xref: /openbmc/linux/arch/s390/kvm/kvm-s390.c (revision 08283d30)
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 int kvm_arch_check_processor_compat(void)
231 {
232 	return 0;
233 }
234 
235 static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
236 			      unsigned long end);
237 
238 static void kvm_clock_sync_scb(struct kvm_s390_sie_block *scb, u64 delta)
239 {
240 	u8 delta_idx = 0;
241 
242 	/*
243 	 * The TOD jumps by delta, we have to compensate this by adding
244 	 * -delta to the epoch.
245 	 */
246 	delta = -delta;
247 
248 	/* sign-extension - we're adding to signed values below */
249 	if ((s64)delta < 0)
250 		delta_idx = -1;
251 
252 	scb->epoch += delta;
253 	if (scb->ecd & ECD_MEF) {
254 		scb->epdx += delta_idx;
255 		if (scb->epoch < delta)
256 			scb->epdx += 1;
257 	}
258 }
259 
260 /*
261  * This callback is executed during stop_machine(). All CPUs are therefore
262  * temporarily stopped. In order not to change guest behavior, we have to
263  * disable preemption whenever we touch the epoch of kvm and the VCPUs,
264  * so a CPU won't be stopped while calculating with the epoch.
265  */
266 static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
267 			  void *v)
268 {
269 	struct kvm *kvm;
270 	struct kvm_vcpu *vcpu;
271 	int i;
272 	unsigned long long *delta = v;
273 
274 	list_for_each_entry(kvm, &vm_list, vm_list) {
275 		kvm_for_each_vcpu(i, vcpu, kvm) {
276 			kvm_clock_sync_scb(vcpu->arch.sie_block, *delta);
277 			if (i == 0) {
278 				kvm->arch.epoch = vcpu->arch.sie_block->epoch;
279 				kvm->arch.epdx = vcpu->arch.sie_block->epdx;
280 			}
281 			if (vcpu->arch.cputm_enabled)
282 				vcpu->arch.cputm_start += *delta;
283 			if (vcpu->arch.vsie_block)
284 				kvm_clock_sync_scb(vcpu->arch.vsie_block,
285 						   *delta);
286 		}
287 	}
288 	return NOTIFY_OK;
289 }
290 
291 static struct notifier_block kvm_clock_notifier = {
292 	.notifier_call = kvm_clock_sync,
293 };
294 
295 int kvm_arch_hardware_setup(void)
296 {
297 	gmap_notifier.notifier_call = kvm_gmap_notifier;
298 	gmap_register_pte_notifier(&gmap_notifier);
299 	vsie_gmap_notifier.notifier_call = kvm_s390_vsie_gmap_notifier;
300 	gmap_register_pte_notifier(&vsie_gmap_notifier);
301 	atomic_notifier_chain_register(&s390_epoch_delta_notifier,
302 				       &kvm_clock_notifier);
303 	return 0;
304 }
305 
306 void kvm_arch_hardware_unsetup(void)
307 {
308 	gmap_unregister_pte_notifier(&gmap_notifier);
309 	gmap_unregister_pte_notifier(&vsie_gmap_notifier);
310 	atomic_notifier_chain_unregister(&s390_epoch_delta_notifier,
311 					 &kvm_clock_notifier);
312 }
313 
314 static void allow_cpu_feat(unsigned long nr)
315 {
316 	set_bit_inv(nr, kvm_s390_available_cpu_feat);
317 }
318 
319 static inline int plo_test_bit(unsigned char nr)
320 {
321 	register unsigned long r0 asm("0") = (unsigned long) nr | 0x100;
322 	int cc;
323 
324 	asm volatile(
325 		/* Parameter registers are ignored for "test bit" */
326 		"	plo	0,0,0,0(0)\n"
327 		"	ipm	%0\n"
328 		"	srl	%0,28\n"
329 		: "=d" (cc)
330 		: "d" (r0)
331 		: "cc");
332 	return cc == 0;
333 }
334 
335 static __always_inline void __insn32_query(unsigned int opcode, u8 *query)
336 {
337 	register unsigned long r0 asm("0") = 0;	/* query function */
338 	register unsigned long r1 asm("1") = (unsigned long) query;
339 
340 	asm volatile(
341 		/* Parameter regs are ignored */
342 		"	.insn	rrf,%[opc] << 16,2,4,6,0\n"
343 		:
344 		: "d" (r0), "a" (r1), [opc] "i" (opcode)
345 		: "cc", "memory");
346 }
347 
348 #define INSN_SORTL 0xb938
349 #define INSN_DFLTCC 0xb939
350 
351 static void kvm_s390_cpu_feat_init(void)
352 {
353 	int i;
354 
355 	for (i = 0; i < 256; ++i) {
356 		if (plo_test_bit(i))
357 			kvm_s390_available_subfunc.plo[i >> 3] |= 0x80 >> (i & 7);
358 	}
359 
360 	if (test_facility(28)) /* TOD-clock steering */
361 		ptff(kvm_s390_available_subfunc.ptff,
362 		     sizeof(kvm_s390_available_subfunc.ptff),
363 		     PTFF_QAF);
364 
365 	if (test_facility(17)) { /* MSA */
366 		__cpacf_query(CPACF_KMAC, (cpacf_mask_t *)
367 			      kvm_s390_available_subfunc.kmac);
368 		__cpacf_query(CPACF_KMC, (cpacf_mask_t *)
369 			      kvm_s390_available_subfunc.kmc);
370 		__cpacf_query(CPACF_KM, (cpacf_mask_t *)
371 			      kvm_s390_available_subfunc.km);
372 		__cpacf_query(CPACF_KIMD, (cpacf_mask_t *)
373 			      kvm_s390_available_subfunc.kimd);
374 		__cpacf_query(CPACF_KLMD, (cpacf_mask_t *)
375 			      kvm_s390_available_subfunc.klmd);
376 	}
377 	if (test_facility(76)) /* MSA3 */
378 		__cpacf_query(CPACF_PCKMO, (cpacf_mask_t *)
379 			      kvm_s390_available_subfunc.pckmo);
380 	if (test_facility(77)) { /* MSA4 */
381 		__cpacf_query(CPACF_KMCTR, (cpacf_mask_t *)
382 			      kvm_s390_available_subfunc.kmctr);
383 		__cpacf_query(CPACF_KMF, (cpacf_mask_t *)
384 			      kvm_s390_available_subfunc.kmf);
385 		__cpacf_query(CPACF_KMO, (cpacf_mask_t *)
386 			      kvm_s390_available_subfunc.kmo);
387 		__cpacf_query(CPACF_PCC, (cpacf_mask_t *)
388 			      kvm_s390_available_subfunc.pcc);
389 	}
390 	if (test_facility(57)) /* MSA5 */
391 		__cpacf_query(CPACF_PRNO, (cpacf_mask_t *)
392 			      kvm_s390_available_subfunc.ppno);
393 
394 	if (test_facility(146)) /* MSA8 */
395 		__cpacf_query(CPACF_KMA, (cpacf_mask_t *)
396 			      kvm_s390_available_subfunc.kma);
397 
398 	if (test_facility(155)) /* MSA9 */
399 		__cpacf_query(CPACF_KDSA, (cpacf_mask_t *)
400 			      kvm_s390_available_subfunc.kdsa);
401 
402 	if (test_facility(150)) /* SORTL */
403 		__insn32_query(INSN_SORTL, kvm_s390_available_subfunc.sortl);
404 
405 	if (test_facility(151)) /* DFLTCC */
406 		__insn32_query(INSN_DFLTCC, kvm_s390_available_subfunc.dfltcc);
407 
408 	if (MACHINE_HAS_ESOP)
409 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_ESOP);
410 	/*
411 	 * We need SIE support, ESOP (PROT_READ protection for gmap_shadow),
412 	 * 64bit SCAO (SCA passthrough) and IDTE (for gmap_shadow unshadowing).
413 	 */
414 	if (!sclp.has_sief2 || !MACHINE_HAS_ESOP || !sclp.has_64bscao ||
415 	    !test_facility(3) || !nested)
416 		return;
417 	allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIEF2);
418 	if (sclp.has_64bscao)
419 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_64BSCAO);
420 	if (sclp.has_siif)
421 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_SIIF);
422 	if (sclp.has_gpere)
423 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GPERE);
424 	if (sclp.has_gsls)
425 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_GSLS);
426 	if (sclp.has_ib)
427 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IB);
428 	if (sclp.has_cei)
429 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_CEI);
430 	if (sclp.has_ibs)
431 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_IBS);
432 	if (sclp.has_kss)
433 		allow_cpu_feat(KVM_S390_VM_CPU_FEAT_KSS);
434 	/*
435 	 * KVM_S390_VM_CPU_FEAT_SKEY: Wrong shadow of PTE.I bits will make
436 	 * all skey handling functions read/set the skey from the PGSTE
437 	 * instead of the real storage key.
438 	 *
439 	 * KVM_S390_VM_CPU_FEAT_CMMA: Wrong shadow of PTE.I bits will make
440 	 * pages being detected as preserved although they are resident.
441 	 *
442 	 * KVM_S390_VM_CPU_FEAT_PFMFI: Wrong shadow of PTE.I bits will
443 	 * have the same effect as for KVM_S390_VM_CPU_FEAT_SKEY.
444 	 *
445 	 * For KVM_S390_VM_CPU_FEAT_SKEY, KVM_S390_VM_CPU_FEAT_CMMA and
446 	 * KVM_S390_VM_CPU_FEAT_PFMFI, all PTE.I and PGSTE bits have to be
447 	 * correctly shadowed. We can do that for the PGSTE but not for PTE.I.
448 	 *
449 	 * KVM_S390_VM_CPU_FEAT_SIGPIF: Wrong SCB addresses in the SCA. We
450 	 * cannot easily shadow the SCA because of the ipte lock.
451 	 */
452 }
453 
454 int kvm_arch_init(void *opaque)
455 {
456 	int rc;
457 
458 	kvm_s390_dbf = debug_register("kvm-trace", 32, 1, 7 * sizeof(long));
459 	if (!kvm_s390_dbf)
460 		return -ENOMEM;
461 
462 	if (debug_register_view(kvm_s390_dbf, &debug_sprintf_view)) {
463 		rc = -ENOMEM;
464 		goto out_debug_unreg;
465 	}
466 
467 	kvm_s390_cpu_feat_init();
468 
469 	/* Register floating interrupt controller interface. */
470 	rc = kvm_register_device_ops(&kvm_flic_ops, KVM_DEV_TYPE_FLIC);
471 	if (rc) {
472 		pr_err("A FLIC registration call failed with rc=%d\n", rc);
473 		goto out_debug_unreg;
474 	}
475 
476 	rc = kvm_s390_gib_init(GAL_ISC);
477 	if (rc)
478 		goto out_gib_destroy;
479 
480 	return 0;
481 
482 out_gib_destroy:
483 	kvm_s390_gib_destroy();
484 out_debug_unreg:
485 	debug_unregister(kvm_s390_dbf);
486 	return rc;
487 }
488 
489 void kvm_arch_exit(void)
490 {
491 	kvm_s390_gib_destroy();
492 	debug_unregister(kvm_s390_dbf);
493 }
494 
495 /* Section: device related */
496 long kvm_arch_dev_ioctl(struct file *filp,
497 			unsigned int ioctl, unsigned long arg)
498 {
499 	if (ioctl == KVM_S390_ENABLE_SIE)
500 		return s390_enable_sie();
501 	return -EINVAL;
502 }
503 
504 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
505 {
506 	int r;
507 
508 	switch (ext) {
509 	case KVM_CAP_S390_PSW:
510 	case KVM_CAP_S390_GMAP:
511 	case KVM_CAP_SYNC_MMU:
512 #ifdef CONFIG_KVM_S390_UCONTROL
513 	case KVM_CAP_S390_UCONTROL:
514 #endif
515 	case KVM_CAP_ASYNC_PF:
516 	case KVM_CAP_SYNC_REGS:
517 	case KVM_CAP_ONE_REG:
518 	case KVM_CAP_ENABLE_CAP:
519 	case KVM_CAP_S390_CSS_SUPPORT:
520 	case KVM_CAP_IOEVENTFD:
521 	case KVM_CAP_DEVICE_CTRL:
522 	case KVM_CAP_S390_IRQCHIP:
523 	case KVM_CAP_VM_ATTRIBUTES:
524 	case KVM_CAP_MP_STATE:
525 	case KVM_CAP_IMMEDIATE_EXIT:
526 	case KVM_CAP_S390_INJECT_IRQ:
527 	case KVM_CAP_S390_USER_SIGP:
528 	case KVM_CAP_S390_USER_STSI:
529 	case KVM_CAP_S390_SKEYS:
530 	case KVM_CAP_S390_IRQ_STATE:
531 	case KVM_CAP_S390_USER_INSTR0:
532 	case KVM_CAP_S390_CMMA_MIGRATION:
533 	case KVM_CAP_S390_AIS:
534 	case KVM_CAP_S390_AIS_MIGRATION:
535 		r = 1;
536 		break;
537 	case KVM_CAP_S390_HPAGE_1M:
538 		r = 0;
539 		if (hpage && !kvm_is_ucontrol(kvm))
540 			r = 1;
541 		break;
542 	case KVM_CAP_S390_MEM_OP:
543 		r = MEM_OP_MAX_SIZE;
544 		break;
545 	case KVM_CAP_NR_VCPUS:
546 	case KVM_CAP_MAX_VCPUS:
547 	case KVM_CAP_MAX_VCPU_ID:
548 		r = KVM_S390_BSCA_CPU_SLOTS;
549 		if (!kvm_s390_use_sca_entries())
550 			r = KVM_MAX_VCPUS;
551 		else if (sclp.has_esca && sclp.has_64bscao)
552 			r = KVM_S390_ESCA_CPU_SLOTS;
553 		break;
554 	case KVM_CAP_S390_COW:
555 		r = MACHINE_HAS_ESOP;
556 		break;
557 	case KVM_CAP_S390_VECTOR_REGISTERS:
558 		r = MACHINE_HAS_VX;
559 		break;
560 	case KVM_CAP_S390_RI:
561 		r = test_facility(64);
562 		break;
563 	case KVM_CAP_S390_GS:
564 		r = test_facility(133);
565 		break;
566 	case KVM_CAP_S390_BPB:
567 		r = test_facility(82);
568 		break;
569 	default:
570 		r = 0;
571 	}
572 	return r;
573 }
574 
575 static void kvm_s390_sync_dirty_log(struct kvm *kvm,
576 				    struct kvm_memory_slot *memslot)
577 {
578 	int i;
579 	gfn_t cur_gfn, last_gfn;
580 	unsigned long gaddr, vmaddr;
581 	struct gmap *gmap = kvm->arch.gmap;
582 	DECLARE_BITMAP(bitmap, _PAGE_ENTRIES);
583 
584 	/* Loop over all guest segments */
585 	cur_gfn = memslot->base_gfn;
586 	last_gfn = memslot->base_gfn + memslot->npages;
587 	for (; cur_gfn <= last_gfn; cur_gfn += _PAGE_ENTRIES) {
588 		gaddr = gfn_to_gpa(cur_gfn);
589 		vmaddr = gfn_to_hva_memslot(memslot, cur_gfn);
590 		if (kvm_is_error_hva(vmaddr))
591 			continue;
592 
593 		bitmap_zero(bitmap, _PAGE_ENTRIES);
594 		gmap_sync_dirty_log_pmd(gmap, bitmap, gaddr, vmaddr);
595 		for (i = 0; i < _PAGE_ENTRIES; i++) {
596 			if (test_bit(i, bitmap))
597 				mark_page_dirty(kvm, cur_gfn + i);
598 		}
599 
600 		if (fatal_signal_pending(current))
601 			return;
602 		cond_resched();
603 	}
604 }
605 
606 /* Section: vm related */
607 static void sca_del_vcpu(struct kvm_vcpu *vcpu);
608 
609 /*
610  * Get (and clear) the dirty memory log for a memory slot.
