linux_misc.c revision 293565
1/*- 2 * Copyright (c) 2002 Doug Rabson 3 * Copyright (c) 1994-1995 S��ren Schmidt 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer 11 * in this position and unchanged. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. The name of the author may not be used to endorse or promote products 16 * derived from this software without specific prior written permission 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 28 */ 29 30#include <sys/cdefs.h> 31__FBSDID("$FreeBSD: stable/10/sys/compat/linux/linux_misc.c 293565 2016-01-09 17:11:25Z dchagin $"); 32 33#include "opt_compat.h" 34#include "opt_kdtrace.h" 35 36#include <sys/param.h> 37#include <sys/blist.h> 38#include <sys/fcntl.h> 39#if defined(__i386__) 40#include <sys/imgact_aout.h> 41#endif 42#include <sys/jail.h> 43#include <sys/kernel.h> 44#include <sys/limits.h> 45#include <sys/lock.h> 46#include <sys/malloc.h> 47#include <sys/mman.h> 48#include <sys/mount.h> 49#include <sys/mutex.h> 50#include <sys/namei.h> 51#include <sys/priv.h> 52#include <sys/proc.h> 53#include <sys/reboot.h> 54#include <sys/racct.h> 55#include <sys/resourcevar.h> 56#include <sys/sched.h> 57#include <sys/sdt.h> 58#include <sys/signalvar.h> 59#include <sys/stat.h> 60#include <sys/syscallsubr.h> 61#include <sys/sysctl.h> 62#include <sys/sysproto.h> 63#include <sys/systm.h> 64#include <sys/time.h> 65#include <sys/vmmeter.h> 66#include <sys/vnode.h> 67#include <sys/wait.h> 68#include <sys/cpuset.h> 69 70#include <security/mac/mac_framework.h> 71 72#include <vm/vm.h> 73#include <vm/pmap.h> 74#include <vm/vm_kern.h> 75#include <vm/vm_map.h> 76#include <vm/vm_extern.h> 77#include <vm/vm_object.h> 78#include <vm/swap_pager.h> 79 80#ifdef COMPAT_LINUX32 81#include <machine/../linux32/linux.h> 82#include <machine/../linux32/linux32_proto.h> 83#else 84#include <machine/../linux/linux.h> 85#include <machine/../linux/linux_proto.h> 86#endif 87 88#include <compat/linux/linux_dtrace.h> 89#include <compat/linux/linux_file.h> 90#include <compat/linux/linux_mib.h> 91#include <compat/linux/linux_signal.h> 92#include <compat/linux/linux_timer.h> 93#include <compat/linux/linux_util.h> 94#include <compat/linux/linux_sysproto.h> 95#include <compat/linux/linux_emul.h> 96#include <compat/linux/linux_misc.h> 97 98/** 99 * Special DTrace provider for the linuxulator. 100 * 101 * In this file we define the provider for the entire linuxulator. All 102 * modules (= files of the linuxulator) use it. 103 * 104 * We define a different name depending on the emulated bitsize, see 105 * ../../<ARCH>/linux{,32}/linux.h, e.g.: 106 * native bitsize = linuxulator 107 * amd64, 32bit emulation = linuxulator32 108 */ 109LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE); 110 111int stclohz; /* Statistics clock frequency */ 112 113static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 114 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 115 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 116 RLIMIT_MEMLOCK, RLIMIT_AS 117}; 118 119struct l_sysinfo { 120 l_long uptime; /* Seconds since boot */ 121 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 122#define LINUX_SYSINFO_LOADS_SCALE 65536 123 l_ulong totalram; /* Total usable main memory size */ 124 l_ulong freeram; /* Available memory size */ 125 l_ulong sharedram; /* Amount of shared memory */ 126 l_ulong bufferram; /* Memory used by buffers */ 127 l_ulong totalswap; /* Total swap space size */ 128 l_ulong freeswap; /* swap space still available */ 129 l_ushort procs; /* Number of current processes */ 130 l_ushort pads; 131 l_ulong totalbig; 132 l_ulong freebig; 133 l_uint mem_unit; 134 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 135}; 136 137struct l_pselect6arg { 138 l_uintptr_t ss; 139 l_size_t ss_len; 140}; 141 142int 143linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 144{ 145 struct l_sysinfo sysinfo; 146 vm_object_t object; 147 int i, j; 148 struct timespec ts; 149 150 getnanouptime(&ts); 151 if (ts.tv_nsec != 0) 152 ts.tv_sec++; 153 sysinfo.uptime = ts.tv_sec; 154 155 /* Use the information from the mib to get our load averages */ 156 for (i = 0; i < 3; i++) 157 sysinfo.loads[i] = averunnable.ldavg[i] * 158 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 159 160 sysinfo.totalram = physmem * PAGE_SIZE; 161 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE; 162 163 sysinfo.sharedram = 0; 164 mtx_lock(&vm_object_list_mtx); 165 TAILQ_FOREACH(object, &vm_object_list, object_list) 166 if (object->shadow_count > 1) 167 sysinfo.sharedram += object->resident_page_count; 168 mtx_unlock(&vm_object_list_mtx); 169 170 sysinfo.sharedram *= PAGE_SIZE; 171 sysinfo.bufferram = 0; 172 173 swap_pager_status(&i, &j); 174 sysinfo.totalswap = i * PAGE_SIZE; 175 sysinfo.freeswap = (i - j) * PAGE_SIZE; 176 177 sysinfo.procs = nprocs; 178 179 /* The following are only present in newer Linux kernels. */ 180 sysinfo.totalbig = 0; 181 sysinfo.freebig = 0; 182 sysinfo.mem_unit = 1; 183 184 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 185} 186 187int 188linux_alarm(struct thread *td, struct linux_alarm_args *args) 189{ 190 struct itimerval it, old_it; 191 u_int secs; 192 int error; 193 194#ifdef DEBUG 195 if (ldebug(alarm)) 196 printf(ARGS(alarm, "%u"), args->secs); 197#endif 198 199 secs = args->secs; 200 201 if (secs > INT_MAX) 202 secs = INT_MAX; 203 204 it.it_value.tv_sec = (long) secs; 205 it.it_value.tv_usec = 0; 206 it.it_interval.tv_sec = 0; 207 it.it_interval.tv_usec = 0; 208 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 209 if (error) 210 return (error); 211 if (timevalisset(&old_it.it_value)) { 212 if (old_it.it_value.tv_usec != 0) 213 old_it.it_value.tv_sec++; 214 td->td_retval[0] = old_it.it_value.tv_sec; 215 } 216 return (0); 217} 218 219int 220linux_brk(struct thread *td, struct linux_brk_args *args) 221{ 222 struct vmspace *vm = td->td_proc->p_vmspace; 223 vm_offset_t new, old; 224 struct obreak_args /* { 225 char * nsize; 226 } */ tmp; 227 228#ifdef DEBUG 229 if (ldebug(brk)) 230 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 231#endif 232 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 233 new = (vm_offset_t)args->dsend; 234 tmp.nsize = (char *)new; 235 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp)) 236 td->td_retval[0] = (long)new; 237 else 238 td->td_retval[0] = (long)old; 239 240 return (0); 241} 242 243#if defined(__i386__) 244/* XXX: what about amd64/linux32? */ 245 246int 247linux_uselib(struct thread *td, struct linux_uselib_args *args) 248{ 249 struct nameidata ni; 250 struct vnode *vp; 251 struct exec *a_out; 252 struct vattr attr; 253 vm_offset_t vmaddr; 254 unsigned long file_offset; 255 unsigned long bss_size; 256 char *library; 257 ssize_t aresid; 258 int error, locked, writecount; 259 260 LCONVPATHEXIST(td, args->library, &library); 261 262#ifdef DEBUG 263 if (ldebug(uselib)) 264 printf(ARGS(uselib, "%s"), library); 265#endif 266 267 a_out = NULL; 268 locked = 0; 269 vp = NULL; 270 271 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 272 UIO_SYSSPACE, library, td); 273 error = namei(&ni); 274 LFREEPATH(library); 275 if (error) 276 goto cleanup; 277 278 vp = ni.ni_vp; 279 NDFREE(&ni, NDF_ONLY_PNBUF); 280 281 /* 282 * From here on down, we have a locked vnode that must be unlocked. 283 * XXX: The code below largely duplicates exec_check_permissions(). 