subr_syscall.c revision 1321
1234285Sdim/*- 2234285Sdim * Copyright (c) 1990 The Regents of the University of California. 3234285Sdim * All rights reserved. 4234285Sdim * 5234285Sdim * This code is derived from software contributed to Berkeley by 6234285Sdim * the University of Utah, and William Jolitz. 7234285Sdim * 8234285Sdim * Redistribution and use in source and binary forms, with or without 9234285Sdim * modification, are permitted provided that the following conditions 10234285Sdim * are met: 11234285Sdim * 1. Redistributions of source code must retain the above copyright 12234285Sdim * notice, this list of conditions and the following disclaimer. 13234285Sdim * 2. Redistributions in binary form must reproduce the above copyright 14234285Sdim * notice, this list of conditions and the following disclaimer in the 15234285Sdim * documentation and/or other materials provided with the distribution. 16234285Sdim * 3. All advertising materials mentioning features or use of this software 17234285Sdim * must display the following acknowledgement: 18234285Sdim * This product includes software developed by the University of 19239462Sdim * California, Berkeley and its contributors. 20234285Sdim * 4. Neither the name of the University nor the names of its contributors 21234285Sdim * may be used to endorse or promote products derived from this software 22234285Sdim * without specific prior written permission. 23234285Sdim * 24234285Sdim * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25234285Sdim * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26234285Sdim * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27234285Sdim * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28234285Sdim * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29234285Sdim * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30234285Sdim * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31234285Sdim * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32234285Sdim * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33234285Sdim * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34234285Sdim * SUCH DAMAGE. 35234285Sdim * 36234285Sdim * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 37234285Sdim * $Id: trap.c,v 1.20 1994/03/24 23:12:34 davidg Exp $ 38234285Sdim */ 39234285Sdim 40234285Sdim/* 41234285Sdim * 386 Trap and System call handleing 42234285Sdim */ 43234285Sdim 44234285Sdim#include "isa.h" 45234285Sdim#include "npx.h" 46234285Sdim#include "ddb.h" 47234285Sdim#include "machine/cpu.h" 48234285Sdim#include "machine/psl.h" 49234285Sdim#include "machine/reg.h" 50234285Sdim#include "machine/eflags.h" 51234285Sdim 52234285Sdim#include "param.h" 53234285Sdim#include "systm.h" 54234285Sdim#include "proc.h" 55234285Sdim#include "user.h" 56234285Sdim#include "acct.h" 57234285Sdim#include "kernel.h" 58234285Sdim#ifdef KTRACE 59234285Sdim#include "ktrace.h" 60234285Sdim#endif 61234285Sdim 62234285Sdim#include "vm/vm_param.h" 63234285Sdim#include "vm/pmap.h" 64234285Sdim#include "vm/vm_map.h" 65234285Sdim#include "vm/vm_user.h" 66234285Sdim#include "vm/vm_page.h" 67234285Sdim#include "sys/vmmeter.h" 68234285Sdim 69234285Sdim#include "machine/trap.h" 70234285Sdim 71234285Sdim#ifdef __GNUC__ 72234285Sdim 73234285Sdim/* 74234285Sdim * The "r" contraint could be "rm" except for fatal bugs in gas. As usual, 75234285Sdim * we omit the size from the mov instruction to avoid nonfatal bugs in gas. 76234285Sdim */ 77234285Sdim#define