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/reboot.h>
55 #include <sys/racct.h>
56 #include <sys/random.h>
57 #include <sys/resourcevar.h>
58 #include <sys/sched.h>
60 #include <sys/signalvar.h>
62 #include <sys/syscallsubr.h>
63 #include <sys/sysctl.h>
64 #include <sys/sysproto.h>
65 #include <sys/systm.h>
67 #include <sys/vmmeter.h>
68 #include <sys/vnode.h>
70 #include <sys/cpuset.h>
73 #include <security/mac/mac_framework.h>
77 #include <vm/vm_kern.h>
78 #include <vm/vm_map.h>
79 #include <vm/vm_extern.h>
80 #include <vm/vm_object.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;
156 bzero(&sysinfo, sizeof(sysinfo));
160 sysinfo.uptime = ts.tv_sec;
162 /* Use the information from the mib to get our load averages */
163 for (i = 0; i < 3; i++)
164 sysinfo.loads[i] = averunnable.ldavg[i] *
165 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
167 sysinfo.totalram = physmem * PAGE_SIZE;
168 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
170 sysinfo.sharedram = 0;
171 mtx_lock(&vm_object_list_mtx);
172 TAILQ_FOREACH(object, &vm_object_list, object_list)
173 if (object->shadow_count > 1)
174 sysinfo.sharedram += object->resident_page_count;
175 mtx_unlock(&vm_object_list_mtx);
177 sysinfo.sharedram *= PAGE_SIZE;
178 sysinfo.bufferram = 0;
180 swap_pager_status(&i, &j);
181 sysinfo.totalswap = i * PAGE_SIZE;
182 sysinfo.freeswap = (i - j) * PAGE_SIZE;
184 sysinfo.procs = nprocs;
186 /* The following are only present in newer Linux kernels. */
187 sysinfo.totalbig = 0;
189 sysinfo.mem_unit = 1;
191 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
194 #ifdef LINUX_LEGACY_SYSCALLS
196 linux_alarm(struct thread *td, struct linux_alarm_args *args)
198 struct itimerval it, old_it;
204 printf(ARGS(alarm, "%u"), args->secs);
208 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
209 * to match kern_setitimer()'s limit to avoid error from it.
211 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
214 if (secs > INT32_MAX / 2)
215 secs = INT32_MAX / 2;
217 it.it_value.tv_sec = secs;
218 it.it_value.tv_usec = 0;
219 timevalclear(&it.it_interval);
220 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
221 KASSERT(error == 0, ("kern_setitimer returns %d", error));
223 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
224 old_it.it_value.tv_usec >= 500000)
225 old_it.it_value.tv_sec++;
226 td->td_retval[0] = old_it.it_value.tv_sec;
232 linux_brk(struct thread *td, struct linux_brk_args *args)
234 struct vmspace *vm = td->td_proc->p_vmspace;
235 vm_offset_t new, old;
236 struct break_args /* {
242 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
244 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
245 new = (vm_offset_t)args->dsend;
246 tmp.nsize = (char *)new;
247 if (((caddr_t)new > vm->vm_daddr) && !sys_break(td, &tmp))
248 td->td_retval[0] = (long)new;
250 td->td_retval[0] = (long)old;
255 #if defined(__i386__)
256 /* XXX: what about amd64/linux32? */
259 linux_uselib(struct thread *td, struct linux_uselib_args *args)
266 unsigned long file_offset;
267 unsigned long bss_size;
270 int error, locked, writecount;
272 LCONVPATHEXIST(td, args->library, &library);
276 printf(ARGS(uselib, "%s"), library);
283 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
284 UIO_SYSSPACE, library, td);
291 NDFREE(&ni, NDF_ONLY_PNBUF);
294 * From here on down, we have a locked vnode that must be unlocked.
295 * XXX: The code below largely duplicates exec_check_permissions().
300 error = VOP_GET_WRITECOUNT(vp, &writecount);
303 if (writecount != 0) {
309 error = VOP_GETATTR(vp, &attr, td->td_ucred);
313 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
314 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
315 /* EACCESS is what exec(2) returns. */
321 if (attr.va_size == 0) {
326 /* Can we access it? */
327 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
332 * XXX: This should use vn_open() so that it is properly authorized,
333 * and to reduce code redundancy all over the place here.
334 * XXX: Not really, it duplicates far more of exec_check_permissions()
338 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
342 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
346 /* Pull in executable header into exec_map */
347 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
348 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
352 /* Is it a Linux binary ? */
353 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
359 * While we are here, we should REALLY do some more checks
362 /* Set file/virtual offset based on a.out variant. */
363 switch ((int)(a_out->a_magic & 0xffff)) {
364 case 0413: /* ZMAGIC */
367 case 0314: /* QMAGIC */
375 bss_size = round_page(a_out->a_bss);
377 /* Check various fields in header for validity/bounds. */
378 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
383 /* text + data can't exceed file size */
384 if (a_out->a_data + a_out->a_text > attr.va_size) {
390 * text/data/bss must not exceed limits
391 * XXX - this is not complete. it should check current usage PLUS
392 * the resources needed by this library.
394 PROC_LOCK(td->td_proc);
395 if (a_out->a_text > maxtsiz ||
396 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
397 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
399 PROC_UNLOCK(td->td_proc);
403 PROC_UNLOCK(td->td_proc);
406 * Prevent more writers.
