2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License, version 2, as
4 * published by the Free Software Foundation.
6 * This program is distributed in the hope that it will be useful,
7 * but WITHOUT ANY WARRANTY; without even the implied warranty of
8 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9 * GNU General Public License for more details.
11 * You should have received a copy of the GNU General Public License
12 * along with this program; if not, write to the Free Software
13 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
15 * Copyright IBM Corp. 2007
17 * Authors: Hollis Blanchard <hollisb@us.ibm.com>
20 #ifndef __POWERPC_KVM_HOST_H__
21 #define __POWERPC_KVM_HOST_H__
23 #include <linux/mutex.h>
24 #include <linux/hrtimer.h>
25 #include <linux/interrupt.h>
26 #include <linux/types.h>
27 #include <linux/kvm_types.h>
28 #include <linux/threads.h>
29 #include <linux/spinlock.h>
30 #include <linux/kvm_para.h>
31 #include <linux/list.h>
32 #include <linux/atomic.h>
33 #include <asm/kvm_asm.h>
34 #include <asm/processor.h>
36 #include <asm/cacheflush.h>
37 #include <asm/hvcall.h>
40 #define KVM_MAX_VCPUS NR_CPUS
41 #define KVM_MAX_VCORES NR_CPUS
42 #define KVM_USER_MEM_SLOTS 512
44 #include <asm/cputhreads.h>
45 #define KVM_MAX_VCPU_ID (threads_per_subcore * KVM_MAX_VCORES)
47 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
49 #define KVM_HALT_POLL_NS_DEFAULT 10000 /* 10 us */
51 /* These values are internal and can be increased later */
52 #define KVM_NR_IRQCHIPS 1
53 #define KVM_IRQCHIP_NUM_PINS 256
55 /* PPC-specific vcpu->requests bit members */
56 #define KVM_REQ_WATCHDOG KVM_ARCH_REQ(0)
57 #define KVM_REQ_EPR_EXIT KVM_ARCH_REQ(1)
59 #include <linux/mmu_notifier.h>
61 #define KVM_ARCH_WANT_MMU_NOTIFIER
63 extern int kvm_unmap_hva_range(struct kvm
*kvm
,
64 unsigned long start
, unsigned long end
);
65 extern int kvm_age_hva(struct kvm
*kvm
, unsigned long start
, unsigned long end
);
66 extern int kvm_test_age_hva(struct kvm
*kvm
, unsigned long hva
);
67 extern void kvm_set_spte_hva(struct kvm
*kvm
, unsigned long hva
, pte_t pte
);
69 #define HPTEG_CACHE_NUM (1 << 15)
70 #define HPTEG_HASH_BITS_PTE 13
71 #define HPTEG_HASH_BITS_PTE_LONG 12
72 #define HPTEG_HASH_BITS_VPTE 13
73 #define HPTEG_HASH_BITS_VPTE_LONG 5
74 #define HPTEG_HASH_BITS_VPTE_64K 11
75 #define HPTEG_HASH_NUM_PTE (1 << HPTEG_HASH_BITS_PTE)
76 #define HPTEG_HASH_NUM_PTE_LONG (1 << HPTEG_HASH_BITS_PTE_LONG)
77 #define HPTEG_HASH_NUM_VPTE (1 << HPTEG_HASH_BITS_VPTE)
