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/reboot.h> 55 #include <sys/racct.h> 56 #include <sys/random.h> 57 #include <sys/resourcevar.h> 58 #include <sys/sched.h> 59 #include <sys/sdt.h> 60 #include <sys/signalvar.h> 61 #include <sys/stat.h> 62 #include <sys/syscallsubr.h> 63 #include <sys/sysctl.h> 64 #include <sys/sysproto.h> 65 #include <sys/systm.h> 66 #include <sys/time.h> 67 #include <sys/vmmeter.h> 68 #include <sys/vnode.h> 69 #include <sys/wait.h> 70 #include <sys/cpuset.h> 71 #include <sys/uio.h> 72 73 #include <security/mac/mac_framework.h> 74 75 #include <vm/vm.h> 76 #include <vm/pmap.h> 77 #include <vm/vm_kern.h> 78 #include <vm/vm_map.h> 79 #include <vm/vm_extern.h> 80 #include <vm/vm_object.h> 81 #include <vm/swap_pager.h> 82 83 #ifdef COMPAT_LINUX32 84 #include <machine/../linux32/linux.h> 85 #include <machine/../linux32/linux32_proto.h> 86 #else 87 #include <machine/../linux/linux.h> 88 #include <machine/../linux/linux_proto.h> 89 #endif 90 91 #include <compat/linux/linux_dtrace.h> 92 #include <compat/linux/linux_file.h> 93 #include <compat/linux/linux_mib.h> 94 #include <compat/linux/linux_signal.h> 95 #include <compat/linux/linux_timer.h> 96 #include <compat/linux/linux_util.h> 97 #include <compat/linux/linux_sysproto.h> 98 #include <compat/linux/linux_emul.h> 99 #include <compat/linux/linux_misc.h> 100 101 /** 102 * Special DTrace provider for the linuxulator. 103 * 104 * In this file we define the provider for the entire linuxulator. All 105 * modules (= files of the linuxulator) use it. 106 * 107 * We define a different name depending on the emulated bitsize, see 108 * ../../<ARCH>/linux{,32}/linux.h, e.g.: 109 * native bitsize = linuxulator 110 * amd64, 32bit emulation = linuxulator32 111 */ 112 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE); 113 114 int stclohz; /* Statistics clock frequency */ 115 116 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 117 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 118 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 119 RLIMIT_MEMLOCK, RLIMIT_AS 120 }; 121 122 struct l_sysinfo { 123 l_long uptime; /* Seconds since boot */ 124 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 125 #define LINUX_SYSINFO_LOADS_SCALE 65536 126 l_ulong totalram; /* Total usable main memory size */ 127 l_ulong freeram; /* Available memory size */ 128 l_ulong sharedram; /* Amount of shared memory */ 129 l_ulong bufferram; /* Memory used by buffers */ 130 l_ulong totalswap; /* Total swap space size */ 131 l_ulong freeswap; /* swap space still available */ 132 l_ushort procs; /* Number of current processes */ 133 l_ushort pads; 134 l_ulong totalbig; 135 l_ulong freebig; 136 l_uint mem_unit; 137 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 138 }; 139 140 struct l_pselect6arg { 141 l_uintptr_t ss; 142 l_size_t ss_len; 143 }; 144 145 static int linux_utimensat_nsec_valid(l_long); 146 147 148 int 149 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 150 { 151 struct l_sysinfo sysinfo; 152 vm_object_t object; 153 int i, j; 154 struct timespec ts; 155 156 bzero(&sysinfo, sizeof(sysinfo)); 157 getnanouptime(&ts); 158 if (ts.tv_nsec != 0) 159 ts.tv_sec++; 160 sysinfo.uptime = ts.tv_sec; 161 162 /* Use the information from the mib to get our load averages */ 163 for (i = 0; i < 3; i++) 164 sysinfo.loads[i] = averunnable.ldavg[i] * 165 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 166 167 sysinfo.totalram = physmem * PAGE_SIZE; 168 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE; 169 170 sysinfo.sharedram = 0; 171 mtx_lock(&vm_object_list_mtx); 172 TAILQ_FOREACH(object, &vm_object_list, object_list) 173 if (object->shadow_count > 1) 174 sysinfo.sharedram += object->resident_page_count; 175 mtx_unlock(&vm_object_list_mtx); 176 177 sysinfo.sharedram *= PAGE_SIZE; 178 sysinfo.bufferram = 0; 179 180 swap_pager_status(&i, &j); 181 sysinfo.totalswap = i * PAGE_SIZE; 182 sysinfo.freeswap = (i - j) * PAGE_SIZE; 183 184 sysinfo.procs = nprocs; 185 186 /* The following are only present in newer Linux kernels. */ 187 sysinfo.totalbig = 0; 188 sysinfo.freebig = 0; 189 sysinfo.mem_unit = 1; 190 191 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 192 } 193 194 #ifdef LINUX_LEGACY_SYSCALLS 195 int 196 linux_alarm(struct thread *td, struct linux_alarm_args *args) 197 { 198 struct itimerval it, old_it; 199 u_int secs; 200 int error; 201 202 #ifdef DEBUG 203 if (ldebug(alarm)) 204 printf(ARGS(alarm, "%u"), args->secs); 205 #endif 206 secs = args->secs; 207 /* 208 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2 209 * to match kern_setitimer()'s limit to avoid error from it. 210 * 211 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit 212 * platforms. 213 */ 214 if (secs > INT32_MAX / 2) 215 secs = INT32_MAX / 2; 216 217 it.it_value.tv_sec = secs; 218 it.it_value.tv_usec = 0; 219 timevalclear(&it.it_interval); 220 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 221 KASSERT(error == 0, ("kern_setitimer returns %d", error)); 222 223 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) || 224 old_it.it_value.tv_usec >= 500000) 225 old_it.it_value.tv_sec++; 226 td->td_retval[0] = old_it.it_value.tv_sec; 227 return (0); 228 } 229 #endif 230 231 int 232 linux_brk(struct thread *td, struct linux_brk_args *args) 233 { 234 struct vmspace *vm = td->td_proc->p_vmspace; 235 uintptr_t new, old; 236 237 #ifdef DEBUG 238 if (ldebug(brk)) 239 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 240 #endif 241 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize); 242 new = (uintptr_t)args->dsend; 243 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new)) 244 td->td_retval[0] = (register_t)new; 245 else 246 td->td_retval[0] = (register_t)old; 247 248 return (0); 249 } 250 251 #if defined(__i386__) 252 /* XXX: what about amd64/linux32? */ 253 254 int 255 linux_uselib(struct thread *td, struct linux_uselib_args *args) 256 { 257 struct nameidata ni; 258 struct vnode *vp; 259 struct exec *a_out; 260 struct vattr attr; 261 vm_offset_t vmaddr; 262 unsigned long file_offset; 263 unsigned long bss_size; 264 char *library; 265 ssize_t aresid; 266 int error, locked, writecount; 267 268 LCONVPATHEXIST(td, args->library, &library); 269 270 #ifdef DEBUG 271 if (ldebug(uselib)) 272 printf(ARGS(uselib, "%s"), library); 273 #endif 274 275 a_out = NULL; 276 locked = 0; 277 vp = NULL; 278 279 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 280 UIO_SYSSPACE, library, td); 281 error = namei(&ni); 282 LFREEPATH(library); 283 if (error) 284 goto cleanup; 285 286 vp = ni.ni_vp; 287 NDFREE(&ni, NDF_ONLY_PNBUF); 288 289 /* 290 * From here on down, we have a locked vnode that must be unlocked. 291 * XXX: The code below largely duplicates exec_check_permissions(). 292 */ 293 locked = 1; 294 295 /* Writable? */ 296 error = VOP_GET_WRITECOUNT(vp, &writecount); 297 if (error != 0) 298 goto cleanup; 299 if (writecount != 0) { 300 error = ETXTBSY; 301 goto cleanup; 302 } 303 304 /* Executable? */ 305 error = VOP_GETATTR(vp, &attr, td->td_ucred); 306 if (error) 307 goto cleanup; 308 309 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 310 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 311 /* EACCESS is what exec(2) returns. */ 312 error = ENOEXEC; 313 goto cleanup; 314 } 315 316 /* Sensible size? */ 317 if (attr.va_size == 0) { 318 error = ENOEXEC; 319 goto cleanup; 320 } 321 322 /* Can we access it? */ 323 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 324 if (error) 325 goto cleanup; 326 327 /* 328 * XXX: This should use vn_open() so that it is properly authorized, 329 * and to reduce code redundancy all over the place here. 330 * XXX: Not really, it duplicates far more of exec_check_permissions() 331 * than vn_open(). 