linux_misc.c revision 293591
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 293591 2016-01-09 17:53:32Z 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 201 202 secs = args->secs; 203 204 if (secs > INT_MAX) 205 secs = INT_MAX; 206 207 it.it_value.tv_sec = (long) secs; 208 it.it_value.tv_usec = 0; 209 it.it_interval.tv_sec = 0; 210 it.it_interval.tv_usec = 0; 211 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 212 if (error) 213 return (error); 214 if (timevalisset(&old_it.it_value)) { 215 if (old_it.it_value.tv_usec != 0) 216 old_it.it_value.tv_sec++; 217 td->td_retval[0] = old_it.it_value.tv_sec; 218 } 219 return (0); 220} 221 222int 223linux_brk(struct thread *td, struct linux_brk_args *args) 224{ 225 struct vmspace *vm = td->td_proc->p_vmspace; 226 vm_offset_t new, old; 227 struct obreak_args /* { 228 char * nsize; 229 } */ tmp; 230 231#ifdef DEBUG 232 if (ldebug(brk)) 233 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 234#endif 235 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 236 new = (vm_offset_t)args->dsend; 237 tmp.nsize = (char *)new; 238 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp)) 239 td->td_retval[0] = (long)new; 240 else 241 td->td_retval[0] = (long)old; 242 243 return (0); 244} 245 246#if defined(__i386__) 247/* XXX: what about amd64/linux32? */ 248 249int 250linux_uselib(struct thread *td, struct linux_uselib_args *args) 251{ 252 struct nameidata ni; 253 struct vnode *vp; 254 struct exec *a_out; 255 struct vattr attr; 256 vm_offset_t vmaddr; 257 unsigned long file_offset; 258 unsigned long bss_size; 259 char *library; 260 ssize_t aresid; 261 int error, locked, writecount; 262 263 LCONVPATHEXIST(td, args->library, &library); 264 265#ifdef DEBUG 266 if (ldebug(uselib)) 267 printf(ARGS(uselib, "%s"), library); 268#endif 269 270 a_out = NULL; 271 locked = 0; 272 vp = NULL; 273 274 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 275 UIO_SYSSPACE, library, td); 276 error = namei(&ni); 277 LFREEPATH(library); 278 if (error) 279 goto cleanup; 280 281 vp = ni.ni_vp; 282 NDFREE(&ni, NDF_ONLY_PNBUF); 283 284 /* 285 * From here on down, we have a locked vnode that must be unlocked. 286 * XXX: The code below largely duplicates exec_check_permissions(). 287 */ 288 locked = 1; 289 290 /* Writable? */ 291 error = VOP_GET_WRITECOUNT(vp, &writecount); 292 if (error != 0) 293 goto cleanup; 294 if (writecount != 0) { 295 error = ETXTBSY; 296 goto cleanup; 297 } 298 299 /* Executable? */ 300 error = VOP_GETATTR(vp, &attr, td->td_ucred); 301 if (error) 302 goto cleanup; 303 304 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 305 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 306 /* EACCESS is what exec(2) returns. */ 307 error = ENOEXEC; 308 goto cleanup; 309 } 310 311 /* Sensible size? */ 312 if (attr.va_size == 0) { 313 error = ENOEXEC; 314 goto cleanup; 315 } 316 317 /* Can we access it? */ 318 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 319 if (error) 320 goto cleanup; 321 322 /* 323 * XXX: This should use vn_open() so that it is properly authorized, 324 * and to reduce code redundancy all over the place here. 325 * XXX: Not really, it duplicates far more of exec_check_permissions() 326 * than vn_open(). 327 */ 328#ifdef MAC 329 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 330 if (error) 331 goto cleanup; 332#endif 333 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 334 if (error) 335 goto cleanup; 336 337 /* Pull in executable header into exec_map */ 338 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 339 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 340 if (error) 341 goto cleanup; 342 343 /* Is it a Linux binary ? */ 344 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 345 error = ENOEXEC; 346 goto cleanup; 347 } 348 349 /* 350 * While we are here, we should REALLY do some more checks 351 */ 352 353 /* Set file/virtual offset based on a.out variant. */ 354 switch ((int)(a_out->a_magic & 0xffff)) { 355 case 0413: /* ZMAGIC */ 356 file_offset = 1024; 357 break; 358 case 0314: /* QMAGIC */ 359 file_offset = 0; 360 break; 361 default: 362 error = ENOEXEC; 363 goto cleanup; 364 } 365 366 bss_size = round_page(a_out->a_bss); 367 368 /* Check various fields in header for validity/bounds. */ 369 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 370 error = ENOEXEC; 371 goto cleanup; 372 } 373 374 /* text + data can't exceed file size */ 375 if (a_out->a_data + a_out->a_text > attr.va_size) { 376 error = EFAULT; 377 goto cleanup; 378 } 379 380 /* 381 * text/data/bss must not exceed limits 382 * XXX - this is not complete. it should check current usage PLUS 383 * the resources needed by this library. 384 */ 385 PROC_LOCK(td->td_proc); 386 if (a_out->a_text > maxtsiz || 387 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) || 388 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 389 bss_size) != 0) { 390 PROC_UNLOCK(td->td_proc); 391 error = ENOMEM; 392 goto cleanup; 393 } 394 PROC_UNLOCK(td->td_proc); 395 396 /* 397 * Prevent more writers. 398 * XXX: Note that if any of the VM operations fail below we don't 399 * clear this flag. 400 */ 401 VOP_SET_TEXT(vp); 402 403 /* 404 * Lock no longer needed 405 */ 406 locked = 0; 407 VOP_UNLOCK(vp, 0); 408 409 /* 410 * Check if file_offset page aligned. Currently we cannot handle 411 * misalinged file offsets, and so we read in the entire image 412 * (what a waste). 413 */ 414 if (file_offset & PAGE_MASK) { 415#ifdef DEBUG 416 printf("uselib: Non page aligned binary %lu\n", file_offset); 417#endif 418 /* Map text+data read/write/execute */ 419 420 /* a_entry is the load address and is page aligned */ 421 vmaddr = trunc_page(a_out->a_entry); 422 423 /* get anon user mapping, read+write+execute */ 424 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 425 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 426 VM_PROT_ALL, VM_PROT_ALL, 0); 427 if (error) 428 goto cleanup; 429 430 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 431 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 432 td->td_ucred, NOCRED, &aresid, td); 433 if (error != 0) 434 goto cleanup; 435 if (aresid != 0) { 436 error = ENOEXEC; 437 goto cleanup; 438 } 439 } else { 440#ifdef DEBUG 441 printf("uselib: Page aligned binary %lu\n", file_offset); 442#endif 443 /* 444 * for QMAGIC, a_entry is 20 bytes beyond the load address 445 * to skip the executable header 446 */ 447 vmaddr = trunc_page(a_out->a_entry); 448 449 /* 450 * Map it all into the process's space as a single 451 * copy-on-write "data" segment. 