subr_syscall.c revision 9545
1/*- 2 * Copyright (C) 1994, David Greenman 3 * Copyright (c) 1990, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * the University of Utah, and William Jolitz. 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 by the University of 20 * California, Berkeley and its contributors. 21 * 4. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 38 * $Id: trap.c,v 1.55 1995/07/16 05:39:22 davidg Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/proc.h> 48#include <sys/user.h> 49#include <sys/acct.h> 50#include <sys/kernel.h> 51#include <sys/syscall.h> 52#include <sys/sysent.h> 53#ifdef KTRACE 54#include <sys/ktrace.h> 55#endif 56 57#include <vm/vm_param.h> 58#include <vm/pmap.h> 59#include <vm/vm_kern.h> 60#include <vm/vm_map.h> 61#include <vm/vm_page.h> 62 63#include <machine/cpu.h> 64#include <machine/md_var.h> 65#include <machine/psl.h> 66#include <machine/reg.h> 67#include <machine/trap.h> 68#include <machine/../isa/isa_device.h> 69 70#ifdef POWERFAIL_NMI 71# include <syslog.h> 72# include <machine/clock.h> 73#endif 74 75#include "isa.h" 76#include "npx.h" 77 78int trap_pfault __P((struct trapframe *, int)); 79void trap_fatal __P((struct trapframe *)); 80 81#define MAX_TRAP_MSG 27 82char *trap_msg[] = { 83 "", /* 0 unused */ 84 "privileged instruction fault", /* 1 T_PRIVINFLT */ 85 "", /* 2 unused */ 86 "breakpoint instruction fault", /* 3 T_BPTFLT */ 87 "", /* 4 unused */ 88 "", /* 5 unused */ 89 "arithmetic trap", /* 6 T_ARITHTRAP */ 90 "system forced exception", /* 7 T_ASTFLT */ 91 "", /* 8 unused */ 92 "general protection fault", /* 9 T_PROTFLT */ 93 "trace trap", /* 10 T_TRCTRAP */ 94 "", /* 11 unused */ 95 "page fault", /* 12 T_PAGEFLT */ 96 "", /* 13 unused */ 97 "alignment fault", /* 14 T_ALIGNFLT */ 98 "", /* 15 unused */ 99 "", /* 16 unused */ 100 "", /* 17 unused */ 101 "integer divide fault", /* 18 T_DIVIDE */ 102 "non-maskable interrupt trap", /* 19 T_NMI */ 103 "overflow trap", /* 20 T_OFLOW */ 104 "FPU bounds check fault", /* 21 T_BOUND */ 105 "FPU device not available", /* 22 T_DNA */ 106 "double fault", /* 23 T_DOUBLEFLT */ 107 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 108 "invalid TSS fault", /* 25 T_TSSFLT */ 109 "segment not present fault", /* 26 T_SEGNPFLT */ 110 "stack fault", /* 27 T_STKFLT */ 111}; 112 113static inline void 114userret(p, frame, oticks) 115 struct proc *p; 116 struct trapframe *frame; 117 u_quad_t oticks; 118{ 119 int sig, s; 120 121 while ((sig = CURSIG(p)) != 0) 122 postsig(sig); 123 p->p_priority = p->p_usrpri; 124 if (want_resched) { 125 /* 126 * Since we are curproc, clock will normally just change 127 * our priority without moving us from one queue to another 128 * (since the running process is not on a queue.) 129 * If that happened after we setrunqueue ourselves but before we 130 * mi_switch()'ed, we might not be on the queue indicated by 131 * our priority. 132 */ 133 s = splclock(); 134 setrunqueue(p); 135 p->p_stats->p_ru.ru_nivcsw++; 136 mi_switch(); 137 splx(s); 138 while ((sig = CURSIG(p)) != 0) 139 postsig(sig); 140 } 141 /* 142 * Charge system time if profiling. 