1 /*- 2 * Copyright (c) 1994-1995 S�ren Schmidt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_mac.h" 33 34 #include <sys/param.h> 35 #include <sys/blist.h> 36 #include <sys/fcntl.h> 37 #include <sys/imgact_aout.h> 38 #include <sys/jail.h> 39 #include <sys/kernel.h> 40 #include <sys/limits.h> 41 #include <sys/lock.h> 42 #include <sys/mac.h> 43 #include <sys/malloc.h> 44 #include <sys/mman.h> 45 #include <sys/mount.h> 46 #include <sys/mutex.h> 47 #include <sys/namei.h> 48 #include <sys/proc.h> 49 #include <sys/reboot.h> 50 #include <sys/resourcevar.h> 51 #include <sys/signalvar.h> 52 #include <sys/stat.h> 53 #include <sys/syscallsubr.h> 54 #include <sys/sysctl.h> 55 #include <sys/sysproto.h> 56 #include <sys/systm.h> 57 #include <sys/time.h> 58 #include <sys/vmmeter.h> 59 #include <sys/vnode.h> 60 #include <sys/wait.h> 61 62 #include <vm/vm.h> 63 #include <vm/pmap.h> 64 #include <vm/vm_kern.h> 65 #include <vm/vm_map.h> 66 #include <vm/vm_extern.h> 67 #include <vm/vm_object.h> 68 #include <vm/swap_pager.h> 69 70 #include <posix4/sched.h> 71 72 #include <machine/../linux/linux.h> 73 #include <machine/../linux/linux_proto.h> 74 75 #include <compat/linux/linux_mib.h> 76 #include <compat/linux/linux_util.h> 77 78 #ifdef __i386__ 79 #include <machine/cputypes.h> 80 #endif 81 82 #ifdef __alpha__ 83 #define BSD_TO_LINUX_SIGNAL(sig) (sig) 84 #else 85 #define BSD_TO_LINUX_SIGNAL(sig) \ 86 (((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig) 87 #endif 88 89 #ifndef __alpha__ 90 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 91 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 92 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 93 RLIMIT_MEMLOCK, -1 94 }; 95 #endif /*!__alpha__*/ 96 97 struct l_sysinfo { 98 l_long uptime; /* Seconds since boot */ 99 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 100 #define LINUX_SYSINFO_LOADS_SCALE 65536 101 l_ulong totalram; /* Total usable main memory size */ 102 l_ulong freeram; /* Available memory size */ 103 l_ulong sharedram; /* Amount of shared memory */ 104 l_ulong bufferram; /* Memory used by buffers */ 105 l_ulong totalswap; /* Total swap space size */ 106 l_ulong freeswap; /* swap space still available */ 107 l_ushort procs; /* Number of current processes */ 108 l_ulong totalbig; 109 l_ulong freebig; 110 l_uint mem_unit; 111 char _f[6]; /* Pads structure to 64 bytes */ 112 }; 113 #ifndef __alpha__ 114 int 115 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 116 { 117 struct l_sysinfo sysinfo; 118 vm_object_t object; 119 int i, j; 120 struct timespec ts; 121 122 /* Uptime is copied out of print_uptime() in kern_shutdown.c */ 123 getnanouptime(&ts); 124 i = 0; 125 if (ts.tv_sec >= 86400) { 126 ts.tv_sec %= 86400; 127 i = 1; 128 } 129 if (i || ts.tv_sec >= 3600) { 130 ts.tv_sec %= 3600; 131 i = 1; 132 } 133 if (i || ts.tv_sec >= 60) { 134 ts.tv_sec %= 60; 135 i = 1; 136 } 137 sysinfo.uptime=ts.tv_sec; 138 139 /* Use the information from the mib to get our load averages */ 140 for (i = 0; i < 3; i++) 141 sysinfo.loads[i] = averunnable.ldavg[i] * 142 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 143 144 sysinfo.totalram = physmem * PAGE_SIZE; 145 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE; 146 147 sysinfo.sharedram = 0; 148 for (object = TAILQ_FIRST(&vm_object_list); object != NULL; 149 object = TAILQ_NEXT(object, object_list)) 150 if (object->shadow_count > 1) 151 sysinfo.sharedram += object->resident_page_count; 152 153 sysinfo.sharedram *= PAGE_SIZE; 154 sysinfo.bufferram = 0; 155 156 swap_pager_status(&i, &j); 157 sysinfo.totalswap= i * PAGE_SIZE; 158 sysinfo.freeswap = (i - j) * PAGE_SIZE; 159 160 sysinfo.procs = nprocs; 161 162 /* The following are only present in newer Linux kernels. */ 163 sysinfo.totalbig = 0; 164 sysinfo.freebig = 0; 165 sysinfo.mem_unit = 1; 166 167 return copyout(&sysinfo, args->info, sizeof(sysinfo)); 168 } 169 #endif /*!