linux_misc.c revision 293521
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: stable/10/sys/compat/linux/linux_misc.c 293521 2016-01-09 15:54:55Z dchagin $"); 32 33#include "opt_compat.h" 34#include "opt_kdtrace.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/malloc.h> 47#include <sys/mman.h> 48#include <sys/mount.h> 49#include <sys/mutex.h> 50#include <sys/namei.h> 51#include <sys/priv.h> 52#include <sys/proc.h> 53#include <sys/reboot.h> 54#include <sys/racct.h> 55#include <sys/resourcevar.h> 56#include <sys/sched.h> 57#include <sys/signalvar.h> 58#include <sys/stat.h> 59#include <sys/syscallsubr.h> 60#include <sys/sysctl.h> 61#include <sys/sysproto.h> 62#include <sys/systm.h> 63#include <sys/time.h> 64#include <sys/vmmeter.h> 65#include <sys/vnode.h> 66#include <sys/wait.h> 67#include <sys/cpuset.h> 68 69#include <security/mac/mac_framework.h> 70 71#include <vm/vm.h> 72#include <vm/pmap.h> 73#include <vm/vm_kern.h> 74#include <vm/vm_map.h> 75#include <vm/vm_extern.h> 76#include <vm/vm_object.h> 77#include <vm/swap_pager.h> 78 79#ifdef COMPAT_LINUX32 80#include <machine/../linux32/linux.h> 81#include <machine/../linux32/linux32_proto.h> 82#else 83#include <machine/../linux/linux.h> 84#include <machine/../linux/linux_proto.h> 85#endif 86 87#include <compat/linux/linux_file.h> 88#include <compat/linux/linux_mib.h> 89#include <compat/linux/linux_signal.h> 90#include <compat/linux/linux_util.h> 91#include <compat/linux/linux_sysproto.h> 92#include <compat/linux/linux_emul.h> 93#include <compat/linux/linux_misc.h> 94 95int stclohz; /* Statistics clock frequency */ 96 97static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 98 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 99 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 100 RLIMIT_MEMLOCK, RLIMIT_AS 101}; 102 103struct l_sysinfo { 104 l_long uptime; /* Seconds since boot */ 105 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 106#define LINUX_SYSINFO_LOADS_SCALE 65536 107 l_ulong totalram; /* Total usable main memory size */ 108 l_ulong freeram; /* Available memory size */ 109 l_ulong sharedram; /* Amount of shared memory */ 110 l_ulong bufferram; /* Memory used by buffers */ 111 l_ulong totalswap; /* Total swap space size */ 112 l_ulong freeswap; /* swap space still available */ 113 l_ushort procs; /* Number of current processes */ 114 l_ushort pads; 115 l_ulong totalbig; 116 l_ulong freebig; 117 l_uint mem_unit; 118 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 119}; 120 121struct l_pselect6arg { 122 l_uintptr_t ss; 123 l_size_t ss_len; 124}; 125 126int 127linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 128{ 129 struct l_sysinfo sysinfo; 130 vm_object_t object; 131 int i, j; 132 struct timespec ts; 133 134 getnanouptime(&ts); 135 if (ts.tv_nsec != 0) 136 ts.tv_sec++; 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 mtx_lock(&vm_object_list_mtx); 149 TAILQ_FOREACH(object, &vm_object_list, object_list) 150 if (object->shadow_count > 1) 151 sysinfo.sharedram += object->resident_page_count; 152 mtx_unlock(&vm_object_list_mtx); 153 154 sysinfo.sharedram *= PAGE_SIZE; 155 sysinfo.bufferram = 0; 156 157 swap_pager_status(&i, &j); 158 sysinfo.totalswap = i * PAGE_SIZE; 159 sysinfo.freeswap = (i - j) * PAGE_SIZE; 160 161 sysinfo.procs = nprocs; 162 163 /* The following are only present in newer Linux kernels. */ 164 sysinfo.totalbig = 0; 165 sysinfo.freebig = 0; 166 sysinfo.mem_unit = 1; 167 168 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 169} 170 171int 172linux_alarm(struct thread *td, struct linux_alarm_args *args) 173{ 174 struct itimerval it, old_it; 175 u_int secs; 176 int error; 177 178#ifdef DEBUG 179 if (ldebug(alarm)) 180 printf(ARGS(alarm, "%u"), args->secs); 181#endif 182 183 secs = args->secs; 184 185 if (secs > INT_MAX) 186 secs = INT_MAX; 187 188 it.it_value.tv_sec = (long) secs; 189 it.it_value.tv_usec = 0; 190 it.it_interval.tv_sec = 0; 191 it.it_interval.tv_usec = 0; 192 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 193 if (error) 194 return (error); 195 if (timevalisset(&old_it.it_value)) { 196 if (old_it.it_value.tv_usec != 0) 197 old_it.it_value.tv_sec++; 198 td->td_retval[0] = old_it.it_value.tv_sec; 199 } 200 return (0); 201} 202 203int 204linux_brk(struct thread *td, struct linux_brk_args *args) 205{ 206 struct vmspace *vm = td->td_proc->p_vmspace; 207 vm_offset_t new, old; 208 struct obreak_args /* { 209 char * nsize; 210 } */ tmp; 211 212#ifdef DEBUG 213 if (ldebug(brk)) 214 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 215#endif 216 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 217 new = (vm_offset_t)args->dsend; 218 tmp.nsize = (char *)new; 219 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp)) 220 td->td_retval[0] = (long)new; 221 else 222 td->td_retval[0] = (long)old; 223 224 return (0); 225} 226 227#if defined(__i386__) 228/* XXX: what about amd64/linux32? */ 229 230int 231linux_uselib(struct thread *td, struct linux_uselib_args *args) 232{ 233 struct nameidata ni; 234 struct vnode *vp; 235 struct exec *a_out; 236 struct vattr attr; 237 vm_offset_t vmaddr; 238 unsigned long file_offset; 239 unsigned long bss_size; 240 char *library; 241 ssize_t aresid; 242 int error, locked, writecount; 243 244 LCONVPATHEXIST(td, args->library, &library); 245 246#ifdef DEBUG 247 if (ldebug(uselib)) 248 printf(ARGS(uselib, "%s"), library); 249#endif 250 251 a_out = NULL; 252 locked = 0; 253 vp = NULL; 254 255 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 256 UIO_SYSSPACE, library, td); 257 error = namei(&ni); 258 LFREEPATH(library); 259 if (error) 260 goto cleanup; 261 262 vp = ni.ni_vp; 263 NDFREE(&ni, NDF_ONLY_PNBUF); 264 265 /* 266 * From here on down, we have a locked vnode that must be unlocked. 267 * XXX: The code below largely duplicates exec_check_permissions(). 268 */ 269 locked = 1; 270 271 /* Writable? */ 272 error = VOP_GET_WRITECOUNT(vp, &writecount); 273 if (error != 0) 274 goto cleanup; 275 if (writecount != 0) { 276 error = ETXTBSY; 277 goto cleanup; 278 } 279 280 /* Executable? */ 281 error = VOP_GETATTR(vp, &attr, td->td_ucred); 282 if (error) 283 goto cleanup; 284 285 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 286 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 287 /* EACCESS is what exec(2) returns. */ 288 error = ENOEXEC; 289 goto cleanup; 290 } 291 292 /* Sensible size? */ 293 if (attr.va_size == 0) { 294 error = ENOEXEC; 295 goto cleanup; 296 } 297 298 /* Can we access it? */ 299 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 300 if (error) 301 goto cleanup; 302 303 /* 304 * XXX: This should use vn_open() so that it is properly authorized, 305 * and to reduce code redundancy all over the place here. 306 * XXX: Not really, it duplicates far more of exec_check_permissions() 307 * than vn_open(). 