407 * XXX: Note that if any of the VM operations fail below we don't
413 * Lock no longer needed
419 * Check if file_offset page aligned. Currently we cannot handle
420 * misalinged file offsets, and so we read in the entire image
423 if (file_offset & PAGE_MASK) {
425 printf("uselib: Non page aligned binary %lu\n", file_offset);
427 /* Map text+data read/write/execute */
429 /* a_entry is the load address and is page aligned */
430 vmaddr = trunc_page(a_out->a_entry);
432 /* get anon user mapping, read+write+execute */
433 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
434 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
435 VM_PROT_ALL, VM_PROT_ALL, 0);
439 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
440 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
441 td->td_ucred, NOCRED, &aresid, td);
450 printf("uselib: Page aligned binary %lu\n", file_offset);
453 * for QMAGIC, a_entry is 20 bytes beyond the load address
454 * to skip the executable header
456 vmaddr = trunc_page(a_out->a_entry);
459 * Map it all into the process's space as a single
460 * copy-on-write "data" segment.
462 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
463 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
464 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
469 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
470 ((long *)vmaddr)[1]);
473 /* Calculate BSS start address */
474 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
477 /* allocate some 'anon' space */
478 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
479 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
486 /* Unlock vnode if needed */
490 /* Release the temporary mapping. */
492 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
497 #endif /* __i386__ */
499 #ifdef LINUX_LEGACY_SYSCALLS
501 linux_select(struct thread *td, struct linux_select_args *args)
504 struct timeval tv0, tv1, utv, *tvp;
509 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
510 (void *)args->readfds, (void *)args->writefds,
511 (void *)args->exceptfds, (void *)args->timeout);
515 * Store current time for computation of the amount of
519 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
521 utv.tv_sec = ltv.tv_sec;
522 utv.tv_usec = ltv.tv_usec;
525 printf(LMSG("incoming timeout (%jd/%ld)"),
526 (intmax_t)utv.tv_sec, utv.tv_usec);
529 if (itimerfix(&utv)) {
531 * The timeval was invalid. Convert it to something
532 * valid that will act as it does under Linux.
534 utv.tv_sec += utv.tv_usec / 1000000;
535 utv.tv_usec %= 1000000;
536 if (utv.tv_usec < 0) {
538 utv.tv_usec += 1000000;
548 error = kern_select(td, args->nfds, args->readfds, args->writefds,
549 args->exceptfds, tvp, LINUX_NFDBITS);
553 printf(LMSG("real select returns %d"), error);
559 if (td->td_retval[0]) {
561 * Compute how much time was left of the timeout,
562 * by subtracting the current time and the time
563 * before we started the call, and subtracting
564 * that result from the user-supplied value.
567 timevalsub(&tv1, &tv0);
568 timevalsub(&utv, &tv1);
575 printf(LMSG("outgoing timeout (%jd/%ld)"),
576 (intmax_t)utv.tv_sec, utv.tv_usec);
578 ltv.tv_sec = utv.tv_sec;
579 ltv.tv_usec = utv.tv_usec;
580 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
587 printf(LMSG("select_out -> %d"), error);
594 linux_mremap(struct thread *td, struct linux_mremap_args *args)
602 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
603 (void *)(uintptr_t)args->addr,
604 (unsigned long)args->old_len,
605 (unsigned long)args->new_len,
606 (unsigned long)args->flags);
609 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
610 td->td_retval[0] = 0;
615 * Check for the page alignment.
616 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
618 if (args->addr & PAGE_MASK) {
619 td->td_retval[0] = 0;
623 args->new_len = round_page(args->new_len);
624 args->old_len = round_page(args->old_len);
626 if (args->new_len > args->old_len) {
627 td->td_retval[0] = 0;
631 if (args->new_len < args->old_len) {
632 addr = args->addr + args->new_len;
633 len = args->old_len - args->new_len;
634 error = kern_munmap(td, addr, len);
637 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
641 #define LINUX_MS_ASYNC 0x0001
642 #define LINUX_MS_INVALIDATE 0x0002
643 #define LINUX_MS_SYNC 0x0004
646 linux_msync(struct thread *td, struct linux_msync_args *args)
649 return (kern_msync(td, args->addr, args->len,
650 args->fl & ~LINUX_MS_SYNC));
653 #ifdef LINUX_LEGACY_SYSCALLS
655 linux_time(struct thread *td, struct linux_time_args *args)
663 printf(ARGS(time, "*"));
668 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
670 td->td_retval[0] = tm;
675 struct l_times_argv {
678 l_clock_t tms_cutime;
679 l_clock_t tms_cstime;
684 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
685 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
686 * auxiliary vector entry.
