subr_syscall.c revision 33108
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 1.120 1998/01/31 05:00:15 eivind Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include "opt_cpu.h" 46#include "opt_ddb.h" 47#include "opt_diagnostic.h" 48#include "opt_ktrace.h" 49#include "opt_trap.h" 50#include "opt_vm86.h" 51 52#include <sys/param.h> 53#include <sys/systm.h> 54#include <sys/proc.h> 55#include <sys/pioctl.h> 56#include <sys/kernel.h> 57#include <sys/resourcevar.h> 58#include <sys/signalvar.h> 59#include <sys/syscall.h> 60#include <sys/sysent.h> 61#include <sys/vmmeter.h> 62#ifdef KTRACE 63#include <sys/ktrace.h> 64#endif 65 66#include <vm/vm.h> 67#include <vm/vm_param.h> 68#include <vm/vm_prot.h> 69#include <sys/lock.h> 70#include <vm/pmap.h> 71#include <vm/vm_kern.h> 72#include <vm/vm_map.h> 73#include <vm/vm_page.h> 74#include <vm/vm_extern.h> 75 76#include <machine/cpu.h> 77#include <machine/ipl.h> 78#include <machine/md_var.h> 79#include <machine/pcb.h> 80#ifdef SMP 81#include <machine/smp.h> 82#endif 83#include <machine/tss.h> 84 85#include <i386/isa/intr_machdep.h> 86 87#ifdef POWERFAIL_NMI 88#include <sys/syslog.h> 89#include <machine/clock.h> 90#endif 91 92#ifdef VM86 93#include <machine/vm86.h> 94#endif 95 96#include "isa.h" 97#include "npx.h" 98 99extern struct i386tss common_tss; 100 101int (*pmath_emulate) __P((struct trapframe *)); 102 103extern void trap __P((struct trapframe frame)); 104extern int trapwrite __P((unsigned addr)); 105extern void syscall __P((struct trapframe frame)); 106 107static int trap_pfault __P((struct trapframe *, int)); 108static void trap_fatal __P((struct trapframe *)); 109void dblfault_handler __P((void)); 110 111extern inthand_t IDTVEC(syscall); 112 113#define MAX_TRAP_MSG 28 114static char *trap_msg[] = { 115 "", /* 0 unused */ 116 "privileged instruction fault", /* 1 T_PRIVINFLT */ 117 "", /* 2 unused */ 118 "breakpoint instruction fault", /* 3 T_BPTFLT */ 119 "", /* 4 unused */ 120 "", /* 5 unused */ 121 "arithmetic trap", /* 6 T_ARITHTRAP */ 122 "system forced exception", /* 7 T_ASTFLT */ 123 "", /* 8 unused */ 124 "general protection fault", /* 9 T_PROTFLT */ 125 "trace trap", /* 10 T_TRCTRAP */ 126 "", /* 11 unused */ 127 "page fault", /* 12 T_PAGEFLT */ 128 "", /* 13 unused */ 129 "alignment fault", /* 14 T_ALIGNFLT */ 130 "", /* 15 unused */ 131 "", /* 16 unused */ 132 "", /* 17 unused */ 133 "integer divide fault", /* 18 T_DIVIDE */ 134 "non-maskable interrupt trap", /* 19 T_NMI */ 135 "overflow trap", /* 20 T_OFLOW */ 136 "FPU bounds check fault", /* 21 T_BOUND */ 137 "FPU device not available", /* 22 T_DNA */ 138 "double fault", /* 23 T_DOUBLEFLT */ 139 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 140 "invalid TSS fault", /* 25 T_TSSFLT */ 141 "segment not present fault", /* 26 T_SEGNPFLT */ 142 "stack fault", /* 27 T_STKFLT */ 143 "machine check trap", /* 28 T_MCHK */ 144}; 145 146static void userret __P((struct proc *p, struct trapframe *frame, 147 u_quad_t oticks)); 148 149#if defined(I586_CPU) && !defined(NO_F00F_HACK) 150extern struct gate_descriptor *t_idt; 151extern int has_f00f_bug; 152#endif 153 154static inline void 155userret(p, frame, oticks) 156 struct proc *p; 157 struct trapframe *frame; 158 u_quad_t oticks; 159{ 160 int