2 * Copyright (c) 2002 Doug Rabson
3 * Copyright (c) 1994-1995 Søren Schmidt
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer
11 * in this position and unchanged.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote products
16 * derived from this software without specific prior written permission
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
33 #include "opt_compat.h"
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
39 #include <sys/imgact_aout.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
45 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
52 #include <sys/reboot.h>
53 #include <sys/racct.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
57 #include <sys/signalvar.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysproto.h>
62 #include <sys/systm.h>
64 #include <sys/vmmeter.h>
65 #include <sys/vnode.h>
67 #include <sys/cpuset.h>
69 #include <security/mac/mac_framework.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
87 #include <compat/linux/linux_dtrace.h>
88 #include <compat/linux/linux_file.h>
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_util.h>
92 #include <compat/linux/linux_sysproto.h>
93 #include <compat/linux/linux_emul.h>
94 #include <compat/linux/linux_misc.h>
97 * Special DTrace provider for the linuxulator.
99 * In this file we define the provider for the entire linuxulator. All
100 * modules (= files of the linuxulator) use it.
102 * We define a different name depending on the emulated bitsize, see
103 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
104 * native bitsize = linuxulator
105 * amd64, 32bit emulation = linuxulator32
107 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
109 int stclohz; /* Statistics clock frequency */
111 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
112 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
113 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
114 RLIMIT_MEMLOCK, RLIMIT_AS
118 l_long uptime; /* Seconds since boot */
119 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
120 #define LINUX_SYSINFO_LOADS_SCALE 65536
121 l_ulong totalram; /* Total usable main memory size */
122 l_ulong freeram; /* Available memory size */
123 l_ulong sharedram; /* Amount of shared memory */
124 l_ulong bufferram; /* Memory used by buffers */
125 l_ulong totalswap; /* Total swap space size */
126 l_ulong freeswap; /* swap space still available */
127 l_ushort procs; /* Number of current processes */
132 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
135 struct l_pselect6arg {
141 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
143 struct l_sysinfo sysinfo;
151 sysinfo.uptime = ts.tv_sec;
153 /* Use the information from the mib to get our load averages */
154 for (i = 0; i < 3; i++)
155 sysinfo.loads[i] = averunnable.ldavg[i] *
156 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
158 sysinfo.totalram = physmem * PAGE_SIZE;
159 sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
161 sysinfo.sharedram = 0;
162 mtx_lock(&vm_object_list_mtx);
163 TAILQ_FOREACH(object, &vm_object_list, object_list)
164 if (object->shadow_count > 1)
165 sysinfo.sharedram += object->resident_page_count;
166 mtx_unlock(&vm_object_list_mtx);
168 sysinfo.sharedram *= PAGE_SIZE;
169 sysinfo.bufferram = 0;
171 swap_pager_status(&i, &j);
172 sysinfo.totalswap = i * PAGE_SIZE;
173 sysinfo.freeswap = (i - j) * PAGE_SIZE;
175 sysinfo.procs = nprocs;
177 /* The following are only present in newer Linux kernels. */
178 sysinfo.totalbig = 0;
180 sysinfo.mem_unit = 1;
182 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
186 linux_alarm(struct thread *td, struct linux_alarm_args *args)
188 struct itimerval it, old_it;
194 printf(ARGS(alarm, "%u"), args->secs);
202 it.it_value.tv_sec = (long) secs;
203 it.it_value.tv_usec = 0;
204 it.it_interval.tv_sec = 0;
205 it.it_interval.tv_usec = 0;
206 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
209 if (timevalisset(&old_it.it_value)) {
210 if (old_it.it_value.tv_usec != 0)
211 old_it.it_value.tv_sec++;
212 td->td_retval[0] = old_it.it_value.tv_sec;
218 linux_brk(struct thread *td, struct linux_brk_args *args)
220 struct vmspace *vm = td->td_proc->p_vmspace;
221 vm_offset_t new, old;
222 struct obreak_args /* {
228 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
230 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
231 new = (vm_offset_t)args->dsend;
232 tmp.nsize = (char *)new;
233 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
234 td->td_retval[0] = (long)new;
236 td->td_retval[0] = (long)old;
241 #if defined(__i386__)
242 /* XXX: what about amd64/linux32? */
245 linux_uselib(struct thread *td, struct linux_uselib_args *args)
252 unsigned long file_offset;
253 unsigned long bss_size;
256 int error, locked, writecount;
258 LCONVPATHEXIST(td, args->library, &library);
262 printf(ARGS(uselib, "%s"), library);
269 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
270 UIO_SYSSPACE, library, td);
277 NDFREE(&ni, NDF_ONLY_PNBUF);
280 * From here on down, we have a locked vnode that must be unlocked.
281 * XXX: The code below largely duplicates exec_check_permissions().
286 error = VOP_GET_WRITECOUNT(vp, &writecount);
289 if (writecount != 0) {
295 error = VOP_GETATTR(vp, &attr, td->td_ucred);
299 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
300 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
301 /* EACCESS is what exec(2) returns. */
307 if (attr.va_size == 0) {
312 /* Can we access it? */
313 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
318 * XXX: This should use vn_open() so that it is properly authorized,
319 * and to reduce code redundancy all over the place here.
