subr_syscall.c revision 26265
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 1.95 1997/05/07 20:08:53 peter 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 <sys/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#include <machine/smp.h> 80 81#ifdef POWERFAIL_NMI 82#include <sys/syslog.h> 83#include <machine/clock.h> 84#endif 85 86#include "isa.h" 87#include "npx.h" 88 89#ifdef SMP 90extern struct i386tss *SMPcommon_tss_ptr[]; 91#else 92extern struct i386tss common_tss; 93#endif 94 95int (*pmath_emulate) __P((struct trapframe *)); 96 97extern void trap __P((struct trapframe frame)); 98extern int trapwrite __P((unsigned addr)); 99extern void syscall __P((struct trapframe frame)); 100 101static int trap_pfault __P((struct trapframe *, int)); 102static void trap_fatal __P((struct trapframe *)); 103void dblfault_handler __P((void)); 104 105extern inthand_t IDTVEC(syscall); 106 107#define MAX_TRAP_MSG 28 108static char *trap_msg[] = { 109 "", /* 0 unused */ 110 "privileged instruction fault", /* 1 T_PRIVINFLT */ 111 "", /* 2 unused */ 112 "breakpoint instruction fault", /* 3 T_BPTFLT */ 113 "", /* 4 unused */ 114 "", /* 5 unused */ 115 "arithmetic trap", /* 6 T_ARITHTRAP */ 116 "system forced exception", /* 7 T_ASTFLT */ 117 "", /* 8 unused */ 118 "general protection fault", /* 9 T_PROTFLT */ 119 "trace trap", /* 10 T_TRCTRAP */ 120 "", /* 11 unused */ 121 "page fault", /* 12 T_PAGEFLT */ 122 "", /* 13 unused */ 123 "alignment fault", /* 14 T_ALIGNFLT */ 124 "", /* 15 unused */ 125 "", /* 16 unused */ 126 "", /* 17 unused */ 127 "integer divide fault", /* 18 T_DIVIDE */ 128 "non-maskable interrupt trap", /* 19 T_NMI */ 129 "overflow trap", /* 20 T_OFLOW */ 130 "FPU bounds check fault", /* 21 T_BOUND */ 131 "FPU device not available", /* 22 T_DNA */ 132 "double fault", /* 23 T_DOUBLEFLT */ 133 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 134 "invalid TSS fault", /* 25 T_TSSFLT */ 135 "segment not present fault", /* 26 T_SEGNPFLT */ 136 "stack fault", /* 27 T_STKFLT */ 137 "machine check trap", /* 28 T_MCHK */ 138}; 139 140static void userret __P((struct proc *p, struct trapframe *frame, 141 u_quad_t oticks)); 142 143static inline void 144userret(p, frame, oticks) 145 struct proc *p; 146 struct trapframe *frame; 147 u_quad_t oticks; 148{ 149 int sig, s; 150 151 while ((sig = CURSIG(p)) != 0) 152 postsig(sig); 153 p->p_priority = p->p_usrpri; 154 if (want_resched) { 155 /* 156 * Since we are curproc, clock will normally just change 157 * our priority without moving us from one queue to another 158 * (since the running process is not on a queue.) 159 * If that happened after we setrunqueue ourselves but before we 160 * mi_switch()'ed, we might not be on the queue indicated by 161 * our priority. 162 */ 163 s = splhigh(); 164 setrunqueue(p); 165 p->p_stats->p_ru.ru_nivcsw++; 166 mi_switch(); 167 splx(s); 168 while ((sig = CURSIG(p)) != 0) 169 postsig(sig); 170 } 171 /* 172 * Charge system time if profiling. 173 */ 174 if (p->p_flag & P_PROFIL) 175 addupc_task(p, frame->tf_eip, 176 (u_int)(p->p_sticks - oticks) * psratio); 177 178 curpriority = p->p_priority; 179} 180 181/* 182 * Exception, fault, and trap interface to the FreeBSD kernel. 