linux_misc.c revision 293582
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 293582 2016-01-09 17:41:00Z 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_utimensat(struct thread *td, struct linux_utimensat_args *args) 828{ 829 struct l_timespec l_times[2]; 830 struct timespec times[2], *timesp = NULL; 831 char *path = NULL; 832 int error, dfd, flags = 0; 833 834 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 835 836#ifdef DEBUG 837 if (ldebug(utimensat)) 838 printf(ARGS(utimensat, "%d, *"), dfd); 839#endif 840 841 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW) 842 return (EINVAL); 843 844 if (args->times != NULL) { 845 error = copyin(args->times, l_times, sizeof(l_times)); 846 if (error != 0) 847 return (error); 848 849 if (l_times[0].tv_nsec > 999999999 || 850 l_times[1].tv_nsec > 999999999) 851 return (EINVAL); 852 853 times[0].tv_sec = l_times[0].tv_sec; 854 switch (l_times[0].tv_nsec) 855 { 856 case LINUX_UTIME_OMIT: 857 times[0].tv_nsec = UTIME_OMIT; 858 break; 859 case LINUX_UTIME_NOW: 860 times[0].tv_nsec = UTIME_NOW; 861 break; 862 default: 863 times[0].tv_nsec = l_times[0].tv_nsec; 864 } 865 866 times[1].tv_sec = l_times[1].tv_sec; 867 switch (l_times[1].tv_nsec) 868 { 869 case LINUX_UTIME_OMIT: 870 times[1].tv_nsec = UTIME_OMIT; 871 break; 872 case LINUX_UTIME_NOW: 873 times[1].tv_nsec = UTIME_NOW; 874 break; 875 default: 876 times[1].tv_nsec = l_times[1].tv_nsec; 877 break; 878 } 879 timesp = times; 880 } 881 882 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT) 883 /* This breaks POSIX, but is what the Linux kernel does 884 * _on purpose_ (documented in the man page for utimensat(2)), 885 * so we must follow that behaviour. */ 886 return (0); 887 888 if (args->pathname != NULL) 889 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); 890 else if (args->flags != 0) 891 return (EINVAL); 892 893 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW) 894 flags |= AT_SYMLINK_NOFOLLOW; 895 896 if (path == NULL) 897 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE); 898 else { 899 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp, 900 UIO_SYSSPACE, flags); 901 LFREEPATH(path); 902 } 903 904 return (error); 905} 906 907int 908linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 909{ 910 l_timeval ltv[2]; 911 struct timeval tv[2], *tvp = NULL; 912 char *fname; 913 int error, dfd; 914 915 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 916 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 917 918#ifdef DEBUG 919 if (ldebug(futimesat)) 920 printf(ARGS(futimesat, "%s, *"), fname); 921#endif 922 923 if (args->utimes != NULL) { 924 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 925 LFREEPATH(fname); 926 return (error); 927 } 928 tv[0].tv_sec = ltv[0].tv_sec; 929 tv[0].tv_usec = ltv[0].tv_usec; 930 tv[1].tv_sec = ltv[1].tv_sec; 931 tv[1].tv_usec = ltv[1].tv_usec; 932 tvp = tv; 933 } 934 935 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 936 LFREEPATH(fname); 937 return (error); 938} 939 940int 941linux_common_wait(struct thread *td, int pid, int *status, 942 int options, struct rusage *ru) 943{ 944 int error, tmpstat; 945 946 error = kern_wait(td, pid, &tmpstat, options, ru); 947 if (error) 948 return (error); 949 950 if (status) { 951 tmpstat &= 0xffff; 952 if (WIFSIGNALED(tmpstat)) 953 tmpstat = (tmpstat & 0xffffff80) | 954 bsd_to_linux_signal(WTERMSIG(tmpstat)); 955 else if (WIFSTOPPED(tmpstat)) 956 tmpstat = (tmpstat & 0xffff00ff) | 957 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8); 958 else if (WIFCONTINUED(tmpstat)) 959 tmpstat = 0xffff; 960 error = copyout(&tmpstat, status, sizeof(int)); 961 } 962 963 return (error); 964} 965 966#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 967int 968linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 969{ 970 struct linux_wait4_args wait4_args; 971 972#ifdef DEBUG 973 if (ldebug(waitpid)) 974 printf(ARGS(waitpid, "%d, %p, %d"), 975 args->pid, (void *)args->status, args->options); 976#endif 977 978 wait4_args.pid = args->pid; 979 wait4_args.status = args->status; 980 wait4_args.options = args->options; 981 wait4_args.rusage = NULL; 982 983 return (linux_wait4(td, &wait4_args)); 984} 985#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 986 987int 988linux_wait4(struct thread *td, struct linux_wait4_args *args) 989{ 990 int error, options; 991 struct rusage ru, *rup; 992 993#ifdef DEBUG 994 if (ldebug(wait4)) 995 printf(ARGS(wait4, "%d, %p, %d, %p"), 996 args->pid, (void *)args->status, args->options, 997 (void *)args->rusage); 998#endif 999 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG | 1000 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL)) 1001 return (EINVAL); 1002 1003 options = WEXITED; 1004 linux_to_bsd_waitopts(args->options, &options); 1005 1006 if (args->rusage != NULL) 1007 rup = &ru; 1008 else 1009 rup = NULL; 1010 error = linux_common_wait(td, args->pid, args->status, options, rup); 1011 if (error != 0) 1012 return (error); 1013 if (args->rusage != NULL) 1014 error = linux_copyout_rusage(&ru, args->rusage); 1015 return (error); 1016} 1017 1018int 1019linux_waitid(struct thread *td, struct linux_waitid_args *args) 1020{ 1021 int status, options, sig; 