611  */
612 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
613 			       struct kvm_dirty_log *log)
614 {
615 	int r;
616 	unsigned long n;
617 	struct kvm_memslots *slots;
618 	struct kvm_memory_slot *memslot;
619 	int is_dirty = 0;
620 
621 	if (kvm_is_ucontrol(kvm))
622 		return -EINVAL;
623 
624 	mutex_lock(&kvm->slots_lock);
625 
626 	r = -EINVAL;
627 	if (log->slot >= KVM_USER_MEM_SLOTS)
628 		goto out;
629 
630 	slots = kvm_memslots(kvm);
631 	memslot = id_to_memslot(slots, log->slot);
632 	r = -ENOENT;
633 	if (!memslot->dirty_bitmap)
634 		goto out;
635 
636 	kvm_s390_sync_dirty_log(kvm, memslot);
637 	r = kvm_get_dirty_log(kvm, log, &is_dirty);
638 	if (r)
639 		goto out;
640 
641 	/* Clear the dirty log */
642 	if (is_dirty) {
643 		n = kvm_dirty_bitmap_bytes(memslot);
644 		memset(memslot->dirty_bitmap, 0, n);
645 	}
646 	r = 0;
647 out:
648 	mutex_unlock(&kvm->slots_lock);
649 	return r;
650 }
651 
652 static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
653 {
654 	unsigned int i;
655 	struct kvm_vcpu *vcpu;
656 
657 	kvm_for_each_vcpu(i, vcpu, kvm) {
658 		kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
659 	}
660 }
661 
662 int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
663 {
664 	int r;
665 
666 	if (cap->flags)
667 		return -EINVAL;
668 
669 	switch (cap->cap) {
670 	case KVM_CAP_S390_IRQCHIP:
671 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_IRQCHIP");
672 		kvm->arch.use_irqchip = 1;
673 		r = 0;
674 		break;
675 	case KVM_CAP_S390_USER_SIGP:
676 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_SIGP");
677 		kvm->arch.user_sigp = 1;
678 		r = 0;
679 		break;
680 	case KVM_CAP_S390_VECTOR_REGISTERS:
681 		mutex_lock(&kvm->lock);
682 		if (kvm->created_vcpus) {
683 			r = -EBUSY;
684 		} else if (MACHINE_HAS_VX) {
685 			set_kvm_facility(kvm->arch.model.fac_mask, 129);
686 			set_kvm_facility(kvm->arch.model.fac_list, 129);
687 			if (test_facility(134)) {
688 				set_kvm_facility(kvm->arch.model.fac_mask, 134);
689 				set_kvm_facility(kvm->arch.model.fac_list, 134);
690 			}
691 			if (test_facility(135)) {
692 				set_kvm_facility(kvm->arch.model.fac_mask, 135);
693 				set_kvm_facility(kvm->arch.model.fac_list, 135);
694 			}
695 			if (test_facility(148)) {
696 				set_kvm_facility(kvm->arch.model.fac_mask, 148);
697 				set_kvm_facility(kvm->arch.model.fac_list, 148);
698 			}
699 			if (test_facility(152)) {
700 				set_kvm_facility(kvm->arch.model.fac_mask, 152);
701 				set_kvm_facility(kvm->arch.model.fac_list, 152);
702 			}
703 			r = 0;
704 		} else
705 			r = -EINVAL;
706 		mutex_unlock(&kvm->lock);
707 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_VECTOR_REGISTERS %s",
708 			 r ? "(not available)" : "(success)");
709 		break;
710 	case KVM_CAP_S390_RI:
711 		r = -EINVAL;
712 		mutex_lock(&kvm->lock);
713 		if (kvm->created_vcpus) {
714 			r = -EBUSY;
715 		} else if (test_facility(64)) {
716 			set_kvm_facility(kvm->arch.model.fac_mask, 64);
717 			set_kvm_facility(kvm->arch.model.fac_list, 64);
718 			r = 0;
719 		}
720 		mutex_unlock(&kvm->lock);
721 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_RI %s",
722 			 r ? "(not available)" : "(success)");
723 		break;
724 	case KVM_CAP_S390_AIS:
725 		mutex_lock(&kvm->lock);
726 		if (kvm->created_vcpus) {
727 			r = -EBUSY;
728 		} else {
729 			set_kvm_facility(kvm->arch.model.fac_mask, 72);
730 			set_kvm_facility(kvm->arch.model.fac_list, 72);
731 			r = 0;
732 		}
733 		mutex_unlock(&kvm->lock);
734 		VM_EVENT(kvm, 3, "ENABLE: AIS %s",
735 			 r ? "(not available)" : "(success)");
736 		break;
737 	case KVM_CAP_S390_GS:
738 		r = -EINVAL;
739 		mutex_lock(&kvm->lock);
740 		if (kvm->created_vcpus) {
741 			r = -EBUSY;
742 		} else if (test_facility(133)) {
743 			set_kvm_facility(kvm->arch.model.fac_mask, 133);
744 			set_kvm_facility(kvm->arch.model.fac_list, 133);
745 			r = 0;
746 		}
747 		mutex_unlock(&kvm->lock);
748 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_GS %s",
749 			 r ? "(not available)" : "(success)");
750 		break;
751 	case KVM_CAP_S390_HPAGE_1M:
752 		mutex_lock(&kvm->lock);
753 		if (kvm->created_vcpus)
754 			r = -EBUSY;
755 		else if (!hpage || kvm->arch.use_cmma || kvm_is_ucontrol(kvm))
756 			r = -EINVAL;
757 		else {
758 			r = 0;
759 			down_write(&kvm->mm->mmap_sem);
760 			kvm->mm->context.allow_gmap_hpage_1m = 1;
761 			up_write(&kvm->mm->mmap_sem);
762 			/*
763 			 * We might have to create fake 4k page
764 			 * tables. To avoid that the hardware works on
765 			 * stale PGSTEs, we emulate these instructions.
766 			 */
767 			kvm->arch.use_skf = 0;
768 			kvm->arch.use_pfmfi = 0;
769 		}
770 		mutex_unlock(&kvm->lock);
771 		VM_EVENT(kvm, 3, "ENABLE: CAP_S390_HPAGE %s",
772 			 r ? "(not available)" : "(success)");
773 		break;
774 	case KVM_CAP_S390_USER_STSI:
775 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_STSI");
776 		kvm->arch.user_stsi = 1;
777 		r = 0;
778 		break;
779 	case KVM_CAP_S390_USER_INSTR0:
780 		VM_EVENT(kvm, 3, "%s", "ENABLE: CAP_S390_USER_INSTR0");
781 		kvm->arch.user_instr0 = 1;
782 		icpt_operexc_on_all_vcpus(kvm);
783 		r = 0;
784 		break;
785 	default:
786 		r = -EINVAL;
787 		break;
788 	}
789 	return r;
790 }
791 
792 static int kvm_s390_get_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
793 {
794 	int ret;
795 
796 	switch (attr->attr) {
797 	case KVM_S390_VM_MEM_LIMIT_SIZE:
798 		ret = 0;
799 		VM_EVENT(kvm, 3, "QUERY: max guest memory: %lu bytes",
800 			 kvm->arch.mem_limit);
801 		if (put_user(kvm->arch.mem_limit, (u64 __user *)attr->addr))
802 			ret = -EFAULT;
803 		break;
804 	default:
805 		ret = -ENXIO;
806 		break;
807 	}
808 	return ret;
809 }
810 
811 static int kvm_s390_set_mem_control(struct kvm *kvm, struct kvm_device_attr *attr)
812 {
813 	int ret;
814 	unsigned int idx;
815 	switch (attr->attr) {
816 	case KVM_S390_VM_MEM_ENABLE_CMMA:
817 		ret = -ENXIO;
818 		if (!sclp.has_cmma)
819 			break;
820 
821 		VM_EVENT(kvm, 3, "%s", "ENABLE: CMMA support");
822 		mutex_lock(&kvm->lock);
823 		if (kvm->created_vcpus)
824 			ret = -EBUSY;
825 		else if (kvm->mm->context.allow_gmap_hpage_1m)
826 			ret = -EINVAL;
827 		else {
828 			kvm->arch.use_cmma = 1;
829 			/* Not compatible with cmma. */
830 			kvm->arch.use_pfmfi = 0;
831 			ret = 0;
832 		}
833 		mutex_unlock(&kvm->lock);
834 		break;
835 	case KVM_S390_VM_MEM_CLR_CMMA:
836 		ret = -ENXIO;
837 		if (!sclp.has_cmma)
838 			break;
839 		ret = -EINVAL;
840 		if (!kvm->arch.use_cmma)
841 			break;
842 
843 		VM_EVENT(kvm, 3, "%s", "RESET: CMMA states");
844 		mutex_lock(&kvm->lock);
845 		idx = srcu_read_lock(&kvm->srcu);
846 		s390_reset_cmma(kvm->arch.gmap->mm);
847 		srcu_read_unlock(&kvm->srcu, idx);
848 		mutex_unlock(&kvm->lock);
849 		ret = 0;
850 		break;
851 	case KVM_S390_VM_MEM_LIMIT_SIZE: {
852 		unsigned long new_limit;
853 
854 		if (kvm_is_ucontrol(kvm))
855 			return -EINVAL;
856 
857 		if (get_user(new_limit, (u64 __user *)attr->addr))
858 			return -EFAULT;
859 
860 		if (kvm->arch.mem_limit != KVM_S390_NO_MEM_LIMIT &&
861 		    new_limit > kvm->arch.mem_limit)
862 			return -E2BIG;
863 
864 		if (!new_limit)
865 			return -EINVAL;
866 
867 		/* gmap_create takes last usable address */
868 		if (new_limit != KVM_S390_NO_MEM_LIMIT)
869 			new_limit -= 1;
870 
871 		ret = -EBUSY;
872 		mutex_lock(&kvm->lock);
873 		if (!kvm->created_vcpus) {
874 			/* gmap_create will round the limit up */
875 			struct gmap *new = gmap_create(current->mm, new_limit);
876 
877 			if (!new) {
878 				ret = -ENOMEM;
879 			} else {
880 				gmap_remove(kvm->arch.gmap);
881 				new->private = kvm;
882 				kvm->arch.gmap = new;
883 				ret = 0;
884 			}
885 		}
886 		mutex_unlock(&kvm->lock);
887 		VM_EVENT(kvm, 3, "SET: max guest address: %lu", new_limit);
888 		VM_EVENT(kvm, 3, "New guest asce: 0x%pK",
889 			 (void *) kvm->arch.gmap->asce);
890 		break;
891 	}
892 	default:
893 		ret = -ENXIO;
894 		break;
895 	}
896 	return ret;
897 }
898 
899 static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
900 
901 void kvm_s390_vcpu_crypto_reset_all(struct kvm *kvm)
902 {
903 	struct kvm_vcpu *vcpu;
904 	int i;
905 
906 	kvm_s390_vcpu_block_all(kvm);
907 
908 	kvm_for_each_vcpu(i, vcpu, kvm) {
909 		kvm_s390_vcpu_crypto_setup(vcpu);
910 		/* recreate the shadow crycb by leaving the VSIE handler */
911 		kvm_s390_sync_request(KVM_REQ_VSIE_RESTART, vcpu);
912 	}
913 
914 	kvm_s390_vcpu_unblock_all(kvm);
915 }
916 
917 static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
918 {
919 	mutex_lock(&kvm->lock);
920 	switch (attr->attr) {
921 	case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
922 		if (!test_kvm_facility(kvm, 76)) {
923 			mutex_unlock(&kvm->lock);
924 			return -EINVAL;
925 		}
926 		get_random_bytes(
927 			kvm->arch.crypto.crycb->aes_wrapping_key_mask,
928 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
929 		kvm->arch.crypto.aes_kw = 1;
930 		VM_EVENT(kvm, 3, "%s", "ENABLE: AES keywrapping support");
931 		break;
932 	case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
933 		if (!test_kvm_facility(kvm, 76)) {
934 			mutex_unlock(&kvm->lock);
935 			return -EINVAL;
936 		}
937 		get_random_bytes(
938 			kvm->arch.crypto.crycb->dea_wrapping_key_mask,
939 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
940 		kvm->arch.crypto.dea_kw = 1;
941 		VM_EVENT(kvm, 3, "%s", "ENABLE: DEA keywrapping support");
942 		break;
943 	case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
944 		if (!test_kvm_facility(kvm, 76)) {
945 			mutex_unlock(&kvm->lock);
946 			return -EINVAL;
947 		}
948 		kvm->arch.crypto.aes_kw = 0;
949 		memset(kvm->arch.crypto.crycb->aes_wrapping_key_mask, 0,
950 			sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
951 		VM_EVENT(kvm, 3, "%s", "DISABLE: AES keywrapping support");
952 		break;
953 	case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
954 		if (!test_kvm_facility(kvm, 76)) {
955 			mutex_unlock(&kvm->lock);
956 			return -EINVAL;
957 		}
958 		kvm->arch.crypto.dea_kw = 0;
959 		memset(kvm->arch.crypto.crycb->dea_wrapping_key_mask, 0,
960 			sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
961 		VM_EVENT(kvm, 3, "%s", "DISABLE: DEA keywrapping support");
962 		break;
963 	case KVM_S390_VM_CRYPTO_ENABLE_APIE:
964 		if (!ap_instructions_available()) {
965 			mutex_unlock(&kvm->lock);
966 			return -EOPNOTSUPP;
967 		}
968 		kvm->arch.crypto.apie = 1;
969 		break;
970 	case KVM_S390_VM_CRYPTO_DISABLE_APIE:
971 		if (!ap_instructions_available()) {
972 			mutex_unlock(&kvm->lock);
973 			return -EOPNOTSUPP;
974 		}
975 		kvm->arch.crypto.apie = 0;
976 		break;
977 	default:
978 		mutex_unlock(&kvm->lock);
979 		return -ENXIO;
980 	}
981 
982 	kvm_s390_vcpu_crypto_reset_all(kvm);
983 	mutex_unlock(&kvm->lock);
984 	return 0;
985 }
986 
987 static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
988 {
989 	int cx;
990 	struct kvm_vcpu *vcpu;
991 
992 	kvm_for_each_vcpu(cx, vcpu, kvm)
993 		kvm_s390_sync_request(req, vcpu);
994 }
995 
996 /*
997  * Must be called with kvm->srcu held to avoid races on memslots, and with
998  * kvm->slots_lock to avoid races with ourselves and kvm_s390_vm_stop_migration.
