1 // SPDX-License-Identifier: GPL-2.0
3 * guest access functions
5 * Copyright IBM Corp. 2014
9 #include <linux/vmalloc.h>
10 #include <linux/mm_types.h>
11 #include <linux/err.h>
12 #include <linux/pgtable.h>
13 #include <linux/bitfield.h>
18 #include <asm/switch_to.h>
23 unsigned long origin
: 52; /* Region- or Segment-Table Origin */
25 unsigned long g
: 1; /* Subspace Group Control */
26 unsigned long p
: 1; /* Private Space Control */
27 unsigned long s
: 1; /* Storage-Alteration-Event Control */
28 unsigned long x
: 1; /* Space-Switch-Event Control */
29 unsigned long r
: 1; /* Real-Space Control */
31 unsigned long dt
: 2; /* Designation-Type Control */
32 unsigned long tl
: 2; /* Region- or Segment-Table Length */
37 ASCE_TYPE_SEGMENT
= 0,
38 ASCE_TYPE_REGION3
= 1,
39 ASCE_TYPE_REGION2
= 2,
43 union region1_table_entry
{
46 unsigned long rto
: 52;/* Region-Table Origin */
48 unsigned long p
: 1; /* DAT-Protection Bit */
50 unsigned long tf
: 2; /* Region-Second-Table Offset */
51 unsigned long i
: 1; /* Region-Invalid Bit */
53 unsigned long tt
: 2; /* Table-Type Bits */
54 unsigned long tl
: 2; /* Region-Second-Table Length */
58 union region2_table_entry
{
61 unsigned long rto
: 52;/* Region-Table Origin */
63 unsigned long p
: 1; /* DAT-Protection Bit */
65 unsigned long tf
: 2; /* Region-Third-Table Offset */
66 unsigned long i
: 1; /* Region-Invalid Bit */
68 unsigned long tt
: 2; /* Table-Type Bits */
69 unsigned long tl
: 2; /* Region-Third-Table Length */
73 struct region3_table_entry_fc0
{
74 unsigned long sto
: 52;/* Segment-Table Origin */
76 unsigned long fc
: 1; /* Format-Control */
77 unsigned long p
: 1; /* DAT-Protection Bit */
79 unsigned long tf
: 2; /* Segment-Table Offset */
80 unsigned long i
: 1; /* Region-Invalid Bit */
81 unsigned long cr
: 1; /* Common-Region Bit */
82 unsigned long tt
: 2; /* Table-Type Bits */
83 unsigned long tl
: 2; /* Segment-Table Length */
86 struct region3_table_entry_fc1
{
87 unsigned long rfaa
: 33; /* Region-Frame Absolute Address */
89 unsigned long av
: 1; /* ACCF-Validity Control */
90 unsigned long acc
: 4; /* Access-Control Bits */
91 unsigned long f
: 1; /* Fetch-Protection Bit */
92 unsigned long fc
: 1; /* Format-Control */
93 unsigned long p
: 1; /* DAT-Protection Bit */
94 unsigned long iep
: 1; /* Instruction-Execution-Protection */
96 unsigned long i
: 1; /* Region-Invalid Bit */
97 unsigned long cr
: 1; /* Common-Region Bit */
98 unsigned long tt
: 2; /* Table-Type Bits */
102 union region3_table_entry
{
104 struct region3_table_entry_fc0 fc0
;
105 struct region3_table_entry_fc1 fc1
;
108 unsigned long fc
: 1; /* Format-Control */
110 unsigned long i
: 1; /* Region-Invalid Bit */
111 unsigned long cr
: 1; /* Common-Region Bit */
112 unsigned long tt
: 2; /* Table-Type Bits */
117 struct segment_entry_fc0
{
118 unsigned long pto
: 53;/* Page-Table Origin */
119 unsigned long fc
: 1; /* Format-Control */
120 unsigned long p
: 1; /* DAT-Protection Bit */
122 unsigned long i
: 1; /* Segment-Invalid Bit */
123 unsigned long cs
: 1; /* Common-Segment Bit */
124 unsigned long tt
: 2; /* Table-Type Bits */
128 struct segment_entry_fc1
{
129 unsigned long sfaa
: 44; /* Segment-Frame Absolute Address */
131 unsigned long av
: 1; /* ACCF-Validity Control */
132 unsigned long acc
: 4; /* Access-Control Bits */
133 unsigned long f
: 1; /* Fetch-Protection Bit */
134 unsigned long fc
: 1; /* Format-Control */
135 unsigned long p
: 1; /* DAT-Protection Bit */
136 unsigned long iep
: 1; /* Instruction-Execution-Protection */
138 unsigned long i
: 1; /* Segment-Invalid Bit */
139 unsigned long cs
: 1; /* Common-Segment Bit */
140 unsigned long tt
: 2; /* Table-Type Bits */
144 union segment_table_entry
{
146 struct segment_entry_fc0 fc0
;
147 struct segment_entry_fc1 fc1
;
150 unsigned long fc
: 1; /* Format-Control */
152 unsigned long i
: 1; /* Segment-Invalid Bit */
153 unsigned long cs
: 1; /* Common-Segment Bit */
154 unsigned long tt
: 2; /* Table-Type Bits */
160 TABLE_TYPE_SEGMENT
= 0,
161 TABLE_TYPE_REGION3
= 1,
162 TABLE_TYPE_REGION2
= 2,
163 TABLE_TYPE_REGION1
= 3
166 union page_table_entry
{
169 unsigned long pfra
: 52; /* Page-Frame Real Address */
170 unsigned long z
: 1; /* Zero Bit */
171 unsigned long i
: 1; /* Page-Invalid Bit */
172 unsigned long p
: 1; /* DAT-Protection Bit */
173 unsigned long iep
: 1; /* Instruction-Execution-Protection */
179 * vaddress union in order to easily decode a virtual address into its
180 * region first index, region second index etc. parts.
185 unsigned long rfx
: 11;
186 unsigned long rsx
: 11;
187 unsigned long rtx
: 11;
188 unsigned long sx
: 11;
189 unsigned long px
: 8;
190 unsigned long bx
: 12;
193 unsigned long rfx01
: 2;
195 unsigned long rsx01
: 2;
197 unsigned long rtx01
: 2;
199 unsigned long sx01
: 2;
205 * raddress union which will contain the result (real or absolute address)
206 * after a page table walk. The rfaa, sfaa and pfra members are used to
207 * simply assign them the value of a region, segment or page table entry.
