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/vm_object.h>
83 #include <vm/swap_pager.h>
86 #include <machine/../linux32/linux.h>
87 #include <machine/../linux32/linux32_proto.h>
89 #include <machine/../linux/linux.h>
90 #include <machine/../linux/linux_proto.h>
93 #include <compat/linux/linux_dtrace.h>
94 #include <compat/linux/linux_file.h>
95 #include <compat/linux/linux_mib.h>
96 #include <compat/linux/linux_signal.h>
97 #include <compat/linux/linux_timer.h>
98 #include <compat/linux/linux_util.h>
99 #include <compat/linux/linux_sysproto.h>
100 #include <compat/linux/linux_emul.h>
101 #include <compat/linux/linux_misc.h>
104 * Special DTrace provider for the linuxulator.
106 * In this file we define the provider for the entire linuxulator. All
107 * modules (= files of the linuxulator) use it.
109 * We define a different name depending on the emulated bitsize, see
110 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
111 * native bitsize = linuxulator
112 * amd64, 32bit emulation = linuxulator32
114 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
116 int stclohz; /* Statistics clock frequency */
118 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
119 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
120 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
121 RLIMIT_MEMLOCK, RLIMIT_AS
125 l_long uptime; /* Seconds since boot */
126 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
127 #define LINUX_SYSINFO_LOADS_SCALE 65536
128 l_ulong totalram; /* Total usable main memory size */
129 l_ulong freeram; /* Available memory size */
130 l_ulong sharedram; /* Amount of shared memory */
131 l_ulong bufferram; /* Memory used by buffers */
132 l_ulong totalswap; /* Total swap space size */
133 l_ulong freeswap; /* swap space still available */
134 l_ushort procs; /* Number of current processes */
139 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
142 struct l_pselect6arg {
147 static bool map_sched_prio = true;
148 SYSCTL_BOOL(_compat_linux, OID_AUTO, map_sched_prio, CTLFLAG_RDTUN,
149 &map_sched_prio, 0, "Map scheduler priorities to Linux priorities "
150 "(not POSIX compliant)");
152 static int linux_utimensat_nsec_valid(l_long);
156 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
158 struct l_sysinfo sysinfo;
163 bzero(&sysinfo, sizeof(sysinfo));
167 sysinfo.uptime = ts.tv_sec;
169 /* Use the information from the mib to get our load averages */
170 for (i = 0; i < 3; i++)
171 sysinfo.loads[i] = averunnable.ldavg[i] *
172 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
174 sysinfo.totalram = physmem * PAGE_SIZE;
175 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
177 sysinfo.sharedram = 0;
178 mtx_lock(&vm_object_list_mtx);
179 TAILQ_FOREACH(object, &vm_object_list, object_list)
180 if (object->shadow_count > 1)
181 sysinfo.sharedram += object->resident_page_count;
182 mtx_unlock(&vm_object_list_mtx);
184 sysinfo.sharedram *= PAGE_SIZE;
185 sysinfo.bufferram = 0;
187 swap_pager_status(&i, &j);
188 sysinfo.totalswap = i * PAGE_SIZE;
189 sysinfo.freeswap = (i - j) * PAGE_SIZE;
191 sysinfo.procs = nprocs;
193 /* The following are only present in newer Linux kernels. */
194 sysinfo.totalbig = 0;
196 sysinfo.mem_unit = 1;
198 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
201 #ifdef LINUX_LEGACY_SYSCALLS
203 linux_alarm(struct thread *td, struct linux_alarm_args *args)
205 struct itimerval it, old_it;
211 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
212 * to match kern_setitimer()'s limit to avoid error from it.
214 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
217 if (secs > INT32_MAX / 2)
218 secs = INT32_MAX / 2;
220 it.it_value.tv_sec = secs;
221 it.it_value.tv_usec = 0;
222 timevalclear(&it.it_interval);
223 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
224 KASSERT(error == 0, ("kern_setitimer returns %d", error));
226 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
227 old_it.it_value.tv_usec >= 500000)
228 old_it.it_value.tv_sec++;
229 td->td_retval[0] = old_it.it_value.tv_sec;
235 linux_brk(struct thread *td, struct linux_brk_args *args)
237 struct vmspace *vm = td->td_proc->p_vmspace;
240 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
241 new = (uintptr_t)args->dsend;
242 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
243 td->td_retval[0] = (register_t)new;
245 td->td_retval[0] = (register_t)old;
250 #if defined(__i386__)
251 /* XXX: what about amd64/linux32? */
254 linux_uselib(struct thread *td, struct linux_uselib_args *args)
260 vm_map_entry_t entry;
263 unsigned long file_offset;
264 unsigned long bss_size;
268 bool locked, opened, textset;
270 LCONVPATHEXIST(td, args->library, &library);
278 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
279 UIO_SYSSPACE, library, td);
286 NDFREE(&ni, NDF_ONLY_PNBUF);
289 * From here on down, we have a locked vnode that must be unlocked.
290 * XXX: The code below largely duplicates exec_check_permissions().
295 error = VOP_GETATTR(vp, &attr, td->td_ucred);
299 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
300 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
301 /* EACCESS is what exec(2) returns. */
307 if (attr.va_size == 0) {
312 /* Can we access it? */
313 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
318 * XXX: This should use vn_open() so that it is properly authorized,
319 * and to reduce code redundancy all over the place here.
