linux_misc.c revision 297304
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 297304 2016-03-27 06:52:40Z 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 } 898 899 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) 900 /* This breaks POSIX, but is what the Linux kernel does 901 * _on purpose_ (documented in the man page for utimensat(2)), 902 * so we must follow that behaviour. */ 903 return (0); 904 905 if (args->pathname != NULL) 906 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 907 else if (args->flags != 0) 908 return (EINVAL); 909 910 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 911 flags |= AT_SYMLINK_NOFOLLOW; 912 913 if (path == NULL) 914 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 915 else { 916 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 917 UIO_SYSSPACE, flags); 918 LFREEPATH(path); 919 } 920 921 return (error); 922} 923 924int 925linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 926{ 927 l_timeval ltv[2]; 928 struct timeval tv[2], *tvp = NULL; 929 char *fname; 930 int error, dfd; 931 932 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 933 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 934 935#ifdef DEBUG 936 if (ldebug(futimesat)) 937 printf(ARGS(futimesat, "%s, *"), fname); 938#endif 939 940 if (args->utimes != NULL) { 941 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 942 LFREEPATH(fname); 943 return (error); 944 } 945 tv[0].tv_sec = ltv[0].tv_sec; 946 tv[0].tv_usec = ltv[0].tv_usec; 947 tv[1].tv_sec = ltv[1].tv_sec; 948 tv[1].tv_usec = ltv[1].tv_usec; 949 tvp = tv; 950 } 951 952 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 953 LFREEPATH(fname); 954 return (error); 955} 956 957int 958linux_common_wait(struct thread *td, int pid, int *status, 959 int options, struct rusage *ru) 960{ 961 int error, tmpstat; 962 963 error = kern_wait(td, pid, &tmpstat, options, ru); 964 if (error) 965 return (error); 966 967 if (status) { 968 tmpstat &= 0xffff; 969 if (WIFSIGNALED(tmpstat)) 970 tmpstat = (tmpstat & 0xffffff80) | 971 bsd_to_linux_signal(WTERMSIG(tmpstat)); 972 else if (WIFSTOPPED(tmpstat)) 973 tmpstat = (tmpstat & 0xffff00ff) | 974 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 975 else if (WIFCONTINUED(tmpstat)) 976 tmpstat = 0xffff; 977 error = copyout(&tmpstat, status, sizeof(int)); 978 } 979 980 return (error); 981} 982 983#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 984int 985linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 986{ 987 struct linux_wait4_args wait4_args; 988 989#ifdef DEBUG 990 if (ldebug(waitpid)) 991 printf(ARGS(waitpid, "%d, %p, %d"), 992 args->pid, (void *)args->status, args->options); 993#endif 994 995 wait4_args.pid = args->pid; 996 wait4_args.status = args->status; 997 wait4_args.options = args->options; 998 wait4_args.rusage = NULL; 999 1000 return (linux_wait4(td, &wait4_args)); 1001} 1002#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1003 1004int 1005linux_wait4(struct thread *td, struct linux_wait4_args *args) 1006{ 1007 int error, options; 1008 struct rusage ru, *rup; 1009 1010#ifdef DEBUG 1011 if (ldebug(wait4)) 1012 printf(ARGS(wait4, "%d, %p, %d, %p"), 1013 args->pid, (void *)args->status, args->options, 1014 (void *)args->rusage); 1015#endif 1016 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 1017 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 1018 return (EINVAL); 1019 1020 options = WEXITED; 1021 linux_to_bsd_waitopts(args->options, &options); 1022 1023 if (args->rusage != NULL) 1024 rup = &ru; 1025 else 1026 rup = NULL; 1027 error = linux_common_wait(td, args->pid, args->status, options, rup); 1028 if (error != 0) 1029 return (error); 1030 if (args->rusage != NULL) 1031 error = linux_copyout_rusage(&ru, args->rusage); 1032 return (error); 1033} 1034 1035int 1036linux_waitid(struct thread *td, struct linux_waitid_args *args) 1037{ 1038 int status, options, sig; 1039 struct __wrusage wru; 1040 siginfo_t siginfo; 1041 l_siginfo_t lsi; 1042 idtype_t idtype; 1043 struct proc *p; 1044 int error; 1045 1046 options = 0; 1047 linux_to_bsd_waitopts(args->options, &options); 1048 1049 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1050 return (EINVAL); 1051 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1052 return (EINVAL); 1053 1054 switch (args->idtype) { 1055 case LINUX_P_ALL: 1056 idtype = P_ALL; 1057 break; 1058 case LINUX_P_PID: 1059 if (args->id <= 0) 1060 return (EINVAL); 1061 idtype = P_PID; 1062 break; 1063 case LINUX_P_PGID: 1064 if (args->id <= 0) 1065 return (EINVAL); 1066 idtype = P_PGID; 1067 break; 1068 default: 1069 return (EINVAL); 1070 } 1071 1072 error = kern_wait6(td, idtype, args->id, &status, options, 1073 &wru, &siginfo); 1074 if (error != 0) 1075 return (error); 1076 if (args->rusage != NULL) { 1077 error = linux_copyout_rusage(&wru.wru_children, 1078 args->rusage); 1079 if (error != 0) 1080 return (error); 1081 } 1082 if (args->info != NULL) { 1083 p = td->td_proc; 1084 if (td->td_retval[0] == 0) 1085 bzero(&lsi, sizeof(lsi)); 1086 else { 1087 sig = bsd_to_linux_signal(siginfo.si_signo); 