2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2002 Doug Rabson
5 * Copyright (c) 1994-1995 Søren Schmidt
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer
13 * in this position and unchanged.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. The name of the author may not be used to endorse or promote products
18 * derived from this software without specific prior written permission
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_compat.h"
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
41 #include <sys/imgact_aout.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
47 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
51 #include <sys/namei.h>
54 #include <sys/procctl.h>
55 #include <sys/reboot.h>
56 #include <sys/racct.h>
57 #include <sys/random.h>
58 #include <sys/resourcevar.h>
59 #include <sys/sched.h>
61 #include <sys/signalvar.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/sysproto.h>
66 #include <sys/systm.h>
68 #include <sys/vmmeter.h>
69 #include <sys/vnode.h>
71 #include <sys/cpuset.h>
74 #include <security/mac/mac_framework.h>
78 #include <vm/vm_kern.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
81 #include <vm/swap_pager.h>
84 #include <machine/../linux32/linux.h>
85 #include <machine/../linux32/linux32_proto.h>
87 #include <machine/../linux/linux.h>
88 #include <machine/../linux/linux_proto.h>
91 #include <compat/linux/linux_dtrace.h>
92 #include <compat/linux/linux_file.h>
93 #include <compat/linux/linux_mib.h>
94 #include <compat/linux/linux_signal.h>
95 #include <compat/linux/linux_timer.h>
96 #include <compat/linux/linux_util.h>
97 #include <compat/linux/linux_sysproto.h>
98 #include <compat/linux/linux_emul.h>
99 #include <compat/linux/linux_misc.h>
102 * Special DTrace provider for the linuxulator.
104 * In this file we define the provider for the entire linuxulator. All
105 * modules (= files of the linuxulator) use it.
107 * We define a different name depending on the emulated bitsize, see
108 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
109 * native bitsize = linuxulator
110 * amd64, 32bit emulation = linuxulator32
112 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
114 int stclohz; /* Statistics clock frequency */
116 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
117 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
118 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
119 RLIMIT_MEMLOCK, RLIMIT_AS
123 l_long uptime; /* Seconds since boot */
124 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
125 #define LINUX_SYSINFO_LOADS_SCALE 65536
126 l_ulong totalram; /* Total usable main memory size */
127 l_ulong freeram; /* Available memory size */
128 l_ulong sharedram; /* Amount of shared memory */
129 l_ulong bufferram; /* Memory used by buffers */
130 l_ulong totalswap; /* Total swap space size */
131 l_ulong freeswap; /* swap space still available */
132 l_ushort procs; /* Number of current processes */
137 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
140 struct l_pselect6arg {
145 static int linux_utimensat_nsec_valid(l_long);
149 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
151 struct l_sysinfo sysinfo;
155 bzero(&sysinfo, sizeof(sysinfo));
159 sysinfo.uptime = ts.tv_sec;
161 /* Use the information from the mib to get our load averages */
162 for (i = 0; i < 3; i++)
163 sysinfo.loads[i] = averunnable.ldavg[i] *
164 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
166 sysinfo.totalram = physmem * PAGE_SIZE;
167 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
170 * sharedram counts pages allocated to named, swap-backed objects such
171 * as shared memory segments and tmpfs files. There is no cheap way to
172 * compute this, so just leave the field unpopulated. Linux itself only
173 * started setting this field in the 3.x timeframe.
175 sysinfo.sharedram = 0;
176 sysinfo.bufferram = 0;
178 swap_pager_status(&i, &j);
179 sysinfo.totalswap = i * PAGE_SIZE;
180 sysinfo.freeswap = (i - j) * PAGE_SIZE;
182 sysinfo.procs = nprocs;
184 /* The following are only present in newer Linux kernels. */
185 sysinfo.totalbig = 0;
187 sysinfo.mem_unit = 1;
189 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
192 #ifdef LINUX_LEGACY_SYSCALLS
194 linux_alarm(struct thread *td, struct linux_alarm_args *args)
196 struct itimerval it, old_it;
202 printf(ARGS(alarm, "%u"), args->secs);
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;
237 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
239 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
240 new = (uintptr_t)args->dsend;
241 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
242 td->td_retval[0] = (register_t)new;
244 td->td_retval[0] = (register_t)old;
249 #if defined(__i386__)
250 /* XXX: what about amd64/linux32? */
253 linux_uselib(struct thread *td, struct linux_uselib_args *args)
259 vm_map_entry_t entry;
262 unsigned long file_offset;
263 unsigned long bss_size;
267 bool locked, opened, textset;
269 LCONVPATHEXIST(td, args->library, &library);
273 printf(ARGS(uselib, "%s"), library);
282 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
283 UIO_SYSSPACE, library, td);
290 NDFREE(&ni, NDF_ONLY_PNBUF);
293 * From here on down, we have a locked vnode that must be unlocked.
294 * XXX: The code below largely duplicates exec_check_permissions().
299 error = VOP_GETATTR(vp, &attr, td->td_ucred);
303 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
304 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
305 /* EACCESS is what exec(2) returns. */
311 if (attr.va_size == 0) {
316 /* Can we access it? */
317 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
322 * XXX: This should use vn_open() so that it is properly authorized,
323 * and to reduce code redundancy all over the place here.
