vm_machdep.c revision 140
1/*- 2 * Copyright (c) 1982, 1986 The Regents of the University of California. 3 * Copyright (c) 1989, 1990 William Jolitz 4 * All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * the Systems Programming Group of the University of Utah Computer 8 * Science Department, and William Jolitz. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91 39 * 40 * PATCHES MAGIC LEVEL PATCH THAT GOT US HERE 41 * -------------------- ----- ---------------------- 42 * CURRENT PATCH LEVEL: 1 00154 43 * -------------------- ----- ---------------------- 44 * 45 * 20 Apr 93 Bruce Evans New npx-0.5 code 46 * 47 */ 48 49/* 50 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$ 51 */ 52static char rcsid[] = "$Header: /a/cvs/386BSD/src/sys.386bsd/i386/i386/vm_machdep.c,v 1.1.1.1 1993/06/12 14:58:05 rgrimes Exp $"; 53 54#include "param.h" 55#include "systm.h" 56#include "proc.h" 57#include "malloc.h" 58#include "buf.h" 59#include "user.h" 60 61#include "../include/cpu.h" 62 63#include "vm/vm.h" 64#include "vm/vm_kern.h" 65 66/* 67 * Finish a fork operation, with process p2 nearly set up. 68 * Copy and update the kernel stack and pcb, making the child 69 * ready to run, and marking it so that it can return differently 70 * than the parent. Returns 1 in the child process, 0 in the parent. 71 * We currently double-map the user area so that the stack is at the same 72 * address in each process; in the future we will probably relocate 73 * the frame pointers on the stack after copying. 74 */ 75cpu_fork(p1, p2) 76 register struct proc *p1, *p2; 77{ 78 register struct user *up = p2->p_addr; 79 int foo, offset, addr, i; 80 extern char kstack[]; 81 extern int mvesp(); 82 83 /* 84 * Copy pcb and stack from proc p1 to p2. 85 * We do this as cheaply as possible, copying only the active 86 * part of the stack. The stack and pcb need to agree; 87 * this is tricky, as the final pcb is constructed by savectx, 88 * but its frame isn't yet on the stack when the stack is copied. 89 * swtch compensates for this when the child eventually runs. 90 * This should be done differently, with a single call 91 * that copies and updates the pcb+stack, 92 * replacing the bcopy and savectx. 93 */ 94 p2->p_addr->u_pcb = p1->p_addr->u_pcb; 95 offset = mvesp() - (int)kstack; 96 bcopy((caddr_t)kstack + offset, (caddr_t)p2->p_addr + offset, 97 (unsigned) ctob(UPAGES) - offset); 98 p2->p_regs = p1->p_regs; 99 100 /* 101 * Wire top of address space of child to it's kstack. 102 * First, fault in a page of pte's to map it. 103 */ 104 addr = trunc_page((u_int)vtopte(kstack)); 105 vm_map_pageable(&p2->p_vmspace->vm_map, addr, addr+NBPG, FALSE); 106 for (i=0; i < UPAGES; i++) 107 pmap_enter(&p2->p_vmspace->vm_pmap, kstack+i*NBPG, 108 pmap_extract(kernel_pmap, ((int)p2->p_addr)+i*NBPG), VM_PROT_READ, 1); 109 110 pmap_activate(&p2->p_vmspace->vm_pmap, &up->u_pcb); 111 112 /* 113 * 114 * Arrange for a non-local goto when the new process 115 * is started, to resume here, returning nonzero from setjmp. 116 */ 117 if (savectx(up, 1)) { 118 /* 119 * Return 1 in child. 120 */ 121 return (1); 122 } 123 return (0); 124} 125 126#ifdef notyet 127/* 128 * cpu_exit is called as the last action during exit. 129 * 130 * We change to an inactive address space and a "safe" stack, 131 * passing thru an argument to the new stack. Now, safely isolated 132 * from the resources we're shedding, we release the address space 133 * and any remaining machine-dependent resources, including the 134 * memory for the user structure and kernel stack. 