linux_pipe.c revision 1.16
1/* $NetBSD: linux_pipe.c,v 1.16 1995/08/21 23:15:51 fvdl Exp $ */ 2 3/* 4 * Copyright (c) 1995 Frank van der Linden 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed for the NetBSD Project 18 * by Frank van der Linden 19 * 4. The name of the author may not be used to endorse or promote products 20 * derived from this software without specific prior written permission 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 23 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 24 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 25 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 27 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 31 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34/* 35 * Linux compatibility module. Try to deal with various Linux system calls. 36 */ 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/namei.h> 41#include <sys/proc.h> 42#include <sys/dir.h> 43#include <sys/file.h> 44#include <sys/stat.h> 45#include <sys/filedesc.h> 46#include <sys/ioctl.h> 47#include <sys/kernel.h> 48#include <sys/malloc.h> 49#include <sys/mbuf.h> 50#include <sys/mman.h> 51#include <sys/mount.h> 52#include <sys/ptrace.h> 53#include <sys/resource.h> 54#include <sys/resourcevar.h> 55#include <sys/signal.h> 56#include <sys/signalvar.h> 57#include <sys/socket.h> 58#include <sys/time.h> 59#include <sys/times.h> 60#include <sys/vnode.h> 61#include <sys/uio.h> 62#include <sys/wait.h> 63#include <sys/utsname.h> 64#include <sys/unistd.h> 65 66#include <sys/syscallargs.h> 67 68#include <vm/vm.h> 69#include <vm/vm_param.h> 70 71#include <compat/linux/linux_types.h> 72#include <compat/linux/linux_fcntl.h> 73#include <compat/linux/linux_mmap.h> 74#include <compat/linux/linux_signal.h> 75#include <compat/linux/linux_syscallargs.h> 76#include <compat/linux/linux_util.h> 77#include <compat/linux/linux_dirent.h> 78 79/* 80 * The information on a terminated (or stopped) process needs 81 * to be converted in order for Linux binaries to get a valid signal 82 * number out of it. 83 */ 84static int 85bsd_to_linux_wstat(status) 86 int *status; 87{ 88 if (WIFSIGNALED(*status)) 89 *status = (*status & ~0177) | 90 bsd_to_linux_sig[WTERMSIG(*status)]; 91 else if (WIFSTOPPED(*status)) 92 *status = (*status & ~0xff00) | 93 (bsd_to_linux_sig[WSTOPSIG(*status)] << 8); 94} 95 96/* 97 * waitpid(2). Passed on to the NetBSD call, surrounded by code to 98 * reserve some space for a NetBSD-style wait status, and converting 99 * it to what Linux wants. 100 */ 101int 102linux_waitpid(p, uap, retval) 103 struct proc *p; 104 struct linux_waitpid_args /* { 105 syscallarg(int) pid; 106 syscallarg(int *) status; 107 syscallarg(int) options; 108 } */ *uap; 109 register_t *retval; 110{ 111 struct wait4_args w4a; 112 int error, *status, tstat; 113 caddr_t sg; 114 115 if (SCARG(uap, status) != NULL) { 116 sg = stackgap_init(p->p_emul); 117 status = (int *) stackgap_alloc(&sg, sizeof status); 118 } else 119 status = NULL; 120 121 SCARG(&w4a, pid) = SCARG(uap, pid); 122 SCARG(&w4a, status) = status; 123 SCARG(&w4a, options) = SCARG(uap, options); 124 SCARG(&w4a, rusage) = NULL; 125 126 if ((error = wait4(p, &w4a, retval))) 127 return error; 128 129 if (status != NULL) { 130 if ((error = copyin(status, &tstat, sizeof tstat))) 131 return error; 132 133 bsd_to_linux_wstat(&tstat); 134 135 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 136 } 137 138 return 0; 139} 140 141/* 142 * This is very much the same as waitpid() 143 */ 