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>
39 #define KVM_MAX_VCPUS NR_CPUS
40 #define KVM_MAX_VCORES NR_CPUS
41 #define KVM_USER_MEM_SLOTS 512
43 #include <asm/cputhreads.h>
44 #define KVM_MAX_VCPU_ID (threads_per_subcore * KVM_MAX_VCORES)
46 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
48 #define KVM_HALT_POLL_NS_DEFAULT 10000 /* 10 us */
50 /* These values are internal and can be increased later */
51 #define KVM_NR_IRQCHIPS 1
52 #define KVM_IRQCHIP_NUM_PINS 256
54 /* PPC-specific vcpu->requests bit members */
55 #define KVM_REQ_WATCHDOG 8
56 #define KVM_REQ_EPR_EXIT 9
58 #include <linux/mmu_notifier.h>
60 #define KVM_ARCH_WANT_MMU_NOTIFIER
62 extern int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
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 static inline void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
70 unsigned long address)
74 #define HPTEG_CACHE_NUM (1 << 15)
75 #define HPTEG_HASH_BITS_PTE 13
76 #define HPTEG_HASH_BITS_PTE_LONG 12
77 #define HPTEG_HASH_BITS_VPTE 13
78 #define HPTEG_HASH_BITS_VPTE_LONG 5
79 #define HPTEG_HASH_BITS_VPTE_64K 11
80 #define HPTEG_HASH_NUM_PTE (1 << HPTEG_HASH_BITS_PTE)
81 #define HPTEG_HASH_NUM_PTE_LONG (1 << HPTEG_HASH_BITS_PTE_LONG)
82 #define HPTEG_HASH_NUM_VPTE (1 << HPTEG_HASH_BITS_VPTE)
83 #define HPTEG_HASH_NUM_VPTE_LONG (1 << HPTEG_HASH_BITS_VPTE_LONG)
84 #define HPTEG_HASH_NUM_VPTE_64K (1 << HPTEG_HASH_BITS_VPTE_64K)
86 /* Physical Address Mask - allowed range of real mode RAM access */
87 #define KVM_PAM 0x0fffffffffffffffULL
93 struct kvmppc_vcpu_book3s;
94 struct kvmppc_book3s_shadow_vcpu;
97 ulong remote_tlb_flush;
100 struct kvm_vcpu_stat {
105 /* Account for special types of light exits: */
106 u64 itlb_real_miss_exits;
107 u64 itlb_virt_miss_exits;
108 u64 dtlb_real_miss_exits;
109 u64 dtlb_virt_miss_exits;
113 u64 emulated_inst_exits;
116 u64 halt_poll_success_ns;
117 u64 halt_poll_fail_ns;
119 u64 halt_successful_poll;
120 u64 halt_attempted_poll;
121 u64 halt_successful_wait;
122 u64 halt_poll_invalid;
128 #ifdef CONFIG_PPC_BOOK3S
142 enum kvm_exit_types {
145 ITLB_REAL_MISS_EXITS,
146 ITLB_VIRT_MISS_EXITS,
147 DTLB_REAL_MISS_EXITS,
148 DTLB_VIRT_MISS_EXITS,
153 EMULATED_MTMSRWE_EXITS,
154 EMULATED_WRTEE_EXITS,
155 EMULATED_MTSPR_EXITS,
156 EMULATED_MFSPR_EXITS,
157 EMULATED_MTMSR_EXITS,
158 EMULATED_MFMSR_EXITS,
159 EMULATED_TLBSX_EXITS,
160 EMULATED_TLBWE_EXITS,
173 __NUMBER_OF_KVM_EXIT_TYPES
176 /* allow access to big endian 32bit upper/lower parts and 64bit var */
177 struct kvmppc_exit_timing {
186 struct kvmppc_pginfo {
191 struct kvmppc_spapr_tce_iommu_table {
193 struct list_head next;
194 struct iommu_table *tbl;
198 struct kvmppc_spapr_tce_table {
199 struct list_head list;
204 u64 offset; /* in pages */
205 u64 size; /* window size in pages */
206 struct list_head iommu_tables;
207 struct page *pages[0];
210 /* XICS components, defined in book3s_xics.c */
213 extern struct kvm_device_ops kvm_xics_ops;
215 /* XIVE components, defined in book3s_xive.c */
217 struct kvmppc_xive_vcpu;
218 extern struct kvm_device_ops kvm_xive_ops;
220 struct kvmppc_passthru_irqmap;
223 * The reverse mapping array has one entry for each HPTE,
224 * which stores the guest's view of the second word of the HPTE
225 * (including the guest physical address of the mapping),
226 * plus forward and backward pointers in a doubly-linked ring
227 * of HPTEs that map the same host page. The pointers in this
228 * ring are 32-bit HPTE indexes, to save space.
