1 /* SPDX-License-Identifier: GPL-2.0-only */
4 * Copyright SUSE Linux Products GmbH 2010
6 * Authors: Alexander Graf <agraf@suse.de>
9 #ifndef __ASM_KVM_BOOK3S_64_H__
10 #define __ASM_KVM_BOOK3S_64_H__
12 #include <linux/string.h>
13 #include <asm/bitops.h>
14 #include <asm/book3s/64/mmu-hash.h>
15 #include <asm/cpu_has_feature.h>
16 #include <asm/ppc-opcode.h>
17 #include <asm/pte-walk.h>
19 #ifdef CONFIG_PPC_PSERIES
20 static inline bool kvmhv_on_pseries(void)
22 return !cpu_has_feature(CPU_FTR_HVMODE
);
25 static inline bool kvmhv_on_pseries(void)
32 * Structure for a nested guest, that is, for a guest that is managed by
35 struct kvm_nested_guest
{
36 struct kvm
*l1_host
; /* L1 VM that owns this nested guest */
37 int l1_lpid
; /* lpid L1 guest thinks this guest is */
38 int shadow_lpid
; /* real lpid of this nested guest */
39 pgd_t
*shadow_pgtable
; /* our page table for this guest */
40 u64 l1_gr_to_hr
; /* L1's addr of part'n-scoped table */
41 u64 process_table
; /* process table entry for this guest */
42 u64 hfscr
; /* HFSCR that the L1 requested for this nested guest */
43 long refcnt
; /* number of pointers to this struct */
44 struct mutex tlb_lock
; /* serialize page faults and tlbies */
45 struct kvm_nested_guest
*next
;
46 cpumask_t need_tlb_flush
;
47 cpumask_t cpu_in_guest
;
48 short prev_cpu
[NR_CPUS
];
49 u8 radix
; /* is this nested guest radix */
53 * We define a nested rmap entry as a single 64-bit quantity
54 * 0xFFF0000000000000 12-bit lpid field
55 * 0x000FFFFFFFFFF000 40-bit guest 4k page frame number
56 * 0x0000000000000001 1-bit single entry flag
58 #define RMAP_NESTED_LPID_MASK 0xFFF0000000000000UL
59 #define RMAP_NESTED_LPID_SHIFT (52)
60 #define RMAP_NESTED_GPA_MASK 0x000FFFFFFFFFF000UL
61 #define RMAP_NESTED_IS_SINGLE_ENTRY 0x0000000000000001UL
63 /* Structure for a nested guest rmap entry */
65 struct llist_node list
;
70 * for_each_nest_rmap_safe - iterate over the list of nested rmap entries
71 * safe against removal of the list entry or NULL list
72 * @pos: a (struct rmap_nested *) to use as a loop cursor
73 * @node: pointer to the first entry
74 * NOTE: this can be NULL
75 * @rmapp: an (unsigned long *) in which to return the rmap entries on each
77 * NOTE: this must point to already allocated memory
79 * The nested_rmap is a llist of (struct rmap_nested) entries pointed to by the
80 * rmap entry in the memslot. The list is always terminated by a "single entry"
81 * stored in the list element of the final entry of the llist. If there is ONLY
82 * a single entry then this is itself in the rmap entry of the memslot, not a
85 * Note that the iterator below assumes that a nested rmap entry is always
86 * non-zero. This is true for our usage because the LPID field is always
87 * non-zero (zero is reserved for the host).
89 * This should be used to iterate over the list of rmap_nested entries with
90 * processing done on the u64 rmap value given by each iteration. This is safe
91 * against removal of list entries and it is always safe to call free on (pos).
