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, vfslocked, writecount;
252 LCONVPATHEXIST(td, args->library, &library);
256 printf(ARGS(uselib, "%s"), library);
264 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
265 UIO_SYSSPACE, library, td);
272 vfslocked = NDHASGIANT(&ni);
273 NDFREE(&ni, NDF_ONLY_PNBUF);
276 * From here on down, we have a locked vnode that must be unlocked.
277 * XXX: The code below largely duplicates exec_check_permissions().
282 error = VOP_GET_WRITECOUNT(vp, &writecount);
285 if (writecount != 0) {
291 error = VOP_GETATTR(vp, &attr, td->td_ucred);
295 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
296 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
297 /* EACCESS is what exec(2) returns. */
303 if (attr.va_size == 0) {
308 /* Can we access it? */
309 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
314 * XXX: This should use vn_open() so that it is properly authorized,
315 * and to reduce code redundancy all over the place here.
316 * XXX: Not really, it duplicates far more of exec_check_permissions()
320 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
324 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
328 /* Pull in executable header into exec_map */
329 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
330 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
334 /* Is it a Linux binary ? */
335 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
341 * While we are here, we should REALLY do some more checks
344 /* Set file/virtual offset based on a.out variant. */
345 switch ((int)(a_out->a_magic & 0xffff)) {
346 case 0413: /* ZMAGIC */
349 case 0314: /* QMAGIC */
357 bss_size = round_page(a_out->a_bss);
359 /* Check various fields in header for validity/bounds. */
360 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
365 /* text + data can't exceed file size */
366 if (a_out->a_data + a_out->a_text > attr.va_size) {
372 * text/data/bss must not exceed limits
373 * XXX - this is not complete. it should check current usage PLUS
374 * the resources needed by this library.
376 PROC_LOCK(td->td_proc);
377 if (a_out->a_text > maxtsiz ||
378 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
379 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
381 PROC_UNLOCK(td->td_proc);
385 PROC_UNLOCK(td->td_proc);
388 * Prevent more writers.
389 * XXX: Note that if any of the VM operations fail below we don't
395 * Lock no longer needed
399 VFS_UNLOCK_GIANT(vfslocked);
402 * Check if file_offset page aligned. Currently we cannot handle
403 * misalinged file offsets, and so we read in the entire image
406 if (file_offset & PAGE_MASK) {
408 printf("uselib: Non page aligned binary %lu\n", file_offset);
410 /* Map text+data read/write/execute */
412 /* a_entry is the load address and is page aligned */
413 vmaddr = trunc_page(a_out->a_entry);
415 /* get anon user mapping, read+write+execute */
416 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
417 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
422 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
423 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
424 td->td_ucred, NOCRED, &aresid, td);
433 printf("uselib: Page aligned binary %lu\n", file_offset);
436 * for QMAGIC, a_entry is 20 bytes beyond the load address
437 * to skip the executable header
439 vmaddr = trunc_page(a_out->a_entry);
442 * Map it all into the process's space as a single
443 * copy-on-write "data" segment.
445 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
446 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
447 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
452 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
453 ((long *)vmaddr)[1]);
456 /* Calculate BSS start address */
457 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
460 /* allocate some 'anon' space */
461 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
462 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
468 /* Unlock vnode if needed */
471 VFS_UNLOCK_GIANT(vfslocked);
474 /* Release the temporary mapping. */
476 kmem_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
481 #endif /* __i386__ */
484 linux_select(struct thread *td, struct linux_select_args *args)
487 struct timeval tv0, tv1, utv, *tvp;
492 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
493 (void *)args->readfds, (void *)args->writefds,
494 (void *)args->exceptfds, (void *)args->timeout);
498 * Store current time for computation of the amount of
502 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
504 utv.tv_sec = ltv.tv_sec;
505 utv.tv_usec = ltv.tv_usec;
508 printf(LMSG("incoming timeout (%jd/%ld)"),
509 (intmax_t)utv.tv_sec, utv.tv_usec);
512 if (itimerfix(&utv)) {
514 * The timeval was invalid. Convert it to something
515 * valid that will act as it does under Linux.