1088 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1089 } 1090 error = copyout(&lsi, args->info, sizeof(lsi)); 1091 } 1092 td->td_retval[0] = 0; 1093 1094 return (error); 1095} 1096 1097int 1098linux_mknod(struct thread *td, struct linux_mknod_args *args) 1099{ 1100 char *path; 1101 int error; 1102 1103 LCONVPATHCREAT(td, args->path, &path); 1104 1105#ifdef DEBUG 1106 if (ldebug(mknod)) 1107 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1108 (uintmax_t)args->dev); 1109#endif 1110 1111 switch (args->mode & S_IFMT) { 1112 case S_IFIFO: 1113 case S_IFSOCK: 1114 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1115 break; 1116 1117 case S_IFCHR: 1118 case S_IFBLK: 1119 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1120 args->dev); 1121 break; 1122 1123 case S_IFDIR: 1124 error = EPERM; 1125 break; 1126 1127 case 0: 1128 args->mode |= S_IFREG; 1129 /* FALLTHROUGH */ 1130 case S_IFREG: 1131 error = kern_open(td, path, UIO_SYSSPACE, 1132 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1133 if (error == 0) 1134 kern_close(td, td->td_retval[0]); 1135 break; 1136 1137 default: 1138 error = EINVAL; 1139 break; 1140 } 1141 LFREEPATH(path); 1142 return (error); 1143} 1144 1145int 1146linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1147{ 1148 char *path; 1149 int error, dfd; 1150 1151 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1152 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1153 1154#ifdef DEBUG 1155 if (ldebug(mknodat)) 1156 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1157#endif 1158 1159 switch (args->mode & S_IFMT) { 1160 case S_IFIFO: 1161 case S_IFSOCK: 1162 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1163 break; 1164 1165 case S_IFCHR: 1166 case S_IFBLK: 1167 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1168 args->dev); 1169 break; 1170 1171 case S_IFDIR: 1172 error = EPERM; 1173 break; 1174 1175 case 0: 1176 args->mode |= S_IFREG; 1177 /* FALLTHROUGH */ 1178 case S_IFREG: 1179 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1180 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1181 if (error == 0) 1182 kern_close(td, td->td_retval[0]); 1183 break; 1184 1185 default: 1186 error = EINVAL; 1187 break; 1188 } 1189 LFREEPATH(path); 1190 return (error); 1191} 1192 1193/* 1194 * UGH! This is just about the dumbest idea I've ever heard!! 1195 */ 1196int 1197linux_personality(struct thread *td, struct linux_personality_args *args) 1198{ 1199#ifdef DEBUG 1200 if (ldebug(personality)) 1201 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1202#endif 1203 if (args->per != 0) 1204 return (EINVAL); 1205 1206 /* Yes Jim, it's still a Linux... */ 1207 td->td_retval[0] = 0; 1208 return (0); 1209} 1210 1211struct l_itimerval { 1212 l_timeval it_interval; 1213 l_timeval it_value; 1214}; 1215 1216#define B2L_ITIMERVAL(bip, lip) \ 1217 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1218 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1219 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1220 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1221 1222int 1223linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1224{ 1225 int error; 1226 struct l_itimerval ls; 1227 struct itimerval aitv, oitv; 1228 1229#ifdef DEBUG 1230 if (ldebug(setitimer)) 1231 printf(ARGS(setitimer, "%p, %p"), 1232 (void *)uap->itv, (void *)uap->oitv); 1233#endif 1234 1235 if (uap->itv == NULL) { 1236 uap->itv = uap->oitv; 1237 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1238 } 1239 1240 error = copyin(uap->itv, &ls, sizeof(ls)); 1241 if (error != 0) 1242 return (error); 1243 B2L_ITIMERVAL(&aitv, &ls); 1244#ifdef DEBUG 1245 if (ldebug(setitimer)) { 1246 printf("setitimer: value: sec: %jd, usec: %ld\n", 1247 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1248 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1249 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1250 } 1251#endif 1252 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1253 if (error != 0 || uap->oitv == NULL) 1254 return (error); 1255 B2L_ITIMERVAL(&ls, &oitv); 1256 1257 return (copyout(&ls, uap->oitv, sizeof(ls))); 1258} 1259 1260int 1261linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1262{ 1263 int error; 1264 struct l_itimerval ls; 1265 struct itimerval aitv; 1266 1267#ifdef DEBUG 1268 if (ldebug(getitimer)) 1269 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1270#endif 1271 error = kern_getitimer(td, uap->which, &aitv); 1272 if (error != 0) 1273 return (error); 1274 B2L_ITIMERVAL(&ls, &aitv); 1275 return (copyout(&ls, uap->itv, sizeof(ls))); 1276} 1277 1278#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1279int 1280linux_nice(struct thread *td, struct linux_nice_args *args) 1281{ 1282 struct setpriority_args bsd_args; 1283 1284 bsd_args.which = PRIO_PROCESS; 1285 bsd_args.who = 0; /* current process */ 1286 bsd_args.prio = args->inc; 1287 return (sys_setpriority(td, &bsd_args)); 1288} 1289#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1290 1291int 1292linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1293{ 