1 /* $NetBSD: pte.h,v 1.27 2011/02/01 20:09:08 chuck Exp $ */ 2 3 /* 4 * Copyright (c) 2001 Wasabi Systems, Inc. 5 * All rights reserved. 6 * 7 * Written by Frank van der Linden for Wasabi Systems, Inc. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 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 * 3. All advertising materials mentioning features or use of this software 18 * must display the following acknowledgement: 19 * This product includes software developed for the NetBSD Project by 20 * Wasabi Systems, Inc. 21 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 22 * or promote products derived from this software without specific prior 23 * written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35 * POSSIBILITY OF SUCH DAMAGE. 36 */ 37 38 /* 39 * Copyright (c) 1997 Charles D. Cranor and Washington University. 40 * All rights reserved. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 51 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 52 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 53 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 54 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 55 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 56 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 57 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 58 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 59 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 60 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 61 */ 62 63 /* 64 * pte.h rewritten by chuck based on the jolitz version, plus random 65 * info on the pentium and other processors found on the net. the 66 * goal of this rewrite is to provide enough documentation on the MMU 67 * hardware that the reader will be able to understand it without having 68 * to refer to a hardware manual. 69 */ 70 71 #ifndef _I386_PTE_H_ 72 #define _I386_PTE_H_ 73 #ifdef _KERNEL_OPT 74 #include "opt_xen.h" 75 #endif 76 77 /* 78 * i386 MMU hardware structure (without PAE extension): 79 * 80 * the i386 MMU is a two-level MMU which maps 4GB of virtual memory. 81 * the pagesize is 4K (4096 [0x1000] bytes), although newer pentium 82 * processors can support a 4MB pagesize as well. 83 * 84 * the first level table (segment table?) is called a "page directory" 85 * and it contains 1024 page directory entries (PDEs). each PDE is 86 * 4 bytes (an int), so a PD fits in a single 4K page. this page is 87 * the page directory page (PDP). each PDE in a PDP maps 4MB of space 88 * (1024 * 4MB = 4GB). a PDE contains the physical address of the 89 * second level table: the page table. or, if 4MB pages are being used, 90 * then the PDE contains the PA of the 4MB page being mapped. 91 * 92 * a page table consists of 1024 page table entries (PTEs). each PTE is 93 * 4 bytes (an int), so a page table also fits in a single 4K page. a 94 * 4K page being used as a page table is called a page table page (PTP). 95 * each PTE in a PTP maps one 4K page (1024 * 4K = 4MB). a PTE contains 96 * the physical address of the page it maps and some flag bits (described 97 * below). 98 * 99 * the processor has a special register, "cr3", which points to the 100 * the PDP which is currently controlling the mappings of the virtual 101 * address space. 102 * 103 * the following picture shows the translation process for a 4K page: 104 * 105 * %cr3 register [PA of PDP] 106 * | 107 * | 108 * | bits <31-22> of VA bits <21-12> of VA bits <11-0> 109 * | index the PDP (0 - 1023) index the PTP are the page offset 110 * | | | | 111 * | v | | 112 * +--->+----------+ | | 113 * | PD Page | PA of v | 114 * | |---PTP-------->+------------+ | 115 * | 1024 PDE | | page table |--PTE--+ | 116 * | entries | | (aka PTP) | | | 117 * +----------+ | 1024 PTE | | | 118 * | entries | | | 119 * +------------+ | | 120 * | | 121 * bits <31-12> bits <11-0> 122 * p h y s i c a l a d d r 123 * 124 * the i386 caches PTEs in a TLB. it is important to flush out old 125 * TLB mappings when making a change to a mappings. writing to the 126 * %cr3 will flush the entire TLB. newer processors also have an 127 * instruction that will invalidate the mapping of a single page (which 128 * is useful if you are changing a single mappings because it preserves 129 * all the cached TLB entries). 130 * 131 * as shows, bits 31-12 of the PTE contain PA of the page being mapped. 