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/msgbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
55 #include <sys/procctl.h>
56 #include <sys/reboot.h>
57 #include <sys/racct.h>
58 #include <sys/random.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sched.h>
62 #include <sys/signalvar.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/systm.h>
69 #include <sys/vmmeter.h>
70 #include <sys/vnode.h>
72 #include <sys/cpuset.h>
75 #include <security/mac/mac_framework.h>
79 #include <vm/vm_kern.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.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;
156 bzero(&sysinfo, sizeof(sysinfo));
160 sysinfo.uptime = ts.tv_sec;
162 /* Use the information from the mib to get our load averages */
163 for (i = 0; i < 3; i++)
164 sysinfo.loads[i] = averunnable.ldavg[i] *
165 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
167 sysinfo.totalram = physmem * PAGE_SIZE;
168 sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
171 * sharedram counts pages allocated to named, swap-backed objects such
172 * as shared memory segments and tmpfs files. There is no cheap way to
173 * compute this, so just leave the field unpopulated. Linux itself only
174 * started setting this field in the 3.x timeframe.
176 sysinfo.sharedram = 0;
177 sysinfo.bufferram = 0;
179 swap_pager_status(&i, &j);
180 sysinfo.totalswap = i * PAGE_SIZE;
181 sysinfo.freeswap = (i - j) * PAGE_SIZE;
183 sysinfo.procs = nprocs;
186 * Platforms supported by the emulation layer do not have a notion of
189 sysinfo.totalhigh = 0;
190 sysinfo.freehigh = 0;
192 sysinfo.mem_unit = 1;
194 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
197 #ifdef LINUX_LEGACY_SYSCALLS
199 linux_alarm(struct thread *td, struct linux_alarm_args *args)
201 struct itimerval it, old_it;
207 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
208 * to match kern_setitimer()'s limit to avoid error from it.
210 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
213 if (secs > INT32_MAX / 2)
214 secs = INT32_MAX / 2;
216 it.it_value.tv_sec = secs;
217 it.it_value.tv_usec = 0;
218 timevalclear(&it.it_interval);
219 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
220 KASSERT(error == 0, ("kern_setitimer returns %d", error));
222 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
223 old_it.it_value.tv_usec >= 500000)
224 old_it.it_value.tv_sec++;
225 td->td_retval[0] = old_it.it_value.tv_sec;
231 linux_brk(struct thread *td, struct linux_brk_args *args)
233 struct vmspace *vm = td->td_proc->p_vmspace;
236 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
237 new = (uintptr_t)args->dsend;
238 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
239 td->td_retval[0] = (register_t)new;
241 td->td_retval[0] = (register_t)old;
246 #if defined(__i386__)
247 /* XXX: what about amd64/linux32? */
250 linux_uselib(struct thread *td, struct linux_uselib_args *args)
256 vm_map_entry_t entry;
259 unsigned long file_offset;
260 unsigned long bss_size;
264 bool locked, opened, textset;
266 LCONVPATHEXIST(td, args->library, &library);
274 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
275 UIO_SYSSPACE, library, td);
282 NDFREE(&ni, NDF_ONLY_PNBUF);
285 * From here on down, we have a locked vnode that must be unlocked.
286 * XXX: The code below largely duplicates exec_check_permissions().
291 error = VOP_GETATTR(vp, &attr, td->td_ucred);
295 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
296 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
297 /* EACCESS is what exec(2) returns. */
303 if (attr.va_size == 0) {
308 /* Can we access it? */
309 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
314 * XXX: This should use vn_open() so that it is properly authorized,
315 * and to reduce code redundancy all over the place here.
316 * XXX: Not really, it duplicates far more of exec_check_permissions()
320 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
324 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
329 /* Pull in executable header into exec_map */
330 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
331 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
335 /* Is it a Linux binary ? */
336 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
342 * While we are here, we should REALLY do some more checks
345 /* Set file/virtual offset based on a.out variant. */
346 switch ((int)(a_out->a_magic & 0xffff)) {
347 case 0413: /* ZMAGIC */
350 case 0314: /* QMAGIC */
358 bss_size = round_page(a_out->a_bss);
360 /* Check various fields in header for validity/bounds. */
361 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
366 /* text + data can't exceed file size */
367 if (a_out->a_data + a_out->a_text > attr.va_size) {
373 * text/data/bss must not exceed limits
374 * XXX - this is not complete. it should check current usage PLUS
375 * the resources needed by this library.
377 PROC_LOCK(td->td_proc);
378 if (a_out->a_text > maxtsiz ||
379 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
380 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
382 PROC_UNLOCK(td->td_proc);
386 PROC_UNLOCK(td->td_proc);
389 * Prevent more writers.
