linux_misc.c revision 293524
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 293524 2016-01-09 15:58:42Z 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, %ju"), path, args->mode, 997 (uintmax_t)args->dev); 998#endif 999 1000 switch (args->mode & S_IFMT) { 1001 case S_IFIFO: 1002 case S_IFSOCK: 1003 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1004 break; 1005 1006 case S_IFCHR: 1007 case S_IFBLK: 1008 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1009 args->dev); 1010 break; 1011 1012 case S_IFDIR: 1013 error = EPERM; 1014 break; 1015 1016 case 0: 1017 args->mode |= S_IFREG; 1018 /* FALLTHROUGH */ 1019 case S_IFREG: 1020 error = kern_open(td, path, UIO_SYSSPACE, 1021 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1022 if (error == 0) 1023 kern_close(td, td->td_retval[0]); 1024 break; 1025 1026 default: 1027 error = EINVAL; 1028 break; 1029 } 1030 LFREEPATH(path); 1031 return (error); 1032} 1033 1034int 1035linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1036{ 1037 char *path; 1038 int error, dfd; 1039 1040 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1041 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1042 1043#ifdef DEBUG 1044 if (ldebug(mknodat)) 1045 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1046#endif 1047 1048 switch (args->mode & S_IFMT) { 1049 case S_IFIFO: 1050 case S_IFSOCK: 1051 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1052 break; 1053 1054 case S_IFCHR: 1055 case S_IFBLK: 1056 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1057 args->dev); 1058 break; 1059 1060 case S_IFDIR: 1061 error = EPERM; 1062 break; 1063 1064 case 0: 1065 args->mode |= S_IFREG; 1066 /* FALLTHROUGH */ 1067 case S_IFREG: 1068 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1069 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1070 if (error == 0) 1071 kern_close(td, td->td_retval[0]); 1072 break; 1073 1074 default: 1075 error = EINVAL; 1076 break; 1077 } 1078 LFREEPATH(path); 1079 return (error); 1080} 1081 1082/* 1083 * UGH! This is just about the dumbest idea I've ever heard!! 1084 */ 1085int 1086linux_personality(struct thread *td, struct linux_personality_args *args) 1087{ 1088#ifdef DEBUG 1089 if (ldebug(personality)) 1090 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1091#endif 1092 if (args->per != 0) 1093 return (EINVAL); 1094 1095 /* Yes Jim, it's still a Linux... */ 1096 td->td_retval[0] = 0; 1097 return (0); 1098} 1099 1100struct l_itimerval { 1101 l_timeval it_interval; 1102 l_timeval it_value; 1103}; 1104 1105#define B2L_ITIMERVAL(bip, lip) \ 1106 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1107 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1108 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1109 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1110 1111int 1112linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1113{ 1114 int error; 1115 struct l_itimerval ls; 1116 struct itimerval aitv, oitv; 1117 1118#ifdef DEBUG 1119 if (ldebug(setitimer)) 1120 printf(ARGS(setitimer, "%p, %p"), 1121 (void *)uap->itv, (void *)uap->oitv); 1122#endif 1123 1124 if (uap->itv == NULL) { 1125 uap->itv = uap->oitv; 1126 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1127 } 1128 1129 error = copyin(uap->itv, &ls, sizeof(ls)); 1130 if (error != 0) 1131 return (error); 1132 B2L_ITIMERVAL(&aitv, &ls); 1133#ifdef DEBUG 1134 if (ldebug(setitimer)) { 1135 printf("setitimer: value: sec: %jd, usec: %ld\n", 1136 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1137 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1138 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1139 } 1140#endif 1141 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1142 if (error != 0 || uap->oitv == NULL) 1143 return (error); 1144 B2L_ITIMERVAL(&ls, &oitv); 1145 1146 return (copyout(&ls, uap->oitv, sizeof(ls))); 1147} 1148 1149int 1150linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1151{ 1152 int error; 1153 struct l_itimerval ls; 1154 struct itimerval aitv; 1155 1156#ifdef