2 * Copyright (c) 1991 Regents of the University of California.
5 * This code is derived from software contributed to Berkeley by
6 * the Systems Programming Group of the University of Utah Computer
7 * Science Department and William Jolitz of UUNET Technologies Inc.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * Derived from hp300 version by Mike Hibler, this version by William
34 * Jolitz uses a recursive map [a pde points to the page directory] to
35 * map the page tables using the pagetables themselves. This is done to
36 * reduce the impact on kernel virtual memory for lots of sparse address
37 * space, and to reduce the cost of memory to each process.
39 * from: hp300: @(#)pmap.h 7.2 (Berkeley) 12/16/90
40 * from: @(#)pmap.h 7.4 (Berkeley) 5/12/91
44 #ifndef _MACHINE_PMAP_H_
45 #define _MACHINE_PMAP_H_
48 * Page-directory and page-table entries follow this format, with a few
49 * of the fields not present here and there, depending on a lot of things.
51 /* ---- Intel Nomenclature ---- */
52 #define PG_V 0x001 /* P Valid */
53 #define PG_RW 0x002 /* R/W Read/Write */
54 #define PG_U 0x004 /* U/S User/Supervisor */
55 #define PG_NC_PWT 0x008 /* PWT Write through */
56 #define PG_NC_PCD 0x010 /* PCD Cache disable */
57 #define PG_A 0x020 /* A Accessed */
58 #define PG_M 0x040 /* D Dirty */
59 #define PG_PS 0x080 /* PS Page size (0=4k,1=4M) */
60 #define PG_PTE_PAT 0x080 /* PAT PAT index */
61 #define PG_G 0x100 /* G Global */
62 #define PG_AVAIL1 0x200 /* / Available for system */
63 #define PG_AVAIL2 0x400 /* < programmers use */
64 #define PG_AVAIL3 0x800 /* \ */
65 #define PG_PDE_PAT 0x1000 /* PAT PAT index */
67 #define PG_NX (1ull<<63) /* No-execute */
71 /* Our various interpretations of the above */
72 #define PG_W PG_AVAIL1 /* "Wired" pseudoflag */
73 #define PG_MANAGED PG_AVAIL2
75 #define PG_FRAME (0x000ffffffffff000ull)
76 #define PG_PS_FRAME (0x000fffffffe00000ull)
78 #define PG_FRAME (~PAGE_MASK)
79 #define PG_PS_FRAME (0xffc00000)
81 #define PG_PROT (PG_RW|PG_U) /* all protection bits . */
82 #define PG_N (PG_NC_PWT|PG_NC_PCD) /* Non-cacheable */
84 /* Page level cache control fields used to determine the PAT type */
85 #define PG_PDE_CACHE (PG_PDE_PAT | PG_NC_PWT | PG_NC_PCD)
86 #define PG_PTE_CACHE (PG_PTE_PAT | PG_NC_PWT | PG_NC_PCD)
89 * Promotion to a 2 or 4MB (PDE) page mapping requires that the corresponding
90 * 4KB (PTE) page mappings have identical settings for the following fields:
92 #define PG_PTE_PROMOTE (PG_MANAGED | PG_W | PG_G | PG_PTE_PAT | \
93 PG_M | PG_A | PG_NC_PCD | PG_NC_PWT | PG_U | PG_RW | PG_V)
96 * Page Protection Exception bits
99 #define PGEX_P 0x01 /* Protection violation vs. not present */
100 #define PGEX_W 0x02 /* during a Write cycle */
101 #define PGEX_U 0x04 /* access from User mode (UPL) */
102 #define PGEX_RSV 0x08 /* reserved PTE field is non-zero */
103 #define PGEX_I 0x10 /* during an instruction fetch */
106 * Size of Kernel address space. This is the number of page table pages
107 * (4MB each) to use for the kernel. 256 pages == 1 Gigabyte.
108 * This **MUST** be a multiple of 4 (eg: 252, 256, 260, etc).