452 */ 453 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 454 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 455 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 456 if (error) 457 goto cleanup; 458 } 459#ifdef DEBUG 460 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 461 ((long *)vmaddr)[1]); 462#endif 463 if (bss_size != 0) { 464 /* Calculate BSS start address */ 465 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 466 a_out->a_data; 467 468 /* allocate some 'anon' space */ 469 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 470 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 471 VM_PROT_ALL, 0); 472 if (error) 473 goto cleanup; 474 } 475 476cleanup: 477 /* Unlock vnode if needed */ 478 if (locked) 479 VOP_UNLOCK(vp, 0); 480 481 /* Release the temporary mapping. */ 482 if (a_out) 483 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 484 485 return (error); 486} 487 488#endif /* __i386__ */ 489 490int 491linux_select(struct thread *td, struct linux_select_args *args) 492{ 493 l_timeval ltv; 494 struct timeval tv0, tv1, utv, *tvp; 495 int error; 496 497#ifdef DEBUG 498 if (ldebug(select)) 499 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 500 (void *)args->readfds, (void *)args->writefds, 501 (void *)args->exceptfds, (void *)args->timeout); 502#endif 503 504 /* 505 * Store current time for computation of the amount of 506 * time left. 507 */ 508 if (args->timeout) { 509 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 510 goto select_out; 511 utv.tv_sec = ltv.tv_sec; 512 utv.tv_usec = ltv.tv_usec; 513#ifdef DEBUG 514 if (ldebug(select)) 515 printf(LMSG("incoming timeout (%jd/%ld)"), 516 (intmax_t)utv.tv_sec, utv.tv_usec); 517#endif 518 519 if (itimerfix(&utv)) { 520 /* 521 * The timeval was invalid. Convert it to something 522 * valid that will act as it does under Linux. 523 */ 524 utv.tv_sec += utv.tv_usec / 1000000; 525 utv.tv_usec %= 1000000; 526 if (utv.tv_usec < 0) { 527 utv.tv_sec -= 1; 528 utv.tv_usec += 1000000; 529 } 530 if (utv.tv_sec < 0) 531 timevalclear(&utv); 532 } 533 microtime(&tv0); 534 tvp = &utv; 535 } else 536 tvp = NULL; 537 538 error = kern_select(td, args->nfds, args->readfds, args->writefds, 539 args->exceptfds, tvp, LINUX_NFDBITS); 540 541#ifdef DEBUG 542 if (ldebug(select)) 543 printf(LMSG("real select returns %d"), error); 544#endif 545 if (error) 546 goto select_out; 547 548 if (args->timeout) { 549 if (td->td_retval[0]) { 550 /* 551 * Compute how much time was left of the timeout, 552 * by subtracting the current time and the time 553 * before we started the call, and subtracting 554 * that result from the user-supplied value. 555 */ 556 microtime(&tv1); 557 timevalsub(&tv1, &tv0); 558 timevalsub(&utv, &tv1); 559 if (utv.tv_sec < 0) 560 timevalclear(&utv); 561 } else 562 timevalclear(&utv); 563#ifdef DEBUG 564 if (ldebug(select)) 565 printf(LMSG("outgoing timeout (%jd/%ld)"), 566 (intmax_t)utv.tv_sec, utv.tv_usec); 567#endif 568 ltv.tv_sec = utv.tv_sec; 569 ltv.tv_usec = utv.tv_usec; 570 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 571 goto select_out; 572 } 573 574select_out: 575#ifdef DEBUG 576 if (ldebug(select)) 577 printf(LMSG("select_out -> %d"), error); 578#endif 579 return (error); 580} 581 582int 583linux_mremap(struct thread *td, struct linux_mremap_args *args) 584{ 585 struct munmap_args /* { 586 void *addr; 587 size_t len; 588 } */ bsd_args; 589 int error = 0; 590 591#ifdef DEBUG 592 if (ldebug(mremap)) 593 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 594 (void *)(uintptr_t)args->addr, 595 (unsigned long)args->old_len, 596 (unsigned long)args->new_len, 597 (unsigned long)args->flags); 598#endif 599 600 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 601 td->td_retval[0] = 0; 602 return (EINVAL); 603 } 604 605 /* 606 * Check for the page alignment. 607 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 608 */ 609 if (args->addr & PAGE_MASK) { 610 td->td_retval[0] = 0; 611 return (EINVAL); 612 } 613 614 args->new_len = round_page(args->new_len); 615 args->old_len = round_page(args->old_len); 616 617 if (args->new_len > args->old_len) { 618 td->td_retval[0] = 0; 619 return (ENOMEM); 620 } 621 622 if (args->new_len < args->old_len) { 623 bsd_args.addr = 624 (caddr_t)((uintptr_t)args->addr + args->new_len); 625 bsd_args.len = args->old_len - args->new_len; 626 error = sys_munmap(td, &bsd_args); 627 } 628 629 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 630 return (error); 631} 632 633#define LINUX_MS_ASYNC 0x0001 634#define LINUX_MS_INVALIDATE 0x0002 635#define LINUX_MS_SYNC 0x0004 636 637int 638linux_msync(struct thread *td, struct linux_msync_args *args) 639{ 640 struct msync_args bsd_args; 641 642 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 643 bsd_args.len = (uintptr_t)args->len; 644 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 645 646 return (sys_msync(td, &bsd_args)); 647} 648 649int 650linux_time(struct thread *td, struct linux_time_args *args) 651{ 652 struct timeval tv; 653 l_time_t tm; 654 int error; 655 656#ifdef DEBUG 657 if (ldebug(time)) 658 printf(ARGS(time, "*")); 659#endif 660 661 microtime(&tv); 662 tm = tv.tv_sec; 663 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 664 return (error); 665 td->td_retval[0] = tm; 666 return (0); 667} 668 669struct l_times_argv { 670 l_clock_t tms_utime; 671 l_clock_t tms_stime; 672 l_clock_t tms_cutime; 673 l_clock_t tms_cstime; 674}; 675 676 677/* 678 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 679 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 680 * auxiliary vector entry. 681 */ 682#define CLK_TCK 100 683 684#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 685#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 686 687#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 688 CONVNTCK(r) : CONVOTCK(r)) 689 690int 691linux_times(struct thread *td, struct linux_times_args *args) 692{ 693 struct timeval tv, utime, stime, cutime, cstime; 694 struct l_times_argv tms; 695 struct proc *p; 696 int error; 697 698#ifdef DEBUG 699 if (ldebug(times)) 700 printf(ARGS(times, "*")); 701#endif 702 703 if (args->buf != NULL) { 704 p = td->td_proc; 705 PROC_LOCK(p); 706 PROC_STATLOCK(p); 707 calcru(p, &utime, &stime); 708 PROC_STATUNLOCK(p); 709 calccru(p, &cutime, &cstime); 710 PROC_UNLOCK(p); 711 712 tms.tms_utime = CONVTCK(utime); 713 tms.tms_stime = CONVTCK(stime); 714 715 tms.tms_cutime = CONVTCK(cutime); 716 tms.tms_cstime = CONVTCK(cstime); 717 718 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 719 return (error); 720 } 721 722 microuptime(&tv); 723 td->td_retval[0] = (int)CONVTCK(tv); 724 return (0); 725} 726 727int 728linux_newuname(struct thread *td, struct linux_newuname_args *args) 729{ 730 struct l_new_utsname utsname; 731 char osname[LINUX_MAX_UTSNAME]; 732 char osrelease[LINUX_MAX_UTSNAME]; 733 char *p; 734 735#ifdef DEBUG 736 if (ldebug(newuname)) 737 printf(ARGS(newuname, "*")); 738#endif 739 740 linux_get_osname(td, osname); 741 linux_get_osrelease(td, osrelease); 742 743 bzero(&utsname, sizeof(utsname)); 744 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 745 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 746 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 747 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 748 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 749 for (p = utsname.version; *p != '\0'; ++p) 750 if (*p == '\n') { 751 *p = '\0'; 752 break; 753 } 754 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 755 756 return (copyout(&utsname, args->buf, sizeof(utsname))); 757} 758 759struct l_utimbuf { 760 l_time_t l_actime; 761 l_time_t l_modtime; 762}; 763 764int 765linux_utime(struct thread *td, struct linux_utime_args *args) 766{ 767 struct timeval tv[2], *tvp; 768 struct l_utimbuf lut; 769 char *fname; 770 int error; 771 772 LCONVPATHEXIST(td, args->fname, &fname); 773 774#ifdef DEBUG 775 if (ldebug(utime)) 776 printf(ARGS(utime, "%s, *"), fname); 777#endif 778 779 if (args->times) { 780 if ((error = copyin(args->times, &lut, sizeof lut))) { 781 LFREEPATH(fname); 782 return (error); 783 } 784 tv[0].tv_sec = lut.l_actime; 785 tv[0].tv_usec = 0; 786 tv[1].tv_sec = lut.l_modtime; 787 tv[1].tv_usec = 0; 788 tvp = tv; 789 } else 790 tvp = NULL; 791 792 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 793 LFREEPATH(fname); 794 return (error); 795} 796 797int 798linux_utimes(struct thread *td, struct linux_utimes_args *args) 799{ 800 l_timeval ltv[2]; 801 struct timeval tv[2], *tvp = NULL; 802 char *fname; 803 int error; 804 805 LCONVPATHEXIST(td, args->fname, &fname); 806 807#ifdef DEBUG 808 if (ldebug(utimes)) 809 printf(ARGS(utimes, "%s, *"), fname); 810#endif 811 812 if (args->tptr != NULL) { 813 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 814 LFREEPATH(fname); 815 return (error); 816 } 817 tv[0].tv_sec = ltv[0].tv_sec; 818 tv[0].tv_usec = ltv[0].tv_usec; 819 tv[1].tv_sec = ltv[1].tv_sec; 820 tv[1].tv_usec = ltv[1].tv_usec; 821 tvp = tv; 822 } 823 824 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 825 LFREEPATH(fname); 826 return (error); 827} 828 829static int 830linux_utimensat_nsec_valid(l_long nsec) 831{ 832 833 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW) 834 return (0); 835 if (nsec >= 0 && nsec <= 999999999) 836 return (0); 837 return (1); 838} 839 840int 841linux_utimensat(struct thread *td, struct linux_utimensat_args *args) 842{ 843 struct l_timespec l_times[2]; 844 struct timespec times[2], *timesp = NULL; 845 char *path = NULL; 846 int error, dfd, flags = 0; 847 848 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 849 850#ifdef DEBUG 851 if (ldebug(utimensat)) 852 printf(ARGS(utimensat, "%d, *"), dfd); 853#endif 854 855 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 856 return (EINVAL); 857 858 if (args->times != NULL) { 859 error = copyin(args->times, l_times, sizeof(l_times)); 860 if (error != 0) 861 return (error); 862 863 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 || 864 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0) 865 return (EINVAL); 866 867 times[0].tv_sec = l_times[0].tv_sec; 868 switch (l_times[0].tv_nsec) 869 { 870 case LINUX_UTIME_OMIT: 871 times[0].tv_nsec = UTIME_OMIT; 872 break; 873 case LINUX_UTIME_NOW: 874 times[0].tv_nsec = UTIME_NOW; 875 break; 876 default: 877 times[0].tv_nsec = l_times[0].tv_nsec; 878 } 879 880 times[1].tv_sec = l_times[1].tv_sec; 881 switch (l_times[1].tv_nsec) 882 { 883 case LINUX_UTIME_OMIT: 884 times[1].tv_nsec = UTIME_OMIT; 885 break; 886 case LINUX_UTIME_NOW: 887 times[1].tv_nsec = UTIME_NOW; 888 break; 889 default: 890 times[1].tv_nsec = l_times[1].tv_nsec; 891 break; 892 } 893 timesp = times; 894 } 895 896 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) 897 /* This breaks POSIX, but is what the Linux kernel does 898 * _on purpose_ (documented in the man page for utimensat(2)), 899 * so we must follow that behaviour. */ 900 return (0); 901 902 if (args->pathname != NULL) 903 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 904 else if (args->flags != 0) 905 return (EINVAL); 906 907 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 908 flags |= AT_SYMLINK_NOFOLLOW; 909 910 if (path == NULL) 911 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 912 else { 913 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 914 UIO_SYSSPACE, flags); 915 LFREEPATH(path); 916 } 917 918 return (error); 919} 920 921int 922linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 923{ 924 l_timeval ltv[2]; 925 struct timeval tv[2], *tvp = NULL; 926 char *fname; 927 int error, dfd; 928 929 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 930 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 931 932#ifdef DEBUG 933 if (ldebug(futimesat)) 934 printf(ARGS(futimesat, "%s, *"), fname); 935#endif 936 937 if (args->utimes != NULL) { 938 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 939 LFREEPATH(fname); 940 return (error); 941 } 942 tv[0].tv_sec = ltv[0].tv_sec; 943 tv[0].tv_usec = ltv[0].tv_usec; 944 tv[1].tv_sec = ltv[1].tv_sec; 945 tv[1].tv_usec = ltv[1].tv_usec; 946 tvp = tv; 947 } 948 949 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 950 LFREEPATH(fname); 951 return (error); 952} 953 954int 955linux_common_wait(struct thread *td, int pid, int *status, 956 int options, struct rusage *ru) 957{ 958 int error, tmpstat; 959 960 error = kern_wait(td, pid, &tmpstat, options, ru); 961 if (error) 962 return (error); 963 964 if (status) { 965 tmpstat &= 0xffff; 966 if (WIFSIGNALED(tmpstat)) 967 tmpstat = (tmpstat & 0xffffff80) | 968 bsd_to_linux_signal(WTERMSIG(tmpstat)); 969 else if (WIFSTOPPED(tmpstat)) 970 tmpstat = (tmpstat & 0xffff00ff) | 971 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 972 else if (WIFCONTINUED(tmpstat)) 973 tmpstat = 0xffff; 974 error = copyout(&tmpstat, status, sizeof(int)); 975 } 976 977 return (error); 978} 979 980#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 981int 982linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 983{ 984 struct linux_wait4_args wait4_args; 985 986#ifdef DEBUG 987 if (ldebug(waitpid)) 988 printf(ARGS(waitpid, "%d, %p, %d"), 989 args->pid, (void *)args->status, args->options); 990#endif 991 992 wait4_args.pid = args->pid; 993 wait4_args.status = args->status; 994 wait4_args.options = args->options; 995 wait4_args.rusage = NULL; 996 997 return (linux_wait4(td, &wait4_args)); 998} 999#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1000 1001int 1002linux_wait4(struct thread *td, struct linux_wait4_args *args) 1003{ 1004 int error, options; 1005 struct rusage ru, *rup; 1006 1007#ifdef DEBUG 1008 if (ldebug(wait4)) 1009 printf(ARGS(wait4, "%d, %p, %d, %p"), 1010 args->pid, (void *)args->status, args->options, 1011 (void *)args->rusage); 1012#endif 1013 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 1014 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 1015 return (EINVAL); 1016 1017 options = WEXITED; 1018 linux_to_bsd_waitopts(args->options, &options); 1019 1020 if (args->rusage != NULL) 1021 rup = &ru; 1022 else 1023 rup = NULL; 1024 error = linux_common_wait(td, args->pid, args->status, options, rup); 1025 if (error != 0) 1026 return (error); 1027 if (args->rusage != NULL) 1028 error = linux_copyout_rusage(&ru, args->rusage); 1029 return (error); 1030} 1031 1032int 1033linux_waitid(struct thread *td, struct linux_waitid_args *args) 1034{ 1035 int status, options, sig; 1036 struct __wrusage wru; 1037 siginfo_t siginfo; 1038 l_siginfo_t lsi; 1039 idtype_t idtype; 1040 struct proc *p; 1041 int error; 1042 1043 options = 0; 1044 linux_to_bsd_waitopts(args->options, &options); 1045 1046 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1047 return (EINVAL); 1048 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1049 return (EINVAL); 1050 1051 switch (args->idtype) { 1052 case LINUX_P_ALL: 1053 idtype = P_ALL; 1054 break; 1055 case LINUX_P_PID: 1056 if (args->id <= 0) 1057 return (EINVAL); 1058 idtype = P_PID; 1059 break; 1060 