1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2002 Doug Rabson 5 * Copyright (c) 1994-1995 Søren Schmidt 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer 13 * in this position and unchanged. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. The name of the author may not be used to endorse or promote products 18 * derived from this software without specific prior written permission 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 23 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_compat.h" 36 37 #include <sys/param.h> 38 #include <sys/blist.h> 39 #include <sys/fcntl.h> 40 #if defined(__i386__) 41 #include <sys/imgact_aout.h> 42 #endif 43 #include <sys/jail.h> 44 #include <sys/kernel.h> 45 #include <sys/limits.h> 46 #include <sys/lock.h> 47 #include <sys/malloc.h> 48 #include <sys/mman.h> 49 #include <sys/mount.h> 50 #include <sys/mutex.h> 51 #include <sys/namei.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/procctl.h> 55 #include <sys/reboot.h> 56 #include <sys/racct.h> 57 #include <sys/random.h> 58 #include <sys/resourcevar.h> 59 #include <sys/sched.h> 60 #include <sys/sdt.h> 61 #include <sys/signalvar.h> 62 #include <sys/stat.h> 63 #include <sys/syscallsubr.h> 64 #include <sys/sysctl.h> 65 #include <sys/sysproto.h> 66 #include <sys/systm.h> 67 #include <sys/time.h> 68 #include <sys/vmmeter.h> 69 #include <sys/vnode.h> 70 #include <sys/wait.h> 71 #include <sys/cpuset.h> 72 #include <sys/uio.h> 73 74 #include <security/mac/mac_framework.h> 75 76 #include <vm/vm.h> 77 #include <vm/pmap.h> 78 #include <vm/vm_kern.h> 79 #include <vm/vm_map.h> 80 #include <vm/vm_extern.h> 81 #include <vm/vm_object.h> 82 #include <vm/swap_pager.h> 83 84 #ifdef COMPAT_LINUX32 85 #include <machine/../linux32/linux.h> 86 #include <machine/../linux32/linux32_proto.h> 87 #else 88 #include <machine/../linux/linux.h> 89 #include <machine/../linux/linux_proto.h> 90 #endif 91 92 #include <compat/linux/linux_dtrace.h> 93 #include <compat/linux/linux_file.h> 94 #include <compat/linux/linux_mib.h> 95 #include <compat/linux/linux_signal.h> 96 #include <compat/linux/linux_timer.h> 97 #include <compat/linux/linux_util.h> 98 #include <compat/linux/linux_sysproto.h> 99 #include <compat/linux/linux_emul.h> 100 #include <compat/linux/linux_misc.h> 101 102 /** 103 * Special DTrace provider for the linuxulator. 104 * 105 * In this file we define the provider for the entire linuxulator. All 106 * modules (= files of the linuxulator) use it. 107 * 108 * We define a different name depending on the emulated bitsize, see 109 * ../../<ARCH>/linux{,32}/linux.h, e.g.: 110 * native bitsize = linuxulator 111 * amd64, 32bit emulation = linuxulator32 112 */ 113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE); 114 115 int stclohz; /* Statistics clock frequency */ 116 117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 118 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 119 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 120 RLIMIT_MEMLOCK, RLIMIT_AS 121 }; 122 123 struct l_sysinfo { 124 l_long uptime; /* Seconds since boot */ 125 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 126 #define LINUX_SYSINFO_LOADS_SCALE 65536 127 l_ulong totalram; /* Total usable main memory size */ 128 l_ulong freeram; /* Available memory size */ 129 l_ulong sharedram; /* Amount of shared memory */ 130 l_ulong bufferram; /* Memory used by buffers */ 131 l_ulong totalswap; /* Total swap space size */ 132 l_ulong freeswap; /* swap space still available */ 133 l_ushort procs; /* Number of current processes */ 134 l_ushort pads; 135 l_ulong totalbig; 136 l_ulong freebig; 137 l_uint mem_unit; 138 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 139 }; 140 141 struct l_pselect6arg { 142 l_uintptr_t ss; 143 l_size_t ss_len; 144 }; 145 146 static int linux_utimensat_nsec_valid(l_long); 147 148 149 int 150 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 151 { 152 struct l_sysinfo sysinfo; 153 vm_object_t object; 154 int i, j; 155 struct timespec ts; 156 157 bzero(&sysinfo, sizeof(sysinfo)); 158 getnanouptime(&ts); 159 if (ts.tv_nsec != 0) 160 ts.tv_sec++; 161 sysinfo.uptime = ts.tv_sec; 162 163 /* Use the information from the mib to get our load averages */ 164 for (i = 0; i < 3; i++) 165 sysinfo.loads[i] = averunnable.ldavg[i] * 166 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 167 168 sysinfo.totalram = physmem * PAGE_SIZE; 169 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE; 170 171 sysinfo.sharedram = 0; 172 mtx_lock(&vm_object_list_mtx); 173 TAILQ_FOREACH(object, &vm_object_list, object_list) 174 if (object->shadow_count > 1) 175 sysinfo.sharedram += object->resident_page_count; 176 mtx_unlock(&vm_object_list_mtx); 177 178 sysinfo.sharedram *= PAGE_SIZE; 179 sysinfo.bufferram = 0; 180 181 swap_pager_status(&i, &j); 182 sysinfo.totalswap = i * PAGE_SIZE; 183 sysinfo.freeswap = (i - j) * PAGE_SIZE; 184 185 sysinfo.procs = nprocs; 186 187 /* The following are only present in newer Linux kernels. */ 188 sysinfo.totalbig = 0; 189 sysinfo.freebig = 0; 190 sysinfo.mem_unit = 1; 191 192 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 193 } 194 195 #ifdef LINUX_LEGACY_SYSCALLS 196 int 197 linux_alarm(struct thread *td, struct linux_alarm_args *args) 198 { 199 struct itimerval it, old_it; 200 u_int secs; 201 int error; 202 203 secs = args->secs; 204 /* 205 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2 206 * to match kern_setitimer()'s limit to avoid error from it. 207 * 208 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit 209 * platforms. 210 */ 211 if (secs > INT32_MAX / 2) 212 secs = INT32_MAX / 2; 213 214 it.it_value.tv_sec = secs; 215 it.it_value.tv_usec = 0; 216 timevalclear(&it.it_interval); 217 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 218 KASSERT(error == 0, ("kern_setitimer returns %d", error)); 219 220 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) || 221 old_it.it_value.tv_usec >= 500000) 222 old_it.it_value.tv_sec++; 223 td->td_retval[0] = old_it.it_value.tv_sec; 224 return (0); 225 } 226 #endif 227 228 int 229 linux_brk(struct thread *td, struct linux_brk_args *args) 230 { 231 struct vmspace *vm = td->td_proc->p_vmspace; 232 uintptr_t new, old; 233 234 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize); 235 new = (uintptr_t)args->dsend; 236 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new)) 237 td->td_retval[0] = (register_t)new; 238 else 239 td->td_retval[0] = (register_t)old; 240 241 return (0); 242 } 243 244 #if defined(__i386__) 245 /* XXX: what about amd64/linux32? */ 246 247 int 248 linux_uselib(struct thread *td, struct linux_uselib_args *args) 249 { 250 struct nameidata ni; 251 struct vnode *vp; 252 struct exec *a_out; 253 vm_map_t map; 254 vm_map_entry_t entry; 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; 262 bool locked, opened, textset; 263 264 LCONVPATHEXIST(td, args->library, &library); 265 266 a_out = NULL; 267 vp = NULL; 268 locked = false; 269 textset = false; 270 opened = false; 271 272 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 273 UIO_SYSSPACE, library, td); 274 error = namei(&ni); 275 LFREEPATH(library); 276 if (error) 277 goto cleanup; 278 279 vp = ni.ni_vp; 280 NDFREE(&ni, NDF_ONLY_PNBUF); 281 282 /* 283 * From here on down, we have a locked vnode that must be unlocked. 