sig, s; 161 162 while ((sig = CURSIG(p)) != 0) 163 postsig(sig); 164 165#if 0 166 if (!want_resched && 167 (p->p_priority <= p->p_usrpri) && 168 (p->p_rtprio.type == RTP_PRIO_NORMAL)) { 169 int newpriority; 170 p->p_estcpu += 1; 171 newpriority = PUSER + p->p_estcpu / 4 + 2 * p->p_nice; 172 newpriority = min(newpriority, MAXPRI); 173 p->p_usrpri = newpriority; 174 } 175#endif 176 177 p->p_priority = p->p_usrpri; 178 if (want_resched) { 179 /* 180 * Since we are curproc, clock will normally just change 181 * our priority without moving us from one queue to another 182 * (since the running process is not on a queue.) 183 * If that happened after we setrunqueue ourselves but before we 184 * mi_switch()'ed, we might not be on the queue indicated by 185 * our priority. 186 */ 187 s = splhigh(); 188 setrunqueue(p); 189 p->p_stats->p_ru.ru_nivcsw++; 190 mi_switch(); 191 splx(s); 192 while ((sig = CURSIG(p)) != 0) 193 postsig(sig); 194 } 195 /* 196 * Charge system time if profiling. 197 */ 198 if (p->p_flag & P_PROFIL) 199 addupc_task(p, frame->tf_eip, 200 (u_int)(p->p_sticks - oticks) * psratio); 201 202 curpriority = p->p_priority; 203} 204 205/* 206 * Exception, fault, and trap interface to the FreeBSD kernel. 207 * This common code is called from assembly language IDT gate entry 208 * routines that prepare a suitable stack frame, and restore this 209 * frame after the exception has been processed. 210 */ 211 212void 213trap(frame) 214 struct trapframe frame; 215{ 216 struct proc *p = curproc; 217 u_quad_t sticks = 0; 218 int i = 0, ucode = 0, type, code; 219#ifdef DEBUG 220 u_long eva; 221#endif 222 223#if defined(I586_CPU) && !defined(NO_F00F_HACK) 224restart: 225#endif 226 type = frame.tf_trapno; 227 code = frame.tf_err; 228 229 if ((ISPL(frame.tf_cs) == SEL_UPL) || (frame.tf_eflags & PSL_VM)) { 230 /* user trap */ 231 232 sticks = p->p_sticks; 233 p->p_md.md_regs = &frame; 234 235 switch (type) { 236 case T_PRIVINFLT: /* privileged instruction fault */ 237 ucode = type; 238 i = SIGILL; 239 break; 240 241 case T_BPTFLT: /* bpt instruction fault */ 242 case T_TRCTRAP: /* trace trap */ 243 frame.tf_eflags &= ~PSL_T; 244 i = SIGTRAP; 245 break; 246 247 case T_ARITHTRAP: /* arithmetic trap */ 248 ucode = code; 249 i = SIGFPE; 250 break; 251 252 case T_ASTFLT: /* Allow process switch */ 253 astoff(); 254 cnt.v_soft++; 255 if (p->p_flag & P_OWEUPC) { 256 p->p_flag &= ~P_OWEUPC; 257 addupc_task(p, p->p_stats->p_prof.pr_addr, 258 p->p_stats->p_prof.pr_ticks); 259 } 260 goto out; 261 262 /* 263 * The following two traps can happen in 264 * vm86 mode, and, if so, we want to handle 265 * them specially. 266 */ 267 case T_PROTFLT: /* general protection fault */ 268 case T_STKFLT: /* stack fault */ 269#ifdef VM86 270 if (frame.tf_eflags & PSL_VM) { 271 i = vm86_emulate((struct vm86frame *)&frame); 272 if (i == 0) 273 goto out; 274 break; 275 } 276#endif /* VM86 */ 277 /* FALL THROUGH */ 278 279 case T_SEGNPFLT: /* segment not present fault */ 280 case T_TSSFLT: /* invalid TSS fault */ 281 case T_DOUBLEFLT: /* double fault */ 282 default: 283 ucode = code + BUS_SEGM_FAULT ; 284 i = SIGBUS; 285 break; 286 287 case T_PAGEFLT: /* page fault */ 288 i = trap_pfault(&frame, TRUE); 289 if (i == -1) 290 return; 291#if