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 int linux_utimensat_nsec_valid(l_long);
151 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
153 struct l_sysinfo sysinfo;
158 bzero(&sysinfo, sizeof(sysinfo));
162 sysinfo.uptime = ts.tv_sec;
164 /* Use the information from the mib to get our load averages */
165 for (i = 0; i < 3; i++)
166 sysinfo.loads[i] = averunnable.ldavg[i] *
167 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
169 sysinfo.totalram = physmem * PAGE_SIZE;
170 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
172 sysinfo.sharedram = 0;
173 mtx_lock(&vm_object_list_mtx);
174 TAILQ_FOREACH(object, &vm_object_list, object_list)
175 if (object->shadow_count > 1)
176 sysinfo.sharedram += object->resident_page_count;
177 mtx_unlock(&vm_object_list_mtx);
179 sysinfo.sharedram *= PAGE_SIZE;
180 sysinfo.bufferram = 0;
182 swap_pager_status(&i, &j);
183 sysinfo.totalswap = i * PAGE_SIZE;
184 sysinfo.freeswap = (i - j) * PAGE_SIZE;
186 sysinfo.procs = nprocs;
188 /* The following are only present in newer Linux kernels. */
189 sysinfo.totalbig = 0;
191 sysinfo.mem_unit = 1;
193 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
196 #ifdef LINUX_LEGACY_SYSCALLS
198 linux_alarm(struct thread *td, struct linux_alarm_args *args)
200 struct itimerval it, old_it;
206 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
207 * to match kern_setitimer()'s limit to avoid error from it.
209 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
212 if (secs > INT32_MAX / 2)
213 secs = INT32_MAX / 2;
215 it.it_value.tv_sec = secs;
216 it.it_value.tv_usec = 0;
217 timevalclear(&it.it_interval);
218 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
219 KASSERT(error == 0, ("kern_setitimer returns %d", error));
221 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
222 old_it.it_value.tv_usec >= 500000)
223 old_it.it_value.tv_sec++;
224 td->td_retval[0] = old_it.it_value.tv_sec;
230 linux_brk(struct thread *td, struct linux_brk_args *args)
232 struct vmspace *vm = td->td_proc->p_vmspace;
235 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
236 new = (uintptr_t)args->dsend;
237 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
238 td->td_retval[0] = (register_t)new;
240 td->td_retval[0] = (register_t)old;
245 #if defined(__i386__)
246 /* XXX: what about amd64/linux32? */
249 linux_uselib(struct thread *td, struct linux_uselib_args *args)
255 vm_map_entry_t entry;
258 unsigned long file_offset;
259 unsigned long bss_size;
263 bool locked, opened, textset;
265 LCONVPATHEXIST(td, args->library, &library);
273 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
274 UIO_SYSSPACE, library, td);
281 NDFREE(&ni, NDF_ONLY_PNBUF);
284 * From here on down, we have a locked vnode that must be unlocked.
285 * XXX: The code below largely duplicates exec_check_permissions().
290 error = VOP_GETATTR(vp, &attr, td->td_ucred);
294 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
295 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
296 /* EACCESS is what exec(2) returns. */
302 if (attr.va_size == 0) {
307 /* Can we access it? */
308 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
313 * XXX: This should use vn_open() so that it is properly authorized,
314 * and to reduce code redundancy all over the place here.
315 * XXX: Not really, it duplicates far more of exec_check_permissions()
319 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
323 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
328 /* Pull in executable header into exec_map */
329 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
330 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
334 /* Is it a Linux binary ? */
335 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
341 * While we are here, we should REALLY do some more checks
344 /* Set file/virtual offset based on a.out variant. */
345 switch ((int)(a_out->a_magic & 0xffff)) {
346 case 0413: /* ZMAGIC */
349 case 0314: /* QMAGIC */
357 bss_size = round_page(a_out->a_bss);
359 /* Check various fields in header for validity/bounds. */
360 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
365 /* text + data can't exceed file size */
366 if (a_out->a_data + a_out->a_text > attr.va_size) {
372 * text/data/bss must not exceed limits
373 * XXX - this is not complete. it should check current usage PLUS
374 * the resources needed by this library.
376 PROC_LOCK(td->td_proc);
377 if (a_out->a_text > maxtsiz ||
378 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
379 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
381 PROC_UNLOCK(td->td_proc);
385 PROC_UNLOCK(td->td_proc);
388 * Prevent more writers.
390 error = VOP_SET_TEXT(vp);
396 * Lock no longer needed
402 * Check if file_offset page aligned. Currently we cannot handle
403 * misalinged file offsets, and so we read in the entire image
406 if (file_offset & PAGE_MASK) {
407 /* Map text+data read/write/execute */
409 /* a_entry is the load address and is page aligned */
410 vmaddr = trunc_page(a_out->a_entry);
412 /* get anon user mapping, read+write+execute */
413 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
414 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
415 VM_PROT_ALL, VM_PROT_ALL, 0);
419 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
420 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
421 td->td_ucred, NOCRED, &aresid, td);
430 * for QMAGIC, a_entry is 20 bytes beyond the load address
431 * to skip the executable header
433 vmaddr = trunc_page(a_out->a_entry);
436 * Map it all into the process's space as a single
437 * copy-on-write "data" segment.