94 * struct rmap_nested *cursor;
95 * struct llist_node *first;
97 * for_each_nest_rmap_safe(cursor, first, &rmap) {
102 #define for_each_nest_rmap_safe(pos, node, rmapp) \
103 for ((pos) = llist_entry((node), typeof(*(pos)), list); \
105 (*(rmapp) = ((RMAP_NESTED_IS_SINGLE_ENTRY & ((u64) (node))) ? \
106 ((u64) (node)) : ((pos)->rmap))) && \
107 (((node) = ((RMAP_NESTED_IS_SINGLE_ENTRY & ((u64) (node))) ? \
108 ((struct llist_node *) ((pos) = NULL)) : \
109 (pos)->list.next)), true); \
110 (pos) = llist_entry((node), typeof(*(pos)), list))
112 struct kvm_nested_guest
*kvmhv_get_nested(struct kvm
*kvm
, int l1_lpid
,
114 void kvmhv_put_nested(struct kvm_nested_guest
*gp
);
115 int kvmhv_nested_next_lpid(struct kvm
*kvm
, int lpid
);
117 /* Encoding of first parameter for H_TLB_INVALIDATE */
118 #define H_TLBIE_P1_ENC(ric, prs, r) (___PPC_RIC(ric) | ___PPC_PRS(prs) | \
121 /* Power architecture requires HPT is at least 256kiB, at most 64TiB */
122 #define PPC_MIN_HPT_ORDER 18
123 #define PPC_MAX_HPT_ORDER 46
125 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
126 static inline struct kvmppc_book3s_shadow_vcpu
*svcpu_get(struct kvm_vcpu
*vcpu
)
129 return &get_paca()->shadow_vcpu
;
132 static inline void svcpu_put(struct kvmppc_book3s_shadow_vcpu
*svcpu
)
138 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
140 static inline bool kvm_is_radix(struct kvm
*kvm
)
142 return kvm
->arch
.radix
;
145 static inline bool kvmhv_vcpu_is_radix(struct kvm_vcpu
*vcpu
)
149 if (vcpu
->arch
.nested
)
150 radix
= vcpu
->arch
.nested
->radix
;
152 radix
= kvm_is_radix(vcpu
->kvm
);
157 int kvmhv_vcpu_entry_p9(struct kvm_vcpu
*vcpu
, u64 time_limit
, unsigned long lpcr
);
159 #define KVM_DEFAULT_HPT_ORDER 24 /* 16MB HPT by default */
163 * Invalid HDSISR value which is used to indicate when HW has not set the reg.
164 * Used to work around an errata.
166 #define HDSISR_CANARY 0x7fff
169 * We use a lock bit in HPTE dword 0 to synchronize updates and
170 * accesses to each HPTE, and another bit to indicate non-present
173 #define HPTE_V_HVLOCK 0x40UL
174 #define HPTE_V_ABSENT 0x20UL
177 * We use this bit in the guest_rpte field of the revmap entry
178 * to indicate a modified HPTE.
180 #define HPTE_GR_MODIFIED (1ul << 62)
182 /* These bits are reserved in the guest view of the HPTE */
183 #define HPTE_GR_RESERVED HPTE_GR_MODIFIED
185 static inline long try_lock_hpte(__be64
*hpte
, unsigned long bits
)
187 unsigned long tmp
, old
;
188 __be64 be_lockbit
, be_bits
;
191 * We load/store in native endian, but the HTAB is in big endian. If
192 * we byte swap all data we apply on the PTE we're implicitly correct
195 be_lockbit
= cpu_to_be64(HPTE_V_HVLOCK
);
196 be_bits
= cpu_to_be64(bits
);
198 asm volatile(" ldarx %0,0,%2\n"
206 : "=&r" (tmp
), "=&r" (old
)
207 : "r" (hpte
), "r" (be_bits
), "r" (be_lockbit
)
212 static inline void unlock_hpte(__be64
*hpte
, unsigned long hpte_v
)
214 hpte_v
&= ~HPTE_V_HVLOCK
;
215 asm volatile(PPC_RELEASE_BARRIER
"" : : : "memory");
216 hpte
[0] = cpu_to_be64(hpte_v
);
219 /* Without barrier */
220 static inline void __unlock_hpte(__be64
*hpte
, unsigned long hpte_v
)
222 hpte_v
&= ~HPTE_V_HVLOCK
;
223 hpte
[0] = cpu_to_be64(hpte_v
);
227 * These functions encode knowledge of the POWER7/8/9 hardware
228 * interpretations of the HPTE LP (large page size) field.
