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"
34 #include "opt_kdtrace.h"
36 #include <sys/param.h>
37 #include <sys/blist.h>
38 #include <sys/fcntl.h>
40 #include <sys/imgact_aout.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
46 #include <sys/malloc.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/namei.h>
53 #include <sys/reboot.h>
54 #include <sys/racct.h>
55 #include <sys/resourcevar.h>
56 #include <sys/sched.h>
58 #include <sys/signalvar.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysproto.h>
63 #include <sys/systm.h>
65 #include <sys/vmmeter.h>
66 #include <sys/vnode.h>
68 #include <sys/cpuset.h>
70 #include <security/mac/mac_framework.h>
74 #include <vm/vm_kern.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
77 #include <vm/vm_object.h>
78 #include <vm/swap_pager.h>
81 #include <machine/../linux32/linux.h>
82 #include <machine/../linux32/linux32_proto.h>
84 #include <machine/../linux/linux.h>
85 #include <machine/../linux/linux_proto.h>
88 #include <compat/linux/linux_dtrace.h>
89 #include <compat/linux/linux_file.h>
90 #include <compat/linux/linux_mib.h>
91 #include <compat/linux/linux_signal.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
98 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
100 /* Linuxulator-global DTrace probes */
101 LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
102 LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
103 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
104 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
105 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
106 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
108 int stclohz; /* Statistics clock frequency */
110 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
111 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
112 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
113 RLIMIT_MEMLOCK, RLIMIT_AS
117 l_long uptime; /* Seconds since boot */
118 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
119 #define LINUX_SYSINFO_LOADS_SCALE 65536
120 l_ulong totalram; /* Total usable main memory size */
121 l_ulong freeram; /* Available memory size */
122 l_ulong sharedram; /* Amount of shared memory */
123 l_ulong bufferram; /* Memory used by buffers */
124 l_ulong totalswap; /* Total swap space size */
125 l_ulong freeswap; /* swap space still available */
126 l_ushort procs; /* Number of current processes */
131 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
134 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
136 struct l_sysinfo sysinfo;
141 bzero(&sysinfo, sizeof(sysinfo));
145 sysinfo.uptime = ts.tv_sec;
147 /* Use the information from the mib to get our load averages */
148 for (i = 0; i < 3; i++)
149 sysinfo.loads[i] = averunnable.ldavg[i] *
150 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
152 sysinfo.totalram = physmem * PAGE_SIZE;
153 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
155 sysinfo.sharedram = 0;
156 mtx_lock(&vm_object_list_mtx);
157 TAILQ_FOREACH(object, &vm_object_list, object_list)
158 if (object->shadow_count > 1)
159 sysinfo.sharedram += object->resident_page_count;
160 mtx_unlock(&vm_object_list_mtx);
162 sysinfo.sharedram *= PAGE_SIZE;
163 sysinfo.bufferram = 0;
165 swap_pager_status(&i, &j);
166 sysinfo.totalswap = i * PAGE_SIZE;
167 sysinfo.freeswap = (i - j) * PAGE_SIZE;
169 sysinfo.procs = nprocs;
171 /* The following are only present in newer Linux kernels. */
172 sysinfo.totalbig = 0;
174 sysinfo.mem_unit = 1;
176 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
180 linux_alarm(struct thread *td, struct linux_alarm_args *args)
182 struct itimerval it, old_it;
188 printf(ARGS(alarm, "%u"), args->secs);
196 it.it_value.tv_sec = (long) secs;
197 it.it_value.tv_usec = 0;
198 it.it_interval.tv_sec = 0;
199 it.it_interval.tv_usec = 0;
200 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
203 if (timevalisset(&old_it.it_value)) {
204 if (old_it.it_value.tv_usec != 0)
205 old_it.it_value.tv_sec++;
206 td->td_retval[0] = old_it.it_value.tv_sec;
212 linux_brk(struct thread *td, struct linux_brk_args *args)
214 struct vmspace *vm = td->td_proc->p_vmspace;
215 vm_offset_t new, old;
216 struct obreak_args /* {
222 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
224 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
225 new = (vm_offset_t)args->dsend;
226 tmp.nsize = (char *)new;
227 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
228 td->td_retval[0] = (long)new;
230 td->td_retval[0] = (long)old;
235 #if defined(__i386__)
236 /* XXX: what about amd64/linux32? */
239 linux_uselib(struct thread *td, struct linux_uselib_args *args)
246 unsigned long file_offset;
247 unsigned long bss_size;
250 int error, locked, writecount;
252 LCONVPATHEXIST(td, args->library, &library);
256 printf(ARGS(uselib, "%s"), library);
263 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
264 UIO_SYSSPACE, library, td);
271 NDFREE(&ni, NDF_ONLY_PNBUF);
274 * From here on down, we have a locked vnode that must be unlocked.