391 error = VOP_SET_TEXT(vp);
397 * Lock no longer needed
403 * Check if file_offset page aligned. Currently we cannot handle
404 * misalinged file offsets, and so we read in the entire image
407 if (file_offset & PAGE_MASK) {
408 /* Map text+data read/write/execute */
410 /* a_entry is the load address and is page aligned */
411 vmaddr = trunc_page(a_out->a_entry);
413 /* get anon user mapping, read+write+execute */
414 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
415 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
416 VM_PROT_ALL, VM_PROT_ALL, 0);
420 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
421 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
422 td->td_ucred, NOCRED, &aresid, td);
431 * for QMAGIC, a_entry is 20 bytes beyond the load address
432 * to skip the executable header
434 vmaddr = trunc_page(a_out->a_entry);
437 * Map it all into the process's space as a single
438 * copy-on-write "data" segment.
440 map = &td->td_proc->p_vmspace->vm_map;
441 error = vm_mmap(map, &vmaddr,
442 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
443 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
447 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
452 entry->eflags |= MAP_ENTRY_VN_EXEC;
458 /* Calculate BSS start address */
459 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
462 /* allocate some 'anon' space */
463 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
464 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
475 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
480 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
482 VOP_UNSET_TEXT_CHECKED(vp);
487 /* Release the temporary mapping. */
489 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
494 #endif /* __i386__ */
496 #ifdef LINUX_LEGACY_SYSCALLS
498 linux_select(struct thread *td, struct linux_select_args *args)
501 struct timeval tv0, tv1, utv, *tvp;
505 * Store current time for computation of the amount of
509 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
511 utv.tv_sec = ltv.tv_sec;
512 utv.tv_usec = ltv.tv_usec;
514 if (itimerfix(&utv)) {
516 * The timeval was invalid. Convert it to something
517 * valid that will act as it does under Linux.
519 utv.tv_sec += utv.tv_usec / 1000000;
520 utv.tv_usec %= 1000000;
521 if (utv.tv_usec < 0) {
523 utv.tv_usec += 1000000;
533 error = kern_select(td, args->nfds, args->readfds, args->writefds,
534 args->exceptfds, tvp, LINUX_NFDBITS);
539 if (td->td_retval[0]) {
541 * Compute how much time was left of the timeout,
542 * by subtracting the current time and the time
543 * before we started the call, and subtracting
544 * that result from the user-supplied value.
547 timevalsub(&tv1, &tv0);
548 timevalsub(&utv, &tv1);
553 ltv.tv_sec = utv.tv_sec;
554 ltv.tv_usec = utv.tv_usec;
555 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
565 linux_mremap(struct thread *td, struct linux_mremap_args *args)
571 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
572 td->td_retval[0] = 0;
577 * Check for the page alignment.
578 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
580 if (args->addr & PAGE_MASK) {
581 td->td_retval[0] = 0;
585 args->new_len = round_page(args->new_len);
586 args->old_len = round_page(args->old_len);
588 if (args->new_len > args->old_len) {
589 td->td_retval[0] = 0;
593 if (args->new_len < args->old_len) {
594 addr = args->addr + args->new_len;
595 len = args->old_len - args->new_len;
596 error = kern_munmap(td, addr, len);
599 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
603 #define LINUX_MS_ASYNC 0x0001
604 #define LINUX_MS_INVALIDATE 0x0002
605 #define LINUX_MS_SYNC 0x0004
608 linux_msync(struct thread *td, struct linux_msync_args *args)
611 return (kern_msync(td, args->addr, args->len,
612 args->fl & ~LINUX_MS_SYNC));
615 #ifdef LINUX_LEGACY_SYSCALLS
617 linux_time(struct thread *td, struct linux_time_args *args)
625 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
627 td->td_retval[0] = tm;
632 struct l_times_argv {
635 l_clock_t tms_cutime;
636 l_clock_t tms_cstime;
641 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
642 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
643 * auxiliary vector entry.
647 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
648 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
650 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
651 CONVNTCK(r) : CONVOTCK(r))
654 linux_times(struct thread *td, struct linux_times_args *args)
656 struct timeval tv, utime, stime, cutime, cstime;
657 struct l_times_argv tms;
661 if (args->buf != NULL) {
665 calcru(p, &utime, &stime);
667 calccru(p, &cutime, &cstime);
670 tms.tms_utime = CONVTCK(utime);
671 tms.tms_stime = CONVTCK(stime);
673 tms.tms_cutime = CONVTCK(cutime);
674 tms.tms_cstime = CONVTCK(cstime);
676 if ((error = copyout(&tms, args->buf, sizeof(tms))))
681 td->td_retval[0] = (int)CONVTCK(tv);
686 linux_newuname(struct thread *td, struct linux_newuname_args *args)
688 struct l_new_utsname utsname;
689 char osname[LINUX_MAX_UTSNAME];
690 char osrelease[LINUX_MAX_UTSNAME];
693 linux_get_osname(td, osname);
694 linux_get_osrelease(td, osrelease);
696 bzero(&utsname, sizeof(utsname));
697 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
698 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
699 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
700 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
701 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
702 for (p = utsname.version; *p != '\0'; ++p)
707 #if defined(__amd64__)
709 * On amd64, Linux uname(2) needs to return "x86_64"
710 * for both 64-bit and 32-bit applications. On 32-bit,
711 * the string returned by getauxval(AT_PLATFORM) needs
712 * to remain "i686", though.
