2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2003 Peter Wemm.
5 * Copyright (c) 1991 Regents of the University of California.
8 * This code is derived from software contributed to Berkeley by
9 * the Systems Programming Group of the University of Utah Computer
10 * Science Department and William Jolitz of UUNET Technologies Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * Derived from hp300 version by Mike Hibler, this version by William
37 * Jolitz uses a recursive map [a pde points to the page directory] to
38 * map the page tables using the pagetables themselves. This is done to
39 * reduce the impact on kernel virtual memory for lots of sparse address
40 * space, and to reduce the cost of memory to each process.
42 * from: hp300: @(#)pmap.h 7.2 (Berkeley) 12/16/90
43 * from: @(#)pmap.h 7.4 (Berkeley) 5/12/91
47 #ifndef _MACHINE_PMAP_H_
48 #define _MACHINE_PMAP_H_
51 * Page-directory and page-table entries follow this format, with a few
52 * of the fields not present here and there, depending on a lot of things.
54 /* ---- Intel Nomenclature ---- */
55 #define X86_PG_V 0x001 /* P Valid */
56 #define X86_PG_RW 0x002 /* R/W Read/Write */
57 #define X86_PG_U 0x004 /* U/S User/Supervisor */
58 #define X86_PG_NC_PWT 0x008 /* PWT Write through */
59 #define X86_PG_NC_PCD 0x010 /* PCD Cache disable */
60 #define X86_PG_A 0x020 /* A Accessed */
61 #define X86_PG_M 0x040 /* D Dirty */
62 #define X86_PG_PS 0x080 /* PS Page size (0=4k,1=2M) */
63 #define X86_PG_PTE_PAT 0x080 /* PAT PAT index */
64 #define X86_PG_G 0x100 /* G Global */
65 #define X86_PG_AVAIL1 0x200 /* / Available for system */
66 #define X86_PG_AVAIL2 0x400 /* < programmers use */
67 #define X86_PG_AVAIL3 0x800 /* \ */
68 #define X86_PG_PDE_PAT 0x1000 /* PAT PAT index */
69 #define X86_PG_PKU(idx) ((pt_entry_t)idx << 59)
70 #define X86_PG_NX (1ul<<63) /* No-execute */
71 #define X86_PG_AVAIL(x) (1ul << (x))
73 /* Page level cache control fields used to determine the PAT type */
74 #define X86_PG_PDE_CACHE (X86_PG_PDE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD)
75 #define X86_PG_PTE_CACHE (X86_PG_PTE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD)
77 /* Protection keys indexes */
78 #define PMAP_MAX_PKRU_IDX 0xf
79 #define X86_PG_PKU_MASK X86_PG_PKU(PMAP_MAX_PKRU_IDX)
82 * Intel extended page table (EPT) bit definitions.
84 #define EPT_PG_READ 0x001 /* R Read */
85 #define EPT_PG_WRITE 0x002 /* W Write */
86 #define EPT_PG_EXECUTE 0x004 /* X Execute */
87 #define EPT_PG_IGNORE_PAT 0x040 /* IPAT Ignore PAT */
88 #define EPT_PG_PS 0x080 /* PS Page size */
89 #define EPT_PG_A 0x100 /* A Accessed */
90 #define EPT_PG_M 0x200 /* D Dirty */
91 #define EPT_PG_MEMORY_TYPE(x) ((x) << 3) /* MT Memory Type */
94 * Define the PG_xx macros in terms of the bits on x86 PTEs.