308 */ 309#ifdef MAC 310 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 311 if (error) 312 goto cleanup; 313#endif 314 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 315 if (error) 316 goto cleanup; 317 318 /* Pull in executable header into exec_map */ 319 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 320 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 321 if (error) 322 goto cleanup; 323 324 /* Is it a Linux binary ? */ 325 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 326 error = ENOEXEC; 327 goto cleanup; 328 } 329 330 /* 331 * While we are here, we should REALLY do some more checks 332 */ 333 334 /* Set file/virtual offset based on a.out variant. */ 335 switch ((int)(a_out->a_magic & 0xffff)) { 336 case 0413: /* ZMAGIC */ 337 file_offset = 1024; 338 break; 339 case 0314: /* QMAGIC */ 340 file_offset = 0; 341 break; 342 default: 343 error = ENOEXEC; 344 goto cleanup; 345 } 346 347 bss_size = round_page(a_out->a_bss); 348 349 /* Check various fields in header for validity/bounds. */ 350 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 351 error = ENOEXEC; 352 goto cleanup; 353 } 354 355 /* text + data can't exceed file size */ 356 if (a_out->a_data + a_out->a_text > attr.va_size) { 357 error = EFAULT; 358 goto cleanup; 359 } 360 361 /* 362 * text/data/bss must not exceed limits 363 * XXX - this is not complete. it should check current usage PLUS 364 * the resources needed by this library. 365 */ 366 PROC_LOCK(td->td_proc); 367 if (a_out->a_text > maxtsiz || 368 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) || 369 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 370 bss_size) != 0) { 371 PROC_UNLOCK(td->td_proc); 372 error = ENOMEM; 373 goto cleanup; 374 } 375 PROC_UNLOCK(td->td_proc); 376 377 /* 378 * Prevent more writers. 379 * XXX: Note that if any of the VM operations fail below we don't 380 * clear this flag. 381 */ 382 VOP_SET_TEXT(vp); 383 384 /* 385 * Lock no longer needed 386 */ 387 locked = 0; 388 VOP_UNLOCK(vp, 0); 389 390 /* 391 * Check if file_offset page aligned. Currently we cannot handle 392 * misalinged file offsets, and so we read in the entire image 393 * (what a waste). 394 */ 395 if (file_offset & PAGE_MASK) { 396#ifdef DEBUG 397 printf("uselib: Non page aligned binary %lu\n", file_offset); 398#endif 399 /* Map text+data read/write/execute */ 400 401 /* a_entry is the load address and is page aligned */ 402 vmaddr = trunc_page(a_out->a_entry); 403 404 /* get anon user mapping, read+write+execute */ 405 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 406 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 407 VM_PROT_ALL, VM_PROT_ALL, 0); 408 if (error) 409 goto cleanup; 410 411 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 412 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 413 td->td_ucred, NOCRED, &aresid, td); 414 if (error != 0) 415 goto cleanup; 416 if (aresid != 0) { 417 error = ENOEXEC; 418 goto cleanup; 419 } 420 } else { 421#ifdef DEBUG 422 printf("uselib: Page aligned binary %lu\n", file_offset); 423#endif 424 /* 425 * for QMAGIC, a_entry is 20 bytes beyond the load address 426 * to skip the executable header 427 */ 428 vmaddr = trunc_page(a_out->a_entry); 429 430 /* 431 * Map it all into the process's space as a single 432 * copy-on-write "data" segment. 433 */ 434 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 435 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 436 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 437 if (error) 438 goto cleanup; 439 } 440#ifdef DEBUG 441 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 442 ((long *)vmaddr)[1]); 443#endif 444 if (bss_size != 0) { 445 /* Calculate BSS start address */ 446 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 447 a_out->a_data; 448 449 /* allocate some 'anon' space */ 450 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 451 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 452 VM_PROT_ALL, 0); 453 if (error) 454 goto cleanup; 455 } 456 457cleanup: 458 /* Unlock vnode if needed */ 459 if (locked) 460 VOP_UNLOCK(vp, 0); 461 462 /* Release the temporary mapping. */ 463 if (a_out) 464 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 465 466 return (error); 467} 468 469#endif /* __i386__ */ 470 471int 472linux_select(struct thread *td, struct linux_select_args *args) 473{ 474 l_timeval ltv; 475 struct timeval tv0, tv1, utv, *tvp; 476 int error; 477 478#ifdef DEBUG 479 if (ldebug(select)) 480 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 481 (void *)args->readfds, (void *)args->writefds, 482 (void *)args->exceptfds, (void *)args->timeout); 483#endif 484 485 /* 486 * Store current time for computation of the amount of 487 * time left. 488 */ 489 if (args->timeout) { 490 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 491 goto select_out; 492 utv.tv_sec = ltv.tv_sec; 493 utv.tv_usec = ltv.tv_usec; 494#ifdef DEBUG 495 if (ldebug(select)) 496 printf(LMSG("incoming timeout (%jd/%ld)"), 497 (intmax_t)utv.tv_sec, utv.tv_usec); 498#endif 499 500 if (itimerfix(&utv)) { 501 /* 502 * The timeval was invalid. Convert it to something 503 * valid that will act as it does under Linux. 504 */ 505 utv.tv_sec += utv.tv_usec / 1000000; 506 utv.tv_usec %= 1000000; 507 if (utv.tv_usec < 0) { 508 utv.tv_sec -= 1; 509 utv.tv_usec += 1000000; 510 } 511 if (utv.tv_sec < 0) 512 timevalclear(&utv); 513 } 514 microtime(&tv0); 515 tvp = &utv; 516 } else 517 tvp = NULL; 518 519 error = kern_select(td, args->nfds, args->readfds, args->writefds, 520 args->exceptfds, tvp, sizeof(l_int) * 8); 521 522#ifdef DEBUG 523 if (ldebug(select)) 524 printf(LMSG("real select returns %d"), error); 525#endif 526 if (error) 527 goto select_out; 528 529 if (args->timeout) { 530 if (td->td_retval[0]) { 531 /* 532 * Compute how much time was left of the timeout, 533 * by subtracting the current time and the time 534 * before we started the call, and subtracting 535 * that result from the user-supplied value. 536 */ 537 microtime(&tv1); 538 timevalsub(&tv1, &tv0); 539 timevalsub(&utv, &tv1); 540 if (utv.tv_sec < 0) 541 timevalclear(&utv); 542 } else 543 timevalclear(&utv); 544#ifdef DEBUG 545 if (ldebug(select)) 546 printf(LMSG("outgoing timeout (%jd/%ld)"), 547 (intmax_t)utv.tv_sec, utv.tv_usec); 548#endif 549 ltv.tv_sec = utv.tv_sec; 550 ltv.tv_usec = utv.tv_usec; 551 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 552 goto select_out; 553 } 554 555select_out: 556#ifdef DEBUG 557 if (ldebug(select)) 558 printf(LMSG("select_out -> %d"), error); 559#endif 560 return (error); 561} 562 563int 564linux_mremap(struct thread *td, struct linux_mremap_args *args) 565{ 566 struct munmap_args /* { 567 void *addr; 568 size_t len; 569 } */ bsd_args; 570 int error = 0; 571 572#ifdef DEBUG 573 if (ldebug(mremap)) 574 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 575 (void *)(uintptr_t)args->addr, 576 (unsigned long)args->old_len, 577 (unsigned long)args->new_len, 578 (unsigned long)args->flags); 579#endif 580 581 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 582 td->td_retval[0] = 0; 583 return (EINVAL); 584 } 585 586 /* 587 * Check for the page alignment. 