320 * XXX: Not really, it duplicates far more of exec_check_permissions()
324 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
328 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
333 /* Pull in executable header into exec_map */
334 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
335 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
339 /* Is it a Linux binary ? */
340 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
346 * While we are here, we should REALLY do some more checks
349 /* Set file/virtual offset based on a.out variant. */
350 switch ((int)(a_out->a_magic & 0xffff)) {
351 case 0413: /* ZMAGIC */
354 case 0314: /* QMAGIC */
362 bss_size = round_page(a_out->a_bss);
364 /* Check various fields in header for validity/bounds. */
365 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
370 /* text + data can't exceed file size */
371 if (a_out->a_data + a_out->a_text > attr.va_size) {
377 * text/data/bss must not exceed limits
378 * XXX - this is not complete. it should check current usage PLUS
379 * the resources needed by this library.
381 PROC_LOCK(td->td_proc);
382 if (a_out->a_text > maxtsiz ||
383 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
384 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
386 PROC_UNLOCK(td->td_proc);
390 PROC_UNLOCK(td->td_proc);
393 * Prevent more writers.
395 error = VOP_SET_TEXT(vp);
401 * Lock no longer needed
407 * Check if file_offset page aligned. Currently we cannot handle
408 * misalinged file offsets, and so we read in the entire image
411 if (file_offset & PAGE_MASK) {
412 /* Map text+data read/write/execute */
414 /* a_entry is the load address and is page aligned */
415 vmaddr = trunc_page(a_out->a_entry);
417 /* get anon user mapping, read+write+execute */
418 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
419 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
420 VM_PROT_ALL, VM_PROT_ALL, 0);
424 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
425 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
426 td->td_ucred, NOCRED, &aresid, td);
435 * for QMAGIC, a_entry is 20 bytes beyond the load address
436 * to skip the executable header
438 vmaddr = trunc_page(a_out->a_entry);
441 * Map it all into the process's space as a single
442 * copy-on-write "data" segment.
444 map = &td->td_proc->p_vmspace->vm_map;
445 error = vm_mmap(map, &vmaddr,
446 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
447 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
451 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
456 entry->eflags |= MAP_ENTRY_VN_EXEC;
462 /* Calculate BSS start address */
463 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
466 /* allocate some 'anon' space */
467 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
468 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
479 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
484 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
486 VOP_UNSET_TEXT_CHECKED(vp);
491 /* Release the temporary mapping. */
493 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
498 #endif /* __i386__ */
500 #ifdef LINUX_LEGACY_SYSCALLS
502 linux_select(struct thread *td, struct linux_select_args *args)
505 struct timeval tv0, tv1, utv, *tvp;
509 * Store current time for computation of the amount of
513 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
515 utv.tv_sec = ltv.tv_sec;
516 utv.tv_usec = ltv.tv_usec;
518 if (itimerfix(&utv)) {
520 * The timeval was invalid. Convert it to something
521 * valid that will act as it does under Linux.
523 utv.tv_sec += utv.tv_usec / 1000000;
524 utv.tv_usec %= 1000000;
525 if (utv.tv_usec < 0) {
527 utv.tv_usec += 1000000;
537 error = kern_select(td, args->nfds, args->readfds, args->writefds,
538 args->exceptfds, tvp, LINUX_NFDBITS);
543 if (td->td_retval[0]) {
545 * Compute how much time was left of the timeout,
546 * by subtracting the current time and the time
547 * before we started the call, and subtracting
548 * that result from the user-supplied value.
551 timevalsub(&tv1, &tv0);
552 timevalsub(&utv, &tv1);
557 ltv.tv_sec = utv.tv_sec;
558 ltv.tv_usec = utv.tv_usec;
559 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
569 linux_mremap(struct thread *td, struct linux_mremap_args *args)
575 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
576 td->td_retval[0] = 0;
581 * Check for the page alignment.
582 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
584 if (args->addr & PAGE_MASK) {
585 td->td_retval[0] = 0;
589 args->new_len = round_page(args->new_len);
590 args->old_len = round_page(args->old_len);
592 if (args->new_len > args->old_len) {
593 td->td_retval[0] = 0;
597 if (args->new_len < args->old_len) {
598 addr = args->addr + args->new_len;
599 len = args->old_len - args->new_len;
600 error = kern_munmap(td, addr, len);
603 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
607 #define LINUX_MS_ASYNC 0x0001
608 #define LINUX_MS_INVALIDATE 0x0002
609 #define LINUX_MS_SYNC 0x0004
612 linux_msync(struct thread *td, struct linux_msync_args *args)
615 return (kern_msync(td, args->addr, args->len,
616 args->fl & ~LINUX_MS_SYNC));
619 #ifdef LINUX_LEGACY_SYSCALLS
621 linux_time(struct thread *td, struct linux_time_args *args)
629 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
631 td->td_retval[0] = tm;
636 struct l_times_argv {
639 l_clock_t tms_cutime;
640 l_clock_t tms_cstime;
645 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
646 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
647 * auxiliary vector entry.