275 * XXX: The code below largely duplicates exec_check_permissions().
280 error = VOP_GET_WRITECOUNT(vp, &writecount);
283 if (writecount != 0) {
289 error = VOP_GETATTR(vp, &attr, td->td_ucred);
293 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
294 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
295 /* EACCESS is what exec(2) returns. */
301 if (attr.va_size == 0) {
306 /* Can we access it? */
307 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
312 * XXX: This should use vn_open() so that it is properly authorized,
313 * and to reduce code redundancy all over the place here.
314 * XXX: Not really, it duplicates far more of exec_check_permissions()
318 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
322 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
326 /* Pull in executable header into exec_map */
327 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
328 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
332 /* Is it a Linux binary ? */
333 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
339 * While we are here, we should REALLY do some more checks
342 /* Set file/virtual offset based on a.out variant. */
343 switch ((int)(a_out->a_magic & 0xffff)) {
344 case 0413: /* ZMAGIC */
347 case 0314: /* QMAGIC */
355 bss_size = round_page(a_out->a_bss);
357 /* Check various fields in header for validity/bounds. */
358 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
363 /* text + data can't exceed file size */
364 if (a_out->a_data + a_out->a_text > attr.va_size) {
370 * text/data/bss must not exceed limits
371 * XXX - this is not complete. it should check current usage PLUS
372 * the resources needed by this library.
374 PROC_LOCK(td->td_proc);
375 if (a_out->a_text > maxtsiz ||
376 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
377 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
379 PROC_UNLOCK(td->td_proc);
383 PROC_UNLOCK(td->td_proc);
386 * Prevent more writers.
387 * XXX: Note that if any of the VM operations fail below we don't
393 * Lock no longer needed
399 * Check if file_offset page aligned. Currently we cannot handle
400 * misalinged file offsets, and so we read in the entire image
403 if (file_offset & PAGE_MASK) {
405 printf("uselib: Non page aligned binary %lu\n", file_offset);
407 /* Map text+data read/write/execute */
409 /* a_entry is the load address and is page aligned */
410 vmaddr = trunc_page(a_out->a_entry);
412 /* get anon user mapping, read+write+execute */
413 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
414 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
415 VM_PROT_ALL, VM_PROT_ALL, 0);
419 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
420 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
421 td->td_ucred, NOCRED, &aresid, td);
430 printf("uselib: Page aligned binary %lu\n", file_offset);
433 * for QMAGIC, a_entry is 20 bytes beyond the load address
434 * to skip the executable header
436 vmaddr = trunc_page(a_out->a_entry);
439 * Map it all into the process's space as a single
440 * copy-on-write "data" segment.
442 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
443 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
444 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
449 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
450 ((long *)vmaddr)[1]);
453 /* Calculate BSS start address */
454 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
457 /* allocate some 'anon' space */
458 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
459 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
466 /* Unlock vnode if needed */
470 /* Release the temporary mapping. */
472 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
477 #endif /* __i386__ */
480 linux_select(struct thread *td, struct linux_select_args *args)
483 struct timeval tv0, tv1, utv, *tvp;
488 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
489 (void *)args->readfds, (void *)args->writefds,
490 (void *)args->exceptfds, (void *)args->timeout);
494 * Store current time for computation of the amount of
498 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
500 utv.tv_sec = ltv.tv_sec;
501 utv.tv_usec = ltv.tv_usec;
504 printf(LMSG("incoming timeout (%jd/%ld)"),
505 (intmax_t)utv.tv_sec, utv.tv_usec);
508 if (itimerfix(&utv)) {
510 * The timeval was invalid. Convert it to something
511 * valid that will act as it does under Linux.
