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 VOP_UNSET_TEXT_CHECKED(vp); 477 if (locked) 478 VOP_UNLOCK(vp, 0); 479 480 /* Release the temporary mapping. */ 481 if (a_out) 482 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 483 484 return (error); 485 } 486 487 #endif /* __i386__ */ 488 489 #ifdef LINUX_LEGACY_SYSCALLS 490 int 491 linux_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 /* 498 * Store current time for computation of the amount of 499 * time left. 500 */ 501 if (args->timeout) { 502 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 503 goto select_out; 504 utv.tv_sec = ltv.tv_sec; 505 utv.tv_usec = ltv.tv_usec; 506 507 if (itimerfix(&utv)) { 508 /* 509 * The timeval was invalid. Convert it to something 510 * valid that will act as it does under Linux. 511 */ 512 utv.tv_sec += utv.tv_usec / 1000000; 513 utv.tv_usec %= 1000000; 514 if (utv.tv_usec < 0) { 515 utv.tv_sec -= 1; 516 utv.tv_usec += 1000000; 517 } 518 if (utv.tv_sec < 0) 519 timevalclear(&utv); 520 } 521 microtime(&tv0); 522 tvp = &utv; 523 } else 524 tvp = NULL; 525 526 error = kern_select(td, args->nfds, args->readfds, args->writefds, 527 args->exceptfds, tvp, LINUX_NFDBITS); 528 if (error) 529 goto select_out; 530 531 if (args->timeout) { 532 if (td->td_retval[0]) { 533 /* 534 * Compute how much time was left of the timeout, 535 * by subtracting the current time and the time 536 * before we started the call, and subtracting 537 * that result from the user-supplied value. 538 */ 539 microtime(&tv1); 540 timevalsub(&tv1, &tv0); 541 timevalsub(&utv, &tv1); 542 if (utv.tv_sec < 0) 543 timevalclear(&utv); 544 } else 545 timevalclear(&utv); 546 ltv.tv_sec = utv.tv_sec; 547 ltv.tv_usec = utv.tv_usec; 548 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 549 goto select_out; 550 } 551 552 select_out: 553 return (error); 554 } 555 #endif 556 557 int 558 linux_mremap(struct thread *td, struct linux_mremap_args *args) 559 { 560 uintptr_t addr; 561 size_t len; 562 int error = 0; 563 564 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 565 td->td_retval[0] = 0; 566 return (EINVAL); 567 } 568 569 /* 570 * Check for the page alignment. 571 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 572 */ 573 if (args->addr & PAGE_MASK) { 574 td->td_retval[0] = 0; 575 return (EINVAL); 576 } 577 578 args->new_len = round_page(args->new_len); 579 args->old_len = round_page(args->old_len); 580 581 if (args->new_len > args->old_len) { 582 td->td_retval[0] = 0; 583 return (ENOMEM); 584 } 585 586 if (args->new_len < args->old_len) { 587 addr = args->addr + args->new_len; 588 len = args->old_len - args->new_len; 589 error = kern_munmap(td, addr, len); 590 } 591 592 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 593 return (error); 594 } 595 596 #define LINUX_MS_ASYNC 0x0001 597 #define LINUX_MS_INVALIDATE 0x0002 598 #define LINUX_MS_SYNC 0x0004 599 600 int 601 linux_msync(struct thread *td, struct linux_msync_args *args) 602 { 603 604 return (kern_msync(td, args->addr, args->len, 605 args->fl & ~LINUX_MS_SYNC)); 606 } 607 608 #ifdef LINUX_LEGACY_SYSCALLS 609 int 610 linux_time(struct thread *td, struct linux_time_args *args) 611 { 612 struct timeval tv; 613 l_time_t tm; 614 int error; 615 616 microtime(&tv); 617 tm = tv.tv_sec; 618 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 619 return (error); 620 td->td_retval[0] = tm; 621 return (0); 622 } 623 #endif 624 625 struct l_times_argv { 626 l_clock_t tms_utime; 627 l_clock_t tms_stime; 628 l_clock_t tms_cutime; 629 l_clock_t tms_cstime; 630 }; 631 632 633 /* 634 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 635 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 636 * auxiliary vector entry. 637 */ 638 #define CLK_TCK 100 639 640 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 641 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 642 643 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 644 CONVNTCK(r) : CONVOTCK(r)) 645 646 int 647 linux_times(struct thread *td, struct linux_times_args *args) 648 { 649 struct timeval tv, utime, stime, cutime, cstime; 650 struct l_times_argv tms; 651 struct proc *p; 652 int error; 653 654 if (args->buf != NULL) { 655 p = td->td_proc; 656 PROC_LOCK(p); 657 PROC_STATLOCK(p); 658 calcru(p, &utime, &stime); 659 PROC_STATUNLOCK(p); 660 calccru(p, &cutime, &cstime); 661 PROC_UNLOCK(p); 662 663 tms.tms_utime = CONVTCK(utime); 664 tms.tms_stime = CONVTCK(stime); 665 666 tms.tms_cutime = CONVTCK(cutime); 667 tms.tms_cstime = CONVTCK(cstime); 668 669 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 670 return (error); 671 } 672 673 microuptime(&tv); 674 td->td_retval[0] = (int)CONVTCK(tv); 675 return (0); 676 } 677 678 int 679 linux_newuname(struct thread *td, struct linux_newuname_args *args) 680 { 681 struct l_new_utsname utsname; 682 char osname[LINUX_MAX_UTSNAME]; 683 char osrelease[LINUX_MAX_UTSNAME]; 684 char *p; 685 686 linux_get_osname(td, osname); 687 linux_get_osrelease(td, osrelease); 688 689 bzero(&utsname, sizeof(utsname)); 690 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 691 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 692 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 693 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 694 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 695 for (p = utsname.version; *p != '\0'; ++p) 696 if (*p == '\n') { 697 *p = '\0'; 698 break; 699 } 700 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 701 702 return (copyout(&utsname, args->buf, sizeof(utsname))); 703 } 704 705 struct l_utimbuf { 706 l_time_t l_actime; 707 l_time_t l_modtime; 708 }; 709 710 #ifdef