read_gs() ({ u_short gs; __asm("mov %%gs,%0" : "=r" (gs)); gs; }) 78234285Sdim#define write_gs(newgs) __asm("mov %0,%%gs" : : "r" ((u_short) newgs)) 79234285Sdim 80234285Sdim#else /* not __GNUC__ */ 81234285Sdim 82234285Sdimu_short read_gs __P((void)); 83234285Sdimvoid write_gs __P((/* promoted u_short */ int gs)); 84234285Sdim 85234285Sdim#endif /* __GNUC__ */ 86234285Sdim 87234285Sdimextern int grow(struct proc *,int); 88234285Sdim 89234285Sdimstruct sysent sysent[]; 90234285Sdimint nsysent; 91234285Sdim 92234285Sdim#define MAX_TRAP_MSG 27 93234285Sdimchar *trap_msg[] = { 94234285Sdim "reserved addressing fault", /* 0 T_RESADFLT */ 95234285Sdim "privileged instruction fault", /* 1 T_PRIVINFLT */ 96234285Sdim "reserved operand fault", /* 2 T_RESOPFLT */ 97234285Sdim "breakpoint instruction fault", /* 3 T_BPTFLT */ 98234285Sdim "", /* 4 unused */ 99234285Sdim "system call trap", /* 5 T_SYSCALL */ 100234285Sdim "arithmetic trap", /* 6 T_ARITHTRAP */ 101234285Sdim "system forced exception", /* 7 T_ASTFLT */ 102234285Sdim "segmentation (limit) fault", /* 8 T_SEGFLT */ 103234285Sdim "protection fault", /* 9 T_PROTFLT */ 104234285Sdim "trace trap", /* 10 T_TRCTRAP */ 105234285Sdim "", /* 11 unused */ 106234285Sdim "page fault", /* 12 T_PAGEFLT */ 107234285Sdim "page table fault", /* 13 T_TABLEFLT */ 108234285Sdim "alignment fault", /* 14 T_ALIGNFLT */ 109234285Sdim "kernel stack pointer not valid", /* 15 T_KSPNOTVAL */ 110234285Sdim "bus error", /* 16 T_BUSERR */ 111234285Sdim "kernel debugger fault", /* 17 T_KDBTRAP */ 112234285Sdim "integer divide fault", /* 18 T_DIVIDE */ 113234285Sdim "non-maskable interrupt trap", /* 19 T_NMI */ 114234285Sdim "overflow trap", /* 20 T_OFLOW */ 115234285Sdim "FPU bounds check fault", /* 21 T_BOUND */ 116234285Sdim "FPU device not available", /* 22 T_DNA */ 117234285Sdim "double fault", /* 23 T_DOUBLEFLT */ 118234285Sdim "FPU operand fetch fault", /* 24 T_FPOPFLT */ 119234285Sdim "invalid TSS fault", /* 25 T_TSSFLT */ 120234285Sdim "segment not present fault", /* 26 T_SEGNPFLT */ 121234285Sdim "stack fault", /* 27 T_STKFLT */ 122234285Sdim}; 123234285Sdim 124234285Sdim#define pde_v(v) (PTD[((v)>>PD_SHIFT)&1023].pd_v) 125234285Sdim 126234285Sdim/* 127234285Sdim * trap(frame): 128234285Sdim * Exception, fault, and trap interface to BSD kernel. This 129234285Sdim * common code is called from assembly language IDT gate entry 130234285Sdim * routines that prepare a suitable stack frame, and restore this 131234285Sdim * frame after the exception has been processed. Note that the 132234285Sdim * effect is as if the arguments were passed call by reference. 133234285Sdim */ 134234285Sdim 135234285Sdim/*ARGSUSED*/ 136234285Sdimvoid 137234285Sdimtrap(frame) 138234285Sdim struct trapframe frame; 139234285Sdim{ 140234285Sdim register int i; 141239462Sdim register struct proc *p = curproc; 142234285Sdim struct timeval syst; 143234285Sdim int ucode, type, code, eva, fault_type; 144234285Sdim 145234285Sdim frame.tf_eflags &= ~PSL_NT; /* clear nested trap XXX */ 146234285Sdim type = frame.tf_trapno; 147234285Sdim#if NDDB > 0 148234285Sdim if (curpcb && curpcb->pcb_onfault) { 149234285Sdim if (frame.tf_trapno == T_BPTFLT 150234285Sdim || frame.tf_trapno == T_TRCTRAP) 151234285Sdim if (kdb_trap (type, 0, &frame)) 152234285Sdim return; 153234285Sdim } 154234285Sdim#endif 155239462Sdim 156234285Sdim if (curpcb == 0 || curproc == 0) 157234285Sdim goto skiptoswitch; 158234285Sdim if (curpcb->pcb_onfault && frame.tf_trapno != T_PAGEFLT) { 159234285Sdim extern int _udatasel; 160234285Sdim 161234285Sdim if (read_gs() != (u_short) _udatasel) 162234285Sdim /* 163234285Sdim * Some user has corrupted %gs but we depend on it in 164234285Sdim * copyout() etc. Fix it up and retry. 