DEBUG 1157 if (ldebug(getitimer)) 1158 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1159#endif 1160 error = kern_getitimer(td, uap->which, &aitv); 1161 if (error != 0) 1162 return (error); 1163 B2L_ITIMERVAL(&ls, &aitv); 1164 return (copyout(&ls, uap->itv, sizeof(ls))); 1165} 1166 1167#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1168int 1169linux_nice(struct thread *td, struct linux_nice_args *args) 1170{ 1171 struct setpriority_args bsd_args; 1172 1173 bsd_args.which = PRIO_PROCESS; 1174 bsd_args.who = 0; /* current process */ 1175 bsd_args.prio = args->inc; 1176 return (sys_setpriority(td, &bsd_args)); 1177} 1178#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1179 1180int 1181linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1182{ 1183 struct ucred *newcred, *oldcred; 1184 l_gid_t *linux_gidset; 1185 gid_t *bsd_gidset; 1186 int ngrp, error; 1187 struct proc *p; 1188 1189 ngrp = args->gidsetsize; 1190 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1191 return (EINVAL); 1192 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK); 1193 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1194 if (error) 1195 goto out; 1196 newcred = crget(); 1197 p = td->td_proc; 1198 PROC_LOCK(p); 1199 oldcred = crcopysafe(p, newcred); 1200 1201 /* 1202 * cr_groups[0] holds egid. Setting the whole set from 1203 * the supplied set will cause egid to be changed too. 1204 * Keep cr_groups[0] unchanged to prevent that. 1205 */ 1206 1207 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1208 PROC_UNLOCK(p); 1209 crfree(newcred); 1210 goto out; 1211 } 1212 1213 if (ngrp > 0) { 1214 newcred->cr_ngroups = ngrp + 1; 1215 1216 bsd_gidset = newcred->cr_groups; 1217 ngrp--; 1218 while (ngrp >= 0) { 1219 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1220 ngrp--; 1221 } 1222 } else 1223 newcred->cr_ngroups = 1; 1224 1225 setsugid(p); 1226 p->p_ucred = newcred; 1227 PROC_UNLOCK(p); 1228 crfree(oldcred); 1229 error = 0; 1230out: 1231 free(linux_gidset, M_TEMP); 1232 return (error); 1233} 1234 1235int 1236linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1237{ 1238 struct ucred *cred; 1239 l_gid_t *linux_gidset; 1240 gid_t *bsd_gidset; 1241 int bsd_gidsetsz, ngrp, error; 1242 1243 cred = td->td_ucred; 1244 bsd_gidset = cred->cr_groups; 1245 bsd_gidsetsz = cred->cr_ngroups - 1; 1246 1247 /* 1248 * cr_groups[0] holds egid. Returning the whole set 1249 * here will cause a duplicate. Exclude cr_groups[0] 1250 * to prevent that. 1251 */ 1252 1253 if ((ngrp = args->gidsetsize) == 0) { 1254 td->td_retval[0] = bsd_gidsetsz; 1255 return (0); 1256 } 1257 1258 if (ngrp < bsd_gidsetsz) 1259 return (EINVAL); 1260 1261 ngrp = 0; 1262 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1263 M_TEMP, M_WAITOK); 1264 while (ngrp < bsd_gidsetsz) { 1265 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1266 ngrp++; 1267 } 1268 1269 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1270 free(linux_gidset, M_TEMP); 1271 if (error) 1272 return (error); 1273 1274 td->td_retval[0] = ngrp; 1275 return (0); 1276} 1277 1278int 1279linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1280{ 1281 struct rlimit bsd_rlim; 1282 struct l_rlimit rlim; 1283 u_int which; 1284 int error; 1285 1286#ifdef DEBUG 1287 if (ldebug(setrlimit)) 1288 printf(ARGS(setrlimit, "%d, %p"), 1289 args->resource, (void *)args->rlim); 1290#endif 1291 1292 if (args->resource >= LINUX_RLIM_NLIMITS) 1293 return (EINVAL); 1294 1295 which = linux_to_bsd_resource[args->resource]; 1296 if (which == -1) 1297 return (EINVAL); 1298 1299 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1300 if (error) 1301 return (error); 1302 1303 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1304 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1305 return (kern_setrlimit(td, which, &bsd_rlim)); 1306} 1307 1308#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1309int 1310linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1311{ 1312 struct l_rlimit rlim; 1313 struct proc *p = td->td_proc; 1314 