LINUX_LEGACY_SYSCALLS 711 int 712 linux_utime(struct thread *td, struct linux_utime_args *args) 713 { 714 struct timeval tv[2], *tvp; 715 struct l_utimbuf lut; 716 char *fname; 717 int error; 718 719 LCONVPATHEXIST(td, args->fname, &fname); 720 721 if (args->times) { 722 if ((error = copyin(args->times, &lut, sizeof lut))) { 723 LFREEPATH(fname); 724 return (error); 725 } 726 tv[0].tv_sec = lut.l_actime; 727 tv[0].tv_usec = 0; 728 tv[1].tv_sec = lut.l_modtime; 729 tv[1].tv_usec = 0; 730 tvp = tv; 731 } else 732 tvp = NULL; 733 734 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp, 735 UIO_SYSSPACE); 736 LFREEPATH(fname); 737 return (error); 738 } 739 #endif 740 741 #ifdef LINUX_LEGACY_SYSCALLS 742 int 743 linux_utimes(struct thread *td, struct linux_utimes_args *args) 744 { 745 l_timeval ltv[2]; 746 struct timeval tv[2], *tvp = NULL; 747 char *fname; 748 int error; 749 750 LCONVPATHEXIST(td, args->fname, &fname); 751 752 if (args->tptr != NULL) { 753 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 754 LFREEPATH(fname); 755 return (error); 756 } 757 tv[0].tv_sec = ltv[0].tv_sec; 758 tv[0].tv_usec = ltv[0].tv_usec; 759 tv[1].tv_sec = ltv[1].tv_sec; 760 tv[1].tv_usec = ltv[1].tv_usec; 761 tvp = tv; 762 } 763 764 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, 765 tvp, UIO_SYSSPACE); 766 LFREEPATH(fname); 767 return (error); 768 } 769 #endif 770 771 static int 772 linux_utimensat_nsec_valid(l_long nsec) 773 { 774 775 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW) 776 return (0); 777 if (nsec >= 0 && nsec <= 999999999) 778 return (0); 779 return (1); 780 } 781 782 int 783 linux_utimensat(struct thread *td, struct linux_utimensat_args *args) 784 { 785 struct l_timespec l_times[2]; 786 struct timespec times[2], *timesp = NULL; 787 char *path = NULL; 788 int error, dfd, flags = 0; 789 790 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 791 792 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 793 return (EINVAL); 794 795 if (args->times != NULL) { 796 error = copyin(args->times, l_times, sizeof(l_times)); 797 if (error != 0) 798 return (error); 799 800 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 || 801 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0) 802 return (EINVAL); 803 804 times[0].tv_sec = l_times[0].tv_sec; 805 switch (l_times[0].tv_nsec) 806 { 807 case LINUX_UTIME_OMIT: 808 times[0].tv_nsec = UTIME_OMIT; 809 break; 810 case LINUX_UTIME_NOW: 811 times[0].tv_nsec = UTIME_NOW; 812 break; 813 default: 814 times[0].tv_nsec = l_times[0].tv_nsec; 815 } 816 817 times[1].tv_sec = l_times[1].tv_sec; 818 switch (l_times[1].tv_nsec) 819 { 820 case LINUX_UTIME_OMIT: 821 times[1].tv_nsec = UTIME_OMIT; 822 break; 823 case LINUX_UTIME_NOW: 824 times[1].tv_nsec = UTIME_NOW; 825 break; 826 default: 827 times[1].tv_nsec = l_times[1].tv_nsec; 828 break; 829 } 830 timesp = times; 831 832 /* This breaks POSIX, but is what the Linux kernel does 833 * _on purpose_ (documented in the man page for utimensat(2)), 834 * so we must follow that behaviour. */ 835 if (times[0].tv_nsec == UTIME_OMIT && 836 times[1].tv_nsec == UTIME_OMIT) 837 return (0); 838 } 839 840 if (args->pathname != NULL) 841 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 842 else if (args->flags != 0) 843 return (EINVAL); 844 845 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 846 flags |= AT_SYMLINK_NOFOLLOW; 847 848 if (path == NULL) 849 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 850 else { 851 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 852 UIO_SYSSPACE, flags); 853 LFREEPATH(path); 854 } 855 856 return (error); 857 } 858 859 #ifdef LINUX_LEGACY_SYSCALLS 860 int 861 linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 862 { 863 l_timeval ltv[2]; 864 struct timeval tv[2], *tvp = NULL; 865 char *fname; 866 int error, dfd; 867 868 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 869 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 870 871 if (args->utimes != NULL) { 872 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 873 LFREEPATH(fname); 874 return (error); 875 } 876 tv[0].tv_sec = ltv[0].tv_sec; 877 tv[0].tv_usec = ltv[0].tv_usec; 878 tv[1].tv_sec = ltv[1].tv_sec; 879 tv[1].tv_usec = ltv[1].tv_usec; 880 tvp = tv; 881 } 882 883 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 884 LFREEPATH(fname); 885 return (error); 886 } 887 #endif 888 889 static int 890 linux_common_wait(struct thread *td, int pid, int *statusp, 891 int options, struct __wrusage *wrup) 892 { 893 siginfo_t siginfo; 894 idtype_t idtype; 895 id_t id; 896 int error, status, tmpstat; 897 898 if (pid == WAIT_ANY) { 899 idtype = P_ALL; 900 id = 0; 901 } else if (pid < 0) { 902 idtype = P_PGID; 903 id = (id_t)-pid; 904 } else { 905 idtype = P_PID; 906 id = (id_t)pid; 907 } 908 909 /* 910 * For backward compatibility we implicitly add flags WEXITED 911 * and WTRAPPED here. 912 */ 913 options |= WEXITED | WTRAPPED; 914 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo); 915 if (error) 916 return (error); 917 918 if (statusp) { 919 tmpstat = status & 0xffff; 920 if (WIFSIGNALED(tmpstat)) { 921 tmpstat = (tmpstat & 0xffffff80) | 922 bsd_to_linux_signal(WTERMSIG(tmpstat)); 923 } else if (WIFSTOPPED(tmpstat)) { 924 tmpstat = (tmpstat & 0xffff00ff) | 925 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 926 #if defined(__amd64__) && !defined(COMPAT_LINUX32) 927 if (WSTOPSIG(status) == SIGTRAP) { 928 tmpstat = linux_ptrace_status(td, 929 siginfo.si_pid, tmpstat); 930 } 931 #endif 932 } else if (WIFCONTINUED(tmpstat)) { 933 tmpstat = 0xffff; 934 } 935 error = copyout(&tmpstat, statusp, sizeof(int)); 936 } 937 938 return (error); 