211 unsigned long rfaa
: 33; /* Region-Frame Absolute Address */
212 unsigned long sfaa
: 44; /* Segment-Frame Absolute Address */
213 unsigned long pfra
: 52; /* Page-Frame Real Address */
236 unsigned long i
: 1; /* ALEN-Invalid Bit */
238 unsigned long fo
: 1; /* Fetch-Only Bit */
239 unsigned long p
: 1; /* Private Bit */
240 unsigned long alesn
: 8; /* Access-List-Entry Sequence Number */
241 unsigned long aleax
: 16; /* Access-List-Entry Authorization Index */
244 unsigned long asteo
: 25; /* ASN-Second-Table-Entry Origin */
246 unsigned long astesn
: 32; /* ASTE Sequence Number */
250 unsigned long i
: 1; /* ASX-Invalid Bit */
251 unsigned long ato
: 29; /* Authority-Table Origin */
253 unsigned long b
: 1; /* Base-Space Bit */
254 unsigned long ax
: 16; /* Authorization Index */
255 unsigned long atl
: 12; /* Authority-Table Length */
257 unsigned long ca
: 1; /* Controlled-ASN Bit */
258 unsigned long ra
: 1; /* Reusable-ASN Bit */
259 unsigned long asce
: 64; /* Address-Space-Control Element */
260 unsigned long ald
: 32;
261 unsigned long astesn
: 32;
262 /* .. more fields there */
265 int ipte_lock_held(struct kvm
*kvm
)
270 read_lock(&kvm
->arch
.sca_lock
);
271 rc
= kvm_s390_get_ipte_control(kvm
)->kh
!= 0;
272 read_unlock(&kvm
->arch
.sca_lock
);
275 return kvm
->arch
.ipte_lock_count
!= 0;
278 static void ipte_lock_simple(struct kvm
*kvm
)
280 union ipte_control old
, new, *ic
;
282 mutex_lock(&kvm
->arch
.ipte_mutex
);
283 kvm
->arch
.ipte_lock_count
++;
284 if (kvm
->arch
.ipte_lock_count
> 1)
287 read_lock(&kvm
->arch
.sca_lock
);
288 ic
= kvm_s390_get_ipte_control(kvm
);
290 old
= READ_ONCE(*ic
);
292 read_unlock(&kvm
->arch
.sca_lock
);
298 } while (cmpxchg(&ic
->val
, old
.val
, new.val
) != old
.val
);
299 read_unlock(&kvm
->arch
.sca_lock
);
301 mutex_unlock(&kvm
->arch
.ipte_mutex
);
304 static void ipte_unlock_simple(struct kvm
*kvm
)
306 union ipte_control old
, new, *ic
;
308 mutex_lock(&kvm
->arch
.ipte_mutex
);
309 kvm
->arch
.ipte_lock_count
--;
310 if (kvm
->arch
.ipte_lock_count
)
312 read_lock(&kvm
->arch
.sca_lock
);
313 ic
= kvm_s390_get_ipte_control(kvm
);
315 old
= READ_ONCE(*ic
);
318 } while (cmpxchg(&ic
->val
, old
.val
, new.val
) != old
.val
);
319 read_unlock(&kvm
->arch
.sca_lock
);
320 wake_up(&kvm
->arch
.ipte_wq
);
322 mutex_unlock(&kvm
->arch
.ipte_mutex
);
325 static void ipte_lock_siif(struct kvm
*kvm
)
327 union ipte_control old
, new, *ic
;
330 read_lock(&kvm
->arch
.sca_lock
);
331 ic
= kvm_s390_get_ipte_control(kvm
);
333 old
= READ_ONCE(*ic
);
335 read_unlock(&kvm
->arch
.sca_lock
);
342 } while (cmpxchg(&ic
->val
, old
.val
, new.val
) != old
.val
);
343 read_unlock(&kvm
->arch
.sca_lock
);
346 static void ipte_unlock_siif(struct kvm
*kvm
)
348 union ipte_control old
, new, *ic
;
350 read_lock(&kvm
->arch
.sca_lock
);
351 ic
= kvm_s390_get_ipte_control(kvm
);
353 old
= READ_ONCE(*ic
);
358 } while (cmpxchg(&ic
->val
, old
.val
, new.val
) != old
.val
);
359 read_unlock(&kvm
->arch
.sca_lock
);
361 wake_up(&kvm
->arch
.ipte_wq
);
364 void ipte_lock(struct kvm
*kvm
)
369 ipte_lock_simple(kvm
);
372 void ipte_unlock(struct kvm
*kvm
)
375 ipte_unlock_siif(kvm
);
377 ipte_unlock_simple(kvm
);
380 static int ar_translation(struct kvm_vcpu
*vcpu
, union asce
*asce
, u8 ar
,
386 unsigned long ald_addr
, authority_table_addr
;
394 save_access_regs(vcpu
->run
->s
.regs
.acrs
);
395 alet
.val
= vcpu
->run
->s
.regs
.acrs
[ar
];
397 if (ar
== 0 || alet
.val
== 0) {
398 asce
->val
= vcpu
->arch
.sie_block
->gcr
[1];
400 } else if (alet
.val
== 1) {
401 asce
->val
= vcpu
->arch
.sie_block
->gcr
[7];
406 return PGM_ALET_SPECIFICATION
;
409 ald_addr
= vcpu
->arch
.sie_block
->gcr
[5];
411 ald_addr
= vcpu
->arch
.sie_block
->gcr
[2];
412 ald_addr
&= 0x7fffffc0;
414 rc
= read_guest_real(vcpu
, ald_addr
+ 16, &ald
.val
, sizeof(union ald
));
418 if (alet
.alen
/ 8 > ald
.all
)
419 return PGM_ALEN_TRANSLATION
;
421 if (0x7fffffff - ald
.alo
* 128 < alet
.alen
* 16)
422 return PGM_ADDRESSING
;
424 rc
= read_guest_real(vcpu
, ald
.alo
* 128 + alet
.alen
* 16, &ale
,
430 return PGM_ALEN_TRANSLATION
;
431 if (ale
.alesn
!= alet
.alesn
)
432 return PGM_ALE_SEQUENCE
;
434 rc
= read_guest_real(vcpu
, ale
.asteo
* 64, &aste
, sizeof(struct aste
));
439 return PGM_ASTE_VALIDITY
;
440 if (aste
.astesn
!= ale
.astesn
)
441 return PGM_ASTE_SEQUENCE
;
444 eax
= (vcpu
->arch
.sie_block
->gcr
[8] >> 16) & 0xffff;
445 if (ale
.aleax
!= eax
) {
446 if (eax
/ 16 > aste
.atl
)
447 return PGM_EXTENDED_AUTHORITY
;
449 authority_table_addr
= aste
.ato
* 4 + eax
/ 4;
451 rc
= read_guest_real(vcpu
, authority_table_addr
,
457 if ((authority_table
& (0x40 >> ((eax
& 3) * 2))) == 0)
458 return PGM_EXTENDED_AUTHORITY
;
462 if (ale
.fo
== 1 && mode
== GACC_STORE
)
463 return PGM_PROTECTION
;
465 asce
->val
= aste
.asce
;
469 struct trans_exc_code_bits
{
470 unsigned long addr
: 52; /* Translation-exception Address */
471 unsigned long fsi
: 2; /* Access Exception Fetch/Store Indication */
473 unsigned long b56
: 1;
475 unsigned long b60
: 1;
476 unsigned long b61
: 1;
477 unsigned long as
: 2; /* ASCE Identifier */
481 FSI_UNKNOWN
= 0, /* Unknown whether fetch or store */
482 FSI_STORE
= 1, /* Exception was due to store operation */
483 FSI_FETCH
= 2 /* Exception was due to fetch operation */
492 /* Dummy value for passing an initialized value when code != PGM_PROTECTION */
496 static int trans_exc_ending(struct kvm_vcpu
*vcpu
, int code
, unsigned long gva
, u8 ar
,
497 enum gacc_mode mode
, enum prot_type prot
, bool terminate
)
499 struct kvm_s390_pgm_info
*pgm
= &vcpu
->arch
.pgm
;
500 struct trans_exc_code_bits
*tec
;
502 memset(pgm
, 0, sizeof(*pgm
));
504 tec
= (struct trans_exc_code_bits
*)&pgm
->trans_exc_code
;
510 /* We should never get here, acts like termination */
536 case PGM_PAGE_TRANSLATION
:
537 case PGM_REGION_FIRST_TRANS
:
538 case PGM_REGION_SECOND_TRANS
:
539 case PGM_REGION_THIRD_TRANS
:
540 case PGM_SEGMENT_TRANSLATION
:
542 * op_access_id only applies to MOVE_PAGE -> set bit 61
543 * exc_access_id has to be set to 0 for some instructions. Both
544 * cases have to be handled by the caller.