690 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
691 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
693 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
694 CONVNTCK(r) : CONVOTCK(r))
697 linux_times(struct thread *td, struct linux_times_args *args)
699 struct timeval tv, utime, stime, cutime, cstime;
700 struct l_times_argv tms;
706 printf(ARGS(times, "*"));
709 if (args->buf != NULL) {
713 calcru(p, &utime, &stime);
715 calccru(p, &cutime, &cstime);
718 tms.tms_utime = CONVTCK(utime);
719 tms.tms_stime = CONVTCK(stime);
721 tms.tms_cutime = CONVTCK(cutime);
722 tms.tms_cstime = CONVTCK(cstime);
724 if ((error = copyout(&tms, args->buf, sizeof(tms))))
729 td->td_retval[0] = (int)CONVTCK(tv);
734 linux_newuname(struct thread *td, struct linux_newuname_args *args)
736 struct l_new_utsname utsname;
737 char osname[LINUX_MAX_UTSNAME];
738 char osrelease[LINUX_MAX_UTSNAME];
742 if (ldebug(newuname))
743 printf(ARGS(newuname, "*"));
746 linux_get_osname(td, osname);
747 linux_get_osrelease(td, osrelease);
749 bzero(&utsname, sizeof(utsname));
750 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
751 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
752 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
753 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
754 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
755 for (p = utsname.version; *p != '\0'; ++p)
760 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
762 return (copyout(&utsname, args->buf, sizeof(utsname)));
770 #ifdef LINUX_LEGACY_SYSCALLS
772 linux_utime(struct thread *td, struct linux_utime_args *args)
774 struct timeval tv[2], *tvp;
775 struct l_utimbuf lut;
779 LCONVPATHEXIST(td, args->fname, &fname);
783 printf(ARGS(utime, "%s, *"), fname);
787 if ((error = copyin(args->times, &lut, sizeof lut))) {
791 tv[0].tv_sec = lut.l_actime;
793 tv[1].tv_sec = lut.l_modtime;
799 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
806 #ifdef LINUX_LEGACY_SYSCALLS
808 linux_utimes(struct thread *td, struct linux_utimes_args *args)
811 struct timeval tv[2], *tvp = NULL;
815 LCONVPATHEXIST(td, args->fname, &fname);
819 printf(ARGS(utimes, "%s, *"), fname);
822 if (args->tptr != NULL) {
823 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
827 tv[0].tv_sec = ltv[0].tv_sec;
828 tv[0].tv_usec = ltv[0].tv_usec;
829 tv[1].tv_sec = ltv[1].tv_sec;
830 tv[1].tv_usec = ltv[1].tv_usec;
834 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
842 linux_utimensat_nsec_valid(l_long nsec)
845 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
847 if (nsec >= 0 && nsec <= 999999999)
853 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
855 struct l_timespec l_times[2];
856 struct timespec times[2], *timesp = NULL;
858 int error, dfd, flags = 0;
860 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
863 if (ldebug(utimensat))
864 printf(ARGS(utimensat, "%d, *"), dfd);
867 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
870 if (args->times != NULL) {
871 error = copyin(args->times, l_times, sizeof(l_times));
875 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
876 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
879 times[0].tv_sec = l_times[0].tv_sec;
880 switch (l_times[0].tv_nsec)
882 case LINUX_UTIME_OMIT:
883 times[0].tv_nsec = UTIME_OMIT;
885 case LINUX_UTIME_NOW:
886 times[0].tv_nsec = UTIME_NOW;
889 times[0].tv_nsec = l_times[0].tv_nsec;
892 times[1].tv_sec = l_times[1].tv_sec;
893 switch (l_times[1].tv_nsec)
895 case LINUX_UTIME_OMIT:
896 times[1].tv_nsec = UTIME_OMIT;
898 case LINUX_UTIME_NOW:
899 times[1].tv_nsec = UTIME_NOW;
902 times[1].tv_nsec = l_times[1].tv_nsec;
907 /* This breaks POSIX, but is what the Linux kernel does
908 * _on purpose_ (documented in the man page for utimensat(2)),
909 * so we must follow that behaviour. */
910 if (times[0].tv_nsec == UTIME_OMIT &&
911 times[1].tv_nsec == UTIME_OMIT)
915 if (args->pathname != NULL)
916 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
917 else if (args->flags != 0)
920 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
921 flags |= AT_SYMLINK_NOFOLLOW;
924 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
926 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
927 UIO_SYSSPACE, flags);
934 #ifdef LINUX_LEGACY_SYSCALLS
936 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
939 struct timeval tv[2], *tvp = NULL;
943 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
944 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
947 if (ldebug(futimesat))
948 printf(ARGS(futimesat, "%s, *"), fname);
951 if (args->utimes != NULL) {
952 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
956 tv[0].tv_sec = ltv[0].tv_sec;
957 tv[0].tv_usec = ltv[0].tv_usec;
958 tv[1].tv_sec = ltv[1].tv_sec;
959 tv[1].tv_usec = ltv[1].tv_usec;
963 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
970 linux_common_wait(struct thread *td, int pid, int *status,
971 int options, struct rusage *ru)
975 error = kern_wait(td, pid, &tmpstat, options, ru);
981 if (WIFSIGNALED(tmpstat))
982 tmpstat = (tmpstat & 0xffffff80) |
983 bsd_to_linux_signal(WTERMSIG(tmpstat));
984 else if (WIFSTOPPED(tmpstat))
985 tmpstat = (tmpstat & 0xffff00ff) |
986 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
987 else if (WIFCONTINUED(tmpstat))
989 error = copyout(&tmpstat, status, sizeof(int));
995 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
997 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
999 struct linux_wait4_args wait4_args;
1002 if (ldebug(waitpid))
1003 printf(ARGS(waitpid, "%d, %p, %d"),
1004 args->pid, (void *)args->status, args->options);
