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)));
195 linux_alarm(struct thread *td, struct linux_alarm_args *args)
197 struct itimerval it, old_it;
203 printf(ARGS(alarm, "%u"), args->secs);
207 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
208 * to match kern_setitimer()'s limit to avoid error from it.
210 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
213 if (secs > INT32_MAX / 2)
214 secs = INT32_MAX / 2;
216 it.it_value.tv_sec = secs;
217 it.it_value.tv_usec = 0;
218 timevalclear(&it.it_interval);
219 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
220 KASSERT(error == 0, ("kern_setitimer returns %d", error));
222 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
223 old_it.it_value.tv_usec >= 500000)
224 old_it.it_value.tv_sec++;
225 td->td_retval[0] = old_it.it_value.tv_sec;
230 linux_brk(struct thread *td, struct linux_brk_args *args)
232 struct vmspace *vm = td->td_proc->p_vmspace;
233 vm_offset_t new, old;
234 struct obreak_args /* {
240 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
242 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
243 new = (vm_offset_t)args->dsend;
244 tmp.nsize = (char *)new;
245 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
246 td->td_retval[0] = (long)new;
248 td->td_retval[0] = (long)old;
253 #if defined(__i386__)
254 /* XXX: what about amd64/linux32? */
257 linux_uselib(struct thread *td, struct linux_uselib_args *args)
264 unsigned long file_offset;
265 unsigned long bss_size;
268 int error, locked, writecount;
270 LCONVPATHEXIST(td, args->library, &library);
274 printf(ARGS(uselib, "%s"), library);
281 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
282 UIO_SYSSPACE, library, td);
289 NDFREE(&ni, NDF_ONLY_PNBUF);
292 * From here on down, we have a locked vnode that must be unlocked.
293 * XXX: The code below largely duplicates exec_check_permissions().
298 error = VOP_GET_WRITECOUNT(vp, &writecount);
301 if (writecount != 0) {
307 error = VOP_GETATTR(vp, &attr, td->td_ucred);
311 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
312 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
313 /* EACCESS is what exec(2) returns. */
319 if (attr.va_size == 0) {
324 /* Can we access it? */
325 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
330 * XXX: This should use vn_open() so that it is properly authorized,
331 * and to reduce code redundancy all over the place here.
332 * XXX: Not really, it duplicates far more of exec_check_permissions()
336 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
340 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
344 /* Pull in executable header into exec_map */
345 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
346 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
350 /* Is it a Linux binary ? */
351 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
357 * While we are here, we should REALLY do some more checks
360 /* Set file/virtual offset based on a.out variant. */
361 switch ((int)(a_out->a_magic & 0xffff)) {
362 case 0413: /* ZMAGIC */
365 case 0314: /* QMAGIC */
373 bss_size = round_page(a_out->a_bss);
375 /* Check various fields in header for validity/bounds. */
376 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
381 /* text + data can't exceed file size */
382 if (a_out->a_data + a_out->a_text > attr.va_size) {
388 * text/data/bss must not exceed limits
389 * XXX - this is not complete. it should check current usage PLUS
390 * the resources needed by this library.
392 PROC_LOCK(td->td_proc);
393 if (a_out->a_text > maxtsiz ||
394 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
395 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
397 PROC_UNLOCK(td->td_proc);
401 PROC_UNLOCK(td->td_proc);
404 * Prevent more writers.
405 * XXX: Note that if any of the VM operations fail below we don't
411 * Lock no longer needed
417 * Check if file_offset page aligned. Currently we cannot handle
418 * misalinged file offsets, and so we read in the entire image
421 if (file_offset & PAGE_MASK) {
423 printf("uselib: Non page aligned binary %lu\n", file_offset);
425 /* Map text+data read/write/execute */
427 /* a_entry is the load address and is page aligned */
428 vmaddr = trunc_page(a_out->a_entry);
430 /* get anon user mapping, read+write+execute */
431 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
432 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
433 VM_PROT_ALL, VM_PROT_ALL, 0);
437 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
438 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
439 td->td_ucred, NOCRED, &aresid, td);
448 printf("uselib: Page aligned binary %lu\n", file_offset);
451 * for QMAGIC, a_entry is 20 bytes beyond the load address
452 * to skip the executable header
454 vmaddr = trunc_page(a_out->a_entry);
457 * Map it all into the process's space as a single
458 * copy-on-write "data" segment.
460 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
461 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
462 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
467 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
468 ((long *)vmaddr)[1]);
471 /* Calculate BSS start address */
472 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
475 /* allocate some 'anon' space */
476 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
477 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
484 /* Unlock vnode if needed */
488 /* Release the temporary mapping. */
490 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
495 #endif /* __i386__ */
498 linux_select(struct thread *td, struct linux_select_args *args)
501 struct timeval tv0, tv1, utv, *tvp;
506 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
507 (void *)args->readfds, (void *)args->writefds,
508 (void *)args->exceptfds, (void *)args->timeout);
512 * Store current time for computation of the amount of
516 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
518 utv.tv_sec = ltv.tv_sec;
519 utv.tv_usec = ltv.tv_usec;
522 printf(LMSG("incoming timeout (%jd/%ld)"),
523 (intmax_t)utv.tv_sec, utv.tv_usec);
526 if (itimerfix(&utv)) {
528 * The timeval was invalid. Convert it to something
529 * valid that will act as it does under Linux.