1022 struct __wrusage wru; 1023 siginfo_t siginfo; 1024 l_siginfo_t lsi; 1025 idtype_t idtype; 1026 struct proc *p; 1027 int error; 1028 1029 options = 0; 1030 linux_to_bsd_waitopts(args->options, &options); 1031 1032 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 1033 return (EINVAL); 1034 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 1035 return (EINVAL); 1036 1037 switch (args->idtype) { 1038 case LINUX_P_ALL: 1039 idtype = P_ALL; 1040 break; 1041 case LINUX_P_PID: 1042 if (args->id <= 0) 1043 return (EINVAL); 1044 idtype = P_PID; 1045 break; 1046 case LINUX_P_PGID: 1047 if (args->id <= 0) 1048 return (EINVAL); 1049 idtype = P_PGID; 1050 break; 1051 default: 1052 return (EINVAL); 1053 } 1054 1055 error = kern_wait6(td, idtype, args->id, &status, options, 1056 &wru, &siginfo); 1057 if (error != 0) 1058 return (error); 1059 if (args->rusage != NULL) { 1060 error = linux_copyout_rusage(&wru.wru_children, 1061 args->rusage); 1062 if (error != 0) 1063 return (error); 1064 } 1065 if (args->info != NULL) { 1066 p = td->td_proc; 1067 if (td->td_retval[0] == 0) 1068 bzero(&lsi, sizeof(lsi)); 1069 else { 1070 sig = bsd_to_linux_signal(siginfo.si_signo); 1071 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 1072 } 1073 error = copyout(&lsi, args->info, sizeof(lsi)); 1074 } 1075 td->td_retval[0] = 0; 1076 1077 return (error); 1078} 1079 1080int 1081linux_mknod(struct thread *td, struct linux_mknod_args *args) 1082{ 1083 char *path; 1084 int error; 1085 1086 LCONVPATHCREAT(td, args->path, &path); 1087 1088#ifdef DEBUG 1089 if (ldebug(mknod)) 1090 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode, 1091 (uintmax_t)args->dev); 1092#endif 1093 1094 switch (args->mode & S_IFMT) { 1095 case S_IFIFO: 1096 case S_IFSOCK: 1097 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 1098 break; 1099 1100 case S_IFCHR: 1101 case S_IFBLK: 1102 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1103 args->dev); 1104 break; 1105 1106 case S_IFDIR: 1107 error = EPERM; 1108 break; 1109 1110 case 0: 1111 args->mode |= S_IFREG; 1112 /* FALLTHROUGH */ 1113 case S_IFREG: 1114 error = kern_open(td, path, UIO_SYSSPACE, 1115 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1116 if (error == 0) 1117 kern_close(td, td->td_retval[0]); 1118 break; 1119 1120 default: 1121 error = EINVAL; 1122 break; 1123 } 1124 LFREEPATH(path); 1125 return (error); 1126} 1127 1128int 1129linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1130{ 1131 char *path; 1132 int error, dfd; 1133 1134 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1135 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1136 1137#ifdef DEBUG 1138 if (ldebug(mknodat)) 1139 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1140#endif 1141 1142 switch (args->mode & S_IFMT) { 1143 case S_IFIFO: 1144 case S_IFSOCK: 1145 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1146 break; 1147 1148 case S_IFCHR: 1149 case S_IFBLK: 1150 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1151 args->dev); 1152 break; 1153 1154 case S_IFDIR: 1155 error = EPERM; 1156 break; 1157 1158 case 0: 1159 args->mode |= S_IFREG; 1160 /* FALLTHROUGH */ 1161 case S_IFREG: 1162 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1163 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1164 if (error == 0) 1165 kern_close(td, td->td_retval[0]); 1166 break; 1167 1168 default: 1169 error = EINVAL; 1170 break; 1171 } 1172 LFREEPATH(path); 1173 return (error); 1174} 1175 1176/* 1177 * UGH! This is just about the dumbest idea I've ever heard!! 1178 */ 1179int 1180linux_personality(struct thread *td, struct linux_personality_args *args) 1181{ 1182#ifdef DEBUG 1183 if (ldebug(personality)) 1184 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1185#endif 1186 if (args->per != 0) 1187 return (EINVAL); 1188 1189 /* Yes Jim, it's still a Linux... */ 1190 td->td_retval[0] = 0; 1191 return (0); 1192} 1193 1194struct l_itimerval { 1195 l_timeval it_interval; 1196 l_timeval it_value; 1197}; 1198 1199#define B2L_ITIMERVAL(bip, lip) \ 1200 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1201 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1202 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1203 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1204 1205int 1206linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1207{ 1208 int error; 1209 struct l_itimerval ls; 1210 struct itimerval aitv, oitv; 1211 1212#ifdef DEBUG 1213 if (ldebug(setitimer)) 1214 printf(ARGS(setitimer, "%p, %p"), 1215 (void *)uap->itv, (void *)uap->oitv); 1216#endif 1217 1218 if (uap->itv == NULL) { 1219 uap->itv = uap->oitv; 1220 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1221 } 1222 1223 error = copyin(uap->itv, &ls, sizeof(ls)); 1224 if (error != 0) 1225 return (error); 1226 B2L_ITIMERVAL(&aitv, &ls); 1227#ifdef DEBUG 1228 if (ldebug(setitimer)) { 1229 printf("setitimer: value: sec: %jd, usec: %ld\n", 1230 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1231 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1232 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1233 } 1234#endif 1235 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1236 if (error != 0 || uap->oitv == NULL) 1237 return (error); 1238 B2L_ITIMERVAL(&ls, &oitv); 1239 1240 return (copyout(&ls, uap->oitv, sizeof(ls))); 1241} 1242 1243int 1244linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1245{ 1246 int error; 1247 struct l_itimerval ls; 1248 struct itimerval aitv; 1249 1250#ifdef DEBUG 1251 if (ldebug(getitimer)) 1252 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1253#endif 1254 error = kern_getitimer(td, uap->which, &aitv); 1255 if (error != 0) 1256 return (error); 1257 B2L_ITIMERVAL(&ls, &aitv); 1258 return (copyout(&ls, uap->itv, sizeof(ls))); 1259} 1260 1261#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1262int 1263linux_nice(struct thread *td, struct linux_nice_args *args) 1264{ 1265 struct setpriority_args bsd_args; 1266 1267 bsd_args.which = PRIO_PROCESS; 1268 bsd_args.who = 0; /* current process */ 1269 bsd_args.prio = args->inc; 1270 return (sys_setpriority(td, &bsd_args)); 1271} 1272#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1273 1274int 1275linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1276{ 1277 struct ucred *newcred, *oldcred; 1278 l_gid_t *linux_gidset; 1279 gid_t *bsd_gidset; 1280 int ngrp, error; 1281 struct proc *p; 1282 1283 ngrp = args->gidsetsize; 1284 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1285 return (EINVAL); 1286 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK); 1287 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1288 if (error) 1289 goto out; 1290 newcred = crget(); 1291 p = td->td_proc; 1292 PROC_LOCK(p); 1293 oldcred = crcopysafe(p, newcred); 1294 1295 /* 1296 * cr_groups[0] holds egid. Setting the whole set from 1297 * the supplied set will cause egid to be changed too. 1298 * Keep cr_groups[0] unchanged to prevent that. 1299 */ 1300 1301 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1302 PROC_UNLOCK(p); 1303 crfree(newcred); 1304 goto out; 1305 } 1306 1307 if (ngrp > 0) { 1308 newcred->cr_ngroups = ngrp + 1; 1309 1310 bsd_gidset = newcred->cr_groups; 1311 ngrp--; 1312 while (ngrp >= 0) { 1313 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1314 ngrp--; 1315 } 1316 } else 1317 newcred->cr_ngroups = 1; 1318 1319 setsugid(p); 1320 p->p_ucred = newcred; 1321 PROC_UNLOCK(p); 1322 crfree(oldcred); 1323 error = 0; 1324out: 1325 free(linux_gidset, M_LINUX); 1326 return (error); 1327} 1328 1329int 1330linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1331{ 1332 struct ucred *cred; 1333 l_gid_t *linux_gidset; 1334 gid_t *bsd_gidset; 1335 int bsd_gidsetsz, ngrp, error; 1336 1337 cred = td->td_ucred; 1338 bsd_gidset = cred->cr_groups; 1339 bsd_gidsetsz = cred->cr_ngroups - 1; 1340 1341 /* 1342 * cr_groups[0] holds egid. Returning the whole set 1343 * here will cause a duplicate. Exclude cr_groups[0] 1344 * to prevent that. 1345 */ 1346 1347 if ((ngrp = args->gidsetsize) == 0) { 1348 td->td_retval[0] = bsd_gidsetsz; 1349 return (0); 1350 } 1351 1352 if (ngrp < bsd_gidsetsz) 1353 return (EINVAL); 1354 1355 ngrp = 0; 1356 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1357 M_LINUX, M_WAITOK); 1358 while (ngrp < bsd_gidsetsz) { 1359 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1360 ngrp++; 1361 } 1362 1363 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1364 free(linux_gidset, M_LINUX); 1365 if (error) 1366 return (error); 1367 1368 td->td_retval[0] = ngrp; 1369 return (0); 1370} 1371 1372int 1373linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1374{ 1375 struct rlimit bsd_rlim; 1376 struct l_rlimit rlim; 1377 u_int which; 1378 int error; 1379 1380#ifdef DEBUG 1381 if (ldebug(setrlimit)) 1382 printf(ARGS(setrlimit, "%d, %p"), 1383 args->resource, (void *)args->rlim); 1384#endif 1385 1386 if (args->resource >= LINUX_RLIM_NLIMITS) 1387 return (EINVAL); 1388 1389 which = linux_to_bsd_resource[args->resource]; 1390 if (which == -1) 1391 return (EINVAL); 1392 1393 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1394 if (error) 1395 return (error); 1396 1397 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1398 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1399 return (kern_setrlimit(td, which, &bsd_rlim)); 1400} 1401 1402#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 1403int 1404linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1405{ 1406 struct l_rlimit rlim; 1407 struct proc *p = td->td_proc; 1408 struct rlimit bsd_rlim; 1409 u_int which; 1410 1411#ifdef DEBUG 1412 if (ldebug(old_getrlimit)) 1413 printf(ARGS(old_getrlimit, "%d, %p"), 1414 args->resource, (void *)args->rlim); 1415#endif 1416 1417 if (args->resource >= LINUX_RLIM_NLIMITS) 1418 return (EINVAL); 1419 1420 which = linux_to_bsd_resource[args->resource]; 1421 if (which == -1) 1422 return (EINVAL); 1423 1424 PROC_LOCK(p); 1425 lim_rlimit(p, which, &bsd_rlim); 1426 PROC_UNLOCK(p); 1427 1428#ifdef COMPAT_LINUX32 1429 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1430 if (rlim.rlim_cur == UINT_MAX) 1431 rlim.rlim_cur = INT_MAX; 1432 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1433 if (rlim.rlim_max == UINT_MAX) 1434 rlim.rlim_max = INT_MAX; 1435#else 1436 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1437 if (rlim.rlim_cur == ULONG_MAX) 1438 rlim.rlim_cur = LONG_MAX; 1439 