230 struct revmap_entry {
231 unsigned long guest_rpte;
232 unsigned int forw, back;
236 * We use the top bit of each memslot->arch.rmap entry as a lock bit,
237 * and bit 32 as a present flag. The bottom 32 bits are the
238 * index in the guest HPT of a HPTE that points to the page.
240 #define KVMPPC_RMAP_LOCK_BIT 63
241 #define KVMPPC_RMAP_RC_SHIFT 32
242 #define KVMPPC_RMAP_CHG_SHIFT 48
243 #define KVMPPC_RMAP_REFERENCED (HPTE_R_R << KVMPPC_RMAP_RC_SHIFT)
244 #define KVMPPC_RMAP_CHANGED (HPTE_R_C << KVMPPC_RMAP_RC_SHIFT)
245 #define KVMPPC_RMAP_CHG_ORDER (0x3ful << KVMPPC_RMAP_CHG_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 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
271 unsigned int tlb_sets;
272 struct kvm_hpt_info hpt;
273 atomic64_t mmio_update;
274 unsigned int host_lpid;
275 unsigned long host_lpcr;
277 unsigned long host_sdr1;
280 unsigned long vrma_slb_v;
282 atomic_t vcpus_running;
284 atomic_t hpte_mod_interest;
285 cpumask_t need_tlb_flush;
286 cpumask_t cpu_in_guest;
290 struct dentry *debugfs_dir;
291 struct dentry *htab_dentry;
292 struct kvm_resize_hpt *resize_hpt; /* protected by kvm->lock */
293 #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */
294 #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
295 struct mutex hpt_mutex;
297 #ifdef CONFIG_PPC_BOOK3S_64
298 struct list_head spapr_tce_tables;
299 struct list_head rtas_tokens;
300 DECLARE_BITMAP(enabled_hcalls, MAX_HCALL_OPCODE/4 + 1);
302 #ifdef CONFIG_KVM_MPIC
303 struct openpic *mpic;
305 #ifdef CONFIG_KVM_XICS
306 struct kvmppc_xics *xics;
307 struct kvmppc_xive *xive;
308 struct kvmppc_passthru_irqmap *pimap;
310 struct kvmppc_ops *kvm_ops;
311 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
312 /* This array can grow quite large, keep it at the end */
313 struct kvmppc_vcore *vcores[KVM_MAX_VCORES];
317 #define VCORE_ENTRY_MAP(vc) ((vc)->entry_exit_map & 0xff)
318 #define VCORE_EXIT_MAP(vc) ((vc)->entry_exit_map >> 8)
319 #define VCORE_IS_EXITING(vc) (VCORE_EXIT_MAP(vc) != 0)
321 /* This bit is used when a vcore exit is triggered from outside the vcore */
322 #define VCORE_EXIT_REQ 0x10000
325 * Values for vcore_state.
326 * Note that these are arranged such that lower values
327 * (< VCORE_SLEEPING) don't require stolen time accounting
328 * on load/unload, and higher values do.
330 #define VCORE_INACTIVE 0
331 #define VCORE_PREEMPT 1
332 #define VCORE_PIGGYBACK 2
333 #define VCORE_SLEEPING 3
334 #define VCORE_RUNNING 4
335 #define VCORE_EXITING 5
336 #define VCORE_POLLING 6
339 * Struct used to manage memory for a virtual processor area
340 * registered by a PAPR guest. There are three types of area
341 * that a guest can register.