78 #define HPTEG_HASH_NUM_VPTE_LONG (1 << HPTEG_HASH_BITS_VPTE_LONG)
79 #define HPTEG_HASH_NUM_VPTE_64K (1 << HPTEG_HASH_BITS_VPTE_64K)
81 /* Physical Address Mask - allowed range of real mode RAM access */
82 #define KVM_PAM 0x0fffffffffffffffULL
88 struct kvmppc_vcpu_book3s
;
89 struct kvmppc_book3s_shadow_vcpu
;
92 ulong remote_tlb_flush
;
95 struct kvm_vcpu_stat
{
100 /* Account for special types of light exits: */
101 u64 itlb_real_miss_exits
;
102 u64 itlb_virt_miss_exits
;
103 u64 dtlb_real_miss_exits
;
104 u64 dtlb_virt_miss_exits
;
108 u64 emulated_inst_exits
;
111 u64 halt_poll_success_ns
;
112 u64 halt_poll_fail_ns
;
114 u64 halt_successful_poll
;
115 u64 halt_attempted_poll
;
116 u64 halt_successful_wait
;
117 u64 halt_poll_invalid
;
123 #ifdef CONFIG_PPC_BOOK3S
137 enum kvm_exit_types
{
140 ITLB_REAL_MISS_EXITS
,
141 ITLB_VIRT_MISS_EXITS
,
142 DTLB_REAL_MISS_EXITS
,
143 DTLB_VIRT_MISS_EXITS
,
148 EMULATED_MTMSRWE_EXITS
,
149 EMULATED_WRTEE_EXITS
,
150 EMULATED_MTSPR_EXITS
,
151 EMULATED_MFSPR_EXITS
,
152 EMULATED_MTMSR_EXITS
,
153 EMULATED_MFMSR_EXITS
,
154 EMULATED_TLBSX_EXITS
,
155 EMULATED_TLBWE_EXITS
,
168 __NUMBER_OF_KVM_EXIT_TYPES
171 /* allow access to big endian 32bit upper/lower parts and 64bit var */
172 struct kvmppc_exit_timing
{
181 struct kvmppc_pginfo
{
186 struct kvmppc_spapr_tce_iommu_table
{
188 struct list_head next
;
189 struct iommu_table
*tbl
;
193 struct kvmppc_spapr_tce_table
{
194 struct list_head list
;
199 u64 offset
; /* in pages */
200 u64 size
; /* window size in pages */
201 struct list_head iommu_tables
;
202 struct page
*pages
[0];
205 /* XICS components, defined in book3s_xics.c */
208 extern struct kvm_device_ops kvm_xics_ops
;
210 /* XIVE components, defined in book3s_xive.c */
212 struct kvmppc_xive_vcpu
;
213 extern struct kvm_device_ops kvm_xive_ops
;
215 struct kvmppc_passthru_irqmap
;
218 * The reverse mapping array has one entry for each HPTE,
219 * which stores the guest's view of the second word of the HPTE
220 * (including the guest physical address of the mapping),
221 * plus forward and backward pointers in a doubly-linked ring
222 * of HPTEs that map the same host page. The pointers in this
223 * ring are 32-bit HPTE indexes, to save space.
225 struct revmap_entry
{
226 unsigned long guest_rpte
;
227 unsigned int forw
, back
;
231 * We use the top bit of each memslot->arch.rmap entry as a lock bit,
232 * and bit 32 as a present flag. The bottom 32 bits are the
233 * index in the guest HPT of a HPTE that points to the page.