513 utv.tv_sec += utv.tv_usec / 1000000;
514 utv.tv_usec %= 1000000;
515 if (utv.tv_usec < 0) {
517 utv.tv_usec += 1000000;
527 error = kern_select(td, args->nfds, args->readfds, args->writefds,
528 args->exceptfds, tvp, sizeof(l_int) * 8);
532 printf(LMSG("real select returns %d"), error);
538 if (td->td_retval[0]) {
540 * Compute how much time was left of the timeout,
541 * by subtracting the current time and the time
542 * before we started the call, and subtracting
543 * that result from the user-supplied value.
546 timevalsub(&tv1, &tv0);
547 timevalsub(&utv, &tv1);
554 printf(LMSG("outgoing timeout (%jd/%ld)"),
555 (intmax_t)utv.tv_sec, utv.tv_usec);
557 ltv.tv_sec = utv.tv_sec;
558 ltv.tv_usec = utv.tv_usec;
559 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
566 printf(LMSG("select_out -> %d"), error);
572 linux_mremap(struct thread *td, struct linux_mremap_args *args)
574 struct munmap_args /* {
582 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
583 (void *)(uintptr_t)args->addr,
584 (unsigned long)args->old_len,
585 (unsigned long)args->new_len,
586 (unsigned long)args->flags);
589 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
590 td->td_retval[0] = 0;
595 * Check for the page alignment.
596 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
598 if (args->addr & PAGE_MASK) {
599 td->td_retval[0] = 0;
603 args->new_len = round_page(args->new_len);
604 args->old_len = round_page(args->old_len);
606 if (args->new_len > args->old_len) {
607 td->td_retval[0] = 0;
611 if (args->new_len < args->old_len) {
613 (caddr_t)((uintptr_t)args->addr + args->new_len);
614 bsd_args.len = args->old_len - args->new_len;
615 error = sys_munmap(td, &bsd_args);
618 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
622 #define LINUX_MS_ASYNC 0x0001
623 #define LINUX_MS_INVALIDATE 0x0002
624 #define LINUX_MS_SYNC 0x0004
627 linux_msync(struct thread *td, struct linux_msync_args *args)
629 struct msync_args bsd_args;
631 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
632 bsd_args.len = (uintptr_t)args->len;
633 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
635 return (sys_msync(td, &bsd_args));
639 linux_time(struct thread *td, struct linux_time_args *args)
647 printf(ARGS(time, "*"));
652 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
654 td->td_retval[0] = tm;
658 struct l_times_argv {
661 l_clock_t tms_cutime;
662 l_clock_t tms_cstime;
667 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
668 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
669 * auxiliary vector entry.
673 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
674 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
676 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
677 CONVNTCK(r) : CONVOTCK(r))
680 linux_times(struct thread *td, struct linux_times_args *args)
682 struct timeval tv, utime, stime, cutime, cstime;
683 struct l_times_argv tms;
689 printf(ARGS(times, "*"));
692 if (args->buf != NULL) {
696 calcru(p, &utime, &stime);
698 calccru(p, &cutime, &cstime);
701 tms.tms_utime = CONVTCK(utime);
702 tms.tms_stime = CONVTCK(stime);
704 tms.tms_cutime = CONVTCK(cutime);
705 tms.tms_cstime = CONVTCK(cstime);
707 if ((error = copyout(&tms, args->buf, sizeof(tms))))
712 td->td_retval[0] = (int)CONVTCK(tv);
717 linux_newuname(struct thread *td, struct linux_newuname_args *args)
719 struct l_new_utsname utsname;
720 char osname[LINUX_MAX_UTSNAME];
721 char osrelease[LINUX_MAX_UTSNAME];
725 if (ldebug(newuname))
726 printf(ARGS(newuname, "*"));
729 linux_get_osname(td, osname);
730 linux_get_osrelease(td, osrelease);
732 bzero(&utsname, sizeof(utsname));
733 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
734 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
735 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
736 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
737 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
738 for (p = utsname.version; *p != '\0'; ++p)
743 strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
745 return (copyout(&utsname, args->buf, sizeof(utsname)));
748 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
755 linux_utime(struct thread *td, struct linux_utime_args *args)
757 struct timeval tv[2], *tvp;
758 struct l_utimbuf lut;
762 LCONVPATHEXIST(td, args->fname, &fname);
766 printf(ARGS(utime, "%s, *"), fname);
770 if ((error = copyin(args->times, &lut, sizeof lut))) {
774 tv[0].tv_sec = lut.l_actime;
776 tv[1].tv_sec = lut.l_modtime;
782 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
788 linux_utimes(struct thread *td, struct linux_utimes_args *args)
791 struct timeval tv[2], *tvp = NULL;
795 LCONVPATHEXIST(td, args->fname, &fname);
799 printf(ARGS(utimes, "%s, *"), fname);
802 if (args->tptr != NULL) {
803 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
807 tv[0].tv_sec = ltv[0].tv_sec;
808 tv[0].tv_usec = ltv[0].tv_usec;
809 tv[1].tv_sec = ltv[1].tv_sec;
810 tv[1].tv_usec = ltv[1].tv_usec;
814 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
820 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
823 struct timeval tv[2], *tvp = NULL;
827 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