344 unsigned long gpa; /* Current guest phys addr */
345 void *pinned_addr; /* Address in kernel linear mapping */
346 void *pinned_end; /* End of region */
347 unsigned long next_gpa; /* Guest phys addr for update */
348 unsigned long len; /* Number of bytes required */
349 u8 update_pending; /* 1 => update pinned_addr from next_gpa */
350 bool dirty; /* true => area has been modified by kernel */
359 bool may_execute : 1;
361 u8 page_size; /* MMU_PAGE_xxx */
366 void (*slbmte)(struct kvm_vcpu *vcpu, u64 rb, u64 rs);
367 u64 (*slbmfee)(struct kvm_vcpu *vcpu, u64 slb_nr);
368 u64 (*slbmfev)(struct kvm_vcpu *vcpu, u64 slb_nr);
369 void (*slbie)(struct kvm_vcpu *vcpu, u64 slb_nr);
370 void (*slbia)(struct kvm_vcpu *vcpu);
372 void (*mtsrin)(struct kvm_vcpu *vcpu, u32 srnum, ulong value);
373 u32 (*mfsrin)(struct kvm_vcpu *vcpu, u32 srnum);
374 int (*xlate)(struct kvm_vcpu *vcpu, gva_t eaddr,
375 struct kvmppc_pte *pte, bool data, bool iswrite);
376 void (*reset_msr)(struct kvm_vcpu *vcpu);
377 void (*tlbie)(struct kvm_vcpu *vcpu, ulong addr, bool large);
378 int (*esid_to_vsid)(struct kvm_vcpu *vcpu, ulong esid, u64 *vsid);
379 u64 (*ea_to_vp)(struct kvm_vcpu *vcpu, gva_t eaddr, bool data);
380 bool (*is_dcbz32)(struct kvm_vcpu *vcpu);
392 bool large : 1; /* PTEs are 16MB */
393 bool tb : 1; /* 1TB segment */
395 u8 base_page_size; /* MMU_PAGE_xxx */
398 /* Struct used to accumulate timing information in HV real mode code */
399 struct kvmhv_tb_accumulator {
400 u64 seqcount; /* used to synchronize access, also count * 2 */
401 u64 tb_total; /* total time in timebase ticks */
402 u64 tb_min; /* min time */
403 u64 tb_max; /* max time */
406 #ifdef CONFIG_PPC_BOOK3S_64
407 struct kvmppc_irq_map {
410 struct irq_desc *desc;
413 #define KVMPPC_PIRQ_MAPPED 1024
414 struct kvmppc_passthru_irqmap {
416 struct kvmppc_irq_map mapped[KVMPPC_PIRQ_MAPPED];
420 # ifdef CONFIG_PPC_FSL_BOOK3E
421 #define KVMPPC_BOOKE_IAC_NUM 2
422 #define KVMPPC_BOOKE_DAC_NUM 2
424 #define KVMPPC_BOOKE_IAC_NUM 4
425 #define KVMPPC_BOOKE_DAC_NUM 2
427 #define KVMPPC_BOOKE_MAX_IAC 4
428 #define KVMPPC_BOOKE_MAX_DAC 2
430 /* KVMPPC_EPR_USER takes precedence over KVMPPC_EPR_KERNEL */
431 #define KVMPPC_EPR_NONE 0 /* EPR not supported */
432 #define KVMPPC_EPR_USER 1 /* exit to userspace to fill EPR */
433 #define KVMPPC_EPR_KERNEL 2 /* in-kernel irqchip */
435 #define KVMPPC_IRQ_DEFAULT 0
436 #define KVMPPC_IRQ_MPIC 1
437 #define KVMPPC_IRQ_XICS 2 /* Includes a XIVE option */
439 #define MMIO_HPTE_CACHE_SIZE 4
441 struct mmio_hpte_cache_entry {
442 unsigned long hpte_v;
443 unsigned long hpte_r;
445 unsigned long pte_index;