284 */ 285 locked = 1; 286 287 /* Writable? */ 288 error = VOP_GET_WRITECOUNT(vp, &writecount); 289 if (error != 0) 290 goto cleanup; 291 if (writecount != 0) { 292 error = ETXTBSY; 293 goto cleanup; 294 } 295 296 /* Executable? */ 297 error = VOP_GETATTR(vp, &attr, td->td_ucred); 298 if (error) 299 goto cleanup; 300 301 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 302 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 303 /* EACCESS is what exec(2) returns. */ 304 error = ENOEXEC; 305 goto cleanup; 306 } 307 308 /* Sensible size? */ 309 if (attr.va_size == 0) { 310 error = ENOEXEC; 311 goto cleanup; 312 } 313 314 /* Can we access it? */ 315 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 316 if (error) 317 goto cleanup; 318 319 /* 320 * XXX: This should use vn_open() so that it is properly authorized, 321 * and to reduce code redundancy all over the place here. 322 * XXX: Not really, it duplicates far more of exec_check_permissions() 323 * than vn_open(). 324 */ 325#ifdef MAC 326 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 327 if (error) 328 goto cleanup; 329#endif 330 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 331 if (error) 332 goto cleanup; 333 334 /* Pull in executable header into exec_map */ 335 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 336 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 337 if (error) 338 goto cleanup; 339 340 /* Is it a Linux binary ? */ 341 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 342 error = ENOEXEC; 343 goto cleanup; 344 } 345 346 /* 347 * While we are here, we should REALLY do some more checks 348 */ 349 350 /* Set file/virtual offset based on a.out variant. */ 351 switch ((int)(a_out->a_magic & 0xffff)) { 352 case 0413: /* ZMAGIC */ 353 file_offset = 1024; 354 break; 355 case 0314: /* QMAGIC */ 356 file_offset = 0; 357 break; 358 default: 359 error = ENOEXEC; 360 goto cleanup; 361 } 362 363 bss_size = round_page(a_out->a_bss); 364 365 /* Check various fields in header for validity/bounds. */ 366 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 367 error = ENOEXEC; 368 goto cleanup; 369 } 370 371 /* text + data can't exceed file size */ 372 if (a_out->a_data + a_out->a_text > attr.va_size) { 373 error = EFAULT; 374 goto cleanup; 375 } 376 377 /* 378 * text/data/bss must not exceed limits 379 * XXX - this is not complete. it should check current usage PLUS 380 * the resources needed by this library. 381 */ 382 PROC_LOCK(td->td_proc); 383 if (a_out->a_text > maxtsiz || 384 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) || 385 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 386 bss_size) != 0) { 387 PROC_UNLOCK(td->td_proc); 388 error = ENOMEM; 389 goto cleanup; 390 } 391 PROC_UNLOCK(td->td_proc); 392 393 /* 394 * Prevent more writers. 395 * XXX: Note that if any of the VM operations fail below we don't 396 * clear this flag. 397 */ 398 VOP_SET_TEXT(vp); 399 400 /* 401 * Lock no longer needed 402 */ 403 locked = 0; 404 VOP_UNLOCK(vp, 0); 405 406 /* 407 * Check if file_offset page aligned. Currently we cannot handle 408 * misalinged file offsets, and so we read in the entire image 409 * (what a waste). 410 */ 411 if (file_offset & PAGE_MASK) { 412#ifdef DEBUG 413 printf("uselib: Non page aligned binary %lu\n", file_offset); 414#endif 415 /* Map text+data read/write/execute */ 416 417 /* a_entry is the load address and is page aligned */ 418 vmaddr = trunc_page(a_out->a_entry); 419 420 /* get anon user mapping, read+write+execute */ 421 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 422 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 423 VM_PROT_ALL, VM_PROT_ALL, 0); 424 if (error) 425 goto cleanup; 426 427 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 428 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 429 td->td_ucred, NOCRED, &aresid, td); 430 if (error != 0) 431 goto cleanup; 432 if (aresid != 0) { 433 error = ENOEXEC; 434 goto cleanup; 435 } 436 } else { 437#ifdef DEBUG 438 printf("uselib: Page aligned binary %lu\n", file_offset); 439#endif 440 /* 441 * for QMAGIC, a_entry is 20 bytes beyond the load address 442 * to skip the executable header 443 */ 444 vmaddr = trunc_page(a_out->a_entry); 445 446 /* 447 * Map it all into the process's space as a single 448 * copy-on-write "data" segment. 449 */ 450 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 451 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 452 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 453 if (error) 454 goto cleanup; 455 } 456#ifdef DEBUG 457 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 458 ((long *)vmaddr)[1]); 459#endif 460 if (bss_size != 0) { 461 /* Calculate BSS start address */ 462 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 463 a_out->a_data; 464 465 /* allocate some 'anon' space */ 466 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 467 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 468 VM_PROT_ALL, 0); 469 if (error) 470 goto cleanup; 471 } 472 473cleanup: 474 /* Unlock vnode if needed */ 475 if (locked) 476 VOP_UNLOCK(vp, 0); 477 478 /* Release the temporary mapping. */ 479 if (a_out) 480 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 481 482 return (error); 483} 484 485#endif /* __i386__ */ 486 487int 488linux_select(struct thread *td, struct linux_select_args *args) 489{ 490 l_timeval ltv; 491 struct timeval tv0, tv1, utv, *tvp; 492 int error; 493 494#ifdef DEBUG 495 if (ldebug(select)) 496 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 497 (void *)args->readfds, (void *)args->writefds, 498 (void *)args->exceptfds, (void *)args->timeout); 499#endif 500 501 /* 502 * Store current time for computation of the amount of 503 * time left. 504 */ 505 if (args->timeout) { 506 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 507 goto select_out; 508 utv.tv_sec = ltv.tv_sec; 509 utv.tv_usec = ltv.tv_usec; 510#ifdef DEBUG 511 if (ldebug(select)) 512 printf(LMSG("incoming timeout (%jd/%ld)"), 513 (intmax_t)utv.tv_sec, utv.tv_usec); 514#endif 515 516 if (itimerfix(&utv)) { 517 /* 518 * The timeval was invalid. Convert it to something 519 * valid that will act as it does under Linux. 520 */ 521 utv.tv_sec += utv.tv_usec / 1000000; 522 utv.tv_usec %= 1000000; 523 if (utv.tv_usec < 0) { 524 utv.tv_sec -= 1; 525 utv.tv_usec += 1000000; 526 } 527 if (utv.tv_sec < 0) 528 timevalclear(&utv); 529 } 530 microtime(&tv0); 531 tvp = &utv; 532 } else 533 tvp = NULL; 534 535 error = kern_select(td, args->nfds, args->readfds, args->writefds, 536 args->exceptfds, tvp, LINUX_NFDBITS); 537 538#ifdef DEBUG 539 if (ldebug(select)) 540 printf(LMSG("real select returns %d"), error); 541#endif 542 if (error) 543 goto select_out; 544 545 if (args->timeout) { 546 if (td->td_retval[0]) { 547 /* 548 * Compute how much time was left of the timeout, 549 * by subtracting the current time and the time 550 * before we started the call, and subtracting 551 * that result from the user-supplied value. 552 */ 553 microtime(&tv1); 554 timevalsub(&tv1, &tv0); 555 timevalsub(&utv, &tv1); 556 if (utv.tv_sec < 0) 557 timevalclear(&utv); 558 } else 559 timevalclear(&utv); 560#ifdef DEBUG 561 if (ldebug(select)) 562 printf(LMSG("outgoing timeout (%jd/%ld)"), 563 (intmax_t)utv.tv_sec, utv.tv_usec); 564#endif 565 ltv.tv_sec = utv.tv_sec; 566 ltv.tv_usec = utv.tv_usec; 567 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 568 goto select_out; 569 } 570 571select_out: 572#ifdef DEBUG 573 if (ldebug(select)) 574 printf(LMSG("select_out -> %d"), error); 575#endif 576 return (error); 577} 578 579int 580linux_mremap(struct thread *td, struct linux_mremap_args *args) 581{ 582 struct munmap_args /* { 583 void *addr; 584 size_t len; 585 } */ bsd_args; 586 int error = 0; 587 588#ifdef DEBUG 589 if (ldebug(mremap)) 590 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 591 (void *)(uintptr_t)args->addr, 592 (unsigned long)args->old_len, 593 (unsigned long)args->new_len, 594 (unsigned long)args->flags); 595#endif 596 597 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 598 td->td_retval[0] = 0; 599 return (EINVAL); 600 } 601 602 /* 603 * Check for the page alignment. 