1/*-
  2 * SPDX-License-Identifier: BSD-3-Clause
  3 *
  4 * Copyright (c) 2003 Peter Wemm.
  5 * Copyright (c) 1991 Regents of the University of California.
  6 * All rights reserved.
  7 *
  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.
 11 *
 12 * Redistribution and use in source and binary forms, with or without
 13 * modification, are permitted provided that the following conditions
 14 * are met:
 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.
 23 *
 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
 34 * SUCH DAMAGE.
 35 *
 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.
 41 *
 42 *	from: hp300: @(#)pmap.h	7.2 (Berkeley) 12/16/90
 43 *	from: @(#)pmap.h	7.4 (Berkeley) 5/12/91
 44 */
 45
 46#ifdef __i386__
 47#include <i386/pmap.h>
 48#else /* !__i386__ */
 49
 50#ifndef _MACHINE_PMAP_H_
 51#define	_MACHINE_PMAP_H_
 52
 53/*
 54 * Page-directory and page-table entries follow this format, with a few
 55 * of the fields not present here and there, depending on a lot of things.
 56 */
 57				/* ---- Intel Nomenclature ---- */
 58#define	X86_PG_V	0x001	/* P	Valid			*/
 59#define	X86_PG_RW	0x002	/* R/W	Read/Write		*/
 60#define	X86_PG_U	0x004	/* U/S  User/Supervisor		*/
 61#define	X86_PG_NC_PWT	0x008	/* PWT	Write through		*/
 62#define	X86_PG_NC_PCD	0x010	/* PCD	Cache disable		*/
 63#define	X86_PG_A	0x020	/* A	Accessed		*/
 64#define	X86_PG_M	0x040	/* D	Dirty			*/
 65#define	X86_PG_PS	0x080	/* PS	Page size (0=4k,1=2M)	*/
 66#define	X86_PG_PTE_PAT	0x080	/* PAT	PAT index		*/
 67#define	X86_PG_G	0x100	/* G	Global			*/
 68#define	X86_PG_AVAIL1	0x200	/*    /	Available for system	*/
 69#define	X86_PG_AVAIL2	0x400	/*   <	programmers use		*/
 70#define	X86_PG_AVAIL3	0x800	/*    \				*/
 71#define	X86_PG_PDE_PAT	0x1000	/* PAT	PAT index		*/
 72#define	X86_PG_PKU(idx)	((pt_entry_t)idx << 59)
 73#define	X86_PG_NX	(1ul<<63) /* No-execute */
 74#define	X86_PG_AVAIL(x)	(1ul << (x))
 75
 76/* Page level cache control fields used to determine the PAT type */
 77#define	X86_PG_PDE_CACHE (X86_PG_PDE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD)
 78#define	X86_PG_PTE_CACHE (X86_PG_PTE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD)
 79
 80/* Protection keys indexes */
 81#define	PMAP_MAX_PKRU_IDX	0xf
 82#define	X86_PG_PKU_MASK		X86_PG_PKU(PMAP_MAX_PKRU_IDX)
 83
 84/*
 85 * Intel extended page table (EPT) bit definitions.
 86 */
 87#define	EPT_PG_READ		0x001	/* R	Read		*/
 88#define	EPT_PG_WRITE		0x002	/* W	Write		*/
 89#define	EPT_PG_EXECUTE		0x004	/* X	Execute		*/
 90#define	EPT_PG_IGNORE_PAT	0x040	/* IPAT	Ignore PAT	*/
 91#define	EPT_PG_PS		0x080	/* PS	Page size	*/
 92#define	EPT_PG_A		0x100	/* A	Accessed	*/
 93#define	EPT_PG_M		0x200	/* D	Dirty		*/
 94#define	EPT_PG_MEMORY_TYPE(x)	((x) << 3) /* MT Memory Type	*/
 95
 96/*
 97 * Define the PG_xx macros in terms of the bits on x86 PTEs.