546 tec
->addr
= gva
>> PAGE_SHIFT
;
547 tec
->fsi
= mode
== GACC_STORE
? FSI_STORE
: FSI_FETCH
;
548 tec
->as
= psw_bits(vcpu
->arch
.sie_block
->gpsw
).as
;
550 case PGM_ALEN_TRANSLATION
:
551 case PGM_ALE_SEQUENCE
:
552 case PGM_ASTE_VALIDITY
:
553 case PGM_ASTE_SEQUENCE
:
554 case PGM_EXTENDED_AUTHORITY
:
556 * We can always store exc_access_id, as it is
557 * undefined for non-ar cases. It is undefined for
558 * most DAT protection exceptions.
560 pgm
->exc_access_id
= ar
;
566 static int trans_exc(struct kvm_vcpu
*vcpu
, int code
, unsigned long gva
, u8 ar
,
567 enum gacc_mode mode
, enum prot_type prot
)
569 return trans_exc_ending(vcpu
, code
, gva
, ar
, mode
, prot
, false);
572 static int get_vcpu_asce(struct kvm_vcpu
*vcpu
, union asce
*asce
,
573 unsigned long ga
, u8 ar
, enum gacc_mode mode
)
576 struct psw_bits psw
= psw_bits(vcpu
->arch
.sie_block
->gpsw
);
584 if ((mode
== GACC_IFETCH
) && (psw
.as
!= PSW_BITS_AS_HOME
))
585 psw
.as
= PSW_BITS_AS_PRIMARY
;
588 case PSW_BITS_AS_PRIMARY
:
589 asce
->val
= vcpu
->arch
.sie_block
->gcr
[1];
591 case PSW_BITS_AS_SECONDARY
:
592 asce
->val
= vcpu
->arch
.sie_block
->gcr
[7];
594 case PSW_BITS_AS_HOME
:
595 asce
->val
= vcpu
->arch
.sie_block
->gcr
[13];
597 case PSW_BITS_AS_ACCREG
:
598 rc
= ar_translation(vcpu
, asce
, ar
, mode
);
600 return trans_exc(vcpu
, rc
, ga
, ar
, mode
, PROT_TYPE_ALC
);
606 static int deref_table(struct kvm
*kvm
, unsigned long gpa
, unsigned long *val
)
608 return kvm_read_guest(kvm
, gpa
, val
, sizeof(*val
));
612 * guest_translate - translate a guest virtual into a guest absolute address
614 * @gva: guest virtual address
615 * @gpa: points to where guest physical (absolute) address should be stored
616 * @asce: effective asce
617 * @mode: indicates the access mode to be used
618 * @prot: returns the type for protection exceptions
620 * Translate a guest virtual address into a guest absolute address by means
621 * of dynamic address translation as specified by the architecture.
622 * If the resulting absolute address is not available in the configuration
623 * an addressing exception is indicated and @gpa will not be changed.
625 * Returns: - zero on success; @gpa contains the resulting absolute address
626 * - a negative value if guest access failed due to e.g. broken
628 * - a positive value if an access exception happened. In this case
629 * the returned value is the program interruption code as defined
630 * by the architecture
632 static unsigned long guest_translate(struct kvm_vcpu
*vcpu
, unsigned long gva
,
633 unsigned long *gpa
, const union asce asce
,
634 enum gacc_mode mode
, enum prot_type
*prot
)
636 union vaddress vaddr
= {.addr
= gva
};
637 union raddress raddr
= {.addr
= gva
};
638 union page_table_entry pte
;
639 int dat_protection
= 0;
640 int iep_protection
= 0;
641 union ctlreg0 ctlreg0
;
643 int edat1
, edat2
, iep
;
645 ctlreg0
.val
= vcpu
->arch
.sie_block
->gcr
[0];
646 edat1
= ctlreg0
.edat
&& test_kvm_facility(vcpu
->kvm
, 8);
647 edat2
= edat1
&& test_kvm_facility(vcpu
->kvm
, 78);
648 iep
= ctlreg0
.iep
&& test_kvm_facility(vcpu
->kvm
, 130);
651 ptr
= asce
.origin
* PAGE_SIZE
;
653 case ASCE_TYPE_REGION1
:
654 if (vaddr
.rfx01
> asce
.tl
)
655 return PGM_REGION_FIRST_TRANS
;
656 ptr
+= vaddr
.rfx
* 8;
658 case ASCE_TYPE_REGION2
:
660 return PGM_ASCE_TYPE
;
661 if (vaddr
.rsx01
> asce
.tl
)
662 return PGM_REGION_SECOND_TRANS
;
663 ptr
+= vaddr
.rsx
* 8;
665 case ASCE_TYPE_REGION3
:
666 if (vaddr
.rfx
|| vaddr
.rsx
)
667 return PGM_ASCE_TYPE
;
668 if (vaddr
.rtx01
> asce
.tl
)
669 return PGM_REGION_THIRD_TRANS
;
670 ptr
+= vaddr
.rtx
* 8;
672 case ASCE_TYPE_SEGMENT
:
673 if (vaddr
.rfx
|| vaddr
.rsx
|| vaddr
.rtx
)
674 return PGM_ASCE_TYPE
;
675 if (vaddr
.sx01
> asce
.tl
)
676 return PGM_SEGMENT_TRANSLATION
;
681 case ASCE_TYPE_REGION1
: {
682 union region1_table_entry rfte
;
684 if (kvm_is_error_gpa(vcpu
->kvm
, ptr
))
685 return PGM_ADDRESSING
;
686 if (deref_table(vcpu
->kvm
, ptr
, &rfte
.val
))
689 return PGM_REGION_FIRST_TRANS
;
690 if (rfte
.tt
!= TABLE_TYPE_REGION1
)
691 return PGM_TRANSLATION_SPEC
;
692 if (vaddr
.rsx01
< rfte
.tf
|| vaddr
.rsx01
> rfte
.tl
)
693 return PGM_REGION_SECOND_TRANS
;
695 dat_protection
|= rfte
.p
;
696 ptr
= rfte
.rto
* PAGE_SIZE
+ vaddr
.rsx
* 8;
699 case ASCE_TYPE_REGION2
: {
700 union region2_table_entry rste
;
702 if (kvm_is_error_gpa(vcpu
->kvm
, ptr
))
703 return PGM_ADDRESSING
;
704 if (deref_table(vcpu
->kvm
, ptr
, &rste
.val
))
707 return PGM_REGION_SECOND_TRANS