__alpha__*/ 170 171 #ifndef __alpha__ 172 int 173 linux_alarm(struct thread *td, struct linux_alarm_args *args) 174 { 175 struct itimerval it, old_it; 176 struct timeval tv; 177 struct proc *p; 178 179 #ifdef DEBUG 180 if (ldebug(alarm)) 181 printf(ARGS(alarm, "%u"), args->secs); 182 #endif 183 184 if (args->secs > 100000000) 185 return EINVAL; 186 187 it.it_value.tv_sec = (long)args->secs; 188 it.it_value.tv_usec = 0; 189 it.it_interval.tv_sec = 0; 190 it.it_interval.tv_usec = 0; 191 p = td->td_proc; 192 PROC_LOCK(p); 193 old_it = p->p_realtimer; 194 getmicrouptime(&tv); 195 if (timevalisset(&old_it.it_value)) 196 callout_stop(&p->p_itcallout); 197 if (it.it_value.tv_sec != 0) { 198 callout_reset(&p->p_itcallout, tvtohz(&it.it_value), 199 realitexpire, p); 200 timevaladd(&it.it_value, &tv); 201 } 202 p->p_realtimer = it; 203 PROC_UNLOCK(p); 204 if (timevalcmp(&old_it.it_value, &tv, >)) { 205 timevalsub(&old_it.it_value, &tv); 206 if (old_it.it_value.tv_usec != 0) 207 old_it.it_value.tv_sec++; 208 td->td_retval[0] = old_it.it_value.tv_sec; 209 } 210 return 0; 211 } 212 #endif /*!__alpha__*/ 213 214 int 215 linux_brk(struct thread *td, struct linux_brk_args *args) 216 { 217 struct vmspace *vm = td->td_proc->p_vmspace; 218 vm_offset_t new, old; 219 struct obreak_args /* { 220 char * nsize; 221 } */ tmp; 222 223 #ifdef DEBUG 224 if (ldebug(brk)) 225 printf(ARGS(brk, "%p"), (void *)args->dsend); 226 #endif 227 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 228 new = (vm_offset_t)args->dsend; 229 tmp.nsize = (char *) new; 230 if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp)) 231 td->td_retval[0] = (long)new; 232 else 233 td->td_retval[0] = (long)old; 234 235 return 0; 236 } 237 238 int 239 linux_uselib(struct thread *td, struct linux_uselib_args *args) 240 { 241 struct nameidata ni; 242 struct vnode *vp; 243 struct exec *a_out; 244 struct vattr attr; 245 vm_offset_t vmaddr; 246 unsigned long file_offset; 247 vm_offset_t buffer; 248 unsigned long bss_size; 249 char *library; 250 int error; 251 int locked; 252 253 LCONVPATHEXIST(td, args->library, &library); 254 255 #ifdef DEBUG 256 if (ldebug(uselib)) 257 printf(ARGS(uselib, "%s"), library); 258 #endif 259 260 a_out = NULL; 261 locked = 0; 262 vp = NULL; 263 264 /* 265 * XXX: This code should make use of vn_open(), rather than doing 266 * all this stuff itself. 267 */ 268 NDINIT(&ni, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, library, td); 269 error = namei(&ni); 270 LFREEPATH(library); 271 if (error) 272 goto cleanup; 273 274 vp = ni.ni_vp; 275 /* 276 * XXX - This looks like a bogus check. A LOCKLEAF namei should not 277 * succeed without returning a vnode. 278 */ 279 if (vp == NULL) { 280 error = ENOEXEC; /* ?? */ 281 goto cleanup; 282 } 283 NDFREE(&ni, NDF_ONLY_PNBUF); 284 285 /* 286 * From here on down, we have a locked vnode that must be unlocked. 287 */ 288 locked++; 289 290 /* Writable? */ 291 if (vp->v_writecount) { 292 error = ETXTBSY; 293 goto cleanup; 294 } 295 296 /* Executable? */ 297 error = VOP_GETATTR(vp, &attr, td->td_ucred, td); 298 if (error) 299 goto cleanup; 300 301 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 302 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 303 error = ENOEXEC; 304 goto cleanup; 305 } 306 307 /* Sensible size? */ 308 if (attr.va_size == 0) { 309 error = ENOEXEC; 310 goto cleanup; 311 } 312 313 /* Can we access it? */ 314 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 315 if (error) 316 goto cleanup; 317 318 /* 319 * XXX: This should use vn_open() so that it is properly authorized, 320 * and to reduce code redundancy all over the place here. 321 */ 322 #ifdef MAC 323 error = mac_check_vnode_open(td->td_ucred, vp, FREAD); 324 if (error) 325 goto cleanup; 326 #endif 327 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, -1); 328 if (error) 329 goto cleanup; 330 331 /* Pull in executable header into kernel_map */ 332 error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE, 333 VM_PROT_READ, VM_PROT_READ, 0, (caddr_t)vp, 0); 334 /* 335 * Lock no longer needed 336 */ 337 locked = 0; 338 VOP_UNLOCK(vp, 0, td); 339 340 if (error) 341 goto cleanup; 342 343 /* Is it a Linux binary ? */ 344 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 345 error = ENOEXEC; 346 goto cleanup; 347 } 348 349 /* 350 * While we are here, we should REALLY do some more checks 351 */ 352 353 /* Set file/virtual offset based on a.out variant. */ 354 switch ((int)(a_out->a_magic & 0xffff)) { 