144int 145linux_wait4(p, uap, retval) 146 struct proc *p; 147 struct linux_wait4_args /* { 148 syscallarg(int) pid; 149 syscallarg(int *) status; 150 syscallarg(int) options; 151 syscallarg(struct rusage *) rusage; 152 } */ *uap; 153 register_t *retval; 154{ 155 struct wait4_args w4a; 156 int error, *status, tstat; 157 caddr_t sg; 158 159 if (SCARG(uap, status) != NULL) { 160 sg = stackgap_init(p->p_emul); 161 status = (int *) stackgap_alloc(&sg, sizeof status); 162 } else 163 status = NULL; 164 165 SCARG(&w4a, pid) = SCARG(uap, pid); 166 SCARG(&w4a, status) = status; 167 SCARG(&w4a, options) = SCARG(uap, options); 168 SCARG(&w4a, rusage) = SCARG(uap, rusage); 169 170 if ((error = wait4(p, &w4a, retval))) 171 return error; 172 173 if (status != NULL) { 174 if ((error = copyin(status, &tstat, sizeof tstat))) 175 return error; 176 177 bsd_to_linux_wstat(&tstat); 178 179 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 180 } 181 182 return 0; 183} 184 185/* 186 * This is the old brk(2) call. I don't think anything in the Linux 187 * world uses this anymore 188 */ 189int 190linux_break(p, uap, retval) 191 struct proc *p; 192 struct linux_brk_args /* { 193 syscallarg(char *) nsize; 194 } */ *uap; 195 register_t *retval; 196{ 197 return ENOSYS; 198} 199 200/* 201 * Linux brk(2). The check if the new address is >= the old one is 202 * done in the kernel in Linux. NetBSD does it in the library. 203 */ 204int 205linux_brk(p, uap, retval) 206 struct proc *p; 207 struct linux_brk_args /* { 208 syscallarg(char *) nsize; 209 } */ *uap; 210 register_t *retval; 211{ 212 char *nbrk = SCARG(uap, nsize); 213 struct obreak_args oba; 214 struct vmspace *vm = p->p_vmspace; 215 int error = 0; 216 caddr_t oldbrk, newbrk; 217 218 oldbrk = vm->vm_daddr + ctob(vm->vm_dsize); 219 /* 220 * XXX inconsistent.. Linux always returns at least the old 221 * brk value, but it will be page-aligned if this fails, 222 * and possibly not page aligned if it succeeds (the user 223 * supplied pointer is returned). 224 */ 225 SCARG(&oba, nsize) = nbrk; 226 227 if ((caddr_t) nbrk > vm->vm_daddr && obreak(p, &oba, retval) == 0) 228 retval[0] = (register_t) nbrk; 229 else 230 retval[0] = (register_t) oldbrk; 231 232 return 0; 233} 234 235/* 236 * I wonder why Linux has gettimeofday() _and_ time().. Still, we 237 * need to deal with it. 238 */ 239int 240linux_time(p, uap, retval) 241 struct proc *p; 242 struct linux_time_args /* { 243 linux_time_t *t; 244 } */ *uap; 245 register_t *retval; 246{ 247 struct timeval atv; 248 linux_time_t tt; 249 int error; 250 251 microtime(&atv); 252 253 tt = atv.tv_sec; 254 if (SCARG(uap, t) && (error = copyout(&tt, SCARG(uap, t), sizeof tt))) 255 return error; 256 257 retval[0] = tt; 258 return 0; 259} 260 261/* 262 * Convert BSD statfs structure to Linux statfs structure. 263 * The Linux structure has less fields, and it also wants 264 * the length of a name in a dir entry in a field, which 265 * we fake (probably the wrong way). 266 */ 267static void 268bsd_to_linux_statfs(bsp, lsp) 269 struct statfs *bsp; 270 struct linux_statfs *lsp; 271{ 272 lsp->l_ftype = bsp->f_type; 273 lsp->l_fbsize = bsp->f_bsize; 274 lsp->l_fblocks = bsp->f_blocks; 275 lsp->l_fbfree = bsp->f_bfree; 276 lsp->l_fbavail = bsp->f_bavail; 277 lsp->l_ffiles = bsp->f_files; 278 lsp->l_fffree = bsp->f_ffree; 279 lsp->l_ffsid.val[0] = bsp->f_fsid.val[0]; 280 lsp->l_ffsid.val[1] = bsp->f_fsid.val[1]; 281 lsp->l_fnamelen = MAXNAMLEN; /* XXX */ 282} 283 284/* 285 * Implement the fs stat functions. Straightforward. 