Loading...
1/*
2 * guest access functions
3 *
4 * Copyright IBM Corp. 2014
5 *
6 */
7
8#include <linux/vmalloc.h>
9#include <linux/err.h>
10#include <asm/pgtable.h>
11#include "kvm-s390.h"
12#include "gaccess.h"
13#include <asm/switch_to.h>
14
15union asce {
16 unsigned long val;
17 struct {
18 unsigned long origin : 52; /* Region- or Segment-Table Origin */
19 unsigned long : 2;
20 unsigned long g : 1; /* Subspace Group Control */
21 unsigned long p : 1; /* Private Space Control */
22 unsigned long s : 1; /* Storage-Alteration-Event Control */
23 unsigned long x : 1; /* Space-Switch-Event Control */
24 unsigned long r : 1; /* Real-Space Control */
25 unsigned long : 1;
26 unsigned long dt : 2; /* Designation-Type Control */
27 unsigned long tl : 2; /* Region- or Segment-Table Length */
28 };
29};
30
31enum {
32 ASCE_TYPE_SEGMENT = 0,
33 ASCE_TYPE_REGION3 = 1,
34 ASCE_TYPE_REGION2 = 2,
35 ASCE_TYPE_REGION1 = 3
36};
37
38union region1_table_entry {
39 unsigned long val;
40 struct {
41 unsigned long rto: 52;/* Region-Table Origin */
42 unsigned long : 2;
43 unsigned long p : 1; /* DAT-Protection Bit */
44 unsigned long : 1;
45 unsigned long tf : 2; /* Region-Second-Table Offset */
46 unsigned long i : 1; /* Region-Invalid Bit */
47 unsigned long : 1;
48 unsigned long tt : 2; /* Table-Type Bits */
49 unsigned long tl : 2; /* Region-Second-Table Length */
50 };
51};
52
53union region2_table_entry {
54 unsigned long val;
55 struct {
56 unsigned long rto: 52;/* Region-Table Origin */
57 unsigned long : 2;
58 unsigned long p : 1; /* DAT-Protection Bit */
59 unsigned long : 1;
60 unsigned long tf : 2; /* Region-Third-Table Offset */
61 unsigned long i : 1; /* Region-Invalid Bit */
62 unsigned long : 1;
63 unsigned long tt : 2; /* Table-Type Bits */
64 unsigned long tl : 2; /* Region-Third-Table Length */
65 };
66};
67
68struct region3_table_entry_fc0 {
69 unsigned long sto: 52;/* Segment-Table Origin */
70 unsigned long : 1;
71 unsigned long fc : 1; /* Format-Control */
72 unsigned long p : 1; /* DAT-Protection Bit */
73 unsigned long : 1;
74 unsigned long tf : 2; /* Segment-Table Offset */
75 unsigned long i : 1; /* Region-Invalid Bit */
76 unsigned long cr : 1; /* Common-Region Bit */
77 unsigned long tt : 2; /* Table-Type Bits */
78 unsigned long tl : 2; /* Segment-Table Length */
79};
80
81struct region3_table_entry_fc1 {
82 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
83 unsigned long : 14;
84 unsigned long av : 1; /* ACCF-Validity Control */
85 unsigned long acc: 4; /* Access-Control Bits */
86 unsigned long f : 1; /* Fetch-Protection Bit */
87 unsigned long fc : 1; /* Format-Control */
88 unsigned long p : 1; /* DAT-Protection Bit */
89 unsigned long co : 1; /* Change-Recording Override */
90 unsigned long : 2;
91 unsigned long i : 1; /* Region-Invalid Bit */
92 unsigned long cr : 1; /* Common-Region Bit */
93 unsigned long tt : 2; /* Table-Type Bits */
94 unsigned long : 2;
95};
96
97union region3_table_entry {
98 unsigned long val;
99 struct region3_table_entry_fc0 fc0;
100 struct region3_table_entry_fc1 fc1;
101 struct {
102 unsigned long : 53;
103 unsigned long fc : 1; /* Format-Control */
104 unsigned long : 4;
105 unsigned long i : 1; /* Region-Invalid Bit */
106 unsigned long cr : 1; /* Common-Region Bit */
107 unsigned long tt : 2; /* Table-Type Bits */
108 unsigned long : 2;
109 };
110};
111
112struct segment_entry_fc0 {
113 unsigned long pto: 53;/* Page-Table Origin */
114 unsigned long fc : 1; /* Format-Control */
115 unsigned long p : 1; /* DAT-Protection Bit */
116 unsigned long : 3;
117 unsigned long i : 1; /* Segment-Invalid Bit */
118 unsigned long cs : 1; /* Common-Segment Bit */
119 unsigned long tt : 2; /* Table-Type Bits */
120 unsigned long : 2;
121};
122
123struct segment_entry_fc1 {
124 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
125 unsigned long : 3;
126 unsigned long av : 1; /* ACCF-Validity Control */
127 unsigned long acc: 4; /* Access-Control Bits */
128 unsigned long f : 1; /* Fetch-Protection Bit */
129 unsigned long fc : 1; /* Format-Control */
130 unsigned long p : 1; /* DAT-Protection Bit */
131 unsigned long co : 1; /* Change-Recording Override */
132 unsigned long : 2;
133 unsigned long i : 1; /* Segment-Invalid Bit */
134 unsigned long cs : 1; /* Common-Segment Bit */
135 unsigned long tt : 2; /* Table-Type Bits */
136 unsigned long : 2;
137};
138
139union segment_table_entry {
140 unsigned long val;
141 struct segment_entry_fc0 fc0;
142 struct segment_entry_fc1 fc1;
143 struct {
144 unsigned long : 53;
145 unsigned long fc : 1; /* Format-Control */
146 unsigned long : 4;
147 unsigned long i : 1; /* Segment-Invalid Bit */
148 unsigned long cs : 1; /* Common-Segment Bit */
149 unsigned long tt : 2; /* Table-Type Bits */
150 unsigned long : 2;
151 };
152};
153
154enum {
155 TABLE_TYPE_SEGMENT = 0,
156 TABLE_TYPE_REGION3 = 1,
157 TABLE_TYPE_REGION2 = 2,
158 TABLE_TYPE_REGION1 = 3
159};
160
161union page_table_entry {
162 unsigned long val;
163 struct {
164 unsigned long pfra : 52; /* Page-Frame Real Address */
165 unsigned long z : 1; /* Zero Bit */
166 unsigned long i : 1; /* Page-Invalid Bit */
167 unsigned long p : 1; /* DAT-Protection Bit */
168 unsigned long co : 1; /* Change-Recording Override */
169 unsigned long : 8;
170 };
171};
172
173/*
174 * vaddress union in order to easily decode a virtual address into its
175 * region first index, region second index etc. parts.
176 */
177union vaddress {
178 unsigned long addr;
179 struct {
180 unsigned long rfx : 11;
181 unsigned long rsx : 11;
182 unsigned long rtx : 11;
183 unsigned long sx : 11;
184 unsigned long px : 8;
185 unsigned long bx : 12;
186 };
187 struct {
188 unsigned long rfx01 : 2;
189 unsigned long : 9;
190 unsigned long rsx01 : 2;
191 unsigned long : 9;
192 unsigned long rtx01 : 2;
193 unsigned long : 9;
194 unsigned long sx01 : 2;
195 unsigned long : 29;
196 };
197};
198
199/*
200 * raddress union which will contain the result (real or absolute address)
201 * after a page table walk. The rfaa, sfaa and pfra members are used to
202 * simply assign them the value of a region, segment or page table entry.