332 */ 333 #ifdef MAC 334 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 335 if (error) 336 goto cleanup; 337 #endif 338 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 339 if (error) 340 goto cleanup; 341 342 /* Pull in executable header into exec_map */ 343 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 344 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 345 if (error) 346 goto cleanup; 347 348 /* Is it a Linux binary ? */ 349 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 350 error = ENOEXEC; 351 goto cleanup; 352 } 353 354 /* 355 * While we are here, we should REALLY do some more checks 356 */ 357 358 /* Set file/virtual offset based on a.out variant. */ 359 switch ((int)(a_out->a_magic & 0xffff)) { 360 case 0413: /* ZMAGIC */ 361 file_offset = 1024; 362 break; 363 case 0314: /* QMAGIC */ 364 file_offset = 0; 365 break; 366 default: 367 error = ENOEXEC; 368 goto cleanup; 369 } 370 371 bss_size = round_page(a_out->a_bss); 372 373 /* Check various fields in header for validity/bounds. */ 374 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 375 error = ENOEXEC; 376 goto cleanup; 377 } 378 379 /* text + data can't exceed file size */ 380 if (a_out->a_data + a_out->a_text > attr.va_size) { 381 error = EFAULT; 382 goto cleanup; 383 } 384 385 /* 386 * text/data/bss must not exceed limits 387 * XXX - this is not complete. it should check current usage PLUS 388 * the resources needed by this library. 389 */ 390 PROC_LOCK(td->td_proc); 391 if (a_out->a_text > maxtsiz || 392 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) || 393 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 394 bss_size) != 0) { 395 PROC_UNLOCK(td->td_proc); 396 error = ENOMEM; 397 goto cleanup; 398 } 399 PROC_UNLOCK(td->td_proc); 400 401 /* 402 * Prevent more writers. 403 * XXX: Note that if any of the VM operations fail below we don't 404 * clear this flag. 405 */ 406 VOP_SET_TEXT(vp); 407 408 /* 409 * Lock no longer needed 410 */ 411 locked = 0; 412 VOP_UNLOCK(vp, 0); 413 414 /* 415 * Check if file_offset page aligned. Currently we cannot handle 416 * misalinged file offsets, and so we read in the entire image 417 * (what a waste). 418 */ 419 if (file_offset & PAGE_MASK) { 420 #ifdef DEBUG 421 printf("uselib: Non page aligned binary %lu\n", file_offset); 422 #endif 423 /* Map text+data read/write/execute */ 424 425 /* a_entry is the load address and is page aligned */ 426 vmaddr = trunc_page(a_out->a_entry); 427 428 /* get anon user mapping, read+write+execute */ 429 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 430 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 431 VM_PROT_ALL, VM_PROT_ALL, 0); 432 if (error) 433 goto cleanup; 434 435 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 436 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 437 td->td_ucred, NOCRED, &aresid, td); 438 if (error != 0) 439 goto cleanup; 440 if (aresid != 0) { 441 error = ENOEXEC; 442 goto cleanup; 443 } 444 } else { 445 #ifdef DEBUG 446 printf("uselib: Page aligned binary %lu\n", file_offset); 447 #endif 448 /* 449 * for QMAGIC, a_entry is 20 bytes beyond the load address 450 * to skip the executable header 451 */ 452 vmaddr = trunc_page(a_out->a_entry); 453 454 /* 455 * Map it all into the process's space as a single 456 * copy-on-write "data" segment. 457 */ 458 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 459 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 460 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 461 if (error) 462 goto cleanup; 463 } 464 #ifdef DEBUG 465 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 466 ((long *)vmaddr)[1]); 467 #endif 468 if (bss_size != 0) { 469 /* Calculate BSS start address */ 470 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 471 a_out->a_data; 472 473 /* allocate some 'anon' space */ 474 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 475 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 476 VM_PROT_ALL, 0); 477 if (error) 478 goto cleanup; 479 } 480 481 cleanup: 482 /* Unlock vnode if needed */ 483 if (locked) 484 VOP_UNLOCK(vp, 0); 485 486 /* Release the temporary mapping. */ 487 if (a_out) 488 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 489 490 return (error); 491 } 492 493 #endif /* __i386__ */ 494 495 #ifdef LINUX_LEGACY_SYSCALLS 496 int 497 linux_select(struct thread *td, struct linux_select_args *args) 498 { 499 l_timeval ltv; 500 struct timeval tv0, tv1, utv, *tvp; 501 int error; 502 503 #ifdef DEBUG 504 if (ldebug(select)) 505 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 506 (void *)args->readfds, (void *)args->writefds, 507 (void *)args->exceptfds, (void *)args->timeout); 508 #endif 509 510 /* 511 * Store current time for computation of the amount of 512 * time left. 513 */ 514 if (args->timeout) { 515 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 516 goto select_out; 517 utv.tv_sec = ltv.tv_sec; 518 utv.tv_usec = ltv.tv_usec; 519 #ifdef DEBUG 520 if (ldebug(select)) 521 printf(LMSG("incoming timeout (%jd/%ld)"), 522 (intmax_t)utv.tv_sec, utv.tv_usec); 523 #endif 524 525 if (itimerfix(&utv)) { 526 /* 527 * The timeval was invalid. Convert it to something 528 * valid that will act as it does under Linux. 529 */ 530 utv.tv_sec += utv.tv_usec / 1000000; 531 utv.tv_usec %= 1000000; 532 if (utv.tv_usec < 0) { 533 utv.tv_sec -= 1; 534 utv.tv_usec += 1000000; 535 } 536 if (utv.tv_sec < 0) 537 timevalclear(&utv); 538 } 539 microtime(&tv0); 540 tvp = &utv; 541 } else 542 tvp = NULL; 543 544 error = kern_select(td, args->nfds, args->readfds, args->writefds, 545 args->exceptfds, tvp, LINUX_NFDBITS); 546 547 #ifdef DEBUG 548 if (ldebug(select)) 549 printf(LMSG("real select returns %d"), error); 550 #endif 551 if (error) 552 goto select_out; 553 554 if (args->timeout) { 555 if (td->td_retval[0]) { 556 /* 557 * Compute how much time was left of the timeout, 558 * by subtracting the current time and the time 559 * before we started the call, and subtracting 560 * that result from the user-supplied value. 561 */ 562 microtime(&tv1); 563 timevalsub(&tv1, &tv0); 564 timevalsub(&utv, &tv1); 565 if (utv.tv_sec < 0) 566 timevalclear(&utv); 567 } else 568 timevalclear(&utv); 569 #ifdef DEBUG 570 if (ldebug(select)) 571 printf(LMSG("outgoing timeout (%jd/%ld)"), 572 (intmax_t)utv.tv_sec, utv.tv_usec); 573 #endif 574 ltv.tv_sec = utv.tv_sec; 575 ltv.tv_usec = utv.tv_usec; 576 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 577 goto select_out; 578 } 579 580 select_out: 581 #ifdef DEBUG 582 if (ldebug(select)) 583 printf(LMSG("select_out -> %d"), error); 584 #endif 585 return (error); 586 } 587 #endif 588 589 int 590 linux_mremap(struct thread *td, struct linux_mremap_args *args) 591 { 592 uintptr_t addr; 593 size_t len; 594 int error = 0; 595 596 #ifdef DEBUG 597 if (ldebug(mremap)) 598 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 599 (void *)(uintptr_t)args->addr, 600 (unsigned long)args->old_len, 601 (unsigned long)args->new_len, 602 (unsigned long)args->flags); 603 #endif 604 605 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 606 td->td_retval[0] = 0; 607 return (EINVAL); 608 } 609 610 /* 611 * Check for the page alignment. 612 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 613 */ 614 if (args->addr & PAGE_MASK) { 615 td->td_retval[0] = 0; 616 return (EINVAL); 617 } 618 619 args->new_len = round_page(args->new_len); 620 args->old_len = round_page(args->old_len); 621 622 if (args->new_len > args->old_len) { 623 td->td_retval[0] = 0; 624 return (ENOMEM); 625 } 626 627 if (args->new_len < args->old_len) { 628 addr = args->addr + args->new_len; 629 len = args->old_len - args->new_len; 630 error = kern_munmap(td, addr, len); 631 } 632 633 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 634 return (error); 635 } 636 637 #define LINUX_MS_ASYNC 0x0001 638 #define LINUX_MS_INVALIDATE 0x0002 639 #define LINUX_MS_SYNC 0x0004 640 641 int 642 linux_msync(struct thread *td, struct linux_msync_args *args) 643 { 644 645 return (kern_msync(td, args->addr, args->len, 646 args->fl & ~LINUX_MS_SYNC)); 647 } 648 649 #ifdef LINUX_LEGACY_SYSCALLS 650 int 651 linux_time(struct thread *td, struct linux_time_args *args) 652 { 653 struct timeval tv; 654 l_time_t tm; 655 int error; 656 657 #ifdef DEBUG 658 if (ldebug(time)) 659 printf(ARGS(time, "*")); 660 #endif 661 662 microtime(&tv); 663 tm = tv.tv_sec; 664 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 665 return (error); 666 td->td_retval[0] = tm; 667 return (0); 668 } 669 #endif 670 671 struct l_times_argv { 672 l_clock_t tms_utime; 673 l_clock_t tms_stime; 674 l_clock_t tms_cutime; 675 l_clock_t tms_cstime; 676 }; 677 678 679 /* 680 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 681 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 682 * auxiliary vector entry. 