324 * XXX: Not really, it duplicates far more of exec_check_permissions()
328 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
332 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
337 /* Pull in executable header into exec_map */
338 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
339 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
343 /* Is it a Linux binary ? */
344 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
350 * While we are here, we should REALLY do some more checks
353 /* Set file/virtual offset based on a.out variant. */
354 switch ((int)(a_out->a_magic & 0xffff)) {
355 case 0413: /* ZMAGIC */
358 case 0314: /* QMAGIC */
366 bss_size = round_page(a_out->a_bss);
368 /* Check various fields in header for validity/bounds. */
369 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
374 /* text + data can't exceed file size */
375 if (a_out->a_data + a_out->a_text > attr.va_size) {
381 * text/data/bss must not exceed limits
382 * XXX - this is not complete. it should check current usage PLUS
383 * the resources needed by this library.
385 PROC_LOCK(td->td_proc);
386 if (a_out->a_text > maxtsiz ||
387 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
388 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
390 PROC_UNLOCK(td->td_proc);
394 PROC_UNLOCK(td->td_proc);
397 * Prevent more writers.
399 error = VOP_SET_TEXT(vp);
405 * Lock no longer needed
411 * Check if file_offset page aligned. Currently we cannot handle
412 * misalinged file offsets, and so we read in the entire image
415 if (file_offset & PAGE_MASK) {
417 printf("uselib: Non page aligned binary %lu\n", file_offset);
419 /* Map text+data read/write/execute */
421 /* a_entry is the load address and is page aligned */
422 vmaddr = trunc_page(a_out->a_entry);
424 /* get anon user mapping, read+write+execute */
425 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
426 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
427 VM_PROT_ALL, VM_PROT_ALL, 0);
431 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
432 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
433 td->td_ucred, NOCRED, &aresid, td);
442 printf("uselib: Page aligned binary %lu\n", file_offset);
445 * for QMAGIC, a_entry is 20 bytes beyond the load address
446 * to skip the executable header
448 vmaddr = trunc_page(a_out->a_entry);
451 * Map it all into the process's space as a single
452 * copy-on-write "data" segment.
454 map = &td->td_proc->p_vmspace->vm_map;
455 error = vm_mmap(map, &vmaddr,
456 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
457 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
461 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
466 entry->eflags |= MAP_ENTRY_VN_EXEC;
471 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
472 ((long *)vmaddr)[1]);
475 /* Calculate BSS start address */
476 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
479 /* allocate some 'anon' space */
480 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
481 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
492 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
497 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
499 VOP_UNSET_TEXT_CHECKED(vp);
504 /* Release the temporary mapping. */
506 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
511 #endif /* __i386__ */
513 #ifdef LINUX_LEGACY_SYSCALLS
515 linux_select(struct thread *td, struct linux_select_args *args)
518 struct timeval tv0, tv1, utv, *tvp;
523 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
524 (void *)args->readfds, (void *)args->writefds,
525 (void *)args->exceptfds, (void *)args->timeout);
529 * Store current time for computation of the amount of
533 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
535 utv.tv_sec = ltv.tv_sec;
536 utv.tv_usec = ltv.tv_usec;
539 printf(LMSG("incoming timeout (%jd/%ld)"),
540 (intmax_t)utv.tv_sec, utv.tv_usec);
543 if (itimerfix(&utv)) {
545 * The timeval was invalid. Convert it to something
546 * valid that will act as it does under Linux.
548 utv.tv_sec += utv.tv_usec / 1000000;
549 utv.tv_usec %= 1000000;
550 if (utv.tv_usec < 0) {
552 utv.tv_usec += 1000000;
562 error = kern_select(td, args->nfds, args->readfds, args->writefds,
563 args->exceptfds, tvp, LINUX_NFDBITS);
567 printf(LMSG("real select returns %d"), error);
573 if (td->td_retval[0]) {
575 * Compute how much time was left of the timeout,
576 * by subtracting the current time and the time
577 * before we started the call, and subtracting
578 * that result from the user-supplied value.
581 timevalsub(&tv1, &tv0);
582 timevalsub(&utv, &tv1);
589 printf(LMSG("outgoing timeout (%jd/%ld)"),
590 (intmax_t)utv.tv_sec, utv.tv_usec);
592 ltv.tv_sec = utv.tv_sec;
593 ltv.tv_usec = utv.tv_usec;
594 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
601 printf(LMSG("select_out -> %d"), error);
608 linux_mremap(struct thread *td, struct linux_mremap_args *args)
616 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
617 (void *)(uintptr_t)args->addr,
618 (unsigned long)args->old_len,
619 (unsigned long)args->new_len,
620 (unsigned long)args->flags);
623 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
624 td->td_retval[0] = 0;
629 * Check for the page alignment.
630 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
632 if (args->addr & PAGE_MASK) {
633 td->td_retval[0] = 0;
637 args->new_len = round_page(args->new_len);
638 args->old_len = round_page(args->old_len);
640 if (args->new_len > args->old_len) {
641 td->td_retval[0] = 0;
645 if (args->new_len < args->old_len) {
646 addr = args->addr + args->new_len;
647 len = args->old_len - args->new_len;
648 error = kern_munmap(td, addr, len);
651 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
655 #define LINUX_MS_ASYNC 0x0001
656 #define LINUX_MS_INVALIDATE 0x0002
657 #define LINUX_MS_SYNC 0x0004
660 linux_msync(struct thread *td, struct linux_msync_args *args)
663 return (kern_msync(td, args->addr, args->len,
664 args->fl & ~LINUX_MS_SYNC));
667 #ifdef LINUX_LEGACY_SYSCALLS
669 linux_time(struct thread *td, struct linux_time_args *args)
677 printf(ARGS(time, "*"));
682 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
684 td->td_retval[0] = tm;
689 struct l_times_argv {
692 l_clock_t tms_cutime;
693 l_clock_t tms_cstime;
698 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
699 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
700 * auxiliary vector entry.