135 * 136 * Next, we assign a dummy context to be written over by swtch, 137 * calling it to send this process off to oblivion. 138 * [The nullpcb allows us to minimize cost in swtch() by not having 139 * a special case]. 140 */ 141struct proc *swtch_to_inactive(); 142volatile void 143cpu_exit(p) 144 register struct proc *p; 145{ 146 static struct pcb nullpcb; /* pcb to overwrite on last swtch */ 147 148#ifdef NPX 149 npxexit(p); 150#endif 151 152 /* move to inactive space and stack, passing arg accross */ 153 p = swtch_to_inactive(p); 154 155 /* drop per-process resources */ 156 vmspace_free(p->p_vmspace); 157 kmem_free(kernel_map, (vm_offset_t)p->p_addr, ctob(UPAGES)); 158 159 p->p_addr = (struct user *) &nullpcb; 160 splclock(); 161 swtch(); 162 /* NOTREACHED */ 163} 164#else 165volatile void 166cpu_exit(p) 167 register struct proc *p; 168{ 169 170#ifdef NPX 171 npxexit(p); 172#endif 173 splclock(); 174 swtch(); 175} 176 177cpu_wait(p) struct proc *p; { 178 179 /* drop per-process resources */ 180 vmspace_free(p->p_vmspace); 181 kmem_free(kernel_map, (vm_offset_t)p->p_addr, ctob(UPAGES)); 182} 183#endif 184 185/* 186 * Set a red zone in the kernel stack after the u. area. 187 */ 188setredzone(pte, vaddr) 189 u_short *pte; 190 caddr_t vaddr; 191{ 192/* eventually do this by setting up an expand-down stack segment 193 for ss0: selector, allowing stack access down to top of u. 194 this means though that protection violations need to be handled 195 thru a double fault exception that must do an integral task 196 switch to a known good context, within which a dump can be 197 taken. a sensible scheme might be to save the initial context 198 used by sched (that has physical memory mapped 1:1 at bottom) 199 and take the dump while still in mapped mode */ 200} 201 202/* 203 * Move pages from one kernel virtual address to another. 204 * Both addresses are assumed to reside in the Sysmap, 205 * and size must be a multiple of CLSIZE. 206 */ 207pagemove(from, to, size) 208 register caddr_t from, to; 209 int size; 210{ 211 register struct pte *fpte, *tpte; 212 213 if (size % CLBYTES) 214 panic("pagemove"); 215 fpte = kvtopte(from); 216 tpte = kvtopte(to); 217 while (size > 0) { 218 *tpte++ = *fpte; 219 *(int *)fpte++ = 0; 220 from += NBPG; 221 to += NBPG; 222 size -= NBPG; 223 } 224 tlbflush(); 225} 226 227/* 228 * Convert kernel VA to physical address 229 */ 230kvtop(addr) 231 register caddr_t addr; 232{ 233 vm_offset_t va; 234 235 va = pmap_extract(kernel_pmap, (vm_offset_t)addr); 236 if (va == 0) 237 panic("kvtop: zero page frame"); 238 return((int)va); 239} 240 241#ifdef notdef 242/* 243 * The probe[rw] routines should probably be redone in assembler 244 * for efficiency. 245 */ 246prober(addr) 247 register u_int addr; 248{ 249 register int page; 250 register struct proc *p; 251 252 if (addr >= USRSTACK) 253 return(0); 254 p = u.u_procp; 255 page = btop(addr); 256 if (page < dptov(p, p->p_dsize) || page > sptov(p, p->p_ssize)) 257 return(1); 258 return(0); 259} 260 261probew(addr) 262 register u_int addr; 263{ 264 register int page; 265 register struct proc *p; 266 267 if (addr >= USRSTACK) 268 return(0); 269 p = u.u_procp; 270 page = btop(addr); 271 if (page < dptov(p, p->p_dsize) || page > sptov(p, p->p_ssize)) 272 return((*(int *)vtopte(p, page) & PG_PROT) == PG_UW); 273 return(0); 274} 275 276/* 277 * NB: assumes a physically contiguous kernel page table 278 * (makes life a LOT simpler). 