subr_syscall.c revision 16725
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 1.78 1996/06/13 07:17:21 asami Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include "opt_ktrace.h" 46#include "opt_ddb.h" 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/proc.h> 51#include <sys/acct.h> 52#include <sys/kernel.h> 53#include <sys/syscall.h> 54#include <sys/sysent.h> 55#include <sys/queue.h> 56#include <sys/vmmeter.h> 57#ifdef KTRACE 58#include <sys/ktrace.h> 59#endif 60 61#include <vm/vm.h> 62#include <vm/vm_param.h> 63#include <vm/vm_prot.h> 64#include <vm/lock.h> 65#include <vm/pmap.h> 66#include <vm/vm_kern.h> 67#include <vm/vm_map.h> 68#include <vm/vm_page.h> 69#include <vm/vm_extern.h> 70 71#include <sys/user.h> 72 73#include <machine/cpu.h> 74#include <machine/md_var.h> 75#include <machine/psl.h> 76#include <machine/reg.h> 77#include <machine/trap.h> 78#include <machine/../isa/isa_device.h> 79 80#ifdef POWERFAIL_NMI 81# include <syslog.h> 82# include <machine/clock.h> 83#endif 84 85#include "isa.h" 86#include "npx.h" 87 88int (*pmath_emulate) __P((struct trapframe *)); 89 90extern void trap __P((struct trapframe frame)); 91extern int trapwrite __P((unsigned addr)); 92extern void syscall __P((struct trapframe frame)); 93 94static int trap_pfault __P((struct trapframe *, int)); 95static void trap_fatal __P((struct trapframe *)); 96void dblfault_handler __P((void)); 97 98extern inthand_t IDTVEC(syscall); 99 100#define MAX_TRAP_MSG 27 101static char *trap_msg[] = { 102 "", /* 0 unused */ 103 "privileged instruction fault", /* 1 T_PRIVINFLT */ 104 "", /* 2 unused */ 105 "breakpoint instruction fault", /* 3 T_BPTFLT */ 106 "", /* 4 unused */ 107 "", /* 5 unused */ 108 "arithmetic trap", /* 6 T_ARITHTRAP */ 109 "system forced exception", /* 7 T_ASTFLT */ 110 "", /* 8 unused */ 111 "general protection fault", /* 9 T_PROTFLT */ 112 "trace trap", /* 10 T_TRCTRAP */ 113 "", /* 11 unused */ 114 "page fault", /* 12 T_PAGEFLT */ 115 "", /* 13 unused */ 116 "alignment fault", /* 14 T_ALIGNFLT */ 117 "", /* 15 unused */ 118 "", /* 16 unused */ 119 "", /* 17 unused */ 120 "integer divide fault", /* 18 T_DIVIDE */ 121 "non-maskable interrupt trap", /* 19 T_NMI */ 122 "overflow trap", /* 20 T_OFLOW */ 123 "FPU bounds check fault", /* 21 T_BOUND */ 124 "FPU device not available", /* 22 T_DNA */ 125 "double fault", /* 23 T_DOUBLEFLT */ 126 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 127 "invalid TSS fault", /* 25 T_TSSFLT */ 128 "segment not present fault", /* 26 T_SEGNPFLT */ 129 "stack fault", /* 27 T_STKFLT */ 130}; 131 132static void userret __P((struct proc *p, struct trapframe *frame, 133 u_quad_t oticks)); 134 135static inline void 136userret(p, frame, oticks) 137 struct proc *p; 138 struct trapframe *frame; 139 u_quad_t oticks; 140{ 141 int sig, s; 142 143 while ((sig = CURSIG(p)) != 0) 144 postsig(sig); 145 p->p_priority = p->p_usrpri; 146 if (want_resched) { 147 /* 148 * Since we are curproc, clock will normally just change 149 * our priority without moving us from one queue to another 150 * (since the running process is not on a queue.) 