320 * XXX: Not really, it duplicates far more of exec_check_permissions()
324 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
328 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
332 /* Pull in executable header into exec_map */
333 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
334 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
338 /* Is it a Linux binary ? */
339 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
345 * While we are here, we should REALLY do some more checks
348 /* Set file/virtual offset based on a.out variant. */
349 switch ((int)(a_out->a_magic & 0xffff)) {
350 case 0413: /* ZMAGIC */
353 case 0314: /* QMAGIC */
361 bss_size = round_page(a_out->a_bss);
363 /* Check various fields in header for validity/bounds. */
364 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
369 /* text + data can't exceed file size */
370 if (a_out->a_data + a_out->a_text > attr.va_size) {
376 * text/data/bss must not exceed limits
377 * XXX - this is not complete. it should check current usage PLUS
378 * the resources needed by this library.
380 PROC_LOCK(td->td_proc);
381 if (a_out->a_text > maxtsiz ||
382 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
383 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
385 PROC_UNLOCK(td->td_proc);
389 PROC_UNLOCK(td->td_proc);
392 * Prevent more writers.
393 * XXX: Note that if any of the VM operations fail below we don't
399 * Lock no longer needed
405 * Check if file_offset page aligned. Currently we cannot handle
406 * misalinged file offsets, and so we read in the entire image
409 if (file_offset & PAGE_MASK) {
411 printf("uselib: Non page aligned binary %lu\n", file_offset);
413 /* Map text+data read/write/execute */
415 /* a_entry is the load address and is page aligned */
416 vmaddr = trunc_page(a_out->a_entry);
418 /* get anon user mapping, read+write+execute */
419 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
420 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
421 VM_PROT_ALL, VM_PROT_ALL, 0);
425 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
426 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
427 td->td_ucred, NOCRED, &aresid, td);
436 printf("uselib: Page aligned binary %lu\n", file_offset);
439 * for QMAGIC, a_entry is 20 bytes beyond the load address
440 * to skip the executable header
442 vmaddr = trunc_page(a_out->a_entry);
445 * Map it all into the process's space as a single
446 * copy-on-write "data" segment.
448 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
449 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
450 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
455 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
456 ((long *)vmaddr)[1]);
459 /* Calculate BSS start address */
460 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
463 /* allocate some 'anon' space */
464 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
465 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
472 /* Unlock vnode if needed */
476 /* Release the temporary mapping. */
478 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
483 #endif /* __i386__ */
486 linux_select(struct thread *td, struct linux_select_args *args)
489 struct timeval tv0, tv1, utv, *tvp;
494 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
495 (void *)args->readfds, (void *)args->writefds,
496 (void *)args->exceptfds, (void *)args->timeout);
500 * Store current time for computation of the amount of
504 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
506 utv.tv_sec = ltv.tv_sec;
507 utv.tv_usec = ltv.tv_usec;
510 printf(LMSG("incoming timeout (%jd/%ld)"),
511 (intmax_t)utv.tv_sec, utv.tv_usec);
514 if (itimerfix(&utv)) {
516 * The timeval was invalid. Convert it to something
517 * valid that will act as it does under Linux.
519 utv.tv_sec += utv.tv_usec / 1000000;
520 utv.tv_usec %= 1000000;
521 if (utv.tv_usec < 0) {
523 utv.tv_usec += 1000000;
533 error = kern_select(td, args->nfds, args->readfds, args->writefds,
534 args->exceptfds, tvp, sizeof(l_int) * 8);
538 printf(LMSG("real select returns %d"), error);
544 if (td->td_retval[0]) {
546 * Compute how much time was left of the timeout,
547 * by subtracting the current time and the time
548 * before we started the call, and subtracting
549 * that result from the user-supplied value.
552 timevalsub(&tv1, &tv0);
553 timevalsub(&utv, &tv1);
560 printf(LMSG("outgoing timeout (%jd/%ld)"),
561 (intmax_t)utv.tv_sec, utv.tv_usec);
563 ltv.tv_sec = utv.tv_sec;
564 ltv.tv_usec = utv.tv_usec;
565 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
572 printf(LMSG("select_out -> %d"), error);
578 linux_mremap(struct thread *td, struct linux_mremap_args *args)
580 struct munmap_args /* {
588 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
589 (void *)(uintptr_t)args->addr,
590 (unsigned long)args->old_len,
591 (unsigned long)args->new_len,
592 (unsigned long)args->flags);
595 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
596 td->td_retval[0] = 0;
601 * Check for the page alignment.