165234285Sdim * 166234285Sdim * (We don't preserve %fs or %gs, so users can change 167234285Sdim * them to either _ucodesel, _udatasel or a not-present 168234285Sdim * selector, possibly ORed with 0 to 3, making them 169234285Sdim * volatile for other users. Not preserving them saves 170234285Sdim * time and doesn't lose functionality or open security 171234285Sdim * holes.) 172234285Sdim */ 173234285Sdim write_gs(_udatasel); 174234285Sdim else 175234285Sdimcopyfault: 176234285Sdim frame.tf_eip = (int)curpcb->pcb_onfault; 177234285Sdim return; 178234285Sdim } 179234285Sdim 180234285Sdim syst = p->p_stime; 181234285Sdim if (ISPL(frame.tf_cs) == SEL_UPL) { 182234285Sdim type |= T_USER; 183234285Sdim p->p_regs = (int *)&frame; 184 } 185 186skiptoswitch: 187 ucode=0; 188 eva = rcr2(); 189 code = frame.tf_err; 190 191 if ((type & ~T_USER) == T_PAGEFLT) 192 goto pfault; 193 194 switch (type) { 195 case T_SEGNPFLT|T_USER: 196 case T_STKFLT|T_USER: 197 case T_PROTFLT|T_USER: /* protection fault */ 198 ucode = code + BUS_SEGM_FAULT ; 199 i = SIGBUS; 200 break; 201 202 case T_PRIVINFLT|T_USER: /* privileged instruction fault */ 203 case T_RESADFLT|T_USER: /* reserved addressing fault */ 204 case T_RESOPFLT|T_USER: /* reserved operand fault */ 205 case T_FPOPFLT|T_USER: /* coprocessor operand fault */ 206 ucode = type &~ T_USER; 207 i = SIGILL; 208 break; 209 210 case T_ASTFLT|T_USER: /* Allow process switch */ 211 astoff(); 212 cnt.v_soft++; 213 if ((p->p_flag & SOWEUPC) && p->p_stats->p_prof.pr_scale) { 214 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 215 p->p_flag &= ~SOWEUPC; 216 } 217 goto out; 218 219 case T_DNA|T_USER: 220#if NNPX > 0 221 /* if a transparent fault (due to context switch "late") */ 222 if (npxdna()) return; 223#endif /* NNPX > 0 */ 224#ifdef MATH_EMULATE 225 i = math_emulate(&frame); 226 if (i == 0) return; 227#else /* MATH_EMULATE */ 228 panic("trap: math emulation necessary!"); 229#endif /* MATH_EMULATE */ 230 ucode = FPE_FPU_NP_TRAP; 231 break; 232 233 case T_BOUND|T_USER: 234 ucode = FPE_SUBRNG_TRAP; 235 i = SIGFPE; 236 break; 237 238 case T_OFLOW|T_USER: 239 ucode = FPE_INTOVF_TRAP; 240 i = SIGFPE; 241 break; 242 243 case T_DIVIDE|T_USER: 244 ucode = FPE_INTDIV_TRAP; 245 i = SIGFPE; 246 break; 247 248 case T_ARITHTRAP|T_USER: 249 ucode = code; 250 i = SIGFPE; 251 break; 252 253 pfault: 254 case T_PAGEFLT: /* allow page faults in kernel mode */ 255 case T_PAGEFLT|T_USER: /* page fault */ 256 { 257 vm_offset_t va; 258 struct vmspace *vm; 259 vm_map_t map = 0; 260 int rv = 0, oldflags; 261 vm_prot_t ftype; 262 unsigned v; 263 extern vm_map_t kernel_map; 264 265 va = trunc_page((vm_offset_t)eva); 266 267 /* 268 * Don't allow user-mode faults in kernel address space 269 */ 270 if ((type == (T_PAGEFLT|T_USER)) && (va >= KERNBASE)) { 271 goto nogo; 272 } 273 274 if ((p == 0) || (type == T_PAGEFLT && va >= KERNBASE)) { 275 vm = 0; 276 map = kernel_map; 277 } else { 278 vm = p->p_vmspace; 279 map = &vm->vm_map; 280 } 281 282 if (code & PGEX_W) 283 ftype = VM_PROT_READ | VM_PROT_WRITE; 284 else 285 ftype = VM_PROT_READ; 286 287 oldflags = p->p_flag; 288 if (map != kernel_map) { 289 vm_offset_t pa; 290 vm_offset_t v = (vm_offset_t) vtopte(va); 291 vm_page_t ptepg; 292 293 /* 294 * Keep swapout from messing with us during this 295 * critical time. 296 */ 297 p->p_flag |= SLOCK; 298 299 /* 300 * Grow the stack if necessary 301 */ 302 if ((caddr_t)va > vm->vm_maxsaddr 303 && (caddr_t)va < (caddr_t)USRSTACK) { 304 if (!grow(p, va)) { 305 rv = KERN_FAILURE; 306 p->p_flag &= ~SLOCK; 307 p->p_flag |= (oldflags & SLOCK); 308 goto nogo; 309 } 310 } 311 312 /* 313 * Check if page table is mapped, if not, 314 * fault it first 315 */ 316 317 /* Fault the pte only if needed: */ 318 *(volatile char *)v += 0; 319 320 ptepg = (vm_page_t) pmap_pte_vm_page(vm_map_pmap(map), v); 321 vm_page_hold(ptepg); 322 323 /* Fault in the user page: */ 324 rv = vm_fault(map, va, ftype, FALSE); 325 326 vm_page_unhold(ptepg); 327 328 /* 329 * page table pages don't need to be kept if they 330 * are not held 331 */ 332 if( ptepg->hold_count == 0 && ptepg->wire_count == 0) { 333 pmap_page_protect( VM_PAGE_TO_PHYS(ptepg), 334 VM_PROT_NONE); 335 if( ptepg->flags & PG_CLEAN) 336 vm_page_free(ptepg); 337 } 338 339 340 p->p_flag &= ~SLOCK; 341 p->p_flag |= (oldflags & SLOCK); 342 } else { 343 /* 344 * Since we know that kernel virtual address addresses 345 * always have pte pages mapped, we just have to fault 346 * the page. 347 */ 348 rv = vm_fault(map, va, ftype, FALSE); 349 } 350 351 if (rv == KERN_SUCCESS) { 352 if (type == T_PAGEFLT) 353 return; 354 goto out; 355 } 356nogo: 357 if (type == T_PAGEFLT) { 358 if (curpcb->pcb_onfault) 359 goto copyfault; 360 361 goto we_re_toast; 362 } 363 i = (rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV; 364 365 /* kludge to pass faulting virtual address to sendsig */ 366 ucode = type &~ T_USER; 367 frame.tf_err = eva; 368 369 break; 370 } 371 372#if NDDB == 0 373 case T_TRCTRAP: /* trace trap -- someone single stepping lcall's */ 374 frame.tf_eflags &= ~PSL_T; 375 376 /* Q: how do we turn it on again? */ 377 return; 378#endif 379 380 case T_BPTFLT|T_USER: /* bpt instruction fault */ 381 case T_TRCTRAP|T_USER: /* trace trap */ 382 frame.tf_eflags &= ~PSL_T; 383 i = SIGTRAP; 384 break; 385 386#if NISA > 0 387 case T_NMI: 388 case T_NMI|T_USER: 389#if NDDB > 0 390 /* NMI can be hooked up to a pushbutton for debugging */ 391 printf ("NMI ... going to debugger\n"); 392 if (kdb_trap (type, 0, &frame)) 393 return; 394#endif 395 /* machine/parity/power fail/"kitchen sink" faults */ 396 if (isa_nmi(code) == 0) return; 397 /* FALL THROUGH */ 398#endif 399 default: 400 we_re_toast: 401 402 fault_type = type & ~T_USER; 403 if (fault_type <= MAX_TRAP_MSG) 404 printf("\n\nFatal trap %d: %s while in %s mode\n", 405 fault_type, trap_msg[fault_type], 406 ISPL(frame.tf_cs) == SEL_UPL ? "user" : "kernel"); 407 if (fault_type == T_PAGEFLT) { 408 printf("fault virtual address = 0x%x\n", eva); 409 printf("fault code = %s %s, %s\n", 410 code & PGEX_U ? "user" : "supervisor", 411 code & PGEX_W ? "write" : "read", 412 code & PGEX_P ? "protection violation" : "page not present"); 413 } 414 printf("instruction pointer = 0x%x\n", frame.tf_eip); 415 printf("processor eflags = "); 416 if (frame.tf_eflags & EFL_TF) 417 printf("trace/trap, "); 418 if (frame.tf_eflags & EFL_IF) 419 printf("interrupt enabled, "); 420 if (frame.tf_eflags & EFL_NT) 421 printf("nested task, "); 422 if (frame.tf_eflags & EFL_RF) 423 printf("resume, "); 424 if (frame.tf_eflags & EFL_VM) 425 printf("vm86, "); 426 printf("IOPL = %d\n", (frame.tf_eflags & EFL_IOPL) >> 12); 427 printf("current process = "); 428 if (curproc) { 429 printf("%d (%s)\n", 430 curproc->p_pid, curproc->p_comm ? 