714 strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
716 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
719 return (copyout(&utsname, args->buf, sizeof(utsname)));
727 #ifdef LINUX_LEGACY_SYSCALLS
729 linux_utime(struct thread *td, struct linux_utime_args *args)
731 struct timeval tv[2], *tvp;
732 struct l_utimbuf lut;
736 LCONVPATHEXIST(td, args->fname, &fname);
739 if ((error = copyin(args->times, &lut, sizeof lut))) {
743 tv[0].tv_sec = lut.l_actime;
745 tv[1].tv_sec = lut.l_modtime;
751 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
758 #ifdef LINUX_LEGACY_SYSCALLS
760 linux_utimes(struct thread *td, struct linux_utimes_args *args)
763 struct timeval tv[2], *tvp = NULL;
767 LCONVPATHEXIST(td, args->fname, &fname);
769 if (args->tptr != NULL) {
770 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
774 tv[0].tv_sec = ltv[0].tv_sec;
775 tv[0].tv_usec = ltv[0].tv_usec;
776 tv[1].tv_sec = ltv[1].tv_sec;
777 tv[1].tv_usec = ltv[1].tv_usec;
781 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
789 linux_utimensat_nsec_valid(l_long nsec)
792 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
794 if (nsec >= 0 && nsec <= 999999999)
800 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
802 struct l_timespec l_times[2];
803 struct timespec times[2], *timesp = NULL;
805 int error, dfd, flags = 0;
807 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
809 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
812 if (args->times != NULL) {
813 error = copyin(args->times, l_times, sizeof(l_times));
817 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
818 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
821 times[0].tv_sec = l_times[0].tv_sec;
822 switch (l_times[0].tv_nsec)
824 case LINUX_UTIME_OMIT:
825 times[0].tv_nsec = UTIME_OMIT;
827 case LINUX_UTIME_NOW:
828 times[0].tv_nsec = UTIME_NOW;
831 times[0].tv_nsec = l_times[0].tv_nsec;
834 times[1].tv_sec = l_times[1].tv_sec;
835 switch (l_times[1].tv_nsec)
837 case LINUX_UTIME_OMIT:
838 times[1].tv_nsec = UTIME_OMIT;
840 case LINUX_UTIME_NOW:
841 times[1].tv_nsec = UTIME_NOW;
844 times[1].tv_nsec = l_times[1].tv_nsec;
849 /* This breaks POSIX, but is what the Linux kernel does
850 * _on purpose_ (documented in the man page for utimensat(2)),
851 * so we must follow that behaviour. */
852 if (times[0].tv_nsec == UTIME_OMIT &&
853 times[1].tv_nsec == UTIME_OMIT)
857 if (args->pathname != NULL)
858 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
859 else if (args->flags != 0)
862 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
863 flags |= AT_SYMLINK_NOFOLLOW;
866 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
868 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
869 UIO_SYSSPACE, flags);
876 #ifdef LINUX_LEGACY_SYSCALLS
878 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
881 struct timeval tv[2], *tvp = NULL;
885 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
886 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
888 if (args->utimes != NULL) {
889 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
893 tv[0].tv_sec = ltv[0].tv_sec;
894 tv[0].tv_usec = ltv[0].tv_usec;
895 tv[1].tv_sec = ltv[1].tv_sec;
896 tv[1].tv_usec = ltv[1].tv_usec;
900 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
907 linux_common_wait(struct thread *td, int pid, int *statusp,
908 int options, struct __wrusage *wrup)
913 int error, status, tmpstat;
915 if (pid == WAIT_ANY) {
918 } else if (pid < 0) {
927 * For backward compatibility we implicitly add flags WEXITED
930 options |= WEXITED | WTRAPPED;
931 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
936 tmpstat = status & 0xffff;
937 if (WIFSIGNALED(tmpstat)) {
938 tmpstat = (tmpstat & 0xffffff80) |
939 bsd_to_linux_signal(WTERMSIG(tmpstat));
940 } else if (WIFSTOPPED(tmpstat)) {
941 tmpstat = (tmpstat & 0xffff00ff) |
942 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
943 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
944 if (WSTOPSIG(status) == SIGTRAP) {
945 tmpstat = linux_ptrace_status(td,
946 siginfo.si_pid, tmpstat);
949 } else if (WIFCONTINUED(tmpstat)) {
952 error = copyout(&tmpstat, statusp, sizeof(int));
958 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
960 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
962 struct linux_wait4_args wait4_args;
964 wait4_args.pid = args->pid;
965 wait4_args.status = args->status;
966 wait4_args.options = args->options;
967 wait4_args.rusage = NULL;
969 return (linux_wait4(td, &wait4_args));
971 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
974 linux_wait4(struct thread *td, struct linux_wait4_args *args)
977 struct __wrusage wru, *wrup;
979 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
980 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
984 linux_to_bsd_waitopts(args->options, &options);
986 if (args->rusage != NULL)
990 error = linux_common_wait(td, args->pid, args->status, options, wrup);
993 if (args->rusage != NULL)
994 error = linux_copyout_rusage(&wru.wru_self, args->rusage);
999 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1001 int status, options, sig;
1002 struct __wrusage wru;
1010 linux_to_bsd_waitopts(args->options, &options);
1012 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1014 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1017 switch (args->idtype) {
1035 error = kern_wait6(td, idtype, args->id, &status, options,
1039 if (args->rusage != NULL) {
1040 error = linux_copyout_rusage(&wru.wru_children,
1045 if (args->info != NULL) {
1047 bzero(&lsi, sizeof(lsi));
1048 if (td->td_retval[0] != 0) {
1049 sig = bsd_to_linux_signal(siginfo.si_signo);
1050 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1052 error = copyout(&lsi, args->info, sizeof(lsi));
1054 td->td_retval[0] = 0;
1059 #ifdef LINUX_LEGACY_SYSCALLS
1061 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1066 LCONVPATHCREAT(td, args->path, &path);
1068 switch (args->mode & S_IFMT) {
1071 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1077 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1078 args->mode, args->dev);
1086 args->mode |= S_IFREG;
1089 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1090 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1092 kern_close(td, td->td_retval[0]);
1105 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1110 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1111 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1113 switch (args->mode & S_IFMT) {
1116 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1121 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1130 args->mode |= S_IFREG;
1133 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1134 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1136 kern_close(td, td->td_retval[0]);
1148 * UGH! This is just about the dumbest idea I've ever heard!!
