linux_misc.c revision 297780
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 297780 2016-04-10 06:36:58Z 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/sdt.h> 58#include <sys/signalvar.h> 59#include <sys/stat.h> 60#include <sys/syscallsubr.h> 61#include <sys/sysctl.h> 62#include <sys/sysproto.h> 63#include <sys/systm.h> 64#include <sys/time.h> 65#include <sys/vmmeter.h> 66#include <sys/vnode.h> 67#include <sys/wait.h> 68#include <sys/cpuset.h> 69 70#include <security/mac/mac_framework.h> 71 72#include <vm/vm.h> 73#include <vm/pmap.h> 74#include <vm/vm_kern.h> 75#include <vm/vm_map.h> 76#include <vm/vm_extern.h> 77#include <vm/vm_object.h> 78#include <vm/swap_pager.h> 79 80#ifdef COMPAT_LINUX32 81#include <machine/../linux32/linux.h> 82#include <machine/../linux32/linux32_proto.h> 83#else 84#include <machine/../linux/linux.h> 85#include <machine/../linux/linux_proto.h> 86#endif 87 88#include <compat/linux/linux_dtrace.h> 89#include <compat/linux/linux_file.h> 90#include <compat/linux/linux_mib.h> 91#include <compat/linux/linux_signal.h> 92#include <compat/linux/linux_timer.h> 93#include <compat/linux/linux_util.h> 94#include <compat/linux/linux_sysproto.h> 95#include <compat/linux/linux_emul.h> 96#include <compat/linux/linux_misc.h> 97 98/** 99 * Special DTrace provider for the linuxulator. 100 * 101 * In this file we define the provider for the entire linuxulator. All 102 * modules (= files of the linuxulator) use it. 103 * 104 * We define a different name depending on the emulated bitsize, see 105 * ../../<ARCH>/linux{,32}/linux.h, e.g.: 106 * native bitsize = linuxulator 107 * amd64, 32bit emulation = linuxulator32 108 */ 109LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE); 110 111int stclohz; /* Statistics clock frequency */ 112 113static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 114 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 115 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 116 RLIMIT_MEMLOCK, RLIMIT_AS 117}; 118 119struct l_sysinfo { 120 l_long uptime; /* Seconds since boot */ 121 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 122#define LINUX_SYSINFO_LOADS_SCALE 65536 123 l_ulong totalram; /* Total usable main memory size */ 124 l_ulong freeram; /* Available memory size */ 125 l_ulong sharedram; /* Amount of shared memory */ 126 l_ulong bufferram; /* Memory used by buffers */ 127 l_ulong totalswap; /* Total swap space size */ 128 l_ulong freeswap; /* swap space still available */ 129 l_ushort procs; /* Number of current processes */ 130 l_ushort pads; 131 l_ulong totalbig; 132 l_ulong freebig; 133 l_uint mem_unit; 134 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 135}; 136 137struct l_pselect6arg { 138 l_uintptr_t ss; 139 l_size_t ss_len; 140}; 141 142static int linux_utimensat_nsec_valid(l_long); 143 144 145int 146linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 147{ 148 struct l_sysinfo sysinfo; 149 vm_object_t object; 150 int i, j; 151 struct timespec ts; 152 153 getnanouptime(&ts); 154 if (ts.tv_nsec != 0) 155 ts.tv_sec++; 156 sysinfo.uptime = ts.tv_sec; 157 158 /* Use the information from the mib to get our load averages */ 159 for (i = 0; i < 3; i++) 160 sysinfo.loads[i] = averunnable.ldavg[i] * 161 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 162 163 sysinfo.totalram = physmem * PAGE_SIZE; 164 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE; 165 166 sysinfo.sharedram = 0; 167 mtx_lock(&vm_object_list_mtx); 168 TAILQ_FOREACH(object, &vm_object_list, object_list) 169 if (object->shadow_count > 1) 170 sysinfo.sharedram += object->resident_page_count; 171 mtx_unlock(&vm_object_list_mtx); 172 173 sysinfo.sharedram *= PAGE_SIZE; 174 sysinfo.bufferram = 0; 175 176 swap_pager_status(&i, &j); 177 sysinfo.totalswap = i * PAGE_SIZE; 178 sysinfo.freeswap = (i - j) * PAGE_SIZE; 179 180 sysinfo.procs = nprocs; 181 182 /* The following are only present in newer Linux kernels. */ 183 sysinfo.totalbig = 0; 184 sysinfo.freebig = 0; 185 sysinfo.mem_unit = 1; 186 187 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 188} 189 190int 191linux_alarm(struct thread *td, struct linux_alarm_args *args) 192{ 193 struct itimerval it, old_it; 194 u_int secs; 195 int error; 196 197#ifdef DEBUG 198 if (ldebug(alarm)) 199 printf(ARGS(alarm, "%u"), args->secs); 200#endif 201 secs = args->secs; 202 /* 203 * Linux alarm() is always successfull. Limit secs to INT32_MAX / 2 204 * to match kern_setitimer()'s limit to avoid error from it. 205 * 206 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit 207 * platforms. 208 */ 209 if (secs > INT32_MAX / 2) 210 secs = INT32_MAX / 2; 211 212 it.it_value.tv_sec = secs; 213 it.it_value.tv_usec = 0; 214 timevalclear(&it.it_interval); 215 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 216 KASSERT(error == 0, ("kern_setitimer returns %d", error)); 217 218 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) || 219 old_it.it_value.tv_usec >= 500000) 220 old_it.it_value.tv_sec++; 221 td->td_retval[0] = old_it.it_value.tv_sec; 222 return (0); 223} 224 225int 226linux_brk(struct thread *td, struct linux_brk_args *args) 227{ 228 struct vmspace *vm = td->td_proc->p_vmspace; 229 vm_offset_t new, old; 230 struct obreak_args /* { 231 char * nsize; 232 } */ tmp; 233 234#ifdef DEBUG 235 if (ldebug(brk)) 236 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 237#endif 238 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 239 new = (vm_offset_t)args->dsend; 240 tmp.nsize = (char *)new; 241 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp)) 242 td->td_retval[0] = (long)new; 243 else 244 td->td_retval[0] = (long)old; 245 246 return (0); 247} 248 249#if defined(__i386__) 250/* XXX: what about amd64/linux32? */ 251 252int 253linux_uselib(struct thread *td, struct linux_uselib_args *args) 254{ 255 struct nameidata ni; 256 struct vnode *vp; 257 struct exec *a_out; 258 struct vattr attr; 259 vm_offset_t vmaddr; 260 unsigned long file_offset; 261 unsigned long bss_size; 262 char *library; 263 ssize_t aresid; 264 int error, locked, writecount; 265 266 LCONVPATHEXIST(td, args->library, &library); 267 268#ifdef DEBUG 269 if (ldebug(uselib)) 270 printf(ARGS(uselib, "%s"), library); 271#endif 272 273 a_out = NULL; 274 locked = 0; 275 vp = NULL; 276 277 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 278 UIO_SYSSPACE, library, td); 279 error = namei(&ni); 280 LFREEPATH(library); 281 if (error) 282 goto cleanup; 283 284 vp = ni.ni_vp; 285 NDFREE(&ni, NDF_ONLY_PNBUF); 286 287 /* 288 * From here on down, we have a locked vnode that must be unlocked. 289 * XXX: The code below largely duplicates exec_check_permissions(). 290 */ 291 locked = 1; 292 293 /* Writable? */ 294 error = VOP_GET_WRITECOUNT(vp, &writecount); 295 if (error != 0) 296 goto cleanup; 297 if (writecount != 0) { 298 error = ETXTBSY; 299 goto cleanup; 300 } 301 302 /* Executable? */ 303 error = VOP_GETATTR(vp, &attr, td->td_ucred); 304 if (error) 305 goto cleanup; 306 307 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 308 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 309 /* EACCESS is what exec(2) returns. */ 310 error = ENOEXEC; 311 goto cleanup; 312 } 313 314 /* Sensible size? */ 315 if (attr.va_size == 0) { 316 error = ENOEXEC; 317 goto cleanup; 318 } 319 320 /* Can we access it? */ 321 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 322 if (error) 323 goto cleanup; 324 325 /* 326 * XXX: This should use vn_open() so that it is properly authorized, 327 * and to reduce code redundancy all over the place here. 328 * XXX: Not really, it duplicates far more of exec_check_permissions() 329 * than vn_open(). 