683 */ 684 #define CLK_TCK 100 685 686 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 687 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 688 689 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 690 CONVNTCK(r) : CONVOTCK(r)) 691 692 int 693 linux_times(struct thread *td, struct linux_times_args *args) 694 { 695 struct timeval tv, utime, stime, cutime, cstime; 696 struct l_times_argv tms; 697 struct proc *p; 698 int error; 699 700 #ifdef DEBUG 701 if (ldebug(times)) 702 printf(ARGS(times, "*")); 703 #endif 704 705 if (args->buf != NULL) { 706 p = td->td_proc; 707 PROC_LOCK(p); 708 PROC_STATLOCK(p); 709 calcru(p, &utime, &stime); 710 PROC_STATUNLOCK(p); 711 calccru(p, &cutime, &cstime); 712 PROC_UNLOCK(p); 713 714 tms.tms_utime = CONVTCK(utime); 715 tms.tms_stime = CONVTCK(stime); 716 717 tms.tms_cutime = CONVTCK(cutime); 718 tms.tms_cstime = CONVTCK(cstime); 719 720 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 721 return (error); 722 } 723 724 microuptime(&tv); 725 td->td_retval[0] = (int)CONVTCK(tv); 726 return (0); 727 } 728 729 int 730 linux_newuname(struct thread *td, struct linux_newuname_args *args) 731 { 732 struct l_new_utsname utsname; 733 char osname[LINUX_MAX_UTSNAME]; 734 char osrelease[LINUX_MAX_UTSNAME]; 735 char *p; 736 737 #ifdef DEBUG 738 if (ldebug(newuname)) 739 printf(ARGS(newuname, "*")); 740 #endif 741 742 linux_get_osname(td, osname); 743 linux_get_osrelease(td, osrelease); 744 745 bzero(&utsname, sizeof(utsname)); 746 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 747 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 748 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 749 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 750 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 751 for (p = utsname.version; *p != '\0'; ++p) 752 if (*p == '\n') { 753 *p = '\0'; 754 break; 755 } 756 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 757 758 return (copyout(&utsname, args->buf, sizeof(utsname))); 759 } 760 761 struct l_utimbuf { 762 l_time_t l_actime; 763 l_time_t l_modtime; 764 }; 765 766 #ifdef LINUX_LEGACY_SYSCALLS 767 int 768 linux_utime(struct thread *td, struct linux_utime_args *args) 769 { 770 struct timeval tv[2], *tvp; 771 struct l_utimbuf lut; 772 char *fname; 773 int error; 774 775 LCONVPATHEXIST(td, args->fname, &fname); 776 777 #ifdef DEBUG 778 if (ldebug(utime)) 779 printf(ARGS(utime, "%s, *"), fname); 780 #endif 781 782 if (args->times) { 783 if ((error = copyin(args->times, &lut, sizeof lut))) { 784 LFREEPATH(fname); 785 return (error); 786 } 787 tv[0].tv_sec = lut.l_actime; 788 tv[0].tv_usec = 0; 789 tv[1].tv_sec = lut.l_modtime; 790 tv[1].tv_usec = 0; 791 tvp = tv; 792 } else 793 tvp = NULL; 794 795 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp, 796 UIO_SYSSPACE); 797 LFREEPATH(fname); 798 return (error); 799 } 800 #endif 801 802 #ifdef LINUX_LEGACY_SYSCALLS 803 int 804 linux_utimes(struct thread *td, struct linux_utimes_args *args) 805 { 806 l_timeval ltv[2]; 807 struct timeval tv[2], *tvp = NULL; 808 char *fname; 809 int error; 810 811 LCONVPATHEXIST(td, args->fname, &fname); 812 813 #ifdef DEBUG 814 if (ldebug(utimes)) 815 printf(ARGS(utimes, "%s, *"), fname); 816 #endif 817 818 if (args->tptr != NULL) { 819 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 820 LFREEPATH(fname); 821 return (error); 822 } 823 tv[0].tv_sec = ltv[0].tv_sec; 824 tv[0].tv_usec = ltv[0].tv_usec; 825 tv[1].tv_sec = ltv[1].tv_sec; 826 tv[1].tv_usec = ltv[1].tv_usec; 827 tvp = tv; 828 } 829 830 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, 831 tvp, UIO_SYSSPACE); 832 LFREEPATH(fname); 833 return (error); 834 } 835 #endif 836 837 static int 838 linux_utimensat_nsec_valid(l_long nsec) 839 { 840 841 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW) 842 return (0); 843 if (nsec >= 0 && nsec <= 999999999) 844 return (0); 845 return (1); 846 } 847 848 int 849 linux_utimensat(struct thread *td, struct linux_utimensat_args *args) 850 { 851 struct l_timespec l_times[2]; 852 struct timespec times[2], *timesp = NULL; 853 char *path = NULL; 854 int error, dfd, flags = 0; 855 856 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 857 858 #ifdef DEBUG 859 if (ldebug(utimensat)) 860 printf(ARGS(utimensat, "%d, *"), dfd); 861 #endif 862 863 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 864 return (EINVAL); 865 866 if (args->times != NULL) { 867 error = copyin(args->times, l_times, sizeof(l_times)); 868 if (error != 0) 869 return (error); 870 871 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 || 872 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0) 873 return (EINVAL); 874 875 times[0].tv_sec = l_times[0].tv_sec; 876 switch (l_times[0].tv_nsec) 877 { 878 case LINUX_UTIME_OMIT: 879 times[0].tv_nsec = UTIME_OMIT; 880 break; 881 case LINUX_UTIME_NOW: 882 times[0].tv_nsec = UTIME_NOW; 883 break; 884 default: 885 times[0].tv_nsec = l_times[0].tv_nsec; 886 } 887 888 times[1].tv_sec = l_times[1].tv_sec; 889 switch (l_times[1].tv_nsec) 890 { 891 case LINUX_UTIME_OMIT: 892 times[1].tv_nsec = UTIME_OMIT; 893 break; 894 case LINUX_UTIME_NOW: 895 times[1].tv_nsec = UTIME_NOW; 896 break; 897 default: 898 times[1].tv_nsec = l_times[1].tv_nsec; 899 break; 900 } 901 timesp = times; 902 903 /* This breaks POSIX, but is what the Linux kernel does 904 * _on purpose_ (documented in the man page for utimensat(2)), 905 * so we must follow that behaviour. */ 906 if (times[0].tv_nsec == UTIME_OMIT && 907 times[1].tv_nsec == UTIME_OMIT) 908 return (0); 909 } 910 911 if (args->pathname != NULL) 912 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 913 else if (args->flags != 0) 914 return (EINVAL); 915 916 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 917 flags |= AT_SYMLINK_NOFOLLOW; 918 919 if (path == NULL) 920 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 921 else { 922 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 923 UIO_SYSSPACE, flags); 924 LFREEPATH(path); 925 } 926 927 return (error); 928 } 929 930 #ifdef LINUX_LEGACY_SYSCALLS 931 int 932 linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 933 { 934 l_timeval ltv[2]; 935 struct timeval tv[2], *tvp = NULL; 936 char *fname; 937 int error, dfd; 938 939 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 940 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 941 942 #ifdef DEBUG 943 if (ldebug(futimesat)) 944 printf(ARGS(futimesat, "%s, *"), fname); 945 #endif 946 947 if (args->utimes != NULL) { 948 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 949 LFREEPATH(fname); 950 return (error); 951 } 952 tv[0].tv_sec = ltv[0].tv_sec; 953 tv[0].tv_usec = ltv[0].tv_usec; 954 tv[1].tv_sec = ltv[1].tv_sec; 955 tv[1].tv_usec = ltv[1].tv_usec; 956 tvp = tv; 957 } 958 959 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 960 LFREEPATH(fname); 961 return (error); 962 } 963 #endif 964 965 int 966 linux_common_wait(struct thread *td, int pid, int *status, 967 int options, struct rusage *ru) 968 { 969 int error, tmpstat; 970 971 error = kern_wait(td, pid, &tmpstat, options, ru); 972 if (error) 973 return (error); 974 975 if (status) { 976 tmpstat &= 0xffff; 977 if (WIFSIGNALED(tmpstat)) 978 tmpstat = (tmpstat & 0xffffff80) | 979 bsd_to_linux_signal(WTERMSIG(tmpstat)); 980 else if (WIFSTOPPED(tmpstat)) 981 tmpstat = (tmpstat & 0xffff00ff) | 982 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 983 else if (WIFCONTINUED(tmpstat)) 984 tmpstat = 0xffff; 985 error = copyout(&tmpstat, status, sizeof(int)); 986 } 987 988 return (error); 989 } 990 991 