439 map = &td->td_proc->p_vmspace->vm_map;
440 error = vm_mmap(map, &vmaddr,
441 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
442 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
446 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
451 entry->eflags |= MAP_ENTRY_VN_EXEC;
457 /* Calculate BSS start address */
458 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
461 /* allocate some 'anon' space */
462 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
463 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
474 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
479 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
481 VOP_UNSET_TEXT_CHECKED(vp);
486 /* Release the temporary mapping. */
488 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
493 #endif /* __i386__ */
495 #ifdef LINUX_LEGACY_SYSCALLS
497 linux_select(struct thread *td, struct linux_select_args *args)
500 struct timeval tv0, tv1, utv, *tvp;
504 * Store current time for computation of the amount of
508 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
510 utv.tv_sec = ltv.tv_sec;
511 utv.tv_usec = ltv.tv_usec;
513 if (itimerfix(&utv)) {
515 * The timeval was invalid. Convert it to something
516 * valid that will act as it does under Linux.
518 utv.tv_sec += utv.tv_usec / 1000000;
519 utv.tv_usec %= 1000000;
520 if (utv.tv_usec < 0) {
522 utv.tv_usec += 1000000;
532 error = kern_select(td, args->nfds, args->readfds, args->writefds,
533 args->exceptfds, tvp, LINUX_NFDBITS);
538 if (td->td_retval[0]) {
540 * Compute how much time was left of the timeout,
541 * by subtracting the current time and the time
542 * before we started the call, and subtracting
543 * that result from the user-supplied value.
546 timevalsub(&tv1, &tv0);
547 timevalsub(&utv, &tv1);
552 ltv.tv_sec = utv.tv_sec;
553 ltv.tv_usec = utv.tv_usec;
554 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
564 linux_mremap(struct thread *td, struct linux_mremap_args *args)
570 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
571 td->td_retval[0] = 0;
576 * Check for the page alignment.
577 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
579 if (args->addr & PAGE_MASK) {
580 td->td_retval[0] = 0;
584 args->new_len = round_page(args->new_len);
585 args->old_len = round_page(args->old_len);
587 if (args->new_len > args->old_len) {
588 td->td_retval[0] = 0;
592 if (args->new_len < args->old_len) {
593 addr = args->addr + args->new_len;
594 len = args->old_len - args->new_len;
595 error = kern_munmap(td, addr, len);
598 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
602 #define LINUX_MS_ASYNC 0x0001
603 #define LINUX_MS_INVALIDATE 0x0002
604 #define LINUX_MS_SYNC 0x0004
607 linux_msync(struct thread *td, struct linux_msync_args *args)
610 return (kern_msync(td, args->addr, args->len,
611 args->fl & ~LINUX_MS_SYNC));
614 #ifdef LINUX_LEGACY_SYSCALLS
616 linux_time(struct thread *td, struct linux_time_args *args)
624 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
626 td->td_retval[0] = tm;
631 struct l_times_argv {
634 l_clock_t tms_cutime;
635 l_clock_t tms_cstime;
640 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
641 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
642 * auxiliary vector entry.
646 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
647 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
649 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
650 CONVNTCK(r) : CONVOTCK(r))
653 linux_times(struct thread *td, struct linux_times_args *args)
655 struct timeval tv, utime, stime, cutime, cstime;
656 struct l_times_argv tms;
660 if (args->buf != NULL) {
664 calcru(p, &utime, &stime);
666 calccru(p, &cutime, &cstime);
669 tms.tms_utime = CONVTCK(utime);
670 tms.tms_stime = CONVTCK(stime);
672 tms.tms_cutime = CONVTCK(cutime);
673 tms.tms_cstime = CONVTCK(cstime);
675 if ((error = copyout(&tms, args->buf, sizeof(tms))))
680 td->td_retval[0] = (int)CONVTCK(tv);
685 linux_newuname(struct thread *td, struct linux_newuname_args *args)
687 struct l_new_utsname utsname;
688 char osname[LINUX_MAX_UTSNAME];
689 char osrelease[LINUX_MAX_UTSNAME];
692 linux_get_osname(td, osname);
693 linux_get_osrelease(td, osrelease);
695 bzero(&utsname, sizeof(utsname));
696 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
697 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
698 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
699 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
700 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
701 for (p = utsname.version; *p != '\0'; ++p)
706 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
708 return (copyout(&utsname, args->buf, sizeof(utsname)));
716 #ifdef LINUX_LEGACY_SYSCALLS
718 linux_utime(struct thread *td, struct linux_utime_args *args)
720 struct timeval tv[2], *tvp;
721 struct l_utimbuf lut;
725 LCONVPATHEXIST(td, args->fname, &fname);
728 if ((error = copyin(args->times, &lut, sizeof lut))) {
732 tv[0].tv_sec = lut.l_actime;
734 tv[1].tv_sec = lut.l_modtime;
740 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
747 #ifdef LINUX_LEGACY_SYSCALLS
749 linux_utimes(struct thread *td, struct linux_utimes_args *args)
752 struct timeval tv[2], *tvp = NULL;
756 LCONVPATHEXIST(td, args->fname, &fname);
758 if (args->tptr != NULL) {
759 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
763 tv[0].tv_sec = ltv[0].tv_sec;
764 tv[0].tv_usec = ltv[0].tv_usec;
765 tv[1].tv_sec = ltv[1].tv_sec;
766 tv[1].tv_usec = ltv[1].tv_usec;
770 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
778 linux_utimensat_nsec_valid(l_long nsec)
781 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
783 if (nsec >= 0 && nsec <= 999999999)