704 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
705 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
707 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
708 CONVNTCK(r) : CONVOTCK(r))
711 linux_times(struct thread *td, struct linux_times_args *args)
713 struct timeval tv, utime, stime, cutime, cstime;
714 struct l_times_argv tms;
720 printf(ARGS(times, "*"));
723 if (args->buf != NULL) {
727 calcru(p, &utime, &stime);
729 calccru(p, &cutime, &cstime);
732 tms.tms_utime = CONVTCK(utime);
733 tms.tms_stime = CONVTCK(stime);
735 tms.tms_cutime = CONVTCK(cutime);
736 tms.tms_cstime = CONVTCK(cstime);
738 if ((error = copyout(&tms, args->buf, sizeof(tms))))
743 td->td_retval[0] = (int)CONVTCK(tv);
748 linux_newuname(struct thread *td, struct linux_newuname_args *args)
750 struct l_new_utsname utsname;
751 char osname[LINUX_MAX_UTSNAME];
752 char osrelease[LINUX_MAX_UTSNAME];
756 if (ldebug(newuname))
757 printf(ARGS(newuname, "*"));
760 linux_get_osname(td, osname);
761 linux_get_osrelease(td, osrelease);
763 bzero(&utsname, sizeof(utsname));
764 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
765 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
766 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
767 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
768 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
769 for (p = utsname.version; *p != '\0'; ++p)
774 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
776 return (copyout(&utsname, args->buf, sizeof(utsname)));
784 #ifdef LINUX_LEGACY_SYSCALLS
786 linux_utime(struct thread *td, struct linux_utime_args *args)
788 struct timeval tv[2], *tvp;
789 struct l_utimbuf lut;
793 LCONVPATHEXIST(td, args->fname, &fname);
797 printf(ARGS(utime, "%s, *"), fname);
801 if ((error = copyin(args->times, &lut, sizeof lut))) {
805 tv[0].tv_sec = lut.l_actime;
807 tv[1].tv_sec = lut.l_modtime;
813 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
820 #ifdef LINUX_LEGACY_SYSCALLS
822 linux_utimes(struct thread *td, struct linux_utimes_args *args)
825 struct timeval tv[2], *tvp = NULL;
829 LCONVPATHEXIST(td, args->fname, &fname);
833 printf(ARGS(utimes, "%s, *"), fname);
836 if (args->tptr != NULL) {
837 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
841 tv[0].tv_sec = ltv[0].tv_sec;
842 tv[0].tv_usec = ltv[0].tv_usec;
843 tv[1].tv_sec = ltv[1].tv_sec;
844 tv[1].tv_usec = ltv[1].tv_usec;
848 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
856 linux_utimensat_nsec_valid(l_long nsec)
859 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
861 if (nsec >= 0 && nsec <= 999999999)
867 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
869 struct l_timespec l_times[2];
870 struct timespec times[2], *timesp = NULL;
872 int error, dfd, flags = 0;
874 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
877 if (ldebug(utimensat))
878 printf(ARGS(utimensat, "%d, *"), dfd);
881 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
884 if (args->times != NULL) {
885 error = copyin(args->times, l_times, sizeof(l_times));
889 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
890 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
893 times[0].tv_sec = l_times[0].tv_sec;
894 switch (l_times[0].tv_nsec)
896 case LINUX_UTIME_OMIT:
897 times[0].tv_nsec = UTIME_OMIT;
899 case LINUX_UTIME_NOW:
900 times[0].tv_nsec = UTIME_NOW;
903 times[0].tv_nsec = l_times[0].tv_nsec;
906 times[1].tv_sec = l_times[1].tv_sec;
907 switch (l_times[1].tv_nsec)
909 case LINUX_UTIME_OMIT:
910 times[1].tv_nsec = UTIME_OMIT;
912 case LINUX_UTIME_NOW:
913 times[1].tv_nsec = UTIME_NOW;
916 times[1].tv_nsec = l_times[1].tv_nsec;
921 /* This breaks POSIX, but is what the Linux kernel does
922 * _on purpose_ (documented in the man page for utimensat(2)),
923 * so we must follow that behaviour. */
924 if (times[0].tv_nsec == UTIME_OMIT &&
925 times[1].tv_nsec == UTIME_OMIT)
929 if (args->pathname != NULL)
930 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
931 else if (args->flags != 0)
934 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
935 flags |= AT_SYMLINK_NOFOLLOW;
938 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
940 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
941 UIO_SYSSPACE, flags);
948 #ifdef LINUX_LEGACY_SYSCALLS
950 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
953 struct timeval tv[2], *tvp = NULL;
957 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
958 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
961 if (ldebug(futimesat))
962 printf(ARGS(futimesat, "%s, *"), fname);
965 if (args->utimes != NULL) {
966 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
970 tv[0].tv_sec = ltv[0].tv_sec;
971 tv[0].tv_usec = ltv[0].tv_usec;
972 tv[1].tv_sec = ltv[1].tv_sec;
973 tv[1].tv_usec = ltv[1].tv_usec;
977 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
984 linux_common_wait(struct thread *td, int pid, int *status,
985 int options, struct rusage *ru)
989 error = kern_wait(td, pid, &tmpstat, options, ru);
995 if (WIFSIGNALED(tmpstat))
996 tmpstat = (tmpstat & 0xffffff80) |
997 bsd_to_linux_signal(WTERMSIG(tmpstat));
998 else if (WIFSTOPPED(tmpstat))
999 tmpstat = (tmpstat & 0xffff00ff) |
1000 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
1001 else if (WIFCONTINUED(tmpstat))
1003 error = copyout(&tmpstat, status, sizeof(int));
1009 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1011 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
1013 struct linux_wait4_args wait4_args;
1016 if (ldebug(waitpid))
1017 printf(ARGS(waitpid, "%d, %p, %d"),
1018 args->pid, (void *)args->status, args->options);