235 #define KVMPPC_RMAP_LOCK_BIT 63
236 #define KVMPPC_RMAP_RC_SHIFT 32
237 #define KVMPPC_RMAP_REFERENCED (HPTE_R_R << KVMPPC_RMAP_RC_SHIFT)
238 #define KVMPPC_RMAP_PRESENT 0x100000000ul
239 #define KVMPPC_RMAP_INDEX 0xfffffffful
241 struct kvm_arch_memory_slot
{
242 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
244 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
247 struct kvm_hpt_info
{
248 /* Host virtual (linear mapping) address of guest HPT */
250 /* Array of reverse mapping entries for each guest HPTE */
251 struct revmap_entry
*rev
;
252 /* Guest HPT size is 2**(order) bytes */
254 /* 1 if HPT allocated with CMA, 0 otherwise */
258 struct kvm_resize_hpt
;
262 unsigned int smt_mode
; /* # vcpus per virtual core */
263 unsigned int emul_smt_mode
; /* emualted SMT mode, on P9 */
264 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
265 unsigned int tlb_sets
;
266 struct kvm_hpt_info hpt
;
267 atomic64_t mmio_update
;
268 unsigned int host_lpid
;
269 unsigned long host_lpcr
;
271 unsigned long host_sdr1
;
274 unsigned long vrma_slb_v
;
276 atomic_t vcpus_running
;
278 atomic_t hpte_mod_interest
;
279 cpumask_t need_tlb_flush
;
280 cpumask_t cpu_in_guest
;
286 struct dentry
*debugfs_dir
;
287 struct dentry
*htab_dentry
;
288 struct kvm_resize_hpt
*resize_hpt
; /* protected by kvm->lock */
289 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
290 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
291 struct mutex hpt_mutex
;
293 #ifdef CONFIG_PPC_BOOK3S_64
294 struct list_head spapr_tce_tables
;
295 struct list_head rtas_tokens
;
296 DECLARE_BITMAP(enabled_hcalls
, MAX_HCALL_OPCODE
/4 + 1);
298 #ifdef CONFIG_KVM_MPIC
299 struct openpic
*mpic
;
301 #ifdef CONFIG_KVM_XICS
302 struct kvmppc_xics
*xics
;
303 struct kvmppc_xive
*xive
;
304 struct kvmppc_passthru_irqmap
*pimap
;
306 struct kvmppc_ops
*kvm_ops
;
307 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
308 /* This array can grow quite large, keep it at the end */
309 struct kvmppc_vcore
*vcores
[KVM_MAX_VCORES
];
313 #define VCORE_ENTRY_MAP(vc) ((vc)->entry_exit_map & 0xff)
314 #define VCORE_EXIT_MAP(vc) ((vc)->entry_exit_map >> 8)
315 #define VCORE_IS_EXITING(vc) (VCORE_EXIT_MAP(vc) != 0)
317 /* This bit is used when a vcore exit is triggered from outside the vcore */
318 #define VCORE_EXIT_REQ 0x10000
321 * Values for vcore_state.
322 * Note that these are arranged such that lower values
323 * (< VCORE_SLEEPING) don't require stolen time accounting
324 * on load/unload, and higher values do.
326 #define VCORE_INACTIVE 0
327 #define VCORE_PREEMPT 1
328 #define VCORE_PIGGYBACK 2
329 #define VCORE_SLEEPING 3
330 #define VCORE_RUNNING 4
331 #define VCORE_EXITING 5
332 #define VCORE_POLLING 6
335 * Struct used to manage memory for a virtual processor area
336 * registered by a PAPR guest. There are three types of area
337 * that a guest can register.
340 unsigned long gpa
; /* Current guest phys addr */
341 void *pinned_addr
; /* Address in kernel linear mapping */
342 void *pinned_end
; /* End of region */
343 unsigned long next_gpa
; /* Guest phys addr for update */