defined(I586_CPU) && !defined(NO_F00F_HACK) 292 if (i == -2) 293 goto restart; 294#endif 295 if (i == 0) 296 goto out; 297 298 ucode = T_PAGEFLT; 299 break; 300 301 case T_DIVIDE: /* integer divide fault */ 302 ucode = FPE_INTDIV_TRAP; 303 i = SIGFPE; 304 break; 305 306#if NISA > 0 307 case T_NMI: 308#ifdef POWERFAIL_NMI 309 goto handle_powerfail; 310#else /* !POWERFAIL_NMI */ 311#ifdef DDB 312 /* NMI can be hooked up to a pushbutton for debugging */ 313 printf ("NMI ... going to debugger\n"); 314 if (kdb_trap (type, 0, &frame)) 315 return; 316#endif /* DDB */ 317 /* machine/parity/power fail/"kitchen sink" faults */ 318 if (isa_nmi(code) == 0) return; 319 panic("NMI indicates hardware failure"); 320#endif /* POWERFAIL_NMI */ 321#endif /* NISA > 0 */ 322 323 case T_OFLOW: /* integer overflow fault */ 324 ucode = FPE_INTOVF_TRAP; 325 i = SIGFPE; 326 break; 327 328 case T_BOUND: /* bounds check fault */ 329 ucode = FPE_SUBRNG_TRAP; 330 i = SIGFPE; 331 break; 332 333 case T_DNA: 334#if NNPX > 0 335 /* if a transparent fault (due to context switch "late") */ 336 if (npxdna()) 337 return; 338#endif 339 if (!pmath_emulate) { 340 i = SIGFPE; 341 ucode = FPE_FPU_NP_TRAP; 342 break; 343 } 344 i = (*pmath_emulate)(&frame); 345 if (i == 0) { 346 if (!(frame.tf_eflags & PSL_T)) 347 return; 348 frame.tf_eflags &= ~PSL_T; 349 i = SIGTRAP; 350 } 351 /* else ucode = emulator_only_knows() XXX */ 352 break; 353 354 case T_FPOPFLT: /* FPU operand fetch fault */ 355 ucode = T_FPOPFLT; 356 i = SIGILL; 357 break; 358 } 359 } else { 360 /* kernel trap */ 361 362 switch (type) { 363 case T_PAGEFLT: /* page fault */ 364 (void) trap_pfault(&frame, FALSE); 365 return; 366 367 case T_DNA: 368#if NNPX > 0 369 /* 370 * The kernel is apparently using npx for copying. 371 * XXX this should be fatal unless the kernel has 372 * registered such use. 373 */ 374 if (npxdna()) 375 return; 376#endif 377 break; 378 379 case T_PROTFLT: /* general protection fault */ 380 case T_SEGNPFLT: /* segment not present fault */ 381 /* 382 * Invalid segment selectors and out of bounds 383 * %eip's and %esp's can be set up in user mode. 384 * This causes a fault in kernel mode when the 385 * kernel tries to return to user mode. We want 386 * to get this fault so that we can fix the 387 * problem here and not have to check all the 388 * selectors and pointers when the user changes 389 * them. 390 */ 391#define MAYBE_DORETI_FAULT(where, whereto) \ 392 do { \ 393 if (frame.tf_eip == (int)where) { \ 394 frame.tf_eip = (int)whereto; \ 395 return; \ 396 } \ 397 } while (0) 398 399 if (intr_nesting_level == 0) { 400 /* 401 * Invalid %fs's and %gs's can be created using 402 * procfs or PT_SETREGS or by invalidating the 403 * underlying LDT entry. This causes a fault 404 * in kernel mode when the kernel attempts to 405 * switch contexts. Lose the bad context 406 * (XXX) so that we can continue, and generate 407 * a signal. 