1 /*- 2 * Copyright (c) 2002 Doug Rabson 3 * Copyright (c) 1994-1995 S�ren Schmidt 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer 11 * in this position and unchanged. 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. The name of the author may not be used to endorse or promote products 16 * derived from this software without specific prior written permission 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 28 */ 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include "opt_compat.h" 34 #include "opt_mac.h" 35 36 #include <sys/param.h> 37 #include <sys/blist.h> 38 #include <sys/fcntl.h> 39 #if defined(__i386__) 40 #include <sys/imgact_aout.h> 41 #endif 42 #include <sys/jail.h> 43 #include <sys/kernel.h> 44 #include <sys/limits.h> 45 #include <sys/lock.h> 46 #include <sys/mac.h> 47 #include <sys/malloc.h> 48 #include <sys/mman.h> 49 #include <sys/mount.h> 50 #include <sys/mutex.h> 51 #include <sys/namei.h> 52 #include <sys/proc.h> 53 #include <sys/reboot.h> 54 #include <sys/resourcevar.h> 55 #include <sys/signalvar.h> 56 #include <sys/stat.h> 57 #include <sys/syscallsubr.h> 58 #include <sys/sysctl.h> 59 #include <sys/sysproto.h> 60 #include <sys/systm.h> 61 #include <sys/time.h> 62 #include <sys/vmmeter.h> 63 #include <sys/vnode.h> 64 #include <sys/wait.h> 65 66 #include <vm/vm.h> 67 #include <vm/pmap.h> 68 #include <vm/vm_kern.h> 69 #include <vm/vm_map.h> 70 #include <vm/vm_extern.h> 71 #include <vm/vm_object.h> 72 #include <vm/swap_pager.h> 73 74 #include <posix4/sched.h> 75 76 #include <compat/linux/linux_sysproto.h> 77 78 #ifdef COMPAT_LINUX32 79 #include <machine/../linux32/linux.h> 80 #include <machine/../linux32/linux32_proto.h> 81 #else 82 #include <machine/../linux/linux.h> 83 #include <machine/../linux/linux_proto.h> 84 #endif 85 86 #include <compat/linux/linux_mib.h> 87 #include <compat/linux/linux_util.h> 88 89 #ifdef __i386__ 90 #include <machine/cputypes.h> 91 #endif 92 93 #define BSD_TO_LINUX_SIGNAL(sig) \ 94 (((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig) 95 96 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 97 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 98 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 99 RLIMIT_MEMLOCK, -1 100 }; 101 102 struct l_sysinfo { 103 l_long uptime; /* Seconds since boot */ 104 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 105 #define LINUX_SYSINFO_LOADS_SCALE 65536 106 l_ulong totalram; /* Total usable main memory size */ 107 l_ulong freeram; /* Available memory size */ 108 l_ulong sharedram; /* Amount of shared memory */ 109 l_ulong bufferram; /* Memory used by buffers */ 110 l_ulong totalswap; /* Total swap space size */ 111 l_ulong freeswap; /* swap space still available */ 112 l_ushort procs; /* Number of current processes */ 113 l_ulong totalbig; 114 l_ulong freebig; 115 l_uint mem_unit; 116 char _f[6]; /* Pads structure to 64 bytes */ 117 }; 118 int 119 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 120 { 121 struct l_sysinfo sysinfo; 122 vm_object_t object; 123 int i, j; 124 struct timespec ts; 125 126 getnanouptime(&ts); 127 if (ts.tv_nsec != 0) 128 ts.tv_sec++; 129 sysinfo.uptime = ts.tv_sec; 130 131 /* Use the information from the mib to get our load averages */ 132 for (i = 0; i < 3; i++) 133 sysinfo.loads[i] = averunnable.ldavg[i] * 134 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 135 136 sysinfo.totalram = physmem * PAGE_SIZE; 137 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE; 138 139 sysinfo.sharedram = 0; 140 mtx_lock(&vm_object_list_mtx); 141 TAILQ_FOREACH(object, &vm_object_list, object_list) 142 if (object->shadow_count > 1) 143 sysinfo.sharedram += object->resident_page_count; 144 mtx_unlock(&vm_object_list_mtx); 145 146 sysinfo.sharedram *= PAGE_SIZE; 147 sysinfo.bufferram = 0; 148 149 swap_pager_status(&i, &j); 150 sysinfo.totalswap= i * PAGE_SIZE; 151 sysinfo.freeswap = (i - j) * PAGE_SIZE; 152 153 sysinfo.procs = nprocs; 154 155 /* The following are only present in newer Linux kernels. */ 156 sysinfo.totalbig = 0; 157 sysinfo.freebig = 0; 158 sysinfo.mem_unit = 1; 159 160 return copyout(&sysinfo, args->info, sizeof(sysinfo)); 161 } 162 163 int 164 linux_alarm(struct thread *td, struct linux_alarm_args *args) 165 { 166 struct itimerval it, old_it; 167 int error; 168 169 #ifdef DEBUG 170 if (ldebug(alarm)) 171 printf(ARGS(alarm, "%u"), args->secs); 172 #endif 173 174 if (args->secs > 100000000) 175 return (EINVAL); 176 177 it.it_value.tv_sec = (long)args->secs; 178 it.it_value.tv_usec = 0; 179 it.it_interval.tv_sec = 0; 180 it.it_interval.tv_usec = 0; 181 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 182 if (error) 183 return (error); 184 if (timevalisset(&old_it.it_value)) { 185 if (old_it.it_value.tv_usec != 0) 186 old_it.it_value.tv_sec++; 187 td->td_retval[0] = old_it.it_value.tv_sec; 188 } 189 return (0); 190 } 191 192 int 193 linux_brk(struct thread *td, struct linux_brk_args *args) 194 { 195 struct vmspace *vm = td->td_proc->p_vmspace; 196 vm_offset_t new, old; 197 struct obreak_args /* { 198 char * nsize; 199 } */ tmp; 200 201 #ifdef DEBUG 202 if (ldebug(brk)) 203 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 204 #endif 205 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 206 new = (vm_offset_t)args->dsend; 207 tmp.nsize = (char *) new; 208 if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp)) 209 td->td_retval[0] = (long)new; 210 else 211 td->td_retval[0] = (long)old; 212 213 return 0; 214 } 215 216 #if defined(__i386__) 217 /* XXX: what about amd64/linux32? */ 218 219 int 220 linux_uselib(struct thread *td, struct linux_uselib_args *args) 221 { 222 struct nameidata ni; 223 struct vnode *vp; 224 struct exec *a_out; 225 struct vattr attr; 226 vm_offset_t vmaddr; 227 unsigned long file_offset; 228 vm_offset_t buffer; 229 unsigned long bss_size; 230 char *library; 231 int error; 232 int locked, vfslocked; 233 234 LCONVPATHEXIST(td, args->library, &library); 235 236 #ifdef DEBUG 237 if (ldebug(uselib)) 238 printf(ARGS(uselib, "%s"), library); 239 #endif 240 241 a_out = NULL; 242 vfslocked = 0; 243 locked = 0; 244 vp = NULL; 245 246 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1, 247 UIO_SYSSPACE, library, td); 248 error = namei(&ni); 249 LFREEPATH(library); 250 if (error) 251 goto cleanup; 252 253 vp = ni.ni_vp; 254 vfslocked = NDHASGIANT(&ni); 255 NDFREE(&ni, NDF_ONLY_PNBUF); 256 257 /* 258 * From here on down, we have a locked vnode that must be unlocked. 259 * XXX: The code below largely duplicates exec_check_permissions(). 260 */ 261 locked = 1; 262 263 /* Writable? */ 264 if (vp->v_writecount) { 265 error = ETXTBSY; 266 goto cleanup; 267 } 268 269 /* Executable? */ 270 error = VOP_GETATTR(vp, &attr, td->td_ucred, td); 271 if (error) 272 goto cleanup; 273 274 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 275 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 276 /* EACCESS is what exec(2) returns. */ 277 error = ENOEXEC; 278 goto cleanup; 279 } 280 281 /* Sensible size? */ 282 if (attr.va_size == 0) { 283 error = ENOEXEC; 284 goto cleanup; 285 } 286 287 /* Can we access it? */ 288 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 289 if (error) 290 goto cleanup; 291 292 /* 293 * XXX: This should use vn_open() so that it is properly authorized, 294 * and to reduce code redundancy all over the place here. 295 * XXX: Not really, it duplicates far more of exec_check_permissions() 296 * than vn_open(). 297 */ 298 #ifdef MAC 299 error = mac_check_vnode_open(td->td_ucred, vp, FREAD); 300 if (error) 301 goto cleanup; 302 #endif 303 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, -1); 304 if (error) 305 goto cleanup; 306 307 /* Pull in executable header into kernel_map */ 308 error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE, 309 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 310 if (error) 311 goto cleanup; 312 313 /* Is it a Linux binary ? */ 314 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 315 error = ENOEXEC; 316 goto cleanup; 317 } 318 319 /* 320 * While we are here, we should REALLY do some more checks 321 */ 322 323 /* Set file/virtual offset based on a.out variant. */ 324 switch ((int)(a_out->a_magic & 0xffff)) { 325 case 0413: /* ZMAGIC */ 326 file_offset = 1024; 327 break; 328 case 0314: /* QMAGIC */ 329 file_offset = 0; 330 break; 331 default: 332 error = ENOEXEC; 333 goto cleanup; 334 } 335 336 bss_size = round_page(a_out->a_bss); 337 338 /* Check various fields in header for validity/bounds. */ 339 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 340 error = ENOEXEC; 341 goto cleanup; 342 } 343 344 /* text + data can't exceed file size */ 345 if (a_out->a_data + a_out->a_text > attr.va_size) { 346 error = EFAULT; 347 goto cleanup; 348 } 349 350 /* 351 * text/data/bss must not exceed limits 352 * XXX - this is not complete. it should check current usage PLUS 353 * the resources needed by this library. 