939 } 940 941 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 942 int 943 linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 944 { 945 struct linux_wait4_args wait4_args; 946 947 wait4_args.pid = args->pid; 948 wait4_args.status = args->status; 949 wait4_args.options = args->options; 950 wait4_args.rusage = NULL; 951 952 return (linux_wait4(td, &wait4_args)); 953 } 954 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 955 956 int 957 linux_wait4(struct thread *td, struct linux_wait4_args *args) 958 { 959 int error, options; 960 struct __wrusage wru, *wrup; 961 962 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 963 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 964 return (EINVAL); 965 966 options = WEXITED; 967 linux_to_bsd_waitopts(args->options, &options); 968 969 if (args->rusage != NULL) 970 wrup = &wru; 971 else 972 wrup = NULL; 973 error = linux_common_wait(td, args->pid, args->status, options, wrup); 974 if (error != 0) 975 return (error); 976 if (args->rusage != NULL) 977 error = linux_copyout_rusage(&wru.wru_self, args->rusage); 978 return (error); 979 } 980 981 int 982 linux_waitid(struct thread *td, struct linux_waitid_args *args) 983 { 984 int status, options, sig; 985 struct __wrusage wru; 986 siginfo_t siginfo; 987 l_siginfo_t lsi; 988 idtype_t idtype; 989 struct proc *p; 990 int error; 991 992 options = 0; 993 linux_to_bsd_waitopts(args->options, &options); 994 995 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 996 return (EINVAL); 997 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 998 return (EINVAL); 999 1000 switch (args->idtype) { 1001 case LINUX_P_ALL: 1002 idtype = P_ALL; 1003 break; 1004 case LINUX_P_PID: 1005 if (args->id <= 0) 1006 return (EINVAL); 1007 idtype = P_PID; 1008 break; 1009 case LINUX_P_PGID: 1010 if (args->id <= 0) 1011 return (EINVAL); 1012 idtype = P_PGID; 1013 break; 1014 default: 1015 return (EINVAL); 1016 } 1017 1018 error = kern_wait6(td, idtype, args->id, &status, options, 1019 &wru, &siginfo); 1020 if (error != 0) 1021 return (error); 1022 if (args->rusage != NULL) { 1023 error = linux_copyout_rusage(&wru.wru_children, 1024 args->rusage); 1025 if (error != 0) 1026 return (error); 1027 } 1028 if (args->info != NULL) { 1029 p = td->td_proc; 1030 bzero(&lsi, sizeof(lsi)); 1031 if (td->td_retval[0] != 0) { 1032 sig = bsd_to_linux_signal(siginfo.si_signo); 1033 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1034 } 1035 error = copyout(&lsi, args->info, sizeof(lsi)); 1036 } 1037 td->td_retval[0] = 0; 1038 1039 return (error); 1040 } 1041 1042 #ifdef LINUX_LEGACY_SYSCALLS 1043 int 1044 linux_mknod(struct thread *td, struct linux_mknod_args *args) 1045 { 1046 char *path; 1047 int error; 1048 1049 LCONVPATHCREAT(td, args->path, &path); 1050 1051 switch (args->mode & S_IFMT) { 1052 case S_IFIFO: 1053 case S_IFSOCK: 1054 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE, 1055 args->mode); 1056 break; 1057 1058 case S_IFCHR: 1059 case S_IFBLK: 1060 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE, 1061 args->mode, args->dev); 1062 break; 1063 1064 case S_IFDIR: 1065 error = EPERM; 1066 break; 1067 1068 case 0: 1069 args->mode |= S_IFREG; 1070 /* FALLTHROUGH */ 1071 case S_IFREG: 1072 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE, 1073 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1074 if (error == 0) 1075 kern_close(td, td->td_retval[0]); 1076 break; 1077 1078 default: 1079 error = EINVAL; 1080 break; 1081 } 1082 LFREEPATH(path); 1083 return (error); 1084 } 1085 #endif 1086 1087 int 1088 linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1089 { 1090 char *path; 1091 int error, dfd; 1092 1093 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1094 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1095 1096 switch (args->mode & S_IFMT) { 1097 case S_IFIFO: 1098 case S_IFSOCK: 1099 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1100 break; 1101 1102 case S_IFCHR: 1103 case S_IFBLK: 1104 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1105 args->dev); 1106 break; 1107 1108 case S_IFDIR: 1109 error = EPERM; 1110 break; 1111 1112 case 0: 1113 args->mode |= S_IFREG; 1114 /* FALLTHROUGH */ 1115 case S_IFREG: 1116 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1117 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1118 if (error == 0) 1119 kern_close(td, td->td_retval[0]); 1120 break; 1121 1122 default: 1123 error = EINVAL; 1124 break; 1125 } 1126 LFREEPATH(path); 1127 return (error); 1128 } 1129 1130 /* 1131 * UGH! This is just about the dumbest idea I've ever heard!! 1132 */ 1133 int 1134 linux_personality(struct thread *td, struct linux_personality_args *args) 1135 { 1136 struct linux_pemuldata *pem; 1137 struct proc *p = td->td_proc; 1138 uint32_t old; 1139 1140 PROC_LOCK(p); 1141 pem = pem_find(p); 1142 old = pem->persona; 1143 if (args->per != 0xffffffff) 1144 pem->persona = args->per; 1145 PROC_UNLOCK(p); 1146 1147 td->td_retval[0] = old; 1148 return (0); 1149 } 1150 1151 struct l_itimerval { 1152 l_timeval it_interval; 1153 l_timeval it_value; 1154 }; 1155 1156 #define B2L_ITIMERVAL(bip, lip) \ 1157 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1158 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1159 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1160 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1161 1162 int 1163 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1164 { 1165 int error; 1166 struct l_itimerval ls; 1167 struct itimerval aitv, oitv; 1168 1169 if (uap->itv == NULL) { 1170 uap->itv = uap->oitv; 1171 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1172 } 1173 