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