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 41 /* 42 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$ 43 */ 44 45 #include "param.h" 46 #include "systm.h" 47 #include "proc.h" 48 #include "malloc.h" 49 #include "buf.h" 50 #include "user.h" 51 52 #include "../include/cpu.h" 53 54 #include "vm/vm.h" 55 #include "vm/vm_kern.h" 56 57 /* 58 * Finish a fork operation, with process p2 nearly set up. 59 * Copy and update the kernel stack and pcb, making the child 60 * ready to run, and marking it so that it can return differently 61 * than the parent. Returns 1 in the child process, 0 in the parent. 62 * We currently double-map the user area so that the stack is at the same 63 * address in each process; in the future we will probably relocate 64 * the frame pointers on the stack after copying. 65 */ 66 cpu_fork(p1, p2) 67 register struct proc *p1, *p2; 68 { 69 register struct user *up = p2->p_addr; 70 int foo, offset, addr, i; 71 extern char kstack[]; 72 extern int mvesp(); 73 74 /* 75 * Copy pcb and stack from proc p1 to p2. 76 * We do this as cheaply as possible, copying only the active 77 * part of the stack. The stack and pcb need to agree; 78 * this is tricky, as the final pcb is constructed by savectx, 79 * but its frame isn't yet on the stack when the stack is copied. 80 * swtch compensates for this when the child eventually runs. 81 * This should be done differently, with a single call 82 * that copies and updates the pcb+stack, 83 * replacing the bcopy and savectx. 84 */ 85 p2->p_addr->u_pcb = p1->p_addr->u_pcb; 86 offset = mvesp() - (int)kstack; 87 bcopy((caddr_t)kstack + offset, (caddr_t)p2->p_addr + offset, 88 (unsigned) ctob(UPAGES) - offset); 89 p2->p_regs = p1->p_regs; 90 91 /* 92 * Wire top of address space of child to it's u. 93 * First, fault in a page of pte's to map it. 94 */ 95 addr = trunc_page((u_int)vtopte(kstack)); 96 (void)vm_map_pageable(&p2->p_vmspace->vm_map, addr, addr+NBPG, FALSE); 97 for (i=0; i < UPAGES; i++) 98 pmap_enter(&p2->p_vmspace->vm_pmap, kstack+i*NBPG, 99 pmap_extract(kernel_pmap, ((int)p2->p_addr)+i*NBPG), 100 VM_PROT_READ, 1); 101 pmap_activate(&p2->p_vmspace->vm_pmap, &up->u_pcb); 102 103 /* 104 * 105 * Arrange for a non-local goto when the new process 106 * is started, to resume here, returning nonzero from setjmp. 107 */ 108 if (savectx(up, 1)) { 109 /* 110 * Return 1 in child. 111 */ 112 return (1); 113 } 114 return (0); 115 } 116 117 extern struct proc *npxproc; 118 119 #ifdef notyet 120 /* 121 * cpu_exit is called as the last action during exit. 122 * 123 * We change to an inactive address space and a "safe" stack, 124 * passing thru an argument to the new stack. Now, safely isolated 125 * from the resources we're shedding, we release the address space 126 * and any remaining machine-dependent resources, including the 127 * memory for the user structure and kernel stack. 128 * 129 * Next, we assign a dummy context to be written over by swtch, 130 * calling it to send this process off to oblivion. 131 * [The nullpcb allows us to minimize cost in swtch() by not having 132 * a special case]. 