1174 error = copyin(uap->itv, &ls, sizeof(ls)); 1175 if (error != 0) 1176 return (error); 1177 B2L_ITIMERVAL(&aitv, &ls); 1178 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1179 if (error != 0 || uap->oitv == NULL) 1180 return (error); 1181 B2L_ITIMERVAL(&ls, &oitv); 1182 1183 return (copyout(&ls, uap->oitv, sizeof(ls))); 1184 } 1185 1186 int 1187 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1188 { 1189 int error; 1190 struct l_itimerval ls; 1191 struct itimerval aitv; 1192 1193 error = kern_getitimer(td, uap->which, &aitv); 1194 if (error != 0) 1195 return (error); 1196 B2L_ITIMERVAL(&ls, &aitv); 1197 return (copyout(&ls, uap->itv, sizeof(ls))); 1198 } 1199 1200 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1201 int 1202 linux_nice(struct thread *td, struct linux_nice_args *args) 1203 { 1204 struct setpriority_args bsd_args; 1205 1206 bsd_args.which = PRIO_PROCESS; 1207 bsd_args.who = 0; /* current process */ 1208 bsd_args.prio = args->inc; 1209 return (sys_setpriority(td, &bsd_args)); 1210 } 1211 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1212 1213 int 1214 linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1215 { 1216 struct ucred *newcred, *oldcred; 1217 l_gid_t *linux_gidset; 1218 gid_t *bsd_gidset; 1219 int ngrp, error; 1220 struct proc *p; 1221 1222 ngrp = args->gidsetsize; 1223 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1224 return (EINVAL); 1225 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1226 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1227 if (error) 1228 goto out; 1229 newcred = crget(); 1230 crextend(newcred, ngrp + 1); 1231 p = td->td_proc; 1232 PROC_LOCK(p); 1233 oldcred = p->p_ucred; 1234 crcopy(newcred, oldcred); 1235 1236 /* 1237 * cr_groups[0] holds egid. Setting the whole set from 1238 * the supplied set will cause egid to be changed too. 1239 * Keep cr_groups[0] unchanged to prevent that. 1240 */ 1241 1242 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) { 1243 PROC_UNLOCK(p); 1244 crfree(newcred); 1245 goto out; 1246 } 1247 1248 if (ngrp > 0) { 1249 newcred->cr_ngroups = ngrp + 1; 1250 1251 bsd_gidset = newcred->cr_groups; 1252 ngrp--; 1253 while (ngrp >= 0) { 1254 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1255 ngrp--; 1256 } 1257 } else 1258 newcred->cr_ngroups = 1; 1259 1260 setsugid(p); 1261 proc_set_cred(p, newcred); 1262 PROC_UNLOCK(p); 1263 crfree(oldcred); 1264 error = 0; 1265 out: 1266 free(linux_gidset, M_LINUX); 1267 return (error); 1268 } 1269 1270 int 1271 linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1272 { 1273 struct ucred *cred; 1274 l_gid_t *linux_gidset; 1275 gid_t *bsd_gidset; 1276 int bsd_gidsetsz, ngrp, error; 1277 1278 cred = td->td_ucred; 1279 bsd_gidset = cred->cr_groups; 1280 bsd_gidsetsz = cred->cr_ngroups - 1; 1281 1282 /* 1283 * cr_groups[0] holds egid. Returning the whole set 1284 * here will cause a duplicate. Exclude cr_groups[0] 1285 * to prevent that. 1286 */ 1287 1288 if ((ngrp = args->gidsetsize) == 0) { 1289 td->td_retval[0] = bsd_gidsetsz; 1290 return (0); 1291 } 1292 1293 if (ngrp < bsd_gidsetsz) 1294 return (EINVAL); 1295 1296 ngrp = 0; 1297 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1298 M_LINUX, M_WAITOK); 1299 while (ngrp < bsd_gidsetsz) { 1300 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1301 ngrp++; 1302 } 1303 1304 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1305 free(linux_gidset, M_LINUX); 1306 if (error) 1307 return (error); 1308 1309 td->td_retval[0] = ngrp; 1310 return (0); 1311 } 1312 1313 int 1314 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1315 { 1316 struct rlimit bsd_rlim; 1317 struct l_rlimit rlim; 1318 u_int which; 1319 int error; 1320 1321 if (args->resource >= LINUX_RLIM_NLIMITS) 1322 return (EINVAL); 1323 1324 which = linux_to_bsd_resource[args->resource]; 1325 if (which == -1) 1326 return (EINVAL); 1327 1328 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1329 if (error) 1330 return (error); 1331 1332 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1333 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1334 return (kern_setrlimit(td, which, &bsd_rlim)); 1335 } 1336 1337 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1338 int 1339 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1340 { 1341 struct l_rlimit rlim; 1342 struct rlimit bsd_rlim; 1343 u_int which; 1344 1345 if (args->resource >= LINUX_RLIM_NLIMITS) 1346 return (EINVAL); 1347 1348 which = linux_to_bsd_resource[args->resource]; 1349 if (which == -1) 1350 return (EINVAL); 1351 1352 lim_rlimit(td, which, &bsd_rlim); 1353 1354 #ifdef COMPAT_LINUX32 1355 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1356 if (rlim.rlim_cur == UINT_MAX) 1357 rlim.rlim_cur = INT_MAX; 1358 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1359 if (rlim.rlim_max == UINT_MAX) 1360 rlim.rlim_max = INT_MAX; 1361 #else 1362 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1363 if (rlim.rlim_cur == ULONG_MAX) 1364 rlim.rlim_cur = LONG_MAX; 1365 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1366 if (rlim.rlim_max == ULONG_MAX) 1367 rlim.rlim_max = LONG_MAX; 1368 #endif 1369 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1370 } 1371 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1372 1373 int 1374 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1375 { 1376 struct l_rlimit rlim; 1377 struct rlimit bsd_rlim; 1378 u_int which; 1379 1380 if (args->resource >= LINUX_RLIM_NLIMITS) 1381 return (EINVAL); 1382 1383 which = linux_to_bsd_resource[args->resource]; 1384 if (which == -1) 1385 return (EINVAL); 1386 1387 lim_rlimit(td, which, &bsd_rlim); 1388 1389 