subr_syscall.c revision 27535
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 1.100 1997/06/22 16:03:37 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/kernel.h> 52#include <sys/syscall.h> 53#include <sys/sysent.h> 54#include <sys/vmmeter.h> 55#ifdef KTRACE 56#include <sys/ktrace.h> 57#endif 58 59#include <vm/vm.h> 60#include <vm/vm_param.h> 61#include <vm/vm_prot.h> 62#include <sys/lock.h> 63#include <vm/pmap.h> 64#include <vm/vm_kern.h> 65#include <vm/vm_map.h> 66#include <vm/vm_page.h> 67#include <vm/vm_extern.h> 68 69#include <sys/user.h> 70 71#include <machine/cpu.h> 72#include <machine/ipl.h> 73#include <machine/md_var.h> 74#include <machine/psl.h> 75#include <machine/../isa/intr_machdep.h> 76#include <machine/smp.h> 77 78#ifdef POWERFAIL_NMI 79#include <sys/syslog.h> 80#include <machine/clock.h> 81#endif 82 83#include "isa.h" 84#include "npx.h" 85 86extern struct i386tss common_tss; 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 28 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 "machine check trap", /* 28 T_MCHK */ 131}; 132 133static void userret __P((struct proc *p, struct trapframe *frame, 134 u_quad_t oticks)); 135 136static inline void 137userret(p, frame, oticks) 138 struct proc *p; 139 struct trapframe *frame; 140 u_quad_t oticks; 141{ 142 int sig, s; 143 144 while ((sig = CURSIG(p)) != 0) 145 postsig(sig); 146 p->p_priority = p->p_usrpri; 147 if (want_resched) { 148 /* 149 * Since we are curproc, clock will normally just change 150 * our priority without moving us from one queue to another 151 * (since the running process is not on a queue.) 152 * If that happened after we setrunqueue ourselves but before we 153 * mi_switch()'ed, we might not be on the queue indicated by 154 * our priority. 155 */ 156 s = splhigh(); 157 setrunqueue(p); 158 p->p_stats->p_ru.ru_nivcsw++; 159 mi_switch(); 160 splx(s); 161 while ((sig = CURSIG(p)) != 0) 162 postsig(sig); 163 } 164 /* 165 * Charge system time if profiling. 166 */ 167 if (p->p_flag & P_PROFIL) 168 addupc_task(p, frame->tf_eip, 169 (u_int)(p->p_sticks - oticks) * psratio); 170 171 curpriority = p->p_priority; 172} 173 174/* 175 * Exception, fault, and trap interface to the FreeBSD kernel. 176 * This common code is called from assembly language IDT gate entry 177 * routines that prepare a suitable stack frame, and restore this 178 * frame after the exception has been processed. 179 */ 180 181void 182trap(frame) 183 struct trapframe frame; 184{ 185 struct proc *p = curproc; 186 u_quad_t sticks = 0; 187 int i = 0, ucode = 0, type, code; 188#ifdef DEBUG 189 u_long eva; 190#endif 191 192 type = frame.tf_trapno; 193 code = frame.tf_err; 194 195 if (ISPL(frame.tf_cs) == SEL_UPL) { 196 /* user trap */ 197 198 sticks = p->p_sticks; 199 p->p_md.md_regs = &frame; 200 201 switch (type) { 202 case T_PRIVINFLT: /* privileged instruction fault */ 203 ucode = type; 204 i = SIGILL; 205 break; 206 207 case T_BPTFLT: /* bpt instruction fault */ 208 case T_TRCTRAP: /* trace trap */ 209 frame.tf_eflags &= ~PSL_T; 210 i = SIGTRAP; 211 break; 212 213 case T_ARITHTRAP: /* arithmetic trap */ 214 ucode = code; 215 i = SIGFPE; 216 break; 217 218 case T_ASTFLT: /* Allow process switch */ 219 astoff(); 220 cnt.v_soft++; 221 if (p->p_flag & P_OWEUPC) { 222 p->p_flag &= ~P_OWEUPC; 223 addupc_task(p, p->p_stats->p_prof.pr_addr, 224 p->p_stats->p_prof.pr_ticks); 225 } 226 goto out; 227 228 case T_PROTFLT: /* general protection fault */ 229 case T_SEGNPFLT: /* segment not present fault */ 230 case T_STKFLT: /* stack fault */ 231 case