602 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
604 if (args->addr & PAGE_MASK) {
605 td->td_retval[0] = 0;
609 args->new_len = round_page(args->new_len);
610 args->old_len = round_page(args->old_len);
612 if (args->new_len > args->old_len) {
613 td->td_retval[0] = 0;
617 if (args->new_len < args->old_len) {
619 (caddr_t)((uintptr_t)args->addr + args->new_len);
620 bsd_args.len = args->old_len - args->new_len;
621 error = sys_munmap(td, &bsd_args);
624 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
628 #define LINUX_MS_ASYNC 0x0001
629 #define LINUX_MS_INVALIDATE 0x0002
630 #define LINUX_MS_SYNC 0x0004
633 linux_msync(struct thread *td, struct linux_msync_args *args)
635 struct msync_args bsd_args;
637 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
638 bsd_args.len = (uintptr_t)args->len;
639 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
641 return (sys_msync(td, &bsd_args));
645 linux_time(struct thread *td, struct linux_time_args *args)
653 printf(ARGS(time, "*"));
658 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
660 td->td_retval[0] = tm;
664 struct l_times_argv {
667 l_clock_t tms_cutime;
668 l_clock_t tms_cstime;
673 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
674 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
675 * auxiliary vector entry.
679 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
680 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
682 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
683 CONVNTCK(r) : CONVOTCK(r))
686 linux_times(struct thread *td, struct linux_times_args *args)
688 struct timeval tv, utime, stime, cutime, cstime;
689 struct l_times_argv tms;
695 printf(ARGS(times, "*"));
698 if (args->buf != NULL) {
702 calcru(p, &utime, &stime);
704 calccru(p, &cutime, &cstime);
707 tms.tms_utime = CONVTCK(utime);
708 tms.tms_stime = CONVTCK(stime);
710 tms.tms_cutime = CONVTCK(cutime);
711 tms.tms_cstime = CONVTCK(cstime);
713 if ((error = copyout(&tms, args->buf, sizeof(tms))))
718 td->td_retval[0] = (int)CONVTCK(tv);
723 linux_newuname(struct thread *td, struct linux_newuname_args *args)
725 struct l_new_utsname utsname;
726 char osname[LINUX_MAX_UTSNAME];
727 char osrelease[LINUX_MAX_UTSNAME];
731 if (ldebug(newuname))
732 printf(ARGS(newuname, "*"));
735 linux_get_osname(td, osname);
736 linux_get_osrelease(td, osrelease);
738 bzero(&utsname, sizeof(utsname));
739 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
740 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
741 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
742 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
743 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
744 for (p = utsname.version; *p != '\0'; ++p)
749 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
751 return (copyout(&utsname, args->buf, sizeof(utsname)));
760 linux_utime(struct thread *td, struct linux_utime_args *args)
762 struct timeval tv[2], *tvp;
763 struct l_utimbuf lut;
767 LCONVPATHEXIST(td, args->fname, &fname);
771 printf(ARGS(utime, "%s, *"), fname);
775 if ((error = copyin(args->times, &lut, sizeof lut))) {
779 tv[0].tv_sec = lut.l_actime;
781 tv[1].tv_sec = lut.l_modtime;
787 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
794 linux_utimes(struct thread *td, struct linux_utimes_args *args)
797 struct timeval tv[2], *tvp = NULL;
801 LCONVPATHEXIST(td, args->fname, &fname);
805 printf(ARGS(utimes, "%s, *"), fname);
808 if (args->tptr != NULL) {
809 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
813 tv[0].tv_sec = ltv[0].tv_sec;
814 tv[0].tv_usec = ltv[0].tv_usec;
815 tv[1].tv_sec = ltv[1].tv_sec;
816 tv[1].tv_usec = ltv[1].tv_usec;
820 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
827 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
830 struct timeval tv[2], *tvp = NULL;
834 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
835 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
838 if (ldebug(futimesat))
839 printf(ARGS(futimesat, "%s, *"), fname);
842 if (args->utimes != NULL) {
843 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
847 tv[0].tv_sec = ltv[0].tv_sec;
848 tv[0].tv_usec = ltv[0].tv_usec;
849 tv[1].tv_sec = ltv[1].tv_sec;
850 tv[1].tv_usec = ltv[1].tv_usec;
854 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
860 linux_common_wait(struct thread *td, int pid, int *status,
861 int options, struct rusage *ru)
865 error = kern_wait(td, pid, &tmpstat, options, ru);
871 if (WIFSIGNALED(tmpstat))
872 tmpstat = (tmpstat & 0xffffff80) |
873 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
874 else if (WIFSTOPPED(tmpstat))
875 tmpstat = (tmpstat & 0xffff00ff) |
876 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
877 else if (WIFCONTINUED(tmpstat))
879 error = copyout(&tmpstat, status, sizeof(int));
885 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
887 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
893 printf(ARGS(waitpid, "%d, %p, %d"),
894 args->pid, (void *)args->status, args->options);
897 * this is necessary because the test in kern_wait doesn't work
898 * because we mess with the options here
900 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
903 options = (args->options & (WNOHANG | WUNTRACED));
904 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
905 if (args->options & __WCLONE)
906 options |= WLINUXCLONE;