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1440 if (rlim.rlim_max == ULONG_MAX) 1441 rlim.rlim_max = LONG_MAX; 1442#endif 1443 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1444} 1445#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 1446 1447int 1448linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1449{ 1450 struct l_rlimit rlim; 1451 struct proc *p = td->td_proc; 1452 struct rlimit bsd_rlim; 1453 u_int which; 1454 1455#ifdef DEBUG 1456 if (ldebug(getrlimit)) 1457 printf(ARGS(getrlimit, "%d, %p"), 1458 args->resource, (void *)args->rlim); 1459#endif 1460 1461 if (args->resource >= LINUX_RLIM_NLIMITS) 1462 return (EINVAL); 1463 1464 which = linux_to_bsd_resource[args->resource]; 1465 if (which == -1) 1466 return (EINVAL); 1467 1468 PROC_LOCK(p); 1469 lim_rlimit(p, which, &bsd_rlim); 1470 PROC_UNLOCK(p); 1471 1472 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1473 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1474 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1475} 1476 1477int 1478linux_sched_setscheduler(struct thread *td, 1479 struct linux_sched_setscheduler_args *args) 1480{ 1481 struct sched_param sched_param; 1482 struct thread *tdt; 1483 int error, policy; 1484 1485#ifdef DEBUG 1486 if (ldebug(sched_setscheduler)) 1487 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1488 args->pid, args->policy, (const void *)args->param); 1489#endif 1490 1491 switch (args->policy) { 1492 case LINUX_SCHED_OTHER: 1493 policy = SCHED_OTHER; 1494 break; 1495 case LINUX_SCHED_FIFO: 1496 policy = SCHED_FIFO; 1497 break; 1498 case LINUX_SCHED_RR: 1499 policy = SCHED_RR; 1500 break; 1501 default: 1502 return (EINVAL); 1503 } 1504 1505 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1506 if (error) 1507 return (error); 1508 1509 tdt = linux_tdfind(td, args->pid, -1); 1510 if (tdt == NULL) 1511 return (ESRCH); 1512 1513 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1514 PROC_UNLOCK(tdt->td_proc); 1515 return (error); 1516} 1517 1518int 1519linux_sched_getscheduler(struct thread *td, 1520 struct linux_sched_getscheduler_args *args) 1521{ 1522 struct thread *tdt; 1523 int error, policy; 1524 1525#ifdef DEBUG 1526 if (ldebug(sched_getscheduler)) 1527 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1528#endif 1529 1530 tdt = linux_tdfind(td, args->pid, -1); 1531 if (tdt == NULL) 1532 return (ESRCH); 1533 1534 error = kern_sched_getscheduler(td, tdt, &policy); 1535 PROC_UNLOCK(tdt->td_proc); 1536 1537 switch (policy) { 1538 case SCHED_OTHER: 1539 td->td_retval[0] = LINUX_SCHED_OTHER; 1540 break; 1541 case SCHED_FIFO: 1542 td->td_retval[0] = LINUX_SCHED_FIFO; 1543 break; 1544 case SCHED_RR: 1545 td->td_retval[0] = LINUX_SCHED_RR; 1546 break; 1547 } 1548 return (error); 1549} 1550 1551int 1552linux_sched_get_priority_max(struct thread *td, 1553 struct linux_sched_get_priority_max_args *args) 1554{ 1555 struct sched_get_priority_max_args bsd; 1556 1557#ifdef DEBUG 1558 if (ldebug(sched_get_priority_max)) 1559 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1560#endif 1561 1562 switch (args->policy) { 1563 case LINUX_SCHED_OTHER: 1564 bsd.policy = SCHED_OTHER; 1565 break; 1566 case LINUX_SCHED_FIFO: 1567 bsd.policy = SCHED_FIFO; 1568 break; 1569 case LINUX_SCHED_RR: 1570 bsd.policy = SCHED_RR; 1571 break; 1572 default: 1573 return (EINVAL); 1574 } 1575 return (sys_sched_get_priority_max(td, &bsd)); 1576} 1577 1578int 1579linux_sched_get_priority_min(struct thread *td, 1580 struct linux_sched_get_priority_min_args *args) 1581{ 1582 struct sched_get_priority_min_args bsd; 1583 1584#ifdef DEBUG 1585 if (ldebug(sched_get_priority_min)) 1586 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1587#endif 1588 1589 switch (args->policy) { 1590 case LINUX_SCHED_OTHER: 1591 bsd.policy = SCHED_OTHER; 1592 break; 1593 case LINUX_SCHED_FIFO: 1594 bsd.policy = SCHED_FIFO; 1595 break; 1596 case LINUX_SCHED_RR: 1597 bsd.policy = SCHED_RR; 1598 break; 1599 default: 1600 return (EINVAL); 1601 } 1602 return (sys_sched_get_priority_min(td, &bsd)); 1603} 1604 1605#define REBOOT_CAD_ON 0x89abcdef 1606#define REBOOT_CAD_OFF 0 1607#define REBOOT_HALT 0xcdef0123 1608#define REBOOT_RESTART 0x01234567 1609#define REBOOT_RESTART2 0xA1B2C3D4 1610#define REBOOT_POWEROFF 0x4321FEDC 1611#define REBOOT_MAGIC1 0xfee1dead 1612#define REBOOT_MAGIC2 0x28121969 1613#define REBOOT_MAGIC2A 0x05121996 1614#define REBOOT_MAGIC2B 0x16041998 1615 1616int 1617linux_reboot(struct thread *td, struct linux_reboot_args *args) 1618{ 1619 struct reboot_args bsd_args; 1620 1621#ifdef DEBUG 1622 if (ldebug(reboot)) 1623 printf(ARGS(reboot, "0x%x"), args->cmd); 1624#endif 1625 1626 if (args->magic1 != REBOOT_MAGIC1) 1627 return (EINVAL); 1628 1629 switch (args->magic2) { 1630 case REBOOT_MAGIC2: 1631 case REBOOT_MAGIC2A: 1632 case REBOOT_MAGIC2B: 1633 break; 1634 default: 1635 return (EINVAL); 1636 } 1637 1638 switch (args->cmd) { 1639 case REBOOT_CAD_ON: 1640 case REBOOT_CAD_OFF: 1641 return (priv_check(td, PRIV_REBOOT)); 1642 case REBOOT_HALT: 1643 bsd_args.opt = RB_HALT; 1644 break; 1645 case REBOOT_RESTART: 1646 case REBOOT_RESTART2: 1647 bsd_args.opt = 0; 1648 break; 1649 case REBOOT_POWEROFF: 1650 bsd_args.opt = RB_POWEROFF; 