999  */
1000 static int kvm_s390_vm_start_migration(struct kvm *kvm)
1001 {
1002 	struct kvm_memory_slot *ms;
1003 	struct kvm_memslots *slots;
1004 	unsigned long ram_pages = 0;
1005 	int slotnr;
1006 
1007 	/* migration mode already enabled */
1008 	if (kvm->arch.migration_mode)
1009 		return 0;
1010 	slots = kvm_memslots(kvm);
1011 	if (!slots || !slots->used_slots)
1012 		return -EINVAL;
1013 
1014 	if (!kvm->arch.use_cmma) {
1015 		kvm->arch.migration_mode = 1;
1016 		return 0;
1017 	}
1018 	/* mark all the pages in active slots as dirty */
1019 	for (slotnr = 0; slotnr < slots->used_slots; slotnr++) {
1020 		ms = slots->memslots + slotnr;
1021 		if (!ms->dirty_bitmap)
1022 			return -EINVAL;
1023 		/*
1024 		 * The second half of the bitmap is only used on x86,
1025 		 * and would be wasted otherwise, so we put it to good
1026 		 * use here to keep track of the state of the storage
1027 		 * attributes.
1028 		 */
1029 		memset(kvm_second_dirty_bitmap(ms), 0xff, kvm_dirty_bitmap_bytes(ms));
1030 		ram_pages += ms->npages;
1031 	}
1032 	atomic64_set(&kvm->arch.cmma_dirty_pages, ram_pages);
1033 	kvm->arch.migration_mode = 1;
1034 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_START_MIGRATION);
1035 	return 0;
1036 }
1037 
1038 /*
1039  * Must be called with kvm->slots_lock to avoid races with ourselves and
1040  * kvm_s390_vm_start_migration.
1041  */
1042 static int kvm_s390_vm_stop_migration(struct kvm *kvm)
1043 {
1044 	/* migration mode already disabled */
1045 	if (!kvm->arch.migration_mode)
1046 		return 0;
1047 	kvm->arch.migration_mode = 0;
1048 	if (kvm->arch.use_cmma)
1049 		kvm_s390_sync_request_broadcast(kvm, KVM_REQ_STOP_MIGRATION);
1050 	return 0;
1051 }
1052 
1053 static int kvm_s390_vm_set_migration(struct kvm *kvm,
1054 				     struct kvm_device_attr *attr)
1055 {
1056 	int res = -ENXIO;
1057 
1058 	mutex_lock(&kvm->slots_lock);
1059 	switch (attr->attr) {
1060 	case KVM_S390_VM_MIGRATION_START:
1061 		res = kvm_s390_vm_start_migration(kvm);
1062 		break;
1063 	case KVM_S390_VM_MIGRATION_STOP:
1064 		res = kvm_s390_vm_stop_migration(kvm);
1065 		break;
1066 	default:
1067 		break;
1068 	}
1069 	mutex_unlock(&kvm->slots_lock);
1070 
1071 	return res;
1072 }
1073 
1074 static int kvm_s390_vm_get_migration(struct kvm *kvm,
1075 				     struct kvm_device_attr *attr)
1076 {
1077 	u64 mig = kvm->arch.migration_mode;
1078 
1079 	if (attr->attr != KVM_S390_VM_MIGRATION_STATUS)
1080 		return -ENXIO;
1081 
1082 	if (copy_to_user((void __user *)attr->addr, &mig, sizeof(mig)))
1083 		return -EFAULT;
1084 	return 0;
1085 }
1086 
1087 static int kvm_s390_set_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1088 {
1089 	struct kvm_s390_vm_tod_clock gtod;
1090 
1091 	if (copy_from_user(&gtod, (void __user *)attr->addr, sizeof(gtod)))
1092 		return -EFAULT;
1093 
1094 	if (!test_kvm_facility(kvm, 139) && gtod.epoch_idx)
1095 		return -EINVAL;
1096 	kvm_s390_set_tod_clock(kvm, &gtod);
1097 
1098 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x, TOD base: 0x%llx",
1099 		gtod.epoch_idx, gtod.tod);
1100 
1101 	return 0;
1102 }
1103 
1104 static int kvm_s390_set_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1105 {
1106 	u8 gtod_high;
1107 
1108 	if (copy_from_user(&gtod_high, (void __user *)attr->addr,
1109 					   sizeof(gtod_high)))
1110 		return -EFAULT;
1111 
1112 	if (gtod_high != 0)
1113 		return -EINVAL;
1114 	VM_EVENT(kvm, 3, "SET: TOD extension: 0x%x", gtod_high);
1115 
1116 	return 0;
1117 }
1118 
1119 static int kvm_s390_set_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1120 {
1121 	struct kvm_s390_vm_tod_clock gtod = { 0 };
1122 
1123 	if (copy_from_user(&gtod.tod, (void __user *)attr->addr,
1124 			   sizeof(gtod.tod)))
1125 		return -EFAULT;
1126 
1127 	kvm_s390_set_tod_clock(kvm, &gtod);
1128 	VM_EVENT(kvm, 3, "SET: TOD base: 0x%llx", gtod.tod);
1129 	return 0;
1130 }
1131 
1132 static int kvm_s390_set_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1133 {
1134 	int ret;
1135 
1136 	if (attr->flags)
1137 		return -EINVAL;
1138 
1139 	switch (attr->attr) {
1140 	case KVM_S390_VM_TOD_EXT:
1141 		ret = kvm_s390_set_tod_ext(kvm, attr);
1142 		break;
1143 	case KVM_S390_VM_TOD_HIGH:
1144 		ret = kvm_s390_set_tod_high(kvm, attr);
1145 		break;
1146 	case KVM_S390_VM_TOD_LOW:
1147 		ret = kvm_s390_set_tod_low(kvm, attr);
1148 		break;
1149 	default:
1150 		ret = -ENXIO;
1151 		break;
1152 	}
1153 	return ret;
1154 }
1155 
1156 static void kvm_s390_get_tod_clock(struct kvm *kvm,
1157 				   struct kvm_s390_vm_tod_clock *gtod)
1158 {
1159 	struct kvm_s390_tod_clock_ext htod;
1160 
1161 	preempt_disable();
1162 
1163 	get_tod_clock_ext((char *)&htod);
1164 
1165 	gtod->tod = htod.tod + kvm->arch.epoch;
1166 	gtod->epoch_idx = 0;
1167 	if (test_kvm_facility(kvm, 139)) {
1168 		gtod->epoch_idx = htod.epoch_idx + kvm->arch.epdx;
1169 		if (gtod->tod < htod.tod)
1170 			gtod->epoch_idx += 1;
1171 	}
1172 
1173 	preempt_enable();
1174 }
1175 
1176 static int kvm_s390_get_tod_ext(struct kvm *kvm, struct kvm_device_attr *attr)
1177 {
1178 	struct kvm_s390_vm_tod_clock gtod;
1179 
1180 	memset(&gtod, 0, sizeof(gtod));
1181 	kvm_s390_get_tod_clock(kvm, &gtod);
1182 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1183 		return -EFAULT;
1184 
1185 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x, TOD base: 0x%llx",
1186 		gtod.epoch_idx, gtod.tod);
1187 	return 0;
1188 }
1189 
1190 static int kvm_s390_get_tod_high(struct kvm *kvm, struct kvm_device_attr *attr)
1191 {
1192 	u8 gtod_high = 0;
1193 
1194 	if (copy_to_user((void __user *)attr->addr, &gtod_high,
1195 					 sizeof(gtod_high)))
1196 		return -EFAULT;
1197 	VM_EVENT(kvm, 3, "QUERY: TOD extension: 0x%x", gtod_high);
1198 
1199 	return 0;
1200 }
1201 
1202 static int kvm_s390_get_tod_low(struct kvm *kvm, struct kvm_device_attr *attr)
1203 {
1204 	u64 gtod;
1205 
1206 	gtod = kvm_s390_get_tod_clock_fast(kvm);
1207 	if (copy_to_user((void __user *)attr->addr, &gtod, sizeof(gtod)))
1208 		return -EFAULT;
1209 	VM_EVENT(kvm, 3, "QUERY: TOD base: 0x%llx", gtod);
1210 
1211 	return 0;
1212 }
1213 
1214 static int kvm_s390_get_tod(struct kvm *kvm, struct kvm_device_attr *attr)
1215 {
1216 	int ret;
1217 
1218 	if (attr->flags)
1219 		return -EINVAL;
1220 
1221 	switch (attr->attr) {
1222 	case KVM_S390_VM_TOD_EXT:
1223 		ret = kvm_s390_get_tod_ext(kvm, attr);
1224 		break;
1225 	case KVM_S390_VM_TOD_HIGH:
1226 		ret = kvm_s390_get_tod_high(kvm, attr);
1227 		break;
1228 	case KVM_S390_VM_TOD_LOW:
1229 		ret = kvm_s390_get_tod_low(kvm, attr);
1230 		break;
1231 	default:
1232 		ret = -ENXIO;
1233 		break;
1234 	}
1235 	return ret;
1236 }
1237 
1238 static int kvm_s390_set_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1239 {
1240 	struct kvm_s390_vm_cpu_processor *proc;
1241 	u16 lowest_ibc, unblocked_ibc;
1242 	int ret = 0;
1243 
1244 	mutex_lock(&kvm->lock);
1245 	if (kvm->created_vcpus) {
1246 		ret = -EBUSY;
1247 		goto out;
1248 	}
1249 	proc = kzalloc(sizeof(*proc), GFP_KERNEL);
1250 	if (!proc) {
1251 		ret = -ENOMEM;
1252 		goto out;
1253 	}
1254 	if (!copy_from_user(proc, (void __user *)attr->addr,
1255 			    sizeof(*proc))) {
1256 		kvm->arch.model.cpuid = proc->cpuid;
1257 		lowest_ibc = sclp.ibc >> 16 & 0xfff;
1258 		unblocked_ibc = sclp.ibc & 0xfff;
1259 		if (lowest_ibc && proc->ibc) {
1260 			if (proc->ibc > unblocked_ibc)
1261 				kvm->arch.model.ibc = unblocked_ibc;
1262 			else if (proc->ibc < lowest_ibc)
1263 				kvm->arch.model.ibc = lowest_ibc;
1264 			else
1265 				kvm->arch.model.ibc = proc->ibc;
1266 		}
1267 		memcpy(kvm->arch.model.fac_list, proc->fac_list,
1268 		       S390_ARCH_FAC_LIST_SIZE_BYTE);
1269 		VM_EVENT(kvm, 3, "SET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1270 			 kvm->arch.model.ibc,
1271 			 kvm->arch.model.cpuid);
1272 		VM_EVENT(kvm, 3, "SET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1273 			 kvm->arch.model.fac_list[0],
1274 			 kvm->arch.model.fac_list[1],
1275 			 kvm->arch.model.fac_list[2]);
1276 	} else
1277 		ret = -EFAULT;
1278 	kfree(proc);
1279 out:
1280 	mutex_unlock(&kvm->lock);
1281 	return ret;
1282 }
1283 
1284 static int kvm_s390_set_processor_feat(struct kvm *kvm,
1285 				       struct kvm_device_attr *attr)
1286 {
1287 	struct kvm_s390_vm_cpu_feat data;
1288 
1289 	if (copy_from_user(&data, (void __user *)attr->addr, sizeof(data)))
1290 		return -EFAULT;
1291 	if (!bitmap_subset((unsigned long *) data.feat,
1292 			   kvm_s390_available_cpu_feat,
1293 			   KVM_S390_VM_CPU_FEAT_NR_BITS))
1294 		return -EINVAL;
1295 
1296 	mutex_lock(&kvm->lock);
1297 	if (kvm->created_vcpus) {
1298 		mutex_unlock(&kvm->lock);
1299 		return -EBUSY;
1300 	}
1301 	bitmap_copy(kvm->arch.cpu_feat, (unsigned long *) data.feat,
1302 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1303 	mutex_unlock(&kvm->lock);
1304 	VM_EVENT(kvm, 3, "SET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1305 			 data.feat[0],
1306 			 data.feat[1],
1307 			 data.feat[2]);
1308 	return 0;
1309 }
1310 
1311 static int kvm_s390_set_processor_subfunc(struct kvm *kvm,
1312 					  struct kvm_device_attr *attr)
1313 {
1314 	mutex_lock(&kvm->lock);
1315 	if (kvm->created_vcpus) {
1316 		mutex_unlock(&kvm->lock);
1317 		return -EBUSY;
1318 	}
1319 
1320 	if (copy_from_user(&kvm->arch.model.subfuncs, (void __user *)attr->addr,
1321 			   sizeof(struct kvm_s390_vm_cpu_subfunc))) {
1322 		mutex_unlock(&kvm->lock);
1323 		return -EFAULT;
1324 	}
1325 	mutex_unlock(&kvm->lock);
1326 
1327 	VM_EVENT(kvm, 3, "SET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1328 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1329 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1330 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1331 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1332 	VM_EVENT(kvm, 3, "SET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1333 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1334 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1335 	VM_EVENT(kvm, 3, "SET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1336 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1337 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1338 	VM_EVENT(kvm, 3, "SET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1339 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1340 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1341 	VM_EVENT(kvm, 3, "SET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1342 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1343 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1344 	VM_EVENT(kvm, 3, "SET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1345 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1346 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1347 	VM_EVENT(kvm, 3, "SET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1348 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1349 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1350 	VM_EVENT(kvm, 3, "SET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1351 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1352 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1353 	VM_EVENT(kvm, 3, "SET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1354 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1355 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1356 	VM_EVENT(kvm, 3, "SET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1357 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1358 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1359 	VM_EVENT(kvm, 3, "SET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1360 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1361 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1362 	VM_EVENT(kvm, 3, "SET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1363 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1364 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1365 	VM_EVENT(kvm, 3, "SET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1366 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1367 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1368 	VM_EVENT(kvm, 3, "SET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1369 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1370 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1371 	VM_EVENT(kvm, 3, "SET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1372 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1373 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1374 	VM_EVENT(kvm, 3, "SET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1375 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1376 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1377 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1378 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1379 	VM_EVENT(kvm, 3, "SET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1380 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1381 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1382 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1383 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1384 