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 992 int 993 linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 994 { 995 struct linux_wait4_args wait4_args; 996 997 #ifdef DEBUG 998 if (ldebug(waitpid)) 999 printf(ARGS(waitpid, "%d, %p, %d"), 1000 args->pid, (void *)args->status, args->options); 1001 #endif 1002 1003 wait4_args.pid = args->pid; 1004 wait4_args.status = args->status; 1005 wait4_args.options = args->options; 1006 wait4_args.rusage = NULL; 1007 1008 return (linux_wait4(td, &wait4_args)); 1009 } 1010 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1011 1012 int 1013 linux_wait4(struct thread *td, struct linux_wait4_args *args) 1014 { 1015 int error, options; 1016 struct rusage ru, *rup; 1017 1018 #ifdef DEBUG 1019 if (ldebug(wait4)) 1020 printf(ARGS(wait4, "%d, %p, %d, %p"), 1021 args->pid, (void *)args->status, args->options, 1022 (void *)args->rusage); 1023 #endif 1024 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 1025 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 1026 return (EINVAL); 1027 1028 options = WEXITED; 1029 linux_to_bsd_waitopts(args->options, &options); 1030 1031 if (args->rusage != NULL) 1032 rup = &ru; 1033 else 1034 rup = NULL; 1035 error = linux_common_wait(td, args->pid, args->status, options, rup); 1036 if (error != 0) 1037 return (error); 1038 if (args->rusage != NULL) 1039 error = linux_copyout_rusage(&ru, args->rusage); 1040 return (error); 1041 } 1042 1043 int 1044 linux_waitid(struct thread *td, struct linux_waitid_args *args) 1045 { 1046 int status, options, sig; 1047 struct __wrusage wru; 1048 siginfo_t siginfo; 1049 l_siginfo_t lsi; 1050 idtype_t idtype; 1051 struct proc *p; 1052 int error; 1053 1054 options = 0; 1055 linux_to_bsd_waitopts(args->options, &options); 1056 1057 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1058 return (EINVAL); 1059 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1060 return (EINVAL); 1061 1062 switch (args->idtype) { 1063 case LINUX_P_ALL: 1064 idtype = P_ALL; 1065 break; 1066 case LINUX_P_PID: 1067 if (args->id <= 0) 1068 return (EINVAL); 1069 idtype = P_PID; 1070 break; 1071 case LINUX_P_PGID: 1072 if (args->id <= 0) 1073 return (EINVAL); 1074 idtype = P_PGID; 1075 break; 1076 default: 1077 return (EINVAL); 1078 } 1079 1080 error = kern_wait6(td, idtype, args->id, &status, options, 1081 &wru, &siginfo); 1082 if (error != 0) 1083 return (error); 1084 if (args->rusage != NULL) { 1085 error = linux_copyout_rusage(&wru.wru_children, 1086 args->rusage); 1087 if (error != 0) 1088 return (error); 1089 } 1090 if (args->info != NULL) { 1091 p = td->td_proc; 1092 bzero(&lsi, sizeof(lsi)); 1093 if (td->td_retval[0] != 0) { 1094 sig = bsd_to_linux_signal(siginfo.si_signo); 1095 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1096 } 1097 error = copyout(&lsi, args->info, sizeof(lsi)); 1098 } 1099 td->td_retval[0] = 0; 1100 1101 return (error); 1102 } 1103 1104 #ifdef LINUX_LEGACY_SYSCALLS 1105 int 1106 linux_mknod(struct thread *td, struct linux_mknod_args *args) 1107 { 1108 char *path; 1109 int error; 1110 1111 LCONVPATHCREAT(td, args->path, &path); 1112 1113 #ifdef DEBUG 1114 if (ldebug(mknod)) 1115 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1116 (uintmax_t)args->dev); 1117 #endif 1118 1119 switch (args->mode & S_IFMT) { 1120 case S_IFIFO: 1121 case S_IFSOCK: 1122 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE, 1123 args->mode); 1124 break; 1125 1126 case S_IFCHR: 1127 case S_IFBLK: 1128 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE, 1129 args->mode, args->dev); 1130 break; 1131 1132 case S_IFDIR: 1133 error = EPERM; 1134 break; 1135 1136 case 0: 1137 args->mode |= S_IFREG; 1138 /* FALLTHROUGH */ 1139 case S_IFREG: 1140 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE, 1141 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1142 if (error == 0) 1143 kern_close(td, td->td_retval[0]); 1144 break; 1145 1146 default: 1147 error = EINVAL; 1148 break; 1149 } 1150 LFREEPATH(path); 1151 return (error); 1152 } 1153 #endif 1154 1155 int 1156 linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1157 { 1158 char *path; 1159 int error, dfd; 1160 1161 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1162 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1163 1164 #ifdef DEBUG 1165 if (ldebug(mknodat)) 1166 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1167 #endif 1168 1169 switch (args->mode & S_IFMT) { 1170 case S_IFIFO: 1171 case S_IFSOCK: 1172 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1173 break; 1174 1175 case S_IFCHR: 1176 case S_IFBLK: 1177 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1178 args->dev); 1179 break; 1180 1181 case S_IFDIR: 1182 error = EPERM; 1183 break; 1184 1185 case 0: 1186 args->mode |= S_IFREG; 1187 /* FALLTHROUGH */ 1188 case S_IFREG: 1189 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1190 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1191 if (error == 0) 1192 kern_close(td, td->td_retval[0]); 1193 break; 1194 1195 default: 1196 error = EINVAL; 1197 break; 1198 } 1199 LFREEPATH(path); 1200 return (error); 1201 } 1202 1203 /* 1204 * UGH! This is just about the dumbest idea I've ever heard!! 1205 */ 1206 int 1207 linux_personality(struct thread *td, struct linux_personality_args *args) 1208 { 1209 struct linux_pemuldata *pem; 1210 struct proc *p = td->td_proc; 1211 uint32_t old; 1212 1213 #ifdef DEBUG 1214 if (ldebug(personality)) 1215 printf(ARGS(personality, "%u"), args->per); 1216 #endif 1217 1218 PROC_LOCK(p); 1219 pem = pem_find(p); 1220 old = pem->persona; 1221 if (args->per != 0xffffffff) 1222 pem->persona = args->per; 1223 PROC_UNLOCK(p); 1224 1225 td->td_retval[0] = old; 1226 return (0); 1227 } 1228 1229 struct l_itimerval { 1230 l_timeval it_interval; 1231 l_timeval it_value; 1232 }; 1233 1234 #define B2L_ITIMERVAL(bip, lip) \ 1235 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1236 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1237 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1238 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1239 1240 int 1241 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1242 { 1243 int error; 1244 struct l_itimerval ls; 1245 struct itimerval aitv, oitv; 1246 1247 #ifdef DEBUG 1248 if (ldebug(setitimer)) 1249 printf(ARGS(setitimer, "%p, %p"), 1250 (void *)uap->itv, (void *)uap->oitv); 1251 #endif 1252 1253 if (uap->itv == NULL) { 1254 uap->itv = uap->oitv; 1255 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1256 } 1257 1258 error = copyin(uap->itv, &ls, sizeof(ls)); 1259 if (error != 0) 1260 return (error); 1261 B2L_ITIMERVAL(&aitv, &ls); 1262 #ifdef DEBUG 1263 if (ldebug(setitimer)) { 1264 printf("setitimer: value: sec: %jd, usec: %ld\n", 1265 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1266 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1267 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1268 } 1269 #endif 1270 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1271 if (error != 0 || uap->oitv == NULL) 1272 return (error); 1273 B2L_ITIMERVAL(&ls, &oitv); 1274 1275 return (copyout(&ls, uap->oitv, sizeof(ls))); 1276 } 1277 1278 int 1279 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1280 { 1281 int error; 1282 struct l_itimerval ls; 1283 struct itimerval aitv; 1284 1285 #ifdef DEBUG 1286 if (ldebug(getitimer)) 1287 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1288 #endif 1289 error = kern_getitimer(td, uap->which, &aitv); 1290 if (error != 0) 1291 return (error); 1292 B2L_ITIMERVAL(&ls, &aitv); 1293 return (copyout(&ls, uap->itv, sizeof(ls))); 1294 } 1295 1296 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1297 int 1298 linux_nice(struct thread *td, struct linux_nice_args *args) 1299 { 1300 struct setpriority_args bsd_args; 1301 1302 bsd_args.which = PRIO_PROCESS; 1303 bsd_args.who = 0; /* current process */ 1304 bsd_args.prio = args->inc; 1305 return (sys_setpriority(td, &bsd_args)); 1306 } 1307 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1308 1309 int 1310 linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1311 { 1312 struct ucred *newcred, *oldcred; 1313 l_gid_t *linux_gidset; 1314 gid_t *bsd_gidset; 1315 int ngrp, error; 1316 struct proc *p; 1317 1318 ngrp = args->gidsetsize; 1319 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1320 return (EINVAL); 1321 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1322 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1323 if (error) 1324 goto out; 1325 newcred = crget(); 1326 crextend(newcred, ngrp + 1); 1327 p = td->td_proc; 1328 PROC_LOCK(p); 1329 oldcred = p->p_ucred; 1330 crcopy(newcred, oldcred); 1331 1332 /* 1333 * cr_groups[0] holds egid. Setting the whole set from 1334 * the supplied set will cause egid to be changed too. 1335 * Keep cr_groups[0] unchanged to prevent that. 1336 */ 1337 1338 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) { 1339 PROC_UNLOCK(p); 1340 crfree(newcred); 1341 goto out; 1342 } 1343 1344 if (ngrp > 0) { 1345 newcred->cr_ngroups = ngrp + 1; 1346 1347 bsd_gidset = newcred->cr_groups; 1348 ngrp--; 1349 while (ngrp >= 0) { 1350 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1351 ngrp--; 1352 } 1353 } else 1354 newcred->cr_ngroups = 1; 1355 1356 setsugid(p); 1357 proc_set_cred(p, newcred); 1358 PROC_UNLOCK(p); 1359 crfree(oldcred); 1360 error = 0; 1361 out: 1362 free(linux_gidset, M_LINUX); 1363 return (error); 1364 } 1365 1366 int 1367 linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1368 { 1369 struct ucred *cred; 1370 l_gid_t *linux_gidset; 1371 gid_t *bsd_gidset; 1372 int bsd_gidsetsz, ngrp, error; 1373 1374 cred = td->td_ucred; 1375 bsd_gidset = cred->cr_groups; 1376 bsd_gidsetsz = cred->cr_ngroups - 1; 1377 1378 /* 1379 * cr_groups[0] holds egid. Returning the whole set 1380 * here will cause a duplicate. Exclude cr_groups[0] 1381 * to prevent that. 1382 */ 1383 1384 if ((ngrp = args->gidsetsize) == 0) { 1385 td->td_retval[0] = bsd_gidsetsz; 1386 return (0); 1387 } 1388 1389 if (ngrp < bsd_gidsetsz) 1390 return (EINVAL); 1391 1392 ngrp = 0; 1393 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1394 M_LINUX, M_WAITOK); 1395 while (ngrp < bsd_gidsetsz) { 1396 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1397 ngrp++; 1398 } 1399 1400 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1401 free(linux_gidset, M_LINUX); 1402 if (error) 1403 return (error); 1404 1405 td->td_retval[0] = ngrp; 1406 return (0); 1407 } 1408 1409 int 1410 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1411 { 1412 struct rlimit bsd_rlim; 1413 struct l_rlimit rlim; 1414 u_int which; 1415 int error; 1416 1417 #ifdef DEBUG 1418 if (ldebug(setrlimit)) 1419 printf(ARGS(setrlimit, "%d, %p"), 1420 args->resource, (void *)args->rlim); 1421 #endif 1422 1423 if (args->resource >= LINUX_RLIM_NLIMITS) 1424 return (EINVAL); 1425 1426 which = linux_to_bsd_resource[args->resource]; 1427 if (which == -1) 1428 return (EINVAL); 1429 1430 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1431 if (error) 1432 return (error); 1433 1434 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1435 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1436 return (kern_setrlimit(td, which, &bsd_rlim)); 1437 } 1438 1439 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1440 int 1441 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1442 { 1443 struct l_rlimit rlim; 1444 struct rlimit bsd_rlim; 1445 u_int which; 1446 1447 #ifdef DEBUG 1448 if (ldebug(old_getrlimit)) 1449 printf(ARGS(old_getrlimit, "%d, %p"), 1450 args->resource, (void *)args->rlim); 1451 #endif 1452 1453 if (args->resource >= LINUX_RLIM_NLIMITS) 1454 return (EINVAL); 1455 1456 which = linux_to_bsd_resource[args->resource]; 1457 if (which == -1) 1458 return (EINVAL); 1459 1460 lim_rlimit(td, which, &bsd_rlim); 1461 1462 #ifdef COMPAT_LINUX32 1463 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1464 if (rlim.rlim_cur == UINT_MAX) 1465 rlim.rlim_cur = INT_MAX; 1466 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1467 if (rlim.rlim_max == UINT_MAX) 1468 rlim.rlim_max = INT_MAX; 1469 #else 1470 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1471 if (rlim.rlim_cur == ULONG_MAX) 1472 rlim.rlim_cur = LONG_MAX; 1473 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1474 if (rlim.rlim_max == ULONG_MAX) 1475 rlim.rlim_max = LONG_MAX; 1476 #endif 1477 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1478 } 1479 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1480 1481 int 1482 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1483 { 1484 struct l_rlimit rlim; 1485 struct rlimit bsd_rlim; 1486 u_int which; 1487 1488 #ifdef DEBUG 1489 if (ldebug(getrlimit)) 1490 printf(ARGS(getrlimit, "%d, %p"), 1491 args->resource, (void *)args->rlim); 1492 #endif 1493 1494 if (args->resource >= LINUX_RLIM_NLIMITS) 1495 return (EINVAL); 1496 1497 which = linux_to_bsd_resource[args->resource]; 1498 if (which == -1) 1499 return (EINVAL); 1500 1501 lim_rlimit(td, which, &bsd_rlim); 1502 1503 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1504 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1505 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1506 } 1507 1508 int 1509 linux_sched_setscheduler(struct thread *td, 1510 struct linux_sched_setscheduler_args *args) 1511 { 1512 struct sched_param sched_param; 1513 struct thread *tdt; 1514 int error, policy; 1515 1516 #ifdef DEBUG 1517 if (ldebug(sched_setscheduler)) 1518 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1519 args->pid, args->policy, (const void *)args->param); 1520 #endif 1521 1522 switch (args->policy) { 1523 case LINUX_SCHED_OTHER: 1524 policy = SCHED_OTHER; 1525 break; 1526 case LINUX_SCHED_FIFO: 1527 policy = SCHED_FIFO; 1528 break; 1529 case LINUX_SCHED_RR: 1530 policy = SCHED_RR; 1531 break; 1532 default: 1533 return (EINVAL); 1534 } 1535 1536 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1537 if (error) 1538 return (error); 1539 1540 tdt = linux_tdfind(td, args->pid, -1); 1541 if (tdt == NULL) 1542 return (ESRCH); 1543 1544 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1545 PROC_UNLOCK(tdt->td_proc); 1546 return (error); 1547 } 1548 1549 int 1550 linux_sched_getscheduler(struct thread *td, 1551 struct linux_sched_getscheduler_args *args) 1552 { 1553 struct thread *tdt; 1554 int error, policy; 1555 1556 #ifdef DEBUG 1557 if (ldebug(sched_getscheduler)) 1558 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1559 #endif 1560 1561 tdt = linux_tdfind(td, args->pid, -1); 1562 if (tdt == NULL) 1563 return (ESRCH); 1564 1565 error = kern_sched_getscheduler(td, tdt, &policy); 1566 PROC_UNLOCK(tdt->td_proc); 1567 1568 switch (policy) { 1569 case SCHED_OTHER: 1570 td->td_retval[0] = LINUX_SCHED_OTHER; 1571 break; 1572 case SCHED_FIFO: 1573 td->td_retval[0] = LINUX_SCHED_FIFO; 1574 break; 1575 case SCHED_RR: 1576 td->td_retval[0] = LINUX_SCHED_RR; 1577 break; 1578 } 1579 return (error); 1580 } 1581 1582 int 1583 linux_sched_get_priority_max(struct thread *td, 1584 struct linux_sched_get_priority_max_args *args) 1585 { 1586 struct sched_get_priority_max_args bsd; 1587 1588 #ifdef DEBUG 1589 if (ldebug(sched_get_priority_max)) 1590 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1591 #endif 1592 1593 switch (args->policy) { 1594 case LINUX_SCHED_OTHER: 1595 bsd.policy = SCHED_OTHER; 1596 break; 1597 case LINUX_SCHED_FIFO: 1598 bsd.policy = SCHED_FIFO; 1599 break; 1600 case LINUX_SCHED_RR: 1601 bsd.policy = SCHED_RR; 1602 break; 1603 default: 1604 return (EINVAL); 1605 } 1606 return (sys_sched_get_priority_max(td, &bsd)); 1607 } 1608 1609 int 1610 linux_sched_get_priority_min(struct thread *td, 1611 struct linux_sched_get_priority_min_args *args) 1612 { 1613 struct sched_get_priority_min_args bsd; 1614 1615 #ifdef DEBUG 1616 if (ldebug(sched_get_priority_min)) 1617 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1618 #endif 1619 1620 switch (args->policy) { 1621 case LINUX_SCHED_OTHER: 1622 bsd.policy = SCHED_OTHER; 1623 break; 1624 case LINUX_SCHED_FIFO: 1625 bsd.policy = SCHED_FIFO; 1626 break; 1627 case