330 */ 331#ifdef MAC 332 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 333 if (error) 334 goto cleanup; 335#endif 336 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 337 if (error) 338 goto cleanup; 339 340 /* Pull in executable header into exec_map */ 341 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 342 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 343 if (error) 344 goto cleanup; 345 346 /* Is it a Linux binary ? */ 347 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 348 error = ENOEXEC; 349 goto cleanup; 350 } 351 352 /* 353 * While we are here, we should REALLY do some more checks 354 */ 355 356 /* Set file/virtual offset based on a.out variant. */ 357 switch ((int)(a_out->a_magic & 0xffff)) { 358 case 0413: /* ZMAGIC */ 359 file_offset = 1024; 360 break; 361 case 0314: /* QMAGIC */ 362 file_offset = 0; 363 break; 364 default: 365 error = ENOEXEC; 366 goto cleanup; 367 } 368 369 bss_size = round_page(a_out->a_bss); 370 371 /* Check various fields in header for validity/bounds. */ 372 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 373 error = ENOEXEC; 374 goto cleanup; 375 } 376 377 /* text + data can't exceed file size */ 378 if (a_out->a_data + a_out->a_text > attr.va_size) { 379 error = EFAULT; 380 goto cleanup; 381 } 382 383 /* 384 * text/data/bss must not exceed limits 385 * XXX - this is not complete. it should check current usage PLUS 386 * the resources needed by this library. 387 */ 388 PROC_LOCK(td->td_proc); 389 if (a_out->a_text > maxtsiz || 390 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) || 391 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 392 bss_size) != 0) { 393 PROC_UNLOCK(td->td_proc); 394 error = ENOMEM; 395 goto cleanup; 396 } 397 PROC_UNLOCK(td->td_proc); 398 399 /* 400 * Prevent more writers. 401 * XXX: Note that if any of the VM operations fail below we don't 402 * clear this flag. 403 */ 404 VOP_SET_TEXT(vp); 405 406 /* 407 * Lock no longer needed 408 */ 409 locked = 0; 410 VOP_UNLOCK(vp, 0); 411 412 /* 413 * Check if file_offset page aligned. Currently we cannot handle 414 * misalinged file offsets, and so we read in the entire image 415 * (what a waste). 416 */ 417 if (file_offset & PAGE_MASK) { 418#ifdef DEBUG 419 printf("uselib: Non page aligned binary %lu\n", file_offset); 420#endif 421 /* Map text+data read/write/execute */ 422 423 /* a_entry is the load address and is page aligned */ 424 vmaddr = trunc_page(a_out->a_entry); 425 426 /* get anon user mapping, read+write+execute */ 427 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 428 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 429 VM_PROT_ALL, VM_PROT_ALL, 0); 430 if (error) 431 goto cleanup; 432 433 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 434 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 435 td->td_ucred, NOCRED, &aresid, td); 436 if (error != 0) 437 goto cleanup; 438 if (aresid != 0) { 439 error = ENOEXEC; 440 goto cleanup; 441 } 442 } else { 443#ifdef DEBUG 444 printf("uselib: Page aligned binary %lu\n", file_offset); 445#endif 446 /* 447 * for QMAGIC, a_entry is 20 bytes beyond the load address 448 * to skip the executable header 449 */ 450 vmaddr = trunc_page(a_out->a_entry); 451 452 /* 453 * Map it all into the process's space as a single 454 * copy-on-write "data" segment. 455 */ 456 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 457 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 458 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 459 if (error) 460 goto cleanup; 461 } 462#ifdef DEBUG 463 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 464 ((long *)vmaddr)[1]); 465#endif 466 if (bss_size != 0) { 467 /* Calculate BSS start address */ 468 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 469 a_out->a_data; 470 471 /* allocate some 'anon' space */ 472 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 473 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 474 VM_PROT_ALL, 0); 475 if (error) 476 goto cleanup; 477 } 478 479cleanup: 480 /* Unlock vnode if needed */ 481 if (locked) 482 VOP_UNLOCK(vp, 0); 483 484 /* Release the temporary mapping. */ 485 if (a_out) 486 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 487 488 return (error); 489} 490 491#endif /* __i386__ */ 492 493int 494linux_select(struct thread *td, struct linux_select_args *args) 495{ 496 l_timeval ltv; 497 struct timeval tv0, tv1, utv, *tvp; 498 int error; 499 500#ifdef DEBUG 501 if (ldebug(select)) 502 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 503 (void *)args->readfds, (void *)args->writefds, 504 (void *)args->exceptfds, (void *)args->timeout); 505#endif 506 507 /* 508 * Store current time for computation of the amount of 509 * time left. 510 */ 511 if (args->timeout) { 512 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 513 goto select_out; 514 utv.tv_sec = ltv.tv_sec; 515 utv.tv_usec = ltv.tv_usec; 516#ifdef DEBUG 517 if (ldebug(select)) 518 printf(LMSG("incoming timeout (%jd/%ld)"), 519 (intmax_t)utv.tv_sec, utv.tv_usec); 520#endif 521 522 if (itimerfix(&utv)) { 523 /* 524 * The timeval was invalid. Convert it to something 525 * valid that will act as it does under Linux. 526 */ 527 utv.tv_sec += utv.tv_usec / 1000000; 528 utv.tv_usec %= 1000000; 529 if (utv.tv_usec < 0) { 530 utv.tv_sec -= 1; 531 utv.tv_usec += 1000000; 532 } 533 if (utv.tv_sec < 0) 534 timevalclear(&utv); 535 } 536 microtime(&tv0); 537 tvp = &utv; 538 } else 539 tvp = NULL; 540 541 error = kern_select(td, args->nfds, args->readfds, args->writefds, 542 args->exceptfds, tvp, LINUX_NFDBITS); 543 544#ifdef DEBUG 545 if (ldebug(select)) 546 printf(LMSG("real select returns %d"), error); 547#endif 548 if (error) 549 goto select_out; 550 551 if (args->timeout) { 552 if (td->td_retval[0]) { 553 /* 554 * Compute how much time was left of the timeout, 555 * by subtracting the current time and the time 556 * before we started the call, and subtracting 557 * that result from the user-supplied value. 558 */ 559 microtime(&tv1); 560 timevalsub(&tv1, &tv0); 561 timevalsub(&utv, &tv1); 562 if (utv.tv_sec < 0) 563 timevalclear(&utv); 564 } else 565 timevalclear(&utv); 566#ifdef DEBUG 567 if (ldebug(select)) 568 printf(LMSG("outgoing timeout (%jd/%ld)"), 569 (intmax_t)utv.tv_sec, utv.tv_usec); 570#endif 571 ltv.tv_sec = utv.tv_sec; 572 ltv.tv_usec = utv.tv_usec; 573 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 574 goto select_out; 575 } 576 577select_out: 578#ifdef DEBUG 579 if (ldebug(select)) 580 printf(LMSG("select_out -> %d"), error); 581#endif 582 return (error); 583} 584 585int 586linux_mremap(struct thread *td, struct linux_mremap_args *args) 587{ 588 struct munmap_args /* { 589 void *addr; 590 size_t len; 591 } */ bsd_args; 592 int error = 0; 593 594#ifdef DEBUG 595 if (ldebug(mremap)) 596 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 597 (void *)(uintptr_t)args->addr, 598 (unsigned long)args->old_len, 599 (unsigned long)args->new_len, 600 (unsigned long)args->flags); 601#endif 602 603 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 604 td->td_retval[0] = 0; 605 return (EINVAL); 606 } 607 608 /* 609 * Check for the page alignment. 