1385 	return 0;
1386 }
1387 
1388 static int kvm_s390_set_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1389 {
1390 	int ret = -ENXIO;
1391 
1392 	switch (attr->attr) {
1393 	case KVM_S390_VM_CPU_PROCESSOR:
1394 		ret = kvm_s390_set_processor(kvm, attr);
1395 		break;
1396 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1397 		ret = kvm_s390_set_processor_feat(kvm, attr);
1398 		break;
1399 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1400 		ret = kvm_s390_set_processor_subfunc(kvm, attr);
1401 		break;
1402 	}
1403 	return ret;
1404 }
1405 
1406 static int kvm_s390_get_processor(struct kvm *kvm, struct kvm_device_attr *attr)
1407 {
1408 	struct kvm_s390_vm_cpu_processor *proc;
1409 	int ret = 0;
1410 
1411 	proc = kzalloc(sizeof(*proc), GFP_KERNEL);
1412 	if (!proc) {
1413 		ret = -ENOMEM;
1414 		goto out;
1415 	}
1416 	proc->cpuid = kvm->arch.model.cpuid;
1417 	proc->ibc = kvm->arch.model.ibc;
1418 	memcpy(&proc->fac_list, kvm->arch.model.fac_list,
1419 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1420 	VM_EVENT(kvm, 3, "GET: guest ibc: 0x%4.4x, guest cpuid: 0x%16.16llx",
1421 		 kvm->arch.model.ibc,
1422 		 kvm->arch.model.cpuid);
1423 	VM_EVENT(kvm, 3, "GET: guest faclist: 0x%16.16llx.%16.16llx.%16.16llx",
1424 		 kvm->arch.model.fac_list[0],
1425 		 kvm->arch.model.fac_list[1],
1426 		 kvm->arch.model.fac_list[2]);
1427 	if (copy_to_user((void __user *)attr->addr, proc, sizeof(*proc)))
1428 		ret = -EFAULT;
1429 	kfree(proc);
1430 out:
1431 	return ret;
1432 }
1433 
1434 static int kvm_s390_get_machine(struct kvm *kvm, struct kvm_device_attr *attr)
1435 {
1436 	struct kvm_s390_vm_cpu_machine *mach;
1437 	int ret = 0;
1438 
1439 	mach = kzalloc(sizeof(*mach), GFP_KERNEL);
1440 	if (!mach) {
1441 		ret = -ENOMEM;
1442 		goto out;
1443 	}
1444 	get_cpu_id((struct cpuid *) &mach->cpuid);
1445 	mach->ibc = sclp.ibc;
1446 	memcpy(&mach->fac_mask, kvm->arch.model.fac_mask,
1447 	       S390_ARCH_FAC_LIST_SIZE_BYTE);
1448 	memcpy((unsigned long *)&mach->fac_list, S390_lowcore.stfle_fac_list,
1449 	       sizeof(S390_lowcore.stfle_fac_list));
1450 	VM_EVENT(kvm, 3, "GET: host ibc:  0x%4.4x, host cpuid:  0x%16.16llx",
1451 		 kvm->arch.model.ibc,
1452 		 kvm->arch.model.cpuid);
1453 	VM_EVENT(kvm, 3, "GET: host facmask:  0x%16.16llx.%16.16llx.%16.16llx",
1454 		 mach->fac_mask[0],
1455 		 mach->fac_mask[1],
1456 		 mach->fac_mask[2]);
1457 	VM_EVENT(kvm, 3, "GET: host faclist:  0x%16.16llx.%16.16llx.%16.16llx",
1458 		 mach->fac_list[0],
1459 		 mach->fac_list[1],
1460 		 mach->fac_list[2]);
1461 	if (copy_to_user((void __user *)attr->addr, mach, sizeof(*mach)))
1462 		ret = -EFAULT;
1463 	kfree(mach);
1464 out:
1465 	return ret;
1466 }
1467 
1468 static int kvm_s390_get_processor_feat(struct kvm *kvm,
1469 				       struct kvm_device_attr *attr)
1470 {
1471 	struct kvm_s390_vm_cpu_feat data;
1472 
1473 	bitmap_copy((unsigned long *) data.feat, kvm->arch.cpu_feat,
1474 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1475 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1476 		return -EFAULT;
1477 	VM_EVENT(kvm, 3, "GET: guest feat: 0x%16.16llx.0x%16.16llx.0x%16.16llx",
1478 			 data.feat[0],
1479 			 data.feat[1],
1480 			 data.feat[2]);
1481 	return 0;
1482 }
1483 
1484 static int kvm_s390_get_machine_feat(struct kvm *kvm,
1485 				     struct kvm_device_attr *attr)
1486 {
1487 	struct kvm_s390_vm_cpu_feat data;
1488 
1489 	bitmap_copy((unsigned long *) data.feat,
1490 		    kvm_s390_available_cpu_feat,
1491 		    KVM_S390_VM_CPU_FEAT_NR_BITS);
1492 	if (copy_to_user((void __user *)attr->addr, &data, sizeof(data)))
1493 		return -EFAULT;
1494 	VM_EVENT(kvm, 3, "GET: host feat:  0x%16.16llx.0x%16.16llx.0x%16.16llx",
1495 			 data.feat[0],
1496 			 data.feat[1],
1497 			 data.feat[2]);
1498 	return 0;
1499 }
1500 
1501 static int kvm_s390_get_processor_subfunc(struct kvm *kvm,
1502 					  struct kvm_device_attr *attr)
1503 {
1504 	if (copy_to_user((void __user *)attr->addr, &kvm->arch.model.subfuncs,
1505 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1506 		return -EFAULT;
1507 
1508 	VM_EVENT(kvm, 3, "GET: guest PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1509 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[0],
1510 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[1],
1511 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[2],
1512 		 ((unsigned long *) &kvm->arch.model.subfuncs.plo)[3]);
1513 	VM_EVENT(kvm, 3, "GET: guest PTFF   subfunc 0x%16.16lx.%16.16lx",
1514 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[0],
1515 		 ((unsigned long *) &kvm->arch.model.subfuncs.ptff)[1]);
1516 	VM_EVENT(kvm, 3, "GET: guest KMAC   subfunc 0x%16.16lx.%16.16lx",
1517 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[0],
1518 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmac)[1]);
1519 	VM_EVENT(kvm, 3, "GET: guest KMC    subfunc 0x%16.16lx.%16.16lx",
1520 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[0],
1521 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmc)[1]);
1522 	VM_EVENT(kvm, 3, "GET: guest KM     subfunc 0x%16.16lx.%16.16lx",
1523 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[0],
1524 		 ((unsigned long *) &kvm->arch.model.subfuncs.km)[1]);
1525 	VM_EVENT(kvm, 3, "GET: guest KIMD   subfunc 0x%16.16lx.%16.16lx",
1526 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[0],
1527 		 ((unsigned long *) &kvm->arch.model.subfuncs.kimd)[1]);
1528 	VM_EVENT(kvm, 3, "GET: guest KLMD   subfunc 0x%16.16lx.%16.16lx",
1529 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[0],
1530 		 ((unsigned long *) &kvm->arch.model.subfuncs.klmd)[1]);
1531 	VM_EVENT(kvm, 3, "GET: guest PCKMO  subfunc 0x%16.16lx.%16.16lx",
1532 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[0],
1533 		 ((unsigned long *) &kvm->arch.model.subfuncs.pckmo)[1]);
1534 	VM_EVENT(kvm, 3, "GET: guest KMCTR  subfunc 0x%16.16lx.%16.16lx",
1535 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[0],
1536 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmctr)[1]);
1537 	VM_EVENT(kvm, 3, "GET: guest KMF    subfunc 0x%16.16lx.%16.16lx",
1538 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[0],
1539 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmf)[1]);
1540 	VM_EVENT(kvm, 3, "GET: guest KMO    subfunc 0x%16.16lx.%16.16lx",
1541 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[0],
1542 		 ((unsigned long *) &kvm->arch.model.subfuncs.kmo)[1]);
1543 	VM_EVENT(kvm, 3, "GET: guest PCC    subfunc 0x%16.16lx.%16.16lx",
1544 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[0],
1545 		 ((unsigned long *) &kvm->arch.model.subfuncs.pcc)[1]);
1546 	VM_EVENT(kvm, 3, "GET: guest PPNO   subfunc 0x%16.16lx.%16.16lx",
1547 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[0],
1548 		 ((unsigned long *) &kvm->arch.model.subfuncs.ppno)[1]);
1549 	VM_EVENT(kvm, 3, "GET: guest KMA    subfunc 0x%16.16lx.%16.16lx",
1550 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[0],
1551 		 ((unsigned long *) &kvm->arch.model.subfuncs.kma)[1]);
1552 	VM_EVENT(kvm, 3, "GET: guest KDSA   subfunc 0x%16.16lx.%16.16lx",
1553 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[0],
1554 		 ((unsigned long *) &kvm->arch.model.subfuncs.kdsa)[1]);
1555 	VM_EVENT(kvm, 3, "GET: guest SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1556 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[0],
1557 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[1],
1558 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[2],
1559 		 ((unsigned long *) &kvm->arch.model.subfuncs.sortl)[3]);
1560 	VM_EVENT(kvm, 3, "GET: guest DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1561 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[0],
1562 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[1],
1563 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[2],
1564 		 ((unsigned long *) &kvm->arch.model.subfuncs.dfltcc)[3]);
1565 
1566 	return 0;
1567 }
1568 
1569 static int kvm_s390_get_machine_subfunc(struct kvm *kvm,
1570 					struct kvm_device_attr *attr)
1571 {
1572 	if (copy_to_user((void __user *)attr->addr, &kvm_s390_available_subfunc,
1573 	    sizeof(struct kvm_s390_vm_cpu_subfunc)))
1574 		return -EFAULT;
1575 
1576 	VM_EVENT(kvm, 3, "GET: host  PLO    subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1577 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[0],
1578 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[1],
1579 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[2],
1580 		 ((unsigned long *) &kvm_s390_available_subfunc.plo)[3]);
1581 	VM_EVENT(kvm, 3, "GET: host  PTFF   subfunc 0x%16.16lx.%16.16lx",
1582 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[0],
1583 		 ((unsigned long *) &kvm_s390_available_subfunc.ptff)[1]);
1584 	VM_EVENT(kvm, 3, "GET: host  KMAC   subfunc 0x%16.16lx.%16.16lx",
1585 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[0],
1586 		 ((unsigned long *) &kvm_s390_available_subfunc.kmac)[1]);
1587 	VM_EVENT(kvm, 3, "GET: host  KMC    subfunc 0x%16.16lx.%16.16lx",
1588 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[0],
1589 		 ((unsigned long *) &kvm_s390_available_subfunc.kmc)[1]);
1590 	VM_EVENT(kvm, 3, "GET: host  KM     subfunc 0x%16.16lx.%16.16lx",
1591 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[0],
1592 		 ((unsigned long *) &kvm_s390_available_subfunc.km)[1]);
1593 	VM_EVENT(kvm, 3, "GET: host  KIMD   subfunc 0x%16.16lx.%16.16lx",
1594 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[0],
1595 		 ((unsigned long *) &kvm_s390_available_subfunc.kimd)[1]);
1596 	VM_EVENT(kvm, 3, "GET: host  KLMD   subfunc 0x%16.16lx.%16.16lx",
1597 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[0],
1598 		 ((unsigned long *) &kvm_s390_available_subfunc.klmd)[1]);
1599 	VM_EVENT(kvm, 3, "GET: host  PCKMO  subfunc 0x%16.16lx.%16.16lx",
1600 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[0],
1601 		 ((unsigned long *) &kvm_s390_available_subfunc.pckmo)[1]);
1602 	VM_EVENT(kvm, 3, "GET: host  KMCTR  subfunc 0x%16.16lx.%16.16lx",
1603 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[0],
1604 		 ((unsigned long *) &kvm_s390_available_subfunc.kmctr)[1]);
1605 	VM_EVENT(kvm, 3, "GET: host  KMF    subfunc 0x%16.16lx.%16.16lx",
1606 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[0],
1607 		 ((unsigned long *) &kvm_s390_available_subfunc.kmf)[1]);
1608 	VM_EVENT(kvm, 3, "GET: host  KMO    subfunc 0x%16.16lx.%16.16lx",
1609 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[0],
1610 		 ((unsigned long *) &kvm_s390_available_subfunc.kmo)[1]);
1611 	VM_EVENT(kvm, 3, "GET: host  PCC    subfunc 0x%16.16lx.%16.16lx",
1612 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[0],
1613 		 ((unsigned long *) &kvm_s390_available_subfunc.pcc)[1]);
1614 	VM_EVENT(kvm, 3, "GET: host  PPNO   subfunc 0x%16.16lx.%16.16lx",
1615 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[0],
1616 		 ((unsigned long *) &kvm_s390_available_subfunc.ppno)[1]);
1617 	VM_EVENT(kvm, 3, "GET: host  KMA    subfunc 0x%16.16lx.%16.16lx",
1618 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[0],
1619 		 ((unsigned long *) &kvm_s390_available_subfunc.kma)[1]);
1620 	VM_EVENT(kvm, 3, "GET: host  KDSA   subfunc 0x%16.16lx.%16.16lx",
1621 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[0],
1622 		 ((unsigned long *) &kvm_s390_available_subfunc.kdsa)[1]);
1623 	VM_EVENT(kvm, 3, "GET: host  SORTL  subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1624 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[0],
1625 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[1],
1626 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[2],