1651 break; 1652 default: 1653 return (EINVAL); 1654 } 1655 return (sys_reboot(td, &bsd_args)); 1656} 1657 1658 1659/* 1660 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1661 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1662 * are assumed to be preserved. The following lightweight syscalls fixes 1663 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1664 * 1665 * linux_getpid() - MP SAFE 1666 * linux_getgid() - MP SAFE 1667 * linux_getuid() - MP SAFE 1668 */ 1669 1670int 1671linux_getpid(struct thread *td, struct linux_getpid_args *args) 1672{ 1673 1674#ifdef DEBUG 1675 if (ldebug(getpid)) 1676 printf(ARGS(getpid, "")); 1677#endif 1678 td->td_retval[0] = td->td_proc->p_pid; 1679 1680 return (0); 1681} 1682 1683int 1684linux_gettid(struct thread *td, struct linux_gettid_args *args) 1685{ 1686 struct linux_emuldata *em; 1687 1688#ifdef DEBUG 1689 if (ldebug(gettid)) 1690 printf(ARGS(gettid, "")); 1691#endif 1692 1693 em = em_find(td); 1694 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1695 1696 td->td_retval[0] = em->em_tid; 1697 1698 return (0); 1699} 1700 1701 1702int 1703linux_getppid(struct thread *td, struct linux_getppid_args *args) 1704{ 1705 1706#ifdef DEBUG 1707 if (ldebug(getppid)) 1708 printf(ARGS(getppid, "")); 1709#endif 1710 1711 PROC_LOCK(td->td_proc); 1712 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1713 PROC_UNLOCK(td->td_proc); 1714 return (0); 1715} 1716 1717int 1718linux_getgid(struct thread *td, struct linux_getgid_args *args) 1719{ 1720 1721#ifdef DEBUG 1722 if (ldebug(getgid)) 1723 printf(ARGS(getgid, "")); 1724#endif 1725 1726 td->td_retval[0] = td->td_ucred->cr_rgid; 1727 return (0); 1728} 1729 1730int 1731linux_getuid(struct thread *td, struct linux_getuid_args *args) 1732{ 1733 1734#ifdef DEBUG 1735 if (ldebug(getuid)) 1736 printf(ARGS(getuid, "")); 1737#endif 1738 1739 td->td_retval[0] = td->td_ucred->cr_ruid; 1740 return (0); 1741} 1742 1743 1744int 1745linux_getsid(struct thread *td, struct linux_getsid_args *args) 1746{ 1747 struct getsid_args bsd; 1748 1749#ifdef DEBUG 1750 if (ldebug(getsid)) 1751 printf(ARGS(getsid, "%i"), args->pid); 1752#endif 1753 1754 bsd.pid = args->pid; 1755 return (sys_getsid(td, &bsd)); 1756} 1757 1758int 1759linux_nosys(struct thread *td, struct nosys_args *ignore) 1760{ 1761 1762 return (ENOSYS); 1763} 1764 1765int 1766linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1767{ 1768 struct getpriority_args bsd_args; 1769 int error; 1770 1771#ifdef DEBUG 1772 if (ldebug(getpriority)) 1773 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1774#endif 1775 1776 bsd_args.which = args->which; 1777 bsd_args.who = args->who; 1778 error = sys_getpriority(td, &bsd_args); 1779 td->td_retval[0] = 20 - td->td_retval[0]; 1780 return (error); 1781} 1782 1783int 1784linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1785{ 1786 int name[2]; 1787 1788#ifdef DEBUG 1789 if (ldebug(sethostname)) 1790 printf(ARGS(sethostname, "*, %i"), args->len); 1791#endif 1792 1793 name[0] = CTL_KERN; 1794 name[1] = KERN_HOSTNAME; 1795 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1796 args->len, 0, 0)); 1797} 1798 1799int 1800linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1801{ 1802 int name[2]; 1803 1804#ifdef DEBUG 1805 if (ldebug(setdomainname)) 1806 printf(ARGS(setdomainname, "*, %i"), args->len); 1807#endif 1808 1809 name[0] = CTL_KERN; 1810 name[1] = KERN_NISDOMAINNAME; 1811 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1812 args->len, 0, 0)); 1813} 1814 1815int 1816linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1817{ 1818 1819#ifdef DEBUG 1820 if (ldebug(exit_group)) 1821 printf(ARGS(exit_group, "%i"), args->error_code); 1822#endif 1823 1824 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1825 args->error_code); 1826 1827 /* 1828 * XXX: we should send a signal to the parent if 1829 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1830 * as it doesnt occur often. 1831 */ 1832 exit1(td, W_EXITCODE(args->error_code, 0)); 1833 /* NOTREACHED */ 1834} 1835 1836#define _LINUX_CAPABILITY_VERSION 0x19980330 1837 1838struct l_user_cap_header { 1839 l_int version; 1840 l_int pid; 1841}; 1842 1843struct l_user_cap_data { 1844 l_int effective; 1845 l_int permitted; 1846 l_int inheritable; 1847}; 1848 1849int 1850linux_capget(struct thread *td, struct linux_capget_args *args) 1851{ 1852 struct l_user_cap_header luch; 1853 struct l_user_cap_data lucd; 1854 int error; 1855 1856 if (args->hdrp == NULL) 1857 return (EFAULT); 1858 1859 error = copyin(args->hdrp, &luch, sizeof(luch)); 1860 if (error != 0) 1861 return (error); 1862 1863 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1864 luch.version = _LINUX_CAPABILITY_VERSION; 1865 error = copyout(&luch, args->hdrp, sizeof(luch)); 1866 if (error) 1867 return (error); 1868 return (EINVAL); 1869 } 1870 1871 if (luch.pid) 1872 return (EPERM); 1873 1874 if (args->datap) { 1875 /* 1876 * The current implementation doesn't support setting 1877 * a capability (it's essentially a stub) so indicate 1878 * that no capabilities are currently set or available 1879 * to request. 