183 * This common code is called from assembly language IDT gate entry 184 * routines that prepare a suitable stack frame, and restore this 185 * frame after the exception has been processed. 186 */ 187 188void 189trap(frame) 190 struct trapframe frame; 191{ 192 struct proc *p = curproc; 193 u_quad_t sticks = 0; 194 int i = 0, ucode = 0, type, code; 195#ifdef DEBUG 196 u_long eva; 197#endif 198 199 type = frame.tf_trapno; 200 code = frame.tf_err; 201 202 if (ISPL(frame.tf_cs) == SEL_UPL) { 203 /* user trap */ 204 205 sticks = p->p_sticks; 206 p->p_md.md_regs = &frame; 207 208 switch (type) { 209 case T_PRIVINFLT: /* privileged instruction fault */ 210 ucode = type; 211 i = SIGILL; 212 break; 213 214 case T_BPTFLT: /* bpt instruction fault */ 215 case T_TRCTRAP: /* trace trap */ 216 frame.tf_eflags &= ~PSL_T; 217 i = SIGTRAP; 218 break; 219 220 case T_ARITHTRAP: /* arithmetic trap */ 221 ucode = code; 222 i = SIGFPE; 223 break; 224 225 case T_ASTFLT: /* Allow process switch */ 226 astoff(); 227 cnt.v_soft++; 228 if (p->p_flag & P_OWEUPC) { 229 p->p_flag &= ~P_OWEUPC; 230 addupc_task(p, p->p_stats->p_prof.pr_addr, 231 p->p_stats->p_prof.pr_ticks); 232 } 233 goto out; 234 235 case T_PROTFLT: /* general protection fault */ 236 case T_SEGNPFLT: /* segment not present fault */ 237 case T_STKFLT: /* stack fault */ 238 case T_TSSFLT: /* invalid TSS fault */ 239 case T_DOUBLEFLT: /* double fault */ 240 default: 241 ucode = code + BUS_SEGM_FAULT ; 242 i = SIGBUS; 243 break; 244 245 case T_PAGEFLT: /* page fault */ 246 i = trap_pfault(&frame, TRUE); 247 if (i == -1) 248 return; 249 if (i == 0) 250 goto out; 251 252 ucode = T_PAGEFLT; 253 break; 254 255 case T_DIVIDE: /* integer divide fault */ 256 ucode = FPE_INTDIV_TRAP; 257 i = SIGFPE; 258 break; 259 260#if NISA > 0 261 case T_NMI: 262#ifdef POWERFAIL_NMI 263 goto handle_powerfail; 264#else /* !POWERFAIL_NMI */ 265#ifdef DDB 266 /* NMI can be hooked up to a pushbutton for debugging */ 267 printf ("NMI ... going to debugger\n"); 268 if (kdb_trap (type, 0, &frame)) 269 return; 270#endif /* DDB */ 271 /* machine/parity/power fail/"kitchen sink" faults */ 272 if (isa_nmi(code) == 0) return; 273 panic("NMI indicates hardware failure"); 274#endif /* POWERFAIL_NMI */ 275#endif /* NISA > 0 */ 276 277 case T_OFLOW: /* integer overflow fault */ 278 ucode = FPE_INTOVF_TRAP; 279 i = SIGFPE; 280 break; 281 282 case T_BOUND: /* bounds check fault */ 283 ucode = FPE_SUBRNG_TRAP; 284 i = SIGFPE; 285 break; 286 287 case T_DNA: 288#if NNPX > 0 289 /* if a transparent fault (due to context switch "late") */ 290 if (npxdna()) 291 return; 292#endif 293 if (!pmath_emulate) { 294 i = SIGFPE; 295 ucode = FPE_FPU_NP_TRAP; 296 break; 297 } 298 i = (*pmath_emulate)(&frame); 299 if (i == 0) { 300 if (!(frame.tf_eflags & PSL_T)) 301 return; 302 frame.tf_eflags &= ~PSL_T; 303 i = SIGTRAP; 304 } 305 /* else ucode = emulator_only_knows() XXX */ 306 break; 307 308 case T_FPOPFLT: /* FPU operand fetch fault */ 309 ucode = T_FPOPFLT; 310 i = SIGILL; 311 break; 312 } 313 } else { 314 /* kernel trap */ 315 316 switch (type) { 317 case T_PAGEFLT: /* page fault */ 318 (void) trap_pfault(&frame, FALSE); 319 return; 320 321 case T_DNA: 322#if NNPX > 0 323 /* 324 * The kernel is apparently using npx for copying. 