133 */ 134 struct proc *swtch_to_inactive(); 135 cpu_exit(p) 136 register struct proc *p; 137 { 138 static struct pcb nullpcb; /* pcb to overwrite on last swtch */ 139 140 /* free cporcessor (if we have it) */ 141 if( p == npxproc) npxproc =0; 142 143 /* move to inactive space and stack, passing arg accross */ 144 p = swtch_to_inactive(p); 145 146 /* drop per-process resources */ 147 vmspace_free(p->p_vmspace); 148 kmem_free(kernel_map, (vm_offset_t)p->p_addr, ctob(UPAGES)); 149 150 p->p_addr = (struct user *) &nullpcb; 151 swtch(); 152 /* NOTREACHED */ 153 } 154 #else 155 cpu_exit(p) 156 register struct proc *p; 157 { 158 159 /* free coprocessor (if we have it) */ 160 if( p == npxproc) npxproc =0; 161 162 curproc = p; 163 swtch(); 164 } 165 166 cpu_wait(p) struct proc *p; { 167 168 /* drop per-process resources */ 169 vmspace_free(p->p_vmspace); 170 kmem_free(kernel_map, (vm_offset_t)p->p_addr, ctob(UPAGES)); 171 } 172 #endif 173 174 /* 175 * Set a red zone in the kernel stack after the u. area. 176 */ 177 setredzone(pte, vaddr) 178 u_short *pte; 179 caddr_t vaddr; 180 { 181 /* eventually do this by setting up an expand-down stack segment 182 for ss0: selector, allowing stack access down to top of u. 183 this means though that protection violations need to be handled 184 thru a double fault exception that must do an integral task 185 switch to a known good context, within which a dump can be 186 taken. a sensible scheme might be to save the initial context 187 used by sched (that has physical memory mapped 1:1 at bottom) 188 and take the dump while still in mapped mode */ 189 } 190 191 /* 192 * Move pages from one kernel virtual address to another. 193 * Both addresses are assumed to reside in the Sysmap, 194 * and size must be a multiple of CLSIZE. 195 */ 196 pagemove(from, to, size) 197 register caddr_t from, to; 198 int size; 199 { 200 register struct pte *fpte, *tpte; 201 202 if (size % CLBYTES) 203 panic("pagemove"); 204 fpte = kvtopte(from); 205 tpte = kvtopte(to); 206 while (size > 0) { 207 *tpte++ = *fpte; 208 *(int *)fpte++ = 0; 209 from += NBPG; 210 to += NBPG; 211 size -= NBPG; 212 } 213 tlbflush(); 214 } 215 216 /* 217 * Convert kernel VA to physical address 218 */ 219 kvtop(addr) 220 register caddr_t addr; 221 { 222 vm_offset_t va; 223 224 va = pmap_extract(kernel_pmap, (vm_offset_t)addr); 225 if (va == 0) 226 panic("kvtop: zero page frame"); 227 return((int)va); 228 } 229 230 #ifdef notdef 231 /* 232 * The probe[rw] routines should probably be redone in assembler 233 * for efficiency. 234 */ 235 prober(addr) 236 register u_int addr; 237 { 238 register int page; 239 register struct proc *p; 240 241 if (addr >= USRSTACK) 242 return(0); 243 p = u.u_procp; 244 page = btop(addr); 245 if (page < dptov(p, p->p_dsize) || page > sptov(p, p->p_ssize)) 246 return(1); 247 return(0); 248 } 249 250 probew(addr) 251 register u_int addr; 252 { 253 register int page; 254 register struct proc *p; 255 256 if (addr >= USRSTACK) 257 return(0); 258 p = u.u_procp; 259 page = btop(addr); 260 if (page < dptov(p, p->p_dsize) || page > sptov(p, p->p_ssize)) 261 return((*(int *)vtopte(p, page) & PG_PROT) == PG_UW); 262 return(0); 263 } 264 265 /* 266 * NB: assumes a physically contiguous kernel page table 267 * (makes life a LOT simpler). 