908 return (linux_common_wait(td, args->pid, args->status, options, NULL));
910 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
913 linux_wait4(struct thread *td, struct linux_wait4_args *args)
916 struct rusage ru, *rup;
920 printf(ARGS(wait4, "%d, %p, %d, %p"),
921 args->pid, (void *)args->status, args->options,
922 (void *)args->rusage);
925 options = (args->options & (WNOHANG | WUNTRACED));
926 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
927 if (args->options & __WCLONE)
928 options |= WLINUXCLONE;
930 if (args->rusage != NULL)
934 error = linux_common_wait(td, args->pid, args->status, options, rup);
937 if (args->rusage != NULL)
938 error = linux_copyout_rusage(&ru, args->rusage);
943 linux_waitid(struct thread *td, struct linux_waitid_args *args)
945 int status, options, sig;
946 struct __wrusage wru;
954 linux_to_bsd_waitopts(args->options, &options);
956 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
958 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
961 switch (args->idtype) {
979 error = kern_wait6(td, idtype, args->id, &status, options,
983 if (args->rusage != NULL) {
984 error = linux_copyout_rusage(&wru.wru_children,
989 if (args->info != NULL) {
991 if (td->td_retval[0] == 0)
992 bzero(&lsi, sizeof(lsi));
994 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo);
995 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
997 error = copyout(&lsi, args->info, sizeof(lsi));
999 td->td_retval[0] = 0;
1005 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1010 LCONVPATHCREAT(td, args->path, &path);
1014 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1015 (uintmax_t)args->dev);
1018 switch (args->mode & S_IFMT) {
1021 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1027 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1028 args->mode, args->dev);
1036 args->mode |= S_IFREG;
1039 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1040 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1042 kern_close(td, td->td_retval[0]);
1054 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1059 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1060 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1063 if (ldebug(mknodat))
1064 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1067 switch (args->mode & S_IFMT) {
1070 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1075 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1084 args->mode |= S_IFREG;
1087 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1088 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1090 kern_close(td, td->td_retval[0]);
1102 * UGH! This is just about the dumbest idea I've ever heard!!
1105 linux_personality(struct thread *td, struct linux_personality_args *args)
1108 if (ldebug(personality))
1109 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1114 /* Yes Jim, it's still a Linux... */
1115 td->td_retval[0] = 0;
1119 struct l_itimerval {
1120 l_timeval it_interval;
1124 #define B2L_ITIMERVAL(bip, lip) \
1125 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1126 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1127 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1128 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1131 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1134 struct l_itimerval ls;
1135 struct itimerval aitv, oitv;
1138 if (ldebug(setitimer))
1139 printf(ARGS(setitimer, "%p, %p"),
1140 (void *)uap->itv, (void *)uap->oitv);
1143 if (uap->itv == NULL) {
1144 uap->itv = uap->oitv;
1145 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1148 error = copyin(uap->itv, &ls, sizeof(ls));
1151 B2L_ITIMERVAL(&aitv, &ls);
1153 if (ldebug(setitimer)) {
1154 printf("setitimer: value: sec: %jd, usec: %ld\n",
1155 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1156 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1157 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1160 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1161 if (error != 0 || uap->oitv == NULL)
1163 B2L_ITIMERVAL(&ls, &oitv);
1165 return (copyout(&ls, uap->oitv, sizeof(ls)));
1169 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1172 struct l_itimerval ls;
1173 struct itimerval aitv;
1176 if (ldebug(getitimer))
1177 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1179 error = kern_getitimer(td, uap->which, &aitv);
1182 B2L_ITIMERVAL(&ls, &aitv);
1183 return (copyout(&ls, uap->itv, sizeof(ls)));
1186 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1188 linux_nice(struct thread *td, struct linux_nice_args *args)
1190 struct setpriority_args bsd_args;
1192 bsd_args.which = PRIO_PROCESS;
1193 bsd_args.who = 0; /* current process */
1194 bsd_args.prio = args->inc;
1195 return (sys_setpriority(td, &bsd_args));
1197 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1200 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1202 struct ucred *newcred, *oldcred;
1203 l_gid_t *linux_gidset;
1208 ngrp = args->gidsetsize;
1209 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1211 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1212 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1218 oldcred = crcopysafe(p, newcred);
1221 * cr_groups[0] holds egid. Setting the whole set from
1222 * the supplied set will cause egid to be changed too.
1223 * Keep cr_groups[0] unchanged to prevent that.