143 */ 144 if (p->p_flag & P_PROFIL) { 145 u_quad_t ticks = p->p_sticks - oticks; 146 147 if (ticks) { 148#ifdef PROFTIMER 149 extern int profscale; 150 addupc(frame->tf_eip, &p->p_stats->p_prof, 151 ticks * profscale); 152#else 153 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 154#endif 155 } 156 } 157 curpriority = p->p_priority; 158} 159 160/* 161 * trap(frame): 162 * Exception, fault, and trap interface to the FreeBSD kernel. 163 * This common code is called from assembly language IDT gate entry 164 * routines that prepare a suitable stack frame, and restore this 165 * frame after the exception has been processed. 166 */ 167 168/*ARGSUSED*/ 169void 170trap(frame) 171 struct trapframe frame; 172{ 173 struct proc *p = curproc; 174 u_quad_t sticks = 0; 175 int i = 0, ucode = 0, type, code; 176#ifdef DIAGNOSTIC 177 u_long eva; 178#endif 179 180 type = frame.tf_trapno; 181 code = frame.tf_err; 182 183 if (ISPL(frame.tf_cs) == SEL_UPL) { 184 /* user trap */ 185 186 sticks = p->p_sticks; 187 p->p_md.md_regs = (int *)&frame; 188 189 switch (type) { 190 case T_PRIVINFLT: /* privileged instruction fault */ 191 ucode = type; 192 i = SIGILL; 193 break; 194 195 case T_BPTFLT: /* bpt instruction fault */ 196 case T_TRCTRAP: /* trace trap */ 197 frame.tf_eflags &= ~PSL_T; 198 i = SIGTRAP; 199 break; 200 201 case T_ARITHTRAP: /* arithmetic trap */ 202 ucode = code; 203 i = SIGFPE; 204 break; 205 206 case T_ASTFLT: /* Allow process switch */ 207 astoff(); 208 cnt.v_soft++; 209 if (p->p_flag & P_OWEUPC) { 210 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 211 p->p_flag &= ~P_OWEUPC; 212 } 213 goto out; 214 215 case T_PROTFLT: /* general protection fault */ 216 case T_SEGNPFLT: /* segment not present fault */ 217 case T_STKFLT: /* stack fault */ 218 case T_TSSFLT: /* invalid TSS fault */ 219 case T_DOUBLEFLT: /* double fault */ 220 default: 221 ucode = code + BUS_SEGM_FAULT ; 222 i = SIGBUS; 223 break; 224 225 case T_PAGEFLT: /* page fault */ 226 i = trap_pfault(&frame, TRUE); 227 if (i == -1) 228 return; 229 if (i == 0) 230 goto out; 231 232 ucode = T_PAGEFLT; 233 break; 234 235 case T_DIVIDE: /* integer divide fault */ 236 ucode = FPE_INTDIV_TRAP; 237 i = SIGFPE; 238 break; 239 240#if NISA > 0 241 case T_NMI: 242#ifdef POWERFAIL_NMI 243 goto handle_powerfail; 244#else /* !POWERFAIL_NMI */ 245#ifdef DDB 246 /* NMI can be hooked up to a pushbutton for debugging */ 247 printf ("NMI ... going to debugger\n"); 248 if (kdb_trap (type, 0, &frame)) 249 return; 250#endif /* DDB */ 251 /* machine/parity/power fail/"kitchen sink" faults */ 252 if (isa_nmi(code) == 0) return; 253 panic("NMI indicates hardware failure"); 254#endif /* POWERFAIL_NMI */ 255#endif /* NISA > 0 */ 256 257 case T_OFLOW: /* integer overflow fault */ 258 ucode = FPE_INTOVF_TRAP; 259 i = SIGFPE; 260 break; 261 262 case T_BOUND: /* bounds check fault */ 263 ucode = FPE_SUBRNG_TRAP; 264 i = SIGFPE; 265 break; 266 267 case T_DNA: 268#if NNPX > 0 269 /* if a transparent fault (due to context switch "late") */ 270 if (npxdna()) 271 return; 272#endif /* NNPX > 0 */ 273 274#if defined(MATH_EMULATE) || defined(GPL_MATH_EMULATE) 275 i = math_emulate(&frame); 276 if (i == 0) { 277 if (!