1151 linux_personality(struct thread *td, struct linux_personality_args *args)
1153 struct linux_pemuldata *pem;
1154 struct proc *p = td->td_proc;
1160 if (args->per != 0xffffffff)
1161 pem->persona = args->per;
1164 td->td_retval[0] = old;
1168 struct l_itimerval {
1169 l_timeval it_interval;
1173 #define B2L_ITIMERVAL(bip, lip) \
1174 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1175 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1176 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1177 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1180 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1183 struct l_itimerval ls;
1184 struct itimerval aitv, oitv;
1186 if (uap->itv == NULL) {
1187 uap->itv = uap->oitv;
1188 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1191 error = copyin(uap->itv, &ls, sizeof(ls));
1194 B2L_ITIMERVAL(&aitv, &ls);
1195 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1196 if (error != 0 || uap->oitv == NULL)
1198 B2L_ITIMERVAL(&ls, &oitv);
1200 return (copyout(&ls, uap->oitv, sizeof(ls)));
1204 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1207 struct l_itimerval ls;
1208 struct itimerval aitv;
1210 error = kern_getitimer(td, uap->which, &aitv);
1213 B2L_ITIMERVAL(&ls, &aitv);
1214 return (copyout(&ls, uap->itv, sizeof(ls)));
1217 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1219 linux_nice(struct thread *td, struct linux_nice_args *args)
1222 return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1224 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1227 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1229 struct ucred *newcred, *oldcred;
1230 l_gid_t *linux_gidset;
1235 ngrp = args->gidsetsize;
1236 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1238 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1239 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1243 crextend(newcred, ngrp + 1);
1246 oldcred = p->p_ucred;
1247 crcopy(newcred, oldcred);
1250 * cr_groups[0] holds egid. Setting the whole set from
1251 * the supplied set will cause egid to be changed too.
1252 * Keep cr_groups[0] unchanged to prevent that.
1255 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1262 newcred->cr_ngroups = ngrp + 1;
1264 bsd_gidset = newcred->cr_groups;
1267 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1271 newcred->cr_ngroups = 1;
1274 proc_set_cred(p, newcred);
1279 free(linux_gidset, M_LINUX);
1284 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1287 l_gid_t *linux_gidset;
1289 int bsd_gidsetsz, ngrp, error;
1291 cred = td->td_ucred;
1292 bsd_gidset = cred->cr_groups;
1293 bsd_gidsetsz = cred->cr_ngroups - 1;
1296 * cr_groups[0] holds egid. Returning the whole set
1297 * here will cause a duplicate. Exclude cr_groups[0]
1301 if ((ngrp = args->gidsetsize) == 0) {
1302 td->td_retval[0] = bsd_gidsetsz;
1306 if (ngrp < bsd_gidsetsz)
1310 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1312 while (ngrp < bsd_gidsetsz) {
1313 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1317 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1318 free(linux_gidset, M_LINUX);
1322 td->td_retval[0] = ngrp;
1327 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1329 struct rlimit bsd_rlim;
1330 struct l_rlimit rlim;
1334 if (args->resource >= LINUX_RLIM_NLIMITS)
1337 which = linux_to_bsd_resource[args->resource];
1341 error = copyin(args->rlim, &rlim, sizeof(rlim));
1345 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1346 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1347 return (kern_setrlimit(td, which, &bsd_rlim));
1350 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1352 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1354 struct l_rlimit rlim;
1355 struct rlimit bsd_rlim;
1358 if (args->resource >= LINUX_RLIM_NLIMITS)
1361 which = linux_to_bsd_resource[args->resource];
1365 lim_rlimit(td, which, &bsd_rlim);
1367 #ifdef COMPAT_LINUX32
1368 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1369 if (rlim.rlim_cur == UINT_MAX)
1370 rlim.rlim_cur = INT_MAX;
1371 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1372 if (rlim.rlim_max == UINT_MAX)
1373 rlim.rlim_max = INT_MAX;
1375 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1376 if (rlim.rlim_cur == ULONG_MAX)
1377 rlim.rlim_cur = LONG_MAX;
1378 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1379 if (rlim.rlim_max == ULONG_MAX)
1380 rlim.rlim_max = LONG_MAX;
1382 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1384 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1387 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1389 struct l_rlimit rlim;
1390 struct rlimit bsd_rlim;
1393 if (args->resource >= LINUX_RLIM_NLIMITS)
1396 which = linux_to_bsd_resource[args->resource];
1400 lim_rlimit(td, which, &bsd_rlim);
1402 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1403 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1404 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1408 linux_sched_setscheduler(struct thread *td,