610 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 611 */ 612 if (args->addr & PAGE_MASK) { 613 td->td_retval[0] = 0; 614 return (EINVAL); 615 } 616 617 args->new_len = round_page(args->new_len); 618 args->old_len = round_page(args->old_len); 619 620 if (args->new_len > args->old_len) { 621 td->td_retval[0] = 0; 622 return (ENOMEM); 623 } 624 625 if (args->new_len < args->old_len) { 626 bsd_args.addr = 627 (caddr_t)((uintptr_t)args->addr + args->new_len); 628 bsd_args.len = args->old_len - args->new_len; 629 error = sys_munmap(td, &bsd_args); 630 } 631 632 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 633 return (error); 634} 635 636#define LINUX_MS_ASYNC 0x0001 637#define LINUX_MS_INVALIDATE 0x0002 638#define LINUX_MS_SYNC 0x0004 639 640int 641linux_msync(struct thread *td, struct linux_msync_args *args) 642{ 643 struct msync_args bsd_args; 644 645 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 646 bsd_args.len = (uintptr_t)args->len; 647 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 648 649 return (sys_msync(td, &bsd_args)); 650} 651 652int 653linux_time(struct thread *td, struct linux_time_args *args) 654{ 655 struct timeval tv; 656 l_time_t tm; 657 int error; 658 659#ifdef DEBUG 660 if (ldebug(time)) 661 printf(ARGS(time, "*")); 662#endif 663 664 microtime(&tv); 665 tm = tv.tv_sec; 666 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 667 return (error); 668 td->td_retval[0] = tm; 669 return (0); 670} 671 672struct l_times_argv { 673 l_clock_t tms_utime; 674 l_clock_t tms_stime; 675 l_clock_t tms_cutime; 676 l_clock_t tms_cstime; 677}; 678 679 680/* 681 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 682 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 683 * auxiliary vector entry. 684 */ 685#define CLK_TCK 100 686 687#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 688#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 689 690#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 691 CONVNTCK(r) : CONVOTCK(r)) 692 693int 694linux_times(struct thread *td, struct linux_times_args *args) 695{ 696 struct timeval tv, utime, stime, cutime, cstime; 697 struct l_times_argv tms; 698 struct proc *p; 699 int error; 700 701#ifdef DEBUG 702 if (ldebug(times)) 703 printf(ARGS(times, "*")); 704#endif 705 706 if (args->buf != NULL) { 707 p = td->td_proc; 708 PROC_LOCK(p); 709 PROC_STATLOCK(p); 710 calcru(p, &utime, &stime); 711 PROC_STATUNLOCK(p); 712 calccru(p, &cutime, &cstime); 713 PROC_UNLOCK(p); 714 715 tms.tms_utime = CONVTCK(utime); 716 tms.tms_stime = CONVTCK(stime); 717 718 tms.tms_cutime = CONVTCK(cutime); 719 tms.tms_cstime = CONVTCK(cstime); 720 721 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 722 return (error); 723 } 724 725 microuptime(&tv); 726 td->td_retval[0] = (int)CONVTCK(tv); 727 return (0); 728} 729 730int 731linux_newuname(struct thread *td, struct linux_newuname_args *args) 732{ 733 struct l_new_utsname utsname; 734 char osname[LINUX_MAX_UTSNAME]; 735 char osrelease[LINUX_MAX_UTSNAME]; 736 char *p; 737 738#ifdef DEBUG 739 if (ldebug(newuname)) 740 printf(ARGS(newuname, "*")); 741#endif 742 743 linux_get_osname(td, osname); 744 linux_get_osrelease(td, osrelease); 745 746 bzero(&utsname, sizeof(utsname)); 747 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 748 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 749 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 750 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 751 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 752 for (p = utsname.version; *p != '\0'; ++p) 753 if (*p == '\n') { 754 *p = '\0'; 755 break; 756 } 757 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 758 759 return (copyout(&utsname, args->buf, sizeof(utsname))); 760} 761 762struct l_utimbuf { 763 l_time_t l_actime; 764 l_time_t l_modtime; 765}; 766 767int 768linux_utime(struct thread *td, struct linux_utime_args *args) 769{ 770 struct timeval tv[2], *tvp; 771 struct l_utimbuf lut; 772 char *fname; 773 int error; 774 775 LCONVPATHEXIST(td, args->fname, &fname); 776 777#ifdef DEBUG 778 if (ldebug(utime)) 779 printf(ARGS(utime, "%s, *"), fname); 780#endif 781 782 if (args->times) { 783 if ((error = copyin(args->times, &lut, sizeof lut))) { 784 LFREEPATH(fname); 785 return (error); 786 } 787 tv[0].tv_sec = lut.l_actime; 788 tv[0].tv_usec = 0; 789 tv[1].tv_sec = lut.l_modtime; 790 tv[1].tv_usec = 0; 791 tvp = tv; 792 } else 793 tvp = NULL; 794 795 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 796 LFREEPATH(fname); 797 return (error); 798} 799 800int 801linux_utimes(struct thread *td, struct linux_utimes_args *args) 802{ 803 l_timeval ltv[2]; 804 struct timeval tv[2], *tvp = NULL; 805 char *fname; 806 int error; 807 808 LCONVPATHEXIST(td, args->fname, &fname); 809 810#ifdef DEBUG 811 if (ldebug(utimes)) 812 printf(ARGS(utimes, "%s, *"), fname); 813#endif 814 815 if (args->tptr != NULL) { 816 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 817 LFREEPATH(fname); 818 return (error); 819 } 820 tv[0].tv_sec = ltv[0].tv_sec; 821 tv[0].tv_usec = ltv[0].tv_usec; 822 tv[1].tv_sec = ltv[1].tv_sec; 823 tv[1].tv_usec = ltv[1].tv_usec; 824 tvp = tv; 825 } 826 827 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 828 LFREEPATH(fname); 829 return (error); 830} 831 832static int 833linux_utimensat_nsec_valid(l_long nsec) 834{ 835 836 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW) 837 return (0); 838 if (nsec >= 0 && nsec <= 999999999) 839 return (0); 840 return (1); 841} 842 843int 844linux_utimensat(struct thread *td, struct linux_utimensat_args *args) 845{ 846 struct l_timespec l_times[2]; 847 struct timespec times[2], *timesp = NULL; 848 char *path = NULL; 849 int error, dfd, flags = 0; 850 851 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 852 853#ifdef DEBUG 854 if (ldebug(utimensat)) 855 printf(ARGS(utimensat, "%d, *"), dfd); 856#endif 857 858 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 859 return (EINVAL); 860 861 if (args->times != NULL) { 862 error = copyin(args->times, l_times, sizeof(l_times)); 863 if (error != 0) 864 return (error); 865 866 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 || 867 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0) 868 return (EINVAL); 869 870 times[0].tv_sec = l_times[0].tv_sec; 871 switch (l_times[0].tv_nsec) 872 { 873 case LINUX_UTIME_OMIT: 874 times[0].tv_nsec = UTIME_OMIT; 875 break; 876 case LINUX_UTIME_NOW: 877 times[0].tv_nsec = UTIME_NOW; 878 break; 879 default: 880 times[0].tv_nsec = l_times[0].tv_nsec; 881 } 882 883 times[1].tv_sec = l_times[1].tv_sec; 884 switch (l_times[1].tv_nsec) 885 { 886 case LINUX_UTIME_OMIT: 887 times[1].tv_nsec = UTIME_OMIT; 888 break; 889 case LINUX_UTIME_NOW: 890 times[1].tv_nsec = UTIME_NOW; 891 break; 892 default: 893 times[1].tv_nsec = l_times[1].tv_nsec; 894 break; 895 } 896 timesp = times; 897 898 /* This breaks POSIX, but is what the Linux kernel does 899 * _on purpose_ (documented in the man page for utimensat(2)), 900 * so we must follow that behaviour. */ 901 if (times[0].tv_nsec == UTIME_OMIT && 902 times[1].tv_nsec == UTIME_OMIT) 903 return (0); 904 } 905 906 if (args->pathname != NULL) 907 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 908 else if (args->flags != 0) 909 return (EINVAL); 910 911 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 912 flags |= AT_SYMLINK_NOFOLLOW; 913 914 if (path == NULL) 915 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 916 else { 917 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 918 UIO_SYSSPACE, flags); 919 LFREEPATH(path); 920 } 921 922 return (error); 923} 924 925int 926linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 927{ 928 l_timeval ltv[2]; 929 struct timeval tv[2], *tvp = NULL; 930 char *fname; 931 int error, dfd; 932 933 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 934 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 935 936#ifdef DEBUG 937 if (ldebug(futimesat)) 938 printf(ARGS(futimesat, "%s, *"), fname); 939#endif 940 941 if (args->utimes != NULL) { 942 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 943 LFREEPATH(fname); 944 return (error); 945 } 946 tv[0].tv_sec = ltv[0].tv_sec; 947 tv[0].tv_usec = ltv[0].tv_usec; 948 tv[1].tv_sec = ltv[1].tv_sec; 949 tv[1].tv_usec = ltv[1].tv_usec; 950 tvp = tv; 951 } 952 953 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 954 LFREEPATH(fname); 955 return (error); 956} 957 958int 959linux_common_wait(struct thread *td, int pid, int *status, 960 int options, struct rusage *ru) 961{ 962 int error, tmpstat; 963 964 error = kern_wait(td, pid, &tmpstat, options, ru); 965 if (error) 966 return (error); 967 968 if (status) { 969 tmpstat &= 0xffff; 970 if (WIFSIGNALED(tmpstat)) 971 tmpstat = (tmpstat & 0xffffff80) | 972 bsd_to_linux_signal(WTERMSIG(tmpstat)); 973 else if (WIFSTOPPED(tmpstat)) 974 tmpstat = (tmpstat & 0xffff00ff) | 975 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 976 else if (WIFCONTINUED(tmpstat)) 977 tmpstat = 0xffff; 978 error = copyout(&tmpstat, status, sizeof(int)); 979 } 980 981 return (error); 982} 983 984#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 985int 986linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 987{ 988 struct linux_wait4_args wait4_args; 989 990#ifdef DEBUG 991 if (ldebug(waitpid)) 992 printf(ARGS(waitpid, "%d, %p, %d"), 993 args->pid, (void *)args->status, args->options); 994#endif 995 996 wait4_args.pid = args->pid; 997 wait4_args.status = args->status; 998 wait4_args.options = args->options; 999 wait4_args.rusage = NULL; 1000 1001 return (linux_wait4(td, &wait4_args)); 1002} 1003#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1004 1005int 1006linux_wait4(struct thread *td, struct linux_wait4_args *args) 1007{ 1008 int error, options; 1009 struct rusage ru, *rup; 1010 1011#ifdef DEBUG 1012 if (ldebug(wait4)) 1013 printf(ARGS(wait4, "%d, %p, %d, %p"), 1014 args->pid, (void *)args->status, args->options, 1015 (void *)args->rusage); 1016#endif 1017 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 1018 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 1019 return (EINVAL); 1020 1021 options = WEXITED; 1022 linux_to_bsd_waitopts(args->options, &options); 1023 1024 if (args->rusage != NULL) 1025 rup = &ru; 1026 else 1027 rup = NULL; 1028 error = linux_common_wait(td, args->pid, args->status, options, rup); 1029 if (error != 0) 1030 return (error); 1031 if (args->rusage != NULL) 1032 error = linux_copyout_rusage(&ru, args->rusage); 1033 return (error); 1034} 1035 1036int 1037linux_waitid(struct thread *td, struct linux_waitid_args *args) 1038{ 1039 int status, options, sig; 1040 struct __wrusage wru; 1041 siginfo_t siginfo; 1042 l_siginfo_t lsi; 1043 idtype_t idtype; 1044 struct proc *p; 1045 int error; 1046 1047 options = 0; 1048 linux_to_bsd_waitopts(args->options, &options); 1049 1050 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1051 return (EINVAL); 1052 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1053 return (EINVAL); 1054 1055 switch (args->idtype) { 1056 case LINUX_P_ALL: 1057 idtype = P_ALL; 1058 break; 1059 case LINUX_P_PID: 1060 if (args->id <= 0) 1061 return (EINVAL); 1062 idtype = P_PID; 1063 break; 1064 case LINUX_P_PGID: 1065 if (args->id <= 0) 1066 return (EINVAL); 1067 idtype = P_PGID; 1068 break; 1069 default: 1070 return (EINVAL); 1071 } 1072 1073 error = kern_wait6(td, idtype, args->id, &status, options, 1074 &wru, &siginfo); 1075 if (error != 0) 1076 return (error); 1077 if (args->rusage != NULL) { 1078 error = linux_copyout_rusage(&wru.wru_children, 1079 args->rusage); 1080 if (error != 0) 1081 return (error); 1082 } 1083 if (args->info != NULL) { 1084 p = td->td_proc; 1085 if (td->td_retval[0] == 0) 1086 bzero(&lsi, sizeof(lsi)); 1087 else { 1088 sig = bsd_to_linux_signal(siginfo.si_signo); 1089 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1090 } 1091 error = copyout(&lsi, args->info, sizeof(lsi)); 1092 } 1093 td->td_retval[0] = 0; 1094 1095 return (error); 1096} 1097 1098int 1099linux_mknod(struct thread *td, struct linux_mknod_args *args) 1100{ 1101 char *path; 1102 int error; 1103 1104 LCONVPATHCREAT(td, args->path, &path); 1105 1106#ifdef DEBUG 1107 if (ldebug(mknod)) 1108 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1109 (uintmax_t)args->dev); 1110#endif 1111 1112 switch (args->mode & S_IFMT) { 1113 case S_IFIFO: 1114 case S_IFSOCK: 1115 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1116 break; 1117 1118 case S_IFCHR: 1119 case S_IFBLK: 1120 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1121 args->dev); 1122 break; 1123 1124 case S_IFDIR: 1125 error = EPERM; 1126 break; 1127 1128 case 0: 1129 args->mode |= S_IFREG; 1130 /* FALLTHROUGH */ 1131 case S_IFREG: 1132 error = kern_open(td, path, UIO_SYSSPACE, 1133 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1134 if (error == 0) 1135 kern_close(td, td->td_retval[0]); 1136 break; 1137 1138 default: 1139 error = EINVAL; 1140 break; 1141 } 1142 LFREEPATH(path); 1143 return (error); 1144} 1145 1146int 1147linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1148{ 1149 char *path; 1150 int error, dfd; 1151 1152 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1153 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1154 1155#ifdef DEBUG 1156 if (ldebug(mknodat)) 1157 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1158#endif 1159 1160 switch (args->mode & S_IFMT) { 1161 case S_IFIFO: 1162 case S_IFSOCK: 1163 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1164 break; 1165 1166 case S_IFCHR: 1167 case S_IFBLK: 1168 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1169 args->dev); 1170 break; 1171 1172 case S_IFDIR: 1173 error = EPERM; 1174 break; 1175 1176 case 0: 1177 args->mode |= S_IFREG; 1178 /* FALLTHROUGH */ 1179 case S_IFREG: 1180 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1181 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1182 if (error == 0) 1183 kern_close(td, td->td_retval[0]); 1184 break; 1185 1186 default: 1187 error = EINVAL; 1188 break; 1189 } 1190 LFREEPATH(path); 1191 return (error); 1192} 1193 1194/* 1195 * UGH! This is just about the dumbest idea I've ever heard!! 1196 */ 1197int 1198linux_personality(struct thread *td, struct linux_personality_args *args) 1199{ 1200#ifdef DEBUG 1201 if (ldebug(personality)) 1202 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1203#endif 1204 if (args->per != 0) 1205 return (EINVAL); 1206 1207 /* Yes Jim, it's still a Linux... */ 1208 td->td_retval[0] = 0; 1209 return (0); 1210} 1211 1212struct l_itimerval { 1213 l_timeval it_interval; 1214 l_timeval it_value; 1215}; 1216 1217#define B2L_ITIMERVAL(bip, lip) \ 1218 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1219 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1220 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1221 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1222 1223int 1224linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1225{ 1226 int error; 1227 struct l_itimerval ls; 1228 struct itimerval aitv, oitv; 1229 1230#ifdef DEBUG 1231 if (ldebug(setitimer)) 1232 printf(ARGS(setitimer, "%p, %p"), 1233 (void *)uap->itv, (void *)uap->oitv); 1234#endif 1235 1236 if (uap->itv == NULL) { 1237 uap->itv = uap->oitv; 1238 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1239 } 1240 1241 error = copyin(uap->itv, &ls, sizeof(ls)); 1242 if (error != 0) 1243 return (error); 1244 B2L_ITIMERVAL(&aitv, &ls); 1245#ifdef DEBUG 1246 if (ldebug(setitimer)) { 1247 printf("setitimer: value: sec: %jd, usec: %ld\n", 1248 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1249 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1250 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1251 } 1252#endif 1253 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1254 if (error != 0 || uap->oitv == NULL) 1255 return (error); 1256 B2L_ITIMERVAL(&ls, &oitv); 1257 1258 return (copyout(&ls, uap->oitv, sizeof(ls))); 1259} 1260 1261int 1262linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1263{ 1264 int error; 1265 struct l_itimerval ls; 1266 struct itimerval aitv; 1267 1268#ifdef DEBUG 1269 if (ldebug(getitimer)) 1270 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1271#endif 1272 error = kern_getitimer(td, uap->which, &aitv); 1273 if (error != 0) 