(frame.tf_eflags & PSL_T)) 278 return; 279 frame.tf_eflags &= ~PSL_T; 280 i = SIGTRAP; 281 } 282 /* else ucode = emulator_only_knows() XXX */ 283#else /* MATH_EMULATE || GPL_MATH_EMULATE */ 284 i = SIGFPE; 285 ucode = FPE_FPU_NP_TRAP; 286#endif /* MATH_EMULATE || GPL_MATH_EMULATE */ 287 break; 288 289 case T_FPOPFLT: /* FPU operand fetch fault */ 290 ucode = T_FPOPFLT; 291 i = SIGILL; 292 break; 293 } 294 } else { 295 /* kernel trap */ 296 297 switch (type) { 298 case T_PAGEFLT: /* page fault */ 299 (void) trap_pfault(&frame, FALSE); 300 return; 301 302 case T_PROTFLT: /* general protection fault */ 303 case T_SEGNPFLT: /* segment not present fault */ 304 /* 305 * Invalid segment selectors and out of bounds 306 * %eip's and %esp's can be set up in user mode. 307 * This causes a fault in kernel mode when the 308 * kernel tries to return to user mode. We want 309 * to get this fault so that we can fix the 310 * problem here and not have to check all the 311 * selectors and pointers when the user changes 312 * them. 313 */ 314#define MAYBE_DORETI_FAULT(where, whereto) \ 315 do { \ 316 if (frame.tf_eip == (int)where) { \ 317 frame.tf_eip = (int)whereto; \ 318 return; \ 319 } \ 320 } while (0) 321 322 if (intr_nesting_level == 0) { 323 MAYBE_DORETI_FAULT(doreti_iret, 324 doreti_iret_fault); 325 MAYBE_DORETI_FAULT(doreti_popl_ds, 326 doreti_popl_ds_fault); 327 MAYBE_DORETI_FAULT(doreti_popl_es, 328 doreti_popl_es_fault); 329 } 330 if (curpcb && curpcb->pcb_onfault) { 331 frame.tf_eip = (int)curpcb->pcb_onfault; 332 return; 333 } 334 break; 335 336 case T_TSSFLT: 337 /* 338 * PSL_NT can be set in user mode and isn't cleared 339 * automatically when the kernel is entered. This 340 * causes a TSS fault when the kernel attempts to 341 * `iret' because the TSS link is uninitialized. We 342 * want to get this fault so that we can fix the 343 * problem here and not every time the kernel is 344 * entered. 345 */ 346 if (frame.tf_eflags & PSL_NT) { 347 frame.tf_eflags &= ~PSL_NT; 348 return; 349 } 350 break; 351 352#ifdef DDB 353 case T_BPTFLT: 354 case T_TRCTRAP: 355 if (kdb_trap (type, 0, &frame)) 356 return; 357 break; 358#else 359 case T_TRCTRAP: /* trace trap -- someone single stepping lcall's */ 360 /* Q: how do we turn it on again? */ 361 frame.tf_eflags &= ~PSL_T; 362 return; 363#endif 364 365#if NISA > 0 366 case T_NMI: 367#ifdef POWERFAIL_NMI 368#ifndef TIMER_FREQ 369# define TIMER_FREQ 1193182 370#endif 371 handle_powerfail: 372 { 373 static unsigned lastalert = 0; 374 375 if(time.tv_sec - lastalert > 10) 376 { 377 log(LOG_WARNING, "NMI: power fail\n"); 378 sysbeep(TIMER_FREQ/880, hz); 379 lastalert = time.tv_sec; 380 } 381 return; 382 } 383#else /* !POWERFAIL_NMI */ 384#ifdef DDB 385 /* NMI can be hooked up to a pushbutton for debugging */ 386 printf ("NMI ... going to debugger\n"); 387 if (kdb_trap (type, 0, &frame)) 388 return; 389#endif /* DDB */ 390 /* machine/parity/power fail/"kitchen sink" faults */ 391 if (isa_nmi(code) == 0) return; 392 /* FALL THROUGH */ 393#endif /* POWERFAIL_NMI */ 394#endif /* NISA > 0 */ 395 } 396 397 trap_fatal(&frame); 398 return; 399 } 400 401 trapsignal(p, i, ucode); 402 403#ifdef DEBUG 404 eva = rcr2(); 405 if (type <= MAX_TRAP_MSG) { 406 uprintf("fatal process exception: %s", 407 trap_msg[type]); 408 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 409 uprintf(", fault VA = 0x%x", eva); 410 uprintf("\n"); 411 } 412#endif 413 414out: 415 userret(p, &frame, sticks); 416} 417 418#ifdef notyet 419/* 420 * This version doesn't allow a page fault to user space while 421 * in the kernel. The rest of the kernel needs to be made "safe" 422 * before this can be used. I