151 * If that happened after we setrunqueue ourselves but before we 152 * mi_switch()'ed, we might not be on the queue indicated by 153 * our priority. 154 */ 155 s = splclock(); 156 setrunqueue(p); 157 p->p_stats->p_ru.ru_nivcsw++; 158 mi_switch(); 159 splx(s); 160 while ((sig = CURSIG(p)) != 0) 161 postsig(sig); 162 } 163 /* 164 * Charge system time if profiling. 165 */ 166 if (p->p_flag & P_PROFIL) 167 addupc_task(p, frame->tf_eip, 168 (u_int)(p->p_sticks - oticks) * psratio); 169 170 curpriority = p->p_priority; 171} 172 173/* 174 * Exception, fault, and trap interface to the FreeBSD kernel. 175 * This common code is called from assembly language IDT gate entry 176 * routines that prepare a suitable stack frame, and restore this 177 * frame after the exception has been processed. 178 */ 179 180void 181trap(frame) 182 struct trapframe frame; 183{ 184 struct proc *p = curproc; 185 u_quad_t sticks = 0; 186 int i = 0, ucode = 0, type, code; 187#ifdef DEBUG 188 u_long eva; 189#endif 190 191 type = frame.tf_trapno; 192 code = frame.tf_err; 193 194 if (ISPL(frame.tf_cs) == SEL_UPL) { 195 /* user trap */ 196 197 sticks = p->p_sticks; 198 p->p_md.md_regs = (int *)&frame; 199 200 switch (type) { 201 case T_PRIVINFLT: /* privileged instruction fault */ 202 ucode = type; 203 i = SIGILL; 204 break; 205 206 case T_BPTFLT: /* bpt instruction fault */ 207 case T_TRCTRAP: /* trace trap */ 208 frame.tf_eflags &= ~PSL_T; 209 i = SIGTRAP; 210 break; 211 212 case T_ARITHTRAP: /* arithmetic trap */ 213 ucode = code; 214 i = SIGFPE; 215 break; 216 217 case T_ASTFLT: /* Allow process switch */ 218 astoff(); 219 cnt.v_soft++; 220 if (p->p_flag & P_OWEUPC) { 221 p->p_flag &= ~P_OWEUPC; 222 addupc_task(p, p->p_stats->p_prof.pr_addr, 223 p->p_stats->p_prof.pr_ticks); 224 } 225 goto out; 226 227 case T_PROTFLT: /* general protection fault */ 228 case T_SEGNPFLT: /* segment not present fault */ 229 case T_STKFLT: /* stack fault */ 230 case T_TSSFLT: /* invalid TSS fault */ 231 case T_DOUBLEFLT: /* double fault */ 232 default: 233 ucode = code + BUS_SEGM_FAULT ; 234 i = SIGBUS; 235 break; 236 237 case T_PAGEFLT: /* page fault */ 238 i = trap_pfault(&frame, TRUE); 239 if (i == -1) 240 return; 241 if (i == 0) 242 goto out; 243 244 ucode = T_PAGEFLT; 245 break; 246 247 case T_DIVIDE: /* integer divide fault */ 248 ucode = FPE_INTDIV_TRAP; 249 i = SIGFPE; 250 break; 251 252#if NISA > 0 253 case T_NMI: 254#ifdef POWERFAIL_NMI 255 goto handle_powerfail; 256#else /* !POWERFAIL_NMI */ 257#ifdef DDB 258 /* NMI can be hooked up to a pushbutton for debugging */ 259 printf ("NMI ... going to debugger\n"); 260 if (kdb_trap (type, 0, &frame)) 261 return; 262#endif /* DDB */ 263 /* machine/parity/power fail/"kitchen sink" faults */ 264 if (isa_nmi(code) == 0) return; 265 panic("NMI indicates hardware failure"); 266#endif /* POWERFAIL_NMI */ 267#endif /* NISA > 0 */ 268 269 case T_OFLOW: /* integer overflow fault */ 270 ucode = FPE_INTOVF_TRAP; 271 i = SIGFPE; 272 break; 273 274 case T_BOUND: /* bounds check fault */ 275 ucode = FPE_SUBRNG_TRAP; 276 i = SIGFPE; 277 break; 278 279 case T_DNA: 280#if NNPX > 0 281 /* if a transparent fault (due to context switch "late") */ 282 if (npxdna()) 283 return; 284#endif /* NNPX > 0 */ 285 286 if (!pmath_emulate) { 287 i = SIGFPE; 288 ucode = FPE_FPU_NP_TRAP; 289 break; 290 } 291 i = (*pmath_emulate)(&frame); 292 if (i == 0) { 293 if (!