354 */ 355 PROC_LOCK(td->td_proc); 356 if (a_out->a_text > maxtsiz || 357 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) { 358 PROC_UNLOCK(td->td_proc); 359 error = ENOMEM; 360 goto cleanup; 361 } 362 PROC_UNLOCK(td->td_proc); 363 364 /* 365 * Prevent more writers. 366 * XXX: Note that if any of the VM operations fail below we don't 367 * clear this flag. 368 */ 369 vp->v_vflag |= VV_TEXT; 370 371 /* 372 * Lock no longer needed 373 */ 374 locked = 0; 375 VOP_UNLOCK(vp, 0, td); 376 VFS_UNLOCK_GIANT(vfslocked); 377 378 /* 379 * Check if file_offset page aligned. Currently we cannot handle 380 * misalinged file offsets, and so we read in the entire image 381 * (what a waste). 382 */ 383 if (file_offset & PAGE_MASK) { 384 #ifdef DEBUG 385 printf("uselib: Non page aligned binary %lu\n", file_offset); 386 #endif 387 /* Map text+data read/write/execute */ 388 389 /* a_entry is the load address and is page aligned */ 390 vmaddr = trunc_page(a_out->a_entry); 391 392 /* get anon user mapping, read+write+execute */ 393 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 394 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL, 395 VM_PROT_ALL, 0); 396 if (error) 397 goto cleanup; 398 399 /* map file into kernel_map */ 400 error = vm_mmap(kernel_map, &buffer, 401 round_page(a_out->a_text + a_out->a_data + file_offset), 402 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 403 trunc_page(file_offset)); 404 if (error) 405 goto cleanup; 406 407 /* copy from kernel VM space to user space */ 408 error = copyout(PTRIN(buffer + file_offset), 409 (void *)vmaddr, a_out->a_text + a_out->a_data); 410 411 /* release temporary kernel space */ 412 vm_map_remove(kernel_map, buffer, buffer + 413 round_page(a_out->a_text + a_out->a_data + file_offset)); 414 415 if (error) 416 goto cleanup; 417 } else { 418 #ifdef DEBUG 419 printf("uselib: Page aligned binary %lu\n", file_offset); 420 #endif 421 /* 422 * for QMAGIC, a_entry is 20 bytes beyond the load address 423 * to skip the executable header 424 */ 425 vmaddr = trunc_page(a_out->a_entry); 426 427 /* 428 * Map it all into the process's space as a single 429 * copy-on-write "data" segment. 430 */ 431 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 432 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 433 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 434 if (error) 435 goto cleanup; 436 } 437 #ifdef DEBUG 438 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long*)vmaddr)[0], 439 ((long*)vmaddr)[1]); 440 #endif 441 if (bss_size != 0) { 442 /* Calculate BSS start address */ 443 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 444 a_out->a_data; 445 446 /* allocate some 'anon' space */ 447 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 448 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0); 449 if (error) 450 goto cleanup; 451 } 452 453 cleanup: 454 /* Unlock vnode if needed */ 455 if (locked) { 456 VOP_UNLOCK(vp, 0, td); 457 VFS_UNLOCK_GIANT(vfslocked); 458 } 459 460 /* Release the kernel mapping. */ 461 if (a_out) 462 vm_map_remove(kernel_map, (vm_offset_t)a_out, 463 (vm_offset_t)a_out + PAGE_SIZE); 464 465 return error; 466 } 467 468 #endif /* __i386__ */ 469 470 int 471 linux_select(struct thread *td, struct linux_select_args *args) 472 { 473 l_timeval ltv; 474 struct timeval tv0, tv1, utv, *tvp; 475 int error; 476 477 #ifdef DEBUG 478 if (ldebug(select)) 479 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 480 (void *)args->readfds, (void *)args->writefds, 481 (void *)args->exceptfds, (void *)args->timeout); 482 #endif 483 484 /* 485 * Store current time for computation of the amount of 486 * time left. 487 */ 488 if (args->timeout) { 489 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 490 goto select_out; 491 utv.tv_sec = ltv.tv_sec; 492 utv.tv_usec = ltv.tv_usec; 493 #ifdef DEBUG 494 if (ldebug(select)) 495 printf(LMSG("incoming timeout (%jd/%ld)"), 496 (intmax_t)utv.tv_sec, utv.tv_usec); 497 #endif 498 499 if (itimerfix(&utv)) { 500 /* 501 * The timeval was invalid. Convert it to something 502 * valid that will act as it does under Linux. 