286 */ 287int 288linux_statfs(p, uap, retval) 289 struct proc *p; 290 struct linux_statfs_args /* { 291 syscallarg(char *) path; 292 syscallarg(struct linux_statfs *) sp; 293 } */ *uap; 294 register_t *retval; 295{ 296 struct statfs btmp, *bsp; 297 struct linux_statfs ltmp; 298 struct statfs_args bsa; 299 caddr_t sg; 300 int error; 301 302 sg = stackgap_init(p->p_emul); 303 bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs)); 304 305 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 306 307 SCARG(&bsa, path) = SCARG(uap, path); 308 SCARG(&bsa, buf) = bsp; 309 310 if ((error = statfs(p, &bsa, retval))) 311 return error; 312 313 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 314 return error; 315 316 bsd_to_linux_statfs(&btmp, <mp); 317 318 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 319} 320 321int 322linux_fstatfs(p, uap, retval) 323 struct proc *p; 324 struct linux_fstatfs_args /* { 325 syscallarg(int) fd; 326 syscallarg(struct linux_statfs *) sp; 327 } */ *uap; 328 register_t *retval; 329{ 330 struct statfs btmp, *bsp; 331 struct linux_statfs ltmp; 332 struct fstatfs_args bsa; 333 caddr_t sg; 334 int error; 335 336 sg = stackgap_init(p->p_emul); 337 bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs)); 338 339 SCARG(&bsa, fd) = SCARG(uap, fd); 340 SCARG(&bsa, buf) = bsp; 341 342 if ((error = statfs(p, &bsa, retval))) 343 return error; 344 345 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 346 return error; 347 348 bsd_to_linux_statfs(&btmp, <mp); 349 350 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 351} 352 353/* 354 * uname(). Just copy the info from the various strings stored in the 355 * kernel, and put it in the Linux utsname structure. That structure 356 * is almost the same as the NetBSD one, only it has fields 65 characters 357 * long, and an extra domainname field. 358 */ 359int 360linux_uname(p, uap, retval) 361 struct proc *p; 362 struct linux_uname_args /* { 363 syscallarg(struct linux_utsname *) up; 364 } */ *uap; 365 register_t *retval; 366{ 367 extern char ostype[], hostname[], osrelease[], version[], machine[], 368 domainname[]; 369 struct linux_utsname luts; 370 int len; 371 char *cp; 372 373 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 374 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 375 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 376 strncpy(luts.l_version, version, sizeof(luts.l_version)); 377 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 378 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 379 380 /* This part taken from the the uname() in libc */ 381 len = sizeof(luts.l_version); 382 for (cp = luts.l_version; len--; ++cp) 383 if (*cp == '\n' || *cp == '\t') 384 if (len > 1) 385 *cp = ' '; 386 else 387 *cp = '\0'; 388 389 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 390} 391 392int 393linux_olduname(p, uap, retval) 394 struct proc *p; 395 struct linux_uname_args /* { 396 syscallarg(struct linux_oldutsname *) up; 397 } */ *uap; 398 register_t *retval; 399{ 400 extern char ostype[], hostname[], osrelease[], version[], machine[]; 401 struct linux_oldutsname luts; 402 int len; 403 char *cp; 404 405 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 406 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 407 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 408 strncpy(luts.l_version, version, sizeof(luts.l_version)); 409 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 