203 */
204union raddress {
205 unsigned long addr;
206 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
207 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
208 unsigned long pfra : 52; /* Page-Frame Real Address */
209};
210
211union alet {
212 u32 val;
213 struct {
214 u32 reserved : 7;
215 u32 p : 1;
216 u32 alesn : 8;
217 u32 alen : 16;
218 };
219};
220
221union ald {
222 u32 val;
223 struct {
224 u32 : 1;
225 u32 alo : 24;
226 u32 all : 7;
227 };
228};
229
230struct ale {
231 unsigned long i : 1; /* ALEN-Invalid Bit */
232 unsigned long : 5;
233 unsigned long fo : 1; /* Fetch-Only Bit */
234 unsigned long p : 1; /* Private Bit */
235 unsigned long alesn : 8; /* Access-List-Entry Sequence Number */
236 unsigned long aleax : 16; /* Access-List-Entry Authorization Index */
237 unsigned long : 32;
238 unsigned long : 1;
239 unsigned long asteo : 25; /* ASN-Second-Table-Entry Origin */
240 unsigned long : 6;
241 unsigned long astesn : 32; /* ASTE Sequence Number */
242} __packed;
243
244struct aste {
245 unsigned long i : 1; /* ASX-Invalid Bit */
246 unsigned long ato : 29; /* Authority-Table Origin */
247 unsigned long : 1;
248 unsigned long b : 1; /* Base-Space Bit */
249 unsigned long ax : 16; /* Authorization Index */
250 unsigned long atl : 12; /* Authority-Table Length */
251 unsigned long : 2;
252 unsigned long ca : 1; /* Controlled-ASN Bit */
253 unsigned long ra : 1; /* Reusable-ASN Bit */
254 unsigned long asce : 64; /* Address-Space-Control Element */
255 unsigned long ald : 32;
256 unsigned long astesn : 32;
257 /* .. more fields there */
258} __packed;
259
260int ipte_lock_held(struct kvm_vcpu *vcpu)
261{
262 if (vcpu->arch.sie_block->eca & 1) {
263 int rc;
264
265 read_lock(&vcpu->kvm->arch.sca_lock);
266 rc = kvm_s390_get_ipte_control(vcpu->kvm)->kh != 0;
267 read_unlock(&vcpu->kvm->arch.sca_lock);
268 return rc;
269 }
270 return vcpu->kvm->arch.ipte_lock_count != 0;
271}
272
273static void ipte_lock_simple(struct kvm_vcpu *vcpu)
274{
275 union ipte_control old, new, *ic;
276
277 mutex_lock(&vcpu->kvm->arch.ipte_mutex);
278 vcpu->kvm->arch.ipte_lock_count++;
279 if (vcpu->kvm->arch.ipte_lock_count > 1)
280 goto out;
281retry:
282 read_lock(&vcpu->kvm->arch.sca_lock);
283 ic = kvm_s390_get_ipte_control(vcpu->kvm);
284 do {
285 old = READ_ONCE(*ic);
286 if (old.k) {
287 read_unlock(&vcpu->kvm->arch.sca_lock);
288 cond_resched();
289 goto retry;
290 }
291 new = old;
292 new.k = 1;
293 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
294 read_unlock(&vcpu->kvm->arch.sca_lock);
295out:
296 mutex_unlock(&vcpu->kvm->arch.ipte_mutex);
297}
298
299static void ipte_unlock_simple(struct kvm_vcpu *vcpu)
300{
301 union ipte_control old, new, *ic;
302
303 mutex_lock(&vcpu->kvm->arch.ipte_mutex);
304 vcpu->kvm->arch.ipte_lock_count--;
305 if (vcpu->kvm->arch.ipte_lock_count)
306 goto out;
307 read_lock(&vcpu->kvm->arch.sca_lock);
308 ic = kvm_s390_get_ipte_control(vcpu->kvm);
309 do {
310 old = READ_ONCE(*ic);
311 new = old;
312 new.k = 0;
313 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
314 read_unlock(&vcpu->kvm->arch.sca_lock);
315 wake_up(&vcpu->kvm->arch.ipte_wq);
316out:
317 mutex_unlock(&vcpu->kvm->arch.ipte_mutex);
318}
319
320static void ipte_lock_siif(struct kvm_vcpu *vcpu)
321{
322 union ipte_control old, new, *ic;
323
324retry:
325 read_lock(&vcpu->kvm->arch.sca_lock);
326 ic = kvm_s390_get_ipte_control(vcpu->kvm);
327 do {
328 old = READ_ONCE(*ic);
329 if (old.kg) {
330 read_unlock(&vcpu->kvm->arch.sca_lock);
331 cond_resched();
332 goto retry;
333 }
334 new = old;
335 new.k = 1;
336 new.kh++;
337 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
338 read_unlock(&vcpu->kvm->arch.sca_lock);
339}
340
341static void ipte_unlock_siif(struct kvm_vcpu *vcpu)
342{
343 union ipte_control old, new, *ic;
344
345 read_lock(&vcpu->kvm->arch.sca_lock);
346 ic = kvm_s390_get_ipte_control(vcpu->kvm);
347 do {
348 old = READ_ONCE(*ic);
349 new = old;
350 new.kh--;
351 if (!new.kh)
352 new.k = 0;
353 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
354 read_unlock(&vcpu->kvm->arch.sca_lock);
355 if (!new.kh)
356 wake_up(&vcpu->kvm->arch.ipte_wq);
357}
358
359void ipte_lock(struct kvm_vcpu *vcpu)
360{
361 if (vcpu->arch.sie_block->eca & 1)
362 ipte_lock_siif(vcpu);
363 else
364 ipte_lock_simple(vcpu);
365}
366
367void ipte_unlock(struct kvm_vcpu *vcpu)
368{
369 if (vcpu->arch.sie_block->eca & 1)
370 ipte_unlock_siif(vcpu);
371 else
372 ipte_unlock_simple(vcpu);
373}
374
375static int ar_translation(struct kvm_vcpu *vcpu, union asce *asce, ar_t ar,
376 enum gacc_mode mode)
377{
378 union alet alet;
379 struct ale ale;
380 struct aste aste;
381 unsigned long ald_addr, authority_table_addr;
382 union ald ald;
383 int eax, rc;
384 u8 authority_table;
385
386 if (ar >= NUM_ACRS)
387 return -EINVAL;
388
389 save_access_regs(vcpu->run->s.regs.acrs);
390 alet.val = vcpu->run->s.regs.acrs[ar];
391
392 if (ar == 0 || alet.val == 0) {
393 asce->val = vcpu->arch.sie_block->gcr[1];
394 return 0;
395 } else if (alet.val == 1) {
396 asce->val = vcpu->arch.sie_block->gcr[7];
397 return 0;
398 }
399
400 if (alet.reserved)
401 return PGM_ALET_SPECIFICATION;
402
403 if (alet.p)
404 ald_addr = vcpu->arch.sie_block->gcr[5];
405 else
406 ald_addr = vcpu->arch.sie_block->gcr[2];
407 ald_addr &= 0x7fffffc0;
408
409 rc = read_guest_real(vcpu, ald_addr + 16, &ald.val, sizeof(union ald));
410 if (rc)
411 return rc;
412
413 if (alet.alen / 8 > ald.all)
414 return PGM_ALEN_TRANSLATION;
415
416 if (0x7fffffff - ald.alo * 128 < alet.alen * 16)
417 return PGM_ADDRESSING;
418
419 rc = read_guest_real(vcpu, ald.alo * 128 + alet.alen * 16, &ale,
420 sizeof(struct ale));
421 if (rc)
422 return rc;
423
424 if (ale.i == 1)
425 return PGM_ALEN_TRANSLATION;
426 if (ale.alesn != alet.alesn)
427 return PGM_ALE_SEQUENCE;
428
429 rc = read_guest_real(vcpu, ale.asteo * 64, &aste, sizeof(struct aste));
430 if (rc)
431 return rc;
432
433 if (aste.i)
434 return PGM_ASTE_VALIDITY;
435 if (aste.astesn != ale.astesn)
436 return PGM_ASTE_SEQUENCE;
437
438 if (ale.p == 1) {
439 eax = (vcpu->arch.sie_block->gcr[8] >> 16) & 0xffff;
440 if (ale.aleax != eax) {
441 if (eax / 16 > aste.atl)
442 return PGM_EXTENDED_AUTHORITY;
443
444 authority_table_addr = aste.ato * 4 + eax / 4;
445
446 rc = read_guest_real(vcpu, authority_table_addr,
447 &authority_table,
448 sizeof(u8));
449 if (rc)
450 return rc;
451
452 if ((authority_table & (0x40 >> ((eax & 3) * 2))) == 0)
453 return PGM_EXTENDED_AUTHORITY;
454 }
455 }
456
457 if (ale.fo == 1 && mode == GACC_STORE)
458 return PGM_PROTECTION;
459
460 asce->val = aste.asce;
461 return 0;
462}
463
464struct trans_exc_code_bits {
465 unsigned long addr : 52; /* Translation-exception Address */
466 unsigned long fsi : 2; /* Access Exception Fetch/Store Indication */
467 unsigned long : 6;
468 unsigned long b60 : 1;
469 unsigned long b61 : 1;
470 unsigned long as : 2; /* ASCE Identifier */
471};
472
473enum {
474 FSI_UNKNOWN = 0, /* Unknown wether fetch or store */
475 FSI_STORE = 1, /* Exception was due to store operation */
476 FSI_FETCH = 2 /* Exception was due to fetch operation */
477};
478
479static int get_vcpu_asce(struct kvm_vcpu *vcpu, union asce *asce,
480 ar_t ar, enum gacc_mode mode)
481{
482 int rc;
483 struct psw_bits psw = psw_bits(vcpu->arch.sie_block->gpsw);
484 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
485 struct trans_exc_code_bits *tec_bits;
486
487 memset(pgm, 0, sizeof(*pgm));
488 tec_bits = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
489 tec_bits->fsi = mode == GACC_STORE ? FSI_STORE : FSI_FETCH;
490 tec_bits->as = psw.as;
491
492 if (!psw.t) {
493 asce->val = 0;
494 asce->r = 1;
495 return 0;
496 }
497
498 if (mode == GACC_IFETCH)
499 psw.as = psw.as == PSW_AS_HOME ? PSW_AS_HOME : PSW_AS_PRIMARY;
500
501 switch (psw.as) {
502 case PSW_AS_PRIMARY:
503 asce->val = vcpu->arch.sie_block->gcr[1];
504 return 0;
505 case PSW_AS_SECONDARY:
506 asce->val = vcpu->arch.sie_block->gcr[7];
507 return 0;
508 case PSW_AS_HOME:
509 asce->val = vcpu->arch.sie_block->gcr[13];
510 return 0;
511 case PSW_AS_ACCREG:
512 rc = ar_translation(vcpu, asce, ar, mode);
513 switch (rc) {
514 case PGM_ALEN_TRANSLATION:
515 case PGM_ALE_SEQUENCE:
516 case PGM_ASTE_VALIDITY:
517 case PGM_ASTE_SEQUENCE:
518 case PGM_EXTENDED_AUTHORITY:
519 vcpu->arch.pgm.exc_access_id = ar;
520 break;
521 case PGM_PROTECTION:
522 tec_bits->b60 = 1;
523 tec_bits->b61 = 1;
524 break;
525 }
526 if (rc > 0)
527 pgm->code = rc;
528 return rc;
529 }
530 return 0;
531}
532
533static int deref_table(struct kvm *kvm, unsigned long gpa, unsigned long *val)
534{
535 return kvm_read_guest(kvm, gpa, val, sizeof(*val));
536}
537
538/**
539 * guest_translate - translate a guest virtual into a guest absolute address
540 * @vcpu: virtual cpu
541 * @gva: guest virtual address
542 * @gpa: points to where guest physical (absolute) address should be stored
543 * @asce: effective asce
544 * @mode: indicates the access mode to be used
545 *
546 * Translate a guest virtual address into a guest absolute address by means
547 * of dynamic address translation as specified by the architecture.
548 * If the resulting absolute address is not available in the configuration
549 * an addressing exception is indicated and @gpa will not be changed.