344 unsigned long len
; /* Number of bytes required */
345 u8 update_pending
; /* 1 => update pinned_addr from next_gpa */
346 bool dirty
; /* true => area has been modified by kernel */
355 bool may_execute
: 1;
357 u8 page_size
; /* MMU_PAGE_xxx */
362 void (*slbmte
)(struct kvm_vcpu
*vcpu
, u64 rb
, u64 rs
);
363 u64 (*slbmfee
)(struct kvm_vcpu
*vcpu
, u64 slb_nr
);
364 u64 (*slbmfev
)(struct kvm_vcpu
*vcpu
, u64 slb_nr
);
365 void (*slbie
)(struct kvm_vcpu
*vcpu
, u64 slb_nr
);
366 void (*slbia
)(struct kvm_vcpu
*vcpu
);
368 void (*mtsrin
)(struct kvm_vcpu
*vcpu
, u32 srnum
, ulong value
);
369 u32 (*mfsrin
)(struct kvm_vcpu
*vcpu
, u32 srnum
);
370 int (*xlate
)(struct kvm_vcpu
*vcpu
, gva_t eaddr
,
371 struct kvmppc_pte
*pte
, bool data
, bool iswrite
);
372 void (*reset_msr
)(struct kvm_vcpu
*vcpu
);
373 void (*tlbie
)(struct kvm_vcpu
*vcpu
, ulong addr
, bool large
);
374 int (*esid_to_vsid
)(struct kvm_vcpu
*vcpu
, ulong esid
, u64
*vsid
);
375 u64 (*ea_to_vp
)(struct kvm_vcpu
*vcpu
, gva_t eaddr
, bool data
);
376 bool (*is_dcbz32
)(struct kvm_vcpu
*vcpu
);
388 bool large
: 1; /* PTEs are 16MB */
389 bool tb
: 1; /* 1TB segment */
391 u8 base_page_size
; /* MMU_PAGE_xxx */
394 /* Struct used to accumulate timing information in HV real mode code */
395 struct kvmhv_tb_accumulator
{
396 u64 seqcount
; /* used to synchronize access, also count * 2 */
397 u64 tb_total
; /* total time in timebase ticks */
398 u64 tb_min
; /* min time */
399 u64 tb_max
; /* max time */
402 #ifdef CONFIG_PPC_BOOK3S_64
403 struct kvmppc_irq_map
{
406 struct irq_desc
*desc
;
409 #define KVMPPC_PIRQ_MAPPED 1024
410 struct kvmppc_passthru_irqmap
{
412 struct kvmppc_irq_map mapped
[KVMPPC_PIRQ_MAPPED
];
416 # ifdef CONFIG_PPC_FSL_BOOK3E
417 #define KVMPPC_BOOKE_IAC_NUM 2
418 #define KVMPPC_BOOKE_DAC_NUM 2
420 #define KVMPPC_BOOKE_IAC_NUM 4
421 #define KVMPPC_BOOKE_DAC_NUM 2
423 #define KVMPPC_BOOKE_MAX_IAC 4
424 #define KVMPPC_BOOKE_MAX_DAC 2
426 /* KVMPPC_EPR_USER takes precedence over KVMPPC_EPR_KERNEL */
427 #define KVMPPC_EPR_NONE 0 /* EPR not supported */
428 #define KVMPPC_EPR_USER 1 /* exit to userspace to fill EPR */
429 #define KVMPPC_EPR_KERNEL 2 /* in-kernel irqchip */
431 #define KVMPPC_IRQ_DEFAULT 0
432 #define KVMPPC_IRQ_MPIC 1
433 #define KVMPPC_IRQ_XICS 2 /* Includes a XIVE option */
435 #define MMIO_HPTE_CACHE_SIZE 4
437 struct mmio_hpte_cache_entry
{
438 unsigned long hpte_v
;
439 unsigned long hpte_r
;
441 unsigned long pte_index
;
445 unsigned int slb_base_pshift
;
448 struct mmio_hpte_cache
{
449 struct mmio_hpte_cache_entry entry
[MMIO_HPTE_CACHE_SIZE
];
453 #define KVMPPC_VSX_COPY_NONE 0
454 #define KVMPPC_VSX_COPY_WORD 1
455 #define KVMPPC_VSX_COPY_DWORD 2
456 #define KVMPPC_VSX_COPY_DWORD_LOAD_DUMP 3
460 /* W0 and W1 of a XIVE thread management context */
475 struct kvm_vcpu_arch
{
478 #ifdef CONFIG_PPC_BOOK3S
479 struct kvmppc_slb slb
[64];
480 int slb_max
; /* 1 + index of last valid entry in slb[] */
481 int slb_nr
; /* total number of entries in SLB */
482 struct kvmppc_mmu mmu
;
483 struct kvmppc_vcpu_book3s
*book3s
;