 98 */
 99#define	PG_V		X86_PG_V
100#define	PG_RW		X86_PG_RW
101#define	PG_U		X86_PG_U
102#define	PG_NC_PWT	X86_PG_NC_PWT
103#define	PG_NC_PCD	X86_PG_NC_PCD
104#define	PG_A		X86_PG_A
105#define	PG_M		X86_PG_M
106#define	PG_PS		X86_PG_PS
107#define	PG_PTE_PAT	X86_PG_PTE_PAT
108#define	PG_G		X86_PG_G
109#define	PG_AVAIL1	X86_PG_AVAIL1
110#define	PG_AVAIL2	X86_PG_AVAIL2
111#define	PG_AVAIL3	X86_PG_AVAIL3
112#define	PG_PDE_PAT	X86_PG_PDE_PAT
113#define	PG_NX		X86_PG_NX
114#define	PG_PDE_CACHE	X86_PG_PDE_CACHE
115#define	PG_PTE_CACHE	X86_PG_PTE_CACHE
116
117/* Our various interpretations of the above */
118#define	PG_W		X86_PG_AVAIL3	/* "Wired" pseudoflag */
119#define	PG_MANAGED	X86_PG_AVAIL2
120#define	EPT_PG_EMUL_V	X86_PG_AVAIL(52)
121#define	EPT_PG_EMUL_RW	X86_PG_AVAIL(53)
122#define	PG_PROMOTED	X86_PG_AVAIL(54)	/* PDE only */
123#define	PG_FRAME	(0x000ffffffffff000ul)
124#define	PG_PS_FRAME	(0x000fffffffe00000ul)
125#define	PG_PS_PDP_FRAME	(0x000fffffc0000000ul)
126
127/*
128 * Promotion to a 2MB (PDE) page mapping requires that the corresponding 4KB
129 * (PTE) page mappings have identical settings for the following fields:
130 */
131#define	PG_PTE_PROMOTE	(PG_NX | PG_MANAGED | PG_W | PG_G | PG_PTE_CACHE | \
132	    PG_M | PG_U | PG_RW | PG_V | PG_PKU_MASK)
133
134/*
135 * Page Protection Exception bits
136 */
137
138#define PGEX_P		0x01	/* Protection violation vs. not present */
139#define PGEX_W		0x02	/* during a Write cycle */
140#define PGEX_U		0x04	/* access from User mode (UPL) */
141#define PGEX_RSV	0x08	/* reserved PTE field is non-zero */
142#define PGEX_I		0x10	/* during an instruction fetch */
143#define	PGEX_PK		0x20	/* protection key violation */
144#define	PGEX_SGX	0x8000	/* SGX-related */
145
146/* 
147 * undef the PG_xx macros that define bits in the regular x86 PTEs that
148 * have a different position in nested PTEs. This is done when compiling
149 * code that needs to be aware of the differences between regular x86 and
150 * nested PTEs.
151 *
152 * The appropriate bitmask will be calculated at runtime based on the pmap
153 * type.
154 */
155#ifdef AMD64_NPT_AWARE
156#undef PG_AVAIL1		/* X86_PG_AVAIL1 aliases with EPT_PG_M */
157#undef PG_G
158#undef PG_A
159#undef PG_M
160#undef PG_PDE_PAT
161#undef PG_PDE_CACHE
162#undef PG_PTE_PAT
163#undef PG_PTE_CACHE
164#undef PG_RW
165#undef PG_V
166#endif
167
168/*
169 * Pte related macros.  This is complicated by having to deal with
170 * the sign extension of the 48th bit.
171 */
172#define KV4ADDR(l4, l3, l2, l1) ( \
173	((unsigned long)-1 << 47) | \
174	((unsigned long)(l4) << PML4SHIFT) | \
175	((unsigned long)(l3) << PDPSHIFT) | \
176	((unsigned long)(l2) << PDRSHIFT) | \
177	((unsigned long)(l1) << PAGE_SHIFT))
178#define KV5ADDR(l5, l4, l3, l2, l1) (		\
179	((unsigned long)-1 << 56) | \
180	((unsigned long)(l5) << PML5SHIFT) | \
181	((unsigned long)(l4) << PML4SHIFT) | \
182	((unsigned long)(l3) << PDPSHIFT) | \
183	((unsigned long)(l2) << PDRSHIFT) | \
184	((unsigned long)(l1) << PAGE_SHIFT))
185
186#define UVADDR(l5, l4, l3, l2, l1) (	     \
187	((unsigned long)(l5) << PML5SHIFT) | \
188	((unsigned long)(l4) << PML4SHIFT) | \
189	((unsigned long)(l3) << PDPSHIFT) | \
190	((unsigned long)(l2) << PDRSHIFT) | \
191	((unsigned long)(l1) << PAGE_SHIFT))
192
193/*
194 * Number of kernel PML4 slots.  Can be anywhere from 1 to 64 or so,
195 * but setting it larger than NDMPML4E makes no sense.