1627 		 ((unsigned long *) &kvm_s390_available_subfunc.sortl)[3]);
1628 	VM_EVENT(kvm, 3, "GET: host  DFLTCC subfunc 0x%16.16lx.%16.16lx.%16.16lx.%16.16lx",
1629 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[0],
1630 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[1],
1631 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[2],
1632 		 ((unsigned long *) &kvm_s390_available_subfunc.dfltcc)[3]);
1633 
1634 	return 0;
1635 }
1636 
1637 static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
1638 {
1639 	int ret = -ENXIO;
1640 
1641 	switch (attr->attr) {
1642 	case KVM_S390_VM_CPU_PROCESSOR:
1643 		ret = kvm_s390_get_processor(kvm, attr);
1644 		break;
1645 	case KVM_S390_VM_CPU_MACHINE:
1646 		ret = kvm_s390_get_machine(kvm, attr);
1647 		break;
1648 	case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1649 		ret = kvm_s390_get_processor_feat(kvm, attr);
1650 		break;
1651 	case KVM_S390_VM_CPU_MACHINE_FEAT:
1652 		ret = kvm_s390_get_machine_feat(kvm, attr);
1653 		break;
1654 	case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1655 		ret = kvm_s390_get_processor_subfunc(kvm, attr);
1656 		break;
1657 	case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1658 		ret = kvm_s390_get_machine_subfunc(kvm, attr);
1659 		break;
1660 	}
1661 	return ret;
1662 }
1663 
1664 static int kvm_s390_vm_set_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1665 {
1666 	int ret;
1667 
1668 	switch (attr->group) {
1669 	case KVM_S390_VM_MEM_CTRL:
1670 		ret = kvm_s390_set_mem_control(kvm, attr);
1671 		break;
1672 	case KVM_S390_VM_TOD:
1673 		ret = kvm_s390_set_tod(kvm, attr);
1674 		break;
1675 	case KVM_S390_VM_CPU_MODEL:
1676 		ret = kvm_s390_set_cpu_model(kvm, attr);
1677 		break;
1678 	case KVM_S390_VM_CRYPTO:
1679 		ret = kvm_s390_vm_set_crypto(kvm, attr);
1680 		break;
1681 	case KVM_S390_VM_MIGRATION:
1682 		ret = kvm_s390_vm_set_migration(kvm, attr);
1683 		break;
1684 	default:
1685 		ret = -ENXIO;
1686 		break;
1687 	}
1688 
1689 	return ret;
1690 }
1691 
1692 static int kvm_s390_vm_get_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1693 {
1694 	int ret;
1695 
1696 	switch (attr->group) {
1697 	case KVM_S390_VM_MEM_CTRL:
1698 		ret = kvm_s390_get_mem_control(kvm, attr);
1699 		break;
1700 	case KVM_S390_VM_TOD:
1701 		ret = kvm_s390_get_tod(kvm, attr);
1702 		break;
1703 	case KVM_S390_VM_CPU_MODEL:
1704 		ret = kvm_s390_get_cpu_model(kvm, attr);
1705 		break;
1706 	case KVM_S390_VM_MIGRATION:
1707 		ret = kvm_s390_vm_get_migration(kvm, attr);
1708 		break;
1709 	default:
1710 		ret = -ENXIO;
1711 		break;
1712 	}
1713 
1714 	return ret;
1715 }
1716 
1717 static int kvm_s390_vm_has_attr(struct kvm *kvm, struct kvm_device_attr *attr)
1718 {
1719 	int ret;
1720 
1721 	switch (attr->group) {
1722 	case KVM_S390_VM_MEM_CTRL:
1723 		switch (attr->attr) {
1724 		case KVM_S390_VM_MEM_ENABLE_CMMA:
1725 		case KVM_S390_VM_MEM_CLR_CMMA:
1726 			ret = sclp.has_cmma ? 0 : -ENXIO;
1727 			break;
1728 		case KVM_S390_VM_MEM_LIMIT_SIZE:
1729 			ret = 0;
1730 			break;
1731 		default:
1732 			ret = -ENXIO;
1733 			break;
1734 		}
1735 		break;
1736 	case KVM_S390_VM_TOD:
1737 		switch (attr->attr) {
1738 		case KVM_S390_VM_TOD_LOW:
1739 		case KVM_S390_VM_TOD_HIGH:
1740 			ret = 0;
1741 			break;
1742 		default:
1743 			ret = -ENXIO;
1744 			break;
1745 		}
1746 		break;
1747 	case KVM_S390_VM_CPU_MODEL:
1748 		switch (attr->attr) {
1749 		case KVM_S390_VM_CPU_PROCESSOR:
1750 		case KVM_S390_VM_CPU_MACHINE:
1751 		case KVM_S390_VM_CPU_PROCESSOR_FEAT:
1752 		case KVM_S390_VM_CPU_MACHINE_FEAT:
1753 		case KVM_S390_VM_CPU_MACHINE_SUBFUNC:
1754 		case KVM_S390_VM_CPU_PROCESSOR_SUBFUNC:
1755 			ret = 0;
1756 			break;
1757 		default:
1758 			ret = -ENXIO;
1759 			break;
1760 		}
1761 		break;
1762 	case KVM_S390_VM_CRYPTO:
1763 		switch (attr->attr) {
1764 		case KVM_S390_VM_CRYPTO_ENABLE_AES_KW:
1765 		case KVM_S390_VM_CRYPTO_ENABLE_DEA_KW:
1766 		case KVM_S390_VM_CRYPTO_DISABLE_AES_KW:
1767 		case KVM_S390_VM_CRYPTO_DISABLE_DEA_KW:
1768 			ret = 0;
1769 			break;
1770 		case KVM_S390_VM_CRYPTO_ENABLE_APIE:
1771 		case KVM_S390_VM_CRYPTO_DISABLE_APIE:
1772 			ret = ap_instructions_available() ? 0 : -ENXIO;
1773 			break;
1774 		default:
1775 			ret = -ENXIO;
1776 			break;
1777 		}
1778 		break;
1779 	case KVM_S390_VM_MIGRATION:
1780 		ret = 0;
1781 		break;
1782 	default:
1783 		ret = -ENXIO;
1784 		break;
1785 	}
1786 
1787 	return ret;
1788 }
1789 
1790 static long kvm_s390_get_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1791 {
1792 	uint8_t *keys;
1793 	uint64_t hva;
1794 	int srcu_idx, i, r = 0;
1795 
1796 	if (args->flags != 0)
1797 		return -EINVAL;
1798 
1799 	/* Is this guest using storage keys? */
1800 	if (!mm_uses_skeys(current->mm))
1801 		return KVM_S390_GET_SKEYS_NONE;
1802 
1803 	/* Enforce sane limit on memory allocation */
1804 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1805 		return -EINVAL;
1806 
1807 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL);
1808 	if (!keys)
1809 		return -ENOMEM;
1810 
1811 	down_read(&current->mm->mmap_sem);
1812 	srcu_idx = srcu_read_lock(&kvm->srcu);
1813 	for (i = 0; i < args->count; i++) {
1814 		hva = gfn_to_hva(kvm, args->start_gfn + i);
1815 		if (kvm_is_error_hva(hva)) {
1816 			r = -EFAULT;
1817 			break;
1818 		}
1819 
1820 		r = get_guest_storage_key(current->mm, hva, &keys[i]);
1821 		if (r)
1822 			break;
1823 	}
1824 	srcu_read_unlock(&kvm->srcu, srcu_idx);
1825 	up_read(&current->mm->mmap_sem);
1826 
1827 	if (!r) {
1828 		r = copy_to_user((uint8_t __user *)args->skeydata_addr, keys,
1829 				 sizeof(uint8_t) * args->count);
1830 		if (r)
1831 			r = -EFAULT;
1832 	}
1833 
1834 	kvfree(keys);
1835 	return r;
1836 }
1837 
1838 static long kvm_s390_set_skeys(struct kvm *kvm, struct kvm_s390_skeys *args)
1839 {
1840 	uint8_t *keys;
1841 	uint64_t hva;
1842 	int srcu_idx, i, r = 0;
1843 	bool unlocked;
1844 
1845 	if (args->flags != 0)
1846 		return -EINVAL;
1847 
1848 	/* Enforce sane limit on memory allocation */
1849 	if (args->count < 1 || args->count > KVM_S390_SKEYS_MAX)
1850 		return -EINVAL;
1851 
1852 	keys = kvmalloc_array(args->count, sizeof(uint8_t), GFP_KERNEL);
1853 	if (!keys)
1854 		return -ENOMEM;
1855 
1856 	r = copy_from_user(keys, (uint8_t __user *)args->skeydata_addr,
1857 			   sizeof(uint8_t) * args->count);
1858 	if (r) {
1859 		r = -EFAULT;
1860 		goto out;
1861 	}
1862 
1863 	/* Enable storage key handling for the guest */
1864 	r = s390_enable_skey();
1865 	if (r)
1866 		goto out;
1867 
1868 	i = 0;
1869 	down_read(&current->mm->mmap_sem);
1870 	srcu_idx = srcu_read_lock(&kvm->srcu);
1871         while (i < args->count) {
1872 		unlocked = false;
1873 		hva = gfn_to_hva(kvm, args->start_gfn + i);
1874 		if (kvm_is_error_hva(hva)) {
1875 			r = -EFAULT;
1876 			break;
1877 		}
1878 
1879 		/* Lowest order bit is reserved */
1880 		if (keys[i] & 0x01) {
1881 			r = -EINVAL;
1882 			break;
1883 		}
1884 
1885 		r = set_guest_storage_key(current->mm, hva, keys[i], 0);
1886 		if (r) {
1887 			r = fixup_user_fault(current, current->mm, hva,
1888 					     FAULT_FLAG_WRITE, &unlocked);
1889 			if (r)
1890 				break;
1891 		}
1892 		if (!r)
1893 			i++;
1894 	}
1895 	srcu_read_unlock(&kvm->srcu, srcu_idx);
1896 	up_read(&current->mm->mmap_sem);
1897 out:
1898 	kvfree(keys);
1899 	return r;
1900 }
1901 
1902 /*
1903  * Base address and length must be sent at the start of each block, therefore
1904  * it's cheaper to send some clean data, as long as it's less than the size of
1905  * two longs.
1906  */
1907 #define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
1908 /* for consistency */
1909 #define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
1910 
1911 /*
1912  * Similar to gfn_to_memslot, but returns the index of a memslot also when the
1913  * address falls in a hole. In that case the index of one of the memslots
1914  * bordering the hole is returned.
1915  */
1916 static int gfn_to_memslot_approx(struct kvm_memslots *slots, gfn_t gfn)
1917 {
1918 	int start = 0, end = slots->used_slots;
1919 	int slot = atomic_read(&slots->lru_slot);
1920 	struct kvm_memory_slot *memslots = slots->memslots;
1921 
1922 	if (gfn >= memslots[slot].base_gfn &&
1923 	    gfn < memslots[slot].base_gfn + memslots[slot].npages)
1924 		return slot;
1925 
1926 	while (start < end) {
1927 		slot = start + (end - start) / 2;
1928 
1929 		if (gfn >= memslots[slot].base_gfn)
1930 			end = slot;
1931 		else
1932 			start = slot + 1;
1933 	}
1934 
1935 	if (gfn >= memslots[start].base_gfn &&
1936 	    gfn < memslots[start].base_gfn + memslots[start].npages) {
1937 		atomic_set(&slots->lru_slot, start);
1938 	}
1939 
1940 	return start;
1941 }
1942 
1943 static int kvm_s390_peek_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
1944 			      u8 *res, unsigned long bufsize)
1945 {
1946 	unsigned long pgstev, hva, cur_gfn = args->start_gfn;
1947 
1948 	args->count = 0;
1949 	while (args->count < bufsize) {
1950 		hva = gfn_to_hva(kvm, cur_gfn);
1951 		/*
1952 		 * We return an error if the first value was invalid, but we
1953 		 * return successfully if at least one value was copied.
1954 		 */
1955 		if (kvm_is_error_hva(hva))
1956 			return args->count ? 0 : -EFAULT;
1957 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
1958 			pgstev = 0;
1959 		res[args->count++] = (pgstev >> 24) & 0x43;
1960 		cur_gfn++;
1961 	}
1962 
1963 	return 0;
1964 }
1965 
1966 static unsigned long kvm_s390_next_dirty_cmma(struct kvm_memslots *slots,
1967 					      unsigned long cur_gfn)
1968 {
1969 	int slotidx = gfn_to_memslot_approx(slots, cur_gfn);
1970 	struct kvm_memory_slot *ms = slots->memslots + slotidx;
1971 	unsigned long ofs = cur_gfn - ms->base_gfn;
1972 
1973 	if (ms->base_gfn + ms->npages <= cur_gfn) {
1974 		slotidx--;
1975 		/* If we are above the highest slot, wrap around */
1976 		if (slotidx < 0)
1977 			slotidx = slots->used_slots - 1;
1978 
1979 		ms = slots->memslots + slotidx;
1980 		ofs = 0;
1981 	}
1982 	ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, ofs);
1983 	while ((slotidx > 0) && (ofs >= ms->npages)) {
1984 		slotidx--;
1985 		ms = slots->memslots + slotidx;
1986 		ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, 0);
1987 	}
1988 	return ms->base_gfn + ofs;
1989 }
1990 
1991 static int kvm_s390_get_cmma(struct kvm *kvm, struct kvm_s390_cmma_log *args,
1992 			     u8 *res, unsigned long bufsize)
1993 {
1994 	unsigned long mem_end, cur_gfn, next_gfn, hva, pgstev;
1995 	struct kvm_memslots *slots = kvm_memslots(kvm);
1996 	struct kvm_memory_slot *ms;
1997 
1998 	cur_gfn = kvm_s390_next_dirty_cmma(slots, args->start_gfn);
1999 	ms = gfn_to_memslot(kvm, cur_gfn);
2000 	args->count = 0;
2001 	args->start_gfn = cur_gfn;
2002 	if (!ms)
2003 		return 0;
2004 	next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2005 	mem_end = slots->memslots[0].base_gfn + slots->memslots[0].npages;
2006 
2007 	while (args->count < bufsize) {
2008 		hva = gfn_to_hva(kvm, cur_gfn);
2009 		if (kvm_is_error_hva(hva))
2010 			return 0;
2011 		/* Decrement only if we actually flipped the bit to 0 */
2012 		if (test_and_clear_bit(cur_gfn - ms->base_gfn, kvm_second_dirty_bitmap(ms)))
2013 			atomic64_dec(&kvm->arch.cmma_dirty_pages);
2014 		if (get_pgste(kvm->mm, hva, &pgstev) < 0)
2015 			pgstev = 0;
2016 		/* Save the value */
2017 		res[args->count++] = (pgstev >> 24) & 0x43;
2018 		/* If the next bit is too far away, stop. */
2019 		if (next_gfn > cur_gfn + KVM_S390_MAX_BIT_DISTANCE)
2020 			return 0;
2021 		/* If we reached the previous "next", find the next one */
2022 		if (cur_gfn == next_gfn)
2023 			next_gfn = kvm_s390_next_dirty_cmma(slots, cur_gfn + 1);
2024 		/* Reached the end of memory or of the buffer, stop */
2025 		if ((next_gfn >= mem_end) ||
2026 		    (next_gfn - args->start_gfn >= bufsize))
2027 			return 0;
2028 		cur_gfn++;
2029 		/* Reached the end of the current memslot, take the next one. */
2030 		if (cur_gfn - ms->base_gfn >= ms->npages) {
2031 			ms = gfn_to_memslot(kvm, cur_gfn);
2032 			if (!ms)
2033 				return 0;
2034 		}
2035 	}
2036 	return 0;
2037 }
2038 
2039 /*
2040  * This function searches for the next page with dirty CMMA attributes, and
2041  * saves the attributes in the buffer up to either the end of the buffer or
2042  * until a block of at least KVM_S390_MAX_BIT_DISTANCE clean bits is found;
2043  * no trailing clean bytes are saved.