1274 return (error); 1275 B2L_ITIMERVAL(&ls, &aitv); 1276 return (copyout(&ls, uap->itv, sizeof(ls))); 1277} 1278 1279#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1280int 1281linux_nice(struct thread *td, struct linux_nice_args *args) 1282{ 1283 struct setpriority_args bsd_args; 1284 1285 bsd_args.which = PRIO_PROCESS; 1286 bsd_args.who = 0; /* current process */ 1287 bsd_args.prio = args->inc; 1288 return (sys_setpriority(td, &bsd_args)); 1289} 1290#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1291 1292int 1293linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1294{ 1295 struct ucred *newcred, *oldcred; 1296 l_gid_t *linux_gidset; 1297 gid_t *bsd_gidset; 1298 int ngrp, error; 1299 struct proc *p; 1300 1301 ngrp = args->gidsetsize; 1302 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1303 return (EINVAL); 1304 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1305 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1306 if (error) 1307 goto out; 1308 newcred = crget(); 1309 crextend(newcred, ngrp + 1); 1310 p = td->td_proc; 1311 PROC_LOCK(p); 1312 oldcred = p->p_ucred; 1313 crcopy(newcred, oldcred); 1314 1315 /* 1316 * cr_groups[0] holds egid. Setting the whole set from 1317 * the supplied set will cause egid to be changed too. 1318 * Keep cr_groups[0] unchanged to prevent that. 1319 */ 1320 1321 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1322 PROC_UNLOCK(p); 1323 crfree(newcred); 1324 goto out; 1325 } 1326 1327 if (ngrp > 0) { 1328 newcred->cr_ngroups = ngrp + 1; 1329 1330 bsd_gidset = newcred->cr_groups; 1331 ngrp--; 1332 while (ngrp >= 0) { 1333 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1334 ngrp--; 1335 } 1336 } else 1337 newcred->cr_ngroups = 1; 1338 1339 setsugid(p); 1340 p->p_ucred = newcred; 1341 PROC_UNLOCK(p); 1342 crfree(oldcred); 1343 error = 0; 1344out: 1345 free(linux_gidset, M_LINUX); 1346 return (error); 1347} 1348 1349int 1350linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1351{ 1352 struct ucred *cred; 1353 l_gid_t *linux_gidset; 1354 gid_t *bsd_gidset; 1355 int bsd_gidsetsz, ngrp, error; 1356 1357 cred = td->td_ucred; 1358 bsd_gidset = cred->cr_groups; 1359 bsd_gidsetsz = cred->cr_ngroups - 1; 1360 1361 /* 1362 * cr_groups[0] holds egid. Returning the whole set 1363 * here will cause a duplicate. Exclude cr_groups[0] 1364 * to prevent that. 1365 */ 1366 1367 if ((ngrp = args->gidsetsize) == 0) { 1368 td->td_retval[0] = bsd_gidsetsz; 1369 return (0); 1370 } 1371 1372 if (ngrp < bsd_gidsetsz) 1373 return (EINVAL); 1374 1375 ngrp = 0; 1376 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1377 M_LINUX, M_WAITOK); 1378 while (ngrp < bsd_gidsetsz) { 1379 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1380 ngrp++; 1381 } 1382 1383 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1384 free(linux_gidset, M_LINUX); 1385 if (error) 1386 return (error); 1387 1388 td->td_retval[0] = ngrp; 1389 return (0); 1390} 1391 1392int 1393linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1394{ 1395 struct rlimit bsd_rlim; 1396 struct l_rlimit rlim; 1397 u_int which; 1398 int error; 1399 1400#ifdef DEBUG 1401 if (ldebug(setrlimit)) 1402 printf(ARGS(setrlimit, "%d, %p"), 1403 args->resource, (void *)args->rlim); 1404#endif 1405 1406 if (args->resource >= LINUX_RLIM_NLIMITS) 1407 return (EINVAL); 1408 1409 which = linux_to_bsd_resource[args->resource]; 1410 if (which == -1) 1411 return (EINVAL); 1412 1413 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1414 if (error) 1415 return (error); 1416 1417 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1418 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1419 return (kern_setrlimit(td, which, &bsd_rlim)); 1420} 1421 1422#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1423int 1424linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1425{ 1426 struct l_rlimit rlim; 1427 struct proc *p = td->td_proc; 1428 struct rlimit bsd_rlim; 1429 u_int which; 1430 1431#ifdef DEBUG 1432 if (ldebug(old_getrlimit)) 1433 printf(ARGS(old_getrlimit, "%d, %p"), 1434 args->resource, (void *)args->rlim); 1435#endif 1436 1437 if (args->resource >= LINUX_RLIM_NLIMITS) 1438 return (EINVAL); 1439 1440 which = linux_to_bsd_resource[args->resource]; 1441 if (which == -1) 1442 return (EINVAL); 1443 1444 PROC_LOCK(p); 1445 lim_rlimit(p, which, &bsd_rlim); 1446 PROC_UNLOCK(p); 1447 1448#ifdef COMPAT_LINUX32 1449 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1450 if (rlim.rlim_cur == UINT_MAX) 1451 rlim.rlim_cur = INT_MAX; 1452 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1453 if (rlim.rlim_max == UINT_MAX) 1454 rlim.rlim_max = INT_MAX; 1455#else 1456 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1457 if (rlim.rlim_cur == ULONG_MAX) 1458 rlim.rlim_cur = LONG_MAX; 1459 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1460 if (rlim.rlim_max == ULONG_MAX) 1461 rlim.rlim_max = LONG_MAX; 1462#endif 1463 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1464} 1465#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1466 1467int 1468linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1469{ 1470 struct l_rlimit rlim; 1471 struct proc *p = td->td_proc; 1472 struct rlimit bsd_rlim; 1473 u_int which; 1474 1475#ifdef DEBUG 1476 if (ldebug(getrlimit)) 1477 printf(ARGS(getrlimit, "%d, %p"), 1478 args->resource, (void *)args->rlim); 1479#endif 1480 1481 if (args->resource >= LINUX_RLIM_NLIMITS) 1482 return (EINVAL); 1483 1484 which = linux_to_bsd_resource[args->resource]; 1485 if (which == -1) 1486 return (EINVAL); 1487 1488 PROC_LOCK(p); 1489 lim_rlimit(p, which, &bsd_rlim); 1490 PROC_UNLOCK(p); 1491 1492 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1493 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1494 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1495} 1496 1497int 1498linux_sched_setscheduler(struct thread *td, 1499 struct linux_sched_setscheduler_args *args) 1500{ 1501 struct sched_param sched_param; 1502 struct thread *tdt; 1503 int error, policy; 1504 1505#ifdef DEBUG 1506 if (ldebug(sched_setscheduler)) 1507 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1508 args->pid, args->policy, (const void *)args->param); 1509#endif 1510 1511 switch (args->policy) { 1512 case LINUX_SCHED_OTHER: 1513 policy = SCHED_OTHER; 1514 break; 1515 case LINUX_SCHED_FIFO: 1516 policy = SCHED_FIFO; 1517 break; 1518 case LINUX_SCHED_RR: 1519 policy = SCHED_RR; 1520 break; 1521 default: 1522 return (EINVAL); 1523 } 1524 1525 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1526 if (error) 1527 return (error); 1528 1529 tdt = linux_tdfind(td, args->pid, -1); 1530 if (tdt == NULL) 1531 return (ESRCH); 1532 1533 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1534 PROC_UNLOCK(tdt->td_proc); 1535 return (error); 1536} 1537 1538int 1539linux_sched_getscheduler(struct thread *td, 1540 struct linux_sched_getscheduler_args *args) 1541{ 1542 struct thread *tdt; 1543 int error, policy; 1544 1545#ifdef DEBUG 1546 if (ldebug(sched_getscheduler)) 1547 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1548#endif 1549 1550 tdt = linux_tdfind(td, args->pid, -1); 1551 if (tdt == NULL) 1552 return (ESRCH); 1553 1554 error = kern_sched_getscheduler(td, tdt, &policy); 1555 PROC_UNLOCK(tdt->td_proc); 1556 1557 switch (policy) { 1558 case SCHED_OTHER: 1559 td->td_retval[0] = LINUX_SCHED_OTHER; 1560 break; 1561 case SCHED_FIFO: 1562 td->td_retval[0] = LINUX_SCHED_FIFO; 1563 break; 1564 case SCHED_RR: 1565 td->td_retval[0] = LINUX_SCHED_RR; 1566 break; 1567 } 1568 return (error); 1569} 1570 1571int 1572linux_sched_get_priority_max(struct thread *td, 1573 struct linux_sched_get_priority_max_args *args) 1574{ 1575 struct sched_get_priority_max_args bsd; 1576 1577#ifdef DEBUG 1578 if (ldebug(sched_get_priority_max)) 