517 utv.tv_sec += utv.tv_usec / 1000000;
518 utv.tv_usec %= 1000000;
519 if (utv.tv_usec < 0) {
521 utv.tv_usec += 1000000;
531 error = kern_select(td, args->nfds, args->readfds, args->writefds,
532 args->exceptfds, tvp, sizeof(l_int) * 8);
536 printf(LMSG("real select returns %d"), error);
542 if (td->td_retval[0]) {
544 * Compute how much time was left of the timeout,
545 * by subtracting the current time and the time
546 * before we started the call, and subtracting
547 * that result from the user-supplied value.
550 timevalsub(&tv1, &tv0);
551 timevalsub(&utv, &tv1);
558 printf(LMSG("outgoing timeout (%jd/%ld)"),
559 (intmax_t)utv.tv_sec, utv.tv_usec);
561 ltv.tv_sec = utv.tv_sec;
562 ltv.tv_usec = utv.tv_usec;
563 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
570 printf(LMSG("select_out -> %d"), error);
576 linux_mremap(struct thread *td, struct linux_mremap_args *args)
578 struct munmap_args /* {
586 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
587 (void *)(uintptr_t)args->addr,
588 (unsigned long)args->old_len,
589 (unsigned long)args->new_len,
590 (unsigned long)args->flags);
593 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
594 td->td_retval[0] = 0;
599 * Check for the page alignment.
600 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
602 if (args->addr & PAGE_MASK) {
603 td->td_retval[0] = 0;
607 args->new_len = round_page(args->new_len);
608 args->old_len = round_page(args->old_len);
610 if (args->new_len > args->old_len) {
611 td->td_retval[0] = 0;
615 if (args->new_len < args->old_len) {
617 (caddr_t)((uintptr_t)args->addr + args->new_len);
618 bsd_args.len = args->old_len - args->new_len;
619 error = sys_munmap(td, &bsd_args);
622 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
626 #define LINUX_MS_ASYNC 0x0001
627 #define LINUX_MS_INVALIDATE 0x0002
628 #define LINUX_MS_SYNC 0x0004
631 linux_msync(struct thread *td, struct linux_msync_args *args)
633 struct msync_args bsd_args;
635 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
636 bsd_args.len = (uintptr_t)args->len;
637 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
639 return (sys_msync(td, &bsd_args));
643 linux_time(struct thread *td, struct linux_time_args *args)
651 printf(ARGS(time, "*"));
656 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
658 td->td_retval[0] = tm;
662 struct l_times_argv {
665 l_clock_t tms_cutime;
666 l_clock_t tms_cstime;
671 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
672 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
673 * auxiliary vector entry.
677 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
678 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
680 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
681 CONVNTCK(r) : CONVOTCK(r))
684 linux_times(struct thread *td, struct linux_times_args *args)
686 struct timeval tv, utime, stime, cutime, cstime;
687 struct l_times_argv tms;
693 printf(ARGS(times, "*"));
696 if (args->buf != NULL) {
700 calcru(p, &utime, &stime);
702 calccru(p, &cutime, &cstime);
705 tms.tms_utime = CONVTCK(utime);
706 tms.tms_stime = CONVTCK(stime);
708 tms.tms_cutime = CONVTCK(cutime);
709 tms.tms_cstime = CONVTCK(cstime);
711 if ((error = copyout(&tms, args->buf, sizeof(tms))))
716 td->td_retval[0] = (int)CONVTCK(tv);
721 linux_newuname(struct thread *td, struct linux_newuname_args *args)
723 struct l_new_utsname utsname;
724 char osname[LINUX_MAX_UTSNAME];
725 char osrelease[LINUX_MAX_UTSNAME];
729 if (ldebug(newuname))
730 printf(ARGS(newuname, "*"));
733 linux_get_osname(td, osname);
734 linux_get_osrelease(td, osrelease);
736 bzero(&utsname, sizeof(utsname));
737 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
738 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
739 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
740 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
741 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
742 for (p = utsname.version; *p != '\0'; ++p)
747 strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
749 return (copyout(&utsname, args->buf, sizeof(utsname)));
752 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
759 linux_utime(struct thread *td, struct linux_utime_args *args)
761 struct timeval tv[2], *tvp;
762 struct l_utimbuf lut;
766 LCONVPATHEXIST(td, args->fname, &fname);
770 printf(ARGS(utime, "%s, *"), fname);
774 if ((error = copyin(args->times, &lut, sizeof lut))) {
778 tv[0].tv_sec = lut.l_actime;
780 tv[1].tv_sec = lut.l_modtime;
786 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
792 linux_utimes(struct thread *td, struct linux_utimes_args *args)
795 struct timeval tv[2], *tvp = NULL;
799 LCONVPATHEXIST(td, args->fname, &fname);
803 printf(ARGS(utimes, "%s, *"), fname);
806 if (args->tptr != NULL) {
807 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
811 tv[0].tv_sec = ltv[0].tv_sec;
812 tv[0].tv_usec = ltv[0].tv_usec;
813 tv[1].tv_sec = ltv[1].tv_sec;
814 tv[1].tv_usec = ltv[1].tv_usec;
818 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
824 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
827 struct timeval tv[2], *tvp = NULL;
831 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