97 #define PG_RW X86_PG_RW
99 #define PG_NC_PWT X86_PG_NC_PWT
100 #define PG_NC_PCD X86_PG_NC_PCD
101 #define PG_A X86_PG_A
102 #define PG_M X86_PG_M
103 #define PG_PS X86_PG_PS
104 #define PG_PTE_PAT X86_PG_PTE_PAT
105 #define PG_G X86_PG_G
106 #define PG_AVAIL1 X86_PG_AVAIL1
107 #define PG_AVAIL2 X86_PG_AVAIL2
108 #define PG_AVAIL3 X86_PG_AVAIL3
109 #define PG_PDE_PAT X86_PG_PDE_PAT
110 #define PG_NX X86_PG_NX
111 #define PG_PDE_CACHE X86_PG_PDE_CACHE
112 #define PG_PTE_CACHE X86_PG_PTE_CACHE
114 /* Our various interpretations of the above */
115 #define PG_W X86_PG_AVAIL3 /* "Wired" pseudoflag */
116 #define PG_MANAGED X86_PG_AVAIL2
117 #define EPT_PG_EMUL_V X86_PG_AVAIL(52)
118 #define EPT_PG_EMUL_RW X86_PG_AVAIL(53)
119 #define PG_PROMOTED X86_PG_AVAIL(54) /* PDE only */
120 #define PG_FRAME (0x000ffffffffff000ul)
121 #define PG_PS_FRAME (0x000fffffffe00000ul)
122 #define PG_PS_PDP_FRAME (0x000fffffc0000000ul)
125 * Promotion to a 2MB (PDE) page mapping requires that the corresponding 4KB
126 * (PTE) page mappings have identical settings for the following fields:
128 #define PG_PTE_PROMOTE (PG_NX | PG_MANAGED | PG_W | PG_G | PG_PTE_CACHE | \
129 PG_M | PG_A | PG_U | PG_RW | PG_V | PG_PKU_MASK)
132 * Page Protection Exception bits
135 #define PGEX_P 0x01 /* Protection violation vs. not present */
136 #define PGEX_W 0x02 /* during a Write cycle */
137 #define PGEX_U 0x04 /* access from User mode (UPL) */
138 #define PGEX_RSV 0x08 /* reserved PTE field is non-zero */
139 #define PGEX_I 0x10 /* during an instruction fetch */
140 #define PGEX_PK 0x20 /* protection key violation */
141 #define PGEX_SGX 0x8000 /* SGX-related */
144 * undef the PG_xx macros that define bits in the regular x86 PTEs that
145 * have a different position in nested PTEs. This is done when compiling
146 * code that needs to be aware of the differences between regular x86 and
149 * The appropriate bitmask will be calculated at runtime based on the pmap
152 #ifdef AMD64_NPT_AWARE
153 #undef PG_AVAIL1 /* X86_PG_AVAIL1 aliases with EPT_PG_M */
166 * Pte related macros. This is complicated by having to deal with
167 * the sign extension of the 48th bit.
169 #define KVADDR(l4, l3, l2, l1) ( \
170 ((unsigned long)-1 << 47) | \
171 ((unsigned long)(l4) << PML4SHIFT) | \
172 ((unsigned long)(l3) << PDPSHIFT) | \
173 ((unsigned long)(l2) << PDRSHIFT) | \
174 ((unsigned long)(l1) << PAGE_SHIFT))
176 #define UVADDR(l4, l3, l2, l1) ( \
177 ((unsigned long)(l4) << PML4SHIFT) | \
178 ((unsigned long)(l3) << PDPSHIFT) | \
179 ((unsigned long)(l2) << PDRSHIFT) | \
180 ((unsigned long)(l1) << PAGE_SHIFT))
183 * Number of kernel PML4 slots. Can be anywhere from 1 to 64 or so,
184 * but setting it larger than NDMPML4E makes no sense.
186 * Each slot provides .5 TB of kernel virtual space.
190 #define NUPML4E (NPML4EPG/2) /* number of userland PML4 pages */
191 #define NUPDPE (NUPML4E*NPDPEPG)/* number of userland PDP pages */
192 #define NUPDE (NUPDPE*NPDEPG) /* number of userland PD entries */
195 * NDMPML4E is the maximum number of PML4 entries that will be
196 * used to implement the direct map. It must be a power of two,
197 * and should generally exceed NKPML4E. The maximum possible
198 * value is 64; using 128 will make the direct map intrude into
199 * the recursive page table map.
204 * These values control the layout of virtual memory. The starting address
205 * of the direct map, which is controlled by DMPML4I, must be a multiple of
206 * its size. (See the PHYS_TO_DMAP() and DMAP_TO_PHYS() macros.)
208 * Note: KPML4I is the index of the (single) level 4 page that maps
209 * the KVA that holds KERNBASE, while KPML4BASE is the index of the
210 * first level 4 page that maps VM_MIN_KERNEL_ADDRESS. If NKPML4E
211 * is 1, these are the same, otherwise KPML4BASE < KPML4I and extra
212 * level 4 PDEs are needed to map from VM_MIN_KERNEL_ADDRESS up to
215 * (KPML4I combines with KPDPI to choose where KERNBASE starts.
