subr_syscall.c revision 12930
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 1.67 1995/12/19 14:30:28 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/acct.h> 49#include <sys/kernel.h> 50#include <sys/syscall.h> 51#include <sys/sysent.h> 52#include <sys/queue.h> 53#include <sys/vmmeter.h> 54#ifdef KTRACE 55#include <sys/ktrace.h> 56#endif 57 58#include <vm/vm.h> 59#include <vm/vm_param.h> 60#include <vm/vm_prot.h> 61#include <vm/lock.h> 62#include <vm/pmap.h> 63#include <vm/vm_kern.h> 64#include <vm/vm_map.h> 65#include <vm/vm_page.h> 66#include <vm/vm_extern.h> 67 68#include <sys/user.h> 69 70#include <machine/cpu.h> 71#include <machine/md_var.h> 72#include <machine/psl.h> 73#include <machine/reg.h> 74#include <machine/trap.h> 75#include <machine/../isa/isa_device.h> 76 77#ifdef POWERFAIL_NMI 78# include <syslog.h> 79# include <machine/clock.h> 80#endif 81 82#include "isa.h" 83#include "npx.h" 84 85int (*pmath_emulate) __P((struct trapframe *)); 86 87extern void trap __P((struct trapframe frame)); 88extern int trapwrite __P((unsigned addr)); 89extern void syscall __P((struct trapframe frame)); 90extern void linux_syscall __P((struct trapframe frame)); 91 92static int trap_pfault __P((struct trapframe *, int)); 93static void trap_fatal __P((struct trapframe *)); 94void dblfault_handler __P((void)); 95 96extern inthand_t IDTVEC(syscall); 97 98#define MAX_TRAP_MSG 27 99static char *trap_msg[] = { 100 "", /* 0 unused */ 101 "privileged instruction fault", /* 1 T_PRIVINFLT */ 102 "", /* 2 unused */ 103 "breakpoint instruction fault", /* 3 T_BPTFLT */ 104 "", /* 4 unused */ 105 "", /* 5 unused */ 106 "arithmetic trap", /* 6 T_ARITHTRAP */ 107 "system forced exception", /* 7 T_ASTFLT */ 108 "", /* 8 unused */ 109 "general protection fault", /* 9 T_PROTFLT */ 110 "trace trap", /* 10 T_TRCTRAP */ 111 "", /* 11 unused */ 112 "page fault", /* 12 T_PAGEFLT */ 113 "", /* 13 unused */ 114 "alignment fault", /* 14 T_ALIGNFLT */ 115 "", /* 15 unused */ 116 "", /* 16 unused */ 117 "", /* 17 unused */ 118 "integer divide fault", /* 18 T_DIVIDE */ 119 "non-maskable interrupt trap", /* 19 T_NMI */ 120 "overflow trap", /* 20 T_OFLOW */ 121 "FPU bounds check fault", /* 21 T_BOUND */ 122 "FPU device not available", /* 22 T_DNA */ 123 "double fault", /* 23 T_DOUBLEFLT */ 124 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 125 "invalid TSS fault", /* 25 T_TSSFLT */ 126 "segment not present fault", /* 26 T_SEGNPFLT */ 127 "stack fault", /* 27 T_STKFLT */ 128}; 129 130static void userret __P((struct proc *p, struct trapframe *frame, 131 u_quad_t oticks)); 132 133static inline void 134userret(p, frame, oticks) 135 struct proc *p; 136 struct trapframe *frame; 137 u_quad_t oticks; 138{ 139 int sig, s; 140 141 while ((sig = CURSIG(p)) != 0) 142 postsig(sig); 143 p->p_priority = p->p_usrpri; 144 if (want_resched) { 145 /* 146 * Since we are curproc, clock will normally just change 147 * our priority without moving us from one queue to another 148 * (since the running process is not on a queue.) 