1880 */ 1881 bzero (&lucd, sizeof(lucd)); 1882 error = copyout(&lucd, args->datap, sizeof(lucd)); 1883 } 1884 1885 return (error); 1886} 1887 1888int 1889linux_capset(struct thread *td, struct linux_capset_args *args) 1890{ 1891 struct l_user_cap_header luch; 1892 struct l_user_cap_data lucd; 1893 int error; 1894 1895 if (args->hdrp == NULL || args->datap == NULL) 1896 return (EFAULT); 1897 1898 error = copyin(args->hdrp, &luch, sizeof(luch)); 1899 if (error != 0) 1900 return (error); 1901 1902 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1903 luch.version = _LINUX_CAPABILITY_VERSION; 1904 error = copyout(&luch, args->hdrp, sizeof(luch)); 1905 if (error) 1906 return (error); 1907 return (EINVAL); 1908 } 1909 1910 if (luch.pid) 1911 return (EPERM); 1912 1913 error = copyin(args->datap, &lucd, sizeof(lucd)); 1914 if (error != 0) 1915 return (error); 1916 1917 /* We currently don't support setting any capabilities. */ 1918 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1919 linux_msg(td, 1920 "capset effective=0x%x, permitted=0x%x, " 1921 "inheritable=0x%x is not implemented", 1922 (int)lucd.effective, (int)lucd.permitted, 1923 (int)lucd.inheritable); 1924 return (EPERM); 1925 } 1926 1927 return (0); 1928} 1929 1930int 1931linux_prctl(struct thread *td, struct linux_prctl_args *args) 1932{ 1933 int error = 0, max_size; 1934 struct proc *p = td->td_proc; 1935 char comm[LINUX_MAX_COMM_LEN]; 1936 struct linux_emuldata *em; 1937 int pdeath_signal; 1938 1939#ifdef DEBUG 1940 if (ldebug(prctl)) 1941 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option, 1942 (uintmax_t)args->arg2, (uintmax_t)args->arg3, 1943 (uintmax_t)args->arg4, (uintmax_t)args->arg5); 1944#endif 1945 1946 switch (args->option) { 1947 case LINUX_PR_SET_PDEATHSIG: 1948 if (!LINUX_SIG_VALID(args->arg2)) 1949 return (EINVAL); 1950 em = em_find(td); 1951 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1952 em->pdeath_signal = args->arg2; 1953 break; 1954 case LINUX_PR_GET_PDEATHSIG: 1955 em = em_find(td); 1956 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1957 pdeath_signal = em->pdeath_signal; 1958 error = copyout(&pdeath_signal, 1959 (void *)(register_t)args->arg2, 1960 sizeof(pdeath_signal)); 1961 break; 1962 case LINUX_PR_GET_KEEPCAPS: 1963 /* 1964 * Indicate that we always clear the effective and 1965 * permitted capability sets when the user id becomes 1966 * non-zero (actually the capability sets are simply 1967 * always zero in the current implementation). 1968 */ 1969 td->td_retval[0] = 0; 1970 break; 1971 case LINUX_PR_SET_KEEPCAPS: 1972 /* 1973 * Ignore requests to keep the effective and permitted 1974 * capability sets when the user id becomes non-zero. 1975 */ 1976 break; 1977 case LINUX_PR_SET_NAME: 1978 /* 1979 * To be on the safe side we need to make sure to not 1980 * overflow the size a linux program expects. We already 1981 * do this here in the copyin, so that we don't need to 1982 * check on copyout. 1983 */ 1984 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1985 error = copyinstr((void *)(register_t)args->arg2, comm, 1986 max_size, NULL); 1987 1988 /* Linux silently truncates the name if it is too long. */ 1989 if (error == ENAMETOOLONG) { 1990 /* 1991 * XXX: copyinstr() isn't documented to populate the 1992 * array completely, so do a copyin() to be on the 1993 * safe side. This should be changed in case 1994 * copyinstr() is changed to guarantee this. 1995 */ 1996 error = copyin((void *)(register_t)args->arg2, comm, 1997 max_size - 1); 1998 comm[max_size - 1] = '\0'; 1999 } 2000 if (error) 2001 return (error); 2002 2003 PROC_LOCK(p); 2004 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 2005 PROC_UNLOCK(p); 2006 break; 2007 case LINUX_PR_GET_NAME: 2008 PROC_LOCK(p); 2009 strlcpy(comm, p->p_comm, sizeof(comm)); 2010 PROC_UNLOCK(p); 2011 error = copyout(comm, (void *)(register_t)args->arg2, 2012 strlen(comm) + 1); 2013 break; 2014 default: 2015 error = EINVAL; 2016 break; 2017 } 2018 2019 return (error); 2020} 2021 2022int 2023linux_sched_setparam(struct thread *td, 2024 struct linux_sched_setparam_args *uap) 2025{ 2026 struct sched_param sched_param; 2027 struct thread *tdt; 2028 int error; 2029 2030#ifdef DEBUG 2031 if (ldebug(sched_setparam)) 2032 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 2033#endif 2034 2035 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 2036 if (error) 2037 return (error); 2038 2039 tdt = linux_tdfind(td, uap->pid, -1); 2040 if (tdt == NULL) 2041 return (ESRCH); 2042 2043 error = kern_sched_setparam(td, tdt, &sched_param); 2044 PROC_UNLOCK(tdt->td_proc); 2045 return (error); 2046} 2047 2048int 2049linux_sched_getparam(struct thread *td, 2050 struct linux_sched_getparam_args *uap) 2051{ 2052 struct sched_param sched_param; 2053 struct thread *tdt; 2054 int error; 2055 2056#ifdef DEBUG 2057 if (ldebug(sched_getparam)) 2058 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 2059#endif 2060 2061 tdt = linux_tdfind(td, uap->pid, -1); 2062 if (tdt == NULL) 2063 return (ESRCH); 2064 2065 error = kern_sched_getparam(td, tdt, &sched_param); 2066 PROC_UNLOCK(tdt->td_proc); 2067 if (error == 0) 2068 error = copyout(&sched_param, uap->param, 2069 sizeof(sched_param)); 2070 return (error); 2071} 2072 2073/* 2074 * Get affinity of a process. 