196 *
197 * Each slot provides .5 TB of kernel virtual space.
198 */
199#define NKPML4E		4
200
201/*
202 * Number of PML4 slots for the KASAN shadow map.  It requires 1 byte of memory
203 * for every 8 bytes of the kernel address space.
204 */
205#define	NKASANPML4E	((NKPML4E + 7) / 8)
206
207/*
208 * Number of PML4 slots for the KMSAN shadow and origin maps.  These are
209 * one-to-one with the kernel map.
210 */
211#define	NKMSANSHADPML4E	NKPML4E
212#define	NKMSANORIGPML4E	NKPML4E
213
214/*
215 * We use the same numbering of the page table pages for 5-level and
216 * 4-level paging structures.
217 */
218#define	NUPML5E		(NPML5EPG / 2)		/* number of userland PML5
219						   pages */
220#define	NUPML4E		(NUPML5E * NPML4EPG)	/* number of userland PML4
221						   pages */
222#define	NUPDPE		(NUPML4E * NPDPEPG)	/* number of userland PDP
223						   pages */
224#define	NUPDE		(NUPDPE * NPDEPG)	/* number of userland PD
225						   entries */
226#define	NUP4ML4E	(NPML4EPG / 2)
227
228/*
229 * NDMPML4E is the maximum number of PML4 entries that will be
230 * used to implement the direct map.  It must be a power of two,
231 * and should generally exceed NKPML4E.  The maximum possible
232 * value is 64; using 128 will make the direct map intrude into
233 * the recursive page table map.
234 */
235#define	NDMPML4E	8
236
237/*
238 * These values control the layout of virtual memory.  The starting address
239 * of the direct map, which is controlled by DMPML4I, must be a multiple of
240 * its size.  (See the PHYS_TO_DMAP() and DMAP_TO_PHYS() macros.)
241 *
242 * Note: KPML4I is the index of the (single) level 4 page that maps
243 * the KVA that holds KERNBASE, while KPML4BASE is the index of the
244 * first level 4 page that maps VM_MIN_KERNEL_ADDRESS.  If NKPML4E
245 * is 1, these are the same, otherwise KPML4BASE < KPML4I and extra
246 * level 4 PDEs are needed to map from VM_MIN_KERNEL_ADDRESS up to
247 * KERNBASE.
248 *
249 * (KPML4I combines with KPDPI to choose where KERNBASE starts.
250 * Or, in other words, KPML4I provides bits 39..47 of KERNBASE,
251 * and KPDPI provides bits 30..38.)
252 */
253#define	PML4PML4I	(NPML4EPG / 2)	/* Index of recursive pml4 mapping */
254#define	PML5PML5I	(NPML5EPG / 2)	/* Index of recursive pml5 mapping */
255
256#define	KPML4BASE	(NPML4EPG-NKPML4E) /* KVM at highest addresses */
257#define	DMPML4I		rounddown(KPML4BASE-NDMPML4E, NDMPML4E) /* Below KVM */
258
259#define	KPML4I		(NPML4EPG-1)
260#define	KPDPI		(NPDPEPG-2)	/* kernbase at -2GB */
261
262#define	KASANPML4I	(DMPML4I - NKASANPML4E) /* Below the direct map */
263
264#define	KMSANSHADPML4I	(KPML4BASE - NKMSANSHADPML4E)
265#define	KMSANORIGPML4I	(DMPML4I - NKMSANORIGPML4E)
266
267/* Large map: index of the first and max last pml4 entry */
268#define	LMSPML4I	(PML4PML4I + 1)
269#define	LMEPML4I	(KASANPML4I - 1)
270
271/*
272 * XXX doesn't really belong here I guess...