828 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
831 if (ldebug(futimesat))
832 printf(ARGS(futimesat, "%s, *"), fname);
835 if (args->utimes != NULL) {
836 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
840 tv[0].tv_sec = ltv[0].tv_sec;
841 tv[0].tv_usec = ltv[0].tv_usec;
842 tv[1].tv_sec = ltv[1].tv_sec;
843 tv[1].tv_usec = ltv[1].tv_usec;
847 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
851 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
854 linux_common_wait(struct thread *td, int pid, int *status,
855 int options, struct rusage *ru)
859 error = kern_wait(td, pid, &tmpstat, options, ru);
865 if (WIFSIGNALED(tmpstat))
866 tmpstat = (tmpstat & 0xffffff80) |
867 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
868 else if (WIFSTOPPED(tmpstat))
869 tmpstat = (tmpstat & 0xffff00ff) |
870 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
871 error = copyout(&tmpstat, status, sizeof(int));
878 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
884 printf(ARGS(waitpid, "%d, %p, %d"),
885 args->pid, (void *)args->status, args->options);
888 * this is necessary because the test in kern_wait doesn't work
889 * because we mess with the options here
891 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
894 options = (args->options & (WNOHANG | WUNTRACED));
895 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
896 if (args->options & __WCLONE)
897 options |= WLINUXCLONE;
899 return (linux_common_wait(td, args->pid, args->status, options, NULL));
904 linux_mknod(struct thread *td, struct linux_mknod_args *args)
909 LCONVPATHCREAT(td, args->path, &path);
913 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
916 switch (args->mode & S_IFMT) {
919 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
924 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
933 args->mode |= S_IFREG;
936 error = kern_open(td, path, UIO_SYSSPACE,
937 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
939 kern_close(td, td->td_retval[0]);
951 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
956 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
957 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
961 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
964 switch (args->mode & S_IFMT) {
967 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
972 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
981 args->mode |= S_IFREG;
984 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
985 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
987 kern_close(td, td->td_retval[0]);
999 * UGH! This is just about the dumbest idea I've ever heard!!
1002 linux_personality(struct thread *td, struct linux_personality_args *args)
1005 if (ldebug(personality))
1006 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1011 /* Yes Jim, it's still a Linux... */
1012 td->td_retval[0] = 0;
1016 struct l_itimerval {
1017 l_timeval it_interval;
1021 #define B2L_ITIMERVAL(bip, lip) \
1022 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1023 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1024 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1025 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1028 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1031 struct l_itimerval ls;
1032 struct itimerval aitv, oitv;
1035 if (ldebug(setitimer))
1036 printf(ARGS(setitimer, "%p, %p"),
1037 (void *)uap->itv, (void *)uap->oitv);
1040 if (uap->itv == NULL) {
1041 uap->itv = uap->oitv;
1042 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1045 error = copyin(uap->itv, &ls, sizeof(ls));
1048 B2L_ITIMERVAL(&aitv, &ls);
1050 if (ldebug(setitimer)) {
1051 printf("setitimer: value: sec: %jd, usec: %ld\n",
1052 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1053 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1054 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1057 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1058 if (error != 0 || uap->oitv == NULL)
1060 B2L_ITIMERVAL(&ls, &oitv);
1062 return (copyout(&ls, uap->oitv, sizeof(ls)));
1066 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1069 struct l_itimerval ls;
1070 struct itimerval aitv;
1073 if (ldebug(getitimer))
1074 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1076 error = kern_getitimer(td, uap->which, &aitv);
1079 B2L_ITIMERVAL(&ls, &aitv);
1080 return (copyout(&ls, uap->itv, sizeof(ls)));
1084 linux_nice(struct thread *td, struct linux_nice_args *args)
1086 struct setpriority_args bsd_args;
1088 bsd_args.which = PRIO_PROCESS;
1089 bsd_args.who = 0; /* current process */
1090 bsd_args.prio = args->inc;
1091 return (sys_setpriority(td, &bsd_args));
1095 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1097 struct ucred *newcred, *oldcred;
1098 l_gid_t *linux_gidset;
1103 ngrp = args->gidsetsize;
1104 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1106 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1107 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1111 crextend(newcred, ngrp + 1);
1114 oldcred = p->p_ucred;
1115 crcopy(newcred, oldcred);
1118 * cr_groups[0] holds egid. Setting the whole set from
1119 * the supplied set will cause egid to be changed too.