149 * If that happened after we setrunqueue ourselves but before we 150 * mi_switch()'ed, we might not be on the queue indicated by 151 * our priority. 152 */ 153 s = splclock(); 154 setrunqueue(p); 155 p->p_stats->p_ru.ru_nivcsw++; 156 mi_switch(); 157 splx(s); 158 while ((sig = CURSIG(p)) != 0) 159 postsig(sig); 160 } 161 /* 162 * Charge system time if profiling. 163 */ 164 if (p->p_flag & P_PROFIL) { 165 u_quad_t ticks = p->p_sticks - oticks; 166 167 if (ticks) { 168#ifdef PROFTIMER 169 extern int profscale; 170 addupc(frame->tf_eip, &p->p_stats->p_prof, 171 ticks * profscale); 172#else 173 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 174#endif 175 } 176 } 177 curpriority = p->p_priority; 178} 179 180/* 181 * Exception, fault, and trap interface to the FreeBSD kernel. 182 * This common code is called from assembly language IDT gate entry 183 * routines that prepare a suitable stack frame, and restore this 184 * frame after the exception has been processed. 185 */ 186 187void 188trap(frame) 189 struct trapframe frame; 190{ 191 struct proc *p = curproc; 192 u_quad_t sticks = 0; 193 int i = 0, ucode = 0, type, code; 194#ifdef DEBUG 195 u_long eva; 196#endif 197 198 type = frame.tf_trapno; 199 code = frame.tf_err; 200 201 if (ISPL(frame.tf_cs) == SEL_UPL) { 202 /* user trap */ 203 204 sticks = p->p_sticks; 205 p->p_md.md_regs = (int *)&frame; 206 207 switch (type) { 208 case T_PRIVINFLT: /* privileged instruction fault */ 209 ucode = type; 210 i = SIGILL; 211 break; 212 213 case T_BPTFLT: /* bpt instruction fault */ 214 case T_TRCTRAP: /* trace trap */ 215 frame.tf_eflags &= ~PSL_T; 216 i = SIGTRAP; 217 break; 218 219 case T_ARITHTRAP: /* arithmetic trap */ 220 ucode = code; 221 i = SIGFPE; 222 break; 223 224 case T_ASTFLT: /* Allow process switch */ 225 astoff(); 226 cnt.v_soft++; 227 if (p->p_flag & P_OWEUPC) { 228 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 229 p->p_flag &= ~P_OWEUPC; 230 } 231 goto out; 232 233 case T_PROTFLT: /* general protection fault */ 234 case T_SEGNPFLT: /* segment not present fault */ 235 case T_STKFLT: /* stack fault */ 236 case T_TSSFLT: /* invalid TSS fault */ 237 case T_DOUBLEFLT: /* double fault */ 238 default: 239 ucode = code + BUS_SEGM_FAULT ; 240 i = SIGBUS; 241 break; 242 243 case T_PAGEFLT: /* page fault */ 244 i = trap_pfault(&frame, TRUE); 245 if (i == -1) 246 return; 247 if (i == 0) 248 goto out; 249 250 ucode = T_PAGEFLT; 251 break; 252 253 case T_DIVIDE: /* integer divide fault */ 254 ucode = FPE_INTDIV_TRAP; 255 i = SIGFPE; 256 break; 257 258#if NISA > 0 259 case T_NMI: 260#ifdef POWERFAIL_NMI 261 goto handle_powerfail; 262#else /* !POWERFAIL_NMI */ 263#ifdef DDB 264 /* NMI can be hooked up to a pushbutton for debugging */ 265 printf ("NMI ... going to debugger\n"); 266 if (kdb_trap (type, 0, &frame)) 267 return; 268#endif /* DDB */ 269 /* machine/parity/power fail/"kitchen sink" faults */ 270 if (isa_nmi(code) == 0) return; 271 panic("NMI indicates hardware failure"); 272#endif /* POWERFAIL_NMI */ 273#endif /* NISA > 0 */ 274 275 case T_OFLOW: /* integer overflow fault */ 276 ucode = FPE_INTOVF_TRAP; 277 i = SIGFPE; 278 break; 279 280 case T_BOUND: /* bounds check fault */ 281 ucode = FPE_SUBRNG_TRAP; 282 i = SIGFPE; 283 break; 284 285 case T_DNA: 286#if NNPX > 0 287 /* if a transparent fault (due to context switch "late") */ 288 if (npxdna()) 289 return; 290#endif /* NNPX > 0 */ 291 292 if (!pmath_emulate) { 293 i = SIGFPE; 294 ucode = FPE_FPU_NP_TRAP; 295 break; 296 } 297 i = (*pmath_emulate)(&frame); 298 if (i == 0) { 299 if (!