325 * XXX this should be fatal unless the kernel has 326 * registered such use. 327 */ 328 if (npxdna()) 329 return; 330#endif 331 break; 332 333 case T_PROTFLT: /* general protection fault */ 334 case T_SEGNPFLT: /* segment not present fault */ 335 /* 336 * Invalid segment selectors and out of bounds 337 * %eip's and %esp's can be set up in user mode. 338 * This causes a fault in kernel mode when the 339 * kernel tries to return to user mode. We want 340 * to get this fault so that we can fix the 341 * problem here and not have to check all the 342 * selectors and pointers when the user changes 343 * them. 344 */ 345#define MAYBE_DORETI_FAULT(where, whereto) \ 346 do { \ 347 if (frame.tf_eip == (int)where) { \ 348 frame.tf_eip = (int)whereto; \ 349 return; \ 350 } \ 351 } while (0) 352 353 if (intr_nesting_level == 0) { 354 MAYBE_DORETI_FAULT(doreti_iret, 355 doreti_iret_fault); 356 MAYBE_DORETI_FAULT(doreti_popl_ds, 357 doreti_popl_ds_fault); 358 MAYBE_DORETI_FAULT(doreti_popl_es, 359 doreti_popl_es_fault); 360 if (curpcb && curpcb->pcb_onfault) { 361 frame.tf_eip = (int)curpcb->pcb_onfault; 362 return; 363 } 364 } 365 break; 366 367 case T_TSSFLT: 368 /* 369 * PSL_NT can be set in user mode and isn't cleared 370 * automatically when the kernel is entered. This 371 * causes a TSS fault when the kernel attempts to 372 * `iret' because the TSS link is uninitialized. We 373 * want to get this fault so that we can fix the 374 * problem here and not every time the kernel is 375 * entered. 376 */ 377 if (frame.tf_eflags & PSL_NT) { 378 frame.tf_eflags &= ~PSL_NT; 379 return; 380 } 381 break; 382 383 case T_TRCTRAP: /* trace trap */ 384 if (frame.tf_eip == (int)IDTVEC(syscall)) { 385 /* 386 * We've just entered system mode via the 387 * syscall lcall. Continue single stepping 388 * silently until the syscall handler has 389 * saved the flags. 390 */ 391 return; 392 } 393 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 394 /* 395 * The syscall handler has now saved the 396 * flags. Stop single stepping it. 397 */ 398 frame.tf_eflags &= ~PSL_T; 399 return; 400 } 401 /* 402 * Fall through. 403 */ 404 case T_BPTFLT: 405 /* 406 * If DDB is enabled, let it handle the debugger trap. 407 * Otherwise, debugger traps "can't happen". 408 */ 409#ifdef DDB 410 if (kdb_trap (type, 0, &frame)) 411 return; 412#endif 413 break; 414 415#if NISA > 0 416 case T_NMI: 417#ifdef POWERFAIL_NMI 418#ifndef TIMER_FREQ 419# define TIMER_FREQ 1193182 420#endif 421 handle_powerfail: 422 { 423 static unsigned lastalert = 0; 424 425 if(time.tv_sec - lastalert > 10) 426 { 427 log(LOG_WARNING, "NMI: power fail\n"); 428 sysbeep(TIMER_FREQ/880, hz); 429 lastalert = time.tv_sec; 430 } 431 return; 432 } 433#else /* !POWERFAIL_NMI */ 434#ifdef DDB 435 /* NMI can be hooked up to a pushbutton for debugging */ 436 printf ("NMI ... going to debugger\n"); 437 if (kdb_trap (type, 0, &frame)) 438 return; 439#endif /* DDB */ 440 /* machine/parity/power fail/"kitchen sink" faults */ 441 if (isa_nmi(code) == 0) return; 442 /* FALL THROUGH */ 443#endif /* POWERFAIL_NMI */ 444#endif /* NISA > 0 */ 445 } 446 447 trap_fatal(&frame); 448 return; 449 } 450 451 trapsignal(p, i, ucode); 452 453#ifdef DEBUG 454 eva = rcr2(); 455 if (type <= MAX_TRAP_MSG) { 456 uprintf("fatal process exception: %s", 457 trap_msg[type]); 458 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 459 uprintf(", fault VA = 0x%x", eva); 460 uprintf("\n"); 461 } 462#endif 463 464out: 465 userret(p, &frame, sticks); 466} 467 468#ifdef notyet 469/* 470 * This version doesn't allow a page fault to user space while 471 * in the kernel. The rest of the kernel needs to be made "safe" 472 * before this can be used. I think the only things remaining 473 * to be made safe are the iBCS2 code and the process tracing/ 474 * debugging code. 475 */ 476static int 477trap_pfault(frame, usermode) 478 struct trapframe *frame; 479 int usermode; 480{ 481 vm_offset_t va; 482 struct vmspace *vm = NULL; 483 vm_map_t map = 0; 484 int rv = 0; 485 vm_prot_t ftype; 486 int eva; 487 struct proc *p = curproc; 488 489 if (frame->tf_err & PGEX_W) 490 ftype = VM_PROT_READ | VM_PROT_WRITE; 491 else 492 ftype = VM_PROT_READ; 493 494 eva = rcr2(); 495 va = trunc_page((vm_offset_t)eva); 496 497 if (va < VM_MIN_KERNEL_ADDRESS) { 498 vm_offset_t v; 499 vm_page_t mpte; 500 501 if (p == NULL || 502 (!usermode && va < VM_MAXUSER_ADDRESS && 503 (intr_nesting_level != 0 || curpcb == NULL || 504 curpcb->pcb_onfault == NULL))) { 505 trap_fatal(frame); 506 return (-1); 507 } 508 509 /* 510 * This is a fault on non-kernel virtual memory. 511 * vm is initialized above to NULL. If curproc is NULL 512 * or curproc->p_vmspace is NULL the fault is fatal. 513 */ 514 vm = p->p_vmspace; 515 if (vm == NULL) 516 goto nogo; 517 518 map = &vm->vm_map; 519 520 /* 521 * Keep swapout from messing with us during this 522 * critical time. 523 */ 524 ++p->p_lock; 525 526 /* 527 * Grow the stack if necessary 528 */ 529 if ((caddr_t)va > vm->vm_maxsaddr 530 && (caddr_t)va < (caddr_t)USRSTACK) { 531 if (!grow(p, va)) { 532 rv = KERN_FAILURE; 533 --p->p_lock; 534 goto nogo; 535 } 536 } 537 538 /* Fault in the user page: */ 539 rv = vm_fault(map, va, ftype, 540 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 541 542 --p->p_lock; 543 } else { 544 /* 545 * Don't allow user-mode faults in kernel address space. 546 */ 547 if (usermode) 548 goto nogo; 549 550 /* 551 * Since we know that kernel virtual address addresses 552 * always have pte pages mapped, we just have to fault 553 * the page. 554 */ 555 rv = vm_fault(kernel_map, va, ftype, FALSE); 556 } 557 558 if (rv == KERN_SUCCESS) 559 return (0); 560nogo: 561 if (!usermode) { 562 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 563 frame->tf_eip = (int)curpcb->pcb_onfault; 564 return (0); 565 } 566 trap_fatal(frame); 567 return (-1); 568 } 569 570 /* kludge to pass faulting virtual address to sendsig */ 571 frame->tf_err = eva; 572 573 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 574} 575#endif 576 577int 578trap_pfault(frame, usermode) 579 struct trapframe *frame; 580 int usermode; 581{ 582 vm_offset_t va; 583 struct vmspace *vm = NULL; 584 vm_map_t map = 0; 585 int rv = 0; 586 vm_prot_t ftype; 587 int eva; 588 struct proc *p = curproc; 589 590 eva = rcr2(); 591 va = trunc_page((vm_offset_t)eva); 592 593 if (va >= KERNBASE) { 594 /* 595 * Don't allow user-mode faults in kernel address space. 