216 * Or, in other words, KPML4I provides bits 39..47 of KERNBASE,
217 * and KPDPI provides bits 30..38.)
219 #define PML4PML4I (NPML4EPG/2) /* Index of recursive pml4 mapping */
221 #define KPML4BASE (NPML4EPG-NKPML4E) /* KVM at highest addresses */
222 #define DMPML4I rounddown(KPML4BASE-NDMPML4E, NDMPML4E) /* Below KVM */
224 #define KPML4I (NPML4EPG-1)
225 #define KPDPI (NPDPEPG-2) /* kernbase at -2GB */
227 /* Large map: index of the first and max last pml4 entry */
228 #define LMSPML4I (PML4PML4I + 1)
229 #define LMEPML4I (DMPML4I - 1)
232 * XXX doesn't really belong here I guess...
234 #define ISA_HOLE_START 0xa0000
235 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
237 #define PMAP_PCID_NONE 0xffffffff
238 #define PMAP_PCID_KERN 0
239 #define PMAP_PCID_OVERMAX 0x1000
240 #define PMAP_PCID_OVERMAX_KERN 0x800
241 #define PMAP_PCID_USER_PT 0x800
243 #define PMAP_NO_CR3 (~0UL)
247 #include <sys/queue.h>
248 #include <sys/_cpuset.h>
249 #include <sys/_lock.h>
250 #include <sys/_mutex.h>
251 #include <sys/_pctrie.h>
252 #include <sys/_rangeset.h>
254 #include <vm/_vm_radix.h>
256 typedef u_int64_t pd_entry_t;
257 typedef u_int64_t pt_entry_t;
258 typedef u_int64_t pdp_entry_t;
259 typedef u_int64_t pml4_entry_t;
262 * Address of current address space page table maps and directories.
265 #define addr_PTmap (KVADDR(PML4PML4I, 0, 0, 0))
266 #define addr_PDmap (KVADDR(PML4PML4I, PML4PML4I, 0, 0))
267 #define addr_PDPmap (KVADDR(PML4PML4I, PML4PML4I, PML4PML4I, 0))
268 #define addr_PML4map (KVADDR(PML4PML4I, PML4PML4I, PML4PML4I, PML4PML4I))
269 #define addr_PML4pml4e (addr_PML4map + (PML4PML4I * sizeof(pml4_entry_t)))
270 #define PTmap ((pt_entry_t *)(addr_PTmap))
271 #define PDmap ((pd_entry_t *)(addr_PDmap))
272 #define PDPmap ((pd_entry_t *)(addr_PDPmap))
273 #define PML4map ((pd_entry_t *)(addr_PML4map))
274 #define PML4pml4e ((pd_entry_t *)(addr_PML4pml4e))
276 extern int nkpt; /* Initial number of kernel page tables */
277 extern u_int64_t KPDPphys; /* physical address of kernel level 3 */
278 extern u_int64_t KPML4phys; /* physical address of kernel level 4 */
281 * virtual address to page table entry and
282 * to physical address.
283 * Note: these work recursively, thus vtopte of a pte will give
284 * the corresponding pde that in turn maps it.
286 pt_entry_t *vtopte(vm_offset_t);
287 #define vtophys(va) pmap_kextract(((vm_offset_t) (va)))
289 #define pte_load_store(ptep, pte) atomic_swap_long(ptep, pte)
290 #define pte_load_clear(ptep) atomic_swap_long(ptep, 0)
291 #define pte_store(ptep, pte) do { \
292 *(u_long *)(ptep) = (u_long)(pte); \
294 #define pte_clear(ptep) pte_store(ptep, 0)
296 #define pde_store(pdep, pde) pte_store(pdep, pde)
298 extern pt_entry_t pg_nx;
313 TAILQ_HEAD(, pv_entry) pv_list; /* (p) */
314 int pv_gen; /* (p) */
319 PT_X86, /* regular x86 page tables */
320 PT_EPT, /* Intel's nested page tables */
321 PT_RVI, /* AMD's nested page tables */
330 * The kernel virtual address (KVA) of the level 4 page table page is always
331 * within the direct map (DMAP) region.