1294 struct ucred *newcred, *oldcred; 1295 l_gid_t *linux_gidset; 1296 gid_t *bsd_gidset; 1297 int ngrp, error; 1298 struct proc *p; 1299 1300 ngrp = args->gidsetsize; 1301 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1302 return (EINVAL); 1303 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1304 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1305 if (error) 1306 goto out; 1307 newcred = crget(); 1308 crextend(newcred, ngrp + 1); 1309 p = td->td_proc; 1310 PROC_LOCK(p); 1311 oldcred = p->p_ucred; 1312 crcopy(newcred, oldcred); 1313 1314 /* 1315 * cr_groups[0] holds egid. Setting the whole set from 1316 * the supplied set will cause egid to be changed too. 1317 * Keep cr_groups[0] unchanged to prevent that. 1318 */ 1319 1320 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1321 PROC_UNLOCK(p); 1322 crfree(newcred); 1323 goto out; 1324 } 1325 1326 if (ngrp > 0) { 1327 newcred->cr_ngroups = ngrp + 1; 1328 1329 bsd_gidset = newcred->cr_groups; 1330 ngrp--; 1331 while (ngrp >= 0) { 1332 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1333 ngrp--; 1334 } 1335 } else 1336 newcred->cr_ngroups = 1; 1337 1338 setsugid(p); 1339 p->p_ucred = newcred; 1340 PROC_UNLOCK(p); 1341 crfree(oldcred); 1342 error = 0; 1343out: 1344 free(linux_gidset, M_LINUX); 1345 return (error); 1346} 1347 1348int 1349linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1350{ 1351 struct ucred *cred; 1352 l_gid_t *linux_gidset; 1353 gid_t *bsd_gidset; 1354 int bsd_gidsetsz, ngrp, error; 1355 1356 cred = td->td_ucred; 1357 bsd_gidset = cred->cr_groups; 1358 bsd_gidsetsz = cred->cr_ngroups - 1; 1359 1360 /* 1361 * cr_groups[0] holds egid. Returning the whole set 1362 * here will cause a duplicate. Exclude cr_groups[0] 1363 * to prevent that. 1364 */ 1365 1366 if ((ngrp = args->gidsetsize) == 0) { 1367 td->td_retval[0] = bsd_gidsetsz; 1368 return (0); 1369 } 1370 1371 if (ngrp < bsd_gidsetsz) 1372 return (EINVAL); 1373 1374 ngrp = 0; 1375 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1376 M_LINUX, M_WAITOK); 1377 while (ngrp < bsd_gidsetsz) { 1378 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1379 ngrp++; 1380 } 1381 1382 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1383 free(linux_gidset, M_LINUX); 1384 if (error) 1385 return (error); 1386 1387 td->td_retval[0] = ngrp; 1388 return (0); 1389} 1390 1391int 1392linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1393{ 1394 struct rlimit bsd_rlim; 1395 struct l_rlimit rlim; 1396 u_int which; 1397 int error; 1398 1399#ifdef DEBUG 1400 if (ldebug(setrlimit)) 1401 printf(ARGS(setrlimit, "%d, %p"), 1402 args->resource, (void *)args->rlim); 1403#endif 1404 1405 if (args->resource >= LINUX_RLIM_NLIMITS) 1406 return (EINVAL); 1407 1408 which = linux_to_bsd_resource[args->resource]; 1409 if (which == -1) 1410 return (EINVAL); 1411 1412 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1413 if (error) 1414 return (error); 1415 1416 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1417 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1418 return (kern_setrlimit(td, which, &bsd_rlim)); 1419} 1420 1421#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1422int 1423linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1424{ 1425 struct l_rlimit rlim; 1426 struct proc *p = td->td_proc; 1427 struct rlimit bsd_rlim; 1428 u_int which; 1429 1430#ifdef DEBUG 1431 if (ldebug(old_getrlimit)) 1432 printf(ARGS(old_getrlimit, "%d, %p"), 1433 args->resource, (void *)args->rlim); 1434#endif 1435 1436 if (args->resource >= LINUX_RLIM_NLIMITS) 1437 return (EINVAL); 1438 1439 which = linux_to_bsd_resource[args->resource]; 1440 if (which == -1) 1441 return (EINVAL); 1442 1443 PROC_LOCK(p); 1444 lim_rlimit(p, which, &bsd_rlim); 1445 PROC_UNLOCK(p); 1446 1447#ifdef COMPAT_LINUX32 1448 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1449 if (rlim.rlim_cur == UINT_MAX) 1450 rlim.rlim_cur = INT_MAX; 1451 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1452 if (rlim.rlim_max == UINT_MAX) 1453 rlim.rlim_max = INT_MAX; 1454#else 1455 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1456 if (rlim.rlim_cur == ULONG_MAX) 1457 rlim.rlim_cur = LONG_MAX; 1458 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1459 if (rlim.rlim_max == ULONG_MAX) 1460 rlim.rlim_max = LONG_MAX; 1461#endif 1462 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1463} 1464#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1465 1466int 1467linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1468{ 1469 struct l_rlimit rlim; 1470 struct proc *p = td->td_proc; 1471 struct rlimit bsd_rlim; 1472 u_int which; 1473 1474#ifdef DEBUG 1475 if (ldebug(getrlimit)) 1476 printf(ARGS(getrlimit, "%d, %p"), 1477 args->resource, (void *)args->rlim); 1478#endif 1479 1480 if (args->resource >= LINUX_RLIM_NLIMITS) 1481 return (EINVAL); 1482 1483 which = linux_to_bsd_resource[args->resource]; 1484 if (which == -1) 1485 return (EINVAL); 1486 1487 PROC_LOCK(p); 1488 lim_rlimit(p, which, &bsd_rlim); 1489 PROC_UNLOCK(p); 1490 1491 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1492 