think the only things remaining 423 * to be made safe are the iBCS2 code and the process tracing/ 424 * debugging code. 425 */ 426int 427trap_pfault(frame, usermode) 428 struct trapframe *frame; 429 int usermode; 430{ 431 vm_offset_t va; 432 struct vmspace *vm = NULL; 433 vm_map_t map = 0; 434 int rv = 0; 435 vm_prot_t ftype; 436 int eva; 437 struct proc *p = curproc; 438 439 if (frame->tf_err & PGEX_W) 440 ftype = VM_PROT_READ | VM_PROT_WRITE; 441 else 442 ftype = VM_PROT_READ; 443 444 eva = rcr2(); 445 va = trunc_page((vm_offset_t)eva); 446 447 if (va < VM_MIN_KERNEL_ADDRESS) { 448 vm_offset_t v; 449 vm_page_t ptepg; 450 451 if ((p == NULL) || 452 (!usermode && va < VM_MAXUSER_ADDRESS && 453 curpcb->pcb_onfault == NULL)) { 454 trap_fatal(frame); 455 return (-1); 456 } 457 458 /* 459 * This is a fault on non-kernel virtual memory. 460 * vm is initialized above to NULL. If curproc is NULL 461 * or curproc->p_vmspace is NULL the fault is fatal. 462 */ 463 vm = p->p_vmspace; 464 if (vm == NULL) 465 goto nogo; 466 467 map = &vm->vm_map; 468 469 /* 470 * Keep swapout from messing with us during this 471 * critical time. 472 */ 473 ++p->p_lock; 474 475 /* 476 * Grow the stack if necessary 477 */ 478 if ((caddr_t)va > vm->vm_maxsaddr 479 && (caddr_t)va < (caddr_t)USRSTACK) { 480 if (!grow(p, va)) { 481 rv = KERN_FAILURE; 482 --p->p_lock; 483 goto nogo; 484 } 485 } 486 487 /* 488 * Check if page table is mapped, if not, 489 * fault it first 490 */ 491 v = (vm_offset_t) vtopte(va); 492 493 /* Fault the pte only if needed: */ 494 if (*((int *)vtopte(v)) == 0) 495 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 496 497 pmap_use_pt( vm_map_pmap(map), va); 498 499 /* Fault in the user page: */ 500 rv = vm_fault(map, va, ftype, FALSE); 501 502 pmap_unuse_pt( vm_map_pmap(map), va); 503 504 --p->p_lock; 505 } else { 506 /* 507 * Don't allow user-mode faults in kernel address space. 508 */ 509 if (usermode) 510 goto nogo; 511 512 /* 513 * Since we know that kernel virtual address addresses 514 * always have pte pages mapped, we just have to fault 515 * the page. 516 */ 517 rv = vm_fault(kernel_map, va, ftype, FALSE); 518 } 519 520 if (rv == KERN_SUCCESS) 521 return (0); 522nogo: 523 if (!usermode) { 524 if (curpcb && curpcb->pcb_onfault) { 525 frame->tf_eip = (int)curpcb->pcb_onfault; 526 return (0); 527 } 528 trap_fatal(frame); 529 return (-1); 530 } 531 532 /* kludge to pass faulting virtual address to sendsig */ 533 frame->tf_err = eva; 534 535 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 536} 537#endif 538 539int 540trap_pfault(frame, usermode) 541 struct trapframe *frame; 542 int usermode; 543{ 544 vm_offset_t va; 545 struct vmspace *vm = NULL; 546 vm_map_t map = 0; 547 int rv = 0; 548 vm_prot_t ftype; 549 int eva; 550 struct proc *p = curproc; 551 552 eva = rcr2(); 553 va = trunc_page((vm_offset_t)eva); 554 555 if (va >= KERNBASE) { 556 /* 557 * Don't allow user-mode faults in kernel address space. 558 */ 559 if (usermode) 560 goto nogo; 561 562 map = kernel_map; 563 } else { 564 /* 565 * This is a fault on non-kernel virtual memory. 566 * vm is initialized above to NULL. If curproc is NULL 567 * or curproc->p_vmspace is NULL the fault is fatal. 