2044  * In case no dirty bits were found, or if CMMA was not enabled or used, the
2045  * output buffer will indicate 0 as length.
2046  */
2047 static int kvm_s390_get_cmma_bits(struct kvm *kvm,
2048 				  struct kvm_s390_cmma_log *args)
2049 {
2050 	unsigned long bufsize;
2051 	int srcu_idx, peek, ret;
2052 	u8 *values;
2053 
2054 	if (!kvm->arch.use_cmma)
2055 		return -ENXIO;
2056 	/* Invalid/unsupported flags were specified */
2057 	if (args->flags & ~KVM_S390_CMMA_PEEK)
2058 		return -EINVAL;
2059 	/* Migration mode query, and we are not doing a migration */
2060 	peek = !!(args->flags & KVM_S390_CMMA_PEEK);
2061 	if (!peek && !kvm->arch.migration_mode)
2062 		return -EINVAL;
2063 	/* CMMA is disabled or was not used, or the buffer has length zero */
2064 	bufsize = min(args->count, KVM_S390_CMMA_SIZE_MAX);
2065 	if (!bufsize || !kvm->mm->context.uses_cmm) {
2066 		memset(args, 0, sizeof(*args));
2067 		return 0;
2068 	}
2069 	/* We are not peeking, and there are no dirty pages */
2070 	if (!peek && !atomic64_read(&kvm->arch.cmma_dirty_pages)) {
2071 		memset(args, 0, sizeof(*args));
2072 		return 0;
2073 	}
2074 
2075 	values = vmalloc(bufsize);
2076 	if (!values)
2077 		return -ENOMEM;
2078 
2079 	down_read(&kvm->mm->mmap_sem);
2080 	srcu_idx = srcu_read_lock(&kvm->srcu);
2081 	if (peek)
2082 		ret = kvm_s390_peek_cmma(kvm, args, values, bufsize);
2083 	else
2084 		ret = kvm_s390_get_cmma(kvm, args, values, bufsize);
2085 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2086 	up_read(&kvm->mm->mmap_sem);
2087 
2088 	if (kvm->arch.migration_mode)
2089 		args->remaining = atomic64_read(&kvm->arch.cmma_dirty_pages);
2090 	else
2091 		args->remaining = 0;
2092 
2093 	if (copy_to_user((void __user *)args->values, values, args->count))
2094 		ret = -EFAULT;
2095 
2096 	vfree(values);
2097 	return ret;
2098 }
2099 
2100 /*
2101  * This function sets the CMMA attributes for the given pages. If the input
2102  * buffer has zero length, no action is taken, otherwise the attributes are
2103  * set and the mm->context.uses_cmm flag is set.
2104  */
2105 static int kvm_s390_set_cmma_bits(struct kvm *kvm,
2106 				  const struct kvm_s390_cmma_log *args)
2107 {
2108 	unsigned long hva, mask, pgstev, i;
2109 	uint8_t *bits;
2110 	int srcu_idx, r = 0;
2111 
2112 	mask = args->mask;
2113 
2114 	if (!kvm->arch.use_cmma)
2115 		return -ENXIO;
2116 	/* invalid/unsupported flags */
2117 	if (args->flags != 0)
2118 		return -EINVAL;
2119 	/* Enforce sane limit on memory allocation */
2120 	if (args->count > KVM_S390_CMMA_SIZE_MAX)
2121 		return -EINVAL;
2122 	/* Nothing to do */
2123 	if (args->count == 0)
2124 		return 0;
2125 
2126 	bits = vmalloc(array_size(sizeof(*bits), args->count));
2127 	if (!bits)
2128 		return -ENOMEM;
2129 
2130 	r = copy_from_user(bits, (void __user *)args->values, args->count);
2131 	if (r) {
2132 		r = -EFAULT;
2133 		goto out;
2134 	}
2135 
2136 	down_read(&kvm->mm->mmap_sem);
2137 	srcu_idx = srcu_read_lock(&kvm->srcu);
2138 	for (i = 0; i < args->count; i++) {
2139 		hva = gfn_to_hva(kvm, args->start_gfn + i);
2140 		if (kvm_is_error_hva(hva)) {
2141 			r = -EFAULT;
2142 			break;
2143 		}
2144 
2145 		pgstev = bits[i];
2146 		pgstev = pgstev << 24;
2147 		mask &= _PGSTE_GPS_USAGE_MASK | _PGSTE_GPS_NODAT;
2148 		set_pgste_bits(kvm->mm, hva, mask, pgstev);
2149 	}
2150 	srcu_read_unlock(&kvm->srcu, srcu_idx);
2151 	up_read(&kvm->mm->mmap_sem);
2152 
2153 	if (!kvm->mm->context.uses_cmm) {
2154 		down_write(&kvm->mm->mmap_sem);
2155 		kvm->mm->context.uses_cmm = 1;
2156 		up_write(&kvm->mm->mmap_sem);
2157 	}
2158 out:
2159 	vfree(bits);
2160 	return r;
2161 }
2162 
2163 long kvm_arch_vm_ioctl(struct file *filp,
2164 		       unsigned int ioctl, unsigned long arg)
2165 {
2166 	struct kvm *kvm = filp->private_data;
2167 	void __user *argp = (void __user *)arg;
2168 	struct kvm_device_attr attr;
2169 	int r;
2170 
2171 	switch (ioctl) {
2172 	case KVM_S390_INTERRUPT: {
2173 		struct kvm_s390_interrupt s390int;
2174 
2175 		r = -EFAULT;
2176 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
2177 			break;
2178 		r = kvm_s390_inject_vm(kvm, &s390int);
2179 		break;
2180 	}
2181 	case KVM_CREATE_IRQCHIP: {
2182 		struct kvm_irq_routing_entry routing;
2183 
2184 		r = -EINVAL;
2185 		if (kvm->arch.use_irqchip) {
2186 			/* Set up dummy routing. */
2187 			memset(&routing, 0, sizeof(routing));
2188 			r = kvm_set_irq_routing(kvm, &routing, 0, 0);
2189 		}
2190 		break;
2191 	}
2192 	case KVM_SET_DEVICE_ATTR: {
2193 		r = -EFAULT;
2194 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2195 			break;
2196 		r = kvm_s390_vm_set_attr(kvm, &attr);
2197 		break;
2198 	}
2199 	case KVM_GET_DEVICE_ATTR: {
2200 		r = -EFAULT;
2201 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2202 			break;
2203 		r = kvm_s390_vm_get_attr(kvm, &attr);
2204 		break;
2205 	}
2206 	case KVM_HAS_DEVICE_ATTR: {
2207 		r = -EFAULT;
2208 		if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
2209 			break;
2210 		r = kvm_s390_vm_has_attr(kvm, &attr);
2211 		break;
2212 	}
2213 	case KVM_S390_GET_SKEYS: {
2214 		struct kvm_s390_skeys args;
2215 
2216 		r = -EFAULT;
2217 		if (copy_from_user(&args, argp,
2218 				   sizeof(struct kvm_s390_skeys)))
2219 			break;
2220 		r = kvm_s390_get_skeys(kvm, &args);
2221 		break;
2222 	}
2223 	case KVM_S390_SET_SKEYS: {
2224 		struct kvm_s390_skeys args;
2225 
2226 		r = -EFAULT;
2227 		if (copy_from_user(&args, argp,
2228 				   sizeof(struct kvm_s390_skeys)))
2229 			break;
2230 		r = kvm_s390_set_skeys(kvm, &args);
2231 		break;
2232 	}
2233 	case KVM_S390_GET_CMMA_BITS: {
2234 		struct kvm_s390_cmma_log args;
2235 
2236 		r = -EFAULT;
2237 		if (copy_from_user(&args, argp, sizeof(args)))
2238 			break;
2239 		mutex_lock(&kvm->slots_lock);
2240 		r = kvm_s390_get_cmma_bits(kvm, &args);
2241 		mutex_unlock(&kvm->slots_lock);
2242 		if (!r) {
2243 			r = copy_to_user(argp, &args, sizeof(args));
2244 			if (r)
2245 				r = -EFAULT;
2246 		}
2247 		break;
2248 	}
2249 	case KVM_S390_SET_CMMA_BITS: {
2250 		struct kvm_s390_cmma_log args;
2251 
2252 		r = -EFAULT;
2253 		if (copy_from_user(&args, argp, sizeof(args)))
2254 			break;
2255 		mutex_lock(&kvm->slots_lock);
2256 		r = kvm_s390_set_cmma_bits(kvm, &args);
2257 		mutex_unlock(&kvm->slots_lock);
2258 		break;
2259 	}
2260 	default:
2261 		r = -ENOTTY;
2262 	}
2263 
2264 	return r;
2265 }
2266 
2267 static int kvm_s390_apxa_installed(void)
2268 {
2269 	struct ap_config_info info;
2270 
2271 	if (ap_instructions_available()) {
2272 		if (ap_qci(&info) == 0)
2273 			return info.apxa;
2274 	}
2275 
2276 	return 0;
2277 }
2278 
2279 /*
2280  * The format of the crypto control block (CRYCB) is specified in the 3 low
2281  * order bits of the CRYCB designation (CRYCBD) field as follows:
2282  * Format 0: Neither the message security assist extension 3 (MSAX3) nor the
2283  *	     AP extended addressing (APXA) facility are installed.
2284  * Format 1: The APXA facility is not installed but the MSAX3 facility is.
2285  * Format 2: Both the APXA and MSAX3 facilities are installed
2286  */
2287 static void kvm_s390_set_crycb_format(struct kvm *kvm)
2288 {
2289 	kvm->arch.crypto.crycbd = (__u32)(unsigned long) kvm->arch.crypto.crycb;
2290 
2291 	/* Clear the CRYCB format bits - i.e., set format 0 by default */
2292 	kvm->arch.crypto.crycbd &= ~(CRYCB_FORMAT_MASK);
2293 
2294 	/* Check whether MSAX3 is installed */
2295 	if (!test_kvm_facility(kvm, 76))
2296 		return;
2297 
2298 	if (kvm_s390_apxa_installed())
2299 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT2;
2300 	else
2301 		kvm->arch.crypto.crycbd |= CRYCB_FORMAT1;
2302 }
2303 
2304 void kvm_arch_crypto_set_masks(struct kvm *kvm, unsigned long *apm,
2305 			       unsigned long *aqm, unsigned long *adm)
2306 {
2307 	struct kvm_s390_crypto_cb *crycb = kvm->arch.crypto.crycb;
2308 
2309 	mutex_lock(&kvm->lock);
2310 	kvm_s390_vcpu_block_all(kvm);
2311 
2312 	switch (kvm->arch.crypto.crycbd & CRYCB_FORMAT_MASK) {
2313 	case CRYCB_FORMAT2: /* APCB1 use 256 bits */
2314 		memcpy(crycb->apcb1.apm, apm, 32);
2315 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx %016lx %016lx %016lx",
2316 			 apm[0], apm[1], apm[2], apm[3]);
2317 		memcpy(crycb->apcb1.aqm, aqm, 32);
2318 		VM_EVENT(kvm, 3, "SET CRYCB: aqm %016lx %016lx %016lx %016lx",
2319 			 aqm[0], aqm[1], aqm[2], aqm[3]);
2320 		memcpy(crycb->apcb1.adm, adm, 32);
2321 		VM_EVENT(kvm, 3, "SET CRYCB: adm %016lx %016lx %016lx %016lx",
2322 			 adm[0], adm[1], adm[2], adm[3]);
2323 		break;
2324 	case CRYCB_FORMAT1:
2325 	case CRYCB_FORMAT0: /* Fall through both use APCB0 */
2326 		memcpy(crycb->apcb0.apm, apm, 8);
2327 		memcpy(crycb->apcb0.aqm, aqm, 2);
2328 		memcpy(crycb->apcb0.adm, adm, 2);
2329 		VM_EVENT(kvm, 3, "SET CRYCB: apm %016lx aqm %04x adm %04x",
2330 			 apm[0], *((unsigned short *)aqm),
2331 			 *((unsigned short *)adm));
2332 		break;
2333 	default:	/* Can not happen */
2334 		break;
2335 	}
2336 
2337 	/* recreate the shadow crycb for each vcpu */
2338 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2339 	kvm_s390_vcpu_unblock_all(kvm);
2340 	mutex_unlock(&kvm->lock);
2341 }
2342 EXPORT_SYMBOL_GPL(kvm_arch_crypto_set_masks);
2343 
2344 void kvm_arch_crypto_clear_masks(struct kvm *kvm)
2345 {
2346 	mutex_lock(&kvm->lock);
2347 	kvm_s390_vcpu_block_all(kvm);
2348 
2349 	memset(&kvm->arch.crypto.crycb->apcb0, 0,
2350 	       sizeof(kvm->arch.crypto.crycb->apcb0));
2351 	memset(&kvm->arch.crypto.crycb->apcb1, 0,
2352 	       sizeof(kvm->arch.crypto.crycb->apcb1));
2353 
2354 	VM_EVENT(kvm, 3, "%s", "CLR CRYCB:");
2355 	/* recreate the shadow crycb for each vcpu */
2356 	kvm_s390_sync_request_broadcast(kvm, KVM_REQ_VSIE_RESTART);
2357 	kvm_s390_vcpu_unblock_all(kvm);
2358 	mutex_unlock(&kvm->lock);
2359 }
2360 EXPORT_SYMBOL_GPL(kvm_arch_crypto_clear_masks);
2361 
2362 static u64 kvm_s390_get_initial_cpuid(void)
2363 {
2364 	struct cpuid cpuid;
2365 
2366 	get_cpu_id(&cpuid);
2367 	cpuid.version = 0xff;
2368 	return *((u64 *) &cpuid);
2369 }
2370 
2371 static void kvm_s390_crypto_init(struct kvm *kvm)
2372 {
2373 	kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
2374 	kvm_s390_set_crycb_format(kvm);
2375 
2376 	if (!test_kvm_facility(kvm, 76))
2377 		return;
2378 
2379 	/* Enable AES/DEA protected key functions by default */
2380 	kvm->arch.crypto.aes_kw = 1;
2381 	kvm->arch.crypto.dea_kw = 1;
2382 	get_random_bytes(kvm->arch.crypto.crycb->aes_wrapping_key_mask,
2383 			 sizeof(kvm->arch.crypto.crycb->aes_wrapping_key_mask));
2384 	get_random_bytes(kvm->arch.crypto.crycb->dea_wrapping_key_mask,
2385 			 sizeof(kvm->arch.crypto.crycb->dea_wrapping_key_mask));
2386 }
2387 
2388 static void sca_dispose(struct kvm *kvm)
2389 {
2390 	if (kvm->arch.use_esca)
2391 		free_pages_exact(kvm->arch.sca, sizeof(struct esca_block));
2392 	else
2393 		free_page((unsigned long)(kvm->arch.sca));
2394 	kvm->arch.sca = NULL;
2395 }
2396 
2397 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
2398 {
2399 	gfp_t alloc_flags = GFP_KERNEL;
2400 	int i, rc;
2401 	char debug_name[16];