1226 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1233 newcred->cr_ngroups = ngrp + 1;
1235 bsd_gidset = newcred->cr_groups;
1238 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1242 newcred->cr_ngroups = 1;
1245 proc_set_cred(p, newcred);
1250 free(linux_gidset, M_LINUX);
1255 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1258 l_gid_t *linux_gidset;
1260 int bsd_gidsetsz, ngrp, error;
1262 cred = td->td_ucred;
1263 bsd_gidset = cred->cr_groups;
1264 bsd_gidsetsz = cred->cr_ngroups - 1;
1267 * cr_groups[0] holds egid. Returning the whole set
1268 * here will cause a duplicate. Exclude cr_groups[0]
1272 if ((ngrp = args->gidsetsize) == 0) {
1273 td->td_retval[0] = bsd_gidsetsz;
1277 if (ngrp < bsd_gidsetsz)
1281 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1283 while (ngrp < bsd_gidsetsz) {
1284 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1288 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1289 free(linux_gidset, M_LINUX);
1293 td->td_retval[0] = ngrp;
1298 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1300 struct rlimit bsd_rlim;
1301 struct l_rlimit rlim;
1306 if (ldebug(setrlimit))
1307 printf(ARGS(setrlimit, "%d, %p"),
1308 args->resource, (void *)args->rlim);
1311 if (args->resource >= LINUX_RLIM_NLIMITS)
1314 which = linux_to_bsd_resource[args->resource];
1318 error = copyin(args->rlim, &rlim, sizeof(rlim));
1322 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1323 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1324 return (kern_setrlimit(td, which, &bsd_rlim));
1327 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1329 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1331 struct l_rlimit rlim;
1332 struct proc *p = td->td_proc;
1333 struct rlimit bsd_rlim;
1337 if (ldebug(old_getrlimit))
1338 printf(ARGS(old_getrlimit, "%d, %p"),
1339 args->resource, (void *)args->rlim);
1342 if (args->resource >= LINUX_RLIM_NLIMITS)
1345 which = linux_to_bsd_resource[args->resource];
1350 lim_rlimit(p, which, &bsd_rlim);
1353 #ifdef COMPAT_LINUX32
1354 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1355 if (rlim.rlim_cur == UINT_MAX)
1356 rlim.rlim_cur = INT_MAX;
1357 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1358 if (rlim.rlim_max == UINT_MAX)
1359 rlim.rlim_max = INT_MAX;
1361 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1362 if (rlim.rlim_cur == ULONG_MAX)
1363 rlim.rlim_cur = LONG_MAX;
1364 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1365 if (rlim.rlim_max == ULONG_MAX)
1366 rlim.rlim_max = LONG_MAX;
1368 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1370 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1373 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1375 struct l_rlimit rlim;
1376 struct proc *p = td->td_proc;
1377 struct rlimit bsd_rlim;
1381 if (ldebug(getrlimit))
1382 printf(ARGS(getrlimit, "%d, %p"),
1383 args->resource, (void *)args->rlim);
1386 if (args->resource >= LINUX_RLIM_NLIMITS)
1389 which = linux_to_bsd_resource[args->resource];
1394 lim_rlimit(p, which, &bsd_rlim);
1397 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1398 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1399 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1403 linux_sched_setscheduler(struct thread *td,
1404 struct linux_sched_setscheduler_args *args)
1406 struct sched_param sched_param;
1411 if (ldebug(sched_setscheduler))
1412 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1413 args->pid, args->policy, (const void *)args->param);
1416 switch (args->policy) {
1417 case LINUX_SCHED_OTHER:
1418 policy = SCHED_OTHER;
1420 case LINUX_SCHED_FIFO:
1421 policy = SCHED_FIFO;
1423 case LINUX_SCHED_RR:
1430 error = copyin(args->param, &sched_param, sizeof(sched_param));
1434 tdt = linux_tdfind(td, args->pid, -1);
1438 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1439 PROC_UNLOCK(tdt->td_proc);
1444 linux_sched_getscheduler(struct thread *td,
1445 struct linux_sched_getscheduler_args *args)
1451 if (ldebug(sched_getscheduler))
1452 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1455 tdt = linux_tdfind(td, args->pid, -1);
1459 error = kern_sched_getscheduler(td, tdt, &policy);
1460 PROC_UNLOCK(tdt->td_proc);
1464 td->td_retval[0] = LINUX_SCHED_OTHER;
1467 td->td_retval[0] = LINUX_SCHED_FIFO;
1470 td->td_retval[0] = LINUX_SCHED_RR;
1477 linux_sched_get_priority_max(struct thread *td,
1478 struct linux_sched_get_priority_max_args *args)
1480 struct sched_get_priority_max_args bsd;
1483 if (ldebug(sched_get_priority_max))
1484 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1487 switch (args->policy) {
1488 case LINUX_SCHED_OTHER:
1489 bsd.policy = SCHED_OTHER;
1491 case LINUX_SCHED_FIFO:
1492 bsd.policy = SCHED_FIFO;
1494 case LINUX_SCHED_RR:
1495 bsd.policy = SCHED_RR;
1500 return (sys_sched_get_priority_max(td, &bsd));
1504 linux_sched_get_priority_min(struct thread *td,
1505 struct linux_sched_get_priority_min_args *args)
1507 struct sched_get_priority_min_args bsd;
1510 if (ldebug(sched_get_priority_min))
1511 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1514 switch (args->policy) {
1515 case LINUX_SCHED_OTHER:
1516 bsd.policy = SCHED_OTHER;
1518 case LINUX_SCHED_FIFO:
1519 bsd.policy = SCHED_FIFO;
1521 case LINUX_SCHED_RR:
1522 bsd.policy = SCHED_RR;
1527 return (sys_sched_get_priority_min(td, &bsd));
1530 #define REBOOT_CAD_ON 0x89abcdef
1531 #define REBOOT_CAD_OFF 0
1532 #define REBOOT_HALT 0xcdef0123
1533 #define REBOOT_RESTART 0x01234567
1534 #define REBOOT_RESTART2 0xA1B2C3D4
1535 #define REBOOT_POWEROFF 0x4321FEDC
1536 #define REBOOT_MAGIC1 0xfee1dead
1537 #define REBOOT_MAGIC2 0x28121969
1538 #define REBOOT_MAGIC2A 0x05121996
1539 #define REBOOT_MAGIC2B 0x16041998