355 case 0413: /* ZMAGIC */ 356 file_offset = 1024; 357 break; 358 case 0314: /* QMAGIC */ 359 file_offset = 0; 360 break; 361 default: 362 error = ENOEXEC; 363 goto cleanup; 364 } 365 366 bss_size = round_page(a_out->a_bss); 367 368 /* Check various fields in header for validity/bounds. */ 369 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 370 error = ENOEXEC; 371 goto cleanup; 372 } 373 374 /* text + data can't exceed file size */ 375 if (a_out->a_data + a_out->a_text > attr.va_size) { 376 error = EFAULT; 377 goto cleanup; 378 } 379 380 /* To protect td->td_proc->p_rlimit in the if condition. */ 381 mtx_assert(&Giant, MA_OWNED); 382 383 /* 384 * text/data/bss must not exceed limits 385 * XXX - this is not complete. it should check current usage PLUS 386 * the resources needed by this library. 387 */ 388 if (a_out->a_text > maxtsiz || 389 a_out->a_data + bss_size > 390 td->td_proc->p_rlimit[RLIMIT_DATA].rlim_cur) { 391 error = ENOMEM; 392 goto cleanup; 393 } 394 395 mp_fixme("Unlocked vflags access."); 396 /* prevent more writers */ 397 vp->v_vflag |= VV_TEXT; 398 399 /* 400 * Check if file_offset page aligned. Currently we cannot handle 401 * misalinged file offsets, and so we read in the entire image 402 * (what a waste). 403 */ 404 if (file_offset & PAGE_MASK) { 405 #ifdef DEBUG 406 printf("uselib: Non page aligned binary %lu\n", file_offset); 407 #endif 408 /* Map text+data read/write/execute */ 409 410 /* a_entry is the load address and is page aligned */ 411 vmaddr = trunc_page(a_out->a_entry); 412 413 /* get anon user mapping, read+write+execute */ 414 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 415 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL, 416 VM_PROT_ALL, 0); 417 if (error) 418 goto cleanup; 419 420 /* map file into kernel_map */ 421 error = vm_mmap(kernel_map, &buffer, 422 round_page(a_out->a_text + a_out->a_data + file_offset), 423 VM_PROT_READ, VM_PROT_READ, 0, (caddr_t)vp, 424 trunc_page(file_offset)); 425 if (error) 426 goto cleanup; 427 428 /* copy from kernel VM space to user space */ 429 error = copyout((void *)(uintptr_t)(buffer + file_offset), 430 (void *)vmaddr, a_out->a_text + a_out->a_data); 431 432 /* release temporary kernel space */ 433 vm_map_remove(kernel_map, buffer, buffer + 434 round_page(a_out->a_text + a_out->a_data + file_offset)); 435 436 if (error) 437 goto cleanup; 438 } else { 439 #ifdef DEBUG 440 printf("uselib: Page aligned binary %lu\n", file_offset); 441 #endif 442 /* 443 * for QMAGIC, a_entry is 20 bytes beyond the load address 444 * to skip the executable header 445 */ 446 vmaddr = trunc_page(a_out->a_entry); 447 448 /* 449 * Map it all into the process's space as a single 450 * copy-on-write "data" segment. 451 */ 452 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 453 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 454 MAP_PRIVATE | MAP_FIXED, (caddr_t)vp, file_offset); 455 if (error) 456 goto cleanup; 457 } 458 #ifdef DEBUG 459 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long*)vmaddr)[0], 460 ((long*)vmaddr)[1]); 461 #endif 462 if (bss_size != 0) { 463 /* Calculate BSS start address */ 464 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 465 a_out->a_data; 466 467 /* allocate some 'anon' space */ 468 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 469 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0); 470 if (error) 471 goto cleanup; 472 } 473 474 cleanup: 475 /* Unlock vnode if needed */ 476 if (locked) 477 VOP_UNLOCK(vp, 0, td); 478 479 /* Release the kernel mapping. */ 480 if (a_out) 481 vm_map_remove(kernel_map, (vm_offset_t)a_out, 482 (vm_offset_t)a_out + PAGE_SIZE); 483 484 return error; 485 } 486 487 int 488 linux_select(struct thread *td, struct linux_select_args *args) 489 { 490 struct timeval tv0, tv1, utv, *tvp; 491 int error; 492 493 #ifdef DEBUG 494 if (ldebug(select)) 495 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 496 (void *)args->readfds, (void *)args->writefds, 497 (void *)args->exceptfds, (void *)args->timeout); 498 #endif 499 500 /* 501 * Store current time for computation of the amount of 502 * time left. 503 */ 504 if (args->timeout) { 505 if ((error = copyin(args->timeout, &utv, sizeof(utv)))) 506 goto select_out; 507 #ifdef DEBUG 508 if (ldebug(select)) 509 printf(LMSG("incoming timeout (%ld/%ld)"), 510 utv.tv_sec, utv.tv_usec); 511 #endif 512 513 if (itimerfix(&utv)) { 514 /* 515 * The timeval was invalid. Convert it to something 516 * valid that will act as it does under Linux. 