531 utv.tv_sec += utv.tv_usec / 1000000;
532 utv.tv_usec %= 1000000;
533 if (utv.tv_usec < 0) {
535 utv.tv_usec += 1000000;
545 error = kern_select(td, args->nfds, args->readfds, args->writefds,
546 args->exceptfds, tvp, LINUX_NFDBITS);
550 printf(LMSG("real select returns %d"), error);
556 if (td->td_retval[0]) {
558 * Compute how much time was left of the timeout,
559 * by subtracting the current time and the time
560 * before we started the call, and subtracting
561 * that result from the user-supplied value.
564 timevalsub(&tv1, &tv0);
565 timevalsub(&utv, &tv1);
572 printf(LMSG("outgoing timeout (%jd/%ld)"),
573 (intmax_t)utv.tv_sec, utv.tv_usec);
575 ltv.tv_sec = utv.tv_sec;
576 ltv.tv_usec = utv.tv_usec;
577 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
584 printf(LMSG("select_out -> %d"), error);
590 linux_mremap(struct thread *td, struct linux_mremap_args *args)
598 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
599 (void *)(uintptr_t)args->addr,
600 (unsigned long)args->old_len,
601 (unsigned long)args->new_len,
602 (unsigned long)args->flags);
605 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
606 td->td_retval[0] = 0;
611 * Check for the page alignment.
612 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
614 if (args->addr & PAGE_MASK) {
615 td->td_retval[0] = 0;
619 args->new_len = round_page(args->new_len);
620 args->old_len = round_page(args->old_len);
622 if (args->new_len > args->old_len) {
623 td->td_retval[0] = 0;
627 if (args->new_len < args->old_len) {
628 addr = args->addr + args->new_len;
629 len = args->old_len - args->new_len;
630 error = kern_munmap(td, addr, len);
633 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
637 #define LINUX_MS_ASYNC 0x0001
638 #define LINUX_MS_INVALIDATE 0x0002
639 #define LINUX_MS_SYNC 0x0004
642 linux_msync(struct thread *td, struct linux_msync_args *args)
645 return (kern_msync(td, args->addr, args->len,
646 args->fl & ~LINUX_MS_SYNC));
650 linux_time(struct thread *td, struct linux_time_args *args)
658 printf(ARGS(time, "*"));
663 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
665 td->td_retval[0] = tm;
669 struct l_times_argv {
672 l_clock_t tms_cutime;
673 l_clock_t tms_cstime;
678 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
679 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
680 * auxiliary vector entry.
684 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
685 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
687 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
688 CONVNTCK(r) : CONVOTCK(r))
691 linux_times(struct thread *td, struct linux_times_args *args)
693 struct timeval tv, utime, stime, cutime, cstime;
694 struct l_times_argv tms;
700 printf(ARGS(times, "*"));
703 if (args->buf != NULL) {
707 calcru(p, &utime, &stime);
709 calccru(p, &cutime, &cstime);
712 tms.tms_utime = CONVTCK(utime);
713 tms.tms_stime = CONVTCK(stime);
715 tms.tms_cutime = CONVTCK(cutime);
716 tms.tms_cstime = CONVTCK(cstime);
718 if ((error = copyout(&tms, args->buf, sizeof(tms))))
723 td->td_retval[0] = (int)CONVTCK(tv);
728 linux_newuname(struct thread *td, struct linux_newuname_args *args)
730 struct l_new_utsname utsname;
731 char osname[LINUX_MAX_UTSNAME];
732 char osrelease[LINUX_MAX_UTSNAME];
736 if (ldebug(newuname))
737 printf(ARGS(newuname, "*"));
740 linux_get_osname(td, osname);
741 linux_get_osrelease(td, osrelease);
743 bzero(&utsname, sizeof(utsname));
744 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
745 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
746 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
747 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
748 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
749 for (p = utsname.version; *p != '\0'; ++p)
754 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
756 return (copyout(&utsname, args->buf, sizeof(utsname)));
765 linux_utime(struct thread *td, struct linux_utime_args *args)
767 struct timeval tv[2], *tvp;
768 struct l_utimbuf lut;
772 LCONVPATHEXIST(td, args->fname, &fname);
776 printf(ARGS(utime, "%s, *"), fname);
780 if ((error = copyin(args->times, &lut, sizeof lut))) {
784 tv[0].tv_sec = lut.l_actime;
786 tv[1].tv_sec = lut.l_modtime;
792 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
799 linux_utimes(struct thread *td, struct linux_utimes_args *args)
802 struct timeval tv[2], *tvp = NULL;
806 LCONVPATHEXIST(td, args->fname, &fname);
810 printf(ARGS(utimes, "%s, *"), fname);
813 if (args->tptr != NULL) {
814 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
818 tv[0].tv_sec = ltv[0].tv_sec;
819 tv[0].tv_usec = ltv[0].tv_usec;
820 tv[1].tv_sec = ltv[1].tv_sec;
821 tv[1].tv_usec = ltv[1].tv_usec;
825 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
832 linux_utimensat_nsec_valid(l_long nsec)
835 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
837 if (nsec >= 0 && nsec <= 999999999)
843 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
845 struct l_timespec l_times[2];
846 struct timespec times[2], *timesp = NULL;
848 int error, dfd, flags = 0;
850 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
853 if (ldebug(utimensat))
854 printf(ARGS(utimensat, "%d, *"), dfd);
857 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
860 if (args->times != NULL) {