284 * XXX: The code below largely duplicates exec_check_permissions(). 285 */ 286 locked = true; 287 288 /* Executable? */ 289 error = VOP_GETATTR(vp, &attr, td->td_ucred); 290 if (error) 291 goto cleanup; 292 293 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 294 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 295 /* EACCESS is what exec(2) returns. */ 296 error = ENOEXEC; 297 goto cleanup; 298 } 299 300 /* Sensible size? */ 301 if (attr.va_size == 0) { 302 error = ENOEXEC; 303 goto cleanup; 304 } 305 306 /* Can we access it? */ 307 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 308 if (error) 309 goto cleanup; 310 311 /* 312 * XXX: This should use vn_open() so that it is properly authorized, 313 * and to reduce code redundancy all over the place here. 314 * XXX: Not really, it duplicates far more of exec_check_permissions() 315 * than vn_open(). 316 */ 317 #ifdef MAC 318 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 319 if (error) 320 goto cleanup; 321 #endif 322 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 323 if (error) 324 goto cleanup; 325 opened = true; 326 327 /* Pull in executable header into exec_map */ 328 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 329 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 330 if (error) 331 goto cleanup; 332 333 /* Is it a Linux binary ? */ 334 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 335 error = ENOEXEC; 336 goto cleanup; 337 } 338 339 /* 340 * While we are here, we should REALLY do some more checks 341 */ 342 343 /* Set file/virtual offset based on a.out variant. */ 344 switch ((int)(a_out->a_magic & 0xffff)) { 345 case 0413: /* ZMAGIC */ 346 file_offset = 1024; 347 break; 348 case 0314: /* QMAGIC */ 349 file_offset = 0; 350 break; 351 default: 352 error = ENOEXEC; 353 goto cleanup; 354 } 355 356 bss_size = round_page(a_out->a_bss); 357 358 /* Check various fields in header for validity/bounds. */ 359 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 360 error = ENOEXEC; 361 goto cleanup; 362 } 363 364 /* text + data can't exceed file size */ 365 if (a_out->a_data + a_out->a_text > attr.va_size) { 366 error = EFAULT; 367 goto cleanup; 368 } 369 370 /* 371 * text/data/bss must not exceed limits 372 * XXX - this is not complete. it should check current usage PLUS 373 * the resources needed by this library. 374 */ 375 PROC_LOCK(td->td_proc); 376 if (a_out->a_text > maxtsiz || 377 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) || 378 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 379 bss_size) != 0) { 380 PROC_UNLOCK(td->td_proc); 381 error = ENOMEM; 382 goto cleanup; 383 } 384 PROC_UNLOCK(td->td_proc); 385 386 /* 387 * Prevent more writers. 388 */ 389 error = VOP_SET_TEXT(vp); 390 if (error != 0) 391 goto cleanup; 392 textset = true; 393 394 /* 395 * Lock no longer needed 396 */ 397 locked = false; 398 VOP_UNLOCK(vp, 0); 399 400 /* 401 * Check if file_offset page aligned. Currently we cannot handle 402 * misalinged file offsets, and so we read in the entire image 403 * (what a waste). 404 */ 405 if (file_offset & PAGE_MASK) { 406 /* Map text+data read/write/execute */ 407 408 /* a_entry is the load address and is page aligned */ 409 vmaddr = trunc_page(a_out->a_entry); 410 411 /* get anon user mapping, read+write+execute */ 412 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 413 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 414 VM_PROT_ALL, VM_PROT_ALL, 0); 415 if (error) 416 goto cleanup; 417 418 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 419 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 420 td->td_ucred, NOCRED, &aresid, td); 421 if (error != 0) 422 goto cleanup; 423 if (aresid != 0) { 424 error = ENOEXEC; 425 goto cleanup; 426 } 427 } else { 428 /* 429 * for QMAGIC, a_entry is 20 bytes beyond the load address 430 * to skip the executable header 431 */ 432 vmaddr = trunc_page(a_out->a_entry); 433 434 /* 435 * Map it all into the process's space as a single 436 * copy-on-write "data" segment. 437 */ 438 map = &td->td_proc->p_vmspace->vm_map; 439 error = vm_mmap(map, &vmaddr, 440 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 441 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 442 if (error) 443 goto cleanup; 444 vm_map_lock(map); 445 if (!vm_map_lookup_entry(map, vmaddr, &entry)) { 446 vm_map_unlock(map); 447 error = EDOOFUS; 448 goto cleanup; 449 } 450 entry->eflags |= MAP_ENTRY_VN_EXEC; 451 vm_map_unlock(map); 452 textset = false; 453 } 454 455 if (bss_size != 0) { 456 /* Calculate BSS start address */ 457 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 458 a_out->a_data; 459 460 /* allocate some 'anon' space */ 461 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 462 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 463 VM_PROT_ALL, 0); 464 if (error) 465 goto cleanup; 466 } 467 468 cleanup: 469 if (opened) { 470 if (locked) 471 VOP_UNLOCK(vp, 0); 472 locked = false; 473 VOP_CLOSE(vp, FREAD, td->td_ucred, td); 474 } 475 if (textset) { 476 if (!locked) { 477 locked = true; 478 VOP_LOCK(vp, LK_SHARED | LK_RETRY); 479 } 480 VOP_UNSET_TEXT_CHECKED(vp); 481 } 482 if (locked) 483 VOP_UNLOCK(vp, 0); 484 485 /* Release the temporary mapping. */ 486 if (a_out) 487 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 488 489 return (error); 490 } 491 492 #endif /* __i386__ */ 493 494 #ifdef LINUX_LEGACY_SYSCALLS 495 int 496 linux_select(struct thread *td, struct linux_select_args *args) 497 { 498 l_timeval ltv; 499 struct timeval tv0, tv1, utv, *tvp; 500 int error; 501 502 /* 503 * Store current time for computation of the amount of 504 * time left. 505 */ 506 if (args->timeout) { 507 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 508 goto select_out; 509 utv.tv_sec = ltv.tv_sec; 510 utv.tv_usec = ltv.tv_usec; 511 512 if (itimerfix(&utv)) { 513 /* 514 * The timeval was invalid. Convert it to something 515 * valid that will act as it does under Linux. 516 */ 517 utv.tv_sec += utv.tv_usec / 1000000; 518 utv.tv_usec %= 1000000; 519 if (utv.tv_usec < 0) { 520 utv.tv_sec -= 1; 521 utv.tv_usec += 1000000; 522 } 523 if (utv.tv_sec < 0) 524 timevalclear(&utv); 525 } 526 microtime(&tv0); 527 tvp = &utv; 528 } else 529 tvp = NULL; 530 531 error = kern_select(td, args->nfds, args->readfds, args->writefds, 532 args->exceptfds, tvp, LINUX_NFDBITS); 533 if (error) 534 goto select_out; 535 536 if (args->timeout) { 537 if (td->td_retval[0]) { 538 /* 539 * Compute how much time was left of the timeout, 540 * by subtracting the current time and the time 541 * before we started the call, and subtracting 542 * that result from the user-supplied value. 543 */ 544 microtime(&tv1); 545 timevalsub(&tv1, &tv0); 546 timevalsub(&utv, &tv1); 547 if (utv.tv_sec < 0) 548 timevalclear(&utv); 549 } else 550 timevalclear(&utv); 551 ltv.tv_sec = utv.tv_sec; 552 ltv.tv_usec = utv.tv_usec; 553 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 554 goto select_out; 555 } 556 557 select_out: 558 return (error); 559 } 560 #endif 561 562 int 563 linux_mremap(struct thread *td, struct linux_mremap_args *args) 564 { 565 uintptr_t addr; 566 size_t len; 567 int error = 0; 568 569 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 570 td->td_retval[0] = 0; 571 return (EINVAL); 572 } 573 574 /* 575 * Check for the page alignment. 