1409 struct linux_sched_setscheduler_args *args)
1411 struct sched_param sched_param;
1415 switch (args->policy) {
1416 case LINUX_SCHED_OTHER:
1417 policy = SCHED_OTHER;
1419 case LINUX_SCHED_FIFO:
1420 policy = SCHED_FIFO;
1422 case LINUX_SCHED_RR:
1429 error = copyin(args->param, &sched_param, sizeof(sched_param));
1433 if (linux_map_sched_prio) {
1436 if (sched_param.sched_priority != 0)
1439 sched_param.sched_priority =
1440 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1444 if (sched_param.sched_priority < 1 ||
1445 sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1449 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1450 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1452 sched_param.sched_priority =
1453 (sched_param.sched_priority - 1) *
1454 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1455 (LINUX_MAX_RT_PRIO - 1);
1460 tdt = linux_tdfind(td, args->pid, -1);
1464 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1465 PROC_UNLOCK(tdt->td_proc);
1470 linux_sched_getscheduler(struct thread *td,
1471 struct linux_sched_getscheduler_args *args)
1476 tdt = linux_tdfind(td, args->pid, -1);
1480 error = kern_sched_getscheduler(td, tdt, &policy);
1481 PROC_UNLOCK(tdt->td_proc);
1485 td->td_retval[0] = LINUX_SCHED_OTHER;
1488 td->td_retval[0] = LINUX_SCHED_FIFO;
1491 td->td_retval[0] = LINUX_SCHED_RR;
1498 linux_sched_get_priority_max(struct thread *td,
1499 struct linux_sched_get_priority_max_args *args)
1501 struct sched_get_priority_max_args bsd;
1503 if (linux_map_sched_prio) {
1504 switch (args->policy) {
1505 case LINUX_SCHED_OTHER:
1506 td->td_retval[0] = 0;
1508 case LINUX_SCHED_FIFO:
1509 case LINUX_SCHED_RR:
1510 td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1517 switch (args->policy) {
1518 case LINUX_SCHED_OTHER:
1519 bsd.policy = SCHED_OTHER;
1521 case LINUX_SCHED_FIFO:
1522 bsd.policy = SCHED_FIFO;
1524 case LINUX_SCHED_RR:
1525 bsd.policy = SCHED_RR;
1530 return (sys_sched_get_priority_max(td, &bsd));
1534 linux_sched_get_priority_min(struct thread *td,
1535 struct linux_sched_get_priority_min_args *args)
1537 struct sched_get_priority_min_args bsd;
1539 if (linux_map_sched_prio) {
1540 switch (args->policy) {
1541 case LINUX_SCHED_OTHER:
1542 td->td_retval[0] = 0;
1544 case LINUX_SCHED_FIFO:
1545 case LINUX_SCHED_RR:
1546 td->td_retval[0] = 1;
1553 switch (args->policy) {
1554 case LINUX_SCHED_OTHER:
1555 bsd.policy = SCHED_OTHER;
1557 case LINUX_SCHED_FIFO:
1558 bsd.policy = SCHED_FIFO;
1560 case LINUX_SCHED_RR:
1561 bsd.policy = SCHED_RR;
1566 return (sys_sched_get_priority_min(td, &bsd));
1569 #define REBOOT_CAD_ON 0x89abcdef
1570 #define REBOOT_CAD_OFF 0
1571 #define REBOOT_HALT 0xcdef0123
1572 #define REBOOT_RESTART 0x01234567
1573 #define REBOOT_RESTART2 0xA1B2C3D4
1574 #define REBOOT_POWEROFF 0x4321FEDC
1575 #define REBOOT_MAGIC1 0xfee1dead
1576 #define REBOOT_MAGIC2 0x28121969
1577 #define REBOOT_MAGIC2A 0x05121996
1578 #define REBOOT_MAGIC2B 0x16041998
1581 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1583 struct reboot_args bsd_args;
1585 if (args->magic1 != REBOOT_MAGIC1)
1588 switch (args->magic2) {
1590 case REBOOT_MAGIC2A:
1591 case REBOOT_MAGIC2B:
1597 switch (args->cmd) {
1599 case REBOOT_CAD_OFF:
1600 return (priv_check(td, PRIV_REBOOT));
1602 bsd_args.opt = RB_HALT;
1604 case REBOOT_RESTART:
1605 case REBOOT_RESTART2:
1608 case REBOOT_POWEROFF:
1609 bsd_args.opt = RB_POWEROFF;
1614 return (sys_reboot(td, &bsd_args));
1619 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1622 td->td_retval[0] = td->td_proc->p_pid;
1628 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1630 struct linux_emuldata *em;
1633 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1635 td->td_retval[0] = em->em_tid;
1642 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1645 td->td_retval[0] = kern_getppid(td);
1650 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1653 td->td_retval[0] = td->td_ucred->cr_rgid;
1658 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1661 td->td_retval[0] = td->td_ucred->cr_ruid;
1666 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1669 return (kern_getsid(td, args->pid));
1673 linux_nosys(struct thread *td, struct nosys_args *ignore)
1680 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1684 error = kern_getpriority(td, args->which, args->who);
1685 td->td_retval[0] = 20 - td->td_retval[0];
1690 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1695 name[1] = KERN_HOSTNAME;
1696 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1701 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1706 name[1] = KERN_NISDOMAINNAME;
1707 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1712 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1715 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1719 * XXX: we should send a signal to the parent if
1720 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1721 * as it doesnt occur often.