(frame.tf_eflags & PSL_T)) 294 return; 295 frame.tf_eflags &= ~PSL_T; 296 i = SIGTRAP; 297 } 298 /* else ucode = emulator_only_knows() XXX */ 299 break; 300 301 case T_FPOPFLT: /* FPU operand fetch fault */ 302 ucode = T_FPOPFLT; 303 i = SIGILL; 304 break; 305 } 306 } else { 307 /* kernel trap */ 308 309 switch (type) { 310 case T_PAGEFLT: /* page fault */ 311 (void) trap_pfault(&frame, FALSE); 312 return; 313 314 case T_DNA: 315#if NNPX > 0 316 /* if a transparent fault (due to context switch "late") */ 317 if (npxdna()) 318 return; 319#endif /* NNPX > 0 */ 320 break; 321 322 case T_PROTFLT: /* general protection fault */ 323 case T_SEGNPFLT: /* segment not present fault */ 324 /* 325 * Invalid segment selectors and out of bounds 326 * %eip's and %esp's can be set up in user mode. 327 * This causes a fault in kernel mode when the 328 * kernel tries to return to user mode. We want 329 * to get this fault so that we can fix the 330 * problem here and not have to check all the 331 * selectors and pointers when the user changes 332 * them. 333 */ 334#define MAYBE_DORETI_FAULT(where, whereto) \ 335 do { \ 336 if (frame.tf_eip == (int)where) { \ 337 frame.tf_eip = (int)whereto; \ 338 return; \ 339 } \ 340 } while (0) 341 342 if (intr_nesting_level == 0) { 343 MAYBE_DORETI_FAULT(doreti_iret, 344 doreti_iret_fault); 345 MAYBE_DORETI_FAULT(doreti_popl_ds, 346 doreti_popl_ds_fault); 347 MAYBE_DORETI_FAULT(doreti_popl_es, 348 doreti_popl_es_fault); 349 } 350 if (curpcb && curpcb->pcb_onfault) { 351 frame.tf_eip = (int)curpcb->pcb_onfault; 352 return; 353 } 354 break; 355 356 case T_TSSFLT: 357 /* 358 * PSL_NT can be set in user mode and isn't cleared 359 * automatically when the kernel is entered. This 360 * causes a TSS fault when the kernel attempts to 361 * `iret' because the TSS link is uninitialized. We 362 * want to get this fault so that we can fix the 363 * problem here and not every time the kernel is 364 * entered. 365 */ 366 if (frame.tf_eflags & PSL_NT) { 367 frame.tf_eflags &= ~PSL_NT; 368 return; 369 } 370 break; 371 372 case T_TRCTRAP: /* trace trap */ 373 if (frame.tf_eip == (int)IDTVEC(syscall)) { 374 /* 375 * We've just entered system mode via the 376 * syscall lcall. Continue single stepping 377 * silently until the syscall handler has 378 * saved the flags. 379 */ 380 return; 381 } 382 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 383 /* 384 * The syscall handler has now saved the 385 * flags. Stop single stepping it. 386 */ 387 frame.tf_eflags &= ~PSL_T; 388 return; 389 } 390 /* 391 * Fall through. 392 */ 393 case T_BPTFLT: 394 /* 395 * If DDB is enabled, let it handle the debugger trap. 396 * Otherwise, debugger traps "can't happen". 