408 */ 409 if (frame.tf_eip == (int)cpu_switch_load_fs) { 410 curpcb->pcb_fs = 0; 411 psignal(p, SIGBUS); 412 return; 413 } 414 if (frame.tf_eip == (int)cpu_switch_load_gs) { 415 curpcb->pcb_gs = 0; 416 psignal(p, SIGBUS); 417 return; 418 } 419 MAYBE_DORETI_FAULT(doreti_iret, 420 doreti_iret_fault); 421 MAYBE_DORETI_FAULT(doreti_popl_ds, 422 doreti_popl_ds_fault); 423 MAYBE_DORETI_FAULT(doreti_popl_es, 424 doreti_popl_es_fault); 425 if (curpcb && curpcb->pcb_onfault) { 426 frame.tf_eip = (int)curpcb->pcb_onfault; 427 return; 428 } 429 } 430 break; 431 432 case T_TSSFLT: 433 /* 434 * PSL_NT can be set in user mode and isn't cleared 435 * automatically when the kernel is entered. This 436 * causes a TSS fault when the kernel attempts to 437 * `iret' because the TSS link is uninitialized. We 438 * want to get this fault so that we can fix the 439 * problem here and not every time the kernel is 440 * entered. 441 */ 442 if (frame.tf_eflags & PSL_NT) { 443 frame.tf_eflags &= ~PSL_NT; 444 return; 445 } 446 break; 447 448 case T_TRCTRAP: /* trace trap */ 449 if (frame.tf_eip == (int)IDTVEC(syscall)) { 450 /* 451 * We've just entered system mode via the 452 * syscall lcall. Continue single stepping 453 * silently until the syscall handler has 454 * saved the flags. 455 */ 456 return; 457 } 458 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 459 /* 460 * The syscall handler has now saved the 461 * flags. Stop single stepping it. 462 */ 463 frame.tf_eflags &= ~PSL_T; 464 return; 465 } 466 /* 467 * Fall through. 468 */ 469 case T_BPTFLT: 470 /* 471 * If DDB is enabled, let it handle the debugger trap. 472 * Otherwise, debugger traps "can't happen". 473 */ 474#ifdef DDB 475 if (kdb_trap (type, 0, &frame)) 476 return; 477#endif 478 break; 479 480#if NISA > 0 481 case T_NMI: 482#ifdef POWERFAIL_NMI 483#ifndef TIMER_FREQ 484# define TIMER_FREQ 1193182 485#endif 486 handle_powerfail: 487 { 488 static unsigned lastalert = 0; 489 490 if(time.tv_sec - lastalert > 10) 491 { 492 log(LOG_WARNING, "NMI: power fail\n"); 493 sysbeep(TIMER_FREQ/880, hz); 494 lastalert = time.tv_sec; 495 } 496 return; 497 } 498#else /* !POWERFAIL_NMI */ 499#ifdef DDB 500 /* NMI can be hooked up to a pushbutton for debugging */ 501 printf ("NMI ... going to debugger\n"); 502 if (kdb_trap (type, 0, &frame)) 503 return; 504#endif /* DDB */ 505 /* machine/parity/power fail/"kitchen sink" faults */ 506 if (isa_nmi(code) == 0) return; 507 /* FALL THROUGH */ 508#endif /* POWERFAIL_NMI */ 509#endif /* NISA > 0 */ 510 } 511 512 trap_fatal(&frame); 513 return; 514 } 515 516 trapsignal(p, i, ucode); 517 518#ifdef DEBUG 519 eva = rcr2(); 520 if (type <= MAX_TRAP_MSG) { 521 uprintf("fatal process exception: %s", 522 trap_msg[type]); 523 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 524 uprintf(", fault VA = 0x%x", eva); 525 uprintf("\n"); 526 } 527#endif 528 529out: 530 userret(p, &frame, sticks); 531} 532 533#ifdef notyet 534/* 535 * This version doesn't allow a page fault to user space while 536 * in the kernel. The rest of the kernel needs to be made "safe" 537 * before this can be used. I think the only things remaining 538 * to be made safe are the iBCS2 code and the process tracing/ 539 * debugging code. 540 */ 541static int 542trap_pfault(frame, usermode) 543 struct trapframe *frame; 544 int usermode; 545{ 546 vm_offset_t va; 547 struct vmspace *vm = NULL; 548 vm_map_t map = 0; 549 int rv = 0; 550 vm_prot_t ftype; 551 int eva; 552 struct proc *p = curproc; 553 554 if (frame->tf_err & PGEX_W) 555 ftype = VM_PROT_READ | VM_PROT_WRITE; 556 else 557 ftype = VM_PROT_READ; 558 559 eva = rcr2(); 560 va = trunc_page((vm_offset_t)eva); 561 562 if (va < VM_MIN_KERNEL_ADDRESS) { 563 vm_offset_t v; 564 vm_page_t mpte; 565 566 if (p == NULL || 567 (!usermode && va < VM_MAXUSER_ADDRESS && 568 (intr_nesting_level != 0 || curpcb == NULL || 569 curpcb->pcb_onfault == NULL))) { 570 trap_fatal(frame); 571 return (-1); 572 } 573 574 /* 575 * This is a fault on non-kernel virtual memory. 