789 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
791 struct l_timespec l_times[2];
792 struct timespec times[2], *timesp = NULL;
794 int error, dfd, flags = 0;
796 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
798 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
801 if (args->times != NULL) {
802 error = copyin(args->times, l_times, sizeof(l_times));
806 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
807 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
810 times[0].tv_sec = l_times[0].tv_sec;
811 switch (l_times[0].tv_nsec)
813 case LINUX_UTIME_OMIT:
814 times[0].tv_nsec = UTIME_OMIT;
816 case LINUX_UTIME_NOW:
817 times[0].tv_nsec = UTIME_NOW;
820 times[0].tv_nsec = l_times[0].tv_nsec;
823 times[1].tv_sec = l_times[1].tv_sec;
824 switch (l_times[1].tv_nsec)
826 case LINUX_UTIME_OMIT:
827 times[1].tv_nsec = UTIME_OMIT;
829 case LINUX_UTIME_NOW:
830 times[1].tv_nsec = UTIME_NOW;
833 times[1].tv_nsec = l_times[1].tv_nsec;
838 /* This breaks POSIX, but is what the Linux kernel does
839 * _on purpose_ (documented in the man page for utimensat(2)),
840 * so we must follow that behaviour. */
841 if (times[0].tv_nsec == UTIME_OMIT &&
842 times[1].tv_nsec == UTIME_OMIT)
846 if (args->pathname != NULL)
847 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
848 else if (args->flags != 0)
851 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
852 flags |= AT_SYMLINK_NOFOLLOW;
855 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
857 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
858 UIO_SYSSPACE, flags);
865 #ifdef LINUX_LEGACY_SYSCALLS
867 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
870 struct timeval tv[2], *tvp = NULL;
874 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
875 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
877 if (args->utimes != NULL) {
878 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
882 tv[0].tv_sec = ltv[0].tv_sec;
883 tv[0].tv_usec = ltv[0].tv_usec;
884 tv[1].tv_sec = ltv[1].tv_sec;
885 tv[1].tv_usec = ltv[1].tv_usec;
889 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
896 linux_common_wait(struct thread *td, int pid, int *statusp,
897 int options, struct __wrusage *wrup)
902 int error, status, tmpstat;
904 if (pid == WAIT_ANY) {
907 } else if (pid < 0) {
916 * For backward compatibility we implicitly add flags WEXITED
919 options |= WEXITED | WTRAPPED;
920 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
925 tmpstat = status & 0xffff;
926 if (WIFSIGNALED(tmpstat)) {
927 tmpstat = (tmpstat & 0xffffff80) |
928 bsd_to_linux_signal(WTERMSIG(tmpstat));
929 } else if (WIFSTOPPED(tmpstat)) {
930 tmpstat = (tmpstat & 0xffff00ff) |
931 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
932 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
933 if (WSTOPSIG(status) == SIGTRAP) {
934 tmpstat = linux_ptrace_status(td,
935 siginfo.si_pid, tmpstat);
938 } else if (WIFCONTINUED(tmpstat)) {
941 error = copyout(&tmpstat, statusp, sizeof(int));
947 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
949 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
951 struct linux_wait4_args wait4_args;
953 wait4_args.pid = args->pid;
954 wait4_args.status = args->status;
955 wait4_args.options = args->options;
956 wait4_args.rusage = NULL;
958 return (linux_wait4(td, &wait4_args));
960 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
963 linux_wait4(struct thread *td, struct linux_wait4_args *args)
966 struct __wrusage wru, *wrup;
968 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
969 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
973 linux_to_bsd_waitopts(args->options, &options);
975 if (args->rusage != NULL)
979 error = linux_common_wait(td, args->pid, args->status, options, wrup);
982 if (args->rusage != NULL)
983 error = linux_copyout_rusage(&wru.wru_self, args->rusage);
988 linux_waitid(struct thread *td, struct linux_waitid_args *args)
990 int status, options, sig;
991 struct __wrusage wru;
999 linux_to_bsd_waitopts(args->options, &options);
1001 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1003 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1006 switch (args->idtype) {
1024 error = kern_wait6(td, idtype, args->id, &status, options,
1028 if (args->rusage != NULL) {
1029 error = linux_copyout_rusage(&wru.wru_children,
1034 if (args->info != NULL) {
1036 bzero(&lsi, sizeof(lsi));
1037 if (td->td_retval[0] != 0) {
1038 sig = bsd_to_linux_signal(siginfo.si_signo);
1039 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1041 error = copyout(&lsi, args->info, sizeof(lsi));
1043 td->td_retval[0] = 0;
1048 #ifdef LINUX_LEGACY_SYSCALLS
1050 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1055 LCONVPATHCREAT(td, args->path, &path);
1057 switch (args->mode & S_IFMT) {
1060 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1066 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1067 args->mode, args->dev);
1075 args->mode |= S_IFREG;
1078 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1079 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1081 kern_close(td, td->td_retval[0]);
1094 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1099 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1100 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1102 switch (args->mode & S_IFMT) {
1105 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1110 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1119 args->mode |= S_IFREG;
1122 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1123 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1125 kern_close(td, td->td_retval[0]);
1137 * UGH! This is just about the dumbest idea I've ever heard!!