1021 wait4_args.pid = args->pid;
1022 wait4_args.status = args->status;
1023 wait4_args.options = args->options;
1024 wait4_args.rusage = NULL;
1026 return (linux_wait4(td, &wait4_args));
1028 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1031 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1034 struct rusage ru, *rup;
1038 printf(ARGS(wait4, "%d, %p, %d, %p"),
1039 args->pid, (void *)args->status, args->options,
1040 (void *)args->rusage);
1042 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1043 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1047 linux_to_bsd_waitopts(args->options, &options);
1049 if (args->rusage != NULL)
1053 error = linux_common_wait(td, args->pid, args->status, options, rup);
1056 if (args->rusage != NULL)
1057 error = linux_copyout_rusage(&ru, args->rusage);
1062 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1064 int status, options, sig;
1065 struct __wrusage wru;
1073 linux_to_bsd_waitopts(args->options, &options);
1075 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1077 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1080 switch (args->idtype) {
1098 error = kern_wait6(td, idtype, args->id, &status, options,
1102 if (args->rusage != NULL) {
1103 error = linux_copyout_rusage(&wru.wru_children,
1108 if (args->info != NULL) {
1110 bzero(&lsi, sizeof(lsi));
1111 if (td->td_retval[0] != 0) {
1112 sig = bsd_to_linux_signal(siginfo.si_signo);
1113 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1115 error = copyout(&lsi, args->info, sizeof(lsi));
1117 td->td_retval[0] = 0;
1122 #ifdef LINUX_LEGACY_SYSCALLS
1124 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1129 LCONVPATHCREAT(td, args->path, &path);
1133 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1134 (uintmax_t)args->dev);
1137 switch (args->mode & S_IFMT) {
1140 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1146 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1147 args->mode, args->dev);
1155 args->mode |= S_IFREG;
1158 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1159 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1161 kern_close(td, td->td_retval[0]);
1174 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1179 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1180 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1183 if (ldebug(mknodat))
1184 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1187 switch (args->mode & S_IFMT) {
1190 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1195 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1204 args->mode |= S_IFREG;
1207 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1208 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1210 kern_close(td, td->td_retval[0]);
1222 * UGH! This is just about the dumbest idea I've ever heard!!
1225 linux_personality(struct thread *td, struct linux_personality_args *args)
1227 struct linux_pemuldata *pem;
1228 struct proc *p = td->td_proc;
1232 if (ldebug(personality))
1233 printf(ARGS(personality, "%u"), args->per);
1239 if (args->per != 0xffffffff)
1240 pem->persona = args->per;
1243 td->td_retval[0] = old;
1247 struct l_itimerval {
1248 l_timeval it_interval;
1252 #define B2L_ITIMERVAL(bip, lip) \
1253 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1254 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1255 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1256 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1259 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1262 struct l_itimerval ls;
1263 struct itimerval aitv, oitv;
1266 if (ldebug(setitimer))
1267 printf(ARGS(setitimer, "%p, %p"),
1268 (void *)uap->itv, (void *)uap->oitv);
1271 if (uap->itv == NULL) {
1272 uap->itv = uap->oitv;
1273 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1276 error = copyin(uap->itv, &ls, sizeof(ls));
1279 B2L_ITIMERVAL(&aitv, &ls);
1281 if (ldebug(setitimer)) {
1282 printf("setitimer: value: sec: %jd, usec: %ld\n",
1283 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1284 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1285 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1288 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1289 if (error != 0 || uap->oitv == NULL)
1291 B2L_ITIMERVAL(&ls, &oitv);
1293 return (copyout(&ls, uap->oitv, sizeof(ls)));
1297 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1300 struct l_itimerval ls;
1301 struct itimerval aitv;
1304 if (ldebug(getitimer))
1305 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1307 error = kern_getitimer(td, uap->which, &aitv);
1310 B2L_ITIMERVAL(&ls, &aitv);
1311 return (copyout(&ls, uap->itv, sizeof(ls)));
1314 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1316 linux_nice(struct thread *td, struct linux_nice_args *args)
1318 struct setpriority_args bsd_args;
1320 bsd_args.which = PRIO_PROCESS;
1321 bsd_args.who = 0; /* current process */
1322 bsd_args.prio = args->inc;
1323 return (sys_setpriority(td, &bsd_args));
1325 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1328 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1330 struct ucred *newcred, *oldcred;
1331 l_gid_t *linux_gidset;
1336 ngrp = args->gidsetsize;
1337 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1339 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1340 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1344 crextend(newcred, ngrp + 1);
1347 oldcred = p->p_ucred;
1348 crcopy(newcred, oldcred);
1351 * cr_groups[0] holds egid. Setting the whole set from
1352 * the supplied set will cause egid to be changed too.
1353 * Keep cr_groups[0] unchanged to prevent that.