1007 wait4_args.pid = args->pid;
1008 wait4_args.status = args->status;
1009 wait4_args.options = args->options;
1010 wait4_args.rusage = NULL;
1012 return (linux_wait4(td, &wait4_args));
1014 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1017 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1020 struct rusage ru, *rup;
1024 printf(ARGS(wait4, "%d, %p, %d, %p"),
1025 args->pid, (void *)args->status, args->options,
1026 (void *)args->rusage);
1028 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1029 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1033 linux_to_bsd_waitopts(args->options, &options);
1035 if (args->rusage != NULL)
1039 error = linux_common_wait(td, args->pid, args->status, options, rup);
1042 if (args->rusage != NULL)
1043 error = linux_copyout_rusage(&ru, args->rusage);
1048 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1050 int status, options, sig;
1051 struct __wrusage wru;
1059 linux_to_bsd_waitopts(args->options, &options);
1061 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1063 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1066 switch (args->idtype) {
1084 error = kern_wait6(td, idtype, args->id, &status, options,
1088 if (args->rusage != NULL) {
1089 error = linux_copyout_rusage(&wru.wru_children,
1094 if (args->info != NULL) {
1096 if (td->td_retval[0] == 0)
1097 bzero(&lsi, sizeof(lsi));
1099 sig = bsd_to_linux_signal(siginfo.si_signo);
1100 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1102 error = copyout(&lsi, args->info, sizeof(lsi));
1104 td->td_retval[0] = 0;
1109 #ifdef LINUX_LEGACY_SYSCALLS
1111 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1116 LCONVPATHCREAT(td, args->path, &path);
1120 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1121 (uintmax_t)args->dev);
1124 switch (args->mode & S_IFMT) {
1127 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1133 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1134 args->mode, args->dev);
1142 args->mode |= S_IFREG;
1145 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1146 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1148 kern_close(td, td->td_retval[0]);
1161 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1166 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1167 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1170 if (ldebug(mknodat))
1171 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1174 switch (args->mode & S_IFMT) {
1177 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1182 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1191 args->mode |= S_IFREG;
1194 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1195 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1197 kern_close(td, td->td_retval[0]);
1209 * UGH! This is just about the dumbest idea I've ever heard!!
1212 linux_personality(struct thread *td, struct linux_personality_args *args)
1214 struct linux_pemuldata *pem;
1215 struct proc *p = td->td_proc;
1219 if (ldebug(personality))
1220 printf(ARGS(personality, "%u"), args->per);
1226 if (args->per != 0xffffffff)
1227 pem->persona = args->per;
1230 td->td_retval[0] = old;
1234 struct l_itimerval {
1235 l_timeval it_interval;
1239 #define B2L_ITIMERVAL(bip, lip) \
1240 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1241 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1242 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1243 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1246 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1249 struct l_itimerval ls;
1250 struct itimerval aitv, oitv;
1253 if (ldebug(setitimer))
1254 printf(ARGS(setitimer, "%p, %p"),
1255 (void *)uap->itv, (void *)uap->oitv);
1258 if (uap->itv == NULL) {
1259 uap->itv = uap->oitv;
1260 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1263 error = copyin(uap->itv, &ls, sizeof(ls));
1266 B2L_ITIMERVAL(&aitv, &ls);
1268 if (ldebug(setitimer)) {
1269 printf("setitimer: value: sec: %jd, usec: %ld\n",
1270 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1271 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1272 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1275 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1276 if (error != 0 || uap->oitv == NULL)
1278 B2L_ITIMERVAL(&ls, &oitv);
1280 return (copyout(&ls, uap->oitv, sizeof(ls)));
1284 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1287 struct l_itimerval ls;
1288 struct itimerval aitv;
1291 if (ldebug(getitimer))
1292 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1294 error = kern_getitimer(td, uap->which, &aitv);
1297 B2L_ITIMERVAL(&ls, &aitv);
1298 return (copyout(&ls, uap->itv, sizeof(ls)));
1301 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1303 linux_nice(struct thread *td, struct linux_nice_args *args)
1305 struct setpriority_args bsd_args;
1307 bsd_args.which = PRIO_PROCESS;
1308 bsd_args.who = 0; /* current process */
1309 bsd_args.prio = args->inc;
1310 return (sys_setpriority(td, &bsd_args));
1312 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1315 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1317 struct ucred *newcred, *oldcred;
1318 l_gid_t *linux_gidset;
1323 ngrp = args->gidsetsize;
1324 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1326 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1327 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1331 crextend(newcred, ngrp + 1);
1334 oldcred = p->p_ucred;
1335 crcopy(newcred, oldcred);
1338 * cr_groups[0] holds egid. Setting the whole set from
1339 * the supplied set will cause egid to be changed too.
1340 * Keep cr_groups[0] unchanged to prevent that.