431 curproc->p_comm : ""); 432 } else { 433 printf("Idle\n"); 434 } 435 printf("interrupt mask = "); 436 if ((cpl & net_imask) == net_imask) 437 printf("net "); 438 if ((cpl & tty_imask) == tty_imask) 439 printf("tty "); 440 if ((cpl & bio_imask) == bio_imask) 441 printf("bio "); 442 if (cpl == 0) 443 printf("none"); 444 printf("\n"); 445 446#ifdef KDB 447 if (kdb_trap(&psl)) 448 return; 449#endif 450#if NDDB > 0 451 if (kdb_trap (type, 0, &frame)) 452 return; 453#endif 454 if (fault_type <= MAX_TRAP_MSG) 455 panic(trap_msg[fault_type]); 456 else 457 panic("unknown/reserved trap"); 458 459 /* NOTREACHED */ 460 } 461 462 trapsignal(p, i, ucode); 463 if ((type & T_USER) == 0) 464 return; 465out: 466 while (i = CURSIG(p)) 467 psig(i); 468 p->p_pri = p->p_usrpri; 469 if (want_resched) { 470 int s; 471 /* 472 * Since we are curproc, clock will normally just change 473 * our priority without moving us from one queue to another 474 * (since the running process is not on a queue.) 475 * If that happened after we setrq ourselves but before we 476 * swtch()'ed, we might not be on the queue indicated by 477 * our priority. 478 */ 479 s = splclock(); 480 setrq(p); 481 p->p_stats->p_ru.ru_nivcsw++; 482 swtch(); 483 splx(s); 484 while (i = CURSIG(p)) 485 psig(i); 486 } 487 if (p->p_stats->p_prof.pr_scale) { 488 int ticks; 489 struct timeval *tv = &p->p_stime; 490 491 ticks = ((tv->tv_sec - syst.tv_sec) * 1000 + 492 (tv->tv_usec - syst.tv_usec) / 1000) / (tick / 1000); 493 if (ticks) { 494#ifdef PROFTIMER 495 extern int profscale; 496 addupc(frame.tf_eip, &p->p_stats->p_prof, 497 ticks * profscale); 498#else 499 addupc(frame.tf_eip, &p->p_stats->p_prof, ticks); 500#endif 501 } 502 } 503 curpri = p->p_pri; 504} 505 506/* 507 * Compensate for 386 brain damage (missing URKR). 508 * This is a little simpler than the pagefault handler in trap() because 509 * it the page tables have already been faulted in and high addresses 510 * are thrown out early for other reasons. 511 */ 512int trapwrite(addr) 513 unsigned addr; 514{ 515 struct proc *p; 516 vm_offset_t va, v; 517 struct vmspace *vm; 518 int oldflags; 519 int rv; 520 521 va = trunc_page((vm_offset_t)addr); 522 /* 523 * XXX - MAX is END. Changed > to >= for temp. fix. 524 */ 525 if (va >= VM_MAXUSER_ADDRESS) 526 return (1); 527 528 p = curproc; 529 vm = p->p_vmspace; 530 531 oldflags = p->p_flag; 532 p->p_flag |= SLOCK; 533 534 if ((caddr_t)va >= vm->vm_maxsaddr 535 && (caddr_t)va < (caddr_t)USRSTACK) { 536 if (!grow(p, va)) { 537 p->p_flag &= ~SLOCK; 538 p->p_flag |= (oldflags & SLOCK); 539 return (1); 540 } 541 } 542 543 v = trunc_page(vtopte(va)); 544 545 /* 546 * wire the pte page 547 */ 548 if (va < USRSTACK) { 549 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 550 } 551 552 /* 553 * fault the data page 554 */ 555 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 556 557 /* 558 * unwire the pte page 559 */ 560 if (va < USRSTACK) { 561 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 562 } 563 564 p->p_flag &= ~SLOCK; 565 p->p_flag |= (oldflags & SLOCK); 566 567 if (rv != KERN_SUCCESS) 568 return 1; 569 570 return (0); 571} 572 573/* 574 * syscall(frame): 575 * System call request from POSIX system call gate interface to kernel. 576 * Like trap(), argument is call by reference. 