1542 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1544 struct reboot_args bsd_args;
1548 printf(ARGS(reboot, "0x%x"), args->cmd);
1551 if (args->magic1 != REBOOT_MAGIC1)
1554 switch (args->magic2) {
1556 case REBOOT_MAGIC2A:
1557 case REBOOT_MAGIC2B:
1563 switch (args->cmd) {
1565 case REBOOT_CAD_OFF:
1566 return (priv_check(td, PRIV_REBOOT));
1568 bsd_args.opt = RB_HALT;
1570 case REBOOT_RESTART:
1571 case REBOOT_RESTART2:
1574 case REBOOT_POWEROFF:
1575 bsd_args.opt = RB_POWEROFF;
1580 return (sys_reboot(td, &bsd_args));
1585 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1586 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1587 * are assumed to be preserved. The following lightweight syscalls fixes
1588 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1590 * linux_getpid() - MP SAFE
1591 * linux_getgid() - MP SAFE
1592 * linux_getuid() - MP SAFE
1596 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1601 printf(ARGS(getpid, ""));
1603 td->td_retval[0] = td->td_proc->p_pid;
1609 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1611 struct linux_emuldata *em;
1615 printf(ARGS(gettid, ""));
1619 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1621 td->td_retval[0] = em->em_tid;
1628 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1632 if (ldebug(getppid))
1633 printf(ARGS(getppid, ""));
1636 PROC_LOCK(td->td_proc);
1637 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1638 PROC_UNLOCK(td->td_proc);
1643 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1648 printf(ARGS(getgid, ""));
1651 td->td_retval[0] = td->td_ucred->cr_rgid;
1656 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1661 printf(ARGS(getuid, ""));
1664 td->td_retval[0] = td->td_ucred->cr_ruid;
1670 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1672 struct getsid_args bsd;
1676 printf(ARGS(getsid, "%i"), args->pid);
1679 bsd.pid = args->pid;
1680 return (sys_getsid(td, &bsd));
1684 linux_nosys(struct thread *td, struct nosys_args *ignore)
1691 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1693 struct getpriority_args bsd_args;
1697 if (ldebug(getpriority))
1698 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1701 bsd_args.which = args->which;
1702 bsd_args.who = args->who;
1703 error = sys_getpriority(td, &bsd_args);
1704 td->td_retval[0] = 20 - td->td_retval[0];
1709 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1714 if (ldebug(sethostname))
1715 printf(ARGS(sethostname, "*, %i"), args->len);
1719 name[1] = KERN_HOSTNAME;
1720 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1725 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1730 if (ldebug(setdomainname))
1731 printf(ARGS(setdomainname, "*, %i"), args->len);
1735 name[1] = KERN_NISDOMAINNAME;
1736 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1741 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1745 if (ldebug(exit_group))
1746 printf(ARGS(exit_group, "%i"), args->error_code);
1749 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1753 * XXX: we should send a signal to the parent if
1754 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1755 * as it doesnt occur often.
1757 exit1(td, W_EXITCODE(args->error_code, 0));
1761 #define _LINUX_CAPABILITY_VERSION 0x19980330
1763 struct l_user_cap_header {
1768 struct l_user_cap_data {
1775 linux_capget(struct thread *td, struct linux_capget_args *args)
1777 struct l_user_cap_header luch;
1778 struct l_user_cap_data lucd;
1781 if (args->hdrp == NULL)
1784 error = copyin(args->hdrp, &luch, sizeof(luch));
1788 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1789 luch.version = _LINUX_CAPABILITY_VERSION;
1790 error = copyout(&luch, args->hdrp, sizeof(luch));
1801 * The current implementation doesn't support setting
1802 * a capability (it's essentially a stub) so indicate
1803 * that no capabilities are currently set or available
1806 bzero (&lucd, sizeof(lucd));
1807 error = copyout(&lucd, args->datap, sizeof(lucd));
1814 linux_capset(struct thread *td, struct linux_capset_args *args)
1816 struct l_user_cap_header luch;
1817 struct l_user_cap_data lucd;
1820 if (args->hdrp == NULL || args->datap == NULL)
1823 error = copyin(args->hdrp, &luch, sizeof(luch));
1827 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1828 luch.version = _LINUX_CAPABILITY_VERSION;
1829 error = copyout(&luch, args->hdrp, sizeof(luch));
1838 error = copyin(args->datap, &lucd, sizeof(lucd));
1842 /* We currently don't support setting any capabilities. */
1843 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1845 "capset effective=0x%x, permitted=0x%x, "
1846 "inheritable=0x%x is not implemented",
1847 (int)lucd.effective, (int)lucd.permitted,
1848 (int)lucd.inheritable);
1856 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1858 int error = 0, max_size;
1859 struct proc *p = td->td_proc;
1860 char comm[LINUX_MAX_COMM_LEN];
1861 struct linux_emuldata *em;
1866 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1867 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1868 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1871 switch (args->option) {
1872 case LINUX_PR_SET_PDEATHSIG:
1873 if (!LINUX_SIG_VALID(args->arg2))
1876 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1877 em->pdeath_signal = args->arg2;
1879 case LINUX_PR_GET_PDEATHSIG:
1881 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1882 pdeath_signal = em->pdeath_signal;
1883 error = copyout(&pdeath_signal,
1884 (void *)(register_t)args->arg2,
1885 sizeof(pdeath_signal));
1887 case LINUX_PR_GET_KEEPCAPS:
1889 * Indicate that we always clear the effective and
1890 * permitted capability sets when the user id becomes
1891 * non-zero (actually the capability sets are simply
1892 * always zero in the current implementation).