1343 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1350 newcred->cr_ngroups = ngrp + 1;
1352 bsd_gidset = newcred->cr_groups;
1355 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1359 newcred->cr_ngroups = 1;
1362 proc_set_cred(p, newcred);
1367 free(linux_gidset, M_LINUX);
1372 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1375 l_gid_t *linux_gidset;
1377 int bsd_gidsetsz, ngrp, error;
1379 cred = td->td_ucred;
1380 bsd_gidset = cred->cr_groups;
1381 bsd_gidsetsz = cred->cr_ngroups - 1;
1384 * cr_groups[0] holds egid. Returning the whole set
1385 * here will cause a duplicate. Exclude cr_groups[0]
1389 if ((ngrp = args->gidsetsize) == 0) {
1390 td->td_retval[0] = bsd_gidsetsz;
1394 if (ngrp < bsd_gidsetsz)
1398 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1400 while (ngrp < bsd_gidsetsz) {
1401 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1405 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1406 free(linux_gidset, M_LINUX);
1410 td->td_retval[0] = ngrp;
1415 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1417 struct rlimit bsd_rlim;
1418 struct l_rlimit rlim;
1423 if (ldebug(setrlimit))
1424 printf(ARGS(setrlimit, "%d, %p"),
1425 args->resource, (void *)args->rlim);
1428 if (args->resource >= LINUX_RLIM_NLIMITS)
1431 which = linux_to_bsd_resource[args->resource];
1435 error = copyin(args->rlim, &rlim, sizeof(rlim));
1439 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1440 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1441 return (kern_setrlimit(td, which, &bsd_rlim));
1444 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1446 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1448 struct l_rlimit rlim;
1449 struct rlimit bsd_rlim;
1453 if (ldebug(old_getrlimit))
1454 printf(ARGS(old_getrlimit, "%d, %p"),
1455 args->resource, (void *)args->rlim);
1458 if (args->resource >= LINUX_RLIM_NLIMITS)
1461 which = linux_to_bsd_resource[args->resource];
1465 lim_rlimit(td, which, &bsd_rlim);
1467 #ifdef COMPAT_LINUX32
1468 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1469 if (rlim.rlim_cur == UINT_MAX)
1470 rlim.rlim_cur = INT_MAX;
1471 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1472 if (rlim.rlim_max == UINT_MAX)
1473 rlim.rlim_max = INT_MAX;
1475 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1476 if (rlim.rlim_cur == ULONG_MAX)
1477 rlim.rlim_cur = LONG_MAX;
1478 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1479 if (rlim.rlim_max == ULONG_MAX)
1480 rlim.rlim_max = LONG_MAX;
1482 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1484 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1487 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1489 struct l_rlimit rlim;
1490 struct rlimit bsd_rlim;
1494 if (ldebug(getrlimit))
1495 printf(ARGS(getrlimit, "%d, %p"),
1496 args->resource, (void *)args->rlim);
1499 if (args->resource >= LINUX_RLIM_NLIMITS)
1502 which = linux_to_bsd_resource[args->resource];
1506 lim_rlimit(td, which, &bsd_rlim);
1508 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1509 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1510 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1514 linux_sched_setscheduler(struct thread *td,
1515 struct linux_sched_setscheduler_args *args)
1517 struct sched_param sched_param;
1522 if (ldebug(sched_setscheduler))
1523 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1524 args->pid, args->policy, (const void *)args->param);
1527 switch (args->policy) {
1528 case LINUX_SCHED_OTHER:
1529 policy = SCHED_OTHER;
1531 case LINUX_SCHED_FIFO:
1532 policy = SCHED_FIFO;
1534 case LINUX_SCHED_RR:
1541 error = copyin(args->param, &sched_param, sizeof(sched_param));
1545 tdt = linux_tdfind(td, args->pid, -1);
1549 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1550 PROC_UNLOCK(tdt->td_proc);
1555 linux_sched_getscheduler(struct thread *td,
1556 struct linux_sched_getscheduler_args *args)
1562 if (ldebug(sched_getscheduler))
1563 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1566 tdt = linux_tdfind(td, args->pid, -1);
1570 error = kern_sched_getscheduler(td, tdt, &policy);
1571 PROC_UNLOCK(tdt->td_proc);
1575 td->td_retval[0] = LINUX_SCHED_OTHER;
1578 td->td_retval[0] = LINUX_SCHED_FIFO;
1581 td->td_retval[0] = LINUX_SCHED_RR;
1588 linux_sched_get_priority_max(struct thread *td,
1589 struct linux_sched_get_priority_max_args *args)
1591 struct sched_get_priority_max_args bsd;
1594 if (ldebug(sched_get_priority_max))
1595 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1598 switch (args->policy) {
1599 case LINUX_SCHED_OTHER:
1600 bsd.policy = SCHED_OTHER;
1602 case LINUX_SCHED_FIFO:
1603 bsd.policy = SCHED_FIFO;
1605 case LINUX_SCHED_RR:
1606 bsd.policy = SCHED_RR;
1611 return (sys_sched_get_priority_max(td, &bsd));
1615 linux_sched_get_priority_min(struct thread *td,
1616 struct linux_sched_get_priority_min_args *args)
1618 struct sched_get_priority_min_args bsd;
1621 if (ldebug(sched_get_priority_min))
1622 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1625 switch (args->policy) {
1626 case LINUX_SCHED_OTHER:
1627 bsd.policy = SCHED_OTHER;
1629 case LINUX_SCHED_FIFO:
1630 bsd.policy = SCHED_FIFO;
1632 case LINUX_SCHED_RR:
1633 bsd.policy = SCHED_RR;
1638 return (sys_sched_get_priority_min(td, &bsd));
1641 #define REBOOT_CAD_ON 0x89abcdef
1642 #define REBOOT_CAD_OFF 0
1643 #define REBOOT_HALT 0xcdef0123
1644 #define REBOOT_RESTART 0x01234567
1645 #define REBOOT_RESTART2 0xA1B2C3D4
1646 #define REBOOT_POWEROFF 0x4321FEDC
1647 #define REBOOT_MAGIC1 0xfee1dead
1648 #define REBOOT_MAGIC2 0x28121969
1649 #define REBOOT_MAGIC2A 0x05121996
1650 #define REBOOT_MAGIC2B 0x16041998