449 unsigned int slb_base_pshift;
452 struct mmio_hpte_cache {
453 struct mmio_hpte_cache_entry entry[MMIO_HPTE_CACHE_SIZE];
457 #define KVMPPC_VSX_COPY_NONE 0
458 #define KVMPPC_VSX_COPY_WORD 1
459 #define KVMPPC_VSX_COPY_DWORD 2
460 #define KVMPPC_VSX_COPY_DWORD_LOAD_DUMP 3
464 /* W0 and W1 of a XIVE thread management context */
479 struct kvm_vcpu_arch {
482 #ifdef CONFIG_PPC_BOOK3S
483 struct kvmppc_slb slb[64];
484 int slb_max; /* 1 + index of last valid entry in slb[] */
485 int slb_nr; /* total number of entries in SLB */
486 struct kvmppc_mmu mmu;
487 struct kvmppc_vcpu_book3s *book3s;
489 #ifdef CONFIG_PPC_BOOK3S_32
490 struct kvmppc_book3s_shadow_vcpu *shadow_vcpu;
495 struct thread_fp_state fp;
503 #ifdef CONFIG_ALTIVEC
504 struct thread_vr_state vr;
507 #ifdef CONFIG_KVM_BOOKE_HV
517 #if defined(CONFIG_BOOKE)
518 #if defined(CONFIG_KVM_BOOKE_HV) || defined(CONFIG_64BIT)
523 #ifdef CONFIG_PPC_BOOK3S
524 /* For Gekko paired singles */
531 #ifdef CONFIG_PPC_BOOK3S
538 #ifdef CONFIG_PPC_BOOK3S
540 ulong guest_owned_ext;
571 u32 vrsave; /* also USPRG0 */
573 /* shadow_msr is unused for BookE HV */
586 /* Time base value when we entered the guest */
591 ulong tsr; /* we need to perform set/clr_bits() which requires ulong */
611 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
627 struct thread_fp_state fp_tm;
629 struct thread_vr_state vr_tm;
630 u32 vrsave_tm; /* also USPRG0 */
634 #ifdef CONFIG_KVM_EXIT_TIMING
635 struct mutex exit_timing_lock;
636 struct kvmppc_exit_timing timing_exit;
637 struct kvmppc_exit_timing timing_last_enter;
639 u32 timing_count_type[__NUMBER_OF_KVM_EXIT_TYPES];
640 u64 timing_sum_duration[__NUMBER_OF_KVM_EXIT_TYPES];
641 u64 timing_sum_quad_duration[__NUMBER_OF_KVM_EXIT_TYPES];
642 u64 timing_min_duration[__NUMBER_OF_KVM_EXIT_TYPES];
643 u64 timing_max_duration[__NUMBER_OF_KVM_EXIT_TYPES];
644 u64 timing_last_exit;
645 struct dentry *debugfs_exit_timing;
648 #ifdef CONFIG_PPC_BOOK3S
651 unsigned long intr_msr;
652 ulong fault_gpa; /* guest real address of page fault (POWER9) */
661 struct timer_list wdt_timer;
670 /* guest debug registers*/
671 struct debug_reg dbg_reg;
673 gpa_t paddr_accessed;
674 gva_t vaddr_accessed;
677 u8 io_gpr; /* GPR used as IO source/target */
678 u8 mmio_host_swabbed;
680 /* conversion between single and double precision */
683 * Number of simulations for vsx.
684 * If we use 2*8bytes to simulate 1*16bytes,
685 * then the number should be 2 and
686 * mmio_vsx_copy_type=KVMPPC_VSX_COPY_DWORD.
687 * If we use 4*4bytes to simulate 1*16bytes,
688 * the number should be 4 and
689 * mmio_vsx_copy_type=KVMPPC_VSX_COPY_WORD.