568 */ 569 if (p != NULL) 570 vm = p->p_vmspace; 571 572 if (vm == NULL) 573 goto nogo; 574 575 map = &vm->vm_map; 576 } 577 578 if (frame->tf_err & PGEX_W) 579 ftype = VM_PROT_READ | VM_PROT_WRITE; 580 else 581 ftype = VM_PROT_READ; 582 583 if (map != kernel_map) { 584 vm_offset_t v; 585 vm_page_t ptepg; 586 587 /* 588 * Keep swapout from messing with us during this 589 * critical time. 590 */ 591 ++p->p_lock; 592 593 /* 594 * Grow the stack if necessary 595 */ 596 if ((caddr_t)va > vm->vm_maxsaddr 597 && (caddr_t)va < (caddr_t)USRSTACK) { 598 if (!grow(p, va)) { 599 rv = KERN_FAILURE; 600 --p->p_lock; 601 goto nogo; 602 } 603 } 604 605 /* 606 * Check if page table is mapped, if not, 607 * fault it first 608 */ 609 v = (vm_offset_t) vtopte(va); 610 611 /* Fault the pte only if needed: */ 612 if (*((int *)vtopte(v)) == 0) 613 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 614 615 pmap_use_pt( vm_map_pmap(map), va); 616 617 /* Fault in the user page: */ 618 rv = vm_fault(map, va, ftype, FALSE); 619 620 pmap_unuse_pt( vm_map_pmap(map), va); 621 622 --p->p_lock; 623 } else { 624 /* 625 * Since we know that kernel virtual address addresses 626 * always have pte pages mapped, we just have to fault 627 * the page. 628 */ 629 rv = vm_fault(map, va, ftype, FALSE); 630 } 631 632 if (rv == KERN_SUCCESS) 633 return (0); 634nogo: 635 if (!usermode) { 636 if (curpcb && curpcb->pcb_onfault) { 637 frame->tf_eip = (int)curpcb->pcb_onfault; 638 return (0); 639 } 640 trap_fatal(frame); 641 return (-1); 642 } 643 644 /* kludge to pass faulting virtual address to sendsig */ 645 frame->tf_err = eva; 646 647 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 648} 649 650void 651trap_fatal(frame) 652 struct trapframe *frame; 653{ 654 int code, type, eva; 655 struct soft_segment_descriptor softseg; 656 657 code = frame->tf_err; 658 type = frame->tf_trapno; 659 eva = rcr2(); 660 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 661 662 if (type <= MAX_TRAP_MSG) 663 printf("\n\nFatal trap %d: %s while in %s mode\n", 664 type, trap_msg[type], 665 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 666 if (type == T_PAGEFLT) { 667 printf("fault virtual address = 0x%x\n", eva); 668 printf("fault code = %s %s, %s\n", 669 code & PGEX_U ? "user" : "supervisor", 670 code & PGEX_W ? "write" : "read", 671 code & PGEX_P ? "protection violation" : "page not present"); 672 } 673 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 674 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 675 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 676 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 677 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 678 printf("processor eflags = "); 679 if (frame->tf_eflags & PSL_T) 680 printf("trace/trap, "); 681 if (frame->tf_eflags & PSL_I) 682 printf("interrupt enabled, "); 683 if (frame->tf_eflags & PSL_NT) 684 printf("nested task, "); 685 if (frame->tf_eflags & PSL_RF) 686 printf("resume, "); 687 if (frame->tf_eflags & PSL_VM) 688 printf("vm86, "); 689 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 690 printf("current process = "); 691 if (curproc) { 692 printf("%lu (%s)\n", 693 (u_long)curproc->p_pid, curproc->p_comm ? 