832 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
835 if (ldebug(futimesat))
836 printf(ARGS(futimesat, "%s, *"), fname);
839 if (args->utimes != NULL) {
840 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
844 tv[0].tv_sec = ltv[0].tv_sec;
845 tv[0].tv_usec = ltv[0].tv_usec;
846 tv[1].tv_sec = ltv[1].tv_sec;
847 tv[1].tv_usec = ltv[1].tv_usec;
851 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
855 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
858 linux_common_wait(struct thread *td, int pid, int *status,
859 int options, struct rusage *ru)
863 error = kern_wait(td, pid, &tmpstat, options, ru);
869 if (WIFSIGNALED(tmpstat))
870 tmpstat = (tmpstat & 0xffffff80) |
871 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
872 else if (WIFSTOPPED(tmpstat))
873 tmpstat = (tmpstat & 0xffff00ff) |
874 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
875 error = copyout(&tmpstat, status, sizeof(int));
882 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
888 printf(ARGS(waitpid, "%d, %p, %d"),
889 args->pid, (void *)args->status, args->options);
892 * this is necessary because the test in kern_wait doesn't work
893 * because we mess with the options here
895 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
898 options = (args->options & (WNOHANG | WUNTRACED));
899 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
900 if (args->options & __WCLONE)
901 options |= WLINUXCLONE;
903 return (linux_common_wait(td, args->pid, args->status, options, NULL));
908 linux_mknod(struct thread *td, struct linux_mknod_args *args)
913 LCONVPATHCREAT(td, args->path, &path);
917 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
920 switch (args->mode & S_IFMT) {
923 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
928 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
937 args->mode |= S_IFREG;
940 error = kern_open(td, path, UIO_SYSSPACE,
941 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
943 kern_close(td, td->td_retval[0]);
955 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
960 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
961 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
965 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
968 switch (args->mode & S_IFMT) {
971 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
976 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
985 args->mode |= S_IFREG;
988 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
989 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
991 kern_close(td, td->td_retval[0]);
1003 * UGH! This is just about the dumbest idea I've ever heard!!
1006 linux_personality(struct thread *td, struct linux_personality_args *args)
1009 if (ldebug(personality))
1010 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1015 /* Yes Jim, it's still a Linux... */
1016 td->td_retval[0] = 0;
1020 struct l_itimerval {
1021 l_timeval it_interval;
1025 #define B2L_ITIMERVAL(bip, lip) \
1026 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1027 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1028 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1029 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1032 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1035 struct l_itimerval ls;
1036 struct itimerval aitv, oitv;
1039 if (ldebug(setitimer))
1040 printf(ARGS(setitimer, "%p, %p"),
1041 (void *)uap->itv, (void *)uap->oitv);
1044 if (uap->itv == NULL) {
1045 uap->itv = uap->oitv;
1046 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1049 error = copyin(uap->itv, &ls, sizeof(ls));
1052 B2L_ITIMERVAL(&aitv, &ls);
1054 if (ldebug(setitimer)) {
1055 printf("setitimer: value: sec: %jd, usec: %ld\n",
1056 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1057 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1058 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1061 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1062 if (error != 0 || uap->oitv == NULL)
1064 B2L_ITIMERVAL(&ls, &oitv);
1066 return (copyout(&ls, uap->oitv, sizeof(ls)));
1070 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1073 struct l_itimerval ls;
1074 struct itimerval aitv;
1077 if (ldebug(getitimer))
1078 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1080 error = kern_getitimer(td, uap->which, &aitv);
1083 B2L_ITIMERVAL(&ls, &aitv);
1084 return (copyout(&ls, uap->itv, sizeof(ls)));
1088 linux_nice(struct thread *td, struct linux_nice_args *args)
1090 struct setpriority_args bsd_args;
1092 bsd_args.which = PRIO_PROCESS;
1093 bsd_args.who = 0; /* current process */
1094 bsd_args.prio = args->inc;
1095 return (sys_setpriority(td, &bsd_args));
1099 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1101 struct ucred *newcred, *oldcred;
1102 l_gid_t *linux_gidset;
1107 ngrp = args->gidsetsize;
1108 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1110 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1111 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1115 crextend(newcred, ngrp + 1);
1118 oldcred = p->p_ucred;
1119 crcopy(newcred, oldcred);
1122 * cr_groups[0] holds egid. Setting the whole set from
1123 * the supplied set will cause egid to be changed too.