503 */ 504 utv.tv_sec += utv.tv_usec / 1000000; 505 utv.tv_usec %= 1000000; 506 if (utv.tv_usec < 0) { 507 utv.tv_sec -= 1; 508 utv.tv_usec += 1000000; 509 } 510 if (utv.tv_sec < 0) 511 timevalclear(&utv); 512 } 513 microtime(&tv0); 514 tvp = &utv; 515 } else 516 tvp = NULL; 517 518 error = kern_select(td, args->nfds, args->readfds, args->writefds, 519 args->exceptfds, tvp); 520 521 #ifdef DEBUG 522 if (ldebug(select)) 523 printf(LMSG("real select returns %d"), error); 524 #endif 525 if (error) { 526 /* 527 * See fs/select.c in the Linux kernel. Without this, 528 * Maelstrom doesn't work. 529 */ 530 if (error == ERESTART) 531 error = EINTR; 532 goto select_out; 533 } 534 535 if (args->timeout) { 536 if (td->td_retval[0]) { 537 /* 538 * Compute how much time was left of the timeout, 539 * by subtracting the current time and the time 540 * before we started the call, and subtracting 541 * that result from the user-supplied value. 542 */ 543 microtime(&tv1); 544 timevalsub(&tv1, &tv0); 545 timevalsub(&utv, &tv1); 546 if (utv.tv_sec < 0) 547 timevalclear(&utv); 548 } else 549 timevalclear(&utv); 550 #ifdef DEBUG 551 if (ldebug(select)) 552 printf(LMSG("outgoing timeout (%jd/%ld)"), 553 (intmax_t)utv.tv_sec, utv.tv_usec); 554 #endif 555 ltv.tv_sec = utv.tv_sec; 556 ltv.tv_usec = utv.tv_usec; 557 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 558 goto select_out; 559 } 560 561 select_out: 562 #ifdef DEBUG 563 if (ldebug(select)) 564 printf(LMSG("select_out -> %d"), error); 565 #endif 566 return error; 567 } 568 569 int 570 linux_mremap(struct thread *td, struct linux_mremap_args *args) 571 { 572 struct munmap_args /* { 573 void *addr; 574 size_t len; 575 } */ bsd_args; 576 int error = 0; 577 578 #ifdef DEBUG 579 if (ldebug(mremap)) 580 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 581 (void *)(uintptr_t)args->addr, 582 (unsigned long)args->old_len, 583 (unsigned long)args->new_len, 584 (unsigned long)args->flags); 585 #endif 586 args->new_len = round_page(args->new_len); 587 args->old_len = round_page(args->old_len); 588 589 if (args->new_len > args->old_len) { 590 td->td_retval[0] = 0; 591 return ENOMEM; 592 } 593 594 if (args->new_len < args->old_len) { 595 bsd_args.addr = 596 (caddr_t)((uintptr_t)args->addr + args->new_len); 597 bsd_args.len = args->old_len - args->new_len; 598 error = munmap(td, &bsd_args); 599 } 600 601 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 602 return error; 603 } 604 605 #define LINUX_MS_ASYNC 0x0001 606 #define LINUX_MS_INVALIDATE 0x0002 607 #define LINUX_MS_SYNC 0x0004 608 609 int 610 linux_msync(struct thread *td, struct linux_msync_args *args) 611 { 612 struct msync_args bsd_args; 613 614 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 615 bsd_args.len = (uintptr_t)args->len; 616 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 617 618 return msync(td, &bsd_args); 619 } 620 621 int 622 linux_time(struct thread *td, struct linux_time_args *args) 623 { 624 struct timeval tv; 625 l_time_t tm; 626 int error; 627 628 #ifdef DEBUG 629 if (ldebug(time)) 630 printf(ARGS(time, "*")); 631 #endif 632 633 microtime(&tv); 634 tm = tv.tv_sec; 635 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 636 return error; 637 td->td_retval[0] = tm; 638 return 0; 639 } 640 641 struct l_times_argv { 642 l_long tms_utime; 643 l_long tms_stime; 644 l_long tms_cutime; 645 l_long tms_cstime; 646 }; 647 648 #define CLK_TCK 100 /* Linux uses 100 */ 649 650 #define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 651 652 int 653 linux_times(struct thread *td, struct linux_times_args *args) 654 { 655 struct timeval tv, utime, stime, cutime, cstime; 656 struct l_times_argv tms; 657 struct proc *p; 658 int error; 659 660 #ifdef DEBUG 661 if (ldebug(times)) 662 printf(ARGS(times, "*")); 663 #endif 664 665 if (args->buf != NULL) { 666 p = td->td_proc; 667 PROC_LOCK(p); 668 calcru(p, &utime, &stime); 669 calccru(p, &cutime, &cstime); 670 PROC_UNLOCK(p); 671 672 tms.tms_utime = CONVTCK(utime); 673 tms.tms_stime = CONVTCK(stime); 674 675 tms.tms_cutime = CONVTCK(cutime); 676 tms.tms_cstime = CONVTCK(cstime); 677 678 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 679 return error; 680 } 681 682 microuptime(&tv); 683 td->td_retval[0] = (int)CONVTCK(tv); 684 return 0; 685 } 686 687 int 688 linux_newuname(struct thread *td, struct linux_newuname_args *args) 689 { 690 struct l_new_utsname utsname; 691 char osname[LINUX_MAX_UTSNAME]; 692 char osrelease[LINUX_MAX_UTSNAME]; 693 char *p; 694 695 #ifdef DEBUG 696 if (ldebug(newuname)) 697 printf(ARGS(newuname, "*")); 698 #endif 699 700 linux_get_osname(td, osname); 701 