604 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 605 */ 606 if (args->addr & PAGE_MASK) { 607 td->td_retval[0] = 0; 608 return (EINVAL); 609 } 610 611 args->new_len = round_page(args->new_len); 612 args->old_len = round_page(args->old_len); 613 614 if (args->new_len > args->old_len) { 615 td->td_retval[0] = 0; 616 return (ENOMEM); 617 } 618 619 if (args->new_len < args->old_len) { 620 bsd_args.addr = 621 (caddr_t)((uintptr_t)args->addr + args->new_len); 622 bsd_args.len = args->old_len - args->new_len; 623 error = sys_munmap(td, &bsd_args); 624 } 625 626 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 627 return (error); 628} 629 630#define LINUX_MS_ASYNC 0x0001 631#define LINUX_MS_INVALIDATE 0x0002 632#define LINUX_MS_SYNC 0x0004 633 634int 635linux_msync(struct thread *td, struct linux_msync_args *args) 636{ 637 struct msync_args bsd_args; 638 639 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 640 bsd_args.len = (uintptr_t)args->len; 641 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 642 643 return (sys_msync(td, &bsd_args)); 644} 645 646int 647linux_time(struct thread *td, struct linux_time_args *args) 648{ 649 struct timeval tv; 650 l_time_t tm; 651 int error; 652 653#ifdef DEBUG 654 if (ldebug(time)) 655 printf(ARGS(time, "*")); 656#endif 657 658 microtime(&tv); 659 tm = tv.tv_sec; 660 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 661 return (error); 662 td->td_retval[0] = tm; 663 return (0); 664} 665 666struct l_times_argv { 667 l_clock_t tms_utime; 668 l_clock_t tms_stime; 669 l_clock_t tms_cutime; 670 l_clock_t tms_cstime; 671}; 672 673 674/* 675 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 676 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 677 * auxiliary vector entry. 678 */ 679#define CLK_TCK 100 680 681#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 682#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 683 684#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 685 CONVNTCK(r) : CONVOTCK(r)) 686 687int 688linux_times(struct thread *td, struct linux_times_args *args) 689{ 690 struct timeval tv, utime, stime, cutime, cstime; 691 struct l_times_argv tms; 692 struct proc *p; 693 int error; 694 695#ifdef DEBUG 696 if (ldebug(times)) 697 printf(ARGS(times, "*")); 698#endif 699 700 if (args->buf != NULL) { 701 p = td->td_proc; 702 PROC_LOCK(p); 703 PROC_STATLOCK(p); 704 calcru(p, &utime, &stime); 705 PROC_STATUNLOCK(p); 706 calccru(p, &cutime, &cstime); 707 PROC_UNLOCK(p); 708 709 tms.tms_utime = CONVTCK(utime); 710 tms.tms_stime = CONVTCK(stime); 711 712 tms.tms_cutime = CONVTCK(cutime); 713 tms.tms_cstime = CONVTCK(cstime); 714 715 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 716 return (error); 717 } 718 719 microuptime(&tv); 720 td->td_retval[0] = (int)CONVTCK(tv); 721 return (0); 722} 723 724int 725linux_newuname(struct thread *td, struct linux_newuname_args *args) 726{ 727 struct l_new_utsname utsname; 728 char osname[LINUX_MAX_UTSNAME]; 729 char osrelease[LINUX_MAX_UTSNAME]; 730 char *p; 731 732#ifdef DEBUG 733 if (ldebug(newuname)) 734 printf(ARGS(newuname, "*")); 735#endif 736 737 linux_get_osname(td, osname); 738 linux_get_osrelease(td, osrelease); 739 740 bzero(&utsname, sizeof(utsname)); 741 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 742 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 743 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 744 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 745 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 746 for (p = utsname.version; *p != '\0'; ++p) 747 if (*p == '\n') { 748 *p = '\0'; 749 break; 750 } 751 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME); 752 753 return (copyout(&utsname, args->buf, sizeof(utsname))); 754} 755 756struct l_utimbuf { 757 l_time_t l_actime; 758 l_time_t l_modtime; 759}; 760 761int 762linux_utime(struct thread *td, struct linux_utime_args *args) 763{ 764 struct timeval tv[2], *tvp; 765 struct l_utimbuf lut; 766 char *fname; 767 int error; 768 769 LCONVPATHEXIST(td, args->fname, &fname); 770 771#ifdef DEBUG 772 if (ldebug(utime)) 773 printf(ARGS(utime, "%s, *"), fname); 774#endif 775 776 if (args->times) { 777 if ((error = copyin(args->times, &lut, sizeof lut))) { 778 LFREEPATH(fname); 779 return (error); 780 } 781 tv[0].tv_sec = lut.l_actime; 782 tv[0].tv_usec = 0; 783 tv[1].tv_sec = lut.l_modtime; 784 tv[1].tv_usec = 0; 785 tvp = tv; 786 } else 787 tvp = NULL; 788 789 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 790 LFREEPATH(fname); 791 return (error); 792} 793 794int 795linux_utimes(struct thread *td, struct linux_utimes_args *args) 796{ 797 l_timeval ltv[2]; 798 struct timeval tv[2], *tvp = NULL; 799 char *fname; 800 int error; 801 802 LCONVPATHEXIST(td, args->fname, &fname); 803 804#ifdef DEBUG 805 if (ldebug(utimes)) 806 printf(ARGS(utimes, "%s, *"), fname); 807#endif 808 809 if (args->tptr != NULL) { 810 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 811 LFREEPATH(fname); 812 return (error); 813 } 814 tv[0].tv_sec = ltv[0].tv_sec; 815 tv[0].tv_usec = ltv[0].tv_usec; 816 tv[1].tv_sec = ltv[1].tv_sec; 817 tv[1].tv_usec = ltv[1].tv_usec; 818 tvp = tv; 819 } 820 821 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 822 LFREEPATH(fname); 823 return (error); 824} 825 826int 827linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 828{ 829 l_timeval ltv[2]; 830 struct timeval tv[2], *tvp = NULL; 831 char *fname; 832 int error, dfd; 833 834 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 835 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 836 837#ifdef DEBUG 838 if (ldebug(futimesat)) 839 printf(ARGS(futimesat, "%s, *"), fname); 840#endif 841 842 if (args->utimes != NULL) { 843 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 844 LFREEPATH(fname); 845 return (error); 846 } 847 tv[0].tv_sec = ltv[0].tv_sec; 848 tv[0].tv_usec = ltv[0].tv_usec; 849 tv[1].tv_sec = ltv[1].tv_sec; 850 tv[1].tv_usec = ltv[1].tv_usec; 851 tvp = tv; 852 } 853 854 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 855 LFREEPATH(fname); 856 return (error); 857} 858 859int 860linux_common_wait(struct thread *td, int pid, int *status, 861 int options, struct rusage *ru) 862{ 863 int error, tmpstat; 864 865 error = kern_wait(td, pid, &tmpstat, options, ru); 866 if (error) 867 return (error); 868 869 if (status) { 870 tmpstat &= 0xffff; 871 if (WIFSIGNALED(tmpstat)) 872 tmpstat = (tmpstat & 0xffffff80) | 873 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 874 else if (WIFSTOPPED(tmpstat)) 875 tmpstat = (tmpstat & 0xffff00ff) | 876 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 877 else if (WIFCONTINUED(tmpstat)) 878 tmpstat = 0xffff; 879 error = copyout(&tmpstat, status, sizeof(int)); 880 } 881 882 return (error); 883} 884 885#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 886int 887linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 888{ 889 struct linux_wait4_args wait4_args; 890 891#ifdef DEBUG 892 if (ldebug(waitpid)) 893 printf(ARGS(waitpid, "%d, %p, %d"), 894 