596 */ 597 if (usermode) 598 goto nogo; 599 600 map = kernel_map; 601 } else { 602 /* 603 * This is a fault on non-kernel virtual memory. 604 * vm is initialized above to NULL. If curproc is NULL 605 * or curproc->p_vmspace is NULL the fault is fatal. 606 */ 607 if (p != NULL) 608 vm = p->p_vmspace; 609 610 if (vm == NULL) 611 goto nogo; 612 613 map = &vm->vm_map; 614 } 615 616 if (frame->tf_err & PGEX_W) 617 ftype = VM_PROT_READ | VM_PROT_WRITE; 618 else 619 ftype = VM_PROT_READ; 620 621 if (map != kernel_map) { 622 /* 623 * Keep swapout from messing with us during this 624 * critical time. 625 */ 626 ++p->p_lock; 627 628 /* 629 * Grow the stack if necessary 630 */ 631 if ((caddr_t)va > vm->vm_maxsaddr 632 && (caddr_t)va < (caddr_t)USRSTACK) { 633 if (!grow(p, va)) { 634 rv = KERN_FAILURE; 635 --p->p_lock; 636 goto nogo; 637 } 638 } 639 640 /* Fault in the user page: */ 641 rv = vm_fault(map, va, ftype, 642 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 643 644 --p->p_lock; 645 } else { 646 /* 647 * Don't have to worry about process locking or stacks in the kernel. 648 */ 649 rv = vm_fault(map, va, ftype, FALSE); 650 } 651 652 if (rv == KERN_SUCCESS) 653 return (0); 654nogo: 655 if (!usermode) { 656 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 657 frame->tf_eip = (int)curpcb->pcb_onfault; 658 return (0); 659 } 660 trap_fatal(frame); 661 return (-1); 662 } 663 664 /* kludge to pass faulting virtual address to sendsig */ 665 frame->tf_err = eva; 666 667 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 668} 669 670static void 671trap_fatal(frame) 672 struct trapframe *frame; 673{ 674 int code, type, eva, ss, esp; 675 struct soft_segment_descriptor softseg; 676 677 code = frame->tf_err; 678 type = frame->tf_trapno; 679 eva = rcr2(); 680 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 681 682 if (type <= MAX_TRAP_MSG) 683 printf("\n\nFatal trap %d: %s while in %s mode\n", 684 type, trap_msg[type], 685 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 686#ifdef SMP 687 printf("cpunumber = %d\n", cpunumber()); 688#endif 689 if (type == T_PAGEFLT) { 690 printf("fault virtual address = 0x%x\n", eva); 691 printf("fault code = %s %s, %s\n", 692 code & PGEX_U ? "user" : "supervisor", 693 code & PGEX_W ? "write" : "read", 694 code & PGEX_P ? "protection violation" : "page not present"); 695 } 696 printf("instruction pointer = 0x%x:0x%x\n", 697 frame->tf_cs & 0xffff, frame->tf_eip); 698 if (ISPL(frame->tf_cs) == SEL_UPL) { 699 ss = frame->tf_ss & 0xffff; 700 esp = frame->tf_esp; 701 } else { 702 ss = GSEL(GDATA_SEL, SEL_KPL); 703 esp = (int)&frame->tf_esp; 704 } 705 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 706 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 707 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 708 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 709 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 710 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 711 softseg.ssd_gran); 712 printf("processor eflags = "); 713 if (frame->tf_eflags & PSL_T) 714 printf("trace trap, "); 715 if (frame->tf_eflags & PSL_I) 716 printf("interrupt enabled, "); 717 if (frame->tf_eflags & PSL_NT) 718 printf("nested task, "); 719 if (frame->tf_eflags & PSL_RF) 720 printf("resume, "); 721 if (frame->tf_eflags & PSL_VM) 722 printf("vm86, "); 723 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 724 printf("current process = "); 725 if (curproc) { 726 printf("%lu (%s)\n", 727 (u_long)curproc->p_pid, curproc->p_comm ? 