1723 exit1(td, args->error_code, 0);
1727 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1728 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1729 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1731 struct l_user_cap_header {
1736 struct l_user_cap_data {
1743 linux_capget(struct thread *td, struct linux_capget_args *uap)
1745 struct l_user_cap_header luch;
1746 struct l_user_cap_data lucd[2];
1749 if (uap->hdrp == NULL)
1752 error = copyin(uap->hdrp, &luch, sizeof(luch));
1756 switch (luch.version) {
1757 case _LINUX_CAPABILITY_VERSION_1:
1760 case _LINUX_CAPABILITY_VERSION_2:
1761 case _LINUX_CAPABILITY_VERSION_3:
1765 luch.version = _LINUX_CAPABILITY_VERSION_1;
1766 error = copyout(&luch, uap->hdrp, sizeof(luch));
1777 * The current implementation doesn't support setting
1778 * a capability (it's essentially a stub) so indicate
1779 * that no capabilities are currently set or available
1782 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1783 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1790 linux_capset(struct thread *td, struct linux_capset_args *uap)
1792 struct l_user_cap_header luch;
1793 struct l_user_cap_data lucd[2];
1796 if (uap->hdrp == NULL || uap->datap == NULL)
1799 error = copyin(uap->hdrp, &luch, sizeof(luch));
1803 switch (luch.version) {
1804 case _LINUX_CAPABILITY_VERSION_1:
1807 case _LINUX_CAPABILITY_VERSION_2:
1808 case _LINUX_CAPABILITY_VERSION_3:
1812 luch.version = _LINUX_CAPABILITY_VERSION_1;
1813 error = copyout(&luch, uap->hdrp, sizeof(luch));
1822 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1826 /* We currently don't support setting any capabilities. */
1827 for (i = 0; i < u32s; i++) {
1828 if (lucd[i].effective || lucd[i].permitted ||
1829 lucd[i].inheritable) {
1831 "capset[%d] effective=0x%x, permitted=0x%x, "
1832 "inheritable=0x%x is not implemented", i,
1833 (int)lucd[i].effective, (int)lucd[i].permitted,
1834 (int)lucd[i].inheritable);
1843 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1845 int error = 0, max_size;
1846 struct proc *p = td->td_proc;
1847 char comm[LINUX_MAX_COMM_LEN];
1850 switch (args->option) {
1851 case LINUX_PR_SET_PDEATHSIG:
1852 if (!LINUX_SIG_VALID(args->arg2))
1854 pdeath_signal = linux_to_bsd_signal(args->arg2);
1855 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1857 case LINUX_PR_GET_PDEATHSIG:
1858 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1862 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1863 return (copyout(&pdeath_signal,
1864 (void *)(register_t)args->arg2,
1865 sizeof(pdeath_signal)));
1867 case LINUX_PR_GET_KEEPCAPS:
1869 * Indicate that we always clear the effective and
1870 * permitted capability sets when the user id becomes
1871 * non-zero (actually the capability sets are simply
1872 * always zero in the current implementation).
1874 td->td_retval[0] = 0;
1876 case LINUX_PR_SET_KEEPCAPS:
1878 * Ignore requests to keep the effective and permitted
1879 * capability sets when the user id becomes non-zero.
1882 case LINUX_PR_SET_NAME:
1884 * To be on the safe side we need to make sure to not
1885 * overflow the size a Linux program expects. We already
1886 * do this here in the copyin, so that we don't need to
1889 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1890 error = copyinstr((void *)(register_t)args->arg2, comm,
1893 /* Linux silently truncates the name if it is too long. */
1894 if (error == ENAMETOOLONG) {
1896 * XXX: copyinstr() isn't documented to populate the
1897 * array completely, so do a copyin() to be on the
1898 * safe side. This should be changed in case
1899 * copyinstr() is changed to guarantee this.