230 static inline int kvmppc_hpte_page_shifts(unsigned long h
, unsigned long l
)
234 if (!(h
& HPTE_V_LARGE
))
236 lphi
= (l
>> 16) & 0xf;
237 switch ((l
>> 12) & 0xf) {
239 return !lphi
? 24 : 0; /* 16MB */
242 return 16; /* 64kB */
245 return !lphi
? 34 : 0; /* 16GB */
248 return (16 << 8) + 12; /* 64kB in 4kB */
252 return (24 << 8) + 16; /* 16MB in 64kkB */
254 return (24 << 8) + 12; /* 16MB in 4kB */
260 static inline int kvmppc_hpte_base_page_shift(unsigned long h
, unsigned long l
)
262 return kvmppc_hpte_page_shifts(h
, l
) & 0xff;
265 static inline int kvmppc_hpte_actual_page_shift(unsigned long h
, unsigned long l
)
267 int tmp
= kvmppc_hpte_page_shifts(h
, l
);
274 static inline unsigned long kvmppc_actual_pgsz(unsigned long v
, unsigned long r
)
276 int shift
= kvmppc_hpte_actual_page_shift(v
, r
);
283 static inline int kvmppc_pgsize_lp_encoding(int base_shift
, int actual_shift
)
285 switch (base_shift
) {
287 switch (actual_shift
) {
297 switch (actual_shift
) {
310 static inline unsigned long compute_tlbie_rb(unsigned long v
, unsigned long r
,
311 unsigned long pte_index
)
313 int a_pgshift
, b_pgshift
;
314 unsigned long rb
= 0, va_low
, sllp
;
316 b_pgshift
= a_pgshift
= kvmppc_hpte_page_shifts(v
, r
);
317 if (a_pgshift
>= 0x100) {
323 * Ignore the top 14 bits of va
324 * v have top two bits covering segment size, hence move
325 * by 16 bits, Also clear the lower HPTE_V_AVPN_SHIFT (7) bits.
326 * AVA field in v also have the lower 23 bits ignored.
327 * For base page size 4K we need 14 .. 65 bits (so need to
328 * collect extra 11 bits)
329 * For others we need 14..14+i
331 /* This covers 14..54 bits of va*/
332 rb
= (v
& ~0x7fUL
) << 16; /* AVA field */
335 * AVA in v had cleared lower 23 bits. We need to derive
336 * that from pteg index
338 va_low
= pte_index
>> 3;
339 if (v
& HPTE_V_SECONDARY
)
342 * get the vpn bits from va_low using reverse of hashing.