T_TSSFLT: /* invalid TSS fault */ 232 case T_DOUBLEFLT: /* double fault */ 233 default: 234 ucode = code + BUS_SEGM_FAULT ; 235 i = SIGBUS; 236 break; 237 238 case T_PAGEFLT: /* page fault */ 239 i = trap_pfault(&frame, TRUE); 240 if (i == -1) 241 return; 242 if (i == 0) 243 goto out; 244 245 ucode = T_PAGEFLT; 246 break; 247 248 case T_DIVIDE: /* integer divide fault */ 249 ucode = FPE_INTDIV_TRAP; 250 i = SIGFPE; 251 break; 252 253#if NISA > 0 254 case T_NMI: 255#ifdef POWERFAIL_NMI 256 goto handle_powerfail; 257#else /* !POWERFAIL_NMI */ 258#ifdef DDB 259 /* NMI can be hooked up to a pushbutton for debugging */ 260 printf ("NMI ... going to debugger\n"); 261 if (kdb_trap (type, 0, &frame)) 262 return; 263#endif /* DDB */ 264 /* machine/parity/power fail/"kitchen sink" faults */ 265 if (isa_nmi(code) == 0) return; 266 panic("NMI indicates hardware failure"); 267#endif /* POWERFAIL_NMI */ 268#endif /* NISA > 0 */ 269 270 case T_OFLOW: /* integer overflow fault */ 271 ucode = FPE_INTOVF_TRAP; 272 i = SIGFPE; 273 break; 274 275 case T_BOUND: /* bounds check fault */ 276 ucode = FPE_SUBRNG_TRAP; 277 i = SIGFPE; 278 break; 279 280 case T_DNA: 281#if NNPX > 0 282 /* if a transparent fault (due to context switch "late") */ 283 if (npxdna()) 284 return; 285#endif 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 /* 317 * The kernel is apparently using npx for copying. 318 * XXX this should be fatal unless the kernel has 319 * registered such use. 320 */ 321 if (npxdna()) 322 return; 323#endif 324 break; 325 326 case T_PROTFLT: /* general protection fault */ 327 case T_SEGNPFLT: /* segment not present fault */ 328 /* 329 * Invalid segment selectors and out of bounds 330 * %eip's and %esp's can be set up in user mode. 331 * This causes a fault in kernel mode when the 332 * kernel tries to return to user mode. We want 333 * to get this fault so that we can fix the 334 * problem here and not have to check all the 335 * selectors and pointers when the user changes 336 * them. 337 */ 338#define MAYBE_DORETI_FAULT(where, whereto) \ 339 do { \ 340 if (frame.tf_eip == (int)where) { \ 341 frame.tf_eip = (int)whereto; \ 342 return; \ 343 } \ 344 } while (0) 345 346 if (intr_nesting_level == 0) { 347 /* 348 * Invalid %fs's and %gs's can be created using 349 * procfs or PT_SETREGS or by invalidating the 350 * underlying LDT entry. This causes a fault 351 * in kernel mode when the kernel attempts to 352 * switch contexts. Lose the bad context 353 * (XXX) so that we can continue, and generate 354 * a signal. 355 */ 356 if (frame.tf_eip == (int)cpu_switch_load_fs) { 357 curpcb->pcb_fs = 0; 358 psignal(p, SIGBUS); 359 return; 360 } 361 if (frame.tf_eip == (int)cpu_switch_load_gs) { 362 curpcb->pcb_gs = 0; 363 psignal(p, SIGBUS); 364 return; 365 } 366 MAYBE_DORETI_FAULT(doreti_iret, 367 doreti_iret_fault); 368 MAYBE_DORETI_FAULT(doreti_popl_ds, 369 doreti_popl_ds_fault); 370 MAYBE_DORETI_FAULT(doreti_popl_es, 371 doreti_popl_es_fault); 372 if (curpcb && curpcb->pcb_onfault) { 373 frame.tf_eip = (int)curpcb->pcb_onfault; 374 return; 375 } 376 } 377 break; 378 379 case T_TSSFLT: 380 /* 381 * PSL_NT can be set in user mode and isn't cleared 382 * automatically when the kernel is entered. This 383 * causes a TSS fault when the kernel attempts to 384 * `iret' because the TSS link is uninitialized. We 385 * want to get this fault so that we can fix the 386 * problem here and not every time the kernel is 387 * entered. 