2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2002 Doug Rabson
5 * Copyright (c) 1994-1995 Søren Schmidt
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer
13 * in this position and unchanged.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. The name of the author may not be used to endorse or promote products
18 * derived from this software without specific prior written permission
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_compat.h"
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
41 #include <sys/imgact_aout.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
47 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
51 #include <sys/namei.h>
54 #include <sys/procctl.h>
55 #include <sys/reboot.h>
56 #include <sys/racct.h>
57 #include <sys/random.h>
58 #include <sys/resourcevar.h>
59 #include <sys/sched.h>
61 #include <sys/signalvar.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/sysproto.h>
66 #include <sys/systm.h>
68 #include <sys/vmmeter.h>
69 #include <sys/vnode.h>
71 #include <sys/cpuset.h>
74 #include <security/mac/mac_framework.h>
78 #include <vm/vm_kern.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_object.h>
82 #include <vm/swap_pager.h>
85 #include <machine/../linux32/linux.h>
86 #include <machine/../linux32/linux32_proto.h>
88 #include <machine/../linux/linux.h>
89 #include <machine/../linux/linux_proto.h>
92 #include <compat/linux/linux_dtrace.h>
93 #include <compat/linux/linux_file.h>
94 #include <compat/linux/linux_mib.h>
95 #include <compat/linux/linux_signal.h>
96 #include <compat/linux/linux_timer.h>
97 #include <compat/linux/linux_util.h>
98 #include <compat/linux/linux_sysproto.h>
99 #include <compat/linux/linux_emul.h>
100 #include <compat/linux/linux_misc.h>
103 * Special DTrace provider for the linuxulator.
105 * In this file we define the provider for the entire linuxulator. All
106 * modules (= files of the linuxulator) use it.
108 * We define a different name depending on the emulated bitsize, see
109 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
110 * native bitsize = linuxulator
111 * amd64, 32bit emulation = linuxulator32
113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
115 int stclohz; /* Statistics clock frequency */
117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
118 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
119 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
120 RLIMIT_MEMLOCK, RLIMIT_AS
124 l_long uptime; /* Seconds since boot */
125 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
126 #define LINUX_SYSINFO_LOADS_SCALE 65536
127 l_ulong totalram; /* Total usable main memory size */
128 l_ulong freeram; /* Available memory size */
129 l_ulong sharedram; /* Amount of shared memory */
130 l_ulong bufferram; /* Memory used by buffers */
131 l_ulong totalswap; /* Total swap space size */
132 l_ulong freeswap; /* swap space still available */
133 l_ushort procs; /* Number of current processes */
138 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
141 struct l_pselect6arg {
146 static int linux_utimensat_nsec_valid(l_long);
150 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
152 struct l_sysinfo sysinfo;
157 bzero(&sysinfo, sizeof(sysinfo));
161 sysinfo.uptime = ts.tv_sec;
163 /* Use the information from the mib to get our load averages */
164 for (i = 0; i < 3; i++)
165 sysinfo.loads[i] = averunnable.ldavg[i] *
166 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
168 sysinfo.totalram = physmem * PAGE_SIZE;
169 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
171 sysinfo.sharedram = 0;
172 mtx_lock(&vm_object_list_mtx);
173 TAILQ_FOREACH(object, &vm_object_list, object_list)
174 if (object->shadow_count > 1)
175 sysinfo.sharedram += object->resident_page_count;
176 mtx_unlock(&vm_object_list_mtx);
178 sysinfo.sharedram *= PAGE_SIZE;
179 sysinfo.bufferram = 0;
181 swap_pager_status(&i, &j);
182 sysinfo.totalswap = i * PAGE_SIZE;
183 sysinfo.freeswap = (i - j) * PAGE_SIZE;
185 sysinfo.procs = nprocs;
187 /* The following are only present in newer Linux kernels. */
188 sysinfo.totalbig = 0;
190 sysinfo.mem_unit = 1;
192 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
195 #ifdef LINUX_LEGACY_SYSCALLS
197 linux_alarm(struct thread *td, struct linux_alarm_args *args)
199 struct itimerval it, old_it;
205 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
206 * to match kern_setitimer()'s limit to avoid error from it.
208 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
211 if (secs > INT32_MAX / 2)
212 secs = INT32_MAX / 2;
214 it.it_value.tv_sec = secs;
215 it.it_value.tv_usec = 0;
216 timevalclear(&it.it_interval);
217 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
218 KASSERT(error == 0, ("kern_setitimer returns %d", error));
220 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
221 old_it.it_value.tv_usec >= 500000)
222 old_it.it_value.tv_sec++;
223 td->td_retval[0] = old_it.it_value.tv_sec;
229 linux_brk(struct thread *td, struct linux_brk_args *args)
231 struct vmspace *vm = td->td_proc->p_vmspace;
234 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
235 new = (uintptr_t)args->dsend;
236 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
237 td->td_retval[0] = (register_t)new;
239 td->td_retval[0] = (register_t)old;
244 #if defined(__i386__)
245 /* XXX: what about amd64/linux32? */
248 linux_uselib(struct thread *td, struct linux_uselib_args *args)
254 vm_map_entry_t entry;
257 unsigned long file_offset;
258 unsigned long bss_size;
262 bool locked, opened, textset;
264 LCONVPATHEXIST(td, args->library, &library);
272 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
273 UIO_SYSSPACE, library, td);
280 NDFREE(&ni, NDF_ONLY_PNBUF);
283 * From here on down, we have a locked vnode that must be unlocked.