struct rlimit bsd_rlim; 1315 u_int which; 1316 1317#ifdef DEBUG 1318 if (ldebug(old_getrlimit)) 1319 printf(ARGS(old_getrlimit, "%d, %p"), 1320 args->resource, (void *)args->rlim); 1321#endif 1322 1323 if (args->resource >= LINUX_RLIM_NLIMITS) 1324 return (EINVAL); 1325 1326 which = linux_to_bsd_resource[args->resource]; 1327 if (which == -1) 1328 return (EINVAL); 1329 1330 PROC_LOCK(p); 1331 lim_rlimit(p, which, &bsd_rlim); 1332 PROC_UNLOCK(p); 1333 1334#ifdef COMPAT_LINUX32 1335 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1336 if (rlim.rlim_cur == UINT_MAX) 1337 rlim.rlim_cur = INT_MAX; 1338 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1339 if (rlim.rlim_max == UINT_MAX) 1340 rlim.rlim_max = INT_MAX; 1341#else 1342 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1343 if (rlim.rlim_cur == ULONG_MAX) 1344 rlim.rlim_cur = LONG_MAX; 1345 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1346 if (rlim.rlim_max == ULONG_MAX) 1347 rlim.rlim_max = LONG_MAX; 1348#endif 1349 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1350} 1351#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1352 1353int 1354linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1355{ 1356 struct l_rlimit rlim; 1357 struct proc *p = td->td_proc; 1358 struct rlimit bsd_rlim; 1359 u_int which; 1360 1361#ifdef DEBUG 1362 if (ldebug(getrlimit)) 1363 printf(ARGS(getrlimit, "%d, %p"), 1364 args->resource, (void *)args->rlim); 1365#endif 1366 1367 if (args->resource >= LINUX_RLIM_NLIMITS) 1368 return (EINVAL); 1369 1370 which = linux_to_bsd_resource[args->resource]; 1371 if (which == -1) 1372 return (EINVAL); 1373 1374 PROC_LOCK(p); 1375 lim_rlimit(p, which, &bsd_rlim); 1376 PROC_UNLOCK(p); 1377 1378 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1379 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1380 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1381} 1382 1383int 1384linux_sched_setscheduler(struct thread *td, 1385 struct linux_sched_setscheduler_args *args) 1386{ 1387 struct sched_param sched_param; 1388 struct thread *tdt; 1389 int error, policy; 1390 1391#ifdef DEBUG 1392 if (ldebug(sched_setscheduler)) 1393 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1394 args->pid, args->policy, (const void *)args->param); 1395#endif 1396 1397 switch (args->policy) { 1398 case LINUX_SCHED_OTHER: 1399 policy = SCHED_OTHER; 1400 break; 1401 case LINUX_SCHED_FIFO: 1402 policy = SCHED_FIFO; 1403 break; 1404 case LINUX_SCHED_RR: 1405 policy = SCHED_RR; 1406 break; 1407 default: 1408 return (EINVAL); 1409 } 1410 1411 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1412 if (error) 1413 return (error); 1414 1415 tdt = linux_tdfind(td, args->pid, -1); 1416 if (tdt == NULL) 1417 return (ESRCH); 1418 1419 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1420 PROC_UNLOCK(tdt->td_proc); 1421 return (error); 1422} 1423 1424int 1425linux_sched_getscheduler(struct thread *td, 1426 struct linux_sched_getscheduler_args *args) 1427{ 1428 struct thread *tdt; 1429 int error, policy; 1430 1431#ifdef DEBUG 1432 if (ldebug(sched_getscheduler)) 1433 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1434#endif 1435 1436 tdt = linux_tdfind(td, args->pid, -1); 1437 if (tdt == NULL) 1438 return (ESRCH); 1439 1440 error = kern_sched_getscheduler(td, tdt, &policy); 1441 PROC_UNLOCK(tdt->td_proc); 1442 1443 switch (policy) { 1444 case SCHED_OTHER: 1445 td->td_retval[0] = LINUX_SCHED_OTHER; 1446 break; 1447 case SCHED_FIFO: 1448 td->td_retval[0] = LINUX_SCHED_FIFO; 1449 break; 1450 case SCHED_RR: 1451 td->td_retval[0] = LINUX_SCHED_RR; 1452 break; 1453 } 1454 return (error); 1455} 1456 1457int 1458linux_sched_get_priority_max(struct thread *td, 1459 struct linux_sched_get_priority_max_args *args) 1460{ 1461 struct sched_get_priority_max_args bsd; 1462 1463#ifdef DEBUG 1464 if (ldebug(sched_get_priority_max)) 1465 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1466#endif 1467 1468 switch (args->policy) { 1469 case LINUX_SCHED_OTHER: 1470 bsd.policy = SCHED_OTHER; 1471 break; 