576 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 577 */ 578 if (args->addr & PAGE_MASK) { 579 td->td_retval[0] = 0; 580 return (EINVAL); 581 } 582 583 args->new_len = round_page(args->new_len); 584 args->old_len = round_page(args->old_len); 585 586 if (args->new_len > args->old_len) { 587 td->td_retval[0] = 0; 588 return (ENOMEM); 589 } 590 591 if (args->new_len < args->old_len) { 592 addr = args->addr + args->new_len; 593 len = args->old_len - args->new_len; 594 error = kern_munmap(td, addr, len); 595 } 596 597 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 598 return (error); 599 } 600 601 #define LINUX_MS_ASYNC 0x0001 602 #define LINUX_MS_INVALIDATE 0x0002 603 #define LINUX_MS_SYNC 0x0004 604 605 int 606 linux_msync(struct thread *td, struct linux_msync_args *args) 607 { 608 609 return (kern_msync(td, args->addr, args->len, 610 args->fl & ~LINUX_MS_SYNC)); 611 } 612 613 #ifdef LINUX_LEGACY_SYSCALLS 614 int 615 linux_time(struct thread *td, struct linux_time_args *args) 616 { 617 struct timeval tv; 618 l_time_t tm; 619 int error; 620 621 microtime(&tv); 622 tm = tv.tv_sec; 623 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 624 return (error); 625 td->td_retval[0] = tm; 626 return (0); 627 } 628 #endif 629 630 struct l_times_argv { 631 l_clock_t tms_utime; 632 l_clock_t tms_stime; 633 l_clock_t tms_cutime; 634 l_clock_t tms_cstime; 635 }; 636 637 638 /* 639 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 640 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 641 * auxiliary vector entry. 642 */ 643 #define CLK_TCK 100 644 645 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 646 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 647 648 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 649 CONVNTCK(r) : CONVOTCK(r)) 650 651 int 652 linux_times(struct thread *td, struct linux_times_args *args) 653 { 654 struct timeval tv, utime, stime, cutime, cstime; 655 struct l_times_argv tms; 656 struct proc *p; 657 int error; 658 659 if (args->buf != NULL) { 660 p = td->td_proc; 661 PROC_LOCK(p); 662 PROC_STATLOCK(p); 663 calcru(p, &utime, &stime); 664 PROC_STATUNLOCK(p); 665 calccru(p, &cutime, &cstime); 666 PROC_UNLOCK(p); 667 668 tms.tms_utime = CONVTCK(utime); 669 tms.tms_stime = CONVTCK(stime); 670 671 tms.tms_cutime = CONVTCK(cutime); 672 tms.tms_cstime = CONVTCK(cstime); 673 674 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 675 return (error); 676 } 677 678 microuptime(&tv); 679 td->td_retval[0] = (int)CONVTCK(tv); 680 return (0); 681 } 682 683 int 684 linux_newuname(struct thread *td, struct linux_newuname_args *args) 685 { 686 struct l_new_utsname utsname; 687 char osname[LINUX_MAX_UTSNAME]; 688 char osrelease[LINUX_MAX_UTSNAME]; 689 char *p; 690 691 linux_get_osname(td, osname); 692 linux_get_osrelease(td, osrelease); 693 694 bzero(&utsname, sizeof(utsname)); 695 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 696 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 697 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 698 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 699 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 700 for (p = utsname.version; *p != '\0'; ++p) 701 if (*p == '\n') { 702 *p = '\0'; 703 break; 704 } 705 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 706 707 return (copyout(&utsname, args->buf, sizeof(utsname))); 708 } 709 710 struct l_utimbuf { 711 l_time_t l_actime; 712 l_time_t l_modtime; 713 }; 714 715 #ifdef LINUX_LEGACY_SYSCALLS 716 int 717 linux_utime(struct thread *td, struct linux_utime_args *args) 718 { 719 struct timeval tv[2], *tvp; 720 struct l_utimbuf lut; 721 char *fname; 722 int error; 723 724 LCONVPATHEXIST(td, args->fname, &fname); 725 726 if (args->times) { 727 if ((error = copyin(args->times, &lut, sizeof lut))) { 728 LFREEPATH(fname); 729 return (error); 730 } 731 tv[0].tv_sec = lut.l_actime; 732 tv[0].tv_usec = 0; 733 tv[1].tv_sec = lut.l_modtime; 734 tv[1].tv_usec = 0; 735 tvp = tv; 736 } else 737 tvp = NULL; 738 739 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp, 740 UIO_SYSSPACE); 741 LFREEPATH(fname); 742 return (error); 743 } 744 #endif 745 746 #ifdef LINUX_LEGACY_SYSCALLS 747 int 748 linux_utimes(struct thread *td, struct linux_utimes_args *args) 749 { 750 l_timeval ltv[2]; 751 struct timeval tv[2], *tvp = NULL; 752 char *fname; 753 int error; 754 755 LCONVPATHEXIST(td, args->fname, &fname); 756 757 if (args->tptr != NULL) { 758 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 759 LFREEPATH(fname); 760 return (error); 761 } 762 tv[0].tv_sec = ltv[0].tv_sec; 763 tv[0].tv_usec = ltv[0].tv_usec; 764 tv[1].tv_sec = ltv[1].tv_sec; 765 tv[1].tv_usec = ltv[1].tv_usec; 766 tvp = tv; 767 } 768 769 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, 770 tvp, UIO_SYSSPACE); 771 LFREEPATH(fname); 772 return (error); 773 } 774 #endif 775 776 static int 777 linux_utimensat_nsec_valid(l_long nsec) 778 { 779 780 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW) 781 return (0); 782 if (nsec >= 0 && nsec <= 999999999) 783 return (0); 784 return (1); 785 } 786 787 int 788 linux_utimensat(struct thread *td, struct linux_utimensat_args *args) 789 { 790 struct l_timespec l_times[2]; 791 struct timespec times[2], *timesp = NULL; 792 char *path = NULL; 793 int error, dfd, flags = 0; 794 795 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 796 797 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 798 return (EINVAL); 799 800 if (args->times != NULL) { 801 error = copyin(args->times, l_times, sizeof(l_times)); 802 if (error != 0) 803 return (error); 804 805 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 || 806 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0) 807 return (EINVAL); 808 809 times[0].tv_sec = l_times[0].tv_sec; 810 switch (l_times[0].tv_nsec) 811 { 812 case LINUX_UTIME_OMIT: 813 times[0].tv_nsec = UTIME_OMIT; 814 break; 815 case LINUX_UTIME_NOW: 816 times[0].tv_nsec = UTIME_NOW; 817 break; 818 default: 819 times[0].tv_nsec = l_times[0].tv_nsec; 820 } 821 822 times[1].tv_sec = l_times[1].tv_sec; 823 switch (l_times[1].tv_nsec) 824 { 825 case LINUX_UTIME_OMIT: 826 times[1].tv_nsec = UTIME_OMIT; 827 break; 828 case LINUX_UTIME_NOW: 829 times[1].tv_nsec = UTIME_NOW; 830 break; 831 default: 832 times[1].tv_nsec = l_times[1].tv_nsec; 833 break; 834 } 835 timesp = times; 836 837 /* This breaks POSIX, but is what the Linux kernel does 838 * _on purpose_ (documented in the man page for utimensat(2)), 839 * so we must follow that behaviour. */ 840 if (times[0].tv_nsec == UTIME_OMIT && 841 times[1].tv_nsec == UTIME_OMIT) 842 return (0); 843 } 844 845 if (args->pathname != NULL) 846 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 847 else if (args->flags != 0) 848 return (EINVAL); 849 850 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 851 flags |= AT_SYMLINK_NOFOLLOW; 852 853 if (path == NULL) 854 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 855 else { 856 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 857 UIO_SYSSPACE, flags); 858 LFREEPATH(path); 859 } 860 861 return (error); 862 } 863 864 #ifdef LINUX_LEGACY_SYSCALLS 865 int 866 linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 867 { 868 l_timeval ltv[2]; 869 struct timeval tv[2], *tvp = NULL; 870 char *fname; 871 int error, dfd; 872 873 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 874 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 875 876 if (args->utimes != NULL) { 877 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 878 LFREEPATH(fname); 879 return (error); 880 } 881 tv[0].tv_sec = ltv[0].tv_sec; 882 tv[0].tv_usec = ltv[0].tv_usec; 883 tv[1].tv_sec = ltv[1].tv_sec; 884 tv[1].tv_usec = ltv[1].tv_usec; 885 tvp = tv; 886 } 887 888 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 889 LFREEPATH(fname); 890 return (error); 891 } 892 #endif 893 894 static int 895 linux_common_wait(struct thread *td, int pid, int *statusp, 896 int options, struct __wrusage *wrup) 897 { 898 siginfo_t siginfo; 899 idtype_t idtype; 900 id_t id; 901 int error, status, tmpstat; 902 903 if (pid == WAIT_ANY) { 904 idtype = P_ALL; 905 id = 0; 906 } else if (pid < 0) { 907 idtype = P_PGID; 908 id = (id_t)-pid; 909 } else { 910 idtype = P_PID; 911 id = (id_t)pid; 912 } 913 914 /* 915 * For backward compatibility we implicitly add flags WEXITED 916 * and WTRAPPED here. 917 */ 918 options |= WEXITED | WTRAPPED; 919 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo); 920 if (error) 921 return (error); 922 923 if (statusp) { 924 tmpstat = status & 0xffff; 925 if (WIFSIGNALED(tmpstat)) { 926 tmpstat = (tmpstat & 0xffffff80) | 927 bsd_to_linux_signal(WTERMSIG(tmpstat)); 928 } else if (WIFSTOPPED(tmpstat)) { 929 tmpstat = (tmpstat & 0xffff00ff) | 930 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 931 #if defined(__amd64__) && !defined(COMPAT_LINUX32) 932 if (WSTOPSIG(status) == SIGTRAP) { 933 tmpstat = linux_ptrace_status(td, 934 siginfo.si_pid, tmpstat); 935 } 936 #endif 937 } else if (WIFCONTINUED(tmpstat)) { 938 tmpstat = 0xffff; 939 } 940 error = copyout(&tmpstat, statusp, sizeof(int)); 941 } 942 943 return (error); 944 } 945 946 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 947 int 948 linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 949 { 950 struct linux_wait4_args wait4_args; 951 952 wait4_args.pid = args->pid; 953 wait4_args.status = args->status; 954 wait4_args.options = args->options; 955 wait4_args.rusage = NULL; 956 957 return (linux_wait4(td, &wait4_args)); 958 } 959 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 960 961 int 962 linux_wait4(struct thread *td, struct linux_wait4_args *args) 963 { 964 int error, options; 965 struct __wrusage wru, *wrup; 966 967 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 968 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 969 return (EINVAL); 970 971 options = WEXITED; 972 linux_to_bsd_waitopts(args->options, &options); 973 974 if (args->rusage != NULL) 975 wrup = &wru; 976 else 977 wrup = NULL; 978 error = linux_common_wait(td, args->pid, args->status, options, wrup); 979 if (error != 0) 980 return (error); 981 if (args->rusage != NULL) 982 error = linux_copyout_rusage(&wru.wru_self, args->rusage); 983 return (error); 984 } 985 986 int 987 linux_waitid(struct thread *td, struct linux_waitid_args *args) 988 { 989 int status, options, sig; 990 struct __wrusage wru; 991 siginfo_t siginfo; 992 l_siginfo_t lsi; 993 idtype_t idtype; 994 struct proc *p; 995 int error; 996 997 options = 0; 998 linux_to_bsd_waitopts(args->options, &options); 999 1000 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1001 return (EINVAL); 1002 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1003 return (EINVAL); 1004 1005 switch (args->idtype) { 1006 case LINUX_P_ALL: 1007 idtype = P_ALL; 1008 break; 1009 case LINUX_P_PID: 1010 if (args->id <= 0) 1011 return (EINVAL); 1012 idtype = P_PID; 1013 break; 1014 case LINUX_P_PGID: 1015 if (args->id <= 0) 1016 return (EINVAL); 1017 idtype = P_PGID; 1018 break; 1019 default: 1020 return (EINVAL); 1021 } 1022 1023 error = kern_wait6(td, idtype, args->id, &status, options, 1024 &wru, &siginfo); 1025 if (error != 0) 1026 return (error); 1027 if (args->rusage != NULL) { 1028 error = linux_copyout_rusage(&wru.wru_children, 1029 args->rusage); 1030 if (error != 0) 1031 return (error); 1032 } 1033 if (args->info != NULL) { 1034 p = td->td_proc; 1035 bzero(&lsi, sizeof(lsi)); 1036 if (td->td_retval[0] != 0) { 1037 sig = bsd_to_linux_signal(siginfo.si_signo); 1038 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1039 } 1040 error = copyout(&lsi, args->info, sizeof(lsi)); 1041 } 1042 td->td_retval[0] = 0; 1043 1044 return (error); 1045 } 1046 1047 #ifdef LINUX_LEGACY_SYSCALLS 1048 int 1049 linux_mknod(struct thread *td, struct linux_mknod_args *args) 1050 { 1051 char *path; 1052 int error; 1053 1054 LCONVPATHCREAT(td, args->path, &path); 1055 1056 switch (args->mode & S_IFMT) { 1057 case S_IFIFO: 1058 case S_IFSOCK: 1059 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE, 1060 args->mode); 1061 break; 1062 1063 case S_IFCHR: 1064 case S_IFBLK: 1065 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE, 1066 args->mode, args->dev); 1067 break; 1068 1069 case S_IFDIR: 1070 error = EPERM; 1071 break; 1072 1073 case 0: 1074 args->mode |= S_IFREG; 1075 /* FALLTHROUGH */ 1076 case S_IFREG: 1077 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE, 1078 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1079 if (error == 0) 1080 kern_close(td, td->td_retval[0]); 1081 break; 1082 1083 default: 1084 error = EINVAL; 1085 break; 1086 } 1087 LFREEPATH(path); 1088 return (error); 1089 } 1090 #endif 1091 1092 int 1093 linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1094 { 1095 char *path; 1096 int error, dfd; 1097 1098 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1099 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1100 1101 switch (args->mode & S_IFMT) { 1102 case S_IFIFO: 1103 case S_IFSOCK: 1104 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1105 break; 1106 1107 case S_IFCHR: 1108 case S_IFBLK: 1109 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1110 args->dev); 1111 break; 1112 1113 case S_IFDIR: 1114 error = EPERM; 1115 break; 1116 1117 case 0: 1118 args->mode |= S_IFREG; 1119 /* FALLTHROUGH */ 1120 case S_IFREG: 1121 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1122 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1123 if (error == 0) 1124 kern_close(td, td->td_retval[0]); 1125 break; 1126 1127 default: 1128 error = EINVAL; 1129 break; 1130 } 1131 LFREEPATH(path); 1132 return (error); 1133 } 1134 1135 /* 1136 * UGH! This is just about the dumbest idea I've ever heard!! 