410 411 /* This part taken from the the uname() in libc */ 412 len = sizeof(luts.l_version); 413 for (cp = luts.l_version; len--; ++cp) 414 if (*cp == '\n' || *cp == '\t') 415 if (len > 1) 416 *cp = ' '; 417 else 418 *cp = '\0'; 419 420 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 421} 422 423int 424linux_oldolduname(p, uap, retval) 425 struct proc *p; 426 struct linux_uname_args /* { 427 syscallarg(struct linux_oldoldutsname *) up; 428 } */ *uap; 429 register_t *retval; 430{ 431 extern char ostype[], hostname[], osrelease[], version[], machine[]; 432 struct linux_oldoldutsname luts; 433 int len; 434 char *cp; 435 436 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 437 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 438 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 439 strncpy(luts.l_version, version, sizeof(luts.l_version)); 440 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 441 442 /* This part taken from the the uname() in libc */ 443 len = sizeof(luts.l_version); 444 for (cp = luts.l_version; len--; ++cp) 445 if (*cp == '\n' || *cp == '\t') 446 if (len > 1) 447 *cp = ' '; 448 else 449 *cp = '\0'; 450 451 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 452} 453 454/* 455 * Linux wants to pass everything to a syscall in registers. However, 456 * mmap() has 6 of them. Oops: out of register error. They just pass 457 * everything in a structure. 458 */ 459int 460linux_mmap(p, uap, retval) 461 struct proc *p; 462 struct linux_mmap_args /* { 463 syscallarg(struct linux_mmap *) lmp; 464 } */ *uap; 465 register_t *retval; 466{ 467 struct linux_mmap lmap; 468 struct mmap_args cma; 469 int error, flags; 470 471 if ((error = copyin(SCARG(uap, lmp), &lmap, sizeof lmap))) 472 return error; 473 474 flags = 0; 475 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_SHARED, MAP_SHARED); 476 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_PRIVATE, MAP_PRIVATE); 477 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_FIXED, MAP_FIXED); 478 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_ANON, MAP_ANON); 479 480 SCARG(&cma,addr) = lmap.lm_addr; 481 SCARG(&cma,len) = lmap.lm_len; 482 SCARG(&cma,prot) = lmap.lm_prot; 483 SCARG(&cma,flags) = flags; 484 SCARG(&cma,fd) = lmap.lm_fd; 485 SCARG(&cma,pad) = 0; 486 SCARG(&cma,pos) = lmap.lm_pos; 487 488 return mmap(p, &cma, retval); 489} 490 491/* 492 * Linux doesn't use the retval[1] value to determine whether 493 * we are the child or parent. 494 */ 495int 496linux_fork(p, uap, retval) 497 struct proc *p; 498 void *uap; 499 register_t *retval; 500{ 501 int error; 502 503 if ((error = fork(p, uap, retval))) 504 return error; 505 506 if (retval[1] == 1) 507 retval[0] = 0; 508 509 return 0; 510} 511 512/* 513 * This code is partly stolen from src/lib/libc/compat-43/times.c 514 * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here 515 */ 516 517#define CLK_TCK 100 518#define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 519 520int 521linux_times(p, uap, retval) 522 struct proc *p; 523 struct linux_times_args /* { 524 syscallarg(struct times *) tms; 525 } */ *uap; 526 register_t *retval; 527{ 528 struct timeval t; 529 struct linux_tms ltms; 530 struct rusage ru; 531 int error, s; 532 533 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 534 ltms.ltms_utime = CONVTCK(ru.ru_utime); 535 ltms.ltms_stime = CONVTCK(ru.ru_stime); 536 537 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 538 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 539 540 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 541 return error; 542 543 s = splclock(); 544 timersub(&time, &boottime, &t); 545 splx(s); 546 547 retval[0] = ((linux_clock_t)(CONVTCK(t))); 548 return 0; 549} 550 551/* 552 * NetBSD passes fd[0] in retval[0], and fd[1] in retval[1]. 