284 * XXX: The code below largely duplicates exec_check_permissions().
289 error = VOP_GETATTR(vp, &attr, td->td_ucred);
293 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
294 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
295 /* EACCESS is what exec(2) returns. */
301 if (attr.va_size == 0) {
306 /* Can we access it? */
307 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
312 * XXX: This should use vn_open() so that it is properly authorized,
313 * and to reduce code redundancy all over the place here.
314 * XXX: Not really, it duplicates far more of exec_check_permissions()
318 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
322 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
327 /* Pull in executable header into exec_map */
328 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
329 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
333 /* Is it a Linux binary ? */
334 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
340 * While we are here, we should REALLY do some more checks
343 /* Set file/virtual offset based on a.out variant. */
344 switch ((int)(a_out->a_magic & 0xffff)) {
345 case 0413: /* ZMAGIC */
348 case 0314: /* QMAGIC */
356 bss_size = round_page(a_out->a_bss);
358 /* Check various fields in header for validity/bounds. */
359 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
364 /* text + data can't exceed file size */
365 if (a_out->a_data + a_out->a_text > attr.va_size) {
371 * text/data/bss must not exceed limits
372 * XXX - this is not complete. it should check current usage PLUS
373 * the resources needed by this library.
375 PROC_LOCK(td->td_proc);
376 if (a_out->a_text > maxtsiz ||
377 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
378 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
380 PROC_UNLOCK(td->td_proc);
384 PROC_UNLOCK(td->td_proc);
387 * Prevent more writers.
389 error = VOP_SET_TEXT(vp);
395 * Lock no longer needed
401 * Check if file_offset page aligned. Currently we cannot handle
402 * misalinged file offsets, and so we read in the entire image
405 if (file_offset & PAGE_MASK) {
406 /* Map text+data read/write/execute */
408 /* a_entry is the load address and is page aligned */
409 vmaddr = trunc_page(a_out->a_entry);
411 /* get anon user mapping, read+write+execute */
412 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
413 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
414 VM_PROT_ALL, VM_PROT_ALL, 0);
418 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
419 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
420 td->td_ucred, NOCRED, &aresid, td);
429 * for QMAGIC, a_entry is 20 bytes beyond the load address
430 * to skip the executable header
432 vmaddr = trunc_page(a_out->a_entry);
435 * Map it all into the process's space as a single
436 * copy-on-write "data" segment.
438 map = &td->td_proc->p_vmspace->vm_map;
439 error = vm_mmap(map, &vmaddr,
440 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
441 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
445 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
450 entry->eflags |= MAP_ENTRY_VN_EXEC;
456 /* Calculate BSS start address */
457 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
460 /* allocate some 'anon' space */
461 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
462 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
473 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
476 VOP_UNSET_TEXT_CHECKED(vp);
480 /* Release the temporary mapping. */
482 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
487 #endif /* __i386__ */
489 #ifdef LINUX_LEGACY_SYSCALLS
491 linux_select(struct thread *td, struct linux_select_args *args)
494 struct timeval tv0, tv1, utv, *tvp;
498 * Store current time for computation of the amount of
502 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
504 utv.tv_sec = ltv.tv_sec;
505 utv.tv_usec = ltv.tv_usec;
507 if (itimerfix(&utv)) {
509 * The timeval was invalid. Convert it to something
510 * valid that will act as it does under Linux.
512 utv.tv_sec += utv.tv_usec / 1000000;
513 utv.tv_usec %= 1000000;
514 if (utv.tv_usec < 0) {
516 utv.tv_usec += 1000000;
526 error = kern_select(td, args->nfds, args->readfds, args->writefds,
527 args->exceptfds, tvp, LINUX_NFDBITS);
532 if (td->td_retval[0]) {
534 * Compute how much time was left of the timeout,
535 * by subtracting the current time and the time
536 * before we started the call, and subtracting
537 * that result from the user-supplied value.
540 timevalsub(&tv1, &tv0);
541 timevalsub(&utv, &tv1);
546 ltv.tv_sec = utv.tv_sec;
547 ltv.tv_usec = utv.tv_usec;
548 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
558 linux_mremap(struct thread *td, struct linux_mremap_args *args)
564 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
565 td->td_retval[0] = 0;
570 * Check for the page alignment.