388 */ 389 if (frame.tf_eflags & PSL_NT) { 390 frame.tf_eflags &= ~PSL_NT; 391 return; 392 } 393 break; 394 395 case T_TRCTRAP: /* trace trap */ 396 if (frame.tf_eip == (int)IDTVEC(syscall)) { 397 /* 398 * We've just entered system mode via the 399 * syscall lcall. Continue single stepping 400 * silently until the syscall handler has 401 * saved the flags. 402 */ 403 return; 404 } 405 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 406 /* 407 * The syscall handler has now saved the 408 * flags. Stop single stepping it. 409 */ 410 frame.tf_eflags &= ~PSL_T; 411 return; 412 } 413 /* 414 * Fall through. 415 */ 416 case T_BPTFLT: 417 /* 418 * If DDB is enabled, let it handle the debugger trap. 419 * Otherwise, debugger traps "can't happen". 420 */ 421#ifdef DDB 422 if (kdb_trap (type, 0, &frame)) 423 return; 424#endif 425 break; 426 427#if NISA > 0 428 case T_NMI: 429#ifdef POWERFAIL_NMI 430#ifndef TIMER_FREQ 431# define TIMER_FREQ 1193182 432#endif 433 handle_powerfail: 434 { 435 static unsigned lastalert = 0; 436 437 if(time.tv_sec - lastalert > 10) 438 { 439 log(LOG_WARNING, "NMI: power fail\n"); 440 sysbeep(TIMER_FREQ/880, hz); 441 lastalert = time.tv_sec; 442 } 443 return; 444 } 445#else /* !POWERFAIL_NMI */ 446#ifdef DDB 447 /* NMI can be hooked up to a pushbutton for debugging */ 448 printf ("NMI ... going to debugger\n"); 449 if (kdb_trap (type, 0, &frame)) 450 return; 451#endif /* DDB */ 452 /* machine/parity/power fail/"kitchen sink" faults */ 453 if (isa_nmi(code) == 0) return; 454 /* FALL THROUGH */ 455#endif /* POWERFAIL_NMI */ 456#endif /* NISA > 0 */ 457 } 458 459 trap_fatal(&frame); 460 return; 461 } 462 463 trapsignal(p, i, ucode); 464 465#ifdef DEBUG 466 eva = rcr2(); 467 if (type <= MAX_TRAP_MSG) { 468 uprintf("fatal process exception: %s", 469 trap_msg[type]); 470 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 471 uprintf(", fault VA = 0x%x", eva); 472 uprintf("\n"); 473 } 474#endif 475 476out: 477 userret(p, &frame, sticks); 478} 479 480#ifdef notyet 481/* 482 * This version doesn't allow a page fault to user space while 483 * in the kernel. The rest of the kernel needs to be made "safe" 484 * before this can be used. I think the only things remaining 485 * to be made safe are the iBCS2 code and the process tracing/ 486 * debugging code. 487 */ 488static int 489trap_pfault(frame, usermode) 490 struct trapframe *frame; 491 int usermode; 492{ 493 vm_offset_t va; 494 struct vmspace *vm = NULL; 495 vm_map_t map = 0; 496 int rv = 0; 497 vm_prot_t ftype; 498 int eva; 499 struct proc *p = curproc; 500 501 if (frame->tf_err & PGEX_W) 502 ftype = VM_PROT_READ | VM_PROT_WRITE; 503 else 504 ftype = VM_PROT_READ; 505 506 eva = rcr2(); 507 va = trunc_page((vm_offset_t)eva); 508 509 if (va < VM_MIN_KERNEL_ADDRESS) { 510 vm_offset_t v; 511 vm_page_t mpte; 512 513 if (p == NULL || 514 (!usermode && va < VM_MAXUSER_ADDRESS && 515 (intr_nesting_level != 0 || curpcb == NULL || 516 curpcb->pcb_onfault == NULL))) { 517 trap_fatal(frame); 518 return (-1); 519 } 520 521 /* 522 * This is a fault on non-kernel virtual memory. 523 * vm is initialized above to NULL. If curproc is NULL 524 * or curproc->p_vmspace is NULL the fault is fatal. 525 */ 526 vm = p->p_vmspace; 527 if (vm == NULL) 528 goto nogo; 529 530 map = &vm->vm_map; 531 532 /* 533 * Keep swapout from messing with us during this 534 * critical time. 535 */ 536 ++p->p_lock; 537 538 /* 539 * Grow the stack if necessary 540 */ 541 if ((caddr_t)va > vm->vm_maxsaddr 542 && (caddr_t)va < (caddr_t)USRSTACK) { 543 if (!grow(p, va)) { 544 rv = KERN_FAILURE; 545 --p->p_lock; 546 goto nogo; 547 } 548 } 549 550 /* Fault in the user page: */ 551 rv = vm_fault(map, va, ftype, 552 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 553 554 --p->p_lock; 555 } else { 556 /* 557 * Don't allow user-mode faults in kernel address space. 