fault(vcpu->arch.gmap, gpa,
3615 			  writable ? FAULT_FLAG_WRITE : 0);
3616 }
3617 
3618 static void __kvm_inject_pfault_token(struct kvm_vcpu *vcpu, bool start_token,
3619 				      unsigned long token)
3620 {
3621 	struct kvm_s390_interrupt inti;
3622 	struct kvm_s390_irq irq;
3623 
3624 	if (start_token) {
3625 		irq.u.ext.ext_params2 = token;
3626 		irq.type = KVM_S390_INT_PFAULT_INIT;
3627 		WARN_ON_ONCE(kvm_s390_inject_vcpu(vcpu, &irq));
3628 	} else {
3629 		inti.type = KVM_S390_INT_PFAULT_DONE;
3630 		inti.parm64 = token;
3631 		WARN_ON_ONCE(kvm_s390_inject_vm(vcpu->kvm, &inti));
3632 	}
3633 }
3634 
3635 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
3636 				     struct kvm_async_pf *work)
3637 {
3638 	trace_kvm_s390_pfault_init(vcpu, work->arch.pfault_token);
3639 	__kvm_inject_pfault_token(vcpu, true, work->arch.pfault_token);
3640 }
3641 
3642 void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
3643 				 struct kvm_async_pf *work)
3644 {
3645 	trace_kvm_s390_pfault_done(vcpu, work->arch.pfault_token);
3646 	__kvm_inject_pfault_token(vcpu, false, work->arch.pfault_token);
3647 }
3648 
3649 void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu,
3650 			       struct kvm_async_pf *work)
3651 {
3652 	/* s390 will always inject the page directly */
3653 }
3654 
3655 bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu)
3656 {
3657 	/*
3658 	 * s390 will always inject the page directly,
3659 	 * but we still want check_async_completion to cleanup
3660 	 */
3661 	return true;
3662 }
3663 
3664 static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu)
3665 {
3666 	hva_t hva;
3667 	struct kvm_arch_async_pf arch;
3668 	int rc;
3669 
3670 	if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
3671 		return 0;
3672 	if ((vcpu->arch.sie_block->gpsw.mask & vcpu->arch.pfault_select) !=
3673 	    vcpu->arch.pfault_compare)
3674 		return 0;
3675 	if (psw_extint_disabled(vcpu))
3676 		return 0;
3677 	if (kvm_s390_vcpu_has_irq(vcpu, 0))
3678 		return 0;
3679 	if (!(vcpu->arch.sie_block->gcr[0] & CR0_SERVICE_SIGNAL_SUBMASK))
3680 		return 0;
3681 	if (!vcpu->arch.gmap->pfault_enabled)
3682 		return 0;
3683 
3684 	hva = gfn_to_hva(vcpu->kvm, gpa_to_gfn(current->thread.gmap_addr));
3685 	hva += current->thread.gmap_addr & ~PAGE_MASK;
3686 	if (read_guest_real(vcpu, vcpu->arch.pfault_token, &arch.pfault_token, 8))
3687 		return 0;
3688 
3689 	rc = kvm_setup_async_pf(vcpu, current->thread.gmap_addr, hva, &arch);
3690 	return rc;
3691 }
3692 
3693 static int vcpu_pre_run(struct kvm_vcpu *vcpu)
3694 {
3695 	int rc, cpuflags;
3696 
3697 	/*
3698 	 * On s390 notifications for arriving pages will be delivered directly
3699 	 * to the guest but the house keeping for completed pfaults is
3700 	 * handled outside the worker.
3701 	 */
3702 	kvm_check_async_pf_completion(vcpu);
3703 
3704 	vcpu->arch.sie_block->gg14 = vcpu->run->s.regs.gprs[14];
3705 	vcpu->arch.sie_block->gg15 = vcpu->run->s.regs.gprs[15];
3706 
3707 	if (need_resched())
3708 		schedule();
3709 
3710 	if (test_cpu_flag(CIF_MCCK_PENDING))
3711 		s390_handle_mcck();
3712 
3713 	if (!kvm_is_ucontrol(vcpu->kvm)) {
3714 		rc = kvm_s390_deliver_pending_interrupts(vcpu);
3715 		if (rc)
3716 			return rc;
3717 	}
3718 
3719 	rc = kvm_s390_handle_requests(vcpu);
3720 	if (rc)
3721 		return rc;
3722 
3723 	if (guestdbg_enabled(vcpu)) {
3724 		kvm_s390_backup_guest_per_regs(vcpu);
3725 		kvm_s390_patch_guest_per_regs(vcpu);
3726 	}
3727 
3728 	clear_bit(vcpu->vcpu_id, vcpu->kvm->arch.gisa_int.kicked_mask);
3729 
3730 	vcpu->arch.sie_block->icptcode = 0;
3731 	cpuflags = atomic_read(&vcpu->arch.sie_block->cpuflags);
3732 	VCPU_EVENT(vcpu, 6, "entering sie flags %x", cpuflags);
3733 	trace_kvm_s390_sie_enter(vcpu, cpuflags);
3734 
3735 	return 0;
3736 }
3737 
3738 static int vcpu_post_run_fault_in_sie(struct kvm_vcpu *vcpu)
3739 {
3740 	struct kvm_s390_pgm_info pgm_info = {
3741 		.code = PGM_ADDRESSING,
3742 	};
3743 	u8 opcode, ilen;
3744 	int rc;
3745 
3746 	VCPU_EVENT(vcpu, 3, "%s", "fault in sie instruction");
3747 	trace_kvm_s390_sie_fault(vcpu);
3748 
3749 	/*
3750 	 * We want to inject an addressing exception, which is defined as a
3751 	 * suppressing or terminating exception. However, since we came here
3752 	 * by a DAT access exception, the PSW still points to the faulting
3753 	 * instruction since DAT exceptions are nullifying. So we've got
3754 	 * to look up the current opcode to get the length of the instruction
3755 	 * to be able to forward the PSW.