1137 */ 1138 int 1139 linux_personality(struct thread *td, struct linux_personality_args *args) 1140 { 1141 struct linux_pemuldata *pem; 1142 struct proc *p = td->td_proc; 1143 uint32_t old; 1144 1145 PROC_LOCK(p); 1146 pem = pem_find(p); 1147 old = pem->persona; 1148 if (args->per != 0xffffffff) 1149 pem->persona = args->per; 1150 PROC_UNLOCK(p); 1151 1152 td->td_retval[0] = old; 1153 return (0); 1154 } 1155 1156 struct l_itimerval { 1157 l_timeval it_interval; 1158 l_timeval it_value; 1159 }; 1160 1161 #define B2L_ITIMERVAL(bip, lip) \ 1162 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1163 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1164 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1165 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1166 1167 int 1168 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1169 { 1170 int error; 1171 struct l_itimerval ls; 1172 struct itimerval aitv, oitv; 1173 1174 if (uap->itv == NULL) { 1175 uap->itv = uap->oitv; 1176 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1177 } 1178 1179 error = copyin(uap->itv, &ls, sizeof(ls)); 1180 if (error != 0) 1181 return (error); 1182 B2L_ITIMERVAL(&aitv, &ls); 1183 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1184 if (error != 0 || uap->oitv == NULL) 1185 return (error); 1186 B2L_ITIMERVAL(&ls, &oitv); 1187 1188 return (copyout(&ls, uap->oitv, sizeof(ls))); 1189 } 1190 1191 int 1192 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1193 { 1194 int error; 1195 struct l_itimerval ls; 1196 struct itimerval aitv; 1197 1198 error = kern_getitimer(td, uap->which, &aitv); 1199 if (error != 0) 1200 return (error); 1201 B2L_ITIMERVAL(&ls, &aitv); 1202 return (copyout(&ls, uap->itv, sizeof(ls))); 1203 } 1204 1205 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1206 int 1207 linux_nice(struct thread *td, struct linux_nice_args *args) 1208 { 1209 struct setpriority_args bsd_args; 1210 1211 bsd_args.which = PRIO_PROCESS; 1212 bsd_args.who = 0; /* current process */ 1213 bsd_args.prio = args->inc; 1214 return (sys_setpriority(td, &bsd_args)); 1215 } 1216 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1217 1218 int 1219 linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1220 { 1221 struct ucred *newcred, *oldcred; 1222 l_gid_t *linux_gidset; 1223 gid_t *bsd_gidset; 1224 int ngrp, error; 1225 struct proc *p; 1226 1227 ngrp = args->gidsetsize; 1228 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1229 return (EINVAL); 1230 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1231 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1232 if (error) 1233 goto out; 1234 newcred = crget(); 1235 crextend(newcred, ngrp + 1); 1236 p = td->td_proc; 1237 PROC_LOCK(p); 1238 oldcred = p->p_ucred; 1239 crcopy(newcred, oldcred); 1240 1241 /* 1242 * cr_groups[0] holds egid. Setting the whole set from 1243 * the supplied set will cause egid to be changed too. 1244 * Keep cr_groups[0] unchanged to prevent that. 1245 */ 1246 1247 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) { 1248 PROC_UNLOCK(p); 1249 crfree(newcred); 1250 goto out; 1251 } 1252 1253 if (ngrp > 0) { 1254 newcred->cr_ngroups = ngrp + 1; 1255 1256 bsd_gidset = newcred->cr_groups; 1257 ngrp--; 1258 while (ngrp >= 0) { 1259 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1260 ngrp--; 1261 } 1262 } else 1263 newcred->cr_ngroups = 1; 1264 1265 setsugid(p); 1266 proc_set_cred(p, newcred); 1267 PROC_UNLOCK(p); 1268 crfree(oldcred); 1269 error = 0; 1270 out: 1271 free(linux_gidset, M_LINUX); 1272 return (error); 1273 } 1274 1275 int 1276 linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1277 { 1278 struct ucred *cred; 1279 l_gid_t *linux_gidset; 1280 gid_t *bsd_gidset; 1281 int bsd_gidsetsz, ngrp, error; 1282 1283 cred = td->td_ucred; 1284 bsd_gidset = cred->cr_groups; 1285 bsd_gidsetsz = cred->cr_ngroups - 1; 1286 1287 /* 1288 * cr_groups[0] holds egid. Returning the whole set 1289 * here will cause a duplicate. Exclude cr_groups[0] 1290 * to prevent that. 1291 */ 1292 1293 if ((ngrp = args->gidsetsize) == 0) { 1294 td->td_retval[0] = bsd_gidsetsz; 1295 return (0); 1296 } 1297 1298 if (ngrp < bsd_gidsetsz) 1299 return (EINVAL); 1300 1301 ngrp = 0; 1302 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1303 M_LINUX, M_WAITOK); 1304 while (ngrp < bsd_gidsetsz) { 1305 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1306 ngrp++; 1307 } 1308 1309 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1310 free(linux_gidset, M_LINUX); 1311 if (error) 1312 return (error); 1313 1314 td->td_retval[0] = ngrp; 1315 return (0); 1316 } 1317 1318 int 1319 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1320 { 1321 struct rlimit bsd_rlim; 1322 struct l_rlimit rlim; 1323 u_int which; 1324 int error; 1325 1326 if (args->resource >= LINUX_RLIM_NLIMITS) 1327 return (EINVAL); 1328 1329 which = linux_to_bsd_resource[args->resource]; 1330 if (which == -1) 1331 return (EINVAL); 1332 1333 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1334 if (error) 1335 return (error); 1336 1337 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1338 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1339 return (kern_setrlimit(td, which, &bsd_rlim)); 1340 } 1341 1342 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1343 int 1344 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1345 { 1346 struct l_rlimit rlim; 1347 struct rlimit bsd_rlim; 1348 u_int which; 1349 1350 if (args->resource >= LINUX_RLIM_NLIMITS) 1351 return (EINVAL); 1352 1353 which = linux_to_bsd_resource[args->resource]; 1354 if (which == -1) 1355 return (EINVAL); 1356 1357 lim_rlimit(td, which, &bsd_rlim); 1358 1359 #ifdef COMPAT_LINUX32 1360 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1361 if (rlim.rlim_cur == UINT_MAX) 1362 rlim.rlim_cur = INT_MAX; 1363 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1364 if (rlim.rlim_max == UINT_MAX) 1365 rlim.rlim_max = INT_MAX; 1366 #else 1367 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1368 if (rlim.rlim_cur == ULONG_MAX) 1369 rlim.rlim_cur = LONG_MAX; 1370 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1371 if (rlim.rlim_max == ULONG_MAX) 1372 rlim.rlim_max = LONG_MAX; 1373 #endif 1374 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1375 } 1376 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1377 1378 int 1379 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1380 { 1381 struct l_rlimit rlim; 1382 struct rlimit bsd_rlim; 1383 u_int which; 1384 1385 if (args->resource >= LINUX_RLIM_NLIMITS) 1386 return (EINVAL); 1387 1388 which = linux_to_bsd_resource[args->resource]; 1389 if (which == -1) 1390 return (EINVAL); 1391 1392 lim_rlimit(td, which, &bsd_rlim); 1393 1394 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1395 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1396 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1397 } 1398 1399 int 1400 linux_sched_setscheduler(struct thread *td, 1401 struct linux_sched_setscheduler_args *args) 1402 { 1403 struct sched_param sched_param; 1404 struct thread *tdt; 1405 int error, policy; 1406 1407 switch (args->policy) { 1408 case LINUX_SCHED_OTHER: 1409 policy = SCHED_OTHER; 1410 break; 1411 case LINUX_SCHED_FIFO: 1412 policy = SCHED_FIFO; 1413 break; 1414 case LINUX_SCHED_RR: 1415 policy = SCHED_RR; 1416 break; 1417 default: 1418 return (EINVAL); 1419 } 1420 1421 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1422 if (error) 1423 return (error); 1424 1425 tdt = linux_tdfind(td, args->pid, -1); 