(frame.tf_eflags & PSL_T)) 300 return; 301 frame.tf_eflags &= ~PSL_T; 302 i = SIGTRAP; 303 } 304 /* else ucode = emulator_only_knows() XXX */ 305 break; 306 307 case T_FPOPFLT: /* FPU operand fetch fault */ 308 ucode = T_FPOPFLT; 309 i = SIGILL; 310 break; 311 } 312 } else { 313 /* kernel trap */ 314 315 switch (type) { 316 case T_PAGEFLT: /* page fault */ 317 (void) trap_pfault(&frame, FALSE); 318 return; 319 320 case T_PROTFLT: /* general protection fault */ 321 case T_SEGNPFLT: /* segment not present fault */ 322 /* 323 * Invalid segment selectors and out of bounds 324 * %eip's and %esp's can be set up in user mode. 325 * This causes a fault in kernel mode when the 326 * kernel tries to return to user mode. We want 327 * to get this fault so that we can fix the 328 * problem here and not have to check all the 329 * selectors and pointers when the user changes 330 * them. 331 */ 332#define MAYBE_DORETI_FAULT(where, whereto) \ 333 do { \ 334 if (frame.tf_eip == (int)where) { \ 335 frame.tf_eip = (int)whereto; \ 336 return; \ 337 } \ 338 } while (0) 339 340 if (intr_nesting_level == 0) { 341 MAYBE_DORETI_FAULT(doreti_iret, 342 doreti_iret_fault); 343 MAYBE_DORETI_FAULT(doreti_popl_ds, 344 doreti_popl_ds_fault); 345 MAYBE_DORETI_FAULT(doreti_popl_es, 346 doreti_popl_es_fault); 347 } 348 if (curpcb && curpcb->pcb_onfault) { 349 frame.tf_eip = (int)curpcb->pcb_onfault; 350 return; 351 } 352 break; 353 354 case T_TSSFLT: 355 /* 356 * PSL_NT can be set in user mode and isn't cleared 357 * automatically when the kernel is entered. This 358 * causes a TSS fault when the kernel attempts to 359 * `iret' because the TSS link is uninitialized. We 360 * want to get this fault so that we can fix the 361 * problem here and not every time the kernel is 362 * entered. 363 */ 364 if (frame.tf_eflags & PSL_NT) { 365 frame.tf_eflags &= ~PSL_NT; 366 return; 367 } 368 break; 369 370 case T_TRCTRAP: /* trace trap */ 371 if (frame.tf_eip == (int)IDTVEC(syscall)) { 372 /* 373 * We've just entered system mode via the 374 * syscall lcall. Continue single stepping 375 * silently until the syscall handler has 376 * saved the flags. 377 */ 378 return; 379 } 380 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 381 /* 382 * The syscall handler has now saved the 383 * flags. Stop single stepping it. 384 */ 385 frame.tf_eflags &= ~PSL_T; 386 return; 387 } 388 /* 389 * Fall through. 390 */ 391 case T_BPTFLT: 392 /* 393 * If DDB is enabled, let it handle the debugger trap. 394 * Otherwise, debugger traps "can't happen". 