550 *
551 * Returns: - zero on success; @gpa contains the resulting absolute address
552 * - a negative value if guest access failed due to e.g. broken
553 * guest mapping
554 * - a positve value if an access exception happened. In this case
555 * the returned value is the program interruption code as defined
556 * by the architecture
557 */
558static unsigned long guest_translate(struct kvm_vcpu *vcpu, unsigned long gva,
559 unsigned long *gpa, const union asce asce,
560 enum gacc_mode mode)
561{
562 union vaddress vaddr = {.addr = gva};
563 union raddress raddr = {.addr = gva};
564 union page_table_entry pte;
565 int dat_protection = 0;
566 union ctlreg0 ctlreg0;
567 unsigned long ptr;
568 int edat1, edat2;
569
570 ctlreg0.val = vcpu->arch.sie_block->gcr[0];
571 edat1 = ctlreg0.edat && test_kvm_facility(vcpu->kvm, 8);
572 edat2 = edat1 && test_kvm_facility(vcpu->kvm, 78);
573 if (asce.r)
574 goto real_address;
575 ptr = asce.origin * 4096;
576 switch (asce.dt) {
577 case ASCE_TYPE_REGION1:
578 if (vaddr.rfx01 > asce.tl)
579 return PGM_REGION_FIRST_TRANS;
580 ptr += vaddr.rfx * 8;
581 break;
582 case ASCE_TYPE_REGION2:
583 if (vaddr.rfx)
584 return PGM_ASCE_TYPE;
585 if (vaddr.rsx01 > asce.tl)
586 return PGM_REGION_SECOND_TRANS;
587 ptr += vaddr.rsx * 8;
588 break;
589 case ASCE_TYPE_REGION3:
590 if (vaddr.rfx || vaddr.rsx)
591 return PGM_ASCE_TYPE;
592 if (vaddr.rtx01 > asce.tl)
593 return PGM_REGION_THIRD_TRANS;
594 ptr += vaddr.rtx * 8;
595 break;
596 case ASCE_TYPE_SEGMENT:
597 if (vaddr.rfx || vaddr.rsx || vaddr.rtx)
598 return PGM_ASCE_TYPE;
599 if (vaddr.sx01 > asce.tl)
600 return PGM_SEGMENT_TRANSLATION;
601 ptr += vaddr.sx * 8;
602 break;
603 }
604 switch (asce.dt) {
605 case ASCE_TYPE_REGION1: {
606 union region1_table_entry rfte;
607
608 if (kvm_is_error_gpa(vcpu->kvm, ptr))
609 return PGM_ADDRESSING;
610 if (deref_table(vcpu->kvm, ptr, &rfte.val))
611 return -EFAULT;
612 if (rfte.i)
613 return PGM_REGION_FIRST_TRANS;
614 if (rfte.tt != TABLE_TYPE_REGION1)
615 return PGM_TRANSLATION_SPEC;
616 if (vaddr.rsx01 < rfte.tf || vaddr.rsx01 > rfte.tl)
617 return PGM_REGION_SECOND_TRANS;
618 if (edat1)
619 dat_protection |= rfte.p;
620 ptr = rfte.rto * 4096 + vaddr.rsx * 8;
621 }
622 /* fallthrough */
623 case ASCE_TYPE_REGION2: {
624 union region2_table_entry rste;
625
626 if (kvm_is_error_gpa(vcpu->kvm, ptr))
627 return PGM_ADDRESSING;
628 if (deref_table(vcpu->kvm, ptr, &rste.val))
629 return -EFAULT;
630 if (rste.i)
631 return PGM_REGION_SECOND_TRANS;
632 if (rste.tt != TABLE_TYPE_REGION2)
633 return PGM_TRANSLATION_SPEC;
634 if (vaddr.rtx01 < rste.tf || vaddr.rtx01 > rste.tl)
635 return PGM_REGION_THIRD_TRANS;
636 if (edat1)
637 dat_protection |= rste.p;
638 ptr = rste.rto * 4096 + vaddr.rtx * 8;
639 }
640 /* fallthrough */
641 case ASCE_TYPE_REGION3: {
642 union region3_table_entry rtte;
643
644 if (kvm_is_error_gpa(vcpu->kvm, ptr))
645 return PGM_ADDRESSING;
646 if (deref_table(vcpu->kvm, ptr, &rtte.val))
647 return -EFAULT;
648 if (rtte.i)
649 return PGM_REGION_THIRD_TRANS;
650 if (rtte.tt != TABLE_TYPE_REGION3)
651 return PGM_TRANSLATION_SPEC;
652 if (rtte.cr && asce.p && edat2)
653 return PGM_TRANSLATION_SPEC;
654 if (rtte.fc && edat2) {
655 dat_protection |= rtte.fc1.p;
656 raddr.rfaa = rtte.fc1.rfaa;
657 goto absolute_address;
658 }
659 if (vaddr.sx01 < rtte.fc0.tf)
660 return PGM_SEGMENT_TRANSLATION;
661 if (vaddr.sx01 > rtte.fc0.tl)
662 return PGM_SEGMENT_TRANSLATION;
663 if (edat1)
664 dat_protection |= rtte.fc0.p;
665 ptr = rtte.fc0.sto * 4096 + vaddr.sx * 8;
666 }
667 /* fallthrough */
668 case ASCE_TYPE_SEGMENT: {
669 union segment_table_entry ste;
670
671 if (kvm_is_error_gpa(vcpu->kvm, ptr))
672 return PGM_ADDRESSING;
673 if (deref_table(vcpu->kvm, ptr, &ste.val))
674 return -EFAULT;
675 if (ste.i)
676 return PGM_SEGMENT_TRANSLATION;
677 if (ste.tt != TABLE_TYPE_SEGMENT)
678 return PGM_TRANSLATION_SPEC;
679 if (ste.cs && asce.p)
680 return PGM_TRANSLATION_SPEC;
681 if (ste.fc && edat1) {
682 dat_protection |= ste.fc1.p;
683 raddr.sfaa = ste.fc1.sfaa;
684 goto absolute_address;
685 }
686 dat_protection |= ste.fc0.p;
687 ptr = ste.fc0.pto * 2048 + vaddr.px * 8;
688 }
689 }
690 if (kvm_is_error_gpa(vcpu->kvm, ptr))
691 return PGM_ADDRESSING;
692 if (deref_table(vcpu->kvm, ptr, &pte.val))
693 return -EFAULT;
694 if (pte.i)
695 return PGM_PAGE_TRANSLATION;
696 if (pte.z)
697 return PGM_TRANSLATION_SPEC;
698 if (pte.co && !edat1)
699 return PGM_TRANSLATION_SPEC;
700 dat_protection |= pte.p;
701 raddr.pfra = pte.pfra;
702real_address:
703 raddr.addr = kvm_s390_real_to_abs(vcpu, raddr.addr);
704absolute_address:
705 if (mode == GACC_STORE && dat_protection)
706 return PGM_PROTECTION;
707 if (kvm_is_error_gpa(vcpu->kvm, raddr.addr))
708 return PGM_ADDRESSING;
709 *gpa = raddr.addr;
710 return 0;
711}
712
713static inline int is_low_address(unsigned long ga)
714{
715 /* Check for address ranges 0..511 and 4096..4607 */
716 return (ga & ~0x11fful) == 0;
717}
718
719static int low_address_protection_enabled(struct kvm_vcpu *vcpu,
720 const union asce asce)
721{
722 union ctlreg0 ctlreg0 = {.val = vcpu->arch.sie_block->gcr[0]};
723 psw_t *psw = &vcpu->arch.sie_block->gpsw;
724
725 if (!ctlreg0.lap)
726 return 0;
727 if (psw_bits(*psw).t && asce.p)
728 return 0;
729 return 1;
730}
731
732static int guest_page_range(struct kvm_vcpu *vcpu, unsigned long ga,
733 unsigned long *pages, unsigned long nr_pages,
734 const union asce asce, enum gacc_mode mode)
735{
736 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
737 psw_t *psw = &vcpu->arch.sie_block->gpsw;
738 struct trans_exc_code_bits *tec_bits;
739 int lap_enabled, rc;
740
741 tec_bits = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
742 lap_enabled = low_address_protection_enabled(vcpu, asce);
743 while (nr_pages) {
744 ga = kvm_s390_logical_to_effective(vcpu, ga);
745 tec_bits->addr = ga >> PAGE_SHIFT;
746 if (mode == GACC_STORE && lap_enabled && is_low_address(ga)) {
747 pgm->code = PGM_PROTECTION;
748 return pgm->code;
749 }
750 ga &= PAGE_MASK;
751 if (psw_bits(*psw).t) {
752 rc = guest_translate(vcpu, ga, pages, asce, mode);
753 if (rc < 0)
754 return rc;
755 if (rc == PGM_PROTECTION)
756 tec_bits->b61 = 1;
757 if (rc)
758 pgm->code = rc;
759 } else {
760 *pages = kvm_s390_real_to_abs(vcpu, ga);
761 if (kvm_is_error_gpa(vcpu->kvm, *pages))
762 pgm->code = PGM_ADDRESSING;
763 }
764 if (pgm->code)
765 return pgm->code;
766 ga += PAGE_SIZE;
767 pages++;
768 nr_pages--;
769 }
770 return 0;
771}
772
773int access_guest(struct kvm_vcpu *vcpu, unsigned long ga, ar_t ar, void *data,
774 unsigned long len, enum gacc_mode mode)
775{
776 psw_t *psw = &vcpu->arch.sie_block->gpsw;
777 unsigned long _len, nr_pages, gpa, idx;
778 unsigned long pages_array[2];
779 unsigned long *pages;
780 int need_ipte_lock;
781 union asce asce;
782 int rc;
783
784 if (!len)
785 return 0;
786 rc = get_vcpu_asce(vcpu, &asce, ar, mode);
787 if (rc)
788 return rc;
789 nr_pages = (((ga & ~PAGE_MASK) + len - 1) >> PAGE_SHIFT) + 1;
790 pages = pages_array;
791 if (nr_pages > ARRAY_SIZE(pages_array))
792 pages = vmalloc(nr_pages * sizeof(unsigned long));
793 if (!pages)
794 return -ENOMEM;
795 need_ipte_lock = psw_bits(*psw).t && !asce.r;
796 if (need_ipte_lock)
797 ipte_lock(vcpu);
798 rc = guest_page_range(vcpu, ga, pages, nr_pages, asce, mode);
799 for (idx = 0; idx < nr_pages && !rc; idx++) {
800 gpa = *(pages + idx) + (ga & ~PAGE_MASK);
801 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
802 if (mode == GACC_STORE)
803 rc = kvm_write_guest(vcpu->kvm, gpa, data, _len);
804 else
805 rc = kvm_read_guest(vcpu->kvm, gpa, data, _len);
806 len -= _len;
807 ga += _len;
808 data += _len;
809 }
810 if (need_ipte_lock)
811 ipte_unlock(vcpu);
812 if (nr_pages > ARRAY_SIZE(pages_array))
813 vfree(pages);
814 return rc;
815}
816
817int access_guest_real(struct kvm_vcpu *vcpu, unsigned long gra,
818 void *data, unsigned long len, enum gacc_mode mode)
819{
820 unsigned long _len, gpa;
821 int rc = 0;
822
823 while (len && !rc) {
824 gpa = kvm_s390_real_to_abs(vcpu, gra);
825 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
826 if (mode)
827 rc = write_guest_abs(vcpu, gpa, data, _len);
828 else
829 rc = read_guest_abs(vcpu, gpa, data, _len);
830 len -= _len;
831 gra += _len;
832 data += _len;
833 }
834 return rc;
835}
836
837/**
838 * guest_translate_address - translate guest logical into guest absolute address
839 *
840 * Parameter semantics are the same as the ones from guest_translate.