3756 	 */
3757 	rc = read_guest_instr(vcpu, vcpu->arch.sie_block->gpsw.addr, &opcode, 1);
3758 	ilen = insn_length(opcode);
3759 	if (rc < 0) {
3760 		return rc;
3761 	} else if (rc) {
3762 		/* Instruction-Fetching Exceptions - we can't detect the ilen.
3763 		 * Forward by arbitrary ilc, injection will take care of
3764 		 * nullification if necessary.
3765 		 */
3766 		pgm_info = vcpu->arch.pgm;
3767 		ilen = 4;
3768 	}
3769 	pgm_info.flags = ilen | KVM_S390_PGM_FLAGS_ILC_VALID;
3770 	kvm_s390_forward_psw(vcpu, ilen);
3771 	return kvm_s390_inject_prog_irq(vcpu, &pgm_info);
3772 }
3773 
3774 static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason)
3775 {
3776 	struct mcck_volatile_info *mcck_info;
3777 	struct sie_page *sie_page;
3778 
3779 	VCPU_EVENT(vcpu, 6, "exit sie icptcode %d",
3780 		   vcpu->arch.sie_block->icptcode);
3781 	trace_kvm_s390_sie_exit(vcpu, vcpu->arch.sie_block->icptcode);
3782 
3783 	if (guestdbg_enabled(vcpu))
3784 		kvm_s390_restore_guest_per_regs(vcpu);
3785 
3786 	vcpu->run->s.regs.gprs[14] = vcpu->arch.sie_block->gg14;
3787 	vcpu->run->s.regs.gprs[15] = vcpu->arch.sie_block->gg15;
3788 
3789 	if (exit_reason == -EINTR) {
3790 		VCPU_EVENT(vcpu, 3, "%s", "machine check");
3791 		sie_page = container_of(vcpu->arch.sie_block,
3792 					struct sie_page, sie_block);
3793 		mcck_info = &sie_page->mcck_info;
3794 		kvm_s390_reinject_machine_check(vcpu, mcck_info);
3795 		return 0;
3796 	}
3797 
3798 	if (vcpu->arch.sie_block->icptcode > 0) {
3799 		int rc = kvm_handle_sie_intercept(vcpu);
3800 
3801 		if (rc != -EOPNOTSUPP)
3802 			return rc;
3803 		vcpu->run->exit_reason = KVM_EXIT_S390_SIEIC;
3804 		vcpu->run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode;
3805 		vcpu->run->s390_sieic.ipa = vcpu->arch.sie_block->ipa;
3806 		vcpu->run->s390_sieic.ipb = vcpu->arch.sie_block->ipb;
3807 		return -EREMOTE;
3808 	} else if (exit_reason != -EFAULT) {
3809 		vcpu->stat.exit_null++;
3810 		return 0;
3811 	} else if (kvm_is_ucontrol(vcpu->kvm)) {
3812 		vcpu->run->exit_reason = KVM_EXIT_S390_UCONTROL;
3813 		vcpu->run->s390_ucontrol.trans_exc_code =
3814 						current->thread.gmap_addr;
3815 		vcpu->run->s390_ucontrol.pgm_code = 0x10;
3816 		return -EREMOTE;
3817 	} else if (current->thread.gmap_pfault) {
3818 		trace_kvm_s390_major_guest_pfault(vcpu);
3819 		current->thread.gmap_pfault = 0;
3820 		if (kvm_arch_setup_async_pf(vcpu))
3821 			return 0;
3822 		return kvm_arch_fault_in_page(vcpu, current->thread.gmap_addr, 1);
3823 	}
3824 	return vcpu_post_run_fault_in_sie(vcpu);
3825 }
3826 
3827 static int __vcpu_run(struct kvm_vcpu *vcpu)
3828 {
3829 	int rc, exit_reason;
3830 
3831 	/*
3832 	 * We try to hold kvm->srcu during most of vcpu_run (except when run-
3833 	 * ning the guest), so that memslots (and other stuff) are protected
3834 	 */
3835 	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
3836 
3837 	do {
3838 		rc = vcpu_pre_run(vcpu);
3839 		if (rc)
3840 			break;
3841 
3842 		srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
3843 		/*
3844 		 * As PF_VCPU will be used in fault handler, between
3845 		 * guest_enter and guest_exit should be no uaccess.