1124 * Keep cr_groups[0] unchanged to prevent that.
1127 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1134 newcred->cr_ngroups = ngrp + 1;
1136 bsd_gidset = newcred->cr_groups;
1139 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1143 newcred->cr_ngroups = 1;
1146 p->p_ucred = newcred;
1151 free(linux_gidset, M_TEMP);
1156 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1159 l_gid_t *linux_gidset;
1161 int bsd_gidsetsz, ngrp, error;
1163 cred = td->td_ucred;
1164 bsd_gidset = cred->cr_groups;
1165 bsd_gidsetsz = cred->cr_ngroups - 1;
1168 * cr_groups[0] holds egid. Returning the whole set
1169 * here will cause a duplicate. Exclude cr_groups[0]
1173 if ((ngrp = args->gidsetsize) == 0) {
1174 td->td_retval[0] = bsd_gidsetsz;
1178 if (ngrp < bsd_gidsetsz)
1182 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1184 while (ngrp < bsd_gidsetsz) {
1185 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1189 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1190 free(linux_gidset, M_TEMP);
1194 td->td_retval[0] = ngrp;
1199 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1201 struct rlimit bsd_rlim;
1202 struct l_rlimit rlim;
1207 if (ldebug(setrlimit))
1208 printf(ARGS(setrlimit, "%d, %p"),
1209 args->resource, (void *)args->rlim);
1212 if (args->resource >= LINUX_RLIM_NLIMITS)
1215 which = linux_to_bsd_resource[args->resource];
1219 error = copyin(args->rlim, &rlim, sizeof(rlim));
1223 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1224 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1225 return (kern_setrlimit(td, which, &bsd_rlim));
1229 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1231 struct l_rlimit rlim;
1232 struct proc *p = td->td_proc;
1233 struct rlimit bsd_rlim;
1237 if (ldebug(old_getrlimit))
1238 printf(ARGS(old_getrlimit, "%d, %p"),
1239 args->resource, (void *)args->rlim);
1242 if (args->resource >= LINUX_RLIM_NLIMITS)
1245 which = linux_to_bsd_resource[args->resource];
1250 lim_rlimit(p, which, &bsd_rlim);
1253 #ifdef COMPAT_LINUX32
1254 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1255 if (rlim.rlim_cur == UINT_MAX)
1256 rlim.rlim_cur = INT_MAX;
1257 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1258 if (rlim.rlim_max == UINT_MAX)
1259 rlim.rlim_max = INT_MAX;
1261 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1262 if (rlim.rlim_cur == ULONG_MAX)
1263 rlim.rlim_cur = LONG_MAX;
1264 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1265 if (rlim.rlim_max == ULONG_MAX)
1266 rlim.rlim_max = LONG_MAX;
1268 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1272 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1274 struct l_rlimit rlim;
1275 struct proc *p = td->td_proc;
1276 struct rlimit bsd_rlim;
1280 if (ldebug(getrlimit))
1281 printf(ARGS(getrlimit, "%d, %p"),
1282 args->resource, (void *)args->rlim);
1285 if (args->resource >= LINUX_RLIM_NLIMITS)
1288 which = linux_to_bsd_resource[args->resource];
1293 lim_rlimit(p, which, &bsd_rlim);
1296 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1297 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1298 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1302 linux_sched_setscheduler(struct thread *td,
1303 struct linux_sched_setscheduler_args *args)
1305 struct sched_setscheduler_args bsd;