571 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
573 if (args->addr & PAGE_MASK) {
574 td->td_retval[0] = 0;
578 args->new_len = round_page(args->new_len);
579 args->old_len = round_page(args->old_len);
581 if (args->new_len > args->old_len) {
582 td->td_retval[0] = 0;
586 if (args->new_len < args->old_len) {
587 addr = args->addr + args->new_len;
588 len = args->old_len - args->new_len;
589 error = kern_munmap(td, addr, len);
592 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
596 #define LINUX_MS_ASYNC 0x0001
597 #define LINUX_MS_INVALIDATE 0x0002
598 #define LINUX_MS_SYNC 0x0004
601 linux_msync(struct thread *td, struct linux_msync_args *args)
604 return (kern_msync(td, args->addr, args->len,
605 args->fl & ~LINUX_MS_SYNC));
608 #ifdef LINUX_LEGACY_SYSCALLS
610 linux_time(struct thread *td, struct linux_time_args *args)
618 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
620 td->td_retval[0] = tm;
625 struct l_times_argv {
628 l_clock_t tms_cutime;
629 l_clock_t tms_cstime;
634 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
635 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
636 * auxiliary vector entry.
640 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
641 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
643 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
644 CONVNTCK(r) : CONVOTCK(r))
647 linux_times(struct thread *td, struct linux_times_args *args)
649 struct timeval tv, utime, stime, cutime, cstime;
650 struct l_times_argv tms;
654 if (args->buf != NULL) {
658 calcru(p, &utime, &stime);
660 calccru(p, &cutime, &cstime);
663 tms.tms_utime = CONVTCK(utime);
664 tms.tms_stime = CONVTCK(stime);
666 tms.tms_cutime = CONVTCK(cutime);
667 tms.tms_cstime = CONVTCK(cstime);
669 if ((error = copyout(&tms, args->buf, sizeof(tms))))
674 td->td_retval[0] = (int)CONVTCK(tv);
679 linux_newuname(struct thread *td, struct linux_newuname_args *args)
681 struct l_new_utsname utsname;
682 char osname[LINUX_MAX_UTSNAME];
683 char osrelease[LINUX_MAX_UTSNAME];
686 linux_get_osname(td, osname);
687 linux_get_osrelease(td, osrelease);
689 bzero(&utsname, sizeof(utsname));
690 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
691 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
692 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
693 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
694 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
695 for (p = utsname.version; *p != '\0'; ++p)
700 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
702 return (copyout(&utsname, args->buf, sizeof(utsname)));
710 #ifdef LINUX_LEGACY_SYSCALLS
712 linux_utime(struct thread *td, struct linux_utime_args *args)
714 struct timeval tv[2], *tvp;
715 struct l_utimbuf lut;
719 LCONVPATHEXIST(td, args->fname, &fname);
722 if ((error = copyin(args->times, &lut, sizeof lut))) {
726 tv[0].tv_sec = lut.l_actime;
728 tv[1].tv_sec = lut.l_modtime;
734 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
741 #ifdef LINUX_LEGACY_SYSCALLS
743 linux_utimes(struct thread *td, struct linux_utimes_args *args)
746 struct timeval tv[2], *tvp = NULL;
750 LCONVPATHEXIST(td, args->fname, &fname);
752 if (args->tptr != NULL) {
753 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
757 tv[0].tv_sec = ltv[0].tv_sec;
758 tv[0].tv_usec = ltv[0].tv_usec;
759 tv[1].tv_sec = ltv[1].tv_sec;
760 tv[1].tv_usec = ltv[1].tv_usec;
764 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
772 linux_utimensat_nsec_valid(l_long nsec)
775 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
777 if (nsec >= 0 && nsec <= 999999999)
783 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
785 struct l_timespec l_times[2];
786 struct timespec times[2], *timesp = NULL;
788 int error, dfd, flags = 0;
790 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
792 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
795 if (args->times != NULL) {
796 error = copyin(args->times, l_times, sizeof(l_times));
800 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
801 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
804 times[0].tv_sec = l_times[0].tv_sec;
805 switch (l_times[0].tv_nsec)
807 case LINUX_UTIME_OMIT:
808 times[0].tv_nsec = UTIME_OMIT;
810 case LINUX_UTIME_NOW:
811 times[0].tv_nsec = UTIME_NOW;
814 times[0].tv_nsec = l_times[0].tv_nsec;
817 times[1].tv_sec = l_times[1].tv_sec;
818 switch (l_times[1].tv_nsec)
820 case LINUX_UTIME_OMIT:
821 times[1].tv_nsec = UTIME_OMIT;
823 case LINUX_UTIME_NOW:
824 times[1].tv_nsec = UTIME_NOW;
827 times[1].tv_nsec = l_times[1].tv_nsec;
832 /* This breaks POSIX, but is what the Linux kernel does
833 * _on purpose_ (documented in the man page for utimensat(2)),
834 * so we must follow that behaviour. */
835 if (times[0].tv_nsec == UTIME_OMIT &&
836 times[1].tv_nsec == UTIME_OMIT)
840 if (args->pathname != NULL)
841 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
842 else if (args->flags != 0)
845 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
846 flags |= AT_SYMLINK_NOFOLLOW;
849 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
851 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
852 UIO_SYSSPACE, flags);
859 #ifdef LINUX_LEGACY_SYSCALLS
861 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
864 struct timeval tv[2], *tvp = NULL;
868 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
869 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
871 if (args->utimes != NULL) {
872 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
876 tv[0].tv_sec = ltv[0].tv_sec;
877 tv[0].tv_usec = ltv[0].tv_usec;
878 tv[1].tv_sec = ltv[1].tv_sec;
879 tv[1].tv_usec = ltv[1].tv_usec;
883 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
890 linux_common_wait(struct thread *td, int pid, int *statusp,