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1493 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1494} 1495 1496int 1497linux_sched_setscheduler(struct thread *td, 1498 struct linux_sched_setscheduler_args *args) 1499{ 1500 struct sched_param sched_param; 1501 struct thread *tdt; 1502 int error, policy; 1503 1504#ifdef DEBUG 1505 if (ldebug(sched_setscheduler)) 1506 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1507 args->pid, args->policy, (const void *)args->param); 1508#endif 1509 1510 switch (args->policy) { 1511 case LINUX_SCHED_OTHER: 1512 policy = SCHED_OTHER; 1513 break; 1514 case LINUX_SCHED_FIFO: 1515 policy = SCHED_FIFO; 1516 break; 1517 case LINUX_SCHED_RR: 1518 policy = SCHED_RR; 1519 break; 1520 default: 1521 return (EINVAL); 1522 } 1523 1524 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1525 if (error) 1526 return (error); 1527 1528 tdt = linux_tdfind(td, args->pid, -1); 1529 if (tdt == NULL) 1530 return (ESRCH); 1531 1532 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1533 PROC_UNLOCK(tdt->td_proc); 1534 return (error); 1535} 1536 1537int 1538linux_sched_getscheduler(struct thread *td, 1539 struct linux_sched_getscheduler_args *args) 1540{ 1541 struct thread *tdt; 1542 int error, policy; 1543 1544#ifdef DEBUG 1545 if (ldebug(sched_getscheduler)) 1546 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1547#endif 1548 1549 tdt = linux_tdfind(td, args->pid, -1); 1550 if (tdt == NULL) 1551 return (ESRCH); 1552 1553 error = kern_sched_getscheduler(td, tdt, &policy); 1554 PROC_UNLOCK(tdt->td_proc); 1555 1556 switch (policy) { 1557 case SCHED_OTHER: 1558 td->td_retval[0] = LINUX_SCHED_OTHER; 1559 break; 1560 case SCHED_FIFO: 1561 td->td_retval[0] = LINUX_SCHED_FIFO; 1562 break; 1563 case SCHED_RR: 1564 td->td_retval[0] = LINUX_SCHED_RR; 1565 break; 1566 } 1567 return (error); 1568} 1569 1570int 1571linux_sched_get_priority_max(struct thread *td, 1572 struct linux_sched_get_priority_max_args *args) 1573{ 1574 struct sched_get_priority_max_args bsd; 1575 1576#ifdef DEBUG 1577 if (ldebug(sched_get_priority_max)) 1578 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1579#endif 1580 1581 switch (args->policy) { 1582 case LINUX_SCHED_OTHER: 1583 bsd.policy = SCHED_OTHER; 1584 break; 1585 case LINUX_SCHED_FIFO: 1586 bsd.policy = SCHED_FIFO; 1587 break; 1588 case LINUX_SCHED_RR: 1589 bsd.policy = SCHED_RR; 1590 break; 1591 default: 1592 return (EINVAL); 1593 } 1594 return (sys_sched_get_priority_max(td, &bsd)); 1595} 1596 1597int 1598linux_sched_get_priority_min(struct thread *td, 1599 struct linux_sched_get_priority_min_args *args) 1600{ 1601 struct sched_get_priority_min_args bsd; 1602 1603#ifdef DEBUG 1604 if (ldebug(sched_get_priority_min)) 1605 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1606#endif 1607 1608 switch (args->policy) { 1609 case LINUX_SCHED_OTHER: 1610 bsd.policy = SCHED_OTHER; 1611 break; 1612 case LINUX_SCHED_FIFO: 1613 bsd.policy = SCHED_FIFO; 1614 break; 1615 case LINUX_SCHED_RR: 1616 bsd.policy = SCHED_RR; 1617 break; 1618 default: 1619 return (EINVAL); 1620 } 1621 return (sys_sched_get_priority_min(td, &bsd)); 1622} 1623 1624#define REBOOT_CAD_ON 0x89abcdef 1625#define REBOOT_CAD_OFF 0 1626#define REBOOT_HALT 0xcdef0123 1627#define REBOOT_RESTART 0x01234567 1628#define REBOOT_RESTART2 0xA1B2C3D4 1629#define REBOOT_POWEROFF 0x4321FEDC 1630#define REBOOT_MAGIC1 0xfee1dead 1631#define REBOOT_MAGIC2 0x28121969 1632#define REBOOT_MAGIC2A 0x05121996 1633#define REBOOT_MAGIC2B 0x16041998 1634 1635int 1636linux_reboot(struct thread *td, struct linux_reboot_args *args) 1637{ 1638 struct reboot_args bsd_args; 1639 1640#ifdef DEBUG 1641 if (ldebug(reboot)) 1642 printf(ARGS(reboot, "0x%x"), args->cmd); 1643#endif 1644 1645 if (args->magic1 != REBOOT_MAGIC1) 1646 return (EINVAL); 1647 1648 switch (args->magic2) { 1649 case REBOOT_MAGIC2: 1650 case REBOOT_MAGIC2A: 1651 case REBOOT_MAGIC2B: 1652 break; 1653 default: 1654 return (EINVAL); 1655 } 1656 1657 switch (args->cmd) { 1658 case REBOOT_CAD_ON: 1659 case REBOOT_CAD_OFF: 1660 return (priv_check(td, PRIV_REBOOT)); 1661 case REBOOT_HALT: 1662 bsd_args.opt = RB_HALT; 1663 break; 1664 case REBOOT_RESTART: 1665 case REBOOT_RESTART2: 1666 bsd_args.opt = 0; 1667 break; 1668 case REBOOT_POWEROFF: 1669 bsd_args.opt = RB_POWEROFF; 1670 break; 1671 default: 1672 return (EINVAL); 1673 } 1674 return (sys_reboot(td, &bsd_args)); 1675} 1676 1677 1678/* 1679 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1680 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1681 * are assumed to be preserved. The following lightweight syscalls fixes 1682 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1683 * 1684 * linux_getpid() - MP SAFE 1685 * linux_getgid() - MP SAFE 1686 * linux_getuid() - MP SAFE 1687 */ 1688 1689int 1690linux_getpid(struct thread *td, struct linux_getpid_args *args) 1691{ 1692 1693#ifdef DEBUG 1694 if (ldebug(getpid)) 1695 printf(ARGS(getpid, "")); 1696#endif 1697 td->td_retval[0] = td->td_proc->p_pid; 1698 1699 return (0); 1700} 1701 1702int 1703linux_gettid(struct