397 */ 398#ifdef DDB 399 if (kdb_trap (type, 0, &frame)) 400 return; 401#endif 402 break; 403 404#if NISA > 0 405 case T_NMI: 406#ifdef POWERFAIL_NMI 407#ifndef TIMER_FREQ 408# define TIMER_FREQ 1193182 409#endif 410 handle_powerfail: 411 { 412 static unsigned lastalert = 0; 413 414 if(time.tv_sec - lastalert > 10) 415 { 416 log(LOG_WARNING, "NMI: power fail\n"); 417 sysbeep(TIMER_FREQ/880, hz); 418 lastalert = time.tv_sec; 419 } 420 return; 421 } 422#else /* !POWERFAIL_NMI */ 423#ifdef DDB 424 /* NMI can be hooked up to a pushbutton for debugging */ 425 printf ("NMI ... going to debugger\n"); 426 if (kdb_trap (type, 0, &frame)) 427 return; 428#endif /* DDB */ 429 /* machine/parity/power fail/"kitchen sink" faults */ 430 if (isa_nmi(code) == 0) return; 431 /* FALL THROUGH */ 432#endif /* POWERFAIL_NMI */ 433#endif /* NISA > 0 */ 434 } 435 436 trap_fatal(&frame); 437 return; 438 } 439 440 trapsignal(p, i, ucode); 441 442#ifdef DEBUG 443 eva = rcr2(); 444 if (type <= MAX_TRAP_MSG) { 445 uprintf("fatal process exception: %s", 446 trap_msg[type]); 447 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 448 uprintf(", fault VA = 0x%x", eva); 449 uprintf("\n"); 450 } 451#endif 452 453out: 454 userret(p, &frame, sticks); 455} 456 457#ifdef notyet 458/* 459 * This version doesn't allow a page fault to user space while 460 * in the kernel. The rest of the kernel needs to be made "safe" 461 * before this can be used. I think the only things remaining 462 * to be made safe are the iBCS2 code and the process tracing/ 463 * debugging code. 464 */ 465static int 466trap_pfault(frame, usermode) 467 struct trapframe *frame; 468 int usermode; 469{ 470 vm_offset_t va; 471 struct vmspace *vm = NULL; 472 vm_map_t map = 0; 473 int rv = 0; 474 vm_prot_t ftype; 475 int eva; 476 struct proc *p = curproc; 477 478 if (frame->tf_err & PGEX_W) 479 ftype = VM_PROT_READ | VM_PROT_WRITE; 480 else 481 ftype = VM_PROT_READ; 482 483 eva = rcr2(); 484 va = trunc_page((vm_offset_t)eva); 485 486 if (va < VM_MIN_KERNEL_ADDRESS) { 487 vm_offset_t v; 488 vm_page_t mpte; 489 490 if (p == NULL || 491 (!usermode && va < VM_MAXUSER_ADDRESS && 492 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 493 trap_fatal(frame); 494 return (-1); 495 } 496 497 /* 498 * This is a fault on non-kernel virtual memory. 499 * vm is initialized above to NULL. If curproc is NULL 500 * or curproc->p_vmspace is NULL the fault is fatal. 501 */ 502 vm = p->p_vmspace; 503 if (vm == NULL) 504 goto nogo; 505 506 map = &vm->vm_map; 507 508 /* 509 * Keep swapout from messing with us during this 510 * critical time. 511 */ 512 ++p->p_lock; 513 514 /* 515 * Grow the stack if necessary 516 */ 517 if ((caddr_t)va > vm->vm_maxsaddr 518 && (caddr_t)va < (caddr_t)USRSTACK) { 519 if (!grow(p, va)) { 520 rv = KERN_FAILURE; 521 --p->p_lock; 522 goto nogo; 523 } 524 } 525 526 /* Fault in the user page: */ 527 rv = vm_fault(map, va, ftype, FALSE); 528 529 --p->p_lock; 530 } else { 531 /* 532 * Don't allow user-mode faults in kernel address space. 