335 pml4_entry_t *pm_pml4; /* KVA of level 4 page table */
336 pml4_entry_t *pm_pml4u; /* KVA of user l4 page table */
339 TAILQ_HEAD(,pv_chunk) pm_pvchunk; /* list of mappings in pmap */
340 cpuset_t pm_active; /* active on cpus */
341 enum pmap_type pm_type; /* regular or nested tables */
342 struct pmap_statistics pm_stats; /* pmap statistics */
343 struct vm_radix pm_root; /* spare page table pages */
344 long pm_eptgen; /* EPT pmap generation id */
346 struct pmap_pcids pm_pcids[MAXCPU];
347 struct rangeset pm_pkru;
351 #define PMAP_NESTED_IPIMASK 0xff
352 #define PMAP_PDE_SUPERPAGE (1 << 8) /* supports 2MB superpages */
353 #define PMAP_EMULATE_AD_BITS (1 << 9) /* needs A/D bits emulation */
354 #define PMAP_SUPPORTS_EXEC_ONLY (1 << 10) /* execute only mappings ok */
356 typedef struct pmap *pmap_t;
359 extern struct pmap kernel_pmap_store;
360 #define kernel_pmap (&kernel_pmap_store)
362 #define PMAP_LOCK(pmap) mtx_lock(&(pmap)->pm_mtx)
363 #define PMAP_LOCK_ASSERT(pmap, type) \
364 mtx_assert(&(pmap)->pm_mtx, (type))
365 #define PMAP_LOCK_DESTROY(pmap) mtx_destroy(&(pmap)->pm_mtx)
366 #define PMAP_LOCK_INIT(pmap) mtx_init(&(pmap)->pm_mtx, "pmap", \
367 NULL, MTX_DEF | MTX_DUPOK)
368 #define PMAP_LOCKED(pmap) mtx_owned(&(pmap)->pm_mtx)
369 #define PMAP_MTX(pmap) (&(pmap)->pm_mtx)
370 #define PMAP_TRYLOCK(pmap) mtx_trylock(&(pmap)->pm_mtx)
371 #define PMAP_UNLOCK(pmap) mtx_unlock(&(pmap)->pm_mtx)
373 int pmap_pinit_type(pmap_t pmap, enum pmap_type pm_type, int flags);
374 int pmap_emulate_accessed_dirty(pmap_t pmap, vm_offset_t va, int ftype);
378 * For each vm_page_t, there is a list of all currently valid virtual
379 * mappings of that page. An entry is a pv_entry_t, the list is pv_list.
381 typedef struct pv_entry {
382 vm_offset_t pv_va; /* virtual address for mapping */
383 TAILQ_ENTRY(pv_entry) pv_next;
387 * pv_entries are allocated in chunks per-process. This avoids the
388 * need to track per-pmap assignments.
392 #define PV_CHUNK_HEADER \
394 TAILQ_ENTRY(pv_chunk) pc_list; \
395 uint64_t pc_map[_NPCM]; /* bitmap; 1 = free */ \
396 TAILQ_ENTRY(pv_chunk) pc_lru;
398 struct pv_chunk_header {
404 struct pv_entry pc_pventry[_NPCPV];
409 extern caddr_t CADDR1;
410 extern pt_entry_t *CMAP1;
411 extern vm_offset_t virtual_avail;
412 extern vm_offset_t virtual_end;
413 extern vm_paddr_t dmaplimit;
414 extern int pmap_pcid_enabled;
415 extern int invpcid_works;
417 #define pmap_page_get_memattr(m) ((vm_memattr_t)(m)->md.pat_mode)
418 #define pmap_page_is_write_mapped(m) (((m)->a.flags & PGA_WRITEABLE) != 0)
419 #define pmap_unmapbios(va, sz) pmap_unmapdev((va), (sz))
423 void pmap_activate_boot(pmap_t pmap);
424 void pmap_activate_sw(struct thread *);
425 void pmap_allow_2m_x_ept_recalculate(void);
426 void pmap_bootstrap(vm_paddr_t *);
427 int pmap_cache_bits(pmap_t pmap, int mode, boolean_t is_pde);
428 int pmap_change_attr(vm_offset_t, vm_size_t, int);
429 int pmap_change_prot(vm_offset_t, vm_size_t, vm_prot_t);