343 * In v we have va with 23 bits dropped and then left shifted
344 * HPTE_V_AVPN_SHIFT (7) bits. Now to find vsid we need
345 * right shift it with (SID_SHIFT - (23 - 7))
347 if (!(v
& HPTE_V_1TB_SEG
))
348 va_low
^= v
>> (SID_SHIFT
- 16);
350 va_low
^= v
>> (SID_SHIFT_1T
- 16);
353 if (b_pgshift
<= 12) {
354 if (a_pgshift
> 12) {
355 sllp
= (a_pgshift
== 16) ? 5 : 4;
356 rb
|= sllp
<< 5; /* AP field */
358 rb
|= (va_low
& 0x7ff) << 12; /* remaining 11 bits of AVA */
362 * remaining bits of AVA/LP fields
363 * Also contain the rr bits of LP
365 rb
|= (va_low
<< b_pgshift
) & 0x7ff000;
367 * Now clear not needed LP bits based on actual psize
369 rb
&= ~((1ul << a_pgshift
) - 1);
371 * AVAL field 58..77 - base_page_shift bits of va
372 * we have space for 58..64 bits, Missing bits should
373 * be zero filled. +1 is to take care of L bit shift
375 aval_shift
= 64 - (77 - b_pgshift
) + 1;
376 rb
|= ((va_low
<< aval_shift
) & 0xfe);
378 rb
|= 1; /* L field */
379 rb
|= r
& 0xff000 & ((1ul << a_pgshift
) - 1); /* LP field */
381 rb
|= (v
>> HPTE_V_SSIZE_SHIFT
) << 8; /* B field */
385 static inline unsigned long hpte_rpn(unsigned long ptel
, unsigned long psize
)
387 return ((ptel
& HPTE_R_RPN
) & ~(psize
- 1)) >> PAGE_SHIFT
;
390 static inline int hpte_is_writable(unsigned long ptel
)
392 unsigned long pp
= ptel
& (HPTE_R_PP0
| HPTE_R_PP
);
394 return pp
!= PP_RXRX
&& pp
!= PP_RXXX
;
397 static inline unsigned long hpte_make_readonly(unsigned long ptel
)
399 if ((ptel
& HPTE_R_PP0
) || (ptel
& HPTE_R_PP
) == PP_RWXX
)
400 ptel
= (ptel
& ~HPTE_R_PP
) | PP_RXXX
;
406 static inline bool hpte_cache_flags_ok(unsigned long hptel
, bool is_ci
)
408 unsigned int wimg
= hptel
& HPTE_R_WIMG
;
411 if (wimg
== (HPTE_R_W
| HPTE_R_I
| HPTE_R_M
) &&
412 cpu_has_feature(CPU_FTR_ARCH_206
))
416 return wimg
== HPTE_R_M
;
418 * if host is mapped cache inhibited, make sure hptel also have
421 if (wimg
& HPTE_R_W
) /* FIXME!! is this ok for all guest. ? */
423 return !!(wimg
& HPTE_R_I
);
427 * If it's present and writable, atomically set dirty and referenced bits and
428 * return the PTE, otherwise return 0.
430 static inline pte_t
kvmppc_read_update_linux_pte(pte_t
*ptep
, int writing
)
432 pte_t old_pte
, new_pte
= __pte(0);
436 * Make sure we don't reload from ptep
438 old_pte
= READ_ONCE(*ptep
);
440 * wait until H_PAGE_BUSY is clear then set it atomically
442 if (unlikely(pte_val(old_pte
) & H_PAGE_BUSY
)) {
446 /* If pte is not present return None */
447 if (unlikely(!pte_present(old_pte
)))
450 new_pte
= pte_mkyoung(old_pte
);
451 if (writing
&& pte_write(old_pte
))
452 new_pte
= pte_mkdirty(new_pte
);
454 if (pte_xchg(ptep
, old_pte
, new_pte
))
460 static inline bool hpte_read_permission(unsigned long pp
, unsigned long key
)
463 return PP_RWRX
<= pp
&& pp
<= PP_RXRX
;
467 static inline bool hpte_write_permission(unsigned long pp
, unsigned long key
)
470 return pp
== PP_RWRW
;
471 return pp
<= PP_RWRW
;
474 static inline int hpte_get_skey_perm(unsigned long hpte_r
, unsigned long amr
)
478 skey
= ((hpte_r
& HPTE_R_KEY_HI
) >> 57) |
479 ((hpte_r
& HPTE_R_KEY_LO
) >> 9);
480 return (amr
>> (62 - 2 * skey
)) & 3;
483 static inline void lock_rmap(unsigned long *rmap
)
486 while (test_bit(KVMPPC_RMAP_LOCK_BIT
, rmap
))
488 } while (test_and_set_bit_lock(KVMPPC_RMAP_LOCK_BIT
, rmap
));
491 static inline void unlock_rmap(unsigned long *rmap
)
493 __clear_bit_unlock(KVMPPC_RMAP_LOCK_BIT
, rmap
);
496 static inline bool slot_is_aligned(struct kvm_memory_slot
*memslot
,
497 unsigned long pagesize
)
499 unsigned long mask
= (pagesize
>> PAGE_SHIFT
) - 1;
501 if (pagesize
<= PAGE_SIZE
)
503 return !(memslot
->base_gfn
& mask
) && !(memslot
->npages
& mask
);
507 * This works for 4k, 64k and 16M pages on POWER7,
508 * and 4k and 16M pages on PPC970.