576 * vm is initialized above to NULL. If curproc is NULL 577 * or curproc->p_vmspace is NULL the fault is fatal. 578 */ 579 vm = p->p_vmspace; 580 if (vm == NULL) 581 goto nogo; 582 583 map = &vm->vm_map; 584 585 /* 586 * Keep swapout from messing with us during this 587 * critical time. 588 */ 589 ++p->p_lock; 590 591 /* 592 * Grow the stack if necessary 593 */ 594 if ((caddr_t)va > vm->vm_maxsaddr 595 && (caddr_t)va < (caddr_t)USRSTACK) { 596 if (!grow(p, va)) { 597 rv = KERN_FAILURE; 598 --p->p_lock; 599 goto nogo; 600 } 601 } 602 603 /* Fault in the user page: */ 604 rv = vm_fault(map, va, ftype, 605 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 606 607 --p->p_lock; 608 } else { 609 /* 610 * Don't allow user-mode faults in kernel address space. 611 */ 612 if (usermode) 613 goto nogo; 614 615 /* 616 * Since we know that kernel virtual address addresses 617 * always have pte pages mapped, we just have to fault 618 * the page. 619 */ 620 rv = vm_fault(kernel_map, va, ftype, FALSE); 621 } 622 623 if (rv == KERN_SUCCESS) 624 return (0); 625nogo: 626 if (!usermode) { 627 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 628 frame->tf_eip = (int)curpcb->pcb_onfault; 629 return (0); 630 } 631 trap_fatal(frame); 632 return (-1); 633 } 634 635 /* kludge to pass faulting virtual address to sendsig */ 636 frame->tf_err = eva; 637 638 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 639} 640#endif 641 642int 643trap_pfault(frame, usermode) 644 struct trapframe *frame; 645 int usermode; 646{ 647 vm_offset_t va; 648 struct vmspace *vm = NULL; 649 vm_map_t map = 0; 650 int rv = 0; 651 vm_prot_t ftype; 652 int eva; 653 struct proc *p = curproc; 654 655 eva = rcr2(); 656 va = trunc_page((vm_offset_t)eva); 657 658 if (va >= KERNBASE) { 659 /* 660 * Don't allow user-mode faults in kernel address space. 661 * An exception: if the faulting address is the invalid 662 * instruction entry in the IDT, then the Intel Pentium 663 * F00F bug workaround was triggered, and we need to 664 * treat it is as an illegal instruction, and not a page 665 * fault. 666 */ 667#if defined(I586_CPU) && !defined(NO_F00F_HACK) 668 if ((eva == (unsigned int)&t_idt[6]) && has_f00f_bug) { 669 frame->tf_trapno = T_PRIVINFLT; 670 return -2; 671 } 672#endif 673 if (usermode) 674 goto nogo; 675 676 map = kernel_map; 677 } else { 678 /* 679 * This is a fault on non-kernel virtual memory. 680 * vm is initialized above to NULL. If curproc is NULL 681 * or curproc->p_vmspace is NULL the fault is fatal. 