273 */
274#define ISA_HOLE_START    0xa0000
275#define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
276
277#define	PMAP_PCID_NONE		0xffffffff
278#define	PMAP_PCID_KERN		0
279#define	PMAP_PCID_OVERMAX	0x1000
280#define	PMAP_PCID_OVERMAX_KERN	0x800
281#define	PMAP_PCID_USER_PT	0x800
282
283#define	PMAP_NO_CR3		0xffffffffffffffff
284#define	PMAP_UCR3_NOMASK	0xffffffffffffffff
285
286#ifndef LOCORE
287
288#include <sys/kassert.h>
289#include <sys/queue.h>
290#include <sys/_cpuset.h>
291#include <sys/_lock.h>
292#include <sys/_mutex.h>
293#include <sys/_pctrie.h>
294#include <machine/_pmap.h>
295#include <sys/_pv_entry.h>
296#include <sys/_rangeset.h>
297#include <sys/_smr.h>
298
299#include <vm/_vm_radix.h>
300
301typedef u_int64_t pd_entry_t;
302typedef u_int64_t pt_entry_t;
303typedef u_int64_t pdp_entry_t;
304typedef u_int64_t pml4_entry_t;
305typedef u_int64_t pml5_entry_t;
306
307/*
308 * Address of current address space page table maps and directories.
309 */
310#ifdef _KERNEL
311#define	addr_P4Tmap	(KV4ADDR(PML4PML4I, 0, 0, 0))
312#define	addr_P4Dmap	(KV4ADDR(PML4PML4I, PML4PML4I, 0, 0))
313#define	addr_P4DPmap	(KV4ADDR(PML4PML4I, PML4PML4I, PML4PML4I, 0))
314#define	addr_P4ML4map	(KV4ADDR(PML4PML4I, PML4PML4I, PML4PML4I, PML4PML4I))
315#define	addr_P4ML4pml4e	(addr_PML4map + (PML4PML4I * sizeof(pml4_entry_t)))
316#define	P4Tmap		((pt_entry_t *)(addr_P4Tmap))
317#define	P4Dmap		((pd_entry_t *)(addr_P4Dmap))
318
319#define	addr_P5Tmap	(KV5ADDR(PML5PML5I, 0, 0, 0, 0))
320#define	addr_P5Dmap	(KV5ADDR(PML5PML5I, PML5PML5I, 0, 0, 0))
321#define	addr_P5DPmap	(KV5ADDR(PML5PML5I, PML5PML5I, PML5PML5I, 0, 0))
322#define	addr_P5ML4map	(KV5ADDR(PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I, 0))
323#define	addr_P5ML5map	\
324    (KVADDR(PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I))
325#define	addr_P5ML5pml5e	(addr_P5ML5map + (PML5PML5I * sizeof(pml5_entry_t)))
326#define	P5Tmap		((pt_entry_t *)(addr_P5Tmap))
327#define	P5Dmap		((pd_entry_t *)(addr_P5Dmap))
328
329extern int nkpt;		/* Initial number of kernel page tables */
330extern u_int64_t KPML4phys;	/* physical address of kernel level 4 */
331extern u_int64_t KPML5phys;	/* physical address of kernel level 5 */
332
333/*
334 * virtual address to page table entry and
335 * to physical address.
336 * Note: these work recursively, thus vtopte of a pte will give
337 * the corresponding pde that in turn maps it.
338 */
339pt_entry_t *vtopte(vm_offset_t);
340#define	vtophys(va)	pmap_kextract(((vm_offset_t) (va)))
341
342#define	pte_load_store(ptep, pte)	atomic_swap_long(ptep, pte)
343#define	pte_load_clear(ptep)		atomic_swap_long(ptep, 0)
344#define	pte_store(ptep, pte) do { \
345	*(u_long *)(ptep) = (u_long)(pte); \
346} while (0)
347#define	pte_clear(ptep)			pte_store(ptep, 0)
348
349#define	pde_store(pdep, pde)		pte_store(pdep, pde)
350
351extern pt_entry_t pg_nx;
352
353#endif /* _KERNEL */
354
355/*
356 * Pmap stuff
357 */
358
359/*
360 * Locks
361 * (p) PV list lock
362 */
363struct md_page {
364	TAILQ_HEAD(, pv_entry)	pv_list;  /* (p) */
365	int			pv_gen;   /* (p) */
366	int			pat_mode;
367};
368
369enum pmap_type {
370	PT_X86,			/* regular x86 page tables */
371	PT_EPT,			/* Intel's nested page tables */
372	PT_RVI,			/* AMD's nested page tables */
373};
374
375/*
376 * The kernel virtual address (KVA) of the level 4 page table page is always
377 * within the direct map (DMAP) region.