1120 * Keep cr_groups[0] unchanged to prevent that.
1123 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1130 newcred->cr_ngroups = ngrp + 1;
1132 bsd_gidset = newcred->cr_groups;
1135 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1139 newcred->cr_ngroups = 1;
1142 p->p_ucred = newcred;
1147 free(linux_gidset, M_TEMP);
1152 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1155 l_gid_t *linux_gidset;
1157 int bsd_gidsetsz, ngrp, error;
1159 cred = td->td_ucred;
1160 bsd_gidset = cred->cr_groups;
1161 bsd_gidsetsz = cred->cr_ngroups - 1;
1164 * cr_groups[0] holds egid. Returning the whole set
1165 * here will cause a duplicate. Exclude cr_groups[0]
1169 if ((ngrp = args->gidsetsize) == 0) {
1170 td->td_retval[0] = bsd_gidsetsz;
1174 if (ngrp < bsd_gidsetsz)
1178 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1180 while (ngrp < bsd_gidsetsz) {
1181 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1185 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1186 free(linux_gidset, M_TEMP);
1190 td->td_retval[0] = ngrp;
1195 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1197 struct rlimit bsd_rlim;
1198 struct l_rlimit rlim;
1203 if (ldebug(setrlimit))
1204 printf(ARGS(setrlimit, "%d, %p"),
1205 args->resource, (void *)args->rlim);
1208 if (args->resource >= LINUX_RLIM_NLIMITS)
1211 which = linux_to_bsd_resource[args->resource];
1215 error = copyin(args->rlim, &rlim, sizeof(rlim));
1219 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1220 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1221 return (kern_setrlimit(td, which, &bsd_rlim));
1225 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1227 struct l_rlimit rlim;
1228 struct proc *p = td->td_proc;
1229 struct rlimit bsd_rlim;
1233 if (ldebug(old_getrlimit))
1234 printf(ARGS(old_getrlimit, "%d, %p"),
1235 args->resource, (void *)args->rlim);
1238 if (args->resource >= LINUX_RLIM_NLIMITS)
1241 which = linux_to_bsd_resource[args->resource];
1246 lim_rlimit(p, which, &bsd_rlim);
1249 #ifdef COMPAT_LINUX32
1250 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1251 if (rlim.rlim_cur == UINT_MAX)
1252 rlim.rlim_cur = INT_MAX;
1253 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1254 if (rlim.rlim_max == UINT_MAX)
1255 rlim.rlim_max = INT_MAX;
1257 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1258 if (rlim.rlim_cur == ULONG_MAX)
1259 rlim.rlim_cur = LONG_MAX;
1260 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1261 if (rlim.rlim_max == ULONG_MAX)
1262 rlim.rlim_max = LONG_MAX;
1264 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1268 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1270 struct l_rlimit rlim;
1271 struct proc *p = td->td_proc;
1272 struct rlimit bsd_rlim;
1276 if (ldebug(getrlimit))
1277 printf(ARGS(getrlimit, "%d, %p"),
1278 args->resource, (void *)args->rlim);
1281 if (args->resource >= LINUX_RLIM_NLIMITS)
1284 which = linux_to_bsd_resource[args->resource];
1289 lim_rlimit(p, which, &bsd_rlim);
1292 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1293 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1294 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1298 linux_sched_setscheduler(struct thread *td,
1299 struct linux_sched_setscheduler_args *args)
1301 struct sched_setscheduler_args bsd;