517 */ 518 utv.tv_sec += utv.tv_usec / 1000000; 519 utv.tv_usec %= 1000000; 520 if (utv.tv_usec < 0) { 521 utv.tv_sec -= 1; 522 utv.tv_usec += 1000000; 523 } 524 if (utv.tv_sec < 0) 525 timevalclear(&utv); 526 } 527 microtime(&tv0); 528 tvp = &utv; 529 } else 530 tvp = NULL; 531 532 error = kern_select(td, args->nfds, args->readfds, args->writefds, 533 args->exceptfds, tvp); 534 535 #ifdef DEBUG 536 if (ldebug(select)) 537 printf(LMSG("real select returns %d"), error); 538 #endif 539 if (error) { 540 /* 541 * See fs/select.c in the Linux kernel. Without this, 542 * Maelstrom doesn't work. 543 */ 544 if (error == ERESTART) 545 error = EINTR; 546 goto select_out; 547 } 548 549 if (args->timeout) { 550 if (td->td_retval[0]) { 551 /* 552 * Compute how much time was left of the timeout, 553 * by subtracting the current time and the time 554 * before we started the call, and subtracting 555 * that result from the user-supplied value. 556 */ 557 microtime(&tv1); 558 timevalsub(&tv1, &tv0); 559 timevalsub(&utv, &tv1); 560 if (utv.tv_sec < 0) 561 timevalclear(&utv); 562 } else 563 timevalclear(&utv); 564 #ifdef DEBUG 565 if (ldebug(select)) 566 printf(LMSG("outgoing timeout (%ld/%ld)"), 567 utv.tv_sec, utv.tv_usec); 568 #endif 569 if ((error = copyout(&utv, args->timeout, sizeof(utv)))) 570 goto select_out; 571 } 572 573 select_out: 574 #ifdef DEBUG 575 if (ldebug(select)) 576 printf(LMSG("select_out -> %d"), error); 577 #endif 578 return error; 579 } 580 581 int 582 linux_mremap(struct thread *td, struct linux_mremap_args *args) 583 { 584 struct munmap_args /* { 585 void *addr; 586 size_t len; 587 } */ bsd_args; 588 int error = 0; 589 590 #ifdef DEBUG 591 if (ldebug(mremap)) 592 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 593 (void *)args->addr, 594 (unsigned long)args->old_len, 595 (unsigned long)args->new_len, 596 (unsigned long)args->flags); 597 #endif 598 args->new_len = round_page(args->new_len); 599 args->old_len = round_page(args->old_len); 600 601 if (args->new_len > args->old_len) { 602 td->td_retval[0] = 0; 603 return ENOMEM; 604 } 605 606 if (args->new_len < args->old_len) { 607 bsd_args.addr = (caddr_t)(args->addr + args->new_len); 608 bsd_args.len = args->old_len - args->new_len; 609 error = munmap(td, &bsd_args); 610 } 611 612 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 613 return error; 614 } 615 616 #define LINUX_MS_ASYNC 0x0001 617 #define LINUX_MS_INVALIDATE 0x0002 618 #define LINUX_MS_SYNC 0x0004 619 620 int 621 linux_msync(struct thread *td, struct linux_msync_args *args) 622 { 623 struct msync_args bsd_args; 624 625 bsd_args.addr = (caddr_t)args->addr; 626 bsd_args.len = args->len; 627 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 628 629 return msync(td, &bsd_args); 630 } 631 632 #ifndef __alpha__ 633 int 634 linux_time(struct thread *td, struct linux_time_args *args) 635 { 636 struct timeval tv; 637 l_time_t tm; 638 int error; 639 640 #ifdef DEBUG 641 if (ldebug(time)) 642 printf(ARGS(time, "*")); 643 #endif 644 645 microtime(&tv); 646 tm = tv.tv_sec; 647 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 648 return error; 649 td->td_retval[0] = tm; 650 return 0; 651 } 652 #endif /*!__alpha__*/ 653 654 struct l_times_argv { 655 l_long tms_utime; 656 l_long tms_stime; 657 l_long tms_cutime; 658 l_long tms_cstime; 659 }; 660 661 #ifdef __alpha__ 662 #define CLK_TCK 1024 /* Linux uses 1024 on alpha */ 663 #else 664 #define CLK_TCK 100 /* Linux uses 100 */ 665 #endif 666 667 #define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 668 669 int 670 linux_times(struct thread *td, struct linux_times_args *args) 671 { 672 struct timeval tv; 673 struct l_times_argv tms; 674 struct rusage ru; 675 int error; 676 677 #ifdef DEBUG 678 if (ldebug(times)) 679 printf(ARGS(times, "*")); 680 #endif 681 682 mtx_lock_spin(&sched_lock); 683 calcru(td->td_proc, &ru.ru_utime, &ru.ru_stime, NULL); 