378 */
379struct pmap {
380	struct mtx		pm_mtx;
381	pml4_entry_t		*pm_pmltop;	/* KVA of top level page table */
382	pml4_entry_t		*pm_pmltopu;	/* KVA of user top page table */
383	uint64_t		pm_cr3;
384	uint64_t		pm_ucr3;
385	TAILQ_HEAD(,pv_chunk)	pm_pvchunk;	/* list of mappings in pmap */
386	cpuset_t		pm_active;	/* active on cpus */
387	enum pmap_type		pm_type;	/* regular or nested tables */
388	struct pmap_statistics	pm_stats;	/* pmap statistics */
389	struct vm_radix		pm_root;	/* spare page table pages */
390	long			pm_eptgen;	/* EPT pmap generation id */
391	smr_t			pm_eptsmr;
392	int			pm_flags;
393	struct pmap_pcid	*pm_pcidp;
394	struct rangeset		pm_pkru;
395};
396
397/* flags */
398#define	PMAP_NESTED_IPIMASK	0xff
399#define	PMAP_PDE_SUPERPAGE	(1 << 8)	/* supports 2MB superpages */
400#define	PMAP_EMULATE_AD_BITS	(1 << 9)	/* needs A/D bits emulation */
401#define	PMAP_SUPPORTS_EXEC_ONLY	(1 << 10)	/* execute only mappings ok */
402
403typedef struct pmap	*pmap_t;
404
405#ifdef _KERNEL
406extern struct pmap	kernel_pmap_store;
407#define kernel_pmap	(&kernel_pmap_store)
408
409#define	PMAP_LOCK(pmap)		mtx_lock(&(pmap)->pm_mtx)
410#define	PMAP_LOCK_ASSERT(pmap, type) \
411				mtx_assert(&(pmap)->pm_mtx, (type))
412#define	PMAP_LOCK_DESTROY(pmap)	mtx_destroy(&(pmap)->pm_mtx)
413#define	PMAP_LOCK_INIT(pmap)	mtx_init(&(pmap)->pm_mtx, "pmap", \
414				    NULL, MTX_DEF | MTX_DUPOK)
415#define	PMAP_LOCKED(pmap)	mtx_owned(&(pmap)->pm_mtx)
416#define	PMAP_MTX(pmap)		(&(pmap)->pm_mtx)
417#define	PMAP_TRYLOCK(pmap)	mtx_trylock(&(pmap)->pm_mtx)
418#define	PMAP_UNLOCK(pmap)	mtx_unlock(&(pmap)->pm_mtx)
419
420int	pmap_pinit_type(pmap_t pmap, enum pmap_type pm_type, int flags);
421int	pmap_emulate_accessed_dirty(pmap_t pmap, vm_offset_t va, int ftype);
422
423extern caddr_t	CADDR1;
424extern pt_entry_t *CMAP1;
425extern vm_offset_t virtual_avail;
426extern vm_offset_t virtual_end;
427extern vm_paddr_t dmaplimit;
428extern int pmap_pcid_enabled;
429extern int invpcid_works;
430extern int invlpgb_works;
431extern int invlpgb_maxcnt;
432extern int pmap_pcid_invlpg_workaround;
433extern int pmap_pcid_invlpg_workaround_uena;
434
435#define	pmap_page_get_memattr(m)	((vm_memattr_t)(m)->md.pat_mode)
436#define	pmap_page_is_write_mapped(m)	(((m)->a.flags & PGA_WRITEABLE) != 0)
437#define	pmap_unmapbios(va, sz)		pmap_unmapdev((va), (sz))
438
439#define	pmap_vm_page_alloc_check(m)					\
440	KASSERT(m->phys_addr < kernphys ||				\
441	    m->phys_addr >= kernphys + (vm_offset_t)&_end - KERNSTART,	\
442	    ("allocating kernel page %p pa %#lx kernphys %#lx end %p", \
443	    m, m->phys_addr, kernphys, &_end));
444
445struct thread;
446
447void	pmap_activate_boot(pmap_t pmap);
448void	pmap_activate_sw(struct thread *);
449void	pmap_allow_2m_x_ept_recalculate(void);
450void	pmap_bootstrap(vm_paddr_t *);
451int	pmap_cache_bits(pmap_t pmap, int mode, boolean_t is_pde);
452int	pmap_change_attr(vm_offset_t, vm_size_t, int);