1308 if (ldebug(sched_setscheduler))
1309 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1310 args->pid, args->policy, (const void *)args->param);
1313 switch (args->policy) {
1314 case LINUX_SCHED_OTHER:
1315 bsd.policy = SCHED_OTHER;
1317 case LINUX_SCHED_FIFO:
1318 bsd.policy = SCHED_FIFO;
1320 case LINUX_SCHED_RR:
1321 bsd.policy = SCHED_RR;
1327 bsd.pid = args->pid;
1328 bsd.param = (struct sched_param *)args->param;
1329 return (sys_sched_setscheduler(td, &bsd));
1333 linux_sched_getscheduler(struct thread *td,
1334 struct linux_sched_getscheduler_args *args)
1336 struct sched_getscheduler_args bsd;
1340 if (ldebug(sched_getscheduler))
1341 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1344 bsd.pid = args->pid;
1345 error = sys_sched_getscheduler(td, &bsd);
1347 switch (td->td_retval[0]) {
1349 td->td_retval[0] = LINUX_SCHED_OTHER;
1352 td->td_retval[0] = LINUX_SCHED_FIFO;
1355 td->td_retval[0] = LINUX_SCHED_RR;
1363 linux_sched_get_priority_max(struct thread *td,
1364 struct linux_sched_get_priority_max_args *args)
1366 struct sched_get_priority_max_args bsd;
1369 if (ldebug(sched_get_priority_max))
1370 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1373 switch (args->policy) {
1374 case LINUX_SCHED_OTHER:
1375 bsd.policy = SCHED_OTHER;
1377 case LINUX_SCHED_FIFO:
1378 bsd.policy = SCHED_FIFO;
1380 case LINUX_SCHED_RR:
1381 bsd.policy = SCHED_RR;
1386 return (sys_sched_get_priority_max(td, &bsd));
1390 linux_sched_get_priority_min(struct thread *td,
1391 struct linux_sched_get_priority_min_args *args)
1393 struct sched_get_priority_min_args bsd;
1396 if (ldebug(sched_get_priority_min))
1397 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1400 switch (args->policy) {
1401 case LINUX_SCHED_OTHER:
1402 bsd.policy = SCHED_OTHER;
1404 case LINUX_SCHED_FIFO:
1405 bsd.policy = SCHED_FIFO;
1407 case LINUX_SCHED_RR:
1408 bsd.policy = SCHED_RR;
1413 return (sys_sched_get_priority_min(td, &bsd));
1416 #define REBOOT_CAD_ON 0x89abcdef
1417 #define REBOOT_CAD_OFF 0
1418 #define REBOOT_HALT 0xcdef0123
1419 #define REBOOT_RESTART 0x01234567
1420 #define REBOOT_RESTART2 0xA1B2C3D4
1421 #define REBOOT_POWEROFF 0x4321FEDC
1422 #define REBOOT_MAGIC1 0xfee1dead
1423 #define REBOOT_MAGIC2 0x28121969
1424 #define REBOOT_MAGIC2A 0x05121996
1425 #define REBOOT_MAGIC2B 0x16041998
1428 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1430 struct reboot_args bsd_args;
1434 printf(ARGS(reboot, "0x%x"), args->cmd);
1437 if (args->magic1 != REBOOT_MAGIC1)
1440 switch (args->magic2) {
1442 case REBOOT_MAGIC2A:
1443 case REBOOT_MAGIC2B:
1449 switch (args->cmd) {
1451 case REBOOT_CAD_OFF:
1452 return (priv_check(td, PRIV_REBOOT));
1454 bsd_args.opt = RB_HALT;
1456 case REBOOT_RESTART:
1457 case REBOOT_RESTART2:
1460 case REBOOT_POWEROFF:
1461 bsd_args.opt = RB_POWEROFF;
1466 return (sys_reboot(td, &bsd_args));
1471 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1472 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1473 * are assumed to be preserved. The following lightweight syscalls fixes
1474 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1476 * linux_getpid() - MP SAFE
1477 * linux_getgid() - MP SAFE
1478 * linux_getuid() - MP SAFE
1482 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1484 struct linux_emuldata *em;