577 */ 578/*ARGSUSED*/ 579void 580syscall(frame) 581 volatile struct trapframe frame; 582{ 583 register int *locr0 = ((int *)&frame); 584 register caddr_t params; 585 register int i; 586 register struct sysent *callp; 587 register struct proc *p = curproc; 588 struct timeval syst; 589 int error, opc; 590 int args[8], rval[2]; 591 int code; 592 593#ifdef lint 594 r0 = 0; r0 = r0; r1 = 0; r1 = r1; 595#endif 596 syst = p->p_stime; 597 if (ISPL(frame.tf_cs) != SEL_UPL) 598 panic("syscall"); 599 600 code = frame.tf_eax; 601 p->p_regs = (int *)&frame; 602 params = (caddr_t)frame.tf_esp + sizeof (int) ; 603 604 /* 605 * Reconstruct pc, assuming lcall $X,y is 7 bytes, as it is always. 606 */ 607 opc = frame.tf_eip - 7; 608 if (code == 0) { 609 code = fuword(params); 610 params += sizeof (int); 611 } 612 if (code < 0 || code >= nsysent) 613 callp = &sysent[0]; 614 else 615 callp = &sysent[code]; 616 617 if ((i = callp->sy_narg * sizeof (int)) && 618 (error = copyin(params, (caddr_t)args, (u_int)i))) { 619 frame.tf_eax = error; 620 frame.tf_eflags |= PSL_C; /* carry bit */ 621#ifdef KTRACE 622 if (KTRPOINT(p, KTR_SYSCALL)) 623 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 624#endif 625 goto done; 626 } 627#ifdef KTRACE 628 if (KTRPOINT(p, KTR_SYSCALL)) 629 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 630#endif 631 rval[0] = 0; 632 rval[1] = frame.tf_edx; 633/*pg("%d. s %d\n", p->p_pid, code);*/ 634 error = (*callp->sy_call)(p, args, rval); 635 if (error == ERESTART) 636 frame.tf_eip = opc; 637 else if (error != EJUSTRETURN) { 638 if (error) { 639/*pg("error %d", error);*/ 640 frame.tf_eax = error; 641 frame.tf_eflags |= PSL_C; /* carry bit */ 642 } else { 643 frame.tf_eax = rval[0]; 644 frame.tf_edx = rval[1]; 645 frame.tf_eflags &= ~PSL_C; /* carry bit */ 646 } 647 } 648 /* else if (error == EJUSTRETURN) */ 649 /* nothing to do */ 650done: 651 /* 652 * Reinitialize proc pointer `p' as it may be different 653 * if this is a child returning from fork syscall. 654 */ 655 p = curproc; 656 while (i = CURSIG(p)) 657 psig(i); 658 p->p_pri = p->p_usrpri; 659 if (want_resched) { 660 int s; 661 /* 662 * Since we are curproc, clock will normally just change 663 * our priority without moving us from one queue to another 664 * (since the running process is not on a queue.) 665 * If that happened after we setrq ourselves but before we 666 * swtch()'ed, we might not be on the queue indicated by 667 * our priority. 668 */ 669 s = splclock(); 670 setrq(p); 671 p->p_stats->p_ru.ru_nivcsw++; 672 swtch(); 673 splx(s); 674 while (i = CURSIG(p)) 675 psig(i); 676 } 677 if (p->p_stats->p_prof.pr_scale) { 678 int ticks; 679 struct timeval *tv = &p->p_stime; 680 681 ticks = ((tv->tv_sec - syst.tv_sec) * 1000 + 682 (tv->tv_usec - syst.tv_usec) / 1000) / (tick / 1000); 683 if (ticks) { 684#ifdef PROFTIMER 685 extern int profscale; 686 addupc(frame.tf_eip, &p->p_stats->p_prof, 687 ticks * profscale); 688#else 689 addupc(frame.tf_eip, &p->p_stats->p_prof, ticks); 690#endif 691 } 692 } 693 curpri = p->p_pri; 694#ifdef KTRACE 695 if (KTRPOINT(p, KTR_SYSRET)) 696 ktrsysret(p->p_tracep, code, error, rval[0]); 697#endif 698#ifdef DIAGNOSTICx 699{ extern int _udatasel, _ucodesel; 700 if (frame.tf_ss != _udatasel) 701 printf("ss %x call %d\n", frame.tf_ss, code); 702 if ((frame.tf_cs&0xffff) != _ucodesel) 703 printf("cs %x call %d\n", frame.tf_cs, code); 704 if (frame.tf_eip > VM_MAXUSER_ADDRESS) { 705 printf("eip %x call %d\n", frame.tf_eip, code); 706 frame.tf_eip = 0; 707 } 708} 709#endif 710} 711