1356 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1363 newcred->cr_ngroups = ngrp + 1;
1365 bsd_gidset = newcred->cr_groups;
1368 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1372 newcred->cr_ngroups = 1;
1375 proc_set_cred(p, newcred);
1380 free(linux_gidset, M_LINUX);
1385 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1388 l_gid_t *linux_gidset;
1390 int bsd_gidsetsz, ngrp, error;
1392 cred = td->td_ucred;
1393 bsd_gidset = cred->cr_groups;
1394 bsd_gidsetsz = cred->cr_ngroups - 1;
1397 * cr_groups[0] holds egid. Returning the whole set
1398 * here will cause a duplicate. Exclude cr_groups[0]
1402 if ((ngrp = args->gidsetsize) == 0) {
1403 td->td_retval[0] = bsd_gidsetsz;
1407 if (ngrp < bsd_gidsetsz)
1411 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1413 while (ngrp < bsd_gidsetsz) {
1414 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1418 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1419 free(linux_gidset, M_LINUX);
1423 td->td_retval[0] = ngrp;
1428 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1430 struct rlimit bsd_rlim;
1431 struct l_rlimit rlim;
1436 if (ldebug(setrlimit))
1437 printf(ARGS(setrlimit, "%d, %p"),
1438 args->resource, (void *)args->rlim);
1441 if (args->resource >= LINUX_RLIM_NLIMITS)
1444 which = linux_to_bsd_resource[args->resource];
1448 error = copyin(args->rlim, &rlim, sizeof(rlim));
1452 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1453 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1454 return (kern_setrlimit(td, which, &bsd_rlim));
1457 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1459 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1461 struct l_rlimit rlim;
1462 struct rlimit bsd_rlim;
1466 if (ldebug(old_getrlimit))
1467 printf(ARGS(old_getrlimit, "%d, %p"),
1468 args->resource, (void *)args->rlim);
1471 if (args->resource >= LINUX_RLIM_NLIMITS)
1474 which = linux_to_bsd_resource[args->resource];
1478 lim_rlimit(td, which, &bsd_rlim);
1480 #ifdef COMPAT_LINUX32
1481 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1482 if (rlim.rlim_cur == UINT_MAX)
1483 rlim.rlim_cur = INT_MAX;
1484 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1485 if (rlim.rlim_max == UINT_MAX)
1486 rlim.rlim_max = INT_MAX;
1488 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1489 if (rlim.rlim_cur == ULONG_MAX)
1490 rlim.rlim_cur = LONG_MAX;
1491 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1492 if (rlim.rlim_max == ULONG_MAX)
1493 rlim.rlim_max = LONG_MAX;
1495 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1497 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1500 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1502 struct l_rlimit rlim;
1503 struct rlimit bsd_rlim;
1507 if (ldebug(getrlimit))
1508 printf(ARGS(getrlimit, "%d, %p"),
1509 args->resource, (void *)args->rlim);
1512 if (args->resource >= LINUX_RLIM_NLIMITS)
1515 which = linux_to_bsd_resource[args->resource];
1519 lim_rlimit(td, which, &bsd_rlim);
1521 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1522 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1523 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1527 linux_sched_setscheduler(struct thread *td,
1528 struct linux_sched_setscheduler_args *args)
1530 struct sched_param sched_param;
1535 if (ldebug(sched_setscheduler))
1536 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1537 args->pid, args->policy, (const void *)args->param);
1540 switch (args->policy) {
1541 case LINUX_SCHED_OTHER:
1542 policy = SCHED_OTHER;
1544 case LINUX_SCHED_FIFO:
1545 policy = SCHED_FIFO;
1547 case LINUX_SCHED_RR:
1554 error = copyin(args->param, &sched_param, sizeof(sched_param));
1558 tdt = linux_tdfind(td, args->pid, -1);
1562 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1563 PROC_UNLOCK(tdt->td_proc);
1568 linux_sched_getscheduler(struct thread *td,
1569 struct linux_sched_getscheduler_args *args)
1575 if (ldebug(sched_getscheduler))
1576 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1579 tdt = linux_tdfind(td, args->pid, -1);
1583 error = kern_sched_getscheduler(td, tdt, &policy);
1584 PROC_UNLOCK(tdt->td_proc);
1588 td->td_retval[0] = LINUX_SCHED_OTHER;
1591 td->td_retval[0] = LINUX_SCHED_FIFO;
1594 td->td_retval[0] = LINUX_SCHED_RR;
1601 linux_sched_get_priority_max(struct thread *td,
1602 struct linux_sched_get_priority_max_args *args)
1604 struct sched_get_priority_max_args bsd;
1607 if (ldebug(sched_get_priority_max))
1608 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1611 switch (args->policy) {
1612 case LINUX_SCHED_OTHER:
1613 bsd.policy = SCHED_OTHER;
1615 case LINUX_SCHED_FIFO:
1616 bsd.policy = SCHED_FIFO;
1618 case LINUX_SCHED_RR:
1619 bsd.policy = SCHED_RR;
1624 return (sys_sched_get_priority_max(td, &bsd));
1628 linux_sched_get_priority_min(struct thread *td,
1629 struct linux_sched_get_priority_min_args *args)
1631 struct sched_get_priority_min_args bsd;
1634 if (ldebug(sched_get_priority_min))
1635 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1638 switch (args->policy) {
1639 case LINUX_SCHED_OTHER:
1640 bsd.policy = SCHED_OTHER;
1642 case LINUX_SCHED_FIFO:
1643 bsd.policy = SCHED_FIFO;
1645 case LINUX_SCHED_RR:
1646 bsd.policy = SCHED_RR;
1651 return (sys_sched_get_priority_min(td, &bsd));
1654 #define REBOOT_CAD_ON 0x89abcdef
1655 #define REBOOT_CAD_OFF 0
1656 #define REBOOT_HALT 0xcdef0123
1657 #define REBOOT_RESTART 0x01234567
1658 #define REBOOT_RESTART2 0xA1B2C3D4
1659 #define REBOOT_POWEROFF 0x4321FEDC
1660 #define REBOOT_MAGIC1 0xfee1dead
1661 #define REBOOT_MAGIC2 0x28121969
1662 #define REBOOT_MAGIC2A 0x05121996
1663 #define REBOOT_MAGIC2B 0x16041998
1666 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1668 struct reboot_args bsd_args;