1304 if (ldebug(sched_setscheduler))
1305 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1306 args->pid, args->policy, (const void *)args->param);
1309 switch (args->policy) {
1310 case LINUX_SCHED_OTHER:
1311 bsd.policy = SCHED_OTHER;
1313 case LINUX_SCHED_FIFO:
1314 bsd.policy = SCHED_FIFO;
1316 case LINUX_SCHED_RR:
1317 bsd.policy = SCHED_RR;
1323 bsd.pid = args->pid;
1324 bsd.param = (struct sched_param *)args->param;
1325 return (sys_sched_setscheduler(td, &bsd));
1329 linux_sched_getscheduler(struct thread *td,
1330 struct linux_sched_getscheduler_args *args)
1332 struct sched_getscheduler_args bsd;
1336 if (ldebug(sched_getscheduler))
1337 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1340 bsd.pid = args->pid;
1341 error = sys_sched_getscheduler(td, &bsd);
1343 switch (td->td_retval[0]) {
1345 td->td_retval[0] = LINUX_SCHED_OTHER;
1348 td->td_retval[0] = LINUX_SCHED_FIFO;
1351 td->td_retval[0] = LINUX_SCHED_RR;
1359 linux_sched_get_priority_max(struct thread *td,
1360 struct linux_sched_get_priority_max_args *args)
1362 struct sched_get_priority_max_args bsd;
1365 if (ldebug(sched_get_priority_max))
1366 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1369 switch (args->policy) {
1370 case LINUX_SCHED_OTHER:
1371 bsd.policy = SCHED_OTHER;
1373 case LINUX_SCHED_FIFO:
1374 bsd.policy = SCHED_FIFO;
1376 case LINUX_SCHED_RR:
1377 bsd.policy = SCHED_RR;
1382 return (sys_sched_get_priority_max(td, &bsd));
1386 linux_sched_get_priority_min(struct thread *td,
1387 struct linux_sched_get_priority_min_args *args)
1389 struct sched_get_priority_min_args bsd;
1392 if (ldebug(sched_get_priority_min))
1393 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1396 switch (args->policy) {
1397 case LINUX_SCHED_OTHER:
1398 bsd.policy = SCHED_OTHER;
1400 case LINUX_SCHED_FIFO:
1401 bsd.policy = SCHED_FIFO;
1403 case LINUX_SCHED_RR:
1404 bsd.policy = SCHED_RR;
1409 return (sys_sched_get_priority_min(td, &bsd));
1412 #define REBOOT_CAD_ON 0x89abcdef
1413 #define REBOOT_CAD_OFF 0
1414 #define REBOOT_HALT 0xcdef0123
1415 #define REBOOT_RESTART 0x01234567
1416 #define REBOOT_RESTART2 0xA1B2C3D4
1417 #define REBOOT_POWEROFF 0x4321FEDC
1418 #define REBOOT_MAGIC1 0xfee1dead
1419 #define REBOOT_MAGIC2 0x28121969
1420 #define REBOOT_MAGIC2A 0x05121996
1421 #define REBOOT_MAGIC2B 0x16041998
1424 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1426 struct reboot_args bsd_args;
1430 printf(ARGS(reboot, "0x%x"), args->cmd);
1433 if (args->magic1 != REBOOT_MAGIC1)
1436 switch (args->magic2) {
1438 case REBOOT_MAGIC2A:
1439 case REBOOT_MAGIC2B:
1445 switch (args->cmd) {
1447 case REBOOT_CAD_OFF:
1448 return (priv_check(td, PRIV_REBOOT));
1450 bsd_args.opt = RB_HALT;
1452 case REBOOT_RESTART:
1453 case REBOOT_RESTART2:
1456 case REBOOT_POWEROFF:
1457 bsd_args.opt = RB_POWEROFF;
1462 return (sys_reboot(td, &bsd_args));
1467 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1468 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1469 * are assumed to be preserved. The following lightweight syscalls fixes
1470 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1472 * linux_getpid() - MP SAFE
1473 * linux_getgid() - MP SAFE
1474 * linux_getuid() - MP SAFE
1478 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1480 struct linux_emuldata *em;