891 int options, struct __wrusage *wrup)
896 int error, status, tmpstat;
898 if (pid == WAIT_ANY) {
901 } else if (pid < 0) {
910 * For backward compatibility we implicitly add flags WEXITED
913 options |= WEXITED | WTRAPPED;
914 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
919 tmpstat = status & 0xffff;
920 if (WIFSIGNALED(tmpstat)) {
921 tmpstat = (tmpstat & 0xffffff80) |
922 bsd_to_linux_signal(WTERMSIG(tmpstat));
923 } else if (WIFSTOPPED(tmpstat)) {
924 tmpstat = (tmpstat & 0xffff00ff) |
925 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
926 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
927 if (WSTOPSIG(status) == SIGTRAP) {
928 tmpstat = linux_ptrace_status(td,
929 siginfo.si_pid, tmpstat);
932 } else if (WIFCONTINUED(tmpstat)) {
935 error = copyout(&tmpstat, statusp, sizeof(int));
941 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
943 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
945 struct linux_wait4_args wait4_args;
947 wait4_args.pid = args->pid;
948 wait4_args.status = args->status;
949 wait4_args.options = args->options;
950 wait4_args.rusage = NULL;
952 return (linux_wait4(td, &wait4_args));
954 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
957 linux_wait4(struct thread *td, struct linux_wait4_args *args)
960 struct __wrusage wru, *wrup;
962 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
963 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
967 linux_to_bsd_waitopts(args->options, &options);
969 if (args->rusage != NULL)
973 error = linux_common_wait(td, args->pid, args->status, options, wrup);
976 if (args->rusage != NULL)
977 error = linux_copyout_rusage(&wru.wru_self, args->rusage);
982 linux_waitid(struct thread *td, struct linux_waitid_args *args)
984 int status, options, sig;
985 struct __wrusage wru;
993 linux_to_bsd_waitopts(args->options, &options);
995 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
997 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1000 switch (args->idtype) {
1018 error = kern_wait6(td, idtype, args->id, &status, options,
1022 if (args->rusage != NULL) {
1023 error = linux_copyout_rusage(&wru.wru_children,
1028 if (args->info != NULL) {
1030 bzero(&lsi, sizeof(lsi));
1031 if (td->td_retval[0] != 0) {
1032 sig = bsd_to_linux_signal(siginfo.si_signo);
1033 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1035 error = copyout(&lsi, args->info, sizeof(lsi));
1037 td->td_retval[0] = 0;
1042 #ifdef LINUX_LEGACY_SYSCALLS
1044 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1049 LCONVPATHCREAT(td, args->path, &path);
1051 switch (args->mode & S_IFMT) {
1054 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1060 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1061 args->mode, args->dev);
1069 args->mode |= S_IFREG;
1072 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1073 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1075 kern_close(td, td->td_retval[0]);
1088 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1093 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1094 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1096 switch (args->mode & S_IFMT) {
1099 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1104 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1113 args->mode |= S_IFREG;
1116 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1117 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1119 kern_close(td, td->td_retval[0]);
1131 * UGH! This is just about the dumbest idea I've ever heard!!
1134 linux_personality(struct thread *td, struct linux_personality_args *args)
1136 struct linux_pemuldata *pem;
1137 struct proc *p = td->td_proc;
1143 if (args->per != 0xffffffff)
1144 pem->persona = args->per;
1147 td->td_retval[0] = old;
1151 struct l_itimerval {
1152 l_timeval it_interval;
1156 #define B2L_ITIMERVAL(bip, lip) \
1157 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1158 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1159 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1160 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1163 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1166 struct l_itimerval ls;
1167 struct itimerval aitv, oitv;
1169 if (uap->itv == NULL) {
1170 uap->itv = uap->oitv;
1171 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1174 error = copyin(uap->itv, &ls, sizeof(ls));
1177 B2L_ITIMERVAL(&aitv, &ls);
1178 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1179 if (error != 0 || uap->oitv == NULL)
1181 B2L_ITIMERVAL(&ls, &oitv);
1183 return (copyout(&ls, uap->oitv, sizeof(ls)));
1187 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1190 struct l_itimerval ls;
1191 struct itimerval aitv;
1193 error = kern_getitimer(td, uap->which, &aitv);
1196 B2L_ITIMERVAL(&ls, &aitv);
1197 return (copyout(&ls, uap->itv, sizeof(ls)));
1200 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1202 linux_nice(struct thread *td, struct linux_nice_args *args)
1204 struct setpriority_args bsd_args;
1206 bsd_args.which = PRIO_PROCESS;
1207 bsd_args.who = 0; /* current process */
1208 bsd_args.prio = args->inc;
1209 return (sys_setpriority(td, &bsd_args));
1211 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1214 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1216 struct ucred *newcred, *oldcred;
1217 l_gid_t *linux_gidset;
1222 ngrp = args->gidsetsize;
1223 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1225 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1226 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1230 crextend(newcred, ngrp + 1);
1233 oldcred = p->p_ucred;
1234 crcopy(newcred, oldcred);
1237 * cr_groups[0] holds egid. Setting the whole set from
1238 * the supplied set will cause egid to be changed too.