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1390 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1391 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1392 } 1393 1394 int 1395 linux_sched_setscheduler(struct thread *td, 1396 struct linux_sched_setscheduler_args *args) 1397 { 1398 struct sched_param sched_param; 1399 struct thread *tdt; 1400 int error, policy; 1401 1402 switch (args->policy) { 1403 case LINUX_SCHED_OTHER: 1404 policy = SCHED_OTHER; 1405 break; 1406 case LINUX_SCHED_FIFO: 1407 policy = SCHED_FIFO; 1408 break; 1409 case LINUX_SCHED_RR: 1410 policy = SCHED_RR; 1411 break; 1412 default: 1413 return (EINVAL); 1414 } 1415 1416 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1417 if (error) 1418 return (error); 1419 1420 tdt = linux_tdfind(td, args->pid, -1); 1421 if (tdt == NULL) 1422 return (ESRCH); 1423 1424 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1425 PROC_UNLOCK(tdt->td_proc); 1426 return (error); 1427 } 1428 1429 int 1430 linux_sched_getscheduler(struct thread *td, 1431 struct linux_sched_getscheduler_args *args) 1432 { 1433 struct thread *tdt; 1434 int error, policy; 1435 1436 tdt = linux_tdfind(td, args->pid, -1); 1437 if (tdt == NULL) 1438 return (ESRCH); 1439 1440 error = kern_sched_getscheduler(td, tdt, &policy); 1441 PROC_UNLOCK(tdt->td_proc); 1442 1443 switch (policy) { 1444 case SCHED_OTHER: 1445 td->td_retval[0] = LINUX_SCHED_OTHER; 1446 break; 1447 case SCHED_FIFO: 1448 td->td_retval[0] = LINUX_SCHED_FIFO; 1449 break; 1450 case SCHED_RR: 1451 td->td_retval[0] = LINUX_SCHED_RR; 1452 break; 1453 } 1454 return (error); 1455 } 1456 1457 int 1458 linux_sched_get_priority_max(struct thread *td, 1459 struct linux_sched_get_priority_max_args *args) 1460 { 1461 struct sched_get_priority_max_args bsd; 1462 1463 switch (args->policy) { 1464 case LINUX_SCHED_OTHER: 1465 bsd.policy = SCHED_OTHER; 1466 break; 1467 case LINUX_SCHED_FIFO: 1468 bsd.policy = SCHED_FIFO; 1469 break; 1470 case LINUX_SCHED_RR: 1471 bsd.policy = SCHED_RR; 1472 break; 1473 default: 1474 return (EINVAL); 1475 } 1476 return (sys_sched_get_priority_max(td, &bsd)); 1477 } 1478 1479 int 1480 linux_sched_get_priority_min(struct thread *td, 1481 struct linux_sched_get_priority_min_args *args) 1482 { 1483 struct sched_get_priority_min_args bsd; 1484 1485 switch (args->policy) { 1486 case LINUX_SCHED_OTHER: 1487 bsd.policy = SCHED_OTHER; 1488 break; 1489 case LINUX_SCHED_FIFO: 1490 bsd.policy = SCHED_FIFO; 1491 break; 1492 case LINUX_SCHED_RR: 1493 bsd.policy = SCHED_RR; 1494 break; 1495 default: 1496 return (EINVAL); 1497 } 1498 return (sys_sched_get_priority_min(td, &bsd)); 1499 } 1500 1501 #define REBOOT_CAD_ON 0x89abcdef 1502 #define REBOOT_CAD_OFF 0 1503 #define REBOOT_HALT 0xcdef0123 1504 #define REBOOT_RESTART 0x01234567 1505 #define REBOOT_RESTART2 0xA1B2C3D4 1506 #define REBOOT_POWEROFF 0x4321FEDC 1507 #define REBOOT_MAGIC1 0xfee1dead 1508 #define REBOOT_MAGIC2 0x28121969 1509 #define REBOOT_MAGIC2A 0x05121996 1510 #define REBOOT_MAGIC2B 0x16041998 1511 1512 int 1513 linux_reboot(struct thread *td, struct linux_reboot_args *args) 1514 { 1515 struct reboot_args bsd_args; 1516 1517 if (args->magic1 != REBOOT_MAGIC1) 1518 return (EINVAL); 1519 1520 switch (args->magic2) { 1521 case REBOOT_MAGIC2: 1522 case REBOOT_MAGIC2A: 1523 case REBOOT_MAGIC2B: 1524 break; 1525 default: 1526 return (EINVAL); 1527 } 1528 1529 switch (args->cmd) { 1530 case REBOOT_CAD_ON: 1531 case REBOOT_CAD_OFF: 1532 return (priv_check(td, PRIV_REBOOT)); 1533 case REBOOT_HALT: 1534 bsd_args.opt = RB_HALT; 1535 break; 1536 case REBOOT_RESTART: 1537 case REBOOT_RESTART2: 1538 bsd_args.opt = 0; 1539 break; 1540 case REBOOT_POWEROFF: 1541 bsd_args.opt = RB_POWEROFF; 1542 break; 1543 default: 1544 return (EINVAL); 1545 } 1546 return (sys_reboot(td, &bsd_args)); 1547 } 1548 1549 1550 /* 1551 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1552 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1553 * are assumed to be preserved. The following lightweight syscalls fixes 1554 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1555 * 1556 * linux_getpid() - MP SAFE 1557 * linux_getgid() - MP SAFE 1558 * linux_getuid() - MP SAFE 1559 */ 1560 1561 int 1562 linux_getpid(struct thread *td, struct linux_getpid_args *args) 1563 { 1564 1565 td->td_retval[0] = td->td_proc->p_pid; 1566 1567 return (0); 1568 } 1569 1570 int 1571 linux_gettid(struct thread *td, struct linux_gettid_args *args) 1572 { 1573 struct linux_emuldata *em; 1574 1575 em = em_find(td); 1576 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1577 1578 td->td_retval[0] = em->em_tid; 1579 1580 return (0); 1581 } 1582 1583 1584 int 1585 linux_getppid(struct thread *td, struct linux_getppid_args *args) 1586 { 1587 1588 td->td_retval[0] = kern_getppid(td); 1589 return (0); 1590 } 1591 1592 int 1593 linux_getgid(struct thread *td, struct linux_getgid_args *args) 1594 { 1595 1596 td->td_retval[0] = td->td_ucred->cr_rgid; 1597 return (0); 1598 } 1599 1600 int 1601 linux_getuid(struct thread *td, struct linux_getuid_args *args) 1602 { 1603 1604 td->td_retval[0] = td->td_ucred->cr_ruid; 1605 return (0); 1606 } 1607 1608 1609 int 1610 linux_getsid(struct thread *td, struct linux_getsid_args *args) 1611 { 1612 struct getsid_args bsd; 1613 1614 bsd.pid = args->pid; 1615 return (sys_getsid(td, &bsd)); 1616 } 1617 1618 int 1619 linux_nosys(struct thread *td, struct nosys_args *ignore) 1620 { 1621 1622 return (ENOSYS); 1623 } 1624 1625 int 1626 linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1627 { 1628 struct getpriority_args bsd_args; 1629 int error; 1630 1631 bsd_args.which = args->which; 1632 