510 static inline unsigned long slb_pgsize_encoding(unsigned long psize
)
512 unsigned long senc
= 0;
514 if (psize
> 0x1000) {
516 if (psize
== 0x10000)
517 senc
|= SLB_VSID_LP_01
;
522 static inline int is_vrma_hpte(unsigned long hpte_v
)
524 return (hpte_v
& ~0xffffffUL
) ==
525 (HPTE_V_1TB_SEG
| (VRMA_VSID
<< (40 - 16)));
528 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
530 * Note modification of an HPTE; set the HPTE modified bit
531 * if anyone is interested.
533 static inline void note_hpte_modification(struct kvm
*kvm
,
534 struct revmap_entry
*rev
)
536 if (atomic_read(&kvm
->arch
.hpte_mod_interest
))
537 rev
->guest_rpte
|= HPTE_GR_MODIFIED
;
541 * Like kvm_memslots(), but for use in real mode when we can't do
542 * any RCU stuff (since the secondary threads are offline from the
543 * kernel's point of view), and we can't print anything.
544 * Thus we use rcu_dereference_raw() rather than rcu_dereference_check().
546 static inline struct kvm_memslots
*kvm_memslots_raw(struct kvm
*kvm
)
548 return rcu_dereference_raw_check(kvm
->memslots
[0]);
551 extern void kvmppc_mmu_debugfs_init(struct kvm
*kvm
);
552 extern void kvmhv_radix_debugfs_init(struct kvm
*kvm
);
554 extern void kvmhv_rm_send_ipi(int cpu
);
556 static inline unsigned long kvmppc_hpt_npte(struct kvm_hpt_info
*hpt
)
558 /* HPTEs are 2**4 bytes long */
559 return 1UL << (hpt
->order
- 4);
562 static inline unsigned long kvmppc_hpt_mask(struct kvm_hpt_info
*hpt
)
564 /* 128 (2**7) bytes in each HPTEG */
565 return (1UL << (hpt
->order
- 7)) - 1;
568 /* Set bits in a dirty bitmap, which is in LE format */
569 static inline void set_dirty_bits(unsigned long *map
, unsigned long i
,
570 unsigned long npages
)
574 memset((char *)map
+ i
/ 8, 0xff, npages
/ 8);
576 for (; npages
; ++i
, --npages
)
577 __set_bit_le(i
, map
);
580 static inline void set_dirty_bits_atomic(unsigned long *map
, unsigned long i
,
581 unsigned long npages
)
584 memset((char *)map
+ i
/ 8, 0xff, npages
/ 8);
586 for (; npages
; ++i
, --npages
)
590 static inline u64
sanitize_msr(u64 msr
)
597 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
598 static inline void copy_from_checkpoint(struct kvm_vcpu
*vcpu
)
600 vcpu
->arch
.regs
.ccr
= vcpu
->arch
.cr_tm
;
601 vcpu
->arch
.regs
.xer
= vcpu
->arch
.xer_tm
;
602 vcpu
->arch
.regs
.link
= vcpu
->arch
.lr_tm
;
603 vcpu
->arch
.regs
.ctr
= vcpu
->arch
.ctr_tm
;
604 vcpu
->arch
.amr
= vcpu
->arch
.amr_tm
;
605 vcpu
->arch
.ppr
= vcpu
->arch
.ppr_tm
;
606 vcpu
->arch
.dscr
= vcpu
->arch
.dscr_tm
;
607 vcpu
->arch
.tar
= vcpu
->arch
.tar_tm
;
608 memcpy(vcpu
->arch
.regs
.gpr
, vcpu
->arch