588 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 589 */ 590 if (args->addr & PAGE_MASK) { 591 td->td_retval[0] = 0; 592 return (EINVAL); 593 } 594 595 args->new_len = round_page(args->new_len); 596 args->old_len = round_page(args->old_len); 597 598 if (args->new_len > args->old_len) { 599 td->td_retval[0] = 0; 600 return (ENOMEM); 601 } 602 603 if (args->new_len < args->old_len) { 604 bsd_args.addr = 605 (caddr_t)((uintptr_t)args->addr + args->new_len); 606 bsd_args.len = args->old_len - args->new_len; 607 error = sys_munmap(td, &bsd_args); 608 } 609 610 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 611 return (error); 612} 613 614#define LINUX_MS_ASYNC 0x0001 615#define LINUX_MS_INVALIDATE 0x0002 616#define LINUX_MS_SYNC 0x0004 617 618int 619linux_msync(struct thread *td, struct linux_msync_args *args) 620{ 621 struct msync_args bsd_args; 622 623 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 624 bsd_args.len = (uintptr_t)args->len; 625 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 626 627 return (sys_msync(td, &bsd_args)); 628} 629 630int 631linux_time(struct thread *td, struct linux_time_args *args) 632{ 633 struct timeval tv; 634 l_time_t tm; 635 int error; 636 637#ifdef DEBUG 638 if (ldebug(time)) 639 printf(ARGS(time, "*")); 640#endif 641 642 microtime(&tv); 643 tm = tv.tv_sec; 644 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 645 return (error); 646 td->td_retval[0] = tm; 647 return (0); 648} 649 650struct l_times_argv { 651 l_clock_t tms_utime; 652 l_clock_t tms_stime; 653 l_clock_t tms_cutime; 654 l_clock_t tms_cstime; 655}; 656 657 658/* 659 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 660 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 661 * auxiliary vector entry. 662 */ 663#define CLK_TCK 100 664 665#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 666#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 667 668#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 669 CONVNTCK(r) : CONVOTCK(r)) 670 671int 672linux_times(struct thread *td, struct linux_times_args *args) 673{ 674 struct timeval tv, utime, stime, cutime, cstime; 675 struct l_times_argv tms; 676 struct proc *p; 677 int error; 678 679#ifdef DEBUG 680 if (ldebug(times)) 681 printf(ARGS(times, "*")); 682#endif 683 684 if (args->buf != NULL) { 685 p = td->td_proc; 686 PROC_LOCK(p); 687 PROC_STATLOCK(p); 688 calcru(p, &utime, &stime); 689 PROC_STATUNLOCK(p); 690 calccru(p, &cutime, &cstime); 691 PROC_UNLOCK(p); 692 693 tms.tms_utime = CONVTCK(utime); 694 tms.tms_stime = CONVTCK(stime); 695 696 tms.tms_cutime = CONVTCK(cutime); 697 tms.tms_cstime = CONVTCK(cstime); 698 699 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 700 return (error); 701 } 702 703 microuptime(&tv); 704 td->td_retval[0] = (int)CONVTCK(tv); 705 return (0); 706} 707 708int 709linux_newuname(struct thread *td, struct linux_newuname_args *args) 710{ 711 struct l_new_utsname utsname; 712 char osname[LINUX_MAX_UTSNAME]; 713 char osrelease[LINUX_MAX_UTSNAME]; 714 char *p; 715 716#ifdef DEBUG 717 if (ldebug(newuname)) 718 printf(ARGS(newuname, "*")); 719#endif 720 721 linux_get_osname(td, osname); 722 linux_get_osrelease(td, osrelease); 723 724 bzero(&utsname, sizeof(utsname)); 725 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 726 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 727 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 728 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 729 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 730 for (p = utsname.version; *p != '\0'; ++p) 731 if (*p == '\n') { 732 *p = '\0'; 733 break; 734 } 735 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 736 737 return (copyout(&utsname, args->buf, sizeof(utsname))); 738} 739 740struct l_utimbuf { 741 l_time_t l_actime; 742 l_time_t l_modtime; 743}; 744 745int 746linux_utime(struct thread *td, struct linux_utime_args *args) 747{ 748 struct timeval tv[2], *tvp; 749 struct l_utimbuf lut; 750 char *fname; 751 int error; 752 753 LCONVPATHEXIST(td, args->fname, &fname); 754 755#ifdef DEBUG 756 if (ldebug(utime)) 757 printf(ARGS(utime, "%s, *"), fname); 758#endif 759 760 if (args->times) { 761 if ((error = copyin(args->times, &lut, sizeof lut))) { 762 LFREEPATH(fname); 763 return (error); 764 } 765 tv[0].tv_sec = lut.l_actime; 766 tv[0].tv_usec = 0; 767 tv[1].tv_sec = lut.l_modtime; 768 tv[1].tv_usec = 0; 769 tvp = tv; 770 } else 771 tvp = NULL; 772 773 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 774 LFREEPATH(fname); 775 return (error); 776} 777 778int 779linux_utimes(struct thread *td, struct linux_utimes_args *args) 780{ 781 l_timeval ltv[2]; 782 struct timeval tv[2], *tvp = NULL; 783 char *fname; 784 int error; 785 786 LCONVPATHEXIST(td, args->fname, &fname); 787 788#ifdef DEBUG 789 if (ldebug(utimes)) 790 printf(ARGS(utimes, "%s, *"), fname); 791#endif 792 793 if (args->tptr != NULL) { 794 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 795 LFREEPATH(fname); 796 return (error); 797 } 798 tv[0].tv_sec = ltv[0].tv_sec; 799 tv[0].tv_usec = ltv[0].tv_usec; 800 tv[1].tv_sec = ltv[1].tv_sec; 801 tv[1].tv_usec = ltv[1].tv_usec; 802 tvp = tv; 803 } 804 805 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 806 LFREEPATH(fname); 807 return (error); 808} 809 810int 811linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 812{ 813 l_timeval ltv[2]; 814 struct timeval tv[2], *tvp = NULL; 815 char *fname; 816 int error, dfd; 817 818 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 819 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 820 821#ifdef DEBUG 822 if (ldebug(futimesat)) 823 printf(ARGS(futimesat, "%s, *"), fname); 824#endif 825 826 if (args->utimes != NULL) { 827 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 828 LFREEPATH(fname); 829 return (error); 830 } 831 tv[0].tv_sec = ltv[0].tv_sec; 832 tv[0].tv_usec = ltv[0].tv_usec; 833 tv[1].tv_sec = ltv[1].tv_sec; 834 tv[1].tv_usec = ltv[1].tv_usec; 835 tvp = tv; 836 } 837 838 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 839 LFREEPATH(fname); 840 return (error); 841} 842 843int 844linux_common_wait(struct thread *td, int pid, int *status, 845 int options, struct rusage *ru) 846{ 847 int error, tmpstat; 848 849 error = kern_wait(td, pid, &tmpstat, options, ru); 850 if (error) 851 return (error); 852 853 if (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, status, sizeof(int)); 862 } 863 864 return (error); 865} 866 867#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 868int 869linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 870{ 871 int options; 872 873#ifdef DEBUG 874 if (ldebug(waitpid)) 875 printf(ARGS(waitpid, "%d, %p, %d"), 876 args->pid, (void *)args->status, args->options); 877#endif 878 /* 879 * this is necessary because the test in kern_wait doesn't work 880 * because we mess with the options here 881 */ 882 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE)) 883 