1901 error = copyin((void *)(register_t)args->arg2, comm,
1903 comm[max_size - 1] = '\0';
1909 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1912 case LINUX_PR_GET_NAME:
1914 strlcpy(comm, p->p_comm, sizeof(comm));
1916 error = copyout(comm, (void *)(register_t)args->arg2,
1928 linux_sched_setparam(struct thread *td,
1929 struct linux_sched_setparam_args *uap)
1931 struct sched_param sched_param;
1935 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1939 tdt = linux_tdfind(td, uap->pid, -1);
1943 if (linux_map_sched_prio) {
1944 error = kern_sched_getscheduler(td, tdt, &policy);
1950 if (sched_param.sched_priority != 0) {
1954 sched_param.sched_priority =
1955 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1959 if (sched_param.sched_priority < 1 ||
1960 sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1965 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1966 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1968 sched_param.sched_priority =
1969 (sched_param.sched_priority - 1) *
1970 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1971 (LINUX_MAX_RT_PRIO - 1);
1976 error = kern_sched_setparam(td, tdt, &sched_param);
1977 out: PROC_UNLOCK(tdt->td_proc);
1982 linux_sched_getparam(struct thread *td,
1983 struct linux_sched_getparam_args *uap)
1985 struct sched_param sched_param;
1989 tdt = linux_tdfind(td, uap->pid, -1);
1993 error = kern_sched_getparam(td, tdt, &sched_param);
1995 PROC_UNLOCK(tdt->td_proc);
1999 if (linux_map_sched_prio) {
2000 error = kern_sched_getscheduler(td, tdt, &policy);
2001 PROC_UNLOCK(tdt->td_proc);
2007 sched_param.sched_priority = 0;
2012 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2013 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2015 sched_param.sched_priority =
2016 (sched_param.sched_priority *
2017 (LINUX_MAX_RT_PRIO - 1) +
2018 (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2019 (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2023 PROC_UNLOCK(tdt->td_proc);
2025 error = copyout(&sched_param, uap->param, sizeof(sched_param));
2030 * Get affinity of a process.
2033 linux_sched_getaffinity(struct thread *td,
2034 struct linux_sched_getaffinity_args *args)
2039 if (args->len < sizeof(cpuset_t))
2042 tdt = linux_tdfind(td, args->pid, -1);
2046 PROC_UNLOCK(tdt->td_proc);
2048 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2049 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2051 td->td_retval[0] = sizeof(cpuset_t);
2057 * Set affinity of a process.
2060 linux_sched_setaffinity(struct thread *td,
2061 struct linux_sched_setaffinity_args *args)
2065 if (args->len < sizeof(cpuset_t))
2068 tdt = linux_tdfind(td, args->pid, -1);
2072 PROC_UNLOCK(tdt->td_proc);
2074 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2075 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2078 struct linux_rlimit64 {
2084 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2086 struct rlimit rlim, nrlim;
2087 struct linux_rlimit64 lrlim;
2093 if (args->resource >= LINUX_RLIM_NLIMITS)
2096 which = linux_to_bsd_resource[args->resource];
2100 if (args->new != NULL) {
2102 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2103 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2104 * as INFINITY so we do not need a conversion even.
2106 error = copyin(args->new, &nrlim, sizeof(nrlim));
2111 flags = PGET_HOLD | PGET_NOTWEXIT;
2112 if (args->new != NULL)
2113 flags |= PGET_CANDEBUG;
2115 flags |= PGET_CANSEE;
2116 if (args->pid == 0) {
2120 error = pget(args->pid, flags, &p);
2124 if (args->old != NULL) {
2126 lim_rlimit_proc(p, which, &rlim);
2128 if (rlim.rlim_cur == RLIM_INFINITY)
2129 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2131 lrlim.rlim_cur = rlim.rlim_cur;
2132 if (rlim.rlim_max == RLIM_INFINITY)
2133 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2135 lrlim.rlim_max = rlim.rlim_max;
2136 error = copyout(&lrlim, args->old, sizeof(lrlim));
2141 if (args->new != NULL)
2142 error = kern_proc_setrlimit(td, p, which, &nrlim);
2150 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2152 struct timeval utv, tv0, tv1, *tvp;
2153 struct l_pselect6arg lpse6;
2154 struct l_timespec lts;
2155 struct timespec uts;
2162 if (args->sig != NULL) {
2163 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2166 if (lpse6.ss_len != sizeof(l_ss))
2168 if (lpse6.ss != 0) {
2169 error = copyin(PTRIN(lpse6.ss), &l_ss,
2173 linux_to_bsd_sigset(&l_ss, &ss);
2179 * Currently glibc changes nanosecond number to microsecond.
2180 * This mean losing precision but for now it is hardly seen.
2182 if (args->tsp != NULL) {
2183 error = copyin(args->tsp, <s, sizeof(lts));
2186 error = linux_to_native_timespec(&uts, <s);
2190 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2191 if (itimerfix(&utv))
2199 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2200 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2202 if (error == 0 && args->tsp != NULL) {
2203 if (td->td_retval[0] != 0) {
2205 * Compute how much time was left of the timeout,
2206 * by subtracting the current time and the time
2207 * before we started the call, and subtracting
2208 * that result from the user-supplied value.