485 #ifdef CONFIG_PPC_BOOK3S_32
486 struct kvmppc_book3s_shadow_vcpu
*shadow_vcpu
;
491 struct thread_fp_state fp
;
499 #ifdef CONFIG_ALTIVEC
500 struct thread_vr_state vr
;
503 #ifdef CONFIG_KVM_BOOKE_HV
513 #if defined(CONFIG_BOOKE)
514 #if defined(CONFIG_KVM_BOOKE_HV) || defined(CONFIG_64BIT)
519 #ifdef CONFIG_PPC_BOOK3S
520 /* For Gekko paired singles */
527 #ifdef CONFIG_PPC_BOOK3S
534 #ifdef CONFIG_PPC_BOOK3S
536 ulong guest_owned_ext
;
568 u32 vrsave
; /* also USPRG0 */
570 /* shadow_msr is unused for BookE HV */
583 /* Time base value when we entered the guest */
588 ulong tsr
; /* we need to perform set/clr_bits() which requires ulong */
608 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
625 struct thread_fp_state fp_tm
;
627 struct thread_vr_state vr_tm
;
628 u32 vrsave_tm
; /* also USPRG0 */
632 #ifdef CONFIG_KVM_EXIT_TIMING
633 struct mutex exit_timing_lock
;
634 struct kvmppc_exit_timing timing_exit
;
635 struct kvmppc_exit_timing timing_last_enter
;
637 u32 timing_count_type
[__NUMBER_OF_KVM_EXIT_TYPES
];
638 u64 timing_sum_duration
[__NUMBER_OF_KVM_EXIT_TYPES
];
639 u64 timing_sum_quad_duration
[__NUMBER_OF_KVM_EXIT_TYPES
];
640 u64 timing_min_duration
[__NUMBER_OF_KVM_EXIT_TYPES
];
641 u64 timing_max_duration
[__NUMBER_OF_KVM_EXIT_TYPES
];
642 u64 timing_last_exit
;
643 struct dentry
*debugfs_exit_timing
;
646 #ifdef CONFIG_PPC_BOOK3S
649 unsigned long intr_msr
;
650 ulong fault_gpa
; /* guest real address of page fault (POWER9) */
659 struct timer_list wdt_timer
;
668 /* guest debug registers*/
669 struct debug_reg dbg_reg
;
671 gpa_t paddr_accessed
;
672 gva_t vaddr_accessed
;
675 u8 io_gpr
; /* GPR used as IO source/target */
676 u8 mmio_host_swabbed
;
678 /* conversion between single and double precision */
681 * Number of simulations for vsx.
682 * If we use 2*8bytes to simulate 1*16bytes,
683 * then the number should be 2 and
684 * mmio_vsx_copy_type=KVMPPC_VSX_COPY_DWORD.
685 * If we use 4*4bytes to simulate 1*16bytes,
686 * the number should be 4 and
687 * mmio_vsx_copy_type=KVMPPC_VSX_COPY_WORD.
689 u8 mmio_vsx_copy_nums
;
691 u8 mmio_vsx_copy_type
;
692 u8 mmio_vsx_tx_sx_enabled
;
693 u8 mmio_vmx_copy_nums
;
701 u8 epr_flags
; /* KVMPPC_EPR_xxx */
704 u32 cpr0_cfgaddr
; /* holds the last set cpr0_cfgaddr */
706 struct hrtimer dec_timer
;
709 unsigned long pending_exceptions
;
713 u8 irq_pending
; /* Used by XIVE to signal pending guest irqs */
716 struct swait_queue_head
*wqp
;
717 struct kvmppc_vcore
*vcore
;
725 wait_queue_head_t cpu_run
;
726 struct machine_check_event mce_evt
; /* Valid if trap == 0x200 */
728 struct kvm_vcpu_arch_shared
*shared
;
729 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
730 bool shared_big_endian
;
732 unsigned long magic_page_pa
; /* phys addr to map the magic page to */
733 unsigned long magic_page_ea
; /* effect. addr to map the magic page to */
734 bool disable_kernel_nx
;
736 int irq_type
; /* one of KVM_IRQ_* */
738 struct openpic
*mpic
; /* KVM_IRQ_MPIC */
739 #ifdef CONFIG_KVM_XICS
740 struct kvmppc_icp
*icp
; /* XICS presentation controller */
741 struct kvmppc_xive_vcpu