3846 		 */
3847 		local_irq_disable();
3848 		guest_enter_irqoff();
3849 		__disable_cpu_timer_accounting(vcpu);
3850 		local_irq_enable();
3851 		exit_reason = sie64a(vcpu->arch.sie_block,
3852 				     vcpu->run->s.regs.gprs);
3853 		local_irq_disable();
3854 		__enable_cpu_timer_accounting(vcpu);
3855 		guest_exit_irqoff();
3856 		local_irq_enable();
3857 		vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
3858 
3859 		rc = vcpu_post_run(vcpu, exit_reason);
3860 	} while (!signal_pending(current) && !guestdbg_exit_pending(vcpu) && !rc);
3861 
3862 	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
3863 	return rc;
3864 }
3865 
3866 static void sync_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3867 {
3868 	struct runtime_instr_cb *riccb;
3869 	struct gs_cb *gscb;
3870 
3871 	riccb = (struct runtime_instr_cb *) &kvm_run->s.regs.riccb;
3872 	gscb = (struct gs_cb *) &kvm_run->s.regs.gscb;
3873 	vcpu->arch.sie_block->gpsw.mask = kvm_run->psw_mask;
3874 	vcpu->arch.sie_block->gpsw.addr = kvm_run->psw_addr;
3875 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_PREFIX)
3876 		kvm_s390_set_prefix(vcpu, kvm_run->s.regs.prefix);
3877 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_CRS) {
3878 		memcpy(&vcpu->arch.sie_block->gcr, &kvm_run->s.regs.crs, 128);
3879 		/* some control register changes require a tlb flush */
3880 		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
3881 	}
3882 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_ARCH0) {
3883 		kvm_s390_set_cpu_timer(vcpu, kvm_run->s.regs.cputm);
3884 		vcpu->arch.sie_block->ckc = kvm_run->s.regs.ckc;
3885 		vcpu->arch.sie_block->todpr = kvm_run->s.regs.todpr;
3886 		vcpu->arch.sie_block->pp = kvm_run->s.regs.pp;
3887 		vcpu->arch.sie_block->gbea = kvm_run->s.regs.gbea;
3888 	}
3889 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_PFAULT) {
3890 		vcpu->arch.pfault_token = kvm_run->s.regs.pft;
3891 		vcpu->arch.pfault_select = kvm_run->s.regs.pfs;
3892 		vcpu->arch.pfault_compare = kvm_run->s.regs.pfc;
3893 		if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
3894 			kvm_clear_async_pf_completion_queue(vcpu);
3895 	}
3896 	/*
3897 	 * If userspace sets the riccb (e.g. after migration) to a valid state,
3898 	 * we should enable RI here instead of doing the lazy enablement.
3899 	 */
3900 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_RICCB) &&
3901 	    test_kvm_facility(vcpu->kvm, 64) &&
3902 	    riccb->v &&
3903 	    !(vcpu->arch.sie_block->ecb3 & ECB3_RI)) {
3904 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: RI (sync_regs)");
3905 		vcpu->arch.sie_block->ecb3 |= ECB3_RI;
3906 	}
3907 	/*
3908 	 * If userspace sets the gscb (e.g. after migration) to non-zero,
3909 	 * we should enable GS here instead of doing the lazy enablement.