109 * For PAE, the page table page unit size is 2MB. This means that 512 pages
110 * is 1 Gigabyte. Double everything. It must be a multiple of 8 for PAE.
114 #define KVA_PAGES 512
116 #define KVA_PAGES 256
123 #define VADDR(pdi, pti) ((vm_offset_t)(((pdi)<<PDRSHIFT)|((pti)<<PAGE_SHIFT)))
125 /* Initial number of kernel page tables. */
128 /* 152 page tables needed to map 16G (76B "struct vm_page", 2M page tables). */
131 /* 18 page tables needed to map 4G (72B "struct vm_page", 4M page tables). */
137 #define NKPDE (KVA_PAGES) /* number of page tables/pde's */
141 * The *PTDI values control the layout of virtual memory
143 * XXX This works for now, but I am not real happy with it, I'll fix it
144 * right after I fix locore.s and the magic 28K hole
146 #define KPTDI (NPDEPTD-NKPDE) /* start of kernel virtual pde's */
147 #define PTDPTDI (KPTDI-NPGPTD) /* ptd entry that points to ptd! */
150 * XXX doesn't really belong here I guess...
152 #define ISA_HOLE_START 0xa0000
153 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
157 #include <sys/queue.h>
158 #include <sys/_cpuset.h>
159 #include <sys/_lock.h>
160 #include <sys/_mutex.h>
162 #include <vm/_vm_radix.h>
166 typedef uint64_t pdpt_entry_t;
167 typedef uint64_t pd_entry_t;
168 typedef uint64_t pt_entry_t;
175 typedef uint32_t pd_entry_t;
176 typedef uint32_t pt_entry_t;
184 * Address of current address space page table maps and directories.
187 extern pt_entry_t PTmap[];
188 extern pd_entry_t PTD[];
189 extern pd_entry_t PTDpde[];
192 extern pdpt_entry_t *IdlePDPT;
194 extern pd_entry_t *IdlePTD; /* physical address of "Idle" state directory */
197 * Translate a virtual address to the kernel virtual address of its page table
198 * entry (PTE). This can be used recursively. If the address of a PTE as
199 * previously returned by this macro is itself given as the argument, then the
200 * address of the page directory entry (PDE) that maps the PTE will be
203 * This macro may be used before pmap_bootstrap() is called.
205 #define vtopte(va) (PTmap + i386_btop(va))
208 * Translate a virtual address to its physical address.
210 * This macro may be used before pmap_bootstrap() is called.
212 #define vtophys(va) pmap_kextract((vm_offset_t)(va))
215 #include <sys/param.h>
217 #include <xen/xen-os.h>
219 #include <machine/xen/xenvar.h>
220 #include <machine/xen/xenpmap.h>
222 extern pt_entry_t pg_nx;
224 #define PG_KERNEL (PG_V | PG_A | PG_RW | PG_M)
226 #define MACH_TO_VM_PAGE(ma) PHYS_TO_VM_PAGE(xpmap_mtop((ma)))
227 #define VM_PAGE_TO_MACH(m) xpmap_ptom(VM_PAGE_TO_PHYS((m)))
229 #define VTOM(va) xpmap_ptom(VTOP(va))
231 static __inline vm_paddr_t
232 pmap_kextract_ma(vm_offset_t va)
235 if ((ma = PTD[va >> PDRSHIFT]) & PG_PS) {
236 ma = (ma & ~(NBPDR - 1)) | (va & (NBPDR - 1));
238 ma = (*vtopte(va) & PG_FRAME) | (va & PAGE_MASK);
243 static __inline vm_paddr_t
244 pmap_kextract(vm_offset_t va)
246 return xpmap_mtop(pmap_kextract_ma(va));
248 #define vtomach(va) pmap_kextract_ma(((vm_offset_t) (va)))
250 vm_paddr_t pmap_extract_ma(struct pmap *pmap, vm_offset_t va);