return (EINVAL); 884 885 options = (args->options & (WNOHANG | WUNTRACED)); 886 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 887 if (args->options & __WCLONE) 888 options |= WLINUXCLONE; 889 890 return (linux_common_wait(td, args->pid, args->status, options, NULL)); 891} 892#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 893 894int 895linux_wait4(struct thread *td, struct linux_wait4_args *args) 896{ 897 int error, options; 898 struct rusage ru, *rup; 899 900#ifdef DEBUG 901 if (ldebug(wait4)) 902 printf(ARGS(wait4, "%d, %p, %d, %p"), 903 args->pid, (void *)args->status, args->options, 904 (void *)args->rusage); 905#endif 906 907 options = (args->options & (WNOHANG | WUNTRACED)); 908 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 909 if (args->options & __WCLONE) 910 options |= WLINUXCLONE; 911 912 if (args->rusage != NULL) 913 rup = &ru; 914 else 915 rup = NULL; 916 error = linux_common_wait(td, args->pid, args->status, options, rup); 917 if (error != 0) 918 return (error); 919 if (args->rusage != NULL) 920 error = linux_copyout_rusage(&ru, args->rusage); 921 return (error); 922} 923 924int 925linux_waitid(struct thread *td, struct linux_waitid_args *args) 926{ 927 int status, options, sig; 928 struct __wrusage wru; 929 siginfo_t siginfo; 930 l_siginfo_t lsi; 931 idtype_t idtype; 932 struct proc *p; 933 int error; 934 935 options = 0; 936 linux_to_bsd_waitopts(args->options, &options); 937 938 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 939 return (EINVAL); 940 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 941 return (EINVAL); 942 943 switch (args->idtype) { 944 case LINUX_P_ALL: 945 idtype = P_ALL; 946 break; 947 case LINUX_P_PID: 948 if (args->id <= 0) 949 return (EINVAL); 950 idtype = P_PID; 951 break; 952 case LINUX_P_PGID: 953 if (args->id <= 0) 954 return (EINVAL); 955 idtype = P_PGID; 956 break; 957 default: 958 return (EINVAL); 959 } 960 961 error = kern_wait6(td, idtype, args->id, &status, options, 962 &wru, &siginfo); 963 if (error != 0) 964 return (error); 965 if (args->rusage != NULL) { 966 error = linux_copyout_rusage(&wru.wru_children, 967 args->rusage); 968 if (error != 0) 969 return (error); 970 } 971 if (args->info != NULL) { 972 p = td->td_proc; 973 if (td->td_retval[0] == 0) 974 bzero(&lsi, sizeof(lsi)); 975 else { 976 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo); 977 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 978 } 979 error = copyout(&lsi, args->info, sizeof(lsi)); 980 } 981 td->td_retval[0] = 0; 982 983 return (error); 984} 985 986int 987linux_mknod(struct thread *td, struct linux_mknod_args *args) 988{ 989 char *path; 990 int error; 991 992 LCONVPATHCREAT(td, args->path, &path); 993 994#ifdef DEBUG 995 if (ldebug(mknod)) 996 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev); 997#endif 998 999 switch (args->mode & S_IFMT) { 1000 case S_IFIFO: 1001 case S_IFSOCK: 1002 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1003 break; 1004 1005 case S_IFCHR: 1006 case S_IFBLK: 1007 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1008 args->dev); 1009 break; 1010 1011 case S_IFDIR: 1012 error = EPERM; 1013 break; 1014 1015 case 0: 1016 args->mode |= S_IFREG; 1017 /* FALLTHROUGH */ 1018 case S_IFREG: 1019 error = kern_open(td, path, UIO_SYSSPACE, 1020 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1021 if (error == 0) 1022 kern_close(td, td->td_retval[0]); 1023 break; 1024 1025 default: 1026 error = EINVAL; 1027 break; 1028 } 1029 LFREEPATH(path); 1030 return (error); 1031} 1032 1033int 1034linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1035{ 1036 char *path; 1037 int error, dfd; 1038 1039 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1040 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1041 1042#ifdef DEBUG 1043 if (ldebug(mknodat)) 1044 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1045#endif 1046 1047 switch (args->mode & S_IFMT) { 1048 case S_IFIFO: 1049 case S_IFSOCK: 1050 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1051 break; 1052 1053 case S_IFCHR: 1054 case S_IFBLK: 1055 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1056 args->dev); 1057 break; 1058 1059 case S_IFDIR: 1060 error = EPERM; 1061 break; 1062 1063 case 0: 1064 args->mode |= S_IFREG; 1065 /* FALLTHROUGH */ 1066 case S_IFREG: 1067 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1068 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1069 if (error == 0) 1070 kern_close(td, td->td_retval[0]); 1071 break; 1072 1073 default: 1074 error = EINVAL; 1075 break; 1076 } 1077 LFREEPATH(path); 1078 return (error); 1079} 1080 1081/* 1082 * UGH! This is just about the dumbest idea I've ever heard!! 1083 */ 1084int 1085linux_personality(struct thread *td, struct linux_personality_args *args) 1086{ 1087#ifdef DEBUG 1088 if (ldebug(personality)) 1089 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1090#endif 1091 if (args->per != 0) 1092 return (EINVAL); 1093 1094 /* Yes Jim, it's still a Linux... */ 1095 td->td_retval[0] = 0; 1096 return (0); 1097} 1098 1099struct l_itimerval { 1100 l_timeval it_interval; 1101 l_timeval it_value; 1102}; 1103 1104#define B2L_ITIMERVAL(bip, lip) \ 1105 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1106 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1107 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1108 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1109 1110int 1111linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1112{ 1113 int error; 1114 struct l_itimerval ls; 1115 struct itimerval aitv, oitv; 1116 1117#ifdef DEBUG 1118 if (ldebug(setitimer)) 1119 printf(ARGS(setitimer, "%p, %p"), 1120 (void *)uap->itv, (void *)uap->oitv); 1121#endif 1122 1123 if (uap->itv == NULL) { 1124 uap->itv = uap->oitv; 1125 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1126 } 1127 1128 error = copyin(uap->itv, &ls, sizeof(ls)); 1129 if (error != 0) 1130 return (error); 1131 B2L_ITIMERVAL(&aitv, &ls); 1132#ifdef DEBUG 1133 if (ldebug(setitimer)) { 1134 printf("setitimer: value: sec: %jd, usec: %ld\n", 1135 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1136 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1137 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1138 } 1139#endif 1140 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1141 if (error != 0 || uap->oitv == NULL) 1142 return (error); 1143 B2L_ITIMERVAL(&ls, &oitv); 1144 1145 return (copyout(&ls, uap->oitv, sizeof(ls))); 1146} 1147 1148int 1149linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1150{ 1151 int error; 1152 struct l_itimerval ls; 1153 struct itimerval aitv; 1154 1155#ifdef DEBUG 1156 if (ldebug(getitimer)) 1157 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1158#endif 1159 error = kern_getitimer(td, uap->which, &aitv); 1160 if (error != 0) 1161 return (error); 1162 B2L_ITIMERVAL(&ls, &aitv); 1163 return (copyout(&ls, uap->itv, sizeof(ls))); 1164} 1165 1166#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1167int 