1140 linux_personality(struct thread *td, struct linux_personality_args *args)
1142 struct linux_pemuldata *pem;
1143 struct proc *p = td->td_proc;
1149 if (args->per != 0xffffffff)
1150 pem->persona = args->per;
1153 td->td_retval[0] = old;
1157 struct l_itimerval {
1158 l_timeval it_interval;
1162 #define B2L_ITIMERVAL(bip, lip) \
1163 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1164 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1165 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1166 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1169 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1172 struct l_itimerval ls;
1173 struct itimerval aitv, oitv;
1175 if (uap->itv == NULL) {
1176 uap->itv = uap->oitv;
1177 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1180 error = copyin(uap->itv, &ls, sizeof(ls));
1183 B2L_ITIMERVAL(&aitv, &ls);
1184 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1185 if (error != 0 || uap->oitv == NULL)
1187 B2L_ITIMERVAL(&ls, &oitv);
1189 return (copyout(&ls, uap->oitv, sizeof(ls)));
1193 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1196 struct l_itimerval ls;
1197 struct itimerval aitv;
1199 error = kern_getitimer(td, uap->which, &aitv);
1202 B2L_ITIMERVAL(&ls, &aitv);
1203 return (copyout(&ls, uap->itv, sizeof(ls)));
1206 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1208 linux_nice(struct thread *td, struct linux_nice_args *args)
1211 return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1213 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1216 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1218 struct ucred *newcred, *oldcred;
1219 l_gid_t *linux_gidset;
1224 ngrp = args->gidsetsize;
1225 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1227 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1228 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1232 crextend(newcred, ngrp + 1);
1235 oldcred = p->p_ucred;
1236 crcopy(newcred, oldcred);
1239 * cr_groups[0] holds egid. Setting the whole set from
1240 * the supplied set will cause egid to be changed too.
1241 * Keep cr_groups[0] unchanged to prevent that.
1244 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1251 newcred->cr_ngroups = ngrp + 1;
1253 bsd_gidset = newcred->cr_groups;
1256 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1260 newcred->cr_ngroups = 1;
1263 proc_set_cred(p, newcred);
1268 free(linux_gidset, M_LINUX);
1273 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1276 l_gid_t *linux_gidset;
1278 int bsd_gidsetsz, ngrp, error;
1280 cred = td->td_ucred;
1281 bsd_gidset = cred->cr_groups;
1282 bsd_gidsetsz = cred->cr_ngroups - 1;
1285 * cr_groups[0] holds egid. Returning the whole set
1286 * here will cause a duplicate. Exclude cr_groups[0]
1290 if ((ngrp = args->gidsetsize) == 0) {
1291 td->td_retval[0] = bsd_gidsetsz;
1295 if (ngrp < bsd_gidsetsz)
1299 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1301 while (ngrp < bsd_gidsetsz) {
1302 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1306 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1307 free(linux_gidset, M_LINUX);
1311 td->td_retval[0] = ngrp;
1316 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1318 struct rlimit bsd_rlim;
1319 struct l_rlimit rlim;
1323 if (args->resource >= LINUX_RLIM_NLIMITS)
1326 which = linux_to_bsd_resource[args->resource];
1330 error = copyin(args->rlim, &rlim, sizeof(rlim));
1334 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1335 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1336 return (kern_setrlimit(td, which, &bsd_rlim));
1339 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1341 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1343 struct l_rlimit rlim;
1344 struct rlimit bsd_rlim;
1347 if (args->resource >= LINUX_RLIM_NLIMITS)
1350 which = linux_to_bsd_resource[args->resource];
1354 lim_rlimit(td, which, &bsd_rlim);
1356 #ifdef COMPAT_LINUX32
1357 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1358 if (rlim.rlim_cur == UINT_MAX)
1359 rlim.rlim_cur = INT_MAX;
1360 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1361 if (rlim.rlim_max == UINT_MAX)
1362 rlim.rlim_max = INT_MAX;
1364 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1365 if (rlim.rlim_cur == ULONG_MAX)
1366 rlim.rlim_cur = LONG_MAX;
1367 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1368 if (rlim.rlim_max == ULONG_MAX)
1369 rlim.rlim_max = LONG_MAX;
1371 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1373 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1376 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1378 struct l_rlimit rlim;
1379 struct rlimit bsd_rlim;
1382 if (args->resource >= LINUX_RLIM_NLIMITS)
1385 which = linux_to_bsd_resource[args->resource];
1389 lim_rlimit(td, which, &bsd_rlim);
1391 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1392 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1393 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1397 linux_sched_setscheduler(struct thread *td,