1488 printf(ARGS(getpid, ""));
1491 if (linux_use26(td)) {
1492 em = em_find(td->td_proc, EMUL_DONTLOCK);
1493 KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1494 td->td_retval[0] = em->shared->group_pid;
1496 td->td_retval[0] = td->td_proc->p_pid;
1503 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1508 printf(ARGS(gettid, ""));
1511 td->td_retval[0] = td->td_proc->p_pid;
1517 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1519 struct linux_emuldata *em;
1520 struct proc *p, *pp;
1523 if (ldebug(getppid))
1524 printf(ARGS(getppid, ""));
1527 if (!linux_use26(td)) {
1528 PROC_LOCK(td->td_proc);
1529 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1530 PROC_UNLOCK(td->td_proc);
1534 em = em_find(td->td_proc, EMUL_DONTLOCK);
1536 KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1538 /* find the group leader */
1539 p = pfind(em->shared->group_pid);
1543 printf(LMSG("parent process not found.\n"));
1548 pp = p->p_pptr; /* switch to parent */
1552 /* if its also linux process */
1553 if (pp->p_sysent == &elf_linux_sysvec) {
1554 em = em_find(pp, EMUL_DONTLOCK);
1555 KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1557 td->td_retval[0] = em->shared->group_pid;
1559 td->td_retval[0] = pp->p_pid;
1567 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1572 printf(ARGS(getgid, ""));
1575 td->td_retval[0] = td->td_ucred->cr_rgid;
1580 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1585 printf(ARGS(getuid, ""));
1588 td->td_retval[0] = td->td_ucred->cr_ruid;
1594 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1596 struct getsid_args bsd;
1600 printf(ARGS(getsid, "%i"), args->pid);
1603 bsd.pid = args->pid;
1604 return (sys_getsid(td, &bsd));
1608 linux_nosys(struct thread *td, struct nosys_args *ignore)
1615 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1617 struct getpriority_args bsd_args;
1621 if (ldebug(getpriority))
1622 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1625 bsd_args.which = args->which;
1626 bsd_args.who = args->who;
1627 error = sys_getpriority(td, &bsd_args);
1628 td->td_retval[0] = 20 - td->td_retval[0];
1633 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1638 if (ldebug(sethostname))
1639 printf(ARGS(sethostname, "*, %i"), args->len);
1643 name[1] = KERN_HOSTNAME;
1644 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1649 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1654 if (ldebug(setdomainname))
1655 printf(ARGS(setdomainname, "*, %i"), args->len);
1659 name[1] = KERN_NISDOMAINNAME;
1660 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1665 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1667 struct linux_emuldata *em;
1670 if (ldebug(exit_group))
1671 printf(ARGS(exit_group, "%i"), args->error_code);
1674 em = em_find(td->td_proc, EMUL_DONTLOCK);
1675 if (em->shared->refs > 1) {
1676 EMUL_SHARED_WLOCK(&emul_shared_lock);
1677 em->shared->flags |= EMUL_SHARED_HASXSTAT;
1678 em->shared->xstat = W_EXITCODE(args->error_code, 0);
1679 EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1680 if (linux_use26(td))
1681 linux_kill_threads(td, SIGKILL);
1685 * XXX: we should send a signal to the parent if
1686 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1687 * as it doesnt occur often.