thread *td, struct linux_gettid_args *args) 1704{ 1705 struct linux_emuldata *em; 1706 1707#ifdef DEBUG 1708 if (ldebug(gettid)) 1709 printf(ARGS(gettid, "")); 1710#endif 1711 1712 em = em_find(td); 1713 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1714 1715 td->td_retval[0] = em->em_tid; 1716 1717 return (0); 1718} 1719 1720 1721int 1722linux_getppid(struct thread *td, struct linux_getppid_args *args) 1723{ 1724 1725#ifdef DEBUG 1726 if (ldebug(getppid)) 1727 printf(ARGS(getppid, "")); 1728#endif 1729 1730 PROC_LOCK(td->td_proc); 1731 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1732 PROC_UNLOCK(td->td_proc); 1733 return (0); 1734} 1735 1736int 1737linux_getgid(struct thread *td, struct linux_getgid_args *args) 1738{ 1739 1740#ifdef DEBUG 1741 if (ldebug(getgid)) 1742 printf(ARGS(getgid, "")); 1743#endif 1744 1745 td->td_retval[0] = td->td_ucred->cr_rgid; 1746 return (0); 1747} 1748 1749int 1750linux_getuid(struct thread *td, struct linux_getuid_args *args) 1751{ 1752 1753#ifdef DEBUG 1754 if (ldebug(getuid)) 1755 printf(ARGS(getuid, "")); 1756#endif 1757 1758 td->td_retval[0] = td->td_ucred->cr_ruid; 1759 return (0); 1760} 1761 1762 1763int 1764linux_getsid(struct thread *td, struct linux_getsid_args *args) 1765{ 1766 struct getsid_args bsd; 1767 1768#ifdef DEBUG 1769 if (ldebug(getsid)) 1770 printf(ARGS(getsid, "%i"), args->pid); 1771#endif 1772 1773 bsd.pid = args->pid; 1774 return (sys_getsid(td, &bsd)); 1775} 1776 1777int 1778linux_nosys(struct thread *td, struct nosys_args *ignore) 1779{ 1780 1781 return (ENOSYS); 1782} 1783 1784int 1785linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1786{ 1787 struct getpriority_args bsd_args; 1788 int error; 1789 1790#ifdef DEBUG 1791 if (ldebug(getpriority)) 1792 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1793#endif 1794 1795 bsd_args.which = args->which; 1796 bsd_args.who = args->who; 1797 error = sys_getpriority(td, &bsd_args); 1798 td->td_retval[0] = 20 - td->td_retval[0]; 1799 return (error); 1800} 1801 1802int 1803linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1804{ 1805 int name[2]; 1806 1807#ifdef DEBUG 1808 if (ldebug(sethostname)) 1809 printf(ARGS(sethostname, "*, %i"), args->len); 1810#endif 1811 1812 name[0] = CTL_KERN; 1813 name[1] = KERN_HOSTNAME; 1814 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1815 args->len, 0, 0)); 1816} 1817 1818int 1819linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1820{ 1821 int name[2]; 1822 1823#ifdef DEBUG 1824 if (ldebug(setdomainname)) 1825 printf(ARGS(setdomainname, "*, %i"), args->len); 1826#endif 1827 1828 name[0] = CTL_KERN; 1829 name[1] = KERN_NISDOMAINNAME; 1830 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1831 args->len, 0, 0)); 1832} 1833 1834int 1835linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1836{ 1837 1838#ifdef DEBUG 1839 if (ldebug(exit_group)) 1840 printf(ARGS(exit_group, "%i"), args->error_code); 1841#endif 1842 1843 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1844 args->error_code); 1845 1846 /* 1847 * XXX: we should send a signal to the parent if 1848 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1849 * as it doesnt occur often. 1850 */ 1851 exit1(td, W_EXITCODE(args->error_code, 0)); 1852 /* NOTREACHED */ 1853} 1854 1855#define _LINUX_CAPABILITY_VERSION 0x19980330 1856 1857struct l_user_cap_header { 1858 l_int version; 1859 l_int pid; 1860}; 1861 1862struct l_user_cap_data { 1863 l_int effective; 1864 l_int permitted; 1865 l_int inheritable; 1866}; 1867 1868int 1869linux_capget(struct thread *td, struct linux_capget_args *args) 1870{ 1871 struct l_user_cap_header luch; 1872 struct l_user_cap_data lucd; 1873 int error; 1874 1875 if (args->hdrp == NULL) 1876 return (EFAULT); 1877 1878 error = copyin(args->hdrp, &luch, sizeof(luch)); 1879 if (error != 0) 1880 return (error); 1881 1882 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1883 luch.version = _LINUX_CAPABILITY_VERSION; 1884 error = copyout(&luch, args->hdrp, sizeof(luch)); 1885 if (error) 1886 return (error); 1887 return (EINVAL); 1888 } 1889 1890 if (luch.pid) 1891 return (EPERM); 1892 1893 if (args->datap) { 1894 /* 1895 * The current implementation doesn't support setting 1896 * a capability (it's essentially a stub) so indicate 1897 * that no capabilities are currently set or available 1898 * to request. 