1484 printf(ARGS(getpid, ""));
1487 if (linux_use26(td)) {
1488 em = em_find(td->td_proc, EMUL_DONTLOCK);
1489 KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1490 td->td_retval[0] = em->shared->group_pid;
1492 td->td_retval[0] = td->td_proc->p_pid;
1499 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1504 printf(ARGS(gettid, ""));
1507 td->td_retval[0] = td->td_proc->p_pid;
1513 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1515 struct linux_emuldata *em;
1516 struct proc *p, *pp;
1519 if (ldebug(getppid))
1520 printf(ARGS(getppid, ""));
1523 if (!linux_use26(td)) {
1524 PROC_LOCK(td->td_proc);
1525 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1526 PROC_UNLOCK(td->td_proc);
1530 em = em_find(td->td_proc, EMUL_DONTLOCK);
1532 KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1534 /* find the group leader */
1535 p = pfind(em->shared->group_pid);
1539 printf(LMSG("parent process not found.\n"));
1544 pp = p->p_pptr; /* switch to parent */
1548 /* if its also linux process */
1549 if (pp->p_sysent == &elf_linux_sysvec) {
1550 em = em_find(pp, EMUL_DONTLOCK);
1551 KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1553 td->td_retval[0] = em->shared->group_pid;
1555 td->td_retval[0] = pp->p_pid;
1563 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1568 printf(ARGS(getgid, ""));
1571 td->td_retval[0] = td->td_ucred->cr_rgid;
1576 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1581 printf(ARGS(getuid, ""));
1584 td->td_retval[0] = td->td_ucred->cr_ruid;
1590 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1592 struct getsid_args bsd;
1596 printf(ARGS(getsid, "%i"), args->pid);
1599 bsd.pid = args->pid;
1600 return (sys_getsid(td, &bsd));
1604 linux_nosys(struct thread *td, struct nosys_args *ignore)
1611 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1613 struct getpriority_args bsd_args;
1617 if (ldebug(getpriority))
1618 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1621 bsd_args.which = args->which;
1622 bsd_args.who = args->who;
1623 error = sys_getpriority(td, &bsd_args);
1624 td->td_retval[0] = 20 - td->td_retval[0];
1629 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1634 if (ldebug(sethostname))
1635 printf(ARGS(sethostname, "*, %i"), args->len);
1639 name[1] = KERN_HOSTNAME;
1640 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1645 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1650 if (ldebug(setdomainname))
1651 printf(ARGS(setdomainname, "*, %i"), args->len);
1655 name[1] = KERN_NISDOMAINNAME;
1656 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1661 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1663 struct linux_emuldata *em;
1666 if (ldebug(exit_group))
1667 printf(ARGS(exit_group, "%i"), args->error_code);
1670 em = em_find(td->td_proc, EMUL_DONTLOCK);
1671 if (em->shared->refs > 1) {
1672 EMUL_SHARED_WLOCK(&emul_shared_lock);
1673 em->shared->flags |= EMUL_SHARED_HASXSTAT;
1674 em->shared->xstat = W_EXITCODE(args->error_code, 0);
1675 EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1676 if (linux_use26(td))
1677 linux_kill_threads(td, SIGKILL);
1681 * XXX: we should send a signal to the parent if
1682 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1683 * as it doesnt occur often.