1239 * Keep cr_groups[0] unchanged to prevent that.
1242 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1249 newcred->cr_ngroups = ngrp + 1;
1251 bsd_gidset = newcred->cr_groups;
1254 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1258 newcred->cr_ngroups = 1;
1261 proc_set_cred(p, newcred);
1266 free(linux_gidset, M_LINUX);
1271 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1274 l_gid_t *linux_gidset;
1276 int bsd_gidsetsz, ngrp, error;
1278 cred = td->td_ucred;
1279 bsd_gidset = cred->cr_groups;
1280 bsd_gidsetsz = cred->cr_ngroups - 1;
1283 * cr_groups[0] holds egid. Returning the whole set
1284 * here will cause a duplicate. Exclude cr_groups[0]
1288 if ((ngrp = args->gidsetsize) == 0) {
1289 td->td_retval[0] = bsd_gidsetsz;
1293 if (ngrp < bsd_gidsetsz)
1297 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1299 while (ngrp < bsd_gidsetsz) {
1300 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1304 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1305 free(linux_gidset, M_LINUX);
1309 td->td_retval[0] = ngrp;
1314 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1316 struct rlimit bsd_rlim;
1317 struct l_rlimit rlim;
1321 if (args->resource >= LINUX_RLIM_NLIMITS)
1324 which = linux_to_bsd_resource[args->resource];
1328 error = copyin(args->rlim, &rlim, sizeof(rlim));
1332 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1333 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1334 return (kern_setrlimit(td, which, &bsd_rlim));
1337 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1339 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1341 struct l_rlimit rlim;
1342 struct rlimit bsd_rlim;
1345 if (args->resource >= LINUX_RLIM_NLIMITS)
1348 which = linux_to_bsd_resource[args->resource];
1352 lim_rlimit(td, which, &bsd_rlim);
1354 #ifdef COMPAT_LINUX32
1355 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1356 if (rlim.rlim_cur == UINT_MAX)
1357 rlim.rlim_cur = INT_MAX;
1358 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1359 if (rlim.rlim_max == UINT_MAX)
1360 rlim.rlim_max = INT_MAX;
1362 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1363 if (rlim.rlim_cur == ULONG_MAX)
1364 rlim.rlim_cur = LONG_MAX;
1365 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1366 if (rlim.rlim_max == ULONG_MAX)
1367 rlim.rlim_max = LONG_MAX;
1369 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1371 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1374 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1376 struct l_rlimit rlim;
1377 struct rlimit bsd_rlim;
1380 if (args->resource >= LINUX_RLIM_NLIMITS)
1383 which = linux_to_bsd_resource[args->resource];
1387 lim_rlimit(td, which, &bsd_rlim);
1389 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1390 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1391 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1395 linux_sched_setscheduler(struct thread *td,
1396 struct linux_sched_setscheduler_args *args)
1398 struct sched_param sched_param;
1402 switch (args->policy) {
1403 case LINUX_SCHED_OTHER:
1404 policy = SCHED_OTHER;
1406 case LINUX_SCHED_FIFO:
1407 policy = SCHED_FIFO;
1409 case LINUX_SCHED_RR:
1416 error = copyin(args->param, &sched_param, sizeof(sched_param));
1420 tdt = linux_tdfind(td, args->pid, -1);
1424 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1425 PROC_UNLOCK(tdt->td_proc);
1430 linux_sched_getscheduler(struct thread *td,
1431 struct linux_sched_getscheduler_args *args)
1436 tdt = linux_tdfind(td, args->pid, -1);
1440 error = kern_sched_getscheduler(td, tdt, &policy);
1441 PROC_UNLOCK(tdt->td_proc);
1445 td->td_retval[0] = LINUX_SCHED_OTHER;
1448 td->td_retval[0] = LINUX_SCHED_FIFO;
1451 td->td_retval[0] = LINUX_SCHED_RR;
1458 linux_sched_get_priority_max(struct thread *td,
1459 struct linux_sched_get_priority_max_args *args)
1461 struct sched_get_priority_max_args bsd;
1463 switch (args->policy) {
1464 case LINUX_SCHED_OTHER:
1465 bsd.policy = SCHED_OTHER;
1467 case LINUX_SCHED_FIFO:
1468 bsd.policy = SCHED_FIFO;
1470 case LINUX_SCHED_RR:
1471 bsd.policy = SCHED_RR;
1476 return (sys_sched_get_priority_max(td, &bsd));
1480 linux_sched_get_priority_min(struct thread *td,
1481 struct linux_sched_get_priority_min_args *args)
1483 struct sched_get_priority_min_args bsd;
1485 switch (args->policy) {
1486 case LINUX_SCHED_OTHER:
1487 bsd.policy = SCHED_OTHER;
1489 case LINUX_SCHED_FIFO:
1490 bsd.policy = SCHED_FIFO;
1492 case LINUX_SCHED_RR:
1493 bsd.policy = SCHED_RR;
1498 return (sys_sched_get_priority_min(td, &bsd));
1501 #define REBOOT_CAD_ON 0x89abcdef
1502 #define REBOOT_CAD_OFF 0
1503 #define REBOOT_HALT 0xcdef0123
1504 #define REBOOT_RESTART 0x01234567
1505 #define REBOOT_RESTART2 0xA1B2C3D4
1506 #define REBOOT_POWEROFF 0x4321FEDC
1507 #define REBOOT_MAGIC1 0xfee1dead
1508 #define REBOOT_MAGIC2 0x28121969
1509 #define REBOOT_MAGIC2A 0x05121996
1510 #define REBOOT_MAGIC2B 0x16041998
1513 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1515 struct reboot_args bsd_args;
1517 if (args->magic1 != REBOOT_MAGIC1)
1520 switch (args->magic2) {
1522 case REBOOT_MAGIC2A:
1523 case REBOOT_MAGIC2B:
1529 switch (args->cmd) {
1531 case REBOOT_CAD_OFF:
1532 return (priv_check(td, PRIV_REBOOT));
1534 bsd_args.opt = RB_HALT;
1536 case REBOOT_RESTART:
1537 case REBOOT_RESTART2:
1540 case REBOOT_POWEROFF:
1541 bsd_args.opt = RB_POWEROFF;
1546 return (sys_reboot(td, &bsd_args));
1551 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1552 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1553 * are assumed to be preserved. The following lightweight syscalls fixes
1554 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1556 * linux_getpid() - MP SAFE
1557 * linux_getgid() - MP SAFE
1558 * linux_getuid() - MP SAFE
1562 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1565 td->td_retval[0] = td->td_proc->p_pid;
1571 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1573 struct linux_emuldata *em;
1576 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1578 td->td_retval[0] = em->em_tid;
1585 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1588 td->td_retval[0] = kern_getppid(td);
1593 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1596 td->td_retval[0] = td->td_ucred->cr_rgid;
1601 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1604 td->td_retval[0] = td->td_ucred->cr_ruid;
1610 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1612 struct getsid_args bsd;
1614 bsd.pid = args->pid;
1615 return (sys_getsid(td, &bsd));
1619 linux_nosys(struct thread *td, struct nosys_args *ignore)
1626 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1628 struct getpriority_args bsd_args;
1631 bsd_args.which = args->which;
1632 bsd_args.who = args->who;
1633 error = sys_getpriority(td, &bsd_args);
1634 td->td_retval[0] = 20 - td->td_retval[0];
1639 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1644 name[1] = KERN_HOSTNAME;
1645 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1650 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1655 name[1] = KERN_NISDOMAINNAME;
1656 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1661 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1664 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1668 * XXX: we should send a signal to the parent if
1669 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1670 * as it doesnt occur often.
1672 exit1(td, args->error_code, 0);
1676 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1677 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1678 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1680 struct l_user_cap_header {
1685 struct l_user_cap_data {
1692 linux_capget(struct thread *td, struct linux_capget_args *uap)
1694 struct l_user_cap_header luch;
1695 struct l_user_cap_data lucd[2];
1698 if (uap->hdrp == NULL)
1701 error = copyin(uap->hdrp, &luch, sizeof(luch));
1705 switch (luch.version) {
1706 case _LINUX_CAPABILITY_VERSION_1:
1709 case _LINUX_CAPABILITY_VERSION_2:
1710 case _LINUX_CAPABILITY_VERSION_3:
1714 luch.version = _LINUX_CAPABILITY_VERSION_1;
1715 error = copyout(&luch, uap->hdrp, sizeof(luch));
1726 * The current implementation doesn't support setting
1727 * a capability (it's essentially a stub) so indicate
1728 * that no capabilities are currently set or available
1731 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1732 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1739 linux_capset(struct thread *td, struct linux_capset_args *uap)
1741 struct l_user_cap_header luch;
1742 struct l_user_cap_data lucd[2];
1745 if (uap->hdrp == NULL || uap->datap == NULL)
1748 error = copyin(uap->hdrp, &luch, sizeof(luch));
1752 switch (luch.version) {
1753 case _LINUX_CAPABILITY_VERSION_1:
1756 case _LINUX_CAPABILITY_VERSION_2:
1757 case _LINUX_CAPABILITY_VERSION_3:
1761 luch.version = _LINUX_CAPABILITY_VERSION_1;
1762 error = copyout(&luch, uap->hdrp, sizeof(luch));
1771 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1775 /* We currently don't support setting any capabilities. */
1776 for (i = 0; i < u32s; i++) {
1777 if (lucd[i].effective || lucd[i].permitted ||
1778 lucd[i].inheritable) {
1780 "capset[%d] effective=0x%x, permitted=0x%x, "
1781 "inheritable=0x%x is not implemented", i,
1782 (int)lucd[i].effective, (int)lucd[i].permitted,
1783 (int)lucd[i].inheritable);
1792 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1794 int error = 0, max_size;
1795 struct proc *p = td->td_proc;
1796 char comm[LINUX_MAX_COMM_LEN];
1799 switch (args->option) {
1800 case LINUX_PR_SET_PDEATHSIG:
1801 if (!LINUX_SIG_VALID(args->arg2))
1803 pdeath_signal = linux_to_bsd_signal(args->arg2);
1804 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1806 case LINUX_PR_GET_PDEATHSIG:
1807 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1811 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1812 return (copyout(&pdeath_signal,
1813 (void *)(register_t)args->arg2,
1814 sizeof(pdeath_signal)));
1816 case LINUX_PR_GET_KEEPCAPS:
1818 * Indicate that we always clear the effective and
1819 * permitted capability sets when the user id becomes
1820 * non-zero (actually the capability sets are simply
1821 * always zero in the current implementation).