682 */ 683 if (p != NULL) 684 vm = p->p_vmspace; 685 686 if (vm == NULL) 687 goto nogo; 688 689 map = &vm->vm_map; 690 } 691 692 if (frame->tf_err & PGEX_W) 693 ftype = VM_PROT_READ | VM_PROT_WRITE; 694 else 695 ftype = VM_PROT_READ; 696 697 if (map != kernel_map) { 698 /* 699 * Keep swapout from messing with us during this 700 * critical time. 701 */ 702 ++p->p_lock; 703 704 /* 705 * Grow the stack if necessary 706 */ 707 if ((caddr_t)va > vm->vm_maxsaddr 708 && (caddr_t)va < (caddr_t)USRSTACK) { 709 if (!grow(p, va)) { 710 rv = KERN_FAILURE; 711 --p->p_lock; 712 goto nogo; 713 } 714 } 715 716 /* Fault in the user page: */ 717 rv = vm_fault(map, va, ftype, 718 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 719 720 --p->p_lock; 721 } else { 722 /* 723 * Don't have to worry about process locking or stacks in the kernel. 724 */ 725 rv = vm_fault(map, va, ftype, FALSE); 726 } 727 728 if (rv == KERN_SUCCESS) 729 return (0); 730nogo: 731 if (!usermode) { 732 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 733 frame->tf_eip = (int)curpcb->pcb_onfault; 734 return (0); 735 } 736 trap_fatal(frame); 737 return (-1); 738 } 739 740 /* kludge to pass faulting virtual address to sendsig */ 741 frame->tf_err = eva; 742 743 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 744} 745 746static void 747trap_fatal(frame) 748 struct trapframe *frame; 749{ 750 int code, type, eva, ss, esp; 751 struct soft_segment_descriptor softseg; 752 753 code = frame->tf_err; 754 type = frame->tf_trapno; 755 eva = rcr2(); 756 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 757 758 if (type <= MAX_TRAP_MSG) 759 printf("\n\nFatal trap %d: %s while in %s mode\n", 760 type, trap_msg[type], 761 frame->tf_eflags & PSL_VM ? "vm86" : 762 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 763#ifdef SMP 764 /* three seperate prints in case of a trap on an unmapped page */ 765 printf("mp_lock = %08x; ", mp_lock); 766 printf("cpuid = %d; ", cpuid); 767 printf("lapic.id = %08x\n", lapic.id); 768#endif 769 if (type == T_PAGEFLT) { 770 printf("fault virtual address = 0x%x\n", eva); 771 printf("fault code = %s %s, %s\n", 772 code & PGEX_U ? "user" : "supervisor", 773 code & PGEX_W ? "write" : "read", 774 code & PGEX_P ? "protection violation" : "page not present"); 775 } 776 printf("instruction pointer = 0x%x:0x%x\n", 777 frame->tf_cs & 0xffff, frame->tf_eip); 778 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) { 779 ss = frame->tf_ss & 0xffff; 780 esp = frame->tf_esp; 781 } else { 782 ss = GSEL(GDATA_SEL, SEL_KPL); 783 esp = (int)&frame->tf_esp; 784 } 785 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 786 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 787 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 788 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 789 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 790 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 791 softseg.ssd_gran); 792 printf("processor eflags = "); 793 if (frame->tf_eflags & PSL_T) 794 printf("trace trap, "); 795 if (frame->tf_eflags & PSL_I) 796 printf("interrupt enabled, "); 797 if (frame->tf_eflags & PSL_NT) 798 printf("nested task, "); 799 if (frame->tf_eflags & PSL_RF) 800 printf("resume, "); 801 if (frame->tf_eflags & PSL_VM) 802 printf("vm86, "); 803 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 804 printf("current process = "); 805 if (curproc) { 806 printf("%lu (%s)\n", 807 (u_long)curproc->p_pid, curproc->p_comm ? 