395 */ 396#ifdef DDB 397 if (kdb_trap (type, 0, &frame)) 398 return; 399#endif 400 break; 401 402#if NISA > 0 403 case T_NMI: 404#ifdef POWERFAIL_NMI 405#ifndef TIMER_FREQ 406# define TIMER_FREQ 1193182 407#endif 408 handle_powerfail: 409 { 410 static unsigned lastalert = 0; 411 412 if(time.tv_sec - lastalert > 10) 413 { 414 log(LOG_WARNING, "NMI: power fail\n"); 415 sysbeep(TIMER_FREQ/880, hz); 416 lastalert = time.tv_sec; 417 } 418 return; 419 } 420#else /* !POWERFAIL_NMI */ 421#ifdef DDB 422 /* NMI can be hooked up to a pushbutton for debugging */ 423 printf ("NMI ... going to debugger\n"); 424 if (kdb_trap (type, 0, &frame)) 425 return; 426#endif /* DDB */ 427 /* machine/parity/power fail/"kitchen sink" faults */ 428 if (isa_nmi(code) == 0) return; 429 /* FALL THROUGH */ 430#endif /* POWERFAIL_NMI */ 431#endif /* NISA > 0 */ 432 } 433 434 trap_fatal(&frame); 435 return; 436 } 437 438 trapsignal(p, i, ucode); 439 440#ifdef DEBUG 441 eva = rcr2(); 442 if (type <= MAX_TRAP_MSG) { 443 uprintf("fatal process exception: %s", 444 trap_msg[type]); 445 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 446 uprintf(", fault VA = 0x%x", eva); 447 uprintf("\n"); 448 } 449#endif 450 451out: 452 userret(p, &frame, sticks); 453} 454 455#ifdef notyet 456/* 457 * This version doesn't allow a page fault to user space while 458 * in the kernel. The rest of the kernel needs to be made "safe" 459 * before this can be used. I think the only things remaining 460 * to be made safe are the iBCS2 code and the process tracing/ 461 * debugging code. 462 */ 463static int 464trap_pfault(frame, usermode) 465 struct trapframe *frame; 466 int usermode; 467{ 468 vm_offset_t va; 469 struct vmspace *vm = NULL; 470 vm_map_t map = 0; 471 int rv = 0; 472 vm_prot_t ftype; 473 int eva; 474 struct proc *p = curproc; 475 476 if (frame->tf_err & PGEX_W) 477 ftype = VM_PROT_READ | VM_PROT_WRITE; 478 else 479 ftype = VM_PROT_READ; 480 481 eva = rcr2(); 482 va = trunc_page((vm_offset_t)eva); 483 484 if (va < VM_MIN_KERNEL_ADDRESS) { 485 vm_offset_t v; 486 vm_page_t ptepg; 487 488 if (p == NULL || 489 (!usermode && va < VM_MAXUSER_ADDRESS && 490 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 491 trap_fatal(frame); 492 return (-1); 493 } 494 495 /* 496 * This is a fault on non-kernel virtual memory. 497 * vm is initialized above to NULL. If curproc is NULL 498 * or curproc->p_vmspace is NULL the fault is fatal. 499 */ 500 vm = p->p_vmspace; 501 if (vm == NULL) 502 goto nogo; 503 504 map = &vm->vm_map; 505 506 /* 507 * Keep swapout from messing with us during this 508 * critical time. 509 */ 510 ++p->p_lock; 511 512 /* 513 * Grow the stack if necessary 514 */ 515 if ((caddr_t)va > vm->vm_maxsaddr 516 && (caddr_t)va < (caddr_t)USRSTACK) { 517 if (!grow(p, va)) { 518 rv = KERN_FAILURE; 519 --p->p_lock; 520 goto nogo; 521 } 522 } 523 524 /* 525 * Check if page table is mapped, if not, 526 * fault it first 527 */ 528 v = (vm_offset_t) vtopte(va); 529 530 /* Fault the pte only if needed: */ 531 if (*((int *)vtopte(v)) == 0) 532 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 533 534 pmap_use_pt( vm_map_pmap(map), va); 535 536 /* Fault in the user page: */ 537 rv = vm_fault(map, va, ftype, FALSE); 538 539 pmap_unuse_pt( vm_map_pmap(map), va); 540 541 --p->p_lock; 542 } else { 543 /* 544 * Don't allow user-mode faults in kernel address space. 