684 mtx_unlock_spin(&sched_lock); 685 686 tms.tms_utime = CONVTCK(ru.ru_utime); 687 tms.tms_stime = CONVTCK(ru.ru_stime); 688 689 tms.tms_cutime = CONVTCK(td->td_proc->p_stats->p_cru.ru_utime); 690 tms.tms_cstime = CONVTCK(td->td_proc->p_stats->p_cru.ru_stime); 691 692 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 693 return error; 694 695 microuptime(&tv); 696 td->td_retval[0] = (int)CONVTCK(tv); 697 return 0; 698 } 699 700 int 701 linux_newuname(struct thread *td, struct linux_newuname_args *args) 702 { 703 struct l_new_utsname utsname; 704 char osname[LINUX_MAX_UTSNAME]; 705 char osrelease[LINUX_MAX_UTSNAME]; 706 char *p; 707 708 #ifdef DEBUG 709 if (ldebug(newuname)) 710 printf(ARGS(newuname, "*")); 711 #endif 712 713 linux_get_osname(td, osname); 714 linux_get_osrelease(td, osrelease); 715 716 bzero(&utsname, sizeof(utsname)); 717 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 718 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 719 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 720 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 721 for (p = utsname.version; *p != '\0'; ++p) 722 if (*p == '\n') { 723 *p = '\0'; 724 break; 725 } 726 #ifdef __i386__ 727 { 728 const char *class; 729 switch (cpu_class) { 730 case CPUCLASS_686: 731 class = "i686"; 732 break; 733 case CPUCLASS_586: 734 class = "i586"; 735 break; 736 case CPUCLASS_486: 737 class = "i486"; 738 break; 739 default: 740 class = "i386"; 741 } 742 strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME); 743 } 744 #else 745 strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME); 746 #endif 747 strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME); 748 749 return (copyout(&utsname, args->buf, sizeof(utsname))); 750 } 751 752 #if defined(__i386__) 753 struct l_utimbuf { 754 l_time_t l_actime; 755 l_time_t l_modtime; 756 }; 757 758 int 759 linux_utime(struct thread *td, struct linux_utime_args *args) 760 { 761 struct timeval tv[2], *tvp; 762 struct l_utimbuf lut; 763 char *fname; 764 int error; 765 766 LCONVPATHEXIST(td, args->fname, &fname); 767 768 #ifdef DEBUG 769 if (ldebug(utime)) 770 printf(ARGS(utime, "%s, *"), fname); 771 #endif 772 773 if (args->times) { 774 if ((error = copyin(args->times, &lut, sizeof lut))) { 775 LFREEPATH(fname); 776 return error; 777 } 778 tv[0].tv_sec = lut.l_actime; 779 tv[0].tv_usec = 0; 780 tv[1].tv_sec = lut.l_modtime; 781 tv[1].tv_usec = 0; 782 tvp = tv; 783 } else 784 tvp = NULL; 785 786 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 787 LFREEPATH(fname); 788 return (error); 789 } 790 #endif /* __i386__ */ 791 792 #define __WCLONE 0x80000000 793 794 #ifndef __alpha__ 795 int 796 linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 797 { 798 struct wait_args /* { 799 int pid; 800 int *status; 801 int options; 802 struct rusage *rusage; 803 } */ tmp; 804 int error, tmpstat; 805 806 #ifdef DEBUG 807 if (ldebug(waitpid)) 808 printf(ARGS(waitpid, "%d, %p, %d"), 809 args->pid, (void *)args->status, args->options); 810 #endif 811 812 tmp.pid = args->pid; 813 tmp.status = args->status; 814 tmp.options = (args->options & (WNOHANG | WUNTRACED)); 815 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 816 if (args->options & __WCLONE) 817 tmp.options |= WLINUXCLONE; 818 tmp.rusage = NULL; 819 820 if ((error = wait4(td, &tmp)) != 0) 821 return error; 822 823 if (args->status) { 824 if ((error = copyin(args->status, &tmpstat, sizeof(int))) != 0) 825 return error; 826 tmpstat &= 0xffff; 827 if (WIFSIGNALED(tmpstat)) 828 tmpstat = (tmpstat & 0xffffff80) | 829 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 830 else if (WIFSTOPPED(tmpstat)) 831 tmpstat = (tmpstat & 0xffff00ff) | 832 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 833 return copyout(&tmpstat, args->status, sizeof(int)); 834 } 835 836 return 0; 837 } 838 #endif /*!