861 error = copyin(args->times, l_times, sizeof(l_times));
865 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
866 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
869 times[0].tv_sec = l_times[0].tv_sec;
870 switch (l_times[0].tv_nsec)
872 case LINUX_UTIME_OMIT:
873 times[0].tv_nsec = UTIME_OMIT;
875 case LINUX_UTIME_NOW:
876 times[0].tv_nsec = UTIME_NOW;
879 times[0].tv_nsec = l_times[0].tv_nsec;
882 times[1].tv_sec = l_times[1].tv_sec;
883 switch (l_times[1].tv_nsec)
885 case LINUX_UTIME_OMIT:
886 times[1].tv_nsec = UTIME_OMIT;
888 case LINUX_UTIME_NOW:
889 times[1].tv_nsec = UTIME_NOW;
892 times[1].tv_nsec = l_times[1].tv_nsec;
897 /* This breaks POSIX, but is what the Linux kernel does
898 * _on purpose_ (documented in the man page for utimensat(2)),
899 * so we must follow that behaviour. */
900 if (times[0].tv_nsec == UTIME_OMIT &&
901 times[1].tv_nsec == UTIME_OMIT)
905 if (args->pathname != NULL)
906 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
907 else if (args->flags != 0)
910 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
911 flags |= AT_SYMLINK_NOFOLLOW;
914 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
916 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
917 UIO_SYSSPACE, flags);
925 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
928 struct timeval tv[2], *tvp = NULL;
932 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
933 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
936 if (ldebug(futimesat))
937 printf(ARGS(futimesat, "%s, *"), fname);
940 if (args->utimes != NULL) {
941 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
945 tv[0].tv_sec = ltv[0].tv_sec;
946 tv[0].tv_usec = ltv[0].tv_usec;
947 tv[1].tv_sec = ltv[1].tv_sec;
948 tv[1].tv_usec = ltv[1].tv_usec;
952 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
958 linux_common_wait(struct thread *td, int pid, int *status,
959 int options, struct rusage *ru)
963 error = kern_wait(td, pid, &tmpstat, options, ru);
969 if (WIFSIGNALED(tmpstat))
970 tmpstat = (tmpstat & 0xffffff80) |
971 bsd_to_linux_signal(WTERMSIG(tmpstat));
972 else if (WIFSTOPPED(tmpstat))
973 tmpstat = (tmpstat & 0xffff00ff) |
974 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
975 else if (WIFCONTINUED(tmpstat))
977 error = copyout(&tmpstat, status, sizeof(int));
983 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
985 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
987 struct linux_wait4_args wait4_args;
991 printf(ARGS(waitpid, "%d, %p, %d"),
992 args->pid, (void *)args->status, args->options);
995 wait4_args.pid = args->pid;
996 wait4_args.status = args->status;
997 wait4_args.options = args->options;
998 wait4_args.rusage = NULL;
1000 return (linux_wait4(td, &wait4_args));
1002 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1005 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1008 struct rusage ru, *rup;
1012 printf(ARGS(wait4, "%d, %p, %d, %p"),
1013 args->pid, (void *)args->status, args->options,
1014 (void *)args->rusage);
1016 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1017 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1021 linux_to_bsd_waitopts(args->options, &options);
1023 if (args->rusage != NULL)
1027 error = linux_common_wait(td, args->pid, args->status, options, rup);
1030 if (args->rusage != NULL)
1031 error = linux_copyout_rusage(&ru, args->rusage);
1036 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1038 int status, options, sig;
1039 struct __wrusage wru;
1047 linux_to_bsd_waitopts(args->options, &options);
1049 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1051 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1054 switch (args->idtype) {
1072 error = kern_wait6(td, idtype, args->id, &status, options,
1076 if (args->rusage != NULL) {
1077 error = linux_copyout_rusage(&wru.wru_children,
1082 if (args->info != NULL) {
1084 if (td->td_retval[0] == 0)
1085 bzero(&lsi, sizeof(lsi));
1087 sig = bsd_to_linux_signal(siginfo.si_signo);
1088 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1090 error = copyout(&lsi, args->info, sizeof(lsi));
1092 td->td_retval[0] = 0;
1098 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1103 LCONVPATHCREAT(td, args->path, &path);
1107 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1108 (uintmax_t)args->dev);
1111 switch (args->mode & S_IFMT) {
1114 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1120 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1121 args->mode, args->dev);
1129 args->mode |= S_IFREG;
1132 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1133 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1135 kern_close(td, td->td_retval[0]);
1147 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1152 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1153 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1156 if (ldebug(mknodat))
1157 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1160 switch (args->mode & S_IFMT) {
1163 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1168 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1177 args->mode |= S_IFREG;
1180 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1181 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1183 kern_close(td, td->td_retval[0]);
1195 * UGH! This is just about the dumbest idea I've ever heard!!