545 */ 546 if (usermode) 547 goto nogo; 548 549 /* 550 * Since we know that kernel virtual address addresses 551 * always have pte pages mapped, we just have to fault 552 * the page. 553 */ 554 rv = vm_fault(kernel_map, va, ftype, FALSE); 555 } 556 557 if (rv == KERN_SUCCESS) 558 return (0); 559nogo: 560 if (!usermode) { 561 if (curpcb && curpcb->pcb_onfault) { 562 frame->tf_eip = (int)curpcb->pcb_onfault; 563 return (0); 564 } 565 trap_fatal(frame); 566 return (-1); 567 } 568 569 /* kludge to pass faulting virtual address to sendsig */ 570 frame->tf_err = eva; 571 572 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 573} 574#endif 575 576int 577trap_pfault(frame, usermode) 578 struct trapframe *frame; 579 int usermode; 580{ 581 vm_offset_t va; 582 struct vmspace *vm = NULL; 583 vm_map_t map = 0; 584 int rv = 0; 585 vm_prot_t ftype; 586 int eva; 587 struct proc *p = curproc; 588 589 eva = rcr2(); 590 va = trunc_page((vm_offset_t)eva); 591 592 if (va >= KERNBASE) { 593 /* 594 * Don't allow user-mode faults in kernel address space. 595 */ 596 if (usermode) 597 goto nogo; 598 599 map = kernel_map; 600 } else { 601 /* 602 * This is a fault on non-kernel virtual memory. 603 * vm is initialized above to NULL. If curproc is NULL 604 * or curproc->p_vmspace is NULL the fault is fatal. 605 */ 606 if (p != NULL) 607 vm = p->p_vmspace; 608 609 if (vm == NULL) 610 goto nogo; 611 612 map = &vm->vm_map; 613 } 614 615 if (frame->tf_err & PGEX_W) 616 ftype = VM_PROT_READ | VM_PROT_WRITE; 617 else 618 ftype = VM_PROT_READ; 619 620 if (map != kernel_map) { 621 vm_offset_t v; 622 623 /* 624 * Keep swapout from messing with us during this 625 * critical time. 626 */ 627 ++p->p_lock; 628 629 /* 630 * Grow the stack if necessary 631 */ 632 if ((caddr_t)va > vm->vm_maxsaddr 633 && (caddr_t)va < (caddr_t)USRSTACK) { 634 if (!grow(p, va)) { 635 rv = KERN_FAILURE; 636 --p->p_lock; 637 goto nogo; 638 } 639 } 640 641 /* 642 * Check if page table is mapped, if not, 643 * fault it first 644 */ 645 v = (vm_offset_t) vtopte(va); 646 647 /* Fault the pte only if needed: */ 648 if (*((int *)vtopte(v)) == 0) 649 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 650 651 pmap_use_pt( vm_map_pmap(map), va); 652 653 /* Fault in the user page: */ 654 rv = vm_fault(map, va, ftype, FALSE); 655 656 pmap_unuse_pt( vm_map_pmap(map), va); 657 658 --p->p_lock; 659 } else { 660 /* 661 * Since we know that kernel virtual address addresses 662 * always have pte pages mapped, we just have to fault 663 * the page. 664 */ 665 rv = vm_fault(map, va, ftype, FALSE); 666 } 667 668 if (rv == KERN_SUCCESS) 669 return (0); 670nogo: 671 if (!usermode) { 672 if (curpcb && curpcb->pcb_onfault) { 673 frame->tf_eip = (int)curpcb->pcb_onfault; 674 return (0); 675 } 676 trap_fatal(frame); 677 return (-1); 678 } 679 680 /* kludge to pass faulting