__alpha__*/ 839 840 int 841 linux_wait4(struct thread *td, struct linux_wait4_args *args) 842 { 843 struct wait_args /* { 844 int pid; 845 int *status; 846 int options; 847 struct rusage *rusage; 848 } */ tmp; 849 int error, tmpstat; 850 struct proc *p; 851 852 #ifdef DEBUG 853 if (ldebug(wait4)) 854 printf(ARGS(wait4, "%d, %p, %d, %p"), 855 args->pid, (void *)args->status, args->options, 856 (void *)args->rusage); 857 #endif 858 859 tmp.pid = args->pid; 860 tmp.status = args->status; 861 tmp.options = (args->options & (WNOHANG | WUNTRACED)); 862 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 863 if (args->options & __WCLONE) 864 tmp.options |= WLINUXCLONE; 865 tmp.rusage = (struct rusage *)args->rusage; 866 867 if ((error = wait4(td, &tmp)) != 0) 868 return error; 869 870 p = td->td_proc; 871 PROC_LOCK(p); 872 SIGDELSET(p->p_siglist, SIGCHLD); 873 PROC_UNLOCK(p); 874 875 if (args->status) { 876 if ((error = copyin(args->status, &tmpstat, sizeof(int))) != 0) 877 return error; 878 tmpstat &= 0xffff; 879 if (WIFSIGNALED(tmpstat)) 880 tmpstat = (tmpstat & 0xffffff80) | 881 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 882 else if (WIFSTOPPED(tmpstat)) 883 tmpstat = (tmpstat & 0xffff00ff) | 884 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 885 return copyout(&tmpstat, args->status, sizeof(int)); 886 } 887 888 return 0; 889 } 890 891 int 892 linux_mknod(struct thread *td, struct linux_mknod_args *args) 893 { 894 char *path; 895 int error; 896 897 LCONVPATHCREAT(td, args->path, &path); 898 899 #ifdef DEBUG 900 if (ldebug(mknod)) 901 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev); 902 #endif 903 904 if (args->mode & S_IFIFO) 905 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 906 else 907 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 908 args->dev); 909 LFREEPATH(path); 910 return (error); 911 } 912 913 /* 914 * UGH! This is just about the dumbest idea I've ever heard!! 915 */ 916 int 917 linux_personality(struct thread *td, struct linux_personality_args *args) 918 { 919 #ifdef DEBUG 920 if (ldebug(personality)) 921 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 922 #endif 923 #ifndef __alpha__ 924 if (args->per != 0) 925 return EINVAL; 926 #endif 927 928 /* Yes Jim, it's still a Linux... */ 929 td->td_retval[0] = 0; 930 return 0; 931 } 932 933 /* 934 * Wrappers for get/setitimer for debugging.. 935 */ 936 int 937 linux_setitimer(struct thread *td, struct linux_setitimer_args *args) 938 { 939 struct setitimer_args bsa; 940 struct itimerval foo; 941 int error; 942 943 #ifdef DEBUG 944 if (ldebug(setitimer)) 945 printf(ARGS(setitimer, "%p, %p"), 946 (void *)args->itv, (void *)args->oitv); 947 #endif 948 bsa.which = args->which; 949 bsa.itv = (struct itimerval *)args->itv; 950 bsa.oitv = (struct itimerval *)args->oitv; 951 if (args->itv) { 952 if ((error = copyin(args->itv, &foo, sizeof(foo)))) 953 return error; 954 #ifdef DEBUG 955 if (ldebug(setitimer)) { 956 printf("setitimer: value: sec: %ld, usec: %ld\n", 957 foo.it_value.tv_sec, foo.it_value.tv_usec); 958 printf("setitimer: interval: sec: %ld, usec: %ld\n", 959 foo.it_interval.tv_sec, foo.it_interval.tv_usec); 960 } 961 #endif 962 } 963 return setitimer(td, &bsa); 964 } 965 966 int 967 linux_getitimer(struct thread *td, struct linux_getitimer_args *args) 968 { 969 struct getitimer_args bsa; 970 #ifdef DEBUG 971 if (ldebug(getitimer)) 972 printf(ARGS(getitimer, "%p"), (void *)args->itv); 973 #endif 974 bsa.which = args->which; 975 bsa.itv = (struct itimerval *)args->itv; 976 return getitimer(td, &bsa); 977 } 978 979 #ifndef __alpha__ 980 int 981 linux_nice(struct thread *td, struct linux_nice_args *args) 982 { 983 struct setpriority_args bsd_args; 984 985 bsd_args.which = PRIO_PROCESS; 986 bsd_args.who = 0; /* current process */ 987 bsd_args.prio = args->inc; 988 return setpriority(td, &bsd_args); 989 } 990 #endif /*!__alpha__*/ 991 992 int 993 linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 994 { 995 struct ucred *newcred, *oldcred; 996 l_gid_t linux_gidset[NGROUPS]; 997 gid_t *bsd_gidset; 998 int ngrp, error; 999 struct proc *p; 1000 1001 ngrp = args->gidsetsize; 1002 if (ngrp < 0 || ngrp >= NGROUPS) 1003 return (EINVAL); 1004 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1005 if (error) 1006 return (error); 1007 newcred = crget(); 1008 p = td->td_proc; 1009 PROC_LOCK(p); 1010 oldcred = p->p_ucred; 1011 1012 /* 1013 * cr_groups[0] holds egid. Setting the whole set from 1014 * the supplied set will cause egid to be changed