.gpr_tm
,
609 sizeof(vcpu
->arch
.regs
.gpr
));
610 vcpu
->arch
.fp
= vcpu
->arch
.fp_tm
;
611 vcpu
->arch
.vr
= vcpu
->arch
.vr_tm
;
612 vcpu
->arch
.vrsave
= vcpu
->arch
.vrsave_tm
;
615 static inline void copy_to_checkpoint(struct kvm_vcpu
*vcpu
)
617 vcpu
->arch
.cr_tm
= vcpu
->arch
.regs
.ccr
;
618 vcpu
->arch
.xer_tm
= vcpu
->arch
.regs
.xer
;
619 vcpu
->arch
.lr_tm
= vcpu
->arch
.regs
.link
;
620 vcpu
->arch
.ctr_tm
= vcpu
->arch
.regs
.ctr
;
621 vcpu
->arch
.amr_tm
= vcpu
->arch
.amr
;
622 vcpu
->arch
.ppr_tm
= vcpu
->arch
.ppr
;
623 vcpu
->arch
.dscr_tm
= vcpu
->arch
.dscr
;
624 vcpu
->arch
.tar_tm
= vcpu
->arch
.tar
;
625 memcpy(vcpu
->arch
.gpr_tm
, vcpu
->arch
.regs
.gpr
,
626 sizeof(vcpu
->arch
.regs
.gpr
));
627 vcpu
->arch
.fp_tm
= vcpu
->arch
.fp
;
628 vcpu
->arch
.vr_tm
= vcpu
->arch
.vr
;
629 vcpu
->arch
.vrsave_tm
= vcpu
->arch
.vrsave
;
631 #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
633 extern int kvmppc_create_pte(struct kvm
*kvm
, pgd_t
*pgtable
, pte_t pte
,
634 unsigned long gpa
, unsigned int level
,
635 unsigned long mmu_seq
, unsigned int lpid
,
636 unsigned long *rmapp
, struct rmap_nested
**n_rmap
);
637 extern void kvmhv_insert_nest_rmap(struct kvm
*kvm
, unsigned long *rmapp
,
638 struct rmap_nested
**n_rmap
);
639 extern void kvmhv_update_nest_rmap_rc_list(struct kvm
*kvm
, unsigned long *rmapp
,
640 unsigned long clr
, unsigned long set
,
641 unsigned long hpa
, unsigned long nbytes
);
642 extern void kvmhv_remove_nest_rmap_range(struct kvm
*kvm
,
643 const struct kvm_memory_slot
*memslot
,
644 unsigned long gpa
, unsigned long hpa
,
645 unsigned long nbytes
);
647 static inline pte_t
*
648 find_kvm_secondary_pte_unlocked(struct kvm
*kvm
, unsigned long ea
,
653 pte
= __find_linux_pte(kvm
->arch
.pgtable
, ea
, NULL
, hshift
);
657 static inline pte_t
*find_kvm_secondary_pte(struct kvm
*kvm
, unsigned long ea
,
662 VM_WARN(!spin_is_locked(&kvm
->mmu_lock
),
663 "%s called with kvm mmu_lock not held \n", __func__
);
664 pte
= __find_linux_pte(kvm
->arch
.pgtable
, ea
, NULL
, hshift
);
669 static inline pte_t
*find_kvm_host_pte(struct kvm
*kvm
, unsigned long mmu_seq
,
670 unsigned long ea
, unsigned *hshift
)
674 VM_WARN(!spin_is_locked(&kvm
->mmu_lock
),
675 "%s called with kvm mmu_lock not held \n", __func__
);
677 if (mmu_notifier_retry(kvm
, mmu_seq
))
680 pte
= __find_linux_pte(kvm
->mm
->pgd
, ea
, NULL
, hshift
);
685 extern pte_t
*find_kvm_nested_guest_pte(struct kvm
*kvm
, unsigned long lpid
,
686 unsigned long ea
, unsigned *hshift
);
688 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
690 #endif /* __ASM_KVM_BOOK3S_64_H__ */