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>
46 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
47 #include <asm/kvm_book3s_asm.h> /* for MAX_SMT_THREADS */
48 #define KVM_MAX_VCPU_ID (MAX_SMT_THREADS * KVM_MAX_VCORES)
51 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
52 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
54 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
56 #define KVM_HALT_POLL_NS_DEFAULT 10000 /* 10 us */
58 /* These values are internal and can be increased later */
59 #define KVM_NR_IRQCHIPS 1
60 #define KVM_IRQCHIP_NUM_PINS 256
62 /* PPC-specific vcpu->requests bit members */
63 #define KVM_REQ_WATCHDOG KVM_ARCH_REQ(0)
64 #define KVM_REQ_EPR_EXIT KVM_ARCH_REQ(1)
66 #include <linux/mmu_notifier.h>
68 #define KVM_ARCH_WANT_MMU_NOTIFIER
70 extern int kvm_unmap_hva_range(struct kvm *kvm,
71 unsigned long start, unsigned long end,
73 extern int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
74 extern int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
75 extern void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
77 #define HPTEG_CACHE_NUM (1 << 15)
78 #define HPTEG_HASH_BITS_PTE 13
79 #define HPTEG_HASH_BITS_PTE_LONG 12
80 #define HPTEG_HASH_BITS_VPTE 13
81 #define HPTEG_HASH_BITS_VPTE_LONG 5
82 #define HPTEG_HASH_BITS_VPTE_64K 11
83 #define HPTEG_HASH_NUM_PTE (1 << HPTEG_HASH_BITS_PTE)
84 #define HPTEG_HASH_NUM_PTE_LONG (1 << HPTEG_HASH_BITS_PTE_LONG)
85 #define HPTEG_HASH_NUM_VPTE (1 << HPTEG_HASH_BITS_VPTE)
86 #define HPTEG_HASH_NUM_VPTE_LONG (1 << HPTEG_HASH_BITS_VPTE_LONG)
87 #define HPTEG_HASH_NUM_VPTE_64K (1 << HPTEG_HASH_BITS_VPTE_64K)
89 /* Physical Address Mask - allowed range of real mode RAM access */
90 #define KVM_PAM 0x0fffffffffffffffULL
96 struct kvmppc_vcpu_book3s;
97 struct kvmppc_book3s_shadow_vcpu;
100 ulong remote_tlb_flush;
103 struct kvm_vcpu_stat {
108 /* Account for special types of light exits: */
109 u64 itlb_real_miss_exits;
110 u64 itlb_virt_miss_exits;
111 u64 dtlb_real_miss_exits;
112 u64 dtlb_virt_miss_exits;
116 u64 emulated_inst_exits;
119 u64 halt_poll_success_ns;
120 u64 halt_poll_fail_ns;
122 u64 halt_successful_poll;
123 u64 halt_attempted_poll;
124 u64 halt_successful_wait;
125 u64 halt_poll_invalid;
131 #ifdef CONFIG_PPC_BOOK3S
145 enum kvm_exit_types {
148 ITLB_REAL_MISS_EXITS,
149 ITLB_VIRT_MISS_EXITS,
150 DTLB_REAL_MISS_EXITS,
151 DTLB_VIRT_MISS_EXITS,
156 EMULATED_MTMSRWE_EXITS,
157 EMULATED_WRTEE_EXITS,
158 EMULATED_MTSPR_EXITS,
159 EMULATED_MFSPR_EXITS,
160 EMULATED_MTMSR_EXITS,
161 EMULATED_MFMSR_EXITS,
162 EMULATED_TLBSX_EXITS,
163 EMULATED_TLBWE_EXITS,
176 __NUMBER_OF_KVM_EXIT_TYPES
179 /* allow access to big endian 32bit upper/lower parts and 64bit var */
180 struct kvmppc_exit_timing {
189 struct kvmppc_pginfo {
194 struct kvmppc_spapr_tce_iommu_table {
196 struct list_head next;
197 struct iommu_table *tbl;
201 struct kvmppc_spapr_tce_table {
202 struct list_head list;
207 u64 offset; /* in pages */
208 u64 size; /* window size in pages */
209 struct list_head iommu_tables;
210 struct page *pages[0];
213 /* XICS components, defined in book3s_xics.c */
216 extern struct kvm_device_ops kvm_xics_ops;
218 /* XIVE components, defined in book3s_xive.c */
220 struct kvmppc_xive_vcpu;
221 extern struct kvm_device_ops kvm_xive_ops;
223 struct kvmppc_passthru_irqmap;
226 * The reverse mapping array has one entry for each HPTE,
227 * which stores the guest's view of the second word of the HPTE
228 * (including the guest physical address of the mapping),
229 * plus forward and backward pointers in a doubly-linked ring
230 * of HPTEs that map the same host page. The pointers in this
231 * ring are 32-bit HPTE indexes, to save space.