1168linux_nice(struct thread *td, struct linux_nice_args *args) 1169{ 1170 struct setpriority_args bsd_args; 1171 1172 bsd_args.which = PRIO_PROCESS; 1173 bsd_args.who = 0; /* current process */ 1174 bsd_args.prio = args->inc; 1175 return (sys_setpriority(td, &bsd_args)); 1176} 1177#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1178 1179int 1180linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1181{ 1182 struct ucred *newcred, *oldcred; 1183 l_gid_t *linux_gidset; 1184 gid_t *bsd_gidset; 1185 int ngrp, error; 1186 struct proc *p; 1187 1188 ngrp = args->gidsetsize; 1189 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1190 return (EINVAL); 1191 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK); 1192 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1193 if (error) 1194 goto out; 1195 newcred = crget(); 1196 p = td->td_proc; 1197 PROC_LOCK(p); 1198 oldcred = crcopysafe(p, newcred); 1199 1200 /* 1201 * cr_groups[0] holds egid. Setting the whole set from 1202 * the supplied set will cause egid to be changed too. 1203 * Keep cr_groups[0] unchanged to prevent that. 1204 */ 1205 1206 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1207 PROC_UNLOCK(p); 1208 crfree(newcred); 1209 goto out; 1210 } 1211 1212 if (ngrp > 0) { 1213 newcred->cr_ngroups = ngrp + 1; 1214 1215 bsd_gidset = newcred->cr_groups; 1216 ngrp--; 1217 while (ngrp >= 0) { 1218 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1219 ngrp--; 1220 } 1221 } else 1222 newcred->cr_ngroups = 1; 1223 1224 setsugid(p); 1225 p->p_ucred = newcred; 1226 PROC_UNLOCK(p); 1227 crfree(oldcred); 1228 error = 0; 1229out: 1230 free(linux_gidset, M_TEMP); 1231 return (error); 1232} 1233 1234int 1235linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1236{ 1237 struct ucred *cred; 1238 l_gid_t *linux_gidset; 1239 gid_t *bsd_gidset; 1240 int bsd_gidsetsz, ngrp, error; 1241 1242 cred = td->td_ucred; 1243 bsd_gidset = cred->cr_groups; 1244 bsd_gidsetsz = cred->cr_ngroups - 1; 1245 1246 /* 1247 * cr_groups[0] holds egid. Returning the whole set 1248 * here will cause a duplicate. Exclude cr_groups[0] 1249 * to prevent that. 1250 */ 1251 1252 if ((ngrp = args->gidsetsize) == 0) { 1253 td->td_retval[0] = bsd_gidsetsz; 1254 return (0); 1255 } 1256 1257 if (ngrp < bsd_gidsetsz) 1258 return (EINVAL); 1259 1260 ngrp = 0; 1261 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1262 M_TEMP, M_WAITOK); 1263 while (ngrp < bsd_gidsetsz) { 1264 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1265 ngrp++; 1266 } 1267 1268 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1269 free(linux_gidset, M_TEMP); 1270 if (error) 1271 return (error); 1272 1273 td->td_retval[0] = ngrp; 1274 return (0); 1275} 1276 1277int 1278linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1279{ 1280 struct rlimit bsd_rlim; 1281 struct l_rlimit rlim; 1282 u_int which; 1283 int error; 1284 1285#ifdef DEBUG 1286 if (ldebug(setrlimit)) 1287 printf(ARGS(setrlimit, "%d, %p"), 1288 args->resource, (void *)args->rlim); 1289#endif 1290 1291 if (args->resource >= LINUX_RLIM_NLIMITS) 1292 return (EINVAL); 1293 1294 which = linux_to_bsd_resource[args->resource]; 1295 if (which == -1) 1296 return (EINVAL); 1297 1298 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1299 if (error) 1300 return (error); 1301 1302 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1303 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1304 return (kern_setrlimit(td, which, &bsd_rlim)); 1305} 1306 1307#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1308int 1309linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1310{ 1311 struct l_rlimit rlim; 1312 struct proc *p = td->td_proc; 1313 struct rlimit bsd_rlim; 1314 u_int which; 1315 1316#ifdef DEBUG 1317 if (ldebug(old_getrlimit)) 1318 printf(ARGS(old_getrlimit, "%d, %p"), 1319 args->resource, (void *)args->rlim); 1320#endif 1321 1322 if (args->resource >= LINUX_RLIM_NLIMITS) 1323 return (EINVAL); 1324 1325 which = linux_to_bsd_resource[args->resource]; 1326 if (which == -1) 1327 return (EINVAL); 1328 1329 PROC_LOCK(p); 1330 lim_rlimit(p, which, &bsd_rlim); 1331 PROC_UNLOCK(p); 1332 1333#ifdef COMPAT_LINUX32 1334 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1335 if (rlim.rlim_cur == UINT_MAX) 1336 rlim.rlim_cur = INT_MAX; 1337 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1338 if (rlim.rlim_max == UINT_MAX) 1339 rlim.rlim_max = INT_MAX; 1340#else 1341 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1342 if (rlim.rlim_cur == ULONG_MAX) 1343 rlim.rlim_cur = LONG_MAX; 1344 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1345 if (rlim.rlim_max == ULONG_MAX) 1346 rlim.rlim_max = LONG_MAX; 1347#endif 1348 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1349} 1350#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1351 1352int 1353linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1354{ 1355 struct l_rlimit rlim; 1356 struct proc *p = td->td_proc; 1357 struct rlimit bsd_rlim; 1358 u_int which; 1359 1360#ifdef DEBUG 1361 if (ldebug(getrlimit)) 1362 printf(ARGS(getrlimit, "%d, %p"), 1363 args->resource, (void *)args->rlim); 1364#endif 1365 1366 if (args->resource >= LINUX_RLIM_NLIMITS) 1367 return (EINVAL); 1368 1369 which = linux_to_bsd_resource[args->resource]; 1370 if (which == -1) 1371 return (EINVAL); 1372 1373 PROC_LOCK(p); 1374 lim_rlimit(p, which, &bsd_rlim); 1375 PROC_UNLOCK(p); 1376 1377 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1378 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1379 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1380} 1381 1382int 1383linux_sched_setscheduler(struct thread *td, 1384 struct linux_sched_setscheduler_args *args) 1385{ 1386 struct sched_param sched_param; 1387 struct thread *tdt; 1388 int error, policy; 1389 1390#ifdef DEBUG 1391 if (ldebug(sched_setscheduler)) 1392 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1393 args->pid, args->policy, (const void *)args->param); 1394#endif 1395 1396 switch (args->policy) { 1397 case LINUX_SCHED_OTHER: 1398 policy = SCHED_OTHER; 1399 break; 1400 case LINUX_SCHED_FIFO: 1401 policy = SCHED_FIFO; 1402 break; 1403 case LINUX_SCHED_RR: 1404 policy = SCHED_RR; 1405 break; 1406 default: 1407 return (EINVAL); 1408 } 1409 1410 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1411 if (error) 1412 return (error); 1413 1414 tdt = linux_tdfind(td, args->pid, -1); 1415 if (tdt == NULL) 1416 return (ESRCH); 1417 1418 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1419 PROC_UNLOCK(tdt->td_proc); 1420 return (error); 1421} 1422 1423int 1424linux_sched_getscheduler(struct thread *td, 1425 struct linux_sched_getscheduler_args *args) 1426{ 1427 struct thread *tdt; 1428 int error, policy; 1429 1430#ifdef DEBUG 1431 if (ldebug(sched_getscheduler)) 1432 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1433#endif 1434 1435 tdt = linux_tdfind(td, args->pid, -1); 1436 if (tdt == NULL) 1437 return (ESRCH); 1438 1439 