1653 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1655 struct reboot_args bsd_args;
1659 printf(ARGS(reboot, "0x%x"), args->cmd);
1662 if (args->magic1 != REBOOT_MAGIC1)
1665 switch (args->magic2) {
1667 case REBOOT_MAGIC2A:
1668 case REBOOT_MAGIC2B:
1674 switch (args->cmd) {
1676 case REBOOT_CAD_OFF:
1677 return (priv_check(td, PRIV_REBOOT));
1679 bsd_args.opt = RB_HALT;
1681 case REBOOT_RESTART:
1682 case REBOOT_RESTART2:
1685 case REBOOT_POWEROFF:
1686 bsd_args.opt = RB_POWEROFF;
1691 return (sys_reboot(td, &bsd_args));
1696 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1697 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1698 * are assumed to be preserved. The following lightweight syscalls fixes
1699 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1701 * linux_getpid() - MP SAFE
1702 * linux_getgid() - MP SAFE
1703 * linux_getuid() - MP SAFE
1707 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1712 printf(ARGS(getpid, ""));
1714 td->td_retval[0] = td->td_proc->p_pid;
1720 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1722 struct linux_emuldata *em;
1726 printf(ARGS(gettid, ""));
1730 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1732 td->td_retval[0] = em->em_tid;
1739 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1743 if (ldebug(getppid))
1744 printf(ARGS(getppid, ""));
1747 td->td_retval[0] = kern_getppid(td);
1752 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1757 printf(ARGS(getgid, ""));
1760 td->td_retval[0] = td->td_ucred->cr_rgid;
1765 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1770 printf(ARGS(getuid, ""));
1773 td->td_retval[0] = td->td_ucred->cr_ruid;
1779 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1781 struct getsid_args bsd;
1785 printf(ARGS(getsid, "%i"), args->pid);
1788 bsd.pid = args->pid;
1789 return (sys_getsid(td, &bsd));
1793 linux_nosys(struct thread *td, struct nosys_args *ignore)
1800 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1802 struct getpriority_args bsd_args;
1806 if (ldebug(getpriority))
1807 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1810 bsd_args.which = args->which;
1811 bsd_args.who = args->who;
1812 error = sys_getpriority(td, &bsd_args);
1813 td->td_retval[0] = 20 - td->td_retval[0];
1818 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1823 if (ldebug(sethostname))
1824 printf(ARGS(sethostname, "*, %i"), args->len);
1828 name[1] = KERN_HOSTNAME;
1829 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1834 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1839 if (ldebug(setdomainname))
1840 printf(ARGS(setdomainname, "*, %i"), args->len);
1844 name[1] = KERN_NISDOMAINNAME;
1845 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1850 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1854 if (ldebug(exit_group))
1855 printf(ARGS(exit_group, "%i"), args->error_code);
1858 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1862 * XXX: we should send a signal to the parent if
1863 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1864 * as it doesnt occur often.
1866 exit1(td, args->error_code, 0);
1870 #define _LINUX_CAPABILITY_VERSION 0x19980330
1872 struct l_user_cap_header {
1877 struct l_user_cap_data {
1884 linux_capget(struct thread *td, struct linux_capget_args *args)
1886 struct l_user_cap_header luch;
1887 struct l_user_cap_data lucd;
1890 if (args->hdrp == NULL)
1893 error = copyin(args->hdrp, &luch, sizeof(luch));
1897 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1898 luch.version = _LINUX_CAPABILITY_VERSION;
1899 error = copyout(&luch, args->hdrp, sizeof(luch));
1910 * The current implementation doesn't support setting
1911 * a capability (it's essentially a stub) so indicate
1912 * that no capabilities are currently set or available
1915 bzero (&lucd, sizeof(lucd));
1916 error = copyout(&lucd, args->datap, sizeof(lucd));
1923 linux_capset(struct thread *td, struct linux_capset_args *args)
1925 struct l_user_cap_header luch;
1926 struct l_user_cap_data lucd;
1929 if (args->hdrp == NULL || args->datap == NULL)
1932 error = copyin(args->hdrp, &luch, sizeof(luch));
1936 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1937 luch.version = _LINUX_CAPABILITY_VERSION;
1938 error = copyout(&luch, args->hdrp, sizeof(luch));
1947 error = copyin(args->datap, &lucd, sizeof(lucd));
1951 /* We currently don't support setting any capabilities. */
1952 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1954 "capset effective=0x%x, permitted=0x%x, "
1955 "inheritable=0x%x is not implemented",
1956 (int)lucd.effective, (int)lucd.permitted,
1957 (int)lucd.inheritable);
1965 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1967 int error = 0, max_size;
1968 struct proc *p = td->td_proc;
1969 char comm[LINUX_MAX_COMM_LEN];
1970 struct linux_emuldata *em;
1975 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1976 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1977 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1980 switch (args->option) {
1981 case LINUX_PR_SET_PDEATHSIG:
1982 if (!LINUX_SIG_VALID(args->arg2))
1985 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1986 em->pdeath_signal = args->arg2;
1988 case LINUX_PR_GET_PDEATHSIG:
1990 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1991 pdeath_signal = em->pdeath_signal;
1992 error = copyout(&pdeath_signal,
1993 (void *)(register_t)args->arg2,
1994 sizeof(pdeath_signal));
1996 case LINUX_PR_GET_KEEPCAPS:
1998 * Indicate that we always clear the effective and
1999 * permitted capability sets when the user id becomes
2000 * non-zero (actually the capability sets are simply
2001 * always zero in the current implementation).
