1 /* $NetBSD: pmap_private.h,v 1.4 2022/08/21 09:12:43 riastradh Exp $ */ 2 3 /* 4 * Copyright (c) 1997 Charles D. Cranor and Washington University. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 /* 29 * Copyright (c) 2001 Wasabi Systems, Inc. 30 * All rights reserved. 31 * 32 * Written by Frank van der Linden for Wasabi Systems, Inc. 33 * 34 * Redistribution and use in source and binary forms, with or without 35 * modification, are permitted provided that the following conditions 36 * are met: 37 * 1. Redistributions of source code must retain the above copyright 38 * notice, this list of conditions and the following disclaimer. 39 * 2. Redistributions in binary form must reproduce the above copyright 40 * notice, this list of conditions and the following disclaimer in the 41 * documentation and/or other materials provided with the distribution. 42 * 3. All advertising materials mentioning features or use of this software 43 * must display the following acknowledgement: 44 * This product includes software developed for the NetBSD Project by 45 * Wasabi Systems, Inc. 46 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 47 * or promote products derived from this software without specific prior 48 * written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 52 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 53 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 54 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 55 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 56 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 57 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 58 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 59 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 60 * POSSIBILITY OF SUCH DAMAGE. 61 */ 62 63 #ifndef _I386_PMAP_H_ 64 #define _I386_PMAP_H_ 65 66 #if defined(_KERNEL_OPT) 67 #include "opt_xen.h" 68 #endif 69 70 #include <sys/atomic.h> 71 72 #include <i386/pte.h> 73 #include <i386/vmparam.h> 74 #include <machine/segments.h> 75 #if defined(_KERNEL) 76 #include <machine/cpufunc.h> 77 #endif 78 79 #include <uvm/uvm_object.h> 80 #ifdef XENPV 81 #include <xen/xenfunc.h> 82 #include <xen/xenpmap.h> 83 #endif /* XENPV */ 84 85 /* 86 * see pte.h for a description of i386 MMU terminology and hardware 87 * interface. 88 * 89 * a pmap describes a processes' 4GB virtual address space. when PAE 90 * is not in use, this virtual address space can be broken up into 1024 4MB 91 * regions which are described by PDEs in the PDP. the PDEs are defined as 92 * follows: 93 * 94 * (ranges are inclusive -> exclusive, just like vm_map_entry start/end) 95 * (the following assumes that KERNBASE is 0xc0000000) 96 * 97 * PDE#s VA range usage 98 * 0->766 0x0 -> 0xbfc00000 user address space 99 * 767 0xbfc00000-> recursive mapping of PDP (used for 100 * 0xc0000000 linear mapping of PTPs) 101 * 768->1023 0xc0000000-> kernel address space (constant 102 * 0xffc00000 across all pmap's/processes) 103 * <end> 104 * 105 * 106 * note: a recursive PDP mapping provides a way to map all the PTEs for 107 * a 4GB address space into a linear chunk of virtual memory. in other 108 * words, the PTE for page 0 is the first int mapped into the 4MB recursive 109 * area. the PTE for page 1 is the second int. the very last int in the 110 * 4MB range is the PTE that maps VA 0xfffff000 (the last page in a 4GB 111 * address). 112 * 113 * all pmap's PD's must have the same values in slots 768->1023 so that 114 * the kernel is always mapped in every process. these values are loaded 115 * into the PD at pmap creation time. 116 * 117 * at any one time only one pmap can be active on a processor. this is 118 * the pmap whose PDP is pointed to by processor register %cr3. this pmap 119 * will have all its PTEs mapped into memory at the recursive mapping 120 * point (slot #767 as show above). when the pmap code wants to find the 121 * PTE for a virtual address, all it has to do is the following: 122 * 123 * address of PTE = (767 * 4MB) + (VA / PAGE_SIZE) * sizeof(pt_entry_t) 124 * = 0xbfc00000 + (VA / 4096) * 4 125 * 126 * what happens if the pmap layer is asked to perform an operation 127 * on a pmap that is not the one which is currently active? in that 128 * case we temporarily load this pmap, perform the operation, and mark 129 * the currently active one as pending lazy reload. 