error = kern_sched_getscheduler(td, tdt, &policy); 1440 PROC_UNLOCK(tdt->td_proc); 1441 1442 switch (policy) { 1443 case SCHED_OTHER: 1444 td->td_retval[0] = LINUX_SCHED_OTHER; 1445 break; 1446 case SCHED_FIFO: 1447 td->td_retval[0] = LINUX_SCHED_FIFO; 1448 break; 1449 case SCHED_RR: 1450 td->td_retval[0] = LINUX_SCHED_RR; 1451 break; 1452 } 1453 return (error); 1454} 1455 1456int 1457linux_sched_get_priority_max(struct thread *td, 1458 struct linux_sched_get_priority_max_args *args) 1459{ 1460 struct sched_get_priority_max_args bsd; 1461 1462#ifdef DEBUG 1463 if (ldebug(sched_get_priority_max)) 1464 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1465#endif 1466 1467 switch (args->policy) { 1468 case LINUX_SCHED_OTHER: 1469 bsd.policy = SCHED_OTHER; 1470 break; 1471 case LINUX_SCHED_FIFO: 1472 bsd.policy = SCHED_FIFO; 1473 break; 1474 case LINUX_SCHED_RR: 1475 bsd.policy = SCHED_RR; 1476 break; 1477 default: 1478 return (EINVAL); 1479 } 1480 return (sys_sched_get_priority_max(td, &bsd)); 1481} 1482 1483int 1484linux_sched_get_priority_min(struct thread *td, 1485 struct linux_sched_get_priority_min_args *args) 1486{ 1487 struct sched_get_priority_min_args bsd; 1488 1489#ifdef DEBUG 1490 if (ldebug(sched_get_priority_min)) 1491 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1492#endif 1493 1494 switch (args->policy) { 1495 case LINUX_SCHED_OTHER: 1496 bsd.policy = SCHED_OTHER; 1497 break; 1498 case LINUX_SCHED_FIFO: 1499 bsd.policy = SCHED_FIFO; 1500 break; 1501 case LINUX_SCHED_RR: 1502 bsd.policy = SCHED_RR; 1503 break; 1504 default: 1505 return (EINVAL); 1506 } 1507 return (sys_sched_get_priority_min(td, &bsd)); 1508} 1509 1510#define REBOOT_CAD_ON 0x89abcdef 1511#define REBOOT_CAD_OFF 0 1512#define REBOOT_HALT 0xcdef0123 1513#define REBOOT_RESTART 0x01234567 1514#define REBOOT_RESTART2 0xA1B2C3D4 1515#define REBOOT_POWEROFF 0x4321FEDC 1516#define REBOOT_MAGIC1 0xfee1dead 1517#define REBOOT_MAGIC2 0x28121969 1518#define REBOOT_MAGIC2A 0x05121996 1519#define REBOOT_MAGIC2B 0x16041998 1520 1521int 1522linux_reboot(struct thread *td, struct linux_reboot_args *args) 1523{ 1524 struct reboot_args bsd_args; 1525 1526#ifdef DEBUG 1527 if (ldebug(reboot)) 1528 printf(ARGS(reboot, "0x%x"), args->cmd); 1529#endif 1530 1531 if (args->magic1 != REBOOT_MAGIC1) 1532 return (EINVAL); 1533 1534 switch (args->magic2) { 1535 case REBOOT_MAGIC2: 1536 case REBOOT_MAGIC2A: 1537 case REBOOT_MAGIC2B: 1538 break; 1539 default: 1540 return (EINVAL); 1541 } 1542 1543 switch (args->cmd) { 1544 case REBOOT_CAD_ON: 1545 case REBOOT_CAD_OFF: 1546 return (priv_check(td, PRIV_REBOOT)); 1547 case REBOOT_HALT: 1548 bsd_args.opt = RB_HALT; 1549 break; 1550 case REBOOT_RESTART: 1551 case REBOOT_RESTART2: 1552 bsd_args.opt = 0; 1553 break; 1554 case REBOOT_POWEROFF: 1555 bsd_args.opt = RB_POWEROFF; 1556 break; 1557 default: 1558 return (EINVAL); 1559 } 1560 return (sys_reboot(td, &bsd_args)); 1561} 1562 1563 1564/* 1565 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1566 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1567 * are assumed to be preserved. The following lightweight syscalls fixes 1568 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1569 * 1570 * linux_getpid() - MP SAFE 1571 * linux_getgid() - MP SAFE 1572 * linux_getuid() - MP SAFE 1573 */ 1574 1575int 1576linux_getpid(struct thread *td, struct linux_getpid_args *args) 1577{ 1578 1579#ifdef DEBUG 1580 if (ldebug(getpid)) 1581 printf(ARGS(getpid, "")); 1582#endif 1583 td->td_retval[0] = td->td_proc->p_pid; 1584 1585 return (0); 1586} 1587 1588int 1589linux_gettid(struct thread *td, struct linux_gettid_args *args) 1590{ 1591 struct linux_emuldata *em; 1592 1593#ifdef DEBUG 1594 if (ldebug(gettid)) 1595 printf(ARGS(gettid, "")); 1596#endif 1597 1598 em = em_find(td); 1599 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1600 1601 td->td_retval[0] = em->em_tid; 1602 1603 return (0); 1604} 1605 1606 1607int 1608linux_getppid(struct thread *td, struct linux_getppid_args *args) 1609{ 1610 1611#ifdef DEBUG 1612 if (ldebug(getppid)) 1613 printf(ARGS(getppid, "")); 1614#endif 1615 1616 PROC_LOCK(td->td_proc); 1617 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1618 PROC_UNLOCK(td->td_proc); 1619 return (0); 1620} 1621 1622int 1623linux_getgid(struct thread *td, struct linux_getgid_args *args) 1624{ 1625 1626#ifdef DEBUG 1627 if (ldebug(getgid)) 1628 printf(ARGS(getgid, "")); 1629#endif 1630 1631 td->td_retval[0] = td->td_ucred->cr_rgid; 1632 return (0); 1633} 1634 1635int 1636linux_getuid(struct thread *td, struct linux_getuid_args *args) 1637{ 1638 1639#ifdef DEBUG 1640 if (ldebug(getuid)) 1641 printf(ARGS(getuid, "")); 1642#endif 1643 1644 td->td_retval[0] = td->td_ucred->cr_ruid; 1645 return (0); 1646} 1647 1648 1649int 1650linux_getsid(struct thread *td, struct linux_getsid_args *args) 1651{ 1652 struct getsid_args bsd; 1653 1654#ifdef DEBUG 1655 if (ldebug(getsid)) 1656 printf(ARGS(getsid, "%i"), args->pid); 1657#endif 1658 1659 bsd.pid = args->pid; 1660 return (sys_getsid(td, &bsd)); 1661} 1662 1663int 1664linux_nosys(struct thread *td, struct nosys_args *ignore) 1665{ 1666 1667 return (ENOSYS); 1668} 1669 1670int 1671linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1672{ 1673 struct getpriority_args bsd_args; 1674 int error; 1675 1676#ifdef DEBUG 1677 if (ldebug(getpriority)) 1678 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1679#endif 1680 1681 bsd_args.which = args->which; 1682 bsd_args.who = args->who; 1683 error = sys_getpriority(td, &bsd_args); 1684 td->td_retval[0] = 20 - td->td_retval[0]; 1685 return (error); 1686} 1687 1688int 1689linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1690{ 1691 int name[2]; 1692 1693#ifdef DEBUG 1694 if (ldebug(sethostname)) 1695 printf(ARGS(sethostname, "*, %i"), args->len); 1696#endif 1697 1698 name[0] = CTL_KERN; 1699 name[1] = KERN_HOSTNAME; 1700 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1701 args->len, 0, 0)); 1702} 1703 1704int 1705linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1706{ 1707 int name[2]; 1708 1709#ifdef DEBUG 1710 if (ldebug(setdomainname)) 1711 printf(ARGS(setdomainname, "*, %i"), args->len); 1712#endif 1713 1714 name[0] = CTL_KERN; 1715 name[1] = KERN_NISDOMAINNAME; 1716 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1717 args->len, 0, 0)); 1718} 1719 1720int 1721linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1722{ 1723 1724#ifdef DEBUG 1725 if (ldebug(exit_group)) 1726 printf(ARGS(exit_group, "%i"), args->error_code); 1727#endif 1728 1729 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1730 args->error_code); 1731 1732 /* 1733 * XXX: we should send a signal to the parent if 1734 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1735 * as it doesnt occur often. 