1899 */ 1900 bzero (&lucd, sizeof(lucd)); 1901 error = copyout(&lucd, args->datap, sizeof(lucd)); 1902 } 1903 1904 return (error); 1905} 1906 1907int 1908linux_capset(struct thread *td, struct linux_capset_args *args) 1909{ 1910 struct l_user_cap_header luch; 1911 struct l_user_cap_data lucd; 1912 int error; 1913 1914 if (args->hdrp == NULL || args->datap == NULL) 1915 return (EFAULT); 1916 1917 error = copyin(args->hdrp, &luch, sizeof(luch)); 1918 if (error != 0) 1919 return (error); 1920 1921 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1922 luch.version = _LINUX_CAPABILITY_VERSION; 1923 error = copyout(&luch, args->hdrp, sizeof(luch)); 1924 if (error) 1925 return (error); 1926 return (EINVAL); 1927 } 1928 1929 if (luch.pid) 1930 return (EPERM); 1931 1932 error = copyin(args->datap, &lucd, sizeof(lucd)); 1933 if (error != 0) 1934 return (error); 1935 1936 /* We currently don't support setting any capabilities. */ 1937 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1938 linux_msg(td, 1939 "capset effective=0x%x, permitted=0x%x, " 1940 "inheritable=0x%x is not implemented", 1941 (int)lucd.effective, (int)lucd.permitted, 1942 (int)lucd.inheritable); 1943 return (EPERM); 1944 } 1945 1946 return (0); 1947} 1948 1949int 1950linux_prctl(struct thread *td, struct linux_prctl_args *args) 1951{ 1952 int error = 0, max_size; 1953 struct proc *p = td->td_proc; 1954 char comm[LINUX_MAX_COMM_LEN]; 1955 struct linux_emuldata *em; 1956 int pdeath_signal; 1957 1958#ifdef DEBUG 1959 if (ldebug(prctl)) 1960 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1961 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1962 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1963#endif 1964 1965 switch (args->option) { 1966 case LINUX_PR_SET_PDEATHSIG: 1967 if (!LINUX_SIG_VALID(args->arg2)) 1968 return (EINVAL); 1969 em = em_find(td); 1970 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1971 em->pdeath_signal = args->arg2; 1972 break; 1973 case LINUX_PR_GET_PDEATHSIG: 1974 em = em_find(td); 1975 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1976 pdeath_signal = em->pdeath_signal; 1977 error = copyout(&pdeath_signal, 1978 (void *)(register_t)args->arg2, 1979 sizeof(pdeath_signal)); 1980 break; 1981 case LINUX_PR_GET_KEEPCAPS: 1982 /* 1983 * Indicate that we always clear the effective and 1984 * permitted capability sets when the user id becomes 1985 * non-zero (actually the capability sets are simply 1986 * always zero in the current implementation). 1987 */ 1988 td->td_retval[0] = 0; 1989 break; 1990 case LINUX_PR_SET_KEEPCAPS: 1991 /* 1992 * Ignore requests to keep the effective and permitted 1993 * capability sets when the user id becomes non-zero. 1994 */ 1995 break; 1996 case LINUX_PR_SET_NAME: 1997 /* 1998 * To be on the safe side we need to make sure to not 1999 * overflow the size a linux program expects. We already 2000 * do this here in the copyin, so that we don't need to 2001 * check on copyout. 2002 */ 2003 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 2004 error = copyinstr((void *)(register_t)args->arg2, comm, 2005 max_size, NULL); 2006 2007 /* Linux silently truncates the name if it is too long. */ 2008 if (error == ENAMETOOLONG) { 2009 /* 2010 * XXX: copyinstr() isn't documented to populate the 2011 * array completely, so do a copyin() to be on the 2012 * safe side. This should be changed in case 2013 * copyinstr() is changed to guarantee this. 2014 */ 2015 error = copyin((void *)(register_t)args->arg2, comm, 2016 max_size - 1); 2017 comm[max_size - 1] = '\0'; 2018 } 2019 if (error) 2020 return (error); 2021 2022 PROC_LOCK(p); 2023 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 2024 PROC_UNLOCK(p); 2025 break; 2026 case LINUX_PR_GET_NAME: 2027 PROC_LOCK(p); 2028 strlcpy(comm, p->p_comm, sizeof(comm)); 2029 PROC_UNLOCK(p); 2030 error = copyout(comm, (void *)(register_t)args->arg2, 2031 strlen(comm) + 1); 2032 break; 2033 default: 2034 error = EINVAL; 2035 break; 2036 } 2037 2038 return (error); 2039} 2040 2041int 2042linux_sched_setparam(struct thread *td, 2043 struct linux_sched_setparam_args *uap) 2044{ 2045 struct sched_param sched_param; 2046 struct thread *tdt; 2047 int error; 2048 2049#ifdef DEBUG 2050 if (ldebug(sched_setparam)) 2051 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 2052#endif 2053 2054 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 2055 if (error) 2056 return (error); 2057 2058 tdt = linux_tdfind(td, uap->pid, -1); 2059 if (tdt == NULL) 2060 return (ESRCH); 2061 2062 error = kern_sched_setparam(td, tdt, &sched_param); 2063 PROC_UNLOCK(tdt->td_proc); 2064 return (error); 2065} 2066 2067int 2068linux_sched_getparam(struct thread *td, 2069 struct linux_sched_getparam_args *uap) 2070{ 2071 struct sched_param sched_param; 2072 struct thread *tdt; 2073 int error; 2074 2075#ifdef DEBUG 2076 if (ldebug(sched_getparam)) 2077 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 2078#endif 2079 2080 tdt = linux_tdfind(td, uap->pid, -1); 2081 if (tdt == NULL) 2082 return (ESRCH); 2083 2084 error = kern_sched_getparam(td, tdt, &sched_param); 2085 PROC_UNLOCK(tdt->td_proc); 2086 if (error == 0) 2087 error = copyout(&sched_param, uap->param, 2088 sizeof(sched_param)); 2089 return (error); 2090} 2091 2092/* 2093 * Get affinity of a process. 2094 */ 2095int 2096linux_sched_getaffinity(struct thread *td, 2097 struct linux_sched_getaffinity_args *args) 2098{ 2099 int error; 2100 struct thread *tdt; 2101 struct cpuset_getaffinity_args cga; 2102 2103#ifdef DEBUG 2104 if (ldebug(sched_getaffinity)) 2105 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2106 args->len); 2107#endif 2108 if (args->len < sizeof(cpuset_t)) 2109 return (EINVAL); 2110 2111 tdt = linux_tdfind(td, args->pid, -1); 2112 if (tdt == NULL) 2113 return (ESRCH); 2114 2115 PROC_UNLOCK(tdt->td_proc); 2116 cga.level = CPU_LEVEL_WHICH; 2117 cga.which = CPU_WHICH_TID; 2118 cga.id = tdt->td_tid; 2119 cga.cpusetsize = sizeof(cpuset_t); 2120 cga.mask = (cpuset_t *) args->user_mask_ptr; 2121 2122 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2123 td->td_retval[0] = sizeof(cpuset_t); 2124 2125 return (error); 2126} 2127 2128/* 2129 * Set affinity of a process. 