*xive_vcpu
; /* XIVE virtual CPU data */
742 __be32 xive_cam_word
; /* Cooked W2 in proper endian with valid bit */
743 u8 xive_pushed
; /* Is the VP pushed on the physical CPU ? */
744 u8 xive_esc_on
; /* Is the escalation irq enabled ? */
745 union xive_tma_w01 xive_saved_state
; /* W0..1 of XIVE thread state */
746 u64 xive_esc_raddr
; /* Escalation interrupt ESB real addr */
747 u64 xive_esc_vaddr
; /* Escalation interrupt ESB virt addr */
750 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
751 struct kvm_vcpu_arch_shared shregs
;
753 struct mmio_hpte_cache mmio_cache
;
754 unsigned long pgfault_addr
;
756 unsigned long pgfault_hpte
[2];
757 struct mmio_hpte_cache_entry
*pgfault_cache
;
759 struct task_struct
*run_task
;
760 struct kvm_run
*kvm_run
;
762 spinlock_t vpa_update_lock
;
763 struct kvmppc_vpa vpa
;
764 struct kvmppc_vpa dtl
;
765 struct dtl_entry
*dtl_ptr
;
766 unsigned long dtl_index
;
768 struct kvmppc_vpa slb_shadow
;
770 spinlock_t tbacct_lock
;
777 #ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
778 struct kvmhv_tb_accumulator
*cur_activity
; /* What we're timing */
779 u64 cur_tb_start
; /* when it started */
780 struct kvmhv_tb_accumulator rm_entry
; /* real-mode entry code */
781 struct kvmhv_tb_accumulator rm_intr
; /* real-mode intr handling */
782 struct kvmhv_tb_accumulator rm_exit
; /* real-mode exit code */
783 struct kvmhv_tb_accumulator guest_time
; /* guest execution */
784 struct kvmhv_tb_accumulator cede_time
; /* time napping inside guest */
786 struct dentry
*debugfs_dir
;
787 struct dentry
*debugfs_timings
;
788 #endif /* CONFIG_KVM_BOOK3S_HV_EXIT_TIMING */
791 #define VCPU_FPR(vcpu, i) (vcpu)->arch.fp.fpr[i][TS_FPROFFSET]
792 #define VCPU_VSX_FPR(vcpu, i, j) ((vcpu)->arch.fp.fpr[i][j])
793 #define VCPU_VSX_VR(vcpu, i) ((vcpu)->arch.vr.vr[i])
795 /* Values for vcpu->arch.state */
796 #define KVMPPC_VCPU_NOTREADY 0
797 #define KVMPPC_VCPU_RUNNABLE 1
798 #define KVMPPC_VCPU_BUSY_IN_HOST 2
800 /* Values for vcpu->arch.io_gpr */
801 #define KVM_MMIO_REG_MASK 0x001f
802 #define KVM_MMIO_REG_EXT_MASK 0xffe0
803 #define KVM_MMIO_REG_GPR 0x0000
804 #define KVM_MMIO_REG_FPR 0x0020
805 #define KVM_MMIO_REG_QPR 0x0040
806 #define KVM_MMIO_REG_FQPR 0x0060
807 #define KVM_MMIO_REG_VSX 0x0080
808 #define KVM_MMIO_REG_VMX 0x00c0
810 #define __KVM_HAVE_ARCH_WQP
811 #define __KVM_HAVE_CREATE_DEVICE
813 static inline void kvm_arch_hardware_disable(void) {}
814 static inline void kvm_arch_hardware_unsetup(void) {}
815 static inline void kvm_arch_sync_events(struct kvm
*kvm
) {}
816 static inline void kvm_arch_memslots_updated(struct kvm
*kvm
, struct kvm_memslots
*slots
) {}
817 static inline void kvm_arch_flush_shadow_all(struct kvm
*kvm
) {}
818 static inline void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
) {}
819 static inline void kvm_arch_exit(void) {}
820 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
) {}
821 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
) {}
822 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu
*vcpu
) {}
824 #endif /* __POWERPC_KVM_HOST_H__ */