1579 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1580#endif 1581 1582 switch (args->policy) { 1583 case LINUX_SCHED_OTHER: 1584 bsd.policy = SCHED_OTHER; 1585 break; 1586 case LINUX_SCHED_FIFO: 1587 bsd.policy = SCHED_FIFO; 1588 break; 1589 case LINUX_SCHED_RR: 1590 bsd.policy = SCHED_RR; 1591 break; 1592 default: 1593 return (EINVAL); 1594 } 1595 return (sys_sched_get_priority_max(td, &bsd)); 1596} 1597 1598int 1599linux_sched_get_priority_min(struct thread *td, 1600 struct linux_sched_get_priority_min_args *args) 1601{ 1602 struct sched_get_priority_min_args bsd; 1603 1604#ifdef DEBUG 1605 if (ldebug(sched_get_priority_min)) 1606 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1607#endif 1608 1609 switch (args->policy) { 1610 case LINUX_SCHED_OTHER: 1611 bsd.policy = SCHED_OTHER; 1612 break; 1613 case LINUX_SCHED_FIFO: 1614 bsd.policy = SCHED_FIFO; 1615 break; 1616 case LINUX_SCHED_RR: 1617 bsd.policy = SCHED_RR; 1618 break; 1619 default: 1620 return (EINVAL); 1621 } 1622 return (sys_sched_get_priority_min(td, &bsd)); 1623} 1624 1625#define REBOOT_CAD_ON 0x89abcdef 1626#define REBOOT_CAD_OFF 0 1627#define REBOOT_HALT 0xcdef0123 1628#define REBOOT_RESTART 0x01234567 1629#define REBOOT_RESTART2 0xA1B2C3D4 1630#define REBOOT_POWEROFF 0x4321FEDC 1631#define REBOOT_MAGIC1 0xfee1dead 1632#define REBOOT_MAGIC2 0x28121969 1633#define REBOOT_MAGIC2A 0x05121996 1634#define REBOOT_MAGIC2B 0x16041998 1635 1636int 1637linux_reboot(struct thread *td, struct linux_reboot_args *args) 1638{ 1639 struct reboot_args bsd_args; 1640 1641#ifdef DEBUG 1642 if (ldebug(reboot)) 1643 printf(ARGS(reboot, "0x%x"), args->cmd); 1644#endif 1645 1646 if (args->magic1 != REBOOT_MAGIC1) 1647 return (EINVAL); 1648 1649 switch (args->magic2) { 1650 case REBOOT_MAGIC2: 1651 case REBOOT_MAGIC2A: 1652 case REBOOT_MAGIC2B: 1653 break; 1654 default: 1655 return (EINVAL); 1656 } 1657 1658 switch (args->cmd) { 1659 case REBOOT_CAD_ON: 1660 case REBOOT_CAD_OFF: 1661 return (priv_check(td, PRIV_REBOOT)); 1662 case REBOOT_HALT: 1663 bsd_args.opt = RB_HALT; 1664 break; 1665 case REBOOT_RESTART: 1666 case REBOOT_RESTART2: 1667 bsd_args.opt = 0; 1668 break; 1669 case REBOOT_POWEROFF: 1670 bsd_args.opt = RB_POWEROFF; 1671 break; 1672 default: 1673 return (EINVAL); 1674 } 1675 return (sys_reboot(td, &bsd_args)); 1676} 1677 1678 1679/* 1680 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1681 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1682 * are assumed to be preserved. The following lightweight syscalls fixes 1683 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1684 * 1685 * linux_getpid() - MP SAFE 1686 * linux_getgid() - MP SAFE 1687 * linux_getuid() - MP SAFE 1688 */ 1689 1690int 1691linux_getpid(struct thread *td, struct linux_getpid_args *args) 1692{ 1693 1694#ifdef DEBUG 1695 if (ldebug(getpid)) 1696 printf(ARGS(getpid, "")); 1697#endif 1698 td->td_retval[0] = td->td_proc->p_pid; 1699 1700 return (0); 1701} 1702 1703int 1704linux_gettid(struct thread *td, struct linux_gettid_args *args) 1705{ 1706 struct linux_emuldata *em; 1707 1708#ifdef DEBUG 1709 if (ldebug(gettid)) 1710 printf(ARGS(gettid, "")); 1711#endif 1712 1713 em = em_find(td); 1714 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1715 1716 td->td_retval[0] = em->em_tid; 1717 1718 return (0); 1719} 1720 1721 1722int 1723linux_getppid(struct thread *td, struct linux_getppid_args *args) 1724{ 1725 1726#ifdef DEBUG 1727 if (ldebug(getppid)) 1728 printf(ARGS(getppid, "")); 1729#endif 1730 1731 PROC_LOCK(td->td_proc); 1732 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1733 PROC_UNLOCK(td->td_proc); 1734 return (0); 1735} 1736 1737int 1738linux_getgid(struct thread *td, struct linux_getgid_args *args) 1739{ 1740 1741#ifdef DEBUG 1742 if (ldebug(getgid)) 1743 printf(ARGS(getgid, "")); 1744#endif 1745 1746 td->td_retval[0] = td->td_ucred->cr_rgid; 1747 return (0); 1748} 1749 1750int 1751linux_getuid(struct thread *td, struct linux_getuid_args *args) 1752{ 1753 1754#ifdef DEBUG 1755 if (ldebug(getuid)) 1756 printf(ARGS(getuid, "")); 1757#endif 1758 1759 td->td_retval[0] = td->td_ucred->cr_ruid; 1760 return (0); 1761} 1762 1763 1764int 1765linux_getsid(struct thread *td, struct linux_getsid_args *args) 1766{ 1767 struct getsid_args bsd; 1768 1769#ifdef DEBUG 1770 if (ldebug(getsid)) 1771 printf(ARGS(getsid, "%i"), args->pid); 1772#endif 1773 1774 bsd.pid = args->pid; 1775 return (sys_getsid(td, &bsd)); 1776} 1777 1778int 1779linux_nosys(struct thread *td, struct nosys_args *ignore) 1780{ 1781 1782 return (ENOSYS); 1783} 1784 1785int 1786linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1787{ 1788 struct getpriority_args bsd_args; 1789 int error; 1790 1791#ifdef DEBUG 1792 if (ldebug(getpriority)) 1793 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1794#endif 1795 1796 bsd_args.which = args->which; 1797 bsd_args.who = args->who; 1798 error = sys_getpriority(td, &bsd_args); 1799 td->td_retval[0] = 20 - td->td_retval[0]; 1800 return (error); 1801} 1802 1803int 1804linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1805{ 1806 int name[2]; 1807 1808#ifdef DEBUG 1809 if (ldebug(sethostname)) 1810 printf(ARGS(sethostname, "*, %i"), args->len); 1811#endif 1812 1813 name[0] = CTL_KERN; 1814 name[1] = KERN_HOSTNAME; 1815 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1816 args->len, 0, 0)); 1817} 1818 1819int 1820linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1821{ 1822 int name[2]; 1823 1824#ifdef DEBUG 1825 if (ldebug(setdomainname)) 1826 printf(ARGS(setdomainname, "*, %i"), args->len); 1827#endif 1828 1829 name[0] = CTL_KERN; 1830 name[1] = KERN_NISDOMAINNAME; 1831 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1832 args->len, 0, 0)); 1833} 1834 1835int 1836linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1837{ 1838 1839#ifdef DEBUG 1840 if (ldebug(exit_group)) 1841 printf(ARGS(exit_group, "%i"), args->error_code); 1842#endif 1843 1844 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1845 args->error_code); 1846 1847 /* 1848 * XXX: we should send a signal to the parent if 1849 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1850 * as it doesnt occur often. 1851 */ 1852 exit1(td, W_EXITCODE(args->error_code, 0)); 1853 /* NOTREACHED */ 1854} 1855 1856#define _LINUX_CAPABILITY_VERSION 0x19980330 1857 1858struct l_user_cap_header { 1859 l_int version; 1860 l_int pid; 1861}; 1862 1863struct l_user_cap_data { 1864 l_int effective; 1865 l_int permitted; 1866 l_int inheritable; 1867}; 1868 1869int 1870linux_capget(struct thread *td, struct linux_capget_args *args) 1871{ 1872 struct l_user_cap_header luch; 1873 struct l_user_cap_data lucd; 1874 int error; 1875 1876 if (args->hdrp == NULL) 1877 return (EFAULT); 1878 1879 error = copyin(args->hdrp, &luch, sizeof(luch)); 1880 if (error != 0) 1881 return (error); 1882 1883 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1884 luch.version = _LINUX_CAPABILITY_VERSION; 1885 error = copyout(&luch, args->hdrp, sizeof(luch)); 1886 if (error) 1887 return (error); 1888 return (EINVAL); 1889 } 1890 1891 if (luch.pid) 1892 return (EPERM); 1893 1894 if (args->datap) { 1895 /* 1896 * The current implementation doesn't support setting 1897 * a capability (it's essentially a stub) so indicate 1898 * that no capabilities are currently set or available 1899 * to request. 