3910 	 */
3911 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_GSCB) &&
3912 	    test_kvm_facility(vcpu->kvm, 133) &&
3913 	    gscb->gssm &&
3914 	    !vcpu->arch.gs_enabled) {
3915 		VCPU_EVENT(vcpu, 3, "%s", "ENABLE: GS (sync_regs)");
3916 		vcpu->arch.sie_block->ecb |= ECB_GS;
3917 		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
3918 		vcpu->arch.gs_enabled = 1;
3919 	}
3920 	if ((kvm_run->kvm_dirty_regs & KVM_SYNC_BPBC) &&
3921 	    test_kvm_facility(vcpu->kvm, 82)) {
3922 		vcpu->arch.sie_block->fpf &= ~FPF_BPBC;
3923 		vcpu->arch.sie_block->fpf |= kvm_run->s.regs.bpbc ? FPF_BPBC : 0;
3924 	}
3925 	save_access_regs(vcpu->arch.host_acrs);
3926 	restore_access_regs(vcpu->run->s.regs.acrs);
3927 	/* save host (userspace) fprs/vrs */
3928 	save_fpu_regs();
3929 	vcpu->arch.host_fpregs.fpc = current->thread.fpu.fpc;
3930 	vcpu->arch.host_fpregs.regs = current->thread.fpu.regs;
3931 	if (MACHINE_HAS_VX)
3932 		current->thread.fpu.regs = vcpu->run->s.regs.vrs;
3933 	else
3934 		current->thread.fpu.regs = vcpu->run->s.regs.fprs;
3935 	current->thread.fpu.fpc = vcpu->run->s.regs.fpc;
3936 	if (test_fp_ctl(current->thread.fpu.fpc))
3937 		/* User space provided an invalid FPC, let's clear it */
3938 		current->thread.fpu.fpc = 0;
3939 	if (MACHINE_HAS_GS) {
3940 		preempt_disable();
3941 		__ctl_set_bit(2, 4);
3942 		if (current->thread.gs_cb) {
3943 			vcpu->arch.host_gscb = current->thread.gs_cb;
3944 			save_gs_cb(vcpu->arch.host_gscb);
3945 		}
3946 		if (vcpu->arch.gs_enabled) {
3947 			current->thread.gs_cb = (struct gs_cb *)
3948 						&vcpu->run->s.regs.gscb;
3949 			restore_gs_cb(current->thread.gs_cb);
3950 		}
3951 		preempt_enable();
3952 	}
3953 	/* SIE will load etoken directly from SDNX and therefore kvm_run */
3954 
3955 	kvm_run->kvm_dirty_regs = 0;
3956 }
3957 
3958 static void store_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3959 {
3960 	kvm_run->psw_mask = vcpu->arch.sie_block->gpsw.mask;
3961 	kvm_run->psw_addr = vcpu->arch.sie_block->gpsw.addr;
3962 	kvm_run->s.regs.prefix = kvm_s390_get_prefix(vcpu);
3963 	memcpy(&kvm_run->s.regs.crs, &vcpu->arch.sie_block->gcr, 128);
3964 	kvm_run->s.regs.cputm = kvm_s390_get_cpu_timer(vcpu);
3965 	kvm_run->s.regs.ckc = vcpu->arch.sie_block->ckc;
3966 	kvm_run->s.regs.todpr = vcpu->arch.sie_block->todpr;
3967 	kvm_run->s.regs.pp = vcpu->arch.sie_block->pp;
3968 	kvm_run->s.regs.gbea = vcpu->arch.sie_block->gbea;
3969 	kvm_run->s.regs.pft = vcpu->arch.pfault_token;
3970 	kvm_run->s.regs.pfs = vcpu->arch.pfault_select;
3971 	kvm_run->s.regs.pfc = vcpu->arch.pfault_compare;
3972 	kvm_run->s.regs.bpbc = (vcpu->arch.sie_block->fpf & FPF_BPBC) == FPF_BPBC;
3973 	save_access_regs(vcpu->run->s.regs.acrs);
3974 	restore_access_regs(vcpu->arch.host_acrs);
3975 	/* Save guest register state */
3976 	save_fpu_regs();
3977 	vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
3978 	/* Restore will be done lazily at return */
3979 	current->thread.fpu.fpc = vcpu->arch.host_fpregs.fpc;
3980 	current->thread.fpu.regs = vcpu->arch.host_fpregs.regs;
3981 	if (MACHINE_HAS_GS) {
3982 		__ctl_set_bit(2, 4);
3983 		if (vcpu->arch.gs_enabled)
3984 			save_gs_cb(current->thread.gs_cb);
3985 		preempt_disable();
3986 		current->thread.gs_cb = vcpu->arch.host_gscb;
3987 		restore_gs_cb(vcpu->arch.host_gscb);
3988 		preempt_enable();
3989 		if (!vcpu->arch.host_gscb)
3990 			__ctl_clear_bit(2, 4);
3991 		vcpu->arch.host_gscb = NULL;
3992 	}
3993 	/* SIE will save etoken directly into SDNX and therefore kvm_run */
3994 }
3995 
3996 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3997 {
3998 	int rc;
3999 
4000 	if (kvm_run->immediate_exit)
4001 		return -EINTR;
4002 
4003 	vcpu_load(vcpu);
4004 
4005 	if (guestdbg_exit_pending(vcpu)) {
4006 		kvm_s390_prepare_debug_exit(vcpu);
4007 		rc = 0;
4008 		goto out;
4009 	}
4010 
4011 	kvm_sigset_activate(vcpu);
4012 
4013 	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm)) {
4014 		kvm_s390_vcpu_start(vcpu);
4015 	} else if (is_vcpu_stopped(vcpu)) {
4016 		pr_err_ratelimited("can't run stopped vcpu %d\n",
4017 				   vcpu->vcpu_id);
4018 		rc = -EINVAL;
4019 		goto out;
4020 	}
4021 
4022 	sync_regs(vcpu, kvm_run);
4023 	enable_cpu_timer_accounting(vcpu);
4024 
4025 	might_fault();
4026 	rc = __vcpu_run(vcpu);
4027 
4028 	if (signal_pending(current) && !rc) {
4029 		kvm_run->exit_reason = KVM_EXIT_INTR;
4030 		rc = -EINTR;
4031 	}
4032 
4033 	if (guestdbg_exit_pending(vcpu) && !rc)  {
4034 		kvm_s390_prepare_debug_exit(vcpu);
4035 		rc = 0;
4036 	}
4037 
4038 	if (rc == -EREMOTE) {