virtual address to sendsig */ 681 frame->tf_err = eva; 682 683 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 684} 685 686static void 687trap_fatal(frame) 688 struct trapframe *frame; 689{ 690 int code, type, eva; 691 struct soft_segment_descriptor softseg; 692 693 code = frame->tf_err; 694 type = frame->tf_trapno; 695 eva = rcr2(); 696 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 697 698 if (type <= MAX_TRAP_MSG) 699 printf("\n\nFatal trap %d: %s while in %s mode\n", 700 type, trap_msg[type], 701 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 702 if (type == T_PAGEFLT) { 703 printf("fault virtual address = 0x%x\n", eva); 704 printf("fault code = %s %s, %s\n", 705 code & PGEX_U ? "user" : "supervisor", 706 code & PGEX_W ? "write" : "read", 707 code & PGEX_P ? "protection violation" : "page not present"); 708 } 709 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 710 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 711 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 712 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 713 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 714 printf("processor eflags = "); 715 if (frame->tf_eflags & PSL_T) 716 printf("trace/trap, "); 717 if (frame->tf_eflags & PSL_I) 718 printf("interrupt enabled, "); 719 if (frame->tf_eflags & PSL_NT) 720 printf("nested task, "); 721 if (frame->tf_eflags & PSL_RF) 722 printf("resume, "); 723 if (frame->tf_eflags & PSL_VM) 724 printf("vm86, "); 725 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 726 printf("current process = "); 727 if (curproc) { 728 printf("%lu (%s)\n", 729 (u_long)curproc->p_pid, curproc->p_comm ? 730 curproc->p_comm : ""); 731 } else { 732 printf("Idle\n"); 733 } 734 printf("interrupt mask = "); 735 if ((cpl & net_imask) == net_imask) 736 printf("net "); 737 if ((cpl & tty_imask) == tty_imask) 738 printf("tty "); 739 if ((cpl & bio_imask) == bio_imask) 740 printf("bio "); 741 if (cpl == 0) 742 printf("none"); 743 printf("\n"); 744 745#ifdef KDB 746 if (kdb_trap(&psl)) 747 return; 748#endif 749#ifdef DDB 750 if (kdb_trap (type, 0, frame)) 751 return; 752#endif 753 if (type <= MAX_TRAP_MSG) 754 panic(trap_msg[type]); 755 else 756 panic("unknown/reserved trap"); 757} 758 759/* 760 * Double fault handler. Called when a fault occurs while writing 761 * a frame for a trap/exception onto the stack. This usually occurs 762 * when the stack overflows (such is the case with infinite recursion, 763 * for example). 764 * 765 * XXX Note that the current PTD gets replaced by IdlePTD when the 766 * task switch occurs. This means that the stack that was active at 767 * the time of the double fault is not available at <kstack> unless 768 * the machine was idle when the double fault occurred. The downside 769 * of this is that "trace <ebp>" in ddb won't work. 770 */ 771void 772dblfault_handler() 773{ 774 struct pcb *pcb = curpcb; 775 776 if (pcb != NULL) { 777 printf("\nFatal double fault:\n"); 778 printf("eip = 0x%x\n", pcb->pcb_tss.tss_eip); 779 printf("esp = 0x%x\n", pcb->pcb_tss.tss_esp); 780 printf("ebp = 0x%x\n", pcb->pcb_tss.tss_ebp); 781 } 782 783 panic("double fault"); 784} 785 786/* 787 * Compensate for 386 brain damage (missing URKR). 