841 * The memory contents at the guest address are not changed.
842 *
843 * Note: The IPTE lock is not taken during this function, so the caller
844 * has to take care of this.
845 */
846int guest_translate_address(struct kvm_vcpu *vcpu, unsigned long gva, ar_t ar,
847 unsigned long *gpa, enum gacc_mode mode)
848{
849 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
850 psw_t *psw = &vcpu->arch.sie_block->gpsw;
851 struct trans_exc_code_bits *tec;
852 union asce asce;
853 int rc;
854
855 gva = kvm_s390_logical_to_effective(vcpu, gva);
856 tec = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
857 rc = get_vcpu_asce(vcpu, &asce, ar, mode);
858 tec->addr = gva >> PAGE_SHIFT;
859 if (rc)
860 return rc;
861 if (is_low_address(gva) && low_address_protection_enabled(vcpu, asce)) {
862 if (mode == GACC_STORE) {
863 rc = pgm->code = PGM_PROTECTION;
864 return rc;
865 }
866 }
867
868 if (psw_bits(*psw).t && !asce.r) { /* Use DAT? */
869 rc = guest_translate(vcpu, gva, gpa, asce, mode);
870 if (rc > 0) {
871 if (rc == PGM_PROTECTION)
872 tec->b61 = 1;
873 pgm->code = rc;
874 }
875 } else {
876 rc = 0;
877 *gpa = kvm_s390_real_to_abs(vcpu, gva);
878 if (kvm_is_error_gpa(vcpu->kvm, *gpa))
879 rc = pgm->code = PGM_ADDRESSING;
880 }
881
882 return rc;
883}
884
885/**
886 * check_gva_range - test a range of guest virtual addresses for accessibility
887 */
888int check_gva_range(struct kvm_vcpu *vcpu, unsigned long gva, ar_t ar,
889 unsigned long length, enum gacc_mode mode)
890{
891 unsigned long gpa;
892 unsigned long currlen;
893 int rc = 0;
894
895 ipte_lock(vcpu);
896 while (length > 0 && !rc) {
897 currlen = min(length, PAGE_SIZE - (gva % PAGE_SIZE));
898 rc = guest_translate_address(vcpu, gva, ar, &gpa, mode);
899 gva += currlen;
900 length -= currlen;
901 }
902 ipte_unlock(vcpu);
903
904 return rc;
905}
906
907/**
908 * kvm_s390_check_low_addr_prot_real - check for low-address protection
909 * @gra: Guest real address
910 *
911 * Checks whether an address is subject to low-address protection and set
912 * up vcpu->arch.pgm accordingly if necessary.
913 *
914 * Return: 0 if no protection exception, or PGM_PROTECTION if protected.
915 */
916int kvm_s390_check_low_addr_prot_real(struct kvm_vcpu *vcpu, unsigned long gra)
917{
918 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
919 psw_t *psw = &vcpu->arch.sie_block->gpsw;
920 struct trans_exc_code_bits *tec_bits;
921 union ctlreg0 ctlreg0 = {.val = vcpu->arch.sie_block->gcr[0]};
922
923 if (!ctlreg0.lap || !is_low_address(gra))
924 return 0;
925
926 memset(pgm, 0, sizeof(*pgm));
927 tec_bits = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
928 tec_bits->fsi = FSI_STORE;
929 tec_bits->as = psw_bits(*psw).as;
930 tec_bits->addr = gra >> PAGE_SHIFT;
931 pgm->code = PGM_PROTECTION;
932
933 return pgm->code;
934}
1/*
2 * guest access functions
3 *
4 * Copyright IBM Corp. 2014
5 *
6 */
7
8#include <linux/vmalloc.h>
9#include <linux/err.h>
10#include <asm/pgtable.h>
11#include <asm/gmap.h>
12#include "kvm-s390.h"
13#include "gaccess.h"
14#include <asm/switch_to.h>
15
16union asce {
17 unsigned long val;
18 struct {
19 unsigned long origin : 52; /* Region- or Segment-Table Origin */
20 unsigned long : 2;
21 unsigned long g : 1; /* Subspace Group Control */
22 unsigned long p : 1; /* Private Space Control */
23 unsigned long s : 1; /* Storage-Alteration-Event Control */
24 unsigned long x : 1; /* Space-Switch-Event Control */
25 unsigned long r : 1; /* Real-Space Control */
26 unsigned long : 1;
27 unsigned long dt : 2; /* Designation-Type Control */
28 unsigned long tl : 2; /* Region- or Segment-Table Length */
29 };
30};
31
32enum {
33 ASCE_TYPE_SEGMENT = 0,
34 ASCE_TYPE_REGION3 = 1,
35 ASCE_TYPE_REGION2 = 2,
36 ASCE_TYPE_REGION1 = 3
37};
38
39union region1_table_entry {
40 unsigned long val;
41 struct {
42 unsigned long rto: 52;/* Region-Table Origin */
43 unsigned long : 2;
44 unsigned long p : 1; /* DAT-Protection Bit */
45 unsigned long : 1;
46 unsigned long tf : 2; /* Region-Second-Table Offset */
47 unsigned long i : 1; /* Region-Invalid Bit */
48 unsigned long : 1;
49 unsigned long tt : 2; /* Table-Type Bits */
50 unsigned long tl : 2; /* Region-Second-Table Length */
51 };
52};
53
54union region2_table_entry {
55 unsigned long val;
56 struct {
57 unsigned long rto: 52;/* Region-Table Origin */
58 unsigned long : 2;
59 unsigned long p : 1; /* DAT-Protection Bit */
60 unsigned long : 1;
61 unsigned long tf : 2; /* Region-Third-Table Offset */
62 unsigned long i : 1; /* Region-Invalid Bit */
63 unsigned long : 1;
64 unsigned long tt : 2; /* Table-Type Bits */
65 unsigned long tl : 2; /* Region-Third-Table Length */
66 };
67};
68
69struct region3_table_entry_fc0 {
70 unsigned long sto: 52;/* Segment-Table Origin */
71 unsigned long : 1;
72 unsigned long fc : 1; /* Format-Control */
73 unsigned long p : 1; /* DAT-Protection Bit */
74 unsigned long : 1;
75 unsigned long tf : 2; /* Segment-Table Offset */
76 unsigned long i : 1; /* Region-Invalid Bit */
77 unsigned long cr : 1; /* Common-Region Bit */
78 unsigned long tt : 2; /* Table-Type Bits */
79 unsigned long tl : 2; /* Segment-Table Length */
80};
81
82struct region3_table_entry_fc1 {
83 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
84 unsigned long : 14;
85 unsigned long av : 1; /* ACCF-Validity Control */
86 unsigned long acc: 4; /* Access-Control Bits */
87 unsigned long f : 1; /* Fetch-Protection Bit */
88 unsigned long fc : 1; /* Format-Control */
89 unsigned long p : 1; /* DAT-Protection Bit */
90 unsigned long co : 1; /* Change-Recording Override */
91 unsigned long : 2;
92 unsigned long i : 1; /* Region-Invalid Bit */
93 unsigned long cr : 1; /* Common-Region Bit */
94 unsigned long tt : 2; /* Table-Type Bits */
95 unsigned long : 2;
96};
97
98union region3_table_entry {
99 unsigned long val;
100 struct region3_table_entry_fc0 fc0;
101 struct region3_table_entry_fc1 fc1;
102 struct {
103 unsigned long : 53;
104 unsigned long fc : 1; /* Format-Control */
105 unsigned long : 4;
106 unsigned long i : 1; /* Region-Invalid Bit */
107 unsigned long cr : 1; /* Common-Region Bit */
108 unsigned long tt : 2; /* Table-Type Bits */
109 unsigned long : 2;
110 };
111};
112
113struct segment_entry_fc0 {
114 unsigned long pto: 53;/* Page-Table Origin */
115 unsigned long fc : 1; /* Format-Control */
116 unsigned long p : 1; /* DAT-Protection Bit */
117 unsigned long : 3;
118 unsigned long i : 1; /* Segment-Invalid Bit */
119 unsigned long cs : 1; /* Common-Segment Bit */
120 unsigned long tt : 2; /* Table-Type Bits */
121 unsigned long : 2;
122};
123
124struct segment_entry_fc1 {
125 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
126 unsigned long : 3;
127 unsigned long av : 1; /* ACCF-Validity Control */
128 unsigned long acc: 4; /* Access-Control Bits */
129 unsigned long f : 1; /* Fetch-Protection Bit */
130 unsigned long fc : 1; /* Format-Control */
131 unsigned long p : 1; /* DAT-Protection Bit */
132 unsigned long co : 1; /* Change-Recording Override */
133 unsigned long : 2;
134 unsigned long i : 1; /* Segment-Invalid Bit */
135 unsigned long cs : 1; /* Common-Segment Bit */
136 unsigned long tt : 2; /* Table-Type Bits */
137 unsigned long : 2;
138};
139
140union segment_table_entry {
141 unsigned long val;
142 struct segment_entry_fc0 fc0;
143 struct segment_entry_fc1 fc1;
144 struct {
145 unsigned long : 53;
146 unsigned long fc : 1; /* Format-Control */
147 unsigned long : 4;
148 unsigned long i : 1; /* Segment-Invalid Bit */
149 unsigned long cs : 1; /* Common-Segment Bit */
150 unsigned long tt : 2; /* Table-Type Bits */
151 unsigned long : 2;
152 };
153};
154
155enum {
156 TABLE_TYPE_SEGMENT = 0,
157 TABLE_TYPE_REGION3 = 1,
158 TABLE_TYPE_REGION2 = 2,
159 TABLE_TYPE_REGION1 = 3
160};
161
162union page_table_entry {
163 unsigned long val;
164 struct {
165 unsigned long pfra : 52; /* Page-Frame Real Address */
166 unsigned long z : 1; /* Zero Bit */
167 unsigned long i : 1; /* Page-Invalid Bit */
168 unsigned long p : 1; /* DAT-Protection Bit */
169 unsigned long co : 1; /* Change-Recording Override */
170 unsigned long : 8;
171 };
172};
173
174/*
175 * vaddress union in order to easily decode a virtual address into its
176 * region first index, region second index etc. parts.