;
708 if (rste
.tt
!= TABLE_TYPE_REGION2
)
709 return PGM_TRANSLATION_SPEC
;
710 if (vaddr
.rtx01
< rste
.tf
|| vaddr
.rtx01
> rste
.tl
)
711 return PGM_REGION_THIRD_TRANS
;
713 dat_protection
|= rste
.p
;
714 ptr
= rste
.rto
* PAGE_SIZE
+ vaddr
.rtx
* 8;
717 case ASCE_TYPE_REGION3
: {
718 union region3_table_entry rtte
;
720 if (kvm_is_error_gpa(vcpu
->kvm
, ptr
))
721 return PGM_ADDRESSING
;
722 if (deref_table(vcpu
->kvm
, ptr
, &rtte
.val
))
725 return PGM_REGION_THIRD_TRANS
;
726 if (rtte
.tt
!= TABLE_TYPE_REGION3
)
727 return PGM_TRANSLATION_SPEC
;
728 if (rtte
.cr
&& asce
.p
&& edat2
)
729 return PGM_TRANSLATION_SPEC
;
730 if (rtte
.fc
&& edat2
) {
731 dat_protection
|= rtte
.fc1
.p
;
732 iep_protection
= rtte
.fc1
.iep
;
733 raddr
.rfaa
= rtte
.fc1
.rfaa
;
734 goto absolute_address
;
736 if (vaddr
.sx01
< rtte
.fc0
.tf
)
737 return PGM_SEGMENT_TRANSLATION
;
738 if (vaddr
.sx01
> rtte
.fc0
.tl
)
739 return PGM_SEGMENT_TRANSLATION
;
741 dat_protection
|= rtte
.fc0
.p
;
742 ptr
= rtte
.fc0
.sto
* PAGE_SIZE
+ vaddr
.sx
* 8;
745 case ASCE_TYPE_SEGMENT
: {
746 union segment_table_entry ste
;
748 if (kvm_is_error_gpa(vcpu
->kvm
, ptr
))
749 return PGM_ADDRESSING
;
750 if (deref_table(vcpu
->kvm
, ptr
, &ste
.val
))
753 return PGM_SEGMENT_TRANSLATION
;
754 if (ste
.tt
!= TABLE_TYPE_SEGMENT
)
755 return PGM_TRANSLATION_SPEC
;
756 if (ste
.cs
&& asce
.p
)
757 return PGM_TRANSLATION_SPEC
;
758 if (ste
.fc
&& edat1
) {
759 dat_protection
|= ste
.fc1
.p
;
760 iep_protection
= ste
.fc1
.iep
;
761 raddr
.sfaa
= ste
.fc1
.sfaa
;
762 goto absolute_address
;
764 dat_protection
|= ste
.fc0
.p
;
765 ptr
= ste
.fc0
.pto
* (PAGE_SIZE
/ 2) + vaddr
.px
* 8;
768 if (kvm_is_error_gpa(vcpu
->kvm
, ptr
))
769 return PGM_ADDRESSING
;
770 if (deref_table(vcpu
->kvm
, ptr
, &pte
.val
))
773 return PGM_PAGE_TRANSLATION
;
775 return PGM_TRANSLATION_SPEC
;
776 dat_protection
|= pte
.p
;
777 iep_protection
= pte
.iep
;
778 raddr
.pfra
= pte
.pfra
;
780 raddr
.addr
= kvm_s390_real_to_abs(vcpu
, raddr
.addr
);
782 if (mode
== GACC_STORE
&& dat_protection
) {
783 *prot
= PROT_TYPE_DAT
;
784 return PGM_PROTECTION
;
786 if (mode
== GACC_IFETCH
&& iep_protection
&& iep
) {
787 *prot
= PROT_TYPE_IEP
;
788 return PGM_PROTECTION
;
790 if (kvm_is_error_gpa(vcpu
->kvm
, raddr
.addr
))
791 return PGM_ADDRESSING
;
796 static inline int is_low_address(unsigned long ga
)
798 /* Check for address ranges 0..511 and 4096..4607 */
799 return (ga
& ~0x11fful
) == 0;
802 static int low_address_protection_enabled(struct kvm_vcpu
*vcpu
,
803 const union asce asce
)
805 union ctlreg0 ctlreg0
= {.val
= vcpu
->arch
.sie_block
->gcr
[0]};
806 psw_t
*psw
= &vcpu
->arch
.sie_block
->gpsw
;
810 if (psw_bits(*psw
).dat
&& asce
.p
)
815 static int vm_check_access_key(struct kvm
*kvm
, u8 access_key
,
816 enum gacc_mode mode
, gpa_t gpa
)
818 u8 storage_key
, access_control
;
819 bool fetch_protected
;
826 hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
827 if (kvm_is_error_hva(hva
))
828 return PGM_ADDRESSING
;
830 mmap_read_lock(current
->mm
);
831 r
= get_guest_storage_key(current
->mm
, hva
, &storage_key
);
832 mmap_read_unlock(current
->mm
);
835 access_control
= FIELD_GET(_PAGE_ACC_BITS
, storage_key
);
836 if (access_control
== access_key
)
838 fetch_protected
= storage_key
& _PAGE_FP_BIT
;
839 if ((mode
== GACC_FETCH
|| mode
== GACC_IFETCH
) && !fetch_protected
)
841 return PGM_PROTECTION
;
844 static bool fetch_prot_override_applicable(struct kvm_vcpu
*vcpu
, enum gacc_mode mode
,
847 psw_t
*psw
= &vcpu
->arch
.sie_block
->gpsw
;
848 unsigned long override
;
850 if (mode
== GACC_FETCH
|| mode
== GACC_IFETCH
) {
851 /* check if fetch protection override enabled */
852 override
= vcpu
->arch
.sie_block
->gcr
[0];
853 override
&= CR0_FETCH_PROTECTION_OVERRIDE
;
854 /* not applicable if subject to DAT && private space */
855 override
= override
&& !(psw_bits(*psw
).dat
&& asce
.p
);
861 static bool fetch_prot_override_applies(unsigned long ga
, unsigned int len
)
863 return ga
< 2048 && ga
+ len
<= 2048;
866 static bool storage_prot_override_applicable(struct kvm_vcpu
*vcpu
)
868 /* check if storage protection override enabled */
869 return vcpu
->arch
.sie_block
->gcr
[0] & CR0_STORAGE_PROTECTION_OVERRIDE
;
872 static bool storage_prot_override_applies(u8 access_control
)
874 /* matches special storage protection override key (9) -> allow */
875 return access_control
== PAGE_SPO_ACC
;
878 static int vcpu_check_access_key(struct kvm_vcpu
*vcpu
, u8 access_key