args->pid, (void *)args->status, args->options); 895#endif 896 897 wait4_args.pid = args->pid; 898 wait4_args.status = args->status; 899 wait4_args.options = args->options; 900 wait4_args.rusage = NULL; 901 902 return (linux_wait4(td, &wait4_args)); 903} 904#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 905 906int 907linux_wait4(struct thread *td, struct linux_wait4_args *args) 908{ 909 int error, options; 910 struct rusage ru, *rup; 911 912#ifdef DEBUG 913 if (ldebug(wait4)) 914 printf(ARGS(wait4, "%d, %p, %d, %p"), 915 args->pid, (void *)args->status, args->options, 916 (void *)args->rusage); 917#endif 918 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 919 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 920 return (EINVAL); 921 922 options = WEXITED; 923 linux_to_bsd_waitopts(args->options, &options); 924 925 if (args->rusage != NULL) 926 rup = &ru; 927 else 928 rup = NULL; 929 error = linux_common_wait(td, args->pid, args->status, options, rup); 930 if (error != 0) 931 return (error); 932 if (args->rusage != NULL) 933 error = linux_copyout_rusage(&ru, args->rusage); 934 return (error); 935} 936 937int 938linux_waitid(struct thread *td, struct linux_waitid_args *args) 939{ 940 int status, options, sig; 941 struct __wrusage wru; 942 siginfo_t siginfo; 943 l_siginfo_t lsi; 944 idtype_t idtype; 945 struct proc *p; 946 int error; 947 948 options = 0; 949 linux_to_bsd_waitopts(args->options, &options); 950 951 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 952 return (EINVAL); 953 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 954 return (EINVAL); 955 956 switch (args->idtype) { 957 case LINUX_P_ALL: 958 idtype = P_ALL; 959 break; 960 case LINUX_P_PID: 961 if (args->id <= 0) 962 return (EINVAL); 963 idtype = P_PID; 964 break; 965 case LINUX_P_PGID: 966 if (args->id <= 0) 967 return (EINVAL); 968 idtype = P_PGID; 969 break; 970 default: 971 return (EINVAL); 972 } 973 974 error = kern_wait6(td, idtype, args->id, &status, options, 975 &wru, &siginfo); 976 if (error != 0) 977 return (error); 978 if (args->rusage != NULL) { 979 error = linux_copyout_rusage(&wru.wru_children, 980 args->rusage); 981 if (error != 0) 982 return (error); 983 } 984 if (args->info != NULL) { 985 p = td->td_proc; 986 if (td->td_retval[0] == 0) 987 bzero(&lsi, sizeof(lsi)); 988 else { 989 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo); 990 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 991 } 992 error = copyout(&lsi, args->info, sizeof(lsi)); 993 } 994 td->td_retval[0] = 0; 995 996 return (error); 997} 998 999int 1000linux_mknod(struct thread *td, struct linux_mknod_args *args) 1001{ 1002 char *path; 1003 int error; 1004 1005 LCONVPATHCREAT(td, args->path, &path); 1006 1007#ifdef DEBUG 1008 if (ldebug(mknod)) 1009 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1010 (uintmax_t)args->dev); 1011#endif 1012 1013 switch (args->mode & S_IFMT) { 1014 case S_IFIFO: 1015 case S_IFSOCK: 1016 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1017 break; 1018 1019 case S_IFCHR: 1020 case S_IFBLK: 1021 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1022 args->dev); 1023 break; 1024 1025 case S_IFDIR: 1026 error = EPERM; 1027 break; 1028 1029 case 0: 1030 args->mode |= S_IFREG; 1031 /* FALLTHROUGH */ 1032 case S_IFREG: 1033 error = kern_open(td, path, UIO_SYSSPACE, 1034 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1035 if (error == 0) 1036 kern_close(td, td->td_retval[0]); 1037 break; 1038 1039 default: 1040 error = EINVAL; 1041 break; 1042 } 1043 LFREEPATH(path); 1044 return (error); 1045} 1046 1047int 1048linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1049{ 1050 char *path; 1051 int error, dfd; 1052 1053 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1054 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1055 1056#ifdef DEBUG 1057 if (ldebug(mknodat)) 1058 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1059#endif 1060 1061 switch (args->mode & S_IFMT) { 1062 case S_IFIFO: 1063 case S_IFSOCK: 1064 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1065 break; 1066 1067 case S_IFCHR: 1068 case S_IFBLK: 1069 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1070 args->dev); 1071 break; 1072 1073 case S_IFDIR: 1074 error = EPERM; 1075 break; 1076 1077 case 0: 1078 args->mode |= S_IFREG; 1079 /* FALLTHROUGH */ 1080 case S_IFREG: 1081 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1082 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1083 if (error == 0) 1084 kern_close(td, td->td_retval[0]); 1085 break; 1086 1087 default: 1088 error = EINVAL; 1089 break; 1090 } 1091 LFREEPATH(path); 1092 return (error); 1093} 1094 1095/* 1096 * UGH! This is just about the dumbest idea I've ever heard!! 1097 */ 1098int 1099linux_personality(struct thread *td, struct linux_personality_args *args) 1100{ 1101#ifdef DEBUG 1102 if (ldebug(personality)) 1103 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1104#endif 1105 if (args->per != 0) 1106 return (EINVAL); 1107 1108 /* Yes Jim, it's still a Linux... */ 1109 td->td_retval[0] = 0; 1110 return (0); 1111} 1112 1113struct l_itimerval { 1114 l_timeval it_interval; 1115 l_timeval it_value; 1116}; 1117 1118#define B2L_ITIMERVAL(bip, lip) \ 1119 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1120 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1121 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1122 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1123 1124int 1125linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1126{ 1127 int error; 1128 struct l_itimerval ls; 1129 struct itimerval aitv, oitv; 1130 1131#ifdef DEBUG 1132 if (ldebug(setitimer)) 1133 printf(ARGS(setitimer, "%p, %p"), 1134 (void *)uap->itv, (void *)uap->oitv); 1135#endif 1136 1137 if (uap->itv == NULL) { 1138 uap->itv = uap->oitv; 1139 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1140 } 1141 1142 error = copyin(uap->itv, &ls, sizeof(ls)); 1143 if (error != 0) 1144 return (error); 1145 B2L_ITIMERVAL(&aitv, &ls); 1146#ifdef DEBUG 1147 if (ldebug(setitimer)) { 1148 printf("setitimer: value: sec: %jd, usec: %ld\n", 1149 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1150 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1151 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1152 } 1153#endif 1154 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1155 if (error != 0 || uap->oitv == NULL) 1156 return (error); 1157 B2L_ITIMERVAL(&ls, &oitv); 1158 1159 return (copyout(&ls, uap->oitv, sizeof(ls))); 1160} 1161 1162int 1163linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1164{ 1165 int error; 1166 struct l_itimerval ls; 1167 struct itimerval aitv; 1168 1169#ifdef DEBUG 1170 if (ldebug(getitimer)) 1171 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1172#endif 1173 error = kern_getitimer(td, uap->which, &aitv); 1174 if (error != 0) 1175 return (error); 1176 B2L_ITIMERVAL(&ls, &aitv); 1177 return (copyout(&ls, uap->itv, sizeof(ls))); 1178} 1179 1180#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1181int 1182linux_nice(struct thread *td, struct linux_nice_args *args) 1183{ 1184 struct setpriority_args bsd_args; 1185 1186 bsd_args.which = PRIO_PROCESS; 1187 bsd_args.who = 0; /* current process */ 1188 bsd_args.prio = args->inc; 1189 return (sys_setpriority(td, &bsd_args)); 1190} 1191#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1192 1193int 1194linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1195{ 1196 struct ucred *newcred, *oldcred; 1197 l_gid_t *linux_gidset; 1198 gid_t *bsd_gidset; 1199 int ngrp, error; 1200 struct proc *p; 1201 1202 ngrp = args->gidsetsize; 1203 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1204 return (EINVAL); 1205 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1206 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1207 if (error) 1208 goto out; 1209 newcred = crget(); 1210 p = td->td_proc; 1211 PROC_LOCK(p); 1212 oldcred = crcopysafe(p, newcred); 1213 1214 /* 1215 * cr_groups[0] holds egid. Setting the whole set from 1216 * the supplied set will cause egid to be changed too. 1217 * Keep cr_groups[0] unchanged to prevent that. 1218 */ 1219 1220 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1221 PROC_UNLOCK(p); 1222 crfree(newcred); 1223 goto out; 1224 } 1225 1226 if (ngrp > 0) { 1227 newcred->cr_ngroups = ngrp + 1; 1228 1229 bsd_gidset = newcred->cr_groups; 1230 ngrp--; 1231 while (ngrp >= 0) { 1232 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1233 ngrp--; 1234 } 1235 } else 1236 newcred->cr_ngroups = 1; 1237 1238 setsugid(p); 1239 p->p_ucred = newcred; 1240 PROC_UNLOCK(p); 1241 crfree(oldcred); 1242 error = 0; 1243out: 1244 free(linux_gidset, M_LINUX); 1245 return (error); 1246} 1247 1248int 1249linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1250{ 1251 struct ucred *cred; 1252 l_gid_t *linux_gidset; 1253 gid_t *bsd_gidset; 1254 int bsd_gidsetsz, ngrp, error; 1255 1256 cred = td->td_ucred; 1257 bsd_gidset = cred->cr_groups; 1258 bsd_gidsetsz = cred->cr_ngroups - 1; 1259 1260 /* 1261 * cr_groups[0] holds egid. Returning the whole set 1262 * here will cause a duplicate. Exclude cr_groups[0] 1263 * to prevent that. 1264 */ 1265 1266 if ((ngrp = args->gidsetsize) == 0) { 1267 td->td_retval[0] = bsd_gidsetsz; 1268 return (0); 1269 } 1270 1271 if (ngrp < bsd_gidsetsz) 1272 return (EINVAL); 1273 1274 ngrp = 0; 1275 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1276 M_LINUX, M_WAITOK); 1277 while (ngrp < bsd_gidsetsz) { 1278 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1279 ngrp++; 1280 } 1281 1282 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1283 free(linux_gidset, M_LINUX); 1284 if (error) 1285 return (error); 1286 1287 td->td_retval[0] = ngrp; 1288 return (0); 1289} 1290 1291int 1292linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1293{ 1294 struct rlimit bsd_rlim; 1295 struct l_rlimit rlim; 1296 u_int which; 1297 int error; 1298 1299#ifdef DEBUG 1300 if (ldebug(setrlimit)) 1301 printf(ARGS(setrlimit, "%d, %p"), 1302 args->resource, (void *)args->rlim); 1303#endif 1304 1305 if (args->resource >= LINUX_RLIM_NLIMITS) 1306 return (EINVAL); 1307 1308 which = linux_to_bsd_resource[args->resource]; 1309 if (which == -1) 1310 return (EINVAL); 1311 1312 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1313 if (error) 1314 return (error); 1315 1316 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1317 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1318 return (kern_setrlimit(td, which, &bsd_rlim)); 1319} 1320 1321#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1322int 1323linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1324{ 1325 struct l_rlimit rlim; 1326 struct proc *p = td->td_proc; 1327 struct rlimit bsd_rlim; 1328 u_int which; 1329 1330#ifdef DEBUG 1331 if (ldebug(old_getrlimit)) 1332 printf(ARGS(old_getrlimit, "%d, %p"), 1333 args->resource, (void *)args->rlim); 1334#endif 1335 1336 if (args->resource >= LINUX_RLIM_NLIMITS) 1337 return (EINVAL); 1338 1339 which = linux_to_bsd_resource[args->resource]; 1340 if (which == -1) 1341 return (EINVAL); 1342 1343 PROC_LOCK(p); 1344 lim_rlimit(p, which, &bsd_rlim); 1345 PROC_UNLOCK(p); 1346 1347#ifdef COMPAT_LINUX32 1348 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1349 if (rlim.rlim_cur == UINT_MAX) 1350 rlim.rlim_cur = INT_MAX; 1351 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1352 if (rlim.rlim_max == UINT_MAX) 1353 rlim.rlim_max = INT_MAX; 1354#else 1355 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1356 if (rlim.rlim_cur == ULONG_MAX) 1357 rlim.rlim_cur = LONG_MAX; 1358 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1359 if (rlim.rlim_max == ULONG_MAX) 1360 rlim.rlim_max = LONG_MAX; 1361#endif 1362 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1363} 1364#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1365 1366int 1367linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1368{ 1369 struct l_rlimit rlim; 1370 struct proc *p = td->td_proc; 1371 struct rlimit bsd_rlim; 1372 u_int which; 1373 1374#ifdef DEBUG 1375 if (ldebug(getrlimit)) 1376 printf(ARGS(getrlimit, "%d, %p"), 1377 args->resource, (void *)args->rlim); 1378#endif 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 PROC_LOCK(p); 1388 lim_rlimit(p, which, &bsd_rlim); 1389 PROC_UNLOCK(p); 1390 1391 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1392 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1393 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1394} 1395 1396int 1397linux_sched_setscheduler(struct thread *td, 1398 struct linux_sched_setscheduler_args *args) 1399{ 1400 struct sched_param sched_param; 1401 struct thread *tdt; 1402 int error, policy; 1403 1404#ifdef DEBUG 1405 if (ldebug(sched_setscheduler)) 1406 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1407 args->pid, args->policy, (const void *)args->param); 1408#endif 1409 1410 switch (args->policy) { 1411 case LINUX_SCHED_OTHER: 1412 policy = SCHED_OTHER; 1413 break; 1414 case LINUX_SCHED_FIFO: 1415 policy = SCHED_FIFO; 1416 break; 1417 case LINUX_SCHED_RR: 1418 policy = SCHED_RR; 1419 break; 1420 default: 1421 return (EINVAL); 1422 } 1423 1424 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1425 if (error) 1426 return (error); 1427 1428 tdt = linux_tdfind(td, args->pid, -1); 1429 if (tdt == NULL) 1430 return (ESRCH); 1431 1432 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1433 PROC_UNLOCK(tdt->td_proc); 1434 return (error); 1435} 1436 1437int 1438linux_sched_getscheduler(struct thread *td, 1439 struct linux_sched_getscheduler_args *args) 1440{ 1441 struct thread *tdt; 1442 int error, policy; 1443 1444#ifdef DEBUG 1445 if (ldebug(sched_getscheduler)) 1446 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1447#endif 1448 1449 tdt = linux_tdfind(td, args->pid, -1); 1450 if (tdt == NULL) 1451 return (ESRCH); 1452 1453 error = kern_sched_getscheduler(td, tdt, &policy); 1454 PROC_UNLOCK(tdt->td_proc); 1455 1456 switch (policy) { 1457 case SCHED_OTHER: 1458 td->td_retval[0] = LINUX_SCHED_OTHER; 1459 break; 1460 case SCHED_FIFO: 1461 td->td_retval[0] = LINUX_SCHED_FIFO; 1462 break; 1463 case SCHED_RR: 1464 td->td_retval[0] = LINUX_SCHED_RR; 1465 break; 1466 } 1467 return (error); 1468} 1469 1470int 1471linux_sched_get_priority_max(struct thread *td, 1472 struct linux_sched_get_priority_max_args *args) 1473{ 1474 struct sched_get_priority_max_args bsd; 1475 1476#ifdef DEBUG 1477 if (ldebug(sched_get_priority_max)) 1478 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1479#endif 1480 1481 switch (args->policy) { 1482 case LINUX_SCHED_OTHER: 1483 bsd.policy = SCHED_OTHER; 1484 break; 1485 case LINUX_SCHED_FIFO: 1486 bsd.policy = SCHED_FIFO; 1487 break; 1488 case LINUX_SCHED_RR: 1489 bsd.policy = SCHED_RR; 1490 break; 1491 default: 1492 return (EINVAL); 1493 } 1494 return (sys_sched_get_priority_max(td, &bsd)); 1495} 1496 1497int 1498linux_sched_get_priority_min(struct thread *td, 1499 struct linux_sched_get_priority_min_args *args) 1500{ 1501 struct sched_get_priority_min_args