553 * Linux directly passes the pointer. 554 */ 555int 556linux_pipe(p, uap, retval) 557 struct proc *p; 558 struct linux_pipe_args /* { 559 syscallarg(int *) pfds; 560 } */ *uap; 561 register_t *retval; 562{ 563 int error; 564 565 if ((error = pipe(p, 0, retval))) 566 return error; 567 568 /* Assumes register_t is an int */ 569 570 if ((error = copyout(retval, SCARG(uap, pfds), 2 * sizeof (int)))) 571 return error; 572 573 retval[0] = 0; 574 return 0; 575} 576 577/* 578 * Alarm. This is a libc call which used setitimer(2) in NetBSD. 579 * Fiddle with the timers to make it work. 580 */ 581int 582linux_alarm(p, uap, retval) 583 struct proc *p; 584 struct linux_alarm_args /* { 585 syscallarg(unsigned int) secs; 586 } */ *uap; 587 register_t *retval; 588{ 589 int error, s; 590 struct itimerval *itp, it; 591 592 itp = &p->p_realtimer; 593 s = splclock(); 594 /* 595 * Clear any pending timer alarms. 596 */ 597 untimeout(realitexpire, p); 598 timerclear(&itp->it_interval); 599 if (timerisset(&itp->it_value) && 600 timercmp(&itp->it_value, &time, >)) 601 timersub(&itp->it_value, &time, &itp->it_value); 602 /* 603 * Return how many seconds were left (rounded up) 604 */ 605 retval[0] = itp->it_value.tv_sec; 606 if (itp->it_value.tv_usec) 607 retval[0]++; 608 609 /* 610 * alarm(0) just resets the timer. 611 */ 612 if (SCARG(uap, secs) == 0) { 613 timerclear(&itp->it_value); 614 splx(s); 615 return 0; 616 } 617 618 /* 619 * Check the new alarm time for sanity, and set it. 620 */ 621 timerclear(&it.it_interval); 622 it.it_value.tv_sec = SCARG(uap, secs); 623 it.it_value.tv_usec = 0; 624 if (itimerfix(&it.it_value) || itimerfix(&it.it_interval)) { 625 splx(s); 626 return (EINVAL); 627 } 628 629 if (timerisset(&it.it_value)) { 630 timeradd(&it.it_value, &time, &it.it_value); 631 timeout(realitexpire, p, hzto(&it.it_value)); 632 } 633 p->p_realtimer = it; 634 splx(s); 635 636 return 0; 637} 638 639/* 640 * utime(). Do conversion to things that utimes() understands, 641 * and pass it on. 642 */ 643int 644linux_utime(p, uap, retval) 645 struct proc *p; 646 struct linux_utime_args /* { 647 syscallarg(char *) path; 648 syscallarg(struct linux_utimbuf *)times; 649 } */ *uap; 650 register_t *retval; 651{ 652 caddr_t sg; 653 int error; 654 struct utimes_args ua; 655 struct timeval tv[2], *tvp; 656 struct linux_utimbuf lut; 657 658 sg = stackgap_init(p->p_emul); 659 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 660 661 SCARG(&ua, path) = SCARG(uap, path); 662 663 if (SCARG(uap, times) != NULL) { 664 if ((error = copyin(SCARG(uap, times), &lut, sizeof lut))) 665 return error; 666 tv[0].tv_usec = tv[1].tv_usec = 0; 667 tv[0].tv_sec = lut.l_actime; 668 tv[1].tv_sec = lut.l_modtime; 669 tvp = (struct timeval *) stackgap_alloc(&sg, sizeof(tv)); 670 if ((error = copyout(tv, tvp, sizeof tv))) 671 return error; 672 SCARG(&ua, tptr) = tvp; 673 } 674 else 675 SCARG(&ua, tptr) = NULL; 676 677 return utimes(p, uap, retval); 678} 679 680/* 681 * Linux 'readdir' call. This code is mostly taken from the 682 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 683 * an attempt has been made to keep it a little cleaner (failing 684 * miserably, because of the cruft needed if count 1 is passed). 