177 */
178union vaddress {
179 unsigned long addr;
180 struct {
181 unsigned long rfx : 11;
182 unsigned long rsx : 11;
183 unsigned long rtx : 11;
184 unsigned long sx : 11;
185 unsigned long px : 8;
186 unsigned long bx : 12;
187 };
188 struct {
189 unsigned long rfx01 : 2;
190 unsigned long : 9;
191 unsigned long rsx01 : 2;
192 unsigned long : 9;
193 unsigned long rtx01 : 2;
194 unsigned long : 9;
195 unsigned long sx01 : 2;
196 unsigned long : 29;
197 };
198};
199
200/*
201 * raddress union which will contain the result (real or absolute address)
202 * after a page table walk. The rfaa, sfaa and pfra members are used to
203 * simply assign them the value of a region, segment or page table entry.
204 */
205union raddress {
206 unsigned long addr;
207 unsigned long rfaa : 33; /* Region-Frame Absolute Address */
208 unsigned long sfaa : 44; /* Segment-Frame Absolute Address */
209 unsigned long pfra : 52; /* Page-Frame Real Address */
210};
211
212union alet {
213 u32 val;
214 struct {
215 u32 reserved : 7;
216 u32 p : 1;
217 u32 alesn : 8;
218 u32 alen : 16;
219 };
220};
221
222union ald {
223 u32 val;
224 struct {
225 u32 : 1;
226 u32 alo : 24;
227 u32 all : 7;
228 };
229};
230
231struct ale {
232 unsigned long i : 1; /* ALEN-Invalid Bit */
233 unsigned long : 5;
234 unsigned long fo : 1; /* Fetch-Only Bit */
235 unsigned long p : 1; /* Private Bit */
236 unsigned long alesn : 8; /* Access-List-Entry Sequence Number */
237 unsigned long aleax : 16; /* Access-List-Entry Authorization Index */
238 unsigned long : 32;
239 unsigned long : 1;
240 unsigned long asteo : 25; /* ASN-Second-Table-Entry Origin */
241 unsigned long : 6;
242 unsigned long astesn : 32; /* ASTE Sequence Number */
243} __packed;
244
245struct aste {
246 unsigned long i : 1; /* ASX-Invalid Bit */
247 unsigned long ato : 29; /* Authority-Table Origin */
248 unsigned long : 1;
249 unsigned long b : 1; /* Base-Space Bit */
250 unsigned long ax : 16; /* Authorization Index */
251 unsigned long atl : 12; /* Authority-Table Length */
252 unsigned long : 2;
253 unsigned long ca : 1; /* Controlled-ASN Bit */
254 unsigned long ra : 1; /* Reusable-ASN Bit */
255 unsigned long asce : 64; /* Address-Space-Control Element */
256 unsigned long ald : 32;
257 unsigned long astesn : 32;
258 /* .. more fields there */
259} __packed;
260
261int ipte_lock_held(struct kvm_vcpu *vcpu)
262{
263 if (vcpu->arch.sie_block->eca & 1) {
264 int rc;
265
266 read_lock(&vcpu->kvm->arch.sca_lock);
267 rc = kvm_s390_get_ipte_control(vcpu->kvm)->kh != 0;
268 read_unlock(&vcpu->kvm->arch.sca_lock);
269 return rc;
270 }
271 return vcpu->kvm->arch.ipte_lock_count != 0;
272}
273
274static void ipte_lock_simple(struct kvm_vcpu *vcpu)
275{
276 union ipte_control old, new, *ic;
277
278 mutex_lock(&vcpu->kvm->arch.ipte_mutex);
279 vcpu->kvm->arch.ipte_lock_count++;
280 if (vcpu->kvm->arch.ipte_lock_count > 1)
281 goto out;
282retry:
283 read_lock(&vcpu->kvm->arch.sca_lock);
284 ic = kvm_s390_get_ipte_control(vcpu->kvm);
285 do {
286 old = READ_ONCE(*ic);
287 if (old.k) {
288 read_unlock(&vcpu->kvm->arch.sca_lock);
289 cond_resched();
290 goto retry;
291 }
292 new = old;
293 new.k = 1;
294 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
295 read_unlock(&vcpu->kvm->arch.sca_lock);
296out:
297 mutex_unlock(&vcpu->kvm->arch.ipte_mutex);
298}
299
300static void ipte_unlock_simple(struct kvm_vcpu *vcpu)
301{
302 union ipte_control old, new, *ic;
303
304 mutex_lock(&vcpu->kvm->arch.ipte_mutex);
305 vcpu->kvm->arch.ipte_lock_count--;
306 if (vcpu->kvm->arch.ipte_lock_count)
307 goto out;
308 read_lock(&vcpu->kvm->arch.sca_lock);
309 ic = kvm_s390_get_ipte_control(vcpu->kvm);
310 do {
311 old = READ_ONCE(*ic);
312 new = old;
313 new.k = 0;
314 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
315 read_unlock(&vcpu->kvm->arch.sca_lock);
316 wake_up(&vcpu->kvm->arch.ipte_wq);
317out:
318 mutex_unlock(&vcpu->kvm->arch.ipte_mutex);
319}
320
321static void ipte_lock_siif(struct kvm_vcpu *vcpu)
322{
323 union ipte_control old, new, *ic;
324
325retry:
326 read_lock(&vcpu->kvm->arch.sca_lock);
327 ic = kvm_s390_get_ipte_control(vcpu->kvm);
328 do {
329 old = READ_ONCE(*ic);
330 if (old.kg) {
331 read_unlock(&vcpu->kvm->arch.sca_lock);
332 cond_resched();
333 goto retry;
334 }
335 new = old;
336 new.k = 1;
337 new.kh++;
338 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
339 read_unlock(&vcpu->kvm->arch.sca_lock);
340}
341
342static void ipte_unlock_siif(struct kvm_vcpu *vcpu)
343{
344 union ipte_control old, new, *ic;
345
346 read_lock(&vcpu->kvm->arch.sca_lock);
347 ic = kvm_s390_get_ipte_control(vcpu->kvm);
348 do {
349 old = READ_ONCE(*ic);
350 new = old;
351 new.kh--;
352 if (!new.kh)
353 new.k = 0;
354 } while (cmpxchg(&ic->val, old.val, new.val) != old.val);
355 read_unlock(&vcpu->kvm->arch.sca_lock);
356 if (!new.kh)
357 wake_up(&vcpu->kvm->arch.ipte_wq);
358}
359
360void ipte_lock(struct kvm_vcpu *vcpu)
361{
362 if (vcpu->arch.sie_block->eca & 1)
363 ipte_lock_siif(vcpu);
364 else
365 ipte_lock_simple(vcpu);
366}
367
368void ipte_unlock(struct kvm_vcpu *vcpu)
369{
370 if (vcpu->arch.sie_block->eca & 1)
371 ipte_unlock_siif(vcpu);
372 else
373 ipte_unlock_simple(vcpu);
374}
375
376static int ar_translation(struct kvm_vcpu *vcpu, union asce *asce, ar_t ar,
377 enum gacc_mode mode)
378{
379 union alet alet;
380 struct ale ale;
381 struct aste aste;
382 unsigned long ald_addr, authority_table_addr;
383 union ald ald;
384 int eax, rc;
385 u8 authority_table;
386
387 if (ar >= NUM_ACRS)
388 return -EINVAL;
389
390 save_access_regs(vcpu->run->s.regs.acrs);
391 alet.val = vcpu->run->s.regs.acrs[ar];
392
393 if (ar == 0 || alet.val == 0) {
394 asce->val = vcpu->arch.sie_block->gcr[1];
395 return 0;
396 } else if (alet.val == 1) {
397 asce->val = vcpu->arch.sie_block->gcr[7];
398 return 0;
399 }
400
401 if (alet.reserved)
402 return PGM_ALET_SPECIFICATION;
403
404 if (alet.p)
405 ald_addr = vcpu->arch.sie_block->gcr[5];
406 else
407 ald_addr = vcpu->arch.sie_block->gcr[2];
408 ald_addr &= 0x7fffffc0;
409
410 rc = read_guest_real(vcpu, ald_addr + 16, &ald.val, sizeof(union ald));
411 if (rc)
412 return rc;
413
414 if (alet.alen / 8 > ald.all)
415 return PGM_ALEN_TRANSLATION;
416
417 if (0x7fffffff - ald.alo * 128 < alet.alen * 16)
418 return PGM_ADDRESSING;
419
420 rc = read_guest_real(vcpu, ald.alo * 128 + alet.alen * 16, &ale,
421 sizeof(struct ale));
422 if (rc)
423 return rc;
424
425 if (ale.i == 1)
426 return PGM_ALEN_TRANSLATION;
427 if (ale.alesn != alet.alesn)
428 return PGM_ALE_SEQUENCE;
429
430 rc = read_guest_real(vcpu, ale.asteo * 64, &aste, sizeof(struct aste));
431 if (rc)
432 return rc;
433
434 if (aste.i)
435 return PGM_ASTE_VALIDITY;
436 if (aste.astesn != ale.astesn)
437 return PGM_ASTE_SEQUENCE;
438
439 if (ale.p == 1) {
440 eax = (vcpu->arch.sie_block->gcr[8] >> 16) & 0xffff;
441 if (ale.aleax != eax) {
442 if (eax / 16 > aste.atl)
443 return PGM_EXTENDED_AUTHORITY;
444
445 authority_table_addr = aste.ato * 4 + eax / 4;
446
447 rc = read_guest_real(vcpu, authority_table_addr,
448 &authority_table,
449 sizeof(u8));
450 if (rc)
451 return rc;
452
453 if ((authority_table & (0x40 >> ((eax & 3) * 2))) == 0)
454 return PGM_EXTENDED_AUTHORITY;
455 }
456 }
457
458 if (ale.fo == 1 && mode == GACC_STORE)
459 return PGM_PROTECTION;
460
461 asce->val = aste.asce;
462 return 0;
463}
464
465struct trans_exc_code_bits {
466 unsigned long addr : 52; /* Translation-exception Address */
467 unsigned long fsi : 2; /* Access Exception Fetch/Store Indication */
468 unsigned long : 6;
469 unsigned long b60 : 1;
470 unsigned long b61 : 1;
471 unsigned long as : 2; /* ASCE Identifier */
472};
473
474enum {
475 FSI_UNKNOWN = 0, /* Unknown wether fetch or store */
476 FSI_STORE = 1, /* Exception was due to store operation */
477 FSI_FETCH = 2 /* Exception was due to fetch operation */
478};
479
480enum prot_type {
481 PROT_TYPE_LA = 0,
482 PROT_TYPE_KEYC = 1,
483 PROT_TYPE_ALC = 2,
484 PROT_TYPE_DAT = 3,
485};
486
487static int trans_exc(struct kvm_vcpu *vcpu, int code, unsigned long gva,
488 ar_t ar, enum gacc_mode mode, enum prot_type prot)
489{
490 struct kvm_s390_pgm_info *pgm = &vcpu->arch.pgm;
491 struct trans_exc_code_bits *tec;
492
493 memset(pgm, 0, sizeof(*pgm));
494 pgm->code = code;
495 tec = (struct trans_exc_code_bits *)&pgm->trans_exc_code;
496
497 switch (code) {
498 case PGM_PROTECTION:
499 switch (prot) {
500 case PROT_TYPE_ALC:
501 tec->b60 = 1;
502 /* FALL THROUGH */
503 case PROT_TYPE_DAT:
504 tec->b61 = 1;
505 break;
506 default: /* LA and KEYC set b61 to 0, other params undefined */
507 return code;
508 }
509 /* FALL THROUGH */
510 case PGM_ASCE_TYPE:
511 case PGM_PAGE_TRANSLATION:
512 case PGM_REGION_FIRST_TRANS:
513 case PGM_REGION_SECOND_TRANS:
514 case PGM_REGION_THIRD_TRANS:
515 case PGM_SEGMENT_TRANSLATION:
516 /*
517 * op_access_id only applies to MOVE_PAGE -> set bit 61
518 * exc_access_id has to be set to 0 for some instructions. Both
519 * cases have to be handled by the caller.