1398 struct linux_sched_setscheduler_args *args)
1400 struct sched_param sched_param;
1404 switch (args->policy) {
1405 case LINUX_SCHED_OTHER:
1406 policy = SCHED_OTHER;
1408 case LINUX_SCHED_FIFO:
1409 policy = SCHED_FIFO;
1411 case LINUX_SCHED_RR:
1418 error = copyin(args->param, &sched_param, sizeof(sched_param));
1422 if (linux_map_sched_prio) {
1425 if (sched_param.sched_priority != 0)
1428 sched_param.sched_priority =
1429 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1433 if (sched_param.sched_priority < 1 ||
1434 sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1438 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1439 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1441 sched_param.sched_priority =
1442 (sched_param.sched_priority - 1) *
1443 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1444 (LINUX_MAX_RT_PRIO - 1);
1449 tdt = linux_tdfind(td, args->pid, -1);
1453 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1454 PROC_UNLOCK(tdt->td_proc);
1459 linux_sched_getscheduler(struct thread *td,
1460 struct linux_sched_getscheduler_args *args)
1465 tdt = linux_tdfind(td, args->pid, -1);
1469 error = kern_sched_getscheduler(td, tdt, &policy);
1470 PROC_UNLOCK(tdt->td_proc);
1474 td->td_retval[0] = LINUX_SCHED_OTHER;
1477 td->td_retval[0] = LINUX_SCHED_FIFO;
1480 td->td_retval[0] = LINUX_SCHED_RR;
1487 linux_sched_get_priority_max(struct thread *td,
1488 struct linux_sched_get_priority_max_args *args)
1490 struct sched_get_priority_max_args bsd;
1492 if (linux_map_sched_prio) {
1493 switch (args->policy) {
1494 case LINUX_SCHED_OTHER:
1495 td->td_retval[0] = 0;
1497 case LINUX_SCHED_FIFO:
1498 case LINUX_SCHED_RR:
1499 td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1506 switch (args->policy) {
1507 case LINUX_SCHED_OTHER:
1508 bsd.policy = SCHED_OTHER;
1510 case LINUX_SCHED_FIFO:
1511 bsd.policy = SCHED_FIFO;
1513 case LINUX_SCHED_RR:
1514 bsd.policy = SCHED_RR;
1519 return (sys_sched_get_priority_max(td, &bsd));
1523 linux_sched_get_priority_min(struct thread *td,
1524 struct linux_sched_get_priority_min_args *args)
1526 struct sched_get_priority_min_args bsd;
1528 if (linux_map_sched_prio) {
1529 switch (args->policy) {
1530 case LINUX_SCHED_OTHER:
1531 td->td_retval[0] = 0;
1533 case LINUX_SCHED_FIFO:
1534 case LINUX_SCHED_RR:
1535 td->td_retval[0] = 1;
1542 switch (args->policy) {
1543 case LINUX_SCHED_OTHER:
1544 bsd.policy = SCHED_OTHER;
1546 case LINUX_SCHED_FIFO:
1547 bsd.policy = SCHED_FIFO;
1549 case LINUX_SCHED_RR:
1550 bsd.policy = SCHED_RR;
1555 return (sys_sched_get_priority_min(td, &bsd));
1558 #define REBOOT_CAD_ON 0x89abcdef
1559 #define REBOOT_CAD_OFF 0
1560 #define REBOOT_HALT 0xcdef0123
1561 #define REBOOT_RESTART 0x01234567
1562 #define REBOOT_RESTART2 0xA1B2C3D4
1563 #define REBOOT_POWEROFF 0x4321FEDC
1564 #define REBOOT_MAGIC1 0xfee1dead
1565 #define REBOOT_MAGIC2 0x28121969
1566 #define REBOOT_MAGIC2A 0x05121996
1567 #define REBOOT_MAGIC2B 0x16041998
1570 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1572 struct reboot_args bsd_args;
1574 if (args->magic1 != REBOOT_MAGIC1)
1577 switch (args->magic2) {
1579 case REBOOT_MAGIC2A:
1580 case REBOOT_MAGIC2B:
1586 switch (args->cmd) {
1588 case REBOOT_CAD_OFF:
1589 return (priv_check(td, PRIV_REBOOT));
1591 bsd_args.opt = RB_HALT;
1593 case REBOOT_RESTART:
1594 case REBOOT_RESTART2:
1597 case REBOOT_POWEROFF:
1598 bsd_args.opt = RB_POWEROFF;
1603 return (sys_reboot(td, &bsd_args));
1608 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1611 td->td_retval[0] = td->td_proc->p_pid;
1617 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1619 struct linux_emuldata *em;
1622 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1624 td->td_retval[0] = em->em_tid;
1631 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1634 td->td_retval[0] = kern_getppid(td);
1639 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1642 td->td_retval[0] = td->td_ucred->cr_rgid;
1647 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1650 td->td_retval[0] = td->td_ucred->cr_ruid;
1655 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1658 return (kern_getsid(td, args->pid));
1662 linux_nosys(struct thread *td, struct nosys_args *ignore)
1669 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1673 error = kern_getpriority(td, args->which, args->who);
1674 td->td_retval[0] = 20 - td->td_retval[0];
1679 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1684 name[1] = KERN_HOSTNAME;
1685 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1690 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1695 name[1] = KERN_NISDOMAINNAME;
1696 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1701 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1704 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1708 * XXX: we should send a signal to the parent if
1709 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1710 * as it doesnt occur often.