4039 		/* userspace support is needed, kvm_run has been prepared */
4040 		rc = 0;
4041 	}
4042 
4043 	disable_cpu_timer_accounting(vcpu);
4044 	store_regs(vcpu, kvm_run);
4045 
4046 	kvm_sigset_deactivate(vcpu);
4047 
4048 	vcpu->stat.exit_userspace++;
4049 out:
4050 	vcpu_put(vcpu);
4051 	return rc;
4052 }
4053 
4054 /*
4055  * store status at address
4056  * we use have two special cases:
4057  * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
4058  * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
4059  */
4060 int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
4061 {
4062 	unsigned char archmode = 1;
4063 	freg_t fprs[NUM_FPRS];
4064 	unsigned int px;
4065 	u64 clkcomp, cputm;
4066 	int rc;
4067 
4068 	px = kvm_s390_get_prefix(vcpu);
4069 	if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
4070 		if (write_guest_abs(vcpu, 163, &archmode, 1))
4071 			return -EFAULT;
4072 		gpa = 0;
4073 	} else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
4074 		if (write_guest_real(vcpu, 163, &archmode, 1))
4075 			return -EFAULT;
4076 		gpa = px;
4077 	} else
4078 		gpa -= __LC_FPREGS_SAVE_AREA;
4079 
4080 	/* manually convert vector registers if necessary */
4081 	if (MACHINE_HAS_VX) {
4082 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
4083 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4084 				     fprs, 128);
4085 	} else {
4086 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4087 				     vcpu->run->s.regs.fprs, 128);
4088 	}
4089 	rc |= write_guest_abs(vcpu, gpa + __LC_GPREGS_SAVE_AREA,
4090 			      vcpu->run->s.regs.gprs, 128);
4091 	rc |= write_guest_abs(vcpu, gpa + __LC_PSW_SAVE_AREA,
4092 			      &vcpu->arch.sie_block->gpsw, 16);
4093 	rc |= write_guest_abs(vcpu, gpa + __LC_PREFIX_SAVE_AREA,
4094 			      &px, 4);
4095 	rc |= write_guest_abs(vcpu, gpa + __LC_FP_CREG_SAVE_AREA,
4096 			      &vcpu->run->s.regs.fpc, 4);
4097 	rc |= write_guest_abs(vcpu, gpa + __LC_TOD_PROGREG_SAVE_AREA,
4098 			      &vcpu->arch.sie_block->todpr, 4);
4099 	cputm = kvm_s390_get_cpu_timer(vcpu);
4100 	rc |= write_guest_abs(vcpu, gpa + __LC_CPU_TIMER_SAVE_AREA,
4101 			      &cputm, 8);
4102 	clkcomp = vcpu->arch.sie_block->ckc >> 8;
4103 	rc |= write_guest_abs(vcpu, gpa + __LC_CLOCK_COMP_SAVE_AREA,
4104 			      &clkcomp, 8);
4105 	rc |= write_guest_abs(vcpu, gpa + __LC_AREGS_SAVE_AREA,
4106 			      &vcpu->run->s.regs.acrs, 64);
4107 	rc |= write_guest_abs(vcpu, gpa + __LC_CREGS_SAVE_AREA,
4108 			      &vcpu->arch.sie_block->gcr, 128);
4109 	return rc ? -EFAULT : 0;
4110 }
4111 
4112 int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
4113 {
4114 	/*
4115 	 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
4116 	 * switch in the run ioctl. Let's update our copies before we save
4117 	 * it into the save area
4118 	 */
4119 	save_fpu_regs();
4120 	vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
4121 	save_access_regs(vcpu->run->s.regs.acrs);
4122 
4123 	return kvm_s390_store_status_unloaded(vcpu, addr);
4124 }
4125 
4126 static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4127 {
4128 	kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu);
4129 	kvm_s390_sync_request(KVM_REQ_DISABLE_IBS, vcpu);
4130 }
4131 
4132 static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
4133 {
4134 	unsigned int i;
4135 	struct kvm_vcpu *vcpu;
4136 
4137 	kvm_for_each_vcpu(i, vcpu, kvm) {
4138 		__disable_ibs_on_vcpu(vcpu);
4139 	}
4140 }
4141 
4142 static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4143 {
4144 	if (!sclp.has_ibs)
4145 		return;
4146 	kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu);
4147 	kvm_s390_sync_request(KVM_REQ_ENABLE_IBS, vcpu);
4148 }
4149 
4150 void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
4151 {
4152 	int i, online_vcpus, started_vcpus = 0;
4153 
4154 	if (!is_vcpu_stopped(vcpu))
4155 		return;
4156 
4157 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
4158 	/* Only one cpu at a time may enter/leave the STOPPED state. */
4159 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
4160 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4161 
4162 	for (i = 0; i < online_vcpus; i++) {
4163 		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
4164 			started_vcpus++;
4165 	}
4166 
4167 	if (started_vcpus == 0) {
4168 		/* we're the only active VCPU -> speed it up */
4169 		__enable_ibs_on_vcpu(vcpu);
4170 	} else if (started_vcpus == 1) {
4171 		/*
4172 		 * As we are starting a second VCPU, we have to disable
4173 		 * the IBS facility on all VCPUs to remove potentially
4174 		 * oustanding ENABLE requests.