453int	pmap_change_prot(vm_offset_t, vm_size_t, vm_prot_t);
454void	pmap_demote_DMAP(vm_paddr_t base, vm_size_t len, boolean_t invalidate);
455void	pmap_flush_cache_range(vm_offset_t, vm_offset_t);
456void	pmap_flush_cache_phys_range(vm_paddr_t, vm_paddr_t, vm_memattr_t);
457void	pmap_init_pat(void);
458void	pmap_kenter(vm_offset_t va, vm_paddr_t pa);
459void	*pmap_kenter_temporary(vm_paddr_t pa, int i);
460vm_paddr_t pmap_kextract(vm_offset_t);
461void	pmap_kremove(vm_offset_t);
462int	pmap_large_map(vm_paddr_t, vm_size_t, void **, vm_memattr_t);
463void	pmap_large_map_wb(void *sva, vm_size_t len);
464void	pmap_large_unmap(void *sva, vm_size_t len);
465void	*pmap_mapbios(vm_paddr_t, vm_size_t);
466void	*pmap_mapdev(vm_paddr_t, vm_size_t);
467void	*pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
468void	*pmap_mapdev_pciecfg(vm_paddr_t pa, vm_size_t size);
469bool	pmap_not_in_di(void);
470boolean_t pmap_page_is_mapped(vm_page_t m);
471void	pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
472void	pmap_page_set_memattr_noflush(vm_page_t m, vm_memattr_t ma);
473void	pmap_pinit_pml4(vm_page_t);
474void	pmap_pinit_pml5(vm_page_t);
475bool	pmap_ps_enabled(pmap_t pmap);
476void	pmap_unmapdev(void *, vm_size_t);
477void	pmap_invalidate_page(pmap_t, vm_offset_t);
478void	pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
479void	pmap_invalidate_all(pmap_t);
480void	pmap_invalidate_cache(void);
481void	pmap_invalidate_cache_pages(vm_page_t *pages, int count);
482void	pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
483void	pmap_force_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
484void	pmap_get_mapping(pmap_t pmap, vm_offset_t va, uint64_t *ptr, int *num);
485bool	pmap_map_io_transient(vm_page_t *, vm_offset_t *, int, bool);
486void	pmap_unmap_io_transient(vm_page_t *, vm_offset_t *, int, bool);
487void	pmap_map_delete(pmap_t, vm_offset_t, vm_offset_t);
488void	pmap_pti_add_kva(vm_offset_t sva, vm_offset_t eva, bool exec);
489void	pmap_pti_remove_kva(vm_offset_t sva, vm_offset_t eva);
490void	pmap_pti_pcid_invalidate(uint64_t ucr3, uint64_t kcr3);
491void	pmap_pti_pcid_invlpg(uint64_t ucr3, uint64_t kcr3, vm_offset_t va);
492void	pmap_pti_pcid_invlrng(uint64_t ucr3, uint64_t kcr3, vm_offset_t sva,
493	    vm_offset_t eva);
494int	pmap_pkru_clear(pmap_t pmap, vm_offset_t sva, vm_offset_t eva);
495int	pmap_pkru_set(pmap_t pmap, vm_offset_t sva, vm_offset_t eva,
496	    u_int keyidx, int flags);
497void	pmap_thread_init_invl_gen(struct thread *td);
498int	pmap_vmspace_copy(pmap_t dst_pmap, pmap_t src_pmap);
499void	pmap_page_array_startup(long count);
500vm_page_t pmap_page_alloc_below_4g(bool zeroed);
501
502#if defined(KASAN) || defined(KMSAN)
503void	pmap_san_enter(vm_offset_t);
504#endif
505
506/*
507 * Returns a pointer to a set of CPUs on which the pmap is currently active.
508 * Note that the set can be modified without any mutual exclusion, so a copy
509 * must be made if a stable value is required.