1894 td->td_retval[0] = 0;
1896 case LINUX_PR_SET_KEEPCAPS:
1898 * Ignore requests to keep the effective and permitted
1899 * capability sets when the user id becomes non-zero.
1902 case LINUX_PR_SET_NAME:
1904 * To be on the safe side we need to make sure to not
1905 * overflow the size a linux program expects. We already
1906 * do this here in the copyin, so that we don't need to
1909 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1910 error = copyinstr((void *)(register_t)args->arg2, comm,
1913 /* Linux silently truncates the name if it is too long. */
1914 if (error == ENAMETOOLONG) {
1916 * XXX: copyinstr() isn't documented to populate the
1917 * array completely, so do a copyin() to be on the
1918 * safe side. This should be changed in case
1919 * copyinstr() is changed to guarantee this.
1921 error = copyin((void *)(register_t)args->arg2, comm,
1923 comm[max_size - 1] = '\0';
1929 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1932 case LINUX_PR_GET_NAME:
1934 strlcpy(comm, p->p_comm, sizeof(comm));
1936 error = copyout(comm, (void *)(register_t)args->arg2,
1948 linux_sched_setparam(struct thread *td,
1949 struct linux_sched_setparam_args *uap)
1951 struct sched_param sched_param;
1956 if (ldebug(sched_setparam))
1957 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
1960 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1964 tdt = linux_tdfind(td, uap->pid, -1);
1968 error = kern_sched_setparam(td, tdt, &sched_param);
1969 PROC_UNLOCK(tdt->td_proc);
1974 linux_sched_getparam(struct thread *td,
1975 struct linux_sched_getparam_args *uap)
1977 struct sched_param sched_param;
1982 if (ldebug(sched_getparam))
1983 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
1986 tdt = linux_tdfind(td, uap->pid, -1);
1990 error = kern_sched_getparam(td, tdt, &sched_param);
1991 PROC_UNLOCK(tdt->td_proc);
1993 error = copyout(&sched_param, uap->param,
1994 sizeof(sched_param));
1999 * Get affinity of a process.
2002 linux_sched_getaffinity(struct thread *td,
2003 struct linux_sched_getaffinity_args *args)
2007 struct cpuset_getaffinity_args cga;
2010 if (ldebug(sched_getaffinity))
2011 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2014 if (args->len < sizeof(cpuset_t))
2017 tdt = linux_tdfind(td, args->pid, -1);
2021 PROC_UNLOCK(tdt->td_proc);
2022 cga.level = CPU_LEVEL_WHICH;
2023 cga.which = CPU_WHICH_TID;
2024 cga.id = tdt->td_tid;
2025 cga.cpusetsize = sizeof(cpuset_t);
2026 cga.mask = (cpuset_t *) args->user_mask_ptr;
2028 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2029 td->td_retval[0] = sizeof(cpuset_t);
2035 * Set affinity of a process.
2038 linux_sched_setaffinity(struct thread *td,
2039 struct linux_sched_setaffinity_args *args)
2041 struct cpuset_setaffinity_args csa;
2045 if (ldebug(sched_setaffinity))
2046 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2049 if (args->len < sizeof(cpuset_t))
2052 tdt = linux_tdfind(td, args->pid, -1);
2056 PROC_UNLOCK(tdt->td_proc);
2057 csa.level = CPU_LEVEL_WHICH;
2058 csa.which = CPU_WHICH_TID;
2059 csa.id = tdt->td_tid;
2060 csa.cpusetsize = sizeof(cpuset_t);
2061 csa.mask = (cpuset_t *) args->user_mask_ptr;
2063 return (sys_cpuset_setaffinity(td, &csa));
2066 struct linux_rlimit64 {
2072 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2074 struct rlimit rlim, nrlim;
2075 struct linux_rlimit64 lrlim;
2082 if (ldebug(prlimit64))
2083 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2084 args->resource, (void *)args->new, (void *)args->old);
2087 if (args->resource >= LINUX_RLIM_NLIMITS)
2090 which = linux_to_bsd_resource[args->resource];
2094 if (args->new != NULL) {
2096 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2097 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2098 * as INFINITY so we do not need a conversion even.