533 */ 534 if (usermode) 535 goto nogo; 536 537 /* 538 * Since we know that kernel virtual address addresses 539 * always have pte pages mapped, we just have to fault 540 * the page. 541 */ 542 rv = vm_fault(kernel_map, va, ftype, FALSE); 543 } 544 545 if (rv == KERN_SUCCESS) 546 return (0); 547nogo: 548 if (!usermode) { 549 if (curpcb && curpcb->pcb_onfault) { 550 frame->tf_eip = (int)curpcb->pcb_onfault; 551 return (0); 552 } 553 trap_fatal(frame); 554 return (-1); 555 } 556 557 /* kludge to pass faulting virtual address to sendsig */ 558 frame->tf_err = eva; 559 560 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 561} 562#endif 563 564int 565trap_pfault(frame, usermode) 566 struct trapframe *frame; 567 int usermode; 568{ 569 vm_offset_t va; 570 struct vmspace *vm = NULL; 571 vm_map_t map = 0; 572 int rv = 0; 573 vm_prot_t ftype; 574 int eva; 575 struct proc *p = curproc; 576 577 eva = rcr2(); 578 va = trunc_page((vm_offset_t)eva); 579 580 if (va >= KERNBASE) { 581 /* 582 * Don't allow user-mode faults in kernel address space. 583 */ 584 if (usermode) 585 goto nogo; 586 587 map = kernel_map; 588 } else { 589 /* 590 * This is a fault on non-kernel virtual memory. 591 * vm is initialized above to NULL. If curproc is NULL 592 * or curproc->p_vmspace is NULL the fault is fatal. 593 */ 594 if (p != NULL) 595 vm = p->p_vmspace; 596 597 if (vm == NULL) 598 goto nogo; 599 600 map = &vm->vm_map; 601 } 602 603 if (frame->tf_err & PGEX_W) 604 ftype = VM_PROT_READ | VM_PROT_WRITE; 605 else 606 ftype = VM_PROT_READ; 607 608 if (map != kernel_map) { 609 /* 610 * Keep swapout from messing with us during this 611 * critical time. 612 */ 613 ++p->p_lock; 614 615 /* 616 * Grow the stack if necessary 617 */ 618 if ((caddr_t)va > vm->vm_maxsaddr 619 && (caddr_t)va < (caddr_t)USRSTACK) { 620 if (!grow(p, va)) { 621 rv = KERN_FAILURE; 622 --p->p_lock; 623 goto nogo; 624 } 625 } 626 627 /* Fault in the user page: */ 628 rv = vm_fault(map, va, ftype, FALSE); 629 630 --p->p_lock; 631 } else { 632 /* 633 * Since we know that kernel virtual address addresses 634 * always have pte pages mapped, we just have to fault 635 * the page. 636 */ 637 rv = vm_fault(map, va, ftype, FALSE); 638 } 639 640 if (rv == KERN_SUCCESS) 641 return (0); 642nogo: 643 if (!usermode) { 644 if (curpcb && curpcb->pcb_onfault) { 645 frame->tf_eip = (int)curpcb->pcb_onfault; 646 return (0); 647 } 648 trap_fatal(frame); 649 return (-1); 650 } 651 652 /* kludge to pass faulting virtual address to sendsig */ 653 frame->tf_err = eva; 654 655 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 656} 657 658static void 659trap_fatal(frame) 660 struct trapframe *frame; 661{ 662 int code, type, eva, ss, esp; 663 struct soft_segment_descriptor softseg; 664 665 code = frame->tf_err; 666 type = frame->tf_trapno; 667 eva = rcr2(); 668 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 669 670 if (type <= MAX_TRAP_MSG) 671 printf("\n\nFatal trap %d: %s while in %s mode\n", 672 type, trap_msg[type], 673 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 674 if (type == T_PAGEFLT) { 675 printf("fault virtual address = 0x%x\n", eva); 676 printf("fault code = %s %s, %s\n", 677 code & PGEX_U ? "user" : "supervisor", 678 code & PGEX_W ? "write" : "read", 679 code & PGEX_P ? "protection violation" : "page not present"); 680 } 681 printf("instruction pointer = 0x%x:0x%x\n", 682 frame->tf_cs & 0xffff, frame->tf_eip); 683 if (ISPL(frame->tf_cs) == SEL_UPL) { 684 ss = frame->tf_ss & 0xffff; 685 esp = frame->tf_esp; 686 } else { 687 ss = GSEL(GDATA_SEL, SEL_KPL); 