651 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
652 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
654 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
655 CONVNTCK(r) : CONVOTCK(r))
658 linux_times(struct thread *td, struct linux_times_args *args)
660 struct timeval tv, utime, stime, cutime, cstime;
661 struct l_times_argv tms;
665 if (args->buf != NULL) {
669 calcru(p, &utime, &stime);
671 calccru(p, &cutime, &cstime);
674 tms.tms_utime = CONVTCK(utime);
675 tms.tms_stime = CONVTCK(stime);
677 tms.tms_cutime = CONVTCK(cutime);
678 tms.tms_cstime = CONVTCK(cstime);
680 if ((error = copyout(&tms, args->buf, sizeof(tms))))
685 td->td_retval[0] = (int)CONVTCK(tv);
690 linux_newuname(struct thread *td, struct linux_newuname_args *args)
692 struct l_new_utsname utsname;
693 char osname[LINUX_MAX_UTSNAME];
694 char osrelease[LINUX_MAX_UTSNAME];
697 linux_get_osname(td, osname);
698 linux_get_osrelease(td, osrelease);
700 bzero(&utsname, sizeof(utsname));
701 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
702 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
703 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
704 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
705 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
706 for (p = utsname.version; *p != '\0'; ++p)
711 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
713 return (copyout(&utsname, args->buf, sizeof(utsname)));
721 #ifdef LINUX_LEGACY_SYSCALLS
723 linux_utime(struct thread *td, struct linux_utime_args *args)
725 struct timeval tv[2], *tvp;
726 struct l_utimbuf lut;
730 LCONVPATHEXIST(td, args->fname, &fname);
733 if ((error = copyin(args->times, &lut, sizeof lut))) {
737 tv[0].tv_sec = lut.l_actime;
739 tv[1].tv_sec = lut.l_modtime;
745 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
752 #ifdef LINUX_LEGACY_SYSCALLS
754 linux_utimes(struct thread *td, struct linux_utimes_args *args)
757 struct timeval tv[2], *tvp = NULL;
761 LCONVPATHEXIST(td, args->fname, &fname);
763 if (args->tptr != NULL) {
764 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
768 tv[0].tv_sec = ltv[0].tv_sec;
769 tv[0].tv_usec = ltv[0].tv_usec;
770 tv[1].tv_sec = ltv[1].tv_sec;
771 tv[1].tv_usec = ltv[1].tv_usec;
775 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
783 linux_utimensat_nsec_valid(l_long nsec)
786 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
788 if (nsec >= 0 && nsec <= 999999999)
794 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
796 struct l_timespec l_times[2];
797 struct timespec times[2], *timesp = NULL;
799 int error, dfd, flags = 0;
801 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
803 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
806 if (args->times != NULL) {
807 error = copyin(args->times, l_times, sizeof(l_times));
811 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
812 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
815 times[0].tv_sec = l_times[0].tv_sec;
816 switch (l_times[0].tv_nsec)
818 case LINUX_UTIME_OMIT:
819 times[0].tv_nsec = UTIME_OMIT;
821 case LINUX_UTIME_NOW:
822 times[0].tv_nsec = UTIME_NOW;
825 times[0].tv_nsec = l_times[0].tv_nsec;
828 times[1].tv_sec = l_times[1].tv_sec;
829 switch (l_times[1].tv_nsec)
831 case LINUX_UTIME_OMIT:
832 times[1].tv_nsec = UTIME_OMIT;
834 case LINUX_UTIME_NOW:
835 times[1].tv_nsec = UTIME_NOW;
838 times[1].tv_nsec = l_times[1].tv_nsec;
843 /* This breaks POSIX, but is what the Linux kernel does
844 * _on purpose_ (documented in the man page for utimensat(2)),
845 * so we must follow that behaviour. */
846 if (times[0].tv_nsec == UTIME_OMIT &&
847 times[1].tv_nsec == UTIME_OMIT)
851 if (args->pathname != NULL)
852 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
853 else if (args->flags != 0)
856 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
857 flags |= AT_SYMLINK_NOFOLLOW;
860 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
862 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
863 UIO_SYSSPACE, flags);
870 #ifdef LINUX_LEGACY_SYSCALLS
872 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
875 struct timeval tv[2], *tvp = NULL;
879 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
880 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
882 if (args->utimes != NULL) {
883 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
887 tv[0].tv_sec = ltv[0].tv_sec;
888 tv[0].tv_usec = ltv[0].tv_usec;
889 tv[1].tv_sec = ltv[1].tv_sec;
890 tv[1].tv_usec = ltv[1].tv_usec;
894 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
901 linux_common_wait(struct thread *td, int pid, int *statusp,
902 int options, struct __wrusage *wrup)
907 int error, status, tmpstat;
909 if (pid == WAIT_ANY) {
912 } else if (pid < 0) {
921 * For backward compatibility we implicitly add flags WEXITED
924 options |= WEXITED | WTRAPPED;
925 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
930 tmpstat = status & 0xffff;
931 if (WIFSIGNALED(tmpstat)) {
932 tmpstat = (tmpstat & 0xffffff80) |
933 bsd_to_linux_signal(WTERMSIG(tmpstat));
934 } else if (WIFSTOPPED(tmpstat)) {
935 tmpstat = (tmpstat & 0xffff00ff) |