233 struct revmap_entry {
234 unsigned long guest_rpte;
235 unsigned int forw, back;
239 * We use the top bit of each memslot->arch.rmap entry as a lock bit,
240 * and bit 32 as a present flag. The bottom 32 bits are the
241 * index in the guest HPT of a HPTE that points to the page.
243 #define KVMPPC_RMAP_LOCK_BIT 63
244 #define KVMPPC_RMAP_RC_SHIFT 32
245 #define KVMPPC_RMAP_REFERENCED (HPTE_R_R << KVMPPC_RMAP_RC_SHIFT)
246 #define KVMPPC_RMAP_PRESENT 0x100000000ul
247 #define KVMPPC_RMAP_INDEX 0xfffffffful
249 struct kvm_arch_memory_slot {
250 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
252 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
255 struct kvm_hpt_info {
256 /* Host virtual (linear mapping) address of guest HPT */
258 /* Array of reverse mapping entries for each guest HPTE */
259 struct revmap_entry *rev;
260 /* Guest HPT size is 2**(order) bytes */
262 /* 1 if HPT allocated with CMA, 0 otherwise */
266 struct kvm_resize_hpt;
270 unsigned int smt_mode; /* # vcpus per virtual core */
271 unsigned int emul_smt_mode; /* emualted SMT mode, on P9 */
272 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
273 unsigned int tlb_sets;
274 struct kvm_hpt_info hpt;
275 atomic64_t mmio_update;
276 unsigned int host_lpid;
277 unsigned long host_lpcr;
279 unsigned long host_sdr1;
281 unsigned long vrma_slb_v;
283 atomic_t vcpus_running;
285 atomic_t hpte_mod_interest;
286 cpumask_t need_tlb_flush;
287 cpumask_t cpu_in_guest;
293 struct dentry *debugfs_dir;
294 struct dentry *htab_dentry;
295 struct kvm_resize_hpt *resize_hpt; /* protected by kvm->lock */
296 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
297 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
298 struct mutex hpt_mutex;
300 #ifdef CONFIG_PPC_BOOK3S_64
301 struct list_head spapr_tce_tables;
302 struct list_head rtas_tokens;
303 struct mutex rtas_token_lock;
304 DECLARE_BITMAP(enabled_hcalls, MAX_HCALL_OPCODE/4 + 1);
306 #ifdef CONFIG_KVM_MPIC
307 struct openpic *mpic;
309 #ifdef CONFIG_KVM_XICS
310 struct kvmppc_xics *xics;
311 struct kvmppc_xive *xive;
312 struct kvmppc_passthru_irqmap *pimap;
314 struct kvmppc_ops *kvm_ops;
315 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
316 /* This array can grow quite large, keep it at the end */
317 struct kvmppc_vcore *vcores[KVM_MAX_VCORES];
321 #define VCORE_ENTRY_MAP(vc) ((vc)->entry_exit_map & 0xff)
322 #define VCORE_EXIT_MAP(vc) ((vc)->entry_exit_map >> 8)
323 #define VCORE_IS_EXITING(vc) (VCORE_EXIT_MAP(vc) != 0)
325 /* This bit is used when a vcore exit is triggered from outside the vcore */
326 #define VCORE_EXIT_REQ 0x10000
329 * Values for vcore_state.
330 * Note that these are arranged such that lower values
331 * (< VCORE_SLEEPING) don't require stolen time accounting
332 * on load/unload, and higher values do.
334 #define VCORE_INACTIVE 0
335 #define VCORE_PREEMPT 1
336 #define VCORE_PIGGYBACK 2
337 #define VCORE_SLEEPING 3
338 #define VCORE_RUNNING 4
339 #define VCORE_EXITING 5
340 #define VCORE_POLLING 6
343 * Struct used to manage memory for a virtual processor area
344 * registered by a PAPR guest. There are three types of area
345 * that a guest can register.