558 */ 559 if (usermode) 560 goto nogo; 561 562 /* 563 * Since we know that kernel virtual address addresses 564 * always have pte pages mapped, we just have to fault 565 * the page. 566 */ 567 rv = vm_fault(kernel_map, va, ftype, FALSE); 568 } 569 570 if (rv == KERN_SUCCESS) 571 return (0); 572nogo: 573 if (!usermode) { 574 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 575 frame->tf_eip = (int)curpcb->pcb_onfault; 576 return (0); 577 } 578 trap_fatal(frame); 579 return (-1); 580 } 581 582 /* kludge to pass faulting virtual address to sendsig */ 583 frame->tf_err = eva; 584 585 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 586} 587#endif 588 589int 590trap_pfault(frame, usermode) 591 struct trapframe *frame; 592 int usermode; 593{ 594 vm_offset_t va; 595 struct vmspace *vm = NULL; 596 vm_map_t map = 0; 597 int rv = 0; 598 vm_prot_t ftype; 599 int eva; 600 struct proc *p = curproc; 601 602 eva = rcr2(); 603 va = trunc_page((vm_offset_t)eva); 604 605 if (va >= KERNBASE) { 606 /* 607 * Don't allow user-mode faults in kernel address space. 608 */ 609 if (usermode) 610 goto nogo; 611 612 map = kernel_map; 613 } else { 614 /* 615 * This is a fault on non-kernel virtual memory. 616 * vm is initialized above to NULL. If curproc is NULL 617 * or curproc->p_vmspace is NULL the fault is fatal. 618 */ 619 if (p != NULL) 620 vm = p->p_vmspace; 621 622 if (vm == NULL) 623 goto nogo; 624 625 map = &vm->vm_map; 626 } 627 628 if (frame->tf_err & PGEX_W) 629 ftype = VM_PROT_READ | VM_PROT_WRITE; 630 else 631 ftype = VM_PROT_READ; 632 633 if (map != kernel_map) { 634 /* 635 * Keep swapout from messing with us during this 636 * critical time. 637 */ 638 ++p->p_lock; 639 640 /* 641 * Grow the stack if necessary 642 */ 643 if ((caddr_t)va > vm->vm_maxsaddr 644 && (caddr_t)va < (caddr_t)USRSTACK) { 645 if (!grow(p, va)) { 646 rv = KERN_FAILURE; 647 --p->p_lock; 648 goto nogo; 649 } 650 } 651 652 /* Fault in the user page: */ 653 rv = vm_fault(map, va, ftype, 654 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 655 656 --p->p_lock; 657 } else { 658 /* 659 * Don't have to worry about process locking or stacks in the kernel. 660 */ 661 rv = vm_fault(map, va, ftype, FALSE); 662 } 663 664 if (rv == KERN_SUCCESS) 665 return (0); 666nogo: 667 if (!usermode) { 668 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 669 frame->tf_eip = (int)curpcb->pcb_onfault; 670 return (0); 671 } 672 trap_fatal(frame); 673 return (-1); 674 } 675 676 /* kludge to pass faulting virtual address to sendsig */ 677 frame->tf_err = eva; 678 679 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 680} 681 682static void 683trap_fatal(frame) 684 struct trapframe *frame; 685{ 686 int code, type, eva, ss, esp; 687 struct soft_segment_descriptor softseg; 688 689 code = frame->tf_err; 690 type = frame->tf_trapno; 691 eva = rcr2(); 692 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 693 694 if (type <= MAX_TRAP_MSG) 695 printf("\n\nFatal trap %d: %s while in %s mode\n", 696 type, trap_msg[type], 697 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 698#ifdef SMP 699 printf("cpuid = %d\n", cpuid); 700#endif 701 if (type == T_PAGEFLT) { 702 printf("fault virtual address = 0x%x\n", eva); 703 printf("fault code = %s %s, %s\n", 704 code & PGEX_U ? "user" : "supervisor", 705 code & PGEX_W ? "write" : "read", 