1736 */ 1737 exit1(td, W_EXITCODE(args->error_code, 0)); 1738 /* NOTREACHED */ 1739} 1740 1741#define _LINUX_CAPABILITY_VERSION 0x19980330 1742 1743struct l_user_cap_header { 1744 l_int version; 1745 l_int pid; 1746}; 1747 1748struct l_user_cap_data { 1749 l_int effective; 1750 l_int permitted; 1751 l_int inheritable; 1752}; 1753 1754int 1755linux_capget(struct thread *td, struct linux_capget_args *args) 1756{ 1757 struct l_user_cap_header luch; 1758 struct l_user_cap_data lucd; 1759 int error; 1760 1761 if (args->hdrp == NULL) 1762 return (EFAULT); 1763 1764 error = copyin(args->hdrp, &luch, sizeof(luch)); 1765 if (error != 0) 1766 return (error); 1767 1768 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1769 luch.version = _LINUX_CAPABILITY_VERSION; 1770 error = copyout(&luch, args->hdrp, sizeof(luch)); 1771 if (error) 1772 return (error); 1773 return (EINVAL); 1774 } 1775 1776 if (luch.pid) 1777 return (EPERM); 1778 1779 if (args->datap) { 1780 /* 1781 * The current implementation doesn't support setting 1782 * a capability (it's essentially a stub) so indicate 1783 * that no capabilities are currently set or available 1784 * to request. 1785 */ 1786 bzero (&lucd, sizeof(lucd)); 1787 error = copyout(&lucd, args->datap, sizeof(lucd)); 1788 } 1789 1790 return (error); 1791} 1792 1793int 1794linux_capset(struct thread *td, struct linux_capset_args *args) 1795{ 1796 struct l_user_cap_header luch; 1797 struct l_user_cap_data lucd; 1798 int error; 1799 1800 if (args->hdrp == NULL || args->datap == NULL) 1801 return (EFAULT); 1802 1803 error = copyin(args->hdrp, &luch, sizeof(luch)); 1804 if (error != 0) 1805 return (error); 1806 1807 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1808 luch.version = _LINUX_CAPABILITY_VERSION; 1809 error = copyout(&luch, args->hdrp, sizeof(luch)); 1810 if (error) 1811 return (error); 1812 return (EINVAL); 1813 } 1814 1815 if (luch.pid) 1816 return (EPERM); 1817 1818 error = copyin(args->datap, &lucd, sizeof(lucd)); 1819 if (error != 0) 1820 return (error); 1821 1822 /* We currently don't support setting any capabilities. */ 1823 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1824 linux_msg(td, 1825 "capset effective=0x%x, permitted=0x%x, " 1826 "inheritable=0x%x is not implemented", 1827 (int)lucd.effective, (int)lucd.permitted, 1828 (int)lucd.inheritable); 1829 return (EPERM); 1830 } 1831 1832 return (0); 1833} 1834 1835int 1836linux_prctl(struct thread *td, struct linux_prctl_args *args) 1837{ 1838 int error = 0, max_size; 1839 struct proc *p = td->td_proc; 1840 char comm[LINUX_MAX_COMM_LEN]; 1841 struct linux_emuldata *em; 1842 int pdeath_signal; 1843 1844#ifdef DEBUG 1845 if (ldebug(prctl)) 1846 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option, 1847 args->arg2, args->arg3, args->arg4, args->arg5); 1848#endif 1849 1850 switch (args->option) { 1851 case LINUX_PR_SET_PDEATHSIG: 1852 if (!LINUX_SIG_VALID(args->arg2)) 1853 return (EINVAL); 1854 em = em_find(td); 1855 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1856 em->pdeath_signal = args->arg2; 1857 break; 1858 case LINUX_PR_GET_PDEATHSIG: 1859 em = em_find(td); 1860 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1861 pdeath_signal = em->pdeath_signal; 1862 error = copyout(&pdeath_signal, 1863 (void *)(register_t)args->arg2, 1864 sizeof(pdeath_signal)); 1865 break; 1866 case LINUX_PR_GET_KEEPCAPS: 1867 /* 1868 * Indicate that we always clear the effective and 1869 * permitted capability sets when the user id becomes 1870 * non-zero (actually the capability sets are simply 1871 * always zero in the current implementation). 1872 */ 1873 td->td_retval[0] = 0; 1874 break; 1875 case LINUX_PR_SET_KEEPCAPS: 1876 /* 1877 * Ignore requests to keep the effective and permitted 1878 * capability sets when the user id becomes non-zero. 1879 */ 1880 break; 1881 case LINUX_PR_SET_NAME: 1882 /* 1883 * To be on the safe side we need to make sure to not 1884 * overflow the size a linux program expects. We already 1885 * do this here in the copyin, so that we don't need to 1886 * check on copyout. 1887 */ 1888 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1889 error = copyinstr((void *)(register_t)args->arg2, comm, 1890 max_size, NULL); 1891 1892 /* Linux silently truncates the name if it is too long. */ 1893 if (error == ENAMETOOLONG) { 1894 /* 1895 * XXX: copyinstr() isn't documented to populate the 1896 * array completely, so do a copyin() to be on the 1897 * safe side. This should be changed in case 1898 * copyinstr() is changed to guarantee this. 1899 */ 1900 error = copyin((void *)(register_t)args->arg2, comm, 1901 max_size - 1); 1902 comm[max_size - 1] = '\0'; 1903 } 1904 if (error) 1905 return (error); 1906 1907 PROC_LOCK(p); 1908 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1909 PROC_UNLOCK(p); 1910 break; 1911 case LINUX_PR_GET_NAME: 1912 PROC_LOCK(p); 1913 strlcpy(comm, p->p_comm, sizeof(comm)); 1914 PROC_UNLOCK(p); 1915 error = copyout(comm, (void *)(register_t)args->arg2, 1916 strlen(comm) + 1); 1917 break; 1918 default: 1919 error = EINVAL; 1920 break; 1921 } 1922 1923 return (error); 1924} 1925 1926int 1927linux_sched_setparam(struct thread *td, 1928 struct linux_sched_setparam_args *uap) 1929{ 1930 struct sched_param sched_param; 1931 struct thread *tdt; 1932 int error; 1933 1934#ifdef DEBUG 1935 if (ldebug(sched_setparam)) 1936 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 1937#endif 1938 1939 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1940 if (error) 1941 return (error); 1942 1943 tdt = linux_tdfind(td, uap->pid, -1); 1944 if (tdt == NULL) 1945 return (ESRCH); 1946 1947 error = kern_sched_setparam(td, tdt, &sched_param); 1948 PROC_UNLOCK(tdt->td_proc); 1949 return (error); 1950} 1951 1952int 1953linux_sched_getparam(struct thread *td, 1954 struct linux_sched_getparam_args *uap) 1955{ 1956 struct sched_param sched_param; 1957 struct thread *tdt; 1958 int error; 1959 1960#ifdef DEBUG 1961 if (ldebug(sched_getparam)) 1962 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 1963#endif 1964 1965 tdt = linux_tdfind(td, uap->pid, -1); 1966 if (tdt == NULL) 1967 return (ESRCH); 1968 1969 error = kern_sched_getparam(td, tdt, &sched_param); 1970 PROC_UNLOCK(tdt->td_proc); 1971 if (error == 0) 1972 error = copyout(&sched_param, uap->param, 1973 sizeof(sched_param)); 1974 return (error); 1975} 1976 1977/* 1978 * Get affinity of a process. 1979 */ 1980int 1981linux_sched_getaffinity(struct thread *td, 1982 struct linux_sched_getaffinity_args *args) 1983{ 1984 int error; 1985 struct thread *tdt; 1986 struct cpuset_getaffinity_args cga; 1987 1988#ifdef DEBUG 1989 if (ldebug(sched_getaffinity)) 1990 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 1991 args->len); 1992#endif 1993 if (args->len < sizeof(cpuset_t)) 1994 return (EINVAL); 1995 1996 tdt = linux_tdfind(td, args->pid, -1); 1997 if (tdt == NULL) 1998 return (ESRCH); 1999 2000 PROC_UNLOCK(tdt->td_proc); 2001 cga.level = CPU_LEVEL_WHICH; 2002 cga.which = CPU_WHICH_TID; 2003 cga.id = tdt->td_tid; 2004 cga.cpusetsize = sizeof(cpuset_t); 2005 cga.mask = (cpuset_t *) args->user_mask_ptr; 2006 2007 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2008 td->td_retval[0] = sizeof(cpuset_t); 2009 2010 return (error); 2011} 2012 2013/* 2014 * Set affinity of a process. 