2212 timevalsub(&tv1, &tv0);
2213 timevalsub(&utv, &tv1);
2219 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2221 error = native_to_linux_timespec(<s, &uts);
2223 error = copyout(<s, args->tsp, sizeof(lts));
2230 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2232 struct timespec ts0, ts1;
2233 struct l_timespec lts;
2234 struct timespec uts, *tsp;
2240 if (args->sset != NULL) {
2241 if (args->ssize != sizeof(l_ss))
2243 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2246 linux_to_bsd_sigset(&l_ss, &ss);
2250 if (args->tsp != NULL) {
2251 error = copyin(args->tsp, <s, sizeof(lts));
2254 error = linux_to_native_timespec(&uts, <s);
2263 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2265 if (error == 0 && args->tsp != NULL) {
2266 if (td->td_retval[0]) {
2268 timespecsub(&ts1, &ts0, &ts1);
2269 timespecsub(&uts, &ts1, &uts);
2271 timespecclear(&uts);
2273 timespecclear(&uts);
2275 error = native_to_linux_timespec(<s, &uts);
2277 error = copyout(<s, args->tsp, sizeof(lts));
2284 linux_sched_rr_get_interval(struct thread *td,
2285 struct linux_sched_rr_get_interval_args *uap)
2288 struct l_timespec lts;
2293 * According to man in case the invalid pid specified
2294 * EINVAL should be returned.
2299 tdt = linux_tdfind(td, uap->pid, -1);
2303 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2304 PROC_UNLOCK(tdt->td_proc);
2307 error = native_to_linux_timespec(<s, &ts);
2310 return (copyout(<s, uap->interval, sizeof(lts)));
2314 * In case when the Linux thread is the initial thread in
2315 * the thread group thread id is equal to the process id.
2316 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2319 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2321 struct linux_emuldata *em;
2326 if (tid == 0 || tid == td->td_tid) {
2328 PROC_LOCK(tdt->td_proc);
2329 } else if (tid > PID_MAX)
2330 tdt = tdfind(tid, pid);
2333 * Initial thread where the tid equal to the pid.
2337 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2339 * p is not a Linuxulator process.
2344 FOREACH_THREAD_IN_PROC(p, tdt) {
2346 if (tid == em->em_tid)
2358 linux_to_bsd_waitopts(int options, int *bsdopts)
2361 if (options & LINUX_WNOHANG)
2362 *bsdopts |= WNOHANG;
2363 if (options & LINUX_WUNTRACED)
2364 *bsdopts |= WUNTRACED;
2365 if (options & LINUX_WEXITED)
2366 *bsdopts |= WEXITED;
2367 if (options & LINUX_WCONTINUED)
2368 *bsdopts |= WCONTINUED;
2369 if (options & LINUX_WNOWAIT)
2370 *bsdopts |= WNOWAIT;
2372 if (options & __WCLONE)
2373 *bsdopts |= WLINUXCLONE;
2377 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2383 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2385 if (args->count > INT_MAX)
2386 args->count = INT_MAX;
2388 iov.iov_base = args->buf;
2389 iov.iov_len = args->count;
2393 uio.uio_resid = iov.iov_len;
2394 uio.uio_segflg = UIO_USERSPACE;
2395 uio.uio_rw = UIO_READ;
2398 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2400 td->td_retval[0] = args->count - uio.uio_resid;
2405 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2408 /* Needs to be page-aligned */
2409 if (args->start & PAGE_MASK)
2411 return (kern_mincore(td, args->start, args->len, args->vec));
2414 #define SYSLOG_TAG "<6>"
2417 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2419 char buf[128], *src, *dst;
2423 if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2424 linux_msg(td, "syslog unsupported type 0x%x", args->type);
2428 if (args->len < 6) {
2429 td->td_retval[0] = 0;
2433 error = priv_check(td, PRIV_MSGBUF);
2437 mtx_lock(&msgbuf_lock);
2438 msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2439 mtx_unlock(&msgbuf_lock);
2442 error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2443 /* The -1 is to skip the trailing '\0'. */
2444 dst += sizeof(SYSLOG_TAG) - 1;
2446 while (error == 0) {
2447 mtx_lock(&msgbuf_lock);
2448 buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2449 mtx_unlock(&msgbuf_lock);
2454 for (src = buf; src < buf + buflen && error == 0; src++) {
2458 if (dst >= args->buf + args->len)
2461 error = copyout(src, dst, 1);
2464 if (*src == '\n' && *(src + 1) != '<' &&
2465 dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2466 error = copyout(&SYSLOG_TAG,
2467 dst, sizeof(SYSLOG_TAG));
2468 dst += sizeof(SYSLOG_TAG) - 1;
2473 td->td_retval[0] = dst - args->buf;
2478 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2480 int cpu, error, node;
2482 cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2484 node = cpuid_to_pcpu[cpu]->pc_domain;
2486 if (args->cpu != NULL)
2487 error = copyout(&cpu, args->cpu, sizeof(l_int));
2488 if (args->node != NULL)
2489 error = copyout(&node, args->node, sizeof(l_int));