430 void pmap_demote_DMAP(vm_paddr_t base, vm_size_t len, boolean_t invalidate);
431 void pmap_flush_cache_range(vm_offset_t, vm_offset_t);
432 void pmap_flush_cache_phys_range(vm_paddr_t, vm_paddr_t, vm_memattr_t);
433 void pmap_init_pat(void);
434 void pmap_kenter(vm_offset_t va, vm_paddr_t pa);
435 void *pmap_kenter_temporary(vm_paddr_t pa, int i);
436 vm_paddr_t pmap_kextract(vm_offset_t);
437 void pmap_kremove(vm_offset_t);
438 int pmap_large_map(vm_paddr_t, vm_size_t, void **, vm_memattr_t);
439 void pmap_large_map_wb(void *sva, vm_size_t len);
440 void pmap_large_unmap(void *sva, vm_size_t len);
441 void *pmap_mapbios(vm_paddr_t, vm_size_t);
442 void *pmap_mapdev(vm_paddr_t, vm_size_t);
443 void *pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
444 void *pmap_mapdev_pciecfg(vm_paddr_t pa, vm_size_t size);
445 bool pmap_not_in_di(void);
446 boolean_t pmap_page_is_mapped(vm_page_t m);
447 void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
448 void pmap_pinit_pml4(vm_page_t);
449 bool pmap_ps_enabled(pmap_t pmap);
450 void pmap_unmapdev(vm_offset_t, vm_size_t);
451 void pmap_invalidate_page(pmap_t, vm_offset_t);
452 void pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
453 void pmap_invalidate_all(pmap_t);
454 void pmap_invalidate_cache(void);
455 void pmap_invalidate_cache_pages(vm_page_t *pages, int count);
456 void pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
457 void pmap_force_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
458 void pmap_get_mapping(pmap_t pmap, vm_offset_t va, uint64_t *ptr, int *num);
459 boolean_t pmap_map_io_transient(vm_page_t *, vm_offset_t *, int, boolean_t);
460 void pmap_unmap_io_transient(vm_page_t *, vm_offset_t *, int, boolean_t);
461 void pmap_pti_add_kva(vm_offset_t sva, vm_offset_t eva, bool exec);
462 void pmap_pti_remove_kva(vm_offset_t sva, vm_offset_t eva);
463 void pmap_pti_pcid_invalidate(uint64_t ucr3, uint64_t kcr3);
464 void pmap_pti_pcid_invlpg(uint64_t ucr3, uint64_t kcr3, vm_offset_t va);
465 void pmap_pti_pcid_invlrng(uint64_t ucr3, uint64_t kcr3, vm_offset_t sva,
467 int pmap_pkru_clear(pmap_t pmap, vm_offset_t sva, vm_offset_t eva);
468 int pmap_pkru_set(pmap_t pmap, vm_offset_t sva, vm_offset_t eva,
469 u_int keyidx, int flags);
470 void pmap_thread_init_invl_gen(struct thread *td);
471 int pmap_vmspace_copy(pmap_t dst_pmap, pmap_t src_pmap);
472 void pmap_page_array_startup(long count);
475 /* Return various clipped indexes for a given VA */
476 static __inline vm_pindex_t
477 pmap_pte_index(vm_offset_t va)
480 return ((va >> PAGE_SHIFT) & ((1ul << NPTEPGSHIFT) - 1));
483 static __inline vm_pindex_t
484 pmap_pde_index(vm_offset_t va)
487 return ((va >> PDRSHIFT) & ((1ul << NPDEPGSHIFT) - 1));
490 static __inline vm_pindex_t
491 pmap_pdpe_index(vm_offset_t va)
494 return ((va >> PDPSHIFT) & ((1ul << NPDPEPGSHIFT) - 1));
497 static __inline vm_pindex_t
498 pmap_pml4e_index(vm_offset_t va)
501 return ((va >> PML4SHIFT) & ((1ul << NPML4EPGSHIFT) - 1));
506 #endif /* !_MACHINE_PMAP_H_ */