808 curproc->p_comm : ""); 809 } else { 810 printf("Idle\n"); 811 } 812 printf("interrupt mask = "); 813 if ((cpl & net_imask) == net_imask) 814 printf("net "); 815 if ((cpl & tty_imask) == tty_imask) 816 printf("tty "); 817 if ((cpl & bio_imask) == bio_imask) 818 printf("bio "); 819 if ((cpl & cam_imask) == cam_imask) 820 printf("cam "); 821 if (cpl == 0) 822 printf("none"); 823#ifdef SMP 824/** 825 * XXX FIXME: 826 * we probably SHOULD have stopped the other CPUs before now! 827 * another CPU COULD have been touching cpl at this moment... 828 */ 829 printf(" <- SMP: XXX"); 830#endif 831 printf("\n"); 832 833#ifdef KDB 834 if (kdb_trap(&psl)) 835 return; 836#endif 837#ifdef DDB 838 if (kdb_trap (type, 0, frame)) 839 return; 840#endif 841 printf("trap number = %d\n", type); 842 if (type <= MAX_TRAP_MSG) 843 panic(trap_msg[type]); 844 else 845 panic("unknown/reserved trap"); 846} 847 848/* 849 * Double fault handler. Called when a fault occurs while writing 850 * a frame for a trap/exception onto the stack. This usually occurs 851 * when the stack overflows (such is the case with infinite recursion, 852 * for example). 853 * 854 * XXX Note that the current PTD gets replaced by IdlePTD when the 855 * task switch occurs. This means that the stack that was active at 856 * the time of the double fault is not available at <kstack> unless 857 * the machine was idle when the double fault occurred. The downside 858 * of this is that "trace <ebp>" in ddb won't work. 859 */ 860void 861dblfault_handler() 862{ 863 printf("\nFatal double fault:\n"); 864 printf("eip = 0x%x\n", common_tss.tss_eip); 865 printf("esp = 0x%x\n", common_tss.tss_esp); 866 printf("ebp = 0x%x\n", common_tss.tss_ebp); 867#ifdef SMP 868 /* three seperate prints in case of a trap on an unmapped page */ 869 printf("mp_lock = %08x; ", mp_lock); 870 printf("cpuid = %d; ", cpuid); 871 printf("lapic.id = %08x\n", lapic.id); 872#endif 873 panic("double fault"); 874} 875 876/* 877 * Compensate for 386 brain damage (missing URKR). 878 * This is a little simpler than the pagefault handler in trap() because 879 * it the page tables have already been faulted in and high addresses 880 * are thrown out early for other reasons. 881 */ 882int trapwrite(addr) 883 unsigned addr; 884{ 885 struct proc *p; 886 vm_offset_t va; 887 struct vmspace *vm; 888 int rv; 889 890 va = trunc_page((vm_offset_t)addr); 891 /* 892 * XXX - MAX is END. Changed > to >= for temp. fix. 893 */ 894 if (va >= VM_MAXUSER_ADDRESS) 895 return (1); 896 897 p = curproc; 898 vm = p->p_vmspace; 899 900 ++p->p_lock; 901 902 if ((caddr_t)va >= vm->vm_maxsaddr 903 && (caddr_t)va < (caddr_t)USRSTACK) { 904 if (!grow(p, va)) { 905 --p->p_lock; 906 return (1); 907 } 908 } 909 910 /* 911 * fault the data page 912 */ 913 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, VM_FAULT_DIRTY); 914 915 --p->p_lock; 916 917 if (rv != KERN_SUCCESS) 918 return 1; 919 920 return (0); 921} 922 923/* 924 * System call request from POSIX system call gate interface to kernel. 925 * Like trap(), argument is call by reference. 926 */ 927void 928syscall(frame) 929 struct trapframe frame; 930{ 931 caddr_t params; 932 int i; 933 struct sysent *callp; 934 struct proc *p = curproc; 935 u_quad_t sticks; 936 int error; 937 int args[8]; 938 u_int code; 939 940#ifdef DIAGNOSTIC 941 if (ISPL(frame.tf_cs) != SEL_UPL) 942 panic("syscall"); 943#endif 944 sticks = p->p_sticks; 945 p->p_md.md_regs = &frame; 946 params = (caddr_t)frame.tf_esp + sizeof(int); 947 code = frame.tf_eax; 948 if (p->p_sysent->sv_prepsyscall) { 949 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 950 } else { 951 /* 952 * Need to check if this is a 32 bit or 64 bit syscall. 