252 void pmap_kenter_ma(vm_offset_t va, vm_paddr_t pa);
253 void pmap_map_readonly(struct pmap *pmap, vm_offset_t va, int len);
254 void pmap_map_readwrite(struct pmap *pmap, vm_offset_t va, int len);
256 static __inline pt_entry_t
257 pte_load_store(pt_entry_t *ptep, pt_entry_t v)
262 PT_SET_VA(ptep, v, TRUE);
266 static __inline pt_entry_t
267 pte_load_store_ma(pt_entry_t *ptep, pt_entry_t v)
272 PT_SET_VA_MA(ptep, v, TRUE);
276 #define pte_load_clear(ptep) pte_load_store((ptep), (pt_entry_t)0ULL)
278 #define pte_store(ptep, pte) pte_load_store((ptep), (pt_entry_t)pte)
279 #define pte_store_ma(ptep, pte) pte_load_store_ma((ptep), (pt_entry_t)pte)
280 #define pde_store_ma(ptep, pte) pte_load_store_ma((ptep), (pt_entry_t)pte)
285 * KPTmap is a linear mapping of the kernel page table. It differs from the
286 * recursive mapping in two ways: (1) it only provides access to kernel page
287 * table pages, and not user page table pages, and (2) it provides access to
288 * a kernel page table page after the corresponding virtual addresses have
289 * been promoted to a 2/4MB page mapping.
291 * KPTmap is first initialized by locore to support just NPKT page table
292 * pages. Later, it is reinitialized by pmap_bootstrap() to allow for
293 * expansion of the kernel page table.
295 extern pt_entry_t *KPTmap;
298 * Extract from the kernel page table the physical address that is mapped by
299 * the given virtual address "va".
301 * This function may be used before pmap_bootstrap() is called.
303 static __inline vm_paddr_t
304 pmap_kextract(vm_offset_t va)
308 if ((pa = PTD[va >> PDRSHIFT]) & PG_PS) {
309 pa = (pa & PG_PS_FRAME) | (va & PDRMASK);
312 * Beware of a concurrent promotion that changes the PDE at
313 * this point! For example, vtopte() must not be used to
314 * access the PTE because it would use the new PDE. It is,
315 * however, safe to use the old PDE because the page table
316 * page is preserved by the promotion.
318 pa = KPTmap[i386_btop(va)];
319 pa = (pa & PG_FRAME) | (va & PAGE_MASK);
326 #define PT_UPDATES_FLUSH()
329 #if defined(PAE) && !defined(XEN)
331 #define pde_cmpset(pdep, old, new) atomic_cmpset_64_i586(pdep, old, new)
332 #define pte_load_store(ptep, pte) atomic_swap_64_i586(ptep, pte)
333 #define pte_load_clear(ptep) atomic_swap_64_i586(ptep, 0)
334 #define pte_store(ptep, pte) atomic_store_rel_64_i586(ptep, pte)
336 extern pt_entry_t pg_nx;
338 #elif !defined(PAE) && !defined(XEN)
340 #define pde_cmpset(pdep, old, new) atomic_cmpset_int(pdep, old, new)
341 #define pte_load_store(ptep, pte) atomic_swap_int(ptep, pte)
342 #define pte_load_clear(ptep) atomic_swap_int(ptep, 0)
343 #define pte_store(ptep, pte) do { \
344 *(u_int *)(ptep) = (u_int)(pte); \
349 #define pte_clear(ptep) pte_store(ptep, 0)
351 #define pde_store(pdep, pde) pte_store(pdep, pde)
362 TAILQ_HEAD(,pv_entry) pv_list;
368 pd_entry_t *pm_pdir; /* KVA of page directory */
369 TAILQ_HEAD(,pv_chunk) pm_pvchunk; /* list of mappings in pmap */
370 cpuset_t pm_active; /* active on cpus */