706 code & PGEX_P ? "protection violation" : "page not present"); 707 } 708 printf("instruction pointer = 0x%x:0x%x\n", 709 frame->tf_cs & 0xffff, frame->tf_eip); 710 if (ISPL(frame->tf_cs) == SEL_UPL) { 711 ss = frame->tf_ss & 0xffff; 712 esp = frame->tf_esp; 713 } else { 714 ss = GSEL(GDATA_SEL, SEL_KPL); 715 esp = (int)&frame->tf_esp; 716 } 717 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 718 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 719 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 720 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 721 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 722 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 723 softseg.ssd_gran); 724 printf("processor eflags = "); 725 if (frame->tf_eflags & PSL_T) 726 printf("trace trap, "); 727 if (frame->tf_eflags & PSL_I) 728 printf("interrupt enabled, "); 729 if (frame->tf_eflags & PSL_NT) 730 printf("nested task, "); 731 if (frame->tf_eflags & PSL_RF) 732 printf("resume, "); 733 if (frame->tf_eflags & PSL_VM) 734 printf("vm86, "); 735 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 736 printf("current process = "); 737 if (curproc) { 738 printf("%lu (%s)\n", 739 (u_long)curproc->p_pid, curproc->p_comm ? 740 curproc->p_comm : ""); 741 } else { 742 printf("Idle\n"); 743 } 744 printf("interrupt mask = "); 745 if ((cpl & net_imask) == net_imask) 746 printf("net "); 747 if ((cpl & tty_imask) == tty_imask) 748 printf("tty "); 749 if ((cpl & bio_imask) == bio_imask) 750 printf("bio "); 751 if (cpl == 0) 752 printf("none"); 753 printf("\n"); 754 755#ifdef KDB 756 if (kdb_trap(&psl)) 757 return; 758#endif 759#ifdef DDB 760 if (kdb_trap (type, 0, frame)) 761 return; 762#endif 763 printf("trap number = %d\n", type); 764 if (type <= MAX_TRAP_MSG) 765 panic(trap_msg[type]); 766 else 767 panic("unknown/reserved trap"); 768} 769 770/* 771 * Double fault handler. Called when a fault occurs while writing 772 * a frame for a trap/exception onto the stack. This usually occurs 773 * when the stack overflows (such is the case with infinite recursion, 774 * for example). 775 * 776 * XXX Note that the current PTD gets replaced by IdlePTD when the 777 * task switch occurs. This means that the stack that was active at 778 * the time of the double fault is not available at <kstack> unless 779 * the machine was idle when the double fault occurred. The downside 780 * of this is that "trace <ebp>" in ddb won't work. 781 */ 782void 783dblfault_handler() 784{ 785 printf("\nFatal double fault:\n"); 786 printf("eip = 0x%x\n", common_tss.tss_eip); 787 printf("esp = 0x%x\n", common_tss.tss_esp); 788 printf("ebp = 0x%x\n", common_tss.tss_ebp); 789#ifdef SMP 790 printf("cpuid = %d\n", cpuid); 791#endif 792 panic("double fault"); 793} 794 795/* 796 * Compensate for 386 brain damage (missing URKR). 797 * This is a little simpler than the pagefault handler in trap() because 798 * it the page tables have already been faulted in and high addresses 799 * are thrown out early for other reasons. 800 */ 801int trapwrite(addr) 802 unsigned addr; 803{ 804 struct proc *p; 805 vm_offset_t va; 806 struct vmspace *vm; 807 int rv; 808 809 va = trunc_page((vm_offset_t)addr); 810 /* 811 * XXX - MAX is END. Changed > to >= for temp. fix. 812 */ 813 if (va >= VM_MAXUSER_ADDRESS) 814 return (1); 815 816 p = curproc; 817 vm = p->p_vmspace; 818 819 ++p->p_lock; 820 821 if ((caddr_t)va >= vm->vm_maxsaddr 822 && (caddr_t)va < (caddr_t)USRSTACK) { 823 if (!grow(p, va)) { 824 --p->p_lock; 825 return (1); 826 } 827 } 828 829 /* 830 * fault the data page 831 */ 832 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, VM_FAULT_DIRTY); 833 834 --p->p_lock; 835 836 if (rv != KERN_SUCCESS) 837 return 1; 838 839 return (0); 840} 841 842/* 843 * System call request from POSIX system call gate interface to kernel. 