1823 td->td_retval[0] = 0;
1825 case LINUX_PR_SET_KEEPCAPS:
1827 * Ignore requests to keep the effective and permitted
1828 * capability sets when the user id becomes non-zero.
1831 case LINUX_PR_SET_NAME:
1833 * To be on the safe side we need to make sure to not
1834 * overflow the size a Linux program expects. We already
1835 * do this here in the copyin, so that we don't need to
1838 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1839 error = copyinstr((void *)(register_t)args->arg2, comm,
1842 /* Linux silently truncates the name if it is too long. */
1843 if (error == ENAMETOOLONG) {
1845 * XXX: copyinstr() isn't documented to populate the
1846 * array completely, so do a copyin() to be on the
1847 * safe side. This should be changed in case
1848 * copyinstr() is changed to guarantee this.
1850 error = copyin((void *)(register_t)args->arg2, comm,
1852 comm[max_size - 1] = '\0';
1858 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1861 case LINUX_PR_GET_NAME:
1863 strlcpy(comm, p->p_comm, sizeof(comm));
1865 error = copyout(comm, (void *)(register_t)args->arg2,
1877 linux_sched_setparam(struct thread *td,
1878 struct linux_sched_setparam_args *uap)
1880 struct sched_param sched_param;
1884 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1888 tdt = linux_tdfind(td, uap->pid, -1);
1892 error = kern_sched_setparam(td, tdt, &sched_param);
1893 PROC_UNLOCK(tdt->td_proc);
1898 linux_sched_getparam(struct thread *td,
1899 struct linux_sched_getparam_args *uap)
1901 struct sched_param sched_param;
1905 tdt = linux_tdfind(td, uap->pid, -1);
1909 error = kern_sched_getparam(td, tdt, &sched_param);
1910 PROC_UNLOCK(tdt->td_proc);
1912 error = copyout(&sched_param, uap->param,
1913 sizeof(sched_param));
1918 * Get affinity of a process.
1921 linux_sched_getaffinity(struct thread *td,
1922 struct linux_sched_getaffinity_args *args)
1927 if (args->len < sizeof(cpuset_t))
1930 tdt = linux_tdfind(td, args->pid, -1);
1934 PROC_UNLOCK(tdt->td_proc);
1936 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1937 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
1939 td->td_retval[0] = sizeof(cpuset_t);
1945 * Set affinity of a process.
1948 linux_sched_setaffinity(struct thread *td,
1949 struct linux_sched_setaffinity_args *args)
1953 if (args->len < sizeof(cpuset_t))
1956 tdt = linux_tdfind(td, args->pid, -1);
1960 PROC_UNLOCK(tdt->td_proc);
1962 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1963 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
1966 struct linux_rlimit64 {
1972 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
1974 struct rlimit rlim, nrlim;
1975 struct linux_rlimit64 lrlim;
1981 if (args->resource >= LINUX_RLIM_NLIMITS)
1984 which = linux_to_bsd_resource[args->resource];
1988 if (args->new != NULL) {
1990 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
1991 * rlim is unsigned 64-bit. FreeBSD treats negative limits
1992 * as INFINITY so we do not need a conversion even.