685 * 686 * Read in BSD-style entries, convert them, and copy them out. 687 * Note that the Linux d_reclen is actually the name length, 688 * and d_off is the reclen. 689 * 690 * Note that this doesn't handle union-mounted filesystems. 691 */ 692int 693linux_readdir(p, uap, retval) 694 struct proc *p; 695 struct linux_readdir_args /* { 696 syscallarg(int) fd; 697 syscallarg(struct linux_dirent *) dent; 698 syscallarg(unsigned int) count; 699 } */ *uap; 700 register_t *retval; 701{ 702 register struct dirent *bdp; 703 struct vnode *vp; 704 caddr_t inp, buf; /* BSD-format */ 705 int len, reclen; /* BSD-format */ 706 caddr_t outp; /* Linux-format */ 707 int resid, linuxreclen; /* Linux-format */ 708 struct file *fp; 709 struct uio auio; 710 struct iovec aiov; 711 struct linux_dirent idb; 712 off_t off; /* true file offset */ 713 linux_off_t soff; /* Linux file offset */ 714 int buflen, error, eofflag, nbytes, justone; 715 struct vattr va; 716 717 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 718 return (error); 719 720 if ((fp->f_flag & FREAD) == 0) 721 return (EBADF); 722 723 vp = (struct vnode *)fp->f_data; 724 725 if (vp->v_type != VDIR) /* XXX vnode readdir op should do this */ 726 return (EINVAL); 727 728 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 729 return error; 730 731 nbytes = SCARG(uap, count); 732 if (nbytes == 1) { /* Need this for older Linux libs, apparently */ 733 nbytes = sizeof (struct linux_dirent); 734 buflen = max(va.va_blocksize, nbytes); 735 justone = 1; 736 } else { 737 buflen = min(MAXBSIZE, nbytes); 738 justone = 0; 739 } 740 buf = malloc(buflen, M_TEMP, M_WAITOK); 741 VOP_LOCK(vp); 742 off = fp->f_offset; 743again: 744 aiov.iov_base = buf; 745 aiov.iov_len = buflen; 746 auio.uio_iov = &aiov; 747 auio.uio_iovcnt = 1; 748 auio.uio_rw = UIO_READ; 749 auio.uio_segflg = UIO_SYSSPACE; 750 auio.uio_procp = p; 751 auio.uio_resid = buflen; 752 auio.uio_offset = off; 753 /* 754 * First we read into the malloc'ed buffer, then 755 * we massage it into user space, one record at a time. 756 */ 757 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, (u_long *)0, 0); 758 if (error) 759 goto out; 760 761 inp = buf; 762 outp = (caddr_t) SCARG(uap, dent); 763 resid = nbytes; 764 if ((len = buflen - auio.uio_resid) == 0) 765 goto eof; 766 767 for (; len > 0; len -= reclen) { 768 bdp = (struct dirent *)inp; 769 reclen = bdp->d_reclen; 770 if (reclen & 3) 771 panic("linux_readdir"); 772 off += reclen; /* each entry points to next */ 773 if (bdp->d_fileno == 0) { 774 inp += reclen; /* it is a hole; squish it out */ 775 continue; 776 } 777 linuxreclen = LINUX_RECLEN(&idb, bdp->d_namlen); 778 if (reclen > len || resid < linuxreclen) { 779 /* entry too big for buffer, so just stop */ 780 outp++; 781 break; 782 } 783 /* 784 * Massage in place to make a Linux-shaped dirent (otherwise 785 * we have to worry about touching user memory outside of 786 * the copyout() call). 787 */ 788 idb.d_ino = (long)bdp->d_fileno; 789 idb.d_off = (linux_off_t)linuxreclen; 790 idb.d_reclen = (u_short)bdp->d_namlen; 791 strcpy(idb.d_name, bdp->d_name); 792 if ((error = copyout((caddr_t)&idb, outp, linuxreclen))) 793 goto out; 794 /* advance past this real entry */ 795 inp += reclen; 796 /* advance output past Linux-shaped entry */ 797 outp += linuxreclen; 798 resid -= linuxreclen; 799 if (justone) 800 break; 801 } 802 803 /* if we squished out the whole block, try again */ 804 if (outp == (caddr_t) SCARG(uap, dent)) 805 goto again; 806 fp->f_offset = off; /* update the vnode offset */ 807 808 if (justone) 809 nbytes = resid + linuxreclen; 810 811eof: 812 *retval = nbytes - resid; 813out: 814 VOP_UNLOCK(vp); 815 free(buf, M_TEMP); 816 return error; 817} 818 819/* 820 * Out of register error once more.. Also, Linux copies the amount of 821 * time left into the user-supplied timeval structure. 