2003 td->td_retval[0] = 0;
2005 case LINUX_PR_SET_KEEPCAPS:
2007 * Ignore requests to keep the effective and permitted
2008 * capability sets when the user id becomes non-zero.
2011 case LINUX_PR_SET_NAME:
2013 * To be on the safe side we need to make sure to not
2014 * overflow the size a Linux program expects. We already
2015 * do this here in the copyin, so that we don't need to
2018 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2019 error = copyinstr((void *)(register_t)args->arg2, comm,
2022 /* Linux silently truncates the name if it is too long. */
2023 if (error == ENAMETOOLONG) {
2025 * XXX: copyinstr() isn't documented to populate the
2026 * array completely, so do a copyin() to be on the
2027 * safe side. This should be changed in case
2028 * copyinstr() is changed to guarantee this.
2030 error = copyin((void *)(register_t)args->arg2, comm,
2032 comm[max_size - 1] = '\0';
2038 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2041 case LINUX_PR_GET_NAME:
2043 strlcpy(comm, p->p_comm, sizeof(comm));
2045 error = copyout(comm, (void *)(register_t)args->arg2,
2057 linux_sched_setparam(struct thread *td,
2058 struct linux_sched_setparam_args *uap)
2060 struct sched_param sched_param;
2065 if (ldebug(sched_setparam))
2066 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2069 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2073 tdt = linux_tdfind(td, uap->pid, -1);
2077 error = kern_sched_setparam(td, tdt, &sched_param);
2078 PROC_UNLOCK(tdt->td_proc);
2083 linux_sched_getparam(struct thread *td,
2084 struct linux_sched_getparam_args *uap)
2086 struct sched_param sched_param;
2091 if (ldebug(sched_getparam))
2092 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2095 tdt = linux_tdfind(td, uap->pid, -1);
2099 error = kern_sched_getparam(td, tdt, &sched_param);
2100 PROC_UNLOCK(tdt->td_proc);
2102 error = copyout(&sched_param, uap->param,
2103 sizeof(sched_param));
2108 * Get affinity of a process.
2111 linux_sched_getaffinity(struct thread *td,
2112 struct linux_sched_getaffinity_args *args)
2118 if (ldebug(sched_getaffinity))
2119 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2122 if (args->len < sizeof(cpuset_t))
2125 tdt = linux_tdfind(td, args->pid, -1);
2129 PROC_UNLOCK(tdt->td_proc);
2131 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2132 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2134 td->td_retval[0] = sizeof(cpuset_t);
2140 * Set affinity of a process.
2143 linux_sched_setaffinity(struct thread *td,
2144 struct linux_sched_setaffinity_args *args)
2149 if (ldebug(sched_setaffinity))
2150 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2153 if (args->len < sizeof(cpuset_t))
2156 tdt = linux_tdfind(td, args->pid, -1);
2160 PROC_UNLOCK(tdt->td_proc);
2162 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2163 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2166 struct linux_rlimit64 {
2172 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2174 struct rlimit rlim, nrlim;
2175 struct linux_rlimit64 lrlim;
2182 if (ldebug(prlimit64))
2183 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2184 args->resource, (void *)args->new, (void *)args->old);
2187 if (args->resource >= LINUX_RLIM_NLIMITS)
2190 which = linux_to_bsd_resource[args->resource];
2194 if (args->new != NULL) {
2196 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2197 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2198 * as INFINITY so we do not need a conversion even.
2200 error = copyin(args->new, &nrlim, sizeof(nrlim));
2205 flags = PGET_HOLD | PGET_NOTWEXIT;
2206 if (args->new != NULL)
2207 flags |= PGET_CANDEBUG;
2209 flags |= PGET_CANSEE;
2210 error = pget(args->pid, flags, &p);
2214 if (args->old != NULL) {
2216 lim_rlimit_proc(p, which, &rlim);
2218 if (rlim.rlim_cur == RLIM_INFINITY)
2219 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2221 lrlim.rlim_cur = rlim.rlim_cur;
2222 if (rlim.rlim_max == RLIM_INFINITY)
2223 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2225 lrlim.rlim_max = rlim.rlim_max;
2226 error = copyout(&lrlim, args->old, sizeof(lrlim));
2231 if (args->new != NULL)
2232 error = kern_proc_setrlimit(td, p, which, &nrlim);
2240 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2242 struct timeval utv, tv0, tv1, *tvp;
2243 struct l_pselect6arg lpse6;
2244 struct l_timespec lts;
2245 struct timespec uts;
2252 if (args->sig != NULL) {
2253 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2256 if (lpse6.ss_len != sizeof(l_ss))
2258 if (lpse6.ss != 0) {
2259 error = copyin(PTRIN(lpse6.ss), &l_ss,
2263 linux_to_bsd_sigset(&l_ss, &ss);
2269 * Currently glibc changes nanosecond number to microsecond.