1994 error = copyin(args->new, &nrlim, sizeof(nrlim));
1999 flags = PGET_HOLD | PGET_NOTWEXIT;
2000 if (args->new != NULL)
2001 flags |= PGET_CANDEBUG;
2003 flags |= PGET_CANSEE;
2004 error = pget(args->pid, flags, &p);
2008 if (args->old != NULL) {
2010 lim_rlimit_proc(p, which, &rlim);
2012 if (rlim.rlim_cur == RLIM_INFINITY)
2013 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2015 lrlim.rlim_cur = rlim.rlim_cur;
2016 if (rlim.rlim_max == RLIM_INFINITY)
2017 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2019 lrlim.rlim_max = rlim.rlim_max;
2020 error = copyout(&lrlim, args->old, sizeof(lrlim));
2025 if (args->new != NULL)
2026 error = kern_proc_setrlimit(td, p, which, &nrlim);
2034 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2036 struct timeval utv, tv0, tv1, *tvp;
2037 struct l_pselect6arg lpse6;
2038 struct l_timespec lts;
2039 struct timespec uts;
2046 if (args->sig != NULL) {
2047 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2050 if (lpse6.ss_len != sizeof(l_ss))
2052 if (lpse6.ss != 0) {
2053 error = copyin(PTRIN(lpse6.ss), &l_ss,
2057 linux_to_bsd_sigset(&l_ss, &ss);
2063 * Currently glibc changes nanosecond number to microsecond.
2064 * This mean losing precision but for now it is hardly seen.
2066 if (args->tsp != NULL) {
2067 error = copyin(args->tsp, <s, sizeof(lts));
2070 error = linux_to_native_timespec(&uts, <s);
2074 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2075 if (itimerfix(&utv))
2083 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2084 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2086 if (error == 0 && args->tsp != NULL) {
2087 if (td->td_retval[0] != 0) {
2089 * Compute how much time was left of the timeout,
2090 * by subtracting the current time and the time
2091 * before we started the call, and subtracting
2092 * that result from the user-supplied value.
2096 timevalsub(&tv1, &tv0);
2097 timevalsub(&utv, &tv1);
2103 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2105 error = native_to_linux_timespec(<s, &uts);
2107 error = copyout(<s, args->tsp, sizeof(lts));
2114 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2116 struct timespec ts0, ts1;
2117 struct l_timespec lts;
2118 struct timespec uts, *tsp;
2124 if (args->sset != NULL) {
2125 if (args->ssize != sizeof(l_ss))
2127 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2130 linux_to_bsd_sigset(&l_ss, &ss);
2134 if (args->tsp != NULL) {
2135 error = copyin(args->tsp, <s, sizeof(lts));
2138 error = linux_to_native_timespec(&uts, <s);
2147 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2149 if (error == 0 && args->tsp != NULL) {
2150 if (td->td_retval[0]) {
2152 timespecsub(&ts1, &ts0, &ts1);
2153 timespecsub(&uts, &ts1, &uts);
2155 timespecclear(&uts);
2157 timespecclear(&uts);
2159 error = native_to_linux_timespec(<s, &uts);
2161 error = copyout(<s, args->tsp, sizeof(lts));
2168 linux_sched_rr_get_interval(struct thread *td,
2169 struct linux_sched_rr_get_interval_args *uap)
2172 struct l_timespec lts;
2177 * According to man in case the invalid pid specified
2178 * EINVAL should be returned.
2183 tdt = linux_tdfind(td, uap->pid, -1);
2187 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2188 PROC_UNLOCK(tdt->td_proc);
2191 error = native_to_linux_timespec(<s, &ts);
2194 return (copyout(<s, uap->interval, sizeof(lts)));
2198 * In case when the Linux thread is the initial thread in
2199 * the thread group thread id is equal to the process id.
2200 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2203 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2205 struct linux_emuldata *em;
2210 if (tid == 0 || tid == td->td_tid) {
2212 PROC_LOCK(tdt->td_proc);
2213 } else if (tid > PID_MAX)
2214 tdt = tdfind(tid, pid);
2217 * Initial thread where the tid equal to the pid.
2221 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2223 * p is not a Linuxulator process.
2228 FOREACH_THREAD_IN_PROC(p, tdt) {
2230 if (tid == em->em_tid)
2242 linux_to_bsd_waitopts(int options, int *bsdopts)
2245 if (options & LINUX_WNOHANG)
2246 *bsdopts |= WNOHANG;
2247 if (options & LINUX_WUNTRACED)
2248 *bsdopts |= WUNTRACED;
2249 if (options & LINUX_WEXITED)
2250 *bsdopts |= WEXITED;
2251 if (options & LINUX_WCONTINUED)
2252 *bsdopts |= WCONTINUED;
2253 if (options & LINUX_WNOWAIT)
2254 *bsdopts |= WNOWAIT;
2256 if (options & __WCLONE)
2257 *bsdopts |= WLINUXCLONE;
2261 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2267 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2269 if (args->count > INT_MAX)
2270 args->count = INT_MAX;
2272 iov.iov_base = args->buf;
2273 iov.iov_len = args->count;
2277 uio.uio_resid = iov.iov_len;
2278 uio.uio_segflg = UIO_USERSPACE;
2279 uio.uio_rw = UIO_READ;
2282 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2284 td->td_retval[0] = args->count - uio.uio_resid;
2289 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2292 /* Needs to be page-aligned */
2293 if (args->start & PAGE_MASK)
2295 return (kern_mincore(td, args->start, args->len, args->vec));