1426 if (tdt == NULL) 1427 return (ESRCH); 1428 1429 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1430 PROC_UNLOCK(tdt->td_proc); 1431 return (error); 1432 } 1433 1434 int 1435 linux_sched_getscheduler(struct thread *td, 1436 struct linux_sched_getscheduler_args *args) 1437 { 1438 struct thread *tdt; 1439 int error, policy; 1440 1441 tdt = linux_tdfind(td, args->pid, -1); 1442 if (tdt == NULL) 1443 return (ESRCH); 1444 1445 error = kern_sched_getscheduler(td, tdt, &policy); 1446 PROC_UNLOCK(tdt->td_proc); 1447 1448 switch (policy) { 1449 case SCHED_OTHER: 1450 td->td_retval[0] = LINUX_SCHED_OTHER; 1451 break; 1452 case SCHED_FIFO: 1453 td->td_retval[0] = LINUX_SCHED_FIFO; 1454 break; 1455 case SCHED_RR: 1456 td->td_retval[0] = LINUX_SCHED_RR; 1457 break; 1458 } 1459 return (error); 1460 } 1461 1462 int 1463 linux_sched_get_priority_max(struct thread *td, 1464 struct linux_sched_get_priority_max_args *args) 1465 { 1466 struct sched_get_priority_max_args bsd; 1467 1468 switch (args->policy) { 1469 case LINUX_SCHED_OTHER: 1470 bsd.policy = SCHED_OTHER; 1471 break; 1472 case LINUX_SCHED_FIFO: 1473 bsd.policy = SCHED_FIFO; 1474 break; 1475 case LINUX_SCHED_RR: 1476 bsd.policy = SCHED_RR; 1477 break; 1478 default: 1479 return (EINVAL); 1480 } 1481 return (sys_sched_get_priority_max(td, &bsd)); 1482 } 1483 1484 int 1485 linux_sched_get_priority_min(struct thread *td, 1486 struct linux_sched_get_priority_min_args *args) 1487 { 1488 struct sched_get_priority_min_args bsd; 1489 1490 switch (args->policy) { 1491 case LINUX_SCHED_OTHER: 1492 bsd.policy = SCHED_OTHER; 1493 break; 1494 case LINUX_SCHED_FIFO: 1495 bsd.policy = SCHED_FIFO; 1496 break; 1497 case LINUX_SCHED_RR: 1498 bsd.policy = SCHED_RR; 1499 break; 1500 default: 1501 return (EINVAL); 1502 } 1503 return (sys_sched_get_priority_min(td, &bsd)); 1504 } 1505 1506 #define REBOOT_CAD_ON 0x89abcdef 1507 #define REBOOT_CAD_OFF 0 1508 #define REBOOT_HALT 0xcdef0123 1509 #define REBOOT_RESTART 0x01234567 1510 #define REBOOT_RESTART2 0xA1B2C3D4 1511 #define REBOOT_POWEROFF 0x4321FEDC 1512 #define REBOOT_MAGIC1 0xfee1dead 1513 #define REBOOT_MAGIC2 0x28121969 1514 #define REBOOT_MAGIC2A 0x05121996 1515 #define REBOOT_MAGIC2B 0x16041998 1516 1517 int 1518 linux_reboot(struct thread *td, struct linux_reboot_args *args) 1519 { 1520 struct reboot_args bsd_args; 1521 1522 if (args->magic1 != REBOOT_MAGIC1) 1523 return (EINVAL); 1524 1525 switch (args->magic2) { 1526 case REBOOT_MAGIC2: 1527 case REBOOT_MAGIC2A: 1528 case REBOOT_MAGIC2B: 1529 break; 1530 default: 1531 return (EINVAL); 1532 } 1533 1534 switch (args->cmd) { 1535 case REBOOT_CAD_ON: 1536 case REBOOT_CAD_OFF: 1537 return (priv_check(td, PRIV_REBOOT)); 1538 case REBOOT_HALT: 1539 bsd_args.opt = RB_HALT; 1540 break; 1541 case REBOOT_RESTART: 1542 case REBOOT_RESTART2: 1543 bsd_args.opt = 0; 1544 break; 1545 case REBOOT_POWEROFF: 1546 bsd_args.opt = RB_POWEROFF; 1547 break; 1548 default: 1549 return (EINVAL); 1550 } 1551 return (sys_reboot(td, &bsd_args)); 1552 } 1553 1554 1555 int 1556 linux_getpid(struct thread *td, struct linux_getpid_args *args) 1557 { 1558 1559 td->td_retval[0] = td->td_proc->p_pid; 1560 1561 return (0); 1562 } 1563 1564 int 1565 linux_gettid(struct thread *td, struct linux_gettid_args *args) 1566 { 1567 struct linux_emuldata *em; 1568 1569 em = em_find(td); 1570 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1571 1572 td->td_retval[0] = em->em_tid; 1573 1574 return (0); 1575 } 1576 1577 1578 int 1579 linux_getppid(struct thread *td, struct linux_getppid_args *args) 1580 { 1581 1582 td->td_retval[0] = kern_getppid(td); 1583 return (0); 1584 } 1585 1586 int 1587 linux_getgid(struct thread *td, struct linux_getgid_args *args) 1588 { 1589 1590 td->td_retval[0] = td->td_ucred->cr_rgid; 1591 return (0); 1592 } 1593 1594 int 1595 linux_getuid(struct thread *td, struct linux_getuid_args *args) 1596 { 1597 1598 td->td_retval[0] = td->td_ucred->cr_ruid; 1599 return (0); 1600 } 1601 1602 1603 int 1604 linux_getsid(struct thread *td, struct linux_getsid_args *args) 1605 { 1606 struct getsid_args bsd; 1607 1608 bsd.pid = args->pid; 1609 return (sys_getsid(td, &bsd)); 1610 } 1611 1612 int 1613 linux_nosys(struct thread *td, struct nosys_args *ignore) 1614 { 1615 1616 return (ENOSYS); 1617 } 1618 1619 int 1620 linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1621 { 1622 struct getpriority_args bsd_args; 1623 int error; 1624 1625 bsd_args.which = args->which; 1626 bsd_args.who = args->who; 1627 error = sys_getpriority(td, &bsd_args); 1628 td->td_retval[0] = 20 - td->td_retval[0]; 1629 return (error); 1630 } 1631 1632 int 1633 linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1634 { 1635 int name[2]; 1636 1637 name[0] = CTL_KERN; 1638 name[1] = KERN_HOSTNAME; 1639 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1640 args->len, 0, 0)); 1641 } 1642 1643 int 1644 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1645 { 1646 int name[2]; 1647 1648 name[0] = CTL_KERN; 1649 name[1] = KERN_NISDOMAINNAME; 1650 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1651 args->len, 0, 0)); 1652 } 1653 1654 int 1655 linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1656 { 1657 1658 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1659 args->error_code); 1660 1661 /* 1662 * XXX: we should send a signal to the parent if 1663 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1664 * as it doesnt occur often. 1665 */ 1666 exit1(td, args->error_code, 0); 1667 /* NOTREACHED */ 1668 } 1669 1670 #define _LINUX_CAPABILITY_VERSION_1 0x19980330 1671 #define _LINUX_CAPABILITY_VERSION_2 0x20071026 1672 #define _LINUX_CAPABILITY_VERSION_3 0x20080522 1673 1674 struct l_user_cap_header { 1675 l_int version; 1676 l_int pid; 1677 }; 1678 1679 struct l_user_cap_data { 1680 l_int effective; 1681 l_int permitted; 1682 l_int inheritable; 1683 }; 1684 1685 int 1686 linux_capget(struct thread *td, struct linux_capget_args *uap) 1687 { 1688 struct l_user_cap_header luch; 1689 struct l_user_cap_data lucd[2]; 1690 int error, u32s; 1691 1692 if (uap->hdrp == NULL) 1693 return (EFAULT); 1694 1695 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1696 if (error != 0) 1697 return (error); 1698 1699 switch (luch.version) { 1700 case _LINUX_CAPABILITY_VERSION_1: 1701 u32s = 1; 1702 break; 1703 case _LINUX_CAPABILITY_VERSION_2: 1704 case _LINUX_CAPABILITY_VERSION_3: 1705 u32s = 2; 1706 break; 1707 default: 1708 luch.version = _LINUX_CAPABILITY_VERSION_1; 1709 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1710 if (error) 1711 return (error); 1712 return (EINVAL); 1713 } 1714 1715 if (luch.pid) 1716 return (EPERM); 1717 1718 if (uap->datap) { 1719 /* 1720 * The current implementation doesn't support setting 1721 * a capability (it's essentially a stub) so indicate 1722 * that no capabilities are currently set or available 1723 * to request. 