,
879 enum gacc_mode mode
, union asce asce
, gpa_t gpa
,
880 unsigned long ga
, unsigned int len
)
882 u8 storage_key
, access_control
;
886 /* access key 0 matches any storage key -> allow */
890 * caller needs to ensure that gfn is accessible, so we can
891 * assume that this cannot fail
893 hva
= gfn_to_hva(vcpu
->kvm
, gpa_to_gfn(gpa
));
894 mmap_read_lock(current
->mm
);
895 r
= get_guest_storage_key(current
->mm
, hva
, &storage_key
);
896 mmap_read_unlock(current
->mm
);
899 access_control
= FIELD_GET(_PAGE_ACC_BITS
, storage_key
);
900 /* access key matches storage key -> allow */
901 if (access_control
== access_key
)
903 if (mode
== GACC_FETCH
|| mode
== GACC_IFETCH
) {
904 /* it is a fetch and fetch protection is off -> allow */
905 if (!(storage_key
& _PAGE_FP_BIT
))
907 if (fetch_prot_override_applicable(vcpu
, mode
, asce
) &&
908 fetch_prot_override_applies(ga
, len
))
911 if (storage_prot_override_applicable(vcpu
) &&
912 storage_prot_override_applies(access_control
))
914 return PGM_PROTECTION
;
918 * guest_range_to_gpas() - Calculate guest physical addresses of page fragments
919 * covering a logical range
921 * @ga: guest address, start of range
922 * @ar: access register
923 * @gpas: output argument, may be NULL
924 * @len: length of range in bytes
925 * @asce: address-space-control element to use for translation
927 * @access_key: access key to mach the range's storage keys against
929 * Translate a logical range to a series of guest absolute addresses,
930 * such that the concatenation of page fragments starting at each gpa make up
932 * The translation is performed as if done by the cpu for the given @asce, @ar,
933 * @mode and state of the @vcpu.
934 * If the translation causes an exception, its program interruption code is
935 * returned and the &struct kvm_s390_pgm_info pgm member of @vcpu is modified
936 * such that a subsequent call to kvm_s390_inject_prog_vcpu() will inject
937 * a correct exception into the guest.
938 * The resulting gpas are stored into @gpas, unless it is NULL.
940 * Note: All fragments except the first one start at the beginning of a page.
941 * When deriving the boundaries of a fragment from a gpa, all but the last
942 * fragment end at the end of the page.
946 * * <0 - translation could not be performed, for example if guest
947 * memory could not be accessed
948 * * >0 - an access exception occurred. In this case the returned value
949 * is the program interruption code and the contents of pgm may
950 * be used to inject an exception into the guest.
952 static int guest_range_to_gpas(struct kvm_vcpu
*vcpu
, unsigned long ga
, u8 ar
,
953 unsigned long *gpas
, unsigned long len
,
954 const union asce asce
, enum gacc_mode mode
,
957 psw_t
*psw
= &vcpu
->arch
.sie_block
->gpsw
;
958 unsigned int offset
= offset_in_page(ga
);
959 unsigned int fragment_len
;
960 int lap_enabled
, rc
= 0;
964 lap_enabled
= low_address_protection_enabled(vcpu
, asce
);
965 while (min(PAGE_SIZE
- offset
, len
) > 0) {
966 fragment_len
= min(PAGE_SIZE
- offset
, len
);
967 ga
= kvm_s390_logical_to_effective(vcpu
, ga
);
968 if (mode
== GACC_STORE
&& lap_enabled
&& is_low_address(ga
))
969 return trans_exc(vcpu
, PGM_PROTECTION
, ga
, ar
, mode
,
971 if (psw_bits(*psw
).dat
) {
972 rc
= guest_translate(vcpu
, ga
, &gpa
, asce
, mode
, &prot
);
976 gpa
= kvm_s390_real_to_abs(vcpu
, ga
);
977 if (kvm_is_error_gpa(vcpu
->kvm
, gpa
)) {
983 return trans_exc(vcpu
, rc
, ga
, ar
, mode
, prot
);
984 rc
= vcpu_check_access_key(vcpu
, access_key
, mode
, asce
, gpa
, ga
,
987 return trans_exc(vcpu
, rc
, ga
, ar
, mode
, PROT_TYPE_KEYC
);
997 static int access_guest_page(struct kvm
*kvm
, enum gacc_mode mode
, gpa_t gpa
,
998 void *data
, unsigned int len
)
1000 const unsigned int offset
= offset_in_page(gpa
);
1001 const gfn_t gfn
= gpa_to_gfn(gpa
);
1004 if (mode
== GACC_STORE
)
1005 rc
= kvm_write_guest_page(kvm
, gfn
, data
, offset
, len
);
1007 rc
= kvm_read_guest_page(kvm
, gfn
, data
, offset
, len
);
1012 access_guest_page_with_key(struct kvm
*kvm
, enum gacc_mode mode
, gpa_t gpa
,
1013 void *data
, unsigned int len
, u8 access_key
)
1015 struct kvm_memory_slot
*slot
;
1021 gfn
= gpa
>> PAGE_SHIFT
;
1022 slot
= gfn_to_memslot(kvm
, gfn
);
1023 hva
= gfn_to_hva_memslot_prot(slot
, gfn
, &writable
);
1025 if (kvm_is_error_hva(hva
))
1026 return PGM_ADDRESSING
;
1028 * Check if it's a ro memslot, even tho that can't occur (they're unsupported).
1029 * Don't try to actually handle that case.