1712 exit1(td, args->error_code, 0);
1716 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1717 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1718 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1720 struct l_user_cap_header {
1725 struct l_user_cap_data {
1732 linux_capget(struct thread *td, struct linux_capget_args *uap)
1734 struct l_user_cap_header luch;
1735 struct l_user_cap_data lucd[2];
1738 if (uap->hdrp == NULL)
1741 error = copyin(uap->hdrp, &luch, sizeof(luch));
1745 switch (luch.version) {
1746 case _LINUX_CAPABILITY_VERSION_1:
1749 case _LINUX_CAPABILITY_VERSION_2:
1750 case _LINUX_CAPABILITY_VERSION_3:
1754 luch.version = _LINUX_CAPABILITY_VERSION_1;
1755 error = copyout(&luch, uap->hdrp, sizeof(luch));
1766 * The current implementation doesn't support setting
1767 * a capability (it's essentially a stub) so indicate
1768 * that no capabilities are currently set or available
1771 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1772 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1779 linux_capset(struct thread *td, struct linux_capset_args *uap)
1781 struct l_user_cap_header luch;
1782 struct l_user_cap_data lucd[2];
1785 if (uap->hdrp == NULL || uap->datap == NULL)
1788 error = copyin(uap->hdrp, &luch, sizeof(luch));
1792 switch (luch.version) {
1793 case _LINUX_CAPABILITY_VERSION_1:
1796 case _LINUX_CAPABILITY_VERSION_2:
1797 case _LINUX_CAPABILITY_VERSION_3:
1801 luch.version = _LINUX_CAPABILITY_VERSION_1;
1802 error = copyout(&luch, uap->hdrp, sizeof(luch));
1811 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1815 /* We currently don't support setting any capabilities. */
1816 for (i = 0; i < u32s; i++) {
1817 if (lucd[i].effective || lucd[i].permitted ||
1818 lucd[i].inheritable) {
1820 "capset[%d] effective=0x%x, permitted=0x%x, "
1821 "inheritable=0x%x is not implemented", i,
1822 (int)lucd[i].effective, (int)lucd[i].permitted,
1823 (int)lucd[i].inheritable);
1832 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1834 int error = 0, max_size;
1835 struct proc *p = td->td_proc;
1836 char comm[LINUX_MAX_COMM_LEN];
1839 switch (args->option) {
1840 case LINUX_PR_SET_PDEATHSIG:
1841 if (!LINUX_SIG_VALID(args->arg2))
1843 pdeath_signal = linux_to_bsd_signal(args->arg2);
1844 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1846 case LINUX_PR_GET_PDEATHSIG:
1847 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1851 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1852 return (copyout(&pdeath_signal,
1853 (void *)(register_t)args->arg2,
1854 sizeof(pdeath_signal)));
1856 case LINUX_PR_GET_KEEPCAPS:
1858 * Indicate that we always clear the effective and
1859 * permitted capability sets when the user id becomes
1860 * non-zero (actually the capability sets are simply
1861 * always zero in the current implementation).
1863 td->td_retval[0] = 0;
1865 case LINUX_PR_SET_KEEPCAPS:
1867 * Ignore requests to keep the effective and permitted
1868 * capability sets when the user id becomes non-zero.
1871 case LINUX_PR_SET_NAME:
1873 * To be on the safe side we need to make sure to not
1874 * overflow the size a Linux program expects. We already
1875 * do this here in the copyin, so that we don't need to
1878 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1879 error = copyinstr((void *)(register_t)args->arg2, comm,
1882 /* Linux silently truncates the name if it is too long. */
1883 if (error == ENAMETOOLONG) {
1885 * XXX: copyinstr() isn't documented to populate the
1886 * array completely, so do a copyin() to be on the
1887 * safe side. This should be changed in case
1888 * copyinstr() is changed to guarantee this.