1472 case LINUX_SCHED_FIFO: 1473 bsd.policy = SCHED_FIFO; 1474 break; 1475 case LINUX_SCHED_RR: 1476 bsd.policy = SCHED_RR; 1477 break; 1478 default: 1479 return (EINVAL); 1480 } 1481 return (sys_sched_get_priority_max(td, &bsd)); 1482} 1483 1484int 1485linux_sched_get_priority_min(struct thread *td, 1486 struct linux_sched_get_priority_min_args *args) 1487{ 1488 struct sched_get_priority_min_args bsd; 1489 1490#ifdef DEBUG 1491 if (ldebug(sched_get_priority_min)) 1492 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1493#endif 1494 1495 switch (args->policy) { 1496 case LINUX_SCHED_OTHER: 1497 bsd.policy = SCHED_OTHER; 1498 break; 1499 case LINUX_SCHED_FIFO: 1500 bsd.policy = SCHED_FIFO; 1501 break; 1502 case LINUX_SCHED_RR: 1503 bsd.policy = SCHED_RR; 1504 break; 1505 default: 1506 return (EINVAL); 1507 } 1508 return (sys_sched_get_priority_min(td, &bsd)); 1509} 1510 1511#define REBOOT_CAD_ON 0x89abcdef 1512#define REBOOT_CAD_OFF 0 1513#define REBOOT_HALT 0xcdef0123 1514#define REBOOT_RESTART 0x01234567 1515#define REBOOT_RESTART2 0xA1B2C3D4 1516#define REBOOT_POWEROFF 0x4321FEDC 1517#define REBOOT_MAGIC1 0xfee1dead 1518#define REBOOT_MAGIC2 0x28121969 1519#define REBOOT_MAGIC2A 0x05121996 1520#define REBOOT_MAGIC2B 0x16041998 1521 1522int 1523linux_reboot(struct thread *td, struct linux_reboot_args *args) 1524{ 1525 struct reboot_args bsd_args; 1526 1527#ifdef DEBUG 1528 if (ldebug(reboot)) 1529 printf(ARGS(reboot, "0x%x"), args->cmd); 1530#endif 1531 1532 if (args->magic1 != REBOOT_MAGIC1) 1533 return (EINVAL); 1534 1535 switch (args->magic2) { 1536 case REBOOT_MAGIC2: 1537 case REBOOT_MAGIC2A: 1538 case REBOOT_MAGIC2B: 1539 break; 1540 default: 1541 return (EINVAL); 1542 } 1543 1544 switch (args->cmd) { 1545 case REBOOT_CAD_ON: 1546 case REBOOT_CAD_OFF: 1547 return (priv_check(td, PRIV_REBOOT)); 1548 case REBOOT_HALT: 1549 bsd_args.opt = RB_HALT; 1550 break; 1551 case REBOOT_RESTART: 1552 case REBOOT_RESTART2: 1553 bsd_args.opt = 0; 1554 break; 1555 case REBOOT_POWEROFF: 1556 bsd_args.opt = RB_POWEROFF; 1557 break; 1558 default: 1559 return (EINVAL); 1560 } 1561 return (sys_reboot(td, &bsd_args)); 1562} 1563 1564 1565/* 1566 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1567 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1568 * are assumed to be preserved. The following lightweight syscalls fixes 1569 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1570 * 1571 * linux_getpid() - MP SAFE 1572 * linux_getgid() - MP SAFE 1573 * linux_getuid() - MP SAFE 1574 */ 1575 1576int 1577linux_getpid(struct thread *td, struct linux_getpid_args *args) 1578{ 1579 1580#ifdef DEBUG 1581 if (ldebug(getpid)) 1582 printf(ARGS(getpid, "")); 1583#endif 1584 td->td_retval[0] = td->td_proc->p_pid; 1585 1586 return (0); 1587} 1588 1589int 1590linux_gettid(struct thread *td, struct linux_gettid_args *args) 1591{ 1592 struct linux_emuldata *em; 1593 1594#ifdef DEBUG 1595 if (ldebug(gettid)) 1596 printf(ARGS(gettid, "")); 1597#endif 1598 1599 em = em_find(td); 1600 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1601 1602 td->td_retval[0] = em->em_tid; 1603 1604 return (0); 1605} 1606 1607 1608int 1609linux_getppid(struct thread *td, struct linux_getppid_args *args) 1610{ 1611 1612#ifdef DEBUG 1613 if (ldebug(getppid)) 1614 printf(ARGS(getppid, "")); 1615#endif 1616 1617 PROC_LOCK(td->td_proc); 1618 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1619 PROC_UNLOCK(td->td_proc); 1620 return (0); 1621} 1622 1623int 1624linux_getgid(struct thread *td, struct linux_getgid_args *args) 1625{ 1626 1627#ifdef DEBUG 1628 if (ldebug(getgid)) 1629 printf(ARGS(getgid, "")); 1630#endif 1631 1632 td->td_retval[0] = td->td_ucred->cr_rgid; 1633 return (0); 1634} 1635 1636int 1637linux_getuid(struct thread *td, struct linux_getuid_args *args) 1638{ 1639 1640#ifdef DEBUG 1641 if (ldebug(getuid)) 1642 printf(ARGS(getuid, "")); 1643#endif 1644 1645 td->td_retval[0] = td->td_ucred->cr_ruid; 1646 return (0); 1647} 1648 1649 1650int 1651linux_getsid(struct thread *td, struct