279 */ 280kernacc(addr, count, rw) 281 register u_int addr; 282 int count, rw; 283{ 284 register struct pde *pde; 285 register struct pte *pte; 286 register int ix, cnt; 287 extern long Syssize; 288 289 if (count <= 0) 290 return(0); 291 pde = (struct pde *)((u_int)u.u_procp->p_p0br + u.u_procp->p_szpt * NBPG); 292 ix = (addr & PD_MASK) >> PD_SHIFT; 293 cnt = ((addr + count + (1 << PD_SHIFT) - 1) & PD_MASK) >> PD_SHIFT; 294 cnt -= ix; 295 for (pde += ix; cnt; cnt--, pde++) 296 if (pde->pd_v == 0) 297 return(0); 298 ix = btop(addr-0xfe000000); 299 cnt = btop(addr-0xfe000000+count+NBPG-1); 300 if (cnt > (int)&Syssize) 301 return(0); 302 cnt -= ix; 303 for (pte = &Sysmap[ix]; cnt; cnt--, pte++) 304 if (pte->pg_v == 0 /*|| (rw == B_WRITE && pte->pg_prot == 1)*/) 305 return(0); 306 return(1); 307} 308 309useracc(addr, count, rw) 310 register u_int addr; 311 int count, rw; 312{ 313 register int (*func)(); 314 register u_int addr2; 315 extern int prober(), probew(); 316 317 if (count <= 0) 318 return(0); 319 addr2 = addr; 320 addr += count; 321 func = (rw == B_READ) ? prober : probew; 322 do { 323 if ((*func)(addr2) == 0) 324 return(0); 325 addr2 = (addr2 + NBPG) & ~PGOFSET; 326 } while (addr2 < addr); 327 return(1); 328} 329#endif 330 331extern vm_map_t phys_map; 332 333/* 334 * Map an IO request into kernel virtual address space. Requests fall into 335 * one of five catagories: 336 * 337 * B_PHYS|B_UAREA: User u-area swap. 338 * Address is relative to start of u-area (p_addr). 339 * B_PHYS|B_PAGET: User page table swap. 340 * Address is a kernel VA in usrpt (Usrptmap). 341 * B_PHYS|B_DIRTY: Dirty page push. 342 * Address is a VA in proc2's address space. 343 * B_PHYS|B_PGIN: Kernel pagein of user pages. 344 * Address is VA in user's address space. 345 * B_PHYS: User "raw" IO request. 346 * Address is VA in user's address space. 347 * 348 * All requests are (re)mapped into kernel VA space via the useriomap 349 * (a name with only slightly more meaning than "kernelmap") 350 */ 351vmapbuf(bp) 352 register struct buf *bp; 353{ 354 register int npf; 355 register caddr_t addr; 356 register long flags = bp->b_flags; 357 struct proc *p; 358 int off; 359 vm_offset_t kva; 360 register vm_offset_t pa; 361 362 if ((flags & B_PHYS) == 0) 363 panic("vmapbuf"); 364 addr = bp->b_saveaddr = bp->b_un.b_addr; 365 off = (int)addr & PGOFSET; 366 p = bp->b_proc; 367 npf = btoc(round_page(bp->b_bcount + off)); 368 kva = kmem_alloc_wait(phys_map, ctob(npf)); 369 bp->b_un.b_addr = (caddr_t) (kva + off); 370 while (npf--) { 371 pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t)addr); 372 if (pa == 0) 373 panic("vmapbuf: null page frame"); 374 pmap_enter(vm_map_pmap(phys_map), kva, trunc_page(pa), 375 VM_PROT_READ|VM_PROT_WRITE, TRUE); 376 addr += PAGE_SIZE; 377 kva += PAGE_SIZE; 378 } 379} 380 381/* 382 * Free the io map PTEs associated with this IO operation. 383 * We also invalidate the TLB entries and restore the original b_addr. 384 */ 385vunmapbuf(bp) 386 register struct buf *bp; 387{ 388 register int npf; 389 register caddr_t addr = bp->b_un.b_addr; 390 vm_offset_t kva; 391 392 if ((bp->b_flags & B_PHYS) == 0) 393 panic("vunmapbuf"); 394 npf = btoc(round_page(bp->b_bcount + ((int)addr & PGOFSET))); 395 kva = (vm_offset_t)((int)addr & ~PGOFSET); 396 kmem_free_wakeup(phys_map, kva, ctob(npf)); 397 bp->b_un.b_addr = bp->b_saveaddr; 398 bp->b_saveaddr = NULL; 399} 400 401/* 402 * Force reset the processor by invalidating the entire address space! 403 */ 404cpu_reset() { 405 406 /* force a shutdown by unmapping entire address space ! */ 407 bzero((caddr_t) PTD, NBPG); 408 409 /* "good night, sweet prince .... <THUNK!>" */ 410 tlbflush(); 411 /* NOTREACHED */ 412} 413