936 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
937 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
938 if (WSTOPSIG(status) == SIGTRAP) {
939 tmpstat = linux_ptrace_status(td,
940 siginfo.si_pid, tmpstat);
943 } else if (WIFCONTINUED(tmpstat)) {
946 error = copyout(&tmpstat, statusp, sizeof(int));
952 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
954 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
956 struct linux_wait4_args wait4_args;
958 wait4_args.pid = args->pid;
959 wait4_args.status = args->status;
960 wait4_args.options = args->options;
961 wait4_args.rusage = NULL;
963 return (linux_wait4(td, &wait4_args));
965 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
968 linux_wait4(struct thread *td, struct linux_wait4_args *args)
971 struct __wrusage wru, *wrup;
973 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
974 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
978 linux_to_bsd_waitopts(args->options, &options);
980 if (args->rusage != NULL)
984 error = linux_common_wait(td, args->pid, args->status, options, wrup);
987 if (args->rusage != NULL)
988 error = linux_copyout_rusage(&wru.wru_self, args->rusage);
993 linux_waitid(struct thread *td, struct linux_waitid_args *args)
995 int status, options, sig;
996 struct __wrusage wru;
1004 linux_to_bsd_waitopts(args->options, &options);
1006 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1008 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1011 switch (args->idtype) {
1029 error = kern_wait6(td, idtype, args->id, &status, options,
1033 if (args->rusage != NULL) {
1034 error = linux_copyout_rusage(&wru.wru_children,
1039 if (args->info != NULL) {
1041 bzero(&lsi, sizeof(lsi));
1042 if (td->td_retval[0] != 0) {
1043 sig = bsd_to_linux_signal(siginfo.si_signo);
1044 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1046 error = copyout(&lsi, args->info, sizeof(lsi));
1048 td->td_retval[0] = 0;
1053 #ifdef LINUX_LEGACY_SYSCALLS
1055 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1060 LCONVPATHCREAT(td, args->path, &path);
1062 switch (args->mode & S_IFMT) {
1065 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1071 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1072 args->mode, args->dev);
1080 args->mode |= S_IFREG;
1083 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1084 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1086 kern_close(td, td->td_retval[0]);
1099 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1104 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1105 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1107 switch (args->mode & S_IFMT) {
1110 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1115 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1124 args->mode |= S_IFREG;
1127 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1128 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1130 kern_close(td, td->td_retval[0]);
1142 * UGH! This is just about the dumbest idea I've ever heard!!
1145 linux_personality(struct thread *td, struct linux_personality_args *args)
1147 struct linux_pemuldata *pem;
1148 struct proc *p = td->td_proc;
1154 if (args->per != 0xffffffff)
1155 pem->persona = args->per;
1158 td->td_retval[0] = old;
1162 struct l_itimerval {
1163 l_timeval it_interval;
1167 #define B2L_ITIMERVAL(bip, lip) \
1168 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1169 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1170 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1171 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1174 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1177 struct l_itimerval ls;
1178 struct itimerval aitv, oitv;
1180 if (uap->itv == NULL) {
1181 uap->itv = uap->oitv;
1182 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1185 error = copyin(uap->itv, &ls, sizeof(ls));
1188 B2L_ITIMERVAL(&aitv, &ls);
1189 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1190 if (error != 0 || uap->oitv == NULL)
1192 B2L_ITIMERVAL(&ls, &oitv);
1194 return (copyout(&ls, uap->oitv, sizeof(ls)));
1198 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1201 struct l_itimerval ls;
1202 struct itimerval aitv;
1204 error = kern_getitimer(td, uap->which, &aitv);
1207 B2L_ITIMERVAL(&ls, &aitv);
1208 return (copyout(&ls, uap->itv, sizeof(ls)));
1211 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1213 linux_nice(struct thread *td, struct linux_nice_args *args)
1216 return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1218 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1221 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1223 struct ucred *newcred, *oldcred;
1224 l_gid_t *linux_gidset;
1229 ngrp = args->gidsetsize;
1230 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1232 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1233 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1237 crextend(newcred, ngrp + 1);
1240 oldcred = p->p_ucred;
1241 crcopy(newcred, oldcred);
1244 * cr_groups[0] holds egid. Setting the whole set from
1245 * the supplied set will cause egid to be changed too.
1246 * Keep cr_groups[0] unchanged to prevent that.