788 * This is a little simpler than the pagefault handler in trap() because 789 * it the page tables have already been faulted in and high addresses 790 * are thrown out early for other reasons. 791 */ 792int trapwrite(addr) 793 unsigned addr; 794{ 795 struct proc *p; 796 vm_offset_t va, v; 797 struct vmspace *vm; 798 int rv; 799 800 va = trunc_page((vm_offset_t)addr); 801 /* 802 * XXX - MAX is END. Changed > to >= for temp. fix. 803 */ 804 if (va >= VM_MAXUSER_ADDRESS) 805 return (1); 806 807 p = curproc; 808 vm = p->p_vmspace; 809 810 ++p->p_lock; 811 812 if ((caddr_t)va >= vm->vm_maxsaddr 813 && (caddr_t)va < (caddr_t)USRSTACK) { 814 if (!grow(p, va)) { 815 --p->p_lock; 816 return (1); 817 } 818 } 819 820 v = trunc_page(vtopte(va)); 821 822 /* 823 * wire the pte page 824 */ 825 if (va < USRSTACK) { 826 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 827 } 828 829 /* 830 * fault the data page 831 */ 832 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 833 834 /* 835 * unwire the pte page 836 */ 837 if (va < USRSTACK) { 838 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 839 } 840 841 --p->p_lock; 842 843 if (rv != KERN_SUCCESS) 844 return 1; 845 846 return (0); 847} 848 849/* 850 * System call request from POSIX system call gate interface to kernel. 851 * Like trap(), argument is call by reference. 852 */ 853void 854syscall(frame) 855 struct trapframe frame; 856{ 857 caddr_t params; 858 int i; 859 struct sysent *callp; 860 struct proc *p = curproc; 861 u_quad_t sticks; 862 int error; 863 int args[8], rval[2]; 864 u_int code; 865 866 sticks = p->p_sticks; 867 if (ISPL(frame.tf_cs) != SEL_UPL) 868 panic("syscall"); 869 870 p->p_md.md_regs = (int *)&frame; 871 params = (caddr_t)frame.tf_esp + sizeof(int); 872 code = frame.tf_eax; 873 /* 874 * Need to check if this is a 32 bit or 64 bit syscall. 875 */ 876 if (code == SYS_syscall) { 877 /* 878 * Code is first argument, followed by actual args. 879 */ 880 code = fuword(params); 881 params += sizeof(int); 882 } else if (code == SYS___syscall) { 883 /* 884 * Like syscall, but code is a quad, so as to maintain 885 * quad alignment for the rest of the arguments. 886 */ 887 code = fuword(params); 888 params += sizeof(quad_t); 889 } 890 891 if (p->p_sysent->sv_mask) 892 code &= p->p_sysent->sv_mask; 893 894 if (code >= p->p_sysent->sv_size) 895 callp = &p->p_sysent->sv_table[0]; 896 else 897 callp = &p->p_sysent->sv_table[code]; 898 899 if ((i = callp->sy_narg * sizeof(int)) && 900 (error = copyin(params, (caddr_t)args, (u_int)i))) { 901#ifdef KTRACE 902 if (KTRPOINT(p, KTR_SYSCALL)) 903 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 904#endif 905 goto bad; 906 } 907#ifdef KTRACE 908 if (KTRPOINT(p, KTR_SYSCALL)) 909 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 910#endif 911 rval[0] = 0; 912 rval[1] = frame.tf_edx; 913 914 error = (*callp->sy_call)(p, args, rval); 915 916 switch (error) { 917 918 case 0: 919 /* 920 * Reinitialize proc pointer `p' as it may be different 921 * if this is a child returning from fork syscall. 