1672 printf(ARGS(reboot, "0x%x"), args->cmd);
1675 if (args->magic1 != REBOOT_MAGIC1)
1678 switch (args->magic2) {
1680 case REBOOT_MAGIC2A:
1681 case REBOOT_MAGIC2B:
1687 switch (args->cmd) {
1689 case REBOOT_CAD_OFF:
1690 return (priv_check(td, PRIV_REBOOT));
1692 bsd_args.opt = RB_HALT;
1694 case REBOOT_RESTART:
1695 case REBOOT_RESTART2:
1698 case REBOOT_POWEROFF:
1699 bsd_args.opt = RB_POWEROFF;
1704 return (sys_reboot(td, &bsd_args));
1709 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1714 printf(ARGS(getpid, ""));
1716 td->td_retval[0] = td->td_proc->p_pid;
1722 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1724 struct linux_emuldata *em;
1728 printf(ARGS(gettid, ""));
1732 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1734 td->td_retval[0] = em->em_tid;
1741 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1745 if (ldebug(getppid))
1746 printf(ARGS(getppid, ""));
1749 td->td_retval[0] = kern_getppid(td);
1754 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1759 printf(ARGS(getgid, ""));
1762 td->td_retval[0] = td->td_ucred->cr_rgid;
1767 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1772 printf(ARGS(getuid, ""));
1775 td->td_retval[0] = td->td_ucred->cr_ruid;
1781 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1783 struct getsid_args bsd;
1787 printf(ARGS(getsid, "%i"), args->pid);
1790 bsd.pid = args->pid;
1791 return (sys_getsid(td, &bsd));
1795 linux_nosys(struct thread *td, struct nosys_args *ignore)
1802 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1804 struct getpriority_args bsd_args;
1808 if (ldebug(getpriority))
1809 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1812 bsd_args.which = args->which;
1813 bsd_args.who = args->who;
1814 error = sys_getpriority(td, &bsd_args);
1815 td->td_retval[0] = 20 - td->td_retval[0];
1820 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1825 if (ldebug(sethostname))
1826 printf(ARGS(sethostname, "*, %i"), args->len);
1830 name[1] = KERN_HOSTNAME;
1831 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1836 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1841 if (ldebug(setdomainname))
1842 printf(ARGS(setdomainname, "*, %i"), args->len);
1846 name[1] = KERN_NISDOMAINNAME;
1847 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1852 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1856 if (ldebug(exit_group))
1857 printf(ARGS(exit_group, "%i"), args->error_code);
1860 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1864 * XXX: we should send a signal to the parent if
1865 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1866 * as it doesnt occur often.
1868 exit1(td, args->error_code, 0);
1872 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1873 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1874 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1876 struct l_user_cap_header {
1881 struct l_user_cap_data {
1888 linux_capget(struct thread *td, struct linux_capget_args *uap)
1890 struct l_user_cap_header luch;
1891 struct l_user_cap_data lucd[2];
1894 if (uap->hdrp == NULL)
1897 error = copyin(uap->hdrp, &luch, sizeof(luch));
1901 switch (luch.version) {
1902 case _LINUX_CAPABILITY_VERSION_1:
1905 case _LINUX_CAPABILITY_VERSION_2:
1906 case _LINUX_CAPABILITY_VERSION_3:
1912 printf(LMSG("invalid capget capability version 0x%x"),
1915 luch.version = _LINUX_CAPABILITY_VERSION_1;
1916 error = copyout(&luch, uap->hdrp, sizeof(luch));
1927 * The current implementation doesn't support setting
1928 * a capability (it's essentially a stub) so indicate
1929 * that no capabilities are currently set or available
1932 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1933 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1940 linux_capset(struct thread *td, struct linux_capset_args *uap)
1942 struct l_user_cap_header luch;
1943 struct l_user_cap_data lucd[2];
1946 if (uap->hdrp == NULL || uap->datap == NULL)
1949 error = copyin(uap->hdrp, &luch, sizeof(luch));
1953 switch (luch.version) {
1954 case _LINUX_CAPABILITY_VERSION_1:
1957 case _LINUX_CAPABILITY_VERSION_2:
1958 case _LINUX_CAPABILITY_VERSION_3:
1964 printf(LMSG("invalid capset capability version 0x%x"),
1967 luch.version = _LINUX_CAPABILITY_VERSION_1;
1968 error = copyout(&luch, uap->hdrp, sizeof(luch));
1977 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1981 /* We currently don't support setting any capabilities. */
1982 for (i = 0; i < u32s; i++) {
1983 if (lucd[i].effective || lucd[i].permitted ||
1984 lucd[i].inheritable) {
1986 "capset[%d] effective=0x%x, permitted=0x%x, "
1987 "inheritable=0x%x is not implemented", i,
1988 (int)lucd[i].effective, (int)lucd[i].permitted,
1989 (int)lucd[i].inheritable);
1998 linux_prctl(struct thread *td, struct linux_prctl_args *args)
2000 int error = 0, max_size;
2001 struct proc *p = td->td_proc;
2002 char comm[LINUX_MAX_COMM_LEN];
2007 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
2008 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
2009 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
2012 switch (args->option) {
2013 case LINUX_PR_SET_PDEATHSIG:
2014 if (!LINUX_SIG_VALID(args->arg2))
2016 pdeath_signal = linux_to_bsd_signal(args->arg2);
2017 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
2019 case LINUX_PR_GET_PDEATHSIG:
2020 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
2024 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
2025 return (copyout(&pdeath_signal,
2026 (void *)(register_t)args->arg2,
2027 sizeof(pdeath_signal)));
2029 case LINUX_PR_GET_KEEPCAPS:
2031 * Indicate that we always clear the effective and
2032 * permitted capability sets when the user id becomes
2033 * non-zero (actually the capability sets are simply
2034 * always zero in the current implementation).