268 */ 269 kernacc(addr, count, rw) 270 register u_int addr; 271 int count, rw; 272 { 273 register struct pde *pde; 274 register struct pte *pte; 275 register int ix, cnt; 276 extern long Syssize; 277 278 if (count <= 0) 279 return(0); 280 pde = (struct pde *)((u_int)u.u_procp->p_p0br + u.u_procp->p_szpt * NBPG); 281 ix = (addr & PD_MASK) >> PD_SHIFT; 282 cnt = ((addr + count + (1 << PD_SHIFT) - 1) & PD_MASK) >> PD_SHIFT; 283 cnt -= ix; 284 for (pde += ix; cnt; cnt--, pde++) 285 if (pde->pd_v == 0) 286 return(0); 287 ix = btop(addr-0xfe000000); 288 cnt = btop(addr-0xfe000000+count+NBPG-1); 289 if (cnt > (int)&Syssize) 290 return(0); 291 cnt -= ix; 292 for (pte = &Sysmap[ix]; cnt; cnt--, pte++) 293 if (pte->pg_v == 0 /*|| (rw == B_WRITE && pte->pg_prot == 1)*/) 294 return(0); 295 return(1); 296 } 297 298 useracc(addr, count, rw) 299 register u_int addr; 300 int count, rw; 301 { 302 register int (*func)(); 303 register u_int addr2; 304 extern int prober(), probew(); 305 306 if (count <= 0) 307 return(0); 308 addr2 = addr; 309 addr += count; 310 func = (rw == B_READ) ? prober : probew; 311 do { 312 if ((*func)(addr2) == 0) 313 return(0); 314 addr2 = (addr2 + NBPG) & ~PGOFSET; 315 } while (addr2 < addr); 316 return(1); 317 } 318 #endif 319 320 extern vm_map_t phys_map; 321 322 /* 323 * Map an IO request into kernel virtual address space. Requests fall into 324 * one of five catagories: 325 * 326 * B_PHYS|B_UAREA: User u-area swap. 327 * Address is relative to start of u-area (p_addr). 328 * B_PHYS|B_PAGET: User page table swap. 329 * Address is a kernel VA in usrpt (Usrptmap). 330 * B_PHYS|B_DIRTY: Dirty page push. 331 * Address is a VA in proc2's address space. 332 * B_PHYS|B_PGIN: Kernel pagein of user pages. 333 * Address is VA in user's address space. 334 * B_PHYS: User "raw" IO request. 335 * Address is VA in user's address space. 336 * 337 * All requests are (re)mapped into kernel VA space via the useriomap 338 * (a name with only slightly more meaning than "kernelmap") 339 */ 340 vmapbuf(bp) 341 register struct buf *bp; 342 { 343 register int npf; 344 register caddr_t addr; 345 register long flags = bp->b_flags; 346 struct proc *p; 347 int off; 348 vm_offset_t kva; 349 register vm_offset_t pa; 350 351 if ((flags & B_PHYS) == 0) 352 panic("vmapbuf"); 353 addr = bp->b_saveaddr = bp->b_un.b_addr; 354 off = (int)addr & PGOFSET; 355 p = bp->b_proc; 356 npf = btoc(round_page(bp->b_bcount + off)); 357 kva = kmem_alloc_wait(phys_map, ctob(npf)); 358 bp->b_un.b_addr = (caddr_t) (kva + off); 359 while (npf--) { 360 pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t)addr); 361 if (pa == 0) 362 panic("vmapbuf: null page frame"); 363 pmap_enter(vm_map_pmap(phys_map), kva, trunc_page(pa), 364 VM_PROT_READ|VM_PROT_WRITE, TRUE); 365 addr += PAGE_SIZE; 366 kva += PAGE_SIZE; 367 } 368 } 369 370 /* 371 * Free the io map PTEs associated with this IO operation. 372 * We also invalidate the TLB entries and restore the original b_addr. 373 */ 374 vunmapbuf(bp) 375 register struct buf *bp; 376 { 377 register int npf; 378 register caddr_t addr = bp->b_un.b_addr; 379 vm_offset_t kva; 380 381 if ((bp->b_flags & B_PHYS) == 0) 382 panic("vunmapbuf"); 383 npf = btoc(round_page(bp->b_bcount + ((int)addr & PGOFSET))); 384 kva = (vm_offset_t)((int)addr & ~PGOFSET); 385 kmem_free_wakeup(phys_map, kva, ctob(npf)); 386 bp->b_un.b_addr = bp->b_saveaddr; 387 bp->b_saveaddr = NULL; 388 } 389