2015 */ 2016int 2017linux_sched_setaffinity(struct thread *td, 2018 struct linux_sched_setaffinity_args *args) 2019{ 2020 struct cpuset_setaffinity_args csa; 2021 struct thread *tdt; 2022 2023#ifdef DEBUG 2024 if (ldebug(sched_setaffinity)) 2025 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2026 args->len); 2027#endif 2028 if (args->len < sizeof(cpuset_t)) 2029 return (EINVAL); 2030 2031 tdt = linux_tdfind(td, args->pid, -1); 2032 if (tdt == NULL) 2033 return (ESRCH); 2034 2035 PROC_UNLOCK(tdt->td_proc); 2036 csa.level = CPU_LEVEL_WHICH; 2037 csa.which = CPU_WHICH_TID; 2038 csa.id = tdt->td_tid; 2039 csa.cpusetsize = sizeof(cpuset_t); 2040 csa.mask = (cpuset_t *) args->user_mask_ptr; 2041 2042 return (sys_cpuset_setaffinity(td, &csa)); 2043} 2044 2045struct linux_rlimit64 { 2046 uint64_t rlim_cur; 2047 uint64_t rlim_max; 2048}; 2049 2050int 2051linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2052{ 2053 struct rlimit rlim, nrlim; 2054 struct linux_rlimit64 lrlim; 2055 struct proc *p; 2056 u_int which; 2057 int flags; 2058 int error; 2059 2060#ifdef DEBUG 2061 if (ldebug(prlimit64)) 2062 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2063 args->resource, (void *)args->new, (void *)args->old); 2064#endif 2065 2066 if (args->resource >= LINUX_RLIM_NLIMITS) 2067 return (EINVAL); 2068 2069 which = linux_to_bsd_resource[args->resource]; 2070 if (which == -1) 2071 return (EINVAL); 2072 2073 if (args->new != NULL) { 2074 /* 2075 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2076 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2077 * as INFINITY so we do not need a conversion even. 2078 */ 2079 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2080 if (error != 0) 2081 return (error); 2082 } 2083 2084 flags = PGET_HOLD | PGET_NOTWEXIT; 2085 if (args->new != NULL) 2086 flags |= PGET_CANDEBUG; 2087 else 2088 flags |= PGET_CANSEE; 2089 error = pget(args->pid, flags, &p); 2090 if (error != 0) 2091 return (error); 2092 2093 if (args->old != NULL) { 2094 PROC_LOCK(p); 2095 lim_rlimit(p, which, &rlim); 2096 PROC_UNLOCK(p); 2097 if (rlim.rlim_cur == RLIM_INFINITY) 2098 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2099 else 2100 lrlim.rlim_cur = rlim.rlim_cur; 2101 if (rlim.rlim_max == RLIM_INFINITY) 2102 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2103 else 2104 lrlim.rlim_max = rlim.rlim_max; 2105 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2106 if (error != 0) 2107 goto out; 2108 } 2109 2110 if (args->new != NULL) 2111 error = kern_proc_setrlimit(td, p, which, &nrlim); 2112 2113 out: 2114 PRELE(p); 2115 return (error); 2116} 2117 2118int 2119linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2120{ 2121 struct timeval utv, tv0, tv1, *tvp; 2122 struct l_pselect6arg lpse6; 2123 struct l_timespec lts; 2124 struct timespec uts; 2125 l_sigset_t l_ss; 2126 sigset_t *ssp; 2127 sigset_t ss; 2128 int error; 2129 2130 ssp = NULL; 2131 if (args->sig != NULL) { 2132 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2133 if (error != 0) 2134 return (error); 2135 if (lpse6.ss_len != sizeof(l_ss)) 2136 return (EINVAL); 2137 if (lpse6.ss != 0) { 2138 error = copyin(PTRIN(lpse6.ss), &l_ss, 2139 sizeof(l_ss)); 2140 if (error != 0) 2141 return (error); 2142 linux_to_bsd_sigset(&l_ss, &ss); 2143 ssp = &ss; 2144 } 2145 } 2146 2147 /* 2148 * Currently glibc changes nanosecond number to microsecond. 2149 * This mean losing precision but for now it is hardly seen. 2150 */ 2151 if (args->tsp != NULL) { 2152 error = copyin(args->tsp, <s, sizeof(lts)); 2153 if (error != 0) 2154 return (error); 2155 uts.tv_sec = lts.tv_sec; 2156 uts.tv_nsec = lts.tv_nsec; 2157 2158 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2159 if (itimerfix(&utv)) 2160 return (EINVAL); 2161 2162 microtime(&tv0); 2163 tvp = &utv; 2164 } else 2165 tvp = NULL; 2166 2167 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2168 args->exceptfds, tvp, ssp, sizeof(l_int) * 8); 2169 2170 if (error == 0 && args->tsp != NULL) { 2171 if (td->td_retval[0] != 0) { 2172 /* 2173 * Compute how much time was left of the timeout, 2174 * by subtracting the current time and the time 2175 * before we started the call, and subtracting 2176 * that result from the user-supplied value. 2177 */ 2178 2179 microtime(&tv1); 2180 timevalsub(&tv1, &tv0); 2181 timevalsub(&utv, &tv1); 2182 if (utv.tv_sec < 0) 2183 timevalclear(&utv); 2184 } else 2185 timevalclear(&utv); 2186 2187 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2188 lts.tv_sec = uts.tv_sec; 2189 lts.tv_nsec = uts.tv_nsec; 2190 error = copyout(<s, args->tsp, sizeof(lts)); 2191 } 2192 2193 return (error); 2194} 2195 2196int 2197linux_sched_rr_get_interval(struct thread *td, 2198 struct linux_sched_rr_get_interval_args *uap) 2199{ 2200 struct timespec ts; 2201 struct l_timespec lts; 2202 struct thread *tdt; 2203 int error; 2204 2205 /* 2206 * According to man in case the invalid pid specified 2207 * EINVAL should be returned. 2208 */ 2209 if (uap->pid < 0) 2210 return (EINVAL); 2211 2212 tdt = linux_tdfind(td, uap->pid, -1); 2213 if (tdt == NULL) 2214 return (ESRCH); 2215 2216 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2217 PROC_UNLOCK(tdt->td_proc); 2218 if (error != 0) 2219 return (error); 2220 lts.tv_sec = ts.tv_sec; 2221 lts.tv_nsec = ts.tv_nsec; 2222 return (copyout(<s, uap->interval, sizeof(lts))); 2223} 2224 2225/* 2226 * In case when the Linux thread is the initial thread in 2227 * the thread group thread id is equal to the process id. 2228 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2229 */ 2230struct thread * 2231linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2232{ 2233 struct linux_emuldata *em; 2234 struct thread *tdt; 2235 struct proc *p; 2236 2237 tdt = NULL; 2238 if (tid == 0 || tid == td->td_tid) { 2239 tdt = td; 2240 PROC_LOCK(tdt->td_proc); 2241 } else if (tid > PID_MAX) 2242 tdt = tdfind(tid, pid); 2243 else { 2244 /* 2245 * Initial thread where the tid equal to the pid. 2246 */ 2247 p = pfind(tid); 2248 if (p != NULL) { 2249 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2250 /* 2251 * p is not a Linuxulator process. 2252 */ 2253 PROC_UNLOCK(p); 2254 return (NULL); 2255 } 2256 FOREACH_THREAD_IN_PROC(p, tdt) { 2257 em = em_find(tdt); 2258 if (tid == em->em_tid) 2259 return (tdt); 2260 } 2261 PROC_UNLOCK(p); 2262 } 2263 return (NULL); 2264 } 2265 2266 return (tdt); 2267} 2268 2269void 2270linux_to_bsd_waitopts(int options, int *bsdopts) 2271{ 2272 2273 if (options & LINUX_WNOHANG) 2274 *bsdopts |= WNOHANG; 2275 if (options & LINUX_WUNTRACED) 2276 *bsdopts |= WUNTRACED; 2277 if (options & LINUX_WEXITED) 2278 *bsdopts |= WEXITED; 2279 if (options & LINUX_WCONTINUED) 2280 *bsdopts |= WCONTINUED; 2281 if (options & LINUX_WNOWAIT) 2282 *bsdopts |= WNOWAIT; 2283 2284 if (options & __WCLONE) 2285 *bsdopts |= WLINUXCLONE; 2286} 2287