case LINUX_P_PGID: 1061 if (args->id <= 0) 1062 return (EINVAL); 1063 idtype = P_PGID; 1064 break; 1065 default: 1066 return (EINVAL); 1067 } 1068 1069 error = kern_wait6(td, idtype, args->id, &status, options, 1070 &wru, &siginfo); 1071 if (error != 0) 1072 return (error); 1073 if (args->rusage != NULL) { 1074 error = linux_copyout_rusage(&wru.wru_children, 1075 args->rusage); 1076 if (error != 0) 1077 return (error); 1078 } 1079 if (args->info != NULL) { 1080 p = td->td_proc; 1081 if (td->td_retval[0] == 0) 1082 bzero(&lsi, sizeof(lsi)); 1083 else { 1084 sig = bsd_to_linux_signal(siginfo.si_signo); 1085 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1086 } 1087 error = copyout(&lsi, args->info, sizeof(lsi)); 1088 } 1089 td->td_retval[0] = 0; 1090 1091 return (error); 1092} 1093 1094int 1095linux_mknod(struct thread *td, struct linux_mknod_args *args) 1096{ 1097 char *path; 1098 int error; 1099 1100 LCONVPATHCREAT(td, args->path, &path); 1101 1102#ifdef DEBUG 1103 if (ldebug(mknod)) 1104 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1105 (uintmax_t)args->dev); 1106#endif 1107 1108 switch (args->mode & S_IFMT) { 1109 case S_IFIFO: 1110 case S_IFSOCK: 1111 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1112 break; 1113 1114 case S_IFCHR: 1115 case S_IFBLK: 1116 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1117 args->dev); 1118 break; 1119 1120 case S_IFDIR: 1121 error = EPERM; 1122 break; 1123 1124 case 0: 1125 args->mode |= S_IFREG; 1126 /* FALLTHROUGH */ 1127 case S_IFREG: 1128 error = kern_open(td, path, UIO_SYSSPACE, 1129 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1130 if (error == 0) 1131 kern_close(td, td->td_retval[0]); 1132 break; 1133 1134 default: 1135 error = EINVAL; 1136 break; 1137 } 1138 LFREEPATH(path); 1139 return (error); 1140} 1141 1142int 1143linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1144{ 1145 char *path; 1146 int error, dfd; 1147 1148 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1149 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1150 1151#ifdef DEBUG 1152 if (ldebug(mknodat)) 1153 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1154#endif 1155 1156 switch (args->mode & S_IFMT) { 1157 case S_IFIFO: 1158 case S_IFSOCK: 1159 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1160 break; 1161 1162 case S_IFCHR: 1163 case S_IFBLK: 1164 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1165 args->dev); 1166 break; 1167 1168 case S_IFDIR: 1169 error = EPERM; 1170 break; 1171 1172 case 0: 1173 args->mode |= S_IFREG; 1174 /* FALLTHROUGH */ 1175 case S_IFREG: 1176 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1177 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1178 if (error == 0) 1179 kern_close(td, td->td_retval[0]); 1180 break; 1181 1182 default: 1183 error = EINVAL; 1184 break; 1185 } 1186 LFREEPATH(path); 1187 return (error); 1188} 1189 1190/* 1191 * UGH! This is just about the dumbest idea I've ever heard!! 1192 */ 1193int 1194linux_personality(struct thread *td, struct linux_personality_args *args) 1195{ 1196#ifdef DEBUG 1197 if (ldebug(personality)) 1198 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1199#endif 1200 if (args->per != 0) 1201 return (EINVAL); 1202 1203 /* Yes Jim, it's still a Linux... */ 1204 td->td_retval[0] = 0; 1205 return (0); 1206} 1207 1208struct l_itimerval { 1209 l_timeval it_interval; 1210 l_timeval it_value; 1211}; 1212 1213#define B2L_ITIMERVAL(bip, lip) \ 1214 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1215 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1216 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1217 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1218 1219int 1220linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1221{ 1222 int error; 1223 struct l_itimerval ls; 1224 struct itimerval aitv, oitv; 1225 1226#ifdef DEBUG 1227 if (ldebug(setitimer)) 1228 printf(ARGS(setitimer, "%p, %p"), 1229 (void *)uap->itv, (void *)uap->oitv); 1230#endif 1231 1232 if (uap->itv == NULL) { 1233 uap->itv = uap->oitv; 1234 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1235 } 1236 1237 error = copyin(uap->itv, &ls, sizeof(ls)); 1238 if (error != 0) 1239 return (error); 1240 B2L_ITIMERVAL(&aitv, &ls); 1241#ifdef DEBUG 1242 if (ldebug(setitimer)) { 1243 printf("setitimer: value: sec: %jd, usec: %ld\n", 1244 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1245 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1246 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1247 } 1248#endif 1249 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1250 if (error != 0 || uap->oitv == NULL) 1251 return (error); 1252 B2L_ITIMERVAL(&ls, &oitv); 1253 1254 return (copyout(&ls, uap->oitv, sizeof(ls))); 1255} 1256 1257int 1258linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1259{ 1260 int error; 1261 struct l_itimerval ls; 1262 struct itimerval aitv; 1263 1264#ifdef DEBUG 1265 if (ldebug(getitimer)) 1266 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1267#endif 1268 error = kern_getitimer(td, uap->which, &aitv); 1269 if (error != 0) 1270 return (error); 1271 B2L_ITIMERVAL(&ls, &aitv); 1272 return (copyout(&ls, uap->itv, sizeof(ls))); 1273} 1274 1275#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1276int 1277linux_nice(struct thread *td, struct linux_nice_args *args) 1278{ 1279 struct setpriority_args bsd_args; 1280 1281 bsd_args.which = PRIO_PROCESS; 1282 bsd_args.who = 0; /* current process */ 1283 bsd_args.prio = args->inc; 1284 return (sys_setpriority(td, &bsd_args)); 1285} 1286#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1287 1288int 1289linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1290{ 1291 struct ucred *newcred, *oldcred; 1292 l_gid_t *linux_gidset; 1293 gid_t *bsd_gidset; 1294 int ngrp, error; 1295 struct proc *p; 1296 1297 ngrp = args->gidsetsize; 1298 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1299 return (EINVAL); 1300 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1301 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1302 if (error) 1303 goto out; 1304 newcred = crget(); 1305 p = td->td_proc; 1306 PROC_LOCK(p); 1307 oldcred = crcopysafe(p, newcred); 1308 1309 /* 1310 * cr_groups[0] holds egid. Setting the whole set from 1311 * the supplied set will cause egid to be changed too. 1312 * Keep cr_groups[0] unchanged to prevent that. 1313 */ 1314 1315 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1316 PROC_UNLOCK(p); 1317 crfree(newcred); 1318 goto out; 1319 } 1320 1321 if (ngrp > 0) { 1322 newcred->cr_ngroups = ngrp + 1; 1323 1324 bsd_gidset = newcred->cr_groups; 1325 ngrp--; 1326 while (ngrp >= 0) { 1327 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1328 ngrp--; 1329 } 1330 } else 1331 newcred->cr_ngroups = 1; 1332 1333 setsugid(p); 1334 p->p_ucred = newcred; 1335 PROC_UNLOCK(p); 1336 crfree(oldcred); 1337 error = 0; 1338out: 1339 free(linux_gidset, M_LINUX); 1340 return (error); 1341} 1342 1343int 1344linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1345{ 1346 struct ucred *cred; 1347 l_gid_t *linux_gidset; 1348 gid_t *bsd_gidset; 1349 int bsd_gidsetsz, ngrp, error; 1350 1351 cred = td->td_ucred; 1352 bsd_gidset = cred->cr_groups; 1353 bsd_gidsetsz = cred->cr_ngroups - 1; 1354 1355 /* 1356 * cr_groups[0] holds egid. Returning the whole set 1357 * here will cause a duplicate. Exclude cr_groups[0] 1358 * to prevent that. 1359 */ 1360 1361 if ((ngrp = args->gidsetsize) == 0) { 1362 td->td_retval[0] = bsd_gidsetsz; 1363 return (0); 1364 } 1365 1366 if (ngrp < bsd_gidsetsz) 1367 return (EINVAL); 1368 1369 ngrp = 0; 1370 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1371 M_LINUX, M_WAITOK); 1372 while (ngrp < bsd_gidsetsz) { 1373 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1374 ngrp++; 1375 } 1376 1377 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1378 free(linux_gidset, M_LINUX); 1379 if (error) 1380 return (error); 1381 1382 td->td_retval[0] = ngrp; 1383 return (0); 1384} 1385 1386int 1387linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1388{ 1389 struct rlimit bsd_rlim; 1390 struct l_rlimit rlim; 1391 u_int which; 1392 int error; 1393 1394#ifdef DEBUG 1395 if (ldebug(setrlimit)) 1396 printf(ARGS(setrlimit, "%d, %p"), 1397 args->resource, (void *)args->rlim); 1398#endif 1399 1400 if (args->resource >= LINUX_RLIM_NLIMITS) 1401 return (EINVAL); 1402 1403 which = linux_to_bsd_resource[args->resource]; 1404 if (which == -1) 1405 return (EINVAL); 1406 1407 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1408 if (error) 1409 return (error); 1410 1411 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1412 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1413 return (kern_setrlimit(td, which, &bsd_rlim)); 1414} 1415 1416#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1417int 1418linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1419{ 1420 struct l_rlimit rlim; 1421 struct proc *p = td->td_proc; 1422 struct rlimit bsd_rlim; 1423 u_int which; 1424 1425#ifdef DEBUG 1426 if (ldebug(old_getrlimit)) 1427 printf(ARGS(old_getrlimit, "%d, %p"), 1428 args->resource, (void *)args->rlim); 1429#endif 1430 1431 if (args->resource >= LINUX_RLIM_NLIMITS) 1432 return (EINVAL); 1433 1434 which = linux_to_bsd_resource[args->resource]; 1435 if (which == -1) 1436 return (EINVAL); 1437 1438 PROC_LOCK(p); 1439 lim_rlimit(p, which, &bsd_rlim); 1440 PROC_UNLOCK(p); 1441 1442#ifdef COMPAT_LINUX32 1443 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1444 if (rlim.rlim_cur == UINT_MAX) 1445 rlim.rlim_cur = INT_MAX; 1446 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1447 if (rlim.rlim_max == UINT_MAX) 1448 rlim.rlim_max = INT_MAX; 1449#else 1450 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1451 if (rlim.rlim_cur == ULONG_MAX) 1452 rlim.rlim_cur = LONG_MAX; 1453 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1454 if (rlim.rlim_max == ULONG_MAX) 1455 rlim.rlim_max = LONG_MAX; 1456#endif 1457 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1458} 1459#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1460 1461int 1462linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1463{ 1464 struct l_rlimit rlim; 1465 struct proc *p = td->td_proc; 1466 struct rlimit bsd_rlim; 1467 u_int which; 1468 1469#ifdef DEBUG 1470 if (ldebug(getrlimit)) 1471 printf(ARGS(getrlimit, "%d, %p"), 1472 args->resource, (void *)args->rlim); 1473#endif 1474 1475 if (args->resource >= LINUX_RLIM_NLIMITS) 1476 return (EINVAL); 1477 1478 which = linux_to_bsd_resource[args->resource]; 1479 if (which == -1) 1480 return (EINVAL); 1481 1482 PROC_LOCK(p); 1483 lim_rlimit(p, which, &bsd_rlim); 1484 PROC_UNLOCK(p); 1485 1486 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1487 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1488 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1489} 1490 1491int 1492linux_sched_setscheduler(struct thread *td, 1493 struct linux_sched_setscheduler_args *args) 1494{ 1495 struct sched_param sched_param; 1496 struct thread *tdt; 1497 int error, policy; 1498 1499#ifdef DEBUG 1500 if (ldebug(sched_setscheduler)) 1501 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1502 args->pid, args->policy, (const void *)args->param); 1503#endif 1504 1505 switch (args->policy) { 1506 case LINUX_SCHED_OTHER: 1507 policy = SCHED_OTHER; 1508 break; 1509 case LINUX_SCHED_FIFO: 1510 policy = SCHED_FIFO; 1511 break; 1512 case LINUX_SCHED_RR: 1513 policy = SCHED_RR; 1514 break; 1515 default: 1516 return (EINVAL); 1517 } 1518 1519 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1520 if (error) 1521 return (error); 1522 1523 tdt = linux_tdfind(td, args->pid, -1); 1524 if (tdt == NULL) 1525 return (ESRCH); 1526 1527 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1528 PROC_UNLOCK(tdt->td_proc); 1529 return (error); 1530} 1531 1532int 1533linux_sched_getscheduler(struct thread *td, 1534 struct linux_sched_getscheduler_args *args) 1535{ 1536 struct thread *tdt; 1537 int error, policy; 1538 1539#ifdef DEBUG 1540 if (ldebug(sched_getscheduler)) 1541 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1542#endif 1543 1544 tdt = linux_tdfind(td, args->pid, -1); 1545 if (tdt == NULL) 1546 return (ESRCH); 1547 1548 error = kern_sched_getscheduler(td, tdt, &policy); 1549 PROC_UNLOCK(tdt->td_proc); 1550 1551 switch (policy) { 1552 case SCHED_OTHER: 1553 td->td_retval[0] = LINUX_SCHED_OTHER; 1554 break; 1555 case SCHED_FIFO: 1556 td->td_retval[0] = LINUX_SCHED_FIFO; 1557 break; 1558 case SCHED_RR: 1559 td->td_retval[0] = LINUX_SCHED_RR; 1560 break; 1561 } 1562 return (error); 1563} 1564 1565int 1566linux_sched_get_priority_max(struct thread *td, 1567 struct linux_sched_get_priority_max_args *args) 1568{ 1569 struct sched_get_priority_max_args bsd; 1570 1571#ifdef DEBUG 1572 if (ldebug(sched_get_priority_max)) 1573 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1574#endif 1575 1576 switch (args->policy) { 1577 case LINUX_SCHED_OTHER: 1578 bsd.policy = SCHED_OTHER; 1579 break; 1580 case LINUX_SCHED_FIFO: 1581 bsd.policy = SCHED_FIFO; 1582 break; 1583 case LINUX_SCHED_RR: 1584 bsd.policy = SCHED_RR; 1585 break; 1586 default: 1587 return (EINVAL); 1588 } 1589 return (sys_sched_get_priority_max(td, &bsd)); 1590} 1591 1592int 1593linux_sched_get_priority_min(struct thread *td, 1594 struct linux_sched_get_priority_min_args *args) 1595{ 1596 struct sched_get_priority_min_args bsd; 1597 1598#ifdef DEBUG 1599 if (ldebug(sched_get_priority_min)) 1600 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1601#endif 1602 1603 switch (args->policy) { 1604 case LINUX_SCHED_OTHER: 1605 bsd.policy = SCHED_OTHER; 1606 break; 1607 case LINUX_SCHED_FIFO: 1608 bsd.policy = SCHED_FIFO; 1609 break; 1610 case LINUX_SCHED_RR: 1611 bsd.policy = SCHED_RR; 1612 break; 1613 default: 1614 return (EINVAL); 1615 } 1616 return (sys_sched_get_priority_min(td, &bsd)); 1617} 1618 1619#define REBOOT_CAD_ON 0x89abcdef 1620#define REBOOT_CAD_OFF 0 1621#define REBOOT_HALT 0xcdef0123 1622#define REBOOT_RESTART 0x01234567 1623#define REBOOT_RESTART2 0xA1B2C3D4 1624#define REBOOT_POWEROFF 0x4321FEDC 1625#define REBOOT_MAGIC1 0xfee1dead 1626#define REBOOT_MAGIC2 0x28121969 1627#define REBOOT_MAGIC2A 0x05121996 1628#define REBOOT_MAGIC2B 0x16041998 1629 1630int 1631linux_reboot(struct thread *td, struct linux_reboot_args *args) 1632{ 1633 struct reboot_args bsd_args; 1634 