1724 */ 1725 memset(&lucd, 0, u32s * sizeof(lucd[0])); 1726 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0])); 1727 } 1728 1729 return (error); 1730 } 1731 1732 int 1733 linux_capset(struct thread *td, struct linux_capset_args *uap) 1734 { 1735 struct l_user_cap_header luch; 1736 struct l_user_cap_data lucd[2]; 1737 int error, i, u32s; 1738 1739 if (uap->hdrp == NULL || uap->datap == NULL) 1740 return (EFAULT); 1741 1742 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1743 if (error != 0) 1744 return (error); 1745 1746 switch (luch.version) { 1747 case _LINUX_CAPABILITY_VERSION_1: 1748 u32s = 1; 1749 break; 1750 case _LINUX_CAPABILITY_VERSION_2: 1751 case _LINUX_CAPABILITY_VERSION_3: 1752 u32s = 2; 1753 break; 1754 default: 1755 luch.version = _LINUX_CAPABILITY_VERSION_1; 1756 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1757 if (error) 1758 return (error); 1759 return (EINVAL); 1760 } 1761 1762 if (luch.pid) 1763 return (EPERM); 1764 1765 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0])); 1766 if (error != 0) 1767 return (error); 1768 1769 /* We currently don't support setting any capabilities. */ 1770 for (i = 0; i < u32s; i++) { 1771 if (lucd[i].effective || lucd[i].permitted || 1772 lucd[i].inheritable) { 1773 linux_msg(td, 1774 "capset[%d] effective=0x%x, permitted=0x%x, " 1775 "inheritable=0x%x is not implemented", i, 1776 (int)lucd[i].effective, (int)lucd[i].permitted, 1777 (int)lucd[i].inheritable); 1778 return (EPERM); 1779 } 1780 } 1781 1782 return (0); 1783 } 1784 1785 int 1786 linux_prctl(struct thread *td, struct linux_prctl_args *args) 1787 { 1788 int error = 0, max_size; 1789 struct proc *p = td->td_proc; 1790 char comm[LINUX_MAX_COMM_LEN]; 1791 int pdeath_signal; 1792 1793 switch (args->option) { 1794 case LINUX_PR_SET_PDEATHSIG: 1795 if (!LINUX_SIG_VALID(args->arg2)) 1796 return (EINVAL); 1797 pdeath_signal = linux_to_bsd_signal(args->arg2); 1798 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL, 1799 &pdeath_signal)); 1800 case LINUX_PR_GET_PDEATHSIG: 1801 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS, 1802 &pdeath_signal); 1803 if (error != 0) 1804 return (error); 1805 pdeath_signal = bsd_to_linux_signal(pdeath_signal); 1806 return (copyout(&pdeath_signal, 1807 (void *)(register_t)args->arg2, 1808 sizeof(pdeath_signal))); 1809 break; 1810 case LINUX_PR_GET_KEEPCAPS: 1811 /* 1812 * Indicate that we always clear the effective and 1813 * permitted capability sets when the user id becomes 1814 * non-zero (actually the capability sets are simply 1815 * always zero in the current implementation). 1816 */ 1817 td->td_retval[0] = 0; 1818 break; 1819 case LINUX_PR_SET_KEEPCAPS: 1820 /* 1821 * Ignore requests to keep the effective and permitted 1822 * capability sets when the user id becomes non-zero. 1823 */ 1824 break; 1825 case LINUX_PR_SET_NAME: 1826 /* 1827 * To be on the safe side we need to make sure to not 1828 * overflow the size a Linux program expects. We already 1829 * do this here in the copyin, so that we don't need to 1830 * check on copyout. 1831 */ 1832 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1833 error = copyinstr((void *)(register_t)args->arg2, comm, 1834 max_size, NULL); 1835 1836 /* Linux silently truncates the name if it is too long. */ 1837 if (error == ENAMETOOLONG) { 1838 /* 1839 * XXX: copyinstr() isn't documented to populate the 1840 * array completely, so do a copyin() to be on the 1841 * safe side. This should be changed in case 1842 * copyinstr() is changed to guarantee this. 1843 */ 1844 error = copyin((void *)(register_t)args->arg2, comm, 1845 max_size - 1); 1846 comm[max_size - 1] = '\0'; 1847 } 1848 if (error) 1849 return (error); 1850 1851 PROC_LOCK(p); 1852 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1853 PROC_UNLOCK(p); 1854 break; 1855 case LINUX_PR_GET_NAME: 1856 PROC_LOCK(p); 1857 strlcpy(comm, p->p_comm, sizeof(comm)); 1858 PROC_UNLOCK(p); 1859 error = copyout(comm, (void *)(register_t)args->arg2, 1860 strlen(comm) + 1); 1861 break; 1862 default: 1863 error = EINVAL; 1864 break; 1865 } 1866 1867 return (error); 1868 } 1869 1870 int 1871 linux_sched_setparam(struct thread *td, 1872 struct linux_sched_setparam_args *uap) 1873 { 1874 struct sched_param sched_param; 1875 struct thread *tdt; 1876 int error; 1877 1878 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1879 if (error) 1880 return (error); 1881 1882 tdt = linux_tdfind(td, uap->pid, -1); 1883 if (tdt == NULL) 1884 return (ESRCH); 1885 1886 error = kern_sched_setparam(td, tdt, &sched_param); 1887 PROC_UNLOCK(tdt->td_proc); 1888 return (error); 1889 } 1890 1891 int 1892 linux_sched_getparam(struct thread *td, 1893 struct linux_sched_getparam_args *uap) 1894 { 1895 struct sched_param sched_param; 1896 struct thread *tdt; 1897 int error; 1898 1899 tdt = linux_tdfind(td, uap->pid, -1); 1900 if (tdt == NULL) 1901 return (ESRCH); 1902 1903 error = kern_sched_getparam(td, tdt, &sched_param); 1904 PROC_UNLOCK(tdt->td_proc); 1905 if (error == 0) 1906 error = copyout(&sched_param, uap->param, 1907 sizeof(sched_param)); 1908 return (error); 1909 } 1910 1911 /* 1912 * Get affinity of a process. 1913 */ 1914 int 1915 linux_sched_getaffinity(struct thread *td, 1916 struct linux_sched_getaffinity_args *args) 1917 { 1918 int error; 1919 struct thread *tdt; 1920 1921 if (args->len < sizeof(cpuset_t)) 1922 return (EINVAL); 1923 1924 tdt = linux_tdfind(td, args->pid, -1); 1925 if (tdt == NULL) 1926 return (ESRCH); 1927 1928 PROC_UNLOCK(tdt->td_proc); 1929 1930 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 1931 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr); 1932 if (error == 0) 1933 td->td_retval[0] = sizeof(cpuset_t); 1934 1935 return (error); 1936 } 1937 1938 /* 1939 * Set affinity of a process. 1940 */ 1941 int 1942 linux_sched_setaffinity(struct thread *td, 1943 struct linux_sched_setaffinity_args *args) 1944 { 1945 struct thread *tdt; 1946 1947 if (args->len < sizeof(cpuset_t)) 1948 return (EINVAL); 1949 1950 tdt = linux_tdfind(td, args->pid, -1); 1951 if (tdt == NULL) 1952 return (ESRCH); 1953 1954 PROC_UNLOCK(tdt->td_proc); 1955 1956 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 1957 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr)); 1958 } 1959 1960 struct linux_rlimit64 { 1961 uint64_t rlim_cur; 1962 uint64_t rlim_max; 1963 }; 1964 1965 int 1966 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 1967 { 1968 struct rlimit rlim, nrlim; 1969 struct linux_rlimit64 lrlim; 1970 struct proc *p; 1971 u_int which; 1972 int flags; 1973 int error; 1974 1975 if (args->resource >= LINUX_RLIM_NLIMITS) 1976 return (EINVAL); 1977 1978 which = linux_to_bsd_resource[args->resource]; 1979 if (which == -1) 1980 return (EINVAL); 1981 1982 if (args->new != NULL) { 1983 /* 1984 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 1985 * rlim is unsigned 64-bit. FreeBSD treats negative limits 1986 * as INFINITY so we do not need a conversion even. 1987 */ 1988 error = copyin(args->new, &nrlim, sizeof(nrlim)); 1989 if (error != 0) 1990 return (error); 1991 } 1992 1993 flags = PGET_HOLD | PGET_NOTWEXIT; 1994 if (args->new != NULL) 1995 flags |= PGET_CANDEBUG; 1996 else 1997 flags |= PGET_CANSEE; 1998 if (args->pid == 0) { 1999 p = td->td_proc; 2000 PHOLD(p); 2001 } else { 2002 error = pget(args->pid, flags, &p); 2003 if (error != 0) 2004 return (error); 2005 } 2006 if (args->old != NULL) { 2007 PROC_LOCK(p); 2008 