1249 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1256 newcred->cr_ngroups = ngrp + 1;
1258 bsd_gidset = newcred->cr_groups;
1261 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1265 newcred->cr_ngroups = 1;
1268 proc_set_cred(p, newcred);
1273 free(linux_gidset, M_LINUX);
1278 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1281 l_gid_t *linux_gidset;
1283 int bsd_gidsetsz, ngrp, error;
1285 cred = td->td_ucred;
1286 bsd_gidset = cred->cr_groups;
1287 bsd_gidsetsz = cred->cr_ngroups - 1;
1290 * cr_groups[0] holds egid. Returning the whole set
1291 * here will cause a duplicate. Exclude cr_groups[0]
1295 if ((ngrp = args->gidsetsize) == 0) {
1296 td->td_retval[0] = bsd_gidsetsz;
1300 if (ngrp < bsd_gidsetsz)
1304 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1306 while (ngrp < bsd_gidsetsz) {
1307 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1311 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1312 free(linux_gidset, M_LINUX);
1316 td->td_retval[0] = ngrp;
1321 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1323 struct rlimit bsd_rlim;
1324 struct l_rlimit rlim;
1328 if (args->resource >= LINUX_RLIM_NLIMITS)
1331 which = linux_to_bsd_resource[args->resource];
1335 error = copyin(args->rlim, &rlim, sizeof(rlim));
1339 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1340 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1341 return (kern_setrlimit(td, which, &bsd_rlim));
1344 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1346 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1348 struct l_rlimit rlim;
1349 struct rlimit bsd_rlim;
1352 if (args->resource >= LINUX_RLIM_NLIMITS)
1355 which = linux_to_bsd_resource[args->resource];
1359 lim_rlimit(td, which, &bsd_rlim);
1361 #ifdef COMPAT_LINUX32
1362 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1363 if (rlim.rlim_cur == UINT_MAX)
1364 rlim.rlim_cur = INT_MAX;
1365 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1366 if (rlim.rlim_max == UINT_MAX)
1367 rlim.rlim_max = INT_MAX;
1369 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1370 if (rlim.rlim_cur == ULONG_MAX)
1371 rlim.rlim_cur = LONG_MAX;
1372 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1373 if (rlim.rlim_max == ULONG_MAX)
1374 rlim.rlim_max = LONG_MAX;
1376 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1378 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1381 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1383 struct l_rlimit rlim;
1384 struct rlimit bsd_rlim;
1387 if (args->resource >= LINUX_RLIM_NLIMITS)
1390 which = linux_to_bsd_resource[args->resource];
1394 lim_rlimit(td, which, &bsd_rlim);
1396 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1397 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1398 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1402 linux_sched_setscheduler(struct thread *td,
1403 struct linux_sched_setscheduler_args *args)
1405 struct sched_param sched_param;
1409 switch (args->policy) {
1410 case LINUX_SCHED_OTHER:
1411 policy = SCHED_OTHER;
1413 case LINUX_SCHED_FIFO:
1414 policy = SCHED_FIFO;
1416 case LINUX_SCHED_RR:
1423 error = copyin(args->param, &sched_param, sizeof(sched_param));
1427 if (map_sched_prio) {
1430 if (sched_param.sched_priority != 0)
1433 sched_param.sched_priority =
1434 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1438 if (sched_param.sched_priority < 1 ||
1439 sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1443 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1444 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1446 sched_param.sched_priority =
1447 (sched_param.sched_priority - 1) *
1448 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1449 (LINUX_MAX_RT_PRIO - 1);
1454 tdt = linux_tdfind(td, args->pid, -1);
1458 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1459 PROC_UNLOCK(tdt->td_proc);
1464 linux_sched_getscheduler(struct thread *td,
1465 struct linux_sched_getscheduler_args *args)
1470 tdt = linux_tdfind(td, args->pid, -1);
1474 error = kern_sched_getscheduler(td, tdt, &policy);
1475 PROC_UNLOCK(tdt->td_proc);
1479 td->td_retval[0] = LINUX_SCHED_OTHER;
1482 td->td_retval[0] = LINUX_SCHED_FIFO;
1485 td->td_retval[0] = LINUX_SCHED_RR;
1492 linux_sched_get_priority_max(struct thread *td,
1493 struct linux_sched_get_priority_max_args *args)
1495 struct sched_get_priority_max_args bsd;
1497 if (map_sched_prio) {
1498 switch (args->policy) {
1499 case LINUX_SCHED_OTHER:
1500 td->td_retval[0] = 0;
1502 case LINUX_SCHED_FIFO:
1503 case LINUX_SCHED_RR:
1504 td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1511 switch (args->policy) {
1512 case LINUX_SCHED_OTHER:
1513 bsd.policy = SCHED_OTHER;
1515 case LINUX_SCHED_FIFO:
1516 bsd.policy = SCHED_FIFO;
1518 case LINUX_SCHED_RR:
1519 bsd.policy = SCHED_RR;
1524 return (sys_sched_get_priority_max(td, &bsd));
1528 linux_sched_get_priority_min(struct thread *td,
1529 struct linux_sched_get_priority_min_args *args)
1531 struct sched_get_priority_min_args bsd;
1533 if (map_sched_prio) {
1534 switch (args->policy) {
1535 case LINUX_SCHED_OTHER:
1536 td->td_retval[0] = 0;
1538 case LINUX_SCHED_FIFO:
1539 case LINUX_SCHED_RR:
1540 td->td_retval[0] = 1;
1547 switch (args->policy) {
1548 case LINUX_SCHED_OTHER:
1549 bsd.policy = SCHED_OTHER;
1551 case LINUX_SCHED_FIFO:
1552 bsd.policy = SCHED_FIFO;
1554 case LINUX_SCHED_RR:
1555 bsd.policy = SCHED_RR;
1560 return (sys_sched_get_priority_min(td, &bsd));
1563 #define REBOOT_CAD_ON 0x89abcdef
1564 #define REBOOT_CAD_OFF 0
1565 #define REBOOT_HALT 0xcdef0123
1566 #define REBOOT_RESTART 0x01234567
1567 #define REBOOT_RESTART2 0xA1B2C3D4
1568 #define REBOOT_POWEROFF 0x4321FEDC
1569 #define REBOOT_MAGIC1 0xfee1dead