1689 exit1(td, W_EXITCODE(args->error_code, 0));
1694 #define _LINUX_CAPABILITY_VERSION 0x19980330
1696 struct l_user_cap_header {
1701 struct l_user_cap_data {
1708 linux_capget(struct thread *td, struct linux_capget_args *args)
1710 struct l_user_cap_header luch;
1711 struct l_user_cap_data lucd;
1714 if (args->hdrp == NULL)
1717 error = copyin(args->hdrp, &luch, sizeof(luch));
1721 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1722 luch.version = _LINUX_CAPABILITY_VERSION;
1723 error = copyout(&luch, args->hdrp, sizeof(luch));
1734 * The current implementation doesn't support setting
1735 * a capability (it's essentially a stub) so indicate
1736 * that no capabilities are currently set or available
1739 bzero (&lucd, sizeof(lucd));
1740 error = copyout(&lucd, args->datap, sizeof(lucd));
1747 linux_capset(struct thread *td, struct linux_capset_args *args)
1749 struct l_user_cap_header luch;
1750 struct l_user_cap_data lucd;
1753 if (args->hdrp == NULL || args->datap == NULL)
1756 error = copyin(args->hdrp, &luch, sizeof(luch));
1760 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1761 luch.version = _LINUX_CAPABILITY_VERSION;
1762 error = copyout(&luch, args->hdrp, sizeof(luch));
1771 error = copyin(args->datap, &lucd, sizeof(lucd));
1775 /* We currently don't support setting any capabilities. */
1776 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1778 "capset effective=0x%x, permitted=0x%x, "
1779 "inheritable=0x%x is not implemented",
1780 (int)lucd.effective, (int)lucd.permitted,
1781 (int)lucd.inheritable);
1789 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1791 int error = 0, max_size;
1792 struct proc *p = td->td_proc;
1793 char comm[LINUX_MAX_COMM_LEN];
1794 struct linux_emuldata *em;
1799 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1800 args->arg2, args->arg3, args->arg4, args->arg5);
1803 switch (args->option) {
1804 case LINUX_PR_SET_PDEATHSIG:
1805 if (!LINUX_SIG_VALID(args->arg2))
1807 em = em_find(p, EMUL_DOLOCK);
1808 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1809 em->pdeath_signal = args->arg2;
1810 EMUL_UNLOCK(&emul_lock);
1812 case LINUX_PR_GET_PDEATHSIG:
1813 em = em_find(p, EMUL_DOLOCK);
1814 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1815 pdeath_signal = em->pdeath_signal;
1816 EMUL_UNLOCK(&emul_lock);
1817 error = copyout(&pdeath_signal,
1818 (void *)(register_t)args->arg2,
1819 sizeof(pdeath_signal));
1821 case LINUX_PR_GET_KEEPCAPS:
1823 * Indicate that we always clear the effective and
1824 * permitted capability sets when the user id becomes
1825 * non-zero (actually the capability sets are simply
1826 * always zero in the current implementation).
1828 td->td_retval[0] = 0;
1830 case LINUX_PR_SET_KEEPCAPS:
1832 * Ignore requests to keep the effective and permitted
1833 * capability sets when the user id becomes non-zero.
1836 case LINUX_PR_SET_NAME:
1838 * To be on the safe side we need to make sure to not
1839 * overflow the size a linux program expects. We already
1840 * do this here in the copyin, so that we don't need to
1843 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1844 error = copyinstr((void *)(register_t)args->arg2, comm,
1847 /* Linux silently truncates the name if it is too long. */
1848 if (error == ENAMETOOLONG) {
1850 * XXX: copyinstr() isn't documented to populate the
1851 * array completely, so do a copyin() to be on the
1852 * safe side. This should be changed in case
1853 * copyinstr() is changed to guarantee this.
1855 error = copyin((void *)(register_t)args->arg2, comm,
1857 comm[max_size - 1] = '\0';
1863 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1866 case LINUX_PR_GET_NAME:
1868 strlcpy(comm, p->p_comm, sizeof(comm));
1870 error = copyout(comm, (void *)(register_t)args->arg2,
1882 * Get affinity of a process.
1885 linux_sched_getaffinity(struct thread *td,
1886 struct linux_sched_getaffinity_args *args)
1889 struct cpuset_getaffinity_args cga;
1892 if (ldebug(sched_getaffinity))
1893 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1896 if (args->len < sizeof(cpuset_t))
1899 cga.level = CPU_LEVEL_WHICH;
1900 cga.which = CPU_WHICH_PID;
1902 cga.cpusetsize = sizeof(cpuset_t);
1903 cga.mask = (cpuset_t *) args->user_mask_ptr;
1905 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1906 td->td_retval[0] = sizeof(cpuset_t);
1912 * Set affinity of a process.
1915 linux_sched_setaffinity(struct thread *td,
1916 struct linux_sched_setaffinity_args *args)
1918 struct cpuset_setaffinity_args csa;
1921 if (ldebug(sched_setaffinity))
1922 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1925 if (args->len < sizeof(cpuset_t))
1928 csa.level = CPU_LEVEL_WHICH;
1929 csa.which = CPU_WHICH_PID;
1931 csa.cpusetsize = sizeof(cpuset_t);
1932 csa.mask = (cpuset_t *) args->user_mask_ptr;
1934 return (sys_cpuset_setaffinity(td, &csa));