2402 	static unsigned long sca_offset;
2403 
2404 	rc = -EINVAL;
2405 #ifdef CONFIG_KVM_S390_UCONTROL
2406 	if (type & ~KVM_VM_S390_UCONTROL)
2407 		goto out_err;
2408 	if ((type & KVM_VM_S390_UCONTROL) && (!capable(CAP_SYS_ADMIN)))
2409 		goto out_err;
2410 #else
2411 	if (type)
2412 		goto out_err;
2413 #endif
2414 
2415 	rc = s390_enable_sie();
2416 	if (rc)
2417 		goto out_err;
2418 
2419 	rc = -ENOMEM;
2420 
2421 	if (!sclp.has_64bscao)
2422 		alloc_flags |= GFP_DMA;
2423 	rwlock_init(&kvm->arch.sca_lock);
2424 	/* start with basic SCA */
2425 	kvm->arch.sca = (struct bsca_block *) get_zeroed_page(alloc_flags);
2426 	if (!kvm->arch.sca)
2427 		goto out_err;
2428 	mutex_lock(&kvm_lock);
2429 	sca_offset += 16;
2430 	if (sca_offset + sizeof(struct bsca_block) > PAGE_SIZE)
2431 		sca_offset = 0;
2432 	kvm->arch.sca = (struct bsca_block *)
2433 			((char *) kvm->arch.sca + sca_offset);
2434 	mutex_unlock(&kvm_lock);
2435 
2436 	sprintf(debug_name, "kvm-%u", current->pid);
2437 
2438 	kvm->arch.dbf = debug_register(debug_name, 32, 1, 7 * sizeof(long));
2439 	if (!kvm->arch.dbf)
2440 		goto out_err;
2441 
2442 	BUILD_BUG_ON(sizeof(struct sie_page2) != 4096);
2443 	kvm->arch.sie_page2 =
2444 	     (struct sie_page2 *) get_zeroed_page(GFP_KERNEL | GFP_DMA);
2445 	if (!kvm->arch.sie_page2)
2446 		goto out_err;
2447 
2448 	kvm->arch.sie_page2->kvm = kvm;
2449 	kvm->arch.model.fac_list = kvm->arch.sie_page2->fac_list;
2450 
2451 	for (i = 0; i < kvm_s390_fac_size(); i++) {
2452 		kvm->arch.model.fac_mask[i] = S390_lowcore.stfle_fac_list[i] &
2453 					      (kvm_s390_fac_base[i] |
2454 					       kvm_s390_fac_ext[i]);
2455 		kvm->arch.model.fac_list[i] = S390_lowcore.stfle_fac_list[i] &
2456 					      kvm_s390_fac_base[i];
2457 	}
2458 	kvm->arch.model.subfuncs = kvm_s390_available_subfunc;
2459 
2460 	/* we are always in czam mode - even on pre z14 machines */
2461 	set_kvm_facility(kvm->arch.model.fac_mask, 138);
2462 	set_kvm_facility(kvm->arch.model.fac_list, 138);
2463 	/* we emulate STHYI in kvm */
2464 	set_kvm_facility(kvm->arch.model.fac_mask, 74);
2465 	set_kvm_facility(kvm->arch.model.fac_list, 74);
2466 	if (MACHINE_HAS_TLB_GUEST) {
2467 		set_kvm_facility(kvm->arch.model.fac_mask, 147);
2468 		set_kvm_facility(kvm->arch.model.fac_list, 147);
2469 	}
2470 
2471 	if (css_general_characteristics.aiv && test_facility(65))
2472 		set_kvm_facility(kvm->arch.model.fac_mask, 65);
2473 
2474 	kvm->arch.model.cpuid = kvm_s390_get_initial_cpuid();
2475 	kvm->arch.model.ibc = sclp.ibc & 0x0fff;
2476 
2477 	kvm_s390_crypto_init(kvm);
2478 
2479 	mutex_init(&kvm->arch.float_int.ais_lock);
2480 	spin_lock_init(&kvm->arch.float_int.lock);
2481 	for (i = 0; i < FIRQ_LIST_COUNT; i++)
2482 		INIT_LIST_HEAD(&kvm->arch.float_int.lists[i]);
2483 	init_waitqueue_head(&kvm->arch.ipte_wq);
2484 	mutex_init(&kvm->arch.ipte_mutex);
2485 
2486 	debug_register_view(kvm->arch.dbf, &debug_sprintf_view);
2487 	VM_EVENT(kvm, 3, "vm created with type %lu", type);
2488 
2489 	if (type & KVM_VM_S390_UCONTROL) {
2490 		kvm->arch.gmap = NULL;
2491 		kvm->arch.mem_limit = KVM_S390_NO_MEM_LIMIT;
2492 	} else {
2493 		if (sclp.hamax == U64_MAX)
2494 			kvm->arch.mem_limit = TASK_SIZE_MAX;
2495 		else
2496 			kvm->arch.mem_limit = min_t(unsigned long, TASK_SIZE_MAX,
2497 						    sclp.hamax + 1);
2498 		kvm->arch.gmap = gmap_create(current->mm, kvm->arch.mem_limit - 1);
2499 		if (!kvm->arch.gmap)
2500 			goto out_err;
2501 		kvm->arch.gmap->private = kvm;
2502 		kvm->arch.gmap->pfault_enabled = 0;
2503 	}
2504 
2505 	kvm->arch.use_pfmfi = sclp.has_pfmfi;
2506 	kvm->arch.use_skf = sclp.has_skey;
2507 	spin_lock_init(&kvm->arch.start_stop_lock);
2508 	kvm_s390_vsie_init(kvm);
2509 	kvm_s390_gisa_init(kvm);
2510 	KVM_EVENT(3, "vm 0x%pK created by pid %u", kvm, current->pid);
2511 
2512 	return 0;
2513 out_err:
2514 	free_page((unsigned long)kvm->arch.sie_page2);
2515 	debug_unregister(kvm->arch.dbf);
2516 	sca_dispose(kvm);
2517 	KVM_EVENT(3, "creation of vm failed: %d", rc);
2518 	return rc;
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 	if (kvm_run->kvm_valid_regs & ~KVM_SYNC_S390_VALID_FIELDS ||
4004 	    kvm_run->kvm_dirty_regs & ~KVM_SYNC_S390_VALID_FIELDS)
4005 		return -EINVAL;
4006 
4007 	vcpu_load(vcpu);
4008 
4009 	if (guestdbg_exit_pending(vcpu)) {
4010 		kvm_s390_prepare_debug_exit(vcpu);
4011 		rc = 0;
4012 		goto out;
4013 	}
4014 
4015 	kvm_sigset_activate(vcpu);
4016 
4017 	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm)) {
4018 		kvm_s390_vcpu_start(vcpu);
4019 	} else if (is_vcpu_stopped(vcpu)) {
4020 		pr_err_ratelimited("can't run stopped vcpu %d\n",
4021 				   vcpu->vcpu_id);
4022 		rc = -EINVAL;
4023 		goto out;
4024 	}
4025 
4026 	sync_regs(vcpu, kvm_run);
4027 	enable_cpu_timer_accounting(vcpu);
4028 
4029 	might_fault();
4030 	rc = __vcpu_run(vcpu);
4031 
4032 	if (signal_pending(current) && !rc) {
4033 		kvm_run->exit_reason = KVM_EXIT_INTR;
4034 		rc = -EINTR;
4035 	}
4036 
4037 	if (guestdbg_exit_pending(vcpu) && !rc)  {
4038 		kvm_s390_prepare_debug_exit(vcpu);
4039 		rc = 0;
4040 	}
4041 
4042 	if (rc == -EREMOTE) {
4043 		/* userspace support is needed, kvm_run has been prepared */
4044 		rc = 0;
4045 	}
4046 
4047 	disable_cpu_timer_accounting(vcpu);
4048 	store_regs(vcpu, kvm_run);
4049 
4050 	kvm_sigset_deactivate(vcpu);
4051 
4052 	vcpu->stat.exit_userspace++;
4053 out:
4054 	vcpu_put(vcpu);
4055 	return rc;
4056 }
4057 
4058 /*
4059  * store status at address
4060  * we use have two special cases:
4061  * KVM_S390_STORE_STATUS_NOADDR: -> 0x1200 on 64 bit
4062  * KVM_S390_STORE_STATUS_PREFIXED: -> prefix
4063  */
4064 int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long gpa)
4065 {
4066 	unsigned char archmode = 1;
4067 	freg_t fprs[NUM_FPRS];
4068 	unsigned int px;
4069 	u64 clkcomp, cputm;
4070 	int rc;
4071 
4072 	px = kvm_s390_get_prefix(vcpu);
4073 	if (gpa == KVM_S390_STORE_STATUS_NOADDR) {
4074 		if (write_guest_abs(vcpu, 163, &archmode, 1))
4075 			return -EFAULT;
4076 		gpa = 0;
4077 	} else if (gpa == KVM_S390_STORE_STATUS_PREFIXED) {
4078 		if (write_guest_real(vcpu, 163, &archmode, 1))
4079 			return -EFAULT;
4080 		gpa = px;
4081 	} else
4082 		gpa -= __LC_FPREGS_SAVE_AREA;
4083 
4084 	/* manually convert vector registers if necessary */
4085 	if (MACHINE_HAS_VX) {
4086 		convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
4087 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4088 				     fprs, 128);
4089 	} else {
4090 		rc = write_guest_abs(vcpu, gpa + __LC_FPREGS_SAVE_AREA,
4091 				     vcpu->run->s.regs.fprs, 128);
4092 	}
4093 	rc |= write_guest_abs(vcpu, gpa + __LC_GPREGS_SAVE_AREA,
4094 			      vcpu->run->s.regs.gprs, 128);
4095 	rc |= write_guest_abs(vcpu, gpa + __LC_PSW_SAVE_AREA,
4096 			      &vcpu->arch.sie_block->gpsw, 16);
4097 	rc |= write_guest_abs(vcpu, gpa + __LC_PREFIX_SAVE_AREA,
4098 			      &px, 4);
4099 	rc |= write_guest_abs(vcpu, gpa + __LC_FP_CREG_SAVE_AREA,
4100 			      &vcpu->run->s.regs.fpc, 4);
4101 	rc |= write_guest_abs(vcpu, gpa + __LC_TOD_PROGREG_SAVE_AREA,
4102 			      &vcpu->arch.sie_block->todpr, 4);
4103 	cputm = kvm_s390_get_cpu_timer(vcpu);
4104 	rc |= write_guest_abs(vcpu, gpa + __LC_CPU_TIMER_SAVE_AREA,
4105 			      &cputm, 8);
4106 	clkcomp = vcpu->arch.sie_block->ckc >> 8;
4107 	rc |= write_guest_abs(vcpu, gpa + __LC_CLOCK_COMP_SAVE_AREA,
4108 			      &clkcomp, 8);
4109 	rc |= write_guest_abs(vcpu, gpa + __LC_AREGS_SAVE_AREA,
4110 			      &vcpu->run->s.regs.acrs, 64);
4111 	rc |= write_guest_abs(vcpu, gpa + __LC_CREGS_SAVE_AREA,
4112 			      &vcpu->arch.sie_block->gcr, 128);
4113 	return rc ? -EFAULT : 0;
4114 }
4115 
4116 int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
4117 {
4118 	/*
4119 	 * The guest FPRS and ACRS are in the host FPRS/ACRS due to the lazy
4120 	 * switch in the run ioctl. Let's update our copies before we save
4121 	 * it into the save area
4122 	 */
4123 	save_fpu_regs();
4124 	vcpu->run->s.regs.fpc = current->thread.fpu.fpc;
4125 	save_access_regs(vcpu->run->s.regs.acrs);
4126 
4127 	return kvm_s390_store_status_unloaded(vcpu, addr);
4128 }
4129 
4130 static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4131 {
4132 	kvm_check_request(KVM_REQ_ENABLE_IBS, vcpu);
4133 	kvm_s390_sync_request(KVM_REQ_DISABLE_IBS, vcpu);
4134 }
4135 
4136 static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
4137 {
4138 	unsigned int i;
4139 	struct kvm_vcpu *vcpu;
4140 
4141 	kvm_for_each_vcpu(i, vcpu, kvm) {
4142 		__disable_ibs_on_vcpu(vcpu);
4143 	}
4144 }
4145 
4146 static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
4147 {
4148 	if (!sclp.has_ibs)
4149 		return;
4150 	kvm_check_request(KVM_REQ_DISABLE_IBS, vcpu);
4151 	kvm_s390_sync_request(KVM_REQ_ENABLE_IBS, vcpu);
4152 }
4153 
4154 void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
4155 {
4156 	int i, online_vcpus, started_vcpus = 0;
4157 
4158 	if (!is_vcpu_stopped(vcpu))
4159 		return;
4160 
4161 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
4162 	/* Only one cpu at a time may enter/leave the STOPPED state. */
4163 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
4164 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4165 
4166 	for (i = 0; i < online_vcpus; i++) {
4167 		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
4168 			started_vcpus++;
4169 	}
4170 
4171 	if (started_vcpus == 0) {
4172 		/* we're the only active VCPU -> speed it up */
4173 		__enable_ibs_on_vcpu(vcpu);
4174 	} else if (started_vcpus == 1) {
4175 		/*
4176 		 * As we are starting a second VCPU, we have to disable
4177 		 * the IBS facility on all VCPUs to remove potentially
4178 		 * oustanding ENABLE requests.
4179 		 */
4180 		__disable_ibs_on_all_vcpus(vcpu->kvm);
4181 	}
4182 
4183 	kvm_s390_clear_cpuflags(vcpu, CPUSTAT_STOPPED);
4184 	/*
4185 	 * Another VCPU might have used IBS while we were offline.
4186 	 * Let's play safe and flush the VCPU at startup.
4187 	 */
4188 	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
4189 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4190 	return;
4191 }
4192 
4193 void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
4194 {
4195 	int i, online_vcpus, started_vcpus = 0;
4196 	struct kvm_vcpu *started_vcpu = NULL;
4197 
4198 	if (is_vcpu_stopped(vcpu))
4199 		return;
4200 
4201 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
4202 	/* Only one cpu at a time may enter/leave the STOPPED state. */
4203 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
4204 	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
4205 
4206 	/* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
4207 	kvm_s390_clear_stop_irq(vcpu);
4208 
4209 	kvm_s390_set_cpuflags(vcpu, CPUSTAT_STOPPED);
4210 	__disable_ibs_on_vcpu(vcpu);
4211 
4212 	for (i = 0; i < online_vcpus; i++) {
4213 		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
4214 			started_vcpus++;
4215 			started_vcpu = vcpu->kvm->vcpus[i];
4216 		}
4217 	}
4218 
4219 	if (started_vcpus == 1) {
4220 		/*
4221 		 * As we only have one VCPU left, we want to enable the
4222 		 * IBS facility for that VCPU to speed it up.