520 */
521 tec->addr = gva >> PAGE_SHIFT;
522 tec->fsi = mode == GACC_STORE ? FSI_STORE : FSI_FETCH;
523 tec->as = psw_bits(vcpu->arch.sie_block->gpsw).as;
524 /* FALL THROUGH */
525 case PGM_ALEN_TRANSLATION:
526 case PGM_ALE_SEQUENCE:
527 case PGM_ASTE_VALIDITY:
528 case PGM_ASTE_SEQUENCE:
529 case PGM_EXTENDED_AUTHORITY:
530 /*
531 * We can always store exc_access_id, as it is
532 * undefined for non-ar cases. It is undefined for
533 * most DAT protection exceptions.
534 */
535 pgm->exc_access_id = ar;
536 break;
537 }
538 return code;
539}
540
541static int get_vcpu_asce(struct kvm_vcpu *vcpu, union asce *asce,
542 unsigned long ga, ar_t ar, enum gacc_mode mode)
543{
544 int rc;
545 struct psw_bits psw = psw_bits(vcpu->arch.sie_block->gpsw);
546
547 if (!psw.t) {
548 asce->val = 0;
549 asce->r = 1;
550 return 0;
551 }
552
553 if (mode == GACC_IFETCH)
554 psw.as = psw.as == PSW_AS_HOME ? PSW_AS_HOME : PSW_AS_PRIMARY;
555
556 switch (psw.as) {
557 case PSW_AS_PRIMARY:
558 asce->val = vcpu->arch.sie_block->gcr[1];
559 return 0;
560 case PSW_AS_SECONDARY:
561 asce->val = vcpu->arch.sie_block->gcr[7];
562 return 0;
563 case PSW_AS_HOME:
564 asce->val = vcpu->arch.sie_block->gcr[13];
565 return 0;
566 case PSW_AS_ACCREG:
567 rc = ar_translation(vcpu, asce, ar, mode);
568 if (rc > 0)
569 return trans_exc(vcpu, rc, ga, ar, mode, PROT_TYPE_ALC);
570 return rc;
571 }
572 return 0;
573}
574
575static int deref_table(struct kvm *kvm, unsigned long gpa, unsigned long *val)
576{
577 return kvm_read_guest(kvm, gpa, val, sizeof(*val));
578}
579
580/**
581 * guest_translate - translate a guest virtual into a guest absolute address
582 * @vcpu: virtual cpu
583 * @gva: guest virtual address
584 * @gpa: points to where guest physical (absolute) address should be stored
585 * @asce: effective asce
586 * @mode: indicates the access mode to be used
587 *
588 * Translate a guest virtual address into a guest absolute address by means
589 * of dynamic address translation as specified by the architecture.
590 * If the resulting absolute address is not available in the configuration
591 * an addressing exception is indicated and @gpa will not be changed.
592 *
593 * Returns: - zero on success; @gpa contains the resulting absolute address
594 * - a negative value if guest access failed due to e.g. broken
595 * guest mapping
596 * - a positve value if an access exception happened. In this case
597 * the returned value is the program interruption code as defined
598 * by the architecture
599 */
600static unsigned long guest_translate(struct kvm_vcpu *vcpu, unsigned long gva,
601 unsigned long *gpa, const union asce asce,
602 enum gacc_mode mode)
603{
604 union vaddress vaddr = {.addr = gva};
605 union raddress raddr = {.addr = gva};
606 union page_table_entry pte;
607 int dat_protection = 0;
608 union ctlreg0 ctlreg0;
609 unsigned long ptr;
610 int edat1, edat2;
611
612 ctlreg0.val = vcpu->arch.sie_block->gcr[0];
613 edat1 = ctlreg0.edat && test_kvm_facility(vcpu->kvm, 8);
614 edat2 = edat1 && test_kvm_facility(vcpu->kvm, 78);
615 if (asce.r)
616 goto real_address;
617 ptr = asce.origin * 4096;
618 switch (asce.dt) {
619 case ASCE_TYPE_REGION1:
620 if (vaddr.rfx01 > asce.tl)
621 return PGM_REGION_FIRST_TRANS;
622 ptr += vaddr.rfx * 8;
623 break;
624 case ASCE_TYPE_REGION2:
625 if (vaddr.rfx)
626 return PGM_ASCE_TYPE;
627 if (vaddr.rsx01 > asce.tl)
628 return PGM_REGION_SECOND_TRANS;
629 ptr += vaddr.rsx * 8;
630 break;
631 case ASCE_TYPE_REGION3:
632 if (vaddr.rfx || vaddr.rsx)
633 return PGM_ASCE_TYPE;
634 if (vaddr.rtx01 > asce.tl)
635 return PGM_REGION_THIRD_TRANS;
636 ptr += vaddr.rtx * 8;
637 break;
638 case ASCE_TYPE_SEGMENT:
639 if (vaddr.rfx || vaddr.rsx || vaddr.rtx)
640 return PGM_ASCE_TYPE;
641 if (vaddr.sx01 > asce.tl)
642 return PGM_SEGMENT_TRANSLATION;
643 ptr += vaddr.sx * 8;
644 break;
645 }
646 switch (asce.dt) {
647 case ASCE_TYPE_REGION1: {
648 union region1_table_entry rfte;
649
650 if (kvm_is_error_gpa(vcpu->kvm, ptr))
651 return PGM_ADDRESSING;
652 if (deref_table(vcpu->kvm, ptr, &rfte.val))
653 return -EFAULT;
654 if (rfte.i)
655 return PGM_REGION_FIRST_TRANS;
656 if (rfte.tt != TABLE_TYPE_REGION1)
657 return PGM_TRANSLATION_SPEC;
658 if (vaddr.rsx01 < rfte.tf || vaddr.rsx01 > rfte.tl)
659 return PGM_REGION_SECOND_TRANS;
660 if (edat1)
661 dat_protection |= rfte.p;
662 ptr = rfte.rto * 4096 + vaddr.rsx * 8;
663 }
664 /* fallthrough */
665 case ASCE_TYPE_REGION2: {
666 union region2_table_entry rste;
667
668 if (kvm_is_error_gpa(vcpu->kvm, ptr))
669 return PGM_ADDRESSING;
670 if (deref_table(vcpu->kvm, ptr, &rste.val))
671 return -EFAULT;
672 if (rste.i)
673 return PGM_REGION_SECOND_TRANS;
674 if (rste.tt != TABLE_TYPE_REGION2)
675 return PGM_TRANSLATION_SPEC;
676 if (vaddr.rtx01 < rste.tf || vaddr.rtx01 > rste.tl)
677 return PGM_REGION_THIRD_TRANS;
678 if (edat1)
679 dat_protection |= rste.p;
680 ptr = rste.rto * 4096 + vaddr.rtx * 8;
681 }
682 /* fallthrough */
683 case ASCE_TYPE_REGION3: {
684 union region3_table_entry rtte;
685
686 if (kvm_is_error_gpa(vcpu->kvm, ptr))
687 return PGM_ADDRESSING;
688 if (deref_table(vcpu->kvm, ptr, &rtte.val))
689 return -EFAULT;
690 if (rtte.i)
691 return PGM_REGION_THIRD_TRANS;
692 if (rtte.tt != TABLE_TYPE_REGION3)
693 return PGM_TRANSLATION_SPEC;
694 if (rtte.cr && asce.p && edat2)
695 return PGM_TRANSLATION_SPEC;
696 if (rtte.fc && edat2) {
697 dat_protection |= rtte.fc1.p;
698 raddr.rfaa = rtte.fc1.rfaa;
699 goto absolute_address;
700 }
701 if (vaddr.sx01 < rtte.fc0.tf)
702 return PGM_SEGMENT_TRANSLATION;
703 if (vaddr.sx01 > rtte.fc0.tl)
704 return PGM_SEGMENT_TRANSLATION;
705 if (edat1)
706 dat_protection |= rtte.fc0.p;
707 ptr = rtte.fc0.sto * 4096 + vaddr.sx * 8;
708 }
709 /* fallthrough */
710 case ASCE_TYPE_SEGMENT: {
711 union segment_table_entry ste;
712
713 if (kvm_is_error_gpa(vcpu->kvm, ptr))
714 return PGM_ADDRESSING;
715 if (deref_table(vcpu->kvm, ptr, &ste.val))
716 return -EFAULT;
717 if (ste.i)
718 return PGM_SEGMENT_TRANSLATION;
719 if (ste.tt != TABLE_TYPE_SEGMENT)
720 return PGM_TRANSLATION_SPEC;
721 if (ste.cs && asce.p)
722 return PGM_TRANSLATION_SPEC;
723 if (ste.fc && edat1) {
724 dat_protection |= ste.fc1.p;
725 raddr.sfaa = ste.fc1.sfaa;
726 goto absolute_address;