1198 linux_personality(struct thread *td, struct linux_personality_args *args)
1200 struct linux_pemuldata *pem;
1201 struct proc *p = td->td_proc;
1205 if (ldebug(personality))
1206 printf(ARGS(personality, "%u"), args->per);
1212 if (args->per != 0xffffffff)
1213 pem->persona = args->per;
1216 td->td_retval[0] = old;
1220 struct l_itimerval {
1221 l_timeval it_interval;
1225 #define B2L_ITIMERVAL(bip, lip) \
1226 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1227 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1228 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1229 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1232 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1235 struct l_itimerval ls;
1236 struct itimerval aitv, oitv;
1239 if (ldebug(setitimer))
1240 printf(ARGS(setitimer, "%p, %p"),
1241 (void *)uap->itv, (void *)uap->oitv);
1244 if (uap->itv == NULL) {
1245 uap->itv = uap->oitv;
1246 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1249 error = copyin(uap->itv, &ls, sizeof(ls));
1252 B2L_ITIMERVAL(&aitv, &ls);
1254 if (ldebug(setitimer)) {
1255 printf("setitimer: value: sec: %jd, usec: %ld\n",
1256 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1257 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1258 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1261 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1262 if (error != 0 || uap->oitv == NULL)
1264 B2L_ITIMERVAL(&ls, &oitv);
1266 return (copyout(&ls, uap->oitv, sizeof(ls)));
1270 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1273 struct l_itimerval ls;
1274 struct itimerval aitv;
1277 if (ldebug(getitimer))
1278 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1280 error = kern_getitimer(td, uap->which, &aitv);
1283 B2L_ITIMERVAL(&ls, &aitv);
1284 return (copyout(&ls, uap->itv, sizeof(ls)));
1287 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1289 linux_nice(struct thread *td, struct linux_nice_args *args)
1291 struct setpriority_args bsd_args;
1293 bsd_args.which = PRIO_PROCESS;
1294 bsd_args.who = 0; /* current process */
1295 bsd_args.prio = args->inc;
1296 return (sys_setpriority(td, &bsd_args));
1298 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1301 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1303 struct ucred *newcred, *oldcred;
1304 l_gid_t *linux_gidset;
1309 ngrp = args->gidsetsize;
1310 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1312 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1313 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1317 crextend(newcred, ngrp + 1);
1320 oldcred = p->p_ucred;
1321 crcopy(newcred, oldcred);
1324 * cr_groups[0] holds egid. Setting the whole set from
1325 * the supplied set will cause egid to be changed too.
1326 * Keep cr_groups[0] unchanged to prevent that.
1329 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1336 newcred->cr_ngroups = ngrp + 1;
1338 bsd_gidset = newcred->cr_groups;
1341 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1345 newcred->cr_ngroups = 1;
1348 proc_set_cred(p, newcred);
1353 free(linux_gidset, M_LINUX);
1358 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1361 l_gid_t *linux_gidset;
1363 int bsd_gidsetsz, ngrp, error;
1365 cred = td->td_ucred;
1366 bsd_gidset = cred->cr_groups;
1367 bsd_gidsetsz = cred->cr_ngroups - 1;
1370 * cr_groups[0] holds egid. Returning the whole set
1371 * here will cause a duplicate. Exclude cr_groups[0]
1375 if ((ngrp = args->gidsetsize) == 0) {
1376 td->td_retval[0] = bsd_gidsetsz;
1380 if (ngrp < bsd_gidsetsz)
1384 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1386 while (ngrp < bsd_gidsetsz) {
1387 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1391 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1392 free(linux_gidset, M_LINUX);
1396 td->td_retval[0] = ngrp;
1401 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1403 struct rlimit bsd_rlim;
1404 struct l_rlimit rlim;
1409 if (ldebug(setrlimit))
1410 printf(ARGS(setrlimit, "%d, %p"),
1411 args->resource, (void *)args->rlim);
1414 if (args->resource >= LINUX_RLIM_NLIMITS)
1417 which = linux_to_bsd_resource[args->resource];
1421 error = copyin(args->rlim, &rlim, sizeof(rlim));
1425 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1426 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1427 return (kern_setrlimit(td, which, &bsd_rlim));
1430 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1432 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1434 struct l_rlimit rlim;
1435 struct rlimit bsd_rlim;
1439 if (ldebug(old_getrlimit))
1440 printf(ARGS(old_getrlimit, "%d, %p"),
1441 args->resource, (void *)args->rlim);
1444 if (args->resource >= LINUX_RLIM_NLIMITS)