linux_get_osrelease(td, osrelease); 702 703 bzero(&utsname, sizeof(utsname)); 704 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 705 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 706 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 707 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 708 for (p = utsname.version; *p != '\0'; ++p) 709 if (*p == '\n') { 710 *p = '\0'; 711 break; 712 } 713 #ifdef __i386__ 714 { 715 const char *class; 716 switch (cpu_class) { 717 case CPUCLASS_686: 718 class = "i686"; 719 break; 720 case CPUCLASS_586: 721 class = "i586"; 722 break; 723 case CPUCLASS_486: 724 class = "i486"; 725 break; 726 default: 727 class = "i386"; 728 } 729 strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME); 730 } 731 #elif defined(__amd64__) /* XXX: Linux can change 'personality'. */ 732 #ifdef COMPAT_LINUX32 733 strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME); 734 #else 735 strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME); 736 #endif /* COMPAT_LINUX32 */ 737 #else /* something other than i386 or amd64 - assume we and Linux agree */ 738 strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME); 739 #endif /* __i386__ */ 740 strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME); 741 742 return (copyout(&utsname, args->buf, sizeof(utsname))); 743 } 744 745 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 746 struct l_utimbuf { 747 l_time_t l_actime; 748 l_time_t l_modtime; 749 }; 750 751 int 752 linux_utime(struct thread *td, struct linux_utime_args *args) 753 { 754 struct timeval tv[2], *tvp; 755 struct l_utimbuf lut; 756 char *fname; 757 int error; 758 759 LCONVPATHEXIST(td, args->fname, &fname); 760 761 #ifdef DEBUG 762 if (ldebug(utime)) 763 printf(ARGS(utime, "%s, *"), fname); 764 #endif 765 766 if (args->times) { 767 if ((error = copyin(args->times, &lut, sizeof lut))) { 768 LFREEPATH(fname); 769 return error; 770 } 771 tv[0].tv_sec = lut.l_actime; 772 tv[0].tv_usec = 0; 773 tv[1].tv_sec = lut.l_modtime; 774 tv[1].tv_usec = 0; 775 tvp = tv; 776 } else 777 tvp = NULL; 778 779 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 780 LFREEPATH(fname); 781 return (error); 782 } 783 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 784 785 #define __WCLONE 0x80000000 786 787 int 788 linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 789 { 790 int error, options, tmpstat; 791 792 #ifdef DEBUG 793 if (ldebug(waitpid)) 794 printf(ARGS(waitpid, "%d, %p, %d"), 795 args->pid, (void *)args->status, args->options); 796 #endif 797 798 options = (args->options & (WNOHANG | WUNTRACED)); 799 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 800 if (args->options & __WCLONE) 801 options |= WLINUXCLONE; 802 803 error = kern_wait(td, args->pid, &tmpstat, options, NULL); 804 if (error) 805 return error; 806 807 if (args->status) { 808 tmpstat &= 0xffff; 809 if (WIFSIGNALED(tmpstat)) 810 tmpstat = (tmpstat & 0xffffff80) | 811 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 812 else if (WIFSTOPPED(tmpstat)) 813 tmpstat = (tmpstat & 0xffff00ff) | 814 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 815 return copyout(&tmpstat, args->status, sizeof(int)); 816 } 817 818 return 0; 819 } 820 821 int 822 linux_wait4(struct thread *td, struct linux_wait4_args *args) 823 { 824 int error, options, tmpstat; 825 struct rusage ru, *rup; 826 struct proc *p; 827 828 #ifdef DEBUG 829 if (ldebug(wait4)) 830 printf(ARGS(wait4, "%d, %p, %d, %p"), 831 args->pid, (void *)args->status, args->options, 832 (void *)args->rusage); 833 #endif 834 835 options = (args->options & (WNOHANG | WUNTRACED)); 836 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 837 if (args->options & __WCLONE) 838 options |= WLINUXCLONE; 839 840 if (args->rusage != NULL) 841 rup = &ru; 842 else 843 rup = NULL; 844 error = kern_wait(td, args->pid, &tmpstat, options, rup); 845 if (error) 846 return error; 847 848 p = td->td_proc; 849 PROC_LOCK(p); 850 sigqueue_delete(&p->p_sigqueue, SIGCHLD); 851 PROC_UNLOCK(p); 852 853 if (args->status) { 854 tmpstat &= 0xffff; 855 if (WIFSIGNALED(tmpstat)) 856 tmpstat = (tmpstat & 0xffffff80) | 857 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 858 else if (WIFSTOPPED(tmpstat)) 859 tmpstat = (tmpstat & 0xffff00ff) | 860 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 861 error = copyout(&tmpstat, args->status, sizeof(int)); 862 } 863 if (args->rusage != NULL && error == 0) 864 error = copyout(&ru, args->rusage, sizeof(ru)); 865 866 return (error); 867 } 868 869 int 870 linux_mknod(struct thread *td, struct linux_mknod_args *args) 871 { 872 char *path; 873 int error; 874 875 LCONVPATHCREAT(td, args->path, &path); 876 877 #ifdef DEBUG 878 if (ldebug(mknod)) 879 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev); 880 #endif 881 882 if (args->mode & S_IFIFO) 883 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 884 else 885 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 886 args->dev); 887 LFREEPATH(path); 888 return (error); 889 } 890 891 /* 892 * UGH! This is just about the dumbest idea I've ever heard!! 893 */ 894 int 895 linux_personality(struct thread *td, struct linux_personality_args *args) 896 { 897 #ifdef DEBUG 898 if (ldebug(personality)) 899 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 900 #endif 901 if (args->per != 0) 902 return EINVAL; 903 904 /* Yes Jim, it's still a Linux... */ 905 td->td_retval[0] = 0; 906 return 0; 907 } 908 909 struct l_itimerval { 910 l_timeval it_interval; 911 l_timeval it_value; 912 }; 913 914 #define B2L_ITIMERVAL(bip, lip) \ 915 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 916 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 917 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 918 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 919 920 int 921 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 922 { 923 int error; 924 struct l_itimerval ls; 925 struct itimerval aitv, oitv; 926 927 #ifdef DEBUG 928 if (ldebug(setitimer)) 929 printf(ARGS(setitimer, "%p, %p"), 930 (void *)uap->itv, (void *)uap->oitv); 931 #endif 932 933 if (uap->itv == NULL) { 934 uap->itv = uap->oitv; 935 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 936 } 937 938 error = copyin(uap->itv, &ls, sizeof(ls)); 939 if (error != 0) 940 return (error); 941 B2L_ITIMERVAL(&aitv, &ls); 942 #ifdef DEBUG 943 if (ldebug(setitimer)) { 944 printf("setitimer: value: sec: %jd, usec: %ld\n", 945 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 946 printf("setitimer: interval: sec: %jd, usec: %ld\n", 947 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 948 } 949 #endif 950 error = kern_setitimer(td, uap->which, &aitv, &oitv); 951 if (error != 0 || uap->oitv == NULL) 952 return (error); 953 B2L_ITIMERVAL(&ls, &oitv); 954 955 return (copyout(&ls, uap->oitv, sizeof(ls))); 956 } 957 958 int 959 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 960 { 961 int error; 962 struct l_itimerval ls; 963 struct itimerval aitv; 964 965 #ifdef DEBUG 966 if (ldebug(getitimer)) 967 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 968 #endif 969 error = kern_getitimer(td, uap->which, &aitv); 970 if (error != 0) 971 return (error); 972 B2L_ITIMERVAL(&ls, &aitv); 973 return (copyout(&ls, uap->itv, sizeof(ls))); 974 } 975 976 int 977 linux_nice(struct thread *td, struct linux_nice_args *args) 978 { 979 struct setpriority_args bsd_args; 980 981 bsd_args.which = PRIO_PROCESS; 982 bsd_args.who = 0; /* current process */ 983 bsd_args.prio = args->inc; 984 return setpriority(td, &bsd_args); 985 } 986 987 int 988 linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 989 { 990 struct ucred *newcred, *oldcred; 991 l_gid_t linux_gidset[NGROUPS]; 992 gid_t *bsd_gidset; 993 int ngrp, error; 994 struct proc *p; 995 996 ngrp = args->gidsetsize; 997 if (ngrp < 0 || ngrp >= NGROUPS) 998 return (EINVAL); 999 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1000 if (error) 1001 return (error); 1002 newcred = crget(); 1003 p = td->td_proc; 1004 PROC_LOCK(p); 1005 oldcred = p->p_ucred; 1006 1007 /* 1008 * cr_groups[0] holds egid. Setting the whole set from 1009 * the supplied set will cause egid to be changed too. 1010 * Keep cr_groups[0] unchanged to prevent that. 1011 */ 1012 1013 if ((error = suser_cred(oldcred, SUSER_ALLOWJAIL)) != 0) { 1014 PROC_UNLOCK(p); 1015 crfree(newcred); 1016 return (error); 1017 } 1018 1019 crcopy(newcred, oldcred); 1020 if (ngrp > 0) { 1021 newcred->cr_ngroups = ngrp + 1; 1022 1023 bsd_gidset = newcred->cr_groups; 1024 ngrp--; 1025 while (ngrp >= 0) { 1026 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1027 ngrp--; 1028 } 1029 } 1030 else 1031 newcred->cr_ngroups = 1; 1032 1033 setsugid(p); 1034 p->p_ucred = newcred; 1035 PROC_UNLOCK(p); 1036 crfree(oldcred); 1037 return (0); 1038 } 1039 1040 int 1041 linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1042 { 1043 struct ucred *cred; 1044 l_gid_t linux_gidset[NGROUPS]; 1045 gid_t *bsd_gidset; 1046 int bsd_gidsetsz, ngrp, error; 1047 1048 cred = td->td_ucred; 1049 bsd_gidset = cred->cr_groups; 1050 bsd_gidsetsz = cred->cr_ngroups - 1; 1051 1052 /* 1053 * cr_groups[0] holds egid. Returning the whole set 1054 * here will cause a duplicate. Exclude cr_groups[0] 1055 * to prevent that. 1056 */ 1057 1058 if ((ngrp = args->gidsetsize) == 0) { 1059 td->td_retval[0] = bsd_gidsetsz; 1060 return (0); 1061 } 1062 1063 if (ngrp < bsd_gidsetsz) 1064 return (EINVAL); 1065 1066 ngrp = 0; 1067 while (ngrp < bsd_gidsetsz) { 1068 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1069 ngrp++; 1070 } 1071 1072 if ((error = copyout(linux_gidset, args->grouplist, 1073 ngrp * sizeof(l_gid_t)))) 1074 return (error); 1075 1076 td->td_retval[0] = ngrp; 1077 return (0); 1078 } 1079 1080 int 1081 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1082 { 1083 struct rlimit bsd_rlim; 1084 struct l_rlimit rlim; 1085 u_int which; 1086 int error; 1087 1088 #ifdef DEBUG 1089 if (ldebug(setrlimit)) 1090 printf(ARGS(setrlimit, "%d, %p"), 1091 args->resource, (void *)args->rlim); 1092 #endif 1093 1094 if (args->resource >= LINUX_RLIM_NLIMITS) 1095 return (EINVAL); 1096 1097 which = linux_to_bsd_resource[args->resource]; 1098 if (which == -1) 1099 return (EINVAL); 1100 1101 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1102 if (error) 1103 return (error); 1104 1105 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1106 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1107 return (kern_setrlimit(td, which, &bsd_rlim)); 1108 } 1109 1110 int 1111 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1112 { 1113 struct l_rlimit rlim; 1114 struct proc *p = td->td_proc; 1115 struct rlimit bsd_rlim; 1116 u_int which; 1117 1118 #ifdef DEBUG 1119 if (ldebug(old_getrlimit)) 1120 printf(ARGS(old_getrlimit, "%d, %p"), 1121 args->resource, (void *)args->rlim); 1122 #endif 1123 1124 if (args->resource >= LINUX_RLIM_NLIMITS) 1125 return (EINVAL); 1126 1127 which = linux_to_bsd_resource[args->resource]; 1128 if (which == -1) 1129 return (EINVAL); 1130 1131 PROC_LOCK(p); 1132 lim_rlimit(p, which, &bsd_rlim); 1133 PROC_UNLOCK(p); 1134 1135 #ifdef COMPAT_LINUX32 1136 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1137 if (rlim.rlim_cur == UINT_MAX) 1138 rlim.rlim_cur = INT_MAX; 1139 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1140 if (rlim.rlim_max == UINT_MAX) 1141 rlim.rlim_max = INT_MAX; 1142 #else 1143 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1144 if (rlim.rlim_cur == ULONG_MAX) 1145 rlim.rlim_cur = LONG_MAX; 1146 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1147 if (rlim.rlim_max == ULONG_MAX) 1148 rlim.rlim_max = LONG_MAX; 1149 #endif 1150 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1151 } 1152 1153 int 1154 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1155 { 1156 struct l_rlimit rlim; 1157 struct proc *p = td->td_proc; 1158 struct rlimit bsd_rlim; 1159 u_int which; 1160 1161 #ifdef DEBUG 1162 if (ldebug(getrlimit)) 1163 printf(ARGS(getrlimit, "%d, %p"), 1164 args->resource, (void *)args->rlim); 1165 #endif 1166 1167 if (args->resource >= LINUX_RLIM_NLIMITS) 1168 return (EINVAL); 1169 1170 which = linux_to_bsd_resource[args->resource]; 1171 if (which == -1) 1172 return (EINVAL); 1173 1174 PROC_LOCK(p); 1175 lim_rlimit(p, which, &bsd_rlim); 1176 PROC_UNLOCK(p); 1177 1178 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1179 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1180 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1181 } 1182 1183 int 1184 linux_sched_setscheduler(struct thread *td, 1185 struct linux_sched_setscheduler_args *args) 1186 { 1187 struct sched_setscheduler_args bsd; 1188 1189 #ifdef DEBUG 1190 if (ldebug(sched_setscheduler)) 1191 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1192 args->pid, args->policy, (const void *)args->param); 1193 #endif 1194 1195 switch (args->policy) { 1196 case LINUX_SCHED_OTHER: 1197 bsd.policy = SCHED_OTHER; 1198 break; 1199 case LINUX_SCHED_FIFO: 1200 bsd.policy = SCHED_FIFO; 1201 break; 1202 case LINUX_SCHED_RR: 1203 bsd.policy = SCHED_RR; 1204 break; 1205 default: 1206 return EINVAL; 1207 } 1208 1209 bsd.pid = args->pid; 1210 bsd.param = (struct sched_param *)args->param; 1211 return sched_setscheduler(td, &bsd); 1212 } 1213 1214 int 1215 linux_sched_getscheduler(struct thread *td, 1216 struct linux_sched_getscheduler_args *args) 1217 { 1218 struct sched_getscheduler_args bsd; 1219 int error; 1220 1221 #ifdef DEBUG 1222 if (ldebug(sched_getscheduler)) 1223 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1224 #endif 1225 1226 bsd.pid = args->pid; 1227 error = sched_getscheduler(td, &bsd); 1228 1229 switch (td->td_retval[0]) { 1230 case SCHED_OTHER: 1231 td->td_retval[0] = LINUX_SCHED_OTHER; 1232 break; 1233 case SCHED_FIFO: 1234 td->td_retval[0] = LINUX_SCHED_FIFO; 1235 break; 1236 case SCHED_RR: 1237 td->td_retval[0] = LINUX_SCHED_RR; 1238 break; 1239 } 1240 1241 return error; 1242 } 1243 1244 int 1245 linux_sched_get_priority_max(struct thread *td, 1246 struct linux_sched_get_priority_max_args *args) 1247 { 1248 struct sched_get_priority_max_args bsd; 1249 1250 #ifdef DEBUG 1251 if (ldebug(sched_get_priority_max)) 1252 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1253 #endif 1254 1255 switch (args->policy) { 1256 case LINUX_SCHED_OTHER: 1257 bsd.policy = SCHED_OTHER; 1258 break; 1259 case LINUX_SCHED_FIFO: 1260 bsd.policy = SCHED_FIFO; 1261 break; 1262 case LINUX_SCHED_RR: 1263 bsd.policy = SCHED_RR; 1264 break; 1265 default: 1266 return EINVAL; 1267 } 1268 return sched_get_priority_max(td, &bsd); 1269 } 1270 1271 int 1272 linux_sched_get_priority_min(struct thread *td, 1273 struct linux_sched_get_priority_min_args *args) 1274 { 1275 struct sched_get_priority_min_args bsd; 1276 1277 #ifdef DEBUG 1278 if (ldebug(sched_get_priority_min)) 1279 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1280 #endif 1281 1282 switch (args->policy) { 1283 case LINUX_SCHED_OTHER: 1284 bsd.policy = SCHED_OTHER; 1285 break; 1286 case LINUX_SCHED_FIFO: 1287 bsd.policy = SCHED_FIFO; 1288 break; 1289 case LINUX_SCHED_RR: 1290 bsd.policy = SCHED_RR; 1291 break; 1292 default: 1293 return EINVAL; 1294 } 1295 return sched_get_priority_min(td, &bsd); 1296 } 1297 1298 #define REBOOT_CAD_ON 0x89abcdef 1299 #define REBOOT_CAD_OFF 0 1300 #define REBOOT_HALT 0xcdef0123 1301 1302 int 1303 linux_reboot(struct thread *td, struct linux_reboot_args *args) 1304 { 1305 struct reboot_args bsd_args; 1306 1307 #ifdef DEBUG 1308 if (ldebug(reboot)) 1309 printf(ARGS(reboot, "0x%x"), args->cmd); 1310 #endif 1311 if (args->cmd == REBOOT_CAD_ON || args->cmd == REBOOT_CAD_OFF) 1312 return (0); 1313 bsd_args.opt = (args->cmd == REBOOT_HALT) ? RB_HALT : 0; 1314 return (reboot(td, &bsd_args)); 1315 } 1316 1317 1318 /* 1319 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1320 * td->td_retval[1] when COMPAT_43 is defined. This 1321 * globbers registers that are assumed to be preserved. The following 1322 * lightweight syscalls fixes this. See also linux_getgid16() and 1323 * linux_getuid16() in linux_uid16.c. 1324 * 1325 * linux_getpid() - MP SAFE 1326 * linux_getgid() - MP SAFE 1327 * linux_getuid() - MP SAFE 1328 */ 1329 1330 int 1331 linux_getpid(struct thread *td, struct linux_getpid_args *args) 1332 { 1333 1334 td->td_retval[0] = td->td_proc->p_pid; 1335 return (0); 1336 } 1337 1338 int 1339 linux_getgid(struct thread *td, struct linux_getgid_args *args) 1340 { 1341 1342 td->td_retval[0] = td->td_ucred->cr_rgid; 1343 return (0); 1344 } 1345 1346 int 1347 linux_getuid(struct thread *td, struct linux_getuid_args *args) 1348 { 1349 1350 td->td_retval[0] = td->td_ucred->cr_ruid; 1351 return (0); 1352 } 1353 1354 1355 int 1356 linux_getsid(struct thread *td, struct linux_getsid_args *args) 1357 { 1358 struct getsid_args bsd; 1359 bsd.pid = args->pid; 1360 return getsid(td, &bsd); 1361 } 1362 1363 int 1364 linux_nosys(struct thread *td, struct nosys_args *ignore) 1365 { 1366 1367 return (ENOSYS); 1368 } 1369 1370 int 1371 linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1372 { 1373 struct getpriority_args bsd_args; 1374 int error; 1375 1376 bsd_args.which = args->which; 1377 bsd_args.who = args->who; 1378 error = getpriority(td, &bsd_args); 1379 td->td_retval[0] = 20 - td->td_retval[0]; 1380 return error; 1381 } 1382 1383 int 1384 linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1385 { 1386 int name[2]; 1387 int error; 1388 1389 name[0] = CTL_KERN; 1390 name[1] = KERN_HOSTNAME; 1391 if ((error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL))) 1392 return (error); 1393 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1394 args->len, 0, 0)); 1395 } 1396 1397