1685 exit1(td, W_EXITCODE(args->error_code, 0));
1690 #define _LINUX_CAPABILITY_VERSION 0x19980330
1692 struct l_user_cap_header {
1697 struct l_user_cap_data {
1704 linux_capget(struct thread *td, struct linux_capget_args *args)
1706 struct l_user_cap_header luch;
1707 struct l_user_cap_data lucd;
1710 if (args->hdrp == NULL)
1713 error = copyin(args->hdrp, &luch, sizeof(luch));
1717 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1718 luch.version = _LINUX_CAPABILITY_VERSION;
1719 error = copyout(&luch, args->hdrp, sizeof(luch));
1730 * The current implementation doesn't support setting
1731 * a capability (it's essentially a stub) so indicate
1732 * that no capabilities are currently set or available
1735 bzero (&lucd, sizeof(lucd));
1736 error = copyout(&lucd, args->datap, sizeof(lucd));
1743 linux_capset(struct thread *td, struct linux_capset_args *args)
1745 struct l_user_cap_header luch;
1746 struct l_user_cap_data lucd;
1749 if (args->hdrp == NULL || args->datap == NULL)
1752 error = copyin(args->hdrp, &luch, sizeof(luch));
1756 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1757 luch.version = _LINUX_CAPABILITY_VERSION;
1758 error = copyout(&luch, args->hdrp, sizeof(luch));
1767 error = copyin(args->datap, &lucd, sizeof(lucd));
1771 /* We currently don't support setting any capabilities. */
1772 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1774 "capset effective=0x%x, permitted=0x%x, "
1775 "inheritable=0x%x is not implemented",
1776 (int)lucd.effective, (int)lucd.permitted,
1777 (int)lucd.inheritable);
1785 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1787 int error = 0, max_size;
1788 struct proc *p = td->td_proc;
1789 char comm[LINUX_MAX_COMM_LEN];
1790 struct linux_emuldata *em;
1795 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1796 args->arg2, args->arg3, args->arg4, args->arg5);
1799 switch (args->option) {
1800 case LINUX_PR_SET_PDEATHSIG:
1801 if (!LINUX_SIG_VALID(args->arg2))
1803 em = em_find(p, EMUL_DOLOCK);
1804 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1805 em->pdeath_signal = args->arg2;
1806 EMUL_UNLOCK(&emul_lock);
1808 case LINUX_PR_GET_PDEATHSIG:
1809 em = em_find(p, EMUL_DOLOCK);
1810 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1811 pdeath_signal = em->pdeath_signal;
1812 EMUL_UNLOCK(&emul_lock);
1813 error = copyout(&pdeath_signal,
1814 (void *)(register_t)args->arg2,
1815 sizeof(pdeath_signal));
1817 case LINUX_PR_GET_KEEPCAPS:
1819 * Indicate that we always clear the effective and
1820 * permitted capability sets when the user id becomes
1821 * non-zero (actually the capability sets are simply
1822 * always zero in the current implementation).
1824 td->td_retval[0] = 0;
1826 case LINUX_PR_SET_KEEPCAPS:
1828 * Ignore requests to keep the effective and permitted
1829 * capability sets when the user id becomes non-zero.
1832 case LINUX_PR_SET_NAME:
1834 * To be on the safe side we need to make sure to not
1835 * overflow the size a linux program expects. We already
1836 * do this here in the copyin, so that we don't need to
1839 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1840 error = copyinstr((void *)(register_t)args->arg2, comm,
1843 /* Linux silently truncates the name if it is too long. */
1844 if (error == ENAMETOOLONG) {
1846 * XXX: copyinstr() isn't documented to populate the
1847 * array completely, so do a copyin() to be on the
1848 * safe side. This should be changed in case
1849 * copyinstr() is changed to guarantee this.
1851 error = copyin((void *)(register_t)args->arg2, comm,
1853 comm[max_size - 1] = '\0';
1859 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1862 case LINUX_PR_GET_NAME:
1864 strlcpy(comm, p->p_comm, sizeof(comm));
1866 error = copyout(comm, (void *)(register_t)args->arg2,
1878 * Get affinity of a process.
1881 linux_sched_getaffinity(struct thread *td,
1882 struct linux_sched_getaffinity_args *args)
1885 struct cpuset_getaffinity_args cga;
1888 if (ldebug(sched_getaffinity))
1889 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1892 if (args->len < sizeof(cpuset_t))
1895 cga.level = CPU_LEVEL_WHICH;
1896 cga.which = CPU_WHICH_PID;
1898 cga.cpusetsize = sizeof(cpuset_t);
1899 cga.mask = (cpuset_t *) args->user_mask_ptr;
1901 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1902 td->td_retval[0] = sizeof(cpuset_t);
1908 * Set affinity of a process.
1911 linux_sched_setaffinity(struct thread *td,
1912 struct linux_sched_setaffinity_args *args)
1914 struct cpuset_setaffinity_args csa;
1917 if (ldebug(sched_setaffinity))
1918 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1921 if (args->len < sizeof(cpuset_t))
1924 csa.level = CPU_LEVEL_WHICH;
1925 csa.which = CPU_WHICH_PID;
1927 csa.cpusetsize = sizeof(cpuset_t);
1928 csa.mask = (cpuset_t *) args->user_mask_ptr;
1930 return (sys_cpuset_setaffinity(td, &csa));