2130 */ 2131int 2132linux_sched_setaffinity(struct thread *td, 2133 struct linux_sched_setaffinity_args *args) 2134{ 2135 struct cpuset_setaffinity_args csa; 2136 struct thread *tdt; 2137 2138#ifdef DEBUG 2139 if (ldebug(sched_setaffinity)) 2140 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2141 args->len); 2142#endif 2143 if (args->len < sizeof(cpuset_t)) 2144 return (EINVAL); 2145 2146 tdt = linux_tdfind(td, args->pid, -1); 2147 if (tdt == NULL) 2148 return (ESRCH); 2149 2150 PROC_UNLOCK(tdt->td_proc); 2151 csa.level = CPU_LEVEL_WHICH; 2152 csa.which = CPU_WHICH_TID; 2153 csa.id = tdt->td_tid; 2154 csa.cpusetsize = sizeof(cpuset_t); 2155 csa.mask = (cpuset_t *) args->user_mask_ptr; 2156 2157 return (sys_cpuset_setaffinity(td, &csa)); 2158} 2159 2160struct linux_rlimit64 { 2161 uint64_t rlim_cur; 2162 uint64_t rlim_max; 2163}; 2164 2165int 2166linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2167{ 2168 struct rlimit rlim, nrlim; 2169 struct linux_rlimit64 lrlim; 2170 struct proc *p; 2171 u_int which; 2172 int flags; 2173 int error; 2174 2175#ifdef DEBUG 2176 if (ldebug(prlimit64)) 2177 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2178 args->resource, (void *)args->new, (void *)args->old); 2179#endif 2180 2181 if (args->resource >= LINUX_RLIM_NLIMITS) 2182 return (EINVAL); 2183 2184 which = linux_to_bsd_resource[args->resource]; 2185 if (which == -1) 2186 return (EINVAL); 2187 2188 if (args->new != NULL) { 2189 /* 2190 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2191 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2192 * as INFINITY so we do not need a conversion even. 2193 */ 2194 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2195 if (error != 0) 2196 return (error); 2197 } 2198 2199 flags = PGET_HOLD | PGET_NOTWEXIT; 2200 if (args->new != NULL) 2201 flags |= PGET_CANDEBUG; 2202 else 2203 flags |= PGET_CANSEE; 2204 error = pget(args->pid, flags, &p); 2205 if (error != 0) 2206 return (error); 2207 2208 if (args->old != NULL) { 2209 PROC_LOCK(p); 2210 lim_rlimit(p, which, &rlim); 2211 PROC_UNLOCK(p); 2212 if (rlim.rlim_cur == RLIM_INFINITY) 2213 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2214 else 2215 lrlim.rlim_cur = rlim.rlim_cur; 2216 if (rlim.rlim_max == RLIM_INFINITY) 2217 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2218 else 2219 lrlim.rlim_max = rlim.rlim_max; 2220 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2221 if (error != 0) 2222 goto out; 2223 } 2224 2225 if (args->new != NULL) 2226 error = kern_proc_setrlimit(td, p, which, &nrlim); 2227 2228 out: 2229 PRELE(p); 2230 return (error); 2231} 2232 2233int 2234linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2235{ 2236 struct timeval utv, tv0, tv1, *tvp; 2237 struct l_pselect6arg lpse6; 2238 struct l_timespec lts; 2239 struct timespec uts; 2240 l_sigset_t l_ss; 2241 sigset_t *ssp; 2242 sigset_t ss; 2243 int error; 2244 2245 ssp = NULL; 2246 if (args->sig != NULL) { 2247 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2248 if (error != 0) 2249 return (error); 2250 if (lpse6.ss_len != sizeof(l_ss)) 2251 return (EINVAL); 2252 if (lpse6.ss != 0) { 2253 error = copyin(PTRIN(lpse6.ss), &l_ss, 2254 sizeof(l_ss)); 2255 if (error != 0) 2256 return (error); 2257 linux_to_bsd_sigset(&l_ss, &ss); 2258 ssp = &ss; 2259 } 2260 } 2261 2262 /* 2263 * Currently glibc changes nanosecond number to microsecond. 2264 * This mean losing precision but for now it is hardly seen. 2265 */ 2266 if (args->tsp != NULL) { 2267 error = copyin(args->tsp, <s, sizeof(lts)); 2268 if (error != 0) 2269 return (error); 2270 error = linux_to_native_timespec(&uts, <s); 2271 if (error != 0) 2272 return (error); 2273 2274 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2275 if (itimerfix(&utv)) 2276 return (EINVAL); 2277 2278 microtime(&tv0); 2279 tvp = &utv; 2280 } else 2281 tvp = NULL; 2282 2283 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2284 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2285 2286 if (error == 0 && args->tsp != NULL) { 2287 if (td->td_retval[0] != 0) { 2288 /* 2289 * Compute how much time was left of the timeout, 2290 * by subtracting the current time and the time 2291 * before we started the call, and subtracting 2292 * that result from the user-supplied value. 