1447 which = linux_to_bsd_resource[args->resource];
1451 lim_rlimit(td, which, &bsd_rlim);
1453 #ifdef COMPAT_LINUX32
1454 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1455 if (rlim.rlim_cur == UINT_MAX)
1456 rlim.rlim_cur = INT_MAX;
1457 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1458 if (rlim.rlim_max == UINT_MAX)
1459 rlim.rlim_max = INT_MAX;
1461 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1462 if (rlim.rlim_cur == ULONG_MAX)
1463 rlim.rlim_cur = LONG_MAX;
1464 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1465 if (rlim.rlim_max == ULONG_MAX)
1466 rlim.rlim_max = LONG_MAX;
1468 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1470 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1473 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1475 struct l_rlimit rlim;
1476 struct rlimit bsd_rlim;
1480 if (ldebug(getrlimit))
1481 printf(ARGS(getrlimit, "%d, %p"),
1482 args->resource, (void *)args->rlim);
1485 if (args->resource >= LINUX_RLIM_NLIMITS)
1488 which = linux_to_bsd_resource[args->resource];
1492 lim_rlimit(td, which, &bsd_rlim);
1494 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1495 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1496 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1500 linux_sched_setscheduler(struct thread *td,
1501 struct linux_sched_setscheduler_args *args)
1503 struct sched_param sched_param;
1508 if (ldebug(sched_setscheduler))
1509 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1510 args->pid, args->policy, (const void *)args->param);
1513 switch (args->policy) {
1514 case LINUX_SCHED_OTHER:
1515 policy = SCHED_OTHER;
1517 case LINUX_SCHED_FIFO:
1518 policy = SCHED_FIFO;
1520 case LINUX_SCHED_RR:
1527 error = copyin(args->param, &sched_param, sizeof(sched_param));
1531 tdt = linux_tdfind(td, args->pid, -1);
1535 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1536 PROC_UNLOCK(tdt->td_proc);
1541 linux_sched_getscheduler(struct thread *td,
1542 struct linux_sched_getscheduler_args *args)
1548 if (ldebug(sched_getscheduler))
1549 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1552 tdt = linux_tdfind(td, args->pid, -1);
1556 error = kern_sched_getscheduler(td, tdt, &policy);
1557 PROC_UNLOCK(tdt->td_proc);
1561 td->td_retval[0] = LINUX_SCHED_OTHER;
1564 td->td_retval[0] = LINUX_SCHED_FIFO;
1567 td->td_retval[0] = LINUX_SCHED_RR;
1574 linux_sched_get_priority_max(struct thread *td,
1575 struct linux_sched_get_priority_max_args *args)
1577 struct sched_get_priority_max_args bsd;
1580 if (ldebug(sched_get_priority_max))
1581 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1584 switch (args->policy) {
1585 case LINUX_SCHED_OTHER:
1586 bsd.policy = SCHED_OTHER;
1588 case LINUX_SCHED_FIFO:
1589 bsd.policy = SCHED_FIFO;
1591 case LINUX_SCHED_RR:
1592 bsd.policy = SCHED_RR;
1597 return (sys_sched_get_priority_max(td, &bsd));
1601 linux_sched_get_priority_min(struct thread *td,
1602 struct linux_sched_get_priority_min_args *args)
1604 struct sched_get_priority_min_args bsd;
1607 if (ldebug(sched_get_priority_min))
1608 printf(ARGS(sched_get_priority_min, "%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_min(td, &bsd));
1627 #define REBOOT_CAD_ON 0x89abcdef
1628 #define REBOOT_CAD_OFF 0
1629 #define REBOOT_HALT 0xcdef0123
1630 #define REBOOT_RESTART 0x01234567
1631 #define REBOOT_RESTART2 0xA1B2C3D4
1632 #define REBOOT_POWEROFF 0x4321FEDC
1633 #define REBOOT_MAGIC1 0xfee1dead
1634 #define REBOOT_MAGIC2 0x28121969
1635 #define REBOOT_MAGIC2A 0x05121996
1636 #define REBOOT_MAGIC2B 0x16041998
1639 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1641 struct reboot_args bsd_args;
1645 printf(ARGS(reboot, "0x%x"), args->cmd);
1648 if (args->magic1 != REBOOT_MAGIC1)
1651 switch (args->magic2) {
1653 case REBOOT_MAGIC2A:
1654 case REBOOT_MAGIC2B:
1660 switch (args->cmd) {
1662 case REBOOT_CAD_OFF:
1663 return (priv_check(td, PRIV_REBOOT));
1665 bsd_args.opt = RB_HALT;
1667 case REBOOT_RESTART:
1668 case REBOOT_RESTART2:
1671 case REBOOT_POWEROFF:
1672 bsd_args.opt = RB_POWEROFF;
1677 return (sys_reboot(td, &bsd_args));
1682 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1683 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1684 * are assumed to be preserved. The following lightweight syscalls fixes
1685 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1687 * linux_getpid() - MP SAFE
1688 * linux_getgid() - MP SAFE
1689 * linux_getuid() - MP SAFE
1693 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1698 printf(ARGS(getpid, ""));
1700 td->td_retval[0] = td->td_proc->p_pid;
1706 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1708 struct linux_emuldata *em;
1712 printf(ARGS(gettid, ""));