691 u8 mmio_vsx_copy_nums;
693 u8 mmio_vsx_copy_type;
694 u8 mmio_vsx_tx_sx_enabled;
702 u8 epr_flags; /* KVMPPC_EPR_xxx */
705 u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */
707 struct hrtimer dec_timer;
710 unsigned long pending_exceptions;
715 struct swait_queue_head *wqp;
716 struct kvmppc_vcore *vcore;
724 wait_queue_head_t cpu_run;
726 struct kvm_vcpu_arch_shared *shared;
727 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
728 bool shared_big_endian;
730 unsigned long magic_page_pa; /* phys addr to map the magic page to */
731 unsigned long magic_page_ea; /* effect. addr to map the magic page to */
732 bool disable_kernel_nx;
734 int irq_type; /* one of KVM_IRQ_* */
736 struct openpic *mpic; /* KVM_IRQ_MPIC */
737 #ifdef CONFIG_KVM_XICS
738 struct kvmppc_icp *icp; /* XICS presentation controller */
739 struct kvmppc_xive_vcpu *xive_vcpu; /* XIVE virtual CPU data */
740 __be32 xive_cam_word; /* Cooked W2 in proper endian with valid bit */
741 u32 xive_pushed; /* Is the VP pushed on the physical CPU ? */
742 union xive_tma_w01 xive_saved_state; /* W0..1 of XIVE thread state */
745 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
746 struct kvm_vcpu_arch_shared shregs;
748 struct mmio_hpte_cache mmio_cache;
749 unsigned long pgfault_addr;
751 unsigned long pgfault_hpte[2];
752 struct mmio_hpte_cache_entry *pgfault_cache;
754 struct task_struct *run_task;
755 struct kvm_run *kvm_run;
757 spinlock_t vpa_update_lock;
758 struct kvmppc_vpa vpa;
759 struct kvmppc_vpa dtl;
760 struct dtl_entry *dtl_ptr;
761 unsigned long dtl_index;
763 struct kvmppc_vpa slb_shadow;
765 spinlock_t tbacct_lock;
772 #ifdef CONFIG_KVM_BOOK3S_HV_EXIT_TIMING
773 struct kvmhv_tb_accumulator *cur_activity; /* What we're timing */
774 u64 cur_tb_start; /* when it started */
775 struct kvmhv_tb_accumulator rm_entry; /* real-mode entry code */
776 struct kvmhv_tb_accumulator rm_intr; /* real-mode intr handling */
777 struct kvmhv_tb_accumulator rm_exit; /* real-mode exit code */
778 struct kvmhv_tb_accumulator guest_time; /* guest execution */
779 struct kvmhv_tb_accumulator cede_time; /* time napping inside guest */
781 struct dentry *debugfs_dir;
782 struct dentry *debugfs_timings;
783 #endif /* CONFIG_KVM_BOOK3S_HV_EXIT_TIMING */
786 #define VCPU_FPR(vcpu, i) (vcpu)->arch.fp.fpr[i][TS_FPROFFSET]
787 #define VCPU_VSX_FPR(vcpu, i, j) ((vcpu)->arch.fp.fpr[i][j])
788 #define VCPU_VSX_VR(vcpu, i) ((vcpu)->arch.vr.vr[i])
790 /* Values for vcpu->arch.state */
791 #define KVMPPC_VCPU_NOTREADY 0
792 #define KVMPPC_VCPU_RUNNABLE 1
793 #define KVMPPC_VCPU_BUSY_IN_HOST 2
795 /* Values for vcpu->arch.io_gpr */
796 #define KVM_MMIO_REG_MASK 0x001f
797 #define KVM_MMIO_REG_EXT_MASK 0xffe0
798 #define KVM_MMIO_REG_GPR 0x0000
799 #define KVM_MMIO_REG_FPR 0x0020
800 #define KVM_MMIO_REG_QPR 0x0040
801 #define KVM_MMIO_REG_FQPR 0x0060
802 #define KVM_MMIO_REG_VSX 0x0080
804 #define __KVM_HAVE_ARCH_WQP
805 #define __KVM_HAVE_CREATE_DEVICE
807 static inline void kvm_arch_hardware_disable(void) {}
808 static inline void kvm_arch_hardware_unsetup(void) {}
809 static inline void kvm_arch_sync_events(struct kvm *kvm) {}
810 static inline void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots) {}
811 static inline void kvm_arch_flush_shadow_all(struct kvm *kvm) {}
812 static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
813 static inline void kvm_arch_exit(void) {}
814 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
815 static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
816 static inline void kvm_arch_vcpu_block_finish(struct kvm_vcpu *vcpu) {}
818 #endif /* __POWERPC_KVM_HOST_H__ */