2075 */ 2076int 2077linux_sched_getaffinity(struct thread *td, 2078 struct linux_sched_getaffinity_args *args) 2079{ 2080 int error; 2081 struct thread *tdt; 2082 struct cpuset_getaffinity_args cga; 2083 2084#ifdef DEBUG 2085 if (ldebug(sched_getaffinity)) 2086 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 2087 args->len); 2088#endif 2089 if (args->len < sizeof(cpuset_t)) 2090 return (EINVAL); 2091 2092 tdt = linux_tdfind(td, args->pid, -1); 2093 if (tdt == NULL) 2094 return (ESRCH); 2095 2096 PROC_UNLOCK(tdt->td_proc); 2097 cga.level = CPU_LEVEL_WHICH; 2098 cga.which = CPU_WHICH_TID; 2099 cga.id = tdt->td_tid; 2100 cga.cpusetsize = sizeof(cpuset_t); 2101 cga.mask = (cpuset_t *) args->user_mask_ptr; 2102 2103 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 2104 td->td_retval[0] = sizeof(cpuset_t); 2105 2106 return (error); 2107} 2108 2109/* 2110 * Set affinity of a process. 2111 */ 2112int 2113linux_sched_setaffinity(struct thread *td, 2114 struct linux_sched_setaffinity_args *args) 2115{ 2116 struct cpuset_setaffinity_args csa; 2117 struct thread *tdt; 2118 2119#ifdef DEBUG 2120 if (ldebug(sched_setaffinity)) 2121 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2122 args->len); 2123#endif 2124 if (args->len < sizeof(cpuset_t)) 2125 return (EINVAL); 2126 2127 tdt = linux_tdfind(td, args->pid, -1); 2128 if (tdt == NULL) 2129 return (ESRCH); 2130 2131 PROC_UNLOCK(tdt->td_proc); 2132 csa.level = CPU_LEVEL_WHICH; 2133 csa.which = CPU_WHICH_TID; 2134 csa.id = tdt->td_tid; 2135 csa.cpusetsize = sizeof(cpuset_t); 2136 csa.mask = (cpuset_t *) args->user_mask_ptr; 2137 2138 return (sys_cpuset_setaffinity(td, &csa)); 2139} 2140 2141struct linux_rlimit64 { 2142 uint64_t rlim_cur; 2143 uint64_t rlim_max; 2144}; 2145 2146int 2147linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args) 2148{ 2149 struct rlimit rlim, nrlim; 2150 struct linux_rlimit64 lrlim; 2151 struct proc *p; 2152 u_int which; 2153 int flags; 2154 int error; 2155 2156#ifdef DEBUG 2157 if (ldebug(prlimit64)) 2158 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid, 2159 args->resource, (void *)args->new, (void *)args->old); 2160#endif 2161 2162 if (args->resource >= LINUX_RLIM_NLIMITS) 2163 return (EINVAL); 2164 2165 which = linux_to_bsd_resource[args->resource]; 2166 if (which == -1) 2167 return (EINVAL); 2168 2169 if (args->new != NULL) { 2170 /* 2171 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux 2172 * rlim is unsigned 64-bit. FreeBSD treats negative limits 2173 * as INFINITY so we do not need a conversion even. 2174 */ 2175 error = copyin(args->new, &nrlim, sizeof(nrlim)); 2176 if (error != 0) 2177 return (error); 2178 } 2179 2180 flags = PGET_HOLD | PGET_NOTWEXIT; 2181 if (args->new != NULL) 2182 flags |= PGET_CANDEBUG; 2183 else 2184 flags |= PGET_CANSEE; 2185 error = pget(args->pid, flags, &p); 2186 if (error != 0) 2187 return (error); 2188 2189 if (args->old != NULL) { 2190 PROC_LOCK(p); 2191 lim_rlimit(p, which, &rlim); 2192 PROC_UNLOCK(p); 2193 if (rlim.rlim_cur == RLIM_INFINITY) 2194 lrlim.rlim_cur = LINUX_RLIM_INFINITY; 2195 else 2196 lrlim.rlim_cur = rlim.rlim_cur; 2197 if (rlim.rlim_max == RLIM_INFINITY) 2198 lrlim.rlim_max = LINUX_RLIM_INFINITY; 2199 else 2200 lrlim.rlim_max = rlim.rlim_max; 2201 error = copyout(&lrlim, args->old, sizeof(lrlim)); 2202 if (error != 0) 2203 goto out; 2204 } 2205 2206 if (args->new != NULL) 2207 error = kern_proc_setrlimit(td, p, which, &nrlim); 2208 2209 out: 2210 PRELE(p); 2211 return (error); 2212} 2213 2214int 2215linux_pselect6(struct thread *td, struct linux_pselect6_args *args) 2216{ 2217 struct timeval utv, tv0, tv1, *tvp; 2218 struct l_pselect6arg lpse6; 2219 struct l_timespec lts; 2220 struct timespec uts; 2221 l_sigset_t l_ss; 2222 sigset_t *ssp; 2223 sigset_t ss; 2224 int error; 2225 2226 ssp = NULL; 2227 if (args->sig != NULL) { 2228 error = copyin(args->sig, &lpse6, sizeof(lpse6)); 2229 if (error != 0) 2230 return (error); 2231 if (lpse6.ss_len != sizeof(l_ss)) 2232 return (EINVAL); 2233 if (lpse6.ss != 0) { 2234 error = copyin(PTRIN(lpse6.ss), &l_ss, 2235 sizeof(l_ss)); 2236 if (error != 0) 2237 return (error); 2238 linux_to_bsd_sigset(&l_ss, &ss); 2239 ssp = &ss; 2240 } 2241 } 2242 2243 /* 2244 * Currently glibc changes nanosecond number to microsecond. 2245 * This mean losing precision but for now it is hardly seen. 2246 */ 2247 if (args->tsp != NULL) { 2248 error = copyin(args->tsp, <s, sizeof(lts)); 2249 if (error != 0) 2250 return (error); 2251 error = linux_to_native_timespec(&uts, <s); 2252 if (error != 0) 2253 return (error); 2254 2255 TIMESPEC_TO_TIMEVAL(&utv, &uts); 2256 if (itimerfix(&utv)) 2257 return (EINVAL); 2258 2259 microtime(&tv0); 2260 tvp = &utv; 2261 } else 2262 tvp = NULL; 2263 2264 error = kern_pselect(td, args->nfds, args->readfds, args->writefds, 2265 args->exceptfds, tvp, ssp, LINUX_NFDBITS); 2266 2267 if (error == 0 && args->tsp != NULL) { 2268 if (td->td_retval[0] != 0) { 2269 /* 2270 * Compute how much time was left of the timeout, 2271 * by subtracting the current time and the time 2272 * before we started the call, and subtracting 2273 * that result from the user-supplied value. 