1890 error = copyin((void *)(register_t)args->arg2, comm,
1892 comm[max_size - 1] = '\0';
1898 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1901 case LINUX_PR_GET_NAME:
1903 strlcpy(comm, p->p_comm, sizeof(comm));
1905 error = copyout(comm, (void *)(register_t)args->arg2,
1917 linux_sched_setparam(struct thread *td,
1918 struct linux_sched_setparam_args *uap)
1920 struct sched_param sched_param;
1924 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1928 tdt = linux_tdfind(td, uap->pid, -1);
1932 if (linux_map_sched_prio) {
1933 error = kern_sched_getscheduler(td, tdt, &policy);
1939 if (sched_param.sched_priority != 0) {
1943 sched_param.sched_priority =
1944 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1948 if (sched_param.sched_priority < 1 ||
1949 sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1954 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1955 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1957 sched_param.sched_priority =
1958 (sched_param.sched_priority - 1) *
1959 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1960 (LINUX_MAX_RT_PRIO - 1);
1965 error = kern_sched_setparam(td, tdt, &sched_param);
1966 out: PROC_UNLOCK(tdt->td_proc);
1971 linux_sched_getparam(struct thread *td,
1972 struct linux_sched_getparam_args *uap)
1974 struct sched_param sched_param;
1978 tdt = linux_tdfind(td, uap->pid, -1);
1982 error = kern_sched_getparam(td, tdt, &sched_param);
1984 PROC_UNLOCK(tdt->td_proc);
1988 if (linux_map_sched_prio) {
1989 error = kern_sched_getscheduler(td, tdt, &policy);
1990 PROC_UNLOCK(tdt->td_proc);
1996 sched_param.sched_priority = 0;
2001 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2002 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2004 sched_param.sched_priority =
2005 (sched_param.sched_priority *
2006 (LINUX_MAX_RT_PRIO - 1) +
2007 (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2008 (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2012 PROC_UNLOCK(tdt->td_proc);
2014 error = copyout(&sched_param, uap->param, sizeof(sched_param));
2019 * Get affinity of a process.
2022 linux_sched_getaffinity(struct thread *td,
2023 struct linux_sched_getaffinity_args *args)
2028 if (args->len < sizeof(cpuset_t))
2031 tdt = linux_tdfind(td, args->pid, -1);
2035 PROC_UNLOCK(tdt->td_proc);
2037 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2038 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2040 td->td_retval[0] = sizeof(cpuset_t);
2046 * Set affinity of a process.
2049 linux_sched_setaffinity(struct thread *td,
2050 struct linux_sched_setaffinity_args *args)
2054 if (args->len < sizeof(cpuset_t))
2057 tdt = linux_tdfind(td, args->pid, -1);
2061 PROC_UNLOCK(tdt->td_proc);
2063 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2064 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2067 struct linux_rlimit64 {
2073 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2075 struct rlimit rlim, nrlim;
2076 struct linux_rlimit64 lrlim;
2082 if (args->resource >= LINUX_RLIM_NLIMITS)
2085 which = linux_to_bsd_resource[args->resource];
2089 if (args->new != NULL) {
2091 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2092 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2093 * as INFINITY so we do not need a conversion even.
2095 error = copyin(args->new, &nrlim, sizeof(nrlim));
2100 flags = PGET_HOLD | PGET_NOTWEXIT;
2101 if (args->new != NULL)
2102 flags |= PGET_CANDEBUG;
2104 flags |= PGET_CANSEE;
2105 if (args->pid == 0) {
2109 error = pget(args->pid, flags, &p);
2113 if (args->old != NULL) {
2115 lim_rlimit_proc(p, which, &rlim);
2117 if (rlim.rlim_cur == RLIM_INFINITY)
2118 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2120 lrlim.rlim_cur = rlim.rlim_cur;
2121 if (rlim.rlim_max == RLIM_INFINITY)
2122 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2124 lrlim.rlim_max = rlim.rlim_max;
2125 error = copyout(&lrlim, args->old, sizeof(lrlim));
2130 if (args->new != NULL)
2131 error = kern_proc_setrlimit(td, p, which, &nrlim);
2139 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2141 struct timeval utv, tv0, tv1, *tvp;
2142 struct l_pselect6arg lpse6;
2143 struct l_timespec lts;
2144 struct timespec uts;
2151 if (args->sig != NULL) {
2152 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2155 if (lpse6.ss_len != sizeof(l_ss))
2157 if (lpse6.ss != 0) {
2158 error = copyin(PTRIN(lpse6.ss), &l_ss,
2162 linux_to_bsd_sigset(&l_ss, &ss);
2168 * Currently glibc changes nanosecond number to microsecond.
2169 * This mean losing precision but for now it is hardly seen.
2171 if (args->tsp != NULL) {
2172 error = copyin(args->tsp, <s, sizeof(lts));
2175 error = linux_to_native_timespec(&uts, <s);
2179 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2180 if (itimerfix(&utv))
2188 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2189 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2191 if (error == 0 && args->tsp != NULL) {
2192 if (td->td_retval[0] != 0) {
2194 * Compute how much time was left of the timeout,
2195 * by subtracting the current time and the time
2196 * before we started the call, and subtracting
2197 * that result from the user-supplied value.