130 * 131 * the following figure shows the effects of the recursive PDP mapping: 132 * 133 * PDP (%cr3) 134 * +----+ 135 * | 0| -> PTP#0 that maps VA 0x0 -> 0x400000 136 * | | 137 * | | 138 * | 767| -> points back to PDP (%cr3) mapping VA 0xbfc00000 -> 0xc0000000 139 * | 768| -> first kernel PTP (maps 0xc0000000 -> 0xc0400000) 140 * | | 141 * +----+ 142 * 143 * note that the PDE#767 VA (0xbfc00000) is defined as "PTE_BASE" 144 * 145 * starting at VA 0xbfc00000 the current active PDP (%cr3) acts as a 146 * PTP: 147 * 148 * PTP#767 == PDP(%cr3) => maps VA 0xbfc00000 -> 0xc0000000 149 * +----+ 150 * | 0| -> maps the contents of PTP#0 at VA 0xbfc00000->0xbfc01000 151 * | | 152 * | | 153 * | 767| -> maps contents of PTP#767 (the PDP) at VA 0xbfeff000 154 * | 768| -> maps contents of first kernel PTP 155 * | | 156 * |1023| 157 * +----+ 158 * 159 * note that mapping of the PDP at PTP#767's VA (0xbfeff000) is 160 * defined as "PDP_BASE".... within that mapping there are two 161 * defines: 162 * "PDP_PDE" (0xbfeffbfc) is the VA of the PDE in the PDP 163 * which points back to itself. 164 * 165 * - PAE support - 166 * --------------- 167 * 168 * PAE adds another layer of indirection during address translation, breaking 169 * up the translation process in 3 different levels: 170 * - L3 page directory, containing 4 * 64-bits addresses (index determined by 171 * bits [31:30] from the virtual address). This breaks up the address space 172 * in 4 1GB regions. 173 * - the PD (L2), containing 512 64-bits addresses, breaking each L3 region 174 * in 512 * 2MB regions. 175 * - the PT (L1), also containing 512 64-bits addresses (at L1, the size of 176 * the pages is still 4K). 177 * 178 * The kernel virtual space is mapped by the last entry in the L3 page, 179 * the first 3 entries mapping the user VA space. 180 * 181 * Because the L3 has only 4 entries of 1GB each, we can't use recursive 182 * mappings at this level for PDP_PDE (this would eat up 2 of the 4GB 183 * virtual space). There are also restrictions imposed by Xen on the 184 * last entry of the L3 PD (reference count to this page cannot be 185 * bigger than 1), which makes it hard to use one L3 page per pmap to 186 * switch between pmaps using %cr3. 187 * 188 * As such, each CPU gets its own L3 page that is always loaded into its %cr3 189 * (ci_pae_l3_pd in the associated cpu_info struct). We claim that the VM has 190 * only a 2-level PTP (similar to the non-PAE case). L2 PD is now 4 contiguous 191 * pages long (corresponding to the 4 entries of the L3), and the different 192 * index/slots (like PDP_PDE) are adapted accordingly. 193 * 194 * Kernel space remains in L3[3], L3[0-2] maps the user VA space. Switching 195 * between pmaps consists in modifying the first 3 entries of the CPU's L3 page. 196 * 197 * PTE_BASE will need 4 entries in the L2 PD pages to map the L2 pages 198 * recursively. 199 * 200 * In addition, for Xen, we can't recursively map L3[3] (Xen wants the ref 201 * count on this page to be exactly one), so we use a shadow PD page for 202 * the last L2 PD. The shadow page could be static too, but to make pm_pdir[] 203 * contiguous we'll allocate/copy one page per pmap. 204 */ 205 206 /* 207 * the following defines give the virtual addresses of various MMU 208 * data structures: 209 * PTE_BASE: the base VA of the linear PTE mappings 210 * PDP_BASE: the base VA of the recursive mapping of the PDP 211 * PDP_PDE: the VA of the PDE that points back to the PDP 212 */ 213 214 #define PTE_BASE ((pt_entry_t *) (PDIR_SLOT_PTE * NBPD_L2)) 215 216 #define L1_BASE PTE_BASE 217 218 #define L2_BASE ((pd_entry_t *)((char *)L1_BASE + L2_SLOT_PTE * NBPD_L1)) 219 220 #define PDP_PDE (L2_BASE + PDIR_SLOT_PTE) 221 222 #define PDP_BASE L2_BASE 223 224 #define NPDPG (PAGE_SIZE / sizeof (pd_entry_t)) 225 226 #define PTP_MASK_INITIALIZER { L1_MASK, L2_MASK } 227 #define PTP_FRAME_INITIALIZER { L1_FRAME, L2_FRAME } 228 #define PTP_SHIFT_INITIALIZER { L1_SHIFT, L2_SHIFT } 229 #define NKPTP_INITIALIZER { NKL1_START_ENTRIES, NKL2_START_ENTRIES } 230 #define NKPTPMAX_INITIALIZER { NKL1_MAX_ENTRIES, NKL2_MAX_ENTRIES } 231 #define NBPD_INITIALIZER { NBPD_L1, NBPD_L2 } 232 #define PDES_INITIALIZER { L2_BASE } 233 234 #define PTP_LEVELS 2 235 236 /* 237 * PTE_AVL usage: we make use of the ignored bits of the PTE 238 */ 239 #define PTE_WIRED PTE_AVL1 /* Wired Mapping */ 240 #define PTE_PVLIST PTE_AVL2 /* Mapping has entry on pvlist */ 241 #define PTE_X PTE_AVL3 /* Executable */ 242 243 /* XXX To be deleted. */ 244 #define PG_W PTE_WIRED 245 #define PG_PVLIST PTE_PVLIST 246 #define PG_X PTE_X 247 248 #define _MACHINE_PMAP_PRIVATE_H_X86 249 #include <x86/pmap_private.h> 250 #undef _MACHINE_PMAP_PRIVATE_H_X86 251 252 #ifndef XENPV 253 254 #define pmap_pa2pte(a) (a) 255 #define pmap_pte2pa(a) ((a) & PTE_FRAME) 256 #define pmap_pte_set(p, n) do { *(p) = (n); } while (0) 257 #define pmap_pte_flush() /* nothing */ 258 259 #ifdef PAE 260 #define pmap_pte_cas(p, o, n) atomic_cas_64((p), (o), (n)) 261 #define pmap_pte_testset(p, n) \ 262 atomic_swap_64((volatile uint64_t *)p, n) 263 #define pmap_pte_setbits(p, b) \ 264 atomic_or_64((volatile uint64_t *)p, b) 265 #define pmap_pte_clearbits(p, b) \ 266 atomic_and_64((volatile uint64_t *)p, ~(b)) 267 #else /* PAE */ 268 #define pmap_pte_cas(p, o, n) atomic_cas_32((p), (o), (n)) 269 #define pmap_pte_testset(p, n) \ 270 atomic_swap_ulong((volatile unsigned long *)p, n) 271 #define pmap_pte_setbits(p, b) \ 272 atomic_or_ulong((volatile unsigned long *)p, b) 273 #define pmap_pte_clearbits(p, b) \ 274 atomic_and_ulong((volatile unsigned long *)p, ~(b)) 275 #endif /* PAE */ 276 277 #else /* XENPV */ 278 279 extern kmutex_t pte_lock; 280 281 static __inline pt_entry_t 282 pmap_pa2pte(paddr_t pa) 283 { 284 return (pt_entry_t)xpmap_ptom_masked(pa); 285 } 286 287 static __inline paddr_t 288 pmap_pte2pa(pt_entry_t pte) 289 { 290 return xpmap_mtop_masked(pte & PTE_FRAME); 291 } 292 293 static __inline void 294 pmap_pte_set(pt_entry_t *pte, pt_entry_t npte) 295 { 296 int s = splvm(); 297 xpq_queue_pte_update(xpmap_ptetomach(pte), npte); 298 splx(s); 299 } 300 301 static __inline pt_entry_t 302 pmap_pte_cas(volatile pt_entry_t *ptep, pt_entry_t o, pt_entry_t n) 303 { 304 pt_entry_t opte; 305 306 mutex_enter(&pte_lock); 307 opte = *ptep; 308 if (opte == o) { 309 xpq_queue_pte_update(xpmap_ptetomach(__UNVOLATILE(ptep)), n); 310 xpq_flush_queue(); 311 } 312 mutex_exit(&pte_lock); 313 return opte; 314 } 315 316 static __inline pt_entry_t 317 pmap_pte_testset(volatile pt_entry_t *pte, pt_entry_t npte) 318 { 319 pt_entry_t opte; 320 321 mutex_enter(&pte_lock); 322 opte = *pte; 323 xpq_queue_pte_update(xpmap_ptetomach(__UNVOLATILE(pte)), npte); 324 xpq_flush_queue(); 325 mutex_exit(&pte_lock); 326 return opte; 327 } 328 329 static __inline void 330 pmap_pte_setbits(volatile pt_entry_t *pte, pt_entry_t bits) 331 { 332 mutex_enter(&pte_lock); 333 xpq_queue_pte_update(xpmap_ptetomach(__UNVOLATILE(pte)), (*pte) | bits); 334 xpq_flush_queue(); 335 mutex_exit(&pte_lock); 336 } 337 338 static __inline void 339 pmap_pte_clearbits(volatile pt_entry_t *pte, pt_entry_t bits) 340 { 341 mutex_enter(&pte_lock); 342 xpq_queue_pte_update(xpmap_ptetomach(__UNVOLATILE(pte)), 343 (*pte) & ~bits); 344 xpq_flush_queue(); 345 mutex_exit(&pte_lock); 346 } 347 348 static __inline void 349 pmap_pte_flush(void) 350 { 351 int s = splvm(); 352 xpq_flush_queue(); 353 splx(s); 354 } 355 356 #endif 357 358 struct vm_map; 359 struct trapframe; 360 struct pcb; 361 362 int pmap_exec_fixup(struct vm_map *, struct trapframe *, struct pcb *); 363 364 #endif /* _I386_PMAP_H_ */ 365