922 */ 923 p = curproc; 924 frame.tf_eax = rval[0]; 925 frame.tf_edx = rval[1]; 926 frame.tf_eflags &= ~PSL_C; 927 break; 928 929 case ERESTART: 930 /* 931 * Reconstruct pc, assuming lcall $X,y is 7 bytes. 932 */ 933 frame.tf_eip -= 7; 934 break; 935 936 case EJUSTRETURN: 937 break; 938 939 default: 940bad: 941 if (p->p_sysent->sv_errsize) 942 if (error >= p->p_sysent->sv_errsize) 943 error = -1; /* XXX */ 944 else 945 error = p->p_sysent->sv_errtbl[error]; 946 frame.tf_eax = error; 947 frame.tf_eflags |= PSL_C; 948 break; 949 } 950 951 if (frame.tf_eflags & PSL_T) { 952 /* Traced syscall. */ 953 frame.tf_eflags &= ~PSL_T; 954 trapsignal(p, SIGTRAP, 0); 955 } 956 957 userret(p, &frame, sticks); 958 959#ifdef KTRACE 960 if (KTRPOINT(p, KTR_SYSRET)) 961 ktrsysret(p->p_tracep, code, error, rval[0]); 962#endif 963} 964 965#if defined(COMPAT_LINUX) || defined(LINUX) 966void 967linux_syscall(frame) 968 struct trapframe frame; 969{ 970 struct proc *p = curproc; 971 struct sysent *callp; 972 u_quad_t sticks; 973 int error; 974 int rval[2]; 975 u_int code; 976 struct linux_syscall_args { 977 int arg1; 978 int arg2; 979 int arg3; 980 int arg4; 981 int arg5; 982 } args; 983 984 args.arg1 = frame.tf_ebx; 985 args.arg2 = frame.tf_ecx; 986 args.arg3 = frame.tf_edx; 987 args.arg4 = frame.tf_esi; 988 args.arg5 = frame.tf_edi; 989 990 sticks = p->p_sticks; 991 if (ISPL(frame.tf_cs) != SEL_UPL) 992 panic("linux syscall"); 993 994 p->p_md.md_regs = (int *)&frame; 995 code = frame.tf_eax; 996 997 if (p->p_sysent->sv_mask) 998 code &= p->p_sysent->sv_mask; 999 1000 if (code >= p->p_sysent->sv_size) 1001 callp = &p->p_sysent->sv_table[0]; 1002 else 1003 callp = &p->p_sysent->sv_table[code]; 1004 1005#ifdef KTRACE 1006 if (KTRPOINT(p, KTR_SYSCALL)) 1007 ktrsyscall(p->p_tracep, code, callp->sy_narg, (int *)&args); 1008#endif 1009 1010 rval[0] = 0; 1011 1012 error = (*callp->sy_call)(p, &args, rval); 1013 1014 switch (error) { 1015 1016 case 0: 1017 /* 1018 * Reinitialize proc pointer `p' as it may be different 1019 * if this is a child returning from fork syscall. 1020 */ 1021 p = curproc; 1022 frame.tf_eax = rval[0]; 1023 frame.tf_eflags &= ~PSL_C; 1024 break; 1025 1026 case ERESTART: 1027 /* Reconstruct pc, subtract size of int 0x80 */ 1028 frame.tf_eip -= 2; 1029 break; 1030 1031 case EJUSTRETURN: 1032 break; 1033 1034 default: 1035 if (p->p_sysent->sv_errsize) 1036 if (error >= p->p_sysent->sv_errsize) 1037 error = -1; /* XXX */ 1038 else 1039 error = p->p_sysent->sv_errtbl[error]; 1040 frame.tf_eax = -error; 1041 frame.tf_eflags |= PSL_C; 1042 break; 1043 } 1044 1045 if (frame.tf_eflags & PSL_T) { 1046 /* Traced syscall. */ 1047 frame.tf_eflags &= ~PSL_T; 1048 trapsignal(p, SIGTRAP, 0); 1049 } 1050 1051 userret(p, &frame, sticks); 1052 1053#ifdef KTRACE 1054 if (KTRPOINT(p, KTR_SYSRET)) 1055 ktrsysret(p->p_tracep, code, error, rval[0]); 1056#endif 1057} 1058#endif /* COMPAT_LINUX || LINUX */ 1059