linux_getsid_args *args) 1652{ 1653 struct getsid_args bsd; 1654 1655#ifdef DEBUG 1656 if (ldebug(getsid)) 1657 printf(ARGS(getsid, "%i"), args->pid); 1658#endif 1659 1660 bsd.pid = args->pid; 1661 return (sys_getsid(td, &bsd)); 1662} 1663 1664int 1665linux_nosys(struct thread *td, struct nosys_args *ignore) 1666{ 1667 1668 return (ENOSYS); 1669} 1670 1671int 1672linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1673{ 1674 struct getpriority_args bsd_args; 1675 int error; 1676 1677#ifdef DEBUG 1678 if (ldebug(getpriority)) 1679 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1680#endif 1681 1682 bsd_args.which = args->which; 1683 bsd_args.who = args->who; 1684 error = sys_getpriority(td, &bsd_args); 1685 td->td_retval[0] = 20 - td->td_retval[0]; 1686 return (error); 1687} 1688 1689int 1690linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1691{ 1692 int name[2]; 1693 1694#ifdef DEBUG 1695 if (ldebug(sethostname)) 1696 printf(ARGS(sethostname, "*, %i"), args->len); 1697#endif 1698 1699 name[0] = CTL_KERN; 1700 name[1] = KERN_HOSTNAME; 1701 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1702 args->len, 0, 0)); 1703} 1704 1705int 1706linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1707{ 1708 int name[2]; 1709 1710#ifdef DEBUG 1711 if (ldebug(setdomainname)) 1712 printf(ARGS(setdomainname, "*, %i"), args->len); 1713#endif 1714 1715 name[0] = CTL_KERN; 1716 name[1] = KERN_NISDOMAINNAME; 1717 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1718 args->len, 0, 0)); 1719} 1720 1721int 1722linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1723{ 1724 1725#ifdef DEBUG 1726 if (ldebug(exit_group)) 1727 printf(ARGS(exit_group, "%i"), args->error_code); 1728#endif 1729 1730 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1731 args->error_code); 1732 1733 /* 1734 * XXX: we should send a signal to the parent if 1735 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1736 * as it doesnt occur often. 1737 */ 1738 exit1(td, W_EXITCODE(args->error_code, 0)); 1739 /* NOTREACHED */ 1740} 1741 1742#define _LINUX_CAPABILITY_VERSION 0x19980330 1743 1744struct l_user_cap_header { 1745 l_int version; 1746 l_int pid; 1747}; 1748 1749struct l_user_cap_data { 1750 l_int effective; 1751 l_int permitted; 1752 l_int inheritable; 1753}; 1754 1755int 1756linux_capget(struct thread *td, struct linux_capget_args *args) 1757{ 1758 struct l_user_cap_header luch; 1759 struct l_user_cap_data lucd; 1760 int error; 1761 1762 if (args->hdrp == NULL) 1763 return (EFAULT); 1764 1765 error = copyin(args->hdrp, &luch, sizeof(luch)); 1766 if (error != 0) 1767 return (error); 1768 1769 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1770 luch.version = _LINUX_CAPABILITY_VERSION; 1771 error = copyout(&luch, args->hdrp, sizeof(luch)); 1772 if (error) 1773 return (error); 1774 return (EINVAL); 1775 } 1776 1777 if (luch.pid) 1778 return (EPERM); 1779 1780 if (args->datap) { 1781 /* 1782 * The current implementation doesn't support setting 1783 * a capability (it's essentially a stub) so indicate 1784 * that no capabilities are currently set or available 1785 * to request. 1786 */ 1787 bzero (&lucd, sizeof(lucd)); 1788 error = copyout(&lucd, args->datap, sizeof(lucd)); 1789 } 1790 1791 return (error); 1792} 1793 1794int 1795linux_capset(struct thread *td, struct linux_capset_args *args) 1796{ 1797 struct l_user_cap_header luch; 1798 struct l_user_cap_data lucd; 1799 int error; 1800 1801 if (args->hdrp == NULL || args->datap == NULL) 1802 return (EFAULT); 1803 1804 error = copyin(args->hdrp, &luch, sizeof(luch)); 1805 if (error != 0) 1806 return (error); 1807 1808 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1809 luch.version = _LINUX_CAPABILITY_VERSION; 1810 error = copyout(&luch, args->hdrp, sizeof(luch)); 1811 if (error) 1812 return (error); 1813 return (EINVAL); 1814 } 1815 1816 if (luch.pid) 1817 return (EPERM); 1818 1819 error = copyin(args->datap, &lucd, sizeof(lucd)); 1820 if (error != 0) 1821 return (error); 1822 1823 /* We currently don't support setting