1900 */ 1901 bzero (&lucd, sizeof(lucd)); 1902 error = copyout(&lucd, args->datap, sizeof(lucd)); 1903 } 1904 1905 return (error); 1906} 1907 1908int 1909linux_capset(struct thread *td, struct linux_capset_args *args) 1910{ 1911 struct l_user_cap_header luch; 1912 struct l_user_cap_data lucd; 1913 int error; 1914 1915 if (args->hdrp == NULL || args->datap == NULL) 1916 return (EFAULT); 1917 1918 error = copyin(args->hdrp, &luch, sizeof(luch)); 1919 if (error != 0) 1920 return (error); 1921 1922 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1923 luch.version = _LINUX_CAPABILITY_VERSION; 1924 error = copyout(&luch, args->hdrp, sizeof(luch)); 1925 if (error) 1926 return (error); 1927 return (EINVAL); 1928 } 1929 1930 if (luch.pid) 1931 return (EPERM); 1932 1933 error = copyin(args->datap, &lucd, sizeof(lucd)); 1934 if (error != 0) 1935 return (error); 1936 1937 /* We currently don't support setting any capabilities. */ 1938 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1939 linux_msg(td, 1940 "capset effective=0x%x, permitted=0x%x, " 1941 "inheritable=0x%x is not implemented", 1942 (int)lucd.effective, (int)lucd.permitted, 1943 (int)lucd.inheritable); 1944 return (EPERM); 1945 } 1946 1947 return (0); 1948} 1949 1950int 1951linux_prctl(struct thread *td, struct linux_prctl_args *args) 1952{ 1953 int error = 0, max_size; 1954 struct proc *p = td->td_proc; 1955 char comm[LINUX_MAX_COMM_LEN]; 1956 struct linux_emuldata *em; 1957 int pdeath_signal; 1958 1959#ifdef DEBUG 1960 if (ldebug(prctl)) 1961 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1962 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1963 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1964#endif 1965 1966 switch (args->option) { 1967 case LINUX_PR_SET_PDEATHSIG: 1968 if (!LINUX_SIG_VALID(args->arg2)) 1969 return (EINVAL); 1970 em = em_find(td); 1971 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1972 em->pdeath_signal = args->arg2; 1973 break; 1974 case LINUX_PR_GET_PDEATHSIG: 1975 em = em_find(td); 1976 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1977 pdeath_signal = em->pdeath_signal; 1978 error = copyout(&pdeath_signal, 1979 (void *)(register_t)args->arg2, 1980 sizeof(pdeath_signal)); 1981 break; 1982 case LINUX_PR_GET_KEEPCAPS: 1983 /* 1984 * Indicate that we always clear the effective and 1985 * permitted capability sets when the user id becomes 1986 * non-zero (actually the capability sets are simply 1987 * always zero in the current implementation). 1988 */ 1989 td->td_retval[0] = 0; 1990 break; 1991 case LINUX_PR_SET_KEEPCAPS: 1992 /* 1993 * Ignore requests to keep the effective and permitted 1994 * capability sets when the user id becomes non-zero. 1995 */ 1996 break; 1997 case LINUX_PR_SET_NAME: 1998 /* 1999 * To be on the safe side we need to make sure to not 2000 * overflow the size a linux program expects. We already 2001 * do this here in the copyin, so that we don't need to 2002 * check on copyout. 2003 */ 2004 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 2005 error = copyinstr((void *)(register_t)args->arg2, comm, 2006 max_size, NULL); 2007 2008 /* Linux silently truncates the name if it is too long. */ 2009 if (error == ENAMETOOLONG) { 2010 /* 2011 * XXX: copyinstr() isn't documented to populate the 2012 * array completely, so do a copyin() to be on the 2013 * safe side. This should be changed in case 2014 * copyinstr() is changed to guarantee this. 2015 */ 2016 error = copyin((void *)(register_t)args->arg2, comm, 2017 max_size - 1); 2018 comm[max_size - 1] = '\0'; 2019 } 2020 if (error) 2021 return (error); 2022 2023 PROC_LOCK(p); 2024 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 2025 PROC_UNLOCK(p); 2026 break; 2027 case LINUX_PR_GET_NAME: 2028 PROC_LOCK(p); 2029 strlcpy(comm, p->p_comm, sizeof(comm)); 2030 PROC_UNLOCK(p); 2031 error = copyout(comm, (void *)(register_t)args->arg2, 2032 strlen(comm) + 1); 2033 break; 2034 default: 2035 error = EINVAL; 2036 break; 2037 } 2038 2039 return (error); 2040} 2041 2042int 2043linux_sched_setparam(struct thread *td, 2044 struct linux_sched_setparam_args *uap) 2045{ 2046 struct sched_param sched_param; 2047 struct thread *tdt; 2048 int error; 2049 2050#ifdef DEBUG 2051 if (ldebug(sched_setparam)) 2052 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 2053#endif 2054 2055 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 2056 if (error) 2057 return (error); 2058 2059 tdt = linux_tdfind(td, uap->pid, -1); 2060 if (tdt == NULL) 2061 return (ESRCH); 2062 2063 error = kern_sched_setparam(td, tdt, &sched_param); 2064 PROC_UNLOCK(tdt->td_proc); 2065 return (error); 2066} 2067 2068int 2069linux_sched_getparam(struct thread *td, 2070 struct linux_sched_getparam_args *uap) 2071{ 2072 struct sched_param sched_param; 2073 struct thread *tdt; 2074 int error; 2075 2076#ifdef DEBUG 2077 if (ldebug(sched_getparam)) 2078 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 2079#endif 2080 2081 tdt = linux_tdfind(td, uap->pid, -1); 2082 if (tdt == NULL) 2083 return (ESRCH); 2084 2085 error = kern_sched_getparam(td, tdt, &sched_param); 2086 PROC_UNLOCK(tdt->td_proc); 2087 if (error == 0) 2088 error = copyout(&sched_param, uap->param, 2089 sizeof(sched_param)); 2090 return (error); 2091} 2092 2093/* 2094 * Get affinity of a process. 2095 */ 2096int 2097linux_sched_getaffinity(struct thread *td, 2098 struct linux_sched_getaffinity_args *args) 2099{ 2100 int error; 2101 struct thread *tdt; 2102 struct cpuset_getaffinity_args cga; 2103 2104#ifdef DEBUG 2105 if (ldebug(sched_getaffinity)) 2106 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2107 args->len); 2108#endif 2109 if (args->len < sizeof(cpuset_t)) 2110 return (EINVAL); 2111 2112 tdt = linux_tdfind(td, args->pid, -1); 2113 if (tdt == NULL) 2114 return (ESRCH); 2115 2116 PROC_UNLOCK(tdt->td_proc); 2117 cga.level = CPU_LEVEL_WHICH; 2118 cga.which = CPU_WHICH_TID; 2119 cga.id = tdt->td_tid; 2120 cga.cpusetsize = sizeof(cpuset_t); 2121 cga.mask = (cpuset_t *) args->user_mask_ptr; 2122 2123 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2124 td->td_retval[0] = sizeof(cpuset_t); 2125 2126 return (error); 2127} 2128 2129/* 2130 * Set affinity of a process. 2131 */ 2132int 2133linux_sched_setaffinity(struct thread *td, 2134 struct linux_sched_setaffinity_args *args) 2135{ 2136 struct cpuset_setaffinity_args csa; 2137 struct thread *tdt; 2138 2139#ifdef DEBUG 2140 if (ldebug(sched_setaffinity)) 2141 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2142 args->len); 2143#endif 2144 if (args->len < sizeof(cpuset_t)) 2145 return (EINVAL); 2146 2147 tdt = linux_tdfind(td, args->pid, -1); 2148 if (tdt == NULL) 2149 return (ESRCH); 2150 2151 PROC_UNLOCK(tdt->td_proc); 2152 csa.level = CPU_LEVEL_WHICH; 2153 csa.which = CPU_WHICH_TID; 2154 csa.id = tdt->td_tid; 2155 csa.cpusetsize = sizeof(cpuset_t); 2156 csa.mask = (cpuset_t *) args->user_mask_ptr; 2157 2158 return (sys_cpuset_setaffinity(td, &csa)); 2159} 2160 2161struct linux_rlimit64 { 2162 uint64_t rlim_cur; 2163 uint64_t rlim_max; 2164}; 2165 2166int 2167linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2168{ 2169 struct rlimit rlim, nrlim; 2170 struct linux_rlimit64 lrlim; 2171 struct proc *p; 2172 u_int which; 2173 int flags; 2174 int error; 2175 2176#ifdef DEBUG 2177 if (ldebug(prlimit64)) 2178 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2179 args->resource, (void *)args->new, (void *)args->old); 2180#endif 2181 2182 if (args->resource >= LINUX_RLIM_NLIMITS) 2183 return (EINVAL); 2184 2185 which = linux_to_bsd_resource[args->resource]; 2186 if (which == -1) 2187 return (EINVAL); 2188 2189 if (args->new != NULL) { 2190 /* 2191 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2192 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2193 * as INFINITY so we do not need a conversion even. 2194 */ 2195 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2196 if (error != 0) 2197 return (error); 2198 } 2199 2200 flags = PGET_HOLD | PGET_NOTWEXIT; 2201 if (args->new != NULL) 2202 flags |= PGET_CANDEBUG; 2203 else 2204 flags |= PGET_CANSEE; 2205 error = pget(args->pid, flags, &p); 2206 if (error != 0) 2207 return (error); 2208 2209 if (args->old != NULL) { 2210 