1716 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1718 td->td_retval[0] = em->em_tid;
1725 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1729 if (ldebug(getppid))
1730 printf(ARGS(getppid, ""));
1733 td->td_retval[0] = kern_getppid(td);
1738 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1743 printf(ARGS(getgid, ""));
1746 td->td_retval[0] = td->td_ucred->cr_rgid;
1751 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1756 printf(ARGS(getuid, ""));
1759 td->td_retval[0] = td->td_ucred->cr_ruid;
1765 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1767 struct getsid_args bsd;
1771 printf(ARGS(getsid, "%i"), args->pid);
1774 bsd.pid = args->pid;
1775 return (sys_getsid(td, &bsd));
1779 linux_nosys(struct thread *td, struct nosys_args *ignore)
1786 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1788 struct getpriority_args bsd_args;
1792 if (ldebug(getpriority))
1793 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1796 bsd_args.which = args->which;
1797 bsd_args.who = args->who;
1798 error = sys_getpriority(td, &bsd_args);
1799 td->td_retval[0] = 20 - td->td_retval[0];
1804 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1809 if (ldebug(sethostname))
1810 printf(ARGS(sethostname, "*, %i"), args->len);
1814 name[1] = KERN_HOSTNAME;
1815 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1820 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1825 if (ldebug(setdomainname))
1826 printf(ARGS(setdomainname, "*, %i"), args->len);
1830 name[1] = KERN_NISDOMAINNAME;
1831 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1836 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1840 if (ldebug(exit_group))
1841 printf(ARGS(exit_group, "%i"), args->error_code);
1844 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1848 * XXX: we should send a signal to the parent if
1849 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1850 * as it doesnt occur often.
1852 exit1(td, args->error_code, 0);
1856 #define _LINUX_CAPABILITY_VERSION 0x19980330
1858 struct l_user_cap_header {
1863 struct l_user_cap_data {
1870 linux_capget(struct thread *td, struct linux_capget_args *args)
1872 struct l_user_cap_header luch;
1873 struct l_user_cap_data lucd;
1876 if (args->hdrp == NULL)
1879 error = copyin(args->hdrp, &luch, sizeof(luch));
1883 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1884 luch.version = _LINUX_CAPABILITY_VERSION;
1885 error = copyout(&luch, args->hdrp, sizeof(luch));
1896 * The current implementation doesn't support setting
1897 * a capability (it's essentially a stub) so indicate
1898 * that no capabilities are currently set or available
1901 bzero (&lucd, sizeof(lucd));
1902 error = copyout(&lucd, args->datap, sizeof(lucd));
1909 linux_capset(struct thread *td, struct linux_capset_args *args)
1911 struct l_user_cap_header luch;
1912 struct l_user_cap_data lucd;
1915 if (args->hdrp == NULL || args->datap == NULL)
1918 error = copyin(args->hdrp, &luch, sizeof(luch));
1922 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1923 luch.version = _LINUX_CAPABILITY_VERSION;
1924 error = copyout(&luch, args->hdrp, sizeof(luch));
1933 error = copyin(args->datap, &lucd, sizeof(lucd));
1937 /* We currently don't support setting any capabilities. */
1938 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1940 "capset effective=0x%x, permitted=0x%x, "
1941 "inheritable=0x%x is not implemented",
1942 (int)lucd.effective, (int)lucd.permitted,
1943 (int)lucd.inheritable);
1951 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1953 int error = 0, max_size;
1954 struct proc *p = td->td_proc;
1955 char comm[LINUX_MAX_COMM_LEN];
1956 struct linux_emuldata *em;
1961 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1962 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1963 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1966 switch (args->option) {
1967 case LINUX_PR_SET_PDEATHSIG:
1968 if (!LINUX_SIG_VALID(args->arg2))
1971 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1972 em->pdeath_signal = args->arg2;
1974 case LINUX_PR_GET_PDEATHSIG:
1976 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1977 pdeath_signal = em->pdeath_signal;
1978 error = copyout(&pdeath_signal,
1979 (void *)(register_t)args->arg2,
1980 sizeof(pdeath_signal));
1982 case LINUX_PR_GET_KEEPCAPS:
1984 * Indicate that we always clear the effective and
1985 * permitted capability sets when the user id becomes
1986 * non-zero (actually the capability sets are simply
1987 * always zero in the current implementation).
1989 td->td_retval[0] = 0;
1991 case LINUX_PR_SET_KEEPCAPS:
1993 * Ignore requests to keep the effective and permitted
1994 * capability sets when the user id becomes non-zero.