lim_rlimit_proc(p, which, &rlim); 2009 PROC_UNLOCK(p); 2010 if (rlim.rlim_cur == RLIM_INFINITY) 2011 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2012 else 2013 lrlim.rlim_cur = rlim.rlim_cur; 2014 if (rlim.rlim_max == RLIM_INFINITY) 2015 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2016 else 2017 lrlim.rlim_max = rlim.rlim_max; 2018 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2019 if (error != 0) 2020 goto out; 2021 } 2022 2023 if (args->new != NULL) 2024 error = kern_proc_setrlimit(td, p, which, &nrlim); 2025 2026 out: 2027 PRELE(p); 2028 return (error); 2029 } 2030 2031 int 2032 linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2033 { 2034 struct timeval utv, tv0, tv1, *tvp; 2035 struct l_pselect6arg lpse6; 2036 struct l_timespec lts; 2037 struct timespec uts; 2038 l_sigset_t l_ss; 2039 sigset_t *ssp; 2040 sigset_t ss; 2041 int error; 2042 2043 ssp = NULL; 2044 if (args->sig != NULL) { 2045 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2046 if (error != 0) 2047 return (error); 2048 if (lpse6.ss_len != sizeof(l_ss)) 2049 return (EINVAL); 2050 if (lpse6.ss != 0) { 2051 error = copyin(PTRIN(lpse6.ss), &l_ss, 2052 sizeof(l_ss)); 2053 if (error != 0) 2054 return (error); 2055 linux_to_bsd_sigset(&l_ss, &ss); 2056 ssp = &ss; 2057 } 2058 } 2059 2060 /* 2061 * Currently glibc changes nanosecond number to microsecond. 2062 * This mean losing precision but for now it is hardly seen. 2063 */ 2064 if (args->tsp != NULL) { 2065 error = copyin(args->tsp, <s, sizeof(lts)); 2066 if (error != 0) 2067 return (error); 2068 error = linux_to_native_timespec(&uts, <s); 2069 if (error != 0) 2070 return (error); 2071 2072 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2073 if (itimerfix(&utv)) 2074 return (EINVAL); 2075 2076 microtime(&tv0); 2077 tvp = &utv; 2078 } else 2079 tvp = NULL; 2080 2081 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2082 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2083 2084 if (error == 0 && args->tsp != NULL) { 2085 if (td->td_retval[0] != 0) { 2086 /* 2087 * Compute how much time was left of the timeout, 2088 * by subtracting the current time and the time 2089 * before we started the call, and subtracting 2090 * that result from the user-supplied value. 2091 */ 2092 2093 microtime(&tv1); 2094 timevalsub(&tv1, &tv0); 2095 timevalsub(&utv, &tv1); 2096 if (utv.tv_sec < 0) 2097 timevalclear(&utv); 2098 } else 2099 timevalclear(&utv); 2100 2101 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2102 2103 error = native_to_linux_timespec(<s, &uts); 2104 if (error == 0) 2105 error = copyout(<s, args->tsp, sizeof(lts)); 2106 } 2107 2108 return (error); 2109 } 2110 2111 int 2112 linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2113 { 2114 struct timespec ts0, ts1; 2115 struct l_timespec lts; 2116 struct timespec uts, *tsp; 2117 l_sigset_t l_ss; 2118 sigset_t *ssp; 2119 sigset_t ss; 2120 int error; 2121 2122 if (args->sset != NULL) { 2123 if (args->ssize != sizeof(l_ss)) 2124 return (EINVAL); 2125 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2126 if (error) 2127 return (error); 2128 linux_to_bsd_sigset(&l_ss, &ss); 2129 ssp = &ss; 2130 } else 2131 ssp = NULL; 2132 if (args->tsp != NULL) { 2133 error = copyin(args->tsp, <s, sizeof(lts)); 2134 if (error) 2135 return (error); 2136 error = linux_to_native_timespec(&uts, <s); 2137 if (error != 0) 2138 return (error); 2139 2140 nanotime(&ts0); 2141 tsp = &uts; 2142 } else 2143 tsp = NULL; 2144 2145 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2146 2147 if (error == 0 && args->tsp != NULL) { 2148 if (td->td_retval[0]) { 2149 nanotime(&ts1); 2150 timespecsub(&ts1, &ts0, &ts1); 2151 timespecsub(&uts, &ts1, &uts); 2152 if (uts.tv_sec < 0) 2153 timespecclear(&uts); 2154 } else 2155 timespecclear(&uts); 2156 2157 error = native_to_linux_timespec(<s, &uts); 2158 if (error == 0) 2159 error = copyout(<s, args->tsp, sizeof(lts)); 2160 } 2161 2162 return (error); 2163 } 2164 2165 int 2166 linux_sched_rr_get_interval(struct thread *td, 2167 struct linux_sched_rr_get_interval_args *uap) 2168 { 2169 struct timespec ts; 2170 struct l_timespec lts; 2171 struct thread *tdt; 2172 int error; 2173 2174 /* 2175 * According to man in case the invalid pid specified 2176 * EINVAL should be returned. 2177 */ 2178 if (uap->pid < 0) 2179 return (EINVAL); 2180 2181 tdt = linux_tdfind(td, uap->pid, -1); 2182 if (tdt == NULL) 2183 return (ESRCH); 2184 2185 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2186 PROC_UNLOCK(tdt->td_proc); 2187 if (error != 0) 2188 return (error); 2189 error = native_to_linux_timespec(<s, &ts); 2190 if (error != 0) 2191 return (error); 2192 return (copyout(<s, uap->interval, sizeof(lts))); 2193 } 2194 2195 /* 2196 * In case when the Linux thread is the initial thread in 2197 * the thread group thread id is equal to the process id. 2198 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2199 */ 2200 struct thread * 2201 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2202 { 2203 struct linux_emuldata *em; 2204 struct thread *tdt; 2205 struct proc *p; 2206 2207 tdt = NULL; 2208 if (tid == 0 || tid == td->td_tid) { 2209 tdt = td; 2210 PROC_LOCK(tdt->td_proc); 2211 } else if (tid > PID_MAX) 2212 tdt = tdfind(tid, pid); 2213 else { 2214 /* 2215 * Initial thread where the tid equal to the pid. 2216 */ 2217 p = pfind(tid); 2218 if (p != NULL) { 2219 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2220 /* 2221 * p is not a Linuxulator process. 2222 */ 2223 PROC_UNLOCK(p); 2224 return (NULL); 2225 } 2226 FOREACH_THREAD_IN_PROC(p, tdt) { 2227 em = em_find(tdt); 2228 if (tid == em->em_tid) 2229 return (tdt); 2230 } 2231 PROC_UNLOCK(p); 2232 } 2233 return (NULL); 2234 } 2235 2236 return (tdt); 2237 } 2238 2239 void 2240 linux_to_bsd_waitopts(int options, int *bsdopts) 2241 { 2242 2243 if (options & LINUX_WNOHANG) 2244 *bsdopts |= WNOHANG; 2245 if (options & LINUX_WUNTRACED) 2246 *bsdopts |= WUNTRACED; 2247 if (options & LINUX_WEXITED) 2248 *bsdopts |= WEXITED; 2249 if (options & LINUX_WCONTINUED) 2250 *bsdopts |= WCONTINUED; 2251 if (options & LINUX_WNOWAIT) 2252 *bsdopts |= WNOWAIT; 2253 2254 if (options & __WCLONE) 2255 *bsdopts |= WLINUXCLONE; 2256 } 2257 2258 int 2259 linux_getrandom(struct thread *td, struct linux_getrandom_args *args) 2260 { 2261 struct uio uio; 2262 struct iovec iov; 2263 int error; 2264 2265 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM)) 2266 return (EINVAL); 2267 if (args->count > INT_MAX) 2268 args->count = INT_MAX; 2269 2270 iov.iov_base = args->buf; 2271 iov.iov_len = args->count; 2272 2273 uio.uio_iov = &iov; 2274 uio.uio_iovcnt = 1; 2275 uio.uio_resid = iov.iov_len; 2276 uio.uio_segflg = UIO_USERSPACE; 2277 uio.uio_rw = UIO_READ; 2278 uio.uio_td = td; 2279 2280 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK); 2281 if (error == 0) 2282 td->td_retval[0] = args->count - uio.uio_resid; 2283 return (error); 2284 } 2285 2286 int 2287 linux_mincore(struct thread *td, struct linux_mincore_args *args) 2288 { 2289 2290 /* Needs to be page-aligned */ 2291 if (args->start & PAGE_MASK) 2292 return (EINVAL); 2293 return (kern_mincore(td, args->start, args->len, args->vec)); 2294 } 2295