too. 1015 * Keep cr_groups[0] unchanged to prevent that. 1016 */ 1017 1018 if ((error = suser_cred(oldcred, PRISON_ROOT)) != 0) { 1019 PROC_UNLOCK(p); 1020 crfree(newcred); 1021 return (error); 1022 } 1023 1024 crcopy(newcred, oldcred); 1025 if (ngrp > 0) { 1026 newcred->cr_ngroups = ngrp + 1; 1027 1028 bsd_gidset = newcred->cr_groups; 1029 ngrp--; 1030 while (ngrp >= 0) { 1031 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1032 ngrp--; 1033 } 1034 } 1035 else 1036 newcred->cr_ngroups = 1; 1037 1038 setsugid(p); 1039 p->p_ucred = newcred; 1040 PROC_UNLOCK(p); 1041 crfree(oldcred); 1042 return (0); 1043 } 1044 1045 int 1046 linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1047 { 1048 struct ucred *cred; 1049 l_gid_t linux_gidset[NGROUPS]; 1050 gid_t *bsd_gidset; 1051 int bsd_gidsetsz, ngrp, error; 1052 1053 cred = td->td_ucred; 1054 bsd_gidset = cred->cr_groups; 1055 bsd_gidsetsz = cred->cr_ngroups - 1; 1056 1057 /* 1058 * cr_groups[0] holds egid. Returning the whole set 1059 * here will cause a duplicate. Exclude cr_groups[0] 1060 * to prevent that. 1061 */ 1062 1063 if ((ngrp = args->gidsetsize) == 0) { 1064 td->td_retval[0] = bsd_gidsetsz; 1065 return (0); 1066 } 1067 1068 if (ngrp < bsd_gidsetsz) 1069 return (EINVAL); 1070 1071 ngrp = 0; 1072 while (ngrp < bsd_gidsetsz) { 1073 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1074 ngrp++; 1075 } 1076 1077 if ((error = copyout(linux_gidset, args->grouplist, 1078 ngrp * sizeof(l_gid_t)))) 1079 return (error); 1080 1081 td->td_retval[0] = ngrp; 1082 return (0); 1083 } 1084 1085 #ifndef __alpha__ 1086 int 1087 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1088 { 1089 struct rlimit bsd_rlim; 1090 struct l_rlimit rlim; 1091 u_int which; 1092 int error; 1093 1094 #ifdef DEBUG 1095 if (ldebug(setrlimit)) 1096 printf(ARGS(setrlimit, "%d, %p"), 1097 args->resource, (void *)args->rlim); 1098 #endif 1099 1100 if (args->resource >= LINUX_RLIM_NLIMITS) 1101 return (EINVAL); 1102 1103 which = linux_to_bsd_resource[args->resource]; 1104 if (which == -1) 1105 return (EINVAL); 1106 1107 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1108 if (error) 1109 return (error); 1110 1111 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1112 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1113 return (dosetrlimit(td, which, &bsd_rlim)); 1114 } 1115 1116 int 1117 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1118 { 1119 struct l_rlimit rlim; 1120 struct proc *p = td->td_proc; 1121 struct rlimit *bsd_rlp; 1122 u_int which; 1123 1124 #ifdef DEBUG 1125 if (ldebug(old_getrlimit)) 1126 printf(ARGS(old_getrlimit, "%d, %p"), 1127 args->resource, (void *)args->rlim); 1128 #endif 1129 1130 if (args->resource >= LINUX_RLIM_NLIMITS) 1131 return (EINVAL); 1132 1133 which = linux_to_bsd_resource[args->resource]; 1134 if (which == -1) 1135 return (EINVAL); 1136 bsd_rlp = &p->p_rlimit[which]; 1137 1138 rlim.rlim_cur = (unsigned long)bsd_rlp->rlim_cur; 1139 if (rlim.rlim_cur == ULONG_MAX) 1140 rlim.rlim_cur = LONG_MAX; 1141 rlim.rlim_max = (unsigned long)bsd_rlp->rlim_max; 1142 if (rlim.rlim_max == ULONG_MAX) 1143 rlim.rlim_max = LONG_MAX; 1144 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1145 } 1146 1147 int 1148 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1149 { 1150 struct l_rlimit rlim; 1151 struct proc *p = td->td_proc; 1152 struct rlimit *bsd_rlp; 1153 u_int which; 1154 1155 #ifdef DEBUG 1156 if (ldebug(getrlimit)) 1157 printf(ARGS(getrlimit, "%d, %p"), 1158 args->resource, (void *)args->rlim); 1159 #endif 1160 1161 if (args->resource >= LINUX_RLIM_NLIMITS) 1162 return (EINVAL); 1163 1164 which = linux_to_bsd_resource[args->resource]; 1165 if (which == -1) 1166 return (EINVAL); 1167 bsd_rlp = &p->p_rlimit[which]; 1168 1169 rlim.rlim_cur = (l_ulong)bsd_rlp->rlim_cur; 1170 rlim.rlim_max = (l_ulong)bsd_rlp->rlim_max; 1171 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1172 } 1173 #endif /*!