LINUX_SCHED_RR: 1628 bsd.policy = SCHED_RR; 1629 break; 1630 default: 1631 return (EINVAL); 1632 } 1633 return (sys_sched_get_priority_min(td, &bsd)); 1634 } 1635 1636 #define REBOOT_CAD_ON 0x89abcdef 1637 #define REBOOT_CAD_OFF 0 1638 #define REBOOT_HALT 0xcdef0123 1639 #define REBOOT_RESTART 0x01234567 1640 #define REBOOT_RESTART2 0xA1B2C3D4 1641 #define REBOOT_POWEROFF 0x4321FEDC 1642 #define REBOOT_MAGIC1 0xfee1dead 1643 #define REBOOT_MAGIC2 0x28121969 1644 #define REBOOT_MAGIC2A 0x05121996 1645 #define REBOOT_MAGIC2B 0x16041998 1646 1647 int 1648 linux_reboot(struct thread *td, struct linux_reboot_args *args) 1649 { 1650 struct reboot_args bsd_args; 1651 1652 #ifdef DEBUG 1653 if (ldebug(reboot)) 1654 printf(ARGS(reboot, "0x%x"), args->cmd); 1655 #endif 1656 1657 if (args->magic1 != REBOOT_MAGIC1) 1658 return (EINVAL); 1659 1660 switch (args->magic2) { 1661 case REBOOT_MAGIC2: 1662 case REBOOT_MAGIC2A: 1663 case REBOOT_MAGIC2B: 1664 break; 1665 default: 1666 return (EINVAL); 1667 } 1668 1669 switch (args->cmd) { 1670 case REBOOT_CAD_ON: 1671 case REBOOT_CAD_OFF: 1672 return (priv_check(td, PRIV_REBOOT)); 1673 case REBOOT_HALT: 1674 bsd_args.opt = RB_HALT; 1675 break; 1676 case REBOOT_RESTART: 1677 case REBOOT_RESTART2: 1678 bsd_args.opt = 0; 1679 break; 1680 case REBOOT_POWEROFF: 1681 bsd_args.opt = RB_POWEROFF; 1682 break; 1683 default: 1684 return (EINVAL); 1685 } 1686 return (sys_reboot(td, &bsd_args)); 1687 } 1688 1689 1690 /* 1691 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1692 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1693 * are assumed to be preserved. The following lightweight syscalls fixes 1694 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1695 * 1696 * linux_getpid() - MP SAFE 1697 * linux_getgid() - MP SAFE 1698 * linux_getuid() - MP SAFE 1699 */ 1700 1701 int 1702 linux_getpid(struct thread *td, struct linux_getpid_args *args) 1703 { 1704 1705 #ifdef DEBUG 1706 if (ldebug(getpid)) 1707 printf(ARGS(getpid, "")); 1708 #endif 1709 td->td_retval[0] = td->td_proc->p_pid; 1710 1711 return (0); 1712 } 1713 1714 int 1715 linux_gettid(struct thread *td, struct linux_gettid_args *args) 1716 { 1717 struct linux_emuldata *em; 1718 1719 #ifdef DEBUG 1720 if (ldebug(gettid)) 1721 printf(ARGS(gettid, "")); 1722 #endif 1723 1724 em = em_find(td); 1725 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1726 1727 td->td_retval[0] = em->em_tid; 1728 1729 return (0); 1730 } 1731 1732 1733 int 1734 linux_getppid(struct thread *td, struct linux_getppid_args *args) 1735 { 1736 1737 #ifdef DEBUG 1738 if (ldebug(getppid)) 1739 printf(ARGS(getppid, "")); 1740 #endif 1741 1742 td->td_retval[0] = kern_getppid(td); 1743 return (0); 1744 } 1745 1746 int 1747 linux_getgid(struct thread *td, struct linux_getgid_args *args) 1748 { 1749 1750 #ifdef DEBUG 1751 if (ldebug(getgid)) 1752 printf(ARGS(getgid, "")); 1753 #endif 1754 1755 td->td_retval[0] = td->td_ucred->cr_rgid; 1756 return (0); 1757 } 1758 1759 int 1760 linux_getuid(struct thread *td, struct linux_getuid_args *args) 1761 { 1762 1763 #ifdef DEBUG 1764 if (ldebug(getuid)) 1765 printf(ARGS(getuid, "")); 1766 #endif 1767 1768 td->td_retval[0] = td->td_ucred->cr_ruid; 1769 return (0); 1770 } 1771 1772 1773 int 1774 linux_getsid(struct thread *td, struct linux_getsid_args *args) 1775 { 1776 struct getsid_args bsd; 1777 1778 #ifdef DEBUG 1779 if (ldebug(getsid)) 1780 printf(ARGS(getsid, "%i"), args->pid); 1781 #endif 1782 1783 bsd.pid = args->pid; 1784 return (sys_getsid(td, &bsd)); 1785 } 1786 1787 int 1788 linux_nosys(struct thread *td, struct nosys_args *ignore) 1789 { 1790 1791 return (ENOSYS); 1792 } 1793 1794 int 1795 linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1796 { 1797 struct getpriority_args bsd_args; 1798 int error; 1799 1800 #ifdef DEBUG 1801 if (ldebug(getpriority)) 1802 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1803 #endif 1804 1805 bsd_args.which = args->which; 1806 bsd_args.who = args->who; 1807 error = sys_getpriority(td, &bsd_args); 1808 td->td_retval[0] = 20 - td->td_retval[0]; 1809 return (error); 1810 } 1811 1812 int 1813 linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1814 { 1815 int name[2]; 1816 1817 #ifdef DEBUG 1818 if (ldebug(sethostname)) 1819 printf(ARGS(sethostname, "*, %i"), args->len); 1820 #endif 1821 1822 name[0] = CTL_KERN; 1823 name[1] = KERN_HOSTNAME; 1824 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1825 args->len, 0, 0)); 1826 } 1827 1828 int 1829 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1830 { 1831 int name[2]; 1832 1833 #ifdef DEBUG 1834 if (ldebug(setdomainname)) 1835 printf(ARGS(setdomainname, "*, %i"), args->len); 1836 #endif 1837 1838 name[0] = CTL_KERN; 1839 name[1] = KERN_NISDOMAINNAME; 1840 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1841 args->len, 0, 0)); 1842 } 1843 1844 int 1845 linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1846 { 1847 1848 #ifdef DEBUG 1849 if (ldebug(exit_group)) 1850 printf(ARGS(exit_group, "%i"), args->error_code); 1851 #endif 1852 1853 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1854 args->error_code); 1855 1856 /* 1857 * XXX: we should send a signal to the parent if 1858 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1859 * as it doesnt occur often. 1860 */ 1861 exit1(td, args->error_code, 0); 1862 /* NOTREACHED */ 1863 } 1864 1865 #define _LINUX_CAPABILITY_VERSION_1 0x19980330 1866 #define _LINUX_CAPABILITY_VERSION_2 0x20071026 1867 #define _LINUX_CAPABILITY_VERSION_3 0x20080522 1868 1869 struct l_user_cap_header { 1870 l_int version; 1871 l_int pid; 1872 }; 1873 1874 struct l_user_cap_data { 1875 l_int effective; 1876 l_int permitted; 1877 l_int inheritable; 1878 }; 1879 1880 int 1881 linux_capget(struct thread *td, struct linux_capget_args *uap) 1882 { 1883 struct l_user_cap_header luch; 1884 struct l_user_cap_data lucd[2]; 1885 int error, u32s; 1886 1887 if (uap->hdrp == NULL) 1888 return (EFAULT); 1889 1890 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1891 if (error != 0) 1892 return (error); 1893 1894 switch (luch.version) { 1895 case _LINUX_CAPABILITY_VERSION_1: 1896 u32s = 1; 1897 break; 1898 case _LINUX_CAPABILITY_VERSION_2: 1899 case _LINUX_CAPABILITY_VERSION_3: 1900 u32s = 2; 1901 break; 1902 default: 1903 #ifdef DEBUG 1904 if (ldebug(capget)) 1905 printf(LMSG("invalid capget capability version 0x%x"), 1906 luch.version); 1907 #endif 1908 luch.version = _LINUX_CAPABILITY_VERSION_1; 1909 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1910 if (error) 1911 return (error); 1912 return (EINVAL); 1913 } 1914 1915 if (luch.pid) 1916 return (EPERM); 1917 1918 if (uap->datap) { 1919 /* 1920 * The current implementation doesn't support setting 1921 * a capability (it's essentially a stub) so indicate 1922 * that no capabilities are currently set or available 1923 * to request. 