1997 case LINUX_PR_SET_NAME:
1999 * To be on the safe side we need to make sure to not
2000 * overflow the size a Linux program expects. We already
2001 * do this here in the copyin, so that we don't need to
2004 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2005 error = copyinstr((void *)(register_t)args->arg2, comm,
2008 /* Linux silently truncates the name if it is too long. */
2009 if (error == ENAMETOOLONG) {
2011 * XXX: copyinstr() isn't documented to populate the
2012 * array completely, so do a copyin() to be on the
2013 * safe side. This should be changed in case
2014 * copyinstr() is changed to guarantee this.
2016 error = copyin((void *)(register_t)args->arg2, comm,
2018 comm[max_size - 1] = '\0';
2024 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2027 case LINUX_PR_GET_NAME:
2029 strlcpy(comm, p->p_comm, sizeof(comm));
2031 error = copyout(comm, (void *)(register_t)args->arg2,
2043 linux_sched_setparam(struct thread *td,
2044 struct linux_sched_setparam_args *uap)
2046 struct sched_param sched_param;
2051 if (ldebug(sched_setparam))
2052 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2055 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2059 tdt = linux_tdfind(td, uap->pid, -1);
2063 error = kern_sched_setparam(td, tdt, &sched_param);
2064 PROC_UNLOCK(tdt->td_proc);
2069 linux_sched_getparam(struct thread *td,
2070 struct linux_sched_getparam_args *uap)
2072 struct sched_param sched_param;
2077 if (ldebug(sched_getparam))
2078 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2081 tdt = linux_tdfind(td, uap->pid, -1);
2085 error = kern_sched_getparam(td, tdt, &sched_param);
2086 PROC_UNLOCK(tdt->td_proc);
2088 error = copyout(&sched_param, uap->param,
2089 sizeof(sched_param));
2094 * Get affinity of a process.
2097 linux_sched_getaffinity(struct thread *td,
2098 struct linux_sched_getaffinity_args *args)
2104 if (ldebug(sched_getaffinity))
2105 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2108 if (args->len < sizeof(cpuset_t))
2111 tdt = linux_tdfind(td, args->pid, -1);
2115 PROC_UNLOCK(tdt->td_proc);
2117 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2118 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2120 td->td_retval[0] = sizeof(cpuset_t);
2126 * Set affinity of a process.
2129 linux_sched_setaffinity(struct thread *td,
2130 struct linux_sched_setaffinity_args *args)
2135 if (ldebug(sched_setaffinity))
2136 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2139 if (args->len < sizeof(cpuset_t))
2142 tdt = linux_tdfind(td, args->pid, -1);
2146 PROC_UNLOCK(tdt->td_proc);
2148 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2149 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2152 struct linux_rlimit64 {
2158 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2160 struct rlimit rlim, nrlim;
2161 struct linux_rlimit64 lrlim;
2168 if (ldebug(prlimit64))
2169 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2170 args->resource, (void *)args->new, (void *)args->old);
2173 if (args->resource >= LINUX_RLIM_NLIMITS)
2176 which = linux_to_bsd_resource[args->resource];
2180 if (args->new != NULL) {
2182 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2183 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2184 * as INFINITY so we do not need a conversion even.
2186 error = copyin(args->new, &nrlim, sizeof(nrlim));
2191 flags = PGET_HOLD | PGET_NOTWEXIT;
2192 if (args->new != NULL)
2193 flags |= PGET_CANDEBUG;
2195 flags |= PGET_CANSEE;
2196 error = pget(args->pid, flags, &p);
2200 if (args->old != NULL) {
2202 lim_rlimit_proc(p, which, &rlim);
2204 if (rlim.rlim_cur == RLIM_INFINITY)
2205 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2207 lrlim.rlim_cur = rlim.rlim_cur;
2208 if (rlim.rlim_max == RLIM_INFINITY)
2209 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2211 lrlim.rlim_max = rlim.rlim_max;
2212 error = copyout(&lrlim, args->old, sizeof(lrlim));
2217 if (args->new != NULL)
2218 error = kern_proc_setrlimit(td, p, which, &nrlim);
2226 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2228 struct timeval utv, tv0, tv1, *tvp;
2229 struct l_pselect6arg lpse6;
2230 struct l_timespec lts;
2231 struct timespec uts;
2238 if (args->sig != NULL) {
2239 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2242 if (lpse6.ss_len != sizeof(l_ss))
2244 if (lpse6.ss != 0) {
2245 error = copyin(PTRIN(lpse6.ss), &l_ss,
2249 linux_to_bsd_sigset(&l_ss, &ss);
2255 * Currently glibc changes nanosecond number to microsecond.
2256 * This mean losing precision but for now it is hardly seen.