844 * Like trap(), argument is call by reference. 845 */ 846void 847syscall(frame) 848 struct trapframe frame; 849{ 850 caddr_t params; 851 int i; 852 struct sysent *callp; 853 struct proc *p = curproc; 854 u_quad_t sticks; 855 int error; 856 int args[8], rval[2]; 857 u_int code; 858 859 sticks = p->p_sticks; 860 if (ISPL(frame.tf_cs) != SEL_UPL) 861 panic("syscall"); 862 863 p->p_md.md_regs = &frame; 864 params = (caddr_t)frame.tf_esp + sizeof(int); 865 code = frame.tf_eax; 866 if (p->p_sysent->sv_prepsyscall) { 867 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 868 } else { 869 /* 870 * Need to check if this is a 32 bit or 64 bit syscall. 871 */ 872 if (code == SYS_syscall) { 873 /* 874 * Code is first argument, followed by actual args. 875 */ 876 code = fuword(params); 877 params += sizeof(int); 878 } else if (code == SYS___syscall) { 879 /* 880 * Like syscall, but code is a quad, so as to maintain 881 * quad alignment for the rest of the arguments. 882 */ 883 code = fuword(params); 884 params += sizeof(quad_t); 885 } 886 } 887 888 if (p->p_sysent->sv_mask) 889 code &= p->p_sysent->sv_mask; 890 891 if (code >= p->p_sysent->sv_size) 892 callp = &p->p_sysent->sv_table[0]; 893 else 894 callp = &p->p_sysent->sv_table[code]; 895 896 if (params && (i = callp->sy_narg * sizeof(int)) && 897 (error = copyin(params, (caddr_t)args, (u_int)i))) { 898#ifdef KTRACE 899 if (KTRPOINT(p, KTR_SYSCALL)) 900 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 901#endif 902 goto bad; 903 } 904#ifdef KTRACE 905 if (KTRPOINT(p, KTR_SYSCALL)) 906 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 907#endif 908 rval[0] = 0; 909 rval[1] = frame.tf_edx; 910 911 error = (*callp->sy_call)(p, args, rval); 912 913 switch (error) { 914 915 case 0: 916 /* 917 * Reinitialize proc pointer `p' as it may be different 918 * if this is a child returning from fork syscall. 919 */ 920 p = curproc; 921 frame.tf_eax = rval[0]; 922 frame.tf_edx = rval[1]; 923 frame.tf_eflags &= ~PSL_C; 924 break; 925 926 case ERESTART: 927 /* 928 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 929 * int 0x80 is 2 bytes. We saved this in tf_err. 930 */ 931 frame.tf_eip -= frame.tf_err; 932 break; 933 934 case EJUSTRETURN: 935 break; 936 937 default: 938bad: 939 if (p->p_sysent->sv_errsize) 940 if (error >= p->p_sysent->sv_errsize) 941 error = -1; /* XXX */ 942 else 943 error = p->p_sysent->sv_errtbl[error]; 944 frame.tf_eax = error; 945 frame.tf_eflags |= PSL_C; 946 break; 947 } 948 949 if (frame.tf_eflags & PSL_T) { 950 /* Traced syscall. */ 951 frame.tf_eflags &= ~PSL_T; 952 trapsignal(p, SIGTRAP, 0); 953 } 954 955 userret(p, &frame, sticks); 956 957#ifdef KTRACE 958 if (KTRPOINT(p, KTR_SYSRET)) 959 ktrsysret(p->p_tracep, code, error, rval[0]); 960#endif 961} 962 963/* 964 * Simplified back end of syscall(), used when returning from fork() 965 * directly into user mode. 966 */ 967void 968fork_return(p, frame) 969 struct proc *p; 970 struct trapframe frame; 971{ 972 frame.tf_eax = 0; /* Child returns zero */ 973 frame.tf_eflags &= ~PSL_C; /* success */ 974 frame.tf_edx = 1; 975 976 userret(p, &frame, 0); 977#ifdef KTRACE 978 if (KTRPOINT(p, KTR_SYSRET)) 979 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 980#endif 981} 982