132 * the rest of the PTE is defined as follows: 133 * bit# name use 134 * 11 n/a available for OS use, hardware ignores it 135 * 10 n/a available for OS use, hardware ignores it 136 * 9 n/a available for OS use, hardware ignores it 137 * 8 G global bit (see discussion below) 138 * 7 PS page size [for PDEs] (0=4k, 1=4M <if supported>) 139 * 6 D dirty (modified) page 140 * 5 A accessed (referenced) page 141 * 4 PCD cache disable 142 * 3 PWT prevent write through (cache) 143 * 2 U/S user/supervisor bit (0=supervisor only, 1=both u&s) 144 * 1 R/W read/write bit (0=read only, 1=read-write) 145 * 0 P present (valid) 146 * 147 * notes: 148 * - PS is only supported on newer processors 149 * - PTEs with the G bit are global in the sense that they are not 150 * flushed from the TLB when %cr3 is written (to flush, use the 151 * "flush single page" instruction). this is only supported on 152 * newer processors. this bit can be used to keep the kernel's 153 * TLB entries around while context switching. since the kernel 154 * is mapped into all processes at the same place it does not make 155 * sense to flush these entries when switching from one process' 156 * pmap to another. 157 * 158 * The PAE extension extends the size of the PTE to 64 bits (52bits physical 159 * address) and is compatible with the amd64 PTE format. The first level 160 * maps 2M, the second 1G, so a third level page table is introduced to 161 * map the 4GB virtual address space. This PD has only 4 entries. 162 * We can't use recursive mapping at level 3 to map the PD pages, as this 163 * would eat one GB of address space. In addition, Xen imposes restrictions 164 * on the entries we put in the L3 page (for example, the page pointed to by 165 * the last slot can't be shared among different L3 pages), which makes 166 * handling this L3 page in the same way we do for L2 on i386 (or L4 on amd64) 167 * difficult. For most things we'll just pretend to have only 2 levels, 168 * with the 2 high bits of the L2 index being in fact the index in the 169 * L3. 170 */ 171 172 #if !defined(_LOCORE) 173 174 /* 175 * here we define the data types for PDEs and PTEs 176 */ 177 #ifdef PAE 178 typedef uint64_t pd_entry_t; /* PDE */ 179 typedef uint64_t pt_entry_t; /* PTE */ 180 #else 181 typedef uint32_t pd_entry_t; /* PDE */ 182 typedef uint32_t pt_entry_t; /* PTE */ 183 #endif 184 185 #endif 186 187 /* 188 * now we define various for playing with virtual addresses 189 */ 190 191 #ifdef PAE 192 #define L1_SHIFT 12 193 #define L2_SHIFT 21 194 #define L3_SHIFT 30 195 #define NBPD_L1 (1ULL << L1_SHIFT) /* # bytes mapped by L1 ent (4K) */ 196 #define NBPD_L2 (1ULL << L2_SHIFT) /* # bytes mapped by L2 ent (2MB) */ 197 #define NBPD_L3 (1ULL << L3_SHIFT) /* # bytes mapped by L3 ent (1GB) */ 198 199 #define L3_MASK 0xc0000000 200 #define L2_REALMASK 0x3fe00000 201 #define L2_MASK (L2_REALMASK | L3_MASK) 202 #define L1_MASK 0x001ff000 203 204 #define L3_FRAME (L3_MASK) 205 #define L2_FRAME (L3_FRAME | L2_MASK) 206 #define L1_FRAME (L2_FRAME|L1_MASK) 207 208 #define PG_FRAME 0x000ffffffffff000ULL /* page frame mask */ 209 #define PG_LGFRAME 0x000fffffffe00000ULL /* large (2MB) page frame mask */ 210 211 /* macros to get real L2 and L3 index, from our "extended" L2 index */ 212 #define l2tol3(idx) ((idx) >> (L3_SHIFT - L2_SHIFT)) 213 #define l2tol2(idx) ((idx) & (L2_REALMASK >> L2_SHIFT)) 214 215 #else /* PAE */ 216 217 #define L1_SHIFT 12 218 #define L2_SHIFT 22 219 #define NBPD_L1 (1UL << L1_SHIFT) /* # bytes mapped by L1 ent (4K) */ 220 #define NBPD_L2 (1UL << L2_SHIFT) /* # bytes mapped by L2 ent (4MB) */ 221 222 #define L2_MASK 0xffc00000 223 #define L1_MASK 0x003ff000 224 225 #define L2_FRAME (L2_MASK) 226 #define L1_FRAME (L2_FRAME|L1_MASK) 227 228 #define PG_FRAME 0xfffff000 /* page frame mask */ 229 #define PG_LGFRAME 0xffc00000 /* large (4MB) page frame mask */ 230 231 #endif /* PAE */ 232 /* 233 * here we define the bits of the PDE/PTE, as described above: 234 * 235 * XXXCDC: need to rename these (PG_u == ugly). 236 */ 237 238 #define PG_V 0x00000001 /* valid entry */ 239 #define PG_RO 0x00000000 /* read-only page */ 240 #define PG_RW 0x00000002 /* read-write page */ 241 #define PG_u 0x00000004 /* user accessible page */ 242 #define PG_PROT 0x00000806 /* all protection bits */ 243 #define PG_WT 0x00000008 /* write through */ 244 #define PG_N 0x00000010 /* non-cacheable */ 245 #define PG_U 0x00000020 /* has been used */ 246 #define PG_M 0x00000040 /* has been modified */ 247 #define PG_PAT 0x00000080 /* PAT (on pte) */ 248 #define PG_PS 0x00000080 /* 4MB page size (2MB for PAE) */ 249 #define PG_G 0x00000100 /* global, don't TLB flush */ 250 #define PG_AVAIL1 0x00000200 /* ignored by hardware */ 251 #define PG_AVAIL2 0x00000400 /* ignored by hardware */ 252 #define PG_AVAIL3 0x00000800 /* ignored by hardware */ 253 #define PG_LGPAT 0x00001000 /* PAT on large pages */ 254 255 /* 256 * various short-hand protection codes 257 */ 258 259 #define PG_KR 0x00000000 /* kernel read-only */ 260 #define PG_KW 0x00000002 /* kernel read-write */ 261 262 #ifdef PAE 263 #define PG_NX 0x8000000000000000ULL /* No-execute */ 264 #else 265 #define PG_NX 0 /* dummy */ 266 #endif 267 268 #include <x86/pte.h> 269 270 #endif /* _I386_PTE_H_ */ 271