2270 * This mean losing precision but for now it is hardly seen.
2272 if (args->tsp != NULL) {
2273 error = copyin(args->tsp, <s, sizeof(lts));
2276 error = linux_to_native_timespec(&uts, <s);
2280 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2281 if (itimerfix(&utv))
2289 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2290 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2292 if (error == 0 && args->tsp != NULL) {
2293 if (td->td_retval[0] != 0) {
2295 * Compute how much time was left of the timeout,
2296 * by subtracting the current time and the time
2297 * before we started the call, and subtracting
2298 * that result from the user-supplied value.
2302 timevalsub(&tv1, &tv0);
2303 timevalsub(&utv, &tv1);
2309 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2311 error = native_to_linux_timespec(<s, &uts);
2313 error = copyout(<s, args->tsp, sizeof(lts));
2320 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2322 struct timespec ts0, ts1;
2323 struct l_timespec lts;
2324 struct timespec uts, *tsp;
2330 if (args->sset != NULL) {
2331 if (args->ssize != sizeof(l_ss))
2333 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2336 linux_to_bsd_sigset(&l_ss, &ss);
2340 if (args->tsp != NULL) {
2341 error = copyin(args->tsp, <s, sizeof(lts));
2344 error = linux_to_native_timespec(&uts, <s);
2353 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2355 if (error == 0 && args->tsp != NULL) {
2356 if (td->td_retval[0]) {
2358 timespecsub(&ts1, &ts0);
2359 timespecsub(&uts, &ts1);
2361 timespecclear(&uts);
2363 timespecclear(&uts);
2365 error = native_to_linux_timespec(<s, &uts);
2367 error = copyout(<s, args->tsp, sizeof(lts));
2373 #if defined(DEBUG) || defined(KTR)
2374 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2376 #ifdef COMPAT_LINUX32
2377 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL
2379 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL
2382 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2385 linux_debug(int syscall, int toggle, int global)
2389 char c = toggle ? 0 : 0xff;
2391 memset(linux_debug_map, c, sizeof(linux_debug_map));
2394 if (syscall < 0 || syscall >= L_MAXSYSCALL)
2397 clrbit(linux_debug_map, syscall);
2399 setbit(linux_debug_map, syscall);
2405 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2407 * E.g.: sysctl linux.debug=21.0
2409 * As a special case, syscall "all" will apply to all syscalls globally.
2411 #define LINUX_MAX_DEBUGSTR 16
2413 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2415 char value[LINUX_MAX_DEBUGSTR], *p;
2416 int error, sysc, toggle;
2420 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2421 if (error || req->newptr == NULL)
2423 for (p = value; *p != '\0' && *p != '.'; p++);
2427 sysc = strtol(value, NULL, 0);
2428 toggle = strtol(p, NULL, 0);
2429 if (strcmp(value, "all") == 0)
2431 error = linux_debug(sysc, toggle, global);
2435 #endif /* DEBUG || KTR */
2438 linux_sched_rr_get_interval(struct thread *td,
2439 struct linux_sched_rr_get_interval_args *uap)
2442 struct l_timespec lts;
2447 * According to man in case the invalid pid specified
2448 * EINVAL should be returned.
2453 tdt = linux_tdfind(td, uap->pid, -1);
2457 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2458 PROC_UNLOCK(tdt->td_proc);
2461 error = native_to_linux_timespec(<s, &ts);
2464 return (copyout(<s, uap->interval, sizeof(lts)));
2468 * In case when the Linux thread is the initial thread in
2469 * the thread group thread id is equal to the process id.
2470 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2473 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2475 struct linux_emuldata *em;
2480 if (tid == 0 || tid == td->td_tid) {
2482 PROC_LOCK(tdt->td_proc);
2483 } else if (tid > PID_MAX)
2484 tdt = tdfind(tid, pid);
2487 * Initial thread where the tid equal to the pid.
2491 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2493 * p is not a Linuxulator process.
2498 FOREACH_THREAD_IN_PROC(p, tdt) {
2500 if (tid == em->em_tid)
2512 linux_to_bsd_waitopts(int options, int *bsdopts)
2515 if (options & LINUX_WNOHANG)
2516 *bsdopts |= WNOHANG;
2517 if (options & LINUX_WUNTRACED)
2518 *bsdopts |= WUNTRACED;
2519 if (options & LINUX_WEXITED)
2520 *bsdopts |= WEXITED;
2521 if (options & LINUX_WCONTINUED)
2522 *bsdopts |= WCONTINUED;
2523 if (options & LINUX_WNOWAIT)
2524 *bsdopts |= WNOWAIT;
2526 if (options & __WCLONE)
2527 *bsdopts |= WLINUXCLONE;
2531 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2537 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2539 if (args->count > INT_MAX)
2540 args->count = INT_MAX;
2542 iov.iov_base = args->buf;
2543 iov.iov_len = args->count;
2547 uio.uio_resid = iov.iov_len;
2548 uio.uio_segflg = UIO_USERSPACE;
2549 uio.uio_rw = UIO_READ;
2552 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2554 td->td_retval[0] = args->count - uio.uio_resid;
2559 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2562 /* Needs to be page-aligned */
2563 if (args->start & PAGE_MASK)
2565 return (kern_mincore(td, args->start, args->len, args->vec));