2036 td->td_retval[0] = 0;
2038 case LINUX_PR_SET_KEEPCAPS:
2040 * Ignore requests to keep the effective and permitted
2041 * capability sets when the user id becomes non-zero.
2044 case LINUX_PR_SET_NAME:
2046 * To be on the safe side we need to make sure to not
2047 * overflow the size a Linux program expects. We already
2048 * do this here in the copyin, so that we don't need to
2051 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2052 error = copyinstr((void *)(register_t)args->arg2, comm,
2055 /* Linux silently truncates the name if it is too long. */
2056 if (error == ENAMETOOLONG) {
2058 * XXX: copyinstr() isn't documented to populate the
2059 * array completely, so do a copyin() to be on the
2060 * safe side. This should be changed in case
2061 * copyinstr() is changed to guarantee this.
2063 error = copyin((void *)(register_t)args->arg2, comm,
2065 comm[max_size - 1] = '\0';
2071 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2074 case LINUX_PR_GET_NAME:
2076 strlcpy(comm, p->p_comm, sizeof(comm));
2078 error = copyout(comm, (void *)(register_t)args->arg2,
2090 linux_sched_setparam(struct thread *td,
2091 struct linux_sched_setparam_args *uap)
2093 struct sched_param sched_param;
2098 if (ldebug(sched_setparam))
2099 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2102 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2106 tdt = linux_tdfind(td, uap->pid, -1);
2110 error = kern_sched_setparam(td, tdt, &sched_param);
2111 PROC_UNLOCK(tdt->td_proc);
2116 linux_sched_getparam(struct thread *td,
2117 struct linux_sched_getparam_args *uap)
2119 struct sched_param sched_param;
2124 if (ldebug(sched_getparam))
2125 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2128 tdt = linux_tdfind(td, uap->pid, -1);
2132 error = kern_sched_getparam(td, tdt, &sched_param);
2133 PROC_UNLOCK(tdt->td_proc);
2135 error = copyout(&sched_param, uap->param,
2136 sizeof(sched_param));
2141 * Get affinity of a process.
2144 linux_sched_getaffinity(struct thread *td,
2145 struct linux_sched_getaffinity_args *args)
2151 if (ldebug(sched_getaffinity))
2152 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2155 if (args->len < sizeof(cpuset_t))
2158 tdt = linux_tdfind(td, args->pid, -1);
2162 PROC_UNLOCK(tdt->td_proc);
2164 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2165 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2167 td->td_retval[0] = sizeof(cpuset_t);
2173 * Set affinity of a process.
2176 linux_sched_setaffinity(struct thread *td,
2177 struct linux_sched_setaffinity_args *args)
2182 if (ldebug(sched_setaffinity))
2183 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2186 if (args->len < sizeof(cpuset_t))
2189 tdt = linux_tdfind(td, args->pid, -1);
2193 PROC_UNLOCK(tdt->td_proc);
2195 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2196 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2199 struct linux_rlimit64 {
2205 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2207 struct rlimit rlim, nrlim;
2208 struct linux_rlimit64 lrlim;
2215 if (ldebug(prlimit64))
2216 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2217 args->resource, (void *)args->new, (void *)args->old);
2220 if (args->resource >= LINUX_RLIM_NLIMITS)
2223 which = linux_to_bsd_resource[args->resource];
2227 if (args->new != NULL) {
2229 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2230 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2231 * as INFINITY so we do not need a conversion even.
2233 error = copyin(args->new, &nrlim, sizeof(nrlim));
2238 flags = PGET_HOLD | PGET_NOTWEXIT;
2239 if (args->new != NULL)
2240 flags |= PGET_CANDEBUG;
2242 flags |= PGET_CANSEE;
2243 error = pget(args->pid, flags, &p);
2247 if (args->old != NULL) {
2249 lim_rlimit_proc(p, which, &rlim);
2251 if (rlim.rlim_cur == RLIM_INFINITY)
2252 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2254 lrlim.rlim_cur = rlim.rlim_cur;
2255 if (rlim.rlim_max == RLIM_INFINITY)
2256 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2258 lrlim.rlim_max = rlim.rlim_max;
2259 error = copyout(&lrlim, args->old, sizeof(lrlim));
2264 if (args->new != NULL)
2265 error = kern_proc_setrlimit(td, p, which, &nrlim);
2273 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2275 struct timeval utv, tv0, tv1, *tvp;
2276 struct l_pselect6arg lpse6;
2277 struct l_timespec lts;
2278 struct timespec uts;
2285 if (args->sig != NULL) {
2286 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2289 if (lpse6.ss_len != sizeof(l_ss))
2291 if (lpse6.ss != 0) {
2292 error = copyin(PTRIN(lpse6.ss), &l_ss,
2296 linux_to_bsd_sigset(&l_ss, &ss);
2302 * Currently glibc changes nanosecond number to microsecond.