4223 		 */
4224 		__enable_ibs_on_vcpu(started_vcpu);
4225 	}
4226 
4227 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
4228 	return;
4229 }
4230 
4231 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
4232 				     struct kvm_enable_cap *cap)
4233 {
4234 	int r;
4235 
4236 	if (cap->flags)
4237 		return -EINVAL;
4238 
4239 	switch (cap->cap) {
4240 	case KVM_CAP_S390_CSS_SUPPORT:
4241 		if (!vcpu->kvm->arch.css_support) {
4242 			vcpu->kvm->arch.css_support = 1;
4243 			VM_EVENT(vcpu->kvm, 3, "%s", "ENABLE: CSS support");
4244 			trace_kvm_s390_enable_css(vcpu->kvm);
4245 		}
4246 		r = 0;
4247 		break;
4248 	default:
4249 		r = -EINVAL;
4250 		break;
4251 	}
4252 	return r;
4253 }
4254 
4255 static long kvm_s390_guest_mem_op(struct kvm_vcpu *vcpu,
4256 				  struct kvm_s390_mem_op *mop)
4257 {
4258 	void __user *uaddr = (void __user *)mop->buf;
4259 	void *tmpbuf = NULL;
4260 	int r, srcu_idx;
4261 	const u64 supported_flags = KVM_S390_MEMOP_F_INJECT_EXCEPTION
4262 				    | KVM_S390_MEMOP_F_CHECK_ONLY;
4263 
4264 	if (mop->flags & ~supported_flags || mop->ar >= NUM_ACRS || !mop->size)
4265 		return -EINVAL;
4266 
4267 	if (mop->size > MEM_OP_MAX_SIZE)
4268 		return -E2BIG;
4269 
4270 	if (!(mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY)) {
4271 		tmpbuf = vmalloc(mop->size);
4272 		if (!tmpbuf)
4273 			return -ENOMEM;
4274 	}
4275 
4276 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
4277 
4278 	switch (mop->op) {
4279 	case KVM_S390_MEMOP_LOGICAL_READ:
4280 		if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4281 			r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4282 					    mop->size, GACC_FETCH);
4283 			break;
4284 		}
4285 		r = read_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4286 		if (r == 0) {
4287 			if (copy_to_user(uaddr, tmpbuf, mop->size))
4288 				r = -EFAULT;
4289 		}
4290 		break;
4291 	case KVM_S390_MEMOP_LOGICAL_WRITE:
4292 		if (mop->flags & KVM_S390_MEMOP_F_CHECK_ONLY) {
4293 			r = check_gva_range(vcpu, mop->gaddr, mop->ar,
4294 					    mop->size, GACC_STORE);
4295 			break;
4296 		}
4297 		if (copy_from_user(tmpbuf, uaddr, mop->size)) {
4298 			r = -EFAULT;
4299 			break;
4300 		}
4301 		r = write_guest(vcpu, mop->gaddr, mop->ar, tmpbuf, mop->size);
4302 		break;
4303 	default:
4304 		r = -EINVAL;
4305 	}
4306 
4307 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
4308 
4309 	if (r > 0 && (mop->flags & KVM_S390_MEMOP_F_INJECT_EXCEPTION) != 0)
4310 		kvm_s390_inject_prog_irq(vcpu, &vcpu->arch.pgm);
4311 
4312 	vfree(tmpbuf);
4313 	return r;
4314 }
4315 
4316 long kvm_arch_vcpu_async_ioctl(struct file *filp,
4317 			       unsigned int ioctl, unsigned long arg)
4318 {
4319 	struct kvm_vcpu *vcpu = filp->private_data;
4320 	void __user *argp = (void __user *)arg;
4321 
4322 	switch (ioctl) {
4323 	case KVM_S390_IRQ: {
4324 		struct kvm_s390_irq s390irq;
4325 
4326 		if (copy_from_user(&s390irq, argp, sizeof(s390irq)))
4327 			return -EFAULT;
4328 		return kvm_s390_inject_vcpu(vcpu, &s390irq);
4329 	}
4330 	case KVM_S390_INTERRUPT: {
4331 		struct kvm_s390_interrupt s390int;
4332 		struct kvm_s390_irq s390irq = {};
4333 
4334 		if (copy_from_user(&s390int, argp, sizeof(s390int)))
4335 			return -EFAULT;
4336 		if (s390int_to_s390irq(&s390int, &s390irq))
4337 			return -EINVAL;
4338 		return kvm_s390_inject_vcpu(vcpu, &s390irq);
4339 	}
4340 	}
4341 	return -ENOIOCTLCMD;
4342 }
4343 
4344 long kvm_arch_vcpu_ioctl(struct file *filp,
4345 			 unsigned int ioctl, unsigned long arg)
4346 {
4347 	struct kvm_vcpu *vcpu = filp->private_data;
4348 	void __user *argp = (void __user *)arg;
4349 	int idx;
4350 	long r;
4351 
4352 	vcpu_load(vcpu);
4353 
4354 	switch (ioctl) {
4355 	case KVM_S390_STORE_STATUS:
4356 		idx = srcu_read_lock(&vcpu->kvm->srcu);
4357 		r = kvm_s390_vcpu_store_status(vcpu, arg);
4358 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
4359 		break;
4360 	case KVM_S390_SET_INITIAL_PSW: {
4361 		psw_t psw;
4362 
4363 		r = -EFAULT;
4364 		if (copy_from_user(&psw, argp, sizeof(psw)))
4365 			break;
4366 		r = kvm_arch_vcpu_ioctl_set_initial_psw(vcpu, psw);
4367 		break;
4368 	}
4369 	case KVM_S390_INITIAL_RESET:
4370 		r = kvm_arch_vcpu_ioctl_initial_reset(vcpu);
4371 		break;
4372 	case KVM_SET_ONE_REG:
4373 	case KVM_GET_ONE_REG: {
4374 		struct kvm_one_reg reg;
4375 		r = -EFAULT;
4376 		if (copy_from_user(&reg, argp, sizeof(reg)))
4377 			break;
4378 		if (ioctl == KVM_SET_ONE_REG)
4379 			r = kvm_arch_vcpu_ioctl_set_one_reg(vcpu, &reg);
4380 		else
4381 			r = kvm_arch_vcpu_ioctl_get_one_reg(vcpu, &reg);
4382 		break;
4383 	}
4384 #ifdef CONFIG_KVM_S390_UCONTROL
4385 	case KVM_S390_UCAS_MAP: {
4386 		struct kvm_s390_ucas_mapping ucasmap;
4387 
4388 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4389 			r = -EFAULT;
4390 			break;
4391 		}
4392 
4393 		if (!kvm_is_ucontrol(vcpu->kvm)) {
4394 			r = -EINVAL;
4395 			break;
4396 		}
4397 
4398 		r = gmap_map_segment(vcpu->arch.gmap, ucasmap.user_addr,
4399 				     ucasmap.vcpu_addr, ucasmap.length);
4400 		break;
4401 	}
4402 	case KVM_S390_UCAS_UNMAP: {
4403 		struct kvm_s390_ucas_mapping ucasmap;
4404 
4405 		if (copy_from_user(&ucasmap, argp, sizeof(ucasmap))) {
4406 			r = -EFAULT;
4407 			break;
4408 		}
4409 
4410 		if (!kvm_is_ucontrol(vcpu->kvm)) {
4411 			r = -EINVAL;
4412 			break;
4413 		}
4414 
4415 		r = gmap_unmap_segment(vcpu->arch.gmap, ucasmap.vcpu_addr,
4416 			ucasmap.length);
4417 		break;
4418 	}
4419 #endif
4420 	case KVM_S390_VCPU_FAULT: {
4421 		r = gmap_fault(vcpu->arch.gmap, arg, 0);
4422 		break;
4423 	}
4424 	case KVM_ENABLE_CAP:
4425 	{
4426 		struct kvm_enable_cap cap;
4427 		r = -EFAULT;
4428 		if (copy_from_user(&cap, argp, sizeof(cap)))
4429 			break;
4430 		r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
4431 		break;
4432 	}
4433 	case KVM_S390_MEM_OP: {
4434 		struct kvm_s390_mem_op mem_op;
4435 
4436 		if (copy_from_user(&mem_op, argp, sizeof(mem_op)) == 0)
4437 			r = kvm_s390_guest_mem_op(vcpu, &mem_op);
4438 		else
4439 			r = -EFAULT;
4440 		break;
4441 	}
4442 	case KVM_S390_SET_IRQ_STATE: {
4443 		struct kvm_s390_irq_state irq_state;
4444 
4445 		r = -EFAULT;
4446 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4447 			break;
4448 		if (irq_state.len > VCPU_IRQS_MAX_BUF ||
4449 		    irq_state.len == 0 ||
4450 		    irq_state.len % sizeof(struct kvm_s390_irq) > 0) {
4451 			r = -EINVAL;
4452 			break;
4453 		}
4454 		/* do not use irq_state.flags, it will break old QEMUs */
4455 		r = kvm_s390_set_irq_state(vcpu,
4456 					   (void __user *) irq_state.buf,
4457 					   irq_state.len);
4458 		break;
4459 	}
4460 	case KVM_S390_GET_IRQ_STATE: {
4461 		struct kvm_s390_irq_state irq_state;
4462 
4463 		r = -EFAULT;
4464 		if (copy_from_user(&irq_state, argp, sizeof(irq_state)))
4465 			break;
4466 		if (irq_state.len == 0) {
4467 			r = -EINVAL;
4468 			break;
4469 		}
4470 		/* do not use irq_state.flags, it will break old QEMUs */
4471 		r = kvm_s390_get_irq_state(vcpu,
4472 					   (__u8 __user *)  irq_state.buf,
4473 					   irq_state.len);
4474 		break;
4475 	}
4476 	default:
4477 		r = -ENOTTY;
4478 	}
4479 
4480 	vcpu_put(vcpu);
4481 	return r;
4482 }
4483 
4484 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
4485 {
4486 #ifdef CONFIG_KVM_S390_UCONTROL
4487 	if ((vmf->pgoff == KVM_S390_SIE_PAGE_OFFSET)
4488 		 && (kvm_is_ucontrol(vcpu->kvm))) {
4489 		vmf->page = virt_to_page(vcpu->arch.sie_block);
4490 		get_page(vmf->page);
4491 		return 0;
4492 	}
4493 #endif
4494 	return VM_FAULT_SIGBUS;
4495 }
4496 
4497 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
4498 			    unsigned long npages)
4499 {
4500 	return 0;
4501 }
4502 
4503 /* Section: memory related */
4504 int kvm_arch_prepare_memory_region(struct kvm *kvm,
4505 				   struct kvm_memory_slot *memslot,
4506 				   const struct kvm_userspace_memory_region *mem,
4507 				   enum kvm_mr_change change)
4508 {
4509 	/* A few sanity checks. We can have memory slots which have to be
4510 	   located/ended at a segment boundary (1MB). The memory in userland is
4511 	   ok to be fragmented into various different vmas. It is okay to mmap()
4512 	   and munmap() stuff in this slot after doing this call at any time */
4513 
4514 	if (mem->userspace_addr & 0xffffful)
4515 		return -EINVAL;
4516 
4517 	if (mem->memory_size & 0xffffful)
4518 		return -EINVAL;
4519 
4520 	if (mem->guest_phys_addr + mem->memory_size > kvm->arch.mem_limit)
4521 		return -EINVAL;
4522 
4523 	return 0;
4524 }
4525 
4526 void kvm_arch_commit_memory_region(struct kvm *kvm,
4527 				const struct kvm_userspace_memory_region *mem,
4528 				const struct kvm_memory_slot *old,
4529 				const struct kvm_memory_slot *new,
4530 				enum kvm_mr_change change)
4531 {
4532 	int rc = 0;
4533 
4534 	switch (change) {
4535 	case KVM_MR_DELETE:
4536 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
4537 					old->npages * PAGE_SIZE);
4538 		break;
4539 	case KVM_MR_MOVE:
4540 		rc = gmap_unmap_segment(kvm->arch.gmap, old->base_gfn * PAGE_SIZE,
4541 					old->npages * PAGE_SIZE);
4542 		if (rc)
4543 			break;
4544 		/* FALLTHROUGH */
4545 	case KVM_MR_CREATE:
4546 		rc = gmap_map_segment(kvm->arch.gmap, mem->userspace_addr,
4547 				      mem->guest_phys_addr, mem->memory_size);
4548 		break;
4549 	case KVM_MR_FLAGS_ONLY:
4550 		break;
4551 	default:
4552 		WARN(1, "Unknown KVM MR CHANGE: %d\n", change);
4553 	}
4554 	if (rc)
4555 		pr_warn("failed to commit memory region\n");
4556 	return;
4557 }
4558 
4559 static inline unsigned long nonhyp_mask(int i)
4560 {
4561 	unsigned int nonhyp_fai = (sclp.hmfai << i * 2) >> 30;
4562 
4563 	return 0x0000ffffffffffffUL >> (nonhyp_fai << 4);
4564 }
4565 
4566 void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu)
4567 {
4568 	vcpu->valid_wakeup = false;
4569 }
4570 
4571 static int __init kvm_s390_init(void)
4572 {
4573 	int i;
4574 
4575 	if (!sclp.has_sief2) {
4576 		pr_info("SIE is not available\n");
4577 		return -ENODEV;
4578 	}
4579 
4580 	if (nested && hpage) {
4581 		pr_info("A KVM host that supports nesting cannot back its KVM guests with huge pages\n");
4582 		return -EINVAL;
4583 	}
4584 
4585 	for (i = 0; i < 16; i++)
4586 		kvm_s390_fac_base[i] |=
4587 			S390_lowcore.stfle_fac_list[i] & nonhyp_mask(i);
4588 
4589 	return kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
4590 }
4591 
4592 static void __exit kvm_s390_exit(void)
4593 {
4594 	kvm_exit();
4595 }
4596 
4597 module_init(kvm_s390_init);
4598 module_exit(kvm_s390_exit);
4599 
4600 /*
4601  * Enable autoloading of the kvm module.
4602  * Note that we add the module alias here instead of virt/kvm/kvm_main.c
4603  * since x86 takes a different approach.
4604  */
4605 #include <linux/miscdevice.h>
4606 MODULE_ALIAS_MISCDEV(KVM_MINOR);
4607 MODULE_ALIAS("devname:kvm");
4608