2274 */ 2275 2276 microtime(&tv1); 2277 timevalsub(&tv1, &tv0); 2278 timevalsub(&utv, &tv1); 2279 if (utv.tv_sec < 0) 2280 timevalclear(&utv); 2281 } else 2282 timevalclear(&utv); 2283 2284 TIMEVAL_TO_TIMESPEC(&utv, &uts); 2285 2286 native_to_linux_timespec(<s, &uts); 2287 error = copyout(<s, args->tsp, sizeof(lts)); 2288 } 2289 2290 return (error); 2291} 2292 2293int 2294linux_ppoll(struct thread *td, struct linux_ppoll_args *args) 2295{ 2296 struct timespec ts0, ts1; 2297 struct l_timespec lts; 2298 struct timespec uts, *tsp; 2299 l_sigset_t l_ss; 2300 sigset_t *ssp; 2301 sigset_t ss; 2302 int error; 2303 2304 if (args->sset != NULL) { 2305 if (args->ssize != sizeof(l_ss)) 2306 return (EINVAL); 2307 error = copyin(args->sset, &l_ss, sizeof(l_ss)); 2308 if (error) 2309 return (error); 2310 linux_to_bsd_sigset(&l_ss, &ss); 2311 ssp = &ss; 2312 } else 2313 ssp = NULL; 2314 if (args->tsp != NULL) { 2315 error = copyin(args->tsp, <s, sizeof(lts)); 2316 if (error) 2317 return (error); 2318 error = linux_to_native_timespec(&uts, <s); 2319 if (error != 0) 2320 return (error); 2321 2322 nanotime(&ts0); 2323 tsp = &uts; 2324 } else 2325 tsp = NULL; 2326 2327 error = kern_poll(td, args->fds, args->nfds, tsp, ssp); 2328 2329 if (error == 0 && args->tsp != NULL) { 2330 if (td->td_retval[0]) { 2331 nanotime(&ts1); 2332 timespecsub(&ts1, &ts0); 2333 timespecsub(&uts, &ts1); 2334 if (uts.tv_sec < 0) 2335 timespecclear(&uts); 2336 } else 2337 timespecclear(&uts); 2338 2339 native_to_linux_timespec(<s, &uts); 2340 error = copyout(<s, args->tsp, sizeof(lts)); 2341 } 2342 2343 return (error); 2344} 2345 2346#if defined(DEBUG) || defined(KTR) 2347/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */ 2348 2349u_char linux_debug_map[howmany(LINUX_SYS_MAXSYSCALL, sizeof(u_char))]; 2350 2351static int 2352linux_debug(int syscall, int toggle, int global) 2353{ 2354 2355 if (global) { 2356 char c = toggle ? 0 : 0xff; 2357 2358 memset(linux_debug_map, c, sizeof(linux_debug_map)); 2359 return (0); 2360 } 2361 if (syscall < 0 || syscall >= LINUX_SYS_MAXSYSCALL) 2362 return (EINVAL); 2363 if (toggle) 2364 clrbit(linux_debug_map, syscall); 2365 else 2366 setbit(linux_debug_map, syscall); 2367 return (0); 2368} 2369 2370/* 2371 * Usage: sysctl linux.debug=<syscall_nr>.<0/1> 2372 * 2373 * E.g.: sysctl linux.debug=21.0 2374 * 2375 * As a special case, syscall "all" will apply to all syscalls globally. 2376 */ 2377#define LINUX_MAX_DEBUGSTR 16 2378int 2379linux_sysctl_debug(SYSCTL_HANDLER_ARGS) 2380{ 2381 char value[LINUX_MAX_DEBUGSTR], *p; 2382 int error, sysc, toggle; 2383 int global = 0; 2384 2385 value[0] = '\0'; 2386 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req); 2387 if (error || req->newptr == NULL) 2388 return (error); 2389 for (p = value; *p != '\0' && *p != '.'; p++); 2390 if (*p == '\0') 2391 return (EINVAL); 2392 *p++ = '\0'; 2393 sysc = strtol(value, NULL, 0); 2394 toggle = strtol(p, NULL, 0); 2395 if (strcmp(value, "all") == 0) 2396 global = 1; 2397 error = linux_debug(sysc, toggle, global); 2398 return (error); 2399} 2400 2401#endif /* DEBUG || KTR */ 2402 2403int 2404linux_sched_rr_get_interval(struct thread *td, 2405 struct linux_sched_rr_get_interval_args *uap) 2406{ 2407 struct timespec ts; 2408 struct l_timespec lts; 2409 struct thread *tdt; 2410 int error; 2411 2412 /* 2413 * According to man in case the invalid pid specified 2414 * EINVAL should be returned. 2415 */ 2416 if (uap->pid < 0) 2417 return (EINVAL); 2418 2419 tdt = linux_tdfind(td, uap->pid, -1); 2420 if (tdt == NULL) 2421 return (ESRCH); 2422 2423 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2424 PROC_UNLOCK(tdt->td_proc); 2425 if (error != 0) 2426 return (error); 2427 native_to_linux_timespec(<s, &ts); 2428 return (copyout(<s, uap->interval, sizeof(lts))); 2429} 2430 2431/* 2432 * In case when the Linux thread is the initial thread in 2433 * the thread group thread id is equal to the process id. 2434 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2435 */ 2436struct thread * 2437linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2438{ 2439 struct linux_emuldata *em; 2440 struct thread *tdt; 2441 struct proc *p; 2442 2443 tdt = NULL; 2444 if (tid == 0 || tid == td->td_tid) { 2445 tdt = td; 2446 PROC_LOCK(tdt->td_proc); 2447 } else if (tid > PID_MAX) 2448 tdt = tdfind(tid, pid); 2449 else { 2450 /* 2451 * Initial thread where the tid equal to the pid. 2452 */ 2453 p = pfind(tid); 2454 if (p != NULL) { 2455 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2456 /* 2457 * p is not a Linuxulator process. 2458 */ 2459 PROC_UNLOCK(p); 2460 return (NULL); 2461 } 2462 FOREACH_THREAD_IN_PROC(p, tdt) { 2463 em = em_find(tdt); 2464 if (tid == em->em_tid) 2465 return (tdt); 2466 } 2467 PROC_UNLOCK(p); 2468 } 2469 return (NULL); 2470 } 2471 2472 return (tdt); 2473} 2474 2475void 2476linux_to_bsd_waitopts(int options, int *bsdopts) 2477{ 2478 2479 if (options & LINUX_WNOHANG) 2480 *bsdopts |= WNOHANG; 2481 if (options & LINUX_WUNTRACED) 2482 *bsdopts |= WUNTRACED; 2483 if (options & LINUX_WEXITED) 2484 *bsdopts |= WEXITED; 2485 if (options & LINUX_WCONTINUED) 2486 *bsdopts |= WCONTINUED; 2487 if (options & LINUX_WNOWAIT) 2488 *bsdopts |= WNOWAIT; 2489 2490 if (options & __WCLONE) 2491 *bsdopts |= WLINUXCLONE; 2492} 2493