728 curproc->p_comm : ""); 729 } else { 730 printf("Idle\n"); 731 } 732 printf("interrupt mask = "); 733 if ((cpl & net_imask) == net_imask) 734 printf("net "); 735 if ((cpl & tty_imask) == tty_imask) 736 printf("tty "); 737 if ((cpl & bio_imask) == bio_imask) 738 printf("bio "); 739 if (cpl == 0) 740 printf("none"); 741 printf("\n"); 742 743#ifdef KDB 744 if (kdb_trap(&psl)) 745 return; 746#endif 747#ifdef DDB 748 if (kdb_trap (type, 0, frame)) 749 return; 750#endif 751 printf("trap number = %d\n", type); 752 if (type <= MAX_TRAP_MSG) 753 panic(trap_msg[type]); 754 else 755 panic("unknown/reserved trap"); 756} 757 758/* 759 * Double fault handler. Called when a fault occurs while writing 760 * a frame for a trap/exception onto the stack. This usually occurs 761 * when the stack overflows (such is the case with infinite recursion, 762 * for example). 763 * 764 * XXX Note that the current PTD gets replaced by IdlePTD when the 765 * task switch occurs. This means that the stack that was active at 766 * the time of the double fault is not available at <kstack> unless 767 * the machine was idle when the double fault occurred. The downside 768 * of this is that "trace <ebp>" in ddb won't work. 769 */ 770void 771dblfault_handler() 772{ 773#ifdef SMP 774 int x = cpunumber(); 775#endif 776 777 printf("\nFatal double fault:\n"); 778#ifdef SMP 779 printf("eip = 0x%x\n", SMPcommon_tss_ptr[x]->tss_eip); 780 printf("esp = 0x%x\n", SMPcommon_tss_ptr[x]->tss_esp); 781 printf("ebp = 0x%x\n", SMPcommon_tss_ptr[x]->tss_ebp); 782#else 783 printf("eip = 0x%x\n", common_tss.tss_eip); 784 printf("esp = 0x%x\n", common_tss.tss_esp); 785 printf("ebp = 0x%x\n", common_tss.tss_ebp); 786#endif 787 panic("double fault"); 788} 789 790/* 791 * Compensate for 386 brain damage (missing URKR). 792 * This is a little simpler than the pagefault handler in trap() because 793 * it the page tables have already been faulted in and high addresses 794 * are thrown out early for other reasons. 795 */ 796int trapwrite(addr) 797 unsigned addr; 798{ 799 struct proc *p; 800 vm_offset_t va; 801 struct vmspace *vm; 802 int rv; 803 804 va = trunc_page((vm_offset_t)addr); 805 /* 806 * XXX - MAX is END. Changed > to >= for temp. fix. 807 */ 808 if (va >= VM_MAXUSER_ADDRESS) 809 return (1); 810 811 p = curproc; 812 vm = p->p_vmspace; 813 814 ++p->p_lock; 815 816 if ((caddr_t)va >= vm->vm_maxsaddr 817 && (caddr_t)va < (caddr_t)USRSTACK) { 818 if (!grow(p, va)) { 819 --p->p_lock; 820 return (1); 821 } 822 } 823 824 /* 825 * fault the data page 826 */ 827 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, VM_FAULT_DIRTY); 828 829 --p->p_lock; 830 831 if (rv != KERN_SUCCESS) 832 return 1; 833 834 return (0); 835} 836 837/* 838 * System call request from POSIX system call gate interface to kernel. 839 * Like trap(), argument is call by reference. 