694 curproc->p_comm : ""); 695 } else { 696 printf("Idle\n"); 697 } 698 printf("interrupt mask = "); 699 if ((cpl & net_imask) == net_imask) 700 printf("net "); 701 if ((cpl & tty_imask) == tty_imask) 702 printf("tty "); 703 if ((cpl & bio_imask) == bio_imask) 704 printf("bio "); 705 if (cpl == 0) 706 printf("none"); 707 printf("\n"); 708 709#ifdef KDB 710 if (kdb_trap(&psl)) 711 return; 712#endif 713#ifdef DDB 714 if (kdb_trap (type, 0, frame)) 715 return; 716#endif 717 if (type <= MAX_TRAP_MSG) 718 panic(trap_msg[type]); 719 else 720 panic("unknown/reserved trap"); 721} 722 723/* 724 * Compensate for 386 brain damage (missing URKR). 725 * This is a little simpler than the pagefault handler in trap() because 726 * it the page tables have already been faulted in and high addresses 727 * are thrown out early for other reasons. 728 */ 729int trapwrite(addr) 730 unsigned addr; 731{ 732 struct proc *p; 733 vm_offset_t va, v; 734 struct vmspace *vm; 735 int rv; 736 737 va = trunc_page((vm_offset_t)addr); 738 /* 739 * XXX - MAX is END. Changed > to >= for temp. fix. 740 */ 741 if (va >= VM_MAXUSER_ADDRESS) 742 return (1); 743 744 p = curproc; 745 vm = p->p_vmspace; 746 747 ++p->p_lock; 748 749 if ((caddr_t)va >= vm->vm_maxsaddr 750 && (caddr_t)va < (caddr_t)USRSTACK) { 751 if (!grow(p, va)) { 752 --p->p_lock; 753 return (1); 754 } 755 } 756 757 v = trunc_page(vtopte(va)); 758 759 /* 760 * wire the pte page 761 */ 762 if (va < USRSTACK) { 763 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 764 } 765 766 /* 767 * fault the data page 768 */ 769 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 770 771 /* 772 * unwire the pte page 773 */ 774 if (va < USRSTACK) { 775 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 776 } 777 778 --p->p_lock; 779 780 if (rv != KERN_SUCCESS) 781 return 1; 782 783 return (0); 784} 785 786/* 787 * syscall(frame): 788 * System call request from POSIX system call gate interface to kernel. 789 * Like trap(), argument is call by reference. 790 */ 791/*ARGSUSED*/ 792void 793syscall(frame) 794 struct trapframe frame; 795{ 796 caddr_t params; 797 int i; 798 struct sysent *callp; 799 struct proc *p = curproc; 800 u_quad_t sticks; 801 int error, opc; 802 int args[8], rval[2]; 803 u_int code; 804 805 sticks = p->p_sticks; 806 if (ISPL(frame.tf_cs) != SEL_UPL) 807 panic("syscall"); 808 809 code = frame.tf_eax; 810 p->p_md.md_regs = (int *)&frame; 811 params = (caddr_t)frame.tf_esp + sizeof (int) ; 812 813 /* 814 * Reconstruct pc, assuming lcall $X,y is 7 bytes, as it is always. 815 */ 816 opc = frame.tf_eip - 7; 817 /* 818 * Need to check if this is a 32 bit or 64 bit syscall. 819 */ 820 if (code == SYS_syscall) { 821 /* 822 * Code is first argument, followed by actual args. 823 */ 824 code = fuword(params); 825 params += sizeof (int); 826 } else if (code == SYS___syscall) { 827 /* 828 * Like syscall, but code is a quad, so as to maintain 829 * quad alignment for the rest of the arguments. 830 */ 831 code = fuword(params + _QUAD_LOWWORD * sizeof(int)); 832 params += sizeof(quad_t); 833 } 834 835 if (p->p_sysent->sv_mask) 836 code = code & p->p_sysent->sv_mask; 837 838 if (code >= p->p_sysent->sv_size) 839 callp = &p->p_sysent->sv_table[0]; 840 else 841 callp = &p->p_sysent->sv_table[code]; 842 843 if ((i = callp->sy_narg * sizeof (int)) && 844 (error = copyin(params, (caddr_t)args, (u_int)i))) { 845#ifdef KTRACE 846 if (KTRPOINT(p, KTR_SYSCALL)) 847 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 848#endif 849 goto bad; 850 } 851#ifdef KTRACE 852 if (KTRPOINT(p, KTR_SYSCALL)) 853 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 854#endif 855 rval[0] = 0; 856 rval[1] = frame.tf_edx; 857 858 error = (*callp->sy_call)(p, args, rval); 859 860 switch (error) { 861 862 case 0: 863 /* 864 * Reinitialize proc pointer `p' as it may be different 865 * if this is a child returning from fork syscall. 