727 }
728 dat_protection |= ste.fc0.p;
729 ptr = ste.fc0.pto * 2048 + vaddr.px * 8;
730 }
731 }
732 if (kvm_is_error_gpa(vcpu->kvm, ptr))
733 return PGM_ADDRESSING;
734 if (deref_table(vcpu->kvm, ptr, &pte.val))
735 return -EFAULT;
736 if (pte.i)
737 return PGM_PAGE_TRANSLATION;
738 if (pte.z)
739 return PGM_TRANSLATION_SPEC;
740 if (pte.co && !edat1)
741 return PGM_TRANSLATION_SPEC;
742 dat_protection |= pte.p;
743 raddr.pfra = pte.pfra;
744real_address:
745 raddr.addr = kvm_s390_real_to_abs(vcpu, raddr.addr);
746absolute_address:
747 if (mode == GACC_STORE && dat_protection)
748 return PGM_PROTECTION;
749 if (kvm_is_error_gpa(vcpu->kvm, raddr.addr))
750 return PGM_ADDRESSING;
751 *gpa = raddr.addr;
752 return 0;
753}
754
755static inline int is_low_address(unsigned long ga)
756{
757 /* Check for address ranges 0..511 and 4096..4607 */
758 return (ga & ~0x11fful) == 0;
759}
760
761static int low_address_protection_enabled(struct kvm_vcpu *vcpu,
762 const union asce asce)
763{
764 union ctlreg0 ctlreg0 = {.val = vcpu->arch.sie_block->gcr[0]};
765 psw_t *psw = &vcpu->arch.sie_block->gpsw;
766
767 if (!ctlreg0.lap)
768 return 0;
769 if (psw_bits(*psw).t && asce.p)
770 return 0;
771 return 1;
772}
773
774static int guest_page_range(struct kvm_vcpu *vcpu, unsigned long ga, ar_t ar,
775 unsigned long *pages, unsigned long nr_pages,
776 const union asce asce, enum gacc_mode mode)
777{
778 psw_t *psw = &vcpu->arch.sie_block->gpsw;
779 int lap_enabled, rc = 0;
780
781 lap_enabled = low_address_protection_enabled(vcpu, asce);
782 while (nr_pages) {
783 ga = kvm_s390_logical_to_effective(vcpu, ga);
784 if (mode == GACC_STORE && lap_enabled && is_low_address(ga))
785 return trans_exc(vcpu, PGM_PROTECTION, ga, ar, mode,
786 PROT_TYPE_LA);
787 ga &= PAGE_MASK;
788 if (psw_bits(*psw).t) {
789 rc = guest_translate(vcpu, ga, pages, asce, mode);
790 if (rc < 0)
791 return rc;
792 } else {
793 *pages = kvm_s390_real_to_abs(vcpu, ga);
794 if (kvm_is_error_gpa(vcpu->kvm, *pages))
795 rc = PGM_ADDRESSING;
796 }
797 if (rc)
798 return trans_exc(vcpu, rc, ga, ar, mode, PROT_TYPE_DAT);
799 ga += PAGE_SIZE;
800 pages++;
801 nr_pages--;
802 }
803 return 0;
804}
805
806int access_guest(struct kvm_vcpu *vcpu, unsigned long ga, ar_t ar, void *data,
807 unsigned long len, enum gacc_mode mode)
808{
809 psw_t *psw = &vcpu->arch.sie_block->gpsw;
810 unsigned long _len, nr_pages, gpa, idx;
811 unsigned long pages_array[2];
812 unsigned long *pages;
813 int need_ipte_lock;
814 union asce asce;
815 int rc;
816
817 if (!len)
818 return 0;
819 ga = kvm_s390_logical_to_effective(vcpu, ga);
820 rc = get_vcpu_asce(vcpu, &asce, ga, ar, mode);
821 if (rc)
822 return rc;
823 nr_pages = (((ga & ~PAGE_MASK) + len - 1) >> PAGE_SHIFT) + 1;
824 pages = pages_array;
825 if (nr_pages > ARRAY_SIZE(pages_array))
826 pages = vmalloc(nr_pages * sizeof(unsigned long));
827 if (!pages)
828 return -ENOMEM;
829 need_ipte_lock = psw_bits(*psw).t && !asce.r;
830 if (need_ipte_lock)
831 ipte_lock(vcpu);
832 rc = guest_page_range(vcpu, ga, ar, pages, nr_pages, asce, mode);
833 for (idx = 0; idx < nr_pages && !rc; idx++) {
834 gpa = *(pages + idx) + (ga & ~PAGE_MASK);
835 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
836 if (mode == GACC_STORE)
837 rc = kvm_write_guest(vcpu->kvm, gpa, data, _len);
838 else
839 rc = kvm_read_guest(vcpu->kvm, gpa, data, _len);
840 len -= _len;
841 ga += _len;
842 data += _len;
843 }
844 if (need_ipte_lock)
845 ipte_unlock(vcpu);
846 if (nr_pages > ARRAY_SIZE(pages_array))
847 vfree(pages);
848 return rc;
849}
850
851int access_guest_real(struct kvm_vcpu *vcpu, unsigned long gra,
852 void *data, unsigned long len, enum gacc_mode mode)
853{
854 unsigned long _len, gpa;
855 int rc = 0;
856
857 while (len && !rc) {
858 gpa = kvm_s390_real_to_abs(vcpu, gra);
859 _len = min(PAGE_SIZE - (gpa & ~PAGE_MASK), len);
860 if (mode)
861 rc = write_guest_abs(vcpu, gpa, data, _len);
862 else
863 rc = read_guest_abs(vcpu, gpa, data, _len);
864 len -= _len;
865 gra += _len;
866 data += _len;
867 }
868 return rc;
869}
870
871/**
872 * guest_translate_address - translate guest logical into guest absolute address
873 *
874 * Parameter semantics are the same as the ones from guest_translate.
875 * The memory contents at the guest address are not changed.
876 *
877 * Note: The IPTE lock is not taken during this function, so the caller
878 * has to take care of this.
879 */
880int guest_translate_address(struct kvm_vcpu *vcpu, unsigned long gva, ar_t ar,
881 unsigned long *gpa, enum gacc_mode mode)
882{
883 psw_t *psw = &vcpu->arch.sie_block->gpsw;
884 union asce asce;
885 int rc;
886
887 gva = kvm_s390_logical_to_effective(vcpu, gva);
888 rc = get_vcpu_asce(vcpu, &asce, gva, ar, mode);
889 if (rc)
890 return rc;
891 if (is_low_address(gva) && low_address_protection_enabled(vcpu, asce)) {
892 if (mode == GACC_STORE)
893 return trans_exc(vcpu, PGM_PROTECTION, gva, 0,
894 mode, PROT_TYPE_LA);
895 }
896
897 if (psw_bits(*psw).t && !asce.r) { /* Use DAT? */
898 rc = guest_translate(vcpu, gva, gpa, asce, mode);
899 if (rc > 0)
900 return trans_exc(vcpu, rc, gva, 0, mode, PROT_TYPE_DAT);
901 } else {
902 *gpa = kvm_s390_real_to_abs(vcpu, gva);
903 if (kvm_is_error_gpa(vcpu->kvm, *gpa))
904 return trans_exc(vcpu, rc, gva, PGM_ADDRESSING, mode, 0);
905 }
906
907 return rc;
908}
909
910/**
911 * check_gva_range - test a range of guest virtual addresses for accessibility
912 */
913int check_gva_range(struct kvm_vcpu *vcpu, unsigned long gva, ar_t ar,
914 unsigned long length, enum gacc_mode mode)
915{
916 unsigned long gpa;
917 unsigned long currlen;
918 int rc = 0;
919
920 ipte_lock(vcpu);
921 while (length > 0 && !rc) {
922 currlen = min(length, PAGE_SIZE - (gva % PAGE_SIZE));
923 rc = guest_translate_address(vcpu, gva, ar, &gpa, mode);
924 gva += currlen;
925 length -= currlen;
926 }
927 ipte_unlock(vcpu);
928
929 return rc;
930}
931
932/**
933 * kvm_s390_check_low_addr_prot_real - check for low-address protection
934 * @gra: Guest real address
935 *
936 * Checks whether an address is subject to low-address protection and set
937 * up vcpu->arch.pgm accordingly if necessary.
938 *
939 * Return: 0 if no protection exception, or PGM_PROTECTION if protected.