bsd_args.who = args->who; 1633 error = sys_getpriority(td, &bsd_args); 1634 td->td_retval[0] = 20 - td->td_retval[0]; 1635 return (error); 1636 } 1637 1638 int 1639 linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1640 { 1641 int name[2]; 1642 1643 name[0] = CTL_KERN; 1644 name[1] = KERN_HOSTNAME; 1645 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1646 args->len, 0, 0)); 1647 } 1648 1649 int 1650 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1651 { 1652 int name[2]; 1653 1654 name[0] = CTL_KERN; 1655 name[1] = KERN_NISDOMAINNAME; 1656 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1657 args->len, 0, 0)); 1658 } 1659 1660 int 1661 linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1662 { 1663 1664 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1665 args->error_code); 1666 1667 /* 1668 * XXX: we should send a signal to the parent if 1669 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1670 * as it doesnt occur often. 1671 */ 1672 exit1(td, args->error_code, 0); 1673 /* NOTREACHED */ 1674 } 1675 1676 #define _LINUX_CAPABILITY_VERSION_1 0x19980330 1677 #define _LINUX_CAPABILITY_VERSION_2 0x20071026 1678 #define _LINUX_CAPABILITY_VERSION_3 0x20080522 1679 1680 struct l_user_cap_header { 1681 l_int version; 1682 l_int pid; 1683 }; 1684 1685 struct l_user_cap_data { 1686 l_int effective; 1687 l_int permitted; 1688 l_int inheritable; 1689 }; 1690 1691 int 1692 linux_capget(struct thread *td, struct linux_capget_args *uap) 1693 { 1694 struct l_user_cap_header luch; 1695 struct l_user_cap_data lucd[2]; 1696 int error, u32s; 1697 1698 if (uap->hdrp == NULL) 1699 return (EFAULT); 1700 1701 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1702 if (error != 0) 1703 return (error); 1704 1705 switch (luch.version) { 1706 case _LINUX_CAPABILITY_VERSION_1: 1707 u32s = 1; 1708 break; 1709 case _LINUX_CAPABILITY_VERSION_2: 1710 case _LINUX_CAPABILITY_VERSION_3: 1711 u32s = 2; 1712 break; 1713 default: 1714 luch.version = _LINUX_CAPABILITY_VERSION_1; 1715 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1716 if (error) 1717 return (error); 1718 return (EINVAL); 1719 } 1720 1721 if (luch.pid) 1722 return (EPERM); 1723 1724 if (uap->datap) { 1725 /* 1726 * The current implementation doesn't support setting 1727 * a capability (it's essentially a stub) so indicate 1728 * that no capabilities are currently set or available 1729 * to request. 1730 */ 1731 memset(&lucd, 0, u32s * sizeof(lucd[0])); 1732 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0])); 1733 } 1734 1735 return (error); 1736 } 1737 1738 int 1739 linux_capset(struct thread *td, struct linux_capset_args *uap) 1740 { 1741 struct l_user_cap_header luch; 1742 struct l_user_cap_data lucd[2]; 1743 int error, i, u32s; 1744 1745 if (uap->hdrp == NULL || uap->datap == NULL) 1746 return (EFAULT); 1747 1748 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1749 if (error != 0) 1750 return (error); 1751 1752 switch (luch.version) { 1753 case _LINUX_CAPABILITY_VERSION_1: 1754 u32s = 1; 1755 break; 1756 case _LINUX_CAPABILITY_VERSION_2: 1757 case _LINUX_CAPABILITY_VERSION_3: 1758 u32s = 2; 1759 break; 1760 default: 1761 luch.version = _LINUX_CAPABILITY_VERSION_1; 1762 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1763 if (error) 1764 return (error); 1765 return (EINVAL); 1766 } 1767 1768 if (luch.pid) 1769 return (EPERM); 1770 1771 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0])); 1772 if (error != 0) 1773 return (error); 1774 1775 /* We currently don't support setting any capabilities. */ 1776 for (i = 0; i < u32s; i++) { 1777 if (lucd[i].effective || lucd[i].permitted || 1778 lucd[i].inheritable) { 1779 linux_msg(td, 1780 "capset[%d] effective=0x%x, permitted=0x%x, " 1781 "inheritable=0x%x is not implemented", i, 1782 (int)lucd[i].effective, (int)lucd[i].permitted, 1783 (int)lucd[i].inheritable); 1784 return (EPERM); 1785 } 1786 } 1787 1788 return (0); 1789 } 1790 1791 int 1792 linux_prctl(struct thread *td, struct linux_prctl_args *args) 1793 { 1794 int error = 0, max_size; 1795 struct proc *p = td->td_proc; 1796 char comm[LINUX_MAX_COMM_LEN]; 1797 int pdeath_signal; 1798 1799 switch (args->option) { 1800 case LINUX_PR_SET_PDEATHSIG: 1801 if (!LINUX_SIG_VALID(args->arg2)) 1802 return (EINVAL); 1803 pdeath_signal = linux_to_bsd_signal(args->arg2); 1804 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL, 1805 &pdeath_signal)); 1806 case LINUX_PR_GET_PDEATHSIG: 1807 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS, 1808 &pdeath_signal); 1809 if (error != 0) 1810 return (error); 1811 pdeath_signal = bsd_to_linux_signal(pdeath_signal); 1812 return (copyout(&pdeath_signal, 1813 (void *)(register_t)args->arg2, 1814 sizeof(pdeath_signal))); 1815 break; 1816 case LINUX_PR_GET_KEEPCAPS: 1817 /* 1818 * Indicate that we always clear the effective and 1819 * permitted capability sets when the user id becomes 1820 * non-zero (actually the capability sets are simply 1821 * always zero in the current implementation). 1822 */ 1823 td->td_retval[0] = 0; 1824 break; 1825 case LINUX_PR_SET_KEEPCAPS: 1826 /* 1827 * Ignore requests to keep the effective and permitted 1828 * capability sets when the user id becomes non-zero. 1829 */ 1830 break; 1831 case LINUX_PR_SET_NAME: 1832 /* 1833 * To be on the safe side we need to make sure to not 1834 * overflow the size a Linux program expects. We already 1835 * do this here in the copyin, so that we don't need to 1836 * check on copyout. 