1635#ifdef DEBUG 1636 if (ldebug(reboot)) 1637 printf(ARGS(reboot, "0x%x"), args->cmd); 1638#endif 1639 1640 if (args->magic1 != REBOOT_MAGIC1) 1641 return (EINVAL); 1642 1643 switch (args->magic2) { 1644 case REBOOT_MAGIC2: 1645 case REBOOT_MAGIC2A: 1646 case REBOOT_MAGIC2B: 1647 break; 1648 default: 1649 return (EINVAL); 1650 } 1651 1652 switch (args->cmd) { 1653 case REBOOT_CAD_ON: 1654 case REBOOT_CAD_OFF: 1655 return (priv_check(td, PRIV_REBOOT)); 1656 case REBOOT_HALT: 1657 bsd_args.opt = RB_HALT; 1658 break; 1659 case REBOOT_RESTART: 1660 case REBOOT_RESTART2: 1661 bsd_args.opt = 0; 1662 break; 1663 case REBOOT_POWEROFF: 1664 bsd_args.opt = RB_POWEROFF; 1665 break; 1666 default: 1667 return (EINVAL); 1668 } 1669 return (sys_reboot(td, &bsd_args)); 1670} 1671 1672 1673/* 1674 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1675 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1676 * are assumed to be preserved. The following lightweight syscalls fixes 1677 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1678 * 1679 * linux_getpid() - MP SAFE 1680 * linux_getgid() - MP SAFE 1681 * linux_getuid() - MP SAFE 1682 */ 1683 1684int 1685linux_getpid(struct thread *td, struct linux_getpid_args *args) 1686{ 1687 1688#ifdef DEBUG 1689 if (ldebug(getpid)) 1690 printf(ARGS(getpid, "")); 1691#endif 1692 td->td_retval[0] = td->td_proc->p_pid; 1693 1694 return (0); 1695} 1696 1697int 1698linux_gettid(struct thread *td, struct linux_gettid_args *args) 1699{ 1700 struct linux_emuldata *em; 1701 1702#ifdef DEBUG 1703 if (ldebug(gettid)) 1704 printf(ARGS(gettid, "")); 1705#endif 1706 1707 em = em_find(td); 1708 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1709 1710 td->td_retval[0] = em->em_tid; 1711 1712 return (0); 1713} 1714 1715 1716int 1717linux_getppid(struct thread *td, struct linux_getppid_args *args) 1718{ 1719 1720#ifdef DEBUG 1721 if (ldebug(getppid)) 1722 printf(ARGS(getppid, "")); 1723#endif 1724 1725 PROC_LOCK(td->td_proc); 1726 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1727 PROC_UNLOCK(td->td_proc); 1728 return (0); 1729} 1730 1731int 1732linux_getgid(struct thread *td, struct linux_getgid_args *args) 1733{ 1734 1735#ifdef DEBUG 1736 if (ldebug(getgid)) 1737 printf(ARGS(getgid, "")); 1738#endif 1739 1740 td->td_retval[0] = td->td_ucred->cr_rgid; 1741 return (0); 1742} 1743 1744int 1745linux_getuid(struct thread *td, struct linux_getuid_args *args) 1746{ 1747 1748#ifdef DEBUG 1749 if (ldebug(getuid)) 1750 printf(ARGS(getuid, "")); 1751#endif 1752 1753 td->td_retval[0] = td->td_ucred->cr_ruid; 1754 return (0); 1755} 1756 1757 1758int 1759linux_getsid(struct thread *td, struct linux_getsid_args *args) 1760{ 1761 struct getsid_args bsd; 1762 1763#ifdef DEBUG 1764 if (ldebug(getsid)) 1765 printf(ARGS(getsid, "%i"), args->pid); 1766#endif 1767 1768 bsd.pid = args->pid; 1769 return (sys_getsid(td, &bsd)); 1770} 1771 1772int 1773linux_nosys(struct thread *td, struct nosys_args *ignore) 1774{ 1775 1776 return (ENOSYS); 1777} 1778 1779int 1780linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1781{ 1782 struct getpriority_args bsd_args; 1783 int error; 1784 1785#ifdef DEBUG 1786 if (ldebug(getpriority)) 1787 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1788#endif 1789 1790 bsd_args.which = args->which; 1791 bsd_args.who = args->who; 1792 error = sys_getpriority(td, &bsd_args); 1793 td->td_retval[0] = 20 - td->td_retval[0]; 1794 return (error); 1795} 1796 1797int 1798linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1799{ 1800 int name[2]; 1801 1802#ifdef DEBUG 1803 if (ldebug(sethostname)) 1804 printf(ARGS(sethostname, "*, %i"), args->len); 1805#endif 1806 1807 name[0] = CTL_KERN; 1808 name[1] = KERN_HOSTNAME; 1809 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1810 args->len, 0, 0)); 1811} 1812 1813int 1814linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1815{ 1816 int name[2]; 1817 1818#ifdef DEBUG 1819 if (ldebug(setdomainname)) 1820 printf(ARGS(setdomainname, "*, %i"), args->len); 1821#endif 1822 1823 name[0] = CTL_KERN; 1824 name[1] = KERN_NISDOMAINNAME; 1825 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1826 args->len, 0, 0)); 1827} 1828 1829int 1830linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1831{ 1832 1833#ifdef DEBUG 1834 if (ldebug(exit_group)) 1835 printf(ARGS(exit_group, "%i"), args->error_code); 1836#endif 1837 1838 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1839 args->error_code); 1840 1841 /* 1842 * XXX: we should send a signal to the parent if 1843 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1844 * as it doesnt occur often. 1845 */ 1846 exit1(td, W_EXITCODE(args->error_code, 0)); 1847 /* NOTREACHED */ 1848} 1849 1850#define _LINUX_CAPABILITY_VERSION 0x19980330 1851 1852struct l_user_cap_header { 1853 l_int version; 1854 l_int pid; 1855}; 1856 1857struct l_user_cap_data { 1858 l_int effective; 1859 l_int permitted; 1860 l_int inheritable; 1861}; 1862 1863int 1864linux_capget(struct thread *td, struct linux_capget_args *args) 1865{ 1866 struct l_user_cap_header luch; 1867 struct l_user_cap_data lucd; 1868 int error; 1869 1870 if (args->hdrp == NULL) 1871 return (EFAULT); 1872 1873 error = copyin(args->hdrp, &luch, sizeof(luch)); 1874 if (error != 0) 1875 return (error); 1876 1877 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1878 luch.version = _LINUX_CAPABILITY_VERSION; 1879 error = copyout(&luch, args->hdrp, sizeof(luch)); 1880 if (error) 1881 return (error); 1882 return (EINVAL); 1883 } 1884 1885 if (luch.pid) 1886 return (EPERM); 1887 1888 if (args->datap) { 1889 /* 1890 * The current implementation doesn't support setting 1891 * a capability (it's essentially a stub) so indicate 1892 * that no capabilities are currently set or available 1893 * to request. 1894 */ 1895 bzero (&lucd, sizeof(lucd)); 1896 error = copyout(&lucd, args->datap, sizeof(lucd)); 1897 } 1898 1899 return (error); 1900} 1901 1902int 1903linux_capset(struct thread *td, struct linux_capset_args *args) 1904{ 1905 struct l_user_cap_header luch; 1906 struct l_user_cap_data lucd; 1907 int error; 1908 1909 if (args->hdrp == NULL || args->datap == NULL) 1910 return (EFAULT); 1911 1912 error = copyin(args->hdrp, &luch, sizeof(luch)); 1913 if (error != 0) 1914 return (error); 1915 1916 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1917 luch.version = _LINUX_CAPABILITY_VERSION; 1918 error = copyout(&luch, args->hdrp, sizeof(luch)); 1919 if (error) 1920 return (error); 1921 return (EINVAL); 1922 } 1923 1924 if (luch.pid) 1925 return (EPERM); 1926 1927 error = copyin(args->datap, &lucd, sizeof(lucd)); 1928 if (error != 0) 1929 return (error); 1930 1931 /* We currently don't support setting any capabilities. */ 1932 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1933 linux_msg(td, 1934 "capset effective=0x%x, permitted=0x%x, " 1935 "inheritable=0x%x is not implemented", 1936 (int)lucd.effective, (int)lucd.permitted, 1937 (int)lucd.inheritable); 1938 return (EPERM); 1939 } 1940 1941 return (0); 1942} 1943 1944int 1945linux_prctl(struct thread *td, struct linux_prctl_args *args) 1946{ 1947 int error = 0, max_size; 1948 struct proc *p = td->td_proc; 1949 char comm[LINUX_MAX_COMM_LEN]; 