510 */
511static __inline volatile cpuset_t *
512pmap_invalidate_cpu_mask(pmap_t pmap)
513{
514	return (&pmap->pm_active);
515}
516
517#if defined(_SYS_PCPU_H_) && defined(_MACHINE_CPUFUNC_H_)
518/*
519 * It seems that AlderLake+ small cores have some microarchitectural
520 * bug, which results in the INVLPG instruction failing to flush all
521 * global TLB entries when PCID is enabled.  Work around it for now,
522 * by doing global invalidation on small cores instead of INVLPG.
523 */
524static __inline void
525pmap_invlpg(pmap_t pmap, vm_offset_t va)
526{
527	if (pmap == kernel_pmap && PCPU_GET(pcid_invlpg_workaround)) {
528		struct invpcid_descr d = { 0 };
529
530		invpcid(&d, INVPCID_CTXGLOB);
531	} else {
532		invlpg(va);
533	}
534}
535#endif /* sys/pcpu.h && machine/cpufunc.h */
536
537#if defined(_SYS_PCPU_H_)
538/* Return pcid for the pmap pmap on current cpu */
539static __inline uint32_t
540pmap_get_pcid(pmap_t pmap)
541{
542	struct pmap_pcid *pcidp;
543
544	MPASS(pmap_pcid_enabled);
545	pcidp = zpcpu_get(pmap->pm_pcidp);
546	return (pcidp->pm_pcid);
547}
548#endif /* sys/pcpu.h */
549
550/*
551 * Invalidation request.  PCPU pc_smp_tlb_op uses u_int instead of the
552 * enum to avoid both namespace and ABI issues (with enums).
553 */
554enum invl_op_codes {
555	INVL_OP_TLB               = 1,
556	INVL_OP_TLB_INVPCID       = 2,
557	INVL_OP_TLB_INVPCID_PTI   = 3,
558	INVL_OP_TLB_PCID          = 4,
559	INVL_OP_PGRNG             = 5,
560	INVL_OP_PGRNG_INVPCID     = 6,
561	INVL_OP_PGRNG_PCID        = 7,
562	INVL_OP_PG                = 8,
563	INVL_OP_PG_INVPCID        = 9,
564	INVL_OP_PG_PCID           = 10,
565	INVL_OP_CACHE             = 11,
566};
567
568typedef void (*smp_invl_local_cb_t)(struct pmap *, vm_offset_t addr1,
569    vm_offset_t addr2);
570typedef void (*smp_targeted_tlb_shootdown_t)(pmap_t, vm_offset_t, vm_offset_t,
571    smp_invl_local_cb_t, enum invl_op_codes);
572
573void smp_targeted_tlb_shootdown_native(pmap_t, vm_offset_t, vm_offset_t,
574    smp_invl_local_cb_t, enum invl_op_codes);
575extern smp_targeted_tlb_shootdown_t smp_targeted_tlb_shootdown;
576
577#endif /* _KERNEL */
578
579/* Return various clipped indexes for a given VA */
580static __inline vm_pindex_t
581pmap_pte_index(vm_offset_t va)
582{
583
584	return ((va >> PAGE_SHIFT) & ((1ul << NPTEPGSHIFT) - 1));
585}
586
587static __inline vm_pindex_t
588pmap_pde_index(vm_offset_t va)
589{
590
591	return ((va >> PDRSHIFT) & ((1ul << NPDEPGSHIFT) - 1));
592}
593
594static __inline vm_pindex_t
595pmap_pdpe_index(vm_offset_t va)
596{
597
598	return ((va >> PDPSHIFT) & ((1ul << NPDPEPGSHIFT) - 1));
599}
600
601static __inline vm_pindex_t
602pmap_pml4e_index(vm_offset_t va)
603{
604
605	return ((va >> PML4SHIFT) & ((1ul << NPML4EPGSHIFT) - 1));
606}
607
608static __inline vm_pindex_t
609pmap_pml5e_index(vm_offset_t va)
610{
611
612	return ((va >> PML5SHIFT) & ((1ul << NPML5EPGSHIFT) - 1));
613}
614
615#endif /* !LOCORE */
616
617#endif /* !_MACHINE_PMAP_H_ */
618
619#endif /* __i386__ */