1031 if (!writable
&& mode
== GACC_STORE
)
1033 hva
+= offset_in_page(gpa
);
1034 if (mode
== GACC_STORE
)
1035 rc
= copy_to_user_key((void __user
*)hva
, data
, len
, access_key
);
1037 rc
= copy_from_user_key(data
, (void __user
*)hva
, len
, access_key
);
1039 return PGM_PROTECTION
;
1040 if (mode
== GACC_STORE
)
1041 mark_page_dirty_in_slot(kvm
, slot
, gfn
);
1045 int access_guest_abs_with_key(struct kvm
*kvm
, gpa_t gpa
, void *data
,
1046 unsigned long len
, enum gacc_mode mode
, u8 access_key
)
1048 int offset
= offset_in_page(gpa
);
1052 while (min(PAGE_SIZE
- offset
, len
) > 0) {
1053 fragment_len
= min(PAGE_SIZE
- offset
, len
);
1054 rc
= access_guest_page_with_key(kvm
, mode
, gpa
, data
, fragment_len
, access_key
);
1058 len
-= fragment_len
;
1059 data
+= fragment_len
;
1060 gpa
+= fragment_len
;
1065 int access_guest_with_key(struct kvm_vcpu
*vcpu
, unsigned long ga
, u8 ar
,
1066 void *data
, unsigned long len
, enum gacc_mode mode
,
1069 psw_t
*psw
= &vcpu
->arch
.sie_block
->gpsw
;
1070 unsigned long nr_pages
, idx
;
1071 unsigned long gpa_array
[2];
1072 unsigned int fragment_len
;
1073 unsigned long *gpas
;
1074 enum prot_type prot
;
1077 bool try_storage_prot_override
;
1078 bool try_fetch_prot_override
;
1083 ga
= kvm_s390_logical_to_effective(vcpu
, ga
);
1084 rc
= get_vcpu_asce(vcpu
, &asce
, ga
, ar
, mode
);
1087 nr_pages
= (((ga
& ~PAGE_MASK
) + len
- 1) >> PAGE_SHIFT
) + 1;
1089 if (nr_pages
> ARRAY_SIZE(gpa_array
))
1090 gpas
= vmalloc(array_size(nr_pages
, sizeof(unsigned long)));
1093 try_fetch_prot_override
= fetch_prot_override_applicable(vcpu
, mode
, asce
);
1094 try_storage_prot_override
= storage_prot_override_applicable(vcpu
);
1095 need_ipte_lock
= psw_bits(*psw
).dat
&& !asce
.r
;
1097 ipte_lock(vcpu
->kvm
);
1099 * Since we do the access further down ultimately via a move instruction
1100 * that does key checking and returns an error in case of a protection
1101 * violation, we don't need to do the check during address translation.
1102 * Skip it by passing access key 0, which matches any storage key,
1103 * obviating the need for any further checks. As a result the check is
1104 * handled entirely in hardware on access, we only need to take care to
1105 * forego key protection checking if fetch protection override applies or
1106 * retry with the special key 9 in case of storage protection override.
1108 rc
= guest_range_to_gpas(vcpu
, ga
, ar
, gpas
, len
, asce
, mode
, 0);
1111 for (idx
= 0; idx
< nr_pages
; idx
++) {
1112 fragment_len
= min(PAGE_SIZE
- offset_in_page(gpas
[idx
]), len
);
1113 if (try_fetch_prot_override
&& fetch_prot_override_applies(ga
, fragment_len
)) {
1114 rc
= access_guest_page(vcpu
->kvm
, mode
, gpas
[idx
],
1115 data
, fragment_len
);
1117 rc
= access_guest_page_with_key(vcpu
->kvm
, mode
, gpas
[idx
],
1118 data
, fragment_len
, access_key
);
1120 if (rc
== PGM_PROTECTION
&& try_storage_prot_override
)
1121 rc
= access_guest_page_with_key(vcpu
->kvm
, mode
, gpas
[idx
],
1122 data
, fragment_len
, PAGE_SPO_ACC
);
1125 len
-= fragment_len
;
1126 data
+= fragment_len
;
1127 ga
= kvm_s390_logical_to_effective(vcpu
, ga
+ fragment_len
);
1130 bool terminate
= (mode
== GACC_STORE
) && (idx
> 0);
1132 if (rc
== PGM_PROTECTION
)
1133 prot
= PROT_TYPE_KEYC
;
1136 rc
= trans_exc_ending(vcpu
, rc
, ga
, ar
, mode
, prot
, terminate
);
1140 ipte_unlock(vcpu
->kvm
);
1141 if (nr_pages
> ARRAY_SIZE(gpa_array
))
1146 int access_guest_real(struct kvm_vcpu
*vcpu
, unsigned long gra
,
1147 void *data
, unsigned long len
, enum gacc_mode mode
)
1149 unsigned int fragment_len
;
1153 while (len
&& !rc
) {
1154 gpa
= kvm_s390_real_to_abs(vcpu
, gra
);
1155 fragment_len
= min(PAGE_SIZE
- offset_in_page(gpa
), len
);
1156 rc
= access_guest_page(vcpu
->kvm
, mode
, gpa
, data
, fragment_len
);
1157 len
-= fragment_len
;
1158 gra
+= fragment_len
;
1159 data
+= fragment_len
;
1165 * cmpxchg_guest_abs_with_key() - Perform cmpxchg on guest absolute address.
1166 * @kvm: Virtual machine instance.
1167 * @gpa: Absolute guest address of the location to be changed.
1168 * @len: Operand length of the cmpxchg, required: 1 <= len <= 16. Providing a
1169 * non power of two will result in failure.
1170 * @old_addr: Pointer to old value. If the location at @gpa contains this value,
1171 * the exchange will succeed. After calling cmpxchg_guest_abs_with_key()
1172 * *@old_addr contains the value at @gpa before the attempt to
1173 * exchange the value.
1174 * @new: The value to place at @gpa.
1175 * @access_key: The access key to use for the guest access.
1176 * @success: output value indicating if an exchange occurred.
1178 * Atomically exchange the value at @gpa by @new, if it contains *@old.
1179 * Honors storage keys.
1181 * Return: * 0: successful exchange
1182 * * >0: a program interruption code indicating the reason cmpxchg could
1184 * * -EINVAL: address misaligned or len not power of two
1185 * * -EAGAIN: transient failure (len 1 or 2)
1186 * * -EOPNOTSUPP: read-only memslot (should never occur)
1188 int cmpxchg_guest_abs_with_key(struct kvm
*kvm
, gpa_t gpa
, int len
,
1189 __uint128_t
*old_addr
, __uint128_t
new,
1190 u8 access_key
, bool *success
)
1192 gfn_t gfn
= gpa_to_gfn(gpa
);
1193 struct kvm_memory_slot
*slot
= gfn_to_memslot(kvm
, gfn
);
1198 if (!IS_ALIGNED(gpa
, len
))
1201 hva
= gfn_to_hva_memslot_prot(slot
, gfn
, &writable
);
1202 if (kvm_is_error_hva(hva
))
1203 return PGM_ADDRESSING
;
1205 * Check if it's a read-only memslot, even though that cannot occur
1206 * since those are unsupported.
1207 * Don't try to actually handle that case.
1212 hva
+= offset_in_page(gpa
);
1214 * The cmpxchg_user_key macro depends on the type of "old", so we need
1215 * a case for each valid length and get some code duplication as long
1216 * as we don't introduce a new macro.