__alpha__*/ 1174 1175 int 1176 linux_sched_setscheduler(struct thread *td, 1177 struct linux_sched_setscheduler_args *args) 1178 { 1179 struct sched_setscheduler_args bsd; 1180 1181 #ifdef DEBUG 1182 if (ldebug(sched_setscheduler)) 1183 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1184 args->pid, args->policy, (const void *)args->param); 1185 #endif 1186 1187 switch (args->policy) { 1188 case LINUX_SCHED_OTHER: 1189 bsd.policy = SCHED_OTHER; 1190 break; 1191 case LINUX_SCHED_FIFO: 1192 bsd.policy = SCHED_FIFO; 1193 break; 1194 case LINUX_SCHED_RR: 1195 bsd.policy = SCHED_RR; 1196 break; 1197 default: 1198 return EINVAL; 1199 } 1200 1201 bsd.pid = args->pid; 1202 bsd.param = (struct sched_param *)args->param; 1203 return sched_setscheduler(td, &bsd); 1204 } 1205 1206 int 1207 linux_sched_getscheduler(struct thread *td, 1208 struct linux_sched_getscheduler_args *args) 1209 { 1210 struct sched_getscheduler_args bsd; 1211 int error; 1212 1213 #ifdef DEBUG 1214 if (ldebug(sched_getscheduler)) 1215 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1216 #endif 1217 1218 bsd.pid = args->pid; 1219 error = sched_getscheduler(td, &bsd); 1220 1221 switch (td->td_retval[0]) { 1222 case SCHED_OTHER: 1223 td->td_retval[0] = LINUX_SCHED_OTHER; 1224 break; 1225 case SCHED_FIFO: 1226 td->td_retval[0] = LINUX_SCHED_FIFO; 1227 break; 1228 case SCHED_RR: 1229 td->td_retval[0] = LINUX_SCHED_RR; 1230 break; 1231 } 1232 1233 return error; 1234 } 1235 1236 int 1237 linux_sched_get_priority_max(struct thread *td, 1238 struct linux_sched_get_priority_max_args *args) 1239 { 1240 struct sched_get_priority_max_args bsd; 1241 1242 #ifdef DEBUG 1243 if (ldebug(sched_get_priority_max)) 1244 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1245 #endif 1246 1247 switch (args->policy) { 1248 case LINUX_SCHED_OTHER: 1249 bsd.policy = SCHED_OTHER; 1250 break; 1251 case LINUX_SCHED_FIFO: 1252 bsd.policy = SCHED_FIFO; 1253 break; 1254 case LINUX_SCHED_RR: 1255 bsd.policy = SCHED_RR; 1256 break; 1257 default: 1258 return EINVAL; 1259 } 1260 return sched_get_priority_max(td, &bsd); 1261 } 1262 1263 int 1264 linux_sched_get_priority_min(struct thread *td, 1265 struct linux_sched_get_priority_min_args *args) 1266 { 1267 struct sched_get_priority_min_args bsd; 1268 1269 #ifdef DEBUG 1270 if (ldebug(sched_get_priority_min)) 1271 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1272 #endif 1273 1274 switch (args->policy) { 1275 case LINUX_SCHED_OTHER: 1276 bsd.policy = SCHED_OTHER; 1277 break; 1278 case LINUX_SCHED_FIFO: 1279 bsd.policy = SCHED_FIFO; 1280 break; 1281 case LINUX_SCHED_RR: 1282 bsd.policy = SCHED_RR; 1283 break; 1284 default: 1285 return EINVAL; 1286 } 1287 return sched_get_priority_min(td, &bsd); 1288 } 1289 1290 #define REBOOT_CAD_ON 0x89abcdef 1291 #define REBOOT_CAD_OFF 0 1292 #define REBOOT_HALT 0xcdef0123 1293 1294 int 1295 linux_reboot(struct thread *td, struct linux_reboot_args *args) 1296 { 1297 struct reboot_args bsd_args; 1298 1299 #ifdef DEBUG 1300 if (ldebug(reboot)) 1301 printf(ARGS(reboot, "0x%x"), args->cmd); 1302 #endif 1303 if (args->cmd == REBOOT_CAD_ON || args->cmd == REBOOT_CAD_OFF) 1304 return (0); 1305 bsd_args.opt = (args->cmd == REBOOT_HALT) ? RB_HALT : 0; 1306 return (reboot(td, &bsd_args)); 1307 } 1308 1309 #ifndef __alpha__ 1310 1311 /* 1312 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1313 * td->td_retval[1] when COMPAT_43 or COMPAT_SUNOS is defined. This 1314 * globbers registers that are assumed to be preserved. The following 1315 * lightweight syscalls fixes this. See also linux_getgid16() and 1316 * linux_getuid16() in linux_uid16.c. 1317 * 1318 * linux_getpid() - MP SAFE 1319 * linux_getgid() - MP SAFE 1320 * linux_getuid() - MP SAFE 1321 */ 1322 1323 int 1324 linux_getpid(struct thread *td, struct linux_getpid_args *args) 1325 { 1326 1327 td->td_retval[0] = td->td_proc->p_pid; 1328 return (0); 1329 } 1330 1331 int 1332 linux_getgid(struct thread *td, struct linux_getgid_args *args) 1333 { 1334 1335 td->td_retval[0] = td->td_ucred->cr_rgid; 1336 return (0); 1337 } 1338 1339 int 1340 linux_getuid(struct thread *td, struct linux_getuid_args *args) 1341 { 1342 1343 td->td_retval[0] = td->td_ucred->cr_ruid; 1344 return (0); 1345 } 1346 1347 #endif /*!__alpha__*/ 1348 1349 int 1350 linux_getsid(struct thread *td, struct linux_getsid_args *args) 1351 { 1352 struct getsid_args bsd; 1353 bsd.pid = args->pid; 1354 return getsid(td, &bsd); 1355 } 1356