1837 */ 1838 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1839 error = copyinstr((void *)(register_t)args->arg2, comm, 1840 max_size, NULL); 1841 1842 /* Linux silently truncates the name if it is too long. */ 1843 if (error == ENAMETOOLONG) { 1844 /* 1845 * XXX: copyinstr() isn't documented to populate the 1846 * array completely, so do a copyin() to be on the 1847 * safe side. This should be changed in case 1848 * copyinstr() is changed to guarantee this. 1849 */ 1850 error = copyin((void *)(register_t)args->arg2, comm, 1851 max_size - 1); 1852 comm[max_size - 1] = '\0'; 1853 } 1854 if (error) 1855 return (error); 1856 1857 PROC_LOCK(p); 1858 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1859 PROC_UNLOCK(p); 1860 break; 1861 case LINUX_PR_GET_NAME: 1862 PROC_LOCK(p); 1863 strlcpy(comm, p->p_comm, sizeof(comm)); 1864 PROC_UNLOCK(p); 1865 error = copyout(comm, (void *)(register_t)args->arg2, 1866 strlen(comm) + 1); 1867 break; 1868 default: 1869 error = EINVAL; 1870 break; 1871 } 1872 1873 return (error); 1874 } 1875 1876 int 1877 linux_sched_setparam(struct thread *td, 1878 struct linux_sched_setparam_args *uap) 1879 { 1880 struct sched_param sched_param; 1881 struct thread *tdt; 1882 int error; 1883 1884 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1885 if (error) 1886 return (error); 1887 1888 tdt = linux_tdfind(td, uap->pid, -1); 1889 if (tdt == NULL) 1890 return (ESRCH); 1891 1892 error = kern_sched_setparam(td, tdt, &sched_param); 1893 PROC_UNLOCK(tdt->td_proc); 1894 return (error); 1895 } 1896 1897 int 1898 linux_sched_getparam(struct thread *td, 1899 struct linux_sched_getparam_args *uap) 1900 { 1901 struct sched_param sched_param; 1902 struct thread *tdt; 1903 int error; 1904 1905 tdt = linux_tdfind(td, uap->pid, -1); 1906 if (tdt == NULL) 1907 return (ESRCH); 1908 1909 error = kern_sched_getparam(td, tdt, &sched_param); 1910 PROC_UNLOCK(tdt->td_proc); 1911 if (error == 0) 1912 error = copyout(&sched_param, uap->param, 1913 sizeof(sched_param)); 1914 return (error); 1915 } 1916 1917 /* 1918 * Get affinity of a process. 1919 */ 1920 int 1921 linux_sched_getaffinity(struct thread *td, 1922 struct linux_sched_getaffinity_args *args) 1923 { 1924 int error; 1925 struct thread *tdt; 1926 1927 if (args->len < sizeof(cpuset_t)) 1928 return (EINVAL); 1929 1930 tdt = linux_tdfind(td, args->pid, -1); 1931 if (tdt == NULL) 1932 return (ESRCH); 1933 1934 PROC_UNLOCK(tdt->td_proc); 1935 1936 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 1937 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr); 1938 if (error == 0) 1939 td->td_retval[0] = sizeof(cpuset_t); 1940 1941 return (error); 1942 } 1943 1944 /* 1945 * Set affinity of a process. 1946 */ 1947 int 1948 linux_sched_setaffinity(struct thread *td, 1949 struct linux_sched_setaffinity_args *args) 1950 { 1951 struct thread *tdt; 1952 1953 if (args->len < sizeof(cpuset_t)) 1954 return (EINVAL); 1955 1956 tdt = linux_tdfind(td, args->pid, -1); 1957 if (tdt == NULL) 1958 return (ESRCH); 1959 1960 PROC_UNLOCK(tdt->td_proc); 1961 1962 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 1963 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr)); 1964 } 1965 1966 struct linux_rlimit64 { 1967 uint64_t rlim_cur; 1968 uint64_t rlim_max; 1969 }; 1970 1971 int 1972 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 1973 { 1974 struct rlimit rlim, nrlim; 1975 struct linux_rlimit64 lrlim; 1976 struct proc *p; 1977 u_int which; 1978 int flags; 1979 int error; 1980 1981 if (args->resource >= LINUX_RLIM_NLIMITS) 1982 return (EINVAL); 1983 1984 which = linux_to_bsd_resource[args->resource]; 1985 if (which == -1) 1986 return (EINVAL); 1987 1988 if (args->new != NULL) { 1989 /* 1990 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 1991 * rlim is unsigned 64-bit. FreeBSD treats negative limits 1992 * as INFINITY so we do not need a conversion even. 1993 */ 1994 error = copyin(args->new, &nrlim, sizeof(nrlim)); 1995 if (error != 0) 1996 return (error); 1997 } 1998 1999 flags = PGET_HOLD | PGET_NOTWEXIT; 2000 if (args->new != NULL) 2001 flags |= PGET_CANDEBUG; 2002 else 2003 flags |= PGET_CANSEE; 2004 error = pget(args->pid, flags, &p); 2005 if (error != 0) 2006 return (error); 2007 2008 if (args->old != NULL) { 2009 PROC_LOCK(p); 2010 lim_rlimit_proc(p, which, &rlim); 2011 PROC_UNLOCK(p); 2012 if (rlim.rlim_cur == RLIM_INFINITY) 2013 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2014 else 2015 lrlim.rlim_cur = rlim.rlim_cur; 2016 if (rlim.rlim_max == RLIM_INFINITY) 2017 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2018 else 2019 lrlim.rlim_max = rlim.rlim_max; 2020 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2021 if (error != 0) 2022 goto out; 2023 } 2024 2025 if (args->new != NULL) 2026 error = kern_proc_setrlimit(td, p, which, &nrlim); 2027 2028 out: 2029 PRELE(p); 2030 return (error); 2031 } 2032 2033 int 2034 linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2035 { 2036 struct timeval utv, tv0, tv1, *tvp; 2037 struct l_pselect6arg lpse6; 2038 struct l_timespec lts; 2039 struct timespec uts; 2040 l_sigset_t l_ss; 2041 sigset_t *ssp; 2042 sigset_t ss; 2043 int error; 2044 2045 ssp = NULL; 2046 if (args->sig != NULL) { 2047 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2048 if (error != 0) 2049 return (error); 2050 if (lpse6.ss_len != sizeof(l_ss)) 2051 return (EINVAL); 2052 if (lpse6.ss != 0) { 2053 error = copyin(PTRIN(lpse6.ss), &l_ss, 2054 sizeof(l_ss)); 2055 if (error != 0) 2056 return (error); 2057 linux_to_bsd_sigset(&l_ss, &ss); 2058 ssp = &ss; 2059 } 2060 } 2061 2062 /* 2063 * Currently glibc changes nanosecond number to microsecond. 