1222 ret
= cmpxchg_user_key((u8 __user
*)hva
, &old
, *old_addr
, new, access_key
);
1223 *success
= !ret
&& old
== *old_addr
;
1230 ret
= cmpxchg_user_key((u16 __user
*)hva
, &old
, *old_addr
, new, access_key
);
1231 *success
= !ret
&& old
== *old_addr
;
1238 ret
= cmpxchg_user_key((u32 __user
*)hva
, &old
, *old_addr
, new, access_key
);
1239 *success
= !ret
&& old
== *old_addr
;
1246 ret
= cmpxchg_user_key((u64 __user
*)hva
, &old
, *old_addr
, new, access_key
);
1247 *success
= !ret
&& old
== *old_addr
;
1254 ret
= cmpxchg_user_key((__uint128_t __user
*)hva
, &old
, *old_addr
, new, access_key
);
1255 *success
= !ret
&& old
== *old_addr
;
1263 mark_page_dirty_in_slot(kvm
, slot
, gfn
);
1265 * Assume that the fault is caused by protection, either key protection
1266 * or user page write protection.
1269 ret
= PGM_PROTECTION
;
1274 * guest_translate_address_with_key - translate guest logical into guest absolute address
1275 * @vcpu: virtual cpu
1276 * @gva: Guest virtual address
1277 * @ar: Access register
1278 * @gpa: Guest physical address
1279 * @mode: Translation access mode
1280 * @access_key: access key to mach the storage key with
1282 * Parameter semantics are the same as the ones from guest_translate.
1283 * The memory contents at the guest address are not changed.
1285 * Note: The IPTE lock is not taken during this function, so the caller
1286 * has to take care of this.
1288 int guest_translate_address_with_key(struct kvm_vcpu
*vcpu
, unsigned long gva
, u8 ar
,
1289 unsigned long *gpa
, enum gacc_mode mode
,
1295 gva
= kvm_s390_logical_to_effective(vcpu
, gva
);
1296 rc
= get_vcpu_asce(vcpu
, &asce
, gva
, ar
, mode
);
1299 return guest_range_to_gpas(vcpu
, gva
, ar
, gpa
, 1, asce
, mode
,
1304 * check_gva_range - test a range of guest virtual addresses for accessibility
1305 * @vcpu: virtual cpu
1306 * @gva: Guest virtual address
1307 * @ar: Access register
1308 * @length: Length of test range
1309 * @mode: Translation access mode
1310 * @access_key: access key to mach the storage keys with
1312 int check_gva_range(struct kvm_vcpu
*vcpu
, unsigned long gva
, u8 ar
,
1313 unsigned long length
, enum gacc_mode mode
, u8 access_key
)
1318 rc
= get_vcpu_asce(vcpu
, &asce
, gva
, ar
, mode
);
1321 ipte_lock(vcpu
->kvm
);
1322 rc
= guest_range_to_gpas(vcpu
, gva
, ar
, NULL
, length
, asce
, mode
,
1324 ipte_unlock(vcpu
->kvm
);
1330 * check_gpa_range - test a range of guest physical addresses for accessibility
1331 * @kvm: virtual machine instance
1332 * @gpa: guest physical address
1333 * @length: length of test range
1334 * @mode: access mode to test, relevant for storage keys
1335 * @access_key: access key to mach the storage keys with
1337 int check_gpa_range(struct kvm
*kvm
, unsigned long gpa
, unsigned long length
,
1338 enum gacc_mode mode
, u8 access_key
)
1340 unsigned int fragment_len
;
1343 while (length
&& !rc
) {
1344 fragment_len
= min(PAGE_SIZE
- offset_in_page(gpa
), length
);
1345 rc
= vm_check_access_key(kvm
, access_key
, mode
, gpa
);
1346 length
-= fragment_len
;
1347 gpa
+= fragment_len
;
1353 * kvm_s390_check_low_addr_prot_real - check for low-address protection
1354 * @vcpu: virtual cpu
1355 * @gra: Guest real address
1357 * Checks whether an address is subject to low-address protection and set
1358 * up vcpu->arch.pgm accordingly if necessary.
1360 * Return: 0 if no protection exception, or PGM_PROTECTION if protected.
1362 int kvm_s390_check_low_addr_prot_real(struct kvm_vcpu
*vcpu
, unsigned long gra
)
1364 union ctlreg0 ctlreg0
= {.val
= vcpu
->arch
.sie_block
->gcr
[0]};
1366 if (!ctlreg0
.lap
|| !is_low_address(gra
))
1368 return trans_exc(vcpu
, PGM_PROTECTION
, gra
, 0, GACC_STORE
, PROT_TYPE_LA
);
1372 * kvm_s390_shadow_tables - walk the guest page table and create shadow tables
1373 * @sg: pointer to the shadow guest address space structure
1374 * @saddr: faulting address in the shadow gmap
1375 * @pgt: pointer to the beginning of the page table for the given address if
1376 * successful (return value 0), or to the first invalid DAT entry in
1377 * case of exceptions (return value > 0)
1378 * @dat_protection: referenced memory is write protected
1379 * @fake: pgt references contiguous guest memory block, not a pgtable
1381 static int kvm_s390_shadow_tables(struct gmap
*sg
, unsigned long saddr
,
1382 unsigned long *pgt
, int *dat_protection
,
1386 struct gmap
*parent
;
1388 union vaddress vaddr
;
1393 *dat_protection
= 0;
1395 parent
= sg
->parent
;
1397 asce
.val
= sg
->orig_asce
;
1398 ptr
= asce
.origin
* PAGE_SIZE
;
1402 asce
.dt
= ASCE_TYPE_REGION1
;
1405 case ASCE_TYPE_REGION1
:
1406 if (vaddr
.rfx01
> asce
.tl
&& !*fake
)
1407 return PGM_REGION_FIRST_TRANS
;
1409 case ASCE_TYPE_REGION2
:
1411 return PGM_ASCE_TYPE
;
1412 if (vaddr
.rsx01
> asce
.tl
)
1413 return PGM_REGION_SECOND_TRANS
;
1415 case ASCE_TYPE_REGION3
:
1416 if (vaddr
.rfx
|| vaddr
.rsx
)
1417 return PGM_ASCE_TYPE
;
1418 if (vaddr
.rtx01
> asce
.tl
)
1419 return PGM_REGION_THIRD_TRANS