1950 struct linux_emuldata *em; 1951 int pdeath_signal; 1952 1953#ifdef DEBUG 1954 if (ldebug(prctl)) 1955 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1956 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1957 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1958#endif 1959 1960 switch (args->option) { 1961 case LINUX_PR_SET_PDEATHSIG: 1962 if (!LINUX_SIG_VALID(args->arg2)) 1963 return (EINVAL); 1964 em = em_find(td); 1965 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1966 em->pdeath_signal = args->arg2; 1967 break; 1968 case LINUX_PR_GET_PDEATHSIG: 1969 em = em_find(td); 1970 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1971 pdeath_signal = em->pdeath_signal; 1972 error = copyout(&pdeath_signal, 1973 (void *)(register_t)args->arg2, 1974 sizeof(pdeath_signal)); 1975 break; 1976 case LINUX_PR_GET_KEEPCAPS: 1977 /* 1978 * Indicate that we always clear the effective and 1979 * permitted capability sets when the user id becomes 1980 * non-zero (actually the capability sets are simply 1981 * always zero in the current implementation). 1982 */ 1983 td->td_retval[0] = 0; 1984 break; 1985 case LINUX_PR_SET_KEEPCAPS: 1986 /* 1987 * Ignore requests to keep the effective and permitted 1988 * capability sets when the user id becomes non-zero. 1989 */ 1990 break; 1991 case LINUX_PR_SET_NAME: 1992 /* 1993 * To be on the safe side we need to make sure to not 1994 * overflow the size a linux program expects. We already 1995 * do this here in the copyin, so that we don't need to 1996 * check on copyout. 1997 */ 1998 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1999 error = copyinstr((void *)(register_t)args->arg2, comm, 2000 max_size, NULL); 2001 2002 /* Linux silently truncates the name if it is too long. */ 2003 if (error == ENAMETOOLONG) { 2004 /* 2005 * XXX: copyinstr() isn't documented to populate the 2006 * array completely, so do a copyin() to be on the 2007 * safe side. This should be changed in case 2008 * copyinstr() is changed to guarantee this. 2009 */ 2010 error = copyin((void *)(register_t)args->arg2, comm, 2011 max_size - 1); 2012 comm[max_size - 1] = '\0'; 2013 } 2014 if (error) 2015 return (error); 2016 2017 PROC_LOCK(p); 2018 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 2019 PROC_UNLOCK(p); 2020 break; 2021 case LINUX_PR_GET_NAME: 2022 PROC_LOCK(p); 2023 strlcpy(comm, p->p_comm, sizeof(comm)); 2024 PROC_UNLOCK(p); 2025 error = copyout(comm, (void *)(register_t)args->arg2, 2026 strlen(comm) + 1); 2027 break; 2028 default: 2029 error = EINVAL; 2030 break; 2031 } 2032 2033 return (error); 2034} 2035 2036int 2037linux_sched_setparam(struct thread *td, 2038 struct linux_sched_setparam_args *uap) 2039{ 2040 struct sched_param sched_param; 2041 struct thread *tdt; 2042 int error; 2043 2044#ifdef DEBUG 2045 if (ldebug(sched_setparam)) 2046 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 2047#endif 2048 2049 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 2050 if (error) 2051 return (error); 2052 2053 tdt = linux_tdfind(td, uap->pid, -1); 2054 if (tdt == NULL) 2055 return (ESRCH); 2056 2057 error = kern_sched_setparam(td, tdt, &sched_param); 2058 PROC_UNLOCK(tdt->td_proc); 2059 return (error); 2060} 2061 2062int 2063linux_sched_getparam(struct thread *td, 2064 struct linux_sched_getparam_args *uap) 2065{ 2066 struct sched_param sched_param; 2067 struct thread *tdt; 2068 int error; 2069 2070#ifdef DEBUG 2071 if (ldebug(sched_getparam)) 2072 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 2073#endif 2074 2075 tdt = linux_tdfind(td, uap->pid, -1); 2076 if (tdt == NULL) 2077 return (ESRCH); 2078 2079 error = kern_sched_getparam(td, tdt, &sched_param); 2080 PROC_UNLOCK(tdt->td_proc); 2081 if (error == 0) 2082 error = copyout(&sched_param, uap->param, 2083 sizeof(sched_param)); 2084 return (error); 2085} 2086 2087/* 2088 * Get affinity of a process. 2089 */ 2090int 2091linux_sched_getaffinity(struct thread *td, 2092 struct linux_sched_getaffinity_args *args) 2093{ 2094 int error; 2095 struct thread *tdt; 2096 struct cpuset_getaffinity_args cga; 2097 2098#ifdef DEBUG 2099 if (ldebug(sched_getaffinity)) 2100 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2101 args->len); 2102#endif 2103 if (args->len < sizeof(cpuset_t)) 2104 return (EINVAL); 2105 2106 tdt = linux_tdfind(td, args->pid, -1); 2107 if (tdt == NULL) 2108 return (ESRCH); 2109 2110 PROC_UNLOCK(tdt->td_proc); 2111 cga.level = CPU_LEVEL_WHICH; 2112 cga.which = CPU_WHICH_TID; 2113 cga.id = tdt->td_tid; 2114 cga.cpusetsize = sizeof(cpuset_t); 2115 cga.mask = (cpuset_t *) args->user_mask_ptr; 2116 2117 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2118 td->td_retval[0] = sizeof(cpuset_t); 2119 2120 return (error); 2121} 2122 2123/* 2124 * Set affinity of a process. 2125 */ 2126int 2127linux_sched_setaffinity(struct thread *td, 2128 struct linux_sched_setaffinity_args *args) 2129{ 2130 struct cpuset_setaffinity_args csa; 2131 struct thread *tdt; 2132 2133#ifdef DEBUG 2134 if (ldebug(sched_setaffinity)) 2135 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2136 args->len); 2137#endif 2138 if (args->len < sizeof(cpuset_t)) 2139 return (EINVAL); 2140 2141 tdt = linux_tdfind(td, args->pid, -1); 2142 if (tdt == NULL) 2143 return (ESRCH); 2144 2145 PROC_UNLOCK(tdt->td_proc); 2146 csa.level = CPU_LEVEL_WHICH; 2147 csa.which = CPU_WHICH_TID; 2148 csa.id = tdt->td_tid; 2149 csa.cpusetsize = sizeof(cpuset_t); 2150 csa.mask = (cpuset_t *) args->user_mask_ptr; 2151 2152 return (sys_cpuset_setaffinity(td, &csa)); 2153} 2154 2155struct linux_rlimit64 { 2156 uint64_t rlim_cur; 2157 uint64_t rlim_max; 2158}; 2159 2160int 2161linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2162{ 2163 struct rlimit rlim, nrlim; 2164 struct linux_rlimit64 lrlim; 2165 struct proc *p; 2166 u_int which; 2167 int flags; 2168 int error; 2169 2170#ifdef DEBUG 2171 if (ldebug(prlimit64)) 2172 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2173 args->resource, (void *)args->new, (void *)args->old); 2174#endif 2175 2176 if (args->resource >= LINUX_RLIM_NLIMITS) 2177 return (EINVAL); 2178 2179 which = linux_to_bsd_resource[args->resource]; 2180 if (which == -1) 2181 return (EINVAL); 2182 2183 if (args->new != NULL) { 2184 /* 2185 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2186 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2187 * as INFINITY so we do not need a conversion even. 2188 */ 2189 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2190 if (error != 0) 2191 return (error); 2192 } 2193 2194 flags = PGET_HOLD | PGET_NOTWEXIT; 2195 if (args->new != NULL) 2196 flags |= PGET_CANDEBUG; 2197 else 2198 flags |= PGET_CANSEE; 2199 error = pget(args->pid, flags, &p); 2200 if (error != 0) 2201 return (error); 2202 2203 if (args->old != NULL) { 2204 PROC_LOCK(p); 2205 lim_rlimit(p, which, &rlim); 2206 PROC_UNLOCK(p); 2207 if (rlim.rlim_cur == RLIM_INFINITY) 2208 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2209 else 2210 lrlim.rlim_cur = rlim.rlim_cur; 2211 if (rlim.rlim_max == RLIM_INFINITY) 2212 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2213 else 2214 lrlim.rlim_max = rlim.rlim_max; 2215 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2216 if (error != 0) 2217 goto out; 2218 } 2219 2220 if (args->new != NULL) 2221 error = kern_proc_setrlimit(td, p, which, &nrlim); 2222 2223 out: 2224 PRELE(p); 2225 return (error); 2226} 2227 2228int 2229linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2230{ 2231 struct timeval utv, tv0, tv1, *tvp; 2232 struct l_pselect6arg lpse6; 2233 struct l_timespec lts; 2234 struct timespec uts; 2235 l_sigset_t l_ss; 2236 sigset_t *ssp; 2237 sigset_t ss; 2238 int error; 2239 2240 ssp = NULL; 2241 if (args->sig != NULL) { 2242 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2243 if (error != 0) 2244 return (error); 2245 if (lpse6.ss_len != sizeof(l_ss)) 2246 return (EINVAL); 2247 if (lpse6.ss != 0) { 2248 error = copyin(PTRIN(lpse6.ss), &l_ss, 2249 sizeof(l_ss)); 2250 if (error != 0) 2251 return (error); 2252 linux_to_bsd_sigset(&l_ss, &ss); 2253 ssp = &ss; 2254 } 2255 } 2256 2257 /* 2258 * Currently glibc changes nanosecond number to microsecond. 