822 */ 823int 824linux_select(p, uap, retval) 825 struct proc *p; 826 struct linux_select_args /* { 827 syscallarg(struct linux_select *) lsp; 828 } */ *uap; 829 register_t *retval; 830{ 831 struct linux_select ls; 832 struct select_args bsa; 833 struct timeval tv0, tv1, utv, *tvp; 834 caddr_t sg; 835 int error; 836 837 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof(ls)))) 838 return error; 839 840 SCARG(&bsa, nd) = ls.nfds; 841 SCARG(&bsa, in) = ls.readfds; 842 SCARG(&bsa, ou) = ls.writefds; 843 SCARG(&bsa, ex) = ls.exceptfds; 844 SCARG(&bsa, tv) = ls.timeout; 845 846 /* 847 * Store current time for computation of the amount of 848 * time left. 849 */ 850 if (ls.timeout) { 851 if ((error = copyin(ls.timeout, &utv, sizeof(utv)))) 852 return error; 853 if (itimerfix(&utv)) { 854 /* 855 * The timeval was invalid. Convert it to something 856 * valid that will act as it does under Linux. 857 */ 858 sg = stackgap_init(p->p_emul); 859 tvp = stackgap_alloc(&sg, sizeof(utv)); 860 utv.tv_sec += utv.tv_usec / 1000000; 861 utv.tv_usec %= 1000000; 862 if (utv.tv_usec < 0) { 863 utv.tv_sec -= 1; 864 utv.tv_usec += 1000000; 865 } 866 if (utv.tv_sec < 0) 867 timerclear(&utv); 868 if ((error = copyout(&utv, tvp, sizeof(utv)))) 869 return error; 870 SCARG(&bsa, tv) = tvp; 871 } 872 microtime(&tv0); 873 } 874 875 error = select(p, &bsa, retval); 876 if (error) { 877 /* 878 * See fs/select.c in the Linux kernel. Without this, 879 * Maelstrom doesn't work. 880 */ 881 if (error == ERESTART) 882 error = EINTR; 883 return error; 884 } 885 886 if (ls.timeout) { 887 if (*retval) { 888 /* 889 * Compute how much time was left of the timeout, 890 * by subtracting the current time and the time 891 * before we started the call, and subtracting 892 * that result from the user-supplied value. 893 */ 894 microtime(&tv1); 895 timersub(&tv1, &tv0, &tv1); 896 timersub(&utv, &tv1, &utv); 897 if (utv.tv_sec < 0) 898 timerclear(&utv); 899 } else 900 timerclear(&utv); 901 if ((error = copyout(&utv, ls.timeout, sizeof(utv)))) 902 return error; 903 } 904 905 return 0; 906} 907 908/* 909 * Get the process group of a certain process. Look it up 910 * and return the value. 911 */ 912int 913linux_getpgid(p, uap, retval) 914 struct proc *p; 915 struct linux_getpgid_args /* { 916 syscallarg(int) pid; 917 } */ *uap; 918 register_t *retval; 919{ 920 struct proc *targp; 921 922 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) 923 if ((targp = pfind(SCARG(uap, pid))) == 0) 924 return ESRCH; 925 else 926 targp = p; 927 928 retval[0] = targp->p_pgid; 929 return 0; 930} 931 932/* 933 * Set the 'personality' (emulation mode) for the current process. Only 934 * accept the Linux personality here (0). This call is needed because 935 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 936 * ELF binaries run in Linux mode, not SVR4 mode. 937 */ 938int 939linux_personality(p, uap, retval) 940 struct proc *p; 941 struct linux_personality_args /* P 942 syscallarg(int) per; 943 } */ *uap; 944 register_t *retval; 945{ 946 if (SCARG(uap, per) != 0) 947 return EINVAL; 948 retval[0] = 0; 949 return 0; 950} 951