371 struct pmap_statistics pm_stats; /* pmap statistics */
372 LIST_ENTRY(pmap) pm_list; /* List of all pmaps */
374 pdpt_entry_t *pm_pdpt; /* KVA of page director pointer
377 struct vm_radix pm_root; /* spare page table pages */
380 typedef struct pmap *pmap_t;
383 extern struct pmap kernel_pmap_store;
384 #define kernel_pmap (&kernel_pmap_store)
386 #define PMAP_LOCK(pmap) mtx_lock(&(pmap)->pm_mtx)
387 #define PMAP_LOCK_ASSERT(pmap, type) \
388 mtx_assert(&(pmap)->pm_mtx, (type))
389 #define PMAP_LOCK_DESTROY(pmap) mtx_destroy(&(pmap)->pm_mtx)
390 #define PMAP_LOCK_INIT(pmap) mtx_init(&(pmap)->pm_mtx, "pmap", \
391 NULL, MTX_DEF | MTX_DUPOK)
392 #define PMAP_LOCKED(pmap) mtx_owned(&(pmap)->pm_mtx)
393 #define PMAP_MTX(pmap) (&(pmap)->pm_mtx)
394 #define PMAP_TRYLOCK(pmap) mtx_trylock(&(pmap)->pm_mtx)
395 #define PMAP_UNLOCK(pmap) mtx_unlock(&(pmap)->pm_mtx)
399 * For each vm_page_t, there is a list of all currently valid virtual
400 * mappings of that page. An entry is a pv_entry_t, the list is pv_list.
402 typedef struct pv_entry {
403 vm_offset_t pv_va; /* virtual address for mapping */
404 TAILQ_ENTRY(pv_entry) pv_next;
408 * pv_entries are allocated in chunks per-process. This avoids the
409 * need to track per-pmap assignments.
415 TAILQ_ENTRY(pv_chunk) pc_list;
416 uint32_t pc_map[_NPCM]; /* bitmap; 1 = free */
417 TAILQ_ENTRY(pv_chunk) pc_lru;
418 struct pv_entry pc_pventry[_NPCPV];
423 extern caddr_t CADDR3;
424 extern pt_entry_t *CMAP3;
425 extern vm_paddr_t phys_avail[];
426 extern vm_paddr_t dump_avail[];
429 extern char *ptvmmap; /* poor name! */
430 extern vm_offset_t virtual_avail;
431 extern vm_offset_t virtual_end;
433 #define pmap_page_get_memattr(m) ((vm_memattr_t)(m)->md.pat_mode)
434 #define pmap_page_is_write_mapped(m) (((m)->aflags & PGA_WRITEABLE) != 0)
435 #define pmap_unmapbios(va, sz) pmap_unmapdev((va), (sz))
438 * Only the following functions or macros may be used before pmap_bootstrap()
439 * is called: pmap_kenter(), pmap_kextract(), pmap_kremove(), vtophys(), and
442 void pmap_bootstrap(vm_paddr_t);
443 int pmap_cache_bits(int mode, boolean_t is_pde);
444 int pmap_change_attr(vm_offset_t, vm_size_t, int);
445 void pmap_init_pat(void);
446 void pmap_kenter(vm_offset_t va, vm_paddr_t pa);
447 void *pmap_kenter_temporary(vm_paddr_t pa, int i);
448 void pmap_kremove(vm_offset_t);
449 void *pmap_mapbios(vm_paddr_t, vm_size_t);
450 void *pmap_mapdev(vm_paddr_t, vm_size_t);
451 void *pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
452 boolean_t pmap_page_is_mapped(vm_page_t m);
453 void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
454 void pmap_unmapdev(vm_offset_t, vm_size_t);
455 pt_entry_t *pmap_pte(pmap_t, vm_offset_t) __pure2;
456 void pmap_invalidate_page(pmap_t, vm_offset_t);
457 void pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
458 void pmap_invalidate_all(pmap_t);
459 void pmap_invalidate_cache(void);
460 void pmap_invalidate_cache_pages(vm_page_t *pages, int count);
461 void pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
467 #endif /* !_MACHINE_PMAP_H_ */