4175 		 */
4176 		__disable_ibs_on_all_vcpus(vcpu->kvm);
4177 	}
4178 
4179 	kvm_s390_clear_cpuflags(vcpu, CPUSTAT_STOPPED);
4180 	/*
4181 	 * Another VCPU might have used IBS while we were offline.
4182 	 * Let's play safe and flush the VCPU at startup.
4183 	 */
4184 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4185 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4186 	return;
4187 }
4188 
4189 void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
4190 {
4191 	int i, online_vcpus, started_vcpus = 0;
4192 	struct kvm_vcpu *started_vcpu = NULL;
4193 
4194 	if (is_vcpu_stopped(vcpu))
4195 		return;
4196 
4197 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
4198 	/* Only one cpu at a time may enter/leave the STOPPED state. */
4199 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
4200 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4201 
4202 	/* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
4203 	kvm_s390_clear_stop_irq(vcpu);
4204 
4205 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOPPED);
4206 	__disable_ibs_on_vcpu(vcpu);
4207 
4208 	for (i = 0; i < online_vcpus; i++) {
4209 		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
4210 			started_vcpus++;
4211 			started_vcpu = vcpu->kvm->vcpus[i];
4212 		}
4213 	}
4214 
4215 	if (started_vcpus == 1) {
4216 		/*
4217 		 * As we only have one VCPU left, we want to enable the
4218 		 * IBS facility for that VCPU to speed it up.
4219 		 */
4220 		__enable_ibs_on_vcpu(started_vcpu);
4221 	}
4222 
4223 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4224 	return;
4225 }
4226 
4227 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
4228 				     struct kvm_enable_cap *cap)
4229 {
4230 	int r;
4231 
4232 	if (cap->flags)
4233 		return -EINVAL;
4234 
4235 	switch (cap->cap) {
4236 	case KVM_CAP_S390_CSS_SUPPORT:
4237 		if (!vcpu->kvm->arch.css_support) {
4238 			vcpu->kvm->arch.css_support = 1;
4239 			VM_EVENT(vcpu->kvm, 3, "%s", "ENABLE: CSS support");
4240 			trace_kvm_s390_enable_css(vcpu->kvm);
4241 		}
4242 		r = 0;
4243 		break;
4244 	default:
4245 		r = -EINVAL;
4246 		break;
4247 	}
4248 	return r;
4249 }
4250 
4251 static long kvm_s390_guest_mem_op(struct kvm_vcpu *vcpu,
4252 				  struct kvm_s390_mem_op *mop)
4253 {
4254 	void __user *uaddr = (void __user *)mop->buf;
4255 	void *tmpbuf = NULL;
4256 	int r, srcu_idx;
4257 	const u64 supported_flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION
4258 				    | KVM_S390_MEMOP_F_CHECK_ONLY;
4259 
4260 	if (mop->flags & ~supported_flags)
4261 		return -EINVAL;
4262 
4263 	if (mop->size > MEM_OP_MAX_SIZE)
4264 		return -E2BIG;
4265 
4266 	if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
4267 		tmpbuf = vmalloc(mop->size);
4268 		if (!tmpbuf)
4269 			return -ENOMEM;
4270 	}
4271 
4272 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
4273 
4274 	switch (mop->op) {
4275 	case KVM_S390_MEMOP_LOGICAL_READ:
4276 		if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4277 			r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4278 					    mop->size, GACC_FETCH);
4279 			break;
4280 		}
4281 		r = read_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4282 		if (r == 0) {
4283 			if (copy_to_user(uaddr, tmpbuf, mop->size))
4284 				r = -EFAULT;
4285 		}
4286 		break;
4287 	case KVM_S390_MEMOP_LOGICAL_WRITE:
4288 		if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4289 			r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4290 					    mop->size, GACC_STORE);
4291 			break;
4292 		}
4293 		if (copy_from_user(tmpbuf, uaddr, mop->size)) {
4294 			r = -EFAULT;
4295 			break;
4296 		}
4297 		r = write_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4298 		break;
4299 	default:
4300 		r = -EINVAL;
4301 	}
4302 
4303 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
4304 
4305 	if (r > 0 && (mop->flags & KVM_S390_MEMOP_F_INJECT_EXCEPTION) != 0)
4306 		kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
4307 
4308 	vfree(tmpbuf);
4309 	return r;
4310 }
4311 
4312 long kvm_arch_vcpu_async_ioctl(struct file *filp,