688 esp = (int)&frame->tf_esp; 689 } 690 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 691 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 692 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 693 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 694 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 695 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 696 softseg.ssd_gran); 697 printf("processor eflags = "); 698 if (frame->tf_eflags & PSL_T) 699 printf("trace trap, "); 700 if (frame->tf_eflags & PSL_I) 701 printf("interrupt enabled, "); 702 if (frame->tf_eflags & PSL_NT) 703 printf("nested task, "); 704 if (frame->tf_eflags & PSL_RF) 705 printf("resume, "); 706 if (frame->tf_eflags & PSL_VM) 707 printf("vm86, "); 708 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 709 printf("current process = "); 710 if (curproc) { 711 printf("%lu (%s)\n", 712 (u_long)curproc->p_pid, curproc->p_comm ? 713 curproc->p_comm : ""); 714 } else { 715 printf("Idle\n"); 716 } 717 printf("interrupt mask = "); 718 if ((cpl & net_imask) == net_imask) 719 printf("net "); 720 if ((cpl & tty_imask) == tty_imask) 721 printf("tty "); 722 if ((cpl & bio_imask) == bio_imask) 723 printf("bio "); 724 if (cpl == 0) 725 printf("none"); 726 printf("\n"); 727 728#ifdef KDB 729 if (kdb_trap(&psl)) 730 return; 731#endif 732#ifdef DDB 733 if (kdb_trap (type, 0, frame)) 734 return; 735#endif 736 if (type <= MAX_TRAP_MSG) 737 panic(trap_msg[type]); 738 else 739 panic("unknown/reserved trap"); 740} 741 742/* 743 * Double fault handler. Called when a fault occurs while writing 744 * a frame for a trap/exception onto the stack. This usually occurs 745 * when the stack overflows (such is the case with infinite recursion, 746 * for example). 747 * 748 * XXX Note that the current PTD gets replaced by IdlePTD when the 749 * task switch occurs. This means that the stack that was active at 750 * the time of the double fault is not available at <kstack> unless 751 * the machine was idle when the double fault occurred. The downside 752 * of this is that "trace <ebp>" in ddb won't work. 753 */ 754void 755dblfault_handler() 756{ 757 struct pcb *pcb = curpcb; 758 759 if (pcb != NULL) { 760 printf("\nFatal double fault:\n"); 761 printf("eip = 0x%x\n", pcb->pcb_tss.tss_eip); 762 printf("esp = 0x%x\n", pcb->pcb_tss.tss_esp); 763 printf("ebp = 0x%x\n", pcb->pcb_tss.tss_ebp); 764 } 765 766 panic("double fault"); 767} 768 769/* 770 * Compensate for 386 brain damage (missing URKR). 771 * This is a little simpler than the pagefault handler in trap() because 772 * it the page tables have already been faulted in and high addresses 773 * are thrown out early for other reasons. 774 */ 775int trapwrite(addr) 776 unsigned addr; 777{ 778 struct proc *p; 779 vm_offset_t va, v; 780 struct vmspace *vm; 781 int rv; 782 783 va = trunc_page((vm_offset_t)addr); 784 /* 785 * XXX - MAX is END. Changed > to >= for temp. fix. 786 */ 787 if (va >= VM_MAXUSER_ADDRESS) 788 return (1); 789 790 p = curproc; 791 vm = p->p_vmspace; 792 793 ++p->p_lock; 794 795 if ((caddr_t)va >= vm->vm_maxsaddr 796 && (caddr_t)va < (caddr_t)USRSTACK) { 797 if (!grow(p, va)) { 798 --p->p_lock; 799 return (1); 800 } 801 } 802 803 v = trunc_page(vtopte(va)); 804 805 /* 806 * fault the data page 807 */ 808 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 809 810 --p->p_lock; 811 812 if (rv != KERN_SUCCESS) 813 return 1; 814 815 return (0); 816} 817 818/* 819 * System call request from POSIX system call gate interface to kernel. 