2303 * This mean losing precision but for now it is hardly seen.
2305 if (args->tsp != NULL) {
2306 error = copyin(args->tsp, <s, sizeof(lts));
2309 error = linux_to_native_timespec(&uts, <s);
2313 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2314 if (itimerfix(&utv))
2322 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2323 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2325 if (error == 0 && args->tsp != NULL) {
2326 if (td->td_retval[0] != 0) {
2328 * Compute how much time was left of the timeout,
2329 * by subtracting the current time and the time
2330 * before we started the call, and subtracting
2331 * that result from the user-supplied value.
2335 timevalsub(&tv1, &tv0);
2336 timevalsub(&utv, &tv1);
2342 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2344 error = native_to_linux_timespec(<s, &uts);
2346 error = copyout(<s, args->tsp, sizeof(lts));
2353 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2355 struct timespec ts0, ts1;
2356 struct l_timespec lts;
2357 struct timespec uts, *tsp;
2363 if (args->sset != NULL) {
2364 if (args->ssize != sizeof(l_ss))
2366 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2369 linux_to_bsd_sigset(&l_ss, &ss);
2373 if (args->tsp != NULL) {
2374 error = copyin(args->tsp, <s, sizeof(lts));
2377 error = linux_to_native_timespec(&uts, <s);
2386 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2388 if (error == 0 && args->tsp != NULL) {
2389 if (td->td_retval[0]) {
2391 timespecsub(&ts1, &ts0, &ts1);
2392 timespecsub(&uts, &ts1, &uts);
2394 timespecclear(&uts);
2396 timespecclear(&uts);
2398 error = native_to_linux_timespec(<s, &uts);
2400 error = copyout(<s, args->tsp, sizeof(lts));
2406 #if defined(DEBUG) || defined(KTR)
2407 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2409 #ifdef COMPAT_LINUX32
2410 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL
2412 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL
2415 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2418 linux_debug(int syscall, int toggle, int global)
2422 char c = toggle ? 0 : 0xff;
2424 memset(linux_debug_map, c, sizeof(linux_debug_map));
2427 if (syscall < 0 || syscall >= L_MAXSYSCALL)
2430 clrbit(linux_debug_map, syscall);
2432 setbit(linux_debug_map, syscall);
2438 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2440 * E.g.: sysctl linux.debug=21.0
2442 * As a special case, syscall "all" will apply to all syscalls globally.
2444 #define LINUX_MAX_DEBUGSTR 16
2446 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2448 char value[LINUX_MAX_DEBUGSTR], *p;
2449 int error, sysc, toggle;
2453 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2454 if (error || req->newptr == NULL)
2456 for (p = value; *p != '\0' && *p != '.'; p++);
2460 sysc = strtol(value, NULL, 0);
2461 toggle = strtol(p, NULL, 0);
2462 if (strcmp(value, "all") == 0)
2464 error = linux_debug(sysc, toggle, global);
2468 #endif /* DEBUG || KTR */
2471 linux_sched_rr_get_interval(struct thread *td,
2472 struct linux_sched_rr_get_interval_args *uap)
2475 struct l_timespec lts;
2480 * According to man in case the invalid pid specified
2481 * EINVAL should be returned.
2486 tdt = linux_tdfind(td, uap->pid, -1);
2490 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2491 PROC_UNLOCK(tdt->td_proc);
2494 error = native_to_linux_timespec(<s, &ts);
2497 return (copyout(<s, uap->interval, sizeof(lts)));
2501 * In case when the Linux thread is the initial thread in
2502 * the thread group thread id is equal to the process id.
2503 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2506 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2508 struct linux_emuldata *em;
2513 if (tid == 0 || tid == td->td_tid) {
2515 PROC_LOCK(tdt->td_proc);
2516 } else if (tid > PID_MAX)
2517 tdt = tdfind(tid, pid);
2520 * Initial thread where the tid equal to the pid.
2524 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2526 * p is not a Linuxulator process.
2531 FOREACH_THREAD_IN_PROC(p, tdt) {
2533 if (tid == em->em_tid)
2545 linux_to_bsd_waitopts(int options, int *bsdopts)
2548 if (options & LINUX_WNOHANG)
2549 *bsdopts |= WNOHANG;
2550 if (options & LINUX_WUNTRACED)
2551 *bsdopts |= WUNTRACED;
2552 if (options & LINUX_WEXITED)
2553 *bsdopts |= WEXITED;
2554 if (options & LINUX_WCONTINUED)
2555 *bsdopts |= WCONTINUED;
2556 if (options & LINUX_WNOWAIT)
2557 *bsdopts |= WNOWAIT;
2559 if (options & __WCLONE)
2560 *bsdopts |= WLINUXCLONE;
2564 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2570 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2572 if (args->count > INT_MAX)
2573 args->count = INT_MAX;
2575 iov.iov_base = args->buf;
2576 iov.iov_len = args->count;
2580 uio.uio_resid = iov.iov_len;
2581 uio.uio_segflg = UIO_USERSPACE;
2582 uio.uio_rw = UIO_READ;
2585 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2587 td->td_retval[0] = args->count - uio.uio_resid;
2592 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2595 /* Needs to be page-aligned */
2596 if (args->start & PAGE_MASK)
2598 return (kern_mincore(td, args->start, args->len, args->vec));