348 unsigned long gpa; /* Current guest phys addr */
349 void *pinned_addr; /* Address in kernel linear mapping */
350 void *pinned_end; /* End of region */
351 unsigned long next_gpa; /* Guest phys addr for update */
352 unsigned long len; /* Number of bytes required */
353 u8 update_pending; /* 1 => update pinned_addr from next_gpa */
354 bool dirty; /* true => area has been modified by kernel */
363 bool may_execute : 1;
365 u8 page_size; /* MMU_PAGE_xxx */
370 void (*slbmte)(struct kvm_vcpu *vcpu, u64 rb, u64 rs);
371 u64 (*slbmfee)(struct kvm_vcpu *vcpu, u64 slb_nr);
372 u64 (*slbmfev)(struct kvm_vcpu *vcpu, u64 slb_nr);
373 void (*slbie)(struct kvm_vcpu *vcpu, u64 slb_nr);
374 void (*slbia)(struct kvm_vcpu *vcpu);
376 void (*mtsrin)(struct kvm_vcpu *vcpu, u32 srnum, ulong value);
377 u32 (*mfsrin)(struct kvm_vcpu *vcpu, u32 srnum);
378 int (*xlate)(struct kvm_vcpu *vcpu, gva_t eaddr,
379 struct kvmppc_pte *pte, bool data, bool iswrite);
380 void (*reset_msr)(struct kvm_vcpu *vcpu);
381 void (*tlbie)(struct kvm_vcpu *vcpu, ulong addr, bool large);
382 int (*esid_to_vsid)(struct kvm_vcpu *vcpu, ulong esid, u64 *vsid);
383 u64 (*ea_to_vp)(struct kvm_vcpu *vcpu, gva_t eaddr, bool data);
384 bool (*is_dcbz32)(struct kvm_vcpu *vcpu);
396 bool large : 1; /* PTEs are 16MB */
397 bool tb : 1; /* 1TB segment */
399 u8 base_page_size; /* MMU_PAGE_xxx */
402 /* Struct used to accumulate timing information in HV real mode code */
403 struct kvmhv_tb_accumulator {
404 u64 seqcount; /* used to synchronize access, also count * 2 */
405 u64 tb_total; /* total time in timebase ticks */
406 u64 tb_min; /* min time */
407 u64 tb_max; /* max time */
410 #ifdef CONFIG_PPC_BOOK3S_64
411 struct kvmppc_irq_map {
414 struct irq_desc *desc;
417 #define KVMPPC_PIRQ_MAPPED 1024
418 struct kvmppc_passthru_irqmap {
420 struct kvmppc_irq_map mapped[KVMPPC_PIRQ_MAPPED];
424 # ifdef CONFIG_PPC_FSL_BOOK3E
425 #define KVMPPC_BOOKE_IAC_NUM 2
426 #define KVMPPC_BOOKE_DAC_NUM 2
428 #define KVMPPC_BOOKE_IAC_NUM 4
429 #define KVMPPC_BOOKE_DAC_NUM 2
431 #define KVMPPC_BOOKE_MAX_IAC 4
432 #define KVMPPC_BOOKE_MAX_DAC 2
434 /* KVMPPC_EPR_USER takes precedence over KVMPPC_EPR_KERNEL */
435 #define KVMPPC_EPR_NONE 0 /* EPR not supported */
436 #define KVMPPC_EPR_USER 1 /* exit to userspace to fill EPR */
437 #define KVMPPC_EPR_KERNEL 2 /* in-kernel irqchip */
439 #define KVMPPC_IRQ_DEFAULT 0
440 #define KVMPPC_IRQ_MPIC 1
441 #define KVMPPC_IRQ_XICS 2 /* Includes a XIVE option */
443 #define MMIO_HPTE_CACHE_SIZE 4
445 struct mmio_hpte_cache_entry {
446 unsigned long hpte_v;
447 unsigned long hpte_r;
449 unsigned long pte_index;
453 unsigned int slb_base_pshift;
456 struct mmio_hpte_cache {