any capabilities. */ 1824 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1825 linux_msg(td, 1826 "capset effective=0x%x, permitted=0x%x, " 1827 "inheritable=0x%x is not implemented", 1828 (int)lucd.effective, (int)lucd.permitted, 1829 (int)lucd.inheritable); 1830 return (EPERM); 1831 } 1832 1833 return (0); 1834} 1835 1836int 1837linux_prctl(struct thread *td, struct linux_prctl_args *args) 1838{ 1839 int error = 0, max_size; 1840 struct proc *p = td->td_proc; 1841 char comm[LINUX_MAX_COMM_LEN]; 1842 struct linux_emuldata *em; 1843 int pdeath_signal; 1844 1845#ifdef DEBUG 1846 if (ldebug(prctl)) 1847 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1848 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1849 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1850#endif 1851 1852 switch (args->option) { 1853 case LINUX_PR_SET_PDEATHSIG: 1854 if (!LINUX_SIG_VALID(args->arg2)) 1855 return (EINVAL); 1856 em = em_find(td); 1857 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1858 em->pdeath_signal = args->arg2; 1859 break; 1860 case LINUX_PR_GET_PDEATHSIG: 1861 em = em_find(td); 1862 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1863 pdeath_signal = em->pdeath_signal; 1864 error = copyout(&pdeath_signal, 1865 (void *)(register_t)args->arg2, 1866 sizeof(pdeath_signal)); 1867 break; 1868 case LINUX_PR_GET_KEEPCAPS: 1869 /* 1870 * Indicate that we always clear the effective and 1871 * permitted capability sets when the user id becomes 1872 * non-zero (actually the capability sets are simply 1873 * always zero in the current implementation). 1874 */ 1875 td->td_retval[0] = 0; 1876 break; 1877 case LINUX_PR_SET_KEEPCAPS: 1878 /* 1879 * Ignore requests to keep the effective and permitted 1880 * capability sets when the user id becomes non-zero. 1881 */ 1882 break; 1883 case LINUX_PR_SET_NAME: 1884 /* 1885 * To be on the safe side we need to make sure to not 1886 * overflow the size a linux program expects. We already 1887 * do this here in the copyin, so that we don't need to 1888 * check on copyout. 1889 */ 1890 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1891 error = copyinstr((void *)(register_t)args->arg2, comm, 1892 max_size, NULL); 1893 1894 /* Linux silently truncates the name if it is too long. */ 1895 if (error == ENAMETOOLONG) { 1896 /* 1897 * XXX: copyinstr() isn't documented to populate the 1898 * array completely, so do a copyin() to be on the 1899 * safe side. This should be changed in case 1900 * copyinstr() is changed to guarantee this. 1901 */ 1902 error = copyin((void *)(register_t)args->arg2, comm, 1903 max_size - 1); 1904 comm[max_size - 1] = '\0'; 1905 } 1906 if (error) 1907 return (error); 1908 1909 PROC_LOCK(p); 1910 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1911 PROC_UNLOCK(p); 1912 break; 1913 case LINUX_PR_GET_NAME: 1914 PROC_LOCK(p); 1915 strlcpy(comm, p->p_comm, sizeof(comm)); 1916 PROC_UNLOCK(p); 1917 error = copyout(comm, (void *)(register_t)args->arg2, 1918 strlen(comm) + 1); 1919 break; 1920 default: 1921 error = EINVAL; 1922 break; 1923 } 1924 1925 return (error); 1926} 1927 1928int 1929linux_sched_setparam(struct thread *td, 1930 struct linux_sched_setparam_args *uap) 1931{ 1932 struct sched_param sched_param; 1933 struct thread *tdt; 1934 int error; 1935 1936#ifdef DEBUG 1937 if (ldebug(sched_setparam)) 1938 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 1939#endif 1940 1941 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1942 if (error) 1943 return (error); 1944 1945 tdt = linux_tdfind(td, uap->pid, -1); 1946 if (tdt == NULL) 1947 return (ESRCH); 1948 1949 error = kern_sched_setparam(td, tdt, &sched_param); 1950 PROC_UNLOCK(tdt->td_proc); 1951 return (error); 1952} 1953 1954int 1955linux_sched_getparam(struct thread *td, 1956 struct linux_sched_getparam_args *uap) 1957{ 1958 struct sched_param sched_param; 1959 struct thread *tdt; 1960 int error; 1961 1962#ifdef DEBUG 1963 if (ldebug(sched_getparam)) 1964 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 1965#endif 1966 1967 tdt = linux_tdfind(td, uap->pid, -1); 1968 if (tdt == NULL) 1969 return (ESRCH); 1970 1971 error = kern_sched_getparam(td, tdt, &sched_param); 1972 PROC_UNLOCK(tdt->td_proc); 1973 if (error == 0) 1974 error = copyout(&sched_param, uap->param, 1975 sizeof(sched_param)); 1976 return (error); 1977} 1978 1979/* 1980 * Get affinity of a process. 