2293 */ 2294 2295 microtime(&tv1); 2296 timevalsub(&tv1, &tv0); 2297 timevalsub(&utv, &tv1); 2298 if (utv.tv_sec < 0) 2299 timevalclear(&utv); 2300 } else 2301 timevalclear(&utv); 2302 2303 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2304 2305 native_to_linux_timespec(<s, &uts); 2306 error = copyout(<s, args->tsp, sizeof(lts)); 2307 } 2308 2309 return (error); 2310} 2311 2312int 2313linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2314{ 2315 struct timespec ts0, ts1; 2316 struct l_timespec lts; 2317 struct timespec uts, *tsp; 2318 l_sigset_t l_ss; 2319 sigset_t *ssp; 2320 sigset_t ss; 2321 int error; 2322 2323 if (args->sset != NULL) { 2324 if (args->ssize != sizeof(l_ss)) 2325 return (EINVAL); 2326 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2327 if (error) 2328 return (error); 2329 linux_to_bsd_sigset(&l_ss, &ss); 2330 ssp = &ss; 2331 } else 2332 ssp = NULL; 2333 if (args->tsp != NULL) { 2334 error = copyin(args->tsp, <s, sizeof(lts)); 2335 if (error) 2336 return (error); 2337 error = linux_to_native_timespec(&uts, <s); 2338 if (error != 0) 2339 return (error); 2340 2341 nanotime(&ts0); 2342 tsp = &uts; 2343 } else 2344 tsp = NULL; 2345 2346 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2347 2348 if (error == 0 && args->tsp != NULL) { 2349 if (td->td_retval[0]) { 2350 nanotime(&ts1); 2351 timespecsub(&ts1, &ts0); 2352 timespecsub(&uts, &ts1); 2353 if (uts.tv_sec < 0) 2354 timespecclear(&uts); 2355 } else 2356 timespecclear(&uts); 2357 2358 native_to_linux_timespec(<s, &uts); 2359 error = copyout(<s, args->tsp, sizeof(lts)); 2360 } 2361 2362 return (error); 2363} 2364 2365#if defined(DEBUG) || defined(KTR) 2366/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2367 2368#ifdef COMPAT_LINUX32 2369#define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL 2370#else 2371#define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL 2372#endif 2373 2374u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))]; 2375 2376static int 2377linux_debug(int syscall, int toggle, int global) 2378{ 2379 2380 if (global) { 2381 char c = toggle ? 0 : 0xff; 2382 2383 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2384 return (0); 2385 } 2386 if (syscall < 0 || syscall >= L_MAXSYSCALL) 2387 return (EINVAL); 2388 if (toggle) 2389 clrbit(linux_debug_map, syscall); 2390 else 2391 setbit(linux_debug_map, syscall); 2392 return (0); 2393} 2394#undef L_MAXSYSCALL 2395 2396/* 2397 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2398 * 2399 * E.g.: sysctl linux.debug=21.0 2400 * 2401 * As a special case, syscall "all" will apply to all syscalls globally. 2402 */ 2403#define LINUX_MAX_DEBUGSTR 16 2404int 2405linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2406{ 2407 char value[LINUX_MAX_DEBUGSTR], *p; 2408 int error, sysc, toggle; 2409 int global = 0; 2410 2411 value[0] = '\0'; 2412 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2413 if (error || req->newptr == NULL) 2414 return (error); 2415 for (p = value; *p != '\0' && *p != '.'; p++); 2416 if (*p == '\0') 2417 return (EINVAL); 2418 *p++ = '\0'; 2419 sysc = strtol(value, NULL, 0); 2420 toggle = strtol(p, NULL, 0); 2421 if (strcmp(value, "all") == 0) 2422 global = 1; 2423 error = linux_debug(sysc, toggle, global); 2424 return (error); 2425} 2426 2427#endif /* DEBUG || KTR */ 2428 2429int 2430linux_sched_rr_get_interval(struct thread *td, 2431 struct linux_sched_rr_get_interval_args *uap) 2432{ 2433 struct timespec ts; 2434 struct l_timespec lts; 2435 struct thread *tdt; 2436 int error; 2437 2438 /* 2439 * According to man in case the invalid pid specified 2440 * EINVAL should be returned. 2441 */ 2442 if (uap->pid < 0) 2443 return (EINVAL); 2444 2445 tdt = linux_tdfind(td, uap->pid, -1); 2446 if (tdt == NULL) 2447 return (ESRCH); 2448 2449 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2450 PROC_UNLOCK(tdt->td_proc); 2451 if (error != 0) 2452 return (error); 2453 native_to_linux_timespec(<s, &ts); 2454 return (copyout(<s, uap->interval, sizeof(lts))); 2455} 2456 2457/* 2458 * In case when the Linux thread is the initial thread in 2459 * the thread group thread id is equal to the process id. 2460 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2461 */ 2462struct thread * 2463linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2464{ 2465 struct linux_emuldata *em; 2466 struct thread *tdt; 2467 struct proc *p; 2468 2469 tdt = NULL; 2470 if (tid == 0 || tid == td->td_tid) { 2471 tdt = td; 2472 PROC_LOCK(tdt->td_proc); 2473 } else if (tid > PID_MAX) 2474 tdt = tdfind(tid, pid); 2475 else { 2476 /* 2477 * Initial thread where the tid equal to the pid. 2478 */ 2479 p = pfind(tid); 2480 if (p != NULL) { 2481 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2482 /* 2483 * p is not a Linuxulator process. 2484 */ 2485 PROC_UNLOCK(p); 2486 return (NULL); 2487 } 2488 FOREACH_THREAD_IN_PROC(p, tdt) { 2489 em = em_find(tdt); 2490 if (tid == em->em_tid) 2491 return (tdt); 2492 } 2493 PROC_UNLOCK(p); 2494 } 2495 return (NULL); 2496 } 2497 2498 return (tdt); 2499} 2500 2501void 2502linux_to_bsd_waitopts(int options, int *bsdopts) 2503{ 2504 2505 if (options & LINUX_WNOHANG) 2506 *bsdopts |= WNOHANG; 2507 if (options & LINUX_WUNTRACED) 2508 *bsdopts |= WUNTRACED; 2509 if (options & LINUX_WEXITED) 2510 *bsdopts |= WEXITED; 2511 if (options & LINUX_WCONTINUED) 2512 *bsdopts |= WCONTINUED; 2513 if (options & LINUX_WNOWAIT) 2514 *bsdopts |= WNOWAIT; 2515 2516 if (options & __WCLONE) 2517 *bsdopts |= WLINUXCLONE; 2518} 2519