bsd; 1502 1503#ifdef DEBUG 1504 if (ldebug(sched_get_priority_min)) 1505 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1506#endif 1507 1508 switch (args->policy) { 1509 case LINUX_SCHED_OTHER: 1510 bsd.policy = SCHED_OTHER; 1511 break; 1512 case LINUX_SCHED_FIFO: 1513 bsd.policy = SCHED_FIFO; 1514 break; 1515 case LINUX_SCHED_RR: 1516 bsd.policy = SCHED_RR; 1517 break; 1518 default: 1519 return (EINVAL); 1520 } 1521 return (sys_sched_get_priority_min(td, &bsd)); 1522} 1523 1524#define REBOOT_CAD_ON 0x89abcdef 1525#define REBOOT_CAD_OFF 0 1526#define REBOOT_HALT 0xcdef0123 1527#define REBOOT_RESTART 0x01234567 1528#define REBOOT_RESTART2 0xA1B2C3D4 1529#define REBOOT_POWEROFF 0x4321FEDC 1530#define REBOOT_MAGIC1 0xfee1dead 1531#define REBOOT_MAGIC2 0x28121969 1532#define REBOOT_MAGIC2A 0x05121996 1533#define REBOOT_MAGIC2B 0x16041998 1534 1535int 1536linux_reboot(struct thread *td, struct linux_reboot_args *args) 1537{ 1538 struct reboot_args bsd_args; 1539 1540#ifdef DEBUG 1541 if (ldebug(reboot)) 1542 printf(ARGS(reboot, "0x%x"), args->cmd); 1543#endif 1544 1545 if (args->magic1 != REBOOT_MAGIC1) 1546 return (EINVAL); 1547 1548 switch (args->magic2) { 1549 case REBOOT_MAGIC2: 1550 case REBOOT_MAGIC2A: 1551 case REBOOT_MAGIC2B: 1552 break; 1553 default: 1554 return (EINVAL); 1555 } 1556 1557 switch (args->cmd) { 1558 case REBOOT_CAD_ON: 1559 case REBOOT_CAD_OFF: 1560 return (priv_check(td, PRIV_REBOOT)); 1561 case REBOOT_HALT: 1562 bsd_args.opt = RB_HALT; 1563 break; 1564 case REBOOT_RESTART: 1565 case REBOOT_RESTART2: 1566 bsd_args.opt = 0; 1567 break; 1568 case REBOOT_POWEROFF: 1569 bsd_args.opt = RB_POWEROFF; 1570 break; 1571 default: 1572 return (EINVAL); 1573 } 1574 return (sys_reboot(td, &bsd_args)); 1575} 1576 1577 1578/* 1579 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1580 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1581 * are assumed to be preserved. The following lightweight syscalls fixes 1582 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1583 * 1584 * linux_getpid() - MP SAFE 1585 * linux_getgid() - MP SAFE 1586 * linux_getuid() - MP SAFE 1587 */ 1588 1589int 1590linux_getpid(struct thread *td, struct linux_getpid_args *args) 1591{ 1592 1593#ifdef DEBUG 1594 if (ldebug(getpid)) 1595 printf(ARGS(getpid, "")); 1596#endif 1597 td->td_retval[0] = td->td_proc->p_pid; 1598 1599 return (0); 1600} 1601 1602int 1603linux_gettid(struct thread *td, struct linux_gettid_args *args) 1604{ 1605 struct linux_emuldata *em; 1606 1607#ifdef DEBUG 1608 if (ldebug(gettid)) 1609 printf(ARGS(gettid, "")); 1610#endif 1611 1612 em = em_find(td); 1613 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1614 1615 td->td_retval[0] = em->em_tid; 1616 1617 return (0); 1618} 1619 1620 1621int 1622linux_getppid(struct thread *td, struct linux_getppid_args *args) 1623{ 1624 1625#ifdef DEBUG 1626 if (ldebug(getppid)) 1627 printf(ARGS(getppid, "")); 1628#endif 1629 1630 PROC_LOCK(td->td_proc); 1631 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1632 PROC_UNLOCK(td->td_proc); 1633 return (0); 1634} 1635 1636int 1637linux_getgid(struct thread *td, struct linux_getgid_args *args) 1638{ 1639 1640#ifdef DEBUG 1641 if (ldebug(getgid)) 1642 printf(ARGS(getgid, "")); 1643#endif 1644 1645 td->td_retval[0] = td->td_ucred->cr_rgid; 1646 return (0); 1647} 1648 1649int 1650linux_getuid(struct thread *td, struct linux_getuid_args *args) 1651{ 1652 1653#ifdef DEBUG 1654 if (ldebug(getuid)) 1655 printf(ARGS(getuid, "")); 1656#endif 1657 1658 td->td_retval[0] = td->td_ucred->cr_ruid; 1659 return (0); 1660} 1661 1662 1663int 1664linux_getsid(struct thread *td, struct linux_getsid_args *args) 1665{ 1666 struct getsid_args bsd; 1667 1668#ifdef DEBUG 1669 if (ldebug(getsid)) 1670 printf(ARGS(getsid, "%i"), args->pid); 1671#endif 1672 1673 bsd.pid = args->pid; 1674 return (sys_getsid(td, &bsd)); 1675} 1676 1677int 1678linux_nosys(struct thread *td, struct nosys_args *ignore) 1679{ 1680 1681 return (ENOSYS); 1682} 1683 1684int 1685linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1686{ 1687 struct getpriority_args bsd_args; 1688 int error; 1689 1690#ifdef DEBUG 1691 if (ldebug(getpriority)) 1692 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1693#endif 1694 1695 bsd_args.which = args->which; 1696 bsd_args.who = args->who; 1697 error = sys_getpriority(td, &bsd_args); 1698 td->td_retval[0] = 20 - td->td_retval[0]; 1699 return (error); 1700} 1701 1702int 1703linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1704{ 1705 int name[2]; 1706 1707#ifdef DEBUG 1708 if (ldebug(sethostname)) 1709 printf(ARGS(sethostname, "*, %i"), args->len); 1710#endif 1711 1712 name[0] = CTL_KERN; 1713 name[1] = KERN_HOSTNAME; 1714 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1715 args->len, 0, 0)); 1716} 1717 1718int 1719linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1720{ 1721 int name[2]; 1722 1723#ifdef DEBUG 1724 if (ldebug(setdomainname)) 1725 printf(ARGS(setdomainname, "*, %i"), args->len); 1726#endif 1727 1728 name[0] = CTL_KERN; 1729 name[1] = KERN_NISDOMAINNAME; 1730 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1731 args->len, 0, 0)); 1732} 1733 1734int 1735linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1736{ 1737 1738#ifdef DEBUG 1739 if (ldebug(exit_group)) 1740 printf(ARGS(exit_group, "%i"), args->error_code); 1741#endif 1742 1743 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1744 args->error_code); 1745 1746 /* 1747 * XXX: we should send a signal to the parent if 1748 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1749 * as it doesnt occur often. 1750 */ 1751 exit1(td, W_EXITCODE(args->error_code, 0)); 1752 /* NOTREACHED */ 1753} 1754 1755#define _LINUX_CAPABILITY_VERSION 0x19980330 1756 1757struct l_user_cap_header { 1758 l_int version; 1759 l_int pid; 1760}; 1761 1762struct l_user_cap_data { 1763 l_int effective; 1764 l_int permitted; 1765 l_int inheritable; 1766}; 1767 1768int 1769linux_capget(struct thread *td, struct linux_capget_args *args) 1770{ 1771 struct l_user_cap_header luch; 1772 struct l_user_cap_data lucd; 1773 int error; 1774 1775 if (args->hdrp == NULL) 1776 return (EFAULT); 1777 1778 error = copyin(args->hdrp, &luch, sizeof(luch)); 1779 if (error != 0) 1780 return (error); 1781 1782 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1783 luch.version = _LINUX_CAPABILITY_VERSION; 1784 error = copyout(&luch, args->hdrp, sizeof(luch)); 1785 if (error) 1786 return (error); 1787 return (EINVAL); 1788 } 1789 1790 if (luch.pid) 1791 return (EPERM); 1792 1793 if (args->datap) { 1794 /* 1795 * The current implementation doesn't support setting 1796 * a capability (it's essentially a stub) so indicate 1797 * that no capabilities are currently set or available 1798 * to request. 