457 struct mmio_hpte_cache_entry entry[MMIO_HPTE_CACHE_SIZE];
461 #define KVMPPC_VSX_COPY_NONE 0
462 #define KVMPPC_VSX_COPY_WORD 1
463 #define KVMPPC_VSX_COPY_DWORD 2
464 #define KVMPPC_VSX_COPY_DWORD_LOAD_DUMP 3
465 #define KVMPPC_VSX_COPY_WORD_LOAD_DUMP 4
467 #define KVMPPC_VMX_COPY_BYTE 8
468 #define KVMPPC_VMX_COPY_HWORD 9
469 #define KVMPPC_VMX_COPY_WORD 10
470 #define KVMPPC_VMX_COPY_DWORD 11
474 /* W0 and W1 of a XIVE thread management context */
489 struct kvm_vcpu_arch {
492 #ifdef CONFIG_PPC_BOOK3S
493 struct kvmppc_slb slb[64];
494 int slb_max; /* 1 + index of last valid entry in slb[] */
495 int slb_nr; /* total number of entries in SLB */
496 struct kvmppc_mmu mmu;
497 struct kvmppc_vcpu_book3s *book3s;
499 #ifdef CONFIG_PPC_BOOK3S_32
500 struct kvmppc_book3s_shadow_vcpu *shadow_vcpu;
505 struct thread_fp_state fp;
513 #ifdef CONFIG_ALTIVEC
514 struct thread_vr_state vr;
517 #ifdef CONFIG_KVM_BOOKE_HV
527 #if defined(CONFIG_BOOKE)
528 #if defined(CONFIG_KVM_BOOKE_HV) || defined(CONFIG_64BIT)
533 #ifdef CONFIG_PPC_BOOK3S
534 /* For Gekko paired singles */
538 #ifdef CONFIG_PPC_BOOK3S
542 #ifdef CONFIG_PPC_BOOK3S
544 ulong guest_owned_ext;
576 u32 vrsave; /* also USPRG0 */
578 /* shadow_msr is unused for BookE HV */
591 /* Time base value when we entered the guest */
596 ulong tsr; /* we need to perform set/clr_bits() which requires ulong */
616 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
633 struct thread_fp_state fp_tm;
635 struct thread_vr_state vr_tm;
636 u32 vrsave_tm; /* also USPRG0 */
639 #ifdef CONFIG_KVM_EXIT_TIMING
640 struct mutex exit_timing_lock;
641 struct kvmppc_exit_timing timing_exit;
642 struct kvmppc_exit_timing timing_last_enter;
644 u32 timing_count_type[__NUMBER_OF_KVM_EXIT_TYPES];
645 u64 timing_sum_duration[__NUMBER_OF_KVM_EXIT_TYPES];
646 u64 timing_sum_quad_duration[__NUMBER_OF_KVM_EXIT_TYPES];
647 u64 timing_min_duration[__NUMBER_OF_KVM_EXIT_TYPES];
648 u64 timing_max_duration[__NUMBER_OF_KVM_EXIT_TYPES];
649 u64 timing_last_exit;
650 struct dentry *debugfs_exit_timing;
653 #ifdef CONFIG_PPC_BOOK3S
656 unsigned long intr_msr;
657 ulong fault_gpa; /* guest real address of page fault (POWER9) */
666 struct timer_list wdt_timer;
675 /* guest debug registers*/
676 struct debug_reg dbg_reg;
678 gpa_t paddr_accessed;
679 gva_t vaddr_accessed;
682 u16 io_gpr; /* GPR used as IO source/target */
683 u8 mmio_host_swabbed;
685 /* conversion between single and double precision */
688 * Number of simulations for vsx.
689 * If we use 2*8bytes to simulate 1*16bytes,
690 * then the number should be 2 and
691 * mmio_copy_type=KVMPPC_VSX_COPY_DWORD.
692 * If we use 4*4bytes to simulate 1*16bytes,
693 * the number should be 4 and
694 * mmio_vsx_copy_type=KVMPPC_VSX_COPY_WORD.