4313 			       unsigned int ioctl, unsigned long arg)
4314 {
4315 	struct kvm_vcpu *vcpu = filp->private_data;
4316 	void __user *argp = (void __user *)arg;
4317 
4318 	switch (ioctl) {
4319 	case KVM_S390_IRQ: {
4320 		struct kvm_s390_irq s390irq;
4321 
4322 		if (copy_from_user(&s390irq, argp, sizeof(s390irq)))
4323 			return -EFAULT;
4324 		return kvm_s390_inject_vcpu(vcpu, &s390irq);
4325 	}
4326 	case KVM_S390_INTERRUPT: {
4327 		struct kvm_s390_interrupt s390int;
4328 		struct kvm_s390_irq s390irq;
4329 
4330 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
4331 			return -EFAULT;
4332 		if (s390int_to_s390irq(&s390int, &s390irq))
4333 			return -EINVAL;
4334 		return kvm_s390_inject_vcpu(vcpu, &s390irq);
4335 	}
4336 	}
4337 	return -ENOIOCTLCMD;
4338 }
4339 
4340 long kvm_arch_vcpu_ioctl(struct file *filp,
4341 			 unsigned int ioctl, unsigned long arg)
4342 {
4343 	struct kvm_vcpu *vcpu = filp->private_data;
4344 	void __user *argp = (void __user *)arg;
4345 	int idx;
4346 	long r;
4347 
4348 	vcpu_load(vcpu);
4349 
4350 	switch (ioctl) {
4351 	case KVM_S390_STORE_STATUS:
4352 		idx = srcu_read_lock(&vcpu->kvm->srcu);
4353 		r = kvm_s390_vcpu_store_status(vcpu, arg);
4354 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
4355 		break;
4356 	case KVM_S390_SET_INITIAL_PSW: {
4357 		psw_t psw;
4358 
4359 		r = -EFAULT;
4360 		if (copy_from_user(&psw, argp, sizeof(psw)))
4361 			break;
4362 		r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw);
4363 		break;
4364 	}
4365 	case KVM_S390_INITIAL_RESET:
4366 		r = kvm_arch_vcpu_ioctl_initial_reset(vcpu);
4367 		break;
4368 	case KVM_SET_ONE_REG:
4369 	case KVM_GET_ONE_REG: {
4370 		struct kvm_one_reg reg;
4371 		r = -EFAULT;
4372 		if (copy_from_user(&reg, argp, sizeof(reg)))
4373 			break;
4374 		if (ioctl == KVM_SET_ONE_REG)
4375 			r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, &reg);
4376 		else
4377 			r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, &reg);
4378 		break;
4379 	}
4380 #ifdef CONFIG_KVM_S390_UCONTROL
4381 	case KVM_S390_UCAS_MAP: {
4382 		struct kvm_s390_ucas_mapping ucasmap;
4383 
4384 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4385 			r = -EFAULT;
4386 			break;
4387 		}
4388 
4389 		if (!kvm_is_ucontrol(vcpu->kvm)) {
4390 			r = -EINVAL;
4391 			break;
4392 		}
4393 
4394 		r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr,
4395 				     ucasmap.vcpu_addr, ucasmap.length);
4396 		break;
4397 	}
4398 	case KVM_S390_UCAS_UNMAP: {
4399 		struct kvm_s390_ucas_mapping ucasmap;
4400 
4401 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4402 			r = -EFAULT;
4403 			break;
4404 		}
4405 
4406 		if (!kvm_is_ucontrol(vcpu->kvm)) {
4407 			r = -EINVAL;
4408 			break;
4409 		}
4410 
4411 		r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr,
4412 			ucasmap.length);
4413 		break;
4414 	}
4415 #endif
4416 	case KVM_S390_VCPU_FAULT: {
4417 		r = gmap_fault(vcpu->arch.gmap, arg, 0);
4418 		break;
4419 	}
4420 	case KVM_ENABLE_CAP:
4421 	{
4422 		struct kvm_enable_cap cap;
4423 		r = -EFAULT;
4424 		if (copy_from_user(&cap, argp, sizeof(cap)))
4425 			break;
4426 		r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
4427 		break;
4428 	}
4429 	case KVM_S390_MEM_OP: {
4430 		struct kvm_s390_mem_op mem_op;
4431 
4432 		if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
4433 			r = kvm_s390_guest_mem_op(vcpu, &mem_op);
4434 		else
4435 			r = -EFAULT;
4436 		break;
4437 	}
4438 	case KVM_S390_SET_IRQ_STATE: {
4439 		struct kvm_s390_irq_state irq_state;
4440 
4441 		r = -EFAULT;
4442 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4443 			break;
4444 		if (irq_state.len > VCPU_IRQS_MAX_BUF ||
4445 		    irq_state.len == 0 ||
4446 		    irq_state.len % sizeof(struct kvm_s390_irq) > 0) {
4447 			r = -EINVAL;
4448 			break;
4449 		}
4450 		/* do not use irq_state.flags, it will break old QEMUs */
4451 		r = kvm_s390_set_irq_state(vcpu,
4452 					   (void __user *) irq_state.buf,
4453 					   irq_state.len);
4454 		break;
4455 	}
4456 	case KVM_S390_GET_IRQ_STATE: {
4457 		struct kvm_s390_irq_state irq_state;
4458 