2201 timevalsub(&tv1, &tv0);
2202 timevalsub(&utv, &tv1);
2208 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2210 error = native_to_linux_timespec(<s, &uts);
2212 error = copyout(<s, args->tsp, sizeof(lts));
2219 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2221 struct timespec ts0, ts1;
2222 struct l_timespec lts;
2223 struct timespec uts, *tsp;
2229 if (args->sset != NULL) {
2230 if (args->ssize != sizeof(l_ss))
2232 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2235 linux_to_bsd_sigset(&l_ss, &ss);
2239 if (args->tsp != NULL) {
2240 error = copyin(args->tsp, <s, sizeof(lts));
2243 error = linux_to_native_timespec(&uts, <s);
2252 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2254 if (error == 0 && args->tsp != NULL) {
2255 if (td->td_retval[0]) {
2257 timespecsub(&ts1, &ts0, &ts1);
2258 timespecsub(&uts, &ts1, &uts);
2260 timespecclear(&uts);
2262 timespecclear(&uts);
2264 error = native_to_linux_timespec(<s, &uts);
2266 error = copyout(<s, args->tsp, sizeof(lts));
2273 linux_sched_rr_get_interval(struct thread *td,
2274 struct linux_sched_rr_get_interval_args *uap)
2277 struct l_timespec lts;
2282 * According to man in case the invalid pid specified
2283 * EINVAL should be returned.
2288 tdt = linux_tdfind(td, uap->pid, -1);
2292 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2293 PROC_UNLOCK(tdt->td_proc);
2296 error = native_to_linux_timespec(<s, &ts);
2299 return (copyout(<s, uap->interval, sizeof(lts)));
2303 * In case when the Linux thread is the initial thread in
2304 * the thread group thread id is equal to the process id.
2305 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2308 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2310 struct linux_emuldata *em;
2315 if (tid == 0 || tid == td->td_tid) {
2317 PROC_LOCK(tdt->td_proc);
2318 } else if (tid > PID_MAX)
2319 tdt = tdfind(tid, pid);
2322 * Initial thread where the tid equal to the pid.
2326 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2328 * p is not a Linuxulator process.
2333 FOREACH_THREAD_IN_PROC(p, tdt) {
2335 if (tid == em->em_tid)
2347 linux_to_bsd_waitopts(int options, int *bsdopts)
2350 if (options & LINUX_WNOHANG)
2351 *bsdopts |= WNOHANG;
2352 if (options & LINUX_WUNTRACED)
2353 *bsdopts |= WUNTRACED;
2354 if (options & LINUX_WEXITED)
2355 *bsdopts |= WEXITED;
2356 if (options & LINUX_WCONTINUED)
2357 *bsdopts |= WCONTINUED;
2358 if (options & LINUX_WNOWAIT)
2359 *bsdopts |= WNOWAIT;
2361 if (options & __WCLONE)
2362 *bsdopts |= WLINUXCLONE;
2366 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2372 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2374 if (args->count > INT_MAX)
2375 args->count = INT_MAX;
2377 iov.iov_base = args->buf;
2378 iov.iov_len = args->count;
2382 uio.uio_resid = iov.iov_len;
2383 uio.uio_segflg = UIO_USERSPACE;
2384 uio.uio_rw = UIO_READ;
2387 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2389 td->td_retval[0] = args->count - uio.uio_resid;
2394 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2397 /* Needs to be page-aligned */
2398 if (args->start & PAGE_MASK)
2400 return (kern_mincore(td, args->start, args->len, args->vec));
2403 #define SYSLOG_TAG "<6>"
2406 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2408 char buf[128], *src, *dst;
2412 if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2413 linux_msg(td, "syslog unsupported type 0x%x", args->type);
2417 if (args->len < 6) {
2418 td->td_retval[0] = 0;
2422 error = priv_check(td, PRIV_MSGBUF);
2426 mtx_lock(&msgbuf_lock);
2427 msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2428 mtx_unlock(&msgbuf_lock);
2431 error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2432 /* The -1 is to skip the trailing '\0'. */
2433 dst += sizeof(SYSLOG_TAG) - 1;
2435 while (error == 0) {
2436 mtx_lock(&msgbuf_lock);
2437 buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2438 mtx_unlock(&msgbuf_lock);
2443 for (src = buf; src < buf + buflen && error == 0; src++) {
2447 if (dst >= args->buf + args->len)
2450 error = copyout(src, dst, 1);
2453 if (*src == '\n' && *(src + 1) != '<' &&
2454 dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2455 error = copyout(&SYSLOG_TAG,
2456 dst, sizeof(SYSLOG_TAG));
2457 dst += sizeof(SYSLOG_TAG) - 1;
2462 td->td_retval[0] = dst - args->buf;
2467 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2469 int cpu, error, node;
2471 cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2473 node = cpuid_to_pcpu[cpu]->pc_domain;
2475 if (args->cpu != NULL)
2476 error = copyout(&cpu, args->cpu, sizeof(l_int));
2477 if (args->node != NULL)
2478 error = copyout(&node, args->node, sizeof(l_int));