2258 if (args->tsp != NULL) {
2259 error = copyin(args->tsp, <s, sizeof(lts));
2262 error = linux_to_native_timespec(&uts, <s);
2266 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2267 if (itimerfix(&utv))
2275 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2276 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2278 if (error == 0 && args->tsp != NULL) {
2279 if (td->td_retval[0] != 0) {
2281 * Compute how much time was left of the timeout,
2282 * by subtracting the current time and the time
2283 * before we started the call, and subtracting
2284 * that result from the user-supplied value.
2288 timevalsub(&tv1, &tv0);
2289 timevalsub(&utv, &tv1);
2295 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2297 error = native_to_linux_timespec(<s, &uts);
2299 error = copyout(<s, args->tsp, sizeof(lts));
2306 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2308 struct timespec ts0, ts1;
2309 struct l_timespec lts;
2310 struct timespec uts, *tsp;
2316 if (args->sset != NULL) {
2317 if (args->ssize != sizeof(l_ss))
2319 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2322 linux_to_bsd_sigset(&l_ss, &ss);
2326 if (args->tsp != NULL) {
2327 error = copyin(args->tsp, <s, sizeof(lts));
2330 error = linux_to_native_timespec(&uts, <s);
2339 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2341 if (error == 0 && args->tsp != NULL) {
2342 if (td->td_retval[0]) {
2344 timespecsub(&ts1, &ts0);
2345 timespecsub(&uts, &ts1);
2347 timespecclear(&uts);
2349 timespecclear(&uts);
2351 error = native_to_linux_timespec(<s, &uts);
2353 error = copyout(<s, args->tsp, sizeof(lts));
2359 #if defined(DEBUG) || defined(KTR)
2360 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2362 #ifdef COMPAT_LINUX32
2363 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL
2365 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL
2368 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2371 linux_debug(int syscall, int toggle, int global)
2375 char c = toggle ? 0 : 0xff;
2377 memset(linux_debug_map, c, sizeof(linux_debug_map));
2380 if (syscall < 0 || syscall >= L_MAXSYSCALL)
2383 clrbit(linux_debug_map, syscall);
2385 setbit(linux_debug_map, syscall);
2391 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2393 * E.g.: sysctl linux.debug=21.0
2395 * As a special case, syscall "all" will apply to all syscalls globally.
2397 #define LINUX_MAX_DEBUGSTR 16
2399 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2401 char value[LINUX_MAX_DEBUGSTR], *p;
2402 int error, sysc, toggle;
2406 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2407 if (error || req->newptr == NULL)
2409 for (p = value; *p != '\0' && *p != '.'; p++);
2413 sysc = strtol(value, NULL, 0);
2414 toggle = strtol(p, NULL, 0);
2415 if (strcmp(value, "all") == 0)
2417 error = linux_debug(sysc, toggle, global);
2421 #endif /* DEBUG || KTR */
2424 linux_sched_rr_get_interval(struct thread *td,
2425 struct linux_sched_rr_get_interval_args *uap)
2428 struct l_timespec lts;
2433 * According to man in case the invalid pid specified
2434 * EINVAL should be returned.
2439 tdt = linux_tdfind(td, uap->pid, -1);
2443 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2444 PROC_UNLOCK(tdt->td_proc);
2447 error = native_to_linux_timespec(<s, &ts);
2450 return (copyout(<s, uap->interval, sizeof(lts)));
2454 * In case when the Linux thread is the initial thread in
2455 * the thread group thread id is equal to the process id.
2456 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2459 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2461 struct linux_emuldata *em;
2466 if (tid == 0 || tid == td->td_tid) {
2468 PROC_LOCK(tdt->td_proc);
2469 } else if (tid > PID_MAX)
2470 tdt = tdfind(tid, pid);
2473 * Initial thread where the tid equal to the pid.
2477 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2479 * p is not a Linuxulator process.
2484 FOREACH_THREAD_IN_PROC(p, tdt) {
2486 if (tid == em->em_tid)
2498 linux_to_bsd_waitopts(int options, int *bsdopts)
2501 if (options & LINUX_WNOHANG)
2502 *bsdopts |= WNOHANG;
2503 if (options & LINUX_WUNTRACED)
2504 *bsdopts |= WUNTRACED;
2505 if (options & LINUX_WEXITED)
2506 *bsdopts |= WEXITED;
2507 if (options & LINUX_WCONTINUED)
2508 *bsdopts |= WCONTINUED;
2509 if (options & LINUX_WNOWAIT)
2510 *bsdopts |= WNOWAIT;
2512 if (options & __WCLONE)
2513 *bsdopts |= WLINUXCLONE;
2517 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2523 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2525 if (args->count > INT_MAX)
2526 args->count = INT_MAX;
2528 iov.iov_base = args->buf;
2529 iov.iov_len = args->count;
2533 uio.uio_resid = iov.iov_len;
2534 uio.uio_segflg = UIO_USERSPACE;
2535 uio.uio_rw = UIO_READ;
2538 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2540 td->td_retval[0] = args->count - uio.uio_resid;
2545 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2548 /* Needs to be page-aligned */
2549 if (args->start & PAGE_MASK)
2551 return (kern_mincore(td, args->start, args->len, args->vec));