PROC_LOCK(p); 2211 lim_rlimit(p, which, &rlim); 2212 PROC_UNLOCK(p); 2213 if (rlim.rlim_cur == RLIM_INFINITY) 2214 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2215 else 2216 lrlim.rlim_cur = rlim.rlim_cur; 2217 if (rlim.rlim_max == RLIM_INFINITY) 2218 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2219 else 2220 lrlim.rlim_max = rlim.rlim_max; 2221 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2222 if (error != 0) 2223 goto out; 2224 } 2225 2226 if (args->new != NULL) 2227 error = kern_proc_setrlimit(td, p, which, &nrlim); 2228 2229 out: 2230 PRELE(p); 2231 return (error); 2232} 2233 2234int 2235linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2236{ 2237 struct timeval utv, tv0, tv1, *tvp; 2238 struct l_pselect6arg lpse6; 2239 struct l_timespec lts; 2240 struct timespec uts; 2241 l_sigset_t l_ss; 2242 sigset_t *ssp; 2243 sigset_t ss; 2244 int error; 2245 2246 ssp = NULL; 2247 if (args->sig != NULL) { 2248 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2249 if (error != 0) 2250 return (error); 2251 if (lpse6.ss_len != sizeof(l_ss)) 2252 return (EINVAL); 2253 if (lpse6.ss != 0) { 2254 error = copyin(PTRIN(lpse6.ss), &l_ss, 2255 sizeof(l_ss)); 2256 if (error != 0) 2257 return (error); 2258 linux_to_bsd_sigset(&l_ss, &ss); 2259 ssp = &ss; 2260 } 2261 } 2262 2263 /* 2264 * Currently glibc changes nanosecond number to microsecond. 2265 * This mean losing precision but for now it is hardly seen. 2266 */ 2267 if (args->tsp != NULL) { 2268 error = copyin(args->tsp, <s, sizeof(lts)); 2269 if (error != 0) 2270 return (error); 2271 error = linux_to_native_timespec(&uts, <s); 2272 if (error != 0) 2273 return (error); 2274 2275 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2276 if (itimerfix(&utv)) 2277 return (EINVAL); 2278 2279 microtime(&tv0); 2280 tvp = &utv; 2281 } else 2282 tvp = NULL; 2283 2284 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2285 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2286 2287 if (error == 0 && args->tsp != NULL) { 2288 if (td->td_retval[0] != 0) { 2289 /* 2290 * Compute how much time was left of the timeout, 2291 * by subtracting the current time and the time 2292 * before we started the call, and subtracting 2293 * that result from the user-supplied value. 2294 */ 2295 2296 microtime(&tv1); 2297 timevalsub(&tv1, &tv0); 2298 timevalsub(&utv, &tv1); 2299 if (utv.tv_sec < 0) 2300 timevalclear(&utv); 2301 } else 2302 timevalclear(&utv); 2303 2304 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2305 2306 native_to_linux_timespec(<s, &uts); 2307 error = copyout(<s, args->tsp, sizeof(lts)); 2308 } 2309 2310 return (error); 2311} 2312 2313int 2314linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2315{ 2316 struct timespec ts0, ts1; 2317 struct l_timespec lts; 2318 struct timespec uts, *tsp; 2319 l_sigset_t l_ss; 2320 sigset_t *ssp; 2321 sigset_t ss; 2322 int error; 2323 2324 if (args->sset != NULL) { 2325 if (args->ssize != sizeof(l_ss)) 2326 return (EINVAL); 2327 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2328 if (error) 2329 return (error); 2330 linux_to_bsd_sigset(&l_ss, &ss); 2331 ssp = &ss; 2332 } else 2333 ssp = NULL; 2334 if (args->tsp != NULL) { 2335 error = copyin(args->tsp, <s, sizeof(lts)); 2336 if (error) 2337 return (error); 2338 error = linux_to_native_timespec(&uts, <s); 2339 if (error != 0) 2340 return (error); 2341 2342 nanotime(&ts0); 2343 tsp = &uts; 2344 } else 2345 tsp = NULL; 2346 2347 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2348 2349 if (error == 0 && args->tsp != NULL) { 2350 if (td->td_retval[0]) { 2351 nanotime(&ts1); 2352 timespecsub(&ts1, &ts0); 2353 timespecsub(&uts, &ts1); 2354 if (uts.tv_sec < 0) 2355 timespecclear(&uts); 2356 } else 2357 timespecclear(&uts); 2358 2359 native_to_linux_timespec(<s, &uts); 2360 error = copyout(<s, args->tsp, sizeof(lts)); 2361 } 2362 2363 return (error); 2364} 2365 2366#if defined(DEBUG) || defined(KTR) 2367/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2368 2369#ifdef COMPAT_LINUX32 2370#define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL 2371#else 2372#define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL 2373#endif 2374 2375u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))]; 2376 2377static int 2378linux_debug(int syscall, int toggle, int global) 2379{ 2380 2381 if (global) { 2382 char c = toggle ? 0 : 0xff; 2383 2384 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2385 return (0); 2386 } 2387 if (syscall < 0 || syscall >= L_MAXSYSCALL) 2388 return (EINVAL); 2389 if (toggle) 2390 clrbit(linux_debug_map, syscall); 2391 else 2392 setbit(linux_debug_map, syscall); 2393 return (0); 2394} 2395#undef L_MAXSYSCALL 2396 2397/* 2398 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2399 * 2400 * E.g.: sysctl linux.debug=21.0 2401 * 2402 * As a special case, syscall "all" will apply to all syscalls globally. 2403 */ 2404#define LINUX_MAX_DEBUGSTR 16 2405int 2406linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2407{ 2408 char value[LINUX_MAX_DEBUGSTR], *p; 2409 int error, sysc, toggle; 2410 int global = 0; 2411 2412 value[0] = '\0'; 2413 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2414 if (error || req->newptr == NULL) 2415 return (error); 2416 for (p = value; *p != '\0' && *p != '.'; p++); 2417 if (*p == '\0') 2418 return (EINVAL); 2419 *p++ = '\0'; 2420 sysc = strtol(value, NULL, 0); 2421 toggle = strtol(p, NULL, 0); 2422 if (strcmp(value, "all") == 0) 2423 global = 1; 2424 error = linux_debug(sysc, toggle, global); 2425 return (error); 2426} 2427 2428#endif /* DEBUG || KTR */ 2429 2430int 2431linux_sched_rr_get_interval(struct thread *td, 2432 struct linux_sched_rr_get_interval_args *uap) 2433{ 2434 struct timespec ts; 2435 struct l_timespec lts; 2436 struct thread *tdt; 2437 int error; 2438 2439 /* 2440 * According to man in case the invalid pid specified 2441 * EINVAL should be returned. 2442 */ 2443 if (uap->pid < 0) 2444 return (EINVAL); 2445 2446 tdt = linux_tdfind(td, uap->pid, -1); 2447 if (tdt == NULL) 2448 return (ESRCH); 2449 2450 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2451 PROC_UNLOCK(tdt->td_proc); 2452 if (error != 0) 2453 return (error); 2454 native_to_linux_timespec(<s, &ts); 2455 return (copyout(<s, uap->interval, sizeof(lts))); 2456} 2457 2458/* 2459 * In case when the Linux thread is the initial thread in 2460 * the thread group thread id is equal to the process id. 2461 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2462 */ 2463struct thread * 2464linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2465{ 2466 struct linux_emuldata *em; 2467 struct thread *tdt; 2468 struct proc *p; 2469 2470 tdt = NULL; 2471 if (tid == 0 || tid == td->td_tid) { 2472 tdt = td; 2473 PROC_LOCK(tdt->td_proc); 2474 } else if (tid > PID_MAX) 2475 tdt = tdfind(tid, pid); 2476 else { 2477 /* 2478 * Initial thread where the tid equal to the pid. 2479 */ 2480 p = pfind(tid); 2481 if (p != NULL) { 2482 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2483 /* 2484 * p is not a Linuxulator process. 2485 */ 2486 PROC_UNLOCK(p); 2487 return (NULL); 2488 } 2489 FOREACH_THREAD_IN_PROC(p, tdt) { 2490 em = em_find(tdt); 2491 if (tid == em->em_tid) 2492 return (tdt); 2493 } 2494 PROC_UNLOCK(p); 2495 } 2496 return (NULL); 2497 } 2498 2499 return (tdt); 2500} 2501 2502void 2503linux_to_bsd_waitopts(int options, int *bsdopts) 2504{ 2505 2506 if (options & LINUX_WNOHANG) 2507 *bsdopts |= WNOHANG; 2508 if (options & LINUX_WUNTRACED) 2509 *bsdopts |= WUNTRACED; 2510 if (options & LINUX_WEXITED) 2511 *bsdopts |= WEXITED; 2512 if (options & LINUX_WCONTINUED) 2513 *bsdopts |= WCONTINUED; 2514 if (options & LINUX_WNOWAIT) 2515 *bsdopts |= WNOWAIT; 2516 2517 if (options & __WCLONE) 2518 *bsdopts |= WLINUXCLONE; 2519} 2520