2064 * This mean losing precision but for now it is hardly seen. 2065 */ 2066 if (args->tsp != NULL) { 2067 error = copyin(args->tsp, <s, sizeof(lts)); 2068 if (error != 0) 2069 return (error); 2070 error = linux_to_native_timespec(&uts, <s); 2071 if (error != 0) 2072 return (error); 2073 2074 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2075 if (itimerfix(&utv)) 2076 return (EINVAL); 2077 2078 microtime(&tv0); 2079 tvp = &utv; 2080 } else 2081 tvp = NULL; 2082 2083 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2084 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2085 2086 if (error == 0 && args->tsp != NULL) { 2087 if (td->td_retval[0] != 0) { 2088 /* 2089 * Compute how much time was left of the timeout, 2090 * by subtracting the current time and the time 2091 * before we started the call, and subtracting 2092 * that result from the user-supplied value. 2093 */ 2094 2095 microtime(&tv1); 2096 timevalsub(&tv1, &tv0); 2097 timevalsub(&utv, &tv1); 2098 if (utv.tv_sec < 0) 2099 timevalclear(&utv); 2100 } else 2101 timevalclear(&utv); 2102 2103 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2104 2105 error = native_to_linux_timespec(<s, &uts); 2106 if (error == 0) 2107 error = copyout(<s, args->tsp, sizeof(lts)); 2108 } 2109 2110 return (error); 2111 } 2112 2113 int 2114 linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2115 { 2116 struct timespec ts0, ts1; 2117 struct l_timespec lts; 2118 struct timespec uts, *tsp; 2119 l_sigset_t l_ss; 2120 sigset_t *ssp; 2121 sigset_t ss; 2122 int error; 2123 2124 if (args->sset != NULL) { 2125 if (args->ssize != sizeof(l_ss)) 2126 return (EINVAL); 2127 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2128 if (error) 2129 return (error); 2130 linux_to_bsd_sigset(&l_ss, &ss); 2131 ssp = &ss; 2132 } else 2133 ssp = NULL; 2134 if (args->tsp != NULL) { 2135 error = copyin(args->tsp, <s, sizeof(lts)); 2136 if (error) 2137 return (error); 2138 error = linux_to_native_timespec(&uts, <s); 2139 if (error != 0) 2140 return (error); 2141 2142 nanotime(&ts0); 2143 tsp = &uts; 2144 } else 2145 tsp = NULL; 2146 2147 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2148 2149 if (error == 0 && args->tsp != NULL) { 2150 if (td->td_retval[0]) { 2151 nanotime(&ts1); 2152 timespecsub(&ts1, &ts0, &ts1); 2153 timespecsub(&uts, &ts1, &uts); 2154 if (uts.tv_sec < 0) 2155 timespecclear(&uts); 2156 } else 2157 timespecclear(&uts); 2158 2159 error = native_to_linux_timespec(<s, &uts); 2160 if (error == 0) 2161 error = copyout(<s, args->tsp, sizeof(lts)); 2162 } 2163 2164 return (error); 2165 } 2166 2167 int 2168 linux_sched_rr_get_interval(struct thread *td, 2169 struct linux_sched_rr_get_interval_args *uap) 2170 { 2171 struct timespec ts; 2172 struct l_timespec lts; 2173 struct thread *tdt; 2174 int error; 2175 2176 /* 2177 * According to man in case the invalid pid specified 2178 * EINVAL should be returned. 2179 */ 2180 if (uap->pid < 0) 2181 return (EINVAL); 2182 2183 tdt = linux_tdfind(td, uap->pid, -1); 2184 if (tdt == NULL) 2185 return (ESRCH); 2186 2187 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2188 PROC_UNLOCK(tdt->td_proc); 2189 if (error != 0) 2190 return (error); 2191 error = native_to_linux_timespec(<s, &ts); 2192 if (error != 0) 2193 return (error); 2194 return (copyout(<s, uap->interval, sizeof(lts))); 2195 } 2196 2197 /* 2198 * In case when the Linux thread is the initial thread in 2199 * the thread group thread id is equal to the process id. 2200 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2201 */ 2202 struct thread * 2203 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2204 { 2205 struct linux_emuldata *em; 2206 struct thread *tdt; 2207 struct proc *p; 2208 2209 tdt = NULL; 2210 if (tid == 0 || tid == td->td_tid) { 2211 tdt = td; 2212 PROC_LOCK(tdt->td_proc); 2213 } else if (tid > PID_MAX) 2214 tdt = tdfind(tid, pid); 2215 else { 2216 /* 2217 * Initial thread where the tid equal to the pid. 2218 */ 2219 p = pfind(tid); 2220 if (p != NULL) { 2221 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2222 /* 2223 * p is not a Linuxulator process. 2224 */ 2225 PROC_UNLOCK(p); 2226 return (NULL); 2227 } 2228 FOREACH_THREAD_IN_PROC(p, tdt) { 2229 em = em_find(tdt); 2230 if (tid == em->em_tid) 2231 return (tdt); 2232 } 2233 PROC_UNLOCK(p); 2234 } 2235 return (NULL); 2236 } 2237 2238 return (tdt); 2239 } 2240 2241 void 2242 linux_to_bsd_waitopts(int options, int *bsdopts) 2243 { 2244 2245 if (options & LINUX_WNOHANG) 2246 *bsdopts |= WNOHANG; 2247 if (options & LINUX_WUNTRACED) 2248 *bsdopts |= WUNTRACED; 2249 if (options & LINUX_WEXITED) 2250 *bsdopts |= WEXITED; 2251 if (options & LINUX_WCONTINUED) 2252 *bsdopts |= WCONTINUED; 2253 if (options & LINUX_WNOWAIT) 2254 *bsdopts |= WNOWAIT; 2255 2256 if (options & __WCLONE) 2257 *bsdopts |= WLINUXCLONE; 2258 } 2259 2260 int 2261 linux_getrandom(struct thread *td, struct linux_getrandom_args *args) 2262 { 2263 struct uio uio; 2264 struct iovec iov; 2265 int error; 2266 2267 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM)) 2268 return (EINVAL); 2269 if (args->count > INT_MAX) 2270 args->count = INT_MAX; 2271 2272 iov.iov_base = args->buf; 2273 iov.iov_len = args->count; 2274 2275 uio.uio_iov = &iov; 2276 uio.uio_iovcnt = 1; 2277 uio.uio_resid = iov.iov_len; 2278 uio.uio_segflg = UIO_USERSPACE; 2279 uio.uio_rw = UIO_READ; 2280 uio.uio_td = td; 2281 2282 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK); 2283 if (error == 0) 2284 td->td_retval[0] = args->count - uio.uio_resid; 2285 return (error); 2286 } 2287 2288 int 2289 linux_mincore(struct thread *td, struct linux_mincore_args *args) 2290 { 2291 2292 /* Needs to be page-aligned */ 2293 if (args->start & PAGE_MASK) 2294 return (EINVAL); 2295 return (kern_mincore(td, args->start, args->len, args->vec)); 2296 } 2297