1799 */ 1800 bzero (&lucd, sizeof(lucd)); 1801 error = copyout(&lucd, args->datap, sizeof(lucd)); 1802 } 1803 1804 return (error); 1805} 1806 1807int 1808linux_capset(struct thread *td, struct linux_capset_args *args) 1809{ 1810 struct l_user_cap_header luch; 1811 struct l_user_cap_data lucd; 1812 int error; 1813 1814 if (args->hdrp == NULL || args->datap == NULL) 1815 return (EFAULT); 1816 1817 error = copyin(args->hdrp, &luch, sizeof(luch)); 1818 if (error != 0) 1819 return (error); 1820 1821 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1822 luch.version = _LINUX_CAPABILITY_VERSION; 1823 error = copyout(&luch, args->hdrp, sizeof(luch)); 1824 if (error) 1825 return (error); 1826 return (EINVAL); 1827 } 1828 1829 if (luch.pid) 1830 return (EPERM); 1831 1832 error = copyin(args->datap, &lucd, sizeof(lucd)); 1833 if (error != 0) 1834 return (error); 1835 1836 /* We currently don't support setting any capabilities. */ 1837 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1838 linux_msg(td, 1839 "capset effective=0x%x, permitted=0x%x, " 1840 "inheritable=0x%x is not implemented", 1841 (int)lucd.effective, (int)lucd.permitted, 1842 (int)lucd.inheritable); 1843 return (EPERM); 1844 } 1845 1846 return (0); 1847} 1848 1849int 1850linux_prctl(struct thread *td, struct linux_prctl_args *args) 1851{ 1852 int error = 0, max_size; 1853 struct proc *p = td->td_proc; 1854 char comm[LINUX_MAX_COMM_LEN]; 1855 struct linux_emuldata *em; 1856 int pdeath_signal; 1857 1858#ifdef DEBUG 1859 if (ldebug(prctl)) 1860 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1861 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1862 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1863#endif 1864 1865 switch (args->option) { 1866 case LINUX_PR_SET_PDEATHSIG: 1867 if (!LINUX_SIG_VALID(args->arg2)) 1868 return (EINVAL); 1869 em = em_find(td); 1870 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1871 em->pdeath_signal = args->arg2; 1872 break; 1873 case LINUX_PR_GET_PDEATHSIG: 1874 em = em_find(td); 1875 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1876 pdeath_signal = em->pdeath_signal; 1877 error = copyout(&pdeath_signal, 1878 (void *)(register_t)args->arg2, 1879 sizeof(pdeath_signal)); 1880 break; 1881 case LINUX_PR_GET_KEEPCAPS: 1882 /* 1883 * Indicate that we always clear the effective and 1884 * permitted capability sets when the user id becomes 1885 * non-zero (actually the capability sets are simply 1886 * always zero in the current implementation). 1887 */ 1888 td->td_retval[0] = 0; 1889 break; 1890 case LINUX_PR_SET_KEEPCAPS: 1891 /* 1892 * Ignore requests to keep the effective and permitted 1893 * capability sets when the user id becomes non-zero. 1894 */ 1895 break; 1896 case LINUX_PR_SET_NAME: 1897 /* 1898 * To be on the safe side we need to make sure to not 1899 * overflow the size a linux program expects. We already 1900 * do this here in the copyin, so that we don't need to 1901 * check on copyout. 1902 */ 1903 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1904 error = copyinstr((void *)(register_t)args->arg2, comm, 1905 max_size, NULL); 1906 1907 /* Linux silently truncates the name if it is too long. */ 1908 if (error == ENAMETOOLONG) { 1909 /* 1910 * XXX: copyinstr() isn't documented to populate the 1911 * array completely, so do a copyin() to be on the 1912 * safe side. This should be changed in case 1913 * copyinstr() is changed to guarantee this. 1914 */ 1915 error = copyin((void *)(register_t)args->arg2, comm, 1916 max_size - 1); 1917 comm[max_size - 1] = '\0'; 1918 } 1919 if (error) 1920 return (error); 1921 1922 PROC_LOCK(p); 1923 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1924 PROC_UNLOCK(p); 1925 break; 1926 case LINUX_PR_GET_NAME: 1927 PROC_LOCK(p); 1928 strlcpy(comm, p->p_comm, sizeof(comm)); 1929 PROC_UNLOCK(p); 1930 error = copyout(comm, (void *)(register_t)args->arg2, 1931 strlen(comm) + 1); 1932 break; 1933 default: 1934 error = EINVAL; 1935 break; 1936 } 1937 1938 return (error); 1939} 1940 1941int 1942linux_sched_setparam(struct thread *td, 1943 struct linux_sched_setparam_args *uap) 1944{ 1945 struct sched_param sched_param; 1946 struct thread *tdt; 1947 int error; 1948 1949#ifdef DEBUG 1950 if (ldebug(sched_setparam)) 1951 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 1952#endif 1953 1954 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1955 if (error) 1956 return (error); 1957 1958 tdt = linux_tdfind(td, uap->pid, -1); 1959 if (tdt == NULL) 1960 return (ESRCH); 1961 1962 error = kern_sched_setparam(td, tdt, &sched_param); 1963 PROC_UNLOCK(tdt->td_proc); 1964 return (error); 1965} 1966 1967int 1968linux_sched_getparam(struct thread *td, 1969 struct linux_sched_getparam_args *uap) 1970{ 1971 struct sched_param sched_param; 1972 struct thread *tdt; 1973 int error; 1974 1975#ifdef DEBUG 1976 if (ldebug(sched_getparam)) 1977 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 1978#endif 1979 1980 tdt = linux_tdfind(td, uap->pid, -1); 1981 if (tdt == NULL) 1982 return (ESRCH); 1983 1984 error = kern_sched_getparam(td, tdt, &sched_param); 1985 PROC_UNLOCK(tdt->td_proc); 1986 if (error == 0) 1987 error = copyout(&sched_param, uap->param, 1988 sizeof(sched_param)); 1989 return (error); 1990} 1991 1992/* 1993 * Get affinity of a process. 1994 */ 1995int 1996linux_sched_getaffinity(struct thread *td, 1997 struct linux_sched_getaffinity_args *args) 1998{ 1999 int error; 2000 struct thread *tdt; 2001 struct cpuset_getaffinity_args cga; 2002 2003#ifdef DEBUG 2004 if (ldebug(sched_getaffinity)) 2005 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2006 args->len); 2007#endif 2008 if (args->len < sizeof(cpuset_t)) 2009 return (EINVAL); 2010 2011 tdt = linux_tdfind(td, args->pid, -1); 2012 if (tdt == NULL) 2013 return (ESRCH); 2014 2015 PROC_UNLOCK(tdt->td_proc); 2016 cga.level = CPU_LEVEL_WHICH; 2017 cga.which = CPU_WHICH_TID; 2018 cga.id = tdt->td_tid; 2019 cga.cpusetsize = sizeof(cpuset_t); 2020 cga.mask = (cpuset_t *) args->user_mask_ptr; 2021 2022 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2023 td->td_retval[0] = sizeof(cpuset_t); 2024 2025 return (error); 2026} 2027 2028/* 2029 * Set affinity of a process. 2030 */ 2031int 2032linux_sched_setaffinity(struct thread *td, 2033 struct linux_sched_setaffinity_args *args) 2034{ 2035 struct cpuset_setaffinity_args csa; 2036 struct thread *tdt; 2037 2038#ifdef DEBUG 2039 if (ldebug(sched_setaffinity)) 2040 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2041 args->len); 2042#endif 2043 if (args->len < sizeof(cpuset_t)) 2044 return (EINVAL); 2045 2046 tdt = linux_tdfind(td, args->pid, -1); 2047 if (tdt == NULL) 2048 return (ESRCH); 2049 2050 PROC_UNLOCK(tdt->td_proc); 2051 csa.level = CPU_LEVEL_WHICH; 2052 csa.which = CPU_WHICH_TID; 2053 csa.id = tdt->td_tid; 2054 csa.cpusetsize = sizeof(cpuset_t); 2055 csa.mask = (cpuset_t *) args->user_mask_ptr; 2056 2057 return (sys_cpuset_setaffinity(td, &csa)); 2058} 2059 2060struct linux_rlimit64 { 2061 uint64_t rlim_cur; 2062 uint64_t rlim_max; 2063}; 2064 2065int 2066linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2067{ 2068 struct rlimit rlim, nrlim; 2069 struct linux_rlimit64 lrlim; 2070 struct proc *p; 2071 u_int which; 2072 int flags; 2073 int error; 2074 2075#ifdef DEBUG 2076 if (ldebug(prlimit64)) 2077 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2078 args->resource, (void *)args->new, (void *)args->old); 2079#endif 2080 2081 if (args->resource >= LINUX_RLIM_NLIMITS) 2082 return (EINVAL); 2083 2084 which = linux_to_bsd_resource[args->resource]; 2085 if (which == -1) 2086 return (EINVAL); 2087 2088 if (args->new != NULL) { 2089 /* 2090 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2091 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2092 * as INFINITY so we do not need a conversion even. 2093 */ 2094 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2095 if (error != 0) 2096 return (error); 2097 } 2098 2099 flags = PGET_HOLD | PGET_NOTWEXIT; 2100 if (args->new != NULL) 2101 flags |= PGET_CANDEBUG; 2102 else 2103 flags |= PGET_CANSEE; 2104 error = pget(args->pid, flags, &p); 2105 if (error != 0) 2106 return (error); 2107 2108 if (args->old != NULL) { 2109 PROC_LOCK(p); 2110 lim_rlimit(p, which, &rlim); 2111 PROC_UNLOCK(p); 