2259 * This mean losing precision but for now it is hardly seen. 2260 */ 2261 if (args->tsp != NULL) { 2262 error = copyin(args->tsp, <s, sizeof(lts)); 2263 if (error != 0) 2264 return (error); 2265 error = linux_to_native_timespec(&uts, <s); 2266 if (error != 0) 2267 return (error); 2268 2269 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2270 if (itimerfix(&utv)) 2271 return (EINVAL); 2272 2273 microtime(&tv0); 2274 tvp = &utv; 2275 } else 2276 tvp = NULL; 2277 2278 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2279 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2280 2281 if (error == 0 && args->tsp != NULL) { 2282 if (td->td_retval[0] != 0) { 2283 /* 2284 * Compute how much time was left of the timeout, 2285 * by subtracting the current time and the time 2286 * before we started the call, and subtracting 2287 * that result from the user-supplied value. 2288 */ 2289 2290 microtime(&tv1); 2291 timevalsub(&tv1, &tv0); 2292 timevalsub(&utv, &tv1); 2293 if (utv.tv_sec < 0) 2294 timevalclear(&utv); 2295 } else 2296 timevalclear(&utv); 2297 2298 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2299 2300 native_to_linux_timespec(<s, &uts); 2301 error = copyout(<s, args->tsp, sizeof(lts)); 2302 } 2303 2304 return (error); 2305} 2306 2307int 2308linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2309{ 2310 struct timespec ts0, ts1; 2311 struct l_timespec lts; 2312 struct timespec uts, *tsp; 2313 l_sigset_t l_ss; 2314 sigset_t *ssp; 2315 sigset_t ss; 2316 int error; 2317 2318 if (args->sset != NULL) { 2319 if (args->ssize != sizeof(l_ss)) 2320 return (EINVAL); 2321 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2322 if (error) 2323 return (error); 2324 linux_to_bsd_sigset(&l_ss, &ss); 2325 ssp = &ss; 2326 } else 2327 ssp = NULL; 2328 if (args->tsp != NULL) { 2329 error = copyin(args->tsp, <s, sizeof(lts)); 2330 if (error) 2331 return (error); 2332 error = linux_to_native_timespec(&uts, <s); 2333 if (error != 0) 2334 return (error); 2335 2336 nanotime(&ts0); 2337 tsp = &uts; 2338 } else 2339 tsp = NULL; 2340 2341 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2342 2343 if (error == 0 && args->tsp != NULL) { 2344 if (td->td_retval[0]) { 2345 nanotime(&ts1); 2346 timespecsub(&ts1, &ts0); 2347 timespecsub(&uts, &ts1); 2348 if (uts.tv_sec < 0) 2349 timespecclear(&uts); 2350 } else 2351 timespecclear(&uts); 2352 2353 native_to_linux_timespec(<s, &uts); 2354 error = copyout(<s, args->tsp, sizeof(lts)); 2355 } 2356 2357 return (error); 2358} 2359 2360#if defined(DEBUG) || defined(KTR) 2361/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2362 2363u_char linux_debug_map[howmany(LINUX_SYS_MAXSYSCALL, sizeof(u_char))]; 2364 2365static int 2366linux_debug(int syscall, int toggle, int global) 2367{ 2368 2369 if (global) { 2370 char c = toggle ? 0 : 0xff; 2371 2372 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2373 return (0); 2374 } 2375 if (syscall < 0 || syscall >= LINUX_SYS_MAXSYSCALL) 2376 return (EINVAL); 2377 if (toggle) 2378 clrbit(linux_debug_map, syscall); 2379 else 2380 setbit(linux_debug_map, syscall); 2381 return (0); 2382} 2383 2384/* 2385 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2386 * 2387 * E.g.: sysctl linux.debug=21.0 2388 * 2389 * As a special case, syscall "all" will apply to all syscalls globally. 2390 */ 2391#define LINUX_MAX_DEBUGSTR 16 2392int 2393linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2394{ 2395 char value[LINUX_MAX_DEBUGSTR], *p; 2396 int error, sysc, toggle; 2397 int global = 0; 2398 2399 value[0] = '\0'; 2400 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2401 if (error || req->newptr == NULL) 2402 return (error); 2403 for (p = value; *p != '\0' && *p != '.'; p++); 2404 if (*p == '\0') 2405 return (EINVAL); 2406 *p++ = '\0'; 2407 sysc = strtol(value, NULL, 0); 2408 toggle = strtol(p, NULL, 0); 2409 if (strcmp(value, "all") == 0) 2410 global = 1; 2411 error = linux_debug(sysc, toggle, global); 2412 return (error); 2413} 2414 2415#endif /* DEBUG || KTR */ 2416 2417int 2418linux_sched_rr_get_interval(struct thread *td, 2419 struct linux_sched_rr_get_interval_args *uap) 2420{ 2421 struct timespec ts; 2422 struct l_timespec lts; 2423 struct thread *tdt; 2424 int error; 2425 2426 /* 2427 * According to man in case the invalid pid specified 2428 * EINVAL should be returned. 2429 */ 2430 if (uap->pid < 0) 2431 return (EINVAL); 2432 2433 tdt = linux_tdfind(td, uap->pid, -1); 2434 if (tdt == NULL) 2435 return (ESRCH); 2436 2437 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2438 PROC_UNLOCK(tdt->td_proc); 2439 if (error != 0) 2440 return (error); 2441 native_to_linux_timespec(<s, &ts); 2442 return (copyout(<s, uap->interval, sizeof(lts))); 2443} 2444 2445/* 2446 * In case when the Linux thread is the initial thread in 2447 * the thread group thread id is equal to the process id. 2448 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2449 */ 2450struct thread * 2451linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2452{ 2453 struct linux_emuldata *em; 2454 struct thread *tdt; 2455 struct proc *p; 2456 2457 tdt = NULL; 2458 if (tid == 0 || tid == td->td_tid) { 2459 tdt = td; 2460 PROC_LOCK(tdt->td_proc); 2461 } else if (tid > PID_MAX) 2462 tdt = tdfind(tid, pid); 2463 else { 2464 /* 2465 * Initial thread where the tid equal to the pid. 2466 */ 2467 p = pfind(tid); 2468 if (p != NULL) { 2469 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2470 /* 2471 * p is not a Linuxulator process. 2472 */ 2473 PROC_UNLOCK(p); 2474 return (NULL); 2475 } 2476 FOREACH_THREAD_IN_PROC(p, tdt) { 2477 em = em_find(tdt); 2478 if (tid == em->em_tid) 2479 return (tdt); 2480 } 2481 PROC_UNLOCK(p); 2482 } 2483 return (NULL); 2484 } 2485 2486 return (tdt); 2487} 2488 2489void 2490linux_to_bsd_waitopts(int options, int *bsdopts) 2491{ 2492 2493 if (options & LINUX_WNOHANG) 2494 *bsdopts |= WNOHANG; 2495 if (options & LINUX_WUNTRACED) 2496 *bsdopts |= WUNTRACED; 2497 if (options & LINUX_WEXITED) 2498 *bsdopts |= WEXITED; 2499 if (options & LINUX_WCONTINUED) 2500 *bsdopts |= WCONTINUED; 2501 if (options & LINUX_WNOWAIT) 2502 *bsdopts |= WNOWAIT; 2503 2504 if (options & __WCLONE) 2505 *bsdopts |= WLINUXCLONE; 2506} 2507