2100 error = copyin(args->new, &nrlim, sizeof(nrlim));
2105 flags = PGET_HOLD | PGET_NOTWEXIT;
2106 if (args->new != NULL)
2107 flags |= PGET_CANDEBUG;
2109 flags |= PGET_CANSEE;
2110 error = pget(args->pid, flags, &p);
2114 if (args->old != NULL) {
2116 lim_rlimit(p, which, &rlim);
2118 if (rlim.rlim_cur == RLIM_INFINITY)
2119 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2121 lrlim.rlim_cur = rlim.rlim_cur;
2122 if (rlim.rlim_max == RLIM_INFINITY)
2123 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2125 lrlim.rlim_max = rlim.rlim_max;
2126 error = copyout(&lrlim, args->old, sizeof(lrlim));
2131 if (args->new != NULL)
2132 error = kern_proc_setrlimit(td, p, which, &nrlim);
2140 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2142 struct timeval utv, tv0, tv1, *tvp;
2143 struct l_pselect6arg lpse6;
2144 struct l_timespec lts;
2145 struct timespec uts;
2152 if (args->sig != NULL) {
2153 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2156 if (lpse6.ss_len != sizeof(l_ss))
2158 if (lpse6.ss != 0) {
2159 error = copyin(PTRIN(lpse6.ss), &l_ss,
2163 linux_to_bsd_sigset(&l_ss, &ss);
2169 * Currently glibc changes nanosecond number to microsecond.
2170 * This mean losing precision but for now it is hardly seen.
2172 if (args->tsp != NULL) {
2173 error = copyin(args->tsp, <s, sizeof(lts));
2176 uts.tv_sec = lts.tv_sec;
2177 uts.tv_nsec = lts.tv_nsec;
2179 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2180 if (itimerfix(&utv))
2188 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2189 args->exceptfds, tvp, ssp, sizeof(l_int) * 8);
2191 if (error == 0 && args->tsp != NULL) {
2192 if (td->td_retval[0] != 0) {
2194 * Compute how much time was left of the timeout,
2195 * by subtracting the current time and the time
2196 * before we started the call, and subtracting
2197 * that result from the user-supplied value.
2201 timevalsub(&tv1, &tv0);
2202 timevalsub(&utv, &tv1);
2208 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2209 lts.tv_sec = uts.tv_sec;
2210 lts.tv_nsec = uts.tv_nsec;
2211 error = copyout(<s, args->tsp, sizeof(lts));
2217 #if defined(DEBUG) || defined(KTR)
2218 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2220 u_char linux_debug_map[howmany(LINUX_SYS_MAXSYSCALL, sizeof(u_char))];
2223 linux_debug(int syscall, int toggle, int global)
2227 char c = toggle ? 0 : 0xff;
2229 memset(linux_debug_map, c, sizeof(linux_debug_map));
2232 if (syscall < 0 || syscall >= LINUX_SYS_MAXSYSCALL)
2235 clrbit(linux_debug_map, syscall);
2237 setbit(linux_debug_map, syscall);
2242 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2244 * E.g.: sysctl linux.debug=21.0
2246 * As a special case, syscall "all" will apply to all syscalls globally.
2248 #define LINUX_MAX_DEBUGSTR 16
2250 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2252 char value[LINUX_MAX_DEBUGSTR], *p;
2253 int error, sysc, toggle;
2257 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2258 if (error || req->newptr == NULL)
2260 for (p = value; *p != '\0' && *p != '.'; p++);
2264 sysc = strtol(value, NULL, 0);
2265 toggle = strtol(p, NULL, 0);
2266 if (strcmp(value, "all") == 0)
2268 error = linux_debug(sysc, toggle, global);
2272 #endif /* DEBUG || KTR */
2275 linux_sched_rr_get_interval(struct thread *td,
2276 struct linux_sched_rr_get_interval_args *uap)
2279 struct l_timespec lts;
2284 * According to man in case the invalid pid specified
2285 * EINVAL should be returned.
2290 tdt = linux_tdfind(td, uap->pid, -1);
2294 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2295 PROC_UNLOCK(tdt->td_proc);
2298 lts.tv_sec = ts.tv_sec;
2299 lts.tv_nsec = ts.tv_nsec;
2300 return (copyout(<s, uap->interval, sizeof(lts)));
2304 * In case when the Linux thread is the initial thread in
2305 * the thread group thread id is equal to the process id.
2306 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2309 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2311 struct linux_emuldata *em;
2316 if (tid == 0 || tid == td->td_tid) {
2318 PROC_LOCK(tdt->td_proc);
2319 } else if (tid > PID_MAX)
2320 tdt = tdfind(tid, pid);
2323 * Initial thread where the tid equal to the pid.
2327 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2329 * p is not a Linuxulator process.
2334 FOREACH_THREAD_IN_PROC(p, tdt) {
2336 if (tid == em->em_tid)
2348 linux_to_bsd_waitopts(int options, int *bsdopts)
2351 if (options & LINUX_WNOHANG)
2352 *bsdopts |= WNOHANG;
2353 if (options & LINUX_WUNTRACED)
2354 *bsdopts |= WUNTRACED;
2355 if (options & LINUX_WEXITED)
2356 *bsdopts |= WEXITED;
2357 if (options & LINUX_WCONTINUED)
2358 *bsdopts |= WCONTINUED;
2359 if (options & LINUX_WNOWAIT)
2360 *bsdopts |= WNOWAIT;
2362 if (options & __WCLONE)
2363 *bsdopts |= WLINUXCLONE;