1924 */ 1925 memset(&lucd, 0, u32s * sizeof(lucd[0])); 1926 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0])); 1927 } 1928 1929 return (error); 1930 } 1931 1932 int 1933 linux_capset(struct thread *td, struct linux_capset_args *uap) 1934 { 1935 struct l_user_cap_header luch; 1936 struct l_user_cap_data lucd[2]; 1937 int error, i, u32s; 1938 1939 if (uap->hdrp == NULL || uap->datap == NULL) 1940 return (EFAULT); 1941 1942 error = copyin(uap->hdrp, &luch, sizeof(luch)); 1943 if (error != 0) 1944 return (error); 1945 1946 switch (luch.version) { 1947 case _LINUX_CAPABILITY_VERSION_1: 1948 u32s = 1; 1949 break; 1950 case _LINUX_CAPABILITY_VERSION_2: 1951 case _LINUX_CAPABILITY_VERSION_3: 1952 u32s = 2; 1953 break; 1954 default: 1955 #ifdef DEBUG 1956 if (ldebug(capset)) 1957 printf(LMSG("invalid capset capability version 0x%x"), 1958 luch.version); 1959 #endif 1960 luch.version = _LINUX_CAPABILITY_VERSION_1; 1961 error = copyout(&luch, uap->hdrp, sizeof(luch)); 1962 if (error) 1963 return (error); 1964 return (EINVAL); 1965 } 1966 1967 if (luch.pid) 1968 return (EPERM); 1969 1970 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0])); 1971 if (error != 0) 1972 return (error); 1973 1974 /* We currently don't support setting any capabilities. */ 1975 for (i = 0; i < u32s; i++) { 1976 if (lucd[i].effective || lucd[i].permitted || 1977 lucd[i].inheritable) { 1978 linux_msg(td, 1979 "capset[%d] effective=0x%x, permitted=0x%x, " 1980 "inheritable=0x%x is not implemented", i, 1981 (int)lucd[i].effective, (int)lucd[i].permitted, 1982 (int)lucd[i].inheritable); 1983 return (EPERM); 1984 } 1985 } 1986 1987 return (0); 1988 } 1989 1990 int 1991 linux_prctl(struct thread *td, struct linux_prctl_args *args) 1992 { 1993 int error = 0, max_size; 1994 struct proc *p = td->td_proc; 1995 char comm[LINUX_MAX_COMM_LEN]; 1996 struct linux_emuldata *em; 1997 int pdeath_signal; 1998 1999 #ifdef DEBUG 2000 if (ldebug(prctl)) 2001 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 2002 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 2003 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 2004 #endif 2005 2006 switch (args->option) { 2007 case LINUX_PR_SET_PDEATHSIG: 2008 if (!LINUX_SIG_VALID(args->arg2)) 2009 return (EINVAL); 2010 em = em_find(td); 2011 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 2012 em->pdeath_signal = args->arg2; 2013 break; 2014 case LINUX_PR_GET_PDEATHSIG: 2015 em = em_find(td); 2016 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 2017 pdeath_signal = em->pdeath_signal; 2018 error = copyout(&pdeath_signal, 2019 (void *)(register_t)args->arg2, 2020 sizeof(pdeath_signal)); 2021 break; 2022 case LINUX_PR_GET_KEEPCAPS: 2023 /* 2024 * Indicate that we always clear the effective and 2025 * permitted capability sets when the user id becomes 2026 * non-zero (actually the capability sets are simply 2027 * always zero in the current implementation). 2028 */ 2029 td->td_retval[0] = 0; 2030 break; 2031 case LINUX_PR_SET_KEEPCAPS: 2032 /* 2033 * Ignore requests to keep the effective and permitted 2034 * capability sets when the user id becomes non-zero. 2035 */ 2036 break; 2037 case LINUX_PR_SET_NAME: 2038 /* 2039 * To be on the safe side we need to make sure to not 2040 * overflow the size a Linux program expects. We already 2041 * do this here in the copyin, so that we don't need to 2042 * check on copyout. 2043 */ 2044 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 2045 error = copyinstr((void *)(register_t)args->arg2, comm, 2046 max_size, NULL); 2047 2048 /* Linux silently truncates the name if it is too long. */ 2049 if (error == ENAMETOOLONG) { 2050 /* 2051 * XXX: copyinstr() isn't documented to populate the 2052 * array completely, so do a copyin() to be on the 2053 * safe side. This should be changed in case 2054 * copyinstr() is changed to guarantee this. 2055 */ 2056 error = copyin((void *)(register_t)args->arg2, comm, 2057 max_size - 1); 2058 comm[max_size - 1] = '\0'; 2059 } 2060 if (error) 2061 return (error); 2062 2063 PROC_LOCK(p); 2064 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 2065 PROC_UNLOCK(p); 2066 break; 2067 case LINUX_PR_GET_NAME: 2068 PROC_LOCK(p); 2069 strlcpy(comm, p->p_comm, sizeof(comm)); 2070 PROC_UNLOCK(p); 2071 error = copyout(comm, (void *)(register_t)args->arg2, 2072 strlen(comm) + 1); 2073 break; 2074 default: 2075 error = EINVAL; 2076 break; 2077 } 2078 2079 return (error); 2080 } 2081 2082 int 2083 linux_sched_setparam(struct thread *td, 2084 struct linux_sched_setparam_args *uap) 2085 { 2086 struct sched_param sched_param; 2087 struct thread *tdt; 2088 int error; 2089 2090 #ifdef DEBUG 2091 if (ldebug(sched_setparam)) 2092 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 2093 #endif 2094 2095 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 2096 if (error) 2097 return (error); 2098 2099 tdt = linux_tdfind(td, uap->pid, -1); 2100 if (tdt == NULL) 2101 return (ESRCH); 2102 2103 error = kern_sched_setparam(td, tdt, &sched_param); 2104 PROC_UNLOCK(tdt->td_proc); 2105 return (error); 2106 } 2107 2108 int 2109 linux_sched_getparam(struct thread *td, 2110 struct linux_sched_getparam_args *uap) 2111 { 2112 struct sched_param sched_param; 2113 struct thread *tdt; 2114 int error; 2115 2116 #ifdef DEBUG 2117 if (ldebug(sched_getparam)) 2118 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 2119 #endif 2120 2121 tdt = linux_tdfind(td, uap->pid, -1); 2122 if (tdt == NULL) 2123 return (ESRCH); 2124 2125 error = kern_sched_getparam(td, tdt, &sched_param); 2126 PROC_UNLOCK(tdt->td_proc); 2127 if (error == 0) 2128 error = copyout(&sched_param, uap->param, 2129 sizeof(sched_param)); 2130 return (error); 2131 } 2132 2133 /* 2134 * Get affinity of a process. 2135 */ 2136 int 2137 linux_sched_getaffinity(struct thread *td, 2138 struct linux_sched_getaffinity_args *args) 2139 { 2140 int error; 2141 struct thread *tdt; 2142 2143 #ifdef DEBUG 2144 if (ldebug(sched_getaffinity)) 2145 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2146 args->len); 2147 #endif 2148 if (args->len < sizeof(cpuset_t)) 2149 return (EINVAL); 2150 2151 tdt = linux_tdfind(td, args->pid, -1); 2152 if (tdt == NULL) 2153 return (ESRCH); 2154 2155 PROC_UNLOCK(tdt->td_proc); 2156 2157 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 2158 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr); 2159 if (error == 0) 2160 td->td_retval[0] = sizeof(cpuset_t); 2161 2162 return (error); 2163 } 2164 2165 /* 2166 * Set affinity of a process. 2167 */ 2168 int 2169 linux_sched_setaffinity(struct thread *td, 2170 struct linux_sched_setaffinity_args *args) 2171 { 2172 struct thread *tdt; 2173 2174 #ifdef DEBUG 2175 if (ldebug(sched_setaffinity)) 2176 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2177 args->len); 2178 #endif 2179 if (args->len < sizeof(cpuset_t)) 2180 return (EINVAL); 2181 2182 tdt = linux_tdfind(td, args->pid, -1); 2183 if (tdt == NULL) 2184 return (ESRCH); 2185 2186 PROC_UNLOCK(tdt->td_proc); 2187 2188 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID, 2189 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr)); 2190 } 2191 2192 struct linux_rlimit64 { 2193 uint64_t rlim_cur; 2194 uint64_t rlim_max; 2195 }; 2196 2197 int 2198 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2199 { 2200 struct rlimit rlim, nrlim; 2201 struct linux_rlimit64 lrlim; 2202 struct proc *p; 2203 u_int which; 2204 int flags; 2205 int error; 2206 2207 #ifdef DEBUG 2208 if (ldebug(prlimit64)) 2209 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2210 args->resource, (void *)args->new, (void *)args->old); 2211 #endif 2212 2213 if (args->resource >= LINUX_RLIM_NLIMITS) 2214 return (EINVAL); 2215 2216 which = linux_to_bsd_resource[args->resource]; 2217 if (which == -1) 2218 return (EINVAL); 2219 2220 if (args->new != NULL) { 2221 /* 2222 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2223 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2224 * as INFINITY so we do not need a conversion even. 2225 */ 2226 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2227 if (error != 0) 2228 return (error); 2229 } 2230 2231 flags = PGET_HOLD | PGET_NOTWEXIT; 2232 if (args->new != NULL) 2233 flags |= PGET_CANDEBUG; 2234 else 2235 flags |= PGET_CANSEE; 2236 error = pget(args->pid, flags, &p); 2237 if (error != 0) 2238 return (error); 2239 2240 if (args->old != NULL) { 2241 PROC_LOCK(p); 2242 lim_rlimit_proc(p, which, &rlim); 2243 PROC_UNLOCK(p); 2244 if (rlim.rlim_cur == RLIM_INFINITY) 2245 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2246 else 2247 lrlim.rlim_cur = rlim.rlim_cur; 2248 if (rlim.rlim_max == RLIM_INFINITY) 2249 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2250 else 2251 lrlim.rlim_max = rlim.rlim_max; 2252 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2253 if (error != 0) 2254 goto out; 2255 } 2256 2257 if (args->new != NULL) 2258 error = kern_proc_setrlimit(td, p, which, &nrlim); 2259 2260 out: 2261 PRELE(p); 2262 return (error); 2263 } 2264 2265 int 2266 linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2267 { 2268 struct timeval utv, tv0, tv1, *tvp; 2269 struct l_pselect6arg lpse6; 2270 struct l_timespec lts; 2271 struct timespec uts; 2272 l_sigset_t l_ss; 2273 sigset_t *ssp; 2274 sigset_t ss; 2275 int error; 2276 2277 ssp = NULL; 2278 if (args->sig != NULL) { 2279 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2280 if (error != 0) 2281 return (error); 2282 if (lpse6.ss_len != sizeof(l_ss)) 2283 return (EINVAL); 2284 if (lpse6.ss != 0) { 2285 error = copyin(PTRIN(lpse6.ss), &l_ss, 2286 sizeof(l_ss)); 2287 if (error != 0) 2288 return (error); 2289 linux_to_bsd_sigset(&l_ss, &ss); 2290 ssp = &ss; 2291 } 2292 } 2293 2294 /* 2295 * Currently glibc changes nanosecond number to microsecond. 