866 */ 867 p = curproc; 868 frame.tf_eax = rval[0]; 869 frame.tf_edx = rval[1]; 870 frame.tf_eflags &= ~PSL_C; /* carry bit */ 871 break; 872 873 case ERESTART: 874 frame.tf_eip = opc; 875 break; 876 877 case EJUSTRETURN: 878 break; 879 880 default: 881 bad: 882 if (p->p_sysent->sv_errsize) 883 if (error >= p->p_sysent->sv_errsize) 884 error = -1; /* XXX */ 885 else 886 error = p->p_sysent->sv_errtbl[error]; 887 frame.tf_eax = error; 888 frame.tf_eflags |= PSL_C; /* carry bit */ 889 break; 890 } 891 892 userret(p, &frame, sticks); 893 894#ifdef KTRACE 895 if (KTRPOINT(p, KTR_SYSRET)) 896 ktrsysret(p->p_tracep, code, error, rval[0]); 897#endif 898} 899 900#ifdef COMPAT_LINUX 901/* 902 * linux_syscall(frame): 903 */ 904/*ARGSUSED*/ 905void 906linux_syscall(frame) 907 struct trapframe frame; 908{ 909 caddr_t params; 910 int i; 911 struct proc *p = curproc; 912 struct sysent *callp; 913 u_quad_t sticks; 914 int error, opc; 915 int rval[2]; 916 int code; 917 struct linux_syscall_args { 918 int arg1; 919 int arg2; 920 int arg3; 921 int arg4; 922 int arg5; 923 } args; 924 925 args.arg1 = frame.tf_ebx; 926 args.arg2 = frame.tf_ecx; 927 args.arg3 = frame.tf_edx; 928 args.arg4 = frame.tf_esi; 929 args.arg5 = frame.tf_edi; 930 931 sticks = p->p_sticks; 932 if (ISPL(frame.tf_cs) != SEL_UPL) 933 panic("linux syscall"); 934 935 code = frame.tf_eax; 936 p->p_md.md_regs = (int *)&frame; 937 params = (caddr_t)frame.tf_esp + sizeof (int) ; 938 939 /* Reconstruct pc, subtract size of int 0x80 */ 940 opc = frame.tf_eip - 2; 941 if (code == 0) { 942 code = fuword(params); 943 params += sizeof (int); 944 } 945 if (p->p_sysent->sv_mask) 946 code = code & p->p_sysent->sv_mask; 947 948 if (code < 0 || code >= p->p_sysent->sv_size) 949 callp = &p->p_sysent->sv_table[0]; 950 else 951 callp = &p->p_sysent->sv_table[code]; 952 953#ifdef KTRACE 954 if (KTRPOINT(p, KTR_SYSCALL)) 955 ktrsyscall(p->p_tracep, code, callp->sy_narg, &args); 956#endif 957 958#ifdef KTRACE 959 if (KTRPOINT(p, KTR_SYSCALL)) 960 ktrsyscall(p->p_tracep, code, callp->sy_narg, &args); 961#endif 962 rval[0] = 0; 963 rval[1] = frame.tf_edx; 964 965 error = (*callp->sy_call)(p, &args, rval); 966 967 switch (error) { 968 969 case 0: 970 /* 971 * Reinitialize proc pointer `p' as it may be different 972 * if this is a child returning from fork syscall. 973 */ 974 p = curproc; 975 frame.tf_eax = rval[0]; 976 frame.tf_eflags &= ~PSL_C; /* carry bit */ 977 break; 978 979 case ERESTART: 980 frame.tf_eip = opc; 981 break; 982 983 case EJUSTRETURN: 984 break; 985 986 default: 987 bad: 988 if (p->p_sysent->sv_errsize) 989 if (error >= p->p_sysent->sv_errsize) 990 error = -1; /* XXX */ 991 else 992 error = p->p_sysent->sv_errtbl[error]; 993 frame.tf_eax = -error; 994 frame.tf_eflags |= PSL_C; /* carry bit */ 995 break; 996 } 997 998 userret(p, &frame, sticks); 999 1000#ifdef KTRACE 1001 if (KTRPOINT(p, KTR_SYSRET)) 1002 ktrsysret(p->p_tracep, code, error, rval[0]); 1003#endif 1004} 1005#endif /* COMPAT_LINUX */ 1006