840 */ 841void 842syscall(frame) 843 struct trapframe frame; 844{ 845 caddr_t params; 846 int i; 847 struct sysent *callp; 848 struct proc *p = curproc; 849 u_quad_t sticks; 850 int error; 851 int args[8], rval[2]; 852 u_int code; 853 854 sticks = p->p_sticks; 855 if (ISPL(frame.tf_cs) != SEL_UPL) 856 panic("syscall"); 857 858 p->p_md.md_regs = &frame; 859 params = (caddr_t)frame.tf_esp + sizeof(int); 860 code = frame.tf_eax; 861 if (p->p_sysent->sv_prepsyscall) { 862 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 863 } else { 864 /* 865 * Need to check if this is a 32 bit or 64 bit syscall. 866 */ 867 if (code == SYS_syscall) { 868 /* 869 * Code is first argument, followed by actual args. 870 */ 871 code = fuword(params); 872 params += sizeof(int); 873 } else if (code == SYS___syscall) { 874 /* 875 * Like syscall, but code is a quad, so as to maintain 876 * quad alignment for the rest of the arguments. 877 */ 878 code = fuword(params); 879 params += sizeof(quad_t); 880 } 881 } 882 883 if (p->p_sysent->sv_mask) 884 code &= p->p_sysent->sv_mask; 885 886 if (code >= p->p_sysent->sv_size) 887 callp = &p->p_sysent->sv_table[0]; 888 else 889 callp = &p->p_sysent->sv_table[code]; 890 891 if (params && (i = callp->sy_narg * sizeof(int)) && 892 (error = copyin(params, (caddr_t)args, (u_int)i))) { 893#ifdef KTRACE 894 if (KTRPOINT(p, KTR_SYSCALL)) 895 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 896#endif 897 goto bad; 898 } 899#ifdef KTRACE 900 if (KTRPOINT(p, KTR_SYSCALL)) 901 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 902#endif 903 rval[0] = 0; 904 rval[1] = frame.tf_edx; 905 906 error = (*callp->sy_call)(p, args, rval); 907 908 switch (error) { 909 910 case 0: 911 /* 912 * Reinitialize proc pointer `p' as it may be different 913 * if this is a child returning from fork syscall. 914 */ 915 p = curproc; 916 frame.tf_eax = rval[0]; 917 frame.tf_edx = rval[1]; 918 frame.tf_eflags &= ~PSL_C; 919 break; 920 921 case ERESTART: 922 /* 923 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 924 * int 0x80 is 2 bytes. We saved this in tf_err. 925 */ 926 frame.tf_eip -= frame.tf_err; 927 break; 928 929 case EJUSTRETURN: 930 break; 931 932 default: 933bad: 934 if (p->p_sysent->sv_errsize) 935 if (error >= p->p_sysent->sv_errsize) 936 error = -1; /* XXX */ 937 else 938 error = p->p_sysent->sv_errtbl[error]; 939 frame.tf_eax = error; 940 frame.tf_eflags |= PSL_C; 941 break; 942 } 943 944 if (frame.tf_eflags & PSL_T) { 945 /* Traced syscall. */ 946 frame.tf_eflags &= ~PSL_T; 947 trapsignal(p, SIGTRAP, 0); 948 } 949 950 userret(p, &frame, sticks); 951 952#ifdef KTRACE 953 if (KTRPOINT(p, KTR_SYSRET)) 954 ktrsysret(p->p_tracep, code, error, rval[0]); 955#endif 956} 957 958/* 959 * Simplified back end of syscall(), used when returning from fork() 960 * directly into user mode. 961 */ 962void 963fork_return(p, frame) 964 struct proc *p; 965 struct trapframe frame; 966{ 967 frame.tf_eax = 0; /* Child returns zero */ 968 frame.tf_eflags &= ~PSL_C; /* success */ 969 frame.tf_edx = 1; 970 971 userret(p, &frame, 0); 972#ifdef KTRACE 973 if (KTRPOINT(p, KTR_SYSRET)) 974 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 975#endif 976} 977