953 */ 954 if (code == SYS_syscall) { 955 /* 956 * Code is first argument, followed by actual args. 957 */ 958 code = fuword(params); 959 params += sizeof(int); 960 } else if (code == SYS___syscall) { 961 /* 962 * Like syscall, but code is a quad, so as to maintain 963 * quad alignment for the rest of the arguments. 964 */ 965 code = fuword(params); 966 params += sizeof(quad_t); 967 } 968 } 969 970 if (p->p_sysent->sv_mask) 971 code &= p->p_sysent->sv_mask; 972 973 if (code >= p->p_sysent->sv_size) 974 callp = &p->p_sysent->sv_table[0]; 975 else 976 callp = &p->p_sysent->sv_table[code]; 977 978 if (params && (i = callp->sy_narg * sizeof(int)) && 979 (error = copyin(params, (caddr_t)args, (u_int)i))) { 980#ifdef KTRACE 981 if (KTRPOINT(p, KTR_SYSCALL)) 982 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 983#endif 984 goto bad; 985 } 986#ifdef KTRACE 987 if (KTRPOINT(p, KTR_SYSCALL)) 988 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 989#endif 990 p->p_retval[0] = 0; 991 p->p_retval[1] = frame.tf_edx; 992 993 STOPEVENT(p, S_SCE, callp->sy_narg); 994 995 error = (*callp->sy_call)(p, args); 996 997 switch (error) { 998 999 case 0: 1000 /* 1001 * Reinitialize proc pointer `p' as it may be different 1002 * if this is a child returning from fork syscall. 1003 */ 1004 p = curproc; 1005 frame.tf_eax = p->p_retval[0]; 1006 frame.tf_edx = p->p_retval[1]; 1007 frame.tf_eflags &= ~PSL_C; 1008 break; 1009 1010 case ERESTART: 1011 /* 1012 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 1013 * int 0x80 is 2 bytes. We saved this in tf_err. 1014 */ 1015 frame.tf_eip -= frame.tf_err; 1016 break; 1017 1018 case EJUSTRETURN: 1019 break; 1020 1021 default: 1022bad: 1023 if (p->p_sysent->sv_errsize) 1024 if (error >= p->p_sysent->sv_errsize) 1025 error = -1; /* XXX */ 1026 else 1027 error = p->p_sysent->sv_errtbl[error]; 1028 frame.tf_eax = error; 1029 frame.tf_eflags |= PSL_C; 1030 break; 1031 } 1032 1033 if ((frame.tf_eflags & PSL_T) && !(frame.tf_eflags & PSL_VM)) { 1034 /* Traced syscall. */ 1035 frame.tf_eflags &= ~PSL_T; 1036 trapsignal(p, SIGTRAP, 0); 1037 } 1038 1039 userret(p, &frame, sticks); 1040 1041#ifdef KTRACE 1042 if (KTRPOINT(p, KTR_SYSRET)) 1043 ktrsysret(p->p_tracep, code, error, p->p_retval[0]); 1044#endif 1045 1046 /* 1047 * This works because errno is findable through the 1048 * register set. If we ever support an emulation where this 1049 * is not the case, this code will need to be revisited. 1050 */ 1051 STOPEVENT(p, S_SCX, code); 1052 1053} 1054 1055/* 1056 * Simplified back end of syscall(), used when returning from fork() 1057 * directly into user mode. 1058 */ 1059void 1060fork_return(p, frame) 1061 struct proc *p; 1062 struct trapframe frame; 1063{ 1064 frame.tf_eax = 0; /* Child returns zero */ 1065 frame.tf_eflags &= ~PSL_C; /* success */ 1066 frame.tf_edx = 1; 1067 1068 userret(p, &frame, 0); 1069#ifdef KTRACE 1070 if (KTRPOINT(p, KTR_SYSRET)) 1071 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 1072#endif 1073} 1074