1981 */ 1982int 1983linux_sched_getaffinity(struct thread *td, 1984 struct linux_sched_getaffinity_args *args) 1985{ 1986 int error; 1987 struct thread *tdt; 1988 struct cpuset_getaffinity_args cga; 1989 1990#ifdef DEBUG 1991 if (ldebug(sched_getaffinity)) 1992 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 1993 args->len); 1994#endif 1995 if (args->len < sizeof(cpuset_t)) 1996 return (EINVAL); 1997 1998 tdt = linux_tdfind(td, args->pid, -1); 1999 if (tdt == NULL) 2000 return (ESRCH); 2001 2002 PROC_UNLOCK(tdt->td_proc); 2003 cga.level = CPU_LEVEL_WHICH; 2004 cga.which = CPU_WHICH_TID; 2005 cga.id = tdt->td_tid; 2006 cga.cpusetsize = sizeof(cpuset_t); 2007 cga.mask = (cpuset_t *) args->user_mask_ptr; 2008 2009 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2010 td->td_retval[0] = sizeof(cpuset_t); 2011 2012 return (error); 2013} 2014 2015/* 2016 * Set affinity of a process. 2017 */ 2018int 2019linux_sched_setaffinity(struct thread *td, 2020 struct linux_sched_setaffinity_args *args) 2021{ 2022 struct cpuset_setaffinity_args csa; 2023 struct thread *tdt; 2024 2025#ifdef DEBUG 2026 if (ldebug(sched_setaffinity)) 2027 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2028 args->len); 2029#endif 2030 if (args->len < sizeof(cpuset_t)) 2031 return (EINVAL); 2032 2033 tdt = linux_tdfind(td, args->pid, -1); 2034 if (tdt == NULL) 2035 return (ESRCH); 2036 2037 PROC_UNLOCK(tdt->td_proc); 2038 csa.level = CPU_LEVEL_WHICH; 2039 csa.which = CPU_WHICH_TID; 2040 csa.id = tdt->td_tid; 2041 csa.cpusetsize = sizeof(cpuset_t); 2042 csa.mask = (cpuset_t *) args->user_mask_ptr; 2043 2044 return (sys_cpuset_setaffinity(td, &csa)); 2045} 2046 2047struct linux_rlimit64 { 2048 uint64_t rlim_cur; 2049 uint64_t rlim_max; 2050}; 2051 2052int 2053linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2054{ 2055 struct rlimit rlim, nrlim; 2056 struct linux_rlimit64 lrlim; 2057 struct proc *p; 2058 u_int which; 2059 int flags; 2060 int error; 2061 2062#ifdef DEBUG 2063 if (ldebug(prlimit64)) 2064 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2065 args->resource, (void *)args->new, (void *)args->old); 2066#endif 2067 2068 if (args->resource >= LINUX_RLIM_NLIMITS) 2069 return (EINVAL); 2070 2071 which = linux_to_bsd_resource[args->resource]; 2072 if (which == -1) 2073 return (EINVAL); 2074 2075 if (args->new != NULL) { 2076 /* 2077 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2078 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2079 * as INFINITY so we do not need a conversion even. 2080 */ 2081 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2082 if (error != 0) 2083 return (error); 2084 } 2085 2086 flags = PGET_HOLD | PGET_NOTWEXIT; 2087 if (args->new != NULL) 2088 flags |= PGET_CANDEBUG; 2089 else 2090 flags |= PGET_CANSEE; 2091 error = pget(args->pid, flags, &p); 2092 if (error != 0) 2093 return (error); 2094 2095 if (args->old != NULL) { 2096 PROC_LOCK(p); 2097 lim_rlimit(p, which, &rlim); 2098 PROC_UNLOCK(p); 2099 if (rlim.rlim_cur == RLIM_INFINITY) 2100 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2101 else 2102 lrlim.rlim_cur = rlim.rlim_cur; 2103 if (rlim.rlim_max == RLIM_INFINITY) 2104 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2105 else 2106 lrlim.rlim_max = rlim.rlim_max; 2107 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2108 if (error != 0) 2109 goto out; 2110 } 2111 2112 if (args->new != NULL) 2113 error = kern_proc_setrlimit(td, p, which, &nrlim); 2114 2115 out: 2116 PRELE(p); 2117 return (error); 2118} 2119 2120int 2121linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2122{ 2123 struct timeval utv, tv0, tv1, *tvp; 2124 struct l_pselect6arg lpse6; 2125 struct l_timespec lts; 2126 struct timespec uts; 2127 l_sigset_t l_ss; 2128 sigset_t *ssp; 2129 sigset_t ss; 2130 int error; 2131 2132 ssp = NULL; 2133 if (args->sig != NULL) { 2134 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2135 if (error != 0) 2136 return (error); 2137 if (lpse6.ss_len != sizeof(l_ss)) 2138 return (EINVAL); 2139 if (lpse6.ss != 0) { 2140 error = copyin(PTRIN(lpse6.ss), &l_ss, 2141 sizeof(l_ss)); 2142 if (error != 0) 2143 return (error); 2144 linux_to_bsd_sigset(&l_ss, &ss); 2145 ssp = &ss; 2146 } 2147 } 2148 2149 /* 2150 * Currently glibc changes nanosecond number to microsecond. 