4459 		r = -EFAULT;
4460 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4461 			break;
4462 		if (irq_state.len == 0) {
4463 			r = -EINVAL;
4464 			break;
4465 		}
4466 		/* do not use irq_state.flags, it will break old QEMUs */
4467 		r = kvm_s390_get_irq_state(vcpu,
4468 					   (__u8 __user *)  irq_state.buf,
4469 					   irq_state.len);
4470 		break;
4471 	}
4472 	default:
4473 		r = -ENOTTY;
4474 	}
4475 
4476 	vcpu_put(vcpu);
4477 	return r;
4478 }
4479 
4480 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
4481 {
4482 #ifdef CONFIG_KVM_S390_UCONTROL
4483 	if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET)
4484 		 && (kvm_is_ucontrol(vcpu->kvm))) {
4485 		vmf->page = virt_to_page(vcpu->arch.sie_block);
4486 		get_page(vmf->page);
4487 		return 0;
4488 	}
4489 #endif
4490 	return VM_FAULT_SIGBUS;
4491 }
4492 
4493 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
4494 			    unsigned long npages)
4495 {
4496 	return 0;
4497 }
4498 
4499 /* Section: memory related */
4500 int kvm_arch_prepare_memory_region(struct kvm *kvm,
4501 				   struct kvm_memory_slot *memslot,
4502 				   const struct kvm_userspace_memory_region *mem,
4503 				   enum kvm_mr_change change)
4504 {
4505 	/* A few sanity checks. We can have memory slots which have to be
4506 	   located/ended at a segment boundary (1MB). The memory in userland is
4507 	   ok to be fragmented into various different vmas. It is okay to mmap()
4508 	   and munmap() stuff in this slot after doing this call at any time */
4509 
4510 	if (mem->userspace_addr & 0xffffful)
4511 		return -EINVAL;
4512 
4513 	if (mem->memory_size & 0xffffful)
4514 		return -EINVAL;
4515 
4516 	if (mem->guest_phys_addr + mem->memory_size > kvm->arch.mem_limit)
4517 		return -EINVAL;
4518 
4519 	return 0;
4520 }
4521 
4522 void kvm_arch_commit_memory_region(struct kvm *kvm,
4523 				const struct kvm_userspace_memory_region *mem,
4524 				const struct kvm_memory_slot *old,
4525 				const struct kvm_memory_slot *new,
4526 				enum kvm_mr_change change)
4527 {
4528 	int rc = 0;
4529 
4530 	switch (change) {
4531 	case KVM_MR_DELETE:
4532 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
4533 					old->npages * PAGE_SIZE);
4534 		break;
4535 	case KVM_MR_MOVE:
4536 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
4537 					old->npages * PAGE_SIZE);
4538 		if (rc)
4539 			break;
4540 		/* FALLTHROUGH */
4541 	case KVM_MR_CREATE:
4542 		rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr,
4543 				      mem->guest_phys_addr, mem->memory_size);
4544 		break;
4545 	case KVM_MR_FLAGS_ONLY:
4546 		break;
4547 	default:
4548 		WARN(1, "Unknown KVM MR CHANGE: %d\n", change);
4549 	}
4550 	if (rc)
4551 		pr_warn("failed to commit memory region\n");
4552 	return;
4553 }
4554 
4555 static inline unsigned long nonhyp_mask(int i)
4556 {
4557 	unsigned int nonhyp_fai = (sclp.hmfai << i * 2) >> 30;
4558 
4559 	return 0x0000ffffffffffffUL >> (nonhyp_fai << 4);
4560 }
4561 
4562 void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu)
4563 {
4564 	vcpu->valid_wakeup = false;
4565 }
4566 
4567 static int __init kvm_s390_init(void)
4568 {
4569 	int i;
4570 
4571 	if (!sclp.has_sief2) {
4572 		pr_info("SIE is not available\n");
4573 		return -ENODEV;
4574 	}
4575 
4576 	if (nested && hpage) {
4577 		pr_info("A KVM host that supports nesting cannot back its KVM guests with huge pages\n");
4578 		return -EINVAL;
4579 	}
4580 
4581 	for (i = 0; i < 16; i++)
4582 		kvm_s390_fac_base[i] |=
4583 			S390_lowcore.stfle_fac_list[i] & nonhyp_mask(i);
4584 
4585 	return kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
4586 }
4587 
4588 static void __exit kvm_s390_exit(void)
4589 {
4590 	kvm_exit();
4591 }
4592 
4593 module_init(kvm_s390_init);
4594 module_exit(kvm_s390_exit);
4595 
4596 /*
4597  * Enable autoloading of the kvm module.
4598  * Note that we add the module alias here instead of virt/kvm/kvm_main.c
4599  * since x86 takes a different approach.
4600  */
4601 #include <linux/miscdevice.h>
4602 MODULE_ALIAS_MISCDEV(KVM_MINOR);
4603 MODULE_ALIAS("devname:kvm");
4604