1570 #define REBOOT_MAGIC2 0x28121969
1571 #define REBOOT_MAGIC2A 0x05121996
1572 #define REBOOT_MAGIC2B 0x16041998
1575 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1577 struct reboot_args bsd_args;
1579 if (args->magic1 != REBOOT_MAGIC1)
1582 switch (args->magic2) {
1584 case REBOOT_MAGIC2A:
1585 case REBOOT_MAGIC2B:
1591 switch (args->cmd) {
1593 case REBOOT_CAD_OFF:
1594 return (priv_check(td, PRIV_REBOOT));
1596 bsd_args.opt = RB_HALT;
1598 case REBOOT_RESTART:
1599 case REBOOT_RESTART2:
1602 case REBOOT_POWEROFF:
1603 bsd_args.opt = RB_POWEROFF;
1608 return (sys_reboot(td, &bsd_args));
1613 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1616 td->td_retval[0] = td->td_proc->p_pid;
1622 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1624 struct linux_emuldata *em;
1627 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1629 td->td_retval[0] = em->em_tid;
1636 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1639 td->td_retval[0] = kern_getppid(td);
1644 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1647 td->td_retval[0] = td->td_ucred->cr_rgid;
1652 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1655 td->td_retval[0] = td->td_ucred->cr_ruid;
1660 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1663 return (kern_getsid(td, args->pid));
1667 linux_nosys(struct thread *td, struct nosys_args *ignore)
1674 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1678 error = kern_getpriority(td, args->which, args->who);
1679 td->td_retval[0] = 20 - td->td_retval[0];
1684 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1689 name[1] = KERN_HOSTNAME;
1690 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1695 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1700 name[1] = KERN_NISDOMAINNAME;
1701 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1706 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1709 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1713 * XXX: we should send a signal to the parent if
1714 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1715 * as it doesnt occur often.
1717 exit1(td, args->error_code, 0);
1721 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1722 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1723 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1725 struct l_user_cap_header {
1730 struct l_user_cap_data {
1737 linux_capget(struct thread *td, struct linux_capget_args *uap)
1739 struct l_user_cap_header luch;
1740 struct l_user_cap_data lucd[2];
1743 if (uap->hdrp == NULL)
1746 error = copyin(uap->hdrp, &luch, sizeof(luch));
1750 switch (luch.version) {
1751 case _LINUX_CAPABILITY_VERSION_1:
1754 case _LINUX_CAPABILITY_VERSION_2:
1755 case _LINUX_CAPABILITY_VERSION_3:
1759 luch.version = _LINUX_CAPABILITY_VERSION_1;
1760 error = copyout(&luch, uap->hdrp, sizeof(luch));
1771 * The current implementation doesn't support setting
1772 * a capability (it's essentially a stub) so indicate
1773 * that no capabilities are currently set or available
1776 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1777 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1784 linux_capset(struct thread *td, struct linux_capset_args *uap)
1786 struct l_user_cap_header luch;
1787 struct l_user_cap_data lucd[2];
1790 if (uap->hdrp == NULL || uap->datap == NULL)
1793 error = copyin(uap->hdrp, &luch, sizeof(luch));
1797 switch (luch.version) {
1798 case _LINUX_CAPABILITY_VERSION_1:
1801 case _LINUX_CAPABILITY_VERSION_2:
1802 case _LINUX_CAPABILITY_VERSION_3:
1806 luch.version = _LINUX_CAPABILITY_VERSION_1;
1807 error = copyout(&luch, uap->hdrp, sizeof(luch));
1816 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1820 /* We currently don't support setting any capabilities. */
1821 for (i = 0; i < u32s; i++) {
1822 if (lucd[i].effective || lucd[i].permitted ||
1823 lucd[i].inheritable) {
1825 "capset[%d] effective=0x%x, permitted=0x%x, "
1826 "inheritable=0x%x is not implemented", i,
1827 (int)lucd[i].effective, (int)lucd[i].permitted,
1828 (int)lucd[i].inheritable);
1837 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1839 int error = 0, max_size;
1840 struct proc *p = td->td_proc;
1841 char comm[LINUX_MAX_COMM_LEN];
1844 switch (args->option) {
1845 case LINUX_PR_SET_PDEATHSIG:
1846 if (!LINUX_SIG_VALID(args->arg2))
1848 pdeath_signal = linux_to_bsd_signal(args->arg2);
1849 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1851 case LINUX_PR_GET_PDEATHSIG:
1852 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1856 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1857 return (copyout(&pdeath_signal,
1858 (void *)(register_t)args->arg2,
1859 sizeof(pdeath_signal)));
1861 case LINUX_PR_GET_KEEPCAPS:
1863 * Indicate that we always clear the effective and
1864 * permitted capability sets when the user id becomes
1865 * non-zero (actually the capability sets are simply
1866 * always zero in the current implementation).
1868 td->td_retval[0] = 0;
1870 case LINUX_PR_SET_KEEPCAPS:
1872 * Ignore requests to keep the effective and permitted
1873 * capability sets when the user id becomes non-zero.
1876 case LINUX_PR_SET_NAME:
1878 * To be on the safe side we need to make sure to not
1879 * overflow the size a Linux program expects. We already
1880 * do this here in the copyin, so that we don't need to
1883 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1884 error = copyinstr((void *)(register_t)args->arg2, comm,
1887 /* Linux silently truncates the name if it is too long. */
1888 if (error == ENAMETOOLONG) {
1890 * XXX: copyinstr() isn't documented to populate the
1891 * array completely, so do a copyin() to be on the
1892 * safe side. This should be changed in case
1893 * copyinstr() is changed to guarantee this.