2151 * This mean losing precision but for now it is hardly seen. 2152 */ 2153 if (args->tsp != NULL) { 2154 error = copyin(args->tsp, <s, sizeof(lts)); 2155 if (error != 0) 2156 return (error); 2157 uts.tv_sec = lts.tv_sec; 2158 uts.tv_nsec = lts.tv_nsec; 2159 2160 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2161 if (itimerfix(&utv)) 2162 return (EINVAL); 2163 2164 microtime(&tv0); 2165 tvp = &utv; 2166 } else 2167 tvp = NULL; 2168 2169 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2170 args->exceptfds, tvp, ssp, sizeof(l_int) * 8); 2171 2172 if (error == 0 && args->tsp != NULL) { 2173 if (td->td_retval[0] != 0) { 2174 /* 2175 * Compute how much time was left of the timeout, 2176 * by subtracting the current time and the time 2177 * before we started the call, and subtracting 2178 * that result from the user-supplied value. 2179 */ 2180 2181 microtime(&tv1); 2182 timevalsub(&tv1, &tv0); 2183 timevalsub(&utv, &tv1); 2184 if (utv.tv_sec < 0) 2185 timevalclear(&utv); 2186 } else 2187 timevalclear(&utv); 2188 2189 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2190 lts.tv_sec = uts.tv_sec; 2191 lts.tv_nsec = uts.tv_nsec; 2192 error = copyout(<s, args->tsp, sizeof(lts)); 2193 } 2194 2195 return (error); 2196} 2197 2198int 2199linux_sched_rr_get_interval(struct thread *td, 2200 struct linux_sched_rr_get_interval_args *uap) 2201{ 2202 struct timespec ts; 2203 struct l_timespec lts; 2204 struct thread *tdt; 2205 int error; 2206 2207 /* 2208 * According to man in case the invalid pid specified 2209 * EINVAL should be returned. 2210 */ 2211 if (uap->pid < 0) 2212 return (EINVAL); 2213 2214 tdt = linux_tdfind(td, uap->pid, -1); 2215 if (tdt == NULL) 2216 return (ESRCH); 2217 2218 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2219 PROC_UNLOCK(tdt->td_proc); 2220 if (error != 0) 2221 return (error); 2222 lts.tv_sec = ts.tv_sec; 2223 lts.tv_nsec = ts.tv_nsec; 2224 return (copyout(<s, uap->interval, sizeof(lts))); 2225} 2226 2227/* 2228 * In case when the Linux thread is the initial thread in 2229 * the thread group thread id is equal to the process id. 2230 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2231 */ 2232struct thread * 2233linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2234{ 2235 struct linux_emuldata *em; 2236 struct thread *tdt; 2237 struct proc *p; 2238 2239 tdt = NULL; 2240 if (tid == 0 || tid == td->td_tid) { 2241 tdt = td; 2242 PROC_LOCK(tdt->td_proc); 2243 } else if (tid > PID_MAX) 2244 tdt = tdfind(tid, pid); 2245 else { 2246 /* 2247 * Initial thread where the tid equal to the pid. 2248 */ 2249 p = pfind(tid); 2250 if (p != NULL) { 2251 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2252 /* 2253 * p is not a Linuxulator process. 2254 */ 2255 PROC_UNLOCK(p); 2256 return (NULL); 2257 } 2258 FOREACH_THREAD_IN_PROC(p, tdt) { 2259 em = em_find(tdt); 2260 if (tid == em->em_tid) 2261 return (tdt); 2262 } 2263 PROC_UNLOCK(p); 2264 } 2265 return (NULL); 2266 } 2267 2268 return (tdt); 2269} 2270 2271void 2272linux_to_bsd_waitopts(int options, int *bsdopts) 2273{ 2274 2275 if (options & LINUX_WNOHANG) 2276 *bsdopts |= WNOHANG; 2277 if (options & LINUX_WUNTRACED) 2278 *bsdopts |= WUNTRACED; 2279 if (options & LINUX_WEXITED) 2280 *bsdopts |= WEXITED; 2281 if (options & LINUX_WCONTINUED) 2282 *bsdopts |= WCONTINUED; 2283 if (options & LINUX_WNOWAIT) 2284 *bsdopts |= WNOWAIT; 2285 2286 if (options & __WCLONE) 2287 *bsdopts |= WLINUXCLONE; 2288} 2289