2112 if (rlim.rlim_cur == RLIM_INFINITY) 2113 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2114 else 2115 lrlim.rlim_cur = rlim.rlim_cur; 2116 if (rlim.rlim_max == RLIM_INFINITY) 2117 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2118 else 2119 lrlim.rlim_max = rlim.rlim_max; 2120 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2121 if (error != 0) 2122 goto out; 2123 } 2124 2125 if (args->new != NULL) 2126 error = kern_proc_setrlimit(td, p, which, &nrlim); 2127 2128 out: 2129 PRELE(p); 2130 return (error); 2131} 2132 2133int 2134linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2135{ 2136 struct timeval utv, tv0, tv1, *tvp; 2137 struct l_pselect6arg lpse6; 2138 struct l_timespec lts; 2139 struct timespec uts; 2140 l_sigset_t l_ss; 2141 sigset_t *ssp; 2142 sigset_t ss; 2143 int error; 2144 2145 ssp = NULL; 2146 if (args->sig != NULL) { 2147 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2148 if (error != 0) 2149 return (error); 2150 if (lpse6.ss_len != sizeof(l_ss)) 2151 return (EINVAL); 2152 if (lpse6.ss != 0) { 2153 error = copyin(PTRIN(lpse6.ss), &l_ss, 2154 sizeof(l_ss)); 2155 if (error != 0) 2156 return (error); 2157 linux_to_bsd_sigset(&l_ss, &ss); 2158 ssp = &ss; 2159 } 2160 } 2161 2162 /* 2163 * Currently glibc changes nanosecond number to microsecond. 2164 * This mean losing precision but for now it is hardly seen. 2165 */ 2166 if (args->tsp != NULL) { 2167 error = copyin(args->tsp, <s, sizeof(lts)); 2168 if (error != 0) 2169 return (error); 2170 error = linux_to_native_timespec(&uts, <s); 2171 if (error != 0) 2172 return (error); 2173 2174 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2175 if (itimerfix(&utv)) 2176 return (EINVAL); 2177 2178 microtime(&tv0); 2179 tvp = &utv; 2180 } else 2181 tvp = NULL; 2182 2183 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2184 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2185 2186 if (error == 0 && args->tsp != NULL) { 2187 if (td->td_retval[0] != 0) { 2188 /* 2189 * Compute how much time was left of the timeout, 2190 * by subtracting the current time and the time 2191 * before we started the call, and subtracting 2192 * that result from the user-supplied value. 2193 */ 2194 2195 microtime(&tv1); 2196 timevalsub(&tv1, &tv0); 2197 timevalsub(&utv, &tv1); 2198 if (utv.tv_sec < 0) 2199 timevalclear(&utv); 2200 } else 2201 timevalclear(&utv); 2202 2203 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2204 2205 native_to_linux_timespec(<s, &uts); 2206 error = copyout(<s, args->tsp, sizeof(lts)); 2207 } 2208 2209 return (error); 2210} 2211 2212int 2213linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2214{ 2215 struct timespec ts0, ts1; 2216 struct l_timespec lts; 2217 struct timespec uts, *tsp; 2218 l_sigset_t l_ss; 2219 sigset_t *ssp; 2220 sigset_t ss; 2221 int error; 2222 2223 if (args->sset != NULL) { 2224 if (args->ssize != sizeof(l_ss)) 2225 return (EINVAL); 2226 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2227 if (error) 2228 return (error); 2229 linux_to_bsd_sigset(&l_ss, &ss); 2230 ssp = &ss; 2231 } else 2232 ssp = NULL; 2233 if (args->tsp != NULL) { 2234 error = copyin(args->tsp, <s, sizeof(lts)); 2235 if (error) 2236 return (error); 2237 error = linux_to_native_timespec(&uts, <s); 2238 if (error != 0) 2239 return (error); 2240 2241 nanotime(&ts0); 2242 tsp = &uts; 2243 } else 2244 tsp = NULL; 2245 2246 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2247 2248 if (error == 0 && args->tsp != NULL) { 2249 if (td->td_retval[0]) { 2250 nanotime(&ts1); 2251 timespecsub(&ts1, &ts0); 2252 timespecsub(&uts, &ts1); 2253 if (uts.tv_sec < 0) 2254 timespecclear(&uts); 2255 } else 2256 timespecclear(&uts); 2257 2258 native_to_linux_timespec(<s, &uts); 2259 error = copyout(<s, args->tsp, sizeof(lts)); 2260 } 2261 2262 return (error); 2263} 2264 2265#if defined(DEBUG) || defined(KTR) 2266/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2267 2268u_char linux_debug_map[howmany(LINUX_SYS_MAXSYSCALL, sizeof(u_char))]; 2269 2270static int 2271linux_debug(int syscall, int toggle, int global) 2272{ 2273 2274 if (global) { 2275 char c = toggle ? 0 : 0xff; 2276 2277 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2278 return (0); 2279 } 2280 if (syscall < 0 || syscall >= LINUX_SYS_MAXSYSCALL) 2281 return (EINVAL); 2282 if (toggle) 2283 clrbit(linux_debug_map, syscall); 2284 else 2285 setbit(linux_debug_map, syscall); 2286 return (0); 2287} 2288 2289/* 2290 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2291 * 2292 * E.g.: sysctl linux.debug=21.0 2293 * 2294 * As a special case, syscall "all" will apply to all syscalls globally. 2295 */ 2296#define LINUX_MAX_DEBUGSTR 16 2297int 2298linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2299{ 2300 char value[LINUX_MAX_DEBUGSTR], *p; 2301 int error, sysc, toggle; 2302 int global = 0; 2303 2304 value[0] = '\0'; 2305 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2306 if (error || req->newptr == NULL) 2307 return (error); 2308 for (p = value; *p != '\0' && *p != '.'; p++); 2309 if (*p == '\0') 2310 return (EINVAL); 2311 *p++ = '\0'; 2312 sysc = strtol(value, NULL, 0); 2313 toggle = strtol(p, NULL, 0); 2314 if (strcmp(value, "all") == 0) 2315 global = 1; 2316 error = linux_debug(sysc, toggle, global); 2317 return (error); 2318} 2319 2320#endif /* DEBUG || KTR */ 2321 2322int 2323linux_sched_rr_get_interval(struct thread *td, 2324 struct linux_sched_rr_get_interval_args *uap) 2325{ 2326 struct timespec ts; 2327 struct l_timespec lts; 2328 struct thread *tdt; 2329 int error; 2330 2331 /* 2332 * According to man in case the invalid pid specified 2333 * EINVAL should be returned. 2334 */ 2335 if (uap->pid < 0) 2336 return (EINVAL); 2337 2338 tdt = linux_tdfind(td, uap->pid, -1); 2339 if (tdt == NULL) 2340 return (ESRCH); 2341 2342 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2343 PROC_UNLOCK(tdt->td_proc); 2344 if (error != 0) 2345 return (error); 2346 native_to_linux_timespec(<s, &ts); 2347 return (copyout(<s, uap->interval, sizeof(lts))); 2348} 2349 2350/* 2351 * In case when the Linux thread is the initial thread in 2352 * the thread group thread id is equal to the process id. 2353 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2354 */ 2355struct thread * 2356linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2357{ 2358 struct linux_emuldata *em; 2359 struct thread *tdt; 2360 struct proc *p; 2361 2362 tdt = NULL; 2363 if (tid == 0 || tid == td->td_tid) { 2364 tdt = td; 2365 PROC_LOCK(tdt->td_proc); 2366 } else if (tid > PID_MAX) 2367 tdt = tdfind(tid, pid); 2368 else { 2369 /* 2370 * Initial thread where the tid equal to the pid. 2371 */ 2372 p = pfind(tid); 2373 if (p != NULL) { 2374 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2375 /* 2376 * p is not a Linuxulator process. 2377 */ 2378 PROC_UNLOCK(p); 2379 return (NULL); 2380 } 2381 FOREACH_THREAD_IN_PROC(p, tdt) { 2382 em = em_find(tdt); 2383 if (tid == em->em_tid) 2384 return (tdt); 2385 } 2386 PROC_UNLOCK(p); 2387 } 2388 return (NULL); 2389 } 2390 2391 return (tdt); 2392} 2393 2394void 2395linux_to_bsd_waitopts(int options, int *bsdopts) 2396{ 2397 2398 if (options & LINUX_WNOHANG) 2399 *bsdopts |= WNOHANG; 2400 if (options & LINUX_WUNTRACED) 2401 *bsdopts |= WUNTRACED; 2402 if (options & LINUX_WEXITED) 2403 *bsdopts |= WEXITED; 2404 if (options & LINUX_WCONTINUED) 2405 *bsdopts |= WCONTINUED; 2406 if (options & LINUX_WNOWAIT) 2407 *bsdopts |= WNOWAIT; 2408 2409 if (options & __WCLONE) 2410 *bsdopts |= WLINUXCLONE; 2411} 2412