820 * Like trap(), argument is call by reference. 821 */ 822void 823syscall(frame) 824 struct trapframe frame; 825{ 826 caddr_t params; 827 int i; 828 struct sysent *callp; 829 struct proc *p = curproc; 830 u_quad_t sticks; 831 int error; 832 int args[8], rval[2]; 833 u_int code; 834 835 sticks = p->p_sticks; 836 if (ISPL(frame.tf_cs) != SEL_UPL) 837 panic("syscall"); 838 839 p->p_md.md_regs = (int *)&frame; 840 params = (caddr_t)frame.tf_esp + sizeof(int); 841 code = frame.tf_eax; 842 if (p->p_sysent->sv_prepsyscall) { 843 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 844 } else { 845 /* 846 * Need to check if this is a 32 bit or 64 bit syscall. 847 */ 848 if (code == SYS_syscall) { 849 /* 850 * Code is first argument, followed by actual args. 851 */ 852 code = fuword(params); 853 params += sizeof(int); 854 } else if (code == SYS___syscall) { 855 /* 856 * Like syscall, but code is a quad, so as to maintain 857 * quad alignment for the rest of the arguments. 858 */ 859 code = fuword(params); 860 params += sizeof(quad_t); 861 } 862 } 863 864 if (p->p_sysent->sv_mask) 865 code &= p->p_sysent->sv_mask; 866 867 if (code >= p->p_sysent->sv_size) 868 callp = &p->p_sysent->sv_table[0]; 869 else 870 callp = &p->p_sysent->sv_table[code]; 871 872 if (params && (i = callp->sy_narg * sizeof(int)) && 873 (error = copyin(params, (caddr_t)args, (u_int)i))) { 874#ifdef KTRACE 875 if (KTRPOINT(p, KTR_SYSCALL)) 876 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 877#endif 878 goto bad; 879 } 880#ifdef KTRACE 881 if (KTRPOINT(p, KTR_SYSCALL)) 882 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 883#endif 884 rval[0] = 0; 885 rval[1] = frame.tf_edx; 886 887 error = (*callp->sy_call)(p, args, rval); 888 889 switch (error) { 890 891 case 0: 892 /* 893 * Reinitialize proc pointer `p' as it may be different 894 * if this is a child returning from fork syscall. 895 */ 896 p = curproc; 897 frame.tf_eax = rval[0]; 898 frame.tf_edx = rval[1]; 899 frame.tf_eflags &= ~PSL_C; 900 break; 901 902 case ERESTART: 903 /* 904 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 905 * int 0x80 is 2 bytes. We saved this in tf_err. 906 */ 907 frame.tf_eip -= frame.tf_err; 908 break; 909 910 case EJUSTRETURN: 911 break; 912 913 default: 914bad: 915 if (p->p_sysent->sv_errsize) 916 if (error >= p->p_sysent->sv_errsize) 917 error = -1; /* XXX */ 918 else 919 error = p->p_sysent->sv_errtbl[error]; 920 frame.tf_eax = error; 921 frame.tf_eflags |= PSL_C; 922 break; 923 } 924 925 if (frame.tf_eflags & PSL_T) { 926 /* Traced syscall. */ 927 frame.tf_eflags &= ~PSL_T; 928 trapsignal(p, SIGTRAP, 0); 929 } 930 931 userret(p, &frame, sticks); 932 933#ifdef KTRACE 934 if (KTRPOINT(p, KTR_SYSRET)) 935 ktrsysret(p->p_tracep, code, error, rval[0]); 936#endif 937} 938