1895 error = copyin((void *)(register_t)args->arg2, comm,
1897 comm[max_size - 1] = '\0';
1903 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1906 case LINUX_PR_GET_NAME:
1908 strlcpy(comm, p->p_comm, sizeof(comm));
1910 error = copyout(comm, (void *)(register_t)args->arg2,
1922 linux_sched_setparam(struct thread *td,
1923 struct linux_sched_setparam_args *uap)
1925 struct sched_param sched_param;
1929 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1933 tdt = linux_tdfind(td, uap->pid, -1);
1937 if( map_sched_prio ) {
1938 error = kern_sched_getscheduler(td, tdt, &policy);
1944 if (sched_param.sched_priority != 0) {
1948 sched_param.sched_priority =
1949 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1953 if (sched_param.sched_priority < 1 ||
1954 sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1959 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1960 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1962 sched_param.sched_priority =
1963 (sched_param.sched_priority - 1) *
1964 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1965 (LINUX_MAX_RT_PRIO - 1);
1970 error = kern_sched_setparam(td, tdt, &sched_param);
1971 out: PROC_UNLOCK(tdt->td_proc);
1976 linux_sched_getparam(struct thread *td,
1977 struct linux_sched_getparam_args *uap)
1979 struct sched_param sched_param;
1983 tdt = linux_tdfind(td, uap->pid, -1);
1987 error = kern_sched_getparam(td, tdt, &sched_param);
1989 PROC_UNLOCK(tdt->td_proc);
1993 if (map_sched_prio) {
1994 error = kern_sched_getscheduler(td, tdt, &policy);
1995 PROC_UNLOCK(tdt->td_proc);
2001 sched_param.sched_priority = 0;
2006 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2007 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2009 sched_param.sched_priority =
2010 (sched_param.sched_priority *
2011 (LINUX_MAX_RT_PRIO - 1) +
2012 (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2013 (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2017 PROC_UNLOCK(tdt->td_proc);
2019 error = copyout(&sched_param, uap->param, sizeof(sched_param));
2024 * Get affinity of a process.
2027 linux_sched_getaffinity(struct thread *td,
2028 struct linux_sched_getaffinity_args *args)
2033 if (args->len < sizeof(cpuset_t))
2036 tdt = linux_tdfind(td, args->pid, -1);
2040 PROC_UNLOCK(tdt->td_proc);
2042 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2043 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2045 td->td_retval[0] = sizeof(cpuset_t);
2051 * Set affinity of a process.
2054 linux_sched_setaffinity(struct thread *td,
2055 struct linux_sched_setaffinity_args *args)
2059 if (args->len < sizeof(cpuset_t))
2062 tdt = linux_tdfind(td, args->pid, -1);
2066 PROC_UNLOCK(tdt->td_proc);
2068 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2069 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2072 struct linux_rlimit64 {
2078 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2080 struct rlimit rlim, nrlim;
2081 struct linux_rlimit64 lrlim;
2087 if (args->resource >= LINUX_RLIM_NLIMITS)
2090 which = linux_to_bsd_resource[args->resource];
2094 if (args->new != NULL) {
2096 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2097 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2098 * as INFINITY so we do not need a conversion even.
2100 error = copyin(args->new, &nrlim, sizeof(nrlim));
2105 flags = PGET_HOLD | PGET_NOTWEXIT;
2106 if (args->new != NULL)
2107 flags |= PGET_CANDEBUG;
2109 flags |= PGET_CANSEE;
2110 if (args->pid == 0) {
2114 error = pget(args->pid, flags, &p);
2118 if (args->old != NULL) {
2120 lim_rlimit_proc(p, which, &rlim);
2122 if (rlim.rlim_cur == RLIM_INFINITY)
2123 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2125 lrlim.rlim_cur = rlim.rlim_cur;
2126 if (rlim.rlim_max == RLIM_INFINITY)
2127 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2129 lrlim.rlim_max = rlim.rlim_max;
2130 error = copyout(&lrlim, args->old, sizeof(lrlim));
2135 if (args->new != NULL)
2136 error = kern_proc_setrlimit(td, p, which, &nrlim);
2144 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2146 struct timeval utv, tv0, tv1, *tvp;
2147 struct l_pselect6arg lpse6;
2148 struct l_timespec lts;
2149 struct timespec uts;
2156 if (args->sig != NULL) {
2157 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2160 if (lpse6.ss_len != sizeof(l_ss))
2162 if (lpse6.ss != 0) {
2163 error = copyin(PTRIN(lpse6.ss), &l_ss,
2167 linux_to_bsd_sigset(&l_ss, &ss);
2173 * Currently glibc changes nanosecond number to microsecond.
2174 * This mean losing precision but for now it is hardly seen.
2176 if (args->tsp != NULL) {
2177 error = copyin(args->tsp, <s, sizeof(lts));
2180 error = linux_to_native_timespec(&uts, <s);
2184 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2185 if (itimerfix(&utv))
2193 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2194 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2196 if (error == 0 && args->tsp != NULL) {
2197 if (td->td_retval[0] != 0) {
2199 * Compute how much time was left of the timeout,
2200 * by subtracting the current time and the time
2201 * before we started the call, and subtracting
2202 * that result from the user-supplied value.