;
1421 case ASCE_TYPE_SEGMENT
:
1422 if (vaddr
.rfx
|| vaddr
.rsx
|| vaddr
.rtx
)
1423 return PGM_ASCE_TYPE
;
1424 if (vaddr
.sx01
> asce
.tl
)
1425 return PGM_SEGMENT_TRANSLATION
;
1430 case ASCE_TYPE_REGION1
: {
1431 union region1_table_entry rfte
;
1434 ptr
+= vaddr
.rfx
* _REGION1_SIZE
;
1438 *pgt
= ptr
+ vaddr
.rfx
* 8;
1439 rc
= gmap_read_table(parent
, ptr
+ vaddr
.rfx
* 8, &rfte
.val
);
1443 return PGM_REGION_FIRST_TRANS
;
1444 if (rfte
.tt
!= TABLE_TYPE_REGION1
)
1445 return PGM_TRANSLATION_SPEC
;
1446 if (vaddr
.rsx01
< rfte
.tf
|| vaddr
.rsx01
> rfte
.tl
)
1447 return PGM_REGION_SECOND_TRANS
;
1448 if (sg
->edat_level
>= 1)
1449 *dat_protection
|= rfte
.p
;
1450 ptr
= rfte
.rto
* PAGE_SIZE
;
1452 rc
= gmap_shadow_r2t(sg
, saddr
, rfte
.val
, *fake
);
1455 kvm
->stat
.gmap_shadow_r1_entry
++;
1458 case ASCE_TYPE_REGION2
: {
1459 union region2_table_entry rste
;
1462 ptr
+= vaddr
.rsx
* _REGION2_SIZE
;
1466 *pgt
= ptr
+ vaddr
.rsx
* 8;
1467 rc
= gmap_read_table(parent
, ptr
+ vaddr
.rsx
* 8, &rste
.val
);
1471 return PGM_REGION_SECOND_TRANS
;
1472 if (rste
.tt
!= TABLE_TYPE_REGION2
)
1473 return PGM_TRANSLATION_SPEC
;
1474 if (vaddr
.rtx01
< rste
.tf
|| vaddr
.rtx01
> rste
.tl
)
1475 return PGM_REGION_THIRD_TRANS
;
1476 if (sg
->edat_level
>= 1)
1477 *dat_protection
|= rste
.p
;
1478 ptr
= rste
.rto
* PAGE_SIZE
;
1480 rste
.p
|= *dat_protection
;
1481 rc
= gmap_shadow_r3t(sg
, saddr
, rste
.val
, *fake
);
1484 kvm
->stat
.gmap_shadow_r2_entry
++;
1487 case ASCE_TYPE_REGION3
: {
1488 union region3_table_entry rtte
;
1491 ptr
+= vaddr
.rtx
* _REGION3_SIZE
;
1495 *pgt
= ptr
+ vaddr
.rtx
* 8;
1496 rc
= gmap_read_table(parent
, ptr
+ vaddr
.rtx
* 8, &rtte
.val
);
1500 return PGM_REGION_THIRD_TRANS
;
1501 if (rtte
.tt
!= TABLE_TYPE_REGION3
)
1502 return PGM_TRANSLATION_SPEC
;
1503 if (rtte
.cr
&& asce
.p
&& sg
->edat_level
>= 2)
1504 return PGM_TRANSLATION_SPEC
;
1505 if (rtte
.fc
&& sg
->edat_level
>= 2) {
1506 *dat_protection
|= rtte
.fc0
.p
;
1508 ptr
= rtte
.fc1
.rfaa
* _REGION3_SIZE
;
1512 if (vaddr
.sx01
< rtte
.fc0
.tf
|| vaddr
.sx01
> rtte
.fc0
.tl
)
1513 return PGM_SEGMENT_TRANSLATION
;
1514 if (sg
->edat_level
>= 1)
1515 *dat_protection
|= rtte
.fc0
.p
;
1516 ptr
= rtte
.fc0
.sto
* PAGE_SIZE
;
1518 rtte
.fc0
.p
|= *dat_protection
;
1519 rc
= gmap_shadow_sgt(sg
, saddr
, rtte
.val
, *fake
);
1522 kvm
->stat
.gmap_shadow_r3_entry
++;
1525 case ASCE_TYPE_SEGMENT
: {
1526 union segment_table_entry ste
;
1529 ptr
+= vaddr
.sx
* _SEGMENT_SIZE
;
1533 *pgt
= ptr
+ vaddr
.sx
* 8;
1534 rc
= gmap_read_table(parent
, ptr
+ vaddr
.sx
* 8, &ste
.val
);
1538 return PGM_SEGMENT_TRANSLATION
;
1539 if (ste
.tt
!= TABLE_TYPE_SEGMENT
)
1540 return PGM_TRANSLATION_SPEC
;
1541 if (ste
.cs
&& asce
.p
)
1542 return PGM_TRANSLATION_SPEC
;
1543 *dat_protection
|= ste
.fc0
.p
;
1544 if (ste
.fc
&& sg
->edat_level
>= 1) {
1546 ptr
= ste
.fc1
.sfaa
* _SEGMENT_SIZE
;
1550 ptr
= ste
.fc0
.pto
* (PAGE_SIZE
/ 2);
1552 ste
.fc0
.p
|= *dat_protection
;
1553 rc
= gmap_shadow_pgt(sg
, saddr
, ste
.val
, *fake
);
1556 kvm
->stat
.gmap_shadow_sg_entry
++;
1559 /* Return the parent address of the page table */
1565 * kvm_s390_shadow_fault - handle fault on a shadow page table
1566 * @vcpu: virtual cpu
1567 * @sg: pointer to the shadow guest address space structure
1568 * @saddr: faulting address in the shadow gmap
1569 * @datptr: will contain the address of the faulting DAT table entry, or of
1570 * the valid leaf, plus some flags
1572 * Returns: - 0 if the shadow fault was successfully resolved
1573 * - > 0 (pgm exception code) on exceptions while faulting
1574 * - -EAGAIN if the caller can retry immediately
1575 * - -EFAULT when accessing invalid guest addresses
1576 * - -ENOMEM if out of memory
1578 int kvm_s390_shadow_fault(struct kvm_vcpu
*vcpu
, struct gmap
*sg
,
1579 unsigned long saddr
, unsigned long *datptr
)
1581 union vaddress vaddr
;
1582 union page_table_entry pte
;
1583 unsigned long pgt
= 0;
1584 int dat_protection
, fake
;
1587 mmap_read_lock(sg
->mm
);
1589 * We don't want any guest-2 tables to change - so the parent
1590 * tables/pointers we read stay valid - unshadowing is however
1591 * always possible - only guest_table_lock protects us.
1593 ipte_lock(vcpu
->kvm
);
1595 rc
= gmap_shadow_pgt_lookup(sg
, saddr
, &pgt
, &dat_protection
, &fake
);
1597 rc
= kvm_s390_shadow_tables(sg
, saddr
, &pgt
, &dat_protection
,
1602 pte
.val
= pgt
+ vaddr
.px
* PAGE_SIZE
;
1607 case PGM_SEGMENT_TRANSLATION
:
1608 case PGM_REGION_THIRD_TRANS
:
1609 case PGM_REGION_SECOND_TRANS
:
1610 case PGM_REGION_FIRST_TRANS
:
1614 pgt
+= vaddr
.px
* 8;
1615 rc
= gmap_read_table(sg
->parent
, pgt
, &pte
.val
);
1618 *datptr
= pgt
| dat_protection
* PEI_DAT_PROT
;
1620 rc
= PGM_PAGE_TRANSLATION
;
1622 rc
= PGM_TRANSLATION_SPEC
;
1624 pte
.p
|= dat_protection
;
1626 rc
= gmap_shadow_page(sg
, saddr
, __pte(pte
.val
));
1627 vcpu
->kvm
->stat
.gmap_shadow_pg_entry
++;
1628 ipte_unlock(vcpu
->kvm
);
1629 mmap_read_unlock(sg
->mm
);