940 */
941int kvm_s390_check_low_addr_prot_real(struct kvm_vcpu *vcpu, unsigned long gra)
942{
943 union ctlreg0 ctlreg0 = {.val = vcpu->arch.sie_block->gcr[0]};
944
945 if (!ctlreg0.lap || !is_low_address(gra))
946 return 0;
947 return trans_exc(vcpu, PGM_PROTECTION, gra, 0, GACC_STORE, PROT_TYPE_LA);
948}
949
950/**
951 * kvm_s390_shadow_tables - walk the guest page table and create shadow tables
952 * @sg: pointer to the shadow guest address space structure
953 * @saddr: faulting address in the shadow gmap
954 * @pgt: pointer to the page table address result
955 * @fake: pgt references contiguous guest memory block, not a pgtable
956 */
957static int kvm_s390_shadow_tables(struct gmap *sg, unsigned long saddr,
958 unsigned long *pgt, int *dat_protection,
959 int *fake)
960{
961 struct gmap *parent;
962 union asce asce;
963 union vaddress vaddr;
964 unsigned long ptr;
965 int rc;
966
967 *fake = 0;
968 *dat_protection = 0;
969 parent = sg->parent;
970 vaddr.addr = saddr;
971 asce.val = sg->orig_asce;
972 ptr = asce.origin * 4096;
973 if (asce.r) {
974 *fake = 1;
975 asce.dt = ASCE_TYPE_REGION1;
976 }
977 switch (asce.dt) {
978 case ASCE_TYPE_REGION1:
979 if (vaddr.rfx01 > asce.tl && !asce.r)
980 return PGM_REGION_FIRST_TRANS;
981 break;
982 case ASCE_TYPE_REGION2:
983 if (vaddr.rfx)
984 return PGM_ASCE_TYPE;
985 if (vaddr.rsx01 > asce.tl)
986 return PGM_REGION_SECOND_TRANS;
987 break;
988 case ASCE_TYPE_REGION3:
989 if (vaddr.rfx || vaddr.rsx)
990 return PGM_ASCE_TYPE;
991 if (vaddr.rtx01 > asce.tl)
992 return PGM_REGION_THIRD_TRANS;
993 break;
994 case ASCE_TYPE_SEGMENT:
995 if (vaddr.rfx || vaddr.rsx || vaddr.rtx)
996 return PGM_ASCE_TYPE;
997 if (vaddr.sx01 > asce.tl)
998 return PGM_SEGMENT_TRANSLATION;
999 break;
1000 }
1001
1002 switch (asce.dt) {
1003 case ASCE_TYPE_REGION1: {
1004 union region1_table_entry rfte;
1005
1006 if (*fake) {
1007 /* offset in 16EB guest memory block */
1008 ptr = ptr + ((unsigned long) vaddr.rsx << 53UL);
1009 rfte.val = ptr;
1010 goto shadow_r2t;
1011 }
1012 rc = gmap_read_table(parent, ptr + vaddr.rfx * 8, &rfte.val);
1013 if (rc)
1014 return rc;
1015 if (rfte.i)
1016 return PGM_REGION_FIRST_TRANS;
1017 if (rfte.tt != TABLE_TYPE_REGION1)
1018 return PGM_TRANSLATION_SPEC;
1019 if (vaddr.rsx01 < rfte.tf || vaddr.rsx01 > rfte.tl)
1020 return PGM_REGION_SECOND_TRANS;
1021 if (sg->edat_level >= 1)
1022 *dat_protection |= rfte.p;
1023 ptr = rfte.rto << 12UL;
1024shadow_r2t:
1025 rc = gmap_shadow_r2t(sg, saddr, rfte.val, *fake);
1026 if (rc)
1027 return rc;
1028 /* fallthrough */
1029 }
1030 case ASCE_TYPE_REGION2: {
1031 union region2_table_entry rste;
1032
1033 if (*fake) {
1034 /* offset in 8PB guest memory block */
1035 ptr = ptr + ((unsigned long) vaddr.rtx << 42UL);
1036 rste.val = ptr;
1037 goto shadow_r3t;
1038 }
1039 rc = gmap_read_table(parent, ptr + vaddr.rsx * 8, &rste.val);
1040 if (rc)
1041 return rc;
1042 if (rste.i)
1043 return PGM_REGION_SECOND_TRANS;
1044 if (rste.tt != TABLE_TYPE_REGION2)
1045 return PGM_TRANSLATION_SPEC;
1046 if (vaddr.rtx01 < rste.tf || vaddr.rtx01 > rste.tl)
1047 return PGM_REGION_THIRD_TRANS;
1048 if (sg->edat_level >= 1)
1049 *dat_protection |= rste.p;
1050 ptr = rste.rto << 12UL;
1051shadow_r3t:
1052 rste.p |= *dat_protection;
1053 rc = gmap_shadow_r3t(sg, saddr, rste.val, *fake);
1054 if (rc)
1055 return rc;
1056 /* fallthrough */
1057 }
1058 case ASCE_TYPE_REGION3: {
1059 union region3_table_entry rtte;
1060
1061 if (*fake) {
1062 /* offset in 4TB guest memory block */
1063 ptr = ptr + ((unsigned long) vaddr.sx << 31UL);
1064 rtte.val = ptr;
1065 goto shadow_sgt;
1066 }
1067 rc = gmap_read_table(parent, ptr + vaddr.rtx * 8, &rtte.val);
1068 if (rc)
1069 return rc;
1070 if (rtte.i)
1071 return PGM_REGION_THIRD_TRANS;
1072 if (rtte.tt != TABLE_TYPE_REGION3)
1073 return PGM_TRANSLATION_SPEC;
1074 if (rtte.cr && asce.p && sg->edat_level >= 2)
1075 return PGM_TRANSLATION_SPEC;
1076 if (rtte.fc && sg->edat_level >= 2) {
1077 *dat_protection |= rtte.fc0.p;
1078 *fake = 1;
1079 ptr = rtte.fc1.rfaa << 31UL;
1080 rtte.val = ptr;
1081 goto shadow_sgt;
1082 }
1083 if (vaddr.sx01 < rtte.fc0.tf || vaddr.sx01 > rtte.fc0.tl)
1084 return PGM_SEGMENT_TRANSLATION;
1085 if (sg->edat_level >= 1)
1086 *dat_protection |= rtte.fc0.p;
1087 ptr = rtte.fc0.sto << 12UL;
1088shadow_sgt:
1089 rtte.fc0.p |= *dat_protection;
1090 rc = gmap_shadow_sgt(sg, saddr, rtte.val, *fake);
1091 if (rc)
1092 return rc;
1093 /* fallthrough */
1094 }
1095 case ASCE_TYPE_SEGMENT: {
1096 union segment_table_entry ste;
1097
1098 if (*fake) {
1099 /* offset in 2G guest memory block */
1100 ptr = ptr + ((unsigned long) vaddr.sx << 20UL);
1101 ste.val = ptr;
1102 goto shadow_pgt;
1103 }
1104 rc = gmap_read_table(parent, ptr + vaddr.sx * 8, &ste.val);
1105 if (rc)
1106 return rc;
1107 if (ste.i)
1108 return PGM_SEGMENT_TRANSLATION;
1109 if (ste.tt != TABLE_TYPE_SEGMENT)
1110 return PGM_TRANSLATION_SPEC;
1111 if (ste.cs && asce.p)
1112 return PGM_TRANSLATION_SPEC;
1113 *dat_protection |= ste.fc0.p;
1114 if (ste.fc && sg->edat_level >= 1) {
1115 *fake = 1;
1116 ptr = ste.fc1.sfaa << 20UL;
1117 ste.val = ptr;
1118 goto shadow_pgt;
1119 }
1120 ptr = ste.fc0.pto << 11UL;
1121shadow_pgt:
1122 ste.fc0.p |= *dat_protection;
1123 rc = gmap_shadow_pgt(sg, saddr, ste.val, *fake);
1124 if (rc)
1125 return rc;
1126 }
1127 }
1128 /* Return the parent address of the page table */
1129 *pgt = ptr;
1130 return 0;
1131}
1132
1133/**
1134 * kvm_s390_shadow_fault - handle fault on a shadow page table
1135 * @vcpu: virtual cpu
1136 * @sg: pointer to the shadow guest address space structure
1137 * @saddr: faulting address in the shadow gmap
1138 *
1139 * Returns: - 0 if the shadow fault was successfully resolved
1140 * - > 0 (pgm exception code) on exceptions while faulting
1141 * - -EAGAIN if the caller can retry immediately
1142 * - -EFAULT when accessing invalid guest addresses
1143 * - -ENOMEM if out of memory
1144 */
1145int kvm_s390_shadow_fault(struct kvm_vcpu *vcpu, struct gmap *sg,
1146 unsigned long saddr)
1147{
1148 union vaddress vaddr;
1149 union page_table_entry pte;
1150 unsigned long pgt;
1151 int dat_protection, fake;
1152 int rc;
1153
1154 down_read(&sg->mm->mmap_sem);
1155 /*
1156 * We don't want any guest-2 tables to change - so the parent
1157 * tables/pointers we read stay valid - unshadowing is however
1158 * always possible - only guest_table_lock protects us.
1159 */
1160 ipte_lock(vcpu);
1161
1162 rc = gmap_shadow_pgt_lookup(sg, saddr, &pgt, &dat_protection, &fake);
1163 if (rc)
1164 rc = kvm_s390_shadow_tables(sg, saddr, &pgt, &dat_protection,
1165 &fake);
1166
1167 vaddr.addr = saddr;
1168 if (fake) {
1169 /* offset in 1MB guest memory block */
1170 pte.val = pgt + ((unsigned long) vaddr.px << 12UL);
1171 goto shadow_page;
1172 }
1173 if (!rc)
1174 rc = gmap_read_table(sg->parent, pgt + vaddr.px * 8, &pte.val);
1175 if (!rc && pte.i)
1176 rc = PGM_PAGE_TRANSLATION;
1177 if (!rc && (pte.z || (pte.co && sg->edat_level < 1)))
1178 rc = PGM_TRANSLATION_SPEC;
1179shadow_page:
1180 pte.p |= dat_protection;
1181 if (!rc)
1182 rc = gmap_shadow_page(sg, saddr, __pte(pte.val));
1183 ipte_unlock(vcpu);
1184 up_read(&sg->mm->mmap_sem);
1185 return rc;
1186}