2206 timevalsub(&tv1, &tv0);
2207 timevalsub(&utv, &tv1);
2213 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2215 error = native_to_linux_timespec(<s, &uts);
2217 error = copyout(<s, args->tsp, sizeof(lts));
2224 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2226 struct timespec ts0, ts1;
2227 struct l_timespec lts;
2228 struct timespec uts, *tsp;
2234 if (args->sset != NULL) {
2235 if (args->ssize != sizeof(l_ss))
2237 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2240 linux_to_bsd_sigset(&l_ss, &ss);
2244 if (args->tsp != NULL) {
2245 error = copyin(args->tsp, <s, sizeof(lts));
2248 error = linux_to_native_timespec(&uts, <s);
2257 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2259 if (error == 0 && args->tsp != NULL) {
2260 if (td->td_retval[0]) {
2262 timespecsub(&ts1, &ts0, &ts1);
2263 timespecsub(&uts, &ts1, &uts);
2265 timespecclear(&uts);
2267 timespecclear(&uts);
2269 error = native_to_linux_timespec(<s, &uts);
2271 error = copyout(<s, args->tsp, sizeof(lts));
2278 linux_sched_rr_get_interval(struct thread *td,
2279 struct linux_sched_rr_get_interval_args *uap)
2282 struct l_timespec lts;
2287 * According to man in case the invalid pid specified
2288 * EINVAL should be returned.
2293 tdt = linux_tdfind(td, uap->pid, -1);
2297 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2298 PROC_UNLOCK(tdt->td_proc);
2301 error = native_to_linux_timespec(<s, &ts);
2304 return (copyout(<s, uap->interval, sizeof(lts)));
2308 * In case when the Linux thread is the initial thread in
2309 * the thread group thread id is equal to the process id.
2310 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2313 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2315 struct linux_emuldata *em;
2320 if (tid == 0 || tid == td->td_tid) {
2322 PROC_LOCK(tdt->td_proc);
2323 } else if (tid > PID_MAX)
2324 tdt = tdfind(tid, pid);
2327 * Initial thread where the tid equal to the pid.
2331 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2333 * p is not a Linuxulator process.
2338 FOREACH_THREAD_IN_PROC(p, tdt) {
2340 if (tid == em->em_tid)
2352 linux_to_bsd_waitopts(int options, int *bsdopts)
2355 if (options & LINUX_WNOHANG)
2356 *bsdopts |= WNOHANG;
2357 if (options & LINUX_WUNTRACED)
2358 *bsdopts |= WUNTRACED;
2359 if (options & LINUX_WEXITED)
2360 *bsdopts |= WEXITED;
2361 if (options & LINUX_WCONTINUED)
2362 *bsdopts |= WCONTINUED;
2363 if (options & LINUX_WNOWAIT)
2364 *bsdopts |= WNOWAIT;
2366 if (options & __WCLONE)
2367 *bsdopts |= WLINUXCLONE;
2371 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2377 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2379 if (args->count > INT_MAX)
2380 args->count = INT_MAX;
2382 iov.iov_base = args->buf;
2383 iov.iov_len = args->count;
2387 uio.uio_resid = iov.iov_len;
2388 uio.uio_segflg = UIO_USERSPACE;
2389 uio.uio_rw = UIO_READ;
2392 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2394 td->td_retval[0] = args->count - uio.uio_resid;
2399 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2402 /* Needs to be page-aligned */
2403 if (args->start & PAGE_MASK)
2405 return (kern_mincore(td, args->start, args->len, args->vec));
2408 #define SYSLOG_TAG "<6>"
2411 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2413 char buf[128], *src, *dst;
2417 if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2418 linux_msg(td, "syslog unsupported type 0x%x", args->type);
2422 if (args->len < 6) {
2423 td->td_retval[0] = 0;
2427 error = priv_check(td, PRIV_MSGBUF);
2431 mtx_lock(&msgbuf_lock);
2432 msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2433 mtx_unlock(&msgbuf_lock);
2436 error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2437 /* The -1 is to skip the trailing '\0'. */
2438 dst += sizeof(SYSLOG_TAG) - 1;
2440 while (error == 0) {
2441 mtx_lock(&msgbuf_lock);
2442 buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2443 mtx_unlock(&msgbuf_lock);
2448 for (src = buf; src < buf + buflen && error == 0; src++) {
2452 if (dst >= args->buf + args->len)
2455 error = copyout(src, dst, 1);
2458 if (*src == '\n' && *(src + 1) != '<' &&
2459 dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2460 error = copyout(&SYSLOG_TAG,
2461 dst, sizeof(SYSLOG_TAG));
2462 dst += sizeof(SYSLOG_TAG) - 1;
2467 td->td_retval[0] = dst - args->buf;
2472 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2474 int cpu, error, node;
2476 cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2478 node = cpuid_to_pcpu[cpu]->pc_domain;
2480 if (args->cpu != NULL)
2481 error = copyout(&cpu, args->cpu, sizeof(l_int));
2482 if (args->node != NULL)
2483 error = copyout(&node, args->node, sizeof(l_int));