2296 * This mean losing precision but for now it is hardly seen. 2297 */ 2298 if (args->tsp != NULL) { 2299 error = copyin(args->tsp, <s, sizeof(lts)); 2300 if (error != 0) 2301 return (error); 2302 error = linux_to_native_timespec(&uts, <s); 2303 if (error != 0) 2304 return (error); 2305 2306 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2307 if (itimerfix(&utv)) 2308 return (EINVAL); 2309 2310 microtime(&tv0); 2311 tvp = &utv; 2312 } else 2313 tvp = NULL; 2314 2315 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2316 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2317 2318 if (error == 0 && args->tsp != NULL) { 2319 if (td->td_retval[0] != 0) { 2320 /* 2321 * Compute how much time was left of the timeout, 2322 * by subtracting the current time and the time 2323 * before we started the call, and subtracting 2324 * that result from the user-supplied value. 2325 */ 2326 2327 microtime(&tv1); 2328 timevalsub(&tv1, &tv0); 2329 timevalsub(&utv, &tv1); 2330 if (utv.tv_sec < 0) 2331 timevalclear(&utv); 2332 } else 2333 timevalclear(&utv); 2334 2335 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2336 2337 error = native_to_linux_timespec(<s, &uts); 2338 if (error == 0) 2339 error = copyout(<s, args->tsp, sizeof(lts)); 2340 } 2341 2342 return (error); 2343 } 2344 2345 int 2346 linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2347 { 2348 struct timespec ts0, ts1; 2349 struct l_timespec lts; 2350 struct timespec uts, *tsp; 2351 l_sigset_t l_ss; 2352 sigset_t *ssp; 2353 sigset_t ss; 2354 int error; 2355 2356 if (args->sset != NULL) { 2357 if (args->ssize != sizeof(l_ss)) 2358 return (EINVAL); 2359 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2360 if (error) 2361 return (error); 2362 linux_to_bsd_sigset(&l_ss, &ss); 2363 ssp = &ss; 2364 } else 2365 ssp = NULL; 2366 if (args->tsp != NULL) { 2367 error = copyin(args->tsp, <s, sizeof(lts)); 2368 if (error) 2369 return (error); 2370 error = linux_to_native_timespec(&uts, <s); 2371 if (error != 0) 2372 return (error); 2373 2374 nanotime(&ts0); 2375 tsp = &uts; 2376 } else 2377 tsp = NULL; 2378 2379 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2380 2381 if (error == 0 && args->tsp != NULL) { 2382 if (td->td_retval[0]) { 2383 nanotime(&ts1); 2384 timespecsub(&ts1, &ts0, &ts1); 2385 timespecsub(&uts, &ts1, &uts); 2386 if (uts.tv_sec < 0) 2387 timespecclear(&uts); 2388 } else 2389 timespecclear(&uts); 2390 2391 error = native_to_linux_timespec(<s, &uts); 2392 if (error == 0) 2393 error = copyout(<s, args->tsp, sizeof(lts)); 2394 } 2395 2396 return (error); 2397 } 2398 2399 #if defined(DEBUG) || defined(KTR) 2400 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2401 2402 #ifdef COMPAT_LINUX32 2403 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL 2404 #else 2405 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL 2406 #endif 2407 2408 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))]; 2409 2410 static int 2411 linux_debug(int syscall, int toggle, int global) 2412 { 2413 2414 if (global) { 2415 char c = toggle ? 0 : 0xff; 2416 2417 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2418 return (0); 2419 } 2420 if (syscall < 0 || syscall >= L_MAXSYSCALL) 2421 return (EINVAL); 2422 if (toggle) 2423 clrbit(linux_debug_map, syscall); 2424 else 2425 setbit(linux_debug_map, syscall); 2426 return (0); 2427 } 2428 #undef L_MAXSYSCALL 2429 2430 /* 2431 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2432 * 2433 * E.g.: sysctl linux.debug=21.0 2434 * 2435 * As a special case, syscall "all" will apply to all syscalls globally. 2436 */ 2437 #define LINUX_MAX_DEBUGSTR 16 2438 int 2439 linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2440 { 2441 char value[LINUX_MAX_DEBUGSTR], *p; 2442 int error, sysc, toggle; 2443 int global = 0; 2444 2445 value[0] = '\0'; 2446 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2447 if (error || req->newptr == NULL) 2448 return (error); 2449 for (p = value; *p != '\0' && *p != '.'; p++); 2450 if (*p == '\0') 2451 return (EINVAL); 2452 *p++ = '\0'; 2453 sysc = strtol(value, NULL, 0); 2454 toggle = strtol(p, NULL, 0); 2455 if (strcmp(value, "all") == 0) 2456 global = 1; 2457 error = linux_debug(sysc, toggle, global); 2458 return (error); 2459 } 2460 2461 #endif /* DEBUG || KTR */ 2462 2463 int 2464 linux_sched_rr_get_interval(struct thread *td, 2465 struct linux_sched_rr_get_interval_args *uap) 2466 { 2467 struct timespec ts; 2468 struct l_timespec lts; 2469 struct thread *tdt; 2470 int error; 2471 2472 /* 2473 * According to man in case the invalid pid specified 2474 * EINVAL should be returned. 2475 */ 2476 if (uap->pid < 0) 2477 return (EINVAL); 2478 2479 tdt = linux_tdfind(td, uap->pid, -1); 2480 if (tdt == NULL) 2481 return (ESRCH); 2482 2483 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2484 PROC_UNLOCK(tdt->td_proc); 2485 if (error != 0) 2486 return (error); 2487 error = native_to_linux_timespec(<s, &ts); 2488 if (error != 0) 2489 return (error); 2490 return (copyout(<s, uap->interval, sizeof(lts))); 2491 } 2492 2493 /* 2494 * In case when the Linux thread is the initial thread in 2495 * the thread group thread id is equal to the process id. 2496 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2497 */ 2498 struct thread * 2499 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2500 { 2501 struct linux_emuldata *em; 2502 struct thread *tdt; 2503 struct proc *p; 2504 2505 tdt = NULL; 2506 if (tid == 0 || tid == td->td_tid) { 2507 tdt = td; 2508 PROC_LOCK(tdt->td_proc); 2509 } else if (tid > PID_MAX) 2510 tdt = tdfind(tid, pid); 2511 else { 2512 /* 2513 * Initial thread where the tid equal to the pid. 2514 */ 2515 p = pfind(tid); 2516 if (p != NULL) { 2517 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2518 /* 2519 * p is not a Linuxulator process. 2520 */ 2521 PROC_UNLOCK(p); 2522 return (NULL); 2523 } 2524 FOREACH_THREAD_IN_PROC(p, tdt) { 2525 em = em_find(tdt); 2526 if (tid == em->em_tid) 2527 return (tdt); 2528 } 2529 PROC_UNLOCK(p); 2530 } 2531 return (NULL); 2532 } 2533 2534 return (tdt); 2535 } 2536 2537 void 2538 linux_to_bsd_waitopts(int options, int *bsdopts) 2539 { 2540 2541 if (options & LINUX_WNOHANG) 2542 *bsdopts |= WNOHANG; 2543 if (options & LINUX_WUNTRACED) 2544 *bsdopts |= WUNTRACED; 2545 if (options & LINUX_WEXITED) 2546 *bsdopts |= WEXITED; 2547 if (options & LINUX_WCONTINUED) 2548 *bsdopts |= WCONTINUED; 2549 if (options & LINUX_WNOWAIT) 2550 *bsdopts |= WNOWAIT; 2551 2552 if (options & __WCLONE) 2553 *bsdopts |= WLINUXCLONE; 2554 } 2555 2556 int 2557 linux_getrandom(struct thread *td, struct linux_getrandom_args *args) 2558 { 2559 struct uio uio; 2560 struct iovec iov; 2561 int error; 2562 2563 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM)) 2564 return (EINVAL); 2565 if (args->count > INT_MAX) 2566 args->count = INT_MAX; 2567 2568 iov.iov_base = args->buf; 2569 iov.iov_len = args->count; 2570 2571 uio.uio_iov = &iov; 2572 uio.uio_iovcnt = 1; 2573 uio.uio_resid = iov.iov_len; 2574 uio.uio_segflg = UIO_USERSPACE; 2575 uio.uio_rw = UIO_READ; 2576 uio.uio_td = td; 2577 2578 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK); 2579 if (error == 0) 2580 td->td_retval[0] = args->count - uio.uio_resid; 2581 return (error); 2582 } 2583 2584 int 2585 linux_mincore(struct thread *td, struct linux_mincore_args *args) 2586 { 2587 2588 /* Needs to be page-aligned */ 2589 if (args->start & PAGE_MASK) 2590 return (EINVAL); 2591 return (kern_mincore(td, args->start, args->len, args->vec)); 2592 } 2593