696 u8 mmio_vsx_copy_nums;
698 u8 mmio_vmx_copy_nums;
708 u8 epr_flags; /* KVMPPC_EPR_xxx */
711 u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */
713 struct hrtimer dec_timer;
716 unsigned long pending_exceptions;
720 u8 irq_pending; /* Used by XIVE to signal pending guest irqs */
723 struct swait_queue_head *wqp;
724 struct kvmppc_vcore *vcore;
732 wait_queue_head_t cpu_run;
733 struct machine_check_event mce_evt; /* Valid if trap == 0x200 */
735 struct kvm_vcpu_arch_shared *shared;
736 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
737 bool shared_big_endian;
739 unsigned long magic_page_pa; /* phys addr to map the magic page to */
740 unsigned long magic_page_ea; /* effect. addr to map the magic page to */
741 bool disable_kernel_nx;
743 int irq_type; /* one of KVM_IRQ_* */
745 struct openpic *mpic; /* KVM_IRQ_MPIC */
746 #ifdef CONFIG_KVM_XICS
747 struct kvmppc_icp *icp; /* XICS presentation controller */
748 struct kvmppc_xive_vcpu *xive_vcpu; /* XIVE virtual CPU data */
749 __be32 xive_cam_word; /* Cooked W2 in proper endian with valid bit */
750 u8 xive_pushed; /* Is the VP pushed on the physical CPU ? */
751 u8 xive_esc_on; /* Is the escalation irq enabled ? */
752 union xive_tma_w01 xive_saved_state; /* W0..1 of XIVE thread state */
753 u64 xive_esc_raddr; /* Escalation interrupt ESB real addr */
754 u64 xive_esc_vaddr; /* Escalation interrupt ESB virt addr */
757 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
758 struct kvm_vcpu_arch_shared shregs;
760 struct mmio_hpte_cache mmio_cache;
761 unsigned long pgfault_addr;
763 unsigned long pgfault_hpte[2];
764 struct mmio_hpte_cache_entry *pgfault_cache;
766 struct task_struct *run_task;
767 struct kvm_run *kvm_run;
769 spinlock_t vpa_update_lock;
770 struct kvmppc_vpa vpa;
771 struct kvmppc_vpa dtl;
772 struct dtl_entry *dtl_ptr;
773 unsigned long dtl_index;
775 struct kvmppc_vpa slb_shadow;
777 spinlock_t tbacct_lock;
786 #ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
787 struct kvmhv_tb_accumulator *cur_activity; /* What we're timing */
788 u64 cur_tb_start; /* when it started */
789 struct kvmhv_tb_accumulator rm_entry; /* real-mode entry code */
790 struct kvmhv_tb_accumulator rm_intr; /* real-mode intr handling */
791 struct kvmhv_tb_accumulator rm_exit; /* real-mode exit code */
792 struct kvmhv_tb_accumulator guest_time; /* guest execution */
793 struct kvmhv_tb_accumulator cede_time; /* time napping inside guest */
795 struct dentry *debugfs_dir;
796 struct dentry *debugfs_timings;
797 #endif /* CONFIG_KVM_BOOK3S_HV_EXIT_TIMING */
800 #define VCPU_FPR(vcpu, i) (vcpu)->arch.fp.fpr[i][TS_FPROFFSET]
801 #define VCPU_VSX_FPR(vcpu, i, j) ((vcpu)->arch.fp.fpr[i][j])
802 #define VCPU_VSX_VR(vcpu, i) ((vcpu)->arch.vr.vr[i])
804 /* Values for vcpu->arch.state */
805 #define KVMPPC_VCPU_NOTREADY 0
806 #define KVMPPC_VCPU_RUNNABLE 1
807 #define KVMPPC_VCPU_BUSY_IN_HOST 2
809 /* Values for vcpu->arch.io_gpr */
810 #define KVM_MMIO_REG_MASK 0x003f
811 #define KVM_MMIO_REG_EXT_MASK 0xffc0
812 #define KVM_MMIO_REG_GPR 0x0000
813 #define KVM_MMIO_REG_FPR 0x0040
814 #define KVM_MMIO_REG_QPR 0x0080
815 #define KVM_MMIO_REG_FQPR 0x00c0
816 #define KVM_MMIO_REG_VSX 0x0100
817 #define KVM_MMIO_REG_VMX 0x0180
819 #define __KVM_HAVE_ARCH_WQP
820 #define __KVM_HAVE_CREATE_DEVICE
822 static inline void kvm_arch_hardware_disable(void) {}
823 static inline void kvm_arch_hardware_unsetup(void) {}
824 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
825 static inline void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen) {}
826 static inline void kvm_arch_flush_shadow_all(struct kvm *kvm) {}
827 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
828 static inline void kvm_arch_exit(void) {}
829 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
830 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
831 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
833 #endif /* __POWERPC_KVM_HOST_H__ */